Linux-odroid 6.13.1

Tried two videos from Jellyfin Repo unfortunately no vpu for HEVC 8bit or 10bit video. On 8bit mpv shows a blank blue screen then switch to no vpu hw acceleration on my device.

 mpv --hwdec=v4l2request --profile=fast '/home/alarm/Downloads/Test Jellyfin 1080p HEVC 8bit 10M.mp4' 
● Video  --vid=1  (hevc 1920x1080 60 fps) [default]
Using hardware decoding (v4l2request).
VO: [gpu] 1920x1080 drm_prime[nv12]
[ffmpeg] V4L2RequestContext: Failed waiting on capture buffer 3
Error while decoding frame (hardware decoding)!
[ffmpeg] V4L2RequestContext: Failed waiting on capture buffer 5
Error while decoding frame (hardware decoding)!
[ffmpeg] V4L2RequestContext: Failed waiting on capture buffer 0
Error while decoding frame (hardware decoding)!
Attempting next decoding method after failure of hevc-v4l2request.
[ffmpeg/video] hevc: Could not find ref with POC 3
VO: [gpu] 1920x1080 yuv420p
V: 00:00:10 / 00:00:29 (34%)
Exiting... (Quit)
[alarm@opi5plus ~]$ mpv --hwdec=v4l2request --profile=fast '/home/alarm/Downloads/Test Jellyfin 1080p HEVC 8bit 10M.mp4' 
● Video  --vid=1  (hevc 1920x1080 60 fps) [default]
Using hardware decoding (v4l2request).
VO: [gpu] 1920x1080 drm_prime[nv12]
V: 00:00:00 / 00:00:29 (3%) Dropped: 16
Exiting... (Quit)

I think that rockchip doesn’t support it.
For example allwinner h700 (h616/h618)

Thanks. MPV veing able to stream h264 1080p/60 with vpu hw acceleration is a great start (all 8 cores CPU typically 2% to 10% only occassionally greater than 10%).

1 Like

I have two working installations of Manjaro ARM KDE for my Orange PI 5 Plus, one on SDCard (for tests), and another on NVMe which I use for development, 6.14.0 only works when I boot it from SD card, when I try to install it to NVMe it boots to black screen, so I’m stuck with 6.14rc4 on it, any later kernel build fails to boot properly…

When you boot with SD card (kernel-6.14.0) can the system recognizes the NVMe?

Yes, it does recognize NVMe.
I boot from SDCard to recover from “bad” kernel install by chroot’ing to NVMe partitions and reinstalling working kernel.

Interesting, the kernel can can access the NVMe but cannot boot up from the NVMe.

On my device the NVMe has the Joshua Riek’s old Ubuntu 22.04 on it, the boot loader on SPI is also from Joshua Riek’s image.

Manjaro is boot up from SD Card.

1 Like

You will need /boot on enmc or sdcard.

The bootloader cannot read nvme on any rk chip.
I have /boot with bootloader on emmc while / root on ssd drive over sata and that is the only way to make it work for nvme too.

2 Likes

Wrong, old build of linux-rc (6.14rc4) boots directly from NVMe.
I have working setup with only NVMe connected to my Opi5Plus.

1 Like

only possible when you have bootloader on spi or emmc.

1 Like

I haven’t eMMC on my Opi5+, so probably my device was shipped to me with bootloader flashed to SPI.


I have two working installations of Manjaro ARM KDE for my Orange PI 5 Plus, one on SDCard (for tests), and another on NVMe which I use for development, 6.14.0 only works when I boot it from SD card, when I try to install it to NVMe it boots to black screen, so I’m stuck with 6.14rc4 on it, any later kernel build fails to boot properly…

Well, then this makes sense…

  1. When it boots from NVMe, it actually loads u-boot from an SPI, which may have outdated DTB and 6.14 kernel fails to start properly.
  2. When it boots from SDCard, it boots with recent DTB and kernel happily works.

@spikerguy @Darksky WDYT?
Looks like I need to somehow update boot on SPI.

1 Like

linux-6.14.2-1 pushed to unstable when the mirrors sync.

3 Likes

That’s not how dtb works.
If the bootloader is in spi then it will look for certain partition on type of drive. Mostly it is

  1. Nvme (only if the bootloader in spi have this function)
  2. Sd card
  3. Emmc
    If the boot partition is found then it will not into it and read the extlinux and try to execute the command from it which will try to load kernel.

So it is the kernel which have a problem.
Only way you can find out what is the problem is by connection UART connector and get raw logs.

Without that it’s just guess work.

1 Like

Thanks. Upgraded successfully to linux-6.14.2-1. Just found out linux-6.14.2-1 support HDMI1 video output but NO sound on HDMI1. Audio/sound available on HDMI0.

Edit: Installed linux-6.14.2-1 on GT King Pro and it boot successfully. Previously was not able to boot up linux-6.14-rc on GT King Pro.

2 Likes

linux-6.14.3-1 pushed to unstable when the mirrors sync.

4 Likes

GT King Pro. The wifi loading behaviour of linux-6.14.2-1 is different from linux-aml-6.12.21-4.

On linux-aml-6.12.21-4 the wifi will be enabled/loaded immediately when Desktop Environment is booted up.

On linux-6.14.2-1 wifi is only availabe 30 seconds to 1 minute after Desktop Environment is booted up.

Yes, 6.14 tries to load many firmware files, bin, xz, … and this last 1 minute.

1 Like

Hi,

linux-6.14.3-1 kernel update success for the Odroid c2 && n2 && c4 && m1 (ssd disc && nvme disc && emmc module)

Enjoy !

2 Likes

Hi @Darksky & @SickOS
Most likely Samba or nmb.service is delaying the WiFi connection on startup. It is take more than 90 seconds after startup to have WiFi connection on linux-6.14.3-1

Other kernels linux-aml-6.12.21-4 or kernel-6.14.2-1-7Ji (from 7Ji-repo) WiFi connection is available on startup.


 jfl@jfl-gtkpro    systemd-analyze blame
1min 30.098s nmb.service
     22.435s plymouth-quit-wait.service
      2.581s dev-sda2.device
      1.080s ldconfig.service
      1.014s udisks2.service
       849ms NetworkManager.service
       839ms systemd-journal-flush.service
       724ms accounts-daemon.service
       723ms lvm2-monitor.service
       613ms user@1000.service
   jfl@jfl-gtkpro    systemd-analyze critical-chain
The time when unit became active or started is printed after the "@" character.
The time the unit took to start is printed after the "+" character.

graphical.target @1min 37.679s
└─multi-user.target @1min 37.678s
  └─smb.service @1min 37.315s +362ms
    └─network-online.target @7.204s
      └─network.target @7.204s
        └─wpa_supplicant.service @19.765s +153ms
          └─basic.target @6.350s
            └─dbus-broker.service @6.077s +267ms
              └─dbus.socket @6.034s
                └─sysinit.target @6.022s
                  └─systemd-update-done.service @5.937s +83ms
                    └─ldconfig.service @4.850s +1.080s
                      └─systemd-tmpfiles-setup.service @4.351s +495ms
                        └─local-fs.target @4.267s
                          └─run-user-1000-gvfs.mount @11.199s
                            └─run-user-1000.mount @8.734s
                              └─local-fs-pre.target @2.707s
                                └─systemd-tmpfiles-setup-dev.service @2.545s +161ms
                                  └─systemd-sysusers.service @2.313s +217ms
                                    └─systemd-tmpfiles-setup-dev-early.service @1.767s

Is here a way to get WiFi connection first then nmb.service?

Edit: Opi5-Plus boot up successfully with linux-6.14.3-1 as expected.

It has always taken about 90 seconds here also on my vim3pro with the upstream kernel and my rockpro64 always has wifi right away. I intended to look into it but never got around to it until today.

I am thinking smb/nmb has nothing to do with the delay here. I disabled both services here along with networkd (which was taking a long time and never loading) on my vim3pro and it still took @90 seconds for wifi.

vim3pro with smb/nmb/networkd disabled:

graphical.target @4.339s
└─lightdm.service @4.290s +47ms
  └─systemd-user-sessions.service @4.252s +34ms
    └─network.target @4.241s
      └─NetworkManager.service @3.442s +798ms
        └─basic.target @3.439s
          └─dbus-broker.service @3.246s +188ms
            └─dbus.socket @3.216s
              └─sysinit.target @3.210s
                └─systemd-vconsole-setup.service @3.009s +137ms
                  └─systemd-journald.socket @949ms
                    └─system.slice @561ms
                      └─-.slice @561ms