Linux Kernel 6.7 & Nouveau & Sway GSP Error: mmu fault queued

[cartok]

I actually don't know if it's an issue for nouveau or for sway. Link to other (duplicated) issue: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10475

System information

inxi -GSC -xx

System:
  Host: local Kernel: 6.7.0-zen3-1-zen arch: x86_64 bits: 64 compiler: gcc
    v: 13.2.1 Desktop: sway v: 1.8.1 dm: N/A Distro: Arch Linux
CPU:
  Info: 8-core model: Intel Core i9-9900K bits: 64 type: MT MCP
    arch: Coffee Lake rev: D cache: L1: 512 KiB L2: 2 MiB L3: 16 MiB
  Speed (MHz): avg: 5000 high: 5013 min/max: 800/5000 cores: 1: 5000 2: 5005
    3: 5013 4: 5000 5: 5001 6: 5000 7: 5000 8: 4996 9: 5000 10: 4999 11: 5000
    12: 5001 13: 5000 14: 5000 15: 5000 16: 4999 bogomips: 115200
  Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx
Graphics:
  Device-1: NVIDIA AD104 [GeForce RTX 4070] vendor: Gigabyte driver: nouveau
    v: kernel arch: Lovelace pcie: speed: 8 GT/s lanes: 16 ports:
    active: DP-1,DP-2 empty: DP-3,HDMI-A-1 bus-ID: 01:00.0 chip-ID: 10de:2786
  Device-2: Sunplus Innovation Sandberg USB Webcam Pro
    driver: snd-usb-audio,uvcvideo type: USB rev: 2.0 speed: 480 Mb/s lanes: 1
    bus-ID: 1-5.1:5 chip-ID: 1bcf:2650
  Display: wayland server: Xwayland v: 23.2.4 compositor: sway v: 1.8.1
    driver: gpu: nouveau d-rect: 5360x1440 display-ID: 1
  Monitor-1: DP-1 pos: right model: Gigabyte M34WQ res: 3440x1440 dpi: 109
    diag: 867mm (34.1")
  Monitor-2: DP-2 pos: primary,left model: Dell AW2521HFL res: 1920x1080
    dpi: 90 diag: 623mm (24.5")
  API: EGL v: 1.5 platforms: device: 0 drv: nouveau device: 1 drv: swrast
    gbm: drv: nouveau surfaceless: drv: nouveau wayland: drv: nouveau x11:
    drv: nouveau
  API: OpenGL v: 4.5 compat-v: 4.3 vendor: mesa v: 23.3.3-arch1.1 glx-v: 1.4
    direct-render: yes renderer: NV194 device-ID: 10de:2786
  API: Vulkan Message: No Vulkan data available.

• OS: Arch Linux
• GPU: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation AD104 [GeForce RTX 4070] [10de:2786] (rev a1)
• Kernel version: Linux local 6.7.0-zen3-1-zen #1 ZEN SMP PREEMPT_DYNAMIC Sat, 13 Jan 2024 14:36:54 +0000 x86_64 GNU/Linux
• Mesa version: OpenGL version string: 4.3 (Compatibility Profile) Mesa 23.3.3-arch1.1
• Desktop manager and compositor: sway version 1.8.1

Problem Description

I've tried to switch from using my APU to my RTX 4070 using nouveau, now that the GSP firmware is live, but it's unstable on my system. From time to time sway exists, the program does not finish, I don't get the debug output from sway when I run it with sway -d > foo.log, but I get some other info:

Jan 21 14:23:37 local kernel: nouveau 0000:01:00.0: gsp: mmu fault queued
Jan 21 14:23:37 local kernel: nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:16 type:31 scope:1 part:233
Jan 21 14:23:37 local kernel: nouveau 0000:01:00.0: fifo:c00000:0002:0010:[sway[915]] errored - disabling channel
Jan 21 14:23:37 local kernel: nouveau 0000:01:00.0: sway[915]: channel 16 killed!
Jan 21 14:23:37 local systemd[1]: Created slice Slice /system/systemd-coredump.

Not easily reproducible, happens after around an hour or so.

Further System Information

I install nouveau like this:

# Vulkan is not yet working with nouveau, still installing `vulkan-mesa-layers`.
# See: https://wiki.archlinux.org/title/nouveau
pacman -S --needed --noconfirm \
  mesa \
  libva-mesa-driver \
  mesa-vdpau \
  vulkan-mesa-layers

# Add early loading
declare REQUIRED_MODULES="nouveau"
declare MODULES=$(rg $REQUIRED_MODULES $(readlink /etc/mkinitcpio.conf))

if ! [[ $MODULES =~ $REQUIRED_MODULES ]]; then
  sd "(^MODULES=\()" "\$1 $REQUIRED_MODULES" $(readlink /etc/mkinitcpio.conf)
fi

mkinitcpio -P linux-zen

The early loading works. I disabled the APU (i915) in the BIOS. No other drivers are loaded, not i915 nor nvidia and so far everything seems to run fine (chrome, glxgears, eglgears).

Possible Trace

Sway does is not detecting that the 34" Gigabyte Screen actually has 144 Hz, it only detects 100 Hz for the default resolution:

swaymsg -t get_outputs

Output DP-1 'GIGA-BYTE TECHNOLOGY CO., LTD. M34WQ 0x000002FF' (focused)
  Current mode: 3440x1440 @ 59.973 Hz
  Position: 1920,0
  Scale factor: 1.000000
  Scale filter: nearest
  Subpixel hinting: unknown
  Transform: normal
  Workspace: 1 - linux
  Max render time: off
  Adaptive sync: disabled
  Available modes:
    3440x1440 @ 59.973 Hz
    3440x1440 @ 99.982 Hz
    2560x1440 @ 119.998 Hz
    2560x1440 @ 59.951 Hz
    2560x1080 @ 143.945 Hz

The sway output configuration looks like this:

output "GIGA-BYTE TECHNOLOGY CO., LTD. M34WQ 0x000002FF" {
  scale 1
  pos 1920 0
}

output "Dell Inc. AW2521HFL DDR4M23" {
  mode [email protected]
  pos 0 0
  scale 1
}

[bl4ckb0ne]

I don't think Lovelace is supported by nouveau but I might be wrong. Last time I checked they were barely supporting Ampere (the previous gen).

[cartok]

Oh boy, I should've checked that before. Sorry for the fuzz. Thanks for the reply.
Here are the references for anyone else frustrated:

• https://nouveau.freedesktop.org/FeatureMatrix.html

[ghost]

Oh boy, I should've checked that before. Sorry for the fuzz. Thanks for the reply. Here are the references for anyone else frustrated:

* https://nouveau.freedesktop.org/FeatureMatrix.html

* https://nouveau.freedesktop.org/FeatureMatrix.html

The references you've provided state that Ada Lovelace IS supported by nouveau, so I'm not sure what you mean by that?

[cartok]

Thanks. Yes but not the 2D acceleration, thought it might cause the issue, for example in chrome rendering. But I actually have no idea if those unsupported topics matter or not to have it fully functioning.

If not, then I wonder if this is a common issue or not.

Shall I reopen it?

[ghost]

AFAIK if you're getting an MMU fault, then 2D acceleration shouldn't matter, I'd reopen it.

[Zambieslar]

Do you guys know if there is a workaround? I'm running Hyprland and experiencing the same issues using NVK drivers.

[jpxaraujo]

Do you guys know if there is a workaround? I'm running Hyprland and experiencing the same issues using NVK drivers.

Today I tested Fedora 40 beta sway spin, and it has the same problem.

Seems to be the same or similar issue: https://gitlab.freedesktop.org/drm/nouveau/-/issues/346

[whitespace-v]

same with Arch + Hyprland running on: System: Host: ArchLinux Kernel: 6.9.3-arch1-1 arch: x86_64 bits: 64 compiler: gcc v: 14.1.1 Desktop: Hyprland v: 0.40.0 dm: N/A Distro: Arch Linux CPU: Info: 12-core (8-mt/4-st) model: 12th Gen Intel Core i7-12700K bits: 64 type: MST AMCP arch: Alder Lake rev: 2 cache: L1: 1024 KiB L2: 12 MiB L3: 25 MiB Speed (MHz): avg: 3872 high: 4700 min/max: 800/4900:5000:3800 cores: 1: 4700 2: 4700 3: 4567 4: 4700 5: 4700 6: 800 7: 4626 8: 800 9: 4588 10: 800 11: 4700 12: 4610 13: 4684 14: 4699 15: 4689 16: 4688 17: 3601 18: 3601 19: 3599 20: 3600 bogomips: 144440 Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx Graphics: Device-1: Intel AlderLake-S GT1 vendor: Gigabyte driver: i915 v: kernel ports: active: none empty: DP-1,HDMI-A-1,HDMI-A-2 bus-ID: 00:02.0 chip-ID: 8086:4680 Device-2: NVIDIA GA104 [GeForce RTX 3060 Ti Lite Hash Rate] driver: nouveau v: kernel arch: Ampere pcie: speed: 16 GT/s lanes: 16 ports: active: DP-2,DP-3,HDMI-A-3 empty: DP-4 bus-ID: 01:00.0 chip-ID: 10de:2489 Display: wayland server: Xwayland v: 24.1.0 compositor: Hyprland v: 0.40.0 driver: gpu: nouveau display-ID: 1 Monitor-1: DP-2 model: AF24H1 res: 1920x1080 dpi: 92 diag: 609mm (24") Monitor-2: DP-3 model: AF24H1 res: 1920x1080 dpi: 92 diag: 609mm (24") Monitor-3: HDMI-A-3 model: MSI MP243X res: 1920x1080 dpi: 92 diag: 604mm (23.8") API: EGL Message: EGL data requires eglinfo. Check --recommends.

[bl4ckb0ne]

There's a patch here 1, i encourage you to try it

[cartok]

There's a patch here 1, i encourage you to try it

Unfortunately I already changed to AMD and can't test it 🙏

[bl4ckb0ne]

Alright then, I'll close the issue since its not sway.