If two threads both attempt to load the same provider at the same time,
they will first both check to see if the provider already exists. If it
doesn't then they will both then create new provider objects and call the
init function. However only one of the threads will be successful in adding
the provider to the store. For the "losing" thread we should still return
"success", but we should deinitialise and free the no longer required
provider object, and return the object that exists in the store.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15854)
Update use_fallbacks to zero when we add a provider to the store rather
than when we activate it. Its only at the point that we add it to the store
that it is actually usable and visible to other threads.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15854)
Previously we instantiated all the predefined providers at the point that
we create the provider store. Instead we move them to be instantiated as we
need them.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15854)
We were deferring the initial creation of the child providers until the
first fetch. This is a carry over from an earlier iteration of the child
lib ctx development and is no longer necessary. In fact we need to init
the child providers immediately otherwise not all providers quite init
correctly.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15270)
Also remove redundant opt_name() and make names of opt_{i,u}ntmax() consistent.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15111)
A failure to obtain a lock would have resulted in much badness, now it results
in a failure return.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14773)
This corresponds to the |info| field in EVP_PKEY_ASN1_METHOD, as well
as the generic use of OBJ_nid2ln() as a one line description.
We also add the base functionality to make use of this field.
Fixes#14514
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14656)
Some functions that lock things are void, so we just return early.
Also make ossl_namemap_empty return 0 on error. Updated the docs, and added
some code to ossl_namemap_stored() to handle the failure, and updated the
tests to allow for failure.
Fixes: #14230
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14238)
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14319)
If someone calls an EVP_PKEY_get0*() function then we create a legacy
key and cache it in the EVP_PKEY - but it doesn't become an "origin" and
it doesn't ever get updated. This will be documented as a restriction of
the EVP_PKEY_get0*() function with provided keys.
Fixes#14020
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14319)
We need to add something for the 'tests' target to depend on, so a
special syntax for those is introduced:
DEPEND[|tests|]=fipsmodule.cnf
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14320)
We don't want to hold a read lock when calling a user supplied callback.
That callback could do anything so the risk of a deadlock is high.
Instead we collect all the names first inside the read lock, and then
subsequently call the user callback outside the read lock.
Fixes#14225
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14250)
Without this, a provider has no way to know that an application
has finished with the array it returned earlier. A non-caching provider
requires this information.
Fixes#12974
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12974)
Add an argument to PROVIDER_try_load() that permits a provider to be
loaded without changing the fallback status. This is useful when an
additional provider needs to be loaded without perturbing any other setup.
E.g. adding mock providers as part of unit testing.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13652)
This fixes a race condition where the index to the cache location was found
under a read lock and a later write lock set the cache entry. The issue being
that two threads could get the same location index and then fight each other
over writing the cache entry. The most likely outcome is a memory leak,
however it would be possible to set up an invalid cache entry.
The operation cache was a fixed sized array, once full an assertion failed.
The other fix here is to convert this to a stack. The code is simplified and
it avoids a cache overflow condition.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14062)
The EVP_PKEY operation_cache caches references to provider side key
objects that have previously been exported for this EVP_PKEY, and their
associated key managers. The cache may be updated from time to time as the
EVP_PKEY is exported to more providers. Since an EVP_PKEY may be shared by
multiple threads simultaneously we must be careful to ensure the cache
updates are locked.
Fixes#13818
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13987)
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)
Refactor them into inline ossl_ends_with_dirsep function in
internal/cryptlib.h.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13306)
doc/internal/man7/deprecation.pod was unclear in some areas, and
included general documentation that has no place there.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/13240)
This removes a TODO.
This function is not needed since any place that needs to do the
conversion normally has a special case name2nid table.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13202)
