[RESEND RFC PATCH 2/3] Publish tasks's scheduler stats thru the shared structure

[Prakash Sangappa]

Define a 'struct task_schedstat' which contains members corresponding to
scheduler stats that are currently available thru /proc/pid/tasks/pid
/schedstats. Update scheduler stats in this structure in kernel at the
same time stats in the 'struct task_struct' are updated. Add a
TASK_SCHEDSTAT option to task_getshared system call to request these per
thread scheduler stats thru the shared structure.

Signed-off-by: Prakash Sangappa <prakash.sangappa@xxxxxxxxxx>
---
 include/linux/task_shared.h      | 35 ++++++++++++++++++++++++++++++++++-
 include/uapi/linux/task_shared.h | 22 ++++++++++++++++++++++
 kernel/sched/deadline.c          |  1 +
 kernel/sched/fair.c              |  1 +
 kernel/sched/rt.c                |  1 +
 kernel/sched/sched.h             |  1 +
 kernel/sched/stats.h             |  3 +++
 kernel/sched/stop_task.c         |  1 +
 mm/task_shared.c                 | 13 +++++++++++++
 9 files changed, 77 insertions(+), 1 deletion(-)
 create mode 100644 include/uapi/linux/task_shared.h

diff --git a/include/linux/task_shared.h b/include/linux/task_shared.h
index de17849..62793e4 100644
--- a/include/linux/task_shared.h
+++ b/include/linux/task_shared.h
@@ -3,6 +3,7 @@
 #define	__TASK_SHARED_H__
 
 #include <linux/mm_types.h>
+#include <uapi/linux/task_shared.h>
 
 /*
  * Track user-kernel shared pages referred by mm_struct
@@ -18,7 +19,7 @@ struct ushared_pages {
  * fast communication.
  */
 struct task_ushared {
-	long version;
+	struct task_schedstat ts;
 };
 
 /*
@@ -52,6 +53,38 @@ struct task_ushrd_struct {
 	struct ushared_pg *upg;
 };
 
+
+#ifdef CONFIG_SCHED_INFO
+
+#define task_update_exec_runtime(t)					\
+	do {								\
+		struct task_ushrd_struct *shrdp = t->task_ushrd;	\
+		if (shrdp != NULL && shrdp->kaddr != NULL)		\
+			shrdp->kaddr->ts.sum_exec_runtime =		\
+				 t->se.sum_exec_runtime;		\
+	} while (0)
+
+#define task_update_runq_stat(t, p)					\
+	do {								\
+		struct task_ushrd_struct *shrdp = t->task_ushrd;	\
+		if (shrdp != NULL && shrdp->kaddr != NULL) {		\
+			shrdp->kaddr->ts.run_delay =			\
+				 t->sched_info.run_delay;		\
+			if (p) {					\
+				shrdp->kaddr->ts.pcount =		\
+					 t->sched_info.pcount;		\
+			}						\
+		}							\
+	} while (0)
+#else
+
+#define task_update_exec_runtime(t)	do { } while (0)
+#define task_update_runq_stat(t, p)	do { } while (0)
+
+#endif
+
+
+
 extern void task_ushared_free(struct task_struct *t);
 extern void mm_ushared_clear(struct mm_struct *mm);
 #endif /* __TASK_SHARED_H__ */
diff --git a/include/uapi/linux/task_shared.h b/include/uapi/linux/task_shared.h
new file mode 100644
index 0000000..06a8522
--- /dev/null
+++ b/include/uapi/linux/task_shared.h
@@ -0,0 +1,22 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+#ifndef LINUX_TASK_SHARED_H
+#define LINUX_TASK_SHARED_H
+
+/*
+ * Per task user-kernel mapped structure for faster communication.
+ */
+
+/*
+ * Following is the option to request struct task_schedstats shared structure,
+ * in which kernel shares the task's exec time and time on run queue & number
+ * of times it was scheduled to run on a cpu. Requires kernel with
+ * CONFIG_SCHED_INFO enabled.
+ */
+#define TASK_SCHEDSTAT 1
+
+struct task_schedstat {
+	volatile u64	sum_exec_runtime;
+	volatile u64	run_delay;
+	volatile u64	pcount;
+};
+#endif
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index aaacd6c..189c74c 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1270,6 +1270,7 @@ static void update_curr_dl(struct rq *rq)
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
+	task_update_exec_runtime(curr);
 
 	curr->se.exec_start = now;
 	cgroup_account_cputime(curr, delta_exec);
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 44c4520..cbd182b 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -817,6 +817,7 @@ static void update_curr(struct cfs_rq *cfs_rq)
 	if (entity_is_task(curr)) {
 		struct task_struct *curtask = task_of(curr);
 
+		task_update_exec_runtime(curtask);
 		trace_sched_stat_runtime(curtask, delta_exec, curr->vruntime);
 		cgroup_account_cputime(curtask, delta_exec);
 		account_group_exec_runtime(curtask, delta_exec);
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 3daf42a..61082fc 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1014,6 +1014,7 @@ static void update_curr_rt(struct rq *rq)
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
+	task_update_exec_runtime(curr);
 
 	curr->se.exec_start = now;
 	cgroup_account_cputime(curr, delta_exec);
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 14a41a2..4ebbd8f 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -67,6 +67,7 @@
 #include <linux/syscalls.h>
 #include <linux/task_work.h>
 #include <linux/tsacct_kern.h>
+#include <linux/task_shared.h>
 
 #include <asm/tlb.h>
 
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index d8f8eb0..6b2d69c 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 
 #ifdef CONFIG_SCHEDSTATS
+#include <linux/task_shared.h>
 
 /*
  * Expects runqueue lock to be held for atomicity of update
@@ -166,6 +167,7 @@ static inline void sched_info_dequeue(struct rq *rq, struct task_struct *t)
 	delta = rq_clock(rq) - t->sched_info.last_queued;
 	t->sched_info.last_queued = 0;
 	t->sched_info.run_delay += delta;
+	task_update_runq_stat(t, 0);
 
 	rq_sched_info_dequeue(rq, delta);
 }
@@ -188,6 +190,7 @@ static void sched_info_arrive(struct rq *rq, struct task_struct *t)
 	t->sched_info.run_delay += delta;
 	t->sched_info.last_arrival = now;
 	t->sched_info.pcount++;
+	task_update_runq_stat(t, 1);
 
 	rq_sched_info_arrive(rq, delta);
 }
diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c
index f988ebe..7b9b60f 100644
--- a/kernel/sched/stop_task.c
+++ b/kernel/sched/stop_task.c
@@ -82,6 +82,7 @@ static void put_prev_task_stop(struct rq *rq, struct task_struct *prev)
 			max(curr->se.statistics.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
+	task_update_exec_runtime(curr);
 	account_group_exec_runtime(curr, delta_exec);
 
 	curr->se.exec_start = rq_clock_task(rq);
diff --git a/mm/task_shared.c b/mm/task_shared.c
index 3ec5eb6..7169ccd 100644
--- a/mm/task_shared.c
+++ b/mm/task_shared.c
@@ -275,6 +275,14 @@ static long task_getshared(u64 opt, u64 flags, void __user *uaddr)
 {
 	struct task_ushrd_struct *ushrd = current->task_ushrd;
 
+	/* Currently only TASK_SCHEDSTAT supported */
+#ifdef CONFIG_SCHED_INFO
+	if (opt != TASK_SCHEDSTAT)
+		return (-EINVAL);
+#else
+	return (-EOPNOTSUPP);
+#endif
+
 	/* We have address, return. */
 	if (ushrd != NULL && ushrd->upg != NULL) {
 		if (copy_to_user(uaddr, &ushrd->uaddr,
@@ -286,6 +294,11 @@ static long task_getshared(u64 opt, u64 flags, void __user *uaddr)
 	task_ushared_alloc();
 	ushrd = current->task_ushrd;
 	if (ushrd != NULL && ushrd->upg != NULL) {
+		if (opt == TASK_SCHEDSTAT) {
+			/* init current values */
+			task_update_exec_runtime(current);
+			task_update_runq_stat(current, 1);
+		}
 		if (copy_to_user(uaddr, &ushrd->uaddr,
 			sizeof(struct task_ushared *)))
 			return (-EFAULT);
-- 
2.7.4