[GitHub/LineageOS/android_kernel_samsung_universal7580.git] / fs / timerfd.c

/*
 *  fs/timerfd.c
 *
 *  Copyright (C) 2007  Davide Libenzi <davidel@xmailserver.org>
 *
 *
 *  Thanks to Thomas Gleixner for code reviews and useful comments.
 *
 */

#include <linux/alarmtimer.h>
#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 {
	union {
		struct hrtimer tmr;
		struct alarm alarm;
	} t;
	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;
	spinlock_t cancel_lock;
	bool might_cancel;
};

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

static inline bool isalarm(struct timerfd_ctx *ctx)
{
	return ctx->clockid == CLOCK_REALTIME_ALARM ||
		ctx->clockid == CLOCK_BOOTTIME_ALARM;
}

/*
 * 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 void timerfd_triggered(struct timerfd_ctx *ctx)
{
	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);
}

static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
{
	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx,
					       t.tmr);
	timerfd_triggered(ctx);
	return HRTIMER_NORESTART;
}

static enum alarmtimer_restart timerfd_alarmproc(struct alarm *alarm,
	ktime_t now)
{
	struct timerfd_ctx *ctx = container_of(alarm, struct timerfd_ctx,
					       t.alarm);
	timerfd_triggered(ctx);
	return ALARMTIMER_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 void timerfd_remove_cancel(struct timerfd_ctx *ctx)
{
	spin_lock(&ctx->cancel_lock);
	__timerfd_remove_cancel(ctx);
	spin_unlock(&ctx->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)
{
	spin_lock(&ctx->cancel_lock);
	if ((ctx->clockid == CLOCK_REALTIME ||
	     ctx->clockid == CLOCK_REALTIME_ALARM) &&
	    (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 {
		__timerfd_remove_cancel(ctx);
	}
	spin_unlock(&ctx->cancel_lock);
}

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

	if (isalarm(ctx))
		remaining = alarm_expires_remaining(&ctx->t.alarm);
	else
		remaining = hrtimer_expires_remaining(&ctx->t.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);

	if (isalarm(ctx)) {
		alarm_init(&ctx->t.alarm,
			   ctx->clockid == CLOCK_REALTIME_ALARM ?
			   ALARM_REALTIME : ALARM_BOOTTIME,
			   timerfd_alarmproc);
	} else {
		hrtimer_init(&ctx->t.tmr, clockid, htmode);
		hrtimer_set_expires(&ctx->t.tmr, texp);
		ctx->t.tmr.function = timerfd_tmrproc;
	}

	if (texp.tv64 != 0) {
		if (isalarm(ctx)) {
			if (flags & TFD_TIMER_ABSTIME)
				alarm_start(&ctx->t.alarm, texp);
			else
				alarm_start_relative(&ctx->t.alarm, texp);
		} else {
			hrtimer_start(&ctx->t.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);

	if (isalarm(ctx))
		alarm_cancel(&ctx->t.alarm);
	else
		hrtimer_cancel(&ctx->t.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.
			 */
			if (isalarm(ctx)) {
				ticks += alarm_forward_now(
					&ctx->t.alarm, ctx->tintv) - 1;
				alarm_restart(&ctx->t.alarm);
			} else {
				ticks += hrtimer_forward_now(&ctx->t.tmr,
							     ctx->tintv) - 1;
				hrtimer_restart(&ctx->t.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 &&
	     clockid != CLOCK_REALTIME_ALARM &&
	     clockid != CLOCK_BOOTTIME &&
	     clockid != CLOCK_BOOTTIME_ALARM))
		return -EINVAL;

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

	init_waitqueue_head(&ctx->wqh);
	spin_lock_init(&ctx->cancel_lock);
	ctx->clockid = clockid;

	if (isalarm(ctx))
		alarm_init(&ctx->t.alarm,
			   ctx->clockid == CLOCK_REALTIME_ALARM ?
			   ALARM_REALTIME : ALARM_BOOTTIME,
			   timerfd_alarmproc);
	else
		hrtimer_init(&ctx->t.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 (isalarm(ctx)) {
			if (alarm_try_to_cancel(&ctx->t.alarm) >= 0)
				break;
		} else {
			if (hrtimer_try_to_cancel(&ctx->t.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) {
		if (isalarm(ctx))
			alarm_forward_now(&ctx->t.alarm, ctx->tintv);
		else
			hrtimer_forward_now(&ctx->t.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;

		if (isalarm(ctx)) {
			ctx->ticks +=
				alarm_forward_now(
					&ctx->t.alarm, ctx->tintv) - 1;
			alarm_restart(&ctx->t.alarm);
		} else {
			ctx->ticks +=
				hrtimer_forward_now(&ctx->t.tmr, ctx->tintv)
				- 1;
			hrtimer_restart(&ctx->t.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 <davidel@xmailserver.org>
	5	*
	6	*
	7	* Thanks to Thomas Gleixner for code reviews and useful comments.
	8	*
	9	*/
	10
3c2a0909	11	#include <linux/alarmtimer.h>
b215e283 DL	12	#include <linux/file.h>
	13	#include <linux/poll.h>
	14	#include <linux/init.h>
	15	#include <linux/fs.h>
	16	#include <linux/sched.h>
	17	#include <linux/kernel.h>
5a0e3ad6	18	#include <linux/slab.h>
b215e283 DL	19	#include <linux/list.h>
	20	#include <linux/spinlock.h>
	21	#include <linux/time.h>
	22	#include <linux/hrtimer.h>
	23	#include <linux/anon_inodes.h>
	24	#include <linux/timerfd.h>
45cc2b96	25	#include <linux/syscalls.h>
9d94b9e2	26	#include <linux/compat.h>
9ec26907	27	#include <linux/rcupdate.h>
b215e283 DL	28
b215e283 DL	29	struct timerfd_ctx {
3c2a0909 S	30	union {
	31	struct hrtimer tmr;
	32	struct alarm alarm;
	33	} t;
b215e283	34	ktime_t tintv;
99ee5315	35	ktime_t moffs;
b215e283	36	wait_queue_head_t wqh;
4d672e7a	37	u64 ticks;
b215e283	38	int expired;
4d672e7a	39	int clockid;
9ec26907 TG	40	struct rcu_head rcu;
9ec26907 TG	41	struct list_head clist;
ac421a87	42	spinlock_t cancel_lock;
99ee5315	43	bool might_cancel;
b215e283 DL	44	};
b215e283 DL	45
9ec26907 TG	46	static LIST_HEAD(cancel_list);
	47	static DEFINE_SPINLOCK(cancel_lock);
	48
3c2a0909 S	49	static inline bool isalarm(struct timerfd_ctx *ctx)
	50	{
	51	return ctx->clockid == CLOCK_REALTIME_ALARM \|\|
	52	ctx->clockid == CLOCK_BOOTTIME_ALARM;
	53	}
	54
b215e283 DL	55	/*
	56	* This gets called when the timer event triggers. We set the "expired"
	57	* flag, but we do not re-arm the timer (in case it's necessary,
4d672e7a	58	* tintv.tv64 != 0) until the timer is accessed.
b215e283	59	*/
3c2a0909	60	static void timerfd_triggered(struct timerfd_ctx *ctx)
b215e283	61	{
b215e283 DL	62	unsigned long flags;
b215e283 DL	63
18963c01	64	spin_lock_irqsave(&ctx->wqh.lock, flags);
b215e283	65	ctx->expired = 1;
4d672e7a	66	ctx->ticks++;
b215e283	67	wake_up_locked(&ctx->wqh);
18963c01	68	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
3c2a0909	69	}
b215e283	70
3c2a0909 S	71	static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
	72	{
	73	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx,
	74	t.tmr);
	75	timerfd_triggered(ctx);
b215e283 DL	76	return HRTIMER_NORESTART;
	77	}
	78
3c2a0909 S	79	static enum alarmtimer_restart timerfd_alarmproc(struct alarm *alarm,
	80	ktime_t now)
	81	{
	82	struct timerfd_ctx *ctx = container_of(alarm, struct timerfd_ctx,
	83	t.alarm);
	84	timerfd_triggered(ctx);
	85	return ALARMTIMER_NORESTART;
	86	}
	87
9ec26907 TG	88	/*
9ec26907 TG	89	* Called when the clock was set to cancel the timers in the cancel
1123d939 MA	90	* list. This will wake up processes waiting on these timers. The
	91	* wake-up requires ctx->ticks to be non zero, therefore we increment
	92	* it before calling wake_up_locked().
9ec26907 TG	93	*/
9ec26907 TG	94	void timerfd_clock_was_set(void)
4d672e7a	95	{
9ec26907 TG	96	ktime_t moffs = ktime_get_monotonic_offset();
	97	struct timerfd_ctx *ctx;
	98	unsigned long flags;
4d672e7a	99
9ec26907 TG	100	rcu_read_lock();
	101	list_for_each_entry_rcu(ctx, &cancel_list, clist) {
	102	if (!ctx->might_cancel)
	103	continue;
	104	spin_lock_irqsave(&ctx->wqh.lock, flags);
	105	if (ctx->moffs.tv64 != moffs.tv64) {
	106	ctx->moffs.tv64 = KTIME_MAX;
1123d939	107	ctx->ticks++;
9ec26907 TG	108	wake_up_locked(&ctx->wqh);
	109	}
	110	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
	111	}
	112	rcu_read_unlock();
4d672e7a DL	113	}
4d672e7a DL	114
ac421a87	115	static void __timerfd_remove_cancel(struct timerfd_ctx *ctx)
99ee5315	116	{
9ec26907 TG	117	if (ctx->might_cancel) {
	118	ctx->might_cancel = false;
	119	spin_lock(&cancel_lock);
	120	list_del_rcu(&ctx->clist);
	121	spin_unlock(&cancel_lock);
	122	}
	123	}
99ee5315	124
ac421a87 TG	125	static void timerfd_remove_cancel(struct timerfd_ctx *ctx)
	126	{
	127	spin_lock(&ctx->cancel_lock);
	128	__timerfd_remove_cancel(ctx);
	129	spin_unlock(&ctx->cancel_lock);
	130	}
	131
9ec26907 TG	132	static bool timerfd_canceled(struct timerfd_ctx *ctx)
	133	{
	134	if (!ctx->might_cancel \|\| ctx->moffs.tv64 != KTIME_MAX)
99ee5315	135	return false;
9ec26907 TG	136	ctx->moffs = ktime_get_monotonic_offset();
	137	return true;
	138	}
99ee5315	139
9ec26907 TG	140	static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
9ec26907 TG	141	{
ac421a87	142	spin_lock(&ctx->cancel_lock);
3c2a0909 S	143	if ((ctx->clockid == CLOCK_REALTIME \|\|
	144	ctx->clockid == CLOCK_REALTIME_ALARM) &&
	145	(flags & TFD_TIMER_ABSTIME) && (flags & TFD_TIMER_CANCEL_ON_SET)) {
9ec26907 TG	146	if (!ctx->might_cancel) {
	147	ctx->might_cancel = true;
	148	spin_lock(&cancel_lock);
	149	list_add_rcu(&ctx->clist, &cancel_list);
	150	spin_unlock(&cancel_lock);
	151	}
ac421a87 TG	152	} else {
ac421a87 TG	153	__timerfd_remove_cancel(ctx);
9ec26907	154	}
ac421a87	155	spin_unlock(&ctx->cancel_lock);
9ec26907	156	}
99ee5315	157
9ec26907 TG	158	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
	159	{
	160	ktime_t remaining;
99ee5315	161
3c2a0909 S	162	if (isalarm(ctx))
	163	remaining = alarm_expires_remaining(&ctx->t.alarm);
	164	else
	165	remaining = hrtimer_expires_remaining(&ctx->t.tmr);
	166
9ec26907	167	return remaining.tv64 < 0 ? ktime_set(0, 0): remaining;
99ee5315 TG	168	}
	169
	170	static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
	171	const struct itimerspec *ktmr)
b215e283 DL	172	{
	173	enum hrtimer_mode htmode;
	174	ktime_t texp;
99ee5315	175	int clockid = ctx->clockid;
b215e283 DL	176
	177	htmode = (flags & TFD_TIMER_ABSTIME) ?
	178	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	179
	180	texp = timespec_to_ktime(ktmr->it_value);
	181	ctx->expired = 0;
4d672e7a	182	ctx->ticks = 0;
b215e283	183	ctx->tintv = timespec_to_ktime(ktmr->it_interval);
3c2a0909 S	184
	185	if (isalarm(ctx)) {
	186	alarm_init(&ctx->t.alarm,
	187	ctx->clockid == CLOCK_REALTIME_ALARM ?
	188	ALARM_REALTIME : ALARM_BOOTTIME,
	189	timerfd_alarmproc);
	190	} else {
	191	hrtimer_init(&ctx->t.tmr, clockid, htmode);
	192	hrtimer_set_expires(&ctx->t.tmr, texp);
	193	ctx->t.tmr.function = timerfd_tmrproc;
	194	}
	195
99ee5315	196	if (texp.tv64 != 0) {
3c2a0909 S	197	if (isalarm(ctx)) {
	198	if (flags & TFD_TIMER_ABSTIME)
	199	alarm_start(&ctx->t.alarm, texp);
	200	else
	201	alarm_start_relative(&ctx->t.alarm, texp);
	202	} else {
	203	hrtimer_start(&ctx->t.tmr, texp, htmode);
	204	}
	205
99ee5315 TG	206	if (timerfd_canceled(ctx))
	207	return -ECANCELED;
	208	}
	209	return 0;
b215e283 DL	210	}
	211
	212	static int timerfd_release(struct inode inode, struct file file)
	213	{
	214	struct timerfd_ctx *ctx = file->private_data;
	215
9ec26907	216	timerfd_remove_cancel(ctx);
3c2a0909 S	217
	218	if (isalarm(ctx))
	219	alarm_cancel(&ctx->t.alarm);
	220	else
	221	hrtimer_cancel(&ctx->t.tmr);
9ec26907	222	kfree_rcu(ctx, rcu);
b215e283 DL	223	return 0;
	224	}
	225
	226	static unsigned int timerfd_poll(struct file file, poll_table wait)
	227	{
	228	struct timerfd_ctx *ctx = file->private_data;
	229	unsigned int events = 0;
	230	unsigned long flags;
	231
	232	poll_wait(file, &ctx->wqh, wait);
	233
18963c01	234	spin_lock_irqsave(&ctx->wqh.lock, flags);
4d672e7a	235	if (ctx->ticks)
b215e283	236	events \|= POLLIN;
18963c01	237	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	238
	239	return events;
	240	}
	241
	242	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	243	loff_t *ppos)
	244	{
	245	struct timerfd_ctx *ctx = file->private_data;
	246	ssize_t res;
09828402	247	u64 ticks = 0;
b215e283 DL	248
	249	if (count < sizeof(ticks))
	250	return -EINVAL;
18963c01	251	spin_lock_irq(&ctx->wqh.lock);
8120a8aa MN	252	if (file->f_flags & O_NONBLOCK)
	253	res = -EAGAIN;
	254	else
	255	res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);
99ee5315	256
9ec26907 TG	257	/*
	258	* If clock has changed, we do not care about the
	259	* ticks and we do not rearm the timer. Userspace must
	260	* reevaluate anyway.
	261	*/
	262	if (timerfd_canceled(ctx)) {
	263	ctx->ticks = 0;
	264	ctx->expired = 0;
	265	res = -ECANCELED;
	266	}
	267
4d672e7a DL	268	if (ctx->ticks) {
4d672e7a DL	269	ticks = ctx->ticks;
99ee5315	270
4d672e7a	271	if (ctx->expired && ctx->tintv.tv64) {
b215e283 DL	272	/*
	273	* If tintv.tv64 != 0, this is a periodic timer that
	274	* needs to be re-armed. We avoid doing it in the timer
	275	* callback to avoid DoS attacks specifying a very
	276	* short timer period.
	277	*/
3c2a0909 S	278	if (isalarm(ctx)) {
	279	ticks += alarm_forward_now(
	280	&ctx->t.alarm, ctx->tintv) - 1;
	281	alarm_restart(&ctx->t.alarm);
	282	} else {
	283	ticks += hrtimer_forward_now(&ctx->t.tmr,
	284	ctx->tintv) - 1;
	285	hrtimer_restart(&ctx->t.tmr);
	286	}
4d672e7a DL	287	}
	288	ctx->expired = 0;
	289	ctx->ticks = 0;
b215e283	290	}
18963c01	291	spin_unlock_irq(&ctx->wqh.lock);
b215e283	292	if (ticks)
09828402	293	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	294	return res;
	295	}
