antlab-polimi / flowblaze.p4 Goto Github PK

uppercase?
pre-pending a special character?

Configure TravisCI to build and push Docker containers when PR/commits are submitted

Rather then not returning the file to download or returning it with partial transitions we should inform back the user.

Regardless the client-side JS checks something might go wrong in Python

For example:

• program name
• FLOW_SCOPE
• the meta field that can be used in conditions
• EFSM_MATCH_FIELDS

Run use cases in dmoro92/p4mn docker image with runtime configuration coming from the GUI

e.g. forward(5) and _drop()
the action parser in interpret_EFSM() has to be checked

the JS should already offer these two functions!

Right now, the state ID generated by the parser is the order number in the JSON list of states. This means that state 0 might not be translated to the 0 state by the parser. This is a requirement for flowblaze.p4.
Examples:
Wrong:

{
  "x": 473,
  "y": 209,
  "text": "1", <-- STATE 1 is first in the list and it's translated to STATE 0
  "isAcceptState": false
},
{
  "x": 679,
  "y": 247,
  "text": "2",
  "isAcceptState": false
},
{
  "x": 120,
  "y": 175,
  "text": "0", <-- STATE 0 is the last in the list and it's translated to STATE 2
  "isAcceptState": false
}

Correct:

{
  "x": 120,
  "y": 175,
  "text": "0",
  "isAcceptState": false
},
{
  "x": 473,
  "y": 209,
  "text": "1",
  "isAcceptState": false
},
{
  "x": 679,
  "y": 247,
  "text": "2",
  "isAcceptState": false
}

Parse bmv2.json file to retrieve header fields, actions and META to fill-up the GUI and improve GUI usability

Probably the best would using the P4 syntax (i.e. value&&mask)

number of conditions
number of flow/global registers
number of update operations
number of actions (<=1)

FlowBlaze metadata is local to the FlowBlaze loop. There is no need to define them into the global metadata and can be kept local to the FlowBlaze loop (e.g., by defining a flowblaze_t flowblaze_metadata; at the beginning of the FlowBlaze loop). This will remove the need for adding the flowblaze_t into the user-defined metadata.

We should add the possibility of generating the switch config directly starting from the FSM JSON file without opening the GUI.

• Add copyright headers

• Add License file

• General cleanup

• Address issues labeled as Documentation

• #51

• #14

e.g. an update action whose result is @now or @meta
shall we just skip the reg_action or the whole transition?

e.g. a transition with too many update actions or more than 1 pkt action
shall we just skip the reg_action or the whole transition?

-POSSIBLE_PACKET_ACTION in gui/config.py should be set by p4_json_parser.py

-shall the user explicitly use NoAction()?