On Thu, Nov 18, 2010 at 03:16:57PM +0100, Christoph Hellwig wrote: > On Fri, Nov 19, 2010 at 01:13:55AM +1100, Nick Piggin wrote: > > >> Anyway, what if writeback noticed pagecache was cleaned at this point, and then > > >> clears I_DIRTY bits from inode before you test it above? Won't that leave your > > >> metadata not on disk? > > > > I'm not sure if you answered this, though. Can't background writeout go through > > and clear the I_DIRTY_* bits on your inode, so that a subsequent fsync will skip > > required writeout because i_state is clean, but your private dirty > > bits are still set? > > > > That was my (poorly worded) concern. > > Good point - we should just skip the sync_inode_metadata for that case. > Which already is a no-op if no dirty bits are set, so just removing the > I_DIRTY check probably is the best thing we can do here. Hmm, it seems that your dirty bit handling is still broken and racy, along with half the other filesystems. I'll post a couple of patches to fsdevel for comment. -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
