Commit 22a77880 authored by SeongJae Park's avatar SeongJae Park Committed by Andrew Morton
Browse files

mm/damon/vaddr: call damon_update_region_access_rate() always 

When getting mm_struct of the monitoring target process fails, there wil
be no need to increase the access rate counter (nr_accesses) of the
regions for the process.  Hence, damon_va_check_accesses() skips calling
damon_update_region_access_rate() in the case.  This breaks the assumption
that damon_update_region_access_rate() is called for every region, for
every sampling interval.  Call the function for every region even in the
case.  This might increase the overhead in some cases, but such case would
not be frequent, so no significant impact is really expected.

Link: https://lkml.kernel.org/r/20230915025251.72816-3-sj@kernel.org


Signed-off-by: default avatarSeongJae Park <sj@kernel.org>
Cc: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
parent 78fbfb15

mm/damon/vaddr.c

+7 −3
Original line number Diff line number Diff line
@@ -564,6 +564,11 @@ static void __damon_va_check_access(struct mm_struct *mm, 
	static unsigned long last_folio_sz = PAGE_SIZE;

	static bool last_accessed;



	if (!mm) {

		damon_update_region_access_rate(r, false);

		return;

	}



	/* If the region is in the last checked page, reuse the result */

	if (same_target && (ALIGN_DOWN(last_addr, last_folio_sz) ==

				ALIGN_DOWN(r->sampling_addr, last_folio_sz))) {
@@ -587,14 +592,13 @@ static unsigned int damon_va_check_accesses(struct damon_ctx *ctx) 


	damon_for_each_target(t, ctx) {

		mm = damon_get_mm(t);

		if (!mm)

			continue;

		same_target = false;

		damon_for_each_region(r, t) {

			__damon_va_check_access(mm, r, same_target);

			max_nr_accesses = max(r->nr_accesses, max_nr_accesses);

			same_target = true;

		}

		if (mm)

			mmput(mm);

	}