On 7/2/22 11:05, Hans de Goede wrote:
[...]
A few more notes on the Surface Laptop Studio:
A peculiarity of the Surface Laptop Studio is its "slate/tent" mode
(symbolized: user> _/\). In this mode, the screen covers the keyboard
but leaves the touchpad exposed. This is essentially a mode in-between
tablet and laptop, and it is debatable whether tablet-mode should be
enabled in this mode. We therefore let the user decide this via a module
parameter.
In particular, tablet-mode may bring up the on-screen touch keyboard
more easily, which would be desirable in this mode. However, some
user-space software currently also decides to disable keyboard and, more
importantly, touchpad input, while the touchpad is still accessible in
the "slate/tent" mode. Furthermore, this mode shares its identifier with
"slate/flipped" mode where the screen is flipped 180° and the keyboard
points away from the user (symbolized: user> /_). In this mode we would
like to enable auto-rotation, something that user-space software may
only do when tablet-mode is enabled. We therefore default to the
slate-mode enabling the tablet-mode switch.
Note that you could add a libinput quirk to ignore the tablet-switch
for keybd/touchpad event suprression since when not used the keyboard /
touchpad are covered I expect them to not send false input events so
this should be fine. Then the touchpad will keep working in "slate/tent"
mode while still enabling the onscreen-kbd since you are reporting
sw_tablet_mode=1.
The onscreen-kbd and auto-rotation enabling/disabling lives outside of
ibinput. Although gnome-shell does use libinput to read the touchpad and
currently I think the only way to get libinput to ignore the tablet-mode-sw
is to mark it as unreliable, which I guess might also stop it from reporting
it to gnome-shell.
Another option would be to mark the touchpad as "external" with a libinput
quirk (so not build into the hw) then it also won't suppress events based
on the tablet-mode-sw state. TL;DR: this can be fixed with a quirk on
the libinput side, how best to do that needs to be discussed with the libinput
devs.
And once solved at the libinput side, you probably don't need
the module-parameter. Note I'm fine with keeping it, but feel free to
submit a follow-up patch to remove it, explaining that this is fixed
in libinput now...
Thanks for that hint!
Turns out libinput has a quirk for that (ModelTabletModeNoSuspend). I'm
working with someone actually owning a SLS and we will submit an entry
to the libinput hwdb once we've fleshed that out.
I'll yet have to think some more about whether it makes sense to keep
the module parameter or not. I think ideally there would be a way to
notify user-space for more states than just tablet/laptop, but that is
very device specific.
Regards,
Max
