1 #include <linux/ceph/ceph_debug.h>
2 
3 #include <linux/crc32c.h>
4 #include <linux/ctype.h>
5 #include <linux/highmem.h>
6 #include <linux/inet.h>
7 #include <linux/kthread.h>
8 #include <linux/net.h>
9 #include <linux/slab.h>
10 #include <linux/socket.h>
11 #include <linux/string.h>
12 #ifdef	CONFIG_BLOCK
13 #include <linux/bio.h>
14 #endif	/* CONFIG_BLOCK */
15 #include <linux/dns_resolver.h>
16 #include <net/tcp.h>
17 
18 #include <linux/ceph/ceph_features.h>
19 #include <linux/ceph/libceph.h>
20 #include <linux/ceph/messenger.h>
21 #include <linux/ceph/decode.h>
22 #include <linux/ceph/pagelist.h>
23 #include <linux/export.h>
24 
25 #define list_entry_next(pos, member)					\
26 	list_entry(pos->member.next, typeof(*pos), member)
27 
28 /*
29  * Ceph uses the messenger to exchange ceph_msg messages with other
30  * hosts in the system.  The messenger provides ordered and reliable
31  * delivery.  We tolerate TCP disconnects by reconnecting (with
32  * exponential backoff) in the case of a fault (disconnection, bad
33  * crc, protocol error).  Acks allow sent messages to be discarded by
34  * the sender.
35  */
36 
37 /*
38  * We track the state of the socket on a given connection using
39  * values defined below.  The transition to a new socket state is
40  * handled by a function which verifies we aren't coming from an
41  * unexpected state.
42  *
43  *      --------
44  *      | NEW* |  transient initial state
45  *      --------
46  *          | con_sock_state_init()
47  *          v
48  *      ----------
49  *      | CLOSED |  initialized, but no socket (and no
50  *      ----------  TCP connection)
51  *       ^      \
52  *       |       \ con_sock_state_connecting()
53  *       |        ----------------------
54  *       |                              \
55  *       + con_sock_state_closed()       \
56  *       |+---------------------------    \
57  *       | \                          \    \
58  *       |  -----------                \    \
59  *       |  | CLOSING |  socket event;  \    \
60  *       |  -----------  await close     \    \
61  *       |       ^                        \   |
62  *       |       |                         \  |
63  *       |       + con_sock_state_closing() \ |
64  *       |      / \                         | |
65  *       |     /   ---------------          | |
66  *       |    /                   \         v v
67  *       |   /                    --------------
68  *       |  /    -----------------| CONNECTING |  socket created, TCP
69  *       |  |   /                 --------------  connect initiated
70  *       |  |   | con_sock_state_connected()
71  *       |  |   v
72  *      -------------
73  *      | CONNECTED |  TCP connection established
74  *      -------------
75  *
76  * State values for ceph_connection->sock_state; NEW is assumed to be 0.
77  */
78 
79 #define CON_SOCK_STATE_NEW		0	/* -> CLOSED */
80 #define CON_SOCK_STATE_CLOSED		1	/* -> CONNECTING */
81 #define CON_SOCK_STATE_CONNECTING	2	/* -> CONNECTED or -> CLOSING */
82 #define CON_SOCK_STATE_CONNECTED	3	/* -> CLOSING or -> CLOSED */
83 #define CON_SOCK_STATE_CLOSING		4	/* -> CLOSED */
84 
85 /*
86  * connection states
87  */
88 #define CON_STATE_CLOSED        1  /* -> PREOPEN */
89 #define CON_STATE_PREOPEN       2  /* -> CONNECTING, CLOSED */
90 #define CON_STATE_CONNECTING    3  /* -> NEGOTIATING, CLOSED */
91 #define CON_STATE_NEGOTIATING   4  /* -> OPEN, CLOSED */
92 #define CON_STATE_OPEN          5  /* -> STANDBY, CLOSED */
93 #define CON_STATE_STANDBY       6  /* -> PREOPEN, CLOSED */
94 
95 /*
96  * ceph_connection flag bits
97  */
98 #define CON_FLAG_LOSSYTX           0  /* we can close channel or drop
99 				       * messages on errors */
100 #define CON_FLAG_KEEPALIVE_PENDING 1  /* we need to send a keepalive */
101 #define CON_FLAG_WRITE_PENDING	   2  /* we have data ready to send */
102 #define CON_FLAG_SOCK_CLOSED	   3  /* socket state changed to closed */
103 #define CON_FLAG_BACKOFF           4  /* need to retry queuing delayed work */
104 
con_flag_valid(unsigned long con_flag)105 static bool con_flag_valid(unsigned long con_flag)
106 {
107 	switch (con_flag) {
108 	case CON_FLAG_LOSSYTX:
109 	case CON_FLAG_KEEPALIVE_PENDING:
110 	case CON_FLAG_WRITE_PENDING:
111 	case CON_FLAG_SOCK_CLOSED:
112 	case CON_FLAG_BACKOFF:
113 		return true;
114 	default:
115 		return false;
116 	}
117 }
118 
con_flag_clear(struct ceph_connection * con,unsigned long con_flag)119 static void con_flag_clear(struct ceph_connection *con, unsigned long con_flag)
120 {
121 	BUG_ON(!con_flag_valid(con_flag));
122 
123 	clear_bit(con_flag, &con->flags);
124 }
125 
con_flag_set(struct ceph_connection * con,unsigned long con_flag)126 static void con_flag_set(struct ceph_connection *con, unsigned long con_flag)
127 {
128 	BUG_ON(!con_flag_valid(con_flag));
129 
130 	set_bit(con_flag, &con->flags);
131 }
132 
con_flag_test(struct ceph_connection * con,unsigned long con_flag)133 static bool con_flag_test(struct ceph_connection *con, unsigned long con_flag)
134 {
135 	BUG_ON(!con_flag_valid(con_flag));
136 
137 	return test_bit(con_flag, &con->flags);
138 }
139 
con_flag_test_and_clear(struct ceph_connection * con,unsigned long con_flag)140 static bool con_flag_test_and_clear(struct ceph_connection *con,
141 					unsigned long con_flag)
142 {
143 	BUG_ON(!con_flag_valid(con_flag));
144 
145 	return test_and_clear_bit(con_flag, &con->flags);
146 }
147 
con_flag_test_and_set(struct ceph_connection * con,unsigned long con_flag)148 static bool con_flag_test_and_set(struct ceph_connection *con,
149 					unsigned long con_flag)
150 {
151 	BUG_ON(!con_flag_valid(con_flag));
152 
153 	return test_and_set_bit(con_flag, &con->flags);
154 }
155 
156 /* Slab caches for frequently-allocated structures */
157 
158 static struct kmem_cache	*ceph_msg_cache;
159 static struct kmem_cache	*ceph_msg_data_cache;
160 
161 /* static tag bytes (protocol control messages) */
162 static char tag_msg = CEPH_MSGR_TAG_MSG;
163 static char tag_ack = CEPH_MSGR_TAG_ACK;
164 static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE;
165 
166 #ifdef CONFIG_LOCKDEP
167 static struct lock_class_key socket_class;
168 #endif
169 
170 /*
171  * When skipping (ignoring) a block of input we read it into a "skip
172  * buffer," which is this many bytes in size.
173  */
174 #define SKIP_BUF_SIZE	1024
175 
176 static void queue_con(struct ceph_connection *con);
177 static void cancel_con(struct ceph_connection *con);
178 static void con_work(struct work_struct *);
179 static void con_fault(struct ceph_connection *con);
180 
181 /*
182  * Nicely render a sockaddr as a string.  An array of formatted
183  * strings is used, to approximate reentrancy.
184  */
185 #define ADDR_STR_COUNT_LOG	5	/* log2(# address strings in array) */
186 #define ADDR_STR_COUNT		(1 << ADDR_STR_COUNT_LOG)
187 #define ADDR_STR_COUNT_MASK	(ADDR_STR_COUNT - 1)
188 #define MAX_ADDR_STR_LEN	64	/* 54 is enough */
189 
190 static char addr_str[ADDR_STR_COUNT][MAX_ADDR_STR_LEN];
191 static atomic_t addr_str_seq = ATOMIC_INIT(0);
192 
193 static struct page *zero_page;		/* used in certain error cases */
194 
ceph_pr_addr(const struct sockaddr_storage * ss)195 const char *ceph_pr_addr(const struct sockaddr_storage *ss)
196 {
197 	int i;
198 	char *s;
199 	struct sockaddr_in *in4 = (struct sockaddr_in *) ss;
200 	struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss;
201 
202 	i = atomic_inc_return(&addr_str_seq) & ADDR_STR_COUNT_MASK;
203 	s = addr_str[i];
204 
205 	switch (ss->ss_family) {
206 	case AF_INET:
207 		snprintf(s, MAX_ADDR_STR_LEN, "%pI4:%hu", &in4->sin_addr,
208 			 ntohs(in4->sin_port));
209 		break;
210 
211 	case AF_INET6:
212 		snprintf(s, MAX_ADDR_STR_LEN, "[%pI6c]:%hu", &in6->sin6_addr,
213 			 ntohs(in6->sin6_port));
214 		break;
215 
216 	default:
217 		snprintf(s, MAX_ADDR_STR_LEN, "(unknown sockaddr family %hu)",
218 			 ss->ss_family);
219 	}
220 
221 	return s;
222 }
223 EXPORT_SYMBOL(ceph_pr_addr);
224 
encode_my_addr(struct ceph_messenger * msgr)225 static void encode_my_addr(struct ceph_messenger *msgr)
226 {
227 	memcpy(&msgr->my_enc_addr, &msgr->inst.addr, sizeof(msgr->my_enc_addr));
228 	ceph_encode_addr(&msgr->my_enc_addr);
229 }
230 
231 /*
232  * work queue for all reading and writing to/from the socket.
233  */
234 static struct workqueue_struct *ceph_msgr_wq;
235 
ceph_msgr_slab_init(void)236 static int ceph_msgr_slab_init(void)
237 {
238 	BUG_ON(ceph_msg_cache);
239 	ceph_msg_cache = kmem_cache_create("ceph_msg",
240 					sizeof (struct ceph_msg),
241 					__alignof__(struct ceph_msg), 0, NULL);
242 
243 	if (!ceph_msg_cache)
244 		return -ENOMEM;
245 
246 	BUG_ON(ceph_msg_data_cache);
247 	ceph_msg_data_cache = kmem_cache_create("ceph_msg_data",
248 					sizeof (struct ceph_msg_data),
249 					__alignof__(struct ceph_msg_data),
250 					0, NULL);
251 	if (ceph_msg_data_cache)
252 		return 0;
253 
254 	kmem_cache_destroy(ceph_msg_cache);
255 	ceph_msg_cache = NULL;
256 
257 	return -ENOMEM;
258 }
259 
ceph_msgr_slab_exit(void)260 static void ceph_msgr_slab_exit(void)
261 {
262 	BUG_ON(!ceph_msg_data_cache);
263 	kmem_cache_destroy(ceph_msg_data_cache);
264 	ceph_msg_data_cache = NULL;
265 
266 	BUG_ON(!ceph_msg_cache);
267 	kmem_cache_destroy(ceph_msg_cache);
268 	ceph_msg_cache = NULL;
269 }
270 
_ceph_msgr_exit(void)271 static void _ceph_msgr_exit(void)
272 {
273 	if (ceph_msgr_wq) {
274 		destroy_workqueue(ceph_msgr_wq);
275 		ceph_msgr_wq = NULL;
276 	}
277 
278 	ceph_msgr_slab_exit();
279 
280 	BUG_ON(zero_page == NULL);
281 	kunmap(zero_page);
282 	page_cache_release(zero_page);
283 	zero_page = NULL;
284 }
285 
ceph_msgr_init(void)286 int ceph_msgr_init(void)
287 {
288 	BUG_ON(zero_page != NULL);
289 	zero_page = ZERO_PAGE(0);
290 	page_cache_get(zero_page);
291 
292 	if (ceph_msgr_slab_init())
293 		return -ENOMEM;
294 
295 	/*
296 	 * The number of active work items is limited by the number of
297 	 * connections, so leave @max_active at default.
298 	 */
299 	ceph_msgr_wq = alloc_workqueue("ceph-msgr", WQ_MEM_RECLAIM, 0);
300 	if (ceph_msgr_wq)
301 		return 0;
302 
303 	pr_err("msgr_init failed to create workqueue\n");
304 	_ceph_msgr_exit();
305 
306 	return -ENOMEM;
307 }
308 EXPORT_SYMBOL(ceph_msgr_init);
309 
ceph_msgr_exit(void)310 void ceph_msgr_exit(void)
311 {
312 	BUG_ON(ceph_msgr_wq == NULL);
313 
314 	_ceph_msgr_exit();
315 }
316 EXPORT_SYMBOL(ceph_msgr_exit);
317 
ceph_msgr_flush(void)318 void ceph_msgr_flush(void)
319 {
320 	flush_workqueue(ceph_msgr_wq);
321 }
322 EXPORT_SYMBOL(ceph_msgr_flush);
323 
324 /* Connection socket state transition functions */
325 
con_sock_state_init(struct ceph_connection * con)326 static void con_sock_state_init(struct ceph_connection *con)
327 {
328 	int old_state;
329 
330 	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED);
331 	if (WARN_ON(old_state != CON_SOCK_STATE_NEW))
332 		printk("%s: unexpected old state %d\n", __func__, old_state);
333 	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
334 	     CON_SOCK_STATE_CLOSED);
335 }
336 
con_sock_state_connecting(struct ceph_connection * con)337 static void con_sock_state_connecting(struct ceph_connection *con)
338 {
339 	int old_state;
340 
341 	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTING);
342 	if (WARN_ON(old_state != CON_SOCK_STATE_CLOSED))
343 		printk("%s: unexpected old state %d\n", __func__, old_state);
344 	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
345 	     CON_SOCK_STATE_CONNECTING);
346 }
347 
con_sock_state_connected(struct ceph_connection * con)348 static void con_sock_state_connected(struct ceph_connection *con)
349 {
350 	int old_state;
351 
352 	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CONNECTED);
353 	if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING))
354 		printk("%s: unexpected old state %d\n", __func__, old_state);
355 	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
356 	     CON_SOCK_STATE_CONNECTED);
357 }
358 
con_sock_state_closing(struct ceph_connection * con)359 static void con_sock_state_closing(struct ceph_connection *con)
360 {
361 	int old_state;
362 
363 	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSING);
364 	if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTING &&
365 			old_state != CON_SOCK_STATE_CONNECTED &&
366 			old_state != CON_SOCK_STATE_CLOSING))
367 		printk("%s: unexpected old state %d\n", __func__, old_state);
368 	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
369 	     CON_SOCK_STATE_CLOSING);
370 }
371 
con_sock_state_closed(struct ceph_connection * con)372 static void con_sock_state_closed(struct ceph_connection *con)
373 {
374 	int old_state;
375 
376 	old_state = atomic_xchg(&con->sock_state, CON_SOCK_STATE_CLOSED);
377 	if (WARN_ON(old_state != CON_SOCK_STATE_CONNECTED &&
378 		    old_state != CON_SOCK_STATE_CLOSING &&
379 		    old_state != CON_SOCK_STATE_CONNECTING &&
380 		    old_state != CON_SOCK_STATE_CLOSED))
381 		printk("%s: unexpected old state %d\n", __func__, old_state);
382 	dout("%s con %p sock %d -> %d\n", __func__, con, old_state,
383 	     CON_SOCK_STATE_CLOSED);
384 }
385 
386 /*
387  * socket callback functions
388  */
389 
390 /* data available on socket, or listen socket received a connect */
ceph_sock_data_ready(struct sock * sk)391 static void ceph_sock_data_ready(struct sock *sk)
392 {
393 	struct ceph_connection *con = sk->sk_user_data;
394 	if (atomic_read(&con->msgr->stopping)) {
395 		return;
396 	}
397 
398 	if (sk->sk_state != TCP_CLOSE_WAIT) {
399 		dout("%s on %p state = %lu, queueing work\n", __func__,
400 		     con, con->state);
401 		queue_con(con);
402 	}
403 }
404 
405 /* socket has buffer space for writing */
ceph_sock_write_space(struct sock * sk)406 static void ceph_sock_write_space(struct sock *sk)
407 {
408 	struct ceph_connection *con = sk->sk_user_data;
409 
410 	/* only queue to workqueue if there is data we want to write,
411 	 * and there is sufficient space in the socket buffer to accept
412 	 * more data.  clear SOCK_NOSPACE so that ceph_sock_write_space()
413 	 * doesn't get called again until try_write() fills the socket
414 	 * buffer. See net/ipv4/tcp_input.c:tcp_check_space()
415 	 * and net/core/stream.c:sk_stream_write_space().
416 	 */
417 	if (con_flag_test(con, CON_FLAG_WRITE_PENDING)) {
418 		if (sk_stream_is_writeable(sk)) {
419 			dout("%s %p queueing write work\n", __func__, con);
420 			clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
421 			queue_con(con);
422 		}
423 	} else {
424 		dout("%s %p nothing to write\n", __func__, con);
425 	}
426 }
427 
428 /* socket's state has changed */
ceph_sock_state_change(struct sock * sk)429 static void ceph_sock_state_change(struct sock *sk)
430 {
431 	struct ceph_connection *con = sk->sk_user_data;
432 
433 	dout("%s %p state = %lu sk_state = %u\n", __func__,
434 	     con, con->state, sk->sk_state);
435 
436 	switch (sk->sk_state) {
437 	case TCP_CLOSE:
438 		dout("%s TCP_CLOSE\n", __func__);
439 	case TCP_CLOSE_WAIT:
440 		dout("%s TCP_CLOSE_WAIT\n", __func__);
441 		con_sock_state_closing(con);
442 		con_flag_set(con, CON_FLAG_SOCK_CLOSED);
443 		queue_con(con);
444 		break;
445 	case TCP_ESTABLISHED:
446 		dout("%s TCP_ESTABLISHED\n", __func__);
447 		con_sock_state_connected(con);
448 		queue_con(con);
449 		break;
450 	default:	/* Everything else is uninteresting */
451 		break;
452 	}
453 }
454 
455 /*
456  * set up socket callbacks
457  */
set_sock_callbacks(struct socket * sock,struct ceph_connection * con)458 static void set_sock_callbacks(struct socket *sock,
459 			       struct ceph_connection *con)
460 {
461 	struct sock *sk = sock->sk;
462 	sk->sk_user_data = con;
463 	sk->sk_data_ready = ceph_sock_data_ready;
464 	sk->sk_write_space = ceph_sock_write_space;
465 	sk->sk_state_change = ceph_sock_state_change;
466 }
467 
468 
469 /*
470  * socket helpers
471  */
472 
473 /*
474  * initiate connection to a remote socket.
475  */
ceph_tcp_connect(struct ceph_connection * con)476 static int ceph_tcp_connect(struct ceph_connection *con)
477 {
478 	struct sockaddr_storage *paddr = &con->peer_addr.in_addr;
479 	struct socket *sock;
480 	int ret;
481 
482 	BUG_ON(con->sock);
483 	ret = sock_create_kern(con->peer_addr.in_addr.ss_family, SOCK_STREAM,
484 			       IPPROTO_TCP, &sock);
485 	if (ret)
486 		return ret;
487 	sock->sk->sk_allocation = GFP_NOFS;
488 
489 #ifdef CONFIG_LOCKDEP
490 	lockdep_set_class(&sock->sk->sk_lock, &socket_class);
491 #endif
492 
493 	set_sock_callbacks(sock, con);
494 
495 	dout("connect %s\n", ceph_pr_addr(&con->peer_addr.in_addr));
496 
497 	con_sock_state_connecting(con);
498 	ret = sock->ops->connect(sock, (struct sockaddr *)paddr, sizeof(*paddr),
499 				 O_NONBLOCK);
500 	if (ret == -EINPROGRESS) {
501 		dout("connect %s EINPROGRESS sk_state = %u\n",
502 		     ceph_pr_addr(&con->peer_addr.in_addr),
503 		     sock->sk->sk_state);
504 	} else if (ret < 0) {
505 		pr_err("connect %s error %d\n",
506 		       ceph_pr_addr(&con->peer_addr.in_addr), ret);
507 		sock_release(sock);
508 		return ret;
509 	}
510 
511 	if (con->msgr->tcp_nodelay) {
512 		int optval = 1;
513 
514 		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
515 					(char *)&optval, sizeof(optval));
516 		if (ret)
517 			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
518 			       ret);
519 	}
520 
521 	con->sock = sock;
522 	return 0;
523 }
524 
ceph_tcp_recvmsg(struct socket * sock,void * buf,size_t len)525 static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len)
526 {
527 	struct kvec iov = {buf, len};
528 	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
529 	int r;
530 
531 	r = kernel_recvmsg(sock, &msg, &iov, 1, len, msg.msg_flags);
532 	if (r == -EAGAIN)
533 		r = 0;
534 	return r;
535 }
536 
ceph_tcp_recvpage(struct socket * sock,struct page * page,int page_offset,size_t length)537 static int ceph_tcp_recvpage(struct socket *sock, struct page *page,
538 		     int page_offset, size_t length)
539 {
540 	void *kaddr;
541 	int ret;
542 
543 	BUG_ON(page_offset + length > PAGE_SIZE);
544 
545 	kaddr = kmap(page);
546 	BUG_ON(!kaddr);
547 	ret = ceph_tcp_recvmsg(sock, kaddr + page_offset, length);
548 	kunmap(page);
549 
550 	return ret;
551 }
552 
553 /*
554  * write something.  @more is true if caller will be sending more data
555  * shortly.
556  */
ceph_tcp_sendmsg(struct socket * sock,struct kvec * iov,size_t kvlen,size_t len,int more)557 static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov,
558 		     size_t kvlen, size_t len, int more)
559 {
560 	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
561 	int r;
562 
563 	if (more)
564 		msg.msg_flags |= MSG_MORE;
565 	else
566 		msg.msg_flags |= MSG_EOR;  /* superfluous, but what the hell */
567 
568 	r = kernel_sendmsg(sock, &msg, iov, kvlen, len);
569 	if (r == -EAGAIN)
570 		r = 0;
571 	return r;
572 }
573 
__ceph_tcp_sendpage(struct socket * sock,struct page * page,int offset,size_t size,bool more)574 static int __ceph_tcp_sendpage(struct socket *sock, struct page *page,
575 		     int offset, size_t size, bool more)
576 {
577 	int flags = MSG_DONTWAIT | MSG_NOSIGNAL | (more ? MSG_MORE : MSG_EOR);
578 	int ret;
579 
580 	ret = kernel_sendpage(sock, page, offset, size, flags);
581 	if (ret == -EAGAIN)
582 		ret = 0;
583 
584 	return ret;
585 }
586 
ceph_tcp_sendpage(struct socket * sock,struct page * page,int offset,size_t size,bool more)587 static int ceph_tcp_sendpage(struct socket *sock, struct page *page,
588 		     int offset, size_t size, bool more)
589 {
590 	int ret;
591 	struct kvec iov;
592 
593 	/* sendpage cannot properly handle pages with page_count == 0,
594 	 * we need to fallback to sendmsg if that's the case */
595 	if (page_count(page) >= 1)
596 		return __ceph_tcp_sendpage(sock, page, offset, size, more);
597 
598 	iov.iov_base = kmap(page) + offset;
599 	iov.iov_len = size;
600 	ret = ceph_tcp_sendmsg(sock, &iov, 1, size, more);
601 	kunmap(page);
602 
603 	return ret;
604 }
605 
606 /*
607  * Shutdown/close the socket for the given connection.
608  */
con_close_socket(struct ceph_connection * con)609 static int con_close_socket(struct ceph_connection *con)
610 {
611 	int rc = 0;
612 
613 	dout("con_close_socket on %p sock %p\n", con, con->sock);
614 	if (con->sock) {
615 		rc = con->sock->ops->shutdown(con->sock, SHUT_RDWR);
616 		sock_release(con->sock);
617 		con->sock = NULL;
618 	}
619 
620 	/*
621 	 * Forcibly clear the SOCK_CLOSED flag.  It gets set
622 	 * independent of the connection mutex, and we could have
623 	 * received a socket close event before we had the chance to
624 	 * shut the socket down.
625 	 */
626 	con_flag_clear(con, CON_FLAG_SOCK_CLOSED);
627 
628 	con_sock_state_closed(con);
629 	return rc;
630 }
631 
632 /*
633  * Reset a connection.  Discard all incoming and outgoing messages
634  * and clear *_seq state.
635  */
ceph_msg_remove(struct ceph_msg * msg)636 static void ceph_msg_remove(struct ceph_msg *msg)
637 {
638 	list_del_init(&msg->list_head);
639 	BUG_ON(msg->con == NULL);
640 	msg->con->ops->put(msg->con);
641 	msg->con = NULL;
642 
643 	ceph_msg_put(msg);
644 }
ceph_msg_remove_list(struct list_head * head)645 static void ceph_msg_remove_list(struct list_head *head)
646 {
647 	while (!list_empty(head)) {
648 		struct ceph_msg *msg = list_first_entry(head, struct ceph_msg,
649 							list_head);
650 		ceph_msg_remove(msg);
651 	}
652 }
653 
reset_connection(struct ceph_connection * con)654 static void reset_connection(struct ceph_connection *con)
655 {
656 	/* reset connection, out_queue, msg_ and connect_seq */
657 	/* discard existing out_queue and msg_seq */
658 	dout("reset_connection %p\n", con);
659 	ceph_msg_remove_list(&con->out_queue);
660 	ceph_msg_remove_list(&con->out_sent);
661 
662 	if (con->in_msg) {
663 		BUG_ON(con->in_msg->con != con);
664 		con->in_msg->con = NULL;
665 		ceph_msg_put(con->in_msg);
666 		con->in_msg = NULL;
667 		con->ops->put(con);
668 	}
669 
670 	con->connect_seq = 0;
671 	con->out_seq = 0;
672 	if (con->out_msg) {
673 		ceph_msg_put(con->out_msg);
674 		con->out_msg = NULL;
675 	}
676 	con->in_seq = 0;
677 	con->in_seq_acked = 0;
678 
679 	con->out_skip = 0;
680 }
681 
682 /*
683  * mark a peer down.  drop any open connections.
684  */
ceph_con_close(struct ceph_connection * con)685 void ceph_con_close(struct ceph_connection *con)
686 {
687 	mutex_lock(&con->mutex);
688 	dout("con_close %p peer %s\n", con,
689 	     ceph_pr_addr(&con->peer_addr.in_addr));
690 	con->state = CON_STATE_CLOSED;
691 
692 	con_flag_clear(con, CON_FLAG_LOSSYTX);	/* so we retry next connect */
693 	con_flag_clear(con, CON_FLAG_KEEPALIVE_PENDING);
694 	con_flag_clear(con, CON_FLAG_WRITE_PENDING);
695 	con_flag_clear(con, CON_FLAG_BACKOFF);
696 
697 	reset_connection(con);
698 	con->peer_global_seq = 0;
699 	cancel_con(con);
700 	con_close_socket(con);
701 	mutex_unlock(&con->mutex);
702 }
703 EXPORT_SYMBOL(ceph_con_close);
704 
705 /*
706  * Reopen a closed connection, with a new peer address.
707  */
ceph_con_open(struct ceph_connection * con,__u8 entity_type,__u64 entity_num,struct ceph_entity_addr * addr)708 void ceph_con_open(struct ceph_connection *con,
709 		   __u8 entity_type, __u64 entity_num,
710 		   struct ceph_entity_addr *addr)
711 {
712 	mutex_lock(&con->mutex);
713 	dout("con_open %p %s\n", con, ceph_pr_addr(&addr->in_addr));
714 
715 	WARN_ON(con->state != CON_STATE_CLOSED);
716 	con->state = CON_STATE_PREOPEN;
717 
718 	con->peer_name.type = (__u8) entity_type;
719 	con->peer_name.num = cpu_to_le64(entity_num);
720 
721 	memcpy(&con->peer_addr, addr, sizeof(*addr));
722 	con->delay = 0;      /* reset backoff memory */
723 	mutex_unlock(&con->mutex);
724 	queue_con(con);
725 }
726 EXPORT_SYMBOL(ceph_con_open);
727 
728 /*
729  * return true if this connection ever successfully opened
730  */
ceph_con_opened(struct ceph_connection * con)731 bool ceph_con_opened(struct ceph_connection *con)
732 {
733 	return con->connect_seq > 0;
734 }
735 
736 /*
737  * initialize a new connection.
738  */
ceph_con_init(struct ceph_connection * con,void * private,const struct ceph_connection_operations * ops,struct ceph_messenger * msgr)739 void ceph_con_init(struct ceph_connection *con, void *private,
740 	const struct ceph_connection_operations *ops,
741 	struct ceph_messenger *msgr)
742 {
743 	dout("con_init %p\n", con);
744 	memset(con, 0, sizeof(*con));
745 	con->private = private;
746 	con->ops = ops;
747 	con->msgr = msgr;
748 
749 	con_sock_state_init(con);
750 
751 	mutex_init(&con->mutex);
752 	INIT_LIST_HEAD(&con->out_queue);
753 	INIT_LIST_HEAD(&con->out_sent);
754 	INIT_DELAYED_WORK(&con->work, con_work);
755 
756 	con->state = CON_STATE_CLOSED;
757 }
758 EXPORT_SYMBOL(ceph_con_init);
759 
760 
761 /*
762  * We maintain a global counter to order connection attempts.  Get
763  * a unique seq greater than @gt.
764  */
get_global_seq(struct ceph_messenger * msgr,u32 gt)765 static u32 get_global_seq(struct ceph_messenger *msgr, u32 gt)
766 {
767 	u32 ret;
768 
769 	spin_lock(&msgr->global_seq_lock);
770 	if (msgr->global_seq < gt)
771 		msgr->global_seq = gt;
772 	ret = ++msgr->global_seq;
773 	spin_unlock(&msgr->global_seq_lock);
774 	return ret;
775 }
776 
con_out_kvec_reset(struct ceph_connection * con)777 static void con_out_kvec_reset(struct ceph_connection *con)
778 {
779 	BUG_ON(con->out_skip);
780 
781 	con->out_kvec_left = 0;
782 	con->out_kvec_bytes = 0;
783 	con->out_kvec_cur = &con->out_kvec[0];
784 }
785 
con_out_kvec_add(struct ceph_connection * con,size_t size,void * data)786 static void con_out_kvec_add(struct ceph_connection *con,
787 				size_t size, void *data)
788 {
789 	int index = con->out_kvec_left;
790 
791 	BUG_ON(con->out_skip);
792 	BUG_ON(index >= ARRAY_SIZE(con->out_kvec));
793 
794 	con->out_kvec[index].iov_len = size;
795 	con->out_kvec[index].iov_base = data;
796 	con->out_kvec_left++;
797 	con->out_kvec_bytes += size;
798 }
799 
800 /*
801  * Chop off a kvec from the end.  Return residual number of bytes for
802  * that kvec, i.e. how many bytes would have been written if the kvec
803  * hadn't been nuked.
804  */
con_out_kvec_skip(struct ceph_connection * con)805 static int con_out_kvec_skip(struct ceph_connection *con)
806 {
807 	int off = con->out_kvec_cur - con->out_kvec;
808 	int skip = 0;
809 
810 	if (con->out_kvec_bytes > 0) {
811 		skip = con->out_kvec[off + con->out_kvec_left - 1].iov_len;
812 		BUG_ON(con->out_kvec_bytes < skip);
813 		BUG_ON(!con->out_kvec_left);
814 		con->out_kvec_bytes -= skip;
815 		con->out_kvec_left--;
816 	}
817 
818 	return skip;
819 }
820 
821 #ifdef CONFIG_BLOCK
822 
823 /*
824  * For a bio data item, a piece is whatever remains of the next
825  * entry in the current bio iovec, or the first entry in the next
826  * bio in the list.
827  */
ceph_msg_data_bio_cursor_init(struct ceph_msg_data_cursor * cursor,size_t length)828 static void ceph_msg_data_bio_cursor_init(struct ceph_msg_data_cursor *cursor,
829 					size_t length)
830 {
831 	struct ceph_msg_data *data = cursor->data;
832 	struct bio *bio;
833 
834 	BUG_ON(data->type != CEPH_MSG_DATA_BIO);
835 
836 	bio = data->bio;
837 	BUG_ON(!bio);
838 
839 	cursor->resid = min(length, data->bio_length);
840 	cursor->bio = bio;
841 	cursor->bvec_iter = bio->bi_iter;
842 	cursor->last_piece =
843 		cursor->resid <= bio_iter_len(bio, cursor->bvec_iter);
844 }
845 
ceph_msg_data_bio_next(struct ceph_msg_data_cursor * cursor,size_t * page_offset,size_t * length)846 static struct page *ceph_msg_data_bio_next(struct ceph_msg_data_cursor *cursor,
847 						size_t *page_offset,
848 						size_t *length)
849 {
850 	struct ceph_msg_data *data = cursor->data;
851 	struct bio *bio;
852 	struct bio_vec bio_vec;
853 
854 	BUG_ON(data->type != CEPH_MSG_DATA_BIO);
855 
856 	bio = cursor->bio;
857 	BUG_ON(!bio);
858 
859 	bio_vec = bio_iter_iovec(bio, cursor->bvec_iter);
860 
861 	*page_offset = (size_t) bio_vec.bv_offset;
862 	BUG_ON(*page_offset >= PAGE_SIZE);
863 	if (cursor->last_piece) /* pagelist offset is always 0 */
864 		*length = cursor->resid;
865 	else
866 		*length = (size_t) bio_vec.bv_len;
867 	BUG_ON(*length > cursor->resid);
868 	BUG_ON(*page_offset + *length > PAGE_SIZE);
869 
870 	return bio_vec.bv_page;
871 }
872 
ceph_msg_data_bio_advance(struct ceph_msg_data_cursor * cursor,size_t bytes)873 static bool ceph_msg_data_bio_advance(struct ceph_msg_data_cursor *cursor,
874 					size_t bytes)
875 {
876 	struct bio *bio;
877 	struct bio_vec bio_vec;
878 
879 	BUG_ON(cursor->data->type != CEPH_MSG_DATA_BIO);
880 
881 	bio = cursor->bio;
882 	BUG_ON(!bio);
883 
884 	bio_vec = bio_iter_iovec(bio, cursor->bvec_iter);
885 
886 	/* Advance the cursor offset */
887 
888 	BUG_ON(cursor->resid < bytes);
889 	cursor->resid -= bytes;
890 
891 	bio_advance_iter(bio, &cursor->bvec_iter, bytes);
892 
893 	if (bytes < bio_vec.bv_len)
894 		return false;	/* more bytes to process in this segment */
895 
896 	/* Move on to the next segment, and possibly the next bio */
897 
898 	if (!cursor->bvec_iter.bi_size) {
899 		bio = bio->bi_next;
900 		cursor->bio = bio;
901 		if (bio)
902 			cursor->bvec_iter = bio->bi_iter;
903 		else
904 			memset(&cursor->bvec_iter, 0,
905 			       sizeof(cursor->bvec_iter));
906 	}
907 
908 	if (!cursor->last_piece) {
909 		BUG_ON(!cursor->resid);
910 		BUG_ON(!bio);
911 		/* A short read is OK, so use <= rather than == */
912 		if (cursor->resid <= bio_iter_len(bio, cursor->bvec_iter))
913 			cursor->last_piece = true;
914 	}
915 
916 	return true;
917 }
918 #endif /* CONFIG_BLOCK */
919 
920 /*
921  * For a page array, a piece comes from the first page in the array
922  * that has not already been fully consumed.
