2009-03-31 16:45:12 -06:00
|
|
|
// Copyright 2009 The Go Authors. All rights reserved.
|
|
|
|
|
// Use of this source code is governed by a BSD-style
|
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
|
|
|
|
|
|
// System calls and other sys.stuff for 386, Darwin
|
|
|
|
|
// See http://fxr.watson.org/fxr/source/bsd/kern/syscalls.c?v=xnu-1228
|
|
|
|
|
// or /usr/include/sys/syscall.h (on a Mac) for system call numbers.
|
|
|
|
|
|
[dev.cc] runtime: convert assembly files for C to Go transition
The main change is that #include "zasm_GOOS_GOARCH.h"
is now #include "go_asm.h" and/or #include "go_tls.h".
Also, because C StackGuard is now Go _StackGuard,
the assembly name changes from const_StackGuard to
const__StackGuard.
In asm_$GOARCH.s, add new function getg, formerly
implemented in C.
The renamed atomics now have Go wrappers, to get
escape analysis annotations right. Those wrappers
are in CL 174860043.
LGTM=r, aram
R=r, aram
CC=austin, dvyukov, golang-codereviews, iant, khr
https://golang.org/cl/168510043
2014-11-11 15:06:22 -07:00
|
|
|
#include "go_asm.h"
|
|
|
|
|
#include "go_tls.h"
|
2014-09-04 21:05:18 -06:00
|
|
|
#include "textflag.h"
|
2009-06-17 16:15:55 -06:00
|
|
|
|
2009-03-31 16:45:12 -06:00
|
|
|
// Exit the entire program (like C exit)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·exit(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $1, AX
|
|
|
|
|
INT $0x80
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Exit this OS thread (like pthread_exit, which eventually
|
|
|
|
|
// calls __bsdthread_terminate).
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·exit1(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $361, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·open(SB),NOSPLIT,$0
|
2013-03-12 11:47:44 -06:00
|
|
|
MOVL $5, AX
|
|
|
|
|
INT $0x80
|
2015-03-02 21:16:48 -07:00
|
|
|
JAE 2(PC)
|
|
|
|
|
MOVL $-1, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+12(FP)
|
2013-03-12 11:47:44 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·close(SB),NOSPLIT,$0
|
2013-03-12 11:47:44 -06:00
|
|
|
MOVL $6, AX
|
|
|
|
|
INT $0x80
|
2015-03-02 21:16:48 -07:00
|
|
|
JAE 2(PC)
|
|
|
|
|
MOVL $-1, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+4(FP)
|
2013-03-12 11:47:44 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·read(SB),NOSPLIT,$0
|
2013-03-12 11:47:44 -06:00
|
|
|
MOVL $3, AX
|
|
|
|
|
INT $0x80
|
2015-03-02 21:16:48 -07:00
|
|
|
JAE 2(PC)
|
|
|
|
|
MOVL $-1, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+12(FP)
|
2013-03-12 11:47:44 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·write(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $4, AX
|
|
|
|
|
INT $0x80
|
2015-03-02 21:16:48 -07:00
|
|
|
JAE 2(PC)
|
|
|
|
|
MOVL $-1, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+12(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2015-01-14 09:18:24 -07:00
|
|
|
TEXT runtime·raise(SB),NOSPLIT,$0
|
|
|
|
|
// Ideally we'd send the signal to the current thread,
|
|
|
|
|
// not the whole process, but that's too hard on OS X.
|
|
|
|
|
JMP runtime·raiseproc(SB)
|
|
|
|
|
|
|
|
|
|
TEXT runtime·raiseproc(SB),NOSPLIT,$16
|
2013-03-14 23:11:03 -06:00
|
|
|
MOVL $20, AX // getpid
|
|
|
|
|
INT $0x80
|
|
|
|
|
MOVL AX, 4(SP) // pid
|
|
|
|
|
MOVL sig+0(FP), AX
|
|
|
|
|
MOVL AX, 8(SP) // signal
|
|
|
|
|
MOVL $1, 12(SP) // posix
|
|
|
|
|
MOVL $37, AX // kill
|
2011-04-25 14:58:00 -06:00
|
|
|
INT $0x80
|
|
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mmap(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $197, AX
|
|
|
|
|
INT $0x80
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+24(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·madvise(SB),NOSPLIT,$0
|
2012-01-19 13:51:29 -07:00
|
|
|
MOVL $75, AX
|
|
|
|
|
INT $0x80
|
2012-12-22 13:06:28 -07:00
|
|
|
// ignore failure - maybe pages are locked
|
2012-01-19 13:51:29 -07:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·munmap(SB),NOSPLIT,$0
|
2010-09-07 07:57:22 -06:00
|
|
|
MOVL $73, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2010-09-07 07:57:22 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·setitimer(SB),NOSPLIT,$0
|
2011-03-23 09:31:42 -06:00
|
|
|
MOVL $83, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
RET
|
|
|
|
|
|
2012-06-05 14:23:30 -06:00
|
|
|
// OS X comm page time offsets
|
|
|
|
|
// http://www.opensource.apple.com/source/xnu/xnu-1699.26.8/osfmk/i386/cpu_capabilities.h
|
|
|
|
|
#define cpu_capabilities 0x20
|
|
|
|
|
#define nt_tsc_base 0x50
|
|
|
|
|
#define nt_scale 0x58
|
|
|
|
|
#define nt_shift 0x5c
|
|
|
|
|
#define nt_ns_base 0x60
|
|
|
|
|
#define nt_generation 0x68
|
|
|
|
|
#define gtod_generation 0x6c
|
|
|
|
|
#define gtod_ns_base 0x70
|
|
|
|
|
#define gtod_sec_base 0x78
|
|
|
|
|
|
|
|
|
|
// called from assembly
|
|
|
|
|
// 64-bit unix nanoseconds returned in DX:AX.
|
|
|
|
|
// I'd much rather write this in C but we need
|
|
|
|
|
// assembly for the 96-bit multiply and RDTSC.
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·now(SB),NOSPLIT,$40
|
2012-06-05 14:23:30 -06:00
|
|
|
MOVL $0xffff0000, BP /* comm page base */
|
|
|
|
|
|
|
|
|
|
// Test for slow CPU. If so, the math is completely
|
|
|
|
|
// different, and unimplemented here, so use the
|
|
|
|
|
// system call.
|
|
|
|
|
MOVL cpu_capabilities(BP), AX
|
|
|
|
|
TESTL $0x4000, AX
|
|
|
|
|
JNZ systime
|
|
|
|
|
|
|
|
|
|
// Loop trying to take a consistent snapshot
|
|
|
|
|
// of the time parameters.
|
|
|
|
|
timeloop:
|
|
|
|
|
MOVL gtod_generation(BP), BX
|
|
|
|
|
TESTL BX, BX
|
|
|
|
|
JZ systime
|
|
|
|
|
MOVL nt_generation(BP), CX
|
|
|
|
|
TESTL CX, CX
|
|
|
|
|
JZ timeloop
|
|
|
|
|
RDTSC
|
|
|
|
|
MOVL nt_tsc_base(BP), SI
|
|
|
|
|
MOVL (nt_tsc_base+4)(BP), DI
|
|
|
|
|
MOVL SI, 0(SP)
|
|
|
|
|
MOVL DI, 4(SP)
|
|
|
|
|
MOVL nt_scale(BP), SI
|
|
|
|
|
MOVL SI, 8(SP)
|
|
|
|
|
MOVL nt_ns_base(BP), SI
|
|
|
|
|
MOVL (nt_ns_base+4)(BP), DI
|
|
|
|
|
MOVL SI, 12(SP)
|
|
|
|
|
MOVL DI, 16(SP)
|
|
|
|
|
CMPL nt_generation(BP), CX
|
|
|
|
|
JNE timeloop
|
|
|
|
|
MOVL gtod_ns_base(BP), SI
|
|
|
|
|
MOVL (gtod_ns_base+4)(BP), DI
|
|
|
|
|
MOVL SI, 20(SP)
|
|
|
|
|
MOVL DI, 24(SP)
|
|
|
|
|
MOVL gtod_sec_base(BP), SI
|
|
|
|
|
MOVL (gtod_sec_base+4)(BP), DI
|
|
|
|
|
MOVL SI, 28(SP)
|
|
|
|
|
MOVL DI, 32(SP)
|
|
|
|
|
CMPL gtod_generation(BP), BX
|
|
|
|
|
JNE timeloop
|
|
|
|
|
|
|
|
|
|
// Gathered all the data we need. Compute time.
|
|
|
|
|
// ((tsc - nt_tsc_base) * nt_scale) >> 32 + nt_ns_base - gtod_ns_base + gtod_sec_base*1e9
|
|
|
|
|
// The multiply and shift extracts the top 64 bits of the 96-bit product.
|
|
|
|
|
SUBL 0(SP), AX // DX:AX = (tsc - nt_tsc_base)
|
|
|
|
|
SBBL 4(SP), DX
|
|
|
|
|
|
|
|
|
|
// We have x = tsc - nt_tsc_base - DX:AX to be
|
|
|
|
|
// multiplied by y = nt_scale = 8(SP), keeping the top 64 bits of the 96-bit product.
|
|
|
|
|
// x*y = (x&0xffffffff)*y + (x&0xffffffff00000000)*y
|
|
|
|
|
// (x*y)>>32 = ((x&0xffffffff)*y)>>32 + (x>>32)*y
|
|
|
|
|
MOVL DX, CX // SI = (x&0xffffffff)*y >> 32
|
|
|
|
|
MOVL $0, DX
|
|
|
|
|
MULL 8(SP)
|
|
|
|
|
MOVL DX, SI
|
|
|
|
|
|
|
|
|
|
MOVL CX, AX // DX:AX = (x>>32)*y
|
|
|
|
|
MOVL $0, DX
|
|
|
|
|
MULL 8(SP)
|
|
|
|
|
|
|
|
|
|
ADDL SI, AX // DX:AX += (x&0xffffffff)*y >> 32
|
|
|
|
|
ADCL $0, DX
|
|
|
|
|
|
|
|
|
|
// DX:AX is now ((tsc - nt_tsc_base) * nt_scale) >> 32.
|
|
|
|
|
ADDL 12(SP), AX // DX:AX += nt_ns_base
|
|
|
|
|
ADCL 16(SP), DX
|
|
|
|
|
SUBL 20(SP), AX // DX:AX -= gtod_ns_base
|
|
|
|
|
SBBL 24(SP), DX
|
|
|
|
|
MOVL AX, SI // DI:SI = DX:AX
|
|
|
|
|
MOVL DX, DI
|
|
|
|
|
MOVL 28(SP), AX // DX:AX = gtod_sec_base*1e9
|
|
|
|
|
MOVL 32(SP), DX
|
|
|
|
|
MOVL $1000000000, CX
|
|
|
|
|
MULL CX
|
|
|
|
|
ADDL SI, AX // DX:AX += DI:SI
|
|
|
|
|
ADCL DI, DX
|
2011-11-30 09:59:44 -07:00
|
|
|
RET
|
|
|
|
|
|
2012-06-05 14:23:30 -06:00
|
|
|
systime:
|
|
|
|
|
// Fall back to system call (usually first call in this thread)
|
2010-02-08 15:32:22 -07:00
|
|
|
LEAL 12(SP), AX // must be non-nil, unused
|
|
|
|
|
MOVL AX, 4(SP)
|
|
|
|
|
MOVL $0, 8(SP) // time zone pointer
|
|
|
|
|
MOVL $116, AX
|
|
|
|
|
INT $0x80
|
2012-06-05 14:23:30 -06:00
|
|
|
// sec is in AX, usec in DX
|
2011-11-03 15:35:28 -06:00
|
|
|
// convert to DX:AX nsec
|
2012-06-05 14:23:30 -06:00
|
|
|
MOVL DX, BX
|
2011-11-03 15:35:28 -06:00
|
|
|
MOVL $1000000000, CX
|
|
|
|
|
MULL CX
|
|
|
|
|
IMULL $1000, BX
|
|
|
|
|
ADDL BX, AX
|
|
|
|
|
ADCL $0, DX
|
2012-06-05 14:23:30 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// func now() (sec int64, nsec int32)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT time·now(SB),NOSPLIT,$0
|
2012-06-05 14:23:30 -06:00
|
|
|
CALL runtime·now(SB)
|
|
|
|
|
MOVL $1000000000, CX
|
|
|
|
|
DIVL CX
|
|
|
|
|
MOVL AX, sec+0(FP)
|
|
|
|
|
MOVL $0, sec+4(FP)
|
|
|
|
|
MOVL DX, nsec+8(FP)
|
|
|
|
|
RET
|
2012-02-28 14:18:24 -07:00
|
|
|
|
2012-06-05 14:23:30 -06:00
|
|
|
// int64 nanotime(void) so really
|
|
|
|
|
// void nanotime(int64 *nsec)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·nanotime(SB),NOSPLIT,$0
|
2012-06-05 14:23:30 -06:00
|
|
|
CALL runtime·now(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret_lo+0(FP)
|
|
|
|
|
MOVL DX, ret_hi+4(FP)
|
2010-02-08 15:32:22 -07:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·sigprocmask(SB),NOSPLIT,$0
|
2012-02-28 14:18:24 -07:00
|
|
|
MOVL $329, AX // pthread_sigmask (on OS X, sigprocmask==entire process)
|
2012-02-23 12:44:06 -07:00
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2012-02-23 12:44:06 -07:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·sigaction(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $46, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Sigtramp's job is to call the actual signal handler.
|
|
|
|
|
// It is called with the following arguments on the stack:
|
|
|
|
|
// 0(FP) "return address" - ignored
|
|
|
|
|
// 4(FP) actual handler
|
2012-09-04 12:40:49 -06:00
|
|
|
// 8(FP) signal number
|
|
|
|
|
// 12(FP) siginfo style
|
2009-03-31 16:45:12 -06:00
|
|
|
// 16(FP) siginfo
|
|
|
|
|
// 20(FP) context
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·sigtramp(SB),NOSPLIT,$40
|
2010-01-06 18:58:55 -07:00
|
|
|
get_tls(CX)
|
2012-03-12 13:55:18 -06:00
|
|
|
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
// check that g exists
|
|
|
|
|
MOVL g(CX), DI
|
|
|
|
|
CMPL DI, $0
|
2013-07-11 14:39:39 -06:00
|
|
|
JNE 6(PC)
|
2012-09-04 12:40:49 -06:00
|
|
|
MOVL sig+8(FP), BX
|
|
|
|
|
MOVL BX, 0(SP)
|
2013-07-11 14:39:39 -06:00
|
|
|
MOVL $runtime·badsignal(SB), AX
|
|
|
|
|
CALL AX
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
JMP ret
|
2010-06-12 11:48:04 -06:00
|
|
|
|
|
|
|
|
// save g
|
2011-02-23 12:47:42 -07:00
|
|
|
MOVL DI, 20(SP)
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
|
2010-06-12 11:48:04 -06:00
|
|
|
// g = m->gsignal
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVL g_m(DI), BP
|
2009-06-17 16:15:55 -06:00
|
|
|
MOVL m_gsignal(BP), BP
|
2010-01-06 18:58:55 -07:00
|
|
|
MOVL BP, g(CX)
|
2009-03-31 16:45:12 -06:00
|
|
|
|
2011-02-23 12:47:42 -07:00
|
|
|
// copy arguments to sighandler
|
|
|
|
|
MOVL sig+8(FP), BX
|
|
|
|
|
MOVL BX, 0(SP)
|
|
|
|
|
MOVL info+12(FP), BX
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL BX, 4(SP)
|
2011-02-23 12:47:42 -07:00
|
|
|
MOVL context+16(FP), BX
|
|
|
|
|
MOVL BX, 8(SP)
|
|
|
|
|
MOVL DI, 12(SP)
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
|
2011-02-23 12:47:42 -07:00
|
|
|
MOVL handler+0(FP), BX
|
|
|
|
|
CALL BX
|
2009-03-31 16:45:12 -06:00
|
|
|
|
2010-06-12 11:48:04 -06:00
|
|
|
// restore g
|
2010-01-06 18:58:55 -07:00
|
|
|
get_tls(CX)
|
2011-02-23 12:47:42 -07:00
|
|
|
MOVL 20(SP), DI
|
|
|
|
|
MOVL DI, g(CX)
|
2009-11-16 18:51:47 -07:00
|
|
|
|
[dev.power64] cmd/5a, cmd/6a, cmd/8a, cmd/9a: make labels function-scoped
I removed support for jumping between functions years ago,
as part of doing the instruction layout for each function separately.
Given that, it makes sense to treat labels as function-scoped.
This lets each function have its own 'loop' label, for example.
Makes the assembly much cleaner and removes the last
reason anyone would reach for the 123(PC) form instead.
Note that this is on the dev.power64 branch, but it changes all
the assemblers. The change will ship in Go 1.5 (perhaps after
being ported into the new assembler).
Came up as part of CL 167730043.
LGTM=r
R=r
CC=austin, dave, golang-codereviews, minux
https://golang.org/cl/159670043
2014-10-28 19:50:16 -06:00
|
|
|
ret:
|
2011-02-23 12:47:42 -07:00
|
|
|
// call sigreturn
|
2009-11-16 18:51:47 -07:00
|
|
|
MOVL context+16(FP), CX
|
|
|
|
|
MOVL style+4(FP), BX
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $0, 0(SP) // "caller PC" - ignored
|
|
|
|
|
MOVL CX, 4(SP)
|
|
|
|
|
MOVL BX, 8(SP)
|
|
|
|
|
MOVL $184, AX // sigreturn(ucontext, infostyle)
|
|
|
|
|
INT $0x80
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·sigaltstack(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $53, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
2012-03-08 12:03:56 -07:00
|
|
|
MOVL $0xf1, 0xf1 // crash
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·usleep(SB),NOSPLIT,$32
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
MOVL $0, DX
|
|
|
|
|
MOVL usec+0(FP), AX
|
|
|
|
|
MOVL $1000000, CX
|
|
|
|
|
DIVL CX
|
|
|
|
|
MOVL AX, 24(SP) // sec
|
|
|
|
|
MOVL DX, 28(SP) // usec
|
|
|
|
|
|
|
|
|
|
// select(0, 0, 0, 0, &tv)
|
|
|
|
|
MOVL $0, 0(SP) // "return PC" - ignored
|
|
|
|
|
MOVL $0, 4(SP)
|
|
|
|
|
MOVL $0, 8(SP)
|
|
|
|
|
MOVL $0, 12(SP)
|
|
|
|
|
MOVL $0, 16(SP)
|
|
|
|
|
LEAL 24(SP), AX
|
|
|
|
|
MOVL AX, 20(SP)
|
|
|
|
|
MOVL $93, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
RET
|
|
|
|
|
|
2012-12-18 09:30:29 -07:00
|
|
|
// void bsdthread_create(void *stk, M *mp, G *gp, void (*fn)(void))
|
2009-06-04 12:16:03 -06:00
|
|
|
// System call args are: func arg stack pthread flags.
