[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/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? */
static struct per_user_data *port_user[NR_EVENT_CHANNELS];
static DEFINE_SPINLOCK(port_user_lock); /* protects port_user[] and ring_prod */

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 = port_user[port];

	disable_irq_nosync(irq);

	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;
	}

	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++)
		if ((kbuf[i] < NR_EVENT_CHANNELS) && (port_user[kbuf[i]] == u))
			enable_irq(irq_from_evtchn(kbuf[i]));
	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(port_user[port] != NULL);
	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();

	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 (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 (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 (port_user[i] != u)
			continue;

		evtchn_unbind_from_user(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;

	spin_lock_init(&port_user_lock);
	memset(port_user, 0, sizeof(port_user));

	/* 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)
{
	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>
f7116284 IC	41	#include <linux/miscdevice.h>
	42	#include <linux/major.h>
	43	#include <linux/proc_fs.h>
	44	#include <linux/stat.h>
	45	#include <linux/poll.h>
	46	#include <linux/irq.h>
	47	#include <linux/init.h>
	48	#include <linux/gfp.h>
	49	#include <linux/mutex.h>
	50	#include <linux/cpu.h>
	51	#include <xen/events.h>
	52	#include <xen/evtchn.h>
	53	#include <asm/xen/hypervisor.h>
	54
	55	struct per_user_data {
0a4666b5 JF	56	struct mutex bind_mutex; /* serialize bind/unbind operations */
0a4666b5 JF	57
f7116284 IC	58	/* Notification ring, accessed via /dev/xen/evtchn. */
	59	#define EVTCHN_RING_SIZE (PAGE_SIZE / sizeof(evtchn_port_t))
	60	#define EVTCHN_RING_MASK(_i) ((_i)&(EVTCHN_RING_SIZE-1))
	61	evtchn_port_t *ring;
	62	unsigned int ring_cons, ring_prod, ring_overflow;
	63	struct mutex ring_cons_mutex; /* protect against concurrent readers */
	64
	65	/* Processes wait on this queue when ring is empty. */
	66	wait_queue_head_t evtchn_wait;
	67	struct fasync_struct *evtchn_async_queue;
	68	const char *name;
	69	};
	70
	71	/* Who's bound to each port? */
	72	static struct per_user_data *port_user[NR_EVENT_CHANNELS];
0a4666b5	73	static DEFINE_SPINLOCK(port_user_lock); /* protects port_user[] and ring_prod */
f7116284 IC	74
	75	irqreturn_t evtchn_interrupt(int irq, void *data)
	76	{
	77	unsigned int port = (unsigned long)data;
	78	struct per_user_data *u;
	79
	80	spin_lock(&port_user_lock);
	81
	82	u = port_user[port];
	83
	84	disable_irq_nosync(irq);
	85
	86	if ((u->ring_prod - u->ring_cons) < EVTCHN_RING_SIZE) {
	87	u->ring[EVTCHN_RING_MASK(u->ring_prod)] = port;
	88	wmb(); /* Ensure ring contents visible */
	89	if (u->ring_cons == u->ring_prod++) {
	90	wake_up_interruptible(&u->evtchn_wait);
	91	kill_fasync(&u->evtchn_async_queue,
	92	SIGIO, POLL_IN);
	93	}
	94	} else {
	95	u->ring_overflow = 1;
	96	}
	97
	98	spin_unlock(&port_user_lock);
	99
	100	return IRQ_HANDLED;
	101	}
	102
	103	static ssize_t evtchn_read(struct file file, char __user buf,
	104	size_t count, loff_t *ppos)
	105	{
	106	int rc;
	107	unsigned int c, p, bytes1 = 0, bytes2 = 0;
	108	struct per_user_data *u = file->private_data;
	109
	110	/* Whole number of ports. */
	111	count &= ~(sizeof(evtchn_port_t)-1);
	112
	113	if (count == 0)
	114	return 0;
	115
	116	if (count > PAGE_SIZE)
	117	count = PAGE_SIZE;
	118
	119	for (;;) {
	120	mutex_lock(&u->ring_cons_mutex);
	121
	122	rc = -EFBIG;
	123	if (u->ring_overflow)
	124	goto unlock_out;
	125
	126	c = u->ring_cons;
	127	p = u->ring_prod;
	128	if (c != p)
	129	break;
	130
	131	mutex_unlock(&u->ring_cons_mutex);
	132
	133	if (file->f_flags & O_NONBLOCK)
	134	return -EAGAIN;
	135
	136	rc = wait_event_interruptible(u->evtchn_wait,
	137	u->ring_cons != u->ring_prod);
138	if (rc)
139	return rc;
140	}
141
142	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
143	if (((c ^ p) & EVTCHN_RING_SIZE) != 0) {
144	bytes1 = (EVTCHN_RING_SIZE - EVTCHN_RING_MASK(c)) *
145	sizeof(evtchn_port_t);
146	bytes2 = EVTCHN_RING_MASK(p) * sizeof(evtchn_port_t);
147	} else {
148	bytes1 = (p - c) * sizeof(evtchn_port_t);
149	bytes2 = 0;
150	}
151
152	/* Truncate chunks according to caller's maximum byte count. */
153	if (bytes1 > count) {
154	bytes1 = count;
155	bytes2 = 0;
156	} else if ((bytes1 + bytes2) > count) {
157	bytes2 = count - bytes1;
158	}
159
160	rc = -EFAULT;
161	rmb(); /* Ensure that we see the port before we copy it. */
162	if (copy_to_user(buf, &u->ring[EVTCHN_RING_MASK(c)], bytes1) \|\|
163	((bytes2 != 0) &&
164	copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
165	goto unlock_out;
166
167	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
168	rc = bytes1 + bytes2;
169
170	unlock_out:
171	mutex_unlock(&u->ring_cons_mutex);
172	return rc;
173	}
174
175	static ssize_t evtchn_write(struct file file, const char __user buf,
176	size_t count, loff_t *ppos)
177	{
178	int rc, i;
179	evtchn_port_t kbuf = (evtchn_port_t )__get_free_page(GFP_KERNEL);
180	struct per_user_data *u = file->private_data;
181
182	if (kbuf == NULL)
183	return -ENOMEM;
184
185	/* Whole number of ports. */
186	count &= ~(sizeof(evtchn_port_t)-1);
187
188	rc = 0;
189	if (count == 0)
190	goto out;
191
192	if (count > PAGE_SIZE)
193	count = PAGE_SIZE;
194
195	rc = -EFAULT;
196	if (copy_from_user(kbuf, buf, count) != 0)
197	goto out;
198
199	spin_lock_irq(&port_user_lock);
200	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++)
201	if ((kbuf[i] < NR_EVENT_CHANNELS) && (port_user[kbuf[i]] == u))
202	enable_irq(irq_from_evtchn(kbuf[i]));
203	spin_unlock_irq(&port_user_lock);
204
205	rc = count;
206
207	out:
208	free_page((unsigned long)kbuf);
209	return rc;
210	}
211
212	static int evtchn_bind_to_user(struct per_user_data *u, int port)
213	{
f7116284 IC	214	int rc = 0;
f7116284 IC	215
0a4666b5 JF	216	/*
	217	* Ports are never reused, so every caller should pass in a
	218	* unique port.
	219	*
	220	* (Locking not necessary because we haven't registered the
	221	* interrupt handler yet, and our caller has already
	222	* serialized bind operations.)
	223	*/
f7116284	224	BUG_ON(port_user[port] != NULL);
f7116284 IC	225	port_user[port] = u;
f7116284 IC	226
0a4666b5 JF	227	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, IRQF_DISABLED,
	228	u->name, (void *)(unsigned long)port);
	229	if (rc >= 0)
	230	rc = 0;
	231
f7116284 IC	232	return rc;
	233	}
	234
	235	static void evtchn_unbind_from_user(struct per_user_data *u, int port)
	236	{
	237	int irq = irq_from_evtchn(port);
	238
	239	unbind_from_irqhandler(irq, (void *)(unsigned long)port);
0a4666b5 JF	240
	241	/* make sure we unbind the irq handler before clearing the port */
	242	barrier();
	243
f7116284 IC	244	port_user[port] = NULL;
	245	}
	246
	247	static long evtchn_ioctl(struct file *file,
	248	unsigned int cmd, unsigned long arg)
	249	{
	250	int rc;
	251	struct per_user_data *u = file->private_data;
	252	void __user uarg = (void __user ) arg;
	253
0a4666b5 JF	254	/* Prevent bind from racing with unbind */
	255	mutex_lock(&u->bind_mutex);
	256
f7116284 IC	257	switch (cmd) {
	258	case IOCTL_EVTCHN_BIND_VIRQ: {
	259	struct ioctl_evtchn_bind_virq bind;
	260	struct evtchn_bind_virq bind_virq;
	261
	262	rc = -EFAULT;
	263	if (copy_from_user(&bind, uarg, sizeof(bind)))
	264	break;
	265
	266	bind_virq.virq = bind.virq;
	267	bind_virq.vcpu = 0;
	268	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
	269	&bind_virq);
	270	if (rc != 0)
	271	break;
	272
	273	rc = evtchn_bind_to_user(u, bind_virq.port);
	274	if (rc == 0)
	275	rc = bind_virq.port;
	276	break;
	277	}
	278
	279	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
	280	struct ioctl_evtchn_bind_interdomain bind;
	281	struct evtchn_bind_interdomain bind_interdomain;
	282
	283	rc = -EFAULT;
	284	if (copy_from_user(&bind, uarg, sizeof(bind)))
	285	break;
	286
	287	bind_interdomain.remote_dom = bind.remote_domain;
	288	bind_interdomain.remote_port = bind.remote_port;
	289	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
	290	&bind_interdomain);
	291	if (rc != 0)
	292	break;
	293
	294	rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
	295	if (rc == 0)
	296	rc = bind_interdomain.local_port;
	297	break;
	298	}
	299
	300	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
	301	struct ioctl_evtchn_bind_unbound_port bind;
	302	struct evtchn_alloc_unbound alloc_unbound;
	303
	304	rc = -EFAULT;
	305	if (copy_from_user(&bind, uarg, sizeof(bind)))
	306	break;
	307
	308	alloc_unbound.dom = DOMID_SELF;
	309	alloc_unbound.remote_dom = bind.remote_domain;
	310	rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
	311	&alloc_unbound);
	312	if (rc != 0)
	313	break;
	314
	315	rc = evtchn_bind_to_user(u, alloc_unbound.port);
	316	if (rc == 0)
	317	rc = alloc_unbound.port;
	318	break;
	319	}
	320
321	case IOCTL_EVTCHN_UNBIND: {
322	struct ioctl_evtchn_unbind unbind;
323
324	rc = -EFAULT;
325	if (copy_from_user(&unbind, uarg, sizeof(unbind)))
326	break;
327
328	rc = -EINVAL;
329	if (unbind.port >= NR_EVENT_CHANNELS)
330	break;
331
332	spin_lock_irq(&port_user_lock);
333
334	rc = -ENOTCONN;
335	if (port_user[unbind.port] != u) {
336	spin_unlock_irq(&port_user_lock);
337	break;
338	}
339
340	evtchn_unbind_from_user(u, unbind.port);
341
342	spin_unlock_irq(&port_user_lock);
343
344	rc = 0;
345	break;
346	}
347
348	case IOCTL_EVTCHN_NOTIFY: {
349	struct ioctl_evtchn_notify notify;
350
351	rc = -EFAULT;
352	if (copy_from_user(&notify, uarg, sizeof(notify)))
353	break;
354
355	if (notify.port >= NR_EVENT_CHANNELS) {
356	rc = -EINVAL;
357	} else if (port_user[notify.port] != u) {
358	rc = -ENOTCONN;
359	} else {
360	notify_remote_via_evtchn(notify.port);
361	rc = 0;
362	}
363	break;
364	}
365
366	case IOCTL_EVTCHN_RESET: {
367	/* Initialise the ring to empty. Clear errors. */
