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
After fixing support for x32, unlock eventfd support for all CPUs.
Before this patch, it was explicitly limited to 64-bit ones.
You can disable eventfs manually on systems where it's auto-detected:
- cmake: `-DHAVE_EVENTFD=0`
- configure: `export ac_cv_func_eventfd=0`
Ref: c2aa504ab9#16239Closes#16277
Add a 'wanted' major HTTP version bitmask next to the 'allowed' bitmask
in HTTP version negotiation. This will try connections as specified in
'wanted', but enabled Alt-Svc and HTTPS-RR to redirect to other major
HTTP versions, if those are 'allowed'.
Changes libcurl internal default to `CURL_HTTP_VERSION_NONE` and removes
the code in curl that sets `CURL_HTTP_VERSION_2TLS` if the command line
does not say anything else.
Closes#16117
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
The eventfd manpage says:
A write(2) fails with the error EINVAL if the size of the supplied
buffer is less than 8 bytes
When doing x32 on a 64-bit system, pointers are still four bytes so this
code must not use the size of a pointer but the size of a 64-bit type.
Fixes#16237
Reported-by: Jan Engelhardt
Closes#16239
By better sticking to listing the struct members sorted by size, this
struct is now 48 bytes smaller on my fairly maximized build, without
removing anything.
Turned 'connect_only' into two bits instead of an unsigned char with two
magic values.
Also put the 'gssapi_delegation' field within ifdef HAVE_GSSAPI.
Closes#16097
Adds a `follow()` callback to protocol handlers, so they may decide how
to act on a `newurl` after a request has been done. This is optional.
This moves the HTTP code for handling redirects from multi.c to http.c
where it should be. If we ever add a protocol with its own logic, it
would install its own follow function.
Closes#16075
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
The TLS session cache is now held by the multi handle unless it is
shared, so that all easy handles within a multi handle get the benefit
of sharing the same, larger, cache.
The multi handle session cache size is set to 25, unless it is the
internal one used for the easy interface - which still uses only 3.
Closes#15982
Count connections to a host against a possibly configured destination
limit. Trigger multi `connchange` when a connection has been shutdown,
so pending transfers can try to get a connection once again.
Reported-by: baranyaib90 on github
Fixes#15857Closes#15879
- Ensure that CURLM_OK is returned when curl_multi_remove_handle is
called with an already removed easy handle.
Prior to this change and since ba235ab2 which precedes 8.10.0, if
curl_multi_remove_handle was called with an already-removed easy handle
then the return code would be CURLM_OK or CURLM_BAD_EASY_HANDLE
depending respectively on whether the multi did or did not contain other
easy handles.
This change restores the old behavior of returning CURLM_OK in both
cases.
Reported-by: Ralph Sennhauser
Fixes https://github.com/curl/curl/issues/15844
Closes https://github.com/curl/curl/pull/15852
- Make curl_multi_waitfds consistent with the documentation.
Issue Addressed:
- The documentation of curl_multi_waitfds indicates that users should
be able to call curl_multi_waitfds with a NULL ufds. However, before
this change, the function would return CURLM_BAD_FUNCTION_ARGUMENT.
- Additionally, the documentation suggests that users can use this
function to determine the number of file descriptors (fds) needed.
However, the function would stop counting fds if the supplied fds
were exhausted.
Changes Made:
- NULL ufds Handling: curl_multi_waitfds can now accept a NULL ufds if
size is also zero.
- Counting File Descriptors: If curl_multi_waitfds is passed a NULL
ufds, or the size of ufds is insufficient, the output parameter
fd_count will return the number of fds needed. This value may be
higher than actually needed but never lower.
Testing:
- Test 2405 has been updated to cover the usage scenarios described
above.
Fixes https://github.com/curl/curl/issues/15146
Closes https://github.com/curl/curl/pull/15155
`Makefile.mk` supported MS-DOS and Amiga, but `./configure` also
supported them in a better tested and more flexible way.
This patch also adds CMake support for MS-DOS/DJGPP and Amiga OS 3.
`Makefile.mk` was not maintained. Delete it in favour of first-tier
build methods.
Also include some non-MS-DOS/AmigaOS-specific tidy-up, see details at
the end of this message.
Details:
- fix/silence all MS-DOS/DJGPP build warnings and issues.
- add MS-DOS support to cmake.
- default to `ENABLE_THREADED_RESOLVER=OFF` for MS-DOS.
- add support for `WATT_ROOT`.
- use static libcurl with MS-DOS.
- fixup default CMake suffixes/prefixes for DJGPP.
- disable hidden symbols for MS-DOS. Not supported on MS-DOS.
- opt-in MS-DOS into `USE_UNIX_SOCKETS`.
- improve MS-DOS support in autotools.
- default to `--disable-threaded-resolver` for MS-DOS.
- make sure to use `close_s()` (from Watt-32) with autotools and cmake.
`Makefile.mk` used it before this patch.
- GHA: add DJGPP cmake (~30s) and autotools (~60s) build jobs.
Also build tests and examples with cmake.
- improve AmigaOS support in autotools:
- configure: detect `CloseSocket()` when it's a macro.
- configure: fix `IoctlSocket` detection on AmigaOS.
- curl-amissl.m4: pass AmiSSL libs to tests/servers.
- add AmigaOS3 support to cmake:
- cmake: fix `HAVE_IOCTLSOCKET_CAMEL` and
`HAVE_IOCTLSOCKET_CAMEL_FIONBIO` detections.
- set necessary system libs.
- add AmiSSL support.
- inet_ntop, inet_pton: fix using it for AmigaOS. cmake detects them,
and they did not compile with AmigaOS.
- cmake: better sync `gethostname` detection with autotools.
Fixes detection for AmigaOS, where `gethostname` is a macro.
- cmake: fix `sys/utime.h` detection on AmigaOS.
- cmake: force-disable `getaddrinfo` for AmigaOS.
- cmake: tweak threading and static/shared default for AmigaOS.
- cmake: rely on manual variable `AMIGA` to enable the platform.
- GHA: add AmigaOS cmake and autotools (~45s) jobs.
Also build tests and examples with cmake.
- INSTALL: update MS-DOS and AmigaOS build instructions.
- amigaos: fix `-Wpointer-sign` and
`zero or negative size array '_args'` in `Printf()`.
- amigaos: fix `-Wpointer-sign`
- amigaos: fix `-Wredundant-decls` `errno` and `h_errno`.
- amigaos: brute-force silence `lseek()` size warnings.
- amigaos: server/resolve: silence `-Wdiscarded-qualifiers`.
- amigaos: server/resolve: fix `-Wpointer-sign`.
- amigaos: fix `CURL_SA_FAMILY_T` type.
- nonblock: prefer `HAVE_IOCTLSOCKET_CAMEL_FIONBIO` for AmigaOS.
`ioctl` is also detected, but fails when used. Make the above override
it for a successful build.
Authored-by: Darren Banfi
Fixes#15537Closes#15603
- tftpd: prefer `HAVE_IOCTLSOCKET_CAMEL_FIONBIO` for AmigaOS.
- tftpd: tidy-up conditional code.
- curl: set stack size to 16384 for AmigaOS3/4
Overriding the default 4096.
Suggested-by: Darren Banfi
Ref: https://github.com/curl/curl/pull/15543#issuecomment-2498783123
Ref: https://wiki.amigaos.net/wiki/Controlling_Application_Stack
- functypes.h: fix `SEND_QUAL_ARG2` for AmigaOS.
- tftp: add missing cast in sendto() call for AmigaOS.
- getinfo: fix warning with AmigaOS.
- tool_operate: silence warning with AmigaOS
- amigaos: fix building libtests due to missing `RLIMIT_NOFILE`.
- curl_gethostname: silence warning for AmigaOS.
- ftp: silence `-Wtype-limits` for AmigaOS.
- libtest: fix timeval initialization for AmigaOS.
- examples: fix `timeval` initialization for AmigaOS.
- examples: silence warning for AmigaOS.
- configure: fix IPv6 detection for cross-builds.
- netrc: fix to build with AmigaOS cleanly.
- buildinfo: detect and add `DOS` tag for MS-DOS builds.
- buildinfo: add `AMIGA` to buildinfo.txt in auttools.
- build: move `USE_WATT32` macro definition to cmake/configure.
Non-MS-DOS/AmigeOS-specific tidy-ups:
- configure: sync `sa_family_t` detection with cmake.
- configure: sync `ADDRESS_FAMILY` detection signals with cmake.
- doh: use `CURL_SA_FAMILY_T`.
- lib: drop mingw-specific `CURL_SA_FAMILY_T` workaround.
- cmake: extend instead of override check-specific
configurations/requirements.
This allows to honor global requirements added earlier.
Necessary for AmigaOS for example.
- cmake: omit warning on disabled IPv6 for MS-DOS and AmigaOS.
No IPv6 support on these platforms. Also sync with autotools.
- lib1960: use libcurl `inet_pton()` wrapper.
- cmake: detect LibreSSL (to match autotools).
- cmake: say the specific OpenSSL flavour detected.
- hostip: add missing `HAVE_SOCKADDR_IN6_SIN6_SCOPE_ID` guard.
- lib: simplify classic mac feature guards.
Follow-up to a8861b6ccd#9764Closes#15543
Issue is reproducible for me if I have made request with multi handle,
then I make request that will take very long and then I make request
that should be fast again, however what happens it is that it seems
to think that timeout was not changed and it makes it not call initial
`CURLMOPT_TIMERFUNCTION`.
Closes#15627
Introduce five functions named after the state they serve:
- state_connect for MSTATE_CONNECT
- state_do for MSTATE_DO
- state_performing for MSTATE_PERFORMING
- state_ratelimiting for MSTATE_RATELIMITING
- state_resolving for MSTATE_RESOLVING
Closes#15418
It makes the callbacks get different signnatures when used from within
libcurl vs outside of it by libcurl-using applications (such as the
libtests) and this triggers UndefinedBehaviorSanitizer errors.
Closes#15289
When a multi handle is being cleaned up, it can still cause user
callbacks to be fired. Notably Curl_cpool_destroy calls socket_callback
on all pooled connections. It's still possible for the callback to call
curl_multi_assign leading to an assert.
This commit moves clearing of a multi handle magic to a point where the
multi handle stops being a valid object.
Fixes#15201Closes#15206
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
For all states before MSTATE_DO the connect timeout needs to be
considered.
Regression since #13371 (be659030ba) shipped in 8.8.0
Reported-by: Deniz Sökmen
Fixes#15100Closes#15119
The proper alignment is not guaranteed. This function now instead uses
only the first and last byte of the key since they are the ones likely
to change most (one of them, depending on CPU endian) and the hash is
tiny anyway.
Closes#15063
Warn/assert about a possibly stalling transfer only when it
has no timeout set.
The assertion triggered in test 1540 on loaded CI sometimes.
Closes#15028
With higher parallelism in CI, the ASSERT triggered on pause tests.
Strengthen the check. We might want to think about removing
KEEP_RECV_PAUSE|KEEP_SEND_PAUSE altogether.
Closes#14981
By requiring that the multi handle is fine, it can detect bad usage
better and by that avoid crashes. Like in the #14860 case, which is an
application calling curl_multi_assign() with a NULL pointer multi
handle.
Reported-by: Carlo Cabrera
Fixes#14860Closes#14862
This is a better match for what they do and the general "cpool"
var/function prefix works well.
The pool now handles very long hostnames correctly.
The following changes have been made:
* 'struct connectdata', e.g. connections, keep new members
named `destination` and ' destination_len' that fully specifies
interface+port+hostname of where the connection is going to.
This is used in the pool for "bundling" of connections with
the same destination. There is no limit on the length any more.
* Locking: all locks are done inside conncache.c when calling
into the pool and released on return. This eliminates hazards
of the callers keeping track.
* 'struct connectbundle' is now internal to the pool. It is no
longer referenced by a connection.
* 'bundle->multiuse' no longer exists. HTTP/2 and 3 and TLS filters
no longer need to set it. Instead, the multi checks on leaving
MSTATE_CONNECT or MSTATE_CONNECTING if the connection is now
multiplexed and new, e.g. not conn->bits.reuse. In that case
the processing of pending handles is triggered.
* The pool's init is provided with a callback to invoke on all
connections being discarded. This allows the cleanups in
`Curl_disconnect` to run, wherever it is decided to retire
a connection.
* Several pool operations can now be fully done with one call.
Pruning dead connections, upkeep and checks on pool limits
can now directly discard connections and need no longer return
those to the caller for doing that (as we have now the callback
described above).
* Finding a connection for reuse is now done via `Curl_cpool_find()`
and the caller provides callbacks to evaluate the connection
candidates.
* The 'Curl_cpool_check_limits()' now directly uses the max values
that may be set in the transfer's multi. No need to pass them
around. Curl_multi_max_host_connections() and
Curl_multi_max_total_connections() are gone.
* Add method 'Curl_node_llist()' to get the llist a node is in.
Used in cpool to verify connection are indeed in the list (or
not in any list) as they need to.
I left the conncache.[ch] as is for now and also did not touch the
documentation. If we update that outside the feature window, we can
do this in a separate PR.
Multi-thread safety is not achieved by this PR, but since more details
on how pools operate are now "internal" it is a better starting
point to go for this in the future.
Closes#14662
- Renames Curl_readwrite() to Curl_sendrecv() to reflect that it
is mainly about talking to the server, not reads or writes to the
client. Add a `nowp` parameter since the single caller already
has this.
- Curl_sendrecv() now runs all possible operations whenever it is
called and either it had been polling sockets or the 'select_bits'
are set.
POLL_IN/POLL_OUT are not always directly related to send/recv
operations. Filters like HTTP/2, QUIC or TLS may monitor reverse
directions. If a transfer does not want to send (KEEP_SEND), it
will not do so, as before. Same for receives.
- Curl_update_timer() now checks the absolute timestamp of an expiry
and the last/new timeout to determine if the application needs
to stop/start/restart its timer. This fixes edge cases where
updates did not happen as they should have.
- improved --test-event curl_easy_perform() simulation to handle
situations where no sockets are registered but a timeout is
in place.
- fixed bug in events_socket() that complained about removing
a socket that was unknown, when indeed it had removed the socket
just before, only it was the last in the list
- fixed conncache's internal handle to carry the multi instance
(where the cache has one) so that operations on the closure handle
trigger event callbacks correctly.
- fixed conncache to not POLL_REMOVE a socket twice when a conneciton
was closed.
Closes#14561
- add set/get functions for the custom data in a tree node
- use Curl_timediff for time comparisons instead of the custom macro, as they
do the exact same things.
- add asserts to catch mistakes better
- updated test 1309 accordingly
Closes#14562
Previously an error from the callback accidentally made libcurl keep the
hash entry which would lead to the entry remaining and then potentially
getting removed *again* which would lead to internal confusions.
This is an old issue (introduced in 2b3dd01b77), caught by the new
asserts from c0233a35da.
Closes#14557
- converted the Curl_hash_count() macro to a function
- Discourage accessing struct fields directly
- Document the internal API in HASH.md
Closes#14503
`data->id` is unique in *most* situations, but not in all. If a libcurl
application uses more than one connection cache, they will overlap. This
is a rare situations, but libcurl apps do crazy things. However, for
informative things, like tracing, `data->id` is superior, since it
assigns new ids in curl's serial curl_easy_perform() use.
Introduce `data->mid` which is a unique identifer inside one multi
instance, assigned on multi_add_handle() and cleared on
multi_remove_handle().
Use the `mid` in DoH operations and also in h2/h3 stream hashes.
Reported-by: 罗朝辉
Fixes#14414Closes#14499
- Turned them all into functions to also do asserts etc.
- The llist related structs got all their fields renamed in order to make
sure no existing code remains using direct access.
- Each list node struct now points back to the list it "lives in", so
Curl_node_remove() no longer needs the list pointer.
- Rename the node struct and some of the access functions.
- Added lots of ASSERTs to verify API being used correctly
- Fix some cases of API misuse
Add docs/LLIST.md documenting the internal linked list API.
Closes#14485
Instead of having an especially "unique" linked list handler for the
main list of easy handles within the multi handle, this now uses a
regular Curl_llist for this as well.
With this change, it is also clearer that every easy handle added to a
multi handle belongs to one and only one out of three different lists:
process - the general one for normal transfer processing
pending - queued up waiting to get a connection (MSTATE_PENDING)
msgsent - transfer completed (MSTATE_MSGSENT)
An easy handle must therefore be removed from the current list before it
gets added to another.
Closes#14474
Use these words and casing more consistently across text, comments and
one curl tool output:
AIX, ALPN, ANSI, BSD, Cygwin, Darwin, FreeBSD, GitHub, HP-UX, Linux,
macOS, MS-DOS, MSYS, MinGW, NTLM, POSIX, Solaris, UNIX, Unix, Unicode,
WINE, WebDAV, Win32, winbind, WinIDN, Windows, Windows CE, Winsock.
Mostly OS names and a few more.
Also a couple of other minor text fixups.
Closes#14360
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
Replace Curl_resolv_unlock() with Curl_resolv_unlink():
-replace inuse member with refcount in Curl_dns_entry
- pass Curl_dns_entry ** to unlink, so it gets always cleared
- solve potential (but unlikley) UAF in FTP's handling of looked up
Curl_dns_entry. Esp. do not use addr information after unlinking an entry.
In reality, the unlink will not free memory, as the dns entry is still
referenced by the hostcache. But this is not safe and relying on no other
code pruning the cache in the meantime.
- pass permanent flag when adding a dns entry instead of fixing timestamp
afterwards.
url.c: fold several static *resolve_* functions into one.
Closes#14195
- multi.c: when ratelimiting a transfer stops (MSTATE_RATELIMITING ->
MSTATE_PERFORMING), run the MSTATE_PERFORMING state right away
- urldata.h: factor out upload and download progress counters into a
struct, use that for passing these to progress update functions
- progress.c/getinfo.c: change names of moved progress counters
- progress.c: use new structs and a helper struct to factor repeated
calculation into static helpers
Closes#14335
