GTR9 Pro - Linux - System Freezes - amdgpu Pageflip

jotapesse

Hi! I’m using NixOS Linux, kernel 6.17.0, KDE Plasma and I’m getting ocasional system freezes, even on idle system for long, apparently related to the amdgpu kernel driver pageflip handling. Possibly a bug on the amdgpu kernel module driver?

out 09 21:49:52 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:49:52 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:49:52 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
out 09 21:49:53 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:49:53 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:49:53 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
out 09 21:49:53 beelink-gtr9pro kernel: amdgpu 0000:c6:00.0: [drm] *ERROR* [CRTC:86:crtc-0] flip_done timed out
out 09 21:49:55 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:49:55 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:49:55 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
out 09 21:49:58 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:49:58 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:49:58 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
out 09 21:49:59 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:49:59 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:49:59 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
(...)
out 09 21:50:35 beelink-gtr9pro kernel: amdgpu 0000:c6:00.0: [drm] *ERROR* flip_done timed out
out 09 21:50:35 beelink-gtr9pro kernel: amdgpu 0000:c6:00.0: [drm] *ERROR* [CRTC:86:crtc-0] commit wait timed out
(...)
out 09 21:50:45 beelink-gtr9pro kernel: amdgpu 0000:c6:00.0: [drm] *ERROR* flip_done timed out
out 09 21:50:45 beelink-gtr9pro kernel: amdgpu 0000:c6:00.0: [drm] *ERROR* [PLANE:83:plane-7] commit wait timed out
out 09 21:50:45 beelink-gtr9pro kernel: ------------[ cut here ]------------
out 09 21:50:45 beelink-gtr9pro kernel: WARNING: CPU: 16 PID: 1534 at drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:8983 amdgpu_dm_atomic_commit_tail+0x354b/0x35c0 [amdgpu]
out 09 21:50:45 beelink-gtr9pro kernel: Modules linked in: sd_mod scsi_mod scsi_common snd_usb_audio snd_ump snd_usbmidi_lib snd_rawmidi joydev mousedev mc qrtr snd_seq_dummy snd_hrtimer snd_seq snd_seq_device af_packet nls_iso8859_1 nls_cp437 vfat fat amdgpu snd_acp70 snd_acp_i2s snd_acp_pdm snd_acp_pcm snd_sof_amd_acp70 snd_sof_amd_acp63 snd_sof_amd_vangogh snd_sof_amd_rembrandt snd_sof_amd_renoir snd_sof_amd_acp snd_sof_pci snd_sof_xtensa_dsp snd_sof snd_sof_utils snd_pci_ps snd_soc_acpi_amd_match snd_amd_sdw_acpi soundwire_amd mt7925e soundwire_generic_allocation mt7925_common soundwire_bus snd_soc_sdca mt792x_lib mt76_connac_lib snd_soc_core mt76 snd_hda_codec_alc662 snd_hda_codec_realtek_lib snd_hda_codec_atihdmi snd_hda_codec_generic snd_hda_codec_hdmi ixgbe snd_compress amdxcp drm_panel_backlight_quirks ac97_bus drm_buddy mac80211 snd_hda_intel snd_pcm_dmaengine drm_exec drm_suballoc_helper snd_rpl_pci_acp6x xfrm_algo snd_hda_codec sch_fq_codel snd_acp_pci drm_ttm_helper mdio_devres edac_mce_amd of_mdio snd_amd_acpi_mach ttm fixed_phy
out 09 21:50:45 beelink-gtr9pro kernel:  snd_hda_core snd_acp_legacy_common fwnode_mdio edac_core sp5100_tco evdev snd_pci_acp6x amd_atl intel_rapl_msr snd_intel_dspcfg libphy mac_hid wmi_bmof uinput snd_pci_acp5x watchdog intel_rapl_common snd_intel_sdw_acpi drm_display_helper cfg80211 mdio_bus snd_hwdep snd_rn_pci_acp3x polyval_clmulni mdio loop snd_pcm ghash_clmulni_intel snd_acp_config amdxdna cec ptp i2c_piix4 snd_soc_acpi snd_timer pps_core gpu_sched i2c_algo_bit dca rapl snd rfkill i2c_smbus snd_pci_acp3x libarc4 soundcore kvm_amd video thermal tpm_crb rtc_cmos wmi i2c_hid_acpi ccp tpm_tis i2c_hid amd_pmc tpm_tis_core soc_button_array kvm irqbypass atkbd libps2 serio fuse vivaldi_fmap efi_pstore configfs nfnetlink efivarfs dmi_sysfs autofs4 ext4 crc16 mbcache jbd2 dm_crypt encrypted_keys trusted asn1_encoder tee tpm rng_core libaescfb ecdh_generic ecc hid_logitech_hidpp input_leds hid_logitech_dj hid_generic usbhid hid sdhci_pci sdhci_uhs2 sdhci xhci_pci nvme cqhci thunderbolt xhci_hcd nvme_core mmc_core aesni_intel nvme_keyring nvme_auth
out 09 21:50:45 beelink-gtr9pro kernel:  led_class dm_mod dax
out 09 21:50:45 beelink-gtr9pro kernel: CPU: 16 UID: 0 PID: 1534 Comm: systemd-logind Not tainted 6.17.0 #1-NixOS PREEMPT(voluntary) 
out 09 21:50:45 beelink-gtr9pro kernel: Hardware name: AZW GTR Pro/GTR Pro, BIOS GTRP108 09/16/2025
out 09 21:50:45 beelink-gtr9pro kernel: RIP: 0010:amdgpu_dm_atomic_commit_tail+0x354b/0x35c0 [amdgpu]
out 09 21:50:45 beelink-gtr9pro kernel: Code: ff 80 8d a0 fe ff ff 80 e9 33 fe ff ff 90 0f 0b 90 e9 12 f8 ff ff 90 0f 0b 31 f6 e9 6b d0 ff ff 90 0f 0b 90 e9 28 d1 ff ff 90 <0f> 0b 90 e9 15 f8 ff ff 48 c7 85 28 fe ff ff 00 00 00 00 48 c7 85
out 09 21:50:45 beelink-gtr9pro kernel: RSP: 0018:ffffcbe8040f3678 EFLAGS: 00010086
out 09 21:50:45 beelink-gtr9pro kernel: RAX: 0000000000000001 RBX: 0000000000000286 RCX: 0000000000000007
out 09 21:50:45 beelink-gtr9pro kernel: RDX: 0000000000000001 RSI: 0000000000000297 RDI: 0000000000000000
out 09 21:50:45 beelink-gtr9pro kernel: RBP: ffffcbe8040f38d8 R08: 0000000000000001 R09: 0000000000000000
out 09 21:50:45 beelink-gtr9pro kernel: R10: ffff888197f43118 R11: ffffcbe8040f35d4 R12: ffff888197f43118
out 09 21:50:45 beelink-gtr9pro kernel: R13: ffff888147c7a000 R14: ffff888197f43000 R15: 0000000000000000
out 09 21:50:45 beelink-gtr9pro kernel: FS:  00007fb0bd2a7900(0000) GS:ffff888916340000(0000) knlGS:0000000000000000
out 09 21:50:45 beelink-gtr9pro kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
out 09 21:50:45 beelink-gtr9pro kernel: CR2: 00007f890e59e07c CR3: 0000000139243000 CR4: 0000000000f50ef0
out 09 21:50:45 beelink-gtr9pro kernel: PKRU: 55555554
out 09 21:50:45 beelink-gtr9pro kernel: Call Trace:
out 09 21:50:45 beelink-gtr9pro kernel:  <TASK>
out 09 21:50:45 beelink-gtr9pro kernel:  commit_tail+0x91/0x130
out 09 21:50:45 beelink-gtr9pro kernel:  drm_atomic_helper_commit+0x11a/0x140
out 09 21:50:45 beelink-gtr9pro kernel:  drm_atomic_commit+0xae/0xe0
out 09 21:50:45 beelink-gtr9pro kernel:  ? __pfx___drm_printfn_info+0x10/0x10
out 09 21:50:45 beelink-gtr9pro kernel:  drm_client_modeset_commit_atomic+0x201/0x250
out 09 21:50:45 beelink-gtr9pro kernel:  drm_client_modeset_commit_locked+0x5a/0x160
out 09 21:50:45 beelink-gtr9pro kernel:  __drm_fb_helper_restore_fbdev_mode_unlocked+0x5e/0xd0
out 09 21:50:45 beelink-gtr9pro kernel:  drm_fb_helper_set_par+0x30/0x40
out 09 21:50:45 beelink-gtr9pro kernel:  fb_set_var+0x248/0x430
out 09 21:50:45 beelink-gtr9pro kernel:  ? psi_task_switch+0x113/0x290
out 09 21:50:45 beelink-gtr9pro kernel:  ? finish_task_switch.isra.0+0x99/0x2e0
out 09 21:50:45 beelink-gtr9pro kernel:  fbcon_blank+0x236/0x300
out 09 21:50:45 beelink-gtr9pro kernel:  do_unblank_screen+0xc2/0x1b0
out 09 21:50:45 beelink-gtr9pro kernel:  complete_change_console+0x54/0x120
out 09 21:50:45 beelink-gtr9pro kernel:  vt_ioctl+0xee8/0x1480
out 09 21:50:45 beelink-gtr9pro kernel:  ? kernel_termios_to_user_termios_1+0x13/0x20
out 09 21:50:45 beelink-gtr9pro kernel:  ? tty_mode_ioctl+0x198/0x6a0
out 09 21:50:45 beelink-gtr9pro kernel:  tty_ioctl+0xe8/0x8e0
out 09 21:50:45 beelink-gtr9pro kernel:  ? __seccomp_filter+0x42/0x500
out 09 21:50:45 beelink-gtr9pro kernel:  __x64_sys_ioctl+0x93/0xe0
out 09 21:50:45 beelink-gtr9pro kernel:  ? syscall_trace_enter+0x95/0x1a0
out 09 21:50:45 beelink-gtr9pro kernel:  do_syscall_64+0xb7/0x3a0
out 09 21:50:45 beelink-gtr9pro kernel:  entry_SYSCALL_64_after_hwframe+0x77/0x7f
out 09 21:50:45 beelink-gtr9pro kernel: RIP: 0033:0x7fb0bcb1e11d
out 09 21:50:45 beelink-gtr9pro kernel: Code: 04 25 28 00 00 00 48 89 45 c8 31 c0 48 8d 45 10 c7 45 b0 10 00 00 00 48 89 45 b8 48 8d 45 d0 48 89 45 c0 b8 10 00 00 00 0f 05 <89> c2 3d 00 f0 ff ff 77 22 48 8b 45 c8 64 48 2b 04 25 28 00 00 00
out 09 21:50:45 beelink-gtr9pro kernel: RSP: 002b:00007ffdcc698d10 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
out 09 21:50:45 beelink-gtr9pro kernel: RAX: ffffffffffffffda RBX: 0000000000000018 RCX: 00007fb0bcb1e11d
out 09 21:50:45 beelink-gtr9pro kernel: RDX: 0000000000000001 RSI: 0000000000005605 RDI: 0000000000000018
out 09 21:50:45 beelink-gtr9pro kernel: RBP: 00007ffdcc698d60 R08: 0000000000000000 R09: 0000000000000007
out 09 21:50:45 beelink-gtr9pro kernel: R10: 0000000000000007 R11: 0000000000000246 R12: 0000000000000000
out 09 21:50:45 beelink-gtr9pro kernel: R13: 00007ffdcc698e00 R14: 00007ffdcc698df8 R15: 00005648cc5c62a0
out 09 21:50:45 beelink-gtr9pro kernel:  </TASK>
out 09 21:50:45 beelink-gtr9pro kernel: ---[ end trace 0000000000000000 ]---
out 09 21:50:46 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:50:46 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:50:46 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'
out 09 21:50:47 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Pageflip timed out! This is a bug in the amdgpu kernel driver
out 09 21:50:47 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: Please report this at https://gitlab.freedesktop.org/drm/amd/-/issues
out 09 21:50:47 beelink-gtr9pro kwin_wayland[1884]: kwin_wayland_drm: With the output of 'sudo dmesg' and 'journalctl --user-unit plasma-kwin_wayland --boot 0'

