This changes our resumption strategy. Before, we would negotiate ciphers
only on fresh handshakes. On resumption, we would blindly use whatever
was in the session.
Instead, evaluate cipher suite preferences on every handshake.
Resumption requires that the saved cipher suite match the one that would
have been negotiated anyway. If client or server preferences changed
sufficiently, we decline the session.
This is much easier to reason about (we always pick the best cipher
suite), simpler, and avoids getting stuck under old preferences if
tickets are continuously renewed. Notably, although TLS 1.2 ticket
renewal does not work in practice, TLS 1.3 will renew tickets like
there's no tomorrow.
It also means we don't need dedicated code to avoid resuming a cipher
which has since been disabled. (That dedicated code was a little odd
anyway since the mask_k, etc., checks didn't occur. When cert_cb was
skipped on resumption, one could resume without ever configuring a
certificate! So we couldn't know whether to mask off RSA or ECDSA cipher
suites.)
Add tests which assert on this new arrangement.
BUG=116
Change-Id: Id40d851ccd87e06c46c6ec272527fd8ece8abfc6
Reviewed-on: https://boringssl-review.googlesource.com/11847
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>
This is in preparation for determining the cipher suite (which, in TLS
1.2, requires the certificate be known) before resumption.
Note this has caller-visible effects:
- cert_cb is now called whether resumption occurs or not. Our only
consumer which uses this as a server is Node which will require a
patch to fix up their mucking about with SSL_get_session. (But the
patch should be quite upstreamable. More 1.1.0-compatible and
generally saner.)
- cert_cb is now called before new_session_cb and dos_protection_cb.
BUG=116
Change-Id: I6cc745757f63281fad714d4548f23880570204b0
Reviewed-on: https://boringssl-review.googlesource.com/11846
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>
This simplifies a little code around EMS and PSK KE modes, but requires
tweaking the SNI code.
The extensions that are more tightly integrated with the handshake are
still processed inline for now. It does, however, require an extra state
in 1.2 so the asynchronous session callback does not cause extensions to
be processed twice. Tweak a test enforce this.
This and a follow-up to move cert_cb before resumption are done in
preparation for resolving the cipher suite before resumption and only
resuming on match.
Note this has caller-visible effects:
- The legacy SNI callback happens before resumption.
- The ALPN callback happens before resumption.
- Custom extension ClientHello parsing callbacks also cannot depend on
resumption state.
- The DoS protection callback now runs after all the extension callbacks
as it is documented to be called after the resumption decision.
BUG=116
Change-Id: I1281a3b61789b95c370314aaed4f04c1babbc65f
Reviewed-on: https://boringssl-review.googlesource.com/11845
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>
As a client, we must tolerate this to avoid interoperability failures
with allowed server behaviors.
BUG=117
Change-Id: I9c40a2a048282e2e63ab5ee1d40773fc2eda110a
Reviewed-on: https://boringssl-review.googlesource.com/12311
Reviewed-by: David Benjamin <davidben@google.com>
Draft 18 sadly loosens the requirements to only requiring the PRF hash
stay fixed.
BUG=117
Change-Id: Ic94d53fd9cabaee611fcf36b0071558075e10728
Reviewed-on: https://boringssl-review.googlesource.com/12310
Reviewed-by: Nick Harper <nharper@chromium.org>
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 generally much cleaner and makes it possible to implement the
more lax cipher matching in draft 18.
BUG=117
Change-Id: I595d7619d60bc92e598d75b43945286323c0b72b
Reviewed-on: https://boringssl-review.googlesource.com/12309
Reviewed-by: Nick Harper <nharper@chromium.org>
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 a no-op because all affected codepaths are either unreachable or
are fine because ssl_hs_error (intentionally, since C doesn't help us
any) aligns with zero. Still, fix these.
Change-Id: Ieba4e3eec3881a56b5ddcd32abdd2c9dda875eda
Reviewed-on: https://boringssl-review.googlesource.com/12313
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Steven Valdez <svaldez@chromium.org>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
It doesn't particular matter, but AcceptAnySession should only skip the
things that would cause us to note accept a ticket. ExpectTicketAge is
an assertion, not part of protocol logic. Accordingly, fix the text.
Change-Id: I3bea9c58f4d5f912308252ec8834f183287d632f
Reviewed-on: https://boringssl-review.googlesource.com/12308
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Nick Harper <nharper@chromium.org>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
The version check should run if AcceptAnyVersion is *not* set.
Change-Id: I4c137564f91a86cb5e6a26e09fd4670cce8f1dcb
Reviewed-on: https://boringssl-review.googlesource.com/12307
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Nick Harper <nharper@chromium.org>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
When debugging a flaky test, it's useful to be able to run a given test
over and over.
