[linux.git] / fs / timerfd.c

// SPDX-License-Identifier: GPL-2.0
/*
 *  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>
#include <linux/time_namespace.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_poll(&ctx->wqh, EPOLLIN);
	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_poll(&ctx->wqh, EPOLLIN);
		}
		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 itimerspec64 *ktmr)
{
	enum hrtimer_mode htmode;
	ktime_t texp;
	int clockid = ctx->clockid;

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

	texp = timespec64_to_ktime(ktmr->it_value);
	ctx->expired = 0;
	ctx->ticks = 0;
	ctx->tintv = timespec64_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 (flags & TFD_TIMER_ABSTIME)
			texp = timens_ktime_to_host(clockid, texp);
		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 __poll_t timerfd_poll(struct file *file, poll_table *wait)
{
	struct timerfd_ctx *ctx = file->private_data;
	__poll_t events = 0;
	unsigned long flags;

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

	spin_lock_irqsave(&ctx->wqh.lock, flags);
	if (ctx->ticks)
		events |= EPOLLIN;
	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 timespec64 value, interval;

	spin_lock_irq(&ctx->wqh.lock);
	value = ktime_to_timespec64(timerfd_get_remaining(ctx));
	interval = ktime_to_timespec64(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)value.tv_sec,
		   (unsigned long long)value.tv_nsec,
		   (unsigned long long)interval.tv_sec,
		   (unsigned long long)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_poll(&ctx->wqh, EPOLLIN);
		} 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 itimerspec64 *new,
		struct itimerspec64 *old)
{
	struct fd f;
	struct timerfd_ctx *ctx;
	int ret;

	if ((flags & ~TFD_SETTIME_FLAGS) ||
		 !itimerspec64_valid(new))
		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);

		if (isalarm(ctx))
			hrtimer_cancel_wait_running(&ctx->t.alarm.timer);
		else
			hrtimer_cancel_wait_running(&ctx->t.tmr);
	}

	/*
	 * 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_timespec64(timerfd_get_remaining(ctx));
	old->it_interval = ktime_to_timespec64(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 itimerspec64 *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_timespec64(timerfd_get_remaining(ctx));
	t->it_interval = ktime_to_timespec64(ctx->tintv);
	spin_unlock_irq(&ctx->wqh.lock);
	fdput(f);
	return 0;
}

SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
		const struct __kernel_itimerspec __user *, utmr,
		struct __kernel_itimerspec __user *, otmr)
{
	struct itimerspec64 new, old;
	int ret;

	if (get_itimerspec64(&new, utmr))
		return -EFAULT;
	ret = do_timerfd_settime(ufd, flags, &new, &old);
	if (ret)
		return ret;
	if (otmr && put_itimerspec64(&old, otmr))
		return -EFAULT;

	return ret;
}

SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct __kernel_itimerspec __user *, otmr)
{
	struct itimerspec64 kotmr;
	int ret = do_timerfd_gettime(ufd, &kotmr);
	if (ret)
		return ret;
	return put_itimerspec64(&kotmr, otmr) ? -EFAULT : 0;
}

#ifdef CONFIG_COMPAT_32BIT_TIME
SYSCALL_DEFINE4(timerfd_settime32, int, ufd, int, flags,
		const struct old_itimerspec32 __user *, utmr,
		struct old_itimerspec32 __user *, otmr)
{
	struct itimerspec64 new, old;
	int ret;

	if (get_old_itimerspec32(&new, utmr))
		return -EFAULT;
	ret = do_timerfd_settime(ufd, flags, &new, &old);
	if (ret)
		return ret;
	if (otmr && put_old_itimerspec32(&old, otmr))
		return -EFAULT;
	return ret;
}

SYSCALL_DEFINE2(timerfd_gettime32, int, ufd,
		struct old_itimerspec32 __user *, otmr)
{
	struct itimerspec64 kotmr;
	int ret = do_timerfd_gettime(ufd, &kotmr);
	if (ret)
		return ret;
	return put_old_itimerspec32(&kotmr, otmr) ? -EFAULT : 0;
}
#endif
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
b215e283 DL	2	/*
	3	* fs/timerfd.c
	4	*
	5	* Copyright (C) 2007 Davide Libenzi <[email protected]>
	6	*
	7	*
	8	* Thanks to Thomas Gleixner for code reviews and useful comments.
	9	*
	10	*/
	11
11ffa9d6	12	#include <linux/alarmtimer.h>
b215e283 DL	13	#include <linux/file.h>
	14	#include <linux/poll.h>
	15	#include <linux/init.h>
	16	#include <linux/fs.h>
	17	#include <linux/sched.h>
	18	#include <linux/kernel.h>
5a0e3ad6	19	#include <linux/slab.h>
b215e283 DL	20	#include <linux/list.h>
	21	#include <linux/spinlock.h>
	22	#include <linux/time.h>
	23	#include <linux/hrtimer.h>
	24	#include <linux/anon_inodes.h>
	25	#include <linux/timerfd.h>
45cc2b96	26	#include <linux/syscalls.h>
9d94b9e2	27	#include <linux/compat.h>
9ec26907	28	#include <linux/rcupdate.h>
6cd889d4	29	#include <linux/time_namespace.h>
b215e283 DL	30
b215e283 DL	31	struct timerfd_ctx {
11ffa9d6 TP	32	union {
	33	struct hrtimer tmr;
	34	struct alarm alarm;
	35	} t;
b215e283	36	ktime_t tintv;
99ee5315	37	ktime_t moffs;
b215e283	38	wait_queue_head_t wqh;
4d672e7a	39	u64 ticks;
4d672e7a	40	int clockid;
af9c4957 CG	41	short unsigned expired;
af9c4957 CG	42	short unsigned settime_flags; /* to show in fdinfo */
9ec26907 TG	43	struct rcu_head rcu;
9ec26907 TG	44	struct list_head clist;
1e38da30	45	spinlock_t cancel_lock;
99ee5315	46	bool might_cancel;
b215e283 DL	47	};
b215e283 DL	48
9ec26907 TG	49	static LIST_HEAD(cancel_list);
	50	static DEFINE_SPINLOCK(cancel_lock);
	51
11ffa9d6 TP	52	static inline bool isalarm(struct timerfd_ctx *ctx)
	53	{
	54	return ctx->clockid == CLOCK_REALTIME_ALARM \|\|
	55	ctx->clockid == CLOCK_BOOTTIME_ALARM;
	56	}
	57
b215e283 DL	58	/*
	59	* This gets called when the timer event triggers. We set the "expired"
	60	* flag, but we do not re-arm the timer (in case it's necessary,
2456e855	61	* tintv != 0) until the timer is accessed.
b215e283	62	*/
11ffa9d6	63	static void timerfd_triggered(struct timerfd_ctx *ctx)
b215e283	64	{
b215e283 DL	65	unsigned long flags;
b215e283 DL	66
18963c01	67	spin_lock_irqsave(&ctx->wqh.lock, flags);
b215e283	68	ctx->expired = 1;
4d672e7a	69	ctx->ticks++;
7dda7128	70	wake_up_locked_poll(&ctx->wqh, EPOLLIN);
18963c01	71	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
11ffa9d6	72	}
b215e283	73
11ffa9d6 TP	74	static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
	75	{
	76	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx,
	77	t.tmr);
	78	timerfd_triggered(ctx);
b215e283 DL	79	return HRTIMER_NORESTART;
	80	}
	81
11ffa9d6 TP	82	static enum alarmtimer_restart timerfd_alarmproc(struct alarm *alarm,
	83	ktime_t now)
	84	{
	85	struct timerfd_ctx *ctx = container_of(alarm, struct timerfd_ctx,
	86	t.alarm);
	87	timerfd_triggered(ctx);
	88	return ALARMTIMER_NORESTART;
	89	}
	90
9ec26907 TG	91	/*
9ec26907 TG	92	* Called when the clock was set to cancel the timers in the cancel
1123d939 MA	93	* list. This will wake up processes waiting on these timers. The
	94	* wake-up requires ctx->ticks to be non zero, therefore we increment
	95	* it before calling wake_up_locked().
9ec26907 TG	96	*/
9ec26907 TG	97	void timerfd_clock_was_set(void)
4d672e7a	98	{
2456e855	99	ktime_t moffs = ktime_mono_to_real(0);
9ec26907 TG	100	struct timerfd_ctx *ctx;
9ec26907 TG	101	unsigned long flags;
4d672e7a	102
9ec26907 TG	103	rcu_read_lock();
	104	list_for_each_entry_rcu(ctx, &cancel_list, clist) {
	105	if (!ctx->might_cancel)
	106	continue;
	107	spin_lock_irqsave(&ctx->wqh.lock, flags);
2456e855 TG	108	if (ctx->moffs != moffs) {
2456e855 TG	109	ctx->moffs = KTIME_MAX;
1123d939	110	ctx->ticks++;
7dda7128	111	wake_up_locked_poll(&ctx->wqh, EPOLLIN);
9ec26907 TG	112	}
	113	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
	114	}
	115	rcu_read_unlock();
4d672e7a DL	116	}
4d672e7a DL	117
1e38da30	118	static void __timerfd_remove_cancel(struct timerfd_ctx *ctx)
99ee5315	119	{
9ec26907 TG	120	if (ctx->might_cancel) {
	121	ctx->might_cancel = false;
	122	spin_lock(&cancel_lock);
	123	list_del_rcu(&ctx->clist);
	124	spin_unlock(&cancel_lock);
	125	}
	126	}
99ee5315	127
1e38da30 TG	128	static void timerfd_remove_cancel(struct timerfd_ctx *ctx)
	129	{
	130	spin_lock(&ctx->cancel_lock);
	131	__timerfd_remove_cancel(ctx);
	132	spin_unlock(&ctx->cancel_lock);
	133	}
	134
9ec26907 TG	135	static bool timerfd_canceled(struct timerfd_ctx *ctx)
9ec26907 TG	136	{
2456e855	137	if (!ctx->might_cancel \|\| ctx->moffs != KTIME_MAX)
99ee5315	138	return false;
2456e855	139	ctx->moffs = ktime_mono_to_real(0);
9ec26907 TG	140	return true;
9ec26907 TG	141	}
99ee5315	142
9ec26907 TG	143	static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
9ec26907 TG	144	{
1e38da30	145	spin_lock(&ctx->cancel_lock);
11ffa9d6 TP	146	if ((ctx->clockid == CLOCK_REALTIME \|\|
	147	ctx->clockid == CLOCK_REALTIME_ALARM) &&
	148	(flags & TFD_TIMER_ABSTIME) && (flags & TFD_TIMER_CANCEL_ON_SET)) {
9ec26907 TG	149	if (!ctx->might_cancel) {
	150	ctx->might_cancel = true;
	151	spin_lock(&cancel_lock);
	152	list_add_rcu(&ctx->clist, &cancel_list);
	153	spin_unlock(&cancel_lock);
	154	}
1e38da30 TG	155	} else {
1e38da30 TG	156	__timerfd_remove_cancel(ctx);
9ec26907	157	}
1e38da30	158	spin_unlock(&ctx->cancel_lock);
9ec26907	159	}
99ee5315	160
9ec26907 TG	161	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
	162	{
	163	ktime_t remaining;
99ee5315	164
11ffa9d6 TP	165	if (isalarm(ctx))
	166	remaining = alarm_expires_remaining(&ctx->t.alarm);
	167	else
b62526ed	168	remaining = hrtimer_expires_remaining_adjusted(&ctx->t.tmr);
11ffa9d6	169
8b0e1953	170	return remaining < 0 ? 0: remaining;
99ee5315 TG	171	}
	172
	173	static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
bff41203	174	const struct itimerspec64 *ktmr)
b215e283 DL	175	{
	176	enum hrtimer_mode htmode;
	177	ktime_t texp;
99ee5315	178	int clockid = ctx->clockid;
b215e283 DL	179
	180	htmode = (flags & TFD_TIMER_ABSTIME) ?
	181	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	182
bff41203	183	texp = timespec64_to_ktime(ktmr->it_value);
b215e283	184	ctx->expired = 0;
4d672e7a	185	ctx->ticks = 0;
bff41203	186	ctx->tintv = timespec64_to_ktime(ktmr->it_interval);
11ffa9d6 TP	187
	188	if (isalarm(ctx)) {
	189	alarm_init(&ctx->t.alarm,
	190	ctx->clockid == CLOCK_REALTIME_ALARM ?
	191	ALARM_REALTIME : ALARM_BOOTTIME,
	192	timerfd_alarmproc);
	193	} else {
	194	hrtimer_init(&ctx->t.tmr, clockid, htmode);
	195	hrtimer_set_expires(&ctx->t.tmr, texp);
	196	ctx->t.tmr.function = timerfd_tmrproc;
	197	}
	198
2456e855	199	if (texp != 0) {
6cd889d4 AV	200	if (flags & TFD_TIMER_ABSTIME)
6cd889d4 AV	201	texp = timens_ktime_to_host(clockid, texp);
11ffa9d6 TP	202	if (isalarm(ctx)) {
	203	if (flags & TFD_TIMER_ABSTIME)
	204	alarm_start(&ctx->t.alarm, texp);
	205	else
	206	alarm_start_relative(&ctx->t.alarm, texp);
	207	} else {
	208	hrtimer_start(&ctx->t.tmr, texp, htmode);
	209	}
	210
99ee5315 TG	211	if (timerfd_canceled(ctx))
	212	return -ECANCELED;
	213	}
af9c4957 CG	214
af9c4957 CG	215	ctx->settime_flags = flags & TFD_SETTIME_FLAGS;
99ee5315	216	return 0;
b215e283 DL	217	}
	218
	219	static int timerfd_release(struct inode inode, struct file file)
	220	{
	221	struct timerfd_ctx *ctx = file->private_data;
	222
9ec26907	223	timerfd_remove_cancel(ctx);
11ffa9d6 TP	224
	225	if (isalarm(ctx))
	226	alarm_cancel(&ctx->t.alarm);
	227	else
	228	hrtimer_cancel(&ctx->t.tmr);
9ec26907	229	kfree_rcu(ctx, rcu);
b215e283 DL	230	return 0;
b215e283 DL	231	}
a11e1d43 LT	232
a11e1d43 LT	233	static __poll_t timerfd_poll(struct file file, poll_table wait)
b215e283 DL	234	{
b215e283 DL	235	struct timerfd_ctx *ctx = file->private_data;
a11e1d43 LT	236	__poll_t events = 0;
a11e1d43 LT	237	unsigned long flags;
b215e283	238
a11e1d43	239	poll_wait(file, &ctx->wqh, wait);
b215e283	240
a11e1d43 LT	241	spin_lock_irqsave(&ctx->wqh.lock, flags);
	242	if (ctx->ticks)
	243	events \|= EPOLLIN;
	244	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283	245
a11e1d43	246	return events;
b215e283 DL	247	}
	248
	249	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	250	loff_t *ppos)
	251	{
	252	struct timerfd_ctx *ctx = file->private_data;
	253	ssize_t res;
09828402	254	u64 ticks = 0;
b215e283 DL	255
	256	if (count < sizeof(ticks))
	257	return -EINVAL;
18963c01	258	spin_lock_irq(&ctx->wqh.lock);
8120a8aa MN	259	if (file->f_flags & O_NONBLOCK)
	260	res = -EAGAIN;
	261	else
	262	res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);
99ee5315	263
9ec26907 TG	264	/*
	265	* If clock has changed, we do not care about the
	266	* ticks and we do not rearm the timer. Userspace must
	267	* reevaluate anyway.
	268	*/
	269	if (timerfd_canceled(ctx)) {
	270	ctx->ticks = 0;
	271	ctx->expired = 0;
	272	res = -ECANCELED;
	273	}
	274
4d672e7a DL	275	if (ctx->ticks) {
4d672e7a DL	276	ticks = ctx->ticks;
99ee5315	277
2456e855	278	if (ctx->expired && ctx->tintv) {
b215e283	279	/*
2456e855	280	* If tintv != 0, this is a periodic timer that
b215e283 DL	281	* needs to be re-armed. We avoid doing it in the timer
	282	* callback to avoid DoS attacks specifying a very
	283	* short timer period.
	284	*/
11ffa9d6 TP	285	if (isalarm(ctx)) {
	286	ticks += alarm_forward_now(
	287	&ctx->t.alarm, ctx->tintv) - 1;
	288	alarm_restart(&ctx->t.alarm);
	289	} else {
	290	ticks += hrtimer_forward_now(&ctx->t.tmr,
	291	ctx->tintv) - 1;
