When Multi-Gen LRU is enabled, we can't use /sys/kernel/mm/page_idle/bitmap to track memory which is actually been used. Multi-Gen LRU page-table walker clears pte young flag, but it doesn't clear page idle flag. When we use /sys/kernel/mm/page_idle/bitmap to check whether one page is accessed, it would tell us this page is idle, but actually this page has been accessed. For those unmapped filecache pages, page idle flag would not been cleared in folio_mark_accessed if Multi-Gen LRU is enabled. So we couln't use /sys/kernel/mm/page_idle/bitmap to check whether a filecache page is read or written. What's more, /sys/kernel/mm/page_idle/bitmap also clears pte young flag. If one page is accessed, it would set page young flag. Multi-Gen LRU page-table walker should check both page&pte young flags. Changes in V2: We move folio_clear_idle into get_pfn_folio to improve efficiency of read /sys/kernel/mm/page_idle/bitmap. Sometimes memcg of process may be changed, but memcg of page would not be changed. Page idle flag would not be cleared when multi-gen LRU scan this process's page-table. We think it's better to clear page idle flag as soon as possible when walking page-table. For those non-idle pages, it's not necessary to walk rmap and clear pte young(read /sys/kernel/mm/page_idle/bitmap) if multi-gen LRU page-table walker already scanned once before. Henry Huang (1): mm: Multi-Gen LRU: fix use mm/page_idle/bitmap mm/swap.c | 3 +++ mm/vmscan.c | 40 ++++++++++++++++++++++++++++------------ 2 files changed, 31 insertions(+), 12 deletions(-) -- 2.43.0
