Bus fault on http_client_req in ncs 3.1.0

[meraki]

Getting a Precise data bus error - BUS FAULT, when sending a http POST request to an https sever over tls. Cannot decipher the exact instruction address as on enabling CONFIG_DEBUG_OPTIMIZATIONS+Y, my RAM overflows. Tried to trace the flow and see where the error might be getting triggered using breakpoints and my speculation in the sendall api call in http_client.c, http_client_req -> http_send_data->sendall().
When the api declares a struct for file decriptor,

static int sendall(int sock, const void *buf, size_t len,

const k_timepoint_t req_end_timepoint)

{

while (len) {

ssize_t out_len = zsock_send(sock, buf, len, 0);

if ((out_len == 0) || (out_len < 0 && errno == EAGAIN)) {

struct zsock_pollfd pfd;

int pollres;

k_ticks_t req_timeout_ticks =

sys_timepoint_timeout(req_end_timepoint).ticks;

int req_timeout_ms = k_ticks_to_ms_floor32(req_timeout_ticks);

This bus fault is getting triggered in ncs 3.1.0 but noy in ncs 2.9

[Turbo J]

Looks to me like a stack buffer overflow. Data size 9500 bytes is kind of a lot, especially if you have to encrypt it...

[meraki]

increased the stack and checked, doesn't seem like a stack issue, and the same file is successfully sent in ncs 2.9

[Håkon Alseth]

Hi,

 

Set CONFIG_THREAD_NAME=y to get a verbose information on which thread has faulted, and use addr2line as outlined in ncs intermediate course to find where the fault occurred:

https://academy.nordicsemi.com/courses/nrf-connect-sdk-intermediate/lessons/lesson-2-debugging/topic/exercise-2-11/

 

You're sending almost 10kB of data, so you will need to adjust the net buffers accordingly (see symbols CONFIG_NET_PKT_BUF_TX_DATA_POOL_SIZE and CONFIG_NET_PKT_BUF_RX_DATA_POOL_SIZE), as well as the mbedtls heap size (if using TLS).

 

Kind regards,

Håkon