In the past, cgo generated Go code and C code. The C code was linked
into a shared library. The Go code was built into an executable that
dynamically linked against that shared library. C wrappers were
exported from the shared library, and the Go code called them.
It was all a long time ago, but in order to permit C code to call back
into Go, somebody implemented #pragma dynexport (https://golang.org/cl/661043)
to export a Go symbol into the dynamic symbol table. Then that same
person added code to cgo to recognize //export comments
(https://golang.org/cl/853042). The //export comments were implemented
by generating C code, to be compiled by GCC, that would refer to C code,
to be compiled by 6c, that would call the Go code. The GCC code would
go into a shared library. The code compiled by 6c would be in the Go
executable. The GCC code needed to refer to the 6c code, so the 6c
function was marked with #pragma dynexport. The important point here is
that #pragma dynexport was used to expose an internal detail of the
implementation of an exported function, because at the time it was
necessary.
Moving forward to today, cgo no longer generates a shared library and 6c
no longer exists. It's still true that we have a function compiled by
GCC that refers to a wrapper function now written in Go. In the normal
case today we are doing an external link, and we use a
//go:cgo_export_static function to make the Go wrapper function visible
to the C code under a known name.
The #pragma dynexport statement has become a //go:cgo_export_dynamic
comment on the Go code. That comment only takes effect when doing
internal linking. The comment tells the linker to put the symbol in the
dynamic symbol table. That still makes sense for the now unusual case
of using internal linking with a shared library.
However, all the changes to this code have carefully preserved the
property that the //go:cgo_export_dynamic comment refers to an internal
detail of the implementation of an exported function. That was
necessary a long time ago, but no longer makes sense.
This CL changes the code to put the actual C-callable function into the
dynamic symbol table. I considered dropping the comment entirely, but
it turns out that there is even a test for this, so I preserved it.
Change-Id: I66a7958e366e5974363099bfaa6ba862ca327849
Reviewed-on: https://go-review.googlesource.com/17061
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
The lack of this annotation causes Value.SetMapIndex to allocate
when it doesn't need to.
Add comments about why it's safe to do so.
Add a test to make sure we stay allocation-free.
Change-Id: I00826e0d73e317a31bdeae5c7e46bf95b0c6ae6a
Reviewed-on: https://go-review.googlesource.com/17060
Reviewed-by: David Chase <drchase@google.com>
It's intended primarily as a torture test for OS X.
Apparently Windows can't take it.
Updates fix for #12327.
Change-Id: If2af249ea8e2f55bff8f232dce06172e6fef9f49
Reviewed-on: https://go-review.googlesource.com/17073
Reviewed-by: Russ Cox <rsc@golang.org>
sysmon runs without a P. This means it can't interact with the garbage
collector, so write barriers not allowed in anything that sysmon does.
Fixes#10600.
Change-Id: I9de1283900dadee4f72e2ebfc8787123e382ae88
Reviewed-on: https://go-review.googlesource.com/17006
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
allocm is a very unusual function: it is specifically designed to
allocate in contexts where m.p is nil by temporarily taking over a P.
Since allocm is used in many contexts where it would make sense to use
nowritebarrierrec, this commit teaches the nowritebarrierrec analysis
to stop at allocm.
Updates #10600.
Change-Id: I8499629461d4fe25712d861720dfe438df7ada9b
Reviewed-on: https://go-review.googlesource.com/17005
Reviewed-by: Russ Cox <rsc@golang.org>
gentraceback is used in many contexts where write barriers are
disallowed. This currently works because the only write barrier is in
assigning frame.argmap in setArgInfo and in practice frame is always
on the stack, so this write barrier is a no-op.
However, we can easily eliminate this write barrier, which will let us
statically disallow write barriers (using go:nowritebarrierrec
annotations) in many more situations. As a bonus, this makes the code
a little more idiomatic.
Updates #10600.
Change-Id: I45ba5cece83697ff79f8537ee6e43eadf1c18c6d
Reviewed-on: https://go-review.googlesource.com/17003
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
This is a bit of a belt-and-suspenders fix.
On OS X, we now parse the Mach-O file to find the __text section,
which is arguably the more proper fix. But it's a bit worrisome to
depend on a name like __text not changing, so we also read more
of the initial file (now 32 kB, up from 8 kB) and scan that too.
Fixes#12327.
Change-Id: I3a201a3dc278d24707109bb3961c3bdd8b8a0b7b
Reviewed-on: https://go-review.googlesource.com/17038
Reviewed-by: Ian Lance Taylor <iant@golang.org>
My version of bash doesn't know what 'declare -A' means.
Change-Id: Icf6b0e60ebaea3feaa8661ec0423012f213b53e8
Reviewed-on: https://go-review.googlesource.com/17070
Reviewed-by: Ian Lance Taylor <iant@golang.org>
There was no documentation produced by "go doc cmd/asm".
Follow the style set by cmd/compile.
Fixes#13148.
Change-Id: I02e08ce2e7471f855bfafbbecee98ffdb7096995
Reviewed-on: https://go-review.googlesource.com/16997
Reviewed-by: Russ Cox <rsc@golang.org>
The assumption is that there are no nested function calls in complex expressions.
For the most part that assumption is true. It wasn't for these calls inserted during walk.
Fix that.
I looked through all the calls to mkcall in walk and these were the only cases
that emitted calls, that could be part of larger expressions (like not delete),
and that were not already handled.
Fixes#12225.
Change-Id: Iad380683fe2e054d480e7ae4e8faf1078cdd744c
Reviewed-on: https://go-review.googlesource.com/17034
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Because there are now multiple packages that compose the runtime
we need to distinguish between the case where a runtime package
is being compiled versus the case the "runtime" package is being
compiled. In golang.org/cl/14204 I mistakenly used
localpkg.Name == "runtime"
to check against the "runtime" package, but doing this would treat
a package with the path "foo.org/bar/runtime" as the runtime package.
The correct check is
myimportpath == "runtime"
.
Change-Id: If90e95cef768d91206f2df1c06e27be876722e4e
Reviewed-on: https://go-review.googlesource.com/17059
Run-TryBot: Michael Matloob <matloob@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
golang.org/cl/16383 broke android/386 because by a sort of confluence of hacks
no TLS relocations were emitted at all when Flag_shared != 0. The hack in
runtime/cgo works as well in a PIE executable as it does with a position
dependent one, so the simplest fix is to still emit a R_TLS_LE reloc when goos
== "android".
A real fix is to use something more like the IE model code but loading the
offset from %gs to the thread local storage from a global variable rather than
from a location chosen by the system linker (this is how android/arm works).
Issue #9327.
Change-Id: I9fbfc890ec7fe191f80a595b6cf8e2a1fcbe3034
Reviewed-on: https://go-review.googlesource.com/17049
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
This adds a test that runs CPU profiling with a high load of stack
barriers and stack barrier insertion/removal operations and checks
that both 1) the runtime doesn't crash and 2) stackBarrier itself
never appears in a profile. Prior to the fix for gentraceback starting
in the middle of stackBarrier, condition 2 often failed.
Change-Id: Ic28860448859029779844c4bf3bb28ca84611e2c
Reviewed-on: https://go-review.googlesource.com/17037
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
A sigprof during stack barrier insertion or removal can crash if it
detects an inconsistency between the stkbar array and the stack
itself. Currently we protect against this when scanning another G's
stack using stackLock, but we don't protect against it when unwinding
stack barriers for a recover or a memmove to the stack.
This commit cleans up and improves the stack locking code. It
abstracts out the lock and unlock operations. It uses the lock
consistently everywhere we perform stack operations, and pushes the
lock/unlock down closer to where the stack barrier operations happen
to make it more obvious what it's protecting. Finally, it modifies
sigprof so that instead of spinning until it acquires the lock, it
simply doesn't perform a traceback if it can't acquire it. This is
necessary to prevent self-deadlock.
Updates #11863, which introduced stackLock to fix some of these
issues, but didn't go far enough.
Updates #12528.
Change-Id: I9d1fa88ae3744d31ba91500c96c6988ce1a3a349
Reviewed-on: https://go-review.googlesource.com/17036
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Currently, if a profiling signal happens in the middle of
stackBarrier, gentraceback may see inconsistencies between stkbar and
the barriers on the stack and it will certainly get the wrong return
PC for stackBarrier. In most cases, the return PC won't be a PC at all
and this will immediately abort the traceback (which is considered
okay for a sigprof), but if it happens to be a valid PC this may sent
gentraceback down a rabbit hole.
Fix this by detecting when the gentraceback starts in stackBarrier and
simulating the completion of the barrier to get the correct initial
frame.
Change-Id: Ib11f705ac9194925f63fe5dfbfc84013a38333e6
Reviewed-on: https://go-review.googlesource.com/17035
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This CL adds skipped failing tests, showing differences between HTTP/1
and HTTP/2 behavior. They'll be fixed in later commits.
Only a tiny fraction of the net/http tests have been split into their
"_h1" and "_h2" variants. That will also continue. (help welcome)
Updates #6891
Updates #13315
Updates #13316
Updates #13317
Change-Id: I16c3c381dbe267a3098fb266ab0d804c36473a64
Reviewed-on: https://go-review.googlesource.com/17046
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Fix regression from https://golang.org/cl/16829 ("require valid methods
in NewRequest and Transport.RoundTrip").
An empty string is a valid method (it means "GET", per the docs).
Fixes#13311
Change-Id: I26b71dc4ccc146498b5d7e38fbe31ed11dd5a6cf
Reviewed-on: https://go-review.googlesource.com/16952
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
Handling of &(T{}) assumed that the parser would not introduce ()'s.
Also: Better comments around handling of OPAREN syntax tree optimization.
Fixes#13261.
Change-Id: Ifc5047a0448f5e7d74cd42f6608b87dcc9c2f2fb
Reviewed-on: https://go-review.googlesource.com/17040
Reviewed-by: Chris Manghane <cmang@golang.org>
Also:
- better error messages in some cases
- factored out function to produce syntax error at given line number
Fixes#13273.
Change-Id: I0192a94731cc23444680a26bd0656ef663e6da0b
Reviewed-on: https://go-review.googlesource.com/16992
Reviewed-by: Chris Manghane <cmang@golang.org>
Found by cmd/vet.
Change-Id: Id570ecd76c3f1efd9696680ccd9799610217f8f7
Reviewed-on: https://go-review.googlesource.com/17042
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This works by adding a call to __x86.get_pc_thunk.cx immediately before any
instruction that accesses global data and then assembling the instruction to
use the appropriate offset from CX instead of the absolute address. Some forms
cannot be assembled that way and are rewritten to load the address into CX
first.
-buildmode=pie works now, but is not yet tested.
Fixes#13201 (I think)
Change-Id: I32a8561e7fc9dd4ca6ae3b0e57ad78a6c50bf1f5
Reviewed-on: https://go-review.googlesource.com/17014
Reviewed-by: Ian Lance Taylor <iant@golang.org>
I was prodded into doing this in review comments for the ARM version, and it's
going to make shared libs for 386 easier.
Change-Id: Id12de801b1425b8c6b5736fe91b418fc123a4e40
Reviewed-on: https://go-review.googlesource.com/17012
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Does not fix#12327 but nicer anyway.
Change-Id: I4ad730a4ca833d76957b7571895b3a08a6a530d4
Reviewed-on: https://go-review.googlesource.com/16964
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Cgo-created threads transition between having associated Go g's and m's and not.
A signal arriving during the transition could think it was safe and appropriate to
run Go signal handlers when it was in fact not.
Avoid the race by masking all signals during the transition.
Fixes#12277.
Change-Id: Ie9711bc1d098391d58362492197a7e0f5b497d14
Reviewed-on: https://go-review.googlesource.com/16915
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
An internal link may need the C compiler support library, libgcc.a. Add
a -libgcc option to set the name of the compiler support library. If
-libgcc is not used, run the compiler to find it. Permit -libgcc=none
to skip using libgcc at all and hope for the best.
Change cmd/dist to not copy libgcc into the distribution. Add tests to
ensure that all the standard packages that use cgo can be linked in
internal mode without using libgcc. This ensures that somebody with a
Go installation without a C compiler can build programs.
Change-Id: I8ba35fb87ab0dd20e5cc0166b5f4145b04ce52a4
Reviewed-on: https://go-review.googlesource.com/16993
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
This includes the first parts of the general approach to PIC: load PC into CX
whenever it is needed. This is going to lead to large binaries and poor
performance but it's a start and easy to get right.
Change-Id: Ic8bf1d0a74284cca0d94a68cf75024e8ab063b4e
Reviewed-on: https://go-review.googlesource.com/16383
Reviewed-by: Ian Lance Taylor <iant@golang.org>
We might be able to do better than this, but it's a start.
Change-Id: I80ebce9094e084a4746039106ccf1ad9c4b8bb7c
Reviewed-on: https://go-review.googlesource.com/16384
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Mostly by avoiding CX entirely, sometimes by reloading it.
I also vetted the assembly in other packages, it's all fine.
Change-Id: I50059669aaaa04efa303cf22ac228f9d14d83db0
Reviewed-on: https://go-review.googlesource.com/16386
Reviewed-by: Ian Lance Taylor <iant@golang.org>
BER allows the sender to choose either short form or long form where
both are legal, but DER requires the minimal one be used. Enforce this
and add a test. Fix one test which was not minimally-encoded and another
which would not distinguish rejecting the input because the long form
length wasn't minimally-encoded from rejecting it because long form was
chosen when short form was allowed.
Change-Id: I1b56fcca594dcdeddea9378b4fab427cbe7cd26d
Reviewed-on: https://go-review.googlesource.com/16517
Reviewed-by: Adam Langley <agl@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
Otherwise _2006 is treated as _2 and then an error.
Fixes#11334
Change-Id: I40a385b45e279e9f4538bf419baab72781cdb215
Reviewed-on: https://go-review.googlesource.com/16311
Reviewed-by: Rob Pike <r@golang.org>
Compare basepath and targetpath using strings.EqualFold. The absence
of this on Windows causes an unterminating condition in `for` statement
later in the function.
Fixes#13258
Change-Id: Ib5a0caba864ee425dc75ece47b9cf6fb626f47f1
Reviewed-on: https://go-review.googlesource.com/16857
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Only apply RFC 6724's CommonPrefixLen rule for IPv4 source/destination
pairs that are members of the same IPv4 special purpose block.
Fixes#13283.
Change-Id: I2f7c26b408dd4675dfc5c1959e22d05b43bb8241
Reviewed-on: https://go-review.googlesource.com/16995
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
On Windows, Rel emits error messages of the form `Rel: can't make
\windows relative to \windows`. Rather than emitting paths after
stripping volume names, emit the original paths so as to make those of
the form `Rel: can't make d:\windows relative to c:\windows`. Fixed a
test that expected the error message to emit clean path instead of the
original.
Fixes#13259
Change-Id: I3a9bd5b137205f22794ec8046b4e917ee48cf750
Reviewed-on: https://go-review.googlesource.com/16858
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Replace the cross platform but unsafe [4]uintptr type with a OS
specific type, sigset. Most OSes already define sigset, and this
change defines a suitable sigset for the OSes that don't (darwin,
openbsd). The OSes that don't use m.sigmask (windows, plan9, nacl)
now defines sigset as the empty type, struct{}.
The gain is strongly typed access to m.sigmask, saving a dynamic
size sanity check and unsafe.Pointer casting. Also, some storage is
saved for each M, since [4]uinptr was conservative for most OSes.
The cost is that OSes that don't need m.sigmask has to define sigset.
completes ./all.bash with GOOS linux, on amd64
completes ./make.bash with GOOSes openbsd, android, plan9, windows,
darwin, solaris, netbsd, freebsd, dragonfly, all amd64.
With GOOS=nacl ./make.bash failed with a seemingly unrelated error.
R=go1.7
Change-Id: Ib460379f063eb83d393e1c5efe7333a643c1595e
Reviewed-on: https://go-review.googlesource.com/16942
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
