xref: /linux/net/ceph/messenger.c (revision 07fdad3a93756b872da7b53647715c48d0f4a2d0)
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 
86 static 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 
100 void 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 
107 void 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 
114 bool 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 
121 bool 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 
129 bool 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 
139 static struct kmem_cache	*ceph_msg_cache;
140 
141 #ifdef CONFIG_LOCKDEP
142 static struct lock_class_key socket_class;
143 #endif
144 
145 static void queue_con(struct ceph_connection *con);
146 static void cancel_con(struct ceph_connection *con);
147 static void ceph_con_workfn(struct work_struct *);
148 static 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 
159 static char addr_str[ADDR_STR_COUNT][MAX_ADDR_STR_LEN];
160 static atomic_t addr_str_seq = ATOMIC_INIT(0);
161 
162 struct page *ceph_zero_page;		/* used in certain error cases */
163 
164 const 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 }
195 EXPORT_SYMBOL(ceph_pr_addr);
196 
197 void 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  */
209 static struct workqueue_struct *ceph_msgr_wq;
210 
211 static 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 
221 static 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 
228 static 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 
242 int __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 
266 void ceph_msgr_exit(void)
267 {
268 	BUG_ON(ceph_msgr_wq == NULL);
269 
270 	_ceph_msgr_exit();
271 }
272 
273 void ceph_msgr_flush(void)
274 {
275 	flush_workqueue(ceph_msgr_wq);
276 }
277 EXPORT_SYMBOL(ceph_msgr_flush);
278 
279 /* Connection socket state transition functions */
280 
281 static 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 
292 static 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 
303 static 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 
314 static 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 
327 static 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 */
346 static 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 */
364 static 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 */
387 static 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  */
417 static 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  */
435 int 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 *)&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  */
486 int 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 
509 static 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  */
539 static void ceph_msg_remove(struct ceph_msg *msg)
540 {
541 	list_del_init(&msg->list_head);
542 
543 	ceph_msg_put(msg);
544 }
545 
546 static 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 
555 void 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  */
576 void 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 }
593 EXPORT_SYMBOL(ceph_con_close);
594 
595 /*
596  * Reopen a closed connection, with a new peer address.
597  */
598 void 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 }
616 EXPORT_SYMBOL(ceph_con_open);
617 
618 /*
619  * return true if this connection ever successfully opened
620  */
621 bool 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  */
632 void 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 }
651 EXPORT_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  */
657 u32 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  */
672 void 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  */
697 void 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  */
725 static 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 
739 static 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 
751 static 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 
781 static 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 
794 static 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 
806 static 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  */
832 static 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 
852 static struct page *
853 ceph_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 
868 static 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  */
896 static void
897 ceph_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 
920 static struct page *
921 ceph_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 
941 static 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 
973 static 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 
984 static 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 
1013 static 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  */
1037 static 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 
1067 void 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  */
1086 struct 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  */
1127 void 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 
1165 u32 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 
1178 bool 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 }
1193 EXPORT_SYMBOL(ceph_addr_is_blank);
1194 
1195 int 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 
1206 void 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  */
1221 static 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
1243 static 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
1288 static 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  */
1299 static 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  */
1315 int 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 
1393 bad:
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  */
1402 void 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  */
1435 static 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 
1455 static void queue_con(struct ceph_connection *con)
1456 {
1457 	(void) queue_con_delay(con, 0);
1458 }
1459 
1460 static 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 
1468 static 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 
1500 static 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 
1520 static 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  */
1542 static 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  */
1615 static 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 
1660 void 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  */
1670 void 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 
1700 void ceph_messenger_fini(struct ceph_messenger *msgr)
1701 {
1702 	put_net(read_pnet(&msgr->net));
1703 }
1704 
1705 static 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 
1714 static 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  */
1732 void 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 }
1767 EXPORT_SYMBOL(ceph_con_send);
1768 
1769 /*
1770  * Revoke a message that was previously queued for send
1771  */
1772 void 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);
1798 		else
1799 			ceph_con_v1_revoke(con);
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  */
1812 void 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  */
1841 void 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 }
1852 EXPORT_SYMBOL(ceph_con_keepalive);
1853 
1854 bool 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 
1869 static 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 
1875 static 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 
1885 void 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 }
1902 EXPORT_SYMBOL(ceph_msg_data_add_pages);
1903 
1904 void 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 }
1919 EXPORT_SYMBOL(ceph_msg_data_add_pagelist);
1920 
1921 #ifdef	CONFIG_BLOCK
1922 void 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 }
1934 EXPORT_SYMBOL(ceph_msg_data_add_bio);
1935 #endif	/* CONFIG_BLOCK */
1936 
1937 void 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 }
1948 EXPORT_SYMBOL(ceph_msg_data_add_bvecs);
1949 
1950 void 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  */
1966 struct 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 
2007 out2:
2008 	ceph_msg_put(m);
2009 out:
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 }
2020 EXPORT_SYMBOL(ceph_msg_new2);
2021 
2022 struct 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 }
2027 EXPORT_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  */
2036 static 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  */
2067 int 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 
2114 void ceph_con_get_out_msg(struct ceph_connection *con)
2115 {
2116 	struct ceph_msg *msg;
2117 
2118 	BUG_ON(list_empty(&con->out_queue));
2119 	msg = list_first_entry(&con->out_queue, struct ceph_msg, list_head);
2120 	WARN_ON(msg->con != con);
2121 
2122 	/*
2123 	 * Put the message on "sent" list using a ref from ceph_con_send().
2124 	 * It is put when the message is acked or revoked.
2125 	 */
2126 	list_move_tail(&msg->list_head, &con->out_sent);
2127 
2128 	/*
2129 	 * Only assign outgoing seq # if we haven't sent this message
2130 	 * yet.  If it is requeued, resend with it's original seq.
2131 	 */
2132 	if (msg->needs_out_seq) {
2133 		msg->hdr.seq = cpu_to_le64(++con->out_seq);
2134 		msg->needs_out_seq = false;
2135 
2136 		if (con->ops->reencode_message)
2137 			con->ops->reencode_message(msg);
2138 	}
2139 
2140 	/*
2141 	 * Get a ref for out_msg.  It is put when we are done sending the
2142 	 * message or in case of a fault.
2143 	 */
2144 	WARN_ON(con->out_msg);
2145 	con->out_msg = ceph_msg_get(msg);
2146 }
2147 
2148 /*
2149  * Free a generically kmalloc'd message.
2150  */
2151 static void ceph_msg_free(struct ceph_msg *m)
2152 {
2153 	dout("%s %p\n", __func__, m);
2154 	kvfree(m->front.iov_base);
2155 	kfree(m->data);
2156 	kmem_cache_free(ceph_msg_cache, m);
2157 }
2158 
2159 static void ceph_msg_release(struct kref *kref)
2160 {
2161 	struct ceph_msg *m = container_of(kref, struct ceph_msg, kref);
2162 	int i;
2163 
2164 	dout("%s %p\n", __func__, m);
2165 	WARN_ON(!list_empty(&m->list_head));
2166 
2167 	msg_con_set(m, NULL);
2168 
2169 	/* drop middle, data, if any */
2170 	if (m->middle) {
2171 		ceph_buffer_put(m->middle);
2172 		m->middle = NULL;
2173 	}
2174 
2175 	for (i = 0; i < m->num_data_items; i++)
2176 		ceph_msg_data_destroy(&m->data[i]);
2177 
2178 	if (m->pool)
2179 		ceph_msgpool_put(m->pool, m);
2180 	else
2181 		ceph_msg_free(m);
2182 }
2183 
2184 struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
2185 {
2186 	dout("%s %p (was %d)\n", __func__, msg,
2187 	     kref_read(&msg->kref));
2188 	kref_get(&msg->kref);
2189 	return msg;
2190 }
2191 EXPORT_SYMBOL(ceph_msg_get);
2192 
2193 void ceph_msg_put(struct ceph_msg *msg)
2194 {
2195 	dout("%s %p (was %d)\n", __func__, msg,
2196 	     kref_read(&msg->kref));
2197 	kref_put(&msg->kref, ceph_msg_release);
2198 }
2199 EXPORT_SYMBOL(ceph_msg_put);
2200 
2201 void ceph_msg_dump(struct ceph_msg *msg)
2202 {
2203 	pr_debug("msg_dump %p (front_alloc_len %d length %zd)\n", msg,
2204 		 msg->front_alloc_len, msg->data_length);
2205 	print_hex_dump(KERN_DEBUG, "header: ",
2206 		       DUMP_PREFIX_OFFSET, 16, 1,
2207 		       &msg->hdr, sizeof(msg->hdr), true);
2208 	print_hex_dump(KERN_DEBUG, " front: ",
2209 		       DUMP_PREFIX_OFFSET, 16, 1,
2210 		       msg->front.iov_base, msg->front.iov_len, true);
2211 	if (msg->middle)
2212 		print_hex_dump(KERN_DEBUG, "middle: ",
2213 			       DUMP_PREFIX_OFFSET, 16, 1,
2214 			       msg->middle->vec.iov_base,
2215 			       msg->middle->vec.iov_len, true);
2216 	print_hex_dump(KERN_DEBUG, "footer: ",
2217 		       DUMP_PREFIX_OFFSET, 16, 1,
2218 		       &msg->footer, sizeof(msg->footer), true);
2219 }
2220 EXPORT_SYMBOL(ceph_msg_dump);
2221