Cherry Zhang [Sun, 27 Oct 2019 02:48:15 +0000 (22:48 -0400)]
cmd/internal/obj/mips: fix encoding of FCR registers
The asm encoder generally assumes that the lowest 5 bits of the
REG_XX constants match the machine instruction encoding, i.e.
the lowest 5 bits is the register number. This was not true for
FCR registers and M registers. Make it so.
MOV Rx, FCRy was encoded as two machine instructions. The first
is unnecessary. Remove.
Mikhail Fesenko [Mon, 28 Oct 2019 21:51:00 +0000 (21:51 +0000)]
cmd/fix, cmd/go, cmd/gofmt: refactor common code into new internal diff package
Change-Id: Idac8473d1752059bf2f617fd7a781000ee2c3af4
GitHub-Last-Rev: 02a3aa1a3241d3ed4422518f1c954cd54bbe858e
GitHub-Pull-Request: golang/go#35141
Reviewed-on: https://go-review.googlesource.com/c/go/+/203218
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Shenghou Ma [Fri, 25 Oct 2019 01:29:30 +0000 (21:29 -0400)]
cmd/compile/internal/gc: reword "declared and not used" error message
"declared and not used" is technically correct, but might confuse
the user. Switching "and" to "but" will hopefully create the
contrast for the users: they did one thing (declaration), but
not the other --- actually using the variable.
This new message is still not ideal (specifically, declared is not
entirely precise here), but at least it matches the other parsers
and is one step in the right direction.
Change-Id: I725c7c663535f9ab9725c4b0bf35b4fa74b0eb20
Reviewed-on: https://go-review.googlesource.com/c/go/+/203282
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Agniva De Sarker [Wed, 13 Feb 2019 03:19:52 +0000 (08:49 +0530)]
go/ast: fix SortImports to handle block comments (take 2)
This is a 2nd attempt at fixing CL 162337 which had an off-by-one error.
We were unconditionally getting the position of the start of the next line
from the last import without checking whether it is the end of the file or not.
Fix the code to check for that and move the testcase added in CL 190523
to the end of the file for it to trigger the issue properly.
Fixes #18929
Change-Id: I59e77256e256570b160fea6a17bce9ef49e810df
Reviewed-on: https://go-review.googlesource.com/c/go/+/190480
Run-TryBot: Agniva De Sarker <agniva.quicksilver@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
John Papandriopoulos [Sun, 29 Sep 2019 23:59:56 +0000 (16:59 -0700)]
cmd/link: pass-through undefined call targets in external link mode
Allows Go asm calls referencing a function in a .syso file to be
passed through to the external linker, that would have otherwise
raised a "relocation target X not defined" error in cmd/link.
Fixes #33139
Change-Id: I2a8eb6063ebcd05fac96f141acf7652cf9189766
Reviewed-on: https://go-review.googlesource.com/c/go/+/198798
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Constantin Konstantinidis [Sat, 26 Oct 2019 18:01:47 +0000 (20:01 +0200)]
os: remove read-only directories in RemoveAll on Windows
Remove skipping of TestRemoveUnreadableDir on Windows.
Fixes #26295
Change-Id: I364a3caa55406c855ece807759f6298f7e4ddf1e
Reviewed-on: https://go-review.googlesource.com/c/go/+/203599
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Jay Conrod [Wed, 23 Oct 2019 21:14:09 +0000 (17:14 -0400)]
cmd/dist: support GOROOT vendoring
In the second step of make.bash, cmd/dist builds cmd/go by invoking
the compiler, linker, and other tools directly on transitive
dependencies of cmd/go. Essentially, cmd/dist acts as a minimal
version of 'go install' when building go_toolchain.
Until now, cmd/go has had no transitive dependencies in vendor
directories. This changes in CL 202698, where several packages are
deleted and equivalent versions in golang.org/x/mod are used
instead. So this CL adds support to cmd/dist for vendor directories.
Updates #31761
Change-Id: Iab4cdc7e505069a8df296287d16fbaa871944955
Reviewed-on: https://go-review.googlesource.com/c/go/+/203537
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Joshua M. Clulow [Mon, 28 Oct 2019 16:19:48 +0000 (09:19 -0700)]
runtime: make NumCPU respect zone CPU cap on illumos
On illumos systems, check for the "zone.cpu-cap" resource control when
determining how many usable CPUs are available. If the resource control
is not set, or we are unable to read it, ignore the failure and return
the value we used to return; i.e., the CPU count from
sysconf(_SC_NPROCESSORS_ONLN).
Cherry Zhang [Fri, 25 Oct 2019 20:43:08 +0000 (16:43 -0400)]
cmd/compile: delete ZeroAuto
ZeroAuto was used with the ambiguously live logic. The
ambiguously live logic is removed as we switched to stack
objects. It is now never called. Remove.
Lynn Boger [Mon, 28 Oct 2019 13:29:40 +0000 (09:29 -0400)]
runtime: fix textOff for multiple text sections
If a compilation has multiple text sections, code in
textOff must compare the offset argument against the range
for each text section to determine which one it is in.
The comparison looks like this:
if uintptr(off) >= sectaddr && uintptr(off) <= sectaddr+sectlen
If the off value being compared is equal to sectaddr+sectlen then it
is not within the range of the text section but after it. The
comparison should be just '<'.
Lynn Boger [Wed, 20 Mar 2019 13:30:27 +0000 (09:30 -0400)]
crypto/elliptic: add asm implementation for p256 on ppc64le
This adds an asm implementation of the p256 functions used
in crypto/elliptic, utilizing VMX, VSX to improve performance.
On a power9 the improvement is:
This implemenation is based on the s390x implementation, using
comparable instructions for most with some minor changes where the
instructions are not quite the same.
Some changes were also needed since s390x is big endian and ppc64le
is little endian.
Phil Pearl [Sun, 27 Oct 2019 16:05:54 +0000 (16:05 +0000)]
encoding/json: remove allocation when using a Marshaler with value receiver
If we marshal a non-pointer struct field whose type implements Marshaler with
a non-pointer receiver, then we avoid an allocation if we take the address of
the field before casting it to an interface.
name old time/op new time/op delta
EncodeMarshaler-8 104ns ± 1% 92ns ± 2% -11.72% (p=0.001 n=7+7)
name old alloc/op new alloc/op delta
EncodeMarshaler-8 36.0B ± 0% 4.0B ± 0% -88.89% (p=0.000 n=8+8)
name old allocs/op new allocs/op delta
EncodeMarshaler-8 2.00 ± 0% 1.00 ± 0% -50.00% (p=0.000 n=8+8)
Test coverage already looks good enough for this change. TestRefValMarshal
already covers all possible combinations of value & pointer receivers on
value and pointer struct fields.
Change-Id: I6fc7f72396396d98f9a90c3c86e813690f41c099
Reviewed-on: https://go-review.googlesource.com/c/go/+/203608 Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Phil Pearl [Sun, 13 Oct 2019 12:01:58 +0000 (13:01 +0100)]
encoding/json: improve performance of Compact
This change improves performance of Compact by using a sync.Pool to allow re-use
of a scanner. This also has the side-effect of removing an allocation for each
field that implements Marshaler when marshalling JSON.
name old time/op new time/op delta
EncodeMarshaler-8 118ns ± 2% 104ns ± 1% -12.21% (p=0.001 n=7+7)
name old alloc/op new alloc/op delta
EncodeMarshaler-8 100B ± 0% 36B ± 0% -64.00% (p=0.000 n=8+8)
name old allocs/op new allocs/op delta
EncodeMarshaler-8 3.00 ± 0% 2.00 ± 0% -33.33% (p=0.000 n=8+8)
Change-Id: Ic70c61a0a6354823da5220f5aad04b94c054f233
Reviewed-on: https://go-review.googlesource.com/c/go/+/200864 Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Jason A. Donenfeld [Sat, 26 Oct 2019 21:05:22 +0000 (23:05 +0200)]
internal/syscall/windows/registry: remove TestWalkFullRegistry due to false assumptions
This test's existence was predicated upon assumptions about the full
range of known data types and known data into those types. However,
we've learned from Microsoft that there are several undocumented secret
registry types that are in use by various parts of Windows, and we've
learned from inspection that many Microsoft uses of registry types don't
strictly adhere to the recommended value size. It's therefore foolhardy
to make any assumptions about what goes in and out of the registry, and
so this test, as well as its "blacklist", are meaningless.
