etnaviv-gpu 134000.gpu: MMU fault status 0x00000002 on i.XM6 Quad Plus

[Lucas Stach]

Am Dienstag, den 29.08.2017, 14:36 -0300 schrieb Fabio Estevam:
> On Mon, Aug 21, 2017 at 9:05 AM, Luís Mendes <luis.p.mendes at gmail.com> wrote:
> > Hi,
> >
> > I'm a software and hardware engineer and I would like to help fix this
> > issue. I'm however not acquainted with Open GL nor with Vivante GPUs, thus
> > some guidelines would be welcome.
> >
> > Problem details:
> > This issue appears when I login into Ubuntu Mate 17.04. I'm using kernel
> > 4.11.12 with Etnaviv drivers, libdrm 2.4.82 and mesa 17.2.0 RC4.
> > During the login multiple MMU faults occur and the GUI becomes corrupted...
> > the applications menu misses some words or characters, also dialog boxes
> > appearing thereafter are also corrupted, missing some words or characters.
> > The applications, including OpenGL accelerated ones are fine, rendering
> > without detectable errors or artifacts.
> >
> > The error only occurs with GC3000, not with GC2000...
> >
> > Here is the dmesg output of the error and system:
> > [    6.260838] etnaviv gpu-subsystem: bound 134000.gpu (ops gpu_ops
> > [etnaviv])
> > [    6.272870] etnaviv gpu-subsystem: bound 130000.gpu (ops gpu_ops
> > [etnaviv])
> > [    6.273303] etnaviv gpu-subsystem: bound 2204000.gpu (ops gpu_ops
> > [etnaviv])
> > [    6.273328] etnaviv-gpu 134000.gpu: model: GC320, revision: 5303
> > [    6.420783] systemd-fstab-generator[364]: Checking was requested for "/",
> > but it is not a device.
> > [    7.116977] etnaviv-gpu 130000.gpu: model: GC3000, revision: 5450
> > [    8.488007] etnaviv-gpu 2204000.gpu: model: GC355, revision: 1216
> > [    8.488035] etnaviv-gpu 2204000.gpu: Ignoring GPU with VG and FE2.0
> > [    8.497905] [drm] Initialized etnaviv 1.1.0 20151214 for gpu-subsystem on
> > minor 1
> >
> > ...
> >
> > [   47.628983] etnaviv-gpu 134000.gpu: MMU fault status 0x00000002
> > [   47.635159] etnaviv-gpu 134000.gpu: MMU 0 fault addr 0x0a03ffc0
> > [   47.641100] etnaviv-gpu 134000.gpu: MMU 1 fault addr 0x00000000
> > [   47.647034] etnaviv-gpu 134000.gpu: MMU 2 fault addr 0x00000000
> > [   47.652967] etnaviv-gpu 134000.gpu: MMU 3 fault addr 0x00000000
> > [   51.050237] etnaviv-gpu 134000.gpu: hangcheck detected gpu lockup!
> > [   51.056686] etnaviv-gpu 134000.gpu:      completed fence: 828
> > [   51.062592] etnaviv-gpu 134000.gpu:      active fence: 832
> > [   51.069158] etnaviv-gpu 134000.gpu: hangcheck recover!
> 
> Just wanted to confirm that I was able to reproduce this failure on a
> imx6qp sabresd board as well.
> 
> Luís shared his UbuntuMate rootfs and I successfully tested it on imx6q sabresd.
> 
> However on a imx6qp sabresd the Ubuntu login screen does not even
> start.  If I type the password then the MMU faults above are
> triggered.
> 
> Has anyone successfully used X11 on imx6qp with Etnaviv?

Yes, it's running fine for me, but my test-case is a stripped down
embedded X11, not some full-blown desktop distro.

Generally the way to debug this would be to disable individual
acceleration functions within the armada driver, forcing the software
fallback path to get a clue where things are going wrong.

Regards,
Lucas