[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20071119232252.GC3528@suse.de>
Date: Mon, 19 Nov 2007 15:22:52 -0800
From: Greg KH <gregkh@...e.de>
To: Jeremy Fitzhardinge <jeremy@...p.org>
Cc: Greg KH <greg@...ah.com>, David <david@...olicited.net>,
Ingo Molnar <mingo@...e.hu>,
Javier Kohen <jkohen@...rs.sourceforge.net>,
Andrew Morton <akpm@...ux-foundation.org>,
linux-kernel@...r.kernel.org, stable@...nel.org
Subject: Re: [stable] Soft lockups since stable kernel upgrade to 2.6.23.8
On Sat, Nov 17, 2007 at 04:34:56PM -0800, Jeremy Fitzhardinge wrote:
> Greg KH wrote:
> > Great, thanks for tracking this down.
> >
> > Ingo, this corrisponds to changeset
> > a115d5caca1a2905ba7a32b408a6042b20179aaa in mainline. Is that patch
> > incorrect? Should this patch in the -stable tree be reverted?
> >
>
> Hm, I've never observed a problem with this in mainline.
>
> Ah. The significant difference between 2.6.23 and -git is that the
> former used sched_clock as the softlockup timebase, versus cpu_clock in
> git. If sched_clock() is tsc-based, and the tsc isn't stable when using
> cpufreq, then the softlockup with get confused and fire spuriously.
> Ingo's fix to reporting exposed the fact that softlockup is terminally
> broken in that kernel.
>
> I think the best course for now is to revert it, since softlockup is
> hardly a critical feature. The proper fixes would either be to backport
> cpu_clock() to 2.6.23, or make it go back to using ticks.
Can you try applying the patch below to see if that solves the problem
for you?
thanks,
greg k-h
-------------
From: Ingo Molnar <mingo@...e.hu>
Date: Sun, 18 Nov 2007 01:55:38 +0100
Subject: softlockup watchdog fixes and cleanups
To: Greg KH <greg@...ah.com>
Cc: David <david@...olicited.net>, Jeremy Fitzhardinge <jeremy@...p.org>, gregkh@...e.de, Javier Kohen <jkohen@...rs.sourceforge.net>, Andrew Morton <akpm@...ux-foundation.org>, linux-kernel@...r.kernel.org, stable@...nel.org
Message-ID: <20071118005538.GD26865@...e.hu>
Content-Disposition: inline
From: Ingo Molnar <mingo@...e.hu>
This is a merge of commits a5f2ce3c6024a5bb895647b6bd88ecae5001020a and
43581a10075492445f65234384210492ff333eba in mainline to fix a warning in
the 2.6.23.3 kernel release.
softlockup watchdog: style cleanups
kernel/softirq.c grew a few style uncleanlinesses in the past few
months, clean that up. No functional changes:
text data bss dec hex filename
1126 76 4 1206 4b6 softlockup.o.before
1129 76 4 1209 4b9 softlockup.o.after
( the 3 bytes .text increase is due to the "<1>" appended to one of
the printk messages. )
Signed-off-by: Ingo Molnar <mingo@...e.hu>
Signed-off-by: Andrew Morton <akpm@...ux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@...ux-foundation.org>
softlockup: improve debug output
Improve the debuggability of kernel lockups by enhancing the debug
output of the softlockup detector: print the task that causes the lockup
and try to print a more intelligent backtrace.
The old format was:
BUG: soft lockup detected on CPU#1!
[<c0105e4a>] show_trace_log_lvl+0x19/0x2e
[<c0105f43>] show_trace+0x12/0x14
[<c0105f59>] dump_stack+0x14/0x16
[<c015f6bc>] softlockup_tick+0xbe/0xd0
[<c013457d>] run_local_timers+0x12/0x14
[<c01346b8>] update_process_times+0x3e/0x63
[<c0145fb8>] tick_sched_timer+0x7c/0xc0
[<c0140a75>] hrtimer_interrupt+0x135/0x1ba
[<c011bde7>] smp_apic_timer_interrupt+0x6e/0x80
[<c0105aa3>] apic_timer_interrupt+0x33/0x38
[<c0104f8a>] syscall_call+0x7/0xb
=======================
The new format is:
BUG: soft lockup detected on CPU#1! [prctl:2363]
Pid: 2363, comm: prctl
EIP: 0060:[<c013915f>] CPU: 1
EIP is at sys_prctl+0x24/0x18c
EFLAGS: 00000213 Not tainted (2.6.22-cfs-v20 #26)
EAX: 00000001 EBX: 000003e7 ECX: 00000001 EDX: f6df0000
ESI: 000003e7 EDI: 000003e7 EBP: f6df0fb0 DS: 007b ES: 007b FS: 00d8
CR0: 8005003b CR2: 4d8c3340 CR3: 3731d000 CR4: 000006d0
[<c0105e4a>] show_trace_log_lvl+0x19/0x2e
[<c0105f43>] show_trace+0x12/0x14
[<c01040be>] show_regs+0x1ab/0x1b3
[<c015f807>] softlockup_tick+0xef/0x108
[<c013457d>] run_local_timers+0x12/0x14
[<c01346b8>] update_process_times+0x3e/0x63
[<c0145fcc>] tick_sched_timer+0x7c/0xc0
[<c0140a89>] hrtimer_interrupt+0x135/0x1ba
[<c011bde7>] smp_apic_timer_interrupt+0x6e/0x80
[<c0105aa3>] apic_timer_interrupt+0x33/0x38
[<c0104f8a>] syscall_call+0x7/0xb
=======================
Note that in the old format we only knew that some system call locked
up, we didnt know _which_. With the new format we know that it's at a
specific place in sys_prctl(). [which was where i created an artificial
kernel lockup to test the new format.]
