[deliverable/linux.git] / drivers / xen / evtchn.c

/******************************************************************************
 * evtchn.c
 *
 * Driver for receiving and demuxing event-channel signals.
 *
 * Copyright (c) 2004-2005, K A Fraser
 * Multi-process extensions Copyright (c) 2004, Steven Smith
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License version 2
 * as published by the Free Software Foundation; or, when distributed
 * separately from the Linux kernel or incorporated into other
 * software packages, subject to the following license:
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this source file (the "Software"), to deal in the Software without
 * restriction, including without limitation the rights to use, copy, modify,
 * merge, publish, distribute, sublicense, and/or sell copies of the Software,
 * and to permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 */

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/fs.h>
#include <linux/errno.h>
#include <linux/miscdevice.h>
#include <linux/major.h>
#include <linux/proc_fs.h>
#include <linux/stat.h>
#include <linux/poll.h>
#include <linux/irq.h>
#include <linux/init.h>
#include <linux/gfp.h>
#include <linux/mutex.h>
#include <linux/cpu.h>
#include <xen/events.h>
#include <xen/evtchn.h>
#include <asm/xen/hypervisor.h>

struct per_user_data {
	struct mutex bind_mutex; /* serialize bind/unbind operations */

	/* Notification ring, accessed via /dev/xen/evtchn. */
#define EVTCHN_RING_SIZE     (PAGE_SIZE / sizeof(evtchn_port_t))
#define EVTCHN_RING_MASK(_i) ((_i)&(EVTCHN_RING_SIZE-1))
	evtchn_port_t *ring;
	unsigned int ring_cons, ring_prod, ring_overflow;
	struct mutex ring_cons_mutex; /* protect against concurrent readers */

	/* Processes wait on this queue when ring is empty. */
	wait_queue_head_t evtchn_wait;
	struct fasync_struct *evtchn_async_queue;
	const char *name;
};

/*
 * Who's bound to each port?  This is logically an array of struct
 * per_user_data *, but we encode the current enabled-state in bit 0.
 */
static unsigned long *port_user;
static DEFINE_SPINLOCK(port_user_lock); /* protects port_user[] and ring_prod */

static inline struct per_user_data *get_port_user(unsigned port)
{
	return (struct per_user_data *)(port_user[port] & ~1);
}

static inline void set_port_user(unsigned port, struct per_user_data *u)
{
	port_user[port] = (unsigned long)u;
}

static inline bool get_port_enabled(unsigned port)
{
	return port_user[port] & 1;
}

static inline void set_port_enabled(unsigned port, bool enabled)
{
	if (enabled)
		port_user[port] |= 1;
	else
		port_user[port] &= ~1;
}

irqreturn_t evtchn_interrupt(int irq, void *data)
{
	unsigned int port = (unsigned long)data;
	struct per_user_data *u;

	spin_lock(&port_user_lock);

	u = get_port_user(port);

	if (WARN(!get_port_enabled(port),
		 "Interrupt for port %d, but apparently not enabled; per-user %p\n",
		 port, u))
		goto out;

	disable_irq_nosync(irq);
	set_port_enabled(port, false);

	if ((u->ring_prod - u->ring_cons) < EVTCHN_RING_SIZE) {
		u->ring[EVTCHN_RING_MASK(u->ring_prod)] = port;
		wmb(); /* Ensure ring contents visible */
		if (u->ring_cons == u->ring_prod++) {
			wake_up_interruptible(&u->evtchn_wait);
			kill_fasync(&u->evtchn_async_queue,
				    SIGIO, POLL_IN);
		}
	} else
		u->ring_overflow = 1;

out:
	spin_unlock(&port_user_lock);

	return IRQ_HANDLED;
}

static ssize_t evtchn_read(struct file *file, char __user *buf,
			   size_t count, loff_t *ppos)
{
	int rc;
	unsigned int c, p, bytes1 = 0, bytes2 = 0;
	struct per_user_data *u = file->private_data;

