On Tue, Jul 14, 2015 at 10:16 AM, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> On Tue, 14 Jul 2015, Manuel Lauss wrote:
>
>> On Mon, Jul 13, 2015 at 10:46 PM, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>> > bcsr_csc_handler() is a cascading interrupt handler. It has a
>> > disable_irq_nosync()/enable_irq() pair around the generic_handle_irq()
>> > call. The value of this disable/enable is zero because its a complete
>> > noop:
>> >
>> > disable_irq_nosync() merily increments the disable count without
>> > actually masking the interrupt. enable_irq() soleley decrements the
>> > disable count without touching the interrupt chip. The interrupt
>> > cannot arrive again because the complete call chain runs with
>> > interrupts disabled.
>> >
>> > Remove it.
>>
>> Is there another patch this one depends on? The DB1300 board doesn't
>
> No.
>
>> boot (i.e. interrupts from the cpld aren't serviced) with this patch applied:
>> (irq 136 is the first serviced by the bcsr cpld):
>>
>> irq 136: nobody cared (try booting with the "irqpoll" option)
>
> That's weird. Looking deeper, enable_irq() actually calls
> chip->unmask() unconditionally. So it seems the chip is sensitive to
> that.
>
> Does the following patch on top fix things again?
>
> Thanks,
>
> tglx
> ----
> diff --git a/arch/mips/alchemy/devboards/bcsr.c b/arch/mips/alchemy/devboards/bcsr.c
> index 3a24f2d6ecfd..ec47abe580c6 100644
> --- a/arch/mips/alchemy/devboards/bcsr.c
> +++ b/arch/mips/alchemy/devboards/bcsr.c
> @@ -88,8 +88,11 @@ EXPORT_SYMBOL_GPL(bcsr_mod);
> static void bcsr_csc_handler(unsigned int irq, struct irq_desc *d)
> {
> unsigned short bisr = __raw_readw(bcsr_virt + BCSR_REG_INTSTAT);
> + struct irq_chip *chip = irq_desc_get_chip(d);
>
> + chained_irq_enter(chip, d);
> generic_handle_irq(bcsr_csc_base + __ffs(bisr));
> + chained_irq_exit(chip, d);
> }
>
> static void bcsr_irq_mask(struct irq_data *d)
Yes. Add #include <linux/irqchip/chained_irq.h> on top and it works again.
This hardware is problematic, an older variant with identical verilog
code in the cpld's
irq unit works fine without this.
Thanks,
Manuel
