Also IWYU the headers.
BUG=79
Change-Id: Iafee0444c9979496166885db6ba5009cb597cb4d
Reviewed-on: https://boringssl-review.googlesource.com/11225
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
This allows the fuzzer to discover server-side resumption paths by
simply supplying what we'd like the ticket to decrypt to in the clear.
We also have a natural way to get transcripts out of runner. We record
the runner-side transcripts, so all resumption handshakes will replay
the shim-created unencrypted tickets.
BUG=104
Change-Id: Icf9cbf4af520077d38e2c8c2766b6f8bfa3c9ab5
Reviewed-on: https://boringssl-review.googlesource.com/11224
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Both the C and Go code were sampling the real clock. With this, two
successive iterations of runner transcripts give the same output.
Change-Id: I4d9e219e863881bf518c5ac199dce938a49cdfaa
Reviewed-on: https://boringssl-review.googlesource.com/11222
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
We want to ensure -fuzzer passes tests, except for the tests it
intentionally fails on. This ensures that we don't lose our ability to
refresh the fuzzer transcripts.
Change-Id: I761856c30379a3934fd46a24627ef8415b136f93
Reviewed-on: https://boringssl-review.googlesource.com/11221
Reviewed-by: Adam Langley <agl@google.com>
Apparently we never wrote one of those. Also send a decrypt_error alert
to be consistent with all the other signature checks.
Change-Id: Ib5624d098d1e3086245192cdce92f5df26005064
Reviewed-on: https://boringssl-review.googlesource.com/11180
Reviewed-by: David Benjamin <davidben@google.com>
SSL_peek works fine for us, but OpenSSL 1.1.0 regressed this
(https://github.com/openssl/openssl/issues/1563), and we don't have
tests either. Fix this.
SSL_peek can handle all weird events that SSL_read can, so use runner
and tell bssl_shim to do a SSL_peek + SSL_peek + SSL_read instead of
SSL_read. Then add tests for all the events we may discover.
Change-Id: I9e8635e3ca19653a02a883f220ab1332d4412f98
Reviewed-on: https://boringssl-review.googlesource.com/11090
Reviewed-by: Adam Langley <agl@google.com>
Upstream makes 0 mean "min/max supported version". Match that behavior,
although call it "default" instead. It shouldn't get you TLS 1.3 until
we're ready to turn it on everywhere.
BUG=90
Change-Id: I9f122fceb701b7d4de2ff70afbc1ffdf370cb97e
Reviewed-on: https://boringssl-review.googlesource.com/11181
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
We'd previously been assuming we'd want to predict P-256 and X25519 but,
on reflection, that's nonsense. Although, today, P-256 is widespread and
X25519 is less so, that's not the right question to ask. Those servers
are all 1.2.
The right question is whether we believe enough servers will get to TLS
1.3 before X25519 to justify wasting 64 bytes on all other connections.
Given that OpenSSL has already shipped X25519 and Microsoft was doing
interop testing on X25519 around when we were shipping it, I think the
answer is no.
Moreover, if we are wrong, it will be easier to go from predicting one
group to two rather than the inverse (provided we send a fake one with
GREASE). I anticipate prediction-miss HelloRetryRequest logic across the
TLS/TCP ecosystem will be largely untested (no one wants to pay an RTT),
so taking a group out of the predicted set will likely be a risky
operation.
Only predicting one group also makes things a bit simpler. I haven't
done this here, but we'll be able to fold the 1.2 and 1.3 ecdh_ctx's
together, even.
Change-Id: Ie7e42d3105aca48eb9d97e2e05a16c5379aa66a3
Reviewed-on: https://boringssl-review.googlesource.com/10960
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
The old numbers violate a MUST-level requirement in TLS 1.2 to not
advertise anonymous (0x0700 ends in 0x00). The spec has been updated
with new allocations which avoid these.
BUG=webrtc:6342
Change-Id: Ia5663ada98fa1ebf0f8a7f50fe74a0e9206c4194
Reviewed-on: https://boringssl-review.googlesource.com/11131
Reviewed-by: Adam Langley <agl@google.com>
Corpus recorded from runner and merged into existing corpus with
libFuzzer's -merge flag.
BUG=79
Change-Id: I986a50976ffef141b63e31de3a81fdb4ed5c1348
Reviewed-on: https://boringssl-review.googlesource.com/11130
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
Found by libFuzzer and then one more mistake caught by valgrind. Add a
test for this case.
