Re: Questions on ext4 and writeback

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

On Tue, Aug 18, 2009 at 09:57:49AM -0700, Curt Wohlgemuth wrote:
> Hi Ted:
> 
> On Tue, Aug 18, 2009 at 9:39 AM, Theodore Tso<tytso@mit.edu> wrote:
> > On Thu, Aug 13, 2009 at 09:09:58AM -0700, Curt Wohlgemuth wrote:
> >> But the actual submittal of I/O for the previous extent might cause
> >> pages_skipped to be bumped, right?  Removing these increments might cause
> >> the accounting to be incorrect, it seems to me.
> >
> > I don't see where the submission of an extent of pages for I/O would
> > cause page_skipped to be incremented or changed --- am I missing
> > something?
> 
> Probably not.  But it seems to me that a call order of
> 
>   write_cache_pages -> __mpage_da_writepage -> mpage_da_submit_io ->
> ext4_writepage
> 
> can cause pages_skipped to be incremented, either directly in
> ext4_writepage() (page has delayed/unwritted buffers) or in
> __block_write_full_page() (buffer already locked).
> 
> In fact, in mpage_da_submit_io(), pages_written is only incremented if
> pages_skipped hasn't been bumped -- so that routine already knows that
> pages_skipped might be changed on I/O submit.
> 
> If this happens, ext4_da_writepages() will wipe out the fact that
> pages_skipped was changed during submittal, won't it?
> 

In that case we would have set MPAGE_DA_EXTENT_TAIL and we loop again
in the writepages, trying to write again from the beginning and writing the
earlier skipped pages.

In otherwords we retry in writepages untill we are able to write all the
nr_to_write pages.

-aneesh
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html