target/linux/generic/backport-5.15/020-v6.1-13-mm-mglru-don-t-sync-disk-for-each-aging-cycle.patch

   1 From 92d430e8955c976eacb7cc91d7ff849c0dd009af Mon Sep 17 00:00:00 2001
   2 From: Yu Zhao <yuzhao@google.com>
   3 Date: Wed, 28 Sep 2022 13:36:58 -0600
   4 Subject: [PATCH 13/29] mm/mglru: don't sync disk for each aging cycle
   5
   6 wakeup_flusher_threads() was added under the assumption that if a system
   7 runs out of clean cold pages, it might want to write back dirty pages more
   8 aggressively so that they can become clean and be dropped.
   9
  10 However, doing so can breach the rate limit a system wants to impose on
  11 writeback, resulting in early SSD wearout.
  12
  13 Link: https://lkml.kernel.org/r/YzSiWq9UEER5LKup@google.com
  14 Fixes: bd74fdaea146 ("mm: multi-gen LRU: support page table walks")
  15 Signed-off-by: Yu Zhao <yuzhao@google.com>
  16 Reported-by: Axel Rasmussen <axelrasmussen@google.com>
  17 Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
  18 ---
  19  mm/vmscan.c | 2 --
  20  1 file changed, 2 deletions(-)
  21
  22 diff --git a/mm/vmscan.c b/mm/vmscan.c
  23 index b74b334488d8..1c0875e6514a 100644
  24 --- a/mm/vmscan.c
  25 +++ b/mm/vmscan.c
  26 @@ -4165,8 +4165,6 @@ static bool try_to_inc_max_seq(struct lruvec *lruvec, unsigned long max_seq,
  27         if (wq_has_sleeper(&lruvec->mm_state.wait))
  28                 wake_up_all(&lruvec->mm_state.wait);
  29
  30 -       wakeup_flusher_threads(WB_REASON_VMSCAN);
  31 -
  32         return true;
  33  }
  34
  35 --
  36 2.40.0
  37