Re: Go binary panics on amd64-current - SIG*unknown

[Christos Zoulas]

On Jan 7,  9:03pm, al...@yandex.ru (Alexander Nasonov) wrote:
-- Subject: Re: Go binary panics on amd64-current - SIG*unknown


Thanks! Committed (3 x)

christos

Re: Go binary panics on amd64-current - SIG*unknown

[Alexander Nasonov]

Christos Zoulas wrote:
> Index: linux_sigaction.c
> ===
> RCS file: /cvsroot/src/sys/compat/linux/common/linux_sigaction.c,v
> retrieving revision 1.34
> diff -u -u -r1.34 linux_sigaction.c
> --- linux_sigaction.c 17 Oct 2008 20:21:34 -  1.34
> +++ linux_sigaction.c 7 Jan 2018 00:10:43 -
> @@ -84,6 +84,15 @@
>   linux_old_to_native_sigaction(, );
>   }
>   sig = SCARG(uap, signum);
> + /*
> +  * XXX: Linux has 33 realtime signals, the go binary wants to
> +  * reset all of them; nothing else uses the last RT signal, so for
> +  * now ignore it.
> +  */
> + if (sig == LINUX__NSIG) {
> + uprintf("%s: setting signal %d ignored\n", __func__, sig);
> + sig--;  /* back to 63 which is ignored */
> + }
>   if (sig < 0 || sig >= LINUX__NSIG)
>   return (EINVAL);
>   if (sig > 0 && !linux_to_native_signo[sig]) {

it works but you patched the wrong file, this change should go into
linux_signal.c.

-- 
Alex

Re: Go binary panics on amd64-current - SIG*unknown

[Alexander Nasonov]

Ah, I thought I mentioned it in my email. Yes, it’s Linux binary.Sent from Yandex.Mail for mobile: http://m.ya.ru/ymail

Re: Go binary panics on amd64-current - SIG*unknown

[Christos Zoulas]

In article <20180106224309.w2vicky2i6rf6fvu@neva>,
Alexander Nasonov   wrote:
>I downloaded IACA tool from intel.com but I couldn't run it:
>
>$ ktrace ./iaca-v3.0-lin64/iaca
>fatal error: rt_sigaction read failure
>
>runtime stack:
>runtime.throw(0x75de24, 0x19)
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/panic.go:596 +0x95
>runtime.getsig(0x40, 0x441c60)
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/os_linux.go:427 +0x92
>runtime.initsig(0xe59d00)
>   
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/signal_unix.go:79
>+0x98
>runtime.mstart1()
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/proc.go:1175 +0xa4
>runtime.mstart()
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/proc.go:1149 +0x64
>
>goroutine 1 [runnable]:
>runtime.main()
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/proc.go:106
>runtime.goexit()
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/asm_amd64.s:2197 +0x1
>
>goroutine 17 [syscall, locked to thread]:
>runtime.goexit()
>/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/asm_amd64.s:2197 +0x1
>
>
>$ kdump
>
>  5477   5477 iaca CALL  rt_sigaction(SIG*unknown 64*,0,0x7f7fe580,8)
>  5477   5477 iaca RET   rt_sigaction -1 errno -22 Invalid argument
>  5477   5477 iaca CALL  rt_sigaction(SIG*unknown 64*,0,0x7f7fe8b8,8)
>  5477   5477 iaca RET   rt_sigaction -1 errno -22 Invalid argument
> 
>$ uname -a
>NetBSD neva 8.99.7 NetBSD 8.99.7 (GENERIC_KASLR) #0: Sat Nov 18 09:54:53
>GMT 2017 
>alnsn@neva:/home/alnsn/netbsd-current/clean/src/sys/arch/amd64/compile/obj/GENERIC_KASLR
> amd64
>
>I don't know if it matters but I have PAX protections enabled:
>
>$ sysctl -a |grep -w pax | grep enabled
>security.pax.mprotect.enabled = 1
>security.pax.segvguard.enabled = 1
>security.pax.aslr.enabled = 1

Index: linux_sigaction.c
===
RCS file: /cvsroot/src/sys/compat/linux/common/linux_sigaction.c,v
retrieving revision 1.34
diff -u -u -r1.34 linux_sigaction.c
--- linux_sigaction.c   17 Oct 2008 20:21:34 -  1.34
+++ linux_sigaction.c   7 Jan 2018 00:10:43 -
@@ -84,6 +84,15 @@
linux_old_to_native_sigaction(, );
}
sig = SCARG(uap, signum);
+   /*
+* XXX: Linux has 33 realtime signals, the go binary wants to
+* reset all of them; nothing else uses the last RT signal, so for
+* now ignore it.
+*/
+   if (sig == LINUX__NSIG) {
+   uprintf("%s: setting signal %d ignored\n", __func__, sig);
+   sig--;  /* back to 63 which is ignored */
+   }
if (sig < 0 || sig >= LINUX__NSIG)
return (EINVAL);
if (sig > 0 && !linux_to_native_signo[sig]) {

Re: Go binary panics on amd64-current - SIG*unknown

[Stephen Ma]

I think the issue is caused by linux defining signals from 1 to
LINUX__NSIG, while the linux compat code in netbsd recognises signals
from 1 to LINUX__NSIG-1. That's why the binary is aborting when
calling sigaction on signal 64, which is LINUX__NSIG on amd64.

- S

Re: Go binary panics on amd64-current - SIG*unknown

[maya]

On Sat, Jan 06, 2018 at 10:43:09PM +, Alexander Nasonov wrote:
> runtime.getsig(0x40, 0x441c60)
> /nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/os_linux.go:427 
> +0x92

Is that a linux binary?

that seems to come from:
https://github.com/golang/go/blob/eaf603601bf79216772ba36a9b981f21b50ade96/src/runtime/os_linux.go#L427

(btw, there were some discrepancies between go and netbsd along
these lines, which might be related:
https://github.com/golang/go/commit/c6e7cb4a375966f7b034e6dbf439a58ea326104a )

Go binary panics on amd64-current - SIG*unknown

[Alexander Nasonov]

I downloaded IACA tool from intel.com but I couldn't run it:

$ ktrace ./iaca-v3.0-lin64/iaca
fatal error: rt_sigaction read failure

runtime stack:
runtime.throw(0x75de24, 0x19)
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/panic.go:596 +0x95
runtime.getsig(0x40, 0x441c60)
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/os_linux.go:427 +0x92
runtime.initsig(0xe59d00)
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/signal_unix.go:79 
+0x98
runtime.mstart1()
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/proc.go:1175 +0xa4
runtime.mstart()
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/proc.go:1149 +0x64

goroutine 1 [runnable]:
runtime.main()
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/proc.go:106
runtime.goexit()
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/asm_amd64.s:2197 +0x1

goroutine 17 [syscall, locked to thread]:
runtime.goexit()
/nfs/iil/disks/kfw/tools/go/go-latest/src/runtime/asm_amd64.s:2197 +0x1


$ kdump

  5477   5477 iaca CALL  rt_sigaction(SIG*unknown 64*,0,0x7f7fe580,8)
  5477   5477 iaca RET   rt_sigaction -1 errno -22 Invalid argument
  5477   5477 iaca CALL  rt_sigaction(SIG*unknown 64*,0,0x7f7fe8b8,8)
  5477   5477 iaca RET   rt_sigaction -1 errno -22 Invalid argument
 
$ uname -a
NetBSD neva 8.99.7 NetBSD 8.99.7 (GENERIC_KASLR) #0: Sat Nov 18 09:54:53 GMT 
2017  
alnsn@neva:/home/alnsn/netbsd-current/clean/src/sys/arch/amd64/compile/obj/GENERIC_KASLR
 amd64

I don't know if it matters but I have PAX protections enabled:

$ sysctl -a |grep -w pax | grep enabled
security.pax.mprotect.enabled = 1
security.pax.segvguard.enabled = 1
security.pax.aslr.enabled = 1

-- 
Alex