nbd.c

   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 * Network block device - make block devices work over TCP
   4 *
   5 * Note that you can not swap over this thing, yet. Seems to work but
   6 * deadlocks sometimes - you can not swap over TCP in general.
   7 * 
   8 * Copyright 1997-2000, 2008 Pavel Machek <pavel@ucw.cz>
   9 * Parts copyright 2001 Steven Whitehouse <steve@chygwyn.com>
  10 *
  11 * (part of code stolen from loop.c)
  12 */
  13
  14#define pr_fmt(fmt) "nbd: " fmt
  15
  16#include <linux/major.h>
  17
  18#include <linux/blkdev.h>
  19#include <linux/module.h>
  20#include <linux/init.h>
  21#include <linux/sched.h>
  22#include <linux/sched/mm.h>
  23#include <linux/fs.h>
  24#include <linux/bio.h>
  25#include <linux/stat.h>
  26#include <linux/errno.h>
  27#include <linux/file.h>
  28#include <linux/ioctl.h>
  29#include <linux/mutex.h>
  30#include <linux/compiler.h>
  31#include <linux/completion.h>
  32#include <linux/err.h>
  33#include <linux/kernel.h>
  34#include <linux/slab.h>
  35#include <net/sock.h>
  36#include <linux/net.h>
  37#include <linux/kthread.h>
  38#include <linux/types.h>
  39#include <linux/debugfs.h>
  40#include <linux/blk-mq.h>
  41
  42#include <linux/uaccess.h>
  43#include <asm/types.h>
  44
  45#include <linux/nbd.h>
  46#include <linux/nbd-netlink.h>
  47#include <net/genetlink.h>
  48
  49#define CREATE_TRACE_POINTS
  50#include <trace/events/nbd.h>
  51
  52static DEFINE_IDR(nbd_index_idr);
  53static DEFINE_MUTEX(nbd_index_mutex);
  54static struct workqueue_struct *nbd_del_wq;
  55static int nbd_total_devices = 0;
  56
  57struct nbd_sock {
  58	struct socket *sock;
  59	struct mutex tx_lock;
  60	struct request *pending;
  61	int sent;
  62	bool dead;
  63	int fallback_index;
  64	int cookie;
  65	struct work_struct work;
  66};
  67
  68struct recv_thread_args {
  69	struct work_struct work;
  70	struct nbd_device *nbd;
  71	struct nbd_sock *nsock;
  72	int index;
  73};
  74
  75struct link_dead_args {
  76	struct work_struct work;
  77	int index;
  78};
  79
  80#define NBD_RT_TIMEDOUT			0
  81#define NBD_RT_DISCONNECT_REQUESTED	1
  82#define NBD_RT_DISCONNECTED		2
  83#define NBD_RT_HAS_PID_FILE		3
  84#define NBD_RT_HAS_CONFIG_REF		4
  85#define NBD_RT_BOUND			5
  86#define NBD_RT_DISCONNECT_ON_CLOSE	6
  87#define NBD_RT_HAS_BACKEND_FILE		7
  88
  89#define NBD_DESTROY_ON_DISCONNECT	0
  90#define NBD_DISCONNECT_REQUESTED	1
  91
  92struct nbd_config {
  93	u32 flags;
  94	unsigned long runtime_flags;
  95	u64 dead_conn_timeout;
  96
  97	struct nbd_sock **socks;
  98	int num_connections;
  99	atomic_t live_connections;
 100	wait_queue_head_t conn_wait;
 101
 102	atomic_t recv_threads;
 103	wait_queue_head_t recv_wq;
 104	unsigned int blksize_bits;
 105	loff_t bytesize;
 106#if IS_ENABLED(CONFIG_DEBUG_FS)
 107	struct dentry *dbg_dir;
 108#endif
 109};
 110
 111static inline unsigned int nbd_blksize(struct nbd_config *config)
 112{
 113	return 1u << config->blksize_bits;
 114}
 115
 116struct nbd_device {
 117	struct blk_mq_tag_set tag_set;
 118
 119	int index;
 120	refcount_t config_refs;
 121	refcount_t refs;
 122	struct nbd_config *config;
 123	struct mutex config_lock;
 124	struct gendisk *disk;
 125	struct workqueue_struct *recv_workq;
 126	struct work_struct remove_work;
 127
 128	struct list_head list;
 129	struct task_struct *task_setup;
 130
 131	unsigned long flags;
 132	pid_t pid; /* pid of nbd-client, if attached */
 133
 134	char *backend;
 135};
 136
 137#define NBD_CMD_REQUEUED	1
 138/*
 139 * This flag will be set if nbd_queue_rq() succeed, and will be checked and
 140 * cleared in completion. Both setting and clearing of the flag are protected
 141 * by cmd->lock.
 142 */
 143#define NBD_CMD_INFLIGHT	2
 144
 145/* Just part of request header or data payload is sent successfully */
 146#define NBD_CMD_PARTIAL_SEND	3
 147
 148struct nbd_cmd {
 149	struct nbd_device *nbd;
 150	struct mutex lock;
 151	int index;
 152	int cookie;
 153	int retries;
 154	blk_status_t status;
 155	unsigned long flags;
 156	u32 cmd_cookie;
 157};
 158
 159#if IS_ENABLED(CONFIG_DEBUG_FS)
 160static struct dentry *nbd_dbg_dir;
 161#endif
 162
 163#define nbd_name(nbd) ((nbd)->disk->disk_name)
 164
 165#define NBD_DEF_BLKSIZE_BITS 10
 166
 167static unsigned int nbds_max = 16;
 168static int max_part = 16;
 169static int part_shift;
 170
 171static int nbd_dev_dbg_init(struct nbd_device *nbd);
 172static void nbd_dev_dbg_close(struct nbd_device *nbd);
 173static void nbd_config_put(struct nbd_device *nbd);
 174static void nbd_connect_reply(struct genl_info *info, int index);
 175static int nbd_genl_status(struct sk_buff *skb, struct genl_info *info);
 176static void nbd_dead_link_work(struct work_struct *work);
 177static void nbd_disconnect_and_put(struct nbd_device *nbd);
 178
 179static inline struct device *nbd_to_dev(struct nbd_device *nbd)
 180{
 181	return disk_to_dev(nbd->disk);
 182}
 183
 184static void nbd_requeue_cmd(struct nbd_cmd *cmd)
 185{
 186	struct request *req = blk_mq_rq_from_pdu(cmd);
 187
 188	lockdep_assert_held(&cmd->lock);
 189
 190	/*
 191	 * Clear INFLIGHT flag so that this cmd won't be completed in
 192	 * normal completion path
 193	 *
 194	 * INFLIGHT flag will be set when the cmd is queued to nbd next
 195	 * time.
 196	 */
 197	__clear_bit(NBD_CMD_INFLIGHT, &cmd->flags);
 198
 199	if (!test_and_set_bit(NBD_CMD_REQUEUED, &cmd->flags))
 200		blk_mq_requeue_request(req, true);
 201}
 202
 203#define NBD_COOKIE_BITS 32
 204
 205static u64 nbd_cmd_handle(struct nbd_cmd *cmd)
 206{
 207	struct request *req = blk_mq_rq_from_pdu(cmd);
 208	u32 tag = blk_mq_unique_tag(req);
 209	u64 cookie = cmd->cmd_cookie;
 210
 211	return (cookie << NBD_COOKIE_BITS) | tag;
 212}
 213
 214static u32 nbd_handle_to_tag(u64 handle)
 215{
 216	return (u32)handle;
 217}
 218
 219static u32 nbd_handle_to_cookie(u64 handle)
 220{
 221	return (u32)(handle >> NBD_COOKIE_BITS);
 222}
 223
 224static const char *nbdcmd_to_ascii(int cmd)
 225{
 226	switch (cmd) {
 227	case  NBD_CMD_READ: return "read";
 228	case NBD_CMD_WRITE: return "write";
 229	case  NBD_CMD_DISC: return "disconnect";
 230	case NBD_CMD_FLUSH: return "flush";
 231	case  NBD_CMD_TRIM: return "trim/discard";
 232	}
 233	return "invalid";
 234}
 235
 236static ssize_t pid_show(struct device *dev,
 237			struct device_attribute *attr, char *buf)
 238{
 239	struct gendisk *disk = dev_to_disk(dev);
 240	struct nbd_device *nbd = disk->private_data;
 241
 242	return sprintf(buf, "%d\n", nbd->pid);
 243}
 244
 245static const struct device_attribute pid_attr = {
 246	.attr = { .name = "pid", .mode = 0444},
 247	.show = pid_show,
 248};
 249
 250static ssize_t backend_show(struct device *dev,
 251		struct device_attribute *attr, char *buf)
 252{
 253	struct gendisk *disk = dev_to_disk(dev);
 254	struct nbd_device *nbd = disk->private_data;
 255
 256	return sprintf(buf, "%s\n", nbd->backend ?: "");
 257}
 258
 259static const struct device_attribute backend_attr = {
 260	.attr = { .name = "backend", .mode = 0444},
 261	.show = backend_show,
 262};
 263
 264static void nbd_dev_remove(struct nbd_device *nbd)
 265{
 266	struct gendisk *disk = nbd->disk;
 267
 268	del_gendisk(disk);
 269	blk_mq_free_tag_set(&nbd->tag_set);
 270
 271	/*
 272	 * Remove from idr after del_gendisk() completes, so if the same ID is
 273	 * reused, the following add_disk() will succeed.
 274	 */
 275	mutex_lock(&nbd_index_mutex);
 276	idr_remove(&nbd_index_idr, nbd->index);
 277	mutex_unlock(&nbd_index_mutex);
 278	destroy_workqueue(nbd->recv_workq);
 279	put_disk(disk);
 280}
 281
 282static void nbd_dev_remove_work(struct work_struct *work)
 283{
 284	nbd_dev_remove(container_of(work, struct nbd_device, remove_work));
 285}
 286
 287static void nbd_put(struct nbd_device *nbd)
 288{
 289	if (!refcount_dec_and_test(&nbd->refs))
 290		return;
 291
 292	/* Call del_gendisk() asynchrounously to prevent deadlock */
 293	if (test_bit(NBD_DESTROY_ON_DISCONNECT, &nbd->flags))
 294		queue_work(nbd_del_wq, &nbd->remove_work);
 295	else
 296		nbd_dev_remove(nbd);
 297}
 298
 299static int nbd_disconnected(struct nbd_config *config)
 300{
 301	return test_bit(NBD_RT_DISCONNECTED, &config->runtime_flags) ||
 302		test_bit(NBD_RT_DISCONNECT_REQUESTED, &config->runtime_flags);
 303}
 304
 305static void nbd_mark_nsock_dead(struct nbd_device *nbd, struct nbd_sock *nsock,
 306				int notify)
 307{
 308	if (!nsock->dead && notify && !nbd_disconnected(nbd->config)) {
 309		struct link_dead_args *args;
 310		args = kmalloc(sizeof(struct link_dead_args), GFP_NOIO);
 311		if (args) {
 312			INIT_WORK(&args->work, nbd_dead_link_work);
 313			args->index = nbd->index;
 314			queue_work(system_percpu_wq, &args->work);
 315		}
 316	}
 317	if (!nsock->dead) {
 318		kernel_sock_shutdown(nsock->sock, SHUT_RDWR);
 319		if (atomic_dec_return(&nbd->config->live_connections) == 0) {
 320			if (test_and_clear_bit(NBD_RT_DISCONNECT_REQUESTED,
 321					       &nbd->config->runtime_flags)) {
 322				set_bit(NBD_RT_DISCONNECTED,
 323					&nbd->config->runtime_flags);
 324				dev_info(nbd_to_dev(nbd),
 325					"Disconnected due to user request.\n");
 326			}
 327		}
 328	}
 329	nsock->dead = true;
 330	nsock->pending = NULL;
 331	nsock->sent = 0;
 332}
 333
 334static int nbd_set_size(struct nbd_device *nbd, loff_t bytesize, loff_t blksize)
 335{
 336	struct queue_limits lim;
 337	int error;
 338
 339	if (!blksize)
 340		blksize = 1u << NBD_DEF_BLKSIZE_BITS;
 341
 342	if (blk_validate_block_size(blksize))
 343		return -EINVAL;
 344
 345	if (bytesize < 0)
 346		return -EINVAL;
 347
 348	nbd->config->bytesize = bytesize;
 349	nbd->config->blksize_bits = __ffs(blksize);
 350
 351	if (!nbd->pid)
 352		return 0;
 353
 354	lim = queue_limits_start_update(nbd->disk->queue);
 355	if (nbd->config->flags & NBD_FLAG_SEND_TRIM)
 356		lim.max_hw_discard_sectors = UINT_MAX >> SECTOR_SHIFT;
 357	else
 358		lim.max_hw_discard_sectors = 0;
 359	if (!(nbd->config->flags & NBD_FLAG_SEND_FLUSH)) {
 360		lim.features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA);
 361	} else if (nbd->config->flags & NBD_FLAG_SEND_FUA) {
 362		lim.features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA;
 363	} else {
 364		lim.features |= BLK_FEAT_WRITE_CACHE;
 365		lim.features &= ~BLK_FEAT_FUA;
 366	}
 367	if (nbd->config->flags & NBD_FLAG_ROTATIONAL)
 368		lim.features |= BLK_FEAT_ROTATIONAL;
 369	if (nbd->config->flags & NBD_FLAG_SEND_WRITE_ZEROES)
 370		lim.max_write_zeroes_sectors = UINT_MAX >> SECTOR_SHIFT;
 371
 372	lim.logical_block_size = blksize;
 373	lim.physical_block_size = blksize;
 374	error = queue_limits_commit_update_frozen(nbd->disk->queue, &lim);
 375	if (error)
 376		return error;
 377
 378	if (max_part)
 379		set_bit(GD_NEED_PART_SCAN, &nbd->disk->state);
 380	if (!set_capacity_and_notify(nbd->disk, bytesize >> 9))
 381		kobject_uevent(&nbd_to_dev(nbd)->kobj, KOBJ_CHANGE);
 382	return 0;
 383}
 384
 385static void nbd_complete_rq(struct request *req)
 386{
 387	struct nbd_cmd *cmd = blk_mq_rq_to_pdu(req);
 388
 389	dev_dbg(nbd_to_dev(cmd->nbd), "request %p: %s\n", req,
 390		cmd->status ? "failed" : "done");
 391
 392	blk_mq_end_request(req, cmd->status);
 393}
 394
 395/*
 396 * Forcibly shutdown the socket causing all listeners to error
 397 */
 398static void sock_shutdown(struct nbd_device *nbd)
 399{
 400	struct nbd_config *config = nbd->config;
 401	int i;
 402
 403	if (config->num_connections == 0)
 404		return;
 405	if (test_and_set_bit(NBD_RT_DISCONNECTED, &config->runtime_flags))
 406		return;
 407
 408	for (i = 0; i < config->num_connections; i++) {
 409		struct nbd_sock *nsock = config->socks[i];
 410		mutex_lock(&nsock->tx_lock);
 411		nbd_mark_nsock_dead(nbd, nsock, 0);
 412		mutex_unlock(&nsock->tx_lock);
 413	}
 414	dev_warn(disk_to_dev(nbd->disk), "shutting down sockets\n");
 415}
 416
 417static u32 req_to_nbd_cmd_type(struct request *req)
 418{
 419	switch (req_op(req)) {
 420	case REQ_OP_DISCARD:
 421		return NBD_CMD_TRIM;
 422	case REQ_OP_FLUSH:
 423		return NBD_CMD_FLUSH;
 424	case REQ_OP_WRITE:
 425		return NBD_CMD_WRITE;
 426	case REQ_OP_READ:
 427		return NBD_CMD_READ;
 428	case REQ_OP_WRITE_ZEROES:
 429		return NBD_CMD_WRITE_ZEROES;
 430	default:
 431		return U32_MAX;
 432	}
 433}
 434
 435static struct nbd_config *nbd_get_config_unlocked(struct nbd_device *nbd)
 436{
 437	if (refcount_inc_not_zero(&nbd->config_refs)) {
 438		/*
 439		 * Add smp_mb__after_atomic to ensure that reading nbd->config_refs
 440		 * and reading nbd->config is ordered. The pair is the barrier in
 441		 * nbd_alloc_and_init_config(), avoid nbd->config_refs is set
 442		 * before nbd->config.
