Re: 2.6.20-rc3-mm1: umount reiser4 FS stuck in D state

[Edward Shishkin <edward@namesys.com>]

Laurent Riffard wrote:

>
>
> Le 01.02.2007 21:04, Edward Shishkin a écrit :
>
>> Laurent Riffard wrote:
>>
>>> Le 23.01.2007 16:46, Jens Axboe a écrit :
>>>
>>>> On Tue, Jan 23 2007, Vladimir V. Saveliev wrote:
>>>>
>>>>> Hello
>>>>>
>>>>> On Saturday 13 January 2007 01:56, Laurent Riffard wrote:
>>>>>
>>>>>> Le 06.01.2007 19:58, Vladimir V. Saveliev a écrit :
>>>>>>
>>>>>>> Hello
>>>>>>>
>>>>>>> On Saturday 06 January 2007 13:58, Laurent Riffard wrote:
>>>>>>>
>>>>>>>> Hello,
>>>>>>>>
>>>>>>>> got this with 2.6.20-rc3-mm1:
>>>>>>>>
>>>>>>>> =======================
>>>>>>>> SysRq : Show Blocked State
>>>>>>>>
>>>>>>>>                          free                        sibling
>>>>>>>>   task             PC    stack   pid father child younger older
>>>>>>>> umount        D C013135E  6044  1168   1150                     
>>>>>>>> (NOTLB)
>>>>>>>>        de591ae4 00000086 de591abc c013135e dff979c8 c012a6fe 
>>>>>>>> 00000046 00000007        dfd94ac0 128d3000 00000026 00000000 
>>>>>>>> dfd94bcc dff979c8 de591ae4 dffda038        00000002 dff979c0 
>>>>>>>> dff979bc dff979c8 de591b10 c012d600 dff979f8 00000000 Call Trace:
>>>>>>>>  [<c012d600>] synchronize_qrcu+0x70/0x8c
>>>>>>>>  [<c01bede4>] __make_request+0x4c/0x29b
>>>>>>>>  [<c01bd24b>] generic_make_request+0x1b0/0x1de
>>>>>>>>  [<c01bf354>] submit_bio+0xda/0xe2
>>>>>>>>  [<e12674bd>] write_jnodes_to_disk_extent+0x920/0x974 [reiser4]
>>>>>>>>  [<e12678dd>] update_journal_footer+0x29f/0x2b7 [reiser4]
>>>>>>>>  [<e1268b65>] write_tx_back+0x149/0x185 [reiser4]
>>>>>>>>  [<e126a8e7>] reiser4_write_logs+0xea4/0xfd2 [reiser4]
>>>>>>>>  [<e125626a>] try_commit_txnh+0x7e6/0xa4f [reiser4]
>>>>>>>>  [<e125661b>] reiser4_txn_end+0x148/0x3cf [reiser4]
>>>>>>>>  [<e12568ad>] reiser4_txn_restart+0xb/0x1a [reiser4]
>>>>>>>>  [<e125692f>] reiser4_txn_restart_current+0x73/0x75 [reiser4]
>>>>>>>>  [<e1256b89>] force_commit_atom+0x258/0x261 [reiser4]
>>>>>>>>  [<e1257703>] txnmgr_force_commit_all+0x406/0x697 [reiser4]
>>>>>>>>  [<e12e5e08>] release_format40+0x10c/0x193 [reiser4]
>>>>>>>>  [<e1279922>] reiser4_put_super+0x134/0x16a [reiser4]
>>>>>>>>  [<c015c930>] generic_shutdown_super+0x55/0xd8
>>>>>>>>  [<c015c9d3>] kill_block_super+0x20/0x32
>>>>>>>>  [<c015ca75>] deactivate_super+0x3f/0x51
>>>>>>>>  [<c016d903>] mntput_no_expire+0x42/0x5f
>>>>>>>>  [<c0160f37>] path_release_on_umount+0x15/0x18
>>>>>>>>  [<c016df77>] sys_umount+0x1a3/0x1cb
>>>>>>>>  [<c016dfb8>] sys_oldumount+0x19/0x1b
>>>>>>>>  [<c0103ed2>] sysenter_past_esp+0x5f/0x99
>>>>>>>>  =======================
>>>>>>>>
>>>>>>>> Scenario:
>>>>>>>> - umount a reiser4 FS (no need to write something before)
>>>>>>>
>>>>>>>
>>>>>>> Hmm, I can not reproduce this with 2.6.20-rc3-mm1. Probably I 
>>>>>>> need to config the kernel more close to your system.
>>>>>>>
>>>>>>>> Earlier kernels were OK.
>>>>>>>
>>>>>>>
>>>>>> This still happens with 2.6.20-rc4-mm1...
>>>>>>
>>>>>> Should I open a bug report at http://bugzilla.kernel.org?
>>>>>>
>>>>> Which device with reiser4 did you try to umount?  Jens wrote that it
>>>>> could be a barrier related. If there are no multidevices involved -
>>>>> please report to bugzilla.
>>>>
>>>>
>>>>
>>>> Make sure that your kernel contains this fix:
>>>>
>>>> http://git.kernel.dk/?p=linux-2.6-block.git;a=commitdiff;h=4af09c42ee1af70356471f51c1f40c1ff7881b68;hp=036f6008f43b5b4dd8c825365f15434d75005c6d 
>>>>
>>>>
>>>> I think it missed 2.6.20-rc3-mm1. Again, that assumes you are using md
>>>> or dm.
>>>
>>>
>>>
>>> I've got 2 reiser4 FS:
>>> - one with /dev/sdb6
>>> - the other with /dev/vglinux1/ccache (vglinux1 is built on 
>>> /dev/sda4 and /dev/sdb7).
>>> There is no md here, only dm.
>>>
>>> I applied the above patch on top of 2.6.20-rc4-mm1, but the problem 
>>> still happens with the two devices.
>>>
>>> thanks
>>
>>
>> Laurent, would you please try 2.6.20-rc6-mm3 + this patch:
>> http://lkml.org/lkml/diff/2007/2/1/195/1
>
>
> Reiser4 works fine with 2.6.20-rc6-mm2 or 2.6.20-rc6-mm3 without any 
> additional patch (it was broken in rc6-mm1).
>
> FWIW, Andrew removed git-block.patch from 2.6.20-rc6-mm2, and he 
> restored git-block.patch without some problematic CFQ updates in 
> 2.6.20-rc6-mm3.
>
> In this case, does this patch need testing in rc6-mm3 ?

Yes. This is against git-block patch to prevent endless waiting for IO 
completion.
I have reproduced it by ./iozone -B -a -n 524288 -f /mnt/foo
on x86 box with 512M RAM available.