[SOLVED] gpu/kernel issues on AMD PC

derguteweka · 09-27-2022, 05:09 AM

Moin,

Just got this PC:
https://www.pollin.de/p/joy-it-pc-am...1tb-hdd-751497
Installing ubuntu(18something) worked fine with no issues (including GUI).
Also building LFS-11.2 on this system worked fine.
Also booting a Kernel according to config-ok.txt works fine; all i expect from an LFS seems to work OK.
But i have a fancy 80x25 Char text console on my monitor (which is nice, because i don't need my glasses to read it, but um...you know....)
Also because i'd like to proceed with BLFS/xorg, i'd need some drm/gpu thingies enabled in the kernel.
That's where i got stuck. E.g. a kernel built from config-bad.txt gets stuck pretty early during booting. Last lines on my screen are like:

[ 0.301867] pci 0000:00:14:2: PME# supported from D0 D3hot D3cold
[ 0.302258] pci 0000:00:14.3: [1022:780e] type 00 class 0x060100

(the 23 lines above look pretty similar, just other pci numbers, etc.)

Then nothing happens (also for a long time) until i reboot with another kernel.

Similar behaviour, no matter if i chose amd or radeon option. Also no real difference if building as module or in the kernel.
Any hints would be appreciated.

cheers,
WK
PS: There's a diff between the good and bad config:

Code:

2662a2663,2665
> CONFIG_DRM_KMS_HELPER=y
> CONFIG_DRM_FBDEV_EMULATION=y
> CONFIG_DRM_FBDEV_OVERALLOC=100
2663a2667,2683
> CONFIG_DRM_DISPLAY_HELPER=y
> CONFIG_DRM_DISPLAY_DP_HELPER=y
> CONFIG_DRM_DISPLAY_HDMI_HELPER=y
> # CONFIG_DRM_DP_AUX_CHARDEV is not set
> # CONFIG_DRM_DP_CEC is not set
> CONFIG_DRM_TTM=y
> CONFIG_DRM_TTM_HELPER=y
> CONFIG_DRM_SCHED=y
> 
> #
> # I2C encoder or helper chips
> #
> # CONFIG_DRM_I2C_CH7006 is not set
> # CONFIG_DRM_I2C_SIL164 is not set
> # CONFIG_DRM_I2C_NXP_TDA998X is not set
> # CONFIG_DRM_I2C_NXP_TDA9950 is not set
> # end of I2C encoder or helper chips
2671c2691,2708
< # CONFIG_DRM_AMDGPU is not set
---
> CONFIG_DRM_AMDGPU=y
> # CONFIG_DRM_AMDGPU_SI is not set
> # CONFIG_DRM_AMDGPU_CIK is not set
> # CONFIG_DRM_AMDGPU_USERPTR is not set
> 
> #
> # ACP (Audio CoProcessor) Configuration
> #
> # CONFIG_DRM_AMD_ACP is not set
> # end of ACP (Audio CoProcessor) Configuration
> 
> #
> # Display Engine Configuration
> #
> # CONFIG_DRM_AMD_DC is not set
> # end of Display Engine Configuration
> 
> # CONFIG_HSA_AMD is not set
2719a2757,2759
> CONFIG_FB_SYS_FILLRECT=y
> CONFIG_FB_SYS_COPYAREA=y
> CONFIG_FB_SYS_IMAGEBLIT=y
2720a2761,2762
> CONFIG_FB_SYS_FOPS=y
> CONFIG_FB_DEFERRED_IO=y
4340a4383
> CONFIG_INTERVAL_TREE=y

mrmazda · 09-27-2022, 10:27 PM

I don't build, but according to your lspci you have AMD Mullins, so "CONFIG_DRM_AMDGPU_CIK is not set" and "CONFIG_DRM_AMDGPU is not set" look wrong. Mullins is CIK/GCN2, officially experimental, but working on my two GCN2 chips, Kaveri and Godavari, which directly preceded Mullins.

derguteweka · 09-28-2022, 03:25 AM

Moin,

Thanks a lot. Seem's this goes into the right direction.
It just looks like the kernel crashes again, because output stops at around showing timestamp [ 0.717414] within the SATA initialisation on the screen. Picture is frozen, no scrolling, when i hit return key multiple times.
But - great success: Machine is reachable via ssh, and dmesg tells:

Code:

[    4.248757] cik_cp: Failed to load firmware "radeon/MULLINS_pfp.bin"
[    4.248763] [drm:cik_init [radeon]] *ERROR* Failed to load firmware!
[    4.248888] radeon 0000:00:01.0: Fatal error during GPU init
[    4.248895] [drm] radeon: finishing device.
[    4.256142] [drm] radeon: ttm finalized
[    4.256497] radeon: probe of 0000:00:01.0 failed with error -2

yeah, looks like i'm able to proceed now :-)
Thanks again!

Cheers
WK

derguteweka · 09-28-2022, 04:05 AM

Moin,

Yupp, it also needed a couple of firmware blobs, now all looks good (i need my glasses again to read the screen :-)).
I placed all the MULLINS*bin and bonaire*bin files i found at:
https://anduin.linuxfromscratch.org/...rmware/radeon/

to /lib/firmware/radeon.
Seems to take a lot of time to crank up all the gpu crap, have a look at the 26sec gap in dmesg; so being patient is a good strategy here:

Code:

[    1.304264] hid-generic 0003:04D9:1133.0003: input,hidraw2: USB HID v1.10 Mouse [HID 04d9:1133] on usb-0000:00:10.0-2/input0
[    1.385593] tsc: Refined TSC clocksource calibration: 1996.255 MHz
[    1.385658] clocksource: tsc: mask: 0xffffffffffffffff max_cycles: 0x398cb85cd06, max_idle_ns: 881590416994 ns
[    1.385751] clocksource: Switched to clocksource tsc
[   27.204620] udevd[180]: starting version 3.2.11
[   27.453666] udevd[181]: starting eudev-3.2.11
[   27.638802] r8169 0000:01:00.0 enp1s0: renamed from eth0

cheers,
WK