[Xenomai-help] adding PCI support to 16550A.c

[Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

I've tried to make the pci serial controller visable to the 16550
driver by adding:
## FROM 16550A.c
static int rt_16550_interrupt(rtdm_irq_t * irq_context)
{
	struct rt_16550_context *ctx;
	unsigned long base;
	int mode;
	int iir;
	uint64_t timestamp = rtdm_clock_read();
	int rbytes = 0;
	int events = 0;
	int modem;
	int ret = RTDM_IRQ_NONE;
// TEST
struct pci_dev *pdev = NULL;  // from
http://bachman.tor.lindesign.se/tmp/driver/sunix/snx_golden.c
pdev = pci_get_device(0, 0, pdev); // from
http://bachman.tor.lindesign.se/tmp/driver/sunix/snx_golden.c
//
	ctx = rtdm_irq_get_arg(irq_context, struct rt_16550_context);
	base = ctx->base_addr;
	mode = rt_16550_io_mode_from_ctx(ctx);

	rtdm_lock_get(&ctx->lock);

	while (1) {
		iir = rt_16550_reg_in(mode, base, IIR) & IIR_MASK;
		if (testbits(iir, IIR_PIRQ))
			break;


Loading module:
insmod ./xeno_16550A.ko io=0x3f8,0xbc00 irq=4,3 tx_fifo=10,20 start_index=0

running crosslink test:
sandbox:/usr/src/xenomai/examples/rtdm/profiles/serial# ./cross-link
main : write-file opened
main : write-config written
main : read-file opened
main : read-config written
main : write-task created
main : read-task created
main : starting write-task
main : starting read-task
 Nr |   write->irq    |    irq->read    |   write->read   |
-----------------------------------------------------------
read_task: error on RTSER_RTIOC_WAIT_EVENT, Connection timed out
read_task: error on RTSER_RTIOC_WAIT_EVENT, Connection timed out

any ideas?
regards /Bachman

On 09/09/2007, Jan Kiszka <jan.kiszka@domain.hid> wrote:
> > Are there any chance that modifications in the source will get it going?
>
> Maybe we are just lacking some pci_enable_device() on the right pci_dev.
> That would mean extending xeno_16550A to deal with PCI devices. Surely
> feasible, but it takes time and hardware to test - I'm lacking both.
>
> Are you willing to play a bit with it? Then, as a prototype, you could
> pick up the lookup and enabling code from the snx driver and hack it
> into xeno_16550A. If that works, we could discuss how to establish
> something generic for PCI-based adapters.
>
> Note that I haven't read through that driver in details. There is a
> slight risk that some more bits are different. On the other hand, the
> 16650 UART that seems to be on that card is advertised as
> 16550-compatible, thus it /should/ work...
>
> Jan
>
>
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 831 bytes --]

Bachman Kharazmi wrote:
> I've tried to make the pci serial controller visable to the 16550
> driver by adding:
> ## FROM 16550A.c
> static int rt_16550_interrupt(rtdm_irq_t * irq_context)
> {
> 	struct rt_16550_context *ctx;
> 	unsigned long base;
> 	int mode;
> 	int iir;
> 	uint64_t timestamp = rtdm_clock_read();
> 	int rbytes = 0;
> 	int events = 0;
> 	int modem;
> 	int ret = RTDM_IRQ_NONE;
> // TEST
> struct pci_dev *pdev = NULL;  // from
> http://bachman.tor.lindesign.se/tmp/driver/sunix/snx_golden.c
> pdev = pci_get_device(0, 0, pdev); // from
                       ^^^^^^
You need to provide the fitting vendor and device ID here. Then you have
to call pci_device_enable on that pdev - if it's not NULL (ie. not
found). That should tell us if it's already enough to enable the device.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

A guy recommended me to try to get things going with the working
mobo-serialport first before playing with linux drivers and serial
controllers.. probabily a good point to start at.

Anyhow I tried to used your cross-link.c to write ASCII to serialport
rtser0. I connected the cable to the dc motor drive which accepts
ASCII commands with CR(\r\n) at the end of every command line.
The light version of your code that I was testing with looks like:
http://pastebin.ca/690561

I'm trying to write two simple commands in the example, but I couldn't
get it working. Don't know why :/

First when I got that working it would be even nicer to be able to
give a argument to the binary which is a ASCII command(string) written
to rtser0.

thanks
/Bachman
On 10/09/2007, Jan Kiszka <jan.kiszka@domain.hid> wrote:
> Bachman Kharazmi wrote:
> > I've tried to make the pci serial controller visable to the 16550
> > driver by adding:
> > ## FROM 16550A.c
> > static int rt_16550_interrupt(rtdm_irq_t * irq_context)
> > {
> >       struct rt_16550_context *ctx;
> >       unsigned long base;
> >       int mode;
> >       int iir;
> >       uint64_t timestamp = rtdm_clock_read();
> >       int rbytes = 0;
> >       int events = 0;
> >       int modem;
> >       int ret = RTDM_IRQ_NONE;
> > // TEST
> > struct pci_dev *pdev = NULL;  // from
> > http://bachman.tor.lindesign.se/tmp/driver/sunix/snx_golden.c
> > pdev = pci_get_device(0, 0, pdev); // from
>                        ^^^^^^
> You need to provide the fitting vendor and device ID here. Then you have
> to call pci_device_enable on that pdev - if it's not NULL (ie. not
> found). That should tell us if it's already enough to enable the device.
>
> Jan
>
>
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1248 bytes --]

Bachman Kharazmi wrote:
> A guy recommended me to try to get things going with the working
> mobo-serialport first before playing with linux drivers and serial
> controllers.. probabily a good point to start at.

