2012-05-06 23:50:10

by Christoph Bartoschek

Subject: Strange behaviour after uptime of 208-209 days

Hi,

we run kernel 2.6.37.6 from opensuse 11.4. All machines with uptimes more
than 208 days show strange behaviour. The scheduler seems to avoid some
cores. For example on a 12 core machine only 3 cores are used. I see the
following messages in the logfiles

May 6 04:16:50 r1106i14 kernel: [18446743865.627390] BUG: soft lockup -
CPU#1 stuck for 4278190091s! [bonnRoute:12613]
May 6 04:16:51 r1106i14 kernel: [18446743866.001912] BUG: soft lockup -
CPU#6 stuck for 4278190091s! [bonnRoute:25309]
May 6 04:16:51 r1106i14 kernel: [18446743866.676048] BUG: soft lockup -
CPU#15 stuck for 4278190091s! [bonnRoute:28259]
May 6 04:16:43 r1106i11 kernel: [18446743821.077585] BUG: soft lockup -
CPU#2 stuck for 4278190091s! [chipbench:14254]
May 6 04:16:43 r1106i11 kernel: [18446743821.152489] BUG: soft lockup -
CPU#3 stuck for 4278190091s! [chipbench:14246]
May 6 04:16:43 r1106i11 kernel: [18446743821.227393] BUG: soft lockup -
CPU#4 stuck for 4278190091s! [chipbench:14220]
May 6 04:16:44 r1106i11 kernel: [18446743821.302297] BUG: soft lockup -
CPU#5 stuck for 4278190091s! [chipbench:14271]
May 6 04:16:44 r1106i11 kernel: [18446743821.452108] BUG: soft lockup -
CPU#7 stuck for 4278190092s! [chipbench:14190]
May 6 04:16:44 r1106i11 kernel: [18446743821.527011] BUG: soft lockup -
CPU#8 stuck for 4278190092s! [chipbench:14173]
May 6 04:16:44 r1106i11 kernel: [18446743821.601915] BUG: soft lockup -
CPU#9 stuck for 4278190091s! [chipbench:14162]
May 6 04:16:44 r1106i11 kernel: [18446743821.676820] BUG: soft lockup -
CPU#10 stuck for 4278190091s! [chipbench:14296]
May 6 04:16:44 r1106i11 kernel: [18446743821.751724] BUG: soft lockup -
CPU#11 stuck for 4278190091s! [chipbench:14203]


The representation of the kernel time stamp is near to 2^64. Is there an
integer overflow involved?

Could you please tell me which kernel version fixed this bug?

Thanks
Christoph

2012-05-07 04:22:14

by Mike Galbraith

Subject: Re: Strange behaviour after uptime of 208-209 days

On Mon, 2012-05-07 at 01:33 +0200, Christoph Bartoschek wrote:
> Hi,
>
> we run kernel 2.6.37.6 from opensuse 11.4. All machines with uptimes more
> than 208 days show strange behaviour. The scheduler seems to avoid some
> cores. For example on a 12 core machine only 3 cores are used. I see the
> following messages in the logfiles

Ah, 208 day bugfixes went to stable, but not to 2.6.37. Dunno if that
kernel is still being maintained, but please file a bug with opensuse.

-Mike