Re: Assistance Needed with EVL4 Monitoring Data Not Updating

[Philippe Gerum <rpm@xenomai.org>]

YinJie LI <liyinjie0418@gmail.com> writes:

> Hi，all
>
> I recently started working with EVL and wrote a simple piece of code to understand its behavior.
> However, I’ve encountered an issue where the monitoring data does not update as expected, 
> while latmus tests show changes.
>
> Here’s the code snippet:
>
> #include <sys/types.h>
> #include <unistd.h>
> #include <error.h>
> #include <sched.h>
> #include <pthread.h>
> #include <evl/sched.h>
> #include <evl/evl.h>
> #include <evl/thread.h>
>
> int main(int argc, char *const argv[]) {
> 	struct sched_param param;
> 	int ret, tfd;
>
> 	ret = evl_init();
> 	if (ret)
> 		error(1, -ret, "cannot initialize EVL”);
>
> 	param.sched_priority = 8;
> 	ret = pthread_setschedparam(pthread_self(), SCHED_FIFO, &param);
>
> 	tfd = evl_attach_thread(EVL_CLONE_PUBLIC, "evl-example-%d", getpid());
> 	if (tfd < 0)
> 		error(1, -tfd, "evl_attach_self() failed”);
>
> 	for (;;) {
> 		struct evl_version ver_t = evl_get_version();
> 		printf("Hello, EVL! %s \n", ver_t.version_string);

evl_printf(...)

> 		usleep(50);

evl_usleep(50)


> 	}
>
> 	return 0;
> }
>
> The monitoring command (evl ps -tsp) shows the following, which does not change over time:
>
> Every 1.0s: evl ps -tsp                                                                       
>
> CPU   PID   SCHED   PRIO  ISW     CTXSW     SYS       RWA       STAT     TIMEOUT      %CPU      CPUTIME       NAME
>   3       96822  fifo          8       1           1                1             0             X              -                   0.0        0:000.009        evl-example-96822
>
> The relevant versions used are as follows:
> libevl：r46
> Linux-evl: v6.6.23-evl2
> Gcc : 8.4
> Linux : x64 intel
>
> Could anyone help me understand why the monitoring data isn’t updating or suggest what I might be missing?
>
> Thanks in advance for your assistance!

This code loop never runs real-time evl-wise, because it's using plain
so-called "in-band" calls from the regular kernel logic, not evl
services. The first call to printf() downgrades the thread execution
mode to in-band, leaving it there indefinitely since there is no evl
call causing a switch back to out-of-band scheduling. As a result, this
thread does not perform any evl context switch as reported by 'evl ps'
in the CTXSW column, consumes no time in real-time mode and so on.

You may want to read this:
https://v4.xenomai.org/core/user-api/thread/
https://v4.xenomai.org/dovetail/altsched/

-- 
Philippe.