[alsa_midi] fix hotplug device (de)enumeration

[ventosus]

Issues:

• With a running JACK with enabled alsa_midi driver (-X alsa_midi), plugging in
  a new MIDI device has no effect, e.g. no corresponding JACK ports are spawned
• With a running JACK with enabled alsa_midi driver (-X alsa_midi), deplugging
  a MIDI device has no effect, e.g. the corresponding JACK ports stay around

Result:

• JACK only creates JACK ports of ALSA MIDI clients/ports found at startup
• JACK has to be restarted for any ALSA MIDI device (de)enumeration to take
  place

Problem:

• There are some functions defined which actually should accomplish this in the
  alsa_midi driver code (e.g. 'a2j_update_ports' and 'a2j_free_ports'), but they
  are not called from any other function ;-)

Solution:

• Discriminate properly between ALSA PORT_START and PORT_CHANGE events
  • 'a2j_new_ports' function has been added which recycles some code from
    'alsa_input_thread'
• Actually call the already existing hot(de)plugging infrastructure
  • 'a2j_update_ports' and 'a2j_new_ports' get called from the
    'alsa_input_thread'
  • 'a2j_free_ports' gets called from 'alsa_output_thread'
  • 'alsa_out_thread' is woken up by 'a2j_jack_process_internal'
• Cleanup code that is not used:
  • 'port_add' ringbuffer has no function, as 'new_ports' ringbuffer seems to be
    implemented to accomplish the same

Signed-off-by: Hanspeter Portner dev@open-music-kontrollers.ch

[ventosus]

Can be tested with snd_virmidi kernel module, if there should not be some USB MIDI device at hands...

sudo modprobe snd_virmidi
sudo rmmod -f snd_virmidi

[rncbc]

i can confirm this patch is working as advertised: MIDI client ports are (de)listed according to resp. MIDI h/w device (un)plugging.

+1 to merge.

[ventosus]

This seems to have been merged via Git natively, I'll thus close it manually here.