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