Skip to contentSkip to conversation

linux-next RCU complaint during boot

DISCUSSION [email protected] 3 participants 5 messages Sep 1, 2011 – Sep 2, 2011

lore
Share & tools
lore
#0 lore raw

Seen during boot of linux-next 20110831:

[    8.823047] ------------[ cut here ]------------
[    8.823177] WARNING: at kernel/rcutree.c:1821 rcu_needs_cpu_quick_check+0x6e/0x204()
[    8.823314] Hardware name: Latitude E6500
[    8.823418] Modules linked in:
[    8.823601] Pid: 0, comm: swapper Not tainted 3.1.0-rc4-next-20110831-dirty #11
[    8.823724] Call Trace:
[    8.823824]  <IRQ>  [<ffffffff81039682>] warn_slowpath_common+0x7e/0x96
[    8.824015]  [<ffffffff81092467>] ? note_new_gpnum+0x32/0x96
[    8.824034]  [<ffffffff810396af>] warn_slowpath_null+0x15/0x17
[    8.824034]  [<ffffffff810917de>] rcu_needs_cpu_quick_check+0x6e/0x204
[    8.824034]  [<ffffffff81093df8>] rcu_needs_cpu+0x9/0xd
[    8.824034]  [<ffffffff81065933>] tick_nohz_stop_sched_tick+0x20a/0x420
[    8.824034]  [<ffffffff8103fd77>] irq_exit+0xd2/0xe0
[    8.824034]  [<ffffffff81018f91>] smp_apic_timer_interrupt+0x8b/0x99
[    8.824034]  [<ffffffff815a1b30>] apic_timer_interrupt+0x70/0x80
[    8.824034]  <EOI>  [<ffffffff8159dbbf>] ? notifier_call_chain+0x20/0x81
[    8.824034]  [<ffffffff8159dc7e>] __atomic_notifier_call_chain+0x5e/0x90
[    8.824034]  [<ffffffff8159dc20>] ? notifier_call_chain+0x81/0x81
[    8.824034]  [<ffffffff8159dcbf>] atomic_notifier_call_chain+0xf/0x11
[    8.824034]  [<ffffffff8100092d>] cpu_idle+0xaf/0x102
[    8.824034]  [<ffffffff81569ddc>] rest_init+0xd0/0xd7
[    8.824034]  [<ffffffff81569d0c>] ? csum_partial_copy_generic+0x16c/0x16c
[    8.824034]  [<ffffffff81b31b09>] start_kernel+0x375/0x380
[    8.824034]  [<ffffffff81b3129f>] x86_64_start_reservations+0xaf/0xb3
[    8.824034]  [<ffffffff81b31393>] x86_64_start_kernel+0xf0/0xf7
[    8.824034] ---[ end trace f6289633ba846f8f ]---

Any ideas?

#1 lore raw
Quoted 29 lines · On Thu, Sep 01, 2011 at 12:33:37PM -0400, [email protected] wrote:

Seen during boot of linux-next 20110831:

[    8.823047] ------------[ cut here ]------------
[    8.823177] WARNING: at kernel/rcutree.c:1821 rcu_needs_cpu_quick_check+0x6e/0x204()
[    8.823314] Hardware name: Latitude E6500
[    8.823418] Modules linked in:
[    8.823601] Pid: 0, comm: swapper Not tainted 3.1.0-rc4-next-20110831-dirty #11
[    8.823724] Call Trace:
[    8.823824]  <IRQ>  [<ffffffff81039682>] warn_slowpath_common+0x7e/0x96
[    8.824015]  [<ffffffff81092467>] ? note_new_gpnum+0x32/0x96
[    8.824034]  [<ffffffff810396af>] warn_slowpath_null+0x15/0x17
[    8.824034]  [<ffffffff810917de>] rcu_needs_cpu_quick_check+0x6e/0x204
[    8.824034]  [<ffffffff81093df8>] rcu_needs_cpu+0x9/0xd
[    8.824034]  [<ffffffff81065933>] tick_nohz_stop_sched_tick+0x20a/0x420
[    8.824034]  [<ffffffff8103fd77>] irq_exit+0xd2/0xe0
[    8.824034]  [<ffffffff81018f91>] smp_apic_timer_interrupt+0x8b/0x99
[    8.824034]  [<ffffffff815a1b30>] apic_timer_interrupt+0x70/0x80
[    8.824034]  <EOI>  [<ffffffff8159dbbf>] ? notifier_call_chain+0x20/0x81
[    8.824034]  [<ffffffff8159dc7e>] __atomic_notifier_call_chain+0x5e/0x90
[    8.824034]  [<ffffffff8159dc20>] ? notifier_call_chain+0x81/0x81
[    8.824034]  [<ffffffff8159dcbf>] atomic_notifier_call_chain+0xf/0x11
[    8.824034]  [<ffffffff8100092d>] cpu_idle+0xaf/0x102
[    8.824034]  [<ffffffff81569ddc>] rest_init+0xd0/0xd7
[    8.824034]  [<ffffffff81569d0c>] ? csum_partial_copy_generic+0x16c/0x16c
[    8.824034]  [<ffffffff81b31b09>] start_kernel+0x375/0x380
[    8.824034]  [<ffffffff81b3129f>] x86_64_start_reservations+0xaf/0xb3
[    8.824034]  [<ffffffff81b31393>] x86_64_start_kernel+0xf0/0xf7
[    8.824034] ---[ end trace f6289633ba846f8f ]---
Quoted 1 line

Any ideas?

