Re: [PATCH 0/5 v2] fs: Deferred inode reclaim

["Darrick J. Wong" <djwong@kernel.org>]

On Fri, Sep 11, 2026 at 10:55:53AM -0400, Theodore Tso wrote:
> On Fri, Sep 11, 2026 at 10:51:36AM -0500, Jan Kara wrote:
> > 
> > after a long pause here is a second revision of my patches implementing
> > deferred inode reclaim to deal with MM warnings due to GFP_NOFAIL allocations
> > from reclaim paths. This happens because to reclaim some inodes, filesystems
> > have to do IO including complex operations using journalling and forward
> > progress of these depends on successful memory allocations (which is impossible
> > to guarantee from reclaim context).
> 
> Thanks for the patch set!  I wonder if it is worthwhile to wait for
> inode reclaim before freezing a file system for suspend.  On the plus
> side, it leaves the file system in a more cleaned-up state, and
> reduces the amount of memory that might need to be written in a
> hibernation scenario.  On the other side of the argument there's more
> opportunity for freeze/suspend deadlocks, and more complexity.  So
> maybe it's not worth it.
> 
> What do folks think?

xfs doesn't flush any pending inode reclaim work on freeze, since the
log will be dirty and log recovery (if the frozen fs image gets
remounted) will take care of that.  Freezing to make a snapshot is fast,
mounting the snapshot is not.

--D

> 						- Ted
>