[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;
	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 != 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 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 != KTIME_MAX)
		return false;
	ctx->moffs = ktime_mono_to_real(0);
	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_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 ((clockid == CLOCK_REALTIME_ALARM ||
	     clockid == CLOCK_BOOTTIME_ALARM) &&
	    !capable(CAP_WAKE_ALARM))
		return -EPERM;

	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_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 (isalarm(ctx) && !capable(CAP_WAKE_ALARM)) {
		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;
1e38da30	43	spinlock_t cancel_lock;
99ee5315	44	bool might_cancel;
b215e283 DL	45	};
b215e283 DL	46
9ec26907 TG	47	static LIST_HEAD(cancel_list);
	48	static DEFINE_SPINLOCK(cancel_lock);
	49
11ffa9d6 TP	50	static inline bool isalarm(struct timerfd_ctx *ctx)
	51	{
	52	return ctx->clockid == CLOCK_REALTIME_ALARM \|\|
	53	ctx->clockid == CLOCK_BOOTTIME_ALARM;
	54	}
	55
b215e283 DL	56	/*
	57	* This gets called when the timer event triggers. We set the "expired"
	58	* flag, but we do not re-arm the timer (in case it's necessary,
2456e855	59	* tintv != 0) until the timer is accessed.
b215e283	60	*/
11ffa9d6	61	static void timerfd_triggered(struct timerfd_ctx *ctx)
b215e283	62	{
b215e283 DL	63	unsigned long flags;
b215e283 DL	64
18963c01	65	spin_lock_irqsave(&ctx->wqh.lock, flags);
b215e283	66	ctx->expired = 1;
4d672e7a	67	ctx->ticks++;
b215e283	68	wake_up_locked(&ctx->wqh);
18963c01	69	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
11ffa9d6	70	}
b215e283	71
11ffa9d6 TP	72	static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
	73	{
	74	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx,
	75	t.tmr);
	76	timerfd_triggered(ctx);
b215e283 DL	77	return HRTIMER_NORESTART;
	78	}
	79
11ffa9d6 TP	80	static enum alarmtimer_restart timerfd_alarmproc(struct alarm *alarm,
	81	ktime_t now)
	82	{
	83	struct timerfd_ctx *ctx = container_of(alarm, struct timerfd_ctx,
	84	t.alarm);
	85	timerfd_triggered(ctx);
	86	return ALARMTIMER_NORESTART;
	87	}
	88
9ec26907 TG	89	/*
9ec26907 TG	90	* Called when the clock was set to cancel the timers in the cancel
1123d939 MA	91	* list. This will wake up processes waiting on these timers. The
	92	* wake-up requires ctx->ticks to be non zero, therefore we increment
	93	* it before calling wake_up_locked().
9ec26907 TG	94	*/
9ec26907 TG	95	void timerfd_clock_was_set(void)
4d672e7a	96	{
2456e855	97	ktime_t moffs = ktime_mono_to_real(0);
9ec26907 TG	98	struct timerfd_ctx *ctx;
9ec26907 TG	99	unsigned long flags;
4d672e7a	100
9ec26907 TG	101	rcu_read_lock();
	102	list_for_each_entry_rcu(ctx, &cancel_list, clist) {
	103	if (!ctx->might_cancel)
	104	continue;
	105	spin_lock_irqsave(&ctx->wqh.lock, flags);
2456e855 TG	106	if (ctx->moffs != moffs) {
2456e855 TG	107	ctx->moffs = KTIME_MAX;
1123d939	108	ctx->ticks++;
9ec26907 TG	109	wake_up_locked(&ctx->wqh);
	110	}
	111	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
	112	}
	113	rcu_read_unlock();
4d672e7a DL	114	}
4d672e7a DL	115
1e38da30	116	static void __timerfd_remove_cancel(struct timerfd_ctx *ctx)
99ee5315	117	{
9ec26907 TG	118	if (ctx->might_cancel) {
	119	ctx->might_cancel = false;
	120	spin_lock(&cancel_lock);
	121	list_del_rcu(&ctx->clist);
	122	spin_unlock(&cancel_lock);
	123	}
	124	}
99ee5315	125
1e38da30 TG	126	static void timerfd_remove_cancel(struct timerfd_ctx *ctx)
	127	{
	128	spin_lock(&ctx->cancel_lock);
	129	__timerfd_remove_cancel(ctx);
	130	spin_unlock(&ctx->cancel_lock);
	131	}
	132
9ec26907 TG	133	static bool timerfd_canceled(struct timerfd_ctx *ctx)
9ec26907 TG	134	{
2456e855	135	if (!ctx->might_cancel \|\| ctx->moffs != KTIME_MAX)
99ee5315	136	return false;
2456e855	137	ctx->moffs = ktime_mono_to_real(0);
9ec26907 TG	138	return true;
9ec26907 TG	139	}
99ee5315	140
9ec26907 TG	141	static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
9ec26907 TG	142	{
1e38da30	143	spin_lock(&ctx->cancel_lock);
11ffa9d6 TP	144	if ((ctx->clockid == CLOCK_REALTIME \|\|
	145	ctx->clockid == CLOCK_REALTIME_ALARM) &&
	146	(flags & TFD_TIMER_ABSTIME) && (flags & TFD_TIMER_CANCEL_ON_SET)) {
9ec26907 TG	147	if (!ctx->might_cancel) {
	148	ctx->might_cancel = true;
	149	spin_lock(&cancel_lock);
	150	list_add_rcu(&ctx->clist, &cancel_list);
	151	spin_unlock(&cancel_lock);
	152	}
1e38da30 TG	153	} else {
1e38da30 TG	154	__timerfd_remove_cancel(ctx);
9ec26907	155	}
1e38da30	156	spin_unlock(&ctx->cancel_lock);
9ec26907	157	}
99ee5315	158
9ec26907 TG	159	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
	160	{
	161	ktime_t remaining;
99ee5315	162
11ffa9d6 TP	163	if (isalarm(ctx))
	164	remaining = alarm_expires_remaining(&ctx->t.alarm);
	165	else
b62526ed	166	remaining = hrtimer_expires_remaining_adjusted(&ctx->t.tmr);
11ffa9d6	167
8b0e1953	168	return remaining < 0 ? 0: remaining;
99ee5315 TG	169	}
	170
	171	static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
	172	const struct itimerspec *ktmr)
b215e283 DL	173	{
	174	enum hrtimer_mode htmode;
	175	ktime_t texp;
99ee5315	176	int clockid = ctx->clockid;
b215e283 DL	177
	178	htmode = (flags & TFD_TIMER_ABSTIME) ?
	179	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	180
	181	texp = timespec_to_ktime(ktmr->it_value);
	182	ctx->expired = 0;
4d672e7a	183	ctx->ticks = 0;
b215e283	184	ctx->tintv = timespec_to_ktime(ktmr->it_interval);
11ffa9d6 TP	185
	186	if (isalarm(ctx)) {
	187	alarm_init(&ctx->t.alarm,
	188	ctx->clockid == CLOCK_REALTIME_ALARM ?
	189	ALARM_REALTIME : ALARM_BOOTTIME,
	190	timerfd_alarmproc);
	191	} else {
	192	hrtimer_init(&ctx->t.tmr, clockid, htmode);
	193	hrtimer_set_expires(&ctx->t.tmr, texp);
	194	ctx->t.tmr.function = timerfd_tmrproc;
	195	}
	196
2456e855	197	if (texp != 0) {
11ffa9d6 TP	198	if (isalarm(ctx)) {
	199	if (flags & TFD_TIMER_ABSTIME)
	200	alarm_start(&ctx->t.alarm, texp);
	201	else
	202	alarm_start_relative(&ctx->t.alarm, texp);
	203	} else {
	204	hrtimer_start(&ctx->t.tmr, texp, htmode);
	205	}
	206
99ee5315 TG	207	if (timerfd_canceled(ctx))
	208	return -ECANCELED;
	209	}
af9c4957 CG	210
af9c4957 CG	211	ctx->settime_flags = flags & TFD_SETTIME_FLAGS;
99ee5315	212	return 0;
b215e283 DL	213	}
	214
	215	static int timerfd_release(struct inode inode, struct file file)
	216	{
	217	struct timerfd_ctx *ctx = file->private_data;
	218
9ec26907	219	timerfd_remove_cancel(ctx);
11ffa9d6 TP	220
	221	if (isalarm(ctx))
	222	alarm_cancel(&ctx->t.alarm);
	223	else
	224	hrtimer_cancel(&ctx->t.tmr);
9ec26907	225	kfree_rcu(ctx, rcu);
b215e283 DL	226	return 0;
	227	}
	228
	229	static unsigned int timerfd_poll(struct file file, poll_table wait)
	230	{
	231	struct timerfd_ctx *ctx = file->private_data;
	232	unsigned int events = 0;
	233	unsigned long flags;
	234
	235	poll_wait(file, &ctx->wqh, wait);
	236
18963c01	237	spin_lock_irqsave(&ctx->wqh.lock, flags);
4d672e7a	238	if (ctx->ticks)
b215e283	239	events \|= POLLIN;
18963c01	240	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	241
	242	return events;
	243	}
	244
	245	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	246	loff_t *ppos)
	247	{
	248	struct timerfd_ctx *ctx = file->private_data;
	249	ssize_t res;
09828402	250	u64 ticks = 0;
b215e283 DL	251
	252	if (count < sizeof(ticks))
	253	return -EINVAL;
18963c01	254	spin_lock_irq(&ctx->wqh.lock);
8120a8aa MN	255	if (file->f_flags & O_NONBLOCK)
	256	res = -EAGAIN;
	257	else
	258	res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);
99ee5315	259
9ec26907 TG	260	/*
	261	* If clock has changed, we do not care about the
	262	* ticks and we do not rearm the timer. Userspace must
	263	* reevaluate anyway.
	264	*/
	265	if (timerfd_canceled(ctx)) {
	266	ctx->ticks = 0;
	267	ctx->expired = 0;
	268	res = -ECANCELED;
	269	}
	270
4d672e7a DL	271	if (ctx->ticks) {
4d672e7a DL	272	ticks = ctx->ticks;
99ee5315	273
2456e855	274	if (ctx->expired && ctx->tintv) {
b215e283	275	/*
2456e855	276	* If tintv != 0, this is a periodic timer that
b215e283 DL	277	* needs to be re-armed. We avoid doing it in the timer
	278	* callback to avoid DoS attacks specifying a very
	279	* short timer period.
	280	*/
11ffa9d6 TP	281	if (isalarm(ctx)) {
	282	ticks += alarm_forward_now(
	283	&ctx->t.alarm, ctx->tintv) - 1;
	284	alarm_restart(&ctx->t.alarm);
	285	} else {
	286	ticks += hrtimer_forward_now(&ctx->t.tmr,
	287	ctx->tintv) - 1;
	288	hrtimer_restart(&ctx->t.tmr);
	289	}
4d672e7a DL	290	}
	291	ctx->expired = 0;
	292	ctx->ticks = 0;
b215e283	293	}
18963c01	294	spin_unlock_irq(&ctx->wqh.lock);
b215e283	295	if (ticks)
09828402	296	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	297	return res;
	298	}
	299
af9c4957	300	#ifdef CONFIG_PROC_FS
a3816ab0	301	static void timerfd_show(struct seq_file m, struct file file)
af9c4957 CG	302	{
	303	struct timerfd_ctx *ctx = file->private_data;
	304	struct itimerspec t;
	305
	306	spin_lock_irq(&ctx->wqh.lock);
	307	t.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	308	t.it_interval = ktime_to_timespec(ctx->tintv);
	309	spin_unlock_irq(&ctx->wqh.lock);
	310
a3816ab0 JP	311	seq_printf(m,
	312	"clockid: %d\n"
	313	"ticks: %llu\n"
	314	"settime flags: 0%o\n"
	315	"it_value: (%llu, %llu)\n"
	316	"it_interval: (%llu, %llu)\n",
	317	ctx->clockid,
	318	(unsigned long long)ctx->ticks,
	319	ctx->settime_flags,
	320	(unsigned long long)t.it_value.tv_sec,
	321	(unsigned long long)t.it_value.tv_nsec,
	322	(unsigned long long)t.it_interval.tv_sec,
	323	(unsigned long long)t.it_interval.tv_nsec);
af9c4957 CG	324	}
	325	#else
	326	#define timerfd_show NULL
	327	#endif
	328
5442e9fb CG	329	#ifdef CONFIG_CHECKPOINT_RESTORE
	330	static long timerfd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
	331	{
	332	struct timerfd_ctx *ctx = file->private_data;
	333	int ret = 0;
	334
	335	switch (cmd) {
	336	case TFD_IOC_SET_TICKS: {
	337	u64 ticks;
	338
	339	if (copy_from_user(&ticks, (u64 __user *)arg, sizeof(ticks)))
	340	return -EFAULT;
	341	if (!ticks)
	342	return -EINVAL;
	343
	344	spin_lock_irq(&ctx->wqh.lock);
	345	if (!timerfd_canceled(ctx)) {
	346	ctx->ticks = ticks;
88299c9b	347	wake_up_locked(&ctx->wqh);
5442e9fb CG	348	} else
	349	ret = -ECANCELED;
	350	spin_unlock_irq(&ctx->wqh.lock);
	351	break;
	352	}
	353	default:
	354	ret = -ENOTTY;
	355	break;
	356	}
	357
	358	return ret;
	359	}
	360	#else
	361	#define timerfd_ioctl NULL
	362	#endif
	363
b215e283 DL	364	static const struct file_operations timerfd_fops = {
	365	.release = timerfd_release,
	366	.poll = timerfd_poll,
	367	.read = timerfd_read,
6038f373	368	.llseek = noop_llseek,
af9c4957	369	.show_fdinfo = timerfd_show,
5442e9fb	370	.unlocked_ioctl = timerfd_ioctl,
b215e283 DL	371	};
b215e283 DL	372
2903ff01	373	static int timerfd_fget(int fd, struct fd *p)
4d672e7a	374	{
2903ff01 AV	375	struct fd f = fdget(fd);
	376	if (!f.file)
	377	return -EBADF;
	378	if (f.file->f_op != &timerfd_fops) {
	379	fdput(f);
	380	return -EINVAL;
4d672e7a	381	}
2903ff01 AV	382	*p = f;
2903ff01 AV	383	return 0;
4d672e7a DL	384	}
4d672e7a DL	385
836f92ad	386	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	387	{
2030a42c	388	int ufd;
b215e283	389	struct timerfd_ctx *ctx;
b215e283	390
e38b36f3 UD	391	/* Check the TFD_* constants for consistency. */
	392	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	393	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	394
610d18f4 DL	395	if ((flags & ~TFD_CREATE_FLAGS) \|\|
610d18f4 DL	396	(clockid != CLOCK_MONOTONIC &&
11ffa9d6 TP	397	clockid != CLOCK_REALTIME &&
11ffa9d6 TP	398	clockid != CLOCK_REALTIME_ALARM &&
4a2378a9	399	clockid != CLOCK_BOOTTIME &&
11ffa9d6	400	clockid != CLOCK_BOOTTIME_ALARM))
b215e283	401	return -EINVAL;
4d672e7a	402
25b68a8f SS	403	if ((clockid == CLOCK_REALTIME_ALARM \|\|
	404	clockid == CLOCK_BOOTTIME_ALARM) &&
	405	!capable(CAP_WAKE_ALARM))
2895a5e5 EC	406	return -EPERM;
2895a5e5 EC	407
4d672e7a DL	408	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	409	if (!ctx)
	410	return -ENOMEM;
	411
	412	init_waitqueue_head(&ctx->wqh);
1e38da30	413	spin_lock_init(&ctx->cancel_lock);
4d672e7a	414	ctx->clockid = clockid;
11ffa9d6 TP	415
	416	if (isalarm(ctx))
	417	alarm_init(&ctx->t.alarm,
	418	ctx->clockid == CLOCK_REALTIME_ALARM ?
	419	ALARM_REALTIME : ALARM_BOOTTIME,
	420	timerfd_alarmproc);
	421	else
	422	hrtimer_init(&ctx->t.tmr, clockid, HRTIMER_MODE_ABS);
	423
2456e855	424	ctx->moffs = ktime_mono_to_real(0);
4d672e7a	425
11fcb6c1	426	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	427	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	428	if (ufd < 0)
4d672e7a	429	kfree(ctx);
4d672e7a DL	430
	431	return ufd;
	432	}
	433
9d94b9e2 AV	434	static int do_timerfd_settime(int ufd, int flags,
	435	const struct itimerspec *new,
	436	struct itimerspec *old)
4d672e7a	437	{
2903ff01	438	struct fd f;
4d672e7a	439	struct timerfd_ctx *ctx;
2903ff01	440	int ret;
4d672e7a	441
610d18f4	442	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
9d94b9e2 AV	443	!timespec_valid(&new->it_value) \|\|
9d94b9e2 AV	444	!timespec_valid(&new->it_interval))
b215e283 DL	445	return -EINVAL;
b215e283 DL	446
2903ff01 AV	447	ret = timerfd_fget(ufd, &f);
	448	if (ret)
	449	return ret;
	450	ctx = f.file->private_data;
b215e283	451
25b68a8f	452	if (isalarm(ctx) && !capable(CAP_WAKE_ALARM)) {
2895a5e5 EC	453	fdput(f);
	454	return -EPERM;
	455	}
	456
9ec26907 TG	457	timerfd_setup_cancel(ctx, flags);
9ec26907 TG	458
4d672e7a DL	459	/*
	460	* We need to stop the existing timer before reprogramming
	461	* it to the new values.
