[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(2),
	},
	.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);
	put_user_ns(up->user_ns);
}

#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 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) {
			/* possibly resurrect an "almost deleted" object */
			if (atomic_inc_return(&user->__count) == 1)
				cancel_delayed_work(&user->work);
			return user;
		}
	}

	return NULL;
}

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);
}

/* delayed work function to remove sysfs directory for a user and free up
 * corresponding structures.
 */
static void cleanup_user_struct(struct work_struct *w)
{
	struct user_struct *up = container_of(w, struct user_struct, work.work);
	unsigned long flags;
	int remove_user = 0;

	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	 * atomic.
	 */
	uids_mutex_lock();

	spin_lock_irqsave(&uidhash_lock, flags);
	if (atomic_read(&up->__count) == 0) {
		uid_hash_remove(up);
		remove_user = 1;
	}
	spin_unlock_irqrestore(&uidhash_lock, flags);

	if (!remove_user)
		goto done;

	if (up->user_ns == &init_user_ns) {
		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)
{
	INIT_DELAYED_WORK(&up->work, cleanup_user_struct);
	schedule_delayed_work(&up->work, msecs_to_jiffies(1000));
	spin_unlock_irqrestore(&uidhash_lock, flags);
}

#else	/* CONFIG_USER_SCHED && CONFIG_SYSFS */

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;
}

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);
	kmem_cache_free(uid_cachep, up);
}

#endif

#if defined(CONFIG_RT_GROUP_SCHED) && defined(CONFIG_USER_SCHED)
/*
 * We need to check if a setuid can take place. This function should be called
 * before successfully completing the setuid.
 */
int task_can_switch_user(struct user_struct *up, struct task_struct *tsk)
{

	return sched_rt_can_attach(up->tg, tsk);

