Radeon 7900XTX randomly hangs after kernel message "Fence fallback timer expired on ring comp_1.1.1"

Brief summary of the problem:

When I plays in the game "Tom Clancy’s The Division® 2" GPU may randomly hang after kernel message "Fence fallback timer expired on ring comp_1.1.1".

Hardware description:

• CPU: AMD Ryzen 3950X
• GPU: 0c:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 31 [Radeon RX 7900 XT/7900 XTX] [1002:744c] (rev c8)
• System Memory: 64GB
• Display(s): Philips 436M6VBPAB
• Type of Display Connection: DisplayPort

System information:

• Distro name and Version: Fedora Rawhide
• Kernel version: 6.2-rc5
• Custom kernel: commit 7c46948a6e9c + patch #2360 (comment 1739852)
• Mesa commit: mesa/mesa@15e19d04

How to reproduce the issue:

Play in the game "Tom Clancy’s The Division® 2" with DX12 render. For it needed steam client, bleeding edge proton experimental and the game "Tom Clancy’s The Division® 2".

[58796.836478] [drm] Fence fallback timer expired on ring comp_1.1.1
[61160.072879] GpuWatchdog[4725]: segfault at 0 ip 00007fcf06592336 sp 00007fceeddfd3e0 error 6 in libcef.so[7fcf020ef000+776f000] likely on CPU 8 (core 8, socket 0)
[61160.072908] Code: 89 de e8 3d ef 6e ff 80 7d cf 00 79 09 48 8b 7d b8 e8 be 65 2c 03 41 8b 84 24 e0 00 00 00 89 45 b8 48 8d 7d b8 e8 ca d7 b5 fb <c7> 04 25 00 00 00 00 37 13 00 00 48 83 c4 38 5b 41 5c 41 5d 41 5e
[61401.031196] sysrq: Show Blocked State
[61401.032097] task:steamwebhelper  state:D stack:0     pid:4692  ppid:4659   flags:0x00004002
[61401.032119] Call Trace:
[61401.032130]  <TASK>
[61401.032141]  __schedule+0x3a9/0x1380
[61401.032156]  ? __schedule+0x3b1/0x1380
[61401.032169]  schedule+0x5d/0xe0
[61401.032180]  do_exit+0xd7/0xaf0
[61401.032193]  do_group_exit+0x2d/0x80
[61401.032205]  get_signal+0x932/0x970
[61401.032217]  arch_do_signal_or_restart+0x3a/0x280
[61401.032230]  exit_to_user_mode_prepare+0x117/0x1f0
[61401.032243]  syscall_exit_to_user_mode+0x17/0x40
[61401.032255]  do_syscall_64+0x67/0x80
[61401.032268]  ? exc_page_fault+0x70/0x170
[61401.032282]  entry_SYSCALL_64_after_hwframe+0x72/0xdc
[61401.032296] RIP: 0033:0x7fceffb6bc6d
[61401.032329] RSP: 002b:00007fff93976d98 EFLAGS: 00000212 ORIG_RAX: 00000000000000ca
[61401.032341] RAX: fffffffffffffe00 RBX: 00001b2bd8ad022c RCX: 00007fceffb6bc6d
[61401.032352] RDX: 0000000000000002 RSI: 0000000000000009 RDI: 00001b2bd8ad022c
[61401.032362] RBP: 00007fff93976db0 R08: 0000000000000000 R09: ffffffffffffffff
[61401.032372] R10: 0000000000000000 R11: 0000000000000212 R12: 0000000000000000
[61401.032381] R13: 0000000000000001 R14: 00001b2bd8ab3000 R15: 0000000000000000
[61401.032446]  </TASK>
[61401.032508] task:steamwebhe:cs0  state:D stack:0     pid:4695  ppid:4659   flags:0x00004002
[61401.032520] Call Trace:

Screenshots/video files

Log files (for system lockups / game freezes / crashes)

• Dmesg log dmesg.txt

changed the description

mentioned in issue mesa/mesa#7939 (closed)

[Fri Jan 27 22:01:58 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 22:02:00 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 22:07:13 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 22:07:49 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 22:07:57 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 22:12:36 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 22:13:07 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 22:13:30 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 23:33:59 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:34:33 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:39:33 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 23:40:23 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:40:25 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:43:41 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 23:45:17 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:48:25 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:48:47 2023] [drm] Fence fallback timer expired on ring comp_1.1.1
[Fri Jan 27 23:49:09 2023] [drm] Fence fallback timer expired on ring comp_1.1.0
[Fri Jan 27 23:49:13 2023] [drm] Fence fallback timer expired on ring comp_1.1.0

I also noticed that the messages themselves are harmless. And they usually match with time with game stuttering.
Some times the last stutter become endless after that the system GUI stop responsible.

Can you check /proc/interrupts for the GPU? Those messages are usually due to interrupts not being delivered to the driver for some reason.

Content of /proc/interrupts interesting after hang occurs? Or after any stutter?
Below attached content of /proc/interrupts after stutters:
interrupts.txt
interrupts.txt

interrupts1.txt
interrupts2.txt

added hang/freeze label

I have the very same issue while playing the Division 2.

Started with Lutris.

I´m running Fedora 39.

Things I have tried so far. The distribution drivers Mesa 23.3.3-1 Recompiling Libdrm. (needed for Mesa 24) Mesa 24.0 RC2 ( compiled from git )

Kernel 6.6.13 Kernel 6.7.1

The Problem still exists.

My System: ASUS ROG Strix X570 Ryzen 5900X 32GB memory Radeon RX 6900XT (XTXH chip)

IOMMU is disabled on my system.

I don't think the issue is specific to just the 7000 series, I'm running a 6900 XT and I've experienced the same persistent issue of Fence fallback timer expired on ring comp_1.1.1 spam in dmesg while playing Division 2 with DX12 renderer enabled ingame (using proton experimental / vkd3d-proton). Curiously, the problem completely vanishes when switching the game to DX11 mode.

I have exactly the same issue here. For me the following games are affected:

I tested dirt 5 with radv and with amdvlk, same issue.

The divsion 2 in DX12. Avatar (DX12) Dirt 5 (DX12)

The issue does not show up in any of my DX11 games with dxvk. The issues does not show up in windows 10 on the same hardware.

For me this is a longstandig issue already, it already started at the 5.x kernel time.

What i tested so far:

• Switch from Opensuse TW to Arch -> Issue still there
• Bios Upgrade to latest version -> Issue is still there

Kernel parameters:

• amdgpu.msi=0 -> Issue is still there
• noapic -> Issue is still there
• amdgpu.aspm = 0 -> Issue is still there

Bios settings:

• disable iommu -> Issue is still there

I also tried different pm settings for the card:

• power_dpm_force_performance_level auto and manual
• UV and stock
• different pp_power_profile_mode settings

The issue came out of nowhere and decided to stay.

@agd5f Any further ideas on the issue ?

My setup:

Machine: Type: Desktop System: Gigabyte product: X570 AORUS ELITE v: -CF serial: <superuser required> Mobo: Gigabyte model: X570 AORUS ELITE serial: <superuser required> UEFI: American Megatrends LLC. v: F38h date: 12/25/2023 CPU: Info: 16-core model: AMD Ryzen 9 5950X bits: 64 type: MT MCP cache: L2: 8 MiB Speed (MHz): avg: 2403 min/max: 2200/5083 cores: 1: 3400 2: 2800 3: 2200 4: 3400 5: 2200 6: 2200 7: 2200 8: 2200 9: 2200 10: 3400 11: 2200 12: 2720 13: 2200 14: 2200 15: 2200 16: 2200 17: 2200 18: 2200 19: 2200 20: 2200 21: 2200 22: 2200 23: 2200 24: 2200 25: 2200 26: 2200 27: 2200 28: 2200 29: 2200 30: 2200 31: 2800 32: 3400 Graphics: Device-1: AMD Navi 21 [Radeon RX 6800/6800 XT / 6900 XT] driver: amdgpu v: kernel Display: wayland server: X.org v: 1.21.1.11 with: Xwayland v: 23.2.4 compositor: kwin_wayland driver: X: loaded: amdgpu dri: radeonsi gpu: amdgpu resolution: 3440x1440

I think I am facing the same issue, while to I don't experience hangs, I do experience the stutter. https://github.com/HansKristian-Work/vkd3d-proton/issues/1956

Someone else reported the a similar issue on gamescope: https://github.com/ValveSoftware/gamescope/issues/1205#issuecomment-2017835136

I don't know if they are all related to the same issue, but they all have 3 things in common.

1. Huge frametime spike
2. [drm] Fence fallback timer expired on ring comp logged during the spike
3. all use vkd3d

I was previously experiencing the same or similar issue described by the posters above in the games Helldivers 2 and Resident Evil 4 Remake, on an RX 6600 non-XT, with symptoms including

• 500+ ms stutter
• [drm] Fence fallback timer expired in dmesg/journalctl
• only with D3D12, regardless of VKD3D-Proton or Mesa version used

However, I tried building the latest commit at the time of writing (682aef19) of the amd-staging-drm-next kernel at https://gitlab.freedesktop.org/agd5f/linux/-/tree/amd-staging-drm-next, and when running this kernel, have not been able to reproduce the issue on Helldivers 2 across 3 days of frequent play. I have not tested Resident Evil 4 yet. I would suggest those affected try this kernel as well and see if this improves the situation for them.

EDIT: Further testing has shown that I can and do still encounter the issue when using the amd-staging-drm-next kernel and Mesa 24.0.4-2 from the Arch Linux repos. The issue seems to become somewhat less common with mesa git commit 96ee0d6711, but stutter still persists under some conditions.

Sadly that does not help with Dirt 5 here :(.

@lstevens Hi, I managed to stop (for now, hope it doesn't come back) having this issue by using a udev rule found here make sure to use COMPUTE. Can you by chance test it out and see if it helps you?

@pingubot Hi, I know you were still having this issue after using the udev rule, were you playing on x11 or wayland?

@nutta-git Unfortunately, I have tried that udev rule with both 3D_FULL_SCREEN as in the post and COMPUTE, and it has not solved the issue for me. I have also been experimenting with installing other, older distros for testing, as this issue has manifested recently for me (within the last 2-3 months), but with little luck so far.

I've also tested x11 and Wayland sessions of various types and with no luck yet. (kwin wayland, kwin x11, gamescope, sway, labwc.)

Kernel versions, CPU and GPU power profiles, and mesa versions do not seem to make a difference either.

Do let me know if you gain any further insight into the issue.

@nutta-git Its on wayland. But the same issue happend on X11 for me in the not so long ago past.

I tried out gnome 46.1 because it had the explicit sync patches and I was able to reproduce the issue even with the udev rule. I am thinking this issue only seems to affects major DEs?

I'm facing the same issue, currently on a 6950XT and with Horizon Forbidden West. After a few " [drm] Fence fallback timer expired on ring comp" error my GPU doesn't resume anymore either, screen just goes black and I have to hard-reset the PC for it to work again.

What hardware are you using ? MB, cpu etc..

CPU: 7800X3D MB: Asus X670E Pro WiFi 32GB RAM 6950XT is from Sapphire.

Not sure what else would be relevant.

@Yasuhan whats desktop environment? i.e gnome, kde, xfce

@nutta-git I'm using Hyprland, also happens with KDE (tried that because HDR for horizon forbidden west)

@Yasuhan if you are using hyprland, can you try the udev rule? make sure to use COMPUTE

@nutta-git Doesn't change anything, still have the same issue.

I have Fence fallback timer expired on ring comp_1.1.0 or Fence fallback timer expired on ring comp_1.1.1 when playing Horizon Forbidden West on "3D Fullscreen" power profile with default clocks (2489 MHz). No issue on "Compute" or "Virtual reality" profiles (but this profile drains more power). Also no issue on "3D Fullscreen" power profile when I lower the GPU clock manually (2000 MHz).

It causes game stuttering, but not hangs entire PC, I have RX 6900 XT, Mesa 24.1.7, Kernel 6.10.11, Xorg.

Since kernel 6.11.7 the "3D Fullscreen" profile is default, so this issue happens on default settings now.