Skip to contentSkip to conversation

? xfrm: INFO: inconsistent lock state

BUG In review Wei Yongjun 6 participants 9 messages Feb 1, 2010 – Feb 3, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

I got following lock INFO in my Fedora12 box uner 2.6.33-rc6. Is this a BUG?

=================================
[ INFO: inconsistent lock state ]
2.6.33-rc6 #120
---------------------------------
inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage. swapper/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
(&(&x->lock)->rlock){?.-...}, at: [<c072d03f>] xfrm_input+0xdf/0x5c0
{IN-HARDIRQ-W} state was registered at:

  [<c04718dc>] __lock_acquire+0xa9c/0x1890
  [<c047274f>] lock_acquire+0x7f/0xf0
  [<c0762958>] _raw_spin_lock+0x38/0x50
  [<c072b5ca>] xfrm_timer_handler+0x3a/0x260
  [<c0447d9d>] __hrtimer_tasklet_trampoline+0xd/0x10
  [<c04634ce>] hrtimer_run_queues+0x15e/0x2a0
  [<c045146d>] run_local_timers+0xd/0x20
  [<c04514b4>] update_process_times+0x34/0x70
  [<c046ce8a>] tick_periodic+0x2a/0x80
  [<c046cefe>] tick_handle_periodic+0x1e/0x90
  [<c0768377>] smp_apic_timer_interrupt+0x57/0x8b
  [<c076382f>] apic_timer_interrupt+0x2f/0x34
  [<c0401d3b>] cpu_idle+0x4b/0x80
  [<c074e0d7>] rest_init+0x67/0x70
  [<c0956874>] start_kernel+0x30e/0x314
  [<c095609e>] i386_start_kernel+0x9e/0xa5
irq event stamp: 1113374
hardirqs last  enabled at (1113374): [<c0449288>] local_bh_enable_ip+0x68/0xd0
hardirqs last disabled at (1113373): [<c0449248>] local_bh_enable_ip+0x28/0xd0
softirqs last  enabled at (1113338): [<c0449570>] __do_softirq+0x100/0x1c0
softirqs last disabled at (1113343): [<c0404e17>] do_softirq+0x97/0xe0

other info that might help us debug this:
3 locks held by swapper/0:

 #0:  (rcu_read_lock){.+.+..}, at: [<c06b8426>] net_rx_action+0xd6/0x230
 #1:  (rcu_read_lock){.+.+..}, at: [<c06b77c8>] netif_receive_skb+0x138/0x570
 #2:  (rcu_read_lock){.+.+..}, at: [<d1c5d540>] ip6_input_finish+0x0/0x390 [ipv6]
stack backtrace:
Pid: 0, comm: swapper Not tainted 2.6.33-rc6 #120
Call Trace:
 [<c075f984>] ? printk+0x1d/0x21
 [<c046fb8f>] print_usage_bug+0x15f/0x1a0
 [<c046ff8a>] mark_lock+0x3ba/0x5c0
 [<c04701f2>] ? mark_held_locks+0x62/0x90
 [<c0473540>] ? check_usage_backwards+0x0/0xd0
 [<c04717ef>] __lock_acquire+0x9af/0x1890
 [<c05a0cac>] ? trace_hardirqs_on_thunk+0xc/0x10
 [<c07633fc>] ? restore_all_notrace+0x0/0x18
 [<c04701f2>] ? mark_held_locks+0x62/0x90
 [<c047274f>] lock_acquire+0x7f/0xf0
 [<c072d03f>] ? xfrm_input+0xdf/0x5c0
 [<c0762958>] _raw_spin_lock+0x38/0x50
 [<c072d03f>] ? xfrm_input+0xdf/0x5c0
 [<c072d03f>] xfrm_input+0xdf/0x5c0
 [<d1c88929>] xfrm6_rcv_spi+0x29/0x30 [ipv6]
 [<d1c8894c>] xfrm6_rcv+0x1c/0x20 [ipv6]
 [<d1c5d681>] ip6_input_finish+0x141/0x390 [ipv6]
 [<d1c5d540>] ? ip6_input_finish+0x0/0x390 [ipv6]
 [<d1c5d92f>] ip6_input+0x5f/0x70 [ipv6]
 [<d1c5de78>] ipv6_rcv+0x338/0x440 [ipv6]
 [<d1c5db6e>] ? ipv6_rcv+0x2e/0x440 [ipv6]
 [<c06b79df>] netif_receive_skb+0x34f/0x570
 [<c06b77c8>] ? netif_receive_skb+0x138/0x570
 [<c06affde>] ? __alloc_skb+0x4e/0x130
 [<c04080b0>] ? nommu_sync_single_for_device+0x0/0x10
 [<d087ec5e>] pcnet32_poll+0x27e/0x7a0 [pcnet32]
 [<c0450797>] ? run_timer_softirq+0x37/0x2c0
 [<c0763067>] ? _raw_spin_unlock_irq+0x27/0x30
 [<c06b84a0>] net_rx_action+0x150/0x230
 [<c06b8426>] ? net_rx_action+0xd6/0x230
 [<c04494c8>] ? __do_softirq+0x58/0x1c0
 [<c0449510>] __do_softirq+0xa0/0x1c0
 [<c0449470>] ? __do_softirq+0x0/0x1c0
 <IRQ>  [<c0448ce5>] ? irq_exit+0x45/0x50
 [<c076837c>] ? smp_apic_timer_interrupt+0x5c/0x8b
 [<c05a0cbc>] ? trace_hardirqs_off_thunk+0xc/0x10
 [<c076382f>] ? apic_timer_interrupt+0x2f/0x34
 [<c0409de3>] ? default_idle+0x43/0xa0
 [<c0401d3b>] ? cpu_idle+0x4b/0x80
 [<c074e0d7>] ? rest_init+0x67/0x70
 [<c0956874>] ? start_kernel+0x30e/0x314
 [<c0956390>] ? unknown_bootoption+0x0/0x1a0
 [<c095609e>] ? i386_start_kernel+0x9e/0xa5
#1 lore raw
Quoted 2 lines · Wei Yongjun

I got following lock INFO in my Fedora12 box uner 2.6.33-rc6. Is this a BUG?

Apparently.

This is introduced by

commit 9e0d57fd6dad37d72a3ca6db00ca8c76f2215454

Author: Yury Polyanskiy <[email protected]>
Date:   Sun Nov 8 20:58:41 2009 -0800

xfrm: SAD entries do not expire correctly after suspend-resume

To be honest I never liked that patch since the semantic change makes sense for suspend-resume but is totally the wrong thing to do when the system clock is stepped/changed.

Whether this is a real dead-lock depends on whether hrtimers behave as real hard IRQs. If they do then yes it can deadlock.

So my suggestion would be to just revert this change and consider an alternative solution to the original suspend-resume problem.

My preference would be something that is triggered through power- management as that would then allows us to differentiate between the suspend-resume case and the clock change case.

It could even be triggered from user-space as it knows when an resume occurs.

Cheers,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#2 lore raw
Quoted 8 lines · On Tue, 2 Feb 2010 18:49:14 +1100 Herbert Xu <[email protected]>…
Quoted 2 lines · Wei Yongjun

I got following lock INFO in my Fedora12 box uner 2.6.33-rc6. Is this a BUG?

Apparently.

This is introduced by

Quoted 8 lines

commit 9e0d57fd6dad37d72a3ca6db00ca8c76f2215454

Author: Yury Polyanskiy <[email protected]>
Date:   Sun Nov 8 20:58:41 2009 -0800

xfrm: SAD entries do not expire correctly after suspend-resume

Whether this is a real dead-lock depends on whether hrtimers behave as real hard IRQs. If they do then yes it can deadlock.

If hrtimer_tasklet interface functions properly, the xfrm_timer_handler should be called in softirq context (and thus is never in parallel with xfrm_input()). The deadlock isn't possible then.

In this case it seems that for some reason xfrm_timer_handler() is called in the hardirq context. The relevant code in hrtimer_tasklet:

static enum hrtimer_restart __hrtimer_tasklet_trampoline(struct hrtimer *timer)

{
	struct tasklet_hrtimer *ttimer =
		container_of(timer, struct tasklet_hrtimer, timer);

	if (hrtimer_is_hres_active(timer)) {
		tasklet_hi_schedule(&ttimer->tasklet);
		return HRTIMER_NORESTART;
	}
	return ttimer->function(timer);
}

I am copying Peter on this. Peter, how is it possible that ttimer->function() is called in hardirq?

Could it be that switch from hres_active happened after the call to trampoline and before the if() above?

Best,
Yury

#3 lore raw

PATCH 1/1 hrtimer, softirq: Fix hrtimer->softirq trampoline

Quoted 19 lines · Yury Polyanskiy

If hrtimer_tasklet interface functions properly, the xfrm_timer_handler should be called in softirq context (and thus is never in parallel with xfrm_input()). The deadlock isn't possible then.

In this case it seems that for some reason xfrm_timer_handler() is called in the hardirq context. The relevant code in hrtimer_tasklet:

static enum hrtimer_restart __hrtimer_tasklet_trampoline(struct hrtimer *timer)

{
	struct tasklet_hrtimer *ttimer =
		container_of(timer, struct tasklet_hrtimer, timer);

	if (hrtimer_is_hres_active(timer)) {
		tasklet_hi_schedule(&ttimer->tasklet);
		return HRTIMER_NORESTART;
	}
	return ttimer->function(timer);
}
Quoted 5 lines

I am copying Peter on this. Peter, how is it possible that ttimer->function() is called in hardirq?

Could it be that switch from hres_active happened after the call to trampoline and before the if() above?

The original email had more information:

Quoted 17 lines

{IN-HARDIRQ-W} state was registered at:

  [<c04718dc>] __lock_acquire+0xa9c/0x1890
  [<c047274f>] lock_acquire+0x7f/0xf0
  [<c0762958>] _raw_spin_lock+0x38/0x50
  [<c072b5ca>] xfrm_timer_handler+0x3a/0x260
  [<c0447d9d>] __hrtimer_tasklet_trampoline+0xd/0x10
  [<c04634ce>] hrtimer_run_queues+0x15e/0x2a0
  [<c045146d>] run_local_timers+0xd/0x20
  [<c04514b4>] update_process_times+0x34/0x70
  [<c046ce8a>] tick_periodic+0x2a/0x80
  [<c046cefe>] tick_handle_periodic+0x1e/0x90
  [<c0768377>] smp_apic_timer_interrupt+0x57/0x8b
  [<c076382f>] apic_timer_interrupt+0x2f/0x34
  [<c0401d3b>] cpu_idle+0x4b/0x80
  [<c074e0d7>] rest_init+0x67/0x70
  [<c0956874>] start_kernel+0x30e/0x314
  [<c095609e>] i386_start_kernel+0x9e/0xa5

Which indicates we were called from hardirq context, it appears that that hrtimer_is_hres_active() case is indeed faulty. Not sure if I made a mistake when I wrote that or if we changed hrtimer behaviour afterwards, but the hrtimer fallback is still from hardirq context.

Which would seem to suggest the following patch:

---
Subject: hrtimer, softirq: Fix hrtimer->softirq trampoline

hrtimers callbacks are always done from hardirq context, either the jiffy tick interrupt or the hrtimer device interrupt.

---
 kernel/softirq.c |   13 +++----------
 1 files changed, 3 insertions(+), 10 deletions(-)
kernel/softirq.c +3 -10
index a09502e..c1983b7 100644
@@ -500,22 +500,15 @@ EXPORT_SYMBOL(tasklet_kill);
  */
 
 /*
- * The trampoline is called when the hrtimer expires. If this is
- * called from the hrtimer interrupt then we schedule the tasklet as
- * the timer callback function expects to run in softirq context. If
- * it's called in softirq context anyway (i.e. high resolution timers
- * disabled) then the hrtimer callback is called right away.
+ * The trampoline is called when the hrtimer expires. 
  */
 static enum hrtimer_restart __hrtimer_tasklet_trampoline(struct hrtimer *timer)
 {
 	struct tasklet_hrtimer *ttimer =
 		container_of(timer, struct tasklet_hrtimer, timer);
 
-	if (hrtimer_is_hres_active(timer)) {
-		tasklet_hi_schedule(&ttimer->tasklet);
-		return HRTIMER_NORESTART;
-	}
-	return ttimer->function(timer);
+	tasklet_hi_schedule(&ttimer->tasklet);
+	return HRTIMER_NORESTART;
 }
 
 /*
#4 lore raw

Re: [PATCH] hrtimer, softirq: Fix hrtimer->softirq trampoline

Quoted 4 lines · On Tue, 02 Feb 2010 15:20:01 +0100 Peter Zijlstra <[email protected]> wrote:

static enum hrtimer_restart __hrtimer_tasklet_trampoline(struct hrtimer *timer) {

kernel/softirq.c +0 -0
 	struct tasklet_hrtimer *ttimer =

kernel/softirq.c EXPORT_SYMBOL(tasklet_kill); · line 512

Quoted 10 lines
kernel/softirq.c +2 -5
 		container_of(timer, struct tasklet_hrtimer, timer);
 
-	if (hrtimer_is_hres_active(timer)) {
-		tasklet_hi_schedule(&ttimer->tasklet);
-		return HRTIMER_NORESTART;
-	}
-	return ttimer->function(timer);
+	tasklet_hi_schedule(&ttimer->tasklet);
+	return HRTIMER_NORESTART;
 }

Are you totally against if(in_irq())?

Yury

#5 lore raw

Re: [PATCH] hrtimer, softirq: Fix hrtimer->softirq trampoline

Quoted 7 lines · Yury Polyanskiy
Quoted 5 lines · On Tue, 02 Feb 2010 15:20:01 +0100 Peter Zijlstra <[email protected]> wrote:

static enum hrtimer_restart __hrtimer_tasklet_trampoline(struct hrtimer *timer) {

kernel/softirq.c +0 -0
 	struct tasklet_hrtimer *ttimer =
 		container_of(timer, struct tasklet_hrtimer, timer);
Quoted 11 lines

kernel/softirq.c EXPORT_SYMBOL(tasklet_kill); · line 512

Quoted 9 lines
kernel/softirq.c +2 -5
-	if (hrtimer_is_hres_active(timer)) {
-		tasklet_hi_schedule(&ttimer->tasklet);
-		return HRTIMER_NORESTART;
-	}
-	return ttimer->function(timer);
+	tasklet_hi_schedule(&ttimer->tasklet);
+	return HRTIMER_NORESTART;
 }

Are you totally against if(in_irq())?

Yeah, things like that are an indication that you really don't know wtf you're doing and are just patching up.

There is a single site where hrtimer callbacks can indeed be done from softirq, but in that case the above still works correctly, and I've been meaning to get rid of that anyway.

#6 lore raw

Re: [PATCH] hrtimer, softirq: Fix hrtimer->softirq trampoline

In reply to Peter Zijlstra
Quoted 31 lines · Peter Zijlstra

On Tue, 2010-02-02 at 08:51 -0500, Yury Polyanskiy wrote:

The original email had more information:

Quoted 18 lines

{IN-HARDIRQ-W} state was registered at:

  [<c04718dc>] __lock_acquire+0xa9c/0x1890
  [<c047274f>] lock_acquire+0x7f/0xf0
  [<c0762958>] _raw_spin_lock+0x38/0x50
  [<c072b5ca>] xfrm_timer_handler+0x3a/0x260
  [<c0447d9d>] __hrtimer_tasklet_trampoline+0xd/0x10
  [<c04634ce>] hrtimer_run_queues+0x15e/0x2a0
  [<c045146d>] run_local_timers+0xd/0x20
  [<c04514b4>] update_process_times+0x34/0x70
  [<c046ce8a>] tick_periodic+0x2a/0x80
  [<c046cefe>] tick_handle_periodic+0x1e/0x90
  [<c0768377>] smp_apic_timer_interrupt+0x57/0x8b
  [<c076382f>] apic_timer_interrupt+0x2f/0x34
  [<c0401d3b>] cpu_idle+0x4b/0x80
  [<c074e0d7>] rest_init+0x67/0x70
  [<c0956874>] start_kernel+0x30e/0x314
  [<c095609e>] i386_start_kernel+0x9e/0xa5

Which indicates we were called from hardirq context, it appears that that hrtimer_is_hres_active() case is indeed faulty. Not sure if I made a mistake when I wrote that or if we changed hrtimer behaviour afterwards, but the hrtimer fallback is still from hardirq context.

Which would seem to suggest the following patch:

Quoted 9 lines

---
Subject: hrtimer, softirq: Fix hrtimer->softirq trampoline

hrtimers callbacks are always done from hardirq context, either the jiffy tick interrupt or the hrtimer device interrupt.

With this patch, the inconsistent lock state INFO is gone. Thanks.

Wei Yongjun

maintainer Acked-by on v1 1/1
#7 lore raw

Re: [PATCH] hrtimer, softirq: Fix hrtimer->softirq trampoline

In reply to Peter Zijlstra

From: Peter Zijlstra <[email protected]>
Date: Tue, 02 Feb 2010 15:20:01 +0100

Quoted 6 lines

Subject: hrtimer, softirq: Fix hrtimer->softirq trampoline

hrtimers callbacks are always done from hardirq context, either the jiffy tick interrupt or the hrtimer device interrupt.

It would be nice to give mention of the bug reporter et al. in the final commit message.

Thanks.

#8 lore raw

[tip:timers/urgent] hrtimer, softirq: Fix hrtimer->softirq trampoline

In reply to Peter Zijlstra
Commit-ID:  b9c3032277f756e73f6c673419dc414155e04e46
Gitweb:     http://git.kernel.org/tip/b9c3032277f756e73f6c673419dc414155e04e46
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Wed, 3 Feb 2010 18:08:52 +0100
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 3 Feb 2010 18:17:40 +0100

hrtimer, softirq: Fix hrtimer->softirq trampoline

hrtimers callbacks are always done from hardirq context, either the jiffy tick interrupt or the hrtimer device interrupt.

[ there is currently one exception that can still call a hrtimer
callback from softirq, but even in that case this will still work correctly. ]

LKML-Reference: <1265120401.24455.306.camel@laptop>

---
 kernel/softirq.c |   15 +++++----------
 1 files changed, 5 insertions(+), 10 deletions(-)
kernel/softirq.c +5 -10
index a09502e..7c1a67e 100644
@@ -500,22 +500,17 @@ EXPORT_SYMBOL(tasklet_kill);
  */
 
 /*
- * The trampoline is called when the hrtimer expires. If this is
- * called from the hrtimer interrupt then we schedule the tasklet as
- * the timer callback function expects to run in softirq context. If
- * it's called in softirq context anyway (i.e. high resolution timers
- * disabled) then the hrtimer callback is called right away.
+ * The trampoline is called when the hrtimer expires. It schedules a tasklet
+ * to run __tasklet_hrtimer_trampoline() which in turn will call the intended
+ * hrtimer callback, but from softirq context.
  */
 static enum hrtimer_restart __hrtimer_tasklet_trampoline(struct hrtimer *timer)
 {
 	struct tasklet_hrtimer *ttimer =
 		container_of(timer, struct tasklet_hrtimer, timer);
 
-	if (hrtimer_is_hres_active(timer)) {
-		tasklet_hi_schedule(&ttimer->tasklet);
-		return HRTIMER_NORESTART;
-	}
-	return ttimer->function(timer);
+	tasklet_hi_schedule(&ttimer->tasklet);
+	return HRTIMER_NORESTART;
 }
 
 /*