narmstrong changed the topic of #linux-amlogic to: Amlogic mainline kernel development discussion - our wiki http://linux-meson.com/ - ml linux-amlogic@lists.infradead.org - Publicly Logged on https://irclog.whitequark.org/linux-amlogic
Barada has joined #linux-amlogic
chewitt has quit [Quit: Zzz..]
Darkmatter66 has joined #linux-amlogic
chewitt has joined #linux-amlogic
ldevulder has joined #linux-amlogic
afaerber has joined #linux-amlogic
gdeverlant_ has joined #linux-amlogic
<gdeverlant_> hi everyone im back from the dead
chewitt has quit [Quit: Zzz..]
afaerber has quit [Quit: Leaving]
brads has joined #linux-amlogic
chewitt has joined #linux-amlogic
<brads> I did some testing with I2S output on the Odroid C2 using development audio patches from Jerome's git - https://github.com/jeromebrunet/linux/tree/wip/meson/gx-audio using basic pcm5102 codec driver with a uda1380. Audio quality is excellent only after I invert the bitclock and it is working with HDMI / audio patches I grabbed from Neil's yocto git. Look forward to seeing the end result of the Audio drivers for
<brads> meson-gx. Thanks for great work!
gdeverlant_ has quit [Ping timeout: 240 seconds]
gdeverlant_ has joined #linux-amlogic
brads has quit [Read error: Connection reset by peer]
ldevulder has quit [Ping timeout: 252 seconds]
brads has joined #linux-amlogic
rcasa has joined #linux-amlogic
<Darkmatter66> Hi, have anyone tried the video decoder by Maxime Jourdan on boards other than the Le potato ?
<Darkmatter66> I'm testing on an S905 box and ffmpeg just hangs
<Darkmatter66> Can't dump the frames or play the video!
afaerber has joined #linux-amlogic
<mjourdan> Darkmatter66: hi, I test the vdec regularly on an Odroid-C2 (S905). Where did you get the patches and what is your ffmpeg command line ?
<mjourdan> (also, what ffmpeg version?)
<mjourdan> And make sure you got the firmwares! https://github.com/chewitt/meson-firmware, install the meson folder in /lib/firmware
<Darkmatter66> mjourdan: I tried to compile the kernel directly from your github repo and also tried to apply you patches that you sent to the mailing list on top of 4.18 and 4.19 mainline tree
<Darkmatter66> all gives the same problem
<Darkmatter66> ffmpeg hangs at "[h264_v4l2m2m @ 0xaaab147a10c0] output POLLERR"
<Darkmatter66> ffmpeg version is "3.4.4-0ubuntu0.18.04.1"
<Darkmatter66> I used this command "ffmpeg -c:v h264_v4l2m2m -i ~/Desktop/Test.mp4 -f rawvideo output_nv12"
<Darkmatter66> I have the firmwares in /lib/firmware/meson like so
<Darkmatter66> inside gx (h263_mc vh265_mc vh265_mc_mmu vmjpeg_mc vmpeg12_mc vmpeg4_mc_5)
<Darkmatter66> inside gxbb (vh264_mc)
<mjourdan> firmwares are ok
<mjourdan> ffmpeg version is a bit old, some v4l2m2m stuff was fixed in 4.0+
<mjourdan> does dmesg say anything interesting ?
<Darkmatter66> dmesg doesn't say anything, used to complain about missing firmware and I added it
<Darkmatter66> I think this is the last ffmpeg version in ubuntu repos
<Darkmatter66> doo you compile ffmpeg from source ?
<Darkmatter66> this is the full ffmpeg log (https://pastebin.com/x9pH5aJw)
<mjourdan> I build images via yocto which has a ffmpeg 4.0.2 recipe
<mjourdan> I don't think I've ever used something older than 4.0
<Darkmatter66> oh, I will try to compile ffmpeg or build a yocto image
<mjourdan> Can you try
<mjourdan> echo 'module meson_vdec +p" > /sys/kernel/debug/dynamic_debug/control
<mjourdan> dmesg -n 8
<mjourdan> run ffmpeg, then give me the dmesg ?
<Darkmatter66> I based my work on your post in odroid forums on which you mentioned that ffmpeg version should be >3.4
ldevulder has joined #linux-amlogic
<mjourdan> Indeed I did..
<Darkmatter66> Ok I tried and this is the log https://pastebin.com/Epu4EFPZ
<mjourdan> let me fire up my odroid c2 and see what's what. What branch did you se ?
<mjourdan> use*
<mjourdan> (from my repo)
<Darkmatter66> 4.18/amlogic-v4l2-m2m
<Darkmatter66> and used the meson-gxbb-p201 device tree
<mjourdan> Ahh sorry, that branch is a bit old.. I should delete the older ones ..
<mjourdan> If you don't mind trying with 4.18/v4l2-m2m-pr (or https://github.com/Elyotna/linux/commits/4.19/v4l2-m2m-pr), that's be great
<Darkmatter66> Ok I will try it and report back
<mjourdan> The log seems to indicate that there's no irq from the decoder except for the first one, but I think it's better to just try with a more recent branch
<Darkmatter66> Did you test the decoder with gstreamer ? does it work with everything-in-one pipes like playbin ?
<Darkmatter66> something like "gst-launch-1.0 playbin uri=file:///home/mohammad/Desktop/Test.mp4"
<Darkmatter66> Ok I will try with the newer branches but I have to wait a few hours before I can build a new kernel, I will report back here.
<mjourdan> Not really. I *think* I tried gst-play a few months back but don't remember what happened
<narmstrong> Darkmatter66: it will work but gst fails rendering because fbdev sink is crap, and kmssink does not know the meson drm driver... you need to specify a full gst pipeline to make it work
<Darkmatter66> narmstrong: what pipeline do you use ? is specifying the sink as kmssink sufficient ?
<mjourdan> Darkmatter66: I tested on an odroid-c2 (S905) with 4.18/v4l2-m2m-pr branch, decoding a h264 file worked (with ffmpeg 4.0.2). Please poke if it doesn't work out with the new branch.
<narmstrong> gst-launch-1.0 souphttpsrc location=http://download.blender.org/peach/bigbuckbunny_movies/big_buck_bunny_720p_h264.mov ! parsebin ! v4l2video0dec ! videoconvert n-threads=4 ! kmssink driver-name=meson force-modesetting=true connector-id=31 sync=false
<narmstrong> *this doesn't use the drm overlay rendering*
<narmstrong> I failed using the overlay using kmssink...
<narmstrong> maybe ndufresne could help us !
<mjourdan> Although it also works with 4.18/amlogic-v4l2-m2m...
<mjourdan> Can you upload your mp4 sample somewhere ?
<Darkmatter66> narmstrong: what's your gstreamer version ? mine is "1.14.1" but I get this
<Darkmatter66> WARNING: erroneous pipeline: no element "v4l2video0dec"
<Darkmatter66> although I have the good plugins installed
<narmstrong> mine is 1.12
<narmstrong> run gst-inspect and grep on v4l2
<narmstrong> I think the name has changed since 1.14
<mjourdan> It might be v4l2h264dec now..
<Darkmatter66> yeah I think so too
<Darkmatter66> there is now v4l2h264dec and v4l2h263dec
<Darkmatter66> mjourdan: I can upload the sample but I think I also tested on the big_buck_bunny_1080p_H264_AAC_25fps_7200K.mp4 sample with the same problem
<mjourdan> okay, nevermind then :< .
<mjourdan> Btw, are you sure meson-gxbb-p201 corresponds to your device ?
<Darkmatter66> mjourdan: just tested with ffmpeg compiled from git "git-2018-09-08-6e15495" with the big buck bunny
<narmstrong> Darkmatter66: can you send the full kernel boot dmesg ?
<Darkmatter66> the "[h264_v4l2m2m @ 0xaaab147a10c0] output POLLER" is gone but ffmpeg still hangs
<Darkmatter66> on the amlogic kernel my device used to boot with the gxbb-p200 dtb but with mainline I use gxbb-p201 to have ethernet working (on gxbb-p200 there is no ethernet)
<mjourdan> What is your device ?
<Darkmatter66> narmstrong: full dmesg https://pastebin.com/vm9mzTa3
Barada has quit [Quit: Barada]
<Darkmatter66> mjourdan: video strong k2 pro
Barada has joined #linux-amlogic
<Darkmatter66> this is the old device tree I use to boot the 3.14 kernel https://github.com/CoreELEC/device-trees-amlogic/blob/master/gxbb_p200_k2_pro.dts
<narmstrong> it has the same SoC revision as Nanopi K2, I tested it, it works fine
<Darkmatter66> narmstrong: you tested it in yocto ? I did my tests in ubuntu(Armbian) maybe something is wrong in ubuntu userspace ? old libraries maybe?
<Darkmatter66> I probably need to build a Yocto image for testing
<Darkmatter66> I don't have access to my build machine so I can't build Yocto for testing. Can you test with Armbian Images ? 150balbes has already images for all devices
<mjourdan> Darkmatter66: Input buffers get parsed but it looks like no frames are decoded from the dmesg..
<mjourdan> I have one idea of what could go wrong but
<mjourdan> I need to be sure that a sample like big buck bunny fails as well (if you can include dmesg of bbb that fails, would be great)
<Darkmatter66> the fill dmesg I linked above for Neil is from the bbb sample
<Darkmatter66> you can see in the log that I run ffmpeg twice the first was with the ffmpeg 3 and the second with ffmpeg compiled from git
<Darkmatter66> the full*
<mjourdan> that other bbb sample you have doesn't have what interests me to know :P
<mjourdan> ok so with your same branch (4.18/amlogic-v4l2-m2m) I can reproduce with https://samplemedia.linaro.org/H264/big_buck_bunny_1080p_H264_AAC_25fps_7200K.MP4.
<mjourdan> so I'll be able to troubleshoot this
<Darkmatter66> nice
<Darkmatter66> I will test the sunflower sample shortly
<mjourdan> Nope sorry, talked too fast, it works on my end even with the old branch.. (the URL was giving me a timeout instead, had to download the file)
<mjourdan> So currently I don't know what's not working on your end.. :(. Please still try the other bbb file and the newer branches, and then we'll see
<Darkmatter66> trying to scp the sample to the box fails (the network is slow for some reason !) also the box doesn't detect a usb flash drive !
<mjourdan> There's a USB regression that was out of our hands in 4.18 for gxbb, a workaround is to use "lsusb -v", it should force the flash drive to show up.. hopefully
<Darkmatter66> sunflower sample has the same issue
<Darkmatter66> mjourdan: the lsusb workaround worked :)
<Darkmatter66> I can't build from the other branch right now as I don't have access to my build machine, but I will build it later and report back
<mjourdan> Darkmatter66: Thanks for testing the other sample! Definitely seems like an issue with the decoder driver on this particular chip, but I'm clueless :/
gdeverlant_ has quit [Ping timeout: 240 seconds]
<Darkmatter66> That's unfortunate :(
<Darkmatter66> But I think other people in the Armbian community reported that the decoder doesn't work on their boxes, so maybe this is not the only box affected ?!
Xogium has joined #linux-amlogic
<Xogium> hi guys
<Xogium> running kernel 4.17.6 on my odroid c2 here, and I noticed that whenever the board is under some load, like 1.0 load average my keyboard start acting all weird, sending they keys more than once, a good 20 or 30 times before it calmes down, then does it again if I have the misfortune of touching the keyboard… Does anyone else have that sort of problem ?
Barada has quit [Quit: Barada]
<Xogium> happens as well if I download something
gdeverlant_ has joined #linux-amlogic
<Xogium> tried another keyboard and it does it as well, guess it's some weird usb issue ?
gdeverlant_ has quit [Ping timeout: 240 seconds]
nemunaire has quit [Ping timeout: 252 seconds]
nemunaire has joined #linux-amlogic
nashpa has quit [Ping timeout: 245 seconds]
nashpa has joined #linux-amlogic
ldevulder has quit [Quit: Leaving]
vagrantc has joined #linux-amlogic
gdeverlant_ has joined #linux-amlogic
gdeverlant_ has left #linux-amlogic [#linux-amlogic]
gdeverlant_ has joined #linux-amlogic
afaerber has quit [Quit: Leaving]
cottsay has joined #linux-amlogic
cottsay has quit [Quit: TTFN]
cottsay has joined #linux-amlogic
<ndufresne> narmstrong, if you use force-modesetting, you will endup with using a primary plane and page flip
<ndufresne> narmstrong, as of current implementation, to use overlay, you need to do modesetting externally, you can use modetest for that
<ndufresne> and then by default, kmssink will search for an overlay
<Xogium> someone having clue for my issue ? :S
<ndufresne> narmstrong, I'm working with some Xilinx folks to improve this, so that kmssink gets smarter with handling overlays/underlays, if only the zorder was explicit ;-D
<ndufresne> i.e. we have a local hack callled fullscreen-overlay, which does what you think it should do
<narmstrong> ndufresne: thanks for the hint !
<narmstrong> ndufresne: is this hack easy to test ?
<ndufresne> (we are also looking into moving away from legacy drm)
<ndufresne> narmstrong, I would need to extract the patch and place them somewhere public, it's a bit of work as we internally target 1.12
<narmstrong> ndufresne: it would be cool to have, but it’s really not urgent ! I can use modetest for now, I needed to understand before
<ndufresne> let me know
<ndufresne> we already merged the CRTC restoration, which was a dep, so it's coming
<ndufresne> long term we want to make kmssink looks more like the decklinkvideosink, so it would be really for fullscreen streaming over hdmi, sdi, and other output
<narmstrong> ndufresne: ok, thanks for the status !
<Darkmatter66> narmstrong: Hi Neil, why can't I use fetchall when building amlogic-image-headless-sd in bitbake
<Darkmatter66> it says there is no fetchall task
<Darkmatter66> shouldn't this be inherited ?
<narmstrong> No idea what you are talking about
<narmstrong> Do you have a build error ?
<Darkmatter66> well normally when I build image using openembedded I use fetchall to fetch all the files required for the build first
<Darkmatter66> like this: bitbake -c fetchall amlogic-image-headless-sd
<Darkmatter66> but this is failing with the following error:
<Darkmatter66> Task do_fetchall does not exist for target amlogic-image-headless-sd (/home/ubuntu/repos/amlogic-bsp/poky/../meta-meson/recipes-core/images/amlogic-image-headless-sd.bb:do_fetchall). Close matches:
<Darkmatter66> do_fetch
<Darkmatter66> seems like the fetchall task was dropped in a recent patch
<Darkmatter66> now there is a new syntax: bitbake amlogic-image-headless-sd --runall=fetch
<Darkmatter66> sorry for the noise
<Xogium> this is starting to drive me up the wall
<Xogium> I see nothing in dmesg that would confirm or not confirm my idea that once the board is under load the usb starts to freak out
gdeverlant_ has left #linux-amlogic ["ByrdIRC"]
sputnik_ has quit [Remote host closed the connection]
sputnik_ has joined #linux-amlogic
<Xogium> fun thing is my sound card and my hard drive are not going crazy, it's only my keyboard that repeats the key multiple times soon as the board is under a bit of load
jakogut_ has quit [Quit: jakogut_]
jakogut_ has joined #linux-amlogic
aep3 has joined #linux-amlogic
Darkmatter66 has quit [Ping timeout: 252 seconds]
aep2 has quit [Ping timeout: 245 seconds]
vagrantc has quit [Quit: leaving]
rcasa has quit [Remote host closed the connection]
commavir has quit [Quit: leaving]
commavir has joined #linux-amlogic
cottsay has quit [Quit: TTFN]
cottsay has joined #linux-amlogic
<mjourdan> Xogium: sorry, never experienced USB issues except for the regression in 4.18 :<
<Xogium> mjourdan: what regression ? Sorry, not aware on that