Switching fb backend back to default

[Johannes Lundberg]

On Mon, Mar 18, 2019 at 19:28 Pete Wright <pete at nomadlogic.org> wrote:

>
>
> On 3/17/19 2:50 PM, Johannes Lundberg wrote:
> > On Sun, Mar 17, 2019 at 21:35 Emmanuel Vadot <manu at bidouilliste.com>
> wrote:
> >
> >> On Sun, 17 Mar 2019 16:32:43 +0000
> >> Johannes Lundberg <johalun0 at gmail.com> wrote:
> >>
> >>> On 3/17/19 3:34 PM, Greg V wrote:
> >>>>
> >>>> On Sun, Mar 17, 2019 at 3:07 PM, Johannes Lundberg
> >>>> <johalun0 at gmail.com> wrote:
> >>>>> Hi
> >>>>>
> >>>>> I'm working on making i915kms unload properly. I've come to what I
> >> think
> >>>>> is the last issue. The drm driver unloads ok, the "efifb" backend is
> >>>>> restored (according to logs) and vt_efifb_init() is being called but
> >> the
> >>>>> screen (laptop built in display) stays black. The system seems
> >>>>> operational otherwise. If I load i915kms again in this state I get
> >> back
> >>>>> a visible (i915kms) framebuffer.
> >>>>>
> >>>>> Did we ever have this working so it's known to work?
> >>>> Recently on the linux kernel mailing list:
> >>>>
> >>>> http://lkml.iu.edu/hypermail/linux/kernel/1903.1/01162.html
> >>>>
> >>>>> Of course, once native drivers like i915 or radeon take over, such a
> >>>> framebuffer is toast... [6]
> >>>>
> >>>>> [6]
> >> linux/drivers/gpu/drm/i915/i915_drv.c::i915_kick_out_firmware_fb()
> >>>>> linux/drivers/gpu/drm/radeon/radeon_drv.c::radeon_pci_probe()
> >>>> So it seems like efifb is not supposed to work after a driver has been
> >>>> loaded at least once.
> >>>>
> >>>>
> >>> Hmm, well the code is there to handle switching back to the boot time
> >>> fb. What I think is happening is that i915 powers off the displays at
> >>> unload and vt doesn't know how to power on (or that it should).
> >>>
> >>   That and if the display pipeline is de-configured or the resolution
> >> changed you cannot reset it to the original state.
> >>   Unloading drm modules is only useful for testing (and finding leaks).
> >
> > Yeah a normal user would never unload it. Since I mostly ssh to my test
> > machines I think I’m fine personally with losing the display while
> > unloading.
> >
> > Keyboard input still works though and at least it doesn’t crash anymore
> :)
> >
>
> that's awesome, so in theory we will be able to upgrade the drm-kmod and
> use the new driver without a reboot.  i like that as a hacker and
> end-user


You probably have to exit X to unload the driver so I’m not sure it’s that
much better than a reboot :) Either way, it will make simple testing a lot
easier.


>
> -pete
>
> --
> Pete Wright
> pete at nomadlogic.org
> @nomadlogicLA
>
>