sched: add fair-user scheduler
authorSrivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Mon, 15 Oct 2007 15:00:09 +0000 (17:00 +0200)
committerIngo Molnar <mingo@elte.hu>
Mon, 15 Oct 2007 15:00:09 +0000 (17:00 +0200)
Enable user-id based fair group scheduling. This is useful for anyone
who wants to test the group scheduler w/o having to enable
CONFIG_CGROUPS.

A separate scheduling group (i.e struct task_grp) is automatically created for
every new user added to the system. Upon uid change for a task, it is made to
move to the corresponding scheduling group.

A /proc tunable (/proc/root_user_share) is also provided to tune root
user's quota of cpu bandwidth.

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>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
include/linux/sched.h
init/Kconfig
kernel/sched.c
kernel/sched_debug.c
kernel/user.c

index 03c13b6..d0cc583 100644 (file)
@@ -597,6 +597,10 @@ struct user_struct {
        /* Hash table maintenance information */
        struct hlist_node uidhash_node;
        uid_t uid;
+
+#ifdef CONFIG_FAIR_USER_SCHED
+       struct task_grp *tg;
+#endif
 };
 
 extern struct user_struct *find_user(uid_t);
index ef90a15..37711fe 100644 (file)
@@ -289,6 +289,19 @@ config FAIR_GROUP_SCHED
          This feature lets cpu scheduler recognize task groups and control cpu
          bandwidth allocation to such task groups.
 
+choice
+       depends on FAIR_GROUP_SCHED
+       prompt "Basis for grouping tasks"
+       default FAIR_USER_SCHED
+
+       config FAIR_USER_SCHED
+               bool "user id"
+               help
+                 This option will choose userid as the basis for grouping
+                 tasks, thus providing equal cpu bandwidth to each user.
+
+endchoice
+
 config SYSFS_DEPRECATED
        bool "Create deprecated sysfs files"
        default y
index e10c403..f33608e 100644 (file)
@@ -200,7 +200,12 @@ struct task_grp init_task_grp =  {
                                .cfs_rq = init_cfs_rq_p,
                                 };
 
+#ifdef CONFIG_FAIR_USER_SCHED
+#define INIT_TASK_GRP_LOAD     2*NICE_0_LOAD
+#else
 #define INIT_TASK_GRP_LOAD     NICE_0_LOAD
+#endif
+
 static int init_task_grp_load = INIT_TASK_GRP_LOAD;
 
 /* return group to which a task belongs */
@@ -208,7 +213,11 @@ static inline struct task_grp *task_grp(struct task_struct *p)
 {
        struct task_grp *tg;
 
+#ifdef CONFIG_FAIR_USER_SCHED
+       tg = p->user->tg;
+#else
        tg  = &init_task_grp;
+#endif
 
        return tg;
 }
index 3e47e87..57ee9d5 100644 (file)
@@ -212,6 +212,49 @@ static void sysrq_sched_debug_show(void)
        sched_debug_show(NULL, NULL);
 }
 
+#ifdef CONFIG_FAIR_USER_SCHED
+
+static DEFINE_MUTEX(root_user_share_mutex);
+
+static int
+root_user_share_read_proc(char *page, char **start, off_t off, int count,
+                                int *eof, void *data)
+{
+       int len;
+
+       len = sprintf(page, "%d\n", init_task_grp_load);
+
+       return len;
+}
+
+static int
+root_user_share_write_proc(struct file *file, const char __user *buffer,
+                                unsigned long count, void *data)
+{
+       unsigned long shares;
+       char kbuf[sizeof(unsigned long)+1];
+       int rc = 0;
+
+       if (copy_from_user(kbuf, buffer, sizeof(kbuf)))
+               return -EFAULT;
+
+       shares = simple_strtoul(kbuf, NULL, 0);
+
+       if (!shares)
+               shares = NICE_0_LOAD;
+
+       mutex_lock(&root_user_share_mutex);
+
+       init_task_grp_load = shares;
+       rc = sched_group_set_shares(&init_task_grp, shares);
+
+       mutex_unlock(&root_user_share_mutex);
+
+       return (rc < 0 ? rc : count);
+}
+
+#endif /* CONFIG_FAIR_USER_SCHED */
+
 static int sched_debug_open(struct inode *inode, struct file *filp)
 {
        return single_open(filp, sched_debug_show, NULL);
@@ -234,6 +277,15 @@ static int __init init_sched_debug_procfs(void)
 
        pe->proc_fops = &sched_debug_fops;
 
+#ifdef CONFIG_FAIR_USER_SCHED
+       pe = create_proc_entry("root_user_share", 0644, NULL);
+       if (!pe)
+               return -ENOMEM;
+
+       pe->read_proc = root_user_share_read_proc;
+       pe->write_proc = root_user_share_write_proc;
+#endif
+
        return 0;
 }
 
index 9ca2848..c6387fa 100644 (file)
@@ -50,8 +50,41 @@ struct user_struct root_user = {
        .uid_keyring    = &root_user_keyring,
        .session_keyring = &root_session_keyring,
 #endif
+#ifdef CONFIG_FAIR_USER_SCHED
+       .tg             = &init_task_grp,
+#endif
 };
 
+#ifdef CONFIG_FAIR_USER_SCHED
+static void sched_destroy_user(struct user_struct *up)
+{
+       sched_destroy_group(up->tg);
+}
+
+static int sched_create_user(struct user_struct *up)
+{
+       int rc = 0;
+
+       up->tg = sched_create_group();
+       if (IS_ERR(up->tg))
+               rc = -ENOMEM;
+
+       return rc;
+}
+
+static void sched_switch_user(struct task_struct *p)
+{
+       sched_move_task(p);
+}
+
+#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) { }
+
+#endif /* CONFIG_FAIR_USER_SCHED */
+
 /*
  * These routines must be called with the uidhash spinlock held!
  */
@@ -109,6 +142,7 @@ void free_uid(struct user_struct *up)
        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);
@@ -150,6 +184,13 @@ struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
                        return NULL;
                }
 
+               if (sched_create_user(new) < 0) {
+                       key_put(new->uid_keyring);
+                       key_put(new->session_keyring);
+                       kmem_cache_free(uid_cachep, new);
+                       return NULL;
+               }
+
                /*
                 * Before adding this, check whether we raced
                 * on adding the same user already..
@@ -157,6 +198,7 @@ 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);
                        key_put(new->uid_keyring);
                        key_put(new->session_keyring);
                        kmem_cache_free(uid_cachep, new);
@@ -184,6 +226,7 @@ void switch_uid(struct user_struct *new_user)
        atomic_dec(&old_user->processes);
        switch_uid_keyring(new_user);
        current->user = new_user;
+       sched_switch_user(current);
 
        /*
         * We need to synchronize with __sigqueue_alloc()