[PATCH 1/2] Revert "arm64: use cpu_online_mask when using forced irq_set_affinity"

[PATCH 1/2] Revert "arm64: use cpu_online_mask when using forced irq_set_affinity"

[byungchul.park]

From: Byungchul Park <byungchul.park@lge.com>

This reverts commit 601c942176d8ad8334118bddb747e3720bed24f8.

This patch is designed to ensure that the cpu being offlined is not
present in the affinity mask. But it is a bad idea to overwrite the
affinity variable with cpu_online_mask, even in case that the current
affinity already includes onlined cpus.

So revert this patch to replace it with another one doing exactly
what it intends.

Signed-off-by: Byungchul Park <byungchul.park@lge.com>
---
 arch/arm64/kernel/irq.c |   10 +++-------
 1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/arch/arm64/kernel/irq.c b/arch/arm64/kernel/irq.c
index 0f08dfd..473e5db 100644
--- a/arch/arm64/kernel/irq.c
+++ b/arch/arm64/kernel/irq.c
@@ -97,15 +97,11 @@ static bool migrate_one_irq(struct irq_desc *desc)
 	if (irqd_is_per_cpu(d) || !cpumask_test_cpu(smp_processor_id(), affinity))
 		return false;
 
-	if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids)
+	if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids) {
+		affinity = cpu_online_mask;
 		ret = true;
+	}
 
-	/*
-	 * when using forced irq_set_affinity we must ensure that the cpu
-	 * being offlined is not present in the affinity mask, it may be
-	 * selected as the target CPU otherwise
-	 */
-	affinity = cpu_online_mask;
 	c = irq_data_get_irq_chip(d);
 	if (!c->irq_set_affinity)
 		pr_debug("IRQ%u: unable to set affinity\n", d->irq);
-- 
1.7.9.5

Re: [PATCH 1/2] Revert "arm64: use cpu_online_mask when using forced irq_set_affinity"

[Will Deacon]

On Wed, Aug 27, 2014 at 10:30:06AM +0100, byungchul.park@lge.com wrote:
> From: Byungchul Park <byungchul.park@lge.com>
> 
> This reverts commit 601c942176d8ad8334118bddb747e3720bed24f8.
> 
> This patch is designed to ensure that the cpu being offlined is not
> present in the affinity mask. But it is a bad idea to overwrite the
> affinity variable with cpu_online_mask, even in case that the current
> affinity already includes onlined cpus.
> 
> So revert this patch to replace it with another one doing exactly
> what it intends.

Sudeep: what's the right way forward for this? There seems to be general
agreement that the existing code is broken, but a bunch of different
`fixes'. Can we just take a straight port of what tglx proposed for ARM?
(changing force to false)

Will

Re: [PATCH 1/2] Revert "arm64: use cpu_online_mask when using forced irq_set_affinity"

[Sudeep Holla]

On 28/08/14 10:38, Will Deacon wrote:
> On Wed, Aug 27, 2014 at 10:30:06AM +0100, byungchul.park@lge.com wrote:
>> From: Byungchul Park <byungchul.park@lge.com>
>>
>> This reverts commit 601c942176d8ad8334118bddb747e3720bed24f8.
>>
>> This patch is designed to ensure that the cpu being offlined is not
>> present in the affinity mask. But it is a bad idea to overwrite the
>> affinity variable with cpu_online_mask, even in case that the current
>> affinity already includes onlined cpus.
>>
>> So revert this patch to replace it with another one doing exactly
>> what it intends.
>
> Sudeep: what's the right way forward for this? There seems to be general
> agreement that the existing code is broken, but a bunch of different
> `fixes'. Can we just take a straight port of what tglx proposed for ARM?
> (changing force to false)
>

Yes I agree but for that we need agreement from rmk and hence I asked to
wait till we hear from rmk. Main issue raised by rmk is if some other
interrupt controller implementation decide not to migrate away when
force is false(theoretically possible).

Regards,
Sudeep

Re: [PATCH 1/2] Revert "arm64: use cpu_online_mask when using forced irq_set_affinity"

[Will Deacon]

On Thu, Aug 28, 2014 at 10:49:54AM +0100, Sudeep Holla wrote:
> 
> 
> On 28/08/14 10:38, Will Deacon wrote:
> > On Wed, Aug 27, 2014 at 10:30:06AM +0100, byungchul.park@lge.com wrote:
> >> From: Byungchul Park <byungchul.park@lge.com>
> >>
> >> This reverts commit 601c942176d8ad8334118bddb747e3720bed24f8.
> >>
> >> This patch is designed to ensure that the cpu being offlined is not
> >> present in the affinity mask. But it is a bad idea to overwrite the
> >> affinity variable with cpu_online_mask, even in case that the current
> >> affinity already includes onlined cpus.
> >>
> >> So revert this patch to replace it with another one doing exactly
> >> what it intends.
> >
> > Sudeep: what's the right way forward for this? There seems to be general
> > agreement that the existing code is broken, but a bunch of different
> > `fixes'. Can we just take a straight port of what tglx proposed for ARM?
> > (changing force to false)
> >
> 
> Yes I agree but for that we need agreement from rmk and hence I asked to
> wait till we hear from rmk. Main issue raised by rmk is if some other
> interrupt controller implementation decide not to migrate away when
> force is false(theoretically possible).

Okey doke. Whatever solution we take should be the same for arm and arm64,
so I'll leave it with you.

Will

Re: [PATCH 1/2] Revert "arm64: use cpu_online_mask when using forced irq_set_affinity"

[Sudeep Holla]

On 28/08/14 10:50, Will Deacon wrote:
> On Thu, Aug 28, 2014 at 10:49:54AM +0100, Sudeep Holla wrote:
>>
>>
>> On 28/08/14 10:38, Will Deacon wrote:
>>> On Wed, Aug 27, 2014 at 10:30:06AM +0100, byungchul.park@lge.com wrote:
>>>> From: Byungchul Park <byungchul.park@lge.com>
>>>>
>>>> This reverts commit 601c942176d8ad8334118bddb747e3720bed24f8.
>>>>
>>>> This patch is designed to ensure that the cpu being offlined is not
>>>> present in the affinity mask. But it is a bad idea to overwrite the
>>>> affinity variable with cpu_online_mask, even in case that the current
>>>> affinity already includes onlined cpus.
>>>>
>>>> So revert this patch to replace it with another one doing exactly
>>>> what it intends.
>>>
>>> Sudeep: what's the right way forward for this? There seems to be general
>>> agreement that the existing code is broken, but a bunch of different
>>> `fixes'. Can we just take a straight port of what tglx proposed for ARM?
>>> (changing force to false)
>>>
>>
>> Yes I agree but for that we need agreement from rmk and hence I asked to
>> wait till we hear from rmk. Main issue raised by rmk is if some other
>> interrupt controller implementation decide not to migrate away when
>> force is false(theoretically possible).
>
> Okey doke. Whatever solution we take should be the same for arm and arm64,
> so I'll leave it with you.
>

tglx just confirmed that interrupt controller implementation using force
flag must use online_cpumask. So converting to false should be fine.
So once rmk agrees for ARM, we can apply this revert and change to false
for ARM64 also.

Regards,
Sudeep