On Wed, 5 Apr 2000, Alessandro Rubini wrote:
> I'm currently studying the USB kernel subsystem using modules, and I
> noticed how the usage count of specific device never raises above
> zero. Should that be construed as a feature?
>
> I'd personally prefer to have the driver marked as busy when there is
> a device using it. For example, a devault crontab in Debian removes
..
> The best option would be to increment the count when the device is
> actually being read, but the current layout of the USB and input
> subsystems make it impractical to do so, and I wouldn't like such
> feature to be implemented as things are clean enough as they are now.
My USB modules have varying semantics:
Network drivers only have a non-zero usage count when they are up. They
shut down the interface automatically on hardware disconnect.
Character/block drivers only have a non-zero usage count when they are
open. Removing the device does not (can not) automatically close, so the
module is not removed until the last opener finishes.
Mouse and keyboard device drivers have a non-zero usage count whenever
they are plugged in. Any explicit mouse open further increments the usage
count. Keyboards just inject events, and thus they never know if anyone
is listening.
I believe this is close to the semantics of most of the pre-2.4 drivers, but
I'm not certain that it's consistent or documented.
Donald Becker
Scyld Computing Corporation, [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
