wl-screenrec: nvidia (vdpau) drivers gives Segmentation fault (core dumped)
OS: Arch Linux Kernel: 6.3.9-arch1-1 DE: Hyprland GPU: NVIDIA RTX 2070 SUPER
wl-screenrec-git r85.7ceafa9-1 libva-vdpau-driver 0.7.4-6 nvidia-dkms 535.54.03-1 nvidia-utils 535.54.03-1 lib32-nvidia-utils 535.54.03-2 hyprland-nvidia-git r3023.e96fcb31-1
When running wl-screenrec --verbose
[file @ 0x563a0ecd3340] Setting default whitelist 'file,crypto,data'
[AVHWDeviceContext @ 0x563a0ecd3900] libva: VA-API version 1.18.0
[AVHWDeviceContext @ 0x563a0ecd3900] libva: User environment variable requested driver 'nvidia'
[AVHWDeviceContext @ 0x563a0ecd3900] libva: Trying to open /usr/lib/dri/nvidia_drv_video.so
[AVHWDeviceContext @ 0x563a0ecd3900] libva: Found init function __vaDriverInit_1_13
[1] 104588 segmentation fault (core dumped) wl-screenrec --verbose
I also tried wl-screenrec --dri-device /dev/dri/renderD128 --low-power off --verbose
[file @ 0x564d336fcc80] Setting default whitelist 'file,crypto,data'
[AVHWDeviceContext @ 0x564d336fd240] libva: VA-API version 1.18.0
[AVHWDeviceContext @ 0x564d336fd240] libva: User environment variable requested driver 'nvidia'
[AVHWDeviceContext @ 0x564d336fd240] libva: Trying to open /usr/lib/dri/nvidia_drv_video.so
[AVHWDeviceContext @ 0x564d336fd240] libva: Found init function __vaDriverInit_1_13
[1] 104724 segmentation fault (core dumped) wl-screenrec --dri-device /dev/dri/renderD128 --low-power off --verbose
About this issue
- Original URL
- State: closed
- Created 3 years ago
- Comments: 26 (13 by maintainers)
Will be fixed by #95
This project does not use xdp at all–and the issue here isn’t capture, but the encode part.
I believe it is possible to support nvidia gpus, and I’ve started work on it–we’re going to have to not use libva of course
My understanding is there is no libva driver for nvenc–however ffmpeg does support nvenc and hopefully supports importing dmabuf, so this is a feature that could be added.
Is there a reason you want to use your nvidia GPU for encode, especially considering your intel GPU is likely used for compositing so it would probably result in more load to copy the surfaces to the nvidia gpu for encode, considering there is a perfectly fine encoder block on the intel GPU?
The reason “intel” doesn’t work is the name for the intel driver is
iHD. You can see what drivers are present on your system withls /usr/lib/dri/*_video.so