Callers may as well use ASN1_mbstring_ncopy directly, but some code uses
this, so test it. I've intentionally not tested updating entries because
it crashes if you use a built-in one, and updating a dynamic one seems
unnecessary.
Change-Id: If760a751fbdcd1a2f14d5dcb08de2b0f2a8d3549
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49768
Reviewed-by: Adam Langley <agl@google.com>
There's a test in the file under ifdef, but that is not wired up into
the build.
Change-Id: Iec09277c7ce948c33303d12c325207de2188d908
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49766
Reviewed-by: Adam Langley <agl@google.com>
We have a ton of per-file rotation functions, often with generic names
that do not tell you whether they are uint32_t vs uint64_t, or rotl vs
rotr.
Additionally, (x >> r) | (x << (32 - r)) is UB at r = 0.
(x >> r) | (x << ((-r) & 31)) works for 0 <= r < 32, which is what
cast.c does. GCC and Clang recognize this pattern as a rotate, but MSVC
doesn't. MSVC does, however, provide functions for this.
We usually rotate by a non-zero constant, which makes this moot, but
rotation comes up often enough that it's worth extracting out. Some
particular changes to call out:
- I've switched sha256.c from rotl to rotr. There was a comment
explaining why it differed from the specification. Now that we have
both functions, it's simpler to just match the specification.
- I've dropped all the inline assembly from sha512.c. Compilers should
be able to recognize rotations in 2021.
Change-Id: Ia1030e8bfe94dad92514ed1c28777447c48b82f9
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49765
Reviewed-by: Adam Langley <agl@google.com>
This was added in
https://boringssl-review.googlesource.com/c/boringssl/+/12980/, but does
not appear to be used anymore. The corresponding function does not exist
in OpenSSL.
This simplifies the tests slightly, some of which were inadvertently
specifying the boolean and some weren't.
Change-Id: I9b956dcd9f7151910f93f377d207c88273bd9ccf
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49747
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
The OpenSSL X.509 verifier lacks a proper path builder. When there are
two paths available for a certificate, we pick one without looking at
expiry, etc.
In scenarios like one below, X509_V_FLAG_TRUSTED_FIRST will prefer
Leaf -> Intermediate -> Root1. Otherwise, we will prefer
Leaf -> Intermediate -> Root1Cross -> Root2:
Root2
|
Root1 Root1Cross
\ /
Intermediate
|
Leaf
If Root2 is expired, as with Let's Encrypt, X509_V_FLAG_TRUSTED_FIRST
will find the path we want. Same if Root1Cross is expired. (Meanwhile,
if Root1 is expired, TRUSTED_FIRST will break and leaving it off works.
TRUSTED_FIRST does not actually select chains with validity in mind. It
just changes the semi-arbitrary decision.)
OpenSSL 1.1.x now defaults to X509_V_FLAG_TRUSTED_FIRST by default, so
match them. Hopefully the shorter chain is more likely to be correct.
Update-Note: X509_verify_cert will now build slightly different chains
by default. Hopefully, this fixes more issues than it causes, but there
is a risk of trusted_first breaking other scenarios. Those scenarios
will also break OpenSSL 1.1.x defaults, so hopefully this is fine.
Bug: 439
Change-Id: Ie624f1f7e85a9e8c283f1caf24729aef9206ea16
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49746
Reviewed-by: Adam Langley <agl@google.com>
Reviewed-by: Ryan Sleevi <rsleevi@chromium.org>
There are two ways to configure an X509_STORE_CTX after
X509_STORE_CTX_init. One can either modify the already initialized
X509_VERIFY_PARAM or replace it. Modifying the existing one is more
common. Replacing it actually misses some defaults. (See issue #441 for
details.)
In preparation for actually being able to test changes to the default,
switch tests to that model. In doing so, no longer need to explicitly
configure the depth and can test that default. (Though we should write
tests for the depth at some point.)
Bug: 439, 441
Change-Id: I254a82585d70d44eb94920f604891ebfbff4af4c
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49745
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
All the test vectors testing key length greater than the block length
were mislabelled as key length being equal to the block length. Add a
note to these test vectors indicating they are directly from the NIST
tests with the misleading input intact.
Change-Id: I9fe87971265ad48e9b835fccbe92306e1670b4d6
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49705
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Trusty's TLOGE macro nowadays expects TLOG_TAG to be defined
as the log tag to use.
Change-Id: I18121287ba51698d354323027d5382c8406f0b99
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49685
Commit-Queue: Pete Bentley <prb@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
In upstream, these functions take file and line number arguments. Update
ours to match. Guessing almost no one uses these, or we'd have caught
this earlier.
Change-Id: Ic09f8d8274065ac02efa78e70c215b87fa765b9f
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49665
Reviewed-by: Adam Langley <agl@google.com>
Reviewed-by: Benjamin Brittain <bwb@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Node seems uncommonly sensitive to this, so let's write these functions
in a way that stays in sync and test this. See also
https://boringssl-review.googlesource.com/c/boringssl/+/49585
This does incur a cost across all BoringSSL consumers that use these
functions: as a result of Node indiscriminately exposing every cipher,
we end up pulling more and more ciphers into these getters. But that
ship sailed long ago, so, instead, document that EVP_get_cipherby*
should not be used by size-conscious callers.
EVP_get_digestby* probably should have the same warning, but I've left
it alone for now because we don't quite have the same proliferation of
digests as ciphers. (Though there are things in there, like MD4, that
ought to be better disconnected.)
Change-Id: I61ca406c146279bd05a52bed6c57200d1619c5da
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49625
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
There is an obvious bug there: upon entry to 'vpaes_cbc_encrypt'
LR may get signed. However, on the 'cbc_abort' path the LR is
not going to be unsigned before 'ret' is executed.
Found by manual code inspection.
Co-authored-by: Russ Butler <russ.butler@arm.com>
Change-Id: I646cdfaee28db59aafbbd412d4bb6ba022eff15b
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49605
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Linux module signing uses PKCS#7 / CMS because everything is awful and
broken. In order to make the lives of kernel developers easier, support
the calling pattern that the kernel uses to sign modules.
The kernel utility was written at a time when PKCS#7 was hard coded to
use SHA-1 for signing in OpenSSL and it reflects this: you can only
specify “sha1” on the command line, for example. As of OpenSSL 1.1.1, at
least, OpenSSL uses SHA-256 and thus so does this change.
Change-Id: I32b036123a0d8b272ec9e1c0130c45bf3ed0d2c7
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49545
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
I was inspired to look at this again recently and noticed we could do a
bit better. Instead of a tower of selects, rely on all the cases being
mutually exclusive and use the ret |= mask & value formulation without
loss in clarity. We do need to fixup the invalid case slightly, but
since that computation is mostly independent, I'm guessing the CPU and
compiler are able to schedule it effectively.
Before:
Did 251000 base64 decode operations in 2002569us (159.4 MB/sec)
After:
Did 346000 base64 decode operations in 2005426us (219.5 MB/sec) [+37.7%]
Change-Id: I542167202fd4e94c93dd5a2519a97bc388072c89
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49525
Reviewed-by: Adam Langley <agl@google.com>
The i2d functions internally take a tag/class pair of parameters. If tag
is not -1, we override the tag with (tag, class). Otherwise, class is
ignored. (class is inconsistently called aclass or iclass.)
Historically, the remaning bits of class were repurposed to pass extra
flags down the structure. These had to be preserved in all recursive
calls, so the functions take apart and reassemble the two halves of
aclass/iclass. The only such flag was ASN1_TFLG_NDEF, which on certain
types, caused OpenSSL to encode indefinite-length encoding. We removed
this in https://boringssl-review.googlesource.com/c/boringssl/+/43889.
