messenger.c

   1// SPDX-License-Identifier: GPL-2.0
   2#include <linux/ceph/ceph_debug.h>
   3
   4#include <linux/crc32c.h>
   5#include <linux/ctype.h>
   6#include <linux/highmem.h>
   7#include <linux/inet.h>
   8#include <linux/kthread.h>
   9#include <linux/net.h>
  10#include <linux/nsproxy.h>
  11#include <linux/sched/mm.h>
  12#include <linux/slab.h>
  13#include <linux/socket.h>
  14#include <linux/string.h>
  15#ifdef	CONFIG_BLOCK
  16#include <linux/bio.h>
  17#endif	/* CONFIG_BLOCK */
  18#include <linux/dns_resolver.h>
  19#include <net/tcp.h>
  20#include <trace/events/sock.h>
  21
  22#include <linux/ceph/ceph_features.h>
  23#include <linux/ceph/libceph.h>
  24#include <linux/ceph/messenger.h>
  25#include <linux/ceph/decode.h>
  26#include <linux/ceph/pagelist.h>
  27#include <linux/export.h>
  28
  29/*
  30 * Ceph uses the messenger to exchange ceph_msg messages with other
  31 * hosts in the system.  The messenger provides ordered and reliable
  32 * delivery.  We tolerate TCP disconnects by reconnecting (with
  33 * exponential backoff) in the case of a fault (disconnection, bad
  34 * crc, protocol error).  Acks allow sent messages to be discarded by
  35 * the sender.
  36 */
  37
  38/*
  39 * We track the state of the socket on a given connection using
  40 * values defined below.  The transition to a new socket state is
  41 * handled by a function which verifies we aren't coming from an
  42 * unexpected state.
  43 *
  44 *      --------
  45 *      | NEW* |  transient initial state
  46 *      --------
  47 *          | con_sock_state_init()
  48 *          v
  49 *      ----------
  50 *      | CLOSED |  initialized, but no socket (and no
  51 *      ----------  TCP connection)
  52 *       ^      \
  53 *       |       \ con_sock_state_connecting()
  54 *       |        ----------------------
  55 *       |                              \
  56 *       + con_sock_state_closed()       \
  57 *       |+---------------------------    \
  58 *       | \                          \    \
  59 *       |  -----------                \    \
  60 *       |  | CLOSING |  socket event;  \    \
  61 *       |  -----------  await close     \    \
  62 *       |       ^                        \   |
  63 *       |       |                         \  |
  64 *       |       + con_sock_state_closing() \ |
  65 *       |      / \                         | |
  66 *       |     /   ---------------          | |
  67 *       |    /                   \         v v
  68 *       |   /                    --------------
  69 *       |  /    -----------------| CONNECTING |  socket created, TCP
  70 *       |  |   /                 --------------  connect initiated
  71 *       |  |   | con_sock_state_connected()
  72 *       |  |   v
  73 *      -------------
  74 *      | CONNECTED |  TCP connection established
  75 *      -------------
  76 *
  77 * State values for ceph_connection->sock_state; NEW is assumed to be 0.
  78 */
  79
  80#define CON_SOCK_STATE_NEW		0	/* -> CLOSED */
  81#define CON_SOCK_STATE_CLOSED		1	/* -> CONNECTING */
  82#define CON_SOCK_STATE_CONNECTING	2	/* -> CONNECTED or -> CLOSING */
  83#define CON_SOCK_STATE_CONNECTED	3	/* -> CLOSING or -> CLOSED */
  84#define CON_SOCK_STATE_CLOSING		4	/* -> CLOSED */
  85
  86static bool con_flag_valid(unsigned long con_flag)
  87{
  88	switch (con_flag) {
  89	case CEPH_CON_F_LOSSYTX:
  90	case CEPH_CON_F_KEEPALIVE_PENDING:
  91	case CEPH_CON_F_WRITE_PENDING:
  92	case CEPH_CON_F_SOCK_CLOSED:
  93	case CEPH_CON_F_BACKOFF:
  94		return true;
  95	default:
  96		return false;
  97	}
  98}
  99
 100void ceph_con_flag_clear(struct ceph_connection *con, unsigned long con_flag)
 101{
 102	BUG_ON(!con_flag_valid(con_flag));
 103
 104	clear_bit(con_flag, &con->flags);
 105}
 106
 107void ceph_con_flag_set(struct ceph_connection *con, unsigned long con_flag)
 108{
 109	BUG_ON(!con_flag_valid(con_flag));
 110
 111	set_bit(con_flag, &con->flags);
 112}
 113
 114bool ceph_con_flag_test(struct ceph_connection *con, unsigned long con_flag)
 115{
 116	BUG_ON(!con_flag_valid(con_flag));
 117
 118	return test_bit(con_flag, &con->flags);
 119}
 120
 121bool ceph_con_flag_test_and_clear(struct ceph_connection *con,
 122				  unsigned long con_flag)
 123{
 124	BUG_ON(!con_flag_valid(con_flag));
 125
 126	return test_and_clear_bit(con_flag, &con->flags);
 127}
 128
 129bool ceph_con_flag_test_and_set(struct ceph_connection *con,
 130				unsigned long con_flag)
 131{
 132	BUG_ON(!con_flag_valid(con_flag));
 133
 134	return test_and_set_bit(con_flag, &con->flags);
 135}
 136
 137/* Slab caches for frequently-allocated structures */
 138
 139static struct kmem_cache	*ceph_msg_cache;
 140
 141#ifdef CONFIG_LOCKDEP
 142static struct lock_class_key socket_class;
 143#endif
 144
 145static void queue_con(struct ceph_connection *con);
 146static void cancel_con(struct ceph_connection *con);
 147static void ceph_con_workfn(struct work_struct *);
 148static void con_fault(struct ceph_connection *con);
 149
 150/*
 151 * Nicely render a sockaddr as a string.  An array of formatted
 152 * strings is used, to approximate reentrancy.
 153 */
 154#define ADDR_STR_COUNT_LOG	5	/* log2(# address strings in array) */
 155#define ADDR_STR_COUNT		(1 << ADDR_STR_COUNT_LOG)
 156#define ADDR_STR_COUNT_MASK	(ADDR_STR_COUNT - 1)
 157#define MAX_ADDR_STR_LEN	64	/* 54 is enough */
 158
 159static char addr_str[ADDR_STR_COUNT][MAX_ADDR_STR_LEN];
 160static atomic_t addr_str_seq = ATOMIC_INIT(0);
 161
 162struct page *ceph_zero_page;		/* used in certain error cases */
 163
 164const char *ceph_pr_addr(const struct ceph_entity_addr *addr)
 165{
 166	int i;
 167	char *s;
 168	struct sockaddr_storage ss = addr->in_addr; /* align */
 169	struct sockaddr_in *in4 = (struct sockaddr_in *)&ss;
 170	struct sockaddr_in6 *in6 = (struct sockaddr_in6 *)&ss;
 171
 172	i = atomic_inc_return(&addr_str_seq) & ADDR_STR_COUNT_MASK;
 173	s = addr_str[i];
 174
 175	switch (ss.ss_family) {
 176	case AF_INET:
 177		snprintf(s, MAX_ADDR_STR_LEN, "(%d)%pI4:%hu",
 178			 le32_to_cpu(addr->type), &in4->sin_addr,
 179			 ntohs(in4->sin_port));
 180		break;
 181
 182	case AF_INET6:
 183		snprintf(s, MAX_ADDR_STR_LEN, "(%d)[%pI6c]:%hu",
 184			 le32_to_cpu(addr->type), &in6->sin6_addr,
 185			 ntohs(in6->sin6_port));
 186		break;
 187
 188	default:
 189		snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %hu)",
 190			 ss.ss_family);
 191	}
 192
 193	return s;
 194}
 195EXPORT_SYMBOL(ceph_pr_addr);
 196
 197void ceph_encode_my_addr(struct ceph_messenger *msgr)
 198{
 199	if (!ceph_msgr2(from_msgr(msgr))) {
 200		memcpy(&msgr->my_enc_addr, &msgr->inst.addr,
 201		       sizeof(msgr->my_enc_addr));
 202		ceph_encode_banner_addr(&msgr->my_enc_addr);
 203	}
 204}
 205
 206/*
 207 * work queue for all reading and writing to/from the socket.
 208 */
 209static struct workqueue_struct *ceph_msgr_wq;
 210
 211static int ceph_msgr_slab_init(void)
 212{
 213	BUG_ON(ceph_msg_cache);
 214	ceph_msg_cache = KMEM_CACHE(ceph_msg, 0);
 215	if (!ceph_msg_cache)
 216		return -ENOMEM;
 217
 218	return 0;
 219}
 220
 221static void ceph_msgr_slab_exit(void)
 222{
 223	BUG_ON(!ceph_msg_cache);
 224	kmem_cache_destroy(ceph_msg_cache);
 225	ceph_msg_cache = NULL;
 226}
 227
 228static void _ceph_msgr_exit(void)
 229{
 230	if (ceph_msgr_wq) {
 231		destroy_workqueue(ceph_msgr_wq);
 232		ceph_msgr_wq = NULL;
 233	}
 234
 235	BUG_ON(!ceph_zero_page);
 236	put_page(ceph_zero_page);
 237	ceph_zero_page = NULL;
 238
 239	ceph_msgr_slab_exit();
 240}
 241
 242int __init ceph_msgr_init(void)
 243{
 244	if (ceph_msgr_slab_init())
 245		return -ENOMEM;
 246
 247	BUG_ON(ceph_zero_page);
 248	ceph_zero_page = ZERO_PAGE(0);
 249	get_page(ceph_zero_page);
 250
 251	/*
 252	 * The number of active work items is limited by the number of
 253	 * connections, so leave @max_active at default.
 254	 */
 255	ceph_msgr_wq = alloc_workqueue("ceph-msgr",
 256				       WQ_MEM_RECLAIM | WQ_PERCPU, 0);
 257	if (ceph_msgr_wq)
 258		return 0;
 259
 260	pr_err("msgr_init failed to create workqueue\n");
 261	_ceph_msgr_exit();
 262
 263	return -ENOMEM;
 264}
 265
 266void ceph_msgr_exit(void)
 267{
 268	BUG_ON(ceph_msgr_wq == NULL);
 269
 270	_ceph_msgr_exit();
 271}
 272
 273void ceph_msgr_flush(void)
 274{
 275	flush_workqueue(ceph_msgr_wq);
 276}
 277EXPORT_SYMBOL(ceph_msgr_flush);
 278
 279/* Connection socket state transition functions */
 280
 281static void con_sock_state_init(struct ceph_connection *con)
 282{
 283	int old_state;
 284
 285	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED);
 286	if (WARN_ON(old_state != CON_SOCK_STATE_NEW))
 287		printk("%s: unexpected old state %d\n", __func__, old_state);
 288	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
 289	     CON_SOCK_STATE_CLOSED);
 290}
 291
 292static void con_sock_state_connecting(struct ceph_connection *con)
 293{
 294	int old_state;
 295
 296	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTING);
 297	if (WARN_ON(old_state != CON_SOCK_STATE_CLOSED))
 298		printk("%s: unexpected old state %d\n", __func__, old_state);
 299	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
 300	     CON_SOCK_STATE_CONNECTING);
 301}
 302
 303static void con_sock_state_connected(struct ceph_connection *con)
 304{
 305	int old_state;
 306
 307	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTED);
 308	if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING))
 309		printk("%s: unexpected old state %d\n", __func__, old_state);
 310	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
 311	     CON_SOCK_STATE_CONNECTED);
 312}
 313
 314static void con_sock_state_closing(struct ceph_connection *con)
 315{
 316	int old_state;
 317
 318	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSING);
 319	if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING &&
 320			old_state != CON_SOCK_STATE_CONNECTED &&
 321			old_state != CON_SOCK_STATE_CLOSING))
 322		printk("%s: unexpected old state %d\n", __func__, old_state);
 323	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
 324	     CON_SOCK_STATE_CLOSING);
 325}
 326
 327static void con_sock_state_closed(struct ceph_connection *con)
 328{
 329	int old_state;
 330
 331	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED);
 332	if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTED &&
 333		    old_state != CON_SOCK_STATE_CLOSING &&
 334		    old_state != CON_SOCK_STATE_CONNECTING &&
 335		    old_state != CON_SOCK_STATE_CLOSED))
 336		printk("%s: unexpected old state %d\n", __func__, old_state);
 337	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
 338	     CON_SOCK_STATE_CLOSED);
 339}
 340
 341/*
 342 * socket callback functions
 343 */
 344
 345/* data available on socket, or listen socket received a connect */
 346static void ceph_sock_data_ready(struct sock *sk)
 347{
 348	struct ceph_connection *con = sk->sk_user_data;
 349
 350	trace_sk_data_ready(sk);
 351
 352	if (atomic_read(&con->msgr->stopping)) {
 353		return;
 354	}
 355
 356	if (sk->sk_state != TCP_CLOSE_WAIT) {
 357		dout("%s %p state = %d, queueing work\n", __func__,
 358		     con, con->state);
 359		queue_con(con);
 360	}
 361}
 362
 363/* socket has buffer space for writing */
 364static void ceph_sock_write_space(struct sock *sk)
 365{
 366	struct ceph_connection *con = sk->sk_user_data;
 367
 368	/* only queue to workqueue if there is data we want to write,
 369	 * and there is sufficient space in the socket buffer to accept
 370	 * more data.  clear SOCK_NOSPACE so that ceph_sock_write_space()
 371	 * doesn't get called again until try_write() fills the socket
 372	 * buffer. See net/ipv4/tcp_input.c:tcp_check_space()
 373	 * and net/core/stream.c:sk_stream_write_space().
 374	 */
 375	if (ceph_con_flag_test(con, CEPH_CON_F_WRITE_PENDING)) {
 376		if (sk_stream_is_writeable(sk)) {
 377			dout("%s %p queueing write work\n", __func__, con);
 378			clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
 379			queue_con(con);
 380		}
 381	} else {
 382		dout("%s %p nothing to write\n", __func__, con);
 383	}
 384}
 385
 386/* socket's state has changed */
 387static void ceph_sock_state_change(struct sock *sk)
 388{
 389	struct ceph_connection *con = sk->sk_user_data;
 390
 391	dout("%s %p state = %d sk_state = %u\n", __func__,
 392	     con, con->state, sk->sk_state);
 393
 394	switch (sk->sk_state) {
 395	case TCP_CLOSE:
 396		dout("%s TCP_CLOSE\n", __func__);
 397		fallthrough;
 398	case TCP_CLOSE_WAIT:
 399		dout("%s TCP_CLOSE_WAIT\n", __func__);
 400		con_sock_state_closing(con);
 401		ceph_con_flag_set(con, CEPH_CON_F_SOCK_CLOSED);
 402		queue_con(con);
 403		break;
 404	case TCP_ESTABLISHED:
 405		dout("%s TCP_ESTABLISHED\n", __func__);
 406		con_sock_state_connected(con);
 407		queue_con(con);
 408		break;
 409	default:	/* Everything else is uninteresting */
 410		break;
 411	}
 412}
 413
 414/*
 415 * set up socket callbacks
 416 */
 417static void set_sock_callbacks(struct socket *sock,
 418			       struct ceph_connection *con)
 419{
 420	struct sock *sk = sock->sk;
 421	sk->sk_user_data = con;
 422	sk->sk_data_ready = ceph_sock_data_ready;
 423	sk->sk_write_space = ceph_sock_write_space;
 424	sk->sk_state_change = ceph_sock_state_change;
 425}
 426
 427
 428/*
 429 * socket helpers
 430 */
 431
 432/*
 433 * initiate connection to a remote socket.
 434 */
 435int ceph_tcp_connect(struct ceph_connection *con)
 436{
 437	struct sockaddr_storage ss = con->peer_addr.in_addr; /* align */
 438	struct socket *sock;
 439	unsigned int noio_flag;
 440	int ret;
 441
 442	dout("%s con %p peer_addr %s\n", __func__, con,
 443	     ceph_pr_addr(&con->peer_addr));
 444	BUG_ON(con->sock);
 445
 446	/* sock_create_kern() allocates with GFP_KERNEL */
 447	noio_flag = memalloc_noio_save();
 448	ret = sock_create_kern(read_pnet(&con->msgr->net), ss.ss_family,
 449			       SOCK_STREAM, IPPROTO_TCP, &sock);
 450	memalloc_noio_restore(noio_flag);
 451	if (ret)
 452		return ret;
 453	sock->sk->sk_allocation = GFP_NOFS;
 454	sock->sk->sk_use_task_frag = false;
 455
 456#ifdef CONFIG_LOCKDEP
 457	lockdep_set_class(&sock->sk->sk_lock, &socket_class);
 458#endif
 459
 460	set_sock_callbacks(sock, con);
 461
 462	con_sock_state_connecting(con);
 463	ret = kernel_connect(sock, (struct sockaddr_unsized *)&ss, sizeof(ss),
 464			     O_NONBLOCK);
 465	if (ret == -EINPROGRESS) {
 466		dout("connect %s EINPROGRESS sk_state = %u\n",
 467		     ceph_pr_addr(&con->peer_addr),
 468		     sock->sk->sk_state);
 469	} else if (ret < 0) {
 470		pr_err("connect %s error %d\n",
 471		       ceph_pr_addr(&con->peer_addr), ret);
 472		sock_release(sock);
 473		return ret;
 474	}
 475
 476	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
 477		tcp_sock_set_nodelay(sock->sk);
 478
 479	con->sock = sock;
 480	return 0;
 481}
 482
 483/*
 484 * Shutdown/close the socket for the given connection.
 485 */
 486int ceph_con_close_socket(struct ceph_connection *con)
 487{
 488	int rc = 0;
 489
 490	dout("%s con %p sock %p\n", __func__, con, con->sock);
 491	if (con->sock) {
 492		rc = con->sock->ops->shutdown(con->sock, SHUT_RDWR);
 493		sock_release(con->sock);
 494		con->sock = NULL;
 495	}
 496
 497	/*
 498	 * Forcibly clear the SOCK_CLOSED flag.  It gets set
 499	 * independent of the connection mutex, and we could have
 500	 * received a socket close event before we had the chance to
 501	 * shut the socket down.
 502	 */
 503	ceph_con_flag_clear(con, CEPH_CON_F_SOCK_CLOSED);
 504
 505	con_sock_state_closed(con);
 506	return rc;
 507}
 508
 509static void ceph_con_reset_protocol(struct ceph_connection *con)
 510{
 511	dout("%s con %p\n", __func__, con);
 512
 513	ceph_con_close_socket(con);
 514	if (con->in_msg) {
 515		WARN_ON(con->in_msg->con != con);
 516		ceph_msg_put(con->in_msg);
 517		con->in_msg = NULL;
 518	}
 519	if (con->out_msg) {
 520		WARN_ON(con->out_msg->con != con);
 521		ceph_msg_put(con->out_msg);
 522		con->out_msg = NULL;
 523	}
 524	if (con->bounce_page) {
 525		__free_page(con->bounce_page);
 526		con->bounce_page = NULL;
 527	}
 528
 529	if (ceph_msgr2(from_msgr(con->msgr)))
 530		ceph_con_v2_reset_protocol(con);
 531	else
 532		ceph_con_v1_reset_protocol(con);
 533}
 534
 535/*
 536 * Reset a connection.  Discard all incoming and outgoing messages
 537 * and clear *_seq state.
 538 */
 539static void ceph_msg_remove(struct ceph_msg *msg)
 540{
 541	list_del_init(&msg->list_head);
 542
 543	ceph_msg_put(msg);
 544}
 545
 546static void ceph_msg_remove_list(struct list_head *head)
 547{
 548	while (!list_empty(head)) {
 549		struct ceph_msg *msg = list_first_entry(head, struct ceph_msg,
 550							list_head);
 551		ceph_msg_remove(msg);
 552	}
 553}
 554
 555void ceph_con_reset_session(struct ceph_connection *con)
 556{
 557	dout("%s con %p\n", __func__, con);
 558
 559	WARN_ON(con->in_msg);
 560	WARN_ON(con->out_msg);
 561	ceph_msg_remove_list(&con->out_queue);
 562	ceph_msg_remove_list(&con->out_sent);
 563	con->out_seq = 0;
 564	con->in_seq = 0;
 565	con->in_seq_acked = 0;
 566
 567	if (ceph_msgr2(from_msgr(con->msgr)))
 568		ceph_con_v2_reset_session(con);
 569	else
 570		ceph_con_v1_reset_session(con);
 571}
 572
 573/*
 574 * mark a peer down.  drop any open connections.
 575 */
 576void ceph_con_close(struct ceph_connection *con)
 577{
 578	mutex_lock(&con->mutex);
 579	dout("con_close %p peer %s\n", con, ceph_pr_addr(&con->peer_addr));
 580	con->state = CEPH_CON_S_CLOSED;
 581
 582	ceph_con_flag_clear(con, CEPH_CON_F_LOSSYTX);  /* so we retry next
 583							  connect */
 584	ceph_con_flag_clear(con, CEPH_CON_F_KEEPALIVE_PENDING);
 585	ceph_con_flag_clear(con, CEPH_CON_F_WRITE_PENDING);
 586	ceph_con_flag_clear(con, CEPH_CON_F_BACKOFF);
 587
 588	ceph_con_reset_protocol(con);
 589	ceph_con_reset_session(con);
 590	cancel_con(con);
 591	mutex_unlock(&con->mutex);
 592}
 593EXPORT_SYMBOL(ceph_con_close);
 594
 595/*
 596 * Reopen a closed connection, with a new peer address.
 597 */
 598void ceph_con_open(struct ceph_connection *con,
 599		   __u8 entity_type, __u64 entity_num,
 600		   struct ceph_entity_addr *addr)
 601{
 602	mutex_lock(&con->mutex);
 603	dout("con_open %p %s\n", con, ceph_pr_addr(addr));
 604
 605	WARN_ON(con->state != CEPH_CON_S_CLOSED);
 606	con->state = CEPH_CON_S_PREOPEN;
 607
 608	con->peer_name.type = (__u8) entity_type;
 609	con->peer_name.num = cpu_to_le64(entity_num);
 610
 611	memcpy(&con->peer_addr, addr, sizeof(*addr));
 612	con->delay = 0;      /* reset backoff memory */
 613	mutex_unlock(&con->mutex);
 614	queue_con(con);
 615}
 616EXPORT_SYMBOL(ceph_con_open);
 617
 618/*
 619 * return true if this connection ever successfully opened
 620 */
 621bool ceph_con_opened(struct ceph_connection *con)
 622{
 623	if (ceph_msgr2(from_msgr(con->msgr)))
 624		return ceph_con_v2_opened(con);
 625
 626	return ceph_con_v1_opened(con);
 627}
 628
 629/*
 630 * initialize a new connection.
 631 */
 632void ceph_con_init(struct ceph_connection *con, void *private,
 633	const struct ceph_connection_operations *ops,
 634	struct ceph_messenger *msgr)
 635{
 636	dout("con_init %p\n", con);
 637	memset(con, 0, sizeof(*con));
 638	con->private = private;
 639	con->ops = ops;
 640	con->msgr = msgr;
 641
 642	con_sock_state_init(con);
 643
 644	mutex_init(&con->mutex);
 645	INIT_LIST_HEAD(&con->out_queue);
 646	INIT_LIST_HEAD(&con->out_sent);
 647	INIT_DELAYED_WORK(&con->work, ceph_con_workfn);
 648
 649	con->state = CEPH_CON_S_CLOSED;
 650}
 651EXPORT_SYMBOL(ceph_con_init);
 652
 653/*
 654 * We maintain a global counter to order connection attempts.  Get
 655 * a unique seq greater than @gt.
 656 */
 657u32 ceph_get_global_seq(struct ceph_messenger *msgr, u32 gt)
 658{
 659	u32 ret;
 660
 661	spin_lock(&msgr->global_seq_lock);
 662	if (msgr->global_seq < gt)
 663		msgr->global_seq = gt;
 664	ret = ++msgr->global_seq;
 665	spin_unlock(&msgr->global_seq_lock);
 666	return ret;
 667}
 668
 669/*
 670 * Discard messages that have been acked by the server.
 671 */
 672void ceph_con_discard_sent(struct ceph_connection *con, u64 ack_seq)
 673{
 674	struct ceph_msg *msg;
 675	u64 seq;
 676
 677	dout("%s con %p ack_seq %llu\n", __func__, con, ack_seq);
 678	while (!list_empty(&con->out_sent)) {
 679		msg = list_first_entry(&con->out_sent, struct ceph_msg,
 680				       list_head);
 681		WARN_ON(msg->needs_out_seq);
 682		seq = le64_to_cpu(msg->hdr.seq);
 683		if (seq > ack_seq)
 684			break;
 685
 686		dout("%s con %p discarding msg %p seq %llu\n", __func__, con,
 687		     msg, seq);
 688		ceph_msg_remove(msg);
 689	}
 690}
 691
 692/*
 693 * Discard messages that have been requeued in con_fault(), up to
 694 * reconnect_seq.  This avoids gratuitously resending messages that
 695 * the server had received and handled prior to reconnect.
 696 */
 697void ceph_con_discard_requeued(struct ceph_connection *con, u64 reconnect_seq)
 698{
 699	struct ceph_msg *msg;
 700	u64 seq;
 701
 702	dout("%s con %p reconnect_seq %llu\n", __func__, con, reconnect_seq);
 703	while (!list_empty(&con->out_queue)) {
 704		msg = list_first_entry(&con->out_queue, struct ceph_msg,
 705				       list_head);
 706		if (msg->needs_out_seq)
 707			break;
 708		seq = le64_to_cpu(msg->hdr.seq);
 709		if (seq > reconnect_seq)
 710			break;
 711
 712		dout("%s con %p discarding msg %p seq %llu\n", __func__, con,
 713		     msg, seq);
 714		ceph_msg_remove(msg);
 715	}
 716}
 717
 718#ifdef CONFIG_BLOCK
 719
 720/*
 721 * For a bio data item, a piece is whatever remains of the next
 722 * entry in the current bio iovec, or the first entry in the next
 723 * bio in the list.
 724 */
 725static void ceph_msg_data_bio_cursor_init(struct ceph_msg_data_cursor *cursor,
 726					size_t length)
 727{
 728	struct ceph_msg_data *data = cursor->data;
 729	struct ceph_bio_iter *it = &cursor->bio_iter;
 730
 731	cursor->resid = min_t(size_t, length, data->bio_length);
 732	*it = data->bio_pos;
 733	if (cursor->resid < it->iter.bi_size)
 734		it->iter.bi_size = cursor->resid;
 735
 736	BUG_ON(cursor->resid < bio_iter_len(it->bio, it->iter));
 737}
 738
 739static struct page *ceph_msg_data_bio_next(struct ceph_msg_data_cursor *cursor,
 740						size_t *page_offset,
 741						size_t *length)
 742{
 743	struct bio_vec bv = bio_iter_iovec(cursor->bio_iter.bio,
 744					   cursor->bio_iter.iter);
 745
 746	*page_offset = bv.bv_offset;
 747	*length = bv.bv_len;
 748	return bv.bv_page;
 749}
 750
 751static bool ceph_msg_data_bio_advance(struct ceph_msg_data_cursor *cursor,
 752					size_t bytes)
 753{
 754	struct ceph_bio_iter *it = &cursor->bio_iter;
 755	struct page *page = bio_iter_page(it->bio, it->iter);
 756
 757	BUG_ON(bytes > cursor->resid);
 758	BUG_ON(bytes > bio_iter_len(it->bio, it->iter));
 759	cursor->resid -= bytes;
 760	bio_advance_iter(it->bio, &it->iter, bytes);
 761
 762	if (!cursor->resid)
 763		return false;   /* no more data */
 764
 765	if (!bytes || (it->iter.bi_size && it->iter.bi_bvec_done &&
 766		       page == bio_iter_page(it->bio, it->iter)))
 767		return false;	/* more bytes to process in this segment */
 768
 769	if (!it->iter.bi_size) {
 770		it->bio = it->bio->bi_next;
 771		it->iter = it->bio->bi_iter;
 772		if (cursor->resid < it->iter.bi_size)
 773			it->iter.bi_size = cursor->resid;
 774	}
 775
 776	BUG_ON(cursor->resid < bio_iter_len(it->bio, it->iter));
 777	return true;
 778}
 779#endif /* CONFIG_BLOCK */
 780
 781static void ceph_msg_data_bvecs_cursor_init(struct ceph_msg_data_cursor *cursor,
 782					size_t length)
 783{
 784	struct ceph_msg_data *data = cursor->data;
 785	struct bio_vec *bvecs = data->bvec_pos.bvecs;
 786
 787	cursor->resid = min_t(size_t, length, data->bvec_pos.iter.bi_size);
 788	cursor->bvec_iter = data->bvec_pos.iter;
 789	cursor->bvec_iter.bi_size = cursor->resid;
 790
 791	BUG_ON(cursor->resid < bvec_iter_len(bvecs, cursor->bvec_iter));
 792}
 793
 794static struct page *ceph_msg_data_bvecs_next(struct ceph_msg_data_cursor *cursor,
 795						size_t *page_offset,
 796						size_t *length)
 797{
 798	struct bio_vec bv = bvec_iter_bvec(cursor->data->bvec_pos.bvecs,
 799					   cursor->bvec_iter);
 800
 801	*page_offset = bv.bv_offset;
 802	*length = bv.bv_len;
 803	return bv.bv_page;
 804}
 805
 806static bool ceph_msg_data_bvecs_advance(struct ceph_msg_data_cursor *cursor,
 807					size_t bytes)
 808{
 809	struct bio_vec *bvecs = cursor->data->bvec_pos.bvecs;
 810	struct page *page = bvec_iter_page(bvecs, cursor->bvec_iter);
 811
 812	BUG_ON(bytes > cursor->resid);
 813	BUG_ON(bytes > bvec_iter_len(bvecs, cursor->bvec_iter));
 814	cursor->resid -= bytes;
 815	bvec_iter_advance(bvecs, &cursor->bvec_iter, bytes);
 816
 817	if (!cursor->resid)
 818		return false;   /* no more data */
 819
 820	if (!bytes || (cursor->bvec_iter.bi_bvec_done &&
 821		       page == bvec_iter_page(bvecs, cursor->bvec_iter)))
 822		return false;	/* more bytes to process in this segment */
 823
 824	BUG_ON(cursor->resid < bvec_iter_len(bvecs, cursor->bvec_iter));
 825	return true;
 826}
 827
 828/*
 829 * For a page array, a piece comes from the first page in the array
 830 * that has not already been fully consumed.
 831 */
 832static void ceph_msg_data_pages_cursor_init(struct ceph_msg_data_cursor *cursor,
 833					size_t length)
 834{
 835	struct ceph_msg_data *data = cursor->data;
 836	int page_count;
 837
 838	BUG_ON(data->type != CEPH_MSG_DATA_PAGES);
 839
 840	BUG_ON(!data->pages);
 841	BUG_ON(!data->length);
 842
 843	cursor->resid = min(length, data->length);
 844	page_count = calc_pages_for(data->alignment, (u64)data->length);
 845	cursor->page_offset = data->alignment & ~PAGE_MASK;
 846	cursor->page_index = 0;
 847	BUG_ON(page_count > (int)USHRT_MAX);
 848	cursor->page_count = (unsigned short)page_count;
 849	BUG_ON(length > SIZE_MAX - cursor->page_offset);
 850}
 851
 852static struct page *
 853ceph_msg_data_pages_next(struct ceph_msg_data_cursor *cursor,
 854					size_t *page_offset, size_t *length)
 855{
 856	struct ceph_msg_data *data = cursor->data;
 857
 858	BUG_ON(data->type != CEPH_MSG_DATA_PAGES);
 859
 860	BUG_ON(cursor->page_index >= cursor->page_count);
 861	BUG_ON(cursor->page_offset >= PAGE_SIZE);
 862
 863	*page_offset = cursor->page_offset;
 864	*length = min_t(size_t, cursor->resid, PAGE_SIZE - *page_offset);
 865	return data->pages[cursor->page_index];
 866}
 867
 868static bool ceph_msg_data_pages_advance(struct ceph_msg_data_cursor *cursor,
 869						size_t bytes)
 870{
 871	BUG_ON(cursor->data->type != CEPH_MSG_DATA_PAGES);
 872
 873	BUG_ON(cursor->page_offset + bytes > PAGE_SIZE);
 874
 875	/* Advance the cursor page offset */
 876
 877	cursor->resid -= bytes;
 878	cursor->page_offset = (cursor->page_offset + bytes) & ~PAGE_MASK;
 879	if (!bytes || cursor->page_offset)
 880		return false;	/* more bytes to process in the current page */
 881
 882	if (!cursor->resid)
 883		return false;   /* no more data */
 884
 885	/* Move on to the next page; offset is already at 0 */
 886
 887	BUG_ON(cursor->page_index >= cursor->page_count);
 888	cursor->page_index++;
 889	return true;
 890}
 891
 892/*
 893 * For a pagelist, a piece is whatever remains to be consumed in the
 894 * first page in the list, or the front of the next page.
 895 */
 896static void
 897ceph_msg_data_pagelist_cursor_init(struct ceph_msg_data_cursor *cursor,
 898					size_t length)
 899{
 900	struct ceph_msg_data *data = cursor->data;
 901	struct ceph_pagelist *pagelist;
 902	struct page *page;
 903
 904	BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST);
 905
 906	pagelist = data->pagelist;
 907	BUG_ON(!pagelist);
 908
 909	if (!length)
 910		return;		/* pagelist can be assigned but empty */
 911
 912	BUG_ON(list_empty(&pagelist->head));
 913	page = list_first_entry(&pagelist->head, struct page, lru);
 914
 915	cursor->resid = min(length, pagelist->length);
 916	cursor->page = page;
 917	cursor->offset = 0;
 918}
 919
 920static struct page *
 921ceph_msg_data_pagelist_next(struct ceph_msg_data_cursor *cursor,
 922				size_t *page_offset, size_t *length)
 923{
 924	struct ceph_msg_data *data = cursor->data;
 925	struct ceph_pagelist *pagelist;
 926
 927	BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST);
 928
 929	pagelist = data->pagelist;
 930	BUG_ON(!pagelist);
 931
 932	BUG_ON(!cursor->page);
 933	BUG_ON(cursor->offset + cursor->resid != pagelist->length);
 934
 935	/* offset of first page in pagelist is always 0 */
 936	*page_offset = cursor->offset & ~PAGE_MASK;
 937	*length = min_t(size_t, cursor->resid, PAGE_SIZE - *page_offset);
 938	return cursor->page;
 939}
 940
 941static bool ceph_msg_data_pagelist_advance(struct ceph_msg_data_cursor *cursor,
 942						size_t bytes)
 943{
 944	struct ceph_msg_data *data = cursor->data;
 945	struct ceph_pagelist *pagelist;
 946
 947	BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST);
 948
 949	pagelist = data->pagelist;
 950	BUG_ON(!pagelist);
 951
 952	BUG_ON(cursor->offset + cursor->resid != pagelist->length);
 953	BUG_ON((cursor->offset & ~PAGE_MASK) + bytes > PAGE_SIZE);
 954
 955	/* Advance the cursor offset */
 956
 957	cursor->resid -= bytes;
 958	cursor->offset += bytes;
 959	/* offset of first page in pagelist is always 0 */
 960	if (!bytes || cursor->offset & ~PAGE_MASK)
 961		return false;	/* more bytes to process in the current page */
 962
 963	if (!cursor->resid)
 964		return false;   /* no more data */
 965
 966	/* Move on to the next page */
 967
 968	BUG_ON(list_is_last(&cursor->page->lru, &pagelist->head));
 969	cursor->page = list_next_entry(cursor->page, lru);
 970	return true;
 971}
 972
 973static void ceph_msg_data_iter_cursor_init(struct ceph_msg_data_cursor *cursor,
 974					   size_t length)
 975{
 976	struct ceph_msg_data *data = cursor->data;
 977
 978	cursor->iov_iter = data->iter;
 979	cursor->lastlen = 0;
 980	iov_iter_truncate(&cursor->iov_iter, length);
 981	cursor->resid = iov_iter_count(&cursor->iov_iter);
 982}
 983
 984static struct page *ceph_msg_data_iter_next(struct ceph_msg_data_cursor *cursor,
 985					    size_t *page_offset, size_t *length)
 986{
 987	struct page *page;
 988	ssize_t len;
 989
 990	if (cursor->lastlen)
 991		iov_iter_revert(&cursor->iov_iter, cursor->lastlen);
 992
 993	len = iov_iter_get_pages2(&cursor->iov_iter, &page, PAGE_SIZE,
 994				  1, page_offset);
 995	BUG_ON(len < 0);
 996
 997	cursor->lastlen = len;
 998
 999	/*
1000	 * FIXME: The assumption is that the pages represented by the iov_iter
1001	 *	  are pinned, with the references held by the upper-level
1002	 *	  callers, or by virtue of being under writeback. Eventually,
1003	 *	  we'll get an iov_iter_get_pages2 variant that doesn't take
1004	 *	  page refs. Until then, just put the page ref.
1005	 */
1006	VM_BUG_ON_PAGE(!PageWriteback(page) && page_count(page) < 2, page);
1007	put_page(page);
1008
1009	*length = min_t(size_t, len, cursor->resid);
1010	return page;
1011}
1012
1013static bool ceph_msg_data_iter_advance(struct ceph_msg_data_cursor *cursor,
1014				       size_t bytes)
1015{
1016	BUG_ON(bytes > cursor->resid);
1017	cursor->resid -= bytes;
1018
1019	if (bytes < cursor->lastlen) {
1020		cursor->lastlen -= bytes;
1021	} else {
1022		iov_iter_advance(&cursor->iov_iter, bytes - cursor->lastlen);
1023		cursor->lastlen = 0;
1024	}
1025
1026	return cursor->resid;
1027}
1028
1029/*
1030 * Message data is handled (sent or received) in pieces, where each
1031 * piece resides on a single page.  The network layer might not
1032 * consume an entire piece at once.  A data item's cursor keeps
1033 * track of which piece is next to process and how much remains to
1034 * be processed in that piece.  It also tracks whether the current
1035 * piece is the last one in the data item.
1036 */
1037static void __ceph_msg_data_cursor_init(struct ceph_msg_data_cursor *cursor)
1038{
1039	size_t length = cursor->total_resid;
1040
1041	switch (cursor->data->type) {
1042	case CEPH_MSG_DATA_PAGELIST:
1043		ceph_msg_data_pagelist_cursor_init(cursor, length);
1044		break;
1045	case CEPH_MSG_DATA_PAGES:
1046		ceph_msg_data_pages_cursor_init(cursor, length);
1047		break;
1048#ifdef CONFIG_BLOCK
1049	case CEPH_MSG_DATA_BIO:
1050		ceph_msg_data_bio_cursor_init(cursor, length);
1051		break;
1052#endif /* CONFIG_BLOCK */
1053	case CEPH_MSG_DATA_BVECS:
1054		ceph_msg_data_bvecs_cursor_init(cursor, length);
1055		break;
1056	case CEPH_MSG_DATA_ITER:
1057		ceph_msg_data_iter_cursor_init(cursor, length);
1058		break;
1059	case CEPH_MSG_DATA_NONE:
1060	default:
1061		/* BUG(); */
1062		break;
1063	}
1064	cursor->need_crc = true;
1065}
1066
1067void ceph_msg_data_cursor_init(struct ceph_msg_data_cursor *cursor,
1068			       struct ceph_msg *msg, size_t length)
1069{
1070	BUG_ON(!length);
1071	BUG_ON(length > msg->data_length);
1072	BUG_ON(!msg->num_data_items);
1073
1074	cursor->total_resid = length;
1075	cursor->data = msg->data;
1076	cursor->sr_resid = 0;
1077
1078	__ceph_msg_data_cursor_init(cursor);
1079}
1080
1081/*
1082 * Return the page containing the next piece to process for a given
1083 * data item, and supply the page offset and length of that piece.
1084 * Indicate whether this is the last piece in this data item.
