The commit d583d360a620 ("psi: Fix psi state corruption when schedule()
races with cgroup move") fixed a race problem by making cgroup_move_task()
use task->psi_flags instead of looking at the scheduler state.
We can extend task->psi_flags usage to CPU migration, which should be
a minor optimization for performance and code simplicity.
Signed-off-by: Chengming Zhou <[email protected]>
---
include/linux/sched.h | 3 ---
kernel/sched/core.c | 2 +-
kernel/sched/stats.h | 22 ++++------------------
3 files changed, 5 insertions(+), 22 deletions(-)
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 5cdf746988a1..cffce0eaf640 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -890,9 +890,6 @@ struct task_struct {
unsigned sched_reset_on_fork:1;
unsigned sched_contributes_to_load:1;
unsigned sched_migrated:1;
-#ifdef CONFIG_PSI
- unsigned sched_psi_wake_requeue:1;
-#endif
/* Force alignment to the next boundary: */
unsigned :0;
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a7fa9802dfde..2d98b9b5185f 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2056,7 +2056,7 @@ static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
if (!(flags & ENQUEUE_RESTORE)) {
sched_info_enqueue(rq, p);
- psi_enqueue(p, flags & ENQUEUE_WAKEUP);
+ psi_enqueue(p, (flags & ENQUEUE_WAKEUP) && !(flags & ENQUEUE_MIGRATED));
}
uclamp_rq_inc(rq, p);
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index 84a188913cc9..38f3698f5e5b 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -128,11 +128,9 @@ static inline void psi_enqueue(struct task_struct *p, bool wakeup)
if (p->in_memstall)
set |= TSK_MEMSTALL_RUNNING;
- if (!wakeup || p->sched_psi_wake_requeue) {
+ if (!wakeup) {
if (p->in_memstall)
set |= TSK_MEMSTALL;
- if (p->sched_psi_wake_requeue)
- p->sched_psi_wake_requeue = 0;
} else {
if (p->in_iowait)
clear |= TSK_IOWAIT;
@@ -143,8 +141,6 @@ static inline void psi_enqueue(struct task_struct *p, bool wakeup)
static inline void psi_dequeue(struct task_struct *p, bool sleep)
{
- int clear = TSK_RUNNING;
-
if (static_branch_likely(&psi_disabled))
return;
@@ -157,10 +153,7 @@ static inline void psi_dequeue(struct task_struct *p, bool sleep)
if (sleep)
return;
- if (p->in_memstall)
- clear |= (TSK_MEMSTALL | TSK_MEMSTALL_RUNNING);
-
- psi_task_change(p, clear, 0);
+ psi_task_change(p, p->psi_flags, 0);
}
static inline void psi_ttwu_dequeue(struct task_struct *p)
@@ -172,19 +165,12 @@ static inline void psi_ttwu_dequeue(struct task_struct *p)
* deregister its sleep-persistent psi states from the old
* queue, and let psi_enqueue() know it has to requeue.
*/
- if (unlikely(p->in_iowait || p->in_memstall)) {
+ if (unlikely(p->psi_flags)) {
struct rq_flags rf;
struct rq *rq;
- int clear = 0;
-
- if (p->in_iowait)
- clear |= TSK_IOWAIT;
- if (p->in_memstall)
- clear |= TSK_MEMSTALL;
rq = __task_rq_lock(p, &rf);
- psi_task_change(p, clear, 0);
- p->sched_psi_wake_requeue = 1;
+ psi_task_change(p, p->psi_flags, 0);
__task_rq_unlock(rq, &rf);
}
}
--
2.37.2
On Mon, Sep 26, 2022 at 04:19:31PM +0800, Chengming Zhou wrote:
> The commit d583d360a620 ("psi: Fix psi state corruption when schedule()
> races with cgroup move") fixed a race problem by making cgroup_move_task()
> use task->psi_flags instead of looking at the scheduler state.
>
> We can extend task->psi_flags usage to CPU migration, which should be
> a minor optimization for performance and code simplicity.
>
> Signed-off-by: Chengming Zhou <[email protected]>
Nice one!
Acked-by: Johannes Weiner <[email protected]>
The following commit has been merged into the sched/core branch of tip:
Commit-ID: 52b33d87b9197c51e8ffdc61873739d90dd0a16f
Gitweb: https://git.kernel.org/tip/52b33d87b9197c51e8ffdc61873739d90dd0a16f
Author: Chengming Zhou <[email protected]>
AuthorDate: Mon, 26 Sep 2022 16:19:31 +08:00
Committer: Peter Zijlstra <[email protected]>
CommitterDate: Sun, 30 Oct 2022 10:12:15 +01:00
sched/psi: Use task->psi_flags to clear in CPU migration
The commit d583d360a620 ("psi: Fix psi state corruption when schedule()
races with cgroup move") fixed a race problem by making cgroup_move_task()
use task->psi_flags instead of looking at the scheduler state.
We can extend task->psi_flags usage to CPU migration, which should be
a minor optimization for performance and code simplicity.
Signed-off-by: Chengming Zhou <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Johannes Weiner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
---
include/linux/sched.h | 3 ---
kernel/sched/core.c | 2 +-
kernel/sched/stats.h | 22 ++++------------------
3 files changed, 5 insertions(+), 22 deletions(-)
diff --git a/include/linux/sched.h b/include/linux/sched.h
index ffb6eb5..23de7fe 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -888,9 +888,6 @@ struct task_struct {
unsigned sched_reset_on_fork:1;
unsigned sched_contributes_to_load:1;
unsigned sched_migrated:1;
-#ifdef CONFIG_PSI
- unsigned sched_psi_wake_requeue:1;
-#endif
/* Force alignment to the next boundary: */
unsigned :0;
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 87c9cdf..07ac08c 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2053,7 +2053,7 @@ static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
if (!(flags & ENQUEUE_RESTORE)) {
sched_info_enqueue(rq, p);
- psi_enqueue(p, flags & ENQUEUE_WAKEUP);
+ psi_enqueue(p, (flags & ENQUEUE_WAKEUP) && !(flags & ENQUEUE_MIGRATED));
}
uclamp_rq_inc(rq, p);
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index 84a1889..38f3698 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -128,11 +128,9 @@ static inline void psi_enqueue(struct task_struct *p, bool wakeup)
if (p->in_memstall)
set |= TSK_MEMSTALL_RUNNING;
- if (!wakeup || p->sched_psi_wake_requeue) {
+ if (!wakeup) {
if (p->in_memstall)
set |= TSK_MEMSTALL;
- if (p->sched_psi_wake_requeue)
- p->sched_psi_wake_requeue = 0;
} else {
if (p->in_iowait)
clear |= TSK_IOWAIT;
@@ -143,8 +141,6 @@ static inline void psi_enqueue(struct task_struct *p, bool wakeup)
static inline void psi_dequeue(struct task_struct *p, bool sleep)
{
- int clear = TSK_RUNNING;
-
if (static_branch_likely(&psi_disabled))
return;
@@ -157,10 +153,7 @@ static inline void psi_dequeue(struct task_struct *p, bool sleep)
if (sleep)
return;
- if (p->in_memstall)
- clear |= (TSK_MEMSTALL | TSK_MEMSTALL_RUNNING);
-
- psi_task_change(p, clear, 0);
+ psi_task_change(p, p->psi_flags, 0);
}
static inline void psi_ttwu_dequeue(struct task_struct *p)
@@ -172,19 +165,12 @@ static inline void psi_ttwu_dequeue(struct task_struct *p)
* deregister its sleep-persistent psi states from the old
* queue, and let psi_enqueue() know it has to requeue.
*/
- if (unlikely(p->in_iowait || p->in_memstall)) {
+ if (unlikely(p->psi_flags)) {
struct rq_flags rf;
struct rq *rq;
- int clear = 0;
-
- if (p->in_iowait)
- clear |= TSK_IOWAIT;
- if (p->in_memstall)
- clear |= TSK_MEMSTALL;
rq = __task_rq_lock(p, &rf);
- psi_task_change(p, clear, 0);
- p->sched_psi_wake_requeue = 1;
+ psi_task_change(p, p->psi_flags, 0);
__task_rq_unlock(rq, &rf);
}
}