Re: TAKE 949916 - direct read vs delalloc

Re: TAKE 949916 - direct read vs delalloc

[Christoph Hellwig]

On Fri, Mar 17, 2006 at 04:30:19PM +1100, Nathan Scott wrote:
> Flush and invalidate dirty pages at the start of a direct read also,
> else we can hit a delalloc-extents-via-direct-io BUG.

If this is needed it should be done in the filemap.c code.

> 
> Date:  Fri Mar 17 16:29:59 AEDT 2006
> Workarea:  chook.melbourne.sgi.com:/build/nathans/xfs-linux
> Inspected by:  tes
> 
> The following file(s) were checked into:
>   longdrop.melbourne.sgi.com:/isms/xfs-kern/xfs-linux-melb
> 
> 
> Modid:  xfs-linux-melb:xfs-kern:25483a
> linux-2.6/xfs_lrw.c - 1.237 - changed
> http://oss.sgi.com/cgi-bin/cvsweb.cgi/xfs-linux/linux-2.6/xfs_lrw.c.diff?r1=text&tr1=1.237&r2=text&tr2=1.236&f=h
> 
> 
---end quoted text---

Re: TAKE 949916 - direct read vs delalloc

[Nathan Scott]

On Fri, Mar 17, 2006 at 05:57:19PM +0000, Christoph Hellwig wrote:
> On Fri, Mar 17, 2006 at 04:30:19PM +1100, Nathan Scott wrote:
> > Flush and invalidate dirty pages at the start of a direct read also,
> > else we can hit a delalloc-extents-via-direct-io BUG.
> 
> If this is needed it should be done in the filemap.c code.

Hmm, maybe, maybe not.

filemap.c code doesn't know much about delayed allocation, so
this won't bite other filesystems.  We do the same thing on
write and always have, so having it inside XFS is OK by me.

cheers.

-- 
Nathan