Re: [PATCH]: reiser4 [8/8] reiser4 code

[Nikita Danilov]

[[e-mail] is no longer subscribers-only, back into CC]

Alexander Viro writes:
 > 
 > 
 > On Thu, 31 Oct 2002, Nikita Danilov wrote:
 > 
 > > generic_shutdown_super() calls fsync_super(sb) (which will call
 > > ->writepage() on each dirty page) and then invalidate_inodes().
 > > 
 > > Reiser4 has commit out-standing transactions -between- these two points:
 > > after ->writepage() has been called on all dirty pages, but before
 > > inodes were destroyed. Thus, we cannot use
 > > kill_block_super()/generic_shutdown_super().
 > 
 > Why don't you do that from within fsync_super()?  That would be much
 > more natural point for such stuff...
 > 
 > I hadn't looked into akpm's stuff in fs-writeback.c for a while, but
 > if one can't stick such flush point in there I'd argue that this is
 > a bug that needs to be fixed - either there or by providing explicit

->writepages gets (in wbc->sync_mode) enough information to tell
synchronization request like umount or sync from
balance_dirty_pages(). Yes, it looks like better solution,
->writepages(inode, WB_SYNC_ALL) should just synchronously commit all
transactions involving any of inode's pages.

 > callback from fsync_super().

Hmm, I just noted that for now we probably could simply use exported
fsync_bdev(s->s_bdev) in fsync_super(s) stead. How simple. Thank you for
the useful discussion. :-)

Nikita.