mm: write_cache_pages cyclic fix

In write_cache_pages, scanned == 1 is supposed to mean that cyclic writeback has circled through zero, thus we should not circle again. However it gets set to 1 after the first successful pagevec lookup. This leads to cases where not enough data gets written. Counterexample: file with first 10 pages dirty, writeback_index == 5, nr_to_write == 10. Then the 5 last pages will be found, and scanned will be set to 1, after writing those out, we will not cycle back to get the first 5. Rework this logic, now we'll always cycle unless we started off from index 0. When cycling, only write out as far as 1 page before the start page from the first cycle (so we don't write parts of the file twice). Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Chris Mason <chris.mason@oracle.com> Cc: Dave Chinner <david@fromorbit.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: write_cache_pages cyclic fix
In write_cache_pages, scanned == 1 is supposed to mean that cyclic writeback has circled through zero, thus we should not circle again. However it gets set to 1 after the first successful pagevec lookup. This leads to cases where not enough data gets written. Counterexample: file with first 10 pages dirty, writeback_index == 5, nr_to_write == 10. Then the 5 last pages will be found, and scanned will be set to 1, after writing those out, we will not cycle back to get the first 5. Rework this logic, now we'll always cycle unless we started off from index 0. When cycling, only write out as far as 1 page before the start page from the first cycle (so we don't write parts of the file twice). Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Chris Mason <chris.mason@oracle.com> Cc: Dave Chinner <david@fromorbit.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Nick Piggin · Linus Torvalds
1 parent 38c8e61809
Showing 1 changed file with 18 additions and 7 deletions Side-by-side Diff
mm/page-writeback.c
@@ -868,9 +868,10 @@
 	int done = 0;
 	struct pagevec pvec;
 	int nr_pages;
+	pgoff_t uninitialized_var(writeback_index);
 	pgoff_t index;
 	pgoff_t end;		/* Inclusive */
-	int scanned = 0;
+	int cycled;
 	int range_whole = 0;
 	long nr_to_write = wbc->nr_to_write;
  
  
@@ -881,14 +882,19 @@
  
 	pagevec_init(&pvec, 0);
 	if (wbc->range_cyclic) {
-		index = mapping->writeback_index; /* Start from prev offset */
+		writeback_index = mapping->writeback_index; /* prev offset */
+		index = writeback_index;
+		if (index == 0)
+			cycled = 1;
+		else
+			cycled = 0;
 		end = -1;
 	} else {
 		index = wbc->range_start >> PAGE_CACHE_SHIFT;
 		end = wbc->range_end >> PAGE_CACHE_SHIFT;
 		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
 			range_whole = 1;
-		scanned = 1;
+		cycled = 1; /* ignore range_cyclic tests */
 	}
 retry:
 	while (!done && (index <= end) &&
@@ -897,7 +903,6 @@
 					      min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
 		unsigned i;
  
-		scanned = 1;
 		for (i = 0; i < nr_pages; i++) {
 			struct page *page = pvec.pages[i];
  
@@ -915,7 +920,11 @@
 				continue;
 			}
  
-			if (!wbc->range_cyclic && page->index > end) {
+			if (page->index > end) {
+				/*
+				 * can't be range_cyclic (1st pass) because
+				 * end == -1 in that case.
+				 */
 				done = 1;
 				unlock_page(page);
 				continue;
  
  
  
@@ -946,13 +955,15 @@
 		pagevec_release(&pvec);
 		cond_resched();
 	}
-	if (!scanned && !done) {
+	if (!cycled) {
 		/*
+		 * range_cyclic:
 		 * We hit the last page and there is more work to be done: wrap
 		 * back to the start of the file
 		 */
-		scanned = 1;
+		cycled = 1;
 		index = 0;
+		end = writeback_index - 1;
 		goto retry;
 	}
 	if (!wbc->no_nrwrite_index_update) {
...	...	@@ -868,9 +868,10 @@
868	868	int done = 0;
869	869	struct pagevec pvec;
870	870	int nr_pages;
	871	+ pgoff_t uninitialized_var(writeback_index);
871	872	pgoff_t index;
872	873	pgoff_t end; /* Inclusive */
873		- int scanned = 0;
	874	+ int cycled;
874	875	int range_whole = 0;
875	876	long nr_to_write = wbc->nr_to_write;
876	877
877	878
...	...	@@ -881,14 +882,19 @@
881	882
882	883	pagevec_init(&pvec, 0);
883	884	if (wbc->range_cyclic) {
884		- index = mapping->writeback_index; /* Start from prev offset */
	885	+ writeback_index = mapping->writeback_index; /* prev offset */
	886	+ index = writeback_index;
	887	+ if (index == 0)
	888	+ cycled = 1;
	889	+ else
	890	+ cycled = 0;
885	891	end = -1;
886	892	} else {
887	893	index = wbc->range_start >> PAGE_CACHE_SHIFT;
888	894	end = wbc->range_end >> PAGE_CACHE_SHIFT;
889	895	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
890	896	range_whole = 1;
891		- scanned = 1;
	897	+ cycled = 1; /* ignore range_cyclic tests */
892	898	}
893	899	retry:
894	900	while (!done && (index <= end) &&
...	...	@@ -897,7 +903,6 @@
897	903	min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
898	904	unsigned i;
899	905
900		- scanned = 1;
901	906	for (i = 0; i < nr_pages; i++) {
902	907	struct page *page = pvec.pages[i];
903	908
...	...	@@ -915,7 +920,11 @@
915	920	continue;
916	921	}
917	922
918		- if (!wbc->range_cyclic && page->index > end) {
	923	+ if (page->index > end) {
	924	+ /*
	925	+ * can't be range_cyclic (1st pass) because
	926	+ * end == -1 in that case.
	927	+ */
919	928	done = 1;
920	929	unlock_page(page);
921	930	continue;
922	931
923	932
924	933
...	...	@@ -946,13 +955,15 @@
946	955	pagevec_release(&pvec);
947	956	cond_resched();
948	957	}
949		- if (!scanned && !done) {
	958	+ if (!cycled) {
950	959	/*
	960	+ * range_cyclic:
951	961	* We hit the last page and there is more work to be done: wrap
952	962	* back to the start of the file
953	963	*/
954		- scanned = 1;
	964	+ cycled = 1;
955	965	index = 0;
	966	+ end = writeback_index - 1;
956	967	goto retry;
957	968	}
958	969	if (!wbc->no_nrwrite_index_update) {