This changes the alert according to RFC 8446.
Fixes: #25402
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25547)
We use REF_PRINT_COUNT to dump out the value of various reference
counters in our code
However, we commonly use this macro after an increment or decrement. On
increment its fine, but on decrement its not, because the macro
dereferences the object holding the counter value, which may be freed by
another thread, as we've given up our ref count to it prior to using the
macro.
The rule is that we can't reference memory for an object once we've
released our reference, so lets fix this by altering REF_PRINT_COUNT to
accept the value returned by CRYPTO_[UP|DOWN]_REF instead. The
eliminates the need to dereference the memory the object points to an
allows us to use the call after we release our reference count
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25664)
When writing to a blocking quic stream, we sometimes get duplicate
transmitted data. This occurs when a call to quic_write_blocking has to
wait for space to become available in the ring buffer. When we do a
wait, the call sets *written to the value returned in args.total_written
as filled out by the calls to block_until_pred->quic_write_again.
However, the value there is based on the amount we requested, which is
only the remaining data that we didn't append in xso_sstream_write. So
if we call quic_write_blocking with a buffer of length X, and initially
append Y bytes, and write the remainig X-Y bytes via a block_until_pred
call, then *written will return with the value X-Y, even though we wrote
the full X bytes to the ring buffer.
Fix it by recording the initial amount appended into *written, and then
add the args.total_written value if we have to wait on more space
Fixesopenssl/project#924
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/26023)
When calling SSL_stream_reset on a QUIC stream object that has received
all data that is expected to be sent (i.e. when the sender has sent a
STREAM frame with the FIN bit set), we encounter the following segfault:
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7f0bd28 in ossl_quic_sstream_get_final_size (qss=0x0, final_size=0x0) at ssl/quic/quic_sstream.c:273
273 if (!qss->have_final_size)
(gdb) bt
0) 0x00007ffff7f0bd28 in ossl_quic_sstream_get_final_size (qss=0x0, final_size=0x0) at ssl/quic/quic_sstream.c:273
1) 0x00007ffff7ef65bf in quic_validate_for_write (xso=0x5555555efcb0, err=0x7fffffffd5e0) at ssl/quic/quic_impl.c:2513
2) 0x00007ffff7ef8ae3 in ossl_quic_stream_reset (ssl=0x5555555efcb0, args=0x0, args_len=0) at ssl/quic/quic_impl.c:3657
3) 0x00007ffff7ebdaa6 in SSL_stream_reset (s=0x5555555efcb0, args=0x0, args_len=0) at ssl/ssl_lib.c:7635
4) 0x0000555555557527 in build_request_set (
req_list=0x55555555ebd0 "neil1.txt neil2.txt neil3.txt neil4.txt neil5.txt neil6.txt neil7.txt neil8.txt neil9.txt neil10.txt neil11.txt neil12.txt neil13.txt neil14.txt neil15.txt neil16.txt neil17.txt neil18.txt neil19.txt "..., ssl=0x5555555b6f80)
at demos/guide/quic-hq-interop.c:545
5) 0x00005555555587b2 in main (argc=4, argv=0x7fffffffe568) at demos/guide/quic-hq-interop.c:941
This occurs because:
1) When the stream FIN bit is set, the quic stack frees the underlying
stream structures immediately within the QUIC stack
and
2) when SSL_stream_reset is called, the call stack indicates we call
quic_validate_for_write, which attempts to access the
xso->stream->sstream QUIC_SSTREAM object, which was already freed in
(1)
The fix I think is pretty straightforward. On receipt of a STREAM frame
with a FIN bit set, the QUIC stack sets the QUIC_STREAM object state to
QUIC_SSTREAM_STATE_DATA_RECVD, which means we can use that state to
simply assert that the stream is valid for write, which allows it to be
reset properly.
Fixes#25410
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25910)
Free the internal copy of parameter `value` on each early
exit.
Fixes#25906
Signed-off-by: Holger Dengler <dengler@linux.ibm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25926)
Otherwise we will calculate an incorrect header
size for higher stream ids and won't fit the
frame into the packet.
Fixes#25417
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25928)
Calling the functions SSL_CTX_set_cipher_list() or SSL_set_cipher_list() will
return the error "no cipher match" if no TLSv1.2 (or below) ciphers are enabled
after calling them. However this is normal behaviour for QUIC objects which do
not support TLSv1.2 ciphers. Therefore we should suppress that error in this
case.
Fixes#25878
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25886)
When processing a callback within libssl that applies to TLS the original
SSL object may have been created for TLS directly, or for QUIC. When making
the callback we must make sure that we use the correct SSL object. In the
case of QUIC we must not use the internal only SSL object.
Fixes#25788
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25874)
In some cases a QUIC SSL_CONNECTION object needs to get hold of a reference
to the original SSL object as created by the user. We should keep a
reference to it.
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25874)
Fixes#25471
Signed-off-by: Todd Short <todd.short@me.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25845)
In tls_setup_write_buffer() and tls_setup_read_buffer() the calculation
is different. Make them the same.
Fixes#25746
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25764)
Fixes#25625
Several error paths return 0 directly instead of going to err to clean
up the objects.
CLA: trivial
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25636)
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Hugo Landau <hlandau@devever.net>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25297)
1) Convert failures in keylog setup to trace messages for a warning-like
mechanism
2) Convert sslkeylogfile_cb to be a flag used to determine making a
direct call to the internal logging function
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Hugo Landau <hlandau@devever.net>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25297)
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Hugo Landau <hlandau@devever.net>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25297)
* instead of keeping an external reference count, just use the
BIO_up_ref call, and the BIO's callback mechanism to detect the
final free, for which we set keylog_bio to NULL
* Return an error from SSL_CTX_new_ex if the setup of the keylog file
fails
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Hugo Landau <hlandau@devever.net>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25297)
Add a config option for sslkeylog (disabled by default)
When enabled, SSL_CTX_new[_ex] becomes sensitive to the SSLKEYLOGFILE
environment variable. It records keylog callback messages to the file
specified in the environment variable according to the format specified
in https://www.ietf.org/archive/id/draft-thomson-tls-keylogfile-00.html
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Hugo Landau <hlandau@devever.net>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25297)
`sess` is not NULL at this point, and is freed on the success path, but
not on the error path. Fix this by going to the `err` label such that
`SSL_SESSION_free(sess)` is called.
CLA: trivial
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25643)
The quic implementation defined a set of LIST_* macros for list
manipulation, which conflicts with the generally support BSD api found
in the queue.h system header. While this isn't normally a problem, A
report arrived indicating that MacOSX appears to implicitly include
queue.h from another system header which causes definition conflicts.
As the openssl macros are internal only, it seems the most sensible
thing to do is place them in a well known namespace for our library to
avoid the conflict, so add an OSSL_ prefix to all our macros
Fixes#25516
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Saša Nedvědický <sashan@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/25519)
SSL_poll indicates that a stream which has had the fin bit set on it,
should generate SSL_POLL_EVENT_R events, so that applications can detect
stream completion via SSL_read_ex and SSL_get_error returning
SSL_ERROR_ZERO_RETURN.
However, the quic polling code misses on this, as a client that
completely reads a buffer after receipt has its underlying stream buffer
freed, loosing the fin status
We can however detect stream completion still, as a stream which has
been finalized, and had all its data read will be in the
QUIC_RSTREAM_STATE_DATA_READ state, iff the fin bit was set.
Fix it by checking in test_poll_event_r for that state, and generating a
SSL_POLL_EVENT_R if its found to be true, so as to stay in line with the
docs.
Fixesopenssl/private#627
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Sasa Nedvedicky <sashan@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25399)
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24648)
- Remove e_os.h include from "ssl_local.h"
- Added e_os.h into the files that need it now.
- Move e_os.h to be the very first include
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14344)
As reported by Alicja Kario, we ignored excess bytes after the
signature payload in TLS CertificateVerify Messages. These
should not be present.
Fixes: #25298
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25302)
Fixes#7940: Updated the compression check logic to improve protocol compliance. The code now returns `SSL_AD_DECODE_ERROR` when no compression method is provided in the ClientHello message. It returns `SSL_AD_ILLEGAL_PARAMETER` if the “null” compression method (0x00) is missing.
Additionally, refactored the related test code for enhanced readability and maintainability.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25255)
For SHAKE algorithms we now return 0 from EVP_MD_size().
So all the places that check for < 0 needed to change to <= 0
(Otherwise the behaviour will be to digest nothing in most cases).
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25285)
RFC8446 requires we send an illegal_parameter alert if we don't get a
key_share back from the server and our kex_modes require one. We were
instead reporting this as missing_extension.
Fixes#25040
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25059)
CCS records are ignore in TLSv1.3. But we should still call the msg_callback
anyway.
Fixes#25166
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25169)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24821)
PR #18345 added some code for an event queue. It also added a test for it.
Unfortunately this event queue code has never been used for anything.
Additionally the test was never integrated into a test recipe, so it never
actually gets invoked via "make test". This makes the code entirely dead,
unnecessarily bloats the size of libssl and causes a decrease in our
testing code coverage value.
We remove the dead code.
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25100)
If an HRR is sent then it MUST contain supported_versions according to the
RFC. We were sanity checking any supported_versions extension that was sent
but failed to verify that it was actually present.
Fixes#25041
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25068)
-trace option didn't cover early data message which resulted in
misleading logging.
CLA: trivial
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25026)
Add inline qualifier to avoid exporting a function for one unique use
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/24968)
Avoid using a fetched cipher that is decrypt-only
which is the case for 3DES from the fips provider.
Add a decrypt-only parameter to the EVP_CIPHER and test it
in libssl when fetching.
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25028)
The X509_NAME comparison function converts its arguments to DER using
i2d_X509_NAME before comparing the results using memcmp(). For every
invocation of the comparison function (of which there are many when
loading many certificates), it allocates two buffers of the appropriate
size for the DER encoding.
Switching to static buffers (possibly of X509_NAME_MAX size as defined
in crypto/x509/x_name.c) would not work with multithreaded use, e.g.,
when two threads sort two separate STACK_OF(X509_NAME)s at the same
time. A suitable re-usable buffer could have been added to the
STACK_OF(X509_NAME) if sk_X509_NAME_compfunc did have a void* argument,
or a pointer to the STACK_OF(X509_NAME) – but it does not.
Instead, copy the solution chosen in SSL_load_client_CA_file() by
filling an LHASH_OF(X509_NAME) with all existing names in the stack and
using that to deduplicate, rather than relying on sk_X509_NAME_find(),
which ends up being very slow.
Adjust SSL_add_dir_cert_subjects_to_stack() to keep a local
LHASH_OF(X509_NAME)s over the complete directory it is processing.
In a small benchmark that calls SSL_add_dir_cert_subjects_to_stack()
twice, once on a directory with one entry, and once with a directory
with 1000 certificates, and repeats this in a loop 10 times, this change
yields a speed-up of 5.32:
| Benchmark 1: ./bench 10 dir-1 dir-1000
| Time (mean ± σ): 6.685 s ± 0.017 s [User: 6.402 s, System: 0.231 s]
| Range (min … max): 6.658 s … 6.711 s 10 runs
|
| Benchmark 2: LD_LIBRARY_PATH=. ./bench 10 dir-1 dir-1000
| Time (mean ± σ): 1.256 s ± 0.013 s [User: 1.034 s, System: 0.212 s]
| Range (min … max): 1.244 s … 1.286 s 10 runs
|
| Summary
| LD_LIBRARY_PATH=. ./bench 10 dir-1 dir-1000 ran
| 5.32 ± 0.06 times faster than ./bench 10 dir-1 dir-1000
In the worst case scenario where many entries are added to a stack that
is then repeatedly used to add more certificates, and with a larger test
size, the speedup is still very significant. With 15000 certificates,
a single pass to load them, followed by attempting to load a subset of
1000 of these 15000 certificates, followed by a single certificate, the
new approach is ~85 times faster:
| Benchmark 1: ./bench 1 dir-15000 dir-1000 dir-1
| Time (mean ± σ): 176.295 s ± 4.147 s [User: 174.593 s, System: 0.448 s]
| Range (min … max): 173.774 s … 185.594 s 10 runs
|
| Benchmark 2: LD_LIBRARY_PATH=. ./bench 1 dir-15000 dir-1000 dir-1
| Time (mean ± σ): 2.087 s ± 0.034 s [User: 1.679 s, System: 0.393 s]
| Range (min … max): 2.057 s … 2.167 s 10 runs
|
| Summary
| LD_LIBRARY_PATH=. ./bench 1 dir-15000 dir-1000 dir-1 ran
| 84.48 ± 2.42 times faster than ./bench 1 dir-15000 dir-1000 dir-1
Signed-off-by: Clemens Lang <cllang@redhat.com>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/25056)