 443		 */
 444		smp_mb__after_atomic();
 445		return nbd->config;
 446	}
 447
 448	return NULL;
 449}
 450
 451static enum blk_eh_timer_return nbd_xmit_timeout(struct request *req)
 452{
 453	struct nbd_cmd *cmd = blk_mq_rq_to_pdu(req);
 454	struct nbd_device *nbd = cmd->nbd;
 455	struct nbd_config *config;
 456
 457	if (!mutex_trylock(&cmd->lock))
 458		return BLK_EH_RESET_TIMER;
 459
 460	/* partial send is handled in nbd_sock's work function */
 461	if (test_bit(NBD_CMD_PARTIAL_SEND, &cmd->flags)) {
 462		mutex_unlock(&cmd->lock);
 463		return BLK_EH_RESET_TIMER;
 464	}
 465
 466	if (!test_bit(NBD_CMD_INFLIGHT, &cmd->flags)) {
 467		mutex_unlock(&cmd->lock);
 468		return BLK_EH_DONE;
 469	}
 470
 471	config = nbd_get_config_unlocked(nbd);
 472	if (!config) {
 473		cmd->status = BLK_STS_TIMEOUT;
 474		__clear_bit(NBD_CMD_INFLIGHT, &cmd->flags);
 475		mutex_unlock(&cmd->lock);
 476		goto done;
 477	}
 478
 479	if (config->num_connections > 1 ||
 480	    (config->num_connections == 1 && nbd->tag_set.timeout)) {
 481		dev_err_ratelimited(nbd_to_dev(nbd),
 482				    "Connection timed out, retrying (%d/%d alive)\n",
 483				    atomic_read(&config->live_connections),
 484				    config->num_connections);
 485		/*
 486		 * Hooray we have more connections, requeue this IO, the submit
 487		 * path will put it on a real connection. Or if only one
 488		 * connection is configured, the submit path will wait util
 489		 * a new connection is reconfigured or util dead timeout.
 490		 */
 491		if (config->socks) {
 492			if (cmd->index < config->num_connections) {
 493				struct nbd_sock *nsock =
 494					config->socks[cmd->index];
 495				mutex_lock(&nsock->tx_lock);
 496				/* We can have multiple outstanding requests, so
 497				 * we don't want to mark the nsock dead if we've
 498				 * already reconnected with a new socket, so
 499				 * only mark it dead if its the same socket we
 500				 * were sent out on.
 501				 */
 502				if (cmd->cookie == nsock->cookie)
 503					nbd_mark_nsock_dead(nbd, nsock, 1);
 504				mutex_unlock(&nsock->tx_lock);
 505			}
 506			nbd_requeue_cmd(cmd);
 507			mutex_unlock(&cmd->lock);
 508			nbd_config_put(nbd);
 509			return BLK_EH_DONE;
 510		}
 511	}
 512
 513	if (!nbd->tag_set.timeout) {
 514		/*
 515		 * Userspace sets timeout=0 to disable socket disconnection,
 516		 * so just warn and reset the timer.
 517		 */
 518		struct nbd_sock *nsock = config->socks[cmd->index];
 519		cmd->retries++;
 520		dev_info(nbd_to_dev(nbd), "Possible stuck request %p: control (%s@%llu,%uB). Runtime %u seconds\n",
 521			req, nbdcmd_to_ascii(req_to_nbd_cmd_type(req)),
 522			(unsigned long long)blk_rq_pos(req) << 9,
 523			blk_rq_bytes(req), (req->timeout / HZ) * cmd->retries);
 524
 525		mutex_lock(&nsock->tx_lock);
 526		if (cmd->cookie != nsock->cookie) {
 527			nbd_requeue_cmd(cmd);
 528			mutex_unlock(&nsock->tx_lock);
 529			mutex_unlock(&cmd->lock);
 530			nbd_config_put(nbd);
 531			return BLK_EH_DONE;
 532		}
 533		mutex_unlock(&nsock->tx_lock);
 534		mutex_unlock(&cmd->lock);
 535		nbd_config_put(nbd);
 536		return BLK_EH_RESET_TIMER;
 537	}
 538
 539	dev_err_ratelimited(nbd_to_dev(nbd), "Connection timed out\n");
 540	set_bit(NBD_RT_TIMEDOUT, &config->runtime_flags);
 541	cmd->status = BLK_STS_IOERR;
 542	__clear_bit(NBD_CMD_INFLIGHT, &cmd->flags);
 543	mutex_unlock(&cmd->lock);
 544	sock_shutdown(nbd);
 545	nbd_config_put(nbd);
 546done:
 547	blk_mq_complete_request(req);
 548	return BLK_EH_DONE;
 549}
 550
 551static int __sock_xmit(struct nbd_device *nbd, struct socket *sock, int send,
 552		       struct iov_iter *iter, int msg_flags, int *sent)
 553{
 554	int result;
 555	struct msghdr msg = {} ;
 556	unsigned int noreclaim_flag;
 557
 558	if (unlikely(!sock)) {
 559		dev_err_ratelimited(disk_to_dev(nbd->disk),
 560			"Attempted %s on closed socket in sock_xmit\n",
 561			(send ? "send" : "recv"));
 562		return -EINVAL;
 563	}
 564
 565	msg.msg_iter = *iter;
 566
 567	noreclaim_flag = memalloc_noreclaim_save();
 568
 569	scoped_with_kernel_creds() {
 570		do {
 571			sock->sk->sk_allocation = GFP_NOIO | __GFP_MEMALLOC;
 572			sock->sk->sk_use_task_frag = false;
 573			msg.msg_flags = msg_flags | MSG_NOSIGNAL;
 574
 575			if (send)
 576				result = sock_sendmsg(sock, &msg);
 577			else
 578				result = sock_recvmsg(sock, &msg, msg.msg_flags);
 579
 580			if (result <= 0) {
 581				if (result == 0)
 582					result = -EPIPE; /* short read */
 583				break;
 584			}
 585			if (sent)
 586				*sent += result;
 587		} while (msg_data_left(&msg));
 588	}
 589
 590	memalloc_noreclaim_restore(noreclaim_flag);
 591
 592	return result;
 593}
 594
 595/*
 596 *  Send or receive packet. Return a positive value on success and
 597 *  negtive value on failure, and never return 0.
 598 */
 599static int sock_xmit(struct nbd_device *nbd, int index, int send,
 600		     struct iov_iter *iter, int msg_flags, int *sent)
 601{
 602	struct nbd_config *config = nbd->config;
 603	struct socket *sock = config->socks[index]->sock;
 604
 605	return __sock_xmit(nbd, sock, send, iter, msg_flags, sent);
 606}
 607
 608/*
 609 * Different settings for sk->sk_sndtimeo can result in different return values
 610 * if there is a signal pending when we enter sendmsg, because reasons?
 611 */
 612static inline int was_interrupted(int result)
 613{
 614	return result == -ERESTARTSYS || result == -EINTR;
 615}
 616
 617/*
 618 * We've already sent header or part of data payload, have no choice but
 619 * to set pending and schedule it in work.
 620 *
 621 * And we have to return BLK_STS_OK to block core, otherwise this same
 622 * request may be re-dispatched with different tag, but our header has
 623 * been sent out with old tag, and this way does confuse reply handling.
 624 */
 625static void nbd_sched_pending_work(struct nbd_device *nbd,
 626				   struct nbd_sock *nsock,
 627				   struct nbd_cmd *cmd, int sent)
 628{
 629	struct request *req = blk_mq_rq_from_pdu(cmd);
 630
 631	/* pending work should be scheduled only once */
 632	WARN_ON_ONCE(test_bit(NBD_CMD_PARTIAL_SEND, &cmd->flags));
 633
 634	nsock->pending = req;
 635	nsock->sent = sent;
 636	set_bit(NBD_CMD_PARTIAL_SEND, &cmd->flags);
 637	refcount_inc(&nbd->config_refs);
 638	schedule_work(&nsock->work);
 639}
 640
 641/*
 642 * Returns BLK_STS_RESOURCE if the caller should retry after a delay.
 643 * Returns BLK_STS_IOERR if sending failed.
 644 */
 645static blk_status_t nbd_send_cmd(struct nbd_device *nbd, struct nbd_cmd *cmd,
 646				 int index)
 647{
 648	struct request *req = blk_mq_rq_from_pdu(cmd);
 649	struct nbd_config *config = nbd->config;
 650	struct nbd_sock *nsock = config->socks[index];
 651	int result;
 652	struct nbd_request request = {.magic = htonl(NBD_REQUEST_MAGIC)};
 653	struct kvec iov = {.iov_base = &request, .iov_len = sizeof(request)};
 654	struct iov_iter from;
 655	struct bio *bio;
 656	u64 handle;
 657	u32 type;
 658	u32 nbd_cmd_flags = 0;
 659	int sent = nsock->sent, skip = 0;
 660
 661	lockdep_assert_held(&cmd->lock);
 662	lockdep_assert_held(&nsock->tx_lock);
 663
 664	iov_iter_kvec(&from, ITER_SOURCE, &iov, 1, sizeof(request));
 665
 666	type = req_to_nbd_cmd_type(req);
 667	if (type == U32_MAX)
 668		return BLK_STS_IOERR;
 669
 670	if (rq_data_dir(req) == WRITE &&
 671	    (config->flags & NBD_FLAG_READ_ONLY)) {
 672		dev_err_ratelimited(disk_to_dev(nbd->disk),
 673				    "Write on read-only\n");
 674		return BLK_STS_IOERR;
 675	}
 676
 677	if (req->cmd_flags & REQ_FUA)
 678		nbd_cmd_flags |= NBD_CMD_FLAG_FUA;
 679	if ((req->cmd_flags & REQ_NOUNMAP) && (type == NBD_CMD_WRITE_ZEROES))
 680		nbd_cmd_flags |= NBD_CMD_FLAG_NO_HOLE;
 681
 682	/* We did a partial send previously, and we at least sent the whole
 683	 * request struct, so just go and send the rest of the pages in the
 684	 * request.
