[linux-2.6-block.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.
 */

#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt

#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 <linux/mm.h>
#include <linux/vmalloc.h>

#include <xen/xen.h>
#include <xen/events.h>
#include <xen/evtchn.h>
#include <xen/xen-ops.h>
#include <asm/xen/hypervisor.h>

struct per_user_data {
	struct mutex bind_mutex; /* serialize bind/unbind operations */
	struct rb_root evtchns;
	unsigned int nr_evtchns;

	/* Notification ring, accessed via /dev/xen/evtchn. */
	unsigned int ring_size;
	evtchn_port_t *ring;
	unsigned int ring_cons, ring_prod, ring_overflow;
	struct mutex ring_cons_mutex; /* protect against concurrent readers */
	spinlock_t ring_prod_lock; /* product against concurrent interrupts */

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

	domid_t restrict_domid;
};

#define UNRESTRICTED_DOMID ((domid_t)-1)

struct user_evtchn {
	struct rb_node node;
	struct per_user_data *user;
	unsigned port;
	bool enabled;
};

static evtchn_port_t *evtchn_alloc_ring(unsigned int size)
{
	evtchn_port_t *ring;
	size_t s = size * sizeof(*ring);

	ring = kmalloc(s, GFP_KERNEL);
	if (!ring)
		ring = vmalloc(s);

	return ring;
}

static void evtchn_free_ring(evtchn_port_t *ring)
{
	kvfree(ring);
}

static unsigned int evtchn_ring_offset(struct per_user_data *u,
				       unsigned int idx)
{
	return idx & (u->ring_size - 1);
}

static evtchn_port_t *evtchn_ring_entry(struct per_user_data *u,
					unsigned int idx)
{
	return u->ring + evtchn_ring_offset(u, idx);
}

static int add_evtchn(struct per_user_data *u, struct user_evtchn *evtchn)
{
	struct rb_node **new = &(u->evtchns.rb_node), *parent = NULL;

	u->nr_evtchns++;

	while (*new) {
		struct user_evtchn *this;

		this = rb_entry(*new, struct user_evtchn, node);

		parent = *new;
		if (this->port < evtchn->port)
			new = &((*new)->rb_left);
		else if (this->port > evtchn->port)
			new = &((*new)->rb_right);
		else
			return -EEXIST;
	}

	/* Add new node and rebalance tree. */
	rb_link_node(&evtchn->node, parent, new);
	rb_insert_color(&evtchn->node, &u->evtchns);

	return 0;
}

static void del_evtchn(struct per_user_data *u, struct user_evtchn *evtchn)
{
	u->nr_evtchns--;
	rb_erase(&evtchn->node, &u->evtchns);
	kfree(evtchn);
}

static struct user_evtchn *find_evtchn(struct per_user_data *u, unsigned port)
{
	struct rb_node *node = u->evtchns.rb_node;

	while (node) {
		struct user_evtchn *evtchn;

		evtchn = rb_entry(node, struct user_evtchn, node);

		if (evtchn->port < port)
			node = node->rb_left;
		else if (evtchn->port > port)
			node = node->rb_right;
		else
			return evtchn;
	}
	return NULL;
}

static irqreturn_t evtchn_interrupt(int irq, void *data)
{
	struct user_evtchn *evtchn = data;
	struct per_user_data *u = evtchn->user;

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

	disable_irq_nosync(irq);
	evtchn->enabled = false;

	spin_lock(&u->ring_prod_lock);

	if ((u->ring_prod - u->ring_cons) < u->ring_size) {
		*evtchn_ring_entry(u, u->ring_prod) = evtchn->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(&u->ring_prod_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) & u->ring_size) != 0) {
		bytes1 = (u->ring_size - evtchn_ring_offset(u, c)) *
			sizeof(evtchn_port_t);
		bytes2 = evtchn_ring_offset(u, 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, evtchn_ring_entry(u, 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;

	mutex_lock(&u->bind_mutex);

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

		evtchn = find_evtchn(u, port);
		if (evtchn && !evtchn->enabled) {
			evtchn->enabled = true;
			enable_irq(irq_from_evtchn(port));
		}
	}

	mutex_unlock(&u->bind_mutex);

	rc = count;

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

static int evtchn_resize_ring(struct per_user_data *u)
{
	unsigned int new_size;
	evtchn_port_t *new_ring, *old_ring;

