Re: pci-sysfs resource mmap broken (and PATCH)

[Benjamin Herrenschmidt]

On Thu, 2005-04-28 at 00:22 -0700, David S. Miller wrote:
> On Thu, 28 Apr 2005 17:21:19 +1000
> Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx> wrote:
> 
> > I have a real net big performance improvement on X by doing that
> > trick ... the sysfs mmap API doesn't really provide a mean to do
> > it explicitely from userland (unlike the ioctl with the old proc api)
> 
> You can refine your test to "if PCI class is display or VGA" and the
> prefetchability is set in the BAR, then elide the guard PTE
> protection bit.

Yes, but I was also thinking about some of those Myrinet kind of things
who also provide large PCI shared memory region ...

I may end up adding an explicit list of classes/vid/did that are
"allowed" to use the trick to avoid problems.

Ben.


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