 685	 */
 686	if (sent) {
 687		if (sent >= sizeof(request)) {
 688			skip = sent - sizeof(request);
 689
 690			/* initialize handle for tracing purposes */
 691			handle = nbd_cmd_handle(cmd);
 692
 693			goto send_pages;
 694		}
 695		iov_iter_advance(&from, sent);
 696	} else {
 697		cmd->cmd_cookie++;
 698	}
 699	cmd->index = index;
 700	cmd->cookie = nsock->cookie;
 701	cmd->retries = 0;
 702	request.type = htonl(type | nbd_cmd_flags);
 703	if (type != NBD_CMD_FLUSH) {
 704		request.from = cpu_to_be64((u64)blk_rq_pos(req) << 9);
 705		request.len = htonl(blk_rq_bytes(req));
 706	}
 707	handle = nbd_cmd_handle(cmd);
 708	request.cookie = cpu_to_be64(handle);
 709
 710	trace_nbd_send_request(&request, nbd->index, blk_mq_rq_from_pdu(cmd));
 711
 712	dev_dbg(nbd_to_dev(nbd), "request %p: sending control (%s@%llu,%uB)\n",
 713		req, nbdcmd_to_ascii(type),
 714		(unsigned long long)blk_rq_pos(req) << 9, blk_rq_bytes(req));
 715	result = sock_xmit(nbd, index, 1, &from,
 716			(type == NBD_CMD_WRITE) ? MSG_MORE : 0, &sent);
 717	trace_nbd_header_sent(req, handle);
 718	if (result < 0) {
 719		if (was_interrupted(result)) {
 720			/* If we haven't sent anything we can just return BUSY,
 721			 * however if we have sent something we need to make
 722			 * sure we only allow this req to be sent until we are
 723			 * completely done.
 724			 */
 725			if (sent) {
 726				nbd_sched_pending_work(nbd, nsock, cmd, sent);
 727				return BLK_STS_OK;
 728			}
 729			set_bit(NBD_CMD_REQUEUED, &cmd->flags);
 730			return BLK_STS_RESOURCE;
 731		}
 732		dev_err_ratelimited(disk_to_dev(nbd->disk),
 733			"Send control failed (result %d)\n", result);
 734		goto requeue;
 735	}
 736send_pages:
 737	if (type != NBD_CMD_WRITE)
 738		goto out;
 739
 740	bio = req->bio;
 741	while (bio) {
 742		struct bio *next = bio->bi_next;
 743		struct bvec_iter iter;
 744		struct bio_vec bvec;
 745
 746		bio_for_each_segment(bvec, bio, iter) {
 747			bool is_last = !next && bio_iter_last(bvec, iter);
 748			int flags = is_last ? 0 : MSG_MORE;
 749
 750			dev_dbg(nbd_to_dev(nbd), "request %p: sending %d bytes data\n",
 751				req, bvec.bv_len);
 752			iov_iter_bvec(&from, ITER_SOURCE, &bvec, 1, bvec.bv_len);
 753			if (skip) {
 754				if (skip >= iov_iter_count(&from)) {
 755					skip -= iov_iter_count(&from);
 756					continue;
 757				}
 758				iov_iter_advance(&from, skip);
 759				skip = 0;
 760			}
 761			result = sock_xmit(nbd, index, 1, &from, flags, &sent);
 762			if (result < 0) {
 763				if (was_interrupted(result)) {
 764					nbd_sched_pending_work(nbd, nsock, cmd, sent);
 765					return BLK_STS_OK;
 766				}
 767				dev_err(disk_to_dev(nbd->disk),
 768					"Send data failed (result %d)\n",
 769					result);
 770				goto requeue;
 771			}
 772			/*
 773			 * The completion might already have come in,
 774			 * so break for the last one instead of letting
 775			 * the iterator do it. This prevents use-after-free
 776			 * of the bio.
 777			 */
 778			if (is_last)
 779				break;
 780		}
 781		bio = next;
 782	}
 783out:
 784	trace_nbd_payload_sent(req, handle);
 785	nsock->pending = NULL;
 786	nsock->sent = 0;
 787	__set_bit(NBD_CMD_INFLIGHT, &cmd->flags);
 788	return BLK_STS_OK;
 789
 790requeue:
 791	/*
 792	 * Can't requeue in case we are dealing with partial send
 793	 *
 794	 * We must run from pending work function.
 795	 * */
 796	if (test_bit(NBD_CMD_PARTIAL_SEND, &cmd->flags))
 797		return BLK_STS_OK;
 798
 799	/* retry on a different socket */
 800	dev_err_ratelimited(disk_to_dev(nbd->disk),
 801			    "Request send failed, requeueing\n");
 802	nbd_mark_nsock_dead(nbd, nsock, 1);
 803	nbd_requeue_cmd(cmd);
 804	return BLK_STS_OK;
 805}
 806
 807/* handle partial sending */
 808static void nbd_pending_cmd_work(struct work_struct *work)
 809{
 810	struct nbd_sock *nsock = container_of(work, struct nbd_sock, work);
 811	struct request *req = nsock->pending;
 812	struct nbd_cmd *cmd = blk_mq_rq_to_pdu(req);
 813	struct nbd_device *nbd = cmd->nbd;
 814	unsigned long deadline = READ_ONCE(req->deadline);
 815	unsigned int wait_ms = 2;
 816
 817	mutex_lock(&cmd->lock);
 818
 819	WARN_ON_ONCE(test_bit(NBD_CMD_REQUEUED, &cmd->flags));
 820	if (WARN_ON_ONCE(!test_bit(NBD_CMD_PARTIAL_SEND, &cmd->flags)))
 821		goto out;
 822
 823	mutex_lock(&nsock->tx_lock);
 824	while (true) {
 825		nbd_send_cmd(nbd, cmd, cmd->index);
 826		if (!nsock->pending)
 827			break;
 828
 829		/* don't bother timeout handler for partial sending */
 830		if (READ_ONCE(jiffies) + msecs_to_jiffies(wait_ms) >= deadline) {
 831			cmd->status = BLK_STS_IOERR;
 832			blk_mq_complete_request(req);
 833			break;
 834		}
 835		msleep(wait_ms);
 836		wait_ms *= 2;
 837	}
 838	mutex_unlock(&nsock->tx_lock);
 839	clear_bit(NBD_CMD_PARTIAL_SEND, &cmd->flags);
 840out:
 841	mutex_unlock(&cmd->lock);
 842	nbd_config_put(nbd);
 843}
 844
 845static int nbd_read_reply(struct nbd_device *nbd, struct socket *sock,
 846			  struct nbd_reply *reply)
 847{
 848	struct kvec iov = {.iov_base = reply, .iov_len = sizeof(*reply)};
 849	struct iov_iter to;
 850	int result;
 851
 852	reply->magic = 0;
 853	iov_iter_kvec(&to, ITER_DEST, &iov, 1, sizeof(*reply));
 854	result = __sock_xmit(nbd, sock, 0, &to, MSG_WAITALL, NULL);
 855	if (result < 0) {
 856		if (!nbd_disconnected(nbd->config))
 857			dev_err(disk_to_dev(nbd->disk),
 858				"Receive control failed (result %d)\n", result);
 859		return result;
 860	}
 861
 862	if (ntohl(reply->magic) != NBD_REPLY_MAGIC) {
 863		dev_err(disk_to_dev(nbd->disk), "Wrong magic (0x%lx)\n",
 864				(unsigned long)ntohl(reply->magic));
 865		return -EPROTO;
 866	}
 867
 868	return 0;
 869}
 870
 871/* NULL returned = something went wrong, inform userspace */
 872static struct nbd_cmd *nbd_handle_reply(struct nbd_device *nbd, int index,
 873					struct nbd_reply *reply)
 874{
 875	int result;
 876	struct nbd_cmd *cmd;
 877	struct request *req = NULL;
 878	u64 handle;
 879	u16 hwq;
 880	u32 tag;
 881	int ret = 0;
 882
 883	handle = be64_to_cpu(reply->cookie);
 884	tag = nbd_handle_to_tag(handle);
 885	hwq = blk_mq_unique_tag_to_hwq(tag);
 886	if (hwq < nbd->tag_set.nr_hw_queues)
 887		req = blk_mq_tag_to_rq(nbd->tag_set.tags[hwq],
 888				       blk_mq_unique_tag_to_tag(tag));
 889	if (!req || !blk_mq_request_started(req)) {
 890		dev_err(disk_to_dev(nbd->disk), "Unexpected reply (%d) %p\n",
 891			tag, req);
 892		return ERR_PTR(-ENOENT);
 893	}
 894	trace_nbd_header_received(req, handle);
 895	cmd = blk_mq_rq_to_pdu(req);
 896
 897	mutex_lock(&cmd->lock);
 898	if (!test_bit(NBD_CMD_INFLIGHT, &cmd->flags)) {
 899		dev_err(disk_to_dev(nbd->disk), "Suspicious reply %d (status %u flags %lu)",
 900			tag, cmd->status, cmd->flags);
 901		ret = -ENOENT;
 902		goto out;
 903	}
 904	if (cmd->index != index) {
 905		dev_err(disk_to_dev(nbd->disk), "Unexpected reply %d from different sock %d (expected %d)",
 906			tag, index, cmd->index);
 907		ret = -ENOENT;
 908		goto out;
 909	}
 910	if (cmd->cmd_cookie != nbd_handle_to_cookie(handle)) {
 911		dev_err(disk_to_dev(nbd->disk), "Double reply on req %p, cmd_cookie %u, handle cookie %u\n",
 912			req, cmd->cmd_cookie, nbd_handle_to_cookie(handle));
 913		ret = -ENOENT;
 914		goto out;
 915	}
 916	if (cmd->status != BLK_STS_OK) {
 917		dev_err(disk_to_dev(nbd->disk), "Command already handled %p\n",
 918			req);
 919		ret = -ENOENT;
 920		goto out;
 921	}
 922	if (test_bit(NBD_CMD_REQUEUED, &cmd->flags)) {
 923		dev_err(disk_to_dev(nbd->disk), "Raced with timeout on req %p\n",
 924			req);
 925		ret = -ENOENT;
 926		goto out;
 927	}
 928	if (ntohl(reply->error)) {
 929		dev_err(disk_to_dev(nbd->disk), "Other side returned error (%d)\n",
 930			ntohl(reply->error));
 931		cmd->status = BLK_STS_IOERR;
 932		goto out;
 933	}
 934
 935	dev_dbg(nbd_to_dev(nbd), "request %p: got reply\n", req);
 936	if (rq_data_dir(req) != WRITE) {
 937		struct req_iterator iter;
 938		struct bio_vec bvec;
 939		struct iov_iter to;
 940
 941		rq_for_each_segment(bvec, req, iter) {
 942			iov_iter_bvec(&to, ITER_DEST, &bvec, 1, bvec.bv_len);
 943			result = sock_xmit(nbd, index, 0, &to, MSG_WAITALL, NULL);
 944			if (result < 0) {
 945				dev_err(disk_to_dev(nbd->disk), "Receive data failed (result %d)\n",
 946					result);
 947				/*
 948				 * If we've disconnected, we need to make sure we
 949				 * complete this request, otherwise error out
 950				 * and let the timeout stuff handle resubmitting
 951				 * this request onto another connection.
 952				 */
 953				if (nbd_disconnected(nbd->config)) {
 954					cmd->status = BLK_STS_IOERR;
 955					goto out;
 956				}
 957				ret = -EIO;
 958				goto out;
 959			}
 960			dev_dbg(nbd_to_dev(nbd), "request %p: got %d bytes data\n",
 961				req, bvec.bv_len);
 962		}
 963	}
 964out:
 965	trace_nbd_payload_received(req, handle);
 966	mutex_unlock(&cmd->lock);
 967	return ret ? ERR_PTR(ret) : cmd;
 968}
 969
 970static void recv_work(struct work_struct *work)
 971{
 972	struct recv_thread_args *args = container_of(work,
 973						     struct recv_thread_args,
 974						     work);
 975	struct nbd_device *nbd = args->nbd;
 976	struct nbd_config *config = nbd->config;
 977	struct request_queue *q = nbd->disk->queue;
 978	struct nbd_sock *nsock = args->nsock;
 979	struct nbd_cmd *cmd;
 980	struct request *rq;
 981
 982	while (1) {
 983		struct nbd_reply reply;
 984
 985		if (nbd_read_reply(nbd, nsock->sock, &reply))
 986			break;
 987
 988		/*
 989		 * Grab .q_usage_counter so request pool won't go away, then no
 990		 * request use-after-free is possible during nbd_handle_reply().
 991		 * If queue is frozen, there won't be any inflight requests, we
 992		 * needn't to handle the incoming garbage message.
