On Sun, 2011-07-10 at 16:49 -0400, Christoph Hellwig wrote: > Signed-off-by: Christoph Hellwig <hch@xxxxxx> OK, I guess this is fine. I have a suggestion below but I see nothing truly wrong with what you have. Reviewed-by: Alex Elder <aelder@xxxxxxx> > Index: xfs/fs/xfs/xfs_dir2_leaf.c > =================================================================== > --- xfs.orig/fs/xfs/xfs_dir2_leaf.c 2011-07-09 12:07:49.518499801 +0200 > +++ xfs/fs/xfs/xfs_dir2_leaf.c 2011-07-09 12:19:48.171796748 +0200 > @@ -148,6 +148,36 @@ xfs_dir2_block_to_leaf( > return 0; > } > > +STATIC void > +xfs_dir2_leaf_find_stale( > + struct xfs_dir2_leaf *leaf, > + int index, > + int *lowstale, > + int *highstale) > +{ > + /* > + * Find the first stale entry before our index, if any. > + */ > + for (*lowstale = index - 1; > + *lowstale >= 0 && > + leaf->ents[*lowstale].address != > + cpu_to_be32(XFS_DIR2_NULL_DATAPTR); > + --*lowstale) > + continue; > + > + /* > + * Find the first stale entry at or after our index, if any. > + * Stop if the answer would be worse than lowstale. Stop if the result would require moving more entries than using lowstale. (I realize you didn't change this comment, you just moved it into this helper function.) Actually it seems like this searching of the stale entries and moving things around among them could be broken into a few even finer-grained utility routines. It just seems like what this code is doing is simpler than what the code complexity suggests (though I haven't really looked at this stuff much before). > + */ > + for (*highstale = index; > + *highstale < be16_to_cpu(leaf->hdr.count) && > + leaf->ents[*highstale].address != > + cpu_to_be32(XFS_DIR2_NULL_DATAPTR) && > + (*lowstale < 0 || index - *lowstale > *highstale - index); > + ++*highstale) > + continue; > +} > + > struct xfs_dir2_leaf_entry * > xfs_dir2_leaf_find_entry( > xfs_dir2_leaf_t *leaf, /* leaf structure */ . . . _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs