46 lines
1.8 KiB
Diff
46 lines
1.8 KiB
Diff
From b0eaacb0b9b0412916e3ca0c769f8aab34d82725 Mon Sep 17 00:00:00 2001
|
|
From: Quanmin Yan <yanquanmin1@huawei.com>
|
|
Date: Thu, 21 Aug 2025 20:55:55 +0800
|
|
Subject: mm/damon/core: prevent unnecessary overflow in
|
|
damos_set_effective_quota()
|
|
|
|
On 32-bit systems, the throughput calculation in
|
|
damos_set_effective_quota() is prone to unnecessary multiplication
|
|
overflow. Using mult_frac() to fix it.
|
|
|
|
Andrew Paniakin also recently found and privately reported this issue, on
|
|
64 bit systems. This can also happen on 64-bit systems, once the charged
|
|
size exceeds ~17 TiB. On systems running for long time in production,
|
|
this issue can actually happen.
|
|
|
|
More specifically, when a DAMOS scheme having the time quota run for
|
|
longtime, throughput calculation can overflow and set esz too small. As a
|
|
result, speed of the scheme get unexpectedly slow.
|
|
|
|
Link: https://lkml.kernel.org/r/20250821125555.3020951-1-yanquanmin1@huawei.com
|
|
Fixes: 1cd243030059 ("mm/damon/schemes: implement time quota")
|
|
Signed-off-by: Quanmin Yan <yanquanmin1@huawei.com>
|
|
Reported-by: Andrew Paniakin <apanyaki@amazon.com>
|
|
Reviewed-by: SeongJae Park <sj@kernel.org>
|
|
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
|
|
Cc: ze zuo <zuoze1@huawei.com>
|
|
Cc: <stable@vger.kernel.org> [5.16+]
|
|
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
---
|
|
mm/damon/core.c | 4 ++--
|
|
1 file changed, 2 insertions(+), 2 deletions(-)
|
|
|
|
--- a/mm/damon/core.c
|
|
+++ b/mm/damon/core.c
|
|
@@ -2026,8 +2026,8 @@ static void damos_set_effective_quota(st
|
|
|
|
if (quota->ms) {
|
|
if (quota->total_charged_ns)
|
|
- throughput = quota->total_charged_sz * 1000000 /
|
|
- quota->total_charged_ns;
|
|
+ throughput = mult_frac(quota->total_charged_sz, 1000000,
|
|
+ quota->total_charged_ns);
|
|
else
|
|
throughput = PAGE_SIZE * 1024;
|
|
esz = min(throughput * quota->ms, esz);
|