Re: [ 04/16] drm/i915: correctly order the ring init sequence

[Herton Ronaldo Krzesinski]

On Thu, Aug 23, 2012 at 10:10:18AM +0200, Daniel Vetter wrote:
> On Thu, Aug 23, 2012 at 12:44 AM, Herton Ronaldo Krzesinski
> <herton.krzesinski@xxxxxxxxxxxxx> wrote:
> > Really sorry about this, but it was a hardware+bios setting problem
> > here. I finished investigating what was happening here, and is related to
> > the DRAM installed on this machine. The memory installed is DDR3-1333,
> > and on bios it was configured as that, but turns out the chipset (G41)
> > only supports up to DDR3-1066 (as on specs, although the motherboard
> > lists the 1333 as supported, but as "OC", and allows configuring this
> > way automatically or manually).
> >
> > After setting manually back to 1066, I stopped seeing all the weird
> > behaviour, like the writes to the start address don't trigger anymore
> > setting on the head offset. I'm sorry about that, and really it's a
> > brown paper bag time for me :(. The machine worked fine and passed
> > through memtest and never gave any problems, this was the first
> > time (seems only the video device didn't like the memory setting, and
> > the code change probably changed timing in the right way to pop up this
> > now, also never had graphics corruption or other noticeable problems).
> > I retested kernels that gave problems and they run fine now.
> 
> Well, at least we could clear this up.
> 
> Greg, please pick up the following patches for 3.0 (if you haven't already):
> 
> f01db988ef6f6c70a6cc36ee71e4a98a68901229 and
> 0d8957c8a90bbb5d34fab9a304459448a5131e06
> 
> Note that all kernels that need f01db backported also need
> b7884eb45ec98c0d34c7f49005ae9d4b4b4e38f6 (to fix a regression
> introduce by the former).

b7884eb45ec98c0d34c7f49005ae9d4b4b4e38f6 needs some backporting for 3.0,
I did it in advance, build tested and installed here on a sandybridge
based laptop on x86_64: