optimize the logic for handling dirty file folios during reclaim
Subsystems (4)
Baolin Wang
reviewerv2 cover optimize the logic for handling dirty file folios during reclaim
Since we no longer attempt to write back filesystem folios during reclaim, some logic for handling dirty file folios in the reclaim process also needs to be updated. Please check the details in each patch.
Changes from v1:
- Fix the folio_test_reclaim() check.
- Rebase on the mm-new branch.
Baolin Wang (2):
mm: vmscan: filter out the dirty file folios for node_reclaim()
mm: vmscan: simplify the logic for activating dirty file folios
include/linux/mmzone.h | 4 ---- mm/vmscan.c | 33 ++++++++------------------------- 2 files changed, 8 insertions(+), 29 deletions(-)
Signature · 1 line
2.43.7
Baolin Wang
reviewerPATCH v2 1/2 mm: vmscan: filter out the dirty file folios for node_reclaim()
After commit 6b0dfabb3555 ("fs: Remove aops->writepage"), we no longer attempt to write back filesystem folios in pageout(), and only tmpfs/shmem folios and anonymous swapcache folios can be written back. Therefore, we should also filter out the dirty filesystem folios for node_reclaim() to avoid unnecessary LRU scans.
--- mm/vmscan.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
mm/vmscan.c +5 -3
@@ -7601,9 +7601,11 @@ static unsigned long node_pagecache_reclaimable(struct pglist_data *pgdat)else nr_pagecache_reclaimable =node_unmapped_file_pages (pgdat); -/* If we can't clean pages, remove dirty pages from consideration */ -if (!(node_reclaim_mode &RECLAIM_WRITE )) - delta +=node_page_state (pgdat,NR_FILE_DIRTY ); +/* +* Since we can't clean folios through reclaim, remove dirty file +* folios from consideration. +*/ + delta +=node_page_state (pgdat,NR_FILE_DIRTY );/* Watch for any possible underflows due to delta */ if (unlikely (delta > nr_pagecache_reclaimable))
Signature · 1 line
2.43.7
Baolin Wang
reviewerPATCH v2 2/2 mm: vmscan: simplify the logic for activating dirty file folios
In reply to Baolin WangAfter commit 6b0dfabb3555 ("fs: Remove aops->writepage"), we no longer attempt to write back filesystem folios through reclaim.
However, in the shrink_folio_list() function, there still remains some logic related to writeback control of dirty file folios. The original logic was that, for direct reclaim, or when folio_test_reclaim() is false, or the PGDAT_DIRTY flag is not set, the dirty file folios would be directly activated to avoid being scanned again; otherwise, it will try to writeback the dirty file folios. However, since we can no longer perform writeback on dirty folios, the dirty file folios will still be activated.
Additionally, under the original logic, if we continue to try writeback dirty file folios, we will also check the references flag, sc->may_writepage, and may_enter_fs(), which may result in dirty file folios being left in the inactive list. This is unreasonable. Even if these dirty folios are scanned again, we still cannot clean them.
Therefore, the checks on these dirty file folios appear to be redundant and can be removed. Dirty file folios should be directly moved to the active list to avoid being scanned again. Since we set the PG_reclaim flag for the dirty folios, once the writeback is completed, they will be moved back to the tail of the inactive list to be retried for quick reclaim.
--- include/linux/mmzone.h | 4 ---- mm/vmscan.c | 25 +++---------------------- 2 files changed, 3 insertions(+), 26 deletions(-)
include/linux/mmzone.h +0 -4
@@ -1060,10 +1060,6 @@ struct zone { }____cacheline_internodealigned_in_smp ;enum pgdat_flags { -PGDAT_DIRTY ,/* reclaim scanning has recently found -* many dirty file pages at the tail -* of the LRU. -*/ PGDAT_WRITEBACK ,/* reclaim scanning has recently found * many pages under writeback */
mm/vmscan.c +3 -22
@@ -1387,21 +1387,7 @@ static unsigned int shrink_folio_list(struct list_head *folio_list, mapping =folio_mapping (folio);if (folio_test_dirty (folio)) { -/* -* Only kswapd can writeback filesystem folios -* to avoid risk of stack overflow. But avoid -* injecting inefficient single-folio I/O into -* flusher writeback as much as possible: only -* write folios when we've encountered many -* dirty folios, and when we've already scanned -* the rest of the LRU for clean folios and see -* the same dirty folios again (with the reclaim -* flag set). -*/ -if (folio_is_file_lru (folio) && - (!current_is_kswapd () || - !folio_test_reclaim (folio) || - !test_bit (PGDAT_DIRTY , &pgdat->flags))) { +if (folio_is_file_lru (folio)) {/* * Immediately reclaim when written back. * Similar in principle to folio_deactivate()
@@ -1410,7 +1396,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,*/ node_stat_mod_folio (folio,NR_VMSCAN_IMMEDIATE , nr_pages); -folio_set_reclaim (folio); +if (!folio_test_reclaim (folio)) +folio_set_reclaim (folio); goto activate_locked; }
@@ -6105,11 +6092,6 @@ static void shrink_node(pg_data_t *pgdat, struct scan_control *sc)if (sc->nr.writeback && sc->nr.writeback == sc->nr.taken)set_bit (PGDAT_WRITEBACK , &pgdat->flags); -/* Allow kswapd to start writing pages during reclaim.*/ -if (sc->nr.unqueued_dirty && - sc->nr.unqueued_dirty == sc->nr.file_taken) -set_bit (PGDAT_DIRTY , &pgdat->flags); -/* * If kswapd scans pages marked for immediate * reclaim and under writeback (nr_immediate), it
@@ -6850,7 +6832,6 @@ static void clear_pgdat_congested(pg_data_t *pgdat)clear_bit (LRUVEC_NODE_CONGESTED , &lruvec->flags);clear_bit (LRUVEC_CGROUP_CONGESTED , &lruvec->flags); -clear_bit (PGDAT_DIRTY , &pgdat->flags);clear_bit (PGDAT_WRITEBACK , &pgdat->flags); }
Signature · 1 line
2.43.7
Michal Hocko
reviewerRe: [PATCH v2 2/2] mm: vmscan: simplify the logic for activating dirty file folios
Quoted 17 lines · Baolin Wang
After commit
6b0dfabb3555("fs: Removeaops->writepage"), we no longer attempt to write back filesystem folios through reclaim.However, in the
shrink_folio_list()function, there still remains some logic related to writeback control of dirty file folios. The original logic was that, for direct reclaim, or whenfolio_test_reclaim()is false, or thePGDAT_DIRTYflag is not set, the dirty file folios would be directly activated to avoid being scanned again; otherwise, it will try to writeback the dirty file folios. However, since we can no longer perform writeback on dirty folios, the dirty file folios will still be activated.Additionally, under the original logic, if we continue to try writeback dirty file folios, we will also check the references flag,
sc->may_writepage, andmay_enter_fs(), which may result in dirty file folios being left in the inactive list. This is unreasonable. Even if these dirty folios are scanned again, we still cannot clean them.
Quoted 5 lines
Therefore, the checks on these dirty file folios appear to be redundant and can be removed. Dirty file folios should be directly moved to the active list to avoid being scanned again. Since we set the
PG_reclaimflag for the dirty folios, once the writeback is completed, they will be moved back to the tail of the inactive list to be retried for quick reclaim.
Is there any actual problem you are trying to address or is this a code clean up? How have you evaluated this change?
Quoted 81 lines
Signed-off-by Baolin Wang <[email protected]>--- include/linux/mmzone.h | 4 ---- mm/vmscan.c | 25 +++---------------------- 2 files changed, 3 insertions(+), 26 deletions(-)include/linux/mmzone.h +0
-4@@ -1060,10 +1060,6 @@ struct zone { }____cacheline_internodealigned_in_smp ;enum pgdat_flags { -PGDAT_DIRTY ,/* reclaim scanning has recently found -* many dirty file pages at the tail -* of the LRU. -*/ PGDAT_WRITEBACK ,/* reclaim scanning has recently found * many pages under writeback */ mm/vmscan.c +3
-22@@ -1387,21 +1387,7 @@ static unsigned int shrink_folio_list(struct list_head *folio_list, mapping =folio_mapping (folio);if (folio_test_dirty (folio)) { -/* -* Only kswapd can writeback filesystem folios -* to avoid risk of stack overflow. But avoid -* injecting inefficient single-folio I/O into -* flusher writeback as much as possible: only -* write folios when we've encountered many -* dirty folios, and when we've already scanned -* the rest of the LRU for clean folios and see -* the same dirty folios again (with the reclaim -* flag set). -*/ -if (folio_is_file_lru (folio) && - (!current_is_kswapd () || - !folio_test_reclaim (folio) || - !test_bit (PGDAT_DIRTY , &pgdat->flags))) { +if (folio_is_file_lru (folio)) {/* * Immediately reclaim when written back. * Similar in principle to folio_deactivate() @@ -1410,7 +1396,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,*/ node_stat_mod_folio (folio,NR_VMSCAN_IMMEDIATE , nr_pages); -folio_set_reclaim (folio); +if (!folio_test_reclaim (folio)) +folio_set_reclaim (folio); goto activate_locked; }@@ -6105,11 +6092,6 @@ static void shrink_node(pg_data_t *pgdat, struct scan_control *sc)if (sc->nr.writeback && sc->nr.writeback == sc->nr.taken)set_bit (PGDAT_WRITEBACK , &pgdat->flags); -/* Allow kswapd to start writing pages during reclaim.*/ -if (sc->nr.unqueued_dirty && - sc->nr.unqueued_dirty == sc->nr.file_taken) -set_bit (PGDAT_DIRTY , &pgdat->flags); -/* * If kswapd scans pages marked for immediate * reclaim and under writeback (nr_immediate), it @@ -6850,7 +6832,6 @@ static void clear_pgdat_congested(pg_data_t *pgdat)clear_bit (LRUVEC_NODE_CONGESTED , &lruvec->flags);clear_bit (LRUVEC_CGROUP_CONGESTED , &lruvec->flags); -clear_bit (PGDAT_DIRTY , &pgdat->flags);clear_bit (PGDAT_WRITEBACK , &pgdat->flags); }Signature · 1 line
2.43.7
Signature · 2 lines
Michal Hocko SUSE Labs
Baolin Wang
reviewerRe: [PATCH v2 2/2] mm: vmscan: simplify the logic for activating dirty file folios
Quoted 18 lines · Michal Hocko
Quoted 17 lines · Baolin Wang
After commit
6b0dfabb3555("fs: Removeaops->writepage"), we no longer attempt to write back filesystem folios through reclaim.However, in the
shrink_folio_list()function, there still remains some logic related to writeback control of dirty file folios. The original logic was that, for direct reclaim, or whenfolio_test_reclaim()is false, or thePGDAT_DIRTYflag is not set, the dirty file folios would be directly activated to avoid being scanned again; otherwise, it will try to writeback the dirty file folios. However, since we can no longer perform writeback on dirty folios, the dirty file folios will still be activated.Additionally, under the original logic, if we continue to try writeback dirty file folios, we will also check the references flag,
sc->may_writepage, andmay_enter_fs(), which may result in dirty file folios being left in the inactive list. This is unreasonable. Even if these dirty folios are scanned again, we still cannot clean them.
Quoted 8 lines
Quoted 5 lines
Therefore, the checks on these dirty file folios appear to be redundant and can be removed. Dirty file folios should be directly moved to the active list to avoid being scanned again. Since we set the
PG_reclaimflag for the dirty folios, once the writeback is completed, they will be moved back to the tail of the inactive list to be retried for quick reclaim.Is there any actual problem you are trying to address or is this a code clean up? How have you evaluated this change?
This patch is more of a cleanup patch, since dirty file folios are also activated in pageout(), so there are essentially no significant logical changes. Moreover, this patch set is a continuation of the previous cleanup work[1] for dirty file folios, and further cleanup and optimization work for file folios reclamation is still ongoing.
I conducted some evaluations (such as building the kernel in memcg to reclaim file folios and running thpcompact to reclaim file folios), and I did not observe any obvious changes in reclaim efficiency.