Summary:
When ed25519 was introduced in commit 93c7b81b4ea1046bd2f65f4a510d5966786e8d3d,
the ed25519 case was added after the ecdsa case in src/pki.c. The ecdsa case seems to
have relied on falling through to report an error, when HAVE_ECC is not defined.
If HAVE_ECC is not defined, but ecdsa keys are used, with for example,
ssh_pki_import_pubkey_file, the code fallthroughs into the ed25519 case.
Signed-off-by: Juraj Vijtiuk <juraj.vijtiuk@sartura.hr>
Test Plan: Unit tests passed. No memory leaks found with valgrind.
Reviewers: asn
Differential Revision: https://bugs.libssh.org/D13
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Summary:
This patch adds support for mbedTLS as a crypto backend for libssh.
mbedTLS is an SSL/TLS library that has been designed to mainly be used
in embedded systems. It is loosely coupled and has a low memory
footprint. mbedTLS also provides a cryptography library (libmbedcrypto)
that can be used without the TLS modules.
The patch is unfortunately quite big, since several new files had to
be added.
DSA is disabled at compile time, since mbedTLS doesn't support DSA
Patch review and feedback would be appreciated, and if any issues or
suggestions appear, I'm willing to work on them.
Signed-off-by: Juraj Vijtiuk <juraj.vijtiuk@sartura.hr>
Test Plan:
* The patch has been tested with a Debug and MinSizeRel build, with
libssh unit tests, client tests and the pkd tests.
* All the tests have been run with valgrind's memcheck, drd and helgrind
tools.
* The examples/samplessh client works when built with the patch.
Reviewers: asn, aris
Subscribers: simonsj
Differential Revision: https://bugs.libssh.org/D1
Pair-Programmed-With: Jakub Jelen <jjelen@redhat.com>
Signed-off-by: Aris Adamantiadis <aris@0xbadc0de.be>
Signed-off-by: Jakub Jelen <jjelen@redhat.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Summary:
Fix a typo for old libgcrypt builds `k_length` -> `k_len`.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Test Plan: * Spotted this last-minute typo bug in local testing.
Reviewers: asn
Reviewed By: asn
Differential Revision: https://bugs.libssh.org/D10
Summary:
Fix a typo for old libgcrypt builds `k_length` -> `k_len`.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Test Plan: * Spotted this last-minute typo bug in local testing.
Reviewers: asn
Differential Revision: https://bugs.libssh.org/D10
Expose an API 'ssh_server_init_kex' which allows one to change the set of
key exchange, hostkey, ciphers, MACs, and compression algorithms currently
configured for the ssh_session at hand, after having started the
'ssh_handle_key_exchange' process.
One can use this API from the already-existing 'connect_status_function'
callback to dynamically modify the set of algorithms used after having
received the client banner, but before sending out the initial KEXINIT
message.
For example, one might want to prevent advertising the curve25519 key
exchange algorithm for older OpenSSH clients due to interop bugs.
Fixes T25
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Fix a memory leak in the path where parsing returns early due
to seeing a repeated opcode. A testcase is added which
demonstrates the leak and fix with valgrind.
Resolves CID 1374267.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Summary:
Based on Dirkjan's original patch series here:
* https://www.libssh.org/archive/libssh/2015-08/0000029.html
Here the changes are adapted for the current master
branch, and expanded to include libgcrypt support.
Co-Authored-By: Dirkjan Bussink <d.bussink@gmail.com>
Signed-off-by: Jon Simons <jon@jonsimons.org>
Test Plan:
* Ran pkd tests for libcrypto and libgcrypt builds.
* Ran client torture_algorithms.c tests for libcrypto and libgcrypt builds.
* Tested across multiple libgcrypts ("1.6.3" and "1.7.6-beta").
Reviewers: aris, asn
Reviewed By: asn
Tags: #libssh
Differential Revision: https://bugs.libssh.org/D7
Summary:
Based on Dirkjan's original patch series here:
* https://www.libssh.org/archive/libssh/2015-08/0000029.html
Here the changes are adapted for the current master
branch, and expanded to include libgcrypt support.
Co-Authored-By: Dirkjan Bussink <d.bussink@gmail.com>
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Test Plan:
* Ran pkd tests for libcrypto and libgcrypt builds.
* Ran client torture_algorithms.c tests for libcrypto and libgcrypt builds.
* Tested across multiple libgcrypts ("1.6.3" and "1.7.6-beta").
Reviewers: aris, asn
Tags: #libssh
Differential Revision: https://bugs.libssh.org/D7
That way, we will not fail later on key exchange phase when something
unknown is negotiated.
Fixes T37
Signed-off-by: Nikos Mavrogiannopoulos <nmav@redhat.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
On OpenSSL versions prior to 1.1.0, `EVP_CIPHER_CTX_cleanup` will
dereference its argument regardless of whether it is NULL. This
is not a problem on OpenSSL at or beyond 1.1.0, where
`EVP_CIPHER_CTX_cleanup` (macro to `EVP_CIPHER_CTX_reset`) returns
early upon NULL input.
Move the call to `EVP_CIPHER_CTX_cleanup` under the existing NULL
check in `evp_cipher_cleanup` to avoid the problem.
Introduced with this build-break fix:
* e66f370682927ca8bd7ae0e7544754c6f4ac4969
Found in manual testing in an environment with an older OpenSSL.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Relax the cases where `ssh_analyze_banner` fails to extract a
major and minor version from banners which appear like OpenSSH
banners.
Update the tests to demonstrate that now a banner as might be
sent by `ssh-keyscan(1)` ("SSH-2.0-OpenSSH-keyscan") no longer
returns failure.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
With this change, a HAVE_LIBCRYPTO #ifdef is removed from wrapper.c.
Now, the libcrypto-specific logic for EVP_CIPHER_CTX_free is moved
into the ssh_cipher_struct cleanup callback handler for those
ciphers.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Fix a gcrypt build error introduced with
48e7b098f86207f8596651b5ba8242a3b834a868.
The ssh_cipher_struct only contains a `ctx` field on
the libcrypto builds, so it can't be referenced unless
within HAVE_LIBCRYPTO.
This build fix preserves the original spirit of the
change in 48e7b098f86207f8596651b5ba8242a3b834a868:
only call `EVP_CIPHER_CTX_free` when `cipher->ctx`
is non-NULL.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Fix error-checking for `strtoul` in `ssh_analyze_banner`, and
enable some tests which demonstrate the fix before-and-after.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
No changes to code, only whitespace indentation and
an update to the function docs.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
Fix a resource leak in `hmac_final`: say `HMAC_CTX_free` instead
of `HMAC_CTX_reset`. This matches the error handling as done in
`hmac_init`. Introduced with cf1e808e2ffa1f26644fb5d2cb82a919f323deba.
The problem is reproducible running the `pkd_hello` test with:
valgrind --leak-check=full ./pkd_hello -i1 -t torture_pkd_openssh_dsa_rsa_default
Resolves https://red.libssh.org/issues/252.
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
On older OpenSSL versions, the EVP_MD_CTX fields within an HMAC_CTX
structure are contained inlined (change here [1]): be sure to not
try to free those fields on those builds.
Found running the `pkd_hello` test with:
valgrind ./pkd_hello -i1 -t torture_pkd_openssh_dsa_rsa_default
^ valgrind will cite "Invalid free() ..." errors which are present
before this fix and absent after, when building with OpenSSL 1.0.1.
[1] 6e59a892db
Signed-off-by: Jon Simons <jon@jonsimons.org>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
libssh fails to read the configuration from a config file due to a
wrong check in 'ssh_config_parse_line' procedure in 'config.c'; it's
effectively skipping every opcode (and therefore every option) from
the file. The change fixes that behaviour.
Signed-off-by: Artyom V. Poptsov <poptsov.artyom@gmail.com>
Reviewed-by: Andreas Schneider <asn@cryptomilk.org>