This is a bug where someone uses RCU at a time when RCU has been told that the CPU is in dyntick-idle mode. Frederic Weisbecker has some patches that should address this.

Thanx, Paul

#2 lore raw
Quoted 30 lines · Paul E. McKenney
Quoted 29 lines · On Thu, Sep 01, 2011 at 12:33:37PM -0400, [email protected] wrote:

Seen during boot of linux-next 20110831:

[    8.823047] ------------[ cut here ]------------
[    8.823177] WARNING: at kernel/rcutree.c:1821 rcu_needs_cpu_quick_check+0x6e/0x204()
[    8.823314] Hardware name: Latitude E6500
[    8.823418] Modules linked in:
[    8.823601] Pid: 0, comm: swapper Not tainted 3.1.0-rc4-next-20110831-dirty #11
[    8.823724] Call Trace:
[    8.823824]  <IRQ>  [<ffffffff81039682>] warn_slowpath_common+0x7e/0x96
[    8.824015]  [<ffffffff81092467>] ? note_new_gpnum+0x32/0x96
[    8.824034]  [<ffffffff810396af>] warn_slowpath_null+0x15/0x17
[    8.824034]  [<ffffffff810917de>] rcu_needs_cpu_quick_check+0x6e/0x204
[    8.824034]  [<ffffffff81093df8>] rcu_needs_cpu+0x9/0xd
[    8.824034]  [<ffffffff81065933>] tick_nohz_stop_sched_tick+0x20a/0x420
[    8.824034]  [<ffffffff8103fd77>] irq_exit+0xd2/0xe0
[    8.824034]  [<ffffffff81018f91>] smp_apic_timer_interrupt+0x8b/0x99
[    8.824034]  [<ffffffff815a1b30>] apic_timer_interrupt+0x70/0x80
[    8.824034]  <EOI>  [<ffffffff8159dbbf>] ? notifier_call_chain+0x20/0x81
[    8.824034]  [<ffffffff8159dc7e>] __atomic_notifier_call_chain+0x5e/0x90
[    8.824034]  [<ffffffff8159dc20>] ? notifier_call_chain+0x81/0x81
[    8.824034]  [<ffffffff8159dcbf>] atomic_notifier_call_chain+0xf/0x11
[    8.824034]  [<ffffffff8100092d>] cpu_idle+0xaf/0x102
[    8.824034]  [<ffffffff81569ddc>] rest_init+0xd0/0xd7
[    8.824034]  [<ffffffff81569d0c>] ? csum_partial_copy_generic+0x16c/0x16c
[    8.824034]  [<ffffffff81b31b09>] start_kernel+0x375/0x380
[    8.824034]  [<ffffffff81b3129f>] x86_64_start_reservations+0xaf/0xb3
[    8.824034]  [<ffffffff81b31393>] x86_64_start_kernel+0xf0/0xf7
[    8.824034] ---[ end trace f6289633ba846f8f ]---
Quoted 5 lines
Quoted 1 line

Any ideas?

This is a bug where someone uses RCU at a time when RCU has been told that the CPU is in dyntick-idle mode. Frederic Weisbecker has some patches that should address this.

Yep, see there: https://lkml.org/lkml/2011/8/20/81

#3 lore raw
Quoted 7 lines · Frederic Weisbecker
Quoted 3 lines · Paul E. McKenney

This is a bug where someone uses RCU at a time when RCU has been told that the CPU is in dyntick-idle mode. Frederic Weisbecker has some patches that should address this.

Yep, see there: https://lkml.org/lkml/2011/8/20/81

Hmm.. .would test, but didn't apply cleanly to -next from Wednesday. Of course, I was trying to do it before corn flakes, so I probably got the patches out of order. WIll re-try this evening if I get a chance.

#4 lore raw
Quoted 7 lines · On Fri, Sep 02, 2011 at 12:06:43PM -0400, [email protected] wrote:
Quoted 6 lines · Frederic Weisbecker
Quoted 3 lines · Paul E. McKenney

This is a bug where someone uses RCU at a time when RCU has been told that the CPU is in dyntick-idle mode. Frederic Weisbecker has some patches that should address this.

Quoted 5 lines
Quoted 1 line

Yep, see there: https://lkml.org/lkml/2011/8/20/81

Hmm.. .would test, but didn't apply cleanly to -next from Wednesday. Of course, I was trying to do it before corn flakes, so I probably got the patches out of order. WIll re-try this evening if I get a chance.

I need to queue those patches anyway. With luck they will hit -rcu by the end of the weekend.

Thanx, Paul