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

[Rafael J. Wysocki]

On Monday 23 February 2009, Eric Anholt wrote:
> On Mon, 2009-02-23 at 15:46 +0100, Maciej Rutecki wrote:
> > 2.6.29-rc5 and 2.6.29-rc5-git6 work OK. In 2.6.29-rc6 XWindow (KDM)
> > doesn't start, from dmesg:
> > 
> > [   18.170614] [drm] Initialized drm 1.1.0 20060810
> > [   18.247445] pci 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
> > [   18.247455] pci 0000:00:02.0: setting latency timer to 64
> > [   18.249610] mtrr: type mismatch for d0000000,10000000 old:
> > write-back new: write-combining
> > [   18.249621] [drm] MTRR allocation failed
> > [   18.249622] .  Graphics performance may suffer.
> > [   18.249658] pci 0000:00:02.0: irq 27 for MSI/MSI-X
> > [   18.249691] [drm] Initialized i915 1.6.0 20080730 on minor 0
> > [   18.290152] mtrr: type mismatch for d0000000,10000000 old:
> > write-back new: write-combining
> > [   18.567435] set status page addr 0x00043000
> > [   18.567463] [drm:i915_set_status_page] *ERROR* can not ioremap
> > virtual address for G33 hw status page
> 
> The non-GEM G33 HWS support is wrong, and has been more or less broken
> since inception (only now it's failing at startup instead of just being
> slow).  It's doing a mapping of the framebuffer when it should be using
> a cached mapping of the page backing the framebuffer address.
> 
> For a quick fix, converting to ioremap_wc will shut the kernel up and
> get us the old ioremap behavior (whatever the MTRR said, which happens
> to be WC).

This is a serious regression on this particular hardware, though.  We need it
fixed before 2.6.29-final, this way or another.

Thanks,
Rafael
--
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/