[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 kobject uids_kobject; /* represents /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);
}

/* return cpu shares held by the user */
static ssize_t cpu_shares_show(struct kset *kset, char *buffer)
{
	struct user_struct *up = container_of(kset, struct user_struct, kset);

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

/* modify cpu shares held by the user */
static 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 int user_kobject_create(struct user_struct *up)
{
	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);

	kobject_uevent(kobj, KOBJ_ADD);

done:
	return error;
}

/* 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)
{
	int error;

	/* create under /sys/kernel dir */
	uids_kobject.parent = &kernel_subsys.kobj;
	uids_kobject.kset = &kernel_subsys;
	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;
}

/* 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);
	struct kobject *kobj = &up->kset.kobj;
	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;

	sysfs_remove_file(kobj, &up->user_attr.attr);
	kobject_uevent(kobj, KOBJ_REMOVE);
	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 && CONFIG_SYSFS */

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

/*
 * 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;

	/* 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);

	if (!up) {
		struct user_struct *new;

		new = kmem_cache_alloc(uid_cachep, GFP_KERNEL);
		if (!new) {
			uids_mutex_unlock();
			return NULL;
		}

		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) {
			kmem_cache_free(uid_cachep, new);
			uids_mutex_unlock();
			return NULL;
		}

		if (sched_create_user(new) < 0) {
			key_put(new->uid_keyring);
			key_put(new->session_keyring);
			kmem_cache_free(uid_cachep, new);
			uids_mutex_unlock();
			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..
		 */
		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;
}

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
	118	static struct kobject uids_kobject; /* represents /sys/kernel/uids directory */
	119	static DEFINE_MUTEX(uids_mutex);
	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
5cb350ba	131	/* return cpu shares held by the user */
a0f846aa	132	static ssize_t cpu_shares_show(struct kset kset, char buffer)
5cb350ba DG	133	{
5cb350ba DG	134	struct user_struct *up = container_of(kset, struct user_struct, kset);
24e377a8	135
5cb350ba DG	136	return sprintf(buffer, "%lu\n", sched_group_shares(up->tg));
	137	}
	138
	139	/* modify cpu shares held by the user */
a0f846aa AB	140	static ssize_t cpu_shares_store(struct kset kset, const char buffer,
a0f846aa AB	141	size_t size)
5cb350ba DG	142	{
	143	struct user_struct *up = container_of(kset, struct user_struct, kset);
	144	unsigned long shares;
	145	int rc;
	146
	147	sscanf(buffer, "%lu", &shares);
	148
	149	rc = sched_group_set_shares(up->tg, shares);
	150
	151	return (rc ? rc : size);
	152	}
	153
	154	static void user_attr_init(struct subsys_attribute sa, char name, int mode)
	155	{
	156	sa->attr.name = name;
	157	sa->attr.mode = mode;
	158	sa->show = cpu_shares_show;
	159	sa->store = cpu_shares_store;
	160	}
	161
	162	/* Create "/sys/kernel/uids/<uid>" directory and
	163	* "/sys/kernel/uids/<uid>/cpu_share" file for this user.
1da177e4	164	*/
5cb350ba	165	static int user_kobject_create(struct user_struct *up)
1da177e4	166	{
5cb350ba DG	167	struct kset *kset = &up->kset;
	168	struct kobject *kobj = &kset->kobj;
	169	int error;
	170
	171	memset(kset, 0, sizeof(struct kset));
	172	kobj->parent = &uids_kobject; /* create under /sys/kernel/uids dir */
	173	kobject_set_name(kobj, "%d", up->uid);
	174	kset_init(kset);
	175	user_attr_init(&up->user_attr, "cpu_share", 0644);
	176
	177	error = kobject_add(kobj);
	178	if (error)
	179	goto done;
	180
	181	error = sysfs_create_file(kobj, &up->user_attr.attr);
	182	if (error)
	183	kobject_del(kobj);
	184
fb7dde37 SV	185	kobject_uevent(kobj, KOBJ_ADD);
fb7dde37 SV	186
5cb350ba DG	187	done:
5cb350ba DG	188	return error;
1da177e4 LT	189	}
1da177e4 LT	190
5cb350ba DG	191	/* create these in sysfs filesystem:
	192	* "/sys/kernel/uids" directory
	193	* "/sys/kernel/uids/0" directory (for root user)
	194	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	195	*/
	196	int __init uids_kobject_init(void)
1da177e4	197	{
5cb350ba DG	198	int error;
	199
	200	/* create under /sys/kernel dir */
	201	uids_kobject.parent = &kernel_subsys.kobj;
fb7dde37	202	uids_kobject.kset = &kernel_subsys;
5cb350ba DG	203	kobject_set_name(&uids_kobject, "uids");
	204	kobject_init(&uids_kobject);
	205
	206	error = kobject_add(&uids_kobject);
	207	if (!error)
	208	error = user_kobject_create(&root_user);
	209
	210	return error;
1da177e4 LT	211	}
1da177e4 LT	212
5cb350ba DG	213	/* work function to remove sysfs directory for a user and free up
	214	* corresponding structures.
	215	*/
	216	static void remove_user_sysfs_dir(struct work_struct *w)
1da177e4	217	{
5cb350ba DG	218	struct user_struct *up = container_of(w, struct user_struct, work);
	219	struct kobject *kobj = &up->kset.kobj;
	220	unsigned long flags;
	221	int remove_user = 0;
1da177e4	222
5cb350ba DG	223	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	224	* atomic.
	225	*/
	226	uids_mutex_lock();
	227
	228	local_irq_save(flags);
	229
	230	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
	231	uid_hash_remove(up);
	232	remove_user = 1;
	233	spin_unlock_irqrestore(&uidhash_lock, flags);
	234	} else {
	235	local_irq_restore(flags);
1da177e4 LT	236	}
1da177e4 LT	237
5cb350ba DG	238	if (!remove_user)
	239	goto done;
	240
	241	sysfs_remove_file(kobj, &up->user_attr.attr);
fb7dde37	242	kobject_uevent(kobj, KOBJ_REMOVE);
5cb350ba DG	243	kobject_del(kobj);
	244
	245	sched_destroy_user(up);
	246	key_put(up->uid_keyring);
	247	key_put(up->session_keyring);
	248	kmem_cache_free(uid_cachep, up);
	249
	250	done:
	251	uids_mutex_unlock();
	252	}
	253
	254	/* IRQs are disabled and uidhash_lock is held upon function entry.
	255	* IRQ state (as stored in flags) is restored and uidhash_lock released
	256	* upon function exit.
	257	*/
	258	static inline void free_user(struct user_struct *up, unsigned long flags)
	259	{
	260	/* restore back the count */
	261	atomic_inc(&up->__count);
	262	spin_unlock_irqrestore(&uidhash_lock, flags);
	263
	264	INIT_WORK(&up->work, remove_user_sysfs_dir);
	265	schedule_work(&up->work);
1da177e4 LT	266	}
1da177e4 LT	267
b1a8c172	268	#else /* CONFIG_FAIR_USER_SCHED && CONFIG_SYSFS */
5cb350ba	269
5cb350ba DG	270	static inline int user_kobject_create(struct user_struct *up) { return 0; }
	271	static inline void uids_mutex_lock(void) { }
	272	static inline void uids_mutex_unlock(void) { }
	273
	274	/* IRQs are disabled and uidhash_lock is held upon function entry.
	275	* IRQ state (as stored in flags) is restored and uidhash_lock released
	276	* upon function exit.
	277	*/
	278	static inline void free_user(struct user_struct *up, unsigned long flags)
	279	{
	280	uid_hash_remove(up);
	281	spin_unlock_irqrestore(&uidhash_lock, flags);
	282	sched_destroy_user(up);
	283	key_put(up->uid_keyring);
	284	key_put(up->session_keyring);
	285	kmem_cache_free(uid_cachep, up);
	286	}
	287
b1a8c172	288	#endif
5cb350ba	289
1da177e4 LT	290	/*
	291	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	292	* caller must undo that ref with free_uid().
	293	*
	294	* If the user_struct could not be found, return NULL.
	295	*/
	296	struct user_struct *find_user(uid_t uid)
	297	{
	298	struct user_struct *ret;
3fa97c9d	299	unsigned long flags;
acce292c	300	struct user_namespace *ns = current->nsproxy->user_ns;
1da177e4	301
3fa97c9d	302	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	303	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	304	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	305	return ret;
	306	}
	307
	308	void free_uid(struct user_struct *up)
	309	{
3fa97c9d AM	310	unsigned long flags;
3fa97c9d AM	311
36f57413 AM	312	if (!up)
	313	return;
	314
3fa97c9d	315	local_irq_save(flags);
5cb350ba DG	316	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	317	free_user(up, flags);
	318	else
36f57413	319	local_irq_restore(flags);
1da177e4 LT	320	}
1da177e4 LT	321
acce292c	322	struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
1da177e4	323	{
735de223	324	struct hlist_head *hashent = uidhashentry(ns, uid);
1da177e4 LT	325	struct user_struct *up;
1da177e4 LT	326
5cb350ba DG	327	/* Make uid_hash_find() + user_kobject_create() + uid_hash_insert()
	328	* atomic.
	329	*/
	330	uids_mutex_lock();
	331
3fa97c9d	332	spin_lock_irq(&uidhash_lock);
1da177e4	333	up = uid_hash_find(uid, hashent);
3fa97c9d	334	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	335
	336	if (!up) {
	337	struct user_struct *new;
	338
e94b1766	339	new = kmem_cache_alloc(uid_cachep, GFP_KERNEL);
5e8869bb PE	340	if (!new) {
5e8869bb PE	341	uids_mutex_unlock();
1da177e4	342	return NULL;
5e8869bb PE	343	}
5e8869bb PE	344
1da177e4 LT	345	new->uid = uid;
	346	atomic_set(&new->__count, 1);
	347	atomic_set(&new->processes, 0);
	348	atomic_set(&new->files, 0);
	349	atomic_set(&new->sigpending, 0);
2d9048e2	350	#ifdef CONFIG_INOTIFY_USER
0eeca283 RL	351	atomic_set(&new->inotify_watches, 0);
	352	atomic_set(&new->inotify_devs, 0);
	353	#endif
970a8645	354	#ifdef CONFIG_POSIX_MQUEUE
1da177e4	355	new->mq_bytes = 0;
970a8645	356	#endif
1da177e4 LT	357	new->locked_shm = 0;
1da177e4 LT	358
d720024e	359	if (alloc_uid_keyring(new, current) < 0) {
1da177e4	360	kmem_cache_free(uid_cachep, new);
5e8869bb	361	uids_mutex_unlock();
1da177e4 LT	362	return NULL;
	363	}
	364
24e377a8 SV	365	if (sched_create_user(new) < 0) {
	366	key_put(new->uid_keyring);
	367	key_put(new->session_keyring);
	368	kmem_cache_free(uid_cachep, new);
5e8869bb	369	uids_mutex_unlock();
24e377a8 SV	370	return NULL;
	371	}
	372
5cb350ba DG	373	if (user_kobject_create(new)) {
	374	sched_destroy_user(new);
	375	key_put(new->uid_keyring);
	376	key_put(new->session_keyring);
	377	kmem_cache_free(uid_cachep, new);
	378	uids_mutex_unlock();
	379	return NULL;
	380	}
	381
1da177e4 LT	382	/*
	383	* Before adding this, check whether we raced
	384	* on adding the same user already..
	385	*/
3fa97c9d	386	spin_lock_irq(&uidhash_lock);
1da177e4 LT	387	up = uid_hash_find(uid, hashent);
1da177e4 LT	388	if (up) {
5cb350ba DG	389	/* This case is not possible when CONFIG_FAIR_USER_SCHED
	390	* is defined, since we serialize alloc_uid() using
	391	* uids_mutex. Hence no need to call
	392	* sched_destroy_user() or remove_user_sysfs_dir().
	393	*/
1da177e4 LT	394	key_put(new->uid_keyring);
	395	key_put(new->session_keyring);
	396	kmem_cache_free(uid_cachep, new);
	397	} else {
	398	uid_hash_insert(new, hashent);
	399	up = new;
	400	}
3fa97c9d	401	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	402
1da177e4 LT	403	}
5cb350ba DG	404
	405	uids_mutex_unlock();
	406
1da177e4 LT	407	return up;
	408	}
	409
	410	void switch_uid(struct user_struct *new_user)
	411	{
	412	struct user_struct *old_user;
	413
	414	/* What if a process setreuid()'s and this brings the
	415	* new uid over his NPROC rlimit? We can check this now
	416	* cheaply with the new uid cache, so if it matters
	417	* we should be checking for it. -DaveM
	418	*/
	419	old_user = current->user;
	420	atomic_inc(&new_user->processes);
	421	atomic_dec(&old_user->processes);
	422	switch_uid_keyring(new_user);
	423	current->user = new_user;
24e377a8	424	sched_switch_user(current);
45c18b0b LT	425
	426	/*
	427	* We need to synchronize with __sigqueue_alloc()
	428	* doing a get_uid(p->user).. If that saw the old
	429	* user value, we need to wait until it has exited
	430	* its critical region before we can free the old
	431	* structure.
	432	*/
	433	smp_mb();
	434	spin_unlock_wait(&current->sighand->siglock);
	435
1da177e4 LT	436	free_uid(old_user);
	437	suid_keys(current);
	438	}
	439
28f300d2 PE	440	void release_uids(struct user_namespace *ns)
	441	{
	442	int i;
	443	unsigned long flags;
	444	struct hlist_head *head;
	445	struct hlist_node *nd;
	446
	447	spin_lock_irqsave(&uidhash_lock, flags);
	448	/*
	449	* collapse the chains so that the user_struct-s will
	450	* be still alive, but not in hashes. subsequent free_uid()
	451	* will free them.
	452	*/
	453	for (i = 0; i < UIDHASH_SZ; i++) {
	454	head = ns->uidhash_table + i;
	455	while (!hlist_empty(head)) {
	456	nd = head->first;
	457	hlist_del_init(nd);
	458	}
	459	}
	460	spin_unlock_irqrestore(&uidhash_lock, flags);
	461
	462	free_uid(ns->root_user);
	463	}
1da177e4 LT	464
	465	static int __init uid_cache_init(void)
	466	{
	467	int n;
	468
	469	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	470	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	471
1da177e4 LT	472	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	473	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	474
1da177e4 LT	475	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	476	spin_lock_irq(&uidhash_lock);
acce292c	477	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	478	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	479
	480	return 0;
	481	}
	482
	483	module_init(uid_cache_init);