368	mutex_lock(&u->ring_cons_mutex);
369	spin_lock_irq(&port_user_lock);
370	u->ring_cons = u->ring_prod = u->ring_overflow = 0;
371	spin_unlock_irq(&port_user_lock);
372	mutex_unlock(&u->ring_cons_mutex);
373	rc = 0;
374	break;
375	}
376
377	default:
378	rc = -ENOSYS;
379	break;
380	}
0a4666b5	381	mutex_unlock(&u->bind_mutex);
f7116284 IC	382
	383	return rc;
	384	}
	385
	386	static unsigned int evtchn_poll(struct file file, poll_table wait)
	387	{
	388	unsigned int mask = POLLOUT \| POLLWRNORM;
	389	struct per_user_data *u = file->private_data;
	390
	391	poll_wait(file, &u->evtchn_wait, wait);
	392	if (u->ring_cons != u->ring_prod)
	393	mask \|= POLLIN \| POLLRDNORM;
	394	if (u->ring_overflow)
	395	mask = POLLERR;
	396	return mask;
	397	}
	398
	399	static int evtchn_fasync(int fd, struct file *filp, int on)
	400	{
	401	struct per_user_data *u = filp->private_data;
	402	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
	403	}
	404
	405	static int evtchn_open(struct inode inode, struct file filp)
	406	{
	407	struct per_user_data *u;
	408
	409	u = kzalloc(sizeof(*u), GFP_KERNEL);
	410	if (u == NULL)
	411	return -ENOMEM;
	412
	413	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
	414	if (u->name == NULL) {
	415	kfree(u);
	416	return -ENOMEM;
	417	}
	418
	419	init_waitqueue_head(&u->evtchn_wait);
	420
	421	u->ring = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
	422	if (u->ring == NULL) {
	423	kfree(u->name);
	424	kfree(u);
	425	return -ENOMEM;
	426	}
	427
0a4666b5	428	mutex_init(&u->bind_mutex);
f7116284 IC	429	mutex_init(&u->ring_cons_mutex);
	430
	431	filp->private_data = u;
	432
	433	return 0;
	434	}
	435
	436	static int evtchn_release(struct inode inode, struct file filp)
	437	{
	438	int i;
	439	struct per_user_data *u = filp->private_data;
	440
	441	spin_lock_irq(&port_user_lock);
	442
	443	free_page((unsigned long)u->ring);
	444
	445	for (i = 0; i < NR_EVENT_CHANNELS; i++) {
	446	if (port_user[i] != u)
	447	continue;
	448
	449	evtchn_unbind_from_user(port_user[i], i);
	450	}
	451
	452	spin_unlock_irq(&port_user_lock);
	453
	454	kfree(u->name);
	455	kfree(u);
	456
	457	return 0;
	458	}
	459
	460	static const struct file_operations evtchn_fops = {
	461	.owner = THIS_MODULE,
	462	.read = evtchn_read,
	463	.write = evtchn_write,
	464	.unlocked_ioctl = evtchn_ioctl,
	465	.poll = evtchn_poll,
	466	.fasync = evtchn_fasync,
	467	.open = evtchn_open,
	468	.release = evtchn_release,
	469	};
	470
	471	static struct miscdevice evtchn_miscdev = {
	472	.minor = MISC_DYNAMIC_MINOR,
	473	.name = "evtchn",
	474	.fops = &evtchn_fops,
	475	};
	476	static int __init evtchn_init(void)
	477	{
	478	int err;
	479
	480	if (!xen_domain())
	481	return -ENODEV;
	482
	483	spin_lock_init(&port_user_lock);
	484	memset(port_user, 0, sizeof(port_user));
	485
	486	/* Create '/dev/misc/evtchn'. */
	487	err = misc_register(&evtchn_miscdev);
	488	if (err != 0) {
	489	printk(KERN_ALERT "Could not register /dev/misc/evtchn\n");
	490	return err;
	491	}
	492
493	printk(KERN_INFO "Event-channel device installed.\n");
494
495	return 0;
496	}
497
498	static void __exit evtchn_cleanup(void)
499	{
500	misc_deregister(&evtchn_miscdev);
501	}
502
503	module_init(evtchn_init);
504	module_exit(evtchn_cleanup);
505
506	MODULE_LICENSE("GPL");