This removes 16k from a release-mode build of the bssl tool. Now that we've
finished the AEAD refactor, there's no use in keeping this around as a
prototype for "stateful AEADs".
Before:
Did 2264000 RC4-MD5 (16 bytes) seal operations in 1000430us (2263026.9 ops/sec): 36.2 MB/s
Did 266000 RC4-MD5 (1350 bytes) seal operations in 1000984us (265738.5 ops/sec): 358.7 MB/s
Did 50000 RC4-MD5 (8192 bytes) seal operations in 1014209us (49299.5 ops/sec): 403.9 MB/s
After:
Did 1895000 RC4-MD5 (16 bytes) seal operations in 1000239us (1894547.2 ops/sec): 30.3 MB/s
Did 199000 RC4-MD5 (1350 bytes) seal operations in 1001361us (198729.5 ops/sec): 268.3 MB/s
Did 39000 RC4-MD5 (8192 bytes) seal operations in 1014832us (38430.0 ops/sec): 314.8 MB/s
There is a non-trivial performance hit, but this cipher doesn't matter much and
the stitched mode code reaches into MD5_CTX and RC4_KEY in somewhat unfortunate
ways.
Change-Id: I9ecd28d6afb54e90ce61baecc641742af2ae6269
Reviewed-on: https://boringssl-review.googlesource.com/6752
Reviewed-by: Adam Langley <agl@google.com>
We can reuse the HMAC_CTX that stores the key. The API is kind of unfortunate
as, in principle, it should be possible to do an allocation-averse HMAC with a
shared key on multiple threads at once (EVP_AEAD_CTX is normally logically
const). At some point it may be worth rethinking those APIs somewhat. But
these "stateful AEADs" are already stateful in their EVP_CIPHER_CTX, so this is
fine.
Each cipher was run individually to minimize the effect of other ciphers doing
their mallocs. (Although the cost of a malloc is presumably going to depend a
lot on the malloc implementation and what's happened before in the process, so
take these numbers with a bucket of salt. They vary widely even with the same
arguments.)
Taking malloc out of seal/open also helps with the malloc tests. DTLS currently
cannot distinguish a malloc failure (should be fatal) from a decryption failure
(not fatal), so the malloc tests get stuck. But this doesn't completely get us
there since tls_cbc.c mallocs. This also assumes EVP_CIPHER_CTX, EVP_MD_CTX,
and HMAC_CTX are all clever about reusing their allocations when reset (which
they are).
Before:
Did 1315000 AES-128-CBC-SHA1 (16 bytes) seal operations in 1000087us (1314885.6 ops/sec): 21.0 MB/s
Did 181000 AES-128-CBC-SHA1 (1350 bytes) seal operations in 1004918us (180114.2 ops/sec): 243.2 MB/s
Did 34000 AES-128-CBC-SHA1 (8192 bytes) seal operations in 1024250us (33195.0 ops/sec): 271.9 MB/s
After:
Did 1766000 AES-128-CBC-SHA1 (16 bytes) seal operations in 1000319us (1765436.8 ops/sec): 28.2 MB/s
Did 187000 AES-128-CBC-SHA1 (1350 bytes) seal operations in 1004002us (186254.6 ops/sec): 251.4 MB/s
Did 35000 AES-128-CBC-SHA1 (8192 bytes) seal operations in 1014885us (34486.7 ops/sec): 282.5 MB/s
Before:
Did 391000 DES-EDE3-CBC-SHA1 (16 bytes) seal operations in 1000038us (390985.1 ops/sec): 6.3 MB/s
Did 16000 DES-EDE3-CBC-SHA1 (1350 bytes) seal operations in 1060226us (15091.1 ops/sec): 20.4 MB/s
Did 2827 DES-EDE3-CBC-SHA1 (8192 bytes) seal operations in 1035971us (2728.8 ops/sec): 22.4 MB/s
After:
Did 444000 DES-EDE3-CBC-SHA1 (16 bytes) seal operations in 1001814us (443196.0 ops/sec): 7.1 MB/s
