Re: xfstests #078: Mild oops/kernel issue for small partitions

From: Jeff Liu <jeff.liu@oracle.com>
To: Dave Chinner <david@fromorbit.com>
Cc: "Michael L. Semon" <mlsemon35@gmail.com>, xfs@oss.sgi.com
Subject: Re: xfstests #078: Mild oops/kernel issue for small partitions
Date: Sat, 16 Mar 2013 16:52:40 +0800	[thread overview]
Message-ID: <514432D8.1090702@oracle.com> (raw)
In-Reply-To: <20130316072550.GH6369@dastard>

On 03/16/2013 03:25 PM, Dave Chinner wrote:
> On Sat, Mar 16, 2013 at 11:43:17AM +0800, Jeff Liu wrote:
>> Hi Michael,
>>
>> Thanks for the report.
>> On 03/16/2013 10:00 AM, Michael L. Semon wrote:
>>> Hi!  I was making another run with xfstests, using smaller partitions
>>> so I could use the larger ones for other tests, and I got a mild
>>> oops/crash of some sort.  This applies only to small partitions, first
>>> hit using 480MB partitions.  Also, CONFIG_XFS_DEBUG is set here, but I
>>> haven't tested the non-debug case lately.  This has been repeatable
>>> across various 3.8/3.9 kernels used recently.
>>>
>>> PC in this test is my Pentium III boat anchor, 512 MB RAM, 768 MB
>>> swap, no VGA/console, using serial and ssh to communicate.  Output was
>>> taken from the remote syslog facility and trimmed down.
>> So I just tried it against both 480MB and 512MB partitions, but I can not
>> reproduce it on 3.8/3.9-rc1 with xfsprogs 3.1.8.
>>
>> # gdisk -l /dev/sda
>> Number  Start (sector)    End (sector)  Size       Code  Name
>>    7       405682176       406730751   512.0 MiB   8300  Linux filesystem 
>>    8       406732800       407781375   512.0 MiB   8300  Linux filesystem
>>    9       407783424       408766463   480.0 MiB   8300  Linux filesystem
>>   10       408768512       409751551   480.0 MiB   8300  Linux filesystem
>>
>> # ./check 078
>> FSTYP         -- xfs (debug)
>> PLATFORM      -- Linux/x86_64 koala 3.9.0-rc1
> 
> 64 bit system
> 
>> MKFS_OPTIONS  -- -f -bsize=4096 /dev/sda10
>> MOUNT_OPTIONS -- /dev/sda10 /xfstest4
>>
>> 078 46s ... - output mismatch (see 078.out.bad)
>>     --- 078.out	2012-10-16 11:59:14.757395850 +0800
>>     +++ 078.out.bad	2013-03-16 11:39:10.945051414 +0800
>>     @@ -76,6 +76,7 @@
>>      === GROWFS (from 1t to 2500g, 4096 blocksize)
>>      
>>      *** mkfs loop file (size=1t)
>>     +mkfs.xfs: libxfs_device_zero write failed: No space left on device
> 
> it ran out of space creating the log device. That write error is
> likely what triggered the reported oops , which occurred in the
> buffered write error handling routine....
Yes.
> 
>>> This is a crash report.  No fix is requested.  Let me know if this
>>> needs to go somewhere else in the Linux development world as well.
> ....
>>> XFS (loop0): Mounting Filesystem
>>> XFS: Assertion failed: block_offset + from == pos, file:
>>> fs/xfs/xfs_aops.c, line: 1504
>>> ------------[ cut here ]------------
>>> kernel BUG at fs/xfs/xfs_message.c:100!
>>> invalid opcode: 0000 [#1]
>>> Pid: 1203, comm: mkfs.xfs Not tainted 3.9.0-rc1+ #1
>>> EIP: 0060:[<c10fa4f7>] EFLAGS: 00010286 CPU: 0
>>> EIP is at assfail+0x27/0x30
>>> EAX: 00000056 EBX: dfd7bd60 ECX: c141db1c EDX: 00000000
>>> ESI: 1bb72000 EDI: 00000000 EBP: dd037d70 ESP: dd037d5c
> 
> Oops occurred on a 32 bit system. That might be relevant...
Ah, I missed this difference, will build a 32bits VM to verify it again.

Thanks,
-Jeff

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs