scan-build-6 otherwise warns, saying: warning: The left operand of '>='
is a garbage value otherwise, which is false.
Later scan-builds don't claim this on the same code.
Closes#7248
When a connection is disassociated from a transfer, the Session ID entry
should remain.
Regression since 7f4a9a9 (shipped in libcurl 7.77.0)
Reported-by: Gergely Nagy
Reported-by: Paul Groke
Fixes#7222Closes#7230
FD_SETSIZE is irrelevant when using poll. So ensuring that the file
descriptor is smaller than FD_SETSIZE in VALID_SOCK, can cause
multi_wait to ignore perfectly valid file descriptors and simply wait
for 1s to avoid hammering the CPU in a busy loop.
Fixes#7240Closes#7241
The latest GnuTLS-3.7.2 implements disable switch for TLSv1.3 compatible
mode for middle box but it is enabled by default, which is unnecessary
for QUIC.
Fixes#6896Closes#7202
Removing expired cookies needs to be a fast operation since we want to
be able to perform it often and speculatively. By tracking the timestamp
of the next known expiration we can exit early in case the timestamp is
in the future.
Closes: #7172
Reviewed-by: Daniel Stenberg <daniel@haxx.se>
Trying dual-stack on some embedded platform, I noticed that quite
frequently (20%) libCurl starts from IPv4 regardless the Happy Eyeballs
timeout value. After debugging this issue, I noticed that this happens
if c-ares resolver response for IPv6 family comes before IPv4 (which was
randomly happening in my tests).
In such cases, because libCurl puts the last resolver response on top of
the address list, when IPv4 resolver response comes after IPv6 one - the
IPv4 family starts the connection phase instead of IPv6 family.
The solution for this issue is to always put IPv6 addresses on top of
the address list, regardless the order of resolver responses.
Bug: https://curl.se/mail/lib-2021-06/0003.htmlCloses#7188
Commit 1c1d9f1aff removed the last use
for the inet_pton.h headerfile, this removes the inclusion of the
header.
Closes: #7182
Reviewed-by: Daniel Stenberg <daniel@haxx.se>
Fixes potential hang in accept by using select + non-blocking accept.
Fixes potential hang in peer check by replacing the send/recv check with
a getsockname/getpeername check.
Adds length check for returned sockaddr data.
Closes#7144
For options that pass in lists or strings that are subsequently parsed
and must be correct. This broadens the scope for the option previously
known as CURLE_TELNET_OPTION_SYNTAX but the old name is of course still
provided as a #define for existing applications.
Closes#7175
Follow-up to 1a0ebf6632
- Check the return code to Curl_inet_pton() in two instances, even
though we know the input is valid so the functions won't fail.
- Clear the 'struct sockaddr_in' struct before use so that the
'sin_zero' field isn't left uninitialized.
Detected by Coverity.
Assisted-by: Harry Sintonen
Closes#7163
As host names are case insensitive, the use of case sensitive hashing
caused unnecesary cache misses and therefore lost performance. This
lowercases the hash key.
Reported-by: Harry Sintonen
Fixes#7159Closes#7161
This avoids a TCP reset (RST) if the server initiates a connection
shutdown by sending an SSL close notify alert and then closes the TCP
connection.
For SSL connections, usually the server announces that it will close the
connection with an SSL close notify alert. curl should read this alert.
If curl does not read this alert and just closes the connection, some
operating systems close the TCP connection with an RST flag.
See RFC 1122, section 4.2.2.13
If curl reads the close notify alert, the TCP connection is closed
normally with a FIN flag.
The new code is similar to existing code in the "SSL shutdown" function:
try to read an alert (non-blocking), and ignore any read errors.
Closes#7095
This function might get called for an easy handle for which the session
cache hasn't been setup. It now just returns a "miss" in that case.
Reported-by: Christoph M. Becker
Fixes#7148Closes#7153
Resolving the case insensitive host name 'localhost' now returns the
addresses 127.0.0.1 and (if IPv6 is enabled) ::1 without using any
resolver.
This removes the risk that users accidentally resolves 'localhost' to
something else. By making sure 'localhost' is always local, we can
assume a "secure context" for such transfers (for cookies etc).
Closes#7039
Also, use a single function library-wide for detecting if a given hostname is
a numerical IP address.
Reported-by: Harry Sintonen
Fixes#7146Closes#7149
To prevent previous ones to get reused on subsequent requests. Matches
how the built-in HTTP code works. Makes test 90 to 93 work.
Add test 90 to 93 in travis.
Closes#7139
In a3268eca79 this code was changed to use the ALPN_H2 constant
instead of the NGHTTP2_PROTO_ALPN constant. However, these constants are
not the same. The nghttp2 constant included the length of the string,
like this: "\x2h2". The ALPN_H2 constant is just "h2". Therefore we need
to re-add the length of the string to the ALPN buffer.
Closes#7138
When the SecCertificateCopyCommonName function fails, it leaves
common_name in a invalid state so CFStringCompare uses the invalid
result, causing EXC_BAD_ACCESS.
The fix is to check the return value of the function before using the
name.
Closes#7126
Latest version of quiche requires the application to pass the peer
address of received packets, and it provides the address for outgoing
packets back.
Closes#7120
In some situations, it was possible that a transfer was setup to
use an specific IP version, but due do DNS caching or connection
reuse, it ended up using a different IP version from requested.
This commit changes the effect of CURLOPT_IPRESOLVE from simply
restricting address resolution to preventing the wrong connection
type being used, when choosing a connection from the pool, and
to restricting what addresses could be used when establishing
a new connection.
It is important that all addresses versions are resolved, even if
not used in that transfer in particular, because the result is
cached, and could be useful for a different transfer with a
different CURLOPT_IPRESOLVE setting.
Closes#6853
AmiSSL replaces many functions with macros. Curl requires pointer
to some of these functions. Thus, we have to encapsulate these macros:
SHA256_Init, SHA256_Update, SHA256_Final, X509_INFO_free.
Bug: https://github.com/jens-maus/amissl/issues/15
Co-authored-by: Daniel Stenberg <daniel@haxx.se>
Closes#7099
Check the FIRST for all other protocols.
This fixes a timeout in an ftps download. The server sends a TLS
close_notify message in the same packet as the file data. The
close_notify seems to not be handled in the schannel_recv function, so
libcurl is not aware that the server has closed the connection. Thus
libcurl ends up waiting for action on the socket until a timeout is
reached. With the secondary socket check added to the data_pending
function, the close_notify is properly handled, and the ftps transfer
terminates as expected.
Fixes#7068Closes#7069
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
Some of the time, we get a HYPER_TASK_EMPTY response before the status
line, headers, and body have been read. Previously, that would cause us
to poll again, leading to a 1 second timeout.
The HYPER_TASK_EMPTY docs say:
The value of this task is null (does not imply an error).
So, if we receive a HYPER_TASK_EMPTY, continue on with processing the
response.
Reported-by: Kevin Burke
Fixes#7064Closes#7070
As of commit 54e7475, these flags would only be set when using a new
credential handle. When re-using an existing credential handle, the
flags would not be set.
Closes https://github.com/curl/curl/pull/7051
The Curl_resolv() had special code (when built in debug mode) for when
resolving the host name "LocalHost" (using that exact casing). It would
then get the host name from the --interface option instead.
This development-only feature was not used by anything (anymore) and we
have the --resolve feature if we want to play similar tricks properly
going forward.
Closes#7044
Otherwise the old value would linger from a previous use and would mess
up the network speed cap logic.
Reported-by: Ymir1711 on github
Fixes#7042Closes#7043
Writing the cookie file has multiple error conditions, and was using an
int with magic numbers to report the different error (which in turn were
disregarded anyways). This moves reporting to use a CURLcode value.
Lightly-touched-by: Daniel Stenberg
Closes#7037Closes#6749
strstore() is defined as a strdup which ensures to free the target
pointer before duping the source char * into it. Make use of it in
two more cases where it can simplify the code.
Comments in the cookie code were a bit all over the place in terms of
style and wording. This takes a stab at cleaning them up by keeping to
a single style and overall shape. Some comments are moved a little and
some removed alltogether due to being redundant. No functional changes
have been made,
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
Assumed to be a minor coding style improvement with no behavior change.
A modern compiler is expected to have the calculation optimized during
compilation. It may be deemed okay even if that's not the case, since
the added overhead is considered very low.
Closes#7032
Also added 'CURL_SMALLSENDS' to make Curl_write() send short packets,
which helped verifying this even more.
Add test 363 to verify.
Reported-by: ustcqidi on github
Fixes#6950Closes#7024
Previously this logic would cap the send to CURL_MAX_WRITE_SIZE bytes,
but for the situations where a larger upload buffer has been set, this
function can benefit from sending more bytes. With default size used,
this does the same as before.
Also changed the storage of the size to an 'unsigned int' as it is not
allowed to be set larger than 2M.
Also added cautions to the man pages about changing buffer sizes in
run-time.
Closes#7022
A reused transfer handle could otherwise reuse the previous leftover
buffer and havoc would ensue.
Reported-by: sergio-nsk on github
Fixes#7018Closes#7021
