Comment 77 for bug 195009

@blush

This is actually the result of the problems with the loop command. You get there assuming the loop command does work and now is expected to return an interrupt when it is used. If the command is working that should happen, but for at least some of the boards the command does not do anything. A test before will run into a timeout but the code does not take this as an indication that the command is broken. And asking Dmitry he wrote "Unfortunately there are boxes that will loose keyboard if we mark loop as broken if it times out and assume that AUX is present."
Thus the manual workaround with i8042.noloop=1 or putting a board into a quirk table.

@Goswin

No I wasn't expected but if I had spent more time on the code around I might have noticed the whole quirk tables to be valid only for "__i386__" defined which probably isn't quite enough. ;-) Ok, I put another kernel at http://people.ubuntu.com/~smb/bug195009/ (sorry I was a lazy bloke and used the same version number again). I really hope this works this time. And hopefully the other fix (very, very lazy) does not interfere...