[nouveau] xmir black window (mesa 11.0.0) and/or freezing and kernel errors

[dinamic]

wily/unity8/mir

i'm trying to run some x11 apps on unity8/mir but i get a black window, the window size looks right, tested with geany, firefox, etc (windows size is diff depending on the app). tried every Xmir arg no luck

i don't see any warnings or error messages

glamor EGL version: 1.4 (DRI2)
glamor EGL extensions: EGL_EXT_buffer_age EGL_EXT_image_dma_buf_import EGL_KHR_create_context EGL_KHR_get_all_proc_addresses EGL_KHR_gl_renderbuffer_image EGL_KHR_gl_texture_2D_image EGL_KHR_gl_texture_cubemap_image EGL_KHR_image EGL_KHR_image_base EGL_KHR_image_pixmap EGL_KHR_surfaceless_context EGL_MESA_configless_context EGL_MESA_drm_image EGL_MESA_image_dma_buf_export EGL_WL_bind_wayland_display :
glamor GL version: 3.0 Mesa 11.0.0
glamor GL extensions: GL_ARB_multisample GL_EXT_abgr GL_EXT_bgra GL_EXT_blend_color GL_EXT_blend_minmax GL_EXT_blend_subtract GL_EXT_copy_texture GL_EXT_polygon_offset GL_EXT_subtexture GL_EXT_texture_object GL_EXT_vertex_array GL_EXT_compiled_vertex_array GL_EXT_texture GL_EXT_texture3D GL_IBM_rasterpos_clip GL_ARB_point_parameters GL_EXT_draw_range_elements GL_EXT_packed_pixels GL_EXT_point_parameters GL_EXT_rescale_normal GL_EXT_separate_specular_color GL_EXT_texture_edge_clamp GL_SGIS_generate_mipmap GL_SGIS_texture_border_clamp GL_SGIS_texture_edge_clamp GL_SGIS_texture_lod GL_ARB_framebuffer_sRGB GL_ARB_multitexture GL_EXT_framebuffer_sRGB GL_IBM_multimode_draw_arrays GL_IBM_texture_mirrored_repeat GL_ARB_texture_cube_map GL_ARB_texture_env_add GL_ARB_transpose_matrix GL_EXT_blend_func_separate GL_EXT_fog_coord GL_EXT_multi_draw_arrays GL_EXT_secondary_color GL_EXT_texture_env_add GL_EXT_texture_filter_anisotropic GL_EXT_texture_lod_bias GL_INGR_blend_func_separate GL_NV_blend_square GL_NV_light_max_e
glamor GLSL version: 1.30
[DRI2] Setup complete
[DRI2] DRI driver: nouveau
[DRI2] VDPAU driver: nouveau
Resize requested to 1008x728

[Daniel van Vugt (vanvugt)]

dinamic,

Thanks for testing this. We'll have to do some more nvidia testing in the near future.

In the mean time, a major Xmir update is coming soon to wily. Stay tuned and watch for updates...
   https://launchpad.net/ubuntu/+source/xorg-server

I'm not saying that will definitely fix this bug, but a lot has changed and it's not yet released.

[Daniel van Vugt (vanvugt)]

BTW, you may get more success disabling the DRI acceleration that is enabled by default:

Xmir :0 -sw -damage ...

[dinamic (dinamic6661)]

uuuu new stuff :D can't wait

i think there is something wrong with nouveau/mesa 11.0.0, i get all sort of crashes after the last mesa landed

webbrowser-app
qml: Loaded 3 UA override(s) from file:///usr/lib/x86_64-linux-gnu/qt5/qml/Ubuntu/Web/ua-overrides-desktop.js
nouveau: kernel rejected pushbuf: No such file or directory
nouveau: ch0: krec 0 pushes 0 bufs 1 relocs 0
nouveau: ch0: buf 00000000 00000002 00000004 00000004 00000000
nouveau: kernel rejected pushbuf: No such file or directory
nouveau: ch0: krec 0 pushes 2 bufs 25 relocs 0
nouveau: ch0: buf 00000000 00000003 00000004 00000004 00000000
nouveau: kernel rejected pushbuf: No such file or directory
nouveau: ch0: krec 0 pushes 0 bufs 6 relocs 0
nouveau: ch0: buf 00000000 00000003 00000004 00000004 00000000
nouveau: ch0: buf 00000001 00000006 00000004 00000000 00000004
Segmentation fault (core dumped)

