Russ Cox [Tue, 7 Feb 2017 16:59:38 +0000 (11:59 -0500)]
[release-branch.go1.8] crypto/x509: check for new tls-ca-bundle.pem last
We added CentOS 7's /etc/pki/ca-trust/extracted/pem/tls-ca-bundle.pem
to the list in response to #17549 - not being able to find any certs otherwise.
Now we have #18813, where CentOS 6 apparently has both that file
and /etc/pki/tls/certs/ca-bundle.crt, and the latter is complete while
the former is not.
Moving the new CentOS 7 file to the bottom of the list should fix both
problems: the CentOS 7 system that didn't have any of the other files
in the list will still find the new one, and existing systems will still
keep using what they were using instead of preferring the new path
that may or may not be complete on some systems.
Cherry Zhang [Mon, 6 Feb 2017 19:24:16 +0000 (14:24 -0500)]
[release-branch.go1.8] cmd/compile: do not use "oaslit" for global
The compiler did not emit write barrier for assigning global with
struct literal, like global = T{} where T contains pointer.
The relevant code path is:
walkexpr OAS var_ OSTRUCTLIT
oaslit
anylit OSTRUCTLIT
walkexpr OAS var_ nil
return without adding write barrier
return true
break (without adding write barrier)
This CL makes oaslit not apply to globals. See also CL
https://go-review.googlesource.com/c/36355/ for an alternative
fix.
The downside of this is that it generates static data for zeroing
struct now. Also this only covers global. If there is any lurking
bug with implicit zeroing other than globals, this doesn't fix.
Robert Griesemer [Tue, 7 Feb 2017 06:01:07 +0000 (22:01 -0800)]
[release-branch.go1.8] cmd/compile/internal/syntax: avoid follow-up error for incorrect if statement
This is a follow-up on https://go-review.googlesource.com/36470
and leads to a more stable fix. The above CL relied on filtering
of multiple errors on the same line to avoid more than one error
for an `if` statement of the form `if a := 10 {}`. This CL avoids
the secondary error ("missing condition in if statement") in the
first place.
Michael Munday [Fri, 3 Feb 2017 21:47:32 +0000 (16:47 -0500)]
[release-branch.go1.8] cmd/compile: fix type propagation through s390x SSA rules
This CL fixes two issues:
1. Load ops were initially always lowered to unsigned loads, even
for signed types. This was fine by itself however LoadReg ops
(used to re-load spilled values) were lowered to signed loads
for signed types. This meant that spills could invalidate
optimizations that assumed the original unsigned load.
2. Types were not always being maintained correctly through rules
designed to eliminate unnecessary zero and sign extensions.
Updates #18906 and fixes #18958 (backport of CL 36256 to 1.8).
Russ Cox [Mon, 30 Jan 2017 21:55:52 +0000 (16:55 -0500)]
[release-branch.go1.8] all: final merge of master into Go 1.8 release branch
After this, we will merge some of the dev work like
type aliases and inlining into master, so any additional
changes for the Go 1.8 release will need to be cherry-picked,
not merged.
3e55059f cmd/dist: really skip the testsanitizers tests on Android 09496599 runtime: add explicit (void) in C to avoid GCC 7 problem 4cffe2b6 cmd/dist: use the target GOOS to skip the test for issue 18153 6bdb0c11 doc: update go1.8 release notes after TxOptions change 09096bd3 cmd/go: update alldocs after CL 35150 96ea0918 cmd/compile: use CMPWU for 32-bit or smaller unsigned Geq on ppc64{,le} 21a8db1c doc: document go1.7.5
Ian Lance Taylor [Mon, 30 Jan 2017 17:43:25 +0000 (09:43 -0800)]
runtime: add explicit (void) in C to avoid GCC 7 problem
This avoids errors like
./traceback.go:80:2: call of non-function C.f1
I filed https://gcc.gnu.org/PR79289 for the GCC problem. I think this
is a bug in GCC, and it may be fixed before the final GCC 7 release.
This CL is correct either way.
Fixes #18855.
Change-Id: I0785a7b7c5b1d0ca87b454b5eca9079f390fcbd4
Reviewed-on: https://go-review.googlesource.com/35919
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Crawshaw <crawshaw@golang.org>
Chris Broadfoot [Thu, 26 Jan 2017 17:24:20 +0000 (09:24 -0800)]
[release-branch.go1.8] all: merge master into release-branch.go1.8
78860b2ad2 cmd/go: don't reject ./... matching top-level file outside GOPATH 2b283cedef database/sql: fix race when canceling queries immediately 1cf08182f9 go/printer: fix format with leading comments in composite literal b531eb3062 runtime: reorder modules so main.main comes first 165cfbc409 database/sql: let tests wait for db pool to come to expected state ea73649343 doc: update gccgo docs 1db16711f5 doc: clarify what to do with Go 1.4 when installing from source 3717b429f2 doc: note that plugins are not fully baked 98842cabb6 net/http: don't send body on redirects for 301, 302, 303 when GetBody is set 314180e7f6 net/http: fix a nit aad06da2b9 cmd/link: mark DWARF function symbols as reachable be9dcfec29 doc: mention testing.MainStart signature change a96e117a58 runtime: amd64, use 4-byte ops for memmove of 4 bytes 4cce27a3fa cmd/compile: fix constant propagation through s390x MOVDNE instructions 1be957d703 misc/cgo/test: pass current environment to syscall.Exec ec654e2251 misc/cgo/test: fix test when using GCC 7 256a605faa cmd/compile: don't use nilcheck information until the next block e8d5989ed1 cmd/compile: fix compilebench -alloc ea7d9e6a52 runtime: check for nil g and m in msanread
Robert Griesemer [Wed, 25 Jan 2017 23:05:39 +0000 (15:05 -0800)]
go/printer: fix format with leading comments in composite literal
This fix is less pervasive than it seems. The only change affecting
formatting is on printer.go:760. The remaining changes have no effect
on formatting since the value of p.level is ignored except on this
specific line.
The remaining changes are:
- renamed adjBlock to funcBody since that's how it is used
- introduced new printer field 'level' tracking the composite
literal nesting level
- update/restore the composite literal nesting level as needed
David Crawshaw [Wed, 25 Jan 2017 04:19:36 +0000 (20:19 -0800)]
runtime: reorder modules so main.main comes first
Modules appear in the moduledata linked list in the order they are
loaded by the dynamic loader, with one exception: the
firstmoduledata itself the module that contains the runtime.
This is not always the first module (when using -buildmode=shared,
it is typically libstd.so, the second module).
The order matters for typelinksinit, so we swap the first module
with whatever module contains the main function.
Updates #18729
This fixes the test case extracted with -linkshared, and now
go test -linkshared encoding/...
passes. However the original issue about a plugin failure is not
yet fixed.
Change-Id: I9f399ecc3518e22e6b0a350358e90b0baa44ac96
Reviewed-on: https://go-review.googlesource.com/35644
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Daniel Theophanes [Wed, 25 Jan 2017 16:27:45 +0000 (08:27 -0800)]
database/sql: let tests wait for db pool to come to expected state
Slower builders were failing TestQueryContext because the cancel
and return to conn pool happens async. TestQueryContext already
uses a wait method for this reason. Use the same method for
other context tests.
Brad Fitzpatrick [Tue, 24 Jan 2017 17:52:54 +0000 (17:52 +0000)]
net/http: don't send body on redirects for 301, 302, 303 when GetBody is set
The presence of Request.GetBody being set on a request was causing all
redirected requests to have a body, even if the redirect status didn't
warrant one.
This bug came from 307/308 support (https://golang.org/cl/29852) which
removed the line that set req.Body to nil after POST/PUT redirects.
Keith Randall [Mon, 23 Jan 2017 16:22:10 +0000 (08:22 -0800)]
runtime: amd64, use 4-byte ops for memmove of 4 bytes
memmove used to use 2 2-byte load/store pairs to move 4 bytes.
When the result is loaded with a single 4-byte load, it caused
a store to load fowarding stall. To avoid the stall,
special case memmove to use 4 byte ops for the 4 byte copy case.
We already have a special case for 8-byte copies.
386 already specializes 4-byte copies.
I'll do 2-byte copies also, but not for 1.8.
benchmark old ns/op new ns/op delta
BenchmarkIssue18740-8 7567 4799 -36.58%
3-byte copies get a bit slower. Other copies are unchanged.
name old time/op new time/op delta
Memmove/3-8 4.76ns ± 5% 5.26ns ± 3% +10.50% (p=0.000 n=10+10)
Fixes #18740
Change-Id: Iec82cbac0ecfee80fa3c8fc83828f9a1819c3c74
Reviewed-on: https://go-review.googlesource.com/35567
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Michael Munday [Sat, 21 Jan 2017 03:03:13 +0000 (22:03 -0500)]
cmd/compile: fix constant propagation through s390x MOVDNE instructions
The constant propagation rules selected the wrong operand to
propagate. So MOVDNE (move if not equal) propagated operands as if
it were a MOVDEQ (move if equal).
Keith Randall [Fri, 20 Jan 2017 17:54:10 +0000 (09:54 -0800)]
cmd/compile: don't use nilcheck information until the next block
When nilcheck runs, the values in a block are not in any particular
order. So any facts derived from examining the blocks shouldn't be
used until we reach the next block.
This is suboptimal as it won't eliminate nil checks within a block.
But it's probably a better fix for now as it is a much smaller change
than other strategies for fixing this bug.
nilptr3.go changes are mostly because for this pattern:
_ = *p
_ = *p
either nil check is fine to keep, and this CL changes which one
the compiler tends to keep.
There are a few regressions from code like this:
_ = *p
f()
_ = *p
For this pattern, after this CL we issue 2 nil checks instead of one.
(For the curious, this happens because intra-block nil check
elimination now falls to CSE, not nilcheck proper. The former
pattern has two nil checks with the same store argument. The latter
pattern has two nil checks with different store arguments.)
pprof.WriteHeapProfile is shorthand for
pprof.Lookup("heap").WriteTo(f, 0).
The second parameter is debug.
If it is non-zero, pprof writes legacy-format
pprof output, which compilebench can parse.
Chris Broadfoot [Thu, 19 Jan 2017 20:36:22 +0000 (12:36 -0800)]
[release-branch.go1.8] all: merge master into release-branch.go1.8
6593d8650d go/ast: fix Object's doc comment about Data c1730ae424 runtime: force workers out before checking mark roots d10eddcba3 testing: make parallel t.Run safe again 2c8b70eacf crypto/x509: revert SystemCertPool implementation for Windows fcfd91858b doc/go1.8: document Plan 9 requirements 81a61a96c9 runtime: for plugins, don't add duplicate itabs f674537cc9 README.md: update and simplify d8711919db cmd/go: fix bug help message 48d8edb5b2 crypto/tls: disable CBC cipher suites with SHA-256 by default 92ecd78933 cmd/compile: add ZeroWB case in writebarrier 787125abab doc: 2017 is the Year of the Gopher 5b708a6b6a cmd/compile: lvalues are only required for == when calling runtime fns e83d506714 vendor/golang_org/x/crypto/poly1305: revendor to pick up fix for #18673 76f981c8d8 net/http: skip TestServerHijackGetsBackgroundByte on Plan 9 e395e3246a net/http: skip TestServerHijackGetsBackgroundByte_big on Plan 9 6a3c6c0de8 net/http: add another hijack-after-background-read test 467109bf56 all: test adjustments for the iOS builder b2a3b54b95 net/http: make sure Hijack's bufio.Reader includes pre-read background byte 593ea3b360 cmd/go, misc: rework cwd handling for iOS tests 0642b8a2f1 syscall: export Fsid.X__val on s390x 4601eae6ba doc/gdb: mention GOTRACEBACK=crash 4c4c5fc7a3 misc/cgo/testplugin: test that types and itabs are unique 22689c4450 reflect: keep makeFuncImpl live across makeFuncStub 9cf06ed6cd cmd/link: only exclude C-only symbols on darwin 9c3630f578 compress/flate: avoid large stack growth in fillDeflate 4f0aac52d9 cmd/go: add comment about SIGUSR2 on iOS 333f764df3 cmd/go, misc: switch from breakpoint to SIGUSR2 39e31d5ec0 doc/go1.8: update timezone database version 08da8201ca misc/cgo/testshared: test that types and itabs are unique fdde7ba2a2 runtime: avoid clobbering C callee-save register in cgoSigtramp f65abf6ddc cmd/compile: hide testdclstack behind debug flag 641ef2a733 compress/gzip: skip TestGZIPFilesHaveZeroMTimes on non-builders 0724aa813f crypto/dsa: gofmt ac05542985 net/http: deflake TestRetryIdempotentRequestsOnError b842c9aac7 doc: remove inline styles
Austin Clements [Wed, 18 Jan 2017 02:58:10 +0000 (21:58 -0500)]
runtime: force workers out before checking mark roots
Currently we check that all roots are marked as soon as gcMarkDone
decides to transition from mark 1 to mark 2. However, issue #16083
indicates that there may be a race where we try to complete mark 1
while a worker is still scanning a stack, causing the root mark check
to fail.
We don't yet understand this race, but as a simple mitigation, move
the root check to after gcMarkDone performs a ragged barrier, which
will force any remaining workers to finish their current job.
Updates #16083. This may "fix" it, but it would be better to
understand and fix the underlying race.
Keith Randall [Wed, 11 Jan 2017 23:14:06 +0000 (15:14 -0800)]
runtime: for plugins, don't add duplicate itabs
We already do this for shared libraries. Do it for plugins also.
Suggestions on how to test this would be welcome.
I'd like to get this in for 1.8. It could lead to mysterious
hangs when using plugins.
Fixes #18676
Change-Id: I03209b096149090b9ba171c834c5e59087ed0f92
Reviewed-on: https://go-review.googlesource.com/35117 Reviewed-by: David Crawshaw <crawshaw@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
Filippo Valsorda [Mon, 16 Jan 2017 23:54:45 +0000 (23:54 +0000)]
crypto/tls: disable CBC cipher suites with SHA-256 by default
As is, they were fully vulnerable to the Lucky13 attack. The SHA1
variants implement limited countermeasures (see f28cf8346c4) but the
SHA256 ones are apparently used rarely enough (see 8741504888b) that
it's not worth the extra code.
Instead, disable them by default and update the warning.
Updates #13385
Updates #15487
Change-Id: I45b8b716001e2fa0811b17e25be76e2512e5abb2
Reviewed-on: https://go-review.googlesource.com/35290 Reviewed-by: Adam Langley <alangley@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Matt Layher <mdlayher@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Cherry Zhang [Fri, 13 Jan 2017 20:42:50 +0000 (15:42 -0500)]
cmd/compile: add ZeroWB case in writebarrier
It looks like it should be there, although I couldn't find a test
case that fails without it. ZeroWB is probably never generated now:
zeroing an initialized heap object is done by making an autotmp on
stack, zeroing it, and copying (typedmemmove) to heap.
Brad Fitzpatrick [Mon, 16 Jan 2017 01:48:42 +0000 (01:48 +0000)]
doc: 2017 is the Year of the Gopher
Change-Id: Iac713ae1f322f893c92b3fc47fe9b5719052f9eb
Reviewed-on: https://go-review.googlesource.com/35240 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: David Symonds <dsymonds@golang.org> Reviewed-by: Minux Ma <minux@golang.org>
David Crawshaw [Sat, 14 Jan 2017 03:08:27 +0000 (22:08 -0500)]
all: test adjustments for the iOS builder
The working directory is now adjusted to match the typical Go test
working directory in main, as the old trick for adjusting earlier
stopped working with the latest version of LLDB bugs.
That means the small number of places where testdata files are
read before main is called no longer work. This CL adjusts those
reads to happen after main is called. (This has the bonus effect of
not reading some benchmark testdata files in all.bash.)
Fixes compress/bzip2, go/doc, go/parser, os, and time package
tests on the iOS builder.
Brad Fitzpatrick [Fri, 13 Jan 2017 21:43:56 +0000 (21:43 +0000)]
net/http: make sure Hijack's bufio.Reader includes pre-read background byte
Previously, if the Hijack called stopped the background read call
which read a byte, that byte was sitting in memory, buffered, ready to
be Read by Hijack's returned bufio.Reader, but it wasn't yet in the
bufio.Reader's buffer itself, so bufio.Reader.Buffered() reported 1
byte fewer.
This matters for callers who wanted to stitch together any buffered
data (with bufio.Reader.Peek(bufio.Reader.Buffered())) with Hijack's
returned net.Conn. Otherwise there was no way for callers to know a
byte was read.
David Crawshaw [Fri, 13 Jan 2017 12:18:40 +0000 (07:18 -0500)]
cmd/go, misc: rework cwd handling for iOS tests
Another change in behvaior (bug) in LLDB. Despite the fact that
LLDB can dump the symtab of our test binaries and show the function
addresses, it can no longer call the functions. This means the chdir
trick on signal is failing.
This CL uses a new trick. For iOS, the exec script passes the change
in directory as an argument, and it is processed early by the test
harness generated by cmd/go.
For the iOS builders.
Change-Id: I8f5d0f831fe18de99f097761f89c5184d5bf2afb
Reviewed-on: https://go-review.googlesource.com/35152 Reviewed-by: Elias Naur <elias.naur@gmail.com>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Michael Munday [Fri, 30 Sep 2016 15:12:01 +0000 (11:12 -0400)]
syscall: export Fsid.X__val on s390x
mkpost.go replaces all variables prefixed with 'X_' with '_' on s390x
because most of them do not need to be exposed. X__val is being used
by a third party library so it turns out we do need to expose it on
s390x (it is already exposed on all other Linux architectures).
Alberto Donizetti [Mon, 9 Jan 2017 18:11:58 +0000 (19:11 +0100)]
doc/gdb: mention GOTRACEBACK=crash
Also fix a couple of other errors.
Fixes #6877
Change-Id: I94c81c5847cc7b0adab19418e71687bc2ee7fe94
Reviewed-on: https://go-review.googlesource.com/34960 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Austin Clements [Thu, 12 Jan 2017 21:54:42 +0000 (16:54 -0500)]
reflect: keep makeFuncImpl live across makeFuncStub
When traceback sees reflect.makeFuncStub (or reflect.methodValueCall)
on the stack, it expects to be able to get the *reflect.makeFuncImpl
(or *reflect.methodValue) for that call from the first outgoing
argument slot of makeFuncStub/methodValueCall.
However, currently this object isn't necessarily kept live across
makeFuncStub. This means it may get garbage collected while in a
reflect call and reused for something else. If we then try to
traceback, the runtime will see a corrupted makeFuncImpl object and
panic. This was not a problem in previous releases because we always
kept arguments live across the whole function. This became a problem
when we stopped doing this.
Fix this by using reflect.KeepAlive to keep the
makeFuncImpl/methodValue live across all of callReflect/callMethod,
which in turn keeps it live as long as makeFuncStub/methodValueCall
are on the stack.
David Crawshaw [Thu, 12 Jan 2017 22:35:53 +0000 (17:35 -0500)]
cmd/link: only exclude C-only symbols on darwin
C-only symbols are excluded from pclntab because of a quirk of darwin,
where functions are referred to by an exported symbol so dynamic
relocations de-duplicate to the host binary module and break unwinding.
This doesn't happen on ELF systems because the linker always refers to
unexported module-local symbols, so we don't need this condition.
And the current logic for excluding some functions breaks the module
verification code in moduledataverify1. So disable this for plugins
on linux.
(In 1.9, it will probably be necessary to introduce a module-local
symbol reference system on darwin to fix a different bug, so all of
this onlycsymbol code made be short-lived.)
With this CL, the tests in CL 35116 pass.
Change-Id: I517d7ca4427241fa0a91276c462827efb9383be9
Reviewed-on: https://go-review.googlesource.com/35190 Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Joe Tsai [Thu, 12 Jan 2017 07:03:15 +0000 (23:03 -0800)]
compress/flate: avoid large stack growth in fillDeflate
Ranging over an array causes the array to be copied over to the
stack, which cause large re-growths. Instead, we should iterate
over slices of the array.
Also, assigning a large struct literal uses the stack even
though the actual fields being populated are small in comparison
to the entirety of the struct (see #18636).
Fixing the stack growth does not alter CPU-time performance much
since the stack-growth and copying was such a tiny portion of the
compression work:
David Crawshaw [Sun, 8 Jan 2017 01:41:00 +0000 (17:41 -0800)]
cmd/go, misc: switch from breakpoint to SIGUSR2
The iOS test harness has set a breakpoint early in the life of Go
programs so that it can change the current working directory using
information only available from the host debugger. Somewhere in the
upgrade to iOS 10 / XCode 8.2, breakpoints stopped working. This
may be an LLDB bug, or a bug in the ios-deploy LLDB scripts, it's
not clear.
Work around the problem by giving up on breakpoints. Instead, early
in the life of every test binary built for iOS, send (and ignore) a
SIGUSR2 signal. The debugger will catch this, giving the script
go_darwin_arm_exec a chance to change the working directory.
For the iOS builders.
Change-Id: I7476531985217d0c76bc176904c48379210576c2
Reviewed-on: https://go-review.googlesource.com/34926 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Shenghou Ma [Thu, 12 Jan 2017 03:48:34 +0000 (22:48 -0500)]
doc/go1.8: update timezone database version
Fixes #18623.
Change-Id: Ic965f5f7088c3270adbca7162226be486d1b9b4e
Reviewed-on: https://go-review.googlesource.com/35130 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Bryan C. Mills [Wed, 11 Jan 2017 22:39:41 +0000 (17:39 -0500)]
runtime: avoid clobbering C callee-save register in cgoSigtramp
Use R11 (a caller-saved temp register) instead of RBX (a callee-saved
register).
I believe this only affects linux/amd64, since it is the only platform
with a non-trivial cgoSigtramp implementation.
Updates #18328.
Change-Id: I3d35c4512624184d5a8ece653fa09ddf50e079a2
Reviewed-on: https://go-review.googlesource.com/35068 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
David Chase [Tue, 10 Jan 2017 13:29:34 +0000 (08:29 -0500)]
cmd/compile: disable flaky test
The test is inherently racy and vulnerable to starvation,
and within all.bash on some platforms that means it flakes.
Test is kept because it can be useful standalone to verify
behavior of GOEXPERIMENT=preeemptibleloops, and there is
likely to be further development of this feature in the
future.
There's also some question as to why it is flaking, because
though technically this is permitted, it's very odd in this
simple case.
David Chase [Thu, 10 Nov 2016 21:03:47 +0000 (16:03 -0500)]
cmd/compile: insert scheduling checks on loop backedges
Loop breaking with a counter. Benchmarked (see comments),
eyeball checked for sanity on popular loops. This code
ought to handle loops in general, and properly inserts phi
functions in cases where the earlier version might not have.
Includes test, plus modifications to test/run.go to deal with
timeout and killing looping test. Tests broken by the addition
of extra code (branch frequency and live vars) for added
checks turn the check insertion off.
If GOEXPERIMENT=preemptibleloops, the compiler inserts reschedule
checks on every backedge of every reducible loop. Alternately,
specifying GO_GCFLAGS=-d=ssa/insert_resched_checks/on will
enable it for a single compilation, but because the core Go
libraries contain some loops that may run long, this is less
likely to have the desired effect.
This is intended as a tool to help in the study and diagnosis
of GC and other latency problems, now that goal STW GC latency
is on the order of 100 microseconds or less.
Updates #17831.
Updates #10958.
Change-Id: I6206c163a5b0248e3f21eb4fc65f73a179e1f639
Reviewed-on: https://go-review.googlesource.com/33910
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Joe Tsai [Mon, 9 Jan 2017 09:00:36 +0000 (01:00 -0800)]
net/http: preserve original HTTP method when possible
In Go1.7, a 301, 302, or 303 redirect on a HEAD method, would still
cause the following redirects to still use a HEAD.
In CL/29852 this behavior was changed such that those codes always
caused a redirect with the GET method. Fix this such that both
GET and HEAD will preserve the method.
shawnps [Sat, 7 Jan 2017 16:23:11 +0000 (08:23 -0800)]
all: fix misspellings
Change-Id: I429637ca91f7db4144f17621de851a548dc1ce76
Reviewed-on: https://go-review.googlesource.com/34923 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Matthew Dempsky [Fri, 6 Jan 2017 20:25:32 +0000 (12:25 -0800)]
net: disable RFC 6724 Rule 9 for IPv4 addresses
Rule 9 arguably doesn't make sense for IPv4 addresses, and so far it
has only caused problems (#13283, #18518). Disable it until we hear
from users that actually want/need it.
Russ Cox [Fri, 6 Jan 2017 05:54:24 +0000 (00:54 -0500)]
runtime: fix corruption crash/race between select and stack growth
To implement the blocking of a select, a goroutine builds a list of
offers to communicate (pseudo-g's, aka sudog), one for each case,
queues them on the corresponding channels, and waits for another
goroutine to complete one of those cases and wake it up. Obviously it
is not OK for two other goroutines to complete multiple cases and both
wake the goroutine blocked in select. To make sure that only one
branch of the select is chosen, all the sudogs contain a pointer to a
shared (single) 'done uint32', which is atomically cas'ed by any
interested goroutines. The goroutine that wins the cas race gets to
wake up the select. A complication is that 'done uint32' is stored on
the stack of the goroutine running the select, and that stack can move
during the select due to stack growth or stack shrinking.
The relevant ordering to block and unblock in select is:
1. Lock all channels.
2. Create list of sudogs and queue sudogs on all channels.
3. Switch to system stack, mark goroutine as asleep,
unlock all channels.
4. Sleep until woken.
5. Wake up on goroutine stack.
6. Lock all channels.
7. Dequeue sudogs from all channels.
8. Free list of sudogs.
9. Unlock all channels.
There are two kinds of stack moves: stack growth and stack shrinking.
Stack growth happens while the original goroutine is running.
Stack shrinking happens asynchronously, during garbage collection.
While a channel listing a sudog is locked by select in this process,
no other goroutine can attempt to complete communication on that
channel, because that other goroutine doesn't hold the lock and can't
find the sudog. If the stack moves while all the channel locks are
held or when the sudogs are not yet or no longer queued in the
channels, no problem, because no goroutine can get to the sudogs and
therefore to selectdone. We only need to worry about the stack (and
'done uint32') moving with the sudogs queued in unlocked channels.
