[PATCH] ARM: pxa: irq: fix handling of ICMR registers in suspend/resume

[PATCH] ARM: pxa: irq: fix handling of ICMR registers in suspend/resume

[Daniel Mack]

PXA3xx platforms have 56 interrupts that are stored in two ICMR
registers. The code in pxa_irq_suspend() and pxa_irq_resume() however
does a simple division by 32 which only leads to one register being
saved at suspend and restored at resume time. The NAND interrupt
setting, for instance, is lost.

Fix this by using DIV_ROUND_UP() instead.

Signed-off-by: Daniel Mack <daniel@zonque.org>
---
 arch/arm/mach-pxa/irq.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/arm/mach-pxa/irq.c b/arch/arm/mach-pxa/irq.c
index 9c10248fadcc..4e8c2116808e 100644
--- a/arch/arm/mach-pxa/irq.c
+++ b/arch/arm/mach-pxa/irq.c
@@ -185,7 +185,7 @@ static int pxa_irq_suspend(void)
 {
 	int i;
 
-	for (i = 0; i < pxa_internal_irq_nr / 32; i++) {
+	for (i = 0; i < DIV_ROUND_UP(pxa_internal_irq_nr, 32); i++) {
 		void __iomem *base = irq_base(i);
 
 		saved_icmr[i] = __raw_readl(base + ICMR);
@@ -204,7 +204,7 @@ static void pxa_irq_resume(void)
 {
 	int i;
 
-	for (i = 0; i < pxa_internal_irq_nr / 32; i++) {
+	for (i = 0; i < DIV_ROUND_UP(pxa_internal_irq_nr, 32); i++) {
 		void __iomem *base = irq_base(i);
 
 		__raw_writel(saved_icmr[i], base + ICMR);
-- 
2.17.1

[PATCH] ARM: pxa: irq: fix handling of ICMR registers in suspend/resume

[Robert Jarzmik]

Daniel Mack <daniel@zonque.org> writes:

> PXA3xx platforms have 56 interrupts that are stored in two ICMR
> registers. The code in pxa_irq_suspend() and pxa_irq_resume() however
> does a simple division by 32 which only leads to one register being
> saved at suspend and restored at resume time. The NAND interrupt
> setting, for instance, is lost.
>
> Fix this by using DIV_ROUND_UP() instead.
>
> Signed-off-by: Daniel Mack <daniel@zonque.org>
> ---
>  arch/arm/mach-pxa/irq.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm/mach-pxa/irq.c b/arch/arm/mach-pxa/irq.c
> index 9c10248fadcc..4e8c2116808e 100644
> --- a/arch/arm/mach-pxa/irq.c
> +++ b/arch/arm/mach-pxa/irq.c
> @@ -185,7 +185,7 @@ static int pxa_irq_suspend(void)
>  {
>  	int i;
>  
> -	for (i = 0; i < pxa_internal_irq_nr / 32; i++) {
> +	for (i = 0; i < DIV_ROUND_UP(pxa_internal_irq_nr, 32); i++) {
Okay, I get it.  Still, something tells me saved_icmr deserves some love too,
even if the result in the end won't change, it's just for corectness sake.

Of course, while@it I wonder if we couldn't have :
 - pxa_init_irq_common()
 - a kmalloc_array() or similar for saved_icmr[] and saved_ipr[]
That could be in this patch or another one ...

Cheers.

-- 
Robert

[PATCH] ARM: pxa: irq: fix handling of ICMR registers in suspend/resume

[Daniel Mack]

On 07/06/2018 11:28 PM, Robert Jarzmik wrote:
> Daniel Mack <daniel@zonque.org> writes:
> 
>> PXA3xx platforms have 56 interrupts that are stored in two ICMR
>> registers. The code in pxa_irq_suspend() and pxa_irq_resume() however
>> does a simple division by 32 which only leads to one register being
>> saved at suspend and restored at resume time. The NAND interrupt
>> setting, for instance, is lost.
>>
>> Fix this by using DIV_ROUND_UP() instead.
>>
>> Signed-off-by: Daniel Mack <daniel@zonque.org>
>> ---
>>  arch/arm/mach-pxa/irq.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/arm/mach-pxa/irq.c b/arch/arm/mach-pxa/irq.c
>> index 9c10248fadcc..4e8c2116808e 100644
>> --- a/arch/arm/mach-pxa/irq.c
>> +++ b/arch/arm/mach-pxa/irq.c
>> @@ -185,7 +185,7 @@ static int pxa_irq_suspend(void)
>>  {
>>  	int i;
>>  
>> -	for (i = 0; i < pxa_internal_irq_nr / 32; i++) {
>> +	for (i = 0; i < DIV_ROUND_UP(pxa_internal_irq_nr, 32); i++) {
> Okay, I get it.  Still, something tells me saved_icmr deserves some love too,
> even if the result in the end won't change, it's just for corectness sake.
> 
> Of course, while at it I wonder if we couldn't have :
>  - pxa_init_irq_common()
>  - a kmalloc_array() or similar for saved_icmr[] and saved_ipr[]
> That could be in this patch or another one ...

I'd rather do it in a separate patch then and leave this one as-is.


Thanks,
Daniel

[PATCH] ARM: pxa: irq: fix handling of ICMR registers in suspend/resume

[Robert Jarzmik]

Daniel Mack <daniel@zonque.org> writes:

> I'd rather do it in a separate patch then and leave this one as-is.
Queued to pxa/fixes.

Cheers.

-- 
Robert