I'm investigating dynamic device name mapping via udev.
The goal:
On a multi user system, each user can allocate and free devices like usb-serial converters or PCIe cards (FPGAs). If an user allocates a device, the /dev/<user>/<device> nodes shall be created with a dynamic name and permissions for the allocating user. If the user releases the device the devfs nodes need to be removed/renamed and the permissions need to be revoked.
Current situation:
I'm experimenting on /dev/ttyUSB1 – an usb-serial converter. I wrote a Python script which is called from udev while creating the dev-node. It expects the VendorID, ProductID and SerialNo of a usb device and returns the devfs node name: e.g. paebbels/ttyUSB. This works with symlinks, so the original devfs node ttyUSB1 is still present.
How can I use my script to change the original devfs name?
I tried to use the NAME command in my udev rule, but it does not work. The documentation states: NAME can only be used once. Is there a default rule in udev which uses NAME before my rule?
/etc/udev/... does not contain such a rule.
In a latter phase, my script will also return the username, group and permissions for the devfs node.
Secondary question:
Should I create the device files as planned in /dev/<user>/<device> or in ~/fpga1/<device>?
Best Answer
1. Question:
I am not sure, but take a look at
/lib/udev/rules.d/60-persistent-serial.rules2. Question:
If udev itself assigns the device names I would put them all in a subdirectory of
/dev. If not in/run/$UID/devices.