[mirror_ubuntu-eoan-kernel.git] / fs / timerfd.c

// SPDX-License-Identifier: GPL-2.0
/*
 *  fs/timerfd.c
 *
 *  Copyright (C) 2007  Davide Libenzi <davidel@xmailserver.org>
 *
 *
 *  Thanks to Thomas Gleixner for code reviews and useful comments.
 *
 */

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

struct timerfd_ctx {
	union {
		struct hrtimer tmr;
		struct alarm alarm;
	} t;
	ktime_t tintv;
	ktime_t moffs;
	wait_queue_head_t wqh;
	u64 ticks;
	int 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 (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 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_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);
		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_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
COMPAT_SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
		const struct compat_itimerspec __user *, utmr,
		struct compat_itimerspec __user *, otmr)
{
	struct itimerspec64 new, old;
	int ret;

	if (get_compat_itimerspec64(&new, utmr))
		return -EFAULT;
	ret = do_timerfd_settime(ufd, flags, &new, &old);
	if (ret)
		return ret;
	if (otmr && put_compat_itimerspec64(&old, otmr))
		return -EFAULT;
	return ret;
}

COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
		struct compat_itimerspec __user *, otmr)
{
	struct itimerspec64 kotmr;
	int ret = do_timerfd_gettime(ufd, &kotmr);
	if (ret)
		return ret;
	return put_compat_itimerspec64(&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 <davidel@xmailserver.org>
	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>
b215e283 DL	29
b215e283 DL	30	struct timerfd_ctx {
11ffa9d6 TP	31	union {
	32	struct hrtimer tmr;
	33	struct alarm alarm;
	34	} t;
b215e283	35	ktime_t tintv;
99ee5315	36	ktime_t moffs;
b215e283	37	wait_queue_head_t wqh;
4d672e7a	38	u64 ticks;
4d672e7a	39	int clockid;
af9c4957 CG	40	short unsigned expired;
af9c4957 CG	41	short unsigned settime_flags; /* to show in fdinfo */
9ec26907 TG	42	struct rcu_head rcu;
9ec26907 TG	43	struct list_head clist;
1e38da30	44	spinlock_t cancel_lock;
99ee5315	45	bool might_cancel;
b215e283 DL	46	};
b215e283 DL	47
9ec26907 TG	48	static LIST_HEAD(cancel_list);
	49	static DEFINE_SPINLOCK(cancel_lock);
	50
11ffa9d6 TP	51	static inline bool isalarm(struct timerfd_ctx *ctx)
	52	{
	53	return ctx->clockid == CLOCK_REALTIME_ALARM \|\|
	54	ctx->clockid == CLOCK_BOOTTIME_ALARM;
	55	}
	56
b215e283 DL	57	/*
	58	* This gets called when the timer event triggers. We set the "expired"
	59	* flag, but we do not re-arm the timer (in case it's necessary,
2456e855	60	* tintv != 0) until the timer is accessed.
b215e283	61	*/
11ffa9d6	62	static void timerfd_triggered(struct timerfd_ctx *ctx)
b215e283	63	{
b215e283 DL	64	unsigned long flags;
b215e283 DL	65
18963c01	66	spin_lock_irqsave(&ctx->wqh.lock, flags);
b215e283	67	ctx->expired = 1;
4d672e7a	68	ctx->ticks++;
7dda7128	69	wake_up_locked_poll(&ctx->wqh, EPOLLIN);
18963c01	70	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
11ffa9d6	71	}
b215e283	72
11ffa9d6 TP	73	static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
	74	{
	75	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx,
	76	t.tmr);
	77	timerfd_triggered(ctx);
b215e283 DL	78	return HRTIMER_NORESTART;
	79	}
	80
11ffa9d6 TP	81	static enum alarmtimer_restart timerfd_alarmproc(struct alarm *alarm,
	82	ktime_t now)
	83	{
	84	struct timerfd_ctx *ctx = container_of(alarm, struct timerfd_ctx,
	85	t.alarm);
	86	timerfd_triggered(ctx);
	87	return ALARMTIMER_NORESTART;
	88	}
	89
9ec26907 TG	90	/*
9ec26907 TG	91	* Called when the clock was set to cancel the timers in the cancel
1123d939 MA	92	* list. This will wake up processes waiting on these timers. The
	93	* wake-up requires ctx->ticks to be non zero, therefore we increment
	94	* it before calling wake_up_locked().
