Green screen with the attached video sample

backuprepo commented

2025-12-23 10:30:59 +01:00

Owner

Originally created by @kyak on GitHub (Mar 15, 2024).

Originally reported at https://github.com/7Ji-PKGBUILDs/.meta/issues/29

I'm using kodi-mpp-git and ffmpeg-rockchip-git.

With the attached video sample, mpv --profile=fast --hwdec=rkmpp path-to-file displays green screen.

kodi under GBM (https://github.com/hbiyik/ffmpeg-rockchip/wiki/Rendering#kodi-under-gbm) seems to play it, although displays black screen after stopping the video.

After stopping the video, kodi shows black screen (it doesn't freeze, it really is just a blank or black screen).

I see these messages in journalctl:

kernel: [drm:rockchip_vop2_mod_supported] *ERROR* Unsupported format modifier 0x800000000000051

And these messages in kodi log:

2024-03-14 21:28:00.101 T:328297   error <general>: CDRMAtomic::DrmAtomicCommit - test commit failed: (Invalid argument) - falling back to last successful atomic request
2024-03-14 21:28:00.102 T:328297   error <general>:
                                                   Object: plane        ID: 71
                                                     Property: SRC_X    ID: 9   Value: 0
                                                     Property: SRC_Y    ID: 10  Value: 0
                                                     Property: SRC_W    ID: 11  Value: 251658240
                                                     Property: SRC_H    ID: 12  Value: 141557760
                                                     Property: CRTC_X   ID: 13  Value: 0
                                                     Property: CRTC_Y   ID: 14  Value: 0
                                                     Property: CRTC_W   ID: 15  Value: 3840
                                                     Property: CRTC_H   ID: 16  Value: 2160
                                                     Property: FB_ID    ID: 17  Value: 322
                                                     Property: CRTC_ID  ID: 20  Value: 85
2024-03-14 21:28:00.102 T:328297   error <general>: CDRMAtomic::DrmAtomicCommit - atomic commit failed: Invalid argument

Sorry for gzip'ed mkv, couldn't upload to github otherwise :(

out.mkv.gz

Originally created by @kyak on GitHub (Mar 15, 2024). Originally reported at https://github.com/7Ji-PKGBUILDs/.meta/issues/29 I'm using `kodi-mpp-git` and `ffmpeg-rockchip-git`. With the attached video sample, `mpv --profile=fast --hwdec=rkmpp path-to-file` displays green screen. `kodi` under GBM (https://github.com/hbiyik/ffmpeg-rockchip/wiki/Rendering#kodi-under-gbm) seems to play it, although displays black screen after stopping the video. After stopping the video, kodi shows black screen (it doesn't freeze, it really is just a blank or black screen). I see these messages in journalctl: ``` kernel: [drm:rockchip_vop2_mod_supported] *ERROR* Unsupported format modifier 0x800000000000051 ``` And these messages in kodi log: ``` 2024-03-14 21:28:00.101 T:328297 error <general>: CDRMAtomic::DrmAtomicCommit - test commit failed: (Invalid argument) - falling back to last successful atomic request 2024-03-14 21:28:00.102 T:328297 error <general>: Object: plane ID: 71 Property: SRC_X ID: 9 Value: 0 Property: SRC_Y ID: 10 Value: 0 Property: SRC_W ID: 11 Value: 251658240 Property: SRC_H ID: 12 Value: 141557760 Property: CRTC_X ID: 13 Value: 0 Property: CRTC_Y ID: 14 Value: 0 Property: CRTC_W ID: 15 Value: 3840 Property: CRTC_H ID: 16 Value: 2160 Property: FB_ID ID: 17 Value: 322 Property: CRTC_ID ID: 20 Value: 85 2024-03-14 21:28:00.102 T:328297 error <general>: CDRMAtomic::DrmAtomicCommit - atomic commit failed: Invalid argument ``` Sorry for gzip'ed mkv, couldn't upload to github otherwise :( [out.mkv.gz](https://github.com/7Ji-PKGBUILDs/.meta/files/14606172/out.mkv.gz)

backuprepo added the

upstream

label 2025-12-23 10:30:59 +01:00

backuprepo commented

2025-12-23 10:31:00 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

and your kernel is?

@hbiyik commented on GitHub (Mar 15, 2024): and your kernel is?

backuprepo commented

2025-12-23 10:31:00 +01:00

Author

Owner

@kyak commented on GitHub (Mar 15, 2024):

and your kernel is?

It's 5.10.160

@kyak commented on GitHub (Mar 15, 2024): > and your kernel is? It's 5.10.160

backuprepo commented

2025-12-23 10:31:01 +01:00

Author

Owner

@nyanmisaka commented on GitHub (Mar 15, 2024):

This clip itself is abnormal. When I play it on other x86 PCs, it also behaves very choppy. When I use ffmpeg-rockchip to transcode on RK3588, the transcoded clip is still very choppy, but there is no green screen here.

https://github.com/nyanmisaka/ffmpeg-rockchip/assets/14953024/4ae25d5a-38bc-4521-ad40-47c5031c0005

@nyanmisaka commented on GitHub (Mar 15, 2024): This clip itself is abnormal. When I play it on other x86 PCs, it also behaves very choppy. When I use ffmpeg-rockchip to transcode on RK3588, the transcoded clip is still very choppy, but there is no green screen here. https://github.com/nyanmisaka/ffmpeg-rockchip/assets/14953024/4ae25d5a-38bc-4521-ad40-47c5031c0005

backuprepo commented

2025-12-23 10:31:01 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

dont worry, i reproduced the issue,

true that, the file is acting weird, but blue/green screen is because of the dimensions

w = 912,
stride = 57*16 (this is what mpp is using align to 16) = 912
but mesa (EGL/mpv) expect 912 to be 64 dividable, 912/64 = 14.25, so wont work

however gbm should have still worked, seems like there is something weird there, i m trying find out if this is kernel related or xbmc related....

@hbiyik commented on GitHub (Mar 15, 2024): dont worry, i reproduced the issue, true that, the file is acting weird, but blue/green screen is because of the dimensions w = 912, stride = 57*16 (this is what mpp is using align to 16) = 912 but mesa (EGL/mpv) expect 912 to be 64 dividable, 912/64 = 14.25, so wont work however gbm should have still worked, seems like there is something weird there, i m trying find out if this is kernel related or xbmc related....

backuprepo commented

2025-12-23 10:31:02 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

to prevent kodi/gbm issues in the meanwhile you can use without afbc mdoe as below

kodi --windowing=gbm --audio-backend=alsa

it should work (as long the file is not corrupted, the file seems to be corrupted)

do not add env variable FFMPEG_RKMPP_DEC_OPT="afbc=on" , there seems to be an issue there.

@hbiyik commented on GitHub (Mar 15, 2024): to prevent kodi/gbm issues in the meanwhile you can use without afbc mdoe as below `kodi --windowing=gbm --audio-backend=alsa` it should work (as long the file is not corrupted, the file seems to be corrupted) do not add env variable `FFMPEG_RKMPP_DEC_OPT="afbc=on" `, there seems to be an issue there.

backuprepo commented

2025-12-23 10:31:02 +01:00

Author

Owner

@nyanmisaka commented on GitHub (Mar 15, 2024):

I think I know why. I've been using the mpp patch for Jellyfin to avoid 64-aligned issues when using RGA3. Maybe you should add these two to the repository too.

6e76ecb50a
6221ca89d9

@nyanmisaka commented on GitHub (Mar 15, 2024): I think I know why. I've been using the mpp patch for Jellyfin to avoid 64-aligned issues when using RGA3. Maybe you should add these two to the repository too. https://github.com/nyanmisaka/mpp/commit/6e76ecb50ab85b7c3de9368903e687b6aa21aad2 https://github.com/nyanmisaka/mpp/commit/6221ca89d9837935edbc56af46cf651bd3c5b6cf

backuprepo commented

2025-12-23 10:31:03 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

ok what if vp9 or hevc, are they 256 aligned?

and mpeg1/2/4... DVD stuff

@hbiyik commented on GitHub (Mar 15, 2024): ok what if vp9 or hevc, are they 256 aligned? and mpeg1/2/4... DVD stuff

backuprepo commented

2025-12-23 10:31:04 +01:00

Author

Owner

@nyanmisaka commented on GitHub (Mar 15, 2024):

iirc vp9 and hevc should both be 256-aligned. I forgot mpeg1/2/4, will add it later.

@nyanmisaka commented on GitHub (Mar 15, 2024): iirc vp9 and hevc should both be 256-aligned. I forgot mpeg1/2/4, will add it later.

backuprepo commented

2025-12-23 10:31:04 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

i think not always, only when above certain size, at least for either vp9 or hevc afaik...

@hbiyik commented on GitHub (Mar 15, 2024): i think not always, only when above certain size, at least for either vp9 or hevc afaik...

backuprepo commented

2025-12-23 10:31:05 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

linking the issue for future reference https://github.com/rockchip-linux/mpp/issues/507

@hbiyik commented on GitHub (Mar 15, 2024): linking the issue for future reference https://github.com/rockchip-linux/mpp/issues/507

backuprepo commented

2025-12-23 10:31:05 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 15, 2024):

Found the root cause of GBM problem

https://github.com/xbmc/xbmc/pull/24431#issuecomment-2000680604

i will push a fix, in the meanwhile you can disable afbc mode by not passing the env value

@hbiyik commented on GitHub (Mar 15, 2024): Found the root cause of GBM problem https://github.com/xbmc/xbmc/pull/24431#issuecomment-2000680604 i will push a fix, in the meanwhile you can disable afbc mode by not passing the env value

backuprepo commented

2025-12-23 10:31:06 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 17, 2024):

@kyak
i pushed a fix for xbmc here: https://github.com/hbiyik/xbmc/tree/gbm_drm_dynamic_afbc_video_planes_v2
can you test it?

if you do not know how to test it here is a quick recipe.

# download the PKGBUILD of kodi-ext-git to clean directory
yay -G kodi-ext-git 
# go to dir
cd kodi-ext-git
# edit the PKGBUILD
nano PKGBUILD
# replace the source git repo to my test repo
# the line source=("git+https://github.com/xbmc/xbmc.git#branch=$_codename")
# should be source=("git+https://github.com/hbiyik/xbmc.git#branch=gbm_drm_dynamic_afbc_video_planes_v2")

# build the package
makepkg -s

# you dont have to install navigate to build directory
cd src/kodi-build

# and execute the kodi.bin locally
FFMPEG_RKMPP_DEC_OPT="afbc=on" ./kodi.bin --windowing=gbm --audio-backend=alsa

if so far so good, we can move to https://github.com/xbmc/xbmc/pull/24431 for the discussion since the issue is not related to ffmpeg but xbmc

