Btrfs devel - Mar 2012 - attempt to access beyond end of device and livelock

Hi Dongyang, Yan,

When testing BTRFS with RAID 0 metadata on linux-3.3, we see discard
ranges exceeding the end of the block device [1], potentially causing
dataloss; when this occurs, filesystem writeback becomes catatonic due
to continual resubmission.

The reproducer is quite simple [2]. Hope this proves useful...

Thanks,
  Daniel

--- [1]

attempt to access beyond end of device
ram0: rw=129, want=8452072, limit=4096000
...

--- [2]

modprobe brd rd_size=2048000 (or boot with ramdisk_size=2048000)
mkfs.btrfs -m raid0 /dev/ram0 /dev/ram1
mount /dev/ram0 /mnt -o discard
fio testcase
umount /mnt

--- [3] testcase

[global]
directory=/mnt
rw=randread
size=256m
ioengine=libaio
iodepth=4
invalidate=1
direct=1

[bgwriter]
rw=randwrite
iodepth=32

[queryA]
iodepth=1
ioengine=mmap
thinktime=3

[queryB]
iodepth=1
ioengine=mmap
thinktime=1

[bgupdater]
rw=randrw
iodepth=16
thinktime=1
size=32m
-- 
Daniel J Blueman
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hi Josef, Chris,

When testing BTRFS with RAID 0 metadata on linux-3.4-rc1, we see
discard ranges exceeding the end of the block device [1], potentially
causing dataloss; when this occurs, filesystem writeback becomes
catatonic due to continual resubmission.

Simply mounting with discard a raid0 metadata filesystem and copying
some data in [2] provokes the issue.

Thanks,
 Daniel

--- [1]

attempt to access beyond end of device
ram0: rw=129, want=8452072, limit=4096000
...

--- [2]

modprobe brd rd_size=2048000 (or boot with ramdisk_size=2048000)
mkfs.btrfs -m raid0 /dev/ram0 /dev/ram1
mount /dev/ram0 /mnt -o discard
cd /mnt && tar -xvzf linux.tar.gz
<access beyond end of device and livelock>
-- 
Daniel J Blueman

On 04/06/2012 07:36 PM, Daniel J Blueman wrote:
> Hi Josef, Chris,
> 
> When testing BTRFS with RAID 0 metadata on linux-3.4-rc1, we see
> discard ranges exceeding the end of the block device [1], potentially
> causing dataloss; when this occurs, filesystem writeback becomes
> catatonic due to continual resubmission.
> 
> Simply mounting with discard a raid0 metadata filesystem and copying
> some data in [2] provokes the issue.
> 
> Thanks,
>  Daniel
> 
> --- [1]
> 
> attempt to access beyond end of device
> ram0: rw=129, want=8452072, limit=4096000
> ...
> 
> --- [2]
> 
> modprobe brd rd_size=2048000 (or boot with ramdisk_size=2048000)
> mkfs.btrfs -m raid0 /dev/ram0 /dev/ram1
> mount /dev/ram0 /mnt -o discard
> cd /mnt && tar -xvzf linux.tar.gz
> <access beyond end of device and livelock>
Thanks for the report, this bug shows we''ve miscalculated the length of
discard extents.

I''ll send a patch for this soon.

thanks,
-- 
liubo 
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On 8 April 2012 16:49, Liu Bo <liubo2009@cn.fujitsu.com>
wrote:
> On 04/06/2012 07:36 PM, Daniel J Blueman wrote:
>> Hi Josef, Chris,
>>
>> When testing BTRFS with RAID 0 metadata on linux-3.4-rc1, we see
>> discard ranges exceeding the end of the block device [1], potentially
>> causing dataloss; when this occurs, filesystem writeback becomes
>> catatonic due to continual resubmission.
[]
> Thanks for the report, this bug shows we''ve miscalculated the
length of discard extents.
>
> I''ll send a patch for this soon.
The patch test out well here on 3.4-rc2. Thanks Bo!

Daniel
-- 
Daniel J Blueman