9ec26907 TG	95	*/
9ec26907 TG	96	void timerfd_clock_was_set(void)
4d672e7a	97	{
2456e855	98	ktime_t moffs = ktime_mono_to_real(0);
9ec26907 TG	99	struct timerfd_ctx *ctx;
9ec26907 TG	100	unsigned long flags;
4d672e7a	101
9ec26907 TG	102	rcu_read_lock();
	103	list_for_each_entry_rcu(ctx, &cancel_list, clist) {
	104	if (!ctx->might_cancel)
	105	continue;
	106	spin_lock_irqsave(&ctx->wqh.lock, flags);
2456e855 TG	107	if (ctx->moffs != moffs) {
2456e855 TG	108	ctx->moffs = KTIME_MAX;
1123d939	109	ctx->ticks++;
7dda7128	110	wake_up_locked_poll(&ctx->wqh, EPOLLIN);
9ec26907 TG	111	}
	112	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
	113	}
	114	rcu_read_unlock();
4d672e7a DL	115	}
4d672e7a DL	116
1e38da30	117	static void __timerfd_remove_cancel(struct timerfd_ctx *ctx)
99ee5315	118	{
9ec26907 TG	119	if (ctx->might_cancel) {
	120	ctx->might_cancel = false;
	121	spin_lock(&cancel_lock);
	122	list_del_rcu(&ctx->clist);
	123	spin_unlock(&cancel_lock);
	124	}
	125	}
99ee5315	126
1e38da30 TG	127	static void timerfd_remove_cancel(struct timerfd_ctx *ctx)
	128	{
	129	spin_lock(&ctx->cancel_lock);
	130	__timerfd_remove_cancel(ctx);
	131	spin_unlock(&ctx->cancel_lock);
	132	}
	133
9ec26907 TG	134	static bool timerfd_canceled(struct timerfd_ctx *ctx)
9ec26907 TG	135	{
2456e855	136	if (!ctx->might_cancel \|\| ctx->moffs != KTIME_MAX)
99ee5315	137	return false;
2456e855	138	ctx->moffs = ktime_mono_to_real(0);
9ec26907 TG	139	return true;
9ec26907 TG	140	}
99ee5315	141
9ec26907 TG	142	static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags)
9ec26907 TG	143	{
1e38da30	144	spin_lock(&ctx->cancel_lock);
11ffa9d6 TP	145	if ((ctx->clockid == CLOCK_REALTIME \|\|
	146	ctx->clockid == CLOCK_REALTIME_ALARM) &&
	147	(flags & TFD_TIMER_ABSTIME) && (flags & TFD_TIMER_CANCEL_ON_SET)) {
9ec26907 TG	148	if (!ctx->might_cancel) {
	149	ctx->might_cancel = true;
	150	spin_lock(&cancel_lock);
	151	list_add_rcu(&ctx->clist, &cancel_list);
	152	spin_unlock(&cancel_lock);
	153	}
1e38da30 TG	154	} else {
1e38da30 TG	155	__timerfd_remove_cancel(ctx);
9ec26907	156	}
1e38da30	157	spin_unlock(&ctx->cancel_lock);
9ec26907	158	}
99ee5315	159
9ec26907 TG	160	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
	161	{
	162	ktime_t remaining;
99ee5315	163
11ffa9d6 TP	164	if (isalarm(ctx))
	165	remaining = alarm_expires_remaining(&ctx->t.alarm);
	166	else
b62526ed	167	remaining = hrtimer_expires_remaining_adjusted(&ctx->t.tmr);
11ffa9d6	168
8b0e1953	169	return remaining < 0 ? 0: remaining;
99ee5315 TG	170	}
	171
	172	static int timerfd_setup(struct timerfd_ctx *ctx, int flags,
bff41203	173	const struct itimerspec64 *ktmr)
b215e283 DL	174	{
	175	enum hrtimer_mode htmode;
	176	ktime_t texp;
99ee5315	177	int clockid = ctx->clockid;
b215e283 DL	178
	179	htmode = (flags & TFD_TIMER_ABSTIME) ?
	180	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	181
bff41203	182	texp = timespec64_to_ktime(ktmr->it_value);
b215e283	183	ctx->expired = 0;
4d672e7a	184	ctx->ticks = 0;
bff41203	185	ctx->tintv = timespec64_to_ktime(ktmr->it_interval);
11ffa9d6 TP	186
	187	if (isalarm(ctx)) {
	188	alarm_init(&ctx->t.alarm,
	189	ctx->clockid == CLOCK_REALTIME_ALARM ?
	190	ALARM_REALTIME : ALARM_BOOTTIME,
	191	timerfd_alarmproc);
	192	} else {
	193	hrtimer_init(&ctx->t.tmr, clockid, htmode);
	194	hrtimer_set_expires(&ctx->t.tmr, texp);
	195	ctx->t.tmr.function = timerfd_tmrproc;
	196	}
	197
2456e855	198	if (texp != 0) {
11ffa9d6 TP	199	if (isalarm(ctx)) {
	200	if (flags & TFD_TIMER_ABSTIME)
	201	alarm_start(&ctx->t.alarm, texp);
	202	else
	203	alarm_start_relative(&ctx->t.alarm, texp);
	204	} else {
	205	hrtimer_start(&ctx->t.tmr, texp, htmode);
	206	}
	207
99ee5315 TG	208	if (timerfd_canceled(ctx))
	209	return -ECANCELED;
	210	}
af9c4957 CG	211
af9c4957 CG	212	ctx->settime_flags = flags & TFD_SETTIME_FLAGS;
99ee5315	213	return 0;
b215e283 DL	214	}
	215
	216	static int timerfd_release(struct inode inode, struct file file)
	217	{
	218	struct timerfd_ctx *ctx = file->private_data;
	219
9ec26907	220	timerfd_remove_cancel(ctx);
11ffa9d6 TP	221
	222	if (isalarm(ctx))
	223	alarm_cancel(&ctx->t.alarm);
	224	else
	225	hrtimer_cancel(&ctx->t.tmr);
9ec26907	226	kfree_rcu(ctx, rcu);
b215e283 DL	227	return 0;
b215e283 DL	228	}
a11e1d43 LT	229
a11e1d43 LT	230	static __poll_t timerfd_poll(struct file file, poll_table wait)
b215e283 DL	231	{
b215e283 DL	232	struct timerfd_ctx *ctx = file->private_data;
a11e1d43 LT	233	__poll_t events = 0;
a11e1d43 LT	234	unsigned long flags;
b215e283	235
a11e1d43	236	poll_wait(file, &ctx->wqh, wait);
b215e283	237
a11e1d43 LT	238	spin_lock_irqsave(&ctx->wqh.lock, flags);
	239	if (ctx->ticks)
	240	events \|= EPOLLIN;
	241	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283	242
a11e1d43	243	return events;
b215e283 DL	244	}
	245
	246	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	247	loff_t *ppos)
	248	{
	249	struct timerfd_ctx *ctx = file->private_data;
	250	ssize_t res;
09828402	251	u64 ticks = 0;
b215e283 DL	252
	253	if (count < sizeof(ticks))
	254	return -EINVAL;
18963c01	255	spin_lock_irq(&ctx->wqh.lock);
8120a8aa MN	256	if (file->f_flags & O_NONBLOCK)
	257	res = -EAGAIN;
	258	else
	259	res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks);
99ee5315	260
9ec26907 TG	261	/*
	262	* If clock has changed, we do not care about the
	263	* ticks and we do not rearm the timer. Userspace must
	264	* reevaluate anyway.
	265	*/
	266	if (timerfd_canceled(ctx)) {
	267	ctx->ticks = 0;
	268	ctx->expired = 0;
	269	res = -ECANCELED;
	270	}
	271
4d672e7a DL	272	if (ctx->ticks) {
4d672e7a DL	273	ticks = ctx->ticks;
99ee5315	274
2456e855	275	if (ctx->expired && ctx->tintv) {
b215e283	276	/*
2456e855	277	* If tintv != 0, this is a periodic timer that
b215e283 DL	278	* needs to be re-armed. We avoid doing it in the timer
	279	* callback to avoid DoS attacks specifying a very
	280	* short timer period.
	281	*/
11ffa9d6 TP	282	if (isalarm(ctx)) {
	283	ticks += alarm_forward_now(
	284	&ctx->t.alarm, ctx->tintv) - 1;
	285	alarm_restart(&ctx->t.alarm);
	286	} else {
	287	ticks += hrtimer_forward_now(&ctx->t.tmr,
	288	ctx->tintv) - 1;
	289	hrtimer_restart(&ctx->t.tmr);
	290	}
4d672e7a DL	291	}
	292	ctx->expired = 0;
	293	ctx->ticks = 0;
b215e283	294	}
18963c01	295	spin_unlock_irq(&ctx->wqh.lock);
b215e283	296	if (ticks)
09828402	297	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	298	return res;
	299	}
	300
af9c4957	301	#ifdef CONFIG_PROC_FS
a3816ab0	302	static void timerfd_show(struct seq_file m, struct file file)
af9c4957 CG	303	{
	304	struct timerfd_ctx *ctx = file->private_data;
	305	struct itimerspec t;
	306
	307	spin_lock_irq(&ctx->wqh.lock);
	308	t.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	309	t.it_interval = ktime_to_timespec(ctx->tintv);
	310	spin_unlock_irq(&ctx->wqh.lock);
	311
a3816ab0 JP	312	seq_printf(m,
	313	"clockid: %d\n"
	314	"ticks: %llu\n"
	315	"settime flags: 0%o\n"
	316	"it_value: (%llu, %llu)\n"
	317	"it_interval: (%llu, %llu)\n",
	318	ctx->clockid,
	319	(unsigned long long)ctx->ticks,
	320	ctx->settime_flags,
	321	(unsigned long long)t.it_value.tv_sec,
	322	(unsigned long long)t.it_value.tv_nsec,
	323	(unsigned long long)t.it_interval.tv_sec,
	324	(unsigned long long)t.it_interval.tv_nsec);
af9c4957 CG	325	}
	326	#else
	327	#define timerfd_show NULL
	328	#endif
	329
5442e9fb CG	330	#ifdef CONFIG_CHECKPOINT_RESTORE
	331	static long timerfd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
	332	{
	333	struct timerfd_ctx *ctx = file->private_data;
	334	int ret = 0;
	335
	336	switch (cmd) {
	337	case TFD_IOC_SET_TICKS: {
	338	u64 ticks;
	339
	340	if (copy_from_user(&ticks, (u64 __user *)arg, sizeof(ticks)))
	341	return -EFAULT;
	342	if (!ticks)
	343	return -EINVAL;
	344
	345	spin_lock_irq(&ctx->wqh.lock);
	346	if (!timerfd_canceled(ctx)) {
	347	ctx->ticks = ticks;
7dda7128	348	wake_up_locked_poll(&ctx->wqh, EPOLLIN);
5442e9fb CG	349	} else
	350	ret = -ECANCELED;
	351	spin_unlock_irq(&ctx->wqh.lock);
	352	break;
	353	}
	354	default:
	355	ret = -ENOTTY;
	356	break;
	357	}
	358
	359	return ret;
	360	}
	361	#else
	362	#define timerfd_ioctl NULL
	363	#endif
	364
b215e283 DL	365	static const struct file_operations timerfd_fops = {
b215e283 DL	366	.release = timerfd_release,
a11e1d43	367	.poll = timerfd_poll,
b215e283	368	.read = timerfd_read,
6038f373	369	.llseek = noop_llseek,
af9c4957	370	.show_fdinfo = timerfd_show,
5442e9fb	371	.unlocked_ioctl = timerfd_ioctl,
b215e283 DL	372	};
b215e283 DL	373
2903ff01	374	static int timerfd_fget(int fd, struct fd *p)
4d672e7a	375	{
2903ff01 AV	376	struct fd f = fdget(fd);
	377	if (!f.file)
	378	return -EBADF;
	379	if (f.file->f_op != &timerfd_fops) {
	380	fdput(f);
	381	return -EINVAL;
4d672e7a	382	}
2903ff01 AV	383	*p = f;
2903ff01 AV	384	return 0;
4d672e7a DL	385	}
4d672e7a DL	386
836f92ad	387	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	388	{
2030a42c	389	int ufd;
b215e283	390	struct timerfd_ctx *ctx;
b215e283	391
e38b36f3 UD	392	/* Check the TFD_* constants for consistency. */
	393	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	394	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	395
610d18f4 DL	396	if ((flags & ~TFD_CREATE_FLAGS) \|\|
610d18f4 DL	397	(clockid != CLOCK_MONOTONIC &&
11ffa9d6 TP	398	clockid != CLOCK_REALTIME &&
11ffa9d6 TP	399	clockid != CLOCK_REALTIME_ALARM &&
4a2378a9	400	clockid != CLOCK_BOOTTIME &&
11ffa9d6	401	clockid != CLOCK_BOOTTIME_ALARM))
b215e283	402	return -EINVAL;
4d672e7a	403
25b68a8f SS	404	if ((clockid == CLOCK_REALTIME_ALARM \|\|
	405	clockid == CLOCK_BOOTTIME_ALARM) &&
	406	!capable(CAP_WAKE_ALARM))
2895a5e5 EC	407	return -EPERM;
2895a5e5 EC	408
4d672e7a DL	409	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	410	if (!ctx)
	411	return -ENOMEM;
	412
	413	init_waitqueue_head(&ctx->wqh);
1e38da30	414	spin_lock_init(&ctx->cancel_lock);
4d672e7a	415	ctx->clockid = clockid;
11ffa9d6 TP	416
	417	if (isalarm(ctx))
	418	alarm_init(&ctx->t.alarm,
	419	ctx->clockid == CLOCK_REALTIME_ALARM ?