Change-Id: I92773bc1231bafe5fc069e8568d93ac0df4c8acb
Reviewed-on: https://boringssl-review.googlesource.com/11129
Reviewed-by: David Benjamin <davidben@google.com>
Upstream added these functions after we did but decided to change the
names slightly. I'm not sure why they wanted to add the "proto" in
there, but align with them nonetheless so the ecosystem only has one set
of these functions.
BUG=90
Change-Id: Ia9863c58c9734374092051f02952b112806040cc
Reviewed-on: https://boringssl-review.googlesource.com/11123
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This is in preparation for using the supported_versions extension to
experiment with draft TLS 1.3 versions, since we don't wish to restore
the fallback. With versions begin opaque values, we will want
version_from_wire to reject unknown values, not attempt to preserve
order in some way.
This means ClientHello.version processing needs to be separate code.
That's just written out fully in negotiate_version now. It also means
SSL_set_{min,max}_version will notice invalid inputs which aligns us
better with upstream's versions of those APIs.
This CL doesn't replace ssl->version with an internal-representation
version, though follow work should do it once a couple of changes land
in consumers.
BUG=90
Change-Id: Id2f5e1fa72847c823ee7f082e9e69f55e51ce9da
Reviewed-on: https://boringssl-review.googlesource.com/11122
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This will make it a little easier to store the normalized version rather
than the wire version. Also document the V2ClientHello behavior.
Change-Id: I5ce9ccce44ca48be2e60ddf293c0fab6bba1356e
Reviewed-on: https://boringssl-review.googlesource.com/11121
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Passing --quiet makes valgrind only print out errors, so we don't need
to suppress things. Combine that with checking valgrind's dedicated exit
code so we notice errors that happen before the "---DONE---" marker.
This makes that marker unnecessary for valgrind. all_tests.go was not
sensitive to this, but still would do well to have valgrind be silent.
Change-Id: I841edf7de87081137e38990e647e989fd7567295
Reviewed-on: https://boringssl-review.googlesource.com/11128
Reviewed-by: Adam Langley <agl@google.com>
If the test failed due to non-ASan reasons but ASan also had errors,
output those too.
Change-Id: Id908fe2a823c59255c6a9585dfaa894a4fcd9f59
Reviewed-on: https://boringssl-review.googlesource.com/11127
Reviewed-by: Adam Langley <agl@google.com>
Runner needs to implement fuzzer mode as well so we can record
transcripts from it. A bunch of tests were failing:
- C and Go disagreed on what fuzzer mode did to TLS 1.3 padding. So we
fuzz more code, align Go with C. Fuzzer mode TLS 1.3 still pads but
just skips the final AEAD.
- The deterministic RNG should be applied per test, not per exchange. It
turns out, if your RNG is deterministic, one tends to pick the same
session ID over and over which confuses clients. (Resumption is
signaled by echoing the session ID.)
Now the only failing tests are the ones one would expect to fail.
BUG=79
Change-Id: Ica23881a6e726adae71e6767730519214ebcd62a
Reviewed-on: https://boringssl-review.googlesource.com/11126
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
wpa_supplicant in AOSP has now been updated, so these all can go. We're
just left with the AES keywrap business.
Change-Id: Ie4c3e08902a2a1f9b43e1907116c7d85791ad5e9
Reviewed-on: https://boringssl-review.googlesource.com/11160
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
If we see garbage in ClientHello.version and then select static RSA,
that garbage is what goes in the premaster.
Change-Id: I65190a44439745e6b5ffaf7669f063da725c8097
Reviewed-on: https://boringssl-review.googlesource.com/11092
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Plain PSK omits the ServerKeyExchange when there is no hint and includes
it otherwise (it should have always sent it), while other PSK ciphers
like ECDHE_PSK cannot omit the hint. Having different capabilities here
is odd and RFC 4279 5.2 suggests that all PSK ciphers are capable of
"[not] provid[ing] an identity hint".
Interpret this to mean no identity hint and empty identity hint are the
same state. Annoyingly, this gives a plain PSK implementation two
options for spelling an empty hint. The spec isn't clear and this is not
really a battle worth fighting, so I've left both acceptable and added a
test for this case.
See also https://android-review.googlesource.com/c/275217/. This is also
consistent with Android's PskKeyManager API, our only consumer anyway.
https://developer.android.com/reference/android/net/PskKeyManager.html
Change-Id: I8a8e6cc1f7dd1b8b202cdaf3d4f151bebfb4a25b
Reviewed-on: https://boringssl-review.googlesource.com/11087
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
One less field to reset on renego and save a pointer of post-handshake
memory.
Change-Id: Ifc0c3c73072af244ee3848d9a798988d2c8a7c38
Reviewed-on: https://boringssl-review.googlesource.com/11086
Reviewed-by: Adam Langley <agl@google.com>
The MinGW setup on Android already defines this stat macro.
Change-Id: Ia8e89195c06ec01d4b5a2fa7357fb8d2d500aa06
Reviewed-on: https://boringssl-review.googlesource.com/11124
Reviewed-by: Kenny Root <kroot@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Android uses MinGW for some host tools on Windows. That toolchain
doesn't support the #pragma tricks we use for thread-local destructors,
but does appear to support pthreads.
This also lets us remove the INIT_ONCE workaround, although that's
removable anyway since Android's MinGW is now new enough.
Change-Id: I8d1573923fdaac880a50d84acbebbf87461c50d2
Reviewed-on: https://boringssl-review.googlesource.com/11125
Reviewed-by: David Benjamin <davidben@google.com>
Reviewed-by: Kenny Root <kroot@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
On 64-bit systems the SSL structure is 1/16th padding. This change
reorders some fields and changes one to a bitfield in order to reduce
the memory usage a little.
Change-Id: Id7626a44d22652254717d544bdc2e08f1b0d705f
Reviewed-on: https://boringssl-review.googlesource.com/11140
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This isn't hugely important since the hs object will actually be
released at the end of the handshake, but no sense in holding on to them
longer than needed.
Also release |public_key| when we no longer need it and document what
the fields mean.
Change-Id: If677cb4a915c75405dabe7135205630527afd8bc
Reviewed-on: https://boringssl-review.googlesource.com/10360
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
C99 decided that, like PRI* macros, UINT64_C and friends should be
conditioned on __STDC_CONSTANT_MACROS in C++. C++11 then decided this
was ridiculous and overruled this decision. However, Android's headers
in older NDKs mistakenly followed the C99 rules for C++, so work around
this.
This fixes the android_arm bots.
Change-Id: I3b49e8dfc20190ebfa78876909bd0dccd3e210ea
Reviewed-on: https://boringssl-review.googlesource.com/11089
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Android currently implements this manually (see NativeBN_putULongInt) by
reaching into BIGNUM's internals. BN_ULONG is a somewhat unfortunate API
anyway as the size is platform-dependent, so add a platform-independent
way to do this.
The other things Android needs are going to need more work, but this
one's easy.
BUG=97
Change-Id: I4af4dc29f9845bdce0f0663c379b4b5d3e1dc46e
Reviewed-on: https://boringssl-review.googlesource.com/11088
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Code acting generically on an EVP_AEAD_CTX may wish to get at the
underlying EVP_AEAD.
Change-Id: I9cc905522ba76402bda4c255aa1488158323b02c
Reviewed-on: https://boringssl-review.googlesource.com/11085
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Matt Braithwaite <mab@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
clang's integrated assembler does not appear to support debug symbols
yet. (And somehow CMake has another statefulness bug that causes this to
only manifest on a second run, so the try bots aren't noticing.)
Also don't clobber the existing CMAKE_ASM_FLAGS value.
Change-Id: Ia04345f0dc2f9833ac13912d761cfab3479c10cd
Reviewed-on: https://boringssl-review.googlesource.com/11083
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
access_denied is only used to indicate client cert errors and Chrome
maps it to ERR_SSL_BAD_CLIENT_AUTH_CERT accordingly:
access_denied
A valid certificate was received, but when access control was
applied, the sender decided not to proceed with negotiation. This
message is always fatal.
We don't appear to be the cause of Chrome's recent
ERR_SSL_BAD_CLIENT_AUTH_CERT spike, but we should send these correctly
nonetheless.
If the early callback fails, handshake_failure seems the most
appropriate ("I was unable to find suitable parameters"). There isn't
really an alert that matches DoS, but internal_error seems okay?
internal_error
An internal error unrelated to the peer or the correctness of the
protocol (such as a memory allocation failure) makes it impossible
to continue. This message is always fatal.
There's nothing wrong, per se, with your ClientHello, but I just can't
deal with it right now. Please go away.
Change-Id: Icd1c998c09dc42daa4b309c1a4a0f136b85eb69d
Reviewed-on: https://boringssl-review.googlesource.com/11084
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
Reviewed-by: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Conscrypt uses these types. Note that BORINGSSL_MAKE_STACK_DELETER
requires DECLARE_STACK_OF to work. Otherwise the compiler gives some
really confusing error.
Change-Id: I8d194067ea6450937e4a8fcb4acbbf98a2550bce
Reviewed-on: https://boringssl-review.googlesource.com/11082
Reviewed-by: Kenny Root <kroot@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This withdraws support for -DBORINGSSL_ENABLE_RC4_TLS, and removes the
RC4 AEADs.
Change-Id: I1321b76bfe047d180743fa46d1b81c5d70c64e81
Reviewed-on: https://boringssl-review.googlesource.com/10940
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Conscrypt would like to write a CTS test that the callback isn't set
unexpectedly.
Change-Id: I11f987422daf0544e90f5cff4d7aaf557ac1f5a2
Reviewed-on: https://boringssl-review.googlesource.com/11060
Reviewed-by: Kenny Root <kroot@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
A METADATA file provides more structured information about included
third-party code.
Change-Id: I58b285ac2d73139e5b3843ba33d4d923b676afb7
Reviewed-on: https://boringssl-review.googlesource.com/11042
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Since gcm_test now contains variable decls in for loops it needs
-std=c11. However, tests are compiled with C++ test_support files in
Bazel, which doesn't work with -std=c11.
Change-Id: Ife18c2d80b01448bb3b7ee2728412289bf749bd9
Reviewed-on: https://boringssl-review.googlesource.com/11041
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This function (actually a macro in OpenSSL) is used by several projects
(e.g. OpenResty, OpenVPN, ...) so it can useuful to provide it for
compatibility.
However, depending on the semantics of the BIO type (e.g. BIO_pair), the
return value can be meaningless, which might explain why it was removed.
Change-Id: I0e432c92222c267eb994d32b0bc28e999c4b40a7
Reviewed-on: https://boringssl-review.googlesource.com/11020
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
The stuff around i being reused for |len| rounded to a number of blocks
is a little weird.
Change-Id: I6f07a82fe84d077062e5b34ce75cc68250be8a4a
Reviewed-on: https://boringssl-review.googlesource.com/10802
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
I'm not sure what happened here. These are both the same as
MissingKeyShare-Client.
Change-Id: I6601ed378d8639c1b59034f1e96c09a683bb62ca
Reviewed-on: https://boringssl-review.googlesource.com/11007
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
I hear our character set includes such novel symbols as '+'.
Change-Id: I96591a563317e71299748a948d68a849e15b5d60
Reviewed-on: https://boringssl-review.googlesource.com/11009
Commit-Queue: David Benjamin <davidben@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Adam Langley <agl@google.com>
Reason for revert: Right now in TLS 1.3, certificate_auth is exactly
the same as whether we're doing resumption. With the weird reauth
stuff punted to later in the spec, having extra state is just more
room for bugs to creep in.
Original issue's description:
> Determining certificate_auth and key_exchange based on SSL.
>
> This allows us to switch TLS 1.3 to use non-cipher based negotiation
> without needing to use separate functions between 1.3 and below.
>
> BUG=77
>
> Change-Id: I9207e7a6793cb69e8300e2c15afe3548cbf82af2
> Reviewed-on: https://boringssl-review.googlesource.com/10803
> Reviewed-by: David Benjamin <davidben@google.com>
> Commit-Queue: David Benjamin <davidben@google.com>
> CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
>
Change-Id: I240e3ee959ffd1f2481a06eabece3af554d20ffa
Reviewed-on: https://boringssl-review.googlesource.com/11008
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
It's easy to forget to check those. Unfortunately, it's also easy to
forget to check inner structures, which is going to be harder to stress,
but do these to start with. In doing, so fix up and unify some
error-handling, and add a missing check when parsing TLS 1.2
CertificateRequest.
This was also inspired by the recent IETF posting.
Change-Id: I27fe3cd3506258389a75d486036388400f0a33ba
Reviewed-on: https://boringssl-review.googlesource.com/10963
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This will let us use the same test scenarios for testing messages with
trailing garbage or skipped messages.
Change-Id: I9f177983e8dabb6c94d3d8443d224b79a58f40b1
Reviewed-on: https://boringssl-review.googlesource.com/10962
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
We build with them for C files, so include it for the assembly files
too.
Change-Id: I1c595e18314b245330a6f047d0a97a031d55b115
Reviewed-on: https://boringssl-review.googlesource.com/9150
Reviewed-by: Steven Valdez <svaldez@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
This was done just by grepping for 'size_t i;' and 'size_t j;'. I left
everything in crypto/x509 and friends alone.
There's some instances in gcm.c that are non-trivial and pulled into a
separate CL for ease of review.
Change-Id: I6515804e3097f7e90855f1e7610868ee87117223
Reviewed-on: https://boringssl-review.googlesource.com/10801
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Really the only thing we should be doing with these ciphers is hastening
their demise, but it was the weekend and this seemed like fun.