jotapesse

Ok, many people on AMD APUs/GPUs seem to be experiencing the same issues for long, likely a amdgpu kernel driver module bug as reported here. Still unresolved since kernel 6.12.

tbocek

jotapesse After adding this to my kernel parameters, I have not seen this flip issue: amdgpu.dcdebugmask=0×10 (https://wiki.archlinux.org/title/AMDGPU)

jotapesse

tbocek Thanks, I’ll try it. I’ve been reading through those bug reports and source documentation. This seems to be a known issue and affects AMD, Intel and Nvidia GPUs as well. It’s a bit overkill but here are all the possible DC_DEBUG_MASK parameters we can pass to the kernel amdgpu driver module with amdgpu.dcdebugmask.

0x1 = DC_DISABLE_PIPE_SPLIT, If set, disable pipe-splitting
0x2 = DC_DISABLE_STUTTER, If set, disable memory stutter mode
0x4 = DC_DISABLE_DSC, If set, disable display stream compression
0x8 = DC_DISABLE_CLOCK_GATING, If set, disable clock gating optimizations
0x10 = DC_DISABLE_PSR, If set, disable Panel self refresh v1 and PSR-SU
0x20 = DC_FORCE_SUBVP_MCLK_SWITCH, If set, force mclk switch in subvp, even if mclk switch in vblank is possible
0x40 = DC_DISABLE_MPO, If set, disable multi-plane offloading
0x80 = DC_ENABLE_DPIA_TRACE, If set, enable trace logging for DPIA
0x100 = DC_ENABLE_DML2, If set, force usage of DML2, even if the DCN version does not default to it.
0x200 = DC_DISABLE_PSR_SU, If set, disable PSR SU
0x400 = DC_DISABLE_REPLAY, If set, disable all Idle Power States, all the time. If more than one IPS debug bit is set, the lowest bit takes precedence. For example, if DC_FORCE_IPS_ENABLE and DC_DISABLE_IPS_DYNAMIC are set, then DC_DISABLE_IPS_DYNAMIC takes precedence.
0x800 = DC_DISABLE_IPS, If set, disable all IPS, all the time, *except* when driver goes into suspend.
0x1000 = DC_DISABLE_IPS_DYNAMIC, If set, disable IPS2 (IPS1 allowed) if there is an enabled display. Otherwise, enable all IPS.
0x2000 = DC_DISABLE_IPS2_DYNAMIC, If set, force enable all IPS, all the time.
0x8000 = DC_DISABLE_ACPI_EDID, If set, don't attempt to fetch EDID for eDP display from ACPI _DDC method.
0x10000 = DC_DISABLE_HDMI_CEC, If set, disable HDMI-CEC feature in amdgpu driver.
0x20000 = DC_DISABLE_SUBVP_FAMS, If set, disable DCN Sub-Viewport & Firmware Assisted Memory Clock Switching (FAMS) feature in amdgpu driver.
0x40000 = DC_DISABLE_CUSTOM_BRIGHTNESS_CURVE, If set, disable support for custom brightness curves
0x80000 = DC_HDCP_LC_FORCE_FW_ENABLE, If set, use HDCP Locality Check FW path regardless of reported HW capabilities.
0x100000 = DC_HDCP_LC_ENABLE_SW_FALLBACK, If set, upon HDCP Locality Check FW path failure, retry using legacy SW path.
0x200000 = DC_SKIP_DETECTION_LT, If set, skip detection link training

Specifically for this issue some have reported that amdgpu.dcdebugmask=0x10 seems to resolve but I also saw other reports that for them it didn’t resolve. I’ll try it and see how it goes.

References:

jotapesse

❌ I just tried it with amdgpu.dcdebugmask=0x200 to check if that option alone would sort it, but no - the same issue persists. Now trying amdgpu.dcdebugmask=0x10. Will report back if that sorted it for me.

jotapesse

❌ Same issue with amdgpu.dcdebugmask=0x10 that doesn’t sort it for me! 😭

jotapesse

Added References:

jotapesse

❌ Nothing seems to work for me. I also tried using to disable the Window Manager Compositor Direct Scan Out option to no avail, same issue. 😭

Direct Scan Out
Finally, especially for integrated GPU users where none of the above successfully abated the amdgpu freeze, another possible ‘fix’ is to disable “Direct Scan-out”. This is not optimal as direct scanout is meant to increase performance and decrease latency. However this is the only thing that worked for me and is preferable to a full graphics lockup. This can be controlled via a global environment variable (such as set in /etc/environment) and varies with window manager.

For kwin:
KWIN_DRM_NO_DIRECT_SCANOUT=1

For other wayland compositors:
WLR_SCENE_DISABLE_DIRECT_SCANOUT=1

chilango

Hey there, same issue on CachyOS. At the moment it is quite annoying. I mentioned it here and usually can reinitiate the session through an ssh connection and systemctl restart of the kde_wayland service. MEntioned this issue also here already:
https://bbs.bee-link.com/d/7762-gtr-9-pro-ethernet-malfunction-under-load/20

Hope there will be a fix soon, but it seems to appear on some, but not all Strix Halo machines.

jotapesse

Hope there will be a fix soon, but it seems to appear on some, but not all Strix Halo machines.

To be fair this is a well known issue and affects not only AMD but also Intel and Nvidia GPUs as well and it seems to be related to how pageflip timeouts are handled between window managers/compositors, the gpu driver module and the kernel.

I’m now trying another possible solution option, under KDE Plasma Kwin, to disable the Vertical Refresh Rate (VRR) via a global environment variable (such as set in /etc/environment):
KWIN_DRM_NO_VRR=1

❌ ~~I’ll report back how that went.~~ Update: It didn’t work, same issue!

At the moment it is quite annoying. I mentioned it here and usually can reinitiate the session through an ssh connection and systemctl restart of the kde_wayland service.

[ℹ] By the way, most times I get a hard crash and have to power cycle the mini PC to recover from it. But when I don’t get a hard crash and it is still recoverable, alternatively to that SSH connection from a different PC, you can simply disconnect and reconnect the display (HDMI, DP or USB-C). Or change to a different virtual console TTY interface (CTRL-ALT-F2,F3,…) , login and restart Kwin Wayland with: $ setsid kwin_wayland --replace & and then logout and return to the original console TTY interface where KDE Kwin is running. Anyway, it is a pain and really unreliable. 😠

jotapesse

✅ Good news everyone! I think I have a workaround, at least on KDE Plasma. The pageflip timeouts handling between window managers/compositors and kernel seem to be direcly related to how the Window Manager/Wayland APIs handles the Atomic Modesetting to the kernel. This isssue doesn’t seem to happen with X11, only on Wayland. But I would not like to go back to X11 if I can.

Specifically for KDE Plasma KWin window manager with Wayland, we can force it to use the Legacy Atomic Modesetting API (disabling the Modern API which seems to have issues), with the following environment variable (and removing all other above variables or kernel module options/parameters):

KWIN_DRM_NO_AMS=1

Since using it I have not had any other crash anymore. A slight downgrade in performace may be in place but it shouldn’t be noticeable. Hopefully this workarounds it long term until Kwin/Wayland improves their APIs to better handle the Atomic Modesetting to the kernel. 🙏 I’m not sure about other Window Managers though.