DDF should support pseudo-devices

[Jiri Svoboda]

Currently any devman-based device must come into existence by being enumerated by some nexus (bus) device. This is inconvenient for pseudo-devices such as {{file_bd}}.

The distinction here is that for these pseudo-devices, it would be better if the device lifecycle was not controlled by DDF, but rather by the pseudo-device driver itself.

[Jiri Svoboda]

I propose for a pseudo-driver to have the following properties:

• It does not live under the drv directory, it is not started by devman. Pseudo-driver lifecycle is out of scope of the DDF
• It is any server which chooses to connect to devman and register as a driver
• It can create one or more pseudo-devices
• The bottom half of the pseudo-devices is out of scope of the DDF, while the top half (functions) is in scope

I plan looking into this when I am done with ticket #295

[Jiri Svoboda]

I consider this ticket as solved since introduction of the Location service in mainline,1144. Thanks to that (and also since drivers can now handle client connections by themselves), DDF drivers and other servers are indistinguishable for a client.

Thus a pseudo-driver now corresponds to a regular server. The only exception would be a pseudo-nexus driver (such as USB host controller connected over a network). Since we currently don't intend to implement such drivers, we can leave this case for now.