On Thursday 08 January 2015 09:56:39 Arnd Bergmann wrote: > There is another interesting case, which is a USB host controller or > something similar behind a PCI bus. These are quite common and also > need to be handled in some form. Let's do just PCI first for now, but > be aware that this will come next. Should we assume that the ID that > is required for a device is either known from the device node, or > that it comes from a PCI device? That means for the USB case, we will > likely need to have some custom logic. There seems to be an implicit > assumption all over the kernel that all devices have the same IOMMU > instance, but we can't really rely on that here. > Update: I've checked the USB implementation and we are lucky because these interactions are all done by the USB host controller, and a USB device is not itself DMA capable. We don't have to do anything for USB. Arnd -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
