Feature: Add cmdline param sched_uek=[preempt,wakeidle]

Most of the scheduler tunables have been moved to debugfs, and
secureboot lockdown disallows modifications to debugfs. Due to this
limitation, sched feature from commit 57da733cf52f cannot be enabled
at runtime under SecureBoot. Adding a kernel cmdline parameter to
allow this vcpu preemption idle check to be set at the boot time
even under SecureBoot.

Commit 7332dec and commit 806486c from Mel Gorman alters
wakeup behaviors when waking and/or previous CPUs are idle. The
changes don't favor networking workloads under light load. Massive
RDS regressions have been observed on both AMD and Intel platforms.
Also adding a knob to allow users to switch back to previous behaviors
at boot time.

sched_uek=  UEK specific scheduler settings
            preempt  -  enable vcpu preemption idle check
            wakeidle -  disable two Mel Gorman's patches
            Example: sched_uek=preempt,wakeidle

Recent refactoring moves cpu preemption idle check to SCHED_FEAT,
and changes knob setup and its name.

Orabug: 34775543
Signed-off-by: Libo Chen <[email protected]>
Signed-off-by: Mark Nicholson <[email protected]>

Author: Libo Chen

kernel/sched/fair.c

@@ -7368,8 +7368,12 @@ wake_affine_idle(int this_cpu, int prev_cpu, int sync)
 	 * a cpufreq perspective, it's better to have higher utilisation
 	 * on one CPU.
 	 */
-	if (available_idle_cpu(this_cpu) && cpus_share_cache(this_cpu, prev_cpu))
-		return available_idle_cpu(prev_cpu) ? prev_cpu : this_cpu;
+	if (available_idle_cpu(this_cpu)) {
+		if (static_branch_unlikely(&on_exadata))
+			return this_cpu;
+		else if (cpus_share_cache(this_cpu, prev_cpu))
+			return available_idle_cpu(prev_cpu) ? prev_cpu : this_cpu;
+	}
 
 	if (sync && cpu_rq(this_cpu)->nr_running == 1)
 		return this_cpu;