1085 */
1086struct page *ceph_msg_data_next(struct ceph_msg_data_cursor *cursor,
1087				size_t *page_offset, size_t *length)
1088{
1089	struct page *page;
1090
1091	switch (cursor->data->type) {
1092	case CEPH_MSG_DATA_PAGELIST:
1093		page = ceph_msg_data_pagelist_next(cursor, page_offset, length);
1094		break;
1095	case CEPH_MSG_DATA_PAGES:
1096		page = ceph_msg_data_pages_next(cursor, page_offset, length);
1097		break;
1098#ifdef CONFIG_BLOCK
1099	case CEPH_MSG_DATA_BIO:
1100		page = ceph_msg_data_bio_next(cursor, page_offset, length);
1101		break;
1102#endif /* CONFIG_BLOCK */
1103	case CEPH_MSG_DATA_BVECS:
1104		page = ceph_msg_data_bvecs_next(cursor, page_offset, length);
1105		break;
1106	case CEPH_MSG_DATA_ITER:
1107		page = ceph_msg_data_iter_next(cursor, page_offset, length);
1108		break;
1109	case CEPH_MSG_DATA_NONE:
1110	default:
1111		page = NULL;
1112		break;
1113	}
1114
1115	BUG_ON(!page);
1116	BUG_ON(*page_offset + *length > PAGE_SIZE);
1117	BUG_ON(!*length);
1118	BUG_ON(*length > cursor->resid);
1119
1120	return page;
1121}
1122
1123/*
1124 * Returns true if the result moves the cursor on to the next piece
1125 * of the data item.
1126 */
1127void ceph_msg_data_advance(struct ceph_msg_data_cursor *cursor, size_t bytes)
1128{
1129	bool new_piece;
1130
1131	BUG_ON(bytes > cursor->resid);
1132	switch (cursor->data->type) {
1133	case CEPH_MSG_DATA_PAGELIST:
1134		new_piece = ceph_msg_data_pagelist_advance(cursor, bytes);
1135		break;
1136	case CEPH_MSG_DATA_PAGES:
1137		new_piece = ceph_msg_data_pages_advance(cursor, bytes);
1138		break;
1139#ifdef CONFIG_BLOCK
1140	case CEPH_MSG_DATA_BIO:
1141		new_piece = ceph_msg_data_bio_advance(cursor, bytes);
1142		break;
1143#endif /* CONFIG_BLOCK */
1144	case CEPH_MSG_DATA_BVECS:
1145		new_piece = ceph_msg_data_bvecs_advance(cursor, bytes);
1146		break;
1147	case CEPH_MSG_DATA_ITER:
1148		new_piece = ceph_msg_data_iter_advance(cursor, bytes);
1149		break;
1150	case CEPH_MSG_DATA_NONE:
1151	default:
1152		BUG();
1153		break;
1154	}
1155	cursor->total_resid -= bytes;
1156
1157	if (!cursor->resid && cursor->total_resid) {
1158		cursor->data++;
1159		__ceph_msg_data_cursor_init(cursor);
1160		new_piece = true;
1161	}
1162	cursor->need_crc = new_piece;
1163}
1164
1165u32 ceph_crc32c_page(u32 crc, struct page *page, unsigned int page_offset,
1166		     unsigned int length)
1167{
1168	char *kaddr;
1169
1170	kaddr = kmap(page);
1171	BUG_ON(kaddr == NULL);
1172	crc = crc32c(crc, kaddr + page_offset, length);
1173	kunmap(page);
1174
1175	return crc;
1176}
1177
1178bool ceph_addr_is_blank(const struct ceph_entity_addr *addr)
1179{
1180	struct sockaddr_storage ss = addr->in_addr; /* align */
1181	struct in_addr *addr4 = &((struct sockaddr_in *)&ss)->sin_addr;
1182	struct in6_addr *addr6 = &((struct sockaddr_in6 *)&ss)->sin6_addr;
1183
1184	switch (ss.ss_family) {
1185	case AF_INET:
1186		return addr4->s_addr == htonl(INADDR_ANY);
1187	case AF_INET6:
1188		return ipv6_addr_any(addr6);
1189	default:
1190		return true;
1191	}
1192}
1193EXPORT_SYMBOL(ceph_addr_is_blank);
1194
1195int ceph_addr_port(const struct ceph_entity_addr *addr)
1196{
1197	switch (get_unaligned(&addr->in_addr.ss_family)) {
1198	case AF_INET:
1199		return ntohs(get_unaligned(&((struct sockaddr_in *)&addr->in_addr)->sin_port));
1200	case AF_INET6:
1201		return ntohs(get_unaligned(&((struct sockaddr_in6 *)&addr->in_addr)->sin6_port));
1202	}
1203	return 0;
1204}
1205
1206void ceph_addr_set_port(struct ceph_entity_addr *addr, int p)
1207{
1208	switch (get_unaligned(&addr->in_addr.ss_family)) {
1209	case AF_INET:
1210		put_unaligned(htons(p), &((struct sockaddr_in *)&addr->in_addr)->sin_port);
1211		break;
1212	case AF_INET6:
1213		put_unaligned(htons(p), &((struct sockaddr_in6 *)&addr->in_addr)->sin6_port);
1214		break;
1215	}
1216}
1217
1218/*
1219 * Unlike other *_pton function semantics, zero indicates success.
1220 */
1221static int ceph_pton(const char *str, size_t len, struct ceph_entity_addr *addr,
1222		char delim, const char **ipend)
1223{
1224	memset(&addr->in_addr, 0, sizeof(addr->in_addr));
1225
1226	if (in4_pton(str, len, (u8 *)&((struct sockaddr_in *)&addr->in_addr)->sin_addr.s_addr, delim, ipend)) {
1227		put_unaligned(AF_INET, &addr->in_addr.ss_family);
1228		return 0;
1229	}
1230
1231	if (in6_pton(str, len, (u8 *)&((struct sockaddr_in6 *)&addr->in_addr)->sin6_addr.s6_addr, delim, ipend)) {
1232		put_unaligned(AF_INET6, &addr->in_addr.ss_family);
1233		return 0;
1234	}
1235
1236	return -EINVAL;
1237}
1238
1239/*
1240 * Extract hostname string and resolve using kernel DNS facility.
1241 */
1242#ifdef CONFIG_CEPH_LIB_USE_DNS_RESOLVER
1243static int ceph_dns_resolve_name(const char *name, size_t namelen,
1244		struct ceph_entity_addr *addr, char delim, const char **ipend)
1245{
1246	const char *end, *delim_p;
1247	char *colon_p, *ip_addr = NULL;
1248	int ip_len, ret;
1249
1250	/*
1251	 * The end of the hostname occurs immediately preceding the delimiter or
1252	 * the port marker (':') where the delimiter takes precedence.
1253	 */
1254	delim_p = memchr(name, delim, namelen);
1255	colon_p = memchr(name, ':', namelen);
1256
1257	if (delim_p && colon_p)
1258		end = min(delim_p, colon_p);
1259	else if (!delim_p && colon_p)
1260		end = colon_p;
1261	else {
1262		end = delim_p;
1263		if (!end) /* case: hostname:/ */
1264			end = name + namelen;
1265	}
1266
1267	if (end <= name)
1268		return -EINVAL;
1269
1270	/* do dns_resolve upcall */
1271	ip_len = dns_query(current->nsproxy->net_ns,
1272			   NULL, name, end - name, NULL, &ip_addr, NULL, false);
1273	if (ip_len > 0)
1274		ret = ceph_pton(ip_addr, ip_len, addr, -1, NULL);
1275	else
1276		ret = -ESRCH;
1277
1278	kfree(ip_addr);
1279
1280	*ipend = end;
1281
1282	pr_info("resolve '%.*s' (ret=%d): %s\n", (int)(end - name), name,
1283			ret, ret ? "failed" : ceph_pr_addr(addr));
1284
1285	return ret;
1286}
1287#else
1288static inline int ceph_dns_resolve_name(const char *name, size_t namelen,
1289		struct ceph_entity_addr *addr, char delim, const char **ipend)
1290{
1291	return -EINVAL;
1292}
1293#endif
1294
1295/*
1296 * Parse a server name (IP or hostname). If a valid IP address is not found
1297 * then try to extract a hostname to resolve using userspace DNS upcall.
1298 */
1299static int ceph_parse_server_name(const char *name, size_t namelen,
1300		struct ceph_entity_addr *addr, char delim, const char **ipend)
1301{
1302	int ret;
1303
1304	ret = ceph_pton(name, namelen, addr, delim, ipend);
1305	if (ret)
1306		ret = ceph_dns_resolve_name(name, namelen, addr, delim, ipend);
1307
1308	return ret;
1309}
1310
1311/*
1312 * Parse an ip[:port] list into an addr array.  Use the default
1313 * monitor port if a port isn't specified.
1314 */
1315int ceph_parse_ips(const char *c, const char *end,
1316		   struct ceph_entity_addr *addr,
1317		   int max_count, int *count, char delim)
1318{
1319	int i, ret = -EINVAL;
1320	const char *p = c;
1321
1322	dout("parse_ips on '%.*s'\n", (int)(end-c), c);
1323	for (i = 0; i < max_count; i++) {
1324		char cur_delim = delim;
1325		const char *ipend;
1326		int port;
1327
1328		if (*p == '[') {
1329			cur_delim = ']';
1330			p++;
1331		}
1332
1333		ret = ceph_parse_server_name(p, end - p, &addr[i], cur_delim,
1334					     &ipend);
1335		if (ret)
1336			goto bad;
1337		ret = -EINVAL;
1338
1339		p = ipend;
1340
1341		if (cur_delim == ']') {
1342			if (*p != ']') {
1343				dout("missing matching ']'\n");
1344				goto bad;
1345			}
1346			p++;
1347		}
1348
1349		/* port? */
1350		if (p < end && *p == ':') {
1351			port = 0;
1352			p++;
1353			while (p < end && *p >= '0' && *p <= '9') {
1354				port = (port * 10) + (*p - '0');
1355				p++;
1356			}
1357			if (port == 0)
1358				port = CEPH_MON_PORT;
1359			else if (port > 65535)
1360				goto bad;
1361		} else {
1362			port = CEPH_MON_PORT;
1363		}
1364
1365		ceph_addr_set_port(&addr[i], port);
1366		/*
1367		 * We want the type to be set according to ms_mode
1368		 * option, but options are normally parsed after mon
1369		 * addresses.  Rather than complicating parsing, set
1370		 * to LEGACY and override in build_initial_monmap()
1371		 * for mon addresses and ceph_messenger_init() for
1372		 * ip option.
1373		 */
1374		addr[i].type = CEPH_ENTITY_ADDR_TYPE_LEGACY;
1375		addr[i].nonce = 0;
1376
1377		dout("%s got %s\n", __func__, ceph_pr_addr(&addr[i]));
1378
1379		if (p == end)
1380			break;
1381		if (*p != delim)
1382			goto bad;
1383		p++;
1384	}
1385
1386	if (p != end)
1387		goto bad;
1388
1389	if (count)
1390		*count = i + 1;
1391	return 0;
1392
1393bad:
1394	return ret;
1395}
1396
1397/*
1398 * Process message.  This happens in the worker thread.  The callback should
1399 * be careful not to do anything that waits on other incoming messages or it
1400 * may deadlock.