	/*
	 * Ensure the ring is large enough to capture all possible
	 * events. i.e., one free slot for each bound event.
	 */
	if (u->nr_evtchns <= u->ring_size)
		return 0;

	if (u->ring_size == 0)
		new_size = 64;
	else
		new_size = 2 * u->ring_size;

	new_ring = evtchn_alloc_ring(new_size);
	if (!new_ring)
		return -ENOMEM;

	old_ring = u->ring;

	/*
	 * Access to the ring contents is serialized by either the
	 * prod /or/ cons lock so take both when resizing.
	 */
	mutex_lock(&u->ring_cons_mutex);
	spin_lock_irq(&u->ring_prod_lock);

	/*
	 * Copy the old ring contents to the new ring.
	 *
	 * To take care of wrapping, a full ring, and the new index
	 * pointing into the second half, simply copy the old contents
	 * twice.
	 *
	 * +---------+    +------------------+
	 * |34567  12| -> |34567  1234567  12|
	 * +-----p-c-+    +-------c------p---+
	 */
	memcpy(new_ring, old_ring, u->ring_size * sizeof(*u->ring));
	memcpy(new_ring + u->ring_size, old_ring,
	       u->ring_size * sizeof(*u->ring));

	u->ring = new_ring;
	u->ring_size = new_size;

	spin_unlock_irq(&u->ring_prod_lock);
	mutex_unlock(&u->ring_cons_mutex);

	evtchn_free_ring(old_ring);

	return 0;
}

static int evtchn_bind_to_user(struct per_user_data *u, int port)
{
	struct user_evtchn *evtchn;
	struct evtchn_close close;
	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.)
	 */

	evtchn = kzalloc(sizeof(*evtchn), GFP_KERNEL);
	if (!evtchn)
		return -ENOMEM;

	evtchn->user = u;
	evtchn->port = port;
	evtchn->enabled = true; /* start enabled */

	rc = add_evtchn(u, evtchn);
	if (rc < 0)
		goto err;

	rc = evtchn_resize_ring(u);
	if (rc < 0)
		goto err;

	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, 0,
				       u->name, evtchn);
	if (rc < 0)
		goto err;

	rc = evtchn_make_refcounted(port);
	return rc;

err:
	/* bind failed, should close the port now */
	close.port = port;
	if (HYPERVISOR_event_channel_op(EVTCHNOP_close, &close) != 0)
		BUG();
	del_evtchn(u, evtchn);
	return rc;
}

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

	BUG_ON(irq < 0);

	unbind_from_irqhandler(irq, evtchn);

	del_evtchn(u, evtchn);
}

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 = -EACCES;
		if (u->restrict_domid != UNRESTRICTED_DOMID)
			break;

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

		bind_virq.virq = bind.virq;
		bind_virq.vcpu = xen_vcpu_nr(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;

		rc = -EACCES;
		if (u->restrict_domid != UNRESTRICTED_DOMID &&
		    u->restrict_domid != bind.remote_domain)
			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 = -EACCES;
		if (u->restrict_domid != UNRESTRICTED_DOMID)
			break;

		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;
		struct user_evtchn *evtchn;

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

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

		rc = -ENOTCONN;
		evtchn = find_evtchn(u, unbind.port);
		if (!evtchn)
			break;

		disable_irq(irq_from_evtchn(unbind.port));
		evtchn_unbind_from_user(u, evtchn);
		rc = 0;
		break;
	}

	case IOCTL_EVTCHN_NOTIFY: {
		struct ioctl_evtchn_notify notify;
		struct user_evtchn *evtchn;

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

		rc = -ENOTCONN;
		evtchn = find_evtchn(u, notify.port);
		if (evtchn) {
			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(&u->ring_prod_lock);
		u->ring_cons = u->ring_prod = u->ring_overflow = 0;
		spin_unlock_irq(&u->ring_prod_lock);
		mutex_unlock(&u->ring_cons_mutex);
		rc = 0;
		break;
	}

	case IOCTL_EVTCHN_RESTRICT_DOMID: {
		struct ioctl_evtchn_restrict_domid ierd;

		rc = -EACCES;
		if (u->restrict_domid != UNRESTRICTED_DOMID)
			break;

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

		rc = -EINVAL;
		if (ierd.domid == 0 || ierd.domid >= DOMID_FIRST_RESERVED)
			break;

		u->restrict_domid = ierd.domid;
		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);

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

	u->restrict_domid = UNRESTRICTED_DOMID;

	filp->private_data = u;

	return nonseekable_open(inode, filp);
}

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

	while ((node = u->evtchns.rb_node)) {
		struct user_evtchn *evtchn;

		evtchn = rb_entry(node, struct user_evtchn, node);
		disable_irq(irq_from_evtchn(evtchn->port));
		evtchn_unbind_from_user(u, evtchn);
	}

	evtchn_free_ring(u->ring);
	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;

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

	pr_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
283c0972 JP	34	#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
283c0972 JP	35
f7116284 IC	36	#include <linux/module.h>
	37	#include <linux/kernel.h>
	38	#include <linux/sched.h>
	39	#include <linux/slab.h>
	40	#include <linux/string.h>
	41	#include <linux/errno.h>
	42	#include <linux/fs.h>
f7116284 IC	43	#include <linux/miscdevice.h>
	44	#include <linux/major.h>
	45	#include <linux/proc_fs.h>
	46	#include <linux/stat.h>
	47	#include <linux/poll.h>
	48	#include <linux/irq.h>
	49	#include <linux/init.h>
f7116284 IC	50	#include <linux/mutex.h>
f7116284 IC	51	#include <linux/cpu.h>
86200154 DV	52	#include <linux/mm.h>
86200154 DV	53	#include <linux/vmalloc.h>
1ccbf534 JF	54
1ccbf534 JF	55	#include <xen/xen.h>
f7116284 IC	56	#include <xen/events.h>
f7116284 IC	57	#include <xen/evtchn.h>
cbbb4682	58	#include <xen/xen-ops.h>
f7116284 IC	59	#include <asm/xen/hypervisor.h>
	60
	61	struct per_user_data {
0a4666b5	62	struct mutex bind_mutex; /* serialize bind/unbind operations */
73cc4bb0	63	struct rb_root evtchns;
86200154	64	unsigned int nr_evtchns;
0a4666b5	65
f7116284	66	/* Notification ring, accessed via /dev/xen/evtchn. */
86200154	67	unsigned int ring_size;
f7116284 IC	68	evtchn_port_t *ring;
	69	unsigned int ring_cons, ring_prod, ring_overflow;
	70	struct mutex ring_cons_mutex; /* protect against concurrent readers */
73cc4bb0	71	spinlock_t ring_prod_lock; /* product against concurrent interrupts */
f7116284 IC	72
	73	/* Processes wait on this queue when ring is empty. */
	74	wait_queue_head_t evtchn_wait;
	75	struct fasync_struct *evtchn_async_queue;
	76	const char *name;
fbc872c3 DV	77
fbc872c3 DV	78	domid_t restrict_domid;
f7116284 IC	79	};
f7116284 IC	80
fbc872c3 DV	81	#define UNRESTRICTED_DOMID ((domid_t)-1)
fbc872c3 DV	82
73cc4bb0 DV	83	struct user_evtchn {
	84	struct rb_node node;
	85	struct per_user_data *user;
	86	unsigned port;
	87	bool enabled;
	88	};
f7116284	89
86200154 DV	90	static evtchn_port_t *evtchn_alloc_ring(unsigned int size)
	91	{
	92	evtchn_port_t *ring;
	93	size_t s = size * sizeof(*ring);
	94
	95	ring = kmalloc(s, GFP_KERNEL);
	96	if (!ring)
	97	ring = vmalloc(s);
	98
	99	return ring;
	100	}
	101
	102	static void evtchn_free_ring(evtchn_port_t *ring)
	103	{
	104	kvfree(ring);
	105	}
	106
	107	static unsigned int evtchn_ring_offset(struct per_user_data *u,
	108	unsigned int idx)
	109	{
	110	return idx & (u->ring_size - 1);
	111	}
	112
	113	static evtchn_port_t evtchn_ring_entry(struct per_user_data u,
	114	unsigned int idx)
	115	{
	116	return u->ring + evtchn_ring_offset(u, idx);
	117	}
	118
73cc4bb0	119	static int add_evtchn(struct per_user_data u, struct user_evtchn evtchn)
e3cc067b	120	{
73cc4bb0	121	struct rb_node *new = &(u->evtchns.rb_node), parent = NULL;
e3cc067b	122
86200154 DV	123	u->nr_evtchns++;
86200154 DV	124
73cc4bb0 DV	125	while (*new) {
	126	struct user_evtchn *this;
	127
2f60b288	128	this = rb_entry(*new, struct user_evtchn, node);
73cc4bb0 DV	129
	130	parent = *new;
	131	if (this->port < evtchn->port)
	132	new = &((*new)->rb_left);
	133	else if (this->port > evtchn->port)
