On Mon, Aug 9, 2021 at 4:43 PM Huang, Ying <ying.huang@xxxxxxxxx> wrote:
>
> Yang Shi <shy828301@xxxxxxxxx> writes:
>
> > On Fri, Aug 6, 2021 at 1:37 PM Yang Shi <shy828301@xxxxxxxxx> wrote:
> >>
> >> On Thu, Aug 5, 2021 at 11:01 PM Hugh Dickins <hughd@xxxxxxxxxx> wrote:
> >> >
> >> > On Thu, 5 Aug 2021, Yang Shi wrote:
> >> > > On Tue, Aug 3, 2021 at 10:34 PM Hugh Dickins <hughd@xxxxxxxxxx> wrote:
> >> > > >
> >> > > > I've never seen the swapoff race claimed by Miaohe, and don't expect to;
> >> > > > but he's probably right, given the current code. I just dislike adding
> >> > > > unnecessary complexity, and siting it in the wrong place (mm/shmem.c).
> >> > > >
> >> > > > Yang, is it possible that 5.1 commit 8fd2e0b505d1 ("mm: swap: check if
> >> > > > swap backing device is congested or not") was actually developed and
> >> > > > measured on 4.1 or earlier, which still had blk_set_queue_congested()?
> >> > >
> >> > > I forgot the exact version, but definitely not 4.1 or earlier. Maybe
> >> > > 4.19 or earlier. I'm not familiar with how block layer detect
> >> > > congestion, if the logic was changed, hence the optimization doesn't
> >> > > stand anymore nowadays, I'm totally fine to remove it.
> >> >
> >> > You drove me back to look more closely. blk_set_queue_congested()
> >> > vanished from include/linux/blkdev.h in 4.2, but blk_set_congested()
> >> > appeared then in block/blk-core.c to replace it. blk_set_congested()
> >> > vanished (along with all references to "congested" in blk-core.c) in
> >> > 5.0, then your commit (most probably tested on 4.19) went into 5.1 -
> >> > just after it had become redundant!
> >> >
> >> > Thanks, yes, let's revert that and Miaohe's and Huang's, later on.
> >>
> >> It should be easier to revert Huang Ying's , then Miaohe's, then mine.
> >
> > Hi Ying,
> >
> > I just prepared the reverts since I need to revert yours and Miaohe's
> > in order to revert my problematic commit.
>
> If your original commit will be reverted, then mine and Miaohe's can be
> reverted from the race condition point of view.
Yes, this is exactly what I did.
>
> Although I still think it's better to call get/put_swap_device() in
> shmem_swapin_page(), that can be discussed with another patch.
Any additional change could be based off the reverts.
>
> Best Regards,
> Huang, Ying
