On Fri, 25 Oct 2024, Gregory Price wrote:
When numa balancing is enabled with demotion, vmscan will call
migrate_pages when shrinking LRUs. Successful demotions will
cause node vmstat numbers to double-decrement, leading to an
imbalanced page count. The result is dmesg output like such:
$ cat /proc/sys/vm/stat_refresh
[77383.088417] vmstat_refresh: nr_isolated_anon -103212
[77383.088417] vmstat_refresh: nr_isolated_file -899642
This negative value may impact compaction and reclaim throttling.
The double-decrement occurs in the migrate_pages path:
caller to shrink_folio_list decrements the count
shrink_folio_list
demote_folio_list
migrate_pages
migrate_pages_batch
migrate_folio_move
migrate_folio_done
mod_node_page_state(-ve) <- second decrement
This path happens for SUCCESSFUL migrations, not failures. Typically
callers to migrate_pages are required to handle putback/accounting for
failures, but this is already handled in the shrink code.
When accounting for migrations, instead do not decrement the count
when the migration reason is MR_DEMOTION. As of v6.11, this demotion
logic is the only source of MR_DEMOTION.
Signed-off-by: Gregory Price <gourry@xxxxxxxxxx>
Fixes: 26aa2d199d6f2 ("mm/migrate: demote pages during reclaim")
Cc: stable@xxxxxxxxxxxxxxx
Reviewed-by: Davidlohr Bueso <dave@xxxxxxxxxxxx>
