Re: Bisected BUG: using smp_processor_id() in preemptible

From: Joonsoo Kim
Date: Wed Jan 21 2015 - 20:52:54 EST


On Tue, Jan 20, 2015 at 10:38:32AM +0100, Krzysztof Kozlowski wrote:
> Hi,
>
>
> Since next-20150119 booting of Exynos4 based boards is nearly impossible
> because of continuous BUG on messages. The system finally boots... but
> console log is polluted with:
>
> [ 9.700828] BUG: using smp_processor_id() in preemptible [00000000] code: udevd/1656
> [ 9.708525] caller is kfree+0x8c/0x198
> [ 9.712229] CPU: 2 PID: 1656 Comm: udevd Tainted: G W 3.19.0-rc4-00279-gd2dc80750ee0 #1630
> [ 9.721501] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
> [ 9.727602] [<c0014980>] (unwind_backtrace) from [<c0011904>] (show_stack+0x10/0x14)
> [ 9.735324] [<c0011904>] (show_stack) from [<c0585bbc>] (dump_stack+0x70/0xbc)
> [ 9.742525] [<c0585bbc>] (dump_stack) from [<c01e79e0>] (check_preemption_disabled+0xf8/0x128)
> [ 9.751114] [<c01e79e0>] (check_preemption_disabled) from [<c00c501c>] (kfree+0x8c/0x198)
> [ 9.759265] [<c00c501c>] (kfree) from [<c0299b7c>] (uevent_show+0x38/0x104)
> [ 9.766210] [<c0299b7c>] (uevent_show) from [<c0299fb8>] (dev_attr_show+0x1c/0x48)
> [ 9.773763] [<c0299fb8>] (dev_attr_show) from [<c0122424>] (sysfs_kf_seq_show+0x8c/0x10c)
> [ 9.781920] [<c0122424>] (sysfs_kf_seq_show) from [<c0120f90>] (kernfs_seq_show+0x24/0x28)
> [ 9.790172] [<c0120f90>] (kernfs_seq_show) from [<c00e93b0>] (seq_read+0x1ac/0x480)
> [ 9.797806] [<c00e93b0>] (seq_read) from [<c00cacf4>] (__vfs_read+0x18/0x4c)
> [ 9.804833] [<c00cacf4>] (__vfs_read) from [<c00cada4>] (vfs_read+0x7c/0x100)
> [ 9.811950] [<c00cada4>] (vfs_read) from [<c00cae68>] (SyS_read+0x40/0x8c)
> [ 9.818810] [<c00cae68>] (SyS_read) from [<c000f160>] (ret_fast_syscall+0x0/0x30)
>
> I bisected this to:
>
> d2dc80750ee05ceb03c9b13b0531a782116d1ade
> Author: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> Date: Sat Jan 17 11:23:23 2015 +1100
> mm/slub: optimize alloc/free fastpath by removing preemption on/off
>
> Full dmesg and config attached.
>
> Any ideas?

Hello,

This issue will be fixed in next mmotm release.
Following patch is next version of
commit d2dc80750ee05ceb03c9b13b0531a782116d1ade.

https://lkml.org/lkml/2015/1/19/17

Thanks.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/