[linux-2.6-block.git] / drivers / xen / xenbus / xenbus_dev_frontend.c

/*
 * Driver giving user-space access to the kernel's xenbus connection
 * to xenstore.
 *
 * Copyright (c) 2005, Christian Limpach
 * Copyright (c) 2005, Rusty Russell, IBM Corporation
 *
 * 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.
 *
 * Changes:
 * 2008-10-07  Alex Zeffertt    Replaced /proc/xen/xenbus with xenfs filesystem
 *                              and /proc/xen compatibility mount point.
 *                              Turned xenfs into a loadable module.
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/uio.h>
#include <linux/notifier.h>
#include <linux/wait.h>
#include <linux/fs.h>
#include <linux/poll.h>
#include <linux/mutex.h>
#include <linux/sched.h>
#include <linux/spinlock.h>
#include <linux/mount.h>
#include <linux/pagemap.h>
#include <linux/uaccess.h>
#include <linux/init.h>
#include <linux/namei.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/miscdevice.h>

#include <xen/xenbus.h>
#include <xen/xen.h>
#include <asm/xen/hypervisor.h>

#include "xenbus.h"

/*
 * An element of a list of outstanding transactions, for which we're
 * still waiting a reply.
 */
struct xenbus_transaction_holder {
	struct list_head list;
	struct xenbus_transaction handle;
};

/*
 * A buffer of data on the queue.
 */
struct read_buffer {
	struct list_head list;
	unsigned int cons;
	unsigned int len;
	char msg[];
};

struct xenbus_file_priv {
	/*
	 * msgbuffer_mutex is held while partial requests are built up
	 * and complete requests are acted on.  It therefore protects
	 * the "transactions" and "watches" lists, and the partial
	 * request length and buffer.
	 *
	 * reply_mutex protects the reply being built up to return to
	 * usermode.  It nests inside msgbuffer_mutex but may be held
	 * alone during a watch callback.
	 */
	struct mutex msgbuffer_mutex;

	/* In-progress transactions */
	struct list_head transactions;

	/* Active watches. */
	struct list_head watches;

	/* Partial request. */
	unsigned int len;
	union {
		struct xsd_sockmsg msg;
		char buffer[XENSTORE_PAYLOAD_MAX];
	} u;

	/* Response queue. */
	struct mutex reply_mutex;
	struct list_head read_buffers;
	wait_queue_head_t read_waitq;

	struct kref kref;
};

/* Read out any raw xenbus messages queued up. */
static ssize_t xenbus_file_read(struct file *filp,
			       char __user *ubuf,
			       size_t len, loff_t *ppos)
{
	struct xenbus_file_priv *u = filp->private_data;
	struct read_buffer *rb;
	unsigned i;
	int ret;

	mutex_lock(&u->reply_mutex);
again:
	while (list_empty(&u->read_buffers)) {
		mutex_unlock(&u->reply_mutex);
		if (filp->f_flags & O_NONBLOCK)
			return -EAGAIN;

		ret = wait_event_interruptible(u->read_waitq,
					       !list_empty(&u->read_buffers));
		if (ret)
			return ret;
		mutex_lock(&u->reply_mutex);
	}

	rb = list_entry(u->read_buffers.next, struct read_buffer, list);
	i = 0;
	while (i < len) {
		unsigned sz = min((unsigned)len - i, rb->len - rb->cons);

		ret = copy_to_user(ubuf + i, &rb->msg[rb->cons], sz);

		i += sz - ret;
		rb->cons += sz - ret;

		if (ret != 0) {
			if (i == 0)
				i = -EFAULT;
			goto out;
		}

		/* Clear out buffer if it has been consumed */
		if (rb->cons == rb->len) {
			list_del(&rb->list);
			kfree(rb);
			if (list_empty(&u->read_buffers))
				break;
			rb = list_entry(u->read_buffers.next,
					struct read_buffer, list);
		}
	}
	if (i == 0)
		goto again;

out:
	mutex_unlock(&u->reply_mutex);
	return i;
}

/*
 * Add a buffer to the queue.  Caller must hold the appropriate lock
 * if the queue is not local.  (Commonly the caller will build up
 * multiple queued buffers on a temporary local list, and then add it
 * to the appropriate list under lock once all the buffers have een
 * successfully allocated.)
 */
static int queue_reply(struct list_head *queue, const void *data, size_t len)
{
	struct read_buffer *rb;

	if (len == 0)
		return 0;
	if (len > XENSTORE_PAYLOAD_MAX)
		return -EINVAL;

	rb = kmalloc(sizeof(*rb) + len, GFP_KERNEL);
	if (rb == NULL)
		return -ENOMEM;

	rb->cons = 0;
	rb->len = len;

	memcpy(rb->msg, data, len);

	list_add_tail(&rb->list, queue);
	return 0;
}

/*
 * Free all the read_buffer s on a list.
 * Caller must have sole reference to list.
 */
static void queue_cleanup(struct list_head *list)
{
	struct read_buffer *rb;

	while (!list_empty(list)) {
		rb = list_entry(list->next, struct read_buffer, list);
		list_del(list->next);
		kfree(rb);
	}
}

struct watch_adapter {
	struct list_head list;
	struct xenbus_watch watch;
	struct xenbus_file_priv *dev_data;
	char *token;
};

static void free_watch_adapter(struct watch_adapter *watch)
{
	kfree(watch->watch.node);
	kfree(watch->token);
	kfree(watch);
}

static struct watch_adapter *alloc_watch_adapter(const char *path,
						 const char *token)
{
	struct watch_adapter *watch;

	watch = kzalloc(sizeof(*watch), GFP_KERNEL);
	if (watch == NULL)
		goto out_fail;

	watch->watch.node = kstrdup(path, GFP_KERNEL);
	if (watch->watch.node == NULL)
		goto out_free;

	watch->token = kstrdup(token, GFP_KERNEL);
	if (watch->token == NULL)
		goto out_free;

	return watch;

out_free:
	free_watch_adapter(watch);

out_fail:
	return NULL;
}

static void watch_fired(struct xenbus_watch *watch,
			const char *path,
			const char *token)
{
	struct watch_adapter *adap;
	struct xsd_sockmsg hdr;
	const char *token_caller;
	int path_len, tok_len, body_len;
	int ret;
	LIST_HEAD(staging_q);

	adap = container_of(watch, struct watch_adapter, watch);

	token_caller = adap->token;

	path_len = strlen(path) + 1;
	tok_len = strlen(token_caller) + 1;
	body_len = path_len + tok_len;

	hdr.type = XS_WATCH_EVENT;
	hdr.len = body_len;

	mutex_lock(&adap->dev_data->reply_mutex);

	ret = queue_reply(&staging_q, &hdr, sizeof(hdr));
	if (!ret)
		ret = queue_reply(&staging_q, path, path_len);
	if (!ret)
		ret = queue_reply(&staging_q, token_caller, tok_len);

	if (!ret) {
		/* success: pass reply list onto watcher */
		list_splice_tail(&staging_q, &adap->dev_data->read_buffers);
		wake_up(&adap->dev_data->read_waitq);
	} else
		queue_cleanup(&staging_q);

	mutex_unlock(&adap->dev_data->reply_mutex);
}

static void xenbus_file_free(struct kref *kref)
{
	struct xenbus_file_priv *u;
	struct xenbus_transaction_holder *trans, *tmp;
	struct watch_adapter *watch, *tmp_watch;
	struct read_buffer *rb, *tmp_rb;

	u = container_of(kref, struct xenbus_file_priv, kref);

	/*
	 * No need for locking here because there are no other users,
	 * by definition.
	 */

	list_for_each_entry_safe(trans, tmp, &u->transactions, list) {
		xenbus_transaction_end(trans->handle, 1);
		list_del(&trans->list);
		kfree(trans);
	}

	list_for_each_entry_safe(watch, tmp_watch, &u->watches, list) {
		unregister_xenbus_watch(&watch->watch);
		list_del(&watch->list);
		free_watch_adapter(watch);
	}

	list_for_each_entry_safe(rb, tmp_rb, &u->read_buffers, list) {
		list_del(&rb->list);
		kfree(rb);
	}
	kfree(u);
}

static struct xenbus_transaction_holder *xenbus_get_transaction(
	struct xenbus_file_priv *u, uint32_t tx_id)
{
	struct xenbus_transaction_holder *trans;

	list_for_each_entry(trans, &u->transactions, list)
		if (trans->handle.id == tx_id)
			return trans;

	return NULL;
}

void xenbus_dev_queue_reply(struct xb_req_data *req)
{
	struct xenbus_file_priv *u = req->par;
	struct xenbus_transaction_holder *trans = NULL;
	int rc;
	LIST_HEAD(staging_q);

	xs_request_exit(req);

	mutex_lock(&u->msgbuffer_mutex);

	if (req->type == XS_TRANSACTION_START) {
		trans = xenbus_get_transaction(u, 0);
		if (WARN_ON(!trans))
			goto out;
		if (req->msg.type == XS_ERROR) {
			list_del(&trans->list);
			kfree(trans);
		} else {
			rc = kstrtou32(req->body, 10, &trans->handle.id);
			if (WARN_ON(rc))
				goto out;
		}
	} else if (req->type == XS_TRANSACTION_END) {
		trans = xenbus_get_transaction(u, req->msg.tx_id);
		if (WARN_ON(!trans))
			goto out;
		list_del(&trans->list);
		kfree(trans);
	}

	mutex_unlock(&u->msgbuffer_mutex);

	mutex_lock(&u->reply_mutex);
	rc = queue_reply(&staging_q, &req->msg, sizeof(req->msg));
	if (!rc)
		rc = queue_reply(&staging_q, req->body, req->msg.len);
	if (!rc) {
		list_splice_tail(&staging_q, &u->read_buffers);
		wake_up(&u->read_waitq);
	} else {
		queue_cleanup(&staging_q);
	}
	mutex_unlock(&u->reply_mutex);

	kfree(req->body);
	kfree(req);

	kref_put(&u->kref, xenbus_file_free);

	return;

 out:
	mutex_unlock(&u->msgbuffer_mutex);
}

static int xenbus_command_reply(struct xenbus_file_priv *u,
				unsigned int msg_type, const char *reply)
{
	struct {
		struct xsd_sockmsg hdr;
		char body[16];
	} msg;
	int rc;

	msg.hdr = u->u.msg;
	msg.hdr.type = msg_type;
	msg.hdr.len = strlen(reply) + 1;
	if (msg.hdr.len > sizeof(msg.body))
		return -E2BIG;
	memcpy(&msg.body, reply, msg.hdr.len);

	mutex_lock(&u->reply_mutex);
	rc = queue_reply(&u->read_buffers, &msg, sizeof(msg.hdr) + msg.hdr.len);
	wake_up(&u->read_waitq);
	mutex_unlock(&u->reply_mutex);

	if (!rc)
		kref_put(&u->kref, xenbus_file_free);

	return rc;
}

static int xenbus_write_transaction(unsigned msg_type,
				    struct xenbus_file_priv *u)
{
	int rc;
	struct xenbus_transaction_holder *trans = NULL;
	struct {
		struct xsd_sockmsg hdr;
		char body[];
	} *msg = (void *)u->u.buffer;

	if (msg_type == XS_TRANSACTION_START) {
		trans = kzalloc(sizeof(*trans), GFP_KERNEL);
		if (!trans) {
			rc = -ENOMEM;
			goto out;
		}
		list_add(&trans->list, &u->transactions);
	} else if (msg->hdr.tx_id != 0 &&
		   !xenbus_get_transaction(u, msg->hdr.tx_id))
		return xenbus_command_reply(u, XS_ERROR, "ENOENT");
	else if (msg_type == XS_TRANSACTION_END &&
		 !(msg->hdr.len == 2 &&
		   (!strcmp(msg->body, "T") || !strcmp(msg->body, "F"))))
		return xenbus_command_reply(u, XS_ERROR, "EINVAL");

	rc = xenbus_dev_request_and_reply(&msg->hdr, u);
	if (rc && trans) {
		list_del(&trans->list);
		kfree(trans);
	}

out:
	return rc;
}

static int xenbus_write_watch(unsigned msg_type, struct xenbus_file_priv *u)
{
	struct watch_adapter *watch;
	char *path, *token;
	int err, rc;
	LIST_HEAD(staging_q);

	path = u->u.buffer + sizeof(u->u.msg);
	token = memchr(path, 0, u->u.msg.len);
	if (token == NULL) {
		rc = xenbus_command_reply(u, XS_ERROR, "EINVAL");
		goto out;
	}
	token++;
	if (memchr(token, 0, u->u.msg.len - (token - path)) == NULL) {
		rc = xenbus_command_reply(u, XS_ERROR, "EINVAL");
		goto out;
	}

	if (msg_type == XS_WATCH) {
		watch = alloc_watch_adapter(path, token);
		if (watch == NULL) {
			rc = -ENOMEM;
			goto out;
		}

		watch->watch.callback = watch_fired;
		watch->dev_data = u;

		err = register_xenbus_watch(&watch->watch);
		if (err) {
			free_watch_adapter(watch);
			rc = err;
			goto out;
		}
		list_add(&watch->list, &u->watches);
	} else {
		list_for_each_entry(watch, &u->watches, list) {
			if (!strcmp(watch->token, token) &&
			    !strcmp(watch->watch.node, path)) {
				unregister_xenbus_watch(&watch->watch);
				list_del(&watch->list);
				free_watch_adapter(watch);
				break;
			}
		}
	}

	/* Success.  Synthesize a reply to say all is OK. */
	rc = xenbus_command_reply(u, msg_type, "OK");

out:
	return rc;
}

static ssize_t xenbus_file_write(struct file *filp,
				const char __user *ubuf,
				size_t len, loff_t *ppos)
{
	struct xenbus_file_priv *u = filp->private_data;
	uint32_t msg_type;
	int rc = len;
	int ret;
	LIST_HEAD(staging_q);

	/*
	 * We're expecting usermode to be writing properly formed
	 * xenbus messages.  If they write an incomplete message we
	 * buffer it up.  Once it is complete, we act on it.
	 */

	/*
	 * Make sure concurrent writers can't stomp all over each
	 * other's messages and make a mess of our partial message
	 * buffer.  We don't make any attemppt to stop multiple
	 * writers from making a mess of each other's incomplete
	 * messages; we're just trying to guarantee our own internal
	 * consistency and make sure that single writes are handled
	 * atomically.
	 */
	mutex_lock(&u->msgbuffer_mutex);

	/* Get this out of the way early to avoid confusion */
	if (len == 0)
		goto out;

	/* Can't write a xenbus message larger we can buffer */
	if (len > sizeof(u->u.buffer) - u->len) {
		/* On error, dump existing buffer */
		u->len = 0;
		rc = -EINVAL;
		goto out;
	}

	ret = copy_from_user(u->u.buffer + u->len, ubuf, len);

	if (ret != 0) {
		rc = -EFAULT;
		goto out;
	}

	/* Deal with a partial copy. */
	len -= ret;
	rc = len;

	u->len += len;

	/* Return if we haven't got a full message yet */
	if (u->len < sizeof(u->u.msg))
		goto out;	/* not even the header yet */

	/* If we're expecting a message that's larger than we can
	   possibly send, dump what we have and return an error. */
	if ((sizeof(u->u.msg) + u->u.msg.len) > sizeof(u->u.buffer)) {
		rc = -E2BIG;
		u->len = 0;
		goto out;
	}

	if (u->len < (sizeof(u->u.msg) + u->u.msg.len))
		goto out;	/* incomplete data portion */

	/*
	 * OK, now we have a complete message.  Do something with it.
	 */

	kref_get(&u->kref);

	msg_type = u->u.msg.type;

	switch (msg_type) {
	case XS_WATCH:
	case XS_UNWATCH:
		/* (Un)Ask for some path to be watched for changes */
		ret = xenbus_write_watch(msg_type, u);
		break;

	default:
		/* Send out a transaction */
		ret = xenbus_write_transaction(msg_type, u);
		break;
	}
	if (ret != 0) {
		rc = ret;
		kref_put(&u->kref, xenbus_file_free);
	}

	/* Buffered message consumed */
	u->len = 0;

 out:
	mutex_unlock(&u->msgbuffer_mutex);
	return rc;
}

static int xenbus_file_open(struct inode *inode, struct file *filp)
{
	struct xenbus_file_priv *u;

	if (xen_store_evtchn == 0)
		return -ENOENT;

	nonseekable_open(inode, filp);

	filp->f_mode &= ~FMODE_ATOMIC_POS; /* cdev-style semantics */

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

	kref_init(&u->kref);

