[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_KEYS
	.uid_keyring	= &root_user_keyring,
	.session_keyring = &root_session_keyring,
#endif
#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();
	if (IS_ERR(up->tg))
		rc = -ENOMEM;

	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, "%lu\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, "%lu", &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);
#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,
#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_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_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);
}

#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 LT	55	.locked_shm = 0,
	56	#ifdef CONFIG_KEYS
	57	.uid_keyring = &root_user_keyring,
	58	.session_keyring = &root_session_keyring,
	59	#endif
052f1dc7	60	#ifdef CONFIG_USER_SCHED
4cf86d77	61	.tg = &init_task_group,
24e377a8	62	#endif
1da177e4 LT	63	};
1da177e4 LT	64
5cb350ba DG	65	/*
	66	* These routines must be called with the uidhash spinlock held!
	67	*/
40aeb400	68	static void uid_hash_insert(struct user_struct up, struct hlist_head hashent)
5cb350ba DG	69	{
	70	hlist_add_head(&up->uidhash_node, hashent);
	71	}
	72
40aeb400	73	static void uid_hash_remove(struct user_struct *up)
5cb350ba DG	74	{
	75	hlist_del_init(&up->uidhash_node);
	76	}
	77
40aeb400	78	static struct user_struct uid_hash_find(uid_t uid, struct hlist_head hashent)
5cb350ba DG	79	{
	80	struct user_struct *user;
	81	struct hlist_node *h;
	82
	83	hlist_for_each_entry(user, h, hashent, uidhash_node) {
	84	if (user->uid == uid) {
	85	atomic_inc(&user->__count);
	86	return user;
	87	}
	88	}
	89
	90	return NULL;
	91	}
	92
052f1dc7	93	#ifdef CONFIG_USER_SCHED
5cb350ba	94
24e377a8 SV	95	static void sched_destroy_user(struct user_struct *up)
	96	{
	97	sched_destroy_group(up->tg);
	98	}
	99
	100	static int sched_create_user(struct user_struct *up)
	101	{
	102	int rc = 0;
	103
	104	up->tg = sched_create_group();
	105	if (IS_ERR(up->tg))
	106	rc = -ENOMEM;
	107
	108	return rc;
	109	}
	110
	111	static void sched_switch_user(struct task_struct *p)
	112	{
	113	sched_move_task(p);
	114	}
	115
052f1dc7	116	#else /* CONFIG_USER_SCHED */
b1a8c172 DG	117
	118	static void sched_destroy_user(struct user_struct *up) { }
	119	static int sched_create_user(struct user_struct *up) { return 0; }
	120	static void sched_switch_user(struct task_struct *p) { }
	121
052f1dc7	122	#endif /* CONFIG_USER_SCHED */
b1a8c172	123
052f1dc7	124	#if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS)
b1a8c172	125
eb41d946	126	static struct kset uids_kset; / represents the /sys/kernel/uids/ directory */
b1a8c172 DG	127	static DEFINE_MUTEX(uids_mutex);
b1a8c172 DG	128
5cb350ba DG	129	static inline void uids_mutex_lock(void)
	130	{
	131	mutex_lock(&uids_mutex);
	132	}
24e377a8	133
5cb350ba DG	134	static inline void uids_mutex_unlock(void)
	135	{
	136	mutex_unlock(&uids_mutex);
	137	}
24e377a8	138
eb41d946	139	/* uid directory attributes */
052f1dc7	140	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946 KS	141	static ssize_t cpu_shares_show(struct kobject *kobj,
	142	struct kobj_attribute *attr,
	143	char *buf)
5cb350ba	144	{
eb41d946	145	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
24e377a8	146
eb41d946	147	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
5cb350ba DG	148	}
5cb350ba DG	149
eb41d946 KS	150	static ssize_t cpu_shares_store(struct kobject *kobj,
	151	struct kobj_attribute *attr,
	152	const char *buf, size_t size)
5cb350ba	153	{
eb41d946	154	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
5cb350ba DG	155	unsigned long shares;
	156	int rc;
	157
eb41d946	158	sscanf(buf, "%lu", &shares);
5cb350ba DG	159
	160	rc = sched_group_set_shares(up->tg, shares);
	161
	162	return (rc ? rc : size);
	163	}
	164
eb41d946 KS	165	static struct kobj_attribute cpu_share_attr =
eb41d946 KS	166	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
052f1dc7	167	#endif
eb41d946	168
052f1dc7	169	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56 PZ	170	static ssize_t cpu_rt_runtime_show(struct kobject *kobj,
	171	struct kobj_attribute *attr,
	172	char *buf)
	173	{
	174	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	175
	176	return sprintf(buf, "%lu\n", sched_group_rt_runtime(up->tg));
	177	}
	178
	179	static ssize_t cpu_rt_runtime_store(struct kobject *kobj,
	180	struct kobj_attribute *attr,
	181	const char *buf, size_t size)
	182	{
	183	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	184	unsigned long rt_runtime;
	185	int rc;
	186
	187	sscanf(buf, "%lu", &rt_runtime);
	188
	189	rc = sched_group_set_rt_runtime(up->tg, rt_runtime);
	190
	191	return (rc ? rc : size);
	192	}
	193
	194	static struct kobj_attribute cpu_rt_runtime_attr =
	195	__ATTR(cpu_rt_runtime, 0644, cpu_rt_runtime_show, cpu_rt_runtime_store);
052f1dc7	196	#endif
9f0c1e56	197
eb41d946 KS	198	/* default attributes per uid directory */
eb41d946 KS	199	static struct attribute *uids_attributes[] = {
052f1dc7	200	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946	201	&cpu_share_attr.attr,
052f1dc7 PZ	202	#endif
052f1dc7 PZ	203	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56	204	&cpu_rt_runtime_attr.attr,
052f1dc7	205	#endif
eb41d946 KS	206	NULL
	207	};
	208
	209	/* the lifetime of user_struct is not managed by the core (now) */
	210	static void uids_release(struct kobject *kobj)
5cb350ba	211	{
eb41d946	212	return;
5cb350ba DG	213	}
5cb350ba DG	214
eb41d946 KS	215	static struct kobj_type uids_ktype = {
	216	.sysfs_ops = &kobj_sysfs_ops,
	217	.default_attrs = uids_attributes,
	218	.release = uids_release,
	219	};
	220
	221	/* create /sys/kernel/uids/<uid>/cpu_share file for this user */
	222	static int uids_user_create(struct user_struct *up)
1da177e4	223	{
eb41d946	224	struct kobject *kobj = &up->kobj;
5cb350ba DG	225	int error;
5cb350ba DG	226
eb41d946	227	memset(kobj, 0, sizeof(struct kobject));
eb41d946	228	kobj->kset = uids_kset;
cf15126b GKH	229	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	230	if (error) {
	231	kobject_put(kobj);
5cb350ba	232	goto done;
cf15126b	233	}
5cb350ba	234
fb7dde37	235	kobject_uevent(kobj, KOBJ_ADD);
5cb350ba DG	236	done:
5cb350ba DG	237	return error;
1da177e4 LT	238	}
1da177e4 LT	239
eb41d946	240	/* create these entries in sysfs:
5cb350ba DG	241	* "/sys/kernel/uids" directory
	242	* "/sys/kernel/uids/0" directory (for root user)
	243	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	244	*/
eb41d946	245	int __init uids_sysfs_init(void)
1da177e4	246	{
0ff21e46	247	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
eb41d946 KS	248	if (!uids_kset)
eb41d946 KS	249	return -ENOMEM;
5cb350ba	250
eb41d946	251	return uids_user_create(&root_user);
1da177e4 LT	252	}
1da177e4 LT	253
5cb350ba DG	254	/* work function to remove sysfs directory for a user and free up
	255	* corresponding structures.
	256	*/
	257	static void remove_user_sysfs_dir(struct work_struct *w)
1da177e4	258	{
5cb350ba	259	struct user_struct *up = container_of(w, struct user_struct, work);
5cb350ba DG	260	unsigned long flags;
5cb350ba DG	261	int remove_user = 0;
1da177e4	262
5cb350ba DG	263	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	264	* atomic.
	265	*/
	266	uids_mutex_lock();
	267
	268	local_irq_save(flags);
	269
	270	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
	271	uid_hash_remove(up);
	272	remove_user = 1;
	273	spin_unlock_irqrestore(&uidhash_lock, flags);
	274	} else {
	275	local_irq_restore(flags);
1da177e4 LT	276	}
1da177e4 LT	277
5cb350ba DG	278	if (!remove_user)
	279	goto done;
	280
eb41d946 KS	281	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	282	kobject_del(&up->kobj);
	283	kobject_put(&up->kobj);
5cb350ba DG	284
	285	sched_destroy_user(up);
	286	key_put(up->uid_keyring);
	287	key_put(up->session_keyring);
	288	kmem_cache_free(uid_cachep, up);
	289
	290	done:
	291	uids_mutex_unlock();
	292	}
	293
	294	/* IRQs are disabled and uidhash_lock is held upon function entry.
	295	* IRQ state (as stored in flags) is restored and uidhash_lock released
	296	* upon function exit.
	297	*/
	298	static inline void free_user(struct user_struct *up, unsigned long flags)
	299	{
	300	/* restore back the count */
	301	atomic_inc(&up->__count);
	302	spin_unlock_irqrestore(&uidhash_lock, flags);
	303
	304	INIT_WORK(&up->work, remove_user_sysfs_dir);
	305	schedule_work(&up->work);
1da177e4 LT	306	}
1da177e4 LT	307
052f1dc7	308	#else /* CONFIG_USER_SCHED && CONFIG_SYSFS */
5cb350ba	309
eb41d946 KS	310	int uids_sysfs_init(void) { return 0; }
eb41d946 KS	311	static inline int uids_user_create(struct user_struct *up) { return 0; }
5cb350ba DG	312	static inline void uids_mutex_lock(void) { }
	313	static inline void uids_mutex_unlock(void) { }
	314
	315	/* IRQs are disabled and uidhash_lock is held upon function entry.
	316	* IRQ state (as stored in flags) is restored and uidhash_lock released
	317	* upon function exit.
	318	*/
	319	static inline void free_user(struct user_struct *up, unsigned long flags)
	320	{
	321	uid_hash_remove(up);
	322	spin_unlock_irqrestore(&uidhash_lock, flags);
	323	sched_destroy_user(up);
	324	key_put(up->uid_keyring);
	325	key_put(up->session_keyring);
	326	kmem_cache_free(uid_cachep, up);
	327	}
	328
b1a8c172	329	#endif
5cb350ba	330
1da177e4 LT	331	/*
	332	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	333	* caller must undo that ref with free_uid().
	334	*
	335	* If the user_struct could not be found, return NULL.
	336	*/
	337	struct user_struct *find_user(uid_t uid)
	338	{
	339	struct user_struct *ret;
3fa97c9d	340	unsigned long flags;
acce292c	341	struct user_namespace *ns = current->nsproxy->user_ns;
1da177e4	342
3fa97c9d	343	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	344	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	345	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	346	return ret;
	347	}
	348
	349	void free_uid(struct user_struct *up)
	350	{
3fa97c9d AM	351	unsigned long flags;
3fa97c9d AM	352
36f57413 AM	353	if (!up)
	354	return;
	355
3fa97c9d	356	local_irq_save(flags);
5cb350ba DG	357	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	358	free_user(up, flags);
	359	else
36f57413	360	local_irq_restore(flags);
1da177e4 LT	361	}
1da177e4 LT	362
acce292c	363	struct user_struct * alloc_uid(struct user_namespace *ns, uid_t uid)
1da177e4	364	{
735de223	365	struct hlist_head *hashent = uidhashentry(ns, uid);
8eb703e4	366	struct user_struct up, new;
1da177e4	367
eb41d946	368	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
5cb350ba DG	369	* atomic.
	370	*/
	371	uids_mutex_lock();
	372
3fa97c9d	373	spin_lock_irq(&uidhash_lock);
1da177e4	374	up = uid_hash_find(uid, hashent);
3fa97c9d	375	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	376
1da177e4 LT	377	if (!up) {
e94b1766	378	new = kmem_cache_alloc(uid_cachep, GFP_KERNEL);
8eb703e4 PE	379	if (!new)
8eb703e4 PE	380	goto out_unlock;
5e8869bb	381
1da177e4 LT	382	new->uid = uid;
	383	atomic_set(&new->__count, 1);
	384	atomic_set(&new->processes, 0);
	385	atomic_set(&new->files, 0);
	386	atomic_set(&new->sigpending, 0);
2d9048e2	387	#ifdef CONFIG_INOTIFY_USER
0eeca283 RL	388	atomic_set(&new->inotify_watches, 0);
	389	atomic_set(&new->inotify_devs, 0);
	390	#endif
970a8645	391	#ifdef CONFIG_POSIX_MQUEUE
1da177e4	392	new->mq_bytes = 0;
970a8645	393	#endif
1da177e4 LT	394	new->locked_shm = 0;
1da177e4 LT	395
8eb703e4 PE	396	if (alloc_uid_keyring(new, current) < 0)
8eb703e4 PE	397	goto out_free_user;
1da177e4	398
8eb703e4 PE	399	if (sched_create_user(new) < 0)
8eb703e4 PE	400	goto out_put_keys;
24e377a8	401
8eb703e4 PE	402	if (uids_user_create(new))
8eb703e4 PE	403	goto out_destoy_sched;
5cb350ba	404
1da177e4 LT	405	/*
	406	* Before adding this, check whether we raced
	407	* on adding the same user already..
	408	*/
3fa97c9d	409	spin_lock_irq(&uidhash_lock);
1da177e4 LT	410	up = uid_hash_find(uid, hashent);
1da177e4 LT	411	if (up) {
052f1dc7	412	/* This case is not possible when CONFIG_USER_SCHED
5cb350ba DG	413	* is defined, since we serialize alloc_uid() using
	414	* uids_mutex. Hence no need to call
	415	* sched_destroy_user() or remove_user_sysfs_dir().
	416	*/
1da177e4 LT	417	key_put(new->uid_keyring);
	418	key_put(new->session_keyring);
	419	kmem_cache_free(uid_cachep, new);
	420	} else {
	421	uid_hash_insert(new, hashent);
	422	up = new;
	423	}
3fa97c9d	424	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	425
1da177e4 LT	426	}
5cb350ba DG	427
	428	uids_mutex_unlock();
	429
1da177e4	430	return up;
8eb703e4 PE	431
	432	out_destoy_sched:
	433	sched_destroy_user(new);
	434	out_put_keys:
	435	key_put(new->uid_keyring);
	436	key_put(new->session_keyring);
	437	out_free_user:
	438	kmem_cache_free(uid_cachep, new);
	439	out_unlock:
	440	uids_mutex_unlock();
	441	return NULL;
1da177e4 LT	442	}
	443
	444	void switch_uid(struct user_struct *new_user)
	445	{
	446	struct user_struct *old_user;
	447
	448	/* What if a process setreuid()'s and this brings the
	449	* new uid over his NPROC rlimit? We can check this now
	450	* cheaply with the new uid cache, so if it matters
	451	* we should be checking for it. -DaveM
	452	*/
	453	old_user = current->user;
	454	atomic_inc(&new_user->processes);
	455	atomic_dec(&old_user->processes);
	456	switch_uid_keyring(new_user);
	457	current->user = new_user;
24e377a8	458	sched_switch_user(current);
45c18b0b LT	459
	460	/*
	461	* We need to synchronize with __sigqueue_alloc()
	462	* doing a get_uid(p->user).. If that saw the old
	463	* user value, we need to wait until it has exited
	464	* its critical region before we can free the old
	465	* structure.
	466	*/
	467	smp_mb();
	468	spin_unlock_wait(&current->sighand->siglock);
	469
1da177e4 LT	470	free_uid(old_user);
	471	suid_keys(current);
	472	}
	473
aee16ce7	474	#ifdef CONFIG_USER_NS
28f300d2 PE	475	void release_uids(struct user_namespace *ns)
	476	{
	477	int i;
	478	unsigned long flags;
	479	struct hlist_head *head;
	480	struct hlist_node *nd;
	481
	482	spin_lock_irqsave(&uidhash_lock, flags);
	483	/*
	484	* collapse the chains so that the user_struct-s will
	485	* be still alive, but not in hashes. subsequent free_uid()
	486	* will free them.
	487	*/
	488	for (i = 0; i < UIDHASH_SZ; i++) {
	489	head = ns->uidhash_table + i;
	490	while (!hlist_empty(head)) {
	491	nd = head->first;
	492	hlist_del_init(nd);
	493	}
	494	}
	495	spin_unlock_irqrestore(&uidhash_lock, flags);
	496
	497	free_uid(ns->root_user);
	498	}
aee16ce7	499	#endif
1da177e4 LT	500
	501	static int __init uid_cache_init(void)
	502	{
	503	int n;
	504
	505	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	506	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	507
1da177e4 LT	508	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	509	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	510
1da177e4 LT	511	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	512	spin_lock_irq(&uidhash_lock);
acce292c	513	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	514	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	515
	516	return 0;
	517	}
	518
	519	module_init(uid_cache_init);