Re: [PATCH v2] xfs: handle racing deletions in xfs_zone_gc_iter_irec

[Christoph Hellwig <hch@lst.de>]

On Mon, May 18, 2026 at 05:48:22PM +1000, Dave Chinner wrote:
> > -		if (error == -ENOENT) {
> > +		if (error == -ENOENT || error == -EINVAL) {
> >  			iter->rec_idx++;
> >  			goto retry;
> >  		}
> 
> Why did you choose to do this? I was expecting the updated fix to be
> dropping XFS_IGET_UNTRUSTED from the xfs_iget() call because the
> inode numbers are coming from a trusted source (i.e. the rmapbt) and
> this would then return the expected -ENOENT on unlink races instead
> of -EINVAL....

The inode number is not trusted.  It comes from a stale copy of the rmap,
and the inode cluster can have been freed and reallocated.  Dropping
XFS_IGET_UNTRUSTED would be highly dangerous as it could cause iget
to interprete something else as an inode struture in this case and
easily cause kernel crashes.