[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/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?  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;
}

static 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);

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

	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;

	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);
	set_port_enabled(port, true); /* start enabled */

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

	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);

	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;
		}

		disable_irq(irq_from_evtchn(unbind.port));

		spin_unlock_irq(&port_user_lock);

		evtchn_unbind_from_user(u, unbind.port);

		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 nonseekable_open(inode, filp);
}

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;

		disable_irq(irq_from_evtchn(i));
	}

	spin_unlock_irq(&port_user_lock);

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

		evtchn_unbind_from_user(get_port_user(i), i);
	}

	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	 = no_llseek,
};

static struct miscdevice evtchn_miscdev = {
	.minor        = MISC_DYNAMIC_MINOR,
	.name         = "xen/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>
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
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
70697d54	102	static irqreturn_t evtchn_interrupt(int irq, void *data)
f7116284 IC	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);
e3cc067b JF	110
0edce91d JF	111	WARN(!get_port_enabled(port),
	112	"Interrupt for port %d, but apparently not enabled; per-user %p\n",
	113	port, u);
f7116284 IC	114
f7116284 IC	115	disable_irq_nosync(irq);
e3cc067b	116	set_port_enabled(port, false);
f7116284 IC	117
	118	if ((u->ring_prod - u->ring_cons) < EVTCHN_RING_SIZE) {
	119	u->ring[EVTCHN_RING_MASK(u->ring_prod)] = port;
	120	wmb(); /* Ensure ring contents visible */
	121	if (u->ring_cons == u->ring_prod++) {
	122	wake_up_interruptible(&u->evtchn_wait);
	123	kill_fasync(&u->evtchn_async_queue,
	124	SIGIO, POLL_IN);
	125	}
e3cc067b	126	} else
f7116284	127	u->ring_overflow = 1;
f7116284 IC	128
	129	spin_unlock(&port_user_lock);
	130
	131	return IRQ_HANDLED;
	132	}
	133
	134	static ssize_t evtchn_read(struct file file, char __user buf,
	135	size_t count, loff_t *ppos)
	136	{
	137	int rc;
	138	unsigned int c, p, bytes1 = 0, bytes2 = 0;
	139	struct per_user_data *u = file->private_data;
	140
	141	/* Whole number of ports. */
	142	count &= ~(sizeof(evtchn_port_t)-1);
	143
	144	if (count == 0)
	145	return 0;
	146
	147	if (count > PAGE_SIZE)
	148	count = PAGE_SIZE;
	149
	150	for (;;) {
	151	mutex_lock(&u->ring_cons_mutex);
	152
	153	rc = -EFBIG;
	154	if (u->ring_overflow)
	155	goto unlock_out;
	156
	157	c = u->ring_cons;
	158	p = u->ring_prod;
	159	if (c != p)
	160	break;
	161
	162	mutex_unlock(&u->ring_cons_mutex);
	163
	164	if (file->f_flags & O_NONBLOCK)
	165	return -EAGAIN;
	166
	167	rc = wait_event_interruptible(u->evtchn_wait,
	168	u->ring_cons != u->ring_prod);
	169	if (rc)
	170	return rc;
	171	}
	172
	173	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
	174	if (((c ^ p) & EVTCHN_RING_SIZE) != 0) {
	175	bytes1 = (EVTCHN_RING_SIZE - EVTCHN_RING_MASK(c)) *
	176	sizeof(evtchn_port_t);
	177	bytes2 = EVTCHN_RING_MASK(p) * sizeof(evtchn_port_t);
	178	} else {
	179	bytes1 = (p - c) * sizeof(evtchn_port_t);
	180	bytes2 = 0;
	181	}
	182
	183	/* Truncate chunks according to caller's maximum byte count. */
	184	if (bytes1 > count) {
	185	bytes1 = count;
	186	bytes2 = 0;
	187	} else if ((bytes1 + bytes2) > count) {
	188	bytes2 = count - bytes1;
	189	}
	190
	191	rc = -EFAULT;
192	rmb(); /* Ensure that we see the port before we copy it. */
193	if (copy_to_user(buf, &u->ring[EVTCHN_RING_MASK(c)], bytes1) \|\|
194	((bytes2 != 0) &&
195	copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
196	goto unlock_out;
197
198	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
199	rc = bytes1 + bytes2;
200
201	unlock_out:
202	mutex_unlock(&u->ring_cons_mutex);
203	return rc;
204	}
205
206	static ssize_t evtchn_write(struct file file, const char __user buf,
207	size_t count, loff_t *ppos)
208	{
209	int rc, i;
210	evtchn_port_t kbuf = (evtchn_port_t )__get_free_page(GFP_KERNEL);
211	struct per_user_data *u = file->private_data;
212
213	if (kbuf == NULL)
214	return -ENOMEM;
215
216	/* Whole number of ports. */
217	count &= ~(sizeof(evtchn_port_t)-1);
218
219	rc = 0;
220	if (count == 0)
221	goto out;
222
223	if (count > PAGE_SIZE)
224	count = PAGE_SIZE;
225
226	rc = -EFAULT;
227	if (copy_from_user(kbuf, buf, count) != 0)
228	goto out;
229
230	spin_lock_irq(&port_user_lock);
e3cc067b JF	231
	232	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++) {
	233	unsigned port = kbuf[i];
	234
	235	if (port < NR_EVENT_CHANNELS &&
	236	get_port_user(port) == u &&
	237	!get_port_enabled(port)) {
	238	set_port_enabled(port, true);
	239	enable_irq(irq_from_evtchn(port));
	240	}
	241	}
	242
f7116284 IC	243	spin_unlock_irq(&port_user_lock);
	244
	245	rc = count;
	246
	247	out:
	248	free_page((unsigned long)kbuf);
	249	return rc;
	250	}
	251
	252	static int evtchn_bind_to_user(struct per_user_data *u, int port)
	253	{
f7116284 IC	254	int rc = 0;
f7116284 IC	255
0a4666b5 JF	256	/*
	257	* Ports are never reused, so every caller should pass in a
	258	* unique port.
	259	*
	260	* (Locking not necessary because we haven't registered the
	261	* interrupt handler yet, and our caller has already
	262	* serialized bind operations.)
	263	*/
e3cc067b JF	264	BUG_ON(get_port_user(port) != NULL);
e3cc067b JF	265	set_port_user(port, u);
0edce91d	266	set_port_enabled(port, true); /* start enabled */
f7116284	267
0a4666b5 JF	268	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, IRQF_DISABLED,
	269	u->name, (void *)(unsigned long)port);
	270	if (rc >= 0)
420eb554	271	rc = evtchn_make_refcounted(port);
0a4666b5	272
f7116284 IC	273	return rc;
	274	}
	275
	276	static void evtchn_unbind_from_user(struct per_user_data *u, int port)
	277	{
	278	int irq = irq_from_evtchn(port);
	279
	280	unbind_from_irqhandler(irq, (void *)(unsigned long)port);
0a4666b5	281
e3cc067b	282	set_port_user(port, NULL);
f7116284 IC	283	}
	284
	285	static long evtchn_ioctl(struct file *file,
	286	unsigned int cmd, unsigned long arg)
	287	{
	288	int rc;
	289	struct per_user_data *u = file->private_data;
	290	void __user uarg = (void __user ) arg;
	291
0a4666b5 JF	292	/* Prevent bind from racing with unbind */
	293	mutex_lock(&u->bind_mutex);
	294
f7116284 IC	295	switch (cmd) {
	296	case IOCTL_EVTCHN_BIND_VIRQ: {
	297	struct ioctl_evtchn_bind_virq bind;
	298	struct evtchn_bind_virq bind_virq;
	299
	300	rc = -EFAULT;
	301	if (copy_from_user(&bind, uarg, sizeof(bind)))
	302	break;
	303
	304	bind_virq.virq = bind.virq;
	305	bind_virq.vcpu = 0;
	306	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
	307	&bind_virq);
	308	if (rc != 0)
	309	break;
	310
	311	rc = evtchn_bind_to_user(u, bind_virq.port);
	312	if (rc == 0)
	313	rc = bind_virq.port;
	314	break;
	315	}
	316
	317	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
	318	struct ioctl_evtchn_bind_interdomain bind;
	319	struct evtchn_bind_interdomain bind_interdomain;
	320
	321	rc = -EFAULT;
	322	if (copy_from_user(&bind, uarg, sizeof(bind)))
	323	break;
	324
	325	bind_interdomain.remote_dom = bind.remote_domain;
	326	bind_interdomain.remote_port = bind.remote_port;
	327	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
	328	&bind_interdomain);
	329	if (rc != 0)
	330	break;
	331
	332	rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
	333	if (rc == 0)
	334	rc = bind_interdomain.local_port;
	335	break;
	336	}
	337
	338	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
	339	struct ioctl_evtchn_bind_unbound_port bind;
	340	struct evtchn_alloc_unbound alloc_unbound;
	341
	342	rc = -EFAULT;
	343	if (copy_from_user(&bind, uarg, sizeof(bind)))
	344	break;
	345
	346	alloc_unbound.dom = DOMID_SELF;
	347	alloc_unbound.remote_dom = bind.remote_domain;
	348	rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
	349	&alloc_unbound);
	350	if (rc != 0)
	351	break;
	352
	353	rc = evtchn_bind_to_user(u, alloc_unbound.port);
	354	if (rc == 0)
	355	rc = alloc_unbound.port;
	356	break;
	357	}
	358
359	case IOCTL_EVTCHN_UNBIND: {
360	struct ioctl_evtchn_unbind unbind;
361
362	rc = -EFAULT;
363	if (copy_from_user(&unbind, uarg, sizeof(unbind)))
364	break;
365
366	rc = -EINVAL;
367	if (unbind.port >= NR_EVENT_CHANNELS)
368	break;
369
370	spin_lock_irq(&port_user_lock);
371
372	rc = -ENOTCONN;
e3cc067b	373	if (get_port_user(unbind.port) != u) {
f7116284 IC	374	spin_unlock_irq(&port_user_lock);
	375	break;
	376	}
	377
3f5e554f	378	disable_irq(irq_from_evtchn(unbind.port));
f7116284 IC	379
	380	spin_unlock_irq(&port_user_lock);
	381
3f5e554f JF	382	evtchn_unbind_from_user(u, unbind.port);
3f5e554f JF	383
f7116284 IC	384	rc = 0;
	385	break;
	386	}
	387
	388	case IOCTL_EVTCHN_NOTIFY: {
	389	struct ioctl_evtchn_notify notify;
	390
	391	rc = -EFAULT;
	392	if (copy_from_user(&notify, uarg, sizeof(notify)))
	393	break;
	394
	395	if (notify.port >= NR_EVENT_CHANNELS) {
	396	rc = -EINVAL;
e3cc067b	397	} else if (get_port_user(notify.port) != u) {
f7116284 IC	398	rc = -ENOTCONN;
	399	} else {
	400	notify_remote_via_evtchn(notify.port);
	401	rc = 0;
	402	}
	403	break;
	404	}
	405
	406	case IOCTL_EVTCHN_RESET: {
	407	/* Initialise the ring to empty. Clear errors. */
	408	mutex_lock(&u->ring_cons_mutex);
	409	spin_lock_irq(&port_user_lock);
	410	u->ring_cons = u->ring_prod = u->ring_overflow = 0;
	411	spin_unlock_irq(&port_user_lock);
	412	mutex_unlock(&u->ring_cons_mutex);
	413	rc = 0;
	414	break;
	415	}
	416
	417	default:
	418	rc = -ENOSYS;
	419	break;
	420	}
0a4666b5	421	mutex_unlock(&u->bind_mutex);
f7116284 IC	422
	423	return rc;
	424	}
	425
	426	static unsigned int evtchn_poll(struct file file, poll_table wait)
	427	{
	428	unsigned int mask = POLLOUT \| POLLWRNORM;
	429	struct per_user_data *u = file->private_data;
	430
	431	poll_wait(file, &u->evtchn_wait, wait);
	432	if (u->ring_cons != u->ring_prod)
	433	mask \|= POLLIN \| POLLRDNORM;
	434	if (u->ring_overflow)
	435	mask = POLLERR;
	436	return mask;
	437	}
	438
	439	static int evtchn_fasync(int fd, struct file *filp, int on)
	440	{
	441	struct per_user_data *u = filp->private_data;
	442	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
	443	}
	444
	445	static int evtchn_open(struct inode inode, struct file filp)
	446	{
	447	struct per_user_data *u;
	448
	449	u = kzalloc(sizeof(*u), GFP_KERNEL);
	450	if (u == NULL)
	451	return -ENOMEM;
	452
	453	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
	454	if (u->name == NULL) {
	455	kfree(u);
	456	return -ENOMEM;
	457	}
	458
	459	init_waitqueue_head(&u->evtchn_wait);
	460
	461	u->ring = (evtchn_port_t *)__get_free_page(GFP_KERNEL);
	462	if (u->ring == NULL) {
	463	kfree(u->name);
	464	kfree(u);
	465	return -ENOMEM;
	466	}
	467
0a4666b5	468	mutex_init(&u->bind_mutex);
f7116284 IC	469	mutex_init(&u->ring_cons_mutex);
	470
	471	filp->private_data = u;
	472
6eab04a8	473	return nonseekable_open(inode, filp);
f7116284 IC	474	}
	475
	476	static int evtchn_release(struct inode inode, struct file filp)
	477	{
	478	int i;
	479	struct per_user_data *u = filp->private_data;
	480
	481	spin_lock_irq(&port_user_lock);
	482
	483	free_page((unsigned long)u->ring);
	484
	485	for (i = 0; i < NR_EVENT_CHANNELS; i++) {
e3cc067b	486	if (get_port_user(i) != u)
f7116284 IC	487	continue;
f7116284 IC	488
3f5e554f	489	disable_irq(irq_from_evtchn(i));
f7116284 IC	490	}
	491
	492	spin_unlock_irq(&port_user_lock);
	493
3f5e554f JF	494	for (i = 0; i < NR_EVENT_CHANNELS; i++) {
	495	if (get_port_user(i) != u)
	496	continue;
	497
	498	evtchn_unbind_from_user(get_port_user(i), i);
	499	}
	500
f7116284 IC	501	kfree(u->name);
	502	kfree(u);
	503
	504	return 0;
	505	}
	506
	507	static const struct file_operations evtchn_fops = {
	508	.owner = THIS_MODULE,
	509	.read = evtchn_read,
	510	.write = evtchn_write,
	511	.unlocked_ioctl = evtchn_ioctl,
	512	.poll = evtchn_poll,
	513	.fasync = evtchn_fasync,
	514	.open = evtchn_open,
	515	.release = evtchn_release,
bc7fc5e3	516	.llseek = no_llseek,
f7116284 IC	517	};
	518
	519	static struct miscdevice evtchn_miscdev = {
	520	.minor = MISC_DYNAMIC_MINOR,
376d908f	521	.name = "xen/evtchn",
f7116284 IC	522	.fops = &evtchn_fops,
	523	};
	524	static int __init evtchn_init(void)
	525	{
	526	int err;
	527
	528	if (!xen_domain())
	529	return -ENODEV;
	530
93afe0b7 JF	531	port_user = kcalloc(NR_EVENT_CHANNELS, sizeof(*port_user), GFP_KERNEL);
	532	if (port_user == NULL)
	533	return -ENOMEM;
	534
f7116284	535	spin_lock_init(&port_user_lock);
f7116284 IC	536
	537	/* Create '/dev/misc/evtchn'. */
	538	err = misc_register(&evtchn_miscdev);
	539	if (err != 0) {
	540	printk(KERN_ALERT "Could not register /dev/misc/evtchn\n");
	541	return err;
	542	}
	543
	544	printk(KERN_INFO "Event-channel device installed.\n");
	545
	546	return 0;
	547	}
	548
	549	static void __exit evtchn_cleanup(void)
	550	{
93afe0b7 JF	551	kfree(port_user);
	552	port_user = NULL;
	553
f7116284 IC	554	misc_deregister(&evtchn_miscdev);
	555	}
	556
	557	module_init(evtchn_init);
	558	module_exit(evtchn_cleanup);
	559
	560	MODULE_LICENSE("GPL");