4.1-rc7: Xorg broken after resume on thinkpad T40p, radeon problem?

[Paul Bolle]

On Wed, 2015-06-17 at 15:21 +0200, Pavel Machek wrote:
> On Wed 2015-06-17 15:07:41, Christian König wrote:
> > On 17.06.2015 14:49, Pavel Machek wrote:
> > >On Wed 2015-06-17 14:43:47, Pavel Machek wrote:
> > >>Hi!
> > >>
> > >>Video is broken on thinkpad T40p after resume from suspend. (In X
> > >>only, text console still works.)
> > >>
> > >>Maybe this is related?
> > >>
> > >>[ 5430.193071] radeon 0000:01:00.0: putting AGP V2 device into 4x mode
> > >>[ 5430.193107] radeon 0000:01:00.0: GTT: 256M 0xD0000000 - 0xDFFFFFFF
> > >>[ 5430.214670] radeon 0000:01:00.0: WB disabled
> > >>[ 5430.214675] radeon 0000:01:00.0: fence driver on ring 0 use gpu addr 0x00000000d0000000 and cpu addr 0xf841a000
> > >>[ 5430.214721] [drm] radeon: ring at 0x00000000D0001000
> > >>[ 5430.370867] [drm:r100_ring_test] *ERROR* radeon: ring test failed (scratch(0x15E8)=0xCAFEDEAD)
> > >>[ 5430.370869] [drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
> > >>[ 5430.370872] radeon 0000:01:00.0: failed initializing CP (-22).
> > >>
> > >>Hibernation works well here, including X. (Which has small glitch with
> > >>mouse cursor being corrupted until it is changed by application).
> > >>
> > >3.16.0-4-686-pae Debian 3.16.7-cktt11-1 kernel seems to have same
> > >problem, with same scratch(0x15E8)=0xCAFEDEAD message... so this is
> > >not "new" problem...
> > >
> 
> > Looks like CP init doesn't work. Well is there any kernel where that did
> > worked?
> 
> Unfortunately, not as far as I know.
> 
> > If yes it might be a good idea to bisect to narrow down the problem.
> 
> No such luck. I may try something like "3.0" if we are really
> desperate (2.6.X kernels probably won't won't boot with recent
> userland), but I suspect it just never worked.

The above looks very much like the issue that made me write commit
45171002b01b ("radeon: add AGPMode 1 quirk for RV250"). See
https://bugzilla.redhat.com/show_bug.cgi?id=531825 for a lot of
background.

Does booting with radeon.agpmode=1 survive a suspend and resume cycle?

Hope this helps,


Paul Bolle