Ryzen 9950X integrated GPU does not wake from power savings [solved]

graysky · 2024-12-22 23:54:58

Has anyone experienced an integrated GPU fail to wake the monitor from power savings? In this case, it is CPU is a Ryzen 9950X. I ssh'ed in the box and had `journalctl -f` running but did not see anything to indicate a problem. I am using xf86-video-amdgpu-23.0.0-2. I am failing to build the git version from the AUR. Any thoughts are welcomed.

------------[ cut here ]------------
WARNING: CPU: 29 PID: 468 at drivers/gpu/drm/amd/amdgpu/../display/dc/dc_helper.c:100 generic_reg_update_ex+0x1d2/0x290 [amdgpu]
Modules linked in: overlay dm_crypt cbc encrypted_keys trusted asn1_encoder tee amd_atl intel_rapl_msr intel_rapl_common snd_hda_codec_realtek kvm_a>
 hid_microsoft ff_memless hid_generic nvme crc32c_intel nvme_core usbhid nvme_auth amdgpu video wmi amdxcp i2c_algo_bit drm_ttm_helper ttm drm_exec >
CPU: 29 UID: 0 PID: 468 Comm: kworker/29:1H Not tainted 6.12.6-arch1-1 #1 be8168881006593767299fff7299891c69c41600
Hardware name: Micro-Star International Co., Ltd. MS-7E16/X670E GAMING PLUS WIFI (MS-7E16), BIOS 1.93 12/02/2024
Workqueue: events_highpri dm_irq_work_func [amdgpu]
RIP: 0010:generic_reg_update_ex+0x1d2/0x290 [amdgpu]
Code: 97 15 74 4a 41 c7 47 0c 00 00 00 00 48 8d 04 40 49 8d 04 87 44 89 60 15 44 89 68 19 44 89 70 1d 41 83 47 54 01 e9 7b ff ff ff <0f> 0b e9 87 fe>
RSP: 0018:ffffb192c62c7af8 EFLAGS: 00010246
RAX: ffffb192c62c7b20 RBX: ffff98920be36400 RCX: 0000000000000000
RDX: 0000000000000001 RSI: 0000000000000001 RDI: ffff98920be36400
RBP: ffffb192c62c7b78 R08: 0000000000000000 R09: 0000000000000064
R10: 000000000000000c R11: 000000000000000f R12: 0000000000000000
R13: 0000000000000000 R14: ffff98920d800000 R15: ffff989201104800
FS:  0000000000000000(0000) GS:ffff98a87fe80000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000071e3714ff018 CR3: 0000000df4022000 CR4: 0000000000f50ef0
PKRU: 55555554
Call Trace:
 <TASK>
 ? generic_reg_update_ex+0x1d2/0x290 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 ? __warn.cold+0x93/0xf6
 ? generic_reg_update_ex+0x1d2/0x290 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 ? report_bug+0xff/0x140
 ? handle_bug+0x58/0x90
 ? exc_invalid_op+0x17/0x70
 ? asm_exc_invalid_op+0x1a/0x20
 ? generic_reg_update_ex+0x1d2/0x290 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 ? dm_read_reg_func+0x57/0xe0 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 ? generic_reg_get2+0x26/0x50 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 dce_aux_configure_timeout+0x100/0x220 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 try_to_configure_aux_timeout+0x7a/0xe0 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 retrieve_link_cap+0x71/0xda0 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 detect_link_and_local_sink+0xbe4/0x1090 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 link_detect+0x38/0x4e0 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 ? dal_gpio_destroy_irq+0x25/0x40 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 ? query_hpd_status+0x6e/0xa0 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 handle_hpd_irq_helper+0x116/0x190 [amdgpu c91cd080e1155e0debec7c4b8ca50ccbb8b029cb]
 process_one_work+0x17b/0x330
 worker_thread+0x2ce/0x3f0
 ? __pfx_worker_thread+0x10/0x10
 kthread+0xcf/0x100
 ? __pfx_kthread+0x10/0x10
 ret_from_fork+0x31/0x50
 ? __pfx_kthread+0x10/0x10
 ret_from_fork_asm+0x1a/0x30
 </TASK>
---[ end trace 0000000000000000 ]---

Last edited by graysky (2025-01-08 18:05:27)

V1del · 2024-12-23 15:04:01

Tried without xf86-video-amdgpu to just opt for modesetting?

graysky · 2024-12-23 15:29:35

I have not... will give it a whirl.

graysky · 2024-12-23 23:33:59

Did not help... I am not convinced that these 16 core chips are stable with an iGPU.

