[erlang-questions] beam process dies with code 135 (SIGBUS)

[Vlad Dumitrescu]

Hi!

Does anyone have any experience of runtimes dying with code 135,
meaning by signal 7 (SIGBUS) (i.e. Bus error (bad memory access)) on
Linux?

We have seen nodes going down with this reason relatively often (but
randomly) and I don't know how to locate the problem. There doesn't
seem to be any code dump.

Could this be triggered by something internal?

This is a SUSE distribution with
> uname -a
Linux selnx120 2.6.16.53-0.16-bigsmp #1 SMP Tue Oct 2 16:57:49 UTC
2007 i686 i686 i386 GNU/Linux

best regards,
Vlad
_______________________________________________
erlang-questions mailing list
erlang-q...@erlang.org
http://www.erlang.org/mailman/listinfo/erlang-questions

[Richard Andrews]

> Does anyone have any experience of runtimes dying with code 135,
> meaning by signal 7 (SIGBUS) (i.e. Bus error (bad memory access)) on
> Linux?

Have you tried building OTP for i386?

IIRC x86_32 has no alignment restrictions meaning SIGBUS should not ever be generated on the architecture. But I read somewhere recently that SSE2 instructions and x86_64 do have alignment restrictions, hence the i386 suggestion above.

Could something be sending SIGBUS to beam?

Are you mixing 64-bit and 32-bit components?

Make Yahoo!7 your homepage and win a trip to the Quiksilver Pro. Find out more

[Vlad Dumitrescu]

Hi,

On Tue, Feb 17, 2009 at 11:59, Richard Andrews <bbm...@yahoo.com.au> wrote:
>> Does anyone have any experience of runtimes dying with code 135,
>> meaning by signal 7 (SIGBUS) (i.e. Bus error (bad memory access)) on
>> Linux?
>
> Have you tried building OTP for i386?
>
> IIRC x86_32 has no alignment restrictions meaning SIGBUS should not ever be generated on the architecture. But I read somewhere recently that SSE2 instructions and x86_64 do have alignment restrictions, hence the i386 suggestion above.

Unfortunately I can't easily rebuild. This happens in a restricted
environment where I don't have too much freedom.

BTW the Erlang version is R12B-4.

> Could something be sending SIGBUS to beam?

Not that I know of.

> Are you mixing 64-bit and 32-bit components?

That's a good question too. I will check.

Thanks!
regards,
Vlad