@hbiyik commented on GitHub (Mar 17, 2024): @kyak i pushed a fix for xbmc here: https://github.com/hbiyik/xbmc/tree/gbm_drm_dynamic_afbc_video_planes_v2 can you test it? if you do not know how to test it here is a quick recipe. ``` shell # download the PKGBUILD of kodi-ext-git to clean directory yay -G kodi-ext-git # go to dir cd kodi-ext-git # edit the PKGBUILD nano PKGBUILD # replace the source git repo to my test repo # the line source=("git+https://github.com/xbmc/xbmc.git#branch=$_codename") # should be source=("git+https://github.com/hbiyik/xbmc.git#branch=gbm_drm_dynamic_afbc_video_planes_v2") # build the package makepkg -s # you dont have to install navigate to build directory cd src/kodi-build # and execute the kodi.bin locally FFMPEG_RKMPP_DEC_OPT="afbc=on" ./kodi.bin --windowing=gbm --audio-backend=alsa ``` if so far so good, we can move to https://github.com/xbmc/xbmc/pull/24431 for the discussion since the issue is not related to ffmpeg but xbmc

backuprepo commented

2025-12-23 10:31:06 +01:00

Author

Owner

@kyak commented on GitHub (Mar 18, 2024):

@hbiyik although i build kodi-ext-git master branch on a regular basic, i couldn't build from your branch. Maybe it's because your branch is lagging behind the master branch? The error is the following:

[ 33%] Building CXX object build/music/tags/CMakeFiles/music_tags.dir/TagLibVFSStream.cpp.o
In file included from /home/user/build/kodi-ext-git/src/xbmc/xbmc/music/tags/TagLibVFSStream.cpp:8:
/home/user/build/kodi-ext-git/src/xbmc/xbmc/music/tags/TagLibVFSStream.h:40:34: error: «TagLib::ulong» was not declared
   40 |     TagLib::ByteVector readBlock(TagLib::ulong length) override;

....

@kyak commented on GitHub (Mar 18, 2024): @hbiyik although i build `kodi-ext-git` master branch on a regular basic, i couldn't build from your branch. Maybe it's because your branch is lagging behind the master branch? The error is the following: ``` [ 33%] Building CXX object build/music/tags/CMakeFiles/music_tags.dir/TagLibVFSStream.cpp.o In file included from /home/user/build/kodi-ext-git/src/xbmc/xbmc/music/tags/TagLibVFSStream.cpp:8: /home/user/build/kodi-ext-git/src/xbmc/xbmc/music/tags/TagLibVFSStream.h:40:34: error: «TagLib::ulong» was not declared 40 | TagLib::ByteVector readBlock(TagLib::ulong length) override; .... ```

backuprepo commented

2025-12-23 10:31:07 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 18, 2024):

@kyak yeah it is lagging but thats fine.

you can just add

-DENABLE_INTERNAL_TAGLIB=ON

somewhere around line 89~109, in to the _args section as a new line.

also please delete the

 [[ -d kodi-build ]] && rm -rf kodi-build

in line 70, under prepare() this will help to build faster next time, because the PKGBUILD will not delete the previous build and continue where it was left.

@hbiyik commented on GitHub (Mar 18, 2024): @kyak yeah it is lagging but thats fine. you can just add ```shell -DENABLE_INTERNAL_TAGLIB=ON ``` somewhere around line 89~109, in to the `_args` section as a new line. also please delete the ```shell [[ -d kodi-build ]] && rm -rf kodi-build ``` in line 70, under `prepare()` this will help to build faster next time, because the PKGBUILD will not delete the previous build and continue where it was left.

backuprepo commented

2025-12-23 10:31:07 +01:00

Author

Owner

@kyak commented on GitHub (Mar 18, 2024):

@hbiyik thanks for the hints!

Unfortunately, it doesn't really work. Most of the videos don't play (although I found one that plays).

For example, Big_Buck_Bunny_1080_10s_10MB.mp4 displays black screen.

@kyak commented on GitHub (Mar 18, 2024): @hbiyik thanks for the hints! Unfortunately, it doesn't really work. Most of the videos don't play (although I found one that plays). For example, `Big_Buck_Bunny_1080_10s_10MB.mp4` displays black screen.

backuprepo commented

2025-12-23 10:31:07 +01:00

Author

Owner

@kyak commented on GitHub (Mar 18, 2024):

Without FFMPEG_RKMPP_DEC_OPT="afbc=on", it seems to work, albeit weird.

For example, it displays black screen behind OSD (i.e. OSD is drawn on top of black screen). When OSD is not active, the video displays normally.

@kyak commented on GitHub (Mar 18, 2024): Without FFMPEG_RKMPP_DEC_OPT="afbc=on", it seems to work, albeit weird. For example, it displays black screen behind OSD (i.e. OSD is drawn on top of black screen). When OSD is not active, the video displays normally.

backuprepo commented

2025-12-23 10:31:08 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 18, 2024):

Weird tried it last night it was working, i will check again.

Just to confirm, does ur kernel have below path?
0002-vop2_rbga2101010_capability_fix.patch

Which kernel package are you using?

Also you can you provide output of drm_info ?

Drm info pkg is on aur: https://aur.archlinux.org/packages/drm_info

@hbiyik commented on GitHub (Mar 18, 2024): Weird tried it last night it was working, i will check again. Just to confirm, does ur kernel have below path? 0002-vop2_rbga2101010_capability_fix.patch Which kernel package are you using? Also you can you provide output of `drm_info` ? Drm info pkg is on aur: https://aur.archlinux.org/packages/drm_info

backuprepo commented

2025-12-23 10:31:08 +01:00

Author

Owner

@kyak commented on GitHub (Mar 18, 2024):

Weird tried it last night it was working, i will check again.