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·bsdthread_create(SB),NOSPLIT,$32
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $360, AX
|
2009-06-04 12:16:03 -06:00
|
|
|
// 0(SP) is where the caller PC would be; kernel skips it
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL fn+12(FP), BX
|
2009-06-04 12:16:03 -06:00
|
|
|
MOVL BX, 4(SP) // func
|
2009-06-17 16:15:55 -06:00
|
|
|
MOVL mm+4(FP), BX
|
2009-06-04 12:16:03 -06:00
|
|
|
MOVL BX, 8(SP) // arg
|
|
|
|
|
MOVL stk+0(FP), BX
|
|
|
|
|
MOVL BX, 12(SP) // stack
|
2009-06-17 16:15:55 -06:00
|
|
|
MOVL gg+8(FP), BX
|
2009-06-04 12:16:03 -06:00
|
|
|
MOVL BX, 16(SP) // pthread
|
|
|
|
|
MOVL $0x1000000, 20(SP) // flags = PTHREAD_START_CUSTOM
|
2009-03-31 16:45:12 -06:00
|
|
|
INT $0x80
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
JAE 4(PC)
|
2011-01-28 13:03:26 -07:00
|
|
|
NEGL AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+16(FP)
|
2010-04-28 20:36:42 -06:00
|
|
|
RET
|
|
|
|
|
MOVL $0, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+16(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2009-06-04 12:16:03 -06:00
|
|
|
// The thread that bsdthread_create creates starts executing here,
|
|
|
|
|
// because we registered this function using bsdthread_register
|
|
|
|
|
// at startup.
|
|
|
|
|
// AX = "pthread" (= g)
|
|
|
|
|
// BX = mach thread port
|
|
|
|
|
// CX = "func" (= fn)
|
|
|
|
|
// DX = "arg" (= m)
|
|
|
|
|
// DI = stack top
|
|
|
|
|
// SI = flags (= 0x1000000)
|
|
|
|
|
// SP = stack - C_32_STK_ALIGN
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·bsdthread_start(SB),NOSPLIT,$0
|
2009-06-04 12:16:03 -06:00
|
|
|
// set up ldt 7+id to point at m->tls.
|
|
|
|
|
// m->tls is at m+40. newosproc left
|
|
|
|
|
// the m->id in tls[0].
|
2009-06-17 16:15:55 -06:00
|
|
|
LEAL m_tls(DX), BP
|
2009-06-04 12:16:03 -06:00
|
|
|
MOVL 0(BP), DI
|
|
|
|
|
ADDL $7, DI // m0 is LDT#7. count up.
|
|
|
|
|
// setldt(tls#, &tls, sizeof tls)
|
|
|
|
|
PUSHAL // save registers
|
|
|
|
|
PUSHL $32 // sizeof tls
|
|
|
|
|
PUSHL BP // &tls
|
|
|
|
|
PUSHL DI // tls #
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·setldt(SB)
|
2009-06-04 12:16:03 -06:00
|
|
|
POPL AX
|
|
|
|
|
POPL AX
|
|
|
|
|
POPL AX
|
|
|
|
|
POPAL
|
|
|
|
|
|
|
|
|
|
// Now segment is established. Initialize m, g.
|
2010-01-06 18:58:55 -07:00
|
|
|
get_tls(BP)
|
|
|
|
|
MOVL AX, g(BP)
|
all: remove 'extern register M *m' from runtime
The runtime has historically held two dedicated values g (current goroutine)
and m (current thread) in 'extern register' slots (TLS on x86, real registers
backed by TLS on ARM).
This CL removes the extern register m; code now uses g->m.
On ARM, this frees up the register that formerly held m (R9).
This is important for NaCl, because NaCl ARM code cannot use R9 at all.
The Go 1 macrobenchmarks (those with per-op times >= 10 µs) are unaffected:
BenchmarkBinaryTree17 5491374955 5471024381 -0.37%
BenchmarkFannkuch11 4357101311 4275174828 -1.88%
BenchmarkGobDecode 11029957 11364184 +3.03%
BenchmarkGobEncode 6852205 6784822 -0.98%
BenchmarkGzip 650795967 650152275 -0.10%
BenchmarkGunzip 140962363 141041670 +0.06%
BenchmarkHTTPClientServer 71581 73081 +2.10%
BenchmarkJSONEncode 31928079 31913356 -0.05%
BenchmarkJSONDecode 117470065 113689916 -3.22%
BenchmarkMandelbrot200 6008923 5998712 -0.17%
BenchmarkGoParse 6310917 6327487 +0.26%
BenchmarkRegexpMatchMedium_1K 114568 114763 +0.17%
BenchmarkRegexpMatchHard_1K 168977 169244 +0.16%
BenchmarkRevcomp 935294971 914060918 -2.27%
BenchmarkTemplate 145917123 148186096 +1.55%
Minux previous reported larger variations, but these were caused by
run-to-run noise, not repeatable slowdowns.
Actual code changes by Minux.
I only did the docs and the benchmarking.
LGTM=dvyukov, iant, minux
R=minux, josharian, iant, dave, bradfitz, dvyukov
CC=golang-codereviews
https://golang.org/cl/109050043
2014-06-26 09:54:39 -06:00
|
|
|
MOVL DX, g_m(AX)
|
2009-06-17 16:15:55 -06:00
|
|
|
MOVL BX, m_procid(DX) // m->procid = thread port (for debuggers)
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·stackcheck(SB) // smashes AX
|
2009-06-04 12:16:03 -06:00
|
|
|
CALL CX // fn()
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·exit1(SB)
|
2009-04-02 17:48:06 -06:00
|
|
|
RET
|
|
|
|
|
|
2009-06-04 12:16:03 -06:00
|
|
|
// void bsdthread_register(void)
|
|
|
|
|
// registers callbacks for threadstart (see bsdthread_create above
|
|
|
|
|
// and wqthread and pthsize (not used). returns 0 on success.
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·bsdthread_register(SB),NOSPLIT,$40
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $366, AX
|
2009-06-04 12:16:03 -06:00
|
|
|
// 0(SP) is where kernel expects caller PC; ignored
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
MOVL $runtime·bsdthread_start(SB), 4(SP) // threadstart
|
2009-06-04 12:16:03 -06:00
|
|
|
MOVL $0, 8(SP) // wqthread, not used by us
|
|
|
|
|
MOVL $0, 12(SP) // pthsize, not used by us
|
runtime: work around kernel bug in Snow Leopard signal handling
Could not take a signal on threads other than the main thread.
If you look at the spinning binary with dtrace, you can see a
fault happening over and over:
$ dtrace -n '
fbt::user_trap:entry /execname=="boot32" && self->count < 10/
{
self->count++;
printf("%s %x %x %x %x", probefunc, arg1, arg2, arg3, arg4);
stack();
tracemem(arg4, 256);
}'
dtrace: description 'fbt::user_trap:entry ' matched 1 probe
CPU ID FUNCTION:NAME
1 17015 user_trap:entry user_trap 0 10 79af0a0 79af0a0
mach_kernel`lo_alltraps+0x12a
0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef
0: 0e 00 00 00 37 00 00 00 00 00 00 00 1f 00 00 00 ....7...........
10: 1f 00 00 00 a8 33 00 00 00 00 00 01 00 00 00 00 .....3..........
20: 98 ba dc fe 07 09 00 00 00 00 00 00 98 ba dc fe ................
30: 06 00 00 00 0d 00 00 00 34 00 00 00 9e 1c 00 00 ........4.......
40: 17 00 00 00 00 02 00 00 ac 30 00 00 1f 00 00 00 .........0......
50: 00 00 00 00 00 00 00 00 0d 00 00 00 e0 e6 29 00 ..............).
