[PATCH] agp: use scratch page on memory remove and at GATT creation V4

[Jerome Glisse]

On Thu, Apr 22, 2010 at 11:27:11AM +0200, Daniel Vetter wrote:
> On Tue, Apr 20, 2010 at 05:43:34PM +0200, Jerome Glisse wrote:
> > Convert most AGP chipset to use scratch page as default entries.
> > This help avoiding GPU querying 0 address and trigger computer
> > fault. With KMS and memory manager we bind/unbind AGP memory
> > constantly and it seems that some GPU are still doing AGP
> > traffic even after GPU report being idle with the memory segment.
> > 
> > Tested (radeon GPU KMS + Xorg + compiz + glxgears + quake3) on :
> > - SIS 1039:0001 & 1039:0003
> > - Intel 865 8086:2571
> > 
> > Compile tested for other bridges
> > 
> > V2 enable scratch page on uninorth
> > V3 fix unbound check in uninorth insert memory (Michel Dänzer)
> > V4 rebase on top of drm-next branch with the lastest intel AGP
> >    changeset (stable should use version V3 of the patch)
> 
> Nope, v3 still contains the bogus changes to the intel gtt driver (only
> used by intel igds). In this patch, the intel parts look good.
> 
> While looking add this I've found some more stuff to nit-pick over ;)
> Instead of splattering needs_scratch_page = true all over the agp drivers,
> why not do the changes in the agp core (and the few fixups required in the
> drivers) and simply kill this variable? If using a scratch page is
> required by upper layers (drm/radeon), then keeping around this "looks
> optional, but is very much a core requirement" thing lingering around is
> quite a call for trouble, IMHO.
> 
> Yours, Daniel
> -- 
> Daniel Vetter
> Mail: daniel at ffwll.ch
> Mobile: +41 (0)79 365 57 48

V4 is ok it apply on top of drm-next, and no i don't kill
needs_scratch_page because i don't want to touch Alpha AGP
code (i think it should die, i don't think alpha got AGP
hw we care about or even support).

Cheers,
Jerome