Get run-time stack usage by debug trace

Please refer to Use syslog to trace Neo3 for time performance measurement on K81 to add trace before using the functions.

Add pthread_get_stacksize_np and calculate remain stack size in __cyg_profile_func_exit function like below.

void __attribute__((__no_instrument_function__))
__cyg_profile_func_enter(void *this_func, void *call_site)
{
  syslog(LOG_CRIT, "[Start].%p\n", this_func);
}

void __attribute__((__no_instrument_function__))
__cyg_profile_func_exit(void *this_func, void *call_site)
{
  pthread_t currentThread = pthread_self();
  /* Get run-time stack address and size */
  uint8_t *stack_base = pthread_get_stackaddr_np(currentThread);
  size_t stack_size = pthread_get_stacksize_np(currentThread);
  uint8_t *frame_addr = __builtin_frame_address(0);
  size_t remainingStackSize = stack_size - (size_t)(frame_addr - stack_base);
  syslog(LOG_CRIT, "[End:%d].%p\n", remainingStackSize, this_func);
}

After running for a while eg. a ctls transaction, a syslog file should be saved.

Copy the syslog file to Linux PC. Use the script file gethreadstackusage.py to calculate stack usage for each thread like below.

$ python3 gethreadstackusage.py syslog
Thread/TC     Stack Free (bytes)
-------------------------
TC_K81_START  2744
pc_k81start_primary_ss  896
CTLS_transaction_thread  336
pt-CTLS_transaction_thread_level1  200
TC_K81_READER  184
TC_K81_SYSTEM  184
IDG_Server_Request_Handler  168
pc_CTLS_txn_thread_level1  168
pc_k81reader_reader_ss  152
pc_k81system_system_ss  152
pc_k81start_native_ss  144
pc_k81system_pki_stm  144
pt-k81system_system_ui_ss  136
pt-k81system_pki_stm_loop  136
pc_TAL_main_thread  136
k81start  128
pt-k81start_primary_ss  128
k81reader  120
k81system  120

The second column shows the remaind stack size after a ctls transaction 0240.

From the list, thread TC_K81_START remains 2744bytes stack size. So we can say the stack usage of thread TC_K81_START is very little during a ctls transaction.

A syslog trace file for a ctls transaction.