Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Kenta Ishiguro <kentaishiguro@sslab.ics.keio.ac.jp>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>,
        David Hildenbrand <david@redhat.com>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, pl@sslab.ics.keio.ac.jp,
        kono@sslab.ics.keio.ac.jp,
        Kenta Ishiguro <kentaishiguro@sslab.ics.keio.ac.jp>
Subject: [RFC PATCH 1/2] Prevent CFS from ignoring boost requests from KVM
Date:   Thu, 22 Apr 2021 00:08:30 +0900
Message-Id: <20210421150831.60133-2-kentaishiguro@sslab.ics.keio.ac.jp>
In-Reply-To: <20210421150831.60133-1-kentaishiguro@sslab.ics.keio.ac.jp>
References: <20210421150831.60133-1-kentaishiguro@sslab.ics.keio.ac.jp>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

This commit increases the vruntime of yielded vCPU to boost a vCPU instead
of the yielded vCPU when two vCPUs are in the same VM. This change avoids
the situation where scheduling the boosted vCPU is too unfair.

Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Kenta Ishiguro <kentaishiguro@sslab.ics.keio.ac.jp>
---
 kernel/sched/fair.c | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 794c2cb945f8..2908da3f4c77 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7204,9 +7204,36 @@ static void yield_task_fair(struct rq *rq)
 	set_skip_buddy(se);
 }
 
+static void deboost_yield_task_vruntime(struct sched_entity *next_se, struct sched_entity *yield_se)
+{
+	if (wakeup_preempt_entity(next_se, yield_se) < 1)
+		return;
+	yield_se->vruntime = next_se->vruntime - wakeup_gran(yield_se);
+}
+
+static void deboost_yield_task(struct sched_entity *next_se, struct sched_entity *yield_se)
+{
+	struct sched_entity *next_se_base = next_se;
+
+	if (rq_of(cfs_rq_of(yield_se)) != rq_of(cfs_rq_of(next_se)))
+		return;
+
+	for_each_sched_entity(yield_se) {
+		next_se = next_se_base;
+		for_each_sched_entity(next_se) {
+			if (cfs_rq_of(yield_se) == cfs_rq_of(next_se)) {
+				deboost_yield_task_vruntime(next_se, yield_se);
+				return;
+			}
+		}
+	}
+}
+
 static bool yield_to_task_fair(struct rq *rq, struct task_struct *p)
 {
 	struct sched_entity *se = &p->se;
+	struct task_struct *curr;
+	struct sched_entity *yield_se;
 
 	/* throttled hierarchies are not runnable */
 	if (!se->on_rq || throttled_hierarchy(cfs_rq_of(se)))
@@ -7215,6 +7242,10 @@ static bool yield_to_task_fair(struct rq *rq, struct task_struct *p)
 	/* Tell the scheduler that we'd really like pse to run next. */
 	set_next_buddy(se);
 
+	curr = rq->curr;
+	yield_se = &curr->se;
+	deboost_yield_task(se, yield_se);
+
 	yield_task_fair(rq);
 
 	return true;
-- 
2.30.2