Re: + restore-acct_reclaim_writeback-for-folio.patch added to -mm tree

[Matthew Wilcox <willy@xxxxxxxxxxxxx>]

On Thu, Oct 21, 2021 at 03:33:16AM +0100, Matthew Wilcox wrote:
> On Wed, Oct 20, 2021 at 06:08:42PM -0700, akpm@xxxxxxxxxxxxxxxxxxxx wrote:
> > From: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > Subject: restore-acct_reclaim_writeback-for-folio
> > 
> > Make Mel's "mm/vmscan: throttle reclaim and compaction when too may pages
> > are isolated" work for folio changes.
> 
> Mmm.  acct_reclaim_writeback() is going to need to be converted to
> folios -- it accounts a page as a single page instead of as however
> many pages it contains.
> 
> This patch makes sense to apply, so this is just a note that there's
> a fuller fixup to come later.

Later seems to be now.  This patch compiles.  The only non-mechanical
change in here is changing inc_node_page_state() to
node_stat_add_folio() which accounts the number of pages in the folio
instead of 1.

diff --git a/mm/filemap.c b/mm/filemap.c
index 6844c9816a86..daa0e23a6ee6 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1607,7 +1607,7 @@ void folio_end_writeback(struct folio *folio)
 
 	smp_mb__after_atomic();
 	folio_wake(folio, PG_writeback);
-	acct_reclaim_writeback(folio_page(folio, 0));
+	acct_reclaim_writeback(folio);
 	folio_put(folio);
 }
 EXPORT_SYMBOL(folio_end_writeback);
diff --git a/mm/internal.h b/mm/internal.h
index 632c55c5a075..3b79a5c9427a 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -41,15 +41,15 @@ static inline void *folio_raw_mapping(struct folio *folio)
 	return (void *)(mapping & ~PAGE_MAPPING_FLAGS);
 }
 
-void __acct_reclaim_writeback(pg_data_t *pgdat, struct page *page,
+void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
 						int nr_throttled);
-static inline void acct_reclaim_writeback(struct page *page)
+static inline void acct_reclaim_writeback(struct folio *folio)
 {
-	pg_data_t *pgdat = page_pgdat(page);
+	pg_data_t *pgdat = folio_pgdat(folio);
 	int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
 
 	if (nr_throttled)
-		__acct_reclaim_writeback(pgdat, page, nr_throttled);
+		__acct_reclaim_writeback(pgdat, folio, nr_throttled);
 }
 
 static inline void wake_throttle_isolated(pg_data_t *pgdat)
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 59c07ee4220d..fb9584641ac7 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1085,12 +1085,12 @@ void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason)
  * pages to clean. If enough pages have been cleaned since throttling
  * started then wakeup the throttled tasks.
  */
-void __acct_reclaim_writeback(pg_data_t *pgdat, struct page *page,
+void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
 							int nr_throttled)
 {
 	unsigned long nr_written;
 
-	inc_node_page_state(page, NR_THROTTLED_WRITTEN);
+	node_stat_add_folio(folio, NR_THROTTLED_WRITTEN);
 
 	/*
 	 * This is an inaccurate read as the per-cpu deltas may not