Re: [PATCH RFC 2/2] fs: fully sync all fses even for an emergency sync

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

在 2025/11/3 22:34, Christoph Hellwig 写道:
> Forgot to add, but I think the main issue here is that btrfs clears the
> writeback flag on the folio before commiting the metadata for the
> writeback.

During my debug runs, the metadata are properly written back to storage, 
during the sync_inodes_one_sb() calls.

So that doesn't seem to be the problem.

>  I tried to fix this a few years ago, and IIRC Josef tried
> again a few month ago.  If that is fixed, even a non-blocking writeback
> will commit all the updates it actually kicked off, just like for other
> file systems. (It will also sort out all kinds of annoying locking
> issues)

The problem is still there, even if all new metadata are committed, the 
super block is still not, thus after a power loss we are still seeing 
the old metadata, not the newer ones.

Thanks,
Qu