Skip to content

N
net
Switch branch/tag
  1. 24 Oct, 2016 4 commits
    • Tom Bergan's avatar
      http2: reject stream self-dependencies · 65dfc087
      Tom Bergan authored 
      Fix spec bug: Section 5.3.1 says that self-dependencies should be
treated as a stream error of type PROTOCOL_ERROR.

Updates golang/go#16046

Change-Id: I8b5dc8808943dc96aac0c543c7032fa989ab9e0b
Reviewed-on: https://go-review.googlesource.com/31858Reviewed-by: 's avatarBrad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      65dfc087
    • Brad Fitzpatrick's avatar
      http2: optimize server frame writes · e7b14352
      Brad Fitzpatrick authored 
      Don't do async frame writes when the write is known to be small enough
to definitely fit in the write buffer and not cause a flush (which
might block). This avoids starting a new goroutine and doing a channel
send operation for many frame writes.

name                  old time/op    new time/op    delta
ServerGets-4             146µs ± 2%     144µs ± 1%  -1.46%        (p=0.000 n=14+14)
ServerPosts-4            162µs ± 1%     160µs ± 1%  -1.15%        (p=0.000 n=15+15)
Server_GetRequest-4      145µs ± 1%     143µs ± 0%  -1.26%        (p=0.000 n=15+15)
Server_PostRequest-4     160µs ± 1%     158µs ± 1%  -1.32%        (p=0.000 n=15+15)

The headers frame is the main last one which might show some benefit
if there's a cheap enough way to conservatively calculate its size.

Change-Id: I9be2ddbf04689340819d8701ea671fff378d9e79
Reviewed-on: https://go-review.googlesource.com/31495
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarBrad Fitzpatrick <bradfitz@golang.org>
      e7b14352
    • Tom Bergan's avatar
      http2: interface to support pluggable schedulers · 4be9b97e
      Tom Bergan authored 
      This adds an interface to support pluggable schedulers. The interface
is defined in writesched.go. The only type needed by this interface is
FrameWriteRequest, which describes a request to write a frame (this used
to be called frameWriteMsg). The scheduler can be configured with a new
field in http2.Server. Two schedulers are implemented:

1) A random scheduler that is essentially equivalent to the existing
   scheduler. This is currently the default scheduler if none is
   configured. The implementation is in writesched_random.go.

2) A scheduler that uses H2 weights and priorities. The H2 priority tree
   is maintained as a tree of priorityNodes. The next frame is chosen by
   walking this tree in topological order, where sibling nodes are ordered
   by their bandwidth usage relative to their H2 weight. Two optional
   features are added to improve performance -- these are configured with
   PriorityWriteSchedulerConfig.

Fixes golang/go#16168

Change-Id: I97ec93e5c58c2efec35455ba2f3c31e849f706af
Reviewed-on: https://go-review.googlesource.com/25366Reviewed-by: 's avatarBrad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      4be9b97e
    • Nigel Tao's avatar
      publicsuffix: update table to latest list from publicsuffix.org · 5695785b
      Nigel Tao authored 
      Update the list to revision 915565885d0fbd25caf7d8b339cd3478f558da94
(2016-10-19T08:16:09Z).

Change-Id: Ie94e7237015f2d84e0a0d082bd7ff9e04b05ecd9
Reviewed-on: https://go-review.googlesource.com/31530Reviewed-by: 's avatarBrad Fitzpatrick <bradfitz@golang.org>
      5695785b
  3. 23 Oct, 2016 1 commit
  5. 22 Oct, 2016 1 commit
  7. 21 Oct, 2016 2 commits
  9. 20 Oct, 2016 2 commits
  11. 19 Oct, 2016 1 commit
    • Brad Fitzpatrick's avatar
      http2: make Server reuse 64k request body buffer between requests · 3bafa332
      Brad Fitzpatrick authored 
      name           old time/op    new time/op    delta
ServerPosts-4     192µs ± 1%     164µs ± 1%  -14.16%  (p=0.000 n=17+19)

name           old alloc/op   new alloc/op   delta
ServerPosts-4    69.8kB ± 0%     2.8kB ± 0%  -95.95%  (p=0.000 n=18+18)

name           old allocs/op  new allocs/op  delta
ServerPosts-4      42.0 ± 0%      40.0 ± 0%   -4.76%  (p=0.000 n=20+20)

This is a redo of git rev e7da8eda (golang.org/cl/20542) which had a race
and was later reverted in golang.org/cl/21160.

Updates golang/go#14960
Updates grpc/grpc-go#604

Change-Id: Ie216e45730dce4fc0c58f295bcbc669973145599
Reviewed-on: https://go-review.googlesource.com/31447
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      3bafa332
  13. 18 Oct, 2016 1 commit
  15. 13 Oct, 2016 4 commits
  17. 12 Oct, 2016 1 commit
    • Artyom Pervukhin's avatar
      websocket: limit incoming payload size · 6dba816f
      Artyom Pervukhin authored 
      Codec's Receive method calls io.ReadAll of the whole frame payload,
which can be abused by user sending large payloads in order to exhaust
server memory.

Introduce limit on received payload size defined by
Conn.MaxPayloadBytes. If payload size of the message read with
Codec.Receive exceeds limit, ErrFrameTooLarge error is returned; the
connection can still be recovered if required: the next call to Receive
would at first discard leftovers of previous oversized message before
processing the next one.

Fixes golang/go#5082.

Change-Id: Ib04acd7038474fee39a1719324daaec1c0c496b1
Reviewed-on: https://go-review.googlesource.com/23590Reviewed-by: 's avatarBrad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      6dba816f
  19. 11 Oct, 2016 1 commit
  21. 07 Oct, 2016 5 commits
    • Radu Berinde's avatar
      trace: don't allocate all events upfront · f4b625ec
      Radu Berinde authored 
      SetMaxEvent allocates storage for all events, which is very expensive
if we want to set a high value; even with small values, the
unnecessary allocation of the initial slice is measurable.

We improve the performance by having a few events preallocated as part
of the trace, and appending as necessary. We also co-locate the flags
in event which makes it smaller (by a word).

We also add a set of benchmarks; before and after amortized cost
per-event is shown:

name                    old time/op  new time/op  delta
Trace_Default_2-32      1.19µs ± 1%  0.96µs ± 0%  -18.94%
Trace_Default_10-32      579ns ± 0%   635ns ± 1%   +9.82%
Trace_Default_100-32     463ns ± 1%   466ns ± 1%     ~
Trace_Default_1000-32    451ns ± 1%   451ns ± 0%     ~
Trace_Default_10000-32   451ns ± 2%   449ns ± 1%     ~
Trace_10_2-32           1.41µs ± 1%  0.96µs ± 1%  -31.74%
Trace_10_10-32           623ns ± 1%   634ns ± 1%   +1.73%
Trace_10_100-32          469ns ± 1%   466ns ± 1%     ~
Trace_10_1000-32         452ns ± 1%   453ns ± 2%     ~
Trace_10_10000-32        451ns ± 1%   449ns ± 1%     ~
Trace_100_2-32          2.78µs ± 2%  0.97µs ± 1%  -65.28%
Trace_100_10-32          912ns ± 1%   637ns ± 1%  -30.23%
Trace_100_100-32         479ns ± 1%   541ns ± 0%  +12.77%
Trace_100_1000-32        510ns ± 0%   541ns ± 1%   +6.08%
Trace_100_10000-32       514ns ± 1%   540ns ± 0%   +5.14%
Trace_1000_2-32         17.2µs ± 1%   1.0µs ± 1%  -94.39%
Trace_1000_10-32        3.90µs ± 1%  0.64µs ± 0%  -83.68%
Trace_1000_100-32        814ns ± 1%   542ns ± 0%  -33.46%
Trace_1000_1000-32       503ns ± 1%   581ns ± 0%  +15.56%
Trace_1000_10000-32     1.28µs ± 2%  1.03µs ± 1%  -19.68%

Change-Id: I5bbc12153a9f30dff821ef139583cb1c07a11069
Reviewed-on: https://go-review.googlesource.com/30615Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      f4b625ec
    • Brad Fitzpatrick's avatar
      Revert "trace: don't allocate all events upfront" · 9b4f494e
      Brad Fitzpatrick authored 
      This reverts commit If7c68bb1809fb92fa5d06cb6640be5e09e1f131f.