For sure. You may also test against a different PC running Linux or a
Windoz box with terminal programs installed, BTW.

> 
> Anyhow I tried to used your cross-link.c to write ASCII to serialport
> rtser0. I connected the cable to the dc motor drive which accepts
> ASCII commands with CR(\r\n) at the end of every command line.
> The light version of your code that I was testing with looks like:
> http://pastebin.ca/690561

<Quick glance mode> You write strings of different length in lines 152
and 153, but you pass the same size (sz = sizeof(RTIME), ie. 8).

> 
> I'm trying to write two simple commands in the example, but I couldn't
> get it working. Don't know why :/
> 
> First when I got that working it would be even nicer to be able to
> give a argument to the binary which is a ASCII command(string) written
> to rtser0.

Go wild and write such a trivial, command line driver tool to test
serial ports! Would be highly appreciated, and quickly accepted into the
tree (if it's clean).

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

[-- Attachment #1: Type: text/plain, Size: 2064 bytes --]

Now I've a modified crosslink (attached) which take argument and write
to rtser0.
I know it's ugly. But it's working and that's always good ;)

Anyhow. From what I've heard it will be inefficient if my java code
would call "/usr/src/xenomai/example/crosslink string" even time a
string should be written to serial.

And as it looks now I can't even run the binary as user else than root
(since it starts a xenmai thread?).

How can I as a 'user' open a device, write to it and close it without
keeping starting processes or threads every time?

I would like to be able to do something like:
file =  open("/dev/myTestSerialDevice");
while (...)
file.write(...)
and finally close
file.close() when there's no more to write.

/Bachman

On 10/09/2007, Jan Kiszka <jan.kiszka@domain.hid> wrote:
> Bachman Kharazmi wrote:
> > A guy recommended me to try to get things going with the working
> > mobo-serialport first before playing with linux drivers and serial
> > controllers.. probabily a good point to start at.
>
> For sure. You may also test against a different PC running Linux or a
> Windoz box with terminal programs installed, BTW.
>
> >
> > Anyhow I tried to used your cross-link.c to write ASCII to serialport
> > rtser0. I connected the cable to the dc motor drive which accepts
> > ASCII commands with CR(\r\n) at the end of every command line.
> > The light version of your code that I was testing with looks like:
> > http://pastebin.ca/690561
>
> <Quick glance mode> You write strings of different length in lines 152
> and 153, but you pass the same size (sz = sizeof(RTIME), ie. 8).
>
> >
> > I'm trying to write two simple commands in the example, but I couldn't
> > get it working. Don't know why :/
> >
> > First when I got that working it would be even nicer to be able to
> > give a argument to the binary which is a ASCII command(string) written
> > to rtser0.
>
> Go wild and write such a trivial, command line driver tool to test
> serial ports! Would be highly appreciated, and quickly accepted into the
> tree (if it's clean).
>
> Jan
>
>
>

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: cross-link.c --]
[-- Type: text/x-csrc; name="cross-link.c", Size: 6180 bytes --]

/*
 * cross-link.c
 *
 * Userspace test program (Xenomai native skin) for RTDM-based UART drivers
 * Copyright 2005 by Joerg Langenberg <joergel75@gmx.net>
 *
 * Updates by Jan Kiszka <jan.kiszka@web.de>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */
#include <stdio.h>
#include <signal.h>
#include <unistd.h>
#include <sys/mman.h>
#include <string.h>
#include <stdlib.h>
#include <native/task.h>
#include <native/timer.h>

#include <rtdm/rtserial.h>

#define MAIN_PREFIX   "main : "
#define WTASK_PREFIX  "write_task: "
//#define RTASK_PREFIX  "read_task: "

#define WRITE_FILE    "rtser0"

int write_fd = -1;

#define STATE_FILE_OPENED         1
#define STATE_TASK_CREATED        2

unsigned int write_state = 0;

/*                           --s-ms-us-ns */
RTIME write_task_period_ns =    100000000llu;
RT_TASK write_task;

static const struct rtser_config read_config = {
	.config_mask       = 0xFFFF,
	.baud_rate         = 115200,
	.parity            = RTSER_DEF_PARITY,
	.data_bits         = RTSER_DEF_BITS,
	.stop_bits         = RTSER_DEF_STOPB,
	.handshake         = RTSER_DEF_HAND,
	.fifo_depth        = RTSER_DEF_FIFO_DEPTH,
	.rx_timeout        = RTSER_DEF_TIMEOUT,
	.tx_timeout        = RTSER_DEF_TIMEOUT,
	.event_timeout     = 1000000000, /* 1 s */
	.timestamp_history = RTSER_RX_TIMESTAMP_HISTORY,
	.event_mask        = RTSER_EVENT_RXPEND,
};

