[Nouveau] Nouveau: kernel hang on Optimus+Intel+NVidia GeForce 1060m

[Tobias Klausmann]

Hi,

i remember seeing the same error with earlier firmware version with a
similar system (GP106) once in a while on boot, yet it does not happen
with newer versions. Maybe you could try to update the firmware to the
latest version from kernel-firmware.

As a small addition: I remember deeply: you should ignore the
_OSI(Linux) query, as it may break the system in some ways, if you don't
have a specific bug fixed with adding _OSI(Linux), removing it from the
cmdline is a thing to test!

Greetings,

Tobias


On 9/11/17 4:54 PM, Nicolas Mercier wrote:
> Hi,
> I have an Optimus-enabled laptop with a GTX 1060m. I never got it to
> fully work with Nouveau even after Pascal support was added. I need to
> run the kernel with nouveau.runpm=0 to get it to work. Unfortunately
> without proper power mangement support, my laptop will run out of
> battery after about 1h30, so I'd love to get Optimus working.
>
> What I can see is that when the extra GPU is not in use, Nouveau will
> try to shut it off. This seems to work, as the indicator of the laptop
> changes to amber (discrete GPU in use) to blue (discrete GPU powered
> off), and the kernel log (attached) reports that the GPU went off.
>
> If I wake up the GPU (plugging in/out the external monitor, running a
> GL command with DRI_PRIME=1, or even lspci) will make the computer
> unresponsive and only a force shutdown will work, as no graphics
> command seems to be able to execute anymore. Nouveau (likely, vga
> switcheroo) tries to wake up the GPU but it seems to fail. The LED
> indicator does indicate the GPU has power though.
>
> Most of the time (but not often when I have an external monitor
> plugged into the discrete card before I boot the computer) I get a
> timeout during boot or during modprobe (see kernel log). Even with
> runpm=0, I can't seem to be able to run GL commands on it:
>
> yngwe at labarbara: % DRI_PRIME=1 glxinfo
> name of display: :0.0
> nvc0_screen_create:857 - Error allocating PGRAPH context for M2MF: -16
> libGL error: failed to create dri screen
> libGL error: failed to load driver: nouveau
> display: :0  screen: 0
> direct rendering: Yes
> ... follows up using the Intel card ...
>
> the error I get in the kernel in that case:
> [  201.612583] nouveau 0000:01:00.0: gr: FECS falcon already acquired
> by gr!
> [  201.612586] nouveau 0000:01:00.0: gr: init failed, -16
>
> It runs on Debian testing + the firmware from Ubuntu as Debian's
> firmware does not have the nvidia blobs, kernel was either 4.12/4.13
> release candidates from Debian repositories or kernel 4.13.1 self
> compiled. I always had exactly the same symptoms on all these kernels.
>
> /nicolas
>
>
> _______________________________________________
> Nouveau mailing list
> Nouveau at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau