modem: modem chat: alignment((MPU faults and instruction access violations)) issues in modem chat script

[zafersn]

There were crashes (MPU faults and instruction access violations) when the chat script size increased. Debugging showed that the script’s memory alignment was causing issues, leading to invalid instruction execution.

[00:00:09.736,000] <err> os: ***** USAGE FAULT *****
[00:00:09.736,000] <err> os:   Illegal use of the EPSR
[00:00:09.736,000] <err> os: r0/a1:  0x0802ff8c  r1/a2:  0x08033130  r2/a3:  0x00000008
[00:00:09.736,000] <err> os: r3/a4:  0x73253c00 r12/ip:  0xaaaaaaaa r14/lr:  0x0802e051
[00:00:09.736,000] <err> os:  xpsr:  0x2000001e
[00:00:09.736,000] <err> os: Faulting instruction address (r15/pc): 0x73253c00
[00:00:09.736,000] <err> os: >>> ZEPHYR FATAL ERROR 35: Unknown error on CPU 0
[00:00:09.736,000] <err> os: Fault during interrupt handling

[00:00:09.736,000] <err> os: Current thread: 0x20013a20 (idle)
[00:00:09.798,000] <err> os: Halting system

here is the script definations:

MODEM_CHAT_SCRIPT_CMDS_DEFINE(
	swir_hl7812_init_chat_script_cmds, 
	MODEM_CHAT_SCRIPT_CMD_RESP("", at_ready_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+KSREP?", ksrep_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("ATE0", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP_MULT("AT+CGACT=0", allow_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CFUN=4", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CMEE=1", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CREG=1", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CEREG=1", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CREG?", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CEREG?", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CGSN", imei_match), 
	MODEM_CHAT_SCRIPT_CMD_RESP("",ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CGMM", cgmm_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("", ok_match), 
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CGMI",cgmi_match), 
	MODEM_CHAT_SCRIPT_CMD_RESP("", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CGMR", cgmr_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("",ok_match), 
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CIMI", cimi_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("", ok_match), 
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+KSLEEP?",ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CPSMS?", ok_match),
	MODEM_CHAT_SCRIPT_CMD_RESP("AT+CEDRXS?", ok_match),
);

MODEM_CHAT_SCRIPT_DEFINE(swir_hl7812_init_chat_script, swir_hl7812_init_chat_script_cmds,
			 abort_matches, modem_cellular_chat_callback_handler, 10);

ret = modem_chat_run_script(&data->chat, &swir_hl7812_init_chat_script);

Here are my sample stack sizes:

CONFIG_HEAP_MEM_POOL_SIZE=95250
CONFIG_MAIN_STACK_SIZE=24096
CONFIG_SYSTEM_WORKQUEUE_STACK_SIZE=14096
CONFIG_IDLE_STACK_SIZE=14024
CONFIG_ISR_STACK_SIZE=14024

As a solution:

I added __aligned(8) to the MODEM_CHAT_SCRIPT_CMDS_DEFINE and MODEM_CHAT_SCRIPT_DEFINE macros. This ensures proper memory alignment and prevents crashes, even when the script size changes.

Testing:
Without alignment: Crashes when the script grows beyond a certain size.

With __aligned(8): Works fine, even with a larger script.

Updated the macros like this:

__aligned(8) MODEM_CHAT_SCRIPT_CMDS_DEFINE(...)
__aligned(8) MODEM_CHAT_SCRIPT_DEFINE(...)

Can anyone shed some light on this issue? Am I doing something wrong/ missing some configs here?