- Store the c-ares version during global init.
Prior to this change several threads could write the same data to a
static int variable at the same time. Though in practice it's not a
problem ThreadSanitizer may warn.
Reported-by: Nikita Taranov
Assisted-by: Jay Satiro
Fixes#13065Closes#13000
Just a tidy up to contain 'ifdef' pollution of common
code parts with implementation specifics.
- remove the ifdef hyper unpausing in easy.c
- add hyper client reader for CURL_CR_PROTOCOL phase
that implements the unpause method for calling
the hyper waker if it is set
Closes#13075
This is a follow-up for PR #12897.
Add support for SHA-512/256 digest calculation by TLS backends.
Currently only OpenSSL and GnuTLS (actually, nettle) support
SHA-512/256.
Closes#13070
- `struct Curl_cwriter` and `struct Curl_creader` now carry a
`void *ctx` member that points to the instance as allocated.
- using `r->ctx` and `w->ctx` as pointer to the instance specific
struct that has been allocated
Reported-by: Rudi Heitbaum
Fixes#13035Closes#13059
- the change breaks looping in transfer.c receive for transfers that are
speed limited on having gotten *some* bytes.
- the overall speed limit timing is done in multi.c
Reported-by: Dmitry Karpov
Bug: https://curl.se/mail/lib-2024-03/0001.htmlCloses#13050
Add `mime` client reader. Encapsulates reading from mime parts, getting
their length, rewinding and unpausing.
- remove special mime handling from sendf.c and easy.c
- add general "unpause" method to client readers
- use new reader in http/imap/smtp
- make some mime functions static that are now only used internally
In addition:
- remove flag 'forbidchunk' as no longer needed
Closes#13039
- set TIMER_STARTTRANSFER on seeing the first response bytes
in the download client writer, not coming from a CONNECT
- initialized the timer the same way for all protocols
- remove explicit setting of TIMER_STARTTRANSFER in file.c
and c-hyper.c
Closes#13052
If a response without a status line is received, and the connection is
known to use HTTP/1.x (not HTTP/0.9), report the error "Invalid status
line" instead of "Received HTTP/0.9 when not allowed".
Closes#13045
In cases where the connection was fast, curl sometimes failed to open a
connection. This fixes a regression of c2d973627b.
The regression triggered in these steps:
1. Create an smtp connection
2. Use STARTTLS
3. Receive the response
4. We are inside the loop in `smtp_statemachine`, calling
`smtp_state_starttls_resp`
5. In the good flow, we exit the loop, re-enter `smtp_statemachine` and
run `smtp_perform_upgrade_tls` at the start of the function.
In the bad flow, we stay in the while loop, calling
`Curl_pp_readresp`, which reads part of the TLS handshake and things
go wrong.
The reason is that `Curl_pp_moredata` changed behavior and always
returns `true`, so we stay in the loop in `smtp_statemachine`. With a
slow connection `Curl_pp_readresp` cannot read new data and returns
`CURL_AGAIN`, so we leave the loop and re-enter `smtp_statemachine`.
With a fast connection, `Curl_pp_readresp` reads new data from the tcp
connection, which is part of the TLS handshake.
The fix is in `Curl_pp_moredata`, which needs to take the final line
into account and return `false` if only the final line is stored.
Closes#13048
- update client reader documentation
- client reader, add rewind capabilities
- tell creader to rewind on next start
- Curl_client_reset() will keep reader for future rewind if requested
- add Curl_client_cleanup() for freeing all resources independent of
rewinds
- add Curl_client_start() to trigger rewinds
- move rewind code from multi.c to sendf.c and make part of
"cr-in"'s implementation
- http, move the "resume_from" handling into the client readers
- the setup of a HTTP request is reshuffled to follow:
* determine method, target, auth negotiation
* install the client reader(s) for the request, including crlf
conversions and "chunked" encoding
* apply ranges to client reader
* concat request headers, upgrades, cookies, etc.
* complete request by determining Content-Length of installed
readers in combination with method
* send
- add methods for client readers to
* return the overall length they will generate (or -1 when unknown)
* return the amount of data on the CLIENT level, so that
expect-100 can decide if it want to apply itself
* set a "resume_from" offset or fail if unsupported
- struct HTTP has become largely empty now
- rename `Client_reader_*` to `Curl_creader_*`
Closes#13026
Caused by an accidentally duplicated line in
d6825df334.
```
.../lib/vquic/curl_osslq.c:1095:30: warning: implicit conversion loses integer precision: 'curl_socket_t' (aka 'unsigned long long') to 'int' [-Wshorten-64-to-32]
1095 | bio = BIO_new_dgram(ctx->q.sockfd, BIO_NOCLOSE);
| ~~~~~~~~~~~~~ ~~~~~~~^~~~~~
1 warning and 2 errors generated.
```
Reviewed-by: Stefan Eissing
Closes#13043
- rename static functions to avoid duplicate symbols in unity mode.
- windows -> Windows/window in error message and comment.
- fix indentation.
Reviewed-by: Stefan Eissing
Closes#13044
A libpsl install without data and no built-in database is now considered
bad enough to reject all cookies since they cannot be checked. It is
somewhat of a user error, but still.
Reported-by: Dan Fandrich
Closes#13033
- Move all the "upload_done" handling to request.c
- add possibility to abort sending of a request
- add `Curl_req_done_sending()` for checks
- transfer.c: readwrite_upload() now clean
- removing data->state.ulbuf and data->req.upload_fromhere
- as well as data->req.upload_present
- set data->req.upload_done on having read all from
the client and completely flushed the send buffer
- tftp, remove setting of data->req.upload_fromhere
- serves no purpose as `upload_present` is not set
and the data itself is directly `sendto()` anyway
- smtp, make upload EOB conversion a client reader
- xfer_ulbuf addition
- add xfer_ulbuf for borrowing, similar to xfer_buf
- use in file upload
- use in c-hyper body sending
- h1-proxy, remove init of data->state.uilbuf that is never used
- smb, add own send_buf instead of using data->state.ulbuf
Closes#13010
- when unable to obtain a new chunk on a softlimit bufq,
this is an allocation error and needs to be reported as
such.
- writes into a soflimit bufq never must be partial success
Reported-by: Dan Fandrich
Fixes#13020Closes#13023
This fixes miscellaneous typos and duplicated words in the docs, lib
and test comments and a few user facing errorstrings.
Author: RainRat on Github
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Reviewed-by: Dan Fandrich <dan@coneharvesters.com>
Closes: #13019
- 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 (to be made
into a sperate PR, also)
Added as documented [in
CLIENT-READER.md](5b1f31dfba/docs/CLIENT-READERS.md).
- old `Curl_buffer_send()` completely replaced by new `Curl_req_send()`
- old `Curl_fillreadbuffer()` replaced with `Curl_client_read()`
- HTTP chunked uploads are now formatted in a client reader added when
needed.
- FTP line-end conversions are done in a client reader added when
needed.
- when sending requests headers, remaining buffer space is filled with
body data for sending in "one go". This is independent of the request
body size. Resolves#12938 as now small and large requests have the
same code path.
Changes done to test cases:
- test513: now fails before sending request headers as this initial
"client read" triggers the setup fault. Behaves now the same as in
hyper build
- test547, test555, test1620: fix the length check in the lib code to
only fail for reads *smaller* than expected. This was a bug in the
test code that never triggered in the old implementation.
Closes#12969
The curldown conversion accidentally replaced daniel@haxx.se with
just daniel.se. This reverts back to the proper email address in
the curldown docs as well as in a few other stray places where it
was incorrect (while unrelated to curldown).
Reviewed-by: Daniel Stenberg <daniel@haxx.se>
Closes: #12997
When disabling all protocols without enabling any, the resulting
set of allowed protocols remained the default set. Clearing the
allowed set before inspecting the passed value from --proto make
the set empty even in the errorpath of no protocols enabled.
