For now, skip the 1.2 -> 1.1 signal since that will affect shipping
code. We may as well enable it too, but wait until things have settled
down. This implements the version in draft-14 since draft-13's isn't
backwards-compatible.
Change-Id: I46be43e6f4c5203eb4ae006d1c6a2fe7d7a949ec
Reviewed-on: https://boringssl-review.googlesource.com/8724
Reviewed-by: David Benjamin <davidben@google.com>
Rather than blindly select SHA-1 if we can't find a matching one, act as
if the peer advertised rsa_pkcs1_sha1 and ecdsa_sha1. This means that we
will fail the handshake if no common algorithm may be found.
This is done in preparation for removing the SHA-1 default in TLS 1.3.
Change-Id: I3584947909d3d6988b940f9404044cace265b20d
Reviewed-on: https://boringssl-review.googlesource.com/8695
Reviewed-by: David Benjamin <davidben@google.com>
In order to delay the digest of the handshake transcript and unify
around message-based signing callbacks, a copy of the transcript is kept
around until we are sure there is no certificate authentication.
This removes support for SSL_PRIVATE_KEY_METHOD as a client in SSL 3.0.
Change-Id: If8999a19ca021b4ff439319ab91e2cd2103caa64
Reviewed-on: https://boringssl-review.googlesource.com/8561
Reviewed-by: David Benjamin <davidben@google.com>
This also adds a missing check to the C half to ensure fake record types are
always correct, to keep implementations honest.
Change-Id: I1d65272e647ffa67018c721d52c639f8ba47d647
Reviewed-on: https://boringssl-review.googlesource.com/8510
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
The case where ec_group_get_mont_data is NULL is only for arbitrary groups
which we now require to be prime order. BN_mod_exp_mont is fine with a NULL
BN_MONT_CTX. It will just compute it. Saves a bit of special-casing.
Also don't mark p-2 as BN_FLG_CONSTTIME as the exponent is public anyway.
Change-Id: Ie868576d52fc9ae5f5c9f2a4039a729151bf84c7
Reviewed-on: https://boringssl-review.googlesource.com/8307
Reviewed-by: Adam Langley <agl@google.com>
There's a __pragma expression which allows this. Android builds us Windows with
MinGW for some reason, so we actually do have to tolerate non-MSVC-compatible
Windows compilers. (Clang for Windows is much more sensible than MinGW and
intentionally mimicks MSVC.)
MinGW doesn't understand MSVC's pragmas and warns a lot. #pragma warning is
safe to suppress, so wrap those to shush them. This also lets us do away with a
few ifdefs.
Change-Id: I1f5a8bec4940d4b2d947c4c1cc9341bc15ec4972
Reviewed-on: https://boringssl-review.googlesource.com/8236
Reviewed-by: Adam Langley <agl@google.com>
Windows SRWLOCK requires you call different functions here. Split
them up in preparation for switching Windows from CRITICAL_SECTION.
BUG=37
Change-Id: I7b5c6a98eab9ae5bb0734b805cfa1ff334918f35
Reviewed-on: https://boringssl-review.googlesource.com/8080
Reviewed-by: Adam Langley <agl@google.com>
Also add a test.
This is the last of the openssl/asn1.h includes from the directories that are
to be kept in the core libcrypto library. (What remains is to finish sorting
out the crypto/obj stuff. We'll also want to retain a decoupled version of the
PKCS#12 stuff.)
Functions that need to be audited for reuse:
i2d_DHparams
BUG=54
Change-Id: Ibef030a98d3a93ae26e8e56869f14858ec75601b
Reviewed-on: https://boringssl-review.googlesource.com/7900
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Sanity check field lengths and sums to avoid potential overflows and reject
excessively large X509_NAME structures.
Issue reported by Guido Vranken.
(Imported from upstream's 9b08619cb45e75541809b1154c90e1a00450e537.)
Change-Id: Ib2e1e7cd086f9c3f0d689d61947f8ec3e9220049
Reviewed-on: https://boringssl-review.googlesource.com/7842
Reviewed-by: Adam Langley <agl@google.com>
I don't think I ever look at that output. This way our builds are nice and
silent.
Change-Id: Idb215e3702f530a8b8661622c726093530885c91
Reviewed-on: https://boringssl-review.googlesource.com/7700
Reviewed-by: Adam Langley <agl@google.com>
I messed up a few of these.
