Go version
go version go1.21.6 linux/arm
Output of go env in your module/workspace:
GO111MODULE=''
GOARCH='arm'
GOBIN=''
GOCACHE='/home/pi/.cache/go-build'
GOENV='/home/pi/.config/go/env'
GOEXE=''
GOEXPERIMENT=''
GOFLAGS=''
GOHOSTARCH='arm'
GOHOSTOS='linux'
GOINSECURE=''
GOMODCACHE='/home/pi/go/pkg/mod'
GONOPROXY=''
GONOSUMDB=''
GOOS='linux'
GOPATH='/home/pi/go'
GOPRIVATE=''
GOPROXY='https://proxy.golang.org,direct'
GOROOT='/home/pi/go'
GOSUMDB='sum.golang.org'
GOTMPDIR=''
GOTOOLCHAIN='auto'
GOTOOLDIR='/home/pi/go/pkg/tool/linux_arm'
GOVCS=''
GOVERSION='go1.21.6'
GCCGO='gccgo'
GOARM='6'
AR='ar'
CC='gcc'
CXX='g++'
CGO_ENABLED='1'
GOMOD='/dev/null'
GOWORK=''
CGO_CFLAGS='-O2 -g'
CGO_CPPFLAGS=''
CGO_CXXFLAGS='-O2 -g'
CGO_FFLAGS='-O2 -g'
CGO_LDFLAGS='-O2 -g'
PKG_CONFIG='pkg-config'
GOGCCFLAGS='-fPIC -marm -pthread -Wl,--no-gc-sections -fmessage-length=0 -ffile-prefix-map=/tmp/go-build1411701886=/tmp/go-build -gno-record-gcc-switches'
What did you do?
Long running task crashed overnight in vicinity of epoll. No user interaction.
What did you see happen?
We have observed this crash a couple of times over the past year. It is very sporadic and not reproducable. Always on Raspberry Pi. OS version is 11.8.
runtime: netpoll: break fd ready for 5
fatal error: runtime: netpoll: break fd ready for something unexpected
runtime stack:
runtime.throw({0x30936a, 0x39})
/home/pi/go/src/runtime/panic.go:1077 +0x4c fp=0xeaffe4b4 sp=0xeaffe4a0 pc=0x4f0c4
runtime.netpoll(0x0)
/home/pi/go/src/runtime/netpoll_epoll.go:141 +0x3c4 fp=0xeaffed1c sp=0xeaffe4b4 pc=0x4ace0
runtime.findRunnable()
/home/pi/go/src/runtime/proc.go:2978 +0x360 fp=0xeaffedb8 sp=0xeaffed1c pc=0x57ff8
runtime.schedule()
/home/pi/go/src/runtime/proc.go:3582 +0xac fp=0xeaffedd4 sp=0xeaffedb8 pc=0x59d24
runtime.park_m(0x883f00)
/home/pi/go/src/runtime/proc.go:3745 +0x160 fp=0xeaffedf0 sp=0xeaffedd4 pc=0x5a570
traceback: unexpected SPWRITE function runtime.mcall
runtime.mcall()
/home/pi/go/src/runtime/asm_arm.s:265 +0x48 fp=0xeaffedf8 sp=0xeaffedf0 pc=0x84f80
goroutine 1 [IO wait]:
runtime.gopark(0x311298, 0xeeb1bee0, 0x2, 0x1b, 0x5)
/home/pi/go/src/runtime/proc.go:398 +0x104 fp=0x83cd28 sp=0x83cd14 pc=0x52874
runtime.netpollblock(0xeeb1bed0, 0x72, 0x0)
/home/pi/go/src/runtime/netpoll.go:564 +0x100 fp=0x83cd40 sp=0x83cd28 pc=0x49ef8
internal/poll.runtime_pollWait(0xeeb1bed0, 0x72)
/home/pi/go/src/runtime/netpoll.go:343 +0x54 fp=0x83cd54 sp=0x83cd40 pc=0x823cc
internal/poll.(*pollDesc).wait(0x918158, 0x72, 0x0)
/home/pi/go/src/internal/poll/fd_poll_runtime.go:84 +0x30 fp=0x83cd68 sp=0x83cd54 pc=0xd3608
internal/poll.(*pollDesc).waitRead(...)
/home/pi/go/src/internal/poll/fd_poll_runtime.go:89
internal/poll.(*FD).Accept(0x918140)
/home/pi/go/src/internal/poll/fd_unix.go:611 +0x2c0 fp=0x83cdb8 sp=0x83cd68 pc=0xd4a08
net.(*netFD).accept(0x918140)
/home/pi/go/src/net/fd_unix.go:172 +0x20 fp=0x83ce18 sp=0x83cdb8 pc=0x1b82b4
net.(*TCPListener).accept(0x930090)
/home/pi/go/src/net/tcpsock_posix.go:152 +0x20 fp=0x83ce30 sp=0x83ce18 pc=0x1ca670
net.(*TCPListener).Accept(0x930090)
/home/pi/go/src/net/tcpsock.go:315 +0x30 fp=0x83ce4c sp=0x83ce30 pc=0x1c980c
net/http.(*onceCloseListener).Accept(0x914860)
<autogenerated>:1 +0x34 fp=0x83ce64 sp=0x83ce4c pc=0x27bba0
net/http.(*Server).Serve(0x95a000, {0x360df8, 0x930090})
/home/pi/go/src/net/http/server.go:3056 +0x330 fp=0x83cf00 sp=0x83ce64 pc=0x2704c4
net/http.(*Server).ListenAndServe(0x95a000)
/home/pi/go/src/net/http/server.go:2985 +0xa4 fp=0x83cf24 sp=0x83cf00 pc=0x270144
net/http.ListenAndServe(...)
/home/pi/go/src/net/http/server.go:3239
main.main()
/home/pi/foo/cmd/bar/main.go:38 +0x374 fp=0x83cfa0 sp=0x83cf24 pc=0x2907c8
runtime.main()
/home/pi/go/src/runtime/proc.go:267 +0x338 fp=0x83cfec sp=0x83cfa0 pc=0x522dc
runtime.goexit()
/home/pi/go/src/runtime/asm_arm.s:859 +0x4 fp=0x83cfec sp=0x83cfec pc=0x86c4c
What did you expect to see?
No panic.
Comment From: ianlancetaylor
"ready for 5" means that both EPOLLIN and EPOLLOUT are set, but we only expect EPOLLIN.
I'm not sure what is happening but this was probably fixed for the future Go 1.23 release as a side effect of https://go.dev/cl/560615.
Comment From: chalkytoast
Thanks for looking into this. We'll report when 1.23 is released and had time to evaluate.
Comment From: gopherbot
Timed out in state WaitingForInfo. Closing.
(I am just a bot, though. Please speak up if this is a mistake or you have the requested information.)