This fixes an issue with a mix of atexit() usage in DLL and statically linked
libcrypto that came out in the test suite on NonStop, which has slightly
different DLL unload processing semantics compared to Linux. The change
allows a build configuration to select whether to register OPENSSL_cleanup()
with atexit() or not, so avoid situations where atexit() registration causes
SIGSEGV.
INSTALL.md and CHANGES.md have been modified to include and describe this
option.
The no-atexit option has been added to .github/workflows/run-checker-daily.yml.
Fixes: #23135
Signed-of-by: Randall S. Becker <randall.becker@nexbridge.ca>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Todd Short <todd.short@me.com>
(Merged from https://github.com/openssl/openssl/pull/23394)
master/3.2 was never vulnerable to CVE-2023-5678 since it was fixed before
it was released.
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23432)
To demonstrate the use of RCU locks, convert CONF_MOD api to using rcu
rather than RW locks
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22729)
Introduce an RCU lock implementation as an alternative locking mechanism
to openssl. The api is documented in the ossl_rcu.pod
file
Read side implementaiton is comparable to that of RWLOCKS:
ossl_rcu_read_lock(lock);
<
critical section in which data can be accessed via
ossl_derefrence
>
ossl_rcu_read_unlock(lock);
Write side implementation is:
ossl_rcu_write_lock(lock);
<
critical section in which data can be updated via
ossl_assign_pointer
and stale data can optionally be scheduled for removal
via ossl_rcu_call
>
ossl_rcu_write_unlock(lock);
...
ossl_synchronize_rcu(lock);
ossl_rcu_call fixup
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22729)
OSSL_PKEY_PARAM_DERIVE_FROM_PQ must be OSSL_PKEY_PARAM_RSA_DERIVE_FROM_PQ
(note the missing '_RSA').
Signed-off-by: Ingo Franzki <ifranzki@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/23438)
This reverts commit 2b74e75331.
The commit was wrong. With 3.x versions the engines must be themselves
responsible for creating their EVP_PKEYs in a way that they are treated
as legacy - either by using the respective set1 calls or by setting
non-default EVP_PKEY_METHOD.
The workaround has caused more problems than it solved.
Fixes#22945
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23063)
In writing the quic stateless reset test we found that the quic rx code
wasn't checking for stateless reest conditions, as the SRT frames were
getting discarded due to failed lcdim lookups. Move the SRT check above
the lcdim lookup in the rx path to ensure we handle SRT properly in the
client.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23384)
QUIC supports the concept of stateless reset, in which a specially
crafted frame is sent to a client informing it that the QUIC state
information is no longer available, and the connection should be closed
immediately. Test for proper client support here
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23384)
For some reason, those functions (and the _init functions too) would
raise EVP_R_OPERATION_NOT_SUPPORTED_FOR_THIS_KEYTYPE when the passed
ctx is NULL, and then not check if the provider supplied the function
that would support these libcrypto functions.
This corrects the situation, and has all those libcrypto functions
raise ERR_R_PASS_NULL_PARAMETER if ctx is NULL, and then check for the
corresponding provider supplied, and only when that one is missing,
raise EVP_R_OPERATION_NOT_SUPPORTED_FOR_THIS_KEYTYPE.
Because 0 doesn't mean error for EVP_PKEY_verify(), -1 is returned when
ERR_R_PASSED_NULL_PARAMETER is raised. This is done consistently for all
affected functions.
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23411)
bind_afalg calls afalg_aes_cbc which allocates
cipher_handle->_hidden global object(s)
but if one of them fails due to out of memory,
the function bind_afalg relies on the engine destroy
method to be called. But that does not happen
because the dynamic engine object is not destroyed
in the usual way in dynamic_load in this case:
If the bind_engine function fails, there will be no
further calls into the shared object.
See ./crypto/engine/eng_dyn.c near the comment:
/* Copy the original ENGINE structure back */
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23409)
