Re: "xv" video with intel graphics?

["John D. Baker" <jdbaker%consolidated.net@localhost>]

On Sun, 27 Sep 2020, nia wrote:

> i recommend sdl video outputs. if this and opengl don't work

Nope, doesn't work:

  $ mpv -vo sdl xfrm_blast.mpeg
  The legacy option syntax ('-vo value') is deprecated and dangerous.
  Please use '--vo=value'.
  [ffmpeg/demuxer] mpeg: Packet corrupt (stream = 1, dts = 1444422).
   (+) Video --vid=1 (mpeg1video 352x240 29.970fps)
   (+) Audio --aid=1 (mp2 2ch 44100Hz)
  [vo/sdl] Using opengl
  [vo/sdl] Warning: this legacy VO has bad performance. Consider fixing your graphics drivers, or not forcing the sdl VO.
  [ao/netbsd] Opening device /dev/audio
  AO: [netbsd] 48000Hz stereo 2ch s32
  VO: [sdl] 352x240 => 352x262 yuv420p
  Memory fault (core dumped) 

Like the default [gpu], 'mpv' opens a window but does not paint it (shows
snapshot of screen region behind window).  Does not play audio either.

Backtrace is:

[...]
Core was generated by `mpv'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x00006fcf08be4dbc in rebase_depth_stencil (brw=brw@entry=0x6fcf286244b0, 
    irb=irb@entry=0x6fcf2915da40, invalidate=<optimized out>)
    at /x/netbsd-9/src/../xsrc/external/mit/MesaLib/dist/src/mesa/drivers/dri/i965/brw_misc_state.c:143
143	   intel_get_tile_masks(irb->mt->surf.tiling, irb->mt->cpp,
[Current thread is 1 (process 10)]
(gdb) bt
#0  0x00006fcf08be4dbc in rebase_depth_stencil (brw=brw@entry=0x6fcf286244b0, 
    irb=irb@entry=0x6fcf2915da40, invalidate=<optimized out>)
    at /x/netbsd-9/src/../xsrc/external/mit/MesaLib/dist/src/mesa/drivers/dri/i965/brw_misc_state.c:143
#1  0x00006fcf08be5824 in brw_workaround_depthstencil_alignment (
    brw=brw@entry=0x6fcf286244b0, clear_mask=clear_mask@entry=2)
    at /x/netbsd-9/src/../xsrc/external/mit/MesaLib/dist/src/mesa/drivers/dri/i965/brw_misc_state.c:233
#2  0x00006fcf08ba34e5 in brw_clear (ctx=0x6fcf286244b0, mask=2)
    at /x/netbsd-9/src/../xsrc/external/mit/MesaLib/dist/src/mesa/drivers/dri/i965/brw_clear.c:255
#3  0x00006fcf27c47ee5 in GL_RunCommandQueue ()
   from /usr/pkg/lib/libSDL2-2.0.so.0
#4  0x00006fcf27c3cae5 in FlushRenderCommands ()
   from /usr/pkg/lib/libSDL2-2.0.so.0
#5  0x00000000004d316f in draw_image (vo=vo@entry=0x6fcf2ce64540, 
    mpi=0x6fcf296326c0) at ../video/out/vo_sdl.c:879
#6  0x00000000004cef1e in render_frame (vo=0x6fcf2ce64540)
    at ../video/out/vo.c:908
#7  vo_thread (ptr=0x6fcf2ce64540) at ../video/out/vo.c:1034
#8  0x00006fcf2480c072 in pthread__create_tramp (cookie=0x6fcf2cbe2000)
    at /x/netbsd-9/src/lib/libpthread/pthread.c:592
#9  0x00006fcf22687b70 in ?? () from /usr/lib/libc.so.12
Backtrace stopped: Cannot access memory at address 0x6fcf0a5ef000
(gdb) list *0x00006fcf08be4dbc
0x6fcf08be4dbc is in rebase_depth_stencil (/x/netbsd-9/src/../xsrc/external/mit/MesaLib/dist/src/mesa/drivers/dri/i965/brw_misc_state.c:143).
138	{
139	   const struct gen_device_info *devinfo = &brw->screen->devinfo;
140	   struct gl_context *ctx = &brw->ctx;
141	   uint32_t tile_mask_x = 0, tile_mask_y = 0;
142	
143	   intel_get_tile_masks(irb->mt->surf.tiling, irb->mt->cpp,
144	                        &tile_mask_x, &tile_mask_y);
145	   assert(!intel_miptree_level_has_hiz(irb->mt, irb->mt_level));
146	
147	   uint32_t tile_x = irb->draw_x & tile_mask_x;


> something needs fixing system wide, falling back to x11 is

Neither 'mpv' nor 'mplayer' "fall back" to "--vo=x11", they just fail
to display the content.  I have to force "x11" either explicitly or
in the system-wide config file to have them display.

> indicative of a problem. ffplay (another player) should also
> be extremely reliable.

I'll see about trying that.

> my impression is that the "2014" driver should no longer be
> needed - at some point newer NetBSD automatically uses an uxa
> AccelMethod on older intel gpus.  but even software gl with 
> llvmpipe and wsfb should be better than broken-mess-drivers.

I only mentioned the 2014 intel driver as being what was still in
use the last time I remembered using 'mpv' or 'mplayer' without
having to make explicit "--vo=<foo>" selections.

I'm using the 2019 intel driver which now properly falls back to UXA
automagically on hardware for which SNA is completely broken (which is
all x86_64 systems w/intel graphics that I have).

-- 
|/"\ John D. Baker, KN5UKS               NetBSD     Darwin/MacOS X
|\ / jdbaker[snail]consolidated[flyspeck]net  OpenBSD            FreeBSD
| X  No HTML/proprietary data in email.   BSD just sits there and works!
|/ \ GPGkeyID:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645