[net-next-2.6.git] / drivers / block / nbd.c

/*
 * Network block device - make block devices work over TCP
 *
 * Note that you can not swap over this thing, yet. Seems to work but
 * deadlocks sometimes - you can not swap over TCP in general.
 * 
 * Copyright 1997-2000 Pavel Machek <pavel@ucw.cz>
 * Parts copyright 2001 Steven Whitehouse <steve@chygwyn.com>
 *
 * This file is released under GPLv2 or later.
 *
 * (part of code stolen from loop.c)
 */

#include <linux/major.h>

#include <linux/blkdev.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/fs.h>
#include <linux/bio.h>
#include <linux/stat.h>
#include <linux/errno.h>
#include <linux/file.h>
#include <linux/ioctl.h>
#include <linux/compiler.h>
#include <linux/err.h>
#include <linux/kernel.h>
#include <net/sock.h>

#include <asm/uaccess.h>
#include <asm/system.h>
#include <asm/types.h>

#include <linux/nbd.h>

#define LO_MAGIC 0x68797548

#ifdef NDEBUG
#define dprintk(flags, fmt...)
#else /* NDEBUG */
#define dprintk(flags, fmt...) do { \
	if (debugflags & (flags)) printk(KERN_DEBUG fmt); \
} while (0)
#define DBG_IOCTL       0x0004
#define DBG_INIT        0x0010
#define DBG_EXIT        0x0020
#define DBG_BLKDEV      0x0100
#define DBG_RX          0x0200
#define DBG_TX          0x0400
static unsigned int debugflags;
#endif /* NDEBUG */

static unsigned int nbds_max = 16;
static struct nbd_device nbd_dev[MAX_NBD];

/*
 * Use just one lock (or at most 1 per NIC). Two arguments for this:
 * 1. Each NIC is essentially a synchronization point for all servers
 *    accessed through that NIC so there's no need to have more locks
 *    than NICs anyway.
 * 2. More locks lead to more "Dirty cache line bouncing" which will slow
 *    down each lock to the point where they're actually slower than just
 *    a single lock.
 * Thanks go to Jens Axboe and Al Viro for their LKML emails explaining this!
 */
static DEFINE_SPINLOCK(nbd_lock);

#ifndef NDEBUG
static const char *ioctl_cmd_to_ascii(int cmd)
{
	switch (cmd) {
	case NBD_SET_SOCK: return "set-sock";
	case NBD_SET_BLKSIZE: return "set-blksize";
	case NBD_SET_SIZE: return "set-size";
	case NBD_DO_IT: return "do-it";
	case NBD_CLEAR_SOCK: return "clear-sock";
	case NBD_CLEAR_QUE: return "clear-que";
	case NBD_PRINT_DEBUG: return "print-debug";
	case NBD_SET_SIZE_BLOCKS: return "set-size-blocks";
	case NBD_DISCONNECT: return "disconnect";
	case BLKROSET: return "set-read-only";
	case BLKFLSBUF: return "flush-buffer-cache";
	}
	return "unknown";
}

static const char *nbdcmd_to_ascii(int cmd)
{
	switch (cmd) {
	case  NBD_CMD_READ: return "read";
	case NBD_CMD_WRITE: return "write";
	case  NBD_CMD_DISC: return "disconnect";
	}
	return "invalid";
}
#endif /* NDEBUG */

static void nbd_end_request(struct request *req)
{
	int uptodate = (req->errors == 0) ? 1 : 0;
	struct request_queue *q = req->q;
	unsigned long flags;

	dprintk(DBG_BLKDEV, "%s: request %p: %s\n", req->rq_disk->disk_name,
			req, uptodate? "done": "failed");

	spin_lock_irqsave(q->queue_lock, flags);
	if (!end_that_request_first(req, uptodate, req->nr_sectors)) {
		end_that_request_last(req, uptodate);
	}
	spin_unlock_irqrestore(q->queue_lock, flags);
}

/*
 *  Send or receive packet.
 */
static int sock_xmit(struct socket *sock, int send, void *buf, int size,
		int msg_flags)
{
	int result;
	struct msghdr msg;
	struct kvec iov;
	sigset_t blocked, oldset;

	/* Allow interception of SIGKILL only
	 * Don't allow other signals to interrupt the transmission */
	siginitsetinv(&blocked, sigmask(SIGKILL));
	sigprocmask(SIG_SETMASK, &blocked, &oldset);

	do {
		sock->sk->sk_allocation = GFP_NOIO;
		iov.iov_base = buf;
		iov.iov_len = size;
		msg.msg_name = NULL;
		msg.msg_namelen = 0;
		msg.msg_control = NULL;
		msg.msg_controllen = 0;
		msg.msg_flags = msg_flags | MSG_NOSIGNAL;

		if (send)
			result = kernel_sendmsg(sock, &msg, &iov, 1, size);
		else
			result = kernel_recvmsg(sock, &msg, &iov, 1, size, 0);

		if (signal_pending(current)) {
			siginfo_t info;
			printk(KERN_WARNING "nbd (pid %d: %s) got signal %d\n",
				current->pid, current->comm,
				dequeue_signal_lock(current, &current->blocked, &info));
			result = -EINTR;
			break;
		}

		if (result <= 0) {
			if (result == 0)
				result = -EPIPE; /* short read */
			break;
		}
		size -= result;
		buf += result;
	} while (size > 0);

	sigprocmask(SIG_SETMASK, &oldset, NULL);

	return result;
}

static inline int sock_send_bvec(struct socket *sock, struct bio_vec *bvec,
		int flags)
{
	int result;
	void *kaddr = kmap(bvec->bv_page);
	result = sock_xmit(sock, 1, kaddr + bvec->bv_offset, bvec->bv_len,
			flags);
	kunmap(bvec->bv_page);
	return result;
}

static int nbd_send_req(struct nbd_device *lo, struct request *req)
{
	int result, flags;
	struct nbd_request request;
	unsigned long size = req->nr_sectors << 9;
	struct socket *sock = lo->sock;

	request.magic = htonl(NBD_REQUEST_MAGIC);
	request.type = htonl(nbd_cmd(req));
	request.from = cpu_to_be64((u64) req->sector << 9);
	request.len = htonl(size);
	memcpy(request.handle, &req, sizeof(req));

	dprintk(DBG_TX, "%s: request %p: sending control (%s@%llu,%luB)\n",
			lo->disk->disk_name, req,
			nbdcmd_to_ascii(nbd_cmd(req)),
			(unsigned long long)req->sector << 9,
			req->nr_sectors << 9);
	result = sock_xmit(sock, 1, &request, sizeof(request),
			(nbd_cmd(req) == NBD_CMD_WRITE)? MSG_MORE: 0);
	if (result <= 0) {
		printk(KERN_ERR "%s: Send control failed (result %d)\n",
				lo->disk->disk_name, result);
		goto error_out;
	}

	if (nbd_cmd(req) == NBD_CMD_WRITE) {
		struct req_iterator iter;
		struct bio_vec *bvec;
		/*
		 * we are really probing at internals to determine
		 * whether to set MSG_MORE or not...
		 */
		rq_for_each_segment(bvec, req, iter) {
				flags = 0;
				if (!rq_iter_last(req, iter))
					flags = MSG_MORE;
				dprintk(DBG_TX, "%s: request %p: sending %d bytes data\n",
						lo->disk->disk_name, req,
						bvec->bv_len);
				result = sock_send_bvec(sock, bvec, flags);
				if (result <= 0) {
					printk(KERN_ERR "%s: Send data failed (result %d)\n",
							lo->disk->disk_name,
							result);
					goto error_out;
				}
		}
	}
	return 0;

error_out:
	return 1;
}

static struct request *nbd_find_request(struct nbd_device *lo, char *handle)
{
	struct request *req;
	struct list_head *tmp;
	struct request *xreq;
	int err;

	memcpy(&xreq, handle, sizeof(xreq));

	err = wait_event_interruptible(lo->active_wq, lo->active_req != xreq);
	if (unlikely(err))
		goto out;

	spin_lock(&lo->queue_lock);
	list_for_each(tmp, &lo->queue_head) {
		req = list_entry(tmp, struct request, queuelist);
		if (req != xreq)
			continue;
		list_del_init(&req->queuelist);
		spin_unlock(&lo->queue_lock);
		return req;
	}
	spin_unlock(&lo->queue_lock);

	err = -ENOENT;

out:
	return ERR_PTR(err);
}

static inline int sock_recv_bvec(struct socket *sock, struct bio_vec *bvec)
{
	int result;
	void *kaddr = kmap(bvec->bv_page);
	result = sock_xmit(sock, 0, kaddr + bvec->bv_offset, bvec->bv_len,
			MSG_WAITALL);
	kunmap(bvec->bv_page);
	return result;
}

/* NULL returned = something went wrong, inform userspace */
static struct request *nbd_read_stat(struct nbd_device *lo)
{
	int result;
	struct nbd_reply reply;
	struct request *req;
	struct socket *sock = lo->sock;

	reply.magic = 0;
	result = sock_xmit(sock, 0, &reply, sizeof(reply), MSG_WAITALL);
	if (result <= 0) {
		printk(KERN_ERR "%s: Receive control failed (result %d)\n",
				lo->disk->disk_name, result);
		goto harderror;
	}

	if (ntohl(reply.magic) != NBD_REPLY_MAGIC) {
		printk(KERN_ERR "%s: Wrong magic (0x%lx)\n",
				lo->disk->disk_name,
				(unsigned long)ntohl(reply.magic));
		result = -EPROTO;
		goto harderror;
	}

	req = nbd_find_request(lo, reply.handle);
	if (unlikely(IS_ERR(req))) {
		result = PTR_ERR(req);
		if (result != -ENOENT)
			goto harderror;

		printk(KERN_ERR "%s: Unexpected reply (%p)\n",
				lo->disk->disk_name, reply.handle);
		result = -EBADR;
		goto harderror;
	}

	if (ntohl(reply.error)) {
		printk(KERN_ERR "%s: Other side returned error (%d)\n",
				lo->disk->disk_name, ntohl(reply.error));
		req->errors++;
		return req;
	}

	dprintk(DBG_RX, "%s: request %p: got reply\n",
			lo->disk->disk_name, req);
	if (nbd_cmd(req) == NBD_CMD_READ) {
		struct req_iterator iter;
		struct bio_vec *bvec;

		rq_for_each_segment(bvec, req, iter) {
				result = sock_recv_bvec(sock, bvec);
				if (result <= 0) {
					printk(KERN_ERR "%s: Receive data failed (result %d)\n",
							lo->disk->disk_name,
							result);
					req->errors++;
					return req;
				}
				dprintk(DBG_RX, "%s: request %p: got %d bytes data\n",
					lo->disk->disk_name, req, bvec->bv_len);
		}
	}
	return req;
harderror:
	lo->harderror = result;
	return NULL;
}

static ssize_t pid_show(struct gendisk *disk, char *page)
{
	return sprintf(page, "%ld\n",
		(long) ((struct nbd_device *)disk->private_data)->pid);
}

static struct disk_attribute pid_attr = {
	.attr = { .name = "pid", .mode = S_IRUGO },
	.show = pid_show,
};

static int nbd_do_it(struct nbd_device *lo)
{
	struct request *req;
	int ret;

	BUG_ON(lo->magic != LO_MAGIC);

	lo->pid = current->pid;
	ret = sysfs_create_file(&lo->disk->kobj, &pid_attr.attr);
	if (ret) {
		printk(KERN_ERR "nbd: sysfs_create_file failed!");
		return ret;
	}

	while ((req = nbd_read_stat(lo)) != NULL)
		nbd_end_request(req);

	sysfs_remove_file(&lo->disk->kobj, &pid_attr.attr);
	return 0;
}

static void nbd_clear_que(struct nbd_device *lo)
{
	struct request *req;

	BUG_ON(lo->magic != LO_MAGIC);

	/*
	 * Because we have set lo->sock to NULL under the tx_lock, all
	 * modifications to the list must have completed by now.  For
	 * the same reason, the active_req must be NULL.
	 *
	 * As a consequence, we don't need to take the spin lock while
	 * purging the list here.
	 */
	BUG_ON(lo->sock);
	BUG_ON(lo->active_req);

	while (!list_empty(&lo->queue_head)) {
		req = list_entry(lo->queue_head.next, struct request,
				 queuelist);
		list_del_init(&req->queuelist);
		req->errors++;
		nbd_end_request(req);
	}
}

/*
 * We always wait for result of write, for now. It would be nice to make it optional
 * in future
 * if ((rq_data_dir(req) == WRITE) && (lo->flags & NBD_WRITE_NOCHK))
 *   { printk( "Warning: Ignoring result!\n"); nbd_end_request( req ); }
 */

static void do_nbd_request(struct request_queue * q)
{
	struct request *req;
	
	while ((req = elv_next_request(q)) != NULL) {
		struct nbd_device *lo;

		blkdev_dequeue_request(req);
		dprintk(DBG_BLKDEV, "%s: request %p: dequeued (flags=%x)\n",
				req->rq_disk->disk_name, req, req->cmd_type);

		if (!blk_fs_request(req))
			goto error_out;

		lo = req->rq_disk->private_data;

		BUG_ON(lo->magic != LO_MAGIC);

		nbd_cmd(req) = NBD_CMD_READ;
		if (rq_data_dir(req) == WRITE) {
			nbd_cmd(req) = NBD_CMD_WRITE;
			if (lo->flags & NBD_READ_ONLY) {
				printk(KERN_ERR "%s: Write on read-only\n",
						lo->disk->disk_name);
				goto error_out;
			}
		}

		req->errors = 0;
		spin_unlock_irq(q->queue_lock);

		mutex_lock(&lo->tx_lock);
		if (unlikely(!lo->sock)) {
			mutex_unlock(&lo->tx_lock);
			printk(KERN_ERR "%s: Attempted send on closed socket\n",
			       lo->disk->disk_name);
			req->errors++;
			nbd_end_request(req);
			spin_lock_irq(q->queue_lock);
			continue;
		}

		lo->active_req = req;

		if (nbd_send_req(lo, req) != 0) {
			printk(KERN_ERR "%s: Request send failed\n",
					lo->disk->disk_name);
			req->errors++;
			nbd_end_request(req);
		} else {
			spin_lock(&lo->queue_lock);
			list_add(&req->queuelist, &lo->queue_head);
			spin_unlock(&lo->queue_lock);
		}

		lo->active_req = NULL;
		mutex_unlock(&lo->tx_lock);
		wake_up_all(&lo->active_wq);

		spin_lock_irq(q->queue_lock);
		continue;

error_out:
		req->errors++;
		spin_unlock(q->queue_lock);
		nbd_end_request(req);
		spin_lock(q->queue_lock);
	}
	return;
}

static int nbd_ioctl(struct inode *inode, struct file *file,
		     unsigned int cmd, unsigned long arg)
{
	struct nbd_device *lo = inode->i_bdev->bd_disk->private_data;
	int error;
	struct request sreq ;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	BUG_ON(lo->magic != LO_MAGIC);

	/* Anyone capable of this syscall can do *real bad* things */
	dprintk(DBG_IOCTL, "%s: nbd_ioctl cmd=%s(0x%x) arg=%lu\n",
			lo->disk->disk_name, ioctl_cmd_to_ascii(cmd), cmd, arg);

	switch (cmd) {
	case NBD_DISCONNECT:
	        printk(KERN_INFO "%s: NBD_DISCONNECT\n", lo->disk->disk_name);
		sreq.cmd_type = REQ_TYPE_SPECIAL;
		nbd_cmd(&sreq) = NBD_CMD_DISC;
		/*
		 * Set these to sane values in case server implementation
		 * fails to check the request type first and also to keep
		 * debugging output cleaner.
		 */
		sreq.sector = 0;
		sreq.nr_sectors = 0;
                if (!lo->sock)
			return -EINVAL;
                nbd_send_req(lo, &sreq);
                return 0;
 
	case NBD_CLEAR_SOCK:
		error = 0;
		mutex_lock(&lo->tx_lock);
		lo->sock = NULL;
		mutex_unlock(&lo->tx_lock);
		file = lo->file;
		lo->file = NULL;
		nbd_clear_que(lo);
		BUG_ON(!list_empty(&lo->queue_head));
		if (file)
			fput(file);
		return error;
	case NBD_SET_SOCK:
		if (lo->file)
			return -EBUSY;
		error = -EINVAL;
		file = fget(arg);
		if (file) {
			inode = file->f_path.dentry->d_inode;
			if (S_ISSOCK(inode->i_mode)) {
				lo->file = file;
				lo->sock = SOCKET_I(inode);
				error = 0;
			} else {
				fput(file);
			}
		}
		return error;
	case NBD_SET_BLKSIZE:
		lo->blksize = arg;
		lo->bytesize &= ~(lo->blksize-1);
		inode->i_bdev->bd_inode->i_size = lo->bytesize;
		set_blocksize(inode->i_bdev, lo->blksize);
		set_capacity(lo->disk, lo->bytesize >> 9);
		return 0;
	case NBD_SET_SIZE:
		lo->bytesize = arg & ~(lo->blksize-1);
		inode->i_bdev->bd_inode->i_size = lo->bytesize;
		set_blocksize(inode->i_bdev, lo->blksize);
		set_capacity(lo->disk, lo->bytesize >> 9);
		return 0;
	case NBD_SET_SIZE_BLOCKS:
		lo->bytesize = ((u64) arg) * lo->blksize;
		inode->i_bdev->bd_inode->i_size = lo->bytesize;
		set_blocksize(inode->i_bdev, lo->blksize);
		set_capacity(lo->disk, lo->bytesize >> 9);
		return 0;
	case NBD_DO_IT:
		if (!lo->file)
			return -EINVAL;
		error = nbd_do_it(lo);
		if (error)
			return error;
		/* on return tidy up in case we have a signal */
		/* Forcibly shutdown the socket causing all listeners
		 * to error
		 *
		 * FIXME: This code is duplicated from sys_shutdown, but
		 * there should be a more generic interface rather than
		 * calling socket ops directly here */
		mutex_lock(&lo->tx_lock);
		if (lo->sock) {
			printk(KERN_WARNING "%s: shutting down socket\n",
				lo->disk->disk_name);
			lo->sock->ops->shutdown(lo->sock,
				SEND_SHUTDOWN|RCV_SHUTDOWN);
			lo->sock = NULL;
		}
		mutex_unlock(&lo->tx_lock);
		file = lo->file;
		lo->file = NULL;
		nbd_clear_que(lo);
		printk(KERN_WARNING "%s: queue cleared\n", lo->disk->disk_name);
		if (file)
			fput(file);
		return lo->harderror;
	case NBD_CLEAR_QUE:
		/*
		 * This is for compatibility only.  The queue is always cleared
		 * by NBD_DO_IT or NBD_CLEAR_SOCK.
		 */
		BUG_ON(!lo->sock && !list_empty(&lo->queue_head));
		return 0;
	case NBD_PRINT_DEBUG:
		printk(KERN_INFO "%s: next = %p, prev = %p, head = %p\n",
			inode->i_bdev->bd_disk->disk_name,
			lo->queue_head.next, lo->queue_head.prev,
			&lo->queue_head);
		return 0;
	}
	return -EINVAL;
}

static struct block_device_operations nbd_fops =
{
	.owner =	THIS_MODULE,
	.ioctl =	nbd_ioctl,
};

/*
 * And here should be modules and kernel interface 
 *  (Just smiley confuses emacs :-)
 */

static int __init nbd_init(void)
{
	int err = -ENOMEM;
	int i;

	BUILD_BUG_ON(sizeof(struct nbd_request) != 28);

	if (nbds_max > MAX_NBD) {
		printk(KERN_CRIT "nbd: cannot allocate more than %u nbds; %u requested.\n", MAX_NBD,
				nbds_max);
		return -EINVAL;
	}

	for (i = 0; i < nbds_max; i++) {
		struct gendisk *disk = alloc_disk(1);
		if (!disk)
			goto out;
		nbd_dev[i].disk = disk;
		/*
		 * The new linux 2.5 block layer implementation requires
		 * every gendisk to have its very own request_queue struct.
		 * These structs are big so we dynamically allocate them.
		 */
		disk->queue = blk_init_queue(do_nbd_request, &nbd_lock);
		if (!disk->queue) {
			put_disk(disk);
			goto out;
		}
	}

	if (register_blkdev(NBD_MAJOR, "nbd")) {
		err = -EIO;
		goto out;
	}

	printk(KERN_INFO "nbd: registered device at major %d\n", NBD_MAJOR);
	dprintk(DBG_INIT, "nbd: debugflags=0x%x\n", debugflags);

	for (i = 0; i < nbds_max; i++) {
		struct gendisk *disk = nbd_dev[i].disk;
		nbd_dev[i].file = NULL;
		nbd_dev[i].magic = LO_MAGIC;
		nbd_dev[i].flags = 0;
		spin_lock_init(&nbd_dev[i].queue_lock);
		INIT_LIST_HEAD(&nbd_dev[i].queue_head);
		mutex_init(&nbd_dev[i].tx_lock);
		init_waitqueue_head(&nbd_dev[i].active_wq);
		nbd_dev[i].blksize = 1024;
		nbd_dev[i].bytesize = 0x7ffffc00ULL << 10; /* 2TB */
		disk->major = NBD_MAJOR;
		disk->first_minor = i;
		disk->fops = &nbd_fops;
		disk->private_data = &nbd_dev[i];
		disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
		sprintf(disk->disk_name, "nbd%d", i);
		set_capacity(disk, 0x7ffffc00ULL << 1); /* 2 TB */
		add_disk(disk);
	}

	return 0;
out:
	while (i--) {
		blk_cleanup_queue(nbd_dev[i].disk->queue);
		put_disk(nbd_dev[i].disk);
	}
	return err;
}

static void __exit nbd_cleanup(void)
{
	int i;
	for (i = 0; i < nbds_max; i++) {
		struct gendisk *disk = nbd_dev[i].disk;
		nbd_dev[i].magic = 0;
		if (disk) {
			del_gendisk(disk);
			blk_cleanup_queue(disk->queue);
			put_disk(disk);
		}
	}
	unregister_blkdev(NBD_MAJOR, "nbd");
	printk(KERN_INFO "nbd: unregistered device at major %d\n", NBD_MAJOR);
}

module_init(nbd_init);
module_exit(nbd_cleanup);

MODULE_DESCRIPTION("Network Block Device");
MODULE_LICENSE("GPL");

module_param(nbds_max, int, 0444);
MODULE_PARM_DESC(nbds_max, "How many network block devices to initialize.");
#ifndef NDEBUG
module_param(debugflags, int, 0644);
MODULE_PARM_DESC(debugflags, "flags for controlling debug output");
#endif
Commit	Line	Data
1da177e4 LT	1	/*
	2	* Network block device - make block devices work over TCP
	3	*
	4	* Note that you can not swap over this thing, yet. Seems to work but
	5	* deadlocks sometimes - you can not swap over TCP in general.
	6	*
	7	* Copyright 1997-2000 Pavel Machek <pavel@ucw.cz>
	8	* Parts copyright 2001 Steven Whitehouse <steve@chygwyn.com>
	9	*
dbf492d6	10	* This file is released under GPLv2 or later.
1da177e4	11	*
dbf492d6	12	* (part of code stolen from loop.c)
1da177e4 LT	13	*/
	14
	15	#include <linux/major.h>
	16
	17	#include <linux/blkdev.h>
	18	#include <linux/module.h>
	19	#include <linux/init.h>
	20	#include <linux/sched.h>
	21	#include <linux/fs.h>
	22	#include <linux/bio.h>
	23	#include <linux/stat.h>
	24	#include <linux/errno.h>
	25	#include <linux/file.h>
	26	#include <linux/ioctl.h>
4b2f0260 HX	27	#include <linux/compiler.h>
	28	#include <linux/err.h>
	29	#include <linux/kernel.h>
1da177e4 LT	30	#include <net/sock.h>
1da177e4 LT	31
1da177e4	32	#include <asm/uaccess.h>
4b2f0260	33	#include <asm/system.h>
1da177e4 LT	34	#include <asm/types.h>
	35
	36	#include <linux/nbd.h>
	37
	38	#define LO_MAGIC 0x68797548
	39
	40	#ifdef NDEBUG
	41	#define dprintk(flags, fmt...)
	42	#else /* NDEBUG */
	43	#define dprintk(flags, fmt...) do { \
	44	if (debugflags & (flags)) printk(KERN_DEBUG fmt); \
	45	} while (0)
	46	#define DBG_IOCTL 0x0004
	47	#define DBG_INIT 0x0010
	48	#define DBG_EXIT 0x0020
	49	#define DBG_BLKDEV 0x0100
	50	#define DBG_RX 0x0200
	51	#define DBG_TX 0x0400
	52	static unsigned int debugflags;
	53	#endif /* NDEBUG */
	54
9c7a4169	55	static unsigned int nbds_max = 16;
1da177e4 LT	56	static struct nbd_device nbd_dev[MAX_NBD];
	57
	58	/*
	59	* Use just one lock (or at most 1 per NIC). Two arguments for this:
	60	* 1. Each NIC is essentially a synchronization point for all servers
	61	* accessed through that NIC so there's no need to have more locks
	62	* than NICs anyway.
	63	* 2. More locks lead to more "Dirty cache line bouncing" which will slow
	64	* down each lock to the point where they're actually slower than just
	65	* a single lock.
	66	* Thanks go to Jens Axboe and Al Viro for their LKML emails explaining this!
	67	*/
	68	static DEFINE_SPINLOCK(nbd_lock);
	69
	70	#ifndef NDEBUG
	71	static const char *ioctl_cmd_to_ascii(int cmd)
	72	{
	73	switch (cmd) {
	74	case NBD_SET_SOCK: return "set-sock";
	75	case NBD_SET_BLKSIZE: return "set-blksize";
	76	case NBD_SET_SIZE: return "set-size";
	77	case NBD_DO_IT: return "do-it";
	78	case NBD_CLEAR_SOCK: return "clear-sock";
	79	case NBD_CLEAR_QUE: return "clear-que";
	80	case NBD_PRINT_DEBUG: return "print-debug";
	81	case NBD_SET_SIZE_BLOCKS: return "set-size-blocks";
	82	case NBD_DISCONNECT: return "disconnect";
	83	case BLKROSET: return "set-read-only";
	84	case BLKFLSBUF: return "flush-buffer-cache";
	85	}
	86	return "unknown";
	87	}
	88
	89	static const char *nbdcmd_to_ascii(int cmd)
	90	{
	91	switch (cmd) {
	92	case NBD_CMD_READ: return "read";
	93	case NBD_CMD_WRITE: return "write";
	94	case NBD_CMD_DISC: return "disconnect";
	95	}
	96	return "invalid";
	97	}
	98	#endif /* NDEBUG */
	99
	100	static void nbd_end_request(struct request *req)
	101	{
	102	int uptodate = (req->errors == 0) ? 1 : 0;
165125e1	103	struct request_queue *q = req->q;
1da177e4 LT	104	unsigned long flags;
	105
	106	dprintk(DBG_BLKDEV, "%s: request %p: %s\n", req->rq_disk->disk_name,
	107	req, uptodate? "done": "failed");
	108
	109	spin_lock_irqsave(q->queue_lock, flags);
	110	if (!end_that_request_first(req, uptodate, req->nr_sectors)) {
8ffdc655	111	end_that_request_last(req, uptodate);
1da177e4 LT	112	}
	113	spin_unlock_irqrestore(q->queue_lock, flags);
	114	}
	115
	116	/*
	117	* Send or receive packet.
	118	*/
	119	static int sock_xmit(struct socket sock, int send, void buf, int size,
	120	int msg_flags)
	121	{
	122	int result;
	123	struct msghdr msg;
	124	struct kvec iov;
be0ef957	125	sigset_t blocked, oldset;
1da177e4 LT	126
	127	/* Allow interception of SIGKILL only
	128	* Don't allow other signals to interrupt the transmission */
be0ef957 ON	129	siginitsetinv(&blocked, sigmask(SIGKILL));
be0ef957 ON	130	sigprocmask(SIG_SETMASK, &blocked, &oldset);
1da177e4 LT	131
	132	do {
	133	sock->sk->sk_allocation = GFP_NOIO;
	134	iov.iov_base = buf;
	135	iov.iov_len = size;
	136	msg.msg_name = NULL;
	137	msg.msg_namelen = 0;
	138	msg.msg_control = NULL;
	139	msg.msg_controllen = 0;
1da177e4 LT	140	msg.msg_flags = msg_flags \| MSG_NOSIGNAL;
	141
	142	if (send)
	143	result = kernel_sendmsg(sock, &msg, &iov, 1, size);
	144	else
	145	result = kernel_recvmsg(sock, &msg, &iov, 1, size, 0);
	146
	147	if (signal_pending(current)) {
	148	siginfo_t info;
1da177e4	149	printk(KERN_WARNING "nbd (pid %d: %s) got signal %d\n",
be0ef957 ON	150	current->pid, current->comm,
be0ef957 ON	151	dequeue_signal_lock(current, &current->blocked, &info));
1da177e4 LT	152	result = -EINTR;
	153	break;
	154	}
	155
	156	if (result <= 0) {
	157	if (result == 0)
	158	result = -EPIPE; /* short read */
	159	break;
	160	}
	161	size -= result;
	162	buf += result;
	163	} while (size > 0);
	164
be0ef957	165	sigprocmask(SIG_SETMASK, &oldset, NULL);
1da177e4 LT	166
	167	return result;
	168	}
	169
	170	static inline int sock_send_bvec(struct socket sock, struct bio_vec bvec,
	171	int flags)
	172	{
	173	int result;
	174	void *kaddr = kmap(bvec->bv_page);
	175	result = sock_xmit(sock, 1, kaddr + bvec->bv_offset, bvec->bv_len,
	176	flags);
	177	kunmap(bvec->bv_page);
	178	return result;
	179	}
	180
	181	static int nbd_send_req(struct nbd_device lo, struct request req)
	182	{
5705f702	183	int result, flags;
1da177e4 LT	184	struct nbd_request request;
	185	unsigned long size = req->nr_sectors << 9;
	186	struct socket *sock = lo->sock;
	187
	188	request.magic = htonl(NBD_REQUEST_MAGIC);
	189	request.type = htonl(nbd_cmd(req));
	190	request.from = cpu_to_be64((u64) req->sector << 9);
	191	request.len = htonl(size);
	192	memcpy(request.handle, &req, sizeof(req));
	193
1da177e4 LT	194	dprintk(DBG_TX, "%s: request %p: sending control (%s@%llu,%luB)\n",
	195	lo->disk->disk_name, req,
	196	nbdcmd_to_ascii(nbd_cmd(req)),
	197	(unsigned long long)req->sector << 9,
	198	req->nr_sectors << 9);
	199	result = sock_xmit(sock, 1, &request, sizeof(request),
	200	(nbd_cmd(req) == NBD_CMD_WRITE)? MSG_MORE: 0);
	201	if (result <= 0) {
	202	printk(KERN_ERR "%s: Send control failed (result %d)\n",
	203	lo->disk->disk_name, result);
	204	goto error_out;
	205	}
	206
	207	if (nbd_cmd(req) == NBD_CMD_WRITE) {
5705f702 N	208	struct req_iterator iter;
5705f702 N	209	struct bio_vec *bvec;
1da177e4 LT	210	/*
	211	* we are really probing at internals to determine
	212	* whether to set MSG_MORE or not...
	213	*/
5705f702	214	rq_for_each_segment(bvec, req, iter) {
1da177e4	215	flags = 0;
5705f702	216	if (!rq_iter_last(req, iter))
1da177e4 LT	217	flags = MSG_MORE;
	218	dprintk(DBG_TX, "%s: request %p: sending %d bytes data\n",
	219	lo->disk->disk_name, req,
	220	bvec->bv_len);
	221	result = sock_send_bvec(sock, bvec, flags);
	222	if (result <= 0) {
	223	printk(KERN_ERR "%s: Send data failed (result %d)\n",
	224	lo->disk->disk_name,
	225	result);
	226	goto error_out;
	227	}
1da177e4 LT	228	}
1da177e4 LT	229	}
1da177e4 LT	230	return 0;
	231
	232	error_out:
1da177e4 LT	233	return 1;
	234	}
	235
	236	static struct request nbd_find_request(struct nbd_device lo, char *handle)
	237	{
	238	struct request *req;
	239	struct list_head *tmp;
	240	struct request *xreq;
4b2f0260	241	int err;
1da177e4 LT	242
	243	memcpy(&xreq, handle, sizeof(xreq));
	244
4b2f0260 HX	245	err = wait_event_interruptible(lo->active_wq, lo->active_req != xreq);
	246	if (unlikely(err))
	247	goto out;
	248
1da177e4 LT	249	spin_lock(&lo->queue_lock);
	250	list_for_each(tmp, &lo->queue_head) {
	251	req = list_entry(tmp, struct request, queuelist);
	252	if (req != xreq)
	253	continue;
	254	list_del_init(&req->queuelist);
	255	spin_unlock(&lo->queue_lock);
	256	return req;
	257	}
	258	spin_unlock(&lo->queue_lock);
4b2f0260 HX	259
	260	err = -ENOENT;
	261
	262	out:
	263	return ERR_PTR(err);
1da177e4 LT	264	}
	265
	266	static inline int sock_recv_bvec(struct socket sock, struct bio_vec bvec)
	267	{
	268	int result;
	269	void *kaddr = kmap(bvec->bv_page);
	270	result = sock_xmit(sock, 0, kaddr + bvec->bv_offset, bvec->bv_len,
	271	MSG_WAITALL);
	272	kunmap(bvec->bv_page);
	273	return result;
	274	}
	275
	276	/* NULL returned = something went wrong, inform userspace */
	277	static struct request nbd_read_stat(struct nbd_device lo)
	278	{
	279	int result;
	280	struct nbd_reply reply;
	281	struct request *req;
	282	struct socket *sock = lo->sock;
	283
	284	reply.magic = 0;
	285	result = sock_xmit(sock, 0, &reply, sizeof(reply), MSG_WAITALL);
	286	if (result <= 0) {
	287	printk(KERN_ERR "%s: Receive control failed (result %d)\n",
	288	lo->disk->disk_name, result);
	289	goto harderror;
	290	}
e4b57e08 MF	291
	292	if (ntohl(reply.magic) != NBD_REPLY_MAGIC) {
	293	printk(KERN_ERR "%s: Wrong magic (0x%lx)\n",
	294	lo->disk->disk_name,
	295	(unsigned long)ntohl(reply.magic));
	296	result = -EPROTO;
	297	goto harderror;
	298	}
	299
1da177e4	300	req = nbd_find_request(lo, reply.handle);
4b2f0260 HX	301	if (unlikely(IS_ERR(req))) {
	302	result = PTR_ERR(req);
	303	if (result != -ENOENT)
	304	goto harderror;
	305
1da177e4 LT	306	printk(KERN_ERR "%s: Unexpected reply (%p)\n",
	307	lo->disk->disk_name, reply.handle);
	308	result = -EBADR;
	309	goto harderror;
	310	}
	311
1da177e4 LT	312	if (ntohl(reply.error)) {
	313	printk(KERN_ERR "%s: Other side returned error (%d)\n",
	314	lo->disk->disk_name, ntohl(reply.error));
	315	req->errors++;
	316	return req;
	317	}
	318
	319	dprintk(DBG_RX, "%s: request %p: got reply\n",
	320	lo->disk->disk_name, req);
	321	if (nbd_cmd(req) == NBD_CMD_READ) {
5705f702 N	322	struct req_iterator iter;
	323	struct bio_vec *bvec;
	324
	325	rq_for_each_segment(bvec, req, iter) {
1da177e4 LT	326	result = sock_recv_bvec(sock, bvec);
	327	if (result <= 0) {
	328	printk(KERN_ERR "%s: Receive data failed (result %d)\n",
	329	lo->disk->disk_name,
	330	result);
f0df33bc MF	331	req->errors++;
f0df33bc MF	332	return req;
1da177e4 LT	333	}
	334	dprintk(DBG_RX, "%s: request %p: got %d bytes data\n",
	335	lo->disk->disk_name, req, bvec->bv_len);
1da177e4 LT	336	}
	337	}
	338	return req;
	339	harderror:
	340	lo->harderror = result;
	341	return NULL;
	342	}
	343
6b39bb65 PC	344	static ssize_t pid_show(struct gendisk disk, char page)
	345	{
	346	return sprintf(page, "%ld\n",
	347	(long) ((struct nbd_device *)disk->private_data)->pid);
	348	}
	349
	350	static struct disk_attribute pid_attr = {
	351	.attr = { .name = "pid", .mode = S_IRUGO },
	352	.show = pid_show,
	353	};
	354
84963048	355	static int nbd_do_it(struct nbd_device *lo)
1da177e4 LT	356	{
1da177e4 LT	357	struct request *req;
84963048	358	int ret;
1da177e4 LT	359
	360	BUG_ON(lo->magic != LO_MAGIC);
	361
6b39bb65	362	lo->pid = current->pid;
84963048 WC	363	ret = sysfs_create_file(&lo->disk->kobj, &pid_attr.attr);
	364	if (ret) {
	365	printk(KERN_ERR "nbd: sysfs_create_file failed!");
	366	return ret;
	367	}
6b39bb65	368
1da177e4 LT	369	while ((req = nbd_read_stat(lo)) != NULL)
1da177e4 LT	370	nbd_end_request(req);
6b39bb65 PC	371
6b39bb65 PC	372	sysfs_remove_file(&lo->disk->kobj, &pid_attr.attr);
84963048	373	return 0;
1da177e4 LT	374	}
	375
	376	static void nbd_clear_que(struct nbd_device *lo)
	377	{
	378	struct request *req;
	379
	380	BUG_ON(lo->magic != LO_MAGIC);
	381
4b2f0260 HX	382	/*
	383	* Because we have set lo->sock to NULL under the tx_lock, all
	384	* modifications to the list must have completed by now. For
	385	* the same reason, the active_req must be NULL.
	386	*
	387	* As a consequence, we don't need to take the spin lock while
	388	* purging the list here.
	389	*/
	390	BUG_ON(lo->sock);
	391	BUG_ON(lo->active_req);
	392
	393	while (!list_empty(&lo->queue_head)) {
	394	req = list_entry(lo->queue_head.next, struct request,
	395	queuelist);
	396	list_del_init(&req->queuelist);
	397	req->errors++;
	398	nbd_end_request(req);
	399	}
1da177e4 LT	400	}
	401
	402	/*
	403	* We always wait for result of write, for now. It would be nice to make it optional
	404	* in future
e654bc43	405	* if ((rq_data_dir(req) == WRITE) && (lo->flags & NBD_WRITE_NOCHK))
1da177e4 LT	406	* { printk( "Warning: Ignoring result!\n"); nbd_end_request( req ); }
	407	*/
	408
165125e1	409	static void do_nbd_request(struct request_queue * q)
1da177e4 LT	410	{
	411	struct request *req;
	412
	413	while ((req = elv_next_request(q)) != NULL) {
	414	struct nbd_device *lo;
	415
	416	blkdev_dequeue_request(req);
4aff5e23 JA	417	dprintk(DBG_BLKDEV, "%s: request %p: dequeued (flags=%x)\n",
4aff5e23 JA	418	req->rq_disk->disk_name, req, req->cmd_type);
1da177e4	419
4aff5e23	420	if (!blk_fs_request(req))
1da177e4 LT	421	goto error_out;
	422
	423	lo = req->rq_disk->private_data;
	424
	425	BUG_ON(lo->magic != LO_MAGIC);
	426
1da177e4 LT	427	nbd_cmd(req) = NBD_CMD_READ;
	428	if (rq_data_dir(req) == WRITE) {
	429	nbd_cmd(req) = NBD_CMD_WRITE;
	430	if (lo->flags & NBD_READ_ONLY) {
	431	printk(KERN_ERR "%s: Write on read-only\n",
	432	lo->disk->disk_name);
	433	goto error_out;
	434	}
	435	}
	436
	437	req->errors = 0;
	438	spin_unlock_irq(q->queue_lock);
	439
82d4dc5a	440	mutex_lock(&lo->tx_lock);
4b2f0260	441	if (unlikely(!lo->sock)) {
82d4dc5a	442	mutex_unlock(&lo->tx_lock);
4b2f0260 HX	443	printk(KERN_ERR "%s: Attempted send on closed socket\n",
4b2f0260 HX	444	lo->disk->disk_name);
1da177e4 LT	445	req->errors++;
	446	nbd_end_request(req);
	447	spin_lock_irq(q->queue_lock);
	448	continue;
	449	}
	450
4b2f0260	451	lo->active_req = req;
1da177e4 LT	452
	453	if (nbd_send_req(lo, req) != 0) {
	454	printk(KERN_ERR "%s: Request send failed\n",
	455	lo->disk->disk_name);
4b2f0260 HX	456	req->errors++;
	457	nbd_end_request(req);
	458	} else {
	459	spin_lock(&lo->queue_lock);
	460	list_add(&req->queuelist, &lo->queue_head);
	461	spin_unlock(&lo->queue_lock);
1da177e4 LT	462	}
1da177e4 LT	463
4b2f0260	464	lo->active_req = NULL;
82d4dc5a	465	mutex_unlock(&lo->tx_lock);
4b2f0260 HX	466	wake_up_all(&lo->active_wq);
4b2f0260 HX	467
1da177e4 LT	468	spin_lock_irq(q->queue_lock);
	469	continue;
	470
	471	error_out:
	472	req->errors++;
	473	spin_unlock(q->queue_lock);
	474	nbd_end_request(req);
	475	spin_lock(q->queue_lock);
	476	}
	477	return;
	478	}
	479
	480	static int nbd_ioctl(struct inode inode, struct file file,
	481	unsigned int cmd, unsigned long arg)
	482	{
	483	struct nbd_device *lo = inode->i_bdev->bd_disk->private_data;
	484	int error;
	485	struct request sreq ;
	486
	487	if (!capable(CAP_SYS_ADMIN))
	488	return -EPERM;
	489
	490	BUG_ON(lo->magic != LO_MAGIC);
	491
	492	/* Anyone capable of this syscall can do real bad things */
	493	dprintk(DBG_IOCTL, "%s: nbd_ioctl cmd=%s(0x%x) arg=%lu\n",
	494	lo->disk->disk_name, ioctl_cmd_to_ascii(cmd), cmd, arg);
	495
	496	switch (cmd) {
	497	case NBD_DISCONNECT:
	498	printk(KERN_INFO "%s: NBD_DISCONNECT\n", lo->disk->disk_name);
4aff5e23	499	sreq.cmd_type = REQ_TYPE_SPECIAL;
1da177e4 LT	500	nbd_cmd(&sreq) = NBD_CMD_DISC;
	501	/*
	502	* Set these to sane values in case server implementation
	503	* fails to check the request type first and also to keep
	504	* debugging output cleaner.
	505	*/
	506	sreq.sector = 0;
	507	sreq.nr_sectors = 0;
	508	if (!lo->sock)
	509	return -EINVAL;
	510	nbd_send_req(lo, &sreq);
	511	return 0;
	512
	513	case NBD_CLEAR_SOCK:
	514	error = 0;
82d4dc5a	515	mutex_lock(&lo->tx_lock);
1da177e4	516	lo->sock = NULL;
82d4dc5a	517	mutex_unlock(&lo->tx_lock);
1da177e4 LT	518	file = lo->file;
1da177e4 LT	519	lo->file = NULL;
1da177e4	520	nbd_clear_que(lo);
4b2f0260	521	BUG_ON(!list_empty(&lo->queue_head));
1da177e4 LT	522	if (file)
	523	fput(file);
	524	return error;
	525	case NBD_SET_SOCK:
	526	if (lo->file)
	527	return -EBUSY;
	528	error = -EINVAL;
	529	file = fget(arg);
	530	if (file) {
17506041	531	inode = file->f_path.dentry->d_inode;
1da177e4 LT	532	if (S_ISSOCK(inode->i_mode)) {
	533	lo->file = file;
	534	lo->sock = SOCKET_I(inode);
	535	error = 0;
	536	} else {
	537	fput(file);
	538	}
	539	}
	540	return error;
	541	case NBD_SET_BLKSIZE:
	542	lo->blksize = arg;
	543	lo->bytesize &= ~(lo->blksize-1);
	544	inode->i_bdev->bd_inode->i_size = lo->bytesize;
	545	set_blocksize(inode->i_bdev, lo->blksize);
	546	set_capacity(lo->disk, lo->bytesize >> 9);
	547	return 0;
	548	case NBD_SET_SIZE:
	549	lo->bytesize = arg & ~(lo->blksize-1);
	550	inode->i_bdev->bd_inode->i_size = lo->bytesize;
	551	set_blocksize(inode->i_bdev, lo->blksize);
	552	set_capacity(lo->disk, lo->bytesize >> 9);
	553	return 0;
	554	case NBD_SET_SIZE_BLOCKS:
	555	lo->bytesize = ((u64) arg) * lo->blksize;
	556	inode->i_bdev->bd_inode->i_size = lo->bytesize;
	557	set_blocksize(inode->i_bdev, lo->blksize);
	558	set_capacity(lo->disk, lo->bytesize >> 9);
	559	return 0;
	560	case NBD_DO_IT:
	561	if (!lo->file)
	562	return -EINVAL;
84963048 WC	563	error = nbd_do_it(lo);
	564	if (error)
	565	return error;
1da177e4 LT	566	/* on return tidy up in case we have a signal */
	567	/* Forcibly shutdown the socket causing all listeners
	568	* to error
	569	*
	570	* FIXME: This code is duplicated from sys_shutdown, but
	571	* there should be a more generic interface rather than
	572	* calling socket ops directly here */
82d4dc5a	573	mutex_lock(&lo->tx_lock);
1da177e4 LT	574	if (lo->sock) {
	575	printk(KERN_WARNING "%s: shutting down socket\n",
	576	lo->disk->disk_name);
	577	lo->sock->ops->shutdown(lo->sock,
	578	SEND_SHUTDOWN\|RCV_SHUTDOWN);
	579	lo->sock = NULL;
	580	}
82d4dc5a	581	mutex_unlock(&lo->tx_lock);
1da177e4 LT	582	file = lo->file;
1da177e4 LT	583	lo->file = NULL;
1da177e4 LT	584	nbd_clear_que(lo);
	585	printk(KERN_WARNING "%s: queue cleared\n", lo->disk->disk_name);
	586	if (file)
	587	fput(file);
	588	return lo->harderror;
	589	case NBD_CLEAR_QUE:
4b2f0260 HX	590	/*
	591	* This is for compatibility only. The queue is always cleared
	592	* by NBD_DO_IT or NBD_CLEAR_SOCK.
	593	*/
	594	BUG_ON(!lo->sock && !list_empty(&lo->queue_head));
1da177e4 LT	595	return 0;
	596	case NBD_PRINT_DEBUG:
	597	printk(KERN_INFO "%s: next = %p, prev = %p, head = %p\n",
	598	inode->i_bdev->bd_disk->disk_name,
	599	lo->queue_head.next, lo->queue_head.prev,
	600	&lo->queue_head);
	601	return 0;
	602	}
	603	return -EINVAL;
	604	}
	605
	606	static struct block_device_operations nbd_fops =
	607	{
	608	.owner = THIS_MODULE,
	609	.ioctl = nbd_ioctl,
	610	};
	611
	612	/*
	613	* And here should be modules and kernel interface
	614	* (Just smiley confuses emacs :-)
	615	*/
	616
	617	static int __init nbd_init(void)
	618	{
	619	int err = -ENOMEM;
	620	int i;
	621
5b7b18cc	622	BUILD_BUG_ON(sizeof(struct nbd_request) != 28);
1da177e4	623
40be0c28 LMB	624	if (nbds_max > MAX_NBD) {
	625	printk(KERN_CRIT "nbd: cannot allocate more than %u nbds; %u requested.\n", MAX_NBD,
	626	nbds_max);
	627	return -EINVAL;
	628	}
	629
	630	for (i = 0; i < nbds_max; i++) {
1da177e4 LT	631	struct gendisk *disk = alloc_disk(1);
	632	if (!disk)
	633	goto out;
	634	nbd_dev[i].disk = disk;
	635	/*
	636	* The new linux 2.5 block layer implementation requires
	637	* every gendisk to have its very own request_queue struct.
	638	* These structs are big so we dynamically allocate them.
	639	*/
	640	disk->queue = blk_init_queue(do_nbd_request, &nbd_lock);
	641	if (!disk->queue) {
	642	put_disk(disk);
	643	goto out;
	644	}
	645	}
	646
	647	if (register_blkdev(NBD_MAJOR, "nbd")) {
	648	err = -EIO;
	649	goto out;
	650	}
	651
	652	printk(KERN_INFO "nbd: registered device at major %d\n", NBD_MAJOR);
	653	dprintk(DBG_INIT, "nbd: debugflags=0x%x\n", debugflags);
	654
40be0c28	655	for (i = 0; i < nbds_max; i++) {
1da177e4 LT	656	struct gendisk *disk = nbd_dev[i].disk;
	657	nbd_dev[i].file = NULL;
	658	nbd_dev[i].magic = LO_MAGIC;
	659	nbd_dev[i].flags = 0;
	660	spin_lock_init(&nbd_dev[i].queue_lock);
	661	INIT_LIST_HEAD(&nbd_dev[i].queue_head);
82d4dc5a	662	mutex_init(&nbd_dev[i].tx_lock);
4b2f0260	663	init_waitqueue_head(&nbd_dev[i].active_wq);
1da177e4 LT	664	nbd_dev[i].blksize = 1024;
	665	nbd_dev[i].bytesize = 0x7ffffc00ULL << 10; /* 2TB */
	666	disk->major = NBD_MAJOR;
	667	disk->first_minor = i;
	668	disk->fops = &nbd_fops;
	669	disk->private_data = &nbd_dev[i];
	670	disk->flags \|= GENHD_FL_SUPPRESS_PARTITION_INFO;
	671	sprintf(disk->disk_name, "nbd%d", i);
1da177e4 LT	672	set_capacity(disk, 0x7ffffc00ULL << 1); /* 2 TB */
	673	add_disk(disk);
	674	}
	675
	676	return 0;
	677	out:
	678	while (i--) {
	679	blk_cleanup_queue(nbd_dev[i].disk->queue);
	680	put_disk(nbd_dev[i].disk);
	681	}
	682	return err;
	683	}
	684
	685	static void __exit nbd_cleanup(void)
	686	{
	687	int i;
40be0c28	688	for (i = 0; i < nbds_max; i++) {
1da177e4	689	struct gendisk *disk = nbd_dev[i].disk;
40be0c28	690	nbd_dev[i].magic = 0;
1da177e4 LT	691	if (disk) {
	692	del_gendisk(disk);
	693	blk_cleanup_queue(disk->queue);
	694	put_disk(disk);
	695	}
	696	}
1da177e4 LT	697	unregister_blkdev(NBD_MAJOR, "nbd");
	698	printk(KERN_INFO "nbd: unregistered device at major %d\n", NBD_MAJOR);
	699	}
	700
	701	module_init(nbd_init);
	702	module_exit(nbd_cleanup);
	703
	704	MODULE_DESCRIPTION("Network Block Device");
	705	MODULE_LICENSE("GPL");
	706
40be0c28 LMB	707	module_param(nbds_max, int, 0444);
40be0c28 LMB	708	MODULE_PARM_DESC(nbds_max, "How many network block devices to initialize.");
1da177e4 LT	709	#ifndef NDEBUG
	710	module_param(debugflags, int, 0644);
	711	MODULE_PARM_DESC(debugflags, "flags for controlling debug output");
	712	#endif