Re: [Bluez-users] keyboard and mouse lost when bluez does things

Re: [Bluez-users] keyboard and mouse lost when bluez does things

[Mikolaj J. Habryn]

On Fri, 2002-09-13 at 04:19, Maksim (Max) Krasnyanskiy wrote:
> At 02:27 PM 9/12/2002 +1000, Mikolaj J. Habryn wrote:
> >I have a Brainboxes CF card inside a Toshiba Portege 2000. When I do
> >just about anything with it - hciattach, for example - the keyboard and
> >mouse stop responding. The interrupt counts in /proc/interrupts for them
> >stop incrementing (*boggle*). Switching into and out of X fixes it. Any
> >suggestions?
> I never seen this kind of thing before.
> What kernel is this ?

I was afraid of that :P I see it with 2.4.20-pre{5,6}, 2.4.19 (I think),
and 2.5.32. It's a Toshiba Portege 2000, and I guess that means that
something excitingly wacky is happening with the interrupt hardware.
linux-kernel cc'd in case anyone there has ideas.

I did do some instrumenting inside the bluez code, and it appears to
lose the plot within the serial code somewhere - but not necessarily on
any given call into there. The behaviour is quite odd - for example,
starting an l2ping will immediately lock out the keyboard and mouse, and
the first ping response will show an RTT of ~5500 ms, following which
the remainder will come back within ~45ms. If I leave l2ping running
while restoring the keyboard and mouse by switching in and out of X,
they'll almost immediately freeze up again and there'll be another ~5500
ms burp in the l2ping replies.

m.