	296
	297	static const struct file_operations timerfd_fops = {
	298	.release = timerfd_release,
	299	.poll = timerfd_poll,
	300	.read = timerfd_read,
6038f373	301	.llseek = noop_llseek,
b215e283 DL	302	};
b215e283 DL	303
2903ff01	304	static int timerfd_fget(int fd, struct fd *p)
4d672e7a	305	{
2903ff01 AV	306	struct fd f = fdget(fd);
	307	if (!f.file)
	308	return -EBADF;
	309	if (f.file->f_op != &timerfd_fops) {
	310	fdput(f);
	311	return -EINVAL;
4d672e7a	312	}
2903ff01 AV	313	*p = f;
2903ff01 AV	314	return 0;
4d672e7a DL	315	}
4d672e7a DL	316
836f92ad	317	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	318	{
2030a42c	319	int ufd;
b215e283	320	struct timerfd_ctx *ctx;
b215e283	321
e38b36f3 UD	322	/* Check the TFD_* constants for consistency. */
	323	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	324	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	325
610d18f4 DL	326	if ((flags & ~TFD_CREATE_FLAGS) \|\|
610d18f4 DL	327	(clockid != CLOCK_MONOTONIC &&
3c2a0909 S	328	clockid != CLOCK_REALTIME &&
	329	clockid != CLOCK_REALTIME_ALARM &&
	330	clockid != CLOCK_BOOTTIME &&
	331	clockid != CLOCK_BOOTTIME_ALARM))
b215e283	332	return -EINVAL;
4d672e7a DL	333
	334	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	335	if (!ctx)
	336	return -ENOMEM;
	337
	338	init_waitqueue_head(&ctx->wqh);
ac421a87	339	spin_lock_init(&ctx->cancel_lock);
4d672e7a	340	ctx->clockid = clockid;
3c2a0909 S	341
	342	if (isalarm(ctx))
	343	alarm_init(&ctx->t.alarm,
	344	ctx->clockid == CLOCK_REALTIME_ALARM ?
	345	ALARM_REALTIME : ALARM_BOOTTIME,
	346	timerfd_alarmproc);
	347	else
	348	hrtimer_init(&ctx->t.tmr, clockid, HRTIMER_MODE_ABS);
	349
99ee5315	350	ctx->moffs = ktime_get_monotonic_offset();
4d672e7a	351
11fcb6c1	352	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	353	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	354	if (ufd < 0)
4d672e7a	355	kfree(ctx);
4d672e7a DL	356
	357	return ufd;
	358	}
	359
9d94b9e2 AV	360	static int do_timerfd_settime(int ufd, int flags,
	361	const struct itimerspec *new,
	362	struct itimerspec *old)
4d672e7a	363	{
2903ff01	364	struct fd f;
4d672e7a	365	struct timerfd_ctx *ctx;
2903ff01	366	int ret;
4d672e7a	367
610d18f4	368	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
9d94b9e2 AV	369	!timespec_valid(&new->it_value) \|\|
9d94b9e2 AV	370	!timespec_valid(&new->it_interval))
b215e283 DL	371	return -EINVAL;
b215e283 DL	372
2903ff01 AV	373	ret = timerfd_fget(ufd, &f);
	374	if (ret)
	375	return ret;
	376	ctx = f.file->private_data;
b215e283	377
9ec26907 TG	378	timerfd_setup_cancel(ctx, flags);
9ec26907 TG	379
4d672e7a DL	380	/*
	381	* We need to stop the existing timer before reprogramming
	382	* it to the new values.
	383	*/
	384	for (;;) {
	385	spin_lock_irq(&ctx->wqh.lock);
3c2a0909 S	386
	387	if (isalarm(ctx)) {
	388	if (alarm_try_to_cancel(&ctx->t.alarm) >= 0)
	389	break;
	390	} else {
	391	if (hrtimer_try_to_cancel(&ctx->t.tmr) >= 0)
	392	break;
	393	}
18963c01	394	spin_unlock_irq(&ctx->wqh.lock);
4d672e7a	395	cpu_relax();
b215e283 DL	396	}
b215e283 DL	397
4d672e7a DL	398	/*
	399	* If the timer is expired and it's periodic, we need to advance it
	400	* because the caller may want to know the previous expiration time.
	401	* We do not update "ticks" and "expired" since the timer will be
	402	* re-programmed again in the following timerfd_setup() call.
	403	*/
3c2a0909 S	404	if (ctx->expired && ctx->tintv.tv64) {
	405	if (isalarm(ctx))
	406	alarm_forward_now(&ctx->t.alarm, ctx->tintv);
	407	else
	408	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv);
	409	}
b215e283	410
9d94b9e2 AV	411	old->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	412	old->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a DL	413
	414	/*
	415	* Re-program the timer to the new value ...
	416	*/
9d94b9e2	417	ret = timerfd_setup(ctx, flags, new);
4d672e7a DL	418
4d672e7a DL	419	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	420	fdput(f);
99ee5315	421	return ret;
4d672e7a DL	422	}
4d672e7a DL	423
9d94b9e2	424	static int do_timerfd_gettime(int ufd, struct itimerspec *t)
4d672e7a	425	{
2903ff01	426	struct fd f;
4d672e7a	427	struct timerfd_ctx *ctx;
2903ff01 AV	428	int ret = timerfd_fget(ufd, &f);
	429	if (ret)
	430	return ret;
	431	ctx = f.file->private_data;
4d672e7a DL	432
	433	spin_lock_irq(&ctx->wqh.lock);
	434	if (ctx->expired && ctx->tintv.tv64) {
	435	ctx->expired = 0;
3c2a0909 S	436
	437	if (isalarm(ctx)) {
	438	ctx->ticks +=
	439	alarm_forward_now(
	440	&ctx->t.alarm, ctx->tintv) - 1;
	441	alarm_restart(&ctx->t.alarm);
	442	} else {
	443	ctx->ticks +=
	444	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv)
	445	- 1;
	446	hrtimer_restart(&ctx->t.tmr);
	447	}
4d672e7a	448	}
9d94b9e2 AV	449	t->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	450	t->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a	451	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	452	fdput(f);
9d94b9e2 AV	453	return 0;
	454	}
	455
	456	SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
	457	const struct itimerspec __user *, utmr,
	458	struct itimerspec __user *, otmr)
	459	{
	460	struct itimerspec new, old;
	461	int ret;
	462
	463	if (copy_from_user(&new, utmr, sizeof(new)))
	464	return -EFAULT;
	465	ret = do_timerfd_settime(ufd, flags, &new, &old);
	466	if (ret)
	467	return ret;
	468	if (otmr && copy_to_user(otmr, &old, sizeof(old)))
	469	return -EFAULT;
	470
	471	return ret;
	472	}
4d672e7a	473
9d94b9e2 AV	474	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
	475	{
	476	struct itimerspec kotmr;
	477	int ret = do_timerfd_gettime(ufd, &kotmr);
	478	if (ret)
	479	return ret;
4d672e7a	480	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
b215e283 DL	481	}
b215e283 DL	482
0e803baf	483	#ifdef CONFIG_COMPAT
9d94b9e2	484	COMPAT_SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
0e803baf HC	485	const struct compat_itimerspec __user *, utmr,
0e803baf HC	486	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	487	{
	488	struct itimerspec new, old;
	489	int ret;
	490
	491	if (get_compat_itimerspec(&new, utmr))
	492	return -EFAULT;
	493	ret = do_timerfd_settime(ufd, flags, &new, &old);
	494	if (ret)
	495	return ret;
	496	if (otmr && put_compat_itimerspec(otmr, &old))
	497	return -EFAULT;
	498	return ret;
	499	}
	500
	501	COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
0e803baf	502	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	503	{
	504	struct itimerspec kotmr;
	505	int ret = do_timerfd_gettime(ufd, &kotmr);
	506	if (ret)
	507	return ret;
0e803baf	508	return put_compat_itimerspec(otmr, &kotmr) ? -EFAULT: 0;
9d94b9e2 AV	509	}
9d94b9e2 AV	510	#endif