sched: group scheduling, sysfs tunables

Add tunables in sysfs to modify a user's cpu share.

A directory is created in sysfs for each new user in the system.

	/sys/kernel/uids/<uid>/cpu_share

Reading this file returns the cpu shares granted for the user.
Writing into this file modifies the cpu share for the user. Only an
administrator is allowed to modify a user's cpu share.

Ex:
	# cd /sys/kernel/uids/
	# cat 512/cpu_share
	1024
	# echo 2048 > 512/cpu_share
	# cat 512/cpu_share
	2048
	#

Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 210 insertions, 30 deletions

diff --git a/kernel/user.c b/kernel/user.c
index 0c9a7870d08f..74cadea8466f 100644
--- a/kernel/user.c
+++ b/kernel/user.c
@@ -55,7 +55,41 @@ struct user_struct root_user = {
 #endif
 };
 
+/*
+ * These routines must be called with the uidhash spinlock held!
+ */
+static inline void uid_hash_insert(struct user_struct *up,
+						struct hlist_head *hashent)
+{
+	hlist_add_head(&up->uidhash_node, hashent);
+}
+
+static inline void uid_hash_remove(struct user_struct *up)
+{
+	hlist_del_init(&up->uidhash_node);
+}
+
+static inline struct user_struct *uid_hash_find(uid_t uid,
+						struct hlist_head *hashent)
+{
+	struct user_struct *user;
+	struct hlist_node *h;
+
+	hlist_for_each_entry(user, h, hashent, uidhash_node) {
+		if (user->uid == uid) {
+			atomic_inc(&user->__count);
+			return user;
+		}
+	}
+
+	return NULL;
+}
+
 #ifdef CONFIG_FAIR_USER_SCHED
+
+static struct kobject uids_kobject; /* represents /sys/kernel/uids directory */
+static DEFINE_MUTEX(uids_mutex);
+
 static void sched_destroy_user(struct user_struct *up)
 {
 	sched_destroy_group(up->tg);
@@ -77,42 +111,173 @@ static void sched_switch_user(struct task_struct *p)
 	sched_move_task(p);
 }
 
-#else	/* CONFIG_FAIR_USER_SCHED */
+static inline void uids_mutex_lock(void)
+{
+	mutex_lock(&uids_mutex);
+}
 
-static void sched_destroy_user(struct user_struct *up) { }
-static int sched_create_user(struct user_struct *up) { return 0; }
-static void sched_switch_user(struct task_struct *p) { }
+static inline void uids_mutex_unlock(void)
+{
+	mutex_unlock(&uids_mutex);
+}
 
-#endif	/* CONFIG_FAIR_USER_SCHED */
+/* return cpu shares held by the user */
+ssize_t cpu_shares_show(struct kset *kset, char *buffer)
+{
+	struct user_struct *up = container_of(kset, struct user_struct, kset);
 
-/*
- * These routines must be called with the uidhash spinlock held!
+	return sprintf(buffer, "%lu\n", sched_group_shares(up->tg));
+}
+
+/* modify cpu shares held by the user */
+ssize_t cpu_shares_store(struct kset *kset, const char *buffer, size_t size)
+{
+	struct user_struct *up = container_of(kset, struct user_struct, kset);
+	unsigned long shares;
+	int rc;
+
+	sscanf(buffer, "%lu", &shares);
+
+	rc = sched_group_set_shares(up->tg, shares);
+
+	return (rc ? rc : size);
+}
+
+static void user_attr_init(struct subsys_attribute *sa, char *name, int mode)
+{
+	sa->attr.name = name;
+	sa->attr.mode = mode;
+	sa->show = cpu_shares_show;
+	sa->store = cpu_shares_store;
+}
+
+/* Create "/sys/kernel/uids/<uid>" directory and
+ *  "/sys/kernel/uids/<uid>/cpu_share" file for this user.
  */
-static inline void uid_hash_insert(struct user_struct *up, struct hlist_head *hashent)
+static int user_kobject_create(struct user_struct *up)
 {
-	hlist_add_head(&up->uidhash_node, hashent);
+	struct kset *kset = &up->kset;
+	struct kobject *kobj = &kset->kobj;
+	int error;
+
+	memset(kset, 0, sizeof(struct kset));
+	kobj->parent = &uids_kobject;	/* create under /sys/kernel/uids dir */
+	kobject_set_name(kobj, "%d", up->uid);
+	kset_init(kset);
+	user_attr_init(&up->user_attr, "cpu_share", 0644);
+
+	error = kobject_add(kobj);
+	if (error)
+		goto done;
+
+	error = sysfs_create_file(kobj, &up->user_attr.attr);
+	if (error)
+		kobject_del(kobj);
+
+done:
+	return error;
 }
 
-static inline void uid_hash_remove(struct user_struct *up)
+/* create these in sysfs filesystem:
+ * 	"/sys/kernel/uids" directory
+ * 	"/sys/kernel/uids/0" directory (for root user)
+ * 	"/sys/kernel/uids/0/cpu_share" file (for root user)
+ */
+int __init uids_kobject_init(void)
 {
-	hlist_del_init(&up->uidhash_node);
+	int error;
+
+	/* create under /sys/kernel dir */
+	uids_kobject.parent = &kernel_subsys.kobj;
+	kobject_set_name(&uids_kobject, "uids");
+	kobject_init(&uids_kobject);
+
+	error = kobject_add(&uids_kobject);
+	if (!error)
+		error = user_kobject_create(&root_user);
+
+	return error;
 }
 
