If curl_url() returns NULL this should return early to avoid mistakes -
even if right now the subsequent function invokes are all OK.
Coverity (wrongly) pointed out this as a NULL deref.
Closes#8100
While uploading, check for remote window availability in the getsock
function so that we don't wait for a writable socket if no data can be
sent.
Reported-by: Steini2000 on github
Fixes#7821Closes#7839
Regression from 3cb8a74867 (releasde in 7.79.0). That change moved
transfer oriented inits to before the check but also erroneously moved a
few connection oriented ones, which causes problems.
Reported-by: Evangelos Foutras
Fixes#7730Closes#7731
Thus function was written to avoid doing multiple connection data
initializations, which is fine, but since it also initiates stream
related data it is crucial that it doesn't skip those even if called
again for the same connection. Solved by moving the stream
initializations before the "doing-it-again" check.
Reported-by: Inho Oh
Fixes#7630Closes#7692
- the data needs to be "line-based" anyway since it's also passed to the
debug callback/application
- it makes infof() work like failf() and consistency is good
- there's an assert that triggers on newlines in the format string
- Also removes a few instances of "..."
- Removes the code that would append "..." to the end of the data *iff*
it was truncated in infof()
Closes#7357
The libssh2 backend has SSH session associated with the connection but
the callback context is the easy handle, so when a connection gets
attached to a transfer, the protocol handler now allows for a custom
function to get used to set things up correctly.
Reported-by: Michael O'Farrell
Fixes#6898Closes#7078
Since the function is called for any protocol, we can't assume that the
HTTP struct is there without first making sure it is HTTP.
Reported-by: Denis Goleshchikhin
Fixes#7079Closes#7080
This is considered not harmful as a following http2_recv shall be
called very soon.
This is considered helpful in the specific situation where some
servers (e.g. nghttpx v1.43.0) may fulfill stream 1 immediately
following the return of HTTP status 101, other than waiting for
the client-side connection preface to arrive.
Fixes#7036Closes#7040
... detected by Coverity:
Error: RESOURCE_LEAK (CWE-772):
lib/http2.c:532: alloc_fn: Storage is returned from allocation function "duphandle".
lib/http2.c:532: var_assign: Assigning: "newhandle" = storage returned from "duphandle(data)".
lib/http2.c:552: noescape: Resource "newhandle" is not freed or pointed-to in "set_transfer_url".
lib/http2.c:555: leaked_storage: Variable "newhandle" going out of scope leaks the storage it points to.
Closes#6986
... detected by Coverity:
Error: RESOURCE_LEAK (CWE-772):
lib/http2.c:480: alloc_fn: Storage is returned from allocation function "curl_url". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:480: var_assign: Assigning: "u" = storage returned from "curl_url()".
lib/http2.c:486: noescape: Resource "u" is not freed or pointed-to in "curl_url_set". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:488: leaked_storage: Variable "u" going out of scope leaks the storage it points to.
Error: RESOURCE_LEAK (CWE-772):
lib/http2.c:480: alloc_fn: Storage is returned from allocation function "curl_url". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:480: var_assign: Assigning: "u" = storage returned from "curl_url()".
lib/http2.c:493: noescape: Resource "u" is not freed or pointed-to in "curl_url_set". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:495: leaked_storage: Variable "u" going out of scope leaks the storage it points to.
Error: RESOURCE_LEAK (CWE-772):
lib/http2.c:480: alloc_fn: Storage is returned from allocation function "curl_url". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:480: var_assign: Assigning: "u" = storage returned from "curl_url()".
lib/http2.c:500: noescape: Resource "u" is not freed or pointed-to in "curl_url_set". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:502: leaked_storage: Variable "u" going out of scope leaks the storage it points to.
Error: RESOURCE_LEAK (CWE-772):
lib/http2.c:480: alloc_fn: Storage is returned from allocation function "curl_url". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:480: var_assign: Assigning: "u" = storage returned from "curl_url()".
lib/http2.c:505: noescape: Resource "u" is not freed or pointed-to in "curl_url_get". [Note: The source code implementation of the function has been overridden by a builtin model.]
lib/http2.c:507: leaked_storage: Variable "u" going out of scope leaks the storage it points to.
Closes#6986
Storing a stream error in the per-connection struct was an error that lead to
race conditions as subsequent stream handling could overwrite the error code
before it was used for the stream with the actual problem.
Closes#6910
This was this one condition where the stream could be closed due to an
error and the function would still wrongly just return 0 for it.
Reported-by: Gergely Nagy
Fixes#6862Closes#6910
Both were used for the same purposes and there was no logical separation
between them. Combined, this also saves 16 bytes in less holes in my
test build.
Closes#6798
... but instead use a private alternative that points to the "driving
transfer" from the connection. We set the "user data" associated with
the connection to be the connectdata struct, but when we drive transfers
the code still needs to know the pointer to the transfer. We can change
the user data to become the Curl_easy handle, but with older nghttp2
version we cannot dynamically update that pointer properly when
different transfers are used over the same connection.
Closes#6520
... in most cases instead of 'struct connectdata *' but in some cases in
addition to.
- We mostly operate on transfers and not connections.
- We need the transfer handle to log, store data and more. Everything in
libcurl is driven by a transfer (the CURL * in the public API).
- This work clarifies and separates the transfers from the connections
better.
- We should avoid "conn->data". Since individual connections can be used
by many transfers when multiplexing, making sure that conn->data
points to the current and correct transfer at all times is difficult
and has been notoriously error-prone over the years. The goal is to
ultimately remove the conn->data pointer for this reason.
Closes#6425
... as the socket might be readable all the time when paused and thus
causing a busy-loop.
Reported-by: Harry Sintonen
Reviewed-by: Jay Satiro
Fixes#6356Closes#6357
Previously there were several locations that called
nghttp2_session_mem_recv and handled responses slightly differently.
Those have been converted to call the existing
h2_process_pending_input() function.
Moved the end-of-session check to h2_process_pending_input() since the
only place the end-of-session state can change is after nghttp2
processes additional input frames.
This will likely fix the fuzzing error. While I don't have a root cause
the out-of-bounds read seems like a use after free, so moving the
nghttp2_session_check_request_allowed() call to a location with a
guaranteed nghttp2 session seems reasonable.
Also updated a few nghttp2 callsites to include error messages and added
a few additional error checks.
Closes#5648
The previous h2 trailer fix in 54a2b63 was wrong and caused a
regression: it cannot deal with trailers immediately when read since
they may be read off the connection by the wrong 'data' owner.
This change reverts the logic back to gathering all trailers into a
single buffer, like before 54a2b63.
Reported-by: Tadej Vengust
Fixes#5663Closes#5769
Well-behaving HTTP2 servers send two GOAWAY messages. The first
message is a warning that indicates that the server is going to
stop accepting streams. The second one actually closes the stream.
nghttp2 reports this state (and the other state of no more stream
identifiers) via the call nghttp2_session_check_request_allowed().
In this state the client should not create more streams on the
session (tcp connection), and in curl this means that the server
has requested that the connection is closed.
It would be also be possible to put the connclose() call into the
on_http2_frame_recv() function that triggers on the GOAWAY message.
This fixes a bug seen when the client sees the following sequence of
frames:
// advisory GOAWAY
HTTP2 GOAWAY [stream-id = 0, promised-stream-id = -1]
... some additional frames
// final GOAWAY
HTTP2 GOAWAY [stream-id = 0, promised-stream-id = N ]
Before this change, curl will attempt to reuse the connection even
after the last stream, will encounter this error:
* Found bundle for host localhost: 0x5595f0a694e0 [can multiplex]
* Re-using existing connection! (#0) with host localhost
* Connected to localhost (::1) port 10443 (#0)
* Using Stream ID: 9 (easy handle 0x5595f0a72e30)
> GET /index.html?5 HTTP/2
> Host: localhost:10443
> user-agent: curl/7.68.0
> accept: */*
>
* stopped the pause stream!
* Connection #0 to host localhost left intact
curl: (16) Error in the HTTP2 framing layer
This error may posion the connection cache, causing future requests
which resolve to the same curl connection to go through the same error
path.
Closes#5643
Confusingly, nghttp2 has two different error code enums:
- nghttp2_error, to be used with nghttp2_strerror
- nghttp2_error_code, to be used with nghttp2_http2_strerror
Closes#5641
...previously CURLINFO_EFFECTIVE_URL would report the URL of the
original "mother transfer", not the actually pushed resource.
Reported-by: Jonathan Cardoso Machado
Fixes#5589Closes#5591
When the method is updated inside libcurl we must still not change the
method as set by the user as then repeated transfers with that same
handle might not execute the same operation anymore!
This fixes the libcurl part of #5462
Test 1633 added to verify.
Closes#5499
... and free it as soon as the transfer is done. It removes the extra
alloc when a new size is set with setopt() and reduces memory for unused
easy handles.
In addition: the closure_handle now doesn't use an allocated buffer at
all but the smallest supported size as a stack based one.
Closes#5472
Triggered by a crash detected by OSS-Fuzz after the dynbuf introduction in
ed35d6590e. This should make the trailer handling more straight forward and
hopefully less error-prone.
Deliver the trailer header to the callback already at receive-time. No
longer caches the trailers to get delivered at end of stream.
Bug: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=22030Closes#5348
A common set of functions instead of many separate implementations for
creating buffers that can grow when appending data to them. Existing
functionality has been ported over.
In my early basic testing, the total number of allocations seem at
roughly the same amount as before, possibly a few less.
See docs/DYNBUF.md for a description of the API.
Closes#5300
Prior to this change in libcurl debug builds http2 stream closure was
erroneously referred to as connection closure.
Before:
* nread <= 0, server closed connection, bailing
After:
* nread == 0, stream closed, bailing
Closes https://github.com/curl/curl/pull/5118
This reduces the HTTP/2 window size to 32 MB since libcurl might have to
buffer up to this amount of data in memory and yet we don't want it set
lower to potentially impact tranfer performance on high speed networks.
Requires nghttp2 commit b3f85e2daa629
(https://github.com/nghttp2/nghttp2/pull/1444) to work properly, to end
up in the next release after 1.40.0.
Fixes#4939Closes#4940
To simplify our code and since earlier versions lack important function
calls libcurl needs to function correctly.
nghttp2 1.12.0 was relased on June 26, 2016.
Closes#4961
A regression made the code use 'multiplexed' as a boolean instead of the
counter it is intended to be. This made curl try to "over-populate"
connections with new streams.
This regression came with 41fcdf71a1, shipped in curl 7.65.0.
Also, respect the CURLMOPT_MAX_CONCURRENT_STREAMS value in the same
check.
Reported-by: Kunal Ekawde
Fixes#4779Closes#4784
It could accidentally let the connection get used by more than one
thread, leading to double-free and more.
Reported-by: Christopher Reid
Fixes#4544Closes#4557
- Disable warning C4127 "conditional expression is constant" globally
in curl_setup.h for when building with Microsoft's compiler.
This mainly affects building with the Visual Studio project files found
in the projects dir.
Prior to this change the cmake and winbuild build systems already
disabled 4127 globally for when building with Microsoft's compiler.
Also, 4127 was already disabled for all build systems in the limited
circumstance of the WHILE_FALSE macro which disabled the warning
specifically for while(0). This commit removes the WHILE_FALSE macro and
all other cruft in favor of disabling globally in curl_setup.
Background:
We have various macros that cause 0 or 1 to be evaluated, which would
cause warning C4127 in Visual Studio. For example this causes it:
#define Curl_resolver_asynch() 1
Full behavior is not clearly defined and inconsistent across versions.
However it is documented that since VS 2015 Update 3 Microsoft has
addressed this somewhat but not entirely, not warning on while(true) for
example.
Prior to this change some C4127 warnings occurred when I built with
Visual Studio using the generated projects in the projects dir.
Closes https://github.com/curl/curl/pull/4658
To make sure that transfer is being dealt with. Streams without
Content-Length need a final read to notice the end-of-stream state.
Reported-by: Tom van der Woerdt
Fixes#4496
To make sure that the HTTP/2 state is initialized correctly for
duplicated handles. It would otherwise easily generate "spurious"
PRIORITY frames to get sent over HTTP/2 connections when duplicated easy
handles were used.
Reported-by: Daniel Silverstone
Fixes#4303Closes#4442
If the :authority pseudo header field doesn't contain an explicit port,
we assume it is valid for the default port, instead of rejecting the
request for all ports.
Ref: https://curl.haxx.se/mail/lib-2019-09/0041.htmlCloses#4365
It could otherwise return an error even when closed correctly if GOAWAY
had been received previously.
Reported-by: Tom van der Woerdt
Fixes#4267Closes#4268
It was used (intended) to pass in the size of the 'socks' array that is
also passed to these functions, but was rarely actually checked/used and
the array is defined to a fixed size of MAX_SOCKSPEREASYHANDLE entries
that should be used instead.
Closes#4169
There were a leftover few prototypes of Curl_ functions that we used to
export but no longer do, this removes those prototypes and cleans up any
comments still referring to them.
Curl_write32_le(), Curl_strcpy_url(), Curl_strlen_url(), Curl_up_free()
Curl_concat_url(), Curl_detach_connnection(), Curl_http_setup_conn()
were made static in 05b100aee2.
Curl_http_perhapsrewind() made static in 574aecee20.
For the remainder, I didn't trawl the Git logs hard enough to capture
their exact time of deletion, but they were all gone: Curl_splayprint(),
Curl_http2_send_request(), Curl_global_host_cache_dtor(),
Curl_scan_cache_used(), Curl_hostcache_destroy(), Curl_second_connect(),
Curl_http_auth_stage() and Curl_close_connections().
Closes#4096
Reviewed-by: Daniel Stenberg <daniel@haxx.se>
To make sure a HTTP/2 stream registers the end of stream.
Bug #4043 made me find this problem but this fix doesn't correct the
reported issue.
Closes#4068
Various functions called within Curl_http2_done() can have the
side-effect of setting the Easy connection into drain mode (by calling
drain_this()). However, the last time we unset this for a transfer (by
calling drained_transfer()) is at the beginning of Curl_http2_done().
If the Curl_easy is reused for another transfer, it is then stuck in
drain mode permanently, which in practice makes it unable to write any
data in the new transfer.
This fix moves the last call to drained_transfer() to later in
Curl_http2_done(), after the functions that could potentially call for a
drain.
Fixes#3966Closes#3967
Reported-by: Josie-H
They serve very little purpose and mostly just add noise. Most of them
have been around for a very long time. I read them all before removing
or rephrasing them.
Ref: #3876Closes#3883
RFC 7540 says we should verify that the push is for an "authoritative"
server. We make sure of this by only allowing push with an :athority
header that matches the host that was asked for in the URL.
Fixes#3577
Reported-by: Nicolas Grekas
Bug: https://curl.haxx.se/mail/lib-2019-02/0057.htmlCloses#3581
urlapi: turn three local-only functions into statics
conncache: make conncache_find_first_connection static
multi: make detach_connnection static
connect: make getaddressinfo static
curl_ntlm_core: make hmac_md5 static
http2: make two functions static
http: make http_setup_conn static
connect: make tcpnodelay static
tests: make UNITTEST a thing to mark functions with, so they can be static for
normal builds and non-static for unit test builds
... and mark Curl_shuffle_addr accordingly.
url: make up_free static
setopt: make vsetopt static
curl_endian: make write32_le static
rtsp: make rtsp_connisdead static
warnless: remove unused functions
memdebug: remove one unused function, made another static
We use "conn" everywhere to be a pointer to the connection.
Introduces two functions that "attaches" and "detaches" the connection
to and from the transfer.
Going forward, we should favour using "data->conn" (since a transfer
always only has a single connection or none at all) to "conn->data"
(since a connection can have none, one or many transfers associated with
it and updating conn->data to be correct is error prone and a frequent
reason for internal issues).
Closes#3442
This is a companion patch to cbea2fd2c (NTLM: force the connection to
HTTP/1.1, 2018-12-06): with NTLM, we can switch to HTTP/1.1
preemptively. However, with other (Negotiate) authentication it is not
clear to this developer whether there is a way to make it work with
HTTP/2, so let's try HTTP/2 first and fall back in case we encounter the
error HTTP_1_1_REQUIRED.
Note: we will still keep the NTLM workaround, as it avoids an extra
round trip.
Daniel Stenberg helped a lot with this patch, in particular by
suggesting to introduce the Curl_h2_http_1_1_error() function.
Closes#3349
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The function does not return the same value as snprintf() normally does,
so readers may be mislead into thinking the code works differently than
it actually does. A different function name makes this easier to detect.
Reported-by: Tomas Hoger
Assisted-by: Daniel Gustafsson
Fixes#3296Closes#3297
The result of a memory allocation should always be checked, as we may
run under memory pressure where even a small allocation can fail. This
adds checking and error handling to a few cases where the allocation
wasn't checked for success. In the ftp case, the freeing of the path
variable is moved ahead of the allocation since there is little point
in keeping it around across the strdup, and the separation makes for
more readable code. In nwlib, the lock is aslo freed in the error path.
Also bumps the copyright years on affected files.
Closes#3084
Reviewed-by: Jay Satiro <raysatiro@yahoo.com>
Reviewed-by: Daniel Stenberg <daniel@haxx.se>
Add functionality so that protocols can do custom keepalive on their
connections, when an external API function is called.
Add docs for the new options in 7.62.0
Closes#1641
If this is the last stream on this connection, the RST_STREAM might not
get pushed to the wire otherwise.
Fixes#2882Closes#2887
Researched-by: Michael Kaufmann
Deal with tiny "HTTP/0.9" (header-less) responses by checking the
status-line early, even before a full "HTTP/" is received to allow
detecting 0.9 properly.
Test 1266 and 1267 added to verify.
Fixes#2420Closes#2872
This function can get called on a connection that isn't setup enough to
have the 'recv_underlying' function pointer initialized so it would try
to call the NULL pointer.
Reported-by: Dario Weisser
Follow-up to db1b2c7fe9 (never shipped in a release)
Closes#2536
Follow-up to 1514c44655: replace another strstr() call done on a
buffer that might not be zero terminated - with a memchr() call, even if
we know the substring will be found.
Assisted-by: Max Dymond
Detected by OSS-Fuzz
Bug: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=8021Closes#2534
Fuzzing has proven we can reach code in on_frame_recv with status_code
not having been set, so let's detect that in run-time (instead of with
assert) and error error accordingly.
(This should no longer happen with the latest nghttp2)
Detected by OSS-Fuzz
Bug: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=7903Closes#2514
When receiving REFUSED_STREAM, mark the connection for close and retry
streams accordingly on another/fresh connection.
Reported-by: Terry Wu
Fixes#2416Fixes#1618Closes#2510
It's not strictly clear if the API contract allows us to call strstr()
on a string that isn't zero terminated even when we know it will find
the substring, and clang's ASAN check dislikes us for it.
Also added a check of the return code in case it fails, even if I can't
think of a situation how that can trigger.
Detected by OSS-Fuzz
Closes#2513
Bug: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=7760
This triggered an assert if called more than once in debug mode (and a
memory leak if not debug build). With the right sequence of HTTP/2
headers incoming it can happen.
Detected by OSS-Fuzz
Closes#2507
Bug: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=7764
If a connection has received a GOAWAY frame while not being used, the
function now reads frames off the connection before trying to reuse it
to avoid reusing connections the server has told us not to use.
Reported-by: Alex Baines
Fixes#1967Closes#2402
Prior to this change the stored byte count of each trailer was
miscalculated and 1 less than required. It appears any trailer
after the first that was passed to Curl_client_write would be truncated
or corrupted as well as the size. Potentially the size of some
subsequent trailer could be erroneously extracted from the contents of
that trailer, and since that size is used by client write an
out-of-bounds read could occur and cause a crash or be otherwise
processed by client write.
The bug appears to have been born in 0761a51 (precedes 7.49.0).
Closes https://github.com/curl/curl/pull/2231
Add a new type of callback to Curl_handler which performs checks on
the connection. Alter RTSP so that it uses this callback to do its
own check on connection health.
mk-lib1521.pl generates a test program (lib1521.c) that calls
curl_easy_setopt() for every known option with a few typical values to
make sure they work (ignoring the return codes).
Some small changes were necessary to avoid asserts and NULL accesses
when doing this.
The perl script needs to be manually rerun when we add new options.
Closes#1543
... since the total amount is low this is faster, easier and reduces
memory overhead.
Also, Curl_expire_done() can now mark an expire timeout as done so that
it never times out.
Closes#1472
A) reduces the timeout lists drastically
B) prevents a lot of superfluous loops for timers that expires "in vain"
when it has actually already been extended to fire later on
This fixes the following clang warnings:
http2.c:184:27: error: no previous extern declaration for non-static
variable 'Curl_handler_http2' [-Werror,-Wmissing-variable-declarations]
http2.c:204:27: error: no previous extern declaration for non-static
variable 'Curl_handler_http2_ssl'
[-Werror,-Wmissing-variable-declarations]
When removing an easy handler from a multi before it completed its
transfer, and it had pushed streams, it would segfault due to the pushed
counted not being cleared.
Fixed-by: zelinchen@users.noreply.github.comFixes#1249
- In Curl_http2_switched don't call memcpy when src is NULL.
Curl_http2_switched can be called like:
Curl_http2_switched(conn, NULL, 0);
.. and prior to this change memcpy was then called like:
memcpy(dest, NULL, 0)
.. causing address sanitizer to warn:
http2.c:2057:3: runtime error: null pointer passed as argument 2, which
is declared to never be null
... by making sure we don't count down the "upload left" counter when the
uploaded size is unknown and then it can be allowed to continue forever.
Fixes#996
With HTTP/2 each transfer is made in an indivial logical stream over the
connection, making most previous errors that caused the connection to get
forced-closed now instead just kill the stream and not the connection.
Fixes#941
Since the server can at any time send a HTTP/2 frame to us, we need to
wait for the socket to be readable during all transfers so that we can
act on incoming frames even when uploading etc.
Reminded-by: Tatsuhiro Tsujikawa
After a few wasted hours hunting down the reason for slowness during a
TLS handshake that turned out to be because of TCP_NODELAY not being
set, I think we have enough motivation to toggle the default for this
option. We now enable TCP_NODELAY by default and allow applications to
switch it off.
This also makes --tcp-nodelay unnecessary, but --no-tcp-nodelay can be
used to disable it.
Thanks-to: Tim Rühsen
Bug: https://curl.haxx.se/mail/lib-2016-06/0143.html
Previously, passing a timeout of zero to Curl_expire() was a magic code
for clearing all timeouts for the handle. That is now instead made with
the new Curl_expire_clear() function and thus a 0 timeout is fine to set
and will trigger a timeout ASAP.
This will help removing short delays, in particular notable when doing
HTTP/2.
curl's representation of HTTP/2 responses involves transforming the
response to a format that is similar to HTTP/1.1. Prior to this change,
curl would do this by separating header names and values with only a
colon, without introducing a space after the colon.
While this is technically a valid way to represent a HTTP/1.1 header
block, it is much more common to see a space following the colon. This
change introduces that space, to ensure that incautious tools are safely
able to parse the header block.
This also ensures that the difference between the HTTP/1.1 and HTTP/2
response layout is as minimal as possible.
Bug: https://github.com/curl/curl/issues/797Closes#798Fixes#797
curl_printf.h defines printf to curl_mprintf, etc. This can cause
problems with external headers which may use
__attribute__((format(printf, ...))) markers etc.
To avoid that they cause problems with system includes, we include
curl_printf.h after any system headers. That makes the three last
headers to always be, and we keep them in this order:
curl_printf.h
curl_memory.h
memdebug.h
None of them include system headers, they all do funny #defines.
Reported-by: David Benjamin
Fixes#743
- Error if a header line is larger than supported.
- Warn if cumulative header line length may be larger than supported.
- Allow spaces when parsing the path component.
- Make sure each header line ends in \r\n. This fixes an out of bounds.
- Disallow header continuation lines until we decide what to do.
Ref: https://github.com/curl/curl/issues/659
Ref: https://github.com/curl/curl/pull/663
This commit ensures that streams which was closed in on_stream_close
callback gets passed to http2_handle_stream_close. Previously, this
might not happen. To achieve this, we increment drain property to
forcibly call recv function for that stream.
To more accurately check that we have no pending event before shutting
down HTTP/2 session, we sum up drain property into
http_conn.drain_total. We only shutdown session if that value is 0.
With this commit, when stream was closed before reading response
header fields, error code CURLE_HTTP2_STREAM is returned even if
HTTP/2 level error is NO_ERROR. This signals the upper layer that
stream was closed by error just like TCP connection close in HTTP/1.
Ref: https://github.com/curl/curl/issues/659
Ref: https://github.com/curl/curl/pull/663
This commit ensures that data from network are processed before HTTP/2
session is terminated. This is achieved by pausing nghttp2 whenever
different stream than current easy handle receives data.
This commit also fixes the bug that sometimes processing hangs when
multiple HTTP/2 streams are multiplexed.
Ref: https://github.com/curl/curl/issues/659
Ref: https://github.com/curl/curl/pull/663
Previously, when a stream was closed with other than NGHTTP2_NO_ERROR
by RST_STREAM, underlying TCP connection was dropped. This is
undesirable since there may be other streams multiplexed and they are
very much fine. This change introduce new error code
CURLE_HTTP2_STREAM, which indicates stream error that only affects the
relevant stream, and connection should be kept open. The existing
CURLE_HTTP2 means connection error in general.
Ref: https://github.com/curl/curl/issues/659
Ref: https://github.com/curl/curl/pull/663
... but ignore EAGAIN if the stream has ended so that we don't end up in
a loop. This is a follow-up to c8ab613 in order to avoid the problem
d261652 was made to fix.
Reported-by: Jay Satiro
Clues-provided-by: Tatsuhiro Tsujikawa
Discussed in #750
It turns out the google GFE HTTP/2 servers send a PING frame immediately
after a stream ends and its last DATA has been received by curl. So if
we don't drain that from the socket, it makes the socket readable in
subsequent checks and libcurl then (wrongly) assumes the connection is
dead when trying to reuse the connection.
Reported-by: Joonas Kuorilehto
Discussed in #750
It offers extra info from nghttp2 in certain error cases. Like for
example when trying prior-knowledge http2 on a server that doesn't speak
http2 at all. The error message is passed on as a verbose message to
libcurl.
Discussed in #722
The error callback was added in nghttp2 1.9.0
Since commit a5aec58 the handler schemes need to match for the
connections to be reused and for HTTP/2 multiplexing to work, reusing
connections is very important!
Closes#736
Check that the trailer buffer exists before attempting a client write
for trailers on stream close.
Refer to comments in https://github.com/bagder/curl/pull/564
This commit adds trailer support in HTTP/2. In HTTP/1.1, chunked
encoding must be used to send trialer fields. HTTP/2 deprecated any
trandfer-encoding, including chunked. But trailer fields are now
always available.
Since trailer fields are relatively rare these days (gRPC uses them
extensively though), allocating buffer for trailer fields is done when
we detect that HEADERS frame containing trailer fields is started. We
use Curl_add_buffer_* functions to buffer all trailers, just like we
do for regular header fields. And then deliver them when stream is
closed. We have to be careful here so that all data are delivered to
upper layer before sending trailers to the application.
We can deliver trailer field one by one using NGHTTP2_ERR_PAUSE
mechanism, but current method is far more simple.
Another possibility is use chunked encoding internally for HTTP/2
traffic. I have not tested it, but it could add another overhead.
Closes#564
When NGHTTP2_ERR_PAUSE is returned from data_source_read_callback, we
might not process DATA frame fully. Calling nghttp2_session_mem_recv()
again will continue to process DATA frame, but if there is no incoming
frames, then we have to call it again with 0-length data. Without this,
on_stream_close callback will not be called, and stream could be hanged.
Bug: http://curl.haxx.se/mail/lib-2015-11/0103.html
Reported-by: Francisco Moraes
They tend to never get updated anyway so they're frequently inaccurate
and we never go back to revisit them anyway. We document issues to work
on properly in KNOWN_BUGS and TODO instead.
Removed wrong assert()s
The 'conn' passed in as userdata can be used and there can be other
sessionhandles ('data') than the single one this checked for.