This is also useful if the lockup happens in user-space - the user-space
EIP (and other registers) will be printed too. (such a lockup would
either suggest that the task was running at SCHED_FIFO:99 and looping
for more than 10 seconds, or that the softlockup detector has a
false-positive.)
The task name is printed too first, just in case we dont manage to print
a useful backtrace.
[satyam@...radead.org: fix warning]
Signed-off-by: Ingo Molnar <mingo@...e.hu>
Signed-off-by: Satyam Sharma <satyam@...radead.org>
Signed-off-by: Andrew Morton <akpm@...ux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@...ux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@...e.de>
---
kernel/softlockup.c | 37 +++++++++++++++++++++++--------------
1 file changed, 23 insertions(+), 14 deletions(-)
--- a/kernel/softlockup.c
+++ b/kernel/softlockup.c
@@ -15,13 +15,16 @@
#include <linux/notifier.h>
#include <linux/module.h>
+#include <asm/irq_regs.h>
+
static DEFINE_SPINLOCK(print_lock);
static DEFINE_PER_CPU(unsigned long, touch_timestamp);
static DEFINE_PER_CPU(unsigned long, print_timestamp);
static DEFINE_PER_CPU(struct task_struct *, watchdog_task);
-static int did_panic = 0;
+static int did_panic;
+int softlockup_thresh = 10;
static int
softlock_panic(struct notifier_block *this, unsigned long event, void *ptr)
@@ -70,6 +73,7 @@ void softlockup_tick(void)
int this_cpu = smp_processor_id();
unsigned long touch_timestamp = per_cpu(touch_timestamp, this_cpu);
unsigned long print_timestamp;
+ struct pt_regs *regs = get_irq_regs();
unsigned long now;
if (touch_timestamp == 0) {
@@ -99,21 +103,26 @@ void softlockup_tick(void)
wake_up_process(per_cpu(watchdog_task, this_cpu));
/* Warn about unreasonable 10+ seconds delays: */
- if (now > (touch_timestamp + 10)) {
- per_cpu(print_timestamp, this_cpu) = touch_timestamp;
+ if (now <= (touch_timestamp + softlockup_thresh))
+ return;
+
+ per_cpu(print_timestamp, this_cpu) = touch_timestamp;
- spin_lock(&print_lock);
- printk(KERN_ERR "BUG: soft lockup detected on CPU#%d!\n",
- this_cpu);
+ spin_lock(&print_lock);
+ printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %lus! [%s:%d]\n",
+ this_cpu, now - touch_timestamp,
+ current->comm, current->pid);
+ if (regs)
+ show_regs(regs);
+ else
dump_stack();
- spin_unlock(&print_lock);
- }
+ spin_unlock(&print_lock);
}
/*
* The watchdog thread - runs every second and touches the timestamp.
*/
-static int watchdog(void * __bind_cpu)
+static int watchdog(void *__bind_cpu)
{
struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
@@ -151,13 +160,13 @@ cpu_callback(struct notifier_block *nfb,
BUG_ON(per_cpu(watchdog_task, hotcpu));
p = kthread_create(watchdog, hcpu, "watchdog/%d", hotcpu);
if (IS_ERR(p)) {
- printk("watchdog for %i failed\n", hotcpu);
+ printk(KERN_ERR "watchdog for %i failed\n", hotcpu);
return NOTIFY_BAD;
}
- per_cpu(touch_timestamp, hotcpu) = 0;
- per_cpu(watchdog_task, hotcpu) = p;
+ per_cpu(touch_timestamp, hotcpu) = 0;
+ per_cpu(watchdog_task, hotcpu) = p;
kthread_bind(p, hotcpu);
- break;
+ break;
case CPU_ONLINE:
case CPU_ONLINE_FROZEN:
wake_up_process(per_cpu(watchdog_task, hotcpu));
@@ -177,7 +186,7 @@ cpu_callback(struct notifier_block *nfb,
kthread_stop(p);
break;
#endif /* CONFIG_HOTPLUG_CPU */
- }
+ }
return NOTIFY_OK;
}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists