lkml.org 
[lkml]   [2015]   [Jun]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC] panic when reboot the system
On Wed, 10 Jun 2015 15:17:23 +0800 Xishi Qiu <qiuxishi@huawei.com> wrote:

> Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000007
>
> Pid: 1, comm: init Tainted: G R O 3.4.24.19-0.11-default #1
> Call Trace:
> [<ffffffff8144dd24>] panic+0xc1/0x1e2
> [<ffffffff8104483b>] do_exit+0x7db/0x8d0
> [<ffffffff81044c7a>] do_group_exit+0x3a/0xa0
> [<ffffffff8105394b>] get_signal_to_deliver+0x1ab/0x5e0
> [<ffffffff81002270>] do_signal+0x60/0x5f0
> [<ffffffff8145bf97>] ? do_page_fault+0x4a7/0x4d0
> [<ffffffff81170d2c>] ? poll_select_copy_remaining+0xec/0x140
> [<ffffffff81002885>] do_notify_resume+0x65/0x80
> [<ffffffff8124ca7e>] ? trace_hardirqs_on_thunk+0x3a/0x3c
> [<ffffffff814587ab>] retint_signal+0x4d/0x92
>
>
> The system has a little memory left, then reboot it, and get the panic.
> Perhaps this is a bug and trigger it like this and the latest kernel maybe
> also have the problem.
>
> use a lot of memory
> wake up kswapd()
> reclaim some pages from init thread (pid=1)
> reboot
> shutdown the disk
> init thread read data from disk
> page fault, because the page has already reclaimed
> receive SIGBUS, and init thread exit
> trigger the panic
>
>

Interesting. 3.4 is a pretty old kernel but I expect at least some of
this remains.

- Why the heck did the disk get shut down while init still had pages
on it? We shouldn't be able to get that far without having done a
swapoff.

- Ignoring the above, as far as I can tell a regular old I/O error
during pagein of one of init's pages (anon or file-backed) will
result in the delivery of SIGBUS to init. If init isn't catching
SIGBUS (or if init's signal-handling code is also swapped out to a
bad sector??) then we're going to kill init and the system will panic
as above.

I'm not sure what we can do in this situation - init has
permanently lost some text or data and is hence dead. But panicing
the system doesn't seem the correct response.



\
 
 \ /
  Last update: 2015-06-10 23:21    [W:0.052 / U:0.608 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site