jerryg
-
Posts
11 -
Joined
-
Last visited
Content Type
Forums
Store
Crowdfunding
Applications
Events
Raffles
Community Map
Posts posted by jerryg
-
-
Got ver 20.02.0-rc1 20200129 on my NanoPc-T4.
Things work great: wired ethernet, wifi (with martinayotte's fix), bluetooth, nvme
No sound from the native soundcard, but not an issue since I have a USB module for that.
Thanks!
-
Having issues with the NVMe drive with this unified boot process (the FriendlyDesktop image for the NanoPC T4 is able to see the drive, so it is not a hardware issue). I wrote the u-boot-rk3399-tv.img to the SD card, but I am getting:
jerryg@arm:~$ dmesg | grep pcie [ 20.489985] rockchip-pcie f8000000.pcie: no vpcie12v regulator found [ 26.398329] rockchip-pcie f8000000.pcie: no vpcie12v regulator found [ 26.925369] rockchip-pcie f8000000.pcie: PCIe link training gen1 timeout! [ 26.926462] rockchip-pcie: probe of f8000000.pcie failed with error -110
Googling for similar issues seem to indicate that the PCIe bus needs to be powered-on by u-boot in order for the device to be ready when the kernel inits the rockchip-pcie component. What can I do?
-
-
I was actually looking at both encoding and decoding. For decoding, the kernel state can be found here:
https://github.com/torvalds/linux/blob/v5.5-rc2/drivers/staging/media/hantro/rk3399_vpu_hw.c#L64
Notice the .fourcc is: V4L2_PIX_FMT_MPEG2_SLICE
While GStreamer will expect V4L2_PIX_FMT_MPEG2 (notice the missing _SLICE)
I see similar differences for H264 and VP8 decoding.
My guess at this point is that the hantro_vpu driver is providing low-level support for decoding, and that the full decoding that used to be found in the firmware blobs is yet to be submitted. See this thread for other interesting bits: https://forum.armbian.com/topic/4806-the-vpu-driver/
-
On 12/30/2019 at 2:44 PM, usual user said:
For all mainline v4l2 supported devices hardware accelerated video pipelines already work with the gstreamer1 framework. I.e. using a gstreamer based media player will make use of v4l2 codecs and other components like e.g scalers.
Yes, but it seems the hantro_vpu module is a little too barebones at this point in time. Gstreamer will auto-detect what is available, and so far in kernel 5.5-rc2, we only have JPEG encoding supported. Stay tuned for further modules appearing as support improves.
jerryg@arm:~$ gst-inspect-1.0 --plugin /usr/lib/aarch64-linux-gnu/gstreamer-1.0/libgstvideo4linux2.so Plugin Details: Name video4linux2 Description elements for Video 4 Linux Filename /usr/lib/aarch64-linux-gnu/gstreamer-1.0/libgstvideo4linux2.so Version 1.16.1 License LGPL Source module gst-plugins-good Source release date 2019-09-23 Binary package GStreamer Good Plugins (Ubuntu) Origin URL https://launchpad.net/distros/ubuntu/+source/gst-plugins-good1.0 v4l2jpegenc: V4L2 JPEG Encoder v4l2deviceprovider: Video (video4linux2) Device Provider v4l2radio: Radio (video4linux2) Tuner v4l2sink: Video (video4linux2) Sink v4l2src: Video (video4linux2) Source 5 features: +-- 4 elements +-- 1 device providers
Cranking the verbosity shows that the decoders are not yet providing a format that GStreamer can use:
0:00:10.653199799 8177 0xaaaac7b94200 LOG v4l2 gstv4l2.c:92:gst_v4l2_probe_template_caps: description: 'H.264 Parsed Slice Data' 0:00:10.653231882 8177 0xaaaac7b94200 LOG v4l2 gstv4l2.c:93:gst_v4l2_probe_template_caps: pixelformat: S264 0:00:10.653267756 8177 0xaaaac7b94200 DEBUG v4l2 gstv4l2object.c:1556:gst_v4l2_object_v4l2fourcc_to_bare_struct: Unsupported fourcc 0x34363253 S264 0:00:10.653383255 8177 0xaaaac7b94200 LOG v4l2 gstv4l2.c:92:gst_v4l2_probe_template_caps: description: 'MPEG-2 Parsed Slice Data' 0:00:10.653413588 8177 0xaaaac7b94200 LOG v4l2 gstv4l2.c:93:gst_v4l2_probe_template_caps: pixelformat: MG2S 0:00:10.653446838 8177 0xaaaac7b94200 DEBUG v4l2 gstv4l2object.c:1556:gst_v4l2_object_v4l2fourcc_to_bare_struct: Unsupported fourcc 0x5332474d MG2S 0:00:10.653556795 8177 0xaaaac7b94200 LOG v4l2 gstv4l2.c:92:gst_v4l2_probe_template_caps: description: 'VP8 Frame' 0:00:10.653586253 8177 0xaaaac7b94200 LOG v4l2 gstv4l2.c:93:gst_v4l2_probe_template_caps: pixelformat: VP8F 0:00:10.653618919 8177 0xaaaac7b94200 DEBUG v4l2 gstv4l2object.c:1556:gst_v4l2_object_v4l2fourcc_to_bare_struct: Unsupported fourcc 0x46385056 VP8F
So we will have to wait until the module provides H264, VP80, and MPG2 pixelformats or until GStreamer learns to use S264, MG2S and VP8F formats.
-
Well, `mpv --vo=gpu --hwdec=yes --hwdec-codecs=all -v` shows no HW decoding for a MPEG2 video, so there might still be something missing:
Quote[vd] Using underlying hw-decoder 'mpeg2_v4l2m2m'
[ffmpeg/video] mpeg2_v4l2m2m: driver 'hantro-vpu' on card 'rockchip,rk3399-vpu-dec'
[ffmpeg/video] mpeg2_v4l2m2m: driver 'hantro-vpu' on card 'rockchip,rk3399-vpu-enc'
[ffmpeg/video] mpeg2_v4l2m2m: driver 'rockchip-rga' on card 'rockchip-rga'
[ffmpeg/video] mpeg2_v4l2m2m: Could not find a valid device
[ffmpeg/video] mpeg2_v4l2m2m: can't configure decoder
[vd] Could not open codec.
[vd] Falling back to software decoding.Same for a VP8 video. I will have to see if I can get gstreamer to go further.
-
32 minutes ago, NicoD said:
I'll check it out. Seems also vdpau is supported. So maybe/hopefully it could make video to work well in mainline.
The support for the Hantro G1 VPU is implemented via video4linux2 (See Collabora's efforts: https://www.collabora.com/news-and-blog/news-and-events/linux-kernel-5.4.html) I don't think vdpau will help.
With kernel 5.5, native mpv playback of mpeg-2/VP8/VP9 seems to work well at 24fps, but skips frames at 60fps trying to output thru x11. The OpenGL ES playback is slower (--vo=gpu --gpu-api=opengl --opengl-es=yes), so there might still be work to be done to hand decoded frames to the GPU. I will run a few more tests using youtube-dl.
-
Thanks @balbes150, the 5.5 eoan works correctly on my nanopc-t4.
@NicoD: Fastest way to get bleeding edge Mesa on Ubuntu is to use Oibaf's PPA: https://launchpad.net/~oibaf/+archive/ubuntu/graphics-drivers
-
Built latest dev branch on Vagrant. No bluetooth
The hci_uart module is running, but it is not talking to any hci module.
The dmesg output seems to be missing lines along:
[ 6.957330] Bluetooth: hci0: BCM: chip id 101 [ 6.957644] Bluetooth: hci0: BCM: features 0x2f [ 6.961779] Bluetooth: hci0: BCM4354A2 [ 6.961788] Bluetooth: hci0: BCM4356A2 (001.003.015) build 0000 [ 7.965063] Bluetooth: hci0: BCM4356A2 (001.003.015) build 0266
Maybe not that exact chip since the DTS in latest kernel talks about a bcm43438.
Did I miss something in my build command line? Am I missing some firmware?
compile.sh KERNEL_ONLY=no KERNEL_CONFIGURE=no BUILD_DESKTOP=yes BOARD=nanopct4 BRANCH=dev RELEASE=bionic
-
Thanks @hjc I was able to boot Xenial desktop on my T4 from a Vagrant build.
Single Armbian image for RK + AML + AW (aarch64 ARMv8)
in General Chat
Posted
Sorry about that. I was imprecise. Tested HDMI sound this morning and it works perfectly. The 3.5mm jack is completely missing though. Not even listed in the alsamixer UI. As mentioned previously, this is not an issue.
I will try the nvme install this week-end.