Skip to contentSkip to conversation

sched: fair: NULL ptr deref in check_preempt_wakeup

DISCUSSION Sasha Levin 3 participants 13 messages Feb 15, 2014 – Feb 20, 2014 Scheduler

lore
Share & tools
lore
#0 lore raw

Hi folks,

While fuzzing with trinity inside a KVM tools guest running latest -next kernel, I've stumbled on the following:

[  522.645288] BUG: unable to handle kernel NULL pointer dereference at 0000000000000150
[  522.646271] IP: [<ffffffff81186c6f>] check_preempt_wakeup+0x11f/0x210
[  522.646976] PGD b0a79067 PUD ae9cf067 PMD 0
[  522.647494] Oops: 0000 [#1] PREEMPT SMP
[  522.648000] Dumping ftrace buffer:
[  522.648380]    (ftrace buffer empty)
[  522.648775] Modules linked in:
[  522.649125] CPU: 0 PID: 11735 Comm: trinity-c50 Not tainted 3.14.0-rc2-next-20140214-sasha-00008-g95d9d16-dirty #85
[  522.650021] task: ffff8800c00bb000 ti: ffff88007fdb8000 task.ti: ffff88007fdb8000
[  522.650021] RIP: 0010:[<ffffffff81186c6f>]  [<ffffffff81186c6f>] check_preempt_wakeup+0x11f/0x210
[  522.650021] RSP: 0018:ffff880226e03ba8  EFLAGS: 00010046
[  522.650021] RAX: 0000000000000000 RBX: ffff880226fd79c0 RCX: 0000000000000008
[  522.650021] RDX: 0000000000000000 RSI: ffff880211313000 RDI: 000000000000000c
[  522.650021] RBP: ffff880226e03be8 R08: 0000000000000000 R09: 000000000000b4bb
[  522.650021] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
[  522.650021] R13: ffff880211313068 R14: ffff8800c00bb000 R15: 0000000000000000
[  522.650021] FS:  00007f435269f700(0000) GS:ffff880226e00000(0000) knlGS:0000000000000000
[  522.650021] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[  522.650021] CR2: 0000000000000150 CR3: 00000000abd2c000 CR4: 00000000000006f0
[  522.650021] DR0: 0000000000995750 DR1: 0000000000000000 DR2: 0000000000000000
[  522.650021] DR3: 0000000000000000 DR6: 00000000ffff4ff0 DR7: 0000000000000600
[  522.650021] Stack:
[  522.650021]  ffff880211313000 01ff880226fd79c0 ffff880211313000 ffff880226fd79c0
[  522.650021]  ffff880226fd79c0 ffff880211313000 0000000000000000 ffff880226e00000
[  522.650021]  ffff880226e03c08 ffffffff8117361d ffff880226fd79c0 ffff880226fd79c0
[  522.650021] Call Trace:
[  522.650021]  <IRQ>
[  522.650021]  [<ffffffff8117361d>] check_preempt_curr+0x3d/0xb0
[  522.650021]  [<ffffffff81175d88>] ttwu_do_wakeup+0x18/0x130
[  522.650021]  [<ffffffff81175ee4>] T.2248+0x44/0x50
[  522.650021]  [<ffffffff81175f9e>] ttwu_queue+0xae/0xd0
[  522.650021]  [<ffffffff81180224>] ? try_to_wake_up+0x34/0x2a0
[  522.650021]  [<ffffffff81180454>] try_to_wake_up+0x264/0x2a0
[  522.650021]  [<ffffffff811a1672>] ? __lock_acquired+0x2a2/0x2e0
[  522.650021]  [<ffffffff8118049d>] default_wake_function+0xd/0x10
[  522.650021]  [<ffffffff811952f8>] autoremove_wake_function+0x18/0x40
[  522.650021]  [<ffffffff811951b2>] __wake_up_common+0x52/0x90
[  522.650021]  [<ffffffff8119550d>] ? __wake_up+0x2d/0x70
[  522.650021]  [<ffffffff81195523>] __wake_up+0x43/0x70
[  522.650021]  [<ffffffff843119a3>] p9_client_cb+0x43/0x70
[  522.650021]  [<ffffffff84319d05>] req_done+0x105/0x110
[  522.650021]  [<ffffffff81cafca6>] vring_interrupt+0x86/0xa0
[  522.650021]  [<ffffffff811b9a28>] ? handle_irq_event+0x38/0x70
[  522.650021]  [<ffffffff811b9779>] handle_irq_event_percpu+0x129/0x3a0
[  522.650021]  [<ffffffff811b9a33>] handle_irq_event+0x43/0x70
[  522.650021]  [<ffffffff811bd1e8>] handle_edge_irq+0xe8/0x120
[  522.650021]  [<ffffffff81070a34>] handle_irq+0x164/0x180
[  522.650021]  [<ffffffff811833c9>] ? vtime_account_system+0x79/0x90
[  522.650021]  [<ffffffff81183435>] ? vtime_common_account_irq_enter+0x55/0x60
[  522.650021]  [<ffffffff8106f629>] do_IRQ+0x59/0x100
[  522.650021]  [<ffffffff84395e72>] common_interrupt+0x72/0x72
[  522.650021]  <EOI>
[  522.650021]  [<ffffffff812510d5>] ? context_tracking_user_exit+0x1a5/0x1c0
[  522.650021]  [<ffffffff8107cfdd>] syscall_trace_enter+0x2d/0x280
[  522.650021]  [<ffffffff8439f081>] tracesys+0x7e/0xe2
[  522.650021] Code: 0f 1f 40 00 ff c8 4d 8b ad 48 01 00 00 39 d0 7f f3 eb 18 66 0f 1f 84 00 00 00 00 00 4d 8b a4 24 48 01 00 00 4d 8b ad 48 01 00 00 <49> 8b bc 24 50 01 00 00 49 3b bd 50 01 00 00 75 e0 48 85 ff 74
[  522.650021] RIP  [<ffffffff81186c6f>] check_preempt_wakeup+0x11f/0x210
[  522.650021]  RSP <ffff880226e03ba8>
[  522.650021] CR2: 0000000000000150
[  522.650021] ---[ end trace adce75aec8b1b32f ]---

