From: Jerome Marchand <jmarchan@redhat.com>
Date: Tue, 3 Mar 2009 17:27:27 +0100
Subject: [misc] IO accounting: write cancel accounting
Message-id: 20090303162957.826242871@dhcp-0-152.brq.redhat.com
O-Subject: [RHEL5.4 Patch 05/12] IO accounting: write cancel accounting
Bugzilla: 461636
RH-Acked-by: Jiri Pirko <jpirko@redhat.com>

bz461636

Account for the number of byte writes which this process caused to not happen
after all.

Patch is upstream:
commit e08748ce01e02f0ec154b141f392ccb9555333f4

diff --git a/fs/buffer.c b/fs/buffer.c
index 80c238d..fa218ec 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -3293,8 +3293,13 @@ int try_to_free_buffers(struct page *page)
 		 * could encounter a non-uptodate page, which is unresolvable.
 		 * This only applies in the rare case where try_to_free_buffers
 		 * succeeds but the page is not freed.
+		 *
+		 * Also, during truncate, discard_buffer will have marked all
+		 * the page's buffers clean.  We discover that here and clean
+		 * the page also.
 		 */
-		clear_page_dirty(page);
+		if (test_clear_page_dirty(page))
+			task_io_account_cancelled_write(PAGE_CACHE_SIZE);
 	}
 out:
 	if (buffers_to_free) {
diff --git a/mm/truncate.c b/mm/truncate.c
index 7728e45..69394aa 100644
--- a/mm/truncate.c
+++ b/mm/truncate.c
@@ -12,6 +12,7 @@
 #include <linux/module.h>
 #include <linux/pagemap.h>
 #include <linux/pagevec.h>
+#include <linux/task_io_accounting_ops.h>
 #include <linux/buffer_head.h>	/* grr. try_to_release_page,
 				   do_invalidatepage */
 
@@ -42,7 +43,8 @@ truncate_complete_page(struct address_space *mapping, struct page *page)
 	if (PagePrivate(page))
 		do_invalidatepage(page, 0);
 
-	clear_page_dirty(page);
+	if (test_clear_page_dirty(page))
+		task_io_account_cancelled_write(PAGE_CACHE_SIZE);
 	ClearPageUptodate(page);
 	ClearPageMappedToDisk(page);
 	remove_from_page_cache(page);