sslcorrupttest was looking for a "decryption failed or bad record mac"
error in the queue. However if there were multiple errors on the queue
then it would fail to find it. We modify the test to check all errors.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
EVP_CipherUpdate is supposed to return 1 for success or 0 for error.
However for GCM ciphers it was sometimes returning -1 for error.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
We were not correctly passing the provider ctx down the chain during
initialisation of a new cipher ctx. Instead the provider ctx got set to
NULL.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
Other ciphers return the length of the Payload for TLS as a result of an
EVP_DecryptUpdate() operation - but ChaCha20-Poly1305 did not. We change
it so that it does.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
The previous commits separated out the TLS CBC padding code in libssl.
Now we can use that code to directly support TLS CBC padding and MAC
removal in provided ciphers.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
We split these functions out into a separate file because we are
preparing to make this file shared between libssl and providers.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
For CBC ciphersuites using Mac-then-encrypt we have to be careful about
removing the MAC from the record in constant time. Currently that happens
immediately before MAC verification. Instead we move this responsibility
to the various protocol "enc" functions so that MAC removal is handled at
the same time as padding removal.
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12288)
We now check that if libcrypto hasn't loaded the string for some particular
system error, it gives us "reason(nnn)" instead, where 'nnn' is the system
error number in decimal.
We go through all possible error macros that perl serves us, not only the
POSIX ones.
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12343)
test/errtest.c used the system error code 1 for EPERM. However, EPERM
may be coded differently on different systems, so we switch to using
EPERM instead. However, because we know that the ERR sub-system
truncates system error codes that occupy more than 24 bits, we check
that the reason code in the recorded error matches our EPERM, and skip
the test if not.
To be safe (even though the error string for that code is well defined
in POSIX), we also use strerror() to retrieve the string for that
error code instead of using a hard coded value.
Fixes#12276Fixes#12217Fixes#12354
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12343)
Because system errors can be any positive number that fits in an 'int'
according to POSIX, we can't reasonably expect them to be in the 1..127
range, even though that's the most usual.
Instead of packing them into the OpenSSL error code structure, we
recognise them as a special case and mark them as such by storing them
in our error queue with the highest bit set. We make OpenSSL specific
error records have their highest bit cleared, and in doing so, we
shift down the library section of the code by one bit. This still
leaves a very large section for the reason codes.
Of course, we must adapt the error code and reason string extraction
and printing functions accordingly.
With this, we also thrown away the pre-loaded array of system error
strings, and extract them from the system when needed instead, i.e.
when we create error strings.
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12343)
Some ERR_R_ codes overlapped other ERR_R_ codes:
- ERR_R_BUF_LIB vs ERR_R_PASSED_INVALID_ARGUMENT
- ERR_R_DSA_LIB vs ERR_R_INTERRUPTED_OR_CANCELLED
Looking back at history, this was originally not an issue, because
the ERR_R_ codes that weren't ERR_LIB_ aliases had bit 2**6 set.
However, new codes without that bit came in, and we got the overlap
that is mentioned above.
To get rid of the overlap, we repartition the codes as follows:
- ERR_R_{name}_LIB that are aliases for ERR_LIB_{name} are confined to
the range 1..63.
- Other ERR_R_ codes are confined to 64..99
We also expand the reason codes to 24 bits of data, where the 4 top
bits are for reason code flags. We also allocate a "fatal" flag
ERR_RFLAG_FATAL. The reason code ERR_R_FATAL stops acting as a flag,
but is coded in such a way that it still serves as one for code that
happens to use it as such.
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12343)
When ossl_provider_query_operation() returned NULL, the post-condition
callback wasn't called, and could make algorithm_do_this() falsely
tell the caller that there was an error. Because of this, a provider
that answered with NULL for a particular operation identity would
effectively block the same query on all following providers.
Fixes#12293
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12365)
exclude_rule 'MD023' # Headers must start at the beginning of the line
exclude_rule 'MD026' # Trailing punctuation in header
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12109)
Instead appease coverity by marking 1464986 as a false positive.
Coverity is confused by the engine reference counting.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12309)
The strength and max_length DRBG parameters were being cached in the EVP_RAND
layer. This commit removes the caching.
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12321)
The calls to query the DRBG strength, state and maximum output size all used
nested locks. This removes the nesting.
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12321)
CLA: trivial
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/12320)
The shared_target attrribute for Solaris built with gcc wasn't right
and shared libraries couldn't be properly built.
Fixes#12356
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12360)
OpenSSL::config::guess_system() is supposed to return system triplets.
However, for Windows and VMS, it returned the final OpenSSL config
target instead. We move the entries for them to the table that
OpenSSL::config::map_guess() uses, so it can properly convert the
input triplet to an OpenSSL config target.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12339)
Thanks to Michael Mueller on the openssl-users list for the suggested
improvement.
Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/12338)
[extended tests]
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12275)
If an empty password was supplied we ignored it and were trying to use
the fallback method to read the password instead (i.e. read from stdin).
However if that failed (which it always does if the cmp option -batch is
used) then we were reporting that we had successfully read the password
without actually setting one.
Instead, if an empty password is explicitly provided we should use it. If
no password is supplied explicitly and we have no fallback method then we
assume the empty password.
[extended tests]
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12275)
In HTTP_new_bio(), if the host has a trailing '/' we took a copy of the
hostname but failed to terminate it properly.
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/12275)
Make the improved semantics of VFO replace the previous VF and remove VFO
Add warnings about overriding use of HARNESS_VERBOSE* variables
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12279)
This line wasn't properly handled:
SCRIPTS{misc,linkname=tsget}=tsget.pl
It generated an attribute "linkname=tsget" with the value 1, instead of
what it should have, an attribute "linkname" with the value "tsget".
Fixes#12341
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/12344)
'absolutedir' does a thorough job ensuring that we have a "real" path
to both source and build directory, unencumbered by symbolic links.
However, that isn't enough on case insensitive file systems on Unix
flavored platforms, where it's possible to stand in, for example,
/PATH/TO/Work/openssl, and then do this:
perl ../../work/openssl/Configure
... and thereby having it look like the source directory and the build
directory aren't the same.
We solve this by having a closer look at the computed source and build
directories, and making sure they are exactly the same strings if they
are in fact the same directory.
This is especially important when making symbolic links based on this
directories, but may have other ramifications as well.
Fixes#12323
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/12337)
We can run tests in parallel by setting the HARNESS_JOBS environment
variable.
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12326)
The environment variable `HARNESS_JOBS` can be used to control how many
jobs to run in parallel. The default is still to run jobs sequentially.
This commit does not define custom `rules`, and different versions of
`TAP::Harness` come with different strategies regarding the default
`rules` that define which test recipes can be run in parallel.
In recent versions of Perl, unless specified otherwise any task can be
run in parallel.
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
(Merged from https://github.com/openssl/openssl/pull/12326)
Usually it will be freed in tls_early_post_process_client_hello().
However if a ClientHello callback will be used and will return
SSL_CLIENT_HELLO_RETRY then tls_early_post_process_client_hello()
may never come to the point where pre_proc_exts is freed.
Fixes#12194
CLA: trivial
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/12330)
