On Thu, Jul 10, 2014 at 05:20:55PM +0200, Hans de Goede wrote:

Hi,

On 07/10/2014 10:45 AM, Vojtech Pavlik wrote:

quoted

On Thu, Jul 10, 2014 at 09:32:16AM +0200, Hans de Goede wrote:

quoted

quoted

quoted

quoted

I also expect
that most of external PS/2 mice are dead by now, so number of cases when

One would at least hope so. Sadly, internal mice and PS/2 keyboards
aren't going to go away any soon, due to larger power consumption of USB
devices.

Right, but we are not talking about mouse vs keyboard, they use separate
ports anyway, it is touchpad plus external PS/2 mouse case where active
MUX might help.

What about laptops with both a touchpad and a trackpoint ? I think in most
cases the trackpoint works through some sort of pass-through mode of the
touchpad (or is outright part of the touchpad ps/2 device), but are we
sure there are no cases where the trackpoint and touchpad are really
separate ps/2 devices hookedup through an active mux ?

I'm not aware of any current machines using active multiplexing for
that. There are basically two touchpad manufacturers: Synaptics and
ALPS.

And elantech

quoted

Synaptics has a nearly transparent passthrough mode and
touchpoints are connected through that. ALPS basically manufactures a
touchpad+touchpoint combo device and thus doesn't need a fully
transparent passthrough.

Elantech now a days also produces a combo like alps.

They still wire it up internally and route all data through AUX port and
not use active MUX for that.

quoted

Active multiplexing was typically used for external PS/2 ports on
laptops, because the manufacturer couldn't anticipate the protocol of
the externally connected device.

Right, but what about users who still have a laptop with an external
ps/2 port which they happen to also still use ?

The assertion that it works well now, which is a stretch. As far as I
know Windows by default does not activate it, so manufacturers rarely
test it.

quoted

quoted

quoted

quoted

quoted

we have users with PS/2 touchpad + external PS/2 mouse + working active
MUX is exceedingly small.

Let's pull Vojtech in ;)

What I'd prefer is to, based on DMI data, report but not enable by
default Active MUX mode on any machine manufactured after a certain
date. Plus have a DMI-based whitelist for machines that absolutely
needed, if any are found later.

Looking at the changes to nomux blacklist sometimes even trying MUX
messes up KBC. Instead of playing date games I'd rather simply make
i8042.nomux default. I'm fine with having whitelist for boxes that
actually need and support muxing properly.

I'm a bit skeptical about making this change, see above.

I'm not too keen about it either, as it could break existing setups. 

But I have to concede that any working hardware still using both
external and internal PS/2 and thus needing Active Multiplexing is most
likely to be found in museums today.

So the risk of breakage isn't all that big.

One of the nice things about Linux is that in general we've pretty
decent support for older hardware, so if we decide to flip the nomux
default, I think we should really do so based on some BIOS cut-off
date, so as to not break older model laptops.

I think historically IBM and HPs had working MUX support while other's
did might advertise support and fail miserably when we would try
activating it.

Given that now majority of mice are USB even if we select wrong default
on some older box I think it would be acceptable as opposed to a newer
box not having touchpad at all...

Thanks.

-- 
Dmitry