commit 79104becf42b ("sched/fair: Forfeit vruntime on yield") changed
yield_task_fair() to only bump the deadline when the entity is eligible
(vruntime <= avg_vruntime). When the entity is ineligible the yield
becomes a complete no-op from scheduling perspective.

lost_exception_test calls sched_yield() 100,000 times per iteration to
race the EBB exception delivery with a context switch to the eat_cpu
companion process. After enough iterations the test process's vruntime
races ahead of avg_vruntime (each eligible yield bumps vruntime to
deadline, then advances deadline by one slice). Once ineligible,
yield_task_fair() does nothing: so the scheduler won't pick the
eat_cpu child. No context switch occurs, the PMAO race is never
triggered, and ebb_count stays at 0 forever causing the test to
hang until timeout.

Fix by replacing sched_yield() with nanosleep(0, 1ns). nanosleep() goes
through hrtimer_nanosleep() -> do_nanosleep(), which puts the task into
TASK_INTERRUPTIBLE and removes it from the run queue entirely. This
guarantees the scheduler picks the eat_cpu child, restoring the
context-switch guarantee the test requires. The 1ns duration is enough
to engage the hrtimer path while keeping the sleep effectively
instantaneous; the same race window between PMU overflow and context
switch is preserved.

Reported-by: Venkat Rao Bagalkote <[email protected]>
Signed-off-by: Athira Rajeev <[email protected]>
---
 .../selftests/powerpc/pmu/ebb/lost_exception_test.c   | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/powerpc/pmu/ebb/lost_exception_test.c 
b/tools/testing/selftests/powerpc/pmu/ebb/lost_exception_test.c
index ba2681a12cc7..9be5945f3b1f 100644
--- a/tools/testing/selftests/powerpc/pmu/ebb/lost_exception_test.c
+++ b/tools/testing/selftests/powerpc/pmu/ebb/lost_exception_test.c
@@ -8,6 +8,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <sys/mman.h>
+#include <time.h>
 
 #include "ebb.h"
 
@@ -22,6 +23,7 @@ static int test_body(void)
 {
        int i, orig_period, max_period;
        struct event event;
+       struct timespec ts = { .tv_sec = 0, .tv_nsec = 1 };
 
        SKIP_IF(!ebb_is_supported());
 
@@ -57,10 +59,15 @@ static int test_body(void)
                 * kernel to decide our timeslice is up and context switch to
                 * the other thread. When we come back our EBB will have been
                 * lost and we'll spin in this while loop forever.
+                *
+                * Use nanosleep(0) instead of sched_yield() to guarantee a
+                * context switch to the eat_cpu child regardless of the
+                * eligibility state. sched_yield() via yield_task_fair() may
+                * become a no-op when the task is ineligible (vruntime ahead
+                * of avg_vruntime), preventing the required context switch.
                 */
-
                for (i = 0; i < 100000; i++)
-                       sched_yield();
+                       nanosleep(&ts, NULL);
 
                /* Change the sample period slightly to try and hit the race */
                if (sample_period >= (orig_period + 200))
-- 
2.43.0


Reply via email to