923  */
ceph_msg_data_pages_cursor_init(struct ceph_msg_data_cursor * cursor,size_t length)924 static void ceph_msg_data_pages_cursor_init(struct ceph_msg_data_cursor *cursor,
925 					size_t length)
926 {
927 	struct ceph_msg_data *data = cursor->data;
928 	int page_count;
929 
930 	BUG_ON(data->type != CEPH_MSG_DATA_PAGES);
931 
932 	BUG_ON(!data->pages);
933 	BUG_ON(!data->length);
934 
935 	cursor->resid = min(length, data->length);
936 	page_count = calc_pages_for(data->alignment, (u64)data->length);
937 	cursor->page_offset = data->alignment & ~PAGE_MASK;
938 	cursor->page_index = 0;
939 	BUG_ON(page_count > (int)USHRT_MAX);
940 	cursor->page_count = (unsigned short)page_count;
941 	BUG_ON(length > SIZE_MAX - cursor->page_offset);
942 	cursor->last_piece = cursor->page_offset + cursor->resid <= PAGE_SIZE;
943 }
944 
945 static struct page *
ceph_msg_data_pages_next(struct ceph_msg_data_cursor * cursor,size_t * page_offset,size_t * length)946 ceph_msg_data_pages_next(struct ceph_msg_data_cursor *cursor,
947 					size_t *page_offset, size_t *length)
948 {
949 	struct ceph_msg_data *data = cursor->data;
950 
951 	BUG_ON(data->type != CEPH_MSG_DATA_PAGES);
952 
953 	BUG_ON(cursor->page_index >= cursor->page_count);
954 	BUG_ON(cursor->page_offset >= PAGE_SIZE);
955 
956 	*page_offset = cursor->page_offset;
957 	if (cursor->last_piece)
958 		*length = cursor->resid;
959 	else
960 		*length = PAGE_SIZE - *page_offset;
961 
962 	return data->pages[cursor->page_index];
963 }
964 
ceph_msg_data_pages_advance(struct ceph_msg_data_cursor * cursor,size_t bytes)965 static bool ceph_msg_data_pages_advance(struct ceph_msg_data_cursor *cursor,
966 						size_t bytes)
967 {
968 	BUG_ON(cursor->data->type != CEPH_MSG_DATA_PAGES);
969 
970 	BUG_ON(cursor->page_offset + bytes > PAGE_SIZE);
971 
972 	/* Advance the cursor page offset */
973 
974 	cursor->resid -= bytes;
975 	cursor->page_offset = (cursor->page_offset + bytes) & ~PAGE_MASK;
976 	if (!bytes || cursor->page_offset)
977 		return false;	/* more bytes to process in the current page */
978 
979 	if (!cursor->resid)
980 		return false;   /* no more data */
981 
982 	/* Move on to the next page; offset is already at 0 */
983 
984 	BUG_ON(cursor->page_index >= cursor->page_count);
985 	cursor->page_index++;
986 	cursor->last_piece = cursor->resid <= PAGE_SIZE;
987 
988 	return true;
989 }
990 
991 /*
992  * For a pagelist, a piece is whatever remains to be consumed in the
993  * first page in the list, or the front of the next page.
994  */
995 static void
ceph_msg_data_pagelist_cursor_init(struct ceph_msg_data_cursor * cursor,size_t length)996 ceph_msg_data_pagelist_cursor_init(struct ceph_msg_data_cursor *cursor,
997 					size_t length)
998 {
999 	struct ceph_msg_data *data = cursor->data;
1000 	struct ceph_pagelist *pagelist;
1001 	struct page *page;
1002 
1003 	BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST);
1004 
1005 	pagelist = data->pagelist;
1006 	BUG_ON(!pagelist);
1007 
1008 	if (!length)
1009 		return;		/* pagelist can be assigned but empty */
1010 
1011 	BUG_ON(list_empty(&pagelist->head));
1012 	page = list_first_entry(&pagelist->head, struct page, lru);
1013 
1014 	cursor->resid = min(length, pagelist->length);
1015 	cursor->page = page;
1016 	cursor->offset = 0;
1017 	cursor->last_piece = cursor->resid <= PAGE_SIZE;
1018 }
1019 
1020 static struct page *
ceph_msg_data_pagelist_next(struct ceph_msg_data_cursor * cursor,size_t * page_offset,size_t * length)1021 ceph_msg_data_pagelist_next(struct ceph_msg_data_cursor *cursor,
1022 				size_t *page_offset, size_t *length)
1023 {
1024 	struct ceph_msg_data *data = cursor->data;
1025 	struct ceph_pagelist *pagelist;
1026 
1027 	BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST);
1028 
1029 	pagelist = data->pagelist;
1030 	BUG_ON(!pagelist);
1031 
1032 	BUG_ON(!cursor->page);
1033 	BUG_ON(cursor->offset + cursor->resid != pagelist->length);
1034 
1035 	/* offset of first page in pagelist is always 0 */
1036 	*page_offset = cursor->offset & ~PAGE_MASK;
1037 	if (cursor->last_piece)
1038 		*length = cursor->resid;
1039 	else
1040 		*length = PAGE_SIZE - *page_offset;
1041 
1042 	return cursor->page;
1043 }
1044 
ceph_msg_data_pagelist_advance(struct ceph_msg_data_cursor * cursor,size_t bytes)1045 static bool ceph_msg_data_pagelist_advance(struct ceph_msg_data_cursor *cursor,
1046 						size_t bytes)
1047 {
1048 	struct ceph_msg_data *data = cursor->data;
1049 	struct ceph_pagelist *pagelist;
1050 
1051 	BUG_ON(data->type != CEPH_MSG_DATA_PAGELIST);
1052 
1053 	pagelist = data->pagelist;
1054 	BUG_ON(!pagelist);
1055 
1056 	BUG_ON(cursor->offset + cursor->resid != pagelist->length);
1057 	BUG_ON((cursor->offset & ~PAGE_MASK) + bytes > PAGE_SIZE);
1058 
1059 	/* Advance the cursor offset */
1060 
1061 	cursor->resid -= bytes;
1062 	cursor->offset += bytes;
1063 	/* offset of first page in pagelist is always 0 */
1064 	if (!bytes || cursor->offset & ~PAGE_MASK)
1065 		return false;	/* more bytes to process in the current page */
1066 
1067 	if (!cursor->resid)
1068 		return false;   /* no more data */
1069 
1070 	/* Move on to the next page */
1071 
1072 	BUG_ON(list_is_last(&cursor->page->lru, &pagelist->head));
1073 	cursor->page = list_entry_next(cursor->page, lru);
1074 	cursor->last_piece = cursor->resid <= PAGE_SIZE;
1075 
1076 	return true;
1077 }
1078 
1079 /*
1080  * Message data is handled (sent or received) in pieces, where each
1081  * piece resides on a single page.  The network layer might not
1082  * consume an entire piece at once.  A data item's cursor keeps
1083  * track of which piece is next to process and how much remains to
1084  * be processed in that piece.  It also tracks whether the current
1085  * piece is the last one in the data item.
1086  */
__ceph_msg_data_cursor_init(struct ceph_msg_data_cursor * cursor)1087 static void __ceph_msg_data_cursor_init(struct ceph_msg_data_cursor *cursor)
1088 {
1089 	size_t length = cursor->total_resid;
1090 
1091 	switch (cursor->data->type) {
1092 	case CEPH_MSG_DATA_PAGELIST:
1093 		ceph_msg_data_pagelist_cursor_init(cursor, length);
1094 		break;
1095 	case CEPH_MSG_DATA_PAGES:
1096 		ceph_msg_data_pages_cursor_init(cursor, length);
1097 		break;
1098 #ifdef CONFIG_BLOCK
1099 	case CEPH_MSG_DATA_BIO:
1100 		ceph_msg_data_bio_cursor_init(cursor, length);
1101 		break;
1102 #endif /* CONFIG_BLOCK */
1103 	case CEPH_MSG_DATA_NONE:
1104 	default:
1105 		/* BUG(); */
1106 		break;
1107 	}
1108 	cursor->need_crc = true;
1109 }
1110 
ceph_msg_data_cursor_init(struct ceph_msg * msg,size_t length)1111 static void ceph_msg_data_cursor_init(struct ceph_msg *msg, size_t length)
1112 {
1113 	struct ceph_msg_data_cursor *cursor = &msg->cursor;
1114 	struct ceph_msg_data *data;
1115 
1116 	BUG_ON(!length);
1117 	BUG_ON(length > msg->data_length);
1118 	BUG_ON(list_empty(&msg->data));
1119 
1120 	cursor->data_head = &msg->data;
1121 	cursor->total_resid = length;
1122 	data = list_first_entry(&msg->data, struct ceph_msg_data, links);
1123 	cursor->data = data;
1124 
1125 	__ceph_msg_data_cursor_init(cursor);
1126 }
1127 
1128 /*
1129  * Return the page containing the next piece to process for a given
1130  * data item, and supply the page offset and length of that piece.
1131  * Indicate whether this is the last piece in this data item.
1132  */
ceph_msg_data_next(struct ceph_msg_data_cursor * cursor,size_t * page_offset,size_t * length,bool * last_piece)1133 static struct page *ceph_msg_data_next(struct ceph_msg_data_cursor *cursor,
1134 					size_t *page_offset, size_t *length,
1135 					bool *last_piece)
1136 {
1137 	struct page *page;
1138 
1139 	switch (cursor->data->type) {
1140 	case CEPH_MSG_DATA_PAGELIST:
1141 		page = ceph_msg_data_pagelist_next(cursor, page_offset, length);
1142 		break;
1143 	case CEPH_MSG_DATA_PAGES:
1144 		page = ceph_msg_data_pages_next(cursor, page_offset, length);
1145 		break;
1146 #ifdef CONFIG_BLOCK
1147 	case CEPH_MSG_DATA_BIO:
1148 		page = ceph_msg_data_bio_next(cursor, page_offset, length);
1149 		break;
1150 #endif /* CONFIG_BLOCK */
1151 	case CEPH_MSG_DATA_NONE:
1152 	default:
1153 		page = NULL;
1154 		break;
1155 	}
1156 	BUG_ON(!page);
1157 	BUG_ON(*page_offset + *length > PAGE_SIZE);
1158 	BUG_ON(!*length);
1159 	if (last_piece)
1160 		*last_piece = cursor->last_piece;
1161 
1162 	return page;
1163 }
1164 
1165 /*
1166  * Returns true if the result moves the cursor on to the next piece
1167  * of the data item.
1168  */
ceph_msg_data_advance(struct ceph_msg_data_cursor * cursor,size_t bytes)1169 static bool ceph_msg_data_advance(struct ceph_msg_data_cursor *cursor,
1170 				size_t bytes)
1171 {
1172 	bool new_piece;
1173 
1174 	BUG_ON(bytes > cursor->resid);
1175 	switch (cursor->data->type) {
1176 	case CEPH_MSG_DATA_PAGELIST:
1177 		new_piece = ceph_msg_data_pagelist_advance(cursor, bytes);
1178 		break;
1179 	case CEPH_MSG_DATA_PAGES:
1180 		new_piece = ceph_msg_data_pages_advance(cursor, bytes);
1181 		break;
1182 #ifdef CONFIG_BLOCK
1183 	case CEPH_MSG_DATA_BIO:
1184 		new_piece = ceph_msg_data_bio_advance(cursor, bytes);
1185 		break;
1186 #endif /* CONFIG_BLOCK */
1187 	case CEPH_MSG_DATA_NONE:
1188 	default:
1189 		BUG();
1190 		break;
1191 	}
1192 	cursor->total_resid -= bytes;
1193 
1194 	if (!cursor->resid && cursor->total_resid) {
1195 		WARN_ON(!cursor->last_piece);
1196 		BUG_ON(list_is_last(&cursor->data->links, cursor->data_head));
1197 		cursor->data = list_entry_next(cursor->data, links);
1198 		__ceph_msg_data_cursor_init(cursor);
1199 		new_piece = true;
1200 	}
1201 	cursor->need_crc = new_piece;
1202 
1203 	return new_piece;
1204 }
1205 
sizeof_footer(struct ceph_connection * con)1206 static size_t sizeof_footer(struct ceph_connection *con)
1207 {
1208 	return (con->peer_features & CEPH_FEATURE_MSG_AUTH) ?
1209 	    sizeof(struct ceph_msg_footer) :
1210 	    sizeof(struct ceph_msg_footer_old);
1211 }
1212 
prepare_message_data(struct ceph_msg * msg,u32 data_len)1213 static void prepare_message_data(struct ceph_msg *msg, u32 data_len)
1214 {
1215 	BUG_ON(!msg);
1216 	BUG_ON(!data_len);
1217 
1218 	/* Initialize data cursor */
1219 
1220 	ceph_msg_data_cursor_init(msg, (size_t)data_len);
1221 }
1222 
1223 /*
1224  * Prepare footer for currently outgoing message, and finish things
1225  * off.  Assumes out_kvec* are already valid.. we just add on to the end.
1226  */
prepare_write_message_footer(struct ceph_connection * con)1227 static void prepare_write_message_footer(struct ceph_connection *con)
1228 {
1229 	struct ceph_msg *m = con->out_msg;
1230 	int v = con->out_kvec_left;
1231 
1232 	m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE;
1233 
1234 	dout("prepare_write_message_footer %p\n", con);
1235 	con->out_kvec[v].iov_base = &m->footer;
1236 	if (con->peer_features & CEPH_FEATURE_MSG_AUTH) {
1237 		if (con->ops->sign_message)
1238 			con->ops->sign_message(con, m);
1239 		else
1240 			m->footer.sig = 0;
1241 		con->out_kvec[v].iov_len = sizeof(m->footer);
1242 		con->out_kvec_bytes += sizeof(m->footer);
1243 	} else {
1244 		m->old_footer.flags = m->footer.flags;
1245 		con->out_kvec[v].iov_len = sizeof(m->old_footer);
1246 		con->out_kvec_bytes += sizeof(m->old_footer);
1247 	}
1248 	con->out_kvec_left++;
1249 	con->out_more = m->more_to_follow;
1250 	con->out_msg_done = true;
1251 }
1252 
1253 /*
1254  * Prepare headers for the next outgoing message.
1255  */
prepare_write_message(struct ceph_connection * con)1256 static void prepare_write_message(struct ceph_connection *con)
1257 {
1258 	struct ceph_msg *m;
1259 	u32 crc;
1260 
1261 	con_out_kvec_reset(con);
1262 	con->out_msg_done = false;
1263 
1264 	/* Sneak an ack in there first?  If we can get it into the same
1265 	 * TCP packet that's a good thing. */
1266 	if (con->in_seq > con->in_seq_acked) {
1267 		con->in_seq_acked = con->in_seq;
1268 		con_out_kvec_add(con, sizeof (tag_ack), &tag_ack);
1269 		con->out_temp_ack = cpu_to_le64(con->in_seq_acked);
1270 		con_out_kvec_add(con, sizeof (con->out_temp_ack),
1271 			&con->out_temp_ack);
1272 	}
1273 
1274 	BUG_ON(list_empty(&con->out_queue));
1275 	m = list_first_entry(&con->out_queue, struct ceph_msg, list_head);
1276 	con->out_msg = m;
1277 	BUG_ON(m->con != con);
1278 
1279 	/* put message on sent list */
1280 	ceph_msg_get(m);
1281 	list_move_tail(&m->list_head, &con->out_sent);
1282 
1283 	/*
1284 	 * only assign outgoing seq # if we haven't sent this message
1285 	 * yet.  if it is requeued, resend with it's original seq.