-static inline struct user_struct *uid_hash_find(uid_t uid, struct hlist_head *hashent)
+/* work function to remove sysfs directory for a user and free up
+ * corresponding structures.
+ */
+static void remove_user_sysfs_dir(struct work_struct *w)
 {
-	struct user_struct *user;
-	struct hlist_node *h;
+	struct user_struct *up = container_of(w, struct user_struct, work);
+	struct kobject *kobj = &up->kset.kobj;
+	unsigned long flags;
+	int remove_user = 0;
 
-	hlist_for_each_entry(user, h, hashent, uidhash_node) {
-		if(user->uid == uid) {
-			atomic_inc(&user->__count);
-			return user;
-		}
+	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
+	 * atomic.
+	 */
+	uids_mutex_lock();
+
+	local_irq_save(flags);
+
+	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
+		uid_hash_remove(up);
+		remove_user = 1;
+		spin_unlock_irqrestore(&uidhash_lock, flags);
+	} else {
+		local_irq_restore(flags);
 	}
 
-	return NULL;
+	if (!remove_user)
+		goto done;
+
+	sysfs_remove_file(kobj, &up->user_attr.attr);
+	kobject_del(kobj);
+
+	sched_destroy_user(up);
+	key_put(up->uid_keyring);
+	key_put(up->session_keyring);
+	kmem_cache_free(uid_cachep, up);
+
+done:
+	uids_mutex_unlock();
+}
+
+/* IRQs are disabled and uidhash_lock is held upon function entry.
+ * IRQ state (as stored in flags) is restored and uidhash_lock released
+ * upon function exit.
+ */
+static inline void free_user(struct user_struct *up, unsigned long flags)
+{
+	/* restore back the count */
+	atomic_inc(&up->__count);
+	spin_unlock_irqrestore(&uidhash_lock, flags);
+
+	INIT_WORK(&up->work, remove_user_sysfs_dir);
+	schedule_work(&up->work);
 }
 
+#else	/* CONFIG_FAIR_USER_SCHED */
+
+static void sched_destroy_user(struct user_struct *up) { }
+static int sched_create_user(struct user_struct *up) { return 0; }
+static void sched_switch_user(struct task_struct *p) { }
+static inline int user_kobject_create(struct user_struct *up) { return 0; }
+static inline void uids_mutex_lock(void) { }
+static inline void uids_mutex_unlock(void) { }
+
+/* IRQs are disabled and uidhash_lock is held upon function entry.
+ * IRQ state (as stored in flags) is restored and uidhash_lock released
+ * upon function exit.
+ */
+static inline void free_user(struct user_struct *up, unsigned long flags)
+{
+	uid_hash_remove(up);
+	spin_unlock_irqrestore(&uidhash_lock, flags);
+	sched_destroy_user(up);
+	key_put(up->uid_keyring);
+	key_put(up->session_keyring);
+	kmem_cache_free(uid_cachep, up);
+}
+
+#endif	/* CONFIG_FAIR_USER_SCHED */
+
 /*
  * Locate the user_struct for the passed UID.  If found, take a ref on it.  The
  * caller must undo that ref with free_uid().
@@ -139,16 +304,10 @@ void free_uid(struct user_struct *up)
 		return;
 
 	local_irq_save(flags);
-	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
-		uid_hash_remove(up);
-		spin_unlock_irqrestore(&uidhash_lock, flags);
-		sched_destroy_user(up);
-		key_put(up->uid_keyring);
-		key_put(up->session_keyring);
-		kmem_cache_free(uid_cachep, up);
-	} else {
+	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
+		free_user(up, flags);
+	else
 		local_irq_restore(flags);
-	}
 }
 
 struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
@@ -156,6 +315,11 @@ struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
 	struct hlist_head *hashent = uidhashentry(ns, uid);
 	struct user_struct *up;
 
+	/* Make uid_hash_find() + user_kobject_create() + uid_hash_insert()
+	 * atomic.
+	 */
+	uids_mutex_lock();
+
 	spin_lock_irq(&uidhash_lock);
 	up = uid_hash_find(uid, hashent);
 	spin_unlock_irq(&uidhash_lock);
@@ -191,6 +355,15 @@ struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
 			return NULL;
 		}
 
+		if (user_kobject_create(new)) {
+			sched_destroy_user(new);
+			key_put(new->uid_keyring);
+			key_put(new->session_keyring);
+			kmem_cache_free(uid_cachep, new);
+			uids_mutex_unlock();
+			return NULL;
+		}
+
 		/*
 		 * Before adding this, check whether we raced
 		 * on adding the same user already..
@@ -198,7 +371,11 @@ struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
 		spin_lock_irq(&uidhash_lock);
 		up = uid_hash_find(uid, hashent);
 		if (up) {
-			sched_destroy_user(new);
+			/* This case is not possible when CONFIG_FAIR_USER_SCHED
+			 * is defined, since we serialize alloc_uid() using
+			 * uids_mutex. Hence no need to call
+			 * sched_destroy_user() or remove_user_sysfs_dir().
+			 */
 			key_put(new->uid_keyring);
 			key_put(new->session_keyring);
 			kmem_cache_free(uid_cachep, new);
@@ -209,6 +386,9 @@ struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
 		spin_unlock_irq(&uidhash_lock);
 
 	}
+
+	uids_mutex_unlock();
+
 	return up;
 }