[PATCH] irqchip: tegra: fix wrong data being passed as the irqdomain chip data

[PATCH] irqchip: tegra: fix wrong data being passed as the irqdomain chip data

[Lucas Stach]

The irq chip functions use the irq chipdata directly as the base register
address of the controller, so this should be passed in instead of a pointer
to the array address holding the base address.

This fixes Tegra20 CPUidle as now the un-/masking of IRQs at the LIC level
works again, but more importantly it fixes the resulting memory corruption.

Signed-off-by: Lucas Stach <dev@lynxeye.de>
---
This is an important fix and should go into 4.1.
---
 drivers/irqchip/irq-tegra.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/irqchip/irq-tegra.c b/drivers/irqchip/irq-tegra.c
index 51c485d..f67bbd8 100644
--- a/drivers/irqchip/irq-tegra.c
+++ b/drivers/irqchip/irq-tegra.c
@@ -264,7 +264,7 @@ static int tegra_ictlr_domain_alloc(struct irq_domain *domain,
 
 		irq_domain_set_hwirq_and_chip(domain, virq + i, hwirq + i,
 					      &tegra_ictlr_chip,
-					      &info->base[ictlr]);
+					      info->base[ictlr]);
 	}
 
 	parent_args = *args;
-- 
2.1.0

Re: [PATCH] irqchip: tegra: fix wrong data being passed as the irqdomain chip data

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 1368 bytes --]

On Sat, May 09, 2015 at 10:06:54PM +0200, Lucas Stach wrote:
> The irq chip functions use the irq chipdata directly as the base register
> address of the controller, so this should be passed in instead of a pointer
> to the array address holding the base address.
> 
> This fixes Tegra20 CPUidle as now the un-/masking of IRQs at the LIC level
> works again, but more importantly it fixes the resulting memory corruption.
> 
> Signed-off-by: Lucas Stach <dev-8ppwABl0HbeELgA04lAiVw@public.gmane.org>
> ---
> This is an important fix and should go into 4.1.
> ---
>  drivers/irqchip/irq-tegra.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

This does indeed fix a boot regression on Tegra20. Unfortunately this is
not exposed on any platform that uses PCIe because PCIe needs to disable
the CPUidle LP2 state as a workaround for a hardware bug. However, I was
able to reproduce the regression on TrimSlice by disabling PCIe, hence
keeping CPUidle LP2 activated. I reproduced with v4.1-rc3 and applying
this patch on top restores functionality.

Thanks for tracking this down Lucas.

I just noticed that Thomas already applied this while I was testing it,
but here goes anyway:

Tested-by: Thierry Reding <treding-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>
Acked-by: Thierry Reding <treding-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH] irqchip: tegra: fix wrong data being passed as the irqdomain chip data

[Stephen Warren]

On 05/09/2015 02:06 PM, Lucas Stach wrote:
> The irq chip functions use the irq chipdata directly as the base register
> address of the controller, so this should be passed in instead of a pointer
> to the array address holding the base address.
>
> This fixes Tegra20 CPUidle as now the un-/masking of IRQs at the LIC level
> works again, but more importantly it fixes the resulting memory corruption.

Tested-by: Stephen Warren <swarren-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>

Tested on NVIDIA Seaboard (which is affected by the bug in the default 
configuration/DT) with the patch applied on to of 4.1-rc3.