[net-next-2.6.git] / fs / timerfd.c

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

#include <linux/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>

struct timerfd_ctx {
	struct hrtimer tmr;
	ktime_t tintv;
	wait_queue_head_t wqh;
	u64 ticks;
	int expired;
	int clockid;
};

/*
 * This gets called when the timer event triggers. We set the "expired"
 * flag, but we do not re-arm the timer (in case it's necessary,
 * tintv.tv64 != 0) until the timer is accessed.
 */
static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
{
	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx, tmr);
	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);

	return HRTIMER_NORESTART;
}

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

	remaining = hrtimer_expires_remaining(&ctx->tmr);
	return remaining.tv64 < 0 ? ktime_set(0, 0): remaining;
}

static void timerfd_setup(struct timerfd_ctx *ctx, int flags,
			  const struct itimerspec *ktmr)
{
	enum hrtimer_mode htmode;
	ktime_t texp;

	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);
	hrtimer_init(&ctx->tmr, ctx->clockid, htmode);
	hrtimer_set_expires(&ctx->tmr, texp);
	ctx->tmr.function = timerfd_tmrproc;
	if (texp.tv64 != 0)
		hrtimer_start(&ctx->tmr, texp, htmode);
}

static int timerfd_release(struct inode *inode, struct file *file)
{
	struct timerfd_ctx *ctx = file->private_data;

	hrtimer_cancel(&ctx->tmr);
	kfree(ctx);
	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;
	DECLARE_WAITQUEUE(wait, current);

	if (count < sizeof(ticks))
		return -EINVAL;
	spin_lock_irq(&ctx->wqh.lock);
	res = -EAGAIN;
	if (!ctx->ticks && !(file->f_flags & O_NONBLOCK)) {
		__add_wait_queue(&ctx->wqh, &wait);
		for (res = 0;;) {
			set_current_state(TASK_INTERRUPTIBLE);
			if (ctx->ticks) {
				res = 0;
				break;
			}
			if (signal_pending(current)) {
				res = -ERESTARTSYS;
				break;
			}
			spin_unlock_irq(&ctx->wqh.lock);
			schedule();
			spin_lock_irq(&ctx->wqh.lock);
		}
		__remove_wait_queue(&ctx->wqh, &wait);
		__set_current_state(TASK_RUNNING);
	}
	if (ctx->ticks) {
		ticks = ctx->ticks;
		if (ctx->expired && ctx->tintv.tv64) {
			/*
			 * If tintv.tv64 != 0, this is a periodic timer that
			 * needs to be re-armed. We avoid doing it in the timer
			 * callback to avoid DoS attacks specifying a very
			 * short timer period.
			 */
			ticks += hrtimer_forward_now(&ctx->tmr,
						     ctx->tintv) - 1;
			hrtimer_restart(&ctx->tmr);
		}
		ctx->expired = 0;
		ctx->ticks = 0;
	}
	spin_unlock_irq(&ctx->wqh.lock);
	if (ticks)
		res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
	return res;
}

static const struct file_operations timerfd_fops = {
	.release	= timerfd_release,
	.poll		= timerfd_poll,
	.read		= timerfd_read,
};

static struct file *timerfd_fget(int fd)
{
	struct file *file;

	file = fget(fd);
	if (!file)
		return ERR_PTR(-EBADF);
	if (file->f_op != &timerfd_fops) {
		fput(file);
		return ERR_PTR(-EINVAL);
	}

	return file;
}

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))
		return -EINVAL;

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

	init_waitqueue_head(&ctx->wqh);
	ctx->clockid = clockid;
	hrtimer_init(&ctx->tmr, clockid, HRTIMER_MODE_ABS);

	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
			       O_RDWR | (flags & TFD_SHARED_FCNTL_FLAGS));
	if (ufd < 0)
		kfree(ctx);

	return ufd;
}

SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
		const struct itimerspec __user *, utmr,
		struct itimerspec __user *, otmr)
{
	struct file *file;
	struct timerfd_ctx *ctx;
	struct itimerspec ktmr, kotmr;

	if (copy_from_user(&ktmr, utmr, sizeof(ktmr)))
		return -EFAULT;

	if ((flags & ~TFD_SETTIME_FLAGS) ||
	    !timespec_valid(&ktmr.it_value) ||
	    !timespec_valid(&ktmr.it_interval))
		return -EINVAL;

	file = timerfd_fget(ufd);
	if (IS_ERR(file))
		return PTR_ERR(file);
	ctx = file->private_data;

	/*
	 * We need to stop the existing timer before reprogramming
	 * it to the new values.
	 */
	for (;;) {
		spin_lock_irq(&ctx->wqh.lock);
		if (hrtimer_try_to_cancel(&ctx->tmr) >= 0)
			break;
		spin_unlock_irq(&ctx->wqh.lock);
		cpu_relax();
	}

	/*
	 * If the timer is expired and it's periodic, we need to advance it
	 * because the caller may want to know the previous expiration time.
	 * We do not update "ticks" and "expired" since the timer will be
	 * re-programmed again in the following timerfd_setup() call.
	 */
	if (ctx->expired && ctx->tintv.tv64)
		hrtimer_forward_now(&ctx->tmr, ctx->tintv);

	kotmr.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	kotmr.it_interval = ktime_to_timespec(ctx->tintv);

	/*
	 * Re-program the timer to the new value ...
	 */
	timerfd_setup(ctx, flags, &ktmr);

	spin_unlock_irq(&ctx->wqh.lock);
	fput(file);
	if (otmr && copy_to_user(otmr, &kotmr, sizeof(kotmr)))
		return -EFAULT;

	return 0;
}

SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
{
	struct file *file;
	struct timerfd_ctx *ctx;
	struct itimerspec kotmr;

	file = timerfd_fget(ufd);
	if (IS_ERR(file))
		return PTR_ERR(file);
	ctx = file->private_data;

	spin_lock_irq(&ctx->wqh.lock);
	if (ctx->expired && ctx->tintv.tv64) {
		ctx->expired = 0;
		ctx->ticks +=
			hrtimer_forward_now(&ctx->tmr, ctx->tintv) - 1;
		hrtimer_restart(&ctx->tmr);
	}
	kotmr.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	kotmr.it_interval = ktime_to_timespec(ctx->tintv);
	spin_unlock_irq(&ctx->wqh.lock);
	fput(file);

	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
}
Commit	Line	Data
b215e283 DL	1	/*
	2	* fs/timerfd.c
	3	*
	4	* Copyright (C) 2007 Davide Libenzi <davidel@xmailserver.org>
	5	*
	6	*
	7	* Thanks to Thomas Gleixner for code reviews and useful comments.
	8	*
	9	*/
	10
	11	#include <linux/file.h>
	12	#include <linux/poll.h>
	13	#include <linux/init.h>
	14	#include <linux/fs.h>
	15	#include <linux/sched.h>
	16	#include <linux/kernel.h>
5a0e3ad6	17	#include <linux/slab.h>
b215e283 DL	18	#include <linux/list.h>
	19	#include <linux/spinlock.h>
	20	#include <linux/time.h>
	21	#include <linux/hrtimer.h>
	22	#include <linux/anon_inodes.h>
	23	#include <linux/timerfd.h>
45cc2b96	24	#include <linux/syscalls.h>
b215e283 DL	25
	26	struct timerfd_ctx {
	27	struct hrtimer tmr;
	28	ktime_t tintv;
b215e283	29	wait_queue_head_t wqh;
4d672e7a	30	u64 ticks;
b215e283	31	int expired;
4d672e7a	32	int clockid;
b215e283 DL	33	};
	34
	35	/*
	36	* This gets called when the timer event triggers. We set the "expired"
	37	* flag, but we do not re-arm the timer (in case it's necessary,
4d672e7a	38	* tintv.tv64 != 0) until the timer is accessed.
b215e283 DL	39	*/
	40	static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr)
	41	{
	42	struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx, tmr);
	43	unsigned long flags;
	44
18963c01	45	spin_lock_irqsave(&ctx->wqh.lock, flags);
b215e283	46	ctx->expired = 1;
4d672e7a	47	ctx->ticks++;
b215e283	48	wake_up_locked(&ctx->wqh);
18963c01	49	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	50
	51	return HRTIMER_NORESTART;
	52	}
	53
4d672e7a DL	54	static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx)
4d672e7a DL	55	{
76369470	56	ktime_t remaining;
4d672e7a	57
76369470	58	remaining = hrtimer_expires_remaining(&ctx->tmr);
4d672e7a DL	59	return remaining.tv64 < 0 ? ktime_set(0, 0): remaining;
	60	}
	61
	62	static void timerfd_setup(struct timerfd_ctx *ctx, int flags,
b215e283 DL	63	const struct itimerspec *ktmr)
	64	{
	65	enum hrtimer_mode htmode;
	66	ktime_t texp;
	67
	68	htmode = (flags & TFD_TIMER_ABSTIME) ?
	69	HRTIMER_MODE_ABS: HRTIMER_MODE_REL;
	70
	71	texp = timespec_to_ktime(ktmr->it_value);
	72	ctx->expired = 0;
4d672e7a	73	ctx->ticks = 0;
b215e283	74	ctx->tintv = timespec_to_ktime(ktmr->it_interval);
4d672e7a	75	hrtimer_init(&ctx->tmr, ctx->clockid, htmode);
76369470	76	hrtimer_set_expires(&ctx->tmr, texp);
b215e283 DL	77	ctx->tmr.function = timerfd_tmrproc;
	78	if (texp.tv64 != 0)
	79	hrtimer_start(&ctx->tmr, texp, htmode);
	80	}
	81
	82	static int timerfd_release(struct inode inode, struct file file)
	83	{
	84	struct timerfd_ctx *ctx = file->private_data;
	85
	86	hrtimer_cancel(&ctx->tmr);
	87	kfree(ctx);
	88	return 0;
	89	}
	90
	91	static unsigned int timerfd_poll(struct file file, poll_table wait)
	92	{
	93	struct timerfd_ctx *ctx = file->private_data;
	94	unsigned int events = 0;
	95	unsigned long flags;
	96
	97	poll_wait(file, &ctx->wqh, wait);
	98
18963c01	99	spin_lock_irqsave(&ctx->wqh.lock, flags);
4d672e7a	100	if (ctx->ticks)
b215e283	101	events \|= POLLIN;
18963c01	102	spin_unlock_irqrestore(&ctx->wqh.lock, flags);
b215e283 DL	103
	104	return events;
	105	}
	106
	107	static ssize_t timerfd_read(struct file file, char __user buf, size_t count,
	108	loff_t *ppos)
	109	{
	110	struct timerfd_ctx *ctx = file->private_data;
	111	ssize_t res;
09828402	112	u64 ticks = 0;
b215e283 DL	113	DECLARE_WAITQUEUE(wait, current);
	114
	115	if (count < sizeof(ticks))
	116	return -EINVAL;
18963c01	117	spin_lock_irq(&ctx->wqh.lock);
b215e283	118	res = -EAGAIN;
4d672e7a	119	if (!ctx->ticks && !(file->f_flags & O_NONBLOCK)) {
b215e283 DL	120	__add_wait_queue(&ctx->wqh, &wait);
	121	for (res = 0;;) {
	122	set_current_state(TASK_INTERRUPTIBLE);
4d672e7a	123	if (ctx->ticks) {
b215e283 DL	124	res = 0;
	125	break;
	126	}
	127	if (signal_pending(current)) {
	128	res = -ERESTARTSYS;
	129	break;
	130	}
18963c01	131	spin_unlock_irq(&ctx->wqh.lock);
b215e283	132	schedule();
18963c01	133	spin_lock_irq(&ctx->wqh.lock);
b215e283 DL	134	}
	135	__remove_wait_queue(&ctx->wqh, &wait);
	136	__set_current_state(TASK_RUNNING);
	137	}
4d672e7a DL	138	if (ctx->ticks) {
	139	ticks = ctx->ticks;
	140	if (ctx->expired && ctx->tintv.tv64) {
b215e283 DL	141	/*
	142	* If tintv.tv64 != 0, this is a periodic timer that
	143	* needs to be re-armed. We avoid doing it in the timer
	144	* callback to avoid DoS attacks specifying a very
	145	* short timer period.
	146	*/
4d672e7a DL	147	ticks += hrtimer_forward_now(&ctx->tmr,
4d672e7a DL	148	ctx->tintv) - 1;
b215e283	149	hrtimer_restart(&ctx->tmr);
4d672e7a DL	150	}
	151	ctx->expired = 0;
	152	ctx->ticks = 0;
b215e283	153	}
18963c01	154	spin_unlock_irq(&ctx->wqh.lock);
b215e283	155	if (ticks)
09828402	156	res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks);
b215e283 DL	157	return res;
	158	}
	159
	160	static const struct file_operations timerfd_fops = {
	161	.release = timerfd_release,
	162	.poll = timerfd_poll,
	163	.read = timerfd_read,
	164	};
	165
4d672e7a DL	166	static struct file *timerfd_fget(int fd)
	167	{
	168	struct file *file;
	169
	170	file = fget(fd);
	171	if (!file)
	172	return ERR_PTR(-EBADF);
	173	if (file->f_op != &timerfd_fops) {
	174	fput(file);
	175	return ERR_PTR(-EINVAL);
	176	}
	177
	178	return file;
	179	}
	180
836f92ad	181	SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags)
b215e283	182	{
2030a42c	183	int ufd;
b215e283	184	struct timerfd_ctx *ctx;
b215e283	185
e38b36f3 UD	186	/* Check the TFD_* constants for consistency. */
	187	BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC);
	188	BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK);
	189
610d18f4 DL	190	if ((flags & ~TFD_CREATE_FLAGS) \|\|
	191	(clockid != CLOCK_MONOTONIC &&
	192	clockid != CLOCK_REALTIME))
b215e283	193	return -EINVAL;
4d672e7a DL	194
	195	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
	196	if (!ctx)
	197	return -ENOMEM;
	198
	199	init_waitqueue_head(&ctx->wqh);
	200	ctx->clockid = clockid;
	201	hrtimer_init(&ctx->tmr, clockid, HRTIMER_MODE_ABS);
	202
11fcb6c1	203	ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx,
628ff7c1	204	O_RDWR \| (flags & TFD_SHARED_FCNTL_FLAGS));
2030a42c	205	if (ufd < 0)
4d672e7a	206	kfree(ctx);
4d672e7a DL	207
	208	return ufd;
	209	}
	210
836f92ad HC	211	SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags,
	212	const struct itimerspec __user *, utmr,
	213	struct itimerspec __user *, otmr)
4d672e7a DL	214	{
	215	struct file *file;
	216	struct timerfd_ctx *ctx;
	217	struct itimerspec ktmr, kotmr;
	218
	219	if (copy_from_user(&ktmr, utmr, sizeof(ktmr)))
	220	return -EFAULT;
	221
610d18f4 DL	222	if ((flags & ~TFD_SETTIME_FLAGS) \|\|
610d18f4 DL	223	!timespec_valid(&ktmr.it_value) \|\|
b215e283 DL	224	!timespec_valid(&ktmr.it_interval))
	225	return -EINVAL;
	226
4d672e7a DL	227	file = timerfd_fget(ufd);
	228	if (IS_ERR(file))
	229	return PTR_ERR(file);
	230	ctx = file->private_data;
b215e283	231
4d672e7a DL	232	/*
	233	* We need to stop the existing timer before reprogramming
	234	* it to the new values.
	235	*/
	236	for (;;) {
	237	spin_lock_irq(&ctx->wqh.lock);
	238	if (hrtimer_try_to_cancel(&ctx->tmr) >= 0)
	239	break;
18963c01	240	spin_unlock_irq(&ctx->wqh.lock);
4d672e7a	241	cpu_relax();
b215e283 DL	242	}
b215e283 DL	243
4d672e7a DL	244	/*
	245	* If the timer is expired and it's periodic, we need to advance it
	246	* because the caller may want to know the previous expiration time.
	247	* We do not update "ticks" and "expired" since the timer will be
	248	* re-programmed again in the following timerfd_setup() call.
	249	*/
	250	if (ctx->expired && ctx->tintv.tv64)
	251	hrtimer_forward_now(&ctx->tmr, ctx->tintv);
b215e283	252
4d672e7a DL	253	kotmr.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	254	kotmr.it_interval = ktime_to_timespec(ctx->tintv);
	255
	256	/*
	257	* Re-program the timer to the new value ...
	258	*/
	259	timerfd_setup(ctx, flags, &ktmr);
	260
	261	spin_unlock_irq(&ctx->wqh.lock);
	262	fput(file);
	263	if (otmr && copy_to_user(otmr, &kotmr, sizeof(kotmr)))
	264	return -EFAULT;
	265
	266	return 0;
	267	}
	268
d4e82042	269	SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr)
4d672e7a DL	270	{
	271	struct file *file;
	272	struct timerfd_ctx *ctx;
	273	struct itimerspec kotmr;
	274
	275	file = timerfd_fget(ufd);
	276	if (IS_ERR(file))
	277	return PTR_ERR(file);
	278	ctx = file->private_data;
	279
	280	spin_lock_irq(&ctx->wqh.lock);
	281	if (ctx->expired && ctx->tintv.tv64) {
	282	ctx->expired = 0;
	283	ctx->ticks +=
	284	hrtimer_forward_now(&ctx->tmr, ctx->tintv) - 1;
	285	hrtimer_restart(&ctx->tmr);
	286	}
	287	kotmr.it_value = ktime_to_timespec(timerfd_get_remaining(ctx));
	288	kotmr.it_interval = ktime_to_timespec(ctx->tintv);
	289	spin_unlock_irq(&ctx->wqh.lock);
	290	fput(file);
	291
	292	return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0;
b215e283 DL	293	}
b215e283 DL	294