> discussing non-PCI OHCI patches with me have reported
> DMA limitations like "must be in first MByte".  (I think that
> was an ARM CPU, not MIPS or PPC.)

StrongARM SA1100

> If some code wants to take vmalloc memory and do I/O with it,
> the solution gets complicated.  In principle one must pin pages
> and pass their "real" addreses to an I/O layer.  That gets into
> the VM and block I/O layers ... less stable than USB!  :)

vmalloc memory pages might also need copying before the DMA

_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to