[Nouveau] [Bug 67315] System lockup with X consuming all CPU

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=67315

--- Comment #7 from Rick Stevens <ricks at alldigital.com> ---
On 07/26/2013 01:34 PM, bugzilla-daemon at freedesktop.org issued this missive:
> *Comment # 6 <https://bugs.freedesktop.org/show_bug.cgi?id=67315#c6> on
> bug 67315 <https://bugs.freedesktop.org/show_bug.cgi?id=67315> from Emil
> Velikov <mailto:emil.l.velikov at gmail.com> *
>
> Now this is interesting
>
> [19.600630]
> Initially the kernel "drm" seems to pass NULL pointer to the card, while doing
> nv50_fbcon_imageblit()
>
> ch 1 [0x0007cb0000 DRM] subc 3 class 0x502d mthd 0x0860 data 0x00000000
>
> The way I see it, after this point we're on the mercy of the hardware
>
> [19.627513]
> X/the ddx
> MP is still funny and fails to execute/set EDGEFLAG_ENABLE to 1 (which for the
> sake of me I cannot find in the ddx code?)
>
> [55.041877]
> At this point the GPU is completely stuffed fails to execute/set
> NV50_2D_BLIT_SRC_Y_INT, (coming from X).
>
> After that hell breaks loose :P

Emil, thanks for looking at this. It's, uhm, interesting, isn't it?

For clarity sake, this is a bit of an odd system. It's a Shuttle 
motherboard. From "dmidecode":

Handle 0x0002, DMI type 2, 8 bytes
Base Board Information
    Manufacturer: Shuttle Inc
    Product Name: FN78S
    Version: V10
    Serial Number:

> Rick a few interesting notes
> In the last attachment Xorg.log claims that it's starts at ~81.456, whereas
> dmesg states that X was running ~19.627541.
> * Are those logs matching (ie. captured from the same boot/system startup)
> I'm assuming that you've started your screensaver ~55.041877. Is that correct ?
> * Do you recall when did your nouveau started started reporting errors (nouveau
> E) ?

I booted the machine and as soon as the display locked up, I ssh'd to it
from another machine and simply did a "dmesg >/rick/dmesg.txt" to
capture as much as I could. The Xorg.0.log was already generated by the
time I logged in via ssh. I combined the various logs into one file and
shot it off to bugzilla.

As far as the screen saver, I'm running XFCE and the lockup occurred as
soon as I went through the "Applications Menu->Settings->Screensaver"
menu tree. It tried to render the Screensaver window, got as far as
drawing the box around it and everything locked up. After about a
minute, the graphical screen blanked and I started getting the GPU
lockup and going back to fbcon messages appeared on the console. The
GUI screen then reappeared but it was still locked up. Again it
cleared, the GPU lockup message appeared again and around and around we
go. It might do the same thing rendering other windows, but I'm trying
to be consistent to help debug this.

I'll be more than happy to keep tinkering, but I've had like 3 hours
of sleep in the last two days and I'm a bit knackered. I can pick this
back up Monday. I'm in California if that's of any help.

> Note: the following messages are harmless
> nouveau E[    PBUS][0000:02:00.0] MMIO read of * FAULT at 0x1002**

Glad to hear that! That's the first thing that appears after the
plymouth "fill the bubble" screen and before I get the GDM login.
----------------------------------------------------------------------
- Rick Stevens, Systems Engineer, AllDigital    ricks at alldigital.com -
- AIM/Skype: therps2        ICQ: 22643734            Yahoo: origrps2 -
-                                                                    -
-     Squawk!  Pieces of Seven!  Pieces of Seven!  Parity Error!     -
----------------------------------------------------------------------

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/nouveau/attachments/20130726/5cac9682/attachment.html>