[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>

/*
 * 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_KEYS
	.uid_keyring	= &root_user_keyring,
	.session_keyring = &root_session_keyring,
#endif
#ifdef CONFIG_FAIR_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_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 */

#if defined(CONFIG_FAIR_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 */
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);

/* default attributes per uid directory */
static struct attribute *uids_attributes[] = {
	&cpu_share_attr.attr,
	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_FAIR_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_alloc(uid_cachep, GFP_KERNEL);
		if (!new)
			goto out_unlock;

		new->uid = uid;
		atomic_set(&new->__count, 1);
		atomic_set(&new->processes, 0);
		atomic_set(&new->files, 0);
		atomic_set(&new->sigpending, 0);
#ifdef CONFIG_INOTIFY_USER
		atomic_set(&new->inotify_watches, 0);
		atomic_set(&new->inotify_devs, 0);
#endif
#ifdef CONFIG_POSIX_MQUEUE
		new->mq_bytes = 0;
#endif
		new->locked_shm = 0;

		if (alloc_uid_keyring(new, current) < 0)
			goto out_free_user;

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

		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_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);
		} 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_put_keys:
	key_put(new->uid_keyring);
	key_put(new->session_keyring);
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);
}

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);
}

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 LT	19
	20	/*
	21	* UID task count cache, to get fast user lookup in "alloc_uid"
	22	* when changing user ID's (ie setuid() and friends).
	23	*/
	24
1da177e4 LT	25	#define UIDHASH_MASK (UIDHASH_SZ - 1)
1da177e4 LT	26	#define __uidhashfn(uid) (((uid >> UIDHASH_BITS) + uid) & UIDHASH_MASK)
acce292c	27	#define uidhashentry(ns, uid) ((ns)->uidhash_table + __uidhashfn((uid)))
1da177e4	28
e18b890b	29	static struct kmem_cache *uid_cachep;
4021cb27 IM	30
	31	/*
	32	* The uidhash_lock is mostly taken from process context, but it is
	33	* occasionally also taken from softirq/tasklet context, when
	34	* task-structs get RCU-freed. Hence all locking must be softirq-safe.
3fa97c9d AM	35	* But free_uid() is also called with local interrupts disabled, and running
	36	* local_bh_enable() with local interrupts disabled is an error - we'll run
	37	* softirq callbacks, and they can unconditionally enable interrupts, and
	38	* the caller of free_uid() didn't expect that..
4021cb27	39	*/
1da177e4 LT	40	static DEFINE_SPINLOCK(uidhash_lock);
	41
	42	struct user_struct root_user = {
	43	.__count = ATOMIC_INIT(1),
	44	.processes = ATOMIC_INIT(1),
	45	.files = ATOMIC_INIT(0),
	46	.sigpending = ATOMIC_INIT(0),
1da177e4 LT	47	.locked_shm = 0,
	48	#ifdef CONFIG_KEYS
	49	.uid_keyring = &root_user_keyring,
	50	.session_keyring = &root_session_keyring,
	51	#endif
24e377a8	52	#ifdef CONFIG_FAIR_USER_SCHED
4cf86d77	53	.tg = &init_task_group,
24e377a8	54	#endif
1da177e4 LT	55	};
1da177e4 LT	56
5cb350ba DG	57	/*
	58	* These routines must be called with the uidhash spinlock held!
	59	*/
40aeb400	60	static void uid_hash_insert(struct user_struct up, struct hlist_head hashent)
5cb350ba DG	61	{
	62	hlist_add_head(&up->uidhash_node, hashent);
	63	}
	64
40aeb400	65	static void uid_hash_remove(struct user_struct *up)
5cb350ba DG	66	{
	67	hlist_del_init(&up->uidhash_node);
	68	}
	69
40aeb400	70	static struct user_struct uid_hash_find(uid_t uid, struct hlist_head hashent)
5cb350ba DG	71	{
	72	struct user_struct *user;
	73	struct hlist_node *h;
	74
	75	hlist_for_each_entry(user, h, hashent, uidhash_node) {
	76	if (user->uid == uid) {
	77	atomic_inc(&user->__count);
	78	return user;
	79	}
	80	}
	81
	82	return NULL;
	83	}
	84
24e377a8	85	#ifdef CONFIG_FAIR_USER_SCHED
5cb350ba	86
24e377a8 SV	87	static void sched_destroy_user(struct user_struct *up)
	88	{
	89	sched_destroy_group(up->tg);
	90	}
	91
	92	static int sched_create_user(struct user_struct *up)
	93	{
	94	int rc = 0;
	95
	96	up->tg = sched_create_group();
	97	if (IS_ERR(up->tg))
	98	rc = -ENOMEM;
	99
	100	return rc;
	101	}
	102
	103	static void sched_switch_user(struct task_struct *p)
	104	{
	105	sched_move_task(p);
	106	}
	107
b1a8c172 DG	108	#else /* CONFIG_FAIR_USER_SCHED */
	109
	110	static void sched_destroy_user(struct user_struct *up) { }
	111	static int sched_create_user(struct user_struct *up) { return 0; }
	112	static void sched_switch_user(struct task_struct *p) { }
	113
	114	#endif /* CONFIG_FAIR_USER_SCHED */
	115
	116	#if defined(CONFIG_FAIR_USER_SCHED) && defined(CONFIG_SYSFS)
	117
eb41d946	118	static struct kset uids_kset; / represents the /sys/kernel/uids/ directory */
b1a8c172 DG	119	static DEFINE_MUTEX(uids_mutex);
b1a8c172 DG	120
5cb350ba DG	121	static inline void uids_mutex_lock(void)
	122	{
	123	mutex_lock(&uids_mutex);
	124	}
24e377a8	125
5cb350ba DG	126	static inline void uids_mutex_unlock(void)
	127	{
	128	mutex_unlock(&uids_mutex);
	129	}
24e377a8	130
eb41d946 KS	131	/* uid directory attributes */
	132	static ssize_t cpu_shares_show(struct kobject *kobj,
	133	struct kobj_attribute *attr,
	134	char *buf)
5cb350ba	135	{
eb41d946	136	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
24e377a8	137
eb41d946	138	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
5cb350ba DG	139	}
5cb350ba DG	140
eb41d946 KS	141	static ssize_t cpu_shares_store(struct kobject *kobj,
	142	struct kobj_attribute *attr,
	143	const char *buf, size_t size)
5cb350ba	144	{
eb41d946	145	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
5cb350ba DG	146	unsigned long shares;
	147	int rc;
	148
eb41d946	149	sscanf(buf, "%lu", &shares);
5cb350ba DG	150
	151	rc = sched_group_set_shares(up->tg, shares);
	152
	153	return (rc ? rc : size);
	154	}
	155
eb41d946 KS	156	static struct kobj_attribute cpu_share_attr =
	157	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
	158
	159	/* default attributes per uid directory */
	160	static struct attribute *uids_attributes[] = {
	161	&cpu_share_attr.attr,
	162	NULL
	163	};
	164
	165	/* the lifetime of user_struct is not managed by the core (now) */
	166	static void uids_release(struct kobject *kobj)
5cb350ba	167	{
eb41d946	168	return;
5cb350ba DG	169	}
5cb350ba DG	170
eb41d946 KS	171	static struct kobj_type uids_ktype = {
	172	.sysfs_ops = &kobj_sysfs_ops,
	173	.default_attrs = uids_attributes,
	174	.release = uids_release,
	175	};
	176
	177	/* create /sys/kernel/uids/<uid>/cpu_share file for this user */
	178	static int uids_user_create(struct user_struct *up)
1da177e4	179	{
eb41d946	180	struct kobject *kobj = &up->kobj;
5cb350ba DG	181	int error;
5cb350ba DG	182
eb41d946	183	memset(kobj, 0, sizeof(struct kobject));
eb41d946	184	kobj->kset = uids_kset;
cf15126b GKH	185	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	186	if (error) {
	187	kobject_put(kobj);
5cb350ba	188	goto done;
cf15126b	189	}
5cb350ba	190
fb7dde37	191	kobject_uevent(kobj, KOBJ_ADD);
5cb350ba DG	192	done:
5cb350ba DG	193	return error;
1da177e4 LT	194	}
1da177e4 LT	195
eb41d946	196	/* create these entries in sysfs:
5cb350ba DG	197	* "/sys/kernel/uids" directory
	198	* "/sys/kernel/uids/0" directory (for root user)
	199	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	200	*/
eb41d946	201	int __init uids_sysfs_init(void)
1da177e4	202	{
0ff21e46	203	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
eb41d946 KS	204	if (!uids_kset)
eb41d946 KS	205	return -ENOMEM;
5cb350ba	206
eb41d946	207	return uids_user_create(&root_user);
1da177e4 LT	208	}
1da177e4 LT	209
5cb350ba DG	210	/* work function to remove sysfs directory for a user and free up
	211	* corresponding structures.
	212	*/
	213	static void remove_user_sysfs_dir(struct work_struct *w)
1da177e4	214	{
5cb350ba	215	struct user_struct *up = container_of(w, struct user_struct, work);
5cb350ba DG	216	unsigned long flags;
5cb350ba DG	217	int remove_user = 0;
1da177e4	218
5cb350ba DG	219	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	220	* atomic.
	221	*/
	222	uids_mutex_lock();
	223
	224	local_irq_save(flags);
	225
	226	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
	227	uid_hash_remove(up);
	228	remove_user = 1;
	229	spin_unlock_irqrestore(&uidhash_lock, flags);
	230	} else {
	231	local_irq_restore(flags);
1da177e4 LT	232	}
1da177e4 LT	233
5cb350ba DG	234	if (!remove_user)
	235	goto done;
	236
eb41d946 KS	237	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	238	kobject_del(&up->kobj);
	239	kobject_put(&up->kobj);
5cb350ba DG	240
	241	sched_destroy_user(up);
	242	key_put(up->uid_keyring);
	243	key_put(up->session_keyring);
	244	kmem_cache_free(uid_cachep, up);
	245
	246	done:
	247	uids_mutex_unlock();
	248	}
	249
	250	/* IRQs are disabled and uidhash_lock is held upon function entry.
	251	* IRQ state (as stored in flags) is restored and uidhash_lock released
	252	* upon function exit.
	253	*/
	254	static inline void free_user(struct user_struct *up, unsigned long flags)
	255	{
	256	/* restore back the count */
	257	atomic_inc(&up->__count);
	258	spin_unlock_irqrestore(&uidhash_lock, flags);
	259
	260	INIT_WORK(&up->work, remove_user_sysfs_dir);
	261	schedule_work(&up->work);
1da177e4 LT	262	}
1da177e4 LT	263
b1a8c172	264	#else /* CONFIG_FAIR_USER_SCHED && CONFIG_SYSFS */
5cb350ba	265
eb41d946 KS	266	int uids_sysfs_init(void) { return 0; }
eb41d946 KS	267	static inline int uids_user_create(struct user_struct *up) { return 0; }
5cb350ba DG	268	static inline void uids_mutex_lock(void) { }
	269	static inline void uids_mutex_unlock(void) { }
	270
	271	/* IRQs are disabled and uidhash_lock is held upon function entry.
	272	* IRQ state (as stored in flags) is restored and uidhash_lock released
	273	* upon function exit.
	274	*/
	275	static inline void free_user(struct user_struct *up, unsigned long flags)
	276	{
	277	uid_hash_remove(up);
	278	spin_unlock_irqrestore(&uidhash_lock, flags);
	279	sched_destroy_user(up);
	280	key_put(up->uid_keyring);
	281	key_put(up->session_keyring);
	282	kmem_cache_free(uid_cachep, up);
	283	}
	284
b1a8c172	285	#endif
5cb350ba	286
1da177e4 LT	287	/*
	288	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	289	* caller must undo that ref with free_uid().
	290	*
	291	* If the user_struct could not be found, return NULL.
	292	*/
	293	struct user_struct *find_user(uid_t uid)
	294	{
	295	struct user_struct *ret;
3fa97c9d	296	unsigned long flags;
acce292c	297	struct user_namespace *ns = current->nsproxy->user_ns;
1da177e4	298
3fa97c9d	299	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	300	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	301	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	302	return ret;
	303	}
	304
	305	void free_uid(struct user_struct *up)
	306	{
3fa97c9d AM	307	unsigned long flags;
3fa97c9d AM	308
36f57413 AM	309	if (!up)
	310	return;
	311
3fa97c9d	312	local_irq_save(flags);
5cb350ba DG	313	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	314	free_user(up, flags);
	315	else
36f57413	316	local_irq_restore(flags);
1da177e4 LT	317	}
1da177e4 LT	318
acce292c	319	struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
1da177e4	320	{
735de223	321	struct hlist_head *hashent = uidhashentry(ns, uid);
8eb703e4	322	struct user_struct up, new;
1da177e4	323
eb41d946	324	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
5cb350ba DG	325	* atomic.
	326	*/
	327	uids_mutex_lock();
	328
3fa97c9d	329	spin_lock_irq(&uidhash_lock);
1da177e4	330	up = uid_hash_find(uid, hashent);
3fa97c9d	331	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	332
1da177e4 LT	333	if (!up) {
e94b1766	334	new = kmem_cache_alloc(uid_cachep, GFP_KERNEL);
8eb703e4 PE	335	if (!new)
8eb703e4 PE	336	goto out_unlock;
5e8869bb	337
1da177e4 LT	338	new->uid = uid;
	339	atomic_set(&new->__count, 1);
	340	atomic_set(&new->processes, 0);
	341	atomic_set(&new->files, 0);
	342	atomic_set(&new->sigpending, 0);
2d9048e2	343	#ifdef CONFIG_INOTIFY_USER
0eeca283 RL	344	atomic_set(&new->inotify_watches, 0);
	345	atomic_set(&new->inotify_devs, 0);
	346	#endif
970a8645	347	#ifdef CONFIG_POSIX_MQUEUE
1da177e4	348	new->mq_bytes = 0;
970a8645	349	#endif
1da177e4 LT	350	new->locked_shm = 0;
1da177e4 LT	351
8eb703e4 PE	352	if (alloc_uid_keyring(new, current) < 0)
8eb703e4 PE	353	goto out_free_user;
1da177e4	354
8eb703e4 PE	355	if (sched_create_user(new) < 0)
8eb703e4 PE	356	goto out_put_keys;
24e377a8	357
8eb703e4 PE	358	if (uids_user_create(new))
8eb703e4 PE	359	goto out_destoy_sched;
5cb350ba	360
1da177e4 LT	361	/*
	362	* Before adding this, check whether we raced
	363	* on adding the same user already..
	364	*/
3fa97c9d	365	spin_lock_irq(&uidhash_lock);
1da177e4 LT	366	up = uid_hash_find(uid, hashent);
1da177e4 LT	367	if (up) {
5cb350ba DG	368	/* This case is not possible when CONFIG_FAIR_USER_SCHED
	369	* is defined, since we serialize alloc_uid() using
	370	* uids_mutex. Hence no need to call
	371	* sched_destroy_user() or remove_user_sysfs_dir().
	372	*/
1da177e4 LT	373	key_put(new->uid_keyring);
	374	key_put(new->session_keyring);
	375	kmem_cache_free(uid_cachep, new);
	376	} else {
	377	uid_hash_insert(new, hashent);
	378	up = new;
	379	}
3fa97c9d	380	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	381
1da177e4 LT	382	}
5cb350ba DG	383
	384	uids_mutex_unlock();
	385
1da177e4	386	return up;
8eb703e4 PE	387
	388	out_destoy_sched:
	389	sched_destroy_user(new);
	390	out_put_keys:
	391	key_put(new->uid_keyring);
	392	key_put(new->session_keyring);
	393	out_free_user:
	394	kmem_cache_free(uid_cachep, new);
	395	out_unlock:
	396	uids_mutex_unlock();
	397	return NULL;
1da177e4 LT	398	}
	399
	400	void switch_uid(struct user_struct *new_user)
	401	{
	402	struct user_struct *old_user;
	403
	404	/* What if a process setreuid()'s and this brings the
	405	* new uid over his NPROC rlimit? We can check this now
	406	* cheaply with the new uid cache, so if it matters
	407	* we should be checking for it. -DaveM
	408	*/
	409	old_user = current->user;
	410	atomic_inc(&new_user->processes);
	411	atomic_dec(&old_user->processes);
	412	switch_uid_keyring(new_user);
	413	current->user = new_user;
24e377a8	414	sched_switch_user(current);
45c18b0b LT	415
	416	/*
	417	* We need to synchronize with __sigqueue_alloc()
	418	* doing a get_uid(p->user).. If that saw the old
	419	* user value, we need to wait until it has exited
	420	* its critical region before we can free the old
	421	* structure.
	422	*/
	423	smp_mb();
	424	spin_unlock_wait(&current->sighand->siglock);
	425
1da177e4 LT	426	free_uid(old_user);
	427	suid_keys(current);
	428	}
	429
28f300d2 PE	430	void release_uids(struct user_namespace *ns)
	431	{
	432	int i;
	433	unsigned long flags;
	434	struct hlist_head *head;
	435	struct hlist_node *nd;
	436
	437	spin_lock_irqsave(&uidhash_lock, flags);
	438	/*
	439	* collapse the chains so that the user_struct-s will
	440	* be still alive, but not in hashes. subsequent free_uid()
	441	* will free them.
	442	*/
	443	for (i = 0; i < UIDHASH_SZ; i++) {
	444	head = ns->uidhash_table + i;
	445	while (!hlist_empty(head)) {
	446	nd = head->first;
	447	hlist_del_init(nd);
	448	}
	449	}
	450	spin_unlock_irqrestore(&uidhash_lock, flags);
	451
	452	free_uid(ns->root_user);
	453	}
1da177e4 LT	454
	455	static int __init uid_cache_init(void)
	456	{
	457	int n;
	458
	459	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	460	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	461
1da177e4 LT	462	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	463	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	464
1da177e4 LT	465	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	466	spin_lock_irq(&uidhash_lock);
acce292c	467	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	468	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	469
	470	return 0;
	471	}
	472
	473	module_init(uid_cache_init);