On Fri, Dec 20, 2024 at 2:19 PM Chen Ridong <chenridong@xxxxxxxxxxxxxxx> wrote:
>
> From: Chen Ridong <chenridong@xxxxxxxxxx>
>
> The page reclaim isolates a batch of folios from the tail of one of the
> LRU lists and works on those folios one by one. For a suitable
> swap-backed folio, if the swap device is async, it queues that folio for
> writeback. After the page reclaim finishes an entire batch, it puts back
> the folios it queued for writeback to the head of the original LRU list.
>
> In the meantime, the page writeback flushes the queued folios also by
> batches. Its batching logic is independent from that of the page reclaim.
> For each of the folios it writes back, the page writeback calls
> folio_rotate_reclaimable() which tries to rotate a folio to the tail.
>
> folio_rotate_reclaimable() only works for a folio after the page reclaim
> has put it back. If an async swap device is fast enough, the page
> writeback can finish with that folio while the page reclaim is still
> working on the rest of the batch containing it. In this case, that folio
> will remain at the head and the page reclaim will not retry it before
> reaching there.
>
> The commit 359a5e1416ca ("mm: multi-gen LRU: retry folios written back
> while isolated") only fixed the issue for mglru. However, this issue
> also exists in the traditional active/inactive LRU. This issue will be
> worse if THP is split, which makes the list longer and needs longer time
> to finish a batch of folios reclaim.
>
> This issue should be fixed in the same way for the traditional LRU.
> Therefore, the common logic was extracted to the 'find_folios_written_back'
> function firstly, which is then reused in the 'shrink_inactive_list'
> function. Finally, retry reclaiming those folios that may have missed the
> rotation for traditional LRU.
>
> Link: https://lore.kernel.org/linux-kernel/20241010081802.290893-1-chenridong@xxxxxxxxxxxxxxx/
> Link: https://lore.kernel.org/linux-kernel/CAGsJ_4zqL8ZHNRZ44o_CC69kE7DBVXvbZfvmQxMGiFqRxqHQdA@xxxxxxxxxxxxxx/
> Signed-off-by: Chen Ridong <chenridong@xxxxxxxxxx>
> ---
> mm/vmscan.c | 108 ++++++++++++++++++++++++++++++++++------------------
> 1 file changed, 70 insertions(+), 38 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 39886f435ec5..e67e446540ba 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -283,6 +283,39 @@ static void set_task_reclaim_state(struct task_struct *task,
> task->reclaim_state = rs;
> }
>
> +/**
> + * find_folios_written_back - Find and move the written back folios to a new list.
> + * @list: filios list
> + * @clean: the written back folios list
> + * @is_retried: whether the list has already been retried.
> + */
> +static inline void find_folios_written_back(struct list_head *list,
> + struct list_head *clean, bool is_retried)
> +{
> + struct folio *folio;
> + struct folio *next;
> +
> + list_for_each_entry_safe_reverse(folio, next, list, lru) {
> + if (!folio_evictable(folio)) {
> + list_del(&folio->lru);
> + folio_putback_lru(folio);
> + continue;
> + }
> +
> + /* retry folios that may have missed folio_rotate_reclaimable() */
> + if (!is_retried && !folio_test_active(folio) && !folio_mapped(folio) &&
> + !folio_test_dirty(folio) && !folio_test_writeback(folio)) {
> + list_move(&folio->lru, clean);
> + continue;
> + }
> +
> + /* don't add rejected folios to the oldest generation */
> + if (lru_gen_enabled() && !lru_gen_distance(folio, false))
> + set_mask_bits(&folio->flags, LRU_REFS_FLAGS, BIT(PG_active));
> + }
> +
> +}
> +
> /*
> * flush_reclaim_state(): add pages reclaimed outside of LRU-based reclaim to
> * scan_control->nr_reclaimed.
> @@ -1959,14 +1992,18 @@ static unsigned long shrink_inactive_list(unsigned long nr_to_scan,
> enum lru_list lru)
> {
> LIST_HEAD(folio_list);
> + LIST_HEAD(clean_list);
> unsigned long nr_scanned;
> - unsigned int nr_reclaimed = 0;
> + unsigned int nr_reclaimed, total_reclaimed = 0;
> + unsigned int nr_pageout = 0;
> + unsigned int nr_unqueued_dirty = 0;
> unsigned long nr_taken;
> struct reclaim_stat stat;
> bool file = is_file_lru(lru);
> enum vm_event_item item;
> struct pglist_data *pgdat = lruvec_pgdat(lruvec);
> bool stalled = false;
> + bool is_retried = false;
>
> while (unlikely(too_many_isolated(pgdat, file, sc))) {
> if (stalled)
> @@ -2000,22 +2037,47 @@ static unsigned long shrink_inactive_list(unsigned long nr_to_scan,
> if (nr_taken == 0)
> return 0;
>
> +retry:
> nr_reclaimed = shrink_folio_list(&folio_list, pgdat, sc, &stat, false);
>
> + sc->nr.dirty += stat.nr_dirty;
> + sc->nr.congested += stat.nr_congested;
> + sc->nr.unqueued_dirty += stat.nr_unqueued_dirty;
> + sc->nr.writeback += stat.nr_writeback;
> + sc->nr.immediate += stat.nr_immediate;
> + total_reclaimed += nr_reclaimed;
> + nr_pageout += stat.nr_pageout;
> + nr_unqueued_dirty += stat.nr_unqueued_dirty;
> +
> + trace_mm_vmscan_lru_shrink_inactive(pgdat->node_id,
> + nr_scanned, nr_reclaimed, &stat, sc->priority, file);
This is a bit odd, as nr_scanned during a retry still uses the
previous nr_scanned
value. However, I find that mglru shows no difference.
retry:
reclaimed = shrink_folio_list(&list, pgdat, sc, &stat, false);
sc->nr.unqueued_dirty += stat.nr_unqueued_dirty;
sc->nr_reclaimed += reclaimed;
trace_mm_vmscan_lru_shrink_inactive(pgdat->node_id,
scanned, reclaimed, &stat, sc->priority,
type ? LRU_INACTIVE_FILE : LRU_INACTIVE_ANON);
Currently, the active/inactive state aligns with mglru in this trace.
It seems that
the userspace BPF should recognize that the nr_scanned during a retry doesn't
indicate we are isolating new nr_scanned folios. Ideally, the is_retry
flag should
be passed to the trace, allowing userspace to identify that it's a retry and
disregard the nr_scanned value.
It might be worth addressing this in a separate patch. Add Bixuan to clarify
how userspace depends on this trace and if "retry" will break his userspace
BPF for both MGLRU and active/inactive cases.
Otherwise, the patch looks good to me.
> +
> + find_folios_written_back(&folio_list, &clean_list, is_retried);
> +
> spin_lock_irq(&lruvec->lru_lock);
> move_folios_to_lru(lruvec, &folio_list);
>
> __mod_lruvec_state(lruvec, PGDEMOTE_KSWAPD + reclaimer_offset(),
> stat.nr_demoted);
> - __mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
> item = PGSTEAL_KSWAPD + reclaimer_offset();
> if (!cgroup_reclaim(sc))
> __count_vm_events(item, nr_reclaimed);
> __count_memcg_events(lruvec_memcg(lruvec), item, nr_reclaimed);
> __count_vm_events(PGSTEAL_ANON + file, nr_reclaimed);
> +
> + if (!list_empty(&clean_list)) {
> + list_splice_init(&clean_list, &folio_list);
> + is_retried = true;
> + spin_unlock_irq(&lruvec->lru_lock);
> + goto retry;
> + }
> + __mod_node_page_state(pgdat, NR_ISOLATED_ANON + file, -nr_taken);
> spin_unlock_irq(&lruvec->lru_lock);
> + sc->nr.taken += nr_taken;
> + if (file)
> + sc->nr.file_taken += nr_taken;
>
> - lru_note_cost(lruvec, file, stat.nr_pageout, nr_scanned - nr_reclaimed);
> + lru_note_cost(lruvec, file, nr_pageout, nr_scanned - total_reclaimed);
>
> /*
> * If dirty folios are scanned that are not queued for IO, it
> @@ -2028,7 +2090,7 @@ static unsigned long shrink_inactive_list(unsigned long nr_to_scan,
> * the flushers simply cannot keep up with the allocation
> * rate. Nudge the flusher threads in case they are asleep.
