[PATCH 0/3] irqchip/gic-v5: Minor clean-ups/fixes

[PATCH 0/3] irqchip/gic-v5: Minor clean-ups/fixes

[Lorenzo Pieralisi]

A trivial fix for a kbot reported issue and two clean-ups
discovered after the initial series was queued for v6.17;
now that the driver code is merged I am sharing them.

Only patch (1) is a fix, the two others are cleaning the code up
and removing useless constructs.

The series apply cleanly on top of this three-patch series [1] fixing
some exit path issues with the gic-v5 driver that should also
be considered for merging please (I reviewed it and tested it [2]):

[1] https://lore.kernel.org/lkml/670bb1dc-7827-4916-84f3-bb46ee408b20@sabinyo.mountain/
[2] https://lore.kernel.org/lkml/aHoU8sKPuyzEFRIn@lpieralisi/

Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
---
Lorenzo Pieralisi (3):
      irqchip/gic-v5: iwb: Fix iounmap probe failure path
      irqchip/gic-v5: iwb: Remove module boilerplate
      irqchip/gic-v5: Remove IRQD_RESEND_WHEN_IN_PROGRESS for ITS IRQs

 drivers/irqchip/irq-gic-v5-its.c |  1 -
 drivers/irqchip/irq-gic-v5-iwb.c | 14 +++-----------
 2 files changed, 3 insertions(+), 12 deletions(-)
---
base-commit: 260f6f4fda93c8485c8037865c941b42b9cba5d2
change-id: 20250731-gic-v5-fixes-6-17-949eae3c5594

Best regards,
-- 
Lorenzo Pieralisi <lpieralisi@kernel.org>

[PATCH 1/3] irqchip/gic-v5: iwb: Fix iounmap probe failure path

[Lorenzo Pieralisi]

Kbot reported that on the failure path the driver
iounmap() IWB resources that are managed through
devm_ioremap(), which is clearly wrong because
the driver would end up unmapping the MMIO
resource twice on probing failure.

Fix this by removing the error path altogether
and by letting devres manage the iounmapping
on clean-up.

Fixes: 695949d8b16f ("irqchip/gic-v5: Add GICv5 IWB support")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202508010038.N3r4ZmII-lkp@intel.com
Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marc Zyngier <maz@kernel.org>
---
 drivers/irqchip/irq-gic-v5-iwb.c | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

diff --git a/drivers/irqchip/irq-gic-v5-iwb.c b/drivers/irqchip/irq-gic-v5-iwb.c
index ed72fbdd4900..ad9fdc14d1c6 100644
--- a/drivers/irqchip/irq-gic-v5-iwb.c
+++ b/drivers/irqchip/irq-gic-v5-iwb.c
@@ -241,7 +241,6 @@ static int gicv5_iwb_device_probe(struct platform_device *pdev)
 	struct gicv5_iwb_chip_data *iwb_node;
 	void __iomem *iwb_base;
 	struct resource *res;
-	int ret;
 
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	if (!res)
@@ -254,16 +253,10 @@ static int gicv5_iwb_device_probe(struct platform_device *pdev)
 	}
 
 	iwb_node = gicv5_iwb_init_bases(iwb_base, pdev);
-	if (IS_ERR(iwb_node)) {
-		ret = PTR_ERR(iwb_node);
-		goto out_unmap;
-	}
+	if (IS_ERR(iwb_node))
+		return PTR_ERR(iwb_node);
 
 	return 0;
-
-out_unmap:
-	iounmap(iwb_base);
-	return ret;
 }
 
 static const struct of_device_id gicv5_iwb_of_match[] = {

-- 
2.48.0

Re: [PATCH 1/3] irqchip/gic-v5: iwb: Fix iounmap probe failure path

[Marc Zyngier]

On Fri, 01 Aug 2025 08:58:18 +0100,
Lorenzo Pieralisi <lpieralisi@kernel.org> wrote:
> 
> Kbot reported that on the failure path the driver
> iounmap() IWB resources that are managed through
> devm_ioremap(), which is clearly wrong because
> the driver would end up unmapping the MMIO
> resource twice on probing failure.
> 
> Fix this by removing the error path altogether
> and by letting devres manage the iounmapping
> on clean-up.
> 
> Fixes: 695949d8b16f ("irqchip/gic-v5: Add GICv5 IWB support")
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202508010038.N3r4ZmII-lkp@intel.com
> Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Marc Zyngier <maz@kernel.org>

Reviewed-by: Marc Zyngier <maz@kernel.org>

	M.

