[Xenomai-core] Unwanted mode switch.

[Xenomai-core] Unwanted mode switch.

[Gilles Chanteperdrix]

[-- Attachment #1: Type: text/plain, Size: 362 bytes --]


Hi,

in order to test the nocow patch, I wrote the attached test. Despite the
fact that there is no longer any page fault (I have nucleus debugging
on, so I would get a message if there was a fault), there is still an
unwanted mode switch at step 5. Any idea what could be the cause ?


-- 
                                                 Gilles Chanteperdrix

[-- Attachment #2: test_fork.c --]
[-- Type: text/x-csrc, Size: 2663 bytes --]

#include <sys/types.h>
#include <sys/wait.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <time.h>
#include <signal.h>
#include <string.h>
#include <sys/mman.h>
#include <execinfo.h>
#include <native/timer.h>
#include <native/task.h>

unsigned step;
void do_backtrace(void)
{
	static void *buffer[200];
	int levels = backtrace(buffer, sizeof(buffer)/sizeof(void *));
	backtrace_symbols_fd(buffer, levels, STDOUT_FILENO);
}
void mode_switch(int sig)
{
	printf("mode switch at step %d, test failed!\n", step);
	do_backtrace();
	exit(EXIT_FAILURE);
}

int main(void)
{
	RT_TASK task;
	RTIME time;
	pid_t pid;
	int rc;

	if ((SIG_ERR == signal(SIGXCPU, &mode_switch))) {
		perror("signal");
		exit(EXIT_FAILURE);
	}

	pid = fork();

	if (pid < 0) {
		perror("fork");
		exit(EXIT_FAILURE);
	}

	if (pid > 0) {
		int status;
		if (waitpid(pid, &status, 0) == -1) {
			perror("waitpid");
			exit(EXIT_FAILURE);
		}
		status = WEXITSTATUS(status);
		if (status == EXIT_SUCCESS)
			fprintf(stderr, "Test passed\n");
		_exit(status);
	}
	
	if (mlockall(MCL_CURRENT | MCL_FUTURE)) {
		perror("mlockall");
		exit(EXIT_FAILURE);
	}

	step = 1;
	if ((rc = rt_task_shadow(&task, "test_vm1", 1, 0))) {
		fprintf(stderr,
			"rt_task_shadow: %s\n", strerror(-rc));
		exit(EXIT_FAILURE);
	}

	step = 2;
	if ((rc = rt_task_set_mode(0, T_WARNSW, NULL))) {
		fprintf(stderr,
			"rt_task_set_mode_np: %s\n", strerror(-rc));
		exit(EXIT_FAILURE);
	}

	step = 3;
	time = rt_timer_read();

	step = 4;
	if ((rc = rt_task_set_mode(T_PRIMARY, 0, NULL))) {
		fprintf(stderr,
			"rt_task_set_mode_np: %s\n", strerror(-rc));
		exit(EXIT_FAILURE);
	}
	pid = fork();

	if (pid < 0) {
		perror("fork");
		exit(EXIT_FAILURE);
	}

	if (pid > 0) {
		int status;
		step = 5;

		time = rt_timer_read();

		if ((rc = rt_task_set_mode(T_PRIMARY, 0, NULL))) {
			fprintf(stderr,
				"rt_task_set_mode_np: %s\n", strerror(-rc));
			exit(EXIT_FAILURE);
		}

		if (waitpid(pid, &status, 0) == -1) {
			perror("waitpid");
			exit(EXIT_FAILURE);
		}
		_exit(WEXITSTATUS(status));
	}
	
	if (mlockall(MCL_CURRENT | MCL_FUTURE)) {
		perror("mlockall");
		exit(EXIT_FAILURE);
	}

	step = 6;
	if ((rc = rt_task_shadow(&task, "test_vm2", 1, 0))) {
		fprintf(stderr,
			"rt_task_shadow: %s\n", strerror(-rc));
		exit(EXIT_FAILURE);
	}

	step = 7;
	if ((rc = rt_task_set_mode(0, T_WARNSW, NULL))) {
		fprintf(stderr,
			"rt_task_set_mode_np: %s\n", strerror(-rc));
		exit(EXIT_FAILURE);
	}

	step = 8;
	time = rt_timer_read();
	
	if ((rc = rt_task_set_mode(T_PRIMARY, 0, NULL))) {
		fprintf(stderr,
			"rt_task_set_mode_np: %s\n", strerror(-rc));
		exit(EXIT_FAILURE);
	}

	exit(EXIT_SUCCESS);
}

Re: [Xenomai-core] Unwanted mode switch.

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 702 bytes --]

Gilles Chanteperdrix wrote:
> Hi,
> 
> in order to test the nocow patch, I wrote the attached test. Despite the
> fact that there is no longer any page fault (I have nucleus debugging
> on, so I would get a message if there was a fault), there is still an
> unwanted mode switch at step 5. Any idea what could be the cause ?
> 

Put in some xntrace_user_freeze(0) before or after the point-of-interest
in the program, switch on the tracer, and check ipipe/trace/frozen (with
appropriate back_trace_points or post_trace_points) what is going on.
Once you know the particular code path, instrumenting it can also be
done this way: xntrace_special(some_id, some_value_of_interest).

Jan


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

Re: [Xenomai-core] Unwanted mode switch.

[Gilles Chanteperdrix]

Gilles Chanteperdrix wrote:
> Hi,
> 
> in order to test the nocow patch, I wrote the attached test. Despite the
> fact that there is no longer any page fault (I have nucleus debugging
> on, so I would get a message if there was a fault), there is still an
> unwanted mode switch at step 5. Any idea what could be the cause ?

I added a show_stack(NULL, NULL) in xnshadow_relax, and I get the
following stack trace:
xnshadow_relax
hisyscall_event
xnshadow_sys_migrate
losyscall_event

-- 
                                                 Gilles Chanteperdrix

Re: [Xenomai-core] Unwanted mode switch.

[Philippe Gerum]

On Thu, 2007-01-11 at 16:36 +0100, Gilles Chanteperdrix wrote:
> Gilles Chanteperdrix wrote:
> > Hi,
> > 
> > in order to test the nocow patch, I wrote the attached test. Despite the
> > fact that there is no longer any page fault (I have nucleus debugging
> > on, so I would get a message if there was a fault), there is still an
> > unwanted mode switch at step 5. Any idea what could be the cause ?
> 
> I added a show_stack(NULL, NULL) in xnshadow_relax, and I get the
> following stack trace:
> xnshadow_relax
> hisyscall_event

  ^^^ Eeek... Which would mean that we would somehow be running the
hi-stage handler for a task on behalf of, say, xnshadow_harden()? Could
you confirm the backtrace? Any chance the frame unwinding went south
here?

Dumping the muxid/muxop pair for the hisyscall_event call which attempts
to relax the underlying task might be interesting.

> xnshadow_sys_migrate
> losyscall_event
> 
-- 
Philippe.

Re: [Xenomai-core] Unwanted mode switch.

[Philippe Gerum]

On Thu, 2007-01-11 at 16:24 +0100, Gilles Chanteperdrix wrote:
> Hi,
> 
> in order to test the nocow patch, I wrote the attached test. Despite the
> fact that there is no longer any page fault (I have nucleus debugging
> on, so I would get a message if there was a fault), there is still an
> unwanted mode switch at step 5. Any idea what could be the cause ?
> 

I'll have a look asap.

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