sched/mmcid: Drop per CPU CID immediately when switching to per task mode

When a exiting task initiates the switch from per CPU back to per task
mode, it has already dropped its CID and marked itself inactive. But a
leftover from an earlier iteration of the rework then reassigns the per
CPU CID to the exiting task with the transition bit set.

That's wrong as the task is already marked CID inactive, which means it is
inconsistent state. It's harmless because the CID is marked in transit and
therefore dropped back into the pool when the exiting task schedules out
either through preemption or the final schedule().

Simply drop the per CPU CID when the exiting task triggered the transition.

Fixes: fbd0e71dc370 ("sched/mmcid: Provide CID ownership mode fixup functions")
Signed-off-by: Thomas Gleixner <[e-mail]>
Signed-off-by: Peter Zijlstra (Intel) <[e-mail]>
Reviewed-by: Mathieu Desnoyers <[e-mail]>
Link: https://patch.msgid.link/[e-mail]

kernel/sched/core.c

···
 			scoped_guard(raw_spinlock_irq, &mm->mm_cid.lock) {
 				if (!__sched_mm_cid_exit(t))
 					return;
-				/* Mode change required. Transfer currents CID */
-				mm_cid_transit_to_task(current, this_cpu_ptr(mm->mm_cid.pcpu));
+				/*
+				 * Mode change. The task has the CID unset
+				 * already. The CPU CID is still valid and
+				 * does not have MM_CID_TRANSIT set as the
+				 * mode change has just taken effect under
+				 * mm::mm_cid::lock. Drop it.
+				 */
+				mm_drop_cid_on_cpu(mm, this_cpu_ptr(mm->mm_cid.pcpu));
 			}
 			mm_cid_fixup_cpus_to_tasks(mm);
 			return;