Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:6 as permitted sender) client-ip=2620:137:e000::3:6;
From:   Xiaobing Li <xiaobing.li@samsung.com>
To:     mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com,
        vincent.guittot@linaro.org, dietmar.eggemann@arm.com,
        rostedt@goodmis.org, bsegall@google.com, mgorman@suse.de,
        bristot@redhat.com, vschneid@redhat.com, axboe@kernel.dk,
        asml.silence@gmail.com
Cc:     linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        io-uring@vger.kernel.org, kun.dou@samsung.com,
        peiwei.li@samsung.com, joshi.k@samsung.com,
        kundan.kumar@samsung.com, wenwen.chen@samsung.com,
        ruyi.zhang@samsung.com, Xiaobing Li <xiaobing.li@samsung.com>
Subject: [PATCH 1/3] SCHEDULER: Add an interface for counting real
 utilization.
Date:   Thu, 28 Sep 2023 10:22:26 +0800
Message-Id: <20230928022228.15770-2-xiaobing.li@samsung.com>
In-Reply-To: <20230928022228.15770-1-xiaobing.li@samsung.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset="utf-8"
CMS-TYPE: 105P
DLP-Filter: Pass
References: <20230928022228.15770-1-xiaobing.li@samsung.com>
        <CGME20230928023007epcas5p276b6e029a67001a6ed8ab28c05b2be9c@epcas5p2.samsung.com>
Precedence: bulk

Since pelt takes the running time of the thread as utilization, and for
some threads, although they are running, they are not actually
processing any transactions and are in an idling state.
our goal is to count the effective working time of the thread, so as to
Calculate the true utilization of threads.

Signed-off-by: Xiaobing Li <xiaobing.li@samsung.com>
---
 include/linux/kernel.h |  7 ++++++-
 include/linux/sched.h  |  1 +
 kernel/sched/cputime.c | 36 +++++++++++++++++++++++++++++++++++-
 kernel/sched/pelt.c    | 14 ++++++++++++++
 4 files changed, 56 insertions(+), 2 deletions(-)

diff --git a/include/linux/kernel.h b/include/linux/kernel.h
index cee8fe87e9f4..c1557fa9cbbe 100644
--- a/include/linux/kernel.h
+++ b/include/linux/kernel.h
@@ -37,7 +37,8 @@
 #include <uapi/linux/kernel.h>
 
 #define STACK_MAGIC	0xdeadbeef
-
+struct cfs_rq;
+struct sched_entity;
 /**
  * REPEAT_BYTE - repeat the value @x multiple times as an unsigned long value
  * @x: value to repeat
@@ -103,6 +104,10 @@ extern int __cond_resched(void);
 
 #elif defined(CONFIG_PREEMPT_DYNAMIC) && defined(CONFIG_HAVE_PREEMPT_DYNAMIC_CALL)
 
+extern void __update_sq_avg_block(u64 now, struct sched_entity *se);
+
+extern void __update_sq_avg(u64 now, struct sched_entity *se);
+
 extern int __cond_resched(void);
 
 DECLARE_STATIC_CALL(might_resched, __cond_resched);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 77f01ac385f7..403ccb456c9a 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -583,6 +583,7 @@ struct sched_entity {
 	 * collide with read-mostly values above.
 	 */
 	struct sched_avg		avg;
+	struct sched_avg		sq_avg;
 #endif
 };
 
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index af7952f12e6c..824203293fd9 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -479,6 +479,40 @@ void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st)
 
 #else /* !CONFIG_VIRT_CPU_ACCOUNTING_NATIVE: */
 
+void get_sqthread_util(struct task_struct *p)
+{
+	struct task_struct **sqstat = kcpustat_this_cpu->sq_util;
+
+	for (int i = 0; i < MAX_SQ_NUM; i++) {
+		if (sqstat[i] && (task_cpu(sqstat[i]) != task_cpu(p)
+		|| sqstat[i]->__state == TASK_DEAD))
+			sqstat[i] = NULL;
+	}
+
+	if (strncmp(p->comm, "iou-sqp", 7))
+		return;
+
+	if (!kcpustat_this_cpu->flag) {
+		for (int j = 0; j < MAX_SQ_NUM; j++)
+			kcpustat_this_cpu->sq_util[j] = NULL;
+		kcpustat_this_cpu->flag = true;
+	}
+	int index = MAX_SQ_NUM;
+	bool flag = true;
+
+	for (int i = 0; i < MAX_SQ_NUM; i++) {
+		if (sqstat[i] == p)
+			flag = false;
+		if (!sqstat[i] || task_cpu(sqstat[i]) != task_cpu(p)) {
+			sqstat[i] = NULL;
+			if (i < index)
+				index = i;
+		}
+	}
+	if (flag && index < MAX_SQ_NUM)
+		sqstat[index] = p;
+}
+
 /*
  * Account a single tick of CPU time.
  * @p: the process that the CPU time gets accounted to
@@ -487,7 +521,7 @@ void thread_group_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st)
 void account_process_tick(struct task_struct *p, int user_tick)
 {
 	u64 cputime, steal;
-
+	get_sqthread_util(p);
 	if (vtime_accounting_enabled_this_cpu())
 		return;
 
diff --git a/kernel/sched/pelt.c b/kernel/sched/pelt.c
index 0f310768260c..945efe80e08c 100644
--- a/kernel/sched/pelt.c
+++ b/kernel/sched/pelt.c
@@ -266,6 +266,20 @@ ___update_load_avg(struct sched_avg *sa, unsigned long load)
 	WRITE_ONCE(sa->util_avg, sa->util_sum / divider);
 }
 
+void __update_sq_avg_block(u64 now, struct sched_entity *se)
+{
+	if (___update_load_sum(now, &se->sq_avg, 0, 0, 0))
+		___update_load_avg(&se->sq_avg, se_weight(se));
+}
+
+void __update_sq_avg(u64 now, struct sched_entity *se)
+{
+	struct cfs_rq *qcfs_rq = cfs_rq_of(se);
+
+	if (___update_load_sum(now, &se->sq_avg, !!se->on_rq, se_runnable(se), qcfs_rq->curr == se))
+		___update_load_avg(&se->sq_avg, se_weight(se));
+}
+
 /*
  * sched_entity:
  *
-- 
2.34.1