[PATCH] sched: fix a potential divide error

[PATCH] sched: fix a potential divide error

[Xie XiuQi]

We meet a divide error on 3.10.0 kernel, the error message is bellow:

[499992.287996] divide error: 0000 [#1] SMP
[499992.297796] do nothing after die!
[499992.299108] Modules linked in: signo_catch macvlan binfmt_misc
ip_set_hash_netport ip_set_hash_ipport vport_vxlan ipt_REJECT
xt_statistic xt_physdev xt_nat xt_recent xt_mark xt_comment ...
[499992.312751] CPU: 8 PID: 23352 Comm: bash Tainted: G ----V-------   3.10.0+ #1
[499992.314308] Hardware name: OpenStack Foundation OpenStack Nova, BIOS
rel-1.9.1-0-gb3ef39f-20170329_185309-build9a64a246a231 04/01/2014
[499992.317411] task: ffff880033fc9700 ti: ffff8807fed60000 task.ti:ffff8807fed60000
[499992.318967] RIP: 0010:[<ffffffff810c15c2>]  [<ffffffff810c15c2>] task_numa_fault+0x1c2/0xbb0
[499992.320515] RSP: 0000:ffff8807fed63d38  EFLAGS: 00010246
[499992.322018] RAX: 0000002b7efd0000 RBX: ffff880033fc9700 RCX:0000000000000003
[499992.323563] RDX: 0000000000000000 RSI: 0000000000000400 RDI:ffffffff81a80f60
[499992.325052] RBP: ffff8807fed63db8 R08: ffffffff81a80f68 R09:0000000000000000
[499992.326531] R10: ffff88083ffda000 R11: 0000000000000000 R12:0000000000000424
[499992.327987] R13: 00000000002b7efd R14: 0000000000000000 R15:ffffea001ea42a00
[499992.329420] FS:  00007fa01a3b7740(0000) GS:ffff88103ec00000(0000)knlGS:0000000000000000
[499992.330866] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[499992.332302] CR2: 0000000000ff1fb0 CR3: 00000007ff1d1000 CR4:00000000003407e0
[499992.333763] DR0: 0000000000000000 DR1: 0000000000000000 DR2:0000000000000000
[499992.335187] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:0000000000000400
[499992.336595] Stack:
[499992.337974]  0000000000000000 00000001bc9598a8 ffffea001ea42a00 0000000100000001
[499992.339374]  0000000300000001 0000000000000001 ffffea001ea42a00 ffff8807fed63db8
[499992.340768]  0000000000000000 0000000000000000 00000000bc9598a8 0000000000000001
[499992.342148] Call Trace:
[499992.343494]  [<ffffffff8119ab62>] do_numa_page+0x162/0x1f0
[499992.344831]  [<ffffffff8119bde7>] handle_mm_fault+0x627/0xf50
[499992.346145]  [<ffffffff8164e486>] __do_page_fault+0x166/0x470
[499992.347442]  [<ffffffff8164e853>] trace_do_page_fault+0x43/0x110
[499992.348711]  [<ffffffff8164df29>] do_async_page_fault+0x29/0xe0
[499992.349948]  [<ffffffff8164a9f8>] async_page_fault+0x28/0x30
[499992.351149] Code: 00 3d 00 04 00 00 44 0f 4e d8 41 81 fb 00 04 00 00
0f 84 67 07 00 00 4c 89 e8 49 83 c6 01 31 d2 48 c1 e0 10 49 83 c4 01 45
31 c9 <49> f7 f6 48 c7 45 a8 00 00 00 00 48 c7 45 b0 00 00 00 00 49 89
[499992.353707] RIP  [<ffffffff810c15c2>] task_numa_fault+0x1c2/0xbb0
[499992.354927]  RSP <ffff8807fed63d38>
[499992.358114] ---[ end trace 4f2465cac18ff65e ]---
[499992.359304] Kernel panic - not syncing: Fatal exception

sched_clock_cpu may not be consistent bwtwen cpus. If a task migrate
to another cpu, the se.exec_start was set to that cpu's rq_clock_task
by update_stats_curr_start(). Which may not be monotonic.

update_stats_curr_start
  <- set_next_entity
     <- set_curr_task_fair
        <- sched_move_task

So, if  now - p->last_task_numa_placement is -1, then (*period + 1) is
0, and divide error was triggerred at the div operation:
  task_numa_placement:
    runtime = numa_get_avg_runtime(p, &period);
    f_weight = div64_u64(runtime << 16, period + 1);  // divide error here

In this patch, we make sure period is not less than 0 to avoid this
divide error.

Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
Cc: stable@vger.kernel.org
---
 kernel/sched/fair.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 40bd1e27b1b7..f2abb258fc85 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
 	if (p->last_task_numa_placement) {
 		delta = runtime - p->last_sum_exec_runtime;
 		*period = now - p->last_task_numa_placement;
+
+		/* Avoid backward, and prevent potential divide error */
+		if ((s64)*period < 0)
+			*period = 0;
 	} else {
 		delta = p->se.avg.load_sum;
 		*period = LOAD_AVG_MAX;
-- 
2.20.1

Re: [PATCH] sched: fix a potential divide error

[Peter Zijlstra]

On Sat, Apr 20, 2019 at 04:34:16PM +0800, Xie XiuQi wrote:
> We meet a divide error on 3.10.0 kernel, the error message is bellow:

That is a _realllllllyyyy_ old kernel. I would urge you to upgrade.

> [499992.287996] divide error: 0000 [#1] SMP

> sched_clock_cpu may not be consistent bwtwen cpus. If a task migrate
> to another cpu, the se.exec_start was set to that cpu's rq_clock_task
> by update_stats_curr_start(). Which may not be monotonic.
> 
> update_stats_curr_start
>   <- set_next_entity
>      <- set_curr_task_fair
>         <- sched_move_task

That is not in fact a cross-cpu migration path. But I see the point.
Also many migration paths do in fact preserve monotonicity, even when
the clock is busted, but you're right, not all of them.

> So, if  now - p->last_task_numa_placement is -1, then (*period + 1) is
> 0, and divide error was triggerred at the div operation:
>   task_numa_placement:
>     runtime = numa_get_avg_runtime(p, &period);
>     f_weight = div64_u64(runtime << 16, period + 1);  // divide error here
> 
> In this patch, we make sure period is not less than 0 to avoid this
> divide error.
> 
> Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
> Cc: stable@vger.kernel.org
> ---
>  kernel/sched/fair.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 40bd1e27b1b7..f2abb258fc85 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
>  	if (p->last_task_numa_placement) {
>  		delta = runtime - p->last_sum_exec_runtime;
>  		*period = now - p->last_task_numa_placement;
> +
> +		/* Avoid backward, and prevent potential divide error */
> +		if ((s64)*period < 0)
> +			*period = 0;
>  	} else {
>  		delta = p->se.avg.load_sum;
>  		*period = LOAD_AVG_MAX;

Yeah, I suppose that is indeed correct.

I'll try and come up with a better Changelog tomorrow.

Re: [PATCH] sched: fix a potential divide error

[Xie XiuQi]

Hi Peter,

Thanks for your comments.

On 2019/4/24 2:44, Peter Zijlstra wrote:
> On Sat, Apr 20, 2019 at 04:34:16PM +0800, Xie XiuQi wrote:
>> We meet a divide error on 3.10.0 kernel, the error message is bellow:
> 
> That is a _realllllllyyyy_ old kernel. I would urge you to upgrade.

I will.

> 
>> [499992.287996] divide error: 0000 [#1] SMP
> 
>> sched_clock_cpu may not be consistent bwtwen cpus. If a task migrate
>> to another cpu, the se.exec_start was set to that cpu's rq_clock_task
>> by update_stats_curr_start(). Which may not be monotonic.
>>
>> update_stats_curr_start
>>   <- set_next_entity
>>      <- set_curr_task_fair
>>         <- sched_move_task
> 
> That is not in fact a cross-cpu migration path. But I see the point.
> Also many migration paths do in fact preserve monotonicity, even when
> the clock is busted, but you're right, not all of them.
> 
>> So, if  now - p->last_task_numa_placement is -1, then (*period + 1) is
>> 0, and divide error was triggerred at the div operation:
>>   task_numa_placement:
>>     runtime = numa_get_avg_runtime(p, &period);
>>     f_weight = div64_u64(runtime << 16, period + 1);  // divide error here
>>
>> In this patch, we make sure period is not less than 0 to avoid this
>> divide error.
>>
>> Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
>> Cc: stable@vger.kernel.org
>> ---
>>  kernel/sched/fair.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index 40bd1e27b1b7..f2abb258fc85 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
>>  	if (p->last_task_numa_placement) {
>>  		delta = runtime - p->last_sum_exec_runtime;
>>  		*period = now - p->last_task_numa_placement;
>> +
>> +		/* Avoid backward, and prevent potential divide error */
>> +		if ((s64)*period < 0)
>> +			*period = 0;
>>  	} else {
>>  		delta = p->se.avg.load_sum;
>>  		*period = LOAD_AVG_MAX;
> 
> Yeah, I suppose that is indeed correct.
> 
> I'll try and come up with a better Changelog tomorrow.

Thanks.

-- 
Thanks,
Xie XiuQi

Re: [PATCH] sched: fix a potential divide error

[Peter Zijlstra]

On Thu, Apr 25, 2019 at 11:52:28AM +0800, Xie XiuQi wrote:
> On 2019/4/24 2:44, Peter Zijlstra wrote:

> > I'll try and come up with a better Changelog tomorrow.

I actually did, but forgot to send out. I have the below.
Does that work for you?

---
Subject: sched/numa: Fix a possible divide-by-zero
From: Xie XiuQi <xiexiuqi@huawei.com>
Date: Sat, 20 Apr 2019 16:34:16 +0800

sched_clock_cpu() may not be consistent between CPUs. If a task
migrates to another CPU, then se.exec_start is set to that CPU's
rq_clock_task() by update_stats_curr_start(). Specifically, the new
value might be before the old value due to clock skew.

So then if in numa_get_avg_runtime() the expression:

  'now - p->last_task_numa_placement'

ends up as -1, then the divider '*period + 1' in task_numa_placement()
is 0 and things go bang. Similar to update_curr(), check if time goes
backwards to avoid this.

Cc: mingo@redhat.com
Cc: cj.chengjian@huawei.com
Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
[peterz: simplified changelog]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20190420083416.170446-1-xiexiuqi@huawei.com
---
 kernel/sched/fair.c |    4 ++++
 1 file changed, 4 insertions(+)

--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct t
 	if (p->last_task_numa_placement) {
 		delta = runtime - p->last_sum_exec_runtime;
 		*period = now - p->last_task_numa_placement;
+
+		/* Avoid backward, and prevent potential divide error */
+		if (unlikely((s64)*period < 0))
+			*period = 0;
 	} else {
 		delta = p->se.avg.load_sum;
 		*period = LOAD_AVG_MAX;

Re: [PATCH] sched: fix a potential divide error

[Xie XiuQi]

Hi Peter,

On 2019/4/25 16:00, Peter Zijlstra wrote:
> On Thu, Apr 25, 2019 at 11:52:28AM +0800, Xie XiuQi wrote:
>> On 2019/4/24 2:44, Peter Zijlstra wrote:
> 
>>> I'll try and come up with a better Changelog tomorrow.
> 
> I actually did, but forgot to send out. I have the below.
> Does that work for you?

It works for me, thank you very much.

> 
> ---
> Subject: sched/numa: Fix a possible divide-by-zero
> From: Xie XiuQi <xiexiuqi@huawei.com>
> Date: Sat, 20 Apr 2019 16:34:16 +0800
> 
> sched_clock_cpu() may not be consistent between CPUs. If a task
> migrates to another CPU, then se.exec_start is set to that CPU's
> rq_clock_task() by update_stats_curr_start(). Specifically, the new
> value might be before the old value due to clock skew.
> 
> So then if in numa_get_avg_runtime() the expression:
> 
>   'now - p->last_task_numa_placement'
> 
> ends up as -1, then the divider '*period + 1' in task_numa_placement()
> is 0 and things go bang. Similar to update_curr(), check if time goes
> backwards to avoid this.
> 
> Cc: mingo@redhat.com
> Cc: cj.chengjian@huawei.com
> Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
> [peterz: simplified changelog]
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> Link: https://lkml.kernel.org/r/20190420083416.170446-1-xiexiuqi@huawei.com
> ---
>  kernel/sched/fair.c |    4 ++++
>  1 file changed, 4 insertions(+)
> 
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct t
>  	if (p->last_task_numa_placement) {
>  		delta = runtime - p->last_sum_exec_runtime;
>  		*period = now - p->last_task_numa_placement;
> +
> +		/* Avoid backward, and prevent potential divide error */
> +		if (unlikely((s64)*period < 0))
> +			*period = 0;
>  	} else {
>  		delta = p->se.avg.load_sum;
>  		*period = LOAD_AVG_MAX;
> 
> .
> 

-- 
Thanks,
Xie XiuQi

Re: [PATCH] sched: fix a potential divide error

[Ingo Molnar]

