[patch 2/3] intel_idle: disable auto_demotion for hotplugged CPUs

[patch 2/3] intel_idle: disable auto_demotion for hotplugged CPUs

[akpm]

From: Shaohua Li <shaohua.li@intel.com>
Subject: intel_idle: disable auto_demotion for hotplugged CPUs

auto_demotion_disable is called only for online CPUs.  For hotplugged
CPUs, we should disable it too.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 drivers/idle/intel_idle.c |   53 ++++++++++++++++++++++--------------
 1 file changed, 33 insertions(+), 20 deletions(-)

diff -puN drivers/idle/intel_idle.c~intel_idle-disable-auto_demotion-for-hotplugged-cpus drivers/idle/intel_idle.c
--- a/drivers/idle/intel_idle.c~intel_idle-disable-auto_demotion-for-hotplugged-cpus
+++ a/drivers/idle/intel_idle.c
@@ -297,33 +297,42 @@ static void __setup_broadcast_timer(void
 	clockevents_notify(reason, &cpu);
 }
 
-static int setup_broadcast_cpuhp_notify(struct notifier_block *n,
+static void auto_demotion_disable(void *dummy)
+{
+	unsigned long long msr_bits;
+
+	rdmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
+	msr_bits &= ~auto_demotion_disable_flags;
+	wrmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
+}
+
+static void __intel_idle_notify_handler(void *arg)
+{
+	if (auto_demotion_disable_flags)
+		auto_demotion_disable(NULL);
+
+	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE)
+		__setup_broadcast_timer((void *)true);
+}
+
+static int setup_intelidle_cpuhp_notify(struct notifier_block *n,
 		unsigned long action, void *hcpu)
 {
 	int hotcpu = (unsigned long)hcpu;
 
 	switch (action & 0xf) {
 	case CPU_ONLINE:
-		smp_call_function_single(hotcpu, __setup_broadcast_timer,
-			(void *)true, 1);
+		smp_call_function_single(hotcpu, __intel_idle_notify_handler,
+			NULL, 1);
 		break;
 	}
 	return NOTIFY_OK;
 }
 
-static struct notifier_block setup_broadcast_notifier = {
-	.notifier_call = setup_broadcast_cpuhp_notify,
+static struct notifier_block setup_intelidle_notifier = {
+	.notifier_call = setup_intelidle_cpuhp_notify,
 };
 
-static void auto_demotion_disable(void *dummy)
-{
-	unsigned long long msr_bits;
-
-	rdmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
-	msr_bits &= ~auto_demotion_disable_flags;
-	wrmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
-}
-
 /*
  * intel_idle_probe()
  */
@@ -393,10 +402,8 @@ static int intel_idle_probe(void)
 
 	if (boot_cpu_has(X86_FEATURE_ARAT))	/* Always Reliable APIC Timer */
 		lapic_timer_reliable_states = LAPIC_TIMER_ALWAYS_RELIABLE;
-	else {
+	else
 		on_each_cpu(__setup_broadcast_timer, (void *)true, 1);
-		register_cpu_notifier(&setup_broadcast_notifier);
-	}
 
 	pr_debug(PREFIX "v" INTEL_IDLE_VERSION
 		" model 0x%X\n", boot_cpu_data.x86_model);
@@ -559,6 +566,10 @@ static int __init intel_idle_init(void)
 		return retval;
 	}
 
+	if (auto_demotion_disable_flags || lapic_timer_reliable_states !=
+	    LAPIC_TIMER_ALWAYS_RELIABLE)
+		register_cpu_notifier(&setup_intelidle_notifier);
+
 	return 0;
 }
 
@@ -567,10 +578,12 @@ static void __exit intel_idle_exit(void)
 	intel_idle_cpuidle_devices_uninit();
 	cpuidle_unregister_driver(&intel_idle_driver);
 
-	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE) {
+	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE)
 		on_each_cpu(__setup_broadcast_timer, (void *)false, 1);
-		unregister_cpu_notifier(&setup_broadcast_notifier);
-	}
+
+	if (auto_demotion_disable_flags || lapic_timer_reliable_states !=
+	    LAPIC_TIMER_ALWAYS_RELIABLE)
+		unregister_cpu_notifier(&setup_intelidle_notifier);
 
 	return;
 }
_

Re: [patch 2/3] intel_idle: disable auto_demotion for hotplugged CPUs

[Thomas Renninger]

On Wednesday, January 11, 2012 12:48:20 AM akpm@linux-foundation.org wrote:
> From: Shaohua Li <shaohua.li@intel.com>
> Subject: intel_idle: disable auto_demotion for hotplugged CPUs

This one should not be necessary anymore with my intel_idle
patch series I have to rebase and resend (will do that after lunch).

It provides a per cpu intel_idle init function which is called
from processor.ko in physical cpu hotplug case.

>From what I can see this patch does not change anything/much, because:
  - CPUs onlined at bootup are initialized correctly already
  - Soft offlining/onlining shouldn't matter in auto_demotion_disable
    case as it's a HW MSR write and the CPU should still be in the
    same state after off- and onlined again.
  - Physically hotplugged/added CPUs are not registered with cpuidle
    at all (this is what my patch series fixes), because intel_idle
    does not offer any per cpu init() entry function.
    With this patch they might be set into auto_demotion_disable now,
    but no idle driver gets active for them.

I'll pick up patch 1/3 and 3/3, base my patches on top and resend
everything, that should be easiest for everybody?

   Thomas

Re: [patch 2/3] intel_idle: disable auto_demotion for hotplugged CPUs

[Len Brown]

Andrew,
Per reasoning from Thomas below, please drop this one.

thanks,
-Len

On 01/17/2012 05:56 AM, Thomas Renninger wrote:

> On Wednesday, January 11, 2012 12:48:20 AM akpm@linux-foundation.org wrote:
>> From: Shaohua Li <shaohua.li@intel.com>
>> Subject: intel_idle: disable auto_demotion for hotplugged CPUs
> 
> This one should not be necessary anymore with my intel_idle
> patch series I have to rebase and resend (will do that after lunch).
> 
> It provides a per cpu intel_idle init function which is called
> from processor.ko in physical cpu hotplug case.
> 
> From what I can see this patch does not change anything/much, because:
>   - CPUs onlined at bootup are initialized correctly already
>   - Soft offlining/onlining shouldn't matter in auto_demotion_disable
>     case as it's a HW MSR write and the CPU should still be in the
>     same state after off- and onlined again.
>   - Physically hotplugged/added CPUs are not registered with cpuidle
>     at all (this is what my patch series fixes), because intel_idle
>     does not offer any per cpu init() entry function.
>     With this patch they might be set into auto_demotion_disable now,
>     but no idle driver gets active for them.
> 
> I'll pick up patch 1/3 and 3/3, base my patches on top and resend
> everything, that should be easiest for everybody?
> 
>    Thomas
>