lkml.org 
[lkml]   [2010]   [Oct]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] x86: avoid calling arch_trigger_all_cpu_backtrace() at the same time on SMP
On Mon, Oct 18, 2010 at 07:00:15PM +0800, DDD wrote:
> CC to Ingo's mingo@elte.hu and add some code explains for this patch.
>
> Dongdong
>
>
> Dongdong Deng wrote:
> >The spin_lock_debug/rcu_cpu_stall detector uses
> >trigger_all_cpu_backtrace() to dump cpu backtrace.
> >Therefore it is possible that trigger_all_cpu_backtrace()
> >could be called at the same time on different CPUs, which
> >triggers and 'unknown reason NMI' warning. The following case
> >illustrates the problem:

The patch seems reasonable. I can queue it up. Ingo wrote the original
code and knows which spin locks to use better than me, perhaps he can find
a moment to comment.

Cheers,
Don

> >
> > CPU1 CPU2 ... CPU N
> > trigger_all_cpu_backtrace()
> > set "backtrace_mask" to cpu mask
> > |
> >generate NMI interrupts generate NMI interrupts ...
> > \ | /
> > \ | /
> > The "backtrace_mask" will be cleaned by the first NMI interrupt
> > at nmi_watchdog_tick(), then the following NMI interrupts generated
> >by other cpus's arch_trigger_all_cpu_backtrace() will be took as
> >unknown reason NMI interrupts.
> >
> >This patch uses a lock to avoid the problem, and stop the
> >arch_trigger_all_cpu_backtrace() calling to avoid dumping double cpu
> >backtrace info when there is already a trigger_all_cpu_backtrace()
> >in progress.
> >
> >Signed-off-by: Dongdong Deng <dongdong.deng@windriver.com>
> >Reviewed-by: Bruce Ashfield <bruce.ashfield@windriver.com>
> >CC: Thomas Gleixner <tglx@linutronix.de>
> >CC: Ingo Molnar <mingo@redhat.com>
> >CC: "H. Peter Anvin" <hpa@zytor.com>
> >CC: x86@kernel.org
> >CC: linux-kernel@vger.kernel.org
> >---
> > arch/x86/kernel/apic/hw_nmi.c | 14 ++++++++++++++
> > arch/x86/kernel/apic/nmi.c | 14 ++++++++++++++
> > 2 files changed, 28 insertions(+), 0 deletions(-)
> >
> >diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c
> >index cefd694..3aea0a5 100644
> >--- a/arch/x86/kernel/apic/hw_nmi.c
> >+++ b/arch/x86/kernel/apic/hw_nmi.c
> >@@ -29,6 +29,16 @@ u64 hw_nmi_get_sample_period(void)
> > void arch_trigger_all_cpu_backtrace(void)
> > {
> > int i;
> >+ static arch_spinlock_t lock = __ARCH_SPIN_LOCK_UNLOCKED;
>
> Why an arch spin lock vs just using a raw spin lock?
> for example. static DEFINE_RAW_SPINLOCK(lock);
>
> The spin_lock_debug detector was used in raw_spinlock too.
>
> arch_trigger_all_cpu_backtrace() -->
> raw_spin_lock(lock) -->
> _raw_spin_lock(lock) -->
> __raw_spin_lock(lock) -->
> do_raw_spin_lock(lock) -->
> __spin_lock_debug(lock) -->
> trigger_all_cpu_backtrace()
>
> Therefor, we have to use arch spin lock here.
>
>
>
> >+ unsigned long flags;
> >+
> >+ local_irq_save(flags);
>
> Why have to save the irq's here?
>
> When the arch_trigger_all_cpu_backtrace() was triggered by
> "spin_lock()"'s spin_lock_debug detector, it is possible that
> the irq is enabled, thus we have to save and disable it here.
>
>
>
> >+ if (!arch_spin_trylock(&lock))
> >+ /*
> >+ * If there is already a trigger_all_cpu_backtrace()
> >+ * in progress, don't output double cpu dump infos.
> >+ */
> >+ goto out_restore_irq;
> > cpumask_copy(to_cpumask(backtrace_mask), cpu_online_mask);
> >@@ -41,6 +51,10 @@ void arch_trigger_all_cpu_backtrace(void)
> > break;
> > mdelay(1);
> > }
> >+
> >+ arch_spin_unlock(&lock);
> >+out_restore_irq:
> >+ local_irq_restore(flags);
> > }
> > static int __kprobes
> >diff --git a/arch/x86/kernel/apic/nmi.c b/arch/x86/kernel/apic/nmi.c
> >index a43f71c..5fa8a13 100644
> >--- a/arch/x86/kernel/apic/nmi.c
> >+++ b/arch/x86/kernel/apic/nmi.c
> >@@ -552,6 +552,16 @@ int do_nmi_callback(struct pt_regs *regs, int cpu)
> > void arch_trigger_all_cpu_backtrace(void)
> > {
> > int i;
> >+ static arch_spinlock_t lock = __ARCH_SPIN_LOCK_UNLOCKED;
> >+ unsigned long flags;
> >+
> >+ local_irq_save(flags);
> >+ if (!arch_spin_trylock(&lock))
> >+ /*
> >+ * If there is already a trigger_all_cpu_backtrace()
> >+ * in progress, don't output double cpu dump infos.
> >+ */
> >+ goto out_restore_irq;
> > cpumask_copy(to_cpumask(backtrace_mask), cpu_online_mask);
> >@@ -564,4 +574,8 @@ void arch_trigger_all_cpu_backtrace(void)
> > break;
> > mdelay(1);
> > }
> >+
> >+ arch_spin_unlock(&lock);
> >+out_restore_irq:
> >+ local_irq_restore(flags);
> > }
>
>
>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/


\
 
 \ /
  Last update: 2010-10-18 20:07    [W:0.059 / U:2.676 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site