[Bug 194761] New: amdgpu driver breaks on Oland (SI)

[bugzilla-daemon at bugzilla.kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=194761

            Bug ID: 194761
           Summary: amdgpu driver breaks on Oland (SI)
           Product: Drivers
           Version: 2.5
    Kernel Version: 4.10
          Hardware: x86-64
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: Video(DRI - non Intel)
          Assignee: drivers_video-dri at kernel-bugs.osdl.org
          Reporter: jdelvare at suse.de
        Regression: No

Created attachment 255045
  --> https://bugzilla.kernel.org/attachment.cgi?id=255045&action=edit
Checkerboard effect on glxgears

After upgrading my kernel from v4.9.11 to v4.10.0, all 3D applications are
broken. The most visible problem is a checkerboard effect, with blinking of
some areas. See the attached screenshot of glxgears for an example. The bug
only happens when I use the amdgpu driver (with CONFIG_DRM_AMDGPU_SI=y.) If I
use the radeon driver instead then all is fine.

There are no error messages in the kernel log, nor in /var/log/Xorg.0.log.
However I could spot differences between the kernel logs:

-Linux version 4.9.11-1-default (geeko at buildhost) (gcc version 6.3.1 20170202
[gcc-6-branch revision 245119] (SUSE Linux) ) #1 SMP PREEMPT Sat Feb 18
17:59:27 UTC 2017 (cf9c670)
+Linux version 4.10.1-1-default (geeko at buildhost) (gcc version 6.3.1 20170202
[gcc-6-branch revision 245119] (SUSE Linux) ) #1 SMP PREEMPT Sun Feb 26
12:43:10 UTC 2017 (1ecd5af)

 [drm] GPU post is not needed
 amdgpu 0000:01:00.0: VRAM: 1024M 0x0000000000000000 - 0x000000003FFFFFFF
(1024M used)
-amdgpu 0000:01:00.0: GTT: 3991M 0x0000000040000000 - 0x00000001397847FF
+amdgpu 0000:01:00.0: GTT: 1024M 0x0000000040000000 - 0x000000007FFFFFFF
 [drm] Detected VRAM RAM=1024M, BAR=256M
 [drm] RAM width 64bits DDR3
 [drm] amdgpu: 1024M of VRAM memory ready
-[drm] amdgpu: 3991M of GTT memory ready.
-[drm] GART: num cpu pages 1021828, num gpu pages 1021828
-amdgpu 0000:01:00.0: PCIE GART of 3991M enabled (table at 0x0000000000040000).
+[drm] amdgpu: 1024M of GTT memory ready.
+[drm] GART: num cpu pages 262144, num gpu pages 262144
+amdgpu 0000:01:00.0: PCIE GART of 1024M enabled (table at 0x0000000000040000).

-[drm] Initialized amdgpu 3.8.0 20150101 for 0000:01:00.0 on minor 0
+[drm] Initialized amdgpu 3.9.0 20150101 for 0000:01:00.0 on minor 0

So something changed in the detection of the available memory. I don't know if
this is related to the bug. As a comparison point, the radeon driver disagrees
with both the old amdgpu driver and the new amdgpu driver:

radeon 0000:01:00.0: VRAM: 1024M 0x0000000000000000 - 0x000000003FFFFFFF (1024M
used)
radeon 0000:01:00.0: GTT: 2048M 0x0000000040000000 - 0x00000000BFFFFFFF
[drm] Detected VRAM RAM=1024M, BAR=256M
[drm] RAM width 64bits DDR
[drm] radeon: 1024M of VRAM memory ready
[drm] radeon: 2048M of GTT memory ready.

-- 
You are receiving this mail because:
You are watching the assignee of the bug.