container: expose priority and SCHED_OTHER policy

The more explicit API removes the arbitrary limit preventing use of SCHED_OTHER (referred to as SCHED_NORMAL in the kernel). This change also exposes priority value to set.

Signed-off-by: Ophestra <cat@gensokyo.uk>

internal/outcome/shim.go

@@ -274,6 +274,7 @@ func shimEntrypoint(k syscallDispatcher) {
 	cancelContainer.Store(&stop)
 	sp := shimPrivate{k: k, id: state.id}
 	z := container.New(ctx, msg)
+	z.SetScheduler = state.Shim.SchedPolicy > 0
 	z.SchedPolicy = state.Shim.SchedPolicy
 	z.Params = *stateParams.params
 	z.Stdin, z.Stdout, z.Stderr = os.Stdin, os.Stdout, os.Stderr

internal/pkg/exec.go

@@ -39,8 +39,8 @@ type ExecPath struct {
 	W bool
 }
 
-// SchedPolicy is the [container] scheduling policy.
-var SchedPolicy std.SchedPolicy
+// SetSchedIdle is whether to set [std.SCHED_IDLE] scheduling priority.
+var SetSchedIdle bool
 
 // PromoteLayers returns artifacts with identical-by-content layers promoted to
 // the highest priority instance, as if mounted via [ExecPath].
@@ -413,7 +413,8 @@ func (a *execArtifact) cure(f *FContext, hostNet bool) (err error) {
 	z.ParentPerm = 0700
 	z.HostNet = hostNet
 	z.Hostname = "cure"
-	z.SchedPolicy = SchedPolicy
+	z.SetScheduler = SetSchedIdle
+	z.SchedPolicy = std.SCHED_IDLE
 	if z.HostNet {
 		z.Hostname = "cure-net"
 	}