On Thu 15-03-18 19:45:51, Andrey Ryabinin wrote:
> Only kswapd can have non-zero nr_immediate, and current_may_throttle() is
> always true for kswapd (PF_LESS_THROTTLE bit is never set) thus it's
> enough to check stat.nr_immediate only.
OK, so this is a result of some code evolution. We used to check for
dirty pages as well. But then b738d764652d ("Don't trigger congestion
wait on dirty-but-not-writeout pages") removed the nr_unqueued_dirty ==
nr_taken part. I was wondering whether we still have the
PF_LESS_THROTTLE protection in place but then noticed that we still have
if (!sc->hibernation_mode && !current_is_kswapd() &&
current_may_throttle())
wait_iff_congested(pgdat, BLK_RW_ASYNC, HZ/10);
in place, so good.
> Signed-off-by: Andrey Ryabinin <[email protected]>
Acked-by: Michal Hocko <[email protected]>
> ---
> mm/vmscan.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 0d5ab312a7f4..a8f6e4882e00 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1806,7 +1806,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct
> lruvec *lruvec,
> * that pages are cycling through the LRU faster than
> * they are written so also forcibly stall.
> */
> - if (stat.nr_immediate && current_may_throttle())
> + if (stat.nr_immediate)
> congestion_wait(BLK_RW_ASYNC, HZ/10);
> }
>
> --
> 2.16.1
--
Michal Hocko
SUSE Labs
