Re: [PATCH v2] xfs: introduce xfs_validate_stripe_geometry()

[Brian Foster <bfoster@redhat.com>]

On Tue, Oct 13, 2020 at 09:55:37PM +0800, Gao Xiang wrote:
> Hi Brian,
> 
> On Tue, Oct 13, 2020 at 09:44:11AM -0400, Brian Foster wrote:
> > On Tue, Oct 13, 2020 at 11:48:53AM +0800, Gao Xiang wrote:
> > > Introduce a common helper to consolidate stripe validation process.
> > > Also make kernel code xfs_validate_sb_common() use it first.
> > > 
> > > Signed-off-by: Gao Xiang <hsiangkao@redhat.com>
> > > ---
> > > v1: https://lore.kernel.org/r/20201009050546.32174-1-hsiangkao@redhat.com
> > > 
> > > changes since v1:
> > >  - rename the helper to xfs_validate_stripe_geometry() (Brian);
> > >  - drop a new added trailing newline in xfs_sb.c (Brian);
> > >  - add a "bool silent" argument to avoid too many error messages (Brian).
> > > 
> > >  fs/xfs/libxfs/xfs_sb.c | 70 +++++++++++++++++++++++++++++++++++-------
> > >  fs/xfs/libxfs/xfs_sb.h |  3 ++
> > >  2 files changed, 62 insertions(+), 11 deletions(-)
> > > 
> > > diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c
> > > index 5aeafa59ed27..9178715ded45 100644
> > > --- a/fs/xfs/libxfs/xfs_sb.c
> > > +++ b/fs/xfs/libxfs/xfs_sb.c
> > > @@ -360,21 +360,18 @@ xfs_validate_sb_common(
> > >  		}
> > >  	}
> > >  
> > > -	if (sbp->sb_unit) {
> > > -		if (!xfs_sb_version_hasdalign(sbp) ||
> > > -		    sbp->sb_unit > sbp->sb_width ||
> > > -		    (sbp->sb_width % sbp->sb_unit) != 0) {
> > > -			xfs_notice(mp, "SB stripe unit sanity check failed");
> > > -			return -EFSCORRUPTED;
> > > -		}
> > > -	} else if (xfs_sb_version_hasdalign(sbp)) {
> > > +	/*
> > > +	 * Either (sb_unit and !hasdalign) or (!sb_unit and hasdalign)
> > > +	 * would imply the image is corrupted.
> > > +	 */
> > > +	if (!sbp->sb_unit ^ !xfs_sb_version_hasdalign(sbp)) {
> > 
> > This can be simplified to drop the negations (!), right?
> 
> Thanks for the suggestion.
> 
> yet nope, honestly I don't think so, the reason is that sbp->sb_unit is
> an integer here rather than a boolean, so negations cannot be
> simplified and I think it's simpliest now... (some boolean algebra...)
> 

Oh, right. So you'd actually need something like (!!sunit ^ hasdalign())
to avoid the bit operation.

Brian

> > 
> > >  		xfs_notice(mp, "SB stripe alignment sanity check failed");
> 
> ...
> 
> > > +	if (sectorsize && sunit % sectorsize) {
> > > +		if (!silent)
> > > +			xfs_notice(mp,
> > > +"stripe unit (%lld) must be a multiple of the sector size (%d)",
> > > +				   sunit, sectorsize);
> > > +		return false;
> > > +	}
> > > +
> > > +	if (sunit && !swidth) {
> > > +		if (!silent)
> > > +			xfs_notice(mp,
> > > +"invalid stripe unit (%lld) and stripe width of 0", sunit);
> > > +		return false;
> > > +	}
> > > +
> > > +	if (!sunit && swidth) {
> > > +		if (!silent)
> > > +			xfs_notice(mp,
> > > +"invalid stripe width (%lld) and stripe unit of 0", swidth);
> > > +		return false;
> > > +	}
> > > +
> > > +	if (sunit > swidth) {
> > > +		if (!silent)
> > > +			xfs_notice(mp,
> > > +"stripe unit (%lld) is larger than the stripe width (%lld)", sunit, swidth);
> > > +		return false;
> > > +	}
> > > +
> > > +	if (sunit && (swidth % sunit)) {
> > 
> > It might be good to use (or not) params consistently. I.e., the
> > sectorsize check earlier in the function has similar logic structure but
> > drops the params.
> 
> Yeah, that is due to the line was copied from somewhere else... so...
> Anyway, I can resend a quick fix for this if needed. Wait a sec
> for some potential feedback...
> 
> Thanks,
> Gao Xiang
> 
> > 
> > Those nits aside:
> > 
> > Reviewed-by: Brian Foster <bfoster@redhat.com>
> > 
>