<prahal>
liquidAcid: hi , blacklist s5p_jpeg to get s5p-mfc dec at video0
<prahal>
otherwise v4l2video0dec fails (I have not spent enough time to dig where it hardcode the device node :/
<liquidAcid>
prahal, did you try with iommu or without?
<prahal>
not enabled iommu yet ,/
<liquidAcid>
k, thx for testing
<prahal>
upgraded to latest and added most of your patchset
<liquidAcid>
but still on 4.0.y?
<Wizzup_>
what patchset?
<prahal>
a little above : took time to get the proper zpos for the xf86-armsoc HW cursor since set_zpos is no more
<liquidAcid>
prahal, yes, if you've read the ml, i'm planning on rewriting this
<prahal>
not yet , I manage to get hw cursor back (there are two set of overlay at zpos 1 , and I had to select the one from the second set , fun )
<liquidAcid>
it's pretty messy at the moment :D
<prahal>
let me read the ml , this was kind of epic (the dma pl330 I had to tweak to get audio back
<liquidAcid>
also let me update odroid-4.0.y, i think i forgot to do that today
<prahal>
the only thing that postponed a push was that I get an new audio issue when I use gstreamer + s5p-mfc , every 2 seconds a blank (it can take 30 minutes to appear but then it keep doing so
<liquidAcid>
speaking of gstreamer, do you know if they included kmssink at some point?
<liquidAcid>
i'm still looking for some small test application to debug nv12mt output
<prahal>
cannot tell
<liquidAcid>
prahal, what sink do you use for testing?
<prahal>
ximage or cluttersink(snappy or totem) / glimagesink (with gst1.0 plugins bad configure tweak to egl + gles2
<prahal>
I was not tweaking anything gst wise with 3.19 code ...
<liquidAcid>
selecting the overlay plane with largest zpos should work there
<prahal>
I do not know
<prahal>
liquidAcid: the issue was that drm returned 4 overlay : two with zpos 1, two with zpos 2 , hw cursor was invisible it I select the max zpos in the first set
<liquidAcid>
prahal, so 5 planes in total?
<prahal>
with primary ? I have not searched for why I get 4 overlays , could be the primary are dup too
<prahal>
the primary is not returned by drm in default mode