Re: kernel crashes with btrfs and busy database IO - how to debug?

[Qu Wenruo <quwenruo@cn.fujitsu.com>]

Now we can get the full backtrace.
That's a step forward

[45705.854778] BUG: unable to handle kernel NULL pointer dereference at 
0000000000000008
[45705.854824] IP: [<ffffffffc0158b8e>] 
btrfs_wait_pending_ordered+0x5e/0x110 [btrfs]
[45705.855615] Call Trace:
[45705.855637]  [<ffffffffc015addb>] 
btrfs_commit_transaction+0x40b/0xb60 [btrfs]
[45705.855671]  [<ffffffff810c0700>] ? prepare_to_wait_event+0x100/0x100
[45705.855698]  [<ffffffffc0171973>] btrfs_sync_file+0x313/0x380 [btrfs]
[45705.855721]  [<ffffffff81236c26>] vfs_fsync_range+0x46/0xc0
[45705.855740]  [<ffffffff81236cbc>] vfs_fsync+0x1c/0x20
[45705.855758]  [<ffffffff81236cf8>] do_fsync+0x38/0x70
[45705.855777]  [<ffffffff812370d0>] SyS_fsync+0x10/0x20
[45705.855796]  [<ffffffff8180cbb2>] system_call_fastpath+0x16/0x75

Also the hang seems to be highly related to the bug,
would you please send a new mail reporting the hang?

Thanks,
Qu

在 2015年06月14日 15:58, Tomasz Chmielewski 写道:
> On 2015-06-14 09:30, Tomasz Chmielewski wrote:
>> On 2015-06-13 08:23, Tomasz Chmielewski wrote:
>>
>>> I did get it from /var/crash/ though - is it more useful? I don't have
>>> vmlinux for this kernel though, but have just built 4.1-rc7 with the
>>> same config, can try to get the crash there.
>>
>> I've uploaded a crash dump and vmlinux here:
>>
>> http://www.virtall.com/files/temp/201506132321/
>>
>> Let me know if it's anything useful or if you need more info.
>
> I've tried running it the same procedure to get one more crash, but it
> didn't crash this time.
>
> Instead, btrfs is hanged on any writes - any processes trying to write
> get into D state and never return; there is no write activity when
> checking for example with iostat. "sync" command does not return.
>
> Reads from this btrfs filesystem are OK.
>
> I've uploaded the output of "echo w > /proc/sysrq-trigger" here:
>
> http://www.virtall.com/files/temp/dmesg.txt
>
>
> Tomasz Chmielewski
>