romenskiy2012/jemalloc
1
0
Fork 0
mirror of https://github.com/jemalloc/jemalloc.git synced 2026-04-14 14:41:42 +03:00
Code Issues Projects Releases Packages Wiki Activity

Compare commits

base: romenskiy2012:5.3.0
Branches Tags
romenskiy2012:dev
romenskiy2012:master
romenskiy2012:stable-4
romenskiy2012:stable-3
romenskiy2012:5.3.1
romenskiy2012:5.3.0
romenskiy2012:5.2.1
romenskiy2012:5.2.0
romenskiy2012:5.1.0
romenskiy2012:5.0.1
romenskiy2012:5.0.0
romenskiy2012:4.5.0
romenskiy2012:4.4.0
romenskiy2012:4.3.1
romenskiy2012:4.3.0
romenskiy2012:4.2.1
romenskiy2012:4.2.0
romenskiy2012:4.1.1
romenskiy2012:4.1.0
romenskiy2012:4.0.4
romenskiy2012:4.0.3
romenskiy2012:4.0.2
romenskiy2012:4.0.1
romenskiy2012:4.0.0
romenskiy2012:3.6.0
romenskiy2012:3.5.1
romenskiy2012:3.5.0
romenskiy2012:3.4.1
romenskiy2012:3.4.0
romenskiy2012:3.3.1
romenskiy2012:3.3.0
romenskiy2012:3.2.0
romenskiy2012:3.1.0
romenskiy2012:3.0.0
romenskiy2012:2.2.5
romenskiy2012:2.2.4
romenskiy2012:2.2.3
romenskiy2012:2.2.2
romenskiy2012:2.2.1
romenskiy2012:2.2.0
romenskiy2012:2.1.3
romenskiy2012:2.1.2
romenskiy2012:2.1.1
romenskiy2012:2.1.0
romenskiy2012:2.0.1
romenskiy2012:2.0.0
romenskiy2012:1.0.3
romenskiy2012:1.0.2
romenskiy2012:1.0.1
romenskiy2012:1.0.0
romenskiy2012:0.0.0
...
compare: romenskiy2012:dev
Branches Tags
romenskiy2012:dev
romenskiy2012:master
romenskiy2012:stable-4
romenskiy2012:stable-3
romenskiy2012:5.3.1
romenskiy2012:5.3.0
romenskiy2012:5.2.1
romenskiy2012:5.2.0
romenskiy2012:5.1.0
romenskiy2012:5.0.1
romenskiy2012:5.0.0
romenskiy2012:4.5.0
romenskiy2012:4.4.0
romenskiy2012:4.3.1
romenskiy2012:4.3.0
romenskiy2012:4.2.1
romenskiy2012:4.2.0
romenskiy2012:4.1.1
romenskiy2012:4.1.0
romenskiy2012:4.0.4
romenskiy2012:4.0.3
romenskiy2012:4.0.2
romenskiy2012:4.0.1
romenskiy2012:4.0.0
romenskiy2012:3.6.0
romenskiy2012:3.5.1
romenskiy2012:3.5.0
romenskiy2012:3.4.1
romenskiy2012:3.4.0
romenskiy2012:3.3.1
romenskiy2012:3.3.0
romenskiy2012:3.2.0
romenskiy2012:3.1.0
romenskiy2012:3.0.0
romenskiy2012:2.2.5
romenskiy2012:2.2.4
romenskiy2012:2.2.3
romenskiy2012:2.2.2
romenskiy2012:2.2.1
romenskiy2012:2.2.0
romenskiy2012:2.1.3
romenskiy2012:2.1.2
romenskiy2012:2.1.1
romenskiy2012:2.1.0
romenskiy2012:2.0.1
romenskiy2012:2.0.0
romenskiy2012:1.0.3
romenskiy2012:1.0.2
romenskiy2012:1.0.1
romenskiy2012:1.0.0
romenskiy2012:0.0.0

395 commits
5.3.0 ... dev

Author SHA1 Message Date
Guangli Dai
81034ce1f1
Update ChangeLog for release 5.3.1 2026-04-13 17:12:37 -07:00
Ian Ker-Seymer
b8646f4db3 Fix opt.max_background_threads default in docs 2026-04-13 14:46:53 -07:00
Guangli Dai
6515df8cec
Documentation updates (#2869) 
* Document new mallctl interfaces added since 5.3.0

Add documentation for the following new mallctl entries:
- opt.debug_double_free_max_scan: double-free detection scan limit
- opt.prof_bt_max: max profiling backtrace depth
- opt.disable_large_size_classes: page-aligned large allocations
- opt.process_madvise_max_batch: batched process_madvise purging
- thread.tcache.max: per-thread tcache_max control
- thread.tcache.ncached_max.read_sizeclass: query ncached_max
- thread.tcache.ncached_max.write: set ncached_max per size range
- arena.<i>.name: get/set arena names
- arenas.hugepage: hugepage size
- approximate_stats.active: lightweight active bytes estimate

Remove config.prof_frameptr since it still needs more development
and is still experimental.

Co-authored-by: lexprfuncall <carl.shapiro@gmail.com>
 2026-04-07 10:41:44 -07:00
Slobodan Predolac
f265645d02 Emit retained HPA slab stats in JSON 2026-04-01 23:15:19 -04:00
Slobodan Predolac
db7d99703d Add TODO to benchmark possibly better policy 2026-04-01 23:15:19 -04:00
Slobodan Predolac
6281482c39 Nest HPA SEC stats inside hpa_shard JSON 2026-04-01 23:15:19 -04:00
Slobodan Predolac
3cc56d325c Fix large alloc nrequests under-counting on cache misses 2026-04-01 23:15:19 -04:00
Slobodan Predolac
a47fa33b5a Run clang-format on test/unit/tcache_max.c 2026-04-01 23:15:19 -04:00
Slobodan Predolac
b507644cb0 Fix conf_handle_char_p zero-sized dest and remove unused conf_handle_unsigned 2026-04-01 23:15:19 -04:00
Slobodan Predolac
3ac9f96158 Run clang-format on test/unit/conf_parse.c 2026-04-01 23:15:19 -04:00
Slobodan Predolac
5904a42187 Fix memory leak of old curr_reg on san_bump_grow_locked failure 
When san_bump_grow_locked fails, it sets sba->curr_reg to NULL.
The old curr_reg (saved in to_destroy) was never freed or restored,
leaking the virtual memory extent. Restore sba->curr_reg from
to_destroy on failure so the old region remains usable.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
2fceece256 Fix extra size argument in edata_init call in extent_alloc_dss 
An extra 'size' argument was passed where 'slab' (false) should be,
shifting all subsequent arguments: slab got size (nonzero=true),
szind got false (0), and sn got SC_NSIZES instead of a proper serial
number from extent_sn_next(). Match the correct pattern used by the
gap edata_init call above.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
234404d324 Fix wrong loop variable for array index in sz_boot_pind2sz_tab 
The sentinel fill loop used sz_pind2sz_tab[pind] (constant) instead
of sz_pind2sz_tab[i] (loop variable), writing only to the first
entry repeatedly and leaving subsequent entries uninitialized.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
675ab079e7 Fix missing release of acquired neighbor edata in extent_try_coalesce_impl 
When emap_try_acquire_edata_neighbor returned a non-NULL neighbor but
the size check failed, the neighbor was never released from
extent_state_merging, making it permanently invisible to future
allocation and coalescing operations.

Release the neighbor when it doesn't meet the size requirement,
matching the pattern used in extent_recycle_extract.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
3f6e63e86a Fix wrong type for malloc_read_fd return value in prof_stack_range 
Used size_t (unsigned) instead of ssize_t for the return value of
malloc_read_fd, which returns -1 on error. With size_t, -1 becomes
a huge positive value, bypassing the error check and corrupting the
remaining byte count.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
dd30c91eaa Fix wrong fallback value in os_page_detect when sysconf fails 
Returned LG_PAGE (log2 of page size, e.g. 12) instead of PAGE (actual
page size, e.g. 4096) when sysconf(_SC_PAGESIZE) failed. This would
cause os_page to be set to an absurdly small value, breaking all
page-aligned operations.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
3a8bee81f1 Fix pac_mapped stats inflation on allocation failure 
newly_mapped_size was set unconditionally in the ecache_alloc_grow
fallback path, even when the allocation returned NULL. This inflated
pac_mapped stats without a corresponding deallocation to correct them.

Guard the assignment with an edata != NULL check, matching the pattern
used in the batched allocation path above it.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
c2d57040f0 Fix out-of-bounds write in malloc_vsnprintf when size is 0 
When called with size==0, the else branch wrote to str[size-1] which
is str[(size_t)-1], a massive out-of-bounds write. Standard vsnprintf
allows size==0 to mean "compute length only, write nothing".

Add unit test for the size==0 case.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
eab2b29736 Fix off-by-one in stats_arenas_i_bins_j and stats_arenas_i_lextents_j bounds checks 
Same pattern as arenas_bin_i_index: used > instead of >= allowing
access one past the end of bstats[] and lstats[] arrays.

Add unit tests that verify boundary indices return ENOENT.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
a0f2bdf91d Fix missing negation in large_ralloc_no_move usize_min fallback 
The second expansion attempt in large_ralloc_no_move omitted the !
before large_ralloc_no_move_expand(), inverting the return value.
On expansion failure, the function falsely reported success, making
callers believe the allocation was expanded in-place when it was not.
On expansion success, the function falsely reported failure, causing
callers to unnecessarily allocate, copy, and free.

Add unit test that verifies the return value matches actual size change.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
87f9938de5 Fix duplicate "nactive_huge" JSON key in HPA shard stats output 
In both the full_slabs and empty_slabs JSON sections of HPA shard
stats, "nactive_huge" was emitted twice instead of emitting
"ndirty_huge" as the second entry. This caused ndirty_huge to be
missing from the JSON output entirely.

Add a unit test that verifies both sections contain "ndirty_huge".
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
513778bcb1 Fix off-by-one in arenas_bin_i_index and arenas_lextent_i_index bounds checks 
The index validation used > instead of >=, allowing access at index
SC_NBINS (for bins) and SC_NSIZES-SC_NBINS (for lextents), which are
one past the valid range. This caused out-of-bounds reads in bin_infos[]
and sz_index2size_unsafe().

Add unit tests that verify the boundary indices return ENOENT.
 2026-04-01 23:15:19 -04:00
Slobodan Predolac
176ea0a801 Remove experimental.thread.activity_callback 2026-04-01 16:23:41 -07:00
Slobodan Predolac
19bbefe136 Remove dead code: extent_commit_wrapper, large_salloc, tcache_gc_dalloc event waits 
These functions had zero callers anywhere in the codebase:
- extent_commit_wrapper: wrapper never called, _impl used directly
- large_salloc: trivial wrapper never called
- tcache_gc_dalloc_new_event_wait: no header declaration, no callers
- tcache_gc_dalloc_postponed_event_wait: no header declaration, no callers
 2026-04-01 17:48:19 -04:00
Weixie Cui
a87c518bab Fix typo in prof_log_rep_check: use != instead of || for alloc_count 
The condition incorrectly used 'alloc_count || 0' which was likely a typo
for 'alloc_count != 0'. While both evaluate similarly for the zero/non-zero
case, the fix ensures consistency with bt_count and thr_count checks and
uses the correct comparison operator.
 2026-03-26 10:42:29 -07:00
Slobodan Predolac
d758349ca4 Fix psset_pick_purge when last candidate with index 0 dirtiness is ineligible 
psset_pick_purge used max_bit-- after rejecting a time-ineligible
candidate, which caused unnecessary re-scanning of the same bitmap
and makes assert fail in debug mode) and a size_t underflow
when the lowest-index entry was rejected.  Use max_bit = ind - 1
to skip directly past the rejected index.
 2026-03-26 10:39:37 -07:00
Tony Printezis
1d018d8fda improve hpdata_assert_consistent() 
A few ways this consistency check can be improved:
* Print which conditions fail and associated values.
* Accumulate the result so that we can print all conditions that fail.
* Turn hpdata_assert_consistent() into a macro so, when it fails,
  we can get line number where it's called from.
 2026-03-26 10:39:23 -07:00
Carl Shapiro
86b7219213 Add unit tests for conf parsing and its helpers 2026-03-10 18:14:33 -07:00
Carl Shapiro
ad726adf75 Separate out the configuration code from initialization 2026-03-10 18:14:33 -07:00
Carl Shapiro
a056c20d67 Handle tcache init failures gracefully 
tsd_tcache_data_init() returns true on failure but its callers ignore
this return value, leaving the per-thread tcache in an uninitialized
state after a failure.

This change disables the tcache on an initialization failure and logs
an error message.  If opt_abort is true, it will also abort.

New unit tests have been added to test tcache initialization failures.
 2026-03-10 18:14:33 -07:00
Carl Shapiro
a75655badf Add unit test coverage for bin interfaces 2026-03-10 18:14:33 -07:00
Carl Shapiro
0ac9380cf1 Move bin inline functions from arena_inlines_b.h to bin_inlines.h 
This is a continuation of my previous clean-up change, now focusing on
the inline functions defined in header files.
 2026-03-10 18:14:33 -07:00
Carl Shapiro
1cc563f531 Move bin functions from arena.c to bin.c 
This is a clean-up change that gives the bin functions implemented in
the area code a prefix of bin_ and moves them into the bin code.

To further decouple the bin code from the arena code, bin functions
that had taken an arena_t to check arena_is_auto now take an is_auto
parameter instead.
 2026-03-10 18:14:33 -07:00
guangli-dai
c73ab1c2ff Add a test to check the output in JSON-based stats is consistent with mallctl results. 2026-03-10 18:14:33 -07:00
guangli-dai
12b33ed8f1 Fix wrong mutex stats in json-formatted malloc stats 
During mutex stats emit, derived counters are not emitted for json.
Yet the array indexing counter should still be increased to skip
derived elements in the output, which was not. This commit fixes it.
 2026-03-10 18:14:33 -07:00
Carl Shapiro
79cc7dcc82 Guard os_page_id against a NULL address 
While undocumented, the prctl system call will set errno to ENOMEM
when passed NULL as an address.  Under that condition, an assertion
that check for EINVAL as the only possible errno value will fail.  To
avoid the assertion failure, this change skips the call to os_page_id
when address is NULL.  NULL can only occur after mmap fails in which
case there is no mapping to name.
 2026-03-10 18:14:33 -07:00
Yuxuan Chen
a10ef3e1f1 configure: add --with-cxx-stdlib option 
When C++ support is enabled, configure unconditionally probes
`-lstdc++` and keeps it in LIBS if the link test succeeds. On
platforms using libc++, this probe can succeed at compile time (if
libstdc++ headers/libraries happen to be installed) but then cause
runtime failures when configure tries to execute test binaries
because `libstdc++.so.6` isn't actually available.

Add a `--with-cxx-stdlib=<libstdc++|libcxx>` option that lets the
build system specify which C++ standard library to link. When given,
the probe is skipped and the specified library is linked directly.
When not given, the original probe behavior is preserved.
 2026-03-10 18:14:33 -07:00
Tony Printezis
0fa27fd28f Run single subtest from a test file 
Add mechanism to be able to select a test to run from a test file. The test harness will read the JEMALLOC_TEST_NAME env and, if set, it will only run subtests with that name.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
34ace9169b Remove prof_threshold built-in event. It is trivial to implement it as user event if needed 2026-03-10 18:14:33 -07:00
Andrei Pechkurov
4d0ffa075b Fix background thread initialization race 2026-03-10 18:14:33 -07:00
Slobodan Predolac
d4908fe44a Revert "Experimental configuration option for fast path prefetch from cache_bin" 
This reverts commit f9fae9f1f8.
 2026-03-10 18:14:33 -07:00
Carl Shapiro
c51abba131 Determine the page size on Android from NDK header files 
The definition of the PAGE_SIZE macro is used as a signal for a 32-bit
target or a 64-bit target with an older NDK.  Otherwise, a 16KiB page
size is assumed.

Closes: #2657
 2026-03-10 18:14:33 -07:00
Carl Shapiro
5f353dc283 Remove an incorrect use of the address operator 
The address of the local variable created_threads is a different
location than the data it points to.  Incorrectly treating these
values as being the same can cause out-of-bounds writes to the stack.

Closes: facebook/jemalloc#59
 2026-03-10 18:14:33 -07:00
Carl Shapiro
365747bc8d Use the BRE construct \{1,\} for one or more consecutive matches 
This removes duplication introduced by my earlier commit that
eliminating the use of the non-standard "\+" from BREs in the
configure script.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
6016d86c18 [SEC] Make SEC owned by hpa_shard, simplify the code, add stats, lock per bin 2026-03-10 18:14:33 -07:00
Slobodan Predolac
c7690e92da Remove Cirrus CI 2026-03-10 18:14:33 -07:00
Slobodan Predolac
441e840df7 Add a script to generate github actions instead of Travis CI and Cirrus 2026-03-10 18:14:33 -07:00
Guangli Dai
0988583d7c Add a mallctl for users to get an approximate of active bytes. 2026-03-10 18:14:33 -07:00
Slobodan Predolac
8a06b086f3 [EASY] Extract hpa_central component from hpa source file 2026-03-10 18:14:33 -07:00
Slobodan Predolac
355774270d [EASY] Encapsulate better, do not pass hpa_shard when hooks are enough, move shard independent actions to hpa_utils 2026-03-10 18:14:33 -07:00
Slobodan Predolac
47aeff1d08 Add experimental_enforce_hugify 2026-03-10 18:14:33 -07:00
Shirui Cheng
6d4611197e move fill/flush pointer array out of tcache.c 2026-03-10 18:14:33 -07:00
Slobodan Predolac
3678a57c10 When extracting from central, hugify_eager is different than start_as_huge 2026-03-10 18:14:33 -07:00
guangli-dai
2cfa41913e Refactor init_system_thp_mode and print it in malloc stats. 2026-03-10 18:14:33 -07:00
Slobodan Predolac
87555dfbb2 Do not release the hpa_shard->mtx when inserting newly retrieved page from central before allocating from it 2026-03-10 18:14:33 -07:00
Carl Shapiro
f714cd9249 Inline the value of an always false boolean local variable 
Next to its use, which is always as an argument, we include the name
of the parameter in a constant.  This completes a partially
implemented cleanup suggested in an earlier commit.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
5e49c28ef0 [EASY] Spelling in the comments 2026-03-10 18:14:33 -07:00
Slobodan Predolac
7c40be249c Add npurges and npurge_passes to output of pa_benchmark 2026-03-10 18:14:33 -07:00
Slobodan Predolac
707aab0c95 [pa-bench] Add clock to pa benchmark 2026-03-10 18:14:33 -07:00
Slobodan Predolac
a199278f37 [HPA] Add ability to start page as huge and more flexibility for purging 2026-03-10 18:14:33 -07:00
Slobodan Predolac
ace437d26a Running clang-format on two files 2026-03-10 18:14:33 -07:00
Slobodan Predolac
2688047b56 Revert "Do not dehugify when purging" 
This reverts commit 16c5abd1cd.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
de886e05d2 Revert "Remove an unused function and global variable" 
This reverts commit acd85e5359.
 2026-03-10 18:14:33 -07:00
guangli-dai
755735a6bf Remove Travis Windows CI for now since it has infra failures. 2026-03-10 18:14:33 -07:00
Slobodan Predolac
d70882a05d [sdt] Add some tracepoints to sec and hpa modules 2026-03-10 18:14:33 -07:00
Carl Shapiro
67435187d1 Improve the portability of grep patterns in configure.ac 
The configure.ac script uses backslash plus in its grep patterns to
match one or more occurrences.  This is a GNU grep extension to the
Basic Regular Expressions syntax that fails on systems with a more
traditional grep.  This changes fixes grep patterns that use backslash
plus to use a star instead.

Closes: #2777
 2026-03-10 18:14:33 -07:00
guangli-dai
261591f123 Add a page-allocator microbenchmark. 2026-03-10 18:14:33 -07:00
guangli-dai
56cdce8592 Adding trace analysis in preparation for page allocator microbenchmark. 2026-03-10 18:14:33 -07:00
Carl Shapiro
daf44173c5 Replace an instance of indentation with spaces with tabs 2026-03-10 18:14:33 -07:00
Aurélien Brooke
ce02945070 Add missing thread_event_registry.c to Visual Studio projects 
This file was added by b2a35a905f.
 2026-03-10 18:14:33 -07:00
lexprfuncall
c51949ea3e Update config.guess and config.sub to the latest versions 
These files need to be refreshed periodically to support new platform
types.

The following command was used to retrieve the updates

curl -L -O https://git.savannah.gnu.org/cgit/config.git/plain/config.guess
curl -L -O https://git.savannah.gnu.org/cgit/config.git/plain/config.sub

Closes: #2814
 2026-03-10 18:14:33 -07:00
Carl Shapiro
5a634a8d0a Always use pthread_equal to compare thread IDs 
This change replaces direct comparisons of Pthread thread IDs with
calls to pthread_equal.  Directly comparing thread IDs is neither
portable nor reliable since a thread ID is defined as an opaque type
that can be implemented using a structure.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
5d5f76ee01 Remove pidfd_open call handling and rely on PIDFD_SELF 2026-03-10 18:14:33 -07:00
lexprfuncall
9442300cc3 Change the default page size to 64KiB on Aarch64 Linux 
This updates the configuration script to set the default page size to
64KiB on Aarch64 Linux.  This is motivated by compatibility as a build
configured for a 64KiB page will work on kernels that use the smaller
4KiB or 16KiB pages, whereas the reverse is not true.

To make the configured page size setting more visible, the script now
displays the page size when printing the configuration results.

Users that want to override the page size in to choose a smaller value
can still do so with the --with-lg-pagesize configuration option.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
2a66c0be5a [EASY][BUGFIX] Spelling and format 2026-03-10 18:14:33 -07:00
lexprfuncall
38b12427b7 Define malloc_{write,read}_fd as non-inline global functions 
The static inline definition made more sense when these functions just
dispatched to a syscall wrapper.  Since they acquired a retry loop, a
non-inline definition makes more sense.
 2026-03-10 18:14:33 -07:00
lexprfuncall
9fdc1160c5 Handle interruptions and retries of read(2) and write(2) 2026-03-10 18:14:33 -07:00
lexprfuncall
48b4ad60a7 Remove an orphaned comment 
This was left behind when definitions of malloc_open and malloc_close
were abstracted from code that had followed.
 2026-03-10 18:14:33 -07:00
Shirui Cheng
2114349a4e Revert PR #2608: Manually revert commits 70c94d..f9c0b5 
Closes: #2707
 2026-03-10 18:14:33 -07:00
lexprfuncall
ced8b3cffb Fix the compilation check for process madvise 
An include of unistd.h is needed to make the declaration of the
syscall function visible to the compiler.  The include of sys/mman.h
is not used at all.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
5e98585b37 Save and restore errno when calling process_madvise 2026-03-10 18:14:33 -07:00
lexprfuncall
e4fa33148a Remove an unused function and global variable 
When the dehugify functionality was retired in an previous commit, a
dehugify-related function and global variable in a test was
accidentally left in-place causing builds that add -Werror to CFLAGS
to fail.
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
d73de95f72 Experimental configuration option for fast path prefetch from cache_bin 2026-03-10 18:14:33 -07:00
lexprfuncall
9528a2e2dd Use relaxed atomics to access the process madvise pid fd 
Relaxed atomics already provide sequentially consistent access to single
location data structures.
 2026-03-10 18:14:33 -07:00
lexprfuncall
a156e997d7 Do not dehugify when purging 
Giving the advice MADV_DONTNEED to a range of virtual memory backed by
a transparent huge page already causes that range of virtual memory to
become backed by regular pages.
 2026-03-10 18:14:33 -07:00
lexprfuncall
395e63bf7e Fix several spelling errors in comments 2026-03-10 18:14:33 -07:00
Slobodan Predolac
4246475b44 [process_madvise] Make init lazy so that python tests pass. Reset the pidfd on fork 2026-03-10 18:14:33 -07:00
Slobodan Predolac
f87bbab22c Add several USDT probes for hpa 2026-03-10 18:14:33 -07:00
Slobodan Predolac
711fff750c Add experimental support for usdt systemtap probes 2026-03-10 18:14:33 -07:00
guangli-dai
5847516692 Ignore the clang-format changes in the git blame. 2026-03-10 18:14:33 -07:00
guangli-dai
6200e8987f Reformat the codebase with the clang-format 18. 2026-03-10 18:14:33 -07:00
Shirui Cheng
a952a3b8b0 Update the default value for opt_experimental_tcache_gc and opt_calloc_madvise_threshold 2026-03-10 18:14:33 -07:00
Guangli Dai
e350c71571 Remove --enable-limit-usize-gap for cirrus CI since the config-time option is removed. 2026-03-10 18:14:33 -07:00
guangli-dai
95fc091b0f Update appveyor settings. 2026-03-10 18:14:33 -07:00
dzhao.ampere
c5547f9e64 test/unit/psset.c: fix SIGSEGV when PAGESIZE is large 
When hugepage is enabled and PAGESIZE is large, the test could
ask for a stack size larger than user limit. Allocating the
memory instead can avoid the failure.

Closes: #2408
 2026-03-10 18:14:33 -07:00
Slobodan Predolac
015b017973 [thread_event] Add support for user events in thread events when stats are enabled 2026-03-10 18:14:33 -07:00
Slobodan Predolac
e6864c6075 [thread_event] Remove macros from thread_event and replace with dynamic event objects 2026-03-10 18:14:33 -07:00
Qi Wang
1972241cd2 Remove unused options in the batched madvise unit tests. 2025-06-02 11:25:37 -07:00
Jason Evans
27d7960cf9 Revert "Extend purging algorithm with peak demand tracking" 
This reverts commit ad108d50f1.
 2025-06-02 10:44:37 -07:00
guangli-dai
edaab8b3ad Turn clang-format off for codes with multi-line commands in macros 2025-05-28 19:22:21 -07:00
guangli-dai
4531411abe Modify .clang-format to have declarations aligned 2025-05-28 19:22:21 -07:00
guangli-dai
1818170c8d Fix binshard.sh by specifying bin_shards for all sizes. 2025-05-28 19:21:49 -07:00
guangli-dai
fd60645260 Add one more check to double free validation. 2025-05-28 19:21:49 -07:00
Xin Yang
5e460bfea2 Refactor: use the cache_bin_sz_t typedef instead of direct uint16_t 
any future changes to the underlying data type for bin sizes
(such as upgrading from `uint16_t` to `uint32_t`) can be achieved
by modifying only the `cache_bin_sz_t` definition.

Signed-off-by: Xin Yang <yangxin.dev@bytedance.com>
 2025-05-22 10:43:33 -07:00
Xin Yang
9169e9272a Fix: Adjust CACHE_BIN_NFLUSH_BATCH_MAX size to prevent assert failures 
The maximum allowed value for `nflush_batch` is
`CACHE_BIN_NFLUSH_BATCH_MAX`. However, `tcache_bin_flush_impl_small`
could potentially declare an array of `emap_batch_lookup_result_t`
of size `CACHE_BIN_NFLUSH_BATCH_MAX + 1`. leads to a `VARIABLE_ARRAY`
assertion failure, observed when `tcache_nslots_small_max` is
configured to 2048. This patch ensures the array size does not exceed
the allowed maximum.

Signed-off-by: Xin Yang <yangxin.dev@bytedance.com>
 2025-05-22 10:27:09 -07:00
guangli-dai
f19a569216 Ignore formatting commit in blame. 2025-05-20 14:21:08 -07:00
Slobodan Predolac
b6338c4ff6 EASY - be explicit in non-vectorized hpa tests 2025-05-19 16:31:04 -07:00
guangli-dai
554185356b Sample format on tcache_max test 2025-05-19 15:06:13 -07:00
guangli-dai
3cee771cfa Modify .clang-format to make it more aligned with current freebsd style 2025-05-19 15:06:13 -07:00
Jiebin Sun
3c14707b01 To improve reuse efficiency, the maximum coalesced size for large extents 
in the dirty ecache has been limited. This patch was tested with real
workloads using ClickHouse (Clickbench Q35) on a system with 2x240 vCPUs.
The results showed a 2X in query per second (QPS) performance and
a reduction in page faults to 29% of the previous rate. Additionally,
microbenchmark testing involved 256 memory reallocations resizing
from 4KB to 16KB in one arena, which demonstrated a 5X performance
improvement.

Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
 2025-05-12 15:45:36 -07:00
guangli-dai
37bf846cc3 Fixes to prevent static analysis warnings. 2025-05-06 14:47:35 -07:00
guangli-dai
8347f1045a Renaming limit_usize_gap to disable_large_size_classes 2025-05-06 14:47:35 -07:00
Guangli Dai
01e9ecbeb2 Remove build-time configuration 'config_limit_usize_gap' 2025-05-06 14:47:35 -07:00
Slobodan Predolac
852da1be15 Add experimental option force using SYS_process_madvise 2025-04-28 18:45:30 -07:00
Slobodan Predolac
1956a54a43 [process_madvise] Use process_madvise across multiple huge_pages 2025-04-25 19:19:03 -07:00
Slobodan Predolac
0dfb4a5a1a Add output argument to hpa_purge_begin to count dirty ranges 2025-04-25 19:19:03 -07:00
Slobodan Predolac
cfa90dfd80 Refactor hpa purging to prepare for vectorized call across multiple pages 2025-04-25 19:19:03 -07:00
Qi Wang
a3910b9802 Avoid forced purging during thread-arena migration when bg thd is on. 2025-04-25 19:18:20 -07:00
guangli-dai
c23a6bfdf6 Add opt.limit_usize_gap to stats 2025-04-16 10:38:10 -07:00
guangli-dai
c20a63a765 Silence the uninitialized warning from clang. 2025-04-16 10:38:10 -07:00
Qi Wang
f81fb92a89 Remove Travis CI macOS configs (not supported anymore). 2025-04-14 15:27:38 -07:00
Slobodan Predolac
f19f49ef3e if process_madvise is supported, call it when purging hpa 2025-04-04 13:57:42 -07:00
Kaspar M. Rohrer
80e9001af3 Move `extern "C" specifications for C++ to where they are needed 
This should fix errors when compiling C++ code with modules enabled on clang.
 2025-03-31 10:41:51 -07:00
Shirui Cheng
3688dfb5c3 fix assertion error in huge_arena_auto_thp_switch() when b0 is deleted in unit test 2025-03-20 12:45:23 -07:00
Jay Lee
a4defdb854 detect false failure of strerror_r 
See tikv/jemallocator#108.

In a summary, test on `strerror_r` can fail due to reasons other
than `strerror_r` itself, so add an additional test to determine
the failure is expected.

Signed-off-by: Jay Lee <BusyJayLee@gmail.com>
 2025-03-17 17:50:20 -07:00
Shirui Cheng
e1a77ec558 Support THP with Huge Arena in PAC 2025-03-17 16:06:43 -07:00
Audrey Dutcher
86bbabac32 background_thread: add fallback for pthread_create dlsym 
If jemalloc is linked into a shared library, the RTLD_NEXT dlsym call
may fail since RTLD_NEXT is only specified to search all objects after
the current one in the loading order, and the pthread library may be
earlier in the load order. Instead of failing immediately, attempt one
more time to find pthread_create via RTLD_GLOBAL.

Errors cascading from this were observed on FreeBSD 14.1.
 2025-03-17 09:41:04 -07:00
Guangli Dai
81f35e0b55 Modify Travis tests to use frameptr when profiling 2025-03-13 17:15:42 -07:00
Guangli Dai
773b5809f9 Fix frame pointer based unwinder to handle changing stack range 2025-03-13 17:15:42 -07:00
Dmitry Ilvokhin
ad108d50f1 Extend purging algorithm with peak demand tracking 
Implementation inspired by idea described in "Beyond malloc efficiency
to fleet efficiency: a hugepage-aware memory allocator" paper [1].

Primary idea is to track maximum number (peak) of active pages in use
with sliding window and then use this number to decide how many dirty
pages we would like to keep.

We are trying to estimate maximum amount of active memory we'll need in
the near future. We do so by projecting future active memory demand
(based on peak active memory usage we observed in the past within
sliding window) and adding slack on top of it (an overhead is reasonable
to have in exchange of higher hugepages coverage). When peak demand
tracking is off, projection of future active memory is active memory we
are having right now.

Estimation is essentially the same as `nactive_max * (1 + dirty_mult)`.

Peak demand purging algorithm controlled by two config options. Option
`hpa_peak_demand_window_ms` controls duration of sliding window we track
maximum active memory usage in and option `hpa_dirty_mult` controls
amount of slack we are allowed to have as a percent from maximum active
memory usage. By default `hpa_peak_demand_window_ms == 0` now and we
have same behaviour (ratio based purging) that we had before this
commit.

[1]: https://storage.googleapis.com/gweb-research2023-media/pubtools/6170.pdf
 2025-03-13 10:12:22 -07:00
Qi Wang
22440a0207 Implement process_madvise support. 
Add opt.process_madvise_max_batch which determines if process_madvise is enabled
(non-zero) and the max # of regions in each batch.  Added another limiting
factor which is the space to reserve on stack, which results in the max batch of
128.
 2025-03-07 15:32:32 -08:00
Guangli Dai
70f019cd3a Enable limit-usize-gap in CI tests. 
Considering the new usize calculation will be default soon, add the
config option in for Travis, Cirrus and appveyor.
 2025-03-06 15:08:13 -08:00
Guangli Dai
6035d4a8d3 Cache extra extents in the dirty pool from ecache_alloc_grow 2025-03-06 15:08:13 -08:00
guangli-dai
c067a55c79 Introducing a new usize calculation policy 
Converting size to usize is what jemalloc has been done by ceiling
size to the closest size class. However, this causes lots of memory
wastes with HPA enabled.  This commit changes how usize is calculated so
that the gap between two contiguous usize is no larger than a page.
Specifically, this commit includes the following changes:

1. Adding a build-time config option (--enable-limit-usize-gap) and a
runtime one (limit_usize_gap) to guard the changes.
When build-time
config is enabled, some minor CPU overhead is expected because usize
will be stored and accessed apart from index.  When runtime option is
also enabled (it can only be enabled with the build-time config
enabled). a new usize calculation approach wil be employed.  This new
calculation will ceil size to the closest multiple of PAGE for all sizes
larger than USIZE_GROW_SLOW_THRESHOLD instead of using the size classes.
Note when the build-time config is enabled, the runtime option is
default on.

2. Prepare tcache for size to grow by PAGE over GROUP*PAGE.
To prepare for the upcoming changes where size class grows by PAGE when
larger than NGROUP * PAGE, disable the tcache when it is larger than 2 *
NGROUP * PAGE. The threshold for tcache is set higher to prevent perf
regression as much as possible while usizes between NGROUP * PAGE and 2 *
NGROUP * PAGE happen to grow by PAGE.

3. Prepare pac and hpa psset for size to grow by PAGE over GROUP*PAGE
For PAC, to avoid having too many bins, arena bins still have the same
layout.  This means some extra search is needed for a page-level request that
is not aligned with the orginal size class: it should also search the heap
before the current index since the previous heap might also be able to
have some allocations satisfying it.  The same changes apply to HPA's
psset.
This search relies on the enumeration of the heap because not all allocs in
the previous heap are guaranteed to satisfy the request.  To balance the
memory and CPU overhead, we currently enumerate at most a fixed number
of nodes before concluding none can satisfy the request during an
enumeration.

4. Add bytes counter to arena large stats.
To prepare for the upcoming usize changes, stats collected by
multiplying alive allocations and the bin size is no longer accurate.
Thus, add separate counters to record the bytes malloced and dalloced.

5. Change structs use when freeing to avoid using index2size for large sizes.
  - Change the definition of emap_alloc_ctx_t
  - Change the read of both from edata_t.
  - Change the assignment and usage of emap_alloc_ctx_t.
  - Change other callsites of index2size.
Note for the changes in the data structure, i.e., emap_alloc_ctx_t,
will be used when the build-time config (--enable-limit-usize-gap) is
enabled but they will store the same value as index2size(szind) if the
runtime option (opt_limit_usize_gap) is not enabled.

6. Adapt hpa to the usize changes.
Change the settings in sec to limit is usage for sizes larger than
USIZE_GROW_SLOW_THRESHOLD and modify corresponding tests.

7. Modify usize calculation and corresponding tests.
Change the sz_s2u_compute. Note sz_index2size is not always safe now
while sz_size2index still works as expected.
 2025-03-06 15:08:13 -08:00
Guangli Dai
ac279d7e71 Fix profiling sample metadata lookup during xallocx 2025-03-04 14:42:04 -08:00
Qi Wang
f55e0c3f5c Remove unsupported Cirrus CI config 2025-03-03 16:29:04 -08:00
Dmitry Ilvokhin
499f306859 Fix arena 0 deferral_allowed flag init 
Arena 0 have a dedicated initialization path, which differs from
initialization path of other arenas. The main difference for the purpose
of this change is that we initialize arena 0 before we initialize
background threads. HPA shard options have `deferral_allowed` flag which
should be equal to `background_thread_enabled()` return value, but it
wasn't the case before this change, because for arena 0
`background_thread_enabled()` was initialized correctly after arena 0
initialization phase already ended.

Below is initialization sequence for arena 0 after this commit to
illustrate everything still should be initialized correctly.

* `hpa_central_init` initializes HPA Central, before we initialize every
  HPA shard (including arena's 0).
* `background_thread_boot1` initializes `background_thread_enabled()`
  return value.
* `pa_shard_enable_hpa` initializes arena 0 HPA shard.

```
                       malloc_init_hard -------------
                      /           /                  \
                     /           /                    \
                    /           /                      \
malloc_init_hard_a0_locked  background_thread_boot1  pa_shard_enable_hpa
        /                     /                          \
       /                     /                            \
      /                     /                              \
arena_boot       background_thread_enabled_seta         hpa_shard_init
     |
     |
pa_central_init
     |
     |
hpa_central_init
```
 2025-02-18 12:10:35 -08:00
Dmitry Ilvokhin
421b17a622 Remove age_counter from hpa_central 
Before this commit we had two age counters: one global in HPA central
and one local in each HPA shard. We used HPA shard counter, when we are
reused empty pageslab and HPA central counter anywhere else. They
suppose to be comparable, because we use them for allocation placement
decisions, but in reality they are not, there is no ordering guarantees
between them.

At the moment, there is no way for pageslab to migrate between HPA
shards, so we don't actually need HPA central age counter.
 2025-02-13 16:00:41 -08:00
roblabla
c17bf8b368 Disable config from file or envvar with build flag 
This adds a new autoconf flag, --disable-user-config, which disables
reading the configuration from /etc/malloc.conf or the MALLOC_CONF
environment variable. This can be useful when integrating jemalloc in a
binary that internally handles all aspects of the configuration and
shouldn't be impacted by ambient change in the environment.
 2025-02-05 15:01:50 -08:00
Dmitry Ilvokhin
34c823f147 Add autoconf options to enable sanitizers 
This commit allows to enable sanitizers with autoconf options, instead
of modifying `CFLAGS`, `CXXFLAGS` and `LDFLAGS` directly.

* `--enable-tsan` option to enable Thread Sanitizer.
* `--enable-ubsan` option to enable Undefined Behaviour Sanitizer.

End goal is to speedup development by finding problems quickly, early
and easier. Eventually, when all current issues will be fixed, we can
enable sanitizers in CI. Fortunately, there are not a lot of problems we
need to fix.

Address Sanitizer is a bit controversial, because it replaces memory
allocator, so we decided to left it out for a while.

Below are couple of examples of how tests look like under different
sanitizers at the moment.

```
$  ../configure --enable-tsan --enable-debug
<...>
asan               : 0
tsan               : 1
ubsan              : 0
$ make -j`nproc` check
<...>
  Thread T13 (tid=332043, running) created by main thread at:
    #0 pthread_create <null> (libtsan.so.0+0x61748)
    #1 thd_create ../test/src/thd.c:25 (bin_batching+0x5631ca)
    #2 stress_run ../test/unit/bin_batching.c:148
(bin_batching+0x40364c)
    #3 test_races ../test/unit/bin_batching.c:249
(bin_batching+0x403d79)
    #4 p_test_impl ../test/src/test.c:149 (bin_batching+0x562811)
    #5 p_test_no_reentrancy ../test/src/test.c:213
(bin_batching+0x562d35)
    #6 main ../test/unit/bin_batching.c:268 (bin_batching+0x40417e)

SUMMARY: ThreadSanitizer: data race
../include/jemalloc/internal/edata.h:498 in edata_nfree_inc
```

```
$ ../configure --enable-ubsan --enable-debug
<...>
asan               : 0
tsan               : 0
ubsan              : 1
$ make -j`nproc` check
<...>
=== test/unit/hash ===
../test/unit/hash.c:119:16: runtime error: left shift of 176 by 24
places cannot be represented in type 'int'
<...>
```
 2025-02-05 14:28:28 -08:00
Qi Wang
3bc89cfeca Avoid implicit conversion in test/unit/prof_threshold 2025-01-31 10:18:36 -08:00
Qi Wang
1abeae9ebd Fix test/unit/prof_threshold when !config_stats 2025-01-30 10:39:49 -08:00
Shai Duvdevani
257e64b968 Unlike prof_sample which is supported only with profiling mode active, prof_threshold is intended to be an always-supported allocation callback with much less overhead. The usage of the threshold allows performance critical callers to change program execution based on the callback: e.g. drop caches when memory becomes high or to predict the program is about to OOM ahead of time using peak memory watermarks. 2025-01-29 18:55:52 -08:00
Dmitry Ilvokhin
ef8e512e29 Fix bitmap_ffu out of range read 
We tried to load `g` from `bitmap[i]` before checking it is actually a
valid load. Tweaked a loop a bit to `break` early, when we are done
scanning for bits.

Before this commit undefined behaviour sanitizer from GCC 14+ was
unhappy at `test/unit/bitmap` test with following error.

```
../include/jemalloc/internal/bitmap.h:293:5: runtime error: load of
address 0x7bb1c2e08008 with insufficient space for an object of type
'const bitmap_t'
<...>
    #0 0x62671a149954 in bitmap_ffu ../include/jemalloc/internal/bitmap.h:293
    #1 0x62671a149954 in test_bitmap_xfu_body ../test/unit/bitmap.c:275
    #2 0x62671a14b767 in test_bitmap_xfu ../test/unit/bitmap.c:323
    #3 0x62671a376ad1 in p_test_impl ../test/src/test.c:149
    #4 0x62671a377135 in p_test ../test/src/test.c:200
    #5 0x62671a13da06 in main ../test/unit/bitmap.c:336
<...>
```
 2025-01-28 10:42:20 -08:00
Qi Wang
607b866035 Check for 0 input when setting max_background_thread through mallctl. 
Reported by @nc7s.
 2025-01-28 10:38:56 -08:00
Qi Wang
20cc983314 Fix the gettid() detection caught by @mrluanma . 2025-01-22 10:30:53 -08:00
Dmitry Ilvokhin
52fa9577ba Fix integer overflow in test/unit/hash.c 
`final[3]` is `uint8_t`. Integer conversion rank of `uint8_t` is lower
than integer conversion rank of `int`, so `uint8_t` got promoted to
`int`, which is signed integer type. Shift `final[3]` value left on 24,
when leftmost bit is set overflows `int` and it is undefined behaviour.

Before this change Undefined Behaviour Sanitizer was unhappy about it
with the following message.

```
../test/unit/hash.c:119:25: runtime error: left shift of 176 by 24
places cannot be represented in type 'int'
```

After this commit problem is gone.
 2025-01-17 12:54:22 -08:00
Dan Horák
17881ebbfd Add configure check for gettid() presence 
The gettid() function is available on Linux in glibc only since version
2.30. There are supported distributions that still use older glibc
version. Thus add a configure check if the gettid() function is
available and extend the check in src/prof_stack_range.c so it's skipped
also when gettid() isn't available.

Fixes: https://github.com/jemalloc/jemalloc/issues/2740
 2024-12-17 12:40:54 -08:00
appujee
4b88bddbca Conditionally remove unreachable for C23+ 2024-12-17 12:39:00 -08:00
appujee
d8486b2653 Remove unreachable() macro as c23 already defines it. 
Taken from https://android-review.git.corp.google.com/c/platform/external/jemalloc_new/+/3316478

This might need more cleanups to remove the definition of JEMALLOC_INTERNAL_UNREACHABLE.
 2024-12-17 12:39:00 -08:00
Guangli Dai
587676fee8 Disable psset test when hugepage size is too large. 2024-12-17 12:35:35 -08:00
Guangli Dai
a17385a882 Enable large hugepage tests for arm64 on Travis 2024-12-17 12:35:35 -08:00
Guangli Dai
6786934280 Fix ehooks assertion for arena creation 2024-12-11 13:33:32 -08:00
Dmitry Ilvokhin
46690c9ec0 Fix test_retained on boxes with a lot of CPUs 
We are trying to create `ncpus * 2` threads for this test and place them
into `VARIABLE_ARRAY`, but `VARIABLE_ARRAY` can not be more than
`VARIABLE_ARRAY_SIZE_MAX` bytes. When there are a lot of threads on the
box test always fails.

```
$ nproc
176

$ make -j`nproc` tests_unit && ./test/unit/retained
<jemalloc>: ../test/unit/retained.c:123: Failed assertion:
"sizeof(thd_t) * (nthreads) <= VARIABLE_ARRAY_SIZE_MAX"
Aborted (core dumped)
```

There is no need for high concurrency for this test as we are only
checking stats there and it's behaviour is quite stable regarding number
of allocating threads.

Limited number of threads to 16 to save compute resources (on CI for
example) and reduce tests running time.

Before the change (`nproc` is 80 on this box).

```
$ make -j`nproc` tests_unit && time ./test/unit/retained
<...>
real    0m0.372s
user    0m14.236s
sys     0m12.338s
```

After the change (same box).

```
$ make -j`nproc` tests_unit && time ./test/unit/retained
<...>
real    0m0.018s
user    0m0.108s
sys     0m0.068s
```
 2024-12-02 14:12:26 -08:00
Dmitry Ilvokhin
6092c980a6 Expose psset state stats 
When evaluating changes in HPA logic, it is useful to know internal
`hpa_shard` state. Great deal of this state is `psset`. Some of the
`psset` stats was available, but in disaggregated form, which is not
very convenient. This commit exposed `psset` counters to `mallctl`
and malloc stats dumps.

Example of how malloc stats dump will look like after the change.

HPA shard stats:
  Pageslabs: 14899 (4354 huge, 10545 nonhuge)
  Active pages: 6708166 (2228917 huge, 4479249 nonhuge)
  Dirty pages: 233816 (331 huge, 233485 nonhuge)
  Retained pages: 686306
  Purge passes: 8730 (10 / sec)
  Purges: 127501 (146 / sec)
  Hugeifies: 4358 (5 / sec)
  Dehugifies: 4 (0 / sec)

Pageslabs, active pages, dirty pages and retained pages are rows added
by this change.
 2024-11-21 09:23:32 -08:00
Dmitry Ilvokhin
3820e38dc1 Remove validation for HPA ratios 
Config validation was introduced at 3aae792b with main intention to fix
infinite purging loop, but it didn't actually fix the underlying
problem, just masked it. Later 47d69b4ea was merged to address the same
problem.

Options `hpa_dirty_mult` and `hpa_hugification_threshold` have different
application dimensions: `hpa_dirty_mult` applied to active memory on the
shard, but `hpa_hugification_threshold` is a threshold for single
pageslab (hugepage). It doesn't make much sense to sum them up together.

While it is true that too high value of `hpa_dirty_mult` and too low
value of `hpa_hugification_threshold` can lead to pathological
behaviour, it is true for other options as well. Poor configurations
might lead to suboptimal and sometimes completely unacceptable
behaviour and that's OK, that is exactly the reason why they are called
poor.

There are other mechanism exist to prevent extreme behaviour, when we
hugified and then immediately purged page, see
`hpa_hugify_blocked_by_ndirty` function, which exist to prevent exactly
this case.

Lastly, `hpa_dirty_mult + hpa_hugification_threshold >= 1` constraint is
too tight and prevents a lot of valid configurations.
 2024-11-20 18:59:07 -08:00
Dmitry Ilvokhin
0ce13c6fb5 Add opt hpa_hugify_sync to hugify synchronously 
Linux 6.1 introduced `MADV_COLLAPSE` flag to perform a best-effort
synchronous collapse of the native pages mapped by the memory range into
transparent huge pages.

Synchronous hugification might be beneficial for at least two reasons:
we are not relying on khugepaged anymore and get an instant feedback if
range wasn't hugified.

If `hpa_hugify_sync` option is on, we'll try to perform synchronously
collapse and if it wasn't successful, we'll fallback to asynchronous
behaviour.
 2024-11-20 10:52:52 -08:00
Dmitry Ilvokhin
a361e886e2 Move je_cv_thp logic closer to definition 2024-11-20 10:52:52 -08:00
Dmitry Ilvokhin
b82333fdec Split stats_arena_hpa_shard_print function 
Make multiple functions from `stats_arena_hpa_shard_print` for
readability and ease of change in the future.
 2024-11-08 12:18:15 -08:00
Dmitry Ilvokhin
b9758afff0 Add nstime_ms_since to get time since in ms 
Milliseconds are used a lot in hpa, so it is convenient to have
`nstime_ms_since` function instead of dividing to `MILLION` constantly.

For consistency renamed `nstime_msec` to `nstime_ms` as `ms` abbreviation
is used much more commonly across codebase than `msec`.

```
$ grep -Rn '_msec' include src | wc -l
2

$ grep -RPn '_ms( |,|:)' include src | wc -l
72
```

Function `nstime_msec` wasn't used anywhere in the code yet.
 2024-11-08 10:37:28 -08:00
Qi Wang
2a693b83d2 Fix the sized-dealloc safety check abort msg. 2024-10-14 10:34:15 -07:00
Qi Wang
6d625d5e5e Add support for clock_gettime_nsec_np() 
Prefer clock_gettime_nsec_np(CLOCK_UPTIME_RAW) to mach_absolute_time().
 2024-10-14 10:33:27 -07:00
Guangli Dai
397827a27d Updated jeprof with more symbols to filter. 2024-10-14 10:31:58 -07:00
Qi Wang
02251c0070 Update the configure cache file example in INSTALL.md 2024-10-10 16:41:48 -07:00
Qi Wang
8c2b8bcf24 Update doc to reflect muzzy decay is disabled by default. 
It has been disabled since 5.2.0 (in #1421).
 2024-10-10 16:41:23 -07:00
Nathan Slingerland
edc1576f03 Add safe frame-pointer backtrace unwinder 2024-10-01 11:01:56 -07:00
Ben Niu
3a0d9cdadb Use MSVC __declspec(thread) for TSD on Windows 2024-09-30 11:33:44 -07:00
Guangli Dai
1c900088c3 Do not support hpa if HUGEPAGE is too large. 2024-09-27 15:34:13 -07:00
Dmitry Ilvokhin
4f4fd42447 Remove strict_min_purge_interval option 
Option `experimental_hpa_strict_min_purge_interval` was expected to be
temporary to simplify rollout of a bugfix. Now, when bugfix rollout is
complete it is safe to remove this option.
 2024-09-25 11:49:18 -07:00
Qi Wang
6cc42173cb Assert the mutex is locked within malloc_mutex_assert_owner(). 2024-09-23 18:06:07 -07:00
Qi Wang
44db479fad Fix the lock owner sanity checking during background thread boot. 
During boot, some mutexes are not initialized yet, plus there's no point taking
many mutexes while everything is covered by the global init lock, so the locking
assumptions in some functions (e.g. background_thread_enabled_set()) can't be
enforced.  Skip the lock owner check in this case.
 2024-09-23 18:06:07 -07:00
Guangli Dai
0181aaa495 Optimize edata_cmp_summary_compare when __uint128_t is available 2024-09-23 16:23:42 -07:00
roblabla
734f29ce56 Fix compilation with MSVC 2022 
On MSVC, log is an intrinsic that doesn't require libm. However,
AC_SEARCH_LIBS does not successfully detect this, as it will try to
compile a program using the wrong signature for log. Newer versions of
MSVC CL detects this and rejects the program with the following
messages:

conftest.c(40): warning C4391: 'char log()': incorrect return type for intrinsic function, expected 'double'
conftest.c(44): error C2168: 'log': too few actual parameters for intrinsic function

Since log is always available on MSVC (it's been around since the dawn
of time), we simply always assume it's there if MSVC is detected.
 2024-09-23 10:42:31 -07:00
Qi Wang
de5606d0d8 Fix a missing init value warning caught by static analysis. 2024-09-20 16:56:07 -07:00
Qi Wang
1960536b61 Add malloc_mutex_is_locked() sanity checks. 2024-09-20 16:56:07 -07:00
Qi Wang
3eb7a4b53d Fix mutex state tracking around pthread_cond_wait(). 
pthread_cond_wait drops and re-acquires the mutex internally, w/o
going through our wrapper.  Update the locked state explicitly.
 2024-09-20 16:56:07 -07:00
Qi Wang
661fb1e672 Fix the locked flag for malloc_mutex_trylock(). 2024-09-20 16:56:07 -07:00
Guangli Dai
db4f0e7182 Add travis tests for arm64. 2024-09-12 15:40:04 -07:00
Nathan Slingerland
8c2e15d1a5 Add malloc_open() / malloc_close() reentrancy safe helpers 2024-09-12 15:38:08 -07:00
Nathan Slingerland
60f472f367 Fix initialization of pop_attempt_results in bin_batching test 2024-09-12 11:36:17 -07:00
Qi Wang
323ed2e3a8 Optimize fast path to allow static size class computation. 
After inlining at LTO time, many callsites have input size known which means the
index and usable size can be translated at compile time.  However the size-index
lookup table prevents it -- this commit solves that by switching to the compute
approach when the size is detected to be a known const.
 2024-09-12 11:34:09 -07:00
Qi Wang
c1a3ca3755 Adjust the value width in stats output. 
Some of the values are accumulative and can reach high after running for long
periods.
 2024-09-11 14:29:32 -07:00
Qi Wang
3383b98f1b Check if the huge page size is expected when enabling HPA. 2024-09-04 15:43:59 -07:00
Qi Wang
cd05b19f10 Fix the VM over-reservation on aarch64 w/ larger pages. 
HUGEPAGE could be larger on some platforms (e.g. 512M on aarch64 w/ 64K pages),
in which case it would cause grow_retained / exp_grow to over-reserve VMs.

Similarly, make sure the base alloc has a const 2M alignment.
 2024-09-04 15:43:59 -07:00
Shirui Cheng
baa5a90cc6 fix nstime_update_mock in arena_decay unit test 2024-08-29 10:50:33 -07:00
Shirui Cheng
7c99686165 Better handle burst allocation on tcache_alloc_small_hard 2024-08-29 10:50:33 -07:00
Shirui Cheng
0c88be9e0a Regulate GC frequency by requiring a time interval between two consecutive GCs 2024-08-29 10:50:33 -07:00
Shirui Cheng
e2c9f3a9ce Take locality into consideration when doing GC flush 2024-08-29 10:50:33 -07:00
Shirui Cheng
14d5dc136a Allow a range for the nfill passed to arena_cache_bin_fill_small 2024-08-29 10:50:33 -07:00
Shirui Cheng
f68effe4ac Add a runtime option opt_experimental_tcache_gc to guard the new design 2024-08-29 10:50:33 -07:00
Ben Niu
9e123a833c Leverage new Windows API TlsGetValue2 for performance 2024-08-28 16:50:33 -07:00
Qi Wang
e29ac61987 Limit Cirrus CI to freebsd 15 and 14 2024-08-28 16:33:36 -07:00
Qi Wang
bd0a5b0f3b Fix static analysis warnings. 
Newly reported warnings included several reserved macro identifier, and
false-positive used-uninitialized.
 2024-08-28 16:03:53 -07:00
Guangli Dai
5b72ac098a Remove tests for ppc64 on Travic CI. 2024-08-26 09:53:00 -07:00
Shirui Cheng
8c54637f8c Better trigger race condition in bin_batching unit test 2024-08-23 14:10:04 -07:00
Dmitry Ilvokhin
c7ccb8d7e9 Add experimental prefix to hpa_strict_min_purge_interval 
Goal is to make it obvious this option is experimental.
 2024-08-20 10:02:38 -07:00
Dmitry Ilvokhin
aaa29003ab Limit maximum number of purged slabs with option 
Option `experimental_hpa_max_purge_nhp` introduced for backward
compatibility reasons: to make it possible to have behaviour similar
to buggy `hpa_strict_min_purge_interval` implementation.

When `experimental_hpa_max_purge_nhp` is set to -1, there is no limit
to number of slabs we'll purge on each iteration. Otherwise, we'll purge
no more than `experimental_hpa_max_purge_nhp` hugepages (slabs). This in
turn means we might not purge enough dirty pages to satisfy
`hpa_dirty_mult` requirement.

Combination of `hpa_dirty_mult`, `experimental_hpa_max_purge_nhp` and
`hpa_strict_min_purge_interval` options allows us to have steady rate of
pages returned back to the system. This provides a strickier latency
guarantees as number of `madvise` calls is bounded (and hence number of
TLB shootdowns is limited) in exchange to weaker memory usage
guarantees.
 2024-08-20 10:02:38 -07:00
Dmitry Ilvokhin
143f458188 Fix hpa_strict_min_purge_interval option logic 
We update `shard->last_purge` on each call of `hpa_try_purge` if we
purged something. This means, when `hpa_strict_min_purge_interval`
option is set only one slab will be purged, because on the next
call condition for too frequent purge protection
`since_last_purge_ms < shard->opts.min_purge_interval_ms` will always
be true. This is not an intended behaviour.

Instead, we need to check `min_purge_interval_ms` once and purge as many
pages as needed to satisfy requirements for `hpa_dirty_mult` option.

Make possible to count number of actions performed in unit tests (purge,
hugify, dehugify) instead of binary: called/not called. Extended current
unit tests with cases where we need to purge more than one page for a
purge phase.
 2024-08-20 10:02:38 -07:00
Dmitry Ilvokhin
0a9f51d0d8 Simplify hpa_shard_maybe_do_deferred_work 
It doesn't make much sense to repeat purging once we done with
hugification, because we can de-hugify pages that were hugified just
moment ago for no good reason. Let them wait next deferred work phase
instead. And if they still meeting purging conditions then, purge them.
 2024-08-20 10:02:38 -07:00
Amaury Séchet
a25b9b8ba9 Simplify the logic when bumping lg_fill_div. 2024-08-06 13:31:49 -07:00
Shirui Cheng
8fefabd3a4 increase the ncached_max in fill_flush test case to 1024 2024-08-06 13:16:09 -07:00
Shirui Cheng
47c9bcd402 Use a for-loop to fulfill flush requests that are larger than CACHE_BIN_NFLUSH_BATCH_MAX items 2024-08-06 13:16:09 -07:00
Shirui Cheng
48f66cf4a2 add a size check when declare a stack array to be less than 2048 bytes 2024-08-06 13:16:09 -07:00
Burton Li
8dc97b1108 Fix NSTIME_MONOTONIC for win32 implementation 2024-07-30 10:30:41 -07:00
Nathan Slingerland
bc32ddff2d Add usize to prof_sample_hook_t 2024-07-30 10:29:30 -07:00
Dmitry Ilvokhin
b66f689764 Emit long string values without truncation 
There are few long options (`bin_shards` and `slab_sizes` for example)
when they are specified and we emit statistics value gets truncated.

Moved emitting logic for strings into separate `emitter_emit_str`
function. It will try to emit string same way as before and if value is
too long will fallback emiting rest partially with chunks of `BUF_SIZE`.

Justification for long strings (longer than `BUF_SIZE`) is not
supported.
 2024-07-29 13:58:31 -07:00
Danny Lin
c893fcd169 Change macOS mmap tag to fix conflict with CoreMedia 
Tag 101 is assigned to "CoreMedia Capture Data", which makes for confusing output when debugging.

To avoid conflicts, use a tag in the reserved application-specific range from 240–255 (inclusive).

All assigned tags: 94d3b45284/osfmk/mach/vm_statistics.h (L773-L775)
 2024-06-26 14:53:48 -07:00
Shirui Cheng
a1fcbebb18 skip tcache GC for tcache_max unit test 2024-06-25 12:59:45 -07:00
Guangli Dai
8477ec9562 Set dependent as false for all rtree reads without ownership 2024-06-24 10:50:20 -07:00
Guangli Dai
21bcc0a8d4 Make JEMALLOC_CXX_THROW definition compatible with newer C++ versions 2024-06-13 11:03:05 -07:00
Dmitry Ilvokhin
867c6dd7dc Option to guard hpa_min_purge_interval_ms fix 
Change in `hpa_min_purge_interval_ms` handling logic is not backward
compatible as it might increase memory usage. Now this logic guarded by
`hpa_strict_min_purge_interval` option.

When `hpa_strict_min_purge_interval` is true, we will purge no more than
`hpa_min_purge_interval_ms`. When `hpa_strict_min_purge_interval` is
false, old purging logic behaviour is preserved.

Long term strategy migrate all users of hpa to new logic and then delete
`hpa_strict_min_purge_interval` option.
 2024-06-07 10:52:41 -07:00
Dmitry Ilvokhin
91a6d230db Respect hpa_min_purge_interval_ms option 
Currently, hugepages aware allocator backend works together with classic
one as a fallback for not yet supported allocations. When background
threads are enabled wake up time for classic interfere with hpa as there
were no checks inside hpa purging logic to check if we are not purging too
frequently. If background thread is running and `hpa_should_purge`
returns true, then we will purge, even if we purged less than
hpa_min_purge_interval_ms ago.
 2024-06-07 10:52:41 -07:00
Dmitry Ilvokhin
90c627edb7 Export hugepage size with arenas.hugepage 2024-06-05 15:37:41 -07:00
David Goldblatt
f9c0b5f7f8 Bin batching: add some stats. 
This lets us easily see what fraction of flush load is being taken up by the
bins, and helps guide future optimization approaches (for example: should we
prefetch during cache bin fills? It depends on how many objects the average fill
pops out of the batch).
 2024-05-22 10:30:31 -07:00
David Goldblatt
fc615739cb Add batching to arena bins. 
This adds a fast-path for threads freeing a small number of allocations to
bins which are not their "home-base" and which encounter lock contention in
attempting to do so. In producer-consumer workflows, such small lock hold times
can cause lock convoying that greatly increases overall bin mutex contention.
 2024-05-22 10:30:31 -07:00
David Goldblatt
44d91cf243 Tcache flush: Partition by bin before locking. 
This accomplishes two things:
- It avoids a full array scan (and any attendant branch prediction misses, etc.)
  while holding the bin lock.
- It allows us to know the number of items that will be flushed before flushing
  them, which will (in an upcoming commit) let us know if it's safe to use the
  batched flush (in which case we won't acquire the bin mutex).
 2024-05-22 10:30:31 -07:00
David Goldblatt
6e56848850 Tcache: Split up small/large handling. 
The main bits of shared code are the edata filtering and the stats flushing
logic, both of which are fairly simple to read and not so painful to duplicate.
The shared code comes at the cost of guarding all the subtle logic with
`if (small)`, which doesn't feel worth it.
 2024-05-22 10:30:31 -07:00
David Goldblatt
c085530c71 Tcache batching: Plumbing 
In the next commit, we'll start using the batcher to eliminate mutex traffic.
To avoid cluttering up that commit with the random bits of busy-work it entails,
we'll centralize them here.  This commit introduces:
- A batched bin type.
- The ability to mix batched and unbatched bins in the arena.
- Conf parsing to set batches per size and a max batched size.
- mallctl access to the corresponding opt-namespace keys.
- Stats output of the above.
 2024-05-22 10:30:31 -07:00
David Goldblatt
70c94d7474 Add batcher module. 
This can be used to batch up simple operation commands for later use by another
thread.
 2024-05-22 10:30:31 -07:00
David Goldblatt
86f4851f5d Add clang static analyzer suppression macro. 2024-05-22 10:30:31 -07:00
Amaury Séchet
5afff2e44e Simplify the logic in tcache_gc_small. 2024-05-02 18:52:19 -07:00
Qi Wang
8d8379da44 Fix background_thread creation for the oversize_arena. 
Bypassing background thread creation for the oversize_arena used to be an
optimization since that arena had eager purging.  However #2466 changed the
purging policy for the oversize_arena -- specifically it switched to the default
decay time when background_thread is enabled.

This issue is noticable when the number of arenas is low: whenever the total #
of arenas is <= 4 (which is the default max # of background threads), in which
case the purging will be stalled since no background thread is created for the
oversize_arena.
 2024-05-02 14:45:18 -07:00
Dmitry Ilvokhin
47d69b4eab HPA: Fix infinite purging loop 
One of the condition to start purging is `hpa_hugify_blocked_by_ndirty`
function call returns true. This can happen in cases where we have no
dirty memory for this shard at all. In this case purging loop will be an
infinite loop.

`hpa_hugify_blocked_by_ndirty` was introduced at 0f6c420, but at that
time purging loop has different form and additional `break` was not
required. Purging loop form was re-written at 6630c5989, but additional
exit condition wasn't added there at the time.

Repo code was shared by Patrik Dokoupil at [1], I stripped it down to
minimum to reproduce issue in jemalloc unit tests.

[1]: https://github.com/jemalloc/jemalloc/pull/2533
 2024-04-30 13:46:32 -07:00
Qi Wang
fa451de17f Fix the tcache flush sanity checking around ncached and nstashed. 
When there were many items stashed, it's possible that after flushing stashed,
ncached is already lower than the remain, in which case the flush can simply
return at that point.
 2024-04-12 16:01:55 -07:00
debing.sun
630434bb0a Fixed type error with allocated that caused incorrect printing on 32bit 2024-04-09 14:44:43 -07:00
Shirui Cheng
4b555c11a5 Enable heap profiling on MacOS 2024-04-09 12:57:01 -07:00
Daniel Hodges
11038ff762 Add support for namespace pids in heap profile names 
This change adds support for writing pid namespaces to the filename of a
heap profile. When running with namespaces pids may reused across
namespaces and if mounts are shared where profiles are written there is
not a great way to differentiate profiles between pids.

Signed-off-by: Daniel Hodges <hodges.daniel.scott@gmail.com>
Signed-off-by: Daniel Hodges <hodgesd@fb.com>
 2024-04-09 10:27:52 -07:00
Qi Wang
83b075789b rallocx path: only set errno on the realloc case. 2024-04-05 17:41:43 -07:00
Shirui Cheng
5081c16bb4 Experimental calloc implementation with using memset on larger sizes 2024-04-04 15:31:56 -07:00
Juhyung Park
38056fea64 Set errno to ENOMEM on rallocx() OOM failures 
realloc() and rallocx() shares path, and realloc() should set errno to
ENOMEM upon OOM failures.

Fixes: ee961c2310 ("Merge realloc and rallocx pathways.")
Signed-off-by: Juhyung Park <qkrwngud825@gmail.com>
 2024-04-04 15:13:22 -07:00
Dmitry Ilvokhin
268e8ee880 Include HPA ndirty into page allocator ndirty stat 2024-04-04 12:17:30 -07:00
Dmitry Ilvokhin
b2e59a96e1 Introduce getters for page allocator shard stats 
Access nactive, ndirty and nmuzzy throught getters and not directly.
There are no functional change, but getters are required to propagate
HPA's statistics up to Page Allocator's statitics.
 2024-04-04 12:17:30 -07:00
Amaury Séchet
92aa52c062 Reduce nesting in phn_merge_siblings using an early return. 2024-03-14 13:08:17 -07:00
Amaury Séchet
10d713151d Ensure that the root of a heap is always the best element. 2024-03-14 13:07:45 -07:00
Minsoo Choo
1978e5cdac Update acitons/checkout and actions/upload-artifact to v4 2024-03-12 12:59:15 -07:00
XChy
ed9b00a96b Replace unsigned induction variable with size_t in background_threads_enable 
This patch avoids unnecessary vectorizations in clang and missed recognition of memset in gcc. See also https://godbolt.org/z/aoeMsjr4c.
 2024-03-05 14:54:50 -08:00
Shirui Cheng
373884ab48 print out all malloc_conf settings in stats 2024-02-29 12:12:44 -08:00
Qi Wang
1aba4f41a3 Allow zero sized memalign to pass. 
Instead of failing on assertions.  Previously the same change was made for
posix_memalign and aligned_alloc (#1554).  Make memalign behave the same way
even though it's obsolete.
 2024-02-16 13:06:07 -08:00
Qi Wang
6d181bc1b7 Fix Cirrus CI. 
13.0-RELEASE does not exist anymore.  "The resource
'projects/freebsd-org-cloud-dev/global/images/family/freebsd-13-0' was not
found"
 2024-02-16 13:05:40 -08:00
David Goldblatt
f96010b7fa gitignore: Start ignoring clangd dirs. 2024-01-23 17:02:01 -08:00
Qi Wang
a2c5267409 HPA: Allow frequent reused alloc to bypass the slab_max_alloc limit, as long as 
it's within the huge page size.  These requests do not concern internal
fragmentation with huge pages, since the entire range is expected to be
accessed.
 2024-01-18 14:51:04 -08:00
guangli-dai
b1792c80d2 Add LOGs when entrying and exiting free and sdallocx. 2024-01-11 14:37:20 -08:00
Qi Wang
05160258df When safety_check_fail, also embed hint msg in the abort function name 
because there are cases only logging crash stack traces.
 2024-01-11 14:19:54 -08:00
Qi Wang
3a6296e1ef Disable FreeBSD on Travis CI since it's not working. 
Travis CI currently provides only FreeBSD 12 which is EOL.
 2024-01-04 14:47:52 -08:00
Minsoo Choo
d284aad027 Test on more FreeBSD versions 
Added 14.0-RELEASE
Added 15-CURRENT
Added 14-STABLE
Added 13-STABLE

13.0-RELEASE will be updated when 13.3-RELEASE comes out.
 2024-01-04 12:48:24 -08:00
Connor
dfb3260b97 Fix missing cleanup message for collected profiles. 
```
sub cleanup {
  unlink($main::tmpfile_sym);
  unlink(keys %main::tempnames);

  # We leave any collected profiles in $HOME/jeprof in case the user wants
  # to look at them later.  We print a message informing them of this.
  if ((scalar(@main::profile_files) > 0) &&
      defined($main::collected_profile)) {
    if (scalar(@main::profile_files) == 1) {
      print STDERR "Dynamically gathered profile is in $main::collected_profile\n";
    }
    print STDERR "If you want to investigate this profile further, you can do:\n";
    print STDERR "\n";
    print STDERR "  jeprof \\\n";
    print STDERR "    $main::prog \\\n";
    print STDERR "    $main::collected_profile\n";
    print STDERR "\n";
  }
}
```
On cleanup, it would print out a message for the collected profile.
If there is only one collected profile, it would pop by L691, then `scalar(@main::profile_files)` would be 0, and no message would be printed.
 2024-01-03 14:24:38 -08:00
Honggyu Kim
f6fe6abdcb build: Make autogen.sh accept quoted extra options 
The current autogen.sh script doesn't allow receiving quoted extra
options.

If someone wants to pass extra CFLAGS that is split into multiple
options with a whitespace, then a quote is required.

However, the configure inside autogen.sh fails in this case as follows.

  $ ./autogen.sh CFLAGS="-Dmmap=cxl_mmap -Dmunmap=cxl_munmap"
  autoconf
  ./configure --enable-autogen CFLAGS=-Dmmap=cxl_mmap -Dmunmap=cxl_munmap
  configure: error: unrecognized option: `-Dmunmap=cxl_munmap'
  Try `./configure --help' for more information
  Error 0 in ./configure

It's because the quote discarded unexpectedly when calling configure.

This patch is to fix this problem.

Signed-off-by: Honggyu Kim <honggyu.kim@sk.com>
 2024-01-03 14:20:34 -08:00
guangli-dai
eda05b3994 Fix static analysis warnings. 2024-01-03 14:18:52 -08:00
Shirui Cheng
e4817c8d89 Cleanup cache_bin_info_t* info input args 2023-10-25 10:27:31 -07:00
Qi Wang
3025b021b9 Optimize mutex and bin alignment / locality. 2023-10-23 20:28:26 -07:00
guangli-dai
e2cd27132a Change stack_size assertion back to the more compatabile one. 2023-10-23 20:28:26 -07:00
guangli-dai
756d4df2fd Add util.c into vs project file. 2023-10-18 22:11:13 -07:00
Qi Wang
04d1a87b78 Fix a zero-initializer warning on macOS. 2023-10-18 14:12:43 -07:00
guangli-dai
d88fa71bbd Fix nfill = 0 bug when ncached_max is 1 2023-10-18 14:11:46 -07:00
guangli-dai
6fb3b6a8e4 Refactor the tcache initiailization 
1. Pre-generate all default tcache ncached_max in tcache_boot;
2. Add getters returning default ncached_max and ncached_max_set;
3. Refactor tcache init so that it is always init with a given setting.
 2023-10-18 14:11:46 -07:00
guangli-dai
8a22d10b83 Allow setting default ncached_max for each bin through malloc_conf 2023-10-18 14:11:46 -07:00
guangli-dai
867eedfc58 Fix the bug in dalloc promoted allocations. 
An allocation small enough will be promoted so that it does not
share an extent with others.  However, when dalloc, such allocations
may not be dalloc as a promoted one if nbins < SC_NBINS.  This
commit fixes the bug.
 2023-10-17 14:53:23 -07:00
guangli-dai
630f7de952 Add mallctl to set and get ncached_max of each cache_bin. 
1. `thread_tcache_ncached_max_read_sizeclass` allows users to get the
    ncached_max of the bin with the input sizeclass, passed in through
    oldp (will be upper casted if not an exact bin size is given).
2. `thread_tcache_ncached_max_write` takes in a char array
    representing the settings for bins in the tcache.
 2023-10-17 14:53:23 -07:00
guangli-dai
6b197fdd46 Pre-generate ncached_max for all bins for better tcache_max tuning experience. 2023-10-17 14:53:23 -07:00
Shirui Cheng
36becb1302 metadata usage breakdowns: tracking edata and rtree usages 2023-10-11 11:56:01 -07:00
Qi Wang
005f20aa7f Fix comments about malloc_conf to enable logging. 2023-10-04 11:49:10 -07:00
guangli-dai
7a9e4c9073 Mark jemalloc.h as system header to resolve header conflicts. 2023-10-04 11:41:30 -07:00
Qi Wang
72cfdce718 Allocate tcache stack from base allocator 
When using metadata_thp, allocate tcache bin stacks from base0, which means they
will be placed on huge pages along with other metadata, instead of mixed with
other regular allocations.

In order to do so, modified the base allocator to support limited reuse: freed
tcached stacks (from thread termination) will be returned to base0 and made
available for reuse, but no merging will be attempted since they were bump
allocated out of base blocks. These reused base extents are managed using
separately allocated base edata_t -- they are cached in base->edata_avail when
the extent is all allocated.

One tricky part is, stats updating must be skipped for such reused extents
(since they were accounted for already, and there is no purging for base). This
requires tracking the "if is reused" state explicitly and bypass the stats
updates when allocating from them.
 2023-09-18 12:18:32 -07:00
guangli-dai
a442d9b895 Enable per-tcache tcache_max 
1. add tcache_max and nhbins into tcache_t so that they are per-tcache,
   with one auto tcache per thread, it's also per-thread;
2. add mallctl for each thread to set its own tcache_max (of its auto tcache);
3. store the maximum number of items in each bin instead of using a global storage;
4. add tests for the modifications above.
5. Rename `nhbins` and `tcache_maxclass` to `global_do_not_change_nhbins` and `global_do_not_change_tcache_maxclass`.
 2023-09-06 10:47:14 -07:00
guangli-dai
fbca96c433 Remove unnecessary parameters for cache_bin_postincrement. 2023-09-06 10:47:14 -07:00
Evers Chen
7d9eceaf38 Fix array bounds false warning in gcc 12.3.0 
1.error: array subscript 232 is above array bounds of ‘size_t[232]’ in gcc 12.3.0
2.it also optimizer to the code
 2023-09-05 14:33:55 -07:00
BtbN
ce8ce99a4a Expose jemalloc_prefix via pkg-config 2023-09-05 14:30:21 -07:00
BtbN
ed7e6fe71a Expose private library dependencies via pkg-config 
When linking statically, these need to be included for linking to succeed.
 2023-09-05 14:29:33 -07:00
Qi Wang
7d563a8f81 Update safety check message to remove --enable-debug when it's already on. 2023-09-05 14:15:45 -07:00
Qi Wang
b71da25b8a Fix reading CPU id using rdtscp. 
As pointed out in #2527, the correct register containing CPU id should be ecx
instead edx.
 2023-08-28 11:46:39 -07:00
Qi Wang
87c56c8df8 Fix arenas.i.bins.j.mutex link id in manual. 2023-08-28 11:01:13 -07:00
Kevin Svetlitski
da66aa391f Enable a few additional warnings for CI and fix the issues they uncovered 
- `-Wmissing-prototypes` and `-Wmissing-variable-declarations` are
  helpful for finding dead code and/or things that should be `static`
  but aren't marked as such.
- `-Wunused-macros` is of similar utility, but for identifying dead macros.
- `-Wunreachable-code` and `-Wunreachable-code-aggressive` do exactly
  what they say: flag unreachable code.
 2023-08-11 13:56:23 -07:00
Kevin Svetlitski
d2c9ed3d1e Ensure short read(2)s/write(2)s are properly handled by IO utilities 
`read(2)` and `write(2)` may read or write fewer bytes than were
requested. In order to robustly ensure that all of the requested bytes
are read/written, these edge-cases must be handled.
 2023-08-11 13:36:24 -07:00
guangli-dai
254c4847e8 Print colorful reminder for failed tests. 2023-08-08 15:01:07 -07:00
Kevin Svetlitski
4f50f782fa Use compiler-provided assume builtins when available 
There are several benefits to this:
1. It's cleaner and more reliable to use the builtin to
   inform the compiler of assumptions instead of hoping that the
   optimizer understands your intentions.
2. `clang` will warn you if any of your assumptions would produce
   side-effects (which the compiler will discard). [This blog post](https://fastcompression.blogspot.com/2019/01/compiler-checked-contracts.html)
   by Yann Collet highlights that a hazard of using the
   `unreachable()`-based method of signaling assumptions is that it
   can sometimes result in additional instructions being generated (see
   [this Godbolt link](https://godbolt.org/z/lKNMs3) from the blog post
   for an example).
 2023-08-08 14:59:36 -07:00
Kevin Svetlitski
3aae792b10 Fix infinite purging loop in HPA 
As reported in #2449, under certain circumstances it's possible to get
stuck in an infinite loop attempting to purge from the HPA. We now
handle this by validating the HPA settings at the end of
configuration parsing and either normalizing them or aborting depending on
if `abort_conf` is set.
 2023-08-08 14:36:19 -07:00
Kevin Svetlitski
424dd61d57 Issue a warning upon directly accessing an arena's bins 
An arena's bins should normally be accessed via the `arena_get_bin`
function, which properly takes into account bin-shards. To ensure that
we don't accidentally commit code which incorrectly accesses the bins
directly, we mark the field with `__attribute__((deprecated))` with an
appropriate warning message, and suppress the warning in the few places
where directly accessing the bins is allowed.
 2023-08-04 15:47:05 -07:00
Kevin Svetlitski
120abd703a Add support for the deprecated attribute 
This is useful for enforcing the usage of getter/setter functions to
access fields which are considered private or have unique access constraints.
 2023-08-04 15:47:05 -07:00
Kevin Svetlitski
162ff8365d Update the Ubuntu version used by Travis CI 
Update from Ubuntu Focal Fossa to Ubuntu Jammy Jellyfish. Staying up to
date is always good, but I'm also hoping that perhaps this newer release
contains fixes so that PowerPC VMs don't randomly hang indefinitely
while booting anymore, stalling our CI pipeline.
 2023-08-04 15:32:15 -07:00
Kevin Svetlitski
07a2eab3ed Stop over-reporting memory usage from sampled small allocations 
@interwq noticed [while reviewing an earlier PR](https://github.com/jemalloc/jemalloc/pull/2478#discussion_r1256217261)
that I missed modifying this statistics accounting in line with the rest
of the changes from #2459. This is now fixed, such that sampled small
allocations increment the `.nmalloc`/`.ndalloc` of their effective bin
size instead of over-reporting memory usage by attributing all such
allocations to `SC_LARGE_MINCLASS`.
 2023-08-03 16:12:22 -07:00
Kevin Svetlitski
ea5b7bea31 Add configuration option controlling DSS support 
In many environments, the fallback `sbrk(2)` allocation path is never
used even if the system supports the syscall; if you're at the point
where `mmap(2)` is failing, `sbrk(2)` is unlikely to succeed. Without
changing the default, I've added the ability to disable the usage of DSS
altogether, so that you do not need to pay for the additional code size
and handful of extra runtime branches in such environments.
 2023-08-03 11:52:25 -07:00
Qi Wang
6816b23862 Include the unrecognized malloc conf option in the error message. 
Previously the option causing trouble will not be printed, unless the option
key:value pair format is found.
 2023-08-02 10:44:55 -07:00
Kevin Svetlitski
62648c88e5 Ensured sampled allocations are properly deallocated during arena_reset 
Sampled allocations were not being demoted before being deallocated
during an `arena_reset` operation.
 2023-08-01 11:35:37 -07:00
Kevin Svetlitski
b01d496646 Add an override for the compile-time malloc_conf to jemalloc_internal_overrides.h 2023-07-31 14:53:15 -07:00
Kevin Svetlitski
9ba1e1cb37 Make ctl_arena_clear slightly more efficient 
While this function isn't particularly hot, (accounting for just 0.27% of
time spent inside the allocator on average across the fleet), looking
at the generated assembly and performance profiles does show we're dispatching
to multiple different `memset`s when we could instead be just tail-calling
`memset` once, reducing code size and marginally improving performance.
 2023-07-31 14:44:04 -07:00
Kevin Svetlitski
8ff7e7d6c3 Remove errant #includes in public jemalloc.h header 
In an attempt to make all headers self-contained, I inadvertently added
`#include`s which refer to intermediate, generated headers that aren't
included in the final install. Closes #2489.
 2023-07-25 16:26:50 -07:00
Kevin Svetlitski
3e82f357bb Fix all optimization-inhibiting integer-to-pointer casts 
Following from PR #2481, we replace all integer-to-pointer casts [which
hide pointer provenance information (and thus inhibit
optimizations)](https://clang.llvm.org/extra/clang-tidy/checks/performance/no-int-to-ptr.html)
with equivalent operations that preserve this information. I have
enabled the corresponding clang-tidy check in our static analysis CI so
that we do not get bitten by this again in the future.
 2023-07-24 14:40:42 -07:00
Kevin Svetlitski
4827bb17bd Remove vestigial TCACHE_STATE_* macros 2023-07-24 14:40:42 -07:00
Kevin Svetlitski
1431153695 Define SBRK_INVALID instead of using a magic number 2023-07-24 14:40:42 -07:00
Kevin Svetlitski
7e54dd1ddb Define PROF_TCTX_SENTINEL instead of using magic numbers 
This makes the code more readable on its own, and also sets the stage
for more cleanly handling the pointer provenance lints in a following
commit.
 2023-07-24 14:40:42 -07:00
Kevin Svetlitski
c49c17f128 Suppress verbose frame address warnings 
These warnings are not useful, and make the output of some CI jobs
enormous and difficult to read, so let's suppress them.
 2023-07-24 10:44:17 -07:00
Kevin Svetlitski
cdb2c0e02f Implement C23's free_sized and free_aligned_sized 
[N2699 - Sized Memory Deallocation](https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2699.htm)
introduced two new functions which were incorporated into the C23
standard, `free_sized` and `free_aligned_sized`. Both already have
analogues in Jemalloc, all we are doing here is adding the appropriate
wrappers.
 2023-07-20 15:06:41 -07:00
Kevin Svetlitski
41e0b857be Make headers self-contained by fixing #includes 
Header files are now self-contained, which makes the relationships
between the files clearer, and crucially allows LSP tools like `clangd`
to function correctly in all of our header files. I have verified that
the headers are self-contained (aside from the various Windows shims) by
compiling them as if they were C files – in a follow-up commit I plan to
add this to CI to ensure we don't regress on this front.
 2023-07-14 09:06:32 -07:00
Kevin Svetlitski
856db56f6e Move tsd implementation details into tsd_internals.h 
This is a prerequisite to achieving self-contained headers. Previously,
the various tsd implementation headers (`tsd_generic.h`,
`tsd_tls.h`, `tsd_malloc_thread_cleanup.h`, and `tsd_win.h`) relied
implicitly on being included in `tsd.h` after a variety of dependencies
had been defined above them. This commit instead makes these
dependencies explicit by splitting them out into a separate file,
`tsd_internals.h`, which each of the tsd implementation headers includes
directly.
 2023-07-14 09:06:32 -07:00
Kevin Svetlitski
36ca0c1b7d Stop concealing pointer provenance in phn_link_get 
At least for LLVM, [casting from an integer to a pointer hides provenance information](https://clang.llvm.org/extra/clang-tidy/checks/performance/no-int-to-ptr.html)
and inhibits optimizations. Here's a [Godbolt link](https://godbolt.org/z/5bYPcKoWT)
showing how this change removes a couple unnecessary branches in
`phn_merge_siblings`, which is a very hot function. Canary profiles show
only minor improvements (since most of the cost of this function is in
cache misses), but there's no reason we shouldn't take it.
 2023-07-13 15:12:31 -07:00
Kevin Svetlitski
314c073a38 Print the failed assertion before aborting in test cases 
This makes it faster and easier to debug, so that you don't need to fire
up a debugger just to see which assertion triggered in a failing test.
 2023-07-13 15:07:17 -07:00
Kevin Svetlitski
65d3b5989b Print test error messages in color when stderr is a terminal 
When stderr is a terminal and supports color, print error messages
from tests in red to make them stand out from the surrounding output.
 2023-07-13 13:03:23 -07:00
Kevin Svetlitski
1d9e9c2ed6 Fix inconsistent parameter names between definition/declaration pairs 
For the sake of consistency, function definitions and their
corresponding declarations should use the same names for parameters.
I've enabled this check in static analysis to prevent this issue from
occurring again in the future.
 2023-07-13 12:59:47 -07:00
Kevin Svetlitski
5711dc31d8 Only enable -Wstrict-prototypes in CI to unbreak feature detection 
Adding `-Wstrict-prototypes` to the default `CFLAGS` in PR #2473 had the
non-obvious side-effect of breaking configure-time feature detection,
because the [test-program `autoconf` generates for feature
detection](https://www.gnu.org/software/autoconf/manual/autoconf-2.67/html_node/Generating-Sources.html#:~:text=main%20())
defines `main` as:
```c
int main()
```
Which causes all feature checks to fail, since this triggers
`-Wstrict-prototypes` and the feature checks use `-Werror`.

Resolved by only adding `-Wstrict-prototypes` to
`EXTRA_{CFLAGS,CXXFLAGS}` in CI, since these flags are not used during
feature detection and we control which compiler is used.
 2023-07-06 18:03:13 -07:00
Kevin Svetlitski
589c63b424 Make eligible global variables static and/or const 
For better or worse, Jemalloc has a significant number of global
variables. Making all eligible global variables `static` and/or `const`
at least makes it slightly easier to reason about them, as these
qualifications communicate to the programmer restrictions on their use
without having to `grep` the whole codebase.
 2023-07-06 14:15:12 -07:00
Qi Wang
e249d1a2a1 Remove unreachable code. 2023-07-06 12:06:06 -07:00
Qi Wang
602edd7566 Enabled -Wstrict-prototypes and fixed warnings. 2023-07-06 12:00:02 -07:00
Kevin Svetlitski
ebd7e99f5c Add a test-case for small profiled allocations 
Validate that small allocations (i.e. those with `size <= SC_SMALL_MAXCLASS`)
which are sampled for profiling maintain the expected invariants even
though they now take up less space.
 2023-07-03 16:19:06 -07:00
Kevin Svetlitski
5a858c64d6 Reduce the memory overhead of sampled small allocations 
Previously, small allocations which were sampled as part of heap
profiling were rounded up to `SC_LARGE_MINCLASS`. This additional memory
usage becomes problematic when the page size is increased, as noted in #2358.

Small allocations are now rounded up to the nearest multiple of `PAGE`
instead, reducing the memory overhead by a factor of 4 in the most
extreme cases.
 2023-07-03 16:19:06 -07:00
Kevin Svetlitski
e1338703ef Address compiler warnings in the unit tests 2023-07-03 16:06:35 -07:00
Qi Wang
d131331310 Avoid eager purging on the dedicated oversize arena when using bg thds. 
We have observed new workload patterns (namely ML training type) that cycle
through oversized allocations frequently, because 1) the dataset might be sparse
which is faster to go through, and 2) GPU accelerated.  As a result, the eager
purging from the oversize arena becomes a bottleneck.  To offer an easy
solution, allow normal purging of the oversized extents when background threads
are enabled.
 2023-06-27 11:57:41 -07:00
Kevin Svetlitski
46e464a26b Fix downloading LLVM in GitHub Action 
It turns out LLVM does not include a build for every platform in the
assets for every release, just some of them. As such, I've pinned us to
the latest release version with a corresponding build.
 2023-06-23 14:30:49 -07:00
Kevin Svetlitski
f2e00d2fd3 Remove trailing whitespace 
Additionally, added a GitHub Action to ensure no more trailing
whitespace will creep in again in the future.

I'm excluding Markdown files from this check, since trailing whitespace
is significant there, and also excluding `build-aux/install-sh` because
there is significant trailing whitespace on the line that sets
`defaultIFS`.
 2023-06-23 11:58:18 -07:00
Kevin Svetlitski
05385191d4 Add GitHub action which runs static analysis 
Now that all of the various issues that static analysis uncovered have
been fixed (#2431, #2432, #2433, #2436, #2437, #2446), I've added a
GitHub action which will run static analysis for every PR going forward.
When static analysis detects issues with your code, the GitHub action
provides a link to download its findings in a form tailored for human
consumption.

Take a look at [this demonstration of what it looks like when static
analysis issues are
found](https://github.com/Svetlitski/jemalloc/actions/runs/5010245602)
on my fork for an example (make sure to follow the instructions in the
error message to download and inspect the results).
 2023-06-23 11:55:43 -07:00
Kevin Svetlitski
bb0333e745 Fix remaining static analysis warnings 
Fix or suppress the remaining warnings generated by static analysis.
This is a necessary step before we can incorporate static analysis into
CI. Where possible, I've preferred to modify the code itself instead of
just disabling the warning with a magic comment, so that if we decide to
use different static analysis tools in the future we will be covered
against them raising similar warnings.
 2023-06-23 11:50:29 -07:00
Kevin Svetlitski
210f0d0b2b Fix read of uninitialized data in prof_free 
In #2433, I inadvertently introduced a regression which causes the use of
uninitialized data. Namely, the control path I added for the safety
check in `arena_prof_info_get` neglected to set `prof_info->alloc_tctx`
when the check fails, resulting in `prof_info.alloc_tctx` being
uninitialized [when it is read at the end of
`prof_free`](90176f8a87/include/jemalloc/internal/prof_inlines.h (L272)).
 2023-06-15 18:30:05 -07:00
Kevin Svetlitski
90176f8a87 Fix segfault in rb *_tree_remove 
Static analysis flagged this. It's possible to segfault in the
`*_tree_remove` function generated by `rb_gen`, as `nodep` may
still be `NULL` after the initial for loop. I can confirm from reviewing
the fleetwide coredump data that this was in fact being hit in
production, primarily through `tctx_tree_remove`, and much more rarely
through `gctx_tree_remove`.
 2023-06-07 14:48:41 -07:00
Qi Wang
86eb49b478 Fix the arena selection for oversized allocations. 
Use the per-arena oversize_threshold, instead of the global setting.
 2023-06-06 15:03:13 -07:00
Christos Zoulas
5832ef6589 Use a local variable to set the alignment for this particular allocation 
instead of changing mmap_flags which makes the change permanent. This was
enforcing large alignments for allocations that did not need it causing
fragmentation. Reported by Andreas Gustafsson.
 2023-05-31 14:44:24 -07:00
Kevin Svetlitski
6d4aa33753 Extract the calculation of psset heap assignment for an hpdata into a common function 
This is in preparation for upcoming changes I plan to make to this
logic. Extracting it into a common function will make this easier and
less error-prone, and cleans up the existing code regardless.
 2023-05-31 11:44:04 -07:00
Arne Welzel
c1d3ad4674 Prune je_malloc_default and do_rallocx in jeprof 
Running a simple Ruby and Python execution je_malloc_default and
do_rallocx() in the resulting SVG / text output. Prune these, too.

    MALLOC_CONF='stats_print:true,lg_prof_sample:8,prof:true,prof_final:true' \
        python3 -c '[x for x in range(10000000)]'

    MALLOC_CONF='stats_print:true,lg_prof_sample:8,prof:true,prof_final:true' \
        ruby -e 'puts (0..1000).map{"0"}.join(" ")'
 2023-05-31 11:41:09 -07:00
Arne Welzel
d59e30cbc9 Rename fallback_impl to fallbackNewImpl and prune in jeprof 
The existing fallback_impl name seemed a bit generic and given
it's static probably okay to rename.

Closes #2451
 2023-05-31 11:41:09 -07:00
Qi Wang
d577e9b588 Explicitly cast to unsigned for MALLOCX_ARENA and _TCACHE defines. 2023-05-26 11:52:42 -07:00
Qi Wang
a2259f9fa6 Fix the include path of "jemalloc_internal_overrides.h". 2023-05-25 15:22:02 -07:00
Kevin Svetlitski
9c32689e57 Fix bug where hpa_shard was not being destroyed 
It appears that this was a simple mistake where `hpa_shard_disable` was
being called instead of `hpa_shard_destroy`. At present
`hpa_shard_destroy` is not called anywhere at all outside of test-cases,
which further suggests that this is a bug. @davidtgoldblatt noted
however that since HPA is disabled for manual arenas and we don't
support destruction for auto arenas that presently there is no way to
actually trigger this bug. Nonetheless, it should be fixed.
 2023-05-18 14:17:38 -07:00
Kevin Svetlitski
4e6f1e9208 Allow overriding LG_PAGE 
This is useful for our internal builds where we override the
configuration in the header files generated by autoconf.
 2023-05-17 13:55:38 -07:00
Kevin Svetlitski
3e2ba7a651 Remove dead stores detected by static analysis 
None of these are harmful, and they are almost certainly optimized
away by the compiler. The motivation for fixing them anyway is that
we'd like to enable static analysis as part of CI, and the first step
towards that is resolving the warnings it produces at present.
 2023-05-11 20:27:49 -07:00
Kevin Svetlitski
0288126d9c Fix possible NULL pointer dereference from mallctl("prof.prefix", ...) 
Static analysis flagged this issue. Here is a minimal program which
causes a segfault within Jemalloc:
```
#include <jemalloc/jemalloc.h>

const char *malloc_conf = "prof:true";

int main() {
  mallctl("prof.prefix", NULL, NULL, NULL, 0);
}
```

Fixed by checking if `prefix` is `NULL`.
 2023-05-11 14:47:50 -07:00
Qi Wang
d4a2b8bab1 Add the prof_sys_thread_name feature in the prof_recent unit test. 
This tests the combination of the prof_recent and thread_name features.
Verified that it catches the issue being fixed in this PR.

Also explicitly set thread name in test/unit/prof_recent.  This fixes the name
testing when no default thread name is set (e.g. FreeBSD).
 2023-05-11 09:10:57 -07:00
Qi Wang
94ace05832 Fix the prof thread_name reference in prof_recent dump. 
As pointed out in #2434, the thread_name in prof_tdata_t was changed in #2407.
This also requires an update for the prof_recent dump, specifically the emitter
expects a "char **" which is fixed in this commit.
 2023-05-11 09:10:57 -07:00
Qi Wang
6ea8a7e928 Add config detection for JEMALLOC_HAVE_PTHREAD_SET_NAME_NP. 
and use it on the background thread name setting.
 2023-05-11 09:10:57 -07:00
auxten
5bac384970 If ptr present check if alloc_ctx.edata == NULL 2023-05-10 17:18:22 -07:00
auxten
019cccc293 Make arenas_lookup_ctl triable 2023-05-10 17:18:22 -07:00
Kevin Svetlitski
dc0a184f8d Fix possible NULL pointer dereference in VERIFY_READ 
Static analysis flagged this. Fixed by simply checking `oldlenp`
before dereferencing it.
 2023-05-09 10:57:09 -07:00
Kevin Svetlitski
12311fe6c3 Fix segfault in extent_try_coalesce_impl 
Static analysis flagged this. `extent_record` was passing `NULL` as the
value for `coalesced` to `extent_try_coalesce`, which in turn passes
that argument to `extent_try_coalesce_impl`, where it is written to
without checking if it is `NULL`. I can confirm from reviewing the
fleetwide coredump data that this was in fact being hit in production.
 2023-05-09 10:55:44 -07:00
Kevin Svetlitski
70344a2d38 Make eligible functions static 
The codebase is already very disciplined in making any function which
can be `static`, but there are a few that appear to have slipped through
the cracks.
 2023-05-08 15:00:02 -07:00
Kevin Svetlitski
6841110bd6 Make edata_cmp_summary_comp 30% faster 
`edata_cmp_summary_comp` is one of the very hottest functions, taking up
3% of all time spent inside Jemalloc. I noticed that all existing
callsites rely only on the sign of the value returned by this function,
so I came up with this equivalent branchless implementation which
preserves this property. After empirical measurement, I have found that
this implementation is 30% faster, therefore representing a 1% speed-up
to the allocator as a whole.

At @interwq's suggestion, I've applied the same optimization to
`edata_esnead_comp` in case this function becomes hotter in the future.
 2023-05-04 09:59:17 -07:00
Amaury Séchet
f2b28906e6 Some nits in cache_bin.h 2023-05-01 10:21:17 -07:00
Kevin Svetlitski
fc680128e0 Remove errant assert in arena_extent_alloc_large 
This codepath may generate deferred work when the HPA is enabled.
See also [@davidtgoldblatt's relevant comment on the PR which
introduced this](https://github.com/jemalloc/jemalloc/pull/2107#discussion_r699770967)
which prevented a similarly incorrect `assert` from being added elsewhere.
 2023-05-01 10:00:30 -07:00
Eric Mueller
521970fb2e Check for equality instead of assigning in asserts in hpa_from_pai. 
It appears like a simple typo means we're unconditionally overwriting
some fields in hpa_from_pai when asserts are enabled. From hpa_shard_init,
it looks like these fields have these values anyway, so this shouldn't
cause bugs, but if something is wrong it seems better to have these
asserts in place.

See issue #2412.
 2023-04-17 20:57:48 -07:00
guangli-dai
5f64ad60cd Remove locked flag set in malloc_mutex_trylock 
As a hint flag of the lock, parameter locked should be set only
when the lock is gained or freed.
 2023-04-06 10:57:04 -07:00
Qi Wang
434a68e221 Disallow decay during reentrancy. 
Decay should not be triggered during reentrant calls (may cause lock order
reversal / deadlocks).  Added a delay_trigger flag to the tickers to bypass
decay when rentrancy_level is not zero.
 2023-04-05 10:16:37 -07:00
Qi Wang
e62aa478c7 Rearrange the bools in prof_tdata_t to save some bytes. 
This lowered the sizeof(prof_tdata_t) from 200 to 192 which is a round size
class.  Afterwards the tdata_t size remain unchanged with the last commit, which
effectively inlined the storage of thread names for free.
 2023-04-05 10:03:12 -07:00
Qi Wang
ce0b7ab6c8 Inline the storage for thread name in prof_tdata_t. 
The previous approach managed the thread name in a separate buffer, which causes
races because the thread name update (triggered by new samples) can happen at
the same time as prof dumping (which reads the thread names) -- these two
operations are under separate locks to avoid blocking each other.  Implemented
the thread name storage as part of the tdata struct, which resolves the lifetime
issue and also avoids internal alloc / dalloc during prof_sample.
 2023-04-05 10:03:12 -07:00
Qi Wang
6cab460a45 Add a multithreaded test for prof_sys_thread_name. 
Verified that this catches the issue being fixed in 5fd5583.
 2023-04-05 10:03:12 -07:00
Amaury Séchet
5266152d79 Simplify the logic in ph_remove 2023-03-31 14:35:31 -07:00
Amaury Séchet
be6da4f663 Do not maintain root->prev in ph_remove. 2023-03-31 14:34:57 -07:00
Amaury Séchet
543e2d61e6 Simplify the logic in ph_insert 
Also fixes what looks like an off by one error in the lazy aux list
merge part of the code that previously never touched the last node in
the aux list.
 2023-03-31 14:34:24 -07:00
guangli-dai
31e01a98f1 Fix the rdtscp detection bug and add prefix for the macro. 2023-03-23 11:16:19 -07:00
Qi Wang
8b64be3441 Explicit arena assignment in test_tcache_max. 
Otherwise the associated arena could change with percpu arena enabled.
 2023-03-22 15:16:43 -07:00
Qi Wang
8e7353a19b Explicit arena assignment in test_thread_idle. 
Otherwise the associated arena could change with percpu arena enabled.
 2023-03-22 15:16:43 -07:00
Marvin Schmidt
45249cf5a9 Fix exception specification error for hosts using musl libc 
It turns out that the previous commit did not suffice since the
JEMALLOC_SYS_NOTHROW definition also causes the same exception specification
errors as JEMALLOC_USE_CXX_THROW did:
```
x86_64-pc-linux-musl-cc -std=gnu11 -Werror=unknown-warning-option -Wall -Wextra -Wshorten-64-to-32 -Wsign-compare -Wundef -Wno-format-zero-length -Wpointer-
arith -Wno-missing-braces -Wno-missing-field-initializers -pipe -g3 -fvisibility=hidden -Wimplicit-fallthrough -O3 -funroll-loops -march=native -O2 -pipe -c -march=native -O2 -pipe -D_GNU_SOURCE -D_REENTRANT -Iinclude -Iinclude -o src/background_thread.o src/background_thread.c
In file included from src/jemalloc_cpp.cpp:9:
In file included from include/jemalloc/internal/jemalloc_preamble.h:27:
include/jemalloc/internal/../jemalloc.h:254:32: error: exception specification in declaration does not match previous declaration
    void JEMALLOC_SYS_NOTHROW   *je_malloc(size_t size)
                                 ^
include/jemalloc/internal/../jemalloc.h:75:21: note: expanded from macro 'je_malloc'
                    ^
/usr/x86_64-pc-linux-musl/include/stdlib.h:40:7: note: previous declaration is here
void *malloc (size_t);
      ^
```

On systems using the musl C library we have to omit the exception specification
on malloc function family like it's done for MacOS, FreeBSD and OpenBSD.
 2023-03-16 12:11:40 -07:00
Marvin Schmidt
aba1645f2d configure: Handle *-linux-musl* hosts properly 
This is the same as the `*-*-linux*` case with the two exceptions that
we don't set glibc=1 and don't define JEMALLOC_USE_CXX_THROW
 2023-03-16 12:11:40 -07:00
Qi Wang
d503d72129 Add the missing descriptions in AC_DEFINE 2023-03-14 16:47:00 -07:00
Qi Wang
71bc1a3d91 Avoid assuming the arena id in test when percpu_arena is used. 2023-03-13 10:50:10 -07:00
Amaury Séchet
f743690739 Remove unused mutex from hpa_central 2023-03-10 11:25:47 -08:00
Chris Seymour
4edea8eb8e switch to https 2023-03-09 11:44:02 -08:00
guangli-dai
09e4b38fb1 Use asm volatile during benchmarks. 2023-02-24 11:17:48 -08:00
Fernando Pelliccioni
e8b28908de [MSVC] support for Visual Studio 2019 and 2022 2023-02-21 13:39:25 -08:00
barracuda156
4422f88d17 Makefile.in: link with g++ when cxx enabled 2023-02-21 13:26:58 -08:00
Qi Wang
c7805f1eb5 Add a header in HPA stats for the nonfull slabs. 2023-02-17 13:31:27 -08:00
Qi Wang
b6125120ac Add an explicit name to the dedicated oversize arena. 2023-02-17 13:31:09 -08:00
Qi Wang
97b313c7d4 More conservative setting for /test/unit/background_thread_enable. 
Lower the thread and arena count to avoid resource exhaustion on 32-bit.
 2023-02-16 14:42:21 -08:00
Qi Wang
5fd55837bb Fix thread_name updating for heap profiling. 
The current thread name reading path updates the name every time, which requires
both alloc and dalloc -- and the temporary NULL value in the middle causes races
where the prof dump read path gets NULLed in the middle.

Minimize the changes in this commit to isolate the bugfix testing; will also
refactor the whole thread name paths later.
 2023-02-15 17:49:40 -08:00
Qi Wang
8580c65f81 Implement prof sample hooks "experimental.hooks.prof_sample(_free)". 
The added hooks hooks.prof_sample and hooks.prof_sample_free are intended to
allow advanced users to track additional information, to enable new ways of
profiling on top of the jemalloc heap profile and sample features.

The sample hook is invoked after the allocation and backtracing, and forwards
the both the allocation and backtrace to the user hook; the sample_free hook
happens before the actual deallocation, and forwards only the ptr and usz to the
hook.
 2022-12-07 16:06:49 -08:00
guangli-dai
a74acb57e8 Fix dividing 0 error in stress/cpp/microbench 
Summary:
Per issue #2356, some CXX compilers may optimize away the
new/delete operation in stress/cpp/microbench.cpp.
Thus, this commit (1) bumps the time interval to 1 if it is 0, and
(2) modifies the pointers in the microbench to volatile.
 2022-12-06 10:46:14 -08:00
Guangli Dai
e8f9f13811 Inline free and sdallocx into operator delete 2022-11-21 11:14:05 -08:00
guangli-dai
06374d2a6a Benchmark operator delete 
Added the microbenchmark for operator delete.
Also modified bench.h so that it can be used in C++.
 2022-11-21 11:14:05 -08:00
guangli-dai
14ad8205bf Update the ratio display in benchmark 
In bench.h, specify the ratio as the time consumption ratio and
modify the display of the ratio.
 2022-11-21 11:14:05 -08:00
Qi Wang
481bbfc990 Add a configure option --enable-force-getenv. 
Allows the use of getenv() rather than secure_getenv() to read MALLOC_CONF.
This helps in situations where hosts are under full control, and setting
MALLOC_CONF is needed while also setuid.  Disabled by default.
 2022-11-04 13:37:14 -07:00
Qi Wang
143e9c4a2f Enable fast thread locals for dealloc-only threads. 
Previously if a thread does only allocations, it stays on the slow path /
minimal initialized state forever.  However, dealloc-only is a valid pattern for
dedicated reclamation threads -- this means thread cache is disabled (no batched
flush) for them, which causes high overhead and contention.

Added the condition to fully initialize TSD when a fair amount of dealloc
activities are observed.
 2022-10-25 09:54:38 -07:00
Paul Smith
be65438f20 jemalloc_internal_types.h: Use alloca if __STDC_NO_VLA__ is defined 
No currently-available version of Visual Studio C compiler supports
variable length arrays, even if it defines __STDC_VERSION__ >= C99.
As far as I know Microsoft has no plans to ever support VLAs in MSVC.

The C11 standard requires that the __STDC_NO_VLA__ macro be defined if
the compiler doesn't support VLAs, so fall back to alloca() if so.
 2022-10-14 15:48:32 -07:00
divanorama
1897f185d2 Fix safety_check segfault in double free test 2022-10-03 10:55:10 -07:00
Jordan Rome
b04e7666f2 update PROFILING_INTERNALS.md 
Expand the bad example of summing before unbiasing.
 2022-10-03 10:48:29 -07:00
David Carlier
4c95c953e2 fix build for non linux/BSD platforms. 2022-10-03 10:42:09 -07:00
divanorama
3de0c24859 Disable builtin malloc in tests 
With `--with-jemalloc-prefix=` and without `-fno-builtin` or `-O1` both clang and gcc may optimize out `malloc` calls
whose result is unused. Comparing result to NULL also doesn't necessarily count as being used.

This won't be a problem in most client programs as this only concerns really unused pointers, but in
tests it's important to actually execute allocations.
`-fno-builtin` should disable this optimization for both gcc and clang, and applying it only to tests code shouldn't hopefully be an issue.
Another alternative is to force "use" of result but that'd require more changes and may miss some other optimization-related issues.

This should resolve https://github.com/jemalloc/jemalloc/issues/2091
 2022-10-03 10:39:13 -07:00
Lily Wang
c0c9783ec9 Add vcpkg installation instructions 2022-09-19 15:15:28 -07:00
Guangli Dai
c9ac1f4701 Fix a bug in C++ integration test. 2022-09-16 15:04:59 -07:00
Guangli Dai
ba19d2cb78 Add arena-level name. 
An arena-level name can help identify manual arenas.
 2022-09-16 15:04:59 -07:00
Guangli Dai
a0734fd6ee Making jemalloc max stack depth a runtime option 2022-09-12 13:56:22 -07:00
Abael He
56ddbea270 error: implicit declaration of function 'pthread_create_fptr_init' is invalid in C99 
./autogen.sh \
&& ./configure --prefix=/usr/local  --enable-static   --enable-autogen --enable-xmalloc --with-static-libunwind=/usr/local/lib/libunwind.a --enable-lazy-lock --with-jemalloc-prefix='' \
&& make -j16

...
gcc -std=gnu11 -Werror=unknown-warning-option -Wall -Wextra -Wshorten-64-to-32 -Wsign-compare -Wundef -Wno-format-zero-length -Wpointer-arith -Wno-missing-braces -Wno-missing-field-initializers -pipe -g3 -Wimplicit-fallthrough -O3 -funroll-loops -fPIC -DPIC -c -D_REENTRANT -Iinclude -Iinclude -DJEMALLOC_NO_PRIVATE_NAMESPACE -o src/edata_cache.sym.o src/edata_cache.c
src/background_thread.c:768:6: error: implicit declaration of function 'pthread_create_fptr_init' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
            pthread_create_fptr_init()) {
            ^
src/background_thread.c:768:6: note: did you mean 'pthread_create_wrapper_init'?
src/background_thread.c:34:1: note: 'pthread_create_wrapper_init' declared here
pthread_create_wrapper_init(void) {
^
1 error generated.
make: *** [src/background_thread.sym.o] Error 1
make: *** Waiting for unfinished jobs....
 2022-09-07 11:56:41 -07:00
Guangli Dai
ce29b4c3d9 Refactor the remote / cross thread cache bin stats reading 
Refactored cache_bin.h so that only one function is racy.
 2022-09-06 19:41:19 -07:00
Guangli Dai
42daa1ac44 Add double free detection using slab bitmap for debug build 
Add a sanity check for double free issue in the arena in case that the tcache has been flushed.
 2022-09-06 12:54:21 -07:00
Ivan Zaitsev
36366f3c4c Add double free detection in thread cache for debug build 
Add new runtime option `debug_double_free_max_scan` that specifies the max
number of stack entries to scan in the cache bit when trying to detect the
double free bug (currently debug build only).
 2022-08-04 16:58:22 -07:00
David CARLIER
adc70c0511 update travis 2022-07-19 13:23:08 -07:00
David CARLIER
4e12d21c8d enabled percpu_arena settings on macOs. 
follow-up on #2280
 2022-07-19 13:23:08 -07:00
David Carlier
58478412be OpenBSD build fix. still no cpu affinity. 
- enabling pthread_get/pthread_set_name_np api.
- disabling per thread cpu affinity handling, unsupported on this platform.
 2022-07-19 13:20:11 -07:00
Qi Wang
a1c7d9c046 Add the missing opt.cache_oblivious handling. 2022-07-14 22:41:27 -07:00
Jasmin Parent
41a859ef73 Remove duplicated words in documentation 2022-07-11 15:30:16 -07:00
Azat Khuzhin
cb578bbe01 Fix possible "nmalloc >= ndalloc" assertion 
In arena_stats_merge() first nmalloc was read, and after ndalloc.

However with this order, it is possible for some thread to incement
ndalloc in between, and then nmalloc < ndalloc, and assertion will fail,
like again found by ClickHouse CI [1] (even after #2234).

  [1]: https://github.com/ClickHouse/ClickHouse/issues/31531

Swap the order to avoid possible assertion.

Cc: @interwq
Follow-up for: #2234
 2022-07-11 15:27:51 -07:00
David CARLIER
a9215bf18a CI update FreeBSD version. 2022-06-28 11:48:23 -07:00
Alex Lapenkou
3713932836 Update building for Windows instructions 
Explain how to build for Windows in INSTALL.md and remove another readme.txt in
an obscure location.
 2022-06-14 14:04:48 -07:00
David Carlier
4fc5c4fbac New configure option '--enable-pageid' for Linux 
The option makes jemalloc use prctl with PR_SET_VMA to tag memory mappings with
"jemalloc_pg" or "jemalloc_pg_overcommit". This allows to easily identify
jemalloc's mappings in /proc/<pid>/maps. PR_SET_VMA is only available in Linux
5.17 and above.
 2022-06-09 18:54:08 -07:00
Qi Wang
b950934916 Enable retain by default on macOS. 
High number of mappings result in unusually high fork() cost on macOS.  Retain
fixes the issue, at a small cost of extra VM space reserved.
 2022-06-09 11:37:44 -07:00
David Carlier
df8f7d10af Implement malloc_getcpu for amd64 and arm64 macOS 
This enables per CPU arena on MacOS
 2022-06-08 15:13:55 -07:00
Alex Lapenkou
df7ad8a9b6 Revert "Echo installed files via verbose 'install' command" 
This reverts commit f15d8f3b41. "install -v"
turned out to be not portable and not work on NetBSD.
 2022-06-07 12:28:45 -07:00
barracuda156
70e3735f3a jemalloc: fix PowerPC definitions in quantum.h 2022-05-26 10:51:10 -07:00
Alex Lapenkou
5b1f2cc5d7 Implement pvalloc replacement 
Despite being an obsolete function, pvalloc is still present in GLIBC and should
work correctly when jemalloc replaces libc allocator.
 2022-05-18 17:01:09 -07:00
Qi Wang
cd5aaf308a Improve the failure message upon opt_experimental_infallible_new. 2022-05-17 16:07:40 -07:00
Yuriy Chernyshov
70d4102f48 Fix compiling edata.h with MSVC 
At the time an attempt to compile jemalloc 5.3.0 with MSVC 2019 results in the followin error message:

> jemalloc/include/jemalloc/internal/edata.h:660: error C4576: a parenthesized type followed by an initializer list is a non-standard explicit type conversion syntax
 2022-05-09 14:51:07 -07:00
445 changed files with 42535 additions and 22437 deletions
Download patch file Download diff file Expand all files Collapse all files

10
.appveyor.yml
View file

@ -9,6 +9,7 @@ environment:
- MSYSTEM: MINGW64
CPU: x86_64
CONFIG_FLAGS: --enable-debug
EXTRA_CFLAGS: "-fcommon"
- MSYSTEM: MINGW32
CPU: i686
MSVC: x86
@ -16,21 +17,30 @@ environment:
- MSYSTEM: MINGW32
CPU: i686
CONFIG_FLAGS: --enable-debug
EXTRA_CFLAGS: "-fcommon"
- MSYSTEM: MINGW64
CPU: x86_64
MSVC: amd64
CONFIG_FLAGS:
- MSYSTEM: MINGW64
CPU: x86_64
CONFIG_FLAGS:
EXTRA_CFLAGS: "-fcommon"
- MSYSTEM: MINGW32
CPU: i686
MSVC: x86
CONFIG_FLAGS:
- MSYSTEM: MINGW32
CPU: i686
CONFIG_FLAGS:
EXTRA_CFLAGS: "-fcommon"
install:
- set PATH=c:\msys64\%MSYSTEM%\bin;c:\msys64\usr\bin;%PATH%
- if defined MSVC call "c:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\vcvarsall.bat" %MSVC%
- if defined MSVC pacman --noconfirm -Rsc mingw-w64-%CPU%-gcc gcc
- pacman --noconfirm -Syuu
- pacman --noconfirm -S autoconf
build_script:
- bash -c "autoconf"

46
.cirrus.yml
View file

@ -1,46 +0,0 @@
env:
CIRRUS_CLONE_DEPTH: 1
ARCH: amd64
task:
matrix:
env:
DEBUG_CONFIG: --enable-debug
env:
DEBUG_CONFIG: --disable-debug
matrix:
- env:
PROF_CONFIG: --enable-prof
- env:
PROF_CONFIG: --disable-prof
matrix:
- name: 64-bit
env:
CC:
CXX:
- name: 32-bit
env:
CC: cc -m32
CXX: c++ -m32
matrix:
- env:
UNCOMMON_CONFIG:
- env:
UNCOMMON_CONFIG: --with-lg-page=16 --with-malloc-conf=tcache:false
freebsd_instance:
matrix:
image: freebsd-12-3-release-amd64
install_script:
- sed -i.bak -e 's,pkg+http://pkg.FreeBSD.org/\${ABI}/quarterly,pkg+http://pkg.FreeBSD.org/\${ABI}/latest,' /etc/pkg/FreeBSD.conf
- pkg upgrade -y
- pkg install -y autoconf gmake
script:
- autoconf
# We don't perfectly track freebsd stdlib.h definitions. This is fine when
# we count as a system header, but breaks otherwise, like during these
# tests.
- ./configure --with-jemalloc-prefix=ci_ ${DEBUG_CONFIG} ${PROF_CONFIG} ${UNCOMMON_CONFIG}
- export JFLAG=`sysctl -n kern.smp.cpus`
- gmake -j${JFLAG}
- gmake -j${JFLAG} tests
- gmake check

38
.clang-format
View file

@ -4,10 +4,10 @@
# AccessModifierOffset: -2
AlignAfterOpenBracket: DontAlign
AlignConsecutiveAssignments: false
AlignConsecutiveDeclarations: false
AlignConsecutiveDeclarations: true
AlignEscapedNewlines: Right
AlignOperands: false
AlignTrailingComments: false
AlignTrailingComments: true
AllowAllParametersOfDeclarationOnNextLine: true
AllowShortBlocksOnASingleLine: false
AllowShortCaseLabelsOnASingleLine: false
@ -20,16 +20,16 @@ AlwaysBreakBeforeMultilineStrings: true
BinPackArguments: true
BinPackParameters: true
BraceWrapping:
AfterClass: false
AfterControlStatement: false
AfterEnum: false
AfterFunction: false
AfterNamespace: false
AfterObjCDeclaration: false
AfterStruct: false
AfterUnion: false
BeforeCatch: false
BeforeElse: false
AfterClass: true
AfterControlStatement: true
AfterEnum: true
AfterFunction: true
AfterNamespace: true
AfterObjCDeclaration: true
AfterStruct: true
AfterUnion: true
BeforeCatch: true
BeforeElse: true
IndentBraces: false
# BreakAfterJavaFieldAnnotations: true
BreakBeforeBinaryOperators: NonAssignment
@ -43,7 +43,7 @@ ColumnLimit: 80
# CompactNamespaces: true
# ConstructorInitializerAllOnOneLineOrOnePerLine: true
# ConstructorInitializerIndentWidth: 4
ContinuationIndentWidth: 2
ContinuationIndentWidth: 4
Cpp11BracedListStyle: true
DerivePointerAlignment: false
DisableFormat: false
@ -57,7 +57,7 @@ ForEachMacros: [ ql_foreach, qr_foreach, ]
# IncludeIsMainRegex: ''
IndentCaseLabels: false
IndentPPDirectives: AfterHash
IndentWidth: 4
IndentWidth: 8
IndentWrappedFunctionNames: false
# JavaImportGroups: []
# JavaScriptQuotes: Leave
@ -73,8 +73,8 @@ MaxEmptyLinesToKeep: 1
# ObjCSpaceAfterProperty: false
# ObjCSpaceBeforeProtocolList: false
PenaltyBreakAssignment: 2
PenaltyBreakBeforeFirstCallParameter: 1
PenaltyBreakAssignment: 100
PenaltyBreakBeforeFirstCallParameter: 100
PenaltyBreakComment: 300
PenaltyBreakFirstLessLess: 120
PenaltyBreakString: 1000
@ -96,7 +96,7 @@ PointerAlignment: Right
# - 'cpp'
# BasedOnStyle: llvm
# CanonicalDelimiter: 'cc'
ReflowComments: true
ReflowComments: false
SortIncludes: false
SpaceAfterCStyleCast: false
# SpaceAfterTemplateKeyword: true
@ -107,7 +107,7 @@ SpaceBeforeAssignmentOperators: true
SpaceBeforeParens: ControlStatements
# SpaceBeforeRangeBasedForLoopColon: true
SpaceInEmptyParentheses: false
SpacesBeforeTrailingComments: 2
SpacesBeforeTrailingComments: 1
SpacesInAngles: false
SpacesInCStyleCastParentheses: false
# SpacesInContainerLiterals: false
@ -118,5 +118,5 @@ SpacesInSquareBrackets: false
# used by some of the core jemalloc developers.
# StatementMacros: []
TabWidth: 8
UseTab: Never
UseTab: ForIndentation
...

2
.git-blame-ignore-revs Normal file
View file

@ -0,0 +1,2 @@
554185356bf990155df8d72060c4efe993642baf
34f359e0ca613b5f9d970e9b2152a5203c9df8d6

10
.github/workflows/check_formatting.yaml vendored Normal file
View file

@ -0,0 +1,10 @@
name: 'Check Formatting'
on: [pull_request]
jobs:
check-formatting:
runs-on: ubuntu-latest
steps:
- name: Check out repository
uses: actions/checkout@v4
- name: Check for trailing whitespace
run: scripts/check_trailing_whitespace.sh

66
.github/workflows/freebsd-ci.yml vendored Normal file
View file

@ -0,0 +1,66 @@
# This config file is generated by ./scripts/gen_gh_actions.py.
# Do not edit by hand.
name: FreeBSD CI
on:
push:
branches: [ dev, ci_travis ]
pull_request:
branches: [ dev ]
jobs:
test-freebsd:
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
debug: ['--enable-debug', '--disable-debug']
prof: ['--enable-prof', '--disable-prof']
arch: ['64-bit', '32-bit']
uncommon:
- ''
- '--with-lg-page=16 --with-malloc-conf=tcache:false'
name: FreeBSD (${{ matrix.arch }}, debug=${{ matrix.debug }}, prof=${{ matrix.prof }}${{ matrix.uncommon && ', uncommon' || '' }})
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 1
- name: Test on FreeBSD
uses: vmactions/freebsd-vm@v1
with:
release: '15.0'
usesh: true
prepare: |
pkg install -y autoconf gmake
run: |
# Verify we're running in FreeBSD
echo "==== System Information ===="
uname -a
freebsd-version
echo "============================"
# Set compiler flags for 32-bit if needed
if [ "${{ matrix.arch }}" = "32-bit" ]; then
export CC="cc -m32"
export CXX="c++ -m32"
fi
# Generate configure script
autoconf
# Configure with matrix options
./configure --with-jemalloc-prefix=ci_ ${{ matrix.debug }} ${{ matrix.prof }} ${{ matrix.uncommon }}
# Get CPU count for parallel builds
export JFLAG=$(sysctl -n kern.smp.cpus)
gmake -j${JFLAG}
gmake -j${JFLAG} tests
gmake check

695
.github/workflows/linux-ci.yml vendored Normal file
View file

@ -0,0 +1,695 @@
# This config file is generated by ./scripts/gen_gh_actions.py.
# Do not edit by hand.
name: Linux CI
on:
push:
branches: [ dev, ci_travis ]
pull_request:
branches: [ dev ]
jobs:
test-linux:
runs-on: ubuntu-24.04
strategy:
fail-fast: false
matrix:
include:
- env:
CC: gcc
CXX: g++
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: clang
CXX: clang++
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-prof
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-stats
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-libdl
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-opt-safety-checks
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --with-lg-page=16
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: clang
CXX: clang++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: --enable-prof
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: --disable-stats
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: --disable-libdl
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: --enable-opt-safety-checks
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: --with-lg-page=16
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: "--with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: clang
CXX: clang++
CONFIGURE_FLAGS: "--with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: --enable-prof
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: --disable-stats
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: --disable-libdl
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: --enable-opt-safety-checks
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: --with-lg-page=16
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: "--with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
COMPILER_FLAGS: -m32
CONFIGURE_FLAGS: "--with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --enable-prof"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --disable-stats"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --disable-libdl"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --enable-opt-safety-checks"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --with-lg-page=16"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --disable-stats"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --disable-libdl"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-opt-safety-checks"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --with-lg-page=16"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --disable-libdl"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --enable-opt-safety-checks"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --with-lg-page=16"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-stats --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --enable-opt-safety-checks"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --with-lg-page=16"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--disable-libdl --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-opt-safety-checks --with-lg-page=16"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-opt-safety-checks --enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-opt-safety-checks --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-opt-safety-checks --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-opt-safety-checks --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-opt-safety-checks --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr --with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr --with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr --with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr --with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false,dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false,percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false,background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=dss:primary,percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=dss:primary,background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu,background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --disable-cache-oblivious --enable-stats --enable-log --enable-prof"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-debug --enable-experimental-smallocx --enable-stats --enable-prof"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
steps:
- uses: actions/checkout@v4
- name: Show OS version
run: |
echo "=== System Information ==="
uname -a
echo ""
echo "=== Architecture ==="
uname -m
arch
echo ""
echo "=== OS Release ==="
cat /etc/os-release || true
echo ""
echo "=== CPU Info ==="
lscpu | grep -E "Architecture|CPU op-mode|Byte Order|CPU\(s\):" || true
- name: Install dependencies (32-bit)
if: matrix.env.CROSS_COMPILE_32BIT == 'yes'
run: |
sudo dpkg --add-architecture i386
sudo apt-get update
sudo apt-get install -y gcc-multilib g++-multilib libc6-dev-i386
- name: Build and test
env:
CC: ${{ matrix.env.CC }}
CXX: ${{ matrix.env.CXX }}
COMPILER_FLAGS: ${{ matrix.env.COMPILER_FLAGS }}
CONFIGURE_FLAGS: ${{ matrix.env.CONFIGURE_FLAGS }}
EXTRA_CFLAGS: ${{ matrix.env.EXTRA_CFLAGS }}
run: |
# Verify the script generates the same output
./scripts/gen_gh_actions.py > gh_actions_script.yml
# Run autoconf
autoconf
# Configure with flags
if [ -n "$COMPILER_FLAGS" ]; then
./configure CC="${CC} ${COMPILER_FLAGS}" CXX="${CXX} ${COMPILER_FLAGS}" $CONFIGURE_FLAGS
else
./configure $CONFIGURE_FLAGS
fi
# Build
make -j3
make -j3 tests
# Run tests
make check
test-linux-arm64:
runs-on: ubuntu-24.04-arm
strategy:
fail-fast: false
matrix:
include:
- env:
CC: gcc
CXX: g++
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: clang
CXX: clang++
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-prof
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-stats
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-libdl
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-opt-safety-checks
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --with-lg-page=16
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --with-lg-hugepage=29"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--enable-prof --enable-prof-frameptr"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=dss:primary"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=background_thread:true"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
steps:
- uses: actions/checkout@v4
- name: Show OS version
run: |
echo "=== System Information ==="
uname -a
echo ""
echo "=== Architecture ==="
uname -m
arch
echo ""
echo "=== OS Release ==="
cat /etc/os-release || true
echo ""
echo "=== CPU Info ==="
lscpu | grep -E "Architecture|CPU op-mode|Byte Order|CPU\(s\):" || true
- name: Install dependencies (32-bit)
if: matrix.env.CROSS_COMPILE_32BIT == 'yes'
run: |
sudo dpkg --add-architecture i386
sudo apt-get update
sudo apt-get install -y gcc-multilib g++-multilib libc6-dev-i386
- name: Build and test
env:
CC: ${{ matrix.env.CC }}
CXX: ${{ matrix.env.CXX }}
COMPILER_FLAGS: ${{ matrix.env.COMPILER_FLAGS }}
CONFIGURE_FLAGS: ${{ matrix.env.CONFIGURE_FLAGS }}
EXTRA_CFLAGS: ${{ matrix.env.EXTRA_CFLAGS }}
run: |
# Verify the script generates the same output
./scripts/gen_gh_actions.py > gh_actions_script.yml
# Run autoconf
autoconf
# Configure with flags
if [ -n "$COMPILER_FLAGS" ]; then
./configure CC="${CC} ${COMPILER_FLAGS}" CXX="${CXX} ${COMPILER_FLAGS}" $CONFIGURE_FLAGS
else
./configure $CONFIGURE_FLAGS
fi
# Build
make -j3
make -j3 tests
# Run tests
make check

212
.github/workflows/macos-ci.yml vendored Normal file
View file

@ -0,0 +1,212 @@
# This config file is generated by ./scripts/gen_gh_actions.py.
# Do not edit by hand.
name: macOS CI
on:
push:
branches: [ dev, ci_travis ]
pull_request:
branches: [ dev ]
jobs:
test-macos:
runs-on: macos-15-intel
strategy:
fail-fast: false
matrix:
include:
- env:
CC: gcc
CXX: g++
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-stats
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-libdl
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-opt-safety-checks
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --with-lg-page=16
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
steps:
- uses: actions/checkout@v4
- name: Show OS version
run: |
echo "=== macOS Version ==="
sw_vers
echo ""
echo "=== Architecture ==="
uname -m
arch
echo ""
echo "=== CPU Info ==="
sysctl -n machdep.cpu.brand_string
sysctl -n hw.machine
- name: Install dependencies
run: |
brew install autoconf
- name: Build and test
env:
CC: ${{ matrix.env.CC || 'gcc' }}
CXX: ${{ matrix.env.CXX || 'g++' }}
COMPILER_FLAGS: ${{ matrix.env.COMPILER_FLAGS }}
CONFIGURE_FLAGS: ${{ matrix.env.CONFIGURE_FLAGS }}
EXTRA_CFLAGS: ${{ matrix.env.EXTRA_CFLAGS }}
run: |
# Run autoconf
autoconf
# Configure with flags
if [ -n "$COMPILER_FLAGS" ]; then
./configure CC="${CC} ${COMPILER_FLAGS}" CXX="${CXX} ${COMPILER_FLAGS}" $CONFIGURE_FLAGS
else
./configure $CONFIGURE_FLAGS
fi
# Build
make -j3
make -j3 tests
# Run tests
make check
test-macos-arm64:
runs-on: macos-15
strategy:
fail-fast: false
matrix:
include:
- env:
CC: gcc
CXX: g++
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-stats
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --disable-libdl
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-opt-safety-checks
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --with-lg-page=16
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-lg-page=16 --with-lg-hugepage=29"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=tcache:false"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: "--with-malloc-conf=percpu_arena:percpu"
EXTRA_CFLAGS: "-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
steps:
- uses: actions/checkout@v4
- name: Show OS version
run: |
echo "=== macOS Version ==="
sw_vers
echo ""
echo "=== Architecture ==="
uname -m
arch
echo ""
echo "=== CPU Info ==="
sysctl -n machdep.cpu.brand_string
sysctl -n hw.machine
- name: Install dependencies
run: |
brew install autoconf
- name: Build and test
env:
CC: ${{ matrix.env.CC || 'gcc' }}
CXX: ${{ matrix.env.CXX || 'g++' }}
COMPILER_FLAGS: ${{ matrix.env.COMPILER_FLAGS }}
CONFIGURE_FLAGS: ${{ matrix.env.CONFIGURE_FLAGS }}
EXTRA_CFLAGS: ${{ matrix.env.EXTRA_CFLAGS }}
run: |
# Run autoconf
autoconf
# Configure with flags
if [ -n "$COMPILER_FLAGS" ]; then
./configure CC="${CC} ${COMPILER_FLAGS}" CXX="${CXX} ${COMPILER_FLAGS}" $CONFIGURE_FLAGS
else
./configure $CONFIGURE_FLAGS
fi
# Build
make -j3
make -j3 tests
# Run tests
make check

68
.github/workflows/static_analysis.yaml vendored Normal file
View file

@ -0,0 +1,68 @@
name: 'Static Analysis'
on: [pull_request]
jobs:
static-analysis:
runs-on: ubuntu-latest
steps:
# We build libunwind ourselves because sadly the version
# provided by Ubuntu via apt-get is much too old.
- name: Check out libunwind
uses: actions/checkout@v4
with:
repository: libunwind/libunwind
path: libunwind
ref: 'v1.6.2'
github-server-url: 'https://github.com'
- name: Install libunwind
run: |
cd libunwind
autoreconf -i
./configure --prefix=/usr
make -s -j $(nproc) V=0
sudo make -s install V=0
cd ..
rm -rf libunwind
- name: Check out repository
uses: actions/checkout@v4
# We download LLVM directly from the latest stable release
# on GitHub, because this tends to be much newer than the
# version available via apt-get in Ubuntu.
- name: Download LLVM
uses: dsaltares/fetch-gh-release-asset@master
with:
repo: 'llvm/llvm-project'
version: 'tags/llvmorg-16.0.4'
file: 'clang[+]llvm-.*x86_64-linux-gnu.*'
regex: true
target: 'llvm_assets/'
token: ${{ secrets.GITHUB_TOKEN }}
- name: Install prerequisites
id: install_prerequisites
run: |
tar -C llvm_assets -xaf llvm_assets/*.tar* &
sudo apt-get update
sudo apt-get install -y jq bear python3-pip
pip install codechecker
echo "Extracting LLVM from tar" 1>&2
wait
echo "LLVM_BIN_DIR=$(echo llvm_assets/clang*/bin)" >> "$GITHUB_OUTPUT"
- name: Run static analysis
id: run_static_analysis
run: >
PATH="${{ steps.install_prerequisites.outputs.LLVM_BIN_DIR }}:$PATH"
LDFLAGS='-L/usr/lib'
scripts/run_static_analysis.sh static_analysis_results "$GITHUB_OUTPUT"
- name: Upload static analysis results
if: ${{ steps.run_static_analysis.outputs.HAS_STATIC_ANALYSIS_RESULTS }} == '1'
uses: actions/upload-artifact@v4
with:
name: static_analysis_results
path: static_analysis_results
- name: Check static analysis results
run: |
if [[ "${{ steps.run_static_analysis.outputs.HAS_STATIC_ANALYSIS_RESULTS }}" == '1' ]]
then
echo "::error::Static analysis found issues with your code. Download the 'static_analysis_results' artifact from this workflow and view the 'index.html' file contained within it in a web browser locally for detailed results."
exit 1
fi

155
.github/workflows/windows-ci.yml vendored Normal file
View file

@ -0,0 +1,155 @@
# This config file is generated by ./scripts/gen_gh_actions.py.
# Do not edit by hand.
name: Windows CI
on:
push:
branches: [ dev, ci_travis ]
pull_request:
branches: [ dev ]
jobs:
test-windows:
runs-on: windows-latest
strategy:
fail-fast: false
matrix:
include:
- env:
CC: gcc
CXX: g++
EXTRA_CFLAGS: -fcommon
- env:
CC: gcc
CXX: g++
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: -fcommon
- env:
CC: cl.exe
CXX: cl.exe
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
EXTRA_CFLAGS: -fcommon
- env:
CC: cl.exe
CXX: cl.exe
CONFIGURE_FLAGS: --enable-debug
- env:
CC: gcc
CXX: g++
CROSS_COMPILE_32BIT: yes
CONFIGURE_FLAGS: --enable-debug
EXTRA_CFLAGS: -fcommon
- env:
CC: cl.exe
CXX: cl.exe
CROSS_COMPILE_32BIT: yes
- env:
CC: cl.exe
CXX: cl.exe
CROSS_COMPILE_32BIT: yes
CONFIGURE_FLAGS: --enable-debug
steps:
- uses: actions/checkout@v4
- name: Show OS version
shell: cmd
run: |
echo === Windows Version ===
systeminfo | findstr /B /C:"OS Name" /C:"OS Version"
ver
echo.
echo === Architecture ===
echo PROCESSOR_ARCHITECTURE=%PROCESSOR_ARCHITECTURE%
echo.
- name: Setup MSYS2
uses: msys2/setup-msys2@v2
with:
msystem: ${{ matrix.env.CROSS_COMPILE_32BIT == 'yes' && 'MINGW32' || 'MINGW64' }}
update: true
install: >-
autotools
git
pacboy: >-
make:p
gcc:p
binutils:p
- name: Build and test (MinGW-GCC)
if: matrix.env.CC != 'cl.exe'
shell: msys2 {0}
env:
CC: ${{ matrix.env.CC || 'gcc' }}
CXX: ${{ matrix.env.CXX || 'g++' }}
COMPILER_FLAGS: ${{ matrix.env.COMPILER_FLAGS }}
CONFIGURE_FLAGS: ${{ matrix.env.CONFIGURE_FLAGS }}
EXTRA_CFLAGS: ${{ matrix.env.EXTRA_CFLAGS }}
run: |
# Run autoconf
autoconf
# Configure with flags
if [ -n "$COMPILER_FLAGS" ]; then
./configure CC="${CC} ${COMPILER_FLAGS}" CXX="${CXX} ${COMPILER_FLAGS}" $CONFIGURE_FLAGS
else
./configure $CONFIGURE_FLAGS
fi
# Build (mingw32-make is the "make" command in MSYS2)
mingw32-make -j3
mingw32-make tests
# Run tests
mingw32-make -k check
- name: Setup MSVC environment
if: matrix.env.CC == 'cl.exe'
uses: ilammy/msvc-dev-cmd@v1
with:
arch: ${{ matrix.env.CROSS_COMPILE_32BIT == 'yes' && 'x86' || 'x64' }}
- name: Build and test (MSVC)
if: matrix.env.CC == 'cl.exe'
shell: msys2 {0}
env:
CONFIGURE_FLAGS: ${{ matrix.env.CONFIGURE_FLAGS }}
MSYS2_PATH_TYPE: inherit
run: |
# Export MSVC environment variables for configure
export CC=cl.exe
export CXX=cl.exe
export AR=lib.exe
export NM=dumpbin.exe
export RANLIB=:
# Verify cl.exe is accessible (should be in PATH via inherit)
if ! which cl.exe > /dev/null 2>&1; then
echo "cl.exe not found, trying to locate MSVC..."
# Find and add MSVC bin directory to PATH
MSVC_BIN=$(cmd.exe /c "echo %VCToolsInstallDir%" | tr -d '\\r' | sed 's/\\\\\\\\/\//g' | sed 's/C:/\\/c/g')
if [ -n "$MSVC_BIN" ]; then
export PATH="$PATH:$MSVC_BIN/bin/Hostx64/x64:$MSVC_BIN/bin/Hostx86/x86"
fi
fi
# Run autoconf
autoconf
# Configure with MSVC
./configure CC=cl.exe CXX=cl.exe AR=lib.exe $CONFIGURE_FLAGS
# Build (mingw32-make is the "make" command in MSYS2)
mingw32-make -j3
# Build tests sequentially due to PDB file issues
mingw32-make tests
# Run tests
mingw32-make -k check

8
.gitignore vendored
View file

@ -45,6 +45,13 @@
/src/*.[od]
/src/*.sym
# These are semantically meaningful for clangd and related tooling.
/build/
/.cache/
compile_commands.json
/static_analysis_raw_results
/static_analysis_results
/run_tests.out/
/test/test.sh
@ -66,6 +73,7 @@ test/include/test/jemalloc_test_defs.h
/test/stress/[A-Za-z]*
!/test/stress/[A-Za-z]*.*
!/test/stress/pa/
/test/stress/*.[od]
/test/stress/*.out

168
.travis.yml
View file

@ -6,82 +6,10 @@
# Differences are explained here:
# https://docs.travis-ci.com/user/languages/minimal-and-generic/
language: minimal
dist: focal
dist: jammy
jobs:
include:
- os: windows
arch: amd64
env: CC=gcc CXX=g++ EXTRA_CFLAGS="-fcommon"
- os: windows
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug" EXTRA_CFLAGS="-fcommon"
- os: windows
arch: amd64
env: CC=cl.exe CXX=cl.exe
- os: windows
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes EXTRA_CFLAGS="-fcommon"
- os: windows
arch: amd64
env: CC=cl.exe CXX=cl.exe CONFIGURE_FLAGS="--enable-debug"
- os: windows
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-debug" EXTRA_CFLAGS="-fcommon"
- os: windows
arch: amd64
env: CC=cl.exe CXX=cl.exe CROSS_COMPILE_32BIT=yes
- os: windows
arch: amd64
env: CC=cl.exe CXX=cl.exe CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-debug"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-libunwind"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --enable-prof --enable-prof-libunwind"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-debug"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-libunwind --with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-prof --enable-prof-libunwind"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --enable-prof --enable-prof-libunwind --with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-debug --enable-prof --enable-prof-libunwind"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-debug --with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-prof --enable-prof-libunwind --with-lg-page=16 --with-malloc-conf=tcache:false"
- os: freebsd
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes CONFIGURE_FLAGS="--enable-debug --enable-prof --enable-prof-libunwind --with-lg-page=16 --with-malloc-conf=tcache:false"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -109,6 +37,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -142,6 +73,9 @@ jobs:
- os: linux
arch: amd64
env: CC=clang CXX=clang++ CONFIGURE_FLAGS="--with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- os: linux
arch: amd64
env: CC=clang CXX=clang++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- os: linux
arch: amd64
env: CC=clang CXX=clang++ CONFIGURE_FLAGS="--with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
@ -172,6 +106,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes COMPILER_FLAGS="-m32" CONFIGURE_FLAGS="--with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes COMPILER_FLAGS="-m32" CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes COMPILER_FLAGS="-m32" CONFIGURE_FLAGS="--with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -199,6 +136,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -223,6 +163,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -244,6 +187,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-stats --with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-stats --enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-stats --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -262,6 +208,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-libdl --with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-libdl --enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-libdl --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -277,6 +226,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-opt-safety-checks --with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-opt-safety-checks --enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-opt-safety-checks --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -289,6 +241,9 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-opt-safety-checks --with-malloc-conf=background_thread:true" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16 --enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16 --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -301,6 +256,18 @@ jobs:
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16 --with-malloc-conf=background_thread:true" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr --with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr --with-malloc-conf=dss:primary" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr --with-malloc-conf=percpu_arena:percpu" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr --with-malloc-conf=background_thread:true" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=tcache:false,dss:primary" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
@ -320,62 +287,47 @@ jobs:
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=percpu_arena:percpu,background_thread:true" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=clang CXX=clang++ EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes"
- os: linux
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-stats" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-libdl" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-opt-safety-checks" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16 --with-lg-hugepage=29" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-prof --enable-prof-frameptr" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=dss:primary" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=percpu_arena:percpu" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: linux
arch: ppc64le
arch: arm64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=background_thread:true" EXTRA_CFLAGS="-Werror -Wno-array-bounds"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CROSS_COMPILE_32BIT=yes EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-stats" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--disable-libdl" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-opt-safety-checks" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-lg-page=16" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
- os: osx
arch: amd64
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--with-malloc-conf=tcache:false" EXTRA_CFLAGS="-Werror -Wno-array-bounds -Wno-unknown-warning-option -Wno-ignored-attributes -Wno-deprecated-declarations"
# Development build
- os: linux
env: CC=gcc CXX=g++ CONFIGURE_FLAGS="--enable-debug --disable-cache-oblivious --enable-stats --enable-log --enable-prof" EXTRA_CFLAGS="-Werror -Wno-array-bounds"

148
ChangeLog
View file

@ -4,6 +4,154 @@ brevity. Much more detail can be found in the git revision history:
https://github.com/jemalloc/jemalloc
* 5.3.1 (Apr 13, 2026)
This release includes over 390 commits spanning bug fixes, new features,
performance optimizations, and portability improvements. Multiple percent
of system-level metric improvements were measured in tested production
workloads. The release has gone through large-scale production testing
at Meta.
New features:
- Support pvalloc. (@Lapenkov: 5b1f2cc5)
- Add double free detection for the debug build. (@izaitsevfb:
36366f3c, @guangli-dai: 42daa1ac, @divanorama: 1897f185)
- Add compile-time option `--enable-pageid` to enable memory mapping
annotation. (@devnexen: 4fc5c4fb)
- Add runtime option `prof_bt_max` to control the max stack depth for
profiling. (@guangli-dai: a0734fd6)
- Add compile-time option `--enable-force-getenv` to use `getenv` instead
of `secure_getenv`. (@interwq: 481bbfc9)
- Add compile-time option `--disable-dss` to disable the usage of
`sbrk(2)`. (@Svetlitski: ea5b7bea)
- Add runtime option `tcache_ncached_max` to control the number of items
in each size bin in the thread cache. (@guangli-dai: 8a22d10b)
- Add runtime option `calloc_madvise_threshold` to determine if kernel or
memset is used to zero the allocations for calloc. (@nullptr0-0:
5081c16b)
- Add compile-time option `--disable-user-config` to disable reading the
runtime configurations from `/etc/malloc.conf` or environment variable
`MALLOC_CONF`. (@roblabla: c17bf8b3)
- Add runtime option `disable_large_size_classes` to guard the new usable
size calculation, which minimizes the memory overhead for large
allocations, i.e., >= 4 * PAGE. (@guangli-dai: c067a55c, 8347f104)
- Enable process_madvise usage, add runtime option
`process_madvise_max_batch` to control the max # of regions in each
madvise batch. (@interwq: 22440a02, @spredolac: 4246475b)
- Add mallctl interfaces:
+ `opt.prof_bt_max` (@guangli-dai: a0734fd6)
+ `arena.<i>.name` to set and get arena names. (@guangli-dai: ba19d2cb)
+ `thread.tcache.max` to set and get the `tcache_max` of the current
thread. (@guangli-dai: a442d9b8)
+ `thread.tcache.ncached_max.write` and
`thread.tcache.ncached_max.read_sizeclass` to set and get the
`ncached_max` setup of the current thread. (@guangli-dai: 630f7de9,
6b197fdd)
+ `arenas.hugepage` to return the hugepage size used, also exported to
malloc stats. (@ilvokhin: 90c627ed)
+ `approximate_stats.active` to return an estimate of the current active
bytes, which should not be compared with other stats retrieved.
(@guangli-dai: 0988583d)
Bug fixes:
- Prevent potential deadlocks in decaying during reentrancy. (@interwq:
434a68e2)
- Fix segfault in extent coalescing. (@Svetlitski: 12311fe6)
- Add null pointer detections in mallctl calls. (@Svetlitski: dc0a184f,
0288126d)
- Make mallctl `arenas.lookup` triable without crashing on invalid
pointers. (@auxten: 019cccc2, 5bac3849)
- Demote sampled allocations for proper deallocations during
`arena_reset`. (@Svetlitski: 62648c88)
- Fix jemalloc's `read(2)` and `write(2)`. (@Svetlitski: d2c9ed3d, @lexprfuncall:
9fdc1160)
- Fix the pkg-config metadata file. (@BtbN: ed7e6fe7, ce8ce99a)
- Fix the autogen.sh so that it accepts quoted extra options.
(@honggyukim: f6fe6abd)
- Fix `rallocx()` to set errno to ENOMEM upon OOMing. (@arter97: 38056fea,
@interwq: 83b07578)
- Avoid stack overflow for internal variable array usage. (@nullptr0-0:
47c9bcd4, 48f66cf4, @xinydev: 9169e927)
- Fix background thread initialization race. (@puzpuzpuz: 4d0ffa07)
- Guard os_page_id against a NULL address. (@lexprfuncall: 79cc7dcc)
- Handle tcache init failures gracefully. (@lexprfuncall: a056c20d)
- Fix missing release of acquired neighbor edata in
extent_try_coalesce_impl. (@spredolac: 675ab079)
- Fix memory leak of old curr_reg on san_bump_grow_locked failure.
(@spredolac: 5904a421)
- Fix large alloc nrequests under-counting on cache misses. (@spredolac:
3cc56d32)
Portability improvements:
- Fix the build in C99. (@abaelhe: 56ddbea2)
- Add `pthread_setaffinity_np` detection for non Linux/BSD platforms.
(@devnexen: 4c95c953)
- Make `VARIABLE_ARRAY` compatible with compilers not supporting VLA,
i.e., Visual Studio C compiler in C11 or C17 modes. (@madscientist:
be65438f)
- Fix the build on Linux using musl library. (@marv: aba1645f, 45249cf5)
- Reduce the memory overhead in small allocation sampling for systems
with larger page sizes, e.g., ARM. (@Svetlitski: 5a858c64)
- Add C23's `free_sized` and `free_aligned_sized`. (@Svetlitski:
cdb2c0e0)
- Enable heap profiling on MacOS. (@nullptr0-0: 4b555c11)
- Fix incorrect printing on 32bit. (@sundb: 630434bb)
- Make `JEMALLOC_CXX_THROW` compatible with C++ versions newer than
C++17. (@r-barnes, @guangli-dai: 21bcc0a8)
- Fix mmap tag conflicts on MacOS. (@kdrag0n: c893fcd1)
- Fix monotonic timer assumption for win32. (@burtonli: 8dc97b11)
- Fix VM over-reservation on systems with larger pages, e.g., aarch64.
(@interwq: cd05b19f)
- Remove `unreachable()` macro conditionally to prevent definition
conflicts for C23+. (@appujee: d8486b26, 4b88bddb)
- Fix dlsym failure observed on FreeBSD. (@rhelmot: 86bbabac)
- Change the default page size to 64KB on aarch64 Linux. (@lexprfuncall:
9442300c)
- Update config.guess and config.sub to the latest version.
(@lexprfuncall: c51949ea)
- Determine the page size on Android from NDK header files.
(@lexprfuncall: c51abba1)
- Improve the portability of grep patterns in configure.ac.
(@lexprfuncall: 365747bc)
- Add compile-time option `--with-cxx-stdlib` to specify the C++ standard
library. (@yuxuanchen1997: a10ef3e1)
Optimizations and refactors:
- Enable tcache for deallocation-only threads. (@interwq: 143e9c4a)
- Inline to accelerate operator delete. (@guangli-dai: e8f9f138)
- Optimize pairing heap's performance. (@deadalnix: 5266152d, be6da4f6,
543e2d61, 10d71315, 92aa52c0, @Svetlitski: 36ca0c1b)
- Inline the storage for thread name in the profiling data. (@interwq:
ce0b7ab6, e62aa478)
- Optimize a hot function `edata_cmp_summary_comp` to accelerate it.
(@Svetlitski: 6841110b, @guangli-dai: 0181aaa4)
- Allocate thread cache using the base allocator, which enables thread
cache to use thp when `metadata_thp` is turned on. (@interwq:
72cfdce7)
- Allow oversize arena not to purge immediately when background threads
are enabled, although the default decay time is 0 to be back compatible.
(@interwq: d1313313)
- Optimize thread-local storage implementation on Windows. (@mcfi:
9e123a83, 3a0d9cda)
- Optimize fast path to allow static size class computation. (@interwq:
323ed2e3)
- Redesign tcache GC to regulate the frequency and make it
locality-aware. The new design is default on, guarded by option
`experimental_tcache_gc`. (@nullptr0-0: 0c88be9e, e2c9f3a9,
14d5dc13, @deadalnix: 5afff2e4)
- Reduce the arena switching overhead by avoiding forced purging when
background thread is enabled. (@interwq: a3910b98)
- Improve the reuse efficiency by limiting the maximum coalesced size for
large extents. (@jiebinn: 3c14707b)
- Refactor thread events to allow registration of users' thread events
and remove prof_threshold as the built-in event. (@spredolac: e6864c60,
015b0179, 34ace916)
Documentation:
- Update Windows building instructions. (@Lapenkov: 37139328)
- Add vcpkg installation instructions. (@LilyWangLL: c0c9783e)
- Update profiling internals with an example. (@jordalgo: b04e7666)
* 5.3.0 (May 6, 2022)
This release contains many speed and space optimizations, from micro

115
INSTALL.md
View file

@ -139,6 +139,7 @@ any of the following arguments (not a definitive list) to 'configure':
in the following list that appears to function correctly:
+ libunwind (requires --enable-prof-libunwind)
+ frame pointer (requires --enable-prof-frameptr)
+ libgcc (unless --disable-prof-libgcc)
+ gcc intrinsics (unless --disable-prof-gcc)
@ -147,6 +148,12 @@ any of the following arguments (not a definitive list) to 'configure':
Use the libunwind library (http://www.nongnu.org/libunwind/) for stack
backtracing.
* `--enable-prof-frameptr`
Use the optimized frame pointer unwinder for stack backtracing. Safe
to use in mixed code (with and without frame pointers) - but requires
frame pointers to produce meaningful stacks. Linux only.
* `--disable-prof-libgcc`
Disable the use of libgcc's backtracing functionality.
@ -315,13 +322,13 @@ behavior:
'configure' uses this to find programs.
In some cases it may be necessary to work around configuration results that do
not match reality. For example, Linux 4.5 added support for the MADV_FREE flag
to madvise(2), which can cause problems if building on a host with MADV_FREE
support and deploying to a target without. To work around this, use a cache
file to override the relevant configuration variable defined in configure.ac,
e.g.:
not match reality. For example, Linux 3.4 added support for the MADV_DONTDUMP
flag to madvise(2), which can cause problems if building on a host with
MADV_DONTDUMP support and deploying to a target without. To work around this,
use a cache file to override the relevant configuration variable defined in
configure.ac, e.g.:
echo "je_cv_madv_free=no" > config.cache && ./configure -C
echo "je_cv_madv_dontdump=no" > config.cache && ./configure -C
## Advanced compilation
@ -396,6 +403,102 @@ exclusively):
Use this to search for programs used during configuration and building.
## Building for Windows
There are at least two ways to build jemalloc's libraries for Windows. They
differ in their ease of use and flexibility.
### With MSVC solutions
This is the easy, but less flexible approach. It doesn't let you specify
arguments to the `configure` script.
1. Install Cygwin with at least the following packages:
* autoconf
* autogen
* gawk
* grep
* sed
2. Install Visual Studio 2015 or 2017 with Visual C++
3. Add Cygwin\bin to the PATH environment variable
4. Open "x64 Native Tools Command Prompt for VS 2017"
(note: x86/x64 doesn't matter at this point)
5. Generate header files:
sh -c "CC=cl ./autogen.sh"
6. Now the project can be opened and built in Visual Studio:
msvc\jemalloc_vc2017.sln
### With MSYS
This is a more involved approach that offers the same configuration flexibility
as Linux builds. We use it for our CI workflow to test different jemalloc
configurations on Windows.
1. Install the prerequisites
1. MSYS2
2. Chocolatey
3. Visual Studio if you want to compile with MSVC compiler
2. Run your bash emulation. It could be MSYS2 or Git Bash (this manual was
tested on both)
3. Manually and selectively follow
[before_install.sh](https://github.com/jemalloc/jemalloc/blob/dev/scripts/windows/before_install.sh)
script.
1. Skip the `TRAVIS_OS_NAME` check, `rm -rf C:/tools/msys64` and `choco
uninstall/upgrade` part.
2. If using `msys2` shell, add path to `RefreshEnv.cmd` to `PATH`:
`PATH="$PATH:/c/ProgramData/chocolatey/bin"`
3. Assign `msys_shell_cmd`, `msys2`, `mingw32` and `mingw64` as in the
script.
4. Pick `CROSS_COMPILE_32BIT` , `CC` and `USE_MSVC` values depending on
your needs. For instance, if you'd like to build for x86_64 Windows
with `gcc`, then `CROSS_COMPILE_32BIT="no"`, `CC="gcc"` and
`USE_MSVC=""`. If you'd like to build for x86 Windows with `cl.exe`,
then `CROSS_COMPILE_32BIT="yes"`, `CC="cl.exe"`, `USE_MSVC="x86"`.
For x86_64 builds with `cl.exe`, assign `USE_MSVC="amd64"` and
`CROSS_COMPILE_32BIT="no"`.
5. Replace the path to `vcvarsall.bat` with the path on your system. For
instance, on my Windows PC with Visual Studio 17, the path is
`C:\Program Files (x86)\Microsoft Visual
Studio\2017\BuildTools\VC\Auxiliary\Build\vcvarsall.bat`.
6. Execute the rest of the script. It will install the required
dependencies and assign the variable `build_env`, which is a function
that executes following commands with the correct environment
variables set.
4. Use `$build_env <command>` as you would in a Linux shell:
1. `$build_env autoconf`
2. `$build_env ./configure CC="<desired compiler>" <configuration flags>`
3. `$build_env mingw32-make`
If you're having any issues with the above, ensure the following:
5. When you run `cmd //C RefreshEnv.cmd`, you get an output line starting with
`Refreshing` . If it errors saying `RefreshEnv.cmd` is not found, then you
need to add it to your `PATH` as described above in item 3.2
6. When you run `cmd //C $vcvarsall`, it prints a bunch of environment
variables. Otherwise, check the path to the `vcvarsall.bat` in `$vcvarsall`
script and fix it.
### Building from vcpkg
The jemalloc port in vcpkg is kept up to date by Microsoft team members and
community contributors. The url of vcpkg is: https://github.com/Microsoft/vcpkg
. You can download and install jemalloc using the vcpkg dependency manager:
```shell
git clone https://github.com/Microsoft/vcpkg.git
cd vcpkg
./bootstrap-vcpkg.sh # ./bootstrap-vcpkg.bat for Windows
./vcpkg integrate install
./vcpkg install jemalloc
```
If the version is out of date, please [create an issue or pull
request](https://github.com/Microsoft/vcpkg) on the vcpkg repository.
## Development

88
Makefile.in
View file

@ -123,17 +123,19 @@ C_SRCS := $(srcroot)src/jemalloc.c \
$(srcroot)src/san_bump.c \
$(srcroot)src/hook.c \
$(srcroot)src/hpa.c \
$(srcroot)src/hpa_central.c \
$(srcroot)src/hpa_hooks.c \
$(srcroot)src/hpa_utils.c \
$(srcroot)src/hpdata.c \
$(srcroot)src/inspect.c \
$(srcroot)src/large.c \
$(srcroot)src/log.c \
$(srcroot)src/malloc_io.c \
$(srcroot)src/conf.c \
$(srcroot)src/mutex.c \
$(srcroot)src/nstime.c \
$(srcroot)src/pa.c \
$(srcroot)src/pa_extra.c \
$(srcroot)src/pai.c \
$(srcroot)src/pac.c \
$(srcroot)src/pages.c \
$(srcroot)src/peak_event.c \
@ -141,6 +143,7 @@ C_SRCS := $(srcroot)src/jemalloc.c \
$(srcroot)src/prof_data.c \
$(srcroot)src/prof_log.c \
$(srcroot)src/prof_recent.c \
$(srcroot)src/prof_stack_range.c \
$(srcroot)src/prof_stats.c \
$(srcroot)src/prof_sys.c \
$(srcroot)src/psset.c \
@ -153,8 +156,10 @@ C_SRCS := $(srcroot)src/jemalloc.c \
$(srcroot)src/tcache.c \
$(srcroot)src/test_hooks.c \
$(srcroot)src/thread_event.c \
$(srcroot)src/thread_event_registry.c \
$(srcroot)src/ticker.c \
$(srcroot)src/tsd.c \
$(srcroot)src/util.c \
$(srcroot)src/witness.c
ifeq ($(enable_zone_allocator), 1)
C_SRCS += $(srcroot)src/zone.c
@ -201,14 +206,21 @@ TESTS_UNIT := \
$(srcroot)test/unit/atomic.c \
$(srcroot)test/unit/background_thread.c \
$(srcroot)test/unit/background_thread_enable.c \
$(srcroot)test/unit/background_thread_init.c \
$(srcroot)test/unit/base.c \
$(srcroot)test/unit/batch_alloc.c \
$(srcroot)test/unit/bin.c \
$(srcroot)test/unit/binshard.c \
$(srcroot)test/unit/bitmap.c \
$(srcroot)test/unit/bit_util.c \
$(srcroot)test/unit/buf_writer.c \
$(srcroot)test/unit/cache_bin.c \
$(srcroot)test/unit/ckh.c \
$(srcroot)test/unit/conf.c \
$(srcroot)test/unit/conf_init_0.c \
$(srcroot)test/unit/conf_init_1.c \
$(srcroot)test/unit/conf_init_confirm.c \
$(srcroot)test/unit/conf_parse.c \
$(srcroot)test/unit/counter.c \
$(srcroot)test/unit/decay.c \
$(srcroot)test/unit/div.c \
@ -224,6 +236,10 @@ TESTS_UNIT := \
$(srcroot)test/unit/hash.c \
$(srcroot)test/unit/hook.c \
$(srcroot)test/unit/hpa.c \
$(srcroot)test/unit/hpa_sec_integration.c \
$(srcroot)test/unit/hpa_thp_always.c \
$(srcroot)test/unit/hpa_vectorized_madvise.c \
$(srcroot)test/unit/hpa_vectorized_madvise_large_batch.c \
$(srcroot)test/unit/hpa_background_thread.c \
$(srcroot)test/unit/hpdata.c \
$(srcroot)test/unit/huge.c \
@ -231,6 +247,8 @@ TESTS_UNIT := \
$(srcroot)test/unit/junk.c \
$(srcroot)test/unit/junk_alloc.c \
$(srcroot)test/unit/junk_free.c \
$(srcroot)test/unit/json_stats.c \
$(srcroot)test/unit/large_ralloc.c \
$(srcroot)test/unit/log.c \
$(srcroot)test/unit/mallctl.c \
$(srcroot)test/unit/malloc_conf_2.c \
@ -240,6 +258,7 @@ TESTS_UNIT := \
$(srcroot)test/unit/mq.c \
$(srcroot)test/unit/mtx.c \
$(srcroot)test/unit/nstime.c \
$(srcroot)test/unit/ncached_max.c \
$(srcroot)test/unit/oversize_threshold.c \
$(srcroot)test/unit/pa.c \
$(srcroot)test/unit/pack.c \
@ -256,6 +275,7 @@ TESTS_UNIT := \
$(srcroot)test/unit/prof_mdump.c \
$(srcroot)test/unit/prof_recent.c \
$(srcroot)test/unit/prof_reset.c \
$(srcroot)test/unit/prof_small.c \
$(srcroot)test/unit/prof_stats.c \
$(srcroot)test/unit/prof_tctx.c \
$(srcroot)test/unit/prof_thread_name.c \
@ -279,6 +299,7 @@ TESTS_UNIT := \
$(srcroot)test/unit/stats.c \
$(srcroot)test/unit/stats_print.c \
$(srcroot)test/unit/sz.c \
$(srcroot)test/unit/tcache_init.c \
$(srcroot)test/unit/tcache_max.c \
$(srcroot)test/unit/test_hooks.c \
$(srcroot)test/unit/thread_event.c \
@ -332,10 +353,15 @@ TESTS_STRESS := $(srcroot)test/stress/batch_alloc.c \
$(srcroot)test/stress/large_microbench.c \
$(srcroot)test/stress/mallctl.c \
$(srcroot)test/stress/microbench.c
ifeq (@enable_cxx@, 1)
TESTS_STRESS_CPP := $(srcroot)test/stress/cpp/microbench.cpp
else
TESTS_STRESS_CPP :=
endif
TESTS := $(TESTS_UNIT) $(TESTS_INTEGRATION) $(TESTS_INTEGRATION_CPP) \
$(TESTS_ANALYZE) $(TESTS_STRESS)
$(TESTS_ANALYZE) $(TESTS_STRESS) $(TESTS_STRESS_CPP)
PRIVATE_NAMESPACE_HDRS := $(objroot)include/jemalloc/internal/private_namespace.h $(objroot)include/jemalloc/internal/private_namespace_jet.h
PRIVATE_NAMESPACE_GEN_HDRS := $(PRIVATE_NAMESPACE_HDRS:%.h=%.gen.h)
@ -362,9 +388,10 @@ TESTS_INTEGRATION_OBJS := $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%.$(O))
TESTS_INTEGRATION_CPP_OBJS := $(TESTS_INTEGRATION_CPP:$(srcroot)%.cpp=$(objroot)%.$(O))
TESTS_ANALYZE_OBJS := $(TESTS_ANALYZE:$(srcroot)%.c=$(objroot)%.$(O))
TESTS_STRESS_OBJS := $(TESTS_STRESS:$(srcroot)%.c=$(objroot)%.$(O))
TESTS_STRESS_CPP_OBJS := $(TESTS_STRESS_CPP:$(srcroot)%.cpp=$(objroot)%.$(O))
TESTS_OBJS := $(TESTS_UNIT_OBJS) $(TESTS_INTEGRATION_OBJS) $(TESTS_ANALYZE_OBJS) \
$(TESTS_STRESS_OBJS)
TESTS_CPP_OBJS := $(TESTS_INTEGRATION_CPP_OBJS)
TESTS_CPP_OBJS := $(TESTS_INTEGRATION_CPP_OBJS) $(TESTS_STRESS_CPP_OBJS)
.PHONY: all dist build_doc_html build_doc_man build_doc
.PHONY: install_bin install_include install_lib
@ -454,10 +481,13 @@ $(TESTS_INTEGRATION_OBJS): CPPFLAGS += -DJEMALLOC_INTEGRATION_TEST
$(TESTS_INTEGRATION_CPP_OBJS): CPPFLAGS += -DJEMALLOC_INTEGRATION_CPP_TEST
$(TESTS_ANALYZE_OBJS): CPPFLAGS += -DJEMALLOC_ANALYZE_TEST
$(TESTS_STRESS_OBJS): CPPFLAGS += -DJEMALLOC_STRESS_TEST
$(TESTS_STRESS_CPP_OBJS): CPPFLAGS += -DJEMALLOC_STRESS_CPP_TEST
$(TESTS_OBJS): $(objroot)test/%.$(O): $(srcroot)test/%.c
$(TESTS_CPP_OBJS): $(objroot)test/%.$(O): $(srcroot)test/%.cpp
$(TESTS_OBJS): CPPFLAGS += -I$(srcroot)test/include -I$(objroot)test/include
$(TESTS_CPP_OBJS): CPPFLAGS += -I$(srcroot)test/include -I$(objroot)test/include
$(TESTS_OBJS): CFLAGS += -fno-builtin
$(TESTS_CPP_OBJS): CPPFLAGS += -fno-builtin
ifneq ($(IMPORTLIB),$(SO))
$(CPP_OBJS) $(C_SYM_OBJS) $(C_OBJS) $(C_JET_SYM_OBJS) $(C_JET_OBJS): CPPFLAGS += -DDLLEXPORT
endif
@ -472,7 +502,7 @@ $(TESTS_OBJS) $(TESTS_CPP_OBJS): $(objroot)test/include/test/jemalloc_test.h
endif
$(C_OBJS) $(CPP_OBJS) $(C_PIC_OBJS) $(CPP_PIC_OBJS) $(C_TESTLIB_INTEGRATION_OBJS) $(C_UTIL_INTEGRATION_OBJS) $(TESTS_INTEGRATION_OBJS) $(TESTS_INTEGRATION_CPP_OBJS): $(objroot)include/jemalloc/internal/private_namespace.h
$(C_JET_OBJS) $(C_TESTLIB_UNIT_OBJS) $(C_TESTLIB_ANALYZE_OBJS) $(C_TESTLIB_STRESS_OBJS) $(TESTS_UNIT_OBJS) $(TESTS_ANALYZE_OBJS) $(TESTS_STRESS_OBJS): $(objroot)include/jemalloc/internal/private_namespace_jet.h
$(C_JET_OBJS) $(C_TESTLIB_UNIT_OBJS) $(C_TESTLIB_ANALYZE_OBJS) $(C_TESTLIB_STRESS_OBJS) $(TESTS_UNIT_OBJS) $(TESTS_ANALYZE_OBJS) $(TESTS_STRESS_OBJS) $(TESTS_STRESS_CPP_OBJS): $(objroot)include/jemalloc/internal/private_namespace_jet.h
$(C_SYM_OBJS) $(C_OBJS) $(C_PIC_OBJS) $(C_JET_SYM_OBJS) $(C_JET_OBJS) $(C_TESTLIB_OBJS) $(TESTS_OBJS): %.$(O):
@mkdir -p $(@D)
@ -513,7 +543,11 @@ endif
$(objroot)lib/$(LIBJEMALLOC).$(SOREV) : $(if $(PIC_CFLAGS),$(C_PIC_OBJS),$(C_OBJS)) $(if $(PIC_CFLAGS),$(CPP_PIC_OBJS),$(CPP_OBJS))
@mkdir -p $(@D)
ifeq (@enable_cxx@, 1)
$(CXX) $(DSO_LDFLAGS) $(call RPATH,$(RPATH_EXTRA)) $(LDTARGET) $+ $(LDFLAGS) $(LIBS) $(EXTRA_LDFLAGS)
else
$(CC) $(DSO_LDFLAGS) $(call RPATH,$(RPATH_EXTRA)) $(LDTARGET) $+ $(LDFLAGS) $(LIBS) $(EXTRA_LDFLAGS)
endif
$(objroot)lib/$(LIBJEMALLOC)_pic.$(A) : $(C_PIC_OBJS) $(CPP_PIC_OBJS)
$(objroot)lib/$(LIBJEMALLOC).$(A) : $(C_OBJS) $(CPP_OBJS)
@ -543,6 +577,28 @@ $(objroot)test/stress/%$(EXE): $(objroot)test/stress/%.$(O) $(C_JET_OBJS) $(C_TE
@mkdir -p $(@D)
$(CC) $(TEST_LD_MODE) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(LIBS)) $(LM) $(EXTRA_LDFLAGS)
$(objroot)test/stress/pa/pa_data_preprocessor$(EXE): $(objroot)test/stress/pa/pa_data_preprocessor.$(O)
@mkdir -p $(@D)
$(CXX) $(LDTARGET) $(filter %.$(O),$^) $(LDFLAGS) $(filter-out -lm,$(LIBS)) $(LM) $(EXTRA_LDFLAGS)
$(objroot)test/stress/pa/pa_microbench$(EXE): $(objroot)test/stress/pa/pa_microbench.$(O) $(C_JET_OBJS) $(C_TESTLIB_STRESS_OBJS)
@mkdir -p $(@D)
$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(LDFLAGS) $(filter-out -lm,$(LIBS)) $(LM) $(EXTRA_LDFLAGS)
$(objroot)test/stress/pa/%.$(O): $(srcroot)test/stress/pa/%.c
@mkdir -p $(@D)
$(CC) $(CFLAGS) -c $(CPPFLAGS) -DJEMALLOC_STRESS_TEST -I$(srcroot)test/include -I$(objroot)test/include $(CTARGET) $<
ifdef CC_MM
@$(CC) -MM $(CPPFLAGS) -DJEMALLOC_STRESS_TEST -I$(srcroot)test/include -I$(objroot)test/include -MT $@ -o $(@:%.$(O)=%.d) $<
endif
$(objroot)test/stress/pa/%.$(O): $(srcroot)test/stress/pa/%.cpp
@mkdir -p $(@D)
$(CXX) $(CXXFLAGS) -c $(CPPFLAGS) -I$(srcroot)test/include -I$(objroot)test/include $(CTARGET) $<
ifdef CC_MM
@$(CXX) -MM $(CPPFLAGS) -I$(srcroot)test/include -I$(objroot)test/include -MT $@ -o $(@:%.$(O)=%.d) $<
endif
build_lib_shared: $(DSOS)
build_lib_static: $(STATIC_LIBS)
ifeq ($(enable_shared), 1)
@ -555,18 +611,20 @@ endif
install_bin:
$(INSTALL) -d $(BINDIR)
@for b in $(BINS); do \
$(INSTALL) -v -m 755 $$b $(BINDIR); \
echo "$(INSTALL) -m 755 $$b $(BINDIR)"; \
$(INSTALL) -m 755 $$b $(BINDIR); \
done
install_include:
$(INSTALL) -d $(INCLUDEDIR)/jemalloc
@for h in $(C_HDRS); do \
$(INSTALL) -v -m 644 $$h $(INCLUDEDIR)/jemalloc; \
echo "$(INSTALL) -m 644 $$h $(INCLUDEDIR)/jemalloc"; \
$(INSTALL) -m 644 $$h $(INCLUDEDIR)/jemalloc; \
done
install_lib_shared: $(DSOS)
$(INSTALL) -d $(LIBDIR)
$(INSTALL) -v -m 755 $(objroot)lib/$(LIBJEMALLOC).$(SOREV) $(LIBDIR)
$(INSTALL) -m 755 $(objroot)lib/$(LIBJEMALLOC).$(SOREV) $(LIBDIR)
ifneq ($(SOREV),$(SO))
ln -sf $(LIBJEMALLOC).$(SOREV) $(LIBDIR)/$(LIBJEMALLOC).$(SO)
endif
@ -574,13 +632,15 @@ endif
install_lib_static: $(STATIC_LIBS)
$(INSTALL) -d $(LIBDIR)
@for l in $(STATIC_LIBS); do \
$(INSTALL) -v -m 755 $$l $(LIBDIR); \
echo "$(INSTALL) -m 755 $$l $(LIBDIR)"; \
$(INSTALL) -m 755 $$l $(LIBDIR); \
done
install_lib_pc: $(PC)
$(INSTALL) -d $(LIBDIR)/pkgconfig
@for l in $(PC); do \
$(INSTALL) -v -m 644 $$l $(LIBDIR)/pkgconfig; \
echo "$(INSTALL) -m 644 $$l $(LIBDIR)/pkgconfig"; \
$(INSTALL) -m 644 $$l $(LIBDIR)/pkgconfig; \
done
ifeq ($(enable_shared), 1)
@ -594,13 +654,15 @@ install_lib: install_lib_pc
install_doc_html: build_doc_html
$(INSTALL) -d $(DATADIR)/doc/jemalloc$(install_suffix)
@for d in $(DOCS_HTML); do \
$(INSTALL) -v -m 644 $$d $(DATADIR)/doc/jemalloc$(install_suffix); \
echo "$(INSTALL) -m 644 $$d $(DATADIR)/doc/jemalloc$(install_suffix)"; \
$(INSTALL) -m 644 $$d $(DATADIR)/doc/jemalloc$(install_suffix); \
done
install_doc_man: build_doc_man
$(INSTALL) -d $(MANDIR)/man3
@for d in $(DOCS_MAN3); do \
$(INSTALL) -v -m 644 $$d $(MANDIR)/man3; \
echo "$(INSTALL) -m 644 $$d $(MANDIR)/man3"; \
$(INSTALL) -m 644 $$d $(MANDIR)/man3; \
done
install_doc: install_doc_html install_doc_man
@ -656,7 +718,8 @@ endif
tests_unit: $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%$(EXE))
tests_integration: $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%$(EXE)) $(TESTS_INTEGRATION_CPP:$(srcroot)%.cpp=$(objroot)%$(EXE))
tests_analyze: $(TESTS_ANALYZE:$(srcroot)%.c=$(objroot)%$(EXE))
tests_stress: $(TESTS_STRESS:$(srcroot)%.c=$(objroot)%$(EXE))
tests_stress: $(TESTS_STRESS:$(srcroot)%.c=$(objroot)%$(EXE)) $(TESTS_STRESS_CPP:$(srcroot)%.cpp=$(objroot)%$(EXE))
tests_pa: $(objroot)test/stress/pa/pa_data_preprocessor$(EXE) $(objroot)test/stress/pa/pa_microbench$(EXE)
tests: tests_unit tests_integration tests_analyze tests_stress
check_unit_dir:
@ -689,6 +752,7 @@ else
endif
stress: tests_stress stress_dir
$(SHELL) $(objroot)test/test.sh $(TESTS_STRESS:$(srcroot)%.c=$(objroot)%)
$(SHELL) $(objroot)test/test.sh $(TESTS_STRESS_CPP:$(srcroot)%.cpp=$(objroot)%)
check: check_unit check_integration check_integration_decay check_integration_prof
clean:

2
README
View file

@ -17,4 +17,4 @@ jemalloc.
The ChangeLog file contains a brief summary of changes for each release.
URL: http://jemalloc.net/
URL: https://jemalloc.net/

26
TUNING.md
View file

@ -11,9 +11,9 @@ by a few percent, or make favorable trade-offs.
## Notable runtime options for performance tuning
Runtime options can be set via
[malloc_conf](http://jemalloc.net/jemalloc.3.html#tuning).
[malloc_conf](https://jemalloc.net/jemalloc.3.html#tuning).
* [background_thread](http://jemalloc.net/jemalloc.3.html#background_thread)
* [background_thread](https://jemalloc.net/jemalloc.3.html#background_thread)
Enabling jemalloc background threads generally improves the tail latency for
application threads, since unused memory purging is shifted to the dedicated
@ -23,7 +23,7 @@ Runtime options can be set via
Suggested: `background_thread:true` when jemalloc managed threads can be
allowed.
* [metadata_thp](http://jemalloc.net/jemalloc.3.html#opt.metadata_thp)
* [metadata_thp](https://jemalloc.net/jemalloc.3.html#opt.metadata_thp)
Allowing jemalloc to utilize transparent huge pages for its internal
metadata usually reduces TLB misses significantly, especially for programs
@ -35,8 +35,8 @@ Runtime options can be set via
`metadata_thp:always`, which is expected to improve CPU utilization at a
small memory cost.
* [dirty_decay_ms](http://jemalloc.net/jemalloc.3.html#opt.dirty_decay_ms) and
[muzzy_decay_ms](http://jemalloc.net/jemalloc.3.html#opt.muzzy_decay_ms)
* [dirty_decay_ms](https://jemalloc.net/jemalloc.3.html#opt.dirty_decay_ms) and
[muzzy_decay_ms](https://jemalloc.net/jemalloc.3.html#opt.muzzy_decay_ms)
Decay time determines how fast jemalloc returns unused pages back to the
operating system, and therefore provides a fairly straightforward trade-off
@ -46,7 +46,7 @@ Runtime options can be set via
Suggested: tune the values based on the desired trade-offs.
* [narenas](http://jemalloc.net/jemalloc.3.html#opt.narenas)
* [narenas](https://jemalloc.net/jemalloc.3.html#opt.narenas)
By default jemalloc uses multiple arenas to reduce internal lock contention.
However high arena count may also increase overall memory fragmentation,
@ -57,7 +57,7 @@ Runtime options can be set via
Suggested: if low parallelism is expected, try lower arena count while
monitoring CPU and memory usage.
* [percpu_arena](http://jemalloc.net/jemalloc.3.html#opt.percpu_arena)
* [percpu_arena](https://jemalloc.net/jemalloc.3.html#opt.percpu_arena)
Enable dynamic thread to arena association based on running CPU. This has
the potential to improve locality, e.g. when thread to CPU affinity is
@ -100,28 +100,28 @@ aborts immediately on illegal options.
In addition to the runtime options, there are a number of programmatic ways to
improve application performance with jemalloc.
* [Explicit arenas](http://jemalloc.net/jemalloc.3.html#arenas.create)
* [Explicit arenas](https://jemalloc.net/jemalloc.3.html#arenas.create)
Manually created arenas can help performance in various ways, e.g. by
managing locality and contention for specific usages. For example,
applications can explicitly allocate frequently accessed objects from a
dedicated arena with
[mallocx()](http://jemalloc.net/jemalloc.3.html#MALLOCX_ARENA) to improve
[mallocx()](https://jemalloc.net/jemalloc.3.html#MALLOCX_ARENA) to improve
locality. In addition, explicit arenas often benefit from individually
tuned options, e.g. relaxed [decay
time](http://jemalloc.net/jemalloc.3.html#arena.i.dirty_decay_ms) if
time](https://jemalloc.net/jemalloc.3.html#arena.i.dirty_decay_ms) if
frequent reuse is expected.
* [Extent hooks](http://jemalloc.net/jemalloc.3.html#arena.i.extent_hooks)
* [Extent hooks](https://jemalloc.net/jemalloc.3.html#arena.i.extent_hooks)
Extent hooks allow customization for managing underlying memory. One use
case for performance purpose is to utilize huge pages -- for example,
[HHVM](https://github.com/facebook/hhvm/blob/master/hphp/util/alloc.cpp)
[HHVM](httpss://github.com/facebook/hhvm/blob/master/hphp/util/alloc.cpp)
uses explicit arenas with customized extent hooks to manage 1GB huge pages
for frequently accessed data, which reduces TLB misses significantly.
* [Explicit thread-to-arena
binding](http://jemalloc.net/jemalloc.3.html#thread.arena)
binding](https://jemalloc.net/jemalloc.3.html#thread.arena)
It is common for some threads in an application to have different memory
access / allocation patterns. Threads with heavy workloads often benefit

4
autogen.sh
View file

@ -9,8 +9,8 @@ for i in autoconf; do
fi
done
echo "./configure --enable-autogen $@"
./configure --enable-autogen $@
echo "./configure --enable-autogen \"$@\""
./configure --enable-autogen "$@"
if [ $? -ne 0 ]; then
echo "Error $? in ./configure"
exit 1

113
bin/jeprof.in
View file

@ -88,6 +88,7 @@ my %obj_tool_map = (
#"nm_pdb" => "nm-pdb", # for reading windows (PDB-format) executables
#"addr2line_pdb" => "addr2line-pdb", # ditto
#"otool" => "otool", # equivalent of objdump on OS X
#"dyld_info" => "dyld_info", # equivalent of otool on OS X for shared cache
);
# NOTE: these are lists, so you can put in commandline flags if you want.
my @DOT = ("dot"); # leave non-absolute, since it may be in /usr/local
@ -688,15 +689,15 @@ sub Main() {
my $symbol_map = {};
# Read one profile, pick the last item on the list
my $data = ReadProfile($main::prog, pop(@main::profile_files));
my $data = ReadProfile($main::prog, $main::profile_files[0]);
my $profile = $data->{profile};
my $pcs = $data->{pcs};
my $libs = $data->{libs}; # Info about main program and shared libraries
$symbol_map = MergeSymbols($symbol_map, $data->{symbols});
# Add additional profiles, if available.
if (scalar(@main::profile_files) > 0) {
foreach my $pname (@main::profile_files) {
if (scalar(@main::profile_files) > 1) {
foreach my $pname (@main::profile_files[1..$#main::profile_files]) {
my $data2 = ReadProfile($main::prog, $pname);
$profile = AddProfile($profile, $data2->{profile});
$pcs = AddPcs($pcs, $data2->{pcs});
@ -2955,8 +2956,25 @@ sub RemoveUninterestingFrames {
foreach my $name ('@JEMALLOC_PREFIX@calloc',
'cfree',
'@JEMALLOC_PREFIX@malloc',
'je_malloc_default',
'newImpl',
'void* newImpl',
'fallbackNewImpl',
'void* fallbackNewImpl',
'fallback_impl',
'void* fallback_impl',
'imalloc',
'int imalloc',
'imalloc_body',
'int imalloc_body',
'prof_alloc_prep',
'prof_tctx_t *prof_alloc_prep',
'prof_backtrace_impl',
'void prof_backtrace_impl',
'je_prof_backtrace',
'void je_prof_backtrace',
'je_prof_tctx_create',
'prof_tctx_t* prof_tctx_create',
'@JEMALLOC_PREFIX@free',
'@JEMALLOC_PREFIX@memalign',
'@JEMALLOC_PREFIX@posix_memalign',
@ -2965,7 +2983,12 @@ sub RemoveUninterestingFrames {
'@JEMALLOC_PREFIX@valloc',
'@JEMALLOC_PREFIX@realloc',
'@JEMALLOC_PREFIX@mallocx',
'irallocx_prof',
'void *irallocx_prof',
'@JEMALLOC_PREFIX@rallocx',
'do_rallocx',
'ixallocx_prof',
'size_t ixallocx_prof',
'@JEMALLOC_PREFIX@xallocx',
'@JEMALLOC_PREFIX@dallocx',
'@JEMALLOC_PREFIX@sdallocx',
@ -3078,6 +3101,8 @@ sub RemoveUninterestingFrames {
foreach my $a (@addrs) {
if (exists($symbols->{$a})) {
my $func = $symbols->{$a}->[0];
# Remove suffix in the symbols following space when filtering.
$func =~ s/ .*//;
if ($skip{$func} || ($func =~ m/$skip_regexp/)) {
# Throw away the portion of the backtrace seen so far, under the
# assumption that previous frames were for functions internal to the
@ -4500,19 +4525,19 @@ sub FindLibrary {
# For libc libraries, the copy in /usr/lib/debug contains debugging symbols
sub DebuggingLibrary {
my $file = shift;
if ($file !~ m|^/|) {
return undef;
}
# Find debug symbol file if it's named after the library's name.
if (-f "/usr/lib/debug$file") {
if (-f "/usr/lib/debug$file") {
if($main::opt_debug) { print STDERR "found debug info for $file in /usr/lib/debug$file\n"; }
return "/usr/lib/debug$file";
} elsif (-f "/usr/lib/debug$file.debug") {
if($main::opt_debug) { print STDERR "found debug info for $file in /usr/lib/debug$file.debug\n"; }
return "/usr/lib/debug$file.debug";
return "/usr/lib/debug$file.debug";
}
if(!$main::opt_debug_syms_by_id) {
@ -4521,7 +4546,7 @@ sub DebuggingLibrary {
}
# Find debug file if it's named after the library's build ID.
my $readelf = '';
if (!$main::gave_up_on_elfutils) {
$readelf = qx/eu-readelf -n ${file}/;
@ -4657,7 +4682,65 @@ sub ParseTextSectionHeaderFromOtool {
return $r;
}
# Parse text section header of a library in OS X shared cache using dyld_info
sub ParseTextSectionHeaderFromDyldInfo {
my $lib = shift;
my $size = undef;
my $vma;
my $file_offset;
# Get dyld_info output from the library file to figure out how to
# map between mapped addresses and addresses in the library.
my $cmd = ShellEscape($obj_tool_map{"dyld_info"}, "-segments", $lib);
open(DYLD, "$cmd |") || error("$cmd: $!\n");
while (<DYLD>) {
s/\r//g; # turn windows-looking lines into unix-looking lines
# -segments:
# load-address segment section sect-size seg-size perm
# 0x1803E0000 __TEXT 112KB r.x
# 0x1803E4F34 __text 80960
# 0x1803F8B74 __auth_stubs 768
# 0x1803F8E74 __init_offsets 4
# 0x1803F8E78 __gcc_except_tab 1180
my @x = split;
if ($#x >= 2) {
if ($x[0] eq 'load-offset') {
# dyld_info should only be used for the shared lib.
return undef;
} elsif ($x[1] eq '__TEXT') {
$file_offset = $x[0];
} elsif ($x[1] eq '__text') {
$size = $x[2];
$vma = $x[0];
$file_offset = AddressSub($x[0], $file_offset);
last;
}
}
}
close(DYLD);
if (!defined($vma) || !defined($size) || !defined($file_offset)) {
return undef;
}
my $r = {};
$r->{size} = $size;
$r->{vma} = $vma;
$r->{file_offset} = $file_offset;
return $r;
}
sub ParseTextSectionHeader {
# obj_tool_map("dyld_info") is only defined if we're in a Mach-O environment
if (defined($obj_tool_map{"dyld_info"})) {
my $r = ParseTextSectionHeaderFromDyldInfo(@_);
if (defined($r)){
return $r;
}
}
# if dyld_info doesn't work, or we don't have it, fall back to otool
# obj_tool_map("otool") is only defined if we're in a Mach-O environment
if (defined($obj_tool_map{"otool"})) {
my $r = ParseTextSectionHeaderFromOtool(@_);
@ -4698,7 +4781,7 @@ sub ParseLibraries {
$offset = HexExtend($3);
$lib = $4;
$lib =~ s|\\|/|g; # turn windows-style paths into unix-style paths
} elsif ($l =~ /^\s*($h)-($h):\s*(\S+\.so(\.\d+)*)/) {
} elsif ($l =~ /^\s*($h)-($h):\s*(\S+\.(so|dll|dylib|bundle)(\.\d+)*)/) {
# Cooked line from DumpAddressMap. Example:
# 40000000-40015000: /lib/ld-2.3.2.so
$start = HexExtend($1);
@ -4715,6 +4798,15 @@ sub ParseLibraries {
$offset = HexExtend($3);
$lib = $4;
$lib =~ s|\\|/|g; # turn windows-style paths into unix-style paths
} elsif (($l =~ /^\s*($h)-($h):\s*(\S+)/) && ($3 eq $prog)) {
# PIEs and address space randomization do not play well with our
# default assumption that main executable is at lowest
# addresses. So we're detecting main executable from
# DumpAddressMap as well.
$start = HexExtend($1);
$finish = HexExtend($2);
$offset = $zero_offset;
$lib = $3;
}
# FreeBSD 10.0 virtual memory map /proc/curproc/map as defined in
# function procfs_doprocmap (sys/fs/procfs/procfs_map.c)
@ -5245,6 +5337,7 @@ sub ConfigureObjTools {
if ($file_type =~ /Mach-O/) {
# OS X uses otool to examine Mach-O files, rather than objdump.
$obj_tool_map{"otool"} = "otool";
$obj_tool_map{"dyld_info"} = "dyld_info";
$obj_tool_map{"addr2line"} = "false"; # no addr2line
$obj_tool_map{"objdump"} = "false"; # no objdump
}

1345
build-aux/config.guess vendored
View file

File diff suppressed because it is too large Load diff

1015
build-aux/config.sub vendored
View file

File diff suppressed because it is too large Load diff

14
build-aux/install-sh
View file

@ -115,7 +115,7 @@ fi
if [ x"$dir_arg" != x ]; then
dst=$src
src=""
if [ -d $dst ]; then
instcmd=:
else
@ -124,7 +124,7 @@ if [ x"$dir_arg" != x ]; then
else
# Waiting for this to be detected by the "$instcmd $src $dsttmp" command
# might cause directories to be created, which would be especially bad
# might cause directories to be created, which would be especially bad
# if $src (and thus $dsttmp) contains '*'.
if [ -f $src -o -d $src ]
@ -134,7 +134,7 @@ else
echo "install: $src does not exist"
exit 1
fi
if [ x"$dst" = x ]
then
echo "install: no destination specified"
@ -201,17 +201,17 @@ else
# If we're going to rename the final executable, determine the name now.
if [ x"$transformarg" = x ]
if [ x"$transformarg" = x ]
then
dstfile=`basename $dst`
else
dstfile=`basename $dst $transformbasename |
dstfile=`basename $dst $transformbasename |
sed $transformarg`$transformbasename
fi
# don't allow the sed command to completely eliminate the filename
if [ x"$dstfile" = x ]
if [ x"$dstfile" = x ]
then
dstfile=`basename $dst`
else
@ -242,7 +242,7 @@ else
# Now rename the file to the real destination.
$doit $rmcmd -f $dstdir/$dstfile &&
$doit $mvcmd $dsttmp $dstdir/$dstfile
$doit $mvcmd $dsttmp $dstdir/$dstfile
fi &&

498
configure.ac
View file

@ -92,6 +92,32 @@ AC_LANG_POP([C++])
JE_CONCAT_VVV(CXXFLAGS, CONFIGURE_CXXFLAGS, SPECIFIED_CXXFLAGS)
])
CONFIGURE_LDFLAGS=
SPECIFIED_LDFLAGS="${LDFLAGS}"
dnl JE_LDFLAGS_ADD(ldflag)
dnl
dnl LDFLAGS is the concatenation of CONFIGURE_LDFLAGS and SPECIFIED_LDFLAGS
dnl This macro appends to CONFIGURE_LDFLAGS and regenerates LDFLAGS.
AC_DEFUN([JE_LDFLAGS_ADD],
[
AC_MSG_CHECKING([whether linker supports $1])
T_CONFIGURE_LDFLAGS="${CONFIGURE_LDFLAGS}"
JE_APPEND_VS(CONFIGURE_LDFLAGS, $1)
JE_CONCAT_VVV(LDFLAGS, CONFIGURE_LDFLAGS, SPECIFIED_LDFLAGS)
AC_LINK_IFELSE([AC_LANG_PROGRAM(
[[
]], [[
return 0;
]])],
[je_cv_ldflags_added=$1]
AC_MSG_RESULT([yes]),
[je_cv_ldflags_added=]
AC_MSG_RESULT([no])
[CONFIGURE_LDFLAGS="${T_CONFIGURE_LDFLAGS}"]
)
JE_CONCAT_VVV(LDFLAGS, CONFIGURE_LDFLAGS, SPECIFIED_LDFLAGS)
])
dnl JE_COMPILABLE(label, hcode, mcode, rvar)
dnl
dnl Use AC_LINK_IFELSE() rather than AC_COMPILE_IFELSE() so that linker errors
@ -298,6 +324,15 @@ fi
,
enable_cxx="1"
)
AC_ARG_WITH([cxx_stdlib],
[AS_HELP_STRING([--with-cxx-stdlib=<libstdc++|libcxx>],
[Specify the C++ standard library to link (default: probe for libstdc++)])],
[case "${with_cxx_stdlib}" in
libstdc++|libcxx) ;;
*) AC_MSG_ERROR([bad value ${with_cxx_stdlib} for --with-cxx-stdlib]) ;;
esac],
[with_cxx_stdlib=""]
)
if test "x$enable_cxx" = "x1" ; then
dnl Require at least c++14, which is the first version to support sized
dnl deallocation. C++ support is not compiled otherwise.
@ -312,17 +347,28 @@ if test "x$enable_cxx" = "x1" ; then
JE_CXXFLAGS_ADD([-g3])
SAVED_LIBS="${LIBS}"
JE_APPEND_VS(LIBS, -lstdc++)
JE_COMPILABLE([libstdc++ linkage], [
case "${with_cxx_stdlib}" in
libstdc++)
JE_APPEND_VS(LIBS, -lstdc++)
;;
libcxx)
JE_APPEND_VS(LIBS, -lc++)
;;
*)
dnl Probe for libstdc++ (the default when --with-cxx-stdlib is not given).
JE_APPEND_VS(LIBS, -lstdc++)
JE_COMPILABLE([libstdc++ linkage], [
#include <stdlib.h>
], [[
int *arr = (int *)malloc(sizeof(int) * 42);
if (arr == NULL)
return 1;
]], [je_cv_libstdcxx])
if test "x${je_cv_libstdcxx}" = "xno" ; then
LIBS="${SAVED_LIBS}"
fi
if test "x${je_cv_libstdcxx}" = "xno" ; then
LIBS="${SAVED_LIBS}"
fi
;;
esac
else
enable_cxx="0"
fi
@ -510,6 +556,23 @@ typedef unsigned __int32 uint32_t;
else
AC_MSG_ERROR([cannot determine number of significant virtual address bits])
fi
AC_CACHE_CHECK([rdtscp support],
[je_cv_rdtscp],
AC_RUN_IFELSE([AC_LANG_PROGRAM(
[[
#include <stdint.h>
]],
[[
unsigned int dx;
asm volatile("rdtscp" : "=d"(dx) ::);
return 0;
]])],
[je_cv_rdtscp=yes],
[je_cv_rdtscp=no],
[je_cv_rdtscp=no]))
if test "x${je_cv_rdtscp}" = "xyes"; then
AC_DEFINE([JEMALLOC_HAVE_RDTSCP], [ ], [ ])
fi
fi
;;
*)
@ -529,6 +592,37 @@ typedef unsigned __int32 uint32_t;
;;
esac
AC_DEFINE_UNQUOTED([LG_VADDR], [$LG_VADDR], [ ])
AC_CACHE_CHECK([asm volatile support],
[je_cv_asm_volatile],
AC_RUN_IFELSE([AC_LANG_PROGRAM(
[[
]],
[[
void* ptr;
asm volatile("" : "+r"(ptr));
return 0;
]])],
[je_cv_asm_volatile=yes],
[je_cv_asm_volatile=no],
[je_cv_asm_volatile=no]))
if test "x${je_cv_asm_volatile}" = "xyes"; then
AC_DEFINE([JEMALLOC_HAVE_ASM_VOLATILE], [ ], [ ])
fi
AC_CACHE_CHECK([__int128 support],
[je_cv_int128],
AC_RUN_IFELSE([AC_LANG_PROGRAM(
[[
]],
[[
__int128 temp = 0;
return temp;
]])],
[je_cv_int128=yes],
[je_cv_int128=no],
[je_cv_int128=no]))
if test "x${je_cv_int128}" = "xyes"; then
AC_DEFINE([JEMALLOC_HAVE_INT128], [ ], [ ])
fi
LD_PRELOAD_VAR="LD_PRELOAD"
so="so"
@ -578,7 +672,7 @@ AC_ARG_WITH([version],
[AS_HELP_STRING([--with-version=<major>.<minor>.<bugfix>-<nrev>-g<gid>],
[Version string])],
[
echo "${with_version}" | grep ['^[0-9]\+\.[0-9]\+\.[0-9]\+-[0-9]\+-g[0-9a-f]\+$'] 2>&1 1>/dev/null
echo "${with_version}" | grep ['^[0-9]\{1,\}\.[0-9]\{1,\}\.[0-9]\{1,\}-[0-9]\{1,\}-g[0-9a-f]\{1,\}$'] 2>&1 1>/dev/null
if test $? -eq 0 ; then
echo "$with_version" > "${objroot}VERSION"
else
@ -654,6 +748,9 @@ case "${host}" in
SOREV="${rev}.${so}"
sbrk_deprecated="1"
SYM_PREFIX="_"
if test "${LG_SIZEOF_PTR}" = "3"; then
default_retain="1"
fi
;;
*-*-freebsd*)
JE_APPEND_VS(CPPFLAGS, -D_BSD_SOURCE)
@ -687,6 +784,19 @@ case "${host}" in
fi
zero_realloc_default_free="1"
;;
*-*-linux-musl*)
dnl syscall(2) and secure_getenv(3) are exposed by _GNU_SOURCE.
JE_APPEND_VS(CPPFLAGS, -D_GNU_SOURCE)
abi="elf"
AC_DEFINE([JEMALLOC_PURGE_MADVISE_DONTNEED_ZEROS], [ ], [ ])
AC_DEFINE([JEMALLOC_HAS_ALLOCA_H], [ ], [ ])
AC_DEFINE([JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY], [ ], [ ])
AC_DEFINE([JEMALLOC_THREADED_INIT], [ ], [ ])
if test "${LG_SIZEOF_PTR}" = "3"; then
default_retain="1"
fi
zero_realloc_default_free="1"
;;
*-*-linux*)
dnl syscall(2) and secure_getenv(3) are exposed by _GNU_SOURCE.
JE_APPEND_VS(CPPFLAGS, -D_GNU_SOURCE)
@ -829,11 +939,26 @@ AC_SUBST([DUMP_SYMS])
AC_SUBST([CC_MM])
dnl Determine whether libm must be linked to use e.g. log(3).
AC_SEARCH_LIBS([log], [m], , [AC_MSG_ERROR([Missing math functions])])
if test "x$ac_cv_search_log" != "xnone required" ; then
LM="$ac_cv_search_log"
else
# On MSVC, log is an intrinsic that doesn't require libm. However,
# AC_SEARCH_LIBS does not successfully detect this, as it will try to compile
# a program using the wrong signature for log. Newer versions of MSVC CL detects
# this and rejects the program with the following messages.
#
# conftest.c(40): warning C4391: 'char log()': incorrect return type for intrinsic function, expected 'double'
# conftest.c(44): error C2168: 'log': too few actual parameters for intrinsic function
#
# Since log is always available on MSVC (it's been around since the dawn of
# time), we simply always assume it's there if MSVC is detected.
if test "x$je_cv_msvc" = "xyes" ; then
LM=
else
AC_SEARCH_LIBS([log], [m], , [AC_MSG_ERROR([Missing math functions])])
if test "x$ac_cv_search_log" != "xnone required" ; then
LM="$ac_cv_search_log"
else
LM=
fi
fi
AC_SUBST(LM)
@ -939,6 +1064,30 @@ if test "x${je_cv_cold}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_ATTR_COLD], [ ], [ ])
fi
dnl Check for deprecated attribute support.
JE_CFLAGS_SAVE()
JE_CFLAGS_ADD([-Wdeprecated-declarations])
JE_COMPILABLE([deprecated attribute],
[#if !__has_attribute(deprecated)
#error "deprecated attribute not supported"
#endif
struct has_deprecated_field {
int good;
int __attribute__((deprecated("Do not use"))) bad;
};
],
[struct has_deprecated_field instance;
instance.good = 0;
instance.bad = 1;
],
[je_cv_deprecated])
JE_CFLAGS_RESTORE()
if test "x${je_cv_deprecated}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_ATTR_DEPRECATED], [ ], [ ])
JE_CFLAGS_ADD([-Wdeprecated-declarations])
JE_CXXFLAGS_ADD([-Wdeprecated-declarations])
fi
dnl Check for VM_MAKE_TAG for mmap support.
JE_COMPILABLE([vm_make_tag],
[#include <sys/mman.h>
@ -1052,11 +1201,11 @@ AC_SUBST([JEMALLOC_CPREFIX])
AC_ARG_WITH([export],
[AS_HELP_STRING([--without-export], [disable exporting jemalloc public APIs])],
[if test "x$with_export" = "xno"; then
AC_DEFINE([JEMALLOC_EXPORT],[], [ ])
AC_DEFINE([JEMALLOC_EXPORT], [ ], [ ])
fi]
)
public_syms="aligned_alloc calloc dallocx free mallctl mallctlbymib mallctlnametomib malloc malloc_conf malloc_conf_2_conf_harder malloc_message malloc_stats_print malloc_usable_size mallocx smallocx_${jemalloc_version_gid} nallocx posix_memalign rallocx realloc sallocx sdallocx xallocx"
public_syms="aligned_alloc calloc dallocx free free_sized free_aligned_sized mallctl mallctlbymib mallctlnametomib malloc malloc_conf malloc_conf_2_conf_harder malloc_message malloc_stats_print malloc_usable_size mallocx smallocx_${jemalloc_version_gid} nallocx posix_memalign rallocx realloc sallocx sdallocx xallocx"
dnl Check for additional platform-specific public API functions.
AC_CHECK_FUNC([memalign],
[AC_DEFINE([JEMALLOC_OVERRIDE_MEMALIGN], [ ], [ ])
@ -1064,6 +1213,9 @@ AC_CHECK_FUNC([memalign],
AC_CHECK_FUNC([valloc],
[AC_DEFINE([JEMALLOC_OVERRIDE_VALLOC], [ ], [ ])
public_syms="${public_syms} valloc"])
AC_CHECK_FUNC([pvalloc],
[AC_DEFINE([JEMALLOC_OVERRIDE_PVALLOC], [ ], [ ])
public_syms="${public_syms} pvalloc"])
AC_CHECK_FUNC([malloc_size],
[AC_DEFINE([JEMALLOC_HAVE_MALLOC_SIZE], [ ], [ ])
public_syms="${public_syms} malloc_size"])
@ -1077,6 +1229,16 @@ if test "x${JEMALLOC_PREFIX}" = "x" ; then
AC_CHECK_FUNC([__libc_free],
[AC_DEFINE([JEMALLOC_OVERRIDE___LIBC_FREE], [ ], [ ])
wrap_syms="${wrap_syms} __libc_free"])
dnl __libc_free_sized and __libc_free_aligned_sized are here speculatively
dnl under the assumption that glibc will eventually define symbols with these
dnl names. In the event glibc chooses different names for these symbols,
dnl these will need to be amended to match.
AC_CHECK_FUNC([__libc_free_sized],
[AC_DEFINE([JEMALLOC_OVERRIDE___LIBC_FREE_SIZED], [ ], [ ])
wrap_syms="${wrap_syms} __libc_free_sized"])
AC_CHECK_FUNC([__libc_free_aligned_sized],
[AC_DEFINE([JEMALLOC_OVERRIDE___LIBC_FREE_ALIGNED_SIZED], [ ], [ ])
wrap_syms="${wrap_syms} __libc_free_aligned_sized"])
AC_CHECK_FUNC([__libc_malloc],
[AC_DEFINE([JEMALLOC_OVERRIDE___LIBC_MALLOC], [ ], [ ])
wrap_syms="${wrap_syms} __libc_malloc"])
@ -1089,6 +1251,9 @@ if test "x${JEMALLOC_PREFIX}" = "x" ; then
AC_CHECK_FUNC([__libc_valloc],
[AC_DEFINE([JEMALLOC_OVERRIDE___LIBC_VALLOC], [ ], [ ])
wrap_syms="${wrap_syms} __libc_valloc"])
AC_CHECK_FUNC([__libc_pvalloc],
[AC_DEFINE([JEMALLOC_OVERRIDE___LIBC_PVALLOC], [ ], [ ])
wrap_syms="${wrap_syms} __libc_pvalloc"])
AC_CHECK_FUNC([__posix_memalign],
[AC_DEFINE([JEMALLOC_OVERRIDE___POSIX_MEMALIGN], [ ], [ ])
wrap_syms="${wrap_syms} __posix_memalign"])
@ -1256,6 +1421,23 @@ if test "x$enable_stats" = "x1" ; then
fi
AC_SUBST([enable_stats])
dnl Disable reading configuration from file and environment variable
AC_ARG_ENABLE([user_config],
[AS_HELP_STRING([--disable-user-config],
[Do not read malloc config from /etc/malloc.conf or MALLOC_CONF])],
[if test "x$enable_user_config" = "xno" ; then
enable_user_config="0"
else
enable_user_config="1"
fi
],
[enable_user_config="1"]
)
if test "x$enable_user_config" = "x1" ; then
AC_DEFINE([JEMALLOC_CONFIG_ENV], [ ], [ ])
AC_DEFINE([JEMALLOC_CONFIG_FILE], [ ], [ ])
fi
dnl Do not enable smallocx by default.
AC_ARG_ENABLE([experimental_smallocx],
[AS_HELP_STRING([--enable-experimental-smallocx], [Enable experimental smallocx API])],
@ -1329,6 +1511,33 @@ if test "x$backtrace_method" = "x" -a "x$enable_prof_libunwind" = "x1" ; then
fi
fi
if test `uname -s` = "Linux"
then
AC_ARG_ENABLE([prof-frameptr],
[AS_HELP_STRING([--enable-prof-frameptr], [Use optimized frame pointer unwinder for backtracing (Linux only)])],
[if test "x$enable_prof_frameptr" = "xno" ; then
enable_prof_frameptr="0"
else
enable_prof_frameptr="1"
if test "x$enable_prof" = "x0" ; then
AC_MSG_ERROR([--enable-prof-frameptr should only be used with --enable-prof])
fi
fi
],
[enable_prof_frameptr="0"]
)
if test "x$backtrace_method" = "x" -a "x$enable_prof_frameptr" = "x1" \
-a "x$GCC" = "xyes" ; then
JE_CFLAGS_ADD([-fno-omit-frame-pointer])
backtrace_method="frame pointer linux"
AC_DEFINE([JEMALLOC_PROF_FRAME_POINTER], [ ], [ ])
else
enable_prof_frameptr="0"
fi
else
enable_prof_frameptr="0"
fi
AC_ARG_ENABLE([prof-libgcc],
[AS_HELP_STRING([--disable-prof-libgcc],
[Do not use libgcc for backtracing])],
@ -1404,6 +1613,18 @@ if test "x$zero_realloc_default_free" = "x1" ; then
AC_DEFINE([JEMALLOC_ZERO_REALLOC_DEFAULT_FREE], [ ], [ ])
fi
dnl Support allocation from DSS by default
AC_ARG_ENABLE([dss],
[AS_HELP_STRING([--disable-dss], [Disable usage of sbrk(2)])],
[if test "x$enable_dss" = "xno" ; then
enable_dss="0"
else
enable_dss="1"
fi
],
[enable_dss="1"]
)
dnl Enable allocation from DSS if supported by the OS.
have_dss="1"
dnl Check whether the BSD/SUSv1 sbrk() exists. If not, disable DSS support.
@ -1417,7 +1638,7 @@ else
have_dss="0"
fi
if test "x$have_dss" = "x1" ; then
if test "x$have_dss" = "x1" -a "x$enable_dss" = "x1" ; then
AC_DEFINE([JEMALLOC_DSS], [ ], [ ])
fi
@ -1480,6 +1701,55 @@ else
fi
AC_SUBST([enable_utrace])
dnl Disable experimental sdt tracing by default.
AC_ARG_ENABLE([experimental-sdt],
[AS_HELP_STRING([--enable-experimental-sdt], [Enable systemtap USDT probes])],
[if test "x$enable_experimental_sdt" = "xno" ; then
enable_experimental_sdt="0"
else
JE_COMPILABLE([systemtap sdt], [
#include <sys/sdt.h>
], [
void foo(int i, void *p) { STAP_PROBE2(jemalloc, test, i, p); }
],
[je_cv_stap_sdt])
if test "x${je_cv_stap_sdt}" = "xyes" ; then
enable_experimental_sdt="1"
elif test "x${abi}" = "xelf" ; then
case "${host}" in
*-*-linux-android*)
case "${host_cpu}" in aarch64|x86_64)
enable_experimental_sdt="2"
;;
esac
;;
*-*-linux*)
case "${host_cpu}" in x86_64|aarch64|arm*)
enable_experimental_sdt="2"
;;
esac
;;
*)
enable_experimental_sdt="0"
AC_MSG_ERROR([Unsupported sdt on this platform])
;;
esac
else
AC_MSG_ERROR([Unsupported sdt on this platform])
fi
fi
],
[enable_experimental_sdt="0"]
)
if test "x$enable_experimental_sdt" = "x1" ; then
AC_DEFINE([JEMALLOC_EXPERIMENTAL_USDT_STAP], [ ], [ ])
elif test "x$enable_experimental_sdt" = "x2"; then
AC_DEFINE([JEMALLOC_EXPERIMENTAL_USDT_CUSTOM], [ ], [ ])
fi
AC_SUBST([enable_experimental_sdt])
dnl Do not support the xmalloc option by default.
AC_ARG_ENABLE([xmalloc],
[AS_HELP_STRING([--enable-xmalloc], [Support xmalloc option])],
@ -1545,6 +1815,22 @@ if test "x$enable_readlinkat" = "x1" ; then
fi
AC_SUBST([enable_readlinkat])
dnl Do not force getenv by default
AC_ARG_ENABLE([force-getenv],
[AS_HELP_STRING([--enable-force-getenv], [Use getenv over secure_getenv])],
[if test "x$enable_force_getenv" = "xno" ; then
enable_force_getenv="0"
else
enable_force_getenv="1"
fi
],
[enable_force_getenv="0"]
)
if test "x$enable_force_getenv" = "x1" ; then
AC_DEFINE([JEMALLOC_FORCE_GETENV], [ ], [ ])
fi
AC_SUBST([force_getenv])
dnl Avoid extra safety checks by default
AC_ARG_ENABLE([opt-safety-checks],
[AS_HELP_STRING([--enable-opt-safety-checks],
@ -1592,7 +1878,7 @@ fi
[enable_uaf_detection="0"]
)
if test "x$enable_uaf_detection" = "x1" ; then
AC_DEFINE([JEMALLOC_UAF_DETECTION], [ ])
AC_DEFINE([JEMALLOC_UAF_DETECTION], [ ], [ ])
fi
AC_SUBST([enable_uaf_detection])
@ -1694,6 +1980,16 @@ case "${host}" in
LG_PAGE=14
fi
;;
*-*-linux-android)
if test "x$LG_PAGE" = "xdetect"; then
AC_CHECK_DECLS([PAGE_SIZE], [LG_PAGE=12], [LG_PAGE=14], [#include <sys/user.h>])
fi
;;
aarch64-unknown-linux-*)
if test "x$LG_PAGE" = "xdetect"; then
LG_PAGE=16
fi
;;
esac
if test "x$LG_PAGE" = "xdetect"; then
AC_CACHE_CHECK([LG_PAGE],
@ -1758,7 +2054,7 @@ if test "x${je_cv_lg_hugepage}" = "x" ; then
dnl Hugepagesize: 2048 kB
if test -e "/proc/meminfo" ; then
hpsk=[`cat /proc/meminfo 2>/dev/null | \
grep -e '^Hugepagesize:[[:space:]]\+[0-9]\+[[:space:]]kB$' | \
grep '^Hugepagesize:[[:space:]]\{1,\}[0-9]\{1,\}[[:space:]]kB$' | \
awk '{print $2}'`]
if test "x${hpsk}" != "x" ; then
je_cv_lg_hugepage=10
@ -1855,6 +2151,16 @@ dnl Check if we have dlsym support.
if test "x${je_cv_pthread_getname_np}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_PTHREAD_GETNAME_NP], [ ], [ ])
fi
dnl Check if pthread_set_name_np is available with the expected API.
JE_COMPILABLE([pthread_set_name_np(3)], [
#include <pthread.h>
#include <pthread_np.h>
], [
pthread_set_name_np(pthread_self(), "set_name_test");
], [je_cv_pthread_set_name_np])
if test "x${je_cv_pthread_set_name_np}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_PTHREAD_SET_NAME_NP], [ ], [ ])
fi
dnl Check if pthread_get_name_np is not necessarily present despite
dnl the pthread_set_name_np counterpart
JE_COMPILABLE([pthread_get_name_np(3)], [
@ -1942,6 +2248,16 @@ if test "x${je_cv_clock_realtime}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_CLOCK_REALTIME], [ ], [ ])
fi
dnl Check for clock_gettime_nsec_np().
JE_COMPILABLE([clock_gettime_nsec_np()], [
#include <time.h>
], [
clock_gettime_nsec_np(CLOCK_UPTIME_RAW);
], [je_cv_clock_gettime_nsec_np])
if test "x${je_cv_clock_gettime_nsec_np}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_CLOCK_GETTIME_NSEC_NP], [ ], [ ])
fi
dnl Use syscall(2) (if available) by default.
AC_ARG_ENABLE([syscall],
[AS_HELP_STRING([--disable-syscall], [Disable use of syscall(2)])],
@ -1998,6 +2314,15 @@ if test "x$have_sched_setaffinity" = "x1" ; then
AC_DEFINE([JEMALLOC_HAVE_SCHED_SETAFFINITY], [ ], [ ])
fi
dnl Check if the pthread_setaffinity_np function exists.
AC_CHECK_FUNC([pthread_setaffinity_np],
[have_pthread_setaffinity_np="1"],
[have_pthread_setaffinity_np="0"]
)
if test "x$have_pthread_setaffinity_np" = "x1" ; then
AC_DEFINE([JEMALLOC_HAVE_PTHREAD_SETAFFINITY_NP], [ ], [ ])
fi
dnl Check if the Solaris/BSD issetugid function exists.
AC_CHECK_FUNC([issetugid],
[have_issetugid="1"],
@ -2041,6 +2366,14 @@ if test "x$have_memcntl" = "x1" ; then
AC_DEFINE([JEMALLOC_HAVE_MEMCNTL], [ ], [ ])
fi
AC_CHECK_FUNC([prctl],
[have_prctl="1"],
[have_prctl="0"],
)
if test "x$have_prctl" = "x1" ; then
AC_DEFINE([JEMALLOC_HAVE_PRCTL], [ ], [ ])
fi
dnl Disable lazy locking by default.
AC_ARG_ENABLE([lazy_lock],
[AS_HELP_STRING([--enable-lazy-lock],
@ -2203,6 +2536,13 @@ if test "x${je_cv_osatomic}" = "xyes" ; then
fi
dnl ============================================================================
AC_ARG_WITH([experimental_sys_process_madvise],
[AS_HELP_STRING([--with-experimental-sys-process-madvise=<experimental-sys-process-madvise>],
[Force process_madvise and use experimental-sys-process-madvise number when making syscall])],
[je_cv_sys_pmadv_nr="${with_experimental_sys_process_madvise}"],
[je_cv_sys_pmadv_nr=""])
dnl Check for madvise(2).
JE_COMPILABLE([madvise(2)], [
@ -2260,6 +2600,16 @@ if test "x${je_cv_madvise}" = "xyes" ; then
madvise((void *)0, 0, MADV_HUGEPAGE);
madvise((void *)0, 0, MADV_NOHUGEPAGE);
], [je_cv_thp])
case "${host_cpu}" in
arm*)
;;
*)
if test "x${je_cv_thp}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_MADVISE_HUGE], [ ], [ ])
fi
;;
esac
dnl Check for madvise(..., MADV_[NO]CORE).
JE_COMPILABLE([madvise(..., MADV_[[NO]]CORE)], [
#include <sys/mman.h>
@ -2270,15 +2620,35 @@ if test "x${je_cv_madvise}" = "xyes" ; then
if test "x${je_cv_madv_nocore}" = "xyes" ; then
AC_DEFINE([JEMALLOC_MADVISE_NOCORE], [ ], [ ])
fi
case "${host_cpu}" in
arm*)
;;
*)
if test "x${je_cv_thp}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_MADVISE_HUGE], [ ], [ ])
dnl Check for madvise(..., MADV_COLLAPSE).
JE_COMPILABLE([madvise(..., MADV_COLLAPSE)], [
#include <sys/mman.h>
], [
madvise((void *)0, 0, MADV_COLLAPSE);
], [je_cv_madv_collapse])
if test "x${je_cv_madv_collapse}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_MADVISE_COLLAPSE], [ ], [ ])
fi
dnl Check for process_madvise
JE_COMPILABLE([process_madvise(2)], [
#include <sys/pidfd.h>
#include <sys/syscall.h>
#include <unistd.h>
], [
syscall(SYS_process_madvise, PIDFD_SELF, (void *)0, 0, 0, 0);
], [je_cv_process_madvise])
if test "x${je_cv_process_madvise}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_PROCESS_MADVISE], [ ], [ ])
else
if test "x${je_cv_sys_pmadv_nr}" != "x" ; then
dnl Forcing experimental usage of process_madvise
AC_MSG_RESULT([Forcing usage of process_madvise with syscall nr=${je_cv_sys_pmadv_nr}])
AC_DEFINE([JEMALLOC_HAVE_PROCESS_MADVISE], [ ], [ ])
AC_DEFINE_UNQUOTED([EXPERIMENTAL_SYS_PROCESS_MADVISE_NR], [${je_cv_sys_pmadv_nr}], [ ])
fi
fi
;;
esac
else
dnl Check for posix_madvise.
JE_COMPILABLE([posix_madvise], [
@ -2403,12 +2773,62 @@ AC_SUBST([enable_initial_exec_tls])
if test "x${je_cv_tls_model}" = "xyes" -a \
"x${enable_initial_exec_tls}" = "x1" ; then
AC_DEFINE([JEMALLOC_TLS_MODEL],
[__attribute__((tls_model("initial-exec")))],
[__attribute__((tls_model("initial-exec")))],
[ ])
else
AC_DEFINE([JEMALLOC_TLS_MODEL], [ ], [ ])
fi
dnl Do not compile with debugging by default.
AC_ARG_ENABLE([pageid],
[AS_HELP_STRING([--enable-pageid],
[Enable named pages])],
[if test "x$enable_pageid" = "xno" ; then
enable_pageid="0"
else
enable_pageid="1"
fi
],
[enable_pageid="0"]
)
if test "x$enable_pageid" = "x1" ; then
AC_DEFINE([JEMALLOC_PAGEID], [ ], [ ])
fi
AC_ARG_ENABLE([tsan],
[AS_HELP_STRING([--enable-tsan],
[Enable thread sanitizer])],
[if test "x$enable_tsan" = "xno" ; then
enable_tsan="0"
else
enable_tsan="1"
fi
],
[enable_tsan="0"]
)
if test "x$enable_tsan" = "x1" ; then
JE_CFLAGS_ADD([-fsanitize=thread])
JE_CXXFLAGS_ADD([-fsanitize=thread])
JE_LDFLAGS_ADD([-fsanitize=thread])
fi
AC_ARG_ENABLE([ubsan],
[AS_HELP_STRING([--enable-ubsan],
[Enable undefined behavior sanitizer])],
[if test "x$enable_ubsan" = "xno" ; then
enable_ubsan="0"
else
enable_ubsan="1"
fi
],
[enable_ubsan="0"]
)
if test "x$enable_ubsan" = "x1" ; then
JE_CFLAGS_ADD([-fsanitize=undefined])
JE_CXXFLAGS_ADD([-fsanitize=undefined])
JE_LDFLAGS_ADD([-fsanitize=undefined])
fi
dnl ============================================================================
dnl Enable background threads if possible.
@ -2468,6 +2888,15 @@ if test "x${je_cv_pthread_mutex_adaptive_np}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_PTHREAD_MUTEX_ADAPTIVE_NP], [ ], [ ])
fi
JE_COMPILABLE([gettid], [
#include <unistd.h>
], [
int tid = gettid();
], [je_cv_gettid])
if test "x${je_cv_gettid}" = "xyes" ; then
AC_DEFINE([JEMALLOC_HAVE_GETTID], [ ], [ ])
fi
JE_CFLAGS_SAVE()
JE_CFLAGS_ADD([-D_GNU_SOURCE])
JE_CFLAGS_ADD([-Werror])
@ -2482,9 +2911,19 @@ JE_COMPILABLE([strerror_r returns char with gnu source], [
char *error = strerror_r(EINVAL, buffer, 100);
printf("%s\n", error);
], [je_cv_strerror_r_returns_char_with_gnu_source])
if test "x${je_cv_strerror_r_returns_char_with_gnu_source}" = "xno" ; then
JE_COMPILABLE([strerror_r header only], [
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
], [], [je_cv_strerror_r_header_pass])
fi
JE_CFLAGS_RESTORE()
if test "x${je_cv_strerror_r_returns_char_with_gnu_source}" = "xyes" ; then
AC_DEFINE([JEMALLOC_STRERROR_R_RETURNS_CHAR_WITH_GNU_SOURCE], [ ], [ ])
elif test "x${je_cv_strerror_r_header_pass}" = "xno" ; then
AC_MSG_ERROR([cannot determine return type of strerror_r])
fi
dnl ============================================================================
@ -2622,7 +3061,8 @@ AC_MSG_RESULT([CXX : ${CXX}])
AC_MSG_RESULT([CONFIGURE_CXXFLAGS : ${CONFIGURE_CXXFLAGS}])
AC_MSG_RESULT([SPECIFIED_CXXFLAGS : ${SPECIFIED_CXXFLAGS}])
AC_MSG_RESULT([EXTRA_CXXFLAGS : ${EXTRA_CXXFLAGS}])
AC_MSG_RESULT([LDFLAGS : ${LDFLAGS}])
AC_MSG_RESULT([CONFIGURE_LDFLAGS : ${CONFIGURE_LDFLAGS}])
AC_MSG_RESULT([SPECIFIED_LDFLAGS : ${SPECIFIED_LDFLAGS}])
AC_MSG_RESULT([EXTRA_LDFLAGS : ${EXTRA_LDFLAGS}])
AC_MSG_RESULT([DSO_LDFLAGS : ${DSO_LDFLAGS}])
AC_MSG_RESULT([LIBS : ${LIBS}])
@ -2638,6 +3078,8 @@ AC_MSG_RESULT([INCLUDEDIR : ${INCLUDEDIR}])
AC_MSG_RESULT([LIBDIR : ${LIBDIR}])
AC_MSG_RESULT([MANDIR : ${MANDIR}])
AC_MSG_RESULT([])
AC_MSG_RESULT([LG_PAGE : ${LG_PAGE}])
AC_MSG_RESULT([])
AC_MSG_RESULT([srcroot : ${srcroot}])
AC_MSG_RESULT([abs_srcroot : ${abs_srcroot}])
AC_MSG_RESULT([objroot : ${objroot}])
@ -2654,9 +3096,11 @@ AC_MSG_RESULT([static libs : ${enable_static}])
AC_MSG_RESULT([autogen : ${enable_autogen}])
AC_MSG_RESULT([debug : ${enable_debug}])
AC_MSG_RESULT([stats : ${enable_stats}])
AC_MSG_RESULT([user_config : ${enable_user_config}])
AC_MSG_RESULT([experimental_smallocx : ${enable_experimental_smallocx}])
AC_MSG_RESULT([prof : ${enable_prof}])
AC_MSG_RESULT([prof-libunwind : ${enable_prof_libunwind}])
AC_MSG_RESULT([prof-frameptr : ${enable_prof_frameptr}])
AC_MSG_RESULT([prof-libgcc : ${enable_prof_libgcc}])
AC_MSG_RESULT([prof-gcc : ${enable_prof_gcc}])
AC_MSG_RESULT([fill : ${enable_fill}])
@ -2665,5 +3109,9 @@ AC_MSG_RESULT([xmalloc : ${enable_xmalloc}])
AC_MSG_RESULT([log : ${enable_log}])
AC_MSG_RESULT([lazy_lock : ${enable_lazy_lock}])
AC_MSG_RESULT([cache-oblivious : ${enable_cache_oblivious}])
AC_MSG_RESULT([pageid : ${enable_pageid}])
AC_MSG_RESULT([cxx : ${enable_cxx}])
AC_MSG_RESULT([dss : ${enable_dss}])
AC_MSG_RESULT([tsan : ${enable_tsan}])
AC_MSG_RESULT([ubsan : ${enable_ubsan}])
AC_MSG_RESULT([===============================================================================])

216
doc/jemalloc.xml.in
View file

@ -33,6 +33,8 @@
<refname>aligned_alloc</refname>
<refname>realloc</refname>
<refname>free</refname>
<refname>free_sized</refname>
<refname>free_aligned_sized</refname>
<refname>mallocx</refname>
<refname>rallocx</refname>
<refname>xallocx</refname>
@ -89,6 +91,17 @@
<funcdef>void <function>free</function></funcdef>
<paramdef>void *<parameter>ptr</parameter></paramdef>
</funcprototype>
<funcprototype>
<funcdef>void <function>free_sized</function></funcdef>
<paramdef>void *<parameter>ptr</parameter></paramdef>
<paramdef>size_t <parameter>size</parameter></paramdef>
</funcprototype>
<funcprototype>
<funcdef>void <function>free_aligned_sized</function></funcdef>
<paramdef>void *<parameter>ptr</parameter></paramdef>
<paramdef>size_t <parameter>alignment</parameter></paramdef>
<paramdef>size_t <parameter>size</parameter></paramdef>
</funcprototype>
</refsect2>
<refsect2>
<title>Non-standard API</title>
@ -227,6 +240,17 @@
allocated memory referenced by <parameter>ptr</parameter> to be made
available for future allocations. If <parameter>ptr</parameter> is
<constant>NULL</constant>, no action occurs.</para>
<para>The <function>free_sized()</function> function is an extension of
<function>free()</function> with a <parameter>size</parameter> parameter
to allow the caller to pass in the allocation size as an optimization.
</para>
<para>The <function>free_aligned_sized()</function> function accepts a
<parameter>ptr</parameter> which was allocated with a requested
<parameter>size</parameter> and <parameter>alignment</parameter>, causing
the allocated memory referenced by <parameter>ptr</parameter> to be made
available for future allocations.</para>
</refsect2>
<refsect2>
<title>Non-standard API</title>
@ -451,6 +475,24 @@ for (i = 0; i < nbins; i++) {
depended on, since such behavior is entirely implementation-dependent.
</para>
</refsect2>
<refsect2>
<title>Interactions Between the Standard and Non-standard APIs</title>
<para>Generally speaking it is permissible to pass pointers obtained from
the standard API to the non-standard API and vice versa (e.g. calling
<function>free()</function> with a pointer returned by a call to
<function>mallocx()</function>, calling <function>sdallocx()</function>
with a pointer returned by a call to <function>calloc()</function>).
There are however a few exceptions. In keeping with the C23 standard –
which forbids calling <function>free_sized()</function> on a pointer
returned by <function>aligned_alloc()</function>, mandating that either
<function>free_aligned_sized()</function> or <function>free()</function>
be used instead – using any combination of the standard and non-standard
APIs in an equivalent fashion (i.e. taking a pointer which was allocated
with an explicitly requested alignment and attempting to free it via an
API that accepts a size hint, without also providing the alignment hint)
is likewise forbidden.
</para>
</refsect2>
</refsect1>
<refsect1 id="tuning">
<title>TUNING</title>
@ -1095,7 +1137,7 @@ mallctl("arena." STRINGIFY(MALLCTL_ARENAS_ALL) ".decay",
</term>
<listitem><para>Maximum number of background threads that will be created
if <link linkend="background_thread">background_thread</link> is set.
Defaults to number of cpus.</para></listitem>
Defaults to 4.</para></listitem>
</varlistentry>
<varlistentry id="opt.dirty_decay_ms">
@ -1121,9 +1163,7 @@ mallctl("arena." STRINGIFY(MALLCTL_ARENAS_ALL) ".decay",
linkend="arena.i.dirty_decay_ms"><mallctl>arena.&lt;i&gt;.dirty_decay_ms</mallctl></link>
for related dynamic control options. See <link
linkend="opt.muzzy_decay_ms"><mallctl>opt.muzzy_decay_ms</mallctl></link>
for a description of muzzy pages.for a description of muzzy pages. Note
that when the <link
linkend="opt.oversize_threshold"><mallctl>oversize_threshold</mallctl></link>
for a description of muzzy pages. Note that when the <link linkend="opt.oversize_threshold"><mallctl>oversize_threshold</mallctl></link>
feature is enabled, the arenas reserved for oversize requests may have
its own default decay settings.</para></listitem>
</varlistentry>
@ -1145,7 +1185,7 @@ mallctl("arena." STRINGIFY(MALLCTL_ARENAS_ALL) ".decay",
purged according to a sigmoidal decay curve that starts and ends with
zero purge rate. A decay time of 0 causes all unused muzzy pages to be
purged immediately upon creation. A decay time of -1 disables purging.
The default decay time is 10 seconds. See <link
Muzzy decay is disabled by default (with decay time 0). See <link
linkend="arenas.muzzy_decay_ms"><mallctl>arenas.muzzy_decay_ms</mallctl></link>
and <link
linkend="arena.i.muzzy_decay_ms"><mallctl>arena.&lt;i&gt;.muzzy_decay_ms</mallctl></link>
@ -1369,6 +1409,17 @@ malloc_conf = "xmalloc:true";]]></programlisting>
extent hooks.</para></listitem>
</varlistentry>
<varlistentry id="opt.prof_bt_max">
<term>
<mallctl>opt.prof_bt_max</mallctl>
(<type>unsigned</type>)
<literal>r-</literal>
[<option>--enable-prof</option>]
</term>
<listitem><para>Maximum number of stack frames to record in profiling
backtraces. The default is 128.</para></listitem>
</varlistentry>
<varlistentry id="opt.prof">
<term>
<mallctl>opt.prof</mallctl>
@ -1474,6 +1525,23 @@ malloc_conf = "xmalloc:true";]]></programlisting>
by default.</para></listitem>
</varlistentry>
<varlistentry id="opt.prof_pid_namespace">
<term>
<mallctl>opt.prof_pid_namespace</mallctl>
(<type>bool</type>)
<literal>r-</literal>
[<option>--enable-prof</option>]
</term>
<listitem><para>Enable adding the pid namespace to the profile
filename. Profiles are dumped to files named according to the pattern
<filename>&lt;prefix&gt;.&lt;pid_namespace&gt;.&lt;pid&gt;.&lt;seq&gt;.i&lt;iseq&gt;.heap</filename>,
where <literal>&lt;prefix&gt;</literal> is controlled by the <link
linkend="opt.prof_prefix"><mallctl>opt.prof_prefix</mallctl></link> and
<link linkend="prof.prefix"><mallctl>prof.prefix</mallctl></link>
options.
</para></listitem>
</varlistentry>
<varlistentry id="opt.lg_prof_interval">
<term>
<mallctl>opt.lg_prof_interval</mallctl>
@ -1599,6 +1667,53 @@ malloc_conf = "xmalloc:true";]]></programlisting>
testing this behavior.</para></listitem>
</varlistentry>
<varlistentry id="opt.debug_double_free_max_scan">
<term>
<mallctl>opt.debug_double_free_max_scan</mallctl>
(<type>unsigned</type>)
<literal>r-</literal>
[<option>--enable-debug</option>]
</term>
<listitem><para>Maximum number of cached pointers to scan in the
thread cache when checking for double-free errors on deallocation.
When debug is enabled, each deallocation into the tcache scans up to
this many recently cached pointers to detect whether the same pointer
is being freed twice. Setting this to 0 disables the check. This
option is set to 0 and has no effect when debug is not enabled. The
default is 32.</para></listitem>
</varlistentry>
<varlistentry id="opt.disable_large_size_classes">
<term>
<mallctl>opt.disable_large_size_classes</mallctl>
(<type>bool</type>)
<literal>r-</literal>
</term>
<listitem><para>When enabled (the default), large allocations
(i.e. allocations of size &gt;= <constant>SC_LARGE_MINCLASS</constant>)
are rounded up to the nearest page boundary rather than the nearest
large size class. This minimizes memory overhead, especially when
using hugepages, at the cost of disabling the standard large size
class hierarchy.</para></listitem>
</varlistentry>
<varlistentry id="opt.process_madvise_max_batch">
<term>
<mallctl>opt.process_madvise_max_batch</mallctl>
(<type>size_t</type>)
<literal>r-</literal>
</term>
<listitem><para>Maximum number of memory regions to include in each
<citerefentry><refentrytitle>process_madvise</refentrytitle>
<manvolnum>2</manvolnum></citerefentry> batch call. When set to 0
(the default), process_madvise is not used, and the standard
<citerefentry><refentrytitle>madvise</refentrytitle>
<manvolnum>2</manvolnum></citerefentry> is used instead. Setting this
to a positive value enables batched purging via process_madvise, which
can reduce the number of system calls needed for
purging.</para></listitem>
</varlistentry>
<varlistentry id="thread.arena">
<term>
<mallctl>thread.arena</mallctl>
@ -1735,6 +1850,47 @@ malloc_conf = "xmalloc:true";]]></programlisting>
the developer may find manual flushing useful.</para></listitem>
</varlistentry>
<varlistentry id="thread.tcache.max">
<term>
<mallctl>thread.tcache.max</mallctl>
(<type>size_t</type>)
<literal>rw</literal>
</term>
<listitem><para>Get or set the maximum cached size class
(<varname>tcache_max</varname>) for the calling thread's tcache. The
value is clamped to the maximum allowed limit and rounded up to the
nearest size class boundary. Changing this value will resize the
thread cache accordingly.</para></listitem>
</varlistentry>
<varlistentry id="thread.tcache.ncached_max.read_sizeclass">
<term>
<mallctl>thread.tcache.ncached_max.read_sizeclass</mallctl>
(<type>size_t</type>)
<literal>rw</literal>
</term>
<listitem><para>Query the maximum number of cached objects
(<varname>ncached_max</varname>) for a given size class in the calling
thread's tcache. The size class is passed in via
<parameter>newp</parameter>, and the corresponding
<varname>ncached_max</varname> is returned via
<parameter>oldp</parameter>.</para></listitem>
</varlistentry>
<varlistentry id="thread.tcache.ncached_max.write">
<term>
<mallctl>thread.tcache.ncached_max.write</mallctl>
(<type>char *</type>)
<literal>-w</literal>
</term>
<listitem><para>Set the maximum number of cached objects
(<varname>ncached_max</varname>) for size classes in the calling
thread's tcache. The input is a string of pipe-separated settings,
where each setting specifies a size range and a count, in the same
format as the <mallctl>opt.tcache_ncached_max</mallctl> runtime
option.</para></listitem>
</varlistentry>
<varlistentry id="thread.prof.name">
<term>
<mallctl>thread.prof.name</mallctl>
@ -1918,6 +2074,24 @@ malloc_conf = "xmalloc:true";]]></programlisting>
linkend="thread.arena"><mallctl>thread.arena</mallctl></link>.</para></listitem>
</varlistentry>
<varlistentry id="arena.i.name">
<term>
<mallctl>arena.&lt;i&gt;.name</mallctl>
(<type>char *</type>)
<literal>rw</literal>
</term>
<listitem><para>Get or set a descriptive name for arena &lt;i&gt;.
Arena names can be up to 32 characters long (including the null
terminator); longer names are truncated. When reading, the caller
passes a pointer to a pre-allocated buffer (of at least 32 bytes) via
<parameter>oldp</parameter>, and
<parameter>*oldlenp</parameter> must be
<code language="C">sizeof(<type>char *</type>)</code>.
Arena names are also included in the output of <link
linkend="stats_print"><function>malloc_stats_print()</function></link>.
</para></listitem>
</varlistentry>
<varlistentry id="arena.i.dss">
<term>
<mallctl>arena.&lt;i&gt;.dss</mallctl>
@ -2275,6 +2449,18 @@ struct extent_hooks_s {
<listitem><para>Page size.</para></listitem>
</varlistentry>
<varlistentry id="arenas.hugepage">
<term>
<mallctl>arenas.hugepage</mallctl>
(<type>size_t</type>)
<literal>r-</literal>
</term>
<listitem><para>Hugepage size. This value is also reported in the
output of <link
linkend="stats_print"><function>malloc_stats_print()</function></link>.
</para></listitem>
</varlistentry>
<varlistentry id="arenas.tcache_max">
<term>
<mallctl>arenas.tcache_max</mallctl>
@ -2494,6 +2680,24 @@ struct extent_hooks_s {
option for additional information.</para></listitem>
</varlistentry>
<varlistentry id="approximate_stats.active">
<term>
<mallctl>approximate_stats.active</mallctl>
(<type>size_t</type>)
<literal>r-</literal>
</term>
<listitem><para>Return the total number of bytes in active pages
collected in an unsynchronized manner, without requiring an
<link linkend="epoch"><mallctl>epoch</mallctl></link> update.
As a result, this value should NOT be compared with other
stats. For example, the relative ordering between
<mallctl>approximate_stats.active</mallctl> and <link
linkend="stats.active"><mallctl>stats.active</mallctl></link> or <link
linkend="stats.resident"><mallctl>stats.resident</mallctl></link> is
not guaranteed. This interface is intended for lightweight monitoring
where an approximate value is sufficient.</para></listitem>
</varlistentry>
<varlistentry id="stats.allocated">
<term>
<mallctl>stats.allocated</mallctl>
@ -3267,7 +3471,7 @@ struct extent_hooks_s {
<listitem><para>Current number of nonfull slabs.</para></listitem>
</varlistentry>
<varlistentry id="stats.arenas.i.bins.mutex">
<varlistentry id="stats.arenas.i.bins.j.mutex">
<term>
<mallctl>stats.arenas.&lt;i&gt;.bins.&lt;j&gt;.mutex.{counter}</mallctl>
(<type>counter specific type</type>) <literal>r-</literal>

20
doc_internal/PROFILING_INTERNALS.md
View file

@ -99,7 +99,25 @@ Using this approach means that there are a few things users need to be aware of.
If one stack appears twice as often as another, this by itself does not imply that it allocates twice as often. Consider the case in which there are only two types of allocating call stacks in a program. Stack A allocates 8 bytes, and occurs a million times in a program. Stack B allocates 8 MB, and occurs just once in a program. If our sampling rate $R$ is about 1MB, we expect stack A to show up about 8 times, and stack B to show up once. Stack A isn't 8 times more frequent than stack B, though; it's a million times more frequent.
### Aggregation must be done after unbiasing samples
Some tools manually parse heap dump output, and aggregate across stacks (or across program runs) to provide wider-scale data analyses. When doing this aggregation, though, it's important to unbias-and-then-sum, rather than sum-and-then-unbias. Reusing our example from the previous section: suppose we collect heap dumps of the program from a million machines. We then have 8 million occurs of stack A (each of 8 bytes), and a million occurrences of stack B (each of 8 MB). If we sum first, we'll attribute 64 MB to stack A, and 8 TB to stack B. Unbiasing changes these numbers by an infinitesimal amount, so that sum-then-unbias dramatically underreports the amount of memory allocated by stack A.
Some tools manually parse heap dump output, and aggregate across stacks (or across program runs) to provide wider-scale data analyses. When doing this aggregation, though, it's important to unbias-and-then-sum, rather than sum-and-then-unbias. Reusing our example from the previous section: suppose we collect heap dumps of the program from 1 million machines. We then have 8 million samples of stack A (8 per machine, each of 8 bytes), and 1 million samples of stack B (1 per machine, each of 8 MB).
If we sum first then unbias based on this formula: $1 - e^{-Z/R}$ we get:
$$Z = 8,000,000 * 8 bytes = 64MB$$
$$64MB / (1 - e^{-64MB/1MB}) \approx 64MB (Stack A)$$
$$Z = 1,000,000 * 8MB = 8TB$$
$$8TB / (1 - e^{-1TB/1MB}) \approx 8TB (Stack B)$$
Clearly we are unbiasing by an infinitesimal amount, which dramatically underreports the amount of memory allocated by stack A. Whereas if we unbias first and then sum:
$$Z = 8 bytes$$
$$8 bytes / (1 - e^{-8 bytes/1MB}) \approx 1MB$$
$$1MB * 8,000,000 = 8TB (Stack A)$$
$$Z = 8MB$$
$$8MB / (1 - e^{-8MB/1MB}) \approx 8MB$$
$$8MB * 1,000,000 = 8TB (Stack B)$$
## An avenue for future exploration
While the framework we laid out above is pretty general, as an engineering decision we're only interested in fairly simple approaches (i.e. ones for which the chance of an allocation being sampled depends only on its size). Our job is then: for each size class $Z$, pick a probability $p_Z$ that an allocation of that size will be sampled. We made some handwave-y references to statistical distributions to justify our choices, but there's no reason we need to pick them that way. Any set of non-zero probabilities is a valid choice.

23
include/jemalloc/internal/activity_callback.h
View file

@ -1,23 +0,0 @@
#ifndef JEMALLOC_INTERNAL_ACTIVITY_CALLBACK_H
#define JEMALLOC_INTERNAL_ACTIVITY_CALLBACK_H
/*
* The callback to be executed "periodically", in response to some amount of
* allocator activity.
*
* This callback need not be computing any sort of peak (although that's the
* intended first use case), but we drive it from the peak counter, so it's
* keeps things tidy to keep it here.
*
* The calls to this thunk get driven by the peak_event module.
*/
#define ACTIVITY_CALLBACK_THUNK_INITIALIZER {NULL, NULL}
typedef void (*activity_callback_t)(void *uctx, uint64_t allocated,
uint64_t deallocated);
typedef struct activity_callback_thunk_s activity_callback_thunk_t;
struct activity_callback_thunk_s {
activity_callback_t callback;
void *uctx;
};
#endif /* JEMALLOC_INTERNAL_ACTIVITY_CALLBACK_H */

140
include/jemalloc/internal/arena_externs.h
View file

@ -1,8 +1,11 @@
#ifndef JEMALLOC_INTERNAL_ARENA_EXTERNS_H
#define JEMALLOC_INTERNAL_ARENA_EXTERNS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_stats.h"
#include "jemalloc/internal/bin.h"
#include "jemalloc/internal/div.h"
#include "jemalloc/internal/emap.h"
#include "jemalloc/internal/extent_dss.h"
#include "jemalloc/internal/hook.h"
#include "jemalloc/internal/pages.h"
@ -18,104 +21,105 @@ extern ssize_t opt_dirty_decay_ms;
extern ssize_t opt_muzzy_decay_ms;
extern percpu_arena_mode_t opt_percpu_arena;
extern const char *percpu_arena_mode_names[];
extern const char *const percpu_arena_mode_names[];
extern div_info_t arena_binind_div_info[SC_NBINS];
extern malloc_mutex_t arenas_lock;
extern emap_t arena_emap_global;
extern size_t opt_oversize_threshold;
extern size_t oversize_threshold;
extern bool opt_huge_arena_pac_thp;
extern pac_thp_t huge_arena_pac_thp;
/*
* arena_bin_offsets[binind] is the offset of the first bin shard for size class
* binind.
*/
extern uint32_t arena_bin_offsets[SC_NBINS];
void arena_basic_stats_merge(tsdn_t *tsdn, arena_t *arena,
unsigned *nthreads, const char **dss, ssize_t *dirty_decay_ms,
ssize_t *muzzy_decay_ms, size_t *nactive, size_t *ndirty, size_t *nmuzzy);
void arena_basic_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
const char **dss, ssize_t *dirty_decay_ms, ssize_t *muzzy_decay_ms,
size_t *nactive, size_t *ndirty, size_t *nmuzzy);
void arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
const char **dss, ssize_t *dirty_decay_ms, ssize_t *muzzy_decay_ms,
size_t *nactive, size_t *ndirty, size_t *nmuzzy, arena_stats_t *astats,
bin_stats_data_t *bstats, arena_stats_large_t *lstats,
pac_estats_t *estats, hpa_shard_stats_t *hpastats, sec_stats_t *secstats);
bin_stats_data_t *bstats, arena_stats_large_t *lstats, pac_estats_t *estats,
hpa_shard_stats_t *hpastats);
void arena_handle_deferred_work(tsdn_t *tsdn, arena_t *arena);
edata_t *arena_extent_alloc_large(tsdn_t *tsdn, arena_t *arena,
size_t usize, size_t alignment, bool zero);
void arena_extent_dalloc_large_prep(tsdn_t *tsdn, arena_t *arena,
edata_t *edata);
void arena_extent_ralloc_large_shrink(tsdn_t *tsdn, arena_t *arena,
edata_t *edata, size_t oldsize);
void arena_extent_ralloc_large_expand(tsdn_t *tsdn, arena_t *arena,
edata_t *edata, size_t oldsize);
bool arena_decay_ms_set(tsdn_t *tsdn, arena_t *arena, extent_state_t state,
ssize_t decay_ms);
edata_t *arena_extent_alloc_large(
tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment, bool zero);
void arena_extent_dalloc_large_prep(
tsdn_t *tsdn, arena_t *arena, edata_t *edata);
void arena_extent_ralloc_large_shrink(
tsdn_t *tsdn, arena_t *arena, edata_t *edata, size_t oldusize);
void arena_extent_ralloc_large_expand(
tsdn_t *tsdn, arena_t *arena, edata_t *edata, size_t oldusize);
bool arena_decay_ms_set(
tsdn_t *tsdn, arena_t *arena, extent_state_t state, ssize_t decay_ms);
ssize_t arena_decay_ms_get(arena_t *arena, extent_state_t state);
void arena_decay(tsdn_t *tsdn, arena_t *arena, bool is_background_thread,
bool all);
uint64_t arena_time_until_deferred(tsdn_t *tsdn, arena_t *arena);
void arena_do_deferred_work(tsdn_t *tsdn, arena_t *arena);
void arena_reset(tsd_t *tsd, arena_t *arena);
void arena_destroy(tsd_t *tsd, arena_t *arena);
void arena_cache_bin_fill_small(tsdn_t *tsdn, arena_t *arena,
cache_bin_t *cache_bin, cache_bin_info_t *cache_bin_info, szind_t binind,
const unsigned nfill);
void arena_decay(
tsdn_t *tsdn, arena_t *arena, bool is_background_thread, bool all);
uint64_t arena_time_until_deferred(tsdn_t *tsdn, arena_t *arena);
void arena_do_deferred_work(tsdn_t *tsdn, arena_t *arena);
void arena_reset(tsd_t *tsd, arena_t *arena);
void arena_destroy(tsd_t *tsd, arena_t *arena);
cache_bin_sz_t arena_ptr_array_fill_small(tsdn_t *tsdn, arena_t *arena,
szind_t binind, cache_bin_ptr_array_t *arr, const cache_bin_sz_t nfill_min,
const cache_bin_sz_t nfill_max, cache_bin_stats_t merge_stats);
void *arena_malloc_hard(tsdn_t *tsdn, arena_t *arena, size_t size,
szind_t ind, bool zero);
void *arena_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize,
size_t alignment, bool zero, tcache_t *tcache);
void arena_prof_promote(tsdn_t *tsdn, void *ptr, size_t usize);
void arena_dalloc_promoted(tsdn_t *tsdn, void *ptr, tcache_t *tcache,
bool slow_path);
void *arena_malloc_hard(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind,
bool zero, bool slab);
void *arena_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
bool zero, bool slab, tcache_t *tcache);
void arena_prof_promote(
tsdn_t *tsdn, void *ptr, size_t usize, size_t bumped_usize);
void arena_dalloc_promoted(
tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool slow_path);
void arena_slab_dalloc(tsdn_t *tsdn, arena_t *arena, edata_t *slab);
void arena_dalloc_bin_locked_handle_newly_empty(tsdn_t *tsdn, arena_t *arena,
edata_t *slab, bin_t *bin);
void arena_dalloc_bin_locked_handle_newly_nonempty(tsdn_t *tsdn, arena_t *arena,
edata_t *slab, bin_t *bin);
void arena_dalloc_small(tsdn_t *tsdn, void *ptr);
bool arena_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
size_t extra, bool zero, size_t *newsize);
void arena_dalloc_small(tsdn_t *tsdn, void *ptr);
void arena_ptr_array_flush(tsd_t *tsd, szind_t binind,
cache_bin_ptr_array_t *arr, unsigned nflush, bool small,
arena_t *stats_arena, cache_bin_stats_t merge_stats);
bool arena_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
size_t extra, bool zero, size_t *newsize);
void *arena_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t oldsize,
size_t size, size_t alignment, bool zero, tcache_t *tcache,
size_t size, size_t alignment, bool zero, bool slab, tcache_t *tcache,
hook_ralloc_args_t *hook_args);
dss_prec_t arena_dss_prec_get(arena_t *arena);
ehooks_t *arena_get_ehooks(arena_t *arena);
extent_hooks_t *arena_set_extent_hooks(tsd_t *tsd, arena_t *arena,
extent_hooks_t *extent_hooks);
bool arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);
dss_prec_t arena_dss_prec_get(arena_t *arena);
ehooks_t *arena_get_ehooks(arena_t *arena);
extent_hooks_t *arena_set_extent_hooks(
tsd_t *tsd, arena_t *arena, extent_hooks_t *extent_hooks);
bool arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);
void arena_name_get(arena_t *arena, char *name);
void arena_name_set(arena_t *arena, const char *name);
ssize_t arena_dirty_decay_ms_default_get(void);
bool arena_dirty_decay_ms_default_set(ssize_t decay_ms);
bool arena_dirty_decay_ms_default_set(ssize_t decay_ms);
ssize_t arena_muzzy_decay_ms_default_get(void);
bool arena_muzzy_decay_ms_default_set(ssize_t decay_ms);
bool arena_retain_grow_limit_get_set(tsd_t *tsd, arena_t *arena,
size_t *old_limit, size_t *new_limit);
bool arena_muzzy_decay_ms_default_set(ssize_t decay_ms);
bool arena_retain_grow_limit_get_set(
tsd_t *tsd, arena_t *arena, size_t *old_limit, size_t *new_limit);
unsigned arena_nthreads_get(arena_t *arena, bool internal);
void arena_nthreads_inc(arena_t *arena, bool internal);
void arena_nthreads_dec(arena_t *arena, bool internal);
void arena_nthreads_inc(arena_t *arena, bool internal);
void arena_nthreads_dec(arena_t *arena, bool internal);
arena_t *arena_new(tsdn_t *tsdn, unsigned ind, const arena_config_t *config);
bool arena_init_huge(void);
bool arena_is_huge(unsigned arena_ind);
bool arena_init_huge(tsdn_t *tsdn, arena_t *a0);
arena_t *arena_choose_huge(tsd_t *tsd);
bin_t *arena_bin_choose(tsdn_t *tsdn, arena_t *arena, szind_t binind,
unsigned *binshard);
size_t arena_fill_small_fresh(tsdn_t *tsdn, arena_t *arena, szind_t binind,
void **ptrs, size_t nfill, bool zero);
bool arena_boot(sc_data_t *sc_data, base_t *base, bool hpa);
void arena_prefork0(tsdn_t *tsdn, arena_t *arena);
void arena_prefork1(tsdn_t *tsdn, arena_t *arena);
void arena_prefork2(tsdn_t *tsdn, arena_t *arena);
void arena_prefork3(tsdn_t *tsdn, arena_t *arena);
void arena_prefork4(tsdn_t *tsdn, arena_t *arena);
void arena_prefork5(tsdn_t *tsdn, arena_t *arena);
void arena_prefork6(tsdn_t *tsdn, arena_t *arena);
void arena_prefork7(tsdn_t *tsdn, arena_t *arena);
void arena_prefork8(tsdn_t *tsdn, arena_t *arena);
void arena_postfork_parent(tsdn_t *tsdn, arena_t *arena);
void arena_postfork_child(tsdn_t *tsdn, arena_t *arena);
bool arena_boot(sc_data_t *sc_data, base_t *base, bool hpa);
void arena_prefork0(tsdn_t *tsdn, arena_t *arena);
void arena_prefork1(tsdn_t *tsdn, arena_t *arena);
void arena_prefork2(tsdn_t *tsdn, arena_t *arena);
void arena_prefork3(tsdn_t *tsdn, arena_t *arena);
void arena_prefork4(tsdn_t *tsdn, arena_t *arena);
void arena_prefork5(tsdn_t *tsdn, arena_t *arena);
void arena_prefork6(tsdn_t *tsdn, arena_t *arena);
void arena_prefork7(tsdn_t *tsdn, arena_t *arena);
void arena_prefork8(tsdn_t *tsdn, arena_t *arena);
void arena_postfork_parent(tsdn_t *tsdn, arena_t *arena);
void arena_postfork_child(tsdn_t *tsdn, arena_t *arena);
#endif /* JEMALLOC_INTERNAL_ARENA_EXTERNS_H */

3
include/jemalloc/internal/arena_inlines_a.h
View file

@ -1,6 +1,9 @@
#ifndef JEMALLOC_INTERNAL_ARENA_INLINES_A_H
#define JEMALLOC_INTERNAL_ARENA_INLINES_A_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_structs.h"
static inline unsigned
arena_ind_get(const arena_t *arena) {
return arena->ind;

422
include/jemalloc/internal/arena_inlines_b.h
View file

@ -1,20 +1,29 @@
#ifndef JEMALLOC_INTERNAL_ARENA_INLINES_B_H
#define JEMALLOC_INTERNAL_ARENA_INLINES_B_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_externs.h"
#include "jemalloc/internal/arena_structs.h"
#include "jemalloc/internal/bin_inlines.h"
#include "jemalloc/internal/div.h"
#include "jemalloc/internal/emap.h"
#include "jemalloc/internal/jemalloc_internal_inlines_b.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#include "jemalloc/internal/large_externs.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/prof_externs.h"
#include "jemalloc/internal/prof_structs.h"
#include "jemalloc/internal/rtree.h"
#include "jemalloc/internal/safety_check.h"
#include "jemalloc/internal/sc.h"
#include "jemalloc/internal/sz.h"
#include "jemalloc/internal/tcache_inlines.h"
#include "jemalloc/internal/ticker.h"
static inline arena_t *
arena_get_from_edata(edata_t *edata) {
return (arena_t *)atomic_load_p(&arenas[edata_arena_ind_get(edata)],
ATOMIC_RELAXED);
return (arena_t *)atomic_load_p(
&arenas[edata_arena_ind_get(edata)], ATOMIC_RELAXED);
}
JEMALLOC_ALWAYS_INLINE arena_t *
@ -28,14 +37,48 @@ arena_choose_maybe_huge(tsd_t *tsd, arena_t *arena, size_t size) {
* 1) is using auto arena selection (i.e. arena == NULL), and 2) the
* thread is not assigned to a manual arena.
*/
if (unlikely(size >= oversize_threshold)) {
arena_t *tsd_arena = tsd_arena_get(tsd);
if (tsd_arena == NULL || arena_is_auto(tsd_arena)) {
return arena_choose_huge(tsd);
}
arena_t *tsd_arena = tsd_arena_get(tsd);
if (tsd_arena == NULL) {
tsd_arena = arena_choose(tsd, NULL);
}
return arena_choose(tsd, NULL);
size_t threshold = atomic_load_zu(
&tsd_arena->pa_shard.pac.oversize_threshold, ATOMIC_RELAXED);
if (unlikely(size >= threshold) && arena_is_auto(tsd_arena)) {
return arena_choose_huge(tsd);
}
return tsd_arena;
}
JEMALLOC_ALWAYS_INLINE bool
large_dalloc_safety_checks(edata_t *edata, const void *ptr, size_t input_size) {
if (!config_opt_safety_checks) {
return false;
}
/*
* Eagerly detect double free and sized dealloc bugs for large sizes.
* The cost is low enough (as edata will be accessed anyway) to be
* enabled all the time.
*/
if (unlikely(edata == NULL
|| edata_state_get(edata) != extent_state_active)) {
safety_check_fail(
"Invalid deallocation detected: "
"pages being freed (%p) not currently active, "
"possibly caused by double free bugs.",
ptr);
return true;
}
if (unlikely(input_size != edata_usize_get(edata)
|| input_size > SC_LARGE_MAXCLASS)) {
safety_check_fail_sized_dealloc(/* current_dealloc */ true, ptr,
/* true_size */ edata_usize_get(edata), input_size);
return true;
}
return false;
}
JEMALLOC_ALWAYS_INLINE void
@ -46,48 +89,56 @@ arena_prof_info_get(tsd_t *tsd, const void *ptr, emap_alloc_ctx_t *alloc_ctx,
assert(prof_info != NULL);
edata_t *edata = NULL;
bool is_slab;
bool is_slab;
/* Static check. */
if (alloc_ctx == NULL) {
edata = emap_edata_lookup(tsd_tsdn(tsd), &arena_emap_global,
ptr);
edata = emap_edata_lookup(
tsd_tsdn(tsd), &arena_emap_global, ptr);
is_slab = edata_slab_get(edata);
} else if (unlikely(!(is_slab = alloc_ctx->slab))) {
edata = emap_edata_lookup(tsd_tsdn(tsd), &arena_emap_global,
ptr);
edata = emap_edata_lookup(
tsd_tsdn(tsd), &arena_emap_global, ptr);
}
if (unlikely(!is_slab)) {
/* edata must have been initialized at this point. */
assert(edata != NULL);
size_t usize = (alloc_ctx == NULL)
? edata_usize_get(edata)
: emap_alloc_ctx_usize_get(alloc_ctx);
if (reset_recent
&& large_dalloc_safety_checks(edata, ptr, usize)) {
prof_info->alloc_tctx = PROF_TCTX_SENTINEL;
return;
}
large_prof_info_get(tsd, edata, prof_info, reset_recent);
} else {
prof_info->alloc_tctx = (prof_tctx_t *)(uintptr_t)1U;
prof_info->alloc_tctx = PROF_TCTX_SENTINEL;
/*
* No need to set other fields in prof_info; they will never be
* accessed if (uintptr_t)alloc_tctx == (uintptr_t)1U.
* accessed if alloc_tctx == PROF_TCTX_SENTINEL.
*/
}
}
JEMALLOC_ALWAYS_INLINE void
arena_prof_tctx_reset(tsd_t *tsd, const void *ptr,
emap_alloc_ctx_t *alloc_ctx) {
arena_prof_tctx_reset(
tsd_t *tsd, const void *ptr, emap_alloc_ctx_t *alloc_ctx) {
cassert(config_prof);
assert(ptr != NULL);
/* Static check. */
if (alloc_ctx == NULL) {
edata_t *edata = emap_edata_lookup(tsd_tsdn(tsd),
&arena_emap_global, ptr);
edata_t *edata = emap_edata_lookup(
tsd_tsdn(tsd), &arena_emap_global, ptr);
if (unlikely(!edata_slab_get(edata))) {
large_prof_tctx_reset(edata);
}
} else {
if (unlikely(!alloc_ctx->slab)) {
edata_t *edata = emap_edata_lookup(tsd_tsdn(tsd),
&arena_emap_global, ptr);
edata_t *edata = emap_edata_lookup(
tsd_tsdn(tsd), &arena_emap_global, ptr);
large_prof_tctx_reset(edata);
}
}
@ -98,16 +149,16 @@ arena_prof_tctx_reset_sampled(tsd_t *tsd, const void *ptr) {
cassert(config_prof);
assert(ptr != NULL);
edata_t *edata = emap_edata_lookup(tsd_tsdn(tsd), &arena_emap_global,
ptr);
edata_t *edata = emap_edata_lookup(
tsd_tsdn(tsd), &arena_emap_global, ptr);
assert(!edata_slab_get(edata));
large_prof_tctx_reset(edata);
}
JEMALLOC_ALWAYS_INLINE void
arena_prof_info_set(tsd_t *tsd, edata_t *edata, prof_tctx_t *tctx,
size_t size) {
arena_prof_info_set(
tsd_t *tsd, edata_t *edata, prof_tctx_t *tctx, size_t size) {
cassert(config_prof);
assert(!edata_slab_get(edata));
@ -130,8 +181,9 @@ arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks) {
* use a single ticker for all of them.
*/
ticker_geom_t *decay_ticker = tsd_arena_decay_tickerp_get(tsd);
uint64_t *prng_state = tsd_prng_statep_get(tsd);
if (unlikely(ticker_geom_ticks(decay_ticker, prng_state, nticks))) {
uint64_t *prng_state = tsd_prng_statep_get(tsd);
if (unlikely(ticker_geom_ticks(decay_ticker, prng_state, nticks,
tsd_reentrancy_level_get(tsd) > 0))) {
arena_decay(tsdn, arena, false, false);
}
}
@ -143,23 +195,24 @@ arena_decay_tick(tsdn_t *tsdn, arena_t *arena) {
JEMALLOC_ALWAYS_INLINE void *
arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind, bool zero,
tcache_t *tcache, bool slow_path) {
bool slab, tcache_t *tcache, bool slow_path) {
assert(!tsdn_null(tsdn) || tcache == NULL);
if (likely(tcache != NULL)) {
if (likely(size <= SC_SMALL_MAXCLASS)) {
return tcache_alloc_small(tsdn_tsd(tsdn), arena,
tcache, size, ind, zero, slow_path);
if (likely(slab)) {
assert(sz_can_use_slab(size));
return tcache_alloc_small(tsdn_tsd(tsdn), arena, tcache,
size, ind, zero, slow_path);
} else if (likely(ind < tcache_nbins_get(tcache->tcache_slow)
&& !tcache_bin_disabled(ind, &tcache->bins[ind],
tcache->tcache_slow))) {
return tcache_alloc_large(tsdn_tsd(tsdn), arena, tcache,
size, ind, zero, slow_path);
}
if (likely(size <= tcache_maxclass)) {
return tcache_alloc_large(tsdn_tsd(tsdn), arena,
tcache, size, ind, zero, slow_path);
}
/* (size > tcache_maxclass) case falls through. */
assert(size > tcache_maxclass);
/* (size > tcache_max) case falls through. */
}
return arena_malloc_hard(tsdn, arena, size, ind, zero);
return arena_malloc_hard(tsdn, arena, size, ind, zero, slab);
}
JEMALLOC_ALWAYS_INLINE arena_t *
@ -176,7 +229,7 @@ arena_salloc(tsdn_t *tsdn, const void *ptr) {
emap_alloc_ctx_lookup(tsdn, &arena_emap_global, ptr, &alloc_ctx);
assert(alloc_ctx.szind != SC_NSIZES);
return sz_index2size(alloc_ctx.szind);
return emap_alloc_ctx_usize_get(&alloc_ctx);
}
JEMALLOC_ALWAYS_INLINE size_t
@ -191,8 +244,8 @@ arena_vsalloc(tsdn_t *tsdn, const void *ptr) {
*/
emap_full_alloc_ctx_t full_alloc_ctx;
bool missing = emap_full_alloc_ctx_try_lookup(tsdn, &arena_emap_global,
ptr, &full_alloc_ctx);
bool missing = emap_full_alloc_ctx_try_lookup(
tsdn, &arena_emap_global, ptr, &full_alloc_ctx);
if (missing) {
return 0;
}
@ -207,46 +260,24 @@ arena_vsalloc(tsdn_t *tsdn, const void *ptr) {
assert(full_alloc_ctx.szind != SC_NSIZES);
return sz_index2size(full_alloc_ctx.szind);
}
JEMALLOC_ALWAYS_INLINE bool
large_dalloc_safety_checks(edata_t *edata, void *ptr, szind_t szind) {
if (!config_opt_safety_checks) {
return false;
}
/*
* Eagerly detect double free and sized dealloc bugs for large sizes.
* The cost is low enough (as edata will be accessed anyway) to be
* enabled all the time.
*/
if (unlikely(edata == NULL ||
edata_state_get(edata) != extent_state_active)) {
safety_check_fail("Invalid deallocation detected: "
"pages being freed (%p) not currently active, "
"possibly caused by double free bugs.",
(uintptr_t)edata_addr_get(edata));
return true;
}
size_t input_size = sz_index2size(szind);
if (unlikely(input_size != edata_usize_get(edata))) {
safety_check_fail_sized_dealloc(/* current_dealloc */ true, ptr,
/* true_size */ edata_usize_get(edata), input_size);
return true;
}
return false;
return edata_usize_get(full_alloc_ctx.edata);
}
static inline void
arena_dalloc_large_no_tcache(tsdn_t *tsdn, void *ptr, szind_t szind) {
arena_dalloc_large_no_tcache(
tsdn_t *tsdn, void *ptr, szind_t szind, size_t usize) {
/*
* szind is still needed in this function mainly becuase
* szind < SC_NBINS determines not only if this is a small alloc,
* but also if szind is valid (an inactive extent would have
* szind == SC_NSIZES).
*/
if (config_prof && unlikely(szind < SC_NBINS)) {
arena_dalloc_promoted(tsdn, ptr, NULL, true);
} else {
edata_t *edata = emap_edata_lookup(tsdn, &arena_emap_global,
ptr);
if (large_dalloc_safety_checks(edata, ptr, szind)) {
edata_t *edata = emap_edata_lookup(
tsdn, &arena_emap_global, ptr);
if (large_dalloc_safety_checks(edata, ptr, usize)) {
/* See the comment in isfree. */
return;
}
@ -262,42 +293,76 @@ arena_dalloc_no_tcache(tsdn_t *tsdn, void *ptr) {
emap_alloc_ctx_lookup(tsdn, &arena_emap_global, ptr, &alloc_ctx);
if (config_debug) {
edata_t *edata = emap_edata_lookup(tsdn, &arena_emap_global,
ptr);
edata_t *edata = emap_edata_lookup(
tsdn, &arena_emap_global, ptr);
assert(alloc_ctx.szind == edata_szind_get(edata));
assert(alloc_ctx.szind < SC_NSIZES);
assert(alloc_ctx.slab == edata_slab_get(edata));
assert(emap_alloc_ctx_usize_get(&alloc_ctx)
== edata_usize_get(edata));
}
if (likely(alloc_ctx.slab)) {
/* Small allocation. */
arena_dalloc_small(tsdn, ptr);
} else {
arena_dalloc_large_no_tcache(tsdn, ptr, alloc_ctx.szind);
arena_dalloc_large_no_tcache(tsdn, ptr, alloc_ctx.szind,
emap_alloc_ctx_usize_get(&alloc_ctx));
}
}
JEMALLOC_ALWAYS_INLINE void
arena_dalloc_large(tsdn_t *tsdn, void *ptr, tcache_t *tcache, szind_t szind,
bool slow_path) {
if (szind < nhbins) {
if (config_prof && unlikely(szind < SC_NBINS)) {
arena_dalloc_promoted(tsdn, ptr, tcache, slow_path);
} else {
tcache_dalloc_large(tsdn_tsd(tsdn), tcache, ptr, szind,
slow_path);
}
size_t usize, bool slow_path) {
assert(!tsdn_null(tsdn) && tcache != NULL);
bool is_sample_promoted = config_prof && szind < SC_NBINS;
if (unlikely(is_sample_promoted)) {
arena_dalloc_promoted(tsdn, ptr, tcache, slow_path);
} else {
edata_t *edata = emap_edata_lookup(tsdn, &arena_emap_global,
ptr);
if (large_dalloc_safety_checks(edata, ptr, szind)) {
/* See the comment in isfree. */
return;
if (szind < tcache_nbins_get(tcache->tcache_slow)
&& !tcache_bin_disabled(
szind, &tcache->bins[szind], tcache->tcache_slow)) {
tcache_dalloc_large(
tsdn_tsd(tsdn), tcache, ptr, szind, slow_path);
} else {
edata_t *edata = emap_edata_lookup(
tsdn, &arena_emap_global, ptr);
if (large_dalloc_safety_checks(edata, ptr, usize)) {
/* See the comment in isfree. */
return;
}
large_dalloc(tsdn, edata);
}
large_dalloc(tsdn, edata);
}
}
JEMALLOC_ALWAYS_INLINE bool
arena_tcache_dalloc_small_safety_check(tsdn_t *tsdn, void *ptr) {
if (!config_debug) {
return false;
}
edata_t *edata = emap_edata_lookup(tsdn, &arena_emap_global, ptr);
szind_t binind = edata_szind_get(edata);
div_info_t div_info = arena_binind_div_info[binind];
/*
* Calls the internal function bin_slab_regind_impl because the
* safety check does not require a lock.
*/
size_t regind = bin_slab_regind_impl(&div_info, binind, edata, ptr);
slab_data_t *slab_data = edata_slab_data_get(edata);
const bin_info_t *bin_info = &bin_infos[binind];
assert(edata_nfree_get(edata) < bin_info->nregs);
if (unlikely(!bitmap_get(
slab_data->bitmap, &bin_info->bitmap_info, regind))) {
safety_check_fail(
"Invalid deallocation detected: the pointer being freed (%p) not "
"currently active, possibly caused by double free bugs.\n",
ptr);
return true;
}
return false;
}
JEMALLOC_ALWAYS_INLINE void
arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache,
emap_alloc_ctx_t *caller_alloc_ctx, bool slow_path) {
@ -313,26 +378,31 @@ arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache,
if (caller_alloc_ctx != NULL) {
alloc_ctx = *caller_alloc_ctx;
} else {
util_assume(!tsdn_null(tsdn));
emap_alloc_ctx_lookup(tsdn, &arena_emap_global, ptr,
&alloc_ctx);
util_assume(tsdn != NULL);
emap_alloc_ctx_lookup(
tsdn, &arena_emap_global, ptr, &alloc_ctx);
}
if (config_debug) {
edata_t *edata = emap_edata_lookup(tsdn, &arena_emap_global,
ptr);
edata_t *edata = emap_edata_lookup(
tsdn, &arena_emap_global, ptr);
assert(alloc_ctx.szind == edata_szind_get(edata));
assert(alloc_ctx.szind < SC_NSIZES);
assert(alloc_ctx.slab == edata_slab_get(edata));
assert(emap_alloc_ctx_usize_get(&alloc_ctx)
== edata_usize_get(edata));
}
if (likely(alloc_ctx.slab)) {
/* Small allocation. */
tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr,
alloc_ctx.szind, slow_path);
if (arena_tcache_dalloc_small_safety_check(tsdn, ptr)) {
return;
}
tcache_dalloc_small(
tsdn_tsd(tsdn), tcache, ptr, alloc_ctx.szind, slow_path);
} else {
arena_dalloc_large(tsdn, ptr, tcache, alloc_ctx.szind,
slow_path);
emap_alloc_ctx_usize_get(&alloc_ctx), slow_path);
}
}
@ -347,21 +417,22 @@ arena_sdalloc_no_tcache(tsdn_t *tsdn, void *ptr, size_t size) {
* There is no risk of being confused by a promoted sampled
* object, so base szind and slab on the given size.
*/
alloc_ctx.szind = sz_size2index(size);
alloc_ctx.slab = (alloc_ctx.szind < SC_NBINS);
szind_t szind = sz_size2index(size);
emap_alloc_ctx_init(
&alloc_ctx, szind, (szind < SC_NBINS), size);
}
if ((config_prof && opt_prof) || config_debug) {
emap_alloc_ctx_lookup(tsdn, &arena_emap_global, ptr,
&alloc_ctx);
emap_alloc_ctx_lookup(
tsdn, &arena_emap_global, ptr, &alloc_ctx);
assert(alloc_ctx.szind == sz_size2index(size));
assert((config_prof && opt_prof)
|| alloc_ctx.slab == (alloc_ctx.szind < SC_NBINS));
if (config_debug) {
edata_t *edata = emap_edata_lookup(tsdn,
&arena_emap_global, ptr);
edata_t *edata = emap_edata_lookup(
tsdn, &arena_emap_global, ptr);
assert(alloc_ctx.szind == edata_szind_get(edata));
assert(alloc_ctx.slab == edata_slab_get(edata));
}
@ -371,7 +442,8 @@ arena_sdalloc_no_tcache(tsdn_t *tsdn, void *ptr, size_t size) {
/* Small allocation. */
arena_dalloc_small(tsdn, ptr);
} else {
arena_dalloc_large_no_tcache(tsdn, ptr, alloc_ctx.szind);
arena_dalloc_large_no_tcache(tsdn, ptr, alloc_ctx.szind,
emap_alloc_ctx_usize_get(&alloc_ctx));
}
}
@ -391,9 +463,10 @@ arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
if (config_prof && opt_prof) {
if (caller_alloc_ctx == NULL) {
/* Uncommon case and should be a static check. */
emap_alloc_ctx_lookup(tsdn, &arena_emap_global, ptr,
&alloc_ctx);
emap_alloc_ctx_lookup(
tsdn, &arena_emap_global, ptr, &alloc_ctx);
assert(alloc_ctx.szind == sz_size2index(size));
assert(emap_alloc_ctx_usize_get(&alloc_ctx) == size);
} else {
alloc_ctx = *caller_alloc_ctx;
}
@ -407,30 +480,37 @@ arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
}
if (config_debug) {
edata_t *edata = emap_edata_lookup(tsdn, &arena_emap_global,
ptr);
edata_t *edata = emap_edata_lookup(
tsdn, &arena_emap_global, ptr);
assert(alloc_ctx.szind == edata_szind_get(edata));
assert(alloc_ctx.slab == edata_slab_get(edata));
emap_alloc_ctx_init(
&alloc_ctx, alloc_ctx.szind, alloc_ctx.slab, sz_s2u(size));
assert(emap_alloc_ctx_usize_get(&alloc_ctx)
== edata_usize_get(edata));
}
if (likely(alloc_ctx.slab)) {
/* Small allocation. */
tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr,
alloc_ctx.szind, slow_path);
if (arena_tcache_dalloc_small_safety_check(tsdn, ptr)) {
return;
}
tcache_dalloc_small(
tsdn_tsd(tsdn), tcache, ptr, alloc_ctx.szind, slow_path);
} else {
arena_dalloc_large(tsdn, ptr, tcache, alloc_ctx.szind,
slow_path);
sz_s2u(size), slow_path);
}
}
static inline void
arena_cache_oblivious_randomize(tsdn_t *tsdn, arena_t *arena, edata_t *edata,
size_t alignment) {
arena_cache_oblivious_randomize(
tsdn_t *tsdn, arena_t *arena, edata_t *edata, size_t alignment) {
assert(edata_base_get(edata) == edata_addr_get(edata));
if (alignment < PAGE) {
unsigned lg_range = LG_PAGE -
lg_floor(CACHELINE_CEILING(alignment));
unsigned lg_range = LG_PAGE
- lg_floor(CACHELINE_CEILING(alignment));
size_t r;
if (!tsdn_null(tsdn)) {
tsd_t *tsd = tsdn_tsd(tsdn);
@ -440,110 +520,18 @@ arena_cache_oblivious_randomize(tsdn_t *tsdn, arena_t *arena, edata_t *edata,
uint64_t stack_value = (uint64_t)(uintptr_t)&r;
r = (size_t)prng_lg_range_u64(&stack_value, lg_range);
}
uintptr_t random_offset = ((uintptr_t)r) << (LG_PAGE -
lg_range);
edata->e_addr = (void *)((uintptr_t)edata->e_addr +
random_offset);
assert(ALIGNMENT_ADDR2BASE(edata->e_addr, alignment) ==
edata->e_addr);
}
}
/*
* The dalloc bin info contains just the information that the common paths need
* during tcache flushes. By force-inlining these paths, and using local copies
* of data (so that the compiler knows it's constant), we avoid a whole bunch of
* redundant loads and stores by leaving this information in registers.
*/
typedef struct arena_dalloc_bin_locked_info_s arena_dalloc_bin_locked_info_t;
struct arena_dalloc_bin_locked_info_s {
div_info_t div_info;
uint32_t nregs;
uint64_t ndalloc;
};
JEMALLOC_ALWAYS_INLINE size_t
arena_slab_regind(arena_dalloc_bin_locked_info_t *info, szind_t binind,
edata_t *slab, const void *ptr) {
size_t diff, regind;
/* Freeing a pointer outside the slab can cause assertion failure. */
assert((uintptr_t)ptr >= (uintptr_t)edata_addr_get(slab));
assert((uintptr_t)ptr < (uintptr_t)edata_past_get(slab));
/* Freeing an interior pointer can cause assertion failure. */
assert(((uintptr_t)ptr - (uintptr_t)edata_addr_get(slab)) %
(uintptr_t)bin_infos[binind].reg_size == 0);
diff = (size_t)((uintptr_t)ptr - (uintptr_t)edata_addr_get(slab));
/* Avoid doing division with a variable divisor. */
regind = div_compute(&info->div_info, diff);
assert(regind < bin_infos[binind].nregs);
return regind;
}
JEMALLOC_ALWAYS_INLINE void
arena_dalloc_bin_locked_begin(arena_dalloc_bin_locked_info_t *info,
szind_t binind) {
info->div_info = arena_binind_div_info[binind];
info->nregs = bin_infos[binind].nregs;
info->ndalloc = 0;
}
/*
* Does the deallocation work associated with freeing a single pointer (a
* "step") in between a arena_dalloc_bin_locked begin and end call.
*
* Returns true if arena_slab_dalloc must be called on slab. Doesn't do
* stats updates, which happen during finish (this lets running counts get left
* in a register).
*/
JEMALLOC_ALWAYS_INLINE bool
arena_dalloc_bin_locked_step(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
arena_dalloc_bin_locked_info_t *info, szind_t binind, edata_t *slab,
void *ptr) {
const bin_info_t *bin_info = &bin_infos[binind];
size_t regind = arena_slab_regind(info, binind, slab, ptr);
slab_data_t *slab_data = edata_slab_data_get(slab);
assert(edata_nfree_get(slab) < bin_info->nregs);
/* Freeing an unallocated pointer can cause assertion failure. */
assert(bitmap_get(slab_data->bitmap, &bin_info->bitmap_info, regind));
bitmap_unset(slab_data->bitmap, &bin_info->bitmap_info, regind);
edata_nfree_inc(slab);
if (config_stats) {
info->ndalloc++;
}
unsigned nfree = edata_nfree_get(slab);
if (nfree == bin_info->nregs) {
arena_dalloc_bin_locked_handle_newly_empty(tsdn, arena, slab,
bin);
return true;
} else if (nfree == 1 && slab != bin->slabcur) {
arena_dalloc_bin_locked_handle_newly_nonempty(tsdn, arena, slab,
bin);
}
return false;
}
JEMALLOC_ALWAYS_INLINE void
arena_dalloc_bin_locked_finish(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
arena_dalloc_bin_locked_info_t *info) {
if (config_stats) {
bin->stats.ndalloc += info->ndalloc;
assert(bin->stats.curregs >= (size_t)info->ndalloc);
bin->stats.curregs -= (size_t)info->ndalloc;
uintptr_t random_offset = ((uintptr_t)r)
<< (LG_PAGE - lg_range);
edata->e_addr = (void *)((byte_t *)edata->e_addr
+ random_offset);
assert(ALIGNMENT_ADDR2BASE(edata->e_addr, alignment)
== edata->e_addr);
}
}
static inline bin_t *
arena_get_bin(arena_t *arena, szind_t binind, unsigned binshard) {
bin_t *shard0 = (bin_t *)((uintptr_t)arena + arena_bin_offsets[binind]);
bin_t *shard0 = (bin_t *)((byte_t *)arena + arena_bin_offsets[binind]);
return shard0 + binshard;
}

63
include/jemalloc/internal/arena_stats.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_ARENA_STATS_H
#define JEMALLOC_INTERNAL_ARENA_STATS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/lockedint.h"
#include "jemalloc/internal/mutex.h"
@ -13,28 +14,34 @@ JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
typedef struct arena_stats_large_s arena_stats_large_t;
struct arena_stats_large_s {
/*
* Total number of allocation/deallocation requests served directly by
* the arena.
* Total number of large allocation/deallocation requests served directly
* by the arena.
*/
locked_u64_t nmalloc;
locked_u64_t ndalloc;
locked_u64_t nmalloc;
locked_u64_t ndalloc;
/*
* Total large active bytes (allocated - deallocated) served directly
* by the arena.
*/
locked_u64_t active_bytes;
/*
* Number of allocation requests that correspond to this size class.
* This includes requests served by tcache, though tcache only
* periodically merges into this counter.
*/
locked_u64_t nrequests; /* Partially derived. */
locked_u64_t nrequests; /* Partially derived. */
/*
* Number of tcache fills / flushes for large (similarly, periodically
* merged). Note that there is no large tcache batch-fill currently
* (i.e. only fill 1 at a time); however flush may be batched.
*/
locked_u64_t nfills; /* Partially derived. */
locked_u64_t nflushes; /* Partially derived. */
locked_u64_t nfills; /* Partially derived. */
locked_u64_t nflushes; /* Partially derived. */
/* Current number of allocations of this size class. */
size_t curlextents; /* Derived. */
size_t curlextents; /* Derived. */
};
/*
@ -50,38 +57,40 @@ struct arena_stats_s {
* resident includes the base stats -- that's why it lives here and not
* in pa_shard_stats_t.
*/
size_t base; /* Derived. */
size_t resident; /* Derived. */
size_t metadata_thp; /* Derived. */
size_t mapped; /* Derived. */
size_t base; /* Derived. */
size_t metadata_edata; /* Derived. */
size_t metadata_rtree; /* Derived. */
size_t resident; /* Derived. */
size_t metadata_thp; /* Derived. */
size_t mapped; /* Derived. */
atomic_zu_t internal;
atomic_zu_t internal;
size_t allocated_large; /* Derived. */
uint64_t nmalloc_large; /* Derived. */
uint64_t ndalloc_large; /* Derived. */
uint64_t nfills_large; /* Derived. */
uint64_t nflushes_large; /* Derived. */
uint64_t nrequests_large; /* Derived. */
size_t allocated_large; /* Derived. */
uint64_t nmalloc_large; /* Derived. */
uint64_t ndalloc_large; /* Derived. */
uint64_t nfills_large; /* Derived. */
uint64_t nflushes_large; /* Derived. */
uint64_t nrequests_large; /* Derived. */
/*
* The stats logically owned by the pa_shard in the same arena. This
* lives here only because it's convenient for the purposes of the ctl
* module -- it only knows about the single arena_stats.
*/
pa_shard_stats_t pa_shard_stats;
pa_shard_stats_t pa_shard_stats;
/* Number of bytes cached in tcache associated with this arena. */
size_t tcache_bytes; /* Derived. */
size_t tcache_stashed_bytes; /* Derived. */
size_t tcache_bytes; /* Derived. */
size_t tcache_stashed_bytes; /* Derived. */
mutex_prof_data_t mutex_prof_data[mutex_prof_num_arena_mutexes];
/* One element for each large size class. */
arena_stats_large_t lstats[SC_NSIZES - SC_NBINS];
arena_stats_large_t lstats[SC_NSIZES - SC_NBINS];
/* Arena uptime. */
nstime_t uptime;
nstime_t uptime;
};
static inline bool
@ -92,7 +101,7 @@ arena_stats_init(tsdn_t *tsdn, arena_stats_t *arena_stats) {
}
}
if (LOCKEDINT_MTX_INIT(arena_stats->mtx, "arena_stats",
WITNESS_RANK_ARENA_STATS, malloc_mutex_rank_exclusive)) {
WITNESS_RANK_ARENA_STATS, malloc_mutex_rank_exclusive)) {
return true;
}
/* Memory is zeroed, so there is no need to clear stats. */
@ -106,8 +115,8 @@ arena_stats_large_flush_nrequests_add(tsdn_t *tsdn, arena_stats_t *arena_stats,
arena_stats_large_t *lstats = &arena_stats->lstats[szind - SC_NBINS];
locked_inc_u64(tsdn, LOCKEDINT_MTX(arena_stats->mtx),
&lstats->nrequests, nrequests);
locked_inc_u64(tsdn, LOCKEDINT_MTX(arena_stats->mtx),
&lstats->nflushes, 1);
locked_inc_u64(
tsdn, LOCKEDINT_MTX(arena_stats->mtx), &lstats->nflushes, 1);
LOCKEDINT_MTX_UNLOCK(tsdn, arena_stats->mtx);
}

38
include/jemalloc/internal/arena_structs.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_ARENA_STRUCTS_H
#define JEMALLOC_INTERNAL_ARENA_STRUCTS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_stats.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/bin.h"
@ -31,20 +32,20 @@ struct arena_s {
*
* Synchronization: atomic.
*/
atomic_u_t nthreads[2];
atomic_u_t nthreads[2];
/* Next bin shard for binding new threads. Synchronization: atomic. */
atomic_u_t binshard_next;
atomic_u_t binshard_next;
/*
* When percpu_arena is enabled, to amortize the cost of reading /
* updating the current CPU id, track the most recent thread accessing
* this arena, and only read CPU if there is a mismatch.
*/
tsdn_t *last_thd;
tsdn_t *last_thd;
/* Synchronization: internal. */
arena_stats_t stats;
arena_stats_t stats;
/*
* Lists of tcaches and cache_bin_array_descriptors for extant threads
@ -53,28 +54,28 @@ struct arena_s {
*
* Synchronization: tcache_ql_mtx.
*/
ql_head(tcache_slow_t) tcache_ql;
ql_head(cache_bin_array_descriptor_t) cache_bin_array_descriptor_ql;
malloc_mutex_t tcache_ql_mtx;
ql_head(tcache_slow_t) tcache_ql;
ql_head(cache_bin_array_descriptor_t) cache_bin_array_descriptor_ql;
malloc_mutex_t tcache_ql_mtx;
/*
* Represents a dss_prec_t, but atomically.
*
* Synchronization: atomic.
*/
atomic_u_t dss_prec;
atomic_u_t dss_prec;
/*
* Extant large allocations.
*
* Synchronization: large_mtx.
*/
edata_list_active_t large;
edata_list_active_t large;
/* Synchronizes all large allocation/update/deallocation. */
malloc_mutex_t large_mtx;
malloc_mutex_t large_mtx;
/* The page-level allocator shard this arena uses. */
pa_shard_t pa_shard;
pa_shard_t pa_shard;
/*
* A cached copy of base->ind. This can get accessed on hot paths;
@ -87,15 +88,24 @@ struct arena_s {
*
* Synchronization: internal.
*/
base_t *base;
base_t *base;
/* Used to determine uptime. Read-only after initialization. */
nstime_t create_time;
nstime_t create_time;
/* The name of the arena. */
char name[ARENA_NAME_LEN];
/*
* The arena is allocated alongside its bins; really this is a
* dynamically sized array determined by the binshard settings.
* Enforcing cacheline-alignment to minimize the number of cachelines
* touched on the hot paths.
*/
bin_t bins[0];
JEMALLOC_WARN_ON_USAGE(
"Do not use this field directly. "
"Use `arena_get_bin` instead.")
JEMALLOC_ALIGNED(CACHELINE)
bin_t all_bins[0];
};
#endif /* JEMALLOC_INTERNAL_ARENA_STRUCTS_H */

28
include/jemalloc/internal/arena_types.h
View file

@ -1,39 +1,41 @@
#ifndef JEMALLOC_INTERNAL_ARENA_TYPES_H
#define JEMALLOC_INTERNAL_ARENA_TYPES_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/sc.h"
/* Default decay times in milliseconds. */
#define DIRTY_DECAY_MS_DEFAULT ZD(10 * 1000)
#define MUZZY_DECAY_MS_DEFAULT (0)
#define DIRTY_DECAY_MS_DEFAULT ZD(10 * 1000)
#define MUZZY_DECAY_MS_DEFAULT (0)
/* Number of event ticks between time checks. */
#define ARENA_DECAY_NTICKS_PER_UPDATE 1000
#define ARENA_DECAY_NTICKS_PER_UPDATE 1000
/* Maximum length of the arena name. */
#define ARENA_NAME_LEN 32
typedef struct arena_decay_s arena_decay_t;
typedef struct arena_s arena_t;
typedef enum {
percpu_arena_mode_names_base = 0, /* Used for options processing. */
percpu_arena_mode_names_base = 0, /* Used for options processing. */
/*
* *_uninit are used only during bootstrapping, and must correspond
* to initialized variant plus percpu_arena_mode_enabled_base.
*/
percpu_arena_uninit = 0,
per_phycpu_arena_uninit = 1,
percpu_arena_uninit = 0,
per_phycpu_arena_uninit = 1,
/* All non-disabled modes must come after percpu_arena_disabled. */
percpu_arena_disabled = 2,
percpu_arena_disabled = 2,
percpu_arena_mode_names_limit = 3, /* Used for options processing. */
percpu_arena_mode_names_limit = 3, /* Used for options processing. */
percpu_arena_mode_enabled_base = 3,
percpu_arena = 3,
per_phycpu_arena = 4 /* Hyper threads share arena. */
percpu_arena = 3,
per_phycpu_arena = 4 /* Hyper threads share arena. */
} percpu_arena_mode_t;
#define PERCPU_ARENA_ENABLED(m) ((m) >= percpu_arena_mode_enabled_base)
#define PERCPU_ARENA_DEFAULT percpu_arena_disabled
#define PERCPU_ARENA_ENABLED(m) ((m) >= percpu_arena_mode_enabled_base)
#define PERCPU_ARENA_DEFAULT percpu_arena_disabled
/*
* When allocation_size >= oversize_threshold, use the dedicated huge arena

75
include/jemalloc/internal/assert.h
View file

@ -1,3 +1,4 @@
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/malloc_io.h"
#include "jemalloc/internal/util.h"
@ -6,51 +7,57 @@
* assertion failure.
*/
#ifndef assert
#define assert(e) do { \
if (unlikely(config_debug && !(e))) { \
malloc_printf( \
"<jemalloc>: %s:%d: Failed assertion: \"%s\"\n", \
__FILE__, __LINE__, #e); \
abort(); \
} \
} while (0)
# define assert(e) \
do { \
if (unlikely(config_debug && !(e))) { \
malloc_printf( \
"<jemalloc>: %s:%d: Failed assertion: \"%s\"\n", \
__FILE__, __LINE__, #e); \
abort(); \
} \
} while (0)
#endif
#ifndef not_reached
#define not_reached() do { \
if (config_debug) { \
malloc_printf( \
"<jemalloc>: %s:%d: Unreachable code reached\n", \
__FILE__, __LINE__); \
abort(); \
} \
unreachable(); \
} while (0)
# define not_reached() \
do { \
if (config_debug) { \
malloc_printf( \
"<jemalloc>: %s:%d: Unreachable code reached\n", \
__FILE__, __LINE__); \
abort(); \
} \
unreachable(); \
} while (0)
#endif
#ifndef not_implemented
#define not_implemented() do { \
if (config_debug) { \
malloc_printf("<jemalloc>: %s:%d: Not implemented\n", \
__FILE__, __LINE__); \
abort(); \
} \
} while (0)
# define not_implemented() \
do { \
if (config_debug) { \
malloc_printf( \
"<jemalloc>: %s:%d: Not implemented\n", \
__FILE__, __LINE__); \
abort(); \
} \
} while (0)
#endif
#ifndef assert_not_implemented
#define assert_not_implemented(e) do { \
if (unlikely(config_debug && !(e))) { \
not_implemented(); \
} \
} while (0)
# define assert_not_implemented(e) \
do { \
if (unlikely(config_debug && !(e))) { \
not_implemented(); \
} \
} while (0)
#endif
/* Use to assert a particular configuration, e.g., cassert(config_debug). */
#ifndef cassert
#define cassert(c) do { \
if (unlikely(!(c))) { \
not_reached(); \
} \
} while (0)
# define cassert(c) \
do { \
if (unlikely(!(c))) { \
not_reached(); \
} \
} while (0)
#endif

59
include/jemalloc/internal/atomic.h
View file

@ -1,27 +1,29 @@
#ifndef JEMALLOC_INTERNAL_ATOMIC_H
#define JEMALLOC_INTERNAL_ATOMIC_H
#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
#include "jemalloc/internal/jemalloc_preamble.h"
#define JEMALLOC_U8_ATOMICS
#if defined(JEMALLOC_GCC_ATOMIC_ATOMICS)
# include "jemalloc/internal/atomic_gcc_atomic.h"
# if !defined(JEMALLOC_GCC_U8_ATOMIC_ATOMICS)
# undef JEMALLOC_U8_ATOMICS
# endif
# include "jemalloc/internal/atomic_gcc_atomic.h"
# if !defined(JEMALLOC_GCC_U8_ATOMIC_ATOMICS)
# undef JEMALLOC_U8_ATOMICS
# endif
#elif defined(JEMALLOC_GCC_SYNC_ATOMICS)
# include "jemalloc/internal/atomic_gcc_sync.h"
# if !defined(JEMALLOC_GCC_U8_SYNC_ATOMICS)
# undef JEMALLOC_U8_ATOMICS
# endif
# include "jemalloc/internal/atomic_gcc_sync.h"
# if !defined(JEMALLOC_GCC_U8_SYNC_ATOMICS)
# undef JEMALLOC_U8_ATOMICS
# endif
#elif defined(_MSC_VER)
# include "jemalloc/internal/atomic_msvc.h"
# include "jemalloc/internal/atomic_msvc.h"
#elif defined(JEMALLOC_C11_ATOMICS)
# include "jemalloc/internal/atomic_c11.h"
# include "jemalloc/internal/atomic_c11.h"
#else
# error "Don't have atomics implemented on this platform."
# error "Don't have atomics implemented on this platform."
#endif
#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
/*
* This header gives more or less a backport of C11 atomics. The user can write
* JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_sizeof_type); to generate
@ -54,22 +56,19 @@
/*
* Another convenience -- simple atomic helper functions.
*/
#define JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(type, short_type, \
lg_size) \
JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, lg_size) \
ATOMIC_INLINE void \
atomic_load_add_store_##short_type(atomic_##short_type##_t *a, \
type inc) { \
type oldval = atomic_load_##short_type(a, ATOMIC_RELAXED); \
type newval = oldval + inc; \
atomic_store_##short_type(a, newval, ATOMIC_RELAXED); \
} \
ATOMIC_INLINE void \
atomic_load_sub_store_##short_type(atomic_##short_type##_t *a, \
type inc) { \
type oldval = atomic_load_##short_type(a, ATOMIC_RELAXED); \
type newval = oldval - inc; \
atomic_store_##short_type(a, newval, ATOMIC_RELAXED); \
#define JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(type, short_type, lg_size) \
JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, lg_size) \
ATOMIC_INLINE void atomic_load_add_store_##short_type( \
atomic_##short_type##_t *a, type inc) { \
type oldval = atomic_load_##short_type(a, ATOMIC_RELAXED); \
type newval = oldval + inc; \
atomic_store_##short_type(a, newval, ATOMIC_RELAXED); \
} \
ATOMIC_INLINE void atomic_load_sub_store_##short_type( \
atomic_##short_type##_t *a, type inc) { \
type oldval = atomic_load_##short_type(a, ATOMIC_RELAXED); \
type newval = oldval - inc; \
atomic_store_##short_type(a, newval, ATOMIC_RELAXED); \
}
/*
@ -77,7 +76,7 @@
* fact.
*/
#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)
# define JEMALLOC_ATOMIC_U64
# define JEMALLOC_ATOMIC_U64
#endif
JEMALLOC_GENERATE_ATOMICS(void *, p, LG_SIZEOF_PTR)
@ -90,6 +89,8 @@ JEMALLOC_GENERATE_ATOMICS(bool, b, 0)
JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(unsigned, u, LG_SIZEOF_INT)
JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(int, i, LG_SIZEOF_INT)
JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(size_t, zu, LG_SIZEOF_PTR)
JEMALLOC_GENERATE_EXPANDED_INT_ATOMICS(ssize_t, zd, LG_SIZEOF_PTR)

57
include/jemalloc/internal/atomic_c11.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_ATOMIC_C11_H
#define JEMALLOC_INTERNAL_ATOMIC_C11_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include <stdatomic.h>
#define ATOMIC_INIT(...) ATOMIC_VAR_INIT(__VA_ARGS__)
@ -14,6 +15,7 @@
#define atomic_fence atomic_thread_fence
/* clang-format off */
#define JEMALLOC_GENERATE_ATOMICS(type, short_type, \
/* unused */ lg_size) \
typedef _Atomic(type) atomic_##short_type##_t; \
@ -58,40 +60,35 @@ atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a, \
return atomic_compare_exchange_strong_explicit(a, expected, \
desired, success_mo, failure_mo); \
}
/* clang-format on */
/*
* Integral types have some special operations available that non-integral ones
* lack.
*/
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, \
/* unused */ lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
\
ATOMIC_INLINE type \
atomic_fetch_add_##short_type(atomic_##short_type##_t *a, \
type val, atomic_memory_order_t mo) { \
return atomic_fetch_add_explicit(a, val, mo); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_sub_##short_type(atomic_##short_type##_t *a, \
type val, atomic_memory_order_t mo) { \
return atomic_fetch_sub_explicit(a, val, mo); \
} \
ATOMIC_INLINE type \
atomic_fetch_and_##short_type(atomic_##short_type##_t *a, \
type val, atomic_memory_order_t mo) { \
return atomic_fetch_and_explicit(a, val, mo); \
} \
ATOMIC_INLINE type \
atomic_fetch_or_##short_type(atomic_##short_type##_t *a, \
type val, atomic_memory_order_t mo) { \
return atomic_fetch_or_explicit(a, val, mo); \
} \
ATOMIC_INLINE type \
atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, \
type val, atomic_memory_order_t mo) { \
return atomic_fetch_xor_explicit(a, val, mo); \
}
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, /* unused */ lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
\
ATOMIC_INLINE type atomic_fetch_add_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return atomic_fetch_add_explicit(a, val, mo); \
} \
\
ATOMIC_INLINE type atomic_fetch_sub_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return atomic_fetch_sub_explicit(a, val, mo); \
} \
ATOMIC_INLINE type atomic_fetch_and_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return atomic_fetch_and_explicit(a, val, mo); \
} \
ATOMIC_INLINE type atomic_fetch_or_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return atomic_fetch_or_explicit(a, val, mo); \
} \
ATOMIC_INLINE type atomic_fetch_xor_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return atomic_fetch_xor_explicit(a, val, mo); \
}
#endif /* JEMALLOC_INTERNAL_ATOMIC_C11_H */

168
include/jemalloc/internal/atomic_gcc_atomic.h
View file

@ -1,9 +1,13 @@
#ifndef JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H
#define JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
#define ATOMIC_INIT(...) {__VA_ARGS__}
#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
#define ATOMIC_INIT(...) \
{ __VA_ARGS__ }
typedef enum {
atomic_memory_order_relaxed,
@ -36,94 +40,82 @@ atomic_fence(atomic_memory_order_t mo) {
__atomic_thread_fence(atomic_enum_to_builtin(mo));
}
#define JEMALLOC_GENERATE_ATOMICS(type, short_type, \
/* unused */ lg_size) \
typedef struct { \
type repr; \
} atomic_##short_type##_t; \
\
ATOMIC_INLINE type \
atomic_load_##short_type(const atomic_##short_type##_t *a, \
atomic_memory_order_t mo) { \
type result; \
__atomic_load(&a->repr, &result, atomic_enum_to_builtin(mo)); \
return result; \
} \
\
ATOMIC_INLINE void \
atomic_store_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
__atomic_store(&a->repr, &val, atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type \
atomic_exchange_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
type result; \
__atomic_exchange(&a->repr, &val, &result, \
atomic_enum_to_builtin(mo)); \
return result; \
} \
\
ATOMIC_INLINE bool \
atomic_compare_exchange_weak_##short_type(atomic_##short_type##_t *a, \
UNUSED type *expected, type desired, \
atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
return __atomic_compare_exchange(&a->repr, expected, &desired, \
true, atomic_enum_to_builtin(success_mo), \
atomic_enum_to_builtin(failure_mo)); \
} \
\
ATOMIC_INLINE bool \
atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a, \
UNUSED type *expected, type desired, \
atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
return __atomic_compare_exchange(&a->repr, expected, &desired, \
false, \
atomic_enum_to_builtin(success_mo), \
atomic_enum_to_builtin(failure_mo)); \
}
#define JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
typedef struct { \
type repr; \
} atomic_##short_type##_t; \
\
ATOMIC_INLINE type atomic_load_##short_type( \
const atomic_##short_type##_t *a, atomic_memory_order_t mo) { \
type result; \
__atomic_load(&a->repr, &result, atomic_enum_to_builtin(mo)); \
return result; \
} \
\
ATOMIC_INLINE void atomic_store_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
__atomic_store(&a->repr, &val, atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type atomic_exchange_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
type result; \
__atomic_exchange( \
&a->repr, &val, &result, atomic_enum_to_builtin(mo)); \
return result; \
} \
\
ATOMIC_INLINE bool atomic_compare_exchange_weak_##short_type( \
atomic_##short_type##_t *a, UNUSED type *expected, type desired, \
atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
return __atomic_compare_exchange(&a->repr, expected, &desired, \
true, atomic_enum_to_builtin(success_mo), \
atomic_enum_to_builtin(failure_mo)); \
} \
\
ATOMIC_INLINE bool atomic_compare_exchange_strong_##short_type( \
atomic_##short_type##_t *a, UNUSED type *expected, type desired, \
atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
return __atomic_compare_exchange(&a->repr, expected, &desired, \
false, atomic_enum_to_builtin(success_mo), \
atomic_enum_to_builtin(failure_mo)); \
}
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, /* unused */ lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
\
ATOMIC_INLINE type atomic_fetch_add_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __atomic_fetch_add( \
&a->repr, val, atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type atomic_fetch_sub_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __atomic_fetch_sub( \
&a->repr, val, atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type atomic_fetch_and_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __atomic_fetch_and( \
&a->repr, val, atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type atomic_fetch_or_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __atomic_fetch_or( \
&a->repr, val, atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type atomic_fetch_xor_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __atomic_fetch_xor( \
&a->repr, val, atomic_enum_to_builtin(mo)); \
}
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, \
/* unused */ lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
\
ATOMIC_INLINE type \
atomic_fetch_add_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __atomic_fetch_add(&a->repr, val, \
atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_sub_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __atomic_fetch_sub(&a->repr, val, \
atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_and_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __atomic_fetch_and(&a->repr, val, \
atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_or_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __atomic_fetch_or(&a->repr, val, \
atomic_enum_to_builtin(mo)); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __atomic_fetch_xor(&a->repr, val, \
atomic_enum_to_builtin(mo)); \
}
#undef ATOMIC_INLINE
#endif /* JEMALLOC_INTERNAL_ATOMIC_GCC_ATOMIC_H */

100
include/jemalloc/internal/atomic_gcc_sync.h
View file

@ -1,7 +1,12 @@
#ifndef JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H
#define JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H
#define ATOMIC_INIT(...) {__VA_ARGS__}
#include "jemalloc/internal/jemalloc_preamble.h"
#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
#define ATOMIC_INIT(...) \
{ __VA_ARGS__ }
typedef enum {
atomic_memory_order_relaxed,
@ -25,13 +30,13 @@ atomic_fence(atomic_memory_order_t mo) {
return;
}
asm volatile("" ::: "memory");
# if defined(__i386__) || defined(__x86_64__)
#if defined(__i386__) || defined(__x86_64__)
/* This is implicit on x86. */
# elif defined(__ppc64__)
#elif defined(__ppc64__)
asm volatile("lwsync");
# elif defined(__ppc__)
#elif defined(__ppc__)
asm volatile("sync");
# elif defined(__sparc__) && defined(__arch64__)
#elif defined(__sparc__) && defined(__arch64__)
if (mo == atomic_memory_order_acquire) {
asm volatile("membar #LoadLoad | #LoadStore");
} else if (mo == atomic_memory_order_release) {
@ -39,9 +44,9 @@ atomic_fence(atomic_memory_order_t mo) {
} else {
asm volatile("membar #LoadLoad | #LoadStore | #StoreStore");
}
# else
#else
__sync_synchronize();
# endif
#endif
asm volatile("" ::: "memory");
}
@ -64,25 +69,25 @@ atomic_fence(atomic_memory_order_t mo) {
ATOMIC_INLINE void
atomic_pre_sc_load_fence() {
# if defined(__i386__) || defined(__x86_64__) || \
(defined(__sparc__) && defined(__arch64__))
#if defined(__i386__) || defined(__x86_64__) \
|| (defined(__sparc__) && defined(__arch64__))
atomic_fence(atomic_memory_order_relaxed);
# else
#else
atomic_fence(atomic_memory_order_seq_cst);
# endif
#endif
}
ATOMIC_INLINE void
atomic_post_sc_store_fence() {
# if defined(__i386__) || defined(__x86_64__) || \
(defined(__sparc__) && defined(__arch64__))
#if defined(__i386__) || defined(__x86_64__) \
|| (defined(__sparc__) && defined(__arch64__))
atomic_fence(atomic_memory_order_seq_cst);
# else
#else
atomic_fence(atomic_memory_order_relaxed);
# endif
#endif
}
/* clang-format off */
#define JEMALLOC_GENERATE_ATOMICS(type, short_type, \
/* unused */ lg_size) \
typedef struct { \
@ -157,39 +162,36 @@ atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a, \
return false; \
} \
}
/* clang-format on */
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, \
/* unused */ lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
\
ATOMIC_INLINE type \
atomic_fetch_add_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __sync_fetch_and_add(&a->repr, val); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_sub_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __sync_fetch_and_sub(&a->repr, val); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_and_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __sync_fetch_and_and(&a->repr, val); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_or_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __sync_fetch_and_or(&a->repr, val); \
} \
\
ATOMIC_INLINE type \
atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return __sync_fetch_and_xor(&a->repr, val); \
}
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, /* unused */ lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, /* unused */ lg_size) \
\
ATOMIC_INLINE type atomic_fetch_add_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __sync_fetch_and_add(&a->repr, val); \
} \
\
ATOMIC_INLINE type atomic_fetch_sub_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __sync_fetch_and_sub(&a->repr, val); \
} \
\
ATOMIC_INLINE type atomic_fetch_and_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __sync_fetch_and_and(&a->repr, val); \
} \
\
ATOMIC_INLINE type atomic_fetch_or_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __sync_fetch_and_or(&a->repr, val); \
} \
\
ATOMIC_INLINE type atomic_fetch_xor_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return __sync_fetch_and_xor(&a->repr, val); \
}
#undef ATOMIC_INLINE
#endif /* JEMALLOC_INTERNAL_ATOMIC_GCC_SYNC_H */

159
include/jemalloc/internal/atomic_msvc.h
View file

@ -1,7 +1,12 @@
#ifndef JEMALLOC_INTERNAL_ATOMIC_MSVC_H
#define JEMALLOC_INTERNAL_ATOMIC_MSVC_H
#define ATOMIC_INIT(...) {__VA_ARGS__}
#include "jemalloc/internal/jemalloc_preamble.h"
#define ATOMIC_INLINE JEMALLOC_ALWAYS_INLINE
#define ATOMIC_INIT(...) \
{ __VA_ARGS__ }
typedef enum {
atomic_memory_order_relaxed,
@ -11,109 +16,106 @@ typedef enum {
atomic_memory_order_seq_cst
} atomic_memory_order_t;
typedef char atomic_repr_0_t;
typedef short atomic_repr_1_t;
typedef long atomic_repr_2_t;
typedef char atomic_repr_0_t;
typedef short atomic_repr_1_t;
typedef long atomic_repr_2_t;
typedef __int64 atomic_repr_3_t;
ATOMIC_INLINE void
atomic_fence(atomic_memory_order_t mo) {
_ReadWriteBarrier();
# if defined(_M_ARM) || defined(_M_ARM64)
#if defined(_M_ARM) || defined(_M_ARM64)
/* ARM needs a barrier for everything but relaxed. */
if (mo != atomic_memory_order_relaxed) {
MemoryBarrier();
}
# elif defined(_M_IX86) || defined (_M_X64)
#elif defined(_M_IX86) || defined(_M_X64)
/* x86 needs a barrier only for seq_cst. */
if (mo == atomic_memory_order_seq_cst) {
MemoryBarrier();
}
# else
# error "Don't know how to create atomics for this platform for MSVC."
# endif
#else
# error "Don't know how to create atomics for this platform for MSVC."
#endif
_ReadWriteBarrier();
}
#define ATOMIC_INTERLOCKED_REPR(lg_size) atomic_repr_ ## lg_size ## _t
#define ATOMIC_INTERLOCKED_REPR(lg_size) atomic_repr_##lg_size##_t
#define ATOMIC_CONCAT(a, b) ATOMIC_RAW_CONCAT(a, b)
#define ATOMIC_RAW_CONCAT(a, b) a ## b
#define ATOMIC_RAW_CONCAT(a, b) a##b
#define ATOMIC_INTERLOCKED_NAME(base_name, lg_size) ATOMIC_CONCAT( \
base_name, ATOMIC_INTERLOCKED_SUFFIX(lg_size))
#define ATOMIC_INTERLOCKED_NAME(base_name, lg_size) \
ATOMIC_CONCAT(base_name, ATOMIC_INTERLOCKED_SUFFIX(lg_size))
#define ATOMIC_INTERLOCKED_SUFFIX(lg_size) \
ATOMIC_CONCAT(ATOMIC_INTERLOCKED_SUFFIX_, lg_size)
#define ATOMIC_INTERLOCKED_SUFFIX(lg_size) \
ATOMIC_CONCAT(ATOMIC_INTERLOCKED_SUFFIX_, lg_size)
#define ATOMIC_INTERLOCKED_SUFFIX_0 8
#define ATOMIC_INTERLOCKED_SUFFIX_1 16
#define ATOMIC_INTERLOCKED_SUFFIX_2
#define ATOMIC_INTERLOCKED_SUFFIX_3 64
#define JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_size) \
typedef struct { \
ATOMIC_INTERLOCKED_REPR(lg_size) repr; \
} atomic_##short_type##_t; \
\
ATOMIC_INLINE type \
atomic_load_##short_type(const atomic_##short_type##_t *a, \
atomic_memory_order_t mo) { \
ATOMIC_INTERLOCKED_REPR(lg_size) ret = a->repr; \
if (mo != atomic_memory_order_relaxed) { \
atomic_fence(atomic_memory_order_acquire); \
} \
return (type) ret; \
} \
\
ATOMIC_INLINE void \
atomic_store_##short_type(atomic_##short_type##_t *a, \
type val, atomic_memory_order_t mo) { \
if (mo != atomic_memory_order_relaxed) { \
atomic_fence(atomic_memory_order_release); \
} \
a->repr = (ATOMIC_INTERLOCKED_REPR(lg_size)) val; \
if (mo == atomic_memory_order_seq_cst) { \
atomic_fence(atomic_memory_order_seq_cst); \
} \
} \
\
ATOMIC_INLINE type \
atomic_exchange_##short_type(atomic_##short_type##_t *a, type val, \
atomic_memory_order_t mo) { \
return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedExchange, \
lg_size)(&a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val); \
} \
\
ATOMIC_INLINE bool \
atomic_compare_exchange_weak_##short_type(atomic_##short_type##_t *a, \
type *expected, type desired, atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
ATOMIC_INTERLOCKED_REPR(lg_size) e = \
(ATOMIC_INTERLOCKED_REPR(lg_size))*expected; \
ATOMIC_INTERLOCKED_REPR(lg_size) d = \
(ATOMIC_INTERLOCKED_REPR(lg_size))desired; \
ATOMIC_INTERLOCKED_REPR(lg_size) old = \
ATOMIC_INTERLOCKED_NAME(_InterlockedCompareExchange, \
lg_size)(&a->repr, d, e); \
if (old == e) { \
return true; \
} else { \
*expected = (type)old; \
return false; \
} \
} \
\
ATOMIC_INLINE bool \
atomic_compare_exchange_strong_##short_type(atomic_##short_type##_t *a, \
type *expected, type desired, atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
/* We implement the weak version with strong semantics. */ \
return atomic_compare_exchange_weak_##short_type(a, expected, \
desired, success_mo, failure_mo); \
}
#define JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_size) \
typedef struct { \
ATOMIC_INTERLOCKED_REPR(lg_size) repr; \
} atomic_##short_type##_t; \
\
ATOMIC_INLINE type atomic_load_##short_type( \
const atomic_##short_type##_t *a, atomic_memory_order_t mo) { \
ATOMIC_INTERLOCKED_REPR(lg_size) ret = a->repr; \
if (mo != atomic_memory_order_relaxed) { \
atomic_fence(atomic_memory_order_acquire); \
} \
return (type)ret; \
} \
\
ATOMIC_INLINE void atomic_store_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
if (mo != atomic_memory_order_relaxed) { \
atomic_fence(atomic_memory_order_release); \
} \
a->repr = (ATOMIC_INTERLOCKED_REPR(lg_size))val; \
if (mo == atomic_memory_order_seq_cst) { \
atomic_fence(atomic_memory_order_seq_cst); \
} \
} \
\
ATOMIC_INLINE type atomic_exchange_##short_type( \
atomic_##short_type##_t *a, type val, atomic_memory_order_t mo) { \
return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedExchange, \
lg_size)(&a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val); \
} \
\
ATOMIC_INLINE bool atomic_compare_exchange_weak_##short_type( \
atomic_##short_type##_t *a, type *expected, type desired, \
atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
ATOMIC_INTERLOCKED_REPR(lg_size) \
e = (ATOMIC_INTERLOCKED_REPR(lg_size)) * expected; \
ATOMIC_INTERLOCKED_REPR(lg_size) \
d = (ATOMIC_INTERLOCKED_REPR(lg_size))desired; \
ATOMIC_INTERLOCKED_REPR(lg_size) \
old = ATOMIC_INTERLOCKED_NAME( \
_InterlockedCompareExchange, lg_size)(&a->repr, d, e); \
if (old == e) { \
return true; \
} else { \
*expected = (type)old; \
return false; \
} \
} \
\
ATOMIC_INLINE bool atomic_compare_exchange_strong_##short_type( \
atomic_##short_type##_t *a, type *expected, type desired, \
atomic_memory_order_t success_mo, \
atomic_memory_order_t failure_mo) { \
/* We implement the weak version with strong semantics. */ \
return atomic_compare_exchange_weak_##short_type( \
a, expected, desired, success_mo, failure_mo); \
}
/* clang-format off */
#define JEMALLOC_GENERATE_INT_ATOMICS(type, short_type, lg_size) \
JEMALLOC_GENERATE_ATOMICS(type, short_type, lg_size) \
\
@ -154,5 +156,8 @@ atomic_fetch_xor_##short_type(atomic_##short_type##_t *a, \
return (type)ATOMIC_INTERLOCKED_NAME(_InterlockedXor, lg_size)( \
&a->repr, (ATOMIC_INTERLOCKED_REPR(lg_size))val); \
}
/* clang-format on */
#undef ATOMIC_INLINE
#endif /* JEMALLOC_INTERNAL_ATOMIC_MSVC_H */

27
include/jemalloc/internal/background_thread_externs.h
View file

@ -1,26 +1,31 @@
#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
extern bool opt_background_thread;
extern size_t opt_max_background_threads;
extern malloc_mutex_t background_thread_lock;
extern atomic_b_t background_thread_enabled_state;
extern size_t n_background_threads;
extern size_t max_background_threads;
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/background_thread_structs.h"
#include "jemalloc/internal/base.h"
#include "jemalloc/internal/mutex.h"
extern bool opt_background_thread;
extern size_t opt_max_background_threads;
extern malloc_mutex_t background_thread_lock;
extern atomic_b_t background_thread_enabled_state;
extern size_t n_background_threads;
extern size_t max_background_threads;
extern background_thread_info_t *background_thread_info;
bool background_thread_create(tsd_t *tsd, unsigned arena_ind);
bool background_threads_enable(tsd_t *tsd);
bool background_threads_disable(tsd_t *tsd);
bool background_thread_is_started(background_thread_info_t* info);
void background_thread_wakeup_early(background_thread_info_t *info,
nstime_t *remaining_sleep);
bool background_thread_is_started(background_thread_info_t *info);
void background_thread_wakeup_early(
background_thread_info_t *info, nstime_t *remaining_sleep);
void background_thread_prefork0(tsdn_t *tsdn);
void background_thread_prefork1(tsdn_t *tsdn);
void background_thread_postfork_parent(tsdn_t *tsdn);
void background_thread_postfork_child(tsdn_t *tsdn);
bool background_thread_stats_read(tsdn_t *tsdn,
background_thread_stats_t *stats);
bool background_thread_stats_read(
tsdn_t *tsdn, background_thread_stats_t *stats);
void background_thread_ctl_init(tsdn_t *tsdn);
#ifdef JEMALLOC_PTHREAD_CREATE_WRAPPER

20
include/jemalloc/internal/background_thread_inlines.h
View file

@ -1,15 +1,25 @@
#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H
#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_INLINES_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_inlines_a.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/background_thread_externs.h"
JEMALLOC_ALWAYS_INLINE bool
background_thread_enabled(void) {
return atomic_load_b(&background_thread_enabled_state, ATOMIC_RELAXED);
}
JEMALLOC_ALWAYS_INLINE void
background_thread_enabled_set_impl(bool state) {
atomic_store_b(&background_thread_enabled_state, state, ATOMIC_RELAXED);
}
JEMALLOC_ALWAYS_INLINE void
background_thread_enabled_set(tsdn_t *tsdn, bool state) {
malloc_mutex_assert_owner(tsdn, &background_thread_lock);
atomic_store_b(&background_thread_enabled_state, state, ATOMIC_RELAXED);
background_thread_enabled_set_impl(state);
}
JEMALLOC_ALWAYS_INLINE background_thread_info_t *
@ -26,14 +36,14 @@ background_thread_info_get(size_t ind) {
JEMALLOC_ALWAYS_INLINE uint64_t
background_thread_wakeup_time_get(background_thread_info_t *info) {
uint64_t next_wakeup = nstime_ns(&info->next_wakeup);
assert(atomic_load_b(&info->indefinite_sleep, ATOMIC_ACQUIRE) ==
(next_wakeup == BACKGROUND_THREAD_INDEFINITE_SLEEP));
assert(atomic_load_b(&info->indefinite_sleep, ATOMIC_ACQUIRE)
== (next_wakeup == BACKGROUND_THREAD_INDEFINITE_SLEEP));
return next_wakeup;
}
JEMALLOC_ALWAYS_INLINE void
background_thread_wakeup_time_set(tsdn_t *tsdn, background_thread_info_t *info,
uint64_t wakeup_time) {
background_thread_wakeup_time_set(
tsdn_t *tsdn, background_thread_info_t *info, uint64_t wakeup_time) {
malloc_mutex_assert_owner(tsdn, &info->mtx);
atomic_store_b(&info->indefinite_sleep,
wakeup_time == BACKGROUND_THREAD_INDEFINITE_SLEEP, ATOMIC_RELEASE);

29
include/jemalloc/internal/background_thread_structs.h
View file

@ -1,10 +1,13 @@
#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/mutex.h"
/* This file really combines "structs" and "types", but only transitionally. */
#if defined(JEMALLOC_BACKGROUND_THREAD) || defined(JEMALLOC_LAZY_LOCK)
# define JEMALLOC_PTHREAD_CREATE_WRAPPER
# define JEMALLOC_PTHREAD_CREATE_WRAPPER
#endif
#define BACKGROUND_THREAD_INDEFINITE_SLEEP UINT64_MAX
@ -32,33 +35,33 @@ typedef enum {
struct background_thread_info_s {
#ifdef JEMALLOC_BACKGROUND_THREAD
/* Background thread is pthread specific. */
pthread_t thread;
pthread_cond_t cond;
pthread_t thread;
pthread_cond_t cond;
#endif
malloc_mutex_t mtx;
background_thread_state_t state;
malloc_mutex_t mtx;
background_thread_state_t state;
/* When true, it means no wakeup scheduled. */
atomic_b_t indefinite_sleep;
atomic_b_t indefinite_sleep;
/* Next scheduled wakeup time (absolute time in ns). */
nstime_t next_wakeup;
nstime_t next_wakeup;
/*
* Since the last background thread run, newly added number of pages
* that need to be purged by the next wakeup. This is adjusted on
* epoch advance, and is used to determine whether we should signal the
* background thread to wake up earlier.
*/
size_t npages_to_purge_new;
size_t npages_to_purge_new;
/* Stats: total number of runs since started. */
uint64_t tot_n_runs;
uint64_t tot_n_runs;
/* Stats: total sleep time since started. */
nstime_t tot_sleep_time;
nstime_t tot_sleep_time;
};
typedef struct background_thread_info_s background_thread_info_t;
struct background_thread_stats_s {
size_t num_threads;
uint64_t num_runs;
nstime_t run_interval;
size_t num_threads;
uint64_t num_runs;
nstime_t run_interval;
mutex_prof_data_t max_counter_per_bg_thd;
};
typedef struct background_thread_stats_s background_thread_stats_t;

53
include/jemalloc/internal/base.h
View file

@ -1,12 +1,19 @@
#ifndef JEMALLOC_INTERNAL_BASE_H
#define JEMALLOC_INTERNAL_BASE_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/edata.h"
#include "jemalloc/internal/ehooks.h"
#include "jemalloc/internal/mutex.h"
/*
* Alignment when THP is not enabled. Set to constant 2M in case the HUGEPAGE
* value is unexpected high (which would cause VM over-reservation).
*/
#define BASE_BLOCK_MIN_ALIGN ((size_t)2 << 20)
enum metadata_thp_mode_e {
metadata_thp_disabled = 0,
metadata_thp_disabled = 0,
/*
* Lazily enable hugepage for metadata. To avoid high RSS caused by THP
* + low usage arena (i.e. THP becomes a significant percentage), the
@ -15,16 +22,15 @@ enum metadata_thp_mode_e {
* arena), "auto" behaves the same as "always", i.e. madvise hugepage
* right away.
*/
metadata_thp_auto = 1,
metadata_thp_always = 2,
metadata_thp_auto = 1,
metadata_thp_always = 2,
metadata_thp_mode_limit = 3
};
typedef enum metadata_thp_mode_e metadata_thp_mode_t;
#define METADATA_THP_DEFAULT metadata_thp_disabled
extern metadata_thp_mode_t opt_metadata_thp;
extern const char *metadata_thp_mode_names[];
extern const char *const metadata_thp_mode_names[];
/* Embedded at the beginning of every block of base-managed virtual memory. */
typedef struct base_block_s base_block_t;
@ -72,8 +78,13 @@ struct base_s {
/* Heap of extents that track unused trailing space within blocks. */
edata_heap_t avail[SC_NSIZES];
/* Contains reusable base edata (used by tcache_stacks currently). */
edata_avail_t edata_avail;
/* Stats, only maintained if config_stats. */
size_t allocated;
size_t edata_allocated;
size_t rtree_allocated;
size_t resident;
size_t mapped;
/* Number of THP regions touched. */
@ -91,20 +102,24 @@ metadata_thp_enabled(void) {
}
base_t *b0get(void);
base_t *base_new(tsdn_t *tsdn, unsigned ind,
const extent_hooks_t *extent_hooks, bool metadata_use_hooks);
void base_delete(tsdn_t *tsdn, base_t *base);
ehooks_t *base_ehooks_get(base_t *base);
ehooks_t *base_ehooks_get_for_metadata(base_t *base);
extent_hooks_t *base_extent_hooks_set(base_t *base,
extent_hooks_t *extent_hooks);
void *base_alloc(tsdn_t *tsdn, base_t *base, size_t size, size_t alignment);
base_t *base_new(tsdn_t *tsdn, unsigned ind, const extent_hooks_t *extent_hooks,
bool metadata_use_hooks);
void base_delete(tsdn_t *tsdn, base_t *base);
ehooks_t *base_ehooks_get(base_t *base);
ehooks_t *base_ehooks_get_for_metadata(base_t *base);
extent_hooks_t *base_extent_hooks_set(
base_t *base, extent_hooks_t *extent_hooks);
void *base_alloc(tsdn_t *tsdn, base_t *base, size_t size, size_t alignment);
edata_t *base_alloc_edata(tsdn_t *tsdn, base_t *base);
void base_stats_get(tsdn_t *tsdn, base_t *base, size_t *allocated,
size_t *resident, size_t *mapped, size_t *n_thp);
void base_prefork(tsdn_t *tsdn, base_t *base);
void base_postfork_parent(tsdn_t *tsdn, base_t *base);
void base_postfork_child(tsdn_t *tsdn, base_t *base);
bool base_boot(tsdn_t *tsdn);
void *base_alloc_rtree(tsdn_t *tsdn, base_t *base, size_t size);
void *b0_alloc_tcache_stack(tsdn_t *tsdn, size_t size);
void b0_dalloc_tcache_stack(tsdn_t *tsdn, void *tcache_stack);
void base_stats_get(tsdn_t *tsdn, base_t *base, size_t *allocated,
size_t *edata_allocated, size_t *rtree_allocated, size_t *resident,
size_t *mapped, size_t *n_thp);
void base_prefork(tsdn_t *tsdn, base_t *base);
void base_postfork_parent(tsdn_t *tsdn, base_t *base);
void base_postfork_child(tsdn_t *tsdn, base_t *base);
bool base_boot(tsdn_t *tsdn);
#endif /* JEMALLOC_INTERNAL_BASE_H */

51
include/jemalloc/internal/bin.h
View file

@ -1,6 +1,8 @@
#ifndef JEMALLOC_INTERNAL_BIN_H
#define JEMALLOC_INTERNAL_BIN_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/bin_info.h"
#include "jemalloc/internal/bin_stats.h"
#include "jemalloc/internal/bin_types.h"
#include "jemalloc/internal/edata.h"
@ -14,13 +16,13 @@
typedef struct bin_s bin_t;
struct bin_s {
/* All operations on bin_t fields require lock ownership. */
malloc_mutex_t lock;
malloc_mutex_t lock;
/*
* Bin statistics. These get touched every time the lock is acquired,
* so put them close by in the hopes of getting some cache locality.
*/
bin_stats_t stats;
bin_stats_t stats;
/*
* Current slab being used to service allocations of this bin's size
@ -28,17 +30,17 @@ struct bin_s {
* slabcur is reassigned, the previous slab must be deallocated or
* inserted into slabs_{nonfull,full}.
*/
edata_t *slabcur;
edata_t *slabcur;
/*
* Heap of non-full slabs. This heap is used to assure that new
* allocations come from the non-full slab that is oldest/lowest in
* memory.
*/
edata_heap_t slabs_nonfull;
edata_heap_t slabs_nonfull;
/* List used to track full slabs. */
edata_list_active_t slabs_full;
edata_list_active_t slabs_full;
};
/* A set of sharded bins of the same size class. */
@ -48,7 +50,7 @@ struct bins_s {
bin_t *bin_shards;
};
void bin_shard_sizes_boot(unsigned bin_shards[SC_NBINS]);
void bin_shard_sizes_boot(unsigned bin_shard_sizes[SC_NBINS]);
bool bin_update_shard_size(unsigned bin_shards[SC_NBINS], size_t start_size,
size_t end_size, size_t nshards);
@ -60,6 +62,43 @@ void bin_prefork(tsdn_t *tsdn, bin_t *bin);
void bin_postfork_parent(tsdn_t *tsdn, bin_t *bin);
void bin_postfork_child(tsdn_t *tsdn, bin_t *bin);
/* Slab region allocation. */
void *bin_slab_reg_alloc(edata_t *slab, const bin_info_t *bin_info);
void bin_slab_reg_alloc_batch(
edata_t *slab, const bin_info_t *bin_info, unsigned cnt, void **ptrs);
/* Slab list management. */
void bin_slabs_nonfull_insert(bin_t *bin, edata_t *slab);
void bin_slabs_nonfull_remove(bin_t *bin, edata_t *slab);
edata_t *bin_slabs_nonfull_tryget(bin_t *bin);
void bin_slabs_full_insert(bool is_auto, bin_t *bin, edata_t *slab);
void bin_slabs_full_remove(bool is_auto, bin_t *bin, edata_t *slab);
/* Slab association / demotion. */
void bin_dissociate_slab(bool is_auto, edata_t *slab, bin_t *bin);
void bin_lower_slab(tsdn_t *tsdn, bool is_auto, edata_t *slab, bin_t *bin);
/* Deallocation helpers (called under bin lock). */
void bin_dalloc_slab_prepare(tsdn_t *tsdn, edata_t *slab, bin_t *bin);
void bin_dalloc_locked_handle_newly_empty(
tsdn_t *tsdn, bool is_auto, edata_t *slab, bin_t *bin);
void bin_dalloc_locked_handle_newly_nonempty(
tsdn_t *tsdn, bool is_auto, edata_t *slab, bin_t *bin);
/* Slabcur refill and allocation. */
void bin_refill_slabcur_with_fresh_slab(tsdn_t *tsdn, bin_t *bin,
szind_t binind, edata_t *fresh_slab);
void *bin_malloc_with_fresh_slab(tsdn_t *tsdn, bin_t *bin,
szind_t binind, edata_t *fresh_slab);
bool bin_refill_slabcur_no_fresh_slab(tsdn_t *tsdn, bool is_auto,
bin_t *bin);
void *bin_malloc_no_fresh_slab(tsdn_t *tsdn, bool is_auto, bin_t *bin,
szind_t binind);
/* Bin selection. */
bin_t *bin_choose(tsdn_t *tsdn, arena_t *arena, szind_t binind,
unsigned *binshard_p);
/* Stats. */
static inline void
bin_stats_merge(tsdn_t *tsdn, bin_stats_data_t *dst_bin_stats, bin_t *bin) {

11
include/jemalloc/internal/bin_info.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_BIN_INFO_H
#define JEMALLOC_INTERNAL_BIN_INFO_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/bitmap.h"
/*
@ -25,22 +26,22 @@
typedef struct bin_info_s bin_info_t;
struct bin_info_s {
/* Size of regions in a slab for this bin's size class. */
size_t reg_size;
size_t reg_size;
/* Total size of a slab for this bin's size class. */
size_t slab_size;
size_t slab_size;
/* Total number of regions in a slab for this bin's size class. */
uint32_t nregs;
uint32_t nregs;
/* Number of sharded bins in each arena for this size class. */
uint32_t n_shards;
uint32_t n_shards;
/*
* Metadata used to manipulate bitmaps for slabs associated with this
* bin.
*/
bitmap_info_t bitmap_info;
bitmap_info_t bitmap_info;
};
extern bin_info_t bin_infos[SC_NBINS];

112
include/jemalloc/internal/bin_inlines.h Normal file
View file

@ -0,0 +1,112 @@
#ifndef JEMALLOC_INTERNAL_BIN_INLINES_H
#define JEMALLOC_INTERNAL_BIN_INLINES_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/bin.h"
#include "jemalloc/internal/bin_info.h"
#include "jemalloc/internal/bitmap.h"
#include "jemalloc/internal/div.h"
#include "jemalloc/internal/edata.h"
#include "jemalloc/internal/sc.h"
/*
* The dalloc bin info contains just the information that the common paths need
* during tcache flushes. By force-inlining these paths, and using local copies
* of data (so that the compiler knows it's constant), we avoid a whole bunch of
* redundant loads and stores by leaving this information in registers.
*/
typedef struct bin_dalloc_locked_info_s bin_dalloc_locked_info_t;
struct bin_dalloc_locked_info_s {
div_info_t div_info;
uint32_t nregs;
uint64_t ndalloc;
};
/* Find the region index of a pointer within a slab. */
JEMALLOC_ALWAYS_INLINE size_t
bin_slab_regind_impl(
div_info_t *div_info, szind_t binind, edata_t *slab, const void *ptr) {
size_t diff, regind;
/* Freeing a pointer outside the slab can cause assertion failure. */
assert((uintptr_t)ptr >= (uintptr_t)edata_addr_get(slab));
assert((uintptr_t)ptr < (uintptr_t)edata_past_get(slab));
/* Freeing an interior pointer can cause assertion failure. */
assert(((uintptr_t)ptr - (uintptr_t)edata_addr_get(slab))
% (uintptr_t)bin_infos[binind].reg_size
== 0);
diff = (size_t)((uintptr_t)ptr - (uintptr_t)edata_addr_get(slab));
/* Avoid doing division with a variable divisor. */
regind = div_compute(div_info, diff);
assert(regind < bin_infos[binind].nregs);
return regind;
}
JEMALLOC_ALWAYS_INLINE size_t
bin_slab_regind(bin_dalloc_locked_info_t *info, szind_t binind,
edata_t *slab, const void *ptr) {
size_t regind = bin_slab_regind_impl(
&info->div_info, binind, slab, ptr);
return regind;
}
JEMALLOC_ALWAYS_INLINE void
bin_dalloc_locked_begin(
bin_dalloc_locked_info_t *info, szind_t binind) {
info->div_info = arena_binind_div_info[binind];
info->nregs = bin_infos[binind].nregs;
info->ndalloc = 0;
}
/*
* Does the deallocation work associated with freeing a single pointer (a
* "step") in between a bin_dalloc_locked begin and end call.
*
* Returns true if arena_slab_dalloc must be called on slab. Doesn't do
* stats updates, which happen during finish (this lets running counts get left
* in a register).
*/
JEMALLOC_ALWAYS_INLINE bool
bin_dalloc_locked_step(tsdn_t *tsdn, bool is_auto, bin_t *bin,
bin_dalloc_locked_info_t *info, szind_t binind, edata_t *slab,
void *ptr) {
const bin_info_t *bin_info = &bin_infos[binind];
size_t regind = bin_slab_regind(info, binind, slab, ptr);
slab_data_t *slab_data = edata_slab_data_get(slab);
assert(edata_nfree_get(slab) < bin_info->nregs);
/* Freeing an unallocated pointer can cause assertion failure. */
assert(bitmap_get(slab_data->bitmap, &bin_info->bitmap_info, regind));
bitmap_unset(slab_data->bitmap, &bin_info->bitmap_info, regind);
edata_nfree_inc(slab);
if (config_stats) {
info->ndalloc++;
}
unsigned nfree = edata_nfree_get(slab);
if (nfree == bin_info->nregs) {
bin_dalloc_locked_handle_newly_empty(
tsdn, is_auto, slab, bin);
return true;
} else if (nfree == 1 && slab != bin->slabcur) {
bin_dalloc_locked_handle_newly_nonempty(
tsdn, is_auto, slab, bin);
}
return false;
}
JEMALLOC_ALWAYS_INLINE void
bin_dalloc_locked_finish(tsdn_t *tsdn, bin_t *bin,
bin_dalloc_locked_info_t *info) {
if (config_stats) {
bin->stats.ndalloc += info->ndalloc;
assert(bin->stats.curregs >= (size_t)info->ndalloc);
bin->stats.curregs -= (size_t)info->ndalloc;
}
}
#endif /* JEMALLOC_INTERNAL_BIN_INLINES_H */

23
include/jemalloc/internal/bin_stats.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_BIN_STATS_H
#define JEMALLOC_INTERNAL_BIN_STATS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/mutex_prof.h"
typedef struct bin_stats_s bin_stats_t;
@ -11,47 +12,47 @@ struct bin_stats_s {
* many times, resulting many increments to nrequests, but only one
* each to nmalloc and ndalloc.
*/
uint64_t nmalloc;
uint64_t ndalloc;
uint64_t nmalloc;
uint64_t ndalloc;
/*
* Number of allocation requests that correspond to the size of this
* bin. This includes requests served by tcache, though tcache only
* periodically merges into this counter.
*/
uint64_t nrequests;
uint64_t nrequests;
/*
* Current number of regions of this size class, including regions
* currently cached by tcache.
*/
size_t curregs;
size_t curregs;
/* Number of tcache fills from this bin. */
uint64_t nfills;
uint64_t nfills;
/* Number of tcache flushes to this bin. */
uint64_t nflushes;
uint64_t nflushes;
/* Total number of slabs created for this bin's size class. */
uint64_t nslabs;
uint64_t nslabs;
/*
* Total number of slabs reused by extracting them from the slabs heap
* for this bin's size class.
*/
uint64_t reslabs;
uint64_t reslabs;
/* Current number of slabs in this bin. */
size_t curslabs;
size_t curslabs;
/* Current size of nonfull slabs heap in this bin. */
size_t nonfull_slabs;
size_t nonfull_slabs;
};
typedef struct bin_stats_data_s bin_stats_data_t;
struct bin_stats_data_s {
bin_stats_t stats_data;
bin_stats_t stats_data;
mutex_prof_data_t mutex_data;
};
#endif /* JEMALLOC_INTERNAL_BIN_STATS_H */

6
include/jemalloc/internal/bin_types.h
View file

@ -1,13 +1,17 @@
#ifndef JEMALLOC_INTERNAL_BIN_TYPES_H
#define JEMALLOC_INTERNAL_BIN_TYPES_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/sc.h"
#define BIN_SHARDS_MAX (1 << EDATA_BITS_BINSHARD_WIDTH)
#define N_BIN_SHARDS_DEFAULT 1
/* Used in TSD static initializer only. Real init in arena_bind(). */
#define TSD_BINSHARDS_ZERO_INITIALIZER {{UINT8_MAX}}
#define TSD_BINSHARDS_ZERO_INITIALIZER \
{ \
{ UINT8_MAX } \
}
typedef struct tsd_binshards_s tsd_binshards_t;
struct tsd_binshards_s {

61
include/jemalloc/internal/bit_util.h
View file

@ -1,12 +1,13 @@
#ifndef JEMALLOC_INTERNAL_BIT_UTIL_H
#define JEMALLOC_INTERNAL_BIT_UTIL_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
/* Sanity check. */
#if !defined(JEMALLOC_INTERNAL_FFSLL) || !defined(JEMALLOC_INTERNAL_FFSL) \
#if !defined(JEMALLOC_INTERNAL_FFSLL) || !defined(JEMALLOC_INTERNAL_FFSL) \
|| !defined(JEMALLOC_INTERNAL_FFS)
# error JEMALLOC_INTERNAL_FFS{,L,LL} should have been defined by configure
# error JEMALLOC_INTERNAL_FFS{,L,LL} should have been defined by configure
#endif
/*
@ -34,6 +35,7 @@ ffs_u(unsigned x) {
return JEMALLOC_INTERNAL_FFS(x) - 1;
}
/* clang-format off */
#define DO_FLS_SLOW(x, suffix) do { \
util_assume(x != 0); \
x |= (x >> 1); \
@ -57,6 +59,7 @@ ffs_u(unsigned x) {
} \
return ffs_##suffix(x) - 1; \
} while(0)
/* clang-format on */
static inline unsigned
fls_llu_slow(unsigned long long x) {
@ -107,16 +110,19 @@ fls_u(unsigned x) {
}
#elif defined(_MSC_VER)
#if LG_SIZEOF_PTR == 3
#define DO_BSR64(bit, x) _BitScanReverse64(&bit, x)
#else
# if LG_SIZEOF_PTR == 3
# define DO_BSR64(bit, x) _BitScanReverse64(&bit, x)
# else
/*
* This never actually runs; we're just dodging a compiler error for the
* never-taken branch where sizeof(void *) == 8.
*/
#define DO_BSR64(bit, x) bit = 0; unreachable()
#endif
# define DO_BSR64(bit, x) \
bit = 0; \
unreachable()
# endif
/* clang-format off */
#define DO_FLS(x) do { \
if (x == 0) { \
return 8 * sizeof(x); \
@ -143,6 +149,7 @@ fls_u(unsigned x) {
} \
unreachable(); \
} while (0)
/* clang-format on */
static inline unsigned
fls_llu(unsigned long long x) {
@ -159,8 +166,8 @@ fls_u(unsigned x) {
DO_FLS(x);
}
#undef DO_FLS
#undef DO_BSR64
# undef DO_FLS
# undef DO_BSR64
#else
static inline unsigned
@ -180,9 +187,10 @@ fls_u(unsigned x) {
#endif
#if LG_SIZEOF_LONG_LONG > 3
# error "Haven't implemented popcount for 16-byte ints."
# error "Haven't implemented popcount for 16-byte ints."
#endif
/* clang-format off */
#define DO_POPCOUNT(x, type) do { \
/* \
* Algorithm from an old AMD optimization reference manual. \
@ -226,6 +234,7 @@ fls_u(unsigned x) {
x >>= ((sizeof(x) - 1) * 8); \
return (unsigned)x; \
} while(0)
/* clang-format on */
static inline unsigned
popcount_u_slow(unsigned bitmap) {
@ -277,7 +286,7 @@ popcount_llu(unsigned long long bitmap) {
*/
static inline size_t
cfs_lu(unsigned long* bitmap) {
cfs_lu(unsigned long *bitmap) {
util_assume(*bitmap != 0);
size_t bit = ffs_lu(*bitmap);
*bitmap ^= ZU(1) << bit;
@ -293,7 +302,7 @@ ffs_zu(size_t x) {
#elif LG_SIZEOF_PTR == LG_SIZEOF_LONG_LONG
return ffs_llu(x);
#else
#error No implementation for size_t ffs()
# error No implementation for size_t ffs()
#endif
}
@ -306,11 +315,10 @@ fls_zu(size_t x) {
#elif LG_SIZEOF_PTR == LG_SIZEOF_LONG_LONG
return fls_llu(x);
#else
#error No implementation for size_t fls()
# error No implementation for size_t fls()
#endif
}
static inline unsigned
ffs_u64(uint64_t x) {
#if LG_SIZEOF_LONG == 3
@ -318,7 +326,7 @@ ffs_u64(uint64_t x) {
#elif LG_SIZEOF_LONG_LONG == 3
return ffs_llu(x);
#else
#error No implementation for 64-bit ffs()
# error No implementation for 64-bit ffs()
#endif
}
@ -329,7 +337,7 @@ fls_u64(uint64_t x) {
#elif LG_SIZEOF_LONG_LONG == 3
return fls_llu(x);
#else
#error No implementation for 64-bit fls()
# error No implementation for 64-bit fls()
#endif
}
@ -338,9 +346,8 @@ ffs_u32(uint32_t x) {
#if LG_SIZEOF_INT == 2
return ffs_u(x);
#else
#error No implementation for 32-bit ffs()
# error No implementation for 32-bit ffs()
#endif
return ffs_u(x);
}
static inline unsigned
@ -348,9 +355,8 @@ fls_u32(uint32_t x) {
#if LG_SIZEOF_INT == 2
return fls_u(x);
#else
#error No implementation for 32-bit fls()
# error No implementation for 32-bit fls()
#endif
return fls_u(x);
}
static inline uint64_t
@ -370,7 +376,7 @@ pow2_ceil_u64(uint64_t x) {
static inline uint32_t
pow2_ceil_u32(uint32_t x) {
if (unlikely(x <= 1)) {
return x;
return x;
}
size_t msb_on_index = fls_u32(x - 1);
/* As above. */
@ -408,13 +414,16 @@ lg_ceil(size_t x) {
#define LG_FLOOR_2(x) (x < (1ULL << 1) ? LG_FLOOR_1(x) : 1 + LG_FLOOR_1(x >> 1))
#define LG_FLOOR_4(x) (x < (1ULL << 2) ? LG_FLOOR_2(x) : 2 + LG_FLOOR_2(x >> 2))
#define LG_FLOOR_8(x) (x < (1ULL << 4) ? LG_FLOOR_4(x) : 4 + LG_FLOOR_4(x >> 4))
#define LG_FLOOR_16(x) (x < (1ULL << 8) ? LG_FLOOR_8(x) : 8 + LG_FLOOR_8(x >> 8))
#define LG_FLOOR_32(x) (x < (1ULL << 16) ? LG_FLOOR_16(x) : 16 + LG_FLOOR_16(x >> 16))
#define LG_FLOOR_64(x) (x < (1ULL << 32) ? LG_FLOOR_32(x) : 32 + LG_FLOOR_32(x >> 32))
#define LG_FLOOR_16(x) \
(x < (1ULL << 8) ? LG_FLOOR_8(x) : 8 + LG_FLOOR_8(x >> 8))
#define LG_FLOOR_32(x) \
(x < (1ULL << 16) ? LG_FLOOR_16(x) : 16 + LG_FLOOR_16(x >> 16))
#define LG_FLOOR_64(x) \
(x < (1ULL << 32) ? LG_FLOOR_32(x) : 32 + LG_FLOOR_32(x >> 32))
#if LG_SIZEOF_PTR == 2
# define LG_FLOOR(x) LG_FLOOR_32((x))
# define LG_FLOOR(x) LG_FLOOR_32((x))
#else
# define LG_FLOOR(x) LG_FLOOR_64((x))
# define LG_FLOOR(x) LG_FLOOR_64((x))
#endif
#define LG_CEIL(x) (LG_FLOOR(x) + (((x) & ((x) - 1)) == 0 ? 0 : 1))

225
include/jemalloc/internal/bitmap.h
View file

@ -1,26 +1,27 @@
#ifndef JEMALLOC_INTERNAL_BITMAP_H
#define JEMALLOC_INTERNAL_BITMAP_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/bit_util.h"
#include "jemalloc/internal/sc.h"
typedef unsigned long bitmap_t;
#define LG_SIZEOF_BITMAP LG_SIZEOF_LONG
#define LG_SIZEOF_BITMAP LG_SIZEOF_LONG
/* Maximum bitmap bit count is 2^LG_BITMAP_MAXBITS. */
#if SC_LG_SLAB_MAXREGS > LG_CEIL(SC_NSIZES)
/* Maximum bitmap bit count is determined by maximum regions per slab. */
# define LG_BITMAP_MAXBITS SC_LG_SLAB_MAXREGS
# define LG_BITMAP_MAXBITS SC_LG_SLAB_MAXREGS
#else
/* Maximum bitmap bit count is determined by number of extent size classes. */
# define LG_BITMAP_MAXBITS LG_CEIL(SC_NSIZES)
# define LG_BITMAP_MAXBITS LG_CEIL(SC_NSIZES)
#endif
#define BITMAP_MAXBITS (ZU(1) << LG_BITMAP_MAXBITS)
#define BITMAP_MAXBITS (ZU(1) << LG_BITMAP_MAXBITS)
/* Number of bits per group. */
#define LG_BITMAP_GROUP_NBITS (LG_SIZEOF_BITMAP + 3)
#define BITMAP_GROUP_NBITS (1U << LG_BITMAP_GROUP_NBITS)
#define BITMAP_GROUP_NBITS_MASK (BITMAP_GROUP_NBITS-1)
#define LG_BITMAP_GROUP_NBITS (LG_SIZEOF_BITMAP + 3)
#define BITMAP_GROUP_NBITS (1U << LG_BITMAP_GROUP_NBITS)
#define BITMAP_GROUP_NBITS_MASK (BITMAP_GROUP_NBITS - 1)
/*
* Do some analysis on how big the bitmap is before we use a tree. For a brute
@ -28,67 +29,64 @@ typedef unsigned long bitmap_t;
* use a tree instead.
*/
#if LG_BITMAP_MAXBITS - LG_BITMAP_GROUP_NBITS > 3
# define BITMAP_USE_TREE
# define BITMAP_USE_TREE
#endif
/* Number of groups required to store a given number of bits. */
#define BITMAP_BITS2GROUPS(nbits) \
(((nbits) + BITMAP_GROUP_NBITS_MASK) >> LG_BITMAP_GROUP_NBITS)
#define BITMAP_BITS2GROUPS(nbits) \
(((nbits) + BITMAP_GROUP_NBITS_MASK) >> LG_BITMAP_GROUP_NBITS)
/*
* Number of groups required at a particular level for a given number of bits.
*/
#define BITMAP_GROUPS_L0(nbits) \
BITMAP_BITS2GROUPS(nbits)
#define BITMAP_GROUPS_L1(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(nbits))
#define BITMAP_GROUPS_L2(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits))))
#define BITMAP_GROUPS_L3(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS( \
BITMAP_BITS2GROUPS((nbits)))))
#define BITMAP_GROUPS_L4(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS( \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits))))))
#define BITMAP_GROUPS_L0(nbits) BITMAP_BITS2GROUPS(nbits)
#define BITMAP_GROUPS_L1(nbits) BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(nbits))
#define BITMAP_GROUPS_L2(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits))))
#define BITMAP_GROUPS_L3(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS( \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits)))))
#define BITMAP_GROUPS_L4(nbits) \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS( \
BITMAP_BITS2GROUPS(BITMAP_BITS2GROUPS((nbits))))))
/*
* Assuming the number of levels, number of groups required for a given number
* of bits.
*/
#define BITMAP_GROUPS_1_LEVEL(nbits) \
BITMAP_GROUPS_L0(nbits)
#define BITMAP_GROUPS_2_LEVEL(nbits) \
(BITMAP_GROUPS_1_LEVEL(nbits) + BITMAP_GROUPS_L1(nbits))
#define BITMAP_GROUPS_3_LEVEL(nbits) \
(BITMAP_GROUPS_2_LEVEL(nbits) + BITMAP_GROUPS_L2(nbits))
#define BITMAP_GROUPS_4_LEVEL(nbits) \
(BITMAP_GROUPS_3_LEVEL(nbits) + BITMAP_GROUPS_L3(nbits))
#define BITMAP_GROUPS_5_LEVEL(nbits) \
(BITMAP_GROUPS_4_LEVEL(nbits) + BITMAP_GROUPS_L4(nbits))
#define BITMAP_GROUPS_1_LEVEL(nbits) BITMAP_GROUPS_L0(nbits)
#define BITMAP_GROUPS_2_LEVEL(nbits) \
(BITMAP_GROUPS_1_LEVEL(nbits) + BITMAP_GROUPS_L1(nbits))
#define BITMAP_GROUPS_3_LEVEL(nbits) \
(BITMAP_GROUPS_2_LEVEL(nbits) + BITMAP_GROUPS_L2(nbits))
#define BITMAP_GROUPS_4_LEVEL(nbits) \
(BITMAP_GROUPS_3_LEVEL(nbits) + BITMAP_GROUPS_L3(nbits))
#define BITMAP_GROUPS_5_LEVEL(nbits) \
(BITMAP_GROUPS_4_LEVEL(nbits) + BITMAP_GROUPS_L4(nbits))
/*
* Maximum number of groups required to support LG_BITMAP_MAXBITS.
*/
#ifdef BITMAP_USE_TREE
#if LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_1_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_1_LEVEL(BITMAP_MAXBITS)
#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 2
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_2_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_2_LEVEL(BITMAP_MAXBITS)
#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 3
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_3_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_3_LEVEL(BITMAP_MAXBITS)
#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 4
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_4_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_4_LEVEL(BITMAP_MAXBITS)
#elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 5
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_5_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_5_LEVEL(BITMAP_MAXBITS)
#else
# error "Unsupported bitmap size"
#endif
# if LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_1_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_1_LEVEL(BITMAP_MAXBITS)
# elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 2
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_2_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_2_LEVEL(BITMAP_MAXBITS)
# elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 3
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_3_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_3_LEVEL(BITMAP_MAXBITS)
# elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 4
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_4_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_4_LEVEL(BITMAP_MAXBITS)
# elif LG_BITMAP_MAXBITS <= LG_BITMAP_GROUP_NBITS * 5
# define BITMAP_GROUPS(nbits) BITMAP_GROUPS_5_LEVEL(nbits)
# define BITMAP_GROUPS_MAX BITMAP_GROUPS_5_LEVEL(BITMAP_MAXBITS)
# else
# error "Unsupported bitmap size"
# endif
/*
* Maximum number of levels possible. This could be statically computed based
@ -104,42 +102,53 @@ typedef unsigned long bitmap_t;
* unused trailing entries in bitmap_info_t structures; the bitmaps themselves
* are not impacted.
*/
#define BITMAP_MAX_LEVELS 5
# define BITMAP_MAX_LEVELS 5
#define BITMAP_INFO_INITIALIZER(nbits) { \
/* nbits. */ \
nbits, \
/* nlevels. */ \
(BITMAP_GROUPS_L0(nbits) > BITMAP_GROUPS_L1(nbits)) + \
(BITMAP_GROUPS_L1(nbits) > BITMAP_GROUPS_L2(nbits)) + \
(BITMAP_GROUPS_L2(nbits) > BITMAP_GROUPS_L3(nbits)) + \
(BITMAP_GROUPS_L3(nbits) > BITMAP_GROUPS_L4(nbits)) + 1, \
/* levels. */ \
{ \
{0}, \
{BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L1(nbits) + BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L2(nbits) + BITMAP_GROUPS_L1(nbits) + \
BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L3(nbits) + BITMAP_GROUPS_L2(nbits) + \
BITMAP_GROUPS_L1(nbits) + BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L4(nbits) + BITMAP_GROUPS_L3(nbits) + \
BITMAP_GROUPS_L2(nbits) + BITMAP_GROUPS_L1(nbits) \
+ BITMAP_GROUPS_L0(nbits)} \
} \
}
# define BITMAP_INFO_INITIALIZER(nbits) \
{ \
/* nbits. */ \
nbits, /* nlevels. */ \
(BITMAP_GROUPS_L0(nbits) \
> BITMAP_GROUPS_L1(nbits)) \
+ (BITMAP_GROUPS_L1(nbits) \
> BITMAP_GROUPS_L2(nbits)) \
+ (BITMAP_GROUPS_L2(nbits) \
> BITMAP_GROUPS_L3(nbits)) \
+ (BITMAP_GROUPS_L3(nbits) \
> BITMAP_GROUPS_L4(nbits)) \
+ 1, /* levels. */ \
{ \
{0}, {BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L1(nbits) \
+ BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L2(nbits) \
+ BITMAP_GROUPS_L1(nbits) \
+ BITMAP_GROUPS_L0(nbits)}, \
{BITMAP_GROUPS_L3(nbits) \
+ BITMAP_GROUPS_L2(nbits) \
+ BITMAP_GROUPS_L1(nbits) \
+ BITMAP_GROUPS_L0(nbits)}, \
{ \
BITMAP_GROUPS_L4(nbits) \
+ BITMAP_GROUPS_L3(nbits) \
+ BITMAP_GROUPS_L2(nbits) \
+ BITMAP_GROUPS_L1(nbits) \
+ BITMAP_GROUPS_L0(nbits) \
} \
} \
}
#else /* BITMAP_USE_TREE */
#define BITMAP_GROUPS(nbits) BITMAP_BITS2GROUPS(nbits)
#define BITMAP_GROUPS_MAX BITMAP_BITS2GROUPS(BITMAP_MAXBITS)
# define BITMAP_GROUPS(nbits) BITMAP_BITS2GROUPS(nbits)
# define BITMAP_GROUPS_MAX BITMAP_BITS2GROUPS(BITMAP_MAXBITS)
#define BITMAP_INFO_INITIALIZER(nbits) { \
/* nbits. */ \
nbits, \
/* ngroups. */ \
BITMAP_BITS2GROUPS(nbits) \
}
# define BITMAP_INFO_INITIALIZER(nbits) \
{ \
/* nbits. */ \
nbits, /* ngroups. */ \
BITMAP_BITS2GROUPS(nbits) \
}
#endif /* BITMAP_USE_TREE */
@ -160,21 +169,21 @@ typedef struct bitmap_info_s {
* Only the first (nlevels+1) elements are used, and levels are ordered
* bottom to top (e.g. the bottom level is stored in levels[0]).
*/
bitmap_level_t levels[BITMAP_MAX_LEVELS+1];
#else /* BITMAP_USE_TREE */
bitmap_level_t levels[BITMAP_MAX_LEVELS + 1];
#else /* BITMAP_USE_TREE */
/* Number of groups necessary for nbits. */
size_t ngroups;
#endif /* BITMAP_USE_TREE */
} bitmap_info_t;
void bitmap_info_init(bitmap_info_t *binfo, size_t nbits);
void bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo, bool fill);
void bitmap_info_init(bitmap_info_t *binfo, size_t nbits);
void bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo, bool fill);
size_t bitmap_size(const bitmap_info_t *binfo);
static inline bool
bitmap_full(bitmap_t *bitmap, const bitmap_info_t *binfo) {
#ifdef BITMAP_USE_TREE
size_t rgoff = binfo->levels[binfo->nlevels].group_offset - 1;
size_t rgoff = binfo->levels[binfo->nlevels].group_offset - 1;
bitmap_t rg = bitmap[rgoff];
/* The bitmap is full iff the root group is 0. */
return (rg == 0);
@ -192,7 +201,7 @@ bitmap_full(bitmap_t *bitmap, const bitmap_info_t *binfo) {
static inline bool
bitmap_get(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
size_t goff;
size_t goff;
bitmap_t g;
assert(bit < binfo->nbits);
@ -203,9 +212,9 @@ bitmap_get(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
static inline void
bitmap_set(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
size_t goff;
size_t goff;
bitmap_t *gp;
bitmap_t g;
bitmap_t g;
assert(bit < binfo->nbits);
assert(!bitmap_get(bitmap, binfo, bit));
@ -244,12 +253,13 @@ bitmap_ffu(const bitmap_t *bitmap, const bitmap_info_t *binfo, size_t min_bit) {
#ifdef BITMAP_USE_TREE
size_t bit = 0;
for (unsigned level = binfo->nlevels; level--;) {
size_t lg_bits_per_group = (LG_BITMAP_GROUP_NBITS * (level +
1));
bitmap_t group = bitmap[binfo->levels[level].group_offset + (bit
>> lg_bits_per_group)];
unsigned group_nmask = (unsigned)(((min_bit > bit) ? (min_bit -
bit) : 0) >> (lg_bits_per_group - LG_BITMAP_GROUP_NBITS));
size_t lg_bits_per_group = (LG_BITMAP_GROUP_NBITS
* (level + 1));
bitmap_t group = bitmap[binfo->levels[level].group_offset
+ (bit >> lg_bits_per_group)];
unsigned group_nmask =
(unsigned)(((min_bit > bit) ? (min_bit - bit) : 0)
>> (lg_bits_per_group - LG_BITMAP_GROUP_NBITS));
assert(group_nmask <= BITMAP_GROUP_NBITS);
bitmap_t group_mask = ~((1LU << group_nmask) - 1);
bitmap_t group_masked = group & group_mask;
@ -272,25 +282,28 @@ bitmap_ffu(const bitmap_t *bitmap, const bitmap_info_t *binfo, size_t min_bit) {
}
return bitmap_ffu(bitmap, binfo, sib_base);
}
bit += ((size_t)ffs_lu(group_masked)) <<
(lg_bits_per_group - LG_BITMAP_GROUP_NBITS);
bit += ((size_t)ffs_lu(group_masked))
<< (lg_bits_per_group - LG_BITMAP_GROUP_NBITS);
}
assert(bit >= min_bit);
assert(bit < binfo->nbits);
return bit;
#else
size_t i = min_bit >> LG_BITMAP_GROUP_NBITS;
bitmap_t g = bitmap[i] & ~((1LU << (min_bit & BITMAP_GROUP_NBITS_MASK))
- 1);
size_t i = min_bit >> LG_BITMAP_GROUP_NBITS;
bitmap_t g = bitmap[i]
& ~((1LU << (min_bit & BITMAP_GROUP_NBITS_MASK)) - 1);
size_t bit;
do {
while (1) {
if (g != 0) {
bit = ffs_lu(g);
return (i << LG_BITMAP_GROUP_NBITS) + bit;
}
i++;
if (i >= binfo->ngroups) {
break;
}
g = bitmap[i];
} while (i < binfo->ngroups);
}
return binfo->nbits;
#endif
}
@ -298,7 +311,7 @@ bitmap_ffu(const bitmap_t *bitmap, const bitmap_info_t *binfo, size_t min_bit) {
/* sfu: set first unset. */
static inline size_t
bitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo) {
size_t bit;
size_t bit;
bitmap_t g;
unsigned i;
@ -328,9 +341,9 @@ bitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo) {
static inline void
bitmap_unset(bitmap_t *bitmap, const bitmap_info_t *binfo, size_t bit) {
size_t goff;
bitmap_t *gp;
bitmap_t g;
size_t goff;
bitmap_t *gp;
bitmap_t g;
UNUSED bool propagate;
assert(bit < binfo->nbits);

28
include/jemalloc/internal/buf_writer.h
View file

@ -1,6 +1,10 @@
#ifndef JEMALLOC_INTERNAL_BUF_WRITER_H
#define JEMALLOC_INTERNAL_BUF_WRITER_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#include "jemalloc/internal/tsd_types.h"
/*
* Note: when using the buffered writer, cbopaque is passed to write_cb only
* when the buffer is flushed. It would make a difference if cbopaque points
@ -12,21 +16,21 @@
typedef struct {
write_cb_t *write_cb;
void *cbopaque;
char *buf;
size_t buf_size;
size_t buf_end;
bool internal_buf;
void *cbopaque;
char *buf;
size_t buf_size;
size_t buf_end;
bool internal_buf;
} buf_writer_t;
bool buf_writer_init(tsdn_t *tsdn, buf_writer_t *buf_writer,
write_cb_t *write_cb, void *cbopaque, char *buf, size_t buf_len);
void buf_writer_flush(buf_writer_t *buf_writer);
bool buf_writer_init(tsdn_t *tsdn, buf_writer_t *buf_writer,
write_cb_t *write_cb, void *cbopaque, char *buf, size_t buf_len);
void buf_writer_flush(buf_writer_t *buf_writer);
write_cb_t buf_writer_cb;
void buf_writer_terminate(tsdn_t *tsdn, buf_writer_t *buf_writer);
void buf_writer_terminate(tsdn_t *tsdn, buf_writer_t *buf_writer);
typedef ssize_t (read_cb_t)(void *read_cbopaque, void *buf, size_t limit);
void buf_writer_pipe(buf_writer_t *buf_writer, read_cb_t *read_cb,
void *read_cbopaque);
typedef ssize_t(read_cb_t)(void *read_cbopaque, void *buf, size_t limit);
void buf_writer_pipe(
buf_writer_t *buf_writer, read_cb_t *read_cb, void *read_cbopaque);
#endif /* JEMALLOC_INTERNAL_BUF_WRITER_H */

385
include/jemalloc/internal/cache_bin.h
View file

@ -1,7 +1,10 @@
#ifndef JEMALLOC_INTERNAL_CACHE_BIN_H
#define JEMALLOC_INTERNAL_CACHE_BIN_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/jemalloc_internal_externs.h"
#include "jemalloc/internal/ql.h"
#include "jemalloc/internal/safety_check.h"
#include "jemalloc/internal/sz.h"
/*
@ -20,16 +23,20 @@
*/
typedef uint16_t cache_bin_sz_t;
#define JUNK_ADDR ((uintptr_t)0x7a7a7a7a7a7a7a7aULL)
/*
* Leave a noticeable mark pattern on the cache bin stack boundaries, in case a
* bug starts leaking those. Make it look like the junk pattern but be distinct
* from it.
*/
static const uintptr_t cache_bin_preceding_junk =
(uintptr_t)0x7a7a7a7a7a7a7a7aULL;
/* Note: a7 vs. 7a above -- this tells you which pointer leaked. */
static const uintptr_t cache_bin_trailing_junk =
(uintptr_t)0xa7a7a7a7a7a7a7a7ULL;
static const uintptr_t cache_bin_preceding_junk = JUNK_ADDR;
/* Note: JUNK_ADDR vs. JUNK_ADDR + 1 -- this tells you which pointer leaked. */
static const uintptr_t cache_bin_trailing_junk = JUNK_ADDR + 1;
/*
* A pointer used to initialize a fake stack_head for disabled small bins
* so that the enabled/disabled assessment does not rely on ncached_max.
*/
extern const uintptr_t disabled_bin;
/*
* That implies the following value, for the maximum number of items in any
@ -38,8 +45,8 @@ static const uintptr_t cache_bin_trailing_junk =
* 1 << (sizeof(cache_bin_sz_t) * 8)
* bytes spread across pointer sized objects to get the maximum.
*/
#define CACHE_BIN_NCACHED_MAX (((size_t)1 << sizeof(cache_bin_sz_t) * 8) \
/ sizeof(void *) - 1)
#define CACHE_BIN_NCACHED_MAX \
(((size_t)1 << sizeof(cache_bin_sz_t) * 8) / sizeof(void *) - 1)
/*
* This lives inside the cache_bin (for locality reasons), and is initialized
@ -101,7 +108,7 @@ struct cache_bin_s {
* Since the stack grows down, this is a higher address than
* low_bits_full.
*/
uint16_t low_bits_low_water;
cache_bin_sz_t low_bits_low_water;
/*
* The low bits of the value that stack_head will take on when the array
@ -112,7 +119,7 @@ struct cache_bin_s {
* Recall that since the stack grows down, this is the lowest available
* address in the array for caching. Only adjusted when stashing items.
*/
uint16_t low_bits_full;
cache_bin_sz_t low_bits_full;
/*
* The low bits of the value that stack_head will take on when the array
@ -121,7 +128,10 @@ struct cache_bin_s {
* The stack grows down -- this is one past the highest address in the
* array. Immutable after initialization.
*/
uint16_t low_bits_empty;
cache_bin_sz_t low_bits_empty;
/* The maximum number of cached items in the bin. */
cache_bin_info_t bin_info;
};
/*
@ -142,8 +152,8 @@ struct cache_bin_array_descriptor_s {
};
static inline void
cache_bin_array_descriptor_init(cache_bin_array_descriptor_t *descriptor,
cache_bin_t *bins) {
cache_bin_array_descriptor_init(
cache_bin_array_descriptor_t *descriptor, cache_bin_t *bins) {
ql_elm_new(descriptor, link);
descriptor->bins = bins;
}
@ -168,10 +178,41 @@ cache_bin_nonfast_aligned(const void *ptr) {
return ((uintptr_t)ptr & san_cache_bin_nonfast_mask) == 0;
}
static inline const void *
cache_bin_disabled_bin_stack(void) {
return &disabled_bin;
}
/*
* If a cache bin was zero initialized (either because it lives in static or
* thread-local storage, or was memset to 0), this function indicates whether or
* not cache_bin_init was called on it.
*/
static inline bool
cache_bin_still_zero_initialized(cache_bin_t *bin) {
return bin->stack_head == NULL;
}
static inline bool
cache_bin_disabled(cache_bin_t *bin) {
bool disabled = (bin->stack_head == cache_bin_disabled_bin_stack());
if (disabled) {
assert((uintptr_t)(*bin->stack_head) == JUNK_ADDR);
}
return disabled;
}
/* Gets ncached_max without asserting that the bin is enabled. */
static inline cache_bin_sz_t
cache_bin_ncached_max_get_unsafe(cache_bin_t *bin) {
return bin->bin_info.ncached_max;
}
/* Returns ncached_max: Upper limit on ncached. */
static inline cache_bin_sz_t
cache_bin_info_ncached_max(cache_bin_info_t *info) {
return info->ncached_max;
cache_bin_ncached_max_get(cache_bin_t *bin) {
assert(!cache_bin_disabled(bin));
return cache_bin_ncached_max_get_unsafe(bin);
}
/*
@ -181,7 +222,8 @@ cache_bin_info_ncached_max(cache_bin_info_t *info) {
* with later.
*/
static inline void
cache_bin_assert_earlier(cache_bin_t *bin, uint16_t earlier, uint16_t later) {
cache_bin_assert_earlier(
cache_bin_t *bin, cache_bin_sz_t earlier, cache_bin_sz_t later) {
if (earlier > later) {
assert(bin->low_bits_full > bin->low_bits_empty);
}
@ -193,28 +235,19 @@ cache_bin_assert_earlier(cache_bin_t *bin, uint16_t earlier, uint16_t later) {
* Does difference calculations that handle wraparound correctly. Earlier must
* be associated with the position earlier in memory.
*/
static inline uint16_t
cache_bin_diff(cache_bin_t *bin, uint16_t earlier, uint16_t later, bool racy) {
/*
* When it's racy, bin->low_bits_full can be modified concurrently. It
* can cross the uint16_t max value and become less than
* bin->low_bits_empty at the time of the check.
*/
if (!racy) {
cache_bin_assert_earlier(bin, earlier, later);
}
static inline cache_bin_sz_t
cache_bin_diff(cache_bin_t *bin, cache_bin_sz_t earlier, cache_bin_sz_t later) {
cache_bin_assert_earlier(bin, earlier, later);
return later - earlier;
}
/*
* Number of items currently cached in the bin, without checking ncached_max.
* We require specifying whether or not the request is racy or not (i.e. whether
* or not concurrent modifications are possible).
*/
static inline cache_bin_sz_t
cache_bin_ncached_get_internal(cache_bin_t *bin, bool racy) {
cache_bin_ncached_get_internal(cache_bin_t *bin) {
cache_bin_sz_t diff = cache_bin_diff(bin,
(uint16_t)(uintptr_t)bin->stack_head, bin->low_bits_empty, racy);
(cache_bin_sz_t)(uintptr_t)bin->stack_head, bin->low_bits_empty);
cache_bin_sz_t n = diff / sizeof(void *);
/*
* We have undefined behavior here; if this function is called from the
@ -225,7 +258,7 @@ cache_bin_ncached_get_internal(cache_bin_t *bin, bool racy) {
* fast paths. This should still be "safe" in the sense of generating
* the correct assembly for the foreseeable future, though.
*/
assert(n == 0 || *(bin->stack_head) != NULL || racy);
assert(n == 0 || *(bin->stack_head) != NULL);
return n;
}
@ -235,10 +268,9 @@ cache_bin_ncached_get_internal(cache_bin_t *bin, bool racy) {
* possible.
*/
static inline cache_bin_sz_t
cache_bin_ncached_get_local(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin,
/* racy */ false);
assert(n <= cache_bin_info_ncached_max(info));
cache_bin_ncached_get_local(cache_bin_t *bin) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin);
assert(n <= cache_bin_ncached_max_get(bin));
return n;
}
@ -253,10 +285,9 @@ cache_bin_ncached_get_local(cache_bin_t *bin, cache_bin_info_t *info) {
static inline void **
cache_bin_empty_position_get(cache_bin_t *bin) {
cache_bin_sz_t diff = cache_bin_diff(bin,
(uint16_t)(uintptr_t)bin->stack_head, bin->low_bits_empty,
/* racy */ false);
uintptr_t empty_bits = (uintptr_t)bin->stack_head + diff;
void **ret = (void **)empty_bits;
(cache_bin_sz_t)(uintptr_t)bin->stack_head, bin->low_bits_empty);
byte_t *empty_bits = (byte_t *)bin->stack_head + diff;
void **ret = (void **)empty_bits;
assert(ret >= bin->stack_head);
@ -273,10 +304,10 @@ cache_bin_empty_position_get(cache_bin_t *bin) {
* multithreaded environment. Currently concurrent access happens only during
* arena statistics collection.
*/
static inline uint16_t
cache_bin_low_bits_low_bound_get(cache_bin_t *bin, cache_bin_info_t *info) {
return (uint16_t)bin->low_bits_empty -
info->ncached_max * sizeof(void *);
static inline cache_bin_sz_t
cache_bin_low_bits_low_bound_get(cache_bin_t *bin) {
return (cache_bin_sz_t)bin->low_bits_empty
- cache_bin_ncached_max_get(bin) * sizeof(void *);
}
/*
@ -285,9 +316,9 @@ cache_bin_low_bits_low_bound_get(cache_bin_t *bin, cache_bin_info_t *info) {
* A pointer to the position with the lowest address of the backing array.
*/
static inline void **
cache_bin_low_bound_get(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t ncached_max = cache_bin_info_ncached_max(info);
void **ret = cache_bin_empty_position_get(bin) - ncached_max;
cache_bin_low_bound_get(cache_bin_t *bin) {
cache_bin_sz_t ncached_max = cache_bin_ncached_max_get(bin);
void **ret = cache_bin_empty_position_get(bin) - ncached_max;
assert(ret <= bin->stack_head);
return ret;
@ -298,8 +329,8 @@ cache_bin_low_bound_get(cache_bin_t *bin, cache_bin_info_t *info) {
* batch fill a nonempty cache bin.
*/
static inline void
cache_bin_assert_empty(cache_bin_t *bin, cache_bin_info_t *info) {
assert(cache_bin_ncached_get_local(bin, info) == 0);
cache_bin_assert_empty(cache_bin_t *bin) {
assert(cache_bin_ncached_get_local(bin) == 0);
assert(cache_bin_empty_position_get(bin) == bin->stack_head);
}
@ -310,18 +341,19 @@ cache_bin_assert_empty(cache_bin_t *bin, cache_bin_info_t *info) {
*/
static inline cache_bin_sz_t
cache_bin_low_water_get_internal(cache_bin_t *bin) {
return cache_bin_diff(bin, bin->low_bits_low_water,
bin->low_bits_empty, /* racy */ false) / sizeof(void *);
return cache_bin_diff(bin, bin->low_bits_low_water, bin->low_bits_empty)
/ sizeof(void *);
}
/* Returns the numeric value of low water in [0, ncached]. */
static inline cache_bin_sz_t
cache_bin_low_water_get(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_low_water_get(cache_bin_t *bin) {
cache_bin_sz_t low_water = cache_bin_low_water_get_internal(bin);
assert(low_water <= cache_bin_info_ncached_max(info));
assert(low_water <= cache_bin_ncached_get_local(bin, info));
assert(low_water <= cache_bin_ncached_max_get(bin));
assert(low_water <= cache_bin_ncached_get_local(bin));
cache_bin_assert_earlier(bin, (uint16_t)(uintptr_t)bin->stack_head,
cache_bin_assert_earlier(bin,
(cache_bin_sz_t)(uintptr_t)bin->stack_head,
bin->low_bits_low_water);
return low_water;
@ -333,12 +365,14 @@ cache_bin_low_water_get(cache_bin_t *bin, cache_bin_info_t *info) {
*/
static inline void
cache_bin_low_water_set(cache_bin_t *bin) {
bin->low_bits_low_water = (uint16_t)(uintptr_t)bin->stack_head;
assert(!cache_bin_disabled(bin));
bin->low_bits_low_water = (cache_bin_sz_t)(uintptr_t)bin->stack_head;
}
static inline void
cache_bin_low_water_adjust(cache_bin_t *bin) {
if (cache_bin_ncached_get_internal(bin, /* racy */ false)
assert(!cache_bin_disabled(bin));
if (cache_bin_ncached_get_internal(bin)
< cache_bin_low_water_get_internal(bin)) {
cache_bin_low_water_set(bin);
}
@ -358,9 +392,9 @@ cache_bin_alloc_impl(cache_bin_t *bin, bool *success, bool adjust_low_water) {
* This may read from the empty position; however the loaded value won't
* be used. It's safe because the stack has one more slot reserved.
*/
void *ret = *bin->stack_head;
uint16_t low_bits = (uint16_t)(uintptr_t)bin->stack_head;
void **new_head = bin->stack_head + 1;
void *ret = *bin->stack_head;
cache_bin_sz_t low_bits = (cache_bin_sz_t)(uintptr_t)bin->stack_head;
void **new_head = bin->stack_head + 1;
/*
* Note that the low water mark is at most empty; if we pass this check,
@ -382,7 +416,7 @@ cache_bin_alloc_impl(cache_bin_t *bin, bool *success, bool adjust_low_water) {
*/
if (likely(low_bits != bin->low_bits_empty)) {
bin->stack_head = new_head;
bin->low_bits_low_water = (uint16_t)(uintptr_t)new_head;
bin->low_bits_low_water = (cache_bin_sz_t)(uintptr_t)new_head;
*success = true;
return ret;
}
@ -410,8 +444,7 @@ cache_bin_alloc(cache_bin_t *bin, bool *success) {
JEMALLOC_ALWAYS_INLINE cache_bin_sz_t
cache_bin_alloc_batch(cache_bin_t *bin, size_t num, void **out) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin,
/* racy */ false);
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin);
if (n > num) {
n = (cache_bin_sz_t)num;
}
@ -424,7 +457,37 @@ cache_bin_alloc_batch(cache_bin_t *bin, size_t num, void **out) {
JEMALLOC_ALWAYS_INLINE bool
cache_bin_full(cache_bin_t *bin) {
return ((uint16_t)(uintptr_t)bin->stack_head == bin->low_bits_full);
return (
(cache_bin_sz_t)(uintptr_t)bin->stack_head == bin->low_bits_full);
}
/*
* Scans the allocated area of the cache_bin for the given pointer up to limit.
* Fires safety_check_fail if the ptr is found and returns true.
*/
JEMALLOC_ALWAYS_INLINE bool
cache_bin_dalloc_safety_checks(cache_bin_t *bin, void *ptr) {
if (!config_debug || opt_debug_double_free_max_scan == 0) {
return false;
}
cache_bin_sz_t ncached = cache_bin_ncached_get_internal(bin);
unsigned max_scan = opt_debug_double_free_max_scan < ncached
? opt_debug_double_free_max_scan
: ncached;
void **cur = bin->stack_head;
void **limit = cur + max_scan;
for (; cur < limit; cur++) {
if (*cur == ptr) {
safety_check_fail(
"Invalid deallocation detected: double free of "
"pointer %p\n",
ptr);
return true;
}
}
return false;
}
/*
@ -436,10 +499,14 @@ cache_bin_dalloc_easy(cache_bin_t *bin, void *ptr) {
return false;
}
if (unlikely(cache_bin_dalloc_safety_checks(bin, ptr))) {
return true;
}
bin->stack_head--;
*bin->stack_head = ptr;
cache_bin_assert_earlier(bin, bin->low_bits_full,
(uint16_t)(uintptr_t)bin->stack_head);
(cache_bin_sz_t)(uintptr_t)bin->stack_head);
return true;
}
@ -452,11 +519,12 @@ cache_bin_stash(cache_bin_t *bin, void *ptr) {
}
/* Stash at the full position, in the [full, head) range. */
uint16_t low_bits_head = (uint16_t)(uintptr_t)bin->stack_head;
cache_bin_sz_t low_bits_head = (cache_bin_sz_t)(uintptr_t)
bin->stack_head;
/* Wraparound handled as well. */
uint16_t diff = cache_bin_diff(bin, bin->low_bits_full, low_bits_head,
/* racy */ false);
*(void **)((uintptr_t)bin->stack_head - diff) = ptr;
cache_bin_sz_t diff = cache_bin_diff(
bin, bin->low_bits_full, low_bits_head);
*(void **)((byte_t *)bin->stack_head - diff) = ptr;
assert(!cache_bin_full(bin));
bin->low_bits_full += sizeof(void *);
@ -465,67 +533,101 @@ cache_bin_stash(cache_bin_t *bin, void *ptr) {
return true;
}
/*
* Get the number of stashed pointers.
*
* When called from a thread not owning the TLS (i.e. racy = true), it's
* important to keep in mind that 'bin->stack_head' and 'bin->low_bits_full' can
* be modified concurrently and almost none assertions about their values can be
* made.
*/
/* Get the number of stashed pointers. */
JEMALLOC_ALWAYS_INLINE cache_bin_sz_t
cache_bin_nstashed_get_internal(cache_bin_t *bin, cache_bin_info_t *info,
bool racy) {
cache_bin_sz_t ncached_max = cache_bin_info_ncached_max(info);
uint16_t low_bits_low_bound = cache_bin_low_bits_low_bound_get(bin,
info);
cache_bin_nstashed_get_internal(cache_bin_t *bin) {
cache_bin_sz_t ncached_max = cache_bin_ncached_max_get(bin);
cache_bin_sz_t low_bits_low_bound = cache_bin_low_bits_low_bound_get(
bin);
cache_bin_sz_t n = cache_bin_diff(bin, low_bits_low_bound,
bin->low_bits_full, racy) / sizeof(void *);
cache_bin_sz_t n = cache_bin_diff(
bin, low_bits_low_bound, bin->low_bits_full)
/ sizeof(void *);
assert(n <= ncached_max);
if (!racy) {
if (config_debug && n != 0) {
/* Below are for assertions only. */
void **low_bound = cache_bin_low_bound_get(bin, info);
void **low_bound = cache_bin_low_bound_get(bin);
assert((uint16_t)(uintptr_t)low_bound == low_bits_low_bound);
assert(
(cache_bin_sz_t)(uintptr_t)low_bound == low_bits_low_bound);
void *stashed = *(low_bound + n - 1);
bool aligned = cache_bin_nonfast_aligned(stashed);
bool aligned = cache_bin_nonfast_aligned(stashed);
#ifdef JEMALLOC_JET
/* Allow arbitrary pointers to be stashed in tests. */
aligned = true;
#endif
assert(n == 0 || (stashed != NULL && aligned));
assert(stashed != NULL && aligned);
}
return n;
}
JEMALLOC_ALWAYS_INLINE cache_bin_sz_t
cache_bin_nstashed_get_local(cache_bin_t *bin, cache_bin_info_t *info) {
cache_bin_sz_t n = cache_bin_nstashed_get_internal(bin, info,
/* racy */ false);
assert(n <= cache_bin_info_ncached_max(info));
cache_bin_nstashed_get_local(cache_bin_t *bin) {
cache_bin_sz_t n = cache_bin_nstashed_get_internal(bin);
assert(n <= cache_bin_ncached_max_get(bin));
return n;
}
/*
* Obtain a racy view of the number of items currently in the cache bin, in the
* presence of possible concurrent modifications.
*
* Note that this is the only racy function in this header. Any other functions
* are assumed to be non-racy. The "racy" term here means accessed from another
* thread (that is not the owner of the specific cache bin). This only happens
* when gathering stats (read-only). The only change because of the racy
* condition is that assertions based on mutable fields are omitted.
*
* It's important to keep in mind that 'bin->stack_head' and
* 'bin->low_bits_full' can be modified concurrently and almost no assertions
* about their values can be made.
*
* This function should not call other utility functions because the racy
* condition may cause unexpected / undefined behaviors in unverified utility
* functions. Currently, this function calls two utility functions
* cache_bin_ncached_max_get and cache_bin_low_bits_low_bound_get because
* they help access values that will not be concurrently modified.
*/
static inline void
cache_bin_nitems_get_remote(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_sz_t *ncached, cache_bin_sz_t *nstashed) {
cache_bin_sz_t n = cache_bin_ncached_get_internal(bin, /* racy */ true);
assert(n <= cache_bin_info_ncached_max(info));
cache_bin_nitems_get_remote(
cache_bin_t *bin, cache_bin_sz_t *ncached, cache_bin_sz_t *nstashed) {
/* Racy version of cache_bin_ncached_get_internal. */
cache_bin_sz_t diff = bin->low_bits_empty
- (cache_bin_sz_t)(uintptr_t)bin->stack_head;
cache_bin_sz_t n = diff / sizeof(void *);
*ncached = n;
n = cache_bin_nstashed_get_internal(bin, info, /* racy */ true);
assert(n <= cache_bin_info_ncached_max(info));
/* Racy version of cache_bin_nstashed_get_internal. */
cache_bin_sz_t low_bits_low_bound = cache_bin_low_bits_low_bound_get(
bin);
n = (bin->low_bits_full - low_bits_low_bound) / sizeof(void *);
*nstashed = n;
/* Note that cannot assert ncached + nstashed <= ncached_max (racy). */
/*
* Note that cannot assert anything regarding ncached_max because
* it can be configured on the fly and is thus racy.
*/
}
/*
* For small bins, used to calculate how many items to fill at a time.
* The final nfill is calculated by (ncached_max >> (base - offset)).
*/
typedef struct cache_bin_fill_ctl_s cache_bin_fill_ctl_t;
struct cache_bin_fill_ctl_s {
uint8_t base;
uint8_t offset;
};
/*
* Limit how many items can be flushed in a batch (Which is the upper bound
* for the nflush parameter in tcache_bin_flush_impl()).
* This is to avoid stack overflow when we do batch edata look up, which
* reserves a nflush * sizeof(emap_batch_lookup_result_t) stack variable.
*/
#define CACHE_BIN_NFLUSH_BATCH_MAX \
((VARIABLE_ARRAY_SIZE_MAX >> LG_SIZEOF_PTR) - 1)
/*
* Filling and flushing are done in batch, on arrays of void *s. For filling,
* the arrays go forward, and can be accessed with ordinary array arithmetic.
@ -546,7 +648,7 @@ cache_bin_nitems_get_remote(cache_bin_t *bin, cache_bin_info_t *info,
typedef struct cache_bin_ptr_array_s cache_bin_ptr_array_t;
struct cache_bin_ptr_array_s {
cache_bin_sz_t n;
void **ptr;
void **ptr;
};
/*
@ -558,18 +660,18 @@ struct cache_bin_ptr_array_s {
* representations is easy (since they'll require an alloca in the calling
* frame).
*/
#define CACHE_BIN_PTR_ARRAY_DECLARE(name, nval) \
cache_bin_ptr_array_t name; \
name.n = (nval)
#define CACHE_BIN_PTR_ARRAY_DECLARE(name, nval) \
cache_bin_ptr_array_t name; \
name.n = (nval)
/*
* Start a fill. The bin must be empty, and This must be followed by a
* finish_fill call before doing any alloc/dalloc operations on the bin.
*/
static inline void
cache_bin_init_ptr_array_for_fill(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nfill) {
cache_bin_assert_empty(bin, info);
cache_bin_init_ptr_array_for_fill(
cache_bin_t *bin, cache_bin_ptr_array_t *arr, cache_bin_sz_t nfill) {
cache_bin_assert_empty(bin);
arr->ptr = cache_bin_empty_position_get(bin) - nfill;
}
@ -579,15 +681,19 @@ cache_bin_init_ptr_array_for_fill(cache_bin_t *bin, cache_bin_info_t *info,
* case of OOM.
*/
static inline void
cache_bin_finish_fill(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nfilled) {
cache_bin_assert_empty(bin, info);
cache_bin_finish_fill(
cache_bin_t *bin, cache_bin_ptr_array_t *arr, cache_bin_sz_t nfilled) {
cache_bin_assert_empty(bin);
void **empty_position = cache_bin_empty_position_get(bin);
if (nfilled < arr->n) {
memmove(empty_position - nfilled, empty_position - arr->n,
nfilled * sizeof(void *));
}
bin->stack_head = empty_position - nfilled;
/* Reset the bin stats as it's merged during fill. */
if (config_stats) {
bin->tstats.nrequests = 0;
}
}
/*
@ -595,55 +701,61 @@ cache_bin_finish_fill(cache_bin_t *bin, cache_bin_info_t *info,
* everything we give them.
*/
static inline void
cache_bin_init_ptr_array_for_flush(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nflush) {
cache_bin_init_ptr_array_for_flush(
cache_bin_t *bin, cache_bin_ptr_array_t *arr, cache_bin_sz_t nflush) {
arr->ptr = cache_bin_empty_position_get(bin) - nflush;
assert(cache_bin_ncached_get_local(bin, info) == 0
|| *arr->ptr != NULL);
assert(cache_bin_ncached_get_local(bin) == 0 || *arr->ptr != NULL);
}
static inline void
cache_bin_finish_flush(cache_bin_t *bin, cache_bin_info_t *info,
cache_bin_ptr_array_t *arr, cache_bin_sz_t nflushed) {
unsigned rem = cache_bin_ncached_get_local(bin, info) - nflushed;
memmove(bin->stack_head + nflushed, bin->stack_head,
rem * sizeof(void *));
bin->stack_head = bin->stack_head + nflushed;
cache_bin_finish_flush(
cache_bin_t *bin, cache_bin_ptr_array_t *arr, cache_bin_sz_t nflushed) {
unsigned rem = cache_bin_ncached_get_local(bin) - nflushed;
memmove(
bin->stack_head + nflushed, bin->stack_head, rem * sizeof(void *));
bin->stack_head += nflushed;
cache_bin_low_water_adjust(bin);
/* Reset the bin stats as it's merged during flush. */
if (config_stats) {
bin->tstats.nrequests = 0;
}
}
static inline void
cache_bin_init_ptr_array_for_stashed(cache_bin_t *bin, szind_t binind,
cache_bin_info_t *info, cache_bin_ptr_array_t *arr,
cache_bin_sz_t nstashed) {
cache_bin_ptr_array_t *arr, cache_bin_sz_t nstashed) {
assert(nstashed > 0);
assert(cache_bin_nstashed_get_local(bin, info) == nstashed);
assert(cache_bin_nstashed_get_local(bin) == nstashed);
void **low_bound = cache_bin_low_bound_get(bin, info);
void **low_bound = cache_bin_low_bound_get(bin);
arr->ptr = low_bound;
assert(*arr->ptr != NULL);
}
static inline void
cache_bin_finish_flush_stashed(cache_bin_t *bin, cache_bin_info_t *info) {
void **low_bound = cache_bin_low_bound_get(bin, info);
cache_bin_finish_flush_stashed(cache_bin_t *bin) {
void **low_bound = cache_bin_low_bound_get(bin);
/* Reset the bin local full position. */
bin->low_bits_full = (uint16_t)(uintptr_t)low_bound;
assert(cache_bin_nstashed_get_local(bin, info) == 0);
assert(cache_bin_nstashed_get_local(bin) == 0);
/* Reset the bin stats as it's merged during flush. */
if (config_stats) {
bin->tstats.nrequests = 0;
}
}
/*
* Initialize a cache_bin_info to represent up to the given number of items in
* the cache_bins it is associated with.
*/
void cache_bin_info_init(cache_bin_info_t *bin_info,
cache_bin_sz_t ncached_max);
void cache_bin_info_init(
cache_bin_info_t *bin_info, cache_bin_sz_t ncached_max);
/*
* Given an array of initialized cache_bin_info_ts, determine how big an
* allocation is required to initialize a full set of cache_bin_ts.
*/
void cache_bin_info_compute_alloc(cache_bin_info_t *infos, szind_t ninfos,
void cache_bin_info_compute_alloc(const cache_bin_info_t *infos, szind_t ninfos,
size_t *size, size_t *alignment);
/*
@ -653,18 +765,13 @@ void cache_bin_info_compute_alloc(cache_bin_info_t *infos, szind_t ninfos,
* cache_bin_postincrement. *alloc_cur will then point immediately past the end
* of the allocation.
*/
void cache_bin_preincrement(cache_bin_info_t *infos, szind_t ninfos,
void cache_bin_preincrement(const cache_bin_info_t *infos, szind_t ninfos,
void *alloc, size_t *cur_offset);
void cache_bin_postincrement(cache_bin_info_t *infos, szind_t ninfos,
void *alloc, size_t *cur_offset);
void cache_bin_init(cache_bin_t *bin, cache_bin_info_t *info, void *alloc,
void cache_bin_postincrement(void *alloc, size_t *cur_offset);
void cache_bin_init(cache_bin_t *bin, const cache_bin_info_t *info, void *alloc,
size_t *cur_offset);
void cache_bin_init_disabled(cache_bin_t *bin, cache_bin_sz_t ncached_max);
/*
* If a cache bin was zero initialized (either because it lives in static or
* thread-local storage, or was memset to 0), this function indicates whether or
* not cache_bin_init was called on it.
*/
bool cache_bin_still_zero_initialized(cache_bin_t *bin);
bool cache_bin_stack_use_thp(void);
#endif /* JEMALLOC_INTERNAL_CACHE_BIN_H */

11
include/jemalloc/internal/ckh.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_CKH_H
#define JEMALLOC_INTERNAL_CKH_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/tsd.h"
/* Cuckoo hashing implementation. Skip to the end for the interface. */
@ -21,8 +22,8 @@
#define LG_CKH_BUCKET_CELLS (LG_CACHELINE - LG_SIZEOF_PTR - 1)
/* Typedefs to allow easy function pointer passing. */
typedef void ckh_hash_t (const void *, size_t[2]);
typedef bool ckh_keycomp_t (const void *, const void *);
typedef void ckh_hash_t(const void *, size_t[2]);
typedef bool ckh_keycomp_t(const void *, const void *);
/* Hash table cell. */
typedef struct {
@ -55,7 +56,7 @@ typedef struct {
unsigned lg_curbuckets;
/* Hash and comparison functions. */
ckh_hash_t *hash;
ckh_hash_t *hash;
ckh_keycomp_t *keycomp;
/* Hash table with 2^lg_curbuckets buckets. */
@ -88,8 +89,8 @@ bool ckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data);
* the key and value, and doesn't do any lifetime management.
*/
bool ckh_insert(tsd_t *tsd, ckh_t *ckh, const void *key, const void *data);
bool ckh_remove(tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key,
void **data);
bool ckh_remove(
tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key, void **data);
bool ckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data);
/* Some useful hash and comparison functions for strings and pointers. */

23
include/jemalloc/internal/conf.h Normal file
View file

@ -0,0 +1,23 @@
#ifndef JEMALLOC_INTERNAL_CONF_H
#define JEMALLOC_INTERNAL_CONF_H
#include "jemalloc/internal/sc.h"
void malloc_conf_init(sc_data_t *sc_data, unsigned bin_shard_sizes[SC_NBINS],
char readlink_buf[PATH_MAX + 1]);
void malloc_abort_invalid_conf(void);
#ifdef JEMALLOC_JET
extern bool had_conf_error;
bool conf_next(char const **opts_p, char const **k_p, size_t *klen_p,
char const **v_p, size_t *vlen_p);
void conf_error(
const char *msg, const char *k, size_t klen, const char *v, size_t vlen);
bool conf_handle_bool(const char *v, size_t vlen, bool *result);
bool conf_handle_signed(const char *v, size_t vlen, intmax_t min, intmax_t max,
bool check_min, bool check_max, bool clip, intmax_t *result);
bool conf_handle_char_p(const char *v, size_t vlen, char *dest, size_t dest_sz);
#endif
#endif /* JEMALLOC_INTERNAL_CONF_H */

4
include/jemalloc/internal/counter.h
View file

@ -1,12 +1,14 @@
#ifndef JEMALLOC_INTERNAL_COUNTER_H
#define JEMALLOC_INTERNAL_COUNTER_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/lockedint.h"
#include "jemalloc/internal/mutex.h"
typedef struct counter_accum_s {
LOCKEDINT_MTX_DECLARE(mtx)
locked_u64_t accumbytes;
uint64_t interval;
uint64_t interval;
} counter_accum_t;
JEMALLOC_ALWAYS_INLINE bool

141
include/jemalloc/internal/ctl.h
View file

@ -1,6 +1,10 @@
#ifndef JEMALLOC_INTERNAL_CTL_H
#define JEMALLOC_INTERNAL_CTL_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_stats.h"
#include "jemalloc/internal/background_thread_structs.h"
#include "jemalloc/internal/bin_stats.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#include "jemalloc/internal/malloc_io.h"
#include "jemalloc/internal/mutex_prof.h"
@ -9,50 +13,52 @@
#include "jemalloc/internal/stats.h"
/* Maximum ctl tree depth. */
#define CTL_MAX_DEPTH 7
#define CTL_MAX_DEPTH 7
#define CTL_MULTI_SETTING_MAX_LEN 1000
typedef struct ctl_node_s {
bool named;
} ctl_node_t;
typedef struct ctl_named_node_s {
ctl_node_t node;
ctl_node_t node;
const char *name;
/* If (nchildren == 0), this is a terminal node. */
size_t nchildren;
size_t nchildren;
const ctl_node_t *children;
int (*ctl)(tsd_t *, const size_t *, size_t, void *, size_t *, void *,
size_t);
int (*ctl)(
tsd_t *, const size_t *, size_t, void *, size_t *, void *, size_t);
} ctl_named_node_t;
typedef struct ctl_indexed_node_s {
struct ctl_node_s node;
const ctl_named_node_t *(*index)(tsdn_t *, const size_t *, size_t,
size_t);
const ctl_named_node_t *(*index)(
tsdn_t *, const size_t *, size_t, size_t);
} ctl_indexed_node_t;
typedef struct ctl_arena_stats_s {
arena_stats_t astats;
/* Aggregate stats for small size classes, based on bin stats. */
size_t allocated_small;
size_t allocated_small;
uint64_t nmalloc_small;
uint64_t ndalloc_small;
uint64_t nrequests_small;
uint64_t nfills_small;
uint64_t nflushes_small;
bin_stats_data_t bstats[SC_NBINS];
bin_stats_data_t bstats[SC_NBINS];
arena_stats_large_t lstats[SC_NSIZES - SC_NBINS];
pac_estats_t estats[SC_NPSIZES];
hpa_shard_stats_t hpastats;
sec_stats_t secstats;
pac_estats_t estats[SC_NPSIZES];
hpa_shard_stats_t hpastats;
} ctl_arena_stats_t;
typedef struct ctl_stats_s {
size_t allocated;
size_t active;
size_t metadata;
size_t metadata_edata;
size_t metadata_rtree;
size_t metadata_thp;
size_t resident;
size_t mapped;
@ -65,17 +71,17 @@ typedef struct ctl_stats_s {
typedef struct ctl_arena_s ctl_arena_t;
struct ctl_arena_s {
unsigned arena_ind;
bool initialized;
bool initialized;
ql_elm(ctl_arena_t) destroyed_link;
/* Basic stats, supported even if !config_stats. */
unsigned nthreads;
unsigned nthreads;
const char *dss;
ssize_t dirty_decay_ms;
ssize_t muzzy_decay_ms;
size_t pactive;
size_t pdirty;
size_t pmuzzy;
ssize_t dirty_decay_ms;
ssize_t muzzy_decay_ms;
size_t pactive;
size_t pdirty;
size_t pmuzzy;
/* NULL if !config_stats. */
ctl_arena_stats_t *astats;
@ -100,60 +106,67 @@ int ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp,
int ctl_nametomib(tsd_t *tsd, const char *name, size_t *mibp, size_t *miblenp);
int ctl_bymib(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
size_t *oldlenp, void *newp, size_t newlen);
int ctl_mibnametomib(tsd_t *tsd, size_t *mib, size_t miblen, const char *name,
size_t *miblenp);
int ctl_bymibname(tsd_t *tsd, size_t *mib, size_t miblen, const char *name,
size_t *miblenp, void *oldp, size_t *oldlenp, void *newp, size_t newlen);
int ctl_mibnametomib(
tsd_t *tsd, size_t *mib, size_t miblen, const char *name, size_t *miblenp);
int ctl_bymibname(tsd_t *tsd, size_t *mib, size_t miblen, const char *name,
size_t *miblenp, void *oldp, size_t *oldlenp, void *newp, size_t newlen);
bool ctl_boot(void);
void ctl_prefork(tsdn_t *tsdn);
void ctl_postfork_parent(tsdn_t *tsdn);
void ctl_postfork_child(tsdn_t *tsdn);
void ctl_mtx_assert_held(tsdn_t *tsdn);
#define xmallctl(name, oldp, oldlenp, newp, newlen) do { \
if (je_mallctl(name, oldp, oldlenp, newp, newlen) \
!= 0) { \
malloc_printf( \
"<jemalloc>: Failure in xmallctl(\"%s\", ...)\n", \
name); \
abort(); \
} \
} while (0)
#define xmallctl(name, oldp, oldlenp, newp, newlen) \
do { \
if (je_mallctl(name, oldp, oldlenp, newp, newlen) != 0) { \
malloc_printf( \
"<jemalloc>: Failure in xmallctl(\"%s\", ...)\n", \
name); \
abort(); \
} \
} while (0)
#define xmallctlnametomib(name, mibp, miblenp) do { \
if (je_mallctlnametomib(name, mibp, miblenp) != 0) { \
malloc_printf("<jemalloc>: Failure in " \
"xmallctlnametomib(\"%s\", ...)\n", name); \
abort(); \
} \
} while (0)
#define xmallctlnametomib(name, mibp, miblenp) \
do { \
if (je_mallctlnametomib(name, mibp, miblenp) != 0) { \
malloc_printf( \
"<jemalloc>: Failure in " \
"xmallctlnametomib(\"%s\", ...)\n", \
name); \
abort(); \
} \
} while (0)
#define xmallctlbymib(mib, miblen, oldp, oldlenp, newp, newlen) do { \
if (je_mallctlbymib(mib, miblen, oldp, oldlenp, newp, \
newlen) != 0) { \
malloc_write( \
"<jemalloc>: Failure in xmallctlbymib()\n"); \
abort(); \
} \
} while (0)
#define xmallctlbymib(mib, miblen, oldp, oldlenp, newp, newlen) \
do { \
if (je_mallctlbymib(mib, miblen, oldp, oldlenp, newp, newlen) \
!= 0) { \
malloc_write( \
"<jemalloc>: Failure in xmallctlbymib()\n"); \
abort(); \
} \
} while (0)
#define xmallctlmibnametomib(mib, miblen, name, miblenp) do { \
if (ctl_mibnametomib(tsd_fetch(), mib, miblen, name, miblenp) \
!= 0) { \
malloc_write( \
"<jemalloc>: Failure in ctl_mibnametomib()\n"); \
abort(); \
} \
} while (0)
#define xmallctlmibnametomib(mib, miblen, name, miblenp) \
do { \
if (ctl_mibnametomib(tsd_fetch(), mib, miblen, name, miblenp) \
!= 0) { \
malloc_write( \
"<jemalloc>: Failure in ctl_mibnametomib()\n"); \
abort(); \
} \
} while (0)
#define xmallctlbymibname(mib, miblen, name, miblenp, oldp, oldlenp, \
newp, newlen) do { \
if (ctl_bymibname(tsd_fetch(), mib, miblen, name, miblenp, \
oldp, oldlenp, newp, newlen) != 0) { \
malloc_write( \
"<jemalloc>: Failure in ctl_bymibname()\n"); \
abort(); \
} \
} while (0)
#define xmallctlbymibname( \
mib, miblen, name, miblenp, oldp, oldlenp, newp, newlen) \
do { \
if (ctl_bymibname(tsd_fetch(), mib, miblen, name, miblenp, \
oldp, oldlenp, newp, newlen) \
!= 0) { \
malloc_write( \
"<jemalloc>: Failure in ctl_bymibname()\n"); \
abort(); \
} \
} while (0)
#endif /* JEMALLOC_INTERNAL_CTL_H */

16
include/jemalloc/internal/decay.h
View file

@ -1,9 +1,11 @@
#ifndef JEMALLOC_INTERNAL_DECAY_H
#define JEMALLOC_INTERNAL_DECAY_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/smoothstep.h"
#define DECAY_UNBOUNDED_TIME_TO_PURGE ((uint64_t)-1)
#define DECAY_UNBOUNDED_TIME_TO_PURGE ((uint64_t) - 1)
/*
* The decay_t computes the number of pages we should purge at any given time.
@ -166,12 +168,12 @@ void decay_reinit(decay_t *decay, nstime_t *cur_time, ssize_t decay_ms);
/*
* Compute how many of 'npages_new' pages we would need to purge in 'time'.
*/
uint64_t decay_npages_purge_in(decay_t *decay, nstime_t *time,
size_t npages_new);
uint64_t decay_npages_purge_in(
decay_t *decay, nstime_t *time, size_t npages_new);
/* Returns true if the epoch advanced and there are pages to purge. */
bool decay_maybe_advance_epoch(decay_t *decay, nstime_t *new_time,
size_t current_npages);
bool decay_maybe_advance_epoch(
decay_t *decay, nstime_t *new_time, size_t current_npages);
/*
* Calculates wait time until a number of pages in the interval
@ -180,7 +182,7 @@ bool decay_maybe_advance_epoch(decay_t *decay, nstime_t *new_time,
* Returns number of nanoseconds or DECAY_UNBOUNDED_TIME_TO_PURGE in case of
* indefinite wait.
*/
uint64_t decay_ns_until_purge(decay_t *decay, size_t npages_current,
uint64_t npages_threshold);
uint64_t decay_ns_until_purge(
decay_t *decay, size_t npages_current, uint64_t npages_threshold);
#endif /* JEMALLOC_INTERNAL_DECAY_H */

1
include/jemalloc/internal/div.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_DIV_H
#define JEMALLOC_INTERNAL_DIV_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
/*

19
include/jemalloc/internal/ecache.h
View file

@ -1,15 +1,16 @@
#ifndef JEMALLOC_INTERNAL_ECACHE_H
#define JEMALLOC_INTERNAL_ECACHE_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/eset.h"
#include "jemalloc/internal/san.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/san.h"
typedef struct ecache_s ecache_t;
struct ecache_s {
malloc_mutex_t mtx;
eset_t eset;
eset_t guarded_eset;
eset_t eset;
eset_t guarded_eset;
/* All stored extents must be in the same state. */
extent_state_t state;
/* The index of the ehooks the ecache is associated with. */
@ -23,22 +24,22 @@ struct ecache_s {
static inline size_t
ecache_npages_get(ecache_t *ecache) {
return eset_npages_get(&ecache->eset) +
eset_npages_get(&ecache->guarded_eset);
return eset_npages_get(&ecache->eset)
+ eset_npages_get(&ecache->guarded_eset);
}
/* Get the number of extents in the given page size index. */
static inline size_t
ecache_nextents_get(ecache_t *ecache, pszind_t ind) {
return eset_nextents_get(&ecache->eset, ind) +
eset_nextents_get(&ecache->guarded_eset, ind);
return eset_nextents_get(&ecache->eset, ind)
+ eset_nextents_get(&ecache->guarded_eset, ind);
}
/* Get the sum total bytes of the extents in the given page size index. */
static inline size_t
ecache_nbytes_get(ecache_t *ecache, pszind_t ind) {
return eset_nbytes_get(&ecache->eset, ind) +
eset_nbytes_get(&ecache->guarded_eset, ind);
return eset_nbytes_get(&ecache->eset, ind)
+ eset_nbytes_get(&ecache->guarded_eset, ind);
}
static inline unsigned

389
include/jemalloc/internal/edata.h
View file

@ -1,12 +1,14 @@
#ifndef JEMALLOC_INTERNAL_EDATA_H
#define JEMALLOC_INTERNAL_EDATA_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/bin_info.h"
#include "jemalloc/internal/bit_util.h"
#include "jemalloc/internal/hpdata.h"
#include "jemalloc/internal/nstime.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/prof_types.h"
#include "jemalloc/internal/ql.h"
#include "jemalloc/internal/sc.h"
#include "jemalloc/internal/slab_data.h"
@ -19,10 +21,18 @@
*/
#define EDATA_ALIGNMENT 128
/*
* Defines how many nodes visited when enumerating the heap to search for
* qualified extents. More nodes visited may result in better choices at
* the cost of longer search time. This size should not exceed 2^16 - 1
* because we use uint16_t for accessing the queue needed for enumeration.
*/
#define ESET_ENUMERATE_MAX_NUM 32
enum extent_state_e {
extent_state_active = 0,
extent_state_dirty = 1,
extent_state_muzzy = 2,
extent_state_active = 0,
extent_state_dirty = 1,
extent_state_muzzy = 2,
extent_state_retained = 3,
extent_state_transition = 4, /* States below are intermediate. */
extent_state_merging = 5,
@ -32,7 +42,7 @@ typedef enum extent_state_e extent_state_t;
enum extent_head_state_e {
EXTENT_NOT_HEAD,
EXTENT_IS_HEAD /* See comments in ehooks_default_merge_impl(). */
EXTENT_IS_HEAD /* See comments in ehooks_default_merge_impl(). */
};
typedef enum extent_head_state_e extent_head_state_t;
@ -40,25 +50,22 @@ typedef enum extent_head_state_e extent_head_state_t;
* Which implementation of the page allocator interface, (PAI, defined in
* pai.h) owns the given extent?
*/
enum extent_pai_e {
EXTENT_PAI_PAC = 0,
EXTENT_PAI_HPA = 1
};
enum extent_pai_e { EXTENT_PAI_PAC = 0, EXTENT_PAI_HPA = 1 };
typedef enum extent_pai_e extent_pai_t;
struct e_prof_info_s {
/* Time when this was allocated. */
nstime_t e_prof_alloc_time;
nstime_t e_prof_alloc_time;
/* Allocation request size. */
size_t e_prof_alloc_size;
size_t e_prof_alloc_size;
/* Points to a prof_tctx_t. */
atomic_p_t e_prof_tctx;
atomic_p_t e_prof_tctx;
/*
* Points to a prof_recent_t for the allocation; NULL
* means the recent allocation record no longer exists.
* Protected by prof_recent_alloc_mtx.
*/
atomic_p_t e_prof_recent_alloc;
atomic_p_t e_prof_recent_alloc;
};
typedef struct e_prof_info_s e_prof_info_t;
@ -75,20 +82,20 @@ typedef struct e_prof_info_s e_prof_info_t;
*/
typedef struct edata_map_info_s edata_map_info_t;
struct edata_map_info_s {
bool slab;
bool slab;
szind_t szind;
};
typedef struct edata_cmp_summary_s edata_cmp_summary_t;
struct edata_cmp_summary_s {
uint64_t sn;
uint64_t sn;
uintptr_t addr;
};
/* Extent (span of pages). Use accessor functions for e_* fields. */
typedef struct edata_s edata_t;
ph_structs(edata_avail, edata_t);
ph_structs(edata_heap, edata_t);
ph_structs(edata_avail, edata_t, ESET_ENUMERATE_MAX_NUM);
ph_structs(edata_heap, edata_t, ESET_ENUMERATE_MAX_NUM);
struct edata_s {
/*
* Bitfield containing several fields:
@ -139,55 +146,72 @@ struct edata_s {
*
* bin_shard: the shard of the bin from which this extent came.
*/
uint64_t e_bits;
#define MASK(CURRENT_FIELD_WIDTH, CURRENT_FIELD_SHIFT) ((((((uint64_t)0x1U) << (CURRENT_FIELD_WIDTH)) - 1)) << (CURRENT_FIELD_SHIFT))
uint64_t e_bits;
#define MASK(CURRENT_FIELD_WIDTH, CURRENT_FIELD_SHIFT) \
((((((uint64_t)0x1U) << (CURRENT_FIELD_WIDTH)) - 1)) \
<< (CURRENT_FIELD_SHIFT))
#define EDATA_BITS_ARENA_WIDTH MALLOCX_ARENA_BITS
#define EDATA_BITS_ARENA_SHIFT 0
#define EDATA_BITS_ARENA_MASK MASK(EDATA_BITS_ARENA_WIDTH, EDATA_BITS_ARENA_SHIFT)
#define EDATA_BITS_ARENA_WIDTH MALLOCX_ARENA_BITS
#define EDATA_BITS_ARENA_SHIFT 0
#define EDATA_BITS_ARENA_MASK \
MASK(EDATA_BITS_ARENA_WIDTH, EDATA_BITS_ARENA_SHIFT)
#define EDATA_BITS_SLAB_WIDTH 1
#define EDATA_BITS_SLAB_SHIFT (EDATA_BITS_ARENA_WIDTH + EDATA_BITS_ARENA_SHIFT)
#define EDATA_BITS_SLAB_MASK MASK(EDATA_BITS_SLAB_WIDTH, EDATA_BITS_SLAB_SHIFT)
#define EDATA_BITS_SLAB_WIDTH 1
#define EDATA_BITS_SLAB_SHIFT (EDATA_BITS_ARENA_WIDTH + EDATA_BITS_ARENA_SHIFT)
#define EDATA_BITS_SLAB_MASK MASK(EDATA_BITS_SLAB_WIDTH, EDATA_BITS_SLAB_SHIFT)
#define EDATA_BITS_COMMITTED_WIDTH 1
#define EDATA_BITS_COMMITTED_SHIFT (EDATA_BITS_SLAB_WIDTH + EDATA_BITS_SLAB_SHIFT)
#define EDATA_BITS_COMMITTED_MASK MASK(EDATA_BITS_COMMITTED_WIDTH, EDATA_BITS_COMMITTED_SHIFT)
#define EDATA_BITS_COMMITTED_WIDTH 1
#define EDATA_BITS_COMMITTED_SHIFT \
(EDATA_BITS_SLAB_WIDTH + EDATA_BITS_SLAB_SHIFT)
#define EDATA_BITS_COMMITTED_MASK \
MASK(EDATA_BITS_COMMITTED_WIDTH, EDATA_BITS_COMMITTED_SHIFT)
#define EDATA_BITS_PAI_WIDTH 1
#define EDATA_BITS_PAI_SHIFT (EDATA_BITS_COMMITTED_WIDTH + EDATA_BITS_COMMITTED_SHIFT)
#define EDATA_BITS_PAI_MASK MASK(EDATA_BITS_PAI_WIDTH, EDATA_BITS_PAI_SHIFT)
#define EDATA_BITS_PAI_WIDTH 1
#define EDATA_BITS_PAI_SHIFT \
(EDATA_BITS_COMMITTED_WIDTH + EDATA_BITS_COMMITTED_SHIFT)
#define EDATA_BITS_PAI_MASK MASK(EDATA_BITS_PAI_WIDTH, EDATA_BITS_PAI_SHIFT)
#define EDATA_BITS_ZEROED_WIDTH 1
#define EDATA_BITS_ZEROED_SHIFT (EDATA_BITS_PAI_WIDTH + EDATA_BITS_PAI_SHIFT)
#define EDATA_BITS_ZEROED_MASK MASK(EDATA_BITS_ZEROED_WIDTH, EDATA_BITS_ZEROED_SHIFT)
#define EDATA_BITS_ZEROED_WIDTH 1
#define EDATA_BITS_ZEROED_SHIFT (EDATA_BITS_PAI_WIDTH + EDATA_BITS_PAI_SHIFT)
#define EDATA_BITS_ZEROED_MASK \
MASK(EDATA_BITS_ZEROED_WIDTH, EDATA_BITS_ZEROED_SHIFT)
#define EDATA_BITS_GUARDED_WIDTH 1
#define EDATA_BITS_GUARDED_SHIFT (EDATA_BITS_ZEROED_WIDTH + EDATA_BITS_ZEROED_SHIFT)
#define EDATA_BITS_GUARDED_MASK MASK(EDATA_BITS_GUARDED_WIDTH, EDATA_BITS_GUARDED_SHIFT)
#define EDATA_BITS_GUARDED_WIDTH 1
#define EDATA_BITS_GUARDED_SHIFT \
(EDATA_BITS_ZEROED_WIDTH + EDATA_BITS_ZEROED_SHIFT)
#define EDATA_BITS_GUARDED_MASK \
MASK(EDATA_BITS_GUARDED_WIDTH, EDATA_BITS_GUARDED_SHIFT)
#define EDATA_BITS_STATE_WIDTH 3
#define EDATA_BITS_STATE_SHIFT (EDATA_BITS_GUARDED_WIDTH + EDATA_BITS_GUARDED_SHIFT)
#define EDATA_BITS_STATE_MASK MASK(EDATA_BITS_STATE_WIDTH, EDATA_BITS_STATE_SHIFT)
#define EDATA_BITS_STATE_WIDTH 3
#define EDATA_BITS_STATE_SHIFT \
(EDATA_BITS_GUARDED_WIDTH + EDATA_BITS_GUARDED_SHIFT)
#define EDATA_BITS_STATE_MASK \
MASK(EDATA_BITS_STATE_WIDTH, EDATA_BITS_STATE_SHIFT)
#define EDATA_BITS_SZIND_WIDTH LG_CEIL(SC_NSIZES)
#define EDATA_BITS_SZIND_SHIFT (EDATA_BITS_STATE_WIDTH + EDATA_BITS_STATE_SHIFT)
#define EDATA_BITS_SZIND_MASK MASK(EDATA_BITS_SZIND_WIDTH, EDATA_BITS_SZIND_SHIFT)
#define EDATA_BITS_SZIND_WIDTH LG_CEIL(SC_NSIZES)
#define EDATA_BITS_SZIND_SHIFT (EDATA_BITS_STATE_WIDTH + EDATA_BITS_STATE_SHIFT)
#define EDATA_BITS_SZIND_MASK \
MASK(EDATA_BITS_SZIND_WIDTH, EDATA_BITS_SZIND_SHIFT)
#define EDATA_BITS_NFREE_WIDTH (SC_LG_SLAB_MAXREGS + 1)
#define EDATA_BITS_NFREE_SHIFT (EDATA_BITS_SZIND_WIDTH + EDATA_BITS_SZIND_SHIFT)
#define EDATA_BITS_NFREE_MASK MASK(EDATA_BITS_NFREE_WIDTH, EDATA_BITS_NFREE_SHIFT)
#define EDATA_BITS_NFREE_WIDTH (SC_LG_SLAB_MAXREGS + 1)
#define EDATA_BITS_NFREE_SHIFT (EDATA_BITS_SZIND_WIDTH + EDATA_BITS_SZIND_SHIFT)
#define EDATA_BITS_NFREE_MASK \
MASK(EDATA_BITS_NFREE_WIDTH, EDATA_BITS_NFREE_SHIFT)
#define EDATA_BITS_BINSHARD_WIDTH 6
#define EDATA_BITS_BINSHARD_SHIFT (EDATA_BITS_NFREE_WIDTH + EDATA_BITS_NFREE_SHIFT)
#define EDATA_BITS_BINSHARD_MASK MASK(EDATA_BITS_BINSHARD_WIDTH, EDATA_BITS_BINSHARD_SHIFT)
#define EDATA_BITS_BINSHARD_WIDTH 6
#define EDATA_BITS_BINSHARD_SHIFT \
(EDATA_BITS_NFREE_WIDTH + EDATA_BITS_NFREE_SHIFT)
#define EDATA_BITS_BINSHARD_MASK \
MASK(EDATA_BITS_BINSHARD_WIDTH, EDATA_BITS_BINSHARD_SHIFT)
#define EDATA_BITS_IS_HEAD_WIDTH 1
#define EDATA_BITS_IS_HEAD_SHIFT (EDATA_BITS_BINSHARD_WIDTH + EDATA_BITS_BINSHARD_SHIFT)
#define EDATA_BITS_IS_HEAD_MASK MASK(EDATA_BITS_IS_HEAD_WIDTH, EDATA_BITS_IS_HEAD_SHIFT)
#define EDATA_BITS_IS_HEAD_SHIFT \
(EDATA_BITS_BINSHARD_WIDTH + EDATA_BITS_BINSHARD_SHIFT)
#define EDATA_BITS_IS_HEAD_MASK \
MASK(EDATA_BITS_IS_HEAD_WIDTH, EDATA_BITS_IS_HEAD_SHIFT)
/* Pointer to the extent that this structure is responsible for. */
void *e_addr;
void *e_addr;
union {
/*
@ -197,16 +221,16 @@ struct edata_s {
*
* ssssssss [...] ssssssss ssssnnnn nnnnnnnn
*/
size_t e_size_esn;
#define EDATA_SIZE_MASK ((size_t)~(PAGE-1))
#define EDATA_ESN_MASK ((size_t)PAGE-1)
size_t e_size_esn;
#define EDATA_SIZE_MASK ((size_t) ~(PAGE - 1))
#define EDATA_ESN_MASK ((size_t)PAGE - 1)
/* Base extent size, which may not be a multiple of PAGE. */
size_t e_bsize;
size_t e_bsize;
};
/*
* If this edata is a user allocation from an HPA, it comes out of some
* pageslab (we don't yet support huegpage allocations that don't fit
* pageslab (we don't yet support hugepage allocations that don't fit
* into pageslabs). This tracks it.
*/
hpdata_t *e_ps;
@ -222,7 +246,7 @@ struct edata_s {
* List linkage used when the edata_t is active; either in
* arena's large allocations or bin_t's slabs_full.
*/
ql_elm(edata_t) ql_link_active;
ql_elm(edata_t) ql_link_active;
/*
* Pairing heap linkage. Used whenever the extent is inactive
* (in the page allocators), or when it is active and in
@ -230,7 +254,7 @@ struct edata_s {
* extent and sitting in an edata_cache.
*/
union {
edata_heap_link_t heap_link;
edata_heap_link_t heap_link;
edata_avail_link_t avail_link;
};
};
@ -243,10 +267,10 @@ struct edata_s {
*/
ql_elm(edata_t) ql_link_inactive;
/* Small region slab metadata. */
slab_data_t e_slab_data;
slab_data_t e_slab_data;
/* Profiling data, used for large objects. */
e_prof_info_t e_prof_info;
e_prof_info_t e_prof_info;
};
};
@ -255,8 +279,8 @@ TYPED_LIST(edata_list_inactive, edata_t, ql_link_inactive)
static inline unsigned
edata_arena_ind_get(const edata_t *edata) {
unsigned arena_ind = (unsigned)((edata->e_bits &
EDATA_BITS_ARENA_MASK) >> EDATA_BITS_ARENA_SHIFT);
unsigned arena_ind = (unsigned)((edata->e_bits & EDATA_BITS_ARENA_MASK)
>> EDATA_BITS_ARENA_SHIFT);
assert(arena_ind < MALLOCX_ARENA_LIMIT);
return arena_ind;
@ -264,8 +288,8 @@ edata_arena_ind_get(const edata_t *edata) {
static inline szind_t
edata_szind_get_maybe_invalid(const edata_t *edata) {
szind_t szind = (szind_t)((edata->e_bits & EDATA_BITS_SZIND_MASK) >>
EDATA_BITS_SZIND_SHIFT);
szind_t szind = (szind_t)((edata->e_bits & EDATA_BITS_SZIND_MASK)
>> EDATA_BITS_SZIND_SHIFT);
assert(szind <= SC_NSIZES);
return szind;
}
@ -279,13 +303,61 @@ edata_szind_get(const edata_t *edata) {
static inline size_t
edata_usize_get(const edata_t *edata) {
return sz_index2size(edata_szind_get(edata));
assert(edata != NULL);
/*
* When sz_large_size_classes_disabled() is true, two cases:
* 1. if usize_from_ind is not smaller than SC_LARGE_MINCLASS,
* usize_from_size is accurate;
* 2. otherwise, usize_from_ind is accurate.
*
* When sz_large_size_classes_disabled() is not true, the two should be the
* same when usize_from_ind is not smaller than SC_LARGE_MINCLASS.
*
* Note sampled small allocs will be promoted. Their extent size is
* recorded in edata_size_get(edata), while their szind reflects the
* true usize. Thus, usize retrieved here is still accurate for
* sampled small allocs.
*/
szind_t szind = edata_szind_get(edata);
#ifdef JEMALLOC_JET
/*
* Double free is invalid and results in undefined behavior. However,
* for double free tests to end gracefully, return an invalid usize
* when szind shows the edata is not active, i.e., szind == SC_NSIZES.
*/
if (unlikely(szind == SC_NSIZES)) {
return SC_LARGE_MAXCLASS + 1;
}
#endif
if (!sz_large_size_classes_disabled() || szind < SC_NBINS) {
size_t usize_from_ind = sz_index2size(szind);
if (!sz_large_size_classes_disabled()
&& usize_from_ind >= SC_LARGE_MINCLASS) {
size_t size = (edata->e_size_esn & EDATA_SIZE_MASK);
assert(size > sz_large_pad);
size_t usize_from_size = size - sz_large_pad;
assert(usize_from_ind == usize_from_size);
}
return usize_from_ind;
}
size_t size = (edata->e_size_esn & EDATA_SIZE_MASK);
assert(size > sz_large_pad);
size_t usize_from_size = size - sz_large_pad;
/*
* no matter large size classes disabled or not, usize retrieved from
* size is not accurate when smaller than SC_LARGE_MINCLASS.
*/
assert(usize_from_size >= SC_LARGE_MINCLASS);
return usize_from_size;
}
static inline unsigned
edata_binshard_get(const edata_t *edata) {
unsigned binshard = (unsigned)((edata->e_bits &
EDATA_BITS_BINSHARD_MASK) >> EDATA_BITS_BINSHARD_SHIFT);
unsigned binshard = (unsigned)((edata->e_bits
& EDATA_BITS_BINSHARD_MASK)
>> EDATA_BITS_BINSHARD_SHIFT);
assert(binshard < bin_infos[edata_szind_get(edata)].n_shards);
return binshard;
}
@ -297,58 +369,58 @@ edata_sn_get(const edata_t *edata) {
static inline extent_state_t
edata_state_get(const edata_t *edata) {
return (extent_state_t)((edata->e_bits & EDATA_BITS_STATE_MASK) >>
EDATA_BITS_STATE_SHIFT);
return (extent_state_t)((edata->e_bits & EDATA_BITS_STATE_MASK)
>> EDATA_BITS_STATE_SHIFT);
}
static inline bool
edata_guarded_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_GUARDED_MASK) >>
EDATA_BITS_GUARDED_SHIFT);
return (bool)((edata->e_bits & EDATA_BITS_GUARDED_MASK)
>> EDATA_BITS_GUARDED_SHIFT);
}
static inline bool
edata_zeroed_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_ZEROED_MASK) >>
EDATA_BITS_ZEROED_SHIFT);
return (bool)((edata->e_bits & EDATA_BITS_ZEROED_MASK)
>> EDATA_BITS_ZEROED_SHIFT);
}
static inline bool
edata_committed_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_COMMITTED_MASK) >>
EDATA_BITS_COMMITTED_SHIFT);
return (bool)((edata->e_bits & EDATA_BITS_COMMITTED_MASK)
>> EDATA_BITS_COMMITTED_SHIFT);
}
static inline extent_pai_t
edata_pai_get(const edata_t *edata) {
return (extent_pai_t)((edata->e_bits & EDATA_BITS_PAI_MASK) >>
EDATA_BITS_PAI_SHIFT);
return (extent_pai_t)((edata->e_bits & EDATA_BITS_PAI_MASK)
>> EDATA_BITS_PAI_SHIFT);
}
static inline bool
edata_slab_get(const edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_SLAB_MASK) >>
EDATA_BITS_SLAB_SHIFT);
return (bool)((edata->e_bits & EDATA_BITS_SLAB_MASK)
>> EDATA_BITS_SLAB_SHIFT);
}
static inline unsigned
edata_nfree_get(const edata_t *edata) {
assert(edata_slab_get(edata));
return (unsigned)((edata->e_bits & EDATA_BITS_NFREE_MASK) >>
EDATA_BITS_NFREE_SHIFT);
return (unsigned)((edata->e_bits & EDATA_BITS_NFREE_MASK)
>> EDATA_BITS_NFREE_SHIFT);
}
static inline void *
edata_base_get(const edata_t *edata) {
assert(edata->e_addr == PAGE_ADDR2BASE(edata->e_addr) ||
!edata_slab_get(edata));
assert(edata->e_addr == PAGE_ADDR2BASE(edata->e_addr)
|| !edata_slab_get(edata));
return PAGE_ADDR2BASE(edata->e_addr);
}
static inline void *
edata_addr_get(const edata_t *edata) {
assert(edata->e_addr == PAGE_ADDR2BASE(edata->e_addr) ||
!edata_slab_get(edata));
assert(edata->e_addr == PAGE_ADDR2BASE(edata->e_addr)
|| !edata_slab_get(edata));
return edata->e_addr;
}
@ -375,19 +447,19 @@ edata_ps_get(const edata_t *edata) {
static inline void *
edata_before_get(const edata_t *edata) {
return (void *)((uintptr_t)edata_base_get(edata) - PAGE);
return (void *)((byte_t *)edata_base_get(edata) - PAGE);
}
static inline void *
edata_last_get(const edata_t *edata) {
return (void *)((uintptr_t)edata_base_get(edata) +
edata_size_get(edata) - PAGE);
return (void *)((byte_t *)edata_base_get(edata) + edata_size_get(edata)
- PAGE);
}
static inline void *
edata_past_get(const edata_t *edata) {
return (void *)((uintptr_t)edata_base_get(edata) +
edata_size_get(edata));
return (
void *)((byte_t *)edata_base_get(edata) + edata_size_get(edata));
}
static inline slab_data_t *
@ -404,8 +476,8 @@ edata_slab_data_get_const(const edata_t *edata) {
static inline prof_tctx_t *
edata_prof_tctx_get(const edata_t *edata) {
return (prof_tctx_t *)atomic_load_p(&edata->e_prof_info.e_prof_tctx,
ATOMIC_ACQUIRE);
return (prof_tctx_t *)atomic_load_p(
&edata->e_prof_info.e_prof_tctx, ATOMIC_ACQUIRE);
}
static inline const nstime_t *
@ -426,16 +498,16 @@ edata_prof_recent_alloc_get_dont_call_directly(const edata_t *edata) {
static inline void
edata_arena_ind_set(edata_t *edata, unsigned arena_ind) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_ARENA_MASK) |
((uint64_t)arena_ind << EDATA_BITS_ARENA_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_ARENA_MASK)
| ((uint64_t)arena_ind << EDATA_BITS_ARENA_SHIFT);
}
static inline void
edata_binshard_set(edata_t *edata, unsigned binshard) {
/* The assertion assumes szind is set already. */
assert(binshard < bin_infos[edata_szind_get(edata)].n_shards);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_BINSHARD_MASK) |
((uint64_t)binshard << EDATA_BITS_BINSHARD_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_BINSHARD_MASK)
| ((uint64_t)binshard << EDATA_BITS_BINSHARD_SHIFT);
}
static inline void
@ -451,8 +523,8 @@ edata_size_set(edata_t *edata, size_t size) {
static inline void
edata_esn_set(edata_t *edata, size_t esn) {
edata->e_size_esn = (edata->e_size_esn & ~EDATA_ESN_MASK) | (esn &
EDATA_ESN_MASK);
edata->e_size_esn = (edata->e_size_esn & ~EDATA_ESN_MASK)
| (esn & EDATA_ESN_MASK);
}
static inline void
@ -469,25 +541,26 @@ edata_ps_set(edata_t *edata, hpdata_t *ps) {
static inline void
edata_szind_set(edata_t *edata, szind_t szind) {
assert(szind <= SC_NSIZES); /* SC_NSIZES means "invalid". */
edata->e_bits = (edata->e_bits & ~EDATA_BITS_SZIND_MASK) |
((uint64_t)szind << EDATA_BITS_SZIND_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_SZIND_MASK)
| ((uint64_t)szind << EDATA_BITS_SZIND_SHIFT);
}
static inline void
edata_nfree_set(edata_t *edata, unsigned nfree) {
assert(edata_slab_get(edata));
edata->e_bits = (edata->e_bits & ~EDATA_BITS_NFREE_MASK) |
((uint64_t)nfree << EDATA_BITS_NFREE_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_NFREE_MASK)
| ((uint64_t)nfree << EDATA_BITS_NFREE_SHIFT);
}
static inline void
edata_nfree_binshard_set(edata_t *edata, unsigned nfree, unsigned binshard) {
/* The assertion assumes szind is set already. */
assert(binshard < bin_infos[edata_szind_get(edata)].n_shards);
edata->e_bits = (edata->e_bits &
(~EDATA_BITS_NFREE_MASK & ~EDATA_BITS_BINSHARD_MASK)) |
((uint64_t)binshard << EDATA_BITS_BINSHARD_SHIFT) |
((uint64_t)nfree << EDATA_BITS_NFREE_SHIFT);
edata->e_bits = (edata->e_bits
& (~EDATA_BITS_NFREE_MASK
& ~EDATA_BITS_BINSHARD_MASK))
| ((uint64_t)binshard << EDATA_BITS_BINSHARD_SHIFT)
| ((uint64_t)nfree << EDATA_BITS_NFREE_SHIFT);
}
static inline void
@ -515,38 +588,38 @@ edata_sn_set(edata_t *edata, uint64_t sn) {
static inline void
edata_state_set(edata_t *edata, extent_state_t state) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_STATE_MASK) |
((uint64_t)state << EDATA_BITS_STATE_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_STATE_MASK)
| ((uint64_t)state << EDATA_BITS_STATE_SHIFT);
}
static inline void
edata_guarded_set(edata_t *edata, bool guarded) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_GUARDED_MASK) |
((uint64_t)guarded << EDATA_BITS_GUARDED_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_GUARDED_MASK)
| ((uint64_t)guarded << EDATA_BITS_GUARDED_SHIFT);
}
static inline void
edata_zeroed_set(edata_t *edata, bool zeroed) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_ZEROED_MASK) |
((uint64_t)zeroed << EDATA_BITS_ZEROED_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_ZEROED_MASK)
| ((uint64_t)zeroed << EDATA_BITS_ZEROED_SHIFT);
}
static inline void
edata_committed_set(edata_t *edata, bool committed) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_COMMITTED_MASK) |
((uint64_t)committed << EDATA_BITS_COMMITTED_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_COMMITTED_MASK)
| ((uint64_t)committed << EDATA_BITS_COMMITTED_SHIFT);
}
static inline void
edata_pai_set(edata_t *edata, extent_pai_t pai) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_PAI_MASK) |
((uint64_t)pai << EDATA_BITS_PAI_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_PAI_MASK)
| ((uint64_t)pai << EDATA_BITS_PAI_SHIFT);
}
static inline void
edata_slab_set(edata_t *edata, bool slab) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_SLAB_MASK) |
((uint64_t)slab << EDATA_BITS_SLAB_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_SLAB_MASK)
| ((uint64_t)slab << EDATA_BITS_SLAB_SHIFT);
}
static inline void
@ -565,22 +638,22 @@ edata_prof_alloc_size_set(edata_t *edata, size_t size) {
}
static inline void
edata_prof_recent_alloc_set_dont_call_directly(edata_t *edata,
prof_recent_t *recent_alloc) {
edata_prof_recent_alloc_set_dont_call_directly(
edata_t *edata, prof_recent_t *recent_alloc) {
atomic_store_p(&edata->e_prof_info.e_prof_recent_alloc, recent_alloc,
ATOMIC_RELAXED);
}
static inline bool
edata_is_head_get(edata_t *edata) {
return (bool)((edata->e_bits & EDATA_BITS_IS_HEAD_MASK) >>
EDATA_BITS_IS_HEAD_SHIFT);
return (bool)((edata->e_bits & EDATA_BITS_IS_HEAD_MASK)
>> EDATA_BITS_IS_HEAD_SHIFT);
}
static inline void
edata_is_head_set(edata_t *edata, bool is_head) {
edata->e_bits = (edata->e_bits & ~EDATA_BITS_IS_HEAD_MASK) |
((uint64_t)is_head << EDATA_BITS_IS_HEAD_SHIFT);
edata->e_bits = (edata->e_bits & ~EDATA_BITS_IS_HEAD_MASK)
| ((uint64_t)is_head << EDATA_BITS_IS_HEAD_SHIFT);
}
static inline bool
@ -619,7 +692,8 @@ edata_init(edata_t *edata, unsigned arena_ind, void *addr, size_t size,
}
static inline void
edata_binit(edata_t *edata, void *addr, size_t bsize, uint64_t sn) {
edata_binit(
edata_t *edata, void *addr, size_t bsize, uint64_t sn, bool reused) {
edata_arena_ind_set(edata, (1U << MALLOCX_ARENA_BITS) - 1);
edata_addr_set(edata, addr);
edata_bsize_set(edata, bsize);
@ -627,7 +701,8 @@ edata_binit(edata_t *edata, void *addr, size_t bsize, uint64_t sn) {
edata_szind_set(edata, SC_NSIZES);
edata_sn_set(edata, sn);
edata_state_set(edata, extent_state_active);
edata_guarded_set(edata, false);
/* See comments in base_edata_is_reused. */
edata_guarded_set(edata, reused);
edata_zeroed_set(edata, true);
edata_committed_set(edata, true);
/*
@ -656,20 +731,47 @@ edata_ead_comp(const edata_t *a, const edata_t *b) {
static inline edata_cmp_summary_t
edata_cmp_summary_get(const edata_t *edata) {
return (edata_cmp_summary_t){edata_sn_get(edata),
(uintptr_t)edata_addr_get(edata)};
edata_cmp_summary_t result;
result.sn = edata_sn_get(edata);
result.addr = (uintptr_t)edata_addr_get(edata);
return result;
}
#ifdef JEMALLOC_HAVE_INT128
JEMALLOC_ALWAYS_INLINE unsigned __int128
edata_cmp_summary_encode(edata_cmp_summary_t src) {
return ((unsigned __int128)src.sn << 64) | src.addr;
}
static inline int
edata_cmp_summary_comp(edata_cmp_summary_t a, edata_cmp_summary_t b) {
int ret;
ret = (a.sn > b.sn) - (a.sn < b.sn);
if (ret != 0) {
return ret;
}
ret = (a.addr > b.addr) - (a.addr < b.addr);
return ret;
unsigned __int128 a_encoded = edata_cmp_summary_encode(a);
unsigned __int128 b_encoded = edata_cmp_summary_encode(b);
if (a_encoded < b_encoded)
return -1;
if (a_encoded == b_encoded)
return 0;
return 1;
}
#else
static inline int
edata_cmp_summary_comp(edata_cmp_summary_t a, edata_cmp_summary_t b) {
/*
* Logically, what we're doing here is comparing based on `.sn`, and
* falling back to comparing on `.addr` in the case that `a.sn == b.sn`.
* We accomplish this by multiplying the result of the `.sn` comparison
* by 2, so that so long as it is not 0, it will dominate the `.addr`
* comparison in determining the sign of the returned result value.
* The justification for doing things this way is that this is
* branchless - all of the branches that would be present in a
* straightforward implementation are common cases, and thus the branch
* prediction accuracy is not great. As a result, this implementation
* is measurably faster (by around 30%).
*/
return (2 * ((a.sn > b.sn) - (a.sn < b.sn)))
+ ((a.addr > b.addr) - (a.addr < b.addr));
}
#endif
static inline int
edata_snad_comp(const edata_t *a, const edata_t *b) {
@ -681,18 +783,13 @@ edata_snad_comp(const edata_t *a, const edata_t *b) {
static inline int
edata_esnead_comp(const edata_t *a, const edata_t *b) {
int ret;
ret = edata_esn_comp(a, b);
if (ret != 0) {
return ret;
}
ret = edata_ead_comp(a, b);
return ret;
/*
* Similar to `edata_cmp_summary_comp`, we've opted for a
* branchless implementation for the sake of performance.
*/
return (2 * edata_esn_comp(a, b)) + edata_ead_comp(a, b);
}
ph_proto(, edata_avail, edata_t)
ph_proto(, edata_heap, edata_t)
ph_proto(, edata_avail, edata_t) ph_proto(, edata_heap, edata_t)
#endif /* JEMALLOC_INTERNAL_EDATA_H */

17
include/jemalloc/internal/edata_cache.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_EDATA_CACHE_H
#define JEMALLOC_INTERNAL_EDATA_CACHE_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/base.h"
/* For tests only. */
@ -14,13 +15,13 @@
typedef struct edata_cache_s edata_cache_t;
struct edata_cache_s {
edata_avail_t avail;
atomic_zu_t count;
edata_avail_t avail;
atomic_zu_t count;
malloc_mutex_t mtx;
base_t *base;
base_t *base;
};
bool edata_cache_init(edata_cache_t *edata_cache, base_t *base);
bool edata_cache_init(edata_cache_t *edata_cache, base_t *base);
edata_t *edata_cache_get(tsdn_t *tsdn, edata_cache_t *edata_cache);
void edata_cache_put(tsdn_t *tsdn, edata_cache_t *edata_cache, edata_t *edata);
@ -36,14 +37,14 @@ void edata_cache_postfork_child(tsdn_t *tsdn, edata_cache_t *edata_cache);
typedef struct edata_cache_fast_s edata_cache_fast_t;
struct edata_cache_fast_s {
edata_list_inactive_t list;
edata_cache_t *fallback;
bool disabled;
edata_cache_t *fallback;
bool disabled;
};
void edata_cache_fast_init(edata_cache_fast_t *ecs, edata_cache_t *fallback);
edata_t *edata_cache_fast_get(tsdn_t *tsdn, edata_cache_fast_t *ecs);
void edata_cache_fast_put(tsdn_t *tsdn, edata_cache_fast_t *ecs,
edata_t *edata);
void edata_cache_fast_put(
tsdn_t *tsdn, edata_cache_fast_t *ecs, edata_t *edata);
void edata_cache_fast_disable(tsdn_t *tsdn, edata_cache_fast_t *ecs);
#endif /* JEMALLOC_INTERNAL_EDATA_CACHE_H */

44
include/jemalloc/internal/ehooks.h
View file

@ -1,8 +1,11 @@
#ifndef JEMALLOC_INTERNAL_EHOOKS_H
#define JEMALLOC_INTERNAL_EHOOKS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/extent_mmap.h"
#include "jemalloc/internal/tsd.h"
#include "jemalloc/internal/tsd_types.h"
/*
* This module is the internal interface to the extent hooks (both
@ -43,17 +46,17 @@ extern const extent_hooks_t ehooks_default_extent_hooks;
*/
void *ehooks_default_alloc_impl(tsdn_t *tsdn, void *new_addr, size_t size,
size_t alignment, bool *zero, bool *commit, unsigned arena_ind);
bool ehooks_default_dalloc_impl(void *addr, size_t size);
void ehooks_default_destroy_impl(void *addr, size_t size);
bool ehooks_default_commit_impl(void *addr, size_t offset, size_t length);
bool ehooks_default_decommit_impl(void *addr, size_t offset, size_t length);
bool ehooks_default_dalloc_impl(void *addr, size_t size);
void ehooks_default_destroy_impl(void *addr, size_t size);
bool ehooks_default_commit_impl(void *addr, size_t offset, size_t length);
bool ehooks_default_decommit_impl(void *addr, size_t offset, size_t length);
#ifdef PAGES_CAN_PURGE_LAZY
bool ehooks_default_purge_lazy_impl(void *addr, size_t offset, size_t length);
#endif
#ifdef PAGES_CAN_PURGE_FORCED
bool ehooks_default_purge_forced_impl(void *addr, size_t offset, size_t length);
#endif
bool ehooks_default_split_impl();
bool ehooks_default_split_impl(void);
/*
* Merge is the only default extent hook we declare -- see the comment in
* ehooks_merge.
@ -113,8 +116,8 @@ ehooks_get_extent_hooks_ptr(ehooks_t *ehooks) {
static inline bool
ehooks_are_default(ehooks_t *ehooks) {
return ehooks_get_extent_hooks_ptr(ehooks) ==
&ehooks_default_extent_hooks;
return ehooks_get_extent_hooks_ptr(ehooks)
== &ehooks_default_extent_hooks;
}
/*
@ -186,16 +189,15 @@ ehooks_debug_zero_check(void *addr, size_t size) {
}
}
static inline void *
ehooks_alloc(tsdn_t *tsdn, ehooks_t *ehooks, void *new_addr, size_t size,
size_t alignment, bool *zero, bool *commit) {
bool orig_zero = *zero;
void *ret;
bool orig_zero = *zero;
void *ret;
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ret = ehooks_default_alloc_impl(tsdn, new_addr, size,
alignment, zero, commit, ehooks_ind_get(ehooks));
ret = ehooks_default_alloc_impl(tsdn, new_addr, size, alignment,
zero, commit, ehooks_ind_get(ehooks));
} else {
ehooks_pre_reentrancy(tsdn);
ret = extent_hooks->alloc(extent_hooks, new_addr, size,
@ -211,8 +213,8 @@ ehooks_alloc(tsdn_t *tsdn, ehooks_t *ehooks, void *new_addr, size_t size,
}
static inline bool
ehooks_dalloc(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
bool committed) {
ehooks_dalloc(
tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size, bool committed) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
return ehooks_default_dalloc_impl(addr, size);
@ -228,8 +230,8 @@ ehooks_dalloc(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
}
static inline void
ehooks_destroy(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
bool committed) {
ehooks_destroy(
tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size, bool committed) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
ehooks_default_destroy_impl(addr, size);
@ -247,15 +249,15 @@ static inline bool
ehooks_commit(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size,
size_t offset, size_t length) {
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
bool err;
bool err;
if (extent_hooks == &ehooks_default_extent_hooks) {
err = ehooks_default_commit_impl(addr, offset, length);
} else if (extent_hooks->commit == NULL) {
err = true;
} else {
ehooks_pre_reentrancy(tsdn);
err = extent_hooks->commit(extent_hooks, addr, size,
offset, length, ehooks_ind_get(ehooks));
err = extent_hooks->commit(extent_hooks, addr, size, offset,
length, ehooks_ind_get(ehooks));
ehooks_post_reentrancy(tsdn);
}
if (!err) {
@ -381,7 +383,7 @@ ehooks_zero(tsdn_t *tsdn, ehooks_t *ehooks, void *addr, size_t size) {
static inline bool
ehooks_guard(tsdn_t *tsdn, ehooks_t *ehooks, void *guard1, void *guard2) {
bool err;
bool err;
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {
@ -396,7 +398,7 @@ ehooks_guard(tsdn_t *tsdn, ehooks_t *ehooks, void *guard1, void *guard2) {
static inline bool
ehooks_unguard(tsdn_t *tsdn, ehooks_t *ehooks, void *guard1, void *guard2) {
bool err;
bool err;
extent_hooks_t *extent_hooks = ehooks_get_extent_hooks_ptr(ehooks);
if (extent_hooks == &ehooks_default_extent_hooks) {

130
include/jemalloc/internal/emap.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_EMAP_H
#define JEMALLOC_INTERNAL_EMAP_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/base.h"
#include "jemalloc/internal/rtree.h"
@ -9,9 +10,9 @@
* EMAP_DECLARE_RTREE_CTX;
* in uses will avoid empty-statement warnings.
*/
#define EMAP_DECLARE_RTREE_CTX \
rtree_ctx_t rtree_ctx_fallback; \
rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback)
#define EMAP_DECLARE_RTREE_CTX \
rtree_ctx_t rtree_ctx_fallback; \
rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback)
typedef struct emap_s emap_t;
struct emap_s {
@ -19,26 +20,27 @@ struct emap_s {
};
/* Used to pass rtree lookup context down the path. */
typedef struct emap_alloc_ctx_t emap_alloc_ctx_t;
struct emap_alloc_ctx_t {
typedef struct emap_alloc_ctx_s emap_alloc_ctx_t;
struct emap_alloc_ctx_s {
size_t usize;
szind_t szind;
bool slab;
bool slab;
};
typedef struct emap_full_alloc_ctx_s emap_full_alloc_ctx_t;
struct emap_full_alloc_ctx_s {
szind_t szind;
bool slab;
szind_t szind;
bool slab;
edata_t *edata;
};
bool emap_init(emap_t *emap, base_t *base, bool zeroed);
void emap_remap(tsdn_t *tsdn, emap_t *emap, edata_t *edata, szind_t szind,
bool slab);
void emap_remap(
tsdn_t *tsdn, emap_t *emap, edata_t *edata, szind_t szind, bool slab);
void emap_update_edata_state(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
extent_state_t state);
void emap_update_edata_state(
tsdn_t *tsdn, emap_t *emap, edata_t *edata, extent_state_t state);
/*
* The two acquire functions below allow accessing neighbor edatas, if it's safe
@ -60,16 +62,16 @@ edata_t *emap_try_acquire_edata_neighbor(tsdn_t *tsdn, emap_t *emap,
bool forward);
edata_t *emap_try_acquire_edata_neighbor_expand(tsdn_t *tsdn, emap_t *emap,
edata_t *edata, extent_pai_t pai, extent_state_t expected_state);
void emap_release_edata(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
extent_state_t new_state);
void emap_release_edata(
tsdn_t *tsdn, emap_t *emap, edata_t *edata, extent_state_t new_state);
/*
* Associate the given edata with its beginning and end address, setting the
* szind and slab info appropriately.
* Returns true on error (i.e. resource exhaustion).
*/
bool emap_register_boundary(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
szind_t szind, bool slab);
bool emap_register_boundary(
tsdn_t *tsdn, emap_t *emap, edata_t *edata, szind_t szind, bool slab);
/*
* Does the same thing, but with the interior of the range, for slab
@ -90,8 +92,8 @@ bool emap_register_boundary(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
* touched, so no allocation is necessary to fill the interior once the boundary
* has been touched.
*/
void emap_register_interior(tsdn_t *tsdn, emap_t *emap, edata_t *edata,
szind_t szind);
void emap_register_interior(
tsdn_t *tsdn, emap_t *emap, edata_t *edata, szind_t szind);
void emap_deregister_boundary(tsdn_t *tsdn, emap_t *emap, edata_t *edata);
void emap_deregister_interior(tsdn_t *tsdn, emap_t *emap, edata_t *edata);
@ -159,8 +161,8 @@ emap_edata_in_transition(tsdn_t *tsdn, emap_t *emap, edata_t *edata) {
emap_assert_mapped(tsdn, emap, edata);
EMAP_DECLARE_RTREE_CTX;
rtree_contents_t contents = rtree_read(tsdn, &emap->rtree, rtree_ctx,
(uintptr_t)edata_base_get(edata));
rtree_contents_t contents = rtree_read(
tsdn, &emap->rtree, rtree_ctx, (uintptr_t)edata_base_get(edata));
return edata_state_in_transition(contents.metadata.state);
}
@ -185,16 +187,16 @@ emap_edata_is_acquired(tsdn_t *tsdn, emap_t *emap, edata_t *edata) {
*/
EMAP_DECLARE_RTREE_CTX;
rtree_leaf_elm_t *elm = rtree_leaf_elm_lookup(tsdn, &emap->rtree,
rtree_ctx, (uintptr_t)edata_base_get(edata), /* dependent */ true,
rtree_ctx, (uintptr_t)edata_base_get(edata), /* dependent */ false,
/* init_missing */ false);
if (elm == NULL) {
return true;
}
rtree_contents_t contents = rtree_leaf_elm_read(tsdn, &emap->rtree, elm,
/* dependent */ true);
if (contents.edata == NULL ||
contents.metadata.state == extent_state_active ||
edata_state_in_transition(contents.metadata.state)) {
/* dependent */ false);
if (contents.edata == NULL
|| contents.metadata.state == extent_state_active
|| edata_state_in_transition(contents.metadata.state)) {
return true;
}
@ -209,8 +211,8 @@ extent_assert_can_coalesce(const edata_t *inner, const edata_t *outer) {
assert(edata_state_get(inner) == extent_state_active);
assert(edata_state_get(outer) == extent_state_merging);
assert(!edata_guarded_get(inner) && !edata_guarded_get(outer));
assert(edata_base_get(inner) == edata_past_get(outer) ||
edata_base_get(outer) == edata_past_get(inner));
assert(edata_base_get(inner) == edata_past_get(outer)
|| edata_base_get(outer) == edata_past_get(inner));
}
JEMALLOC_ALWAYS_INLINE void
@ -229,16 +231,46 @@ emap_edata_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr) {
return rtree_read(tsdn, &emap->rtree, rtree_ctx, (uintptr_t)ptr).edata;
}
JEMALLOC_ALWAYS_INLINE void
emap_alloc_ctx_init(
emap_alloc_ctx_t *alloc_ctx, szind_t szind, bool slab, size_t usize) {
alloc_ctx->szind = szind;
alloc_ctx->slab = slab;
alloc_ctx->usize = usize;
assert(
sz_large_size_classes_disabled() || usize == sz_index2size(szind));
}
JEMALLOC_ALWAYS_INLINE size_t
emap_alloc_ctx_usize_get(emap_alloc_ctx_t *alloc_ctx) {
assert(alloc_ctx->szind < SC_NSIZES);
if (alloc_ctx->slab) {
assert(alloc_ctx->usize == sz_index2size(alloc_ctx->szind));
return sz_index2size(alloc_ctx->szind);
}
assert(sz_large_size_classes_disabled()
|| alloc_ctx->usize == sz_index2size(alloc_ctx->szind));
assert(alloc_ctx->usize <= SC_LARGE_MAXCLASS);
return alloc_ctx->usize;
}
/* Fills in alloc_ctx with the info in the map. */
JEMALLOC_ALWAYS_INLINE void
emap_alloc_ctx_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
emap_alloc_ctx_t *alloc_ctx) {
emap_alloc_ctx_lookup(
tsdn_t *tsdn, emap_t *emap, const void *ptr, emap_alloc_ctx_t *alloc_ctx) {
EMAP_DECLARE_RTREE_CTX;
rtree_metadata_t metadata = rtree_metadata_read(tsdn, &emap->rtree,
rtree_ctx, (uintptr_t)ptr);
alloc_ctx->szind = metadata.szind;
alloc_ctx->slab = metadata.slab;
rtree_contents_t contents = rtree_read(
tsdn, &emap->rtree, rtree_ctx, (uintptr_t)ptr);
/*
* If the alloc is invalid, do not calculate usize since edata
* could be corrupted.
*/
emap_alloc_ctx_init(alloc_ctx, contents.metadata.szind,
contents.metadata.slab,
(contents.metadata.szind == SC_NSIZES || contents.edata == NULL)
? 0
: edata_usize_get(contents.edata));
}
/* The pointer must be mapped. */
@ -247,8 +279,8 @@ emap_full_alloc_ctx_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
emap_full_alloc_ctx_t *full_alloc_ctx) {
EMAP_DECLARE_RTREE_CTX;
rtree_contents_t contents = rtree_read(tsdn, &emap->rtree, rtree_ctx,
(uintptr_t)ptr);
rtree_contents_t contents = rtree_read(
tsdn, &emap->rtree, rtree_ctx, (uintptr_t)ptr);
full_alloc_ctx->edata = contents.edata;
full_alloc_ctx->szind = contents.metadata.szind;
full_alloc_ctx->slab = contents.metadata.slab;
@ -265,8 +297,8 @@ emap_full_alloc_ctx_try_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
EMAP_DECLARE_RTREE_CTX;
rtree_contents_t contents;
bool err = rtree_read_independent(tsdn, &emap->rtree, rtree_ctx,
(uintptr_t)ptr, &contents);
bool err = rtree_read_independent(
tsdn, &emap->rtree, rtree_ctx, (uintptr_t)ptr, &contents);
if (err) {
return true;
}
@ -281,19 +313,26 @@ emap_full_alloc_ctx_try_lookup(tsdn_t *tsdn, emap_t *emap, const void *ptr,
* fast path, e.g. when the metadata key is not cached.
*/
JEMALLOC_ALWAYS_INLINE bool
emap_alloc_ctx_try_lookup_fast(tsd_t *tsd, emap_t *emap, const void *ptr,
emap_alloc_ctx_t *alloc_ctx) {
emap_alloc_ctx_try_lookup_fast(
tsd_t *tsd, emap_t *emap, const void *ptr, emap_alloc_ctx_t *alloc_ctx) {
/* Use the unsafe getter since this may gets called during exit. */
rtree_ctx_t *rtree_ctx = tsd_rtree_ctxp_get_unsafe(tsd);
rtree_metadata_t metadata;
bool err = rtree_metadata_try_read_fast(tsd_tsdn(tsd), &emap->rtree,
rtree_ctx, (uintptr_t)ptr, &metadata);
bool err = rtree_metadata_try_read_fast(
tsd_tsdn(tsd), &emap->rtree, rtree_ctx, (uintptr_t)ptr, &metadata);
if (err) {
return true;
}
/*
* Small allocs using the fastpath can always use index to get the
* usize. Therefore, do not set alloc_ctx->usize here.
*/
alloc_ctx->szind = metadata.szind;
alloc_ctx->slab = metadata.slab;
if (config_debug) {
alloc_ctx->usize = SC_LARGE_MAXCLASS + 1;
}
return false;
}
@ -308,11 +347,12 @@ typedef const void *(*emap_ptr_getter)(void *ctx, size_t ind);
* This allows size-checking assertions, which we can only do while we're in the
* process of edata lookups.
*/
typedef void (*emap_metadata_visitor)(void *ctx, emap_full_alloc_ctx_t *alloc_ctx);
typedef void (*emap_metadata_visitor)(
void *ctx, emap_full_alloc_ctx_t *alloc_ctx);
typedef union emap_batch_lookup_result_u emap_batch_lookup_result_t;
union emap_batch_lookup_result_u {
edata_t *edata;
edata_t *edata;
rtree_leaf_elm_t *rtree_leaf;
};
@ -338,8 +378,8 @@ emap_edata_lookup_batch(tsd_t *tsd, emap_t *emap, size_t nptrs,
for (size_t i = 0; i < nptrs; i++) {
rtree_leaf_elm_t *elm = result[i].rtree_leaf;
rtree_contents_t contents = rtree_leaf_elm_read(tsd_tsdn(tsd),
&emap->rtree, elm, /* dependent */ true);
rtree_contents_t contents = rtree_leaf_elm_read(
tsd_tsdn(tsd), &emap->rtree, elm, /* dependent */ true);
result[i].edata = contents.edata;
emap_full_alloc_ctx_t alloc_ctx;
/*

148
include/jemalloc/internal/emitter.h
View file

@ -1,6 +1,10 @@
#ifndef JEMALLOC_INTERNAL_EMITTER_H
#define JEMALLOC_INTERNAL_EMITTER_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#include "jemalloc/internal/malloc_io.h"
#include "jemalloc/internal/ql.h"
typedef enum emitter_output_e emitter_output_t;
@ -40,18 +44,18 @@ typedef struct emitter_col_s emitter_col_t;
struct emitter_col_s {
/* Filled in by the user. */
emitter_justify_t justify;
int width;
emitter_type_t type;
int width;
emitter_type_t type;
union {
bool bool_val;
int int_val;
unsigned unsigned_val;
uint32_t uint32_val;
uint32_t uint32_t_val;
uint64_t uint64_val;
uint64_t uint64_t_val;
size_t size_val;
ssize_t ssize_val;
bool bool_val;
int int_val;
unsigned unsigned_val;
uint32_t uint32_val;
uint32_t uint32_t_val;
uint64_t uint64_val;
uint64_t uint64_t_val;
size_t size_val;
ssize_t ssize_val;
const char *str_val;
};
@ -69,8 +73,8 @@ struct emitter_s {
emitter_output_t output;
/* The output information. */
write_cb_t *write_cb;
void *cbopaque;
int nesting_depth;
void *cbopaque;
int nesting_depth;
/* True if we've already emitted a value at the given depth. */
bool item_at_depth;
/* True if we emitted a key and will emit corresponding value next. */
@ -79,8 +83,8 @@ struct emitter_s {
static inline bool
emitter_outputs_json(emitter_t *emitter) {
return emitter->output == emitter_output_json ||
emitter->output == emitter_output_json_compact;
return emitter->output == emitter_output_json
|| emitter->output == emitter_output_json_compact;
}
/* Internal convenience function. Write to the emitter the given string. */
@ -94,26 +98,57 @@ emitter_printf(emitter_t *emitter, const char *format, ...) {
va_end(ap);
}
static inline const char * JEMALLOC_FORMAT_ARG(3)
emitter_gen_fmt(char *out_fmt, size_t out_size, const char *fmt_specifier,
emitter_justify_t justify, int width) {
static inline const char *
JEMALLOC_FORMAT_ARG(3) emitter_gen_fmt(char *out_fmt, size_t out_size,
const char *fmt_specifier, emitter_justify_t justify, int width) {
size_t written;
fmt_specifier++;
if (justify == emitter_justify_none) {
written = malloc_snprintf(out_fmt, out_size,
"%%%s", fmt_specifier);
written = malloc_snprintf(
out_fmt, out_size, "%%%s", fmt_specifier);
} else if (justify == emitter_justify_left) {
written = malloc_snprintf(out_fmt, out_size,
"%%-%d%s", width, fmt_specifier);
written = malloc_snprintf(
out_fmt, out_size, "%%-%d%s", width, fmt_specifier);
} else {
written = malloc_snprintf(out_fmt, out_size,
"%%%d%s", width, fmt_specifier);
written = malloc_snprintf(
out_fmt, out_size, "%%%d%s", width, fmt_specifier);
}
/* Only happens in case of bad format string, which *we* choose. */
assert(written < out_size);
assert(written < out_size);
return out_fmt;
}
static inline void
emitter_emit_str(emitter_t *emitter, emitter_justify_t justify, int width,
char *fmt, size_t fmt_size, const char *str) {
#define BUF_SIZE 256
char buf[BUF_SIZE];
size_t str_written = malloc_snprintf(buf, BUF_SIZE, "\"%s\"", str);
emitter_printf(
emitter, emitter_gen_fmt(fmt, fmt_size, "%s", justify, width), buf);
if (str_written < BUF_SIZE) {
return;
}
/*
* There is no support for long string justification at the moment as
* we output them partially with multiple malloc_snprintf calls and
* justufication will work correctly only withing one call.
* Fortunately this is not a big concern as we don't use justufication
* with long strings right now.
*
* We emitted leading quotation mark and trailing '\0', hence need to
* exclude extra characters from str shift.
*/
str += BUF_SIZE - 2;
do {
str_written = malloc_snprintf(buf, BUF_SIZE, "%s\"", str);
str += str_written >= BUF_SIZE ? BUF_SIZE - 1 : str_written;
emitter_printf(emitter,
emitter_gen_fmt(fmt, fmt_size, "%s", justify, width), buf);
} while (str_written >= BUF_SIZE);
#undef BUF_SIZE
}
/*
* Internal. Emit the given value type in the relevant encoding (so that the
* bool true gets mapped to json "true", but the string "true" gets mapped to
@ -124,8 +159,6 @@ emitter_gen_fmt(char *out_fmt, size_t out_size, const char *fmt_specifier,
static inline void
emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width,
emitter_type_t value_type, const void *value) {
size_t str_written;
#define BUF_SIZE 256
#define FMT_SIZE 10
/*
* We dynamically generate a format string to emit, to let us use the
@ -134,18 +167,17 @@ emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width,
* cases.
*/
char fmt[FMT_SIZE];
char buf[BUF_SIZE];
#define EMIT_SIMPLE(type, format) \
emitter_printf(emitter, \
emitter_gen_fmt(fmt, FMT_SIZE, format, justify, width), \
#define EMIT_SIMPLE(type, format) \
emitter_printf(emitter, \
emitter_gen_fmt(fmt, FMT_SIZE, format, justify, width), \
*(const type *)value);
switch (value_type) {
case emitter_type_bool:
emitter_printf(emitter,
emitter_gen_fmt(fmt, FMT_SIZE, "%s", justify, width),
*(const bool *)value ? "true" : "false");
*(const bool *)value ? "true" : "false");
break;
case emitter_type_int:
EMIT_SIMPLE(int, "%d")
@ -163,15 +195,8 @@ emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width,
EMIT_SIMPLE(size_t, "%zu")
break;
case emitter_type_string:
str_written = malloc_snprintf(buf, BUF_SIZE, "\"%s\"",
emitter_emit_str(emitter, justify, width, fmt, FMT_SIZE,
*(const char *const *)value);
/*
* We control the strings we output; we shouldn't get anything
* anywhere near the fmt size.
*/
assert(str_written < BUF_SIZE);
emitter_printf(emitter,
emitter_gen_fmt(fmt, FMT_SIZE, "%s", justify, width), buf);
break;
case emitter_type_uint32:
EMIT_SIMPLE(uint32_t, "%" FMTu32)
@ -185,11 +210,9 @@ emitter_print_value(emitter_t *emitter, emitter_justify_t justify, int width,
default:
unreachable();
}
#undef BUF_SIZE
#undef FMT_SIZE
}
/* Internal functions. In json mode, tracks nesting state. */
static inline void
emitter_nest_inc(emitter_t *emitter) {
@ -205,7 +228,7 @@ emitter_nest_dec(emitter_t *emitter) {
static inline void
emitter_indent(emitter_t *emitter) {
int amount = emitter->nesting_depth;
int amount = emitter->nesting_depth;
const char *indent_str;
assert(emitter->output != emitter_output_json_compact);
if (emitter->output == emitter_output_json) {
@ -267,12 +290,12 @@ emitter_json_key(emitter_t *emitter, const char *json_key) {
}
static inline void
emitter_json_value(emitter_t *emitter, emitter_type_t value_type,
const void *value) {
emitter_json_value(
emitter_t *emitter, emitter_type_t value_type, const void *value) {
if (emitter_outputs_json(emitter)) {
emitter_json_key_prefix(emitter);
emitter_print_value(emitter, emitter_justify_none, -1,
value_type, value);
emitter_print_value(
emitter, emitter_justify_none, -1, value_type, value);
emitter->item_at_depth = true;
}
}
@ -343,7 +366,6 @@ emitter_json_object_end(emitter_t *emitter) {
}
}
/******************************************************************************/
/* Table public API. */
@ -365,14 +387,13 @@ emitter_table_dict_end(emitter_t *emitter) {
static inline void
emitter_table_kv_note(emitter_t *emitter, const char *table_key,
emitter_type_t value_type, const void *value,
const char *table_note_key, emitter_type_t table_note_value_type,
const void *table_note_value) {
emitter_type_t value_type, const void *value, const char *table_note_key,
emitter_type_t table_note_value_type, const void *table_note_value) {
if (emitter->output == emitter_output_table) {
emitter_indent(emitter);
emitter_printf(emitter, "%s: ", table_key);
emitter_print_value(emitter, emitter_justify_none, -1,
value_type, value);
emitter_print_value(
emitter, emitter_justify_none, -1, value_type, value);
if (table_note_key != NULL) {
emitter_printf(emitter, " (%s: ", table_note_key);
emitter_print_value(emitter, emitter_justify_none, -1,
@ -391,7 +412,6 @@ emitter_table_kv(emitter_t *emitter, const char *table_key,
emitter_type_bool, NULL);
}
/* Write to the emitter the given string, but only in table mode. */
JEMALLOC_FORMAT_PRINTF(2, 3)
static inline void
@ -399,7 +419,8 @@ emitter_table_printf(emitter_t *emitter, const char *format, ...) {
if (emitter->output == emitter_output_table) {
va_list ap;
va_start(ap, format);
malloc_vcprintf(emitter->write_cb, emitter->cbopaque, format, ap);
malloc_vcprintf(
emitter->write_cb, emitter->cbopaque, format, ap);
va_end(ap);
}
}
@ -410,7 +431,7 @@ emitter_table_row(emitter_t *emitter, emitter_row_t *row) {
return;
}
emitter_col_t *col;
ql_foreach(col, &row->cols, link) {
ql_foreach (col, &row->cols, link) {
emitter_print_value(emitter, col->justify, col->width,
col->type, (const void *)&col->bool_val);
}
@ -428,7 +449,6 @@ emitter_col_init(emitter_col_t *col, emitter_row_t *row) {
ql_tail_insert(&row->cols, col, link);
}
/******************************************************************************/
/*
* Generalized public API. Emits using either JSON or table, according to
@ -440,9 +460,8 @@ emitter_col_init(emitter_col_t *col, emitter_row_t *row) {
*/
static inline void
emitter_kv_note(emitter_t *emitter, const char *json_key, const char *table_key,
emitter_type_t value_type, const void *value,
const char *table_note_key, emitter_type_t table_note_value_type,
const void *table_note_value) {
emitter_type_t value_type, const void *value, const char *table_note_key,
emitter_type_t table_note_value_type, const void *table_note_value) {
if (emitter_outputs_json(emitter)) {
emitter_json_key(emitter, json_key);
emitter_json_value(emitter, value_type, value);
@ -461,8 +480,8 @@ emitter_kv(emitter_t *emitter, const char *json_key, const char *table_key,
}
static inline void
emitter_dict_begin(emitter_t *emitter, const char *json_key,
const char *table_header) {
emitter_dict_begin(
emitter_t *emitter, const char *json_key, const char *table_header) {
if (emitter_outputs_json(emitter)) {
emitter_json_key(emitter, json_key);
emitter_json_object_begin(emitter);
@ -502,8 +521,9 @@ emitter_end(emitter_t *emitter) {
if (emitter_outputs_json(emitter)) {
assert(emitter->nesting_depth == 1);
emitter_nest_dec(emitter);
emitter_printf(emitter, "%s", emitter->output ==
emitter_output_json_compact ? "}" : "\n}\n");
emitter_printf(emitter, "%s",
emitter->output == emitter_output_json_compact ? "}"
: "\n}\n");
}
}

3
include/jemalloc/internal/eset.h
View file

@ -1,9 +1,10 @@
#ifndef JEMALLOC_INTERNAL_ESET_H
#define JEMALLOC_INTERNAL_ESET_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/fb.h"
#include "jemalloc/internal/edata.h"
#include "jemalloc/internal/fb.h"
#include "jemalloc/internal/mutex.h"
/*

6
include/jemalloc/internal/exp_grow.h
View file

@ -1,6 +1,8 @@
#ifndef JEMALLOC_INTERNAL_EXP_GROW_H
#define JEMALLOC_INTERNAL_EXP_GROW_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/sz.h"
typedef struct exp_grow_s exp_grow_t;
struct exp_grow_s {
/*
@ -25,8 +27,7 @@ exp_grow_size_prepare(exp_grow_t *exp_grow, size_t alloc_size_min,
*r_alloc_size = sz_pind2sz(exp_grow->next + *r_skip);
while (*r_alloc_size < alloc_size_min) {
(*r_skip)++;
if (exp_grow->next + *r_skip >=
sz_psz2ind(SC_LARGE_MAXCLASS)) {
if (exp_grow->next + *r_skip >= sz_psz2ind(SC_LARGE_MAXCLASS)) {
/* Outside legal range. */
return true;
}
@ -42,7 +43,6 @@ exp_grow_size_commit(exp_grow_t *exp_grow, pszind_t skip) {
} else {
exp_grow->next = exp_grow->limit;
}
}
void exp_grow_init(exp_grow_t *exp_grow);

71
include/jemalloc/internal/extent.h
View file

@ -1,8 +1,10 @@
#ifndef JEMALLOC_INTERNAL_EXTENT_H
#define JEMALLOC_INTERNAL_EXTENT_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/ecache.h"
#include "jemalloc/internal/ehooks.h"
#include "jemalloc/internal/pac.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/rtree.h"
@ -19,50 +21,58 @@
#define LG_EXTENT_MAX_ACTIVE_FIT_DEFAULT 6
extern size_t opt_lg_extent_max_active_fit;
#define PROCESS_MADVISE_MAX_BATCH_DEFAULT 0
extern size_t opt_process_madvise_max_batch;
#ifdef JEMALLOC_HAVE_PROCESS_MADVISE
/* The iovec is on stack. Limit the max batch to avoid stack overflow. */
# define PROCESS_MADVISE_MAX_BATCH_LIMIT \
(VARIABLE_ARRAY_SIZE_MAX / sizeof(struct iovec))
#else
# define PROCESS_MADVISE_MAX_BATCH_LIMIT 0
#endif
edata_t *ecache_alloc(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, edata_t *expand_edata, size_t size, size_t alignment,
bool zero, bool guarded);
edata_t *ecache_alloc_grow(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, edata_t *expand_edata, size_t size, size_t alignment,
bool zero, bool guarded);
void ecache_dalloc(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, edata_t *edata);
void ecache_dalloc(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, ecache_t *ecache,
edata_t *edata);
edata_t *ecache_evict(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
ecache_t *ecache, size_t npages_min);
void extent_gdump_add(tsdn_t *tsdn, const edata_t *edata);
void extent_record(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, ecache_t *ecache,
edata_t *edata);
void extent_dalloc_gap(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata);
void extent_dalloc_gap(
tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, edata_t *edata);
edata_t *extent_alloc_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
void *new_addr, size_t size, size_t alignment, bool zero, bool *commit,
bool growing_retained);
void extent_dalloc_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata);
void extent_destroy_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata);
bool extent_commit_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
bool extent_decommit_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
void extent_dalloc_wrapper(
tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, edata_t *edata);
void extent_dalloc_wrapper_purged(
tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, edata_t *edata);
void extent_destroy_wrapper(
tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, edata_t *edata);
bool extent_purge_lazy_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
bool extent_purge_forced_wrapper(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
size_t offset, size_t length);
edata_t *extent_split_wrapper(tsdn_t *tsdn, pac_t *pac,
ehooks_t *ehooks, edata_t *edata, size_t size_a, size_t size_b,
bool holding_core_locks);
bool extent_merge_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *a, edata_t *b);
bool extent_commit_zero(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
bool commit, bool zero, bool growing_retained);
edata_t *extent_split_wrapper(tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks,
edata_t *edata, size_t size_a, size_t size_b, bool holding_core_locks);
bool extent_merge_wrapper(
tsdn_t *tsdn, pac_t *pac, ehooks_t *ehooks, edata_t *a, edata_t *b);
bool extent_commit_zero(tsdn_t *tsdn, ehooks_t *ehooks, edata_t *edata,
bool commit, bool zero, bool growing_retained);
size_t extent_sn_next(pac_t *pac);
bool extent_boot(void);
bool extent_boot(void);
JEMALLOC_ALWAYS_INLINE bool
extent_neighbor_head_state_mergeable(bool edata_is_head,
bool neighbor_is_head, bool forward) {
extent_neighbor_head_state_mergeable(
bool edata_is_head, bool neighbor_is_head, bool forward) {
/*
* Head states checking: disallow merging if the higher addr extent is a
* head extent. This helps preserve first-fit, and more importantly
@ -90,8 +100,8 @@ extent_can_acquire_neighbor(edata_t *edata, rtree_contents_t contents,
}
/* It's not safe to access *neighbor yet; must verify states first. */
bool neighbor_is_head = contents.metadata.is_head;
if (!extent_neighbor_head_state_mergeable(edata_is_head_get(edata),
neighbor_is_head, forward)) {
if (!extent_neighbor_head_state_mergeable(
edata_is_head_get(edata), neighbor_is_head, forward)) {
return false;
}
extent_state_t neighbor_state = contents.metadata.state;
@ -100,8 +110,9 @@ extent_can_acquire_neighbor(edata_t *edata, rtree_contents_t contents,
return false;
}
/* From this point, it's safe to access *neighbor. */
if (!expanding && (edata_committed_get(edata) !=
edata_committed_get(neighbor))) {
if (!expanding
&& (edata_committed_get(edata)
!= edata_committed_get(neighbor))) {
/*
* Some platforms (e.g. Windows) require an explicit
* commit step (and writing to uncommitted memory is not
@ -121,11 +132,11 @@ extent_can_acquire_neighbor(edata_t *edata, rtree_contents_t contents,
return false;
}
if (opt_retain) {
assert(edata_arena_ind_get(edata) ==
edata_arena_ind_get(neighbor));
assert(edata_arena_ind_get(edata)
== edata_arena_ind_get(neighbor));
} else {
if (edata_arena_ind_get(edata) !=
edata_arena_ind_get(neighbor)) {
if (edata_arena_ind_get(edata)
!= edata_arena_ind_get(neighbor)) {
return false;
}
}

24
include/jemalloc/internal/extent_dss.h
View file

@ -1,26 +1,30 @@
#ifndef JEMALLOC_INTERNAL_EXTENT_DSS_H
#define JEMALLOC_INTERNAL_EXTENT_DSS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_types.h"
#include "jemalloc/internal/tsd_types.h"
typedef enum {
dss_prec_disabled = 0,
dss_prec_primary = 1,
dss_prec_disabled = 0,
dss_prec_primary = 1,
dss_prec_secondary = 2,
dss_prec_limit = 3
dss_prec_limit = 3
} dss_prec_t;
#define DSS_PREC_DEFAULT dss_prec_secondary
#define DSS_DEFAULT "secondary"
extern const char *dss_prec_names[];
extern const char *const dss_prec_names[];
extern const char *opt_dss;
dss_prec_t extent_dss_prec_get(void);
bool extent_dss_prec_set(dss_prec_t dss_prec);
void *extent_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr,
size_t size, size_t alignment, bool *zero, bool *commit);
bool extent_in_dss(void *addr);
bool extent_dss_mergeable(void *addr_a, void *addr_b);
void extent_dss_boot(void);
bool extent_dss_prec_set(dss_prec_t dss_prec);
void *extent_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr,
size_t size, size_t alignment, bool *zero, bool *commit);
bool extent_in_dss(void *addr);
bool extent_dss_mergeable(void *addr_a, void *addr_b);
void extent_dss_boot(void);
#endif /* JEMALLOC_INTERNAL_EXTENT_DSS_H */

6
include/jemalloc/internal/extent_mmap.h
View file

@ -1,10 +1,12 @@
#ifndef JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H
#define JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H
#include "jemalloc/internal/jemalloc_preamble.h"
extern bool opt_retain;
void *extent_alloc_mmap(void *new_addr, size_t size, size_t alignment,
bool *zero, bool *commit);
void *extent_alloc_mmap(
void *new_addr, size_t size, size_t alignment, bool *zero, bool *commit);
bool extent_dalloc_mmap(void *addr, size_t size);
#endif /* JEMALLOC_INTERNAL_EXTENT_MMAP_EXTERNS_H */

33
include/jemalloc/internal/fb.h
View file

@ -1,6 +1,10 @@
#ifndef JEMALLOC_INTERNAL_FB_H
#define JEMALLOC_INTERNAL_FB_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
#include "jemalloc/internal/bit_util.h"
/*
* The flat bitmap module. This has a larger API relative to the bitmap module
* (supporting things like backwards searches, and searching for both set and
@ -11,8 +15,8 @@
typedef unsigned long fb_group_t;
#define FB_GROUP_BITS (ZU(1) << (LG_SIZEOF_LONG + 3))
#define FB_NGROUPS(nbits) ((nbits) / FB_GROUP_BITS \
+ ((nbits) % FB_GROUP_BITS == 0 ? 0 : 1))
#define FB_NGROUPS(nbits) \
((nbits) / FB_GROUP_BITS + ((nbits) % FB_GROUP_BITS == 0 ? 0 : 1))
static inline void
fb_init(fb_group_t *fb, size_t nbits) {
@ -71,7 +75,6 @@ fb_unset(fb_group_t *fb, size_t nbits, size_t bit) {
fb[group_ind] &= ~((fb_group_t)1 << bit_ind);
}
/*
* Some implementation details. This visitation function lets us apply a group
* visitor to each group in the bitmap (potentially modifying it). The mask
@ -90,7 +93,8 @@ fb_visit_impl(fb_group_t *fb, size_t nbits, fb_group_visitor_t visit, void *ctx,
* to from bit 0.
*/
size_t first_group_cnt = (start_bit_ind + cnt > FB_GROUP_BITS
? FB_GROUP_BITS - start_bit_ind : cnt);
? FB_GROUP_BITS - start_bit_ind
: cnt);
/*
* We can basically split affected words into:
* - The first group, where we touch only the high bits
@ -100,8 +104,8 @@ fb_visit_impl(fb_group_t *fb, size_t nbits, fb_group_visitor_t visit, void *ctx,
* this can lead to bad codegen for those middle words.
*/
/* First group */
fb_group_t mask = ((~(fb_group_t)0)
>> (FB_GROUP_BITS - first_group_cnt))
fb_group_t mask =
((~(fb_group_t)0) >> (FB_GROUP_BITS - first_group_cnt))
<< start_bit_ind;
visit(ctx, &fb[group_ind], mask);
@ -172,12 +176,12 @@ fb_ucount(fb_group_t *fb, size_t nbits, size_t start, size_t cnt) {
* Returns the number of bits in the bitmap if no such bit exists.
*/
JEMALLOC_ALWAYS_INLINE ssize_t
fb_find_impl(fb_group_t *fb, size_t nbits, size_t start, bool val,
bool forward) {
fb_find_impl(
fb_group_t *fb, size_t nbits, size_t start, bool val, bool forward) {
assert(start < nbits);
size_t ngroups = FB_NGROUPS(nbits);
size_t ngroups = FB_NGROUPS(nbits);
ssize_t group_ind = start / FB_GROUP_BITS;
size_t bit_ind = start % FB_GROUP_BITS;
size_t bit_ind = start % FB_GROUP_BITS;
fb_group_t maybe_invert = (val ? 0 : (fb_group_t)-1);
@ -261,8 +265,8 @@ fb_iter_range_impl(fb_group_t *fb, size_t nbits, size_t start, size_t *r_begin,
return false;
}
/* Half open range; the set bits are [begin, end). */
ssize_t next_range_end = fb_find_impl(fb, nbits, next_range_begin, !val,
forward);
ssize_t next_range_end = fb_find_impl(
fb, nbits, next_range_begin, !val, forward);
if (forward) {
*r_begin = next_range_begin;
*r_len = next_range_end - next_range_begin;
@ -320,8 +324,9 @@ fb_range_longest_impl(fb_group_t *fb, size_t nbits, bool val) {
size_t begin = 0;
size_t longest_len = 0;
size_t len = 0;
while (begin < nbits && fb_iter_range_impl(fb, nbits, begin, &begin,
&len, val, /* forward */ true)) {
while (begin < nbits
&& fb_iter_range_impl(
fb, nbits, begin, &begin, &len, val, /* forward */ true)) {
if (len > longest_len) {
longest_len = len;
}

5
include/jemalloc/internal/fxp.h
View file

@ -1,6 +1,9 @@
#ifndef JEMALLOC_INTERNAL_FXP_H
#define JEMALLOC_INTERNAL_FXP_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
/*
* A simple fixed-point math implementation, supporting only unsigned values
* (with overflow being an error).
@ -86,7 +89,7 @@ fxp_round_down(fxp_t a) {
static inline uint32_t
fxp_round_nearest(fxp_t a) {
uint32_t fractional_part = (a & ((1U << 16) - 1));
uint32_t fractional_part = (a & ((1U << 16) - 1));
uint32_t increment = (uint32_t)(fractional_part >= (1U << 15));
return (a >> 16) + increment;
}

317
include/jemalloc/internal/hash.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_HASH_H
#define JEMALLOC_INTERNAL_HASH_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
/*
@ -24,7 +25,7 @@ hash_rotl_64(uint64_t x, int8_t r) {
static inline uint32_t
hash_get_block_32(const uint32_t *p, int i) {
/* Handle unaligned read. */
if (unlikely((uintptr_t)p & (sizeof(uint32_t)-1)) != 0) {
if (unlikely((uintptr_t)p & (sizeof(uint32_t) - 1)) != 0) {
uint32_t ret;
memcpy(&ret, (uint8_t *)(p + i), sizeof(uint32_t));
@ -37,7 +38,7 @@ hash_get_block_32(const uint32_t *p, int i) {
static inline uint64_t
hash_get_block_64(const uint64_t *p, int i) {
/* Handle unaligned read. */
if (unlikely((uintptr_t)p & (sizeof(uint64_t)-1)) != 0) {
if (unlikely((uintptr_t)p & (sizeof(uint64_t) - 1)) != 0) {
uint64_t ret;
memcpy(&ret, (uint8_t *)(p + i), sizeof(uint64_t));
@ -71,8 +72,8 @@ hash_fmix_64(uint64_t k) {
static inline uint32_t
hash_x86_32(const void *key, int len, uint32_t seed) {
const uint8_t *data = (const uint8_t *) key;
const int nblocks = len / 4;
const uint8_t *data = (const uint8_t *)key;
const int nblocks = len / 4;
uint32_t h1 = seed;
@ -81,8 +82,8 @@ hash_x86_32(const void *key, int len, uint32_t seed) {
/* body */
{
const uint32_t *blocks = (const uint32_t *) (data + nblocks*4);
int i;
const uint32_t *blocks = (const uint32_t *)(data + nblocks * 4);
int i;
for (i = -nblocks; i; i++) {
uint32_t k1 = hash_get_block_32(blocks, i);
@ -93,21 +94,29 @@ hash_x86_32(const void *key, int len, uint32_t seed) {
h1 ^= k1;
h1 = hash_rotl_32(h1, 13);
h1 = h1*5 + 0xe6546b64;
h1 = h1 * 5 + 0xe6546b64;
}
}
/* tail */
{
const uint8_t *tail = (const uint8_t *) (data + nblocks*4);
const uint8_t *tail = (const uint8_t *)(data + nblocks * 4);
uint32_t k1 = 0;
switch (len & 3) {
case 3: k1 ^= tail[2] << 16; JEMALLOC_FALLTHROUGH;
case 2: k1 ^= tail[1] << 8; JEMALLOC_FALLTHROUGH;
case 1: k1 ^= tail[0]; k1 *= c1; k1 = hash_rotl_32(k1, 15);
k1 *= c2; h1 ^= k1;
case 3:
k1 ^= tail[2] << 16;
JEMALLOC_FALLTHROUGH;
case 2:
k1 ^= tail[1] << 8;
JEMALLOC_FALLTHROUGH;
case 1:
k1 ^= tail[0];
k1 *= c1;
k1 = hash_rotl_32(k1, 15);
k1 *= c2;
h1 ^= k1;
}
}
@ -120,10 +129,9 @@ hash_x86_32(const void *key, int len, uint32_t seed) {
}
static inline void
hash_x86_128(const void *key, const int len, uint32_t seed,
uint64_t r_out[2]) {
const uint8_t * data = (const uint8_t *) key;
const int nblocks = len / 16;
hash_x86_128(const void *key, const int len, uint32_t seed, uint64_t r_out[2]) {
const uint8_t *data = (const uint8_t *)key;
const int nblocks = len / 16;
uint32_t h1 = seed;
uint32_t h2 = seed;
@ -137,95 +145,161 @@ hash_x86_128(const void *key, const int len, uint32_t seed,
/* body */
{
const uint32_t *blocks = (const uint32_t *) (data + nblocks*16);
int i;
const uint32_t *blocks = (const uint32_t *)(data
+ nblocks * 16);
int i;
for (i = -nblocks; i; i++) {
uint32_t k1 = hash_get_block_32(blocks, i*4 + 0);
uint32_t k2 = hash_get_block_32(blocks, i*4 + 1);
uint32_t k3 = hash_get_block_32(blocks, i*4 + 2);
uint32_t k4 = hash_get_block_32(blocks, i*4 + 3);
uint32_t k1 = hash_get_block_32(blocks, i * 4 + 0);
uint32_t k2 = hash_get_block_32(blocks, i * 4 + 1);
uint32_t k3 = hash_get_block_32(blocks, i * 4 + 2);
uint32_t k4 = hash_get_block_32(blocks, i * 4 + 3);
k1 *= c1; k1 = hash_rotl_32(k1, 15); k1 *= c2; h1 ^= k1;
k1 *= c1;
k1 = hash_rotl_32(k1, 15);
k1 *= c2;
h1 ^= k1;
h1 = hash_rotl_32(h1, 19); h1 += h2;
h1 = h1*5 + 0x561ccd1b;
h1 = hash_rotl_32(h1, 19);
h1 += h2;
h1 = h1 * 5 + 0x561ccd1b;
k2 *= c2; k2 = hash_rotl_32(k2, 16); k2 *= c3; h2 ^= k2;
k2 *= c2;
k2 = hash_rotl_32(k2, 16);
k2 *= c3;
h2 ^= k2;
h2 = hash_rotl_32(h2, 17); h2 += h3;
h2 = h2*5 + 0x0bcaa747;
h2 = hash_rotl_32(h2, 17);
h2 += h3;
h2 = h2 * 5 + 0x0bcaa747;
k3 *= c3; k3 = hash_rotl_32(k3, 17); k3 *= c4; h3 ^= k3;
k3 *= c3;
k3 = hash_rotl_32(k3, 17);
k3 *= c4;
h3 ^= k3;
h3 = hash_rotl_32(h3, 15); h3 += h4;
h3 = h3*5 + 0x96cd1c35;
h3 = hash_rotl_32(h3, 15);
h3 += h4;
h3 = h3 * 5 + 0x96cd1c35;
k4 *= c4; k4 = hash_rotl_32(k4, 18); k4 *= c1; h4 ^= k4;
k4 *= c4;
k4 = hash_rotl_32(k4, 18);
k4 *= c1;
h4 ^= k4;
h4 = hash_rotl_32(h4, 13); h4 += h1;
h4 = h4*5 + 0x32ac3b17;
h4 = hash_rotl_32(h4, 13);
h4 += h1;
h4 = h4 * 5 + 0x32ac3b17;
}
}
/* tail */
{
const uint8_t *tail = (const uint8_t *) (data + nblocks*16);
uint32_t k1 = 0;
uint32_t k2 = 0;
uint32_t k3 = 0;
uint32_t k4 = 0;
const uint8_t *tail = (const uint8_t *)(data + nblocks * 16);
uint32_t k1 = 0;
uint32_t k2 = 0;
uint32_t k3 = 0;
uint32_t k4 = 0;
switch (len & 15) {
case 15: k4 ^= tail[14] << 16; JEMALLOC_FALLTHROUGH;
case 14: k4 ^= tail[13] << 8; JEMALLOC_FALLTHROUGH;
case 13: k4 ^= tail[12] << 0;
k4 *= c4; k4 = hash_rotl_32(k4, 18); k4 *= c1; h4 ^= k4;
case 15:
k4 ^= tail[14] << 16;
JEMALLOC_FALLTHROUGH;
case 12: k3 ^= (uint32_t) tail[11] << 24; JEMALLOC_FALLTHROUGH;
case 11: k3 ^= tail[10] << 16; JEMALLOC_FALLTHROUGH;
case 10: k3 ^= tail[ 9] << 8; JEMALLOC_FALLTHROUGH;
case 9: k3 ^= tail[ 8] << 0;
k3 *= c3; k3 = hash_rotl_32(k3, 17); k3 *= c4; h3 ^= k3;
case 14:
k4 ^= tail[13] << 8;
JEMALLOC_FALLTHROUGH;
case 8: k2 ^= (uint32_t) tail[ 7] << 24; JEMALLOC_FALLTHROUGH;
case 7: k2 ^= tail[ 6] << 16; JEMALLOC_FALLTHROUGH;
case 6: k2 ^= tail[ 5] << 8; JEMALLOC_FALLTHROUGH;
case 5: k2 ^= tail[ 4] << 0;
k2 *= c2; k2 = hash_rotl_32(k2, 16); k2 *= c3; h2 ^= k2;
case 13:
k4 ^= tail[12] << 0;
k4 *= c4;
k4 = hash_rotl_32(k4, 18);
k4 *= c1;
h4 ^= k4;
JEMALLOC_FALLTHROUGH;
case 4: k1 ^= (uint32_t) tail[ 3] << 24; JEMALLOC_FALLTHROUGH;
case 3: k1 ^= tail[ 2] << 16; JEMALLOC_FALLTHROUGH;
case 2: k1 ^= tail[ 1] << 8; JEMALLOC_FALLTHROUGH;
case 1: k1 ^= tail[ 0] << 0;
k1 *= c1; k1 = hash_rotl_32(k1, 15); k1 *= c2; h1 ^= k1;
case 12:
k3 ^= (uint32_t)tail[11] << 24;
JEMALLOC_FALLTHROUGH;
case 11:
k3 ^= tail[10] << 16;
JEMALLOC_FALLTHROUGH;
case 10:
k3 ^= tail[9] << 8;
JEMALLOC_FALLTHROUGH;
case 9:
k3 ^= tail[8] << 0;
k3 *= c3;
k3 = hash_rotl_32(k3, 17);
k3 *= c4;
h3 ^= k3;
JEMALLOC_FALLTHROUGH;
case 8:
k2 ^= (uint32_t)tail[7] << 24;
JEMALLOC_FALLTHROUGH;
case 7:
k2 ^= tail[6] << 16;
JEMALLOC_FALLTHROUGH;
case 6:
k2 ^= tail[5] << 8;
JEMALLOC_FALLTHROUGH;
case 5:
k2 ^= tail[4] << 0;
k2 *= c2;
k2 = hash_rotl_32(k2, 16);
k2 *= c3;
h2 ^= k2;
JEMALLOC_FALLTHROUGH;
case 4:
k1 ^= (uint32_t)tail[3] << 24;
JEMALLOC_FALLTHROUGH;
case 3:
k1 ^= tail[2] << 16;
JEMALLOC_FALLTHROUGH;
case 2:
k1 ^= tail[1] << 8;
JEMALLOC_FALLTHROUGH;
case 1:
k1 ^= tail[0] << 0;
k1 *= c1;
k1 = hash_rotl_32(k1, 15);
k1 *= c2;
h1 ^= k1;
break;
}
}
/* finalization */
h1 ^= len; h2 ^= len; h3 ^= len; h4 ^= len;
h1 ^= len;
h2 ^= len;
h3 ^= len;
h4 ^= len;
h1 += h2; h1 += h3; h1 += h4;
h2 += h1; h3 += h1; h4 += h1;
h1 += h2;
h1 += h3;
h1 += h4;
h2 += h1;
h3 += h1;
h4 += h1;
h1 = hash_fmix_32(h1);
h2 = hash_fmix_32(h2);
h3 = hash_fmix_32(h3);
h4 = hash_fmix_32(h4);
h1 += h2; h1 += h3; h1 += h4;
h2 += h1; h3 += h1; h4 += h1;
h1 += h2;
h1 += h3;
h1 += h4;
h2 += h1;
h3 += h1;
h4 += h1;
r_out[0] = (((uint64_t) h2) << 32) | h1;
r_out[1] = (((uint64_t) h4) << 32) | h3;
r_out[0] = (((uint64_t)h2) << 32) | h1;
r_out[1] = (((uint64_t)h4) << 32) | h3;
}
static inline void
hash_x64_128(const void *key, const int len, const uint32_t seed,
uint64_t r_out[2]) {
const uint8_t *data = (const uint8_t *) key;
const int nblocks = len / 16;
hash_x64_128(
const void *key, const int len, const uint32_t seed, uint64_t r_out[2]) {
const uint8_t *data = (const uint8_t *)key;
const int nblocks = len / 16;
uint64_t h1 = seed;
uint64_t h2 = seed;
@ -235,56 +309,99 @@ hash_x64_128(const void *key, const int len, const uint32_t seed,
/* body */
{
const uint64_t *blocks = (const uint64_t *) (data);
int i;
const uint64_t *blocks = (const uint64_t *)(data);
int i;
for (i = 0; i < nblocks; i++) {
uint64_t k1 = hash_get_block_64(blocks, i*2 + 0);
uint64_t k2 = hash_get_block_64(blocks, i*2 + 1);
uint64_t k1 = hash_get_block_64(blocks, i * 2 + 0);
uint64_t k2 = hash_get_block_64(blocks, i * 2 + 1);
k1 *= c1; k1 = hash_rotl_64(k1, 31); k1 *= c2; h1 ^= k1;
k1 *= c1;
k1 = hash_rotl_64(k1, 31);
k1 *= c2;
h1 ^= k1;
h1 = hash_rotl_64(h1, 27); h1 += h2;
h1 = h1*5 + 0x52dce729;
h1 = hash_rotl_64(h1, 27);
h1 += h2;
h1 = h1 * 5 + 0x52dce729;
k2 *= c2; k2 = hash_rotl_64(k2, 33); k2 *= c1; h2 ^= k2;
k2 *= c2;
k2 = hash_rotl_64(k2, 33);
k2 *= c1;
h2 ^= k2;
h2 = hash_rotl_64(h2, 31); h2 += h1;
h2 = h2*5 + 0x38495ab5;
h2 = hash_rotl_64(h2, 31);
h2 += h1;
h2 = h2 * 5 + 0x38495ab5;
}
}
/* tail */
{
const uint8_t *tail = (const uint8_t*)(data + nblocks*16);
uint64_t k1 = 0;
uint64_t k2 = 0;
const uint8_t *tail = (const uint8_t *)(data + nblocks * 16);
uint64_t k1 = 0;
uint64_t k2 = 0;
switch (len & 15) {
case 15: k2 ^= ((uint64_t)(tail[14])) << 48; JEMALLOC_FALLTHROUGH;
case 14: k2 ^= ((uint64_t)(tail[13])) << 40; JEMALLOC_FALLTHROUGH;
case 13: k2 ^= ((uint64_t)(tail[12])) << 32; JEMALLOC_FALLTHROUGH;
case 12: k2 ^= ((uint64_t)(tail[11])) << 24; JEMALLOC_FALLTHROUGH;
case 11: k2 ^= ((uint64_t)(tail[10])) << 16; JEMALLOC_FALLTHROUGH;
case 10: k2 ^= ((uint64_t)(tail[ 9])) << 8; JEMALLOC_FALLTHROUGH;
case 9: k2 ^= ((uint64_t)(tail[ 8])) << 0;
k2 *= c2; k2 = hash_rotl_64(k2, 33); k2 *= c1; h2 ^= k2;
case 15:
k2 ^= ((uint64_t)(tail[14])) << 48;
JEMALLOC_FALLTHROUGH;
case 8: k1 ^= ((uint64_t)(tail[ 7])) << 56; JEMALLOC_FALLTHROUGH;
case 7: k1 ^= ((uint64_t)(tail[ 6])) << 48; JEMALLOC_FALLTHROUGH;
case 6: k1 ^= ((uint64_t)(tail[ 5])) << 40; JEMALLOC_FALLTHROUGH;
case 5: k1 ^= ((uint64_t)(tail[ 4])) << 32; JEMALLOC_FALLTHROUGH;
case 4: k1 ^= ((uint64_t)(tail[ 3])) << 24; JEMALLOC_FALLTHROUGH;
case 3: k1 ^= ((uint64_t)(tail[ 2])) << 16; JEMALLOC_FALLTHROUGH;
case 2: k1 ^= ((uint64_t)(tail[ 1])) << 8; JEMALLOC_FALLTHROUGH;
case 1: k1 ^= ((uint64_t)(tail[ 0])) << 0;
k1 *= c1; k1 = hash_rotl_64(k1, 31); k1 *= c2; h1 ^= k1;
case 14:
k2 ^= ((uint64_t)(tail[13])) << 40;
JEMALLOC_FALLTHROUGH;
case 13:
k2 ^= ((uint64_t)(tail[12])) << 32;
JEMALLOC_FALLTHROUGH;
case 12:
k2 ^= ((uint64_t)(tail[11])) << 24;
JEMALLOC_FALLTHROUGH;
case 11:
k2 ^= ((uint64_t)(tail[10])) << 16;
JEMALLOC_FALLTHROUGH;
case 10:
k2 ^= ((uint64_t)(tail[9])) << 8;
JEMALLOC_FALLTHROUGH;
case 9:
k2 ^= ((uint64_t)(tail[8])) << 0;
k2 *= c2;
k2 = hash_rotl_64(k2, 33);
k2 *= c1;
h2 ^= k2;
JEMALLOC_FALLTHROUGH;
case 8:
k1 ^= ((uint64_t)(tail[7])) << 56;
JEMALLOC_FALLTHROUGH;
case 7:
k1 ^= ((uint64_t)(tail[6])) << 48;
JEMALLOC_FALLTHROUGH;
case 6:
k1 ^= ((uint64_t)(tail[5])) << 40;
JEMALLOC_FALLTHROUGH;
case 5:
k1 ^= ((uint64_t)(tail[4])) << 32;
JEMALLOC_FALLTHROUGH;
case 4:
k1 ^= ((uint64_t)(tail[3])) << 24;
JEMALLOC_FALLTHROUGH;
case 3:
k1 ^= ((uint64_t)(tail[2])) << 16;
JEMALLOC_FALLTHROUGH;
case 2:
k1 ^= ((uint64_t)(tail[1])) << 8;
JEMALLOC_FALLTHROUGH;
case 1:
k1 ^= ((uint64_t)(tail[0])) << 0;
k1 *= c1;
k1 = hash_rotl_64(k1, 31);
k1 *= c2;
h1 ^= k1;
break;
}
}
/* finalization */
h1 ^= len; h2 ^= len;
h1 ^= len;
h2 ^= len;
h1 += h2;
h2 += h1;

26
include/jemalloc/internal/hook.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_HOOK_H
#define JEMALLOC_INTERNAL_HOOK_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/tsd.h"
/*
@ -55,6 +56,7 @@ enum hook_alloc_e {
hook_alloc_calloc,
hook_alloc_memalign,
hook_alloc_valloc,
hook_alloc_pvalloc,
hook_alloc_mallocx,
/* The reallocating functions have both alloc and dalloc variants */
@ -81,7 +83,6 @@ enum hook_dalloc_e {
};
typedef enum hook_dalloc_e hook_dalloc_t;
enum hook_expand_e {
hook_expand_realloc,
hook_expand_rallocx,
@ -89,23 +90,22 @@ enum hook_expand_e {
};
typedef enum hook_expand_e hook_expand_t;
typedef void (*hook_alloc)(
void *extra, hook_alloc_t type, void *result, uintptr_t result_raw,
uintptr_t args_raw[3]);
typedef void (*hook_alloc)(void *extra, hook_alloc_t type, void *result,
uintptr_t result_raw, uintptr_t args_raw[3]);
typedef void (*hook_dalloc)(
void *extra, hook_dalloc_t type, void *address, uintptr_t args_raw[3]);
typedef void (*hook_expand)(
void *extra, hook_expand_t type, void *address, size_t old_usize,
size_t new_usize, uintptr_t result_raw, uintptr_t args_raw[4]);
typedef void (*hook_expand)(void *extra, hook_expand_t type, void *address,
size_t old_usize, size_t new_usize, uintptr_t result_raw,
uintptr_t args_raw[4]);
typedef struct hooks_s hooks_t;
struct hooks_s {
hook_alloc alloc_hook;
hook_alloc alloc_hook;
hook_dalloc dalloc_hook;
hook_expand expand_hook;
void *extra;
void *extra;
};
/*
@ -143,9 +143,9 @@ struct hook_ralloc_args_s {
* Returns an opaque handle to be used when removing the hook. NULL means that
* we couldn't install the hook.
*/
bool hook_boot();
bool hook_boot(void);
void *hook_install(tsdn_t *tsdn, hooks_t *hooks);
void *hook_install(tsdn_t *tsdn, hooks_t *to_install);
/* Uninstalls the hook with the handle previously returned from hook_install. */
void hook_remove(tsdn_t *tsdn, void *opaque);
@ -154,8 +154,8 @@ void hook_remove(tsdn_t *tsdn, void *opaque);
void hook_invoke_alloc(hook_alloc_t type, void *result, uintptr_t result_raw,
uintptr_t args_raw[3]);
void hook_invoke_dalloc(hook_dalloc_t type, void *address,
uintptr_t args_raw[3]);
void hook_invoke_dalloc(
hook_dalloc_t type, void *address, uintptr_t args_raw[3]);
void hook_invoke_expand(hook_expand_t type, void *address, size_t old_usize,
size_t new_usize, uintptr_t result_raw, uintptr_t args_raw[4]);

87
include/jemalloc/internal/hpa.h
View file

@ -1,42 +1,18 @@
#ifndef JEMALLOC_INTERNAL_HPA_H
#define JEMALLOC_INTERNAL_HPA_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/base.h"
#include "jemalloc/internal/edata_cache.h"
#include "jemalloc/internal/emap.h"
#include "jemalloc/internal/exp_grow.h"
#include "jemalloc/internal/hpa_central.h"
#include "jemalloc/internal/hpa_hooks.h"
#include "jemalloc/internal/hpa_opts.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/pai.h"
#include "jemalloc/internal/psset.h"
typedef struct hpa_central_s hpa_central_t;
struct hpa_central_s {
/*
* The mutex guarding most of the operations on the central data
* structure.
*/
malloc_mutex_t mtx;
/*
* Guards expansion of eden. We separate this from the regular mutex so
* that cheaper operations can still continue while we're doing the OS
* call.
*/
malloc_mutex_t grow_mtx;
/*
* Either NULL (if empty), or some integer multiple of a
* hugepage-aligned number of hugepages. We carve them off one at a
* time to satisfy new pageslab requests.
*
* Guarded by grow_mtx.
*/
void *eden;
size_t eden_len;
/* Source for metadata. */
base_t *base;
/* Number of grow operations done on this hpa_central_t. */
uint64_t age_counter;
/* The HPA hooks. */
hpa_hooks_t hooks;
};
#include "jemalloc/internal/sec.h"
typedef struct hpa_shard_nonderived_stats_s hpa_shard_nonderived_stats_t;
struct hpa_shard_nonderived_stats_s {
@ -61,6 +37,14 @@ struct hpa_shard_nonderived_stats_s {
* Guarded by mtx.
*/
uint64_t nhugifies;
/*
* The number of times we've tried to hugify a pageslab, but failed.
*
* Guarded by mtx.
*/
uint64_t nhugify_failures;
/*
* The number of times we've dehugified a pageslab.
*
@ -72,8 +56,9 @@ struct hpa_shard_nonderived_stats_s {
/* Completely derived; only used by CTL. */
typedef struct hpa_shard_stats_s hpa_shard_stats_t;
struct hpa_shard_stats_s {
psset_stats_t psset_stats;
psset_stats_t psset_stats;
hpa_shard_nonderived_stats_t nonderived_stats;
sec_stats_t secstats;
};
typedef struct hpa_shard_s hpa_shard_t;
@ -86,14 +71,17 @@ struct hpa_shard_s {
/* The central allocator we get our hugepages from. */
hpa_central_t *central;
/* Protects most of this shard's state. */
malloc_mutex_t mtx;
/*
* Guards the shard's access to the central allocator (preventing
* multiple threads operating on this shard from accessing the central
* allocator).
*/
malloc_mutex_t grow_mtx;
/* The base metadata allocator. */
base_t *base;
@ -104,6 +92,9 @@ struct hpa_shard_s {
*/
edata_cache_fast_t ecf;
/* Small extent cache (not guarded by mtx) */
JEMALLOC_ALIGNED(CACHELINE) sec_t sec;
psset_t psset;
/*
@ -141,22 +132,31 @@ struct hpa_shard_s {
* Last time we performed purge on this shard.
*/
nstime_t last_purge;
/*
* Last time when we attempted work (purging or hugifying). If deferral
* of the work is allowed (we have background thread), this is the time
* when background thread checked if purging or hugifying needs to be
* done. If deferral is not allowed, this is the time of (hpa_alloc or
* hpa_dalloc) activity in the shard.
*/
nstime_t last_time_work_attempted;
};
bool hpa_hugepage_size_exceeds_limit(void);
/*
* Whether or not the HPA can be used given the current configuration. This is
* is not necessarily a guarantee that it backs its allocations by hugepages,
* just that it can function properly given the system it's running on.
*/
bool hpa_supported();
bool hpa_central_init(hpa_central_t *central, base_t *base, const hpa_hooks_t *hooks);
bool hpa_shard_init(hpa_shard_t *shard, hpa_central_t *central, emap_t *emap,
base_t *base, edata_cache_t *edata_cache, unsigned ind,
const hpa_shard_opts_t *opts);
bool hpa_supported(void);
bool hpa_shard_init(tsdn_t *tsdn, hpa_shard_t *shard, hpa_central_t *central,
emap_t *emap, base_t *base, edata_cache_t *edata_cache, unsigned ind,
const hpa_shard_opts_t *opts, const sec_opts_t *sec_opts);
void hpa_shard_stats_accum(hpa_shard_stats_t *dst, hpa_shard_stats_t *src);
void hpa_shard_stats_merge(tsdn_t *tsdn, hpa_shard_t *shard,
hpa_shard_stats_t *dst);
void hpa_shard_stats_merge(
tsdn_t *tsdn, hpa_shard_t *shard, hpa_shard_stats_t *dst);
/*
* Notify the shard that we won't use it for allocations much longer. Due to
@ -165,15 +165,18 @@ void hpa_shard_stats_merge(tsdn_t *tsdn, hpa_shard_t *shard,
*/
void hpa_shard_disable(tsdn_t *tsdn, hpa_shard_t *shard);
void hpa_shard_destroy(tsdn_t *tsdn, hpa_shard_t *shard);
/* Flush caches that shard may be using */
void hpa_shard_flush(tsdn_t *tsdn, hpa_shard_t *shard);
void hpa_shard_set_deferral_allowed(tsdn_t *tsdn, hpa_shard_t *shard,
bool deferral_allowed);
void hpa_shard_set_deferral_allowed(
tsdn_t *tsdn, hpa_shard_t *shard, bool deferral_allowed);
void hpa_shard_do_deferred_work(tsdn_t *tsdn, hpa_shard_t *shard);
/*
* We share the fork ordering with the PA and arena prefork handling; that's why
* these are 3 and 4 rather than 0 and 1.
* these are 2, 3 and 4 rather than 0 and 1.
*/
void hpa_shard_prefork2(tsdn_t *tsdn, hpa_shard_t *shard);
void hpa_shard_prefork3(tsdn_t *tsdn, hpa_shard_t *shard);
void hpa_shard_prefork4(tsdn_t *tsdn, hpa_shard_t *shard);
void hpa_shard_postfork_parent(tsdn_t *tsdn, hpa_shard_t *shard);

41
include/jemalloc/internal/hpa_central.h Normal file
View file

@ -0,0 +1,41 @@
#ifndef JEMALLOC_INTERNAL_HPA_CENTRAL_H
#define JEMALLOC_INTERNAL_HPA_CENTRAL_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/base.h"
#include "jemalloc/internal/hpa_hooks.h"
#include "jemalloc/internal/hpdata.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/tsd_types.h"
typedef struct hpa_central_s hpa_central_t;
struct hpa_central_s {
/*
* Guards expansion of eden. We separate this from the regular mutex so
* that cheaper operations can still continue while we're doing the OS
* call.
*/
malloc_mutex_t grow_mtx;
/*
* Either NULL (if empty), or some integer multiple of a
* hugepage-aligned number of hugepages. We carve them off one at a
* time to satisfy new pageslab requests.
*
* Guarded by grow_mtx.
*/
void *eden;
size_t eden_len;
/* Source for metadata. */
base_t *base;
/* The HPA hooks. */
hpa_hooks_t hooks;
};
bool hpa_central_init(
hpa_central_t *central, base_t *base, const hpa_hooks_t *hooks);
hpdata_t *hpa_central_extract(tsdn_t *tsdn, hpa_central_t *central, size_t size,
uint64_t age, bool hugify_eager, bool *oom);
#endif /* JEMALLOC_INTERNAL_HPA_CENTRAL_H */

8
include/jemalloc/internal/hpa_hooks.h
View file

@ -1,17 +1,21 @@
#ifndef JEMALLOC_INTERNAL_HPA_HOOKS_H
#define JEMALLOC_INTERNAL_HPA_HOOKS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/nstime.h"
typedef struct hpa_hooks_s hpa_hooks_t;
struct hpa_hooks_s {
void *(*map)(size_t size);
void (*unmap)(void *ptr, size_t size);
void (*purge)(void *ptr, size_t size);
void (*hugify)(void *ptr, size_t size);
bool (*hugify)(void *ptr, size_t size, bool sync);
void (*dehugify)(void *ptr, size_t size);
void (*curtime)(nstime_t *r_time, bool first_reading);
uint64_t (*ms_since)(nstime_t *r_time);
bool (*vectorized_purge)(void *vec, size_t vlen, size_t nbytes);
};
extern hpa_hooks_t hpa_hooks_default;
extern const hpa_hooks_t hpa_hooks_default;
#endif /* JEMALLOC_INTERNAL_HPA_HOOKS_H */

118
include/jemalloc/internal/hpa_opts.h
View file

@ -1,13 +1,66 @@
#ifndef JEMALLOC_INTERNAL_HPA_OPTS_H
#define JEMALLOC_INTERNAL_HPA_OPTS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/fxp.h"
/*
* This file is morally part of hpa.h, but is split out for header-ordering
* reasons.
*
* All of these hpa_shard_opts below are experimental. We are exploring more
* efficient packing, hugifying, and purging approaches to make efficient
* trade-offs between CPU, memory, latency, and usability. This means all of
* them are at the risk of being deprecated and corresponding configurations
* should be updated once the final version settles.
*/
/*
* This enum controls how jemalloc hugifies/dehugifies pages. Each style may be
* more suitable depending on deployment environments.
*
* hpa_hugify_style_none
* Using this means that jemalloc will not be hugifying or dehugifying pages,
* but will let the kernel make those decisions. This style only makes sense
* when deploying on systems where THP are enabled in 'always' mode. With this
* style, you most likely want to have no purging at all (dirty_mult=-1) or
* purge_threshold=HUGEPAGE bytes (2097152 for 2Mb page), although other
* thresholds may work well depending on kernel settings of your deployment
* targets.
*
* hpa_hugify_style_eager
* This style results in jemalloc giving hugepage advice, if needed, to
* anonymous memory immediately after it is mapped, so huge pages can be backing
* that memory at page-fault time. This is usually more efficient than doing
* it later, and it allows us to benefit from the hugepages from the start.
* Same options for purging as for the style 'none' are good starting choices:
* no purging, or purge_threshold=HUGEPAGE, some min_purge_delay_ms that allows
* for page not to be purged quickly, etc. This is a good choice if you can
* afford extra memory and your application gets performance increase from
* transparent hughepages.
*
* hpa_hugify_style_lazy
* This style is suitable when you purge more aggressively (you sacrifice CPU
* performance for less memory). When this style is chosen, jemalloc will
* hugify once hugification_threshold is reached, and dehugify before purging.
* If the kernel is configured to use direct compaction you may experience some
* allocation latency when using this style. The best is to measure what works
* better for your application needs, and in the target deployment environment.
* This is a good choice for apps that cannot afford a lot of memory regression,
* but would still like to benefit from backing certain memory regions with
* hugepages.
*/
enum hpa_hugify_style_e {
hpa_hugify_style_auto = 0,
hpa_hugify_style_none = 1,
hpa_hugify_style_eager = 2,
hpa_hugify_style_lazy = 3,
hpa_hugify_style_limit = hpa_hugify_style_lazy + 1
};
typedef enum hpa_hugify_style_e hpa_hugify_style_t;
extern const char *const hpa_hugify_style_names[];
typedef struct hpa_shard_opts_s hpa_shard_opts_t;
struct hpa_shard_opts_s {
/*
@ -44,12 +97,64 @@ struct hpa_shard_opts_s {
*/
uint64_t hugify_delay_ms;
/*
* Hugify pages synchronously (hugify will happen even if hugify_style
* is not hpa_hugify_style_lazy).
*/
bool hugify_sync;
/*
* Minimum amount of time between purges.
*/
uint64_t min_purge_interval_ms;
/*
* Maximum number of hugepages to purge on each purging attempt.
*/
ssize_t experimental_max_purge_nhp;
/*
* Minimum number of inactive bytes needed for a non-empty page to be
* considered purgable.
*
* When the number of touched inactive bytes on non-empty hugepage is
* >= purge_threshold, the page is purgable. Empty pages are always
* purgable. Setting this to HUGEPAGE bytes would only purge empty
* pages if using hugify_style_eager and the purges would be exactly
* HUGEPAGE bytes. Depending on your kernel settings, this may result
* in better performance.
*
* Please note, when threshold is reached, we will purge all the dirty
* bytes, and not just up to the threshold. If this is PAGE bytes, then
* all the pages that have any dirty bytes are purgable. We treat
* purgability constraint for purge_threshold as stronger than
* dirty_mult, IOW, if no page meets purge_threshold, we will not purge
* even if we are above dirty_mult.
*/
size_t purge_threshold;
/*
* Minimum number of ms that needs to elapse between HP page becoming
* eligible for purging and actually getting purged.
*
* Setting this to a larger number would give better chance of reusing
* that memory. Setting it to 0 means that page is eligible for purging
* as soon as it meets the purge_threshold. The clock resets when
* purgability of the page changes (page goes from being non-purgable to
* purgable). When using eager style you probably want to allow for
* some delay, to avoid purging the page too quickly and give it time to
* be used.
*/
uint64_t min_purge_delay_ms;
/*
* Style of hugification/dehugification (see comment at
* hpa_hugify_style_t for options).
*/
hpa_hugify_style_t hugify_style;
};
/* clang-format off */
#define HPA_SHARD_OPTS_DEFAULT { \
/* slab_max_alloc */ \
64 * 1024, \
@ -67,8 +172,19 @@ struct hpa_shard_opts_s {
false, \
/* hugify_delay_ms */ \
10 * 1000, \
/* hugify_sync */ \
false, \
/* min_purge_interval_ms */ \
5 * 1000 \
5 * 1000, \
/* experimental_max_purge_nhp */ \
-1, \
/* size_t purge_threshold */ \
PAGE, \
/* min_purge_delay_ms */ \
0, \
/* hugify_style */ \
hpa_hugify_style_lazy \
}
/* clang-format on */
#endif /* JEMALLOC_INTERNAL_HPA_OPTS_H */

161
include/jemalloc/internal/hpa_utils.h Normal file
View file

@ -0,0 +1,161 @@
#ifndef JEMALLOC_INTERNAL_HPA_UTILS_H
#define JEMALLOC_INTERNAL_HPA_UTILS_H
#include "jemalloc/internal/hpa.h"
#include "jemalloc/internal/extent.h"
#define HPA_MIN_VAR_VEC_SIZE 8
/*
* This is used for jemalloc internal tuning and may change in the future based
* on production traffic.
*
* This value protects two things:
* 1. Stack size
* 2. Number of huge pages that are being purged in a batch as we do not
* allow allocations while making madvise syscall.
*/
#define HPA_PURGE_BATCH_MAX 16
#ifdef JEMALLOC_HAVE_PROCESS_MADVISE
typedef struct iovec hpa_io_vector_t;
#else
typedef struct {
void *iov_base;
size_t iov_len;
} hpa_io_vector_t;
#endif
static inline size_t
hpa_process_madvise_max_iovec_len(void) {
assert(
opt_process_madvise_max_batch <= PROCESS_MADVISE_MAX_BATCH_LIMIT);
return opt_process_madvise_max_batch == 0
? HPA_MIN_VAR_VEC_SIZE
: opt_process_madvise_max_batch;
}
/* Actually invoke hooks. If we fail vectorized, use single purges */
static void
hpa_try_vectorized_purge(
hpa_hooks_t *hooks, hpa_io_vector_t *vec, size_t vlen, size_t nbytes) {
bool success = opt_process_madvise_max_batch > 0
&& !hooks->vectorized_purge(vec, vlen, nbytes);
if (!success) {
/* On failure, it is safe to purge again (potential perf
* penalty) If kernel can tell exactly which regions
* failed, we could avoid that penalty.
*/
for (size_t i = 0; i < vlen; ++i) {
hooks->purge(vec[i].iov_base, vec[i].iov_len);
}
}
}
/*
* This structure accumulates the regions for process_madvise. It invokes the
* hook when batch limit is reached.
*/
typedef struct {
hpa_io_vector_t *vp;
size_t cur;
size_t total_bytes;
size_t capacity;
} hpa_range_accum_t;
static inline void
hpa_range_accum_init(hpa_range_accum_t *ra, hpa_io_vector_t *v, size_t sz) {
ra->vp = v;
ra->capacity = sz;
ra->total_bytes = 0;
ra->cur = 0;
}
static inline void
hpa_range_accum_flush(hpa_range_accum_t *ra, hpa_hooks_t *hooks) {
assert(ra->total_bytes > 0 && ra->cur > 0);
hpa_try_vectorized_purge(hooks, ra->vp, ra->cur, ra->total_bytes);
ra->cur = 0;
ra->total_bytes = 0;
}
static inline void
hpa_range_accum_add(
hpa_range_accum_t *ra, void *addr, size_t sz, hpa_hooks_t *hooks) {
assert(ra->cur < ra->capacity);
ra->vp[ra->cur].iov_base = addr;
ra->vp[ra->cur].iov_len = sz;
ra->total_bytes += sz;
ra->cur++;
if (ra->cur == ra->capacity) {
hpa_range_accum_flush(ra, hooks);
}
}
static inline void
hpa_range_accum_finish(hpa_range_accum_t *ra, hpa_hooks_t *hooks) {
if (ra->cur > 0) {
hpa_range_accum_flush(ra, hooks);
}
}
/*
* For purging more than one page we use batch of these items
*/
typedef struct {
hpdata_purge_state_t state;
hpdata_t *hp;
bool dehugify;
} hpa_purge_item_t;
typedef struct hpa_purge_batch_s hpa_purge_batch_t;
struct hpa_purge_batch_s {
hpa_purge_item_t *items;
size_t items_capacity;
/* Number of huge pages to purge in current batch */
size_t item_cnt;
/* Number of ranges to purge in current batch */
size_t nranges;
/* Total number of dirty pages in current batch*/
size_t ndirty_in_batch;
/* Max number of huge pages to purge */
size_t max_hp;
/*
* Once we are above this watermark we should not add more pages
* to the same batch. This is because while we want to minimize
* number of madvise calls we also do not want to be preventing
* allocations from too many huge pages (which we have to do
* while they are being purged)
*/
size_t range_watermark;
size_t npurged_hp_total;
};
static inline bool
hpa_batch_full(hpa_purge_batch_t *b) {
/* It's okay for ranges to go above */
return b->npurged_hp_total == b->max_hp
|| b->item_cnt == b->items_capacity
|| b->nranges >= b->range_watermark;
}
static inline void
hpa_batch_pass_start(hpa_purge_batch_t *b) {
b->item_cnt = 0;
b->nranges = 0;
b->ndirty_in_batch = 0;
}
static inline bool
hpa_batch_empty(hpa_purge_batch_t *b) {
return b->item_cnt == 0;
}
/* Purge pages in a batch using given hooks */
void hpa_purge_batch(
hpa_hooks_t *hooks, hpa_purge_item_t *batch, size_t batch_sz);
#endif /* JEMALLOC_INTERNAL_HPA_UTILS_H */

149
include/jemalloc/internal/hpdata.h
View file

@ -1,7 +1,10 @@
#ifndef JEMALLOC_INTERNAL_HPDATA_H
#define JEMALLOC_INTERNAL_HPDATA_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/fb.h"
#include "jemalloc/internal/nstime.h"
#include "jemalloc/internal/pages.h"
#include "jemalloc/internal/ph.h"
#include "jemalloc/internal/ql.h"
#include "jemalloc/internal/typed_list.h"
@ -17,8 +20,14 @@
* an observable property of any given region of address space). It's just
* hugepage-sized and hugepage-aligned; it's *potentially* huge.
*/
/*
* The max enumeration num should not exceed 2^16 - 1, see comments in edata.h
* for ESET_ENUMERATE_MAX_NUM for more details.
*/
#define PSSET_ENUMERATE_MAX_NUM 32
typedef struct hpdata_s hpdata_t;
ph_structs(hpdata_age_heap, hpdata_t);
ph_structs(hpdata_age_heap, hpdata_t, PSSET_ENUMERATE_MAX_NUM);
struct hpdata_s {
/*
* We likewise follow the edata convention of mangling names and forcing
@ -64,7 +73,7 @@ struct hpdata_s {
bool h_hugify_allowed;
/* When we became a hugification candidate. */
nstime_t h_time_hugify_allowed;
bool h_in_psset_hugify_container;
bool h_in_psset_hugify_container;
/* Whether or not a purge or hugify is currently happening. */
bool h_mid_purge;
@ -115,6 +124,12 @@ struct hpdata_s {
/* The touched pages (using the same definition as above). */
fb_group_t touched_pages[FB_NGROUPS(HUGEPAGE_PAGES)];
/* Time when this extent (hpdata) becomes eligible for purging */
nstime_t h_time_purge_allowed;
/* True if the extent was huge and empty last time when it was purged */
bool h_purged_when_empty_and_huge;
};
TYPED_LIST(hpdata_empty_list, hpdata_t, ql_link_empty)
@ -177,8 +192,8 @@ hpdata_purge_allowed_get(const hpdata_t *hpdata) {
static inline void
hpdata_purge_allowed_set(hpdata_t *hpdata, bool purge_allowed) {
assert(purge_allowed == false || !hpdata->h_mid_purge);
hpdata->h_purge_allowed = purge_allowed;
assert(purge_allowed == false || !hpdata->h_mid_purge);
hpdata->h_purge_allowed = purge_allowed;
}
static inline bool
@ -241,7 +256,6 @@ hpdata_changing_state_get(const hpdata_t *hpdata) {
return hpdata->h_mid_purge || hpdata->h_mid_hugify;
}
static inline bool
hpdata_updating_get(const hpdata_t *hpdata) {
return hpdata->h_updating;
@ -276,17 +290,17 @@ hpdata_longest_free_range_set(hpdata_t *hpdata, size_t longest_free_range) {
}
static inline size_t
hpdata_nactive_get(hpdata_t *hpdata) {
hpdata_nactive_get(const hpdata_t *hpdata) {
return hpdata->h_nactive;
}
static inline size_t
hpdata_ntouched_get(hpdata_t *hpdata) {
hpdata_ntouched_get(const hpdata_t *hpdata) {
return hpdata->h_ntouched;
}
static inline size_t
hpdata_ndirty_get(hpdata_t *hpdata) {
hpdata_ndirty_get(const hpdata_t *hpdata) {
return hpdata->h_ntouched - hpdata->h_nactive;
}
@ -295,6 +309,26 @@ hpdata_nretained_get(hpdata_t *hpdata) {
return HUGEPAGE_PAGES - hpdata->h_ntouched;
}
static inline void
hpdata_time_purge_allowed_set(hpdata_t *hpdata, const nstime_t *v) {
nstime_copy(&hpdata->h_time_purge_allowed, v);
}
static inline const nstime_t *
hpdata_time_purge_allowed_get(const hpdata_t *hpdata) {
return &hpdata->h_time_purge_allowed;
}
static inline bool
hpdata_purged_when_empty_and_huge_get(const hpdata_t *hpdata) {
return hpdata->h_purged_when_empty_and_huge;
}
static inline void
hpdata_purged_when_empty_and_huge_set(hpdata_t *hpdata, bool v) {
hpdata->h_purged_when_empty_and_huge = v;
}
static inline void
hpdata_assert_empty(hpdata_t *hpdata) {
assert(fb_empty(hpdata->active_pages, HUGEPAGE_PAGES));
@ -308,58 +342,95 @@ hpdata_assert_empty(hpdata_t *hpdata) {
*/
static inline bool
hpdata_consistent(hpdata_t *hpdata) {
if(fb_urange_longest(hpdata->active_pages, HUGEPAGE_PAGES)
!= hpdata_longest_free_range_get(hpdata)) {
return false;
bool res = true;
const size_t active_urange_longest = fb_urange_longest(
hpdata->active_pages, HUGEPAGE_PAGES);
const size_t longest_free_range = hpdata_longest_free_range_get(hpdata);
if (active_urange_longest != longest_free_range) {
malloc_printf(
"<jemalloc>: active_fb_urange_longest=%zu != hpdata_longest_free_range=%zu\n",
active_urange_longest, longest_free_range);
res = false;
}
if (fb_scount(hpdata->active_pages, HUGEPAGE_PAGES, 0, HUGEPAGE_PAGES)
!= hpdata->h_nactive) {
return false;
const size_t active_scount = fb_scount(
hpdata->active_pages, HUGEPAGE_PAGES, 0, HUGEPAGE_PAGES);
if (active_scount != hpdata->h_nactive) {
malloc_printf(
"<jemalloc>: active_fb_scount=%zu != hpdata_nactive=%zu\n",
active_scount, hpdata->h_nactive);
res = false;
}
if (fb_scount(hpdata->touched_pages, HUGEPAGE_PAGES, 0, HUGEPAGE_PAGES)
!= hpdata->h_ntouched) {
return false;
const size_t touched_scount = fb_scount(
hpdata->touched_pages, HUGEPAGE_PAGES, 0, HUGEPAGE_PAGES);
if (touched_scount != hpdata->h_ntouched) {
malloc_printf(
"<jemalloc>: touched_fb_scount=%zu != hpdata_ntouched=%zu\n",
touched_scount, hpdata->h_ntouched);
res = false;
}
if (hpdata->h_ntouched < hpdata->h_nactive) {
return false;
malloc_printf(
"<jemalloc>: hpdata_ntouched=%zu < hpdata_nactive=%zu\n",
hpdata->h_ntouched, hpdata->h_nactive);
res = false;
}
if (hpdata->h_huge && hpdata->h_ntouched != HUGEPAGE_PAGES) {
return false;
if (hpdata->h_huge && (hpdata->h_ntouched != HUGEPAGE_PAGES)) {
malloc_printf(
"<jemalloc>: hpdata_huge=%d && (hpdata_ntouched=%zu != hugepage_pages=%zu)\n",
hpdata->h_huge, hpdata->h_ntouched, HUGEPAGE_PAGES);
res = false;
}
if (hpdata_changing_state_get(hpdata)
&& ((hpdata->h_purge_allowed) || hpdata->h_hugify_allowed)) {
return false;
const bool changing_state = hpdata_changing_state_get(hpdata);
if (changing_state
&& (hpdata->h_purge_allowed || hpdata->h_hugify_allowed)) {
malloc_printf(
"<jemalloc>: hpdata_changing_state=%d && (hpdata_purge_allowed=%d || hpdata_hugify_allowed=%d)\n",
changing_state, hpdata->h_purge_allowed,
hpdata->h_hugify_allowed);
res = false;
}
if (hpdata_hugify_allowed_get(hpdata)
!= hpdata_in_psset_hugify_container_get(hpdata)) {
return false;
malloc_printf(
"<jemalloc>: hpdata_hugify_allowed=%d != hpdata_in_psset_hugify_container=%d\n",
hpdata_hugify_allowed_get(hpdata),
hpdata_in_psset_hugify_container_get(hpdata));
res = false;
}
return true;
return res;
}
static inline void
hpdata_assert_consistent(hpdata_t *hpdata) {
assert(hpdata_consistent(hpdata));
}
#define hpdata_assert_consistent(hpdata) \
do { \
assert(hpdata_consistent(hpdata)); \
} while (0)
static inline bool
hpdata_empty(hpdata_t *hpdata) {
hpdata_empty(const hpdata_t *hpdata) {
return hpdata->h_nactive == 0;
}
static inline bool
hpdata_full(hpdata_t *hpdata) {
hpdata_full(const hpdata_t *hpdata) {
return hpdata->h_nactive == HUGEPAGE_PAGES;
}
void hpdata_init(hpdata_t *hpdata, void *addr, uint64_t age);
void hpdata_init(hpdata_t *hpdata, void *addr, uint64_t age, bool is_huge);
/*
* Given an hpdata which can serve an allocation request, pick and reserve an
* offset within that allocation.
*/
void *hpdata_reserve_alloc(hpdata_t *hpdata, size_t sz);
void hpdata_unreserve(hpdata_t *hpdata, void *begin, size_t sz);
void hpdata_unreserve(hpdata_t *hpdata, void *addr, size_t sz);
/*
* The hpdata_purge_prepare_t allows grabbing the metadata required to purge
@ -368,10 +439,10 @@ void hpdata_unreserve(hpdata_t *hpdata, void *begin, size_t sz);
*/
typedef struct hpdata_purge_state_s hpdata_purge_state_t;
struct hpdata_purge_state_s {
size_t npurged;
size_t ndirty_to_purge;
size_t npurged;
size_t ndirty_to_purge;
fb_group_t to_purge[FB_NGROUPS(HUGEPAGE_PAGES)];
size_t next_purge_search_begin;
size_t next_purge_search_begin;
};
/*
@ -386,9 +457,11 @@ struct hpdata_purge_state_s {
* until you're done, and then end. Allocating out of an hpdata undergoing
* purging is not allowed.
*
* Returns the number of dirty pages that will be purged.
* Returns the number of dirty pages that will be purged and sets nranges
* to number of ranges with dirty pages that will be purged.
*/
size_t hpdata_purge_begin(hpdata_t *hpdata, hpdata_purge_state_t *purge_state);
size_t hpdata_purge_begin(
hpdata_t *hpdata, hpdata_purge_state_t *purge_state, size_t *nranges);
/*
* If there are more extents to purge, sets *r_purge_addr and *r_purge_size to

13
include/jemalloc/internal/inspect.h
View file

@ -1,6 +1,9 @@
#ifndef JEMALLOC_INTERNAL_INSPECT_H
#define JEMALLOC_INTERNAL_INSPECT_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/tsd_types.h"
/*
* This module contains the heap introspection capabilities. For now they are
* exposed purely through mallctl APIs in the experimental namespace, but this
@ -23,7 +26,7 @@ typedef struct inspect_extent_util_stats_verbose_s
inspect_extent_util_stats_verbose_t;
struct inspect_extent_util_stats_verbose_s {
void *slabcur_addr;
void *slabcur_addr;
size_t nfree;
size_t nregs;
size_t size;
@ -31,10 +34,10 @@ struct inspect_extent_util_stats_verbose_s {
size_t bin_nregs;
};
void inspect_extent_util_stats_get(tsdn_t *tsdn, const void *ptr,
size_t *nfree, size_t *nregs, size_t *size);
void inspect_extent_util_stats_get(
tsdn_t *tsdn, const void *ptr, size_t *nfree, size_t *nregs, size_t *size);
void inspect_extent_util_stats_verbose_get(tsdn_t *tsdn, const void *ptr,
size_t *nfree, size_t *nregs, size_t *size,
size_t *bin_nfree, size_t *bin_nregs, void **slabcur_addr);
size_t *nfree, size_t *nregs, size_t *size, size_t *bin_nfree,
size_t *bin_nregs, void **slabcur_addr);
#endif /* JEMALLOC_INTERNAL_INSPECT_H */

142
include/jemalloc/internal/jemalloc_internal_decls.h
View file

@ -3,64 +3,65 @@
#include <math.h>
#ifdef _WIN32
# include <windows.h>
# include "msvc_compat/windows_extra.h"
# include "msvc_compat/strings.h"
# ifdef _WIN64
# if LG_VADDR <= 32
# error Generate the headers using x64 vcargs
# endif
# else
# if LG_VADDR > 32
# undef LG_VADDR
# define LG_VADDR 32
# endif
# endif
# include <windows.h>
# include "msvc_compat/windows_extra.h"
# include "msvc_compat/strings.h"
# ifdef _WIN64
# if LG_VADDR <= 32
# error Generate the headers using x64 vcargs
# endif
# else
# if LG_VADDR > 32
# undef LG_VADDR
# define LG_VADDR 32
# endif
# endif
#else
# include <sys/param.h>
# include <sys/mman.h>
# if !defined(__pnacl__) && !defined(__native_client__)
# include <sys/syscall.h>
# if !defined(SYS_write) && defined(__NR_write)
# define SYS_write __NR_write
# endif
# if defined(SYS_open) && defined(__aarch64__)
/* Android headers may define SYS_open to __NR_open even though
# include <sys/param.h>
# include <sys/mman.h>
# if !defined(__pnacl__) && !defined(__native_client__)
# include <sys/syscall.h>
# if !defined(SYS_write) && defined(__NR_write)
# define SYS_write __NR_write
# endif
# if defined(SYS_open) && defined(__aarch64__)
/* Android headers may define SYS_open to __NR_open even though
* __NR_open may not exist on AArch64 (superseded by __NR_openat). */
# undef SYS_open
# endif
# include <sys/uio.h>
# endif
# include <pthread.h>
# if defined(__FreeBSD__) || defined(__DragonFly__)
# include <pthread_np.h>
# include <sched.h>
# if defined(__FreeBSD__)
# define cpu_set_t cpuset_t
# endif
# endif
# include <signal.h>
# ifdef JEMALLOC_OS_UNFAIR_LOCK
# include <os/lock.h>
# endif
# ifdef JEMALLOC_GLIBC_MALLOC_HOOK
# include <sched.h>
# endif
# include <errno.h>
# include <sys/time.h>
# include <time.h>
# ifdef JEMALLOC_HAVE_MACH_ABSOLUTE_TIME
# include <mach/mach_time.h>
# endif
# undef SYS_open
# endif
# include <sys/uio.h>
# endif
# include <pthread.h>
# if defined(__FreeBSD__) || defined(__DragonFly__) \
|| defined(__OpenBSD__)
# include <pthread_np.h>
# include <sched.h>
# if defined(__FreeBSD__)
# define cpu_set_t cpuset_t
# endif
# endif
# include <signal.h>
# ifdef JEMALLOC_OS_UNFAIR_LOCK
# include <os/lock.h>
# endif
# ifdef JEMALLOC_GLIBC_MALLOC_HOOK
# include <sched.h>
# endif
# include <errno.h>
# include <sys/time.h>
# include <time.h>
# ifdef JEMALLOC_HAVE_MACH_ABSOLUTE_TIME
# include <mach/mach_time.h>
# endif
#endif
#include <sys/types.h>
#include <limits.h>
#ifndef SIZE_T_MAX
# define SIZE_T_MAX SIZE_MAX
# define SIZE_T_MAX SIZE_MAX
#endif
#ifndef SSIZE_MAX
# define SSIZE_MAX ((ssize_t)(SIZE_T_MAX >> 1))
# define SSIZE_MAX ((ssize_t)(SIZE_T_MAX >> 1))
#endif
#include <stdarg.h>
#include <stdbool.h>
@ -69,30 +70,30 @@
#include <stdint.h>
#include <stddef.h>
#ifndef offsetof
# define offsetof(type, member) ((size_t)&(((type *)NULL)->member))
# define offsetof(type, member) ((size_t) & (((type *)NULL)->member))
#endif
#include <string.h>
#include <strings.h>
#include <ctype.h>
#ifdef _MSC_VER
# include <io.h>
# include <io.h>
typedef intptr_t ssize_t;
# define PATH_MAX 1024
# define STDERR_FILENO 2
# define __func__ __FUNCTION__
# ifdef JEMALLOC_HAS_RESTRICT
# define restrict __restrict
# endif
# define PATH_MAX 1024
# define STDERR_FILENO 2
# define __func__ __FUNCTION__
# ifdef JEMALLOC_HAS_RESTRICT
# define restrict __restrict
# endif
/* Disable warnings about deprecated system functions. */
# pragma warning(disable: 4996)
#if _MSC_VER < 1800
# pragma warning(disable : 4996)
# if _MSC_VER < 1800
static int
isblank(int c) {
return (c == '\t' || c == ' ');
}
#endif
# endif
#else
# include <unistd.h>
# include <unistd.h>
#endif
#include <fcntl.h>
@ -102,7 +103,24 @@ isblank(int c) {
* classes.
*/
#ifdef small
# undef small
# undef small
#endif
/*
* Oftentimes we'd like to perform some kind of arithmetic to obtain
* a pointer from another pointer but with some offset or mask applied.
* Naively you would accomplish this by casting the source pointer to
* `uintptr_t`, performing all of the relevant arithmetic, and then casting
* the result to the desired pointer type. However, this has the unfortunate
* side-effect of concealing pointer provenance, hiding useful information for
* optimization from the compiler (see here for details:
* https://clang.llvm.org/extra/clang-tidy/checks/performance/no-int-to-ptr.html
* )
* Instead what one should do is cast the source pointer to `char *` and perform
* the equivalent arithmetic (since `char` of course represents one byte). But
* because `char *` has the semantic meaning of "string", we define this typedef
* simply to make it clearer where we are performing such pointer arithmetic.
*/
typedef char byte_t;
#endif /* JEMALLOC_INTERNAL_H */

67
include/jemalloc/internal/jemalloc_internal_defs.h.in
View file

@ -14,10 +14,13 @@
*/
#undef JEMALLOC_OVERRIDE___LIBC_CALLOC
#undef JEMALLOC_OVERRIDE___LIBC_FREE
#undef JEMALLOC_OVERRIDE___LIBC_FREE_SIZED
#undef JEMALLOC_OVERRIDE___LIBC_FREE_ALIGNED_SIZED
#undef JEMALLOC_OVERRIDE___LIBC_MALLOC
#undef JEMALLOC_OVERRIDE___LIBC_MEMALIGN
#undef JEMALLOC_OVERRIDE___LIBC_REALLOC
#undef JEMALLOC_OVERRIDE___LIBC_VALLOC
#undef JEMALLOC_OVERRIDE___LIBC_PVALLOC
#undef JEMALLOC_OVERRIDE___POSIX_MEMALIGN
/*
@ -88,6 +91,9 @@
/* Defined if pthread_getname_np(3) is available. */
#undef JEMALLOC_HAVE_PTHREAD_GETNAME_NP
/* Defined if pthread_set_name_np(3) is available. */
#undef JEMALLOC_HAVE_PTHREAD_SET_NAME_NP
/* Defined if pthread_get_name_np(3) is available. */
#undef JEMALLOC_HAVE_PTHREAD_GET_NAME_NP
@ -111,6 +117,11 @@
*/
#undef JEMALLOC_HAVE_CLOCK_REALTIME
/*
* Defined if clock_gettime_nsec_np(CLOCK_UPTIME_RAW) is available.
*/
#undef JEMALLOC_HAVE_CLOCK_GETTIME_NSEC_NP
/*
* Defined if _malloc_thread_cleanup() exists. At least in the case of
* FreeBSD, pthread_key_create() allocates, which if used during malloc
@ -161,6 +172,15 @@
/* Use gcc intrinsics for profile backtracing if defined. */
#undef JEMALLOC_PROF_GCC
/* Use frame pointer for profile backtracing if defined. Linux only. */
#undef JEMALLOC_PROF_FRAME_POINTER
/* JEMALLOC_PAGEID enabled page id */
#undef JEMALLOC_PAGEID
/* JEMALLOC_HAVE_PRCTL checks prctl */
#undef JEMALLOC_HAVE_PRCTL
/*
* JEMALLOC_DSS enables use of sbrk(2) to allocate extents from the data storage
* segment (DSS).
@ -259,6 +279,12 @@
*/
#undef JEMALLOC_READLINKAT
/*
* If defined, use getenv() (instead of secure_getenv() or
* alternatives) to access MALLOC_CONF.
*/
#undef JEMALLOC_FORCE_GETENV
/*
* Darwin (OS X) uses zones to work around Mach-O symbol override shortcomings.
*/
@ -282,6 +308,13 @@
*/
#undef JEMALLOC_HAVE_MADVISE_HUGE
/*
* Defined if best-effort synchronous collapse of the native
* pages mapped by the memory range into transparent huge pages is supported
* via MADV_COLLAPSE arguments to madvise(2).
*/
#undef JEMALLOC_HAVE_MADVISE_COLLAPSE
/*
* Methods for purging unused pages differ between operating systems.
*
@ -312,9 +345,23 @@
*/
#undef JEMALLOC_MADVISE_NOCORE
/* Defined if process_madvise(2) is available. */
#undef JEMALLOC_HAVE_PROCESS_MADVISE
#undef EXPERIMENTAL_SYS_PROCESS_MADVISE_NR
/* Defined if mprotect(2) is available. */
#undef JEMALLOC_HAVE_MPROTECT
/* Defined if sys/sdt.h is available and sdt tracing enabled */
#undef JEMALLOC_EXPERIMENTAL_USDT_STAP
/*
* Defined if sys/sdt.h is unavailable, sdt tracing enabled, and
* platform is supported
*/
#undef JEMALLOC_EXPERIMENTAL_USDT_CUSTOM
/*
* Defined if transparent huge pages (THPs) are supported via the
* MADV_[NO]HUGEPAGE arguments to madvise(2), and THP support is enabled.
@ -378,12 +425,18 @@
/* Adaptive mutex support in pthreads. */
#undef JEMALLOC_HAVE_PTHREAD_MUTEX_ADAPTIVE_NP
/* gettid() support */
#undef JEMALLOC_HAVE_GETTID
/* GNU specific sched_getcpu support */
#undef JEMALLOC_HAVE_SCHED_GETCPU
/* GNU specific sched_setaffinity support */
#undef JEMALLOC_HAVE_SCHED_SETAFFINITY
/* pthread_setaffinity_np support */
#undef JEMALLOC_HAVE_PTHREAD_SETAFFINITY_NP
/*
* If defined, all the features necessary for background threads are present.
*/
@ -424,4 +477,18 @@
/* If defined, realloc(ptr, 0) defaults to "free" instead of "alloc". */
#undef JEMALLOC_ZERO_REALLOC_DEFAULT_FREE
/* If defined, use volatile asm during benchmarks. */
#undef JEMALLOC_HAVE_ASM_VOLATILE
/*
* If defined, support the use of rdtscp to get the time stamp counter
* and the processor ID.
*/
#undef JEMALLOC_HAVE_RDTSCP
/* If defined, use __int128 for optimization. */
#undef JEMALLOC_HAVE_INT128
#include "jemalloc/internal/jemalloc_internal_overrides.h"
#endif /* JEMALLOC_INTERNAL_DEFS_H_ */

86
include/jemalloc/internal/jemalloc_internal_externs.h
View file

@ -1,39 +1,52 @@
#ifndef JEMALLOC_INTERNAL_EXTERNS_H
#define JEMALLOC_INTERNAL_EXTERNS_H
#include "jemalloc/internal/arena_types.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/fxp.h"
#include "jemalloc/internal/hpa_opts.h"
#include "jemalloc/internal/nstime.h"
#include "jemalloc/internal/sec_opts.h"
#include "jemalloc/internal/tsd_types.h"
#include "jemalloc/internal/nstime.h"
/* TSD checks this to set thread local slow state accordingly. */
extern bool malloc_slow;
/* Run-time options. */
extern bool opt_abort;
extern bool opt_abort_conf;
extern bool opt_trust_madvise;
extern bool opt_confirm_conf;
extern bool opt_hpa;
extern bool opt_abort;
extern bool opt_abort_conf;
extern bool opt_trust_madvise;
extern bool opt_experimental_hpa_start_huge_if_thp_always;
extern bool opt_experimental_hpa_enforce_hugify;
extern bool opt_confirm_conf;
extern bool opt_hpa;
extern hpa_shard_opts_t opt_hpa_opts;
extern sec_opts_t opt_hpa_sec_opts;
extern sec_opts_t opt_hpa_sec_opts;
extern const char *opt_junk;
extern bool opt_junk_alloc;
extern bool opt_junk_free;
extern void (*junk_free_callback)(void *ptr, size_t size);
extern void (*junk_alloc_callback)(void *ptr, size_t size);
extern bool opt_utrace;
extern bool opt_xmalloc;
extern bool opt_experimental_infallible_new;
extern bool opt_zero;
extern unsigned opt_narenas;
extern bool opt_junk_alloc;
extern bool opt_junk_free;
extern void (*JET_MUTABLE junk_free_callback)(void *ptr, size_t size);
extern void (*JET_MUTABLE junk_alloc_callback)(void *ptr, size_t size);
extern void (*JET_MUTABLE invalid_conf_abort)(void);
extern bool opt_utrace;
extern bool opt_xmalloc;
extern bool opt_experimental_infallible_new;
extern bool opt_experimental_tcache_gc;
extern bool opt_zero;
extern unsigned opt_narenas;
extern fxp_t opt_narenas_ratio;
extern zero_realloc_action_t opt_zero_realloc_action;
extern malloc_init_t malloc_init_state;
extern const char *zero_realloc_mode_names[];
extern atomic_zu_t zero_realloc_count;
extern bool opt_cache_oblivious;
extern malloc_init_t malloc_init_state;
extern const char *const zero_realloc_mode_names[];
extern atomic_zu_t zero_realloc_count;
extern bool opt_cache_oblivious;
extern unsigned opt_debug_double_free_max_scan;
extern size_t opt_calloc_madvise_threshold;
extern bool opt_disable_large_size_classes;
extern const char *opt_malloc_conf_symlink;
extern const char *opt_malloc_conf_env_var;
/* Escape free-fastpath when ptr & mask == 0 (for sanitization purpose). */
extern uintptr_t san_cache_bin_nonfast_mask;
@ -53,23 +66,26 @@ extern unsigned manual_arena_base;
*/
extern atomic_p_t arenas[];
void *a0malloc(size_t size);
void a0dalloc(void *ptr);
void *bootstrap_malloc(size_t size);
void *bootstrap_calloc(size_t num, size_t size);
void bootstrap_free(void *ptr);
void arena_set(unsigned ind, arena_t *arena);
extern unsigned huge_arena_ind;
void *a0malloc(size_t size);
void a0dalloc(void *ptr);
void *bootstrap_malloc(size_t size);
void *bootstrap_calloc(size_t num, size_t size);
void bootstrap_free(void *ptr);
void arena_set(unsigned ind, arena_t *arena);
unsigned narenas_total_get(void);
arena_t *arena_init(tsdn_t *tsdn, unsigned ind, const arena_config_t *config);
arena_t *arena_choose_hard(tsd_t *tsd, bool internal);
void arena_migrate(tsd_t *tsd, arena_t *oldarena, arena_t *newarena);
void iarena_cleanup(tsd_t *tsd);
void arena_cleanup(tsd_t *tsd);
size_t batch_alloc(void **ptrs, size_t num, size_t size, int flags);
void jemalloc_prefork(void);
void jemalloc_postfork_parent(void);
void jemalloc_postfork_child(void);
void je_sdallocx_noflags(void *ptr, size_t size);
void *malloc_default(size_t size);
void arena_migrate(tsd_t *tsd, arena_t *oldarena, arena_t *newarena);
void iarena_cleanup(tsd_t *tsd);
void arena_cleanup(tsd_t *tsd);
size_t batch_alloc(void **ptrs, size_t num, size_t size, int flags);
void jemalloc_prefork(void);
void jemalloc_postfork_parent(void);
void jemalloc_postfork_child(void);
void sdallocx_default(void *ptr, size_t size, int flags);
void free_default(void *ptr);
void *malloc_default(size_t size);
#endif /* JEMALLOC_INTERNAL_EXTERNS_H */

17
include/jemalloc/internal/jemalloc_internal_inlines_a.h
View file

@ -1,10 +1,14 @@
#ifndef JEMALLOC_INTERNAL_INLINES_A_H
#define JEMALLOC_INTERNAL_INLINES_A_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_externs.h"
#include "jemalloc/internal/arena_types.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/bit_util.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#include "jemalloc/internal/sc.h"
#include "jemalloc/internal/tcache_externs.h"
#include "jemalloc/internal/ticker.h"
JEMALLOC_ALWAYS_INLINE malloc_cpuid_t
@ -14,6 +18,15 @@ malloc_getcpu(void) {
return GetCurrentProcessorNumber();
#elif defined(JEMALLOC_HAVE_SCHED_GETCPU)
return (malloc_cpuid_t)sched_getcpu();
#elif defined(JEMALLOC_HAVE_RDTSCP)
unsigned int ecx;
asm volatile("rdtscp" : "=c"(ecx)::"eax", "edx");
return (malloc_cpuid_t)(ecx & 0xfff);
#elif defined(__aarch64__) && defined(__APPLE__)
/* Other oses most likely use tpidr_el0 instead */
uintptr_t c;
asm volatile("mrs %x0, tpidrro_el0" : "=r"(c)::"memory");
return (malloc_cpuid_t)(c & (1 << 3) - 1);
#else
not_reached();
return -1;
@ -29,8 +42,8 @@ percpu_arena_choose(void) {
assert(cpuid >= 0);
unsigned arena_ind;
if ((opt_percpu_arena == percpu_arena) || ((unsigned)cpuid < ncpus /
2)) {
if ((opt_percpu_arena == percpu_arena)
|| ((unsigned)cpuid < ncpus / 2)) {
arena_ind = cpuid;
} else {
assert(opt_percpu_arena == per_phycpu_arena);

27
include/jemalloc/internal/jemalloc_internal_inlines_b.h
View file

@ -1,7 +1,10 @@
#ifndef JEMALLOC_INTERNAL_INLINES_B_H
#define JEMALLOC_INTERNAL_INLINES_B_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_inlines_a.h"
#include "jemalloc/internal/extent.h"
#include "jemalloc/internal/jemalloc_internal_inlines_a.h"
static inline void
percpu_arena_update(tsd_t *tsd, unsigned cpu) {
@ -20,13 +23,13 @@ percpu_arena_update(tsd_t *tsd, unsigned cpu) {
tcache_t *tcache = tcache_get(tsd);
if (tcache != NULL) {
tcache_slow_t *tcache_slow = tsd_tcache_slowp_get(tsd);
tcache_arena_reassociate(tsd_tsdn(tsd), tcache_slow,
tcache, newarena);
assert(tcache_slow->arena != NULL);
tcache_arena_reassociate(
tsd_tsdn(tsd), tcache_slow, tcache, newarena);
}
}
}
/* Choose an arena based on a per-thread value. */
static inline arena_t *
arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal) {
@ -47,18 +50,18 @@ arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal) {
assert(ret);
if (tcache_available(tsd)) {
tcache_slow_t *tcache_slow = tsd_tcache_slowp_get(tsd);
tcache_t *tcache = tsd_tcachep_get(tsd);
tcache_t *tcache = tsd_tcachep_get(tsd);
if (tcache_slow->arena != NULL) {
/* See comments in tsd_tcache_data_init().*/
assert(tcache_slow->arena ==
arena_get(tsd_tsdn(tsd), 0, false));
assert(tcache_slow->arena
== arena_get(tsd_tsdn(tsd), 0, false));
if (tcache_slow->arena != ret) {
tcache_arena_reassociate(tsd_tsdn(tsd),
tcache_slow, tcache, ret);
}
} else {
tcache_arena_associate(tsd_tsdn(tsd),
tcache_slow, tcache, ret);
tcache_arena_associate(
tsd_tsdn(tsd), tcache_slow, tcache, ret);
}
}
}
@ -68,10 +71,10 @@ arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal) {
* auto percpu arena range, (i.e. thread is assigned to a manually
* managed arena), then percpu arena is skipped.
*/
if (have_percpu_arena && PERCPU_ARENA_ENABLED(opt_percpu_arena) &&
!internal && (arena_ind_get(ret) <
percpu_arena_ind_limit(opt_percpu_arena)) && (ret->last_thd !=
tsd_tsdn(tsd))) {
if (have_percpu_arena && PERCPU_ARENA_ENABLED(opt_percpu_arena)
&& !internal
&& (arena_ind_get(ret) < percpu_arena_ind_limit(opt_percpu_arena))
&& (ret->last_thd != tsd_tsdn(tsd))) {
unsigned ind = percpu_arena_choose();
if (arena_ind_get(ret) != ind) {
percpu_arena_update(tsd, ind);

364
include/jemalloc/internal/jemalloc_internal_inlines_c.h
View file

@ -1,6 +1,10 @@
#ifndef JEMALLOC_INTERNAL_INLINES_C_H
#define JEMALLOC_INTERNAL_INLINES_C_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/arena_externs.h"
#include "jemalloc/internal/arena_inlines_b.h"
#include "jemalloc/internal/emap.h"
#include "jemalloc/internal/hook.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#include "jemalloc/internal/log.h"
@ -8,6 +12,15 @@
#include "jemalloc/internal/thread_event.h"
#include "jemalloc/internal/witness.h"
/*
* These correspond to the macros in jemalloc/jemalloc_macros.h. Broadly, we
* should have one constant here per magic value there. Note however that the
* representations need not be related.
*/
#define TCACHE_IND_NONE ((unsigned)-1)
#define TCACHE_IND_AUTOMATIC ((unsigned)-2)
#define ARENA_IND_AUTOMATIC ((unsigned)-1)
/*
* Translating the names of the 'i' functions:
* Abbreviations used in the first part of the function name (before
@ -41,24 +54,35 @@ isalloc(tsdn_t *tsdn, const void *ptr) {
}
JEMALLOC_ALWAYS_INLINE void *
iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero, tcache_t *tcache,
bool is_internal, arena_t *arena, bool slow_path) {
iallocztm_explicit_slab(tsdn_t *tsdn, size_t size, szind_t ind, bool zero,
bool slab, tcache_t *tcache, bool is_internal, arena_t *arena,
bool slow_path) {
void *ret;
assert(!slab || sz_can_use_slab(size)); /* slab && large is illegal */
assert(!is_internal || tcache == NULL);
assert(!is_internal || arena == NULL || arena_is_auto(arena));
if (!tsdn_null(tsdn) && tsd_reentrancy_level_get(tsdn_tsd(tsdn)) == 0) {
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
}
ret = arena_malloc(tsdn, arena, size, ind, zero, tcache, slow_path);
ret = arena_malloc(
tsdn, arena, size, ind, zero, slab, tcache, slow_path);
if (config_stats && is_internal && likely(ret != NULL)) {
arena_internal_add(iaalloc(tsdn, ret), isalloc(tsdn, ret));
}
return ret;
}
JEMALLOC_ALWAYS_INLINE void *
iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero, tcache_t *tcache,
bool is_internal, arena_t *arena, bool slow_path) {
bool slab = sz_can_use_slab(size);
return iallocztm_explicit_slab(
tsdn, size, ind, zero, slab, tcache, is_internal, arena, slow_path);
}
JEMALLOC_ALWAYS_INLINE void *
ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero, bool slow_path) {
return iallocztm(tsd_tsdn(tsd), size, ind, zero, tcache_get(tsd), false,
@ -66,18 +90,19 @@ ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero, bool slow_path) {
}
JEMALLOC_ALWAYS_INLINE void *
ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
tcache_t *tcache, bool is_internal, arena_t *arena) {
ipallocztm_explicit_slab(tsdn_t *tsdn, size_t usize, size_t alignment,
bool zero, bool slab, tcache_t *tcache, bool is_internal, arena_t *arena) {
void *ret;
assert(!slab || sz_can_use_slab(usize)); /* slab && large is illegal */
assert(usize != 0);
assert(usize == sz_sa2u(usize, alignment));
assert(!is_internal || tcache == NULL);
assert(!is_internal || arena == NULL || arena_is_auto(arena));
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
ret = arena_palloc(tsdn, arena, usize, alignment, zero, tcache);
ret = arena_palloc(tsdn, arena, usize, alignment, zero, slab, tcache);
assert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret);
if (config_stats && is_internal && likely(ret != NULL)) {
arena_internal_add(iaalloc(tsdn, ret), isalloc(tsdn, ret));
@ -85,12 +110,26 @@ ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
return ret;
}
JEMALLOC_ALWAYS_INLINE void *
ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
tcache_t *tcache, bool is_internal, arena_t *arena) {
return ipallocztm_explicit_slab(tsdn, usize, alignment, zero,
sz_can_use_slab(usize), tcache, is_internal, arena);
}
JEMALLOC_ALWAYS_INLINE void *
ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
tcache_t *tcache, arena_t *arena) {
return ipallocztm(tsdn, usize, alignment, zero, tcache, false, arena);
}
JEMALLOC_ALWAYS_INLINE void *
ipalloct_explicit_slab(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
bool slab, tcache_t *tcache, arena_t *arena) {
return ipallocztm_explicit_slab(
tsdn, usize, alignment, zero, slab, tcache, false, arena);
}
JEMALLOC_ALWAYS_INLINE void *
ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero) {
return ipallocztm(tsd_tsdn(tsd), usize, alignment, zero,
@ -108,13 +147,13 @@ idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache,
assert(ptr != NULL);
assert(!is_internal || tcache == NULL);
assert(!is_internal || arena_is_auto(iaalloc(tsdn, ptr)));
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
if (config_stats && is_internal) {
arena_internal_sub(iaalloc(tsdn, ptr), isalloc(tsdn, ptr));
}
if (!is_internal && !tsdn_null(tsdn) &&
tsd_reentrancy_level_get(tsdn_tsd(tsdn)) != 0) {
if (!is_internal && !tsdn_null(tsdn)
&& tsd_reentrancy_level_get(tsdn_tsd(tsdn)) != 0) {
assert(tcache == NULL);
}
arena_dalloc(tsdn, ptr, tcache, alloc_ctx, slow_path);
@ -128,25 +167,26 @@ idalloc(tsd_t *tsd, void *ptr) {
JEMALLOC_ALWAYS_INLINE void
isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
emap_alloc_ctx_t *alloc_ctx, bool slow_path) {
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
arena_sdalloc(tsdn, ptr, size, tcache, alloc_ctx, slow_path);
}
JEMALLOC_ALWAYS_INLINE void *
iralloct_realign(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
size_t alignment, bool zero, tcache_t *tcache, arena_t *arena,
size_t alignment, bool zero, bool slab, tcache_t *tcache, arena_t *arena,
hook_ralloc_args_t *hook_args) {
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
void *p;
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
void *p;
size_t usize, copysize;
usize = sz_sa2u(size, alignment);
if (unlikely(usize == 0 || usize > SC_LARGE_MAXCLASS)) {
return NULL;
}
p = ipalloct(tsdn, usize, alignment, zero, tcache, arena);
p = ipalloct_explicit_slab(
tsdn, usize, alignment, zero, slab, tcache, arena);
if (p == NULL) {
return NULL;
}
@ -156,11 +196,12 @@ iralloct_realign(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
*/
copysize = (size < oldsize) ? size : oldsize;
memcpy(p, ptr, copysize);
hook_invoke_alloc(hook_args->is_realloc
? hook_alloc_realloc : hook_alloc_rallocx, p, (uintptr_t)p,
hook_args->args);
hook_invoke_dalloc(hook_args->is_realloc
? hook_dalloc_realloc : hook_dalloc_rallocx, ptr, hook_args->args);
hook_invoke_alloc(
hook_args->is_realloc ? hook_alloc_realloc : hook_alloc_rallocx, p,
(uintptr_t)p, hook_args->args);
hook_invoke_dalloc(
hook_args->is_realloc ? hook_dalloc_realloc : hook_dalloc_rallocx,
ptr, hook_args->args);
isdalloct(tsdn, ptr, oldsize, tcache, NULL, true);
return p;
}
@ -173,33 +214,42 @@ iralloct_realign(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
* passed-around anywhere.
*/
JEMALLOC_ALWAYS_INLINE void *
iralloct(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t alignment,
bool zero, tcache_t *tcache, arena_t *arena, hook_ralloc_args_t *hook_args)
{
iralloct_explicit_slab(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
size_t alignment, bool zero, bool slab, tcache_t *tcache, arena_t *arena,
hook_ralloc_args_t *hook_args) {
assert(ptr != NULL);
assert(size != 0);
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))
!= 0) {
if (alignment != 0
&& ((uintptr_t)ptr & ((uintptr_t)alignment - 1)) != 0) {
/*
* Existing object alignment is inadequate; allocate new space
* and copy.
*/
return iralloct_realign(tsdn, ptr, oldsize, size, alignment,
zero, tcache, arena, hook_args);
zero, slab, tcache, arena, hook_args);
}
return arena_ralloc(tsdn, arena, ptr, oldsize, size, alignment, zero,
tcache, hook_args);
slab, tcache, hook_args);
}
JEMALLOC_ALWAYS_INLINE void *
iralloct(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t alignment,
size_t usize, bool zero, tcache_t *tcache, arena_t *arena,
hook_ralloc_args_t *hook_args) {
bool slab = sz_can_use_slab(usize);
return iralloct_explicit_slab(tsdn, ptr, oldsize, size, alignment, zero,
slab, tcache, arena, hook_args);
}
JEMALLOC_ALWAYS_INLINE void *
iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
bool zero, hook_ralloc_args_t *hook_args) {
return iralloct(tsd_tsdn(tsd), ptr, oldsize, size, alignment, zero,
tcache_get(tsd), NULL, hook_args);
size_t usize, bool zero, hook_ralloc_args_t *hook_args) {
return iralloct(tsd_tsdn(tsd), ptr, oldsize, size, alignment, usize,
zero, tcache_get(tsd), NULL, hook_args);
}
JEMALLOC_ALWAYS_INLINE bool
@ -207,29 +257,27 @@ ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t extra,
size_t alignment, bool zero, size_t *newsize) {
assert(ptr != NULL);
assert(size != 0);
witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
WITNESS_RANK_CORE, 0);
witness_assert_depth_to_rank(
tsdn_witness_tsdp_get(tsdn), WITNESS_RANK_CORE, 0);
if (alignment != 0 && ((uintptr_t)ptr & ((uintptr_t)alignment-1))
!= 0) {
if (alignment != 0
&& ((uintptr_t)ptr & ((uintptr_t)alignment - 1)) != 0) {
/* Existing object alignment is inadequate. */
*newsize = oldsize;
return true;
}
return arena_ralloc_no_move(tsdn, ptr, oldsize, size, extra, zero,
newsize);
return arena_ralloc_no_move(
tsdn, ptr, oldsize, size, extra, zero, newsize);
}
JEMALLOC_ALWAYS_INLINE void
fastpath_success_finish(tsd_t *tsd, uint64_t allocated_after,
cache_bin_t *bin, void *ret) {
fastpath_success_finish(
tsd_t *tsd, uint64_t allocated_after, cache_bin_t *bin, void *ret) {
thread_allocated_set(tsd, allocated_after);
if (config_stats) {
bin->tstats.nrequests++;
}
LOG("core.malloc.exit", "result: %p", ret);
}
JEMALLOC_ALWAYS_INLINE bool
@ -256,7 +304,6 @@ malloc_initialized(void) {
*/
JEMALLOC_ALWAYS_INLINE void *
imalloc_fastpath(size_t size, void *(fallback_alloc)(size_t)) {
LOG("core.malloc.entry", "size: %zu", size);
if (tsd_get_allocates() && unlikely(!malloc_initialized())) {
return fallback_alloc(size);
}
@ -284,8 +331,8 @@ imalloc_fastpath(size_t size, void *(fallback_alloc)(size_t)) {
sz_size2index_usize_fastpath(size, &ind, &usize);
/* Fast path relies on size being a bin. */
assert(ind < SC_NBINS);
assert((SC_LOOKUP_MAXCLASS < SC_SMALL_MAXCLASS) &&
(size <= SC_SMALL_MAXCLASS));
assert((SC_LOOKUP_MAXCLASS < SC_SMALL_MAXCLASS)
&& (size <= SC_SMALL_MAXCLASS));
uint64_t allocated, threshold;
te_malloc_fastpath_ctx(tsd, &allocated, &threshold);
@ -314,7 +361,9 @@ imalloc_fastpath(size_t size, void *(fallback_alloc)(size_t)) {
tcache_t *tcache = tsd_tcachep_get(tsd);
assert(tcache == tcache_get(tsd));
cache_bin_t *bin = &tcache->bins[ind];
bool tcache_success;
/* Suppress spurious warning from static analysis */
assert(bin != NULL);
bool tcache_success;
void *ret;
/*
@ -337,4 +386,215 @@ imalloc_fastpath(size_t size, void *(fallback_alloc)(size_t)) {
return fallback_alloc(size);
}
JEMALLOC_ALWAYS_INLINE tcache_t *
tcache_get_from_ind(tsd_t *tsd, unsigned tcache_ind, bool slow, bool is_alloc) {
tcache_t *tcache;
if (tcache_ind == TCACHE_IND_AUTOMATIC) {
if (likely(!slow)) {
/* Getting tcache ptr unconditionally. */
tcache = tsd_tcachep_get(tsd);
assert(tcache == tcache_get(tsd));
} else if (is_alloc
|| likely(tsd_reentrancy_level_get(tsd) == 0)) {
tcache = tcache_get(tsd);
} else {
tcache = NULL;
}
} else {
/*
* Should not specify tcache on deallocation path when being
* reentrant.
*/
assert(is_alloc || tsd_reentrancy_level_get(tsd) == 0
|| tsd_state_nocleanup(tsd));
if (tcache_ind == TCACHE_IND_NONE) {
tcache = NULL;
} else {
tcache = tcaches_get(tsd, tcache_ind);
}
}
return tcache;
}
JEMALLOC_ALWAYS_INLINE bool
maybe_check_alloc_ctx(tsd_t *tsd, void *ptr, emap_alloc_ctx_t *alloc_ctx) {
if (config_opt_size_checks) {
emap_alloc_ctx_t dbg_ctx;
emap_alloc_ctx_lookup(
tsd_tsdn(tsd), &arena_emap_global, ptr, &dbg_ctx);
if (alloc_ctx->szind != dbg_ctx.szind) {
safety_check_fail_sized_dealloc(
/* current_dealloc */ true, ptr,
/* true_size */ emap_alloc_ctx_usize_get(&dbg_ctx),
/* input_size */
emap_alloc_ctx_usize_get(alloc_ctx));
return true;
}
if (alloc_ctx->slab != dbg_ctx.slab) {
safety_check_fail(
"Internal heap corruption detected: "
"mismatch in slab bit");
return true;
}
}
return false;
}
JEMALLOC_ALWAYS_INLINE bool
prof_sample_aligned(const void *ptr) {
return ((uintptr_t)ptr & PROF_SAMPLE_ALIGNMENT_MASK) == 0;
}
JEMALLOC_ALWAYS_INLINE bool
free_fastpath_nonfast_aligned(void *ptr, bool check_prof) {
/*
* free_fastpath do not handle two uncommon cases: 1) sampled profiled
* objects and 2) sampled junk & stash for use-after-free detection.
* Both have special alignments which are used to escape the fastpath.
*
* prof_sample is page-aligned, which covers the UAF check when both
* are enabled (the assertion below). Avoiding redundant checks since
* this is on the fastpath -- at most one runtime branch from this.
*/
if (config_debug && cache_bin_nonfast_aligned(ptr)) {
assert(prof_sample_aligned(ptr));
}
if (config_prof && check_prof) {
/* When prof is enabled, the prof_sample alignment is enough. */
if (prof_sample_aligned(ptr)) {
return true;
} else {
return false;
}
}
if (config_uaf_detection) {
if (cache_bin_nonfast_aligned(ptr)) {
return true;
} else {
return false;
}
}
return false;
}
/* Returns whether or not the free attempt was successful. */
JEMALLOC_ALWAYS_INLINE
bool
free_fastpath(void *ptr, size_t size, bool size_hint) {
tsd_t *tsd = tsd_get(false);
/* The branch gets optimized away unless tsd_get_allocates(). */
if (unlikely(tsd == NULL)) {
return false;
}
/*
* The tsd_fast() / initialized checks are folded into the branch
* testing (deallocated_after >= threshold) later in this function.
* The threshold will be set to 0 when !tsd_fast.
*/
assert(tsd_fast(tsd)
|| *tsd_thread_deallocated_next_event_fastp_get_unsafe(tsd) == 0);
emap_alloc_ctx_t alloc_ctx JEMALLOC_CC_SILENCE_INIT({0, 0, false});
size_t usize;
if (!size_hint) {
bool err = emap_alloc_ctx_try_lookup_fast(
tsd, &arena_emap_global, ptr, &alloc_ctx);
/* Note: profiled objects will have alloc_ctx.slab set */
if (unlikely(err || !alloc_ctx.slab
|| free_fastpath_nonfast_aligned(ptr,
/* check_prof */ false))) {
return false;
}
assert(alloc_ctx.szind != SC_NSIZES);
usize = sz_index2size(alloc_ctx.szind);
} else {
/*
* Check for both sizes that are too large, and for sampled /
* special aligned objects. The alignment check will also check
* for null ptr.
*/
if (unlikely(size > SC_LOOKUP_MAXCLASS
|| free_fastpath_nonfast_aligned(ptr,
/* check_prof */ true))) {
return false;
}
sz_size2index_usize_fastpath(size, &alloc_ctx.szind, &usize);
/* Max lookup class must be small. */
assert(alloc_ctx.szind < SC_NBINS);
/* This is a dead store, except when opt size checking is on. */
alloc_ctx.slab = true;
}
/*
* Currently the fastpath only handles small sizes. The branch on
* SC_LOOKUP_MAXCLASS makes sure of it. This lets us avoid checking
* tcache szind upper limit (i.e. tcache_max) as well.
*/
assert(alloc_ctx.slab);
uint64_t deallocated, threshold;
te_free_fastpath_ctx(tsd, &deallocated, &threshold);
uint64_t deallocated_after = deallocated + usize;
/*
* Check for events and tsd non-nominal (fast_threshold will be set to
* 0) in a single branch. Note that this handles the uninitialized case
* as well (TSD init will be triggered on the non-fastpath). Therefore
* anything depends on a functional TSD (e.g. the alloc_ctx sanity check
* below) needs to be after this branch.
*/
if (unlikely(deallocated_after >= threshold)) {
return false;
}
assert(tsd_fast(tsd));
bool fail = maybe_check_alloc_ctx(tsd, ptr, &alloc_ctx);
if (fail) {
/* See the comment in isfree. */
return true;
}
tcache_t *tcache = tcache_get_from_ind(tsd, TCACHE_IND_AUTOMATIC,
/* slow */ false, /* is_alloc */ false);
cache_bin_t *bin = &tcache->bins[alloc_ctx.szind];
/*
* If junking were enabled, this is where we would do it. It's not
* though, since we ensured above that we're on the fast path. Assert
* that to double-check.
*/
assert(!opt_junk_free);
if (!cache_bin_dalloc_easy(bin, ptr)) {
return false;
}
*tsd_thread_deallocatedp_get(tsd) = deallocated_after;
return true;
}
JEMALLOC_ALWAYS_INLINE void JEMALLOC_NOTHROW
je_sdallocx_noflags(void *ptr, size_t size) {
if (!free_fastpath(ptr, size, true)) {
sdallocx_default(ptr, size, 0);
}
}
JEMALLOC_ALWAYS_INLINE void JEMALLOC_NOTHROW
je_sdallocx_impl(void *ptr, size_t size, int flags) {
if (flags != 0 || !free_fastpath(ptr, size, true)) {
sdallocx_default(ptr, size, flags);
}
}
JEMALLOC_ALWAYS_INLINE void JEMALLOC_NOTHROW
je_free_impl(void *ptr) {
if (!free_fastpath(ptr, 0, false)) {
free_default(ptr);
}
}
#endif /* JEMALLOC_INTERNAL_INLINES_C_H */

155
include/jemalloc/internal/jemalloc_internal_macros.h
View file

@ -2,43 +2,46 @@
#define JEMALLOC_INTERNAL_MACROS_H
#ifdef JEMALLOC_DEBUG
# define JEMALLOC_ALWAYS_INLINE static inline
# define JEMALLOC_ALWAYS_INLINE static inline
#else
# ifdef _MSC_VER
# define JEMALLOC_ALWAYS_INLINE static __forceinline
# else
# define JEMALLOC_ALWAYS_INLINE JEMALLOC_ATTR(always_inline) static inline
# endif
# ifdef _MSC_VER
# define JEMALLOC_ALWAYS_INLINE static __forceinline
# else
# define JEMALLOC_ALWAYS_INLINE \
JEMALLOC_ATTR(always_inline) static inline
# endif
#endif
#ifdef _MSC_VER
# define inline _inline
# define inline _inline
#endif
#define UNUSED JEMALLOC_ATTR(unused)
#define ZU(z) ((size_t)z)
#define ZD(z) ((ssize_t)z)
#define QU(q) ((uint64_t)q)
#define QD(q) ((int64_t)q)
#define ZU(z) ((size_t)z)
#define ZD(z) ((ssize_t)z)
#define QU(q) ((uint64_t)q)
#define QD(q) ((int64_t)q)
#define KZU(z) ZU(z##ULL)
#define KZD(z) ZD(z##LL)
#define KQU(q) QU(q##ULL)
#define KQD(q) QI(q##LL)
#define KZU(z) ZU(z##ULL)
#define KZD(z) ZD(z##LL)
#define KQU(q) QU(q##ULL)
#define KQD(q) QI(q##LL)
#ifndef __DECONST
# define __DECONST(type, var) ((type)(uintptr_t)(const void *)(var))
# define __DECONST(type, var) ((type)(uintptr_t)(const void *)(var))
#endif
#if !defined(JEMALLOC_HAS_RESTRICT) || defined(__cplusplus)
# define restrict
# define restrict
#endif
/* Various function pointers are static and immutable except during testing. */
#ifdef JEMALLOC_JET
# define JET_MUTABLE
# define JET_MUTABLE
# define JET_EXTERN extern
#else
# define JET_MUTABLE const
# define JET_MUTABLE const
# define JET_EXTERN static
#endif
#define JEMALLOC_VA_ARGS_HEAD(head, ...) head
@ -46,62 +49,94 @@
/* Diagnostic suppression macros */
#if defined(_MSC_VER) && !defined(__clang__)
# define JEMALLOC_DIAGNOSTIC_PUSH __pragma(warning(push))
# define JEMALLOC_DIAGNOSTIC_POP __pragma(warning(pop))
# define JEMALLOC_DIAGNOSTIC_IGNORE(W) __pragma(warning(disable:W))
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
# define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
# define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
# define JEMALLOC_DIAGNOSTIC_PUSH __pragma(warning(push))
# define JEMALLOC_DIAGNOSTIC_POP __pragma(warning(pop))
# define JEMALLOC_DIAGNOSTIC_IGNORE(W) __pragma(warning(disable : W))
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
# define JEMALLOC_DIAGNOSTIC_IGNORE_FRAME_ADDRESS
# define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
# define JEMALLOC_DIAGNOSTIC_IGNORE_DEPRECATED
# define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
/* #pragma GCC diagnostic first appeared in gcc 4.6. */
#elif (defined(__GNUC__) && ((__GNUC__ > 4) || ((__GNUC__ == 4) && \
(__GNUC_MINOR__ > 5)))) || defined(__clang__)
#elif (defined(__GNUC__) \
&& ((__GNUC__ > 4) || ((__GNUC__ == 4) && (__GNUC_MINOR__ > 5)))) \
|| defined(__clang__)
/*
* The JEMALLOC_PRAGMA__ macro is an implementation detail of the GCC and Clang
* diagnostic suppression macros and should not be used anywhere else.
*/
# define JEMALLOC_PRAGMA__(X) _Pragma(#X)
# define JEMALLOC_DIAGNOSTIC_PUSH JEMALLOC_PRAGMA__(GCC diagnostic push)
# define JEMALLOC_DIAGNOSTIC_POP JEMALLOC_PRAGMA__(GCC diagnostic pop)
# define JEMALLOC_DIAGNOSTIC_IGNORE(W) \
JEMALLOC_PRAGMA__(GCC diagnostic ignored W)
# define JEMALLOC_PRAGMA__(X) _Pragma(#X)
# define JEMALLOC_DIAGNOSTIC_PUSH JEMALLOC_PRAGMA__(GCC diagnostic push)
# define JEMALLOC_DIAGNOSTIC_POP JEMALLOC_PRAGMA__(GCC diagnostic pop)
# define JEMALLOC_DIAGNOSTIC_IGNORE(W) \
JEMALLOC_PRAGMA__(GCC diagnostic ignored W)
/*
* The -Wmissing-field-initializers warning is buggy in GCC versions < 5.1 and
* all clang versions up to version 7 (currently trunk, unreleased). This macro
* suppresses the warning for the affected compiler versions only.
*/
# if ((defined(__GNUC__) && !defined(__clang__)) && (__GNUC__ < 5)) || \
defined(__clang__)
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wmissing-field-initializers")
# else
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
# endif
# if ((defined(__GNUC__) && !defined(__clang__)) && (__GNUC__ < 5)) \
|| defined(__clang__)
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS \
JEMALLOC_DIAGNOSTIC_IGNORE( \
"-Wmissing-field-initializers")
# else
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
# endif
# define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wtype-limits")
# define JEMALLOC_DIAGNOSTIC_IGNORE_UNUSED_PARAMETER \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wunused-parameter")
# if defined(__GNUC__) && !defined(__clang__) && (__GNUC__ >= 7)
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN \
JEMALLOC_DIAGNOSTIC_IGNORE("-Walloc-size-larger-than=")
# else
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
# endif
# define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS \
JEMALLOC_DIAGNOSTIC_PUSH \
JEMALLOC_DIAGNOSTIC_IGNORE_UNUSED_PARAMETER
# define JEMALLOC_DIAGNOSTIC_IGNORE_FRAME_ADDRESS \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wframe-address")
# define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wtype-limits")
# define JEMALLOC_DIAGNOSTIC_IGNORE_UNUSED_PARAMETER \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wunused-parameter")
# if defined(__GNUC__) && !defined(__clang__) && (__GNUC__ >= 7)
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN \
JEMALLOC_DIAGNOSTIC_IGNORE("-Walloc-size-larger-than=")
# else
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
# endif
# ifdef JEMALLOC_HAVE_ATTR_DEPRECATED
# define JEMALLOC_DIAGNOSTIC_IGNORE_DEPRECATED \
JEMALLOC_DIAGNOSTIC_IGNORE("-Wdeprecated-declarations")
# else
# define JEMALLOC_DIAGNOSTIC_IGNORE_DEPRECATED
# endif
# define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS \
JEMALLOC_DIAGNOSTIC_PUSH \
JEMALLOC_DIAGNOSTIC_IGNORE_UNUSED_PARAMETER
#else
# define JEMALLOC_DIAGNOSTIC_PUSH
# define JEMALLOC_DIAGNOSTIC_POP
# define JEMALLOC_DIAGNOSTIC_IGNORE(W)
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
# define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
# define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
# define JEMALLOC_DIAGNOSTIC_PUSH
# define JEMALLOC_DIAGNOSTIC_POP
# define JEMALLOC_DIAGNOSTIC_IGNORE(W)
# define JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
# define JEMALLOC_DIAGNOSTIC_IGNORE_FRAME_ADDRESS
# define JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
# define JEMALLOC_DIAGNOSTIC_IGNORE_ALLOC_SIZE_LARGER_THAN
# define JEMALLOC_DIAGNOSTIC_IGNORE_DEPRECATED
# define JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
#endif
#ifdef __clang_analyzer__
# define JEMALLOC_CLANG_ANALYZER
#endif
#ifdef JEMALLOC_CLANG_ANALYZER
# define JEMALLOC_CLANG_ANALYZER_SUPPRESS __attribute__((suppress))
# define JEMALLOC_CLANG_ANALYZER_SILENCE_INIT(v) = v
#else
# define JEMALLOC_CLANG_ANALYZER_SUPPRESS
# define JEMALLOC_CLANG_ANALYZER_SILENCE_INIT(v)
#endif
#define JEMALLOC_SUPPRESS_WARN_ON_USAGE(...) \
JEMALLOC_DIAGNOSTIC_PUSH \
JEMALLOC_DIAGNOSTIC_IGNORE_DEPRECATED \
__VA_ARGS__ \
JEMALLOC_DIAGNOSTIC_POP
/*
* Disables spurious diagnostics for all headers. Since these headers are not
* included by users directly, it does not affect their diagnostic settings.

22
include/jemalloc/internal/jemalloc_internal_overrides.h Normal file
View file

@ -0,0 +1,22 @@
#ifndef JEMALLOC_INTERNAL_OVERRIDES_H
#define JEMALLOC_INTERNAL_OVERRIDES_H
/*
* Under normal circumstances this header serves no purpose, as these settings
* can be customized via the corresponding autoconf options at configure-time.
* Overriding in this fashion is useful when the header files generated by
* autoconf are used as input for another build system.
*/
#ifdef JEMALLOC_OVERRIDE_LG_PAGE
# undef LG_PAGE
# define LG_PAGE JEMALLOC_OVERRIDE_LG_PAGE
#endif
#ifdef JEMALLOC_OVERRIDE_JEMALLOC_CONFIG_MALLOC_CONF
# undef JEMALLOC_CONFIG_MALLOC_CONF
# define JEMALLOC_CONFIG_MALLOC_CONF \
JEMALLOC_OVERRIDE_JEMALLOC_CONFIG_MALLOC_CONF
#endif
#endif /* JEMALLOC_INTERNAL_OVERRIDES_H */

136
include/jemalloc/internal/jemalloc_internal_types.h
View file

@ -18,13 +18,13 @@ enum zero_realloc_action_e {
typedef enum zero_realloc_action_e zero_realloc_action_t;
/* Signature of write callback. */
typedef void (write_cb_t)(void *, const char *);
typedef void(write_cb_t)(void *, const char *);
enum malloc_init_e {
malloc_init_uninitialized = 3,
malloc_init_a0_initialized = 2,
malloc_init_recursible = 1,
malloc_init_initialized = 0 /* Common case --> jnz. */
malloc_init_uninitialized = 3,
malloc_init_a0_initialized = 2,
malloc_init_recursible = 1,
malloc_init_initialized = 0 /* Common case --> jnz. */
};
typedef enum malloc_init_e malloc_init_t;
@ -39,48 +39,46 @@ typedef enum malloc_init_e malloc_init_t;
*
* aaaaaaaa aaaatttt tttttttt 0znnnnnn
*/
#define MALLOCX_ARENA_BITS 12
#define MALLOCX_TCACHE_BITS 12
#define MALLOCX_LG_ALIGN_BITS 6
#define MALLOCX_ARENA_SHIFT 20
#define MALLOCX_TCACHE_SHIFT 8
#define MALLOCX_ARENA_MASK \
(((1 << MALLOCX_ARENA_BITS) - 1) << MALLOCX_ARENA_SHIFT)
#define MALLOCX_ARENA_BITS 12
#define MALLOCX_TCACHE_BITS 12
#define MALLOCX_LG_ALIGN_BITS 6
#define MALLOCX_ARENA_SHIFT 20
#define MALLOCX_TCACHE_SHIFT 8
#define MALLOCX_ARENA_MASK \
((unsigned)(((1U << MALLOCX_ARENA_BITS) - 1) << MALLOCX_ARENA_SHIFT))
/* NB: Arena index bias decreases the maximum number of arenas by 1. */
#define MALLOCX_ARENA_LIMIT ((1 << MALLOCX_ARENA_BITS) - 1)
#define MALLOCX_TCACHE_MASK \
(((1 << MALLOCX_TCACHE_BITS) - 1) << MALLOCX_TCACHE_SHIFT)
#define MALLOCX_TCACHE_MAX ((1 << MALLOCX_TCACHE_BITS) - 3)
#define MALLOCX_LG_ALIGN_MASK ((1 << MALLOCX_LG_ALIGN_BITS) - 1)
#define MALLOCX_ARENA_LIMIT ((unsigned)((1U << MALLOCX_ARENA_BITS) - 1))
#define MALLOCX_TCACHE_MASK \
((unsigned)(((1U << MALLOCX_TCACHE_BITS) - 1) << MALLOCX_TCACHE_SHIFT))
#define MALLOCX_TCACHE_MAX ((unsigned)((1U << MALLOCX_TCACHE_BITS) - 3))
#define MALLOCX_LG_ALIGN_MASK ((1 << MALLOCX_LG_ALIGN_BITS) - 1)
/* Use MALLOCX_ALIGN_GET() if alignment may not be specified in flags. */
#define MALLOCX_ALIGN_GET_SPECIFIED(flags) \
(ZU(1) << (flags & MALLOCX_LG_ALIGN_MASK))
#define MALLOCX_ALIGN_GET(flags) \
(MALLOCX_ALIGN_GET_SPECIFIED(flags) & (SIZE_T_MAX-1))
#define MALLOCX_ZERO_GET(flags) \
((bool)(flags & MALLOCX_ZERO))
#define MALLOCX_ALIGN_GET_SPECIFIED(flags) \
(ZU(1) << (flags & MALLOCX_LG_ALIGN_MASK))
#define MALLOCX_ALIGN_GET(flags) \
(MALLOCX_ALIGN_GET_SPECIFIED(flags) & (SIZE_T_MAX - 1))
#define MALLOCX_ZERO_GET(flags) ((bool)(flags & MALLOCX_ZERO))
#define MALLOCX_TCACHE_GET(flags) \
(((unsigned)((flags & MALLOCX_TCACHE_MASK) >> MALLOCX_TCACHE_SHIFT)) - 2)
#define MALLOCX_ARENA_GET(flags) \
(((unsigned)(((unsigned)flags) >> MALLOCX_ARENA_SHIFT)) - 1)
#define MALLOCX_TCACHE_GET(flags) \
(((unsigned)((flags & MALLOCX_TCACHE_MASK) >> MALLOCX_TCACHE_SHIFT)) \
- 2)
#define MALLOCX_ARENA_GET(flags) \
(((unsigned)(((unsigned)flags) >> MALLOCX_ARENA_SHIFT)) - 1)
/* Smallest size class to support. */
#define TINY_MIN (1U << LG_TINY_MIN)
#define TINY_MIN (1U << LG_TINY_MIN)
#define LONG ((size_t)(1U << LG_SIZEOF_LONG))
#define LONG_MASK (LONG - 1)
#define LONG ((size_t)(1U << LG_SIZEOF_LONG))
#define LONG_MASK (LONG - 1)
/* Return the smallest long multiple that is >= a. */
#define LONG_CEILING(a) \
(((a) + LONG_MASK) & ~LONG_MASK)
#define LONG_CEILING(a) (((a) + LONG_MASK) & ~LONG_MASK)
#define SIZEOF_PTR (1U << LG_SIZEOF_PTR)
#define PTR_MASK (SIZEOF_PTR - 1)
#define SIZEOF_PTR (1U << LG_SIZEOF_PTR)
#define PTR_MASK (SIZEOF_PTR - 1)
/* Return the smallest (void *) multiple that is >= a. */
#define PTR_CEILING(a) \
(((a) + PTR_MASK) & ~PTR_MASK)
#define PTR_CEILING(a) (((a) + PTR_MASK) & ~PTR_MASK)
/*
* Maximum size of L1 cache line. This is used to avoid cache line aliasing.
@ -89,42 +87,62 @@ typedef enum malloc_init_e malloc_init_t;
* CACHELINE cannot be based on LG_CACHELINE because __declspec(align()) can
* only handle raw constants.
*/
#define LG_CACHELINE 6
#define CACHELINE 64
#define CACHELINE_MASK (CACHELINE - 1)
#define LG_CACHELINE 6
#define CACHELINE 64
#define CACHELINE_MASK (CACHELINE - 1)
/* Return the smallest cacheline multiple that is >= s. */
#define CACHELINE_CEILING(s) \
(((s) + CACHELINE_MASK) & ~CACHELINE_MASK)
#define CACHELINE_CEILING(s) (((s) + CACHELINE_MASK) & ~CACHELINE_MASK)
/* Return the nearest aligned address at or below a. */
#define ALIGNMENT_ADDR2BASE(a, alignment) \
((void *)((uintptr_t)(a) & ((~(alignment)) + 1)))
#define ALIGNMENT_ADDR2BASE(a, alignment) \
((void *)(((byte_t *)(a)) \
- (((uintptr_t)(a)) - ((uintptr_t)(a) & ((~(alignment)) + 1)))))
/* Return the offset between a and the nearest aligned address at or below a. */
#define ALIGNMENT_ADDR2OFFSET(a, alignment) \
#define ALIGNMENT_ADDR2OFFSET(a, alignment) \
((size_t)((uintptr_t)(a) & (alignment - 1)))
/* Return the smallest alignment multiple that is >= s. */
#define ALIGNMENT_CEILING(s, alignment) \
#define ALIGNMENT_CEILING(s, alignment) \
(((s) + (alignment - 1)) & ((~(alignment)) + 1))
/*
* Return the nearest aligned address at or above a.
*
* While at first glance this would appear to be merely a more complicated
* way to perform the same computation as `ALIGNMENT_CEILING`,
* this has the important additional property of not concealing pointer
* provenance from the compiler. See the block-comment on the
* definition of `byte_t` for more details.
*/
#define ALIGNMENT_ADDR2CEILING(a, alignment) \
((void *)(((byte_t *)(a)) \
+ (((((uintptr_t)(a)) + (alignment - 1)) & ((~(alignment)) + 1)) \
- ((uintptr_t)(a)))))
/* Declare a variable-length array. */
#if __STDC_VERSION__ < 199901L
# ifdef _MSC_VER
# include <malloc.h>
# define alloca _alloca
# else
# ifdef JEMALLOC_HAS_ALLOCA_H
# include <alloca.h>
# else
# include <stdlib.h>
# endif
# endif
# define VARIABLE_ARRAY(type, name, count) \
type *name = alloca(sizeof(type) * (count))
#if __STDC_VERSION__ < 199901L || defined(__STDC_NO_VLA__)
# ifdef _MSC_VER
# include <malloc.h>
# define alloca _alloca
# else
# ifdef JEMALLOC_HAS_ALLOCA_H
# include <alloca.h>
# else
# include <stdlib.h>
# endif
# endif
# define VARIABLE_ARRAY_UNSAFE(type, name, count) \
type *name = alloca(sizeof(type) * (count))
#else
# define VARIABLE_ARRAY(type, name, count) type name[(count)]
# define VARIABLE_ARRAY_UNSAFE(type, name, count) type name[(count)]
#endif
#define VARIABLE_ARRAY_SIZE_MAX 2048
#define VARIABLE_ARRAY(type, name, count) \
assert(sizeof(type) * (count) <= VARIABLE_ARRAY_SIZE_MAX); \
VARIABLE_ARRAY_UNSAFE(type, name, count)
#define CALLOC_MADVISE_THRESHOLD_DEFAULT (((size_t)1) << 23) /* 8 MB */
#endif /* JEMALLOC_INTERNAL_TYPES_H */

27
include/jemalloc/internal/jemalloc_preamble.h.in
View file

@ -1,7 +1,7 @@
#ifndef JEMALLOC_PREAMBLE_H
#define JEMALLOC_PREAMBLE_H
#include "jemalloc_internal_defs.h"
#include "jemalloc/internal/jemalloc_internal_defs.h"
#include "jemalloc/internal/jemalloc_internal_decls.h"
#if defined(JEMALLOC_UTRACE) || defined(JEMALLOC_UTRACE_LABEL)
@ -57,6 +57,15 @@
# define JEMALLOC_MADV_FREE 8
#endif
/*
* Can be defined at compile time, in cases, when it is known
* madvise(..., MADV_COLLAPSE) feature is supported, but MADV_COLLAPSE
* constant is not defined.
*/
#ifdef JEMALLOC_DEFINE_MADVISE_COLLAPSE
# define JEMALLOC_MADV_COLLAPSE 25
#endif
static const bool config_debug =
#ifdef JEMALLOC_DEBUG
true
@ -78,6 +87,13 @@ static const bool have_madvise_huge =
false
#endif
;
static const bool have_process_madvise =
#ifdef JEMALLOC_HAVE_PROCESS_MADVISE
true
#else
false
#endif
;
static const bool config_fill =
#ifdef JEMALLOC_FILL
true
@ -114,6 +130,13 @@ static const bool config_prof_libunwind =
false
#endif
;
static const bool config_prof_frameptr =
#ifdef JEMALLOC_PROF_FRAME_POINTER
true
#else
false
#endif
;
static const bool maps_coalesce =
#ifdef JEMALLOC_MAPS_COALESCE
true
@ -215,7 +238,7 @@ static const bool config_enable_cxx =
#endif
;
#if defined(_WIN32) || defined(JEMALLOC_HAVE_SCHED_GETCPU)
#if defined(_WIN32) || defined(__APPLE__) || defined(JEMALLOC_HAVE_SCHED_GETCPU)
/* Currently percpu_arena depends on sched_getcpu. */
#define JEMALLOC_PERCPU_ARENA
#endif

49
include/jemalloc/internal/jemalloc_probe.h Normal file
View file

@ -0,0 +1,49 @@
#ifndef JEMALLOC_INTERNAL_JEMALLOC_PROBE_H
#define JEMALLOC_INTERNAL_JEMALLOC_PROBE_H
#include <jemalloc/internal/jemalloc_preamble.h>
#ifdef JEMALLOC_EXPERIMENTAL_USDT_STAP
#include <jemalloc/internal/jemalloc_probe_stap.h>
#elif defined(JEMALLOC_EXPERIMENTAL_USDT_CUSTOM)
#include <jemalloc/internal/jemalloc_probe_custom.h>
#elif defined(_MSC_VER)
#define JE_USDT(name, N, ...) /* Nothing */
#else /* no USDT, just check the args */
#define JE_USDT(name, N, ...) _JE_USDT_CHECK_ARG##N(__VA_ARGS__)
#define _JE_USDT_CHECK_ARG1(a) \
do { \
(void)(a); \
} while (0)
#define _JE_USDT_CHECK_ARG2(a, b) \
do { \
(void)(a); \
(void)(b); \
} while (0)
#define _JE_USDT_CHECK_ARG3(a, b, c) \
do { \
(void)(a); \
(void)(b); \
(void)(c); \
} while (0)
#define _JE_USDT_CHECK_ARG4(a, b, c, d) \
do { \
(void)(a); \
(void)(b); \
(void)(c); \
(void)(d); \
} while (0)
#define _JE_USDT_CHECK_ARG5(a, b, c, d, e) \
do { \
(void)(a); \
(void)(b); \
(void)(c); \
(void)(d); \
(void)(e); \
} while (0)
#endif /* JEMALLOC_EXPERIMENTAL_USDT_* */
#endif /* JEMALLOC_INTERNAL_JEMALLOC_PROBE_H */

148
include/jemalloc/internal/jemalloc_probe_custom.h Normal file
View file

@ -0,0 +1,148 @@
#ifndef JEMALLOC_INTERNAL_JEMALLOC_PROBE_CUSTOM_H
#define JEMALLOC_INTERNAL_JEMALLOC_PROBE_CUSTOM_H
/* clang-format off */
/*
* This section is based on sys/sdt.h and
* https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation
*/
/* Emit NOP for the probe. */
#if (defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(__arm__)) && defined(__linux__)
#define JE_SDT_NOP nop
#else
#error "Architecture not supported"
#endif
/* Assembly macros */
#define JE_SDT_S(x) #x
#define JE_SDT_ASM_1(x) JE_SDT_S(x) "\n"
#define JE_SDT_ASM_2(x, y) \
JE_SDT_S(x) "," JE_SDT_S(y) "\n"
#define JE_SDT_ASM_3(x, y, z) \
JE_SDT_S(x) "," JE_SDT_S(y) "," JE_SDT_S(z) "\n"
#define JE_SDT_ASM_3(x, y, z) \
JE_SDT_S(x) "," JE_SDT_S(y) "," JE_SDT_S(z) "\n"
#define JE_SDT_ASM_4(x, y, z, p) \
JE_SDT_S(x) "," JE_SDT_S(y) "," JE_SDT_S(z) "," JE_SDT_S(p) "\n"
#define JE_SDT_ASM_5(x, y, z, p, q) \
JE_SDT_S(x) "," JE_SDT_S(y) "," JE_SDT_S(z) "," JE_SDT_S(p) "," \
JE_SDT_S(q) "\n"
/* Arg size */
#ifdef __LP64__
#define JE_SDT_ASM_ADDR .8byte
#else
#define JE_SDT_ASM_ADDR .4byte
#endif
#define JE_SDT_NOTE_NAME "stapsdt"
#define JE_SDT_NOTE_TYPE 3
#define JE_SDT_SEMAPHORE_NONE(provider, name) \
JE_SDT_ASM_1(JE_SDT_ASM_ADDR 0) /* No Semaphore support */
#define JE_SDT_SEMAPHORE_OPERAND(provider, name) \
[__sdt_semaphore] "ip" (0) /* No Semaphore */
#define JE_SDT_ASM_STRING(x) JE_SDT_ASM_1(.asciz JE_SDT_S(x))
#define JE_SDT_NOTE(provider, name, arg_template) \
JE_SDT_ASM_1(990: JE_SDT_NOP) \
JE_SDT_ASM_3( .pushsection .note.stapsdt,"?","note") \
JE_SDT_ASM_1( .balign 4) \
JE_SDT_ASM_3( .4byte 992f-991f, 994f-993f, JE_SDT_NOTE_TYPE) \
JE_SDT_ASM_1(991: .asciz JE_SDT_NOTE_NAME) \
JE_SDT_ASM_1(992: .balign 4) \
JE_SDT_ASM_1(993: JE_SDT_ASM_ADDR 990b) \
JE_SDT_ASM_1( JE_SDT_ASM_ADDR _.stapsdt.base) \
JE_SDT_SEMAPHORE_NONE(provider, name) \
JE_SDT_ASM_STRING(provider) \
JE_SDT_ASM_STRING(name) \
JE_SDT_ASM_STRING(arg_template) \
JE_SDT_ASM_1(994: .balign 4) \
JE_SDT_ASM_1( .popsection)
#define JE_SDT_BASE \
JE_SDT_ASM_1( .ifndef _.stapsdt.base) \
JE_SDT_ASM_5( .pushsection .stapsdt.base, "aG", "progbits", \
.stapsdt.base,comdat) \
JE_SDT_ASM_1( .weak _.stapsdt.base) \
JE_SDT_ASM_1( .hidden _.stapsdt.base) \
JE_SDT_ASM_1( _.stapsdt.base: .space 1) \
JE_SDT_ASM_2( .size _.stapsdt.base, 1) \
JE_SDT_ASM_1( .popsection) \
JE_SDT_ASM_1( .endif)
/*
* Default constraint for probes arguments.
* See https://gcc.gnu.org/onlinedocs/gcc/Constraints.html
*/
#ifndef JE_SDT_ARG_CONSTRAINT
#define JE_SDT_ARG_CONSTRAINT "nor"
#endif
#define JE_SDT_ARGARRAY(x) ((__builtin_classify_type(x) == 14) || \
(__builtin_classify_type(x) == 5))
#define JE_SDT_ARGSIZE(x) (JE_SDT_ARGARRAY(x) ? sizeof(void*) : sizeof(x))
/*
* Format of each probe argument as operand. Size tagged with JE_SDT_Sn,
* with "n" constraint. Value is tagged with JE_SDT_An with configured
* constraint.
*/
#define JE_SDT_ARG(n, x) \
[JE_SDT_S##n] "n" ((size_t)JE_SDT_ARGSIZE(x)), \
[JE_SDT_A##n] JE_SDT_ARG_CONSTRAINT(x)
/* Templates to append arguments as operands. */
#define JE_SDT_OPERANDS_0() [__sdt_dummy] "g" (0)
#define JE_SDT_OPERANDS_1(_1) JE_SDT_ARG(1, _1)
#define JE_SDT_OPERANDS_2(_1, _2) JE_SDT_OPERANDS_1(_1), JE_SDT_ARG(2, _2)
#define JE_SDT_OPERANDS_3(_1, _2, _3) JE_SDT_OPERANDS_2(_1, _2), JE_SDT_ARG(3, _3)
#define JE_SDT_OPERANDS_4(_1, _2, _3, _4) \
JE_SDT_OPERANDS_3(_1, _2, _3), JE_SDT_ARG(4, _4)
#define JE_SDT_OPERANDS_5(_1, _2, _3, _4, _5) \
JE_SDT_OPERANDS_4(_1, _2, _3, _4), JE_SDT_ARG(5, _5)
#define JE_SDT_OPERANDS_6(_1, _2, _3, _4, _5, _6) \
JE_SDT_OPERANDS_5(_1, _2, _3, _4, _5), JE_SDT_ARG(6, _6)
#define JE_SDT_OPERANDS_7(_1, _2, _3, _4, _5, _6, _7) \
JE_SDT_OPERANDS_6(_1, _2, _3, _4, _5, _6), JE_SDT_ARG(7, _7)
/* Templates to reference the arguments from operands. */
#define JE_SDT_ARGFMT(num) %n[JE_SDT_S##num]@%[JE_SDT_A##num]
#define JE_SDT_ARG_TEMPLATE_0 /* No args */
#define JE_SDT_ARG_TEMPLATE_1 JE_SDT_ARGFMT(1)
#define JE_SDT_ARG_TEMPLATE_2 JE_SDT_ARG_TEMPLATE_1 JE_SDT_ARGFMT(2)
#define JE_SDT_ARG_TEMPLATE_3 JE_SDT_ARG_TEMPLATE_2 JE_SDT_ARGFMT(3)
#define JE_SDT_ARG_TEMPLATE_4 JE_SDT_ARG_TEMPLATE_3 JE_SDT_ARGFMT(4)
#define JE_SDT_ARG_TEMPLATE_5 JE_SDT_ARG_TEMPLATE_4 JE_SDT_ARGFMT(5)
#define JE_SDT_ARG_TEMPLATE_6 JE_SDT_ARG_TEMPLATE_5 JE_SDT_ARGFMT(6)
#define JE_SDT_ARG_TEMPLATE_7 JE_SDT_ARG_TEMPLATE_6 JE_SDT_ARGFMT(7)
#define JE_SDT_PROBE( \
provider, name, n, arglist) \
do { \
__asm__ __volatile__( \
JE_SDT_NOTE(provider, name, \
JE_SDT_ARG_TEMPLATE_##n) \
:: JE_SDT_SEMAPHORE_OPERAND(provider, name), \
JE_SDT_OPERANDS_##n arglist); \
__asm__ __volatile__(JE_SDT_BASE); \
} while (0)
#define JE_USDT(name, N, ...) \
JE_SDT_PROBE(jemalloc, name, N, (__VA_ARGS__))
#endif /* JEMALLOC_INTERNAL_JEMALLOC_PROBE_CUSTOM_H */
/* clang-format on */

11
include/jemalloc/internal/jemalloc_probe_stap.h Normal file
View file

@ -0,0 +1,11 @@
#ifndef JEMALLOC_INTERNAL_JEMALLOC_PROBE_STAP_H
#define JEMALLOC_INTERNAL_JEMALLOC_PROBE_STAP_H
#include <sys/sdt.h>
#define JE_USDT(name, N, ...) JE_USDT_PROBE_N(name, N, ##__VA_ARGS__)
#define JE_USDT_PROBE_N(name, N, ...) \
STAP_PROBE##N(jemalloc, name, ##__VA_ARGS__)
#endif /* JEMALLOC_INTERNAL_JEMALLOC_PROBE_STAP_H */

21
include/jemalloc/internal/large_externs.h
View file

@ -1,23 +1,24 @@
#ifndef JEMALLOC_INTERNAL_LARGE_EXTERNS_H
#define JEMALLOC_INTERNAL_LARGE_EXTERNS_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/edata.h"
#include "jemalloc/internal/hook.h"
void *large_malloc(tsdn_t *tsdn, arena_t *arena, size_t usize, bool zero);
void *large_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
bool zero);
bool large_ralloc_no_move(tsdn_t *tsdn, edata_t *edata, size_t usize_min,
size_t usize_max, bool zero);
void *large_palloc(
tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment, bool zero);
bool large_ralloc_no_move(tsdn_t *tsdn, edata_t *edata, size_t usize_min,
size_t usize_max, bool zero);
void *large_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t usize,
size_t alignment, bool zero, tcache_t *tcache,
hook_ralloc_args_t *hook_args);
void large_dalloc_prep_locked(tsdn_t *tsdn, edata_t *edata);
void large_dalloc_finish(tsdn_t *tsdn, edata_t *edata);
void large_dalloc(tsdn_t *tsdn, edata_t *edata);
size_t large_salloc(tsdn_t *tsdn, const edata_t *edata);
void large_prof_info_get(tsd_t *tsd, edata_t *edata, prof_info_t *prof_info,
bool reset_recent);
void large_dalloc_prep_locked(tsdn_t *tsdn, edata_t *edata);
void large_dalloc_finish(tsdn_t *tsdn, edata_t *edata);
void large_dalloc(tsdn_t *tsdn, edata_t *edata);
void large_prof_info_get(
tsd_t *tsd, edata_t *edata, prof_info_t *prof_info, bool reset_recent);
void large_prof_tctx_reset(edata_t *edata);
void large_prof_info_set(edata_t *edata, prof_tctx_t *tctx, size_t size);

68
include/jemalloc/internal/lockedint.h
View file

@ -1,6 +1,11 @@
#ifndef JEMALLOC_INTERNAL_LOCKEDINT_H
#define JEMALLOC_INTERNAL_LOCKEDINT_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/mutex.h"
#include "jemalloc/internal/tsd_types.h"
/*
* In those architectures that support 64-bit atomics, we use atomic updates for
* our 64-bit values. Otherwise, we use a plain uint64_t and synchronize
@ -25,33 +30,34 @@ struct locked_zu_s {
};
#ifndef JEMALLOC_ATOMIC_U64
# define LOCKEDINT_MTX_DECLARE(name) malloc_mutex_t name;
# define LOCKEDINT_MTX_INIT(mu, name, rank, rank_mode) \
malloc_mutex_init(&(mu), name, rank, rank_mode)
# define LOCKEDINT_MTX(mtx) (&(mtx))
# define LOCKEDINT_MTX_LOCK(tsdn, mu) malloc_mutex_lock(tsdn, &(mu))
# define LOCKEDINT_MTX_UNLOCK(tsdn, mu) malloc_mutex_unlock(tsdn, &(mu))
# define LOCKEDINT_MTX_PREFORK(tsdn, mu) malloc_mutex_prefork(tsdn, &(mu))
# define LOCKEDINT_MTX_POSTFORK_PARENT(tsdn, mu) \
malloc_mutex_postfork_parent(tsdn, &(mu))
# define LOCKEDINT_MTX_POSTFORK_CHILD(tsdn, mu) \
malloc_mutex_postfork_child(tsdn, &(mu))
# define LOCKEDINT_MTX_DECLARE(name) malloc_mutex_t name;
# define LOCKEDINT_MTX_INIT(mu, name, rank, rank_mode) \
malloc_mutex_init(&(mu), name, rank, rank_mode)
# define LOCKEDINT_MTX(mtx) (&(mtx))
# define LOCKEDINT_MTX_LOCK(tsdn, mu) malloc_mutex_lock(tsdn, &(mu))
# define LOCKEDINT_MTX_UNLOCK(tsdn, mu) malloc_mutex_unlock(tsdn, &(mu))
# define LOCKEDINT_MTX_PREFORK(tsdn, mu) \
malloc_mutex_prefork(tsdn, &(mu))
# define LOCKEDINT_MTX_POSTFORK_PARENT(tsdn, mu) \
malloc_mutex_postfork_parent(tsdn, &(mu))
# define LOCKEDINT_MTX_POSTFORK_CHILD(tsdn, mu) \
malloc_mutex_postfork_child(tsdn, &(mu))
#else
# define LOCKEDINT_MTX_DECLARE(name)
# define LOCKEDINT_MTX(mtx) NULL
# define LOCKEDINT_MTX_INIT(mu, name, rank, rank_mode) false
# define LOCKEDINT_MTX_LOCK(tsdn, mu)
# define LOCKEDINT_MTX_UNLOCK(tsdn, mu)
# define LOCKEDINT_MTX_PREFORK(tsdn, mu)
# define LOCKEDINT_MTX_POSTFORK_PARENT(tsdn, mu)
# define LOCKEDINT_MTX_POSTFORK_CHILD(tsdn, mu)
# define LOCKEDINT_MTX_DECLARE(name)
# define LOCKEDINT_MTX(mtx) NULL
# define LOCKEDINT_MTX_INIT(mu, name, rank, rank_mode) false
# define LOCKEDINT_MTX_LOCK(tsdn, mu)
# define LOCKEDINT_MTX_UNLOCK(tsdn, mu)
# define LOCKEDINT_MTX_PREFORK(tsdn, mu)
# define LOCKEDINT_MTX_POSTFORK_PARENT(tsdn, mu)
# define LOCKEDINT_MTX_POSTFORK_CHILD(tsdn, mu)
#endif
#ifdef JEMALLOC_ATOMIC_U64
# define LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx) assert((mtx) == NULL)
# define LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx) assert((mtx) == NULL)
#else
# define LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx) \
malloc_mutex_assert_owner(tsdn, (mtx))
# define LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx) \
malloc_mutex_assert_owner(tsdn, (mtx))
#endif
static inline uint64_t
@ -65,8 +71,7 @@ locked_read_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p) {
}
static inline void
locked_inc_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
uint64_t x) {
locked_inc_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p, uint64_t x) {
LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx);
#ifdef JEMALLOC_ATOMIC_U64
atomic_fetch_add_u64(&p->val, x, ATOMIC_RELAXED);
@ -76,8 +81,7 @@ locked_inc_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
}
static inline void
locked_dec_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
uint64_t x) {
locked_dec_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p, uint64_t x) {
LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx);
#ifdef JEMALLOC_ATOMIC_U64
uint64_t r = atomic_fetch_sub_u64(&p->val, x, ATOMIC_RELAXED);
@ -94,7 +98,7 @@ locked_inc_mod_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
const uint64_t x, const uint64_t modulus) {
LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx);
uint64_t before, after;
bool overflow;
bool overflow;
#ifdef JEMALLOC_ATOMIC_U64
before = atomic_load_u64(&p->val, ATOMIC_RELAXED);
do {
@ -104,8 +108,8 @@ locked_inc_mod_u64(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_u64_t *p,
if (overflow) {
after %= modulus;
}
} while (!atomic_compare_exchange_weak_u64(&p->val, &before, after,
ATOMIC_RELAXED, ATOMIC_RELAXED));
} while (!atomic_compare_exchange_weak_u64(
&p->val, &before, after, ATOMIC_RELAXED, ATOMIC_RELAXED));
#else
before = p->val;
after = before + x;
@ -162,8 +166,7 @@ locked_read_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p) {
}
static inline void
locked_inc_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p,
size_t x) {
locked_inc_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p, size_t x) {
LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx);
#ifdef JEMALLOC_ATOMIC_U64
atomic_fetch_add_zu(&p->val, x, ATOMIC_RELAXED);
@ -174,8 +177,7 @@ locked_inc_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p,
}
static inline void
locked_dec_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p,
size_t x) {
locked_dec_zu(tsdn_t *tsdn, malloc_mutex_t *mtx, locked_zu_t *p, size_t x) {
LOCKEDINT_MTX_ASSERT_INTERNAL(tsdn, mtx);
#ifdef JEMALLOC_ATOMIC_U64
size_t r = atomic_fetch_sub_zu(&p->val, x, ATOMIC_RELAXED);

72
include/jemalloc/internal/log.h
View file

@ -1,14 +1,15 @@
#ifndef JEMALLOC_INTERNAL_LOG_H
#define JEMALLOC_INTERNAL_LOG_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/malloc_io.h"
#include "jemalloc/internal/mutex.h"
#ifdef JEMALLOC_LOG
# define JEMALLOC_LOG_VAR_BUFSIZE 1000
# define JEMALLOC_LOG_VAR_BUFSIZE 1000
#else
# define JEMALLOC_LOG_VAR_BUFSIZE 1
# define JEMALLOC_LOG_VAR_BUFSIZE 1
#endif
#define JEMALLOC_LOG_BUFSIZE 4096
@ -26,16 +27,16 @@
* log("extent.a", "log msg for extent.a"); // 5
* log("extent.b", "log msg for extent.b"); // 6
*
* And your malloc_conf option is "log=arena.a|extent", then lines 2, 4, 5, and
* And your malloc_conf option is "log:arena.a|extent", then lines 2, 4, 5, and
* 6 will print at runtime. You can enable logging from all log vars by
* writing "log=.".
* writing "log:.".
*
* None of this should be regarded as a stable API for right now. It's intended
* as a debugging interface, to let us keep around some of our printf-debugging
* statements.
*/
extern char log_var_names[JEMALLOC_LOG_VAR_BUFSIZE];
extern char log_var_names[JEMALLOC_LOG_VAR_BUFSIZE];
extern atomic_b_t log_init_done;
typedef struct log_var_s log_var_t;
@ -44,7 +45,7 @@ struct log_var_s {
* Lowest bit is "inited", second lowest is "enabled". Putting them in
* a single word lets us avoid any fences on weak architectures.
*/
atomic_u_t state;
atomic_u_t state;
const char *name;
};
@ -52,7 +53,8 @@ struct log_var_s {
#define LOG_INITIALIZED_NOT_ENABLED 1U
#define LOG_ENABLED 2U
#define LOG_VAR_INIT(name_str) {ATOMIC_INIT(LOG_NOT_INITIALIZED), name_str}
#define LOG_VAR_INIT(name_str) \
{ ATOMIC_INIT(LOG_NOT_INITIALIZED), name_str }
/*
* Returns the value we should assume for state (which is not necessarily
@ -62,21 +64,21 @@ struct log_var_s {
unsigned log_var_update_state(log_var_t *log_var);
/* We factor out the metadata management to allow us to test more easily. */
#define log_do_begin(log_var) \
if (config_log) { \
unsigned log_state = atomic_load_u(&(log_var).state, \
ATOMIC_RELAXED); \
if (unlikely(log_state == LOG_NOT_INITIALIZED)) { \
log_state = log_var_update_state(&(log_var)); \
assert(log_state != LOG_NOT_INITIALIZED); \
} \
if (log_state == LOG_ENABLED) { \
{
/* User code executes here. */
#define log_do_end(log_var) \
} \
} \
}
#define log_do_begin(log_var) \
if (config_log) { \
unsigned log_state = atomic_load_u( \
&(log_var).state, ATOMIC_RELAXED); \
if (unlikely(log_state == LOG_NOT_INITIALIZED)) { \
log_state = log_var_update_state(&(log_var)); \
assert(log_state != LOG_NOT_INITIALIZED); \
} \
if (log_state == LOG_ENABLED) { \
{
/* User code executes here. */
#define log_do_end(log_var) \
} \
} \
}
/*
* MSVC has some preprocessor bugs in its expansion of __VA_ARGS__ during
@ -87,29 +89,29 @@ if (config_log) { \
*/
static inline void
log_impl_varargs(const char *name, ...) {
char buf[JEMALLOC_LOG_BUFSIZE];
char buf[JEMALLOC_LOG_BUFSIZE];
va_list ap;
va_start(ap, name);
const char *format = va_arg(ap, const char *);
size_t dst_offset = 0;
size_t dst_offset = 0;
dst_offset += malloc_snprintf(buf, JEMALLOC_LOG_BUFSIZE, "%s: ", name);
dst_offset += malloc_vsnprintf(buf + dst_offset,
JEMALLOC_LOG_BUFSIZE - dst_offset, format, ap);
dst_offset += malloc_snprintf(buf + dst_offset,
JEMALLOC_LOG_BUFSIZE - dst_offset, "\n");
dst_offset += malloc_vsnprintf(
buf + dst_offset, JEMALLOC_LOG_BUFSIZE - dst_offset, format, ap);
malloc_snprintf(
buf + dst_offset, JEMALLOC_LOG_BUFSIZE - dst_offset, "\n");
va_end(ap);
malloc_write(buf);
}
/* Call as log("log.var.str", "format_string %d", arg_for_format_string); */
#define LOG(log_var_str, ...) \
do { \
static log_var_t log_var = LOG_VAR_INIT(log_var_str); \
log_do_begin(log_var) \
log_impl_varargs((log_var).name, __VA_ARGS__); \
log_do_end(log_var) \
} while (0)
#define LOG(log_var_str, ...) \
do { \
static log_var_t log_var = LOG_VAR_INIT(log_var_str); \
log_do_begin(log_var) \
log_impl_varargs((log_var).name, __VA_ARGS__); \
log_do_end(log_var) \
} while (0)
#endif /* JEMALLOC_INTERNAL_LOG_H */

120
include/jemalloc/internal/malloc_io.h
View file

@ -1,105 +1,101 @@
#ifndef JEMALLOC_INTERNAL_MALLOC_IO_H
#define JEMALLOC_INTERNAL_MALLOC_IO_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/jemalloc_internal_types.h"
#ifdef _WIN32
# ifdef _WIN64
# define FMT64_PREFIX "ll"
# define FMTPTR_PREFIX "ll"
# else
# define FMT64_PREFIX "ll"
# define FMTPTR_PREFIX ""
# endif
# define FMTd32 "d"
# define FMTu32 "u"
# define FMTx32 "x"
# define FMTd64 FMT64_PREFIX "d"
# define FMTu64 FMT64_PREFIX "u"
# define FMTx64 FMT64_PREFIX "x"
# define FMTdPTR FMTPTR_PREFIX "d"
# define FMTuPTR FMTPTR_PREFIX "u"
# define FMTxPTR FMTPTR_PREFIX "x"
# ifdef _WIN64
# define FMT64_PREFIX "ll"
# define FMTPTR_PREFIX "ll"
# else
# define FMT64_PREFIX "ll"
# define FMTPTR_PREFIX ""
# endif
# define FMTd32 "d"
# define FMTu32 "u"
# define FMTx32 "x"
# define FMTd64 FMT64_PREFIX "d"
# define FMTu64 FMT64_PREFIX "u"
# define FMTx64 FMT64_PREFIX "x"
# define FMTdPTR FMTPTR_PREFIX "d"
# define FMTuPTR FMTPTR_PREFIX "u"
# define FMTxPTR FMTPTR_PREFIX "x"
#else
# include <inttypes.h>
# define FMTd32 PRId32
# define FMTu32 PRIu32
# define FMTx32 PRIx32
# define FMTd64 PRId64
# define FMTu64 PRIu64
# define FMTx64 PRIx64
# define FMTdPTR PRIdPTR
# define FMTuPTR PRIuPTR
# define FMTxPTR PRIxPTR
# include <inttypes.h>
# define FMTd32 PRId32
# define FMTu32 PRIu32
# define FMTx32 PRIx32
# define FMTd64 PRId64
# define FMTu64 PRIu64
# define FMTx64 PRIx64
# define FMTdPTR PRIdPTR
# define FMTuPTR PRIuPTR
# define FMTxPTR PRIxPTR
#endif
/* Size of stack-allocated buffer passed to buferror(). */
#define BUFERROR_BUF 64
#define BUFERROR_BUF 64
/*
* Size of stack-allocated buffer used by malloc_{,v,vc}printf(). This must be
* large enough for all possible uses within jemalloc.
*/
#define MALLOC_PRINTF_BUFSIZE 4096
#define MALLOC_PRINTF_BUFSIZE 4096
write_cb_t wrtmessage;
int buferror(int err, char *buf, size_t buflen);
uintmax_t malloc_strtoumax(const char *restrict nptr, char **restrict endptr,
int base);
int buferror(int err, char *buf, size_t buflen);
uintmax_t malloc_strtoumax(
const char *restrict nptr, char **restrict endptr, int base);
void malloc_write(const char *s);
/*
* malloc_vsnprintf() supports a subset of snprintf(3) that avoids floating
* point math.
*/
size_t malloc_vsnprintf(char *str, size_t size, const char *format,
va_list ap);
size_t malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap);
size_t malloc_snprintf(char *str, size_t size, const char *format, ...)
JEMALLOC_FORMAT_PRINTF(3, 4);
/*
* The caller can set write_cb to null to choose to print with the
* je_malloc_message hook.
*/
void malloc_vcprintf(write_cb_t *write_cb, void *cbopaque, const char *format,
va_list ap);
void malloc_vcprintf(
write_cb_t *write_cb, void *cbopaque, const char *format, va_list ap);
void malloc_cprintf(write_cb_t *write_cb, void *cbopaque, const char *format,
...) JEMALLOC_FORMAT_PRINTF(3, 4);
void malloc_printf(const char *format, ...) JEMALLOC_FORMAT_PRINTF(1, 2);
static inline ssize_t
malloc_write_fd(int fd, const void *buf, size_t count) {
#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_write)
/*
* Use syscall(2) rather than write(2) when possible in order to avoid
* the possibility of memory allocation within libc. This is necessary
* on FreeBSD; most operating systems do not have this problem though.
*
* syscall() returns long or int, depending on platform, so capture the
* result in the widest plausible type to avoid compiler warnings.
*/
long result = syscall(SYS_write, fd, buf, count);
ssize_t malloc_write_fd(int fd, const void *buf, size_t count);
ssize_t malloc_read_fd(int fd, void *buf, size_t count);
static inline int
malloc_open(const char *path, int flags) {
#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_open)
return (int)syscall(SYS_open, path, flags);
#elif defined(JEMALLOC_USE_SYSCALL) && defined(SYS_openat)
return (int)syscall(SYS_openat, AT_FDCWD, path, flags);
#else
ssize_t result = (ssize_t)write(fd, buf,
#ifdef _WIN32
(unsigned int)
return open(path, flags);
#endif
count);
#endif
return (ssize_t)result;
}
static inline ssize_t
malloc_read_fd(int fd, void *buf, size_t count) {
#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_read)
long result = syscall(SYS_read, fd, buf, count);
static inline int
malloc_close(int fd) {
#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_close)
return (int)syscall(SYS_close, fd);
#else
ssize_t result = read(fd, buf,
#ifdef _WIN32
(unsigned int)
return close(fd);
#endif
count);
}
static inline off_t
malloc_lseek(int fd, off_t offset, int whence) {
#if defined(JEMALLOC_USE_SYSCALL) && defined(SYS_lseek)
return (off_t)syscall(SYS_lseek, fd, offset, whence);
#else
return lseek(fd, offset, whence);
#endif
return (ssize_t)result;
}
#endif /* JEMALLOC_INTERNAL_MALLOC_IO_H */

3
include/jemalloc/internal/mpsc_queue.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_MPSC_QUEUE_H
#define JEMALLOC_INTERNAL_MPSC_QUEUE_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
/*
@ -25,6 +26,7 @@
* two-stack tricks reverses orders in the lock-free first stack).
*/
/* clang-format off */
#define mpsc_queue(a_type) \
struct { \
atomic_p_t tail; \
@ -130,5 +132,6 @@ a_prefix##pop_batch(a_queue_type *queue, a_list_type *dst) { \
} \
ql_concat(dst, &reversed, a_link); \
}
/* clang-format on */
#endif /* JEMALLOC_INTERNAL_MPSC_QUEUE_H */

227
include/jemalloc/internal/mutex.h
View file

@ -1,6 +1,7 @@
#ifndef JEMALLOC_INTERNAL_MUTEX_H
#define JEMALLOC_INTERNAL_MUTEX_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/mutex_prof.h"
#include "jemalloc/internal/tsd.h"
@ -30,26 +31,29 @@ struct malloc_mutex_s {
* avoid prefetching a modified cacheline (for the
* unlocking thread).
*/
mutex_prof_data_t prof_data;
#ifdef _WIN32
# if _WIN32_WINNT >= 0x0600
SRWLOCK lock;
# else
CRITICAL_SECTION lock;
# endif
#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
os_unfair_lock lock;
#elif (defined(JEMALLOC_MUTEX_INIT_CB))
pthread_mutex_t lock;
malloc_mutex_t *postponed_next;
#else
pthread_mutex_t lock;
#endif
mutex_prof_data_t prof_data;
/*
* Hint flag to avoid exclusive cache line contention
* during spin waiting
* during spin waiting. Placed along with prof_data
* since it's always modified even with no contention.
* Modified by the lock owner only (after acquired, and
* before release), and may be read by other threads.
*/
atomic_b_t locked;
atomic_b_t locked;
#ifdef _WIN32
# if _WIN32_WINNT >= 0x0600
SRWLOCK lock;
# else
CRITICAL_SECTION lock;
# endif
#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
os_unfair_lock lock;
#elif (defined(JEMALLOC_MUTEX_INIT_CB))
pthread_mutex_t lock;
malloc_mutex_t *postponed_next;
#else
pthread_mutex_t lock;
#endif
};
/*
* We only touch witness when configured w/ debug. However we
@ -58,82 +62,118 @@ struct malloc_mutex_s {
* memory cost.
*/
#if !defined(JEMALLOC_DEBUG)
witness_t witness;
malloc_mutex_lock_order_t lock_order;
witness_t witness;
malloc_mutex_lock_order_t lock_order;
#endif
};
#if defined(JEMALLOC_DEBUG)
witness_t witness;
malloc_mutex_lock_order_t lock_order;
witness_t witness;
malloc_mutex_lock_order_t lock_order;
#endif
};
#ifdef _WIN32
# if _WIN32_WINNT >= 0x0600
# define MALLOC_MUTEX_LOCK(m) AcquireSRWLockExclusive(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) ReleaseSRWLockExclusive(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) (!TryAcquireSRWLockExclusive(&(m)->lock))
# else
# define MALLOC_MUTEX_LOCK(m) EnterCriticalSection(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) LeaveCriticalSection(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) (!TryEnterCriticalSection(&(m)->lock))
# endif
# if _WIN32_WINNT >= 0x0600
# define MALLOC_MUTEX_LOCK(m) AcquireSRWLockExclusive(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) \
ReleaseSRWLockExclusive(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) \
(!TryAcquireSRWLockExclusive(&(m)->lock))
# else
# define MALLOC_MUTEX_LOCK(m) EnterCriticalSection(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) LeaveCriticalSection(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) \
(!TryEnterCriticalSection(&(m)->lock))
# endif
#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
# define MALLOC_MUTEX_LOCK(m) os_unfair_lock_lock(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) os_unfair_lock_unlock(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) (!os_unfair_lock_trylock(&(m)->lock))
# define MALLOC_MUTEX_LOCK(m) os_unfair_lock_lock(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) os_unfair_lock_unlock(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) (!os_unfair_lock_trylock(&(m)->lock))
#else
# define MALLOC_MUTEX_LOCK(m) pthread_mutex_lock(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) pthread_mutex_unlock(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) (pthread_mutex_trylock(&(m)->lock) != 0)
# define MALLOC_MUTEX_LOCK(m) pthread_mutex_lock(&(m)->lock)
# define MALLOC_MUTEX_UNLOCK(m) pthread_mutex_unlock(&(m)->lock)
# define MALLOC_MUTEX_TRYLOCK(m) (pthread_mutex_trylock(&(m)->lock) != 0)
#endif
#define LOCK_PROF_DATA_INITIALIZER \
{NSTIME_ZERO_INITIALIZER, NSTIME_ZERO_INITIALIZER, 0, 0, 0, \
ATOMIC_INIT(0), 0, NULL, 0}
#define LOCK_PROF_DATA_INITIALIZER \
{ \
NSTIME_ZERO_INITIALIZER, NSTIME_ZERO_INITIALIZER, 0, 0, 0, \
ATOMIC_INIT(0), 0, NULL, 0 \
}
#ifdef _WIN32
# define MALLOC_MUTEX_INITIALIZER
# define MALLOC_MUTEX_INITIALIZER
#elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
# if defined(JEMALLOC_DEBUG)
# define MALLOC_MUTEX_INITIALIZER \
{{{LOCK_PROF_DATA_INITIALIZER, OS_UNFAIR_LOCK_INIT, ATOMIC_INIT(false)}}, \
WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT), 0}
# else
# define MALLOC_MUTEX_INITIALIZER \
{{{LOCK_PROF_DATA_INITIALIZER, OS_UNFAIR_LOCK_INIT, ATOMIC_INIT(false)}}, \
WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT)}
# endif
# if defined(JEMALLOC_DEBUG)
# define MALLOC_MUTEX_INITIALIZER \
{ \
{{LOCK_PROF_DATA_INITIALIZER, \
ATOMIC_INIT(false), OS_UNFAIR_LOCK_INIT}}, \
WITNESS_INITIALIZER( \
"mutex", WITNESS_RANK_OMIT), \
0 \
}
# else
# define MALLOC_MUTEX_INITIALIZER \
{ \
{{LOCK_PROF_DATA_INITIALIZER, \
ATOMIC_INIT(false), OS_UNFAIR_LOCK_INIT}}, \
WITNESS_INITIALIZER( \
"mutex", WITNESS_RANK_OMIT) \
}
# endif
#elif (defined(JEMALLOC_MUTEX_INIT_CB))
# if (defined(JEMALLOC_DEBUG))
# define MALLOC_MUTEX_INITIALIZER \
{{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, NULL, ATOMIC_INIT(false)}}, \
WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT), 0}
# else
# define MALLOC_MUTEX_INITIALIZER \
{{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, NULL, ATOMIC_INIT(false)}}, \
WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT)}
# endif
# if (defined(JEMALLOC_DEBUG))
# define MALLOC_MUTEX_INITIALIZER \
{ \
{{LOCK_PROF_DATA_INITIALIZER, \
ATOMIC_INIT(false), \
PTHREAD_MUTEX_INITIALIZER, NULL}}, \
WITNESS_INITIALIZER( \
"mutex", WITNESS_RANK_OMIT), \
0 \
}
# else
# define MALLOC_MUTEX_INITIALIZER \
{ \
{{LOCK_PROF_DATA_INITIALIZER, \
ATOMIC_INIT(false), \
PTHREAD_MUTEX_INITIALIZER, NULL}}, \
WITNESS_INITIALIZER( \
"mutex", WITNESS_RANK_OMIT) \
}
# endif
#else
# define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT
# if defined(JEMALLOC_DEBUG)
# define MALLOC_MUTEX_INITIALIZER \
{{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, ATOMIC_INIT(false)}}, \
WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT), 0}
# else
# define MALLOC_MUTEX_INITIALIZER \
{{{LOCK_PROF_DATA_INITIALIZER, PTHREAD_MUTEX_INITIALIZER, ATOMIC_INIT(false)}}, \
WITNESS_INITIALIZER("mutex", WITNESS_RANK_OMIT)}
# endif
# define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT
# if defined(JEMALLOC_DEBUG)
# define MALLOC_MUTEX_INITIALIZER \
{ \
{{LOCK_PROF_DATA_INITIALIZER, \
ATOMIC_INIT(false), \
PTHREAD_MUTEX_INITIALIZER}}, \
WITNESS_INITIALIZER( \
"mutex", WITNESS_RANK_OMIT), \
0 \
}
# else
# define MALLOC_MUTEX_INITIALIZER \
{ \
{{LOCK_PROF_DATA_INITIALIZER, \
ATOMIC_INIT(false), \
PTHREAD_MUTEX_INITIALIZER}}, \
WITNESS_INITIALIZER( \
"mutex", WITNESS_RANK_OMIT) \
}
# endif
#endif
#ifdef JEMALLOC_LAZY_LOCK
extern bool isthreaded;
#else
# undef isthreaded /* Undo private_namespace.h definition. */
# define isthreaded true
# undef isthreaded /* Undo private_namespace.h definition. */
# define isthreaded true
#endif
bool malloc_mutex_init(malloc_mutex_t *mutex, const char *name,
@ -154,7 +194,12 @@ malloc_mutex_lock_final(malloc_mutex_t *mutex) {
static inline bool
malloc_mutex_trylock_final(malloc_mutex_t *mutex) {
return MALLOC_MUTEX_TRYLOCK(mutex);
bool failed = MALLOC_MUTEX_TRYLOCK(mutex);
if (!failed) {
atomic_store_b(&mutex->locked, true, ATOMIC_RELAXED);
}
return failed;
}
static inline void
@ -169,15 +214,21 @@ mutex_owner_stats_update(tsdn_t *tsdn, malloc_mutex_t *mutex) {
}
}
static inline bool
malloc_mutex_is_locked(malloc_mutex_t *mutex) {
/* Used for sanity checking only. */
return atomic_load_b(&mutex->locked, ATOMIC_RELAXED);
}
/* Trylock: return false if the lock is successfully acquired. */
static inline bool
malloc_mutex_trylock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
witness_assert_not_owner(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
if (isthreaded) {
if (malloc_mutex_trylock_final(mutex)) {
atomic_store_b(&mutex->locked, true, ATOMIC_RELAXED);
return true;
}
assert(malloc_mutex_is_locked(mutex));
mutex_owner_stats_update(tsdn, mutex);
}
witness_lock(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
@ -199,12 +250,12 @@ malloc_mutex_prof_merge(mutex_prof_data_t *sum, mutex_prof_data_t *data) {
if (sum->max_n_thds < data->max_n_thds) {
sum->max_n_thds = data->max_n_thds;
}
uint32_t cur_n_waiting_thds = atomic_load_u32(&sum->n_waiting_thds,
ATOMIC_RELAXED);
uint32_t new_n_waiting_thds = cur_n_waiting_thds + atomic_load_u32(
&data->n_waiting_thds, ATOMIC_RELAXED);
atomic_store_u32(&sum->n_waiting_thds, new_n_waiting_thds,
ATOMIC_RELAXED);
uint32_t cur_n_waiting_thds = atomic_load_u32(
&sum->n_waiting_thds, ATOMIC_RELAXED);
uint32_t new_n_waiting_thds = cur_n_waiting_thds
+ atomic_load_u32(&data->n_waiting_thds, ATOMIC_RELAXED);
atomic_store_u32(
&sum->n_waiting_thds, new_n_waiting_thds, ATOMIC_RELAXED);
sum->n_owner_switches += data->n_owner_switches;
sum->n_lock_ops += data->n_lock_ops;
}
@ -215,8 +266,8 @@ malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
if (isthreaded) {
if (malloc_mutex_trylock_final(mutex)) {
malloc_mutex_lock_slow(mutex);
atomic_store_b(&mutex->locked, true, ATOMIC_RELAXED);
}
assert(malloc_mutex_is_locked(mutex));
mutex_owner_stats_update(tsdn, mutex);
}
witness_lock(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
@ -224,9 +275,10 @@ malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
static inline void
malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
atomic_store_b(&mutex->locked, false, ATOMIC_RELAXED);
witness_unlock(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
if (isthreaded) {
assert(malloc_mutex_is_locked(mutex));
atomic_store_b(&mutex->locked, false, ATOMIC_RELAXED);
MALLOC_MUTEX_UNLOCK(mutex);
}
}
@ -234,6 +286,9 @@ malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex) {
static inline void
malloc_mutex_assert_owner(tsdn_t *tsdn, malloc_mutex_t *mutex) {
witness_assert_owner(tsdn_witness_tsdp_get(tsdn), &mutex->witness);
if (isthreaded) {
assert(malloc_mutex_is_locked(mutex));
}
}
static inline void
@ -255,16 +310,16 @@ malloc_mutex_prof_copy(mutex_prof_data_t *dst, mutex_prof_data_t *source) {
/* Copy the prof data from mutex for processing. */
static inline void
malloc_mutex_prof_read(tsdn_t *tsdn, mutex_prof_data_t *data,
malloc_mutex_t *mutex) {
malloc_mutex_prof_read(
tsdn_t *tsdn, mutex_prof_data_t *data, malloc_mutex_t *mutex) {
/* Can only read holding the mutex. */
malloc_mutex_assert_owner(tsdn, mutex);
malloc_mutex_prof_copy(data, &mutex->prof_data);
}
static inline void
malloc_mutex_prof_accum(tsdn_t *tsdn, mutex_prof_data_t *data,
malloc_mutex_t *mutex) {
malloc_mutex_prof_accum(
tsdn_t *tsdn, mutex_prof_data_t *data, malloc_mutex_t *mutex) {
mutex_prof_data_t *source = &mutex->prof_data;
/* Can only read holding the mutex. */
malloc_mutex_assert_owner(tsdn, mutex);
@ -286,8 +341,8 @@ malloc_mutex_prof_accum(tsdn_t *tsdn, mutex_prof_data_t *data,
/* Compare the prof data and update to the maximum. */
static inline void
malloc_mutex_prof_max_update(tsdn_t *tsdn, mutex_prof_data_t *data,
malloc_mutex_t *mutex) {
malloc_mutex_prof_max_update(
tsdn_t *tsdn, mutex_prof_data_t *data, malloc_mutex_t *mutex) {
mutex_prof_data_t *source = &mutex->prof_data;
/* Can only read holding the mutex. */
malloc_mutex_assert_owner(tsdn, mutex);

113
include/jemalloc/internal/mutex_prof.h
View file

@ -1,80 +1,81 @@
#ifndef JEMALLOC_INTERNAL_MUTEX_PROF_H
#define JEMALLOC_INTERNAL_MUTEX_PROF_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/atomic.h"
#include "jemalloc/internal/nstime.h"
#include "jemalloc/internal/tsd_types.h"
#define MUTEX_PROF_GLOBAL_MUTEXES \
OP(background_thread) \
OP(max_per_bg_thd) \
OP(ctl) \
OP(prof) \
OP(prof_thds_data) \
OP(prof_dump) \
OP(prof_recent_alloc) \
OP(prof_recent_dump) \
OP(prof_stats)
#define MUTEX_PROF_GLOBAL_MUTEXES \
OP(background_thread) \
OP(max_per_bg_thd) \
OP(ctl) \
OP(prof) \
OP(prof_thds_data) \
OP(prof_dump) \
OP(prof_recent_alloc) \
OP(prof_recent_dump) \
OP(prof_stats)
typedef enum {
#define OP(mtx) global_prof_mutex_##mtx,
MUTEX_PROF_GLOBAL_MUTEXES
#undef OP
mutex_prof_num_global_mutexes
mutex_prof_num_global_mutexes
} mutex_prof_global_ind_t;
#define MUTEX_PROF_ARENA_MUTEXES \
OP(large) \
OP(extent_avail) \
OP(extents_dirty) \
OP(extents_muzzy) \
OP(extents_retained) \
OP(decay_dirty) \
OP(decay_muzzy) \
OP(base) \
OP(tcache_list) \
OP(hpa_shard) \
OP(hpa_shard_grow) \
OP(hpa_sec)
#define MUTEX_PROF_ARENA_MUTEXES \
OP(large) \
OP(extent_avail) \
OP(extents_dirty) \
OP(extents_muzzy) \
OP(extents_retained) \
OP(decay_dirty) \
OP(decay_muzzy) \
OP(base) \
OP(tcache_list) \
OP(hpa_shard) \
OP(hpa_shard_grow) \
OP(hpa_sec)
typedef enum {
#define OP(mtx) arena_prof_mutex_##mtx,
MUTEX_PROF_ARENA_MUTEXES
#undef OP
mutex_prof_num_arena_mutexes
mutex_prof_num_arena_mutexes
} mutex_prof_arena_ind_t;
/*
* The forth parameter is a boolean value that is true for derived rate counters
* and false for real ones.
*/
#define MUTEX_PROF_UINT64_COUNTERS \
OP(num_ops, uint64_t, "n_lock_ops", false, num_ops) \
OP(num_ops_ps, uint64_t, "(#/sec)", true, num_ops) \
OP(num_wait, uint64_t, "n_waiting", false, num_wait) \
OP(num_wait_ps, uint64_t, "(#/sec)", true, num_wait) \
OP(num_spin_acq, uint64_t, "n_spin_acq", false, num_spin_acq) \
OP(num_spin_acq_ps, uint64_t, "(#/sec)", true, num_spin_acq) \
OP(num_owner_switch, uint64_t, "n_owner_switch", false, num_owner_switch) \
OP(num_owner_switch_ps, uint64_t, "(#/sec)", true, num_owner_switch) \
OP(total_wait_time, uint64_t, "total_wait_ns", false, total_wait_time) \
OP(total_wait_time_ps, uint64_t, "(#/sec)", true, total_wait_time) \
OP(max_wait_time, uint64_t, "max_wait_ns", false, max_wait_time)
#define MUTEX_PROF_UINT64_COUNTERS \
OP(num_ops, uint64_t, "n_lock_ops", false, num_ops) \
OP(num_ops_ps, uint64_t, "(#/sec)", true, num_ops) \
OP(num_wait, uint64_t, "n_waiting", false, num_wait) \
OP(num_wait_ps, uint64_t, "(#/sec)", true, num_wait) \
OP(num_spin_acq, uint64_t, "n_spin_acq", false, num_spin_acq) \
OP(num_spin_acq_ps, uint64_t, "(#/sec)", true, num_spin_acq) \
OP(num_owner_switch, uint64_t, "n_owner_switch", false, \
num_owner_switch) \
OP(num_owner_switch_ps, uint64_t, "(#/sec)", true, num_owner_switch) \
OP(total_wait_time, uint64_t, "total_wait_ns", false, total_wait_time) \
OP(total_wait_time_ps, uint64_t, "(#/sec)", true, total_wait_time) \
OP(max_wait_time, uint64_t, "max_wait_ns", false, max_wait_time)
#define MUTEX_PROF_UINT32_COUNTERS \
OP(max_num_thds, uint32_t, "max_n_thds", false, max_num_thds)
#define MUTEX_PROF_UINT32_COUNTERS \
OP(max_num_thds, uint32_t, "max_n_thds", false, max_num_thds)
#define MUTEX_PROF_COUNTERS \
MUTEX_PROF_UINT64_COUNTERS \
MUTEX_PROF_UINT32_COUNTERS
#define MUTEX_PROF_COUNTERS \
MUTEX_PROF_UINT64_COUNTERS \
MUTEX_PROF_UINT32_COUNTERS
#define OP(counter, type, human, derived, base_counter) mutex_counter_##counter,
#define COUNTER_ENUM(counter_list, t) \
typedef enum { \
counter_list \
mutex_prof_num_##t##_counters \
} mutex_prof_##t##_counter_ind_t;
#define COUNTER_ENUM(counter_list, t) \
typedef enum { \
counter_list mutex_prof_num_##t##_counters \
} mutex_prof_##t##_counter_ind_t;
COUNTER_ENUM(MUTEX_PROF_UINT64_COUNTERS, uint64_t)
COUNTER_ENUM(MUTEX_PROF_UINT32_COUNTERS, uint32_t)
@ -88,17 +89,17 @@ typedef struct {
* contention. We update them once we have the lock.
*/
/* Total time (in nano seconds) spent waiting on this mutex. */
nstime_t tot_wait_time;
nstime_t tot_wait_time;
/* Max time (in nano seconds) spent on a single lock operation. */
nstime_t max_wait_time;
nstime_t max_wait_time;
/* # of times have to wait for this mutex (after spinning). */
uint64_t n_wait_times;
uint64_t n_wait_times;
/* # of times acquired the mutex through local spinning. */
uint64_t n_spin_acquired;
uint64_t n_spin_acquired;
/* Max # of threads waiting for the mutex at the same time. */
uint32_t max_n_thds;
uint32_t max_n_thds;
/* Current # of threads waiting on the lock. Atomic synced. */
atomic_u32_t n_waiting_thds;
atomic_u32_t n_waiting_thds;
/*
* Data touched on the fast path. These are modified right after we
@ -107,11 +108,11 @@ typedef struct {
* cacheline.
*/
/* # of times the mutex holder is different than the previous one. */
uint64_t n_owner_switches;
uint64_t n_owner_switches;
/* Previous mutex holder, to facilitate n_owner_switches. */
tsdn_t *prev_owner;
tsdn_t *prev_owner;
/* # of lock() operations in total. */
uint64_t n_lock_ops;
uint64_t n_lock_ops;
} mutex_prof_data_t;
#endif /* JEMALLOC_INTERNAL_MUTEX_PROF_H */

51
include/jemalloc/internal/nstime.h
View file

@ -1,14 +1,19 @@
#ifndef JEMALLOC_INTERNAL_NSTIME_H
#define JEMALLOC_INTERNAL_NSTIME_H
#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/assert.h"
/* Maximum supported number of seconds (~584 years). */
#define NSTIME_SEC_MAX KQU(18446744072)
#define NSTIME_MAGIC ((uint32_t)0xb8a9ce37)
#ifdef JEMALLOC_DEBUG
# define NSTIME_ZERO_INITIALIZER {0, NSTIME_MAGIC}
# define NSTIME_ZERO_INITIALIZER \
{ 0, NSTIME_MAGIC }
#else
# define NSTIME_ZERO_INITIALIZER {0}
# define NSTIME_ZERO_INITIALIZER \
{ 0 }
#endif
typedef struct {
@ -20,43 +25,43 @@ typedef struct {
static const nstime_t nstime_zero = NSTIME_ZERO_INITIALIZER;
void nstime_init(nstime_t *time, uint64_t ns);
void nstime_init2(nstime_t *time, uint64_t sec, uint64_t nsec);
void nstime_init(nstime_t *time, uint64_t ns);
void nstime_init2(nstime_t *time, uint64_t sec, uint64_t nsec);
uint64_t nstime_ns(const nstime_t *time);
uint64_t nstime_ms(const nstime_t *time);
uint64_t nstime_sec(const nstime_t *time);
uint64_t nstime_msec(const nstime_t *time);
uint64_t nstime_nsec(const nstime_t *time);
void nstime_copy(nstime_t *time, const nstime_t *source);
int nstime_compare(const nstime_t *a, const nstime_t *b);
void nstime_add(nstime_t *time, const nstime_t *addend);
void nstime_iadd(nstime_t *time, uint64_t addend);
void nstime_subtract(nstime_t *time, const nstime_t *subtrahend);
void nstime_isubtract(nstime_t *time, uint64_t subtrahend);
void nstime_imultiply(nstime_t *time, uint64_t multiplier);
void nstime_idivide(nstime_t *time, uint64_t divisor);
void nstime_copy(nstime_t *time, const nstime_t *source);
int nstime_compare(const nstime_t *a, const nstime_t *b);
void nstime_add(nstime_t *time, const nstime_t *addend);
void nstime_iadd(nstime_t *time, uint64_t addend);
void nstime_subtract(nstime_t *time, const nstime_t *subtrahend);
void nstime_isubtract(nstime_t *time, uint64_t subtrahend);
void nstime_imultiply(nstime_t *time, uint64_t multiplier);
void nstime_idivide(nstime_t *time, uint64_t divisor);
uint64_t nstime_divide(const nstime_t *time, const nstime_t *divisor);
uint64_t nstime_ns_between(const nstime_t *earlier, const nstime_t *later);
uint64_t nstime_ms_between(const nstime_t *earlier, const nstime_t *later);
uint64_t nstime_ns_since(const nstime_t *past);
uint64_t nstime_ms_since(const nstime_t *past);
typedef bool (nstime_monotonic_t)(void);
typedef bool(nstime_monotonic_t)(void);
extern nstime_monotonic_t *JET_MUTABLE nstime_monotonic;
typedef void (nstime_update_t)(nstime_t *);
typedef void(nstime_update_t)(nstime_t *);
extern nstime_update_t *JET_MUTABLE nstime_update;
typedef void (nstime_prof_update_t)(nstime_t *);
typedef void(nstime_prof_update_t)(nstime_t *);
extern nstime_prof_update_t *JET_MUTABLE nstime_prof_update;
void nstime_init_update(nstime_t *time);
void nstime_prof_init_update(nstime_t *time);
enum prof_time_res_e {
prof_time_res_default = 0,
prof_time_res_high = 1
};
enum prof_time_res_e { prof_time_res_default = 0, prof_time_res_high = 1 };
typedef enum prof_time_res_e prof_time_res_t;
extern prof_time_res_t opt_prof_time_res;
extern const char *prof_time_res_mode_names[];
extern prof_time_res_t opt_prof_time_res;
extern const char *const prof_time_res_mode_names[];
JEMALLOC_ALWAYS_INLINE void
nstime_init_zero(nstime_t *time) {
@ -64,7 +69,7 @@ nstime_init_zero(nstime_t *time) {
}
JEMALLOC_ALWAYS_INLINE bool
nstime_equals_zero(nstime_t *time) {
nstime_equals_zero(const nstime_t *time) {
int diff = nstime_compare(time, &nstime_zero);
assert(diff >= 0);
return diff == 0;

Some files were not shown because too many files have changed in this diff Show more