Re: [PATCH 3/3 v2] arm64: dts: renesas: add MOST device

From: Geert Uytterhoeven
Date: Wed Jan 12 2022 - 05:15:35 EST


Hi Nikita,

On Wed, Jan 12, 2022 at 9:56 AM Nikita Yushchenko
<nikita.yoush@xxxxxxxxxxxxxxxxxx> wrote:
> >> + reg = <0 0xec520000 0 0x800>;
> >> + interrupts = <GIC_SPI 384 IRQ_TYPE_LEVEL_HIGH>,
> >> + <GIC_SPI 385 IRQ_TYPE_LEVEL_HIGH>,
> >> + <GIC_SPI 386 IRQ_TYPE_LEVEL_HIGH>,
> >> + <GIC_SPI 387 IRQ_TYPE_LEVEL_HIGH>,
> >> + <GIC_SPI 388 IRQ_TYPE_LEVEL_HIGH>;
> >
> > What is the purpose of the various interrupts?
> > Perhaps you need interrupt-names?
> > The driver seems to use only the first two, which is strange, as
> > the second and third interrupt handle different channels.
>
> Maybe Christian Gromm (the original driver author) can comment here?
>
> As far as I understand:
> - interrupts are: mlb, ahb0, ahb1, ch0rx, ch1rx
> - of those, the first 3 are from dim2 itself, and the last two are from renesas-specific logic around dim2
> - in the interrupt assignment tables for gen3 SoCs, renesas documents all 5 interrupts, however in the
> mlb section, renesas mentions only mlb, ahb0 and ch0rx interrupts
> - moreover, renesas explicitly denies access dim2 registers responsible for channels 32..63 - which
> renders ahb1 interrupt useless; and renesas does not document any registers related to "async rx
> response" on channels 32..63 - which renders chrx1 interrupt useless
> - anyway, dim2 driver registers only 32 channels (for all use cases, not only for renesas), and thus
> uses only ahb0 interrupt
> - dim2 driver does not implement renesas-specific processing logic and thus does not use ch0rx interrupt
>
> I'm not sure how to proceed here.
> Is it better to define only two interrupts (mlb, ahb0) in device trees?
>
> Regarding 'interrupt-names' - dim2 driver currently uses platform_get_irq() and thus depends on numeric
> positions (mlb interrupt at index 0 and ahb0 interrupt at index 1). I'm not sure about current use cases
> of the driver other than with rcar-gen3, and if it is ok to use of_get_irq_byname() instead. And without
> using of_get_irq_byname(), interrupt-names looks somewhat useless.

As the driver is under staging, I think we can make any changes we want.

> > But without any DT binding documentation
> > for this hardware block, this is hard to validate, and not yet ready for
> > upstream integration.
>
> Christian, are you going to provide DT binding documentation for dim2?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds