What version of Go are you using (go version)?

$ go version
go version go1.10 linux/amd64

Does this issue reproduce with the latest release?

Cannot reproduce on go1.10, let alone the latest release.

What operating system and processor architecture are you using (go env)?

Here is the compiling env. Running env is out of reach.

$ go env
GOARCH="amd64"
GOBIN=""
GOCACHE="/root/.cache/go-build"
GOEXE=""
GOHOSTARCH="amd64"
GOHOSTOS="linux"
GOOS="linux"
GOPATH="/root/go"
GORACE=""
GOROOT="/opt/go"
GOTMPDIR=""
GOTOOLDIR="/opt/go/pkg/tool/linux_amd64"
GCCGO="gccgo"
CC="gcc"
CXX="g++"
CGO_ENABLED="1"
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -m64 -pthread -fmessage-length=0 -fdebug-prefix-map=/tmp/go-build089432846=/tmp/go-build"

What did you do?

It is a C++ server program, calling a go lib.

What did you expect to see?

Program continues running for a long term.

What did you see instead?

Program crashed after running for several days, with linux log

Dec 20 18:24:03 hostname kernel: myprogram[27064]: segfault at 119 ip 00007f5bb82c8bd2 sp 00007f5bb4840f70 error 4 in libgowrapper.so[7f5bb7fc3000+8e7000]

My analysis

hex(0x7f5bb82c8bd2 - 0x7f5bb7fc3000)= 0x305bd2

(gdb) info sym 0x305bd2
runtime.gentraceback + 5522 in section .text
(gdb) break *0x305bd2  
Breakpoint 1 at 0x305bd2: file /opt/go/src/runtime/traceback.go, line 290.

Link to the line

func gentraceback(...) int {
			flr = findfunc(frame.lr)
			if !flr.valid() {
				// This happens if you get a profiling interrupt at just the wrong time. ...
				doPrint := printing
				if doPrint && gp.m.incgo { // *** Line 290
					// We can inject sigpanic...
					doPrint = false
				}

unsafe.Offsetof(m.incgo) gives 0x119 so the crash point should be exact: m==nil thus accessing m.incgo cause the segfault.

doPrint should be true, so pcbuf == nil && callback == nil is true. gentraceback() should be called for 'stack prints'.

My questiones:

• How to reproduce 'get a profiling interrupt at just the wrong time'?
• Is gp.m==nil a go-side problem, or it could be caused by user-side code?