Reason for revert: breaks the build for Go 1.6 or below due to use of Go 1.7+ only features.

Original change's description:
> trace: don't allocate all events upfront
> 
> SetMaxEvent allocates storage for all events, which is very expensive
> if we want to set a high value; even with small values, the
> unnecessary allocation of the initial slice is measurable.
> 
> We improve the performance by having a few events preallocated as part
> of the trace, and appending as necessary. We also co-locate the flags
> in event which makes it smaller (by a word).
> 
> We also add a set of benchmarks; before and after amortized cost
> per-event is shown:
> 
> name                 old time/op  new time/op  delta
> Trace/0-2-32         1.19µs ± 1%  0.96µs ± 0%  -18.94%  (p=0.008 n=5+5)
> Trace/0-10-32         579ns ± 0%   635ns ± 1%   +9.82%  (p=0.008 n=5+5)
> Trace/0-100-32        463ns ± 1%   466ns ± 1%     ~     (p=0.190 n=5+5)
> Trace/0-1000-32       451ns ± 1%   451ns ± 0%     ~     (p=0.984 n=5+5)
> Trace/0-10000-32      451ns ± 2%   449ns ± 1%     ~     (p=0.492 n=5+5)
> Trace/10-2-32        1.41µs ± 1%  0.96µs ± 1%  -31.74%  (p=0.008 n=5+5)
> Trace/10-10-32        623ns ± 1%   634ns ± 1%   +1.73%  (p=0.008 n=5+5)
> Trace/10-100-32       469ns ± 1%   466ns ± 1%     ~     (p=0.357 n=5+5)
> Trace/10-1000-32      452ns ± 1%   453ns ± 2%     ~     (p=0.913 n=5+5)
> Trace/10-10000-32     451ns ± 1%   449ns ± 1%     ~     (p=0.175 n=5+5)
> Trace/100-2-32       2.78µs ± 2%  0.97µs ± 1%  -65.28%  (p=0.008 n=5+5)
> Trace/100-10-32       912ns ± 1%   637ns ± 1%  -30.23%  (p=0.008 n=5+5)
> Trace/100-100-32      479ns ± 1%   541ns ± 0%  +12.77%  (p=0.008 n=5+5)
> Trace/100-1000-32     510ns ± 0%   541ns ± 1%   +6.08%  (p=0.008 n=5+5)
> Trace/100-10000-32    514ns ± 1%   540ns ± 0%   +5.14%  (p=0.008 n=5+5)
> Trace/1000-2-32      17.2µs ± 1%   1.0µs ± 1%  -94.39%  (p=0.008 n=5+5)
> Trace/1000-10-32     3.90µs ± 1%  0.64µs ± 0%  -83.68%  (p=0.008 n=5+5)
> Trace/1000-100-32     814ns ± 1%   542ns ± 0%  -33.46%  (p=0.008 n=5+5)
> Trace/1000-1000-32    503ns ± 1%   581ns ± 0%  +15.56%  (p=0.008 n=5+5)
> Trace/1000-10000-32  1.28µs ± 2%  1.03µs ± 1%  -19.68%  (p=0.008 n=5+5)
> 
> Change-Id: If7c68bb1809fb92fa5d06cb6640be5e09e1f131f
> Reviewed-on: https://go-review.googlesource.com/30374
> Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
> TryBot-Result: Gobot Gobot <gobot@golang.org>
> Reviewed-by: Ian Lance Taylor <iant@golang.org>
> 

Change-Id: I7cf25464ae836489e0af4a59a4d534a5d25ee352
Reviewed-on: https://go-review.googlesource.com/30650Reviewed-by: 's avatarBrad Fitzpatrick <bradfitz@golang.org>
      9b4f494e
    • Mikio Hara's avatar
      lif: rename internal types and constants generated by cgo · 084869ab
      Mikio Hara authored 
      To match up with other packages that work with IP protocol stack.