ASN1_R_UNSUPPORTED_ALGORITHM doesn't exist. X509_R_UNSUPPORTED_ALGORITHM does
exist as part of X509_PUBKEY_set, but the SPKI parser doesn't emit this. (I
don't mind the legacy code having really weird errors, but since EVP is now
limited to things we like, let's try to keep that clean.) To avoid churn in
Conscrypt, we'll keep defining X509_R_UNSUPPORTED_ALGORITHM, but not actually
do anything with it anymore. Conscrypt was already aware of
EVP_R_UNSUPPORTED_ALGORITHM, so this should be fine. (I don't expect
EVP_R_UNSUPPORTED_ALGORITHM to go away. The SPKI parsers we like live in EVP
now.)
A few other ASN1_R_* values didn't quite match upstream, so make those match
again. Finally, I got some of the rsa_pss.c values wrong. Each of those
corresponds to an (overly specific) RSA_R_* value in upstream. However, those
were gone in BoringSSL since even the initial commit. We placed the RSA <-> EVP
glue in crypto/evp (so crypto/rsa wouldn't depend on crypto/evp) while upstream
placed them in crypto/rsa.
Since no one seemed to notice the loss of RSA_R_INVALID_SALT_LENGTH, let's undo
all the cross-module errors inserted in crypto/rsa. Instead, since that kind of
specificity is not useful, funnel it all into X509_R_INVALID_PSS_PARAMETERS
(formerly EVP_R_INVALID_PSS_PARAMETERS, formerly RSA_R_INVALID_PSS_PARAMETERS).
Reset the error codes for all affected modules.
(That our error code story means error codes are not stable across this kind of
refactoring is kind of a problem. Hopefully this will be the last of it.)
Change-Id: Ibfb3a0ac340bfc777bc7de6980ef3ddf0a8c84bc
Reviewed-on: https://boringssl-review.googlesource.com/7458
Reviewed-by: Emily Stark (Dunn) <estark@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
People seem to condition on these a lot. Since this code has now been moved
twice, just make them all cross-module errors rather than leave a trail of
renamed error codes in our wake.
Change-Id: Iea18ab3d320f03cf29a64a27acca119768c4115c
Reviewed-on: https://boringssl-review.googlesource.com/7431
Reviewed-by: Emily Stark (Dunn) <estark@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
A number of values have fallen off now that code's been shuffled
around.
Change-Id: I5eac1d3fa4a9335c6aa72b9876d37bb9a9a029ac
Reviewed-on: https://boringssl-review.googlesource.com/7029
Reviewed-by: Adam Langley <agl@google.com>
All the signature algorithm logic depends on X509_ALGOR. This also
removes the X509_ALGOR-based EVP functions which are no longer used
externally. I think those APIs were a mistake on my part. The use in
Chromium was unnecessary (and has since been removed anyway). The new
X.509 stack will want to process the signatureAlgorithm itself to be
able to enforce policies on it.
This also moves the RSA_PSS_PARAMS bits to crypto/x509 from crypto/rsa.
That struct is also tied to crypto/x509. Any new RSA-PSS code would
have to use something else anyway.
BUG=499653
Change-Id: I6c4b4573b2800a2e0f863d35df94d048864b7c41
Reviewed-on: https://boringssl-review.googlesource.com/7025
Reviewed-by: Adam Langley <agl@google.com>
Calling SSL_shutdown while in init previously gave a "1" response,
meaning everything was successfully closed down (even though it
wasn't). Better is to send our close_notify, but fail when trying to
receive one.
The problem with doing a shutdown while in the middle of a handshake
is that once our close_notify is sent we shouldn't really do anything
else (including process handshake/CCS messages) until we've received a
close_notify back from the peer. However the peer might send a CCS
before acting on our close_notify - so we won't be able to read it
because we're not acting on CCS messages!
(Imported from upstream's f73c737c7ac908c5d6407c419769123392a3b0a9)
Change-Id: Iaad5c5e38983456d3697c955522a89919628024b
Reviewed-on: https://boringssl-review.googlesource.com/7207
Reviewed-by: David Benjamin <davidben@google.com>
Many consumers need SPKI support (X.509, TLS, QUIC, WebCrypto), each
with different ways to set signature parameters. SPKIs themselves can
get complex with id-RSASSA-PSS keys which come with various constraints
in the key parameters. This suggests we want a common in-library
representation of an SPKI.