Did 17000 DES-EDE3-CBC-SHA1 (1350 bytes) seal operations in 1042535us (16306.4 ops/sec): 22.0 MB/s
Did 2590 DES-EDE3-CBC-SHA1 (8192 bytes) seal operations in 1012378us (2558.3 ops/sec): 21.0 MB/s
Before:
Did 1316000 AES-256-CBC-SHA1 (16 bytes) seal operations in 1000510us (1315329.2 ops/sec): 21.0 MB/s
Did 157000 AES-256-CBC-SHA1 (1350 bytes) seal operations in 1002944us (156539.1 ops/sec): 211.3 MB/s
Did 29000 AES-256-CBC-SHA1 (8192 bytes) seal operations in 1030284us (28147.6 ops/sec): 230.6 MB/s
After:
Did 1645000 AES-256-CBC-SHA1 (16 bytes) seal operations in 1000313us (1644485.3 ops/sec): 26.3 MB/s
Did 162000 AES-256-CBC-SHA1 (1350 bytes) seal operations in 1003060us (161505.8 ops/sec): 218.0 MB/s
Did 36000 AES-256-CBC-SHA1 (8192 bytes) seal operations in 1014819us (35474.3 ops/sec): 290.6 MB/s
Before:
Did 1435000 RC4-SHA1 (16 bytes) seal operations in 1000245us (1434648.5 ops/sec): 23.0 MB/s
Did 207000 RC4-SHA1 (1350 bytes) seal operations in 1004675us (206036.8 ops/sec): 278.1 MB/s
Did 38000 RC4-SHA1 (8192 bytes) seal operations in 1022712us (37156.1 ops/sec): 304.4 MB/s
After:
Did 1853000 RC4-SHA1 (16 bytes) seal operations in 1000433us (1852198.0 ops/sec): 29.6 MB/s
Did 206000 RC4-SHA1 (1350 bytes) seal operations in 1002370us (205512.9 ops/sec): 277.4 MB/s
Did 42000 RC4-SHA1 (8192 bytes) seal operations in 1024209us (41007.3 ops/sec): 335.9 MB/s
Change-Id: I0edb89bddf146cf91a8e7a99c56b2278c8f38094
Reviewed-on: https://boringssl-review.googlesource.com/6751
Reviewed-by: Adam Langley <agl@google.com>
For completeness. In so far as we care about legacy ciphers' performance at
all, we should have the others too.
Change-Id: Idd2d93345f3af8b6ac5772a1cb3c201f84fe3197
Reviewed-on: https://boringssl-review.googlesource.com/6750
Reviewed-by: Adam Langley <agl@google.com>
Only ECDHE-based ciphers are implemented. To ease the transition, the
pre-standard cipher shares a name with the standard one. The cipher rule parser
is hacked up to match the name to both ciphers. From the perspective of the
cipher suite configuration language, there is only one cipher.
This does mean it is impossible to disable the old variant without a code
change, but this situation will be very short-lived, so this is fine.
Also take this opportunity to make the CK and TXT names align with convention.
Change-Id: Ie819819c55bce8ff58e533f1dbc8bef5af955c21
Reviewed-on: https://boringssl-review.googlesource.com/6686
Reviewed-by: Adam Langley <agl@google.com>
This will be used to test the C implementation against.
Change-Id: I2d396d27630937ea610144e381518eae76f78dab
Reviewed-on: https://boringssl-review.googlesource.com/6685
Reviewed-by: Adam Langley <agl@google.com>
In preparation for a Go implementation of the new TLS ciphers to test
against, implement the AEAD primitive.
Change-Id: I69b5b51257c3de16bdd36912ed2bc9d91ac853c8
Reviewed-on: https://boringssl-review.googlesource.com/6684
Reviewed-by: Adam Langley <agl@google.com>
Handy to test servers with misbehaving client auth.
Change-Id: I93f7b77c35e223761edade648bc03d1f97ed82fd
Reviewed-on: https://boringssl-review.googlesource.com/6614
Reviewed-by: Adam Langley <agl@google.com>
There were a couple more asm lines to turn into __asm__ when the patches got
reordered slightly.