}
#else
int task_can_switch_user(struct user_struct *up, struct task_struct *tsk)
{
	return 1;
}
#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 = {
1d1e9756	23	.refcount = ATOMIC_INIT(2),
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	{
5cb350ba DG	74	hlist_del_init(&up->uidhash_node);
fb5ae64f	75	put_user_ns(up->user_ns);
5cb350ba DG	76	}
5cb350ba DG	77
052f1dc7	78	#ifdef CONFIG_USER_SCHED
5cb350ba	79
24e377a8 SV	80	static void sched_destroy_user(struct user_struct *up)
	81	{
	82	sched_destroy_group(up->tg);
	83	}
	84
	85	static int sched_create_user(struct user_struct *up)
	86	{
	87	int rc = 0;
	88
eff766a6	89	up->tg = sched_create_group(&root_task_group);
24e377a8 SV	90	if (IS_ERR(up->tg))
	91	rc = -ENOMEM;
	92
6c415b92 AB	93	set_tg_uid(up);
6c415b92 AB	94
24e377a8 SV	95	return rc;
	96	}
	97
052f1dc7	98	#else /* CONFIG_USER_SCHED */
b1a8c172 DG	99
	100	static void sched_destroy_user(struct user_struct *up) { }
	101	static int sched_create_user(struct user_struct *up) { return 0; }
b1a8c172	102
052f1dc7	103	#endif /* CONFIG_USER_SCHED */
b1a8c172	104
052f1dc7	105	#if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS)
b1a8c172	106
3959214f KS	107	static struct user_struct uid_hash_find(uid_t uid, struct hlist_head hashent)
	108	{
	109	struct user_struct *user;
	110	struct hlist_node *h;
	111
	112	hlist_for_each_entry(user, h, hashent, uidhash_node) {
	113	if (user->uid == uid) {
	114	/* possibly resurrect an "almost deleted" object */
	115	if (atomic_inc_return(&user->__count) == 1)
	116	cancel_delayed_work(&user->work);
	117	return user;
	118	}
	119	}
	120
	121	return NULL;
	122	}
	123
eb41d946	124	static struct kset uids_kset; / represents the /sys/kernel/uids/ directory */
b1a8c172 DG	125	static DEFINE_MUTEX(uids_mutex);
b1a8c172 DG	126
5cb350ba DG	127	static inline void uids_mutex_lock(void)
	128	{
	129	mutex_lock(&uids_mutex);
	130	}
24e377a8	131
5cb350ba DG	132	static inline void uids_mutex_unlock(void)
	133	{
	134	mutex_unlock(&uids_mutex);
	135	}
24e377a8	136
eb41d946	137	/* uid directory attributes */
052f1dc7	138	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946 KS	139	static ssize_t cpu_shares_show(struct kobject *kobj,
	140	struct kobj_attribute *attr,
	141	char *buf)
5cb350ba	142	{
eb41d946	143	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
24e377a8	144
eb41d946	145	return sprintf(buf, "%lu\n", sched_group_shares(up->tg));
5cb350ba DG	146	}
5cb350ba DG	147
eb41d946 KS	148	static ssize_t cpu_shares_store(struct kobject *kobj,
	149	struct kobj_attribute *attr,
	150	const char *buf, size_t size)
5cb350ba	151	{
eb41d946	152	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
5cb350ba DG	153	unsigned long shares;
	154	int rc;
	155
eb41d946	156	sscanf(buf, "%lu", &shares);
5cb350ba DG	157
	158	rc = sched_group_set_shares(up->tg, shares);
	159
	160	return (rc ? rc : size);
	161	}
	162
eb41d946 KS	163	static struct kobj_attribute cpu_share_attr =
eb41d946 KS	164	__ATTR(cpu_share, 0644, cpu_shares_show, cpu_shares_store);
052f1dc7	165	#endif
eb41d946	166
052f1dc7	167	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56 PZ	168	static ssize_t cpu_rt_runtime_show(struct kobject *kobj,
	169	struct kobj_attribute *attr,
	170	char *buf)
	171	{
	172	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	173
af4491e5	174	return sprintf(buf, "%ld\n", sched_group_rt_runtime(up->tg));
9f0c1e56 PZ	175	}
	176
	177	static ssize_t cpu_rt_runtime_store(struct kobject *kobj,
