please post the bug report as @mattiaformichetti mentioned so there is a higher chance of nvidia team replying.
They are already there.
1 Like
Arch Linux
Kernel: Linux 6.10.4-arch2-1
Iāve just tested driver version 560.31.02 and Iām still getting the same crash as detailed on this thread:
Graphics crash / reset on using xrandr or kscreen-doctor to alter the geometry of the displays. The crash doesnāt happen when only switching refresh rates and not altering geometry.
nvidia-bug-report.log.gz (1.0 MB)
Let me know if there is any other information I can provide that would help.
Iām currently stuck on driver 550.107.02 where the crashing does not occur.
Thanks.
Linux kernel 6.10-5-xanmod.
NixOS.
Nvidia 560.31.02 open driver.
After waking up from sleep, I have this errors in dmesg:
sudo dmesg | grep nvidia
[17841.323602] CPU: 1 PID: 72886 Comm: nvidia-sleep.sh Tainted: G W O 6.10.5-Veil-LTO #1-NixOS
[17841.323665] nv_revoke_gpu_mappings_locked+0x4a/0x70 [nvidia]
[17841.323766] nv_set_system_power_state+0x1bf/0x480 [nvidia]
[17841.323827] nv_procfs_write_suspend+0x129/0x170 [nvidia]
[17841.324084] evdev input_leds tiny_power_button mousedev rapl wmi_bmof mxm_wmi acpi_cpufreq k10temp i2c_piix4 crc16 razermouse(O) rfkill dca soundcore led_class joydev mac_hid button i915 drm_buddy mdev ttm drm_display_helper cec intel_gtt i2c_algo_bit tun tap macvlan bridge stp llc af_packet wireguard ip6_udp_tunnel udp_tunnel libchacha20poly1305 poly1305_x86_64 chacha_x86_64 curve25519_x86_64 libcurve25519_generic libchacha uhid tls vfio_pci vfio_pci_core vfio_iommu_type1 vfio iommufd msr z3fold lz4 lz4_compress fuse efi_pstore configfs nfnetlink dmi_sysfs nls_iso8859_1 nls_cp437 vfat fat dm_mod dax sd_mod btrfs hid_generic libcrc32c crc32c_generic nvidia_uvm(O) usbhid nvidia_drm(O) raid6_pq xor nvidia_modeset(O) kvm_amd nvme ahci libahci nvme_core nvme_auth t10_pi libata kvm video tpm_crb xhci_pci crc64_rocksoft xhci_pci_renesas crc64 crc_t10dif crct10dif_generic scsi_mod tpm_tis crct10dif_pclmul crc32c_intel xhci_hcd ccp scsi_common backlight crct10dif_common rtc_cmos wmi tpm_tis_core nvidia(O) hid efivarfs tpm
[17841.324149] CPU: 1 PID: 72886 Comm: nvidia-sleep.sh Tainted: G W O 6.10.5-Veil-LTO #1-NixOS
[17841.324200] nv_revoke_gpu_mappings_locked+0x4a/0x70 [nvidia]
[17841.324263] nv_set_system_power_state+0x1bf/0x480 [nvidia]
[17841.324322] nv_procfs_write_suspend+0x129/0x170 [nvidia]
[17841.324514] evdev input_leds tiny_power_button mousedev rapl wmi_bmof mxm_wmi acpi_cpufreq k10temp i2c_piix4 crc16 razermouse(O) rfkill dca soundcore led_class joydev mac_hid button i915 drm_buddy mdev ttm drm_display_helper cec intel_gtt i2c_algo_bit tun tap macvlan bridge stp llc af_packet wireguard ip6_udp_tunnel udp_tunnel libchacha20poly1305 poly1305_x86_64 chacha_x86_64 curve25519_x86_64 libcurve25519_generic libchacha uhid tls vfio_pci vfio_pci_core vfio_iommu_type1 vfio iommufd msr z3fold lz4 lz4_compress fuse efi_pstore configfs nfnetlink dmi_sysfs nls_iso8859_1 nls_cp437 vfat fat dm_mod dax sd_mod btrfs hid_generic libcrc32c crc32c_generic nvidia_uvm(O) usbhid nvidia_drm(O) raid6_pq xor nvidia_modeset(O) kvm_amd nvme ahci libahci nvme_core nvme_auth t10_pi libata kvm video tpm_crb xhci_pci crc64_rocksoft xhci_pci_renesas crc64 crc_t10dif crct10dif_generic scsi_mod tpm_tis crct10dif_pclmul crc32c_intel xhci_hcd ccp scsi_common backlight crct10dif_common rtc_cmos wmi tpm_tis_core nvidia(O) hid efivarfs tpm
[17841.324582] CPU: 1 PID: 72886 Comm: nvidia-sleep.sh Tainted: G W O 6.10.5-Veil-LTO #1-NixOS
[17841.324646] nv_revoke_gpu_mappings_locked+0x4a/0x70 [nvidia]
[17841.324738] nv_set_system_power_state+0x1bf/0x480 [nvidia]
[17841.324830] nv_procfs_write_suspend+0x129/0x170 [nvidia]
This errors also happens for me in 555 driver. Iām not sure how exactly it affects system, because SDDM and KDE Wayland session works after sleep.
Also, any news about this bug?
nvidia-bug-report.log.gz (507.4 KB)
Still seeing black screen tearing when screen sharing under Nvidia Wayland on Gnome, even with the 560 driver. (GTX 1060)
NVIDIA, did you guys reproduce this? any fix expected?
2 Likes
Drivers do not pull all listed resolutions from EDID for Apple Pro Display XDR.
Just curious if anyone has a deeper understanding of what may be in the works, or wrong with, the nvidia drivers as they exist today. Even though an edid-decode shows proper resolutions for my monitor, wlr-randr, xrandr, and nvidia donāt seem to load them properly, in X or Wayland.
I have a 6k apple XDR monitor, and it wonāt go to the 6016x3384 res (goes to 5120x2880 max). I tried a custom edid, custom config, all the stuff Iāve found through various threads, but nothing seems to work. Seems like an nvidia driver limitation? Itās a long shot, but thought Iād ask if anyone has any insights here.
xdr-bin.tar.gz (477 Bytes)
Tested on cachyos, 560.31.02 - but hasnāt ever worked on fedora or other distros as well
Will egl-wayland 1.1.15 be bundled with the next driver release?
560.31.02 has egl-wayland 1.1.13 bundled with it which is seemingly why everything is a crash-y mess.
Hyprlandās newest update on 560 with arch using waybar:
Aug 18 13:10:54 archdesk kernel: [drm:nv_drm_atomic_apply_modeset_config.isra.0 [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Failed to initialize semaphore for plane fence
Aug 18 13:10:54 archdesk kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Failed to apply atomic modeset. Error code: -11
Aug 18 13:11:02 archdesk kernel: [drm:nv_drm_atomic_apply_modeset_config.isra.0 [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Failed to initialize semaphore for plane fence
Aug 18 13:11:02 archdesk kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Failed to apply atomic modeset. Error code: -11
Aug 18 13:11:12 archdesk kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Flip event timeout on head 0
Aug 18 13:11:15 archdesk kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Flip event timeout on head 1
Aug 18 13:11:18 archdesk kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Flip event timeout on head 0
Aug 18 13:11:21 archdesk kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Flip event timeout on head 1
1 Like
@jrgiacone I found very similar logs using sway-git that I posted above 560 release feedback & discussion - #151 by krakow20 and also experienced the same behaviour on hyprland but did not look at the logs.
- Were you using multiple monitors?
- Did each screen freeze when you moved the mouse?
Youāre the first person Iāve seen post the same issue so Iām wondering if itās some environment variables.
Hereās what I have set:
~/.config/hypr/hyprland.conf
env = LIBVA_DRIVER_NAME,nvidia
env = GBM_BACKEND,nvidia-drm
env = __GLX_VENDOR_LIBRARY_NAME,nvidia
env = __GL_GSYNC_ALLOWED, 1
env = __GL_VRR_ALLOWED, 1
#env = WLR_NO_HARDWARE_CURSORS,1
env = WLR_DRM_NO_ATOMIC,1
#env = WLR_RENDERER,vulkan
/etc/environment
GBM_BACKEND=nvidia-drm
__GLX_VENDOR_LIBRARY_NAME=nvidia
~/.bash_profile
export __GL_SYNC_DISPLAY_DEVICE=DP-2
Most of these are left over x11 hacks, GBM_BACKEND=nvidia-drm is for the nvidia-vaapi-driver
Notably I am not using the WLR_RENDERER=vulkan env var on sway, going to try that out right now and report back.