Co-authored-by: Dan Fandrich <dan@telarity.com>
Reported-by: Dan Fandrich <dan@telarity.com>
Reviewed-by: Daniel Stenberg <daniel@haxx.se>
Closes: #13004
- 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
If the easy handle that is being added to a multi handle has previously
been used for curl_easy_perform(), there is a private multi handle here
that we can kill off. While it flushes some caches etc for the easy
handle would it be used for an easy interface transfer again after being
used in the multi stack, this cleanup simplifies behavior and uses less
memory.
Closes#12992
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
In order to make MSAN happy:
==2200945==WARNING: MemorySanitizer: use-of-uninitialized-value
#0 0x596f3b3ed246 in curlx_strtoofft [...]/libcurl/src/lib/strtoofft.c:239:11
#1 0x596f3b402156 in Curl_httpchunk_read [...]/libcurl/src/lib/http_chunks.c:149:12
#2 0x596f3b348550 in readwrite_data [...]/libcurl/src/lib/transfer.c:607:11
[...]
==2202041==WARNING: MemorySanitizer: use-of-uninitialized-value
#0 0x5a3fab66a72a in Curl_parse_port [...]/libcurl/src/lib/urlapi.c:547:8
#1 0x5a3fab650645 in parse_authority [...]/libcurl/src/lib/urlapi.c:796:12
#2 0x5a3fab6740f6 in parseurl [...]/libcurl/src/lib/urlapi.c:1176:16
#3 0x5a3fab664fc5 in parseurl_and_replace [...]/libcurl/src/lib/urlapi.c:1342:12
[...]
==2202320==WARNING: MemorySanitizer: use-of-uninitialized-value
#0 0x569076a0d6b0 in ipv4_normalize [...]/libcurl/src/lib/urlapi.c:683:12
#1 0x5690769f2820 in parse_authority [...]/libcurl/src/lib/urlapi.c:803:10
#2 0x569076a160f6 in parseurl [...]/libcurl/src/lib/urlapi.c:1176:16
#3 0x569076a06fc5 in parseurl_and_replace [...]/libcurl/src/lib/urlapi.c:1342:12
[...]
Signed-off-by: Louis Solofrizzo <lsolofrizzo@scaleway.com>
Closes#12995
Refactoring of the client writer that passes the data to the
client/application's callback functions.
- split out into own source cw-out.[ch] from sendf.c
- move tempwrite and tempcount from data->state into the context of the
client writer
- redesign the 3 tempwrite dynbufs as a linked list of dynbufs. On
paused transfers, this allows to "record" interleaved HEADER/BODY
chunks to be "played back" in the same order on unpausing.
- keep the overall size limit of all buffered data to DYN_PAUSE_BUFFER.
On exceeding that, return CURLE_TOO_LARGE instead of
CURLE_OUT_OF_MEMORY as before.
- add method to be called when a transfer is DONE to allow writing of
any data still buffered
- when paused, record HEADER writes exactly as they come for later
playback. HEADERs are documented to be written one-by-one.
Closes#12898
- from `conn->bits.authneg` to `data->req.authneg`
- this is a property of the request about to be made
and not a property of the connection
- in multiuse connections, transfer could step on each others
toes here potentially.
Closes#12949
- add a client writer that does "push" response
headers written to the client if the headers api
is enabled
- remove special handling in sendf.c
- needs to be installed very early on connection
setup to catch CONNECT response headers
Closes#12880
Remove curl_mimepart object from UserDefined structure when
CURL_DISABLE_MIME flag is active. Reduce size of UserDefined structure.
Also remove unreachable code: when CURL_DISABLE_MIME is set, httpreq can
never have HTTPREQ_POST_MIME value and the same goes for the
CURL_DISABLE_FORM_API flag and the HTTPREQ_POST_FORM value
Closes#12948
... to not rely on wrapping, since it is an undefined behavior that is
not what always might happen. This is in our private strtoff() parser
function, used only on platforms without a native version.
Reported-by: vulnerabilityspotter on hackerone
Closes#12990
Prior to this change CURLOPT_PROXY_TLSAUTH_TYPE would return
CURLE_BAD_FUNCTION_ARGUMENT on any type other than NULL. Since there is
only one type of TLS auth and it is also the default (SRP) the TLS auth
would work anyway.
Closes https://github.com/curl/curl/pull/12981
- Support I32 & I64 (eg: %I64d) for all Win32 builds.
Prior to this change mprintf support for the I format prefix, which is a
Microsoft extension, was dependent on the compiler used.
When Borland compiler support was removed in fd7ef00f the prefix was
then no longer supported for that compiler; however since it's still
possible to build with Borland I'm restoring support for the prefix in
this way.
Reported-by: Paweł Witas
Fixes https://github.com/curl/curl/issues/12944
Closes https://github.com/curl/curl/pull/12950
- OR the event bitmask to data->state.select_bits instead of overwriting
them. They are cleared again on use.
Reported-by: 5533asdg on github
Fixes#12971Closes#12972
- refs #12397 where it is dicussed how to en-/disable verbose output
of DoH operations
- introducing `struct curl_trc_feat` to track a curl feature for
tracing
- adding `data->state.feat` optionally pointing to the feature a
transfer belongs to
- adding trace functions and verbosity checks on features
- using trace feature in DoH code
- documenting `doh` as feature for `--trace-config`
Closes#12411
- when data arrived in several chunks, the collection into
the passed buffer always started at offset 0, overwriting
the data already there.
adding test_20_07 to verify fix
- debug environment var CURL_WS_CHUNK_SIZE can be used to
influence the buffer chunk size used for en-/decoding.
Closes#12945
Also fix the tests. New implementation tested with GNU libmicrohttpd.
The new numbers in tests are real SHA-512/256 numbers (not just some
random ;) numbers ).
- set our idle timeout as transport parameter
- query negotiated idle timeout for connection alive checks
- query number of available bidi streams on a connection
- use write_ex2 with SSL_WRITE_FLAG_CONCLUDE to signal
EOF on last chunk write, so stream close does not
require an additional QUIC packet
Closes#12933
- Change the 1 second SO_SNDBUF update limit from per transfer to per
connection.
Prior to this change many transfers over the same connection could cause
many SO_SNDBUF updates made to that connection per second, which was
unnecessary.
Closes https://github.com/curl/curl/pull/12911
- Treat TLS connection close (either due to a close_notify from the
server or just closed due to receiving 0) as pending data.
This is because in some cases schannel_recv knows the connection is
closed but has to return actual pending data so it can't return 0 or an
error to indicate no more data. In this case schannel_recv must be
called again, which only happens if readwrite_data sees that there is
still pending data.
Prior to this change if the total size of the body that libcurl expected
to receive from the server was unknown then it was possible under some
network conditions that libcurl would hang waiting to receive more data,
when in fact a close_notify alert indicating no more data would be sent
was already processed.
Fixes https://github.com/curl/curl/issues/12894
Closes https://github.com/curl/curl/pull/12910
- when waiting on the data connection, always add the control socket to
the pollset on state STOP or let the pingpong add the socket according
to its needs.
Reported-by: Fabian Vogt
Fixes#12901Closes#12913
- can be borrowed by transfer during recv-write operation
- needs to be released before borrowing again
- adjustis size to `data->set.buffer_size`
- used in transfer.c readwrite_data()
Closes#12805
- remove the ftp special handling from sendf.c
- let ftp_do() add a client writer that does
the linened conversions
- change the lineend conversion to no longer
modify the passed buffer, but write smaller
chunks to the next cwriter instead. The
inefficiency of this will be mitigated once
we add output buffering for all client writes.
Closes#12878
- add test_05_04 for requests using http/1.0, http/1.1 and h2 against an
Apache resource that does an unclean TLS shutdown.