	178	struct kobj_attribute *attr,
	179	const char *buf, size_t size)
	180	{
	181	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	182	unsigned long rt_runtime;
	183	int rc;
	184
af4491e5	185	sscanf(buf, "%ld", &rt_runtime);
9f0c1e56 PZ	186
	187	rc = sched_group_set_rt_runtime(up->tg, rt_runtime);
	188
	189	return (rc ? rc : size);
	190	}
	191
	192	static struct kobj_attribute cpu_rt_runtime_attr =
	193	__ATTR(cpu_rt_runtime, 0644, cpu_rt_runtime_show, cpu_rt_runtime_store);
d0b27fa7 PZ	194
	195	static ssize_t cpu_rt_period_show(struct kobject *kobj,
	196	struct kobj_attribute *attr,
	197	char *buf)
	198	{
	199	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	200
	201	return sprintf(buf, "%lu\n", sched_group_rt_period(up->tg));
	202	}
	203
	204	static ssize_t cpu_rt_period_store(struct kobject *kobj,
	205	struct kobj_attribute *attr,
	206	const char *buf, size_t size)
	207	{
	208	struct user_struct *up = container_of(kobj, struct user_struct, kobj);
	209	unsigned long rt_period;
	210	int rc;
	211
	212	sscanf(buf, "%lu", &rt_period);
	213
	214	rc = sched_group_set_rt_period(up->tg, rt_period);
	215
	216	return (rc ? rc : size);
	217	}
	218
	219	static struct kobj_attribute cpu_rt_period_attr =
	220	__ATTR(cpu_rt_period, 0644, cpu_rt_period_show, cpu_rt_period_store);
052f1dc7	221	#endif
9f0c1e56	222
eb41d946 KS	223	/* default attributes per uid directory */
eb41d946 KS	224	static struct attribute *uids_attributes[] = {
052f1dc7	225	#ifdef CONFIG_FAIR_GROUP_SCHED
eb41d946	226	&cpu_share_attr.attr,
052f1dc7 PZ	227	#endif
052f1dc7 PZ	228	#ifdef CONFIG_RT_GROUP_SCHED
9f0c1e56	229	&cpu_rt_runtime_attr.attr,
d0b27fa7	230	&cpu_rt_period_attr.attr,
052f1dc7	231	#endif
eb41d946 KS	232	NULL
	233	};
	234
	235	/* the lifetime of user_struct is not managed by the core (now) */
	236	static void uids_release(struct kobject *kobj)
5cb350ba	237	{
eb41d946	238	return;
5cb350ba DG	239	}
5cb350ba DG	240
eb41d946 KS	241	static struct kobj_type uids_ktype = {
	242	.sysfs_ops = &kobj_sysfs_ops,
	243	.default_attrs = uids_attributes,
	244	.release = uids_release,
	245	};
	246
94d6a5f7 SH	247	/*
	248	* Create /sys/kernel/uids/<uid>/cpu_share file for this user
	249	* We do not create this file for users in a user namespace (until
	250	* sysfs tagging is implemented).
	251	*
	252	* See Documentation/scheduler/sched-design-CFS.txt for ramifications.
	253	*/
eb41d946	254	static int uids_user_create(struct user_struct *up)
1da177e4	255	{
eb41d946	256	struct kobject *kobj = &up->kobj;
5cb350ba DG	257	int error;
5cb350ba DG	258
eb41d946	259	memset(kobj, 0, sizeof(struct kobject));
c37bbb0f SH	260	if (up->user_ns != &init_user_ns)
c37bbb0f SH	261	return 0;
eb41d946	262	kobj->kset = uids_kset;
cf15126b GKH	263	error = kobject_init_and_add(kobj, &uids_ktype, NULL, "%d", up->uid);
	264	if (error) {
	265	kobject_put(kobj);
5cb350ba	266	goto done;
cf15126b	267	}
5cb350ba	268
fb7dde37	269	kobject_uevent(kobj, KOBJ_ADD);
5cb350ba DG	270	done:
5cb350ba DG	271	return error;
1da177e4 LT	272	}
1da177e4 LT	273
eb41d946	274	/* create these entries in sysfs:
5cb350ba DG	275	* "/sys/kernel/uids" directory
	276	* "/sys/kernel/uids/0" directory (for root user)
	277	* "/sys/kernel/uids/0/cpu_share" file (for root user)
	278	*/
eb41d946	279	int __init uids_sysfs_init(void)
1da177e4	280	{
0ff21e46	281	uids_kset = kset_create_and_add("uids", NULL, kernel_kobj);
eb41d946 KS	282	if (!uids_kset)
eb41d946 KS	283	return -ENOMEM;
5cb350ba	284
eb41d946	285	return uids_user_create(&root_user);
1da177e4 LT	286	}
1da177e4 LT	287
3959214f	288	/* delayed work function to remove sysfs directory for a user and free up
5cb350ba DG	289	* corresponding structures.
5cb350ba DG	290	*/
be50b834	291	static void cleanup_user_struct(struct work_struct *w)
1da177e4	292	{
3959214f	293	struct user_struct *up = container_of(w, struct user_struct, work.work);
5cb350ba DG	294	unsigned long flags;
5cb350ba DG	295	int remove_user = 0;
1da177e4	296
5cb350ba DG	297	/* Make uid_hash_remove() + sysfs_remove_file() + kobject_del()
	298	* atomic.
	299	*/
	300	uids_mutex_lock();
	301
3959214f KS	302	spin_lock_irqsave(&uidhash_lock, flags);
3959214f KS	303	if (atomic_read(&up->__count) == 0) {
5cb350ba DG	304	uid_hash_remove(up);
5cb350ba DG	305	remove_user = 1;
1da177e4	306	}
3959214f	307	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4	308
5cb350ba DG	309	if (!remove_user)
	310	goto done;
	311
be50b834 DG	312	if (up->user_ns == &init_user_ns) {
	313	kobject_uevent(&up->kobj, KOBJ_REMOVE);
	314	kobject_del(&up->kobj);
	315	kobject_put(&up->kobj);
	316	}
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	332	{
3959214f KS	333	INIT_DELAYED_WORK(&up->work, cleanup_user_struct);
3959214f KS	334	schedule_delayed_work(&up->work, msecs_to_jiffies(1000));
b00bc0b2	335	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	336	}
1da177e4 LT	337
052f1dc7	338	#else /* CONFIG_USER_SCHED && CONFIG_SYSFS */
5cb350ba	339
3959214f KS	340	static struct user_struct uid_hash_find(uid_t uid, struct hlist_head hashent)
	341	{
	342	struct user_struct *user;
	343	struct hlist_node *h;
	344
	345	hlist_for_each_entry(user, h, hashent, uidhash_node) {
	346	if (user->uid == uid) {
	347	atomic_inc(&user->__count);
	348	return user;
	349	}
	350	}
	351
	352	return NULL;
	353	}
	354
eb41d946 KS	355	int uids_sysfs_init(void) { return 0; }
eb41d946 KS	356	static inline int uids_user_create(struct user_struct *up) { return 0; }
5cb350ba DG	357	static inline void uids_mutex_lock(void) { }
	358	static inline void uids_mutex_unlock(void) { }
	359
	360	/* IRQs are disabled and uidhash_lock is held upon function entry.
	361	* IRQ state (as stored in flags) is restored and uidhash_lock released
	362	* upon function exit.
	363	*/
18b6e041	364	static void free_user(struct user_struct *up, unsigned long flags)
5cb350ba DG	365	{
	366	uid_hash_remove(up);
	367	spin_unlock_irqrestore(&uidhash_lock, flags);
	368	sched_destroy_user(up);
	369	key_put(up->uid_keyring);
	370	key_put(up->session_keyring);
	371	kmem_cache_free(uid_cachep, up);
	372	}
	373
b1a8c172	374	#endif
5cb350ba	375
54e99124 DG	376	#if defined(CONFIG_RT_GROUP_SCHED) && defined(CONFIG_USER_SCHED)
	377	/*
	378	* We need to check if a setuid can take place. This function should be called
	379	* before successfully completing the setuid.
	380	*/
	381	int task_can_switch_user(struct user_struct up, struct task_struct tsk)
	382	{
	383
	384	return sched_rt_can_attach(up->tg, tsk);
	385
	386	}
	387	#else
	388	int task_can_switch_user(struct user_struct up, struct task_struct tsk)
	389	{
	390	return 1;
	391	}
	392	#endif
	393
1da177e4 LT	394	/*
	395	* Locate the user_struct for the passed UID. If found, take a ref on it. The
	396	* caller must undo that ref with free_uid().
	397	*
	398	* If the user_struct could not be found, return NULL.
	399	*/
	400	struct user_struct *find_user(uid_t uid)
	401	{
	402	struct user_struct *ret;
3fa97c9d	403	unsigned long flags;
6ded6ab9	404	struct user_namespace *ns = current_user_ns();
1da177e4	405
3fa97c9d	406	spin_lock_irqsave(&uidhash_lock, flags);
acce292c	407	ret = uid_hash_find(uid, uidhashentry(ns, uid));
3fa97c9d	408	spin_unlock_irqrestore(&uidhash_lock, flags);
1da177e4 LT	409	return ret;
	410	}
	411
	412	void free_uid(struct user_struct *up)
	413	{
3fa97c9d AM	414	unsigned long flags;
3fa97c9d AM	415
36f57413 AM	416	if (!up)
	417	return;
	418
3fa97c9d	419	local_irq_save(flags);
5cb350ba DG	420	if (atomic_dec_and_lock(&up->__count, &uidhash_lock))
	421	free_user(up, flags);
	422	else
36f57413	423	local_irq_restore(flags);
1da177e4 LT	424	}
1da177e4 LT	425
354a1f4d	426	struct user_struct alloc_uid(struct user_namespace ns, uid_t uid)
1da177e4	427	{
735de223	428	struct hlist_head *hashent = uidhashentry(ns, uid);
8eb703e4	429	struct user_struct up, new;
1da177e4	430
eb41d946	431	/* Make uid_hash_find() + uids_user_create() + uid_hash_insert()
5cb350ba DG	432	* atomic.
	433	*/
	434	uids_mutex_lock();
	435
3fa97c9d	436	spin_lock_irq(&uidhash_lock);
1da177e4	437	up = uid_hash_find(uid, hashent);
3fa97c9d	438	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	439
1da177e4 LT	440	if (!up) {
354a1f4d	441	new = kmem_cache_zalloc(uid_cachep, GFP_KERNEL);
8eb703e4 PE	442	if (!new)
8eb703e4 PE	443	goto out_unlock;
5e8869bb	444
1da177e4 LT	445	new->uid = uid;
1da177e4 LT	446	atomic_set(&new->__count, 1);
1da177e4	447
8eb703e4	448	if (sched_create_user(new) < 0)
69664cf1	449	goto out_free_user;
24e377a8	450
18b6e041 SH	451	new->user_ns = get_user_ns(ns);
18b6e041 SH	452
8eb703e4 PE	453	if (uids_user_create(new))
8eb703e4 PE	454	goto out_destoy_sched;
5cb350ba	455
1da177e4 LT	456	/*
	457	* Before adding this, check whether we raced
	458	* on adding the same user already..
	459	*/
3fa97c9d	460	spin_lock_irq(&uidhash_lock);
1da177e4 LT	461	up = uid_hash_find(uid, hashent);
1da177e4 LT	462	if (up) {
052f1dc7	463	/* This case is not possible when CONFIG_USER_SCHED
5cb350ba DG	464	* is defined, since we serialize alloc_uid() using
	465	* uids_mutex. Hence no need to call
	466	* sched_destroy_user() or remove_user_sysfs_dir().
	467	*/
1da177e4 LT	468	key_put(new->uid_keyring);
	469	key_put(new->session_keyring);
	470	kmem_cache_free(uid_cachep, new);
	471	} else {
	472	uid_hash_insert(new, hashent);
	473	up = new;
	474	}
3fa97c9d	475	spin_unlock_irq(&uidhash_lock);
1da177e4	476	}
5cb350ba DG	477
	478	uids_mutex_unlock();
	479
1da177e4	480	return up;
8eb703e4 PE	481
	482	out_destoy_sched:
	483	sched_destroy_user(new);
18b6e041	484	put_user_ns(new->user_ns);
8eb703e4 PE	485	out_free_user:
	486	kmem_cache_free(uid_cachep, new);
	487	out_unlock:
	488	uids_mutex_unlock();
	489	return NULL;
1da177e4 LT	490	}
1da177e4 LT	491
1da177e4 LT	492	static int __init uid_cache_init(void)
	493	{
	494	int n;
	495
	496	uid_cachep = kmem_cache_create("uid_cache", sizeof(struct user_struct),
20c2df83	497	0, SLAB_HWCACHE_ALIGN\|SLAB_PANIC, NULL);
1da177e4 LT	498
1da177e4 LT	499	for(n = 0; n < UIDHASH_SZ; ++n)
735de223	500	INIT_HLIST_HEAD(init_user_ns.uidhash_table + n);
1da177e4 LT	501
1da177e4 LT	502	/* Insert the root user immediately (init already runs as root) */
3fa97c9d	503	spin_lock_irq(&uidhash_lock);
acce292c	504	uid_hash_insert(&root_user, uidhashentry(&init_user_ns, 0));
3fa97c9d	505	spin_unlock_irq(&uidhash_lock);
1da177e4 LT	506
	507	return 0;
	508	}
	509
	510	module_init(uid_cache_init);