Change-Id: I44be5caee6d09bb3db5dea4791592b12d175822c
Reviewed-on: https://boringssl-review.googlesource.com/6741
Reviewed-by: Adam Langley <agl@google.com>
In preparation for implementing the RFC 7539 variant to test against.
Change-Id: I0ce5e856906e00925ad1d849017f9e7fda087a8e
Reviewed-on: https://boringssl-review.googlesource.com/6683
Reviewed-by: Adam Langley <agl@google.com>
The consumers have all been updated, so we can move EVP_aead_chacha20_poly1305
to its final state. Unfortunately, the _rfc7539-suffixed version will need to
stick around for just a hair longer. Also the tls1.h macros, but the remaining
consumers are okay with that changing underneath them.
Change-Id: Ibbb70ec1860d6ac6a7e1d7b45e70fe692bf5ebe5
Reviewed-on: https://boringssl-review.googlesource.com/6600
Reviewed-by: Adam Langley <agl@google.com>
https://boringssl-review.googlesource.com/6101 was mismerged from *ring* and
lost some tests. Also add the corresponding tag truncation tests for the new
construction. So long as we have that feature, we should have tests for it.
(Although, do we actually need to support it?)
Change-Id: I70784cbac345e0ad11b496102856c53932b7362e
Reviewed-on: https://boringssl-review.googlesource.com/6682
Reviewed-by: Adam Langley <agl@google.com>
dh_tmp can only contain parameters, now that DHE always generates keys fresh
for each connection.
Change-Id: I56dad4cbec7e21326360d79df211031fd9734004
Reviewed-on: https://boringssl-review.googlesource.com/6702
Reviewed-by: Adam Langley <agl@google.com>
clang scan-build is annoyed it's not obvious the sizeof line matches the
pointer type. This is easy to fix and makes it be quiet.
Change-Id: Iec80d2a087f81179c88cae300f56d3f76b32b347
Reviewed-on: https://boringssl-review.googlesource.com/6701
Reviewed-by: Adam Langley <agl@google.com>
Rather than the length of the top-level CBB, which is kind of odd when ASN.1
length prefixes are not yet determined, return the number of bytes written to
the CBB so far. This can be computed without increasing the size of CBB at all.
Have offset and pending_*.
This means functions which take in a CBB as argument will not be sensitive to
whether the CBB is a top-level or child CBB. The extensions logic had to be
careful to only ever compare differences of lengths, which was awkward.
The reversal will also allow for the following pattern in the future, once
CBB_add_space is split into, say, CBB_reserve and CBB_did_write and we add a
CBB_data:
uint8_t *signature;
size_t signature_len = 0;
if (!CBB_add_asn1(out, &cert, CBB_ASN1_SEQUENCE) ||
/* Emit the TBSCertificate. */
!CBB_add_asn1(&cert, &tbs_cert, CBS_ASN1_SEQUENCE) ||
!CBB_add_tbs_cert_stuff(&tbs_cert, stuff) ||
!CBB_flush(&cert) ||
/* Feed it into md_ctx. */
!EVP_DigestSignInit(&md_ctx, NULL, EVP_sha256(), NULL, pkey) ||
!EVP_DigestSignUpdate(&md_ctx, CBB_data(&cert), CBB_len(&cert)) ||
/* Emit the signature algorithm. */
!CBB_add_asn1(&cert, &sig_alg, CBS_ASN1_SEQUENCE) ||
!CBB_add_sigalg_stuff(&sig_alg, other_stuff) ||
/* Emit the signature. */
!EVP_DigestSignFinal(&md_ctx, NULL, &signature_len) ||
!CBB_reserve(&cert, &signature, signature_len) ||
!EVP_DigestSignFinal(&md_ctx, signature, &signature_len) ||
!CBB_did_write(&cert, signature_len)) {
goto err;
}
(Were TBSCertificate not the first field, we'd still have to sample
CBB_len(&cert), but at least that's reasonable straight-forward. The
alternative would be if CBB_data and CBB_len somehow worked on
recently-invalidated CBBs, but that would go wrong once the invalidated CBB's
parent flushed and possibly shifts everything.)
