Re: [PATCH] dw_dmac: apply default dma_mask if needed

[Arnd Bergmann]

On Tuesday 29 January 2013, Viresh Kumar wrote:
> Adding Arnd in cc.
> 
> On 29 January 2013 20:36, Andy Shevchenko
> <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> > In some cases we got the device without dma_mask configured. We have to apply
> > the default value to avoid crashes during memory mapping.
> >
> > Signed-off-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> > ---
> >  drivers/dma/dw_dmac.c |    6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/drivers/dma/dw_dmac.c b/drivers/dma/dw_dmac.c
> > index e8d0679..a572a1e 100644
> > --- a/drivers/dma/dw_dmac.c
> > +++ b/drivers/dma/dw_dmac.c
> > @@ -1673,6 +1673,12 @@ static int dw_probe(struct platform_device *pdev)
> >         if (IS_ERR(regs))
> >                 return PTR_ERR(regs);
> >
> > +       /* Apply default dma_mask if needed */
> > +       if (!pdev->dev.dma_mask) {
> > +               pdev->dev.dma_mask = &pdev->dev.coherent_dma_mask;
> > +               pdev->dev.coherent_dma_mask = DMA_BIT_MASK(32);
> > +       }
> > +
> 
> Hmm... Why is it required for a DMA controller? What kind of crash do you
> get?

A valid dma mask is required for any device that is passed into dma_map_*
and other dma-mapping.h interfaces. The question is more about is
responsible for setting up the mask. Traditionally we'd do that from the
platform definition in the place where the platform_device is created,
but this now comes from the device tree, which is a bit inconsistent
with the DMA masks at the moment.

We do set the dev.coherent_dma_mask to DMA_BIT_MASK(32) for all platform
devices instatiated from DT, but I cannot find the code that sets the
dma_mask for a platform_device. For an amba_device, we set both to
the same value. Maybe Grant can remember if the difference here is
intentional, or if the dma_mask is set up elsewhere for a platform_device
coming from DT.

> @Arnd: Is this change recommended?

For all I can tell, it cannot hurt to do it this way.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/