Just to confirm, does ur kernel have below path? 0002-vop2_rbga2101010_capability_fix.patch

Which kernel package are you using?

I'm using linux-aarch64-rockchip-rk3588-bsp5.10-orangepi-git from 7Ji's repo.

Also you can you provide output of drm_info ?

Drm info pkg is on aur: https://aur.archlinux.org/packages/drm_info

drm_info.txt

@kyak commented on GitHub (Mar 18, 2024): > Weird tried it last night it was working, i will check again. > > Just to confirm, does ur kernel have below path? 0002-vop2_rbga2101010_capability_fix.patch > > Which kernel package are you using? I'm using `linux-aarch64-rockchip-rk3588-bsp5.10-orangepi-git` from 7Ji's repo. > > Also you can you provide output of `drm_info` ? > > Drm info pkg is on aur: https://aur.archlinux.org/packages/drm_info [drm_info.txt](https://github.com/nyanmisaka/ffmpeg-rockchip/files/14641345/drm_info.txt)

backuprepo commented

2025-12-23 10:31:08 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 18, 2024):

ok your kernel is not patched correctly

    │       ├───"IN_FORMATS" (immutable): blob = 58
    │       │   ├───ARM_AFBC(BLOCK_SIZE = 16x16) (0x800000000000001)
    │       │   │   ├───XRGB2101010 (0x30335258)
    │       │   │   ├───ARGB2101010 (0x30335241)
    │       │   │   ├───XBGR2101010 (0x30334258)
    │       │   │   ├───ABGR2101010 (0x30334241)

there should be no XRGB2101010 in drm formats... This is a bug confirmed by rockchip https://github.com/nyanmisaka/ffmpeg-rockchip/pull/4#issuecomment-1892934552

now we need a new ticket in 7ji-pkgbuilds. :) i will open it for you :)

@hbiyik commented on GitHub (Mar 18, 2024): ok your kernel is not patched correctly ``` │ ├───"IN_FORMATS" (immutable): blob = 58 │ │ ├───ARM_AFBC(BLOCK_SIZE = 16x16) (0x800000000000001) │ │ │ ├───XRGB2101010 (0x30335258) │ │ │ ├───ARGB2101010 (0x30335241) │ │ │ ├───XBGR2101010 (0x30334258) │ │ │ ├───ABGR2101010 (0x30334241) ``` there should be no XRGB2101010 in drm formats... This is a bug confirmed by rockchip https://github.com/nyanmisaka/ffmpeg-rockchip/pull/4#issuecomment-1892934552 now we need a new ticket in 7ji-pkgbuilds. :) i will open it for you :)

backuprepo commented

2025-12-23 10:31:09 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 18, 2024):

@kyak https://github.com/7Ji-PKGBUILDs/.meta/issues/31

@hbiyik commented on GitHub (Mar 18, 2024): @kyak https://github.com/7Ji-PKGBUILDs/.meta/issues/31

backuprepo commented

2025-12-23 10:31:09 +01:00

Author

Owner

@kyak commented on GitHub (Mar 19, 2024):

@kyak 7Ji-PKGBUILDs/.meta#31

@hbiyik back to you here. As mentioned, i'm now running the patched kernel:

    │       ├───"IN_FORMATS" (immutable): blob = 58
    │       │   ├───ARM_AFBC(BLOCK_SIZE = 16x16) (0x800000000000001)
    │       │   │   ├───XRGB8888 (0x34325258)
    │       │   │   ├───ARGB8888 (0x34325241)
    │       │   │   ├───XBGR8888 (0x34324258)
    │       │   │   ├───ABGR8888 (0x34324241)
    │       │   │   ├───RGB888 (0x34324752)
    │       │   │   ├───BGR888 (0x34324742)
    │       │   │   ├───RGB565 (0x36314752)
    │       │   │   ├───BGR565 (0x36314742)
    │       │   │   ├───YUV420_8BIT (0x38305559)
    │       │   │   ├───YUV420_10BIT (0x30315559)
    │       │   │   ├───YUYV (0x56595559)
    │       │   │   └───Y210 (0x30313259)

I need to mention that kodi-mpp-git from 7Ji's repo doesn't work with FFMPEG_RKMPP_DEC_OPT="afbc=on" nicely. It displays black screen instead of OSD and this is above the video. But I guess this is expected. Without the env var, it works fine.

As for kodi-ext-git built from your branch, it still doesn't work well. With the env var, depending on video, it displays OSD on top of black screen or a spinner (audio plays, video doesn't). Without the env var, it plays some videos fine (those with OSD on black screen with env variable) and shows spinner with audio on other videos.

This message from the kernel log repeats:

kernel: [drm:vop2_plane_atomic_check] *ERROR* Esmart0-win0 is invisible(src: pos[0, 0] rect[3840 x 2160] dst: pos[0, 0] rect[0 x 0]

