[linux.git] / fs / timerfd.c

/*
 *  fs/timerfd.c
 *
 *  Copyright (C) 2007  Davide Libenzi <[email protected]>
 *
 *
 *  Thanks to Thomas Gleixner for code reviews and useful comments.
 *
 */

#include <linux/file.h>
#include <linux/poll.h>
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/time.h>
#include <linux/hrtimer.h>
#include <linux/anon_inodes.h>
#include <linux/timerfd.h>
#include <linux/syscalls.h>
#include <linux/compat.h>
#include <linux/rcupdate.h>

struct timerfd_ctx {
	struct hrtimer tmr;
	ktime_t tintv;
	ktime_t moffs;
	wait_queue_head_t wqh;
	u64 ticks;
	int expired;
	int clockid;
	struct rcu_head rcu;
	struct list_head clist;
	bool might_cancel;
};

static LIST_HEAD(cancel_list);
static DEFINE_SPINLOCK(cancel_lock);

/*
 * This gets called when the timer event triggers. We set the "expired"
 * flag, but we do not re-arm the timer (in case it's necessary,
 * tintv.tv64 != 0) until the timer is accessed.
 */
static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
{
	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx, tmr);
	unsigned long flags;

	spin_lock_irqsave(&ctx->wqh.lock, flags);
	ctx->expired = 1;
	ctx->ticks++;
	wake_up_locked(&ctx->wqh);
	spin_unlock_irqrestore(&ctx->wqh.lock, flags);

	return HRTIMER_NORESTART;
}

/*
 * Called when the clock was set to cancel the timers in the cancel
 * list. This will wake up processes waiting on these timers. The
 * wake-up requires ctx->ticks to be non zero, therefore we increment
 * it before calling wake_up_locked().
 */
void timerfd_clock_was_set(void)
{
	ktime_t moffs = ktime_get_monotonic_offset();
	struct timerfd_ctx *ctx;
	unsigned long flags;

	rcu_read_lock();
	list_for_each_entry_rcu(ctx, &cancel_list, clist) {
		if (!ctx->might_cancel)
			continue;
		spin_lock_irqsave(&ctx->wqh.lock, flags);
		if (ctx->moffs.tv64 != moffs.tv64) {
			ctx->moffs.tv64 = KTIME_MAX;
			ctx->ticks++;
			wake_up_locked(&ctx->wqh);
		}
		spin_unlock_irqrestore(&ctx->wqh.lock, flags);
	}
	rcu_read_unlock();
}

static void timerfd_remove_cancel(struct timerfd_ctx *ctx)
{
	if (ctx->might_cancel) {
		ctx->might_cancel = false;
		spin_lock(&cancel_lock);
		list_del_rcu(&ctx->clist);
		spin_unlock(&cancel_lock);
	}
}

static bool timerfd_canceled(struct timerfd_ctx *ctx)
{
	if (!ctx->might_cancel || ctx->moffs.tv64 != KTIME_MAX)
		return false;
	ctx->moffs = ktime_get_monotonic_offset();
	return true;
}

static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
{
	if (ctx->clockid == CLOCK_REALTIME && (flags & TFD_TIMER_ABSTIME) &&
	    (flags & TFD_TIMER_CANCEL_ON_SET)) {
		if (!ctx->might_cancel) {
			ctx->might_cancel = true;
			spin_lock(&cancel_lock);
			list_add_rcu(&ctx->clist, &cancel_list);
			spin_unlock(&cancel_lock);
		}
	} else if (ctx->might_cancel) {
		timerfd_remove_cancel(ctx);
	}
}

static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
{
	ktime_t remaining;

	remaining = hrtimer_expires_remaining(&ctx->tmr);
	return remaining.tv64 < 0 ? ktime_set(0, 0): remaining;
}

static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
			 const struct itimerspec *ktmr)
{
	enum hrtimer_mode htmode;
	ktime_t texp;
	int clockid = ctx->clockid;

	htmode = (flags & TFD_TIMER_ABSTIME) ?
		HRTIMER_MODE_ABS: HRTIMER_MODE_REL;

	texp = timespec_to_ktime(ktmr->it_value);
	ctx->expired = 0;
	ctx->ticks = 0;
	ctx->tintv = timespec_to_ktime(ktmr->it_interval);
	hrtimer_init(&ctx->tmr, clockid, htmode);
	hrtimer_set_expires(&ctx->tmr, texp);
	ctx->tmr.function = timerfd_tmrproc;
	if (texp.tv64 != 0) {
		hrtimer_start(&ctx->tmr, texp, htmode);
		if (timerfd_canceled(ctx))
			return -ECANCELED;
	}
	return 0;
}

static int timerfd_release(struct inode *inode, struct file *file)
{
	struct timerfd_ctx *ctx = file->private_data;

	timerfd_remove_cancel(ctx);
	hrtimer_cancel(&ctx->tmr);
	kfree_rcu(ctx, rcu);
	return 0;
}

static unsigned int timerfd_poll(struct file *file, poll_table *wait)
{
	struct timerfd_ctx *ctx = file->private_data;
	unsigned int events = 0;
	unsigned long flags;

	poll_wait(file, &ctx->wqh, wait);

	spin_lock_irqsave(&ctx->wqh.lock, flags);
	if (ctx->ticks)
		events |= POLLIN;
	spin_unlock_irqrestore(&ctx->wqh.lock, flags);

	return events;
}

static ssize_t timerfd_read(struct file *file, char __user *buf, size_t count,
			    loff_t *ppos)
{
	struct timerfd_ctx *ctx = file->private_data;
	ssize_t res;
	u64 ticks = 0;

	if (count < sizeof(ticks))
		return -EINVAL;
	spin_lock_irq(&ctx->wqh.lock);
	if (file->f_flags & O_NONBLOCK)
		res = -EAGAIN;
	else
		res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);

	/*
	 * If clock has changed, we do not care about the
	 * ticks and we do not rearm the timer. Userspace must
	 * reevaluate anyway.
	 */
	if (timerfd_canceled(ctx)) {
		ctx->ticks = 0;
		ctx->expired = 0;
		res = -ECANCELED;
	}

	if (ctx->ticks) {
		ticks = ctx->ticks;

		if (ctx->expired && ctx->tintv.tv64) {
			/*
			 * If tintv.tv64 != 0, this is a periodic timer that
			 * needs to be re-armed. We avoid doing it in the timer
			 * callback to avoid DoS attacks specifying a very
			 * short timer period.
			 */
			ticks += hrtimer_forward_now(&ctx->tmr,
						     ctx->tintv) - 1;
			hrtimer_restart(&ctx->tmr);
		}
		ctx->expired = 0;
		ctx->ticks = 0;
	}
	spin_unlock_irq(&ctx->wqh.lock);
	if (ticks)
		res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
	return res;
}

static const struct file_operations timerfd_fops = {
	.release	= timerfd_release,
	.poll		= timerfd_poll,
	.read		= timerfd_read,
	.llseek		= noop_llseek,
};

static int timerfd_fget(int fd, struct fd *p)
{
	struct fd f = fdget(fd);
	if (!f.file)
		return -EBADF;
	if (f.file->f_op != &timerfd_fops) {
		fdput(f);
		return -EINVAL;
	}
	*p = f;
	return 0;
}

SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
{
	int ufd;
	struct timerfd_ctx *ctx;

	/* Check the TFD_* constants for consistency.  */
	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);

	if ((flags & ~TFD_CREATE_FLAGS) ||
	    (clockid != CLOCK_MONOTONIC &&
	     clockid != CLOCK_REALTIME))
		return -EINVAL;

	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	if (!ctx)
		return -ENOMEM;

	init_waitqueue_head(&ctx->wqh);
	ctx->clockid = clockid;
	hrtimer_init(&ctx->tmr, clockid, HRTIMER_MODE_ABS);
	ctx->moffs = ktime_get_monotonic_offset();

	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
			       O_RDWR | (flags & TFD_SHARED_FCNTL_FLAGS));
	if (ufd < 0)
		kfree(ctx);

	return ufd;
}

static int do_timerfd_settime(int ufd, int flags, 
		const struct itimerspec *new,
		struct itimerspec *old)
{
	struct fd f;
	struct timerfd_ctx *ctx;
	int ret;

	if ((flags & ~TFD_SETTIME_FLAGS) ||
	    !timespec_valid(&new->it_value) ||
	    !timespec_valid(&new->it_interval))
		return -EINVAL;

	ret = timerfd_fget(ufd, &f);
	if (ret)
		return ret;
	ctx = f.file->private_data;

	timerfd_setup_cancel(ctx, flags);

	/*
	 * We need to stop the existing timer before reprogramming
	 * it to the new values.
	 */
	for (;;) {
		spin_lock_irq(&ctx->wqh.lock);
		if (hrtimer_try_to_cancel(&ctx->tmr) >= 0)
			break;
		spin_unlock_irq(&ctx->wqh.lock);
		cpu_relax();
	}

	/*
	 * If the timer is expired and it's periodic, we need to advance it
	 * because the caller may want to know the previous expiration time.
	 * We do not update "ticks" and "expired" since the timer will be
	 * re-programmed again in the following timerfd_setup() call.
	 */
	if (ctx->expired && ctx->tintv.tv64)
		hrtimer_forward_now(&ctx->tmr, ctx->tintv);

	old->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	old->it_interval = ktime_to_timespec(ctx->tintv);

	/*
	 * Re-program the timer to the new value ...
	 */
	ret = timerfd_setup(ctx, flags, new);

	spin_unlock_irq(&ctx->wqh.lock);
	fdput(f);
	return ret;
}

static int do_timerfd_gettime(int ufd, struct itimerspec *t)
{
	struct fd f;
	struct timerfd_ctx *ctx;
	int ret = timerfd_fget(ufd, &f);
	if (ret)
		return ret;
	ctx = f.file->private_data;

	spin_lock_irq(&ctx->wqh.lock);
	if (ctx->expired && ctx->tintv.tv64) {
		ctx->expired = 0;
		ctx->ticks +=
			hrtimer_forward_now(&ctx->tmr, ctx->tintv) - 1;
		hrtimer_restart(&ctx->tmr);
	}
	t->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	t->it_interval = ktime_to_timespec(ctx->tintv);
	spin_unlock_irq(&ctx->wqh.lock);
	fdput(f);
	return 0;
}

SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
		const struct itimerspec __user *, utmr,
		struct itimerspec __user *, otmr)
{
	struct itimerspec new, old;
	int ret;

	if (copy_from_user(&new, utmr, sizeof(new)))
		return -EFAULT;
	ret = do_timerfd_settime(ufd, flags, &new, &old);
	if (ret)
		return ret;
	if (otmr && copy_to_user(otmr, &old, sizeof(old)))
		return -EFAULT;

	return ret;
}

SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
{
	struct itimerspec kotmr;
	int ret = do_timerfd_gettime(ufd, &kotmr);
	if (ret)
		return ret;
	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
}

#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
		const struct compat_itimerspec __user *, utmr,
		struct compat_itimerspec __user *, otmr)
{
	struct itimerspec new, old;
	int ret;

	if (get_compat_itimerspec(&new, utmr))
		return -EFAULT;
	ret = do_timerfd_settime(ufd, flags, &new, &old);
	if (ret)
		return ret;
	if (otmr && put_compat_itimerspec(otmr, &old))
		return -EFAULT;
	return ret;
}

COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
		struct compat_itimerspec __user *, otmr)
{
	struct itimerspec kotmr;
	int ret = do_timerfd_gettime(ufd, &kotmr);
	if (ret)
		return ret;
	return put_compat_itimerspec(otmr, &kotmr) ? -EFAULT: 0;
}
#endif
Commit	Line	Data
b215e283 DL	1	/*
	2	* fs/timerfd.c
	3	*
	4	* Copyright (C) 2007 Davide Libenzi <[email protected]>
	5	*
	6	*
	7	* Thanks to Thomas Gleixner for code reviews and useful comments.
	8	*
	9	*/
	10
	11	#include <linux/file.h>
	12	#include <linux/poll.h>
	13	#include <linux/init.h>
	14	#include <linux/fs.h>
	15	#include <linux/sched.h>
	16	#include <linux/kernel.h>
5a0e3ad6	17	#include <linux/slab.h>
b215e283 DL	18	#include <linux/list.h>
	19	#include <linux/spinlock.h>
	20	#include <linux/time.h>
	21	#include <linux/hrtimer.h>
	22	#include <linux/anon_inodes.h>
	23	#include <linux/timerfd.h>
45cc2b96	24	#include <linux/syscalls.h>
9d94b9e2	25	#include <linux/compat.h>
9ec26907	26	#include <linux/rcupdate.h>
b215e283 DL	27
	28	struct timerfd_ctx {
	29	struct hrtimer tmr;
	30	ktime_t tintv;
99ee5315	31	ktime_t moffs;
b215e283	32	wait_queue_head_t wqh;
4d672e7a	33	u64 ticks;
b215e283	34	int expired;
4d672e7a	35	int clockid;
9ec26907 TG	36	struct rcu_head rcu;
9ec26907 TG	37	struct list_head clist;
99ee5315	38	bool might_cancel;
b215e283 DL	39	};
b215e283 DL	40
9ec26907 TG	41	static LIST_HEAD(cancel_list);
	42	static DEFINE_SPINLOCK(cancel_lock);
	43
b215e283 DL	44	/*
	45	* This gets called when the timer event triggers. We set the "expired"
	46	* flag, but we do not re-arm the timer (in case it's necessary,
4d672e7a	47	* tintv.tv64 != 0) until the timer is accessed.
b215e283 DL	48	*/
	49	static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
	50	{
	51	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx, tmr);
	52	unsigned long flags;
	53
18963c01	54	spin_lock_irqsave(&ctx->wqh.lock, flags);
b215e283	55	ctx->expired = 1;
4d672e7a	56	ctx->ticks++;
b215e283	57	wake_up_locked(&ctx->wqh);
18963c01	58	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	59
	60	return HRTIMER_NORESTART;
	61	}
	62
9ec26907 TG	63	/*
9ec26907 TG	64	* Called when the clock was set to cancel the timers in the cancel
1123d939 MA	65	* list. This will wake up processes waiting on these timers. The
	66	* wake-up requires ctx->ticks to be non zero, therefore we increment
	67	* it before calling wake_up_locked().
9ec26907 TG	68	*/
9ec26907 TG	69	void timerfd_clock_was_set(void)
4d672e7a	70	{
9ec26907 TG	71	ktime_t moffs = ktime_get_monotonic_offset();
	72	struct timerfd_ctx *ctx;
	73	unsigned long flags;
4d672e7a	74
9ec26907 TG	75	rcu_read_lock();
	76	list_for_each_entry_rcu(ctx, &cancel_list, clist) {
	77	if (!ctx->might_cancel)
	78	continue;
	79	spin_lock_irqsave(&ctx->wqh.lock, flags);
	80	if (ctx->moffs.tv64 != moffs.tv64) {
	81	ctx->moffs.tv64 = KTIME_MAX;
1123d939	82	ctx->ticks++;
9ec26907 TG	83	wake_up_locked(&ctx->wqh);
	84	}
	85	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
	86	}
	87	rcu_read_unlock();
4d672e7a DL	88	}
4d672e7a DL	89
9ec26907	90	static void timerfd_remove_cancel(struct timerfd_ctx *ctx)
99ee5315	91	{
9ec26907 TG	92	if (ctx->might_cancel) {
	93	ctx->might_cancel = false;
	94	spin_lock(&cancel_lock);
	95	list_del_rcu(&ctx->clist);
	96	spin_unlock(&cancel_lock);
	97	}
	98	}
99ee5315	99
9ec26907 TG	100	static bool timerfd_canceled(struct timerfd_ctx *ctx)
	101	{
	102	if (!ctx->might_cancel \|\| ctx->moffs.tv64 != KTIME_MAX)
99ee5315	103	return false;
9ec26907 TG	104	ctx->moffs = ktime_get_monotonic_offset();
	105	return true;
	106	}
99ee5315	107
9ec26907 TG	108	static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
	109	{
	110	if (ctx->clockid == CLOCK_REALTIME && (flags & TFD_TIMER_ABSTIME) &&
	111	(flags & TFD_TIMER_CANCEL_ON_SET)) {
	112	if (!ctx->might_cancel) {
	113	ctx->might_cancel = true;
	114	spin_lock(&cancel_lock);
	115	list_add_rcu(&ctx->clist, &cancel_list);
	116	spin_unlock(&cancel_lock);
	117	}
	118	} else if (ctx->might_cancel) {
	119	timerfd_remove_cancel(ctx);
	120	}
	121	}
99ee5315	122
9ec26907 TG	123	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
	124	{
	125	ktime_t remaining;
99ee5315	126
9ec26907 TG	127	remaining = hrtimer_expires_remaining(&ctx->tmr);
9ec26907 TG	128	return remaining.tv64 < 0 ? ktime_set(0, 0): remaining;
99ee5315 TG	129	}
	130
	131	static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
	132	const struct itimerspec *ktmr)
b215e283 DL	133	{
	134	enum hrtimer_mode htmode;
	135	ktime_t texp;
99ee5315	136	int clockid = ctx->clockid;
b215e283 DL	137
	138	htmode = (flags & TFD_TIMER_ABSTIME) ?
	139	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	140
	141	texp = timespec_to_ktime(ktmr->it_value);
	142	ctx->expired = 0;
4d672e7a	143	ctx->ticks = 0;
b215e283	144	ctx->tintv = timespec_to_ktime(ktmr->it_interval);
99ee5315	145	hrtimer_init(&ctx->tmr, clockid, htmode);
76369470	146	hrtimer_set_expires(&ctx->tmr, texp);
b215e283	147	ctx->tmr.function = timerfd_tmrproc;
99ee5315	148	if (texp.tv64 != 0) {
b215e283	149	hrtimer_start(&ctx->tmr, texp, htmode);
99ee5315 TG	150	if (timerfd_canceled(ctx))
	151	return -ECANCELED;
	152	}
	153	return 0;
b215e283 DL	154	}
	155
	156	static int timerfd_release(struct inode inode, struct file file)
	157	{
	158	struct timerfd_ctx *ctx = file->private_data;
	159
9ec26907	160	timerfd_remove_cancel(ctx);
b215e283	161	hrtimer_cancel(&ctx->tmr);
9ec26907	162	kfree_rcu(ctx, rcu);
b215e283 DL	163	return 0;
	164	}
	165
	166	static unsigned int timerfd_poll(struct file file, poll_table wait)
	167	{
	168	struct timerfd_ctx *ctx = file->private_data;
	169	unsigned int events = 0;
	170	unsigned long flags;
	171
	172	poll_wait(file, &ctx->wqh, wait);
	173
18963c01	174	spin_lock_irqsave(&ctx->wqh.lock, flags);
4d672e7a	175	if (ctx->ticks)
b215e283	176	events \|= POLLIN;
18963c01	177	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	178
	179	return events;
	180	}
	181
	182	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	183	loff_t *ppos)
	184	{
	185	struct timerfd_ctx *ctx = file->private_data;
	186	ssize_t res;
09828402	187	u64 ticks = 0;
b215e283 DL	188
	189	if (count < sizeof(ticks))
	190	return -EINVAL;
18963c01	191	spin_lock_irq(&ctx->wqh.lock);
8120a8aa MN	192	if (file->f_flags & O_NONBLOCK)
	193	res = -EAGAIN;
	194	else
	195	res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);
99ee5315	196
9ec26907 TG	197	/*
	198	* If clock has changed, we do not care about the
	199	* ticks and we do not rearm the timer. Userspace must
	200	* reevaluate anyway.
	201	*/
	202	if (timerfd_canceled(ctx)) {
	203	ctx->ticks = 0;
	204	ctx->expired = 0;
	205	res = -ECANCELED;
	206	}
	207
4d672e7a DL	208	if (ctx->ticks) {
4d672e7a DL	209	ticks = ctx->ticks;
99ee5315	210
4d672e7a	211	if (ctx->expired && ctx->tintv.tv64) {
b215e283 DL	212	/*
	213	* If tintv.tv64 != 0, this is a periodic timer that
	214	* needs to be re-armed. We avoid doing it in the timer
	215	* callback to avoid DoS attacks specifying a very
	216	* short timer period.
	217	*/
4d672e7a DL	218	ticks += hrtimer_forward_now(&ctx->tmr,
4d672e7a DL	219	ctx->tintv) - 1;
b215e283	220	hrtimer_restart(&ctx->tmr);
4d672e7a DL	221	}
	222	ctx->expired = 0;
	223	ctx->ticks = 0;
b215e283	224	}
18963c01	225	spin_unlock_irq(&ctx->wqh.lock);
b215e283	226	if (ticks)
09828402	227	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	228	return res;
	229	}
	230
	231	static const struct file_operations timerfd_fops = {
	232	.release = timerfd_release,
	233	.poll = timerfd_poll,
	234	.read = timerfd_read,
6038f373	235	.llseek = noop_llseek,
b215e283 DL	236	};
b215e283 DL	237
2903ff01	238	static int timerfd_fget(int fd, struct fd *p)
4d672e7a	239	{
2903ff01 AV	240	struct fd f = fdget(fd);
	241	if (!f.file)
	242	return -EBADF;
	243	if (f.file->f_op != &timerfd_fops) {
	244	fdput(f);
	245	return -EINVAL;
4d672e7a	246	}
2903ff01 AV	247	*p = f;
2903ff01 AV	248	return 0;
4d672e7a DL	249	}
4d672e7a DL	250
836f92ad	251	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	252	{
2030a42c	253	int ufd;
b215e283	254	struct timerfd_ctx *ctx;
b215e283	255
e38b36f3 UD	256	/* Check the TFD_* constants for consistency. */
	257	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	258	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	259
610d18f4 DL	260	if ((flags & ~TFD_CREATE_FLAGS) \|\|
	261	(clockid != CLOCK_MONOTONIC &&
	262	clockid != CLOCK_REALTIME))
b215e283	263	return -EINVAL;
4d672e7a DL	264
	265	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	266	if (!ctx)
	267	return -ENOMEM;
	268
	269	init_waitqueue_head(&ctx->wqh);
	270	ctx->clockid = clockid;
	271	hrtimer_init(&ctx->tmr, clockid, HRTIMER_MODE_ABS);
99ee5315	272	ctx->moffs = ktime_get_monotonic_offset();
4d672e7a	273
11fcb6c1	274	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	275	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	276	if (ufd < 0)
4d672e7a	277	kfree(ctx);
4d672e7a DL	278
	279	return ufd;
	280	}
	281
9d94b9e2 AV	282	static int do_timerfd_settime(int ufd, int flags,
	283	const struct itimerspec *new,
	284	struct itimerspec *old)
4d672e7a	285	{
2903ff01	286	struct fd f;
4d672e7a	287	struct timerfd_ctx *ctx;
2903ff01	288	int ret;
4d672e7a	289
610d18f4	290	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
9d94b9e2 AV	291	!timespec_valid(&new->it_value) \|\|
9d94b9e2 AV	292	!timespec_valid(&new->it_interval))
b215e283 DL	293	return -EINVAL;
b215e283 DL	294
2903ff01 AV	295	ret = timerfd_fget(ufd, &f);
	296	if (ret)
	297	return ret;
	298	ctx = f.file->private_data;
b215e283	299
9ec26907 TG	300	timerfd_setup_cancel(ctx, flags);
9ec26907 TG	301
4d672e7a DL	302	/*
	303	* We need to stop the existing timer before reprogramming
	304	* it to the new values.
	305	*/
	306	for (;;) {
	307	spin_lock_irq(&ctx->wqh.lock);
	308	if (hrtimer_try_to_cancel(&ctx->tmr) >= 0)
	309	break;
18963c01	310	spin_unlock_irq(&ctx->wqh.lock);
4d672e7a	311	cpu_relax();
b215e283 DL	312	}
b215e283 DL	313
4d672e7a DL	314	/*
	315	* If the timer is expired and it's periodic, we need to advance it
	316	* because the caller may want to know the previous expiration time.
	317	* We do not update "ticks" and "expired" since the timer will be
	318	* re-programmed again in the following timerfd_setup() call.
	319	*/
	320	if (ctx->expired && ctx->tintv.tv64)
	321	hrtimer_forward_now(&ctx->tmr, ctx->tintv);
b215e283	322
9d94b9e2 AV	323	old->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	324	old->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a DL	325
	326	/*
	327	* Re-program the timer to the new value ...
	328	*/
9d94b9e2	329	ret = timerfd_setup(ctx, flags, new);
4d672e7a DL	330
4d672e7a DL	331	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	332	fdput(f);
99ee5315	333	return ret;
4d672e7a DL	334	}
4d672e7a DL	335
9d94b9e2	336	static int do_timerfd_gettime(int ufd, struct itimerspec *t)
4d672e7a	337	{
2903ff01	338	struct fd f;
4d672e7a	339	struct timerfd_ctx *ctx;
2903ff01 AV	340	int ret = timerfd_fget(ufd, &f);
	341	if (ret)
	342	return ret;
	343	ctx = f.file->private_data;
4d672e7a DL	344
	345	spin_lock_irq(&ctx->wqh.lock);
	346	if (ctx->expired && ctx->tintv.tv64) {
	347	ctx->expired = 0;
	348	ctx->ticks +=
	349	hrtimer_forward_now(&ctx->tmr, ctx->tintv) - 1;
	350	hrtimer_restart(&ctx->tmr);
	351	}
9d94b9e2 AV	352	t->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	353	t->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a	354	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	355	fdput(f);
9d94b9e2 AV	356	return 0;
	357	}
	358
	359	SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
	360	const struct itimerspec __user *, utmr,
	361	struct itimerspec __user *, otmr)
	362	{
	363	struct itimerspec new, old;
	364	int ret;
	365
	366	if (copy_from_user(&new, utmr, sizeof(new)))
	367	return -EFAULT;
	368	ret = do_timerfd_settime(ufd, flags, &new, &old);
	369	if (ret)
	370	return ret;
	371	if (otmr && copy_to_user(otmr, &old, sizeof(old)))
	372	return -EFAULT;
	373
	374	return ret;
	375	}
4d672e7a	376
9d94b9e2 AV	377	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
	378	{
	379	struct itimerspec kotmr;
	380	int ret = do_timerfd_gettime(ufd, &kotmr);
	381	if (ret)
	382	return ret;
4d672e7a	383	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
b215e283 DL	384	}
b215e283 DL	385
0e803baf	386	#ifdef CONFIG_COMPAT
9d94b9e2	387	COMPAT_SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
0e803baf HC	388	const struct compat_itimerspec __user *, utmr,
0e803baf HC	389	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	390	{
	391	struct itimerspec new, old;
	392	int ret;
	393
	394	if (get_compat_itimerspec(&new, utmr))
	395	return -EFAULT;
	396	ret = do_timerfd_settime(ufd, flags, &new, &old);
	397	if (ret)
	398	return ret;
	399	if (otmr && put_compat_itimerspec(otmr, &old))
	400	return -EFAULT;
	401	return ret;
	402	}
	403
	404	COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
0e803baf	405	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	406	{
	407	struct itimerspec kotmr;
	408	int ret = do_timerfd_gettime(ufd, &kotmr);
	409	if (ret)
	410	return ret;
0e803baf	411	return put_compat_itimerspec(otmr, &kotmr) ? -EFAULT: 0;
9d94b9e2 AV	412	}
9d94b9e2 AV	413	#endif