Re: [PATCH] libxfs: disallow filesystems with reverse mapping and reflink and realtime

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Wed, Sep 30, 2020 at 10:31:21AM -0500, Eric Sandeen wrote:
> On 9/30/20 9:58 AM, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> > 
> > Neither the kernel nor the code in xfsprogs support filesystems that
> > have (either reverse mapping btrees or reflink) enabled and a realtime
> > volume configured.  The kernel rejects such combinations and mkfs
> > refuses to format such a config, but xfsprogs doesn't check and can do
> > Bad Things, so port those checks before someone shreds their filesystem.
> > 
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> 
> seems fine in general but a couple thoughts...
> 
> > ---
> >  libxfs/init.c |   14 ++++++++++++++
> >  1 file changed, 14 insertions(+)
> > 
> > diff --git a/libxfs/init.c b/libxfs/init.c
> > index cb8967bc77d4..1a966084ffea 100644
> > --- a/libxfs/init.c
> > +++ b/libxfs/init.c
> > @@ -724,6 +724,20 @@ libxfs_mount(
> >  		exit(1);
> >  	}
> >  
> > +	if (xfs_sb_version_hasreflink(sbp) && sbp->sb_rblocks) {
> 
> Hm, we really don't use xfs_sb_version_hasrealtime() very
> consistently, but it might be worth doing here?

Nah, I'll move it to rtmount_init.

> I wish we had a feature flag to cross-ref against, a corruption in
> sb_rblocks will lead to an untouchable filesystem, but I guess there's
> nothing we can do about that.

I guess xfs_repair could add a -E killrt=1 flag that would read the sb,
zero out sb_rblocks, and pass that to libxfs_mount.

> Actually, would it help to cross-check against the rtdev arg as well?
> Should we do anything different if the user actually specified a
> realtime device on the commandline?

I doubt it?  I mean, the fs allege it has an rt volume and some
unsupported feature; it doesn't matter if the user did or didn't pass an
rtdev.

> I mean, I suppose 

you suppose...?

> 
> > +		fprintf(stderr,
> > +	_("%s: Reflink not compatible with realtime device. Please try a newer xfsprogs.\n"),
> 
> I like this optimism.  ;)

Optimism?

/me now has an xfsprogs that /does/ support rt rmap and reflink, though
at current patch review rates it won't hit the list until 2024, and
that's assuming I can keep ahead of all the bitrot in rtrmap...

--D

> 
> 
> > +				progname);
> > +		exit(1);
> > +	}
> > +
> > +	if (xfs_sb_version_hasrmapbt(sbp) && sbp->sb_rblocks) {
> > +		fprintf(stderr,
> > +	_("%s: Reverse mapping btree not compatible with realtime device. Please try a newer xfsprogs.\n"),
> > +				progname);
> > +		exit(1);
> > +	}
> > +
> >  	xfs_da_mount(mp);
> >  
> >  	if (xfs_sb_version_hasattr2(&mp->m_sb))
> >