Re: [tip: irq/core] irqchip/xilinx: Enable generic irq multi handler

[Michal Simek]

On 30. 03. 20 11:14, Marc Zyngier wrote:
> On 2020-03-30 10:03, Michal Simek wrote:
>> On 30. 03. 20 10:45, Marc Zyngier wrote:
>>> On 2020-03-30 09:32, Michal Simek wrote:
>>>> Hi Thomas and Marc,
>>>>
>>>> On 29. 03. 20 22:26, tip-bot2 for Michal Simek wrote:
>>>>> The following commit has been merged into the irq/core branch of tip:
>>>>>
>>>>> Commit-ID:ÂÂÂÂ a0789993bf8266e62fea6b4613945ba081c71e7d
>>>>> Gitweb:ÂÂÂÂÂÂÂ
>>>>> https://git.kernel.org/tip/a0789993bf8266e62fea6b4613945ba081c71e7d
>>>>> Author:ÂÂÂÂÂÂÂ Michal Simek <michal.simek@xxxxxxxxxx>
>>>>> AuthorDate:ÂÂÂ Tue, 17 Mar 2020 18:25:59 +05:30
>>>>> Committer:ÂÂÂÂ Marc Zyngier <maz@xxxxxxxxxx>
>>>>> CommitterDate: Sun, 22 Mar 2020 11:52:53
>>>>>
>>>>> irqchip/xilinx: Enable generic irq multi handler
>>>>>
>>>>> Register default arch handler via driver instead of directly
>>>>> pointing to
>>>>> xilinx intc controller. This patch makes architecture code more
>>>>> generic.
>>>>>
>>>>> Driver calls generic domain specific irq handler which does the
>>>>> most of
>>>>> things self. Also get rid of concurrent_irq counting which hasn't been
>>>>> exported anywhere.
>>>>> Based on this loop was also optimized by using do/while loop
>>>>> instead of
>>>>> goto loop.
>>>>>
>>>>> Signed-off-by: Michal Simek <michal.simek@xxxxxxxxxx>
>>>>> Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
>>>>> Reviewed-by: Stefan Asserhall <stefan.asserhall@xxxxxxxxxx>
>>>>> Link:
>>>>> https://lore.kernel.org/r/20200317125600.15913-4-mubin.usman.sayyed@xxxxxxxxxx
>>>>>
>>>>>
>>>>> ---
>>>>> Âarch/microblaze/KconfigÂÂÂÂÂÂÂÂÂÂ |Â 2 ++-
>>>>> Âarch/microblaze/include/asm/irq.h |Â 3 +---
>>>>> Âarch/microblaze/kernel/irq.cÂÂÂÂÂ | 21 +-------------------
>>>>> Âdrivers/irqchip/irq-xilinx-intc.c | 34 +++++++++++++++++-------------
>>>>> Â4 files changed, 23 insertions(+), 37 deletions(-)
>>>>>
>>>>> diff --git a/arch/microblaze/Kconfig b/arch/microblaze/Kconfig
>>>>> index 6a331bd..242f58e 100644
>>>>> --- a/arch/microblaze/Kconfig
>>>>> +++ b/arch/microblaze/Kconfig
>>>>> @@ -47,6 +47,8 @@ config MICROBLAZE
>>>>> ÂÂÂÂ select CPU_NO_EFFICIENT_FFS
>>>>> ÂÂÂÂ select MMU_GATHER_NO_RANGE if MMU
>>>>> ÂÂÂÂ select SPARSE_IRQ
>>>>> +ÂÂÂ select GENERIC_IRQ_MULTI_HANDLER
>>>>> +ÂÂÂ select HANDLE_DOMAIN_IRQ
>>>>>
>>>>> Â# Endianness selection
>>>>> Âchoice
>>>>> diff --git a/arch/microblaze/include/asm/irq.h
>>>>> b/arch/microblaze/include/asm/irq.h
>>>>> index eac2fb4..5166f08 100644
>>>>> --- a/arch/microblaze/include/asm/irq.h
>>>>> +++ b/arch/microblaze/include/asm/irq.h
>>>>> @@ -14,7 +14,4 @@
>>>>> Âstruct pt_regs;
>>>>> Âextern void do_IRQ(struct pt_regs *regs);
>>>>>
>>>>> -/* should be defined in each interrupt controller driver */
>>>>> -extern unsigned int xintc_get_irq(void);
>>>>> -
>>>>> Â#endif /* _ASM_MICROBLAZE_IRQ_H */
>>>>> diff --git a/arch/microblaze/kernel/irq.c
>>>>> b/arch/microblaze/kernel/irq.c
>>>>> index 903dad8..0b37dde 100644
>>>>> --- a/arch/microblaze/kernel/irq.c
>>>>> +++ b/arch/microblaze/kernel/irq.c
>>>>> @@ -20,29 +20,10 @@
>>>>> Â#include <linux/irqchip.h>
>>>>> Â#include <linux/of_irq.h>
>>>>>
>>>>> -static u32 concurrent_irq;
>>>>> -
>>>>> Âvoid __irq_entry do_IRQ(struct pt_regs *regs)
>>>>> Â{
>>>>> -ÂÂÂ unsigned int irq;
>>>>> -ÂÂÂ struct pt_regs *old_regs = set_irq_regs(regs);
>>>>> ÂÂÂÂ trace_hardirqs_off();
>>>>> -
>>>>> -ÂÂÂ irq_enter();
>>>>> -ÂÂÂ irq = xintc_get_irq();
>>>>> -next_irq:
>>>>> -ÂÂÂ BUG_ON(!irq);
>>>>> -ÂÂÂ generic_handle_irq(irq);
>>>>> -
>>>>> -ÂÂÂ irq = xintc_get_irq();
>>>>> -ÂÂÂ if (irq != -1U) {
>>>>> -ÂÂÂÂÂÂÂ pr_debug("next irq: %d\n", irq);
>>>>> -ÂÂÂÂÂÂÂ ++concurrent_irq;
>>>>> -ÂÂÂÂÂÂÂ goto next_irq;
>>>>> -ÂÂÂ }
>>>>> -
>>>>> -ÂÂÂ irq_exit();
>>>>> -ÂÂÂ set_irq_regs(old_regs);
>>>>> +ÂÂÂ handle_arch_irq(regs);
>>>>> ÂÂÂÂ trace_hardirqs_on();
>>>>> Â}
>>>>>
>>>>> diff --git a/drivers/irqchip/irq-xilinx-intc.c
>>>>> b/drivers/irqchip/irq-xilinx-intc.c
>>>>> index 1d3d273..ea74181 100644
>>>>> --- a/drivers/irqchip/irq-xilinx-intc.c
>>>>> +++ b/drivers/irqchip/irq-xilinx-intc.c
>>>>> @@ -124,20 +124,6 @@ static unsigned int xintc_get_irq_local(struct
>>>>> xintc_irq_chip *irqc)
>>>>> ÂÂÂÂ return irq;
>>>>> Â}
>>>>>
>>>>> -unsigned int xintc_get_irq(void)
>>>>> -{
>>>>> -ÂÂÂ unsigned int irq = -1;
>>>>> -ÂÂÂ u32 hwirq;
>>>>> -
>>>>> -ÂÂÂ hwirq = xintc_read(primary_intc, IVR);
>>>>> -ÂÂÂ if (hwirq != -1U)
>>>>> -ÂÂÂÂÂÂÂ irq = irq_find_mapping(primary_intc->root_domain, hwirq);
>>>>> -
>>>>> -ÂÂÂ pr_debug("irq-xilinx: hwirq=%d, irq=%d\n", hwirq, irq);
>>>>> -
>>>>> -ÂÂÂ return irq;
>>>>> -}
>>>>> -
>>>>> Âstatic int xintc_map(struct irq_domain *d, unsigned int irq,
>>>>> irq_hw_number_t hw)
>>>>> Â{
>>>>> ÂÂÂÂ struct xintc_irq_chip *irqc = d->host_data;
>>>>> @@ -177,6 +163,25 @@ static void xil_intc_irq_handler(struct irq_desc
>>>>> *desc)
>>>>> ÂÂÂÂ chained_irq_exit(chip, desc);
>>>>> Â}
>>>>>
>>>>> +static void xil_intc_handle_irq(struct pt_regs *regs)
>>>>> +{
>>>>> +ÂÂÂ u32 hwirq;
>>>>> +ÂÂÂ struct xintc_irq_chip *irqc = primary_intc;
>>>>> +
>>>>> +ÂÂÂ do {
>>>>> +ÂÂÂÂÂÂÂ hwirq = xintc_read(irqc, IVR);
>>>>> +ÂÂÂÂÂÂÂ if (likely(hwirq != -1U)) {
>>>>> +ÂÂÂÂÂÂÂÂÂÂÂ int ret;
>>>>> +
>>>>> +ÂÂÂÂÂÂÂÂÂÂÂ ret = handle_domain_irq(irqc->root_domain, hwirq, regs);
>>>>> +ÂÂÂÂÂÂÂÂÂÂÂ WARN_ONCE(ret, "Unhandled HWIRQ %d\n", hwirq);
>>>>> +ÂÂÂÂÂÂÂÂÂÂÂ continue;
>>>>> +ÂÂÂÂÂÂÂ }
>>>>> +
>>>>> +ÂÂÂÂÂÂÂ break;
>>>>> +ÂÂÂ } while (1);
>>>>> +}
>>>>> +
>>>>> Âstatic int __init xilinx_intc_of_init(struct device_node *intc,
>>>>> ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ struct device_node *parent)
>>>>> Â{
>>>>> @@ -246,6 +251,7 @@ static int __init xilinx_intc_of_init(struct
>>>>> device_node *intc,
>>>>> ÂÂÂÂ } else {
>>>>> ÂÂÂÂÂÂÂÂ primary_intc = irqc;
>>>>> ÂÂÂÂÂÂÂÂ irq_set_default_host(primary_intc->root_domain);
>>>>> +ÂÂÂÂÂÂÂ set_handle_irq(xil_intc_handle_irq);
>>>>> ÂÂÂÂ }
>>>>>
>>>>> ÂÂÂÂ return 0;
>>>>>
>>>>
>>>> Stephen reported compilation issue when this patch is applied
>>>> because of
>>>> removal of xintc_get_irq() which is also used by ancient ppc405/ppc440
>>>> xilinx platforms. I have reported this twice to Marc already last week.
>>>
>>> Did you? I can't possibly find these emails.
>>
>> Stephen was using your arm.com email.
>>
>> https://lore.kernel.org/linux-next/48d3232d-0f1d-42ea-3109-f44bbabfa2e8@xxxxxxxxxx/
>>
> 
> Yeah, no chance I could read that, unfortunately. I've sent a separate
> email to Stephen to update my email address.
> 
>>>
>>>> On Friday I have also send v1 of removal of that platforms (need to
>>>> send v2)
>>>> https://lore.kernel.org/alsa-devel/cover.1585311091.git.michal.simek@xxxxxxxxxx/
>>>>
>>>>
>>>
>>> You want to remove exising platforms two days before the start of the
>>> merge window?
>>> I don't think this is acceptable with such short notice.
>>
>> I also don't think that will go through.
>>
>>>
>>>> That's why please really consider next steps and let us know.
>>>
>>> I think the only option is to revert (at least partially) the Xilinx
>>> series.
>>
>> Unfortunately.
> 
> Can you please check that we're OK reverting just the last two patches?
> Or what is the minimal revert that can be done for everything to keep
> working?

It looks like that none has any issue to remove these platforms. It
means we can just get back that function with comment and remove it
later when both xilinx ppc platforms are removed. What do you think?
Better to revert one patch instead of two.

>From Stephen:
Caused by commit

  a0789993bf82 ("irqchip/xilinx: Enable generic irq multi handler")

I have reverted that commit (and commit

  9c2d4f525c00 ("irqchip/xilinx: Do not call irq_set_default_host()")

that conflicted with the other revert).

Thanks,
Michal