[net-next-2.6.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;

	set_tg_uid(up);

	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
 * We do not create this file for users in a user namespace (until
 * sysfs tagging is implemented).
 *
 * See Documentation/scheduler/sched-design-CFS.txt for ramifications.
 */
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
6c415b92 AB	107	set_tg_uid(up);
6c415b92 AB	108
24e377a8 SV	109	return rc;
	110	}
	111
052f1dc7	112	#else /* CONFIG_USER_SCHED */
b1a8c172 DG	113
	114	static void sched_destroy_user(struct user_struct *up) { }
	115	static int sched_create_user(struct user_struct *up) { return 0; }
b1a8c172	116
052f1dc7	117	#endif /* CONFIG_USER_SCHED */
b1a8c172	118
052f1dc7	119	#if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS)
b1a8c172	120
eb41d946	121	static struct kset uids_kset; / represents the /sys/kernel/uids/ directory */
b1a8c172 DG	122	static DEFINE_MUTEX(uids_mutex);
b1a8c172 DG	123
5cb350ba DG	124	static inline void uids_mutex_lock(void)
	125	{
	126	mutex_lock(&uids_mutex);
	127	}
24e377a8	128
5cb350ba DG	129	static inline void uids_mutex_unlock(void)
	130	{
	131	mutex_unlock(&uids_mutex);
	132	}
24e377a8	133
eb41d946	134	/* uid directory attributes */
052f1dc7	135	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946 KS	136	static ssize_t cpu_shares_show(struct kobject *kobj,
	137	struct kobj_attribute *attr,
	138	char *buf)
5cb350ba	139	{
eb41d946	140	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
24e377a8	141
eb41d946	142	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
5cb350ba DG	143	}
5cb350ba DG	144
eb41d946 KS	145	static ssize_t cpu_shares_store(struct kobject *kobj,
	146	struct kobj_attribute *attr,
	147	const char *buf, size_t size)
5cb350ba	148	{
eb41d946	149	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
5cb350ba DG	150	unsigned long shares;
	151	int rc;
	152
eb41d946	153	sscanf(buf, "%lu", &shares);
5cb350ba DG	154
	155	rc = sched_group_set_shares(up->tg, shares);
	156
	157	return (rc ? rc : size);
	158	}
	159
eb41d946 KS	160	static struct kobj_attribute cpu_share_attr =
eb41d946 KS	161	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
052f1dc7	162	#endif
eb41d946	163
052f1dc7	164	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56 PZ	165	static ssize_t cpu_rt_runtime_show(struct kobject *kobj,
	166	struct kobj_attribute *attr,
	167	char *buf)
	168	{
	169	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	170
af4491e5	171	return sprintf(buf, "%ld\n", sched_group_rt_runtime(up->tg));
9f0c1e56 PZ	172	}
	173
	174	static ssize_t cpu_rt_runtime_store(struct kobject *kobj,
	175	struct kobj_attribute *attr,
	176	const char *buf, size_t size)
	177	{
	178	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	179	unsigned long rt_runtime;
	180	int rc;
	181
af4491e5	182	sscanf(buf, "%ld", &rt_runtime);
9f0c1e56 PZ	183
	184	rc = sched_group_set_rt_runtime(up->tg, rt_runtime);
	185
	186	return (rc ? rc : size);
	187	}
	188
	189	static struct kobj_attribute cpu_rt_runtime_attr =
	190	__ATTR(cpu_rt_runtime, 0644, cpu_rt_runtime_show, cpu_rt_runtime_store);
d0b27fa7 PZ	191
	192	static ssize_t cpu_rt_period_show(struct kobject *kobj,
	193	struct kobj_attribute *attr,
	194	char *buf)
	195	{
	196	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	197
	198	return sprintf(buf, "%lu\n", sched_group_rt_period(up->tg));
	199	}
	200
	201	static ssize_t cpu_rt_period_store(struct kobject *kobj,
	202	struct kobj_attribute *attr,
	203	const char *buf, size_t size)
	204	{
	205	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	206	unsigned long rt_period;
	207	int rc;
	208
	209	sscanf(buf, "%lu", &rt_period);
	210
	211	rc = sched_group_set_rt_period(up->tg, rt_period);
	212
	213	return (rc ? rc : size);
	214	}
	215
	216	static struct kobj_attribute cpu_rt_period_attr =
	217	__ATTR(cpu_rt_period, 0644, cpu_rt_period_show, cpu_rt_period_store);
052f1dc7	218	#endif
9f0c1e56	219
eb41d946 KS	220	/* default attributes per uid directory */
eb41d946 KS	221	static struct attribute *uids_attributes[] = {
052f1dc7	222	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946	223	&cpu_share_attr.attr,
052f1dc7 PZ	224	#endif
052f1dc7 PZ	225	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56	226	&cpu_rt_runtime_attr.attr,
d0b27fa7	227	&cpu_rt_period_attr.attr,
052f1dc7	228	#endif
eb41d946 KS	229	NULL
	230	};
	231
	232	/* the lifetime of user_struct is not managed by the core (now) */
	233	static void uids_release(struct kobject *kobj)
5cb350ba	234	{
eb41d946	235	return;
5cb350ba DG	236	}
5cb350ba DG	237
eb41d946 KS	238	static struct kobj_type uids_ktype = {
	239	.sysfs_ops = &kobj_sysfs_ops,
	240	.default_attrs = uids_attributes,
	241	.release = uids_release,
	242	};
	243
94d6a5f7 SH	244	/*
	245	* Create /sys/kernel/uids/<uid>/cpu_share file for this user
	246	* We do not create this file for users in a user namespace (until
	247	* sysfs tagging is implemented).
