On Tue 20-04-10 12:41:53, Dave Chinner wrote:
> From: Dave Chinner <dchinner@xxxxxxxxxx>
>
> If a filesystem writes more than one page in ->writepage, write_cache_pages
> fails to notice this and continues to attempt writeback when wbc->nr_to_write
> has gone negative - this trace was captured from XFS:
>
>
> wbc_writeback_start: towrt=1024
> wbc_writepage: towrt=1024
> wbc_writepage: towrt=0
> wbc_writepage: towrt=-1
> wbc_writepage: towrt=-5
> wbc_writepage: towrt=-21
> wbc_writepage: towrt=-85
>
> This has adverse effects on filesystem writeback behaviour. write_cache_pages()
> needs to terminate after a certain number of pages are written, not after a
> certain number of calls to ->writepage are made. Make it observe the current
> value of wbc->nr_to_write and treat a value of <= 0 as though it is a either a
> termination condition or a trigger to reset to MAX_WRITEḆACK_PAGES for data
> integrity syncs.
>
> Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> ---
> fs/fs-writeback.c | 9 ---------
> include/linux/writeback.h | 9 +++++++++
> include/trace/events/writeback.h | 1 +
> mm/page-writeback.c | 20 +++++++++++++++++++-
> 4 files changed, 29 insertions(+), 10 deletions(-)
>
<snip>
> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> index d45f59e..e22af84 100644
> --- a/mm/page-writeback.c
> +++ b/mm/page-writeback.c
> @@ -917,6 +917,7 @@ continue_unlock:
> if (!clear_page_dirty_for_io(page))
> goto continue_unlock;
>
> + trace_wbc_writepage(wbc);
> ret = (*writepage)(page, wbc, data);
> if (unlikely(ret)) {
> if (ret == AOP_WRITEPAGE_ACTIVATE) {
> @@ -935,7 +936,7 @@ continue_unlock:
> done = 1;
> break;
> }
> - }
> + }
>
> if (nr_to_write > 0) {
> nr_to_write--;
> @@ -955,6 +956,23 @@ continue_unlock:
> break;
> }
> }
> +
> + /*
> + * Some filesystems will write multiple pages in
> + * ->writepage, so wbc->nr_to_write can change much,
> + * much faster than nr_to_write. Check this as an exit
> + * condition, or if we are doing a data integrity sync,
> + * reset the wbc to MAX_WRITEBACK_PAGES so that such
> + * filesystems can do optimal writeout here.
> + */
> + if (wbc->nr_to_write <= 0) {
> + if (wbc->sync_mode == WB_SYNC_NONE) {
> + done = 1;
> + nr_to_write = 0;
> + break;
> + }
> + wbc->nr_to_write = MAX_WRITEBACK_PAGES;
> + }
Honestly, this is an ugly hack. I'd rather work towards ignoring
nr_to_write completely in WB_SYNC_ALL mode since it doesn't really make
any sence to say "write me *safely* 5 pages".
Honza
--
Jan Kara <jack@xxxxxxx>
SUSE Labs, CR
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
