Re: Shrinking XFS - is that happening?

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

On Mon, Jan 12, 2026 at 08:05:12PM +0530, Nirjhar Roy (IBM) wrote:
> 
> On 8/11/25 20:55, Darrick J. Wong wrote:
> > On Mon, Aug 11, 2025 at 03:48:18AM -0700, Christoph Hellwig wrote:
> > > On Tue, Aug 05, 2025 at 10:20:15AM +0530, Nirjhar Roy (IBM) wrote:
> > > > On Mon, 2025-08-04 at 17:13 +0200, Roy Sigurd Karlsbakk wrote:
> > > > > Hi all!
> > > > > 
> > > > > I beleive I heard something from someone some time back about work in progress on shrinking xfs filesystems. Is this something that's been worked with or have I been lied to or just had a nice dream?
> > > > > 
> > > > > roy
> > > > I have recently posted an RFC[1]. The work is based/inspired from an old RFC[2] by Gao and ideas
> > > > given by Dave Chinner.
> > > Like the previous attempts it doesn't seem to include an attempt to
> > > address the elephant in the room:  moving inodes out of the to be
> > > removed AGs or tail blocks of an AG.
> > Anyone who wants to finish the evacuation part is welcome to pick this
> > up:
> > https://lore.kernel.org/linux-xfs/173568777852.2709794.6356870909327619205.stgit@frogsfrogsfrogs/
> 
> Hi Darrick,
> 
> I am working on the extending my online shrink fs work [1] to realtime fs
> shrink and as a part of that project, I am looking into the patchbomb above
> - specifically " [PATCHSET 3/5] xfsprogs: defragment free space". I have a
> high level questions before I delve deep into the patches above:
> 
> 1. Can you please elaborate a bit on what do you mean by "finish the
> evacuation part"? I took a very high level look at the patch, and it seems
> it is already adding support to move data from one rtg/AG to another?

It can evacuate fully mapped blocks, but it can't do that for the EOF
block if the EOF is in the middle of that block.

--D

> --NR
> 
> > 
> > --D
> 
> -- 
> Nirjhar Roy
> Linux Kernel Developer
> IBM, Bangalore
> 
>