This adds the set of GOEXPERIMENTs to the build version if it differs
from the default set of experiments. This exposes the experiment
settings via runtime.Version() and "go version <binary>".
Change-Id: I143dbbc50f66a4cf175469199974e18848075af6
Reviewed-on: https://go-review.googlesource.com/c/go/+/307820
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Currently, we have boolean and integral constants for GOEXPERIMENTs in
various places. Consolidate these into automatically generated
constants in the internal/goexperiment package.
Change-Id: I42a49aba2a3b4c722fedea23a613162cd8a67bee
Reviewed-on: https://go-review.googlesource.com/c/go/+/307818
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Currently, the objabi.Experiment fields use Go-standard CamelCase, the
GOEXPERIMENT environment variable flags and build tags use all
lowercase, and the asm macros use upper-case with underscores.
This CL makes asm use the lowercase names for macros so there is one
less spelling, e.g., GOEXPERIMENT_regabiargs. This also makes them
consistent with the GOOS_* and GOARCH_* macros, which also use lower
case.
Change-Id: I305cd89af5e8cd1a89cc148746c034bcfd76db3c
Reviewed-on: https://go-review.googlesource.com/c/go/+/307816
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
We have ways to statically access experiments now, so we don't need a
relatively clunky string-parsing dynamic way to do it.
Change-Id: I5d75480916eef4bde2c30d5fe30593180da77ff2
Reviewed-on: https://go-review.googlesource.com/c/go/+/307815
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
update comment cause gc/reflect.go has been moved to reflectdata/reflect.go
In the commit (attach below), gc/reflect.go is moved to reflectdata/reflect.go
So the comment referring gc/reflect.go should be updated to reflectdata/reflect.go
There maybe other places that refers gc/reflect.go that should be updated.
I would work around it soon.
commit:
de65151e50e4895ab4c0
Change-Id: Ieed5c48049ffe6889c08e164972fc7825653ac05
GitHub-Last-Rev: eec9c2328d
GitHub-Pull-Request: golang/go#45421
Reviewed-on: https://go-review.googlesource.com/c/go/+/307930
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
A non-trivial Cgo program may need to use callbacks and interact with
go objects per goroutine. Because of the rules for passing pointers
between Go and C, such a program needs to store handles to associated
Go values. This often causes much extra effort to figure out a way to
correctly deal with: 1) map collision; 2) identifying leaks and 3)
concurrency.
This CL implements a Handle representation in runtime/cgo package, and
related methods such as Value, Delete, etc. which allows Go users can
use a standard way to handle the above difficulties.
In addition, the CL allows a Go value to have multiple handles, and the
NewHandle always returns a different handle compare to the previously
returned handles. In comparison, CL 294670 implements a different
behavior of NewHandle that returns a unique handle when the Go value is
referring to the same object.
Benchmark:
name time/op
Handle/non-concurrent-16 487ns ± 1%
Handle/concurrent-16 674ns ± 1%
Fixes#37033
Change-Id: I0eadb9d44332fffef8fb567c745246a49dd6d4c1
Reviewed-on: https://go-review.googlesource.com/c/go/+/295369
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Cherry Zhang <cherryyz@google.com>
runtime.raceread/racewrite/racewriterange are functions that are
called from compiler instrumented code, follwoing ABIInternal.
They are assembly functions defined as ABIInternal in the runtime,
in order to avoid wrappers because they need to get the caller's
PC. This CL makes them to use the actual internal ABI.
Change-Id: Id91d73cf257f7b11a858958d85c38c4aa904d9c3
Reviewed-on: https://go-review.googlesource.com/c/go/+/307812
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
This is a simple workaround for a bug where runtime.GC() can return
before finishing a full sweep, causing gcTestIsReachable to throw. The
right thing is to fix runtime.GC(), but this should get this test
passing reliably in the meantime.
Updates #45315.
Change-Id: Iae141e6dbb26a9c2649497c1feedd4aaeaf540c7
Reviewed-on: https://go-review.googlesource.com/c/go/+/307809
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
In TestMorestack, on macOS, for some reason it got most of the
samples in synchronization (e.g. pthread_cond_signal and
pthread_cond_wait) and sometimes in other "syscalls" (usleep,
nanotime1), and very few samples in stack copying, sometimes 0,
which causes the test to fail. Maybe synchronization is slower on
macOS? (It doesn't seem so to me.) Or it is the OS's accounting
problem, where it is more likely to trigger a profiling signal
at a syscall (or certain kinds of syscalls)?
As the test is really about whether it can connect stack copying
with the function that grows the stack, this CL makes it spend
more time in copying stack than synchronization. Now it's getting
~100 samples for stack copying on a 5 second interval on my
machine, and the test passes reliably.
Fixes#44418.
Change-Id: I3a462c8c39766f2d67d697598f8641bbe64f16ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/307730
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
reflectcall expects the ABI0 calling convention, but it's marked as
ABIInternal. When it gets called this way, naturally it doesn't work
very well.
For #40724.
Change-Id: Ic76237420cd8c72f5df1c1ac7972ad6f989f8402
Reviewed-on: https://go-review.googlesource.com/c/go/+/306931
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
This change eliminates the use of funcPC to determine if an PC is in
abort. Using funcPC for this purpose is problematic when using plugins
because symbols no longer have unique PCs. funcPC also grabs the wrapper
for runtime.abort which isn't what we want for the new register ABI, so
rather than mark runtime.abort as ABIInternal, use funcID.
For #40724.
Change-Id: I2730e99fe6f326d22d64a10384828b94f04d101a
Reviewed-on: https://go-review.googlesource.com/c/go/+/307391
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Austin Clements <austin@google.com>
Currently, when converting a float (say float64), we use convT64
function. In the runtime convT64 expects a uint64 argument. In
the compiler, convT64 is defined as taking an "any" argument (so
it works with also uint64-like types such as [1]uint64). The "any"
type is instantiated with the concrete type in walk. So the
backend will see instances such as convT64([1]uint64).
Currently, float64 is treated as uint64-like. So the backend will
see convT64(float64). With a memory-based calling convention this
is fine. With a register-based calling convention, however, it
will pass the argument in a floating point register, whereas the
runtime expects the argument in an integer register (as it is
declared as uint64).
To fix this, this CL introduces runtime functions convT32F and
convT64F. They behave the same as convT32/convT64, but with a
float argument. In the compiler, use convT32F/convT64F to convert
float-like type to interface.
With this, "GOEXPERIMENT=regabi,regabiargs go test math fmt"
works.
Updates #40724.
Change-Id: I8b2e232096a95e4a7c4ab81795d77ef224ffaab3
Reviewed-on: https://go-review.googlesource.com/c/go/+/307232
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
For stack frames larger than StackBig, the stack split prologue needs
to guard against potential wraparound. Currently, it carefully
arranges to avoid underflow, but this is complicated and requires a
special check for StackPreempt. StackPreempt is no longer the only
stack poison value, so this check will incorrectly succeed if the
stack bound is poisoned with any other value.
This CL simplifies the logic of the check, reduces its length, and
accounts for any possible poison value by directly checking for
underflow.
Change-Id: I917a313102d6a21895ef7c4b0f304fb84b292c81
Reviewed-on: https://go-review.googlesource.com/c/go/+/307010
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
The CV add changes according to TODO in Go source-code.
Internal atomic set does not comply with sync/atomic library and has shortage
operations for signed integers.
This patch extend internal atomic set by Int32 and Int64 operations. It's
implemented new aliases and asm versions of operations. As a result Cas64 was
replaced by Casint64 in findRunnableGCWorker without type casting.
Another purpose is unified structure of internal atomics' source code. Before,
assembly impementations for different archs were in different files. For
example, filename for AMD64 was asm_amd64.s, but filename for RISC-V was
atomic_riscv64.s. Some arches have both files without any meaning. So, assembly
files were merged and renamed to atomic_{$ARCH}.s filenames.
Change-Id: I29a05a7cbf5f4a9cc146e8315536c038af545677
Reviewed-on: https://go-review.googlesource.com/c/go/+/289152
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Most of the time, the pprof tests are passing, except
for the builder. The reason is still unknown but I'd rather release
the builder to avoid missing other more important bugs.
Updates #45170
Change-Id: I667543ee1ae309b7319c5b3676a0901b4d0ecf2e
Reviewed-on: https://go-review.googlesource.com/c/go/+/306489
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
Internally the compiler uses a0,a1,a3 string, not [3]string,
and this lead to different parameter passing (memory, versus registers)
which of course did not work.
Updates #40724.
Change-Id: I1dbf479b88134559ba54b4b00a042b9a0fd128b8
Reviewed-on: https://go-review.googlesource.com/c/go/+/306910
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
This change finishes off functionality register ABI for the reflect
package.
Specifically, it implements a call on a MakeFunc'd value by performing
the reverse process that reflect.Value.Call does, using the same ABI
steps. It implements a call on a method value created by reflect by
translating between the method value's ABI to the method's ABI.
Tests are added for both cases.
For #40724.
Change-Id: I302820b61fc0a8f94c5525a002bc02776aef41af
Reviewed-on: https://go-review.googlesource.com/c/go/+/298670
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Currently, gcTestMoveStackOnNextCall doubles the stack allocation on
each call because stack movement always doubles the stack. That's
rather unfortunate if you're doing a bunch of stack movement tests in
a row that don't actually have to grow the stack because you'll
quickly hit the stack size limit even though you're hardly using any
of the stack.
Fix this by adding a special stack poison value for
gcTestMoveStackOnNextCall that newstack recognizes and inhibits the
allocation doubling.
Change-Id: Iace7055a0f33cb48dc97b8f4b46e45304bee832c
Reviewed-on: https://go-review.googlesource.com/c/go/+/306672
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
gcTestMoveStackOnNextCall can fail to move the stack in very rare
cases if there's an unfortunately timed preemption that clobbers the
stack guard. This won't happen multiple times in quick succession, so
make the test just retry a few times.
Change-Id: I247dc0551514e269e7132cee7945291429b0e865
Reviewed-on: https://go-review.googlesource.com/c/go/+/306671
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
This change causes finalizers, reflect calls, and Windows syscall
callbacks to assume the register ABI when GOEXPERIMENT=regabiargs is
set. That is, when all Go functions are using the new ABI by default,
these features should assume the new ABI too.
For #40724.
Change-Id: Ie4ee66b8085b692e1ff675f01134c9a4703ae1b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/306571
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
The covers three kinds of uses:
1. Calls of closures from assembly. These are always ABIInternal calls
without wrappers. I went through every indirect call in the runtime
and I think mcall is the only case of assembly calling a Go closure in
a way that's affected by ABIInternal. systemstack also calls a
closure, but it takes no arguments.
2. Calls of Go functions that expect raw ABIInternal pointers. I also
only found one of these: callbackasm1 -> cgocallback on Windows. These
are trickier to find, though.
3. Finally, I found one case on NetBSD where new OS threads were
directly calling the Go runtime entry-point from assembly via a PC,
rather than going through a wrapper. This meant new threads may not
have special registers set up. In this case, a change on all other
OSes had already forced new thread entry to go through an ABI wrapper,
so I just caught NetBSD up with that change.
With this change, I'm able to run a "hello world" with
GOEXPERIMENT=regabi,regabiargs.
For #40724.
Change-Id: I2a6d0e530c4fd4edf13484d923891c6160d683aa
Reviewed-on: https://go-review.googlesource.com/c/go/+/305669
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
This change modifies the system that allows Go functions to be set as
callbacks in various Windows systems to support the new register ABI.
For #40724.
Change-Id: Ie067f9e8a76c96d56177d7aa88f89cbe7223e12e
Reviewed-on: https://go-review.googlesource.com/c/go/+/300113
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
With GOEXPERIMENT=regabidefer, all deferred functions take no
arguments and have no results (their signature is always func()).
Since the signature is fixed, we can replace all of the reflectcalls
in the defer code with direct closure calls.
For #40724.
Change-Id: I3acd6742fe665610608a004c675f473b9d0e65ee
Reviewed-on: https://go-review.googlesource.com/c/go/+/306010
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Change-Id: I78c5744bb01988f1f599569703d83fd21542ac7a
Reviewed-on: https://go-review.googlesource.com/c/go/+/305911
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
exitsyscall0 contains two G variables: _g_ and gp. _g_ is the active G,
g0, while gp is the G to run (which just exited from a syscall).
It is passing _g_ to schedEnabled, which is incorrect; we are about to
execute gp, so that is what we should be checking the schedulability of.
While this is incorrect and should be fixed, I don't think it has ever
caused a problem in practice:
* g0 does not have g.startpc set, so schedEnabled simplifies to
just !sched.disable.user.
* This is correct provided gp is never a system goroutine.
* As far as I know, system goroutines never use entersyscall /
exitsyscall.
As far I can tell, this was a simple copy/paste error from exitsyscall,
where variable _g_ is the G to run.
While we are here, eliminate _g_ entirely, as the one other use is
identical to using gp.
Change-Id: I5df98a34569238b89ab13ff7012cd756fefb10dc
Reviewed-on: https://go-review.googlesource.com/c/go/+/291329
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
When using cgo, some of the frames can be provided by cgoTraceback, a
cgo-provided function to generate C tracebacks. Unlike Go tracebacks,
cgoTraceback has no particular guarantees that it produces valid
tracebacks.
If one of the (invalid) frames happens to put the PC in the alignment
region at the end of a function (filled with int 3's on amd64), then
Frames.Next will find a valid funcInfo for the PC, but pcdatavalue will
panic because PCDATA doesn't cover this PC.
Tolerate this case by doing a non-strict PCDATA lookup. We'll still show
a bogus frame, but at least avoid throwing.
Fixes#44971
Change-Id: I9eed728470d6f264179a7615bd19845c941db78c
Reviewed-on: https://go-review.googlesource.com/c/go/+/301369
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
This CL adds a set of helper functions for testing GC interactions.
These are intended for use in the regabi signature fuzzer, but are
generally useful for GC tests, so we make them generally available to
runtime tests.
These provide:
1. An easy way to force stack movement, for testing stack copying.
2. A simple and robust way to check the reachability of a set of
pointers.
3. A way to check what general category of memory a pointer points to,
mostly so tests can make sure they're testing what they mean to.
For #40724, but generally useful.
Change-Id: I15d33ccb3f5a792c0472a19c2cc9a8b4a9356a66
Reviewed-on: https://go-review.googlesource.com/c/go/+/305330
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
The specials processing loop in mspan.sweep is about to get more
complicated and I'm too allergic to list manipulation to open code
more of it there.
Change-Id: I767a0889739da85fb2878fc06a5c55b73bf2ba7d
Reviewed-on: https://go-review.googlesource.com/c/go/+/305551
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
Change-Id: I18b9508904f19d5aa68355c937c30b5fdf35442c
Reviewed-on: https://go-review.googlesource.com/c/go/+/305249
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
On ARM64, (external) linker generated trampoline may clobber R16
and R17. In CL 183842 we change Duff's devices not to use those
registers. However, this is not enough. The register allocator
also needs to know that these registers may be clobbered in any
calls that don't follow the standard Go calling convention. This
include Duff's devices and the write barrier.
Fixes#32773, second attempt.
Change-Id: Ia52a891d9bbb8515c927617dd53aee5af5bd9aa4
Reviewed-on: https://go-review.googlesource.com/c/go/+/184437
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Meng Zhuo <mzh@golangcn.org>
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Meng Zhuo <mzh@golangcn.org>
On Windows, when calling into needm in cgocallback on a new thread that
is unknown to the Go runtime, we currently call through an ABI wrapper.
The ABI wrapper tries to restore the G register from TLS.
On other platforms, TLS is set up just enough that the wrapper will
simply load a nil g from TLS, but on Windows TLS isn't set up at all, so
there's nowhere for the wrapper to load from.
So, bypass the wrapper in the call to needm. needm takes no arguments
and returns no results so there are no special ABI considerations,
except that we must clear X15 which is used as a zero register in Go
code (a function normally performed by the ABI wrapper). needm is also
otherwise already special and carefully crafted to avoid doing anything
that would require a valid G or M, at least until it is able to create
one.
While we're here, this change simplifies setg so that it doesn't set up
TLS on Windows and instead provides an OS-specific osSetupTLS to do
that.
The result of this is that setg(nil) no longer clears the TLS space
pointer on Windows. There's exactly one place this is used (dropm) where
it doesn't matter anymore, and an empty TLS means that setg's wrapper
will crash on the return path. Another result is that the G slot in the
TLS will be properly cleared, however, which isn't true today.
For #40724.
Change-Id: I65c3d924a3b16abe667b06fd91d467d6d5da31d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/303070
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
When converting a type T to a non-empty interface I, we build the
itab which contains the code pointers of the methods. Currently,
this brings those methods live (if the itab is live), even if the
interface method is never used. This CL changes the itab to use
weak references, so the methods can be pruned if not otherwise
live.
Fixes#42421.
Change-Id: Iee5de2ba11d603c5a102a2ba60440d839a7f9702
Reviewed-on: https://go-review.googlesource.com/c/go/+/268479
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Adds code to the compiler's "order" phase to rewrite go and defer
statements to always be argument-less. E.g.
defer f(x,y) => x1, y1 := x, y
defer func() { f(x1, y1) }
This transformation is not beneficial on its own, but it helps
simplify runtime defer handling for the new register ABI (when
invoking deferred functions on the panic path, the runtime doesn't
need to manage the complexity of determining which args to pass in
register vs memory).
This feature is currently enabled by default if GOEXPERIMENT=regabi or
GOEXPERIMENT=regabidefer is in effect.
Included in this CL are some workarounds in the runtime to insure that
"go" statement targets in the runtime are argument-less already (since
wrapping them can potentially introduce heap-allocated closures, which
are currently not allowed). The expectation is that these workarounds
will be temporary, and can go away once we either A) change the rules
about heap-allocated closures, or B) implement some other scheme for
handling go statements.
Change-Id: I01060d79a6b140c6f0838d6e6813f807ccdca319
Reviewed-on: https://go-review.googlesource.com/c/go/+/298669
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
Windows 10 >= 1607 allows CreateFile and friends to use long paths if
bit 0x80 of the PEB's BitField member is set.
In time this means we'll be able to entirely drop our long path hacks,
which have never really worked right (see bugs below). Until that point,
we'll simply have things working well on recent Windows.
Updates #41734.
Updates #21782.
Updates #36375.
Change-Id: I765de6ea4859dd4e4b8ca80af7f337994734118e
Reviewed-on: https://go-review.googlesource.com/c/go/+/291291
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
runtime.save_g adds X18 to runtime.tls_g in order to have a pointer to
thread local storage. X18 represents a pointer to the TEB on ARM64 and
runtime.tls_g is set in runtime.wintls at initialization time. This
function calls TlsAlloc to allocate a "TLS slot", which is supposed to
index into NtCurrentTeb()->TlsSlots. So the full calculation we want is:
X18 + offsetof(TEB, TlsSlots) + 8*TlsAllocReturnValue
It makes sense to store the complete value of "offsetof(TEB,
TlsSlots) + TlsAllocReturnValue" into runtime.tls_g so that the
calculation can simplify to:
X18 + runtime.tls_g
But, instead of computing that, we're currently doing something kind of
strange, in which we:
- call TlsAlloc, which puts its return value into X0
- make sure X0 is less than 64, so we don't overflow
- set runtime.tls_g to 8*X1 + offsetof(TEB, TlsSlots)
The question is: why are we using X1 instead of X0? What is in X1?
Probably it was, by luck, zero before, and TlsAlloc returned zero, so
there was no problem. But on recent versions of Windows, X1 is some
other garbage value and not zero, so we eventually crash when trying to
dereference X18 + runtime.tls_g.
This commit fixes the problem by just computing:
runtime.tls_g = 8*X0 + offsetof(TEB, TlsSlots)
Fixes#45138.
Change-Id: I560426bae7468217bd183ac6c6eb4b56a3815b09
Reviewed-on: https://go-review.googlesource.com/c/go/+/303273
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Windows amd64 calling convention requires 16-bytes aligned
stack pointer. Before this patch, the real frame size is
0x48 (frame size) + 0x10 (frame pointer & return address),
which does not satisfy the alignment requirement.
_cgo_sys_thread_create eventually calls NtCreateThread,
which receives a pointer to a ThreadContext structure
allocated from (mis-aligned) stack, and may fail with
STATUS_DATATYPE_MISALIGNMENT on some implementations.
BP is saved/restored by prolog/epilog.
AX, CX, DX are volatile, no need to save and restore.
Fixes#41075
Change-Id: I01c0a22b4bf3b4cfdebf4df587445aa46c667973
GitHub-Last-Rev: 15d2bd740e
GitHub-Pull-Request: golang/go#44524
Reviewed-on: https://go-review.googlesource.com/c/go/+/295329
Trust: Alex Brainman <alex.brainman@gmail.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Christopher Nelson <nadiasvertex@gmail.com>
When -clobberdeadreg flag is set, the compiler inserts code that
clobbers integer registers at call sites. This may be helpful for
debugging register ABI.
Only implemented on AMD64 for now.
Change-Id: Ia203d3f891c30fd95d0103489056fe01d63a2899
Reviewed-on: https://go-review.googlesource.com/c/go/+/302809
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
In the runtime there are Windows-specific assembly routines that are
address-taken via funcPC and are not intended to be called through a
wrapper. Mark them as ABIInternal so that we don't grab the wrapper,
because that will break in all sorts of contexts.
For #40724.
For #44065.
Change-Id: I12a728786786f423e5b229f8622e4a80ec27a31c
Reviewed-on: https://go-review.googlesource.com/c/go/+/302109
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
This changes makes it so that nanotimeQPC calls nanotime1 without an ABI
wrapper by specifying the ABIInternal version directly. The reason why
this is necessary is because ABI wrappers typically require additional
stack space, and nanotimeQPC is used deep within nosplit contexts,
and with the ABI wrappers now enabled, this exhausts the stack guard
space held for nosplit functions. Rather than increase the stack guard,
we choose to do this.
For #40724.
Change-Id: Ia9173ca903335a9d6f380f57f4a45e49b58da6bb
Reviewed-on: https://go-review.googlesource.com/c/go/+/303069
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Previously, the runtime had to understand the full syntax of the
GOEXPERIMENT environment variable. Now, sys.GOEXPERIMENT is the
pre-processed experiment list produced by objabi, so we can simplify
the runtime parser.
Change-Id: I0d113a4347dde50a35b8b1f2b0110c88fe802921
Reviewed-on: https://go-review.googlesource.com/c/go/+/303049
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
This moves all remaining GOEXPERIMENT flags into the objabi.Experiment
struct, drops the "_enabled" from their name, and makes them all bool
typed.
We also drop DebugFlags.Fieldtrack because the previous CL shifted the
one test that used it to use GOEXPERIMENT instead.
Change-Id: I3406fe62b1c300bb4caeaffa6ca5ce56a70497fe
Reviewed-on: https://go-review.googlesource.com/c/go/+/302389
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
This patch provides changes according to TODO. Since writeMutex and
writeBlock functions have a lot of code in common, it is better to
move this code to one function.
Change-Id: I81aaad067b0cb1647824909f3b5f6861add3a7ba
Reviewed-on: https://go-review.googlesource.com/c/go/+/280152
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Hyang-Ah Hana Kim <hyangah@gmail.com>
This separates GOEXPERIMENT=regabi into five sub-experiments:
regabiwrappers, regabig, regabireflect, regabidefer, and regabiargs.
Setting GOEXPERIMENT=regabi now implies the working subset of these
(currently, regabiwrappers, regabig, and regabireflect).
This simplifies testing, helps derisk the register ABI project,
and will also help with performance comparisons.
This replaces the -abiwrap flag to the compiler and linker with
the regabiwrappers experiment.
As part of this, regabiargs now enables registers for all calls
in the compiler. Previously, this was statically disabled in
regabiEnabledForAllCompilation, but now that we can control it
independently, this isn't necessary.
For #40724.
Change-Id: I5171e60cda6789031f2ef034cc2e7c5d62459122
Reviewed-on: https://go-review.googlesource.com/c/go/+/302070
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
This CL resurrects the clobberdead debugging mode (CL 23924).
When -clobberdead flag is set (TODO: make it GOEXPERIMENT?), the
compiler inserts code that clobbers all dead stack slots that
contains pointers.
Mark windows syscall functions cgo_unsafe_args, as the code
actually does that, by taking the address of one argument and
passing it to cgocall.
Change-Id: Ie09a015f4bd14ae6053cc707866e30ae509b9d6f
Reviewed-on: https://go-review.googlesource.com/c/go/+/301791
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Austin Clements
Makdon
Changkun Ou
Cherry Zhang
Michael Anthony Knyszek
Ruslan Andreev
Clément Chigot
David Chase
Ian Lance Taylor
Michael Pratt
Lizzzcai
Meng Zhuo
徐志伟
qcrao
Cuong Manh Le
Than McIntosh
Jason A. Donenfeld
Youfu Zhang
Elvina Yakubova