Comments (17)
gschintgen
commented on June 15, 2024
1
I'll file a PR for the AppImage part of the report:
libva info: VA-API version 1.7.0
libva info: Trying to open /tmp/.mount_sunshiq9efzd/usr/lib/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_7
libva error: /tmp/.mount_sunshiq9efzd/usr/lib/iHD_drv_video.so init failed
libva info: va_openDriver() returns 1
libva info: Trying to open /tmp/.mount_sunshiq9efzd/usr/lib/i965_drv_video.so
libva info: Found init function __vaDriverInit_1_6
libva info: va_openDriver() returns 0
[2024:04:16:00:42:30]: Info:
[2024:04:16:00:42:30]: Info: // Ignore any errors mentioned above, they are not relevant. //
[2024:04:16:00:42:30]: Info:
[2024:04:16:00:42:30]: Debug: ------ h264 ------
[2024:04:16:00:42:30]: Debug: PASSED: supported
[2024:04:16:00:42:30]: Debug: REF_FRAMES_RESTRICT: supported
[2024:04:16:00:42:30]: Debug: CBR: supported
[2024:04:16:00:42:30]: Debug: DYNAMIC_RANGE: unsupported
[2024:04:16:00:42:30]: Debug: VUI_PARAMETERS: supported
[2024:04:16:00:42:30]: Debug: -------------------
[2024:04:16:00:42:30]: Info: Found H.264 encoder: h264_vaapi [vaapi]
from sunshine.
gschintgen
commented on June 15, 2024
1
Thank you for testing and providing the log!
The relevant part seems to be here:
[2024:04:17:11:44:01]: Info: dlopen of /tmp/.mount_sunshigiryh4/usr/lib/radeonsi_drv_video.so failed: /tmp/.mount_sunshigiryh4/usr/lib/radeonsi_drv_video.so: undefined symbol: amdgpu_va_get_start_addr
[2024:04:17:11:44:01]: Error: Couldn't initialize va display: unknown libva error
There are a number of very recent reports of Mesa breakage referring to that same symbol: https://www.google.com/search?q=%22amdgpu_va_get_start_addr%22
That amdgpu_va_get_start_addr symbol seems to be rather recent: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/338. It seems to refer to the amdgpu kernel driver.
As I'm now noticing the PR is not working either on my own AMD system (RX6650XT), but with different errors. (While preparing the PR I was using an Intel iGPU-based laptop and it fixed the issue just fine, so I was cautiously optimistic about it.)
The reason that I get past the amdgpu_va_get_start_addr issue might well be that I'm using a PPA with newer mainline kernel images.
I'll see how it goes if I'm adding the stable kisak mesa PPA instead of the "fresh"er variant. (I thought the more bleeding edge PPA might be advantageous for users with more recent hardware, i.e. RX7xxx.)
For completeness, here's the output on my system:
libva info: VA-API version 1.14.0
libva info: Trying to open /tmp/.mount_sunshiVQNzQT/usr/lib/radeonsi_drv_video.so
libva info: Found init function __vaDriverInit_1_14
libva info: va_openDriver() returns 0
amd: LLVM doesn't support , bailing out...
[2024:04:17:21:45:43]: Debug: EGL: [Mesa Project]: version [1.5]
[2024:04:17:21:45:43]: Debug: API's supported: [OpenGL OpenGL_ES ]
[2024:04:17:21:45:43]: Debug: GL: vendor: Mesa
[2024:04:17:21:45:43]: Debug: GL: renderer: llvmpipe (LLVM 15.0.7, 256 bits)
[2024:04:17:21:45:43]: Debug: GL: version: 4.5 (Compatibility Profile) Mesa 23.3.6 - kisak-mesa PPA
[2024:04:17:21:45:43]: Debug: GL: shader: 4.50
[2024:04:17:21:45:43]: Info: SDR color coding [Rec. 601]
[2024:04:17:21:45:43]: Info: Color depth: 8-bit
[2024:04:17:21:45:43]: Info: Color range: [JPEG]
libva info: VA-API version 1.14.0
libva info: Trying to open /tmp/.mount_sunshiVQNzQT/usr/lib/radeonsi_drv_video.so
libva info: Found init function __vaDriverInit_1_14
libva info: va_openDriver() returns 0
[2024:04:17:21:45:43]: Debug: vaapi vendor: Mesa Gallium driver 24.0.5 - kisak-mesa PPA for AMD Radeon RX 6650 XT (radeonsi, navi23, LLVM 15.0.7, DRM 3.57, 6.8.4-zabbly+)
[2024:04:17:21:45:43]: Debug: [AVHWDeviceContext @ 0x563d0a6c0240] VAAPI driver: Mesa Gallium driver 24.0.5 - kisak-mesa PPA for AMD Radeon RX 6650 XT (radeonsi, navi23, LLVM 15.0.7, DRM 3.57, 6.8.4-zabbly+).
[2024:04:17:21:45:43]: Debug: [AVHWDeviceContext @ 0x563d0a6c0240] Driver not found in known nonstandard list, using standard behaviour.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a4e1f80] Input surface format is nv12.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a4e1f80] Using VAAPI profile VAProfileH264High (7).
[2024:04:17:21:45:43]: Error: [h264_vaapi @ 0x563d0a4e1f80] No usable encoding entrypoint found for profile VAProfileH264High (7).
[2024:04:17:21:45:43]: Info: Retrying with fallback configuration options for [h264_vaapi] after error: Function not implemented
libva info: VA-API version 1.14.0
libva info: Trying to open /tmp/.mount_sunshiVQNzQT/usr/lib/radeonsi_drv_video.so
libva info: Found init function __vaDriverInit_1_14
libva info: va_openDriver() returns 0
[2024:04:17:21:45:43]: Debug: vaapi vendor: Mesa Gallium driver 24.0.5 - kisak-mesa PPA for AMD Radeon RX 6650 XT (radeonsi, navi23, LLVM 15.0.7, DRM 3.57, 6.8.4-zabbly+)
[2024:04:17:21:45:43]: Debug: [AVHWDeviceContext @ 0x563d0a8addc0] VAAPI driver: Mesa Gallium driver 24.0.5 - kisak-mesa PPA for AMD Radeon RX 6650 XT (radeonsi, navi23, LLVM 15.0.7, DRM 3.57, 6.8.4-zabbly+).
[2024:04:17:21:45:43]: Debug: [AVHWDeviceContext @ 0x563d0a8addc0] Driver not found in known nonstandard list, using standard behaviour.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Input surface format is nv12.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Using VAAPI profile VAProfileH264High (7).
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Using VAAPI entrypoint VAEntrypointEncSlice (6).
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Using VAAPI render target format YUV420 (0x1).
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] RC mode: CBR.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] RC target: 100% of 1000000 bps over 1000 ms.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] RC buffer: 1000000 bits, initial fullness 750000 bits.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] RC framerate: 60/1 (60.00 fps).
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Driver does not report any additional prediction constraints.
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Using intra and P-frames (supported references: 1 / 1).
[2024:04:17:21:45:43]: Warning: [h264_vaapi @ 0x563d0a6f1d80] Driver does not support some wanted packed headers (wanted 0xd, found 0x1).
[2024:04:17:21:45:43]: Debug: [h264_vaapi @ 0x563d0a6f1d80] Using level 4.2.
/tmp/.mount_sunshiVQNzQT/AppRun.wrapped: line 92: 76507 Segmentation fault (core dumped) LIBVA_DRIVERS_PATH="$HERE/usr/lib" "$SUNSHINE_BIN_HERE" $@
If I have a build that's working on my AMD box I'll let you know.
from sunshine.
gschintgen
commented on June 15, 2024
This report documents two independent(?) issues that I can both confirm for the latest nightly AppImage on Ubuntu 22.04 with Intel graphics. I can happily provide more logs if needed.
Issue 1. Sunshine AppImage is indeed unable to load VA-API. The error messages for my Intel laptop are similar except that it fails loading iHD_drv_video.so and i965_drv_video.so.
Issue 2. KMS capture is not working due to missing capabilities. The AppImage is unable to set the capabilities. I'm not sure there's a workaround for this (see AppImage/AppImageKit#881), except to just run the AppImage as root e.g. using sudo -E in order to use the regular user's ~/.config/sunshine directory.
from sunshine.
ReenigneArcher
commented on June 15, 2024
@gschintgen I'm trying to work around issue 2 in #2300. I haven't tested the code yet, but the idea is to actually copy the binaries out of the AppImage. This has likely been an issue forever, and the custom AppRun probably only ever worked when the AppImage was extracted. sudo -E might be easier though.
For issue 1, we probably need to manually add those libraries to the AppImage. Here's an example of how we did this in the past (for different libraries though). https://github.com/LizardByte/Sunshine/blame/974c4bd4a1d53a4ed09ba0d17add0845551d7c61/.github/workflows/CI.yml#L460
from sunshine.
gschintgen
commented on June 15, 2024
but the idea is to actually copy the binaries out of the AppImage.
Hmm, I thought of such an approach too, but then it's even stranger than having an AppImage with an --install command. (I'm not using AppImages that often though; maybe it's more common than I imagine.) It almost becomes a classic installer instead. But if this is what it takes, then so be it.
Since CAP_SYS_ADMIN is almost root anyway I tried chowning the AppImage to root and setting the setuid bit, but for a mysterious reason it then immediately fails to start with open dir error: Permission denied. (no other output)
As for the other issue, bundling hardware drivers seemed slightly unusual at first, but apparently that's quite normal:
locate i965_drv_video.so
/snap/gnome-3-38-2004/119/usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
/snap/gnome-3-38-2004/143/usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
/snap/gnome-42-2204/141/usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
/snap/gnome-42-2204/176/usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
/snap/moonlight/2330/usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
/usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
and
ls -l /snap/moonlight/2330/usr/lib/x86_64-linux-gnu/dri
total 521619
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 crocus_dri.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 d3d12_dri.so
-rw-r--r-- 5 root root 15101240 Jan 24 13:01 d3d12_drv_video.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 i915_dri.so
-rw-r--r-- 1 root root 1764040 Jul 11 2020 i965_drv_video.so
-rw-r--r-- 1 root root 32892408 Feb 13 2023 iHD_drv_video.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 iris_dri.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 kms_swrast_dri.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 nouveau_dri.so
-rw-r--r-- 5 root root 15101240 Jan 24 13:01 nouveau_drv_video.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 r300_dri.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 r600_dri.so
-rw-r--r-- 5 root root 15101240 Jan 24 13:01 r600_drv_video.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 radeonsi_dri.so
-rw-r--r-- 5 root root 15101240 Jan 24 13:01 radeonsi_drv_video.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 swrast_dri.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 virtio_gpu_dri.so
-rw-r--r-- 5 root root 15101240 Jan 24 13:01 virtio_gpu_drv_video.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 vmwgfx_dri.so
-rw-r--r-- 13 root root 32612968 Jan 24 13:01 zink_dri.so
I don't have the slightest idea though how self-contained those libraries are... and also why the AppImage is unable to load the system libraries. What I can tell is that those drivers have a versioned Init function:
$ vainfo
libva info: VA-API version 1.14.0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_14
libva error: /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so init failed
libva info: va_openDriver() returns 1
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so
libva info: Found init function __vaDriverInit_1_10
libva info: va_openDriver() returns 0
$ nm -D /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so | grep Init
000000000005f180 T __vaDriverInit_1_10
Sunshine.AppImage seems to look only for __vaDriverInit_1_0.
from sunshine.
gschintgen
commented on June 15, 2024
The VAAPI issue seems to be due to some incoherent library versions. It can be "fixed" by removing the bundled libva*.so files.
Here's what I tried (from memory):
mkdir tmpapp
mkdir tmplibs
cd tmpapp
../sunshine.AppImage --appimage-extract
# check that the extracted version is "working" (with broken VAAPI):
cd squashfs-root
./AppRun
# now move libva out of the way:
mv usr/lib/libva*.so ../../tmplibs
# check again, now it should be able to load VAAPI:
./AppRun
It seems to be an all or nothing situation: either don't bundle libva.so file(s) so that the system one is used, which in turn loads the system *_video.so files, or bundle a complete set of Mesa/VAAPI files. Since the AppImage's raison d'être is portability, it's probably best to bundle all VAAPI related files.
from sunshine.
ReenigneArcher
commented on June 15, 2024
it's probably best to bundle all VAAPI related files
I'd agree with this
from sunshine.
gschintgen
commented on June 15, 2024
@RickAndTired : Would you mind testing a build of PR #2429? The AppImage can be downloaded here:
https://github.com/LizardByte/Sunshine/actions/runs/8711575720/artifacts/1419787335
(You must be logged in; don't use wget from the CLI...)
from sunshine.
RickAndTired
commented on June 15, 2024
@RickAndTired : Would you mind testing a build of PR #2429? The AppImage can be downloaded here: https://github.com/LizardByte/Sunshine/actions/runs/8711575720/artifacts/1419787335 (You must be logged in; don't use wget from the CLI...)
VAAPI didn't work with this build, here's the log:
from sunshine.
gschintgen
commented on June 15, 2024
I downgraded the mesa PPA in the AppImage. Now it seems to be working fine on my AMD system. (Logs look good, basic desktop streaming as perfect as it should be.)
I'll post the link as soon as there's a new build available.
(I tested it with the build in my own fork: https://github.com/gschintgen/Sunshine/actions/runs/8728427369/artifacts/1423643042 )
from sunshine.
gschintgen
commented on June 15, 2024
https://github.com/LizardByte/Sunshine/actions/runs/8728292196/artifacts/1423740737
from sunshine.
RickAndTired
commented on June 15, 2024
https://github.com/LizardByte/Sunshine/actions/runs/8728292196/artifacts/1423740737
This one crashes
from sunshine.
gschintgen
commented on June 15, 2024
Bummer.
It might be related to linuxdeploy's blacklist, notably https://github.com/AppImageCommunity/pkg2appimage/blob/ac4d7b73bd98be7d14644775e9058d09109b2fa5/excludelist#L55.
But the libdrm on your Ubuntu 23.10 is not that old either.
At this point I can only guess. All the more reason to just bundle everything.
from sunshine.
Related Issues (20)
- CMAKE_INSTALL_PREFIX / SUNSHINE_ASSETS_DIR is not documented in compilation instructions and defaults to incorrect values HOT 1
- Can't render display in PikaOs on and igpu with drivers under Proxmox with igpu passthrough
- Brew install sunshine but "CMake Generate step failed." HOT 15
- Troubleshooting Streaming Issues with Sunshine 0.22: Freezes, Resolution Changes, and Connectivity Problems
- 0.23.0: CMake error when trying to build with tests disabled (-DBUILD_TESTS:BOOL=FALSE) HOT 5
- Stuttering and video decoding time shot up when waking up from standby. HOT 1
- Black screen when streaming. Error: GL: /build/sunshine/src/platform/linux/graphics.cpp:570: [00000502] HOT 11
- [Win] Service not starting if temp is on another drive HOT 10
- Scaling issues on strange resolutions and how to fix them.
- Wrong setcap command in displayed error message on home screen
- How can I get client's information? or whether the client connected or not? HOT 1
- Web interface claims there is a new version, while running latest available HOT 3
- Sound and Input working, screen seems to be stuck/frozen HOT 1
- Duplicate "fastest" option in QuickSync Preset dropdown
- KMS Capture does not work on second monitor HOT 5
- Can't close the stream when running full desktop stream HOT 2
- M1 Sunshine is not Responding To Keyboard and Mouse Events from Android and PC Moonlight Apps HOT 2
- [Contains a solution.]Android Bluetooth mouse and keyboard remote access to PC, after a period of time, the mouse and keyboard input becomes unresponsive, but the screen transmission remains normal. HOT 1
- Wrong screen orientation on Steam Deck (Linux) using KMS HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from sunshine.