On Sat, Nov 01, 2008 at 08:54:30AM +1100, Dave Chinner wrote:
> Effectively, yes.
> Currently we iterate inodes for data and "metadata" sync, and the
> only other concept is writing superblocks. I think most filesystems
> have more types of metadata than this, so it makes sense for sync to
> work on abstracts sync as data and metadata rather than data, inodes
> and superblocks...
Yes, absolutely. And for those that have inodes as primary / only
metadata besides superblock we can still provide a generic_sync_inodes
helper that just takes a callback to apply to every inode. Which we
probably want anyway as XFS is the only intree-filesystem that currently
has a more efficient way to iterate inodes.
> > And as we found out it's not just sync that gets it wrong, it's also
> > fsync (which isn't part of the above picture as it's per-inode) that
> > gets this utterly wrong, as well as all kinds of syncs, not just the
> > unmount one.
> Async writeback (write_inode()) has the same problem as fsync -
> writing the inode before waiting for data I/O to complete - which
> means we've got to jump through hoops in the filesystem to avoid
> blocking on inodes that can't be immediately flushed, and often we
> end up writing the inode multiple times and having to issue log
> forces whenw e shouldn't need to. Effectively we have to tell the
> VFS to "try again later" the entire time data is being flushed
> before we can write the inode and it's exceedingly inefficient.....
Yes, that was the couple of sync functions I meant above as the whole
inode writeback path is extremly convoluted - mostly due to the dirty
data vs metadata mixup mess.