And similar for signing ServerKeyExchange.
Change-Id: I7761e492ae472d7632875b5666b6088970261b14
Reviewed-on: https://boringssl-review.googlesource.com/6681
Reviewed-by: Adam Langley <agl@google.com>
I skipped a patch when landing and so 793c21e2 caused a build failure
when platform-specific versions of these macros were used.
Change-Id: I8ed6dbb92a511ef306d45087c3eb87781fdfed31
Reviewed-on: https://boringssl-review.googlesource.com/6740
Reviewed-by: Adam Langley <agl@google.com>
This is a remnant from when the headers in include/ where still
symlinks.
Change-Id: Ice27c412c0cdcc43312f5297119678091dcd5d38
Reviewed-on: https://boringssl-review.googlesource.com/6670
Reviewed-by: Adam Langley <agl@google.com>
It's only used in one file. No sense in polluting the namespace here.
Change-Id: Iaf3870a4be2d2cad950f4d080e25fe7f0d3929c7
Reviewed-on: https://boringssl-review.googlesource.com/6660
Reviewed-by: Adam Langley <agl@google.com>
Nothing ever uses the return value. It'd be better off discarding it rather
than make callers stick (void) everywhere.
Change-Id: Ia28c970a1e5a27db441e4511249589d74408849b
Reviewed-on: https://boringssl-review.googlesource.com/6653
Reviewed-by: Adam Langley <agl@google.com>
The uint32_t likely dates to them using HASH_LONG everywhere. Nothing ever
touches c->data as a uint32_t, only bytes. (Which makes sense seeing as it
stores the partial block.)
Change-Id: I634cb7f2b6306523aa663f8697b7dc92aa491320
Reviewed-on: https://boringssl-review.googlesource.com/6651
Reviewed-by: Adam Langley <agl@google.com>
I would hope any sensible compiler would recognize the rotation. (If
not, we should at least pull this into crypto/internal.h.) Confirmed
that clang at least produces the exact same instructions for
sha256_block_data_order for release + NO_ASM. This is also mostly moot
as SHA-1 and SHA-256 both have assembly versions on x86 that sidestep
most of this.
For the digests, take it out of md32_common.h since it doesn't use the
macro. md32_common.h isn't sure whether it's a multiply-included header
or not. It should be, but it has an #include guard (doesn't quite do
what you'd want) and will get HOST_c2l, etc., confused if one tries to
include it twice.
Change-Id: I1632801de6473ffd2c6557f3412521ec5d6b305c
Reviewed-on: https://boringssl-review.googlesource.com/6650
Reviewed-by: Adam Langley <agl@google.com>
Manual tweaks and then clang-formatted again.
Change-Id: I809fdb71b2135343e5c1264dd659b464780fc54a
Reviewed-on: https://boringssl-review.googlesource.com/6649
Reviewed-by: Adam Langley <agl@google.com>
We've tweaked it already and upstream's using a different indentation
style now anyway. This is the first of two commits. For verifiability,
this is the output of clang-format with no modifications.
Change-Id: Ia30f20bee0cc8046aedf9ac7106cc4630e8d93e6
Reviewed-on: https://boringssl-review.googlesource.com/6648
Reviewed-by: Adam Langley <agl@google.com>
We can slightly simplify tls1_P_hash. (Confirmed md32_common.h emits
code with the check.)
Change-Id: I0293ceaaee261a7ac775b42a639f7a9f67705456
Reviewed-on: https://boringssl-review.googlesource.com/6647
Reviewed-by: Adam Langley <agl@google.com>
It is redundant given the other state in the connection.
Change-Id: I5dc71627132659ab4316a5ea360c9ca480fb7c6c
Reviewed-on: https://boringssl-review.googlesource.com/6646
Reviewed-by: Adam Langley <agl@google.com>
These have been unused since we unified everything on EVP_AEAD. I must
have missed them when clearing out dead state. This shaves 136 bytes of
per-connection state.
Change-Id: I705f8de389fd34ab4524554ee9e4b1d6be198994
Reviewed-on: https://boringssl-review.googlesource.com/6645
Reviewed-by: Adam Langley <agl@google.com>
There's no need to track consumed bytes, so rr->data and rr->off may be
merged together.