60: 34 00 00 00 00 00 00 00 9e 1c 00 00 00 00 00 00 4...............
70: 17 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 ................
80: ac 30 00 00 00 00 00 00 1f 00 00 00 00 00 00 00 .0..............
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
a0: 48 00 00 00 10 00 00 00 85 00 00 00 a0 f2 29 00 H.............).
b0: 69 01 00 02 00 00 00 00 e6 93 04 82 ff 7f 00 00 i...............
c0: 2f 00 00 00 00 00 00 00 06 02 00 00 00 00 00 00 /...............
d0: 78 ee 42 01 01 00 00 00 1f 00 00 00 00 00 00 00 x.B.............
e0: 00 ed 9a 07 00 00 00 00 00 00 00 00 00 00 00 00 ................
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
...
The memory dump shows a 32-bit exception frame:
x86_saved_state32
gs = 0x37
fs = 0
es = 0x1f
ds = 0x1f
edi = 0x33a8
esi = 0x01000000
ebp = 0
cr2 = 0xfedcba98
ebx = 0x0907
edx = 0
ecx = 0xfedcba98
eax = 0x06
trapno = 0x0d
err = 0x34
eip = 0x1c9e
cs = 0x17
efl = 0x0200
uesp = 0x30ac
ss = 0x1f
The cr2 of 0xfedcba98 is the address that the new thread read
to cause the fault, but note that the trap is now a GP fault with
error code 0x34, meaning it's moved past the cr2 problem and on
to an invaild segment selector. The 0x34 is suspiciously similar
to the 0x37 in gs, and sure enough, OS X forces gs to have
that value in the signal handler, and if your thread hasn't set
up that segment (known as USER_CTHREAD), you'll fault on the IRET
into the signal handler and never be able to handle a signal.
The kernel bug is that it forces segment 0x37 without making sure
it is a valid segment. Leopard also forced 0x37 but had the courtesy
to set it up first.
Since OS X requires us to set up that segment (using the
thread_fast_set_cthread_self system call), we might as well
use it instead of the more complicated i386_set_ldt call to
set up our per-OS thread storage.
Also add some more zeros to bsdthread_register for new arguments
in Snow Leopard (apparently unnecessary, but being careful).
Fixes #510.
R=r
CC=golang-dev
https://golang.org/cl/824046
2010-04-08 14:24:37 -06:00
|
|
|
MOVL $0, 16(SP) // dummy_value [sic]
|
|
|
|
|
MOVL $0, 20(SP) // targetconc_ptr
|
|
|
|
|
MOVL $0, 24(SP) // dispatchqueue_offset
|
2009-03-31 16:45:12 -06:00
|
|
|
INT $0x80
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
JAE 4(PC)
|
2012-04-30 13:55:07 -06:00
|
|
|
NEGL AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+0(FP)
|
2012-04-30 13:55:07 -06:00
|
|
|
RET
|
|
|
|
|
MOVL $0, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+0(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Invoke Mach system call.
|
|
|
|
|
// Assumes system call number in AX,
|
|
|
|
|
// caller PC on stack, caller's caller PC next,
|
|
|
|
|
// and then the system call arguments.
|
|
|
|
|
//
|
|
|
|
|
// Can be used for BSD too, but we don't,
|
|
|
|
|
// because if you use this interface the BSD
|
|
|
|
|
// system call numbers need an extra field
|
|
|
|
|
// in the high 16 bits that seems to be the
|
|
|
|
|
// argument count in bytes but is not always.
|
|
|
|
|
// INT $0x80 works fine for those.
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·sysenter(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
POPL DX
|
|
|
|
|
MOVL SP, CX
|
|
|
|
|
BYTE $0x0F; BYTE $0x34; // SYSENTER
|
|
|
|
|
// returns to DX with SP set to CX
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_msg_trap(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-31, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+28(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_reply_port(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-26, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+0(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_task_self(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-28, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+0(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// Mach provides trap versions of the semaphore ops,
|
|
|
|
|
// instead of requiring the use of RPC.
|
|
|
|
|
|
|
|
|
|
// uint32 mach_semaphore_wait(uint32)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_semaphore_wait(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-36, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+4(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// uint32 mach_semaphore_timedwait(uint32, uint32, uint32)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_semaphore_timedwait(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-38, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+12(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// uint32 mach_semaphore_signal(uint32)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_semaphore_signal(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-33, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+4(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// uint32 mach_semaphore_signal_all(uint32)
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·mach_semaphore_signal_all(SB),NOSPLIT,$0
|
2009-03-31 16:45:12 -06:00
|
|
|
MOVL $-34, AX
|
runtime: ,s/[a-zA-Z0-9_]+/runtime·&/g, almost
Prefix all external symbols in runtime by runtime·,
to avoid conflicts with possible symbols of the same
name in linked-in C libraries. The obvious conflicts
are printf, malloc, and free, but hide everything to
avoid future pain.
The symbols left alone are:
** known to cgo **
_cgo_free
_cgo_malloc
libcgo_thread_start
initcgo
ncgocall
** known to linker **
_rt0_$GOARCH
_rt0_$GOARCH_$GOOS
text
etext
data
end
pclntab
epclntab
symtab
esymtab
** known to C compiler **
_divv
_modv
_div64by32
etc (arch specific)
Tested on darwin/386, darwin/amd64, linux/386, linux/amd64.
Built (but not tested) for freebsd/386, freebsd/amd64, linux/arm, windows/386.
R=r, PeterGo
CC=golang-dev
https://golang.org/cl/2899041
2010-11-04 12:00:19 -06:00
|
|
|
CALL runtime·sysenter(SB)
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+4(FP)
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// setldt(int entry, int address, int limit)
|
runtime: work around kernel bug in Snow Leopard signal handling
Could not take a signal on threads other than the main thread.
If you look at the spinning binary with dtrace, you can see a
fault happening over and over:
$ dtrace -n '
fbt::user_trap:entry /execname=="boot32" && self->count < 10/
{
self->count++;
printf("%s %x %x %x %x", probefunc, arg1, arg2, arg3, arg4);
stack();
tracemem(arg4, 256);
}'
dtrace: description 'fbt::user_trap:entry ' matched 1 probe
CPU ID FUNCTION:NAME
1 17015 user_trap:entry user_trap 0 10 79af0a0 79af0a0
mach_kernel`lo_alltraps+0x12a
0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef
0: 0e 00 00 00 37 00 00 00 00 00 00 00 1f 00 00 00 ....7...........
10: 1f 00 00 00 a8 33 00 00 00 00 00 01 00 00 00 00 .....3..........
20: 98 ba dc fe 07 09 00 00 00 00 00 00 98 ba dc fe ................
30: 06 00 00 00 0d 00 00 00 34 00 00 00 9e 1c 00 00 ........4.......
40: 17 00 00 00 00 02 00 00 ac 30 00 00 1f 00 00 00 .........0......
50: 00 00 00 00 00 00 00 00 0d 00 00 00 e0 e6 29 00 ..............).
60: 34 00 00 00 00 00 00 00 9e 1c 00 00 00 00 00 00 4...............
70: 17 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 ................
80: ac 30 00 00 00 00 00 00 1f 00 00 00 00 00 00 00 .0..............
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
a0: 48 00 00 00 10 00 00 00 85 00 00 00 a0 f2 29 00 H.............).
b0: 69 01 00 02 00 00 00 00 e6 93 04 82 ff 7f 00 00 i...............
c0: 2f 00 00 00 00 00 00 00 06 02 00 00 00 00 00 00 /...............
d0: 78 ee 42 01 01 00 00 00 1f 00 00 00 00 00 00 00 x.B.............
e0: 00 ed 9a 07 00 00 00 00 00 00 00 00 00 00 00 00 ................
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
...
The memory dump shows a 32-bit exception frame:
x86_saved_state32
gs = 0x37
fs = 0
es = 0x1f
ds = 0x1f
edi = 0x33a8
esi = 0x01000000
ebp = 0
cr2 = 0xfedcba98
ebx = 0x0907
edx = 0
ecx = 0xfedcba98
eax = 0x06
trapno = 0x0d
err = 0x34
eip = 0x1c9e
cs = 0x17
efl = 0x0200
uesp = 0x30ac
ss = 0x1f
The cr2 of 0xfedcba98 is the address that the new thread read
to cause the fault, but note that the trap is now a GP fault with
error code 0x34, meaning it's moved past the cr2 problem and on
to an invaild segment selector. The 0x34 is suspiciously similar
to the 0x37 in gs, and sure enough, OS X forces gs to have
that value in the signal handler, and if your thread hasn't set
up that segment (known as USER_CTHREAD), you'll fault on the IRET
into the signal handler and never be able to handle a signal.
The kernel bug is that it forces segment 0x37 without making sure
it is a valid segment. Leopard also forced 0x37 but had the courtesy
to set it up first.
Since OS X requires us to set up that segment (using the
thread_fast_set_cthread_self system call), we might as well
use it instead of the more complicated i386_set_ldt call to
set up our per-OS thread storage.
Also add some more zeros to bsdthread_register for new arguments
in Snow Leopard (apparently unnecessary, but being careful).
Fixes #510.
R=r
CC=golang-dev
https://golang.org/cl/824046
2010-04-08 14:24:37 -06:00
|
|
|
// entry and limit are ignored.
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·setldt(SB),NOSPLIT,$32
|
2009-10-03 11:37:12 -06:00
|
|
|
MOVL address+4(FP), BX // aka base
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* When linking against the system libraries,
|
|
|
|
|
* we use its pthread_create and let it set up %gs
|
|
|
|
|
* for us. When we do that, the private storage
|
|
|
|
|
* we get is not at 0(GS) but at 0x468(GS).
|
liblink: introduce TLS register on 386 and amd64
When I did the original 386 ports on Linux and OS X, I chose to
define GS-relative expressions like 4(GS) as relative to the actual
thread-local storage base, which was usually GS but might not be
(it might be FS, or it might be a different constant offset from GS or FS).
The original scope was limited but since then the rewrites have
gotten out of control. Sometimes GS is rewritten, sometimes FS.
Some ports do other rewrites to enable shared libraries and
other linking. At no point in the code is it clear whether you are
looking at the real GS/FS or some synthesized thing that will be
rewritten. The code manipulating all these is duplicated in many
places.
The first step to fixing issue 7719 is to make the code intelligible
again.
This CL adds an explicit TLS pseudo-register to the 386 and amd64.
As a register, TLS refers to the thread-local storage base, and it
can only be loaded into another register:
MOVQ TLS, AX
An offset from the thread-local storage base is written off(reg)(TLS*1).
Semantically it is off(reg), but the (TLS*1) annotation marks this as
indexing from the loaded TLS base. This emits a relocation so that
if the linker needs to adjust the offset, it can. For example:
MOVQ TLS, AX
MOVQ 8(AX)(TLS*1), CX // load m into CX
On systems that support direct access to the TLS memory, this
pair of instructions can be reduced to a direct TLS memory reference:
MOVQ 8(TLS), CX // load m into CX
The 2-instruction and 1-instruction forms correspond roughly to
ELF TLS initial exec mode and ELF TLS local exec mode, respectively.
Liblink applies this rewrite on systems that support the 1-instruction form.
The decision is made using only the operating system (and probably
the -shared flag, eventually), not the link mode. If some link modes
on a particular operating system require the 2-instruction form,
then all builds for that operating system will use the 2-instruction
form, so that the link mode decision can be delayed to link time.
Obviously it is late to be making changes like this, but I despair
of correcting issue 7719 and issue 7164 without it. To make sure
I am not changing existing behavior, I built a "hello world" program
for every GOOS/GOARCH combination we have and then worked
to make sure that the rewrite generates exactly the same binaries,
byte for byte. There are a handful of TODOs in the code marking
kludges to get the byte-for-byte property, but at least now I can
explain exactly how each binary is handled.
The targets I tested this way are:
darwin-386
darwin-amd64
dragonfly-386
dragonfly-amd64
freebsd-386
freebsd-amd64
freebsd-arm
linux-386
linux-amd64
linux-arm
nacl-386
nacl-amd64p32
netbsd-386
netbsd-amd64
openbsd-386
openbsd-amd64
plan9-386
plan9-amd64
solaris-amd64
windows-386
windows-amd64
There were four exceptions to the byte-for-byte goal:
windows-386 and windows-amd64 have a time stamp
at bytes 137 and 138 of the header.
darwin-386 and plan9-386 have five or six modified
bytes in the middle of the Go symbol table, caused by
editing comments in runtime/sys_{darwin,plan9}_386.s.
Fixes #7164.
LGTM=iant
R=iant, aram, minux.ma, dave
CC=golang-codereviews
https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
|
|
|
* 8l rewrites 0(TLS) into 0x468(GS) for us.
|
2009-10-03 11:37:12 -06:00
|
|
|
* To accommodate that rewrite, we translate the
|
|
|
|
|
* address and limit here so that 0x468(GS) maps to 0(address).
|
|
|
|
|
*
|
2012-01-19 15:13:33 -07:00
|
|
|
* See cgo/gcc_darwin_386.c:/468 for the derivation
|
2009-10-03 11:37:12 -06:00
|
|
|
* of the constant.
|
|
|
|
|
*/
|
|
|
|
|
SUBL $0x468, BX
|
2009-03-31 16:45:12 -06:00
|
|
|
|
runtime: work around kernel bug in Snow Leopard signal handling
Could not take a signal on threads other than the main thread.
