Skip to contentSkip to conversation

mm/oom_kill, proc: replace magic number 1000 with OOM_SCORE_ADJ_MAX

PATCH Applied Ye Liu 4 participants 7 messages Aug 11 – Aug 31

Subsystems (4)
lore
Share & tools
lore

Ye Liu

author
#0 lore raw

PATCH 1/1 mm/oom_kill, proc: replace magic number 1000 with OOM_SCORE_ADJ_MAX

From: Ye Liu <[email protected]>

In oom_badness() and proc_oom_score(), the oom_score_adj normalization uses a hardcoded 1000, which is the value of OOM_SCORE_ADJ_MAX defined in include/uapi/linux/oom.h. Other code in the kernel (e.g. fs/proc/base.c oom_adj handling) already uses OOM_SCORE_ADJ_MAX for the same purpose.

Replace the magic number with the macro for consistency and readability. No functional change.

---
 fs/proc/base.c | 3 ++-
 mm/oom_kill.c  | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)
fs/proc/base.c +2 -1
index 6a39de424f62..58be38942460 100644
@@ -594,7 +594,8 @@ static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 	 * exporting for a long time so userspace might depend on it.
 	 */
 	if (badness != LONG_MIN)
-		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
+		points = (OOM_SCORE_ADJ_MAX +
+			  badness * OOM_SCORE_ADJ_MAX / (long)totalpages) * 2 / 3;
 
 	seq_printf(m, "%lu\n", points);
 
mm/oom_kill.c +1 -1
index 5f372f6e26fa..08bff7a55db8 100644
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;
 	points += adj;
 
 	return points;
Signature · 1 line
2.25.1
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 8 lines · Ye Liu

From: Ye Liu <[email protected]>

In oom_badness() and proc_oom_score(), the oom_score_adj normalization uses a hardcoded 1000, which is the value of OOM_SCORE_ADJ_MAX defined in include/uapi/linux/oom.h. Other code in the kernel (e.g. fs/proc/base.c oom_adj handling) already uses OOM_SCORE_ADJ_MAX for the same purpose.

Quoted 4 lines

Replace the magic number with the macro for consistency and readability. No functional change.

I am not really sure this adds to the readability much TBH but no fundamental objections from me.

Quoted 34 lines
---
 fs/proc/base.c | 3 ++-
 mm/oom_kill.c  | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)
fs/proc/base.c +2 -1
index 6a39de424f62..58be38942460 100644
@@ -594,7 +594,8 @@ static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 	 * exporting for a long time so userspace might depend on it.
 	 */
 	if (badness != LONG_MIN)
-		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
+		points = (OOM_SCORE_ADJ_MAX +
+			  badness * OOM_SCORE_ADJ_MAX / (long)totalpages) * 2 / 3;
 
 	seq_printf(m, "%lu\n", points);
 
mm/oom_kill.c +1 -1
index 5f372f6e26fa..08bff7a55db8 100644
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;
 	points += adj;
 
 	return points;
Signature · 1 line
2.25.1
Signature · 2 lines
Michal Hocko
SUSE Labs
#2 lore raw
In reply to Ye Liu

mm/oom_kill.c long oom_badness(struct task_struct *p, unsigned long totalpages) · line 233

Quoted 8 lines · Ye Liu
mm/oom_kill.c +1 -1
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;

One thing this line hides: for a memcg OOM, totalpages is mem_cgroup_get_max(), which can be below 1000 pages when the container limit is under 4M. The division then yields 0, the whole oom_score_adj contribution goes away, and a task protected with -997 scores the same as a best-effort task with 1000. The -1000 exemption is checked separately above and still works.

DIV_ROUND_UP(totalpages, OOM_SCORE_ADJ_MAX) would preserve the adj weighting for small limits and change nothing meaningful for large ones. This is an edge case, so probably fine to leave as is - noting it here since the line is being touched anyway.

author
#3 lore raw
Quoted 17 lines · Song Hu

mm/oom_kill.c long oom_badness(struct task_struct *p, unsigned long totalpages) · line 233

Quoted 8 lines · Ye Liu
mm/oom_kill.c +1 -1
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;

One thing this line hides: for a memcg OOM, totalpages is mem_cgroup_get_max(), which can be below 1000 pages when the container limit is under 4M. The division then yields 0, the whole oom_score_adj contribution goes away, and a task protected with -997 scores the same as a best-effort task with 1000. The -1000 exemption is checked separately above and still works.

Quoted 4 lines

DIV_ROUND_UP(totalpages, OOM_SCORE_ADJ_MAX) would preserve the adj weighting for small limits and change nothing meaningful for large ones. This is an edge case, so probably fine to leave as is - noting it here since the line is being touched anyway.

Good catch. Yes, the truncation for totalpages < 1000 is real, but as you noted, it's an existing edge case. I'll keep this patch as a mechanical replacement and won't address it here. Out of curiosity, are sub-4MB memcg limits actually used in practice?

Signature · 2 lines
Thanks,
Ye Liu
maintainer
#4 lore raw
Quoted 25 lines · Ye Liu
Quoted 17 lines · Song Hu

mm/oom_kill.c long oom_badness(struct task_struct *p, unsigned long totalpages) · line 233

Quoted 8 lines · Ye Liu
mm/oom_kill.c +1 -1
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;

One thing this line hides: for a memcg OOM, totalpages is mem_cgroup_get_max(), which can be below 1000 pages when the container limit is under 4M. The division then yields 0, the whole oom_score_adj contribution goes away, and a task protected with -997 scores the same as a best-effort task with 1000. The -1000 exemption is checked separately above and still works.

Quoted 4 lines

DIV_ROUND_UP(totalpages, OOM_SCORE_ADJ_MAX) would preserve the adj weighting for small limits and change nothing meaningful for large ones. This is an edge case, so probably fine to leave as is - noting it here since the line is being touched anyway.

Quoted 4 lines

Good catch. Yes, the truncation for totalpages < 1000 is real, but as you noted, it's an existing edge case. I'll keep this patch as a mechanical replacement and won't address it here. Out of curiosity, are sub-4MB memcg limits actually used in practice?

I have seen containers as small as 20MB and they were suffering from quite some problems - e.g. charge caching on different leyers. I would generally discourage people from running containers that small unless they exactly know what they are doing.

Signature · 2 lines
Michal Hocko
SUSE Labs
maintainer
#5 lore raw
In reply to Ye Liu
Quoted 30 lines · On Tue, 11 Aug 2026 11:36:08 +0800 Ye Liu <[email protected]> wrote:

From: Ye Liu <[email protected]>

In oom_badness() and proc_oom_score(), the oom_score_adj normalization uses a hardcoded 1000, which is the value of OOM_SCORE_ADJ_MAX defined in include/uapi/linux/oom.h. Other code in the kernel (e.g. fs/proc/base.c oom_adj handling) already uses OOM_SCORE_ADJ_MAX for the same purpose.

Replace the magic number with the macro for consistency and readability. No functional change.

...

fs/proc/base.c +2 -1
@@ -594,7 +594,8 @@ static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 	 * exporting for a long time so userspace might depend on it.
 	 */
 	if (badness != LONG_MIN)
-		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
+		points = (OOM_SCORE_ADJ_MAX +
+			  badness * OOM_SCORE_ADJ_MAX / (long)totalpages) * 2 / 3;
 
 	seq_printf(m, "%lu\n", points);
 
mm/oom_kill.c +0 -0
index 5f372f6e26fa..08bff7a55db8 100644

mm/oom_kill.c long oom_badness(struct task_struct *p, unsigned long totalpages) · line 236

Quoted 8 lines
mm/oom_kill.c +1 -1
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;
 	points += adj;
 
 	return points;

AI review suggests that this is a misinterpretation of "1000"?

	https://sashiko.dev/#/patchset/[email protected]
maintainer
#6 lore raw
Quoted 39 lines · Andrew Morton
Quoted 37 lines · On Tue, 11 Aug 2026 11:36:08 +0800 Ye Liu <[email protected]> wrote:

From: Ye Liu <[email protected]>

In oom_badness() and proc_oom_score(), the oom_score_adj normalization uses a hardcoded 1000, which is the value of OOM_SCORE_ADJ_MAX defined in include/uapi/linux/oom.h. Other code in the kernel (e.g. fs/proc/base.c oom_adj handling) already uses OOM_SCORE_ADJ_MAX for the same purpose.

Replace the magic number with the macro for consistency and readability. No functional change.

...

fs/proc/base.c +2 -1
@@ -594,7 +594,8 @@ static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
 	 * exporting for a long time so userspace might depend on it.
 	 */
 	if (badness != LONG_MIN)
-		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
+		points = (OOM_SCORE_ADJ_MAX +
+			  badness * OOM_SCORE_ADJ_MAX / (long)totalpages) * 2 / 3;
 
 	seq_printf(m, "%lu\n", points);
 
mm/oom_kill.c +1 -1
index 5f372f6e26fa..08bff7a55db8 100644
@@ -230,7 +230,7 @@ long oom_badness(struct task_struct *p, unsigned long totalpages)
 	task_unlock(p);
 
 	/* Normalize to oom_score_adj units */
-	adj *= totalpages / 1000;
+	adj *= totalpages / OOM_SCORE_ADJ_MAX;
 	points += adj;
 
Quoted 5 lines

mm/oom_kill.c long oom_badness(struct task_struct *p, unsigned long totalpages) · line 236

Quoted 1 line
mm/oom_kill.c +0 -0
 	return points;

AI review suggests that this is a misinterpretation of "1000"?

	https://sashiko.dev/#/patchset/[email protected]

Sashiko is trying to be clever and it is simply wrong here. The scaling fact is indeed the same as the OOM_SCORE_ADJ_MAX.

Signature · 2 lines
Michal Hocko
SUSE Labs