> */
> - if (stat.nr_unqueued_dirty == nr_taken) {
> + if (nr_unqueued_dirty == nr_taken) {
> wakeup_flusher_threads(WB_REASON_VMSCAN);
> /*
> * For cgroupv1 dirty throttling is achieved by waking up
> @@ -2043,18 +2105,7 @@ static unsigned long shrink_inactive_list(unsigned long nr_to_scan,
> reclaim_throttle(pgdat, VMSCAN_THROTTLE_WRITEBACK);
> }
>
> - sc->nr.dirty += stat.nr_dirty;
> - sc->nr.congested += stat.nr_congested;
> - sc->nr.unqueued_dirty += stat.nr_unqueued_dirty;
> - sc->nr.writeback += stat.nr_writeback;
> - sc->nr.immediate += stat.nr_immediate;
> - sc->nr.taken += nr_taken;
> - if (file)
> - sc->nr.file_taken += nr_taken;
> -
> - trace_mm_vmscan_lru_shrink_inactive(pgdat->node_id,
> - nr_scanned, nr_reclaimed, &stat, sc->priority, file);
> - return nr_reclaimed;
> + return total_reclaimed;
> }
>
> /*
> @@ -4585,12 +4636,10 @@ static int evict_folios(struct lruvec *lruvec, struct scan_control *sc, int swap
> int reclaimed;
> LIST_HEAD(list);
> LIST_HEAD(clean);
> - struct folio *folio;
> - struct folio *next;
> enum vm_event_item item;
> struct reclaim_stat stat;
> struct lru_gen_mm_walk *walk;
> - bool skip_retry = false;
> + bool is_retried = false;
> struct lru_gen_folio *lrugen = &lruvec->lrugen;
> struct mem_cgroup *memcg = lruvec_memcg(lruvec);
> struct pglist_data *pgdat = lruvec_pgdat(lruvec);
> @@ -4616,24 +4665,7 @@ static int evict_folios(struct lruvec *lruvec, struct scan_control *sc, int swap
> scanned, reclaimed, &stat, sc->priority,
> type ? LRU_INACTIVE_FILE : LRU_INACTIVE_ANON);
>
> - list_for_each_entry_safe_reverse(folio, next, &list, lru) {
> - if (!folio_evictable(folio)) {
> - list_del(&folio->lru);
> - folio_putback_lru(folio);
> - continue;
> - }
> -
> - /* retry folios that may have missed folio_rotate_reclaimable() */
> - if (!skip_retry && !folio_test_active(folio) && !folio_mapped(folio) &&
> - !folio_test_dirty(folio) && !folio_test_writeback(folio)) {
> - list_move(&folio->lru, &clean);
> - continue;
> - }
> -
> - /* don't add rejected folios to the oldest generation */
> - if (!lru_gen_distance(folio, false))
> - set_mask_bits(&folio->flags, LRU_REFS_FLAGS, BIT(PG_active));
> - }
> + find_folios_written_back(&list, &clean, is_retried);
>
> spin_lock_irq(&lruvec->lru_lock);
>
> @@ -4656,7 +4688,7 @@ static int evict_folios(struct lruvec *lruvec, struct scan_control *sc, int swap
> list_splice_init(&clean, &list);
>
> if (!list_empty(&list)) {
> - skip_retry = true;
> + is_retried = true;
> goto retry;
> }
>
> --
> 2.34.1
>
Thanks
barry
