On Wed, May 05, 2021 at 01:38:06PM +0200, Sebastien Marie wrote:
> On Wed, May 05, 2021 at 12:52:54PM +0200, Hiltjo Posthuma wrote:
> >
> > Hi,
> >
> > I can (still) reproduce this issue. I don't use Iris though.
> >
> > Using LD_PRELOAD still crashes for me too. Did you make other changes to the
> > mpv code?
> >
>
> the library to use in LD_PRELOAD will depend of the gpu. on mine
> system, it is radeonsi_dri.so for example.
>
> you could try with LD_DEBUG=1 to see the ld.so activity:
>
> $ LD_DEBUG=1 mpv --vo=gpu file.mp4
> [...]
> (+) Video --vid=1 (*) (h264 640x360 25.000fps)
> (+) Audio --aid=1 (*) (aac 2ch 48000Hz)
> tib new=0x790502c1000
> dlopen: loading: /usr/X11R6/lib/modules/dri/radeonsi_dri.so
> flags /usr/X11R6/lib/modules/dri/radeonsi_dri.so = 0x0
> [...]
> Exiting... (Quit)
> free tib=0x790685f6800
> free tib=0x790502c1a00
> free tib=0x790685f6400
> free tib=0x7908242bc00
> pthread_mutex_destroy on mutex with waiters!
> free tib=0x790502c1600
> free tib=0x78fe0bb9800
> free tib=0x790502c1400
> free tib=0x790502c1200
> doing dtors obj 0x79020c0cc00 @0x790ca41c700:
> [/usr/X11R6/lib/modules/dri/radeonsi_dri.so]
> doing dtors obj 0x7908242a400 @0x7907e37c3d0: [/usr/lib/libelf.so.3.0]
> doing dtors obj 0x79042c59000 @0x790cd9b4ad0: [/usr/lib/libLLVM.so.5.0]
> doing dtors obj 0x79017a94c00 @0x7908e87a520:
> [/usr/X11R6/lib/libdrm_radeon.so.4.0]
> doing dtors obj 0x79042c59400 @0x79026e63b70:
> [/usr/X11R6/lib/libdrm_amdgpu.so.1.10]
> unload_shlib called on /usr/X11R6/lib/modules/dri/radeonsi_dri.so
> unload_shlib called on /usr/X11R6/lib/libglapi.so.0.3
> unload_shlib called on /usr/lib/libelf.so.3.0
> unload_shlib unloading on /usr/lib/libelf.so.3.0
> unload_shlib called on /usr/lib/libLLVM.so.5.0
> unload_shlib unloading on /usr/lib/libLLVM.so.5.0
> unload_shlib called on /usr/lib/libexpat.so.12.0
> unload_shlib called on /usr/lib/libz.so.5.0
> unload_shlib called on /usr/lib/libm.so.10.1
> unload_shlib called on /usr/X11R6/lib/libdrm.so.7.9
> unload_shlib called on /usr/X11R6/lib/libdrm_radeon.so.4.0
> unload_shlib called on /usr/X11R6/lib/libdrm.so.7.9
> unload_shlib unloading on /usr/X11R6/lib/libdrm_radeon.so.4.0
> unload_shlib called on /usr/X11R6/lib/libdrm_amdgpu.so.1.10
> unload_shlib called on /usr/X11R6/lib/libdrm.so.7.9
> unload_shlib unloading on /usr/X11R6/lib/libdrm_amdgpu.so.1.10
> unload_shlib called on /usr/lib/libc++.so.8.0
> unload_shlib called on /usr/lib/libc++abi.so.5.0
> unload_shlib called on /usr/lib/libpthread.so.26.1
> unload_shlib unloading on /usr/X11R6/lib/modules/dri/radeonsi_dri.so
> Segmentation fault
>
>
> I wonder if /usr/X11R6/lib/modules/dri/*.so libraries should be linked
> with -znodelete to keep them after loading. Another possibility would
> be using RTLD_NODELETE on dlopen() call, but it needs to be
> implemented first (it should be simple to do as we support
> DF_1_NODELETE Elf flag).
>
> It would be interesting to know what Linux is doing here.
>
> Thanks.
> --
> Sebastien Marie
Thanks that worked! I used the wrong .so. Now it also doesn't crash anymore.
For completeness I used:
LD_PRELOAD=/usr/X11R6/lib/modules/dri/i965_dri.so LD_DEBUG=1 \
mpv -vo=gpu 'https://www.youtube.com/embed/QTk688cXTsA'
--
Kind regards,
Hiltjo
