Skip to contentSkip to conversation

sched_ext: Make scx_group_set_weight() always update tg->scx.weight

PATCH In review Tejun Heo 2 participants 5 messages Jun 13, 2025 – Jun 17, 2025 Scheduler

lore
Share & tools
lore

Tejun Heo

author
#0 lore raw

PATCH 1/2 sched_ext: Make scx_group_set_weight() always update tg->scx.weight

Otherwise, tg->scx.weight can go out of sync while scx_cgroup is not enabled and ops.cgroup_init() may be called with a stale weight value.

Fixes 819513666966 ("sched_ext: Add cgroup support")
Cc [email protected] # v6.12+
---
 kernel/sched/ext.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
kernel/sched/ext.c +6 -6
@@ -4241,12 +4241,12 @@ void scx_group_set_weight(struct task_gr
 
 	percpu_down_read(&scx_cgroup_rwsem);
 
-	if (scx_cgroup_enabled && tg->scx_weight != weight) {
-		if (SCX_HAS_OP(sch, cgroup_set_weight))
-			SCX_CALL_OP(sch, SCX_KF_UNLOCKED, cgroup_set_weight, NULL,
-				    tg_cgrp(tg), weight);
-		tg->scx_weight = weight;
-	}
+	if (scx_cgroup_enabled && SCX_HAS_OP(sch, cgroup_set_weight) &&
+	    tg->scx_weight != weight)
+		SCX_CALL_OP(sch, SCX_KF_UNLOCKED, cgroup_set_weight, NULL,
+			    tg_cgrp(tg), weight);
+
+	tg->scx_weight = weight;
 
 	percpu_up_read(&scx_cgroup_rwsem);
 }
author
#1 lore raw

PATCH 2/2 sched_ext, sched/core: Don't call scx_group_set_weight() prematurely from sched_create_group()

During task_group creation, sched_create_group() calls scx_group_set_weight() with CGROUP_WEIGHT_DFL to initialize the sched_ext portion. This is premature and ends up calling ops.cgroup_set_weight() with an incorrect @cgrp before ops.cgroup_init() is called.

sched_create_group() should just initialize SCX related fields in the new task_group. Fix it by factoring out scx_tg_init() from sched_init() and making sched_create_group() call that function instead of scx_group_set_weight().

Fixes 819513666966 ("sched_ext: Add cgroup support")
Cc [email protected] # v6.12+

---
Ingo, Peter, while this touches kernel/sched/core.c, the changes are trivial and only affect sched_ext. I'm planning to route it through sched_ext/for-6.14-fixes. Please holler if you want it routed through tip.

Thanks.

 kernel/sched/core.c |    9 ++++-----
 kernel/sched/ext.c  |    5 +++++
 kernel/sched/ext.h  |    2 ++
 3 files changed, 11 insertions(+), 5 deletions(-)
kernel/sched/core.c +4 -5
@@ -8544,17 +8544,16 @@ void __init sched_init(void)
 		root_task_group.shares = ROOT_TASK_GROUP_LOAD;
 		init_cfs_bandwidth(&root_task_group.cfs_bandwidth, NULL);
 #endif /* CONFIG_FAIR_GROUP_SCHED */
-#ifdef CONFIG_EXT_GROUP_SCHED
-		root_task_group.scx_weight = CGROUP_WEIGHT_DFL;
-#endif /* CONFIG_EXT_GROUP_SCHED */
+
 #ifdef CONFIG_RT_GROUP_SCHED
 		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
 		ptr += nr_cpu_ids * sizeof(void **);
 
 		root_task_group.rt_rq = (struct rt_rq **)ptr;
 		ptr += nr_cpu_ids * sizeof(void **);
-
 #endif /* CONFIG_RT_GROUP_SCHED */
+
+		scx_tg_init(&root_task_group);
 	}
 
 #ifdef CONFIG_SMP
@@ -8985,7 +8984,7 @@ struct task_group *sched_create_group(st
 	if (!alloc_rt_sched_group(tg, parent))
 		goto err;
 
-	scx_group_set_weight(tg, CGROUP_WEIGHT_DFL);
+	scx_tg_init(tg);
 	alloc_uclamp_sched_group(tg, parent);
 
 	return tg;
kernel/sched/ext.c +5 -0
@@ -4092,6 +4092,11 @@ bool scx_can_stop_tick(struct rq *rq)
 DEFINE_STATIC_PERCPU_RWSEM(scx_cgroup_rwsem);
 static bool scx_cgroup_enabled;
 
+void scx_tg_init(struct task_group *tg)
+{
+	tg->scx_weight = CGROUP_WEIGHT_DFL;
+}
+
 int scx_tg_online(struct task_group *tg)
 {
 	struct scx_sched *sch = scx_root;
kernel/sched/ext.h +2 -0
@@ -79,6 +79,7 @@ static inline void scx_update_idle(struc
 
 #ifdef CONFIG_CGROUP_SCHED
 #ifdef CONFIG_EXT_GROUP_SCHED
+void scx_tg_init(struct task_group *tg);
 int scx_tg_online(struct task_group *tg);
 void scx_tg_offline(struct task_group *tg);
 int scx_cgroup_can_attach(struct cgroup_taskset *tset);
@@ -88,6 +89,7 @@ void scx_cgroup_cancel_attach(struct cgr
 void scx_group_set_weight(struct task_group *tg, unsigned long cgrp_weight);
 void scx_group_set_idle(struct task_group *tg, bool idle);
 #else	/* CONFIG_EXT_GROUP_SCHED */
+static inline void scx_tg_init(struct task_group *tg) {}
 static inline int scx_tg_online(struct task_group *tg) { return 0; }
 static inline void scx_tg_offline(struct task_group *tg) {}
 static inline int scx_cgroup_can_attach(struct cgroup_taskset *tset) { return 0; }
author
#2 lore raw

PATCH v2 2/2 sched_ext, sched/core: Don't call scx_group_set_weight() prematurely from sched_create_group()

During task_group creation, sched_create_group() calls scx_group_set_weight() with CGROUP_WEIGHT_DFL to initialize the sched_ext portion. This is premature and ends up calling ops.cgroup_set_weight() with an incorrect @cgrp before ops.cgroup_init() is called.

sched_create_group() should just initialize SCX related fields in the new task_group. Fix it by factoring out scx_tg_init() from sched_init() and making sched_create_group() call that function instead of scx_group_set_weight().

v2: Retain CONFIG_EXT_GROUP_SCHED ifdef in sched_init() as removing it leads
to build failures on !CONFIG_GROUP_SCHED configs.

Fixes 819513666966 ("sched_ext: Add cgroup support")
Cc [email protected] # v6.12+
---
 kernel/sched/core.c |    4 ++--
 kernel/sched/ext.c  |    5 +++++
 kernel/sched/ext.h  |    2 ++
 3 files changed, 9 insertions(+), 2 deletions(-)
kernel/sched/core.c +2 -2
@@ -8437,7 +8437,7 @@ void __init sched_init(void)
 		init_cfs_bandwidth(&root_task_group.cfs_bandwidth, NULL);
 #endif /* CONFIG_FAIR_GROUP_SCHED */
 #ifdef CONFIG_EXT_GROUP_SCHED
-		root_task_group.scx_weight = CGROUP_WEIGHT_DFL;
+		scx_tg_init(&root_task_group);
 #endif /* CONFIG_EXT_GROUP_SCHED */
 #ifdef CONFIG_RT_GROUP_SCHED
 		root_task_group.rt_se = (struct sched_rt_entity **)ptr;
@@ -8872,7 +8872,7 @@ struct task_group *sched_create_group(st
 	if (!alloc_rt_sched_group(tg, parent))
 		goto err;
 
-	scx_group_set_weight(tg, CGROUP_WEIGHT_DFL);
+	scx_tg_init(tg);
 	alloc_uclamp_sched_group(tg, parent);
 
 	return tg;
kernel/sched/ext.c +5 -0
@@ -4056,6 +4056,11 @@ bool scx_can_stop_tick(struct rq *rq)
 DEFINE_STATIC_PERCPU_RWSEM(scx_cgroup_rwsem);
 static bool scx_cgroup_enabled;
 
+void scx_tg_init(struct task_group *tg)
+{
+	tg->scx_weight = CGROUP_WEIGHT_DFL;
+}
+
 int scx_tg_online(struct task_group *tg)
 {
 	struct scx_sched *sch = scx_root;
kernel/sched/ext.h +2 -0
@@ -95,6 +95,7 @@ static inline void scx_update_idle(struc
 
 #ifdef CONFIG_CGROUP_SCHED
 #ifdef CONFIG_EXT_GROUP_SCHED
+void scx_tg_init(struct task_group *tg);
 int scx_tg_online(struct task_group *tg);
 void scx_tg_offline(struct task_group *tg);
 int scx_cgroup_can_attach(struct cgroup_taskset *tset);
@@ -104,6 +105,7 @@ void scx_cgroup_cancel_attach(struct cgr
 void scx_group_set_weight(struct task_group *tg, unsigned long cgrp_weight);
 void scx_group_set_idle(struct task_group *tg, bool idle);
 #else	/* CONFIG_EXT_GROUP_SCHED */
+static inline void scx_tg_init(struct task_group *tg) {}
 static inline int scx_tg_online(struct task_group *tg) { return 0; }
 static inline void scx_tg_offline(struct task_group *tg) {}
 static inline int scx_cgroup_can_attach(struct cgroup_taskset *tset) { return 0; }
author
#3 lore raw

Re: [PATCH v2 2/2 sched_ext/for-6.16-fixes] sched_ext, sched/core: Don't call scx_group_set_weight() prematurely from sched_create_group()

Quoted 10 lines · Tejun Heo

During task_group creation, sched_create_group() calls scx_group_set_weight() with CGROUP_WEIGHT_DFL to initialize the sched_ext portion. This is premature and ends up calling ops.cgroup_set_weight() with an incorrect @cgrp before ops.cgroup_init() is called.

sched_create_group() should just initialize SCX related fields in the new task_group. Fix it by factoring out scx_tg_init() from sched_init() and making sched_create_group() call that function instead of scx_group_set_weight().

Quoted 6 lines

v2: Retain CONFIG_EXT_GROUP_SCHED ifdef in sched_init() as removing it leads
to build failures on !CONFIG_GROUP_SCHED configs.

Fixes 819513666966 ("sched_ext: Add cgroup support")
Cc [email protected] # v6.12+

Applied 1-2 to sched_ext/for-6.16-fixes.

Thanks.

Signature · 1 line
tejun
maintainer
#4 lore raw

Re: [PATCH 2/2 sched_ext/for-6.16-fixes] sched_ext, sched/core: Don't call scx_group_set_weight() prematurely from sched_create_group()

In reply to Tejun Heo
Quoted 10 lines · Tejun Heo

During task_group creation, sched_create_group() calls scx_group_set_weight() with CGROUP_WEIGHT_DFL to initialize the sched_ext portion. This is premature and ends up calling ops.cgroup_set_weight() with an incorrect @cgrp before ops.cgroup_init() is called.

sched_create_group() should just initialize SCX related fields in the new task_group. Fix it by factoring out scx_tg_init() from sched_init() and making sched_create_group() call that function instead of scx_group_set_weight().

Quoted 7 lines
Fixes 819513666966 ("sched_ext: Add cgroup support")
Cc [email protected] # v6.12+

---
Ingo, Peter, while this touches kernel/sched/core.c, the changes are trivial and only affect sched_ext. I'm planning to route it through sched_ext/for-6.14-fixes. Please holler if you want it routed through tip.

Seems fine, go ahead and take it.