Am Mittwoch, 6. August 2003 10:33 schrieb Petko Manolov:
> On Wed, 6 Aug 2003, Oliver Neukum wrote:
>
> > Am Mittwoch, 6. August 2003 08:24 schrieb Petko Manolov:
> > > On Wed, 6 Aug 2003, Oliver Neukum wrote:
> > >
> > > > GFP_DMA has no place in USB drivers, as its meaning is inconsistent
> > > > across architectures.
> > >
> > > The patch looks ok to me, although GFP_DMA used to mean that the allocated
> > > memory will be contiguous and taken from the dma-able memory. If this is
> > > no longer needed then it's better if GFP_DMA go away.
> >
> > GFP_DMA means capable of DMA on an ISA bus.
>
> Well, not necessarily. I used to work with a MIPS based SOC where the
> only dma-able memory was sram which was on the same die. It had nothing
> to do with ISA or any other bus.
> Anyway, GFP_DMA was doing the right thing there.
Well, we can't use a flag with inconsistent meaning across architectures.
kmalloc with GFP_KERNEL is defined as DMAable by generic hardware.
If you want to care about strange chips which would need bounce buffers
you must use usb_buffer_alloc(). GFP_DMA has no place at all for USB
device drivers.
Regards
Oliver
-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
