[Issue] FW13 Fedora 41/42 (kernel 6.14) hard freeze

I recently started encountering hard freezes while using my FW13 with Fedora. I usually do some programming with Chrome and maybe Slack, and suddenly everything stops, including the mouse and the keyboard, but audio still works for some reason.

It started a few weeks ago with Fedora 41, but I don’t remember exactly after what kernel version update (my guess would be after upgrading to 6.13 maybe?). Upgraded to Fedora 42 yesterday, and the issue persists even after upgrading the kernel to 6.14.2

I guess it’s a kernel issue, because it started after upgrading the kernel and persists across two different distro/DE versions.

Which Linux distro are you using?

Happened on Fedora 41 and Fedora 42

Which kernel are you using?

Currently, 6.14.2

Noticed the issue with the last few versions. I don’t remember at what version exactly this started happening.

Which BIOS version are you using?

3.07

Which Framework Laptop 13 model are you using?

AMD Ryzen 7 7840U Radeon 780M
120hz
64GB RAM

6 Likes

I’m having the same issue with Ubuntu 25.04, Framework 13 AMD Ryzen 7, tried kernels 6.14 and 6.14.2.

It looks like https://gitlab.freedesktop.org/drm/amd/-/issues/4141
Can you also report your linux-firmware version, as well as kernel messages from last boot? (if it’s this issue you should be able to initiate a normal shutdown using your power button, it’ll take a while depending on how fast the hanging processes get killed)

I noticed yesterday that there were some very recent updates to amdgpu firmwares, not tagged yet. I am trying them out currently, and I haven’t had a freeze since :crossed_fingers:. (EDIT: it hasn’t resolved the issue, still getting the same freezes)

I downgraded to 6.13.6 (6.13.12 froze), and it’s working fine.

I can try 6.14 (if I still have it) again and get the logs if needed, but that would probably be around the weekend.

Can you share how to get these logs?

I don’t use a systemd-based distro myself, but journalctl -k -b -1 should give you the kernel messages (-k) of last boot (-b -1).

1 Like

firmware

$ rpm -qa | grep linux-firmware
linux-firmware-whence-20250410-1.fc42.noarch
linux-firmware-20250410-1.fc42.noarch

I will update this with the logs once it crashes again.

Edit: Crashed. logs here: logs - Pastebin.com

same (hard freeze) here:
Framework 13 AMD Ryzen 7840 BIOS 3.08 32GB
Fedora Silverblue 6.14.3-300.fc42.x86_64

I am wondering if this might be related to Graphical Corruption in Fedora 41 on AMD (BIOS3.06, Linux 6.13.5)? It seems the latest mesa version available in the Fedora repo (since a few hours ago) fixed that issue. However, I have not tested long enough to be sure it fixed this one as well.

I doubt that it’s a specific lib causing the issue. I tried the following kernel versions:

6.14.[0|1|2]: hard freeze
6.13.12: same issue
6.15.0 rc2: same issue
6.13.6: works fine

So it looks like a kernel issue on AMD hardware.

But I do run into the same graphics issue mentioned in that thread on 6.13.6 though.

The freeze happened again, it seems the answer is “no”.

1 Like

AMD 7840U
Fedora Silverblue 42:
Firmware 3.07

Edit: It’s not too surprising, but this issue is also present on R9 AI HX 370 :frowning:

~ $ rpm -qa | grep -E 'linux-firmware-2|mesa-libGL-|kernel-6'
linux-firmware-20250410-1.fc42.noarch
mesa-libGL-25.0.4-2.fc42.x86_64
kernel-6.14.3-300.fc42.x86_64

I’ve been seeing this recently and sometimes it doesn’t recover, but sometimes it will recover leaving something like the following in dmesg:

[ 8789.798065] amdgpu 0000:c1:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:24 vmid:7 pasid:32791)
[ 8789.798073] amdgpu 0000:c1:00.0: amdgpu:  in process code pid 9538 thread code:cs0 pid 9561)
[ 8789.798077] amdgpu 0000:c1:00.0: amdgpu:   in page starting at address 0x000000003f800000 from client 10
[ 8789.798079] amdgpu 0000:c1:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00701430
[ 8789.798082] amdgpu 0000:c1:00.0: amdgpu: 	 Faulty UTCL2 client ID: SQC (data) (0xa)
[ 8789.798084] amdgpu 0000:c1:00.0: amdgpu: 	 MORE_FAULTS: 0x0
[ 8789.798086] amdgpu 0000:c1:00.0: amdgpu: 	 WALKER_ERROR: 0x0
[ 8789.798087] amdgpu 0000:c1:00.0: amdgpu: 	 PERMISSION_FAULTS: 0x3
[ 8789.798089] amdgpu 0000:c1:00.0: amdgpu: 	 MAPPING_ERROR: 0x0
[ 8789.798091] amdgpu 0000:c1:00.0: amdgpu: 	 RW: 0x0
[ 8799.843140] amdgpu 0000:c1:00.0: amdgpu: Dumping IP State
[ 8799.845008] amdgpu 0000:c1:00.0: amdgpu: Dumping IP State Completed
[ 8799.845537] amdgpu 0000:c1:00.0: amdgpu: ring gfx_0.0.0 timeout, but soft recovered
1 Like

Several hard freezes recently, including two today.

I’ve noticed that if I catch the problem quickly (usually when I’m reading something and the page stops scrolling), and switch workspaces and then back again, it’ll resume normal functionality rather than completely locking up. This occurs pretty frequently; at least a few times a day.

AMD 7640U
Fedora Workstation 41
BIOS 3.07
Kernel 6.14.3-200.fc41.x86_64 (though it’s been going on on previous iterations of 6.14 as well)

Im also playing around with my new AI 300 device - looking for the best settings kernel etc.

NO idea how reproduceable my just now experience is. I just had ~3 crashes/hangs in a row basically during startup and reopening all my tools → full system load including gpu hardware decoding of a video.

My system was very stable before … what did i change? BIOS iGPU memory allocation.

I changed it from Medium [16GB] to Minimum [0.5GB] and had the crashes 3 times in a row within 10 mins. Im writing this post back on Medium settings and no crashes for the last 10 mins.

MAYBE this helps others.

TLDR; try to increse your iGPU Memory Allocation in the BIOS.

That said … for more the other setting - AMD software defined does not persist, on a reboot its back to the setting before - which hence seems to be a BIOS bug.

I’ve been using my HX 370 and I observe the same thing, this occurs very frequently with “Minimum” and less often with “Medium” (my system is also 64GB meaning “Medium” = 16GB). I’d like to not have to dedicate that much RAM to the GPU but if it makes the machine more usable I’m okay with it as a workaround.

Complete uneducated conjecture: this is feeling like a bug in how amdgpu (the kmod) manages GPU-shared memory? I know people have said that Mesa 25.0.4 fixes the issue but it feels like that’s only part of the problem, shouldn’t the kernel prevent userspace from crashing the GPU in the first place? (Am I naïve in thinking GPU drivers are supposed to be stable?)

Maybe I’m making this up, but I too notice that if I start going wild with three-fingers-down gesture in GNOME (“activities view”) as soon as it freezes, that seems to result in soft-reset saving my session more often?

@Matt_Hartley this starts to become an interesting thread in my opinion. We at least need a new option in the bios that is somewhere in the middle of 0.5GB and 16 GB :wink:

To add some more hints while thinking about things.

  • With the iGPU’s VRAM locked to just 0.5 GB via that BIOS reservation, I’m getting wild glitches - imagine random, colorful 30×30-pixel blocks all over the screen (no screenshot, sorry) and some random flickering.

  • i maybe got more screen real estate hooked up than usual, so that tiny reserved VRAM gets crushed trying to drive everything

xrandr relevant lines

Screen 0: minimum 16 x 16, current 7680 x 1440, maximum 32767 x 32767
DP-2 connected primary 5120x1440+0+0 (normal left inverted right x axis y axis) 1190mm x 340mm
   5120x1440     69.95*+
DP-3 connected 2560x1440+5120+0 (normal left inverted right x axis y axis) 600mm x 340mm
   2560x1440     59.91*+

https://knowledgebase.frame.work/en_us/expansion-card-functionality-on-framework-laptop-13-amd-ryzen-ai-300-series-Hy5SfMRs1l

one display on USB-C [port 3] -> DisplayPort 
the other on big  hub USB-C [port 1] -> HUB -> HDMI

I haven’t had any issues since Fedora (Silverblue) 42 updated to kernel 6.14.5. Still running the same Mesa version.