target/linux/generic/backport-5.15/020-v6.3-28-mm-multi-gen-LRU-simplify-arch_has_hw_pte_young-chec.patch

   1 From cf3297e4c7a928da8b2b2f0baff2f9c69ea57952 Mon Sep 17 00:00:00 2001
   2 From: Yu Zhao <yuzhao@google.com>
   3 Date: Wed, 21 Dec 2022 21:19:06 -0700
   4 Subject: [PATCH 28/29] mm: multi-gen LRU: simplify arch_has_hw_pte_young()
   5  check
   6
   7 Scanning page tables when hardware does not set the accessed bit has
   8 no real use cases.
   9
  10 Link: https://lkml.kernel.org/r/20221222041905.2431096-9-yuzhao@google.com
  11 Signed-off-by: Yu Zhao <yuzhao@google.com>
  12 Cc: Johannes Weiner <hannes@cmpxchg.org>
  13 Cc: Jonathan Corbet <corbet@lwn.net>
  14 Cc: Michael Larabel <Michael@MichaelLarabel.com>
  15 Cc: Michal Hocko <mhocko@kernel.org>
  16 Cc: Mike Rapoport <rppt@kernel.org>
  17 Cc: Roman Gushchin <roman.gushchin@linux.dev>
  18 Cc: Suren Baghdasaryan <surenb@google.com>
  19 Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
  20 ---
  21  mm/vmscan.c | 2 +-
  22  1 file changed, 1 insertion(+), 1 deletion(-)
  23
  24 diff --git a/mm/vmscan.c b/mm/vmscan.c
  25 index 4bcb93df316c..3f6874a69886 100644
  26 --- a/mm/vmscan.c
  27 +++ b/mm/vmscan.c
  28 @@ -4144,7 +4144,7 @@ static bool try_to_inc_max_seq(struct lruvec *lruvec, unsigned long max_seq,
  29          * handful of PTEs. Spreading the work out over a period of time usually
  30          * is less efficient, but it avoids bursty page faults.
  31          */
  32 -       if (!force_scan && !(arch_has_hw_pte_young() && get_cap(LRU_GEN_MM_WALK))) {
  33 +       if (!arch_has_hw_pte_young() || !get_cap(LRU_GEN_MM_WALK)) {
  34                 success = iterate_mm_list_nowalk(lruvec, max_seq);
  35                 goto done;
  36         }
  37 --
  38 2.40.0
  39