	292	hrtimer_restart(&ctx->t.tmr);
	293	}
4d672e7a DL	294	}
	295	ctx->expired = 0;
	296	ctx->ticks = 0;
b215e283	297	}
18963c01	298	spin_unlock_irq(&ctx->wqh.lock);
b215e283	299	if (ticks)
09828402	300	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	301	return res;
	302	}
	303
af9c4957	304	#ifdef CONFIG_PROC_FS
a3816ab0	305	static void timerfd_show(struct seq_file m, struct file file)
af9c4957 CG	306	{
af9c4957 CG	307	struct timerfd_ctx *ctx = file->private_data;
bde9e963	308	struct timespec64 value, interval;
af9c4957 CG	309
af9c4957 CG	310	spin_lock_irq(&ctx->wqh.lock);
bde9e963 AB	311	value = ktime_to_timespec64(timerfd_get_remaining(ctx));
bde9e963 AB	312	interval = ktime_to_timespec64(ctx->tintv);
af9c4957 CG	313	spin_unlock_irq(&ctx->wqh.lock);
af9c4957 CG	314
a3816ab0 JP	315	seq_printf(m,
	316	"clockid: %d\n"
	317	"ticks: %llu\n"
	318	"settime flags: 0%o\n"
	319	"it_value: (%llu, %llu)\n"
	320	"it_interval: (%llu, %llu)\n",
	321	ctx->clockid,
	322	(unsigned long long)ctx->ticks,
	323	ctx->settime_flags,
bde9e963 AB	324	(unsigned long long)value.tv_sec,
	325	(unsigned long long)value.tv_nsec,
	326	(unsigned long long)interval.tv_sec,
	327	(unsigned long long)interval.tv_nsec);
af9c4957 CG	328	}
	329	#else
	330	#define timerfd_show NULL
	331	#endif
	332
5442e9fb CG	333	#ifdef CONFIG_CHECKPOINT_RESTORE
	334	static long timerfd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
	335	{
	336	struct timerfd_ctx *ctx = file->private_data;
	337	int ret = 0;
	338
	339	switch (cmd) {
	340	case TFD_IOC_SET_TICKS: {
	341	u64 ticks;
	342
	343	if (copy_from_user(&ticks, (u64 __user *)arg, sizeof(ticks)))
	344	return -EFAULT;
	345	if (!ticks)
	346	return -EINVAL;
	347
	348	spin_lock_irq(&ctx->wqh.lock);
	349	if (!timerfd_canceled(ctx)) {
	350	ctx->ticks = ticks;
7dda7128	351	wake_up_locked_poll(&ctx->wqh, EPOLLIN);
5442e9fb CG	352	} else
	353	ret = -ECANCELED;
	354	spin_unlock_irq(&ctx->wqh.lock);
	355	break;
	356	}
	357	default:
	358	ret = -ENOTTY;
	359	break;
	360	}
	361
	362	return ret;
	363	}
	364	#else
	365	#define timerfd_ioctl NULL
	366	#endif
	367
b215e283 DL	368	static const struct file_operations timerfd_fops = {
b215e283 DL	369	.release = timerfd_release,
a11e1d43	370	.poll = timerfd_poll,
b215e283	371	.read = timerfd_read,
6038f373	372	.llseek = noop_llseek,
af9c4957	373	.show_fdinfo = timerfd_show,
5442e9fb	374	.unlocked_ioctl = timerfd_ioctl,
b215e283 DL	375	};
b215e283 DL	376
2903ff01	377	static int timerfd_fget(int fd, struct fd *p)
4d672e7a	378	{
2903ff01 AV	379	struct fd f = fdget(fd);
	380	if (!f.file)
	381	return -EBADF;
	382	if (f.file->f_op != &timerfd_fops) {
	383	fdput(f);
	384	return -EINVAL;
4d672e7a	385	}
2903ff01 AV	386	*p = f;
2903ff01 AV	387	return 0;
4d672e7a DL	388	}
4d672e7a DL	389
836f92ad	390	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	391	{
2030a42c	392	int ufd;
b215e283	393	struct timerfd_ctx *ctx;
b215e283	394
e38b36f3 UD	395	/* Check the TFD_* constants for consistency. */
	396	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	397	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	398
610d18f4 DL	399	if ((flags & ~TFD_CREATE_FLAGS) \|\|
610d18f4 DL	400	(clockid != CLOCK_MONOTONIC &&
11ffa9d6 TP	401	clockid != CLOCK_REALTIME &&
11ffa9d6 TP	402	clockid != CLOCK_REALTIME_ALARM &&
4a2378a9	403	clockid != CLOCK_BOOTTIME &&
11ffa9d6	404	clockid != CLOCK_BOOTTIME_ALARM))
b215e283	405	return -EINVAL;
4d672e7a	406
25b68a8f SS	407	if ((clockid == CLOCK_REALTIME_ALARM \|\|
	408	clockid == CLOCK_BOOTTIME_ALARM) &&
	409	!capable(CAP_WAKE_ALARM))
2895a5e5 EC	410	return -EPERM;
2895a5e5 EC	411
4d672e7a DL	412	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	413	if (!ctx)
	414	return -ENOMEM;
	415
	416	init_waitqueue_head(&ctx->wqh);
1e38da30	417	spin_lock_init(&ctx->cancel_lock);
4d672e7a	418	ctx->clockid = clockid;
11ffa9d6 TP	419
	420	if (isalarm(ctx))
	421	alarm_init(&ctx->t.alarm,
	422	ctx->clockid == CLOCK_REALTIME_ALARM ?
	423	ALARM_REALTIME : ALARM_BOOTTIME,
	424	timerfd_alarmproc);
	425	else
	426	hrtimer_init(&ctx->t.tmr, clockid, HRTIMER_MODE_ABS);
	427
2456e855	428	ctx->moffs = ktime_mono_to_real(0);
4d672e7a	429
11fcb6c1	430	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	431	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	432	if (ufd < 0)
4d672e7a	433	kfree(ctx);
4d672e7a DL	434
	435	return ufd;
	436	}
	437
9d94b9e2	438	static int do_timerfd_settime(int ufd, int flags,
bff41203 DD	439	const struct itimerspec64 *new,
bff41203 DD	440	struct itimerspec64 *old)
4d672e7a	441	{
2903ff01	442	struct fd f;
4d672e7a	443	struct timerfd_ctx *ctx;
2903ff01	444	int ret;
4d672e7a	445
610d18f4	446	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
bff41203	447	!itimerspec64_valid(new))
b215e283 DL	448	return -EINVAL;
b215e283 DL	449
2903ff01 AV	450	ret = timerfd_fget(ufd, &f);
	451	if (ret)
	452	return ret;
	453	ctx = f.file->private_data;
b215e283	454
25b68a8f	455	if (isalarm(ctx) && !capable(CAP_WAKE_ALARM)) {
2895a5e5 EC	456	fdput(f);
	457	return -EPERM;
	458	}
	459
9ec26907 TG	460	timerfd_setup_cancel(ctx, flags);
9ec26907 TG	461
4d672e7a DL	462	/*
	463	* We need to stop the existing timer before reprogramming
	464	* it to the new values.
	465	*/
	466	for (;;) {
	467	spin_lock_irq(&ctx->wqh.lock);
11ffa9d6 TP	468
	469	if (isalarm(ctx)) {
	470	if (alarm_try_to_cancel(&ctx->t.alarm) >= 0)
	471	break;
	472	} else {
	473	if (hrtimer_try_to_cancel(&ctx->t.tmr) >= 0)
	474	break;
	475	}
18963c01	476	spin_unlock_irq(&ctx->wqh.lock);
a125ecc1 AMG	477
	478	if (isalarm(ctx))
	479	hrtimer_cancel_wait_running(&ctx->t.alarm.timer);
	480	else
	481	hrtimer_cancel_wait_running(&ctx->t.tmr);
b215e283 DL	482	}
b215e283 DL	483
4d672e7a DL	484	/*
	485	* If the timer is expired and it's periodic, we need to advance it
	486	* because the caller may want to know the previous expiration time.
	487	* We do not update "ticks" and "expired" since the timer will be
	488	* re-programmed again in the following timerfd_setup() call.
	489	*/
2456e855	490	if (ctx->expired && ctx->tintv) {
11ffa9d6 TP	491	if (isalarm(ctx))
	492	alarm_forward_now(&ctx->t.alarm, ctx->tintv);
	493	else
	494	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv);
	495	}
b215e283	496
bff41203 DD	497	old->it_value = ktime_to_timespec64(timerfd_get_remaining(ctx));
bff41203 DD	498	old->it_interval = ktime_to_timespec64(ctx->tintv);
4d672e7a DL	499
	500	/*
	501	* Re-program the timer to the new value ...
	502	*/
9d94b9e2	503	ret = timerfd_setup(ctx, flags, new);
4d672e7a DL	504
4d672e7a DL	505	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	506	fdput(f);
99ee5315	507	return ret;
4d672e7a DL	508	}
4d672e7a DL	509
bff41203	510	static int do_timerfd_gettime(int ufd, struct itimerspec64 *t)
4d672e7a	511	{
2903ff01	512	struct fd f;
4d672e7a	513	struct timerfd_ctx *ctx;
2903ff01 AV	514	int ret = timerfd_fget(ufd, &f);
	515	if (ret)
	516	return ret;
	517	ctx = f.file->private_data;
4d672e7a DL	518
4d672e7a DL	519	spin_lock_irq(&ctx->wqh.lock);
2456e855	520	if (ctx->expired && ctx->tintv) {
4d672e7a	521	ctx->expired = 0;
11ffa9d6 TP	522
	523	if (isalarm(ctx)) {
	524	ctx->ticks +=
	525	alarm_forward_now(
	526	&ctx->t.alarm, ctx->tintv) - 1;
	527	alarm_restart(&ctx->t.alarm);
	528	} else {
	529	ctx->ticks +=
	530	hrtimer_forward_now(&ctx->t.tmr, ctx->tintv)
	531	- 1;
	532	hrtimer_restart(&ctx->t.tmr);
	533	}
4d672e7a	534	}
bff41203 DD	535	t->it_value = ktime_to_timespec64(timerfd_get_remaining(ctx));
bff41203 DD	536	t->it_interval = ktime_to_timespec64(ctx->tintv);
4d672e7a	537	spin_unlock_irq(&ctx->wqh.lock);
2903ff01	538	fdput(f);
9d94b9e2 AV	539	return 0;
	540	}
	541
	542	SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
6ff84735 DD	543	const struct __kernel_itimerspec __user *, utmr,
6ff84735 DD	544	struct __kernel_itimerspec __user *, otmr)
9d94b9e2	545	{
bff41203	546	struct itimerspec64 new, old;
9d94b9e2 AV	547	int ret;
9d94b9e2 AV	548
bff41203	549	if (get_itimerspec64(&new, utmr))
9d94b9e2 AV	550	return -EFAULT;
	551	ret = do_timerfd_settime(ufd, flags, &new, &old);
	552	if (ret)
	553	return ret;
bff41203	554	if (otmr && put_itimerspec64(&old, otmr))
9d94b9e2 AV	555	return -EFAULT;
	556
	557	return ret;
	558	}
4d672e7a	559
6ff84735	560	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct __kernel_itimerspec __user *, otmr)
9d94b9e2	561	{
bff41203	562	struct itimerspec64 kotmr;
9d94b9e2 AV	563	int ret = do_timerfd_gettime(ufd, &kotmr);
	564	if (ret)
	565	return ret;
bff41203	566	return put_itimerspec64(&kotmr, otmr) ? -EFAULT : 0;
b215e283 DL	567	}
b215e283 DL	568
6ff84735	569	#ifdef CONFIG_COMPAT_32BIT_TIME
8dabe724	570	SYSCALL_DEFINE4(timerfd_settime32, int, ufd, int, flags,
9afc5eee AB	571	const struct old_itimerspec32 __user *, utmr,
9afc5eee AB	572	struct old_itimerspec32 __user *, otmr)
9d94b9e2	573	{
bff41203	574	struct itimerspec64 new, old;
9d94b9e2 AV	575	int ret;
9d94b9e2 AV	576
9afc5eee	577	if (get_old_itimerspec32(&new, utmr))
9d94b9e2 AV	578	return -EFAULT;
	579	ret = do_timerfd_settime(ufd, flags, &new, &old);
	580	if (ret)
	581	return ret;
9afc5eee	582	if (otmr && put_old_itimerspec32(&old, otmr))
9d94b9e2 AV	583	return -EFAULT;
	584	return ret;
	585	}
	586
8dabe724	587	SYSCALL_DEFINE2(timerfd_gettime32, int, ufd,
9afc5eee	588	struct old_itimerspec32 __user *, otmr)
9d94b9e2	589	{
bff41203	590	struct itimerspec64 kotmr;
9d94b9e2 AV	591	int ret = do_timerfd_gettime(ufd, &kotmr);
	592	if (ret)
	593	return ret;
9afc5eee	594	return put_old_itimerspec32(&kotmr, otmr) ? -EFAULT : 0;
9d94b9e2 AV	595	}
9d94b9e2 AV	596	#endif