Video PM Methods

[Peter Jones]

On Wed, 2006-05-17 at 11:23 +0100, Matthew Garrett wrote:
> On Tue, May 16, 2006 at 03:15:48PM -0400, Peter Jones wrote:
> 
> > Are you sure you really want this, rather than pci
> > vendor:device:subvendor:subdevice for the video card?  BenH was telling
> > me the other day that he's had good experience moving to the pci IDs
> > rather than the DMI system-type field for the whitelisting of pm stuff
> > in the radeonfb code.
> 
> If you're going to work at the subvendor:subdevice level then you're 
> basically at the DMI level anyway[1]. The bits of video BIOS code that 
> interact with the hardware in more "interesting" ways tend to be written 
> by the hardware developers rather than the chipset developers - Intel 
> provide their BIOSes with some hooks for machine-specific code.

*nod*.  The biggest advantage of using pci ids is that they're very easy
to discover without having to e.g. read /dev/mem .  I think that's a
*big* win.

> [1] Except with less support for machines from vendors who haven't 
> bothered to distinguish some of their platforms at the PCI ID 
> level

Well, that's not really true.  Most of the machine where they haven't
done that, the DMI data is also pretty much junk.  In many cases you can
also allow for some basic wildcards; see my other post.  And actually,
I'm about to update the code in the other post with a possible answer
for these cases, though it's not as pretty as I'd like.

-- 
  Peter