The RSA decryption as implemented before required very careful handling
of both the exit code returned by OpenSSL and the potentially returned
ciphertext. Looking at the recent security vulnerabilities
(CVE-2020-25659 and CVE-2020-25657) it is unlikely that most users of
OpenSSL do it correctly.
Given that correct code requires side channel secure programming in
application code, we can classify the existing RSA decryption methods
as CWE-676, which in turn likely causes CWE-208 and CWE-385 in
application code.
To prevent that, we can use a technique called "implicit rejection".
For that we generate a random message to be returned in case the
padding check fails. We generate the message based on static secret
data (the private exponent) and the provided ciphertext (so that the
attacker cannot determine that the returned value is randomly generated
instead of result of decryption and de-padding). We return it in case
any part of padding check fails.
The upshot of this approach is that then not only is the length of the
returned message useless as the Bleichenbacher oracle, so are the
actual bytes of the returned message. So application code doesn't have
to perform any operations on the returned message in side-channel free
way to remain secure against Bleichenbacher attacks.
Note: this patch implements a specific algorithm, shared with Mozilla
NSS, so that the attacker cannot use one library as an oracle against the
other in heterogeneous environments.
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/13817)
FIPS 186-4 section 5 "The RSA Digital Signature Algorithm", subsection
5.5 "PKCS #1" says: "For RSASSA-PSS […] the length (in bytes) of the
salt (sLen) shall satisfy 0 <= sLen <= hLen, where hLen is the length of
the hash function output block (in bytes)."
Introduce a new option RSA_PSS_SALTLEN_AUTO_DIGEST_MAX and make it the
default. The new value will behave like RSA_PSS_SALTLEN_AUTO, but will
not use more than the digest length when signing, so that FIPS 186-4 is
not violated. This value has two advantages when compared with
RSA_PSS_SALTLEN_DIGEST: (1) It will continue to do auto-detection when
verifying signatures for maximum compatibility, where
RSA_PSS_SALTLEN_DIGEST would fail for other digest sizes. (2) It will
work for combinations where the maximum salt length is smaller than the
digest size, which typically happens with large digest sizes (e.g.,
SHA-512) and small RSA keys.
J.-S. Coron shows in "Optimal Security Proofs for PSS and Other
Signature Schemes. Advances in Cryptology – Eurocrypt 2002, volume 2332
of Lecture Notes in Computer Science, pp. 272 – 287. Springer Verlag,
2002." that longer salts than the output size of modern hash functions
do not increase security: "For example,for an application in which at
most one billion signatures will be generated, k0 = 30 bits of random
salt are actually sufficient to guarantee the same level of security as
RSA, and taking a larger salt does not increase the security level."
Signed-off-by: Clemens Lang <cllang@redhat.com>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19724)
Do not accept password-based if expected signature-based and no secret is available and
do not accept signature-based if expected password-based and no trust anchors available.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/19729)
This PR is based off the contributions in PR #9223 by Jemmy1228.
It has been modified and reworked to:
(1) Work with providers
(2) Support ECDSA and DSA
(3) Add a KDF HMAC_DRBG implementation that shares code with the RAND HMAC_DRBG.
A nonce_type is passed around inside the Signing API's, in order to support any
future deterministic algorithms.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18809)
Signed-off-by: Xu Yizhou <xuyizhou1@huawei.com>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19619)
Add the following OID:
SM4-XTS: 1.2.156.10197.1.104.10
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19619)
This supports all the modes, suites and export mechanisms defined
in RFC9180 and should be relatively easily extensible if/as new
suites are added. The APIs are based on the pseudo-code from the
RFC, e.g. OSS_HPKE_encap() roughly maps to SetupBaseS(). External
APIs are defined in include/openssl/hpke.h and documented in
doc/man3/OSSL_HPKE_CTX_new.pod. Tests (test/hpke_test.c) include
verifying a number of the test vectors from the RFC as well as
round-tripping for all the modes and suites. We have demonstrated
interoperability with other HPKE implementations via a fork [1]
that implements TLS Encrypted ClientHello (ECH) which uses HPKE.
@slontis provided huge help in getting this done and this makes
extensive use of the KEM handling code from his PR#19068.
[1] https://github.com/sftcd/openssl/tree/ECH-draft-13c
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/17172)
This is needed for building with '-march=i386 no-threads', on platforms
where libatomic is not available (djgpp, specifically). The
implementation now falls back to 'CRYPTO_atomic_add()', which performs
plain lock-free addition in a 'no-threads' build.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19751)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/19216)
On this occasion, replace magic constants by mnemonic ones; update doc
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Todd Short <todd.short@me.com>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/19205)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/19230)
TLS device offload allows to perform zerocopy sendfile transmissions.
FreeBSD provides this feature by default, and Linux 5.19 introduced it
as an opt-in. Zerocopy improves the TX rate significantly, but has a
side effect: if the underlying file is changed while being transmitted,
and a TCP retransmission happens, the receiver may get a TLS record
containing both new and old data, which leads to an authentication
failure and termination of connection. This effect is the reason Linux
makes a copy on sendfile by default.
This commit adds support for TLS zerocopy sendfile on Linux disabled by
default to avoid any unlikely backward compatibility issues on Linux,
although sacrificing consistency in OpenSSL's behavior on Linux and
FreeBSD. A new option called KTLSTxZerocopySendfile is added to enable
the new zerocopy behavior on Linux. This option should be used when the
the application guarantees that the file is not modified during
transmission, or it doesn't care about breaking the connection.
The related documentation is also added in this commit. The unit test
added doesn't test the actual functionality (it would require specific
hardware and a non-local peer), but solely checks that it's possible to
set the new option flag.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Todd Short <todd.short@me.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18650)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19346)
And so clean a few useless includes
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19721)
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/19715)
Reduce the Miller Rabin counts to the values specified by FIPS 186-5.
The old code was using a fixed value of 64.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19579)
FIPS 186-4 has 5 different algorithms for key generation,
and all of them rely on testing GCD(a,n) == 1 many times.
Cachegrind was showing that during a RSA keygen operation,
the function BN_gcd() was taking a considerable percentage
of the total cycles.
The default provider uses multiprime keygen, which seemed to
be much faster. This is because it uses BN_mod_inverse()
instead.
For a 4096 bit key, the entropy of a key that was taking a
long time to generate was recorded and fed back into subsequent
runs. Roughly 40% of the cycle time was BN_gcd() with most of the
remainder in the prime testing. Changing to use the inverse
resulted in the cycle count being 96% in the prime testing.
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19578)
include/openssl/e_os2.h defines OSSL_SSIZE_MAX in terms of SIZE_MAX as a
fallback. This doesn't work well on platforms where SIZE_MAX isn't defined,
so we must ensure that it's defined by including "internal/numbers.h".
Since this is compensating for operating system discrepancies, it's
reasonable to make this change in include/internal/e_os.h.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19693)
This makes conversion to using list.h easier because the compiler will error
on an unknown field name rather than accepting `head` and `tail` and missing
some changes.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/19377)
The demux and record RX implemented lists internally. This changes them over
to using list.h.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/19377)
Instead of implementing a list internally.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/19377)
This is instead of re-implementing a linked list itself.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/19377)
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)
Added SFRAME_LIST structure and QUIC_RSTREAM object to
manage received stream data.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19351)
This causes a warning since tv_sec is unsigned.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19322)
Add test for `.' overflows, remove the output size argument from
ossl_a2ulabel() since it was never used and greatly complicated the code.
Convert ossl_a2ulabel() to use WPACKET for building the output string.
Update the documentation to match the new definition of ossl_a2ulabel().
x509: let punycode handle the '\0' string termination. Saves a memset(3)
and some size fiddling. Also update to deal with the modified parameters.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/19591)
These functions pass a library content and prop query.
The i2d documentation related to these functions has been corrected since the bio and fp functions always return 0 or 1.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/18427)
Co-authored-by: Randall Steck <rsteck@thinqsoft.com>
Co-authored-by: Mark J. Minnoch <mark@keypair.us>
Co-authored-by: Steve Weymann <steve@keypair.us>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/19510)
The FIPS 140-3 DSA and ECDSA tests need to be known answer tests which means
the entropy needs to be cooked. This permits this.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/19510)
The re-occuring surprise is that in Win32, size_t is 32 bits...
Fixed by changing size_t to uint64_t in QUIC_CC
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19345)
Signed-off-by: Čestmír Kalina <ckalina@redhat.com>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19473)
Multiple concurrent joins with a running thread suffer from a race
condition that allows concurrent join calls to perform concurrent arch
specific join calls, which is UB on POSIX, or to concurrently execute
join and terminate calls.
As soon as a thread T1 exists, one of the threads that joins with T1
is selected to perform the join, the remaining ones await completion.
Once completed, the remaining calls immediately return. If the join
failed, another thread is selected to attempt the join operation.
Forcefully terminating a thread that is in the process of joining
another thread is not supported.
Common code from thread_posix and thread_win was refactored to use
common wrapper that handles synchronization.
Signed-off-by: Čestmír Kalina <ckalina@redhat.com>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19433)
The recent DTLS write record layer code and the certificate compression
code both added new SSL_R_ reason codes. The numbers are conflicting due
to rebase issues and causing make update to fail.
Reviewed-by: Hugo Landau <hlandau@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19457)
The sequence counter was incremented in numerous different ways in
numerous different locations. We introduce a single function to do this
inside the record layer.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19424)