1286 	 */
1287 	if (m->needs_out_seq) {
1288 		m->hdr.seq = cpu_to_le64(++con->out_seq);
1289 		m->needs_out_seq = false;
1290 	}
1291 	WARN_ON(m->data_length != le32_to_cpu(m->hdr.data_len));
1292 
1293 	dout("prepare_write_message %p seq %lld type %d len %d+%d+%zd\n",
1294 	     m, con->out_seq, le16_to_cpu(m->hdr.type),
1295 	     le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len),
1296 	     m->data_length);
1297 	BUG_ON(le32_to_cpu(m->hdr.front_len) != m->front.iov_len);
1298 
1299 	/* tag + hdr + front + middle */
1300 	con_out_kvec_add(con, sizeof (tag_msg), &tag_msg);
1301 	con_out_kvec_add(con, sizeof(con->out_hdr), &con->out_hdr);
1302 	con_out_kvec_add(con, m->front.iov_len, m->front.iov_base);
1303 
1304 	if (m->middle)
1305 		con_out_kvec_add(con, m->middle->vec.iov_len,
1306 			m->middle->vec.iov_base);
1307 
1308 	/* fill in hdr crc and finalize hdr */
1309 	crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc));
1310 	con->out_msg->hdr.crc = cpu_to_le32(crc);
1311 	memcpy(&con->out_hdr, &con->out_msg->hdr, sizeof(con->out_hdr));
1312 
1313 	/* fill in front and middle crc, footer */
1314 	crc = crc32c(0, m->front.iov_base, m->front.iov_len);
1315 	con->out_msg->footer.front_crc = cpu_to_le32(crc);
1316 	if (m->middle) {
1317 		crc = crc32c(0, m->middle->vec.iov_base,
1318 				m->middle->vec.iov_len);
1319 		con->out_msg->footer.middle_crc = cpu_to_le32(crc);
1320 	} else
1321 		con->out_msg->footer.middle_crc = 0;
1322 	dout("%s front_crc %u middle_crc %u\n", __func__,
1323 	     le32_to_cpu(con->out_msg->footer.front_crc),
1324 	     le32_to_cpu(con->out_msg->footer.middle_crc));
1325 	con->out_msg->footer.flags = 0;
1326 
1327 	/* is there a data payload? */
1328 	con->out_msg->footer.data_crc = 0;
1329 	if (m->data_length) {
1330 		prepare_message_data(con->out_msg, m->data_length);
1331 		con->out_more = 1;  /* data + footer will follow */
1332 	} else {
1333 		/* no, queue up footer too and be done */
1334 		prepare_write_message_footer(con);
1335 	}
1336 
1337 	con_flag_set(con, CON_FLAG_WRITE_PENDING);
1338 }
1339 
1340 /*
1341  * Prepare an ack.
1342  */
prepare_write_ack(struct ceph_connection * con)1343 static void prepare_write_ack(struct ceph_connection *con)
1344 {
1345 	dout("prepare_write_ack %p %llu -> %llu\n", con,
1346 	     con->in_seq_acked, con->in_seq);
1347 	con->in_seq_acked = con->in_seq;
1348 
1349 	con_out_kvec_reset(con);
1350 
1351 	con_out_kvec_add(con, sizeof (tag_ack), &tag_ack);
1352 
1353 	con->out_temp_ack = cpu_to_le64(con->in_seq_acked);
1354 	con_out_kvec_add(con, sizeof (con->out_temp_ack),
1355 				&con->out_temp_ack);
1356 
1357 	con->out_more = 1;  /* more will follow.. eventually.. */
1358 	con_flag_set(con, CON_FLAG_WRITE_PENDING);
1359 }
1360 
1361 /*
1362  * Prepare to share the seq during handshake
1363  */
prepare_write_seq(struct ceph_connection * con)1364 static void prepare_write_seq(struct ceph_connection *con)
1365 {
1366 	dout("prepare_write_seq %p %llu -> %llu\n", con,
1367 	     con->in_seq_acked, con->in_seq);
1368 	con->in_seq_acked = con->in_seq;
1369 
1370 	con_out_kvec_reset(con);
1371 
1372 	con->out_temp_ack = cpu_to_le64(con->in_seq_acked);
1373 	con_out_kvec_add(con, sizeof (con->out_temp_ack),
1374 			 &con->out_temp_ack);
1375 
1376 	con_flag_set(con, CON_FLAG_WRITE_PENDING);
1377 }
1378 
1379 /*
1380  * Prepare to write keepalive byte.
1381  */
prepare_write_keepalive(struct ceph_connection * con)1382 static void prepare_write_keepalive(struct ceph_connection *con)
1383 {
1384 	dout("prepare_write_keepalive %p\n", con);
1385 	con_out_kvec_reset(con);
1386 	con_out_kvec_add(con, sizeof (tag_keepalive), &tag_keepalive);
1387 	con_flag_set(con, CON_FLAG_WRITE_PENDING);
1388 }
1389 
1390 /*
1391  * Connection negotiation.
1392  */
1393 
get_connect_authorizer(struct ceph_connection * con,int * auth_proto)1394 static struct ceph_auth_handshake *get_connect_authorizer(struct ceph_connection *con,
1395 						int *auth_proto)
1396 {
1397 	struct ceph_auth_handshake *auth;
1398 
1399 	if (!con->ops->get_authorizer) {
1400 		con->out_connect.authorizer_protocol = CEPH_AUTH_UNKNOWN;
1401 		con->out_connect.authorizer_len = 0;
1402 		return NULL;
1403 	}
1404 
1405 	/* Can't hold the mutex while getting authorizer */
1406 	mutex_unlock(&con->mutex);
1407 	auth = con->ops->get_authorizer(con, auth_proto, con->auth_retry);
1408 	mutex_lock(&con->mutex);
1409 
1410 	if (IS_ERR(auth))
1411 		return auth;
1412 	if (con->state != CON_STATE_NEGOTIATING)
1413 		return ERR_PTR(-EAGAIN);
1414 
1415 	con->auth_reply_buf = auth->authorizer_reply_buf;
1416 	con->auth_reply_buf_len = auth->authorizer_reply_buf_len;
1417 	return auth;
1418 }
1419 
1420 /*
1421  * We connected to a peer and are saying hello.
1422  */
prepare_write_banner(struct ceph_connection * con)1423 static void prepare_write_banner(struct ceph_connection *con)
1424 {
1425 	con_out_kvec_add(con, strlen(CEPH_BANNER), CEPH_BANNER);
1426 	con_out_kvec_add(con, sizeof (con->msgr->my_enc_addr),
1427 					&con->msgr->my_enc_addr);
1428 
1429 	con->out_more = 0;
1430 	con_flag_set(con, CON_FLAG_WRITE_PENDING);
1431 }
1432 
prepare_write_connect(struct ceph_connection * con)1433 static int prepare_write_connect(struct ceph_connection *con)
1434 {
1435 	unsigned int global_seq = get_global_seq(con->msgr, 0);
1436 	int proto;
1437 	int auth_proto;
1438 	struct ceph_auth_handshake *auth;
1439 
1440 	switch (con->peer_name.type) {
1441 	case CEPH_ENTITY_TYPE_MON:
1442 		proto = CEPH_MONC_PROTOCOL;
1443 		break;
1444 	case CEPH_ENTITY_TYPE_OSD:
1445 		proto = CEPH_OSDC_PROTOCOL;
1446 		break;
1447 	case CEPH_ENTITY_TYPE_MDS:
1448 		proto = CEPH_MDSC_PROTOCOL;
1449 		break;
1450 	default:
1451 		BUG();
1452 	}
1453 
1454 	dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con,
1455 	     con->connect_seq, global_seq, proto);
1456 
1457 	con->out_connect.features = cpu_to_le64(con->msgr->supported_features);
1458 	con->out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT);
1459 	con->out_connect.connect_seq = cpu_to_le32(con->connect_seq);
1460 	con->out_connect.global_seq = cpu_to_le32(global_seq);
1461 	con->out_connect.protocol_version = cpu_to_le32(proto);
1462 	con->out_connect.flags = 0;
1463 
1464 	auth_proto = CEPH_AUTH_UNKNOWN;
1465 	auth = get_connect_authorizer(con, &auth_proto);
1466 	if (IS_ERR(auth))
1467 		return PTR_ERR(auth);
1468 
1469 	con->out_connect.authorizer_protocol = cpu_to_le32(auth_proto);
1470 	con->out_connect.authorizer_len = auth ?
1471 		cpu_to_le32(auth->authorizer_buf_len) : 0;
1472 
1473 	con_out_kvec_add(con, sizeof (con->out_connect),
1474 					&con->out_connect);
1475 	if (auth && auth->authorizer_buf_len)
1476 		con_out_kvec_add(con, auth->authorizer_buf_len,
1477 					auth->authorizer_buf);
1478 
1479 	con->out_more = 0;
1480 	con_flag_set(con, CON_FLAG_WRITE_PENDING);
1481 
1482 	return 0;
1483 }
1484 
1485 /*
1486  * write as much of pending kvecs to the socket as we can.
1487  *  1 -> done
1488  *  0 -> socket full, but more to do
1489  * <0 -> error
1490  */
write_partial_kvec(struct ceph_connection * con)1491 static int write_partial_kvec(struct ceph_connection *con)
1492 {
1493 	int ret;
1494 
1495 	dout("write_partial_kvec %p %d left\n", con, con->out_kvec_bytes);
1496 	while (con->out_kvec_bytes > 0) {
1497 		ret = ceph_tcp_sendmsg(con->sock, con->out_kvec_cur,
1498 				       con->out_kvec_left, con->out_kvec_bytes,
1499 				       con->out_more);
1500 		if (ret <= 0)
1501 			goto out;
1502 		con->out_kvec_bytes -= ret;
1503 		if (con->out_kvec_bytes == 0)
1504 			break;            /* done */
1505 
1506 		/* account for full iov entries consumed */
1507 		while (ret >= con->out_kvec_cur->iov_len) {
1508 			BUG_ON(!con->out_kvec_left);
1509 			ret -= con->out_kvec_cur->iov_len;
1510 			con->out_kvec_cur++;
1511 			con->out_kvec_left--;
1512 		}
1513 		/* and for a partially-consumed entry */
1514 		if (ret) {
1515 			con->out_kvec_cur->iov_len -= ret;
1516 			con->out_kvec_cur->iov_base += ret;
1517 		}
1518 	}
1519 	con->out_kvec_left = 0;
1520 	ret = 1;
1521 out:
1522 	dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con,
1523 	     con->out_kvec_bytes, con->out_kvec_left, ret);
1524 	return ret;  /* done! */
1525 }
1526 
ceph_crc32c_page(u32 crc,struct page * page,unsigned int page_offset,unsigned int length)1527 static u32 ceph_crc32c_page(u32 crc, struct page *page,
1528 				unsigned int page_offset,
1529 				unsigned int length)
1530 {
1531 	char *kaddr;
1532 
1533 	kaddr = kmap(page);
1534 	BUG_ON(kaddr == NULL);
1535 	crc = crc32c(crc, kaddr + page_offset, length);
1536 	kunmap(page);
1537 
1538 	return crc;
1539 }
1540 /*
1541  * Write as much message data payload as we can.  If we finish, queue
1542  * up the footer.
1543  *  1 -> done, footer is now queued in out_kvec[].
1544  *  0 -> socket full, but more to do
1545  * <0 -> error
1546  */
write_partial_message_data(struct ceph_connection * con)1547 static int write_partial_message_data(struct ceph_connection *con)
1548 {
1549 	struct ceph_msg *msg = con->out_msg;
1550 	struct ceph_msg_data_cursor *cursor = &msg->cursor;
1551 	bool do_datacrc = !con->msgr->nocrc;
1552 	u32 crc;
1553 
1554 	dout("%s %p msg %p\n", __func__, con, msg);
1555 
1556 	if (list_empty(&msg->data))
1557 		return -EINVAL;
1558 
1559 	/*
1560 	 * Iterate through each page that contains data to be
1561 	 * written, and send as much as possible for each.
1562 	 *
1563 	 * If we are calculating the data crc (the default), we will
1564 	 * need to map the page.  If we have no pages, they have
1565 	 * been revoked, so use the zero page.
1566 	 */
1567 	crc = do_datacrc ? le32_to_cpu(msg->footer.data_crc) : 0;
1568 	while (cursor->resid) {
1569 		struct page *page;
1570 		size_t page_offset;
1571 		size_t length;
1572 		bool last_piece;
1573 		bool need_crc;
1574 		int ret;
1575 
1576 		page = ceph_msg_data_next(&msg->cursor, &page_offset, &length,
1577 							&last_piece);
1578 		ret = ceph_tcp_sendpage(con->sock, page, page_offset,
1579 				      length, last_piece);
1580 		if (ret <= 0) {
1581 			if (do_datacrc)
1582 				msg->footer.data_crc = cpu_to_le32(crc);
1583 
1584 			return ret;
1585 		}
1586 		if (do_datacrc && cursor->need_crc)
1587 			crc = ceph_crc32c_page(crc, page, page_offset, length);
1588 		need_crc = ceph_msg_data_advance(&msg->cursor, (size_t)ret);
1589 	}
1590 
1591 	dout("%s %p msg %p done\n", __func__, con, msg);
1592 
1593 	/* prepare and queue up footer, too */
1594 	if (do_datacrc)
1595 		msg->footer.data_crc = cpu_to_le32(crc);
1596 	else
1597 		msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC;
1598 	con_out_kvec_reset(con);
1599 	prepare_write_message_footer(con);
1600 
1601 	return 1;	/* must return > 0 to indicate success */
1602 }
1603 
1604 /*
1605  * write some zeros
1606  */
write_partial_skip(struct ceph_connection * con)1607 static int write_partial_skip(struct ceph_connection *con)
1608 {
1609 	int ret;
1610 
1611 	dout("%s %p %d left\n", __func__, con, con->out_skip);
1612 	while (con->out_skip > 0) {
1613 		size_t size = min(con->out_skip, (int) PAGE_CACHE_SIZE);
1614 
1615 		ret = ceph_tcp_sendpage(con->sock, zero_page, 0, size, true);
1616 		if (ret <= 0)
1617 			goto out;
1618 		con->out_skip -= ret;
1619 	}
1620 	ret = 1;
1621 out:
1622 	return ret;
1623 }
1624 
1625 /*
1626  * Prepare to read connection handshake, or an ack.
1627  */
prepare_read_banner(struct ceph_connection * con)1628 static void prepare_read_banner(struct ceph_connection *con)
1629 {
1630 	dout("prepare_read_banner %p\n", con);
1631 	con->in_base_pos = 0;
1632 }
1633 
prepare_read_connect(struct ceph_connection * con)1634 static void prepare_read_connect(struct ceph_connection *con)
1635 {
1636 	dout("prepare_read_connect %p\n", con);
1637 	con->in_base_pos = 0;
1638 }
1639 
prepare_read_ack(struct ceph_connection * con)1640 static void prepare_read_ack(struct ceph_connection *con)
1641 {
1642 	dout("prepare_read_ack %p\n", con);
1643 	con->in_base_pos = 0;
1644 }
1645 
prepare_read_seq(struct ceph_connection * con)1646 static void prepare_read_seq(struct ceph_connection *con)
1647 {
1648 	dout("prepare_read_seq %p\n", con);
1649 	con->in_base_pos = 0;
1650 	con->in_tag = CEPH_MSGR_TAG_SEQ;
1651 }
1652 
prepare_read_tag(struct ceph_connection * con)1653 static void prepare_read_tag(struct ceph_connection *con)
1654 {
1655 	dout("prepare_read_tag %p\n", con);
1656 	con->in_base_pos = 0;
1657 	con->in_tag = CEPH_MSGR_TAG_READY;
1658 }
1659 
1660 /*
1661  * Prepare to read a message.
