[PATCH libevl] tests: skip sched-tp-overrun in virtual environments

[PATCH libevl] tests: skip sched-tp-overrun in virtual environments

[Tobias Schaffner]

Do not try to relax sched-tp-overrun on virtualized environments
but skip the whole test since the test relies on timing guarantees
that are not reliable under virtualization and further relaxations
do not make any sense.

Signed-off-by: Tobias Schaffner <tobias.schaffner@siemens.com>
---
 tests/sched-tp-overrun.c | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/tests/sched-tp-overrun.c b/tests/sched-tp-overrun.c
index 3bb6a21..a3ef27f 100644
--- a/tests/sched-tp-overrun.c
+++ b/tests/sched-tp-overrun.c
@@ -128,16 +128,9 @@ static void *tp_thread(void *arg)
 	 */
 	__Texpr_assert(nf.event.val == part * 2);
 
-	/*
-	 * Perform the following check on real hardware only, it is
-	 * likely to be wrong in a virtualized environment for timing
-	 * accuracy reason.
-	 */
-	if (!running_on_vm()) {
-		/* Check that a single notification was sent. */
-		__Fcall_assert(ret, evl_read_observable(tfd, &nf, 1));
-		__Texpr_assert(ret == -EAGAIN);
-	}
+	/* Check that a single notification was sent. */
+	__Fcall_assert(ret, evl_read_observable(tfd, &nf, 1));
+	__Texpr_assert(ret == -EAGAIN);
 
 	return NULL;
 }
@@ -177,6 +170,15 @@ int main(int argc, char *argv[])
 		return 1;
 	}
 
+	/*
+	 * Skip on virtualized environments since the test relies on
+	 * timing guarantees that are not reliable under virtualization.
+	 */
+	if (running_on_vm()) {
+		emit_info("unchecked (vm)");
+		return EXIT_NO_STATUS;
+	}
+
 	param.sched_priority = HIGH_PRIO;
 	__Texpr_assert(pthread_setschedparam(pthread_self(),
 				SCHED_FIFO, &param) == 0);
-- 
2.43.0

Re: [PATCH libevl] tests: skip sched-tp-overrun in virtual environments

[Philippe Gerum]

Tobias Schaffner <tobias.schaffner@siemens.com> writes:

> Do not try to relax sched-tp-overrun on virtualized environments
> but skip the whole test since the test relies on timing guarantees
> that are not reliable under virtualization and further relaxations
> do not make any sense.
>
> Signed-off-by: Tobias Schaffner <tobias.schaffner@siemens.com>
> ---
>  tests/sched-tp-overrun.c | 22 ++++++++++++----------
>  1 file changed, 12 insertions(+), 10 deletions(-)
>
> diff --git a/tests/sched-tp-overrun.c b/tests/sched-tp-overrun.c
> index 3bb6a21..a3ef27f 100644
> --- a/tests/sched-tp-overrun.c
> +++ b/tests/sched-tp-overrun.c
> @@ -128,16 +128,9 @@ static void *tp_thread(void *arg)
>  	 */
>  	__Texpr_assert(nf.event.val == part * 2);
>  
> -	/*
> -	 * Perform the following check on real hardware only, it is
> -	 * likely to be wrong in a virtualized environment for timing
> -	 * accuracy reason.
> -	 */
> -	if (!running_on_vm()) {
> -		/* Check that a single notification was sent. */
> -		__Fcall_assert(ret, evl_read_observable(tfd, &nf, 1));
> -		__Texpr_assert(ret == -EAGAIN);
> -	}
> +	/* Check that a single notification was sent. */
> +	__Fcall_assert(ret, evl_read_observable(tfd, &nf, 1));
> +	__Texpr_assert(ret == -EAGAIN);
>  
>  	return NULL;
>  }
> @@ -177,6 +170,15 @@ int main(int argc, char *argv[])
>  		return 1;
>  	}
>  
> +	/*
> +	 * Skip on virtualized environments since the test relies on
> +	 * timing guarantees that are not reliable under virtualization.
> +	 */
> +	if (running_on_vm()) {
> +		emit_info("unchecked (vm)");
> +		return EXIT_NO_STATUS;
> +	}
> +
>  	param.sched_priority = HIGH_PRIO;
>  	__Texpr_assert(pthread_setschedparam(pthread_self(),
>  				SCHED_FIFO, &param) == 0);

Merged, thanks.

-- 
Philippe.