Since it's pretty inlined, the code points to:

	check_preempt_wakeup()
		find_matching_se()
			find_matching_se()
				check_preempt_wakeup()


	static inline struct cfs_rq *
	is_same_group(struct sched_entity *se, struct sched_entity *pse)
	{
	        if (se->cfs_rq == pse->cfs_rq)	<=== HERE
	                return se->cfs_rq;
	
	        return NULL;
	}

Thanks,
Sasha

#1 lore raw
Quoted 4 lines · Sasha Levin

Hi folks,

While fuzzing with trinity inside a KVM tools guest running latest -next kernel, I've stumbled on the following:

As soon as I've finished writing that mail I've hit it again, with a different (but similar) stack trace.

[  770.993016] BUG: unable to handle kernel NULL pointer dereference at 0000000000000150
[  770.993865] IP: [<ffffffff8118ef99>] pick_next_task_fair+0x109/0x290
[  770.994531] PGD 1addee067 PUD 1addef067 PMD 0
[  770.995018] Oops: 0000 [#1] PREEMPT SMP
[  770.995573] Dumping ftrace buffer:
[  770.995928]    (ftrace buffer empty)
[  770.996304] Modules linked in:
[  770.996661] CPU: 0 PID: 13754 Comm: trinity-c155 Not tainted 3.14.0-rc2-next-20140214
[  770.997646] task: ffff88021151b000 ti: ffff88016b9f4000 task.ti: ffff88016b9f4000
[  770.998384] RIP: 0010:[<ffffffff8118ef99>]  [<ffffffff8118ef99>] pick_next_task_fair+
[  770.999254] RSP: 0018:ffff88016b9f5bc8  EFLAGS: 00010097
[  770.999787] RAX: 000000004caed01b RBX: ffff880226fd79c0 RCX: 000000000004ccca
[  771.000035] RDX: 0000000000a7076b RSI: ffff880060ff8028 RDI: ffff88008b998078
[  771.000035] RBP: ffff88016b9f5c08 R08: 0000000000000000 R09: 0000000000000001
[  771.000035] R10: 0000000000000000 R11: 0000000000000000 R12: ffff88008b998000
[  771.000035] R13: 0000000000000000 R14: ffff880226fd7a88 R15: ffff880060ffb7c8
[  771.000035] FS:  00007f6e01002700(0000) GS:ffff880226e00000(0000) knlGS:0000000000000
[  771.000035] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  771.000035] CR2: 0000000000000150 CR3: 00000001feeef000 CR4: 00000000000006f0
[  771.000035] DR0: 00007f6e009b2000 DR1: 0000000000000000 DR2: 0000000000000000
[  771.000035] DR3: 0000000000000000 DR6: 00000000ffff4ff0 DR7: 0000000000000600
[  771.000035] Stack:
[  771.000035]  ffff880100000001 ffffffff00000001 ffff88016b9f5c08 ffff880226fd79c0
[  771.000035]  0000000000000000 ffff88021151b990 0000000000000282 00000000ffffffff
[  771.000035]  ffff88016b9f5c88 ffffffff8438ef35 ffff88016b9f5c78 ffffffff811a19a5
[  771.000035] Call Trace:
[  771.000035]  [<ffffffff8438ef35>] __schedule+0x2a5/0x840
[  771.000035]  [<ffffffff811a19a5>] ? __lock_contended+0x205/0x240
[  771.000035]  [<ffffffff8438f795>] schedule+0x65/0x70
[  771.000035]  [<ffffffff8438fb73>] schedule_preempt_disabled+0x13/0x20
[  771.000035]  [<ffffffff8439105d>] mutex_lock_nested+0x2ad/0x510
[  771.000035]  [<ffffffff812ed326>] ? lookup_slow+0x46/0xd0
[  771.000035]  [<ffffffff812ed70d>] ? unlazy_walk+0x16d/0x1e0
[  771.000035]  [<ffffffff812ed326>] ? lookup_slow+0x46/0xd0
[  771.000035]  [<ffffffff812ed326>] lookup_slow+0x46/0xd0
[  771.000035]  [<ffffffff812efbe5>] path_lookupat+0xe5/0x660
[  771.000035]  [<ffffffff812b97ea>] ? kmem_cache_alloc+0x1fa/0x300
[  771.000035]  [<ffffffff812eb497>] ? getname_flags+0x57/0x1c0
[  771.000035]  [<ffffffff812f018f>] filename_lookup+0x2f/0xd0
[  771.000035]  [<ffffffff812f155c>] user_path_at_empty+0x6c/0xb0
[  771.000035]  [<ffffffff812510b5>] ? context_tracking_user_exit+0x185/0x1c0
[  771.000035]  [<ffffffff811a3ccd>] ? trace_hardirqs_on+0xd/0x10
[  771.000035]  [<ffffffff812f15ac>] user_path_at+0xc/0x10
[  771.000035]  [<ffffffff812de913>] do_sys_truncate+0x43/0xc0
[  771.000035]  [<ffffffff812de9a9>] SyS_truncate+0x9/0x10
[  771.000035]  [<ffffffff8439f0e0>] tracesys+0xdd/0xe2
[  771.000035] Code: 4d 8b ad 48 01 00 00 39 c2 7c 19 4d 8b b7 50 01 00 00 4c 89 fe 4c 89 f7 e8 55 98 ff ff 4d 8b bf 48 01 00 00 4d 8b b7 50 01 00 00 <49> 8b bd 50 01 00 00 49 39 fe 75 a3 4d 85 f6 74 9e 4c 89 ee 4c
[  771.000035] RIP  [<ffffffff8118ef99>] pick_next_task_fair+0x109/0x290
[  771.000035]  RSP <ffff88016b9f5bc8>
[  771.000035] CR2: 0000000000000150
[  771.000035] ---[ end trace 408e14968ec7dd7a ]---

Thanks,
Sasha

maintainer
#2 lore raw
In reply to Sasha Levin
Quoted 10 lines · Sasha Levin

Hi folks,

While fuzzing with trinity inside a KVM tools guest running latest -next kernel, I've stumbled on the following:

[  522.645288] BUG: unable to handle kernel NULL pointer dereference at 0000000000000150
[  522.646271] IP: [<ffffffff81186c6f>] check_preempt_wakeup+0x11f/0x210

Since it's pretty inlined, the code points to:

Quoted 14 lines
	check_preempt_wakeup()
		find_matching_se()
			find_matching_se()
				check_preempt_wakeup()


	static inline struct cfs_rq *
	is_same_group(struct sched_entity *se, struct sched_entity *pse)
	{
	        if (se->cfs_rq == pse->cfs_rq)	<=== HERE
	                return se->cfs_rq;
	
	        return NULL;
	}

Hrm.. that means we got se->depth wrong. I'll have a poke tomorrow.

#3 lore raw
In reply to Sasha Levin

Hi, Sasha

Quoted 5 lines · Sasha Levin

Hi folks,

While fuzzing with trinity inside a KVM tools guest running latest -next kernel, I've
stumbled on the following:

I've reproduced the same issue with tip/master, and below patch fixed the problem on my box along with some rcu stall info disappeared, would you like to have a try?

BTW, I reproduced it by steps:
1. change current to RT
2. move to a different depth cpu-cgroup
3. change it back to FAIR

Seems like it was caused by that RT has no task_move_group() implemented which could maintain depth, and that lead to a wrong depth after switched back to FAIR...

Regards,
Michael Wang

kernel/sched/fair.c +5 -1
index 235cfa7..4445e56 100644
@@ -7317,7 +7317,11 @@ static void switched_from_fair(struct rq *rq, struct task_struct *p)
  */
 static void switched_to_fair(struct rq *rq, struct task_struct *p)
 {
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
 