 993		 */
 994		if (!percpu_ref_tryget(&q->q_usage_counter)) {
 995			dev_err(disk_to_dev(nbd->disk), "%s: no io inflight\n",
 996				__func__);
 997			break;
 998		}
 999
1000		cmd = nbd_handle_reply(nbd, args->index, &reply);
1001		if (IS_ERR(cmd)) {
1002			percpu_ref_put(&q->q_usage_counter);
1003			break;
1004		}
1005
1006		rq = blk_mq_rq_from_pdu(cmd);
1007		if (likely(!blk_should_fake_timeout(rq->q))) {
1008			bool complete;
1009
1010			mutex_lock(&cmd->lock);
1011			complete = __test_and_clear_bit(NBD_CMD_INFLIGHT,
1012							&cmd->flags);
1013			mutex_unlock(&cmd->lock);
1014			if (complete)
1015				blk_mq_complete_request(rq);
1016		}
1017		percpu_ref_put(&q->q_usage_counter);
1018	}
1019
1020	mutex_lock(&nsock->tx_lock);
1021	nbd_mark_nsock_dead(nbd, nsock, 1);
1022	mutex_unlock(&nsock->tx_lock);
1023
1024	atomic_dec(&config->recv_threads);
1025	wake_up(&config->recv_wq);
1026	nbd_config_put(nbd);
1027	kfree(args);
1028}
1029
1030static bool nbd_clear_req(struct request *req, void *data)
1031{
1032	struct nbd_cmd *cmd = blk_mq_rq_to_pdu(req);
1033
1034	/* don't abort one completed request */
1035	if (blk_mq_request_completed(req))
1036		return true;
1037
1038	mutex_lock(&cmd->lock);
1039	if (!__test_and_clear_bit(NBD_CMD_INFLIGHT, &cmd->flags)) {
1040		mutex_unlock(&cmd->lock);
1041		return true;
1042	}
1043	cmd->status = BLK_STS_IOERR;
1044	mutex_unlock(&cmd->lock);
1045
1046	blk_mq_complete_request(req);
1047	return true;
1048}
1049
1050static void nbd_clear_que(struct nbd_device *nbd)
1051{
1052	blk_mq_quiesce_queue(nbd->disk->queue);
1053	blk_mq_tagset_busy_iter(&nbd->tag_set, nbd_clear_req, NULL);
1054	blk_mq_unquiesce_queue(nbd->disk->queue);
1055	dev_dbg(disk_to_dev(nbd->disk), "queue cleared\n");
1056}
1057
1058static int find_fallback(struct nbd_device *nbd, int index)
1059{
1060	struct nbd_config *config = nbd->config;
1061	int new_index = -1;
1062	struct nbd_sock *nsock = config->socks[index];
1063	int fallback = nsock->fallback_index;
1064
1065	if (test_bit(NBD_RT_DISCONNECTED, &config->runtime_flags))
1066		return new_index;
1067
1068	if (config->num_connections <= 1) {
1069		dev_err_ratelimited(disk_to_dev(nbd->disk),
1070				    "Dead connection, failed to find a fallback\n");
1071		return new_index;
1072	}
1073
1074	if (fallback >= 0 && fallback < config->num_connections &&
1075	    !config->socks[fallback]->dead)
1076		return fallback;
1077
1078	if (nsock->fallback_index < 0 ||
1079	    nsock->fallback_index >= config->num_connections ||
1080	    config->socks[nsock->fallback_index]->dead) {
1081		int i;
1082		for (i = 0; i < config->num_connections; i++) {
1083			if (i == index)
1084				continue;
1085			if (!config->socks[i]->dead) {
1086				new_index = i;
1087				break;
1088			}
1089		}
1090		nsock->fallback_index = new_index;
1091		if (new_index < 0) {
1092			dev_err_ratelimited(disk_to_dev(nbd->disk),
1093					    "Dead connection, failed to find a fallback\n");
1094			return new_index;
1095		}
1096	}
1097	new_index = nsock->fallback_index;
1098	return new_index;
1099}
1100
1101static int wait_for_reconnect(struct nbd_device *nbd)
1102{
1103	struct nbd_config *config = nbd->config;
1104	if (!config->dead_conn_timeout)
1105		return 0;
1106
1107	if (!wait_event_timeout(config->conn_wait,
1108				test_bit(NBD_RT_DISCONNECTED,
1109					 &config->runtime_flags) ||
1110				atomic_read(&config->live_connections) > 0,
1111				config->dead_conn_timeout))
1112		return 0;
1113
1114	return !test_bit(NBD_RT_DISCONNECTED, &config->runtime_flags);
1115}
1116
1117static blk_status_t nbd_handle_cmd(struct nbd_cmd *cmd, int index)
1118{
1119	struct request *req = blk_mq_rq_from_pdu(cmd);
1120	struct nbd_device *nbd = cmd->nbd;
1121	struct nbd_config *config;
1122	struct nbd_sock *nsock;
1123	blk_status_t ret;
1124
1125	lockdep_assert_held(&cmd->lock);
1126
1127	config = nbd_get_config_unlocked(nbd);
1128	if (!config) {
1129		dev_err_ratelimited(disk_to_dev(nbd->disk),
1130				    "Socks array is empty\n");
1131		return BLK_STS_IOERR;
1132	}
1133
1134	if (index >= config->num_connections) {
1135		dev_err_ratelimited(disk_to_dev(nbd->disk),
1136				    "Attempted send on invalid socket\n");
1137		nbd_config_put(nbd);
1138		return BLK_STS_IOERR;
1139	}
1140	cmd->status = BLK_STS_OK;
1141again:
1142	nsock = config->socks[index];
1143	mutex_lock(&nsock->tx_lock);
1144	if (nsock->dead) {
1145		int old_index = index;
1146		index = find_fallback(nbd, index);
1147		mutex_unlock(&nsock->tx_lock);
1148		if (index < 0) {
1149			if (wait_for_reconnect(nbd)) {
1150				index = old_index;
1151				goto again;
1152			}
1153			/* All the sockets should already be down at this point,
1154			 * we just want to make sure that DISCONNECTED is set so
1155			 * any requests that come in that were queue'ed waiting
1156			 * for the reconnect timer don't trigger the timer again
1157			 * and instead just error out.
1158			 */
1159			sock_shutdown(nbd);
1160			nbd_config_put(nbd);
1161			return BLK_STS_IOERR;
1162		}
1163		goto again;
1164	}
1165
1166	/* Handle the case that we have a pending request that was partially
1167	 * transmitted that _has_ to be serviced first.  We need to call requeue
1168	 * here so that it gets put _after_ the request that is already on the
1169	 * dispatch list.
1170	 */
1171	blk_mq_start_request(req);
1172	if (unlikely(nsock->pending && nsock->pending != req)) {
1173		nbd_requeue_cmd(cmd);
1174		ret = BLK_STS_OK;
1175		goto out;
1176	}
1177	ret = nbd_send_cmd(nbd, cmd, index);
1178out:
1179	mutex_unlock(&nsock->tx_lock);
1180	nbd_config_put(nbd);
1181	return ret;
1182}
1183
1184static blk_status_t nbd_queue_rq(struct blk_mq_hw_ctx *hctx,
1185			const struct blk_mq_queue_data *bd)
1186{
1187	struct nbd_cmd *cmd = blk_mq_rq_to_pdu(bd->rq);
1188	blk_status_t ret;
1189
1190	/*
1191	 * Since we look at the bio's to send the request over the network we
1192	 * need to make sure the completion work doesn't mark this request done
1193	 * before we are done doing our send.  This keeps us from dereferencing
1194	 * freed data if we have particularly fast completions (ie we get the
1195	 * completion before we exit sock_xmit on the last bvec) or in the case
1196	 * that the server is misbehaving (or there was an error) before we're
1197	 * done sending everything over the wire.
1198	 */
1199	mutex_lock(&cmd->lock);
1200	clear_bit(NBD_CMD_REQUEUED, &cmd->flags);
1201
1202	/* We can be called directly from the user space process, which means we
1203	 * could possibly have signals pending so our sendmsg will fail.  In
1204	 * this case we need to return that we are busy, otherwise error out as
1205	 * appropriate.
1206	 */
1207	ret = nbd_handle_cmd(cmd, hctx->queue_num);
1208	mutex_unlock(&cmd->lock);
1209
1210	return ret;
1211}
1212
1213static struct socket *nbd_get_socket(struct nbd_device *nbd, unsigned long fd,
1214				     int *err)
1215{
1216	struct socket *sock;
1217
1218	*err = 0;
1219	sock = sockfd_lookup(fd, err);
1220	if (!sock)
1221		return NULL;
1222
1223	if (!sk_is_tcp(sock->sk) &&
1224	    !sk_is_stream_unix(sock->sk)) {
1225		dev_err(disk_to_dev(nbd->disk), "Unsupported socket: should be TCP or UNIX.\n");
1226		*err = -EINVAL;
1227		sockfd_put(sock);
1228		return NULL;
1229	}
1230
1231	if (sock->ops->shutdown == sock_no_shutdown) {
1232		dev_err(disk_to_dev(nbd->disk), "Unsupported socket: shutdown callout must be supported.\n");
1233		*err = -EINVAL;
1234		sockfd_put(sock);
1235		return NULL;
1236	}
1237
1238	return sock;
1239}
1240
1241static int nbd_add_socket(struct nbd_device *nbd, unsigned long arg,
1242			  bool netlink)
1243{
1244	struct nbd_config *config = nbd->config;
1245	struct socket *sock;
1246	struct nbd_sock **socks;
1247	struct nbd_sock *nsock;
1248	unsigned int memflags;
1249	int err;
1250
1251	/* Arg will be cast to int, check it to avoid overflow */
1252	if (arg > INT_MAX)
1253		return -EINVAL;
1254	sock = nbd_get_socket(nbd, arg, &err);
1255	if (!sock)
1256		return err;
1257
1258	/*
1259	 * We need to make sure we don't get any errant requests while we're
1260	 * reallocating the ->socks array.
1261	 */
1262	memflags = blk_mq_freeze_queue(nbd->disk->queue);
1263
1264	if (!netlink && !nbd->task_setup &&
1265	    !test_bit(NBD_RT_BOUND, &config->runtime_flags))
1266		nbd->task_setup = current;
1267
1268	if (!netlink &&
1269	    (nbd->task_setup != current ||
1270	     test_bit(NBD_RT_BOUND, &config->runtime_flags))) {
1271		dev_err(disk_to_dev(nbd->disk),
1272			"Device being setup by another task");
1273		err = -EBUSY;
1274		goto put_socket;
1275	}
1276
1277	nsock = kzalloc(sizeof(*nsock), GFP_KERNEL);
1278	if (!nsock) {
1279		err = -ENOMEM;
1280		goto put_socket;
1281	}
1282
1283	socks = krealloc(config->socks, (config->num_connections + 1) *
1284			 sizeof(struct nbd_sock *), GFP_KERNEL);
1285	if (!socks) {
1286		kfree(nsock);
1287		err = -ENOMEM;
1288		goto put_socket;
1289	}
1290
1291	config->socks = socks;
1292
1293	nsock->fallback_index = -1;
1294	nsock->dead = false;
1295	mutex_init(&nsock->tx_lock);
1296	nsock->sock = sock;
1297	nsock->pending = NULL;
1298	nsock->sent = 0;
1299	nsock->cookie = 0;
1300	INIT_WORK(&nsock->work, nbd_pending_cmd_work);
1301	socks[config->num_connections++] = nsock;
1302	atomic_inc(&config->live_connections);
1303	blk_mq_unfreeze_queue(nbd->disk->queue, memflags);
1304
1305	return 0;
1306
1307put_socket:
1308	blk_mq_unfreeze_queue(nbd->disk->queue, memflags);
1309	sockfd_put(sock);
1310	return err;
1311}
1312
1313static int nbd_reconnect_socket(struct nbd_device *nbd, unsigned long arg)
1314{
1315	struct nbd_config *config = nbd->config;
1316	struct socket *sock, *old;
1317	struct recv_thread_args *args;
1318	int i;
1319	int err;
1320
1321	sock = nbd_get_socket(nbd, arg, &err);
1322	if (!sock)
1323		return err;
1324
1325	args = kzalloc(sizeof(*args), GFP_KERNEL);
1326	if (!args) {
1327		sockfd_put(sock);
1328		return -ENOMEM;
1329	}
1330
1331	for (i = 0; i < config->num_connections; i++) {
1332		struct nbd_sock *nsock = config->socks[i];
1333
1334		if (!nsock->dead)
1335			continue;
1336
1337		mutex_lock(&nsock->tx_lock);
1338		if (!nsock->dead) {
1339			mutex_unlock(&nsock->tx_lock);
1340			continue;
1341		}
1342		sk_set_memalloc(sock->sk);
1343		if (nbd->tag_set.timeout)
1344			sock->sk->sk_sndtimeo = nbd->tag_set.timeout;
1345		atomic_inc(&config->recv_threads);
1346		refcount_inc(&nbd->config_refs);
1347		old = nsock->sock;
1348		nsock->fallback_index = -1;
1349		nsock->sock = sock;
1350		nsock->dead = false;
1351		INIT_WORK(&args->work, recv_work);
1352		args->index = i;
1353		args->nbd = nbd;
1354		args->nsock = nsock;
1355		nsock->cookie++;
1356		mutex_unlock(&nsock->tx_lock);
1357		sockfd_put(old);
1358
1359		clear_bit(NBD_RT_DISCONNECTED, &config->runtime_flags);
1360
1361		/* We take the tx_mutex in an error path in the recv_work, so we
1362		 * need to queue_work outside of the tx_mutex.