	INIT_LIST_HEAD(&u->transactions);
	INIT_LIST_HEAD(&u->watches);
	INIT_LIST_HEAD(&u->read_buffers);
	init_waitqueue_head(&u->read_waitq);

	mutex_init(&u->reply_mutex);
	mutex_init(&u->msgbuffer_mutex);

	filp->private_data = u;

	return 0;
}

static int xenbus_file_release(struct inode *inode, struct file *filp)
{
	struct xenbus_file_priv *u = filp->private_data;

	kref_put(&u->kref, xenbus_file_free);

	return 0;
}

static __poll_t xenbus_file_poll(struct file *file, poll_table *wait)
{
	struct xenbus_file_priv *u = file->private_data;

	poll_wait(file, &u->read_waitq, wait);
	if (!list_empty(&u->read_buffers))
		return EPOLLIN | EPOLLRDNORM;
	return 0;
}

const struct file_operations xen_xenbus_fops = {
	.read = xenbus_file_read,
	.write = xenbus_file_write,
	.open = xenbus_file_open,
	.release = xenbus_file_release,
	.poll = xenbus_file_poll,
	.llseek = no_llseek,
};
EXPORT_SYMBOL_GPL(xen_xenbus_fops);

static struct miscdevice xenbus_dev = {
	.minor = MISC_DYNAMIC_MINOR,
	.name = "xen/xenbus",
	.fops = &xen_xenbus_fops,
};

static int __init xenbus_init(void)
{
	int err;

	if (!xen_domain())
		return -ENODEV;

	err = misc_register(&xenbus_dev);
	if (err)
		pr_err("Could not register xenbus frontend device\n");
	return err;
}
device_initcall(xenbus_init);
Commit	Line	Data
1107ba88 AZ	1	/*
	2	* Driver giving user-space access to the kernel's xenbus connection
	3	* to xenstore.
	4	*
	5	* Copyright (c) 2005, Christian Limpach
	6	* Copyright (c) 2005, Rusty Russell, IBM Corporation
	7	*
	8	* This program is free software; you can redistribute it and/or
	9	* modify it under the terms of the GNU General Public License version 2
	10	* as published by the Free Software Foundation; or, when distributed
	11	* separately from the Linux kernel or incorporated into other
	12	* software packages, subject to the following license:
	13	*
	14	* Permission is hereby granted, free of charge, to any person obtaining a copy
	15	* of this source file (the "Software"), to deal in the Software without
	16	* restriction, including without limitation the rights to use, copy, modify,
	17	* merge, publish, distribute, sublicense, and/or sell copies of the Software,
	18	* and to permit persons to whom the Software is furnished to do so, subject to
	19	* the following conditions:
	20	*
	21	* The above copyright notice and this permission notice shall be included in
	22	* all copies or substantial portions of the Software.
	23	*
	24	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	25	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	26	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
	27	* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
	28	* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
	29	* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
	30	* IN THE SOFTWARE.
	31	*
	32	* Changes:
	33	* 2008-10-07 Alex Zeffertt Replaced /proc/xen/xenbus with xenfs filesystem
	34	* and /proc/xen compatibility mount point.
	35	* Turned xenfs into a loadable module.
	36	*/
	37
283c0972 JP	38	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
283c0972 JP	39
1107ba88 AZ	40	#include <linux/kernel.h>
	41	#include <linux/errno.h>
	42	#include <linux/uio.h>
	43	#include <linux/notifier.h>
	44	#include <linux/wait.h>
	45	#include <linux/fs.h>
	46	#include <linux/poll.h>
	47	#include <linux/mutex.h>
a99bbaf5	48	#include <linux/sched.h>
1107ba88 AZ	49	#include <linux/spinlock.h>
	50	#include <linux/mount.h>
	51	#include <linux/pagemap.h>
	52	#include <linux/uaccess.h>
	53	#include <linux/init.h>
	54	#include <linux/namei.h>
	55	#include <linux/string.h>
5a0e3ad6	56	#include <linux/slab.h>
2fb3683e	57	#include <linux/miscdevice.h>
1107ba88	58
1107ba88	59	#include <xen/xenbus.h>
b79d2ff9	60	#include <xen/xen.h>
1107ba88 AZ	61	#include <asm/xen/hypervisor.h>
1107ba88 AZ	62
332f791d JG	63	#include "xenbus.h"
332f791d JG	64
1107ba88 AZ	65	/*
	66	* An element of a list of outstanding transactions, for which we're
	67	* still waiting a reply.
	68	*/
	69	struct xenbus_transaction_holder {
	70	struct list_head list;
	71	struct xenbus_transaction handle;
	72	};
	73
	74	/*
	75	* A buffer of data on the queue.
	76	*/
	77	struct read_buffer {
	78	struct list_head list;
	79	unsigned int cons;
	80	unsigned int len;
	81	char msg[];
	82	};
	83
	84	struct xenbus_file_priv {
	85	/*
	86	* msgbuffer_mutex is held while partial requests are built up
	87	* and complete requests are acted on. It therefore protects
	88	* the "transactions" and "watches" lists, and the partial
	89	* request length and buffer.
	90	*
	91	* reply_mutex protects the reply being built up to return to
	92	* usermode. It nests inside msgbuffer_mutex but may be held
	93	* alone during a watch callback.
	94	*/
	95	struct mutex msgbuffer_mutex;
	96
	97	/* In-progress transactions */
	98	struct list_head transactions;
	99
	100	/* Active watches. */
	101	struct list_head watches;
	102
	103	/* Partial request. */
	104	unsigned int len;
	105	union {
	106	struct xsd_sockmsg msg;
50bf7379	107	char buffer[XENSTORE_PAYLOAD_MAX];
1107ba88 AZ	108	} u;
	109
	110	/* Response queue. */
	111	struct mutex reply_mutex;
	112	struct list_head read_buffers;
	113	wait_queue_head_t read_waitq;
	114
fd8aa909	115	struct kref kref;
1107ba88 AZ	116	};
	117
	118	/* Read out any raw xenbus messages queued up. */
	119	static ssize_t xenbus_file_read(struct file *filp,
	120	char __user *ubuf,
	121	size_t len, loff_t *ppos)
	122	{
	123	struct xenbus_file_priv *u = filp->private_data;
	124	struct read_buffer *rb;
	125	unsigned i;
	126	int ret;
	127
	128	mutex_lock(&u->reply_mutex);
7808121b	129	again:
1107ba88 AZ	130	while (list_empty(&u->read_buffers)) {
1107ba88 AZ	131	mutex_unlock(&u->reply_mutex);
6280f190 PB	132	if (filp->f_flags & O_NONBLOCK)
	133	return -EAGAIN;
	134
1107ba88 AZ	135	ret = wait_event_interruptible(u->read_waitq,
	136	!list_empty(&u->read_buffers));
	137	if (ret)
	138	return ret;
	139	mutex_lock(&u->reply_mutex);
	140	}
	141
	142	rb = list_entry(u->read_buffers.next, struct read_buffer, list);
	143	i = 0;
	144	while (i < len) {
	145	unsigned sz = min((unsigned)len - i, rb->len - rb->cons);
	146
	147	ret = copy_to_user(ubuf + i, &rb->msg[rb->cons], sz);
	148
	149	i += sz - ret;
	150	rb->cons += sz - ret;
	151
fb27cfbc	152	if (ret != 0) {
1107ba88 AZ	153	if (i == 0)
	154	i = -EFAULT;
	155	goto out;
	156	}
	157
	158	/* Clear out buffer if it has been consumed */
	159	if (rb->cons == rb->len) {
	160	list_del(&rb->list);
	161	kfree(rb);
	162	if (list_empty(&u->read_buffers))
	163	break;
	164	rb = list_entry(u->read_buffers.next,
	165	struct read_buffer, list);
	166	}
	167	}
7808121b DDG	168	if (i == 0)
7808121b DDG	169	goto again;
1107ba88 AZ	170
	171	out:
	172	mutex_unlock(&u->reply_mutex);
	173	return i;
	174	}
	175
	176	/*
	177	* Add a buffer to the queue. Caller must hold the appropriate lock
	178	* if the queue is not local. (Commonly the caller will build up
	179	* multiple queued buffers on a temporary local list, and then add it
	180	* to the appropriate list under lock once all the buffers have een
	181	* successfully allocated.)
	182	*/
	183	static int queue_reply(struct list_head queue, const void data, size_t len)
	184	{
	185	struct read_buffer *rb;
	186
	187	if (len == 0)
	188	return 0;
85c0a87c IY	189	if (len > XENSTORE_PAYLOAD_MAX)
85c0a87c IY	190	return -EINVAL;
1107ba88 AZ	191
	192	rb = kmalloc(sizeof(*rb) + len, GFP_KERNEL);
	193	if (rb == NULL)
	194	return -ENOMEM;
	195
	196	rb->cons = 0;
	197	rb->len = len;
	198
	199	memcpy(rb->msg, data, len);
	200
	201	list_add_tail(&rb->list, queue);
	202	return 0;
	203	}
	204
	205	/*
	206	* Free all the read_buffer s on a list.
	207	* Caller must have sole reference to list.
	208	*/
	209	static void queue_cleanup(struct list_head *list)
	210	{
	211	struct read_buffer *rb;
	212
	213	while (!list_empty(list)) {
	214	rb = list_entry(list->next, struct read_buffer, list);
	215	list_del(list->next);
	216	kfree(rb);
	217	}
	218	}
	219
	220	struct watch_adapter {
	221	struct list_head list;
	222	struct xenbus_watch watch;
	223	struct xenbus_file_priv *dev_data;
	224	char *token;
	225	};
	226
	227	static void free_watch_adapter(struct watch_adapter *watch)
	228	{
	229	kfree(watch->watch.node);
	230	kfree(watch->token);
	231	kfree(watch);
	232	}
	233
	234	static struct watch_adapter alloc_watch_adapter(const char path,
	235	const char *token)
	236	{
	237	struct watch_adapter *watch;
	238
	239	watch = kzalloc(sizeof(*watch), GFP_KERNEL);
	240	if (watch == NULL)
	241	goto out_fail;
	242
	243	watch->watch.node = kstrdup(path, GFP_KERNEL);
	244	if (watch->watch.node == NULL)
	245	goto out_free;
	246
	247	watch->token = kstrdup(token, GFP_KERNEL);
	248	if (watch->token == NULL)
	249	goto out_free;
	250
	251	return watch;
	252
	253	out_free:
	254	free_watch_adapter(watch);
255
256	out_fail:
257	return NULL;
258	}
259
260	static void watch_fired(struct xenbus_watch *watch,
5584ea25 JG	261	const char *path,
5584ea25 JG	262	const char *token)
1107ba88 AZ	263	{
	264	struct watch_adapter *adap;
	265	struct xsd_sockmsg hdr;
5584ea25 JG	266	const char *token_caller;
5584ea25 JG	267	int path_len, tok_len, body_len;
1107ba88 AZ	268	int ret;
	269	LIST_HEAD(staging_q);
	270
	271	adap = container_of(watch, struct watch_adapter, watch);
	272
5584ea25	273	token_caller = adap->token;
1107ba88 AZ	274
1107ba88 AZ	275	path_len = strlen(path) + 1;
5584ea25 JG	276	tok_len = strlen(token_caller) + 1;
5584ea25 JG	277	body_len = path_len + tok_len;
1107ba88 AZ	278
	279	hdr.type = XS_WATCH_EVENT;
	280	hdr.len = body_len;
	281
	282	mutex_lock(&adap->dev_data->reply_mutex);
	283
	284	ret = queue_reply(&staging_q, &hdr, sizeof(hdr));
	285	if (!ret)
	286	ret = queue_reply(&staging_q, path, path_len);
	287	if (!ret)
5584ea25	288	ret = queue_reply(&staging_q, token_caller, tok_len);
1107ba88 AZ	289
	290	if (!ret) {
	291	/* success: pass reply list onto watcher */
	292	list_splice_tail(&staging_q, &adap->dev_data->read_buffers);
	293	wake_up(&adap->dev_data->read_waitq);
	294	} else
	295	queue_cleanup(&staging_q);
	296
	297	mutex_unlock(&adap->dev_data->reply_mutex);
	298	}
	299
fd8aa909 JG	300	static void xenbus_file_free(struct kref *kref)
	301	{
	302	struct xenbus_file_priv *u;
	303	struct xenbus_transaction_holder trans, tmp;
	304	struct watch_adapter watch, tmp_watch;
	305	struct read_buffer rb, tmp_rb;
	306
	307	u = container_of(kref, struct xenbus_file_priv, kref);
	308
	309	/*
	310	* No need for locking here because there are no other users,
	311	* by definition.
	312	*/
	313
	314	list_for_each_entry_safe(trans, tmp, &u->transactions, list) {
	315	xenbus_transaction_end(trans->handle, 1);
	316	list_del(&trans->list);
	317	kfree(trans);
	318	}
	319
	320	list_for_each_entry_safe(watch, tmp_watch, &u->watches, list) {
	321	unregister_xenbus_watch(&watch->watch);
	322	list_del(&watch->list);
	323	free_watch_adapter(watch);
	324	}
	325
	326	list_for_each_entry_safe(rb, tmp_rb, &u->read_buffers, list) {
	327	list_del(&rb->list);
	328	kfree(rb);
	329	}
	330	kfree(u);
	331	}
	332
	333	static struct xenbus_transaction_holder *xenbus_get_transaction(
	334	struct xenbus_file_priv *u, uint32_t tx_id)
	335	{