- revert special workarund in openssl.c for suppressing shutdown errors
on multiplexed connections
- vlts.c restore to its state before 9a90c9dd64Fixes#12885Fixes#12844Closes#12848
Create the line in a dynbuf. Aborts the reading of the file on
errors. Avoids having to always allocate maximum amount from the
start. Avoids direct malloc.
Closes#12846
- improve info logging when peer verification fails to indicate
if DNS name or ip address has been tried to match
- add test case for contacting https proxy with ip address
- add pytest env check on loaded credentials and re-issue
when they are no longer valid
- disable proxy ip address test for bearssl, since not supported there
Ref: #12831Closes#12838
The psl_check_version_number() API was added in libpsl 0.11.0. CentOS 7
ships with version 0.7.0 which lacks this API. Revert to using the older
versioning API if we detect an old libpsl version.
Follow-up to 72bd88adde
Bug: https://curl.se/mail/archive-2024-02/0004.html
Reported-by: Scott Mutter
Closes#12872
- use the new `Curl_xfer_write_resp()` to write incoming responses
directly to the client
- eliminates `stream->recvbuf`
- memory consumption on parallel transfers minimized
Closes#12828
and mark the stream for close, but return OK since the response this far
was ok - if headers were received. Partly because this is what curl has
done traditionally.
Test 499 verifies. Updates test 689.
Reported-by: Sergey Bronnikov
Bug: https://curl.se/mail/lib-2024-02/0000.htmlCloses#12842
For active mode transfers.
Due to some interesting timing, curl can sometimes get the 226 (transfer
complete) over the control channel first, before the data connection
signals readability. If this happens, use that as a signal to check the
data connection.
Additionally, set the socket filter in listen mode *before* the
PORT/EPRT command is issued, to reduce the risk that the little time gap
could interfere.
This issue never reproduced for me on Debian and takes several hundred
rounds for me to trigger on my mac.
Reported-by: Stefan Eissing
Fixes#12823Closes#12841
Since the QUIC/h3 code has no knowledge or handling of multissl it might
bring unintended consequences if we allow it.
configure, cmake and curl_setup.h all now reject this combination.
Assisted-by: Viktor Szakats
Assisted-by: Gisle Vanem
Ref: #12806Closes#12807
- delete redundant warning suppressions for `-Wformat-nonliteral`.
This now relies on `CURL_PRINTF()` and it's theoratically possible
that this macro isn't active but the warning is. We're ignoring this
as a corner-case here.
- replace two pragmas with code changes to avoid the warnings.
Follow-up to aee4ebe591#12803
Follow-up to 0923012758#12540
Follow-up to 3829759bd0#12489
Reviewed-by: Daniel Stenberg
Closes#12812
- Use http authentication mechanisms as a default, not a preset.
Consider http authentication options which are mapped to SASL options as
a default (overriding the hardcoded default mask for the protocol) that
is ignored if a login option string is given.
Prior to this change, if some HTTP auth options were given, sasl mapped
http authentication options to sasl ones but merged them with the login
options.
That caused problems with the cli tool that sets the http login option
CURLAUTH_BEARER as a side-effect of --oauth2-bearer, because this flag
maps to more than one sasl mechanisms and the latter cannot be cleared
individually by the login options string.
New test 992 checks this.
Fixes https://github.com/curl/curl/issues/10259
Closes https://github.com/curl/curl/pull/12790
When checking if the user wants to replace the header, the check should
be case insensitive.
Adding test 461 to verify
Found-by: Dan Fandrich
Ref: #12782Closes#12784
The pingpong logic now uses its own dynbuf for receiving command
response data.
When the "final" response header for a commanad has been received, that
final line is left first in the recvbuf for the protocols to parse at
will. If there is additional data behind the final response line, the
'overflow' counter is indicate how many bytes.
Closes#12757
Rework CMake options for building/using curl tool and libcurl manuals.
- rename `ENABLE_MANUAL` to `ENABLE_CURL_MANUAL`, meaning:
to build man page and built-in manual for curl tool.
- rename `BUILD_DOCS` to `BUILD_LIBCURL_DOCS`, meaning:
to build man pages for libcurl.
- `BUILD_LIBCURL_DOCS` now works without having to enable
`ENABLE_CURL_MANUAL` too.
- drop support for existing CMake-level `USE_MANUAL` option to avoid
confusion. (It used to work with the effect of current
`ENABLE_CURL_MANUAL`, but only by accident.)
Assisted-by: Richard Levitte
Ref: #12771Closes#12773
This assert triggers wrongly when CURLU_GUESS_SCHEME and
CURLU_NO_AUTHORITY are both set and the URL is a single path.
I think this assert has played out its role. It was introduced in a
rather big refactor.
Follow-up to 4cfa5bcc9a
Reported-by: promptfuzz_ on hackerone
Closes#12775
Since we only support using a single TLS library at any one time, we
know that the TLS library for QUIC is the same that is also shown for
regular TLS.
Fixes#12763
Reported-by: Viktor Szakats
Closes#12767
To prevent that it gets used in a subsequent transfer that skips the
verifystatus check since that check can't be done when the session id is
reused.
Reported-by: Hiroki Kurosawa
Closes#12760
To avoid a local hack to pass function pointers and to avoid
deprecation warnings when building with libssh2 v1.11.1 or newer:
```
lib/vssh/libssh2.c:3324:5: warning: 'libssh2_session_callback_set' is deprecated: since libssh2 1.11.1. Use libssh2_session_callback_set2() [-Wdeprecated-declarations]
lib/vssh/libssh2.c:3326:5: warning: 'libssh2_session_callback_set' is deprecated: since libssh2 1.11.1. Use libssh2_session_callback_set2() [-Wdeprecated-declarations]
```
Ref: https://github.com/curl/curl-for-win/actions/runs/7609484879/job/20720821100#step:3:4982
Ref: https://github.com/libssh2/libssh2/pull/1285
Ref: c0f69548be
Reviewed-by: Daniel Stenberg
Closes#12754
- HTTP/3 for curl using OpenSSL's own QUIC stack together
with nghttp3
- configure with `--with-openssl-quic` to enable curl to
build this. This requires the nghttp3 library
- implementation with the following restrictions:
* macOS has to use an unconnected UDP socket due to an
issue in OpenSSL's datagram implementation
See https://github.com/openssl/openssl/issues/23251
This makes connections to non-reponsive servers hang.
* GET requests will send the indicator that they have
no body in a separate QUIC packet. This may result
in processing delays or Transfer-Encodings on proxied
requests
* uploads that encounter blocks will use 100% cpu as
detection of these flow control issue is not working
(we have not figured out to pry that from OpenSSL).
Closes#12734
- in en- and decoding, check the websocket frame payload lengths for
negative values (from curl_off_t) and error the operation in that case
- add test 2307 to verify
Closes#12707
- enforce a response body length of 0, if the
response has no Content-lenght. This is according
to the RTSP spec.
- excess bytes in a response body are forwarded to
the client writers which will report and fail the
transfer
Follow-up to d7b6ce6Fixes#12701Closes#12706
The libpsl version output otherwise also includes version number for its
dependencies, like IDN lib, but since libcurl does not use libpsl's IDN
functionality those components are not important.
Ref: https://github.com/curl/curl-for-win/issues/63Closes#12700
This clarifies the handling of server responses by folding the code for
the complicated protocols into their protocol handlers. This concerns
mainly HTTP and its bastard sibling RTSP.
The terms "read" and "write" are often used without clear context if
they refer to the connect or the client/application side of a
transfer. This PR uses "read/write" for operations on the client side
and "send/receive" for the connection, e.g. server side. If this is
considered useful, we can revisit renaming of further methods in another
PR.
Curl's protocol handler `readwrite()` method been changed:
```diff
- CURLcode (*readwrite)(struct Curl_easy *data, struct connectdata *conn,
- const char *buf, size_t blen,
- size_t *pconsumed, bool *readmore);
+ CURLcode (*write_resp)(struct Curl_easy *data, const char *buf, size_t blen,
+ bool is_eos, bool *done);
```
The name was changed to clarify that this writes reponse data to the
client side. The parameter changes are:
* `conn` removed as it always operates on `data->conn`
* `pconsumed` removed as the method needs to handle all data on success
* `readmore` removed as no longer necessary
* `is_eos` as indicator that this is the last call for the transfer
response (end-of-stream).
* `done` TRUE on return iff the transfer response is to be treated as
finished
This change affects many files only because of updated comments in
handlers that provide no implementation. The real change is that the
HTTP protocol handlers now provide an implementation.
The HTTP protocol handlers `write_resp()` implementation will get passed
**all** raw data of a server response for the transfer. The HTTP/1.x
formatted status and headers, as well as the undecoded response
body. `Curl_http_write_resp_hds()` is used internally to parse the
response headers and pass them on. This method is public as the RTSP
protocol handler also uses it.
HTTP/1.1 "chunked" transport encoding is now part of the general
*content encoding* writer stack, just like other encodings. A new flag
`CLIENTWRITE_EOS` was added for the last client write. This allows
writers to verify that they are in a valid end state. The chunked
decoder will check if it indeed has seen the last chunk.
The general response handling in `transfer.c:466` happens in function
`readwrite_data()`. This mainly operates now like:
```
static CURLcode readwrite_data(data, ...)
{
do {
Curl_xfer_recv_resp(data, buf)
...
Curl_xfer_write_resp(data, buf)
...
} while(interested);
...
}
```
All the response data handling is implemented in
`Curl_xfer_write_resp()`. It calls the protocol handler's `write_resp()`
implementation if available, or does the default behaviour.
All raw response data needs to pass through this function. Which also
means that anyone in possession of such data may call
`Curl_xfer_write_resp()`.
Closes#12480
The total timer is properly reset in MSTATE_INIT. MSTATE_CONNECT starts
with resetting the timer that is a start point for further multi states.
If file://, MSTATE_DO calls file_do() that should not reset the total
timer. Otherwise, the total time is always less than the pre-transfer
and the start transfer times.
Closes#12682
- `conn->sockfd` is set by `Curl_setup_transfer()`, but that
is called *after* the connection has been established
- use `conn->sock[FIRSTSOCKET]` instead
Follow-up to a0f94800d5Closes#12664
- 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
do not add a socket for POLLIN when the transfer does not want to send
(for example is paused).
Follow-up to 47f5b1a
Reported-by: bubbleguuum on github
Fixes#12632Closes#12633
```
lib/vtls/schannel.c:1201:22: warning: conversion to 'unsigned int' from 'int' may change the sign of the result [-Warith-conversion]
1201 | *extension_len = *list_len +
| ^
```
Closes#12616
Prior to this change if a send failed on a stream in an invalid state
(according to quiche) and not marked as closed (according to libcurl)
then the send function would return CURLE_SEND_ERROR.
We already have similar code for ngtcp2 to return CURLE_HTTP3 in this
case.
Caught by test test_07_upload.py: test_07_22_upload_parallel_fail.
Fixes https://github.com/curl/curl/issues/12590
Closes https://github.com/curl/curl/pull/12597
It was not working anymore since introduction of connection filters.
Also do not attempt to recover from a failing TLS negotiation with
CURLUSESSL_TRY.
Closes#12610
- Use CURLX_FUNCTION_CAST to suppress a function pointer assignment
warning.
a6bbc87f added lookups of some Windows API functions and then cast them
like `*(FARPROC*)&Curl_funcname = address`. Some versions of gcc warn
about that as breaking strict-aliasing rules so this PR changes those
assignments to use CURLX_FUNCTION_CAST.
Bug: https://github.com/curl/curl/pull/12581#issuecomment-1869804317
Reported-by: Marcel Raad
Closes https://github.com/curl/curl/pull/12602