1363		 */
1364		queue_work(nbd->recv_workq, &args->work);
1365
1366		atomic_inc(&config->live_connections);
1367		wake_up(&config->conn_wait);
1368		return 0;
1369	}
1370	sockfd_put(sock);
1371	kfree(args);
1372	return -ENOSPC;
1373}
1374
1375static void nbd_bdev_reset(struct nbd_device *nbd)
1376{
1377	if (disk_openers(nbd->disk) > 1)
1378		return;
1379	set_capacity(nbd->disk, 0);
1380}
1381
1382static void nbd_parse_flags(struct nbd_device *nbd)
1383{
1384	if (nbd->config->flags & NBD_FLAG_READ_ONLY)
1385		set_disk_ro(nbd->disk, true);
1386	else
1387		set_disk_ro(nbd->disk, false);
1388}
1389
1390static void send_disconnects(struct nbd_device *nbd)
1391{
1392	struct nbd_config *config = nbd->config;
1393	struct nbd_request request = {
1394		.magic = htonl(NBD_REQUEST_MAGIC),
1395		.type = htonl(NBD_CMD_DISC),
1396	};
1397	struct kvec iov = {.iov_base = &request, .iov_len = sizeof(request)};
1398	struct iov_iter from;
1399	int i, ret;
1400
1401	for (i = 0; i < config->num_connections; i++) {
1402		struct nbd_sock *nsock = config->socks[i];
1403
1404		iov_iter_kvec(&from, ITER_SOURCE, &iov, 1, sizeof(request));
1405		mutex_lock(&nsock->tx_lock);
1406		ret = sock_xmit(nbd, i, 1, &from, 0, NULL);
1407		if (ret < 0)
1408			dev_err(disk_to_dev(nbd->disk),
1409				"Send disconnect failed %d\n", ret);
1410		mutex_unlock(&nsock->tx_lock);
1411	}
1412}
1413
1414static int nbd_disconnect(struct nbd_device *nbd)
1415{
1416	struct nbd_config *config = nbd->config;
1417
1418	dev_info(disk_to_dev(nbd->disk), "NBD_DISCONNECT\n");
1419	set_bit(NBD_RT_DISCONNECT_REQUESTED, &config->runtime_flags);
1420	set_bit(NBD_DISCONNECT_REQUESTED, &nbd->flags);
1421	send_disconnects(nbd);
1422	return 0;
1423}
1424
1425static void nbd_clear_sock(struct nbd_device *nbd)
1426{
1427	sock_shutdown(nbd);
1428	nbd_clear_que(nbd);
1429	nbd->task_setup = NULL;
1430}
1431
1432static void nbd_config_put(struct nbd_device *nbd)
1433{
1434	if (refcount_dec_and_mutex_lock(&nbd->config_refs,
1435					&nbd->config_lock)) {
1436		struct nbd_config *config = nbd->config;
1437		nbd_dev_dbg_close(nbd);
1438		invalidate_disk(nbd->disk);
1439		if (nbd->config->bytesize)
1440			kobject_uevent(&nbd_to_dev(nbd)->kobj, KOBJ_CHANGE);
1441		if (test_and_clear_bit(NBD_RT_HAS_PID_FILE,
1442				       &config->runtime_flags))
1443			device_remove_file(disk_to_dev(nbd->disk), &pid_attr);
1444		nbd->pid = 0;
1445		if (test_and_clear_bit(NBD_RT_HAS_BACKEND_FILE,
1446				       &config->runtime_flags)) {
1447			device_remove_file(disk_to_dev(nbd->disk), &backend_attr);
1448			kfree(nbd->backend);
1449			nbd->backend = NULL;
1450		}
1451		nbd_clear_sock(nbd);
1452		if (config->num_connections) {
1453			int i;
1454			for (i = 0; i < config->num_connections; i++) {
1455				sockfd_put(config->socks[i]->sock);
1456				kfree(config->socks[i]);
1457			}
1458			kfree(config->socks);
1459		}
1460		kfree(nbd->config);
1461		nbd->config = NULL;
1462
1463		nbd->tag_set.timeout = 0;
1464
1465		mutex_unlock(&nbd->config_lock);
1466		nbd_put(nbd);
1467		module_put(THIS_MODULE);
1468	}
1469}
1470
1471static int nbd_start_device(struct nbd_device *nbd)
1472{
1473	struct nbd_config *config = nbd->config;
1474	int num_connections = config->num_connections;
1475	int error = 0, i;
1476
1477	if (nbd->pid)
1478		return -EBUSY;
1479	if (!config->socks)
1480		return -EINVAL;
1481	if (num_connections > 1 &&
1482	    !(config->flags & NBD_FLAG_CAN_MULTI_CONN)) {
1483		dev_err(disk_to_dev(nbd->disk), "server does not support multiple connections per device.\n");
1484		return -EINVAL;
1485	}
1486
1487retry:
1488	mutex_unlock(&nbd->config_lock);
1489	blk_mq_update_nr_hw_queues(&nbd->tag_set, num_connections);
1490	mutex_lock(&nbd->config_lock);
1491
1492	/* if another code path updated nr_hw_queues, retry until succeed */
1493	if (num_connections != config->num_connections) {
1494		num_connections = config->num_connections;
1495		goto retry;
1496	}
1497
1498	nbd->pid = task_pid_nr(current);
1499
1500	nbd_parse_flags(nbd);
1501
1502	error = device_create_file(disk_to_dev(nbd->disk), &pid_attr);
1503	if (error) {
1504		dev_err(disk_to_dev(nbd->disk), "device_create_file failed for pid!\n");
1505		return error;
1506	}
1507	set_bit(NBD_RT_HAS_PID_FILE, &config->runtime_flags);
1508
1509	nbd_dev_dbg_init(nbd);
1510	for (i = 0; i < num_connections; i++) {
1511		struct recv_thread_args *args;
1512
1513		args = kzalloc(sizeof(*args), GFP_KERNEL);
1514		if (!args) {
1515			sock_shutdown(nbd);
1516			/*
1517			 * If num_connections is m (2 < m),
1518			 * and NO.1 ~ NO.n(1 < n < m) kzallocs are successful.
1519			 * But NO.(n + 1) failed. We still have n recv threads.
1520			 * So, add flush_workqueue here to prevent recv threads
1521			 * dropping the last config_refs and trying to destroy
1522			 * the workqueue from inside the workqueue.
1523			 */
1524			if (i)
1525				flush_workqueue(nbd->recv_workq);
1526			return -ENOMEM;
1527		}
1528		sk_set_memalloc(config->socks[i]->sock->sk);
1529		if (nbd->tag_set.timeout)
1530			config->socks[i]->sock->sk->sk_sndtimeo =
1531				nbd->tag_set.timeout;
1532		atomic_inc(&config->recv_threads);
1533		refcount_inc(&nbd->config_refs);
1534		INIT_WORK(&args->work, recv_work);
1535		args->nbd = nbd;
1536		args->nsock = config->socks[i];
1537		args->index = i;
1538		queue_work(nbd->recv_workq, &args->work);
1539	}
1540	return nbd_set_size(nbd, config->bytesize, nbd_blksize(config));
1541}
1542
1543static int nbd_start_device_ioctl(struct nbd_device *nbd)
1544{
1545	struct nbd_config *config = nbd->config;
1546	int ret;
1547
1548	ret = nbd_start_device(nbd);
1549	if (ret)
1550		return ret;
1551
1552	if (max_part)
1553		set_bit(GD_NEED_PART_SCAN, &nbd->disk->state);
1554	mutex_unlock(&nbd->config_lock);
1555	ret = wait_event_interruptible(config->recv_wq,
1556					 atomic_read(&config->recv_threads) == 0);
1557	if (ret) {
1558		sock_shutdown(nbd);
1559		nbd_clear_que(nbd);
1560	}
1561
1562	flush_workqueue(nbd->recv_workq);
1563	mutex_lock(&nbd->config_lock);
1564	nbd_bdev_reset(nbd);
1565	/* user requested, ignore socket errors */
1566	if (test_bit(NBD_RT_DISCONNECT_REQUESTED, &config->runtime_flags))
1567		ret = 0;
1568	if (test_bit(NBD_RT_TIMEDOUT, &config->runtime_flags))
1569		ret = -ETIMEDOUT;
1570	return ret;
1571}
1572
1573static void nbd_clear_sock_ioctl(struct nbd_device *nbd)
1574{
1575	nbd_clear_sock(nbd);
1576	disk_force_media_change(nbd->disk);
1577	nbd_bdev_reset(nbd);
1578	if (test_and_clear_bit(NBD_RT_HAS_CONFIG_REF,
1579			       &nbd->config->runtime_flags))
1580		nbd_config_put(nbd);
1581}
1582
1583static void nbd_set_cmd_timeout(struct nbd_device *nbd, u64 timeout)
1584{
1585	nbd->tag_set.timeout = timeout * HZ;
1586	if (timeout)
1587		blk_queue_rq_timeout(nbd->disk->queue, timeout * HZ);
1588	else
1589		blk_queue_rq_timeout(nbd->disk->queue, 30 * HZ);
1590}
1591
1592/* Must be called with config_lock held */
1593static int __nbd_ioctl(struct block_device *bdev, struct nbd_device *nbd,
1594		       unsigned int cmd, unsigned long arg)
1595{
1596	struct nbd_config *config = nbd->config;
1597	loff_t bytesize;
1598
1599	switch (cmd) {
1600	case NBD_DISCONNECT:
1601		return nbd_disconnect(nbd);
1602	case NBD_CLEAR_SOCK:
1603		nbd_clear_sock_ioctl(nbd);
1604		return 0;
1605	case NBD_SET_SOCK:
1606		return nbd_add_socket(nbd, arg, false);
1607	case NBD_SET_BLKSIZE:
1608		return nbd_set_size(nbd, config->bytesize, arg);
1609	case NBD_SET_SIZE:
1610		return nbd_set_size(nbd, arg, nbd_blksize(config));
1611	case NBD_SET_SIZE_BLOCKS:
1612		if (check_shl_overflow(arg, config->blksize_bits, &bytesize))
1613			return -EINVAL;
1614		return nbd_set_size(nbd, bytesize, nbd_blksize(config));
1615	case NBD_SET_TIMEOUT:
1616		nbd_set_cmd_timeout(nbd, arg);
1617		return 0;
1618
1619	case NBD_SET_FLAGS:
1620		config->flags = arg;
1621		return 0;
1622	case NBD_DO_IT:
1623		return nbd_start_device_ioctl(nbd);
1624	case NBD_CLEAR_QUE:
1625		/*
1626		 * This is for compatibility only.  The queue is always cleared
1627		 * by NBD_DO_IT or NBD_CLEAR_SOCK.
1628		 */
1629		return 0;
1630	case NBD_PRINT_DEBUG:
1631		/*
1632		 * For compatibility only, we no longer keep a list of
1633		 * outstanding requests.
1634		 */
1635		return 0;
1636	}
1637	return -ENOTTY;
1638}
1639
1640static int nbd_ioctl(struct block_device *bdev, blk_mode_t mode,
1641		     unsigned int cmd, unsigned long arg)
1642{
1643	struct nbd_device *nbd = bdev->bd_disk->private_data;
1644	struct nbd_config *config = nbd->config;
1645	int error = -EINVAL;
1646
1647	if (!capable(CAP_SYS_ADMIN))
1648		return -EPERM;
1649
1650	/* The block layer will pass back some non-nbd ioctls in case we have
1651	 * special handling for them, but we don't so just return an error.
1652	 */
1653	if (_IOC_TYPE(cmd) != 0xab)
1654		return -EINVAL;
1655
1656	mutex_lock(&nbd->config_lock);
1657
1658	/* Don't allow ioctl operations on a nbd device that was created with
1659	 * netlink, unless it's DISCONNECT or CLEAR_SOCK, which are fine.
1660	 */
1661	if (!test_bit(NBD_RT_BOUND, &config->runtime_flags) ||
1662	    (cmd == NBD_DISCONNECT || cmd == NBD_CLEAR_SOCK))
1663		error = __nbd_ioctl(bdev, nbd, cmd, arg);
1664	else
1665		dev_err(nbd_to_dev(nbd), "Cannot use ioctl interface on a netlink controlled device.\n");
1666	mutex_unlock(&nbd->config_lock);
1667	return error;
1668}
1669
1670static int nbd_alloc_and_init_config(struct nbd_device *nbd)
1671{
1672	struct nbd_config *config;
1673
1674	if (WARN_ON(nbd->config))
1675		return -EINVAL;
1676
1677	if (!try_module_get(THIS_MODULE))
1678		return -ENODEV;
1679
1680	config = kzalloc(sizeof(struct nbd_config), GFP_NOFS);
1681	if (!config) {
1682		module_put(THIS_MODULE);
1683		return -ENOMEM;
1684	}
1685
1686	atomic_set(&config->recv_threads, 0);
1687	init_waitqueue_head(&config->recv_wq);
1688	init_waitqueue_head(&config->conn_wait);
1689	config->blksize_bits = NBD_DEF_BLKSIZE_BITS;
1690	atomic_set(&config->live_connections, 0);
1691
1692	nbd->config = config;
1693	/*
1694	 * Order refcount_set(&nbd->config_refs, 1) and nbd->config assignment,
1695	 * its pair is the barrier in nbd_get_config_unlocked().
1696	 * So nbd_get_config_unlocked() won't see nbd->config as null after
1697	 * refcount_inc_not_zero() succeed.
1698	 */
1699	smp_mb__before_atomic();
1700	refcount_set(&nbd->config_refs, 1);
1701
1702	return 0;
1703}
1704
1705static int nbd_open(struct gendisk *disk, blk_mode_t mode)
1706{
1707	struct nbd_device *nbd;
1708	struct nbd_config *config;
1709	int ret = 0;
1710
1711	mutex_lock(&nbd_index_mutex);
1712	nbd = disk->private_data;
1713	if (!nbd) {
1714		ret = -ENXIO;
1715		goto out;
1716	}
1717	if (!refcount_inc_not_zero(&nbd->refs)) {
1718		ret = -ENXIO;
1719		goto out;
1720	}
1721
1722	config = nbd_get_config_unlocked(nbd);
1723	if (!config) {
1724		mutex_lock(&nbd->config_lock);
1725		if (refcount_inc_not_zero(&nbd->config_refs)) {
1726			mutex_unlock(&nbd->config_lock);
1727			goto out;
1728		}
1729		ret = nbd_alloc_and_init_config(nbd);
1730		if (ret) {
1731			mutex_unlock(&nbd->config_lock);
1732			goto out;
1733		}
1734
1735		refcount_inc(&nbd->refs);
1736		mutex_unlock(&nbd->config_lock);
1737		if (max_part)
1738			set_bit(GD_NEED_PART_SCAN, &disk->state);
1739	} else if (nbd_disconnected(config)) {
1740		if (max_part)
1741			set_bit(GD_NEED_PART_SCAN, &disk->state);
1742	}
1743out:
1744	mutex_unlock(&nbd_index_mutex);
1745	return ret;
1746}
1747
1748static void nbd_release(struct gendisk *disk)
1749{
1750	struct nbd_device *nbd = disk->private_data;
1751
1752	if (test_bit(NBD_RT_DISCONNECT_ON_CLOSE, &nbd->config->runtime_flags) &&
1753			disk_openers(disk) == 0)
1754		nbd_disconnect_and_put(nbd);
1755
1756	nbd_config_put(nbd);
1757	nbd_put(nbd);
1758}
1759
1760static void nbd_free_disk(struct gendisk *disk)
1761{
1762	struct nbd_device *nbd = disk->private_data;
1763
1764	kfree(nbd);
1765}
1766
1767static const struct block_device_operations nbd_fops =
1768{
1769	.owner =	THIS_MODULE,
1770	.open =		nbd_open,
1771	.release =	nbd_release,
1772	.ioctl =	nbd_ioctl,
1773	.compat_ioctl =	nbd_ioctl,
1774	.free_disk =	nbd_free_disk,
1775};
1776
1777#if IS_ENABLED(CONFIG_DEBUG_FS)
1778
1779static int nbd_dbg_tasks_show(struct seq_file *s, void *unused)
1780{
1781	struct nbd_device *nbd = s->private;
1782
1783	if (nbd->pid)
1784		seq_printf(s, "recv: %d\n", nbd->pid);
1785
1786	return 0;
1787}
1788
1789DEFINE_SHOW_ATTRIBUTE(nbd_dbg_tasks);
1790
1791static int nbd_dbg_flags_show(struct seq_file *s, void *unused)
1792{
1793	struct nbd_device *nbd = s->private;
1794	u32 flags = nbd->config->flags;
1795
1796	seq_printf(s, "Hex: 0x%08x\n\n", flags);
1797
1798	seq_puts(s, "Known flags:\n");
1799
1800	if (flags & NBD_FLAG_HAS_FLAGS)
1801		seq_puts(s, "NBD_FLAG_HAS_FLAGS\n");
1802	if (flags & NBD_FLAG_READ_ONLY)
1803		seq_puts(s, "NBD_FLAG_READ_ONLY\n");
1804	if (flags & NBD_FLAG_SEND_FLUSH)
1805		seq_puts(s, "NBD_FLAG_SEND_FLUSH\n");
1806	if (flags & NBD_FLAG_SEND_FUA)
1807		seq_puts(s, "NBD_FLAG_SEND_FUA\n");
1808	if (flags & NBD_FLAG_SEND_TRIM)
1809		seq_puts(s, "NBD_FLAG_SEND_TRIM\n");
1810	if (flags & NBD_FLAG_SEND_WRITE_ZEROES)
1811		seq_puts(s, "NBD_FLAG_SEND_WRITE_ZEROES\n");
1812	if (flags & NBD_FLAG_ROTATIONAL)
1813		seq_puts(s, "NBD_FLAG_ROTATIONAL\n");
1814
1815	return 0;
1816}
1817
1818DEFINE_SHOW_ATTRIBUTE(nbd_dbg_flags);
1819
1820static int nbd_dev_dbg_init(struct nbd_device *nbd)
1821{
1822	struct dentry *dir;
1823	struct nbd_config *config = nbd->config;
1824
1825	if (!nbd_dbg_dir)
1826		return -EIO;
1827
1828	dir = debugfs_create_dir(nbd_name(nbd), nbd_dbg_dir);
1829	if (IS_ERR(dir)) {
1830		dev_err(nbd_to_dev(nbd), "Failed to create debugfs dir for '%s'\n",
1831			nbd_name(nbd));
1832		return -EIO;
1833	}
1834	config->dbg_dir = dir;
1835
1836	debugfs_create_file("tasks", 0444, dir, nbd, &nbd_dbg_tasks_fops);
1837	debugfs_create_u64("size_bytes", 0444, dir, &config->bytesize);
1838	debugfs_create_u32("timeout", 0444, dir, &nbd->tag_set.timeout);
1839	debugfs_create_u32("blocksize_bits", 0444, dir, &config->blksize_bits);
1840	debugfs_create_file("flags", 0444, dir, nbd, &nbd_dbg_flags_fops);
1841
1842	return 0;
1843}
1844
1845static void nbd_dev_dbg_close(struct nbd_device *nbd)
1846{
1847	debugfs_remove_recursive(nbd->config->dbg_dir);
1848}
1849
1850static int nbd_dbg_init(void)
1851{
1852	struct dentry *dbg_dir;
1853
1854	dbg_dir = debugfs_create_dir("nbd", NULL);
1855	if (IS_ERR(dbg_dir))
1856		return -EIO;
1857
1858	nbd_dbg_dir = dbg_dir;
1859
1860	return 0;
1861}
1862
1863static void nbd_dbg_close(void)
1864{
1865	debugfs_remove_recursive(nbd_dbg_dir);
1866}
1867
1868#else  /* IS_ENABLED(CONFIG_DEBUG_FS) */
1869
1870static int nbd_dev_dbg_init(struct nbd_device *nbd)
1871{
1872	return 0;
1873}
1874
1875static void nbd_dev_dbg_close(struct nbd_device *nbd)
1876{
1877}
1878
1879static int nbd_dbg_init(void)
1880{
1881	return 0;
1882}
1883
1884static void nbd_dbg_close(void)
1885{
1886}
1887
1888#endif
1889
1890static int nbd_init_request(struct blk_mq_tag_set *set, struct request *rq,
1891			    unsigned int hctx_idx, unsigned int numa_node)
1892{
1893	struct nbd_cmd *cmd = blk_mq_rq_to_pdu(rq);
1894	cmd->nbd = set->driver_data;
1895	cmd->flags = 0;
1896	mutex_init(&cmd->lock);
1897	return 0;
1898}
1899
1900static const struct blk_mq_ops nbd_mq_ops = {
1901	.queue_rq	= nbd_queue_rq,
1902	.complete	= nbd_complete_rq,
1903	.init_request	= nbd_init_request,
1904	.timeout	= nbd_xmit_timeout,
1905};
1906
1907static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
1908{
1909	struct queue_limits lim = {
1910		.max_hw_sectors		= 65536,
1911		.io_opt			= 256 << SECTOR_SHIFT,
1912		.max_segments		= USHRT_MAX,
1913		.max_segment_size	= UINT_MAX,
1914	};
1915	struct nbd_device *nbd;
1916	struct gendisk *disk;
1917	int err = -ENOMEM;
1918
1919	nbd = kzalloc(sizeof(struct nbd_device), GFP_KERNEL);
1920	if (!nbd)
1921		goto out;
1922
1923	nbd->tag_set.ops = &nbd_mq_ops;
1924	nbd->tag_set.nr_hw_queues = 1;
1925	nbd->tag_set.queue_depth = 128;
1926	nbd->tag_set.numa_node = NUMA_NO_NODE;
1927	nbd->tag_set.cmd_size = sizeof(struct nbd_cmd);
1928	nbd->tag_set.flags = BLK_MQ_F_BLOCKING;
1929	nbd->tag_set.driver_data = nbd;
1930	INIT_WORK(&nbd->remove_work, nbd_dev_remove_work);
1931	nbd->backend = NULL;
1932
1933	err = blk_mq_alloc_tag_set(&nbd->tag_set);
1934	if (err)
1935		goto out_free_nbd;
1936
1937	mutex_lock(&nbd_index_mutex);
1938	if (index >= 0) {
1939		err = idr_alloc(&nbd_index_idr, nbd, index, index + 1,
1940				GFP_KERNEL);
1941		if (err == -ENOSPC)
1942			err = -EEXIST;
1943	} else {
1944		err = idr_alloc(&nbd_index_idr, nbd, 0,
1945				(MINORMASK >> part_shift) + 1, GFP_KERNEL);
1946		if (err >= 0)
1947			index = err;
1948	}
1949	nbd->index = index;
1950	mutex_unlock(&nbd_index_mutex);
1951	if (err < 0)
1952		goto out_free_tags;
1953
1954	disk = blk_mq_alloc_disk(&nbd->tag_set, &lim, NULL);
1955	if (IS_ERR(disk)) {
1956		err = PTR_ERR(disk);
1957		goto out_free_idr;
1958	}
1959	nbd->disk = disk;
1960
1961	nbd->recv_workq = alloc_workqueue("nbd%d-recv",
1962					  WQ_MEM_RECLAIM | WQ_HIGHPRI |
1963					  WQ_UNBOUND, 0, nbd->index);
1964	if (!nbd->recv_workq) {
1965		dev_err(disk_to_dev(nbd->disk), "Could not allocate knbd recv work queue.\n");
1966		err = -ENOMEM;
1967		goto out_err_disk;
1968	}
1969
1970	mutex_init(&nbd->config_lock);
1971	refcount_set(&nbd->config_refs, 0);
1972	/*
1973	 * Start out with a zero references to keep other threads from using
1974	 * this device until it is fully initialized.
1975	 */
1976	refcount_set(&nbd->refs, 0);
1977	INIT_LIST_HEAD(&nbd->list);
1978	disk->major = NBD_MAJOR;
1979	disk->first_minor = index << part_shift;
1980	disk->minors = 1 << part_shift;
1981	disk->fops = &nbd_fops;
1982	disk->private_data = nbd;
1983	sprintf(disk->disk_name, "nbd%d", index);
1984	err = add_disk(disk);
1985	if (err)
1986		goto out_free_work;
1987
1988	/*
1989	 * Now publish the device.
1990	 */
1991	refcount_set(&nbd->refs, refs);
1992	nbd_total_devices++;
1993	return nbd;
1994
1995out_free_work:
1996	destroy_workqueue(nbd->recv_workq);
1997out_err_disk:
1998	put_disk(disk);
1999out_free_idr:
2000	mutex_lock(&nbd_index_mutex);
2001	idr_remove(&nbd_index_idr, index);
2002	mutex_unlock(&nbd_index_mutex);
2003out_free_tags:
2004	blk_mq_free_tag_set(&nbd->tag_set);
2005out_free_nbd:
2006	kfree(nbd);
2007out:
2008	return ERR_PTR(err);
2009}
2010
2011static struct nbd_device *nbd_find_get_unused(void)
2012{
2013	struct nbd_device *nbd;
2014	int id;
2015
2016	lockdep_assert_held(&nbd_index_mutex);
2017
2018	idr_for_each_entry(&nbd_index_idr, nbd, id) {
2019		if (refcount_read(&nbd->config_refs) ||
2020		    test_bit(NBD_DESTROY_ON_DISCONNECT, &nbd->flags))
2021			continue;
2022		if (refcount_inc_not_zero(&nbd->refs))
2023			return nbd;
2024	}
2025
2026	return NULL;
2027}
2028
2029/* Netlink interface. */
2030static const struct nla_policy nbd_attr_policy[NBD_ATTR_MAX + 1] = {
2031	[NBD_ATTR_INDEX]		=	{ .type = NLA_U32 },
2032	[NBD_ATTR_SIZE_BYTES]		=	{ .type = NLA_U64 },
2033	[NBD_ATTR_BLOCK_SIZE_BYTES]	=	{ .type = NLA_U64 },
2034	[NBD_ATTR_TIMEOUT]		=	{ .type = NLA_U64 },
2035	[NBD_ATTR_SERVER_FLAGS]		=	{ .type = NLA_U64 },
2036	[NBD_ATTR_CLIENT_FLAGS]		=	{ .type = NLA_U64 },
2037	[NBD_ATTR_SOCKETS]		=	{ .type = NLA_NESTED},
2038	[NBD_ATTR_DEAD_CONN_TIMEOUT]	=	{ .type = NLA_U64 },
2039	[NBD_ATTR_DEVICE_LIST]		=	{ .type = NLA_NESTED},
2040	[NBD_ATTR_BACKEND_IDENTIFIER]	=	{ .type = NLA_STRING},
2041};
2042
2043static const struct nla_policy nbd_sock_policy[NBD_SOCK_MAX + 1] = {
2044	[NBD_SOCK_FD]			=	{ .type = NLA_U32 },
2045};
2046
2047/* We don't use this right now since we don't parse the incoming list, but we
2048 * still want it here so userspace knows what to expect.
2049 */
2050static const struct nla_policy __attribute__((unused))
2051nbd_device_policy[NBD_DEVICE_ATTR_MAX + 1] = {
2052	[NBD_DEVICE_INDEX]		=	{ .type = NLA_U32 },
2053	[NBD_DEVICE_CONNECTED]		=	{ .type = NLA_U8 },
2054};
2055
2056static int nbd_genl_size_set(struct genl_info *info, struct nbd_device *nbd)
2057{
2058	struct nbd_config *config = nbd->config;
2059	u64 bsize = nbd_blksize(config);
2060	u64 bytes = config->bytesize;
2061
2062	if (info->attrs[NBD_ATTR_SIZE_BYTES])
2063		bytes = nla_get_u64(info->attrs[NBD_ATTR_SIZE_BYTES]);
2064
2065	if (info->attrs[NBD_ATTR_BLOCK_SIZE_BYTES])
2066		bsize = nla_get_u64(info->attrs[NBD_ATTR_BLOCK_SIZE_BYTES]);
2067
2068	if (bytes != config->bytesize || bsize != nbd_blksize(config))
2069		return nbd_set_size(nbd, bytes, bsize);
2070	return 0;
2071}
2072
2073static int nbd_genl_connect(struct sk_buff *skb, struct genl_info *info)
2074{
2075	struct nbd_device *nbd;
2076	struct nbd_config *config;
2077	int index = -1;
2078	int ret;
2079	bool put_dev = false;
2080
2081	if (!netlink_capable(skb, CAP_SYS_ADMIN))
2082		return -EPERM;
2083
2084	if (info->attrs[NBD_ATTR_INDEX]) {
2085		index = nla_get_u32(info->attrs[NBD_ATTR_INDEX]);
2086
2087		/*
2088		 * Too big first_minor can cause duplicate creation of
2089		 * sysfs files/links, since index << part_shift might overflow, or
2090		 * MKDEV() expect that the max bits of first_minor is 20.
2091		 */
2092		if (index < 0 || index > MINORMASK >> part_shift) {
2093			pr_err("illegal input index %d\n", index);
2094			return -EINVAL;
2095		}
2096	}
2097	if (GENL_REQ_ATTR_CHECK(info, NBD_ATTR_SOCKETS)) {
2098		pr_err("must specify at least one socket\n");
2099		return -EINVAL;
2100	}
2101	if (GENL_REQ_ATTR_CHECK(info, NBD_ATTR_SIZE_BYTES)) {
2102		pr_err("must specify a size in bytes for the device\n");
2103		return -EINVAL;
2104	}
2105again:
2106	mutex_lock(&nbd_index_mutex);
2107	if (index == -1) {
2108		nbd = nbd_find_get_unused();
2109	} else {
2110		nbd = idr_find(&nbd_index_idr, index);
2111		if (nbd) {
2112			if ((test_bit(NBD_DESTROY_ON_DISCONNECT, &nbd->flags) &&
2113			     test_bit(NBD_DISCONNECT_REQUESTED, &nbd->flags)) ||
2114			    !refcount_inc_not_zero(&nbd->refs)) {
2115				mutex_unlock(&nbd_index_mutex);
2116				pr_err("device at index %d is going down\n",
2117					index);
2118				return -EINVAL;
2119			}
2120		}
2121	}
2122	mutex_unlock(&nbd_index_mutex);
2123
2124	if (!nbd) {
2125		nbd = nbd_dev_add(index, 2);
2126		if (IS_ERR(nbd)) {
2127			pr_err("failed to add new device\n");
2128			return PTR_ERR(nbd);
2129		}
2130	}
2131
2132	mutex_lock(&nbd->config_lock);
2133	if (refcount_read(&nbd->config_refs)) {
2134		mutex_unlock(&nbd->config_lock);
2135		nbd_put(nbd);
2136		if (index == -1)
2137			goto again;
2138		pr_err("nbd%d already in use\n", index);
2139		return -EBUSY;
2140	}
2141
2142	ret = nbd_alloc_and_init_config(nbd);
2143	if (ret) {
2144		mutex_unlock(&nbd->config_lock);
2145		nbd_put(nbd);
2146		pr_err("couldn't allocate config\n");
2147		return ret;
2148	}
2149
2150	config = nbd->config;
2151	set_bit(NBD_RT_BOUND, &config->runtime_flags);
2152	ret = nbd_genl_size_set(info, nbd);
2153	if (ret)
2154		goto out;
2155
2156	if (info->attrs[NBD_ATTR_TIMEOUT])
2157		nbd_set_cmd_timeout(nbd,
2158				    nla_get_u64(info->attrs[NBD_ATTR_TIMEOUT]));
2159	if (info->attrs[NBD_ATTR_DEAD_CONN_TIMEOUT]) {
2160		config->dead_conn_timeout =
2161			nla_get_u64(info->attrs[NBD_ATTR_DEAD_CONN_TIMEOUT]);
2162		config->dead_conn_timeout *= HZ;
2163	}
2164	if (info->attrs[NBD_ATTR_SERVER_FLAGS])
2165		config->flags =
2166			nla_get_u64(info->attrs[NBD_ATTR_SERVER_FLAGS]);
2167	if (info->attrs[NBD_ATTR_CLIENT_FLAGS]) {
2168		u64 flags = nla_get_u64(info->attrs[NBD_ATTR_CLIENT_FLAGS]);
2169		if (flags & NBD_CFLAG_DESTROY_ON_DISCONNECT) {
2170			/*
2171			 * We have 1 ref to keep the device around, and then 1
2172			 * ref for our current operation here, which will be
2173			 * inherited by the config.  If we already have
2174			 * DESTROY_ON_DISCONNECT set then we know we don't have
2175			 * that extra ref already held so we don't need the
2176			 * put_dev.
2177			 */
2178			if (!test_and_set_bit(NBD_DESTROY_ON_DISCONNECT,
2179					      &nbd->flags))
2180				put_dev = true;
2181		} else {
2182			if (test_and_clear_bit(NBD_DESTROY_ON_DISCONNECT,
2183					       &nbd->flags))
2184				refcount_inc(&nbd->refs);
2185		}
2186		if (flags & NBD_CFLAG_DISCONNECT_ON_CLOSE) {
2187			set_bit(NBD_RT_DISCONNECT_ON_CLOSE,
2188				&config->runtime_flags);
2189		}
2190	}
2191
2192	if (info->attrs[NBD_ATTR_SOCKETS]) {
2193		struct nlattr *attr;
2194		int rem, fd;
2195
2196		nla_for_each_nested(attr, info->attrs[NBD_ATTR_SOCKETS],
2197				    rem) {
2198			struct nlattr *socks[NBD_SOCK_MAX+1];
2199
2200			if (nla_type(attr) != NBD_SOCK_ITEM) {
2201				pr_err("socks must be embedded in a SOCK_ITEM attr\n");
2202				ret = -EINVAL;
2203				goto out;
2204			}
2205			ret = nla_parse_nested_deprecated(socks, NBD_SOCK_MAX,
2206							  attr,
2207							  nbd_sock_policy,
2208							  info->extack);
2209			if (ret != 0) {
2210				pr_err("error processing sock list\n");
2211				ret = -EINVAL;
2212				goto out;
2213			}
2214			if (!socks[NBD_SOCK_FD])
2215				continue;
2216			fd = (int)nla_get_u32(socks[NBD_SOCK_FD]);
2217			ret = nbd_add_socket(nbd, fd, true);
2218			if (ret)
2219				goto out;
2220		}
2221	}
2222
2223	if (info->attrs[NBD_ATTR_BACKEND_IDENTIFIER]) {
2224		nbd->backend = nla_strdup(info->attrs[NBD_ATTR_BACKEND_IDENTIFIER],
2225					  GFP_KERNEL);
2226		if (!nbd->backend) {
2227			ret = -ENOMEM;
2228			goto out;
2229		}
2230	}
2231	ret = device_create_file(disk_to_dev(nbd->disk), &backend_attr);
2232	if (ret) {
2233		dev_err(disk_to_dev(nbd->disk),
2234			"device_create_file failed for backend!\n");
2235		goto out;
2236	}
2237	set_bit(NBD_RT_HAS_BACKEND_FILE, &config->runtime_flags);
2238
2239	ret = nbd_start_device(nbd);
2240out:
2241	if (!ret) {
2242		set_bit(NBD_RT_HAS_CONFIG_REF, &config->runtime_flags);
2243		refcount_inc(&nbd->config_refs);
2244		nbd_connect_reply(info, nbd->index);
2245	}
2246	mutex_unlock(&nbd->config_lock);
2247
2248	nbd_config_put(nbd);
2249	if (put_dev)
2250		nbd_put(nbd);
2251	return ret;
2252}
2253
2254static void nbd_disconnect_and_put(struct nbd_device *nbd)
2255{
2256	mutex_lock(&nbd->config_lock);
2257	nbd_disconnect(nbd);
2258	sock_shutdown(nbd);
2259	wake_up(&nbd->config->conn_wait);
2260	/*
2261	 * Make sure recv thread has finished, we can safely call nbd_clear_que()
2262	 * to cancel the inflight I/Os.
2263	 */
2264	flush_workqueue(nbd->recv_workq);
2265	nbd_clear_que(nbd);
2266	nbd->task_setup = NULL;
2267	clear_bit(NBD_RT_BOUND, &nbd->config->runtime_flags);
2268	mutex_unlock(&nbd->config_lock);
2269
2270	if (test_and_clear_bit(NBD_RT_HAS_CONFIG_REF,
2271			       &nbd->config->runtime_flags))
2272		nbd_config_put(nbd);
2273}
2274
2275static int nbd_genl_disconnect(struct sk_buff *skb, struct genl_info *info)
2276{
2277	struct nbd_device *nbd;
2278	int index;
2279
2280	if (!netlink_capable(skb, CAP_SYS_ADMIN))
2281		return -EPERM;
2282
2283	if (GENL_REQ_ATTR_CHECK(info, NBD_ATTR_INDEX)) {
2284		pr_err("must specify an index to disconnect\n");
2285		return -EINVAL;
2286	}
2287	index = nla_get_u32(info->attrs[NBD_ATTR_INDEX]);
2288	mutex_lock(&nbd_index_mutex);
2289	nbd = idr_find(&nbd_index_idr, index);
2290	if (!nbd) {
2291		mutex_unlock(&nbd_index_mutex);
2292		pr_err("couldn't find device at index %d\n", index);
2293		return -EINVAL;
2294	}
2295	if (!refcount_inc_not_zero(&nbd->refs)) {
2296		mutex_unlock(&nbd_index_mutex);
2297		pr_err("device at index %d is going down\n", index);
2298		return -EINVAL;
2299	}
2300	mutex_unlock(&nbd_index_mutex);
2301	if (!refcount_inc_not_zero(&nbd->config_refs))
2302		goto put_nbd;
2303	nbd_disconnect_and_put(nbd);
2304	nbd_config_put(nbd);
2305put_nbd:
2306	nbd_put(nbd);
2307	return 0;
2308}
2309
2310static int nbd_genl_reconfigure(struct sk_buff *skb, struct genl_info *info)
2311{
2312	struct nbd_device *nbd = NULL;
2313	struct nbd_config *config;
2314	int index;
2315	int ret = 0;
2316	bool put_dev = false;
2317
2318	if (!netlink_capable(skb, CAP_SYS_ADMIN))
2319		return -EPERM;
2320
2321	if (GENL_REQ_ATTR_CHECK(info, NBD_ATTR_INDEX)) {
2322		pr_err("must specify a device to reconfigure\n");
2323		return -EINVAL;
2324	}
2325	index = nla_get_u32(info->attrs[NBD_ATTR_INDEX]);
2326	mutex_lock(&nbd_index_mutex);
2327	nbd = idr_find(&nbd_index_idr, index);
2328	if (!nbd) {
2329		mutex_unlock(&nbd_index_mutex);
2330		pr_err("couldn't find a device at index %d\n", index);
2331		return -EINVAL;
2332	}
2333	if (nbd->backend) {
2334		if (info->attrs[NBD_ATTR_BACKEND_IDENTIFIER]) {
2335			if (nla_strcmp(info->attrs[NBD_ATTR_BACKEND_IDENTIFIER],
2336				       nbd->backend)) {
2337				mutex_unlock(&nbd_index_mutex);
2338				dev_err(nbd_to_dev(nbd),
2339					"backend image doesn't match with %s\n",
2340					nbd->backend);
2341				return -EINVAL;
2342			}
2343		} else {
2344			mutex_unlock(&nbd_index_mutex);
2345			dev_err(nbd_to_dev(nbd), "must specify backend\n");
2346			return -EINVAL;
2347		}
2348	}
2349	if (!refcount_inc_not_zero(&nbd->refs)) {
2350		mutex_unlock(&nbd_index_mutex);
2351		pr_err("device at index %d is going down\n", index);
2352		return -EINVAL;
2353	}
2354	mutex_unlock(&nbd_index_mutex);
2355
2356	config = nbd_get_config_unlocked(nbd);
2357	if (!config) {
2358		dev_err(nbd_to_dev(nbd),
2359			"not configured, cannot reconfigure\n");
2360		nbd_put(nbd);
2361		return -EINVAL;
2362	}
2363
2364	mutex_lock(&nbd->config_lock);
2365	if (!test_bit(NBD_RT_BOUND, &config->runtime_flags) ||
2366	    !nbd->pid) {
2367		dev_err(nbd_to_dev(nbd),
2368			"not configured, cannot reconfigure\n");
2369		ret = -EINVAL;
2370		goto out;
2371	}
2372
2373	ret = nbd_genl_size_set(info, nbd);
2374	if (ret)
2375		goto out;
2376
2377	if (info->attrs[NBD_ATTR_TIMEOUT])
2378		nbd_set_cmd_timeout(nbd,
2379				    nla_get_u64(info->attrs[NBD_ATTR_TIMEOUT]));
2380	if (info->attrs[NBD_ATTR_DEAD_CONN_TIMEOUT]) {
2381		config->dead_conn_timeout =
2382			nla_get_u64(info->attrs[NBD_ATTR_DEAD_CONN_TIMEOUT]);
2383		config->dead_conn_timeout *= HZ;
2384	}
2385	if (info->attrs[NBD_ATTR_CLIENT_FLAGS]) {
2386		u64 flags = nla_get_u64(info->attrs[NBD_ATTR_CLIENT_FLAGS]);
2387		if (flags & NBD_CFLAG_DESTROY_ON_DISCONNECT) {
2388			if (!test_and_set_bit(NBD_DESTROY_ON_DISCONNECT,
2389					      &nbd->flags))
2390				put_dev = true;
2391		} else {
2392			if (test_and_clear_bit(NBD_DESTROY_ON_DISCONNECT,
2393					       &nbd->flags))
2394				refcount_inc(&nbd->refs);
2395		}
2396
2397		if (flags & NBD_CFLAG_DISCONNECT_ON_CLOSE) {
2398			set_bit(NBD_RT_DISCONNECT_ON_CLOSE,
2399					&config->runtime_flags);
2400		} else {
2401			clear_bit(NBD_RT_DISCONNECT_ON_CLOSE,
2402					&config->runtime_flags);
2403		}
2404	}
2405
2406	if (info->attrs[NBD_ATTR_SOCKETS]) {
