AMD RDNA4: Ollama crashes when having service enabled, which crashes the boot process and makes a gpu reset

[afterglow1284]

After boot, GDM completely freezes without allowing any user interaction, including cursor movement, I have managed to make it work once by changing to a tty and back, but I cannot reproduce this anymore.
I can confirm that this behavior isn't present when using the lts version of the kernel.

This is the log for GDM.service:

dic 02 08:05:08 computer systemd[1]: Starting GNOME Display Manager...
dic 02 08:05:08 computer systemd[1]: Started GNOME Display Manager.
dic 02 08:06:54 computer gdm-password][7175]: gkr-pam: unable to locate daemon control file
dic 02 08:06:54 computer gdm-password][7175]: gkr-pam: stashed password to try later in open session
dic 02 08:06:54 computer gdm-password][7175]: pam_unix(gdm-password:session): session opened for user username(uid=1000) by username(uid=0)
dic 02 08:06:54 computer gdm-password][7175]: gkr-pam: unlocked login keyring
dic 02 08:06:55 computer gdm[962]: Gdm: Child process -5718 was already dead.
dic 02 12:28:19 computer systemd[1]: Stopping GNOME Display Manager...
dic 02 12:28:19 computer gdm[962]: Gdm: Failed to list cached users: GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name is not activatable
dic 02 12:28:20 computer systemd[1]: gdm.service: Main process exited, code=exited, status=1/FAILURE
dic 02 12:28:20 computer systemd[1]: gdm.service: Failed with result 'exit-code'.
dic 02 12:28:20 computer systemd[1]: Stopped GNOME Display Manager.
dic 02 12:28:20 computer systemd[1]: gdm.service: Triggering OnFailure= dependencies.
dic 02 12:28:20 computer systemd[1]: gdm.service: Failed to enqueue OnFailure=plymouth-quit.service job, ignoring: Transaction for plymouth-quit.service/start is destructive (systemd-journal-flush.service has 'stop' job queued, but 'start' is included in transaction).

[ptr1337]

Can you provide more informaition what kind of hardware this is?

[afterglow1284]

sure, here is the output of inxi -Fxz:

System:
  Kernel: 6.12.60-2-cachyos-lts arch: x86_64 bits: 64 compiler: gcc v: 15.2.1
  Desktop: GNOME v: 49.2 Distro: CachyOS base: Arch Linux
Machine:
  Type: Desktop Mobo: Gigabyte model: B650 EAGLE AX v: x.x
    serial: <superuser required> Firmware: UEFI vendor: American Megatrends LLC.
    v: F35 date: 07/01/2025
CPU:
  Info: 8-core model: AMD Ryzen 7 9700X bits: 64 type: MT MCP arch: Zen 5
    rev: 0 cache: L1: 640 KiB L2: 8 MiB L3: 32 MiB
  Speed (MHz): avg: 3713 min/max: 600/5581 boost: enabled cores: 1: 3713
    2: 3713 3: 3713 4: 3713 5: 3713 6: 3713 7: 3713 8: 3713 9: 3713 10: 3713
    11: 3713 12: 3713 13: 3713 14: 3713 15: 3713 16: 3713 bogomips: 121364
  Flags-basic: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a
    ssse3 svm
Graphics:
  Device-1: Advanced Micro Devices [AMD/ATI] Navi 48 [Radeon RX 9070/9070
    XT/9070 GRE] vendor: Tul / PowerColor Reaper driver: amdgpu v: kernel
    arch: RDNA-4 bus-ID: 03:00.0
  Device-2: Advanced Micro Devices [AMD/ATI] Granite Ridge [Radeon Graphics]
    vendor: Gigabyte driver: amdgpu v: kernel arch: RDNA-2 bus-ID: 12:00.0
    temp: 33.0 C
  Display: wayland server: X.Org v: 24.1.9 with: Xwayland v: 24.1.9
    compositor: gnome-shell driver: X: loaded: amdgpu unloaded: modesetting
    dri: radeonsi gpu: amdgpu resolution: 2560x1440~60Hz
  API: EGL v: 1.5 drivers: radeonsi,swrast platforms:
    active: gbm,wayland,x11,surfaceless,device inactive: N/A
  API: OpenGL v: 4.6 compat-v: 4.5 vendor: amd mesa v: 25.2.7-cachyos1.2
    glx-v: 1.4 direct-render: yes renderer: AMD Radeon RX 9070 XT (radeonsi
    gfx1201 LLVM 21.1.5 DRM 3.61 6.12.60-2-cachyos-lts)
  API: Vulkan v: 1.4.328 drivers: radv surfaces: N/A devices: 2
  Info: Tools: api: eglinfo, glxinfo, vulkaninfo gpu: lact
    x11: xdpyinfo, xprop, xrandr
Audio:
  Device-1: Advanced Micro Devices [AMD/ATI] Navi 48 HDMI/DP Audio
    driver: snd_hda_intel v: kernel bus-ID: 03:00.1
  Device-2: Advanced Micro Devices [AMD/ATI] Radeon High Definition Audio
    [Rembrandt/Strix] driver: snd_hda_intel v: kernel bus-ID: 12:00.1
  Device-3: Advanced Micro Devices [AMD] Family 17h/19h/1ah HD Audio
    vendor: Gigabyte driver: snd_hda_intel v: kernel bus-ID: 12:00.6
  API: ALSA v: k6.12.60-2-cachyos-lts status: kernel-api
  Server-1: sndiod v: N/A status: off
  Server-2: JACK v: 1.9.22 status: off
  Server-3: PipeWire v: 1.4.9 status: active
Network:
  Device-1: Realtek RTL8111/8168/8211/8411 PCI Express Gigabit Ethernet
    vendor: Gigabyte driver: r8169 v: kernel port: e000 bus-ID: 08:00.0
  IF: enp8s0 state: up speed: 1000 Mbps duplex: full mac: <filter>
  Device-2: Intel Wi-Fi 6E AX210/AX1675 2x2 [Typhoon Peak] driver: iwlwifi
    v: kernel bus-ID: 09:00.0
  IF: wlan0 state: down mac: <filter>
  IF-ID-1: tailscale0 state: unknown speed: -1 duplex: full mac: N/A
  IF-ID-2: tun0 state: unknown speed: 10000 Mbps duplex: full mac: N/A
Bluetooth:
  Device-1: Intel AX210 Bluetooth driver: btusb v: 0.8 type: USB bus-ID: 1-7:3
  Report: btmgmt ID: hci0 rfk-id: 0 state: up address: <filter> bt-v: 5.4
    lmp-v: 13
Drives:
  Local Storage: total: 3.67 TiB used: 1 TiB (27.4%)
  ID-1: /dev/nvme0n1 vendor: Western Digital model: WD BLACK SN850X 1000GB
    size: 931.51 GiB temp: 36.9 C
  ID-2: /dev/sda vendor: Samsung model: SSD 860 EVO 1TB size: 931.51 GiB
  ID-3: /dev/sdb vendor: Seagate model: ST1000DM003-1SB102 size: 931.51 GiB
  ID-4: /dev/sdc vendor: Seagate model: ST1000DM003-1ER162 size: 931.51 GiB
  ID-5: /dev/sdd vendor: SanDisk model: Ultra USB 3.0 size: 28.65 GiB
    type: USB
Partition:
  ID-1: / size: 839.87 GiB used: 278.02 GiB (33.1%) fs: btrfs
    dev: /dev/nvme0n1p2
  ID-2: /boot size: 2 GiB used: 255.3 MiB (12.5%) fs: vfat
    dev: /dev/nvme0n1p1
  ID-3: /home size: 839.87 GiB used: 278.02 GiB (33.1%) fs: btrfs
    dev: /dev/nvme0n1p2
  ID-4: /var/log size: 839.87 GiB used: 278.02 GiB (33.1%) fs: btrfs
    dev: /dev/nvme0n1p2
  ID-5: /var/tmp size: 839.87 GiB used: 278.02 GiB (33.1%) fs: btrfs
    dev: /dev/nvme0n1p2
Swap:
  ID-1: swap-1 type: zram size: 30.46 GiB used: 0 KiB (0.0%) dev: /dev/zram0
Sensors:
  System Temperatures: cpu: 40.6 C mobo: 32.0 C
  Fan Speeds (rpm): N/A
  GPU: device: amdgpu temp: 33.0 C device: amdgpu temp: 36.0 C fan: 0
Info:
  Memory: total: 32 GiB note: est. available: 30.46 GiB used: 4.83 GiB (15.9%)
  Processes: 480 Uptime: 18m Init: systemd
  Packages: 1558 Compilers: clang: 21.1.6 gcc: 15.2.1 Shell: Zsh v: 5.9
    inxi: 3.3.40

