AMD vega crashing

[Joey0]

Hi all,

I have moved my desktop to freebsd 14.0 for a few weeks now but my system keeps crashing every few days. It used to happen more frequently when i had a 1440p monitor plugged in, and the most recent time it crashed was when i was trying to get a llama file to use gpu acceleration so i believe the cause is my graphics card.

I have a AMD Vega 64 and i have installed the graphics drivers according to the freebsd handbook, i can load a desktop fine and any apps but after a few days of being on it will just hang and i can no longer ssh to it or when i manage to get physical access it will not respond to any input and has to be hard power cycled.

I have ran sysrc kld_list+=amdgpu and can see the drm and relevant vega drivers are being loaded ok.

Id Refs Address                Size Name
 1  120 0xffffffff80200000  1d34598 kernel
 2    1 0xffffffff81f35000    1c3a8 geom_eli.ko
 3    1 0xffffffff81f52000     7718 cryptodev.ko
 4    1 0xffffffff81f5a000   5d51c8 zfs.ko
 5    1 0xffffffff83200000   504958 amdgpu.ko
 6    2 0xffffffff83110000    7c050 drm.ko
 7    1 0xffffffff8318d000     22b8 iic.ko
 8    3 0xffffffff83190000     3080 linuxkpi_hdmi.ko
 9    3 0xffffffff83194000     6350 dmabuf.ko
10    3 0xffffffff8319b000     3378 lindebugfs.ko
11    1 0xffffffff8319f000     b360 ttm.ko
12    1 0xffffffff831ab000     2220 amdgpu_vega10_gpu_info_bin.ko
13    1 0xffffffff831ae000     64e0 amdgpu_vega10_sdma_bin.ko
14    1 0xffffffff831b5000     64e0 amdgpu_vega10_sdma1_bin.ko
15    1 0xffffffff831bc000    2ac70 amdgpu_vega10_sos_bin.ko
16    1 0xffffffff83705000    2c2e0 amdgpu_vega10_asd_bin.ko
17    1 0xffffffff831e7000     7560 amdgpu_vega10_pfp_bin.ko
18    1 0xffffffff831ef000     6560 amdgpu_vega10_me_bin.ko
19    1 0xffffffff831f6000     4560 amdgpu_vega10_ce_bin.ko
20    1 0xffffffff83732000     63e0 amdgpu_vega10_rlc_bin.ko
21    1 0xffffffff83739000    43800 amdgpu_vega10_mec_bin.ko
22    1 0xffffffff8377d000    43800 amdgpu_vega10_mec2_bin.ko
23    1 0xffffffff837c1000    423e0 amdgpu_vega10_acg_smc_bin.ko
24    1 0xffffffff83804000    5f080 amdgpu_vega10_uvd_bin.ko
25    1 0xffffffff83864000    2c800 amdgpu_vega10_vce_bin.ko
26    1 0xffffffff831fb000     3390 acpi_wmi.ko
27    1 0xffffffff83891000     3220 intpm.ko
28    1 0xffffffff83895000     2178 smbus.ko
29    1 0xffffffff83898000     2260 pflog.ko
30    1 0xffffffff8389b000    4d038 pf.ko
31    1 0xffffffff838e9000     3360 uhid.ko
32    1 0xffffffff838ed000     4364 ums.ko
33    1 0xffffffff838f2000     33c0 usbhid.ko
34    1 0xffffffff838f6000     3380 hidbus.ko
35    1 0xffffffff838fa000     3360 wmt.ko

These are the drm versions installed currently

drm-515-kmod-5.15.160          DRM drivers modules
drm-kmod-20220907_3            Metaport of DRM modules for the linuxkpi-based KMS components
gpu-firmware-kmod-20240401,1   Firmware modules for the drm-kmod drivers
libdrm-2.4.122,1               Direct Rendering Manager library and headers

Any suggestions of how to fix or help diagnose this would be much appreciated

 

[Joey0]

Here is my dmesg output, as well as current xorg log

 

[gotnull]

Hi,

RX Vega 64 owner here too, indeed your situation doesn't feel right.
I can't tell for sure that will work with 14.1-RELEASE because I currently use 13.3-RELEASE, but I think that it might help a lot if you build those:
graphics/gpu-firmware-kmod, graphics/drm-kmod, graphics/drm-515-kmod from port directly instead of installing them from pkg, why that? hopefully someone more experienced than me will jump in to explain it to you more correctly than I could ever do but if you allow me to summarize they are outdated.
So at this point there are two choices:
_ use the port collection, compile and install the port yourself.
_ use a builder called ports-mgmt/poudriere which will create a package that can be installed via pkg.
Personally I use the 2nd solution.