On Thu, Mar 21, 2024 at 08:25:26AM -0700, Nhat Pham wrote:
> On Thu, Mar 21, 2024 at 2:28 AM Chengming Zhou <chengming.zhou@xxxxxxxxx> wrote:
> >
> > On 2024/3/21 14:36, Zhongkun He wrote:
> > > On Thu, Mar 21, 2024 at 1:24 PM Chengming Zhou <chengming.zhou@xxxxxxxxx> wrote:
> > >>
> > >> On 2024/3/21 13:09, Zhongkun He wrote:
> > >>> On Thu, Mar 21, 2024 at 12:42 PM Chengming Zhou
> > >>> <chengming.zhou@xxxxxxxxx> wrote:
> > >>>>
> > >>>> On 2024/3/21 12:34, Zhongkun He wrote:
> > >>>>> Hey folks,
> > >>>>>
> > >>>>> Recently, I tested the zswap with memory reclaiming in the mainline
> > >>>>> (6.8) and found a memory corruption issue related to exclusive loads.
> > >>>>
> > >>>> Is this fix included? 13ddaf26be32 ("mm/swap: fix race when skipping swapcache")
> > >>>> This fix avoids concurrent swapin using the same swap entry.
> > >>>>
> > >>>
> > >>> Yes, This fix avoids concurrent swapin from different cpu, but the
> > >>> reported issue occurs
> > >>> on the same cpu.
> > >>
> > >> I think you may misunderstand the race description in this fix changelog,
> > >> the CPU0 and CPU1 just mean two concurrent threads, not real two CPUs.
> > >>
> > >> Could you verify if the problem still exists with this fix?
> > >
> > > Yes，I'm sure the problem still exists with this patch.
> > > There is some debug info, not mainline.
> > >
> > > bpftrace -e'k:swap_readpage {printf("%lld, %lld,%ld,%ld,%ld\n%s",
> > > ((struct page *)arg0)->private,nsecs,tid,pid,cpu,kstack)}' --include
> > > linux/mm_types.h
> >
> > Ok, this problem seems only happen on SWP_SYNCHRONOUS_IO swap backends,
> > which now include zram, ramdisk, pmem, nvdimm.
> >
> > It maybe not good to use zswap on these swap backends?
>
> My gut reaction is to say yes, but I'll refrain from making sweeping
> statements about backends I'm not too familiar with. Let's see:
>
> 1. zram: I don't even know why we're putting a compressed cache... in
> front of a compressed faux swap device? Ramdisk == other in-memory
> swap backend right?
I personally use it for testing because it's easy, but I doubt any prod
setups actually do that. That being said, I don't think we need to
disable zswap completely for these swap backends just to address this
bug.
> 2. I looked it up, and it seemed SWP_SYNCHRONOUS_IO was introduced for
> fast swap storage (see the original patch series [1]). If this is the
> case, one could argue there are diminishing returns for applying zswap
> on top of this.
>
> [1]: https://lore.kernel.org/linux-mm/1505886205-9671-1-git-send-email-minchan@xxxxxxxxxx/
>
> >
> > The problem here is the page fault handler tries to skip swapcache to
> > swapin the folio (swap entry count == 1), but then it can't install folio
> > to pte entry since some changes happened such as concurrent fork of entry.
> >
> > Maybe we should writeback that folio in this special case.
>
> But yes, if this is simple maybe we can do this first to fix the bug?
Can we just enforce using the swapcache if zswap is in-use? We cannot
simply check if zswap is enabled, because it could be the case that we
stored some pages into zswap then disabled it.
Perhaps we could keep track of whether zswap was ever enabled or if any
pages were ever stored in zswap, and skip the no swap cache optimization
then?
