prj.conf

# Board-neutral AES67 app config. Anything board-specific lives in
# boards/<board>.conf (automatically merged by Zephyr when that board
# is selected).

# ---- Shell & logging ----
CONFIG_SHELL=y
CONFIG_LOG=y
CONFIG_EARLY_CONSOLE=y
CONFIG_PRINTK=y

# ---- Threads / stacks ----
CONFIG_MAIN_STACK_SIZE=4096
# The network stack runs ACD/ARP probing, IGMP and TCP timers here, and the
# eth_litex link supervision drives whole carrier transitions from it. Deep,
# and an overflow on Xtensa is a silent lockup — keep real headroom.
CONFIG_SYSTEM_WORKQUEUE_STACK_SIZE=8192
CONFIG_THREAD_MONITOR=y
CONFIG_THREAD_NAME=y

# main.c reads cb->info in the DHCP-bound event handler. Was previously
# selected implicitly by NET_CONNECTION_MANAGER (dropped — unused).
CONFIG_NET_MGMT_EVENT_INFO=y

# ---- IPv4 link-local / Zeroconf (RFC 3927) ----
# Deliberately NOT CONFIG_NET_IPV4_AUTO: it force-selects CONFIG_NET_IPV4_ACD,
# and ACD has a lock inversion (ipv4_acd_timeout() holds the ACD mutex while
# calling into net_if, net_if_ipv4_addr_rm() holds net_if_lock while calling
# net_ipv4_acd_cancel()). Any address removal racing an ACD timer tick — a
# link-down carrier teardown, a net_dhcpv4_stop() from another thread —
# deadlocks both threads silently: no panic, no log, the node just stops.
# main.c configures the 169.254.x.y address itself instead (MAC-derived, so
# it is stable across reboots) and Zephyr patch 0013 makes the stack use it
# as a source address. The boards raise NET_IF_MAX_IPV4_COUNT to 2 so the
# fallback address and a DHCP lease can coexist.

# Zephyr JSON library (config_json.c + webapi request bodies)
CONFIG_JSON_LIBRARY=y