Change-Id: I8842d005665ea8b4d4a0cced941f3373872cdac4
Reviewed-on: https://boringssl-review.googlesource.com/6644
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>
This uses ssl3_read_bytes for now. We still need to dismantle that
function and then invert the handshake state machine, but this gets
things closer to the right shape as an intermediate step and is a large
chunk in itself. It simplifies a lot of the CCS/handshake
synchronization as a lot of the invariants much more clearly follow from
the handshake itself.
Tests need to be adjusted since this changes some error codes. Now all
the CCS/Handshake checks fall through to the usual
SSL_R_UNEXPECTED_RECORD codepath. Most of what used to be a special-case
falls out naturally. (If half of Finished was in the same record as the
pre-CCS message, that part of the handshake record would have been left
unconsumed, so read_change_cipher_spec would have noticed, just like
read_app_data would have noticed.)
Change-Id: I15c7501afe523d5062f0e24a3b65f053008d87be
Reviewed-on: https://boringssl-review.googlesource.com/6642
Reviewed-by: Adam Langley <agl@google.com>
With server-side renegotiation gone, handshake_fragment's only purpose
in life is to handle a fragmented HelloRequest (we probably do need to
support those if some server does 1/n-1 record-splitting on handshake
records). The logic to route the data into
ssl3_read_bytes(SSL3_RT_HANDSHAKE) never happens, and the contents are
always a HelloRequest prefix.
This also trims a tiny bit of per-connection state.
Change-Id: Ia1b0dda5b7e79d817c28da1478640977891ebc97
Reviewed-on: https://boringssl-review.googlesource.com/6641
Reviewed-by: Adam Langley <agl@google.com>
See also upstream's 06cf881a3a10d5af3c1255c08cfd0c6ddb5f1cc3,
9f040d6decca7930e978784c917f731e5c45e8f0, and
9f6795e7d2d1e35668ad70ba0afc480062be4e2e.
Change-Id: I27d90e382867a5fe988d152b31f8494e001a6a9f
Reviewed-on: https://boringssl-review.googlesource.com/6628
Reviewed-by: Adam Langley <agl@google.com>
update.py used to be used only on Windows until very recently, but
Windows and non-Windows have been at the same clang revision for
a while now. So even a few months ago update.py and update.sh
would've contained the same clang revision.
BUG=chromium:494442
Change-Id: Ie9127a1c49e31a7810ee431f8e662350c245917c
Reviewed-on: https://boringssl-review.googlesource.com/6620
Reviewed-by: David Benjamin <davidben@chromium.org>
Reviewed-by: Adam Langley <agl@google.com>
Sometimes BadRSAClientKeyExchange-1 fails with DATA_TOO_LARGE_FOR_MODULUS if
the corruption brings the ciphertext above the RSA modulus. Ensure this does
not happen.
Change-Id: I0d8ea6887dfcab946fdf5d38f5b196f5a927c4a9
Reviewed-on: https://boringssl-review.googlesource.com/6731
Reviewed-by: Adam Langley <agl@google.com>
Avoids bouncing on the lock, but it doesn't really matter since it's all
taking read locks. If we're declaring that callbacks don't get to see
every object being created, they shouldn't see every object being
destroyed.
CRYPTO_dup_ex_data also already had this optimization, though it wasn't
documented.
BUG=391192
Change-Id: I5b8282335112bca3850a7c0168f8bd7f7d4a2d57
Reviewed-on: https://boringssl-review.googlesource.com/6626
Reviewed-by: Adam Langley <agl@google.com>
This callback is never used. The one caller I've ever seen is in Android
code which isn't built with BoringSSL and it was a no-op.
It also doesn't actually make much sense. A callback cannot reasonably
assume that it sees every, say, SSL_CTX created because the index may be
registered after the first SSL_CTX is created. Nor is there any point in
an EX_DATA consumer in one file knowing about an SSL_CTX created in
completely unrelated code.
