[deliverable/linux.git] / kernel / user.c

/*
 * The "user cache".
 *
 * (C) Copyright 1991-2000 Linus Torvalds
 *
 * We have a per-user structure to keep track of how many
 * processes, files etc the user has claimed, in order to be
 * able to have per-user limits for system resources. 
 */

#include <linux/init.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/bitops.h>
#include <linux/key.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/user_namespace.h>

struct user_namespace init_user_ns = {
	.kref = {
		.refcount	= ATOMIC_INIT(2),
	},
	.root_user = &root_user,
};
EXPORT_SYMBOL_GPL(init_user_ns);

/*
 * UID task count cache, to get fast user lookup in "alloc_uid"
 * when changing user ID's (ie setuid() and friends).
 */

#define UIDHASH_MASK		(UIDHASH_SZ - 1)
#define __uidhashfn(uid)	(((uid >> UIDHASH_BITS) + uid) & UIDHASH_MASK)
#define uidhashentry(ns, uid)	((ns)->uidhash_table + __uidhashfn((uid)))

static struct kmem_cache *uid_cachep;

/*
 * The uidhash_lock is mostly taken from process context, but it is
 * occasionally also taken from softirq/tasklet context, when
 * task-structs get RCU-freed. Hence all locking must be softirq-safe.
 * But free_uid() is also called with local interrupts disabled, and running
 * local_bh_enable() with local interrupts disabled is an error - we'll run
 * softirq callbacks, and they can unconditionally enable interrupts, and
 * the caller of free_uid() didn't expect that..
 */
static DEFINE_SPINLOCK(uidhash_lock);

struct user_struct root_user = {
	.__count	= ATOMIC_INIT(1),
	.processes	= ATOMIC_INIT(1),
	.files		= ATOMIC_INIT(0),
	.sigpending	= ATOMIC_INIT(0),
	.locked_shm     = 0,
#ifdef CONFIG_USER_SCHED
	.tg		= &init_task_group,
#endif
};

/*
 * These routines must be called with the uidhash spinlock held!
 */
static void uid_hash_insert(struct user_struct *up, struct hlist_head *hashent)
{
	hlist_add_head(&up->uidhash_node, hashent);
}

static void uid_hash_remove(struct user_struct *up)
{
	hlist_del_init(&up->uidhash_node);
}

static 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_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(&root_task_group);
	if (IS_ERR(up->tg))
		rc = -ENOMEM;

	set_tg_uid(up);

	return rc;
}

static void sched_switch_user(struct task_struct *p)
{
	sched_move_task(p);
}

#else	/* CONFIG_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_USER_SCHED */

#if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS)

static struct kset *uids_kset; /* represents the /sys/kernel/uids/ directory */
static DEFINE_MUTEX(uids_mutex);

static inline void uids_mutex_lock(void)
{
	mutex_lock(&uids_mutex);
}

static inline void uids_mutex_unlock(void)
{
	mutex_unlock(&uids_mutex);
}

/* uid directory attributes */
#ifdef CONFIG_FAIR_GROUP_SCHED
static ssize_t cpu_shares_show(struct kobject *kobj,
			       struct kobj_attribute *attr,
			       char *buf)
{
	struct user_struct *up = container_of(kobj, struct user_struct, kobj);

	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
}

static ssize_t cpu_shares_store(struct kobject *kobj,
				struct kobj_attribute *attr,
				const char *buf, size_t size)
{
	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	unsigned long shares;
	int rc;

	sscanf(buf, "%lu", &shares);

	rc = sched_group_set_shares(up->tg, shares);

	return (rc ? rc : size);
}

static struct kobj_attribute cpu_share_attr =
	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
#endif

#ifdef CONFIG_RT_GROUP_SCHED
static ssize_t cpu_rt_runtime_show(struct kobject *kobj,
				   struct kobj_attribute *attr,
				   char *buf)
{
	struct user_struct *up = container_of(kobj, struct user_struct, kobj);

	return sprintf(buf, "%ld\n", sched_group_rt_runtime(up->tg));
}

static ssize_t cpu_rt_runtime_store(struct kobject *kobj,
				    struct kobj_attribute *attr,
				    const char *buf, size_t size)
{
	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	unsigned long rt_runtime;
	int rc;

	sscanf(buf, "%ld", &rt_runtime);

	rc = sched_group_set_rt_runtime(up->tg, rt_runtime);

	return (rc ? rc : size);
}

static struct kobj_attribute cpu_rt_runtime_attr =
	__ATTR(cpu_rt_runtime, 0644, cpu_rt_runtime_show, cpu_rt_runtime_store);

static ssize_t cpu_rt_period_show(struct kobject *kobj,
				   struct kobj_attribute *attr,
				   char *buf)
{
	struct user_struct *up = container_of(kobj, struct user_struct, kobj);

	return sprintf(buf, "%lu\n", sched_group_rt_period(up->tg));
}

static ssize_t cpu_rt_period_store(struct kobject *kobj,
				    struct kobj_attribute *attr,
				    const char *buf, size_t size)
{
	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	unsigned long rt_period;
	int rc;

	sscanf(buf, "%lu", &rt_period);

	rc = sched_group_set_rt_period(up->tg, rt_period);

	return (rc ? rc : size);
}

static struct kobj_attribute cpu_rt_period_attr =
	__ATTR(cpu_rt_period, 0644, cpu_rt_period_show, cpu_rt_period_store);
#endif

/* default attributes per uid directory */
static struct attribute *uids_attributes[] = {
#ifdef CONFIG_FAIR_GROUP_SCHED
	&cpu_share_attr.attr,
#endif
#ifdef CONFIG_RT_GROUP_SCHED
	&cpu_rt_runtime_attr.attr,
	&cpu_rt_period_attr.attr,
#endif
	NULL
};

/* the lifetime of user_struct is not managed by the core (now) */
static void uids_release(struct kobject *kobj)
{
	return;
}

static struct kobj_type uids_ktype = {
	.sysfs_ops = &kobj_sysfs_ops,
	.default_attrs = uids_attributes,
	.release = uids_release,
};

/* create /sys/kernel/uids/<uid>/cpu_share file for this user */
static int uids_user_create(struct user_struct *up)
{
	struct kobject *kobj = &up->kobj;
	int error;

	memset(kobj, 0, sizeof(struct kobject));
	kobj->kset = uids_kset;
	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	if (error) {
		kobject_put(kobj);
		goto done;
	}

	kobject_uevent(kobj, KOBJ_ADD);
done:
	return error;
}

/* create these entries in sysfs:
 * 	"/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_sysfs_init(void)
{
	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
	if (!uids_kset)
		return -ENOMEM;

	return uids_user_create(&root_user);
}

/* 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 *up = container_of(w, struct user_struct, work);
	unsigned long flags;
	int remove_user = 0;

	/* 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);
	}

	if (!remove_user)
		goto done;

	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	kobject_del(&up->kobj);
	kobject_put(&up->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_USER_SCHED && CONFIG_SYSFS */

int uids_sysfs_init(void) { return 0; }
static inline int uids_user_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

/*
 * Locate the user_struct for the passed UID.  If found, take a ref on it.  The
 * caller must undo that ref with free_uid().
 *
 * If the user_struct could not be found, return NULL.
 */
struct user_struct *find_user(uid_t uid)
{
	struct user_struct *ret;
	unsigned long flags;
	struct user_namespace *ns = current->nsproxy->user_ns;

	spin_lock_irqsave(&uidhash_lock, flags);
	ret = uid_hash_find(uid, uidhashentry(ns, uid));
	spin_unlock_irqrestore(&uidhash_lock, flags);
	return ret;
}

void free_uid(struct user_struct *up)
{
	unsigned long flags;

	if (!up)
		return;

	local_irq_save(flags);
	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)
{
	struct hlist_head *hashent = uidhashentry(ns, uid);
	struct user_struct *up, *new;

	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
	 * atomic.
	 */
	uids_mutex_lock();

	spin_lock_irq(&uidhash_lock);
	up = uid_hash_find(uid, hashent);
	spin_unlock_irq(&uidhash_lock);

	if (!up) {
		new = kmem_cache_zalloc(uid_cachep, GFP_KERNEL);
		if (!new)
			goto out_unlock;

		new->uid = uid;
		atomic_set(&new->__count, 1);

		if (sched_create_user(new) < 0)
			goto out_free_user;

		if (uids_user_create(new))
			goto out_destoy_sched;

		/*
		 * Before adding this, check whether we raced
		 * on adding the same user already..
		 */
		spin_lock_irq(&uidhash_lock);
		up = uid_hash_find(uid, hashent);
		if (up) {
			/* This case is not possible when CONFIG_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);
		} else {
			uid_hash_insert(new, hashent);
			up = new;
		}
		spin_unlock_irq(&uidhash_lock);

	}

	uids_mutex_unlock();

	return up;

out_destoy_sched:
	sched_destroy_user(new);
out_free_user:
	kmem_cache_free(uid_cachep, new);
out_unlock:
	uids_mutex_unlock();
	return NULL;
}

void switch_uid(struct user_struct *new_user)
{
	struct user_struct *old_user;

	/* What if a process setreuid()'s and this brings the
	 * new uid over his NPROC rlimit?  We can check this now
	 * cheaply with the new uid cache, so if it matters
	 * we should be checking for it.  -DaveM
	 */
	old_user = current->user;
	atomic_inc(&new_user->processes);
	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()
	 * doing a get_uid(p->user).. If that saw the old
	 * user value, we need to wait until it has exited
	 * its critical region before we can free the old
	 * structure.
	 */
	smp_mb();
	spin_unlock_wait(&current->sighand->siglock);

	free_uid(old_user);
	suid_keys(current);
}

#ifdef CONFIG_USER_NS
void release_uids(struct user_namespace *ns)
{
	int i;
	unsigned long flags;
	struct hlist_head *head;
	struct hlist_node *nd;

	spin_lock_irqsave(&uidhash_lock, flags);
	/*
	 * collapse the chains so that the user_struct-s will
	 * be still alive, but not in hashes. subsequent free_uid()
	 * will free them.
	 */
	for (i = 0; i < UIDHASH_SZ; i++) {
		head = ns->uidhash_table + i;
		while (!hlist_empty(head)) {
			nd = head->first;
			hlist_del_init(nd);
		}
	}
	spin_unlock_irqrestore(&uidhash_lock, flags);

	free_uid(ns->root_user);
}
#endif

static int __init uid_cache_init(void)
{
	int n;

	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
			0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);

	for(n = 0; n < UIDHASH_SZ; ++n)
		INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);

	/* Insert the root user immediately (init already runs as root) */
	spin_lock_irq(&uidhash_lock);
	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
	spin_unlock_irq(&uidhash_lock);

	return 0;
}

module_init(uid_cache_init);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* The "user cache".
	3	*
	4	* (C) Copyright 1991-2000 Linus Torvalds
	5	*
	6	* We have a per-user structure to keep track of how many
	7	* processes, files etc the user has claimed, in order to be
	8	* able to have per-user limits for system resources.
	9	*/
	10
	11	#include <linux/init.h>
	12	#include <linux/sched.h>
	13	#include <linux/slab.h>
	14	#include <linux/bitops.h>
	15	#include <linux/key.h>
4021cb27	16	#include <linux/interrupt.h>
acce292c CLG	17	#include <linux/module.h>
acce292c CLG	18	#include <linux/user_namespace.h>
1da177e4	19
aee16ce7 PE	20	struct user_namespace init_user_ns = {
	21	.kref = {
	22	.refcount = ATOMIC_INIT(2),
	23	},
	24	.root_user = &root_user,
	25	};
	26	EXPORT_SYMBOL_GPL(init_user_ns);
	27
1da177e4 LT	28	/*
	29	* UID task count cache, to get fast user lookup in "alloc_uid"
	30	* when changing user ID's (ie setuid() and friends).
	31	*/
	32
1da177e4 LT	33	#define UIDHASH_MASK (UIDHASH_SZ - 1)
1da177e4 LT	34	#define __uidhashfn(uid) (((uid >> UIDHASH_BITS) + uid) & UIDHASH_MASK)
acce292c	35	#define uidhashentry(ns, uid) ((ns)->uidhash_table + __uidhashfn((uid)))
1da177e4	36
e18b890b	37	static struct kmem_cache *uid_cachep;
4021cb27 IM	38
	39	/*
	40	* The uidhash_lock is mostly taken from process context, but it is
	41	* occasionally also taken from softirq/tasklet context, when
	42	* task-structs get RCU-freed. Hence all locking must be softirq-safe.
3fa97c9d AM	43	* But free_uid() is also called with local interrupts disabled, and running
	44	* local_bh_enable() with local interrupts disabled is an error - we'll run
	45	* softirq callbacks, and they can unconditionally enable interrupts, and
	46	* the caller of free_uid() didn't expect that..
4021cb27	47	*/
1da177e4 LT	48	static DEFINE_SPINLOCK(uidhash_lock);
	49
	50	struct user_struct root_user = {
	51	.__count = ATOMIC_INIT(1),
	52	.processes = ATOMIC_INIT(1),
	53	.files = ATOMIC_INIT(0),
	54	.sigpending = ATOMIC_INIT(0),
1da177e4	55	.locked_shm = 0,
052f1dc7	56	#ifdef CONFIG_USER_SCHED
4cf86d77	57	.tg = &init_task_group,
24e377a8	58	#endif
1da177e4 LT	59	};
1da177e4 LT	60
5cb350ba DG	61	/*
	62	* These routines must be called with the uidhash spinlock held!
	63	*/
40aeb400	64	static void uid_hash_insert(struct user_struct up, struct hlist_head hashent)
5cb350ba DG	65	{
	66	hlist_add_head(&up->uidhash_node, hashent);
	67	}
	68
40aeb400	69	static void uid_hash_remove(struct user_struct *up)
5cb350ba DG	70	{
	71	hlist_del_init(&up->uidhash_node);
	72	}
	73
40aeb400	74	static struct user_struct uid_hash_find(uid_t uid, struct hlist_head hashent)
5cb350ba DG	75	{
	76	struct user_struct *user;
	77	struct hlist_node *h;
	78
	79	hlist_for_each_entry(user, h, hashent, uidhash_node) {
	80	if (user->uid == uid) {
	81	atomic_inc(&user->__count);
	82	return user;
	83	}
	84	}
	85
	86	return NULL;
	87	}
	88
052f1dc7	89	#ifdef CONFIG_USER_SCHED
5cb350ba	90
24e377a8 SV	91	static void sched_destroy_user(struct user_struct *up)
	92	{
	93	sched_destroy_group(up->tg);
	94	}
	95
	96	static int sched_create_user(struct user_struct *up)
	97	{
	98	int rc = 0;
	99
eff766a6	100	up->tg = sched_create_group(&root_task_group);
24e377a8 SV	101	if (IS_ERR(up->tg))
	102	rc = -ENOMEM;
	103
6c415b92 AB	104	set_tg_uid(up);
6c415b92 AB	105
24e377a8 SV	106	return rc;
	107	}
	108
	109	static void sched_switch_user(struct task_struct *p)
	110	{
	111	sched_move_task(p);
	112	}
	113
052f1dc7	114	#else /* CONFIG_USER_SCHED */
b1a8c172 DG	115
	116	static void sched_destroy_user(struct user_struct *up) { }
	117	static int sched_create_user(struct user_struct *up) { return 0; }
	118	static void sched_switch_user(struct task_struct *p) { }
	119
052f1dc7	120	#endif /* CONFIG_USER_SCHED */
b1a8c172	121
052f1dc7	122	#if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS)
b1a8c172	123
eb41d946	124	static struct kset uids_kset; / represents the /sys/kernel/uids/ directory */
b1a8c172 DG	125	static DEFINE_MUTEX(uids_mutex);
b1a8c172 DG	126
5cb350ba DG	127	static inline void uids_mutex_lock(void)
	128	{
	129	mutex_lock(&uids_mutex);
	130	}
24e377a8	131
5cb350ba DG	132	static inline void uids_mutex_unlock(void)
	133	{
	134	mutex_unlock(&uids_mutex);
	135	}
24e377a8	136
eb41d946	137	/* uid directory attributes */
052f1dc7	138	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946 KS	139	static ssize_t cpu_shares_show(struct kobject *kobj,
	140	struct kobj_attribute *attr,
	141	char *buf)
5cb350ba	142	{
eb41d946	143	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
24e377a8	144
eb41d946	145	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
5cb350ba DG	146	}
5cb350ba DG	147
eb41d946 KS	148	static ssize_t cpu_shares_store(struct kobject *kobj,
	149	struct kobj_attribute *attr,
	150	const char *buf, size_t size)
5cb350ba	151	{
eb41d946	152	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
5cb350ba DG	153	unsigned long shares;
	154	int rc;
	155
eb41d946	156	sscanf(buf, "%lu", &shares);
5cb350ba DG	157
	158	rc = sched_group_set_shares(up->tg, shares);
	159
	160	return (rc ? rc : size);
	161	}
	162
eb41d946 KS	163	static struct kobj_attribute cpu_share_attr =
eb41d946 KS	164	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
052f1dc7	165	#endif
eb41d946	166
052f1dc7	167	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56 PZ	168	static ssize_t cpu_rt_runtime_show(struct kobject *kobj,
	169	struct kobj_attribute *attr,
	170	char *buf)
	171	{
	172	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	173
af4491e5	174	return sprintf(buf, "%ld\n", sched_group_rt_runtime(up->tg));
9f0c1e56 PZ	175	}
	176
	177	static ssize_t cpu_rt_runtime_store(struct kobject *kobj,
	178	struct kobj_attribute *attr,
	179	const char *buf, size_t size)
	180	{
	181	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	182	unsigned long rt_runtime;
	183	int rc;
	184
af4491e5	185	sscanf(buf, "%ld", &rt_runtime);
9f0c1e56 PZ	186
	187	rc = sched_group_set_rt_runtime(up->tg, rt_runtime);
	188
	189	return (rc ? rc : size);
	190	}
	191
	192	static struct kobj_attribute cpu_rt_runtime_attr =
	193	__ATTR(cpu_rt_runtime, 0644, cpu_rt_runtime_show, cpu_rt_runtime_store);
d0b27fa7 PZ	194
	195	static ssize_t cpu_rt_period_show(struct kobject *kobj,
	196	struct kobj_attribute *attr,
	197	char *buf)
	198	{
	199	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	200
	201	return sprintf(buf, "%lu\n", sched_group_rt_period(up->tg));
	202	}
	203
	204	static ssize_t cpu_rt_period_store(struct kobject *kobj,
	205	struct kobj_attribute *attr,
	206	const char *buf, size_t size)
	207	{
	208	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	209	unsigned long rt_period;
	210	int rc;
	211
	212	sscanf(buf, "%lu", &rt_period);
	213
	214	rc = sched_group_set_rt_period(up->tg, rt_period);
	215
	216	return (rc ? rc : size);
	217	}
	218
	219	static struct kobj_attribute cpu_rt_period_attr =
	220	__ATTR(cpu_rt_period, 0644, cpu_rt_period_show, cpu_rt_period_store);
052f1dc7	221	#endif
9f0c1e56	222
eb41d946 KS	223	/* default attributes per uid directory */
eb41d946 KS	224	static struct attribute *uids_attributes[] = {
052f1dc7	225	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946	226	&cpu_share_attr.attr,
052f1dc7 PZ	227	#endif
052f1dc7 PZ	228	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56	229	&cpu_rt_runtime_attr.attr,
d0b27fa7	230	&cpu_rt_period_attr.attr,
052f1dc7	231	#endif
eb41d946 KS	232	NULL
	233	};
	234
	235	/* the lifetime of user_struct is not managed by the core (now) */
	236	static void uids_release(struct kobject *kobj)
5cb350ba	237	{
eb41d946	238	return;
5cb350ba DG	239	}
5cb350ba DG	240
eb41d946 KS	241	static struct kobj_type uids_ktype = {
	242	.sysfs_ops = &kobj_sysfs_ops,
	243	.default_attrs = uids_attributes,
	244	.release = uids_release,
	245	};
	246
	247	/* create /sys/kernel/uids/<uid>/cpu_share file for this user */
	248	static int uids_user_create(struct user_struct *up)
1da177e4	249	{
eb41d946	250	struct kobject *kobj = &up->kobj;
5cb350ba DG	251	int error;
5cb350ba DG	252
eb41d946	253	memset(kobj, 0, sizeof(struct kobject));
eb41d946	254	kobj->kset = uids_kset;
cf15126b GKH	255	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	256	if (error) {
	257	kobject_put(kobj);
5cb350ba	258	goto done;
cf15126b	259	}
5cb350ba	260
fb7dde37	261	kobject_uevent(kobj, KOBJ_ADD);
5cb350ba DG	262	done:
5cb350ba DG	263	return error;
1da177e4 LT	264	}
1da177e4 LT	265
eb41d946	266	/* create these entries in sysfs:
5cb350ba DG	267	* "/sys/kernel/uids" directory
	268	* "/sys/kernel/uids/0" directory (for root user)
	269	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	270	*/
eb41d946	271	int __init uids_sysfs_init(void)
1da177e4	272	{
0ff21e46	273	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
eb41d946 KS	274	if (!uids_kset)
eb41d946 KS	275	return -ENOMEM;
5cb350ba	276
eb41d946	277	return uids_user_create(&root_user);
1da177e4 LT	278	}
1da177e4 LT	279
5cb350ba DG	280	/* work function to remove sysfs directory for a user and free up
	281	* corresponding structures.
	282	*/
	283	static void remove_user_sysfs_dir(struct work_struct *w)
1da177e4	284	{
5cb350ba	285	struct user_struct *up = container_of(w, struct user_struct, work);
5cb350ba DG	286	unsigned long flags;
5cb350ba DG	287	int remove_user = 0;
1da177e4	288
5cb350ba DG	289	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	290	* atomic.
	291	*/
	292	uids_mutex_lock();
	293
	294	local_irq_save(flags);
	295
	296	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
	297	uid_hash_remove(up);
	298	remove_user = 1;
	299	spin_unlock_irqrestore(&uidhash_lock, flags);
	300	} else {
	301	local_irq_restore(flags);
1da177e4 LT	302	}
1da177e4 LT	303
5cb350ba DG	304	if (!remove_user)
	305	goto done;
	306
eb41d946 KS	307	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	308	kobject_del(&up->kobj);
	309	kobject_put(&up->kobj);
5cb350ba DG	310
	311	sched_destroy_user(up);
	312	key_put(up->uid_keyring);
	313	key_put(up->session_keyring);
	314	kmem_cache_free(uid_cachep, up);
	315
	316	done:
	317	uids_mutex_unlock();
	318	}
	319
	320	/* IRQs are disabled and uidhash_lock is held upon function entry.
	321	* IRQ state (as stored in flags) is restored and uidhash_lock released
	322	* upon function exit.
	323	*/
	324	static inline void free_user(struct user_struct *up, unsigned long flags)
	325	{
	326	/* restore back the count */
	327	atomic_inc(&up->__count);
	328	spin_unlock_irqrestore(&uidhash_lock, flags);
	329
	330	INIT_WORK(&up->work, remove_user_sysfs_dir);
	331	schedule_work(&up->work);
1da177e4 LT	332	}
1da177e4 LT	333
052f1dc7	334	#else /* CONFIG_USER_SCHED && CONFIG_SYSFS */
5cb350ba	335
eb41d946 KS	336	int uids_sysfs_init(void) { return 0; }
eb41d946 KS	337	static inline int uids_user_create(struct user_struct *up) { return 0; }
5cb350ba DG	338	static inline void uids_mutex_lock(void) { }
	339	static inline void uids_mutex_unlock(void) { }
	340
	341	/* IRQs are disabled and uidhash_lock is held upon function entry.
	342	* IRQ state (as stored in flags) is restored and uidhash_lock released
	343	* upon function exit.
	344	*/
	345	static inline void free_user(struct user_struct *up, unsigned long flags)
	346	{
	347	uid_hash_remove(up);
	348	spin_unlock_irqrestore(&uidhash_lock, flags);
	349	sched_destroy_user(up);
	350	key_put(up->uid_keyring);
	351	key_put(up->session_keyring);
	352	kmem_cache_free(uid_cachep, up);
	353	}
	354
b1a8c172	355	#endif
5cb350ba	356
1da177e4 LT	357	/*
	358	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	359	* caller must undo that ref with free_uid().
	360	*
	361	* If the user_struct could not be found, return NULL.
	362	*/
	363	struct user_struct *find_user(uid_t uid)
	364	{
	365	struct user_struct *ret;
3fa97c9d	366	unsigned long flags;
acce292c	367	struct user_namespace *ns = current->nsproxy->user_ns;
1da177e4	368
3fa97c9d	369	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	370	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	371	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	372	return ret;
	373	}
	374
	375	void free_uid(struct user_struct *up)
	376	{
3fa97c9d AM	377	unsigned long flags;
3fa97c9d AM	378
36f57413 AM	379	if (!up)
	380	return;
	381
3fa97c9d	382	local_irq_save(flags);
5cb350ba DG	383	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	384	free_user(up, flags);
	385	else
36f57413	386	local_irq_restore(flags);
1da177e4 LT	387	}
1da177e4 LT	388
354a1f4d	389	struct user_struct alloc_uid(struct user_namespace ns, uid_t uid)
1da177e4	390	{
735de223	391	struct hlist_head *hashent = uidhashentry(ns, uid);
8eb703e4	392	struct user_struct up, new;
1da177e4	393
eb41d946	394	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
5cb350ba DG	395	* atomic.
	396	*/
	397	uids_mutex_lock();
	398
3fa97c9d	399	spin_lock_irq(&uidhash_lock);
1da177e4	400	up = uid_hash_find(uid, hashent);
3fa97c9d	401	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	402
1da177e4 LT	403	if (!up) {
354a1f4d	404	new = kmem_cache_zalloc(uid_cachep, GFP_KERNEL);
8eb703e4 PE	405	if (!new)
8eb703e4 PE	406	goto out_unlock;
5e8869bb	407
1da177e4 LT	408	new->uid = uid;
1da177e4 LT	409	atomic_set(&new->__count, 1);
1da177e4	410
8eb703e4	411	if (sched_create_user(new) < 0)
69664cf1	412	goto out_free_user;
24e377a8	413
8eb703e4 PE	414	if (uids_user_create(new))
8eb703e4 PE	415	goto out_destoy_sched;
5cb350ba	416
1da177e4 LT	417	/*
	418	* Before adding this, check whether we raced
	419	* on adding the same user already..
	420	*/
3fa97c9d	421	spin_lock_irq(&uidhash_lock);
1da177e4 LT	422	up = uid_hash_find(uid, hashent);
1da177e4 LT	423	if (up) {
052f1dc7	424	/* This case is not possible when CONFIG_USER_SCHED
5cb350ba DG	425	* is defined, since we serialize alloc_uid() using
	426	* uids_mutex. Hence no need to call
	427	* sched_destroy_user() or remove_user_sysfs_dir().
	428	*/
1da177e4 LT	429	key_put(new->uid_keyring);
	430	key_put(new->session_keyring);
	431	kmem_cache_free(uid_cachep, new);
	432	} else {
	433	uid_hash_insert(new, hashent);
	434	up = new;
	435	}
3fa97c9d	436	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	437
1da177e4 LT	438	}
5cb350ba DG	439
	440	uids_mutex_unlock();
	441
1da177e4	442	return up;
8eb703e4 PE	443
	444	out_destoy_sched:
	445	sched_destroy_user(new);
8eb703e4 PE	446	out_free_user:
	447	kmem_cache_free(uid_cachep, new);
	448	out_unlock:
	449	uids_mutex_unlock();
	450	return NULL;
1da177e4 LT	451	}
	452
	453	void switch_uid(struct user_struct *new_user)
	454	{
	455	struct user_struct *old_user;
	456
	457	/* What if a process setreuid()'s and this brings the
	458	* new uid over his NPROC rlimit? We can check this now
	459	* cheaply with the new uid cache, so if it matters
	460	* we should be checking for it. -DaveM
	461	*/
	462	old_user = current->user;
	463	atomic_inc(&new_user->processes);
	464	atomic_dec(&old_user->processes);
	465	switch_uid_keyring(new_user);
	466	current->user = new_user;
24e377a8	467	sched_switch_user(current);
45c18b0b LT	468
	469	/*
	470	* We need to synchronize with __sigqueue_alloc()
	471	* doing a get_uid(p->user).. If that saw the old
	472	* user value, we need to wait until it has exited
	473	* its critical region before we can free the old
	474	* structure.
	475	*/
	476	smp_mb();
	477	spin_unlock_wait(&current->sighand->siglock);
	478
1da177e4 LT	479	free_uid(old_user);
	480	suid_keys(current);
	481	}
	482
aee16ce7	483	#ifdef CONFIG_USER_NS
28f300d2 PE	484	void release_uids(struct user_namespace *ns)
	485	{
	486	int i;
	487	unsigned long flags;
	488	struct hlist_head *head;
	489	struct hlist_node *nd;
	490
	491	spin_lock_irqsave(&uidhash_lock, flags);
	492	/*
	493	* collapse the chains so that the user_struct-s will
	494	* be still alive, but not in hashes. subsequent free_uid()
	495	* will free them.
	496	*/
	497	for (i = 0; i < UIDHASH_SZ; i++) {
	498	head = ns->uidhash_table + i;
	499	while (!hlist_empty(head)) {
	500	nd = head->first;
	501	hlist_del_init(nd);
	502	}
	503	}
	504	spin_unlock_irqrestore(&uidhash_lock, flags);
	505
	506	free_uid(ns->root_user);
	507	}
aee16ce7	508	#endif
1da177e4 LT	509
	510	static int __init uid_cache_init(void)
	511	{
	512	int n;
	513
	514	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	515	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	516
1da177e4 LT	517	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	518	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	519
1da177e4 LT	520	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	521	spin_lock_irq(&uidhash_lock);
acce292c	522	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	523	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	524
	525	return 0;
	526	}
	527
	528	module_init(uid_cache_init);