AVR UART 读取一个字节两次

Question

I am currently writing a UART receive code that reads and parses commands from another machine using AVR ATtiny87. The idea is to detect the start character and store it in the buffer, and keep storing the UART byte until 0x0a (FL) is received. I have no issue with doing so, but for some reason, my code reads each byte twice. Following is my function, which is called by my kernel loop.

void vehicle_recv(void) {
uint8_t n = 0;
char byte;

byte = LINDAT; //Reads and stores the content of the UART data register. 

if(compass_packet.state == BUFFER_RX_IDLE) {
    if(byte == '*' || byte == '#') {
        compass_packet.buffer[0] = byte;
        compass_packet.index = 1;
        compass_packet.state = BUFFER_RX_IN_PROG;
    }
}
if(compass_packet.state == BUFFER_RX_IN_PROG) {
    compass_packet.buffer[compass_packet.index] = byte;
    (compass_packet.index)++;
    if(byte == 0x0a) {
        compass_packet.buffer[compass_packet.index] = byte;
        (compass_packet.index)++;
        compass_packet.size = compass_packet.index;
        compass_packet.state = BUFFER_RX_DONE;
    }
}
if(compass_packet.state == BUFFER_RX_DONE) {
    decode_vehicle_command(&compass_packet);
    compass_packet.state = BUFFER_RX_IDLE;
}
}


uint8_t decode_vehicle_command(struct compass_comm_packet *RX_buffer) {

debugChar(debug_str);
sendChar(RX_buffer->buffer[0]);
sendCRLF();
sendChar(RX_buffer->buffer[1]);
sendCRLF();
sendChar(RX_buffer->buffer[2]);
sendCRLF();
sendChar(RX_buffer->buffer[3]);
sendCRLF();
sendChar(RX_buffer->buffer[4]);
sendCRLF();
sendChar(RX_buffer->buffer[5]);
sendCRLF();
sendChar(RX_buffer->buffer[6]);
sendCRLF();
sendChar(RX_buffer->buffer[7]);
sendCRLF();

uint8_t return_value = 0;


if(RX_buffer->buffer[0] == '*') {
    switch(RX_buffer->buffer[1]) {

        case 'H':
            strcpy(debug_str, "Heading\r\n");
            debugChar(debug_str);
            break;
        case 'R':
            strcpy(debug_str, "Reset\r\n");
            debugChar(debug_str);
            break;
        case 'S':
            strcpy(debug_str, "Stop\r\n");
            debugChar(debug_str);
            break;
        case 'C':
            strcpy(debug_str, "Calibrate\r\n");
            debugChar(debug_str);
            break;
    }
}

When I send *H(CR)(FL), I expect the decode_vehicle_command() function to spit out *H(CR)(FL). However, I keep seeing **HH(CR)(CR)(FL)(FL). I can work my way around this issue pretty simply by using RX_buffer->buffer[2] instead of RX_buffer->buffer[1], but I am curious what exactly I am doing wrong here.

Answer 1

Looking at the function vehicle_recv() , you have:

if(compass_packet.state == BUFFER_RX_IDLE) {
    if(byte == '*' || byte == '#') {
        compass_packet.buffer[0] = byte;
        compass_packet.index = 1;
        compass_packet.state = BUFFER_RX_IN_PROG;
    }
}
if(compass_packet.state == BUFFER_RX_IN_PROG) {
    compass_packet.buffer[compass_packet.index] = byte;
    (compass_packet.index)++;
    if(byte == 0x0a) {
        compass_packet.buffer[compass_packet.index] = byte;
        (compass_packet.index)++;
        compass_packet.size = compass_packet.index;
        compass_packet.state = BUFFER_RX_DONE;
    }
}
if(compass_packet.state == BUFFER_RX_DONE) {
    decode_vehicle_command(&compass_packet);
    compass_packet.state = BUFFER_RX_IDLE;
}

After you set the compass_packet.state = BUFFER_RX_IN_PROG inside the first conditional, you enter the second one as you've just set the state to do so. Inside the second conditional you are again saving byte to the buffer---the same byte, per

if(compass_packet.state == BUFFER_RX_IN_PROG) {
    compass_packet.buffer[compass_packet.index] = byte;
    ...

Generally, I use a switch (or if ... else ) for those conditionals; it may be your need too:

switch(compass_packet.state) {
    case BUFFER_RX_IDLE: 
        if(byte == '*' || byte == '#') {
            compass_packet.buffer[0] = byte;
            compass_packet.index = 1;
            compass_packet.state = BUFFER_RX_IN_PROG;
        }
        break;
    case BUFFER_RX_IN_PROG:
        compass_packet.buffer[compass_packet.index] = byte;
        (compass_packet.index)++;
        if(byte == 0x0a) {
            compass_packet.buffer[compass_packet.index] = byte;
            (compass_packet.index)++;
            compass_packet.size = compass_packet.index;
            compass_packet.state = BUFFER_RX_DONE;
        }
        break;
    case BUFFER_RX_DONE:
        decode_vehicle_command(&compass_packet);
        compass_packet.state = BUFFER_RX_IDLE;
        break;
    default:
        /* WTF? */
}

Answer 2

Part of the problem is that you set a state and then act on it in the same pass. Change your code to either exit the routine once you handle a particular character, or change the second and third if statements to else if .

Step through it with a debugger to see what I mean.

Set the state to idle and the received character to '*' and you'll get the following sequence:

if(compass_packet.state == BUFFER_RX_IDLE) {
    // TRUE
    if(byte == '*' || byte == '#') {
        // TRUE
        compass_packet.buffer[0] = byte;
        compass_packet.index = 1;
        compass_packet.state = BUFFER_RX_IN_PROG;
    }
}

At this point the leading character is stored in the buffer, the index is 1 and your state is "RX in progress".

if(compass_packet.state == BUFFER_RX_IN_PROG) {
    // TRUE because you just set it in the previous block
    compass_packet.buffer[compass_packet.index] = byte;
    // here you've now stored the leading character '*' again
    (compass_packet.index)++;

    if(byte == 0x0a) {
        compass_packet.buffer[compass_packet.index] = byte;
        (compass_packet.index)++;
        compass_packet.size = compass_packet.index;
        compass_packet.state = BUFFER_RX_DONE;
    }
}

You have a similar problem for the line feed terminating character.

Try:

if(compass_packet.state == BUFFER_RX_IDLE) {
    if(byte == '*' || byte == '#') {
        compass_packet.buffer[0] = byte;
        compass_packet.index = 1;
        compass_packet.state = BUFFER_RX_IN_PROG;
    }

} else if(compass_packet.state == BUFFER_RX_IN_PROG) {
    compass_packet.buffer[compass_packet.index] = byte;
    (compass_packet.index)++;

    if( byte == 0x0a) {
        compass_packet.size = compass_packet.index;
        compass_packet.state = BUFFER_RX_DONE;

        decode_vehicle_command(&compass_packet);
        compass_packet.state = BUFFER_RX_IDLE;
    }
}

You have other holes in your state machine, btw. What happens if you get too many characters before the end-of-line mark?