Skip to contentSkip to conversation

scheduling while atomic: swapper/0/0x10000002

DISCUSSION Christian Hoffmann 3 participants 6 messages Sep 10, 2011 – Sep 12, 2011

lore
Share & tools
lore
#0 lore raw

Hi,

I can see the following bug report in the kernel 3.1 rc5 dmesg:

[    0.000008] BUG: scheduling while atomic: swapper/0/0x10000002
[    0.000012] Modules linked in:
[    0.000015] Pid: 0, comm: swapper Not tainted 3.1.0-rc5-ch+ #2
[    0.000017] Call Trace:
[    0.000024]  [<ffffffff81052cb6>] __schedule_bug+0x66/0x70
[    0.000028]  [<ffffffff8160329d>] schedule+0x99d/0x9b0
[    0.000032]  [<ffffffff81cf22b5>] ? pidmap_init+0x9f/0xdf
[    0.000035]  [<ffffffff8105cc7a>] __cond_resched+0x2a/0x40
[    0.000038]  [<ffffffff81603541>] _cond_resched+0x31/0x40
[    0.000041]  [<ffffffff8115fb63>] kmem_cache_alloc_trace+0x53/0x160
[    0.000043]  [<ffffffff81cf22b5>] pidmap_init+0x9f/0xdf
[    0.000046]  [<ffffffff81cd9b8a>] start_kernel+0x333/0x3c8
[    0.000049]  [<ffffffff81cd9347>] x86_64_start_reservations+0x132/0x136
[    0.000051]  [<ffffffff81cd944c>] x86_64_start_kernel+0x101/0x110

Full dmesg output:
http://dl.dropbox.com/u/21820416/dmsg.log

Can I provide additional input?

Kind regards,
Chris

#1 lore raw
Quoted 4 lines · Christian Hoffmann

Hi,

I can see the following bug report in the kernel 3.1 rc5 dmesg:

Quoted 14 lines

[    0.000008] BUG: scheduling while atomic: swapper/0/0x10000002 [    0.000012] Modules linked in:
[    0.000015] Pid: 0, comm: swapper Not tainted 3.1.0-rc5-ch+ #2 [    0.000017] Call Trace:
[    0.000024]  [<ffffffff81052cb6>] __schedule_bug+0x66/0x70 [    0.000028]  [<ffffffff8160329d>] schedule+0x99d/0x9b0
[    0.000032]  [<ffffffff81cf22b5>] ? pidmap_init+0x9f/0xdf
[    0.000035]  [<ffffffff8105cc7a>] __cond_resched+0x2a/0x40 [    0.000038]  [<ffffffff81603541>] _cond_resched+0x31/0x40
[    0.000041]  [<ffffffff8115fb63>] kmem_cache_alloc_trace+0x53/0x160 [    0.000043]  [<ffffffff81cf22b5>] pidmap_init+0x9f/0xdf
[    0.000046]  [<ffffffff81cd9b8a>] start_kernel+0x333/0x3c8 [    0.000049]  [<ffffffff81cd9347>] x86_64_start_reservations+0x132/0x136 [    0.000051]  [<ffffffff81cd944c>] x86_64_start_kernel+0x101/0x110

Paul McKenney has a patch that solves this oops. You can find it here:

http://marc.info/?l=linux-kernel&m=131537669921538&w=2

We had discussed getting this into 3.1, but apparently that didn't happen (yet?). I thought it had been sent upstream with my tested-by, but I don't see it included in Linus' tree.

josh

#2 lore raw
Quoted 21 lines · Josh Boyer
Quoted 4 lines · Christian Hoffmann

Hi,

I can see the following bug report in the kernel 3.1 rc5 dmesg:

Quoted 14 lines
[    0.000008] BUG: scheduling while atomic: swapper/0/0x10000002
[    0.000012] Modules linked in:
[    0.000015] Pid: 0, comm: swapper Not tainted 3.1.0-rc5-ch+ #2
[    0.000017] Call Trace:
[    0.000024]  [<ffffffff81052cb6>] __schedule_bug+0x66/0x70
[    0.000028]  [<ffffffff8160329d>] schedule+0x99d/0x9b0
[    0.000032]  [<ffffffff81cf22b5>] ? pidmap_init+0x9f/0xdf
[    0.000035]  [<ffffffff8105cc7a>] __cond_resched+0x2a/0x40
[    0.000038]  [<ffffffff81603541>] _cond_resched+0x31/0x40
[    0.000041]  [<ffffffff8115fb63>] kmem_cache_alloc_trace+0x53/0x160
[    0.000043]  [<ffffffff81cf22b5>] pidmap_init+0x9f/0xdf
[    0.000046]  [<ffffffff81cd9b8a>] start_kernel+0x333/0x3c8
[    0.000049]  [<ffffffff81cd9347>] x86_64_start_reservations+0x132/0x136
[    0.000051]  [<ffffffff81cd944c>] x86_64_start_kernel+0x101/0x110
Quoted 7 lines

Paul McKenney has a patch that solves this oops. You can find it here:

http://marc.info/?l=linux-kernel&m=131537669921538&w=2

We had discussed getting this into 3.1, but apparently that didn't happen (yet?). I thought it had been sent upstream with my tested-by, but I don't see it included in Linus' tree.

Hi,

the patch up there doesn't seem to apply cleanly to 3.1 rc5, as it misses the rename done in "[PATCH tip/core/rcu 23/55] rcu: Simplify quiescent-state accounting" (http://marc.info/?l=linux-kernel&m=131537654621392&w=2).

Kind regards,
Chris

#3 lore raw
Quoted 30 lines · Christian Hoffmann
Quoted 21 lines · Josh Boyer
Quoted 4 lines · Christian Hoffmann

Hi,

I can see the following bug report in the kernel 3.1 rc5 dmesg:

Quoted 14 lines
[    0.000008] BUG: scheduling while atomic: swapper/0/0x10000002
[    0.000012] Modules linked in:
[    0.000015] Pid: 0, comm: swapper Not tainted 3.1.0-rc5-ch+ #2
[    0.000017] Call Trace:
[    0.000024]  [<ffffffff81052cb6>] __schedule_bug+0x66/0x70
[    0.000028]  [<ffffffff8160329d>] schedule+0x99d/0x9b0
[    0.000032]  [<ffffffff81cf22b5>] ? pidmap_init+0x9f/0xdf
[    0.000035]  [<ffffffff8105cc7a>] __cond_resched+0x2a/0x40
[    0.000038]  [<ffffffff81603541>] _cond_resched+0x31/0x40
[    0.000041]  [<ffffffff8115fb63>] kmem_cache_alloc_trace+0x53/0x160
[    0.000043]  [<ffffffff81cf22b5>] pidmap_init+0x9f/0xdf
[    0.000046]  [<ffffffff81cd9b8a>] start_kernel+0x333/0x3c8
[    0.000049]  [<ffffffff81cd9347>] x86_64_start_reservations+0x132/0x136
[    0.000051]  [<ffffffff81cd944c>] x86_64_start_kernel+0x101/0x110
Quoted 7 lines

Paul McKenney has a patch that solves this oops. You can find it here:

http://marc.info/?l=linux-kernel&m=131537669921538&w=2

We had discussed getting this into 3.1, but apparently that didn't happen (yet?). I thought it had been sent upstream with my tested-by, but I don't see it included in Linus' tree.

Quoted 6 lines

Hi,

the patch up there doesn't seem to apply cleanly to 3.1 rc5, as it misses the rename done in "[PATCH tip/core/rcu 23/55] rcu: Simplify quiescent-state accounting"
(http://marc.info/?l=linux-kernel&m=131537654621392&w=2).

For 3.1-rc5, you will be wanting this one:

	https://lkml.org/lkml/2011/8/19/355

							Thanx, Paul
#4 lore raw
Quoted 33 lines · Paul E. McKenney
Quoted 32 lines · Christian Hoffmann
Quoted 21 lines · Josh Boyer
Quoted 4 lines · Christian Hoffmann

Hi,

I can see the following bug report in the kernel 3.1 rc5 dmesg:

Quoted 14 lines
[    0.000008] BUG: scheduling while atomic: swapper/0/0x10000002
[    0.000012] Modules linked in:
[    0.000015] Pid: 0, comm: swapper Not tainted 3.1.0-rc5-ch+ #2
[    0.000017] Call Trace:
[    0.000024]  [<ffffffff81052cb6>] __schedule_bug+0x66/0x70
[    0.000028]  [<ffffffff8160329d>] schedule+0x99d/0x9b0
[    0.000032]  [<ffffffff81cf22b5>] ? pidmap_init+0x9f/0xdf
[    0.000035]  [<ffffffff8105cc7a>] __cond_resched+0x2a/0x40
[    0.000038]  [<ffffffff81603541>] _cond_resched+0x31/0x40
[    0.000041]  [<ffffffff8115fb63>] kmem_cache_alloc_trace+0x53/0x160
[    0.000043]  [<ffffffff81cf22b5>] pidmap_init+0x9f/0xdf
[    0.000046]  [<ffffffff81cd9b8a>] start_kernel+0x333/0x3c8
[    0.000049]  [<ffffffff81cd9347>] x86_64_start_reservations+0x132/0x136
[    0.000051]  [<ffffffff81cd944c>] x86_64_start_kernel+0x101/0x110
Quoted 7 lines

Paul McKenney has a patch that solves this oops. You can find it here:

http://marc.info/?l=linux-kernel&m=131537669921538&w=2

We had discussed getting this into 3.1, but apparently that didn't happen (yet?). I thought it had been sent upstream with my tested-by, but I don't see it included in Linus' tree.

Hi,

Quoted 8 lines
Quoted 4 lines

the patch up there doesn't seem to apply cleanly to 3.1 rc5, as it misses the rename done in "[PATCH tip/core/rcu 23/55] rcu: Simplify quiescent-state accounting"
(http://marc.info/?l=linux-kernel&m=131537654621392&w=2).

For 3.1-rc5, you will be wanting this one:

	https://lkml.org/lkml/2011/8/19/355

Hi,

yes, this applies and makes the stack disappear.

Thanks,
Chris

#5 lore raw
Quoted 39 lines · Christian Hoffmann
Quoted 38 lines · Paul E. McKenney
Quoted 36 lines · Christian Hoffmann
Quoted 21 lines · Josh Boyer
Quoted 4 lines · Christian Hoffmann

Hi,

I can see the following bug report in the kernel 3.1 rc5 dmesg:

Quoted 14 lines
[    0.000008] BUG: scheduling while atomic: swapper/0/0x10000002
[    0.000012] Modules linked in:
[    0.000015] Pid: 0, comm: swapper Not tainted 3.1.0-rc5-ch+ #2
[    0.000017] Call Trace:
[    0.000024]  [<ffffffff81052cb6>] __schedule_bug+0x66/0x70
[    0.000028]  [<ffffffff8160329d>] schedule+0x99d/0x9b0
[    0.000032]  [<ffffffff81cf22b5>] ? pidmap_init+0x9f/0xdf
[    0.000035]  [<ffffffff8105cc7a>] __cond_resched+0x2a/0x40
[    0.000038]  [<ffffffff81603541>] _cond_resched+0x31/0x40
[    0.000041]  [<ffffffff8115fb63>] kmem_cache_alloc_trace+0x53/0x160
[    0.000043]  [<ffffffff81cf22b5>] pidmap_init+0x9f/0xdf
[    0.000046]  [<ffffffff81cd9b8a>] start_kernel+0x333/0x3c8
[    0.000049]  [<ffffffff81cd9347>] x86_64_start_reservations+0x132/0x136
[    0.000051]  [<ffffffff81cd944c>] x86_64_start_kernel+0x101/0x110
Quoted 7 lines

Paul McKenney has a patch that solves this oops. You can find it here:

http://marc.info/?l=linux-kernel&m=131537669921538&w=2

We had discussed getting this into 3.1, but apparently that didn't happen (yet?). I thought it had been sent upstream with my tested-by, but I don't see it included in Linus' tree.

Hi,

the patch up there doesn't seem to apply cleanly to 3.1 rc5, as it misses the rename done in "[PATCH tip/core/rcu 23/55] rcu: Simplify quiescent-state accounting"
(http://marc.info/?l=linux-kernel&m=131537654621392&w=2).

Quoted 7 lines
Quoted 3 lines

For 3.1-rc5, you will be wanting this one:

	https://lkml.org/lkml/2011/8/19/355

Hi,

yes, this applies and makes the stack disappear.

Very good! I have added your Tested-by, along with Josh Boyer's.

Thanx, Paul