[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/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 clockid;
	short unsigned expired;
	short unsigned settime_flags;	/* to show in fdinfo */
	struct rcu_head rcu;
	struct list_head clist;
	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 != 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_mono_to_real(0);
	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 != moffs) {
			ctx->moffs = 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 != KTIME_MAX)
		return false;
	ctx->moffs = ktime_mono_to_real(0);
	return true;
}

static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
{
	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 if (ctx->might_cancel) {
		timerfd_remove_cancel(ctx);
	}
}

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_adjusted(&ctx->t.tmr);

	return remaining < 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 != 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;
	}

	ctx->settime_flags = flags & TFD_SETTIME_FLAGS;
	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) {
			/*
			 * If tintv != 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;
}

#ifdef CONFIG_PROC_FS
static void timerfd_show(struct seq_file *m, struct file *file)
{
	struct timerfd_ctx *ctx = file->private_data;
	struct itimerspec t;

	spin_lock_irq(&ctx->wqh.lock);
	t.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	t.it_interval = ktime_to_timespec(ctx->tintv);
	spin_unlock_irq(&ctx->wqh.lock);

	seq_printf(m,
		   "clockid: %d\n"
		   "ticks: %llu\n"
		   "settime flags: 0%o\n"
		   "it_value: (%llu, %llu)\n"
		   "it_interval: (%llu, %llu)\n",
		   ctx->clockid,
		   (unsigned long long)ctx->ticks,
		   ctx->settime_flags,
		   (unsigned long long)t.it_value.tv_sec,
		   (unsigned long long)t.it_value.tv_nsec,
		   (unsigned long long)t.it_interval.tv_sec,
		   (unsigned long long)t.it_interval.tv_nsec);
}
#else
#define timerfd_show NULL
#endif

#ifdef CONFIG_CHECKPOINT_RESTORE
static long timerfd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
	struct timerfd_ctx *ctx = file->private_data;
	int ret = 0;

	switch (cmd) {
	case TFD_IOC_SET_TICKS: {
		u64 ticks;

		if (copy_from_user(&ticks, (u64 __user *)arg, sizeof(ticks)))
			return -EFAULT;
		if (!ticks)
			return -EINVAL;

		spin_lock_irq(&ctx->wqh.lock);
		if (!timerfd_canceled(ctx)) {
			ctx->ticks = ticks;
			wake_up_locked(&ctx->wqh);
		} else
			ret = -ECANCELED;
		spin_unlock_irq(&ctx->wqh.lock);
		break;
	}
	default:
		ret = -ENOTTY;
		break;
	}

	return ret;
}
#else
#define timerfd_ioctl NULL
#endif

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

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;

	if (!capable(CAP_WAKE_ALARM) &&
	    (clockid == CLOCK_REALTIME_ALARM ||
	     clockid == CLOCK_BOOTTIME_ALARM))
		return -EPERM;

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

	init_waitqueue_head(&ctx->wqh);
	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_mono_to_real(0);

	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;

	if (!capable(CAP_WAKE_ALARM) && isalarm(ctx)) {
		fdput(f);
		return -EPERM;
	}

	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) {
		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) {
		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 <[email protected]>
	5	*
	6	*
	7	* Thanks to Thomas Gleixner for code reviews and useful comments.
	8	*
	9	*/
	10
11ffa9d6	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 {
11ffa9d6 TP	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;
4d672e7a	38	int clockid;
af9c4957 CG	39	short unsigned expired;
af9c4957 CG	40	short unsigned settime_flags; /* to show in fdinfo */
9ec26907 TG	41	struct rcu_head rcu;
9ec26907 TG	42	struct list_head clist;
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
11ffa9d6 TP	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,
2456e855	58	* tintv != 0) until the timer is accessed.
b215e283	59	*/
11ffa9d6	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);
11ffa9d6	69	}
b215e283	70
11ffa9d6 TP	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
11ffa9d6 TP	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	{
2456e855	96	ktime_t moffs = ktime_mono_to_real(0);
9ec26907 TG	97	struct timerfd_ctx *ctx;
9ec26907 TG	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);
2456e855 TG	105	if (ctx->moffs != moffs) {
2456e855 TG	106	ctx->moffs = 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
9ec26907	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
9ec26907 TG	125	static bool timerfd_canceled(struct timerfd_ctx *ctx)
9ec26907 TG	126	{
2456e855	127	if (!ctx->might_cancel \|\| ctx->moffs != KTIME_MAX)
99ee5315	128	return false;
2456e855	129	ctx->moffs = ktime_mono_to_real(0);
9ec26907 TG	130	return true;
9ec26907 TG	131	}
99ee5315	132
9ec26907 TG	133	static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
9ec26907 TG	134	{
11ffa9d6 TP	135	if ((ctx->clockid == CLOCK_REALTIME \|\|
	136	ctx->clockid == CLOCK_REALTIME_ALARM) &&
	137	(flags & TFD_TIMER_ABSTIME) && (flags & TFD_TIMER_CANCEL_ON_SET)) {
9ec26907 TG	138	if (!ctx->might_cancel) {
	139	ctx->might_cancel = true;
	140	spin_lock(&cancel_lock);
	141	list_add_rcu(&ctx->clist, &cancel_list);
	142	spin_unlock(&cancel_lock);
	143	}
	144	} else if (ctx->might_cancel) {
	145	timerfd_remove_cancel(ctx);
	146	}
	147	}
99ee5315	148
9ec26907 TG	149	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
	150	{
	151	ktime_t remaining;
99ee5315	152
11ffa9d6 TP	153	if (isalarm(ctx))
	154	remaining = alarm_expires_remaining(&ctx->t.alarm);
	155	else
b62526ed	156	remaining = hrtimer_expires_remaining_adjusted(&ctx->t.tmr);
11ffa9d6	157
8b0e1953	158	return remaining < 0 ? 0: remaining;
99ee5315 TG	159	}
	160
	161	static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
	162	const struct itimerspec *ktmr)
b215e283 DL	163	{
	164	enum hrtimer_mode htmode;
	165	ktime_t texp;
99ee5315	166	int clockid = ctx->clockid;
b215e283 DL	167
	168	htmode = (flags & TFD_TIMER_ABSTIME) ?
	169	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	170
	171	texp = timespec_to_ktime(ktmr->it_value);
	172	ctx->expired = 0;
4d672e7a	173	ctx->ticks = 0;
b215e283	174	ctx->tintv = timespec_to_ktime(ktmr->it_interval);
11ffa9d6 TP	175
	176	if (isalarm(ctx)) {
	177	alarm_init(&ctx->t.alarm,
	178	ctx->clockid == CLOCK_REALTIME_ALARM ?
	179	ALARM_REALTIME : ALARM_BOOTTIME,
	180	timerfd_alarmproc);
	181	} else {
	182	hrtimer_init(&ctx->t.tmr, clockid, htmode);
	183	hrtimer_set_expires(&ctx->t.tmr, texp);
	184	ctx->t.tmr.function = timerfd_tmrproc;
	185	}
	186
2456e855	187	if (texp != 0) {
11ffa9d6 TP	188	if (isalarm(ctx)) {
	189	if (flags & TFD_TIMER_ABSTIME)
	190	alarm_start(&ctx->t.alarm, texp);
	191	else
	192	alarm_start_relative(&ctx->t.alarm, texp);
	193	} else {
	194	hrtimer_start(&ctx->t.tmr, texp, htmode);
	195	}
	196
99ee5315 TG	197	if (timerfd_canceled(ctx))
	198	return -ECANCELED;
	199	}
af9c4957 CG	200
af9c4957 CG	201	ctx->settime_flags = flags & TFD_SETTIME_FLAGS;
99ee5315	202	return 0;
b215e283 DL	203	}
	204
	205	static int timerfd_release(struct inode inode, struct file file)
	206	{
	207	struct timerfd_ctx *ctx = file->private_data;
	208
9ec26907	209	timerfd_remove_cancel(ctx);
11ffa9d6 TP	210
	211	if (isalarm(ctx))
	212	alarm_cancel(&ctx->t.alarm);
	213	else
	214	hrtimer_cancel(&ctx->t.tmr);
9ec26907	215	kfree_rcu(ctx, rcu);
b215e283 DL	216	return 0;
	217	}
	218
	219	static unsigned int timerfd_poll(struct file file, poll_table wait)
	220	{
	221	struct timerfd_ctx *ctx = file->private_data;
	222	unsigned int events = 0;
	223	unsigned long flags;
	224
	225	poll_wait(file, &ctx->wqh, wait);
	226
18963c01	227	spin_lock_irqsave(&ctx->wqh.lock, flags);
4d672e7a	228	if (ctx->ticks)
b215e283	229	events \|= POLLIN;
18963c01	230	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	231
	232	return events;
	233	}
	234
	235	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	236	loff_t *ppos)
	237	{
	238	struct timerfd_ctx *ctx = file->private_data;
	239	ssize_t res;
09828402	240	u64 ticks = 0;
b215e283 DL	241
	242	if (count < sizeof(ticks))
	243	return -EINVAL;
18963c01	244	spin_lock_irq(&ctx->wqh.lock);
8120a8aa MN	245	if (file->f_flags & O_NONBLOCK)
	246	res = -EAGAIN;
	247	else
	248	res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);
99ee5315	249
9ec26907 TG	250	/*
	251	* If clock has changed, we do not care about the
	252	* ticks and we do not rearm the timer. Userspace must
	253	* reevaluate anyway.
	254	*/
	255	if (timerfd_canceled(ctx)) {
	256	ctx->ticks = 0;
	257	ctx->expired = 0;
	258	res = -ECANCELED;
	259	}
	260
4d672e7a DL	261	if (ctx->ticks) {
4d672e7a DL	262	ticks = ctx->ticks;
99ee5315	263
2456e855	264	if (ctx->expired && ctx->tintv) {
b215e283	265	/*
2456e855	266	* If tintv != 0, this is a periodic timer that
b215e283 DL	267	* needs to be re-armed. We avoid doing it in the timer
	268	* callback to avoid DoS attacks specifying a very
	269	* short timer period.
	270	*/
11ffa9d6 TP	271	if (isalarm(ctx)) {
	272	ticks += alarm_forward_now(
	273	&ctx->t.alarm, ctx->tintv) - 1;
	274	alarm_restart(&ctx->t.alarm);
	275	} else {
	276	ticks += hrtimer_forward_now(&ctx->t.tmr,
	277	ctx->tintv) - 1;
	278	hrtimer_restart(&ctx->t.tmr);
	279	}
4d672e7a DL	280	}
	281	ctx->expired = 0;
	282	ctx->ticks = 0;
b215e283	283	}
18963c01	284	spin_unlock_irq(&ctx->wqh.lock);
b215e283	285	if (ticks)
09828402	286	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	287	return res;
	288	}
	289
af9c4957	290	#ifdef CONFIG_PROC_FS
a3816ab0	291	static void timerfd_show(struct seq_file m, struct file file)
af9c4957 CG	292	{
	293	struct timerfd_ctx *ctx = file->private_data;
	294	struct itimerspec t;
	295
	296	spin_lock_irq(&ctx->wqh.lock);
	297	t.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	298	t.it_interval = ktime_to_timespec(ctx->tintv);
	299	spin_unlock_irq(&ctx->wqh.lock);
	300
a3816ab0 JP	301	seq_printf(m,
	302	"clockid: %d\n"
	303	"ticks: %llu\n"
	304	"settime flags: 0%o\n"
	305	"it_value: (%llu, %llu)\n"
	306	"it_interval: (%llu, %llu)\n",
	307	ctx->clockid,
	308	(unsigned long long)ctx->ticks,
	309	ctx->settime_flags,
	310	(unsigned long long)t.it_value.tv_sec,
	311	(unsigned long long)t.it_value.tv_nsec,
	312	(unsigned long long)t.it_interval.tv_sec,
	313	(unsigned long long)t.it_interval.tv_nsec);
af9c4957 CG	314	}
	315	#else
	316	#define timerfd_show NULL
	317	#endif
	318
5442e9fb CG	319	#ifdef CONFIG_CHECKPOINT_RESTORE
	320	static long timerfd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
	321	{
	322	struct timerfd_ctx *ctx = file->private_data;
	323	int ret = 0;
	324
	325	switch (cmd) {
	326	case TFD_IOC_SET_TICKS: {
	327	u64 ticks;
	328
	329	if (copy_from_user(&ticks, (u64 __user *)arg, sizeof(ticks)))
	330	return -EFAULT;
	331	if (!ticks)
	332	return -EINVAL;
	333
	334	spin_lock_irq(&ctx->wqh.lock);
	335	if (!timerfd_canceled(ctx)) {
	336	ctx->ticks = ticks;
88299c9b	337	wake_up_locked(&ctx->wqh);
5442e9fb CG	338	} else
	339	ret = -ECANCELED;
	340	spin_unlock_irq(&ctx->wqh.lock);
	341	break;
	342	}
	343	default:
	344	ret = -ENOTTY;
	345	break;
	346	}
	347
	348	return ret;
	349	}
	350	#else
	351	#define timerfd_ioctl NULL
	352	#endif
	353
b215e283 DL	354	static const struct file_operations timerfd_fops = {
	355	.release = timerfd_release,
	356	.poll = timerfd_poll,
	357	.read = timerfd_read,
6038f373	358	.llseek = noop_llseek,
af9c4957	359	.show_fdinfo = timerfd_show,
5442e9fb	360	.unlocked_ioctl = timerfd_ioctl,
b215e283 DL	361	};
b215e283 DL	362
2903ff01	363	static int timerfd_fget(int fd, struct fd *p)
4d672e7a	364	{
2903ff01 AV	365	struct fd f = fdget(fd);
	366	if (!f.file)
	367	return -EBADF;
	368	if (f.file->f_op != &timerfd_fops) {
	369	fdput(f);
	370	return -EINVAL;
4d672e7a	371	}
2903ff01 AV	372	*p = f;
2903ff01 AV	373	return 0;
4d672e7a DL	374	}
4d672e7a DL	375
836f92ad	376	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	377	{
2030a42c	378	int ufd;
b215e283	379	struct timerfd_ctx *ctx;
b215e283	380
e38b36f3 UD	381	/* Check the TFD_* constants for consistency. */
	382	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	383	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	384
610d18f4 DL	385	if ((flags & ~TFD_CREATE_FLAGS) \|\|
610d18f4 DL	386	(clockid != CLOCK_MONOTONIC &&
11ffa9d6 TP	387	clockid != CLOCK_REALTIME &&
11ffa9d6 TP	388	clockid != CLOCK_REALTIME_ALARM &&
4a2378a9	389	clockid != CLOCK_BOOTTIME &&
11ffa9d6	390	clockid != CLOCK_BOOTTIME_ALARM))
b215e283	391	return -EINVAL;
4d672e7a	392
2895a5e5 EC	393	if (!capable(CAP_WAKE_ALARM) &&
	394	(clockid == CLOCK_REALTIME_ALARM \|\|
	395	clockid == CLOCK_BOOTTIME_ALARM))
	396	return -EPERM;
	397
4d672e7a DL	398	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	399	if (!ctx)
	400	return -ENOMEM;
	401
	402	init_waitqueue_head(&ctx->wqh);
	403	ctx->clockid = clockid;
11ffa9d6 TP	404
	405	if (isalarm(ctx))
	406	alarm_init(&ctx->t.alarm,
	407	ctx->clockid == CLOCK_REALTIME_ALARM ?
	408	ALARM_REALTIME : ALARM_BOOTTIME,
	409	timerfd_alarmproc);
	410	else
	411	hrtimer_init(&ctx->t.tmr, clockid, HRTIMER_MODE_ABS);
	412
2456e855	413	ctx->moffs = ktime_mono_to_real(0);
4d672e7a	414
11fcb6c1	415	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	416	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	417	if (ufd < 0)
4d672e7a	418	kfree(ctx);
4d672e7a DL	419
	420	return ufd;
	421	}
	422
9d94b9e2 AV	423	static int do_timerfd_settime(int ufd, int flags,
	424	const struct itimerspec *new,
	425	struct itimerspec *old)
4d672e7a	426	{
2903ff01	427	struct fd f;
4d672e7a	428	struct timerfd_ctx *ctx;
2903ff01	429	int ret;
4d672e7a	430
610d18f4	431	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
9d94b9e2 AV	432	!timespec_valid(&new->it_value) \|\|
9d94b9e2 AV	433	!timespec_valid(&new->it_interval))
b215e283 DL	434	return -EINVAL;
b215e283 DL	435
2903ff01 AV	436	ret = timerfd_fget(ufd, &f);
	437	if (ret)
	438	return ret;
	439	ctx = f.file->private_data;
b215e283	440
2895a5e5 EC	441	if (!capable(CAP_WAKE_ALARM) && isalarm(ctx)) {
	442	fdput(f);
	443	return -EPERM;
	444	}
	445
9ec26907 TG	446	timerfd_setup_cancel(ctx, flags);
9ec26907 TG	447
4d672e7a DL	448	/*
	449	* We need to stop the existing timer before reprogramming
	450	* it to the new values.
	451	*/
	452	for (;;) {
	453	spin_lock_irq(&ctx->wqh.lock);
11ffa9d6 TP	454
	455	if (isalarm(ctx)) {
	456	if (alarm_try_to_cancel(&ctx->t.alarm) >= 0)
	457	break;