If you look at the spinning binary with dtrace, you can see a
fault happening over and over:
$ dtrace -n '
fbt::user_trap:entry /execname=="boot32" && self->count < 10/
{
self->count++;
printf("%s %x %x %x %x", probefunc, arg1, arg2, arg3, arg4);
stack();
tracemem(arg4, 256);
}'
dtrace: description 'fbt::user_trap:entry ' matched 1 probe
CPU ID FUNCTION:NAME
1 17015 user_trap:entry user_trap 0 10 79af0a0 79af0a0
mach_kernel`lo_alltraps+0x12a
0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef
0: 0e 00 00 00 37 00 00 00 00 00 00 00 1f 00 00 00 ....7...........
10: 1f 00 00 00 a8 33 00 00 00 00 00 01 00 00 00 00 .....3..........
20: 98 ba dc fe 07 09 00 00 00 00 00 00 98 ba dc fe ................
30: 06 00 00 00 0d 00 00 00 34 00 00 00 9e 1c 00 00 ........4.......
40: 17 00 00 00 00 02 00 00 ac 30 00 00 1f 00 00 00 .........0......
50: 00 00 00 00 00 00 00 00 0d 00 00 00 e0 e6 29 00 ..............).
60: 34 00 00 00 00 00 00 00 9e 1c 00 00 00 00 00 00 4...............
70: 17 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 ................
80: ac 30 00 00 00 00 00 00 1f 00 00 00 00 00 00 00 .0..............
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
a0: 48 00 00 00 10 00 00 00 85 00 00 00 a0 f2 29 00 H.............).
b0: 69 01 00 02 00 00 00 00 e6 93 04 82 ff 7f 00 00 i...............
c0: 2f 00 00 00 00 00 00 00 06 02 00 00 00 00 00 00 /...............
d0: 78 ee 42 01 01 00 00 00 1f 00 00 00 00 00 00 00 x.B.............
e0: 00 ed 9a 07 00 00 00 00 00 00 00 00 00 00 00 00 ................
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
...
The memory dump shows a 32-bit exception frame:
x86_saved_state32
gs = 0x37
fs = 0
es = 0x1f
ds = 0x1f
edi = 0x33a8
esi = 0x01000000
ebp = 0
cr2 = 0xfedcba98
ebx = 0x0907
edx = 0
ecx = 0xfedcba98
eax = 0x06
trapno = 0x0d
err = 0x34
eip = 0x1c9e
cs = 0x17
efl = 0x0200
uesp = 0x30ac
ss = 0x1f
The cr2 of 0xfedcba98 is the address that the new thread read
to cause the fault, but note that the trap is now a GP fault with
error code 0x34, meaning it's moved past the cr2 problem and on
to an invaild segment selector. The 0x34 is suspiciously similar
to the 0x37 in gs, and sure enough, OS X forces gs to have
that value in the signal handler, and if your thread hasn't set
up that segment (known as USER_CTHREAD), you'll fault on the IRET
into the signal handler and never be able to handle a signal.
The kernel bug is that it forces segment 0x37 without making sure
it is a valid segment. Leopard also forced 0x37 but had the courtesy
to set it up first.
Since OS X requires us to set up that segment (using the
thread_fast_set_cthread_self system call), we might as well
use it instead of the more complicated i386_set_ldt call to
set up our per-OS thread storage.
Also add some more zeros to bsdthread_register for new arguments
in Snow Leopard (apparently unnecessary, but being careful).
Fixes #510.
R=r
CC=golang-dev
https://golang.org/cl/824046
2010-04-08 14:24:37 -06:00
|
|
|
/*
|
|
|
|
|
* Must set up as USER_CTHREAD segment because
|
|
|
|
|
* Darwin forces that value into %gs for signal handlers,
|
|
|
|
|
* and if we don't set one up, we'll get a recursive
|
|
|
|
|
* fault trying to get into the signal handler.
|
|
|
|
|
* Since we have to set one up anyway, it might as
|
|
|
|
|
* well be the value we want. So don't bother with
|
|
|
|
|
* i386_set_ldt.
|
|
|
|
|
*/
|
|
|
|
|
MOVL BX, 4(SP)
|
|
|
|
|
MOVL $3, AX // thread_fast_set_cthread_self - machdep call #3
|
|
|
|
|
INT $0x82 // sic: 0x82, not 0x80, for machdep call
|
2009-03-31 16:45:12 -06:00
|
|
|
|
runtime: work around kernel bug in Snow Leopard signal handling
Could not take a signal on threads other than the main thread.
If you look at the spinning binary with dtrace, you can see a
fault happening over and over:
$ dtrace -n '
fbt::user_trap:entry /execname=="boot32" && self->count < 10/
{
self->count++;
printf("%s %x %x %x %x", probefunc, arg1, arg2, arg3, arg4);
stack();
tracemem(arg4, 256);
}'
dtrace: description 'fbt::user_trap:entry ' matched 1 probe
CPU ID FUNCTION:NAME
1 17015 user_trap:entry user_trap 0 10 79af0a0 79af0a0
mach_kernel`lo_alltraps+0x12a
0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef
0: 0e 00 00 00 37 00 00 00 00 00 00 00 1f 00 00 00 ....7...........
10: 1f 00 00 00 a8 33 00 00 00 00 00 01 00 00 00 00 .....3..........
20: 98 ba dc fe 07 09 00 00 00 00 00 00 98 ba dc fe ................
30: 06 00 00 00 0d 00 00 00 34 00 00 00 9e 1c 00 00 ........4.......
40: 17 00 00 00 00 02 00 00 ac 30 00 00 1f 00 00 00 .........0......
50: 00 00 00 00 00 00 00 00 0d 00 00 00 e0 e6 29 00 ..............).
60: 34 00 00 00 00 00 00 00 9e 1c 00 00 00 00 00 00 4...............
70: 17 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 ................
80: ac 30 00 00 00 00 00 00 1f 00 00 00 00 00 00 00 .0..............
90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
a0: 48 00 00 00 10 00 00 00 85 00 00 00 a0 f2 29 00 H.............).
b0: 69 01 00 02 00 00 00 00 e6 93 04 82 ff 7f 00 00 i...............
c0: 2f 00 00 00 00 00 00 00 06 02 00 00 00 00 00 00 /...............
d0: 78 ee 42 01 01 00 00 00 1f 00 00 00 00 00 00 00 x.B.............
e0: 00 ed 9a 07 00 00 00 00 00 00 00 00 00 00 00 00 ................
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
...
The memory dump shows a 32-bit exception frame:
x86_saved_state32
gs = 0x37
fs = 0
es = 0x1f
ds = 0x1f
edi = 0x33a8
esi = 0x01000000
ebp = 0
cr2 = 0xfedcba98
ebx = 0x0907
edx = 0
ecx = 0xfedcba98
eax = 0x06
trapno = 0x0d
err = 0x34
eip = 0x1c9e
cs = 0x17
efl = 0x0200
uesp = 0x30ac
ss = 0x1f
The cr2 of 0xfedcba98 is the address that the new thread read
to cause the fault, but note that the trap is now a GP fault with
error code 0x34, meaning it's moved past the cr2 problem and on
to an invaild segment selector. The 0x34 is suspiciously similar
to the 0x37 in gs, and sure enough, OS X forces gs to have
that value in the signal handler, and if your thread hasn't set
up that segment (known as USER_CTHREAD), you'll fault on the IRET
into the signal handler and never be able to handle a signal.
The kernel bug is that it forces segment 0x37 without making sure
it is a valid segment. Leopard also forced 0x37 but had the courtesy
to set it up first.
Since OS X requires us to set up that segment (using the
thread_fast_set_cthread_self system call), we might as well
use it instead of the more complicated i386_set_ldt call to
set up our per-OS thread storage.
Also add some more zeros to bsdthread_register for new arguments
in Snow Leopard (apparently unnecessary, but being careful).
Fixes #510.
R=r
CC=golang-dev
https://golang.org/cl/824046
2010-04-08 14:24:37 -06:00
|
|
|
XORL AX, AX
|
|
|
|
|
MOVW GS, AX
|
2009-03-31 16:45:12 -06:00
|
|
|
RET
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·sysctl(SB),NOSPLIT,$0
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
MOVL $202, AX
|
|
|
|
|
INT $0x80
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
JAE 4(PC)
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
NEGL AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+24(FP)
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
RET
|
|
|
|
|
MOVL $0, AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+24(FP)
|
runtime: parallelize garbage collector mark + sweep
Running test/garbage/parser.out.
On a 4-core Lenovo X201s (Linux):
31.12u 0.60s 31.74r 1 cpu, no atomics
32.27u 0.58s 32.86r 1 cpu, atomic instructions
33.04u 0.83s 27.47r 2 cpu
On a 16-core Xeon (Linux):
33.08u 0.65s 33.80r 1 cpu, no atomics
34.87u 1.12s 29.60r 2 cpu
36.00u 1.87s 28.43r 3 cpu
36.46u 2.34s 27.10r 4 cpu
38.28u 3.85s 26.92r 5 cpu
37.72u 5.25s 26.73r 6 cpu
39.63u 7.11s 26.95r 7 cpu
39.67u 8.10s 26.68r 8 cpu
On a 2-core MacBook Pro Core 2 Duo 2.26 (circa 2009, MacBookPro5,5):
39.43u 1.45s 41.27r 1 cpu, no atomics
43.98u 2.95s 38.69r 2 cpu
On a 2-core Mac Mini Core 2 Duo 1.83 (circa 2008; Macmini2,1):
48.81u 2.12s 51.76r 1 cpu, no atomics
57.15u 4.72s 51.54r 2 cpu
The handoff algorithm is really only good for two cores.
Beyond that we will need to so something more sophisticated,
like have each core hand off to the next one, around a circle.
Even so, the code is a good checkpoint; for now we'll limit the
number of gc procs to at most 2.
R=dvyukov
CC=golang-dev
https://golang.org/cl/4641082
2011-09-30 07:40:01 -06:00
|
|
|
RET
|
2013-03-14 00:38:37 -06:00
|
|
|
|
|
|
|
|
// int32 runtime·kqueue(void);
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·kqueue(SB),NOSPLIT,$0
|
2013-03-14 00:38:37 -06:00
|
|
|
MOVL $362, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
|
|
|
|
NEGL AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+0(FP)
|
2013-03-14 00:38:37 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// int32 runtime·kevent(int kq, Kevent *changelist, int nchanges, Kevent *eventlist, int nevents, Timespec *timeout);
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·kevent(SB),NOSPLIT,$0
|
2013-03-14 00:38:37 -06:00
|
|
|
MOVL $363, AX
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
|
|
|
|
NEGL AX
|
cmd/cc, runtime: convert C compilers to use Go calling convention
To date, the C compilers and Go compilers differed only in how
values were returned from functions. This made it difficult to call
Go from C or C from Go if return values were involved. It also made
assembly called from Go and assembly called from C different.
This CL changes the C compiler to use the Go conventions, passing
results on the stack, after the arguments.
[Exception: this does not apply to C ... functions, because you can't
know where on the stack the arguments end.]
By doing this, the CL makes it possible to rewrite C functions into Go
one at a time, without worrying about which languages call that
function or which languages it calls.
This CL also updates all the assembly files in package runtime to use
the new conventions. Argument references of the form 40(SP) have
been rewritten to the form name+10(FP) instead, and there are now
Go func prototypes for every assembly function called from C or Go.
This means that 'go vet runtime' checks effectively every assembly
function, and go vet's output was used to automate the bulk of the
conversion.
Some functions, like seek and nsec on Plan 9, needed to be rewritten.
Many assembly routines called from C were reading arguments
incorrectly, using MOVL instead of MOVQ or vice versa, especially on
the less used systems like openbsd.
These were found by go vet and have been corrected too.
If we're lucky, this may reduce flakiness on those systems.
Tested on:
darwin/386
darwin/amd64
linux/arm
linux/386
linux/amd64
If this breaks another system, the bug is almost certainly in the
sys_$GOOS_$GOARCH.s file, since the rest of the CL is tested
by the combination of the above systems.
LGTM=dvyukov, iant
R=golang-codereviews, 0intro, dave, alex.brainman, dvyukov, iant
CC=golang-codereviews, josharian, r
https://golang.org/cl/135830043
2014-08-27 09:32:17 -06:00
|
|
|
MOVL AX, ret+24(FP)
|
2013-03-14 00:38:37 -06:00
|
|
|
RET
|
|
|
|
|
|
|
|
|
|
// int32 runtime·closeonexec(int32 fd);
|
2013-08-07 13:20:05 -06:00
|
|
|
TEXT runtime·closeonexec(SB),NOSPLIT,$32
|
2013-03-14 00:38:37 -06:00
|
|
|
MOVL $92, AX // fcntl
|
|
|
|
|
// 0(SP) is where the caller PC would be; kernel skips it
|
|
|
|
|
MOVL fd+0(FP), BX
|
|
|
|
|
MOVL BX, 4(SP) // fd
|
|
|
|
|
MOVL $2, 8(SP) // F_SETFD
|
|
|
|
|
MOVL $1, 12(SP) // FD_CLOEXEC
|
|
|
|
|
INT $0x80
|
|
|
|
|
JAE 2(PC)
|
|
|
|
|
NEGL AX
|
|
|
|
|
RET
|
