!$disabled{mdc2} was used to determine if DES files should be included
in providers/liblegacy.a. Use !$disabled{des} instead.
Fixes#13865
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13866)
When a fetch is attempted simultaneously from multiple threads then both
threads can attempt to construct the method. However only one of those
will get added to the global evp method store. The one that "lost" the
race to add the method to the global evp method store ended up with the
fetch call returning NULL, instead of returning the method that was
already available.
Fixes#13682
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13660)
The above function was running while holding the store lock with a read
lock. Unfortunately it actually modifies the store, so a write lock is
required instead.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13660)
The primary DRBG may be shared across multiple threads and therefore
we must use locking to access it. Previously we were enabling that locking
lazily when we attempted to obtain one of the child DRBGs. Part of the
process of enabling the lock, is to create the lock. But if we create the
lock lazily then it is too late - we may race with other threads where each
thread is independently attempting to enable the locking. This results
in multiple locks being created - only one of which "sticks" and the rest
are leaked.
Instead we enable locking on the primary when we first create it. This is
already locked and therefore we cannot race.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13660)
The function ossl_lib_ctx_generic_new() modifies the exdata. This may
be simultaneously being modified by other threads and therefore we need
to make sure we take the lock before doing so.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13660)
The provider operation_bits array can see concurrent access by multiple
threads and can be reallocated at any time. Therefore we need to ensure
that it is appropriately locked.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13660)
The FIPS implementation of the ossl_ctx_thread_stop function needs to
use an OSSL_LIB_CTX - but gets passed a provctx as an argument. It was
assuming that these are the same thing (which was true at one point
during development) - but that is no longer the case. The fix is to
get the OSSL_LIB_CTX out of the provctx.
Fixes#13469
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13660)
Also point out in the documenting comment that a non-expired issuer is preferred.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13805)
when possible using the getauxval equivalent which has similar ids as Linux, instead of bad instructions catch approach.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13650)
In some edge cases BN_CTX_end was being called without first calling
BN_CTX_start. This creates a situation where the state of the big
number allocator is corrupted and may lead to crashes.
Fixes#13812
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13813)
...in case the candidate issuer cert is identical to the target cert.
This is the v3.0.0 variant of #13749 fixing #13739 for v1.1.1.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13762)
We've spread around FETCH_FAILED errors in quite a few places, and
that gives somewhat crude error records, as there's no way to tell if
the error was unavailable algorithms or some other error at such high
levels.
As an alternative, we take recording of these kinds of errors down to
the fetching functions, which are in a much better place to tell what
kind of error it was, thereby relieving the higher level calls from
having to guess.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13467)
The openssl code base has only a few occurrences of 'unsigned const char'
(15 occurrences), compared to the more common 'const unsigned char' (4420
occurrences).
While the former is not illegal C, mixing the 'const' keyword (a 'type
qualifier') in between 'unsigned' and 'char' (both 'type specifiers') is a
bit odd.
The background for writing this patch is not to be pedantic, but because
the 'opmock' program (used to mock headers for unit tests) does not accept
the 'unsigned const char' construct. While this definitely is a bug in
opmock or one of its dependencies, openssl is the only piece of software we
are using in combination with opmock that has this construct.
CLA: trivial
Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/13722)
According to the relevant standards, the valid range for SM2 private
keys is [1, n-1), where n is the order of the curve generator.
For this reason we cannot reuse the EC validation function as it is, and
we introduce a new internal function `sm2_key_private_check()`.
Partially fixes https://github.com/openssl/openssl/issues/8435
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13359)
CreateFiber and ConvertThreadToFiber are not allowed in Windows Store
(Universal Windows Platform) apps since they have been replaced by
their Ex variants which have a new dwFlags parameter.
This flag allows the fiber to do floating-point arithmetic in the
fiber on x86, which would silently cause corruption otherwise since
the floating-point state is not switched by default.
Switch to these "new" APIs which were added in Vista.
See: https://docs.microsoft.com/en-us/windows/win32/api/winbase/nf-winbase-createfiberex#parameters
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12400)
If everything has already been initialised we can check this with a
single test at the beginning of OPENSSL_init_crypto() and therefore
reduce the amount of time spent in this function. Since this is called
via very many codepaths this should have significant performance benefits.
Partially fixes#13725 and #13578
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13733)
We add an implementation for CRYPTO_atomic_or() and CRYPTO_atomic_load()
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13733)
The EVP lib was calling EVP_CIPHER_CTX_block_size(), which in turn calls
EVP_CIPHER_block_size() in order to find the block_size in every
EVP_EncryptUpdate() call. This adds a surprising amount of overhead when
using speed to test aes-128-cbc. Since we're in the EVP lib itself, we can
just directly access this value.
To test performance I ran the command:
openssl speed -evp aes-128-cbc -bytes 16 -seconds 30
For the before and after, I ran this twice and discarded the first result
to "warm up" my machine.
Before:
aes-128-cbc 716949.71k
After:
aes-128-cbc 742807.11k
This represents a performance improvement of about 4%
Partially fixes#13407
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13734)
Most of the time we don't have any explicit settings when loading a
config file. Therefore we optimise things so that we don't need to use
a lock in that instance.
Partially addresses performance issues in #13725
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13731)
EVP_CIPHER already caches certain constants so that we don't have to
query the provider every time. We do the same thing with EVP_MD constants.
Without this we can get performance issues, e.g. running "speed" with
small blocks of data to digest can spend a long time in EVP_MD_size(),
which should be quick.
Partialy fixes#13578
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13730)
Previously we cached the cipher constants in EVP_CIPHER_fetch(). However,
this means we do the caching every time we call that function, even if
the core has previusly fetched the cipher and cached it already. This
means we can end up re-caching the constants even though they are already
present. This also means we could be updating these constants from
multiple threads at the same time.
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/13730)
When X509_ATTRIBUTE_create() receives an invalid NID (e.g., -1), return
failure rather than silently constructing a broken X509_ATTRIBUTE object
that might cause NULL pointer accesses later on. This matters because
X509_ATTRIBUTE_create() is used by API functions like PKCS7_add_attribute(3)
and the NID comes straight from the user.
This bug was found while working on LibreSSL documentation.
Reviewed-by: Theo Buehler <tb@openbsd.org>
CLA: trivial
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12052)
The configuration option 'no-rsa' was dropped with OpenSSL 1.1.0, so
this is simply a cleanup of the remains.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13700)
Some of the handling of no-deprecated stuff wasn't quite complete, or
even plain wrong.
This restores i2d_PublicKey() to be able to handle EVP_PKEYs with
legacy internal keys.
This also refactors the DSA key tests in test/evp_extra_test.c to use
EVP functionality entirely.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13706)
The params[0].data is set to a non-NULL value, but params[0].data_size
is always zero. This confuses get_string_internal, which creates 1 byte
string with uninitialized content.
When OSSL_PARAM_construct_utf8_string is used, the data_size is set
correctly.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13699)
In most error cases, EVP_PKEY_CTX_dup() would only free the EVP_PKEY_CTX
without freeing the duplicated contents.
Fixes#13503
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13661)
This introduces a separate activation counter, and the function
ossl_provider_deactivate() for provider deactivation.
Something to be noted is that if the reference count goes down to
zero, we don't care if the activation count is non-zero (i.e. someone
forgot to call ossl_provider_deactivate()). Since there are no more
references to the provider, it doesn't matter.
The important thing is that deactivation doesn't remove the provider
as long as there are references to it, for example because there are
live methods associated with that provider, but still makes the
provider unavailable to create new methods from.
Fixes#13503Fixes#12157
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13661)
Because decoders are coupled with keymgmts from the same provider,
ours need to produce provider side keys the same way. Since our
keymgmts create key data with the provider library context, so must
our decoders.
We solve with functions to adjust the library context of decoded keys,
and use them.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13661)
So far, the MSBLOB and PVK writers could only handle EVP_PKEYs with
legacy internal keys.
Specially to be able to compile the loader_attic engine, we use the C
macro OPENSSL_NO_PROVIDER_CODE to avoid building the provider specific
things when we don't need them. The alternative is to suck half of
crypto/evp/ into loader_attic, and that's just not feasible.
Fixes#13503
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13661)
OSSL_DECODER_CTX_new_by_EVP_PKEY() would keep copies of all the
EVP_KEYMGMTs it finds.
This turns out to be fragile in certain circumstances, so we switch to
fetch the appropriate EVP_KEYMGMT when it's time to construct an
EVP_PKEY from the decoded data instead. This has the added benefit
that we now actually use the property query string that was given by
the caller for these fetches.
Fixes#13503
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13661)
