Video and Keymap Quirks

[Richard Hughes]

On Thu, 2007-07-12 at 16:59 -0500, Michael E Brown wrote:
> On Thu, Jul 12, 2007 at 09:20:56PM +0100, Richard Hughes wrote:
> > On Thu, 2007-07-12 at 13:00 -0500, Michael E Brown wrote:
> > > I havent had a lot of time to pursue it, but I have seen several people
> > > working on things to handle killswitches. The problem I've seen is that
> > > several of them dont take into account features that the Dell notebooks
> > > have: software configurable killswitches, and killswitch for cellular in
> > > particular. 
> > 
> > Why do we need to configure a killswitch? Surely a kill switch is just
> > something that kills _all_ wireless for aircraft. If you want to power
> > down individual devices surely this should be done in sysfs.
> 
> The word "surely" implies policy.

No the work surely was intended to mean "it should".

> The dell implementation provides
> mechanism, the user provides policy. Aircraft is one policy.
> I'm-sitting-at-starbucks-and-dont-want-somebody-to-hack-my-wireless-
> while-I-upload-pictures-from-my-phone-to-my-laptop-over-bluetooth
> is another policy.

Then you don't use a kill switch, you just don't set your bluetooth
adapter to discoverable. Bluetooth adapters use lots of power - the
device should be powered down when not in use and only powered up when
needed. But that's not my point.

> The user can decide which policy they want the
> killswitch to provide via either BIOS configuration options or through
> dellWirelessCtl.

It seems to me the default action should be kill _everything_ and only
lets really advanced users who know what thay are doing can use
dellWirelessCtl.

> Writing software that just assumes that the killswitch can only ever do
> an all-or-nothing operation is encoding bad policy in our software.

I disagree. Getting on a plane and switching the kill switch to off,
only to find it's only disabled the bluetooth and not the wlan is a
totally bad design decision in my opinion.

> Also, sysfs, as I've mentioned elsewhere, is a crappy interface for this
> specific functionality. dellWirelessCtl is implemented in userspace, and
> it doesnt make sense to try to write a kernel module to do it. 

Unfortunately I think Dell hardware is the odd-one-out here - in all the
hardware I've used the kill switch is either implemented in hardware or
involves writing low level commands or poking values in registers, and
hence handled in kernel drivers.

Richard.