	134	new = &((*new)->rb_right);
	135	else
	136	return -EEXIST;
	137	}
	138
	139	/* Add new node and rebalance tree. */
	140	rb_link_node(&evtchn->node, parent, new);
	141	rb_insert_color(&evtchn->node, &u->evtchns);
	142
	143	return 0;
e3cc067b JF	144	}
e3cc067b JF	145
73cc4bb0	146	static void del_evtchn(struct per_user_data u, struct user_evtchn evtchn)
e3cc067b	147	{
86200154	148	u->nr_evtchns--;
73cc4bb0 DV	149	rb_erase(&evtchn->node, &u->evtchns);
73cc4bb0 DV	150	kfree(evtchn);
e3cc067b JF	151	}
e3cc067b JF	152
73cc4bb0	153	static struct user_evtchn find_evtchn(struct per_user_data u, unsigned port)
e3cc067b	154	{
73cc4bb0 DV	155	struct rb_node *node = u->evtchns.rb_node;
	156
	157	while (node) {
	158	struct user_evtchn *evtchn;
	159
2f60b288	160	evtchn = rb_entry(node, struct user_evtchn, node);
73cc4bb0 DV	161
	162	if (evtchn->port < port)
	163	node = node->rb_left;
	164	else if (evtchn->port > port)
	165	node = node->rb_right;
	166	else
	167	return evtchn;
	168	}
	169	return NULL;
e3cc067b JF	170	}
e3cc067b JF	171
70697d54	172	static irqreturn_t evtchn_interrupt(int irq, void *data)
f7116284	173	{
73cc4bb0 DV	174	struct user_evtchn *evtchn = data;
73cc4bb0 DV	175	struct per_user_data *u = evtchn->user;
e3cc067b	176
73cc4bb0	177	WARN(!evtchn->enabled,
0edce91d	178	"Interrupt for port %d, but apparently not enabled; per-user %p\n",
73cc4bb0	179	evtchn->port, u);
f7116284 IC	180
f7116284 IC	181	disable_irq_nosync(irq);
73cc4bb0 DV	182	evtchn->enabled = false;
	183
	184	spin_lock(&u->ring_prod_lock);
f7116284	185
86200154 DV	186	if ((u->ring_prod - u->ring_cons) < u->ring_size) {
86200154 DV	187	*evtchn_ring_entry(u, u->ring_prod) = evtchn->port;
f7116284 IC	188	wmb(); /* Ensure ring contents visible */
	189	if (u->ring_cons == u->ring_prod++) {
	190	wake_up_interruptible(&u->evtchn_wait);
	191	kill_fasync(&u->evtchn_async_queue,
	192	SIGIO, POLL_IN);
	193	}
e3cc067b	194	} else
f7116284	195	u->ring_overflow = 1;
f7116284	196
73cc4bb0	197	spin_unlock(&u->ring_prod_lock);
f7116284 IC	198
	199	return IRQ_HANDLED;
	200	}
	201
	202	static ssize_t evtchn_read(struct file file, char __user buf,
	203	size_t count, loff_t *ppos)
	204	{
	205	int rc;
	206	unsigned int c, p, bytes1 = 0, bytes2 = 0;
	207	struct per_user_data *u = file->private_data;
	208
	209	/* Whole number of ports. */
	210	count &= ~(sizeof(evtchn_port_t)-1);
	211
	212	if (count == 0)
	213	return 0;
	214
	215	if (count > PAGE_SIZE)
	216	count = PAGE_SIZE;
	217
	218	for (;;) {
	219	mutex_lock(&u->ring_cons_mutex);
	220
	221	rc = -EFBIG;
	222	if (u->ring_overflow)
	223	goto unlock_out;
	224
	225	c = u->ring_cons;
	226	p = u->ring_prod;
	227	if (c != p)
	228	break;
	229
	230	mutex_unlock(&u->ring_cons_mutex);
	231
	232	if (file->f_flags & O_NONBLOCK)
	233	return -EAGAIN;
	234
	235	rc = wait_event_interruptible(u->evtchn_wait,
	236	u->ring_cons != u->ring_prod);
	237	if (rc)
	238	return rc;
	239	}
	240
	241	/* Byte lengths of two chunks. Chunk split (if any) is at ring wrap. */
86200154 DV	242	if (((c ^ p) & u->ring_size) != 0) {
86200154 DV	243	bytes1 = (u->ring_size - evtchn_ring_offset(u, c)) *
f7116284	244	sizeof(evtchn_port_t);
86200154	245	bytes2 = evtchn_ring_offset(u, p) * sizeof(evtchn_port_t);
f7116284 IC	246	} else {
	247	bytes1 = (p - c) * sizeof(evtchn_port_t);
	248	bytes2 = 0;
	249	}
	250
	251	/* Truncate chunks according to caller's maximum byte count. */
	252	if (bytes1 > count) {
	253	bytes1 = count;
	254	bytes2 = 0;
	255	} else if ((bytes1 + bytes2) > count) {
	256	bytes2 = count - bytes1;
	257	}
	258
	259	rc = -EFAULT;
	260	rmb(); /* Ensure that we see the port before we copy it. */
86200154	261	if (copy_to_user(buf, evtchn_ring_entry(u, c), bytes1) \|\|
f7116284 IC	262	((bytes2 != 0) &&
	263	copy_to_user(&buf[bytes1], &u->ring[0], bytes2)))
	264	goto unlock_out;
	265
	266	u->ring_cons += (bytes1 + bytes2) / sizeof(evtchn_port_t);
	267	rc = bytes1 + bytes2;
	268
	269	unlock_out:
	270	mutex_unlock(&u->ring_cons_mutex);
	271	return rc;
	272	}
	273
	274	static ssize_t evtchn_write(struct file file, const char __user buf,
	275	size_t count, loff_t *ppos)
	276	{
	277	int rc, i;
	278	evtchn_port_t kbuf = (evtchn_port_t )__get_free_page(GFP_KERNEL);
	279	struct per_user_data *u = file->private_data;
	280
	281	if (kbuf == NULL)
	282	return -ENOMEM;
	283
	284	/* Whole number of ports. */
	285	count &= ~(sizeof(evtchn_port_t)-1);
	286
	287	rc = 0;
	288	if (count == 0)
	289	goto out;
	290
	291	if (count > PAGE_SIZE)
	292	count = PAGE_SIZE;
	293
	294	rc = -EFAULT;
	295	if (copy_from_user(kbuf, buf, count) != 0)
	296	goto out;
	297
73cc4bb0	298	mutex_lock(&u->bind_mutex);
e3cc067b JF	299
	300	for (i = 0; i < (count/sizeof(evtchn_port_t)); i++) {
	301	unsigned port = kbuf[i];
73cc4bb0	302	struct user_evtchn *evtchn;
e3cc067b	303
73cc4bb0 DV	304	evtchn = find_evtchn(u, port);
	305	if (evtchn && !evtchn->enabled) {
	306	evtchn->enabled = true;
e3cc067b JF	307	enable_irq(irq_from_evtchn(port));
	308	}
	309	}
	310
73cc4bb0	311	mutex_unlock(&u->bind_mutex);
f7116284 IC	312
	313	rc = count;
	314
	315	out:
	316	free_page((unsigned long)kbuf);
	317	return rc;
	318	}
	319
86200154 DV	320	static int evtchn_resize_ring(struct per_user_data *u)
	321	{
	322	unsigned int new_size;
	323	evtchn_port_t new_ring, old_ring;
86200154 DV	324
	325	/*
	326	* Ensure the ring is large enough to capture all possible
	327	* events. i.e., one free slot for each bound event.
	328	*/
	329	if (u->nr_evtchns <= u->ring_size)
	330	return 0;
	331
	332	if (u->ring_size == 0)
	333	new_size = 64;
	334	else
	335	new_size = 2 * u->ring_size;
	336
	337	new_ring = evtchn_alloc_ring(new_size);
	338	if (!new_ring)
	339	return -ENOMEM;
	340
	341	old_ring = u->ring;
	342
	343	/*
	344	* Access to the ring contents is serialized by either the
	345	* prod /or/ cons lock so take both when resizing.
	346	*/
	347	mutex_lock(&u->ring_cons_mutex);
	348	spin_lock_irq(&u->ring_prod_lock);
	349
	350	/*
	351	* Copy the old ring contents to the new ring.
	352	*
27e0e638 JB	353	* To take care of wrapping, a full ring, and the new index
	354	* pointing into the second half, simply copy the old contents
	355	* twice.
86200154 DV	356	*
86200154 DV	357	* +---------+ +------------------+
27e0e638 JB	358	* \|34567 12\| -> \|34567 1234567 12\|
27e0e638 JB	359	* +-----p-c-+ +-------c------p---+
86200154	360	*/
27e0e638 JB	361	memcpy(new_ring, old_ring, u->ring_size * sizeof(*u->ring));