static const struct rtser_config write_config = {
	.config_mask       = RTSER_SET_BAUD | RTSER_SET_TIMESTAMP_HISTORY,
	.baud_rate         = 115200,
	.timestamp_history = RTSER_DEF_TIMESTAMP_HISTORY,
	/* the rest implicitely remains default */
};

static int close_file( int fd, char *name)
{
	int err, i=0;
	
	do {
		i++;
		err = rt_dev_close(fd);
		switch (err) {
		case -EAGAIN:
			printf(MAIN_PREFIX "%s -> EAGAIN (%d times)\n",
			       name, i);
			rt_task_sleep(50000); /* wait 50us */
			break;
		case 0:
			printf(MAIN_PREFIX "%s -> closed\n", name);
			break;
		default:
			printf(MAIN_PREFIX "%s -> %s\n", name,
			       strerror(-err));
			break;
		}
	} while (err == -EAGAIN && i < 10);

	return err;
}

void cleanup_all(void)
{
	if (write_state & STATE_FILE_OPENED) {
		close_file(write_fd, WRITE_FILE " (write)");
		write_state &= ~STATE_FILE_OPENED;
	}
	
	if (write_state & STATE_TASK_CREATED) {
		printf(MAIN_PREFIX "delete write_task\n");
		rt_task_delete(&write_task);
		write_state &= ~STATE_TASK_CREATED;
	}
	
}

void catch_signal(int sig)
{
	cleanup_all();
	printf(MAIN_PREFIX "exit\n");
	return;
}

//void write_task_proc(void *arg){
void write_task_proc(void *arg){
char* myString1=arg;
//printf("%s\n",myString1);
	int err;
//	RTIME write_time;
	int write_time;
	ssize_t sz = sizeof(RTIME);
	ssize_t written = 0;
	int counter=0;

	err = rt_task_set_periodic(NULL, TM_NOW,
				   rt_timer_ns2ticks(write_task_period_ns));
	if (err) {
		printf(WTASK_PREFIX "error on set periodic, %s\n",
		       strerror(-err));
		goto exit_write_task;
	}

	while (counter==0) {
		err = rt_task_wait_period(NULL);
		if (err) {
			printf(WTASK_PREFIX
			       "error on rt_task_wait_period, %s\n",
			       strerror(-err));
			break;
		}
	
		//write_time = rt_timer_read();
		//write_time = 1;
		
	//	char *mystringa = "EN\r\n";
	//	char *mystringb = "V0\r\n";
		//int test=3;
		written = rt_dev_write(write_fd, myString1, strlen(myString1));
//		written = rt_dev_write(write_fd, mystringb, strlen(mystringb));
		if (written < 0 ) {
			printf(WTASK_PREFIX "error on rt_dev_write, %s\n",
			       strerror(-err));
			break;
		} else if (written != sz) {
//			printf(WTASK_PREFIX "only %d / %d byte transmitted\n",
//			       written, sz);
			break;
		}
	counter++;
	}
	exit(0);
 exit_write_task:
	if ((write_state & STATE_FILE_OPENED) &&
	    close_file(write_fd, WRITE_FILE " (write)") == 0)
		write_state &= ~STATE_FILE_OPENED;

//	printf(WTASK_PREFIX "exit\n");
}


int main(int argc , char *argv[]){
	int err = 0;
        //printf("%s\n",argv[1]);
	char *myString = argv[1];
	//printf("%s\n",myString);

//printf("%s\n",myString);
//char finalString = strcat( char *myString, char *cr );
char *finalString = (char*) malloc(sizeof(char)*(strlen(myString)+3));
sprintf(finalString, "%s\r\n", myString);
//
	signal(SIGTERM, catch_signal);
	signal(SIGINT, catch_signal);

	/* no memory-swapping for this programm */
	mlockall(MCL_CURRENT | MCL_FUTURE);

	/* open rtser0 */
	write_fd = rt_dev_open( WRITE_FILE, 0);
	if (write_fd < 0) {
		printf(MAIN_PREFIX "can't open %s (write), %s\n", WRITE_FILE,
		       strerror(-write_fd));
		goto error;
	}
	write_state |= STATE_FILE_OPENED;
//	printf(MAIN_PREFIX "write-file opened\n");

	/* writing write-config */
	err = rt_dev_ioctl(write_fd, RTSER_RTIOC_SET_CONFIG, &write_config);
	if (err) {
		printf(MAIN_PREFIX "error while RTSER_RTIOC_SET_CONFIG, %s\n",
		       strerror(-err));
		goto error;
	}
//	printf(MAIN_PREFIX "write-config written\n");

	/* create write_task */
	err = rt_task_create(&write_task, "write_task", 0, 50, 0);
	if (err) {
		printf(MAIN_PREFIX "failed to create write_task, %s\n",
		       strerror(-err));
		goto error;
	}
	write_state |= STATE_TASK_CREATED;
//	printf(MAIN_PREFIX "write-task created\n");

	/* start write_task */
//	printf(MAIN_PREFIX "starting write-task\n");
	//err = rt_task_start(&write_task, &write_task_proc, NULL);
	err = rt_task_start(&write_task, &write_task_proc, finalString);	

	if (err) {
		printf(MAIN_PREFIX "failed to start write_task, %s\n",
		       strerror(-err));
		goto error;
	}

	pause();
	return 0;

 error:
	cleanup_all();
	return err;
}

Re: [Xenomai-help] adding PCI support to 16550A.c

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1989 bytes --]

Bachman Kharazmi wrote:
> Now I've a modified crosslink (attached) which take argument and write
> to rtser0.
> I know it's ugly. But it's working and that's always good ;)
> 
> Anyhow. From what I've heard it will be inefficient if my java code
> would call "/usr/src/xenomai/example/crosslink string" even time a
> string should be written to serial.

Yes, it is better to avoid continuous process generations/destructions
for such kind of work, rather write a native Java binding to Xenomai in
order to talk to the serial port.

Hmm... Java... that makes me wonder again if your control loop is ought
to be closed over Java and where you got/will get the RT-capable Java
engine from. Could you tell us more about the architecture?

> 
> And as it looks now I can't even run the binary as user else than root
> (since it starts a xenmai thread?).

There is a module parameter of the nucleus called xenomai_gid. Provide a
group ID, and all users of that group will gain "Xenomai-power" (means
also the power to toast the box by spawning a CPU hog etc.). That allows
you to do at least /some/ kind of privilege separation between RT users
and root, though not a strictly secure one.

Did we already document this feature outside the code? Can't remember
right now though I hacked it... :-/

> 
> How can I as a 'user' open a device, write to it and close it without
> keeping starting processes or threads every time?
> 
> I would like to be able to do something like:
> file =  open("/dev/myTestSerialDevice");
> while (...)
> file.write(...)
> and finally close
> file.close() when there's no more to write.

You can copy this code almost unmodified into an application and compile
it against the Xenomai POSIX skin. The serial port configuration apart
(same under Linux, though), it will open "rtser<whatever>", even if
given like "/dev/rtser...", and work on that port as if it where a Linux
one. Life can be simple, sometimes.

Jan



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

> You can copy this code almost unmodified into an application and compile
> it against the Xenomai POSIX skin. The serial port configuration apart
> (same under Linux, though), it will open "rtser<whatever>", even if
> given like "/dev/rtser...", and work on that port as if it where a Linux
> one. Life can be simple, sometimes.

Simple? ;)
I've tried to write a c-program which basiclly should open the device
rtser0 and write 100x to it and then close.

The simple program can be found here: http://pastebin.ca/693873

I've connected the serial cable to a reading terminal, but there's no
data at all, when I run the binary. :/

/Bachman

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

another example which confuse me:
##
void write_task_proc(void *arg){
	int err;
	RTIME write_time;
	ssize_t sz = sizeof(RTIME);
	ssize_t written = 0;
	err = rt_task_set_periodic(NULL,
TM_NOW,rt_timer_ns2ticks(write_task_period_ns));
	printf("%d\n",err);
if (err) {
		printf(WTASK_PREFIX "error on set periodic, %s\n",
			strerror(-err));
	}
	while (1) {
		printf("test1\n");
		err = rt_task_wait_period(NULL);
		printf("%d\n",err);
		printf("test2\n");
		if (err) {
			printf(WTASK_PREFIX "error on rt_task_wait_period, %s\n", strerror(-err));
			break;
		}
....
##
the code does compile, test1 is printed but neither err or test2.
so what makes my code stoping executing there?
The documentation sais "If NULL, this count will never be copied
back." about the argument, but exactlly what does that mean?

Why does the work in other code examples I have?

/Bachman
On 12/09/2007, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > You can copy this code almost unmodified into an application and compile
> > it against the Xenomai POSIX skin. The serial port configuration apart
> > (same under Linux, though), it will open "rtser<whatever>", even if
> > given like "/dev/rtser...", and work on that port as if it where a Linux
> > one. Life can be simple, sometimes.
>
> Simple? ;)
> I've tried to write a c-program which basiclly should open the device
> rtser0 and write 100x to it and then close.
>
> The simple program can be found here: http://pastebin.ca/693873
>
> I've connected the serial cable to a reading terminal, but there's no
> data at all, when I run the binary. :/
>
> /Bachman
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/12/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> another example which confuse me:
> ##
> void write_task_proc(void *arg){
>         int err;
>         RTIME write_time;
>         ssize_t sz = sizeof(RTIME);
>         ssize_t written = 0;
>         err = rt_task_set_periodic(NULL,
> TM_NOW,rt_timer_ns2ticks(write_task_period_ns));

How much is writ_task_period_ns ? If it is 10000000011, then you
should wait around 10 s before "test2" is displayed.


-- 
                                               Gilles Chanteperdrix

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/12/07, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
> On 9/12/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > another example which confuse me:
> > ##
> > void write_task_proc(void *arg){
> >         int err;
> >         RTIME write_time;
> >         ssize_t sz = sizeof(RTIME);
> >         ssize_t written = 0;
> >         err = rt_task_set_periodic(NULL,
> > TM_NOW,rt_timer_ns2ticks(write_task_period_ns));
>
> How much is writ_task_period_ns ? If it is 10000000011, then you
> should wait around 10 s before "test2" is displayed.

Sorry, I confused 1 with l.
-- 
                                               Gilles Chanteperdrix

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

Sorry I didn't paste my code, http://pastebin.ca/694397
The output from it is:
sandbox:/home/bkw/code/driver# ./simpledriver
0
test1

And then it exits, no test2 printed at all.. :/
And nothing in the while loop seem to write to the serialport as I expected.

what am I doing wrong?
/Bachman
On 12/09/2007, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
> On 9/12/07, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
> > On 9/12/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > > another example which confuse me:
> > > ##
> > > void write_task_proc(void *arg){
> > >         int err;
> > >         RTIME write_time;
> > >         ssize_t sz = sizeof(RTIME);
> > >         ssize_t written = 0;
> > >         err = rt_task_set_periodic(NULL,
> > > TM_NOW,rt_timer_ns2ticks(write_task_period_ns));
> >
> > How much is writ_task_period_ns ? If it is 10000000011, then you
> > should wait around 10 s before "test2" is displayed.
>
> Sorry, I confused 1 with l.
> --
>                                                Gilles Chanteperdrix
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

Bachman Kharazmi wrote:
 > Sorry I didn't paste my code, http://pastebin.ca/694397
 > The output from it is:
 > sandbox:/home/bkw/code/driver# ./simpledriver
 > 0
 > test1
 > 
 > And then it exits, no test2 printed at all.. :/
 > And nothing in the while loop seem to write to the serialport as I expected.
 > 
 > what am I doing wrong?

You are returning from the main thread, which terminates the whole application.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/12/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > You can copy this code almost unmodified into an application and compile
> > it against the Xenomai POSIX skin. The serial port configuration apart
> > (same under Linux, though), it will open "rtser<whatever>", even if
> > given like "/dev/rtser...", and work on that port as if it where a Linux
> > one. Life can be simple, sometimes.
>
> Simple? ;)
> I've tried to write a c-program which basiclly should open the device
> rtser0 and write 100x to it and then close.
>
> The simple program can be found here: http://pastebin.ca/693873
>
> I've connected the serial cable to a reading terminal, but there's no
> data at all, when I run the binary. :/

You are missing a call to rt_task_shadow. Note that with the POSIX
skin, this would not happen, because the main thread is automatically
shadowed.

>
> /Bachman
>
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help
>


-- 
                                               Gilles Chanteperdrix

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

http://pastebin.ca/694372
I did add "err = rt_task_shadow(&task_desc,"mytask",20,T_CPU(0));" and
_something_ is sent out on the serialport at least, but not what I
expect.
What I do get on the other side of my nullmodem looks like:
00 00 00 00 00 00
000000a0: 00 00 00 00 00 00 00 00   00 00 00 00 00 00 00 00
000000b0: 00 00 00 00 00 00

not what I expected.. :(

thanks.
/Bachman

On 12/09/2007, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
> On 9/12/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > > You can copy this code almost unmodified into an application and compile
> > > it against the Xenomai POSIX skin. The serial port configuration apart
> > > (same under Linux, though), it will open "rtser<whatever>", even if
> > > given like "/dev/rtser...", and work on that port as if it where a Linux
> > > one. Life can be simple, sometimes.
> >
> > Simple? ;)
> > I've tried to write a c-program which basiclly should open the device
> > rtser0 and write 100x to it and then close.
> >
> > The simple program can be found here: http://pastebin.ca/693873
> >
> > I've connected the serial cable to a reading terminal, but there's no
> > data at all, when I run the binary. :/
>
> You are missing a call to rt_task_shadow. Note that with the POSIX
> skin, this would not happen, because the main thread is automatically
> shadowed.
>
> >
> > /Bachman
> >
> > _______________________________________________
> > Xenomai-help mailing list
> > Xenomai-help@domain.hid
> > https://mail.gna.org/listinfo/xenomai-help
> >
>
>
> --
>                                                Gilles Chanteperdrix
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1876 bytes --]

Bachman Kharazmi wrote:
>> You can copy this code almost unmodified into an application and compile
>> it against the Xenomai POSIX skin. The serial port configuration apart
>> (same under Linux, though), it will open "rtser<whatever>", even if
>> given like "/dev/rtser...", and work on that port as if it where a Linux
>> one. Life can be simple, sometimes.
> 
> Simple? ;)
> I've tried to write a c-program which basiclly should open the device
> rtser0 and write 100x to it and then close.
> 
> The simple program can be found here: http://pastebin.ca/693873
> 
> I've connected the serial cable to a reading terminal, but there's no
> data at all, when I run the binary. :/
> 

[Untested late-night hack after some Maß of beer - please excuse
potential nonsense.]

POSIX skin
----------
(use xeno-config --posix-cflags --posix-ldflags as compiler args)

#include bla bla bla

static const struct rtser_config write_config = {
        .config_mask       = RTSER_SET_BAUD,
        .baud_rate         = 115200,
};

int main(int argc, char *argv[])
{
	int fd;

	fd = open("/dev/rtser0", O_RDWR);
	ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
	write(fd, argv[1], strlen(argv[1]));
	close(fd);

	return 0;
}


Native skin
-----------

...
int main(int argc, char *argv[])
{
	int fd;
	RT_TASK task;

	fd = rt_dev_open("/dev/rtser0", O_RDWR);
	rt_dev_ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);

	rt_task_shadow(&task, NULL, 0, 0); /* may be done earlier,
                                              but not later */

	rt_dev_write(fd, argv[1], strlen(argv[1]));
	rt_dev_close(fd);

	return 0;
}


Similar is imaginable for reading from the serial port.

Completed and tested versions, maybe also taking the baudrate as
argument, will be happily accepted to the examples or tools repos.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] adding PCI support to 16550A.c

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 2169 bytes --]

Jan Kiszka wrote:
> Bachman Kharazmi wrote:
>>> You can copy this code almost unmodified into an application and compile
>>> it against the Xenomai POSIX skin. The serial port configuration apart
>>> (same under Linux, though), it will open "rtser<whatever>", even if
>>> given like "/dev/rtser...", and work on that port as if it where a Linux
>>> one. Life can be simple, sometimes.
>> Simple? ;)
>> I've tried to write a c-program which basiclly should open the device
>> rtser0 and write 100x to it and then close.
>>
>> The simple program can be found here: http://pastebin.ca/693873
>>
>> I've connected the serial cable to a reading terminal, but there's no
>> data at all, when I run the binary. :/
>>
> 
> [Untested late-night hack after some Maß of beer - please excuse
> potential nonsense.]
> 
> POSIX skin
> ----------
> (use xeno-config --posix-cflags --posix-ldflags as compiler args)
> 
> #include bla bla bla
> 
> static const struct rtser_config write_config = {
>         .config_mask       = RTSER_SET_BAUD,
>         .baud_rate         = 115200,
> };
> 
> int main(int argc, char *argv[])
> {
> 	int fd;
> 
> 	fd = open("/dev/rtser0", O_RDWR);
> 	ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
> 	write(fd, argv[1], strlen(argv[1]));
> 	close(fd);
> 
> 	return 0;
> }
> 
> 
> Native skin
> -----------
> 
> ...
> int main(int argc, char *argv[])
> {
> 	int fd;
> 	RT_TASK task;
> 
> 	fd = rt_dev_open("/dev/rtser0", O_RDWR);

The first error, copy&paste: This must be just "rtser0".

> 	rt_dev_ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
> 
> 	rt_task_shadow(&task, NULL, 0, 0); /* may be done earlier,
>                                               but not later */
> 
> 	rt_dev_write(fd, argv[1], strlen(argv[1]));
> 	rt_dev_close(fd);
> 
> 	return 0;
> }
> 
> 
> Similar is imaginable for reading from the serial port.
> 
> Completed and tested versions, maybe also taking the baudrate as
> argument, will be happily accepted to the examples or tools repos.

BTW, a generic serial dump tool should, of course, also take the port
number as parameter.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

int main(int argc)
{
       int fd;
	char* myString = "helloworld\n";
       fd = open("rtser0", O_RDWR);
       ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
       write(fd, myString, strlen(myString));
       close(fd);
       return 0;
}
Is supposed to be the POSIX skin. It does compile, but nothing sent on
my serial port.


On 13/09/2007, Jan Kiszka <jan.kiszka@domain.hid> wrote:
> Jan Kiszka wrote:
> > Bachman Kharazmi wrote:
> >>> You can copy this code almost unmodified into an application and compile
> >>> it against the Xenomai POSIX skin. The serial port configuration apart
> >>> (same under Linux, though), it will open "rtser<whatever>", even if
> >>> given like "/dev/rtser...", and work on that port as if it where a Linux
> >>> one. Life can be simple, sometimes.
> >> Simple? ;)
> >> I've tried to write a c-program which basiclly should open the device
> >> rtser0 and write 100x to it and then close.
> >>
> >> The simple program can be found here: http://pastebin.ca/693873
> >>
> >> I've connected the serial cable to a reading terminal, but there's no
> >> data at all, when I run the binary. :/
> >>
> >
> > [Untested late-night hack after some Maß of beer - please excuse
> > potential nonsense.]
> >
> > POSIX skin
> > ----------
> > (use xeno-config --posix-cflags --posix-ldflags as compiler args)
> >
> > #include bla bla bla
> >
> > static const struct rtser_config write_config = {
> >         .config_mask       = RTSER_SET_BAUD,
> >         .baud_rate         = 115200,
> > };
> >
> > int main(int argc, char *argv[])
> > {
> >       int fd;
> >
> >       fd = open("/dev/rtser0", O_RDWR);
> >       ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
> >       write(fd, argv[1], strlen(argv[1]));
> >       close(fd);
> >
> >       return 0;
> > }
> >
> >
> > Native skin
> > -----------
> >
> > ...
> > int main(int argc, char *argv[])
> > {
> >       int fd;
> >       RT_TASK task;
> >
> >       fd = rt_dev_open("/dev/rtser0", O_RDWR);
>
> The first error, copy&paste: This must be just "rtser0".
>
> >       rt_dev_ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
> >
> >       rt_task_shadow(&task, NULL, 0, 0); /* may be done earlier,
> >                                               but not later */
> >
> >       rt_dev_write(fd, argv[1], strlen(argv[1]));
> >       rt_dev_close(fd);
> >
> >       return 0;
> > }
> >
> >
> > Similar is imaginable for reading from the serial port.
> >
> > Completed and tested versions, maybe also taking the baudrate as
> > argument, will be happily accepted to the examples or tools repos.
>
> BTW, a generic serial dump tool should, of course, also take the port
> number as parameter.
>
> Jan
>
>
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/13/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> int main(int argc)
> {
>        int fd;
>         char* myString = "helloworld\n";
>        fd = open("rtser0", O_RDWR);
>        ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
>        write(fd, myString, strlen(myString));
>        close(fd);
>        return 0;
> }
> Is supposed to be the POSIX skin. It does compile, but nothing sent on
> my serial port.