	458	} else {
	459	if (hrtimer_try_to_cancel(&ctx->t.tmr) >= 0)
	460	break;
	461	}
18963c01	462	spin_unlock_irq(&ctx->wqh.lock);
4d672e7a	463	cpu_relax();
b215e283 DL	464	}
b215e283 DL	465
4d672e7a DL	466	/*
	467	* If the timer is expired and it's periodic, we need to advance it
	468	* because the caller may want to know the previous expiration time.
	469	* We do not update "ticks" and "expired" since the timer will be
	470	* re-programmed again in the following timerfd_setup() call.
	471	*/
2456e855	472	if (ctx->expired && ctx->tintv) {
11ffa9d6 TP	473	if (isalarm(ctx))
	474	alarm_forward_now(&ctx->t.alarm, ctx->tintv);
	475	else
	476	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv);
	477	}
b215e283	478
9d94b9e2 AV	479	old->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	480	old->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a DL	481
	482	/*
	483	* Re-program the timer to the new value ...
	484	*/
9d94b9e2	485	ret = timerfd_setup(ctx, flags, new);
4d672e7a DL	486
4d672e7a DL	487	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	488	fdput(f);
99ee5315	489	return ret;
4d672e7a DL	490	}
4d672e7a DL	491
9d94b9e2	492	static int do_timerfd_gettime(int ufd, struct itimerspec *t)
4d672e7a	493	{
2903ff01	494	struct fd f;
4d672e7a	495	struct timerfd_ctx *ctx;
2903ff01 AV	496	int ret = timerfd_fget(ufd, &f);
	497	if (ret)
	498	return ret;
	499	ctx = f.file->private_data;
4d672e7a DL	500
4d672e7a DL	501	spin_lock_irq(&ctx->wqh.lock);
2456e855	502	if (ctx->expired && ctx->tintv) {
4d672e7a	503	ctx->expired = 0;
11ffa9d6 TP	504
	505	if (isalarm(ctx)) {
	506	ctx->ticks +=
	507	alarm_forward_now(
	508	&ctx->t.alarm, ctx->tintv) - 1;
	509	alarm_restart(&ctx->t.alarm);
	510	} else {
	511	ctx->ticks +=
	512	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv)
	513	- 1;
	514	hrtimer_restart(&ctx->t.tmr);
	515	}
4d672e7a	516	}
9d94b9e2 AV	517	t->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	518	t->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a	519	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	520	fdput(f);
9d94b9e2 AV	521	return 0;
	522	}
	523
	524	SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
	525	const struct itimerspec __user *, utmr,
	526	struct itimerspec __user *, otmr)
	527	{
	528	struct itimerspec new, old;
	529	int ret;
	530
	531	if (copy_from_user(&new, utmr, sizeof(new)))
	532	return -EFAULT;
	533	ret = do_timerfd_settime(ufd, flags, &new, &old);
	534	if (ret)
	535	return ret;
	536	if (otmr && copy_to_user(otmr, &old, sizeof(old)))
	537	return -EFAULT;
	538
	539	return ret;
	540	}
4d672e7a	541
9d94b9e2 AV	542	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
	543	{
	544	struct itimerspec kotmr;
	545	int ret = do_timerfd_gettime(ufd, &kotmr);
	546	if (ret)
	547	return ret;
4d672e7a	548	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
b215e283 DL	549	}
b215e283 DL	550
0e803baf	551	#ifdef CONFIG_COMPAT
9d94b9e2	552	COMPAT_SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
0e803baf HC	553	const struct compat_itimerspec __user *, utmr,
0e803baf HC	554	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	555	{
	556	struct itimerspec new, old;
	557	int ret;
	558
	559	if (get_compat_itimerspec(&new, utmr))
	560	return -EFAULT;
	561	ret = do_timerfd_settime(ufd, flags, &new, &old);
	562	if (ret)
	563	return ret;
	564	if (otmr && put_compat_itimerspec(otmr, &old))
	565	return -EFAULT;
	566	return ret;
	567	}
	568
	569	COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
0e803baf	570	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	571	{
	572	struct itimerspec kotmr;
	573	int ret = do_timerfd_gettime(ufd, &kotmr);
	574	if (ret)
	575	return ret;
0e803baf	576	return put_compat_itimerspec(otmr, &kotmr) ? -EFAULT: 0;
9d94b9e2 AV	577	}
9d94b9e2 AV	578	#endif