libva error: va_getDriverName() failed with unknown libva error,driver_name=(null) #278
Comments
|
I havent tried VM before, we need more information, could you help to enable LIBVA_TRACE and attach the report? |
|
I don't think that any libva's driver was designed to work virtualized. Perhaps somebody should try to develop a libva's driver that proxies to the call to the host libva's and composite with the virtualized video driver... but it sounds hard |
|
if the HW is exposed, and can be access, work with i915, I think UMD driver can work with VM. it is transparent for UMD driver. from intel/media-driver#520 some one tried VM with virtual mediated pass-through GPU (Intel GVT-g). it works with i965. |
|
of course, it also depend on which kind of VM @LucasZanella are using. just support several , such as KVM |
|
I am experiencing this issue also, and I'm NOT using a VM! I experience it using VLC both 3.0.x and 4.0-dev VLC has three plugins for connecting to vaapi: wayland, x11 and drm; wayland and x11 have higher priority and so are tried first before falling back to drm. On my system, the wayland one ends early because the VLC window is not wayland (I don't think wayland support is complete yet); then the x11 one is tried, which experiences the issue reported here and thus fails; it then tries the drm one which works. the following is the relevant portion of output from VLC trying the x11 one (with since when VLC tries the drm one it overwrites the trace file, I modified VLC 4.0-dev to deliberately break the drm one to stop it doing so thus I could see what it contained (as you asked for one); unfortunately it only contains this: if I try using with the same useless trace contrastingly the output from the drm based attempt gave the following and works successfully (with just and this with |
|
I've simliar issue (chromium won't start) same error and I'm not on the Virtual machine, btw where can find the info about driver - i.e to specify driver name? I've intel UHD 620 |
|
here's my output does anything here seem to be wrong? |
|
Any updates on this? I'm experiencing this issue on wayland. |
|
hi @pavel-the-best not sure whether my understanding correct. there are two issue
|
|
I am getting: This is with kernel 5.8-rc2+, Debian Bullseye, Mesa 20.1.1, gcc 9.3.0 and amdgpu driver. Both Chromium and VLC have problems, Chromium won't show start page while VLC segfaults when playing video. Edit: building and installing 2.8.0 fixes VLC but Chromium is still broken in 5.8-rc2+: it can't show even settings, just error. |
|
@ipr Just want to clarify it. Chromium will broken when use VLC plugin to playback. Right? IMHO, Google doesn't provide media HW acceleration directly in chromium and there is one non-official patch from community. So I guess you use VLC plugin in chromium. |
|
Same issue here running OpenShot on Debian Unstable. Trying to export using the VA-API and it can't find the hardware device. |
|
I have this issue with an amd card. I am using mpv's vaapi video output driver for testing. Manually exporting To reproduce I do I get the following error message: Trace file: https://0x0.st/i7c1.thd-0x0 Here is what does work: Trace file: https://0x0.st/i7cI.thd-0x0 This file is the output of I can manually cause vainfo to fail, with a different error, by doing Trace file: https://0x0.st/i7cU.thd-0x0 I'm using libva 2.9.1 and libva-utils 2.9.1 EDIT: On my laptop with an intel GPU, I do not get this issue. libva is properly detecting iHD and loading it. |
|
Some more information, The issue here seems to be that, under Xwayland, libva cannot load radeonsi, whereas it loads fine in wayland or drm. |
|
Same error on Windows running WSL2 Ubuntu 20.04 using VcXSrv: $ uname -a Same error: Tried: Same output at running Node/electron app: |
|
Same error for Google chrome. |
how about unset LIBVA_TRACE? |
|
@LaserEyess about AMD , it should because the AMD new driver name "radeonsi" |
|
@gauravguptabits , could you help to share more information, what's your gpu? seems it is nvidia, how about your kmd , it is also nouveau? |
|
Trying to run the Freezer from the terminal in Fedora 34 throws a similar error.
|
|
hi @PranavBhattarai , how about the system configuration, it looks like it could not find the compatible driver name against KMD. but you always could use LIBVA_DRIVER_NAME to specify the driver name. |
I don't think I will see more error logs from now onwards except when I launch Freezer from terminal which I think use Nautilus as u can see the last journal log. I have use sudo dnf autoremove which cleaned up a lot of things. But when I type the command which you can see below, I still see Libva as installed package. Intel Chips inside my Dell:
Don't know how to use the LIBVA_DRIVER_NAME environment variable since I'm noob (let me know in simple steps). If there is anything that I can help to solve this 2 YEAR old issue, do let me know. |
|
could you help to try this command line? |
|
Not worked, sorry :( |
|
could you help also do the same thing like: |
|
same error on env $ |
|
@XinfengZhang I think I started to see these new bugs after when I added the environment variable like you asked, using set -U LIBVA_DRIVER_NAME iHD command. Probably because iHD might not be the right value I should use. What value I should set instead of iHD? How do I know which one to write?
Edit: Depending upon which graphics, it might differ. I have this graphics "Mesa Intel® HD Graphics 5500 (BDW GT2)". There some of something other libva driver name for this value! |
|
it is driver name, you could refer to this: |
I dont think you could run libva on WSL2, maybe you could try to check whether there are device node in /dev/dri/ firstly |
As far as I know WSL exposes "/dev/dxg". Even if you rename /dev/dxg to /dev/dri/card0 it hardly will work because media driver relies on specific behavior of i915 kernel mode driver while /dev/dxg represents a different KMD. |
|
|
@MuhammedTaha02 ， your config is same as PranavBhattarai's , so suppose his solution should works, how about your platforms? what's the display adaptor ? maybe you could use "lspci -x" to check “VGA compatible controller” |
|
I think I found a final solution to my problem. It is VGA compatible. And the kernel driver I'm using is i915. |
@PranavBhattarai No. i915 is a Intel GPU kernel-mode-driver (KMD) for Linux. LIBVA/VAAPI drivers are user-mode-drivers (UMD)... which eventually interact with the GPU KMD via DRM (libdrm). Only i965 and iHD are valid for Intel LIBVA/VAAPI UMD. |
|
I too am getting this error:
|
|
I used to have this issue on wayland and obviously setting |
|
yes, vaGetDisplay for X11, vaGetDisplayWl for Wayland , vaGetDisplayDRM for drm, vaGetDisplayGLX for glx . the sample is https://github.com/intel/libva-utils/blob/master/common/va_display.c#L43-L52, each init function have different vaGetDisplayXXX |
|
I'm trying to start chrome on rocky linux 8 guest on a VM VirtualBox and i have the same error I have resolved the libva erorr with maybe you need to add this driver name to the list? Now I have only this error |
|
I've had this since maybe june on manjaro. Happens with any chromium based browser including electron. It was working fine before that. Anyone have any ideas? |
|
@xenoterracide *odds are you would have to track down which component and version caused the regression. If it turns out to be driver specific (aka not in libva itself), do report it on the respective project i965 or iHD. If it's opera specific you'll have to find their bugtracker. |
|
Actually the problem turned out to be that if I had both of those drivers installed the problem happens turns out I needed the media driver's package but if the other package was installed it got confused regardless of what I set the environment variable to be. I didn't used to need the media's driver package so when I upgraded... They switched out packages on me. |
|
as I am not sure which of those components is the root cause I also want to point to a current issue with chromium (and chromium based) browsers: |
|
I am definitely not using a VM, I am using Debian 12 bookworm. When I run teams like
I see this error periodically:
It does not stop teams or chrome from working, but it is noise in the terminal. |
|
The latest libva release includes a bunch of fixes, so make sure you use that - not sure if Debian 12 ships it or not. I've opened a few more MRs that help, although landing anything in libva (and related projects) is as easy as milking a rock |
|
hi, i have same error, my hardware specs: |
|
I didn't put down the stack trace because we have seen enough of those. Obviously, mine crashed as well. Here's how I looked up the correct parameters and created a command that would allow it to run on my system. used the inxi command to see what drivers are being used on the computer. Mine was obviously nvidia 1. export LIBVA_DRIVER_NAME=nvidia LIBVA_TRACE=1 And it worked. This is on a machine with a nvidia card obviously but you should be able to see what driver is being used, export it and then run the edex file with the listed parameters or customize to your personal setup. |
|
I am getting this exact same error on Ubuntu 22.04.5 64-bit with GNOME Version 42.9 and X11 windowing. I'm not really a Linux expert and am trying to run some eye tracking software (Tobii Pro Eye Tracker Manager) on this system. Trying to run it in this system returns this trace: deborah@tobii-pro-spectrum:~$ tobiiproeyetrackermanager
[2025-01-07T04:19:51.541Z] INFO: Main/47339 on tobii-pro-spectrum: APP VERSION: 2.7.2
[2025-01-07T04:19:51.699Z] INFO: Main/47339 on tobii-pro-spectrum: Service initialization done pid: [47378] id: [IpcStreamServer] type: [EtmProcess.IpcServer]
[2025-01-07T04:19:51.713Z] INFO: Main/47339 on tobii-pro-spectrum: Service initialization done pid: [47377] id: [IpcComServer] type: [EtmProcess.IpcServer]
[2025-01-07T04:19:51.717Z] INFO: Main/47339 on tobii-pro-spectrum: IPC Client Socket connected for: EtmModule.ElectronMain-518d05f0-7821-4f35-96e9-73ff82b1fe76
[2025-01-07T04:19:51.931Z] INFO: Main/47339 on tobii-pro-spectrum: Service initialization done pid: [47404] id: [EtmModule.DataPersistence] type: [EtmProcess.EtmModule]
[2025-01-07T04:19:51.932Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.DataPersistence MODULE_INITIALIZED
[2025-01-07T04:19:51.974Z] INFO: Main/47339 on tobii-pro-spectrum: Service initialization done pid: [47403] id: [EtmModule.UpdatesService] type: [EtmProcess.EtmModule]
[2025-01-07T04:19:51.974Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.UpdatesService MODULE_INITIALIZED
[2025-01-07T04:19:52.065Z] INFO: Main/47339 on tobii-pro-spectrum: Service initialization done pid: [47405] id: [EtmModule.SdkService] type: [EtmProcess.Executable]
[2025-01-07T04:19:52.065Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.SdkService MODULE_INITIALIZED
[2025-01-07T04:19:52.072Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.UpdatesService MODULE_INITIALIZED
[2025-01-07T04:19:52.072Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.DataPersistence MODULE_INITIALIZED
[2025-01-07T04:19:52.072Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.SdkService MODULE_INITIALIZED
[2025-01-07T04:19:52.476Z] INFO: Main/47339 on tobii-pro-spectrum: Service initialization done pid: [47452] id: [EtmModule.Telemetry] type: [EtmProcess.EtmModule]
[2025-01-07T04:19:52.476Z] INFO: Main/47339 on tobii-pro-spectrum: EtmModule.Telemetry MODULE_INITIALIZED
[2025-01-07T04:19:52.476Z] INFO: Main/47339 on tobii-pro-spectrum: ALL_ESSENTIAL_MODULES_LOADED
[2025-01-07T04:19:52.477Z] INFO: Main/47339 on tobii-pro-spectrum: Creating main window.
[2025-01-07T04:19:52.525Z] INFO: Main/47339 on tobii-pro-spectrum: **** PATH INDEX **** /opt/TobiiProEyeTrackerManager/resources/app.asar/index.html
[2025-01-07T04:19:52.536Z] ERROR: Main/47339 on tobii-pro-spectrum:
Child process encountered an error: pid: [47405] id: [EtmModule.SdkService] type: [EtmProcess.Executable] SDK:: Recieved unhandled message from parent process: <Buffer@0x6b58410 7b 22 65 76 65 6e 74 22 3a 22 61 6c 6c 2d 65 73 73 65 6e 74 69 61 6c 2d 6d 6f 64 75 6c 65 73 2d 6c 6f 61 64 65 64 22 7d>``
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
libva error: vaGetDriverNameByIndex() failed with unknown libva error, driver_name = (null)
[47339:0107/151953.929332:FATAL:gpu_data_manager_impl_private.cc(415)] GPU process isn't usable. Goodbye.
Trace/breakpoint trap (core dumped)
I have tried all of the fixes on this thread and nothing has worked so far. Further info as requested on threads above: deborah@tobii-pro-spectrum:~$ xdpyinfo | grep DRI
DRI2
DRI3
deborah@tobii-pro-spectrum:~$ echo $XDG_SESSION_TYPE
x11If I run export LIBVA_DRIVER_NAME=X11, I get the same output as above, but minus the message that it failed to get the driver name; instead I get: [51468:0107/160808.677241:FATAL:gpu_data_manager_impl_private.cc(415)] GPU process isn't usable. GoodbyeI have seen a lot of uninformative error messages in my time, but this one really takes the cake. I have asked the Tobii tech team for help, and they referred me to this thread, but to be honest I have not yet found it very helpful, so any assistance would be gratefully appreciated. Best Deborah. |
and others
I'm doing a QT program and I'm getting this error on my Ubuntu (it's a VM):
libva error: va_getDriverName() failed with unknown libva error,driver_name=(null)I tried all possible solutions on web and they didn't work.
I found that
export LIBVA_DRIVER_NAME=i965could work, but with it, I have the following errors:What can be done?
The text was updated successfully, but these errors were encountered: