Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
sched/ext: Implement cgroup_set_idle() callback
Implement the missing cgroup_set_idle() callback that was marked as a
TODO. This allows BPF schedulers to be notified when a cgroup's idle
state changes, enabling them to adjust their scheduling behavior
accordingly.
The implementation follows the same pattern as other cgroup callbacks
like cgroup_set_weight() and cgroup_set_bandwidth(). It checks if the
BPF scheduler has implemented the callback and invokes it with the
appropriate parameters.
Fixes a spelling error in the cgroup_set_bandwidth() documentation.
tj: s/scx_cgroup_rwsem/scx_cgroup_ops_rwsem/ to fix build breakage.
Signed-off-by: zhidao su <soolaugust@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
+28
-2
+1
include/linux/sched/ext.h
+1
include/linux/sched/ext.h
+15
-1
kernel/sched/ext.c
+15
-1
kernel/sched/ext.c
···
3066
3066
tg->scx.weight = CGROUP_WEIGHT_DFL;
3067
3067
tg->scx.bw_period_us = default_bw_period_us();
3068
3068
tg->scx.bw_quota_us = RUNTIME_INF;
3069
+
tg->scx.idle = false;
3069
3070
}
3070
3071
3071
3072
int scx_tg_online(struct task_group *tg)
···
3215
3214
3216
3215
void scx_group_set_idle(struct task_group *tg, bool idle)
3217
3216
{
3218
-
/* TODO: Implement ops->cgroup_set_idle() */
3217
+
struct scx_sched *sch = scx_root;
3218
+
3219
+
percpu_down_read(&scx_cgroup_ops_rwsem);
3220
+
3221
+
if (scx_cgroup_enabled && SCX_HAS_OP(sch, cgroup_set_idle))
3222
+
SCX_CALL_OP(sch, SCX_KF_UNLOCKED, cgroup_set_idle, NULL,
3223
+
tg_cgrp(tg), idle);
3224
+
3225
+
/* Update the task group's idle state */
3226
+
tg->scx.idle = idle;
3227
+
3228
+
percpu_up_read(&scx_cgroup_ops_rwsem);
3219
3229
}
3220
3230
3221
3231
void scx_group_set_bandwidth(struct task_group *tg,
···
5029
5017
static void sched_ext_ops__cgroup_cancel_move(struct task_struct *p, struct cgroup *from, struct cgroup *to) {}
5030
5018
static void sched_ext_ops__cgroup_set_weight(struct cgroup *cgrp, u32 weight) {}
5031
5019
static void sched_ext_ops__cgroup_set_bandwidth(struct cgroup *cgrp, u64 period_us, u64 quota_us, u64 burst_us) {}
5020
+
static void sched_ext_ops__cgroup_set_idle(struct cgroup *cgrp, bool idle) {}
5032
5021
#endif
5033
5022
static void sched_ext_ops__cpu_online(s32 cpu) {}
5034
5023
static void sched_ext_ops__cpu_offline(s32 cpu) {}
···
5068
5055
.cgroup_cancel_move = sched_ext_ops__cgroup_cancel_move,
5069
5056
.cgroup_set_weight = sched_ext_ops__cgroup_set_weight,
5070
5057
.cgroup_set_bandwidth = sched_ext_ops__cgroup_set_bandwidth,
5058
+
.cgroup_set_idle = sched_ext_ops__cgroup_set_idle,
5071
5059
#endif
5072
5060
.cpu_online = sched_ext_ops__cpu_online,
5073
5061
.cpu_offline = sched_ext_ops__cpu_offline,
+12
-1
kernel/sched/ext_internal.h
+12
-1
kernel/sched/ext_internal.h
···
697
697
* 2_500_000. @cgrp is entitled to 2.5 CPUs. @burst_us can be
698
698
* interpreted in the same fashion and specifies how much @cgrp can
699
699
* burst temporarily. The specific control mechanism and thus the
700
-
* interpretation of @period_us and burstiness is upto to the BPF
700
+
* interpretation of @period_us and burstiness is up to the BPF
701
701
* scheduler.
702
702
*/
703
703
void (*cgroup_set_bandwidth)(struct cgroup *cgrp,
704
704
u64 period_us, u64 quota_us, u64 burst_us);
705
+
706
+
/**
707
+
* @cgroup_set_idle: A cgroup's idle state is being changed
708
+
* @cgrp: cgroup whose idle state is being updated
709
+
* @idle: whether the cgroup is entering or exiting idle state
710
+
*
711
+
* Update @cgrp's idle state to @idle. This callback is invoked when
712
+
* a cgroup transitions between idle and non-idle states, allowing the
713
+
* BPF scheduler to adjust its behavior accordingly.
714
+
*/
715
+
void (*cgroup_set_idle)(struct cgroup *cgrp, bool idle);
705
716
706
717
#endif /* CONFIG_EXT_GROUP_SCHED */
707
718