lkml.org 
[lkml]   [2021]   [Aug]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [f2fs-dev] f2fs do DIO write make file corruption
From
Date
On 2021/8/12 10:55, Wu Bo wrote:
>
> 在 2021/8/11 20:57, Chao Yu 写道:
>> On 2021/8/11 11:28, Wu Bo wrote:
>>>
>>> 在 2021/8/11 11:03, Chao Yu 写道:
>>>> On 2021/8/11 10:48, Wu Bo wrote:
>>>>> I use the following command to create a file, the file may got
>>>>> corruption:
>>>>>      f2fs_io write 2 0 512 inc_num dio $path
>>>>>
>>>>> And when I use bio or to set the chunk size to 1 block, the file is
>>>>> normal. The commands as following:
>>>>>      f2fs_io write 2 0 512 inc_num buffered $path
>>>>>      f2fs_io write 1 0 512 inc_num dio $path
>>>>>
>>>>> I find this bug on old kernel version 4.14.117, and not find on version
>>>>> 4.19.152. So this bug is fixed. Can anyone can tell me which patch
>>>>> fixed
>>>>> this bug?
>>>>
>>>> Not sure,
>>>>
>>>> https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-4.19.y&id=7bae8b6b73e46c307fa355ce086800b7ad6610f8
>>>>
>>
>> I didn't see this patch in 4.14 stable kernel of mailine, so do you mean
>> 4.14 kernel
>> maintained by Android?
>>
>> If so, f2fs codes in between 4.14 and 4.19 Android kernel are almost the
>> same,
>> see below link:
>>
>> https://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs-stable.git/
>>
>
> Sorry, it's our old Android project code.So the kernel version may have
> no reference value.
> This project code last sync to Android common kernel is:
> 9f892496fc0b:f2fs: readahead encrypted block during GC[2018-09-05]
> The code link is:
> https://android.googlesource.com/kernel/common/+/9f892496fc0b
> And after that, we applied some fixup patches.

I think you can pick up patches from your 4.19 f2fs codes which are not in
your 4.14 f2fs codes, and check each patch you picked to see whether there
is a patch can fix that problem.

Thanks,

>
>> Thanks,
>>
>>>>
>>>
>>> This patch is applied. The issue occurs when f2fs dio try to preallocate
>>> multiple blocks and got scattered disk blocks. The log as following:
>>>         my_f2fs_io-13425 [002] ....   395.583637: el0_irq_naked:
>>> 1:type:1,ino:40132,off:768,old_blk:0,new_blk:185764
>>>         my_f2fs_io-13425 [002] ....   395.583710: el0_irq_naked:
>>> 1:type:1,ino:40132,off:769,old_blk:0,new_blk:205824
>>>         my_f2fs_io-13425 [002] ....   395.583721: f2fs_map_blocks: dev =
>>> (259,23), ino = 40132, file offset = 768, start blkaddr = 0x2d5a4, len =
>>> 0x2, err = 0
>>>         my_f2fs_io-13425 [002] ....   395.583735: f2fs_map_blocks: dev =
>>> (259,23), ino = 40132, file offset = 768, start blkaddr = 0x2d5a4, len =
>>> 0x1, err = 0
>>>         my_f2fs_io-13425 [002] ....   395.583741: f2fs_map_blocks: dev =
>>> (259,23), ino = 40132, file offset = 769, start blkaddr = 0x32400, len =
>>> 0x1, err = 0
>>>
>>> And if the blocks are continuously, the file data is normal:
>>>       my_f2fs_io-13425 [002] ....   395.584037: el0_irq_naked:
>>> 1:type:1,ino:40132,off:770,old_blk:0,new_blk:205825
>>>         my_f2fs_io-13425 [002] ....   395.584066: el0_irq_naked:
>>> 1:type:1,ino:40132,off:771,old_blk:0,new_blk:205826
>>>         my_f2fs_io-13425 [002] ....   395.584077: f2fs_map_blocks: dev =
>>> (259,23), ino = 40132, file offset = 770, start blkaddr = 0x32401, len =
>>> 0x2, err = 0
>>>         my_f2fs_io-13425 [002] ....   395.584091: f2fs_map_blocks: dev =
>>> (259,23), ino = 40132, file offset = 770, start blkaddr = 0x32401, len =
>>> 0x2, err = 0
>>>
>>>>
>>>> Thanks,
>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Linux-f2fs-devel mailing list
>>>>> Linux-f2fs-devel@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>>>>>
>>>>
>>>
>>>
>>> _______________________________________________
>>> Linux-f2fs-devel mailing list
>>> Linux-f2fs-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>>>
>>

\
 
 \ /
  Last update: 2021-08-12 14:19    [W:0.034 / U:1.716 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site