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)
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)
SSL_clear() explicitly clears an SSL object to enable it to be reused.
You can have a similar effect by calling SSL_set_accept_state() or
SSL_set_connect_state(). We extend the testing of SSL_clear() to use these
other methods. We also ensure we test the case where we have unread
bufferred data that needs to be cleared.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23256)
Coverity issue 1453632 noted a missing null check in kdf_test_ctrl
recently. If a malformed value is passed in from the test file that
does not contain a ':' character, the p variable will be NULL, leading
to a NULL derefence prepare_from_text
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
(Merged from https://github.com/openssl/openssl/pull/23398)
The failure would be caught later on, so this went unnoticed, until someone
tried with just one hex digit, which was simply ignored.
Fixes#23373
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23374)
PKCS7 ContentInfo fields held within a PKCS12 file can be NULL, even if the
type has been set to a valid value. CVE-2024-0727 is a result of OpenSSL
attempting to dereference the NULL pointer as a result of this.
We add test for various instances of this problem.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23361)
make sure that we get the expected error codes when we do bad things,
rather than a crash
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22995)
The EVP_CIPHER api currently assumes that calls made into several APIs
have already initalized the cipher in a given context via a call to
EVP_CipherInit[_ex[2]]. If that hasnt been done, instead of an error,
the result is typically a SIGSEGV.
Correct that by adding missing NULL checks in the apropriate apis prior
to using ctx->cipher
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22995)
array"key" is uninitialized and it is being read directly in function SipHash_Init() as per the below statements making a way for the garbage values :
uint64_t k0 = U8TO64_LE(k);
uint64_t k1 = U8TO64_LE(k + 8);
CLA: trivial
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23298)
this needs a sanitized 64 bit time_t build to be detected (or possibly
valgrind, trapv or similar)
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22976)
Check that we can write and read back long app data records when using
KTLS.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Neil Horman <nhorman@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23182)
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/19948)
X509_STORE_get0_objects returns a pointer to the X509_STORE's storage,
but this function is a bit deceptive. It is practically unusable in a
multi-threaded program. See, for example, RUSTSEC-2023-0072, a security
vulnerability caused by this OpenSSL API.
One might think that, if no other threads are mutating the X509_STORE,
it is safe to read the resulting list. However, the documention does not
mention that other logically-const operations on the X509_STORE, notably
certifcate verifications when a hash_dir is installed, will, under a
lock, write to the X509_STORE. The X509_STORE also internally re-sorts
the list on the first query.
If the caller knows to call X509_STORE_lock and X509_STORE_unlock, it
can work around this. But this is not obvious, and the documentation
does not discuss how X509_STORE_lock is very rarely safe to use. E.g.
one cannot call any APIs like X509_STORE_add_cert or
X509_STORE_CTX_get1_issuer while holding the lock because those
functions internally expect to take the lock. (X509_STORE_lock is
another such API which is not safe to export as public API.)
Rather than leave all this to the caller to figure out, the API should
have returned a shallow copy of the list, refcounting the values. Then
it could be internally locked and the caller can freely inspect the
result without synchronization with the X509_STORE.
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23224)
Fixes CVE-2023-6237
If a large and incorrect RSA public key is checked with
EVP_PKEY_public_check() the computation could take very long time
due to no limit being applied to the RSA public key size and
unnecessarily high number of Miller-Rabin algorithm rounds
used for non-primality check of the modulus.
Now the keys larger than 16384 bits (OPENSSL_RSA_MAX_MODULUS_BITS)
will fail the check with RSA_R_MODULUS_TOO_LARGE error reason.
Also the number of Miller-Rabin rounds was set to 5.
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23243)
This changeset adds the counterpart to the '-subj' option to allow overriding
the Issuer. For consistency, the `-subj` option is aliased to `-set_subject`.
The issuer can be specified as following apps/openssl x509 -new -set_issuer
'/CN=example-nro-ta' -subj '/CN=2a7dd1d787d793e4c8af56e197d4eed92af6ba13' ...
This is useful in constructing specific test-cases or rechaining PKI trees
Joint work with George Michaelson (@geeohgeegeeoh)
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23257)
When parsing the stable section of a config such as this:
openssl_conf = openssl_init
[openssl_init]
stbl_section = mstbl
[mstbl]
id-tc26 = min
Can lead to a SIGSEGV, as the parsing code doesnt recognize min as a
proper section name without a trailing colon to associate it with a
value. As a result the stack of configuration values has an entry with
a null value in it, which leads to the SIGSEGV in do_tcreate when we
attempt to pass NULL to strtoul.
Fix it by skipping any entry in the config name/value list that has a
null value, prior to passing it to stroul
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22988)
Add test case for re-using a cipher context with the same key, iv and
cipher. It detects, if the hardware-specific cipher context is reset
correctly, like reported in issue #23175.
This test has encrypt and decrypt iterations for cfb128 and
ofb128. All iteations use the same key, iv and plaintext.
Signed-off-by: Holger Dengler <dengler@linux.ibm.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23201)
Even in the good case there was memory leak here.
Add a simple test case to have at least some test coverage.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23234)
The base type OSSL_PARAM getters will NULL deref if they are initalized
as null. Add NULL checks for those parameters that have no expectation
of returning null (int32/64/uint32/64/BN). Other types can be left as
allowing NULL, as a NULL setting may be meaningful (string, utf8str,
octet string, etc).
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23083)
It would be helpful to be able to generate RSA's dmp1/dmq1/iqmp values
when not provided in the param list to EVP_PKEY_fromdata. Augment the
provider in ossl_rsa_fromdata to preform this generation iff:
a) At least p q n e and e are provided
b) the new parameter OSSL_PARAM_RSA_DERIVE_PQ is set to 1
Fixes#21826
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21875)
There are several points during x509 extension creation which rely on
configuration options which may have been incorrectly parsed due to
invalid settings. Preform a value check for null in those locations to
avoid various crashes/undefined behaviors
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23183)
The ssl_old_test has not been fully converted to the test framework but
it still reuses some test framework utilities. Notably it was creating
it's own copy of the global bio_err object directly (which is normally
created and owned by the test framework). This causes a problem because
ever since commit 2fa9044 access to the bio_err object is controlled by
a lock. Since ssl_old_test was circumventing the normal creation and
destruction of bio_err, the lock was not being created resulting in a
crash under certain error conditions.
We fix this by creating and destroying the bio_err object using the
test framework functions designed for that purpose.
Fixes#23184
Reviewed-by: Neil Horman <nhorman@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23187)
When using pbkdf1 key deriviation, it is possible to request a key
length larger than the maximum digest size a given digest can produce,
leading to a read of random stack memory.
fix it by returning an error if the requested key size n is larger than
the EVP_MD_size of the digest
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23174)
If a name is passed to EVP_<OBJ>_fetch of the form:
name1:name2:name3
The names are parsed on the separator ':' and added to the store, but
during the lookup in inner_evp_generic_fetch, the subsequent search of
the store uses the full name1:name2:name3 string, which fails lookup,
and causes subsequent assertion failures in evp_method_id.
instead catch the failure in inner_evp_generic_fetch and return an error
code if the name_id against a colon separated list of names fails. This
provides a graceful error return path without asserts, and leaves room
for a future feature in which such formatted names can be parsed and
searched for iteratively
Add a simple test to verify that providing a colon separated name
results in an error indicating an invalid lookup.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Todd Short <todd.short@me.com>
(Merged from https://github.com/openssl/openssl/pull/23110)
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/20727)
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/20727)
If a malformed config file is provided such as the following:
openssl_conf = openssl_init
[openssl_init]
providers = provider_sect
[provider_sect]
= provider_sect
The config parsing library will crash overflowing the stack, as it
recursively parses the same provider_sect ad nauseum.
Prevent this by maintaing a list of visited nodes as we recurse through
referenced sections, and erroring out in the event we visit any given
section node more than once.
Note, adding the test for this revealed that our diagnostic code
inadvertently pops recorded errors off the error stack because
provider_conf_load returns success even in the event that a
configuration parse failed. The call path to provider_conf_load has been
updated in this commit to address that shortcoming, allowing recorded
errors to be visibile to calling applications.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22898)
Currently, a provider is activated from our config file using the
activate parameter. However, the presence of the config parameter is
sufficient to trigger activation, leading to a counterintuitive
situation in which setting "activate = 0" still activates the provider
Make activation more intuitive by requiring that activate be set to one
of yes|true|1 to trigger activation. Any other value, as well as
omitting the parameter entirely, prevents activation (and also maintains
backward compatibility.
It seems a bit heavyweight to create a test specifically to validate the
plurality of these settings. Instead, modify the exiting openssl config
files in the test directory to use variants of these settings, and
augment the default.cnf file to include a provider section that is
explicitly disabled
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22906)
When verbosity isn't set to 1 or higher, suppress indirect leaks (i.e.
only print direct leaks) to make output more human-readable. Setting
V=1 on make test produces all leaks (direct and indirect)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/22678)
