[..] > > > This is only a problem when we skip the swapcache during swapin. > > Otherwise the swapcache synchronizes this. I wonder how much does > > skipping the swapcache buy us on recent kernels? This optimization was > > introduced a long time ago. > > Still performs quite good. according to kairui's data: > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=13ddaf26be324a7f951891ecd9ccd04466d27458 > > Before: 10934698 us > After: 11157121 us > Cached: 13155355 us (Dropping SWP_SYNCHRONOUS_IO flag) That's 17.9% slower if we use the swapcache. Commit 0bcac06f27d7 ("mm, swap: skip swapcache for swapin of synchronous device") says: """ This patch aims to reduce swap-in latency by skipping swapcache if the swap device is synchronous device like rw_page based device. It enhances 45% my swapin test(5G sequential swapin, no readahead, from 2.41sec to 1.64sec). """ So went from 45% to 17.9%, we are getting close :) > > BTW, zram+zswap seems pointless from the first beginning. it seems a wrong > configuration for users. if this case is really happening, could we > simply fix it Yeah we can do that. Zhongkun said they have a use case for zram+zswap tho, but I don't know what it is. If we can find a better (yet still simple) fix, then I would prefer that. Otherwise enforcing the use of the swapcache when zswap is enabled sounds like a viable option. > by: > > diff --git a/mm/memory.c b/mm/memory.c > index b7cab8be8632..6742d1428373 100644 > --- a/mm/memory.c > +++ b/mm/memory.c > @@ -3999,7 +3999,7 @@ vm_fault_t do_swap_page(struct vm_fault *vmf) > swapcache = folio; > > if (!folio) { > - if (data_race(si->flags & SWP_SYNCHRONOUS_IO) && > + if (!is_zswap_enabled() && data_race(si->flags & > SWP_SYNCHRONOUS_IO) && > __swap_count(entry) == 1) { > /* > * Prevent parallel swapin from proceeding with
