On Sat, 9 Sep 2023 13:33:08 +0800 Jiexun Wang <wangjiexun@xxxxxxxxxxx> wrote:
> Currently the madvise_cold_or_pageout_pte_range() function exhibits
> significant latency under memory pressure, which can be effectively
> reduced by adding cond_resched() within the loop.
>
> When the batch_count reaches SWAP_CLUSTER_MAX, we reschedule
> the task to ensure fairness and avoid long lock holding times.
>
> ...
>
> @@ -441,6 +443,13 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *pmd,
> arch_enter_lazy_mmu_mode();
> for (; addr < end; pte++, addr += PAGE_SIZE) {
> ptent = ptep_get(pte);
> +
> + if (++batch_count == SWAP_CLUSTER_MAX) {
> + pte_unmap_unlock(start_pte, ptl);
> + cond_resched();
> + start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
> + batch_count = 0;
> + }
>
> if (pte_none(ptent))
> continue;
I doubt if we can simply drop the lock like this then proceed as if
nothing has changed while the lock was released.
Could be that something along these lines:
@@ -434,6 +436,7 @@ huge_unlock:
regular_folio:
#endif
tlb_change_page_size(tlb, PAGE_SIZE);
+restart:
start_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
if (!start_pte)
return 0;
@@ -441,6 +444,15 @@ regular_folio:
arch_enter_lazy_mmu_mode();
for (; addr < end; pte++, addr += PAGE_SIZE) {
ptent = ptep_get(pte);
+
+ if (++batch_count == SWAP_CLUSTER_MAX) {
+ batch_count = 0;
+ if (need_resched()) {
+ pte_unmap_unlock(start_pte, ptl);
+ cond_resched();
+ goto restart;
+ }
+ }
if (pte_none(ptent))
continue;
would work, but more analysis would be needed.
