Hi,
Grygorii Strashko <grygorii.strashko@xxxxxx> writes:
> Felipe, Seems the problem might be deeper than on first look :(
>
> Lets see what grep says below. Even if we descope all SoC's drivers,
> there still will be few USB core components which manipulate with DMA parameters:
>
> ---
> ./drivers/usb/core/hcd.c: hcd->self.uses_dma = (dev->dma_mask != NULL);
> usb_create_shared_hcd()
> hcd->self.uses_dma = (dev->dma_mask != NULL);
>
> ---
> ./drivers/usb/core/usb.c: dev->dev.dma_mask = bus->controller->dma_mask;
> usb_alloc_dev()
> dev->dev.dma_mask = bus->controller->dma_mask;
>
> ---
> ./drivers/usb/core/message.c: intf->dev.dma_mask = dev->dev.dma_mask;
> usb_set_configuration()
> intf->dev.dma_mask = dev->dev.dma_mask;
>
> ^ why it is here is total mystery :(
interfaces are devices in their own right. A USB device might be
composite, meaning that it has several "devices" (functions or classes
might be a better wording) bundled into a single device/package.
> ---
> ./drivers/usb/dwc3/core.c: dev->dma_mask = dev->parent->dma_mask;
>
> ---
> ./drivers/usb/dwc2/hcd.c: /* Check if the bus driver or platform code has setup a dma_mask */
> dwc2_hcd_init()
> dma_set_mask(hsotg->dev, DMA_BIT_MASK(32)) < 0)
>
> ---
> ./drivers/usb/host/xhci.c:
> xhci_gen_setup()
> !dma_set_mask(dev, DMA_BIT_MASK(64)
> ./drivers/usb/host/xhci-plat.c: if (WARN_ON(!pdev->dev.dma_mask))
>
> ---
> ./drivers/usb/musb/musb_core.c: if (use_dma && dev->dma_mask) {
>
> ---
> ./drivers/usb/storage/scsiglue.c: if (!us->pusb_dev->bus->controller->dma_mask)
>
> It is big secret as for me (I'm USB noob:) which device is used for
> DMA transfers and when :(
heh, seems like we need to audit the usb subsystem (peripheral and host) :-)
--
balbi
Attachment:
signature.asc
Description: PGP signature
