> Only if the queue depth is not bound. Queue depths are bound and so
> the distance we can go over the threshold is limited. This is the
> fundamental principle on which the throttling is based.....
>
> Hence, if the queue is not full, then we will have either written
> dirty pages to it (i.e wbc->nr_write != write_chunk so we will throttle
> or continue normally if write_chunk was written) or we have no more
> dirty pages left.
>
> Having no dirty pages left on the bdi and it not being congested
> means we effectively have a clean, idle bdi. We should not be trying
> to throttle writeback here - we can't do anything to improve the
> situation by continuing to try to do writeback on this bdi, so we
> may as well give up and let the writer continue. Once we have dirty
> pages on the bdi, we'll get throttled appropriately.
OK, you convinced me.
How about this patch? I introduced a new wbc counter, that sums the
number of dirty pages encountered, including ones already under
writeback.
Dave, big thanks for your insights.
Miklos
Index: linux/include/linux/writeback.h
===================================================================
--- linux.orig/include/linux/writeback.h 2007-03-14 22:43:42.000000000 +0100
+++ linux/include/linux/writeback.h 2007-03-14 22:58:56.000000000 +0100
@@ -44,6 +44,7 @@ struct writeback_control {
long nr_to_write; /* Write this many pages, and decrement
this for each page written */
long pages_skipped; /* Pages which were not written */
+ long nr_dirty; /* Number of dirty pages encountered */
/*
* For a_ops->writepages(): is start or end are non-zero then this is
Index: linux/mm/page-writeback.c
===================================================================
--- linux.orig/mm/page-writeback.c 2007-03-14 22:41:01.000000000 +0100
+++ linux/mm/page-writeback.c 2007-03-14 23:00:20.000000000 +0100
@@ -220,6 +220,17 @@ static void balance_dirty_pages(struct a
pages_written += write_chunk - wbc.nr_to_write;
if (pages_written >= write_chunk)
break; /* We've done our duty */
+
+ /*
+ * If just a few dirty pages were encountered, and
+ * the queue is not congested, then allow this dirty
+ * producer to continue. This resolves the deadlock
+ * that happens when one filesystem writes back data
+ * through another. It should also help when a slow
+ * device is completely blocking other writes.
+ */
+ if (wbc.nr_dirty < 8 && !bdi_write_congested(bdi))
+ break;
}
congestion_wait(WRITE, HZ/10);
}
@@ -612,6 +623,7 @@ retry:
min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
unsigned i;
+ wbc->nr_dirty += nr_pages;
scanned = 1;
for (i = 0; i < nr_pages; i++) {
struct page *page = pvec.pages[i];
-
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
