On Tue, Mar 05, 2013 at 07:51:57PM +0530, Kishon Vijay Abraham I wrote:
> While creating the child devices, *of_platform_populate* sets only
> coherent_dma_mask but USBHCD sets *uses_dma* (determines whether the
> controller is DMA'able) based on dma_mask. So If we haven't explicitly set
> dma_mask, the HCD thinks the controller is not DMA'able and the
> controller will fail. So explicity set dma_mask for dwc3 after
> *of_platform_populate*.
>
> Signed-off-by: Kishon Vijay Abraham I <kishon@xxxxxx>
> ---
> drivers/usb/dwc3/dwc3-omap.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
> index 35b9673..19c6e72 100644
> --- a/drivers/usb/dwc3/dwc3-omap.c
> +++ b/drivers/usb/dwc3/dwc3-omap.c
> @@ -277,6 +277,14 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
> dwc3_omap_writel(omap->base, USBOTGSS_IRQENABLE_SET_0, 0x00);
> }
>
> +static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
> +
> +static int dwc3_omap_set_dmamask(struct device *dev, void *c)
> +{
> + dev->dma_mask = &dwc3_omap_dma_mask;
> + return 0;
> +}
> +
> static int dwc3_omap_probe(struct platform_device *pdev)
> {
> struct device_node *node = pdev->dev.of_node;
> @@ -381,6 +389,8 @@ static int dwc3_omap_probe(struct platform_device *pdev)
> return ret;
> }
>
> + device_for_each_child(&pdev->dev, NULL, dwc3_omap_set_dmamask);
I don't think this is correct. Child should be setting its own by
copying from parent. In summary I think we should add this to dwc3 core:
dma_set_coherent_mask(&pdev->dev, pdev->dev.parent->coherent_dma_mask);
pdev->dev.dma_mask = pdev->dev.parent->dma_mask;
pdev->dev.dma_parms = pdev->dev.parent->dma_parms;
And add the same thing to xhci-plat.c btw.
--
balbi
Attachment:
signature.asc
Description: Digital signature
