On Tue, Jun 02, 2015 at 05:19:14PM -0700, Omar Sandoval wrote:
> I also can't figure out what that shrink_dcache_sb() is doing there.
> d_invalidate() already prunes the dentry cache under the deleted
> subvolume, but this clears the dcache for the whole filesystem, which
> could incur unnecessary overhead.
Correct analysis and the penalty is noticeable. In a directory where the
new/deleted subvolume flux is high. A 'ls' can take very long because all
the cached dentries are forcibly dropped and the metadata blocks have to be
read again.
> The call was added by efefb1438be2
> ("Btrfs: remove negative dentry when deleting subvolumne"), which fixes
> a problem in btrfs_dentry_delete(), but the commit message doesn't
> explain what shrink_dcache_sb() had to do with it. I'll send in an
> updated version with d_invalidate() moved and shrink_dcache_sb() removed
> and see if anyone can enlighten me.
I think it's a material for a separate patch.
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html