> usbdevs(8) uses atoi(3) to convert a device address given from a > command line argument. > But atoi(3) will return - zero in a case of non-integer value and such > device will be used. > This is kind of unpredictable behaviour to use a device by 0 addr in a > case when wrong > address was given to usbdevs.
For what it's worth. Zero is never a legal address for an operating USB device. (It can appear transiently during enumeration, but there's no way to operate a device in that state; so it make no sense for address zero to be used in usermode). A device at address zero is always under the administration of the hub driver and is about to be moved to a non-zero address. Valid USB device addresses are always in the range [1..127]. So (independent of the advisability of using atoi()), it's completely reasonable to use 0 as an illegal USB device address, and reject any attempt to specify that. If looking into `usbdevs`, it would make sense to check this kind of thing. The man page for usbdevs is not clear on what is meant by "address"; if there are multiple host controllers, there may be many devices with the same numerical address, each on a different host controller/bus. Generally an address should include a host controller (or bus ID) in order to be useful. I'm not sure what `usbdevs` is trying to do. For better or worse, `lsusb` on a typical Linux userland does take all this into account, and is the way most people think of this these days; if selecting device by address you also include the bus. Best regards, --Terry