EVP_tls_cbc_copy_mac needs to rotate a buffer by a secret amount. (It
extracts the MAC, but rotated.) We have two codepaths for this. If
CBC_MAC_ROTATE_IN_PLACE is defined (always on), we make some assumptions
abuot cache lines, play games with volatile, and hope that doesn't leak
anything. Otherwise, we do O(N^2) work to constant-time select the
rotation incidences.
But we can do O(N lg N). Rotate by powers of two and constant-time
select by the offset's bit positions. (Handwaivy lower-bound: an array
position has N possible values, so, armed with only a constant-time
select, we need O(lg N) work to resolve it. There's N array positions,
so O(N lg N).)
A microbenchmark of EVP_tls_cbc_copy_mac shows this is 27% faster than
the old one, but still 32% slower than the in-place version.
in-place:
Did 15724000 CopyFromMAC operations in 20000744us (786170.8 ops/sec)
N^2:
Did 8443000 CopyFromMAC operations in 20001582us (422116.6 ops/sec)
N lg N:
Did 10718000 CopyFromMAC operations in 20000763us (535879.6 ops/sec)
This results in the following the CBC ciphers. I measured
AES-128-CBC-SHA1 and AES-256-CBC-SHA384 which are, respectively, the
cipher where the other bits are the fastest and the cipher where N is
largest.
in-place:
Did 2634000 AES-128-CBC-SHA1 (16 bytes) open operations in 10000739us (263380.5 ops/sec): 4.2 MB/s
Did 1424000 AES-128-CBC-SHA1 (1350 bytes) open operations in 10002782us (142360.4 ops/sec): 192.2 MB/s
Did 531000 AES-128-CBC-SHA1 (8192 bytes) open operations in 10002460us (53086.9 ops/sec): 434.9 MB/s
N^2:
Did 2529000 AES-128-CBC-SHA1 (16 bytes) open operations in 10001474us (252862.7 ops/sec): 4.0 MB/s
Did 1392000 AES-128-CBC-SHA1 (1350 bytes) open operations in 10006659us (139107.4 ops/sec): 187.8 MB/s
Did 528000 AES-128-CBC-SHA1 (8192 bytes) open operations in 10001276us (52793.3 ops/sec): 432.5 MB/s
N lg N:
Did 2531000 AES-128-CBC-SHA1 (16 bytes) open operations in 10003057us (253022.7 ops/sec): 4.0 MB/s
Did 1390000 AES-128-CBC-SHA1 (1350 bytes) open operations in 10003287us (138954.3 ops/sec): 187.6 MB/s
Did 531000 AES-128-CBC-SHA1 (8192 bytes) open operations in 10002448us (53087.0 ops/sec): 434.9 MB/s
in-place:
Did 1249000 AES-256-CBC-SHA384 (16 bytes) open operations in 10001767us (124877.9 ops/sec): 2.0 MB/s
Did 879000 AES-256-CBC-SHA384 (1350 bytes) open operations in 10009244us (87818.8 ops/sec): 118.6 MB/s
Did 344000 AES-256-CBC-SHA384 (8192 bytes) open operations in 10025897us (34311.1 ops/sec): 281.1 MB/s
N^2:
Did 1072000 AES-256-CBC-SHA384 (16 bytes) open operations in 10008090us (107113.3 ops/sec): 1.7 MB/s
Did 780000 AES-256-CBC-SHA384 (1350 bytes) open operations in 10007787us (77939.3 ops/sec): 105.2 MB/s
Did 333000 AES-256-CBC-SHA384 (8192 bytes) open operations in 10016332us (33245.7 ops/sec): 272.3 MB/s
N lg N:
Did 1168000 AES-256-CBC-SHA384 (16 bytes) open operations in 10007671us (116710.5 ops/sec): 1.9 MB/s
Did 836000 AES-256-CBC-SHA384 (1350 bytes) open operations in 10001536us (83587.2 ops/sec): 112.8 MB/s
Did 339000 AES-256-CBC-SHA384 (8192 bytes) open operations in 10018522us (33837.3 ops/sec): 277.2 MB/s
TLS CBC performance isn't as important as it was before, and the costs
aren't that high, so avoid making assumptions about cache lines. (If we
care much about CBC open performance, we probably should get the malloc
out of EVP_tls_cbc_digest_record at the end.)
Change-Id: Ib8d8271be4b09e5635062cd3b039e1e96f0d9d3d
Reviewed-on: https://boringssl-review.googlesource.com/11003
Reviewed-by: Adam Langley <agl@google.com>
Commit-Queue: Adam Langley <agl@google.com>
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