Fixes #35084
Change-Id: I6c3fe5fb0e740e88858321b3b042c0ff1a23284e
Reviewed-on: https://go-review.googlesource.com/c/go/+/203604
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com> Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Giovanni Bajo [Fri, 27 Sep 2019 22:05:54 +0000 (00:05 +0200)]
cmd/compile: in poset, implement path collapsing
Sometimes, poset needs to collapse a path making all nodes in
the path aliases. For instance, we know that A<=N1<=B and we
learn that B<=A, we can deduce A==N1==B, and thus we can
collapse all paths from A to B into a single aliased node.
Currently, this is a TODO. This CL implements the path-collapsing
primitive by doing a DFS walk to build a bitset of all nodes
across all paths, and then calling the new aliasnodes that allow
to mark multiple nodes as aliases of a single master node.
This helps only 4 times in std+cmd, but it will be fundamental
when we will rely on poset to calculate numerical limits, to
calculate the correct values.
This also fixes #35157, a bug uncovered by a previous CL in this
serie. A testcase will be added soon.
Change-Id: I5fc54259711769d7bd7c2d166a5abc1cddc26350
Reviewed-on: https://go-review.googlesource.com/c/go/+/200861
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Giovanni Bajo [Fri, 27 Sep 2019 21:39:42 +0000 (23:39 +0200)]
cmd/compile: in poset, allow multiple aliases in a single pass
Change aliasnode into aliasnodes, to allow for recording
multiple aliases in a single pass. The nodes being aliased
are passed as bitset for performance reason (O(1) lookups).
It does look worse in the existing case of SetEqual where
we now need to allocate a bitset just for a single node,
but the new API will allow to fully implement a path-collapsing
primitive in next CL.
No functional changes, passes toolstash -cmp.
Change-Id: I06259610e8ef478106b36852464ed2caacd29ab5
Reviewed-on: https://go-review.googlesource.com/c/go/+/200860 Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Giovanni Bajo [Fri, 27 Sep 2019 22:08:45 +0000 (00:08 +0200)]
cmd/compile: in poset, refactor aliasnode
In preparation for allowing to make multiple nodes as aliases
in a single pass, refactor aliasnode splitting out the case
in which one of the nodes is not in the post into a new
funciton (aliasnewnode).
No functional changes, passes toolstash -cmp
Change-Id: I19ca6ef8426f8aec9f2622b6151c5c617dbb25b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/200859 Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ben Shi [Tue, 25 Jun 2019 10:38:21 +0000 (10:38 +0000)]
runtime: save/restore callee-save registers in arm's sigtramp
ARM's R4-R8 & R10-R11 are callee-save registers, and R9
may be callee-save or not. This CL saves them at the beginning
of sigtramp and restores them in the end.
Austin Clements [Mon, 14 Oct 2019 21:05:56 +0000 (17:05 -0400)]
runtime: M-targeted signals for BSDs
For these, we split up the existing runtime.raise assembly
implementation into its constituent "get thread ID" and "signal
thread" parts. This lets us implement signalM and reimplement raise in
pure Go. (NetBSD conveniently already had lwp_self.)
We also change minit to store the procid directly, rather than
depending on newosproc to do so. This is because newosproc isn't
called for the bootstrap M, but we need a procid for every M. This is
also simpler overall.
Cuong Manh Le [Wed, 23 Oct 2019 07:31:29 +0000 (14:31 +0700)]
runtime: mark findObject nosplit
findObject takes the pointer argument as uintptr. If the pointer is to
the local stack and calling findObject happens to require the stack to
be reallocated, then spanOf is called for the old pointer.
runtime: only shrink stacks at synchronous safe points
We're about to introduce asynchronous safe points, where we won't have
precise pointer maps for all stack frames. That's okay for scanning
the stack (conservatively), but not for shrinking the stack.
Hence, this CL prepares for this by only shrinking the stack as part
of the stack scan if the goroutine is stopped at a synchronous safe
point. Otherwise, it queues up the stack shrink for the next
synchronous safe point.
We already have one condition under which we can't shrink the stack
for very similar reasons: syscalls. Currently, we just give up on
shrinking the stack if it's in a syscall. But with this mechanism, we
defer that stack shrink until the next synchronous safe point.
runtime: make copystack/sudog synchronization more explicit
When we copy a stack of a goroutine blocked in a channel operation, we
have to be very careful because other goroutines may be writing to
that goroutine's stack. To handle this, stack copying acquires the
locks for the channels a goroutine is waiting on.
One complication is that stack growth may happen while a goroutine
holds these locks, in which case stack copying must *not* acquire
these locks because that would self-deadlock.
Currently, stack growth never acquires these locks because stack
growth only happens when a goroutine is running, which means it's
either not blocking on a channel or it's holding the channel locks
already. Stack shrinking always acquires these locks because shrinking
happens asynchronously, so the goroutine is never running, so there
are either no locks or they've been released by the goroutine.
However, we're about to change when stack shrinking can happen, which
is going to break the current rules. Rather than find a new way to
derive whether to acquire these locks or not, this CL simply adds a
flag to the g struct that indicates that stack copying should acquire
channel locks. This flag is set while the goroutine is blocked on a
channel op.
Currently, gcscanvalid is used to resolve a race between attempts to
scan a stack. Now that there's a clear owner of the stack scan
operation, there's no longer any danger of racing or attempting to
scan a stack more than once, so this CL eliminates gcscanvalid.
I double-checked my reasoning by first adding a throw if gcscanvalid
was set in scanstack and verifying that all.bash still passed.
Currently, the process of suspending a goroutine is tied to stack
scanning. In preparation for non-cooperative preemption, this CL
abstracts this into general purpose suspendG/resumeG functions.
suspendG and resumeG closely follow the existing scang and restartg
functions with one exception: the addition of a _Gpreempted status.
Currently, preemption tasks (stack scanning) are carried out by the
target goroutine if it's in _Grunning. In this new approach, the task
is always carried out by the goroutine that called suspendG. Thus, we
need a reliable way to drive the target goroutine out of _Grunning
until the requesting goroutine is ready to resume it. The new
_Gpreempted state provides the handshake: when a runnable goroutine
responds to a preemption request, it now parks itself and enters
_Gpreempted. The requesting goroutine races to put it in _Gwaiting,
which gives it ownership, but also the responsibility to start it
again.
This CL adds several TODOs about improving the synchronization on the
G status. The existing code already has these problems; we're just
taking note of them.
The next CL will remove the now-dead scang and preemptscan.
Tobias Klauser [Fri, 25 Oct 2019 18:48:07 +0000 (20:48 +0200)]
runtime: define emptyfunc as static function in assembly for freebsd/arm64
CL 198544 broke the linux/arm64 build because it declares emptyfunc for
GOARCH=arm64, but only freebsd/arm64 defines it. Make it a static
assembly function specific for freebsd/arm64 and remove the stub.
Fixes #35160
Change-Id: I5fd94249b60c6fd259c251407b6eccc8fa512934
Reviewed-on: https://go-review.googlesource.com/c/go/+/203418 Reviewed-by: Bryan C. Mills <bcmills@google.com>
Bryan C. Mills [Fri, 25 Oct 2019 17:55:10 +0000 (13:55 -0400)]
os/signal: derive TestAtomicStop timeout from overall test timeout
Previously, TestAtomicStop used a hard-coded 2-second timeout.
That empirically is not long enough on certain builders. Rather than
adjusting it to a different arbitrary value, use a slice of the
overall timeout for the test binary. If everything is working, we
won't block nearly that long anyway.
runtime: ensure _Grunning Gs have a valid g.m and g.m.p
We already claim on the documentation for _Grunning that this is case,
but execute transitions to _Grunning before assigning g.m. Fix this
and make the documentation even more explicit.
runtime: make m.libcallsp check in shrinkstack panic
Currently, shrinkstack will not shrink a stack on Windows if
gp.m.libcallsp != 0. In general, we can't shrink stacks in syscalls
because the syscall may hold pointers into the stack, and in principle
this is supposed to be preventing that for libcall-based syscalls
(which are direct syscalls from the runtime). But this test is
actually broken and has been for a long time. That turns out to be
okay because it also appears it's not necessary.
This test is racy. g.m points to whatever M the G was last running on,
even if the G is in a blocked state, and that M could be doing
anything, including making libcalls. Hence, observing that libcallsp
== 0 at one moment in shrinkstack is no guarantee that it won't become
non-zero while we're shrinking the stack, and vice-versa.
It's also weird that this check is only performed on Windows, given
that we now use libcalls on macOS, Solaris, and AIX.
This check was added when stack shrinking was first implemented in CL 69580044. The history of that CL (though not the final version)
suggests this was necessary for libcalls that happened on Go user
stacks, which we never do now because of the limited stack space.
It could also be defending against user stack pointers passed to
libcall system calls from blocked Gs. But the runtime isn't allowed to
keep pointers into the user stack for blocked Gs on any OS, so it's
not clear this would be of any value.
Hence, this checks seems to be simply unnecessary.
Rather than simply remove it, this CL makes it defensive. We can't do
anything about blocked Gs, since it doesn't even make sense to look at
their M, but if a G tries to shrink its own stack while in a libcall,
that indicates a bug in the libcall code. This CL makes shrinkstack
panic in this case.
For #10958, #24543, since those are going to rearrange how we decide
that it's safe to shrink a stack.
Austin Clements [Sat, 12 Oct 2019 22:35:49 +0000 (18:35 -0400)]
cmd/internal/obj/x86: correct pcsp for ADJSP
The x86 assembler supports an "ADJSP" pseudo-op that compiles to an
ADD/SUB from SP. Unfortunately, while this seems perfect for an
instruction that would allow obj to continue to track the SP/FP delta,
obj currently doesn't do that. As a result, FP-relative references
won't work and, perhaps worse, the pcsp table will have the wrong
frame size.
We don't currently use this instruction in any assembly or generate it
in the compiler, but this is a perfect instruction for solving a
problem in #24543.
This CL makes ADJSP useful by incorporating it into the SP delta
logic.
One subtlety is that we do generate ADJSP in obj itself to open a
function's stack frame. Currently, when preprocess enters the loop to
compute the SP delta, it may or may not start at this ADJSP
instruction depending on various factors. We clean this up by instead
always starting the SP delta at 0 and always starting this loop at the
entry to the function.
Why not just recognize ADD/SUB of SP? The danger is that could change
the meaning of existing code. For example, walltime1 in
sys_linux_amd64.s saves SP, SUBs from it, and aligns it. Later, it
restores the saved copy and then does a few FP-relative references.
Currently obj doesn't know any of this is happening, but that's fine
once it gets to the FP-relative references. If we taught obj to
recognize the SUB, it would start to miscompile this code. An
alternative would be to recognize unknown instructions that write to
SP and refuse subsequent FP-relative references, but that's kind of
annoying.
This passes toolstash -cmp for std on both amd64 and 386.
Jason A. Donenfeld [Wed, 23 Oct 2019 20:27:29 +0000 (22:27 +0200)]
internal/syscall/windows/registry: allow for non-null terminated strings
According to MSDN, "If the data has the REG_SZ, REG_MULTI_SZ or
REG_EXPAND_SZ type, this size includes any terminating null character or
characters unless the data was stored without them. [...] If the data
has the REG_SZ, REG_MULTI_SZ or REG_EXPAND_SZ type, the string may not
have been stored with the proper terminating null characters. Therefore,
even if the function returns ERROR_SUCCESS, the application should
ensure that the string is properly terminated before using it;
otherwise, it may overwrite a buffer."
It's therefore dangerous to pass it off unbounded as we do, and in fact
this led to crashes on real systems.
Change-Id: I6d786211814656f036b87fd78631466634cd764a
Reviewed-on: https://go-review.googlesource.com/c/go/+/202937
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Jay Conrod [Tue, 22 Oct 2019 15:30:20 +0000 (11:30 -0400)]
cmd/go: add -modfile flag that sets go.mod file to read/write
This change adds the -modfile flag to module aware build commands and
to 'go mod' subcommands. -modfile may be set to a path to an alternate
go.mod file to be read and written. A real go.mod file must still
exist and is used to set the module root directory. However, it is not
opened.
When -modfile is set, the effective location of the go.sum file is
also changed to the -modfile with the ".mod" suffix trimmed (if
present) and ".sum" added.
Updates #34506
Change-Id: I2d1e044e18af55505a4f24bbff09b73bb9c908b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/202564
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Jay Conrod [Fri, 18 Oct 2019 21:39:31 +0000 (17:39 -0400)]
doc: add skeleton module documentation with headings
Sections will be filled in with individual CLs before Go 1.14.
NOTE: This document is currently in Markdown for ease of writing /
reviewing. Before Go 1.14, we will either ensure that x/website
can render Markdown (flavor TBD) or check in a rendered HTML file that
can be displayed directly.
Updates #33637
Change-Id: Icd43fa2bdb7d256b28a56b93214b70343f43492e
Reviewed-on: https://go-review.googlesource.com/c/go/+/202081 Reviewed-by: Bryan C. Mills <bcmills@google.com>
Bryan C. Mills [Thu, 24 Oct 2019 13:17:54 +0000 (09:17 -0400)]
cmd/go: re-enable 'go list -m' with -mod=vendor for limited patterns
I had prohibited 'go list -m' with -mod=vendor because the module
graph is incomplete, but I've realized that many queries do not
actually require the full graph — and may, in fact, be driven using
modules previously reported by 'go list' for specific, vendored
packages. Queries for those modules should succeed.
Updates #33848
Change-Id: I1000b4cf586a830bb78faf620ebf62d73a3cb300
Reviewed-on: https://go-review.googlesource.com/c/go/+/203138
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Dan Scales [Mon, 24 Jun 2019 19:59:22 +0000 (12:59 -0700)]
cmd/compile, cmd/link, runtime: make defers low-cost through inline code and extra funcdata
Generate inline code at defer time to save the args of defer calls to unique
(autotmp) stack slots, and generate inline code at exit time to check which defer
calls were made and make the associated function/method/interface calls. We
remember that a particular defer statement was reached by storing in the deferBits
variable (always stored on the stack). At exit time, we check the bits of the
deferBits variable to determine which defer function calls to make (in reverse
order). These low-cost defers are only used for functions where no defers
appear in loops. In addition, we don't do these low-cost defers if there are too
many defer statements or too many exits in a function (to limit code increase).
When a function uses open-coded defers, we produce extra
FUNCDATA_OpenCodedDeferInfo information that specifies the number of defers, and
for each defer, the stack slots where the closure and associated args have been
stored. The funcdata also includes the location of the deferBits variable.
Therefore, for panics, we can use this funcdata to determine exactly which defers
are active, and call the appropriate functions/methods/closures with the correct
arguments for each active defer.
In order to unwind the stack correctly after a recover(), we need to add an extra
code segment to functions with open-coded defers that simply calls deferreturn()
and returns. This segment is not reachable by the normal function, but is returned
to by the runtime during recovery. We set the liveness information of this
deferreturn() to be the same as the liveness at the first function call during the
last defer exit code (so all return values and all stack slots needed by the defer
calls will be live).
I needed to increase the stackguard constant from 880 to 896, because of a small
amount of new code in deferreturn().
The -N flag disables open-coded defers. '-d defer' prints out the kind of defer
being used at each defer statement (heap-allocated, stack-allocated, or
open-coded).
Cost of defer statement [ go test -run NONE -bench BenchmarkDefer$ runtime ]
With normal (stack-allocated) defers only: 35.4 ns/op
With open-coded defers: 5.6 ns/op
Cost of function call alone (remove defer keyword): 4.4 ns/op
Text size increase (including funcdata) for go binary without/with open-coded defers: 0.09%
The average size increase (including funcdata) for only the functions that use
open-coded defers is 1.1%.
The cost of a panic followed by a recover got noticeably slower, since panic
processing now requires a scan of the stack for open-coded defer frames. This scan
is required, even if no frames are using open-coded defers:
Cost of panic and recover [ go test -run NONE -bench BenchmarkPanicRecover runtime ]
Without open-coded defers: 62.0 ns/op
With open-coded defers: 255 ns/op
A CGO Go-to-C-to-Go benchmark got noticeably faster because of open-coded defers:
CGO Go-to-C-to-Go benchmark [cd misc/cgo/test; go test -run NONE -bench BenchmarkCGoCallback ]
Without open-coded defers: 443 ns/op
With open-coded defers: 347 ns/op
Bryan C. Mills [Wed, 23 Oct 2019 20:10:55 +0000 (16:10 -0400)]
cmd/go/internal/list: ensure that cfg.BuildMod is initialized before reading it in 'go list -m'
The default value of cfg.BuildMod depends on the 'go' version in the
go.mod file. The go.mod file is read and parsed, and its settings are
applied, in modload.InitMod.
As it turns out, modload.Enabled does not invoke InitMod, so
cfg.BuildMod is not necessarily set even if modload.Enabled returns
true.
Updates #33848
Change-Id: I13a4dd80730528e6f1a5acc492fcfe07cb59d94e
Reviewed-on: https://go-review.googlesource.com/c/go/+/202917
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
Robert Griesemer [Wed, 23 Oct 2019 23:44:51 +0000 (16:44 -0700)]
math/big: make Rat.Denom side-effect free
A Rat is represented via a quotient a/b where a and b are Int values.
To make it possible to use an uninitialized Rat value (with a and b
uninitialized and thus == 0), the implementation treats a 0 denominator
as 1.
Rat.Num and Rat.Denom return pointers to these values a and b. Because
b may be 0, Rat.Denom used to first initialize it to 1 and thus produce
an undesirable side-effect (by changing the Rat's denominator).
This CL changes Denom to return a new (not shared) *Int with value 1
in the rare case where the Rat was not initialized. This eliminates
the side effect and returns the correct denominator value.
While this is changing behavior of the API, the impact should now be
minor because together with (prior) CL https://golang.org/cl/202997,
which initializes Rats ASAP, Denom is unlikely used to access the
denominator of an uninitialized (and thus 0) Rat. Any operation that
will somehow set a Rat value will ensure that the denominator is not 0.
Robert Griesemer [Wed, 23 Oct 2019 21:22:32 +0000 (14:22 -0700)]
math/big: normalize unitialized denominators ASAP
A Rat is represented via a quotient a/b where a and b are Int values.
To make it possible to use an uninitialized Rat value (with a and b
uninitialized and thus == 0), the implementation treats a 0 denominator
as 1.
For each operation we check if the denominator is 0, and then treat
it as 1 (if necessary). Operations that create a new Rat result,
normalize that value such that a result denominator 1 is represened
as 0 again.
This CL changes this behavior slightly: 0 denominators are still
interpreted as 1, but whenever we (safely) can, we set an uninitialized
0 denominator to 1. This simplifies the code overall.
Cherry Zhang [Wed, 23 Oct 2019 15:42:23 +0000 (11:42 -0400)]
runtime: save/fetch g register during VDSO on ARM and ARM64
On ARM and ARM64, during a VDSO call, the g register may be
temporarily clobbered by the VDSO code. If a signal is received
during the execution of VDSO code, we may not find a valid g
reading the g register. In CL 192937, we conservatively assume
g is nil. But this approach has a problem: we cannot handle
the signal in this case. Further, if the signal is not a
profiling signal, we'll call badsignal, which calls needm, which
wants to get an extra m, but we don't have one in a non-cgo
binary, which cuases the program to hang.
This is even more of a problem with async preemption, where we
will receive more signals than before. I ran into this problem
while working on async preemption support on ARM64.
In this CL, before making a VDSO call, we save the g on the
gsignal stack. When we receive a signal, we will be running on
the gsignal stack, so we can fetch the g from there and move on.
We probably want to do the same for PPC64. Currently we rely on
that the VDSO code doesn't actually clobber the g register, but
this is not guaranteed and we don't have control with.
Dmitri Shuralyov [Wed, 23 Oct 2019 17:50:15 +0000 (13:50 -0400)]
misc: delete benchcmp forwarding script
benchcmp was moved out of misc into x/tools in CL 60100043 in 2014,
and then replaced by a forwarding script in CL 82710043.
Five years have since passed, and the forwarding script has outlived
its usefulness. It's now more confusing than helpful. Delete it.
Jason A. Donenfeld [Tue, 22 Oct 2019 20:39:30 +0000 (22:39 +0200)]
internal/syscall/windows/registry: blacklist certain registry keys in TestWalkFullRegistry
It turns out that Windows has "legitimate" keys that have bogus type
values or bogus lengths that don't correspond with their type. On up to
date Windows 10 systems, this test always fails for this reason. These
keys exist because of bugs in Microsoft's code. This commit works around
the problem by simply blacklisting known instances. It also expands the
error message a bit so that we can make adjustments should the problem
ever happen again, and reformats the messages so that it makes copy and
pasting into the blacklist easier.
Updates #35084
Change-Id: I50322828c0eb0ccecbb62d6bf4f9c726fa0b3c27
Reviewed-on: https://go-review.googlesource.com/c/go/+/202897
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Jay Conrod [Wed, 23 Oct 2019 17:47:36 +0000 (13:47 -0400)]
cmd/go: ignore '@' when cleaning local and absolute file path args
Since CL 194600, search.CleanPaths preserves characters after '@' in
each argument. This was done so that paths could be cleaned while
version queries were preserved. However, local and absolute file paths
may contain '@' characters.
With this change, '@' is treated as a normal character by
search.CleanPaths in local and absolute paths.
Fixes #35115
Change-Id: Ia7d37e0a2737442d4f1796cc2fc3a59237a8ddfe
Reviewed-on: https://go-review.googlesource.com/c/go/+/202761
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Jason A. Donenfeld [Wed, 23 Oct 2019 11:08:46 +0000 (13:08 +0200)]
syscall: reenable sysctl on iOS
This was disabled due to a report that the App Store rejects the symbol
__sysctl. However, we use the sysctl symbol, which is fine. The __sysctl
symbol is used by x/sys/unix, which needs fixing instead. So, this
commit reenables sysctl on iOS, so that things like net.InterfaceByName
can work again.
This reverts CL 193843, CL 193844, CL 193845, and CL 193846.
Fixes #35101
Updates #34133
Updates #35103
Change-Id: Ib8eb9f87b81db24965b0de29d99eb52887c7c60a
Reviewed-on: https://go-review.googlesource.com/c/go/+/202778
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com> Reviewed-by: David Crawshaw <crawshaw@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Thu, 11 Apr 2019 21:20:54 +0000 (14:20 -0700)]
runtime: add race detector support for new timers
Since the new timers run on g0, which does not have a race context,
we add a race context field to the P, and use that for timer functions.
This works since all timer functions are in the standard library.
Updates #27707
Change-Id: I8a5b727b4ddc8ca6fc60eb6d6f5e9819245e395b
Reviewed-on: https://go-review.googlesource.com/c/go/+/171882
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Jason A. Donenfeld [Tue, 22 Oct 2019 20:39:30 +0000 (22:39 +0200)]
internal/syscall/windows/registry: fix strict assumptions in TestWalkFullRegistry
It turns out that Windows has "legitimate" keys that have bogus type
values or bogus lengths that don't correspond with their type. On up to
date Windows 10 systems, this test always fails for this reason.
So, this commit alters the test to simply log the discrepancy and move
on.
Fixes #35084
Change-Id: I56e12cc62aff49cfcc38ff01a19dfe53153976a0
Reviewed-on: https://go-review.googlesource.com/c/go/+/202678
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Matthew Dempsky [Thu, 17 Oct 2019 20:24:34 +0000 (13:24 -0700)]
cmd/compile: enable -d=checkptr when -race or -msan is specified
It can still be manually disabled again using -d=checkptr=0.
It's also still disabled by default for GOOS=windows, because the
Windows standard library code has a lot of unsafe pointer conversions
that need updating.
Matthew Dempsky [Tue, 22 Oct 2019 18:59:00 +0000 (11:59 -0700)]
runtime: fix -d=checkptr failure for testing/quick
This CL extends checkptrBase to recognize pointers into the stack and
data/bss sections. I was meaning to do this eventually anyway, but
it's also an easy way to workaround #35068.