Re: [PATCH] spi: pic32: Retire dma_request_slave_channel_compat()

[Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>]

Hi Peter,

On Thu, Nov 21, 2019 at 9:40 AM Peter Ujfalusi <peter.ujfalusi@xxxxxx> wrote:
> On 21/11/2019 10.30, Geert Uytterhoeven wrote:
> > On Thu, Nov 21, 2019 at 9:11 AM Peter Ujfalusi <peter.ujfalusi@xxxxxx> wrote:
> >> There is no reason to use the dma_request_slave_channel_compat() as no
> >> filter function and parameter is provided.
> >>
> >> Switch the driver to use dma_request_chan() instead.
> >>
> >> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@xxxxxx>
> >
> >> --- a/drivers/spi/spi-pic32.c
> >> +++ b/drivers/spi/spi-pic32.c
> >> @@ -609,22 +609,18 @@ static void pic32_spi_cleanup(struct spi_device *spi)
> >>  static void pic32_spi_dma_prep(struct pic32_spi *pic32s, struct device *dev)
> >>  {
> >>         struct spi_master *master = pic32s->master;
> >> -       dma_cap_mask_t mask;
> >>
> >> -       dma_cap_zero(mask);
> >> -       dma_cap_set(DMA_SLAVE, mask);
> >> -
> >> -       master->dma_rx = dma_request_slave_channel_compat(mask, NULL, NULL,
> >> -                                                         dev, "spi-rx");
> >> -       if (!master->dma_rx) {
> >> +       master->dma_rx = dma_request_chan(dev, "spi-rx");
> >
> > Why not dma_request_slave_channel()?
>
> The longer term plan is to retire dma_request_slave_channel() as well.
> With dma_request_chan() deferred probing against DMA drivers is possible
> and it also supports legacy boot with dma_slave_map.
>
> At the end we should be left with only dma_request_chan() for slave
> channels in the kernel.

Thank you, deferred probing is a valid argument.
You may want to add that to the patch description.

Reviewed-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>

> > That way you...
> >
> >> +       if (IS_ERR(master->dma_rx)) {
> >
> > ... don't have to change the NULL check here, and...
> >
> >>                 dev_warn(dev, "RX channel not found.\n");
> >> +               master->dma_rx = NULL;
> >
> > ...  don't have to override by NULL here.
>
> It is a small sacrifice, true, but if anyone cares the driver can
> support deferred probing with dma_request_chan().

Deferred probing in case of missing DMA controllers is always a bit
tricky, as there are multiple options:
  1. Defer probing of the slave driver,
  2a. Continue probing of the slave driver, fall back to PIO (which is
      what this driver does)
  2b. and Retry DMA setup later.

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