Since the content varies during connection setup and while doing it
(eyeballing), remove these strcut from `connectdata` and replace use
with querying the connection filters. Those keep that information
already.
Change the info logging of established connections to also give the
local address and port.
Closes#17960
Add a connection filter query to obtained the negotiated ALPN
protocol to check in setup/protocols how the connection needs
to behave.
Remove the members `alpn` and `proxy_alpn` from `connectdata`.
Closes#17947
When a multiplex connection (h2/h3) is shutdown by the server, the
reported number of parallel transfers allowed drops to 0.
Determine that when the last transfer is done and terminate the
connection instead of keeping it in the cache.
We detect the drop to 0 also when we try to reuse such a connection, but
if we know this at the time the last transfer is done, we better
terminate it right away.
Have a consistent trace logging to this with the connections current
hostname and port. Adjust test expectations to carry port numbers.
Closes#17884
The `transport` to use for a transfer, e.g. TCP/QUIC/UNIX/UDP, is
initially selected by options and protocol used. This is set at the
`struct connectdata` as `transport` member.
During connection establishment, this transport may change due to
Alt-Svc or Happy-Eyeballing. Most common is the switch from TCP to QUIC.
Rename the connection member to `transport_wanted` and add a way to
query the connection for the transport in use via a new connection
filter query.
The filter query can also be used in the happy eyeballing attempts when
code needs to know which transport is used by the "filter below". This
happens in wolfssl initialization, as one example.
Closes#17923
Replace the old Curl_ssl_get_internals() with a new connection filter
query to retrieve the information. Implement that filter query for TCP
and QUIC TLS filter types.
Add tests in client tls_session_reuse to use the info option and check
that pointers are returned.
Reported-by: Larry Campbell
Fixes#17801Closes#17809
Change the signature of `bufq` functions from
* `ssize_t Curl_bufq_*(..., CURLcode *err)` to
* `CURLcode Curl_bufq_*(..., size_t *pn)`
This allows us to write slightly less code and avoids the ssize_t/size_t
conversions in many cases. Also, it gets the function in line with all
the other send/recv signatures.
Added helper functions in `cfilters.h` for sending from/receving into
a bufq.
Fuzzer now fails to build due to these changes and its testing of
the bufq API.
Closes#17396
Connection filters had a method `get_host()` which had not really been
documented. Since then, the cf had the `query()` method added. Replace
the separate get_host with query.
Add `CF_QUERY_HOST_PORT` as query to connection filters to retrieve
which remote hostname and port the filter (or its sub-filter) is talking
to. The query is implemented by HTTP and SOCKS filters, all others pass
it through.
Add `Curl_conn_get_current_host()` to retrieve the remote host and port
for a connection. During connect, this will return the host the
connection is talking to right now. Before/After connect, this will
return `conn->host.name`.
This is used by SASL authentication.
Closes#17419
Used to be a pointer set (and cleared) by the socket connection filters
to a struct in their contexts. Instead, add a filter query method to
obtain the pointer when needed.
Closes#17385
When multi creates the pollset of a transfer, it checks now if
a connection (FIRST/SECONDARY) socket waits on POLLIN and has input data
pending in filters (relevant to OpenSSL's new read ahead). If so, it
triggers a timeout on the transfer via EXPIRE_RUN_NOW.
This fixes sporadic stalls in test 988 when running event based.
Closes#17636
cfilter/conn: change send/recv function signatures. Unify the
calling/return conventions in our send/receive handling.
Curl_conn_recv(), adjust pnread type
Parameter `pnread` was a `ssize_t *`, but `size_t *` is better since the
function returns any error in its `CURLcode` return value.
Closes#17546
When ftp_done() is called to terminate the transfer, it needs to tear
down any open SECONDARY filter chain. The condition on when to do that
was relying on there to be a valid socket. This is not sufficient as the
socket is only set *after* happy eyeballing has decided on one.
Instead of checking for a valid conn->sock, check if any connection
filter is installed.
Fixes#17482
Reported-by: Rasmus Melchior Jacobsen
Closes#17491
Move curlx_ functions into its own subdir.
The idea is to use the curlx_ prefix proper on these functions, and use
these same function names both in tool, lib and test suite source code.
Stop the previous special #define setup for curlx_ names.
The printf defines are now done for the library alone. Tests no longer
use the printf defines. The tool code sets its own defines. The printf
functions are not curlx, they are publicly available.
The strcase defines are not curlx_ functions and should not be used by
tool or server code.
dynbuf, warnless, base64, strparse, timeval, timediff are now proper
curlx functions.
When libcurl is built statically, the functions from the library can be
used as-is. The key is then that the functions must work as-is, without
having to be recompiled for use in tool/tests. This avoids symbol
collisions - when libcurl is built statically, we use those functions
directly when building the tool/tests. When libcurl is shared, we
build/link them separately for the tool/tests.
Assisted-by: Jay Satiro
Closes#17253
Further testing with timeouts in event based processing revealed that
our current shutdown handling in the connection pool was not clear
enough. Graceful shutdowns can only happen inside a multi handle and it
was confusing to track in the code which situation actually applies. It
seems better to split the shutdown handling off and have that code
always be part of a multi handle.
Add `cshutdn.[ch]` with its own struct to maintain connections being
shut down. A `cshutdn` always belongs to a multi handle and uses that
for socket/timeout monitoring.
The `cpool`, which can be part of a multi or share, either passes
connections to a `cshutdn` or terminates them with a one-time, best
effort.
Add an `admin` easy handle to each multi and share. This is used to
perform all maintenance operations where no "real" easy handle is
available. This solves the problem that the multi admin handle requires
some additional initialisation (e.g. timeout list).
The share needs its admin handle as it is often cleaned up when no other
transfer or multi handle exists any more. But we need a `data` in almost
every call.
Fix file:// handling of errors when adding a new connection to the pool.
Changes in `curl` itself:
- for parallel transfers, do not set a connection pool in the share,
rely on the multi's connection pool instead. While not a requirement
for the new `cshutdn` to work, this is
a) helpful in testing to trigger graceful shutdowns
b) a broader code coverage of libcurl via the curl tool
- on test_event with uv, cleanup the multi handle before returning from
parallel_event(). The uv struct is on the stack, cleanup of the multi
later will crash when it tries to register sockets. This is a "eat
your own dogfood" related fix.
Closes#16508
Rework the event based handling of transfers and connections to
be "localized" into a single source file with clearer dependencies.
- add multi_ev.c and multi_ev.h
- add docs/internal/MULTI-EV.md to explain the overall workings
- only do event handling book keeping when the socket callback
is set
- add handling for "connection only" event tracking, when internal
easy handles are used that are not really tied to a connection.
Used in connection pool.
- remove transfer member "last_poll" and connections "shutdown_poll"
and keep all that internal to multi_ev.c
- add CURL_TRC_M() for tracing of "multi" related things, including
event handling and connection pool operations. Add new trace
feature "multi" for trace config.
multi traces will show exactly what is going on in regard to
event handling.
- multi: trace transfers "mstate" in every CURL_TRC_M() call
- make internal trace buffer 2048 bytes and end the silliness
with +n here -m there. Adjust test 1652 expectations of resulting
length and input edge cases.
- add trace feature "lib-ids" to perfix libcurl traces with transfer
and connection ids. Useful for debugging libcurl applications.
Closes#16308
Remove `blocking` argument from cfilter's connect method.
Implement blocking behaviour in Curl_conn_connect() instead for all
filter chains.
Update filters implementations. Several of which did never use the
paramter (QUIC for example). Simplifies connect handling in TLS filters
that no longer need to loop
Fixed a blocking connect call in FTP when waiting on a socket accept()
which only worked because the filter did not implement it.
Closes#16397
Around 2.3x speed-up parsing many large hexadecimal numbers. The decimal and
octal parser get marginally faster.
Still very readable, compact and easy to follow code.
Tweaks
- combine the max and the overflow check, gains 3ns/num (use a separate
check outside of the loop instead for max < base)
- one less indirection in the pointer, gains 3ns/num
- using the table lookup for hex nums, gains 5ns/num
- unfold the num_digit() macro, gains 3s/num
- use the hexasciitable unconditionally, gains 2ns/num
- use post-increment pointer in the table lookup, gains 1ns/num
- improved valid_digit() using the table for the hex case,
gains 26 ns/num
- use "max char" in valid_digit(), gains 3ns/num
Behavior changes:
- no longer returns STRE_TOO_BIG - only STRE_OVERFLOW
- does not move the char ** on error, which is probably better
Updated and extended test 1664 (significantly).
Closes#16374
Translate the `data->set.httpwant` which is one of the consts from the
public API (CURL_HTTP_VERSION_*) into a major version mask plus
additional flags for internal handling.
`Curl_http_neg_init()` does the translation and flags setting in http.c,
using new internal consts CURL_HTTP_V1x, CURL_HTTP_V2x and CURL_HTTP_V3x
for the major versions. The flags are
- only_10: when the application explicity asked fro HTTP/1.0
- h2_upgrade: when the application asks for upgrading 1.1 to 2.
- h2_prior_knowledge: when directly talking h2 without ALPN
- accept_09: when a HTTP/0.9 response is acceptable.
The Alt-Svc and HTTPS RR redirections from one ALPN to another obey the
allowed major versions. If a transfer has only h3 enabled, Alt-Svc
redirection to h2 is ignored.
This is the current implementation. It can be debated if Alt-Svc should
be able to override the allowed major versions. Added test_12_06 to
verify the current restriction.
Closes#16100
- add hex and octal parsers to the Curl_str_* family
- make curlx_strtoofft use these parsers
- remove all use of strtol() and strtoul() in library code
- generally use Curl_str_* more than strtoofft, for stricter parsing
- supports 64-bit universally, instead of 'long' which differs in size
between platforms
Extended the unit test 1664 to verify hex and octal parsing.
Closes#16336
The variable `conn->httpversion` was used for several purposes and it
was unclear at which time the value represents what.
- rename `conn->httpversion` to `conn->httpversion_seen`
This makes clear that the variable only records the last
HTTP version seen on the connection - if any. And that it
no longer is an indication of what version to use.
- Change Alt-Svc handling to no longer modify `conn->httpversion`
but set `data->state.httpwant` for influencing the HTTP version
to use on a transfer.
- Add `data->req.httpversion_sent` to have a record of what
HTTP version was sent in a request
- Add connection filter type CF_TYPE_HTTP
- Add filter query `CF_QUERY_HTTP_VERSION` to ask what HTTP
filter version is in place
- Lookup filters HTTP version instead of using `conn->httpversion`
Test test_12_05 now switches to HTTP/1.1 correctly and the
expectations have been fixed.
Removed the connection fitler "is_httpN()" checks and using
the version query instead.
Closes#16073
Make transfer attach/detach to/from connections chepaer.
- the "attach" event was no longer implemented by any filter
- the "detach" did the same as the "done" event for the filters
who still implemented it. It should be superfluous as the "done"
must always happen.
Closes#16067
booleans should use the type 'bool' and set the value to TRUE/FALSE
non-booleans should not be 'bool' and should not set the value to
TRUE/FALSE
Closes#15123
When ending an FTP upload, we shut down the connection gracefully, since
the server should be notified we had send all bytes. Mostly, this is a
NOP without TLS involved. With TLS, close-notify messages should be
exchanged.
As reported in #14843, not all servers seem to do that. Since it is the
server's responsiblity to check it has received everything, we just log
the timeout and proceed as if everything is fine.
In the receive direction, we still fail the transfer if the server does
not shut down its direction properly.
Fixes#14843
Reported-by: Rasmus Melchior Jacobsen
Closes#14848
Output the 'Connected to...' info message when the connection has been
fully established and all information is available.
Due to our happy eyeballing, we should not emit info messages in
filters, because they may be part of an eyeballing attempt and may be
discarded later for another chain.
Closes#14897
Update IP related information at the connection and the transfer in two
places only: once the filter chain connects and when a transfer is added
to a connection. The latter only updates on reuse when the filters
already are connected.
The only user of that information before a full connect is the HAProxy
filter. Add cfilter CF_QUERY_IP_INFO query to let it find the
information from the filters "below".
This solves two issues with the previous version:
- updates where often done twice with the same info
- happy eyeballing filter "forks" could overwrite each others
updates before the full winner was determined.
Closes#14699
- replace the counting of upload lengths with the new eos send flag
- improve frequency of stream draining to happen less on events where it
is not needed
- this PR is based on #14220
http2, cf-h2-proxy: fix EAGAINed out buffer
- in adjust pollset and shutdown handling, a non-empty `ctx->outbufq`
must trigger send polling, irregardless of http/2 flow control
- in http2, fix retry handling of blocked GOAWAY frame
test case improvement:
- let client 'upload-pausing' handle http versions
Closes#14253
Since data can be held in connection filter buffers when sending gives
EAGAIN, add methods to query this and perform flushing of those buffers.
The transfer loop will continue sending until all upload data is
processed and the connection is flushed.
- add `CF_QUERY_SEND_PENDING` to query filters
- add `CF_CTRL_DATA_SEND_FLUSH` to flush filters
- change `Curl_req_want_send()` to query the connection
if it needs flushing
- use `Curl_req_want_send()` to determine the POLLOUT
in the PERFORMING multi state
- implement flush handling in the HTTP/2 connection filter
Closes#14271
Adds a `bool eos` flag to send methods to indicate that the data
is the last chunk the invovled transfer wants to send to the server.
This will help protocol filters like HTTP/2 and 3 to forward the
stream's EOF flag and also allow to EAGAIN such calls when buffers
are not yet fully flushed.
Closes#14220
Adds a `bool eos` flag to send methods to indicate that the data is the
last chunk the invovled transfer wants to send to the server.
This will help protocol filters like HTTP/2 and 3 to forward the
stream's EOF flag and also allow to EAGAIN such calls when buffers are
not yet fully flushed.
Closes#14220
When libcurl discards a connection there are two phases this may go
through: "shutdown" and "closing". If a connection is aborted, the
shutdown phase is skipped and it is closed right away.
The connection filters attached to the connection implement the phases
in their `do_shutdown()` and `do_close()` callbacks. Filters carry now a
`shutdown` flags next to `connected` to keep track of the shutdown
operation.
Filters are shut down from top to bottom. If a filter is not connected,
its shutdown is skipped. Notable filters that *do* something during
shutdown are HTTP/2 and TLS. HTTP/2 sends the GOAWAY frame. TLS sends
its close notify and expects to receive a close notify from the server.
As sends and receives may EAGAIN on the network, a shutdown is often not
successful right away and needs to poll the connection's socket(s). To
facilitate this, such connections are placed on a new shutdown list
inside the connection cache.
Since managing this list requires the cooperation of a multi handle,
only the connection cache belonging to a multi handle is used. If a
connection was in another cache when being discarded, it is removed
there and added to the multi's cache. If no multi handle is available at
that time, the connection is shutdown and closed in a one-time,
best-effort attempt.
When a multi handle is destroyed, all connection still on the shutdown
list are discarded with a final shutdown attempt and close. In curl
debug builds, the environment variable `CURL_GRACEFUL_SHUTDOWN` can be
set to make this graceful with a timeout in milliseconds given by the
variable.
The shutdown list is limited to the max number of connections configured
for a multi cache. Set via CURLMOPT_MAX_TOTAL_CONNECTIONS. When the
limit is reached, the oldest connection on the shutdown list is
discarded.
- In multi_wait() and multi_waitfds(), collect all connection caches
involved (each transfer might carry its own) into a temporary list.
Let each connection cache on the list contribute sockets and
POLLIN/OUT events it's connections are waiting for.
- in multi_perform() collect the connection caches the same way and let
them peform their maintenance. This will make another non-blocking
attempt to shutdown all connections on its shutdown list.
- for event based multis (multi->socket_cb set), add the sockets and
their poll events via the callback. When `multi_socket()` is invoked
for a socket not known by an active transfer, forward this to the
multi's cache for processing. On closing a connection, remove its
socket(s) via the callback.
TLS connection filters MUST NOT send close nofity messages in their
`do_close()` implementation. The reason is that a TLS close notify
signals a success. When a connection is aborted and skips its shutdown
phase, the server needs to see a missing close notify to detect
something has gone wrong.
A graceful shutdown of FTP's data connection is performed implicitly
before regarding the upload/download as complete and continuing on the
control connection. For FTP without TLS, there is just the socket close
happening. But with TLS, the sent/received close notify signals that the
transfer is complete and healthy. Servers like `vsftpd` verify that and
reject uploads without a TLS close notify.
- added test_19_* for shutdown related tests
- test_19_01 and test_19_02 test for TCP RST packets
which happen without a graceful shutdown and should
no longer appear otherwise.
- add test_19_03 for handling shutdowns by the server
- add test_19_04 for handling shutdowns by curl
- add test_19_05 for event based shutdowny by server
- add test_30_06/07 and test_31_06/07 for shutdown checks
on FTP up- and downloads.
Closes#13976
- clarify Curl_xfer_setup() with RECV/SEND flags and different calls for
which socket they operate on. Add a shutdown flag for secondary
sockets
- change Curl_xfer_setup() calls to new functions
- implement non-blocking connection shutdown at the end of receiving or
sending a transfer
Closes#13913
This adds connection shutdown infrastructure and first use for FTP. FTP
data connections, when not encountering an error, are now shut down in a
blocking way with a 2sec timeout.
- add cfilter `Curl_cft_shutdown` callback
- keep a shutdown start timestamp and timeout at connectdata
- provide shutdown timeout default and member in
`data->set.shutdowntimeout`.