	/* Whole number of ports. */
	count &= ~(sizeof(evtchn_port_t)-1);

	if (count == 0)
		return 0;

	if (count > PAGE_SIZE)
		count = PAGE_SIZE;

	for (;;) {
		mutex_lock(&u->ring_cons_mutex);

		rc = -EFBIG;
		if (u->ring_overflow)
			goto unlock_out;

		c = u->ring_cons;
		p = u->ring_prod;
		if (c != p)
			break;

		mutex_unlock(&u->ring_cons_mutex);

		if (file->f_flags & O_NONBLOCK)
			return -EAGAIN;

		rc = wait_event_interruptible(u->evtchn_wait,
					      u->ring_cons != u->ring_prod);
		if (rc)
			return rc;
	}

	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
	if (((c ^ p) & EVTCHN_RING_SIZE) != 0) {
		bytes1 = (EVTCHN_RING_SIZE - EVTCHN_RING_MASK(c)) *
			sizeof(evtchn_port_t);
		bytes2 = EVTCHN_RING_MASK(p) * sizeof(evtchn_port_t);
	} else {
		bytes1 = (p - c) * sizeof(evtchn_port_t);
		bytes2 = 0;
	}

	/* Truncate chunks according to caller's maximum byte count. */
	if (bytes1 > count) {
		bytes1 = count;
		bytes2 = 0;
	} else if ((bytes1 + bytes2) > count) {
		bytes2 = count - bytes1;
	}

	rc = -EFAULT;
	rmb(); /* Ensure that we see the port before we copy it. */
	if (copy_to_user(buf, &u->ring[EVTCHN_RING_MASK(c)], bytes1) ||
	    ((bytes2 != 0) &&
	     copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
		goto unlock_out;

	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
	rc = bytes1 + bytes2;

 unlock_out:
	mutex_unlock(&u->ring_cons_mutex);
	return rc;
}

static ssize_t evtchn_write(struct file *file, const char __user *buf,
			    size_t count, loff_t *ppos)
{
	int rc, i;
	evtchn_port_t *kbuf = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
	struct per_user_data *u = file->private_data;

	if (kbuf == NULL)
		return -ENOMEM;

	/* Whole number of ports. */
	count &= ~(sizeof(evtchn_port_t)-1);

	rc = 0;
	if (count == 0)
		goto out;

	if (count > PAGE_SIZE)
		count = PAGE_SIZE;

	rc = -EFAULT;
	if (copy_from_user(kbuf, buf, count) != 0)
		goto out;

	spin_lock_irq(&port_user_lock);

	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++) {
		unsigned port = kbuf[i];

		if (port < NR_EVENT_CHANNELS &&
		    get_port_user(port) == u &&
		    !get_port_enabled(port)) {
			set_port_enabled(port, true);
			enable_irq(irq_from_evtchn(port));
		}
	}

	spin_unlock_irq(&port_user_lock);

	rc = count;

 out:
	free_page((unsigned long)kbuf);
	return rc;
}

static int evtchn_bind_to_user(struct per_user_data *u, int port)
{
	int rc = 0;

	/*
	 * Ports are never reused, so every caller should pass in a
	 * unique port.
	 *
	 * (Locking not necessary because we haven't registered the
	 * interrupt handler yet, and our caller has already
	 * serialized bind operations.)
	 */
	BUG_ON(get_port_user(port) != NULL);
	set_port_user(port, u);

	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, IRQF_DISABLED,
				       u->name, (void *)(unsigned long)port);
	if (rc >= 0)
		rc = 0;

	return rc;
}

static void evtchn_unbind_from_user(struct per_user_data *u, int port)
{
	int irq = irq_from_evtchn(port);

	unbind_from_irqhandler(irq, (void *)(unsigned long)port);

	/* make sure we unbind the irq handler before clearing the port */
	barrier();

	set_port_user(port, NULL);
}

static long evtchn_ioctl(struct file *file,
			 unsigned int cmd, unsigned long arg)
{
	int rc;
	struct per_user_data *u = file->private_data;
	void __user *uarg = (void __user *) arg;

	/* Prevent bind from racing with unbind */
	mutex_lock(&u->bind_mutex);

	switch (cmd) {
	case IOCTL_EVTCHN_BIND_VIRQ: {
		struct ioctl_evtchn_bind_virq bind;
		struct evtchn_bind_virq bind_virq;

		rc = -EFAULT;
		if (copy_from_user(&bind, uarg, sizeof(bind)))
			break;

		bind_virq.virq = bind.virq;
		bind_virq.vcpu = 0;
		rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
						 &bind_virq);
		if (rc != 0)
			break;

		rc = evtchn_bind_to_user(u, bind_virq.port);
		if (rc == 0)
			rc = bind_virq.port;
		break;
	}

	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
		struct ioctl_evtchn_bind_interdomain bind;
		struct evtchn_bind_interdomain bind_interdomain;

		rc = -EFAULT;
		if (copy_from_user(&bind, uarg, sizeof(bind)))
			break;

		bind_interdomain.remote_dom  = bind.remote_domain;
		bind_interdomain.remote_port = bind.remote_port;
		rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
						 &bind_interdomain);
		if (rc != 0)
			break;

		rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
		if (rc == 0)
			rc = bind_interdomain.local_port;
		break;
	}

	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
		struct ioctl_evtchn_bind_unbound_port bind;
		struct evtchn_alloc_unbound alloc_unbound;

		rc = -EFAULT;
		if (copy_from_user(&bind, uarg, sizeof(bind)))
			break;

		alloc_unbound.dom        = DOMID_SELF;
		alloc_unbound.remote_dom = bind.remote_domain;
		rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
						 &alloc_unbound);
		if (rc != 0)
			break;

		rc = evtchn_bind_to_user(u, alloc_unbound.port);
		if (rc == 0)
			rc = alloc_unbound.port;
		break;
	}

	case IOCTL_EVTCHN_UNBIND: {
		struct ioctl_evtchn_unbind unbind;

		rc = -EFAULT;
		if (copy_from_user(&unbind, uarg, sizeof(unbind)))
			break;

		rc = -EINVAL;
		if (unbind.port >= NR_EVENT_CHANNELS)
			break;

		spin_lock_irq(&port_user_lock);

		rc = -ENOTCONN;
		if (get_port_user(unbind.port) != u) {
			spin_unlock_irq(&port_user_lock);
			break;
		}

		evtchn_unbind_from_user(u, unbind.port);

		spin_unlock_irq(&port_user_lock);

		rc = 0;
		break;
	}

	case IOCTL_EVTCHN_NOTIFY: {
		struct ioctl_evtchn_notify notify;

		rc = -EFAULT;
		if (copy_from_user(&notify, uarg, sizeof(notify)))
			break;

		if (notify.port >= NR_EVENT_CHANNELS) {
			rc = -EINVAL;
		} else if (get_port_user(notify.port) != u) {
			rc = -ENOTCONN;
		} else {
			notify_remote_via_evtchn(notify.port);
			rc = 0;
		}
		break;
	}

	case IOCTL_EVTCHN_RESET: {
		/* Initialise the ring to empty. Clear errors. */
		mutex_lock(&u->ring_cons_mutex);
		spin_lock_irq(&port_user_lock);
		u->ring_cons = u->ring_prod = u->ring_overflow = 0;
		spin_unlock_irq(&port_user_lock);
		mutex_unlock(&u->ring_cons_mutex);
		rc = 0;
		break;
	}

	default:
		rc = -ENOSYS;
		break;
	}
	mutex_unlock(&u->bind_mutex);

	return rc;
}

static unsigned int evtchn_poll(struct file *file, poll_table *wait)
{
	unsigned int mask = POLLOUT | POLLWRNORM;
	struct per_user_data *u = file->private_data;

	poll_wait(file, &u->evtchn_wait, wait);
	if (u->ring_cons != u->ring_prod)
		mask |= POLLIN | POLLRDNORM;
	if (u->ring_overflow)
		mask = POLLERR;
	return mask;
}

static int evtchn_fasync(int fd, struct file *filp, int on)
{
	struct per_user_data *u = filp->private_data;
	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
}

static int evtchn_open(struct inode *inode, struct file *filp)
{
	struct per_user_data *u;

	u = kzalloc(sizeof(*u), GFP_KERNEL);
	if (u == NULL)
		return -ENOMEM;

	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
	if (u->name == NULL) {
		kfree(u);
		return -ENOMEM;
	}

	init_waitqueue_head(&u->evtchn_wait);

	u->ring = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
	if (u->ring == NULL) {
		kfree(u->name);
		kfree(u);
		return -ENOMEM;
	}

	mutex_init(&u->bind_mutex);
	mutex_init(&u->ring_cons_mutex);

	filp->private_data = u;

	return 0;
}

static int evtchn_release(struct inode *inode, struct file *filp)
{
	int i;
	struct per_user_data *u = filp->private_data;

	spin_lock_irq(&port_user_lock);

	free_page((unsigned long)u->ring);

	for (i = 0; i < NR_EVENT_CHANNELS; i++) {
		if (get_port_user(i) != u)
			continue;

		evtchn_unbind_from_user(get_port_user(i), i);
	}

	spin_unlock_irq(&port_user_lock);

	kfree(u->name);
	kfree(u);

	return 0;
}

static const struct file_operations evtchn_fops = {
	.owner   = THIS_MODULE,
	.read    = evtchn_read,
	.write   = evtchn_write,
	.unlocked_ioctl = evtchn_ioctl,
	.poll    = evtchn_poll,
	.fasync  = evtchn_fasync,
	.open    = evtchn_open,
	.release = evtchn_release,
};

static struct miscdevice evtchn_miscdev = {
	.minor        = MISC_DYNAMIC_MINOR,
	.name         = "evtchn",
	.fops         = &evtchn_fops,
};
static int __init evtchn_init(void)
{
	int err;

	if (!xen_domain())
		return -ENODEV;

	port_user = kcalloc(NR_EVENT_CHANNELS, sizeof(*port_user), GFP_KERNEL);
	if (port_user == NULL)
		return -ENOMEM;

	spin_lock_init(&port_user_lock);

	/* Create '/dev/misc/evtchn'. */
	err = misc_register(&evtchn_miscdev);
	if (err != 0) {
		printk(KERN_ALERT "Could not register /dev/misc/evtchn\n");
		return err;
	}

	printk(KERN_INFO "Event-channel device installed.\n");

	return 0;
}

static void __exit evtchn_cleanup(void)
{
	kfree(port_user);
	port_user = NULL;

	misc_deregister(&evtchn_miscdev);
}

module_init(evtchn_init);
module_exit(evtchn_cleanup);

MODULE_LICENSE("GPL");
Commit	Line	Data
f7116284 IC	1	/******************************************************************************
	2	* evtchn.c
	3	*
	4	* Driver for receiving and demuxing event-channel signals.
	5	*
	6	* Copyright (c) 2004-2005, K A Fraser
	7	* Multi-process extensions Copyright (c) 2004, Steven Smith
	8	*
	9	* This program is free software; you can redistribute it and/or
	10	* modify it under the terms of the GNU General Public License version 2
	11	* as published by the Free Software Foundation; or, when distributed
	12	* separately from the Linux kernel or incorporated into other
	13	* software packages, subject to the following license:
	14	*
	15	* Permission is hereby granted, free of charge, to any person obtaining a copy
	16	* of this source file (the "Software"), to deal in the Software without
	17	* restriction, including without limitation the rights to use, copy, modify,
	18	* merge, publish, distribute, sublicense, and/or sell copies of the Software,
	19	* and to permit persons to whom the Software is furnished to do so, subject to
	20	* the following conditions:
	21	*
	22	* The above copyright notice and this permission notice shall be included in
	23	* all copies or substantial portions of the Software.
	24	*
	25	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	26	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	27	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
	28	* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
	29	* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
	30	* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
	31	* IN THE SOFTWARE.
	32	*/
	33
	34	#include <linux/module.h>
	35	#include <linux/kernel.h>
	36	#include <linux/sched.h>
	37	#include <linux/slab.h>
	38	#include <linux/string.h>
	39	#include <linux/errno.h>
	40	#include <linux/fs.h>
	41	#include <linux/errno.h>
	42	#include <linux/miscdevice.h>
	43	#include <linux/major.h>
	44	#include <linux/proc_fs.h>
	45	#include <linux/stat.h>
	46	#include <linux/poll.h>
	47	#include <linux/irq.h>
	48	#include <linux/init.h>
	49	#include <linux/gfp.h>
	50	#include <linux/mutex.h>
	51	#include <linux/cpu.h>
	52	#include <xen/events.h>
	53	#include <xen/evtchn.h>
	54	#include <asm/xen/hypervisor.h>
	55
	56	struct per_user_data {
0a4666b5 JF	57	struct mutex bind_mutex; /* serialize bind/unbind operations */
0a4666b5 JF	58
f7116284 IC	59	/* Notification ring, accessed via /dev/xen/evtchn. */
	60	#define EVTCHN_RING_SIZE (PAGE_SIZE / sizeof(evtchn_port_t))
	61	#define EVTCHN_RING_MASK(_i) ((_i)&(EVTCHN_RING_SIZE-1))
	62	evtchn_port_t *ring;
	63	unsigned int ring_cons, ring_prod, ring_overflow;
	64	struct mutex ring_cons_mutex; /* protect against concurrent readers */
	65
	66	/* Processes wait on this queue when ring is empty. */
	67	wait_queue_head_t evtchn_wait;
	68	struct fasync_struct *evtchn_async_queue;
	69	const char *name;
	70	};
	71
e3cc067b JF	72	/*
	73	* Who's bound to each port? This is logically an array of struct
	74	* per_user_data *, but we encode the current enabled-state in bit 0.
	75	*/
93afe0b7	76	static unsigned long *port_user;
0a4666b5	77	static DEFINE_SPINLOCK(port_user_lock); /* protects port_user[] and ring_prod */
f7116284	78
e3cc067b JF	79	static inline struct per_user_data *get_port_user(unsigned port)
	80	{
	81	return (struct per_user_data *)(port_user[port] & ~1);
	82	}
	83
	84	static inline void set_port_user(unsigned port, struct per_user_data *u)
	85	{
	86	port_user[port] = (unsigned long)u;
	87	}
	88
	89	static inline bool get_port_enabled(unsigned port)
	90	{
	91	return port_user[port] & 1;
	92	}
	93
	94	static inline void set_port_enabled(unsigned port, bool enabled)
	95	{
	96	if (enabled)
	97	port_user[port] \|= 1;
	98	else
	99	port_user[port] &= ~1;
	100	}
	101
f7116284 IC	102	irqreturn_t evtchn_interrupt(int irq, void *data)
	103	{
	104	unsigned int port = (unsigned long)data;
	105	struct per_user_data *u;
	106
	107	spin_lock(&port_user_lock);
	108
e3cc067b JF	109	u = get_port_user(port);
	110
	111	if (WARN(!get_port_enabled(port),
	112	"Interrupt for port %d, but apparently not enabled; per-user %p\n",
	113	port, u))
	114	goto out;
f7116284 IC	115
f7116284 IC	116	disable_irq_nosync(irq);
e3cc067b	117	set_port_enabled(port, false);
f7116284 IC	118
	119	if ((u->ring_prod - u->ring_cons) < EVTCHN_RING_SIZE) {
	120	u->ring[EVTCHN_RING_MASK(u->ring_prod)] = port;
	121	wmb(); /* Ensure ring contents visible */
	122	if (u->ring_cons == u->ring_prod++) {
	123	wake_up_interruptible(&u->evtchn_wait);
	124	kill_fasync(&u->evtchn_async_queue,
	125	SIGIO, POLL_IN);
	126	}
e3cc067b	127	} else
f7116284	128	u->ring_overflow = 1;
f7116284	129
e3cc067b	130	out:
f7116284 IC	131	spin_unlock(&port_user_lock);
	132
	133	return IRQ_HANDLED;
	134	}
	135
	136	static ssize_t evtchn_read(struct file file, char __user buf,
	137	size_t count, loff_t *ppos)
	138	{
	139	int rc;
	140	unsigned int c, p, bytes1 = 0, bytes2 = 0;
	141	struct per_user_data *u = file->private_data;
	142
	143	/* Whole number of ports. */
	144	count &= ~(sizeof(evtchn_port_t)-1);
	145
	146	if (count == 0)
	147	return 0;
	148
	149	if (count > PAGE_SIZE)
	150	count = PAGE_SIZE;
	151
	152	for (;;) {
	153	mutex_lock(&u->ring_cons_mutex);
	154
	155	rc = -EFBIG;
	156	if (u->ring_overflow)
	157	goto unlock_out;
	158
	159	c = u->ring_cons;
	160	p = u->ring_prod;
	161	if (c != p)
	162	break;
	163
	164	mutex_unlock(&u->ring_cons_mutex);
	165
	166	if (file->f_flags & O_NONBLOCK)
	167	return -EAGAIN;
	168
	169	rc = wait_event_interruptible(u->evtchn_wait,
	170	u->ring_cons != u->ring_prod);
	171	if (rc)
	172	return rc;
	173	}
	174
	175	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
	176	if (((c ^ p) & EVTCHN_RING_SIZE) != 0) {
	177	bytes1 = (EVTCHN_RING_SIZE - EVTCHN_RING_MASK(c)) *
	178	sizeof(evtchn_port_t);
	179	bytes2 = EVTCHN_RING_MASK(p) * sizeof(evtchn_port_t);
	180	} else {
	181	bytes1 = (p - c) * sizeof(evtchn_port_t);
	182	bytes2 = 0;
	183	}
	184
	185	/* Truncate chunks according to caller's maximum byte count. */
	186	if (bytes1 > count) {
	187	bytes1 = count;
	188	bytes2 = 0;
	189	} else if ((bytes1 + bytes2) > count) {
	190	bytes2 = count - bytes1;
	191	}
	192
	193	rc = -EFAULT;
	194	rmb(); /* Ensure that we see the port before we copy it. */
195	if (copy_to_user(buf, &u->ring[EVTCHN_RING_MASK(c)], bytes1) \|\|
196	((bytes2 != 0) &&
197	copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
198	goto unlock_out;
199
200	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
201	rc = bytes1 + bytes2;
202
203	unlock_out:
204	mutex_unlock(&u->ring_cons_mutex);
205	return rc;
206	}
207
208	static ssize_t evtchn_write(struct file file, const char __user buf,
209	size_t count, loff_t *ppos)
210	{
211	int rc, i;
212	evtchn_port_t kbuf = (evtchn_port_t )__get_free_page(GFP_KERNEL);
213	struct per_user_data *u = file->private_data;
214
215	if (kbuf == NULL)
216	return -ENOMEM;
217
218	/* Whole number of ports. */
219	count &= ~(sizeof(evtchn_port_t)-1);
220
221	rc = 0;
222	if (count == 0)
223	goto out;
224
225	if (count > PAGE_SIZE)
226	count = PAGE_SIZE;
227
228	rc = -EFAULT;
229	if (copy_from_user(kbuf, buf, count) != 0)
230	goto out;
231
232	spin_lock_irq(&port_user_lock);
e3cc067b JF	233
	234	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++) {
	235	unsigned port = kbuf[i];
	236
	237	if (port < NR_EVENT_CHANNELS &&
	238	get_port_user(port) == u &&
	239	!get_port_enabled(port)) {
	240	set_port_enabled(port, true);
	241	enable_irq(irq_from_evtchn(port));
	242	}
	243	}
	244
f7116284 IC	245	spin_unlock_irq(&port_user_lock);
	246
	247	rc = count;
	248
	249	out:
	250	free_page((unsigned long)kbuf);
	251	return rc;
	252	}
	253
	254	static int evtchn_bind_to_user(struct per_user_data *u, int port)
	255	{
f7116284 IC	256	int rc = 0;
f7116284 IC	257
0a4666b5 JF	258	/*
	259	* Ports are never reused, so every caller should pass in a
	260	* unique port.
	261	*
	262	* (Locking not necessary because we haven't registered the
	263	* interrupt handler yet, and our caller has already
	264	* serialized bind operations.)
	265	*/
e3cc067b JF	266	BUG_ON(get_port_user(port) != NULL);
e3cc067b JF	267	set_port_user(port, u);
f7116284	268
0a4666b5 JF	269	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, IRQF_DISABLED,
	270	u->name, (void *)(unsigned long)port);
	271	if (rc >= 0)
	272	rc = 0;
	273
f7116284 IC	274	return rc;
	275	}
	276
	277	static void evtchn_unbind_from_user(struct per_user_data *u, int port)
	278	{
	279	int irq = irq_from_evtchn(port);
	280
	281	unbind_from_irqhandler(irq, (void *)(unsigned long)port);
0a4666b5 JF	282
	283	/* make sure we unbind the irq handler before clearing the port */
	284	barrier();
	285
e3cc067b	286	set_port_user(port, NULL);
f7116284 IC	287	}
	288
	289	static long evtchn_ioctl(struct file *file,
	290	unsigned int cmd, unsigned long arg)
	291	{
	292	int rc;
	293	struct per_user_data *u = file->private_data;
	294	void __user uarg = (void __user ) arg;
	295
0a4666b5 JF	296	/* Prevent bind from racing with unbind */
	297	mutex_lock(&u->bind_mutex);
	298
f7116284 IC	299	switch (cmd) {
	300	case IOCTL_EVTCHN_BIND_VIRQ: {
	301	struct ioctl_evtchn_bind_virq bind;
	302	struct evtchn_bind_virq bind_virq;
	303
	304	rc = -EFAULT;
	305	if (copy_from_user(&bind, uarg, sizeof(bind)))
	306	break;
	307
	308	bind_virq.virq = bind.virq;
	309	bind_virq.vcpu = 0;
	310	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
	311	&bind_virq);
	312	if (rc != 0)
	313	break;
	314
	315	rc = evtchn_bind_to_user(u, bind_virq.port);
	316	if (rc == 0)
	317	rc = bind_virq.port;
	318	break;
	319	}
	320
	321	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
	322	struct ioctl_evtchn_bind_interdomain bind;
	323	struct evtchn_bind_interdomain bind_interdomain;
	324
	325	rc = -EFAULT;
	326	if (copy_from_user(&bind, uarg, sizeof(bind)))
	327	break;
	328
	329	bind_interdomain.remote_dom = bind.remote_domain;
	330	bind_interdomain.remote_port = bind.remote_port;
	331	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
	332	&bind_interdomain);
	333	if (rc != 0)
	334	break;
	335
	336	rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
	337	if (rc == 0)
	338	rc = bind_interdomain.local_port;
	339	break;
	340	}
	341
	342	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
	343	struct ioctl_evtchn_bind_unbound_port bind;
	344	struct evtchn_alloc_unbound alloc_unbound;
	345
	346	rc = -EFAULT;
	347	if (copy_from_user(&bind, uarg, sizeof(bind)))
	348	break;
	349
	350	alloc_unbound.dom = DOMID_SELF;
	351	alloc_unbound.remote_dom = bind.remote_domain;
	352	rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
	353	&alloc_unbound);
	354	if (rc != 0)
	355	break;
	356
	357	rc = evtchn_bind_to_user(u, alloc_unbound.port);
	358	if (rc == 0)
	359	rc = alloc_unbound.port;
	360	break;
	361	}
	362
363	case IOCTL_EVTCHN_UNBIND: {
364	struct ioctl_evtchn_unbind unbind;
365
366	rc = -EFAULT;
367	if (copy_from_user(&unbind, uarg, sizeof(unbind)))
368	break;
369
370	rc = -EINVAL;
371	if (unbind.port >= NR_EVENT_CHANNELS)
372	break;
373
374	spin_lock_irq(&port_user_lock);
375
376	rc = -ENOTCONN;
e3cc067b	377	if (get_port_user(unbind.port) != u) {
f7116284 IC	378	spin_unlock_irq(&port_user_lock);
	379	break;
	380	}
	381
	382	evtchn_unbind_from_user(u, unbind.port);
	383
	384	spin_unlock_irq(&port_user_lock);
	385
	386	rc = 0;
	387	break;
	388	}
	389
	390	case IOCTL_EVTCHN_NOTIFY: {
	391	struct ioctl_evtchn_notify notify;
	392
	393	rc = -EFAULT;
	394	if (copy_from_user(&notify, uarg, sizeof(notify)))
	395	break;
	396
	397	if (notify.port >= NR_EVENT_CHANNELS) {
	398	rc = -EINVAL;
e3cc067b	399	} else if (get_port_user(notify.port) != u) {
f7116284 IC	400	rc = -ENOTCONN;
	401	} else {
	402	notify_remote_via_evtchn(notify.port);
	403	rc = 0;
	404	}
	405	break;
	406	}
	407
	408	case IOCTL_EVTCHN_RESET: {
	409	/* Initialise the ring to empty. Clear errors. */
	410	mutex_lock(&u->ring_cons_mutex);
	411	spin_lock_irq(&port_user_lock);
	412	u->ring_cons = u->ring_prod = u->ring_overflow = 0;
	413	spin_unlock_irq(&port_user_lock);
	414	mutex_unlock(&u->ring_cons_mutex);
	415	rc = 0;
	416	break;
	417	}
	418
	419	default:
	420	rc = -ENOSYS;
	421	break;
	422	}
0a4666b5	423	mutex_unlock(&u->bind_mutex);
f7116284 IC	424
	425	return rc;
	426	}
	427
	428	static unsigned int evtchn_poll(struct file file, poll_table wait)
	429	{
	430	unsigned int mask = POLLOUT \| POLLWRNORM;
	431	struct per_user_data *u = file->private_data;
	432
	433	poll_wait(file, &u->evtchn_wait, wait);
	434	if (u->ring_cons != u->ring_prod)
	435	mask \|= POLLIN \| POLLRDNORM;
	436	if (u->ring_overflow)
	437	mask = POLLERR;
	438	return mask;
	439	}
	440
	441	static int evtchn_fasync(int fd, struct file *filp, int on)
	442	{
	443	struct per_user_data *u = filp->private_data;
	444	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
	445	}
	446
	447	static int evtchn_open(struct inode inode, struct file filp)
	448	{
	449	struct per_user_data *u;
	450
	451	u = kzalloc(sizeof(*u), GFP_KERNEL);
	452	if (u == NULL)
	453	return -ENOMEM;
	454
	455	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
	456	if (u->name == NULL) {
	457	kfree(u);
	458	return -ENOMEM;
	459	}
	460
	461	init_waitqueue_head(&u->evtchn_wait);
	462
	463	u->ring = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
	464	if (u->ring == NULL) {
	465	kfree(u->name);
	466	kfree(u);
	467	return -ENOMEM;
	468	}
	469
0a4666b5	470	mutex_init(&u->bind_mutex);
f7116284 IC	471	mutex_init(&u->ring_cons_mutex);
	472
	473	filp->private_data = u;
	474
	475	return 0;
	476	}
	477
	478	static int evtchn_release(struct inode inode, struct file filp)
	479	{
	480	int i;
	481	struct per_user_data *u = filp->private_data;
	482
	483	spin_lock_irq(&port_user_lock);
	484
	485	free_page((unsigned long)u->ring);
	486
	487	for (i = 0; i < NR_EVENT_CHANNELS; i++) {
e3cc067b	488	if (get_port_user(i) != u)
f7116284 IC	489	continue;
f7116284 IC	490
e3cc067b	491	evtchn_unbind_from_user(get_port_user(i), i);
f7116284 IC	492	}
	493
	494	spin_unlock_irq(&port_user_lock);
	495
	496	kfree(u->name);
	497	kfree(u);
	498
	499	return 0;
	500	}
	501
	502	static const struct file_operations evtchn_fops = {
	503	.owner = THIS_MODULE,
	504	.read = evtchn_read,
	505	.write = evtchn_write,
	506	.unlocked_ioctl = evtchn_ioctl,
	507	.poll = evtchn_poll,
	508	.fasync = evtchn_fasync,
	509	.open = evtchn_open,
	510	.release = evtchn_release,
	511	};
	512
	513	static struct miscdevice evtchn_miscdev = {
	514	.minor = MISC_DYNAMIC_MINOR,
	515	.name = "evtchn",
	516	.fops = &evtchn_fops,
	517	};
	518	static int __init evtchn_init(void)
	519	{
	520	int err;
	521
	522	if (!xen_domain())
	523	return -ENODEV;
	524
93afe0b7 JF	525	port_user = kcalloc(NR_EVENT_CHANNELS, sizeof(*port_user), GFP_KERNEL);
	526	if (port_user == NULL)
	527	return -ENOMEM;
	528
f7116284	529	spin_lock_init(&port_user_lock);
f7116284 IC	530
	531	/* Create '/dev/misc/evtchn'. */
	532	err = misc_register(&evtchn_miscdev);
	533	if (err != 0) {
	534	printk(KERN_ALERT "Could not register /dev/misc/evtchn\n");
	535	return err;
	536	}
	537
	538	printk(KERN_INFO "Event-channel device installed.\n");
	539
	540	return 0;
	541	}
	542
	543	static void __exit evtchn_cleanup(void)
	544	{
93afe0b7 JF	545	kfree(port_user);
	546	port_user = NULL;
	547
f7116284 IC	548	misc_deregister(&evtchn_miscdev);
	549	}
	550
	551	module_init(evtchn_init);
	552	module_exit(evtchn_cleanup);
	553
	554	MODULE_LICENSE("GPL");