[PATCH 6/6] sched_ext/scx_qmap: Pick idle CPU for direct dispatch on !wakeup enqueues

[Tejun Heo <tj@kernel.org>]

Because there was no way to directly dispatch to the local DSQ of a remote
CPU from ops.enqueue(), scx_qmap skipped looking for an idle CPU on !wakeup
enqueues. This restriction was removed and schbed_ext now allows
SCX_DSQ_LOCAL_ON verdicts for direct dispatches.

Factor out pick_direct_dispatch_cpu() from ops.select_cpu() and use it to
direct dispatch from ops.enqueue() on !wakeup enqueues.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: David Vernet <void@manifault.com>
Cc: Dan Schatzberg <schatzberg.dan@gmail.com>
Cc: Changwoo Min <changwoo@igalia.com>
Cc: Andrea Righi <righi.andrea@gmail.com>
---
 tools/sched_ext/scx_qmap.bpf.c | 39 ++++++++++++++++++++++++++--------
 tools/sched_ext/scx_qmap.c     |  5 +++--
 2 files changed, 33 insertions(+), 11 deletions(-)

diff --git a/tools/sched_ext/scx_qmap.bpf.c b/tools/sched_ext/scx_qmap.bpf.c
index 27e35066a602..892278f12dce 100644
--- a/tools/sched_ext/scx_qmap.bpf.c
+++ b/tools/sched_ext/scx_qmap.bpf.c
@@ -120,11 +120,26 @@ struct {
 } cpu_ctx_stor SEC(".maps");
 
 /* Statistics */
-u64 nr_enqueued, nr_dispatched, nr_reenqueued, nr_dequeued;
+u64 nr_enqueued, nr_dispatched, nr_reenqueued, nr_dequeued, nr_ddsp_from_enq;
 u64 nr_core_sched_execed;
 u32 cpuperf_min, cpuperf_avg, cpuperf_max;
 u32 cpuperf_target_min, cpuperf_target_avg, cpuperf_target_max;
 
+static s32 pick_direct_dispatch_cpu(struct task_struct *p, s32 prev_cpu)
+{
+	s32 cpu;
+
+	if (p->nr_cpus_allowed == 1 ||
+	    scx_bpf_test_and_clear_cpu_idle(prev_cpu))
+		return prev_cpu;
+
+	cpu = scx_bpf_pick_idle_cpu(p->cpus_ptr, 0);
+	if (cpu >= 0)
+		return cpu;
+
+	return -1;
+}
+
 s32 BPF_STRUCT_OPS(qmap_select_cpu, struct task_struct *p,
 		   s32 prev_cpu, u64 wake_flags)
 {
@@ -137,17 +152,14 @@ s32 BPF_STRUCT_OPS(qmap_select_cpu, struct task_struct *p,
 		return -ESRCH;
 	}
 
-	if (p->nr_cpus_allowed == 1 ||
-	    scx_bpf_test_and_clear_cpu_idle(prev_cpu)) {
+	cpu = pick_direct_dispatch_cpu(p, prev_cpu);
+
+	if (cpu >= 0) {
 		tctx->force_local = true;
+		return cpu;
+	} else {
 		return prev_cpu;
 	}
-
-	cpu = scx_bpf_pick_idle_cpu(p->cpus_ptr, 0);
-	if (cpu >= 0)
-		return cpu;
-
-	return prev_cpu;
 }
 
 static int weight_to_idx(u32 weight)
@@ -172,6 +184,7 @@ void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
 	u32 pid = p->pid;
 	int idx = weight_to_idx(p->scx.weight);
 	void *ring;
+	s32 cpu;
 
 	if (p->flags & PF_KTHREAD) {
 		if (stall_kernel_nth && !(++kernel_cnt % stall_kernel_nth))
@@ -207,6 +220,14 @@ void BPF_STRUCT_OPS(qmap_enqueue, struct task_struct *p, u64 enq_flags)
 		return;
 	}
 
+	/* if !WAKEUP, select_cpu() wasn't called, try direct dispatch */
+	if (!(enq_flags & SCX_ENQ_WAKEUP) &&
+	    (cpu = pick_direct_dispatch_cpu(p, scx_bpf_task_cpu(p))) >= 0) {
+		__sync_fetch_and_add(&nr_ddsp_from_enq, 1);
+		scx_bpf_dispatch(p, SCX_DSQ_LOCAL_ON | cpu, slice_ns, enq_flags);
+		return;
+	}
+
 	/*
 	 * If the task was re-enqueued due to the CPU being preempted by a
 	 * higher priority scheduling class, just re-enqueue the task directly
diff --git a/tools/sched_ext/scx_qmap.c b/tools/sched_ext/scx_qmap.c
index 304f0488a386..c9ca30d62b2b 100644
--- a/tools/sched_ext/scx_qmap.c
+++ b/tools/sched_ext/scx_qmap.c
@@ -116,10 +116,11 @@ int main(int argc, char **argv)
 		long nr_enqueued = skel->bss->nr_enqueued;
 		long nr_dispatched = skel->bss->nr_dispatched;
 
-		printf("stats  : enq=%lu dsp=%lu delta=%ld reenq=%"PRIu64" deq=%"PRIu64" core=%"PRIu64"\n",
+		printf("stats  : enq=%lu dsp=%lu delta=%ld reenq=%"PRIu64" deq=%"PRIu64" core=%"PRIu64" enq_ddsp=%"PRIu64"\n",
 		       nr_enqueued, nr_dispatched, nr_enqueued - nr_dispatched,
 		       skel->bss->nr_reenqueued, skel->bss->nr_dequeued,
-		       skel->bss->nr_core_sched_execed);
+		       skel->bss->nr_core_sched_execed,
+		       skel->bss->nr_ddsp_from_enq);
 		if (__COMPAT_has_ksym("scx_bpf_cpuperf_cur"))
 			printf("cpuperf: cur min/avg/max=%u/%u/%u target min/avg/max=%u/%u/%u\n",
 			       skel->bss->cpuperf_min,
-- 
2.45.2