Schedule slave FSMs with config_requested/config_running in pick-up loop

[Copilot]

Without this fix, ec_fsm_master_state_wait_config spins forever because slaves flagged for parallel configuration are never scheduled — ec_fsm_slave_is_ready() only returns true for request-based operations, not for configuration flags set by ec_fsm_slave_request_config().

Change

master/master.c — ec_master_exec_slave_fsms() pick-up loop

• Expand the pick-up condition to include slaves with config_requested or config_running set, alongside the existing ec_fsm_slave_is_ready() check:

if (ec_fsm_slave_is_ready(&master->fsm_slave->fsm)
        || master->fsm_slave->fsm.config_requested
        || master->fsm_slave->fsm.config_running) {

This is Step 4 of the parallel slave configuration implementation — ec_fsm_slave_exec() already handles these flags correctly; they just needed to be included in the scheduler's pick-up condition.

Original prompt

Problem

This is the final step (Step 4) of the parallel slave configuration implementation described in PARALLEL_SLAVE_CONFIG_IMPL.md.

Steps 1-3 are already complete on the parallel-slave-config branch:

• Step 1: fsm_slave_config already uses external datagram pointers
• Step 2: Configuration FSM embedded in fsm_slave (PR Implement parallel slave configuration (Steps 2–4) #154, bugfix PR Fix NULL datagram segfault in parallel slave config FSM #155)
• Step 3: fsm_master delegates configuration in parallel via ec_fsm_slave_request_config() and ec_fsm_master_state_wait_config

However, Step 4 is missing: The slave FSM pick-up loop in ec_master_exec_slave_fsms() in master/master.c never adds slaves with config_requested or config_running flags to the execution list. This means parallel configuration will hang — ec_fsm_master_state_wait_config will spin forever because the slave FSMs are never scheduled.

Required Change

In master/master.c, in the ec_master_exec_slave_fsms() function, there is a pick-up loop (around line 1336-1358) that scans slaves and adds their FSMs to master->fsm_exec_list:

while (master->fsm_exec_count < EC_EXT_RING_SIZE / 2
        && count < master->slave_count) {

    if (ec_fsm_slave_is_ready(&master->fsm_slave->fsm)) {
        // ... pick up and execute ...
    }

    master->fsm_slave++;
    // ...
    count++;
}

The condition ec_fsm_slave_is_ready() only returns true when the slave FSM state is ec_fsm_slave_state_ready (set via ec_fsm_slave_set_ready()). It does not detect slaves whose config_requested or config_running flags have been set by ec_fsm_slave_request_config().

The fix: Expand the pick-up condition to also check the configuration flags:

if (ec_fsm_slave_is_ready(&master->fsm_slave->fsm)
        || master->fsm_slave->fsm.config_requested
        || master->fsm_slave->fsm.config_running) {

This ensures that when fsm_master sets config_requested=1 on slave FSMs (Step 3), those slaves will be picked up by ec_master_exec_slave_fsms(), given an external datagram, and executed in parallel.

File to modify

• master/master.c — the slave FSM pick-up loop inside ec_master_exec_slave_fsms()

Important context

• The ec_fsm_slave_exec() function (in master/fsm_slave.c) already handles config_requested and config_running correctly — it starts the config FSM, propagates datagrams to sub-FSMs, and clears the flags when done.
• The existing ec_fsm_slave_is_ready() check handles normal request-based slave operations (SDO, FoE, etc.)
• The new condition adds parallel configuration scheduling alongside the existing request scheduling.
• No other files need modification for this step.

This pull request was created from Copilot chat.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.