* Xie XiuQi <xiexiuqi@huawei.com> wrote:

> We meet a divide error on 3.10.0 kernel, the error message is bellow:
> 
> [499992.287996] divide error: 0000 [#1] SMP
> [499992.297796] do nothing after die!
> [499992.299108] Modules linked in: signo_catch macvlan binfmt_misc
> ip_set_hash_netport ip_set_hash_ipport vport_vxlan ipt_REJECT
> xt_statistic xt_physdev xt_nat xt_recent xt_mark xt_comment ...
> [499992.312751] CPU: 8 PID: 23352 Comm: bash Tainted: G ----V-------   3.10.0+ #1
> [499992.314308] Hardware name: OpenStack Foundation OpenStack Nova, BIOS
> rel-1.9.1-0-gb3ef39f-20170329_185309-build9a64a246a231 04/01/2014
> [499992.317411] task: ffff880033fc9700 ti: ffff8807fed60000 task.ti:ffff8807fed60000
> [499992.318967] RIP: 0010:[<ffffffff810c15c2>]  [<ffffffff810c15c2>] task_numa_fault+0x1c2/0xbb0
> [499992.320515] RSP: 0000:ffff8807fed63d38  EFLAGS: 00010246
> [499992.322018] RAX: 0000002b7efd0000 RBX: ffff880033fc9700 RCX:0000000000000003
> [499992.323563] RDX: 0000000000000000 RSI: 0000000000000400 RDI:ffffffff81a80f60
> [499992.325052] RBP: ffff8807fed63db8 R08: ffffffff81a80f68 R09:0000000000000000
> [499992.326531] R10: ffff88083ffda000 R11: 0000000000000000 R12:0000000000000424
> [499992.327987] R13: 00000000002b7efd R14: 0000000000000000 R15:ffffea001ea42a00
> [499992.329420] FS:  00007fa01a3b7740(0000) GS:ffff88103ec00000(0000)knlGS:0000000000000000
> [499992.330866] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [499992.332302] CR2: 0000000000ff1fb0 CR3: 00000007ff1d1000 CR4:00000000003407e0
> [499992.333763] DR0: 0000000000000000 DR1: 0000000000000000 DR2:0000000000000000
> [499992.335187] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:0000000000000400
> [499992.336595] Stack:
> [499992.337974]  0000000000000000 00000001bc9598a8 ffffea001ea42a00 0000000100000001
> [499992.339374]  0000000300000001 0000000000000001 ffffea001ea42a00 ffff8807fed63db8
> [499992.340768]  0000000000000000 0000000000000000 00000000bc9598a8 0000000000000001
> [499992.342148] Call Trace:
> [499992.343494]  [<ffffffff8119ab62>] do_numa_page+0x162/0x1f0
> [499992.344831]  [<ffffffff8119bde7>] handle_mm_fault+0x627/0xf50
> [499992.346145]  [<ffffffff8164e486>] __do_page_fault+0x166/0x470
> [499992.347442]  [<ffffffff8164e853>] trace_do_page_fault+0x43/0x110
> [499992.348711]  [<ffffffff8164df29>] do_async_page_fault+0x29/0xe0
> [499992.349948]  [<ffffffff8164a9f8>] async_page_fault+0x28/0x30
> [499992.351149] Code: 00 3d 00 04 00 00 44 0f 4e d8 41 81 fb 00 04 00 00
> 0f 84 67 07 00 00 4c 89 e8 49 83 c6 01 31 d2 48 c1 e0 10 49 83 c4 01 45
> 31 c9 <49> f7 f6 48 c7 45 a8 00 00 00 00 48 c7 45 b0 00 00 00 00 49 89
> [499992.353707] RIP  [<ffffffff810c15c2>] task_numa_fault+0x1c2/0xbb0
> [499992.354927]  RSP <ffff8807fed63d38>
> [499992.358114] ---[ end trace 4f2465cac18ff65e ]---
> [499992.359304] Kernel panic - not syncing: Fatal exception
> 
> sched_clock_cpu may not be consistent bwtwen cpus. If a task migrate
> to another cpu, the se.exec_start was set to that cpu's rq_clock_task
> by update_stats_curr_start(). Which may not be monotonic.
> 
> update_stats_curr_start
>   <- set_next_entity
>      <- set_curr_task_fair
>         <- sched_move_task
> 
> So, if  now - p->last_task_numa_placement is -1, then (*period + 1) is
> 0, and divide error was triggerred at the div operation:
>   task_numa_placement:
>     runtime = numa_get_avg_runtime(p, &period);
>     f_weight = div64_u64(runtime << 16, period + 1);  // divide error here
> 
> In this patch, we make sure period is not less than 0 to avoid this
> divide error.
> 
> Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
> Cc: stable@vger.kernel.org
> ---
>  kernel/sched/fair.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 40bd1e27b1b7..f2abb258fc85 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
>  	if (p->last_task_numa_placement) {
>  		delta = runtime - p->last_sum_exec_runtime;
>  		*period = now - p->last_task_numa_placement;
> +
> +		/* Avoid backward, and prevent potential divide error */
> +		if ((s64)*period < 0)
> +			*period = 0;
>  	} else {
>  		delta = p->se.avg.load_sum;
>  		*period = LOAD_AVG_MAX;