	362	memcpy(new_ring + u->ring_size, old_ring,
	363	u->ring_size * sizeof(*u->ring));
86200154 DV	364
	365	u->ring = new_ring;
	366	u->ring_size = new_size;
	367
	368	spin_unlock_irq(&u->ring_prod_lock);
	369	mutex_unlock(&u->ring_cons_mutex);
	370
	371	evtchn_free_ring(old_ring);
	372
	373	return 0;
	374	}
	375
f7116284 IC	376	static int evtchn_bind_to_user(struct per_user_data *u, int port)
f7116284 IC	377	{
73cc4bb0 DV	378	struct user_evtchn *evtchn;
73cc4bb0 DV	379	struct evtchn_close close;
f7116284 IC	380	int rc = 0;
f7116284 IC	381
0a4666b5 JF	382	/*
	383	* Ports are never reused, so every caller should pass in a
	384	* unique port.
	385	*
	386	* (Locking not necessary because we haven't registered the
	387	* interrupt handler yet, and our caller has already
	388	* serialized bind operations.)
	389	*/
73cc4bb0 DV	390
	391	evtchn = kzalloc(sizeof(*evtchn), GFP_KERNEL);
	392	if (!evtchn)
	393	return -ENOMEM;
	394
	395	evtchn->user = u;
	396	evtchn->port = port;
	397	evtchn->enabled = true; /* start enabled */
	398
	399	rc = add_evtchn(u, evtchn);
	400	if (rc < 0)
	401	goto err;
f7116284	402
86200154 DV	403	rc = evtchn_resize_ring(u);
	404	if (rc < 0)
	405	goto err;
	406
af09d1a7	407	rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, 0,
73cc4bb0 DV	408	u->name, evtchn);
	409	if (rc < 0)
	410	goto err;
0a4666b5	411
73cc4bb0 DV	412	rc = evtchn_make_refcounted(port);
	413	return rc;
	414
	415	err:
	416	/* bind failed, should close the port now */
	417	close.port = port;
	418	if (HYPERVISOR_event_channel_op(EVTCHNOP_close, &close) != 0)
	419	BUG();
	420	del_evtchn(u, evtchn);
f7116284 IC	421	return rc;
	422	}
	423
73cc4bb0 DV	424	static void evtchn_unbind_from_user(struct per_user_data *u,
73cc4bb0 DV	425	struct user_evtchn *evtchn)
f7116284	426	{
73cc4bb0	427	int irq = irq_from_evtchn(evtchn->port);
f7116284	428
e7e44e44 WL	429	BUG_ON(irq < 0);
e7e44e44 WL	430
73cc4bb0	431	unbind_from_irqhandler(irq, evtchn);
0a4666b5	432
73cc4bb0	433	del_evtchn(u, evtchn);
f7116284 IC	434	}
	435
	436	static long evtchn_ioctl(struct file *file,
	437	unsigned int cmd, unsigned long arg)
	438	{
	439	int rc;
	440	struct per_user_data *u = file->private_data;
	441	void __user uarg = (void __user ) arg;
	442
0a4666b5 JF	443	/* Prevent bind from racing with unbind */
	444	mutex_lock(&u->bind_mutex);
	445
f7116284 IC	446	switch (cmd) {
	447	case IOCTL_EVTCHN_BIND_VIRQ: {
	448	struct ioctl_evtchn_bind_virq bind;
	449	struct evtchn_bind_virq bind_virq;
	450
fbc872c3 DV	451	rc = -EACCES;
	452	if (u->restrict_domid != UNRESTRICTED_DOMID)
	453	break;
	454
f7116284 IC	455	rc = -EFAULT;
	456	if (copy_from_user(&bind, uarg, sizeof(bind)))
	457	break;
	458
	459	bind_virq.virq = bind.virq;
cbbb4682	460	bind_virq.vcpu = xen_vcpu_nr(0);
f7116284 IC	461	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
	462	&bind_virq);
	463	if (rc != 0)
	464	break;
	465
	466	rc = evtchn_bind_to_user(u, bind_virq.port);
	467	if (rc == 0)
	468	rc = bind_virq.port;
	469	break;
	470	}
	471
	472	case IOCTL_EVTCHN_BIND_INTERDOMAIN: {
	473	struct ioctl_evtchn_bind_interdomain bind;
	474	struct evtchn_bind_interdomain bind_interdomain;
	475
	476	rc = -EFAULT;
	477	if (copy_from_user(&bind, uarg, sizeof(bind)))
	478	break;
	479
fbc872c3 DV	480	rc = -EACCES;
	481	if (u->restrict_domid != UNRESTRICTED_DOMID &&
	482	u->restrict_domid != bind.remote_domain)
	483	break;
	484
f7116284 IC	485	bind_interdomain.remote_dom = bind.remote_domain;
	486	bind_interdomain.remote_port = bind.remote_port;
	487	rc = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain,
	488	&bind_interdomain);
	489	if (rc != 0)
	490	break;
	491
	492	rc = evtchn_bind_to_user(u, bind_interdomain.local_port);
	493	if (rc == 0)
	494	rc = bind_interdomain.local_port;
	495	break;
	496	}
	497
	498	case IOCTL_EVTCHN_BIND_UNBOUND_PORT: {
	499	struct ioctl_evtchn_bind_unbound_port bind;
	500	struct evtchn_alloc_unbound alloc_unbound;
	501
fbc872c3 DV	502	rc = -EACCES;
	503	if (u->restrict_domid != UNRESTRICTED_DOMID)
	504	break;
	505
f7116284 IC	506	rc = -EFAULT;
	507	if (copy_from_user(&bind, uarg, sizeof(bind)))
	508	break;
	509
	510	alloc_unbound.dom = DOMID_SELF;
	511	alloc_unbound.remote_dom = bind.remote_domain;
	512	rc = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
	513	&alloc_unbound);
	514	if (rc != 0)
	515	break;
	516
	517	rc = evtchn_bind_to_user(u, alloc_unbound.port);
	518	if (rc == 0)
	519	rc = alloc_unbound.port;
	520	break;
	521	}
	522
	523	case IOCTL_EVTCHN_UNBIND: {
	524	struct ioctl_evtchn_unbind unbind;
73cc4bb0	525	struct user_evtchn *evtchn;
f7116284 IC	526
	527	rc = -EFAULT;
	528	if (copy_from_user(&unbind, uarg, sizeof(unbind)))
	529	break;
	530
	531	rc = -EINVAL;
0dc0064a	532	if (unbind.port >= xen_evtchn_nr_channels())
f7116284 IC	533	break;
f7116284 IC	534
f7116284	535	rc = -ENOTCONN;
73cc4bb0 DV	536	evtchn = find_evtchn(u, unbind.port);
73cc4bb0 DV	537	if (!evtchn)
f7116284	538	break;
f7116284	539
3f5e554f	540	disable_irq(irq_from_evtchn(unbind.port));
73cc4bb0	541	evtchn_unbind_from_user(u, evtchn);
f7116284 IC	542	rc = 0;
	543	break;
	544	}
	545
	546	case IOCTL_EVTCHN_NOTIFY: {
	547	struct ioctl_evtchn_notify notify;
73cc4bb0	548	struct user_evtchn *evtchn;
f7116284 IC	549
	550	rc = -EFAULT;
	551	if (copy_from_user(&notify, uarg, sizeof(notify)))
	552	break;
	553
73cc4bb0 DV	554	rc = -ENOTCONN;
	555	evtchn = find_evtchn(u, notify.port);
	556	if (evtchn) {
f7116284 IC	557	notify_remote_via_evtchn(notify.port);
	558	rc = 0;
	559	}
	560	break;
	561	}
	562
	563	case IOCTL_EVTCHN_RESET: {
	564	/* Initialise the ring to empty. Clear errors. */
	565	mutex_lock(&u->ring_cons_mutex);
73cc4bb0	566	spin_lock_irq(&u->ring_prod_lock);
f7116284	567	u->ring_cons = u->ring_prod = u->ring_overflow = 0;
73cc4bb0	568	spin_unlock_irq(&u->ring_prod_lock);
f7116284 IC	569	mutex_unlock(&u->ring_cons_mutex);
	570	rc = 0;
	571	break;
	572	}
	573
fbc872c3 DV	574	case IOCTL_EVTCHN_RESTRICT_DOMID: {
	575	struct ioctl_evtchn_restrict_domid ierd;
	576
	577	rc = -EACCES;
	578	if (u->restrict_domid != UNRESTRICTED_DOMID)
	579	break;
	580
	581	rc = -EFAULT;
	582	if (copy_from_user(&ierd, uarg, sizeof(ierd)))
	583	break;
	584
	585	rc = -EINVAL;
	586	if (ierd.domid == 0 \|\| ierd.domid >= DOMID_FIRST_RESERVED)
	587	break;
	588
	589	u->restrict_domid = ierd.domid;
	590	rc = 0;
	591
	592	break;
	593	}
	594
f7116284 IC	595	default:
	596	rc = -ENOSYS;
	597	break;
	598	}
0a4666b5	599	mutex_unlock(&u->bind_mutex);
f7116284 IC	600
	601	return rc;
	602	}
	603
	604	static unsigned int evtchn_poll(struct file file, poll_table wait)
	605	{
	606	unsigned int mask = POLLOUT \| POLLWRNORM;
	607	struct per_user_data *u = file->private_data;
	608
	609	poll_wait(file, &u->evtchn_wait, wait);
	610	if (u->ring_cons != u->ring_prod)
	611	mask \|= POLLIN \| POLLRDNORM;
	612	if (u->ring_overflow)
	613	mask = POLLERR;
	614	return mask;
	615	}
	616
	617	static int evtchn_fasync(int fd, struct file *filp, int on)
	618	{
	619	struct per_user_data *u = filp->private_data;
	620	return fasync_helper(fd, filp, on, &u->evtchn_async_queue);
	621	}
	622
	623	static int evtchn_open(struct inode inode, struct file filp)
	624	{
	625	struct per_user_data *u;
	626
	627	u = kzalloc(sizeof(*u), GFP_KERNEL);
	628	if (u == NULL)
	629	return -ENOMEM;
	630
	631	u->name = kasprintf(GFP_KERNEL, "evtchn:%s", current->comm);
	632	if (u->name == NULL) {
	633	kfree(u);
	634	return -ENOMEM;
	635	}
	636
	637	init_waitqueue_head(&u->evtchn_wait);
	638
0a4666b5	639	mutex_init(&u->bind_mutex);
f7116284	640	mutex_init(&u->ring_cons_mutex);
73cc4bb0	641	spin_lock_init(&u->ring_prod_lock);
f7116284	642
fbc872c3 DV	643	u->restrict_domid = UNRESTRICTED_DOMID;
fbc872c3 DV	644
f7116284 IC	645	filp->private_data = u;
f7116284 IC	646
6eab04a8	647	return nonseekable_open(inode, filp);
f7116284 IC	648	}
	649
	650	static int evtchn_release(struct inode inode, struct file filp)
	651	{
f7116284	652	struct per_user_data *u = filp->private_data;
73cc4bb0	653	struct rb_node *node;
f7116284	654
73cc4bb0 DV	655	while ((node = u->evtchns.rb_node)) {
73cc4bb0 DV	656	struct user_evtchn *evtchn;
f7116284	657
73cc4bb0 DV	658	evtchn = rb_entry(node, struct user_evtchn, node);
	659	disable_irq(irq_from_evtchn(evtchn->port));
	660	evtchn_unbind_from_user(u, evtchn);
3f5e554f JF	661	}
3f5e554f JF	662
86200154	663	evtchn_free_ring(u->ring);
f7116284 IC	664	kfree(u->name);
	665	kfree(u);
	666
	667	return 0;
	668	}
	669
	670	static const struct file_operations evtchn_fops = {
	671	.owner = THIS_MODULE,
	672	.read = evtchn_read,
	673	.write = evtchn_write,
	674	.unlocked_ioctl = evtchn_ioctl,
	675	.poll = evtchn_poll,
	676	.fasync = evtchn_fasync,
	677	.open = evtchn_open,
	678	.release = evtchn_release,
bc7fc5e3	679	.llseek = no_llseek,
f7116284 IC	680	};
	681
	682	static struct miscdevice evtchn_miscdev = {
	683	.minor = MISC_DYNAMIC_MINOR,
376d908f	684	.name = "xen/evtchn",
f7116284 IC	685	.fops = &evtchn_fops,
	686	};
	687	static int __init evtchn_init(void)
	688	{
	689	int err;
	690
	691	if (!xen_domain())
	692	return -ENODEV;
	693
18283ea7	694	/* Create '/dev/xen/evtchn'. */
f7116284 IC	695	err = misc_register(&evtchn_miscdev);
f7116284 IC	696	if (err != 0) {
283c0972	697	pr_err("Could not register /dev/xen/evtchn\n");
f7116284 IC	698	return err;
	699	}
	700
283c0972	701	pr_info("Event-channel device installed\n");
f7116284 IC	702
	703	return 0;
	704	}
	705
	706	static void __exit evtchn_cleanup(void)
	707	{
	708	misc_deregister(&evtchn_miscdev);
	709	}
	710
	711	module_init(evtchn_init);
	712	module_exit(evtchn_cleanup);
	713
	714	MODULE_LICENSE("GPL");