Re: [PATCH] sh: Avoid using IRQ0 on SH3 and SH4

On 7/7/23 08:19, John Paul Adrian Glaubitz wrote:
> Hi Guenter!
> 
>> On Jul 7, 2023, at 4:35 PM, Guenter Roeck <[email protected]> wrote:
>>
>> On 7/7/23 02:53, John Paul Adrian Glaubitz wrote:
>>> Hi Geert!
>>> On Fri, 2023-07-07 at 10:48 +0200, Geert Uytterhoeven wrote:
>>>>> I can reproduce the issue with rts7751r2dplus_defconfig, but I may
>>>>> not be able to look into it today...
>>>>
>>>> Disabling CONFIG_USB_OHCI_HCD fixes the hang.
>>> I picked rts7751r2dplus_defconfig, disabled CONFIG_USB_OHCI_HCD but it still
>>> hangs for me. Are you sure it's CONFIG_USB_OHCI_HCD and not something else?
>>
>> It seems to be related. With this patch reverted, ohci_irq() gets a single
>> interrupt, and boot continues. With this patch in place, ohci_irq() does
>> not get any interrupts, and boot hangs with qemu at 100% CPU. I confirmed
>> this by disabling CONFIG_MFD_SM501. After that, the hang is no longer seen.
>> Of course, that also means that OHCI and other emulated sm501 functionality
>> no longer works.
>>
>> My suspicion is that something goes wrong with interrupt routing to
>> SM501 and with it to ohci_irq(), but that is just a wild guess.
> 
> Looking at drivers/mfd/sm501.c, sm501_plat_probe() is explicitly requesting IRQ 0 which is no longer supported on SH:
> 
>  > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mfd/sm501.c <https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mfd/sm501.c>
> 
> So, we need to shift this by 16, I guess.
> 

It isn't interrupt 0, it is interrupt _index_ 0. See platform_get_irq_optional()
which calls platform_get_resource(dev, IORESOURCE_IRQ, num). There are lots
of calls to platform_get_irq() with index 0 in the kernel.

Guenter