[openib-general] [PATCH] mthca: fix registration for giant MRs

[Grant Grundler]

On Wed, Jun 01, 2005 at 04:03:16PM -0700, Roland Dreier wrote:
...
> On ppc64 at least, dma_map_sg() will return 0 if the mapping request
> fails (even if it fails in the middle).  Are other architectures
> different?

Yes. "0" is a valid DMA address.
no portable PCI driver should be expecting 0 for a failure.
ppc64 can return anything it likes. Many arches (including parisc,
alpha, sparc) would just panic when the DMA pool was exhausted.

I've had this discussion several years ago:

| Resent-Date: Fri, 10 Nov 2000 04:39:25 -0700
| From: "David S. Miller" <davem at redhat.com>
...
| Subject: Re: [parisc-linux] tulip DMA mapping
...
|      I've quoted the whole of Grants message below, so you can see the
|      context.  It looks like tulip is treating zero as meaning it
|      doesn't have anything to pci_unmap...

| Thank you for the clarification.
|
| Jeff, this is in fact a bug, please fix :-)

Original problem report is here:
http://lists.parisc-linux.org/pipermail/parisc-linux/2000-November/009850.html

hth,
grant