This linking worked only in CI and local builds, but not during
the build in RPM as it fails on missing symbols that were defined
only in the main library. This is solved as with the other digest
dependencies in external crypto by removing the intermediate
function. We are already linking the md_*.o objects.
The error was like this
sh: symbol lookup error: /path/libssh/libssh-0.12.0-build/libssh-0.12.0/redhat-linux-build/lib/libsntrup761_override.so: undefined symbol: crypto_hash_sha512
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Coverity thought that the best_nlines could underflow, but the best_size is
initialized to 0 before calling this function so its moot. Adjusting the code
to be better understandable to static analyzers by initializing the variable
inside of the function.
Thanks coverity!
CID 1548873
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
... if underlying functions read or write more than expected.
This should never happen, but static analysis tools are inventive.
Thanks coverity!
CID 1548868
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
The `first` is intialized to -1 and if we reach this without setting this, we
needlessly call close(-1). It should be no-op, but better be safe.
Thanks coverity!
CID 1644001
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Checking the session->gssapi is resulting in the very same results, but this
approach is more direct and makes static analysis tools more happy.
Thanks coverity!
CID 1644000
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
... file, even if it was already set before. The options
level handles what was already set.
The proxyJump implementation sets the username from the proxyjump, which
is setting it to NULL, effectively writing the current username to the
new session, which was not possible to override due to the following check.
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
We want to make sure it suceeds because it could fail if
the client tries to send a hostbound public key authentication
request.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
If there is no host key (e.g., because we are doing
gssapi-keyex with "null" host key algorithm), it does not
make sense to use host bound authentication.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
All gssapi-keyex tests have to be disabled in Centos Stream 8
because the KEX is not allowed in FIPS. In Centos Stream 9,
only tests against OpenSSH have to be disabled because
OpenSSH only enables gssapi-keyex since Centos Stream 10.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
And setup a KDC server before pinging the server so we
can connect.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Do not try to verify mic if gssapi-keyex was not performed,
and fix a memory leak of the mic on error.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
These are not strictly necessary because we always check
that we performed GSSAPI KEX, but they won't hurt us.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Reject all GSSAPI-related messages when compiled
without GSSAPI support.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
In particular, gss-nistp256-sha256-* and
gss-curve25519-sha256-*.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
With packet filtering now implemented for type 65,
the current test packet would be rejected, resulting
in failed tests.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Both gcry_mpi_copy and gcry_mpi_set take a pointer to
const gcry_mpi, which const_bignum is not.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
This is necessary to reuse the function
in gss-curve25519-sha256-* KEX.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Make it use the one-shot API of hash functions,
and remove the FIPS restriction for OpenSSL 3.5+
where we can fetch the MD5 implementation from
a non-FIPS provider to use for non-crypto purposes.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Which can be used for non-cryptographic purposes
even in FIPS mode.
Signed-off-by: Pavol Žáčik <pzacik@redhat.com>
Reviewed-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
