[patch] ACPI: Keep TSC stable, when lapic_timer_c2_ok is set

[patch] ACPI: Keep TSC stable, when lapic_timer_c2_ok is set

[Thomas Gleixner]

Len,

can you please integrate the patch into your git tree to make Andrew
happier. I'd like to see this in 2.6.22 if possible.

Thanks,

	tglx

------------------------>

Subject: ACPI: Keep TSC stable, when lapic_timer_c2_ok is set

The local apic timer stop in C2 resp. C3 states is coupled with the
stop of the TSC. When the local apic timer is marked stable in C2
on the kernel commandline, then keep the TSC marked stable in C2 as well.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

---
 drivers/acpi/processor_idle.c |   22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

Index: linux-acpi-2.6/drivers/acpi/processor_idle.c
===================================================================
--- linux-acpi-2.6.orig/drivers/acpi/processor_idle.c
+++ linux-acpi-2.6/drivers/acpi/processor_idle.c
@@ -305,18 +305,23 @@ static void acpi_state_timer_broadcast(s
 				       struct acpi_processor_cx *cx,
 				       int broadcast)
 {
-#ifdef CONFIG_GENERIC_CLOCKEVENTS
-
 	int state = cx - pr->power.states;
 
 	if (state >= pr->power.timer_broadcast_on_state) {
+
+#ifdef CONFIG_GENERIC_CLOCKEVENTS
 		unsigned long reason;
 
 		reason = broadcast ?  CLOCK_EVT_NOTIFY_BROADCAST_ENTER :
 			CLOCK_EVT_NOTIFY_BROADCAST_EXIT;
 		clockevents_notify(reason, &pr->id);
-	}
 #endif
+
+#ifdef CONFIG_GENERIC_TIME
+		/* TSC halts in C2/3, so notify users */
+		mark_tsc_unstable();
+#endif
+	}
 }
 
 #else
@@ -481,10 +489,6 @@ static void acpi_processor_idle(void)
 		/* Get end time (ticks) */
 		t2 = inl(acpi_gbl_FADT.xpm_timer_block.address);
 
-#ifdef CONFIG_GENERIC_TIME
-		/* TSC halts in C2, so notify users */
-		mark_tsc_unstable();
-#endif
 		/* Re-enable interrupts */
 		local_irq_enable();
 		current_thread_info()->status |= TS_POLLING;
@@ -523,10 +527,6 @@ static void acpi_processor_idle(void)
 			acpi_set_register(ACPI_BITREG_ARB_DISABLE, 0);
 		}
 
-#ifdef CONFIG_GENERIC_TIME
-		/* TSC halts in C3, so notify users */
-		mark_tsc_unstable();
-#endif
 		/* Re-enable interrupts */
 		local_irq_enable();
 		current_thread_info()->status |= TS_POLLING;

Re: [patch] ACPI: Keep TSC stable, when lapic_timer_c2_ok is set

[Len Brown]

On Monday 07 May 2007 12:33, Thomas Gleixner wrote:
> Len,
> 
> can you please integrate the patch into your git tree to make Andrew
> happier. I'd like to see this in 2.6.22 if possible.

what tree does this apply to?

Note, however, that the assumption that the TSC and the LAPIC
timer breaking under the same conditions will not always
be true going forward.

In particular, there will be systems with a fully functional TSC
and a broken LAPIC timer.  But I guess we'll cross that bridge
when we come to it...

-Len

> Thanks,
> 
> 	tglx
> 
> ------------------------>
> 
> Subject: ACPI: Keep TSC stable, when lapic_timer_c2_ok is set
> 
> The local apic timer stop in C2 resp. C3 states is coupled with the
> stop of the TSC. When the local apic timer is marked stable in C2
> on the kernel commandline, then keep the TSC marked stable in C2 as well.
> 
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> 
> ---
>  drivers/acpi/processor_idle.c |   22 +++++++++++-----------
>  1 file changed, 11 insertions(+), 11 deletions(-)
> 
> Index: linux-acpi-2.6/drivers/acpi/processor_idle.c
> ===================================================================
> --- linux-acpi-2.6.orig/drivers/acpi/processor_idle.c
> +++ linux-acpi-2.6/drivers/acpi/processor_idle.c
> @@ -305,18 +305,23 @@ static void acpi_state_timer_broadcast(s
>  				       struct acpi_processor_cx *cx,
>  				       int broadcast)
>  {
> -#ifdef CONFIG_GENERIC_CLOCKEVENTS
> -
>  	int state = cx - pr->power.states;
>  
>  	if (state >= pr->power.timer_broadcast_on_state) {
> +
> +#ifdef CONFIG_GENERIC_CLOCKEVENTS
>  		unsigned long reason;
>  
>  		reason = broadcast ?  CLOCK_EVT_NOTIFY_BROADCAST_ENTER :
>  			CLOCK_EVT_NOTIFY_BROADCAST_EXIT;
>  		clockevents_notify(reason, &pr->id);
> -	}
>  #endif
> +
> +#ifdef CONFIG_GENERIC_TIME
> +		/* TSC halts in C2/3, so notify users */
> +		mark_tsc_unstable();
> +#endif
> +	}
>  }
>  
>  #else
> @@ -481,10 +489,6 @@ static void acpi_processor_idle(void)
>  		/* Get end time (ticks) */
>  		t2 = inl(acpi_gbl_FADT.xpm_timer_block.address);
>  
> -#ifdef CONFIG_GENERIC_TIME
> -		/* TSC halts in C2, so notify users */
> -		mark_tsc_unstable();
> -#endif
>  		/* Re-enable interrupts */
>  		local_irq_enable();
>  		current_thread_info()->status |= TS_POLLING;
> @@ -523,10 +527,6 @@ static void acpi_processor_idle(void)
>  			acpi_set_register(ACPI_BITREG_ARB_DISABLE, 0);
>  		}
>  
> -#ifdef CONFIG_GENERIC_TIME
> -		/* TSC halts in C3, so notify users */
> -		mark_tsc_unstable();
> -#endif
>  		/* Re-enable interrupts */
>  		local_irq_enable();
>  		current_thread_info()->status |= TS_POLLING;
> 
> 

Re: [patch] ACPI: Keep TSC stable, when lapic_timer_c2_ok is set

[Thomas Gleixner]

> Note, however, that the assumption that the TSC and the LAPIC
> timer breaking under the same conditions will not always
> be true going forward.
> 
> In particular, there will be systems with a fully functional TSC
> and a broken LAPIC timer.  But I guess we'll cross that bridge
> when we come to it...

Sigh, is there any kind of breakage which is not invented yet ?

I guess we need to have a seperate option for the TSC vs. C2 then.

	tglx