1662  */
prepare_read_message(struct ceph_connection * con)1663 static int prepare_read_message(struct ceph_connection *con)
1664 {
1665 	dout("prepare_read_message %p\n", con);
1666 	BUG_ON(con->in_msg != NULL);
1667 	con->in_base_pos = 0;
1668 	con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0;
1669 	return 0;
1670 }
1671 
1672 
read_partial(struct ceph_connection * con,int end,int size,void * object)1673 static int read_partial(struct ceph_connection *con,
1674 			int end, int size, void *object)
1675 {
1676 	while (con->in_base_pos < end) {
1677 		int left = end - con->in_base_pos;
1678 		int have = size - left;
1679 		int ret = ceph_tcp_recvmsg(con->sock, object + have, left);
1680 		if (ret <= 0)
1681 			return ret;
1682 		con->in_base_pos += ret;
1683 	}
1684 	return 1;
1685 }
1686 
1687 
1688 /*
1689  * Read all or part of the connect-side handshake on a new connection
1690  */
read_partial_banner(struct ceph_connection * con)1691 static int read_partial_banner(struct ceph_connection *con)
1692 {
1693 	int size;
1694 	int end;
1695 	int ret;
1696 
1697 	dout("read_partial_banner %p at %d\n", con, con->in_base_pos);
1698 
1699 	/* peer's banner */
1700 	size = strlen(CEPH_BANNER);
1701 	end = size;
1702 	ret = read_partial(con, end, size, con->in_banner);
1703 	if (ret <= 0)
1704 		goto out;
1705 
1706 	size = sizeof (con->actual_peer_addr);
1707 	end += size;
1708 	ret = read_partial(con, end, size, &con->actual_peer_addr);
1709 	if (ret <= 0)
1710 		goto out;
1711 
1712 	size = sizeof (con->peer_addr_for_me);
1713 	end += size;
1714 	ret = read_partial(con, end, size, &con->peer_addr_for_me);
1715 	if (ret <= 0)
1716 		goto out;
1717 
1718 out:
1719 	return ret;
1720 }
1721 
read_partial_connect(struct ceph_connection * con)1722 static int read_partial_connect(struct ceph_connection *con)
1723 {
1724 	int size;
1725 	int end;
1726 	int ret;
1727 
1728 	dout("read_partial_connect %p at %d\n", con, con->in_base_pos);
1729 
1730 	size = sizeof (con->in_reply);
1731 	end = size;
1732 	ret = read_partial(con, end, size, &con->in_reply);
1733 	if (ret <= 0)
1734 		goto out;
1735 
1736 	size = le32_to_cpu(con->in_reply.authorizer_len);
1737 	end += size;
1738 	ret = read_partial(con, end, size, con->auth_reply_buf);
1739 	if (ret <= 0)
1740 		goto out;
1741 
1742 	dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n",
1743 	     con, (int)con->in_reply.tag,
1744 	     le32_to_cpu(con->in_reply.connect_seq),
1745 	     le32_to_cpu(con->in_reply.global_seq));
1746 out:
1747 	return ret;
1748 
1749 }
1750 
1751 /*
1752  * Verify the hello banner looks okay.
1753  */
verify_hello(struct ceph_connection * con)1754 static int verify_hello(struct ceph_connection *con)
1755 {
1756 	if (memcmp(con->in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) {
1757 		pr_err("connect to %s got bad banner\n",
1758 		       ceph_pr_addr(&con->peer_addr.in_addr));
1759 		con->error_msg = "protocol error, bad banner";
1760 		return -1;
1761 	}
1762 	return 0;
1763 }
1764 
addr_is_blank(struct sockaddr_storage * ss)1765 static bool addr_is_blank(struct sockaddr_storage *ss)
1766 {
1767 	switch (ss->ss_family) {
1768 	case AF_INET:
1769 		return ((struct sockaddr_in *)ss)->sin_addr.s_addr == 0;
1770 	case AF_INET6:
1771 		return
1772 		     ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[0] == 0 &&
1773 		     ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[1] == 0 &&
1774 		     ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[2] == 0 &&
1775 		     ((struct sockaddr_in6 *)ss)->sin6_addr.s6_addr32[3] == 0;
1776 	}
1777 	return false;
1778 }
1779 
addr_port(struct sockaddr_storage * ss)1780 static int addr_port(struct sockaddr_storage *ss)
1781 {
1782 	switch (ss->ss_family) {
1783 	case AF_INET:
1784 		return ntohs(((struct sockaddr_in *)ss)->sin_port);
1785 	case AF_INET6:
1786 		return ntohs(((struct sockaddr_in6 *)ss)->sin6_port);
1787 	}
1788 	return 0;
1789 }
1790 
addr_set_port(struct sockaddr_storage * ss,int p)1791 static void addr_set_port(struct sockaddr_storage *ss, int p)
1792 {
1793 	switch (ss->ss_family) {
1794 	case AF_INET:
1795 		((struct sockaddr_in *)ss)->sin_port = htons(p);
1796 		break;
1797 	case AF_INET6:
1798 		((struct sockaddr_in6 *)ss)->sin6_port = htons(p);
1799 		break;
1800 	}
1801 }
1802 
1803 /*
1804  * Unlike other *_pton function semantics, zero indicates success.
1805  */
ceph_pton(const char * str,size_t len,struct sockaddr_storage * ss,char delim,const char ** ipend)1806 static int ceph_pton(const char *str, size_t len, struct sockaddr_storage *ss,
1807 		char delim, const char **ipend)
1808 {
1809 	struct sockaddr_in *in4 = (struct sockaddr_in *) ss;
1810 	struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) ss;
1811 
1812 	memset(ss, 0, sizeof(*ss));
1813 
1814 	if (in4_pton(str, len, (u8 *)&in4->sin_addr.s_addr, delim, ipend)) {
1815 		ss->ss_family = AF_INET;
1816 		return 0;
1817 	}
1818 
1819 	if (in6_pton(str, len, (u8 *)&in6->sin6_addr.s6_addr, delim, ipend)) {
1820 		ss->ss_family = AF_INET6;
1821 		return 0;
1822 	}
1823 
1824 	return -EINVAL;
1825 }
1826 
1827 /*
1828  * Extract hostname string and resolve using kernel DNS facility.
1829  */
1830 #ifdef CONFIG_CEPH_LIB_USE_DNS_RESOLVER
ceph_dns_resolve_name(const char * name,size_t namelen,struct sockaddr_storage * ss,char delim,const char ** ipend)1831 static int ceph_dns_resolve_name(const char *name, size_t namelen,
1832 		struct sockaddr_storage *ss, char delim, const char **ipend)
1833 {
1834 	const char *end, *delim_p;
1835 	char *colon_p, *ip_addr = NULL;
1836 	int ip_len, ret;
1837 
1838 	/*
1839 	 * The end of the hostname occurs immediately preceding the delimiter or
1840 	 * the port marker (':') where the delimiter takes precedence.
1841 	 */
1842 	delim_p = memchr(name, delim, namelen);
1843 	colon_p = memchr(name, ':', namelen);
1844 
1845 	if (delim_p && colon_p)
1846 		end = delim_p < colon_p ? delim_p : colon_p;
1847 	else if (!delim_p && colon_p)
1848 		end = colon_p;
1849 	else {
1850 		end = delim_p;
1851 		if (!end) /* case: hostname:/ */
1852 			end = name + namelen;
1853 	}
1854 
1855 	if (end <= name)
1856 		return -EINVAL;
1857 
1858 	/* do dns_resolve upcall */
1859 	ip_len = dns_query(NULL, name, end - name, NULL, &ip_addr, NULL);
1860 	if (ip_len > 0)
1861 		ret = ceph_pton(ip_addr, ip_len, ss, -1, NULL);
1862 	else
1863 		ret = -ESRCH;
1864 
1865 	kfree(ip_addr);
1866 
1867 	*ipend = end;
1868 
1869 	pr_info("resolve '%.*s' (ret=%d): %s\n", (int)(end - name), name,
1870 			ret, ret ? "failed" : ceph_pr_addr(ss));
1871 
1872 	return ret;
1873 }
1874 #else
ceph_dns_resolve_name(const char * name,size_t namelen,struct sockaddr_storage * ss,char delim,const char ** ipend)1875 static inline int ceph_dns_resolve_name(const char *name, size_t namelen,
1876 		struct sockaddr_storage *ss, char delim, const char **ipend)
1877 {
1878 	return -EINVAL;
1879 }
1880 #endif
1881 
1882 /*
1883  * Parse a server name (IP or hostname). If a valid IP address is not found
1884  * then try to extract a hostname to resolve using userspace DNS upcall.
1885  */
ceph_parse_server_name(const char * name,size_t namelen,struct sockaddr_storage * ss,char delim,const char ** ipend)1886 static int ceph_parse_server_name(const char *name, size_t namelen,
1887 			struct sockaddr_storage *ss, char delim, const char **ipend)
1888 {
1889 	int ret;
1890 
1891 	ret = ceph_pton(name, namelen, ss, delim, ipend);
1892 	if (ret)
1893 		ret = ceph_dns_resolve_name(name, namelen, ss, delim, ipend);
1894 
1895 	return ret;
1896 }
1897 
1898 /*
1899  * Parse an ip[:port] list into an addr array.  Use the default
1900  * monitor port if a port isn't specified.
1901  */
ceph_parse_ips(const char * c,const char * end,struct ceph_entity_addr * addr,int max_count,int * count)1902 int ceph_parse_ips(const char *c, const char *end,
1903 		   struct ceph_entity_addr *addr,
1904 		   int max_count, int *count)
1905 {
1906 	int i, ret = -EINVAL;
1907 	const char *p = c;
1908 
1909 	dout("parse_ips on '%.*s'\n", (int)(end-c), c);
1910 	for (i = 0; i < max_count; i++) {
1911 		const char *ipend;
1912 		struct sockaddr_storage *ss = &addr[i].in_addr;
1913 		int port;
1914 		char delim = ',';
1915 
1916 		if (*p == '[') {
1917 			delim = ']';
1918 			p++;
1919 		}
1920 
1921 		ret = ceph_parse_server_name(p, end - p, ss, delim, &ipend);
1922 		if (ret)
1923 			goto bad;
1924 		ret = -EINVAL;
1925 
1926 		p = ipend;
1927 
1928 		if (delim == ']') {
1929 			if (*p != ']') {
1930 				dout("missing matching ']'\n");
1931 				goto bad;
1932 			}
1933 			p++;
1934 		}
1935 
1936 		/* port? */
1937 		if (p < end && *p == ':') {
1938 			port = 0;
1939 			p++;
1940 			while (p < end && *p >= '0' && *p <= '9') {
1941 				port = (port * 10) + (*p - '0');
1942 				p++;
1943 			}
1944 			if (port == 0)
1945 				port = CEPH_MON_PORT;
1946 			else if (port > 65535)
1947 				goto bad;
1948 		} else {
1949 			port = CEPH_MON_PORT;
1950 		}
1951 
1952 		addr_set_port(ss, port);
1953 
1954 		dout("parse_ips got %s\n", ceph_pr_addr(ss));
1955 
1956 		if (p == end)
1957 			break;
1958 		if (*p != ',')
1959 			goto bad;
1960 		p++;
1961 	}
1962 
1963 	if (p != end)
1964 		goto bad;
1965 
1966 	if (count)
1967 		*count = i + 1;
1968 	return 0;
1969 
1970 bad:
1971 	pr_err("parse_ips bad ip '%.*s'\n", (int)(end - c), c);
1972 	return ret;
1973 }
1974 EXPORT_SYMBOL(ceph_parse_ips);
1975 
process_banner(struct ceph_connection * con)1976 static int process_banner(struct ceph_connection *con)
1977 {
1978 	dout("process_banner on %p\n", con);
1979 
1980 	if (verify_hello(con) < 0)
1981 		return -1;
1982 
1983 	ceph_decode_addr(&con->actual_peer_addr);
1984 	ceph_decode_addr(&con->peer_addr_for_me);
1985 
1986 	/*
1987 	 * Make sure the other end is who we wanted.  note that the other
1988 	 * end may not yet know their ip address, so if it's 0.0.0.0, give
1989 	 * them the benefit of the doubt.
1990 	 */
1991 	if (memcmp(&con->peer_addr, &con->actual_peer_addr,
1992 		   sizeof(con->peer_addr)) != 0 &&
1993 	    !(addr_is_blank(&con->actual_peer_addr.in_addr) &&
1994 	      con->actual_peer_addr.nonce == con->peer_addr.nonce)) {
1995 		pr_warn("wrong peer, want %s/%d, got %s/%d\n",
1996 			ceph_pr_addr(&con->peer_addr.in_addr),
1997 			(int)le32_to_cpu(con->peer_addr.nonce),
1998 			ceph_pr_addr(&con->actual_peer_addr.in_addr),
1999 			(int)le32_to_cpu(con->actual_peer_addr.nonce));
2000 		con->error_msg = "wrong peer at address";
2001 		return -1;
2002 	}
2003 
2004 	/*
2005 	 * did we learn our address?
2006 	 */
2007 	if (addr_is_blank(&con->msgr->inst.addr.in_addr)) {
2008 		int port = addr_port(&con->msgr->inst.addr.in_addr);
2009 
2010 		memcpy(&con->msgr->inst.addr.in_addr,
2011 		       &con->peer_addr_for_me.in_addr,
2012 		       sizeof(con->peer_addr_for_me.in_addr));
2013 		addr_set_port(&con->msgr->inst.addr.in_addr, port);
2014 		encode_my_addr(con->msgr);
2015 		dout("process_banner learned my addr is %s\n",
2016 		     ceph_pr_addr(&con->msgr->inst.addr.in_addr));
2017 	}
2018 
2019 	return 0;
2020 }
2021 
process_connect(struct ceph_connection * con)2022 static int process_connect(struct ceph_connection *con)
2023 {
2024 	u64 sup_feat = con->msgr->supported_features;
2025 	u64 req_feat = con->msgr->required_features;
2026 	u64 server_feat = ceph_sanitize_features(
2027 				le64_to_cpu(con->in_reply.features));
2028 	int ret;
2029 
2030 	dout("process_connect on %p tag %d\n", con, (int)con->in_tag);
2031 
2032 	switch (con->in_reply.tag) {
2033 	case CEPH_MSGR_TAG_FEATURES:
2034 		pr_err("%s%lld %s feature set mismatch,"
2035 		       " my %llx < server's %llx, missing %llx\n",
2036 		       ENTITY_NAME(con->peer_name),
2037 		       ceph_pr_addr(&con->peer_addr.in_addr),
2038 		       sup_feat, server_feat, server_feat & ~sup_feat);
2039 		con->error_msg = "missing required protocol features";
2040 		reset_connection(con);
2041 		return -1;
2042 
2043 	case CEPH_MSGR_TAG_BADPROTOVER:
2044 		pr_err("%s%lld %s protocol version mismatch,"
2045 		       " my %d != server's %d\n",
2046 		       ENTITY_NAME(con->peer_name),
2047 		       ceph_pr_addr(&con->peer_addr.in_addr),
2048 		       le32_to_cpu(con->out_connect.protocol_version),
2049 		       le32_to_cpu(con->in_reply.protocol_version));
2050 		con->error_msg = "protocol version mismatch";
2051 		reset_connection(con);
2052 		return -1;
2053 
2054 	case CEPH_MSGR_TAG_BADAUTHORIZER:
2055 		con->auth_retry++;
2056 		dout("process_connect %p got BADAUTHORIZER attempt %d\n", con,
2057 		     con->auth_retry);
2058 		if (con->auth_retry == 2) {
2059 			con->error_msg = "connect authorization failure";
2060 			return -1;
2061 		}
2062 		con_out_kvec_reset(con);
2063 		ret = prepare_write_connect(con);
2064 		if (ret < 0)
2065 			return ret;
2066 		prepare_read_connect(con);
2067 		break;
2068 
2069 	case CEPH_MSGR_TAG_RESETSESSION:
2070 		/*
2071 		 * If we connected with a large connect_seq but the peer
2072 		 * has no record of a session with us (no connection, or
2073 		 * connect_seq == 0), they will send RESETSESION to indicate
2074 		 * that they must have reset their session, and may have
2075 		 * dropped messages.
2076 		 */
2077 		dout("process_connect got RESET peer seq %u\n",
2078 		     le32_to_cpu(con->in_reply.connect_seq));
2079 		pr_err("%s%lld %s connection reset\n",
2080 		       ENTITY_NAME(con->peer_name),
2081 		       ceph_pr_addr(&con->peer_addr.in_addr));
2082 		reset_connection(con);
2083 		con_out_kvec_reset(con);
2084 		ret = prepare_write_connect(con);
2085 		if (ret < 0)
2086 			return ret;
2087 		prepare_read_connect(con);
2088 
2089 		/* Tell ceph about it. */
2090 		mutex_unlock(&con->mutex);
2091 		pr_info("reset on %s%lld\n", ENTITY_NAME(con->peer_name));
2092 		if (con->ops->peer_reset)
2093 			con->ops->peer_reset(con);
2094 		mutex_lock(&con->mutex);
2095 		if (con->state != CON_STATE_NEGOTIATING)
2096 			return -EAGAIN;
2097 		break;
2098 
2099 	case CEPH_MSGR_TAG_RETRY_SESSION:
2100 		/*
2101 		 * If we sent a smaller connect_seq than the peer has, try
2102 		 * again with a larger value.
2103 		 */
2104 		dout("process_connect got RETRY_SESSION my seq %u, peer %u\n",
2105 		     le32_to_cpu(con->out_connect.connect_seq),
2106 		     le32_to_cpu(con->in_reply.connect_seq));
2107 		con->connect_seq = le32_to_cpu(con->in_reply.connect_seq);
2108 		con_out_kvec_reset(con);
2109 		ret = prepare_write_connect(con);
2110 		if (ret < 0)
2111 			return ret;
2112 		prepare_read_connect(con);
2113 		break;
2114 
2115 	case CEPH_MSGR_TAG_RETRY_GLOBAL:
2116 		/*
2117 		 * If we sent a smaller global_seq than the peer has, try
2118 		 * again with a larger value.
2119 		 */
2120 		dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n",
2121 		     con->peer_global_seq,
2122 		     le32_to_cpu(con->in_reply.global_seq));
2123 		get_global_seq(con->msgr,
2124 			       le32_to_cpu(con->in_reply.global_seq));
2125 		con_out_kvec_reset(con);
2126 		ret = prepare_write_connect(con);
2127 		if (ret < 0)
2128 			return ret;
2129 		prepare_read_connect(con);
2130 		break;
2131 
2132 	case CEPH_MSGR_TAG_SEQ:
2133 	case CEPH_MSGR_TAG_READY:
2134 		if (req_feat & ~server_feat) {
2135 			pr_err("%s%lld %s protocol feature mismatch,"
2136 			       " my required %llx > server's %llx, need %llx\n",
2137 			       ENTITY_NAME(con->peer_name),
2138 			       ceph_pr_addr(&con->peer_addr.in_addr),
2139 			       req_feat, server_feat, req_feat & ~server_feat);
2140 			con->error_msg = "missing required protocol features";
2141 			reset_connection(con);
2142 			return -1;
2143 		}
2144 
2145 		WARN_ON(con->state != CON_STATE_NEGOTIATING);
2146 		con->state = CON_STATE_OPEN;
2147 		con->auth_retry = 0;    /* we authenticated; clear flag */
2148 		con->peer_global_seq = le32_to_cpu(con->in_reply.global_seq);
2149 		con->connect_seq++;
2150 		con->peer_features = server_feat;
2151 		dout("process_connect got READY gseq %d cseq %d (%d)\n",
2152 		     con->peer_global_seq,
2153 		     le32_to_cpu(con->in_reply.connect_seq),
2154 		     con->connect_seq);
2155 		WARN_ON(con->connect_seq !=
2156 			le32_to_cpu(con->in_reply.connect_seq));
2157 
2158 		if (con->in_reply.flags & CEPH_MSG_CONNECT_LOSSY)
2159 			con_flag_set(con, CON_FLAG_LOSSYTX);
2160 
2161 		con->delay = 0;      /* reset backoff memory */
2162 
2163 		if (con->in_reply.tag == CEPH_MSGR_TAG_SEQ) {
2164 			prepare_write_seq(con);
2165 			prepare_read_seq(con);
2166 		} else {
2167 			prepare_read_tag(con);
2168 		}
2169 		break;
2170 
2171 	case CEPH_MSGR_TAG_WAIT:
2172 		/*
2173 		 * If there is a connection race (we are opening
2174 		 * connections to each other), one of us may just have
2175 		 * to WAIT.  This shouldn't happen if we are the
2176 		 * client.
2177 		 */
2178 		con->error_msg = "protocol error, got WAIT as client";
2179 		return -1;
2180 
2181 	default:
2182 		con->error_msg = "protocol error, garbage tag during connect";
2183 		return -1;
2184 	}
2185 	return 0;
2186 }
2187 
2188 
2189 /*
2190  * read (part of) an ack
2191  */
read_partial_ack(struct ceph_connection * con)2192 static int read_partial_ack(struct ceph_connection *con)
2193 {
2194 	int size = sizeof (con->in_temp_ack);
2195 	int end = size;
2196 
2197 	return read_partial(con, end, size, &con->in_temp_ack);
2198 }
2199 
2200 /*
2201  * We can finally discard anything that's been acked.
2202  */
process_ack(struct ceph_connection * con)2203 static void process_ack(struct ceph_connection *con)
2204 {
2205 	struct ceph_msg *m;
2206 	u64 ack = le64_to_cpu(con->in_temp_ack);
2207 	u64 seq;
2208 
2209 	while (!list_empty(&con->out_sent)) {
2210 		m = list_first_entry(&con->out_sent, struct ceph_msg,
2211 				     list_head);
2212 		seq = le64_to_cpu(m->hdr.seq);
2213 		if (seq > ack)
2214 			break;
2215 		dout("got ack for seq %llu type %d at %p\n", seq,
2216 		     le16_to_cpu(m->hdr.type), m);
2217 		m->ack_stamp = jiffies;
2218 		ceph_msg_remove(m);
2219 	}
2220 	prepare_read_tag(con);
2221 }
2222 
2223 
read_partial_message_section(struct ceph_connection * con,struct kvec * section,unsigned int sec_len,u32 * crc)2224 static int read_partial_message_section(struct ceph_connection *con,
2225 					struct kvec *section,
2226 					unsigned int sec_len, u32 *crc)
2227 {
2228 	int ret, left;
2229 
2230 	BUG_ON(!section);
2231 
2232 	while (section->iov_len < sec_len) {
2233 		BUG_ON(section->iov_base == NULL);
2234 		left = sec_len - section->iov_len;
2235 		ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base +
2236 				       section->iov_len, left);
2237 		if (ret <= 0)
2238 			return ret;
2239 		section->iov_len += ret;
2240 	}
2241 	if (section->iov_len == sec_len)
2242 		*crc = crc32c(0, section->iov_base, section->iov_len);
2243 
2244 	return 1;
2245 }
2246 
read_partial_msg_data(struct ceph_connection * con)2247 static int read_partial_msg_data(struct ceph_connection *con)
2248 {
2249 	struct ceph_msg *msg = con->in_msg;
2250 	struct ceph_msg_data_cursor *cursor = &msg->cursor;
2251 	const bool do_datacrc = !con->msgr->nocrc;
2252 	struct page *page;
2253 	size_t page_offset;
2254 	size_t length;
2255 	u32 crc = 0;
2256 	int ret;
2257 
2258 	BUG_ON(!msg);
2259 	if (list_empty(&msg->data))
2260 		return -EIO;
2261 
2262 	if (do_datacrc)
2263 		crc = con->in_data_crc;
2264 	while (cursor->resid) {
2265 		page = ceph_msg_data_next(&msg->cursor, &page_offset, &length,
2266 							NULL);
2267 		ret = ceph_tcp_recvpage(con->sock, page, page_offset, length);
2268 		if (ret <= 0) {
2269 			if (do_datacrc)
2270 				con->in_data_crc = crc;
2271 
2272 			return ret;
2273 		}
2274 
2275 		if (do_datacrc)
2276 			crc = ceph_crc32c_page(crc, page, page_offset, ret);
2277 		(void) ceph_msg_data_advance(&msg->cursor, (size_t)ret);
2278 	}
2279 	if (do_datacrc)
2280 		con->in_data_crc = crc;
2281 
2282 	return 1;	/* must return > 0 to indicate success */
2283 }
2284 
2285 /*
2286  * read (part of) a message.
2287  */
2288 static int ceph_con_in_msg_alloc(struct ceph_connection *con, int *skip);
2289 
read_partial_message(struct ceph_connection * con)2290 static int read_partial_message(struct ceph_connection *con)
2291 {
2292 	struct ceph_msg *m = con->in_msg;
2293 	int size;
2294 	int end;
2295 	int ret;
2296 	unsigned int front_len, middle_len, data_len;
2297 	bool do_datacrc = !con->msgr->nocrc;
2298 	bool need_sign = (con->peer_features & CEPH_FEATURE_MSG_AUTH);
2299 	u64 seq;
2300 	u32 crc;
2301 
2302 	dout("read_partial_message con %p msg %p\n", con, m);
2303 
2304 	/* header */
2305 	size = sizeof (con->in_hdr);
2306 	end = size;
2307 	ret = read_partial(con, end, size, &con->in_hdr);
2308 	if (ret <= 0)
2309 		return ret;
2310 
2311 	crc = crc32c(0, &con->in_hdr, offsetof(struct ceph_msg_header, crc));
2312 	if (cpu_to_le32(crc) != con->in_hdr.crc) {
2313 		pr_err("read_partial_message bad hdr crc %u != expected %u\n",
2314 		       crc, con->in_hdr.crc);
2315 		return -EBADMSG;
2316 	}
2317 
2318 	front_len = le32_to_cpu(con->in_hdr.front_len);
2319 	if (front_len > CEPH_MSG_MAX_FRONT_LEN)
2320 		return -EIO;
2321 	middle_len = le32_to_cpu(con->in_hdr.middle_len);
2322 	if (middle_len > CEPH_MSG_MAX_MIDDLE_LEN)
2323 		return -EIO;
2324 	data_len = le32_to_cpu(con->in_hdr.data_len);
2325 	if (data_len > CEPH_MSG_MAX_DATA_LEN)
2326 		return -EIO;
2327 
2328 	/* verify seq# */
2329 	seq = le64_to_cpu(con->in_hdr.seq);
2330 	if ((s64)seq - (s64)con->in_seq < 1) {
2331 		pr_info("skipping %s%lld %s seq %lld expected %lld\n",
2332 			ENTITY_NAME(con->peer_name),
2333 			ceph_pr_addr(&con->peer_addr.in_addr),
2334 			seq, con->in_seq + 1);
2335 		con->in_base_pos = -front_len - middle_len - data_len -
2336 			sizeof_footer(con);
2337 		con->in_tag = CEPH_MSGR_TAG_READY;
2338 		return 1;
2339 	} else if ((s64)seq - (s64)con->in_seq > 1) {
2340 		pr_err("read_partial_message bad seq %lld expected %lld\n",
2341 		       seq, con->in_seq + 1);
2342 		con->error_msg = "bad message sequence # for incoming message";
2343 		return -EBADE;
2344 	}
2345 
2346 	/* allocate message? */
2347 	if (!con->in_msg) {
2348 		int skip = 0;
2349 
2350 		dout("got hdr type %d front %d data %d\n", con->in_hdr.type,
2351 		     front_len, data_len);
2352 		ret = ceph_con_in_msg_alloc(con, &skip);
2353 		if (ret < 0)
2354 			return ret;
2355 
2356 		BUG_ON(!con->in_msg ^ skip);
2357 		if (con->in_msg && data_len > con->in_msg->data_length) {
2358 			pr_warn("%s skipping long message (%u > %zd)\n",
2359 				__func__, data_len, con->in_msg->data_length);
2360 			ceph_msg_put(con->in_msg);
2361 			con->in_msg = NULL;
2362 			skip = 1;
2363 		}
2364 		if (skip) {
2365 			/* skip this message */
2366 			dout("alloc_msg said skip message\n");
2367 			con->in_base_pos = -front_len - middle_len - data_len -
2368 				sizeof_footer(con);
2369 			con->in_tag = CEPH_MSGR_TAG_READY;
2370 			con->in_seq++;
2371 			return 1;
2372 		}
2373 
2374 		BUG_ON(!con->in_msg);
2375 		BUG_ON(con->in_msg->con != con);
2376 		m = con->in_msg;
2377 		m->front.iov_len = 0;    /* haven't read it yet */
2378 		if (m->middle)
2379 			m->middle->vec.iov_len = 0;
2380 
2381 		/* prepare for data payload, if any */
2382 
2383 		if (data_len)
2384 			prepare_message_data(con->in_msg, data_len);
2385 	}
2386 
2387 	/* front */
2388 	ret = read_partial_message_section(con, &m->front, front_len,
2389 					   &con->in_front_crc);
2390 	if (ret <= 0)
2391 		return ret;
2392 
2393 	/* middle */
2394 	if (m->middle) {
2395 		ret = read_partial_message_section(con, &m->middle->vec,
2396 						   middle_len,
2397 						   &con->in_middle_crc);
2398 		if (ret <= 0)
2399 			return ret;
2400 	}
2401 
2402 	/* (page) data */
2403 	if (data_len) {
2404 		ret = read_partial_msg_data(con);
2405 		if (ret <= 0)
2406 			return ret;
2407 	}
2408 
2409 	/* footer */
2410 	if (need_sign)
2411 		size = sizeof(m->footer);
2412 	else
2413 		size = sizeof(m->old_footer);
2414 
2415 	end += size;
2416 	ret = read_partial(con, end, size, &m->footer);
2417 	if (ret <= 0)
2418 		return ret;
2419 
2420 	if (!need_sign) {
2421 		m->footer.flags = m->old_footer.flags;
2422 		m->footer.sig = 0;
2423 	}
2424 
2425 	dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n",
2426 	     m, front_len, m->footer.front_crc, middle_len,
2427 	     m->footer.middle_crc, data_len, m->footer.data_crc);
2428 
2429 	/* crc ok? */
2430 	if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) {
2431 		pr_err("read_partial_message %p front crc %u != exp. %u\n",
2432 		       m, con->in_front_crc, m->footer.front_crc);
2433 		return -EBADMSG;
2434 	}
2435 	if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) {
2436 		pr_err("read_partial_message %p middle crc %u != exp %u\n",
2437 		       m, con->in_middle_crc, m->footer.middle_crc);
2438 		return -EBADMSG;
2439 	}
2440 	if (do_datacrc &&
2441 	    (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 &&
2442 	    con->in_data_crc != le32_to_cpu(m->footer.data_crc)) {
2443 		pr_err("read_partial_message %p data crc %u != exp. %u\n", m,
2444 		       con->in_data_crc, le32_to_cpu(m->footer.data_crc));
2445 		return -EBADMSG;
2446 	}
2447 
2448 	if (need_sign && con->ops->check_message_signature &&
2449 	    con->ops->check_message_signature(con, m)) {
2450 		pr_err("read_partial_message %p signature check failed\n", m);
2451 		return -EBADMSG;
2452 	}
2453 
2454 	return 1; /* done! */
2455 }
2456 
2457 /*
2458  * Process message.  This happens in the worker thread.  The callback should
2459  * be careful not to do anything that waits on other incoming messages or it
2460  * may deadlock.
2461  */
process_message(struct ceph_connection * con)2462 static void process_message(struct ceph_connection *con)
2463 {
2464 	struct ceph_msg *msg;
2465 
2466 	BUG_ON(con->in_msg->con != con);
2467 	con->in_msg->con = NULL;
2468 	msg = con->in_msg;
2469 	con->in_msg = NULL;
2470 	con->ops->put(con);
2471 
2472 	/* if first message, set peer_name */
2473 	if (con->peer_name.type == 0)
2474 		con->peer_name = msg->hdr.src;
2475 
2476 	con->in_seq++;
2477 	mutex_unlock(&con->mutex);
2478 
2479 	dout("===== %p %llu from %s%lld %d=%s len %d+%d (%u %u %u) =====\n",
2480 	     msg, le64_to_cpu(msg->hdr.seq),
2481 	     ENTITY_NAME(msg->hdr.src),
2482 	     le16_to_cpu(msg->hdr.type),
2483 	     ceph_msg_type_name(le16_to_cpu(msg->hdr.type)),
2484 	     le32_to_cpu(msg->hdr.front_len),
2485 	     le32_to_cpu(msg->hdr.data_len),
2486 	     con->in_front_crc, con->in_middle_crc, con->in_data_crc);
2487 	con->ops->dispatch(con, msg);
2488 
2489 	mutex_lock(&con->mutex);
2490 }
2491 
2492 
2493 /*
2494  * Write something to the socket.  Called in a worker thread when the
2495  * socket appears to be writeable and we have something ready to send.
2496  */
try_write(struct ceph_connection * con)2497 static int try_write(struct ceph_connection *con)
2498 {
2499 	int ret = 1;
2500 
2501 	dout("try_write start %p state %lu\n", con, con->state);
2502 
2503 more:
2504 	dout("try_write out_kvec_bytes %d\n", con->out_kvec_bytes);
2505 
2506 	/* open the socket first? */
2507 	if (con->state == CON_STATE_PREOPEN) {
2508 		BUG_ON(con->sock);
2509 		con->state = CON_STATE_CONNECTING;
2510 
2511 		con_out_kvec_reset(con);
2512 		prepare_write_banner(con);
2513 		prepare_read_banner(con);
2514 
2515 		BUG_ON(con->in_msg);
2516 		con->in_tag = CEPH_MSGR_TAG_READY;
2517 		dout("try_write initiating connect on %p new state %lu\n",
2518 		     con, con->state);
2519 		ret = ceph_tcp_connect(con);
2520 		if (ret < 0) {
2521 			con->error_msg = "connect error";
2522 			goto out;
2523 		}
2524 	}
2525 
2526 more_kvec:
2527 	/* kvec data queued? */
2528 	if (con->out_kvec_left) {
2529 		ret = write_partial_kvec(con);
2530 		if (ret <= 0)
2531 			goto out;
2532 	}
2533 	if (con->out_skip) {
2534 		ret = write_partial_skip(con);
2535 		if (ret <= 0)
2536 			goto out;
2537 	}
2538 
2539 	/* msg pages? */
2540 	if (con->out_msg) {
2541 		if (con->out_msg_done) {
2542 			ceph_msg_put(con->out_msg);
2543 			con->out_msg = NULL;   /* we're done with this one */
2544 			goto do_next;
2545 		}
2546 
2547 		ret = write_partial_message_data(con);
2548 		if (ret == 1)
2549 			goto more_kvec;  /* we need to send the footer, too! */
2550 		if (ret == 0)
2551 			goto out;
2552 		if (ret < 0) {
2553 			dout("try_write write_partial_message_data err %d\n",
2554 			     ret);
2555 			goto out;
2556 		}
2557 	}
2558 
2559 do_next:
2560 	if (con->state == CON_STATE_OPEN) {
2561 		/* is anything else pending? */
2562 		if (!list_empty(&con->out_queue)) {
2563 			prepare_write_message(con);
2564 			goto more;
2565 		}
2566 		if (con->in_seq > con->in_seq_acked) {
2567 			prepare_write_ack(con);
2568 			goto more;
2569 		}
2570 		if (con_flag_test_and_clear(con, CON_FLAG_KEEPALIVE_PENDING)) {
2571 			prepare_write_keepalive(con);
2572 			goto more;
2573 		}
2574 	}
2575 
2576 	/* Nothing to do! */
2577 	con_flag_clear(con, CON_FLAG_WRITE_PENDING);
2578 	dout("try_write nothing else to write.\n");
2579 	ret = 0;
2580 out:
2581 	dout("try_write done on %p ret %d\n", con, ret);
2582 	return ret;
2583 }
2584 
2585 
2586 
2587 /*
2588  * Read what we can from the socket.