Last edited by graysky (2024-12-25 14:30:45)

graysky · 2024-12-25 14:19:10

I'd like to find a last-good kernel so I can try to bisect and stepped by all the way to 6.6.1 but found the bug there too. I will say that when I used the first step back to 6.9.11, I did not see *anything* in journalctl's output when the monitor goes dead. I also saw nothing in /var/log/Xorg.0.log.

The monitor going dead is absolutely triggered by xfce4-power-manager putting the display to sleep. It sometimes wakes up. Most of the time it just stays in sleep mode until I change to another tty. I see Xorg et al running but I cannot change back to tty-2 on which it is running and have an active signal it seems.

Does anyone have a suggestion for other packages that could be causing this if not the kernel? Love to find a log that has something useful on which to go.

Last edited by graysky (2025-01-06 21:24:43)

graysky · 2025-01-07 16:07:58

Some progress... it seems whatever is to blame is unique to my DE (xfce4) or to xorg-xserver. If I run KDE (which uses wayland), the display goes to sleep and wakes up as expected.

Hypothesis 1: something in xfce4 is the blame
Hypothesis 2: xorg-xserver is to blame

To test hypothesis 1, I went back 1 version of xfce4* using ALA set to 05-Nov-2024 (last version of xfce4 stuff was and 4.18.4-3).

pacman -Qg xfce4 xfce4-goodies | sed ':a;N;$!ba;s/\n/ /g'

Now, I am able to put the monitor to sleep and wake up just fine. I am wondering how to figure out which of the 26 xfce4 packages is to blame in order to make a bug reports upstream. I guess update one-at-a-time and test. Maybe start with xfce4-power-manager.

Last edited by graysky (2025-01-07 16:38:30)

Bicyforecastle · 2025-01-08 15:06:23

I had a very similar problem a couple months ago, and it turned out to be a problem with, of all things, the kernel module for my network adapter.

Check the output of "lspci -k | grep -i network". Do you have a Mediatek MT7922 network adapter? If so, you could be seeing the same problem as me and several others.

Here is a Reddit thread and an Archwiki thread (and another) which suggest some workarounds. Ultimately, what worked for me was to create /usr/lib/systemd/system-sleep/mt7922_suspend_fix.sh with the following contents:

                                                                                                
#!/usr/bin/env bash
case ${1} in
    pre)
    rfkill block all
    echo "Killed wifi/bluetooth before suspend. Reason: workaround mt7922 breaks suspend with kernel 6.11"
    ;;
    post)
    rfkill unblock all
    echo "Started wifi/bluetooth after suspend. Reason: workaround mt7922 breaks suspend with kernel 6.11"
    ;;
esac

graysky · 2025-01-08 15:11:20

Thanks for the reply, this board has RealTek adapters and I think the issue you're referencing is system sleep (ie suspend) whereas this problem is just when the display goes to sleep.

eth0: RTL8125B
eth1: RTL8126A

Last edited by graysky (2025-01-08 15:11:49)

graysky · 2025-01-08 18:04:34

graysky wrote:

Now, I am able to put the monitor to sleep and wake up just fine. I am wondering how to figure out which of the 26 xfce4 packages is to blame in order to make a bug reports upstream. I guess update one-at-a-time and test. Maybe start with xfce4-power-manager.

After some time I determined that xfce4-settings was to blame. Building the git package from the AUR seems to have fixed it. Notable is the following commit report I found through reading git log -- xfsettingsd/displays-x11.c

https://gitlab.xfce.org/xfce/xfce4-sett … issues/567

That damn kernel message in my first post sent me down a few rabbit holes (kernel then video driver). Glad it's fixed now.

Last edited by graysky (2025-01-08 18:05:13)

glenjo · 2025-01-09 03:52:38

I've had the exact same issue on an intermittent basis since the 4.20 upgrade. I've gotten real good at blind logging into VT2 and rebooting.

graysky · 2025-01-09 11:01:09

What video card are you using and what driver? I only see this when using the iGPU in the 9950X. In any case, does using xfce4-settings-git from the AUR fix the issue for you?

Arch Linux

#1 2024-12-22 23:54:58

Ryzen 9950X integrated GPU does not wake from power savings [solved]

#2 2024-12-23 15:04:01

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#3 2024-12-23 15:29:35

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#4 2024-12-23 23:33:59

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#5 2024-12-25 14:19:10

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#6 2025-01-07 16:07:58

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#7 2025-01-08 15:06:23

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#8 2025-01-08 15:11:20

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#9 2025-01-08 18:04:34

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#10 2025-01-09 03:52:38

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

#11 2025-01-09 11:01:09

Re: Ryzen 9950X integrated GPU does not wake from power savings [solved]

Board footer