I've put this into sched/urgent, because this can be triggered in the 
wild it appears.

Thanks,

	Ingo

[tip:sched/urgent] sched/numa: Fix a possible divide-by-zero

[tip-bot for Xie XiuQi]

Commit-ID:  993efef2483bde15fd33ec5fba5464597c2d8124
Gitweb:     https://git.kernel.org/tip/993efef2483bde15fd33ec5fba5464597c2d8124
Author:     Xie XiuQi <xiexiuqi@huawei.com>
AuthorDate: Sat, 20 Apr 2019 16:34:16 +0800
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 25 Apr 2019 19:58:35 +0200

sched/numa: Fix a possible divide-by-zero

sched_clock_cpu() may not be consistent between CPUs. If a task
migrates to another CPU, then se.exec_start is set to that CPU's
rq_clock_task() by update_stats_curr_start(). Specifically, the new
value might be before the old value due to clock skew.

So then if in numa_get_avg_runtime() the expression:

  'now - p->last_task_numa_placement'

ends up as -1, then the divider '*period + 1' in task_numa_placement()
is 0 and things go bang. Similar to update_curr(), check if time goes
backwards to avoid this.

[ peterz: Wrote new changelog. ]
[ mingo: Tweaked the code comment. ]

Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: cj.chengjian@huawei.com
Link: http://lkml.kernel.org/r/20190425080016.GX11158@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/fair.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a4d9e14bf138..35f3ea375084 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
 	if (p->last_task_numa_placement) {
 		delta = runtime - p->last_sum_exec_runtime;
 		*period = now - p->last_task_numa_placement;
+
+		/* Avoid time going backwards, prevent potential divide error: */
+		if (unlikely((s64)*period < 0))
+			*period = 0;
 	} else {
 		delta = p->se.avg.load_sum;
 		*period = LOAD_AVG_MAX;

[tip:sched/urgent] sched/numa: Fix a possible divide-by-zero

[tip-bot for Xie XiuQi]

Commit-ID:  a860fa7b96e1a1c974556327aa1aee852d434c21
Gitweb:     https://git.kernel.org/tip/a860fa7b96e1a1c974556327aa1aee852d434c21
Author:     Xie XiuQi <xiexiuqi@huawei.com>
AuthorDate: Sat, 20 Apr 2019 16:34:16 +0800
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 25 Apr 2019 19:58:54 +0200

sched/numa: Fix a possible divide-by-zero

sched_clock_cpu() may not be consistent between CPUs. If a task
migrates to another CPU, then se.exec_start is set to that CPU's
rq_clock_task() by update_stats_curr_start(). Specifically, the new
value might be before the old value due to clock skew.

So then if in numa_get_avg_runtime() the expression:

  'now - p->last_task_numa_placement'

ends up as -1, then the divider '*period + 1' in task_numa_placement()
is 0 and things go bang. Similar to update_curr(), check if time goes
backwards to avoid this.

[ peterz: Wrote new changelog. ]
[ mingo: Tweaked the code comment. ]

Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: cj.chengjian@huawei.com
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20190425080016.GX11158@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/fair.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a4d9e14bf138..35f3ea375084 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2007,6 +2007,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
 	if (p->last_task_numa_placement) {
 		delta = runtime - p->last_sum_exec_runtime;
 		*period = now - p->last_task_numa_placement;
+
+		/* Avoid time going backwards, prevent potential divide error: */
+		if (unlikely((s64)*period < 0))
+			*period = 0;
 	} else {
 		delta = p->se.avg.load_sum;
 		*period = LOAD_AVG_MAX;