2589  */
try_read(struct ceph_connection * con)2590 static int try_read(struct ceph_connection *con)
2591 {
2592 	int ret = -1;
2593 
2594 more:
2595 	dout("try_read start on %p state %lu\n", con, con->state);
2596 	if (con->state != CON_STATE_CONNECTING &&
2597 	    con->state != CON_STATE_NEGOTIATING &&
2598 	    con->state != CON_STATE_OPEN)
2599 		return 0;
2600 
2601 	BUG_ON(!con->sock);
2602 
2603 	dout("try_read tag %d in_base_pos %d\n", (int)con->in_tag,
2604 	     con->in_base_pos);
2605 
2606 	if (con->state == CON_STATE_CONNECTING) {
2607 		dout("try_read connecting\n");
2608 		ret = read_partial_banner(con);
2609 		if (ret <= 0)
2610 			goto out;
2611 		ret = process_banner(con);
2612 		if (ret < 0)
2613 			goto out;
2614 
2615 		con->state = CON_STATE_NEGOTIATING;
2616 
2617 		/*
2618 		 * Received banner is good, exchange connection info.
2619 		 * Do not reset out_kvec, as sending our banner raced
2620 		 * with receiving peer banner after connect completed.
2621 		 */
2622 		ret = prepare_write_connect(con);
2623 		if (ret < 0)
2624 			goto out;
2625 		prepare_read_connect(con);
2626 
2627 		/* Send connection info before awaiting response */
2628 		goto out;
2629 	}
2630 
2631 	if (con->state == CON_STATE_NEGOTIATING) {
2632 		dout("try_read negotiating\n");
2633 		ret = read_partial_connect(con);
2634 		if (ret <= 0)
2635 			goto out;
2636 		ret = process_connect(con);
2637 		if (ret < 0)
2638 			goto out;
2639 		goto more;
2640 	}
2641 
2642 	WARN_ON(con->state != CON_STATE_OPEN);
2643 
2644 	if (con->in_base_pos < 0) {
2645 		/*
2646 		 * skipping + discarding content.
2647 		 *
2648 		 * FIXME: there must be a better way to do this!
2649 		 */
2650 		static char buf[SKIP_BUF_SIZE];
2651 		int skip = min((int) sizeof (buf), -con->in_base_pos);
2652 
2653 		dout("skipping %d / %d bytes\n", skip, -con->in_base_pos);
2654 		ret = ceph_tcp_recvmsg(con->sock, buf, skip);
2655 		if (ret <= 0)
2656 			goto out;
2657 		con->in_base_pos += ret;
2658 		if (con->in_base_pos)
2659 			goto more;
2660 	}
2661 	if (con->in_tag == CEPH_MSGR_TAG_READY) {
2662 		/*
2663 		 * what's next?
2664 		 */
2665 		ret = ceph_tcp_recvmsg(con->sock, &con->in_tag, 1);
2666 		if (ret <= 0)
2667 			goto out;
2668 		dout("try_read got tag %d\n", (int)con->in_tag);
2669 		switch (con->in_tag) {
2670 		case CEPH_MSGR_TAG_MSG:
2671 			prepare_read_message(con);
2672 			break;
2673 		case CEPH_MSGR_TAG_ACK:
2674 			prepare_read_ack(con);
2675 			break;
2676 		case CEPH_MSGR_TAG_CLOSE:
2677 			con_close_socket(con);
2678 			con->state = CON_STATE_CLOSED;
2679 			goto out;
2680 		default:
2681 			goto bad_tag;
2682 		}
2683 	}
2684 	if (con->in_tag == CEPH_MSGR_TAG_MSG) {
2685 		ret = read_partial_message(con);
2686 		if (ret <= 0) {
2687 			switch (ret) {
2688 			case -EBADMSG:
2689 				con->error_msg = "bad crc";
2690 				/* fall through */
2691 			case -EBADE:
2692 				ret = -EIO;
2693 				break;
2694 			case -EIO:
2695 				con->error_msg = "io error";
2696 				break;
2697 			}
2698 			goto out;
2699 		}
2700 		if (con->in_tag == CEPH_MSGR_TAG_READY)
2701 			goto more;
2702 		process_message(con);
2703 		if (con->state == CON_STATE_OPEN)
2704 			prepare_read_tag(con);
2705 		goto more;
2706 	}
2707 	if (con->in_tag == CEPH_MSGR_TAG_ACK ||
2708 	    con->in_tag == CEPH_MSGR_TAG_SEQ) {
2709 		/*
2710 		 * the final handshake seq exchange is semantically
2711 		 * equivalent to an ACK
2712 		 */
2713 		ret = read_partial_ack(con);
2714 		if (ret <= 0)
2715 			goto out;
2716 		process_ack(con);
2717 		goto more;
2718 	}
2719 
2720 out:
2721 	dout("try_read done on %p ret %d\n", con, ret);
2722 	return ret;
2723 
2724 bad_tag:
2725 	pr_err("try_read bad con->in_tag = %d\n", (int)con->in_tag);
2726 	con->error_msg = "protocol error, garbage tag";
2727 	ret = -1;
2728 	goto out;
2729 }
2730 
2731 
2732 /*
2733  * Atomically queue work on a connection after the specified delay.
2734  * Bump @con reference to avoid races with connection teardown.
2735  * Returns 0 if work was queued, or an error code otherwise.
2736  */
queue_con_delay(struct ceph_connection * con,unsigned long delay)2737 static int queue_con_delay(struct ceph_connection *con, unsigned long delay)
2738 {
2739 	if (!con->ops->get(con)) {
2740 		dout("%s %p ref count 0\n", __func__, con);
2741 		return -ENOENT;
2742 	}
2743 
2744 	if (!queue_delayed_work(ceph_msgr_wq, &con->work, delay)) {
2745 		dout("%s %p - already queued\n", __func__, con);
2746 		con->ops->put(con);
2747 		return -EBUSY;
2748 	}
2749 
2750 	dout("%s %p %lu\n", __func__, con, delay);
2751 	return 0;
2752 }
2753 
queue_con(struct ceph_connection * con)2754 static void queue_con(struct ceph_connection *con)
2755 {
2756 	(void) queue_con_delay(con, 0);
2757 }
2758 
cancel_con(struct ceph_connection * con)2759 static void cancel_con(struct ceph_connection *con)
2760 {
2761 	if (cancel_delayed_work(&con->work)) {
2762 		dout("%s %p\n", __func__, con);
2763 		con->ops->put(con);
2764 	}
2765 }
2766 
con_sock_closed(struct ceph_connection * con)2767 static bool con_sock_closed(struct ceph_connection *con)
2768 {
2769 	if (!con_flag_test_and_clear(con, CON_FLAG_SOCK_CLOSED))
2770 		return false;
2771 
2772 #define CASE(x)								\
2773 	case CON_STATE_ ## x:						\
2774 		con->error_msg = "socket closed (con state " #x ")";	\
2775 		break;
2776 
2777 	switch (con->state) {
2778 	CASE(CLOSED);
2779 	CASE(PREOPEN);
2780 	CASE(CONNECTING);
2781 	CASE(NEGOTIATING);
2782 	CASE(OPEN);
2783 	CASE(STANDBY);
2784 	default:
2785 		pr_warn("%s con %p unrecognized state %lu\n",
2786 			__func__, con, con->state);
2787 		con->error_msg = "unrecognized con state";
2788 		BUG();
2789 		break;
2790 	}
2791 #undef CASE
2792 
2793 	return true;
2794 }
2795 
con_backoff(struct ceph_connection * con)2796 static bool con_backoff(struct ceph_connection *con)
2797 {
2798 	int ret;
2799 
2800 	if (!con_flag_test_and_clear(con, CON_FLAG_BACKOFF))
2801 		return false;
2802 
2803 	ret = queue_con_delay(con, round_jiffies_relative(con->delay));
2804 	if (ret) {
2805 		dout("%s: con %p FAILED to back off %lu\n", __func__,
2806 			con, con->delay);
2807 		BUG_ON(ret == -ENOENT);
2808 		con_flag_set(con, CON_FLAG_BACKOFF);
2809 	}
2810 
2811 	return true;
2812 }
2813 
2814 /* Finish fault handling; con->mutex must *not* be held here */
2815 
con_fault_finish(struct ceph_connection * con)2816 static void con_fault_finish(struct ceph_connection *con)
2817 {
2818 	/*
2819 	 * in case we faulted due to authentication, invalidate our
2820 	 * current tickets so that we can get new ones.
2821 	 */
2822 	if (con->auth_retry && con->ops->invalidate_authorizer) {
2823 		dout("calling invalidate_authorizer()\n");
2824 		con->ops->invalidate_authorizer(con);
2825 	}
2826 
2827 	if (con->ops->fault)
2828 		con->ops->fault(con);
2829 }
2830 
2831 /*
2832  * Do some work on a connection.  Drop a connection ref when we're done.
2833  */
con_work(struct work_struct * work)2834 static void con_work(struct work_struct *work)
2835 {
2836 	struct ceph_connection *con = container_of(work, struct ceph_connection,
2837 						   work.work);
2838 	bool fault;
2839 
2840 	mutex_lock(&con->mutex);
2841 	while (true) {
2842 		int ret;
2843 
2844 		if ((fault = con_sock_closed(con))) {
2845 			dout("%s: con %p SOCK_CLOSED\n", __func__, con);
2846 			break;
2847 		}
2848 		if (con_backoff(con)) {
2849 			dout("%s: con %p BACKOFF\n", __func__, con);
2850 			break;
2851 		}
2852 		if (con->state == CON_STATE_STANDBY) {
2853 			dout("%s: con %p STANDBY\n", __func__, con);
2854 			break;
2855 		}
2856 		if (con->state == CON_STATE_CLOSED) {
2857 			dout("%s: con %p CLOSED\n", __func__, con);
2858 			BUG_ON(con->sock);
2859 			break;
2860 		}
2861 		if (con->state == CON_STATE_PREOPEN) {
2862 			dout("%s: con %p PREOPEN\n", __func__, con);
2863 			BUG_ON(con->sock);
2864 		}
2865 
2866 		ret = try_read(con);
2867 		if (ret < 0) {
2868 			if (ret == -EAGAIN)
2869 				continue;
2870 			if (!con->error_msg)
2871 				con->error_msg = "socket error on read";
2872 			fault = true;
2873 			break;
2874 		}
2875 
2876 		ret = try_write(con);
2877 		if (ret < 0) {
2878 			if (ret == -EAGAIN)
2879 				continue;
2880 			if (!con->error_msg)
2881 				con->error_msg = "socket error on write";
2882 			fault = true;
2883 		}
2884 
2885 		break;	/* If we make it to here, we're done */
2886 	}
2887 	if (fault)
2888 		con_fault(con);
2889 	mutex_unlock(&con->mutex);
2890 
2891 	if (fault)
2892 		con_fault_finish(con);
2893 
2894 	con->ops->put(con);
2895 }
2896 
2897 /*
2898  * Generic error/fault handler.  A retry mechanism is used with
2899  * exponential backoff
2900  */
con_fault(struct ceph_connection * con)2901 static void con_fault(struct ceph_connection *con)
2902 {
2903 	dout("fault %p state %lu to peer %s\n",
2904 	     con, con->state, ceph_pr_addr(&con->peer_addr.in_addr));
2905 
2906 	pr_warn("%s%lld %s %s\n", ENTITY_NAME(con->peer_name),
2907 		ceph_pr_addr(&con->peer_addr.in_addr), con->error_msg);
2908 	con->error_msg = NULL;
2909 
2910 	WARN_ON(con->state != CON_STATE_CONNECTING &&
2911 	       con->state != CON_STATE_NEGOTIATING &&
2912 	       con->state != CON_STATE_OPEN);
2913 
2914 	con_close_socket(con);
2915 
2916 	if (con_flag_test(con, CON_FLAG_LOSSYTX)) {
2917 		dout("fault on LOSSYTX channel, marking CLOSED\n");
2918 		con->state = CON_STATE_CLOSED;
2919 		return;
2920 	}
2921 
2922 	if (con->in_msg) {
2923 		BUG_ON(con->in_msg->con != con);
2924 		con->in_msg->con = NULL;
2925 		ceph_msg_put(con->in_msg);
2926 		con->in_msg = NULL;
2927 		con->ops->put(con);
2928 	}
2929 
2930 	/* Requeue anything that hasn't been acked */
2931 	list_splice_init(&con->out_sent, &con->out_queue);
2932 
2933 	/* If there are no messages queued or keepalive pending, place
2934 	 * the connection in a STANDBY state */
2935 	if (list_empty(&con->out_queue) &&
2936 	    !con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING)) {
2937 		dout("fault %p setting STANDBY clearing WRITE_PENDING\n", con);
2938 		con_flag_clear(con, CON_FLAG_WRITE_PENDING);
2939 		con->state = CON_STATE_STANDBY;
2940 	} else {
2941 		/* retry after a delay. */
2942 		con->state = CON_STATE_PREOPEN;
2943 		if (con->delay == 0)
2944 			con->delay = BASE_DELAY_INTERVAL;
2945 		else if (con->delay < MAX_DELAY_INTERVAL)
2946 			con->delay *= 2;
2947 		con_flag_set(con, CON_FLAG_BACKOFF);
2948 		queue_con(con);
2949 	}
2950 }
2951 
2952 
2953 
2954 /*
2955  * initialize a new messenger instance
2956  */
ceph_messenger_init(struct ceph_messenger * msgr,struct ceph_entity_addr * myaddr,u64 supported_features,u64 required_features,bool nocrc,bool tcp_nodelay)2957 void ceph_messenger_init(struct ceph_messenger *msgr,
2958 			struct ceph_entity_addr *myaddr,
2959 			u64 supported_features,
2960 			u64 required_features,
2961 			bool nocrc,
2962 			bool tcp_nodelay)
2963 {
2964 	msgr->supported_features = supported_features;
2965 	msgr->required_features = required_features;
2966 
2967 	spin_lock_init(&msgr->global_seq_lock);
2968 
2969 	if (myaddr)
2970 		msgr->inst.addr = *myaddr;
2971 
2972 	/* select a random nonce */
2973 	msgr->inst.addr.type = 0;
2974 	get_random_bytes(&msgr->inst.addr.nonce, sizeof(msgr->inst.addr.nonce));
2975 	encode_my_addr(msgr);
2976 	msgr->nocrc = nocrc;
2977 	msgr->tcp_nodelay = tcp_nodelay;
2978 
2979 	atomic_set(&msgr->stopping, 0);
2980 
2981 	dout("%s %p\n", __func__, msgr);
2982 }
2983 EXPORT_SYMBOL(ceph_messenger_init);
2984 
clear_standby(struct ceph_connection * con)2985 static void clear_standby(struct ceph_connection *con)
2986 {
2987 	/* come back from STANDBY? */
2988 	if (con->state == CON_STATE_STANDBY) {
2989 		dout("clear_standby %p and ++connect_seq\n", con);
2990 		con->state = CON_STATE_PREOPEN;
2991 		con->connect_seq++;
2992 		WARN_ON(con_flag_test(con, CON_FLAG_WRITE_PENDING));
2993 		WARN_ON(con_flag_test(con, CON_FLAG_KEEPALIVE_PENDING));
2994 	}
2995 }
2996 
2997 /*
2998  * Queue up an outgoing message on the given connection.
2999  */
ceph_con_send(struct ceph_connection * con,struct ceph_msg * msg)3000 void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg)
3001 {
3002 	/* set src+dst */
3003 	msg->hdr.src = con->msgr->inst.name;
3004 	BUG_ON(msg->front.iov_len != le32_to_cpu(msg->hdr.front_len));
3005 	msg->needs_out_seq = true;
3006 
3007 	mutex_lock(&con->mutex);
3008 
3009 	if (con->state == CON_STATE_CLOSED) {
3010 		dout("con_send %p closed, dropping %p\n", con, msg);
3011 		ceph_msg_put(msg);
3012 		mutex_unlock(&con->mutex);
3013 		return;
3014 	}
3015 
3016 	BUG_ON(msg->con != NULL);
3017 	msg->con = con->ops->get(con);
3018 	BUG_ON(msg->con == NULL);
3019 
3020 	BUG_ON(!list_empty(&msg->list_head));
3021 	list_add_tail(&msg->list_head, &con->out_queue);
3022 	dout("----- %p to %s%lld %d=%s len %d+%d+%d -----\n", msg,
3023 	     ENTITY_NAME(con->peer_name), le16_to_cpu(msg->hdr.type),
3024 	     ceph_msg_type_name(le16_to_cpu(msg->hdr.type)),
3025 	     le32_to_cpu(msg->hdr.front_len),
3026 	     le32_to_cpu(msg->hdr.middle_len),
3027 	     le32_to_cpu(msg->hdr.data_len));
3028 
3029 	clear_standby(con);
3030 	mutex_unlock(&con->mutex);
3031 
3032 	/* if there wasn't anything waiting to send before, queue
3033 	 * new work */
3034 	if (con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0)
3035 		queue_con(con);
3036 }
3037 EXPORT_SYMBOL(ceph_con_send);
3038 
3039 /*
3040  * Revoke a message that was previously queued for send
3041  */
ceph_msg_revoke(struct ceph_msg * msg)3042 void ceph_msg_revoke(struct ceph_msg *msg)
3043 {
3044 	struct ceph_connection *con = msg->con;
3045 
3046 	if (!con)
3047 		return;		/* Message not in our possession */
3048 
3049 	mutex_lock(&con->mutex);
3050 	if (!list_empty(&msg->list_head)) {
3051 		dout("%s %p msg %p - was on queue\n", __func__, con, msg);
3052 		list_del_init(&msg->list_head);
3053 		BUG_ON(msg->con == NULL);
3054 		msg->con->ops->put(msg->con);
3055 		msg->con = NULL;
3056 		msg->hdr.seq = 0;
3057 
3058 		ceph_msg_put(msg);
3059 	}
3060 	if (con->out_msg == msg) {
3061 		BUG_ON(con->out_skip);
3062 		/* footer */
3063 		if (con->out_msg_done) {
3064 			con->out_skip += con_out_kvec_skip(con);
3065 		} else {
3066 			BUG_ON(!msg->data_length);
3067 			if (con->peer_features & CEPH_FEATURE_MSG_AUTH)
3068 				con->out_skip += sizeof(msg->footer);
3069 			else
3070 				con->out_skip += sizeof(msg->old_footer);
3071 		}
3072 		/* data, middle, front */
3073 		if (msg->data_length)
3074 			con->out_skip += msg->cursor.total_resid;
3075 		if (msg->middle)
3076 			con->out_skip += con_out_kvec_skip(con);
3077 		con->out_skip += con_out_kvec_skip(con);
3078 
3079 		dout("%s %p msg %p - was sending, will write %d skip %d\n",
3080 		     __func__, con, msg, con->out_kvec_bytes, con->out_skip);
3081 		msg->hdr.seq = 0;
3082 		con->out_msg = NULL;
3083 		ceph_msg_put(msg);
3084 	}
3085 
3086 	mutex_unlock(&con->mutex);
3087 }
3088 
3089 /*
3090  * Revoke a message that we may be reading data into
3091  */
ceph_msg_revoke_incoming(struct ceph_msg * msg)3092 void ceph_msg_revoke_incoming(struct ceph_msg *msg)
3093 {
3094 	struct ceph_connection *con;
3095 
3096 	BUG_ON(msg == NULL);
3097 	if (!msg->con) {
3098 		dout("%s msg %p null con\n", __func__, msg);
3099 
3100 		return;		/* Message not in our possession */
3101 	}
3102 
3103 	con = msg->con;
3104 	mutex_lock(&con->mutex);
3105 	if (con->in_msg == msg) {
3106 		unsigned int front_len = le32_to_cpu(con->in_hdr.front_len);
3107 		unsigned int middle_len = le32_to_cpu(con->in_hdr.middle_len);
3108 		unsigned int data_len = le32_to_cpu(con->in_hdr.data_len);
3109 
3110 		/* skip rest of message */
3111 		dout("%s %p msg %p revoked\n", __func__, con, msg);
3112 		con->in_base_pos = con->in_base_pos -
3113 				sizeof(struct ceph_msg_header) -
3114 				front_len -
3115 				middle_len -
3116 				data_len -
3117 				sizeof(struct ceph_msg_footer);
3118 		ceph_msg_put(con->in_msg);
3119 		con->in_msg = NULL;
3120 		con->in_tag = CEPH_MSGR_TAG_READY;
3121 		con->in_seq++;
3122 	} else {
3123 		dout("%s %p in_msg %p msg %p no-op\n",
3124 		     __func__, con, con->in_msg, msg);
3125 	}
3126 	mutex_unlock(&con->mutex);
3127 }
3128 
3129 /*
3130  * Queue a keepalive byte to ensure the tcp connection is alive.
3131  */
ceph_con_keepalive(struct ceph_connection * con)3132 void ceph_con_keepalive(struct ceph_connection *con)
3133 {
3134 	dout("con_keepalive %p\n", con);
3135 	mutex_lock(&con->mutex);
3136 	clear_standby(con);
3137 	mutex_unlock(&con->mutex);
3138 	if (con_flag_test_and_set(con, CON_FLAG_KEEPALIVE_PENDING) == 0 &&
3139 	    con_flag_test_and_set(con, CON_FLAG_WRITE_PENDING) == 0)
3140 		queue_con(con);
3141 }
3142 EXPORT_SYMBOL(ceph_con_keepalive);
3143 
ceph_msg_data_create(enum ceph_msg_data_type type)3144 static struct ceph_msg_data *ceph_msg_data_create(enum ceph_msg_data_type type)
3145 {
3146 	struct ceph_msg_data *data;
3147 
3148 	if (WARN_ON(!ceph_msg_data_type_valid(type)))
3149 		return NULL;
3150 
3151 	data = kmem_cache_zalloc(ceph_msg_data_cache, GFP_NOFS);
3152 	if (data)
3153 		data->type = type;
3154 	INIT_LIST_HEAD(&data->links);
3155 
3156 	return data;
3157 }
3158 
ceph_msg_data_destroy(struct ceph_msg_data * data)3159 static void ceph_msg_data_destroy(struct ceph_msg_data *data)
3160 {
3161 	if (!data)
3162 		return;
3163 
3164 	WARN_ON(!list_empty(&data->links));
3165 	if (data->type == CEPH_MSG_DATA_PAGELIST)
3166 		ceph_pagelist_release(data->pagelist);
3167 	kmem_cache_free(ceph_msg_data_cache, data);
3168 }
3169 
ceph_msg_data_add_pages(struct ceph_msg * msg,struct page ** pages,size_t length,size_t alignment)3170 void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
3171 		size_t length, size_t alignment)
3172 {
3173 	struct ceph_msg_data *data;
3174 
3175 	BUG_ON(!pages);
3176 	BUG_ON(!length);
3177 
3178 	data = ceph_msg_data_create(CEPH_MSG_DATA_PAGES);
3179 	BUG_ON(!data);
3180 	data->pages = pages;
3181 	data->length = length;
3182 	data->alignment = alignment & ~PAGE_MASK;
3183 
3184 	list_add_tail(&data->links, &msg->data);
3185 	msg->data_length += length;
3186 }
3187 EXPORT_SYMBOL(ceph_msg_data_add_pages);
3188 
ceph_msg_data_add_pagelist(struct ceph_msg * msg,struct ceph_pagelist * pagelist)3189 void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
3190 				struct ceph_pagelist *pagelist)
3191 {
3192 	struct ceph_msg_data *data;
3193 
3194 	BUG_ON(!pagelist);
3195 	BUG_ON(!pagelist->length);
3196 
3197 	data = ceph_msg_data_create(CEPH_MSG_DATA_PAGELIST);
3198 	BUG_ON(!data);
3199 	data->pagelist = pagelist;
3200 
3201 	list_add_tail(&data->links, &msg->data);
3202 	msg->data_length += pagelist->length;
3203 }
3204 EXPORT_SYMBOL(ceph_msg_data_add_pagelist);
3205 
3206 #ifdef	CONFIG_BLOCK
ceph_msg_data_add_bio(struct ceph_msg * msg,struct bio * bio,size_t length)3207 void ceph_msg_data_add_bio(struct ceph_msg *msg, struct bio *bio,
3208 		size_t length)
3209 {
3210 	struct ceph_msg_data *data;
3211 
3212 	BUG_ON(!bio);
3213 
3214 	data = ceph_msg_data_create(CEPH_MSG_DATA_BIO);
3215 	BUG_ON(!data);
3216 	data->bio = bio;
3217 	data->bio_length = length;
3218 
3219 	list_add_tail(&data->links, &msg->data);
3220 	msg->data_length += length;
3221 }
3222 EXPORT_SYMBOL(ceph_msg_data_add_bio);
3223 #endif	/* CONFIG_BLOCK */
3224 
3225 /*
3226  * construct a new message with given type, size
3227  * the new msg has a ref count of 1.
3228  */
ceph_msg_new(int type,int front_len,gfp_t flags,bool can_fail)3229 struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
3230 			      bool can_fail)
3231 {
3232 	struct ceph_msg *m;
3233 
3234 	m = kmem_cache_zalloc(ceph_msg_cache, flags);
3235 	if (m == NULL)
3236 		goto out;
3237 
3238 	m->hdr.type = cpu_to_le16(type);
3239 	m->hdr.priority = cpu_to_le16(CEPH_MSG_PRIO_DEFAULT);
3240 	m->hdr.front_len = cpu_to_le32(front_len);
3241 
3242 	INIT_LIST_HEAD(&m->list_head);
3243 	kref_init(&m->kref);
3244 	INIT_LIST_HEAD(&m->data);
3245 
3246 	/* front */
3247 	if (front_len) {
3248 		m->front.iov_base = ceph_kvmalloc(front_len, flags);
3249 		if (m->front.iov_base == NULL) {
3250 			dout("ceph_msg_new can't allocate %d bytes\n",
3251 			     front_len);
3252 			goto out2;
3253 		}
3254 	} else {
3255 		m->front.iov_base = NULL;
3256 	}
3257 	m->front_alloc_len = m->front.iov_len = front_len;
3258 
3259 	dout("ceph_msg_new %p front %d\n", m, front_len);
3260 	return m;
3261 
3262 out2:
3263 	ceph_msg_put(m);
3264 out:
3265 	if (!can_fail) {
3266 		pr_err("msg_new can't create type %d front %d\n", type,
3267 		       front_len);
3268 		WARN_ON(1);
3269 	} else {
3270 		dout("msg_new can't create type %d front %d\n", type,
3271 		     front_len);
3272 	}
3273 	return NULL;
3274 }
3275 EXPORT_SYMBOL(ceph_msg_new);
3276 
3277 /*
3278  * Allocate "middle" portion of a message, if it is needed and wasn't
3279  * allocated by alloc_msg.  This allows us to read a small fixed-size
3280  * per-type header in the front and then gracefully fail (i.e.,
3281  * propagate the error to the caller based on info in the front) when
3282  * the middle is too large.
3283  */
ceph_alloc_middle(struct ceph_connection * con,struct ceph_msg * msg)3284 static int ceph_alloc_middle(struct ceph_connection *con, struct ceph_msg *msg)
3285 {
3286 	int type = le16_to_cpu(msg->hdr.type);
3287 	int middle_len = le32_to_cpu(msg->hdr.middle_len);
3288 
3289 	dout("alloc_middle %p type %d %s middle_len %d\n", msg, type,
3290 	     ceph_msg_type_name(type), middle_len);
3291 	BUG_ON(!middle_len);
3292 	BUG_ON(msg->middle);
3293 
3294 	msg->middle = ceph_buffer_new(middle_len, GFP_NOFS);
3295 	if (!msg->middle)
3296 		return -ENOMEM;
3297 	return 0;
3298 }
3299 
3300 /*
3301  * Allocate a message for receiving an incoming message on a
3302  * connection, and save the result in con->in_msg.  Uses the
3303  * connection's private alloc_msg op if available.
3304  *
3305  * Returns 0 on success, or a negative error code.
3306  *
3307  * On success, if we set *skip = 1:
3308  *  - the next message should be skipped and ignored.
3309  *  - con->in_msg == NULL
3310  * or if we set *skip = 0:
3311  *  - con->in_msg is non-null.
3312  * On error (ENOMEM, EAGAIN, ...),
3313  *  - con->in_msg == NULL
3314  */
ceph_con_in_msg_alloc(struct ceph_connection * con,int * skip)3315 static int ceph_con_in_msg_alloc(struct ceph_connection *con, int *skip)
3316 {
3317 	struct ceph_msg_header *hdr = &con->in_hdr;
3318 	int middle_len = le32_to_cpu(hdr->middle_len);
3319 	struct ceph_msg *msg;
3320 	int ret = 0;
3321 
3322 	BUG_ON(con->in_msg != NULL);
3323 	BUG_ON(!con->ops->alloc_msg);
3324 
3325 	mutex_unlock(&con->mutex);
3326 	msg = con->ops->alloc_msg(con, hdr, skip);
3327 	mutex_lock(&con->mutex);
3328 	if (con->state != CON_STATE_OPEN) {
3329 		if (msg)
3330 			ceph_msg_put(msg);
3331 		return -EAGAIN;
3332 	}
3333 	if (msg) {
3334 		BUG_ON(*skip);
3335 		con->in_msg = msg;
3336 		con->in_msg->con = con->ops->get(con);
3337 		BUG_ON(con->in_msg->con == NULL);
3338 	} else {
3339 		/*
3340 		 * Null message pointer means either we should skip
3341 		 * this message or we couldn't allocate memory.  The
3342 		 * former is not an error.
3343 		 */
3344 		if (*skip)
3345 			return 0;
3346 
3347 		con->error_msg = "error allocating memory for incoming message";
3348 		return -ENOMEM;
3349 	}
3350 	memcpy(&con->in_msg->hdr, &con->in_hdr, sizeof(con->in_hdr));
3351 
3352 	if (middle_len && !con->in_msg->middle) {
3353 		ret = ceph_alloc_middle(con, con->in_msg);
3354 		if (ret < 0) {
3355 			ceph_msg_put(con->in_msg);
3356 			con->in_msg = NULL;
3357 		}
3358 	}
3359 
3360 	return ret;
3361 }
3362 
3363 
3364 /*
3365  * Free a generically kmalloc'd message.
3366  */
ceph_msg_free(struct ceph_msg * m)3367 static void ceph_msg_free(struct ceph_msg *m)
3368 {
3369 	dout("%s %p\n", __func__, m);
3370 	kvfree(m->front.iov_base);
3371 	kmem_cache_free(ceph_msg_cache, m);
3372 }
3373 
ceph_msg_release(struct kref * kref)3374 static void ceph_msg_release(struct kref *kref)
3375 {
3376 	struct ceph_msg *m = container_of(kref, struct ceph_msg, kref);
3377 	LIST_HEAD(data);
3378 	struct list_head *links;
3379 	struct list_head *next;
3380 
3381 	dout("%s %p\n", __func__, m);
3382 	WARN_ON(!list_empty(&m->list_head));
3383 
3384 	/* drop middle, data, if any */
3385 	if (m->middle) {
3386 		ceph_buffer_put(m->middle);
3387 		m->middle = NULL;
3388 	}
3389 
3390 	list_splice_init(&m->data, &data);
3391 	list_for_each_safe(links, next, &data) {
3392 		struct ceph_msg_data *data;
3393 
3394 		data = list_entry(links, struct ceph_msg_data, links);
3395 		list_del_init(links);
3396 		ceph_msg_data_destroy(data);
3397 	}
3398 	m->data_length = 0;
3399 
3400 	if (m->pool)
3401 		ceph_msgpool_put(m->pool, m);
3402 	else
3403 		ceph_msg_free(m);
3404 }
3405 
ceph_msg_get(struct ceph_msg * msg)3406 struct ceph_msg *ceph_msg_get(struct ceph_msg *msg)
3407 {
3408 	dout("%s %p (was %d)\n", __func__, msg,
3409 	     atomic_read(&msg->kref.refcount));
3410 	kref_get(&msg->kref);
3411 	return msg;
3412 }
3413 EXPORT_SYMBOL(ceph_msg_get);
3414 
ceph_msg_put(struct ceph_msg * msg)3415 void ceph_msg_put(struct ceph_msg *msg)
3416 {
3417 	dout("%s %p (was %d)\n", __func__, msg,
3418 	     atomic_read(&msg->kref.refcount));
3419 	kref_put(&msg->kref, ceph_msg_release);
3420 }
3421 EXPORT_SYMBOL(ceph_msg_put);
3422 
ceph_msg_dump(struct ceph_msg * msg)3423 void ceph_msg_dump(struct ceph_msg *msg)
3424 {
3425 	pr_debug("msg_dump %p (front_alloc_len %d length %zd)\n", msg,
3426 		 msg->front_alloc_len, msg->data_length);
3427 	print_hex_dump(KERN_DEBUG, "header: ",
3428 		       DUMP_PREFIX_OFFSET, 16, 1,
3429 		       &msg->hdr, sizeof(msg->hdr), true);
3430 	print_hex_dump(KERN_DEBUG, " front: ",
3431 		       DUMP_PREFIX_OFFSET, 16, 1,
3432 		       msg->front.iov_base, msg->front.iov_len, true);
3433 	if (msg->middle)
3434 		print_hex_dump(KERN_DEBUG, "middle: ",
3435 			       DUMP_PREFIX_OFFSET, 16, 1,
3436 			       msg->middle->vec.iov_base,
3437 			       msg->middle->vec.iov_len, true);
3438 	print_hex_dump(KERN_DEBUG, "footer: ",
3439 		       DUMP_PREFIX_OFFSET, 16, 1,
3440 		       &msg->footer, sizeof(msg->footer), true);
3441 }
3442 EXPORT_SYMBOL(ceph_msg_dump);
3443