[PATCH][BUGFIX] Bug in timer_tsc cpufreq callback

[PATCH][BUGFIX] Bug in timer_tsc cpufreq callback

[Pallipadi, Venkatesh]

[-- Attachment #1: Type: text/plain, Size: 1097 bytes --]


Hi,

There is a bug in cpufreq call back funtion in timer_tsc routines,
that can result in system deadlock. The issue is: grabbing the
write_lock on xtime_lock without disabling the interrupts. So, 
if we happen to get a timer interrupt while we are in this code,
system will go into a deadlock.

This bug only effects the kernels that have CONFIG_CPU_FREQ enabled.
Attached patch fixes it. Please apply.


Thanks,
-Venkatesh


--- linux-2.6.0-test6/arch/i386/kernel/timers/timer_tsc.c.org
2003-10-10 14:01:02.000000000 -0700
+++ linux-2.6.0-test6/arch/i386/kernel/timers/timer_tsc.c
2003-10-10 14:01:24.000000000 -0700
@@ -321,7 +321,7 @@ time_cpufreq_notifier(struct notifier_bl
 {
 	struct cpufreq_freqs *freq = data;
 
-	write_seqlock(&xtime_lock);
+	write_seqlock_irq(&xtime_lock);
 	if (!ref_freq) {
 		ref_freq = freq->old;
 		loops_per_jiffy_ref =
cpu_data[freq->cpu].loops_per_jiffy;
@@ -342,7 +342,7 @@ time_cpufreq_notifier(struct notifier_bl
 		}
 #endif
 	}
-	write_sequnlock(&xtime_lock);
+	write_sequnlock_irq(&xtime_lock);
 
 	return 0;
 }

[-- Attachment #2: tsc-deadlock.patch --]
[-- Type: application/octet-stream, Size: 628 bytes --]

--- linux-2.6.0-test6/arch/i386/kernel/timers/timer_tsc.c.org	2003-10-10 14:01:02.000000000 -0700
+++ linux-2.6.0-test6/arch/i386/kernel/timers/timer_tsc.c	2003-10-10 14:01:24.000000000 -0700
@@ -321,7 +321,7 @@ time_cpufreq_notifier(struct notifier_bl
 {
 	struct cpufreq_freqs *freq = data;
 
-	write_seqlock(&xtime_lock);
+	write_seqlock_irq(&xtime_lock);
 	if (!ref_freq) {
 		ref_freq = freq->old;
 		loops_per_jiffy_ref = cpu_data[freq->cpu].loops_per_jiffy;
@@ -342,7 +342,7 @@ time_cpufreq_notifier(struct notifier_bl
 		}
 #endif
 	}
-	write_sequnlock(&xtime_lock);
+	write_sequnlock_irq(&xtime_lock);
 
 	return 0;
 }

RE: [PATCH][BUGFIX] Bug in timer_tsc cpufreq callback

[Nakajima, Jun]

This is not a theory, but we were really caught by this bug.

	Jun
> -----Original Message-----
> From: linux-kernel-owner@vger.kernel.org [mailto:linux-kernel-
> owner@vger.kernel.org] On Behalf Of Pallipadi, Venkatesh
> Sent: Saturday, October 11, 2003 12:22 PM
> To: torvalds@osdl.org; akpm@osdl.org
> Cc: linux-kernel@vger.kernel.org
> Subject: [PATCH][BUGFIX] Bug in timer_tsc cpufreq callback
> 
> 
> Hi,
> 
> There is a bug in cpufreq call back funtion in timer_tsc routines,
> that can result in system deadlock. The issue is: grabbing the
> write_lock on xtime_lock without disabling the interrupts. So,
> if we happen to get a timer interrupt while we are in this code,
> system will go into a deadlock.
> 
> This bug only effects the kernels that have CONFIG_CPU_FREQ enabled.
> Attached patch fixes it. Please apply.
> 
> 
> Thanks,
> -Venkatesh
> 
> 
> --- linux-2.6.0-test6/arch/i386/kernel/timers/timer_tsc.c.org
> 2003-10-10 14:01:02.000000000 -0700
> +++ linux-2.6.0-test6/arch/i386/kernel/timers/timer_tsc.c
> 2003-10-10 14:01:24.000000000 -0700
> @@ -321,7 +321,7 @@ time_cpufreq_notifier(struct notifier_bl
>  {
>  	struct cpufreq_freqs *freq = data;
> 
> -	write_seqlock(&xtime_lock);
> +	write_seqlock_irq(&xtime_lock);
>  	if (!ref_freq) {
>  		ref_freq = freq->old;
>  		loops_per_jiffy_ref =
> cpu_data[freq->cpu].loops_per_jiffy;
> @@ -342,7 +342,7 @@ time_cpufreq_notifier(struct notifier_bl
>  		}
>  #endif
>  	}
> -	write_sequnlock(&xtime_lock);
> +	write_sequnlock_irq(&xtime_lock);
> 
>  	return 0;
>  }