[Bug 88430] New: [i915] from 3.18: 3D Acceleration fails for Intel Ironlake M graphics hardware

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=88430

            Bug ID: 88430
           Summary: [i915] from 3.18: 3D Acceleration fails for Intel
                    Ironlake M graphics hardware
           Product: DRI
           Version: unspecified
          Hardware: Other
                OS: Linux (All)
            Status: NEW
          Severity: normal
          Priority: medium
         Component: DRM/Intel
          Assignee: intel-gfx-bugs at lists.freedesktop.org
          Reporter: vivek at collabora.co.uk
        QA Contact: intel-gfx-bugs at lists.freedesktop.org
                CC: intel-gfx-bugs at lists.freedesktop.org

While working on the latest i915 backport for
https://01.org/linuxgraphics/ (porting the i915 module from 3.18
to the ubuntu utopic kernel) we may have found a problem with the
Ironlake M support:

The symptom is that on the affected hardware we get no 3D
acceleration... in fact something like glxgears or one of the GL
screensavers simply displays a static image.

On our test machine we see this in the xorg log:

[27.437] (II) GLX: Initialized DRI2 GL provider for screen 0
[27.445] (EE) intel(0): Failed to submit rendering commands,
              disabling acceleration.

which appears to be from:

./src/sna/kgem.c:3662:
  "Failed to submit rendering commands, disabling acceleration.\n");

Tracking that back, it looks like a call to do_execbuf is failing.

With drm.debug=0x3, we see this interesting difference in the syslog:

On both kernels:

kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1, I915_GEM_EXECBUFFER2
kernel: [] [drm:i915_gem_execbuffer2] copy 1 exec entries failed 56
kernel: [] [drm:drm_ioctl] ret = -14

Then a little later on the _working_ kernel only:

kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1,
DRM_IOCTL_SET_CLIENT_CAP
kernel: [] [drm:drm_ioctl] ret = -22

A little later, on the working kernel:

kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1, I915_GEM_EXECBUFFER2
kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1, DRM_IOCTL_GEM_CLOSE

Compared to the one that doesn't work:

kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1, I915_GEM_EXECBUFFER2
kernel: [] [drm:drm_ioctl] ret = -16
kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1, I915_GEM_THROTTLE
kernel: [] [drm:drm_ioctl] pid=1550, dev=0xe200, auth=1, DRM_IOCTL_GEM_CLOSE

Which I believe corresponds to the failure that the xorg log is
complaining about.

We're going to build a vanilla 3.18 kernel and test it on the hardware
in question, but the backport was pretty uninvasive, and didn't touch
any hardware specific code paths that I can see, which is why we think 
it might be a regression in the Ironlake M support.

I'll add more info once we've confirmed whether or not it occurs on
a 3.18 kernel on this hardware.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are on the CC list for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20150115/1773ebd5/attachment.html>