Hi,
I just simply change the config file to avoid this. I also have noticed that other branches were fine. I think git-bisect can find the buggy commit easily. I'm too busy these days, or I will report it more hard.

----- Reply message -----
发件人: "Takahiro Akashi" <takahiro.akashi@linaro.org>
收件人: "Zhizhou Zhang (张治洲)" <Zhizhou.Zhang@spreadtrum.com>
抄送: "linaro-android@lists.linaro.org" <linaro-android@lists.linaro.org>, "linaro-kernel@lists.linaro.org" <linaro-kernel@lists.linaro.org>, "Mark Brown" <broonie@linaro.org>
主题: kernel crush because of CONFIG_PREEMPT
日期: 周五, 10月 3 日, 2014 年 10:45

Cc: Mark

I reproduced this problem with the latest branch on armv8 fast model.
But it is not a page fault, but:

Freeing unused kernel memory: 236K (ffffffc000592000 - ffffffc0005cd000)
BUG: scheduling while atomic: init/1/0x00000002
Modules linked in:
CPU: 0 PID: 1 Comm: init Not tainted 3.10.55+ #4
Call trace:
[<ffffffc000088354>] dump_backtrace+0x0/0x130
[<ffffffc000088494>] show_stack+0x10/0x1c
[<ffffffc000413514>] dump_stack+0x1c/0x28
[<ffffffc0000c6dbc>] __schedule_bug+0x48/0x5c
[<ffffffc0004157b0>] __schedule+0x54c/0x63c
[<ffffffc0004158c4>] schedule+0x24/0x68
[<ffffffc000413b54>] schedule_timeout+0x194/0x214
[<ffffffc000414fd4>] wait_for_common+0xa4/0x14c
[<ffffffc00041508c>] wait_for_completion+0x10/0x1c
[<ffffffc0000f8fec>] stop_one_cpu+0x44/0x68
[<ffffffc0000c96e0>] sched_exec+0xa0/0xcc
[<ffffffc00016d2e0>] do_execve_common.isra.32+0x1a4/0x4ac
[<ffffffc00016d600>] do_execve+0x18/0x24
[<ffffffc00040d538>] kernel_init+0x12c/0x140
Kernel panic - not syncing: No init found.  Try passing init= option to kernel. See Linux Documentation/init.txt for guidance.

- Takahiro AKASHI


On 23 September 2014 18:35, Zhizhou Zhang (张治洲) <Zhizhou.Zhang@spreadtrum.com> wrote:
>
> Hello Guys,
>
>
> I'm using branch linux-linaro-lsk-android on lsk(https://git.linaro.org/kernel/linux-linaro-stable.git). My arch is armv8.
>
> When I config CONFIG_PREEMPT, The system panic.
>
> It says "Failed to execute /init". I digged into it and found there was a page fault in atomic path.
>
> And When config CONFIG_PREEMPT_VOLUNTARY or CONFIG_PREEMPT_NONE, it looks well.
>
> Does anyone encounter this situation?
>
>
> --
> Zhizhou
>
> _______________________________________________
> linaro-kernel mailing list
> linaro-kernel@lists.linaro.org
> http://lists.linaro.org/mailman/listinfo/linaro-kernel
>