Hi All,
This patch series aims to drop using platform_get_resource() for IRQ types
in preparation for removal of static setup of IRQ resource from DT core
code.
Dropping usage of platform_get_resource() was agreed based on
the discussion [0].
[0] https://patchwork.kernel.org/project/linux-renesas-soc/
patch/[email protected]/
While I was at it created patch 2/2 to propagate errors upstream.
Cheers,
Prabhakar
Lad Prabhakar (2):
iommu/arm-smmu: Use platform_irq_count() to get the interrupt count
iommu/arm-smmu: Propagate errors from platform_get_irq()
drivers/iommu/arm/arm-smmu/arm-smmu.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)
--
2.17.1
platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
allocation of IRQ resources in DT core code, this causes an issue
when using hierarchical interrupt domains using "interrupts" property
in the node as this bypasses the hierarchical setup and messes up the
irq chaining.
In preparation for removal of static setup of IRQ resource from DT core
code use platform_get_irq_count().
Signed-off-by: Lad Prabhakar <[email protected]>
---
drivers/iommu/arm/arm-smmu/arm-smmu.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
index 4bc75c4ce402..4844cd075644 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
@@ -2105,12 +2105,12 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
if (IS_ERR(smmu))
return PTR_ERR(smmu);
- num_irqs = 0;
- while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, num_irqs))) {
- num_irqs++;
- if (num_irqs > smmu->num_global_irqs)
- smmu->num_context_irqs++;
- }
+ num_irqs = platform_irq_count(pdev);
+ if (num_irqs < 0)
+ return num_irqs;
+
+ if (num_irqs > smmu->num_global_irqs)
+ smmu->num_context_irqs += (num_irqs - smmu->num_global_irqs);
if (!smmu->num_context_irqs) {
dev_err(dev, "found %d interrupts but expected at least %d\n",
--
2.17.1
The driver overrides the error code returned by platform_get_irq() to
-ENODEV. Switch to propagating the error code upstream so that errors
such as -EPROBE_DEFER are handled.
Fixes: 9ec36cafe43b ("of/irq: do irq resolution in platform_get_irq")
Signed-off-by: Lad Prabhakar <[email protected]>
---
drivers/iommu/arm/arm-smmu/arm-smmu.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
index 4844cd075644..6cf5612efcda 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
@@ -2129,7 +2129,7 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
int irq = platform_get_irq(pdev, i);
if (irq < 0)
- return -ENODEV;
+ return irq;
smmu->irqs[i] = irq;
}
--
2.17.1
On 2021-12-23 13:00, Lad Prabhakar wrote:
> platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
> allocation of IRQ resources in DT core code, this causes an issue
> when using hierarchical interrupt domains using "interrupts" property
> in the node as this bypasses the hierarchical setup and messes up the
> irq chaining.
>
> In preparation for removal of static setup of IRQ resource from DT core
> code use platform_get_irq_count().
Nit: platform_irq_count()
> Signed-off-by: Lad Prabhakar <[email protected]>
> ---
> drivers/iommu/arm/arm-smmu/arm-smmu.c | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> index 4bc75c4ce402..4844cd075644 100644
> --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> @@ -2105,12 +2105,12 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
> if (IS_ERR(smmu))
> return PTR_ERR(smmu);
>
> - num_irqs = 0;
> - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, num_irqs))) {
> - num_irqs++;
> - if (num_irqs > smmu->num_global_irqs)
> - smmu->num_context_irqs++;
> - }
> + num_irqs = platform_irq_count(pdev);
> + if (num_irqs < 0)
> + return num_irqs;
> +
> + if (num_irqs > smmu->num_global_irqs)
> + smmu->num_context_irqs += (num_irqs - smmu->num_global_irqs);
This seems a bit overcomplicated. I reckon:
smmu->num_context_irqs = num_irqs - smmu->num_global_irqs;
if (num_irqs <= smmu->num_global_irqs) {
dev_err(...
should do it.
However, FYI I have some patches refactoring most of the IRQ stuff here
that I plan to post next cycle (didn't quite have time to get them done
for 5.17 as I'd hoped...), so unless this needs to go in right now as an
urgent fix, I'm happy to take care of removing platform_get_resource()
as part of that if it's easier.
Thanks,
Robin.
>
> if (!smmu->num_context_irqs) {
> dev_err(dev, "found %d interrupts but expected at least %d\n",
On 2021-12-23 13:00, Lad Prabhakar wrote:
> The driver overrides the error code returned by platform_get_irq() to
> -ENODEV. Switch to propagating the error code upstream so that errors
> such as -EPROBE_DEFER are handled.
I wouldn't usually expect an SMMU to be wired up to a secondary
interrupt controller that could cause deferral, but on the other hand I
don't think there's any good reason *not* to propagate the original
error anyway, so sure, why not.
Reviewed-by: Robin Murphy <[email protected]>
> Fixes: 9ec36cafe43b ("of/irq: do irq resolution in platform_get_irq")
> Signed-off-by: Lad Prabhakar <[email protected]>
> ---
> drivers/iommu/arm/arm-smmu/arm-smmu.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> index 4844cd075644..6cf5612efcda 100644
> --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> @@ -2129,7 +2129,7 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
> int irq = platform_get_irq(pdev, i);
>
> if (irq < 0)
> - return -ENODEV;
> + return irq;
> smmu->irqs[i] = irq;
> }
>
Hi Robin,
Thank you for the review.
On Thu, Dec 23, 2021 at 2:14 PM Robin Murphy <[email protected]> wrote:
>
> On 2021-12-23 13:00, Lad Prabhakar wrote:
> > platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
> > allocation of IRQ resources in DT core code, this causes an issue
> > when using hierarchical interrupt domains using "interrupts" property
> > in the node as this bypasses the hierarchical setup and messes up the
> > irq chaining.
> >
> > In preparation for removal of static setup of IRQ resource from DT core
> > code use platform_get_irq_count().
>
> Nit: platform_irq_count()
>
> > Signed-off-by: Lad Prabhakar <[email protected]>
> > ---
> > drivers/iommu/arm/arm-smmu/arm-smmu.c | 12 ++++++------
> > 1 file changed, 6 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > index 4bc75c4ce402..4844cd075644 100644
> > --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > @@ -2105,12 +2105,12 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
> > if (IS_ERR(smmu))
> > return PTR_ERR(smmu);
> >
> > - num_irqs = 0;
> > - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, num_irqs))) {
> > - num_irqs++;
> > - if (num_irqs > smmu->num_global_irqs)
> > - smmu->num_context_irqs++;
> > - }
> > + num_irqs = platform_irq_count(pdev);
> > + if (num_irqs < 0)
> > + return num_irqs;
> > +
> > + if (num_irqs > smmu->num_global_irqs)
> > + smmu->num_context_irqs += (num_irqs - smmu->num_global_irqs);
>
> This seems a bit overcomplicated. I reckon:
>
> smmu->num_context_irqs = num_irqs - smmu->num_global_irqs;
> if (num_irqs <= smmu->num_global_irqs) {
> dev_err(...
>
> should do it.
>
Agreed.
> However, FYI I have some patches refactoring most of the IRQ stuff here
> that I plan to post next cycle (didn't quite have time to get them done
> for 5.17 as I'd hoped...), so unless this needs to go in right now as an
> urgent fix, I'm happy to take care of removing platform_get_resource()
> as part of that if it's easier.
>
Fine by me, let me know if it gets any later than planned I'll send a v2.
