Skip to contentSkip to conversation

mm/damon/ops-common: handle extreme intervals in damon_hot_score()

PATCH Applied SeongJae Park 2 participants 3 messages Jun 23

Subsystems (3)
lore
Share & tools
lore

SeongJae Park

author maintainer
#0 lore raw

PATCH v2 1/1 mm/damon/ops-common: handle extreme intervals in damon_hot_score()

Fix three issues in damon_hot_score() that comes from wrong handling of extreme (zero or too high) monitoring intervals user setup.

When the user sets sampling interval zero, damon_max_nr_accesses(), which is called from damon_hot_score(), causes a divide-by-zero. Needless to say, it is a problem.

When the user sets the aggregation interval zero, the function returns zero. It is wrong, since the real maximum nr_acceses in the setup should be one. Worse yet, it can cause another divide-by-zero from its caller, damon_hot_score(), since it uses damon_max_nr_accesses() return value as a denominator.

When the user sets the aggregation interval very high, damon_hot_score() could return a value out of [0, DAMOS_MAX_SCORE] range. Since the return value is used as an index to the regions_score_histogram array, which is DAMOS_MAX_SCORE+1 size, it causes out of bounds array access.

The issues can be relatively easily reproduced like below. The sysfs write permission is required, though.

    # ./damo start --damos_action lru_prio --damos_quota_space 100M \
            --damos_quota_interval 1s
    # cd /sys/kernel/mm/damon/admin/kdamonds/0
    # echo 0 > contexts/0/monitoring_attrs/intervals/sample_us
    # echo 0 > contexts/0/monitoring_attrs/intervals/aggr_us
    # echo commit > state
    # dmesg
    [...]
    [  131.329762] Oops: divide error: 0000 [#1] SMP NOPTI
    [...]
    [  131.336089] RIP: 0010:damon_hot_score+0x27/0xd0
    [...]

Fix the divide-by-zero intervals problems by explicitly handling the zero intervals in damon_max_nr_accesses(). Fix the out-of-bound array access by applying [0, DAMOS_MAX_SCORE] bounds before returning from damon_hot_score().

The issue was discovered [1] by Sashiko.

[1] https://lore.kernel.org/[email protected]

Fixes 198f0f4c58b9 ("mm/damon/vaddr,paddr: support pageout prioritization")
Cc <[email protected]> # 5.16.x

---
Changes from RFC v1.3
- RFC v1.3: https://lore.kernel.org/[email protected]
- Drop RFC again.
Changes from RFC v1.2
- RFC v1.2: https://lore.kernel.org/[email protected]
- Drop patch 2 and make patch 1 fixes all damon_hot_score() problems. Changes from v1
- v1: https://lore.kernel.org/[email protected]
- Add out-of-bound array access bug fix as patch 2.
- Add the RFC tag again.
Changes from RFC v1.1
- RFC v1.1: https://lore.kernel.org/[email protected]
- Wordsmith commit message.
- Drop RFC tag.
Changes from RFC v1
- RFC v1: https://lore.kernel.org/[email protected]
- Handle zero aggr_interval case.

 include/linux/damon.h | 8 ++++++--
 mm/damon/ops-common.c | 1 +
 2 files changed, 7 insertions(+), 2 deletions(-)
include/linux/damon.h +6 -2
index 6f7edb3590ef9..888570f55b416 100644
@@ -1065,9 +1065,13 @@ static inline bool damon_target_has_pid(const struct damon_ctx *ctx)
 
 static inline unsigned int damon_max_nr_accesses(const struct damon_attrs *attrs)
 {
-	/* {aggr,sample}_interval are unsigned long, hence could overflow */
-	return min(attrs->aggr_interval / attrs->sample_interval,
+	unsigned long sample_interval;
+	unsigned long max_nr_accesses;
+
+	sample_interval = attrs->sample_interval ? : 1;
+	max_nr_accesses = min(attrs->aggr_interval / sample_interval,
 			(unsigned long)UINT_MAX);
+	return max_nr_accesses ? : 1;
 }
 
 
mm/damon/ops-common.c +1 -0
index 5c93ef2bb8a97..d1842e2b00ef8 100644
@@ -143,6 +143,7 @@ int damon_hot_score(struct damon_ctx *c, struct damon_region *r,
 	 * Transform it to fit in [0, DAMOS_MAX_SCORE]
 	 */
 	hotness = hotness * DAMOS_MAX_SCORE / DAMON_MAX_SUBSCORE;
+	hotness = max(min(hotness, DAMOS_MAX_SCORE), 0);
 
 	return hotness;
 }

base-commit: c12377ad97c98e0bee10870abf4ab1101a946b4c

Signature · 1 line
2.47.3
#1 lore raw
Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] Multiplication in freq_subscore calculation overflows 32-bit unsigned int for extreme intervals, reversing region hotness prioritization.
- [High] User-supplied sysfs weights can overflow, resulting in a bypassed division and Signed Integer Overflow (UB).
--

commit 8323dd74c4905401bc8701431eb5fd1a6b12a452
Author: SeongJae Park <[email protected]>

mm/damon/ops-common: handle extreme intervals in damon_hot_score()

Fix three issues in damon_hot_score() that comes from wrong handling of extreme (zero or too high) monitoring intervals user setup.

mm/damon/ops-common.c int damon_hot_score(struct damon_ctx *c, struct damon_region *r,

Quoted 5 lines
mm/damon/ops-common.c +0 -0
index 5c93ef2bb8a97..d1842e2b00ef8 100644
@@ -143,6 +143,7 @@ int damon_hot_score(struct damon_ctx *c, struct damon_region *r,

[Severity: High]
This isn't a bug introduced by this patch, but could the freq_subscore calculation overflow earlier in this function?

If a user configures a normal sample_interval (e.g., 100us) but a long aggr_interval (e.g., 2 hours), r->nr_accesses can legitimately reach 72,000,000.

mm/damon/ops-common.c:damon_hot_score() {
    ...
    freq_subscore = r->nr_accesses * DAMON_MAX_SUBSCORE /
                    damon_max_nr_accesses(&c->attrs);
    ...
}

The multiplication of r->nr_accesses * DAMON_MAX_SUBSCORE (100) evaluates to 7,200,000,000, which wraps around modulo 2^32, causing extremely hot memory regions to be incorrectly scored as completely cold.

[Severity: High]
This is also a pre-existing issue, but can user-supplied sysfs weights overflow and bypass the averaging division before this transformation?

mm/damon/ops-common.c:damon_hot_score() {
    ...
    if (freq_weight + age_weight)
        hotness /= freq_weight + age_weight;
    ...
}

If a user supplies weights for weight_nr_accesses and weight_age whose sum is exactly UINT_MAX + 1 (e.g., 4,000,000,000 and 294,967,296), the unsigned addition overflows to 0.

This causes the branch to evaluate as false, bypassing the averaging division. The unaveraged weighted sum then evaluates to a large positive integer.

mm/damon/ops-common.c int damon_hot_score(struct damon_ctx *c, struct damon_region *r, · line 145

Quoted 3 lines
mm/damon/ops-common.c +0 -0
 	 * Transform it to fit in [0, DAMOS_MAX_SCORE]
 	 */
 	hotness = hotness * DAMOS_MAX_SCORE / DAMON_MAX_SUBSCORE;

Multiplying it by DAMOS_MAX_SCORE (99) here then overflows the signed int result, resulting in undefined behavior.

Quoted 4 lines
mm/damon/ops-common.c +1 -0
+	hotness = max(min(hotness, DAMOS_MAX_SCORE), 0);
 
 	return hotness;
 }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
author maintainer
#2 lore raw
Quoted 5 lines · On Tue, 23 Jun 2026 14:15:26 +0000 [email protected] wrote:

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] Multiplication in freq_subscore calculation overflows 32-bit unsigned int for extreme intervals, reversing region hotness prioritization.
- [High] User-supplied sysfs weights can overflow, resulting in a bypassed division and Signed Integer Overflow (UB).

Sashiko acknowledges both are pre-existing issue, so no blocker for this patch.

Also, both issues happen only under extreme and never common user setups, and causing only DAMON internal behavioral degradation. No critical issues like panic or memory corruption will happen. These are known low priority issues, not a blocker of this patch.

Thanks,
SJ

[...]