11:35 <urjaman> hanetzer: did you get anywhere with the dwc2 hot plug thing?

11:37 <urjaman> i tried to take a look (and do a google) while updating my kernel so i found this https://www.mail-archive.com/linux-usb@vger.kernel.org/msg104033.html and got the same message ("Session request interrupt - lx_state=2") as they and also similarly forgot to enable verbose debug so doing that now and i'll see what it says...

13:44 <via> what is the best way to get 2d acceleration nowadays on debian testing with a rk3288?

13:45 <via> i don't think i've ever had any acceleration (just using fbdev?) but i just upgraded from stretch to testing and now even console updating is realllly slow

14:50 <via> i compiled armsoc xserver but when it starts up i get:

14:50 <via> [ 229.112] (EE) _CREATE_GEM({height: 768, width: 1366, bpp: 32 buf_type: 0x0}) failed. errno: 22 - Invalid argument

14:50 <via> [ 229.112] (EE) ARMSOC(0): ERROR: Cannot allocate scanout buffer

14:51 <via> https://gist.github.com/via/789a1383918578dcc935f08032c0955c

14:54 <via> this is a mainline kernel (4.18.5) and debian testing's vanilla xserver

16:52 <mmind00> via: the armsoc xserver is somewhat deprecated as it is mainly used to making the mali binary blob work, so in general it won't accelerate your 2d experience much

16:53 <mmind00> via: also it depends on a special ioctl in the drm-driver, which is not part of the mainline kernel

16:53 <mmind00> via: and probably won't be, due to armsoc being somewhat of a stopgap measure

16:56 <mmind00> via: https://github.com/mmind/linux-rockchip/commit/92e4c5196c40c570dbf82fba743b7cf26d476d87#diff-8072ad4bee95165e2224d7cce3750678 for the old ioctl patch

16:57 <via> what should i use instead of armsoc?

17:01 <via> i only care about 2d, and prefer not using a binary blob

17:02 <urjaman> on my c201, the two best are the modesetting one (comes with xorg) - disable glamor though, or fbturbo

17:02 <via> glamor is enabled, and using llvmpipe

17:02 <via> i should disable it?

17:02 <via> this is a c201 as well

17:02 <urjaman> for me it segfaulted on xorg start :P

17:03 <urjaman> but this is arch-arm, not debian

17:03 <sphalerite> modesetting works for me, also on a c201, but on nixos, not debian (idk about debian)

17:03 <urjaman> but yeah a xorg.conf.d fragment with Driver "modesetting" and AccelMethod "none" made it work for me so ...

17:03 <via> cool. yeah i mean its working for me, i just assume its not actually accelerating anything

17:04 <sphalerite> I also didn't need any special config

17:04 <sphalerite> via: I don't think actual acceleration is possible without a blob currently.

17:04 <sphalerite> No idea how to set up the blob though.

17:05 <via> ok

17:05 <via> it did turn out most of what i thought was suddenly things being much slower was specificlaly just my terminal

17:05 <via> and something about xft

17:05 <via> and i fixed it

17:05 <via> so everythings kind of the way it was now

17:06 <sphalerite> via: what desktop are you using? It also makes a lot of sense not to run a compositor :)

17:06 <via> dwm

17:06 <via> def not compositing

17:07 <via> but having some ...intense things in a browser (like slack) work, would be nice

17:10 <sphalerite> one day panfrost will be finished and everything will be wonderful :p

17:12 <via> i was eyeing that

17:13 <via> not much in the way of buildinstructions so i assume its not ready to even try

17:13 <via> and its just 3d isn't it?

17:13 <urjaman> "just"

17:14 <urjaman> (=it should be able to run glamor and thus composite the whole darn display even if the windows are just 2d things... and your browser also uses GL to render if available so...)

17:14 <via> right, ok

17:15 <urjaman> but yeah you should be able to get video scaling from the 2D hardware

17:16 <urjaman> (just to showoff, mpv -vo drm on the command line) and there's some work on the rockchip VPUs too ... so maybe some day hardware video decoding

17:16 <urjaman> though for h264 which isnt what most of eg. youtube is by now but ...

17:21 <via> i don't have a lot of issue with video playback suprisingnly

17:21 <via> i mean, for youtube

17:26 <urjaman> yeah the CPU is enough for 720p (and browsers are surprisingly about the best video player when there is no hardware accel...)

17:27 <urjaman> 720p60 and >=1080p is too much though

18:36 <hanetzer> urjaman: switched to gentoo, applied that host->reset patch, turned off gadget mode (host-only), no dice

18:36 <urjaman> yeah i remember hearing you mention a patch here but wasnt around to see the actual patch :P

18:37 <urjaman> i enabled the verbose debug but got nothing more on device attach

18:57 <urjaman> i think this is just super weird

18:59 <urjaman> an entire class of usb controllers dont hot-plug at all for months and nobody has bothered to get to the bottom of it? ...