[v4,2/6] irqdomain: clear trigger type in irq_domain_push_irq()

Message ID 1504784522-26841-3-git-send-email-yamada.masahiro@socionext.com
State New
Headers show
Series
  • irqdomain, gpio: expand irq_domain_push_irq() for DT use and use it for GPIO
Related show

Commit Message

Masahiro Yamada Sept. 7, 2017, 11:41 a.m.
Prior to the addition of irq_domain_push_irq(), the hierarchy
IRQ domain always allocates IRQs from the outer-most domain.
Each irqchip usually calls irq_domain_alloc_irqs_parent(),
ascending the topology up to the root irqchip.

The brand-new function irq_domain_push_irq() allows us to allocate
IRQs for parent domain first, then add a child irq_data to the
tail of the chain.

For the new use-case, if the parent sets a temporary trigger type,
it may differ from the type requested to the outer-most irqchip,
then irq_create_fwspec_mapping() warns "type mismatch, failed to map..."

Clear the trigger type when a new irq_data is connected to the chain.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
---

Changes in v4:
  - Newly added


 kernel/irq/irqdomain.c | 3 +++
 1 file changed, 3 insertions(+)

Comments

Marc Zyngier Sept. 7, 2017, 12:25 p.m. | #1
On 07/09/17 12:41, Masahiro Yamada wrote:
> Prior to the addition of irq_domain_push_irq(), the hierarchy
> IRQ domain always allocates IRQs from the outer-most domain.
> Each irqchip usually calls irq_domain_alloc_irqs_parent(),
> ascending the topology up to the root irqchip.
> 
> The brand-new function irq_domain_push_irq() allows us to allocate
> IRQs for parent domain first, then add a child irq_data to the
> tail of the chain.
> 
> For the new use-case, if the parent sets a temporary trigger type,
> it may differ from the type requested to the outer-most irqchip,
> then irq_create_fwspec_mapping() warns "type mismatch, failed to map..."
> 
> Clear the trigger type when a new irq_data is connected to the chain.
> 
> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> ---
> 
> Changes in v4:
>   - Newly added
> 
> 
>  kernel/irq/irqdomain.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c
> index da3e0b6..18d11b9 100644
> --- a/kernel/irq/irqdomain.c
> +++ b/kernel/irq/irqdomain.c
> @@ -1532,6 +1532,9 @@ int irq_domain_push_irq(struct irq_domain *domain, int virq, void *arg)
>  	tail_irq_data->chip = NULL;
>  	tail_irq_data->chip_data = NULL;
>  
> +	/* clear the trigger type to avoid "type mismatch" error */
> +	irqd_set_trigger_type(tail_irq_data, IRQ_TYPE_NONE);
> +

This feels wrong. The initial interrupt hierarchy does have a set
trigger, because it has been configured already, and switching it to
NONE is hiding the fact that you're setting it to another, conflicting
value.

Your new fwspec should have a type that is really compatible with with
the underlying interrupt, however "temporary". If it is not, you have a
problem.

Thanks,

	M.
Masahiro Yamada Sept. 8, 2017, 3:09 p.m. | #2
Hi Marc,


2017-09-07 21:25 GMT+09:00 Marc Zyngier <marc.zyngier@arm.com>:
> On 07/09/17 12:41, Masahiro Yamada wrote:
>> Prior to the addition of irq_domain_push_irq(), the hierarchy
>> IRQ domain always allocates IRQs from the outer-most domain.
>> Each irqchip usually calls irq_domain_alloc_irqs_parent(),
>> ascending the topology up to the root irqchip.
>>
>> The brand-new function irq_domain_push_irq() allows us to allocate
>> IRQs for parent domain first, then add a child irq_data to the
>> tail of the chain.
>>
>> For the new use-case, if the parent sets a temporary trigger type,
>> it may differ from the type requested to the outer-most irqchip,
>> then irq_create_fwspec_mapping() warns "type mismatch, failed to map..."
>>
>> Clear the trigger type when a new irq_data is connected to the chain.
>>
>> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
>> ---
>>
>> Changes in v4:
>>   - Newly added
>>
>>
>>  kernel/irq/irqdomain.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c
>> index da3e0b6..18d11b9 100644
>> --- a/kernel/irq/irqdomain.c
>> +++ b/kernel/irq/irqdomain.c
>> @@ -1532,6 +1532,9 @@ int irq_domain_push_irq(struct irq_domain *domain, int virq, void *arg)
>>       tail_irq_data->chip = NULL;
>>       tail_irq_data->chip_data = NULL;
>>
>> +     /* clear the trigger type to avoid "type mismatch" error */
>> +     irqd_set_trigger_type(tail_irq_data, IRQ_TYPE_NONE);
>> +
>
> This feels wrong. The initial interrupt hierarchy does have a set
> trigger, because it has been configured already, and switching it to
> NONE is hiding the fact that you're setting it to another, conflicting
> value.
>
> Your new fwspec should have a type that is really compatible with with
> the underlying interrupt, however "temporary". If it is not, you have a
> problem.
>


My motivation is to describe interrupt connection by
DT property, like follows:

interrupts = <48 4>, <49 4>, <50 4>, <51 4>,
        <52 4>, <53 4>, <54 4>, <55 4>,
        <56 4>, <57 4>, <58 4>, <59 4>,
        <60 4>, <61 4>, <62 4>, <63 4>,
        <154 4>, <155 4>, <156 4>, <157 4>,
        <158 4>;


The number of cells is fixed, so I need to
give something as the second parameter.

The "4" may not be a real trigger type,
it is just a temporal value to fill the space.


The device may pass a different trigger type
when it calls gpio_to_irq() && request_irq().
I cannot know the real trigger type until the device is probed.


Having said that, if this is a bad idea,
I will consider a different approach for my GPIO driver.

Patch

diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c
index da3e0b6..18d11b9 100644
--- a/kernel/irq/irqdomain.c
+++ b/kernel/irq/irqdomain.c
@@ -1532,6 +1532,9 @@  int irq_domain_push_irq(struct irq_domain *domain, int virq, void *arg)
 	tail_irq_data->chip = NULL;
 	tail_irq_data->chip_data = NULL;
 
+	/* clear the trigger type to avoid "type mismatch" error */
+	irqd_set_trigger_type(tail_irq_data, IRQ_TYPE_NONE);
+
 	/* May (probably does) set hwirq, chip, etc. */
 	rv = irq_domain_alloc_irqs_hierarchy(domain, virq, 1, arg);
 	if (rv) {