	336	struct xenbus_transaction_holder *trans;
	337
	338	list_for_each_entry(trans, &u->transactions, list)
	339	if (trans->handle.id == tx_id)
	340	return trans;
	341
	342	return NULL;
	343	}
	344
	345	void xenbus_dev_queue_reply(struct xb_req_data *req)
	346	{
	347	struct xenbus_file_priv *u = req->par;
	348	struct xenbus_transaction_holder *trans = NULL;
	349	int rc;
	350	LIST_HEAD(staging_q);
	351
	352	xs_request_exit(req);
	353
	354	mutex_lock(&u->msgbuffer_mutex);
	355
	356	if (req->type == XS_TRANSACTION_START) {
	357	trans = xenbus_get_transaction(u, 0);
	358	if (WARN_ON(!trans))
	359	goto out;
	360	if (req->msg.type == XS_ERROR) {
	361	list_del(&trans->list);
	362	kfree(trans);
	363	} else {
364	rc = kstrtou32(req->body, 10, &trans->handle.id);
365	if (WARN_ON(rc))
366	goto out;
367	}
2a22ee6c	368	} else if (req->type == XS_TRANSACTION_END) {
fd8aa909 JG	369	trans = xenbus_get_transaction(u, req->msg.tx_id);
	370	if (WARN_ON(!trans))
	371	goto out;
	372	list_del(&trans->list);
	373	kfree(trans);
	374	}
	375
	376	mutex_unlock(&u->msgbuffer_mutex);
	377
	378	mutex_lock(&u->reply_mutex);
	379	rc = queue_reply(&staging_q, &req->msg, sizeof(req->msg));
	380	if (!rc)
	381	rc = queue_reply(&staging_q, req->body, req->msg.len);
	382	if (!rc) {
	383	list_splice_tail(&staging_q, &u->read_buffers);
	384	wake_up(&u->read_waitq);
	385	} else {
	386	queue_cleanup(&staging_q);
	387	}
	388	mutex_unlock(&u->reply_mutex);
	389
	390	kfree(req->body);
	391	kfree(req);
	392
	393	kref_put(&u->kref, xenbus_file_free);
	394
	395	return;
	396
	397	out:
	398	mutex_unlock(&u->msgbuffer_mutex);
	399	}
	400
9a6161fe JG	401	static int xenbus_command_reply(struct xenbus_file_priv *u,
	402	unsigned int msg_type, const char *reply)
	403	{
	404	struct {
	405	struct xsd_sockmsg hdr;
ebf04f33	406	char body[16];
9a6161fe JG	407	} msg;
	408	int rc;
	409
	410	msg.hdr = u->u.msg;
	411	msg.hdr.type = msg_type;
	412	msg.hdr.len = strlen(reply) + 1;
	413	if (msg.hdr.len > sizeof(msg.body))
	414	return -E2BIG;
ebf04f33	415	memcpy(&msg.body, reply, msg.hdr.len);
9a6161fe JG	416
	417	mutex_lock(&u->reply_mutex);
	418	rc = queue_reply(&u->read_buffers, &msg, sizeof(msg.hdr) + msg.hdr.len);
	419	wake_up(&u->read_waitq);
	420	mutex_unlock(&u->reply_mutex);
	421
fd8aa909 JG	422	if (!rc)
	423	kref_put(&u->kref, xenbus_file_free);
	424
9a6161fe JG	425	return rc;
	426	}
	427
1107ba88 AZ	428	static int xenbus_write_transaction(unsigned msg_type,
	429	struct xenbus_file_priv *u)
	430	{
e88a0faa	431	int rc;
1107ba88	432	struct xenbus_transaction_holder *trans = NULL;
8fe5ab41 SG	433	struct {
	434	struct xsd_sockmsg hdr;
	435	char body[];
	436	} msg = (void )u->u.buffer;
1107ba88 AZ	437
1107ba88 AZ	438	if (msg_type == XS_TRANSACTION_START) {
fd8aa909	439	trans = kzalloc(sizeof(*trans), GFP_KERNEL);
1107ba88 AZ	440	if (!trans) {
	441	rc = -ENOMEM;
	442	goto out;
	443	}
fd8aa909	444	list_add(&trans->list, &u->transactions);
8fe5ab41 SG	445	} else if (msg->hdr.tx_id != 0 &&
8fe5ab41 SG	446	!xenbus_get_transaction(u, msg->hdr.tx_id))
fd8aa909	447	return xenbus_command_reply(u, XS_ERROR, "ENOENT");
8fe5ab41 SG	448	else if (msg_type == XS_TRANSACTION_END &&
	449	!(msg->hdr.len == 2 &&
	450	(!strcmp(msg->body, "T") \|\| !strcmp(msg->body, "F"))))
	451	return xenbus_command_reply(u, XS_ERROR, "EINVAL");
1107ba88	452
8fe5ab41	453	rc = xenbus_dev_request_and_reply(&msg->hdr, u);
ac4cde39 JB	454	if (rc && trans) {
ac4cde39 JB	455	list_del(&trans->list);
1107ba88	456	kfree(trans);
ac4cde39	457	}
1107ba88 AZ	458
	459	out:
	460	return rc;
	461	}
	462
	463	static int xenbus_write_watch(unsigned msg_type, struct xenbus_file_priv *u)
	464	{
e1e5b3ff	465	struct watch_adapter *watch;
1107ba88 AZ	466	char path, token;
	467	int err, rc;
	468	LIST_HEAD(staging_q);
	469
	470	path = u->u.buffer + sizeof(u->u.msg);
	471	token = memchr(path, 0, u->u.msg.len);
	472	if (token == NULL) {
9a6161fe	473	rc = xenbus_command_reply(u, XS_ERROR, "EINVAL");
1107ba88 AZ	474	goto out;
	475	}
	476	token++;
a43a5ccd	477	if (memchr(token, 0, u->u.msg.len - (token - path)) == NULL) {
9a6161fe	478	rc = xenbus_command_reply(u, XS_ERROR, "EINVAL");
a43a5ccd JB	479	goto out;
a43a5ccd JB	480	}
1107ba88 AZ	481
	482	if (msg_type == XS_WATCH) {
	483	watch = alloc_watch_adapter(path, token);
	484	if (watch == NULL) {
	485	rc = -ENOMEM;
	486	goto out;
	487	}
	488
	489	watch->watch.callback = watch_fired;
	490	watch->dev_data = u;
	491
	492	err = register_xenbus_watch(&watch->watch);
	493	if (err) {
	494	free_watch_adapter(watch);
	495	rc = err;
	496	goto out;
	497	}
	498	list_add(&watch->list, &u->watches);
	499	} else {
e1e5b3ff	500	list_for_each_entry(watch, &u->watches, list) {
1107ba88 AZ	501	if (!strcmp(watch->token, token) &&
	502	!strcmp(watch->watch.node, path)) {
	503	unregister_xenbus_watch(&watch->watch);
	504	list_del(&watch->list);
	505	free_watch_adapter(watch);
	506	break;
	507	}
	508	}
	509	}
	510
	511	/* Success. Synthesize a reply to say all is OK. */
9a6161fe	512	rc = xenbus_command_reply(u, msg_type, "OK");
1107ba88 AZ	513
	514	out:
	515	return rc;
	516	}
	517
	518	static ssize_t xenbus_file_write(struct file *filp,
	519	const char __user *ubuf,
	520	size_t len, loff_t *ppos)
	521	{
	522	struct xenbus_file_priv *u = filp->private_data;
	523	uint32_t msg_type;
	524	int rc = len;
	525	int ret;
	526	LIST_HEAD(staging_q);
	527
	528	/*
	529	* We're expecting usermode to be writing properly formed
	530	* xenbus messages. If they write an incomplete message we
	531	* buffer it up. Once it is complete, we act on it.
	532	*/
	533
	534	/*
	535	* Make sure concurrent writers can't stomp all over each
	536	* other's messages and make a mess of our partial message
	537	* buffer. We don't make any attemppt to stop multiple
	538	* writers from making a mess of each other's incomplete
	539	* messages; we're just trying to guarantee our own internal
	540	* consistency and make sure that single writes are handled
	541	* atomically.
	542	*/
	543	mutex_lock(&u->msgbuffer_mutex);
	544
	545	/* Get this out of the way early to avoid confusion */
	546	if (len == 0)
	547	goto out;
	548
	549	/* Can't write a xenbus message larger we can buffer */
1bcaba51	550	if (len > sizeof(u->u.buffer) - u->len) {
1107ba88 AZ	551	/* On error, dump existing buffer */
	552	u->len = 0;
	553	rc = -EINVAL;
	554	goto out;
	555	}
	556
	557	ret = copy_from_user(u->u.buffer + u->len, ubuf, len);
	558
fb27cfbc	559	if (ret != 0) {
1107ba88 AZ	560	rc = -EFAULT;
	561	goto out;
	562	}
	563
	564	/* Deal with a partial copy. */
	565	len -= ret;
	566	rc = len;
	567
	568	u->len += len;
	569
	570	/* Return if we haven't got a full message yet */
	571	if (u->len < sizeof(u->u.msg))
	572	goto out; /* not even the header yet */
	573
	574	/* If we're expecting a message that's larger than we can
	575	possibly send, dump what we have and return an error. */
	576	if ((sizeof(u->u.msg) + u->u.msg.len) > sizeof(u->u.buffer)) {
	577	rc = -E2BIG;
	578	u->len = 0;
	579	goto out;
	580	}
	581
	582	if (u->len < (sizeof(u->u.msg) + u->u.msg.len))
	583	goto out; /* incomplete data portion */
	584
	585	/*
	586	* OK, now we have a complete message. Do something with it.
	587	*/
	588
fd8aa909 JG	589	kref_get(&u->kref);
fd8aa909 JG	590
1107ba88 AZ	591	msg_type = u->u.msg.type;
	592
	593	switch (msg_type) {
1107ba88 AZ	594	case XS_WATCH:
	595	case XS_UNWATCH:
	596	/* (Un)Ask for some path to be watched for changes */
	597	ret = xenbus_write_watch(msg_type, u);
	598	break;
	599
	600	default:
6d6df2e4 DO	601	/* Send out a transaction */
6d6df2e4 DO	602	ret = xenbus_write_transaction(msg_type, u);
1107ba88 AZ	603	break;
1107ba88 AZ	604	}
fd8aa909	605	if (ret != 0) {
1107ba88	606	rc = ret;
fd8aa909 JG	607	kref_put(&u->kref, xenbus_file_free);
fd8aa909 JG	608	}
1107ba88 AZ	609
	610	/* Buffered message consumed */
	611	u->len = 0;
	612
	613	out:
	614	mutex_unlock(&u->msgbuffer_mutex);
	615	return rc;
	616	}
	617
	618	static int xenbus_file_open(struct inode inode, struct file filp)
	619	{
	620	struct xenbus_file_priv *u;
	621
	622	if (xen_store_evtchn == 0)
	623	return -ENOENT;
	624
	625	nonseekable_open(inode, filp);
	626
581d21a2 DV	627	filp->f_mode &= ~FMODE_ATOMIC_POS; /* cdev-style semantics */
581d21a2 DV	628
1107ba88 AZ	629	u = kzalloc(sizeof(*u), GFP_KERNEL);
	630	if (u == NULL)
	631	return -ENOMEM;
	632
fd8aa909 JG	633	kref_init(&u->kref);
fd8aa909 JG	634
1107ba88 AZ	635	INIT_LIST_HEAD(&u->transactions);
	636	INIT_LIST_HEAD(&u->watches);
	637	INIT_LIST_HEAD(&u->read_buffers);
	638	init_waitqueue_head(&u->read_waitq);
	639
	640	mutex_init(&u->reply_mutex);
	641	mutex_init(&u->msgbuffer_mutex);
	642
	643	filp->private_data = u;
	644
	645	return 0;
	646	}
	647
	648	static int xenbus_file_release(struct inode inode, struct file filp)
	649	{
	650	struct xenbus_file_priv *u = filp->private_data;
1107ba88	651
fd8aa909	652	kref_put(&u->kref, xenbus_file_free);
1107ba88 AZ	653
	654	return 0;
	655	}
	656
afc9a42b	657	static __poll_t xenbus_file_poll(struct file file, poll_table wait)
1107ba88 AZ	658	{
	659	struct xenbus_file_priv *u = file->private_data;
	660
	661	poll_wait(file, &u->read_waitq, wait);
	662	if (!list_empty(&u->read_buffers))
a9a08845	663	return EPOLLIN \| EPOLLRDNORM;
1107ba88 AZ	664	return 0;
	665	}
	666
2fb3683e	667	const struct file_operations xen_xenbus_fops = {
1107ba88 AZ	668	.read = xenbus_file_read,
	669	.write = xenbus_file_write,
	670	.open = xenbus_file_open,
	671	.release = xenbus_file_release,
	672	.poll = xenbus_file_poll,
6038f373	673	.llseek = no_llseek,
1107ba88	674	};
2fb3683e BB	675	EXPORT_SYMBOL_GPL(xen_xenbus_fops);
	676
	677	static struct miscdevice xenbus_dev = {
	678	.minor = MISC_DYNAMIC_MINOR,
	679	.name = "xen/xenbus",
	680	.fops = &xen_xenbus_fops,
	681	};
	682
	683	static int __init xenbus_init(void)
	684	{
	685	int err;
	686
	687	if (!xen_domain())
	688	return -ENODEV;
	689
	690	err = misc_register(&xenbus_dev);
	691	if (err)
283c0972	692	pr_err("Could not register xenbus frontend device\n");
2fb3683e BB	693	return err;
2fb3683e BB	694	}
ab1241a1	695	device_initcall(xenbus_init);