[PATCH 1/1] mmp2: add handling on PMIC IRQ

[Eric Miao]

On Tue, Jan 26, 2010 at 10:08 PM, Haojian Zhuang
<haojian.zhuang at gmail.com> wrote:
> From 43f04d3f0ab95721b5fd6b9590990ffa1b1eef1d Mon Sep 17 00:00:00 2001
> From: Haojian Zhuang <haojian.zhuang at marvell.com>
> Date: Wed, 27 Jan 2010 08:49:12 -0500
> Subject: [PATCH] mmp2: add handling on PMIC IRQ
>
> Since PMIC INT pin is a special pin of CPU, the status of PMIC INT pin needs
> to be cleared after PMIC IRQ occured. Now append the clear operation in
> irq chip handler.
>
> Signed-off-by: Haojian Zhuang <haojian.zhuang at marvell.com>
> ---
>  arch/arm/mach-mmp/irq-mmp2.c |    7 +++++++
>  1 files changed, 7 insertions(+), 0 deletions(-)
>
> diff --git a/arch/arm/mach-mmp/irq-mmp2.c b/arch/arm/mach-mmp/irq-mmp2.c
> index 41b29f6..761abe3 100644
> --- a/arch/arm/mach-mmp/irq-mmp2.c
> +++ b/arch/arm/mach-mmp/irq-mmp2.c
> @@ -62,7 +62,9 @@ static void _name_##_unmask_irq(unsigned int irq)                     \
>  static void _name_##_irq_demux(unsigned int irq, struct irq_desc *desc)        \
>  {                                                                      \
>        unsigned long status, mask, n;                                  \
> +       unsigned long mfpr_pmic, data;                                  \
>        mask = __raw_readl(prefix##_MASK);                              \
> +       mfpr_pmic = APB_VIRT_BASE + 0x1e000 + 0x2c4;                    \
>        while (1) {                                                     \
>                status = __raw_readl(prefix##_STATUS) & ~mask;          \
>                if (status == 0)                                        \
> @@ -70,6 +72,11 @@ static void _name_##_irq_demux(unsigned int irq,
> struct irq_desc *desc)  \
>                n = find_first_bit(&status, BITS_PER_LONG);             \
>                while (n < BITS_PER_LONG) {                             \
>                        generic_handle_irq(irq_base + n);               \
> +                       if ((irq_base + n) == IRQ_MMP2_PMIC) {          \
> +                               data = __raw_readl(mfpr_pmic);          \
> +                               __raw_writel(data | (1 << 6), mfpr_pmic); \
> +                               __raw_writel(data, mfpr_pmic);          \
> +                       }                                               \
>                        n = find_next_bit(&status, BITS_PER_LONG, n+1); \
>                }                                                       \
>        }                                                               \

I don't think this is probably the right way to go. There could be two
improvements:

1. move mfpr register manipulation into a dedicated function, e.g.
    mmp2_clear_pmic_int() within mmp2.c

2. invent a pmic_mask_ack_irq() to replace the default one at
   run-time, so to save the time that every 2nd level interrupt
   handling has to go through the if (.... == IRQ_MMP2_PMIC) thing,
   provided it does look like an ack operation to me.