[PATCH] Pass the task_struct explicitly to delayacct_blkio_end

[PATCH] Pass the task_struct explicitly to delayacct_blkio_end

[Josh Snyder]

Before e33a9bba85a8, delayacct_blkio_end was called after
context-switching into the task which completed I/O. This resulted in
double counting: the task would account a delay both waiting for I/O and
for time spent in the runqueue.

With e33a9bba85a8, delayacct_blkio_end is called by try_to_wake_up. In
ttwu, we have not yet context-switched. This is more correct, in that the
delay accounting ends when the I/O is complete. But delayacct_blkio_end
relies upon `get_current()`, and we have not yet context-switched into the
task whose I/O completed. This results in the wrong task having its delay
accounting statistics updated.

Instead of doing that, pass the task_struct being woken to
delayacct_blkio_end, so that it can update the statistics of the correct
task_struct.
---
 include/linux/delayacct.h |  8 ++++----
 kernel/delayacct.c        | 42 ++++++++++++++++++++++++++----------------
 kernel/sched/core.c       |  6 +++---
 3 files changed, 33 insertions(+), 23 deletions(-)

diff --git a/include/linux/delayacct.h b/include/linux/delayacct.h
index 4178d24..f2ad868 100644
--- a/include/linux/delayacct.h
+++ b/include/linux/delayacct.h
@@ -71,7 +71,7 @@ extern void delayacct_init(void);
 extern void __delayacct_tsk_init(struct task_struct *);
 extern void __delayacct_tsk_exit(struct task_struct *);
 extern void __delayacct_blkio_start(void);
-extern void __delayacct_blkio_end(void);
+extern void __delayacct_blkio_end(struct task_struct *);
 extern int __delayacct_add_tsk(struct taskstats *, struct task_struct *);
 extern __u64 __delayacct_blkio_ticks(struct task_struct *);
 extern void __delayacct_freepages_start(void);
@@ -122,10 +122,10 @@ static inline void delayacct_blkio_start(void)
 		__delayacct_blkio_start();
 }
 
-static inline void delayacct_blkio_end(void)
+static inline void delayacct_blkio_end(struct task_struct *p)
 {
 	if (current->delays)
-		__delayacct_blkio_end();
+		__delayacct_blkio_end(p);
 	delayacct_clear_flag(DELAYACCT_PF_BLKIO);
 }
 
@@ -169,7 +169,7 @@ static inline void delayacct_tsk_free(struct task_struct *tsk)
 {}
 static inline void delayacct_blkio_start(void)
 {}
-static inline void delayacct_blkio_end(void)
+static inline void delayacct_blkio_end(struct task_struct *p)
 {}
 static inline int delayacct_add_tsk(struct taskstats *d,
 					struct task_struct *tsk)
diff --git a/kernel/delayacct.c b/kernel/delayacct.c
index 4a1c334..e2ec808 100644
--- a/kernel/delayacct.c
+++ b/kernel/delayacct.c
@@ -51,16 +51,16 @@ void __delayacct_tsk_init(struct task_struct *tsk)
  * Finish delay accounting for a statistic using its timestamps (@start),
  * accumalator (@total) and @count
  */
-static void delayacct_end(u64 *start, u64 *total, u32 *count)
+static void delayacct_end(spinlock_t *lock, u64 *start, u64 *total, u32 *count)
 {
 	s64 ns = ktime_get_ns() - *start;
 	unsigned long flags;
 
 	if (ns > 0) {
-		spin_lock_irqsave(&current->delays->lock, flags);
+		spin_lock_irqsave(lock, flags);
 		*total += ns;
 		(*count)++;
-		spin_unlock_irqrestore(&current->delays->lock, flags);
+		spin_unlock_irqrestore(lock, flags);
 	}
 }
 
@@ -69,17 +69,25 @@ void __delayacct_blkio_start(void)
 	current->delays->blkio_start = ktime_get_ns();
 }
 
-void __delayacct_blkio_end(void)
+/*
+ * We cannot rely on the `current` macro, as we haven't yet switched back to
+ * the process being woken.
+ */
+void __delayacct_blkio_end(struct task_struct *p)
 {
-	if (current->delays->flags & DELAYACCT_PF_SWAPIN)
-		/* Swapin block I/O */
-		delayacct_end(&current->delays->blkio_start,
-			&current->delays->swapin_delay,
-			&current->delays->swapin_count);
-	else	/* Other block I/O */
-		delayacct_end(&current->delays->blkio_start,
-			&current->delays->blkio_delay,
-			&current->delays->blkio_count);
+	struct task_delay_info * delays = p->delays;
+	u64 *total;
+	u32 *count;
+
+	if (p->delays->flags & DELAYACCT_PF_SWAPIN) {
+		total = &delays->swapin_delay;
+		count = &delays->swapin_count;
+	} else {
+		total = &delays->blkio_delay;
+		count = &delays->blkio_count;
+	}
+
+	delayacct_end(&delays->lock, &delays->blkio_start, total, count);
 }
 
 int __delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk)
@@ -153,8 +161,10 @@ void __delayacct_freepages_start(void)
 
 void __delayacct_freepages_end(void)
 {
-	delayacct_end(&current->delays->freepages_start,
-			&current->delays->freepages_delay,
-			&current->delays->freepages_count);
+	delayacct_end(
+		&current->delays->lock,
+		&current->delays->freepages_start,
+		&current->delays->freepages_delay,
+		&current->delays->freepages_count);
 }
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 75554f3..deda018 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2056,7 +2056,7 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 	p->state = TASK_WAKING;
 
 	if (p->in_iowait) {
-		delayacct_blkio_end();
+		delayacct_blkio_end(p);
 		atomic_dec(&task_rq(p)->nr_iowait);
 	}
 
@@ -2069,7 +2069,7 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 #else /* CONFIG_SMP */
 
 	if (p->in_iowait) {
-		delayacct_blkio_end();
+		delayacct_blkio_end(p);
 		atomic_dec(&task_rq(p)->nr_iowait);
 	}
 
@@ -2122,7 +2122,7 @@ static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
 
 	if (!task_on_rq_queued(p)) {
 		if (p->in_iowait) {
-			delayacct_blkio_end();
+			delayacct_blkio_end(p);
 			atomic_dec(&rq->nr_iowait);
 		}
 		ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);
-- 
2.7.4

Re: [PATCH] Pass the task_struct explicitly to delayacct_blkio_end

[Peter Zijlstra]

On Fri, Dec 15, 2017 at 06:11:43PM +0000, Josh Snyder wrote:

> Subject: [PATCH] Pass the task_struct explicitly to delayacct_blkio_end

That subject is crap; it needs a subsystem prefix and it needs to
describe what is done, not how its done.

Something like:

  delayacct: Fix double accounting for blk-io

> Before e33a9bba85a8, delayacct_blkio_end was called after
> context-switching into the task which completed I/O. This resulted in
> double counting: the task would account a delay both waiting for I/O and
> for time spent in the runqueue.
> 
> With e33a9bba85a8, delayacct_blkio_end is called by try_to_wake_up. In
> ttwu, we have not yet context-switched. This is more correct, in that the
> delay accounting ends when the I/O is complete. But delayacct_blkio_end
> relies upon `get_current()`, and we have not yet context-switched into the
> task whose I/O completed. This results in the wrong task having its delay
> accounting statistics updated.

The correct way to quote a commit is like:

  e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler")

> Instead of doing that, pass the task_struct being woken to
> delayacct_blkio_end, so that it can update the statistics of the correct
> task_struct.

[PATCH v2] delayacct: Account blkio completion on the correct task

[Josh Snyder]

Before commit e33a9bba85a8 ("sched/core: move IO scheduling accounting from
io_schedule_timeout() into scheduler"), delayacct_blkio_end was called after
context-switching into the task which completed I/O. This resulted in double
counting: the task would account a delay both waiting for I/O and for time
spent in the runqueue.

With e33a9bba85a8, delayacct_blkio_end is called by try_to_wake_up. In
ttwu, we have not yet context-switched. This is more correct, in that the
delay accounting ends when the I/O is complete. But delayacct_blkio_end
relies upon `get_current()`, and we have not yet context-switched into the
task whose I/O completed. This results in the wrong task having its delay
accounting statistics updated.

