Taking sizeof(ptr) is incorrect to determine size of passed in hash and
results in hlen getting set to a very large value since MD5_MAC_LEN >
sizeof(ptr). Provide the actual size of the hash buffer from the caller
to fix this.
tls_key_x_server_params_hash() callers src/tls/tlsv1_client_read.c and
src/tls/tlsv1_server_write.c both pass in a large enough hash (hash[64]
or hash[100]) that this does not appear to have an impact, though it is
still wrong.
Signed-off-by: Glenn Strauss <gstrauss@gluelogic.com>
The highest supported TLS version for pre_master_secret needs to be
limited based on the local configuration for the case where the highest
version number is being explicitly disabled. Without this, the server
would likely detect a downgrade attack.
Signed-off-by: Jouni Malinen <j@w1.fi>
The supported hash algorithms do not use AlgorithmIdentifier parameters.
However, there are implementations that include NULL parameters in
addition to ones that omit the parameters. Previous implementation did
not check the parameters value at all which supported both these cases,
but did not reject any other unexpected information.
Use strict validation of digest algorithm parameters and reject any
unexpected value when validating a signature. This is needed to prevent
potential forging attacks.
Signed-off-by: Jouni Malinen <j@w1.fi>
Digest is within the DigestInfo SEQUENCE and as such, parsing for it
should use the end of that data instead of the end of the decrypted
signature as the end point. Fix this in the PKCS #1 and X.509
implementations to avoid accepting invalid digest data that is
constructed to get the hash value from after the actual DigestInfo
container.
Signed-off-by: Jouni Malinen <j@w1.fi>
The definite long form for the length is allowed only for cases where
the definite short form cannot be used, i.e., if the length is 128 or
greater. This was not previously enforced and as such, multiple
different encoding options for the same length could have been accepted.
Perform more strict checks to reject invalid cases for the definite long
form for the length. This is needed for a compliant implementation and
this is especially important for the case of verifying DER encoded
signatures to prevent potential forging attacks.
Signed-off-by: Jouni Malinen <j@w1.fi>
The extended tag case is allowed only for tag values that are 31 or
larger (i.e., the ones that would not fit in the single octet identifier
case with five bits). Extended tag format was previously accepted even
for the values 0..31 and this would enable multiple different encodings
for the same tag value. That is not allowed for DER.
Perform more strict checks to reject invalid extended tag values. This
is needed for a compliant implementation and this is especially
important for the case of verifying DER encoded signatures to prevent
potential forging attacks.
Signed-off-by: Jouni Malinen <j@w1.fi>
The identifier octet in DER encoding includes three components. Only two
of these (Class and Tag) were checked in most cases when looking for a
specific data type. Also check the Primitive/Constructed bit to avoid
accepting invalid encoding.
This is needed for correct behavior in DER parsing and especially
important for the case of verifying DER encoded signatures to prevent
potential forging attacks.
Signed-off-by: Jouni Malinen <j@w1.fi>
We don't really need to duplicate more of this, so just
move the lib.rules include to the end and do more of the
stuff that's common anyway there.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Derive the library name from the directory name, and let each
library Makefile only declare the objects that are needed.
This reduces duplicate code for the ar call. While at it, also
pretty-print that call.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Now that we no longer leave build artifacts outside the build folder, we
can clean up the gitignore a bit. Also move more things to per-folder
files that we mostly had already anyway.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This is something I hadn't previously done, but there are
cases where it's needed, e.g., building 'wlantest' and then
one of the tests/fuzzing/*/ projects, they use a different
configuration (fuzzing vs. not fuzzing).
Perhaps more importantly, this gets rid of the last thing
that was dumped into the source directories, apart from
the binaries themselves.
Note that due to the use of thin archives, this required
building with absolute paths.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Instead of building in the source tree, put most object
files into the build/ folder at the root, and put each
thing that's being built into a separate folder.
This then allows us to build hostapd and wpa_supplicant
(or other combinations) without "make clean" inbetween.
For the tests keep the objects in place for now (and to
do that, add the build rule) so that we don't have to
rewrite all of that with $(call BUILDOBJS,...) which is
just noise there.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The recent addition of the X.509v3 certificatePolicies parser had a
copy-paste issue on the inner SEQUENCE parser that ended up using
incorrect length for the remaining buffer. Fix that to calculate the
remaining length properly to avoid reading beyond the end of the buffer
in case of corrupted input data.
Credit to OSS-Fuzz: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=20363
Fixes: d165b32f38 ("TLS: TOD-STRICT and TOD-TOFU certificate policies")
Signed-off-by: Jouni Malinen <j@w1.fi>
Allow any pointer to be used as source for encoding and use char * as
the return value from encoding and input value for decoding to reduce
number of type casts needed in the callers.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Some compilers have started to warn about this and the use of two loops
with ix 0..pa-1 and 0..pa loop a bit suspicious, so better make sure the
array is initialized with zeros before extracting the terms from it.
Signed-off-by: Jouni Malinen <j@w1.fi>
This does not need to be specific to X.509, so move the BOOLEAN DER
encoding validation into asn1_get_next() to make it apply for all cases
instead of having to have the caller handle this separately.
Signed-off-by: Jouni Malinen <j@w1.fi>
While BER encoding allows any nonzero value to be used for TRUE, DER is
explicitly allowing only the value 0xff. Enforce this constraint in
X.509 parsing to be more strict with what is acceptable.
Signed-off-by: Jouni Malinen <j@w1.fi>
Handling of the optional pathLenConstraint after cA was not done
properly. The position after cA needs to be compared to the end of the
SEQUENCE, not the end of the available buffer, to determine whether the
optional pathLenConstraint is present. In addition, when parsing
pathLenConstraint, the length of the remaining buffer was calculated
incorrectly by not subtracting the length of the header fields needed
for cA. This could result in reading couple of octets beyond the end of
the buffer before rejecting the ASN.1 data as invalid.
Credit to OSS-Fuzz: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=15408
Signed-off-by: Jouni Malinen <j@w1.fi>
sscanf() can apparently read beyond the end of the buffer even if the
maximum length of the integer is specified in the format string. Replace
this parsing mechanism with helper functions that use sscanf() with NUL
terminated string to avoid this.
Credit to OSS-Fuzz: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=15158
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
If none of the supported name attributes are present, the name string
was nul terminated only at the end. Add an explicit nul termination at
the end of the last written (or beginning of the buffer, if nothing is
written) to avoid writing uninitialized data to debug log.
Signed-off-by: Jouni Malinen <j@w1.fi>
Explicitly check the remaining buffer length before trying to read the
ASN.1 header values. Attempt to parse an ASN.1 header when there was not
enough buffer room for it would have started by reading one or two
octets beyond the end of the buffer before reporting invalid data at the
following explicit check for buffer room.
Signed-off-by: Jouni Malinen <j@w1.fi>
tlsv1_record_receive() did not return error here and as such, &alert was
not set and must not be used. Report internal error instead to avoid use
of uninitialized memory.
Signed-off-by: Jouni Malinen <j@w1.fi>
Add test-tls program that can be used for fuzzing the internal TLS
client and server implementations. This tool can write client or server
messages into a file as an initialization step and for the fuzzing step,
that file (with potential modifications) can be used to replace the
internally generated message contents.
The TEST_FUZZ=y build parameter is used to make a special build where a
hardcoded random number generator and hardcoded timestamp are used to
force deterministic behavior for the TLS operations.
Signed-off-by: Jouni Malinen <j@w1.fi>