@kyak commented on GitHub (Mar 19, 2024): > @kyak [7Ji-PKGBUILDs/.meta#31](https://github.com/7Ji-PKGBUILDs/.meta/issues/31) @hbiyik back to you here. As [mentioned](https://github.com/7Ji-PKGBUILDs/.meta/issues/31#issuecomment-2007938972), i'm now running the patched kernel: ``` │ ├───"IN_FORMATS" (immutable): blob = 58 │ │ ├───ARM_AFBC(BLOCK_SIZE = 16x16) (0x800000000000001) │ │ │ ├───XRGB8888 (0x34325258) │ │ │ ├───ARGB8888 (0x34325241) │ │ │ ├───XBGR8888 (0x34324258) │ │ │ ├───ABGR8888 (0x34324241) │ │ │ ├───RGB888 (0x34324752) │ │ │ ├───BGR888 (0x34324742) │ │ │ ├───RGB565 (0x36314752) │ │ │ ├───BGR565 (0x36314742) │ │ │ ├───YUV420_8BIT (0x38305559) │ │ │ ├───YUV420_10BIT (0x30315559) │ │ │ ├───YUYV (0x56595559) │ │ │ └───Y210 (0x30313259) ``` I need to mention that `kodi-mpp-git` from 7Ji's repo doesn't work with `FFMPEG_RKMPP_DEC_OPT="afbc=on"` nicely. It displays black screen instead of OSD and this is above the video. But I guess this is expected. Without the env var, it works fine. As for `kodi-ext-git` built from your branch, it still doesn't work well. With the env var, depending on video, it displays OSD on top of black screen or a spinner (audio plays, video doesn't). Without the env var, it plays some videos fine (those with OSD on black screen with env variable) and shows spinner with audio on other videos. This message from the kernel log repeats: ``` kernel: [drm:vop2_plane_atomic_check] *ERROR* Esmart0-win0 is invisible(src: pos[0, 0] rect[3840 x 2160] dst: pos[0, 0] rect[0 x 0] ```

backuprepo commented

2025-12-23 10:31:09 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 20, 2024):

thanks for the feedback, i suggest the following

since we are on different hardwares we need to reduce the software variantions to minimum.

To even simplify the cases, lets focus 5.10 kernel, because 6.1 kernel has several differences in display driver of rockchip (vop2) and i dont want possible 6.1 vop2 changes related issues to interfere, so lets make sure 5.10 works first. Later we can reiterate the test on 6.1..

so lets use https://github.com/7Ji-PKGBUILDs/linux-aarch64-rockchip-bsp5.10-joshua/blob/master/PKGBUILD.
first thing to do, patch this with
https://patch-diff.githubusercontent.com/raw/Joshua-Riek/linux-rockchip/pull/8.patch

lets focus on 1 single file and 1 single obvious problem, and that should not be out.mkv.gz, because this file is already broken. So you will see black screen with this file not because rendering is wrong but because the file is broken and ffmpeg cant decode properly.

lets only test https://github.com/hbiyik/xbmc/tree/gbm_drm_dynamic_afbc_video_planes_v2 as xbmc, because i already noticed a problem in the first implementation (the one in the ffmpeg-mpp-git). And for you building this should be reletively easy may be in 2 or 3 mins, we can test a new change. We can forget ffmpeg-mpp-git for now until we clearify all the issues.

If we can reproduce 1 single issue, i assume the root cause of it will fix the rest of the issues.

From my end, i have onyl tested this with 5.10 radxa kernel, so may be i will directly see the issue on joshua 5.10.. we will see

@hbiyik commented on GitHub (Mar 20, 2024): thanks for the feedback, i suggest the following since we are on different hardwares we need to reduce the software variantions to minimum. To even simplify the cases, lets focus 5.10 kernel, because 6.1 kernel has several differences in display driver of rockchip (vop2) and i dont want possible 6.1 vop2 changes related issues to interfere, so lets make sure 5.10 works first. Later we can reiterate the test on 6.1.. 1) so lets use https://github.com/7Ji-PKGBUILDs/linux-aarch64-rockchip-bsp5.10-joshua/blob/master/PKGBUILD. first thing to do, patch this with https://patch-diff.githubusercontent.com/raw/Joshua-Riek/linux-rockchip/pull/8.patch 2) lets focus on 1 single file and 1 single obvious problem, and that should not be [out.mkv.gz](https://github.com/7Ji-PKGBUILDs/.meta/files/14606172/out.mkv.gz), because this file is already broken. So you will see black screen with this file not because rendering is wrong but because the file is broken and ffmpeg cant decode properly. 3) lets only test https://github.com/hbiyik/xbmc/tree/gbm_drm_dynamic_afbc_video_planes_v2 as xbmc, because i already noticed a problem in the first implementation (the one in the ffmpeg-mpp-git). And for you building this should be reletively easy may be in 2 or 3 mins, we can test a new change. We can forget ffmpeg-mpp-git for now until we clearify all the issues. If we can reproduce 1 single issue, i assume the root cause of it will fix the rest of the issues. From my end, i have onyl tested this with 5.10 radxa kernel, so may be i will directly see the issue on joshua 5.10.. we will see

backuprepo commented

2025-12-23 10:31:10 +01:00

Author

Owner

@kyak commented on GitHub (Mar 20, 2024):

thanks for the feedback, i suggest the following

since we are on different hardwares we need to reduce the software variantions to minimum.

I'm running Orange Pi 5 Plus (have both 8 Gb and 16 Gb versions). What is your hardware?

To even simplify the cases, lets focus 5.10 kernel, because 6.1 kernel has several differences in display driver of rockchip (vop2) and i dont want possible 6.1 vop2 changes related issues to interfere, so lets make sure 5.10 works first. Later we can reiterate the test on 6.1..

so lets use https://github.com/7Ji-PKGBUILDs/linux-aarch64-rockchip-bsp5.10-joshua/blob/master/PKGBUILD. first thing to do, patch this with https://patch-diff.githubusercontent.com/raw/Joshua-Riek/linux-rockchip/pull/8.patch

This is clear, will do.

lets focus on 1 single file and 1 single obvious problem, and that should not be out.mkv.gz, because this file is already broken. So you will see black screen with this file not because rendering is wrong but because the file is broken and ffmpeg cant decode properly.

Also clear, I will find you the file that's properly broken :) Thing is, I've split this one from original video using ffmpeg and might've broken something in the video file along the way.

lets only test https://github.com/hbiyik/xbmc/tree/gbm_drm_dynamic_afbc_video_planes_v2 as xbmc, because i already noticed a problem in the first implementation (the one in the ffmpeg-mpp-git). And for you building this should be reletively easy may be in 2 or 3 mins, we can test a new change. We can forget ffmpeg-mpp-git for now until we clearify all the issues.

It's clear I should test kodi-ext-git with your branch gbm_drm_dynamic_afbc_video_planes_v2. But what do you mean that we should forget about ffmpeg-mpp-git? Should I simply keep using the binary package of ffmpeg-mpp-git from 7Ji's repo? Should I still set the env var when testing? Do I need to point kodi to the system ffmpeg-mpp-git or it will use it automatically (since it's -ext package)?

If we can reproduce 1 single issue, i assume the root cause of it will fix the rest of the issues.

From my end, i have onyl tested this with 5.10 radxa kernel, so may be i will directly see the issue on joshua 5.10.. we will see

@kyak commented on GitHub (Mar 20, 2024): > thanks for the feedback, i suggest the following > > since we are on different hardwares we need to reduce the software variantions to minimum. I'm running Orange Pi 5 Plus (have both 8 Gb and 16 Gb versions). What is your hardware? > > To even simplify the cases, lets focus 5.10 kernel, because 6.1 kernel has several differences in display driver of rockchip (vop2) and i dont want possible 6.1 vop2 changes related issues to interfere, so lets make sure 5.10 works first. Later we can reiterate the test on 6.1.. > > > so lets use https://github.com/7Ji-PKGBUILDs/linux-aarch64-rockchip-bsp5.10-joshua/blob/master/PKGBUILD. first thing to do, patch this with https://patch-diff.githubusercontent.com/raw/Joshua-Riek/linux-rockchip/pull/8.patch > This is clear, will do. > > lets focus on 1 single file and 1 single obvious problem, and that should not be [out.mkv.gz](https://github.com/7Ji-PKGBUILDs/.meta/files/14606172/out.mkv.gz), because this file is already broken. So you will see black screen with this file not because rendering is wrong but because the file is broken and ffmpeg cant decode properly. > Also clear, I will find you the file that's properly broken :) Thing is, I've split this one from original video using ffmpeg and might've broken something in the video file along the way. > > lets only test https://github.com/hbiyik/xbmc/tree/gbm_drm_dynamic_afbc_video_planes_v2 as xbmc, because i already noticed a problem in the first implementation (the one in the ffmpeg-mpp-git). And for you building this should be reletively easy may be in 2 or 3 mins, we can test a new change. We can forget ffmpeg-mpp-git for now until we clearify all the issues. > It's clear I should test `kodi-ext-git` with your branch `gbm_drm_dynamic_afbc_video_planes_v2`. But what do you mean that we should forget about `ffmpeg-mpp-git`? Should I simply keep using the binary package of `ffmpeg-mpp-git` from 7Ji's repo? Should I still set the env var when testing? Do I need to point kodi to the system `ffmpeg-mpp-git` or it will use it automatically (since it's `-ext` package)? > If we can reproduce 1 single issue, i assume the root cause of it will fix the rest of the issues. > > From my end, i have onyl tested this with 5.10 radxa kernel, so may be i will directly see the issue on joshua 5.10.. we will see

backuprepo commented

2025-12-23 10:31:10 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 20, 2024):

What is your hardware?

i am on radxa 5b with 4g.

But what do you mean that we should forget about ffmpeg-mpp-git

i mean we should forget about ffmpeg-mpp-git durinf testing gbm workflow, when we fix those issues in gbm_drm_dynamic_afbc_video_planes_v2, i will port them to ffmpeg-mpp-git. You can keep it installed, and use it daily, as far as i know it works without the afbc=on so you can daily use it.

for testing and verifying the kodi-ext-git, point the source to gbm_drm_dynamic_afbc_video_planes_v2 and run make pkg but dont install the generated packages. Instead go to src/kodi-build and run ./kodi.bin locally for testing.

Whenever i push a fix, you will run makepkg again, and it will compile with the latest from my branch so that you can verify that it is fixed in opi5 as well. When we make sure it works fine, i can then port the changes to kodi-mpp-git and mainline xbmc repo with PR saying that it is tested.

@hbiyik commented on GitHub (Mar 20, 2024): > What is your hardware? i am on radxa 5b with 4g. > But what do you mean that we should forget about `ffmpeg-mpp-git` i mean we should forget about ffmpeg-mpp-git durinf testing gbm workflow, when we fix those issues in gbm_drm_dynamic_afbc_video_planes_v2, i will port them to `ffmpeg-mpp-git`. You can keep it installed, and use it daily, as far as i know it works without the `afbc=on` so you can daily use it. for testing and verifying the kodi-ext-git, point the source to gbm_drm_dynamic_afbc_video_planes_v2 and run `make pkg` but dont install the generated packages. Instead go to `src/kodi-build` and run `./kodi.bin` locally for testing. Whenever i push a fix, you will run makepkg again, and it will compile with the latest from my branch so that you can verify that it is fixed in opi5 as well. When we make sure it works fine, i can then port the changes to `kodi-mpp-git` and mainline `xbmc` repo with PR saying that it is tested.

backuprepo commented

2025-12-23 10:31:11 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 20, 2024):

@kyak
here is a patched joshua's 5.10 kernel https://file.io/y7zSI9ZVm1Bl
i have tested and everything seems to work fine with my rock5b

@hbiyik commented on GitHub (Mar 20, 2024): @kyak here is a patched joshua's 5.10 kernel https://file.io/y7zSI9ZVm1Bl i have tested and everything seems to work fine with my rock5b

backuprepo commented

2025-12-23 10:31:11 +01:00

Author

Owner

@kyak commented on GitHub (Mar 20, 2024):

@kyak here is a patched joshua's 5.10 kernel https://file.io/y7zSI9ZVm1Bl i have tested and everything seems to work fine with my rock5b

Don't worry, it's almost built on my side. Will report back tomorrow.

@kyak commented on GitHub (Mar 20, 2024): > @kyak here is a patched joshua's 5.10 kernel https://file.io/y7zSI9ZVm1Bl i have tested and everything seems to work fine with my rock5b Don't worry, it's almost built on my side. Will report back tomorrow.

backuprepo commented

2025-12-23 10:31:11 +01:00

Author

Owner

@kyak commented on GitHub (Mar 21, 2024):

@hbiyik

I'm now running linux-aarch64-rockchip-bsp5.10-joshua with the patch. I also built kodi-ext-git from your branch. I'm running kodi in GBM mode, with Direct to Plane rendering.

With afbc=on, i get black screen instead of video. OSD displays fine.

Without afbc=on, the video plays fine. OSD displays fine on top of the video.

Is this expected?

@kyak commented on GitHub (Mar 21, 2024): @hbiyik I'm now running `linux-aarch64-rockchip-bsp5.10-joshua` with the patch. I also built `kodi-ext-git` from your branch. I'm running kodi in GBM mode, with Direct to Plane rendering. With `afbc=on`, i get black screen instead of video. OSD displays fine. Without `afbc=on`, the video plays fine. OSD displays fine on top of the video. Is this expected?

backuprepo commented

2025-12-23 10:31:11 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 21, 2024):

no this is not expected, it should play in afbc mode aswell.

could you restart the device and provide the full output of drm_info. It is better for me to see the default drm state, therefore i need a clean start.

Then run kodi with afbc mode in gbm, make sure u enabled debug mode . It is very important debug mode to be enabled otherwise i wont see anything.

Run the video which is resulting the black video output. Close kodi, and send the log file at ~/.kodi/temp.kodi.log

Also, for me to reproduce can you send the file that produces black video?

@hbiyik commented on GitHub (Mar 21, 2024): no this is not expected, it should play in afbc mode aswell. could you restart the device and provide the full output of `drm_info`. It is better for me to see the default drm state, therefore i need a clean start. Then run kodi with afbc mode in gbm, make sure u enabled [debug mode ](https://kodi.wiki/view/Log_file/Easy#Steps). It is very important debug mode to be enabled otherwise i wont see anything. Run the video which is resulting the black video output. Close kodi, and send the log file at `~/.kodi/temp.kodi.log` Also, for me to reproduce can you send the file that produces black video?

backuprepo commented

2025-12-23 10:31:12 +01:00

Author

Owner

@kyak commented on GitHub (Mar 21, 2024):

Here's the drm_info output after the reboot:
drm_info.txt

Here's the verbose log:
kodi.log

And here is the video file: https://file.io/QNZMTBNLMCxi

@kyak commented on GitHub (Mar 21, 2024): Here's the drm_info output after the reboot: [drm_info.txt](https://github.com/nyanmisaka/ffmpeg-rockchip/files/14705710/drm_info.txt) Here's the verbose log: [kodi.log](https://github.com/nyanmisaka/ffmpeg-rockchip/files/14705713/kodi.log) And here is the video file: https://file.io/QNZMTBNLMCxi

backuprepo commented

2025-12-23 10:31:12 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 21, 2024):

ok something obviously is wrong according to your logs, let me investigate this a little bit.

CDRMUtils::FindVideoPlane - Can not find a Video Plane plane for format 942691673, modifier 576460752303423553

@hbiyik commented on GitHub (Mar 21, 2024): ok something obviously is wrong according to your logs, let me investigate this a little bit. `CDRMUtils::FindVideoPlane - Can not find a Video Plane plane for format 942691673, modifier 576460752303423553`

backuprepo commented

2025-12-23 10:31:12 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 21, 2024):

@kyak is it possible to boot into https://github.com/7Ji-PKGBUILDs/linux-aarch64-rockchip-bsp6.1-joshua-git kernel, and provide the drm_info output.

I understood the problem, the soc rk3588s (used in opi5) and rk3588 (used in rock 5b) provides different drm layout, this is a little bit weird, to me looks like even a display driver (vop2) bug in kernel.

PS: no need to patch joshua6.1 all patches are already in there
I would like to compare my devices drm tree with your opi5, i think it should give some more hints.

Note to myself: f15552775d most likely this commit might fix a lot issues

@hbiyik commented on GitHub (Mar 21, 2024): @kyak is it possible to boot into https://github.com/7Ji-PKGBUILDs/linux-aarch64-rockchip-bsp6.1-joshua-git kernel, and provide the `drm_info` output. I understood the problem, the soc rk3588s (used in opi5) and rk3588 (used in rock 5b) provides different drm layout, this is a little bit weird, to me looks like even a display driver (vop2) bug in kernel. PS: no need to patch joshua6.1 all patches are already in there I would like to compare my devices drm tree with your opi5, i think it should give some more hints. Note to myself: https://github.com/Joshua-Riek/linux-rockchip/commit/f15552775df1317eafe0ae01d88f50cc91cddb89 most likely this commit might fix a lot issues

backuprepo commented

2025-12-23 10:31:13 +01:00

Author

Owner

@kyak commented on GitHub (Mar 22, 2024):

@hbiyik

Here's the drm_info for 6.1:
drm_info.txt

Please note that i'm running Orange Pi 5 Plus, which is rk3588, not rk3588s.

I also checked the situation with 6.1 kernel and it seems like it didn't change anything.

@kyak commented on GitHub (Mar 22, 2024): @hbiyik Here's the drm_info for 6.1: [drm_info.txt](https://github.com/nyanmisaka/ffmpeg-rockchip/files/14718369/drm_info.txt) Please note that i'm running Orange Pi 5 Plus, which is rk3588, not rk3588s. I also checked the situation with 6.1 kernel and it seems like it didn't change anything.

backuprepo commented

2025-12-23 10:31:13 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 22, 2024):

@kyak
could you replace your dtb with the dtb rk3588-orangepi-5-plus.dtb.tar.gz on joshua's 5.10 kernel, and run drm_info and paste the results?

the location should be:
/boot/dtbs/linux-aarch64-rockchip-bsp5.10-joshua/rockchip/rk3588-orangepi-5-plus.dtb

the actual change is here

PS: I did not mention specificly and thought it was obvious, but in any case you need to restart the device after replacing the dtb :)

@hbiyik commented on GitHub (Mar 22, 2024): @kyak could you replace your dtb with the dtb [rk3588-orangepi-5-plus.dtb.tar.gz](https://github.com/nyanmisaka/ffmpeg-rockchip/files/14727502/rk3588-orangepi-5-plus.dtb.tar.gz) on joshua's 5.10 kernel, and run `drm_info` and paste the results? the location should be: `/boot/dtbs/linux-aarch64-rockchip-bsp5.10-joshua/rockchip/rk3588-orangepi-5-plus.dtb` the actual change is [here](https://github.com/hbiyik/linux-rockchip/commit/83d0b9469244ce2d0a8f141ad973ced46f2240a9) PS: I did not mention specificly and thought it was obvious, but in any case you need to restart the device after replacing the dtb :)

backuprepo commented

2025-12-23 10:31:13 +01:00

Author

Owner

@kyak commented on GitHub (Mar 24, 2024):

@hbiyik thanks so much for your patience.

This is drm_info output from joshua's 5.10 kernel with your modified dtb:
drm_info.txt

Using modified dtb still didn't change anything

@kyak commented on GitHub (Mar 24, 2024): @hbiyik thanks so much for your patience. This is `drm_info` output from joshua's 5.10 kernel with your modified dtb: [drm_info.txt](https://github.com/nyanmisaka/ffmpeg-rockchip/files/14736006/drm_info.txt) Using modified dtb still didn't change anything

backuprepo commented

2025-12-23 10:31:14 +01:00

Author

Owner

@kyak commented on GitHub (Mar 27, 2024):

@hbiyik in case this was lost in discussion, i want to point out that there is a configuration that works fine for me. I.e. all kinds of videos plays fine, OSD displays fine and this is with GBM and Direct to Plane mode of Kodi.

The configuration is:

linux-aarch64-rockchip-rk3588-bsp5.10-orangepi-git with two patches:
0001-rga3_uncompact_fix.patch
0002-vop2_rbga2101010_capability_fix.patch
ffmpeg-mpp-git binary package
kodi-mpp-git binary package, no environment variables

Maybe with can start from this configuration and figure out what the problem is?

@kyak commented on GitHub (Mar 27, 2024): @hbiyik in case this was lost in discussion, i want to point out that there is a configuration that works fine for me. I.e. all kinds of videos plays fine, OSD displays fine and this is with GBM and Direct to Plane mode of Kodi. The configuration is: * `linux-aarch64-rockchip-rk3588-bsp5.10-orangepi-git` with two patches: [0001-rga3_uncompact_fix.patch](https://github.com/hbiyik/ffmpeg-rockchip/wiki/patches/rockchip-kernel/0001-rga3_uncompact_fix.patch) [0002-vop2_rbga2101010_capability_fix.patch](https://github.com/hbiyik/ffmpeg-rockchip/wiki/patches/rockchip-kernel/0002-vop2_rgba2101010_capability_fix.patch) * `ffmpeg-mpp-git` binary package * `kodi-mpp-git` binary package, no environment variables Maybe with can start from this configuration and figure out what the problem is?

backuprepo commented

2025-12-23 10:31:14 +01:00

Author

Owner

@hbiyik commented on GitHub (Mar 27, 2024):

@kyak
Even in this configuration kodi with gbm mode and afbc enabled with FFMPEG_RKMPP_DEC_OPT="afbc=on should cause problems on opi5 variants.

I think i got the underlying problem, and i have something in my mind to solve it with xbmc.

Currently im on vacation may be next week i will prototype something, than i might need some testinv support.

@hbiyik commented on GitHub (Mar 27, 2024): @kyak Even in this configuration kodi with gbm mode and afbc enabled with FFMPEG_RKMPP_DEC_OPT="afbc=on should cause problems on opi5 variants. I think i got the underlying problem, and i have something in my mind to solve it with xbmc. Currently im on vacation may be next week i will prototype something, than i might need some testinv support.

Rows
Columns

Green screen with the attached video sample #30