[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>
#include "cred-internals.h"

struct user_namespace init_user_ns = {
	.kref = {
		.refcount	= ATOMIC_INIT(1),
	},
	.creator = &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);

/* root_user.__count is 2, 1 for init task cred, 1 for init_user_ns->creator */
struct user_struct root_user = {
	.__count	= ATOMIC_INIT(2),
	.processes	= ATOMIC_INIT(1),
	.files		= ATOMIC_INIT(0),
	.sigpending	= ATOMIC_INIT(0),
	.locked_shm     = 0,
	.user_ns	= &init_user_ns,
#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;

	return rc;
}

#else	/* CONFIG_USER_SCHED */

static void sched_destroy_user(struct user_struct *up) { }
static int sched_create_user(struct user_struct *up) { return 0; }

#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));
	if (up->user_ns != &init_user_ns)
		return 0;
	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;

	if (up->user_ns != &init_user_ns)
		return;
	/* 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 void free_user(struct user_struct *up, unsigned long flags)
{
	/* restore back the count */
	atomic_inc(&up->__count);
	spin_unlock_irqrestore(&uidhash_lock, flags);

	put_user_ns(up->user_ns);
	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 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);
	put_user_ns(up->user_ns);
	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_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;

		new->user_ns = get_user_ns(ns);

		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);
	put_user_ns(new->user_ns);
out_free_user:
	kmem_cache_free(uid_cachep, new);
out_unlock:
	uids_mutex_unlock();
	return NULL;
}

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>
d84f4f99	19	#include "cred-internals.h"
1da177e4	20
aee16ce7 PE	21	struct user_namespace init_user_ns = {
aee16ce7 PE	22	.kref = {
18b6e041	23	.refcount = ATOMIC_INIT(1),
aee16ce7	24	},
18b6e041	25	.creator = &root_user,
aee16ce7 PE	26	};
	27	EXPORT_SYMBOL_GPL(init_user_ns);
	28
1da177e4 LT	29	/*
	30	* UID task count cache, to get fast user lookup in "alloc_uid"
	31	* when changing user ID's (ie setuid() and friends).
	32	*/
	33
1da177e4 LT	34	#define UIDHASH_MASK (UIDHASH_SZ - 1)
1da177e4 LT	35	#define __uidhashfn(uid) (((uid >> UIDHASH_BITS) + uid) & UIDHASH_MASK)
acce292c	36	#define uidhashentry(ns, uid) ((ns)->uidhash_table + __uidhashfn((uid)))
1da177e4	37
e18b890b	38	static struct kmem_cache *uid_cachep;
4021cb27 IM	39
	40	/*
	41	* The uidhash_lock is mostly taken from process context, but it is
	42	* occasionally also taken from softirq/tasklet context, when
	43	* task-structs get RCU-freed. Hence all locking must be softirq-safe.
3fa97c9d AM	44	* But free_uid() is also called with local interrupts disabled, and running
	45	* local_bh_enable() with local interrupts disabled is an error - we'll run
	46	* softirq callbacks, and they can unconditionally enable interrupts, and
	47	* the caller of free_uid() didn't expect that..
4021cb27	48	*/
1da177e4 LT	49	static DEFINE_SPINLOCK(uidhash_lock);
1da177e4 LT	50
18b6e041	51	/* root_user.__count is 2, 1 for init task cred, 1 for init_user_ns->creator */
1da177e4	52	struct user_struct root_user = {
18b6e041	53	.__count = ATOMIC_INIT(2),
1da177e4 LT	54	.processes = ATOMIC_INIT(1),
	55	.files = ATOMIC_INIT(0),
	56	.sigpending = ATOMIC_INIT(0),
1da177e4	57	.locked_shm = 0,
18b6e041	58	.user_ns = &init_user_ns,
052f1dc7	59	#ifdef CONFIG_USER_SCHED
4cf86d77	60	.tg = &init_task_group,
24e377a8	61	#endif
1da177e4 LT	62	};
1da177e4 LT	63
5cb350ba DG	64	/*
	65	* These routines must be called with the uidhash spinlock held!
	66	*/
40aeb400	67	static void uid_hash_insert(struct user_struct up, struct hlist_head hashent)
5cb350ba DG	68	{
	69	hlist_add_head(&up->uidhash_node, hashent);
	70	}
	71
40aeb400	72	static void uid_hash_remove(struct user_struct *up)
5cb350ba DG	73	{
	74	hlist_del_init(&up->uidhash_node);
	75	}
	76
40aeb400	77	static struct user_struct uid_hash_find(uid_t uid, struct hlist_head hashent)
5cb350ba DG	78	{
	79	struct user_struct *user;
	80	struct hlist_node *h;
	81
	82	hlist_for_each_entry(user, h, hashent, uidhash_node) {
	83	if (user->uid == uid) {
	84	atomic_inc(&user->__count);
	85	return user;
	86	}
	87	}
	88
	89	return NULL;
	90	}
	91
052f1dc7	92	#ifdef CONFIG_USER_SCHED
5cb350ba	93
24e377a8 SV	94	static void sched_destroy_user(struct user_struct *up)
	95	{
	96	sched_destroy_group(up->tg);
	97	}
	98
	99	static int sched_create_user(struct user_struct *up)
	100	{
	101	int rc = 0;
	102
eff766a6	103	up->tg = sched_create_group(&root_task_group);
24e377a8 SV	104	if (IS_ERR(up->tg))
	105	rc = -ENOMEM;
	106
	107	return rc;
	108	}
	109
052f1dc7	110	#else /* CONFIG_USER_SCHED */
b1a8c172 DG	111
	112	static void sched_destroy_user(struct user_struct *up) { }
	113	static int sched_create_user(struct user_struct *up) { return 0; }
b1a8c172	114
052f1dc7	115	#endif /* CONFIG_USER_SCHED */
b1a8c172	116
052f1dc7	117	#if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS)
b1a8c172	118
eb41d946	119	static struct kset uids_kset; / represents the /sys/kernel/uids/ directory */
b1a8c172 DG	120	static DEFINE_MUTEX(uids_mutex);
b1a8c172 DG	121
5cb350ba DG	122	static inline void uids_mutex_lock(void)
	123	{
	124	mutex_lock(&uids_mutex);
	125	}
24e377a8	126
5cb350ba DG	127	static inline void uids_mutex_unlock(void)
	128	{
	129	mutex_unlock(&uids_mutex);
	130	}
24e377a8	131
eb41d946	132	/* uid directory attributes */
052f1dc7	133	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946 KS	134	static ssize_t cpu_shares_show(struct kobject *kobj,
	135	struct kobj_attribute *attr,
	136	char *buf)
5cb350ba	137	{
eb41d946	138	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
24e377a8	139
eb41d946	140	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
5cb350ba DG	141	}
5cb350ba DG	142
eb41d946 KS	143	static ssize_t cpu_shares_store(struct kobject *kobj,
	144	struct kobj_attribute *attr,
	145	const char *buf, size_t size)
5cb350ba	146	{
eb41d946	147	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
5cb350ba DG	148	unsigned long shares;
	149	int rc;
	150
eb41d946	151	sscanf(buf, "%lu", &shares);
5cb350ba DG	152
	153	rc = sched_group_set_shares(up->tg, shares);
	154
	155	return (rc ? rc : size);
	156	}
	157
eb41d946 KS	158	static struct kobj_attribute cpu_share_attr =
eb41d946 KS	159	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
052f1dc7	160	#endif
eb41d946	161
052f1dc7	162	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56 PZ	163	static ssize_t cpu_rt_runtime_show(struct kobject *kobj,
	164	struct kobj_attribute *attr,
	165	char *buf)
	166	{
	167	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	168
af4491e5	169	return sprintf(buf, "%ld\n", sched_group_rt_runtime(up->tg));
9f0c1e56 PZ	170	}
	171
	172	static ssize_t cpu_rt_runtime_store(struct kobject *kobj,
	173	struct kobj_attribute *attr,
	174	const char *buf, size_t size)
	175	{
	176	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	177	unsigned long rt_runtime;
	178	int rc;
	179
af4491e5	180	sscanf(buf, "%ld", &rt_runtime);
9f0c1e56 PZ	181
	182	rc = sched_group_set_rt_runtime(up->tg, rt_runtime);
	183
	184	return (rc ? rc : size);
	185	}
	186
	187	static struct kobj_attribute cpu_rt_runtime_attr =
	188	__ATTR(cpu_rt_runtime, 0644, cpu_rt_runtime_show, cpu_rt_runtime_store);
d0b27fa7 PZ	189
	190	static ssize_t cpu_rt_period_show(struct kobject *kobj,
	191	struct kobj_attribute *attr,
	192	char *buf)
	193	{
	194	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	195
	196	return sprintf(buf, "%lu\n", sched_group_rt_period(up->tg));
	197	}
	198
	199	static ssize_t cpu_rt_period_store(struct kobject *kobj,
	200	struct kobj_attribute *attr,
	201	const char *buf, size_t size)
	202	{
	203	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	204	unsigned long rt_period;
	205	int rc;
	206
	207	sscanf(buf, "%lu", &rt_period);
	208
	209	rc = sched_group_set_rt_period(up->tg, rt_period);
	210
	211	return (rc ? rc : size);
	212	}
	213
	214	static struct kobj_attribute cpu_rt_period_attr =
	215	__ATTR(cpu_rt_period, 0644, cpu_rt_period_show, cpu_rt_period_store);
052f1dc7	216	#endif
9f0c1e56	217
eb41d946 KS	218	/* default attributes per uid directory */
eb41d946 KS	219	static struct attribute *uids_attributes[] = {
052f1dc7	220	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946	221	&cpu_share_attr.attr,
052f1dc7 PZ	222	#endif
052f1dc7 PZ	223	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56	224	&cpu_rt_runtime_attr.attr,
d0b27fa7	225	&cpu_rt_period_attr.attr,
052f1dc7	226	#endif
eb41d946 KS	227	NULL
	228	};
	229
	230	/* the lifetime of user_struct is not managed by the core (now) */
	231	static void uids_release(struct kobject *kobj)
5cb350ba	232	{
eb41d946	233	return;
5cb350ba DG	234	}
5cb350ba DG	235
eb41d946 KS	236	static struct kobj_type uids_ktype = {
	237	.sysfs_ops = &kobj_sysfs_ops,
	238	.default_attrs = uids_attributes,
	239	.release = uids_release,
	240	};
	241
	242	/* create /sys/kernel/uids/<uid>/cpu_share file for this user */
	243	static int uids_user_create(struct user_struct *up)
1da177e4	244	{
eb41d946	245	struct kobject *kobj = &up->kobj;
5cb350ba DG	246	int error;
5cb350ba DG	247
eb41d946	248	memset(kobj, 0, sizeof(struct kobject));
c37bbb0f SH	249	if (up->user_ns != &init_user_ns)
c37bbb0f SH	250	return 0;
eb41d946	251	kobj->kset = uids_kset;
cf15126b GKH	252	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	253	if (error) {
	254	kobject_put(kobj);
5cb350ba	255	goto done;
cf15126b	256	}
5cb350ba	257
fb7dde37	258	kobject_uevent(kobj, KOBJ_ADD);
5cb350ba DG	259	done:
5cb350ba DG	260	return error;
1da177e4 LT	261	}
1da177e4 LT	262
eb41d946	263	/* create these entries in sysfs:
5cb350ba DG	264	* "/sys/kernel/uids" directory
	265	* "/sys/kernel/uids/0" directory (for root user)
	266	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	267	*/
eb41d946	268	int __init uids_sysfs_init(void)
1da177e4	269	{
0ff21e46	270	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
eb41d946 KS	271	if (!uids_kset)
eb41d946 KS	272	return -ENOMEM;
5cb350ba	273
eb41d946	274	return uids_user_create(&root_user);
1da177e4 LT	275	}
1da177e4 LT	276
5cb350ba DG	277	/* work function to remove sysfs directory for a user and free up
	278	* corresponding structures.
	279	*/
	280	static void remove_user_sysfs_dir(struct work_struct *w)
1da177e4	281	{
5cb350ba	282	struct user_struct *up = container_of(w, struct user_struct, work);
5cb350ba DG	283	unsigned long flags;
5cb350ba DG	284	int remove_user = 0;
1da177e4	285
c37bbb0f SH	286	if (up->user_ns != &init_user_ns)
c37bbb0f SH	287	return;
5cb350ba DG	288	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	289	* atomic.
	290	*/
	291	uids_mutex_lock();
	292
	293	local_irq_save(flags);
	294
	295	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
	296	uid_hash_remove(up);
	297	remove_user = 1;
	298	spin_unlock_irqrestore(&uidhash_lock, flags);
	299	} else {
	300	local_irq_restore(flags);
1da177e4 LT	301	}
1da177e4 LT	302
5cb350ba DG	303	if (!remove_user)
	304	goto done;
	305
eb41d946 KS	306	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	307	kobject_del(&up->kobj);
	308	kobject_put(&up->kobj);
5cb350ba DG	309
	310	sched_destroy_user(up);
	311	key_put(up->uid_keyring);
	312	key_put(up->session_keyring);
	313	kmem_cache_free(uid_cachep, up);
	314
	315	done:
	316	uids_mutex_unlock();
	317	}
	318
	319	/* IRQs are disabled and uidhash_lock is held upon function entry.
	320	* IRQ state (as stored in flags) is restored and uidhash_lock released
	321	* upon function exit.
	322	*/
18b6e041	323	static void free_user(struct user_struct *up, unsigned long flags)
5cb350ba DG	324	{
	325	/* restore back the count */
	326	atomic_inc(&up->__count);
	327	spin_unlock_irqrestore(&uidhash_lock, flags);
	328
18b6e041	329	put_user_ns(up->user_ns);
5cb350ba DG	330	INIT_WORK(&up->work, remove_user_sysfs_dir);
5cb350ba DG	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	*/
18b6e041	345	static void free_user(struct user_struct *up, unsigned long flags)
5cb350ba DG	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);
18b6e041	352	put_user_ns(up->user_ns);
5cb350ba DG	353	kmem_cache_free(uid_cachep, up);
	354	}
	355
b1a8c172	356	#endif
5cb350ba	357
1da177e4 LT	358	/*
	359	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	360	* caller must undo that ref with free_uid().
	361	*
	362	* If the user_struct could not be found, return NULL.
	363	*/
	364	struct user_struct *find_user(uid_t uid)
	365	{
	366	struct user_struct *ret;
3fa97c9d	367	unsigned long flags;
6ded6ab9	368	struct user_namespace *ns = current_user_ns();
1da177e4	369
3fa97c9d	370	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	371	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	372	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	373	return ret;
	374	}
	375
	376	void free_uid(struct user_struct *up)
	377	{
3fa97c9d AM	378	unsigned long flags;
3fa97c9d AM	379
36f57413 AM	380	if (!up)
	381	return;
	382
3fa97c9d	383	local_irq_save(flags);
5cb350ba DG	384	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	385	free_user(up, flags);
	386	else
36f57413	387	local_irq_restore(flags);
1da177e4 LT	388	}
1da177e4 LT	389
354a1f4d	390	struct user_struct alloc_uid(struct user_namespace ns, uid_t uid)
1da177e4	391	{
735de223	392	struct hlist_head *hashent = uidhashentry(ns, uid);
8eb703e4	393	struct user_struct up, new;
1da177e4	394
eb41d946	395	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
5cb350ba DG	396	* atomic.
	397	*/
	398	uids_mutex_lock();
	399
3fa97c9d	400	spin_lock_irq(&uidhash_lock);
1da177e4	401	up = uid_hash_find(uid, hashent);
3fa97c9d	402	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	403
1da177e4 LT	404	if (!up) {
354a1f4d	405	new = kmem_cache_zalloc(uid_cachep, GFP_KERNEL);
8eb703e4 PE	406	if (!new)
8eb703e4 PE	407	goto out_unlock;
5e8869bb	408
1da177e4 LT	409	new->uid = uid;
1da177e4 LT	410	atomic_set(&new->__count, 1);
1da177e4	411
8eb703e4	412	if (sched_create_user(new) < 0)
69664cf1	413	goto out_free_user;
24e377a8	414
18b6e041 SH	415	new->user_ns = get_user_ns(ns);
18b6e041 SH	416
8eb703e4 PE	417	if (uids_user_create(new))
8eb703e4 PE	418	goto out_destoy_sched;
5cb350ba	419
1da177e4 LT	420	/*
	421	* Before adding this, check whether we raced
	422	* on adding the same user already..
	423	*/
3fa97c9d	424	spin_lock_irq(&uidhash_lock);
1da177e4 LT	425	up = uid_hash_find(uid, hashent);
1da177e4 LT	426	if (up) {
052f1dc7	427	/* This case is not possible when CONFIG_USER_SCHED
5cb350ba DG	428	* is defined, since we serialize alloc_uid() using
	429	* uids_mutex. Hence no need to call
	430	* sched_destroy_user() or remove_user_sysfs_dir().
	431	*/
1da177e4 LT	432	key_put(new->uid_keyring);
	433	key_put(new->session_keyring);
	434	kmem_cache_free(uid_cachep, new);
	435	} else {
	436	uid_hash_insert(new, hashent);
	437	up = new;
	438	}
3fa97c9d	439	spin_unlock_irq(&uidhash_lock);
1da177e4	440	}
5cb350ba DG	441
	442	uids_mutex_unlock();
	443
1da177e4	444	return up;
8eb703e4 PE	445
	446	out_destoy_sched:
	447	sched_destroy_user(new);
18b6e041	448	put_user_ns(new->user_ns);
8eb703e4 PE	449	out_free_user:
	450	kmem_cache_free(uid_cachep, new);
	451	out_unlock:
	452	uids_mutex_unlock();
	453	return NULL;
1da177e4 LT	454	}
1da177e4 LT	455
1da177e4 LT	456	static int __init uid_cache_init(void)
	457	{
	458	int n;
	459
	460	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	461	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	462
1da177e4 LT	463	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	464	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	465
1da177e4 LT	466	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	467	spin_lock_irq(&uidhash_lock);
acce292c	468	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	469	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	470
	471	return 0;
	472	}
	473
	474	module_init(uid_cache_init);