Instead of doing that, pass the task_struct being woken to
delayacct_blkio_end, so that it can update the statistics of the correct
task.

Fixes: e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler")
Signed-off-by: Josh Snyder <joshs@netflix.com>
---
 include/linux/delayacct.h |  8 ++++----
 kernel/delayacct.c        | 42 ++++++++++++++++++++++++++----------------
 kernel/sched/core.c       |  6 +++---
 3 files changed, 33 insertions(+), 23 deletions(-)

diff --git a/include/linux/delayacct.h b/include/linux/delayacct.h
index 4178d24..f2ad868 100644
--- a/include/linux/delayacct.h
+++ b/include/linux/delayacct.h
@@ -71,7 +71,7 @@ extern void delayacct_init(void);
 extern void __delayacct_tsk_init(struct task_struct *);
 extern void __delayacct_tsk_exit(struct task_struct *);
 extern void __delayacct_blkio_start(void);
-extern void __delayacct_blkio_end(void);
+extern void __delayacct_blkio_end(struct task_struct *);
 extern int __delayacct_add_tsk(struct taskstats *, struct task_struct *);
 extern __u64 __delayacct_blkio_ticks(struct task_struct *);
 extern void __delayacct_freepages_start(void);
@@ -122,10 +122,10 @@ static inline void delayacct_blkio_start(void)
 		__delayacct_blkio_start();
 }
 
-static inline void delayacct_blkio_end(void)
+static inline void delayacct_blkio_end(struct task_struct *p)
 {
 	if (current->delays)
-		__delayacct_blkio_end();
+		__delayacct_blkio_end(p);
 	delayacct_clear_flag(DELAYACCT_PF_BLKIO);
 }
 
@@ -169,7 +169,7 @@ static inline void delayacct_tsk_free(struct task_struct *tsk)
 {}
 static inline void delayacct_blkio_start(void)
 {}
-static inline void delayacct_blkio_end(void)
+static inline void delayacct_blkio_end(struct task_struct *p)
 {}
 static inline int delayacct_add_tsk(struct taskstats *d,
 					struct task_struct *tsk)
diff --git a/kernel/delayacct.c b/kernel/delayacct.c
index 4a1c334..e2ec808 100644
--- a/kernel/delayacct.c
+++ b/kernel/delayacct.c
@@ -51,16 +51,16 @@ void __delayacct_tsk_init(struct task_struct *tsk)
  * Finish delay accounting for a statistic using its timestamps (@start),
  * accumalator (@total) and @count
  */
-static void delayacct_end(u64 *start, u64 *total, u32 *count)
+static void delayacct_end(spinlock_t *lock, u64 *start, u64 *total, u32 *count)
 {
 	s64 ns = ktime_get_ns() - *start;
 	unsigned long flags;
 
 	if (ns > 0) {
-		spin_lock_irqsave(&current->delays->lock, flags);
+		spin_lock_irqsave(lock, flags);
 		*total += ns;
 		(*count)++;
-		spin_unlock_irqrestore(&current->delays->lock, flags);
+		spin_unlock_irqrestore(lock, flags);
 	}
 }
 
@@ -69,17 +69,25 @@ void __delayacct_blkio_start(void)
 	current->delays->blkio_start = ktime_get_ns();
 }
 
-void __delayacct_blkio_end(void)
+/*
+ * We cannot rely on the `current` macro, as we haven't yet switched back to
+ * the process being woken.
+ */
+void __delayacct_blkio_end(struct task_struct *p)
 {
-	if (current->delays->flags & DELAYACCT_PF_SWAPIN)
-		/* Swapin block I/O */
-		delayacct_end(&current->delays->blkio_start,
-			&current->delays->swapin_delay,
-			&current->delays->swapin_count);
-	else	/* Other block I/O */
-		delayacct_end(&current->delays->blkio_start,
-			&current->delays->blkio_delay,
-			&current->delays->blkio_count);
+	struct task_delay_info *delays = p->delays;
+	u64 *total;
+	u32 *count;
+
+	if (p->delays->flags & DELAYACCT_PF_SWAPIN) {
+		total = &delays->swapin_delay;
+		count = &delays->swapin_count;
+	} else {
+		total = &delays->blkio_delay;
+		count = &delays->blkio_count;
+	}
+
+	delayacct_end(&delays->lock, &delays->blkio_start, total, count);
 }
 
 int __delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk)
@@ -153,8 +161,10 @@ void __delayacct_freepages_start(void)
 
 void __delayacct_freepages_end(void)
 {
-	delayacct_end(&current->delays->freepages_start,
-			&current->delays->freepages_delay,
-			&current->delays->freepages_count);
+	delayacct_end(
+		&current->delays->lock,
+		&current->delays->freepages_start,
+		&current->delays->freepages_delay,
+		&current->delays->freepages_count);
 }
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 75554f3..deda018 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2056,7 +2056,7 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 	p->state = TASK_WAKING;
 
 	if (p->in_iowait) {
-		delayacct_blkio_end();
+		delayacct_blkio_end(p);
 		atomic_dec(&task_rq(p)->nr_iowait);
 	}
 
@@ -2069,7 +2069,7 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 #else /* CONFIG_SMP */
 
 	if (p->in_iowait) {
-		delayacct_blkio_end();
+		delayacct_blkio_end(p);
 		atomic_dec(&task_rq(p)->nr_iowait);
 	}
 
@@ -2122,7 +2122,7 @@ static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
 
 	if (!task_on_rq_queued(p)) {
 		if (p->in_iowait) {
-			delayacct_blkio_end();
+			delayacct_blkio_end(p);
 			atomic_dec(&rq->nr_iowait);
 		}
 		ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);
-- 
2.7.4

Re: [PATCH v2] delayacct: Account blkio completion on the correct task

[Tejun Heo]

Hello,

On Mon, Dec 18, 2017 at 04:15:10PM +0000, Josh Snyder wrote:
> Before commit e33a9bba85a8 ("sched/core: move IO scheduling accounting from
> io_schedule_timeout() into scheduler"), delayacct_blkio_end was called after
> context-switching into the task which completed I/O. This resulted in double
> counting: the task would account a delay both waiting for I/O and for time
> spent in the runqueue.
> 
> With e33a9bba85a8, delayacct_blkio_end is called by try_to_wake_up. In
> ttwu, we have not yet context-switched. This is more correct, in that the
> delay accounting ends when the I/O is complete. But delayacct_blkio_end
> relies upon `get_current()`, and we have not yet context-switched into the
> task whose I/O completed. This results in the wrong task having its delay
> accounting statistics updated.
> 
> Instead of doing that, pass the task_struct being woken to
> delayacct_blkio_end, so that it can update the statistics of the correct
> task.
> 
> Fixes: e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler")
> Signed-off-by: Josh Snyder <joshs@netflix.com>

Acked-by: Tejun Heo <tj@kernel.org>

Thanks.

-- 
tejun

Re: [PATCH v2] delayacct: Account blkio completion on the correct task

[Balbir Singh]

On Mon, Dec 18, 2017 at 9:45 PM, Josh Snyder <joshs@netflix.com> wrote:
> Before commit e33a9bba85a8 ("sched/core: move IO scheduling accounting from
> io_schedule_timeout() into scheduler"), delayacct_blkio_end was called after
> context-switching into the task which completed I/O. This resulted in double
> counting: the task would account a delay both waiting for I/O and for time
> spent in the runqueue.
>

Yes, we included the time spent in the runqueue to delays on account of I/O.

> With e33a9bba85a8, delayacct_blkio_end is called by try_to_wake_up. In
> ttwu, we have not yet context-switched. This is more correct, in that the
> delay accounting ends when the I/O is complete. But delayacct_blkio_end
> relies upon `get_current()`, and we have not yet context-switched into the
> task whose I/O completed. This results in the wrong task having its delay
> accounting statistics updated.
>
> Instead of doing that, pass the task_struct being woken to
> delayacct_blkio_end, so that it can update the statistics of the correct
> task.
>
> Fixes: e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler")
> Signed-off-by: Josh Snyder <joshs@netflix.com>
> ---

Acked-by: Balbir Singh <bsingharora@gmail.com>