[net-next-2.6.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/mutex.h>
#include <linux/cpu.h>

#include <xen/xen.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,
	.llseek = noop_llseek,
};

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>
f7116284 IC	48	#include <linux/mutex.h>
f7116284 IC	49	#include <linux/cpu.h>
1ccbf534 JF	50
1ccbf534 JF	51	#include <xen/xen.h>
f7116284 IC	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
	72	/* Who's bound to each port? */
	73	static struct per_user_data *port_user[NR_EVENT_CHANNELS];
0a4666b5	74	static DEFINE_SPINLOCK(port_user_lock); /* protects port_user[] and ring_prod */
f7116284 IC	75
	76	irqreturn_t evtchn_interrupt(int irq, void *data)
	77	{
	78	unsigned int port = (unsigned long)data;
	79	struct per_user_data *u;
	80
	81	spin_lock(&port_user_lock);
	82
	83	u = port_user[port];
	84
	85	disable_irq_nosync(irq);
	86
	87	if ((u->ring_prod - u->ring_cons) < EVTCHN_RING_SIZE) {
	88	u->ring[EVTCHN_RING_MASK(u->ring_prod)] = port;
	89	wmb(); /* Ensure ring contents visible */
	90	if (u->ring_cons == u->ring_prod++) {
	91	wake_up_interruptible(&u->evtchn_wait);
	92	kill_fasync(&u->evtchn_async_queue,
	93	SIGIO, POLL_IN);
	94	}
	95	} else {
	96	u->ring_overflow = 1;
	97	}
	98
	99	spin_unlock(&port_user_lock);
	100
	101	return IRQ_HANDLED;
	102	}
	103
	104	static ssize_t evtchn_read(struct file file, char __user buf,
	105	size_t count, loff_t *ppos)
	106	{
	107	int rc;
	108	unsigned int c, p, bytes1 = 0, bytes2 = 0;
	109	struct per_user_data *u = file->private_data;
	110
	111	/* Whole number of ports. */
	112	count &= ~(sizeof(evtchn_port_t)-1);
	113
	114	if (count == 0)
	115	return 0;
	116
	117	if (count > PAGE_SIZE)
	118	count = PAGE_SIZE;
	119
	120	for (;;) {
	121	mutex_lock(&u->ring_cons_mutex);
	122
	123	rc = -EFBIG;
	124	if (u->ring_overflow)
	125	goto unlock_out;
	126
	127	c = u->ring_cons;
	128	p = u->ring_prod;
	129	if (c != p)
	130	break;
	131
	132	mutex_unlock(&u->ring_cons_mutex);
	133
	134	if (file->f_flags & O_NONBLOCK)
	135	return -EAGAIN;
	136
	137	rc = wait_event_interruptible(u->evtchn_wait,
	138	u->ring_cons != u->ring_prod);
139	if (rc)
140	return rc;
141	}
142
143	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
144	if (((c ^ p) & EVTCHN_RING_SIZE) != 0) {
145	bytes1 = (EVTCHN_RING_SIZE - EVTCHN_RING_MASK(c)) *
146	sizeof(evtchn_port_t);
147	bytes2 = EVTCHN_RING_MASK(p) * sizeof(evtchn_port_t);
148	} else {
149	bytes1 = (p - c) * sizeof(evtchn_port_t);
150	bytes2 = 0;
151	}
152
153	/* Truncate chunks according to caller's maximum byte count. */
154	if (bytes1 > count) {
155	bytes1 = count;
156	bytes2 = 0;
157	} else if ((bytes1 + bytes2) > count) {
158	bytes2 = count - bytes1;
159	}
160
161	rc = -EFAULT;
162	rmb(); /* Ensure that we see the port before we copy it. */
163	if (copy_to_user(buf, &u->ring[EVTCHN_RING_MASK(c)], bytes1) \|\|
164	((bytes2 != 0) &&
165	copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
166	goto unlock_out;
167
168	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
169	rc = bytes1 + bytes2;
170
171	unlock_out:
172	mutex_unlock(&u->ring_cons_mutex);
173	return rc;
174	}
175
176	static ssize_t evtchn_write(struct file file, const char __user buf,
177	size_t count, loff_t *ppos)
178	{
179	int rc, i;
180	evtchn_port_t kbuf = (evtchn_port_t )__get_free_page(GFP_KERNEL);
181	struct per_user_data *u = file->private_data;
182
183	if (kbuf == NULL)
184	return -ENOMEM;
185
186	/* Whole number of ports. */
187	count &= ~(sizeof(evtchn_port_t)-1);
188
189	rc = 0;
190	if (count == 0)
191	goto out;
192
193	if (count > PAGE_SIZE)
194	count = PAGE_SIZE;
195
196	rc = -EFAULT;
197	if (copy_from_user(kbuf, buf, count) != 0)
198	goto out;
199
200	spin_lock_irq(&port_user_lock);
201	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++)
202	if ((kbuf[i] < NR_EVENT_CHANNELS) && (port_user[kbuf[i]] == u))
203	enable_irq(irq_from_evtchn(kbuf[i]));
204	spin_unlock_irq(&port_user_lock);
205
206	rc = count;
207
208	out:
209	free_page((unsigned long)kbuf);
210	return rc;
211	}
212
213	static int evtchn_bind_to_user(struct per_user_data *u, int port)
214	{
f7116284 IC	215	int rc = 0;
f7116284 IC	216
0a4666b5 JF	217	/*
	218	* Ports are never reused, so every caller should pass in a
	219	* unique port.
	220	*
	221	* (Locking not necessary because we haven't registered the
	222	* interrupt handler yet, and our caller has already
	223	* serialized bind operations.)
	224	*/
f7116284	225	BUG_ON(port_user[port] != NULL);
f7116284 IC	226	port_user[port] = u;
f7116284 IC	227
0a4666b5 JF	228	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, IRQF_DISABLED,
	229	u->name, (void *)(unsigned long)port);
	230	if (rc >= 0)
	231	rc = 0;
	232
f7116284 IC	233	return rc;
	234	}
	235
	236	static void evtchn_unbind_from_user(struct per_user_data *u, int port)
	237	{
	238	int irq = irq_from_evtchn(port);
	239
	240	unbind_from_irqhandler(irq, (void *)(unsigned long)port);
0a4666b5 JF	241
	242	/* make sure we unbind the irq handler before clearing the port */
	243	barrier();
	244
f7116284 IC	245	port_user[port] = NULL;
	246	}
	247
	248	static long evtchn_ioctl(struct file *file,
	249	unsigned int cmd, unsigned long arg)
	250	{
	251	int rc;
	252	struct per_user_data *u = file->private_data;
	253	void __user uarg = (void __user ) arg;
	254
0a4666b5 JF	255	/* Prevent bind from racing with unbind */
	256	mutex_lock(&u->bind_mutex);
	257
f7116284 IC	258	switch (cmd) {
	259	case IOCTL_EVTCHN_BIND_VIRQ: {
	260	struct ioctl_evtchn_bind_virq bind;
	261	struct evtchn_bind_virq bind_virq;
	262
	263	rc = -EFAULT;
	264	if (copy_from_user(&bind, uarg, sizeof(bind)))
	265	break;
	266
	267	bind_virq.virq = bind.virq;
	268	bind_virq.vcpu = 0;
	269	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
	270	&bind_virq);
	271	if (rc != 0)
	272	break;
	273
	274	rc = evtchn_bind_to_user(u, bind_virq.port);
	275	if (rc == 0)
	276	rc = bind_virq.port;
	277	break;
	278	}
	279
	280	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
	281	struct ioctl_evtchn_bind_interdomain bind;
	282	struct evtchn_bind_interdomain bind_interdomain;
	283
	284	rc = -EFAULT;
	285	if (copy_from_user(&bind, uarg, sizeof(bind)))
	286	break;
	287
	288	bind_interdomain.remote_dom = bind.remote_domain;
	289	bind_interdomain.remote_port = bind.remote_port;
	290	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
	291	&bind_interdomain);
	292	if (rc != 0)
	293	break;
	294
	295	rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
	296	if (rc == 0)
	297	rc = bind_interdomain.local_port;
	298	break;
	299	}
	300
	301	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
	302	struct ioctl_evtchn_bind_unbound_port bind;
	303	struct evtchn_alloc_unbound alloc_unbound;
	304
	305	rc = -EFAULT;
	306	if (copy_from_user(&bind, uarg, sizeof(bind)))
	307	break;
	308
	309	alloc_unbound.dom = DOMID_SELF;
	310	alloc_unbound.remote_dom = bind.remote_domain;
	311	rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
	312	&alloc_unbound);
	313	if (rc != 0)
	314	break;
	315
	316	rc = evtchn_bind_to_user(u, alloc_unbound.port);
	317	if (rc == 0)
	318	rc = alloc_unbound.port;
	319	break;
	320	}
	321
322	case IOCTL_EVTCHN_UNBIND: {
323	struct ioctl_evtchn_unbind unbind;
324
325	rc = -EFAULT;
326	if (copy_from_user(&unbind, uarg, sizeof(unbind)))
327	break;
328
329	rc = -EINVAL;
330	if (unbind.port >= NR_EVENT_CHANNELS)
331	break;
332
333	spin_lock_irq(&port_user_lock);
334
335	rc = -ENOTCONN;
336	if (port_user[unbind.port] != u) {
337	spin_unlock_irq(&port_user_lock);
338	break;
339	}
340
341	evtchn_unbind_from_user(u, unbind.port);
342
343	spin_unlock_irq(&port_user_lock);
344
345	rc = 0;
346	break;
347	}
348
349	case IOCTL_EVTCHN_NOTIFY: {
350	struct ioctl_evtchn_notify notify;
351
352	rc = -EFAULT;
353	if (copy_from_user(&notify, uarg, sizeof(notify)))
354	break;
355
356	if (notify.port >= NR_EVENT_CHANNELS) {
357	rc = -EINVAL;
358	} else if (port_user[notify.port] != u) {
359	rc = -ENOTCONN;
360	} else {
361	notify_remote_via_evtchn(notify.port);
362	rc = 0;
363	}
364	break;
365	}
366
367	case IOCTL_EVTCHN_RESET: {
368	/* Initialise the ring to empty. Clear errors. */
369	mutex_lock(&u->ring_cons_mutex);
370	spin_lock_irq(&port_user_lock);
371	u->ring_cons = u->ring_prod = u->ring_overflow = 0;
372	spin_unlock_irq(&port_user_lock);
373	mutex_unlock(&u->ring_cons_mutex);
374	rc = 0;
375	break;
376	}
377
378	default:
379	rc = -ENOSYS;
380	break;
381	}
0a4666b5	382	mutex_unlock(&u->bind_mutex);
f7116284 IC	383
	384	return rc;
	385	}
	386
	387	static unsigned int evtchn_poll(struct file file, poll_table wait)
	388	{
	389	unsigned int mask = POLLOUT \| POLLWRNORM;
	390	struct per_user_data *u = file->private_data;
	391
	392	poll_wait(file, &u->evtchn_wait, wait);
	393	if (u->ring_cons != u->ring_prod)
	394	mask \|= POLLIN \| POLLRDNORM;
	395	if (u->ring_overflow)
	396	mask = POLLERR;
	397	return mask;
	398	}
	399
	400	static int evtchn_fasync(int fd, struct file *filp, int on)
	401	{
	402	struct per_user_data *u = filp->private_data;
	403	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
	404	}
	405
	406	static int evtchn_open(struct inode inode, struct file filp)
	407	{
	408	struct per_user_data *u;
	409
	410	u = kzalloc(sizeof(*u), GFP_KERNEL);
	411	if (u == NULL)
	412	return -ENOMEM;
	413
	414	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
	415	if (u->name == NULL) {
	416	kfree(u);
	417	return -ENOMEM;
	418	}
	419
	420	init_waitqueue_head(&u->evtchn_wait);
	421
	422	u->ring = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
	423	if (u->ring == NULL) {
	424	kfree(u->name);
	425	kfree(u);
	426	return -ENOMEM;
	427	}
	428
0a4666b5	429	mutex_init(&u->bind_mutex);
f7116284 IC	430	mutex_init(&u->ring_cons_mutex);
	431
	432	filp->private_data = u;
	433
	434	return 0;
	435	}
	436
	437	static int evtchn_release(struct inode inode, struct file filp)
	438	{
	439	int i;
	440	struct per_user_data *u = filp->private_data;
	441
	442	spin_lock_irq(&port_user_lock);
	443
	444	free_page((unsigned long)u->ring);
	445
	446	for (i = 0; i < NR_EVENT_CHANNELS; i++) {
	447	if (port_user[i] != u)
	448	continue;
	449
	450	evtchn_unbind_from_user(port_user[i], i);
	451	}
	452
	453	spin_unlock_irq(&port_user_lock);
	454
	455	kfree(u->name);
	456	kfree(u);
	457
	458	return 0;
	459	}
	460
	461	static const struct file_operations evtchn_fops = {
	462	.owner = THIS_MODULE,
	463	.read = evtchn_read,
	464	.write = evtchn_write,
	465	.unlocked_ioctl = evtchn_ioctl,
	466	.poll = evtchn_poll,
	467	.fasync = evtchn_fasync,
	468	.open = evtchn_open,
	469	.release = evtchn_release,
6038f373	470	.llseek = noop_llseek,
f7116284 IC	471	};
	472
	473	static struct miscdevice evtchn_miscdev = {
	474	.minor = MISC_DYNAMIC_MINOR,
	475	.name = "evtchn",
	476	.fops = &evtchn_fops,
	477	};
	478	static int __init evtchn_init(void)
	479	{
	480	int err;
	481
	482	if (!xen_domain())
	483	return -ENODEV;
	484
	485	spin_lock_init(&port_user_lock);
	486	memset(port_user, 0, sizeof(port_user));
	487
	488	/* Create '/dev/misc/evtchn'. */
	489	err = misc_register(&evtchn_miscdev);
	490	if (err != 0) {
	491	printk(KERN_ALERT "Could not register /dev/misc/evtchn\n");
	492	return err;
	493	}
	494
	495	printk(KERN_INFO "Event-channel device installed.\n");
	496
	497	return 0;
	498	}
	499
	500	static void __exit evtchn_cleanup(void)
	501	{
	502	misc_deregister(&evtchn_miscdev);
	503	}
	504
	505	module_init(evtchn_init);
	506	module_exit(evtchn_cleanup);
	507
	508	MODULE_LICENSE("GPL");