Re: [PATCH v14 01/18] irqchip/sifive-plic: Convert PLIC driver into a platform driver

[Lad, Prabhakar]

Hi Samuel and Anup,

On Wed, Apr 3, 2024 at 5:28 PM Samuel Holland <samuel.holland@xxxxxxxxxx> wrote:
>
> Hi Prabhakar,
>
> On 2024-04-03 10:49 AM, Lad, Prabhakar wrote:
> > On Wed, Apr 3, 2024 at 3:17 PM Anup Patel <apatel@xxxxxxxxxxxxxxxx> wrote:
> >>
> >> On Wed, Apr 3, 2024 at 2:01 PM Lad, Prabhakar
> >> <prabhakar.csengg@xxxxxxxxx> wrote:
> >>>
> >>> Hi Anup,
> >>>
> >>> On Thu, Feb 22, 2024 at 9:41 AM Anup Patel <apatel@xxxxxxxxxxxxxxxx> wrote:
> >>>>
> >>>> The PLIC driver does not require very early initialization so convert
> >>>> it into a platform driver.
> >>>>
> >>>> After conversion, the PLIC driver is probed after CPUs are brought-up
> >>>> so setup cpuhp state after context handler of all online CPUs are
> >>>> initialized otherwise PLIC driver crashes for platforms with multiple
> >>>> PLIC instances.
> >>>>
> >>>> Signed-off-by: Anup Patel <apatel@xxxxxxxxxxxxxxxx>
> >>>> ---
> >>>>  drivers/irqchip/irq-sifive-plic.c | 101 ++++++++++++++++++------------
> >>>>  1 file changed, 61 insertions(+), 40 deletions(-)
> >>>>
> >>> This patch seems to have broken things on RZ/Five SoC, after reverting
> >>> this patch I get to boot it back again on v6.9-rc2. Looks like there
> >>> is some probe order issue after switching to platform driver?
> >>
> >> Yes, this is most likely related to probe ordering based on your DT.
> >>
> >> Can you share the failing boot log and DT ?
> >
> > non working case, https://paste.debian.net/1312947/
>
> Looks like you need to add "keep_bootcon" to your kernel command line to get a
> full log here.
>
Thanks for the pointer, that helped me to get to the root cause.

> > after reverting, https://paste.debian.net/1312948/
> > (attached is the DTB)
>
> I don't see anything suspicious between the "riscv-intc" lines and the "Fixed
> dependency cycle(s)" lines that looks like it would depend on the PLIC IRQ
> domain. Maybe there is some driver that does not handle -EPROBE_DEFER? It's hard
> to tell without the full log from the failure case.
>
The clock required for the PLIC wasnt available during the probe of
this driver. This bug got hidden when the PLIC driver was probed
earlier  in boot where it used an incorrect clock source. Ive created
a patch which adds a missing clock for the PLIC.

Sorry for the noise!

Cheers,
Prabhakar