On Thu, 23 Sep 2004, Oliver Neukum wrote:

> That seems reasonable, but how specific to USB is it? I suggest that
> you use bluetooth as the worst case for this problem.

I don't think it's very specific to USB, but it's hard for me to tell.  
And I don't know enough about Bluetooth to say anything useful.  Aren't BT
devices treated pretty much like network interfaces?

> Wouldn't a rwsem per bus not be an adequate, generic solution?

The driver model doesn't include the notion of a bus, only of a bus_type.  
So for example, each USB bus doesn't appear as a separate entity in the 
driver model.  Hence there's no way to use a per-bus anything without 
changing the driver-model core.

If such a change were made, then per-bus semaphores might work okay.  
Wouldn't it be true that whenever a device and one of its ancestors belong
to the same bus, then all the intervening devices would belong to it also?

Things would be a little more complicated, because when going through the 
device tree you would always have to check whether you had crossed a bus 
boundary.  It wouldn't be too bad, though.  Events on each bus would be 
serialized, but events on different buses could run in parallel.

Alan Stern



-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to