Re: Critical error from Tree-checker

[Archange <archange@archlinux.org>]

Le 12/09/2024 à 01:42, Qu Wenruo a écrit :
> 在 2024/9/12 06:53, Qu Wenruo 写道:
>> 在 2024/9/12 06:50, Archange 写道:
>>> Le 12/09/2024 à 00:54, Qu Wenruo a écrit :
>>>> 在 2024/9/12 05:25, Archange 写道:
>>>>> Le 11/09/2024 à 01:37, Qu Wenruo a écrit :
>>>>>> 在 2024/9/11 06:58, Qu Wenruo 写道:
>>>>>>> 在 2024/9/11 06:35, Archange 写道:
>>>>>>> […]
>>>>
>>>> This looks exactly like another report that is caused by inode cache.
>>>>
>>>> So in that case, mind to try the following commands?
>>>>
>>>> # btrfs rescue zero-log <device>
>>>> # btrfs rescue clear-inode-cache <device>
>>>
>>> I supposed the second command was meant to be `clear-ino-cache` (I
>>> remember having to remove `inode_cache` from mount options some time 
>>> ago
>>> as it prevented booting, I had then discovered the related feature had
>>> been deprecated and removed).
>>>
>>> Here are the command outputs:
>>>
>>> # btrfs rescue zero-log /dev/mapper/rootext
>>> Clearing log on /dev/mapper/rootext, previous log_root 0, level 0
>>>
>>> # btrfs rescue clear-ino-cache /dev/mapper/rootext
>>> Successfully cleaned up ino cache for root id: 5
>>> Successfully cleaned up ino cache for root id: 257
>>> Successfully cleaned up ino cache for root id: 258
>>> corrupt node: root=7 block=647369064448 slot=0, invalid level for leaf,
>>> have 1 expect 0
>>
>> This is not expected, I guess I have to double check the inode clearing
>> code to be sure.
>
> It turns out to be a bug in the inode cache clearing code, that it holds
> a path meanwhile the fs can be modified halfway.
>
> I'll fix it soon, and keep you updated, as you may have some inode cache
> left.

OK, thanks.