Re: [PATCH 1/4] xfs_repair: implement custom ifork verifiers

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

On Tue, Mar 20, 2018 at 02:54:30PM -0500, Eric Sandeen wrote:
> On 3/19/18 10:08 PM, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> > 
> > There are a few cases where an early stage of xfs_repair will write an
> > invalid inode fork buffer to signal to a later stage that it needs to
> > correct the value.  This happens in phase 4 when we detect an inline
> > format directory with an invalid .. pointer.  To avoid triggering the
> > ifork verifiers on this, inject a custom verifier for phase 6 that lets
> > this pass for now.
> > 
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > ---
> >  libxfs/libxfs_api_defs.h |    2 +
> >  repair/phase6.c          |   66 +++++++++++++++++++++++++++++++++++++++++++++-
> >  2 files changed, 67 insertions(+), 1 deletion(-)
> > 
> > 
> > diff --git a/libxfs/libxfs_api_defs.h b/libxfs/libxfs_api_defs.h
> > index 5d56340..78daca0 100644
> > --- a/libxfs/libxfs_api_defs.h
> > +++ b/libxfs/libxfs_api_defs.h
> > @@ -150,5 +150,7 @@
> >  #define xfs_rmap_lookup_le_range	libxfs_rmap_lookup_le_range
> >  #define xfs_refc_block			libxfs_refc_block
> >  #define xfs_rmap_compare		libxfs_rmap_compare
> > +#define xfs_dir_get_ops			libxfs_dir_get_ops
> > +#define xfs_default_ifork_ops		libxfs_default_ifork_ops
> >  
> >  #endif /* __LIBXFS_API_DEFS_H__ */
> > diff --git a/repair/phase6.c b/repair/phase6.c
> > index aff83bc..e9189af 100644
> > --- a/repair/phase6.c
> > +++ b/repair/phase6.c
> > @@ -39,6 +39,70 @@ static struct xfs_name		xfs_name_dot = {(unsigned char *)".",
> >  						XFS_DIR3_FT_DIR};
> >  
> >  /*
> > + * When we're checking directory inodes, we're allowed to set a directory's
> 
> (a shortform directory only?)

I think we do it for any directory, but it's only the shortform dirs
that require this fix.

> > + * dotdot entry to zero to signal that the parent needs to be reconnected
> > + * during phase 6.  The ifork verifiers would normally fail that, but we'll
> > + * accept this canary so that we can fix the dir.
> 
> hm we actually just replace it temporarily, potato/potahto?
> 
> > + */
> > +static xfs_failaddr_t
> > +phase6_verify_dir(
> > +	struct xfs_inode		*ip)
> > +{
> > +	struct xfs_mount		*mp = ip->i_mount;
> > +	const struct xfs_dir_ops	*dops;
> > +	struct xfs_ifork		*ifp;
> > +	struct xfs_dir2_sf_hdr		*sfp;
> > +	xfs_failaddr_t			fa;
> > +	xfs_ino_t			old_parent;
> > +	bool				parent_bypass = false;
> > +	int				size;
> > +
> > +	dops = libxfs_dir_get_ops(mp, NULL);
> > +
> > +	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
> > +	sfp = (struct xfs_dir2_sf_hdr *)ifp->if_u1.if_data;
> > +	size = ifp->if_bytes;
> > +
> > +	/* Don't let the NULLFSINO .. entry blow everything up. */
> 
> NULLFSINO is ((xfs_ino_t)-1) not zero, so is this comment accurate?

Oops. :)

> Maybe an explicit comment here about this being for shortform dirs?
> 
> 	/*
> 	 * If this is a shortform directory, phase4 may have set the parent
> 	 * inode to zero to indicate that it must be fixed.  Temporarily
> 	 * set a valid parent so that the directory verifier will pass.
> 	 */

Much better comment, let's go with that.

> > +	if (size > offsetof(struct xfs_dir2_sf_hdr, parent) &&
> > +	    size >= xfs_dir2_sf_hdr_size(sfp->i8count)) {
> > +		old_parent = dops->sf_get_parent_ino(sfp);
> > +		if (old_parent == 0) {
> > +			dops->sf_put_parent_ino(sfp, mp->m_sb.sb_rootino);
> > +			parent_bypass = true;
> > +		}
> > +	}
> > +
> > +	fa = libxfs_default_ifork_ops.verify_dir(ip);
> > +
> > +	/* Put it back. */
> 
> 	/* Put the special parent == 0 back in place */
> 
> > +	if (parent_bypass)
> > +		dops->sf_put_parent_ino(sfp, old_parent);
> > +
> > +	return fa;
> > +}
> > +
> > +static xfs_failaddr_t
> > +phase6_verify_attr(
> > +	struct xfs_inode		*ip)
> > +{
> > +	return libxfs_default_ifork_ops.verify_attr(ip);
> > +}
> 
> Is there a reason for these wrappers vs. just populating the
> custom ifork_ops with xfs_attr_shortform_verify and
> xfs_symlink_shortform_verify?

gcc whines about non-const expressions.  I tried adding const to
everything that touches an ifork_ops but it still wouldn't compile.

> > +
> > +static xfs_failaddr_t
> > +phase6_verify_symlink(
> > +	struct xfs_inode		*ip)
> > +{
> > +	return libxfs_default_ifork_ops.verify_symlink(ip);
> > +}
> > +
> > +struct xfs_ifork_ops phase6_default_ifork_ops = {
> 
> Naming a "custom" verifier "default" seems counterintuitive,
> is there a reason for the "default" semantics I'm missing?  Not
> a huge deal, just makes me go "hmmm...."

-EBADNAME

phase6_ifork_ops, much better.

--D

> > +	.verify_attr	= phase6_verify_attr,
> > +	.verify_dir	= phase6_verify_dir,
> > +	.verify_symlink	= phase6_verify_symlink,
> > +};
> > +
> > +/*
> >   * Data structures used to keep track of directories where the ".."
> >   * entries are updated. These must be rebuilt after the initial pass
> >   */
> > @@ -2833,7 +2897,7 @@ process_dir_inode(
> >  
> >  	ASSERT(!is_inode_refchecked(irec, ino_offset) || dotdot_update);
> >  
> > -	error = -libxfs_iget(mp, NULL, ino, 0, &ip, NULL);
> > +	error = -libxfs_iget(mp, NULL, ino, 0, &ip, &phase6_default_ifork_ops);
> >  	if (error) {
> >  		if (!no_modify)
> >  			do_error(
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html