Hi,
my concerto application is almost running nicely but the lwIP is still giving me some headache, and I reckon it's a little ordering on the calls that is missing.
There's no web-server or file server. The "protocol" I'm using is based on some very simple commands using cmdline.c util and the MCU will receve commands such as READ_EEPROM, READ_PROCESS or CMD 1, CMD 2 and I know those works cause it works through the UART.
on the lwipopts.h I tried disabling a lot of unecessary stuff and open the memory to 16K and got S4 to S7 to .bss to be able to allocate all that. With that I hope to be able to transfer all the data from my responseBuffer directly to the lwIP in just one go and not need to transfering in a later call.
I'm checking the amount of memory available on the lwIP as u16_t buff_size = tcp_sndbuf(pcb); I'm not sure if that is correct.
I've seen a bunch of calls to pbuf_free(p); on the examples and I'm not sure if I was to be doing anything about it, or it was just for the files it was sending with headers, etc, etc.
At the moment it boots fine and I can manually send some commands and it replies the way it should, but then when my desktop application connects to it and starts it's sequence of Connect-> Send Command -> Receive Response -> Disconnect every 500ms... the thing gets a bit crazy and some connetion start polling indefinitely, even if I call tcp_close(pcb), it just keeps polling and stops responding
so I have the logic separated in the usual callbacks, below is the simplified code, it shows all the calls to the lwIP API but to keep it short, it hides some pointers creation, NULL/error checking
Initialisation:
void ethHandler_init(void) {
struct tcp_pcb *pcb;
pcb = tcp_new();
tcp_bind(pcb, IP_ADDR_ANY, 83);
pcb = tcp_listen(pcb);
tcp_accept(pcb, ethHandler_accept);
}
Accept New Connection:
static err_t ethHandler_accept(void *arg, struct tcp_pcb *pcb, err_t err) {
// allocate a struct ethHandler_conn_state and call those as per examples
tcp_arg(pcb, conn);
tcp_recv(pcb, ethHandler_recv);
tcp_err(pcb, ethHandler_conn_err);
tcp_poll(pcb, ethHandler_poll, 4);
tcp_sent(pcb, ethHandler_sent);
}
On receive data:
static err_t ethHandler_recv(void *arg, struct tcp_pcb *pcb, struct pbuf *p, err_t err) {
// Tells lwIP that we read this info and process it
tcp_recved(pcb, p->tot_len);
data = p->payload;
int response = cmdline_parse(data, responseBuffer);
if (response > 0) {
// do a nice scheme to post as much data as possible, but not too much and calls
tcp_write(pcb, ptr->message, buff_size, TCP_WRITE_FLAG_COPY);
tcp_output(pcb);
}else{
tcp_write(pcb, "Bad Command", sizeof("Bad Command"), TCP_WRITE_FLAG_COPY);
tcp_output(pcb);
}
On Sent ok:
static err_t ethHandler_sent(void *arg, struct tcp_pcb *pcb, u16_t len) {
// Do a little checking if there's more data to sent and sent more tcp_write() and tcp_output() if there is or
tcp_close(pcb); // if there's nothing more to send
}
On Poll: (to be honest I'm not sure when or how this poll is supposed to happen)
static err_t ethHandler_poll(void *arg, struct tcp_pcb *pcb) {
// So I'm repeating here kind of the same thing from Sent, if more to send tcp_write() and tcp_output() if there's nothing more to send
if (conn->poll_count++ > 5) {
UARTprintf(UART_DEBUG, " and closing");
tcp_close(pcb);
}
}
please help!!
edit:
I just double checked and whenever the MCU starts polling for ever, after the 5 polls and my code calls tcp_close(pcb); my desktop application gives me a timeout exception, trying to reconnect to the MCU.
but I still don't know what is causing this pooling, why is causing it, and if I should just ignore instead of cancelling.