	248	*
	249	* See Documentation/scheduler/sched-design-CFS.txt for ramifications.
	250	*/
eb41d946	251	static int uids_user_create(struct user_struct *up)
1da177e4	252	{
eb41d946	253	struct kobject *kobj = &up->kobj;
5cb350ba DG	254	int error;
5cb350ba DG	255
eb41d946	256	memset(kobj, 0, sizeof(struct kobject));
c37bbb0f SH	257	if (up->user_ns != &init_user_ns)
c37bbb0f SH	258	return 0;
eb41d946	259	kobj->kset = uids_kset;
cf15126b GKH	260	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	261	if (error) {
	262	kobject_put(kobj);
5cb350ba	263	goto done;
cf15126b	264	}
5cb350ba	265
fb7dde37	266	kobject_uevent(kobj, KOBJ_ADD);
5cb350ba DG	267	done:
5cb350ba DG	268	return error;
1da177e4 LT	269	}
1da177e4 LT	270
eb41d946	271	/* create these entries in sysfs:
5cb350ba DG	272	* "/sys/kernel/uids" directory
	273	* "/sys/kernel/uids/0" directory (for root user)
	274	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	275	*/
eb41d946	276	int __init uids_sysfs_init(void)
1da177e4	277	{
0ff21e46	278	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
eb41d946 KS	279	if (!uids_kset)
eb41d946 KS	280	return -ENOMEM;
5cb350ba	281
eb41d946	282	return uids_user_create(&root_user);
1da177e4 LT	283	}
1da177e4 LT	284
5cb350ba DG	285	/* work function to remove sysfs directory for a user and free up
	286	* corresponding structures.
	287	*/
	288	static void remove_user_sysfs_dir(struct work_struct *w)
1da177e4	289	{
5cb350ba	290	struct user_struct *up = container_of(w, struct user_struct, work);
5cb350ba DG	291	unsigned long flags;
5cb350ba DG	292	int remove_user = 0;
1da177e4	293
c37bbb0f SH	294	if (up->user_ns != &init_user_ns)
c37bbb0f SH	295	return;
5cb350ba DG	296	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	297	* atomic.
	298	*/
	299	uids_mutex_lock();
	300
	301	local_irq_save(flags);
	302
	303	if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) {
	304	uid_hash_remove(up);
	305	remove_user = 1;
	306	spin_unlock_irqrestore(&uidhash_lock, flags);
	307	} else {
	308	local_irq_restore(flags);
1da177e4 LT	309	}
1da177e4 LT	310
5cb350ba DG	311	if (!remove_user)
	312	goto done;
	313
eb41d946 KS	314	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	315	kobject_del(&up->kobj);
	316	kobject_put(&up->kobj);
5cb350ba DG	317
	318	sched_destroy_user(up);
	319	key_put(up->uid_keyring);
	320	key_put(up->session_keyring);
	321	kmem_cache_free(uid_cachep, up);
	322
	323	done:
	324	uids_mutex_unlock();
	325	}
	326
	327	/* IRQs are disabled and uidhash_lock is held upon function entry.
	328	* IRQ state (as stored in flags) is restored and uidhash_lock released
	329	* upon function exit.
	330	*/
18b6e041	331	static void free_user(struct user_struct *up, unsigned long flags)
5cb350ba DG	332	{
	333	/* restore back the count */
	334	atomic_inc(&up->__count);
	335	spin_unlock_irqrestore(&uidhash_lock, flags);
	336
18b6e041	337	put_user_ns(up->user_ns);
5cb350ba DG	338	INIT_WORK(&up->work, remove_user_sysfs_dir);
5cb350ba DG	339	schedule_work(&up->work);
1da177e4 LT	340	}
1da177e4 LT	341
052f1dc7	342	#else /* CONFIG_USER_SCHED && CONFIG_SYSFS */
5cb350ba	343
eb41d946 KS	344	int uids_sysfs_init(void) { return 0; }
eb41d946 KS	345	static inline int uids_user_create(struct user_struct *up) { return 0; }
5cb350ba DG	346	static inline void uids_mutex_lock(void) { }
	347	static inline void uids_mutex_unlock(void) { }
	348
	349	/* IRQs are disabled and uidhash_lock is held upon function entry.
	350	* IRQ state (as stored in flags) is restored and uidhash_lock released
	351	* upon function exit.
	352	*/
18b6e041	353	static void free_user(struct user_struct *up, unsigned long flags)
5cb350ba DG	354	{
	355	uid_hash_remove(up);
	356	spin_unlock_irqrestore(&uidhash_lock, flags);
	357	sched_destroy_user(up);
	358	key_put(up->uid_keyring);
	359	key_put(up->session_keyring);
18b6e041	360	put_user_ns(up->user_ns);
5cb350ba DG	361	kmem_cache_free(uid_cachep, up);
	362	}
	363
b1a8c172	364	#endif
5cb350ba	365
1da177e4 LT	366	/*
	367	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	368	* caller must undo that ref with free_uid().
	369	*
	370	* If the user_struct could not be found, return NULL.
	371	*/
	372	struct user_struct *find_user(uid_t uid)
	373	{
	374	struct user_struct *ret;
3fa97c9d	375	unsigned long flags;
6ded6ab9	376	struct user_namespace *ns = current_user_ns();
1da177e4	377
3fa97c9d	378	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	379	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	380	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	381	return ret;
	382	}
	383
	384	void free_uid(struct user_struct *up)
	385	{
3fa97c9d AM	386	unsigned long flags;
3fa97c9d AM	387
36f57413 AM	388	if (!up)
	389	return;
	390
3fa97c9d	391	local_irq_save(flags);
5cb350ba DG	392	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	393	free_user(up, flags);
	394	else
36f57413	395	local_irq_restore(flags);
1da177e4 LT	396	}
1da177e4 LT	397
354a1f4d	398	struct user_struct alloc_uid(struct user_namespace ns, uid_t uid)
1da177e4	399	{
735de223	400	struct hlist_head *hashent = uidhashentry(ns, uid);
8eb703e4	401	struct user_struct up, new;
1da177e4	402
eb41d946	403	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
5cb350ba DG	404	* atomic.
	405	*/
	406	uids_mutex_lock();
	407
3fa97c9d	408	spin_lock_irq(&uidhash_lock);
1da177e4	409	up = uid_hash_find(uid, hashent);
3fa97c9d	410	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	411
1da177e4 LT	412	if (!up) {
354a1f4d	413	new = kmem_cache_zalloc(uid_cachep, GFP_KERNEL);
8eb703e4 PE	414	if (!new)
8eb703e4 PE	415	goto out_unlock;
5e8869bb	416
1da177e4 LT	417	new->uid = uid;
1da177e4 LT	418	atomic_set(&new->__count, 1);
1da177e4	419
8eb703e4	420	if (sched_create_user(new) < 0)
69664cf1	421	goto out_free_user;
24e377a8	422
18b6e041 SH	423	new->user_ns = get_user_ns(ns);
18b6e041 SH	424
8eb703e4 PE	425	if (uids_user_create(new))
8eb703e4 PE	426	goto out_destoy_sched;
5cb350ba	427
1da177e4 LT	428	/*
	429	* Before adding this, check whether we raced
	430	* on adding the same user already..
	431	*/
3fa97c9d	432	spin_lock_irq(&uidhash_lock);
1da177e4 LT	433	up = uid_hash_find(uid, hashent);
1da177e4 LT	434	if (up) {
052f1dc7	435	/* This case is not possible when CONFIG_USER_SCHED
5cb350ba DG	436	* is defined, since we serialize alloc_uid() using
	437	* uids_mutex. Hence no need to call
	438	* sched_destroy_user() or remove_user_sysfs_dir().
	439	*/
1da177e4 LT	440	key_put(new->uid_keyring);
	441	key_put(new->session_keyring);
	442	kmem_cache_free(uid_cachep, new);
	443	} else {
	444	uid_hash_insert(new, hashent);
	445	up = new;
	446	}
3fa97c9d	447	spin_unlock_irq(&uidhash_lock);
1da177e4	448	}
5cb350ba DG	449
	450	uids_mutex_unlock();
	451
1da177e4	452	return up;
8eb703e4 PE	453
	454	out_destoy_sched:
	455	sched_destroy_user(new);
18b6e041	456	put_user_ns(new->user_ns);
8eb703e4 PE	457	out_free_user:
	458	kmem_cache_free(uid_cachep, new);
	459	out_unlock:
	460	uids_mutex_unlock();
	461	return NULL;
1da177e4 LT	462	}
1da177e4 LT	463
1da177e4 LT	464	static int __init uid_cache_init(void)
	465	{
	466	int n;
	467
	468	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	469	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	470
1da177e4 LT	471	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	472	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	473
1da177e4 LT	474	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	475	spin_lock_irq(&uidhash_lock);
acce292c	476	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	477	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	478
	479	return 0;
	480	}
	481
	482	module_init(uid_cache_init);