2407		struct nlattr *attr;
2408		int rem, fd;
2409
2410		nla_for_each_nested(attr, info->attrs[NBD_ATTR_SOCKETS],
2411				    rem) {
2412			struct nlattr *socks[NBD_SOCK_MAX+1];
2413
2414			if (nla_type(attr) != NBD_SOCK_ITEM) {
2415				pr_err("socks must be embedded in a SOCK_ITEM attr\n");
2416				ret = -EINVAL;
2417				goto out;
2418			}
2419			ret = nla_parse_nested_deprecated(socks, NBD_SOCK_MAX,
2420							  attr,
2421							  nbd_sock_policy,
2422							  info->extack);
2423			if (ret != 0) {
2424				pr_err("error processing sock list\n");
2425				ret = -EINVAL;
2426				goto out;
2427			}
2428			if (!socks[NBD_SOCK_FD])
2429				continue;
2430			fd = (int)nla_get_u32(socks[NBD_SOCK_FD]);
2431			ret = nbd_reconnect_socket(nbd, fd);
2432			if (ret) {
2433				if (ret == -ENOSPC)
2434					ret = 0;
2435				goto out;
2436			}
2437			dev_info(nbd_to_dev(nbd), "reconnected socket\n");
2438		}
2439	}
2440out:
2441	mutex_unlock(&nbd->config_lock);
2442	nbd_config_put(nbd);
2443	nbd_put(nbd);
2444	if (put_dev)
2445		nbd_put(nbd);
2446	return ret;
2447}
2448
2449static const struct genl_small_ops nbd_connect_genl_ops[] = {
2450	{
2451		.cmd	= NBD_CMD_CONNECT,
2452		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2453		.doit	= nbd_genl_connect,
2454	},
2455	{
2456		.cmd	= NBD_CMD_DISCONNECT,
2457		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2458		.doit	= nbd_genl_disconnect,
2459	},
2460	{
2461		.cmd	= NBD_CMD_RECONFIGURE,
2462		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2463		.doit	= nbd_genl_reconfigure,
2464	},
2465	{
2466		.cmd	= NBD_CMD_STATUS,
2467		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2468		.doit	= nbd_genl_status,
2469	},
2470};
2471
2472static const struct genl_multicast_group nbd_mcast_grps[] = {
2473	{ .name = NBD_GENL_MCAST_GROUP_NAME, },
2474};
2475
2476static struct genl_family nbd_genl_family __ro_after_init = {
2477	.hdrsize	= 0,
2478	.name		= NBD_GENL_FAMILY_NAME,
2479	.version	= NBD_GENL_VERSION,
2480	.module		= THIS_MODULE,
2481	.small_ops	= nbd_connect_genl_ops,
2482	.n_small_ops	= ARRAY_SIZE(nbd_connect_genl_ops),
2483	.resv_start_op	= NBD_CMD_STATUS + 1,
2484	.maxattr	= NBD_ATTR_MAX,
2485	.netnsok	= 1,
2486	.policy = nbd_attr_policy,
2487	.mcgrps		= nbd_mcast_grps,
2488	.n_mcgrps	= ARRAY_SIZE(nbd_mcast_grps),
2489};
2490MODULE_ALIAS_GENL_FAMILY(NBD_GENL_FAMILY_NAME);
2491
2492static int populate_nbd_status(struct nbd_device *nbd, struct sk_buff *reply)
2493{
2494	struct nlattr *dev_opt;
2495	u8 connected = 0;
2496	int ret;
2497
2498	/* This is a little racey, but for status it's ok.  The
2499	 * reason we don't take a ref here is because we can't
2500	 * take a ref in the index == -1 case as we would need
2501	 * to put under the nbd_index_mutex, which could
2502	 * deadlock if we are configured to remove ourselves
2503	 * once we're disconnected.
2504	 */
2505	if (refcount_read(&nbd->config_refs))
2506		connected = 1;
2507	dev_opt = nla_nest_start_noflag(reply, NBD_DEVICE_ITEM);
2508	if (!dev_opt)
2509		return -EMSGSIZE;
2510	ret = nla_put_u32(reply, NBD_DEVICE_INDEX, nbd->index);
2511	if (ret)
2512		return -EMSGSIZE;
2513	ret = nla_put_u8(reply, NBD_DEVICE_CONNECTED,
2514			 connected);
2515	if (ret)
2516		return -EMSGSIZE;
2517	nla_nest_end(reply, dev_opt);
2518	return 0;
2519}
2520
2521static int status_cb(int id, void *ptr, void *data)
2522{
2523	struct nbd_device *nbd = ptr;
2524	return populate_nbd_status(nbd, (struct sk_buff *)data);
2525}
2526
2527static int nbd_genl_status(struct sk_buff *skb, struct genl_info *info)
2528{
2529	struct nlattr *dev_list;
2530	struct sk_buff *reply;
2531	void *reply_head;
2532	size_t msg_size;
2533	int index = -1;
2534	int ret = -ENOMEM;
2535
2536	if (info->attrs[NBD_ATTR_INDEX])
2537		index = nla_get_u32(info->attrs[NBD_ATTR_INDEX]);
2538
2539	mutex_lock(&nbd_index_mutex);
2540
2541	msg_size = nla_total_size(nla_attr_size(sizeof(u32)) +
2542				  nla_attr_size(sizeof(u8)));
2543	msg_size *= (index == -1) ? nbd_total_devices : 1;
2544
2545	reply = genlmsg_new(msg_size, GFP_KERNEL);
2546	if (!reply)
2547		goto out;
2548	reply_head = genlmsg_put_reply(reply, info, &nbd_genl_family, 0,
2549				       NBD_CMD_STATUS);
2550	if (!reply_head) {
2551		nlmsg_free(reply);
2552		goto out;
2553	}
2554
2555	dev_list = nla_nest_start_noflag(reply, NBD_ATTR_DEVICE_LIST);
2556	if (!dev_list) {
2557		nlmsg_free(reply);
2558		ret = -EMSGSIZE;
2559		goto out;
2560	}
2561
2562	if (index == -1) {
2563		ret = idr_for_each(&nbd_index_idr, &status_cb, reply);
2564		if (ret) {
2565			nlmsg_free(reply);
2566			goto out;
2567		}
2568	} else {
2569		struct nbd_device *nbd;
2570		nbd = idr_find(&nbd_index_idr, index);
2571		if (nbd) {
2572			ret = populate_nbd_status(nbd, reply);
2573			if (ret) {
2574				nlmsg_free(reply);
2575				goto out;
2576			}
2577		}
2578	}
2579	nla_nest_end(reply, dev_list);
2580	genlmsg_end(reply, reply_head);
2581	ret = genlmsg_reply(reply, info);
2582out:
2583	mutex_unlock(&nbd_index_mutex);
2584	return ret;
2585}
2586
2587static void nbd_connect_reply(struct genl_info *info, int index)
2588{
2589	struct sk_buff *skb;
2590	void *msg_head;
2591	int ret;
2592
2593	skb = genlmsg_new(nla_total_size(sizeof(u32)), GFP_KERNEL);
2594	if (!skb)
2595		return;
2596	msg_head = genlmsg_put_reply(skb, info, &nbd_genl_family, 0,
2597				     NBD_CMD_CONNECT);
2598	if (!msg_head) {
2599		nlmsg_free(skb);
2600		return;
2601	}
2602	ret = nla_put_u32(skb, NBD_ATTR_INDEX, index);
2603	if (ret) {
2604		nlmsg_free(skb);
2605		return;
2606	}
2607	genlmsg_end(skb, msg_head);
2608	genlmsg_reply(skb, info);
2609}
2610
2611static void nbd_mcast_index(int index)
2612{
2613	struct sk_buff *skb;
2614	void *msg_head;
2615	int ret;
2616
2617	skb = genlmsg_new(nla_total_size(sizeof(u32)), GFP_KERNEL);
2618	if (!skb)
2619		return;
2620	msg_head = genlmsg_put(skb, 0, 0, &nbd_genl_family, 0,
2621				     NBD_CMD_LINK_DEAD);
2622	if (!msg_head) {
2623		nlmsg_free(skb);
2624		return;
2625	}
2626	ret = nla_put_u32(skb, NBD_ATTR_INDEX, index);
2627	if (ret) {
2628		nlmsg_free(skb);
2629		return;
2630	}
2631	genlmsg_end(skb, msg_head);
2632	genlmsg_multicast(&nbd_genl_family, skb, 0, 0, GFP_KERNEL);
2633}
2634
2635static void nbd_dead_link_work(struct work_struct *work)
2636{
2637	struct link_dead_args *args = container_of(work, struct link_dead_args,
2638						   work);
2639	nbd_mcast_index(args->index);
2640	kfree(args);
2641}
2642
2643static int __init nbd_init(void)
2644{
2645	int i;
2646
2647	BUILD_BUG_ON(sizeof(struct nbd_request) != 28);
2648
2649	if (max_part < 0) {
2650		pr_err("max_part must be >= 0\n");
2651		return -EINVAL;
2652	}
2653
2654	part_shift = 0;
2655	if (max_part > 0) {
2656		part_shift = fls(max_part);
2657
2658		/*
2659		 * Adjust max_part according to part_shift as it is exported
2660		 * to user space so that user can know the max number of
2661		 * partition kernel should be able to manage.
2662		 *
2663		 * Note that -1 is required because partition 0 is reserved
2664		 * for the whole disk.
2665		 */
2666		max_part = (1UL << part_shift) - 1;
2667	}
2668
2669	if ((1UL << part_shift) > DISK_MAX_PARTS)
2670		return -EINVAL;
2671
2672	if (nbds_max > 1UL << (MINORBITS - part_shift))
2673		return -EINVAL;
2674
2675	if (register_blkdev(NBD_MAJOR, "nbd"))
2676		return -EIO;
2677
2678	nbd_del_wq = alloc_workqueue("nbd-del", WQ_UNBOUND, 0);
2679	if (!nbd_del_wq) {
2680		unregister_blkdev(NBD_MAJOR, "nbd");
2681		return -ENOMEM;
2682	}
2683
2684	if (genl_register_family(&nbd_genl_family)) {
2685		destroy_workqueue(nbd_del_wq);
2686		unregister_blkdev(NBD_MAJOR, "nbd");
2687		return -EINVAL;
2688	}
2689	nbd_dbg_init();
2690
2691	for (i = 0; i < nbds_max; i++)
2692		nbd_dev_add(i, 1);
2693	return 0;
2694}
2695
2696static int nbd_exit_cb(int id, void *ptr, void *data)
2697{
2698	struct list_head *list = (struct list_head *)data;
2699	struct nbd_device *nbd = ptr;
2700
2701	/* Skip nbd that is being removed asynchronously */
2702	if (refcount_read(&nbd->refs))
2703		list_add_tail(&nbd->list, list);
2704
2705	return 0;
2706}
2707
2708static void __exit nbd_cleanup(void)
2709{
2710	struct nbd_device *nbd;
2711	LIST_HEAD(del_list);
2712
2713	/*
2714	 * Unregister netlink interface prior to waiting
2715	 * for the completion of netlink commands.
2716	 */
2717	genl_unregister_family(&nbd_genl_family);
2718
2719	nbd_dbg_close();
2720
2721	mutex_lock(&nbd_index_mutex);
2722	idr_for_each(&nbd_index_idr, &nbd_exit_cb, &del_list);
2723	mutex_unlock(&nbd_index_mutex);
2724
2725	while (!list_empty(&del_list)) {
2726		nbd = list_first_entry(&del_list, struct nbd_device, list);
2727		list_del_init(&nbd->list);
2728		if (refcount_read(&nbd->config_refs))
2729			pr_err("possibly leaking nbd_config (ref %d)\n",
2730					refcount_read(&nbd->config_refs));
2731		if (refcount_read(&nbd->refs) != 1)
2732			pr_err("possibly leaking a device\n");
2733		nbd_put(nbd);
2734	}
2735
2736	/* Also wait for nbd_dev_remove_work() completes */
2737	destroy_workqueue(nbd_del_wq);
2738
2739	idr_destroy(&nbd_index_idr);
2740	unregister_blkdev(NBD_MAJOR, "nbd");
2741}
2742
2743module_init(nbd_init);
2744module_exit(nbd_cleanup);
2745
2746MODULE_DESCRIPTION("Network Block Device");
2747MODULE_LICENSE("GPL");
2748
2749module_param(nbds_max, int, 0444);
2750MODULE_PARM_DESC(nbds_max, "number of network block devices to initialize (default: 16)");
2751module_param(max_part, int, 0444);
2752MODULE_PARM_DESC(max_part, "number of partitions per device (default: 16)");