	462	*/
	463	for (;;) {
	464	spin_lock_irq(&ctx->wqh.lock);
11ffa9d6 TP	465
	466	if (isalarm(ctx)) {
	467	if (alarm_try_to_cancel(&ctx->t.alarm) >= 0)
	468	break;
	469	} else {
	470	if (hrtimer_try_to_cancel(&ctx->t.tmr) >= 0)
	471	break;
	472	}
18963c01	473	spin_unlock_irq(&ctx->wqh.lock);
4d672e7a	474	cpu_relax();
b215e283 DL	475	}
b215e283 DL	476
4d672e7a DL	477	/*
	478	* If the timer is expired and it's periodic, we need to advance it
	479	* because the caller may want to know the previous expiration time.
	480	* We do not update "ticks" and "expired" since the timer will be
	481	* re-programmed again in the following timerfd_setup() call.
	482	*/
2456e855	483	if (ctx->expired && ctx->tintv) {
11ffa9d6 TP	484	if (isalarm(ctx))
	485	alarm_forward_now(&ctx->t.alarm, ctx->tintv);
	486	else
	487	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv);
	488	}
b215e283	489
9d94b9e2 AV	490	old->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	491	old->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a DL	492
	493	/*
	494	* Re-program the timer to the new value ...
	495	*/
9d94b9e2	496	ret = timerfd_setup(ctx, flags, new);
4d672e7a DL	497
4d672e7a DL	498	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	499	fdput(f);
99ee5315	500	return ret;
4d672e7a DL	501	}
4d672e7a DL	502
9d94b9e2	503	static int do_timerfd_gettime(int ufd, struct itimerspec *t)
4d672e7a	504	{
2903ff01	505	struct fd f;
4d672e7a	506	struct timerfd_ctx *ctx;
2903ff01 AV	507	int ret = timerfd_fget(ufd, &f);
	508	if (ret)
	509	return ret;
	510	ctx = f.file->private_data;
4d672e7a DL	511
4d672e7a DL	512	spin_lock_irq(&ctx->wqh.lock);
2456e855	513	if (ctx->expired && ctx->tintv) {
4d672e7a	514	ctx->expired = 0;
11ffa9d6 TP	515
	516	if (isalarm(ctx)) {
	517	ctx->ticks +=
	518	alarm_forward_now(
	519	&ctx->t.alarm, ctx->tintv) - 1;
	520	alarm_restart(&ctx->t.alarm);
	521	} else {
	522	ctx->ticks +=
	523	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv)
	524	- 1;
	525	hrtimer_restart(&ctx->t.tmr);
	526	}
4d672e7a	527	}
9d94b9e2 AV	528	t->it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
9d94b9e2 AV	529	t->it_interval = ktime_to_timespec(ctx->tintv);
4d672e7a	530	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	531	fdput(f);
9d94b9e2 AV	532	return 0;
	533	}
	534
	535	SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
	536	const struct itimerspec __user *, utmr,
	537	struct itimerspec __user *, otmr)
	538	{
	539	struct itimerspec new, old;
	540	int ret;
	541
	542	if (copy_from_user(&new, utmr, sizeof(new)))
	543	return -EFAULT;
	544	ret = do_timerfd_settime(ufd, flags, &new, &old);
	545	if (ret)
	546	return ret;
	547	if (otmr && copy_to_user(otmr, &old, sizeof(old)))
	548	return -EFAULT;
	549
	550	return ret;
	551	}
4d672e7a	552
9d94b9e2 AV	553	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
	554	{
	555	struct itimerspec kotmr;
	556	int ret = do_timerfd_gettime(ufd, &kotmr);
	557	if (ret)
	558	return ret;
4d672e7a	559	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
b215e283 DL	560	}
b215e283 DL	561
0e803baf	562	#ifdef CONFIG_COMPAT
9d94b9e2	563	COMPAT_SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
0e803baf HC	564	const struct compat_itimerspec __user *, utmr,
0e803baf HC	565	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	566	{
	567	struct itimerspec new, old;
	568	int ret;
	569
	570	if (get_compat_itimerspec(&new, utmr))
	571	return -EFAULT;
	572	ret = do_timerfd_settime(ufd, flags, &new, &old);
	573	if (ret)
	574	return ret;
	575	if (otmr && put_compat_itimerspec(otmr, &old))
	576	return -EFAULT;
	577	return ret;
	578	}
	579
	580	COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
0e803baf	581	struct compat_itimerspec __user *, otmr)
9d94b9e2 AV	582	{
	583	struct itimerspec kotmr;
	584	int ret = do_timerfd_gettime(ufd, &kotmr);
	585	if (ret)
	586	return ret;
0e803baf	587	return put_compat_itimerspec(otmr, &kotmr) ? -EFAULT: 0;
9d94b9e2 AV	588	}
9d94b9e2 AV	589	#endif