Replace all the pointers with a typedef to int*. This will ensure code
which passes NULL or 0 continues to compile while breaking code which
passes an actual function.
This simplifies some object creation functions which now needn't worry
about CRYPTO_new_ex_data failing. (Also avoids bouncing on the lock, but
it's taking a read lock, so this doesn't really matter.)
BUG=391192
Change-Id: I02893883c6fa8693682075b7b130aa538a0a1437
Reviewed-on: https://boringssl-review.googlesource.com/6625
Reviewed-by: Adam Langley <agl@google.com>
Then deprecate the old functions. Thanks to upstream's
6977e8ee4a718a76351ba5275a9f0be4e530eab5 for the idea.
Change-Id: I916abd6fca2a3b2a439ec9902d9779707f7e41eb
Reviewed-on: https://boringssl-review.googlesource.com/6622
Reviewed-by: Adam Langley <agl@google.com>
It has no callers. I prepped for its removal earlier with
c05697c2c5
and then completely forgot.
Thanks to upstream's 6f78b9e824c053d062188578635c575017b587c5 for
the reminder. Quoth them:
> This only gets used to set a specific curve without actually checking
> that the peer supports it or not and can therefor result in handshake
> failures that can be avoided by selecting a different cipher.
It's also a very confusing API since it does NOT pass ownership of the
EC_KEY to the caller.
Change-Id: I6a00643b3a2d6746e9e0e228b47c2bc9694b0084
Reviewed-on: https://boringssl-review.googlesource.com/6621
Reviewed-by: Adam Langley <agl@google.com>
Remove the custom copy of those helpers.
Change-Id: I810c3ae8dbf7bc0654d3e9fb9900c425d36f64aa
Reviewed-on: https://boringssl-review.googlesource.com/6611
Reviewed-by: Adam Langley <agl@google.com>
Cover not just the wrong version, but also other mistakes.
Change-Id: I46f05a9a37b7e325adc19084d315a415777d3a46
Reviewed-on: https://boringssl-review.googlesource.com/6610
Reviewed-by: Adam Langley <agl@google.com>
This is a remnant of ssl3_get_client_hello's old DTLS cookie logic, which has
since been removed. (If we ever need HelloVerifyRequest support on the server,
we'll implement something stateless in front.) We can switch this to something
more straightforward now.
See also upstream's 94f98a9019e1c0a3be4ca904b2c27c7af3d937c0,
Change-Id: Ie733030209a381a4915d6744fa12a79ffe972fa5
Reviewed-on: https://boringssl-review.googlesource.com/6601
Reviewed-by: Adam Langley <agl@google.com>
I don't think we're ever going to manage to enforce this, and it doesn't
seem worth the trouble. We don't support application protocols which use
renegotiation outside of the HTTP/1.1 mid-stream client auth hack.
There, it's on the server to reject legacy renegotiations.
This removes the last of SSL_OP_ALL.
Change-Id: I996fdeaabf175b6facb4f687436549c0d3bb0042
Reviewed-on: https://boringssl-review.googlesource.com/6580
Reviewed-by: Adam Langley <agl@google.com>
RFC 5746 forbids a server from downgrading or upgrading
renegotiation_info support. Even with SSL_OP_LEGACY_SERVER_CONNECT set
(the default), we can still enforce a few things.
I do not believe this has practical consequences. The attack variant
where the server half is prefixed does not involve a renegotiation on
the client. The converse where the client sees the renegotiation and
prefix does, but we only support renego for the mid-stream HTTP/1.1
client auth hack, which doesn't do this. (And with triple-handshake,
HTTPS clients should be requiring the certificate be unchanged across
renego which makes this moot.)
Ultimately, an application which makes the mistake of using
renegotiation needs to be aware of what exactly that means and how to
handle connection state changing mid-stream. We make renego opt-in now,
so this is a tenable requirement.
(Also the legacy -> secure direction would have been caught by the
server anyway since we send a non-empty RI extension.)
Change-Id: I915965c342f8a9cf3a4b6b32f0a87a00c3df3559
Reviewed-on: https://boringssl-review.googlesource.com/6559
Reviewed-by: Adam Langley <agl@google.com>
