Re: bin/57591: i915 graphics console goes blank during boot

[Ramiro Aceves <ea1abz%gmail.com@localhost>]

Nice to see that the bug I found in a old  i386 32 bit machine with
Taylor's wisdom fixes problems in newer machines (PR/57833).
Interesting that the same i915 NetBSD10.0_RC3 code did not show any
problems in my Intel 8i7BEH Nuc.


In my total ignorance of all this, I see that it is important to start
old machines even if they are hardly going to be used daily.

Regards.

El mar, 6 feb 2024 a las 16:08, Havard Eidnes (<he%netbsd.org@localhost>) escribió:
>
> >       Booting this PC (HP EliteDesk 800 G1 SFF) which has an i7-4790
> >       CPU and uses the built-in graphics hardware gives one of two
> >       results:
> >
> >       1) with the VGA screen as console, the two screens I have
> >          connected both go blank and enter "energy save mode" before
> >          the boot completes
> >
> >       2) with a serial console, the boot succeeds, and I get a login
> >          prompt on both screens, and X11 can be started the usual
> >          way after login (I'm old-fashioned and use startx).
> >
> >       Incidently, this was the state of NetBSD 10.0_BETA as of Thu
> >       Jan 12, but it remains as current behaviour with this recent
> >       NetBSD 10.0_BETA from Fri Aug 18.
> >
> >       Given that the screen goes blank, it's a little hard to tell
> >       if the kernel has crashed or whether this only affects the
> >       screen output (although I could perhaps wait to see if the
> >       network is configured...).
>
> I got a hint to try with a kernel with DDB_ONPANIC=0, and do a
> second boot using the serial console (which makes it work), to
> collect any kernel messages to narrow down the problem.
>
> However, I also updated to the latest netbsd-10 code at the time
>
> % uname -a
> NetBSD xxxxxx.uninett.no 10.0_RC3 NetBSD 10.0_RC3 (GENERIC) #9: Mon Feb  5 17:38:53 CET 2024  he%xxxxx.uninett.no@localhost:/usr/obj/sys/arch/amd64/compile/GENERIC amd64
> %
>
> and I'm now happy to report that this system no longer needs to
> use the serial console to come properly up.
>
> This update included the recent pull-up which included these
> files:
>
> P sys/external/bsd/drm2/dist/drm/i915/gem/i915_gem_mman.c
> P sys/external/bsd/drm2/dist/drm/i915/gem/i915_gem_pages.c
> P sys/external/bsd/drm2/dist/drm/i915/gem/i915_gem_phys.c
> P sys/external/bsd/drm2/dist/drm/i915/gem/i915_gem_region.c
> P sys/external/bsd/drm2/dist/drm/i915/gem/i915_gem_stolen.c
>
> The last part of dmesg now displays:
>
> [     3.518584] boot device: wd0
> [     3.518584] root on wd0a dumps on wd0b
> [     3.548584] root file system type: ffs
> [     3.558584] kern.module.path=/stand/amd64/10.0/modules
> [     3.568583] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
> [     3.568583] [drm] Driver supports precise vblank timestamp query.
> [     3.568583] i915drmkms0: interrupting at msi5 vec 0 (i915drmkms0)
> [     3.568583] [drm] Initialized i915 1.6.0 20200114 for i915drmkms0 on minor 0
> [     3.608592] i915drmkms0: [drm] Cannot find any crtc or sizes
> [     3.618582] [drm] DRM_I915_DEBUG enabled
> [     3.618582] [drm] DRM_I915_DEBUG_GEM enabled
> [     3.758584] intelfb0 at i915drmkms0
> [     3.758584] warning: /usr/src/sys/external/bsd/drm2/dist/drm/i915/intel_uncore.c:1197: Unclaimed write to register 0x41000
> [     3.758584] warning: /usr/src/sys/external/bsd/drm2/dist/drm/i915/intel_uncore.c:1197: Unclaimed read from register 0x41000
> [     3.758584] intelfb0: framebuffer at 0xe0009000, size 2560x1440, depth 32, stride 10240
> [     3.828583] wsdisplay0 at intelfb0 kbdmux 1: console (default, vt100 emulation), using wskbd0
> [     3.828583] wsmux1: connecting to wsdisplay0
> [    16.228577] wsdisplay0: screen 1 added (default, vt100 emulation)
> [    16.228577] wsdisplay0: screen 2 added (default, vt100 emulation)
> [    16.228577] wsdisplay0: screen 3 added (default, vt100 emulation)
> [    16.228577] wsdisplay0: screen 4 added (default, vt100 emulation)
>
> and it now basically works, currently running dual-headed in-tree
> X11 server, using the "modesetting" driver.
>
> Regards,
>
> - Håvard