[LTP] [PATCH v2] sched/starvation: Choose from available cpus

[LTP] [PATCH v2] sched/starvation: Choose from available cpus

[Edward Liaw via ltp]

From: Edward Liaw via ltp <ltp@lists.linux.it>

Use the first available cpu in the sched_getaffinity set.

This test was failing as a 32bit test for some arm architectures where
cpu 0 does not support 32bit applications.

Reviewed-by: Petr Vorel <pvorel@suse.cz>
Signed-off-by: Edward Liaw <edliaw@google.com>
---
 .../kernel/sched/cfs-scheduler/starvation.c   | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

diff --git a/testcases/kernel/sched/cfs-scheduler/starvation.c b/testcases/kernel/sched/cfs-scheduler/starvation.c
index eb9fd6ff5..0fd45e281 100644
--- a/testcases/kernel/sched/cfs-scheduler/starvation.c
+++ b/testcases/kernel/sched/cfs-scheduler/starvation.c
@@ -49,12 +49,27 @@ again:
 static void setup(void)
 {
 	cpu_set_t mask;
+	int cpu = 0;
+	long ncpus = tst_ncpus_conf();
 
 	CPU_ZERO(&mask);
 
-	CPU_SET(0, &mask);
+	/* Restrict test to a single cpu */
+	TST_EXP_PASS(sched_getaffinity(0, sizeof(mask), &mask));
 
-	TST_EXP_POSITIVE(sched_setaffinity(0, sizeof(mask), &mask));
+	if (CPU_COUNT(&mask) == 0)
+		tst_brk(TBROK, "No cpus available");
+
+	while (CPU_ISSET(cpu, &mask) == 0 && cpu < ncpus)
+		cpu++;
+
+	CPU_ZERO(&mask);
+
+	CPU_SET(cpu, &mask);
+
+	TST_EXP_PASS(sched_setaffinity(0, sizeof(mask), &mask));
+
+	tst_res(TINFO, "Set affinity to CPU %d", cpu);
 
 	if (tst_parse_long(str_loop, &loop, 1, LONG_MAX))
 		tst_brk(TBROK, "Invalid number of loop number '%s'", str_loop);
-- 
2.45.1.467.gbab1589fc0-goog


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH v2] sched/starvation: Choose from available cpus

[Petr Vorel]

Hi Edward, Wei,

> Use the first available cpu in the sched_getaffinity set.

> This test was failing as a 32bit test for some arm architectures where
> cpu 0 does not support 32bit applications.

> Reviewed-by: Petr Vorel <pvorel@suse.cz>
> Signed-off-by: Edward Liaw <edliaw@google.com>
> ---
>  .../kernel/sched/cfs-scheduler/starvation.c   | 19 +++++++++++++++++--
>  1 file changed, 17 insertions(+), 2 deletions(-)

> diff --git a/testcases/kernel/sched/cfs-scheduler/starvation.c b/testcases/kernel/sched/cfs-scheduler/starvation.c
> index eb9fd6ff5..0fd45e281 100644
> --- a/testcases/kernel/sched/cfs-scheduler/starvation.c
> +++ b/testcases/kernel/sched/cfs-scheduler/starvation.c
> @@ -49,12 +49,27 @@ again:
>  static void setup(void)
>  {
>  	cpu_set_t mask;
> +	int cpu = 0;
> +	long ncpus = tst_ncpus_conf();

>  	CPU_ZERO(&mask);

> -	CPU_SET(0, &mask);
> +	/* Restrict test to a single cpu */
> +	TST_EXP_PASS(sched_getaffinity(0, sizeof(mask), &mask));

> -	TST_EXP_POSITIVE(sched_setaffinity(0, sizeof(mask), &mask));

Yeah, Wei was correct that both sched_getaffinity() and sched_setaffinity()
should use TST_EXP_PASS() because return only 0 on success (non-zero value is in
the mask, thus using TST_EXP_POSITIVE() was wrong. Ideally this would be in a
separate commit (before your changes), but I'm ok just to note it in the commit
message.

I was also thinking why test macros are used in setup() function (completely
unrelated to the change), I would just use plain code:

if (!sched_getaffinity(0, sizeof(mask), &mask))
	tst_brk(TBROK | TERRNO, "sched_getaffinity() failed");

> +	if (CPU_COUNT(&mask) == 0)
> +		tst_brk(TBROK, "No cpus available");

I was thinking a while and looking in other software which uses CPU_COUNT().
And agree this should be probably tst_brk(TBROK)

Reviewed-by: Petr Vorel <pvorel@suse.cz>

Kind regards,
Petr

> +
> +	while (CPU_ISSET(cpu, &mask) == 0 && cpu < ncpus)
> +		cpu++;
> +
> +	CPU_ZERO(&mask);
> +
> +	CPU_SET(cpu, &mask);
> +
> +	TST_EXP_PASS(sched_setaffinity(0, sizeof(mask), &mask));
> +
> +	tst_res(TINFO, "Set affinity to CPU %d", cpu);

>  	if (tst_parse_long(str_loop, &loop, 1, LONG_MAX))
>  		tst_brk(TBROK, "Invalid number of loop number '%s'", str_loop);

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp