When the CT extension is enabled but no SCTs are present, the existing
code calls "continue" which causes resizing the data byte slice to be
skipped. In fact, such extensions should be rejected.
Fixes#17958
Change-Id: Iad12da10d1ea72d04ae2e1012c28bb2636f06bcd
Reviewed-on: https://go-review.googlesource.com/33265
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The tree is inconsistent about single l vs double l in those
words in documentation, test messages, and one error value text.
$ git grep -E '[Mm]arshall(|s|er|ers|ed|ing)' | wc -l
42
$ git grep -E '[Mm]arshal(|s|er|ers|ed|ing)' | wc -l
1694
Make it consistently a single l, per earlier decisions. This means
contributors won't be confused by misleading precedence, and it helps
consistency.
Change the spelling in one error value text in newRawAttributes of
crypto/x509 package to be consistent.
This change was generated with:
perl -i -npe 's,([Mm]arshal)l(|s|er|ers|ed|ing),$1$2,' $(git grep -l -E '[Mm]arshall' | grep -v AUTHORS | grep -v CONTRIBUTORS)
Updates #12431.
Follows https://golang.org/cl/14150.
Change-Id: I85d28a2d7692862ccb02d6a09f5d18538b6049a2
Reviewed-on: https://go-review.googlesource.com/33017
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Fix spelling of "original" and "occurred" in new gofmt docs. The same
misspelling of "occurred" was also present in crypto/tls, I fixed it there as
well.
Change-Id: I67b4f1c09bd1a2eb1844207d5514f08a9f525ff9
Reviewed-on: https://go-review.googlesource.com/33138
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
CL 32871 updated the default cipher suites to use AES-GCM in
preference to ChaCha20-Poly1305 on platforms which have hardware
implementations of AES-GCM. This change makes BenchmarkThroughput
use the default cipher suites instead of the test cipher suites to
ensure that the recommended (fastest) algorithms are used.
Updates #17779.
Change-Id: Ib551223e4a00b5ea197d4d73748e1fdd8a47c32d
Reviewed-on: https://go-review.googlesource.com/32838
Run-TryBot: Michael Munday <munday@ca.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
Support for ChaCha20-Poly1305 ciphers was recently added to crypto/tls.
These ciphers are preferable in software, but they cannot beat hardware
support for AES-GCM, if present.
This change moves detection for hardware AES-GCM support into
cipher/internal/cipherhw so that it can be used from crypto/tls. Then,
when AES-GCM hardware is present, the AES-GCM cipher suites are
prioritised by default in crypto/tls. (Some servers, such as Google,
respect the client's preference between AES-GCM and ChaCha20-Poly1305.)
Fixes#17779.
Change-Id: I50de2be486f0b0b8052c4628d3e3205a1d54a646
Reviewed-on: https://go-review.googlesource.com/32871
Run-TryBot: Adam Langley <agl@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
I used the slowtests.go tool as described in
https://golang.org/cl/32684 on packages that stood out.
go test -short std drops from ~56 to ~52 seconds.
This isn't a huge win, but it was mostly an exercise.
Updates #17751
Change-Id: I9f3402e36a038d71e662d06ce2c1d52f6c4b674d
Reviewed-on: https://go-review.googlesource.com/32751
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Currently, the selection of a client certificate done internally based
on the limitations given by the server's request and the certifcates in
the Config. This means that it's not possible for an application to
control that selection based on details of the request.
This change adds a callback, GetClientCertificate, that is called by a
Client during the handshake and which allows applications to select the
best certificate at that time.
(Based on https://golang.org/cl/25570/ by Bernd Fix.)
Fixes#16626.
Change-Id: Ia4cea03235d2aa3c9fd49c99c227593c8e86ddd9
Reviewed-on: https://go-review.googlesource.com/32115
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The SignatureAndHashAlgorithm from TLS 1.2[1] is being changed to
SignatureScheme in TLS 1.3[2]. (The actual values are compatible
however.)
Since we expect to support TLS 1.3 in the future, we're already using
the name and style of SignatureScheme in the recently augmented
ClientHelloInfo. As this is public API, it seems that SignatureScheme
should have its own type and exported values, which is implemented in
this change.
[1] https://tools.ietf.org/html/rfc5246#section-7.4.1.4.1
[2] https://tools.ietf.org/html/draft-ietf-tls-tls13-18#section-4.2.3
Change-Id: I0482755d02bb9a04eaf075c012696103eb806645
Reviewed-on: https://go-review.googlesource.com/32119
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The CloseWrite method sends a close_notify alert record to the other
side of the connection. This record indicates that the sender has
finished sending on the connection. Unlike the Close method, the sender
may still read from the connection until it recieves a close_notify
record (or the underlying connection is closed). This is analogous to a
TCP half-close.
This is a rework of CL 25159 with fixes for the unstable test.
Updates #8579
Change-Id: I47608d2f82a88baff07a90fd64c280ed16a60d5e
Reviewed-on: https://go-review.googlesource.com/31318
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
VerifyPeerCertificate returns an error if the peer should not be
trusted. It will be called after the initial handshake and before
any other verification checks on the cert or chain are performed.
This provides the callee an opportunity to augment the certificate
verification.
If VerifyPeerCertificate is not nil and returns an error,
then the handshake will fail.
Fixes#16363
Change-Id: I6a22f199f0e81b6f5d5f37c54d85ab878216bb22
Reviewed-on: https://go-review.googlesource.com/26654
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Now that we have the Clone method on tls.Config, net/http doesn't need
any custom functions to do that any more.
Change-Id: Ib60707d37f1a7f9a7d7723045f83e59eceffd026
Reviewed-on: https://go-review.googlesource.com/31595
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This change enables the ChaCha20-Poly1305 cipher suites by default. This
changes the default ClientHello and thus requires updating all the
tests.
Change-Id: I6683a2647caaff4a11f9e932babb6f07912cad94
Reviewed-on: https://go-review.googlesource.com/30958
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
GetConfigForClient allows the tls.Config to be updated on a per-client
basis.
Fixes#16066.
Fixes#15707.
Fixes#15699.
Change-Id: I2c675a443d557f969441226729f98502b38901ea
Reviewed-on: https://go-review.googlesource.com/30790
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Although an AEAD, in general, can be used concurrently in both the seal
and open directions, TLS is easier. Since the transport keys are
different for different directions in TLS, an AEAD will only ever be
used in one direction. Thus we don't need separate buffers for seal and
open because they can never happen concurrently.
Also, fix the nonce size to twelve bytes since the fixed-prefix
construction for AEADs is superseded and will never be used for anything
else now.
Change-Id: Ibbf6c6b1da0e639f4ee0e3604410945dc7dcbb46
Reviewed-on: https://go-review.googlesource.com/30959
Run-TryBot: Adam Langley <agl@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This reverts commit c6185aa63217c84a1a73c578c155e7d4dec6cec8. That
commit seems to be causing flaky failures on the builders. See
discussion on the original thread: https://golang.org/cl/25159.
Change-Id: I26e72d962d4efdcee28a0bc61a53f246b046df77
Reviewed-on: https://go-review.googlesource.com/31316
Run-TryBot: Adam Langley <agl@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
This change adds support for the ChaCha20-Poly1305 AEAD to crypto/tls,
as specified in https://tools.ietf.org/html/rfc7905.
Fixes#15499.
Change-Id: Iaa689be90e03f208c40b574eca399e56f3c7ecf1
Reviewed-on: https://go-review.googlesource.com/30957
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The CloseWrite method sends a close_notify alert record to the other
side of the connection. This record indicates that the sender has
finished sending on the connection. Unlike the Close method, the sender
may still read from the connection until it recieves a close_notify
record (or the underlying connection is closed). This is analogous to a
TCP half-close.
Updates #8579
Change-Id: I9c6bc193efcb25cc187f7735ee07170afa7fdde3
Reviewed-on: https://go-review.googlesource.com/25159
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Since this changes the offered curves in the ClientHello, all the test
data needs to be updated too.
Change-Id: I227934711104349c0f0eab11d854e5a2adcbc363
Reviewed-on: https://go-review.googlesource.com/30825
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
X25519 (RFC 7748) is now commonly used for key agreement in TLS
connections, as specified in
https://tools.ietf.org/html/draft-ietf-tls-curve25519-01.
This change adds support for that in crypto/tls, but does not enabled it
by default so that there's less test noise. A future change will enable
it by default and will update all the test data at the same time.
Change-Id: I91802ecd776d73aae5c65bcb653d12e23c413ed4
Reviewed-on: https://go-review.googlesource.com/30824
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
When updating the test data against OpenSSL, the handshake can fail and
the stdout/stderr output of OpenSSL is very useful in finding out why.
However, printing that output has been broken for some time because its
no longer sent to a byte.Buffer. This change fixes that.
Change-Id: I6f846c7dc80f1ccee9fa1be36f0b579b3754e05f
Reviewed-on: https://go-review.googlesource.com/30823
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
We will need OpenSSL 1.1.0 in order to test some of the features
expected for Go 1.8. However, 1.1.0 also disables (by default) some
things that we still want to test, such as RC4, 3DES and SSLv3. Thus
developers wanting to update the crypto/tls test data will need to build
OpenSSL from source.
This change updates the test data with transcripts generated by 1.1.0
(in order to reduce future diffs) and also causes a banner to be printed
if 1.1.0 is not used when updating.
(The test for an ALPN mismatch is removed because OpenSSL now terminates
the connection with a fatal alert if no known ALPN protocols are
offered. There's no point testing against this because it's an OpenSSL
behaviour.)
Change-Id: I957516975e0b8c7def84184f65c81d0b68f1c551
Reviewed-on: https://go-review.googlesource.com/30821
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The aim is to make the decrypt() timing profile constant, irrespective of
the CBC padding length or correctness. The old algorithm, on valid padding,
would only MAC bytes up to the padding length threshold, making CBC
ciphersuites vulnerable to plaintext recovery attacks as presented in the
"Lucky Thirteen" paper.
The new algorithm Write()s to the MAC all supposed payload, performs a
constant time Sum()---which required implementing a constant time Sum() in
crypto/sha1, see the "Lucky Microseconds" paper---and then Write()s the rest
of the data. This is performed whether the padding is good or not.
This should have no explicit secret-dependent timings, but it does NOT
attempt to normalize memory accesses to prevent cache timing leaks.
Updates #13385
Change-Id: I15d91dc3cc6eefc1d44f317f72ff8feb0a9888f7
Reviewed-on: https://go-review.googlesource.com/18130
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Since there's no aspect of key logging that OpenSSL can check for us,
the tests for it might as well just connect to another goroutine as this
is lower-maintainance.
Change-Id: I746d1dbad1b4bbfc8ef6ccf136ee4824dbda021e
Reviewed-on: https://go-review.googlesource.com/30089
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Joonas Kuorilehto <joneskoo@derbian.fi>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
readRecord was not returning early if c.in.decrypt failed and ran
through the rest of the function. It does set c.in.err, so the various
checks in the callers do ultimately notice before acting on the result,
but we should avoid running the rest of the function at all.
Also rename 'err' to 'alertValue' since it isn't actually an error.
Change-Id: I6660924716a85af704bd3fe81521b34766238695
Reviewed-on: https://go-review.googlesource.com/24709
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
After renegotiation support was added (af125a5193c) it's possible for a
Write to block on a Read when racing to complete the handshake:
1. The Write determines that a handshake is needed and tries to
take the neccesary locks in the correct order.
2. The Read also determines that a handshake is needed and wins
the race to take the locks.
3. The Read goroutine completes the handshake and wins a race
to unlock and relock c.in, which it'll hold when waiting for
more network data.
If the application-level protocol requires the Write to complete before
data can be read then the system as a whole will deadlock.
Unfortunately it doesn't appear possible to reverse the locking order of
c.in and handshakeMutex because we might read a renegotiation request at
any point and need to be able to do a handshake without unlocking.
So this change adds a sync.Cond that indicates that a goroutine has
committed to doing a handshake. Other interested goroutines can wait on
that Cond when needed.
The test for this isn't great. I was able to reproduce the deadlock with
it only when building with -race. (Because -race happened to alter the
timing just enough.)
Fixes#17101.
Change-Id: I4e8757f7b82a84e46c9963a977d089f0fb675495
Reviewed-on: https://go-review.googlesource.com/29164
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Since 2a8c81ff handshake messages are not written directly to wire but
buffered. If an error happens at the wrong time the alert will be
written to the buffer but never flushed, causing an EOF on the client
instead of a more descriptive alert.
Thanks to Brendan McMillion for reporting this.
Fixes#17037
Change-Id: Ie093648aa3f754f4bc61c2e98c79962005dd6aa2
Reviewed-on: https://go-review.googlesource.com/28818
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Concurrent use of tls.Config is allowed, and may lead to
KeyLogWriter being written to concurrently. Without a mutex
to protect it, corrupted output may occur. A mutex is added
for correctness.
The mutex is made global to save size of the config struct as
KeyLogWriter is rarely enabled.
Related to #13057.
Change-Id: I5ee55b6d8b43a191ec21f06e2aaae5002a71daef
Reviewed-on: https://go-review.googlesource.com/29016
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
In Go 1.0, the Config struct consisted only of exported fields.
In Go 1.1, it started to grow private, uncopyable fields (sync.Once,
sync.Mutex, etc).
Ever since, people have been writing their own private Config.Clone
methods, or risking it and doing a language-level shallow copy and
copying the unexported sync variables.
Clean this up and export the Config.clone method as Config.Clone.
This matches the convention of Template.Clone from text/template and
html/template at least.
Fixes#15771
Updates #16228 (needs update in x/net/http2 before fixed)
Updates #16492 (not sure whether @agl wants to do more)
Change-Id: I48c2825d4fef55a75d2f99640a7079c56fce39ca
Reviewed-on: https://go-review.googlesource.com/28075
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
Add support for writing TLS client random and master secret
in NSS key log format.
https://developer.mozilla.org/en-US/docs/Mozilla/Projects/NSS/Key_Log_Format
Normally this is enabled by a developer debugging TLS based
applications, especially HTTP/2, by setting the KeyLogWriter
to an open file. The keys negotiated in handshake are then
logged and can be used to decrypt TLS sessions e.g. in Wireshark.
Applications may choose to add support similar to NSS where this
is enabled by environment variable, but no such mechanism is
built in to Go. Instead each application must explicitly enable.
Fixes#13057.
Change-Id: If6edd2d58999903e8390b1674ba4257ecc747ae1
Reviewed-on: https://go-review.googlesource.com/27434
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
If SetSessionTicketKeys was called on a fresh tls.Config, the configured
keys would be overridden with a random key by serverInit.
Fixes#15421.
Change-Id: I5d6cc81fc3e5de4dfa15eb614d102fb886150d1b
Reviewed-on: https://go-review.googlesource.com/27317
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
These were new with TLS 1.2 and, reportedly, some servers require it.
Since it's easy, this change adds suport for three flavours of
AES-128-CBC with SHA-256 MACs.
Other testdata/ files have to be updated because this changes the list
of cipher suites offered by default by the client.
Fixes#15487.
Change-Id: I1b14330c31eeda20185409a37072343552c3464f
Reviewed-on: https://go-review.googlesource.com/27315
Run-TryBot: Adam Langley <agl@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Jonathan Rudenberg <jonathan@titanous.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The RFC is clear that the Parameters in an AlgorithmIdentifer for an RSA
public key must be NULL. BoringSSL enforces this so we have strong
evidence that this is a widely compatible change.
Embarrassingly enough, the major source of violations of this is us. Go
used to get this correct in only one of two places. This was only fixed
in 2013 (with 4874bc9b). That's why lots of test certificates are
updated in this change.
Fixes#16166.
Change-Id: Ib9a4551349354c66e730d44eb8cee4ec402ea8ab
Reviewed-on: https://go-review.googlesource.com/27312
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Moves the state.ServerName assignment to outside the if
statement that checks for handshakeComplete.
Fixes#15571
Change-Id: I6c4131ddb16389aed1c410a975f9aa3b52816965
Reviewed-on: https://go-review.googlesource.com/22862
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
Give *recordingConn the correct WriteTo signature
to be an io.WriterTo. This makes vet happy.
It also means that it'll report errors,
which were previously being ignored.
Updates #11041
Change-Id: I13f171407d63f4b62427679bff362eb74faddca5
Reviewed-on: https://go-review.googlesource.com/27121
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
I believe it's necessary to use a buffer size smaller than 64KB because
(at least some versions of) Window using a TCP receive window less than
64KB. Currently the client and server use buffer sizes of 16KB and 32KB,
respectively (the server uses io.Copy, which defaults to 32KB internally).
Since the server has been using 32KB, it should be safe for the client to
do so as well.
Fixes#15899
Change-Id: I36d44b29f2a5022c03fc086213d3c1adf153e983
Reviewed-on: https://go-review.googlesource.com/24581
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This fixes some 40 warnings from go vet.
Fixes#16134.
Change-Id: Ib9fcba275fe692f027a2a07b581c8cf503b11087
Reviewed-on: https://go-review.googlesource.com/24287
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
This change causes TLS handshake messages to be buffered and written in
a single Write to the underlying net.Conn.
There are two reasons to want to do this:
Firstly, it's slightly preferable to do this in order to save sending
several, small packets over the network where a single one will do.
Secondly, since 37c28759ca46cf381a466e32168a793165d9c9e9 errors from
Write have been returned from a handshake. This means that, if a peer
closes the connection during a handshake, a “broken pipe” error may
result from tls.Conn.Handshake(). This can mask any, more detailed,
fatal alerts that the peer may have sent because a read will never
happen.
Buffering handshake messages means that the peer will not receive, and
possibly reject, any of a flow while it's still being written.
Fixes#15709
Change-Id: I38dcff1abecc06e52b2de647ea98713ce0fb9a21
Reviewed-on: https://go-review.googlesource.com/23609
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This fixes `go test go/types`.
https://golang.org/cl/23487/ introduced this code which contains
two unused variables (declared and assigned to, but never read).
cmd/compile doesn't report the error due open issue #8560 (the
variables are assigned to in a closure), but go/types does. The
build bot only runs go/types tests in -short mode (which doesn't
typecheck the std lib), hence this doesn't show up on the dashboard
either.
We cannot call b.Fatal and friends in the goroutine. Communicating
the error to the invoking function requires a channel or a mutex.
Unless the channel/sycnhronized variable is tested in each iteration
that follows, the iteration blocks if there's a failure. Testing in
each iteration may affect benchmark times.
One could use a time-out but that time depends on the underlying system.
Panicking seems good enough in this unlikely case; better than hanging
or affecting benchmark times.
Change-Id: Idce1172da8058e580fa3b3e398825b0eb4316325
Reviewed-on: https://go-review.googlesource.com/23528
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The Windows builders run the throughput benchmarks really slowly with a
64kb buffer. Lowering it to 16kb brings the performance back into line
with the other builders.
This is a work-around to get the build green until we can figure out why
the Windows builders are slow with the larger buffer size.
Update #15899
Change-Id: I215ebf115e8295295c87f3b3e22a4ef1f9e77f81
Reviewed-on: https://go-review.googlesource.com/23574
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The current code, introduced after Go 1.6 to improve latency on
low-bandwidth connections, sends 1 kB packets until 1 MB has been sent,
and then sends 16 kB packets (the maximum record size).
Unfortunately this decreases throughput for 1-16 MB responses by 20% or so.
Following discussion on #15713, change cutoff to 128 kB sent
and also grow the size allowed for successive packets:
1 kB, 2 kB, 3 kB, ..., 15 kB, 16 kB.
This fixes the throughput problems: the overhead is now closer to 2%.
I hope this still helps with latency but I don't have a great way to test it.
At the least, it's not worse than Go 1.6.
Comparing MaxPacket vs DynamicPacket benchmarks:
name maxpkt time/op dyn. time/op delta
Throughput/1MB-8 5.07ms ± 7% 5.21ms ± 7% +2.73% (p=0.023 n=16+16)
Throughput/2MB-8 15.7ms ±201% 8.4ms ± 5% ~ (p=0.604 n=20+16)
Throughput/4MB-8 14.3ms ± 1% 14.5ms ± 1% +1.53% (p=0.000 n=16+16)
Throughput/8MB-8 26.6ms ± 1% 26.8ms ± 1% +0.47% (p=0.003 n=19+18)
Throughput/16MB-8 51.0ms ± 1% 51.3ms ± 1% +0.47% (p=0.000 n=20+20)
Throughput/32MB-8 100ms ± 1% 100ms ± 1% +0.24% (p=0.033 n=20+20)
Throughput/64MB-8 197ms ± 0% 198ms ± 0% +0.56% (p=0.000 n=18+7)
The small MB runs are bimodal in both cases, probably GC pauses.
But there's clearly no general slowdown anymore.
Fixes#15713.
Change-Id: I5fc44680ba71812d24baac142bceee0e23f2e382
Reviewed-on: https://go-review.googlesource.com/23487
Reviewed-by: Ian Lance Taylor <iant@golang.org>
