Dynamic devices and static configuration [LWN.net]

Dynamic devices and static configuration

Posted Apr 2, 2011 13:23 UTC (Sat) by jimparis (subscriber, #38647)
Parent article: Dynamic devices and static configuration

What are the alternatives? One of those is to do everything in user space, using udev rules. This approach appeals to those who want to see no policy in kernel space, but it is hard to implement in this case; there is no information available to distinguish this wired-down network controller from the traditional USB variety.

That's not true. USB devices can be identified uniquely by physical path -- see "lsusb -t". For for a wired down device, you can figure out which one it is as long as you always solder it to the same spot. This discussion seems to be focusing not on that particular problem, but instead on just extending platform_data to other usb devices for more nuanced reasons, since there are some hardware-specific details that might come up. See e.g. this posting for some examples.

(Log in to post comments)

Dynamic devices and static configuration

Posted Apr 7, 2011 3:32 UTC (Thu) by npitre (subscriber, #5680) [Link]

> That's not true. USB devices can be identified uniquely by physical
> path -- see "lsusb -t". For a wired down device, you can figure out
> which one it is as long as you always solder it to the same spot.

That's the problem. It is not always soldered at the same spot,
if at all, from one board to another similar board. Some parts of
the kernel have no choice but to know differences between such similar
boards already (such as what GPIO is connected to what LED, or what
key, or what IRQ signal, etc). Duplicating part of that knowledge in
user space would increase maintenance costs, and there is no good way
to identify such boards from user space in the first place.