- It is much easier to specify the greybus node address directly for
  development.
- The mdns implementation here is broken as I will need to rewrite and
  get it merged upstream anyway.

Signed-off-by: Ayush Singh <ayush@beagleboard.org>

Add better node disconnection

Closes #6

See merge request !5

- Using TCP Keepalive to detect if node is online

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Can remove and add the node now without much problem.
- Currently relying on the fact that a node is offline if host tries to
  destroy connection with cport 0 of node

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Better to cleanup on node destroy function
- Also simplify svc module removal. This operation should not fail
  unless the node does not exist on kernel. In either case, it needs to
  be cleaned up

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- No point in keeping the node once connection to cport 0 is destroyed.
- Useful when get_manifest or some initial request times out

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- The 2nd array was kind of unnecessary

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Remove unused bits

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Seems better to just follow the spec

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Move from a pull-push model to a push only model.
- Should allow much more idle time to the program

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Incorrect get_manifest_request
- Remove log protocol stuff

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Local node with control port almost done.
- Decided against using greybus log protocol for logging since it can
  become complicated to debug stuff before local node is registered

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Includes defining control port, creating a mainfest, etc
- Allows using greybus protocols for logging, networking, etc

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Make communications with linux kernel greybus spec compiant

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Make things spec compliant on AP side

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Having a wrapper around greybus message is proving more difficult that
  initial assumption

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Much simpler (and less error prone) than using dynamic k_fifo list.
- Everything seems to work

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Just overwrite with the last address

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Use pad bytes to send cport information. This is similar to what is
  done on host side

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Will probably be faster than binary search for n <= 32 due to cache
prediction.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Also switch to normal sdk-next in CI

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Change Control address.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Make HDLC independent from UART

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Add comments
- Run checkpatch

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Running tcp discovery on main thread with highest priority just doesn't
seem like the best idea.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

There reasons are as follows:
1. The removal detection was happening only at an interval which falls
   apart if AP sends message in that delta.
2. Greybus node sometimes fails to respond even if it is online.
3. Caused hit to APBridge performance.

I have also added a cache to allow binary serarching for current nodes.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Using sorted array with binary search for better performance.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Can be calculated from header

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Run checkpatch
- Clang-format

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Was not cleaning up cports properly.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

With this, I am no longer using system heap.

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

No need to allocate HDLC block seperately

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

The final plan is to move away from system heap compleatly

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

- Performed some static analysis

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>

Move greybus messages out of operations

Signed-off-by: Ayush Singh <ayushdevel1325@gmail.com>