Change-Id: I1a7b38792215550b242eb8238214d873d41becb6
Reviewed-on: https://boringssl-review.googlesource.com/12301
Reviewed-by: David Benjamin <davidben@google.com>
The draft 18 implementation did not compute scts_requested correctly. As
a result, it always believed SCTs were requested. Fix this and add tests
for unsolicited OCSP responses and SCTs at all versions.
Thanks to Daniel Hirche for the report.
Change-Id: Ifc59c5c4d7edba5703fa485c6c7a4055b15954b4
Reviewed-on: https://boringssl-review.googlesource.com/12305
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>
Thanks to Eric Rescorla for catching this.
Change-Id: Id0a024d7f705519cfe76d350e0ef2688dbd11a22
Reviewed-on: https://boringssl-review.googlesource.com/12303
Reviewed-by: Nick Harper <nharper@chromium.org>
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>
Having that logic in two different places is a nuisance when we go to
add new checks like resumption stuff. Along the way, this adds missing
tests for the ClientHello cipher/session consistency check. (We'll
eventually get it for free once the cipher/resumption change is
unblocked, but get this working in the meantime.)
This also fixes a bug where the session validity checks happened in the
wrong order relative to whether tickets_supported or renew_ticket was
looked at. Fix that by lifting that logic closer to the handshake.
Change-Id: I3f4b59cfe01064f9125277dc5834e62a36e64aae
Reviewed-on: https://boringssl-review.googlesource.com/12230
Reviewed-by: Adam Langley <agl@google.com>
This was removed a while ago. As of -18, the early data indication
extension is just a boolean.
Change-Id: I328b9abfafad326d4c2a3b5fe981af111f8401ad
Reviewed-on: https://boringssl-review.googlesource.com/12302
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Nick Harper <nharper@chromium.org>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
We missed that the TLS 1.3 code was inconsistent with the TLS 1.2 code.
Only on the server did we push an error code. But consistency between
client and server is probably worthwhile so, fix the 1.2 code to match
for now.
Change-Id: I17952c72048697dc66eacf0f144a66ced9cb3be8
Reviewed-on: https://boringssl-review.googlesource.com/12260
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: I0767cd4801924170ce13b8143a9586485b8f78af
Reviewed-on: https://boringssl-review.googlesource.com/12280
Reviewed-by: Steven Valdez <svaldez@chromium.org>
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: I07c4b67206440d169b314f24e1b3c1c697dda24f
Reviewed-on: https://boringssl-review.googlesource.com/12204
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>
TLS 1.3 adds a number of places with extensions blocks that don't easily
fit into our ClientHello/EncryptedExtensions callbacks. Between
HelloRetryRequest, ServerHello, draft 18 going nuts with Certificate,
and NewSessionTicket when we do 0-RTT, this passes the "abstract things
that are repeated three times" sniff test.
For now, it rejects unknown extensions, but it will probably grow an
allow_unknown parameter for NewSessionTicket.
This involves disabling some MSVC warnings, but they're invalid as of
C99 which we otherwise require. See
https://connect.microsoft.com/VisualStudio/feedback/details/1230248/remove-c99-related-warnings-or-make-them-off-by-default
Change-Id: Iea8bf8ab216270c081dd63e79aaad9ec73b3b550
Reviewed-on: https://boringssl-review.googlesource.com/12233
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>
BUG=112
Change-Id: I88ef17e32e33b091ff1e27b7950f88e1d48f9278
Reviewed-on: https://boringssl-review.googlesource.com/12239
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>
For TLS 1.3 draft 18, it will be useful to get at the full current
message and not just the body. Add a hook to expose it and replace
hash_current_message with a wrapper over it.
BUG=112
Change-Id: Ib9e00dd1b78e8b72e12409d85c80e96c5b411a8b
Reviewed-on: https://boringssl-review.googlesource.com/12238
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>
Certificate chain with intermediate taken from Chromium's tests. Though
it doesn't really matter because the runner tests don't verify
certificates.
BUG=70
Change-Id: I46fd1d4be0f371b5bfd43370b97d2c8053cfad60
Reviewed-on: https://boringssl-review.googlesource.com/12261
CQ-Verified: CQ bot account: commit-bot@chromium.org <commit-bot@chromium.org>
Reviewed-by: Steven Valdez <svaldez@chromium.org>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>
We used to enforce after the version was set, but stopped enforcing with
TLS 1.3. NSS enforces the value for encrypted records, which makes sense
and avoids the problems gating it on have_version. Add tests for this.
Change-Id: I7fb5f94ab4a22e8e3b1c14205aa934952d671727
Reviewed-on: https://boringssl-review.googlesource.com/12143
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 all of one bit, but having it on the SSL object means we need
manually to reset it on renego.
Change-Id: I989dacd430fe0fa63d76451b95f036a942aefcfe
Reviewed-on: https://boringssl-review.googlesource.com/12229
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>
dtls1_finish_message should NULL *out_msg before calling OPENSSL_free,
rather than asking ssl3_complete_message to do it. ssl3_finish_message
has no need to call it at all.
Change-Id: I22054217073690ab391cd19bf9993b1ceada41fd
Reviewed-on: https://boringssl-review.googlesource.com/12231
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>
This is to allow for PSK binders to be munged into the ClientHello as part of
draft 18.
BUG=112
Change-Id: Ic4fd3b70fa45669389b6aaf55e61d5839f296748
Reviewed-on: https://boringssl-review.googlesource.com/12228
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 have AEAD-level coverage for these, but we should also test this in
the TLS stack, and at maximum size per upstream's CVE-2016-7054.
Change-Id: I1f4ad0356e793d6a3eefdc2d55a9c7e05ea08261
Reviewed-on: https://boringssl-review.googlesource.com/12187
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>
This change renames |peer| to |x509_peer| and |cert_chain| to
|x509_chain| in |SSL_SESSION|. It also renames |x509| to |x509_leaf| and
|chain| to |x509_chain| in |CERT|. (All with an eye to maybe making
them lazily initialised in the future).
This a) catches anyone who might be accessing these members directly and
b) makes space for |CRYPTO_BUFFER|-based values to take the unprefixed
names.
Change-Id: I10573304fb7d6f1ea03f9e645f7fc0acdaf71ac2
Reviewed-on: https://boringssl-review.googlesource.com/12162
Reviewed-by: David Benjamin <davidben@google.com>
In transition to removing it altogether, set SSL_MODE_NO_AUTO_CHAIN by
default. If we find some consumer was relying on it, this will allow
them to revert locally with SSL_(CTX_)clear_mode, but hopefully this was
just unused.
BUG=42
Change-Id: Iaf70a436a3324ce02e02dfb18213b6715c034ff2
Reviewed-on: https://boringssl-review.googlesource.com/12180
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
3c6a1ea674 switched what layer handled
the DTLS version mapping but forgot to correct the HelloVerifyRequest
logic to account for this.
Thanks to Jed Davis for noticing this.
Change-Id: I94ea18fc43a7ba15dd7250bfbcf44dbb3361b3ce
Reviewed-on: https://boringssl-review.googlesource.com/11984
Reviewed-by: David Benjamin <davidben@google.com>
This is already manually released at the end of the handshake. With this
change, it can happen implicitly, and SSL3_STATE shrinks further by
another pointer.
Change-Id: I94b9f2e4df55e8f2aa0b3a8799baa3b9a34d7ac1
Reviewed-on: https://boringssl-review.googlesource.com/12121
Reviewed-by: Adam Langley <agl@google.com>
Fuzzer mode explores the handshake, but at the cost of losing coverage
on the record layer. Add a separate build flag and client/server
corpora for this mode.
Note this requires tweaks in consumers' fuzzer build definitions.
BUG=111
Change-Id: I1026dc7301645e165a761068a1daad6eedc9271e
Reviewed-on: https://boringssl-review.googlesource.com/12108
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
The distinction for full handshakes is not meaningful (the timestamp is
currently the start of the handshake), but for renewed sessions, we
currently retain the timestamp of the original issuance.
Instead, when minting or receiving tickets, adjust session->time and
session->timeout so that session->time is the ticket issuance time.
This is still not our final TLS 1.3 behavior (which will need a both
renewable and non-renewable times to honor the server ticket lifetime),
but it gets us closer and unblocks handling ticket_age_add from TLS 1.3
draft 18 and sends the correct NewSessionTicket lifetime.
This fixes the ticket lifetime hint which we emit on the server to
mirror the true ticket lifetime. It also fixes the TLS 1.3 server code
to not set the ticket lifetime hint. There is no need to waste ticket
size with it, it is no longer a "hint" in TLS 1.3, and even in the TLS
1.3 code we didn't fill it in on the server.
Change-Id: I140541f1005a24e53e1b1eaa90996d6dada1c3a1
Reviewed-on: https://boringssl-review.googlesource.com/12105
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>
If there is a malloc failure while assembling the ticket, call
CBB_cleanup. Also return -1 instead of 0; zero means EOF in the old
state machine and -1 means error. (Except enough of the stack gets it
wrong that consumers handle both, but we should fix this.)
Change-Id: I98541a9fa12772ec159f9992d1f9f53e5ca4cc5a
Reviewed-on: https://boringssl-review.googlesource.com/12104
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
There's no sense in flushing twice in one flight. This means when
writing a message is finally synchronous, we don't need the intermediate
state at all.
Change-Id: Iaca60d64917f82dce0456a8b15de4ee00f2d557b
Reviewed-on: https://boringssl-review.googlesource.com/12103
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
TLS 1.3 clarifies that a ticket lifetime of zero means the session is
unusable. We don't currently pay attention to that field (to be fixed in
later changes) but, in preparation for this, switch the >= to a >.
Change-Id: I0e67a0d97bc8def04914f121e84d3e7a2d640d2c
Reviewed-on: https://boringssl-review.googlesource.com/12102
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
These don't make sense and mean some SSL_SESSIONs serialize and
deserialize as different values. If we ever managed to create an
SSL_SESSION without a time, it would never expire because time always
gets set to time(NULL). If we ever created an SSL_SESSION with a zero
timeout, the timeout would be... three? Once we start adjusting
time/timeout to issuance time, driving timeout to zero is actually
plausible, so it should work properly.
Instead, make neither field optional. We always fill both out, so this
shouldn't have any effects. If it does, the only effect would be to
decline to resume some existing tickets which must have been so old that
we'd want them to have expired anyway.
Change-Id: Iee3620658c467dd6d96a2b695fec831721b03b5b
Reviewed-on: https://boringssl-review.googlesource.com/12101
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
The values are long, so check for negative numbers.
Change-Id: I8fc7333edbed50dc058547a4b53bc10b234071b4
Reviewed-on: https://boringssl-review.googlesource.com/12100
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
This business with |ok| is unnecessary. This function is still rather a
mess, but this is a small improvement.
Change-Id: I28fdf1a3687fe6a9d58d81a22cf2f8e7ce5b9b2c
Reviewed-on: https://boringssl-review.googlesource.com/12080
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>
A renewed session does not refresh the timeout. Add tests for this in
preparation for future changes which will revise this logic.
Specifically, TLS 1.3 draft 18's ticket_age_add logic will require some
tweaks in lifetime tracking to record when the ticket was minted. We'll
also likely wish to tweak the parameters for 1.3 to account for (a)
ECDHE-PSK means we're only worried about expiring a short-circuited
authentication rather than forward secrecy and (b) two hours is too
short for a QUIC 0-RTT replacement.
Change-Id: I0f1edd09151e7fcb5aee2742ef8600fbd7080df6
Reviewed-on: https://boringssl-review.googlesource.com/12002
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 only used in one place where we don't take advantage of it being
sorted anyway.
Change-Id: If6f0d04e975db903e8a93c57c869ea4964c0be37
Reviewed-on: https://boringssl-review.googlesource.com/12062
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
This is the last blocker within BoringSSL itself to opaquifying SSL.
(There are still blockers in consumers, of course.)
BUG=6
Change-Id: Ie3b8dcb78eeaa9aea7311406c5431a8625d60401
Reviewed-on: https://boringssl-review.googlesource.com/12061
Reviewed-by: Adam Langley <agl@google.com>
TLS 1.3 ciphers are now always enabled and come with a hard-coded
preference order.
BUG=110
Change-Id: Idd9cb0d75fb6bf2676ecdee27d88893ff974c4a3
Reviewed-on: https://boringssl-review.googlesource.com/12025
Reviewed-by: Adam Langley <agl@google.com>
HTTP/2 requires TLS 1.2 so the negotiated version should be available
during the ALPN callback.
Change-Id: Iea332808b531a6e5c917de5b8c8917c0aa7428a1
Reviewed-on: https://boringssl-review.googlesource.com/12060
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>
They will get very confused about which key they're using. Any caller
using exporters must either (a) leave renegotiation off or (b) be very
aware of when renegotiations happen anyway. (You need to somehow
coordinate with the peer about which epoch's exporter to use.)
Change-Id: I921ad01ac9bdc88f3fd0f8283757ce673a47ec75
Reviewed-on: https://boringssl-review.googlesource.com/12003
Reviewed-by: Adam Langley <agl@google.com>
The existing tests for this codepath require us to reconfigure the shim.
This will not work when TLS 1.3 cipher configuration is detached from
the old cipher language. It also doesn't hit codepaths like sessions
containing a TLS 1.3 version but TLS 1.2 cipher.
Instead, add some logic to the runner to rewrite tickets and build tests
out of that.
Change-Id: I57ac5d49c3069497ed9aaf430afc65c631014bf6
Reviewed-on: https://boringssl-review.googlesource.com/12024
Reviewed-by: Adam Langley <agl@google.com>
It is not ignored.
Change-Id: I2e607a6d6f7444838fc6fa65cd18e9aa142f139f
Reviewed-on: https://boringssl-review.googlesource.com/12023
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
HTTP/2 places requirements on the cipher suite. So that servers can
decline HTTP/2 when these requirements aren't met, defer ALPN
negotiation.
See also b/32553041.
Change-Id: Idbcf049f9c8bda06a8be52a0154fe76e84607268
Reviewed-on: https://boringssl-review.googlesource.com/11982
Reviewed-by: Adam Langley <agl@google.com>
