Partial fix for #17064. Avoid excessive writes to the cache line
containing the refcount for an EVP_MD object to avoid extreme
cache contention when using a single EVP_MD at high frequency on
multiple threads. This changes performance in 3.0 from being double
that of 1.1 to only slightly higher than that of 1.1.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17857)
Also update and complete related documentation.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/17726)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17855)
Supports Linux, MacOS and FreeBSD
Disabled by default, enabled via `enabled-tfo`
Some tests
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8692)
Reviewed-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17848)
If TCP is being used, protocol = 0 is passed to init_client(), then
protocol == IPPROTO_TCP fails when attempting to set BIO_SOCK_NODELAY.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17838)
which replaces the link to the OpenSSL 3.0 Wiki.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17840)
We've had a report of a linker failure on some platforms (this one was
linux ARM) that apparently did not have pthread_atfork. It's strange that
this has not been reported before but the simplest solution is just to
remove this from the library since it isn't really used anyway.
Currently it is called to set up the fork handlers OPENSSL_fork_prepare,
OPENSSL_fork_parent and OPENSSL_fork_child. However all of those functions
are no-ops. This is a remnant from earlier code that got removed. We can
safely remove it now.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17842)
Including running the oqsprovider external test in the
CI external test build.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17832)
Fixes an issue where, when the test suite was being run with the older
Test::Harness package, the test suite would not complete correctly due
to evaluation of $harness->runtests()->has_errors, which is only
available for the newer TAP::Parser::Aggregator code path.
Fixes#17818.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17823)
Two demos are provided: one using RSA-PSS directly in which a digest
must be provided, and one using RSA-PSS with the EVP_DigestSign APIs
which performs the hashing for you.
Fixes#14113.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17800)
This patch will allow the SM4-GCM function to leverage the SM4
high-performance CTR crypto interface already implemented for ARM,
which is faster than current single block cipher routine used
for GCM
It does not address the acceleration of GHASH function of GCM,
which can be a future task, still we can see immediate uplift of
performance (up to 4X)
Before this patch:
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
SM4-GCM 186432.92k 394234.05k 587916.46k 639365.12k 648486.91k 652924.25k
After the patch:
SM4-GCM 193924.87k 860940.35k 1696083.71k 2302548.31k 2580411.73k 2607398.91k
Signed-off-by: Daniel Hu <Daniel.Hu@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17814)
This demo optionally accepts a single command line argument, allowing
the output length to be specified.
Fixes#14106.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17803)
Also check return value of functions that call BIO_new() internally
such as dup_bio_out().
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17421)
Before PR #15310, which reworked how build files (Makefile, ...) were
generated, everything was done when configuring, so configdata.pm
could depend on build file templates and we'd get away with it.
However, since building configdata.pm is now independent of the build
file templates, that dependency is unnecessary, and would lead to
surprises of the build file template is updated, with an unexpected
full reconfiguration as a result, when all that's needed is to run
configdata.pm with no flags to get the build file re-generated.
This change is therefore a completion of what was forgotten in #15310.
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17756)
The EOR3 instruction is implemented with .inst, and the code here is enabled
using run-time detection of the CPU capabilities, so no need to explicitly
ask for the sha3 extension.
Fixes#17773
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17795)
This offers both a known answer test with fixed keys and also
demonstrates a more realistic usage with random keys.
Fixes#14118.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17799)
Fix a bug in `openssl s_server -WWW` where it would attempt to invoke
`SSL_sendfile` if `-ktls -sendfile` was passed on the command line, even
if KTLS has not actually been enabled, for example because it is not
supported by the host. Since `SSL_sendfile` is only supported when KTLS
is actually being used, this resulted in a failure to serve requests.
Fixes#17503.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17788)
Set the environment variable MALLOC_PERTURB_ during tests to perturb the
output from OPENSSL_malloc() calls (see the mallopt man page for details
about this environment variable). This could be a low cost way of spotting
uninit reads in "make test" runs in some situations.
In tests I have found it to be a little unreliable (sometimes it seemed to
not perturb the output for inexplicable reasons) - but since this is easy
to implement I think it is worthwhile.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17768)
FreeBSD does not provide a global timezone variable containing the
offset to UTC. Instead, FreeBSD's libc includes a legacy timezone
function dating back to Version 7 AT&T UNIX. As a result,
asn1_string_to_time_t currently fails to compile on FreeBSD as it
subtracts a function from a time_t value:
../crypto/asn1/a_time.c:625:37: error: invalid operands to binary expression ('time_t' (aka 'long') and 'char *(int, int)')
timestamp_utc = timestamp_local - timezone;
~~~~~~~~~~~~~~~ ^ ~~~~~~~~
1 error generated.
However, FreeBSD's libc does include a non-standard (but widely
available) timegm function which converts a struct tm directly to a
UTC time_t value. Use this on FreeBSD instead of mktime.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17765)
We need to handle signatures with and without digest algs
and we generalize the ossl_cms_ecdsa_dsa_sign() function
to other algorithms that are handled in the same way.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17733)
This fixes a bug whereby BN_mod_exp2_mont can dereference a NULL pointer
if BIGNUM argument m represents zero.
Regression test added. Fixes#17648.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17783)
`openssl req -x509` has code allowing it to generate certificates from CSRs
as a replacement for `openssl x509`, but a bug prevents it from working
properly. -CA and -CAkey can now be passed to generate a CA-signed
certificate as documented in openssl-req(1).
Regression testing has been added to `openssl req`.
Fixes#17736.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17782)
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17763)
`SSL_kECDHE` and `SSL_kEECDH`, and `SSL_kDHE` and `SSL_kEDH` are already
marked as aliases of each other in the headers.
This commit, for each pair, replaces the leftover uses of the latter
synonym with the first one, which is considered more common.
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17763)
Reverting to using swapcontext() when compiling with clang on BTI-enabled
builds fixes the BTI setjmp() failure seen when running asynctest.
The issue with setjmp/longjmp is a known clang bug: see
https://github.com/llvm/llvm-project/issues/48888
Change-Id: I6eeaaa2e15f402789f1b3e742038f84bef846e29
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17698)
Since the potential failure of memory allocation, it
should be better to check the return value of the
OPENSSL_strndup(), like x509v3_add_len_value().
And following the comment of 'if (astrlen < 0)',
return -1 if fails.
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17737)
As the potential failure of the OPENSSL_zalloc(), the OSSL_LIB_CTX_new()
could return NULL.
Therefore, it should be better to check it and return error if fails in
order to guarantee the success of the initialization.
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17738)
This option is no longer set by default from OpenSSL 3.0.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/17748)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17741)