Due to these flags, if tag == -1, class should default to zero. However,
X509_NAME's callbacks pass -1, -1, instead of -1, 0, effectively setting
all flags. This wasn't noticed because none of the types below X509_NAME
pay attention to ASN1_TFLG_NDEF.
This CL does two things: First, it unwinds the remainder of the flags
machinery. If we ever need flags, we should pass it as a distinct
argument. Second, it fixes the X509_NAME calls and asserts that -1 is
always paired with 0.
Change-Id: I285a73a06ad16980617fe23d5ea7f260fc5dbf16
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49385
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CS3 is ciphertext-stealing variant three from SP 800-38A.
Change-Id: I992dc22778c91efad361f25ff65ae5966fc447c6
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49505
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Replace the hardcoded ECH config, which wasn't updated for draft-13,
with a call to SSL_marshal_ech_config.
Bug: 275, oss-fuzz:38054
Change-Id: I10c12b22015c9c0cb90dd6185eb375153a2531f4
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49445
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
Having APIs named "session" and "ID" appears to be far too tempting for
developers, mistaking it as some application-level notion of session.
Update the documentation, in hopes of discouraging this mistake.
Change-Id: Ifd9516287092371d4701114771eff6640df1bcb0
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49405
Reviewed-by: Adam Langley <agl@google.com>
This doesn't affect RSA key generation, which uses
BN_prime_checks_for_generation.
Change-Id: Ibf32c0c4bc9fed369e8f8a1efea72c5bd39185a9
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49426
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
These are a little odd with the ASN1_ENCODING paths. And there were some
bugs previously around CHOICE types. Nothing defines them, inside or
outside BoringSSL, so remove them.
Change-Id: Id2954fef8ee9637f36f7511b51dc0adc2557e3ba
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49352
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
it->funcs is only an ASN1_AUX for ASN1_ITYPE_SEQUENCE and
ASN1_ITYPE_CHOICE. Fortunately, the other possible types for it->funcs
are larger than ASN1_AUX and we don't touch the result when we
shouldn't, so this is merely a strict aliasing violation.
Change-Id: I29e94249e0b137fe8df0b16254366ae6705c8784
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49351
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though
this CL uses a different strategy from upstream. Upstream makes
ASN1_item_ex_i2d continue to allow optionals and checks afterwards at
every non-optional call site. This CL pushes down an optional parameter
and says functions cannot omit items unless explicitly allowed.
I think this is a better default, though it is a larger change. Fields
are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL
flag. Upstream's strategy misses top-level calls.
This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts
where it doesn't make sense. Only fields of SEQUENCEs and SETs may be
OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match
ASN.1 itself. ASN1_TEMPLATE is additionally responsible for
explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms
and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE
but will get confused if marked optional.
As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully"
writing 0 bytes. If we want to allow NULL at the top-level, that's not
too hard to arrange, but our CBB-based i2d functions do not.
Update-Note: Structures with missing mandatory fields can no longer be
encoded. Note that, apart from the cases already handled by preceding
CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main
downstream impact I've seen of this particular change is in combination
with other bugs. Consider a caller that does:
GENERAL_NAME *name = GENERAL_NAME_new();
name->type = GEN_DNS;
name->d.dNSName = DoSomethingComplicated(...);
Suppose DoSomethingComplicated() was actually fallible and returned
NULL, but the caller forgot to check. They'd now construct a
GENERAL_NAME with a missing field. Previously, this would silently
serialize some garbage (omitted field) or empty string. Now we fail to
encode, but the true error was the uncaught DoSomethingComplicated()
failure. (Which likely was itself a bug.)
Bug: 429
Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
See https://github.com/openssl/openssl/issues/16538
Update-Note: A default-constructed object with a required ANY or
string-like CHOICE field cannot be encoded until the field is specified.
