Problem with input device

Joe Shaw joeshaw at
Wed Jun 2 11:25:47 PDT 2004


On Wed, 2004-06-02 at 19:10 +0200, Kristian Høgsberg wrote:
> I can see there is no sysfs device for those devices, the closest thing 
> is the keyboard controller /sys/devices/system/i8042, I guess.  So of 
> course HAL can't see those devices.  What's the best way to change that?

The really correct thing would be to add sysdevice support to the kernel
driver, I think.

Failing that you could probably write an accept function for the input
class device handler which doesn't check for sysdevices and instead
checks the name against a pattern (mouse%d, for example).  You'd also
have to implement a compute_udi method.  That way they'd get added to
the City of Lost Devices.  You can look in net_class_device.c to see how
it's done for PCMCIA cards.

Problem with this approach is that you have absolutely no identifying
information about the device.  I guess you could just assume that it's a
PS/2 mouse.

I'd recommend doing "mouse%d" instead of "event%d" since you really have
no idea what "event%d" is.  It could be a keyboard, mouse, or even a
sound device.  Without identifying info, you'd really just be guessing.

> I'm thinking that maybe you could add a mouse and a keyboard device 
> under /sys/devices/system/i8042/i80420, and 
> /sys/class/input/event0/device should link to that, as it's the case for 
> e.g. USB input devices.  Does that sound reasonable?

Yes, this sounds like the ideal solution, but it requires patching the

> Another problem, which may just be an issue with my setup, is that hald 
> only picks up input.* properties for my usb mouse if I start up hald 
> with the mouse plugged in.  If I plug in the mouse when hald is running, 
> it doesn't get the input properties.  Is this a known problem?

It should work, and it WFM with the two USB mice I just hotplugged.  Is
hald getting any hotplug messages?


hal mailing list
hal at

More information about the Hal mailing list