Re: [PATCH 4/8] xfs_scrub: fix spacemap external log device scan dev key

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

On Fri, Jun 26, 2026 at 07:03:45AM +0200, Christoph Hellwig wrote:
> On Thu, Jun 25, 2026 at 03:55:15PM -0700, Darrick J. Wong wrote:
> > --- a/scrub/spacemap.c
> > +++ b/scrub/spacemap.c
> > @@ -213,7 +213,7 @@ scan_rt_rmaps(
> >  	scan_dev_rmaps(ctx, ctx->fsinfo.fs_rtdev, arg);
> >  }
> >  
> > -/* Iterate all the reverse mappings of the log device. */
> > +/* Iterate all the reverse mappings of the external log device. */
> >  static void
> >  scan_log_rmaps(
> >  	struct workqueue	*wq,
> > @@ -222,7 +222,12 @@ scan_log_rmaps(
> >  {
> >  	struct scrub_ctx	*ctx = (struct scrub_ctx *)wq->wq_ctx;
> >  
> > -	scan_dev_rmaps(ctx, ctx->mnt.fsgeom.rtstart ? 2 : ctx->fsinfo.fs_logdev,
> > +	/*
> > +	 * Internal rt sections (rtstart != 0) means we use synthetic device
> > +	 * keys for external devices.
> > +	 */
> > +	scan_dev_rmaps(ctx, ctx->mnt.fsgeom.rtstart ? XFS_DEV_LOG :
> > +						      ctx->fsinfo.fs_logdev,
> 
> Maybe this should be using to_fsmap_dev instead?  Move the call
> to to_fsmap_dev into scan_dev_rmaps for further simplification.

Yeah.  phase7 could use from_fsmap_dev, so I'll hoist both to spacemap.h
and clean up the callsites.

> scan_rt_rmaps seems to have similar issues.

I don't agree with that since it's only used for pre-rtgroups rt volumes
which can't be internal, but it's trivial to clean it up along with the
other callsites so in it goes.

--D