[afterglow1284]

I have tried waiting for a while and gdm seems to start working after about a minute.
I have also checked on an intel lunar lake laptop and I cannot replicate this issue.
Here the gdm log is:

dic 04 09:49:32 x9-15 systemd[1]: Starting GNOME Display Manager...
dic 04 09:49:32 x9-15 systemd[1]: Started GNOME Display Manager.
dic 04 09:49:50 x9-15 gdm-fingerprint][4180]: pam_unix(gdm-fingerprint:session): session opened for user username(uid=1000) by username(uid=0)
dic 04 09:49:51 x9-15 gdm-fingerprint][4180]: gkr-pam: couldn't unlock the login keyring.
dic 04 09:49:52 x9-15 gdm[809]: Gdm: Child process -3157 was already dead.

The only error is for the login keyring, but it is due to fprintd

[ptr1337]

dic 02 12:28:20 computer systemd[1]: Stopped GNOME Display Manager.
dic 02 12:28:20 computer systemd[1]: gdm.service: Triggering OnFailure= dependencies.
dic 02 12:28:20 computer systemd[1]: gdm.service: Failed to enqueue OnFailure=plymouth-quit.service job, ignoring: Transaction for plymouth-quit.service/start is destructive (systemd-journal-flush.service has 'stop' job queued, but 'start' is included in transaction)

Seems like it does not come up due plymouth

[ptr1337]

Can you provide a full sudo cachyos-bugreport.sh when booting with the 6.18 kernel?

[afterglow1284]

Here is the output of that command: https://paste.cachyos.org/p/7711106.log

[ptr1337]

Here is the output of that command: https://paste.cachyos.org/p/7711106.log

Can you please try "linux-cachyos" instead of "linux-cachyos-bore"

pkgrel should be -3

[afterglow1284]

I have the same issue, it starts frozen and after about a minute the screen turns completely black before showing a working gdm
The gdm log is still:

dic 04 10:05:08 titan systemd[1]: Starting GNOME Display Manager...
dic 04 10:05:08 titan systemd[1]: Started GNOME Display Manager.
dic 04 10:05:46 titan gdm[966]: Gdm: Child process -3299 was already dead.
dic 04 10:05:46 titan gdm[966]: Gdm: Child process -3299 was already dead.
dic 04 10:05:52 titan gdm-password][6624]: gkr-pam: unable to locate daemon control file
dic 04 10:05:52 titan gdm-password][6624]: gkr-pam: stashed password to try later in open session
dic 04 10:05:52 titan gdm-password][6624]: pam_unix(gdm-password:session): session opened for user daniele(uid=1000) by daniele(uid=0)
dic 04 10:05:52 titan gdm-password][6624]: gkr-pam: unlocked login keyring
dic 04 10:05:53 titan gdm[966]: Gdm: Child process -5114 was already dead.

and the only thing marked as an error was this line: gkr-pam: unable to locate daemon control file

[ptr1337]

Would you mind to try to disable the ollama.service for one boot?

[afterglow1284]

now it seems to be working fine, I will just check with other kernels to be sure

[ptr1337]

So, following is happening I think:

• Boot
• Plymouth is coming up
• Ollama.service starts and crashing the gpu
• gdm waiting for plymouth to be done, but plymouth is crashed
• Waiting till it recovers

[afterglow1284]

After checking both linux-cachyos and linux-cachyos-bore it seems like the issue is gone after disabling ollama

[afterglow1284]

it just seems strange that it did not have any issue with the lts kernel

[ptr1337]

Could be just amdgpu regression in 6.18

[afterglow1284]

Indeed, I have just tried doing the same on an Intel igpu with the service and i cannot replicate the issue