https://bugs.launchpad.net/ubuntu/+source/webbrowser-app/+bug/1499419

and kern.log

Sep 25 12:38:13 pixel-MS-7369 kernel: [ 58.943435] ibus-ui-gtk3[1493]: segfault at 31 ip 00007f78664b8bef sp 00007ffd355e2ab0 error 4 in libX11.so.6.3.0[7f7866426000+135000]
Sep 25 12:38:14 pixel-MS-7369 kernel: [ 59.809015] system-crash-no[1489]: segfault at 0 ip 00007f5a2175dd8f sp 00007ffe0bc2d460 error 4 in libmirclient.so.9[7f5a2171a000+81000]
Sep 25 12:42:14 pixel-MS-7369 kernel: [ 300.444656] nouveau E[Mir/IPC[1950]] Unknown handle 0x0000003c
Sep 25 12:42:14 pixel-MS-7369 kernel: [ 300.444688] nouveau E[Mir/IPC[1950]] validate_init
Sep 25 12:42:14 pixel-MS-7369 kernel: [ 300.444691] nouveau E[Mir/IPC[1950]] validate: -2
Sep 25 12:42:39 pixel-MS-7369 kernel: [ 324.756860] nouveau E[Mir/IPC[1950]] Unknown handle 0x0000002c
Sep 25 12:42:39 pixel-MS-7369 kernel: [ 324.756891] nouveau E[Mir/IPC[1950]] validate_init
Sep 25 12:42:39 pixel-MS-7369 kernel: [ 324.756894] nouveau E[Mir/IPC[1950]] validate: -2
Sep 25 12:42:39 pixel-MS-7369 kernel: [ 324.757819] QSGRenderThread[2300]: segfault at 7fbb0f1ce000 ip 00007fbaf9b4f687 sp 00007fbadc8ca898 error 4
Sep 25 12:42:39 pixel-MS-7369 kernel: [ 324.757819] Chrome_InProcGp[2328]: segfault at 7fbb0f1ce000 ip 00007fbaf9b4f687 sp 00007fba9bffd398 error 4 in nouveau_dri.so[7fbaf96c1000+7f7000]
Sep 25 12:42:39 pixel-MS-7369 kernel: [ 324.757832] in nouveau_dri.so[7fbaf96c1000+7f7000]
Sep 25 12:45:59 pixel-MS-7369 kernel: [ 524.621973] nouveau E[ PGRAPH][0000:04:00.0] TRAP_M2MF IN
Sep 25 12:45:59 pixel-MS-7369 kernel: [ 524.621986] nouveau E[ PGRAPH][0000:04:00.0] TRAP_M2MF 00320851 20368000 00000000 04000432
Sep 25 12:45:59 pixel-MS-7369 kernel: [ 524.621993] nouveau E[ PGRAPH][0000:04:00.0] ch 4 [0x000f939000 Mir/IPC[2393]] subc 5 class 0x5039 mthd 0x0234 data 0x00000000
Sep 25 12:45:59 pixel-MS-7369 kernel: [ 524.622002] nouveau E[ PFB][0000:04:00.0] trapped read at 0x002036c000 on channel 0x0000f939 [Mir/IPC[2393]] PGRAPH/DISPATCH/M2M_IN reason: PAGE_NOT_PRESENT
Sep 25 12:45:59 pixel-MS-7369 kernel: [ 524.622014] nouveau E[ PGRAPH][0000:04:00.0] TRAP_M2MF NOTIFY
Sep 25 12:45:59 pixel-MS-7369 kernel: [ 524.622019] nouveau E[ PGRAPH][0000:04:00.0] TRAP_M2MF 00304041 00000000 00000000 06000000
Sep 25 12:45:59 pixel-MS-7369 kernel: [...

[dinamic (dinamic6661)]

with Xmir :0 -sw -damage i occasionally see some random pixels when i resize the window X-) but that's that, still black. for now i'll probably start bugging the nouveau people :D

[Daniel van Vugt (vanvugt)]

I would not bug nouveau people till us Xmir/Mir people have investigated first.

[Daniel van Vugt (vanvugt)]

I get similarly broken behaviour using Unity8 on nouveau:
    nouveau: kernel rejected pushbuf: Device or resource busy
And worse messages appear in the dmesg kernel log. So nouveau is very broken. Nothing new there.

I do find it works better using Xmir -sw. Both geany and firefox start and are visible. However resizing the root window is indeed very buggy with "random pixels". Those random pixels are not specific to nouveau though as you get the same issue with intel graphics.

So you've actually found three separate bugs here and we need to break this one up:
  * Confirmed: nouveau is very unstable, freezes the screen and hangs the system.
  * Confirmed: resizing the root window in Xmir is very buggy -> bug 1417541
  * Unconfirmed: black window (my system usually hangs completely instead)

[Daniel van Vugt (vanvugt)]

Sadly, I must recommend you switch to Intel graphics for now. There's no telling when or if nouveau will ever be stable.

[Daniel van Vugt (vanvugt)]

To clarify: All hope is not lost for nouveau. Nouveau works and it works fine with Mir in general. And mostly fine with Unity8 too. Nouveau is only unusable with Xmir in the default DRI mode.

You can avoid that problem however using -sw mode, and you can avoid the resizing issues using -rootless but only after the new xmir with working rootless gets released. Once that happens you've mostly only got general Unity8 bugs to deal with and can then use:
    Xmir :0 -sw -rootless -flatten

So please wait for the new Xmir to be released and let us know how that goes.

[Daniel van Vugt (vanvugt)]

Xmir stability has improved a lot since I tested this last. Need to retest nvidia hardware again...

[dinamic (dinamic6661)]

hi Daniel, i'm on Xenial now. i still get the black window, i'll wait for more updates :D maybe something will change. here's a screenshot http://i.imgur.com/pSgPc8h.jpg

for my sanity... maybe i'm doing something wrong.. so to launch an X11 apps on Xmir, "Xmir :0 --desktop_file_hint=whatever.desktop". that's how i did it before

the web browser still crashes but the core apps seems to work

[dinamic (dinamic6661)]

the new xmir is not released yet (i'm not on xenial proposed yet), but i'll try "Xmir :0 -sw -rootless -flatten" when it lands

[Daniel van Vugt (vanvugt)]

Unfortunately even xenial still contains the old Xmir code, so you've got a lot to look forward to when the new code eventually arrives.

[Daniel van Vugt (vanvugt)]

Reassigning this bug to the kernel. AFAIK the only problem that was in Xmir itself was resolved as bug 1417541.

The remaining issues, while severe, are general nouveau driver/kernel issues.

[dinamic (dinamic6661)]

Hello daniel, a new xmir landed. I get the same black window but the previous bugs arr gone resizing and random pixels. When i run with rootless flatten i get the loading animations spinning into a loop but nothing happens next.

[dinamic (dinamic6661)]

The browser still crashes and i get nouveau errors so i.ll wait for future nouveau updates