Re: [PATCH] spi: rspi: Fix DMA mode when the DMA controller is built as a module

[Geert Uytterhoeven]

Hi Ovidiu,

On Wed, 23 Sept 2026 at 13:07, Ovidiu Panait
<ovidiu.panait.rb@xxxxxxxxxxx> wrote:
> > On Mon, 21 Sept 2026 at 22:00, Ovidiu Panait
> > <ovidiu.panait.rb@xxxxxxxxxxx> wrote:
> > > On the Renesas RZ/G2L platform, the RSPI driver falls back to PIO mode
> > > instead of using DMA when the DMA controller driver is built as a module
> > > (CONFIG_RZ_DMAC=m):
> > >
> > >     renesas_spi 1004b000.spi: dma_request_slave_channel_compat failed
> > >     renesas_spi 1004b000.spi: DMA not available, using PIO
> > >
> > > The driver requests DMA channels through
> > dma_request_slave_channel_compat()
> > > which returns either a valid pointer or NULL and cannot handle
> > > -EPROBE_DEFER. If the DMA controller has not probed yet, the driver
> > falls
> > > back to PIO directly.
> > >
> > > Fix this by replacing the dma_request_slave_channel_compat() call with a
> > > regular dma_request_chan() call. dma_request_slave_channel_compat() is
> > > always passed a NULL fn_param in all code paths, so it actually behaves
> > > like a plain dma_request_chan() call.
> > >
> > > Fixes: e825b8dd2b36 ("spi: rspi: Add DT support to DMA setup")
> > > Signed-off-by: Ovidiu Panait <ovidiu.panait.rb@xxxxxxxxxxx>
> >
> > Thanks for your patch!
> >
> > > --- a/drivers/spi/spi-rspi.c
> > > +++ b/drivers/spi/spi-rspi.c
> > > @@ -1355,6 +1339,11 @@ static int rspi_probe(struct platform_device
> > *pdev)
> > >         }
> > >
> > >         ret = rspi_request_dma(&pdev->dev, ctlr, res);
> > > +       if (ret == -EPROBE_DEFER) {
> > > +               dev_err_probe(&pdev->dev, ret,
> > > +                             "failed to request DMA channels\n");
> > > +               goto error2;
> >
> > Doesn't this make the probe fail if the DMA driver is not available
> > at all?
>
> Yes, if the "dmas" property is present but the DMA driver is not available
> at all, the probe will keep deferring. This matches other drivers I checked
> (spi-rzv2h-rspi, spi-rockchip, spi-bcm2835, etc).

IMHO that is bad, as it means the driver won't work at all if the DMAC
driver is not available...

> > > +       }
> > >         if (ret < 0)
> > >                 dev_warn(&pdev->dev, "DMA not available, using PIO\n");
> >
> > The driver still falls back to PIO in case of a missing "dmas"
> > property, right?
> >
>
> Yes, without "dmas", the driver still falls back to PIO.

But that's hardware description, so the "dmas" property will be
present anyway, regardless of the availability of the DMAC driver.

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