Re: kmalloc

Matti Aarnio (matti.aarnio@sonera.fi)
Mon, 25 Jan 1999 22:49:58 +0200 (EET)


Alan Cox replied to some questions:
>
> > Is kmalloc guaranteed to alloc a physically contigous buffer? My exact
> Yes
> > problem is: Should I use GFP_DMA when allocating buffers for PCI DMA?
> No

Given that my 64-bit machines are often equiped with memory amounts
counted in gigabytes; sometimes exceeding 4 GB, and as the 32-bit PCI
devices can address "only" up to 4G, there is a problem.
(My largest 64-bit Linux box has 0.5G memory.)

I think we will be needing GFP_DMA32 soon.
(And it will cause *massive* amounts of changes as practically all
drivers will need touching...)

DaveM can tell us how the large SPARC64 systems do it with their heaps
of busses and address spaces. I recall SPARCs have IO-MMU, which may
help in this regard. Large-scale DEC systems have similar gimmics
in their IO controllers mapping 32-bit IO busses to 43-bit cpu/memory
interconnect busses (43 address bits, that is).

/Matti Aarnio <matti.aarnio@sonera.fi>

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/