Including e_os.h with a path from a header file doesn't work well on
certain exotic platform. It simply fails to build.
Since we don't seem to be able to stop ourselves, the better move is
to move e_os.h to an include directory that's part of the inclusion
path given to the compiler.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17641)
The private key for rsa, dsa, dh and ecx was being included when the
selector was just the public key. (ec was working correctly).
This matches the documented behaviour.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17200)
Also update and slightly extend the respective documentation and simplify some code.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/16251)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/16918)
EVP_PKEY_CTX_new_from_pkey() and EVP_CIPHER_CTX_new().
Otherwise may result in memory errors.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/16892)
It is possible to call built-in EVP_PKEY_METHOD functions with a provided
key. For example this might occur if a custom EVP_PKEY_METHOD is in use
that wraps a built-in EVP_PKEY_METHOD. Therefore our EVP_PKEY_METHOD
functions should not assume that we are using a legacy key. Instead we
get the low level key using EVP_PKEY_get0_RSA() or other similar functions.
This "does the right thing" if the key is actually provided.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/16118)
That check was seen as necessary at the time, but other changes have
been made since, so we now have better control on when we're handling
legacy structures and methods, making it safe to run the export_to
function on keys with foreign methods.
The basic message is that foreign methods must set key structure
values according to our standards no matter what, or not set them at
all. This has really always been the case, but was harder to see at
the time because of interaction with other bugs.
Fixes#15927
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15996)
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)
This nicely reduces the number of files considered as fips
provider sources.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15609)
To be reverted once key generation checks are added everywhere and a way to
disable them implemented.
Fixes#15502
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15560)
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)
During key generation, any sized key can be asked for. Attempting to generate
a key with a security strength larger than the RNG strength now fails.
Fixes#15421
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15472)
Also cap the strengths for values under these two because the formula
overestimates around them.
The formula NIST gives doesn't match the table also presented in IG 7.5
Partial fix for #15421
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/15428)
A pairwise test runs only in FIPS mode.
An assumption about the size of the 'to' buffer passed to
RSA_private_decrypt() was incorrect. It needs to be up to RSA_size()
bytes long - so a fixed buffer of 256 bytes was not large enough.
An exiting malloc has increased in size to allocate buffer space for
both the encrypt and decrypt buffer.
The existing test used 2080 bits which was not quite large enough to
trigger the issue. A test using 3072 bits has been added.
Reported by Mark Powers from Acumen.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15447)
The new names are ossl_err_load_xxx_strings.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15446)
We previously took an EVP_KEYMGMT pointer, but now found it necessary
to use a different import function in some cases. Since that's the
only thing we use from EVP_KEYMGMT, we might as well pass the import
function directly, allowing for some flexibility in how export_to is
used.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15293)
Almost all were notes about wanting to deprecate CTRLs/utility functions.
Fixes#15325
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/15328)
The low-level engine and app method based keys have to be treated
as foreign and must be used with old legacy pmeths.
Fixes#14632
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14859)
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)
In 1.1.1 the ctrl EVP_PKEY_CTRL_RSA_KEYGEN_BITS would fail immediately
if the number of bits was too small. In 3.0 it always succeeds, and only
fails later during the key generation stage.
We fix that so that it fails early like it used to in 1.1.1.
Note that in 1.1.1 it fails with a -2 return code. That is not the case
in 3.0 and has not been addressed here (see #14442)
Fixes#14443
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14702)
The legacy implementation would print the ASN.1 value of the trailerfield,
except when it wasn't set (i.e. is default).
For better consistency, we now always print the ASN.1 value, both in the
legacy and the provided implementation.
Fixes#14363
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14676)
with AVX512_IFMA + AVX512_VL instructions, primarily for RSA CRT private key
operations. It uses 256-bit registers to avoid CPU frequency scaling issues.
The performance speedup for RSA2k signature on ICL is ~2x.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13750)
This required refactoring a number of functions from the diverse
EVP_PKEY_ASN1_METHOD implementations to become shared backend
functions. It also meant modifying a few of them to return pointers
to our internal RSA / DSA/ DH / EC_KEY, ... structures instead of
manipulating an EVP_PKEY pointer directly, letting the caller do the
latter.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14314)
CLA: trivial
On line 136, a period is added. I think this is what was intended.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14540)
Fixes#14390
The only caller of this function tests EVP_KEYMGMT_is_a() beforehand
which will fail if the RSA key types do not match. So the test is not
necessary. The assert has been removed when it does the test.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14524)
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)
Fixes#13995
For small keys the MR test on the modulus can return
BN_PRIMETEST_COMPOSITE_WITH_FACTOR status although the modulus
is correct.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14389)
This also fixes the public function RSA_padding_check_SSLv23.
Commit 6555a89 changed the padding check logic in RSA_padding_check_SSLv23
so that padding is rejected if the nul delimiter byte is not immediately
preceded by at least 8 bytes containing 0x03. Prior to that commit the
padding is rejected if it *is* preceded by at least 8 bytes containing 0x03.
Presumably this change was made to be consistent with what it says in
appendix E.3 of RFC 5246. Unfortunately that RFC is in error, and the
original behaviour was correct. This is fixed in later errata issued for
that RFC.
This has no impact on libssl for modern versions of OpenSSL because
there is no protocol support for SSLv2 in these versions. However
applications that call RSA_paddin_check_SSLv23 directly, or use the
RSA_SSLV23_PADDING mode may still be impacted. The effect of the original
error is that an RSA message encrypted by an SSLv2 only client will fail to
be decrypted properly by a TLS capable server, or a message encrypted by a
TLS capable client will fail to decrypt on an SSLv2 only server. Most
significantly an RSA message encrypted by a TLS capable client will be
successfully decrypted by a TLS capable server. This last case should fail
due to a rollback being detected.
Thanks to D. Katz and Joel Luellwitz (both from Trustwave) for reporting
this issue.
CVE-2021-23839
Reviewed-by: Paul Dale <pauli@openssl.org>
Fixes CID #1472393
Previously this switch handled CMS & PCKS7 controls (e.g ANS1_PKEY_CTRL_PKCS7_SIGN)
which fell thru to the dead code to set the X509_ALG.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14163)
Using ERR_LIB_* causes the error output to say 'reason(n)' instead of
the name of the sub-library in question.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14152)
Fixes#13969
- properly handle the mandatory RSA-PSS key parameters
- improve parameter checking when setting the parameters
- compute the algorithm id at the time it is requested so it
reflects the actual parameters set
- when generating keys do not override previously set parameters
with defaults
- tests added to the test_req recipe that should cover the PSS signature
handling
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/13988)
This commit attempts to fix the an issue when generating a key of type
EVP_PKEY_RSA_PSS. Currently, EVP_PKEY_CTX_set_rsa_keygen_bits will
return -1 if the key id is not of type EVP_PKEY_RSA. This commit adds
EVP_PKEY_RSA_PSS to also be accepted.
The macro EVP_PKEY_CTX_set_rsa_pss_keygen_md si converted into a
function and it is now called in legacy_ctrl_to_param.
Fixes#12384
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13099)
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)
This includes error reporting for libcrypto sub-libraries in surprising
places.
This was done using util/err-to-raise
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/13318)
Many of the new types introduced by OpenSSL 3.0 have an OSSL_ prefix,
e.g., OSSL_CALLBACK, OSSL_PARAM, OSSL_ALGORITHM, OSSL_SERIALIZER.
The OPENSSL_CTX type stands out a little by using a different prefix.
For consistency reasons, this type is renamed to OSSL_LIB_CTX.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12621)
Low level algorithm implementations have no business knowing about details
of the higher level CMS concept. This knowledge is therefore moved into the
CMS layer.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/13088)
There is a large amount of CMS sepcific code in the algorithms. This is in
the wrong place and breaks layering. This code should be in the CMS layer.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/13088)
There is quite a large amount of algorithm specific CMS code sitting in
the algorithm directories. However, this seems to break layering.
Algorithms really have no business knowing anything about CMS. Really it
should be the other way around. Where there is algorithm specific CMS code
it is the CMS layer that should know how to handle different algorithms.
Therefore we move this code into the CMS layer.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/13088)
Automatically rename all instances of _with_libctx() to _ex() as per
our coding style.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12970)
Fixes#12635
As discussed in the issue, supporting the set0-like semantics long-term is not necessarily desirable, although necessary for short-term compatibility concerns. So I've deprecated the original method and added an equivalent that is explicitly labelled as set1.
I tried to audit existing usages of the (now-deprecated) API and update them to use set1 if that appeared to align with their expectations.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12917)
rsa_pss_params_30_fromdata() now uses the OSSL_PKEY_PARAM_RSA_DIGEST_PROPS parameter also.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12944)
OCSP_basic_sign_ctx() in ocsp_srv.c , does not check for RSA_METHOD_FLAG_NO_CHECK.
If a key has RSA_METHOD_FLAG_NO_CHECK set, OCSP sign operations can fail
because the X509_check_private_key() can fail.
The check for the RSA_METHOD_FLAG_NO_CHECK was moved to crypto/rsa/rsa_ameth.c
as a common place to check. Checks in ssl_rsa.c were removed.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12419)
(cherry picked from commit 56e8fe0b4e)
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12755)
The RSA key could be a public key, and yet, rsa_todata() always tries
to add the private parts as well. The resulting parameters will look
a bit odd, such as a zero |d|, resulting in an invalid key.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12676)
We have a key in test/recipes/30-test_evp_data/evppkey.txt with bad
PSS parameters (RSA-PSS-BAD), which is supposed to trigger signature
computation faults. However, if this key needs to be exported to the
RSA provider implementation, the result would be an earlier error,
giving the computation that's supposed to be checked n chance to even
be reached.
Either way, the legacy to provider export is no place to validate the
values of the key.
We also ensure that the provider implementation can handle and detect
signed (negative) saltlen values.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12583)
This is needed so RSA keys created from different code paths have a
chance to compare as equal.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12544)
It should never hit this branch of code, so there is no feasible test.
Found due to a similar issue in PR #12176.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12177)
<https://github.com/openssl/openssl/pull/11765> switched the default
code path for keygen.
External testing through TriggerFlow highlighted that in several places
we failed (once more!) to set the `BN_FLG_CONSTTIME` flag on critical
secret values (either long term or temporary values).
This commit tries to make sure that the secret BN values inside the
`rsa struct` are always flagged on creation, and that temporary values
derived from these secrets are flagged when allocated from a BN_CTX.
Acknowledgments
---------------
Thanks to @Voker57, @bbbrumley, @sohhas, @cpereida for the
[OpenSSL Triggerflow CI] ([paper]) through which this defect was detected and
tested, and for providing early feedback to fix the issue!
[OpenSSL Triggerflow CI]: https://gitlab.com/nisec/openssl-triggerflow-ci
[paper]: https://eprint.iacr.org/2019/366
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12167)
For FIPS validation purposes - Automated Cryptographic Validation Protocol (ACVP) tests need to be
performed. (See https://github.com/usnistgov/ACVP). These tests are very similiar to the old CAVS tests.
This PR uses a hardwired subset of these test vectors to perform similiar operations,
to show the usage and prove that the API's are able to perform the required operations.
It may also help with communication with the lab (i.e- The lab could add a test here to show
a unworking use case - which we can then address).
The EVP layer performs these tests instead of calling lower level API's
as was done in the old FOM.
Some of these tests require access to internals that are not normally allowed/required.
The config option 'acvp_tests' (enabled by default) has been added so that this
access may be removed.
The mechanism has been implemented as additional OSSL_PARAM values that can be set and get.
A callback mechanism did not seem to add any additional benefit.
These params will not be added to the gettables lists.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/11572)
Fixes#11742Fixes#11764
The newer RSA sp800-56b algorithm is being used for the normal case of a non multiprime key of at least length 2048.
Insecure key lengths and mutltiprime RSA will use the old method.
Bad public exponents are no longer allowed (i.e values less than 65537 or even). Values such as 2 that would cause a infinite loop now result in an error. The value of 3 has been marked as deprecated but is still allowed for legacy purposes.
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/11765)
As of the previous commit, when a zero-length (string) parameter
is present in the parameters passed to a provider for a given operation,
we will produce an object corresponding to that zero-length parameter,
indicating to the underlying cryptographic operation that the parameter
was passed. However, rsa_cms_decrypt() was relying on the previous
behavior, and unconditionally tried to call
EVP_PKEY_CTX_set0_rsa_oaep_label() even when the implicit default label
was used (and thus the relevant local variable was still NULL).
In the new setup that distinguishes present-but-empty and absent
more clearly, it is an error to attempt to set a NULL parameter,
even if it is zero-length.
Exercise more caution when setting parameters, and do not call
EVP_PKEY_CTX_set0_rsa_oaep_label() when there is not actually a
label provided.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/11920)