1401 */
1402void ceph_con_process_message(struct ceph_connection *con)
1403{
1404	struct ceph_msg *msg = con->in_msg;
1405
1406	BUG_ON(con->in_msg->con != con);
1407	con->in_msg = NULL;
1408
1409	/* if first message, set peer_name */
1410	if (con->peer_name.type == 0)
1411		con->peer_name = msg->hdr.src;
1412
1413	con->in_seq++;
1414	mutex_unlock(&con->mutex);
1415
1416	dout("===== %p %llu from %s%lld %d=%s len %d+%d+%d (%u %u %u) =====\n",
1417	     msg, le64_to_cpu(msg->hdr.seq),
1418	     ENTITY_NAME(msg->hdr.src),
1419	     le16_to_cpu(msg->hdr.type),
1420	     ceph_msg_type_name(le16_to_cpu(msg->hdr.type)),
1421	     le32_to_cpu(msg->hdr.front_len),
1422	     le32_to_cpu(msg->hdr.middle_len),
1423	     le32_to_cpu(msg->hdr.data_len),
1424	     con->in_front_crc, con->in_middle_crc, con->in_data_crc);
1425	con->ops->dispatch(con, msg);
1426
1427	mutex_lock(&con->mutex);
1428}
1429
1430/*
1431 * Atomically queue work on a connection after the specified delay.
1432 * Bump @con reference to avoid races with connection teardown.
1433 * Returns 0 if work was queued, or an error code otherwise.
1434 */
1435static int queue_con_delay(struct ceph_connection *con, unsigned long delay)
1436{
1437	if (!con->ops->get(con)) {
1438		dout("%s %p ref count 0\n", __func__, con);
1439		return -ENOENT;
1440	}
1441
1442	if (delay >= HZ)
1443		delay = round_jiffies_relative(delay);
1444
1445	dout("%s %p %lu\n", __func__, con, delay);
1446	if (!queue_delayed_work(ceph_msgr_wq, &con->work, delay)) {
1447		dout("%s %p - already queued\n", __func__, con);
1448		con->ops->put(con);
1449		return -EBUSY;
1450	}
1451
1452	return 0;
1453}
1454
1455static void queue_con(struct ceph_connection *con)
1456{
1457	(void) queue_con_delay(con, 0);
1458}
1459
1460static void cancel_con(struct ceph_connection *con)
1461{
1462	if (cancel_delayed_work(&con->work)) {
1463		dout("%s %p\n", __func__, con);
1464		con->ops->put(con);
1465	}
1466}
1467
1468static bool con_sock_closed(struct ceph_connection *con)
1469{
1470	if (!ceph_con_flag_test_and_clear(con, CEPH_CON_F_SOCK_CLOSED))
1471		return false;
1472
1473#define CASE(x)								\
1474	case CEPH_CON_S_ ## x:						\
1475		con->error_msg = "socket closed (con state " #x ")";	\
1476		break;
1477
1478	switch (con->state) {
1479	CASE(CLOSED);
1480	CASE(PREOPEN);
1481	CASE(V1_BANNER);
1482	CASE(V1_CONNECT_MSG);
1483	CASE(V2_BANNER_PREFIX);
1484	CASE(V2_BANNER_PAYLOAD);
1485	CASE(V2_HELLO);
1486	CASE(V2_AUTH);
1487	CASE(V2_AUTH_SIGNATURE);
1488	CASE(V2_SESSION_CONNECT);
1489	CASE(V2_SESSION_RECONNECT);
1490	CASE(OPEN);
1491	CASE(STANDBY);
1492	default:
1493		BUG();
1494	}
1495#undef CASE
1496
1497	return true;
1498}
1499
1500static bool con_backoff(struct ceph_connection *con)
1501{
1502	int ret;
1503
1504	if (!ceph_con_flag_test_and_clear(con, CEPH_CON_F_BACKOFF))
1505		return false;
1506
1507	ret = queue_con_delay(con, con->delay);
1508	if (ret) {
1509		dout("%s: con %p FAILED to back off %lu\n", __func__,
1510			con, con->delay);
1511		BUG_ON(ret == -ENOENT);
1512		ceph_con_flag_set(con, CEPH_CON_F_BACKOFF);
1513	}
1514
1515	return true;
1516}
1517
1518/* Finish fault handling; con->mutex must *not* be held here */
1519
1520static void con_fault_finish(struct ceph_connection *con)
1521{
1522	dout("%s %p\n", __func__, con);
1523
1524	/*
1525	 * in case we faulted due to authentication, invalidate our
1526	 * current tickets so that we can get new ones.
1527	 */
1528	if (!ceph_msgr2(from_msgr(con->msgr)) && con->v1.auth_retry) {
1529		dout("auth_retry %d, invalidating\n", con->v1.auth_retry);
1530		if (con->ops->invalidate_authorizer)
1531			con->ops->invalidate_authorizer(con);
1532		con->v1.auth_retry = 0;
1533	}
1534
1535	if (con->ops->fault)
1536		con->ops->fault(con);
1537}
1538
1539/*
1540 * Do some work on a connection.  Drop a connection ref when we're done.
1541 */
1542static void ceph_con_workfn(struct work_struct *work)
1543{
1544	struct ceph_connection *con = container_of(work, struct ceph_connection,
1545						   work.work);
1546	bool fault;
1547
1548	mutex_lock(&con->mutex);
1549	while (true) {
1550		int ret;
1551
1552		if ((fault = con_sock_closed(con))) {
1553			dout("%s: con %p SOCK_CLOSED\n", __func__, con);
1554			break;
1555		}
1556		if (con_backoff(con)) {
1557			dout("%s: con %p BACKOFF\n", __func__, con);
1558			break;
1559		}
1560		if (con->state == CEPH_CON_S_STANDBY) {
1561			dout("%s: con %p STANDBY\n", __func__, con);
1562			break;
1563		}
1564		if (con->state == CEPH_CON_S_CLOSED) {
1565			dout("%s: con %p CLOSED\n", __func__, con);
1566			BUG_ON(con->sock);
1567			break;
1568		}
1569		if (con->state == CEPH_CON_S_PREOPEN) {
1570			dout("%s: con %p PREOPEN\n", __func__, con);
1571			BUG_ON(con->sock);
1572		}
1573
1574		if (ceph_msgr2(from_msgr(con->msgr)))
1575			ret = ceph_con_v2_try_read(con);
1576		else
1577			ret = ceph_con_v1_try_read(con);
1578		if (ret < 0) {
1579			if (ret == -EAGAIN)
1580				continue;
1581			if (!con->error_msg)
1582				con->error_msg = "socket error on read";
1583			fault = true;
1584			break;
1585		}
1586
1587		if (ceph_msgr2(from_msgr(con->msgr)))
1588			ret = ceph_con_v2_try_write(con);
1589		else
1590			ret = ceph_con_v1_try_write(con);
1591		if (ret < 0) {
1592			if (ret == -EAGAIN)
1593				continue;
1594			if (!con->error_msg)
1595				con->error_msg = "socket error on write";
1596			fault = true;
1597		}
1598
1599		break;	/* If we make it to here, we're done */
1600	}
1601	if (fault)
1602		con_fault(con);
1603	mutex_unlock(&con->mutex);
1604
1605	if (fault)
1606		con_fault_finish(con);
1607
1608	con->ops->put(con);
1609}
1610
1611/*
1612 * Generic error/fault handler.  A retry mechanism is used with
1613 * exponential backoff
1614 */
1615static void con_fault(struct ceph_connection *con)
1616{
1617	dout("fault %p state %d to peer %s\n",
1618	     con, con->state, ceph_pr_addr(&con->peer_addr));
1619
1620	pr_warn("%s%lld %s %s\n", ENTITY_NAME(con->peer_name),
1621		ceph_pr_addr(&con->peer_addr), con->error_msg);
1622	con->error_msg = NULL;
1623
1624	WARN_ON(con->state == CEPH_CON_S_STANDBY ||
1625		con->state == CEPH_CON_S_CLOSED);
1626
1627	ceph_con_reset_protocol(con);
1628
1629	if (ceph_con_flag_test(con, CEPH_CON_F_LOSSYTX)) {
1630		dout("fault on LOSSYTX channel, marking CLOSED\n");
1631		con->state = CEPH_CON_S_CLOSED;
1632		return;
1633	}
1634
1635	/* Requeue anything that hasn't been acked */
1636	list_splice_init(&con->out_sent, &con->out_queue);
1637
1638	/* If there are no messages queued or keepalive pending, place
1639	 * the connection in a STANDBY state */
1640	if (list_empty(&con->out_queue) &&
1641	    !ceph_con_flag_test(con, CEPH_CON_F_KEEPALIVE_PENDING)) {
1642		dout("fault %p setting STANDBY clearing WRITE_PENDING\n", con);
1643		ceph_con_flag_clear(con, CEPH_CON_F_WRITE_PENDING);
1644		con->state = CEPH_CON_S_STANDBY;
1645	} else {
1646		/* retry after a delay. */
1647		con->state = CEPH_CON_S_PREOPEN;
1648		if (!con->delay) {
1649			con->delay = BASE_DELAY_INTERVAL;
1650		} else if (con->delay < MAX_DELAY_INTERVAL) {
1651			con->delay *= 2;
1652			if (con->delay > MAX_DELAY_INTERVAL)
1653				con->delay = MAX_DELAY_INTERVAL;
1654		}
1655		ceph_con_flag_set(con, CEPH_CON_F_BACKOFF);
1656		queue_con(con);
1657	}
1658}
1659
1660void ceph_messenger_reset_nonce(struct ceph_messenger *msgr)
1661{
1662	u32 nonce = le32_to_cpu(msgr->inst.addr.nonce) + 1000000;
1663	msgr->inst.addr.nonce = cpu_to_le32(nonce);
1664	ceph_encode_my_addr(msgr);
1665}
1666
1667/*
1668 * initialize a new messenger instance
1669 */
1670void ceph_messenger_init(struct ceph_messenger *msgr,
1671			 struct ceph_entity_addr *myaddr)
1672{
1673	spin_lock_init(&msgr->global_seq_lock);
1674
1675	if (myaddr) {
1676		memcpy(&msgr->inst.addr.in_addr, &myaddr->in_addr,
1677		       sizeof(msgr->inst.addr.in_addr));
1678		ceph_addr_set_port(&msgr->inst.addr, 0);
1679	}
1680
1681	/*
1682	 * Since nautilus, clients are identified using type ANY.
1683	 * For msgr1, ceph_encode_banner_addr() munges it to NONE.
1684	 */
1685	msgr->inst.addr.type = CEPH_ENTITY_ADDR_TYPE_ANY;
1686
1687	/* generate a random non-zero nonce */
1688	do {
1689		get_random_bytes(&msgr->inst.addr.nonce,
1690				 sizeof(msgr->inst.addr.nonce));
1691	} while (!msgr->inst.addr.nonce);
1692	ceph_encode_my_addr(msgr);
1693
1694	atomic_set(&msgr->stopping, 0);
1695	write_pnet(&msgr->net, get_net(current->nsproxy->net_ns));
1696
1697	dout("%s %p\n", __func__, msgr);
1698}
1699
1700void ceph_messenger_fini(struct ceph_messenger *msgr)
1701{
1702	put_net(read_pnet(&msgr->net));
1703}
1704
1705static void msg_con_set(struct ceph_msg *msg, struct ceph_connection *con)
1706{
1707	if (msg->con)
1708		msg->con->ops->put(msg->con);
1709
1710	msg->con = con ? con->ops->get(con) : NULL;
1711	BUG_ON(msg->con != con);
1712}
1713
1714static void clear_standby(struct ceph_connection *con)
1715{
1716	/* come back from STANDBY? */
1717	if (con->state == CEPH_CON_S_STANDBY) {
1718		dout("clear_standby %p\n", con);
1719		con->state = CEPH_CON_S_PREOPEN;
1720		if (!ceph_msgr2(from_msgr(con->msgr)))
1721			con->v1.connect_seq++;
1722		WARN_ON(ceph_con_flag_test(con, CEPH_CON_F_WRITE_PENDING));
1723		WARN_ON(ceph_con_flag_test(con, CEPH_CON_F_KEEPALIVE_PENDING));
1724	}
1725}
1726
1727/*
1728 * Queue up an outgoing message on the given connection.
1729 *
1730 * Consumes a ref on @msg.
1731 */
1732void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg)
1733{
1734	/* set src+dst */
1735	msg->hdr.src = con->msgr->inst.name;
1736	BUG_ON(msg->front.iov_len != le32_to_cpu(msg->hdr.front_len));
1737	msg->needs_out_seq = true;
1738
1739	mutex_lock(&con->mutex);
1740
1741	if (con->state == CEPH_CON_S_CLOSED) {
1742		dout("con_send %p closed, dropping %p\n", con, msg);
1743		ceph_msg_put(msg);
1744		mutex_unlock(&con->mutex);
1745		return;
1746	}
1747
1748	msg_con_set(msg, con);
1749
1750	BUG_ON(!list_empty(&msg->list_head));
1751	list_add_tail(&msg->list_head, &con->out_queue);
1752	dout("----- %p to %s%lld %d=%s len %d+%d+%d -----\n", msg,
1753	     ENTITY_NAME(con->peer_name), le16_to_cpu(msg->hdr.type),
1754	     ceph_msg_type_name(le16_to_cpu(msg->hdr.type)),
1755	     le32_to_cpu(msg->hdr.front_len),
1756	     le32_to_cpu(msg->hdr.middle_len),
1757	     le32_to_cpu(msg->hdr.data_len));
1758
1759	clear_standby(con);
1760	mutex_unlock(&con->mutex);
1761
1762	/* if there wasn't anything waiting to send before, queue
1763	 * new work */
1764	if (!ceph_con_flag_test_and_set(con, CEPH_CON_F_WRITE_PENDING))
1765		queue_con(con);
1766}
1767EXPORT_SYMBOL(ceph_con_send);
1768
1769/*
1770 * Revoke a message that was previously queued for send
1771 */
1772void ceph_msg_revoke(struct ceph_msg *msg)
1773{
1774	struct ceph_connection *con = msg->con;
1775
1776	if (!con) {
1777		dout("%s msg %p null con\n", __func__, msg);
1778		return;		/* Message not in our possession */
1779	}
1780
1781	mutex_lock(&con->mutex);
1782	if (list_empty(&msg->list_head)) {
1783		WARN_ON(con->out_msg == msg);
1784		dout("%s con %p msg %p not linked\n", __func__, con, msg);
1785		mutex_unlock(&con->mutex);
1786		return;
1787	}
1788
1789	dout("%s con %p msg %p was linked\n", __func__, con, msg);
1790	msg->hdr.seq = 0;
1791	ceph_msg_remove(msg);
1792
1793	if (con->out_msg == msg) {
1794		WARN_ON(con->state != CEPH_CON_S_OPEN);
1795		dout("%s con %p msg %p was sending\n", __func__, con, msg);
1796		if (ceph_msgr2(from_msgr(con->msgr)))
1797			ceph_con_v2_revoke(con, msg);
1798		else
1799			ceph_con_v1_revoke(con, msg);
1800		ceph_msg_put(con->out_msg);
1801		con->out_msg = NULL;
1802	} else {
1803		dout("%s con %p msg %p not current, out_msg %p\n", __func__,
1804		     con, msg, con->out_msg);
1805	}
1806	mutex_unlock(&con->mutex);
1807}
1808
1809/*
1810 * Revoke a message that we may be reading data into
1811 */
1812void ceph_msg_revoke_incoming(struct ceph_msg *msg)
1813{
1814	struct ceph_connection *con = msg->con;
1815
1816	if (!con) {
1817		dout("%s msg %p null con\n", __func__, msg);
1818		return;		/* Message not in our possession */
1819	}
1820
1821	mutex_lock(&con->mutex);
1822	if (con->in_msg == msg) {
1823		WARN_ON(con->state != CEPH_CON_S_OPEN);
1824		dout("%s con %p msg %p was recving\n", __func__, con, msg);
1825		if (ceph_msgr2(from_msgr(con->msgr)))
1826			ceph_con_v2_revoke_incoming(con);
1827		else
1828			ceph_con_v1_revoke_incoming(con);
1829		ceph_msg_put(con->in_msg);
1830		con->in_msg = NULL;
1831	} else {
1832		dout("%s con %p msg %p not current, in_msg %p\n", __func__,
1833		     con, msg, con->in_msg);
1834	}
1835	mutex_unlock(&con->mutex);
1836}
1837
1838/*
1839 * Queue a keepalive byte to ensure the tcp connection is alive.
1840 */
1841void ceph_con_keepalive(struct ceph_connection *con)
1842{
1843	dout("con_keepalive %p\n", con);
1844	mutex_lock(&con->mutex);
1845	clear_standby(con);
1846	ceph_con_flag_set(con, CEPH_CON_F_KEEPALIVE_PENDING);
1847	mutex_unlock(&con->mutex);
1848
1849	if (!ceph_con_flag_test_and_set(con, CEPH_CON_F_WRITE_PENDING))
1850		queue_con(con);
1851}
1852EXPORT_SYMBOL(ceph_con_keepalive);
1853
1854bool ceph_con_keepalive_expired(struct ceph_connection *con,
1855			       unsigned long interval)
1856{
1857	if (interval > 0 &&
1858	    (con->peer_features & CEPH_FEATURE_MSGR_KEEPALIVE2)) {
1859		struct timespec64 now;
1860		struct timespec64 ts;
1861		ktime_get_real_ts64(&now);
1862		jiffies_to_timespec64(interval, &ts);
1863		ts = timespec64_add(con->last_keepalive_ack, ts);
1864		return timespec64_compare(&now, &ts) >= 0;
1865	}
1866	return false;
1867}
1868
1869static struct ceph_msg_data *ceph_msg_data_add(struct ceph_msg *msg)
1870{
1871	BUG_ON(msg->num_data_items >= msg->max_data_items);
1872	return &msg->data[msg->num_data_items++];
1873}
1874
1875static void ceph_msg_data_destroy(struct ceph_msg_data *data)
1876{
1877	if (data->type == CEPH_MSG_DATA_PAGES && data->own_pages) {
1878		int num_pages = calc_pages_for(data->alignment, data->length);
1879		ceph_release_page_vector(data->pages, num_pages);
1880	} else if (data->type == CEPH_MSG_DATA_PAGELIST) {
1881		ceph_pagelist_release(data->pagelist);
1882	}
1883}
1884
1885void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
1886			     size_t length, size_t alignment, bool own_pages)
1887{
1888	struct ceph_msg_data *data;
1889
1890	BUG_ON(!pages);
1891	BUG_ON(!length);
1892
1893	data = ceph_msg_data_add(msg);
1894	data->type = CEPH_MSG_DATA_PAGES;
1895	data->pages = pages;
1896	data->length = length;
1897	data->alignment = alignment & ~PAGE_MASK;
1898	data->own_pages = own_pages;
1899
1900	msg->data_length += length;
1901}
1902EXPORT_SYMBOL(ceph_msg_data_add_pages);
1903
1904void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
1905				struct ceph_pagelist *pagelist)
1906{
1907	struct ceph_msg_data *data;
1908
1909	BUG_ON(!pagelist);
1910	BUG_ON(!pagelist->length);
1911
1912	data = ceph_msg_data_add(msg);
1913	data->type = CEPH_MSG_DATA_PAGELIST;
1914	refcount_inc(&pagelist->refcnt);
1915	data->pagelist = pagelist;
1916
1917	msg->data_length += pagelist->length;
1918}
1919EXPORT_SYMBOL(ceph_msg_data_add_pagelist);
1920
1921#ifdef	CONFIG_BLOCK
1922void ceph_msg_data_add_bio(struct ceph_msg *msg, struct ceph_bio_iter *bio_pos,
1923			   u32 length)
1924{
1925	struct ceph_msg_data *data;
1926
1927	data = ceph_msg_data_add(msg);
1928	data->type = CEPH_MSG_DATA_BIO;
1929	data->bio_pos = *bio_pos;
1930	data->bio_length = length;
1931
1932	msg->data_length += length;
1933}
1934EXPORT_SYMBOL(ceph_msg_data_add_bio);
1935#endif	/* CONFIG_BLOCK */
1936
1937void ceph_msg_data_add_bvecs(struct ceph_msg *msg,
1938			     struct ceph_bvec_iter *bvec_pos)
1939{
1940	struct ceph_msg_data *data;
1941
1942	data = ceph_msg_data_add(msg);
1943	data->type = CEPH_MSG_DATA_BVECS;
1944	data->bvec_pos = *bvec_pos;
1945
1946	msg->data_length += bvec_pos->iter.bi_size;
1947}
1948EXPORT_SYMBOL(ceph_msg_data_add_bvecs);
1949
1950void ceph_msg_data_add_iter(struct ceph_msg *msg,
1951			    struct iov_iter *iter)
1952{
1953	struct ceph_msg_data *data;
1954
1955	data = ceph_msg_data_add(msg);
1956	data->type = CEPH_MSG_DATA_ITER;
1957	data->iter = *iter;
1958
1959	msg->data_length += iov_iter_count(&data->iter);
1960}
1961
1962/*
1963 * construct a new message with given type, size
1964 * the new msg has a ref count of 1.
1965 */
1966struct ceph_msg *ceph_msg_new2(int type, int front_len, int max_data_items,
1967			       gfp_t flags, bool can_fail)
1968{
1969	struct ceph_msg *m;
1970
1971	m = kmem_cache_zalloc(ceph_msg_cache, flags);
1972	if (m == NULL)
1973		goto out;
1974
1975	m->hdr.type = cpu_to_le16(type);
1976	m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);
1977	m->hdr.front_len = cpu_to_le32(front_len);
1978
1979	INIT_LIST_HEAD(&m->list_head);
1980	kref_init(&m->kref);
1981
1982	/* front */
1983	if (front_len) {
1984		m->front.iov_base = kvmalloc(front_len, flags);
1985		if (m->front.iov_base == NULL) {
1986			dout("ceph_msg_new can't allocate %d bytes\n",
1987			     front_len);
1988			goto out2;
1989		}
1990	} else {
1991		m->front.iov_base = NULL;
1992	}
1993	m->front_alloc_len = m->front.iov_len = front_len;
1994
1995	if (max_data_items) {
1996		m->data = kmalloc_array(max_data_items, sizeof(*m->data),
1997					flags);
1998		if (!m->data)
1999			goto out2;
2000
2001		m->max_data_items = max_data_items;
2002	}
2003
2004	dout("ceph_msg_new %p front %d\n", m, front_len);
2005	return m;
2006
2007out2:
2008	ceph_msg_put(m);
2009out:
2010	if (!can_fail) {
2011		pr_err("msg_new can't create type %d front %d\n", type,
2012		       front_len);
2013		WARN_ON(1);
2014	} else {
2015		dout("msg_new can't create type %d front %d\n", type,
2016		     front_len);
2017	}
2018	return NULL;
2019}
2020EXPORT_SYMBOL(ceph_msg_new2);
2021
2022struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
2023			      bool can_fail)
2024{
2025	return ceph_msg_new2(type, front_len, 0, flags, can_fail);
2026}
2027EXPORT_SYMBOL(ceph_msg_new);
2028
2029/*
2030 * Allocate "middle" portion of a message, if it is needed and wasn't
2031 * allocated by alloc_msg.  This allows us to read a small fixed-size
2032 * per-type header in the front and then gracefully fail (i.e.,
2033 * propagate the error to the caller based on info in the front) when
2034 * the middle is too large.
2035 */
2036static int ceph_alloc_middle(struct ceph_connection *con, struct ceph_msg *msg)
2037{
2038	int type = le16_to_cpu(msg->hdr.type);
2039	int middle_len = le32_to_cpu(msg->hdr.middle_len);
2040
2041	dout("alloc_middle %p type %d %s middle_len %d\n", msg, type,
2042	     ceph_msg_type_name(type), middle_len);
2043	BUG_ON(!middle_len);
2044	BUG_ON(msg->middle);
2045
2046	msg->middle = ceph_buffer_new(middle_len, GFP_NOFS);
2047	if (!msg->middle)
2048		return -ENOMEM;
2049	return 0;
2050}
2051
2052/*
2053 * Allocate a message for receiving an incoming message on a
2054 * connection, and save the result in con->in_msg.  Uses the
2055 * connection's private alloc_msg op if available.
2056 *
2057 * Returns 0 on success, or a negative error code.
2058 *
2059 * On success, if we set *skip = 1:
2060 *  - the next message should be skipped and ignored.
2061 *  - con->in_msg == NULL
2062 * or if we set *skip = 0:
2063 *  - con->in_msg is non-null.
2064 * On error (ENOMEM, EAGAIN, ...),
2065 *  - con->in_msg == NULL
2066 */
2067int ceph_con_in_msg_alloc(struct ceph_connection *con,
2068			  struct ceph_msg_header *hdr, int *skip)
2069{
2070	int middle_len = le32_to_cpu(hdr->middle_len);
2071	struct ceph_msg *msg;
2072	int ret = 0;
2073
2074	BUG_ON(con->in_msg != NULL);
2075	BUG_ON(!con->ops->alloc_msg);
2076
2077	mutex_unlock(&con->mutex);
2078	msg = con->ops->alloc_msg(con, hdr, skip);
2079	mutex_lock(&con->mutex);
2080	if (con->state != CEPH_CON_S_OPEN) {
2081		if (msg)
2082			ceph_msg_put(msg);
2083		return -EAGAIN;
2084	}
2085	if (msg) {
2086		BUG_ON(*skip);
2087		msg_con_set(msg, con);
2088		con->in_msg = msg;
2089	} else {
2090		/*
2091		 * Null message pointer means either we should skip
2092		 * this message or we couldn't allocate memory.  The
2093		 * former is not an error.
2094		 */
2095		if (*skip)
2096			return 0;
2097
2098		con->error_msg = "error allocating memory for incoming message";
2099		return -ENOMEM;
2100	}
2101	memcpy(&con->in_msg->hdr, hdr, sizeof(*hdr));
2102
2103	if (middle_len && !con->in_msg->middle) {
2104		ret = ceph_alloc_middle(con, con->in_msg);
2105		if (ret < 0) {
2106			ceph_msg_put(con->in_msg);
2107			con->in_msg = NULL;
2108		}
2109	}
2110
2111	return ret;
2112}
2113
2114struct ceph_msg *ceph_con_get_out_msg(struct ceph_connection *con)
2115{
2116	struct ceph_msg *msg;
2117
2118	if (list_empty(&con->out_queue))
2119		return NULL;
2120
2121	msg = list_first_entry(&con->out_queue, struct ceph_msg, list_head);
2122	WARN_ON(msg->con != con);
2123
2124	/*
2125	 * Put the message on "sent" list using a ref from ceph_con_send().
2126	 * It is put when the message is acked or revoked.
2127	 */
2128	list_move_tail(&msg->list_head, &con->out_sent);
2129
2130	/*
2131	 * Only assign outgoing seq # if we haven't sent this message
2132	 * yet.  If it is requeued, resend with it's original seq.
2133	 */
2134	if (msg->needs_out_seq) {
2135		msg->hdr.seq = cpu_to_le64(++con->out_seq);
2136		msg->needs_out_seq = false;
2137
2138		if (con->ops->reencode_message)
2139			con->ops->reencode_message(msg);
2140	}
2141
2142	/*
2143	 * Get a ref for out_msg.  It is put when we are done sending the
2144	 * message or in case of a fault.
2145	 */
2146	WARN_ON(con->out_msg);
2147	return con->out_msg = ceph_msg_get(msg);
2148}
2149
2150/*
2151 * Free a generically kmalloc'd message.
2152 */
2153static void ceph_msg_free(struct ceph_msg *m)
2154{
2155	dout("%s %p\n", __func__, m);
2156	kvfree(m->front.iov_base);
2157	kfree(m->data);
2158	kmem_cache_free(ceph_msg_cache, m);
2159}
2160
2161static void ceph_msg_release(struct kref *kref)
2162{
2163	struct ceph_msg *m = container_of(kref, struct ceph_msg, kref);
2164	int i;
2165
2166	dout("%s %p\n", __func__, m);
2167	WARN_ON(!list_empty(&m->list_head));
2168
2169	msg_con_set(m, NULL);
2170
2171	/* drop middle, data, if any */
2172	if (m->middle) {
2173		ceph_buffer_put(m->middle);
2174		m->middle = NULL;
2175	}
2176
2177	for (i = 0; i < m->num_data_items; i++)
2178		ceph_msg_data_destroy(&m->data[i]);
2179
2180	if (m->pool)
2181		ceph_msgpool_put(m->pool, m);
2182	else
2183		ceph_msg_free(m);
2184}
2185
2186struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
2187{
2188	dout("%s %p (was %d)\n", __func__, msg,
2189	     kref_read(&msg->kref));
2190	kref_get(&msg->kref);
2191	return msg;
2192}
2193EXPORT_SYMBOL(ceph_msg_get);
2194
2195void ceph_msg_put(struct ceph_msg *msg)
2196{
2197	dout("%s %p (was %d)\n", __func__, msg,
2198	     kref_read(&msg->kref));
2199	kref_put(&msg->kref, ceph_msg_release);
2200}
2201EXPORT_SYMBOL(ceph_msg_put);
2202
2203void ceph_msg_dump(struct ceph_msg *msg)
2204{
2205	pr_debug("msg_dump %p (front_alloc_len %d length %zd)\n", msg,
2206		 msg->front_alloc_len, msg->data_length);
2207	print_hex_dump(KERN_DEBUG, "header: ",
2208		       DUMP_PREFIX_OFFSET, 16, 1,
2209		       &msg->hdr, sizeof(msg->hdr), true);
2210	print_hex_dump(KERN_DEBUG, " front: ",
2211		       DUMP_PREFIX_OFFSET, 16, 1,
2212		       msg->front.iov_base, msg->front.iov_len, true);
2213	if (msg->middle)
2214		print_hex_dump(KERN_DEBUG, "middle: ",
2215			       DUMP_PREFIX_OFFSET, 16, 1,
2216			       msg->middle->vec.iov_base,
2217			       msg->middle->vec.iov_len, true);
2218	print_hex_dump(KERN_DEBUG, "footer: ",
2219		       DUMP_PREFIX_OFFSET, 16, 1,
2220		       &msg->footer, sizeof(msg->footer), true);
2221}
2222EXPORT_SYMBOL(ceph_msg_dump);