Currently, there is a potential race between distribute_cfs_runtime()
and assign_cfs_rq_runtime(). Race happens when cfs_b->runtime is read,
distributes without holding lock and finds out there is not enough
runtime to charge against after distribution. Because
assign_cfs_rq_runtime() might be called during distribution, and use
cfs_b->runtime at the same time.
Fibtest is the tool to test this race. Assume all gcfs_rq is throttled
and cfs period timer runs, slow threads might run and sleep, returning
unused cfs_rq runtime and keeping min_cfs_rq_runtime in their local
pool. If all this happens sufficiently quickly, cfs_b->runtime will drop
a lot. If runtime distributed is large too, over-use of runtime happens.
A runtime over-using by about 70 percent of quota is seen when we
test fibtest on a 96-core machine. We run fibtest with 1 fast thread and
95 slow threads in test group, configure 10ms quota for this group and
see the CPU usage of fibtest is 17.0%, which is far from than the
expected 10%.
On a smaller machine with 32 cores, we also run fibtest with 96
threads. CPU usage is more than 12%, which is also more than expected
10%. This shows that on similar workloads, this race do affect CPU
bandwidth control.
Solve this by holding lock inside distribute_cfs_runtime().
Fixes: c06f04c70489 ("sched: Fix potential near-infinite distribute_cfs_runtime() loop")
Signed-off-by: Huaixin Chang <[email protected]>
---
kernel/sched/fair.c | 31 ++++++++++++-------------------
1 file changed, 12 insertions(+), 19 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c1217bfe5e81..c9f0e89fe5da 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4629,11 +4629,11 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
resched_curr(rq);
}
-static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
+static void distribute_cfs_runtime(struct cfs_bandwidth *cfs_b)
{
struct cfs_rq *cfs_rq;
- u64 runtime;
- u64 starting_runtime = remaining;
+ u64 runtime, remaining;
+ unsigned long flags;
rcu_read_lock();
list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq,
@@ -4648,10 +4648,13 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
/* By the above check, this should never be true */
SCHED_WARN_ON(cfs_rq->runtime_remaining > 0);
+ raw_spin_lock_irqsave(&cfs_b->lock, flags);
runtime = -cfs_rq->runtime_remaining + 1;
- if (runtime > remaining)
- runtime = remaining;
- remaining -= runtime;
+ if (runtime > cfs_b->runtime)
+ runtime = cfs_b->runtime;
+ cfs_b->runtime -= runtime;
+ remaining = cfs_b->runtime;
+ raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
cfs_rq->runtime_remaining += runtime;
@@ -4666,8 +4669,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
break;
}
rcu_read_unlock();
-
- return starting_runtime - remaining;
}
/*
@@ -4678,7 +4679,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
*/
static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, unsigned long flags)
{
- u64 runtime;
int throttled;
/* no need to continue the timer with no bandwidth constraint */
@@ -4708,23 +4708,17 @@ static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, u
/*
* This check is repeated as we are holding onto the new bandwidth while
- * we unthrottle. This can potentially race with an unthrottled group
- * trying to acquire new bandwidth from the global pool. This can result
- * in us over-using our runtime if it is all used during this loop, but
- * only by limited amounts in that extreme case.
+ * we unthrottle.
*/
while (throttled && cfs_b->runtime > 0 && !cfs_b->distribute_running) {
- runtime = cfs_b->runtime;
cfs_b->distribute_running = 1;
raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
/* we can't nest cfs_b->lock while distributing bandwidth */
- runtime = distribute_cfs_runtime(cfs_b, runtime);
+ distribute_cfs_runtime(cfs_b);
raw_spin_lock_irqsave(&cfs_b->lock, flags);
cfs_b->distribute_running = 0;
throttled = !list_empty(&cfs_b->throttled_cfs_rq);
-
- lsub_positive(&cfs_b->runtime, runtime);
}
/*
@@ -4858,10 +4852,9 @@ static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b)
if (!runtime)
return;
- runtime = distribute_cfs_runtime(cfs_b, runtime);
+ distribute_cfs_runtime(cfs_b);
raw_spin_lock_irqsave(&cfs_b->lock, flags);
- lsub_positive(&cfs_b->runtime, runtime);
cfs_b->distribute_running = 0;
raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
}
--
2.14.4.44.g2045bb6
Currently, there is a potential race between distribute_cfs_runtime()
and assign_cfs_rq_runtime(). Race happens when cfs_b->runtime is read,
distributes without holding lock and finds out there is not enough
runtime to charge against after distribution. Because
assign_cfs_rq_runtime() might be called during distribution, and use
cfs_b->runtime at the same time.
Fibtest is the tool to test this race. Assume all gcfs_rq is throttled
and cfs period timer runs, slow threads might run and sleep, returning
unused cfs_rq runtime and keeping min_cfs_rq_runtime in their local
pool. If all this happens sufficiently quickly, cfs_b->runtime will drop
a lot. If runtime distributed is large too, over-use of runtime happens.
A runtime over-using by about 70 percent of quota is seen when we
test fibtest on a 96-core machine. We run fibtest with 1 fast thread and
95 slow threads in test group, configure 10ms quota for this group and
see the CPU usage of fibtest is 17.0%, which is far more than the
expected 10%.
On a smaller machine with 32 cores, we also run fibtest with 96
threads. CPU usage is more than 12%, which is also more than expected
10%. This shows that on similar workloads, this race do affect CPU
bandwidth control.
Solve this by holding lock inside distribute_cfs_runtime().
Fixes: c06f04c70489 ("sched: Fix potential near-infinite distribute_cfs_runtime() loop")
Signed-off-by: Huaixin Chang <[email protected]>
---
v2: fix spelling, initialize variable rumaining in distribute_cfs_runtime()
---
kernel/sched/fair.c | 31 ++++++++++++-------------------
1 file changed, 12 insertions(+), 19 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c1217bfe5e81..fd30e06a7ffc 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4629,11 +4629,11 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
resched_curr(rq);
}
-static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
+static void distribute_cfs_runtime(struct cfs_bandwidth *cfs_b)
{
struct cfs_rq *cfs_rq;
- u64 runtime;
- u64 starting_runtime = remaining;
+ u64 runtime, remaining = 1;
+ unsigned long flags;
rcu_read_lock();
list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq,
@@ -4648,10 +4648,13 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
/* By the above check, this should never be true */
SCHED_WARN_ON(cfs_rq->runtime_remaining > 0);
+ raw_spin_lock_irqsave(&cfs_b->lock, flags);
runtime = -cfs_rq->runtime_remaining + 1;
- if (runtime > remaining)
- runtime = remaining;
- remaining -= runtime;
+ if (runtime > cfs_b->runtime)
+ runtime = cfs_b->runtime;
+ cfs_b->runtime -= runtime;
+ remaining = cfs_b->runtime;
+ raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
cfs_rq->runtime_remaining += runtime;
@@ -4666,8 +4669,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
break;
}
rcu_read_unlock();
-
- return starting_runtime - remaining;
}
/*
@@ -4678,7 +4679,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
*/
static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, unsigned long flags)
{
- u64 runtime;
int throttled;
/* no need to continue the timer with no bandwidth constraint */
@@ -4708,23 +4708,17 @@ static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, u
/*
* This check is repeated as we are holding onto the new bandwidth while
- * we unthrottle. This can potentially race with an unthrottled group
- * trying to acquire new bandwidth from the global pool. This can result
- * in us over-using our runtime if it is all used during this loop, but
- * only by limited amounts in that extreme case.
+ * we unthrottle.
*/
while (throttled && cfs_b->runtime > 0 && !cfs_b->distribute_running) {
- runtime = cfs_b->runtime;
cfs_b->distribute_running = 1;
raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
/* we can't nest cfs_b->lock while distributing bandwidth */
- runtime = distribute_cfs_runtime(cfs_b, runtime);
+ distribute_cfs_runtime(cfs_b);
raw_spin_lock_irqsave(&cfs_b->lock, flags);
cfs_b->distribute_running = 0;
throttled = !list_empty(&cfs_b->throttled_cfs_rq);
-
- lsub_positive(&cfs_b->runtime, runtime);
}
/*
@@ -4858,10 +4852,9 @@ static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b)
if (!runtime)
return;
- runtime = distribute_cfs_runtime(cfs_b, runtime);
+ distribute_cfs_runtime(cfs_b);
raw_spin_lock_irqsave(&cfs_b->lock, flags);
- lsub_positive(&cfs_b->runtime, runtime);
cfs_b->distribute_running = 0;
raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
}
--
2.14.4.44.g2045bb6
Huaixin Chang <[email protected]> writes:
> Currently, there is a potential race between distribute_cfs_runtime()
> and assign_cfs_rq_runtime(). Race happens when cfs_b->runtime is read,
> distributes without holding lock and finds out there is not enough
> runtime to charge against after distribution. Because
> assign_cfs_rq_runtime() might be called during distribution, and use
> cfs_b->runtime at the same time.
>
> Fibtest is the tool to test this race. Assume all gcfs_rq is throttled
> and cfs period timer runs, slow threads might run and sleep, returning
> unused cfs_rq runtime and keeping min_cfs_rq_runtime in their local
> pool. If all this happens sufficiently quickly, cfs_b->runtime will drop
> a lot. If runtime distributed is large too, over-use of runtime happens.
>
> A runtime over-using by about 70 percent of quota is seen when we
> test fibtest on a 96-core machine. We run fibtest with 1 fast thread and
> 95 slow threads in test group, configure 10ms quota for this group and
> see the CPU usage of fibtest is 17.0%, which is far more than the
> expected 10%.
>
> On a smaller machine with 32 cores, we also run fibtest with 96
> threads. CPU usage is more than 12%, which is also more than expected
> 10%. This shows that on similar workloads, this race do affect CPU
> bandwidth control.
>
> Solve this by holding lock inside distribute_cfs_runtime().
Some minor requests below, other than that
Reviewed-by: Ben Segall <[email protected]>
>
> Fixes: c06f04c70489 ("sched: Fix potential near-infinite distribute_cfs_runtime() loop")
> Signed-off-by: Huaixin Chang <[email protected]>
> ---
> v2: fix spelling, initialize variable rumaining in distribute_cfs_runtime()
> ---
> kernel/sched/fair.c | 31 ++++++++++++-------------------
> 1 file changed, 12 insertions(+), 19 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index c1217bfe5e81..fd30e06a7ffc 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -4629,11 +4629,11 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
> resched_curr(rq);
> }
>
> -static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
> +static void distribute_cfs_runtime(struct cfs_bandwidth *cfs_b)
> {
> struct cfs_rq *cfs_rq;
> - u64 runtime;
> - u64 starting_runtime = remaining;
> + u64 runtime, remaining = 1;
> + unsigned long flags;
>
> rcu_read_lock();
> list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq,
> @@ -4648,10 +4648,13 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
> /* By the above check, this should never be true */
> SCHED_WARN_ON(cfs_rq->runtime_remaining > 0);
>
> + raw_spin_lock_irqsave(&cfs_b->lock, flags);
No need for _irqsave/_irqrestore, the rqlock already did.
> runtime = -cfs_rq->runtime_remaining + 1;
> - if (runtime > remaining)
> - runtime = remaining;
> - remaining -= runtime;
> + if (runtime > cfs_b->runtime)
> + runtime = cfs_b->runtime;
> + cfs_b->runtime -= runtime;
> + remaining = cfs_b->runtime;
> + raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
>
> cfs_rq->runtime_remaining += runtime;
>
> @@ -4666,8 +4669,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
> break;
> }
> rcu_read_unlock();
> -
> - return starting_runtime - remaining;
> }
>
> /*
> @@ -4678,7 +4679,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
> */
> static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, unsigned long flags)
> {
> - u64 runtime;
> int throttled;
>
> /* no need to continue the timer with no bandwidth constraint */
> @@ -4708,23 +4708,17 @@ static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, u
>
> /*
> * This check is repeated as we are holding onto the new bandwidth while
> - * we unthrottle. This can potentially race with an unthrottled group
> - * trying to acquire new bandwidth from the global pool. This can result
> - * in us over-using our runtime if it is all used during this loop, but
> - * only by limited amounts in that extreme case.
> + * we unthrottle.
"This check is repeated as we release cfs_b->lock while we unthrottle."
or something like that. This code is no longer even holding onto the new
bandwidth on its own.
> */
> while (throttled && cfs_b->runtime > 0 && !cfs_b->distribute_running) {
> - runtime = cfs_b->runtime;
> cfs_b->distribute_running = 1;
> raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
> /* we can't nest cfs_b->lock while distributing bandwidth */
> - runtime = distribute_cfs_runtime(cfs_b, runtime);
> + distribute_cfs_runtime(cfs_b);
> raw_spin_lock_irqsave(&cfs_b->lock, flags);
>
> cfs_b->distribute_running = 0;
> throttled = !list_empty(&cfs_b->throttled_cfs_rq);
> -
> - lsub_positive(&cfs_b->runtime, runtime);
> }
>
> /*
> @@ -4858,10 +4852,9 @@ static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b)
> if (!runtime)
> return;
>
> - runtime = distribute_cfs_runtime(cfs_b, runtime);
> + distribute_cfs_runtime(cfs_b);
>
> raw_spin_lock_irqsave(&cfs_b->lock, flags);
> - lsub_positive(&cfs_b->runtime, runtime);
> cfs_b->distribute_running = 0;
> raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
> }
Currently, there is a potential race between distribute_cfs_runtime()
and assign_cfs_rq_runtime(). Race happens when cfs_b->runtime is read,
distributes without holding lock and finds out there is not enough
runtime to charge against after distribution. Because
assign_cfs_rq_runtime() might be called during distribution, and use
cfs_b->runtime at the same time.
Fibtest is the tool to test this race. Assume all gcfs_rq is throttled
and cfs period timer runs, slow threads might run and sleep, returning
unused cfs_rq runtime and keeping min_cfs_rq_runtime in their local
pool. If all this happens sufficiently quickly, cfs_b->runtime will drop
a lot. If runtime distributed is large too, over-use of runtime happens.
A runtime over-using by about 70 percent of quota is seen when we
test fibtest on a 96-core machine. We run fibtest with 1 fast thread and
95 slow threads in test group, configure 10ms quota for this group and
see the CPU usage of fibtest is 17.0%, which is far more than the
expected 10%.
On a smaller machine with 32 cores, we also run fibtest with 96
threads. CPU usage is more than 12%, which is also more than expected
10%. This shows that on similar workloads, this race do affect CPU
bandwidth control.
Solve this by holding lock inside distribute_cfs_runtime().
Fixes: c06f04c70489 ("sched: Fix potential near-infinite distribute_cfs_runtime() loop")
Signed-off-by: Huaixin Chang <[email protected]>
Reviewed-by: Ben Segall <[email protected]>
Link: https://lore.kernel.org/lkml/[email protected]/
---
kernel/sched/fair.c | 31 +++++++++++--------------------
1 file changed, 11 insertions(+), 20 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c1217bfe5e81..0eaa12d0f1b9 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4629,11 +4629,10 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
resched_curr(rq);
}
-static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
+static void distribute_cfs_runtime(struct cfs_bandwidth *cfs_b)
{
struct cfs_rq *cfs_rq;
- u64 runtime;
- u64 starting_runtime = remaining;
+ u64 runtime, remaining = 1;
rcu_read_lock();
list_for_each_entry_rcu(cfs_rq, &cfs_b->throttled_cfs_rq,
@@ -4648,10 +4647,13 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
/* By the above check, this should never be true */
SCHED_WARN_ON(cfs_rq->runtime_remaining > 0);
+ raw_spin_lock(&cfs_b->lock);
runtime = -cfs_rq->runtime_remaining + 1;
- if (runtime > remaining)
- runtime = remaining;
- remaining -= runtime;
+ if (runtime > cfs_b->runtime)
+ runtime = cfs_b->runtime;
+ cfs_b->runtime -= runtime;
+ remaining = cfs_b->runtime;
+ raw_spin_unlock(&cfs_b->lock);
cfs_rq->runtime_remaining += runtime;
@@ -4666,8 +4668,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
break;
}
rcu_read_unlock();
-
- return starting_runtime - remaining;
}
/*
@@ -4678,7 +4678,6 @@ static u64 distribute_cfs_runtime(struct cfs_bandwidth *cfs_b, u64 remaining)
*/
static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, unsigned long flags)
{
- u64 runtime;
int throttled;
/* no need to continue the timer with no bandwidth constraint */
@@ -4707,24 +4706,17 @@ static int do_sched_cfs_period_timer(struct cfs_bandwidth *cfs_b, int overrun, u
cfs_b->nr_throttled += overrun;
/*
- * This check is repeated as we are holding onto the new bandwidth while
- * we unthrottle. This can potentially race with an unthrottled group
- * trying to acquire new bandwidth from the global pool. This can result
- * in us over-using our runtime if it is all used during this loop, but
- * only by limited amounts in that extreme case.
+ * This check is repeated as we release cfs_b->lock while we unthrottle.
*/
while (throttled && cfs_b->runtime > 0 && !cfs_b->distribute_running) {
- runtime = cfs_b->runtime;
cfs_b->distribute_running = 1;
raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
/* we can't nest cfs_b->lock while distributing bandwidth */
- runtime = distribute_cfs_runtime(cfs_b, runtime);
+ distribute_cfs_runtime(cfs_b);
raw_spin_lock_irqsave(&cfs_b->lock, flags);
cfs_b->distribute_running = 0;
throttled = !list_empty(&cfs_b->throttled_cfs_rq);
-
- lsub_positive(&cfs_b->runtime, runtime);
}
/*
@@ -4858,10 +4850,9 @@ static void do_sched_cfs_slack_timer(struct cfs_bandwidth *cfs_b)
if (!runtime)
return;
- runtime = distribute_cfs_runtime(cfs_b, runtime);
+ distribute_cfs_runtime(cfs_b);
raw_spin_lock_irqsave(&cfs_b->lock, flags);
- lsub_positive(&cfs_b->runtime, runtime);
cfs_b->distribute_running = 0;
raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
}
--
2.14.4.44.g2045bb6
On Fri, Mar 27, 2020 at 11:26:25AM +0800, Huaixin Chang wrote:
> Currently, there is a potential race between distribute_cfs_runtime()
> and assign_cfs_rq_runtime(). Race happens when cfs_b->runtime is read,
> distributes without holding lock and finds out there is not enough
> runtime to charge against after distribution. Because
> assign_cfs_rq_runtime() might be called during distribution, and use
> cfs_b->runtime at the same time.
>
> Fibtest is the tool to test this race. Assume all gcfs_rq is throttled
> and cfs period timer runs, slow threads might run and sleep, returning
> unused cfs_rq runtime and keeping min_cfs_rq_runtime in their local
> pool. If all this happens sufficiently quickly, cfs_b->runtime will drop
> a lot. If runtime distributed is large too, over-use of runtime happens.
>
> A runtime over-using by about 70 percent of quota is seen when we
> test fibtest on a 96-core machine. We run fibtest with 1 fast thread and
> 95 slow threads in test group, configure 10ms quota for this group and
> see the CPU usage of fibtest is 17.0%, which is far more than the
> expected 10%.
>
> On a smaller machine with 32 cores, we also run fibtest with 96
> threads. CPU usage is more than 12%, which is also more than expected
> 10%. This shows that on similar workloads, this race do affect CPU
> bandwidth control.
>
> Solve this by holding lock inside distribute_cfs_runtime().
>
> Fixes: c06f04c70489 ("sched: Fix potential near-infinite distribute_cfs_runtime() loop")
> Signed-off-by: Huaixin Chang <[email protected]>
> Reviewed-by: Ben Segall <[email protected]>
Thanks!