This adds two new functions EVP_parse_public_key and
EVP_marshal_public_key which converts EVP_PKEY to and from SPKI and
implements X509_PUBKEY functions with them. EVP_PKEY seems to have been
intended to be able to express the supported SPKI types with
full-fidelity, so these APIs will continue this.
This means future support for id-RSASSA-PSS would *not* repurpose
EVP_PKEY_RSA. I'm worried about code assuming EVP_PKEY_RSA implies
acting on the RSA* is legal. Instead, it'd add an EVP_PKEY_RSA_PSS and
the data pointer would be some (exposed, so the caller may still check
key size, etc.) RSA_PSS_KEY struct. Internally, the EVP_PKEY_CTX
implementation would enforce the key constraints. If RSA_PSS_KEY would
later need its own API, that code would move there, but that seems
unlikely.
Ideally we'd have a 1:1 correspondence with key OID, although we may
have to fudge things if mistakes happen in standardization. (Whether or
not X.509 reuses id-ecPublicKey for Ed25519, we'll give it a separate
EVP_PKEY type.)
DSA parsing hooks are still implemented, missing parameters and all for
now. This isn't any worse than before.
Decoupling from the giant crypto/obj OID table will be a later task.
BUG=522228
Change-Id: I0e3964edf20cb795a18b0991d17e5ca8bce3e28c
Reviewed-on: https://boringssl-review.googlesource.com/6861
Reviewed-by: Adam Langley <agl@google.com>
Functions which lose object reuse and need auditing:
- d2i_ECParameters
- d2i_ECPrivateKey
This adds a handful of bytestring-based APIs to handle EC key
serialization. Deprecate all the old serialization APIs. Notes:
- An EC_KEY has additional state that controls its encoding, enc_flags
and conv_form. conv_form is left alone, but enc_flags in the new API
is an explicit parameter.
- d2i_ECPrivateKey interpreted its T** argument unlike nearly every
other d2i function. This is an explicit EC_GROUP parameter in the new
function.
- The new specified curve code is much stricter and should parse enough
to uniquely identify the curve.
- I've not bothered with a new version of i2d_ECParameters. It just
writes an OID. This may change later when decoupling from the giant
OID table.
- Likewise, I've not bothered with new APIs for the public key since the
EC_POINT APIs should suffice.
- Previously, d2i_ECPrivateKey would not call EC_KEY_check_key and it
was possible for the imported public and private key to mismatch. It
now calls it.
BUG=499653
Change-Id: I30b4dd2841ae76c56ab0e1808360b2628dee0615
Reviewed-on: https://boringssl-review.googlesource.com/6859
Reviewed-by: Adam Langley <agl@google.com>
There was a TODO to remove it once asn1_mac.h was trimmed. This has now
happened. Remove it and reset error codes for crypto/asn1.
Change-Id: Iaf2f3e75741914415372939471b135618910f95d
Reviewed-on: https://boringssl-review.googlesource.com/6761
Reviewed-by: Adam Langley <agl@google.com>
38 error codes have fallen off the list since the last time we did this.
Change-Id: Id7ee30889a5da2f6ab66957fd8e49e97640c8489
Reviewed-on: https://boringssl-review.googlesource.com/6643
Reviewed-by: Adam Langley <agl@google.com>
dh.c had a 10k-bit limit but it wasn't quite correctly enforced. However,
that's still 1.12s of jank on the IO thread, which is too long. Since the SSL
code consumes DHE groups from the network, it should be responsible for
enforcing what sanity it needs on them.
Costs of various bit lengths on 2013 Macbook Air:
1024 - 1.4ms
2048 - 14ms
3072 - 24ms
4096 - 55ms
5000 - 160ms
10000 - 1.12s
UMA says that DHE groups are 0.2% 4096-bit and otherwise are 5.5% 2048-bit and
94% 1024-bit and some noise. Set the limit to 4096-bit to be conservative,
although that's already quite a lot of jank.
BUG=554295
Change-Id: I8e167748a67e4e1adfb62d73dfff094abfa7d215
Reviewed-on: https://boringssl-review.googlesource.com/6464
Reviewed-by: Adam Langley <agl@google.com>
Removing the function codes continued to sample __func__ for compatibility with
ERR_print_errors_cb, but not ERR_error_string_n. We can just emit
OPENSSL_internal for both. ERR_print_errors_cb already has the file and line
number available which is strictly more information than the function name.
(ERR_error_string_n does not, but we'd already turned that to
OPENSSL_internal.)
This shaves 100kb from a release build of the bssl tool.
In doing so, put an unused function code parameter back into ERR_put_error to
align with OpenSSL. We don't need to pass an additional string in anymore, so
OpenSSL compatibility with anything which uses ERR_LIB_USER or
ERR_get_next_error_library costs nothing. (Not that we need it.)
Change-Id: If6af34628319ade4145190b6f30a0d820e00b20d
Reviewed-on: https://boringssl-review.googlesource.com/6387
Reviewed-by: Adam Langley <agl@google.com>
It's very annoying having to remember the right incant every time I want
to switch around between my build, build-release, build-asan, etc.,
output directories.
Unfortunately, this target is pretty unfriendly without CMake 3.2+ (and
Ninja 1.5+). This combination gives a USES_TERMINAL flag to
add_custom_target which uses Ninja's "console" pool, otherwise the
output buffering gets in the way. Ubuntu LTS is still on an older CMake,
so do a version check in the meantime.
CMake also has its own test mechanism (CTest), but this doesn't use it.
It seems to prefer knowing what all the tests are and then tries to do
its own output management and parallelizing and such. We already have
our own runners. all_tests.go could actually be converted tidily, but
generate_build_files.py also needs to read it, and runner.go has very
specific needs.
Naming the target ninja -C build test would be nice, but CTest squats
that name and CMake grumps when you use a reserved name, so I've gone
with run_tests.
Change-Id: Ibd20ebd50febe1b4e91bb19921f3bbbd9fbcf66c
Reviewed-on: https://boringssl-review.googlesource.com/6270
Reviewed-by: Adam Langley <alangley@gmail.com>
If the two extensions select different next protocols (quite possible since one
is server-selected and the other is client-selected), things will break. This
matches the behavior of NSS (Firefox) and Go.
Change-Id: Ie1da97bf062b91a370c85c12bc61423220a22f36
Reviewed-on: https://boringssl-review.googlesource.com/5780
Reviewed-by: Adam Langley <agl@google.com>
arm_arch.h is included from ARM asm files, but lives in crypto/, not
openssl/include/. Since the asm files are often built from a different
location than their position in the source tree, relative include paths
are unlikely to work so, rather than having crypto/ be a de-facto,
second global include path, this change moves arm_arch.h to
include/openssl/.
It also removes entries from many include paths because they should be
needed as relative includes are always based on the locations of the
source file.
Change-Id: I638ff43d641ca043a4fc06c0d901b11c6ff73542
Reviewed-on: https://boringssl-review.googlesource.com/5746
Reviewed-by: Adam Langley <agl@google.com>
MSVC and clang-cl automatically define |_WIN32| but |WIN32| is only
defined if a Windows header file has been included or if -DWIN32 was
passed on the command line. Thus, it is always better to test |_WIN32|
than |WIN32|. The convention in BoringSSL is to test |OPENSSL_WINDOWS|
instead, except for the place where |OPENSSL_WINDOWS| is defined.
Change-Id: Icf3e03958895be32efe800e689d5ed6a2fed215f
Reviewed-on: https://boringssl-review.googlesource.com/5553
Reviewed-by: David Benjamin <davidben@chromium.org>
Reviewed-by: Adam Langley <agl@google.com>
This change mirrors upstream's custom extension API because we have some
internal users that depend on it.
Change-Id: I408e442de0a55df7b05c872c953ff048cd406513
Reviewed-on: https://boringssl-review.googlesource.com/5471
Reviewed-by: Adam Langley <agl@google.com>
Fastradio was a trick where the ClientHello was padding to at least 1024
bytes in order to trick some mobile radios into entering high-power mode
immediately. After experimentation, the feature is being dropped.
This change also tidies up a bit of the extensions code now that
everything is using the new system.
Change-Id: Icf7892e0ac1fbe5d66a5d7b405ec455c6850a41c
Reviewed-on: https://boringssl-review.googlesource.com/5466
Reviewed-by: Adam Langley <agl@google.com>
Consumers sometimes use ERR_LIB_USER + <favorite number> instead of
ERR_get_next_error_library. To avoid causing them grief, keep ERR_LIB_USER
last.
Change-Id: Id19ae7836c41d5b156044bd20d417daf643bdda2
Reviewed-on: https://boringssl-review.googlesource.com/5290
Reviewed-by: Adam Langley <agl@google.com>
Running make_errors.go every time a function is renamed is incredibly
tedious. Plus we keep getting them wrong.
Instead, sample __func__ (__FUNCTION__ in MSVC) in the OPENSSL_PUT_ERROR macro
and store it alongside file and line number. This doesn't change the format of
ERR_print_errors, however ERR_error_string_n now uses the placeholder
"OPENSSL_internal" rather than an actual function name since that only takes
the uint32_t packed error code as input.
This updates err scripts to not emit the function string table. The
OPENSSL_PUT_ERROR invocations, for now, still include the extra
parameter. That will be removed in a follow-up.
BUG=468039
Change-Id: Iaa2ef56991fb58892fa8a1283b3b8b995fbb308d
Reviewed-on: https://boringssl-review.googlesource.com/5275
Reviewed-by: Adam Langley <agl@google.com>
This removes the version field from RSA and instead handles versioning
as part of parsing. (As a bonus, we now correctly limit multi-prime RSA
to version 1 keys.)
Most consumers are also converted. old_rsa_priv_{de,en}code are left
alone for now. Those hooks are passed in parameters which match the old
d2i/i2d pattern (they're only used in d2i_PrivateKey and
i2d_PrivateKey).
Include a test which, among other things, checks that public keys being
serialized as private keys are handled properly.
BUG=499653
Change-Id: Icdd5f0382c4a84f9c8867024f29756e1a306ba08
Reviewed-on: https://boringssl-review.googlesource.com/5273
Reviewed-by: Adam Langley <agl@google.com>
This is the first structure to be implemented with the new BIGNUM ASN.1
routines. Object reuse in the legacy d2i/i2d functions is implemented by
releasing whatever was in *out before and setting it to the
newly-allocated object. As with the new d2i_SSL_SESSION, this is a
weaker form of object reuse, but should suffice for reasonable callers.
As ECDSA_SIG is more likely to be parsed alone than as part of another
structure (and using CBB is slightly tedious), add convenient functions
which take byte arrays. For consistency with SSL_SESSION, they are named
to/from_bytes. from_bytes, unlike the CBS variant, rejects trailing
data.
Note this changes some test expectations: BER signatures now push an
error code. That they didn't do this was probably a mistake.
BUG=499653
Change-Id: I9ec74db53e70d9a989412cc9e2b599be0454caec
Reviewed-on: https://boringssl-review.googlesource.com/5269
Reviewed-by: Adam Langley <agl@google.com>
RSA and ECDSA will both require being able to convert ASN.1 INTEGERs to
and from DER. Don't bother handling negative BIGNUMs for now. It doesn't
seem necessary and saves bothering with two's-complement vs
sign-and-magnitude.
BUG=499653
Change-Id: I1e80052067ed528809493af73b04f82539d564ff
Reviewed-on: https://boringssl-review.googlesource.com/5268
Reviewed-by: Adam Langley <agl@google.com>
This change also switches the behaviour of the client. Previously the
client would send the SCSV rather than the extension, but now it'll only
do that for SSLv3 connections.
Change-Id: I67a04b8abbef2234747c0dac450458deb6b0cd0a
Reviewed-on: https://boringssl-review.googlesource.com/5143
Reviewed-by: Adam Langley <agl@google.com>
Rather than four massive functions that handle every extension,
organise the code by extension with four smaller functions for each.
Change-Id: I876b31dacb05aca9884ed3ae7c48462e6ffe3b49
Reviewed-on: https://boringssl-review.googlesource.com/5142
Reviewed-by: Adam Langley <agl@google.com>
Platform crypto APIs for PKCS#1 RSA signatures vary between expecting the
caller to prepend the DigestInfo prefix (RSA_sign_raw) and prepending it
internally (RSA_sign). Currently, Chromium implements sign or sign_raw as
appropriate. To avoid needing both variants, the new asynchronous methods will
only expose the higher-level one, sign.
To satisfy ports which previously implemented sign_raw, expose the DigestInfo
prefix as a utility function.
BUG=347404
Change-Id: I04c397b5e9502b2942f6698ecf81662a3c9282e6
Reviewed-on: https://boringssl-review.googlesource.com/4940
Reviewed-by: Adam Langley <agl@google.com>
Mirrors SSL_SESSION_to_bytes. It avoids having to deal with object-reuse, the
non-size_t length parameter, and trailing data. Both it and the object-reuse
variant back onto an unexposed SSL_SESSION_parse which reads a CBS.
Note that this changes the object reuse story slightly. It's now merely an
optional output pointer that frees its old contents. No d2i_SSL_SESSION
consumer in Google that's built does reuse, much less reuse with the assumption
that the top-level object won't be overridden.
Change-Id: I5cb8522f96909bb222cab0f342423f2dd7814282
Reviewed-on: https://boringssl-review.googlesource.com/5121
Reviewed-by: Adam Langley <agl@google.com>
The client and server both have to decide on behaviour when resuming a
session where the EMS state of the session doesn't match the EMS state
as exchanged in the handshake.
Original handshake
| No Yes
------+--------------------------------------------------------------
|
R | Server: ok [1] Server: abort [3]
e No | Client: ok [2] Client: abort [4]
s |
u |
m |
e |
Yes | Server: don't resume No problem
| Client: abort; server
| shouldn't have resumed
[1] Servers want to accept legacy clients. The draft[5] says that
resumptions SHOULD be rejected so that Triple-Handshake can't be done,
but we'll rather enforce that EMS was used when using tls-unique etc.
[2] The draft[5] says that even the initial handshake should be aborted
if the server doesn't support EMS, but we need to be able to talk to the
world.
[3] This is a very weird case where a client has regressed without
flushing the session cache. Hopefully we can be strict and reject these.
[4] This can happen when a server-farm shares a session cache but
frontends are not all updated at once. If Chrome is strict here then
hopefully we can prevent any servers from existing that will try to
resume an EMS session that they don't understand. OpenSSL appears to be
ok here: https://www.ietf.org/mail-archive/web/tls/current/msg16570.html
[5] https://tools.ietf.org/html/draft-ietf-tls-session-hash-05#section-5.2
BUG=492200
Change-Id: Ie1225a3960d49117b05eefa5a36263d8e556e467
Reviewed-on: https://boringssl-review.googlesource.com/4981
Reviewed-by: Adam Langley <agl@google.com>
The SSL_PROTOCOL_METHOD table needs work, but this makes it clearer
exactly what the shared interface between the upper later and TLS/DTLS
is.
BUG=468889
Change-Id: I38931c484aa4ab3f77964d708d38bfd349fac293
Reviewed-on: https://boringssl-review.googlesource.com/4955
Reviewed-by: Adam Langley <agl@google.com>
The functions BN_rshift and BN_lshift shift their arguments to the right or
left by a specified number of bits. Unpredicatable results (including
crashes) can occur if a negative number is supplied for the shift value.
Thanks to Mateusz Kocielski (LogicalTrust), Marek Kroemeke and Filip Palian
for discovering and reporting this issue.
(Imported from upstream's 7cc18d8158b5fc2676393d99b51c30c135502107.)
Change-Id: Ib9f5e410a46df3d7f02a61374807fba209612bd3
Reviewed-on: https://boringssl-review.googlesource.com/4892
Reviewed-by: Adam Langley <agl@google.com>
This isn't exhaustive. There are still failures in some tests which probably
ought to get C++'d first.
Change-Id: Iac58df9d98cdfd94603d54374a531b2559df64c3
Reviewed-on: https://boringssl-review.googlesource.com/4795
Reviewed-by: Adam Langley <agl@google.com>
Currently far from passing and I haven't even tried with a leak checker yet.
Also bn_test is slow.
Change-Id: I4fe2783aa5f7897839ca846062ae7e4a367d2469
Reviewed-on: https://boringssl-review.googlesource.com/4794
Reviewed-by: Adam Langley <agl@google.com>
