Alan Cox wrote:
> > For ISA and EISA like devices, NULL should be passed into
> > pci_alloc_consistent(). I know, we don't even get the
> > interfaces available when !CONFIG_PCI, that is something
> > to fix.
>
> What about devices where ISA space is non coherent. They exist
The other thing that I think needs fixing is that passing NULL is
that you lose information. I think we should define something
like:
#define ISA_DMA ((struct pci_dev *) NULL)
for now, but later we could have something more like:
struct pci_dev { dma_mask: 0xFFFFFF; } __isa_dma;
#define ISA_DMA (&__isa_dma)
The reason is that the new mechanism supports a nice way to
control what part of memory (from the bus's perspective)
via the "pci_dev.dma_mask" value, which is an intellegent
replacement for GFP_DMA. However, if we have multiple
"pci like" busses without per-device instances of
"struct pci_dev" then the current API wouldn't allow these
busses to have different memory needs.
-Mitch
-
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/
This archive was generated by hypermail 2b29 : Mon Feb 07 2000 - 21:00:12 EST