Note this affects i2d_X509: notBefore and notAfter are string-like
CHOICEs in OpenSSL.
Bug: 429
Change-Id: I97d971fa588ab72be25a4c1eb7310ed330f16c4f
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49349
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
asn1_ex_i2c actually does have an error condition, it just wasn't being
handled.
628b3c7f2f, imported from upstream's
f3f8e72f494b36d05e0d04fe418f92b692fbb261, tried to check for OID-less
ASN1_OBJECTs and return an error. But it and the upstream change didn't
actually work. -1 in this function means to omit the object, so OpenSSL
was silently misinterpreting the input structure.
This changes the calling convention for asn1_ex_i2c to support this. It
is, unfortunately, a little messy because:
1. One cannot check for object presense without walking the
ASN1_ITEM/ASN1_TEMPLATE structures. You can *almost* check if *pval
is NULL, but ASN1_BOOLEAN is an int with -1 to indicate an omitted
optional. There are also FBOOLEAN/TBOOLEAN types that omit FALSE/TRUE
for DEFAULT. Thus, without more invasive changes, asn1_ex_i2c must be
able to report an omitted element.
2. While the i2d functions report an omitted element by successfully
writing zero bytes, i2c only writes the contents. It thus must
distinguish between an omitted element and an element with
zero-length contents.
3. i2c_ASN1_INTEGER and i2c_ASN1_BIT_STRING return zero on error rather
than -1. Those error paths are not actually reachable because they
only check for NULL. In fact, OpenSSL has even unexported them. But I
found a few callers. Rather than unwind all this and change the
calling convention, I've just made it handle 0 and map to -1 for now.
It's all a no-op anyway, and hopefully we can redo all this with CBB
later.
I've just added an output parameter for now.
In writing tests, I also noticed that the hand-written i2d_ASN1_OBJECT
and i2d_ASN1_BOOLEAN return the wrong value for errors, so I've fixed
that.
Update-Note: A default-constructed object with a required ASN1_OBJECT
field can no longer be encoded without initializing the ASN1_OBJECT.
Note this affects X509: the signature algorithm is an ASN1_OBJECT. Tests
that try to serialize an X509_new() must fill in all required fields.
(Production code is unlikely to be affected because the output was
unparsable anyway, while tests sometimes wouldn't notice.)
Bug: 429
Change-Id: I04417f5ad6b994cc5ccca540c8a7714b9b3af33d
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49348
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
This handles normal CHOICE types. A follow-up CL will handle MSTRING and
ANY types.
Update-Note: An invalid CHOICE object (e.g. GENERAL_NAME) will now fail
when encoded, rather than be silently omitted. In particular, CHOICE
objects are default-initialized by tasn_new.c in an empty -1 state.
Structures containing a required CHOICE field can no longer be encoded
without filling in the CHOICE.
Bug: 429
Change-Id: I7011deadf518ddc344a56b07a0e268ceaae17fe0
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49347
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
This function forgot to handle errors in ASN1_item_ex_i2d. It also
checked x509_name_canon for ret < 0, when x509_name_canon returns a
boolean. For consistency, I've switched to x509_name_encode to return a
boolean as well. It doesn't actually need to return a length because
it's responsible for filling in a->bytes.
(This is also far from thread-safe, but I'll figure out what to do there
separately.)
Bug: 429
Change-Id: I1dddeab320018be4b837f95001cbeeba4e25f0a1
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49346
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
tasn_enc.c was missing lots of error checks and mixed up 0 and -1
returns. Document all the internal calling conventions, as best as I can
tell, and fix things up.
There are also error cases it forgets to check (it generally does not
notice missing non-OPTIONAL fields). This CL only addresses errors it
already tries to report. Subsequent CLs will add in the missing error
cases. And then if it all sticks, I'm hoping we can rewrite this with
CBB. Rewriting tsan_dec.c to CBS would also be good, but that will be
more difficult as we need to clear out BER first.
Update-Note: Some error cases which were silently misinterpreted as
missing OPTIONAL elements will now cause encoding to fail.
Bug: 429
Change-Id: Ibbb3eba08eb8f8f878930c9456edc8c74479aade
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49345
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
GCC 11.2.1 reportedly warns that CTR_DRBG_init may be passed an
uninitialized personalization buffer. This appears to be a false
positive, because personalization_len will be zero. But it's easy enough
to zero-initialize it, so silence the warning.
Bug: 432
Change-Id: I20f6b74e09f19962e8cae37d45090ff3d1c0215d
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49245
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Later CLs will clean up the ClientHello construction a bit (draft-12
avoids computing ClientHelloOuter twice). I suspect the transcript
handling on the client can also be simpler, but I'll see what's
convenient after I've changed how ClientHelloOuter is constructed.
Changes of note between draft-10 and draft-13:
- There is now an ECH confirmation signal in both HRR and SH. We don't
actually make much use of this in our client right now, but it
resolves a bunch of weird issues around HRR, including edge cases if
HRR applies to one ClientHello but not the other.
- The confirmation signal no longer depends on key_share and PSK, so we
don't have to work around a weird ordering issue.
- ech_is_inner is now folded into the main encrypted_client_hello code
point. This works better with some stuff around HRR.
- Padding is moved from the padding extension, computed with
ClientHelloInner, to something we fill in afterwards. This makes it
easier to pad up the whole thing to a multiple of 32. I've accordingly
updated to the latest recommended padding construction, and updated
the GREASE logic to match.
- ech_outer_extensions is much easier to process because the order is
required to be consistent. We were doing that anyway, and now a simple
linear scan works.
- ClientHelloOuterAAD now uses an all zero placeholder payload of the
same length. This lets us simplify the server code, but, for now, I've
kept the client code the same. I'll follow this up with a CL to avoid
computing ClientHelloOuter twice.
- ClientHelloOuterAAD is allowed to contain a placeholder PSK. I haven't
filled that in and will do it in a follow-up CL.
Bug: 275
Change-Id: I7464345125c53968b2fe692f9268e392120fc2eb
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48912
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
Hopefully it's a little clearer that this may be called whether or not
ECH is offered. (And whether or not it's a server.)
Bug: 275
Change-Id: I39c8ce5758543a0cfda84652b3fc0a5b9669fd0a
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49165
Reviewed-by: Matt Mueller <mattm@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
This unexports X509, X509_CINF, X509_NAME_ENTRY, X509_NAME, X509_OBJECT,
X509_LOOKUP_METHOD, X509_STORE, X509_LOOKUP, and X509_STORE_CTX.
Note this means X509_STORE_CTX can no longer be stack-allocated.
Update-Note: Patch cl/390055173 into the roll that includes this. This
unexports most of the X.509 structs, aligning with OpenSSL. Use the
accessor APIs instead.
Bug: 425
Change-Id: I53e915bfae3b8dc4b67642279d0e54dc606f2297
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48985
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
V_ASN1_APP_CHOOSE has been discouraged by OpenSSL since 2000:
https://git.openssl.org/gitweb/?p=openssl.git;a=blob;f=CHANGES;h=824f421b8d331ba2a2009dbda333a57493bedb1e;hb=fb047ebc87b18bdc4cf9ddee9ee1f5ed93e56aff#l10848
Instead, upstream recommends an MBSTRING_* constant.
https://www.openssl.org/docs/man1.1.1/man3/X509_NAME_add_entry_by_NID.html
This function is a bit overloaded:
MBSTRING_* means "Decode my input from this format and then re-encode it
using whatever string type best suits the NID (usually UTF8String, but
some NIDs require PrintableString)".
V_ASN1_APP_CHOOSE means "This is a Latin-1 string. Without looking at
the NID, pick one of PrintableString, IA5String, or T61String".
The latter is almost certainly not what callers want. If they want a
particular type, they can always force it by passing a particular
V_ASN1_* constant. This removes the only use of ASN1_PRINTABLE_type
within the library, though there is one external use still.
Update-Note: V_ASN1_APP_CHOOSE is removed. I only found one use, which
has been fixed.
Change-Id: Id36376dd0ec68559bbbb366e2305d42be5ddac67
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49067
Reviewed-by: Adam Langley <agl@google.com>
The old loop read one byte past the length. It also stopped the loop
too early on interior NUL. See also upstream's
https://github.com/openssl/openssl/pull/16433, though I've opted to
rewrite the function entirely rather than use their fix.
Also deduplicate the PrintableString check.
Change-Id: Ia8bd282047c2a2ed1d5e71a68a3947c7c108df95
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49066
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
strchr is interprets the trailing NUL as part of the string, so
is_printable thought NUL was allowed. Just write the code in the obvious
way and let the compiler figure it out. (It seems to make a clever
bitmask or something.)
Update-Note: ASN1_mbstring_ncopy will no longer allow PrintableString
for strings containing NUL.
Change-Id: I3675191ceb44c06f0ac7b430f88272cabf392d35
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49065
Reviewed-by: Adam Langley <agl@google.com>
The bulk of RSA_check_key is spent in bn_div_consttime, which is a naive
but constant-time long-division algorithm for the few places that divide
by a secret even divisor: RSA keygen and RSA import. RSA import is
somewhat performance-sensitive, so pick some low-hanging fruit:
The main observation is that, in all but one call site, the bit width of
the divisor is public. That means, for an N-bit divisor, we can skip the
first N-1 iterations of long division because an N-1-bit remainder
cannot exceed the N-bit divisor.
One minor nuisance is bn_lcm_consttime, used in RSA keygen has a case
that does *not* have a public bit width. Apply the optimization there
would leak information. I've implemented this as an optional public
lower bound on num_bits(divisor), which all but that call fills in.
Before:
Did 5060 RSA 2048 private key parse operations in 1058526us (4780.2 ops/sec)
Did 1551 RSA 4096 private key parse operations in 1082343us (1433.0 ops/sec)
After:
Did 11532 RSA 2048 private key parse operations in 1084145us (10637.0 ops/sec) [+122.5%]
Did 3542 RSA 4096 private key parse operations in 1036374us (3417.7 ops/sec) [+138.5%]
Bug: b/192484677
Change-Id: I893ebb8886aeb8200a1a365673b56c49774221a2
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49106
Reviewed-by: Adam Langley <agl@google.com>
We do non-trivial work when parsing RSA private keys (RSA_check_key)
and, in some consumers, this is performance-sensitive.
Bug: b/192484677
Change-Id: Ic27f5f11d8bd030de77dd500a826fb2dd7c5b75d
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49105
Reviewed-by: Adam Langley <agl@google.com>
Although we defined a CBS -> Span<const uint8_t> conversion, MSVC 2015
keeps trying to call the Span(const Container&) constructor. It seems to
not correctly SFINAE the existence of data() and size() members unless
the expression is inlined into the default template argument.
Change-Id: I4e88f820b78ce72ad1b014b5bae0830bc7d099d4
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48945
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
tls13_init_key_schedule calls InitHash internally, but we also call
InitHash earlier at various times. On the client, we do it early to
handle HelloRetryRequest and 0-RTT. ECH draft-12 will also need to do it
early. Apparently we do it early on the server too.
Probably tls13_init_key_schedule doesn't need to call InitHash, but for
now, it is an easy check in SSLTranscript.
Change-Id: I5473047c1f29bdeb60901e4e6e80979e592bd6e9
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48911
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
David Benjamin
Kenny Root
Pete Bentley
Tamas Petz
Adam Langley
jakemas
Shelley Vohr
Adam Langley