Re: [Linux 2.6.29-rc6] [drm:i915_set_status_page] *ERROR* can not ioremap virtual address for G33 hw status page

[Dave Airlie <airlied@redhat.com>]

On Tue, 2009-02-24 at 22:39 +0100, Ingo Molnar wrote:
> * Yinghai Lu <yinghai@kernel.org> wrote:
> 
> > Ingo Molnar wrote:
> > > * Yinghai Lu <yinghai@kernel.org> wrote:
> > > 
> > >> On Mon, Feb 23, 2009 at 7:23 PM, Andy Isaacson <adi@hexapodia.org> wrote:
> > >>> On Mon, Feb 23, 2009 at 05:01:10PM -0800, Yinghai Lu wrote:
> > >>>> CONFIG_MTRR=y
> > >>>> CONFIG_MTRR_SANITIZER=y
> > >>>> CONFIG_MTRR_SANITIZER_ENABLE_DEFAULT=1
> > >>>> CONFIG_MTRR_SANITIZER_SPARE_REG_NR_DEFAULT=1
> > >>>>
> > >>>> should help mtrr ones.
> > >>>>
> > >>>> please post bootlog with those option set.
> > >>> Doesn't help here (Dell E6400).
> > >>>
> > >>> [    0.000000] Initializing cgroup subsys cpuset
> > >>> [    0.000000] Initializing cgroup subsys cpu
> > >>> [    0.000000] Linux version 2.6.29-rc6-dirty (adi@cvp-loaner-01) (gcc version 4.3.2 (Ubuntu 4.3.2-1ubuntu12) ) #14 SMP Mon Feb 23 19:05:31 PST 2009
> > >>> [    0.000000] Command line: root=/dev/sda1 ro
> > >>> [    0.000000] KERNEL supported cpus:
> > >>> [    0.000000]   Intel GenuineIntel
> > >>> [    0.000000]   AMD AuthenticAMD
> > >>> [    0.000000]   Centaur CentaurHauls
> > >>> [    0.000000] BIOS-provided physical RAM map:
> > >>> [    0.000000]  BIOS-e820: 0000000000000000 - 000000000009f000 (usable)
> > >>> [    0.000000]  BIOS-e820: 000000000009f000 - 00000000000a0000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 0000000000100000 - 000000007d04d400 (usable)
> > >>> [    0.000000]  BIOS-e820: 000000007d04d400 - 000000007d04f400 (ACPI NVS)
> > >>> [    0.000000]  BIOS-e820: 000000007d04f400 - 0000000080000000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000f8000000 - 00000000fc000000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000fec00000 - 00000000fec10000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000fed18000 - 00000000fed1c000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000fed20000 - 00000000fed90000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000feda0000 - 00000000feda6000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000fee00000 - 00000000fee10000 (reserved)
> > >>> [    0.000000]  BIOS-e820: 00000000ffe60000 - 0000000100000000 (reserved)
> > >>> [    0.000000] DMI 2.4 present.
> > >>> [    0.000000] last_pfn = 0x7d04d max_arch_pfn = 0x100000000
> > >>> [    0.000000] x86 PAT enabled: cpu 0, old 0x7040600070406, new 0x7010600070106
> > >>> [    0.000000] original variable MTRRs
> > >>> [    0.000000] reg 0, base: 0GB, range: 32GB, type WB
> > >>> [    0.000000] reg 1, base: 3584MB, range: 512MB, type UC
> > >>> [    0.000000] reg 2, base: 2012MB, range: 4MB, type UC
> > >>> [    0.000000] reg 3, base: 2016MB, range: 32MB, type UC
> > >> the BIOS is so sick
> > >> according to MTRR, it said:
> > >> [0,2012M) is WB
> > >> [2048M, 3.5G) is WB too
> > >> [4G, 32G) is WB
> > >>
> > >> but according to e820: about [0,2g) is RAM...
> > >>
> > >> really not how to workaround in BIOS.
> > > 
> > > I suspect the box was tested with other OSs and limped along 
> > > there. Should we perhaps clear non-sensical MTRR entries?
> > > 
> > 
> > according to e820 to update MTRR?
> > 
> > in the previous discussion, some SMI region that is not stated 
> > in e820 could still need to be covered by MTRR with WB
> 
> Indeed. Nasty. Those tend to be in pretty specific places 
> though, typically next to RAM. (they are taken off RAM usually) 
> So if we leave the boundary around real e820 RAM alone, we could 
> zap the other bogosities perhaps?
> 
> Dunno, removing MTRRs does indeed sound somewhat dangerous. A WB 
> MTRR region could also be defined for some PCI card.
> 
> The other question is, why did the ioremap_wc() fail? Why doesnt 
> it fall back to UC transparently?

I'm sort of worried the drivers are going to end up doing, which at
the moment is the only option that will work.

blah = ioremap_wc();
if (!blah)
	blah = ioremap_nocache();

which seems pointless so this is a good point, where is the fallback?


Dave.
> 
> 	Ingo