Bug#732692: loading radeonkms results in unusable screen

[Robert Millan <rmh@debian.org>]

On 07/01/2014 18:17, Steven Chamberlain wrote:
>>> It is most interesting that your HD 2400 / RV610 isn't falling back in
>>> the same way as described above.
>>
>> I haven't tried my patch on RV610 yet. I didn't look at the logs though, only
>> that output is deterministically garbled when firmware is not loaded.
> 
> Could you check if the modesetting sysctl is really registered or not
> if firmware isn't available?

It is.

Uhm, you're right. It seems that r600.c inits successfully without the firmware
blobs. All it does it disabling 3D acceleration.

I just copied this from the si.c driver. 'pfp' as a placeholder but enough to test.
I'll give it a try...

--- kfreebsd-10.orig/sys/dev/drm2/radeon/r600.c 2014-01-07 19:15:37.271947000 +0100
+++ kfreebsd-10/sys/dev/drm2/radeon/r600.c      2014-01-07 19:19:53.349946975 +0100
@@ -3012,6 +3012,15 @@
                rdev->accel_working = false;
        }
 
+       /* Don't start up if the ?? ucode is missing.
+        * The default clocks and voltages before the ?? ucode
+        * is loaded are not suffient for advanced operations.
+        */
+       if (!rdev->pfp_fw) {
+               DRM_ERROR("radeon: ?? ucode required for R600.\n");
+               return -EINVAL;
+       }
+
        return 0;
 }

-- 
Robert Millan