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
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.
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.
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.
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.
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.
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
On deallocation, sampled pointers (specially aligned) get junked and stashed
into tcache (to prevent immediate reuse). The expected behavior is to have
read-after-free corrupted and stopped by the junk-filling, while
write-after-free is checked when flushing the stashed pointers.
Adding guarded extents, which are regular extents surrounded by guard pages
(mprotected). To reduce syscalls, small guarded extents are cached as a
separate eset in ecache, and decay through the dirty / muzzy / retained pipeline
as usual.
Existing backtrace implementations skip native stack frames from runtimes like
Python. The hook allows to augment the backtraces to attribute allocations to
native functions in heap profiles.
There is a race between the doc generation and the doc installation,
so make the install depend on the build for doc.
Signed-off-by: Mingli Yu <mingli.yu@windriver.com>
This is a simple multi-producer, single-consumer queue. The intended use case
is in the HPA, as we begin supporting hpdatas that move between hpa_shards. We
take just a single CAS as the cost to send a message (or a batch of messages) in
the low-contention case, and lock-freedom lets us avoid some lock-ordering
issues.
Using an edata_t both for hugepages and the allocations within those hugepages
was convenient at first, but has outlived its usefulness. Representing
hugepages explicitly, with their own data structure, will make future
development easier.