Stack shrinking can happen any time the goroutine is stopped.
That code already acquires all the channel locks before doing the
stack move, so it avoids this problem.
Stack growth can happen essentially any time the original goroutine is
running on its own stack (not the system stack). In the first half of
the select, all the channels are locked before any sudogs are queued,
and the channels are not unlocked until the goroutine has stopped
executing on its own stack and is asleep, so that part is OK. In the
second half of the select, the goroutine wakes up on its own goroutine
stack and immediately locks all channels. But the actual call to lock
might grow the stack, before acquiring any locks. In that case, the
stack is moving with the sudogs queued in unlocked channels. Not good.
One goroutine has already won a cas on the old stack (that goroutine
woke up the selecting goroutine, moving it out of step 4), and the
fact that done = 1 now should prevent any other goroutines from
completing any other select cases. During the stack move, however,
sudog.selectdone is moved from pointing to the old done variable on
the old stack to a new memory location on the new stack. Another
goroutine might observe the moved pointer before the new memory
location has been initialized. If the new memory word happens to be
zero, that goroutine might win a cas on the new location, thinking it
can now complete the select (again). It will then complete a second
communication (reading from or writing to the goroutine stack
incorrectly) and then attempt to wake up the selecting goroutine,
which is already awake.
The scribbling over the goroutine stack unexpectedly is already bad,
but likely to go unnoticed, at least immediately. As for the second
wakeup, there are a variety of ways it might play out.
* The goroutine might not be asleep.
That will produce a runtime crash (throw) like in #17007:
runtime: gp: gp=0xc0422dcb60, goid=2299, gp->atomicstatus=8
runtime: g: g=0xa5cfe0, goid=0, g->atomicstatus=0
fatal error: bad g->status in ready
Here, atomicstatus=8 is copystack; the second, incorrect wakeup is
observing that the selecting goroutine is in state "Gcopystack"
instead of "Gwaiting".
* The goroutine might be sleeping in a send on a nil chan.
If it wakes up, it will crash with 'fatal error: unreachable'.
* The goroutine might be sleeping in a send on a non-nil chan.
If it wakes up, it will crash with 'fatal error: chansend:
spurious wakeup'.
* The goroutine might be sleeping in a receive on a nil chan.
If it wakes up, it will crash with 'fatal error: unreachable'.
* The goroutine might be sleeping in a receive on a non-nil chan.
If it wakes up, it will silently (incorrectly!) continue as if it
received a zero value from a closed channel, leaving a sudog queued on
the channel pointing at that zero vaue on the goroutine's stack; that
space will be reused as the goroutine executes, and when some other
goroutine finally completes the receive, it will do a stray write into
the goroutine's stack memory, which may cause problems. Then it will
attempt the real wakeup of the goroutine, leading recursively to any
of the cases in this list.
* The goroutine might have been running a select in a finalizer
(I hope not!) and might now be sleeping waiting for more things to
finalize. If it wakes up, as long as it goes back to sleep quickly
(before the real GC code tries to wake it), the spurious wakeup does
no harm (but the stack was still scribbled on).
* The goroutine might be sleeping in gcParkAssist.
If it wakes up, that will let the goroutine continue executing a bit
earlier than we would have liked. Eventually the GC will attempt the
real wakeup of the goroutine, leading recursively to any of the cases
in this list.
* The goroutine cannot be sleeping in bgsweep, because the background
sweepers never use select.
* The goroutine might be sleeping in netpollblock.
If it wakes up, it will crash with 'fatal error: netpollblock:
corrupted state'.
* The goroutine might be sleeping in main as another thread crashes.
If it wakes up, it will exit(0) instead of letting the other thread
crash with a non-zero exit status.
* The goroutine cannot be sleeping in forcegchelper,
because forcegchelper never uses select.
* The goroutine might be sleeping in an empty select - select {}.
If it wakes up, it will return to the next line in the program!
* The goroutine might be sleeping in a non-empty select (again).
In this case, it will wake up spuriously, with gp.param == nil (no
reason for wakeup), but that was fortuitously overloaded for handling
wakeup due to a closing channel and the way it is handled is to rerun
the select, which (accidentally) handles the spurious wakeup
correctly:
if cas == nil {
// This can happen if we were woken up by a close().
// TODO: figure that out explicitly so we don't need this loop.
goto loop
}
Before looping, it will dequeue all the sudogs on all the channels
involved, so that no other goroutine will attempt to wake it.
Since the goroutine was blocked in select before, being blocked in
select again when the spurious wakeup arrives may be quite likely.
In this case, the spurious wakeup does no harm (but the stack was
still scribbled on).
* The goroutine might be sleeping in semacquire (mutex slow path).
If it wakes up, that is taken as a signal to try for the semaphore
again, not a signal that the semaphore is now held, but the next
iteration around the loop will queue the sudog a second time, causing
a cycle in the wakeup list for the given address. If that sudog is the
only one in the list, when it is eventually dequeued, it will
(due to the precise way the code is written) leave the sudog on the
queue inactive with the sudog broken. But the sudog will also be in
the free list, and that will eventually cause confusion.
* The goroutine might be sleeping in notifyListWait, for sync.Cond.
If it wakes up, (*Cond).Wait returns. The docs say "Unlike in other
systems, Wait cannot return unless awoken by Broadcast or Signal,"
so the spurious wakeup is incorrect behavior, but most callers do not
depend on that fact. Eventually the condition will happen, attempting
the real wakeup of the goroutine and leading recursively to any of the
cases in this list.
* The goroutine might be sleeping in timeSleep aka time.Sleep.
If it wakes up, it will continue running, leaving a timer ticking.
When that time bomb goes off, it will try to ready the goroutine
again, leading to any one of the cases in this list.
* The goroutine cannot be sleeping in timerproc,
because timerproc never uses select.
* The goroutine might be sleeping in ReadTrace.
If it wakes up, it will print 'runtime: spurious wakeup of trace
reader' and return nil. All future calls to ReadTrace will print
'runtime: ReadTrace called from multiple goroutines simultaneously'.
Eventually, when trace data is available, a true wakeup will be
attempted, leading to any one of the cases in this list.
None of these fatal errors appear in any of the trybot or dashboard
logs. The 'bad g->status in ready' that happens if the goroutine is
running (the most likely scenario anyway) has happened once on the
dashboard and eight times in trybot logs. Of the eight, five were
atomicstatus=8 during net/http tests, so almost certainly this bug.
The other three were atomicstatus=2, all near code in select,
but in a draft CL by Dmitry that was rewriting select and may or may
not have had its own bugs.
This bug has existed since Go 1.4. Until then the select code was
implemented in C, 'done uint32' was a C stack variable 'uint32 done',
and C stacks never moved. I believe it has become more common recently
because of Brad's work to run more and more tests in net/http in
parallel, which lengthens race windows.
The fix is to run step 6 on the system stack,
avoiding possibility of stack growth.
Fixes #17007 and possibly other mysterious failures.
Austin Clements [Fri, 23 Dec 2016 00:30:23 +0000 (17:30 -0700)]
runtime: update big mgc.go comment
The comment describing the overall GC algorithm at the top of mgc.go
has gotten woefully out-of-date (and was possibly never
correct/complete). Update it to reflect the current workings of the
GC and the set of phases that we now divide it into.
Change-Id: I02143c0ebefe9d4cd7753349dab8045f0973bf95
Reviewed-on: https://go-review.googlesource.com/34711 Reviewed-by: Rick Hudson <rlh@golang.org>
Russ Cox [Fri, 6 Jan 2017 06:17:24 +0000 (01:17 -0500)]
net/http: better failure in TestTransportPersistConnLeak
If one of the c.Get(ts.URL) results in an error, the child goroutine
calls t.Errorf, but the test goroutine gets stuck waiting for <-gotReqCh,
so the test hangs and the program is eventually killed (after 10 minutes!).
Whatever might have been printed to t.Errorf is never seen.
Adjust test so that the test fails cleanly in this case.
Still trying to debug why c.Get might fail.
It seems to have something to do with occasional connection
failures on macOS Sierra.
Austin Clements [Fri, 6 Jan 2017 14:44:41 +0000 (09:44 -0500)]
runtime: use 4K as the boundary of legal pointers
Currently, the check for legal pointers in stack copying uses
_PageSize (8K) as the minimum legal pointer. By default, Linux won't
let you map under 64K, but
1) it's less clear what other OSes allow or will allow in the future;
2) while mapping the first page is a terrible idea, mapping anywhere
above that is arguably more justifiable;
3) the compiler only assumes the first physical page (4K) is never
mapped.
Make the runtime consistent with the compiler and more robust by
changing the bad pointer check to use 4K as the minimum legal pointer.
This came out of discussions on CLs 34663 and 34719.