 	/*
Quoted 108 lines
[  522.645288] BUG: unable to handle kernel NULL pointer dereference at
0000000000000150
[  522.646271] IP: [<ffffffff81186c6f>] check_preempt_wakeup+0x11f/0x210
[  522.646976] PGD b0a79067 PUD ae9cf067 PMD 0
[  522.647494] Oops: 0000 [#1] PREEMPT SMP
[  522.648000] Dumping ftrace buffer:
[  522.648380]    (ftrace buffer empty)
[  522.648775] Modules linked in:
[  522.649125] CPU: 0 PID: 11735 Comm: trinity-c50 Not tainted
3.14.0-rc2-next-20140214-sasha-00008-g95d9d16-dirty #85
[  522.650021] task: ffff8800c00bb000 ti: ffff88007fdb8000 task.ti:
ffff88007fdb8000
[  522.650021] RIP: 0010:[<ffffffff81186c6f>]  [<ffffffff81186c6f>]
check_preempt_wakeup+0x11f/0x210
[  522.650021] RSP: 0018:ffff880226e03ba8  EFLAGS: 00010046
[  522.650021] RAX: 0000000000000000 RBX: ffff880226fd79c0 RCX:
0000000000000008
[  522.650021] RDX: 0000000000000000 RSI: ffff880211313000 RDI:
000000000000000c
[  522.650021] RBP: ffff880226e03be8 R08: 0000000000000000 R09:
000000000000b4bb
[  522.650021] R10: 0000000000000000 R11: 0000000000000000 R12:
0000000000000000
[  522.650021] R13: ffff880211313068 R14: ffff8800c00bb000 R15:
0000000000000000
[  522.650021] FS:  00007f435269f700(0000) GS:ffff880226e00000(0000)
knlGS:0000000000000000
[  522.650021] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[  522.650021] CR2: 0000000000000150 CR3: 00000000abd2c000 CR4:
00000000000006f0
[  522.650021] DR0: 0000000000995750 DR1: 0000000000000000 DR2:
0000000000000000
[  522.650021] DR3: 0000000000000000 DR6: 00000000ffff4ff0 DR7:
0000000000000600
[  522.650021] Stack:
[  522.650021]  ffff880211313000 01ff880226fd79c0 ffff880211313000
ffff880226fd79c0
[  522.650021]  ffff880226fd79c0 ffff880211313000 0000000000000000
ffff880226e00000
[  522.650021]  ffff880226e03c08 ffffffff8117361d ffff880226fd79c0
ffff880226fd79c0
[  522.650021] Call Trace:
[  522.650021]  <IRQ>
[  522.650021]  [<ffffffff8117361d>] check_preempt_curr+0x3d/0xb0
[  522.650021]  [<ffffffff81175d88>] ttwu_do_wakeup+0x18/0x130
[  522.650021]  [<ffffffff81175ee4>] T.2248+0x44/0x50
[  522.650021]  [<ffffffff81175f9e>] ttwu_queue+0xae/0xd0
[  522.650021]  [<ffffffff81180224>] ? try_to_wake_up+0x34/0x2a0
[  522.650021]  [<ffffffff81180454>] try_to_wake_up+0x264/0x2a0
[  522.650021]  [<ffffffff811a1672>] ? __lock_acquired+0x2a2/0x2e0
[  522.650021]  [<ffffffff8118049d>] default_wake_function+0xd/0x10
[  522.650021]  [<ffffffff811952f8>] autoremove_wake_function+0x18/0x40
[  522.650021]  [<ffffffff811951b2>] __wake_up_common+0x52/0x90
[  522.650021]  [<ffffffff8119550d>] ? __wake_up+0x2d/0x70
[  522.650021]  [<ffffffff81195523>] __wake_up+0x43/0x70
[  522.650021]  [<ffffffff843119a3>] p9_client_cb+0x43/0x70
[  522.650021]  [<ffffffff84319d05>] req_done+0x105/0x110
[  522.650021]  [<ffffffff81cafca6>] vring_interrupt+0x86/0xa0
[  522.650021]  [<ffffffff811b9a28>] ? handle_irq_event+0x38/0x70
[  522.650021]  [<ffffffff811b9779>] handle_irq_event_percpu+0x129/0x3a0
[  522.650021]  [<ffffffff811b9a33>] handle_irq_event+0x43/0x70
[  522.650021]  [<ffffffff811bd1e8>] handle_edge_irq+0xe8/0x120
[  522.650021]  [<ffffffff81070a34>] handle_irq+0x164/0x180
[  522.650021]  [<ffffffff811833c9>] ? vtime_account_system+0x79/0x90
[  522.650021]  [<ffffffff81183435>] ?
vtime_common_account_irq_enter+0x55/0x60
[  522.650021]  [<ffffffff8106f629>] do_IRQ+0x59/0x100
[  522.650021]  [<ffffffff84395e72>] common_interrupt+0x72/0x72
[  522.650021]  <EOI>
[  522.650021]  [<ffffffff812510d5>] ?
context_tracking_user_exit+0x1a5/0x1c0
[  522.650021]  [<ffffffff8107cfdd>] syscall_trace_enter+0x2d/0x280
[  522.650021]  [<ffffffff8439f081>] tracesys+0x7e/0xe2
[  522.650021] Code: 0f 1f 40 00 ff c8 4d 8b ad 48 01 00 00 39 d0 7f f3
eb 18 66 0f 1f 84 00 00 00 00 00 4d 8b a4 24 48 01 00 00 4d 8b ad 48 01
00 00 <49> 8b bc 24 50 01 00 00 49 3b bd 50 01 00 00 75 e0 48 85 ff 74
[  522.650021] RIP  [<ffffffff81186c6f>] check_preempt_wakeup+0x11f/0x210
[  522.650021]  RSP <ffff880226e03ba8>
[  522.650021] CR2: 0000000000000150
[  522.650021] ---[ end trace adce75aec8b1b32f ]---

Since it's pretty inlined, the code points to:

    check_preempt_wakeup()
        find_matching_se()
            find_matching_se()
                check_preempt_wakeup()


    static inline struct cfs_rq *
    is_same_group(struct sched_entity *se, struct sched_entity *pse)
    {
            if (se->cfs_rq == pse->cfs_rq)    <=== HERE
                    return se->cfs_rq;
    
            return NULL;
    }

Thanks,
Sasha

Signature · 4 lines
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
maintainer
#4 lore raw
Quoted 17 lines · Michael wang

BTW, I reproduced it by steps:
1. change current to RT
2. move to a different depth cpu-cgroup
3. change it back to FAIR

Seems like it was caused by that RT has no task_move_group() implemented which could maintain depth, and that lead to a wrong depth after switched back to FAIR...

kernel/sched/fair.c +0 -0
index 235cfa7..4445e56 100644
@@ -7317,7 +7317,11 @@ static void switched_from_fair(struct rq *rq, struct task_struct *p)
  */
 static void switched_to_fair(struct rq *rq, struct task_struct *p)

kernel/sched/fair.c static void switched_from_fair(struct rq *rq, struct task_struct *p) · line 7327

Quoted 10 lines
kernel/sched/fair.c +5 -1
 {
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
 
 	/*

Yes indeed. My first idea yesterday was to put it in set_task_rq() to be absolutely sure we catch all; but if this is sufficient its better.

Thanks!

#5 lore raw

On 02/17/2014 05:20 PM, Peter Zijlstra wrote:
[snip]

Quoted 10 lines
Quoted 10 lines

static void switched_to_fair(struct rq *rq, struct task_struct *p) {

kernel/sched/fair.c +5 -1
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
Quoted 4 lines
Quoted 1 line

/*

Yes indeed. My first idea yesterday was to put it in set_task_rq() to be absolutely sure we catch all; but if this is sufficient its better.

Agree, let's wait for Sasha's testing result then :)

Regards,
Michael Wang

Quoted 8 lines

Thanks!
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

author Tested-by
#6 lore raw
Quoted 12 lines · Michael wang

On 02/17/2014 05:20 PM, Peter Zijlstra wrote:
[snip]

Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines

static void switched_to_fair(struct rq *rq, struct task_struct *p) {

kernel/sched/fair.c +5 -1
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
Quoted 5 lines
Quoted 4 lines
Quoted 4 lines
Quoted 1 line
Quoted 1 line

/*

Yes indeed. My first idea yesterday was to put it in set_task_rq() to be absolutely sure we catch all; but if this is sufficient its better.

Agree, let's wait for Sasha's testing result then:)

I took my time with testing it seems I'm hitting new issues with both sched and mm, and I've wanted to confirm I don't see this one any more.

It does seem like this patch fixes the problem for me, so:

Thanks,
Sasha

maintainer
#7 lore raw
Quoted 17 lines · Sasha Levin
Quoted 16 lines · Michael wang

On 02/17/2014 05:20 PM, Peter Zijlstra wrote:
[snip]

Quoted 14 lines
Quoted 14 lines
Quoted 11 lines

kernel/sched/fair.c static void switched_from_fair(struct rq *rq, struct task_struct *p) · line 7327

Quoted 11 lines

static void switched_to_fair(struct rq *rq, struct task_struct *p) {

kernel/sched/fair.c +5 -1
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
 
 	/*

Yes indeed. My first idea yesterday was to put it in set_task_rq() to be absolutely sure we catch all; but if this is sufficient its better.

Quoted 9 lines
Quoted 1 line

Agree, let's wait for Sasha's testing result then:)

I took my time with testing it seems I'm hitting new issues with both sched and mm, and I've wanted to confirm I don't see this one any more.

It does seem like this patch fixes the problem for me, so:

Thanks!

#8 lore raw
In reply to Sasha Levin
Quoted 18 lines · Sasha Levin
Quoted 17 lines · Michael wang

On 02/17/2014 05:20 PM, Peter Zijlstra wrote:
[snip]

Quoted 15 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines

static void switched_to_fair(struct rq *rq, struct task_struct *p) {

kernel/sched/fair.c +5 -1
-    if (!p->se.on_rq)
+    struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+    se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+    if (!se->on_rq)
         return;
Quoted 3 lines
Quoted 1 line
Quoted 1 line
     /*

Yes indeed. My first idea yesterday was to put it in set_task_rq()

to be

Quoted 1 line

absolutely sure we catch all; but if this is sufficient its better.

Quoted 8 lines
Quoted 1 line

Agree, let's wait for Sasha's testing result then:)

I took my time with testing it seems I'm hitting new issues with both sched and mm, and I've wanted to confirm I don't see this one any more.

It does seem like this patch fixes the problem for me, so:

Thanks for the testing :) will post the patch later.

Regards,
Michael Wang

Quoted 5 lines

Thanks,
Sasha

#9 lore raw
In reply to Michael wang
Quoted 13 lines · Michael wang

Hi, Sasha

Quoted 5 lines · Sasha Levin

Hi folks,

While fuzzing with trinity inside a KVM tools guest running latest -next kernel, I've
stumbled on the following:

I've reproduced the same issue with tip/master, and below patch fixed the problem on my box along with some rcu stall info disappeared, would you like to have a try?

Quoted 8 lines

BTW, I reproduced it by steps:
1. change current to RT
2. move to a different depth cpu-cgroup
3. change it back to FAIR

Seems like it was caused by that RT has no task_move_group() implemented which could maintain depth, and that lead to a wrong depth after switched back to FAIR...

I *think* it works. There seems to be another sched issue that causes lockups, so I can't say for certain that this one doesn't occur anymore.

I'm still working on collecting data for the other issue, I'll mail about it soon.

Thanks,
Sasha

#10 lore raw

On 02/18/2014 05:07 AM, Sasha Levin wrote:
[snip]

Quoted 7 lines

I *think* it works. There seems to be another sched issue that causes lockups,
so I can't say for certain that this one doesn't occur anymore.

I'm still working on collecting data for the other issue, I'll mail about it soon.

Thanks for that, looking forward the results :)

Regards,
Michael Wang

Quoted 11 lines

Thanks,
Sasha

Signature · 4 lines
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
maintainer
#11 lore raw
In reply to Michael wang
Quoted 29 lines · Michael wang
Quoted 3 lines

While fuzzing with trinity inside a KVM tools guest running latest -next kernel, I've
stumbled on the following:

I've reproduced the same issue with tip/master, and below patch fixed the problem on my box along with some rcu stall info disappeared, would you like to have a try?

BTW, I reproduced it by steps:
1. change current to RT
2. move to a different depth cpu-cgroup
3. change it back to FAIR

Seems like it was caused by that RT has no task_move_group() implemented which could maintain depth, and that lead to a wrong depth after switched back to FAIR...

Regards,
Michael Wang

kernel/sched/fair.c +0 -0
index 235cfa7..4445e56 100644
@@ -7317,7 +7317,11 @@ static void switched_from_fair(struct rq *rq, struct task_struct *p)
  */
 static void switched_to_fair(struct rq *rq, struct task_struct *p)

kernel/sched/fair.c static void switched_from_fair(struct rq *rq, struct task_struct *p) · line 7327

Quoted 10 lines
kernel/sched/fair.c +5 -1
 {
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
 
 	/*

Michael, do you think you can send a proper patch for this?

#12 lore raw

On 02/20/2014 12:16 AM, Peter Zijlstra wrote:
[snip]

Quoted 19 lines
Quoted 19 lines
kernel/sched/fair.c +5 -1
index 235cfa7..4445e56 100644
@@ -7317,7 +7317,11 @@ static void switched_from_fair(struct rq *rq, struct task_struct *p)
  */
 static void switched_to_fair(struct rq *rq, struct task_struct *p)
 {
-	if (!p->se.on_rq)
+	struct sched_entity *se = &p->se;
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	se->depth = se->parent ? se->parent->depth + 1 : 0;
+#endif
+	if (!se->on_rq)
 		return;
 
Quoted 4 lines

kernel/sched/fair.c static void switched_from_fair(struct rq *rq, struct task_struct *p) · line 7327

Quoted 1 line
kernel/sched/fair.c +0 -0
 	/*

Michael, do you think you can send a proper patch for this?

My pleasure :) will post it later.

Regards,
Michael Wang

Quoted 6 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/