[PATCH v3] selftests: powerpc: Fix online CPU selection

The size of the CPU affinity mask must be large enough for
systems with a very large number of CPUs. Otherwise, tests
which try to determine the first online CPU by calling
sched_getaffinity() will fail. This makes sure that the size
of the allocated affinity mask is dependent on the number of
CPUs as reported by get_nprocs_conf().

Fixes: 3752e453f6ba ("selftests/powerpc: Add tests of PMU EBBs")
Reported-by: Shirisha Ganta <redacted>
Signed-off-by: Sandipan Das <redacted>
Reviewed-by: Kamalesh Babulal <redacted>
---
Previous versions can be found at:
v2: https://lore.kernel.org/linuxppc-dev/20200609073733.997643-1-sandipan@linux.ibm.com/
v1: https://lore.kernel.org/linuxppc-dev/20200608144212.985144-1-sandipan@linux.ibm.com/

Changes in v3:
- Use get_nprocs_conf() instead of get_nprocs() as suggested by
  Srikar to cope with test failures that Michael reported for
  sparse SMT setups.

Changes in v2:
- Added NULL check for the affinity mask as suggested by Kamalesh.
- Changed "cpu set" to "CPU affinity mask" in the commit message.

---
 tools/testing/selftests/powerpc/utils.c | 37 +++++++++++++++++--------
 1 file changed, 25 insertions(+), 12 deletions(-)

diff --git a/tools/testing/selftests/powerpc/utils.c b/tools/testing/selftests/powerpc/utils.c
index 933678f1ed0a0..18b6a773d5c73 100644
--- a/tools/testing/selftests/powerpc/utils.c
+++ b/tools/testing/selftests/powerpc/utils.c

@@ -16,6 +16,7 @@
 #include <string.h>
 #include <sys/ioctl.h>
 #include <sys/stat.h>
+#include <sys/sysinfo.h>
 #include <sys/types.h>
 #include <sys/utsname.h>
 #include <unistd.h>

@@ -88,28 +89,40 @@ void *get_auxv_entry(int type)
 
 int pick_online_cpu(void)
 {
-	cpu_set_t mask;
-	int cpu;
+	int ncpus, cpu = -1;
+	cpu_set_t *mask;
+	size_t size;
+
+	ncpus = get_nprocs_conf();
+	size = CPU_ALLOC_SIZE(ncpus);
+	mask = CPU_ALLOC(ncpus);
+	if (!mask) {
+		perror("malloc");
+		return -1;
+	}
 
-	CPU_ZERO(&mask);
+	CPU_ZERO_S(size, mask);
 
-	if (sched_getaffinity(0, sizeof(mask), &mask)) {
+	if (sched_getaffinity(0, size, mask)) {
 		perror("sched_getaffinity");
-		return -1;
+		goto done;
 	}
 
 	/* We prefer a primary thread, but skip 0 */
-	for (cpu = 8; cpu < CPU_SETSIZE; cpu += 8)
-		if (CPU_ISSET(cpu, &mask))
-			return cpu;
+	for (cpu = 8; cpu < ncpus; cpu += 8)
+		if (CPU_ISSET_S(cpu, size, mask))
+			goto done;
 
 	/* Search for anything, but in reverse */
-	for (cpu = CPU_SETSIZE - 1; cpu >= 0; cpu--)
-		if (CPU_ISSET(cpu, &mask))
-			return cpu;
+	for (cpu = ncpus - 1; cpu >= 0; cpu--)
+		if (CPU_ISSET_S(cpu, size, mask))
+			goto done;
 
 	printf("No cpus in affinity mask?!\n");
-	return -1;
+
+done:
+	CPU_FREE(mask);
+	return cpu;
 }
 
 bool is_ppc64le(void)

-- 
2.25.1

On Thu, 30 Jul 2020 10:38:46 +0530, Sandipan Das wrote:

The size of the CPU affinity mask must be large enough for
systems with a very large number of CPUs. Otherwise, tests
which try to determine the first online CPU by calling
sched_getaffinity() will fail. This makes sure that the size
of the allocated affinity mask is dependent on the number of
CPUs as reported by get_nprocs_conf().

Applied to powerpc/next.

[1/1] selftests/powerpc: Fix online CPU selection
      https://git.kernel.org/powerpc/c/dfa03fff86027e58c8dba5c03ae68150d4e513ad

cheers