When freeing the last reference to an EVP_PKEY there is no point in
taking the lock for the key. It is the last reference and is being freed
so must only be being used by a single thread.
This should not have been the source of any contention so its unclear to
what extent this will improve performance. But we should not be locking
when we don't need to.
Partially fixes#20286
Reviewed-by: Hugo Landau <hlandau@openssl.org>
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/20932)
EVP_PKEY_eq() is used to check, among other things, if a certificate
public key corresponds to a private key. When the private key belongs to
a provider that does not allow to export private keys this currently
fails as the internal functions used to import/export keys ignored the
selection given (which specifies that only the public key needs to be
considered) and instead tries to export everything.
This patch allows to propagate the selection all the way down including
adding it in the cache so that a following operation actually looking
for other selection parameters does not mistakenly pick up an export
containing only partial information.
Signed-off-by: Simo Sorce <simo@redhat.com>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19648)
Since OPENSSL_malloc() and friends report ERR_R_MALLOC_FAILURE, and
at least handle the file name and line number they are called from,
there's no need to report ERR_R_MALLOC_FAILURE where they are called
directly, or when SSLfatal() and RLAYERfatal() is used, the reason
`ERR_R_MALLOC_FAILURE` is changed to `ERR_R_CRYPTO_LIB`.
There were a number of places where `ERR_R_MALLOC_FAILURE` was reported
even though it was a function from a different sub-system that was
called. Those places are changed to report ERR_R_{lib}_LIB, where
{lib} is the name of that sub-system.
Some of them are tricky to get right, as we have a lot of functions
that belong in the ASN1 sub-system, and all the `sk_` calls or from
the CRYPTO sub-system.
Some extra adaptation was necessary where there were custom OPENSSL_malloc()
wrappers, and some bugs are fixed alongside these changes.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19301)
In all initializing functions for functionality that use an EVP_PKEY, the
coded logic was to find an KEYMGMT implementation first, and then try to
find the operation method (for example, SIGNATURE implementation) in the
same provider.
This implies that in providers where there is a KEYMGMT implementation,
there must also be a SIGNATURE implementation, along with a KEYEXCH,
ASYM_CIPHER, etc implementation.
The intended design was, however, the opposite implication, i.e. that
where there is a SIGNATURE implementation, there must also be KEYMGMT.
This change reverses the logic of the code to be closer to the intended
design.
There is a consequence; we now use the query_operation_name function from
the KEYMGMT of the EVP_PKEY given by the EVP_PKEY_CTX (ultimately given by
the application). Previously, we used the query_operation_name function
from the KEYMGMT found alongside the SIGNATURE implementation.
Another minor consequence is that the |keymgmt| field in EVP_PKEY_CTX
is now always a reference to the KEYMGMT of the |pkey| field if that
one is given (|pkey| isn't NULL) and is provided (|pkey->keymgmt|
isn't NULL).
Fixes#16614
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/16725)
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15974)
For functions that exist in 1.1.1 provide a simple aliases via #define.
Fixes#15236
Functions with OSSL_DECODER_, OSSL_ENCODER_, OSSL_STORE_LOADER_,
EVP_KEYEXCH_, EVP_KEM_, EVP_ASYM_CIPHER_, EVP_SIGNATURE_,
EVP_KEYMGMT_, EVP_RAND_, EVP_MAC_, EVP_KDF_, EVP_PKEY_,
EVP_MD_, and EVP_CIPHER_ prefixes are renamed.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15405)
If evp_keymgmt_util_try_import() allocated keydata, and the import
itself fails, it should deallocate keydata.
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/15008)
The TODO suggest a possible refactoring. The refactoring doesn't seem
necessary at this stage. If it is required later it can be done without
affecting external APIs - so just remove the TODO.
Fixes#14397
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14888)
To avoid mutating key data add OSSL_FUNC_KEYMGMT_DUP function
to the provider API and implement it for all asym-key key
managements.
Use it when copying everything to an empty EVP_PKEY
which is the case with EVP_PKEY_dup().
Fixes#14658
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/14793)
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)
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)
When the internal key is changed, we must count it as muted, so that
next time the affected key is considered for an operation, it gets
re-exported to the signing provider. In other words, this will clear
the EVP_PKEY export cache when the next export attempt occurs.
This also updates evp_keymgmt_util_export_to_provider() to actually
look at the dirty count for provider native origin keys, and act
appropriately.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14056)
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 is purely to allow exporting without having to repeatedly specify
the keymgmt and keydata from the EVP_PKEY.
Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12853)
This makes the following functions available for libcrypto code:
evp_keymgmt_util_try_import() - callback function
evp_keymgmt_util_assign_pkey() - assigns keymgmt and keydata to an EVP_PKEY
evp_keymgmt_util_make_pkey() - creates an EVP_PKEY from keymgmt and keydata
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12410)
Even if there is no data to import we should still create an empty key.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11898)
evp_keymgmt_util_copy() didn't treat the case to->keymgmt correctly.
The proper change is to use from->keymgmt when to->keymgmt is NULL.
Fixes coverity #1462553
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/11668)
evp_keymgmt_util_get_deflt_digest_name() is a refactor of the provider
side key part of EVP_PKEY_get_default_digest_name(), that takes
EVP_KEYMGMT and provider keydata pointers instead of an EVP_PKEY
pointer.
We also ensure that it uses SN_undef as the default name if the
provider implementation gave us an empty string, since this is what
EVP_PKEY_get_default_digest_name() responds when getting the digest
name via a EVP_PKEY_ASN1_METHOD ctrl call that returns NID_undef.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11576)
This function intialises an EVP_PKEY to contain a provider side internal
key.
We take the opportunity to also document the older EVP_PKEY_set_type()
and EVP_PKEY_set_type_str().
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/11375)
EVP_PKEY is rather complex, even before provider side keys entered the
stage.
You could have untyped / unassigned keys (pk->type == EVP_PKEY_NONE),
keys that had been assigned a type but no data (pk->pkey.ptr == NULL),
and fully assigned keys (pk->type != EVP_PKEY_NONE && pk->pkey.ptr != NULL).
For provider side keys, the corresponding states weren't well defined,
and the code didn't quite account for all the possibilities.
We also guard most of the legacy fields in EVP_PKEY with FIPS_MODE, so
they don't exist at all in the FIPS module.
Most of all, code needs to adapt to the case where an EVP_PKEY's
|keymgmt| is non-NULL, but its |keydata| is NULL.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/11375)
The following functions are added:
EVP_PKEY_gen_set_params(), replacing the older EVP_PKEY_CTX_ctrl()
EVP_PKEY_gen(), replacing both EVP_PKEY_keygen() and EVP_PKEY_paramgen()
These functions are made to work together with already existing domparams
and key generation functionality: EVP_PKEY_CTX_new_provided(),
EVP_PKEY_paramgen_init(), EVP_PKEY_keygen_init(), etc.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10289)
This adds evp_keymgmt_util_copy() and affects EVP_PKEY_copy_parameters()
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11158)
This adds evp_keymgmt_util_match() and affects EVP_PKEY_cmp() and
EVP_PKEY_cmp_parameters().
The word 'match' was used for the new routines because many associate
'cmp' with comparison functions that allows sorting, i.e. return -1, 0
or 1 depending on the order in which the two compared elements should
be sorted. EVP_PKEY_cmp() and EVP_PKEY_cmp_parameters() don't quite
do that.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11158)
The role of this cache was two-fold:
1. It was a cache of key copies exported to providers with which an
operation was initiated.
2. If the EVP_PKEY didn't have a legacy key, item 0 of the cache was
the corresponding provider side origin, while the rest was the
actual cache.
This dual role for item 0 made the code a bit confusing, so we now
make a separate keymgmt / keydata pair outside of that cache, which is
the provider side "origin" key.
A hard rule is that an EVP_PKEY cannot hold a legacy "origin" and a
provider side "origin" at the same time.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11148)
Previously, evp-keymgmt_util_export_to_provider() took care of all
kinds of exports of EVP_PKEYs to provider side keys, be it from its
legacy key or from another provider side key. This works most of the
times, but there may be cases where the caller wants to be a bit more
in control of what sort of export happens when.
Also, when it's time to remove all legacy stuff, that job will be much
easier if we have a better separation between legacy support and
support of provided stuff, as far as we can take it.
This changes moves the support of legacy key to provider side key
export from evp-keymgmt_util_export_to_provider() to
evp_pkey_make_provided(), and makes sure the latter is called from all
EVP_PKEY functions that handle legacy stuff.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11074)
The KEYMGMT libcrypto <-> provider interface currently makes a few
assumptions:
1. provider side domain parameters and key data isn't mutable. In
other words, as soon as a key has been created in any (loaded,
imported data, ...), it's set in stone.
2. provider side domain parameters can be strictly separated from the
key data.
This does work for the most part, but there are places where that's a
bit too rigid for the functionality that the EVP_PKEY API delivers.
Key data needs to be mutable to allow the flexibility that functions
like EVP_PKEY_copy_parameters promise, as well as to provide the
combinations of data that an EVP_PKEY is generally assumed to be able
to hold:
- domain parameters only
- public key only
- public key + private key
- domain parameters + public key
- domain parameters + public key + private key
To remedy all this, we:
1. let go of the distinction between domain parameters and key
material proper in the libcrypto <-> provider interface.
As a consequence, functions that still need it gain a selection
argument, which is a set of bits that indicate what parts of the
key object are to be considered in a specific call. This allows
a reduction of very similar functions into one.
2. Rework the libcrypto <-> provider interface so provider side key
objects are created and destructed with a separate function, and
get their data filled and extracted in through import and export.
(future work will see other key object constructors and other
functions to fill them with data)
Fixes#10979
squash! Redesign the KEYMGMT libcrypto <-> provider interface - the basics
Remedy 1 needs a rewrite:
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11006)
Some of the evp_keymgmt_ functions are just wrappers around the
EVP_KEYMGMT function pointers. We move those from keymgmt_lib.c to
keymgmt_meth.c.
Other evp_keymgmt_ functions are utility functions to help the rest of
the EVP functions. Since their names are easily confused with the
functions that were moved to keymgmt_meth.c, we rename them so they
all start with evp_keymgmt_util_.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11006)
Some functions went directly on keymgmt function pointers instead of
going through the internal KEYMGMT API, which makes for a confusing
read.
Related to #10962
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10963)
These functions relied entirely on the presence of 'pkey->pmeth',
which is NULL on provider only keys. This adds an interface to get
domparam and key data from a provider, given corresponding provider
data (the actual domparam or key).
The retrieved data is cached in the EVP_PKEY structure (lending the
idea from provided EVP_CIPHER).
Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/10778)
Exporting data from a provider owned domainparams or key is quite an
ordeal, with having to figure out what parameter keys an
implementation supports, call the export function a first time to find
out how large each parameter buffer must be, allocate the necessary
space for it, and call the export function again.
So how about letting the export function build up the key data params
and call back with that? This change implements exactly such a
mechanism.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10414)
This is the EVP operation that corresponds to creating direct RSA, DH
and DSA keys and set their numbers, to then assign them to an EVP_PKEY,
but done entirely using an algorithm agnostic EVP interface.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10187)
It may be that the OSSL_PARAM array we used for getting parameter
values for a key had a few too many entries. These are detected by
their return_size == 0. Before making second export call, we prune
away these items so we only ask for parameters that exist.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10190)