Reporting back: Setting WLR_RENDERER=vulkan makes the issue (screens individually freezing within seconds of moving the mouse) go away! However, it seems like itās because the vulkan renderer does not support explicit sync in the first place as seen here: Draft: render/vulkan: add support for explicit sync (!4768) Ā· Merge requests Ā· wlroots / wlroots Ā· GitLab
@krakow20 If I am reading it right, it looks like someone from Nvidia found the root cause: 560 release feedback & discussion - #210 by abchauhan
1 Like
Hi @alyxk, Can you please post the stack trace of the panic if available? I can check if we have a bug filed already. Thanks
Hi @victorr2007 ,
Was this bug report collected after setting NVreg_PreserveVideoMemoryAllocations=1? Your logs show that it wasnāt applied.
I see corruption on desktop after resuming from suspend when NVreg_PreserveVideoMemoryAllocations is not set. There are no issues with NVreg_PreserveVideoMemoryAllocations=1. Please retest and verify that it gets applied. Please run the following command to verify :-
# cat /proc/driver/nvidia/params | grep PreserveVideoMemoryAllocations
Thank you
2 Likes
Ah the post was edited with more entries! Thanks for pointing that out, I missed that. This is the best feedback I have ever seen from nvidia, really awesome stuff!
3 Likes
Thank you, you were very helpful.
Added path to /etc/modprobe.d/nvidia-power-management.conf file
to the /etc/dracut.conf.d/10-nvidia560.conf file
The new file now looks like this
add_drivers+=" nvidia nvidia-drm nvidia-modeset nvidia-uvm "
install_optional_items+=" /etc/modprobe.d/nvidia-power-management.conf "
install_optional_items+=" /etc/modprobe.d/nvidia560-blacklist-nouveau.conf "
install_optional_items+=" /etc/modprobe.d/nvidia560-wayland.conf "
To the /etc/modprobe.d/nvidia-power-management.conf file
options nvidia NVreg_PreserveVideoMemoryAllocations=1 NVreg_TemporaryFilePath=/var/tmp
To the /etc/modprobe.d/nvidia560-blacklist-nouveau.conf file
blacklist nouveau
To the /etc/modprobe.d/nvidia560-wayland.conf file
options nvidia_drm modeset=1 fbdev=1
Now exiting sleep mode occurs without errors.
1 Like
Well the last beta release is really good. Mozilla works perfectly now in Wayland, also works hardware video decoding. Recently recompiled ffmpeg also to use CUDA optimizations and result is awesome. Thank you a lot, NVidia! I still hope, that nvidia-settings will get fixed also to be +/- same functional as it was in X.org, allowing manual power management policy selection, image quality selection etc.
By the way, any new beta/release planned any time soon?
From time to time, during a work in Ubuntu 24.04, my screen/mouse/keyboard freezes and syslog shows this
2024-08-20T19:35:18.662981+02:00 rog kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
2024-08-20T19:35:21.798987+02:00 rog kernel: [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
A workaround (but only temporary one) is switching to another console by Ctrl + Alt + F1 and back to Ctrl + Alt + F2.
Havenāt tried ver. 555. Iām testing 560.31.02
nvidia-bug-report.log.gz (841.9 KB)
@abchauhan Regarding unstable frametimes on Wayland:
- Happens on at least 555+.
- Wayland only, frametimes on X11 are āfineā (not counting usual X11 stutter shenanigans on desktop).
- Worse at higher refresh rates, but issue still present at 60hz.
- Significantly worse with default dynamic power management. Too sensitive to VRAM clock fluctuations compared to X11.
- Same issue on 3 compositors tried. Kwin, hyprland and Mutter.
- Canāt mantain stable 180hz for longer than ~5-10s without at least dropping a couple of frames. Even at low loads on max clocks.
- Tested on RTX 3090 and RTX 2080 Ti
1 Like
Theyāve already determined root cause for this one, see here: 560 release feedback & discussion - #210 by abchauhan
@mirao - Iāve also experienced this bug with the 555 drivers (I havenāt tried 560 yet), and I can workaround it by disabling Night Light in Ubuntu 24.04. Have you tried toggling Night Light on and off to see if this triggers your issue?
Edit: Ah, have just seen the post above by jrgiacone. Looks like this will be fixed in an upcoming driver release. Great news.