Radeon r6xx with KMS not working

[Mike Lothian]

2009/9/11 Mike Lothian <mike at fireburn.co.uk>:
> 2009/9/9 Alex Deucher <alexdeucher at gmail.com>:
>> On Tue, Sep 8, 2009 at 8:20 PM, Mike Lothian<mike at fireburn.co.uk> wrote:
>>> 2009/9/9 Alex Deucher <alexdeucher at gmail.com>:
>>>> On Tue, Sep 8, 2009 at 7:17 PM, Mike Lothian<mike at fireburn.co.uk> wrote:
>>>>> 2009/9/8 Mike Lothian <mike at fireburn.co.uk>:
>>>>>> 2009/9/8 Alex Deucher <alexdeucher at gmail.com>:
>>>>>>> On Tue, Sep 8, 2009 at 6:34 PM, Mike Lothian<mike at fireburn.co.uk> wrote:
>>>>>>>> Hi
>>>>>>>>
>>>>>>>> When using the new drm-next branch KMS fails (see attached logs) with
>>>>>>>> radeon.modeset=0 passed the computer will hard lockup when KDM is
>>>>>>>> starting (I think this is when the compositing is started)
>>>>>>>
>>>>>>> [drm:r600_ib_test] *ERROR* radeon: fence wait failed (-16).
>>>>>>> [drm:rv770_init] *ERROR* radeon: failled testing IB (-16).
>>>>>>>
>>>>>>> Pull the latest from the drm-next branch.
>>>>>>>
>>>>>>> Alex
>>>>>>>
>>>>>>
>>>>>> Just noticed that after Perry pointed out the fix on Phoronix
>>>>>>
>>>>>> Thanks
>>>>>>
>>>>>> Will let you know how I get on
>>>>>>
>>>>>
>>>>> The good news is that KMS now works, the bad news is KDM still wont
>>>>> start with compositing. Plain X works and glxgears work as does KDM
>>>>> with compositing off
>>>>>
>>>>> Glxgears says:
>>>>> IRQ's not enabled, falling back to busy waits: 2 1
>>>>> do_wait: drmWaitVBlank returned -1, IRQs don't seem to be working correctly.
>>>>>
>>>>> My dmesg says:
>>>>>
>>>>> Unpin not necessary for ffff88007c9a1600 !
>>>>> Unpin not necessary for ffff88006eda6400 !
>>>>> Unpin not necessary for ffff88006b14d400 !
>>>>>
>>>>> I couldn't see anything in my Xorg.0.log file that would explain why
>>>>> KDM keeps rebooting
>>>>>
>>>>> Is there anything else that might be useful?
>>>>>
>>>>
>>>> kde GL compositing doesn't work that well at the moment even without KMS.
>>>>
>>>> Alex
>>>>
>>>
>>> I'm reporting a regression from the r6xx-r7xx-3d tree, the new code
>>> causes X to restart
>>>
>>
>> It's not a regression as KMS did not work until recently;  lots of
>> stuff has problems in dri2 mode with the current r600 3d driver.  Or
>> are you saying that the drm-next code with KMS disabled is causing a
>> problem?
>>
>> Alex
>>
>
> Might I just add that this is repeated constantly in my /var/log/messages
>
> Sep 11 03:43:03 quark kernel: [drm:radeon_ib_get] *ERROR* radeon:
> IB(0:0x0000000040101000:673)
> Sep 11 03:43:03 quark kernel: [drm:radeon_ib_get] *ERROR* radeon: GPU
> lockup detected, fail to get a IB
> Sep 11 03:43:03 quark kernel: [drm:radeon_cs_ioctl] *ERROR* Failed to get ib !
>

With the latest kernel update I know get this in my dmesg repetitively
running glxgears

[drm:radeon_ring_write] *ERROR* radeon: writting more dword to ring
than expected !