This code is crappy. You do not check the return values of functions
which return a result so you have no chance to understand what
happens.

-- 
                                               Gilles Chanteperdrix

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

int main(int argc){
        int fd,err;
         char* myString = "helloworld\n";
        fd = open("rtser0", O_RDWR);
	printf("fd is %d\n", fd);
        ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
        err=write(fd, myString, strlen(myString));
        printf("err is %d\n", err);
	close(fd);
        return 0;
}

gives
sandbox:/home/bkw/code/driver_v4# ./simpledriver_v4
fd is -1
err is -1

Which means that fd didn't open the device successfully? :/
/B

On 13/09/2007, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
> On 9/13/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > int main(int argc)
> > {
> >        int fd;
> >         char* myString = "helloworld\n";
> >        fd = open("rtser0", O_RDWR);
> >        ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
> >        write(fd, myString, strlen(myString));
> >        close(fd);
> >        return 0;
> > }
> > Is supposed to be the POSIX skin. It does compile, but nothing sent on
> > my serial port.
>
> This code is crappy. You do not check the return values of functions
> which return a result so you have no chance to understand what
> happens.
>
> --
>                                                Gilles Chanteperdrix
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/13/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> int main(int argc){
>         int fd,err;
>          char* myString = "helloworld\n";
>         fd = open("rtser0", O_RDWR);
>         printf("fd is %d\n", fd);
>         ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
>         err=write(fd, myString, strlen(myString));
>         printf("err is %d\n", err);
>         close(fd);
>         return 0;
> }
>
> gives
> sandbox:/home/bkw/code/driver_v4# ./simpledriver_v4
> fd is -1
> err is -1
>
> Which means that fd didn't open the device successfully? :/
> /B

This is POSIX, when a function call returns -1, check errno to see the
failure reason. BTW, you did not check the return value of ioctl.

-- 
                                               Gilles Chanteperdrix

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/13/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> int main(int argc){
>         int fd,err;
>          char* myString = "helloworld\n";
>         fd = open("rtser0", O_RDWR);
>         if (fd < 0) {
>                 printf(MAIN_PREFIX "1 : can't open %s (write), %s\n", WRITE_FILE,
> strerror(-fd));
>                 return fd;
>         }
>         err=ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
>         if (err < 0) {
>                 printf(MAIN_PREFIX "2 : error %s (write), %s\n", WRITE_FILE, strerror(-fd));
>                 return err;
>         }
>         err=write(fd, myString, strlen(myString));
>         if (err < 0) {
>                 printf(MAIN_PREFIX "3 : error %s (write), %s\n", WRITE_FILE, strerror(-fd));
>                 return err;
>         }
>         printf("err is %d\n", err);
>         close(fd);
>         return 0;
> }
>
> sandbox:/home/bkw/code/driver_v4# ./simpledriver_v4
> main : 1 : can't open rtser0 (write), Operation not permitted
>
> why that, I'm running as root? :/
> /BK

When a posix function returns -1, the reason for the failure is stored
in errno. You are not checking errno.
If you do not know what errno is, type simply "man errno".

-- 
                                               Gilles Chanteperdrix

Re: [Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

#include <errno.h>
...
int main(int argc){
        int fd,err;
        extern int errno;
	char* myString = "helloworld\n";
        errno=0;
	fd = open("rtser0", O_RDWR);
	printf("errno is : %d\n",errno);
	if (fd < 0) {
		printf(MAIN_PREFIX "1 : can't open %s (write), %s\n", WRITE_FILE,
strerror(-fd));
		return fd;
	}

sandbox:/home/bkw/code/driver_v4# ./simpledriver_v4
errno is : 2
main : 1 : can't open rtser0 (write), Operation not permitted

Correct way of using errno?
Hope so..
/Bachman
On 13/09/2007, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
> On 9/13/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> > int main(int argc){
> >         int fd,err;
> >          char* myString = "helloworld\n";
> >         fd = open("rtser0", O_RDWR);
> >         if (fd < 0) {
> >                 printf(MAIN_PREFIX "1 : can't open %s (write), %s\n", WRITE_FILE,
> > strerror(-fd));
> >                 return fd;
> >         }
> >         err=ioctl(fd, RTSER_RTIOC_SET_CONFIG, &write_config);
> >         if (err < 0) {
> >                 printf(MAIN_PREFIX "2 : error %s (write), %s\n", WRITE_FILE, strerror(-fd));
> >                 return err;
> >         }
> >         err=write(fd, myString, strlen(myString));
> >         if (err < 0) {
> >                 printf(MAIN_PREFIX "3 : error %s (write), %s\n", WRITE_FILE, strerror(-fd));
> >                 return err;
> >         }
> >         printf("err is %d\n", err);
> >         close(fd);
> >         return 0;
> > }
> >
> > sandbox:/home/bkw/code/driver_v4# ./simpledriver_v4
> > main : 1 : can't open rtser0 (write), Operation not permitted
> >
> > why that, I'm running as root? :/
> > /BK
>
> When a posix function returns -1, the reason for the failure is stored
> in errno. You are not checking errno.
> If you do not know what errno is, type simply "man errno".
>
> --
>                                                Gilles Chanteperdrix
>

Re: [Xenomai-help] adding PCI support to 16550A.c

[Gilles Chanteperdrix]

On 9/13/07, Bachman Kharazmi <bahkha@domain.hid> wrote:
> #include <errno.h>
> ...
> int main(int argc){
>         int fd,err;
>         extern int errno;
>         char* myString = "helloworld\n";
>         errno=0;
>         fd = open("rtser0", O_RDWR);
>         printf("errno is : %d\n",errno);
>         if (fd < 0) {
>                 printf(MAIN_PREFIX "1 : can't open %s (write), %s\n", WRITE_FILE,
> strerror(-fd));
>                 return fd;
>         }
>
> sandbox:/home/bkw/code/driver_v4# ./simpledriver_v4
> errno is : 2
> main : 1 : can't open rtser0 (write), Operation not permitted
>
> Correct way of using errno?

No, as indicated in errno manual page, the value of errno is defined
only if a system call indicates a failure. So, the correct way of
using errno is:

fd = open("rtser0", O_RDWR)

if (fd == -1) {
    fprintf(stderr, "open(rtser0): %s\n", strerror(errno));
    exit(EXIT_FAILURE);
}

or:

if (fd == -1) {
   perror("open(rtser0)");
   exit(EXIT_FAILURE);
}

Anyway, 2 is ENOENT strerror(2) is "No such file or directory"
It probably means that you are not compiling your program with the
flags returned by xeno-config.


-- 
                                               Gilles Chanteperdrix

[Xenomai-help] adding PCI support to 16550A.c

[Bachman Kharazmi]

I've tried to make the pci serial controller visable to the 16550
driver by adding:
## FROM 16550A.c
static int rt_16550_interrupt(rtdm_irq_t * irq_context)
{
	struct rt_16550_context *ctx;
	unsigned long base;
	int mode;
	int iir;
	uint64_t timestamp = rtdm_clock_read();
	int rbytes = 0;
	int events = 0;
	int modem;
	int ret = RTDM_IRQ_NONE;
// TEST
struct pci_dev *pdev = NULL;  // from
http://bachman.tor.lindesign.se/tmp/driver/sunix/snx_golden.c
pdev = pci_get_device(0, 0, pdev); // from
http://bachman.tor.lindesign.se/tmp/driver/sunix/snx_golden.c
//
	ctx = rtdm_irq_get_arg(irq_context, struct rt_16550_context);
	base = ctx->base_addr;
	mode = rt_16550_io_mode_from_ctx(ctx);

	rtdm_lock_get(&ctx->lock);

	while (1) {
		iir = rt_16550_reg_in(mode, base, IIR) & IIR_MASK;
		if (testbits(iir, IIR_PIRQ))
			break;


Loading module:
insmod ./xeno_16550A.ko io=0x3f8,0xbc00 irq=4,3 tx_fifo=10,20 start_index=0

running crosslink test:
sandbox:/usr/src/xenomai/examples/rtdm/profiles/serial# ./cross-link
main : write-file opened
main : write-config written
main : read-file opened
main : read-config written
main : write-task created
main : read-task created
main : starting write-task
main : starting read-task
 Nr |   write->irq    |    irq->read    |   write->read   |
-----------------------------------------------------------
read_task: error on RTSER_RTIOC_WAIT_EVENT, Connection timed out
read_task: error on RTSER_RTIOC_WAIT_EVENT, Connection timed out

any ideas?
regards /Bachman

On 09/09/2007, Jan Kiszka <jan.kiszka@domain.hid> wrote:
> > Are there any chance that modifications in the source will get it going?
>
> Maybe we are just lacking some pci_enable_device() on the right pci_dev.
> That would mean extending xeno_16550A to deal with PCI devices. Surely
> feasible, but it takes time and hardware to test - I'm lacking both.
>
> Are you willing to play a bit with it? Then, as a prototype, you could
> pick up the lookup and enabling code from the snx driver and hack it
> into xeno_16550A. If that works, we could discuss how to establish
> something generic for PCI-based adapters.
>
> Note that I haven't read through that driver in details. There is a
> slight risk that some more bits are different. On the other hand, the
> 16650 UART that seems to be on that card is advertised as
> 16550-compatible, thus it /should/ work...
>
> Jan
>
>
>