Removing object reuse makes it dramatically simpler. Along the way, lift
the OID validity checker into crypto/bytestring, so we can use it more
generally. (Although the difference between invalid OID and unknown OID
is pretty academic, so this check isn't that important.)
For now I've preserved the existing behavior, where the OID validity
checker accepts arbitrarily large OID components. Though this results in
an oddity where the OID to string functions reject inputs that the
parser accepts. (There we only allow up to 2^64-1.)
Update-Note: When we removed object-reuse from all the d2i functions, we
missed one d2i_ASN1_OBJECT. See
https://boringssl-review.googlesource.com/c/boringssl/+/56647.
Otherwise, this CL is not expected to change behavior.
Change-Id: If4d2d83d9f3c96abfdc268e156f2cf3a9a903b0c
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58147
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
These functions already don't go through tasn_*.c. Rewrite them to use
CBS and CBB. This removes some dependencies on ASN1_get_object and
ASN1_put_object.
Update-Note: d2i_ASN1_OBJECT and d2i_ASN1_BOOLEAN will no longer accept
non-minimal length prefixes (forbidden in DER). d2i_ASN1_BOOLEAN will
also no longer accept non-canonical representations of TRUE (also
forbidden in DER). This does not affect certificate parsing, as that
still goes through the old template system, though we will make a
similar change to those functions later.
Bug: 354, 548
Change-Id: I0b7aa96f47aca5c31ec4f702e27108b4106311f2
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58145
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
ASN1_TYPE is a union of a bunch of pointer types. Often it takes a
shorthand and accesses a->value.ptr directly. This is allowed in C, but
not C++. Writing the switch/case barely takes more code, so just write
it that way.
Along the way, extract the code for cleaning up an ASN1_TYPE from
tasn_fre.c. This is a small step towards being able to use crypto/asn1's
types without depending on the templates. ASN1_TYPE_free still, for now,
calls into the templates. That will be fixable once tasn_*.c are
rewritten a bit further.
This also removes the weird hack here ASN1_primitive_free (an internal
function) with NULL ASN1_ITEM secretly meant to partially clean up the
ASN1_TYPE. We can just call the function directly now.
Bug: 574
Change-Id: Ie2ba41418801a366ab2f12eccc01e8dadf82c33e
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58126
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
An in-progress rewrite of tasn_dec.c accidentally broke this, so add a
regression test.
Bug: 548
Change-Id: Iac6a23acbc08459187c96a2f6471f0aa97d445a1
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58125
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
Change-Id: I59bcacf10a59ffdf9709785727f5f8b73c992f6e
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58026
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
RDNs are a SET OF attributes which means they should be sorted by
DER encoding length, then lexicographically. We didn't have any test
coverage for this.
Bug: 548
Change-Id: I542196aae26984aeee4f1c6774878b121675b0dc
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58025
Commit-Queue: Bob Beck <bbe@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Guarding the OPENSSL_memset was unnecessary since OPENSSL_memset with
zero length works fine. Also OPENSSL_memset, to workaround a C bug,
internally does the same check anyway. (But also this wasn't a context
where the C bug applied.)
Also don't bother calling EVP_MD_block_size a second time when we
already saved it into block_size.
Finally, don't bother filling key_block up to the whole
EVP_MAX_MD_BLOCK_SIZE size. I could believe the fixed size is easier for
the compiler to optimize, but the XORs in setting up an HMAC cannot
possibly be performance-sensitive, and using the actual length is
clearer.
Also add an assert that the hash's output size fits in the block size.
We're implicitly relying on this when hashing the key down.
Change-Id: I6ce37d41ea5bdbc8890bde7910e1b5651bc35709
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/58027
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
With EVP_PKEY and EVP_PKEY_CTX opaque, these symbols don't appear in any
public APIs anymore. Make them internal, which also opens the door to
renaming them:
- EVP_PKEY_METHOD is more accurately EVP_PKEY_CTX_METHOD
- EVP_PKEY_ASN1_METHOD is more accurately EVP_PKEY_METHOD
Or perhaps the split doesn't mean much and we should fold them together.
Change-Id: I8a0f7c2e07445dc981c7cef697263e59dba7784e
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57885
Commit-Queue: David Benjamin <davidben@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
I did not have "find a bug in the assembler" on my bingo card today, but
here we are.
NASM 2.15, prior to 2.15.04, has a bug where, if a section that already
exists is referenced again with alignment qualifiers, it incorrect adds
padding and mangles the output. See
https://bugzilla.nasm.us/show_bug.cgi?id=3392701.
Work around this by suppressing the perlasm-emitted qualifiers the
second time a section is emitted. We likely don't need these qualifiers
because, for all sections we care about, NASM's defaults are fine, but
perlasm tries to align .text more aggressively than the default, so let
it do that.
Bug: chromium:1422018
Change-Id: Iade5702c139b70772d4957a83c8f9be86c8af97c
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57825
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
ASN1_item_ex_i2d() does not take ownership of the memory pointed at by
*out, so it's the caller's responsibility to free it on error.
Change-Id: Id8cb70e50f280944418629a32b53fd4ca248b0bd
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57805
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Also turn assertions into static_assert where we can.
These should be no-ops with existing assertions. The int assertion is
tighter, but we already assert this in constant_time_declassify_int. We
cannot support 64-bit int because it messes up integer promotion rules.
Change-Id: I628d2d7decdfa8bc01d8c6013bc7c20f927d63b1
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57785
Reviewed-by: Adam Langley <agl@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
This test could simulate the lack of MADV_WIPEONFORK on systems with it,
but couldn't simulate having it if the kernel didn't support it.
This change makes the presence / absence of
BORINGSSL_IGNORE_MADV_WIPEONFORK control whether MADV_WIPEONFORK is
"supported" in the test environment or not, and corrects the test for
the case where it's missing.
Change-Id: I23876788a0e0a4fd2a148f98b6b94e40880b6fc9
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57745
Auto-Submit: Adam Langley <agl@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
We do not expect to support this combination, but other consumers of
BoringSSL may choose to.
Change-Id: Ifdafa6a0032af078343bb9ecd80eea89eee582be
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57705
Auto-Submit: Adam Langley <agl@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
No new tests because they're actually caught by our own tests. I just
forgot to put UBSan on CI! Will fix this shortly.
For BLAKE2, fix this by checking for zero. For c2i_ASN1_INTEGER,
rewrite it with CBS, which has the side effect of avoiding this. (It's
effectively maintaining in->data + start as a temporary, rather than
start itself.)
Bug: fuchsia:46910
Change-Id: I9366f1ba4fd0b0140d64c56e0534d7b060ab90e5
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57687
Commit-Queue: David Benjamin <davidben@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
If another project includes us as a subproject, as gRPC does,
CMAKE_SOURCE_DIR points to the top-level source directory, not ours.
PROJECT_SOURCE_DIR points to ours. Likewise, CMAKE_BINARY_DIR will
point to the top-level one.
gRPC doesn't consume this CMake build, but in preparation for
eventually unifying the two CMake builds, replace CMAKE_SOURCE_DIR and
CMAKE_BINARY_DIR with a combination of CMAKE_CURRENT_{SOURCE,BINARY}_DIR
and PROJECT_SOURCE_DIR.
There's one more CMAKE_SOURCE_DIR which controls some default install
directory. I've left that one alone for now as I'm not sure what to do
with it. Probably the answer is to, like in gRPC, disable the install
target by default when we're not the top-level source directory.
Bug: 542
Change-Id: Iea26bbef8a4637671fd0e7476101512e871e7e18
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57686
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
We used to have a tower of fallbacks to support older Androids that were
missing getauxval. The comments say getauxval is available in Android
API level 20 or higher, but this wasn't right. It's actually API level
18 or higher per the NDK headers and
https://developer.android.com/ndk/guides/cpu-features
Android API level 18 is Android 4.3, or Jelly Bean MR2. Recent versions
of the NDK (starting r24, March 2022) don't even support Jelly Bean,
i.e. the minimum API level is 19, and the usage statistics in the latest
Android Studio stop at KitKat. As far as I know, nothing needs us to
support API levels 17 and below anymore.
Update-Note: BoringSSL now requires API level 18 or later. Projects
needing to support API level of 17 or below will fail to build due to
the use of getauxval. If any such projects exist, please contact
BoringSSL maintainers.
Change-Id: Iedc4836ffd701428ab6d11253d4ebd5a9121e667
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57506
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Setting the first argument to -1 works, but changing the syscall number
is more straightforward. It's doable on aarch64 too, if we use a
different regset.
Cq-Include-Trybots: luci.boringssl.try:android_aarch64
Change-Id: I6c3c2d3dc67c06a44b181f9086cb5c9d343d51bc
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57587
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
At least on the devices we have on CI, PTRACE_SETREGS seems to break
things when passing in a slightly smaller structure. I'm guessing the
other registers get set to zero, so we SIGILL.
Cq-Include-Trybots: luci.boringssl.try:android_aarch64
Change-Id: I09da595dba2b6b70805c9a79c71c797c0f6635c0
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57585
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Reviewed-by: Adam Langley <agl@google.com>
Change-Id: I736c7dc17efcaa5c3d8bd5fdee36d2dcb86ae627
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57567
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
Various constants and strings identifying the authors are currently
misplaced in .text. This change allows using execute-only .text on
platforms that enforce it by default, such as OpenBSD.
Modify x86_64-xlate.pl to replace .rodata with __DATA,__const for macs.
Adapt the nasm/masm path to emit an .rdata segment with alignment of 8.
This last change is not strictly needed but makes things explicit.
Change-Id: If716b892c1faabd85c6c70bdd75e145304841f83
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57445
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Some Android devices in our builder pool are so old that they lack
getrandom. This change attempts to make them happy.
Change-Id: I5eea04f1b1dc599852e3b8448ad829bea05b9fe9
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57527
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Auto-Submit: Adam Langley <agl@google.com>
When the parameters are incorrect, all assumptions of (EC)DSA fly out
the window, including whether the retry loop actually terminates.
While ECDSA is broadly used with fixed, named groups, DSA was
catastrophically mis-specified with arbitrary parameters being the
default and only mode. Cap the number of retries in DSA_do_sign so
invalid DSA groups cannot infinite loop, e.g. if the "generator" is
really nilpotent.
This also caps the iteration count for ECDSA. We do, sadly, support
arbitrary curves via EC_GROUP_new_curve_GFp, to help Conscrypt remain
compatible with a badly-designed Java API. After
https://boringssl-review.googlesource.com/c/boringssl/+/51925, we
documented that untrusted parameters are not supported and may produce
garbage outputs, but we did not document that infinite loops are
possible. I don't have an example where an invalid curve breaks ECDSA,
but as it breaks all preconditions, I cannot be confident it doesn't
exist, so just cap the iterations.
Thanks to Hanno Böck who originally reported an infinite loop on
invalid DSA groups. While that variation did not affect BoringSSL, it
inspired us to find other invalid groups which did.
Thanks also to Guido Vranken who found, in
https://github.com/openssl/openssl/issues/20268, an infinite loop when
the private key is zero. That was fixed in the preceding CL, as it
impacts valid groups too, but the infinite loop is ultimately in the
same place, so this change also would have mitigated the loop.
Update-Note: If signing starts failing with ECDSA_R_INVALID_ITERATIONS,
something went horribly wrong because it should not be possible with
real curves. (Needing even one retry has probability 2^-256 or so.)
Change-Id: If8fb0157055d3d8cb180fe4f27ea7eb349ec2738
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57228
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
DSA private keys cannot be zero. If they are, trying to sign an all
zeros digest loops forever. Thanks to Guido Vranken who reported this in
https://github.com/openssl/openssl/issues/20268
Along the way, because OpenSSL's bad API design made constructing DSA
objects such a mess, just move all the consistency checks to
dsa_check_parameters (now dsa_check_key) so it is consistently checked
everywhere.
Ideally we'd get a better handle on DSA state, like we hope to do for
RSA state (though not there yet), so checks only happen once. But we
consider DSA deprecated so it's not worth putting much effort into it.
Update-Note: Some invalid DSA keys will be rejected by the parser and at
use. Nothing should be using DSA anymore.
Change-Id: I25d3faf145a85389c47abdd9db8e9b0056b37d8a
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57227
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
This reverts commit 8ce0e1c14e48109773f1e94e5f8b020aa1e24dc5.
The original commit didn't work on Android because:
a) urandom_test didn't handle the fact that Android requires
getrandom() and will never fall back to /dev/urandom.
b) Android may open files in /dev/__properties__ which
confused urandom_test.
The original change is patchset 1 so the differences build on that.
Change-Id: Ib840ec20d60cb28d126d3d09271b18fbd9ec1371
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/53705
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Auto-Submit: Adam Langley <agl@google.com>
We already reject values that are out of bounds. Also we were using the
wrong error, so fix that. Zero should additionally be rejected,
otherwise, when signing an all-zero digest (impossible unless your
system signs untrusted, pre-hashed inputs), ECDSA can infinite loop.
Thanks to Guido Vranken who reported an analogous issue with DSA in
https://github.com/openssl/openssl/issues/20268
When EC_KEYs are obtained through the parser, this CL is a no-op. The
corresponding public key is the point at infinity, which we'll reject at
both parse time and in EC_KEY_check_key. But as EC_KEY runs into
OpenSSL's usual API design flaw (mutable, field-by-field setters over
constructor functions for immutable objects), we should reject this in
EC_KEY_set_private_key too.
Update-Note: Systems that manually construct an EC_KEY (i.e. not from
parsing), and either omit the public key or don't call EC_KEY_check_key
will start rejecting the zero private key. If such a system *also* signs
untrusted digests, this fixes an infinite loop in ECDSA.
Change-Id: I3cc9cd2cc59eb6d16826beab3db71d66b23e83ff
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57226
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
OpenSSL's APIs are full of empty objects that aren't captured in the
type system. A DSA object may represent any of nothing, a group, a
public key, or a private key.
Since the type system doesn't capture this, better to check which type
we've got and NULL-check the fields we need for the operation.
Change-Id: I32b09ebaad58fcb2a0bfc9ad56d381f95099bf7b
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57225
Reviewed-by: Adam Langley <agl@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
This extends the support for execute-only memory in AArch64 assembly
and uses adrp and add instead of adr.
Change-Id: I388a13ec754e7f179d7a234516f1bb4ff6a5c919
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57446
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
This convention seems to break with some other tooling we have. Until we
figure out how to resolve that, remove the lines.
This partially reverts 54b04fdc21d540a6e24f9ddb7ddc3e583518e24f but
keeps the fixes to the license header comments.
Change-Id: I4f08a9f3daf65d17b4c78ac6f4ac3de234ec3436
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57366
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
The union isn't actually providing type-safety: nothing checks that we
access the correct arm of the union, and it has a void* arm anyway.
Instead, it's just adding some strict aliasing risk by relying on
type-punning: we usually write to the pointer as void*, via
EVP_PKEY_assign, but then we read from it as the underlying type.
This is allowed in C, but not in C++. And even in C, while that is
allowed, if we ever wrote &pkey->pkey.rsa, it would suddenly be a strict
aliasing violation. Just use a void*, which means we don't type-pun
pointer types against each other.
While I'm here, I made the free callbacks for EVP_PKEYs also NULL the
pointer. The one caller also NULLs it, so its fine, but some did and
some didn't do it, and this seems prudent.
Bug: 301
Change-Id: I74c76ed3984527df66f64bb2d397af44f63920bd
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57106
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
While hiding 'type' isn't such a huge deal, accessing 'pkey' without a
type check is very dangerous. The accessors are type-checked and avoid
this problem. It also gets us slightly closer to not needing to utter
CRYPTO_refcount_t in public headers, as we're currently not quite
declaring it right. And it allows us to remove another union:
https://boringssl-review.googlesource.com/c/boringssl/+/57106
This matches what upstream did in OpenSSL 1.1.0.
Update-Note: Code that reaches into the EVP_PKEY struct will no longer
compile, like in OpenSSL. I believe I've fixed all the cases. If I
missed any, the fix is to switch code to accessors. EVP_PKEY_id(pkey)
for pkey->type is the most common fix.
Change-Id: Ibe8d6b6cb8fbd141ea1cef0d02dc1ae3703e9469
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57105
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Prior to 3.12 (which we won't be requiring until July), OBJECT libraries
cannot be used with target_link_libraries. That means they cannot pick
up INTERFACE_INCLUDE_DIRECTORIES, which makes them pretty unusable in
the "modern CMake" style.
Just switch it to a static library to unbreak the build in CMake 3.10.
For some link ordering reason I don't understand, this also requires
explicitly linking boringssl_gtest to libcxx when we USE_CUSTOM_LIBCXX
is set.
Change-Id: Ia9d8351551f5da060248aa3ca73fe04473bf62aa
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57345
Commit-Queue: Bob Beck <bbe@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Change-Id: I0b1ba546374aa8b0fe79528f56e19f261536e565
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57305
Commit-Queue: Bob Beck <bbe@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Nothing uses this, and the code is somewhat decrepit. Instead of
fixing it and continuing to maintain it as attack surface, we
send this off to the farm where it can run and play all day with
the other unused X.509 extensions.
Update-note: This removes the proxy certificate extension from
the recognized certificate extensions. Previously by default
a certificate with a critical proxy certificate extension would
have been rejected with "proxy certificate not allowed", but
will now be rejected with an unrecognized critical extension
error.
Fixed: 568
Change-Id: I5f838d69c59517254b4fa83f6e2abe6057fa66c7
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57265
Commit-Queue: Bob Beck <bbe@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Auto-Submit: Bob Beck <bbe@google.com>
This was added with the generated symbol-prefixing header. But it
seems to be sufficient for crypto to have a dependency on the
generated header, along with some of the stray bits of delocate.
It's a little unclear from CMake documentation how these are processed;
normally .o files can be built before libraries are built or linked,
only the link step depends on.
But, empirically, if A links B, and B has a dependency on C, then CMake
seems to run C before building any of A. I tested this by making a small
project where the generation step slept for three seconds and running
with enough parallelism that we'd have tripped.
Interestingly, in the Makefile output, the individual object file
targets didn't have that dependency, but the target itself did. But this
was true on both A and B, so I think that just might not work.
Also fix the dependency in the custom target. The old formulation broke
when using an absolute path to the symbols file.
Change-Id: I2053d44949f907d465da403a5ec69c191740268f
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/56928
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Also stick the very verbose default install directory in a variable so
we don't have to repeat it everywhere.
Change-Id: I1a6a85c4d42d3a6e766e52b2d0ecd8e81c6ed4e3
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/56607
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
It's unclear to me whether doing it target-by-target is an improvement
in crypto/fipsmodule, but this otherwise does seem a bit tidier. This
aligns with CMake's documentation and "modern CMake" which prefers this
pattern.
Change-Id: I36c81842bff8b36eeaaf5dd3e0695fb45f3376c9
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/56585
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Compilers are fine at inlining functions nowadays. We can hide the
BN_ULLONG vs. manual carry extraction inside an inline function. I've
patterned the type signatures intentionally after Clang's builtins, in
case we want to use them in the future.
(Previously I wrote in
https://boringssl-review.googlesource.com/c/boringssl/+/56966 that the
builtins weren't good on aarch64. This wasn't quite right. Rather, they
were bad on both x86_64 and aarch64 in LLVM 13, but they're fine on both
in LLVM 14. My machine's Xcode was just a little old.)
Change-Id: I666466dce7a146d5e49e94ff372ea018b610ef34
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57245
Commit-Queue: David Benjamin <davidben@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Also add public APIs for this, now that the specification is no longer
expected to change, and because a project external to the library wishes
to use it.
For now, I've kept the P-256 version using the generic felem_exp, but we
should update that to use the specialized field arithmetic.
Trust Tokens will presumably move to this later and, in the meantime,
another team wants this.
Bug: chromium:1414562
Change-Id: Ie38203b4439ff55659c4fb2070f45d524c55aa2a
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57147
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Steven Valdez <svaldez@google.com>
draft-07 to draft-16 is mostly editorial, but there were a few notable
changes:
- Empty DST values are forbidden.
- The sample implementation for map_to_curve_simple_swu has completely
changed. The new formulation has the same performance (if not a smidge
faster), and aligning with the spec seems generally useful.
- P-384 is now paired with SHA-384, not SHA-512. As this would be a
breaking change for the trust tokens code, I've left that in. A
follow-up CL will add implementations of draft-16, which is expected
to match the final draft.
Before:
Did 77000 hash-to-curve P384_XMD:SHA-512_SSWU_RO_ operations in 4025677us (19127.2 ops/sec)
Did 7156000 hash-to-scalar P384_XMD:SHA-512 operations in 4000385us (1788827.8 ops/sec)
After:
Did 77000 hash-to-curve P384_XMD:SHA-512_SSWU_RO_ operations in 4009708us (19203.4 ops/sec) [+0.4%]
Did 7327000 hash-to-scalar P384_XMD:SHA-512 operations in 4000477us (1831531.6 ops/sec) [+2.4%]
Bug: 1414562
Change-Id: Ic3c37061e325250d5d8723fd9aa263930c6023cf
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57146
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Steven Valdez <svaldez@google.com>
Commit-Queue: Steven Valdez <svaldez@google.com>
The const bool doesn't do anything. While I'm here, make the methods
const.
Change-Id: Id8c31d5fcda6d8bc244c64b02b1d758e4eff6849
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57185
Auto-Submit: David Benjamin <davidben@google.com>
Commit-Queue: Steven Valdez <svaldez@google.com>
Reviewed-by: Steven Valdez <svaldez@google.com>
Remove all the other ERR_R_MALLOC_FAILURES from the
codebase.
Also changes cbb to push to the error stack, to correctly
report cbb failures instead of now only reporting
malloc failures. Previously it turned all cbb failures
into a malloc failure
Bug: 564
Change-Id: Ic13208bf9d9aaa470e83b2f15782fc94946bbc7b
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57046
Auto-Submit: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Decoding from decimal takes quadratic time, and BN_dec2bn will happily
decode however large of input you pass in. This frustrates fuzzers.
I've added a cap to the input length in s2i_ASN1_INTEGER for now, rather
than BN_dec2bn, because we've seen people use BN for surprisingly large
calculator operations, and BN generally doesn't cap inputs to quadratic
(or worse) algorithms beyond memory limits. (We generally rely on
cryptography using fixed parameter sizes, though RSA, DSA, and DH were
misstandardized and need ad-hoc limits everywhere.)
Update-Note: The stringly-typed API for constructing X.509 extensions
now has (very generous) maximum input length for decimal integers of
8,192 digits. If anyone was relying on a higher input, this will break.
This is unlikely and should be caught by unit tests; if a project hits
this outside of tests, that means they are passing untrusted input into
this function, which is a security vulnerability in itself, and means
they especially need this change to avoid a DoS.
Bug: chromium:1415108
Change-Id: I138249d23ca6b1996f8437dba98633349bb3042b
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57205
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Bob Beck <bbe@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
This will let us call ERR and thread_local from OPENSSL_malloc
without creating a circular dependency. We also make
ERR_get_error_line_data add ERR_FLAG_MALLOCED to the returned
flags value, since some projects appear to be making
assumptions about it being there.
Bug: 564
Update-Note: Any recent documentation (in all OpenSSL forks) for the ERR functions
cautions against freeing the returned ERR "data" strings, as freeing them is handled
by the error library. This change can make an existing double free bug more
obvious by being more likely to cause a crash with the double free.
Change-Id: Ie30bd3aee0b506473988b90675c48510969db31a
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57045
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
Auto-Submit: Bob Beck <bbe@google.com>
This includes an internal version which allows a flag to specify
the use of system malloc, or OPENSSL_malloc - this in turn allows
us to use this function in the ERR family of functions and allow
for ERR to not call OPENSSL_malloc with a circular dependency.
Bug: 564
Change-Id: Ifd02d062fda9695cddbb0dbef2e1c1db0802a486
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57005
Auto-Submit: Bob Beck <bbe@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: Bob Beck <bbe@google.com>
The output of ASN1_generate_v3 is *mostly* linear with the input, except
SEQ and SET reference config sections. Sections can be referenced
multiple times, and so the structure grows exponentially.
Cap the total output size to mitigate this. As before, we don't consider
these functions safe to use with untrusted inputs, but unbounded growth
will frustrate fuzzing. This CL sets the limit to 64K, which should be
enough for anyone. (This is the size of a single X.509 extension,
whereas certificates themselves should not get that large.)
While not strictly necessary, this also rearranges the
ASN1_mbstring_copy call to pass in a maximum output. This portion does
scale linearly with the output, so it's fine, but the fuzzer discovered
an input with a 700K-byte input, which, with fuzzer instrumentation and
sanitizers, seems to be a bit slow. This change should help the fuzzer
get past those cases faster.
Update-Note: The stringly-typed API for constructing X.509 extensions
now has a maximum output size. If anyone was constructing an extension
larger than 64K, this will break. This is unlikely and should be caught
by unit tests; if a project hits this outside of tests, that means they
are passing untrusted input into this function, which is a security
vulnerability in itself, and means they especially need this change to
avoid a DoS.
Bug: oss-fuzz:55725
Change-Id: Ibb65854293f44bf48ed5855016ef7cd46d2fae77
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/57125
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Auto-Submit: David Benjamin <davidben@google.com>