Cheers,
Prabhakar
On Thu, Dec 23, 2021 at 02:14:35PM +0000, Robin Murphy wrote:
> On 2021-12-23 13:00, Lad Prabhakar wrote:
> > platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
> > allocation of IRQ resources in DT core code, this causes an issue
> > when using hierarchical interrupt domains using "interrupts" property
> > in the node as this bypasses the hierarchical setup and messes up the
> > irq chaining.
> >
> > In preparation for removal of static setup of IRQ resource from DT core
> > code use platform_get_irq_count().
>
> Nit: platform_irq_count()
>
> > Signed-off-by: Lad Prabhakar <[email protected]>
> > ---
> > drivers/iommu/arm/arm-smmu/arm-smmu.c | 12 ++++++------
> > 1 file changed, 6 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > index 4bc75c4ce402..4844cd075644 100644
> > --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > @@ -2105,12 +2105,12 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
> > if (IS_ERR(smmu))
> > return PTR_ERR(smmu);
> > - num_irqs = 0;
> > - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, num_irqs))) {
> > - num_irqs++;
> > - if (num_irqs > smmu->num_global_irqs)
> > - smmu->num_context_irqs++;
> > - }
> > + num_irqs = platform_irq_count(pdev);
> > + if (num_irqs < 0)
> > + return num_irqs;
> > +
> > + if (num_irqs > smmu->num_global_irqs)
> > + smmu->num_context_irqs += (num_irqs - smmu->num_global_irqs);
>
> This seems a bit overcomplicated. I reckon:
>
> smmu->num_context_irqs = num_irqs - smmu->num_global_irqs;
> if (num_irqs <= smmu->num_global_irqs) {
> dev_err(...
>
> should do it.
>
> However, FYI I have some patches refactoring most of the IRQ stuff here that
> I plan to post next cycle (didn't quite have time to get them done for 5.17
> as I'd hoped...), so unless this needs to go in right now as an urgent fix,
> I'm happy to take care of removing platform_get_resource() as part of that
> if it's easier.
Did you get anywhere with this? December 23rd is long forgotten by now ;)
Will
On Tue, Feb 08, 2022 at 03:28:50PM +0000, Robin Murphy wrote:
> On 2022-02-08 15:19, Will Deacon wrote:
> > On Thu, Dec 23, 2021 at 02:14:35PM +0000, Robin Murphy wrote:
> > > On 2021-12-23 13:00, Lad Prabhakar wrote:
> > > > platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
> > > > allocation of IRQ resources in DT core code, this causes an issue
> > > > when using hierarchical interrupt domains using "interrupts" property
> > > > in the node as this bypasses the hierarchical setup and messes up the
> > > > irq chaining.
> > > >
> > > > In preparation for removal of static setup of IRQ resource from DT core
> > > > code use platform_get_irq_count().
> > >
> > > Nit: platform_irq_count()
> > >
> > > > Signed-off-by: Lad Prabhakar <[email protected]>
> > > > ---
> > > > drivers/iommu/arm/arm-smmu/arm-smmu.c | 12 ++++++------
> > > > 1 file changed, 6 insertions(+), 6 deletions(-)
> > > >
> > > > diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > > > index 4bc75c4ce402..4844cd075644 100644
> > > > --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > > > +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> > > > @@ -2105,12 +2105,12 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
> > > > if (IS_ERR(smmu))
> > > > return PTR_ERR(smmu);
> > > > - num_irqs = 0;
> > > > - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, num_irqs))) {
> > > > - num_irqs++;
> > > > - if (num_irqs > smmu->num_global_irqs)
> > > > - smmu->num_context_irqs++;
> > > > - }
> > > > + num_irqs = platform_irq_count(pdev);
> > > > + if (num_irqs < 0)
> > > > + return num_irqs;
> > > > +
> > > > + if (num_irqs > smmu->num_global_irqs)
> > > > + smmu->num_context_irqs += (num_irqs - smmu->num_global_irqs);
> > >
> > > This seems a bit overcomplicated. I reckon:
> > >
> > > smmu->num_context_irqs = num_irqs - smmu->num_global_irqs;
> > > if (num_irqs <= smmu->num_global_irqs) {
> > > dev_err(...
> > >
> > > should do it.
> > >
> > > However, FYI I have some patches refactoring most of the IRQ stuff here that
> > > I plan to post next cycle (didn't quite have time to get them done for 5.17
> > > as I'd hoped...), so unless this needs to go in right now as an urgent fix,
> > > I'm happy to take care of removing platform_get_resource() as part of that
> > > if it's easier.
> >
> > Did you get anywhere with this? December 23rd is long forgotten by now ;)
>
> Yup: https://gitlab.arm.com/linux-arm/linux-rm/-/commit/b2a40caaf1622eb35c555074a0d72f4f0513cff9
>
> I'm still cleaning up the PMU driver that justifies the whole thing, but I
> can send that patch now if you reckon it's not complete nonsense out of
> context. Otherwise, I'll aim to get the whole lot out next week.
Next week is fine, no rush. I was just trying to work out what to do with
Lad's patches in this thread and it sounds like I should ignore them and
wait for your series instead.
Will
On 2022-02-08 15:19, Will Deacon wrote:
> On Thu, Dec 23, 2021 at 02:14:35PM +0000, Robin Murphy wrote:
>> On 2021-12-23 13:00, Lad Prabhakar wrote:
>>> platform_get_resource(pdev, IORESOURCE_IRQ, ..) relies on static
>>> allocation of IRQ resources in DT core code, this causes an issue
>>> when using hierarchical interrupt domains using "interrupts" property
>>> in the node as this bypasses the hierarchical setup and messes up the
>>> irq chaining.
>>>
>>> In preparation for removal of static setup of IRQ resource from DT core
>>> code use platform_get_irq_count().
>>
>> Nit: platform_irq_count()
>>
>>> Signed-off-by: Lad Prabhakar <[email protected]>
>>> ---
>>> drivers/iommu/arm/arm-smmu/arm-smmu.c | 12 ++++++------
>>> 1 file changed, 6 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
>>> index 4bc75c4ce402..4844cd075644 100644
>>> --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
>>> +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
>>> @@ -2105,12 +2105,12 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
>>> if (IS_ERR(smmu))
>>> return PTR_ERR(smmu);
>>> - num_irqs = 0;
>>> - while ((res = platform_get_resource(pdev, IORESOURCE_IRQ, num_irqs))) {
>>> - num_irqs++;
>>> - if (num_irqs > smmu->num_global_irqs)
>>> - smmu->num_context_irqs++;
>>> - }
>>> + num_irqs = platform_irq_count(pdev);
>>> + if (num_irqs < 0)
>>> + return num_irqs;
>>> +
>>> + if (num_irqs > smmu->num_global_irqs)
>>> + smmu->num_context_irqs += (num_irqs - smmu->num_global_irqs);
>>
>> This seems a bit overcomplicated. I reckon:
>>
>> smmu->num_context_irqs = num_irqs - smmu->num_global_irqs;
>> if (num_irqs <= smmu->num_global_irqs) {
>> dev_err(...
>>
>> should do it.
>>
>> However, FYI I have some patches refactoring most of the IRQ stuff here that
>> I plan to post next cycle (didn't quite have time to get them done for 5.17
>> as I'd hoped...), so unless this needs to go in right now as an urgent fix,
>> I'm happy to take care of removing platform_get_resource() as part of that
>> if it's easier.
>
> Did you get anywhere with this? December 23rd is long forgotten by now ;)
Yup:
https://gitlab.arm.com/linux-arm/linux-rm/-/commit/b2a40caaf1622eb35c555074a0d72f4f0513cff9
I'm still cleaning up the PMU driver that justifies the whole thing, but
I can send that patch now if you reckon it's not complete nonsense out
of context. Otherwise, I'll aim to get the whole lot out next week.
Cheers,
Robin.