__set_cpus_allowed_ptr() makes use of stop_one_cpu() to call migration_cpu_stop() in order to perform migration of a task away from the CPU it's currently running on. If all is as expected then this shouldn't fail, but as the preceding patch shows it's possible for this assumption to be broken fairly subtly.
Add a warning to ensure that if stop_one_cpu() returns an error (-ENOENT is the only one it can currently return) then we warn about it in the kernel log, since this isn't expected to happen. I considered propogating the error upwards, but this would require a change to the return values allowed from the sched_setaffinity() syscall and would require that user programs handle errors other than those caused by their own bad input to the syscall. So for now this patch simply warns, which is an improvement over the silent error & incorrect scheduling we had before. Signed-off-by: Paul Burton <[email protected]> Cc: Ingo Molnar <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Thomas Gleixner <[email protected]> Cc: [email protected] --- kernel/sched/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 2380bc228dd0..cda3affd45b7 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -1127,7 +1127,8 @@ static int __set_cpus_allowed_ptr(struct task_struct *p, struct migration_arg arg = { p, dest_cpu }; /* Need help from migration thread: drop lock and wait. */ task_rq_unlock(rq, p, &rf); - stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg); + ret = stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg); + WARN_ON(ret); tlb_migrate_finish(p->mm); return 0; } else if (task_on_rq_queued(p)) { -- 2.17.0