Change-Id: Idc3ee5b2399c4982dbd9114aac441f6d7452aeef
Reviewed-on: https://go-review.googlesource.com/30576
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      084869ab
    • Mikio Hara's avatar
      ipv6: rename internal types and constants generated by cgo · 30df40a6
      Mikio Hara authored 
      To match up with other packages that work with IP protocol stack.

Change-Id: I1f4eb76051964766113914c9698233022e73d5eb
Reviewed-on: https://go-review.googlesource.com/30575
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      30df40a6
    • Mikio Hara's avatar
      ipv4: rename internal types and constants generated by cgo · a14f429b
      Mikio Hara authored 
      To match up with other packages that work with IP protocol stack.

Change-Id: I2d20e4194617b05b6b46d6fba5df1f6c12ae2793
Reviewed-on: https://go-review.googlesource.com/30574
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      a14f429b
  23. 06 Oct, 2016 5 commits
    • Radu Berinde's avatar
      trace: don't allocate all events upfront · 88c1a61b
      Radu Berinde authored 
      SetMaxEvent allocates storage for all events, which is very expensive
if we want to set a high value; even with small values, the
unnecessary allocation of the initial slice is measurable.

We improve the performance by having a few events preallocated as part
of the trace, and appending as necessary. We also co-locate the flags
in event which makes it smaller (by a word).

We also add a set of benchmarks; before and after amortized cost
per-event is shown:

name                 old time/op  new time/op  delta
Trace/0-2-32         1.19µs ± 1%  0.96µs ± 0%  -18.94%  (p=0.008 n=5+5)
Trace/0-10-32         579ns ± 0%   635ns ± 1%   +9.82%  (p=0.008 n=5+5)
Trace/0-100-32        463ns ± 1%   466ns ± 1%     ~     (p=0.190 n=5+5)
Trace/0-1000-32       451ns ± 1%   451ns ± 0%     ~     (p=0.984 n=5+5)
Trace/0-10000-32      451ns ± 2%   449ns ± 1%     ~     (p=0.492 n=5+5)
Trace/10-2-32        1.41µs ± 1%  0.96µs ± 1%  -31.74%  (p=0.008 n=5+5)
Trace/10-10-32        623ns ± 1%   634ns ± 1%   +1.73%  (p=0.008 n=5+5)
Trace/10-100-32       469ns ± 1%   466ns ± 1%     ~     (p=0.357 n=5+5)
Trace/10-1000-32      452ns ± 1%   453ns ± 2%     ~     (p=0.913 n=5+5)
Trace/10-10000-32     451ns ± 1%   449ns ± 1%     ~     (p=0.175 n=5+5)
Trace/100-2-32       2.78µs ± 2%  0.97µs ± 1%  -65.28%  (p=0.008 n=5+5)
Trace/100-10-32       912ns ± 1%   637ns ± 1%  -30.23%  (p=0.008 n=5+5)
Trace/100-100-32      479ns ± 1%   541ns ± 0%  +12.77%  (p=0.008 n=5+5)
Trace/100-1000-32     510ns ± 0%   541ns ± 1%   +6.08%  (p=0.008 n=5+5)
Trace/100-10000-32    514ns ± 1%   540ns ± 0%   +5.14%  (p=0.008 n=5+5)
Trace/1000-2-32      17.2µs ± 1%   1.0µs ± 1%  -94.39%  (p=0.008 n=5+5)
Trace/1000-10-32     3.90µs ± 1%  0.64µs ± 0%  -83.68%  (p=0.008 n=5+5)
Trace/1000-100-32     814ns ± 1%   542ns ± 0%  -33.46%  (p=0.008 n=5+5)
Trace/1000-1000-32    503ns ± 1%   581ns ± 0%  +15.56%  (p=0.008 n=5+5)
Trace/1000-10000-32  1.28µs ± 2%  1.03µs ± 1%  -19.68%  (p=0.008 n=5+5)

Change-Id: If7c68bb1809fb92fa5d06cb6640be5e09e1f131f
Reviewed-on: https://go-review.googlesource.com/30374
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      88c1a61b
    • Mikio Hara's avatar
      ipv6: don't refer to cgo-generated union field · 2a824cf9
      Mikio Hara authored 
      The size of sockaddr_storage is fixed for interoperability on each
platform. There is no need to depend on changable cgo-generated
offset values.

Change-Id: I54b775b2a8c86d969a6da0398a335075f0cbe3df
Reviewed-on: https://go-review.googlesource.com/30178
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      2a824cf9
    • Mikio Hara's avatar
      ipv4: don't refer to cgo-generated union field · 534e1870
      Mikio Hara authored 
      The size of sockaddr_storage is fixed for interoperability on each
platform. There is no need to depend on changable cgo-generated
offset values.

Change-Id: I7469fe330d49867bbdbdfae128834140244925ec
Reviewed-on: https://go-review.googlesource.com/30177
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      534e1870
    • Mikio Hara's avatar
      ipv6: add support for solaris · a7769702
      Mikio Hara authored 
      This change adds support for Solaris.

To read and write IPv6 ancillary data using ControlMessage, ipv6 pacakge
requires https://go-review.googlesource.com/30171/

Note: Unlike other platforms, Solaris seems to have a few restrictions
on ICMP property access via raw IP sockets. At least applications are
prohibited from using IPV6_CHECKSUM option for ICMP transport.

Fixes golang/go#17324.

Change-Id: Ie014665d94ae6e4955c95d3eea88db90332e20bd
Reviewed-on: https://go-review.googlesource.com/30176
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      a7769702
    • Mikio Hara's avatar
      ipv4: add support for solaris · c8c327cf
      Mikio Hara authored 
      This change adds support for Solaris.

To read and write IPv4 ancillary data using ControlMessage, ipv4 pacakge
requires https://go-review.googlesource.com/30171/

Note: Unlike other platforms, Solaris seems to have a few restrictions
on ICMP property access via raw IP sockets. At least applications are
prohibited from using IP_RECVTTL option for ICMP transport.

Fixes golang/go#17323.

Change-Id: Icb6dfa3c8eced28d14693ddfea4601301999d735
Reviewed-on: https://go-review.googlesource.com/30175
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      c8c327cf
  25. 03 Oct, 2016 4 commits
  27. 01 Oct, 2016 3 commits
  29. 30 Sep, 2016 2 commits
    • Brad Fitzpatrick's avatar
      http2: in Server, disarm connection's ReadTimeout after first request · 69729301
      Brad Fitzpatrick authored 
      Fix a regression from Go 1.5 to Go 1.6: when we introduced automatic
HTTP/2 support in Go 1.6, we never handled Server.ReadTimeout. If a
user set ReadTimeout, the net/http package would set the read deadline
on the connection during the TLS handshake, but then the http2 package
would never disarm it, killing the likely-still-in-use connection
after the timeout period.

This CL changes it to disarm the timeout after the first request
headers, similar to net/http.Server.

Unlike net/http.Server, we don't re-arm it on each idle period,
because the definition of idle is more complicated with HTTP/2.

No tests here for now. Tests will be in the go repo once all the knobs
are in place and this is re-bundled into std, testing both http1 and
http2.

Updates golang/go#16450 (minimal fix for now)
Updates golang/go#14204 (considering a new http1+http2 IdleTimeout knob)

Change-Id: Iaa1570c118efda7dc0a65ba84cd77885699ec8fc
Reviewed-on: https://go-review.googlesource.com/30077Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
      69729301
    • Brad Fitzpatrick's avatar
      http2: add Transport support for IdleConnTimeout · a333c534
      Brad Fitzpatrick authored 
      Tests will be in the go repo's net/http package when this
package is re-bundled into std.

Updates golang/go#16808 (fixes after bundle into std)

Change-Id: Iad31dc120bc008b1e9679bf7b2b988aac9c893c8
Reviewed-on: https://go-review.googlesource.com/30075
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: 's avatarIan Lance Taylor <iant@golang.org>
      a333c534
  31. 29 Sep, 2016 3 commits