[btrfs-devel][patch]fix for btrfs_truncate_in_trans

[Chris Mason]

On Tue, 18 Sep 2007 16:28:41 +0200
Christoph Hellwig <hch at lst.de> wrote:

> On Tue, Sep 18, 2007 at 10:24:56AM -0400, Chris Mason wrote:
> > On Tue, 18 Sep 2007 16:22:33 +0200
> > Christoph Hellwig <hch at lst.de> wrote:
> > 
> > > On Tue, Sep 18, 2007 at 10:13:13AM -0400, Chris Mason wrote:
> > > > try_release_extent_mapping should also be changed not to drop
> > > > things from the extent mapping tree if they are
> > > > EXTENT_MAP_DELALLOC.
> > > 
> > > What XFS does here and what I plan to port over to the extent_map
> > > code is that ->releasepage will start the delalloc conversion so
> > > that the page is a simple dirty page and we can drop it, but not
> > > actually start I/O.
> > > 
> > 
> > Shouldn't this have already been kicked off by writepage?
> 
> delalloc conversion is also kicked of by writepage, but we can get
> into ->releasepage without having called writepage before.

Hmmm, I'm trying to avoid the case where we instantiate the delalloc
blocks just to truncate them away.  Not sure how often it would happen
though.

-chris