-- 
Without deviation from the norm, progress is not possible.

[PATCH 2/3] irqchip/gic-v5: iwb: Remove module boilerplate

[Lorenzo Pieralisi]

The IWB driver cannot be compiled as a module and is as matter of fact a
builtin driver at present.

Make it explicitly so by removing useless boilerplate and by using the
builtin_platform_driver() helper to initialize its registration
functions.

Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marc Zyngier <maz@kernel.org>
---
 drivers/irqchip/irq-gic-v5-iwb.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/irqchip/irq-gic-v5-iwb.c b/drivers/irqchip/irq-gic-v5-iwb.c
index ad9fdc14d1c6..c3a3b30e9f4e 100644
--- a/drivers/irqchip/irq-gic-v5-iwb.c
+++ b/drivers/irqchip/irq-gic-v5-iwb.c
@@ -263,7 +263,6 @@ static const struct of_device_id gicv5_iwb_of_match[] = {
 	{ .compatible = "arm,gic-v5-iwb" },
 	{ /* END */ }
 };
-MODULE_DEVICE_TABLE(of, gicv5_iwb_of_match);
 
 static struct platform_driver gicv5_iwb_platform_driver = {
 	.driver = {
@@ -274,4 +273,4 @@ static struct platform_driver gicv5_iwb_platform_driver = {
 	.probe				= gicv5_iwb_device_probe,
 };
 
-module_platform_driver(gicv5_iwb_platform_driver);
+builtin_platform_driver(gicv5_iwb_platform_driver);

-- 
2.48.0

Re: [PATCH 2/3] irqchip/gic-v5: iwb: Remove module boilerplate

[Marc Zyngier]

On Fri, 01 Aug 2025 08:58:19 +0100,
Lorenzo Pieralisi <lpieralisi@kernel.org> wrote:
> 
> The IWB driver cannot be compiled as a module and is as matter of fact a
> builtin driver at present.

What is blocking that?

> 
> Make it explicitly so by removing useless boilerplate and by using the
> builtin_platform_driver() helper to initialize its registration
> functions.
> 
> Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Marc Zyngier <maz@kernel.org>
> ---
>  drivers/irqchip/irq-gic-v5-iwb.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-gic-v5-iwb.c b/drivers/irqchip/irq-gic-v5-iwb.c
> index ad9fdc14d1c6..c3a3b30e9f4e 100644
> --- a/drivers/irqchip/irq-gic-v5-iwb.c
> +++ b/drivers/irqchip/irq-gic-v5-iwb.c
> @@ -263,7 +263,6 @@ static const struct of_device_id gicv5_iwb_of_match[] = {
>  	{ .compatible = "arm,gic-v5-iwb" },
>  	{ /* END */ }
>  };
> -MODULE_DEVICE_TABLE(of, gicv5_iwb_of_match);
>  
>  static struct platform_driver gicv5_iwb_platform_driver = {
>  	.driver = {
> @@ -274,4 +273,4 @@ static struct platform_driver gicv5_iwb_platform_driver = {
>  	.probe				= gicv5_iwb_device_probe,
>  };
>  
> -module_platform_driver(gicv5_iwb_platform_driver);
> +builtin_platform_driver(gicv5_iwb_platform_driver);
> 

I'd rather we make the driver buildable as a module if at all
possible, instead of forcing it as built-in for everyone. It would
definitely help pipe-cleaning the potential missing dependencies.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.

Re: [PATCH 2/3] irqchip/gic-v5: iwb: Remove module boilerplate

[Lorenzo Pieralisi]

On Fri, Aug 01, 2025 at 01:24:57PM +0100, Marc Zyngier wrote:
> On Fri, 01 Aug 2025 08:58:19 +0100,
> Lorenzo Pieralisi <lpieralisi@kernel.org> wrote:
> > 
> > The IWB driver cannot be compiled as a module and is as matter of fact a
> > builtin driver at present.
> 
> What is blocking that?

Yep, you read my mind I need to test this but AFAICS at the moment
the answer is: nothing.

> > Make it explicitly so by removing useless boilerplate and by using the
> > builtin_platform_driver() helper to initialize its registration
> > functions.
> > 
> > Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
> > Cc: Thomas Gleixner <tglx@linutronix.de>
> > Cc: Marc Zyngier <maz@kernel.org>
> > ---
> >  drivers/irqchip/irq-gic-v5-iwb.c | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> > 
> > diff --git a/drivers/irqchip/irq-gic-v5-iwb.c b/drivers/irqchip/irq-gic-v5-iwb.c
> > index ad9fdc14d1c6..c3a3b30e9f4e 100644
> > --- a/drivers/irqchip/irq-gic-v5-iwb.c
> > +++ b/drivers/irqchip/irq-gic-v5-iwb.c
> > @@ -263,7 +263,6 @@ static const struct of_device_id gicv5_iwb_of_match[] = {
> >  	{ .compatible = "arm,gic-v5-iwb" },
> >  	{ /* END */ }
> >  };
> > -MODULE_DEVICE_TABLE(of, gicv5_iwb_of_match);
> >  
> >  static struct platform_driver gicv5_iwb_platform_driver = {
> >  	.driver = {
> > @@ -274,4 +273,4 @@ static struct platform_driver gicv5_iwb_platform_driver = {
> >  	.probe				= gicv5_iwb_device_probe,
> >  };
> >  
> > -module_platform_driver(gicv5_iwb_platform_driver);
> > +builtin_platform_driver(gicv5_iwb_platform_driver);
> > 
> 
> I'd rather we make the driver buildable as a module if at all
> possible, instead of forcing it as built-in for everyone. It would
> definitely help pipe-cleaning the potential missing dependencies.

Yes - I took a shortcut by removing boilerplate (that the
compiler/linker remove anyway) - it makes sense to rework the
build to make it possible to build as a module instead.

This patch can be dropped then, I will work on the follow-up.

Lorenzo

[PATCH 3/3] irqchip/gic-v5: Remove IRQD_RESEND_WHEN_IN_PROGRESS for ITS IRQs

[Lorenzo Pieralisi]

GICv5 LPI interrupts have an active state hence they cannot retrigger
while the IRQ is being handled.

Therefore setting IRQD_RESEND_WHEN_IN_PROGRESS for GICv5 ITS
interrupts provides no benefit and it is useless (and confusing),
it solves an issue that cannot happen.

Remove it.

Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marc Zyngier <maz@kernel.org>
---
 drivers/irqchip/irq-gic-v5-its.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/irqchip/irq-gic-v5-its.c b/drivers/irqchip/irq-gic-v5-its.c
index 340640fdbdf6..9290ac741949 100644
--- a/drivers/irqchip/irq-gic-v5-its.c
+++ b/drivers/irqchip/irq-gic-v5-its.c
@@ -973,7 +973,6 @@ static int gicv5_its_irq_domain_alloc(struct irq_domain *domain, unsigned int vi
 		irqd = irq_get_irq_data(virq + i);
 		irqd_set_single_target(irqd);
 		irqd_set_affinity_on_activate(irqd);
-		irqd_set_resend_when_in_progress(irqd);
 	}
 
 	return 0;

-- 
2.48.0

Re: [PATCH 3/3] irqchip/gic-v5: Remove IRQD_RESEND_WHEN_IN_PROGRESS for ITS IRQs

[Marc Zyngier]

On Fri, 01 Aug 2025 08:58:20 +0100,
Lorenzo Pieralisi <lpieralisi@kernel.org> wrote:
> 
> GICv5 LPI interrupts have an active state hence they cannot retrigger
> while the IRQ is being handled.
> 
> Therefore setting IRQD_RESEND_WHEN_IN_PROGRESS for GICv5 ITS
> interrupts provides no benefit and it is useless (and confusing),
> it solves an issue that cannot happen.

This doesn't read very well. How about:

"Therefore, setting the IRQD_RESEND_WHEN_IN_PROGRESS flag on LPIs
 is pointless, as the situation this flag caters for cannot happen".

>
> Remove it.
> 
> Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Marc Zyngier <maz@kernel.org>
> ---
>  drivers/irqchip/irq-gic-v5-its.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/drivers/irqchip/irq-gic-v5-its.c b/drivers/irqchip/irq-gic-v5-its.c
> index 340640fdbdf6..9290ac741949 100644
> --- a/drivers/irqchip/irq-gic-v5-its.c
> +++ b/drivers/irqchip/irq-gic-v5-its.c
> @@ -973,7 +973,6 @@ static int gicv5_its_irq_domain_alloc(struct irq_domain *domain, unsigned int vi
>  		irqd = irq_get_irq_data(virq + i);
>  		irqd_set_single_target(irqd);
>  		irqd_set_affinity_on_activate(irqd);
> -		irqd_set_resend_when_in_progress(irqd);
>  	}
>  
>  	return 0;
> 

With that,

Reviewed-by: Marc Zyngier <maz@kernel.org>

	M.

-- 
Without deviation from the norm, progress is not possible.

Re: [PATCH 3/3] irqchip/gic-v5: Remove IRQD_RESEND_WHEN_IN_PROGRESS for ITS IRQs

[Lorenzo Pieralisi]

On Fri, Aug 01, 2025 at 01:30:32PM +0100, Marc Zyngier wrote:
> On Fri, 01 Aug 2025 08:58:20 +0100,
> Lorenzo Pieralisi <lpieralisi@kernel.org> wrote:
> > 
> > GICv5 LPI interrupts have an active state hence they cannot retrigger
> > while the IRQ is being handled.
> > 
> > Therefore setting IRQD_RESEND_WHEN_IN_PROGRESS for GICv5 ITS
> > interrupts provides no benefit and it is useless (and confusing),
> > it solves an issue that cannot happen.
> 
> This doesn't read very well. How about:
> 
> "Therefore, setting the IRQD_RESEND_WHEN_IN_PROGRESS flag on LPIs
>  is pointless, as the situation this flag caters for cannot happen".

Yes it is clearer, I can respin a v2 unless Thomas can update the log
please while applying it ? Just let me know.

Thanks,
Lorenzo

> > Remove it.
> > 
> > Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
> > Cc: Thomas Gleixner <tglx@linutronix.de>
> > Cc: Marc Zyngier <maz@kernel.org>
> > ---
> >  drivers/irqchip/irq-gic-v5-its.c | 1 -
> >  1 file changed, 1 deletion(-)
> > 
> > diff --git a/drivers/irqchip/irq-gic-v5-its.c b/drivers/irqchip/irq-gic-v5-its.c
> > index 340640fdbdf6..9290ac741949 100644
> > --- a/drivers/irqchip/irq-gic-v5-its.c
> > +++ b/drivers/irqchip/irq-gic-v5-its.c
> > @@ -973,7 +973,6 @@ static int gicv5_its_irq_domain_alloc(struct irq_domain *domain, unsigned int vi
> >  		irqd = irq_get_irq_data(virq + i);
> >  		irqd_set_single_target(irqd);
> >  		irqd_set_affinity_on_activate(irqd);
> > -		irqd_set_resend_when_in_progress(irqd);
> >  	}
> >  
> >  	return 0;
> > 
> 
> With that,
> 
> Reviewed-by: Marc Zyngier <maz@kernel.org>
> 
> 	M.
> 
> -- 
> Without deviation from the norm, progress is not possible.

Re: [PATCH 0/3] irqchip/gic-v5: Minor clean-ups/fixes

[Lorenzo Pieralisi]

On Fri, Aug 01, 2025 at 09:58:17AM +0200, Lorenzo Pieralisi wrote:
> A trivial fix for a kbot reported issue and two clean-ups
> discovered after the initial series was queued for v6.17;
> now that the driver code is merged I am sharing them.
> 
> Only patch (1) is a fix, the two others are cleaning the code up
> and removing useless constructs.
> 
> The series apply cleanly on top of this three-patch series [1] fixing
> some exit path issues with the gic-v5 driver that should also
> be considered for merging please (I reviewed it and tested it [2]):

Hi Thomas, Marc,

thank you for reviewing and pulling these fixes, please also consider
fixes in [1] below that are still current, Dan sent them before the
series was queued for -next and I reviewed/tested them (in particular
patch (3) - others are obvious).

Thanks,
Lorenzo

> 
> [1] https://lore.kernel.org/lkml/670bb1dc-7827-4916-84f3-bb46ee408b20@sabinyo.mountain/
> [2] https://lore.kernel.org/lkml/aHoU8sKPuyzEFRIn@lpieralisi/
> 
> Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
> ---
> Lorenzo Pieralisi (3):
>       irqchip/gic-v5: iwb: Fix iounmap probe failure path
>       irqchip/gic-v5: iwb: Remove module boilerplate
>       irqchip/gic-v5: Remove IRQD_RESEND_WHEN_IN_PROGRESS for ITS IRQs
> 
>  drivers/irqchip/irq-gic-v5-its.c |  1 -
>  drivers/irqchip/irq-gic-v5-iwb.c | 14 +++-----------
>  2 files changed, 3 insertions(+), 12 deletions(-)
> ---
> base-commit: 260f6f4fda93c8485c8037865c941b42b9cba5d2
> change-id: 20250731-gic-v5-fixes-6-17-949eae3c5594
> 
> Best regards,
> -- 
> Lorenzo Pieralisi <lpieralisi@kernel.org>
>