	420	ALARM_REALTIME : ALARM_BOOTTIME,
	421	timerfd_alarmproc);
	422	else
	423	hrtimer_init(&ctx->t.tmr, clockid, HRTIMER_MODE_ABS);
	424
2456e855	425	ctx->moffs = ktime_mono_to_real(0);
4d672e7a	426
11fcb6c1	427	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	428	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	429	if (ufd < 0)
4d672e7a	430	kfree(ctx);
4d672e7a DL	431
	432	return ufd;
	433	}
	434
9d94b9e2	435	static int do_timerfd_settime(int ufd, int flags,
bff41203 DD	436	const struct itimerspec64 *new,
bff41203 DD	437	struct itimerspec64 *old)
4d672e7a	438	{
2903ff01	439	struct fd f;
4d672e7a	440	struct timerfd_ctx *ctx;
2903ff01	441	int ret;
4d672e7a	442
610d18f4	443	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
bff41203	444	!itimerspec64_valid(new))
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
bff41203 DD	490	old->it_value = ktime_to_timespec64(timerfd_get_remaining(ctx));
bff41203 DD	491	old->it_interval = ktime_to_timespec64(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
bff41203	503	static int do_timerfd_gettime(int ufd, struct itimerspec64 *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	}
bff41203 DD	528	t->it_value = ktime_to_timespec64(timerfd_get_remaining(ctx));
bff41203 DD	529	t->it_interval = ktime_to_timespec64(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,
6ff84735 DD	536	const struct __kernel_itimerspec __user *, utmr,
6ff84735 DD	537	struct __kernel_itimerspec __user *, otmr)
9d94b9e2	538	{
bff41203	539	struct itimerspec64 new, old;
9d94b9e2 AV	540	int ret;
9d94b9e2 AV	541
bff41203	542	if (get_itimerspec64(&new, utmr))
9d94b9e2 AV	543	return -EFAULT;
	544	ret = do_timerfd_settime(ufd, flags, &new, &old);
	545	if (ret)
	546	return ret;
bff41203	547	if (otmr && put_itimerspec64(&old, otmr))
9d94b9e2 AV	548	return -EFAULT;
	549
	550	return ret;
	551	}
4d672e7a	552
6ff84735	553	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct __kernel_itimerspec __user *, otmr)
9d94b9e2	554	{
bff41203	555	struct itimerspec64 kotmr;
9d94b9e2 AV	556	int ret = do_timerfd_gettime(ufd, &kotmr);
	557	if (ret)
	558	return ret;
bff41203	559	return put_itimerspec64(&kotmr, otmr) ? -EFAULT : 0;
b215e283 DL	560	}
b215e283 DL	561
6ff84735	562	#ifdef CONFIG_COMPAT_32BIT_TIME
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	566	{
bff41203	567	struct itimerspec64 new, old;
9d94b9e2 AV	568	int ret;
9d94b9e2 AV	569
bff41203	570	if (get_compat_itimerspec64(&new, utmr))
9d94b9e2 AV	571	return -EFAULT;
	572	ret = do_timerfd_settime(ufd, flags, &new, &old);
	573	if (ret)
	574	return ret;
bff41203	575	if (otmr && put_compat_itimerspec64(&old, otmr))
9d94b9e2 AV	576	return -EFAULT;
	577	return ret;
	578	}
	579
	580	COMPAT_SYSCALL_DEFINE2(timerfd_gettime, int, ufd,
0e803baf	581	struct compat_itimerspec __user *, otmr)
9d94b9e2	582	{
bff41203	583	struct itimerspec64 kotmr;
9d94b9e2 AV	584	int ret = do_timerfd_gettime(ufd, &kotmr);
	585	if (ret)
	586	return ret;
bff41203	587	return put_compat_itimerspec64(&kotmr, otmr) ? -EFAULT : 0;
9d94b9e2 AV	588	}
9d94b9e2 AV	589	#endif