[v2,2/2] mm/page_io: count submission time as thrashing delay for delayacct

Commit Message

CGEL Aug. 15, 2022, 7:28 a.m. UTC

From: Yang Yang <[e-mail]>

Once upon a time, we only support accounting thrashing of page cache.
Then Joonsoo introduced workingset detection for anonymous pages and
we gained the ability to account thrashing of them[1].

Likes PSI, we count submission time as thrashing delay because when
the device is congested, or the submitting cgroup IO-throttled,
submission can be a significant part of overall IO time.

This patch is based on "delayacct: support re-entrance detection of
thrashing accounting".

[1] commit aae466b0052e ("mm/swap: implement workingset detection for anonymous LRU")

Signed-off-by: Yang Yang <[e-mail]>
Signed-off-by: CGEL ZTE <[e-mail]>
Reviewed-by: Ran Xiaokai <[e-mail]>
Reviewed-by: wangyong <[e-mail]>
Cc: Joonsoo Kim <[e-mail]>
---
Change for v2:
 - use bool in_thrashing instead of unsigned long dflags
---
 mm/page_io.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

Comments

Joonsoo Kim Sept. 14, 2022, 2:49 p.m. UTC | #1

2022년 8월 15일 (월) 오후 4:33, <[e-mail]>님이 작성:
>
> From: Yang Yang <[e-mail]>
>
> Once upon a time, we only support accounting thrashing of page cache.
> Then Joonsoo introduced workingset detection for anonymous pages and
> we gained the ability to account thrashing of them[1].
>
> Likes PSI, we count submission time as thrashing delay because when
> the device is congested, or the submitting cgroup IO-throttled,
> submission can be a significant part of overall IO time.

It's not the main purpose of this patch. Without this patch,
swap thrashing through frontswap or some block device supporting rw_page
operation isn't measured correctly. I think that description change is
required. Please refer to commit 937790699be9c8100e
(mm/page_io.c: annotate refault stalls from swap_readpage).

No objection on the code side.

Thanks.

Patch

diff --git a/mm/page_io.c b/mm/page_io.c
index 68318134dc92..653a2adcfdd9 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -453,18 +453,21 @@  int swap_readpage(struct page *page, bool synchronous,
 	struct swap_info_struct *sis = page_swap_info(page);
 	bool workingset = PageWorkingset(page);
 	unsigned long pflags;
+	bool in_thrashing;
 
 	VM_BUG_ON_PAGE(!PageSwapCache(page) && !synchronous, page);
 	VM_BUG_ON_PAGE(!PageLocked(page), page);
 	VM_BUG_ON_PAGE(PageUptodate(page), page);
 
 	/*
-	 * Count submission time as memory stall. When the device is congested,
-	 * or the submitting cgroup IO-throttled, submission can be a
-	 * significant part of overall IO time.
+	 * Count submission time as memory stall and delay. When the device
+	 * is congested, or the submitting cgroup IO-throttled, submission
+	 * can be a significant part of overall IO time.
 	 */
-	if (workingset)
+	if (workingset) {
+		delayacct_thrashing_start(&in_thrashing);
 		psi_memstall_enter(&pflags);
+	}
 	delayacct_swapin_start();
 
 	if (frontswap_load(page) == 0) {
@@ -513,8 +516,10 @@  int swap_readpage(struct page *page, bool synchronous,
 	bio_put(bio);
 
 out:
-	if (workingset)
+	if (workingset) {
+		delayacct_thrashing_end(&in_thrashing);
 		psi_memstall_leave(&pflags);
+	}
 	delayacct_swapin_end();
 	return ret;
 }