On Tue, Apr 09, 2024 at 11:18:03PM -0700, Christoph Hellwig wrote: > > +static int > > +xfs_attr_ensure_iext( > > + struct xfs_da_args *args, > > + int nr) > > +{ > > + int error; > > + > > + error = xfs_iext_count_may_overflow(args->dp, XFS_ATTR_FORK, nr); > > + if (error == -EFBIG) > > + return xfs_iext_count_upgrade(args->trans, args->dp, nr); > > + return error; > > +} > > I'd rather get my consolidation of these merged instead of adding > a wrapper like this. Just waiting for my RT delalloc and your > exchrange series to hit for-next to resend it. Yeah, I made a mental note to scrub this function out if your patch wins the race. > > +/* > > + * Ensure that the xattr structure maps @args->name to @args->value. > > + * > > + * The caller must have initialized @args, attached dquots, and must not hold > > + * any ILOCKs. Only XATTR_CREATE may be specified in @args->xattr_flags. > > + * Reserved data blocks may be used if @rsvd is set. > > + * > > + * Returns -EEXIST if XATTR_CREATE was specified and the name already exists. > > + */ > > +int > > +xfs_attr_setname( > > Is there any case where we do not want to pass XATTR_CREATE, that > is replace an existing attribute when there is one? Yes, verity setup will use xfs_attr_setname to upsert a merkle tree block into the attr structure and obliterate stale blocks that might already have been there. > > +int > > +xfs_attr_removename( > > + struct xfs_da_args *args, > > + bool rsvd) > > +{ > > Is there a good reason to have a separate remove helper and not > overload a NULL value like we do for the normal xattr interface? xfs_repair uses xfs_parent_unset -> xfs_attr_removename to erase any XFS_ATTR_PARENT attribute that doesn't validate, so it needs to be able to pass in a non-NULL value. Perhaps I'll add a comment about that, since this isn't the first time this has come up. Come to think of it you can't removename a remote parent value, so I guess in that bad case xfs_repair will have to drop the entire attr structure <frown>. /* * Ensure that the xattr structure does not map @args->name to @args->value. * @args->value must be set for XFS_ATTR_PARENT removal (e.g. xfs_repair). * * The caller must have initialized @args, attached dquots, and must not hold * any ILOCKs. Reserved data blocks may be used if @rsvd is set. * * Returns -ENOATTR if the name did not already exist. */ --D