Trouble reading from serial port.

rkjnsn · 12-01-2009, 12:50 PM

I am trying to write a very simple test program that communicates with a serial device. What I want to accomplish is very simple:
1. Wait for a keypress
2. Send a 4 byte command to the device
3. Wait for it to send a byte back
4. Go to step 1

The program successfully sends the command to the device (the device carries out the action), but it never receives the reply byte back, instead blocking on the read forever. I have verified that the device does indeed send the response byte by sending it commands using gtkterm, and the response character is displayed every time.

Frankly, I am quite stumped, as I have looked at many examples and tried several different things. My current code is as follows:

Code:

#include <termios.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <stdio.h>

#define BAUDRATE B9600
#define DEVICE "/dev/ttyUSB0"

main ()
{
	int fd;
	struct termios oldtio, newtio;

	fd = open(DEVICE, O_RDWR | O_NOCTTY);

	tcgetattr(fd, &oldtio);

	memcpy(&newtio, &oldtio, sizeof(struct termios));
	newtio.c_iflag &= ~(IGNBRK | BRKINT | PARMRK | ISTRIP
		| INLCR | IGNCR | ICRNL | IXON);
	newtio.c_oflag &= ~OPOST;
	newtio.c_lflag &= ~(ECHO | ECHONL | ICANON | ISIG | IEXTEN);
	newtio.c_cflag &= ~(CSIZE | PARENB);
	newtio.c_cflag |= CS8;
	newtio.c_cc[VMIN]=1;
	newtio.c_cc[VTIME]=0;
	cfsetispeed(&newtio, B9600);
	cfsetospeed(&newtio, B9600);
	tcsetattr(fd, TCSAFLUSH, &newtio);
	
	while(1)
	{
		char response;
		if (getchar() == 'q')
			break;
		write(fd, "\xA5" "\x5A" "\x0A" "\xF5", 4);
		read(fd, &response, 1)
	}

	tcsetattr(fd, TCSANOW, &oldtio);
	close(fd);

	return 0;
}

P.S.
The settings that work with gtkterm are:
Speed: 9600
Parity: none
Bits: 8
Stopbits: 1
Flow control: none

Aquarius_Girl · 12-02-2009, 06:07 AM

Code:

#define DEVICE "/dev/ttyUSB0"

shouldn't it be something like "/dev/ttyS0" with respect to serial port ?

michaelk · 12-02-2009, 07:44 AM

/dev/ttyUSBx devices are valid and assigned to USB serial port adapters. You should use non-canonical input processing i.e newtio.c_lflag = 0 or asynchronous with a signal handler. Canonical input waits until a line terminating character is received.

http://tldp.org/HOWTO/Serial-Program...OWTO/x115.html

echo36 · 12-02-2009, 06:00 PM

Code:

newtio.c_lflag &= ~(ECHO | ECHONL | ICANON | ISIG | IEXTEN);

It looks like OP is using non-canonical processing.

michaelk · 12-02-2009, 06:18 PM

Your correct I misread that line.... I do see a mistake.

Quote:

read(fd, &response, 1)

No need for the &.

theNbomr · 12-03-2009, 10:20 AM

Quote:

Originally Posted by michaelk

Your correct I misread that line.... I do see a mistake.

No need for the &.

That code looks fine to me. The response buffer is a char scalar, 1 byte deep. read() requires a pointer to it. Ordinarily, one would define the buffer as an array, but here, that is not the case.

With respect to the original problem, I think the cooked mode being used is the problem. Non-canonical processing is probably the solution.
--- rod.

michaelk · 12-03-2009, 10:26 AM

Quote:

Originally Posted by theNbomr

That code looks fine to me. The response buffer is a char scalar, 1 byte deep. read() requires a pointer to it. Ordinarily, one would define the buffer as an array, but here, that is not the case.

Oops.. Your correct... I stand corrected again.... I shall clean my glasses and look again.

michaelk · 12-03-2009, 06:44 PM

Not sure what the problem is anymore other then you might not be sending the correct characters and the getchar() requires the enter key to be pressed i.e if you are not actually using the enter key as your keypress.

rkjnsn · 12-04-2009, 09:30 AM

Quote:

Originally Posted by michaelk

Not sure what the problem is anymore other then you might not be sending the correct characters and the getchar() requires the enter key to be pressed i.e if you are not actually using the enter key as your keypress.

Well, I am indeed using the enter key as my keypress, and I have verified that it is indeed hanging on the read.

Thank you everyone for all of the replies so far.

theNbomr · 12-04-2009, 09:48 AM

How do you know that the response is not received? Your code does nothing with any received characters.
Sorry for having misread the code with respect to the use of non-canonical processing.
--- rod.

rkjnsn · 12-04-2009, 09:53 AM

Quote:

Originally Posted by theNbomr

How do you know that the response is not received? Your code does nothing with any received characters.
Sorry for having misread to code with respect to the use of non-canonical processing.
--- rod.

Well, the main purpose is to just wait until the byte is received, and I don't really care what it is (it's actually always 'R'). But I have indeed tried putting a printf after the read statement, and it never got executed. The key issue is that my program hangs on the read command forever, even though I know that a byte has been sent back.

theNbomr · 12-04-2009, 10:58 AM

Have you tested the return codes from open(), tcsetattr(), write(), and read()? Permissions on the device? If you set VTIME to a non-zero value, does the read() call get any data?
--- rod.

michaelk · 12-04-2009, 11:11 AM

I've tested your code using two laptops with the device being simulated via hyperterminal. Obviously the laptop does not care what characters are being sent to it and in fact I changed them to regular ASCII text. I also added some additional debug print statements but basically the program works for me.

rkjnsn · 12-04-2009, 11:57 PM

Okay, so some more weirdness. Just for grins I tried rewriting the program in python, and it works. I am at a loss as to why it doesn't work in C (since it does work for you).

Code:

import sys
import serial

ser = serial.Serial('/dev/ttyUSB0', 9600)

while 1:
	char = sys.stdin.read(1)
	if char == 'q':
		break;
	ser.write("\xA5\x5A\x0A\xF5")
	ser.read(1)

ser.close()