We only guarantee that the main goroutine runs on the
main OS thread for initialization. Programs that wish to
preserve that property for main.main can call runtime.LockOSThread.
This is what programs used to do before we unleashed
goroutines during init, so it is both a simple fix and keeps
existing programs working.
R=iant, r, dave, dvyukov
CC=golang-dev
https://golang.org/cl/5309070
Fixes#2337.
Unfortunate sequence of events is:
1. maxcpu=2, mcpu=1, grunning=1
2. starttheworld creates an extra M:
maxcpu=2, mcpu=2, grunning=1
4. the goroutine calls runtime.GOMAXPROCS(1)
maxcpu=1, mcpu=2, grunning=1
5. since it sees mcpu>maxcpu, it calls gosched()
6. schedule() deschedules the goroutine:
maxcpu=1, mcpu=1, grunning=0
7. schedule() call getnextandunlock() which
fails to pick up the goroutine again,
because canaddcpu() fails, because mcpu==maxcpu
8. then it sees that grunning==0,
reports deadlock and terminates
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5191044
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
Make the stack traces more readable for new
Go programmers while preserving their utility for old hands.
- Change status number [4] to string.
- Elide frames in runtime package (internal details).
- Swap file:line and arguments.
- Drop 'created by' for main goroutine.
- Show goroutines in order of allocation:
implies main goroutine first if nothing else.
There is no option to get the extra frames back.
Uncomment 'return 1' at the bottom of symtab.c.
$ 6.out
throw: all goroutines are asleep - deadlock!
goroutine 1 [chan send]:
main.main()
/Users/rsc/g/go/src/pkg/runtime/x.go:22 +0x8a
goroutine 2 [select (no cases)]:
main.sel()
/Users/rsc/g/go/src/pkg/runtime/x.go:11 +0x18
created by main.main
/Users/rsc/g/go/src/pkg/runtime/x.go:19 +0x23
goroutine 3 [chan receive]:
main.recv(0xf8400010a0, 0x0)
/Users/rsc/g/go/src/pkg/runtime/x.go:15 +0x2e
created by main.main
/Users/rsc/g/go/src/pkg/runtime/x.go:20 +0x50
goroutine 4 [chan receive (nil chan)]:
main.recv(0x0, 0x0)
/Users/rsc/g/go/src/pkg/runtime/x.go:15 +0x2e
created by main.main
/Users/rsc/g/go/src/pkg/runtime/x.go:21 +0x66
$
$ 6.out index
panic: runtime error: index out of range
goroutine 1 [running]:
main.main()
/Users/rsc/g/go/src/pkg/runtime/x.go:25 +0xb9
$
$ 6.out nil
panic: runtime error: invalid memory address or nil pointer dereference
[signal 0xb code=0x1 addr=0x0 pc=0x22ca]
goroutine 1 [running]:
main.main()
/Users/rsc/g/go/src/pkg/runtime/x.go:28 +0x211
$
$ 6.out panic
panic: panic
goroutine 1 [running]:
main.main()
/Users/rsc/g/go/src/pkg/runtime/x.go:30 +0x101
$
R=golang-dev, qyzhai, n13m3y3r, r
CC=golang-dev
https://golang.org/cl/4907048
Allocate Defer on stack during cgo calls, as suggested
by dvyukov. Also includes some comment corrections.
benchmark old,ns/op new,ns/op
BenchmarkCgoCall 669 330
(Intel Xeon CPU 1.80GHz * 4, Linux 386)
R=dvyukov, rsc
CC=golang-dev
https://golang.org/cl/4910041
The corruption can occur when GOMAXPROCS
is changed from >1 to 1, since GOMAXPROCS=1
does not imply there is only 1 goroutine running,
other goroutines can still be not parked after
the change.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4873050
Replace cas with xadd in scheduler.
Suggested by Dmitriy in last code review.
Verified with Promela model.
When there's actual contention for the atomic word,
this avoids the looping that compare-and-swap requires.
benchmark old ns/op new ns/op delta
runtime_test.BenchmarkSyscall 32 26 -17.08%
runtime_test.BenchmarkSyscall-2 155 59 -61.81%
runtime_test.BenchmarkSyscall-3 112 52 -52.95%
runtime_test.BenchmarkSyscall-4 94 48 -48.57%
runtime_test.BenchmarkSyscallWork 871 872 +0.11%
runtime_test.BenchmarkSyscallWork-2 481 477 -0.83%
runtime_test.BenchmarkSyscallWork-3 338 335 -0.89%
runtime_test.BenchmarkSyscallWork-4 263 256 -2.66%
R=golang-dev, iant
CC=golang-dev
https://golang.org/cl/4800047
Used to use mcpu+msyscall but that's
problematic for packing into a single
atomic word. The running goroutine count
(where running == Go code or syscall)
can be maintained separately, always
manipulated under lock.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/4767041
runtime.goidgen can be quite frequently modified and
shares cache line with the following variables,
it leads to false sharing.
50c6b0 b nfname
50c6b4 b nfunc
50c6b8 b nfunc$17
50c6bc b nhist$17
50c6c0 B runtime.checking
50c6c4 B runtime.gcwaiting
50c6c8 B runtime.goidgen
50c6cc B runtime.gomaxprocs
50c6d0 B runtime.panicking
50c6d4 B strconv.IntSize
50c6d8 B src/pkg/runtime/_xtest_.ss
50c6e0 B src/pkg/runtime/_xtest_.stop
50c6e8 b addrfree
50c6f0 b addrmem
50c6f8 b argv
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4673054
The g->sched.sp saved stack pointer and the
g->stackbase and g->stackguard stack bounds
can change even while "the world is stopped",
because a goroutine has to call functions (and
therefore might split its stack) when exiting a
system call to check whether the world is stopped
(and if so, wait until the world continues).
That means the garbage collector cannot access
those values safely (without a race) for goroutines
executing system calls. Instead, save a consistent
triple in g->gcsp, g->gcstack, g->gcguard during
entersyscall and have the garbage collector refer
to those.
The old code was occasionally seeing (because of
the race) an sp and stk that did not correspond to
each other, so that stk - sp was not the number of
stack bytes following sp. In that case, if sp < stk
then the call scanblock(sp, stk - sp) scanned too
many bytes (anything between the two pointers,
which pointed into different allocation blocks).
If sp > stk then stk - sp wrapped around.
On 32-bit, stk - sp is a uintptr (uint32) converted
to int64 in the call to scanblock, so a large (~4G)
but positive number. Scanblock would try to scan
that many bytes and eventually fault accessing
unmapped memory. On 64-bit, stk - sp is a uintptr (uint64)
promoted to int64 in the call to scanblock, so a negative
number. Scanblock would not scan anything, possibly
causing in-use blocks to be freed.
In short, 32-bit platforms would have seen either
ineffective garbage collection or crashes during garbage
collection, while 64-bit platforms would have seen
either ineffective or incorrect garbage collection.
You can see the invalid arguments to scanblock in the
stack traces in issue 1620.
Fixes#1620.
Fixes#1746.
R=iant, r
CC=golang-dev
https://golang.org/cl/4437075
* Change use of m->g0 stack (aka scheduler stack).
* Provide runtime.mcall(f) to invoke f() on m->g0 stack.
* Replace scheduler loop entry with runtime.mcall(schedule).
Runtime.mcall eliminates the need for fake scheduler states that
exist just to run a bit of code on the m->g0 stack
(Grecovery, Gstackalloc).
The elimination of the scheduler as a loop that stops and
starts using gosave and gogo fixes a bad interaction with the
way cgo uses the m->g0 stack. Cgo runs external (gcc-compiled)
C functions on that stack, and then when calling back into Go,
it sets m->g0->sched.sp below the added call frames, so that
other uses of m->g0's stack will not interfere with those frames.
Unfortunately, gogo (longjmp) back to the scheduler loop at
this point would end up running scheduler with the lower
sp, which no longer points at a valid stack frame for
a call to scheduler. If scheduler then wrote any function call
arguments or local variables to where it expected the stack
frame to be, it would overwrite other data on the stack.
I realized this possibility while debugging a problem with
calling complex Go code in a Go -> C -> Go cgo callback.
This wasn't the bug I was looking for, it turns out, but I believe
it is a real bug nonetheless. Switching to runtime.mcall, which
only adds new frames to the stack and never jumps into
functions running in existing ones, fixes this bug.
* Move cgo-related code out of proc.c into cgocall.c.
* Add very large comment describing cgo call sequences.
* Simpilify, regularize cgo function implementations and names.
* Add test suite as misc/cgo/test.
Now the Go -> C path calls cgocall, which calls asmcgocall,
and the C -> Go path calls cgocallback, which calls cgocallbackg.
The shuffling, which affects mainly the callback case, moves
most of the callback implementation to cgocallback running
on the m->curg stack (not the m->g0 scheduler stack) and
only while accounted for with $GOMAXPROCS (between calls
to exitsyscall and entersyscall).
The previous callback code did not block in startcgocallback's
approximation to exitsyscall, so if, say, the garbage collector
were running, it would still barge in and start doing things
like call malloc. Similarly endcgocallback's approximation of
entersyscall did not call matchmg to kick off new OS threads
when necessary, which caused the bug in issue 1560.
Fixes#1560.
R=iant
CC=golang-dev
https://golang.org/cl/4253054
This functionality might be used in environments
where programs are limited to a single thread,
to simulate a select-driven network server. It is
not exposed via the standard runtime API.
R=r, r2
CC=golang-dev
https://golang.org/cl/4254041
Avoids deadlocks like the one below, in which a stack split happened
in order to call lock(&stacks), but then the stack unsplit cannot run
because stacks is now locked.
The only code calling stackalloc that wasn't on a scheduler
stack already was malg, which creates a new goroutine.
runtime.futex+0x23 /home/rsc/g/go/src/pkg/runtime/linux/amd64/sys.s:139
runtime.futex()
futexsleep+0x50 /home/rsc/g/go/src/pkg/runtime/linux/thread.c:51
futexsleep(0x5b0188, 0x300000003, 0x100020000, 0x4159e2)
futexlock+0x85 /home/rsc/g/go/src/pkg/runtime/linux/thread.c:119
futexlock(0x5b0188, 0x5b0188)
runtime.lock+0x56 /home/rsc/g/go/src/pkg/runtime/linux/thread.c:158
runtime.lock(0x5b0188, 0x7f0d27b4a000)
runtime.stackfree+0x4d /home/rsc/g/go/src/pkg/runtime/malloc.goc:336
runtime.stackfree(0x7f0d27b4a000, 0x1000, 0x8, 0x7fff37e1e218)
runtime.oldstack+0xa6 /home/rsc/g/go/src/pkg/runtime/proc.c:705
runtime.oldstack()
runtime.lessstack+0x22 /home/rsc/g/go/src/pkg/runtime/amd64/asm.s:224
runtime.lessstack()
----- lessstack called from goroutine 2 -----
runtime.lock+0x56 /home/rsc/g/go/src/pkg/runtime/linux/thread.c:158
runtime.lock(0x5b0188, 0x40a5e2)
runtime.stackalloc+0x55 /home/rsc/g/go/src/pkg/runtime/malloc.c:316
runtime.stackalloc(0x1000, 0x4055b0)
runtime.malg+0x3d /home/rsc/g/go/src/pkg/runtime/proc.c:803
runtime.malg(0x1000, 0x40add9)
runtime.newproc1+0x12b /home/rsc/g/go/src/pkg/runtime/proc.c:854
runtime.newproc1(0xf840027440, 0x7f0d27b49230, 0x0, 0x49f238, 0x40, ...)
runtime.newproc+0x2f /home/rsc/g/go/src/pkg/runtime/proc.c:831
runtime.newproc(0x0, 0xf840027440, 0xf800000010, 0x44b059)
...
R=r, r2
CC=golang-dev
https://golang.org/cl/4216045
Fix problems found.
On amd64, various library routines had bigger
stack frames than expected, because large function
calls had been added.
runtime.assertI2T: nosplit stack overflow
120 assumed on entry to runtime.assertI2T
8 after runtime.assertI2T uses 112
0 on entry to runtime.newTypeAssertionError
-8 on entry to runtime.morestack01
runtime.assertE2E: nosplit stack overflow
120 assumed on entry to runtime.assertE2E
16 after runtime.assertE2E uses 104
8 on entry to runtime.panic
0 on entry to runtime.morestack16
-8 after runtime.morestack16 uses 8
runtime.assertE2T: nosplit stack overflow
120 assumed on entry to runtime.assertE2T
16 after runtime.assertE2T uses 104
8 on entry to runtime.panic
0 on entry to runtime.morestack16
-8 after runtime.morestack16 uses 8
runtime.newselect: nosplit stack overflow
120 assumed on entry to runtime.newselect
56 after runtime.newselect uses 64
48 on entry to runtime.printf
8 after runtime.printf uses 40
0 on entry to vprintf
-8 on entry to runtime.morestack16
runtime.selectdefault: nosplit stack overflow
120 assumed on entry to runtime.selectdefault
56 after runtime.selectdefault uses 64
48 on entry to runtime.printf
8 after runtime.printf uses 40
0 on entry to vprintf
-8 on entry to runtime.morestack16
runtime.selectgo: nosplit stack overflow
120 assumed on entry to runtime.selectgo
0 after runtime.selectgo uses 120
-8 on entry to runtime.gosched
On arm, 5c was tagging functions NOSPLIT that should
not have been, like the recursive function printpanics:
printpanics: nosplit stack overflow
124 assumed on entry to printpanics
112 after printpanics uses 12
108 on entry to printpanics
96 after printpanics uses 12
92 on entry to printpanics
80 after printpanics uses 12
76 on entry to printpanics
64 after printpanics uses 12
60 on entry to printpanics
48 after printpanics uses 12
44 on entry to printpanics
32 after printpanics uses 12
28 on entry to printpanics
16 after printpanics uses 12
12 on entry to printpanics
0 after printpanics uses 12
-4 on entry to printpanics
R=r, r2
CC=golang-dev
https://golang.org/cl/4188061
Follow morestack, so that crashes during a stack split
give complete traces. Also mark stack segment boundaries
as an aid to debugging.
Correct various line number bugs with yet another attempt
at interpreting the pc/ln table. This one has a chance at
being correct, because I based it on reading src/cmd/ld/lib.c
instead of on reading the documentation.
Fixes#1138.
Fixes#1430.
Fixes#1461.
throw: runtime: split stack overflow
runtime.throw+0x3e /home/rsc/g/go2/src/pkg/runtime/runtime.c:78
runtime.throw(0x81880af, 0xf75c8b18)
runtime.newstack+0xad /home/rsc/g/go2/src/pkg/runtime/proc.c:728
runtime.newstack()
runtime.morestack+0x4f /home/rsc/g/go2/src/pkg/runtime/386/asm.s:184
runtime.morestack()
----- morestack called from stack: -----
runtime.new+0x1a /home/rsc/g/go2/src/pkg/runtime/malloc.c:288
runtime.new(0x1, 0x0, 0x0)
gongo.makeBoard+0x33 /tmp/Gongo/gongo_robot_test.go:344
gongo.makeBoard(0x809d238, 0x1, 0xf76092c8, 0x1)
----- stack segment boundary -----
gongo.checkEasyScore+0xcc /tmp/Gongo/gongo_robot_test.go:287
gongo.checkEasyScore(0xf764b710, 0x0, 0x809d238, 0x1)
gongo.TestEasyScore+0x8c /tmp/Gongo/gongo_robot_test.go:255
gongo.TestEasyScore(0xf764b710, 0x818a990)
testing.tRunner+0x2f /home/rsc/g/go2/src/pkg/testing/testing.go:132
testing.tRunner(0xf764b710, 0xf763b5dc, 0x0)
runtime.goexit /home/rsc/g/go2/src/pkg/runtime/proc.c:149
runtime.goexit()
R=ken2, r
CC=golang-dev
https://golang.org/cl/4000053
Shame on me: I fixed the same bug in 6l in 8691fcc6a66e
(https://golang.org/cl/2609041) and neglected
to look at 5l and 8l to see if they were affected.
On the positive side, the check I added in that CL is the
one that detected this bug.
Fixes#1457.
R=ken2
CC=golang-dev
https://golang.org/cl/3981052
The old heap maps used a multilevel table, but that
was overkill: there are only 1M entries on a 32-bit
machine and we can arrange to use a dense address
range on a 64-bit machine.
The heap map is in bss. The assumption is that if
we don't touch the pages they won't be mapped in.
Also moved some duplicated memory allocation
code out of the OS-specific files.
R=r
CC=golang-dev
https://golang.org/cl/4118042
The fault was lucky: when it wasn't faulting it was silently
copying a word from some other block and later putting
that same word back. If some other goroutine had changed
that word of memory in the interim, too bad.
The ARM code was inconsistent about whether the
"argument frame" included the saved LR. Including it made
some things more regular but mostly just caused confusion
in the places where the regularity broke. Now the rule
reflects reality: argp is always a pointer to arguments,
never a saved link register.
Renamed struct fields to make meaning clearer.
Running ARM in QEMU, package time's gotest:
* before: 27/58 failed
* after: 0/50
R=r, r2
CC=golang-dev
https://golang.org/cl/3993041
I missed that environment is used during runtime setup,
well before go init() functions run. Implemented os-dependent
runtime.goenvs functions to allow for different unix, plan9 and
windows versions of environment discovery.
R=rsc, paulzhol
CC=golang-dev
https://golang.org/cl/3787046
If we don't do this, then when C code calls back to Go code
which panics, we lose space on the scheduler stack. If that
happens a lot, eventually there is no space left on the
scheduler stack.
R=rsc
CC=golang-dev
https://golang.org/cl/3898042
