Re: [patch 10/26] ACPI: Prefer _CST over FADT for C-state capabilities

[Daniel Drake]

Greg KH wrote:
> -stable review patch.  If anyone has any objections, please let us know.
>
> ------------------
> From: Venkatesh Pallipadi <venkatesh.pallipadi@xxxxxxxxx>
>
> Note: This ACPI standard compliance may cause regression
> on some system, if they have _CST present, but _CST value
> is bogus. "nocst" module parameter should workaround
> that regression.
>
> http://bugzilla.kernel.org/show_bug.cgi?id=5165
>
> (cherry picked from 883baf7f7e81cca26f4683ae0d25ba48f094cc08 commit)
>
> Signed-off-by: Venkatesh Pallipadi<venkatesh.pallipadi@xxxxxxxxx>
> Signed-off-by: Len Brown <len.brown@xxxxxxxxx>
> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxx>
> ---
>  drivers/acpi/processor_idle.c |   10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)

Venkatesh followed up in a private email that a 3rd patch is needed to  solve 
the hyperthreading slowdown issue. This patch is not yet in Linus' tree (it is 
in acpi-test).

Maybe we should drop these patches (10 and 12) until the 3rd patch has been 
merged. I haven't been shipping the 3rd patch in Gentoo (yet) so I'm not able 
to gauge its effect...

Attaching the 3rd patch anyway.

Daniel
From: Len Brown <len.brown@xxxxxxxxx>

Bug fix for bugzilla #5165 http://bugzilla.kernel.org/show_bug.cgi?id=5165

Incremental changes to earlier patch.
* Changing the polarity of plvl2_up
* Skip promotion/demotion code when not needed.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@xxxxxxxxx>
Signed-off-by: Shaohua Li <shaohua.li@xxxxxxxxx>

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
@@ -278,8 +278,6 @@ static void acpi_processor_idle(void)
 		}
 	}
 
-	cx->usage++;
-
 #ifdef CONFIG_HOTPLUG_CPU
 	/*
 	 * Check for P_LVL2_UP flag before entering C2 and above on
@@ -287,9 +285,12 @@ static void acpi_processor_idle(void)
 	 * detection phase, to work cleanly with logical CPU hotplug.
 	 */
 	if ((cx->type != ACPI_STATE_C1) && (num_online_cpus() > 1) && 
-	    !pr->flags.has_cst && acpi_fadt.plvl2_up)
-		cx->type = ACPI_STATE_C1;
+	    !pr->flags.has_cst && !acpi_fadt.plvl2_up)
+		cx = &pr->power.states[ACPI_STATE_C1];
 #endif
+
+	cx->usage++;
+
 	/*
 	 * Sleep:
 	 * ------
@@ -378,6 +379,15 @@ static void acpi_processor_idle(void)
 
 	next_state = pr->power.state;
 
+#ifdef CONFIG_HOTPLUG_CPU
+	/* Don't do promotion/demotion */
+	if ((cx->type == ACPI_STATE_C1) && (num_online_cpus() > 1) &&
+	    !pr->flags.has_cst && !acpi_fadt.plvl2_up) {
+		next_state = cx;
+		goto end;
+	}
+#endif
+
 	/*
 	 * Promotion?
 	 * ----------
@@ -549,7 +559,7 @@ static int acpi_processor_get_power_info
 	 * Check for P_LVL2_UP flag before entering C2 and above on
 	 * an SMP system. 
 	 */
-	if ((num_online_cpus() > 1) && acpi_fadt.plvl2_up)
+	if ((num_online_cpus() > 1) && !acpi_fadt.plvl2_up)
 		return_VALUE(-ENODEV);
 #endif