Erlang scheduler segfault at ffffffffffffffff

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Erlang scheduler segfault at ffffffffffffffff

Gabriele Santomaggio
Hi,
I am having the following the problem:
 
2019-05-06T10:46:16.413758+00:00 d98-f2-b3-c8-b8-65 kernel: [8649388.905125] ../mm/pgtable-generic.c:33: bad pmd ffff883eed530398(80000037db2008e7)
2019-05-06T10:46:16.423251+00:00 d98-f2-b3-c8-b8-65 kernel: [8649388.915722] 34_scheduler[15882]: segfault at ffffffffffffffff ip 0000000040e97113 sp 00007fcffdb7ef78 error 5

it causes a crash without erlang dump.

HIPE is enabled, do you have an idea if is related in some way to hipe?

notes:
1 - Erlang version 20.2.2
2 - 34_scheduler` is one of the erlang schedulers
3-  It is not systematic, it happens randomly 
4- I have some "notice: High CPU load detected: 997.159973"  around the problem
5- I ended up on https://lkml.org/lkml/2017/4/10/152 I am looking if it could be related (??!?).


( let me try also here, I have already asked on the Erlang slack channel :) )

Thank you
--
Gabriele Santomaggio

_______________________________________________
erlang-questions mailing list
[hidden email]
http://erlang.org/mailman/listinfo/erlang-questions
Reply | Threaded
Open this post in threaded view
|

Re: Erlang scheduler segfault at ffffffffffffffff

Mikael Pettersson-5
On Mon, May 20, 2019 at 9:20 PM Gabriele Santomaggio
<[hidden email]> wrote:

>
> Hi,
> I am having the following the problem:
>
> 2019-05-06T10:46:16.413758+00:00 d98-f2-b3-c8-b8-65 kernel: [8649388.905125] ../mm/pgtable-generic.c:33: bad pmd ffff883eed530398(80000037db2008e7)
> 2019-05-06T10:46:16.423251+00:00 d98-f2-b3-c8-b8-65 kernel: [8649388.915722] 34_scheduler[15882]: segfault at ffffffffffffffff ip 0000000040e97113 sp 00007fcffdb7ef78 error 5
>
> it causes a crash without erlang dump.
>
> HIPE is enabled, do you have an idea if is related in some way to hipe?
>
> notes:
> 1 - Erlang version 20.2.2
> 2 - 34_scheduler` is one of the erlang schedulers

This is a kernel bug, plain and simple.  OTP and/or HiPE _might_ do
something that triggers it, but that's just coincidental; a correct
kernel must not crash due to (unprivileged) user-space actions.

You don't say what kernel you're running, but you should consider upgrading it.

(Long shot: may also be a bug in the C compiler used to compile the
kernel, but that's extremely uncommon for vanilla distribution
kernels.)
_______________________________________________
erlang-questions mailing list
[hidden email]
http://erlang.org/mailman/listinfo/erlang-questions
Reply | Threaded
Open this post in threaded view
|

Re: Erlang scheduler segfault at ffffffffffffffff

Gabriele Santomaggio
Thank you Mikael,

We run:

uname -a
Linux [x]  4.4.121-92.98-default #1 SMP Fri Oct 19 07:52:13 UTC 2018 (e4d85ce) x86_64 x86_64 x86_64 GNU/Linux

We will update the kernel.

-
Gabriele Santomaggio


Il giorno mar 21 mag 2019 alle ore 13:26 Mikael Pettersson <[hidden email]> ha scritto:
On Mon, May 20, 2019 at 9:20 PM Gabriele Santomaggio
<[hidden email]> wrote:
>
> Hi,
> I am having the following the problem:
>
> 2019-05-06T10:46:16.413758+00:00 d98-f2-b3-c8-b8-65 kernel: [8649388.905125] ../mm/pgtable-generic.c:33: bad pmd ffff883eed530398(80000037db2008e7)
> 2019-05-06T10:46:16.423251+00:00 d98-f2-b3-c8-b8-65 kernel: [8649388.915722] 34_scheduler[15882]: segfault at ffffffffffffffff ip 0000000040e97113 sp 00007fcffdb7ef78 error 5
>
> it causes a crash without erlang dump.
>
> HIPE is enabled, do you have an idea if is related in some way to hipe?
>
> notes:
> 1 - Erlang version 20.2.2
> 2 - 34_scheduler` is one of the erlang schedulers

This is a kernel bug, plain and simple.  OTP and/or HiPE _might_ do
something that triggers it, but that's just coincidental; a correct
kernel must not crash due to (unprivileged) user-space actions.

You don't say what kernel you're running, but you should consider upgrading it.

(Long shot: may also be a bug in the C compiler used to compile the
kernel, but that's extremely uncommon for vanilla distribution
kernels.)


--
Gabriele Santomaggio

_______________________________________________
erlang-questions mailing list
[hidden email]
http://erlang.org/mailman/listinfo/erlang-questions