runtime: deadlock while stopping the world (most recently openbsd/amd64)

[gopherbot]

#!watchflakes
default <- log ~ `SIGQUIT` && log ~ `stopping the world`

Issue created automatically to collect these failures.

Example (log):

SIGQUIT: quit
PC=0x80b76f7 m=14 sigcode=0

eax    0x4
ebx    0x0
ecx    0x8407b00
edx    0x854aa80
edi    0xb09dfe7c
esi    0xf
ebp    0x3
...
	/tmp/workdir/go/src/runtime/proc.go:1262 +0xc7 fp=0x8438e1c sp=0x8438df4 pc=0x8087577
runtime.stopTheWorldGC(0xa)
	/tmp/workdir/go/src/runtime/proc.go:1310 +0x45 fp=0x8438e34 sp=0x8438e1c pc=0x8087705
runtime.StartTrace()
	/tmp/workdir/go/src/runtime/trace.go:284 +0x1c fp=0x8438e7c sp=0x8438e34 pc=0x80a432c
runtime/trace.Start({0x8237a2c, 0x847e270})
	/tmp/workdir/go/src/runtime/trace/trace.go:125 +0x75 fp=0x8438e98 sp=0x8438e7c pc=0x81105e5
runtime/trace_test.TestTraceFutileWakeup(0x85340f0)
	/tmp/workdir/go/src/runtime/trace/trace_test.go:510 +0xc5 fp=0x8438f98 sp=0x8438e98 pc=0x81b2315
testing.tRunner(0x85340f0, 0x8208078)

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
default <- pkg == "runtime/trace" && test == "TestTraceFutileWakeup"

2023-09-08 18:57 netbsd-386-9_3 go@729e21b2 runtime/trace.TestTraceFutileWakeup (log)

SIGQUIT: quit
PC=0x80b76f7 m=14 sigcode=0

eax    0x4
ebx    0x0
ecx    0x8407b00
edx    0x854aa80
edi    0xb09dfe7c
esi    0xf
ebp    0x3
...
	/tmp/workdir/go/src/runtime/proc.go:1262 +0xc7 fp=0x8438e1c sp=0x8438df4 pc=0x8087577
runtime.stopTheWorldGC(0xa)
	/tmp/workdir/go/src/runtime/proc.go:1310 +0x45 fp=0x8438e34 sp=0x8438e1c pc=0x8087705
runtime.StartTrace()
	/tmp/workdir/go/src/runtime/trace.go:284 +0x1c fp=0x8438e7c sp=0x8438e34 pc=0x80a432c
runtime/trace.Start({0x8237a2c, 0x847e270})
	/tmp/workdir/go/src/runtime/trace/trace.go:125 +0x75 fp=0x8438e98 sp=0x8438e7c pc=0x81105e5
runtime/trace_test.TestTraceFutileWakeup(0x85340f0)
	/tmp/workdir/go/src/runtime/trace/trace_test.go:510 +0xc5 fp=0x8438f98 sp=0x8438e98 pc=0x81b2315
testing.tRunner(0x85340f0, 0x8208078)

— watchflakes

[mknyszek]

This doesn't look like a failure in TestTraceFutileWakeup, but rather a deadlock (or something) when trying to stop-the-world. It's also on NetBSD.

[mknyszek]

I think let's wait for more failures here. If it's a real issue on more platforms I want to track that. If it only happens on NetBSD, then I'll mark it as such.

[mknyszek]

Oh, this may be related to #62524 given that it's a NetBSD failure. Once we have a resolution there, let's close this optimistically.

[mknyszek]

Oh, that does have a resolution. Closing optimistically.