On Thu, Aug 28, 2014 at 07:38:10AM +0800, Wang YanQing wrote: > On Tue, Aug 12, 2014 at 04:02:59PM +0200, Johan Hovold wrote: > > On Sat, Aug 09, 2014 at 02:46:56AM +0800, Wang YanQing wrote: > > > On Fri, Aug 08, 2014 at 09:54:42AM +0200, Johan Hovold wrote: > > > > On Fri, Aug 08, 2014 at 03:10:34AM +0800, Wang YanQing wrote: > > > " > > > You must allocate the buffer dynamically as some platforms cannot do > > > DMA to the stack. > > > " > > > Thanks very much for point out it, could you clarify it? > > > I want to know the reason. > > > > The memory where the stack resides might not be available for DMA, and > > even if it is, there could still be problems with cache coherency. > > It is still vague: > stack memory maybe resident higher place than normal memory, > but I don't think kmalloc could be immune from this problem, unless > we use GFP_DMA? No, you don't need to use GFP_DMA (unless implementing a driver for an ISA device on x86). Have a look at Documentation/DMA-API-HOWTO.txt, specifically the section "What memory is DMA'able?". Johan -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
