Re: [PATCH v4 0/3] console/fbcon: Add support for deferred console takeover

[Hans de Goede]

Hi,

On 27-06-18 11:47, Bartlomiej Zolnierkiewicz wrote:
> On Wednesday, June 27, 2018 11:15:20 AM Daniel Vetter wrote:
> > On Tue, Jun 26, 2018 at 08:36:09PM +0200, Hans de Goede wrote:
> > > Hi All,
> > >
> > > Here is v4 of my patch-set, to delay fbcon taking over the console (and
> > > binding to fbdev devices) until there actually is some text output to the
> > > console. This is intended for use with the "quiet" cmdline option, in
> > > combination with a bootloader which leaves the vendor's logo /
> > > EFI bootgraphics put up by the firmware intact on the EFI framebuffer.
> > >
> > > The end goal here is a boot where the firmware shows its boot graphics
> > > and these stay in place for a couple of seconds until the GUI loads and
> > > the GUI then smoothly takes over the framebuffer without any distruptions.
> > >
> > > This patch-set spans 2 subsystems.
> > >
> > > Petr, the printk subsys change is really trivial (1 line addition) can we
> > > get your Acked-by for merging all 3 patches through the fbdev tree?
> > >
> > > Changelog:
> > >
> > > Changes in v4:
> > > -Keep the comments about which fbcon functions need locks in place
> > >
> > > Changes in v3:
> > > -Export is_console_locke() for use in modules (as fbcon may be built as a .ko)
> > > -Use WARN_CONSOLE_UNLOCKED() in several places in the fbcon code to assert
> > >   proper locking (requested by Daniel)
> > > -Unregister the fbcon-dummycon-output-notifier on fbcon_exit() (req. by Daniel)
> > > -Document the fbcon=nodefer commandline option (req. by Emil)
> > >
> > > Changes in v2:
> > > -Check the whole string when checking for erases in putcs, instead of just
> > >   the first char
> > > -Make dummycon_blank return 1, so that a redraw gets triggered and any text
> > >   rendered while blanked gets output so that it can trigger a deferred
> > >   takeover if one is pending
> >
> > Wrt merging I think it'd be best if we stuff this into drm-misc-next -
> > that will increase testing by gpu drivers a lot, instead of a suprise when
> > the fbdev pull lands in upstream.
> >
> > Bart, is that ok with you?
>
> Not really, since there are efifb changes in the queue which depend
> on this series I would really prefer to merge all patches through
> fbdev tree.
>
> Also fbdev tree is pulled into -next kernels so testing coverage
> should be okay (I assume that everybody are testing -next kernels in
> addition to their own branches :-)..

If you are talking about the "efifb: Copy the ACPI BGRT boot graphics to the
framebuffer" series, I could push those to drm-misc-next too (once acked).

I think most GPU driver developers are running drm-tip and not
-next, so putting things in drm-misc-next would give the changes somewhat
more test-exposure on a wider range of GPUs I believe. Where as -next
testing will likely be more server use-case oriented.

Alternatively you could merge things in the fbdev tree, do an
unmutable branch and then that could be merged into drm-misc-next by
the drm-misc-next maintainers.

Note either way is fine with me. This is up to you and Daniel.

Regards,

Hans