aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorJosh Snyder <joshs@netflix.com>2017-12-18 11:15:10 -0500
committerIngo Molnar <mingo@kernel.org>2018-01-15 21:29:36 -0500
commitc96f5471ce7d2aefd0dda560cc23f08ab00bc65d (patch)
tree84feeca0e7b0819b55216e1fea50a3e0e79d445b
parenta8750ddca918032d6349adbf9a4b6555e7db20da (diff)
delayacct: Account blkio completion on the correct task
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 on '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. Signed-off-by: Josh Snyder <joshs@netflix.com> Acked-by: Tejun Heo <tj@kernel.org> Acked-by: Balbir Singh <bsingharora@gmail.com> Cc: <stable@vger.kernel.org> Cc: Brendan Gregg <bgregg@netflix.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-block@vger.kernel.org Fixes: e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler") Link: http://lkml.kernel.org/r/1513613712-571-1-git-send-email-joshs@netflix.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
-rw-r--r--include/linux/delayacct.h8
-rw-r--r--kernel/delayacct.c42
-rw-r--r--kernel/sched/core.c6
3 files changed, 33 insertions, 23 deletions
diff --git a/include/linux/delayacct.h b/include/linux/delayacct.h
index 4178d2493547..5e335b6203f4 100644
--- a/include/linux/delayacct.h
+++ b/include/linux/delayacct.h
@@ -71,7 +71,7 @@ extern void delayacct_init(void);
71extern void __delayacct_tsk_init(struct task_struct *); 71extern void __delayacct_tsk_init(struct task_struct *);
72extern void __delayacct_tsk_exit(struct task_struct *); 72extern void __delayacct_tsk_exit(struct task_struct *);
73extern void __delayacct_blkio_start(void); 73extern void __delayacct_blkio_start(void);
74extern void __delayacct_blkio_end(void); 74extern void __delayacct_blkio_end(struct task_struct *);
75extern int __delayacct_add_tsk(struct taskstats *, struct task_struct *); 75extern int __delayacct_add_tsk(struct taskstats *, struct task_struct *);
76extern __u64 __delayacct_blkio_ticks(struct task_struct *); 76extern __u64 __delayacct_blkio_ticks(struct task_struct *);
77extern void __delayacct_freepages_start(void); 77extern void __delayacct_freepages_start(void);
@@ -122,10 +122,10 @@ static inline void delayacct_blkio_start(void)
122 __delayacct_blkio_start(); 122 __delayacct_blkio_start();
123} 123}
124 124
125static inline void delayacct_blkio_end(void) 125static inline void delayacct_blkio_end(struct task_struct *p)
126{ 126{
127 if (current->delays) 127 if (current->delays)
128 __delayacct_blkio_end(); 128 __delayacct_blkio_end(p);
129 delayacct_clear_flag(DELAYACCT_PF_BLKIO); 129 delayacct_clear_flag(DELAYACCT_PF_BLKIO);
130} 130}
131 131
@@ -169,7 +169,7 @@ static inline void delayacct_tsk_free(struct task_struct *tsk)
169{} 169{}
170static inline void delayacct_blkio_start(void) 170static inline void delayacct_blkio_start(void)
171{} 171{}
172static inline void delayacct_blkio_end(void) 172static inline void delayacct_blkio_end(struct task_struct *p)
173{} 173{}
174static inline int delayacct_add_tsk(struct taskstats *d, 174static inline int delayacct_add_tsk(struct taskstats *d,
175 struct task_struct *tsk) 175 struct task_struct *tsk)
diff --git a/kernel/delayacct.c b/kernel/delayacct.c
index 4a1c33416b6a..e2764d767f18 100644
--- a/kernel/delayacct.c
+++ b/kernel/delayacct.c
@@ -51,16 +51,16 @@ void __delayacct_tsk_init(struct task_struct *tsk)
51 * Finish delay accounting for a statistic using its timestamps (@start), 51 * Finish delay accounting for a statistic using its timestamps (@start),
52 * accumalator (@total) and @count 52 * accumalator (@total) and @count
53 */ 53 */
54static void delayacct_end(u64 *start, u64 *total, u32 *count) 54static void delayacct_end(spinlock_t *lock, u64 *start, u64 *total, u32 *count)
55{ 55{
56 s64 ns = ktime_get_ns() - *start; 56 s64 ns = ktime_get_ns() - *start;
57 unsigned long flags; 57 unsigned long flags;
58 58
59 if (ns > 0) { 59 if (ns > 0) {
60 spin_lock_irqsave(&current->delays->lock, flags); 60 spin_lock_irqsave(lock, flags);
61 *total += ns; 61 *total += ns;
62 (*count)++; 62 (*count)++;
63 spin_unlock_irqrestore(&current->delays->lock, flags); 63 spin_unlock_irqrestore(lock, flags);
64 } 64 }
65} 65}
66 66
@@ -69,17 +69,25 @@ void __delayacct_blkio_start(void)
69 current->delays->blkio_start = ktime_get_ns(); 69 current->delays->blkio_start = ktime_get_ns();
70} 70}
71 71
72void __delayacct_blkio_end(void) 72/*
73 * We cannot rely on the `current` macro, as we haven't yet switched back to
74 * the process being woken.
75 */
76void __delayacct_blkio_end(struct task_struct *p)
73{ 77{
74 if (current->delays->flags & DELAYACCT_PF_SWAPIN) 78 struct task_delay_info *delays = p->delays;
75 /* Swapin block I/O */ 79 u64 *total;
76 delayacct_end(&current->delays->blkio_start, 80 u32 *count;
77 &current->delays->swapin_delay, 81
78 &current->delays->swapin_count); 82 if (p->delays->flags & DELAYACCT_PF_SWAPIN) {
79 else /* Other block I/O */ 83 total = &delays->swapin_delay;
80 delayacct_end(&current->delays->blkio_start, 84 count = &delays->swapin_count;
81 &current->delays->blkio_delay, 85 } else {
82 &current->delays->blkio_count); 86 total = &delays->blkio_delay;
87 count = &delays->blkio_count;
88 }
89
90 delayacct_end(&delays->lock, &delays->blkio_start, total, count);
83} 91}
84 92
85int __delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk) 93int __delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk)
@@ -153,8 +161,10 @@ void __delayacct_freepages_start(void)
153 161
154void __delayacct_freepages_end(void) 162void __delayacct_freepages_end(void)
155{ 163{
156 delayacct_end(&current->delays->freepages_start, 164 delayacct_end(
157 &current->delays->freepages_delay, 165 &current->delays->lock,
158 &current->delays->freepages_count); 166 &current->delays->freepages_start,
167 &current->delays->freepages_delay,
168 &current->delays->freepages_count);
159} 169}
160 170
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 644fa2e3d993..a7bf32aabfda 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)
2056 p->state = TASK_WAKING; 2056 p->state = TASK_WAKING;
2057 2057
2058 if (p->in_iowait) { 2058 if (p->in_iowait) {
2059 delayacct_blkio_end(); 2059 delayacct_blkio_end(p);
2060 atomic_dec(&task_rq(p)->nr_iowait); 2060 atomic_dec(&task_rq(p)->nr_iowait);
2061 } 2061 }
2062 2062
@@ -2069,7 +2069,7 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
2069#else /* CONFIG_SMP */ 2069#else /* CONFIG_SMP */
2070 2070
2071 if (p->in_iowait) { 2071 if (p->in_iowait) {
2072 delayacct_blkio_end(); 2072 delayacct_blkio_end(p);
2073 atomic_dec(&task_rq(p)->nr_iowait); 2073 atomic_dec(&task_rq(p)->nr_iowait);
2074 } 2074 }
2075 2075
@@ -2122,7 +2122,7 @@ static void try_to_wake_up_local(struct task_struct *p, struct rq_flags *rf)
2122 2122
2123 if (!task_on_rq_queued(p)) { 2123 if (!task_on_rq_queued(p)) {
2124 if (p->in_iowait) { 2124 if (p->in_iowait) {
2125 delayacct_blkio_end(); 2125 delayacct_blkio_end(p);
2126 atomic_dec(&rq->nr_iowait); 2126 atomic_dec(&rq->nr_iowait);
2127 } 2127 }
2128 ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK); 2128 ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_NOCLOCK);