- provide methods for starting, interrogating and clearing
shutdown timers
- provide `Curl_conn_shutdown_blocking()` to shutdown the
`sockindex` filter chain in a blocking way. Use that in FTP.
- add `Curl_conn_cf_poll()` to wait for socket events during
shutdown of a connection filter chain.
This gets the monitoring sockets and events via the filters
"adjust_pollset()" methods. This gives correct behaviour when
shutting down a TLS connection through a HTTP/2 proxy.
- Implement shutdown for all socket filters
- for HTTP/2 and h2 proxying to send GOAWAY
- for TLS backends to the best of their capabilities
- for tcp socket filter to make a final, nonblocking
receive to avoid unwanted RST states
- add shutdown forwarding to happy eyeballers and
https connect ballers when applicable.
Closes#13904
`CURLDEBUG` is meant to enable memory tracking, but in a bunch of cases,
it was protecting debug features that were supposed to be guarded with
`DEBUGBUILD`.
Replace these uses with `DEBUGBUILD`.
This leaves `CURLDEBUG` uses solely for its intended purpose: to enable
the memory tracking debug feature.
Also:
- autotools: rely on `DEBUGBUILD` to enable `checksrc`.
Instead of `CURLDEBUG`, which worked in most cases because debug
builds enable `CURLDEBUG` by default, but it's not accurate.
- include `lib/easyif.h` instead of keeping a copy of a declaration.
- add CI test jobs for the build issues discovered.
Ref: https://github.com/curl/curl/pull/13694#issuecomment-2120311894Closes#13718
- add `Curl_hash_offt` as hashmap between a `curl_off_t` and
an object. Use this in h2+h3 connection filters to associate
`data->id` with the internal stream state.
- changed implementations of all affected connection filters
- removed `h2_ctx*` and `h3_ctx*` from `struct HTTP` and thus
the easy handle
- solves the problem of attaching "foreign protocol" easy handles
during connection shutdown
Test 1616 verifies the new hash functions.
Closes#13204
new struct ip_quadruple for holding local/remote addr+port
- used in data->info and conn and cf-socket.c
- copy back and forth complete struct
- add 'secondary' to conn
- use secondary in reporting success for ftp 2nd connection
Reported-by: DasKutti on github
Fixes#13084Closes#13090
- replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to
clarify when and at what level they operate
- send/recv of transfer related data is now done via
`Curl_xfer_send()/Curl_xfer_recv()` which no longer has
socket/socketindex as parameter. It decides on the transfer
setup of `conn->sockfd` and `conn->writesockfd` on which
connection filter chain to operate.
- send/recv on a specific connection filter chain is done via
`Curl_conn_send()/Curl_conn_recv()` which get the socket index
as parameter.
- rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for
naming consistency
- clarify that the special CURLE_AGAIN hangling to return
`CURLE_OK` with length 0 only applies to `Curl_xfer_send()`
and CURLE_AGAIN is returned by all other send() variants.
- fix a bug in websocket `curl_ws_recv()` that mixed up data
when it arrived in more than a single chunk
The method for sending not just raw bytes, but bytes that are either
"headers" or "body". The send abstraction stack, to to bottom, now is:
* `Curl_req_send()`: has parameter to indicate amount of header bytes,
buffers all data.
* `Curl_xfer_send()`: knows on which socket index to send, returns
amount of bytes sent.
* `Curl_conn_send()`: called with socket index, returns amount of bytes
sent.
In addition there is `Curl_req_flush()` for writing out all buffered
bytes.
`Curl_req_send()` is active for requests without body,
`Curl_buffer_send()` still being used for others. This is because the
special quirks need to be addressed in future parts:
* `expect-100` handling
* `Curl_fillreadbuffer()` needs to add directly to the new
`data->req.sendbuf`
* special body handlings, like `chunked` encodings and line end
conversions will be moved into something like a Client Reader.
In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`,
replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative
values as the returned `CURLcode` already specifies error conditions. This
allows easier handling of lengths without casting.
Closes#12964
Curl_read/Curl_write clarifications
- replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to 1clarify
when and at what level they operate
- send/recv of transfer related data is now done via
`Curl_xfer_send()/Curl_xfer_recv()` which no longer has
socket/socketindex as parameter. It decides on the transfer setup of
`conn->sockfd` and `conn->writesockfd` on which connection filter
chain to operate.
- send/recv on a specific connection filter chain is done via
`Curl_conn_send()/Curl_conn_recv()` which get the socket index as
parameter.
- rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming
consistency
- clarify that the special CURLE_AGAIN handling to return `CURLE_OK`
with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is
returned by all other send() variants.
SingleRequest reshuffling
- move functions into request.[ch]
- differentiate between reset and free
- add Curl_req_done() to perform last actions
- add a send `bufq` to SingleRequest for future use in keeping upload data
Closes#12963
- let `multi_getsock()` initialize the pollset in what the
transfer state requires in regards to SEND/RECV
- change connection filters `adjust_pollset()` implementation
to react on the presence of POLLIN/-OUT in the pollset and
no longer check CURL_WANT_SEND/CURL_WANT_RECV
- cf-socket will no longer add POLLIN on its own
- http2 and http/3 filters will only do adjustments if the
passed pollset wants to POLLIN/OUT for the transfer on
the socket. This is similar to the HTTP/2 proxy filter
and works in stacked filters.
Closes#12640
GCC 14 introduces a new -Walloc-size included in -Wextra which gives:
```
src/tool_operate.c: In function ‘add_per_transfer’:
src/tool_operate.c:213:5: warning: allocation of insufficient size ‘1’ for type ‘struct per_transfer’ with size ‘480’ [-Walloc-size]
213 | p = calloc(sizeof(struct per_transfer), 1);
| ^
src/var.c: In function ‘addvariable’:
src/var.c:361:5: warning: allocation of insufficient size ‘1’ for type ‘struct var’ with size ‘32’ [-Walloc-size]
361 | p = calloc(sizeof(struct var), 1);
| ^
```
The calloc prototype is:
```
void *calloc(size_t nmemb, size_t size);
```
So, just swap the number of members and size arguments to match the
prototype, as we're initialising 1 struct of size `sizeof(struct
...)`. GCC then sees we're not doing anything wrong.
Closes#12292
- fixed libssh.c workaround for a socket being closed by
the library
- eliminate the terrible hack in cf-socket.c to guess when
this happened and try not closing the socket again.
- fixes race in eyeballing when socket could have failed to
be closed for a discarded connect attempt
Closes#12207
Connection filter had a `get_select_socks()` method, inspired by the
various `getsocks` functions involved during the lifetime of a
transfer. These, depending on transfer state (CONNECT/DO/DONE/ etc.),
return sockets to monitor and flag if this shall be done for POLLIN
and/or POLLOUT.
Due to this design, sockets and flags could only be added, not
removed. This led to problems in filters like HTTP/2 where flow control
prohibits the sending of data until the peer increases the flow
window. The general transfer loop wants to write, adds POLLOUT, the
socket is writeable but no data can be written.
This leads to cpu busy loops. To prevent that, HTTP/2 did set the
`SEND_HOLD` flag of such a blocked transfer, so the transfer loop cedes
further attempts. This works if only one such filter is involved. If a
HTTP/2 transfer goes through a HTTP/2 proxy, two filters are
setting/clearing this flag and may step on each other's toes.
Connection filters `get_select_socks()` is replaced by
`adjust_pollset()`. They get passed a `struct easy_pollset` that keeps
up to `MAX_SOCKSPEREASYHANDLE` sockets and their `POLLIN|POLLOUT`
flags. This struct is initialized in `multi_getsock()` by calling the
various `getsocks()` implementations based on transfer state, as before.
After protocol handlers/transfer loop have set the sockets and flags
they want, the `easy_pollset` is *always* passed to the filters. Filters
"higher" in the chain are called first, starting at the first
not-yet-connection one. Each filter may add sockets and/or change
flags. When all flags are removed, the socket itself is removed from the
pollset.
Example:
* transfer wants to send, adds POLLOUT
* http/2 filter has a flow control block, removes POLLOUT and adds
POLLIN (it is waiting on a WINDOW_UPDATE from the server)
* TLS filter is connected and changes nothing
* h2-proxy filter also has a flow control block on its tunnel stream,
removes POLLOUT and adds POLLIN also.
* socket filter is connected and changes nothing
* The resulting pollset is then mixed together with all other transfers
and their pollsets, just as before.
Use of `SEND_HOLD` is no longer necessary in the filters.
All filters are adapted for the changed method. The handling in
`multi.c` has been adjusted, but its state handling the the protocol
handlers' `getsocks` method are untouched.
The most affected filters are http/2, ngtcp2, quiche and h2-proxy. TLS
filters needed to be adjusted for the connecting handshake read/write
handling.
No noticeable difference in performance was detected in local scorecard
runs.
Closes#11833
