xref: /openbmc/linux/net/ceph/messenger_v1.c (revision b04b076f)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3 
4 #include <linux/bvec.h>
5 #include <linux/crc32c.h>
6 #include <linux/net.h>
7 #include <linux/socket.h>
8 #include <net/sock.h>
9 
10 #include <linux/ceph/ceph_features.h>
11 #include <linux/ceph/decode.h>
12 #include <linux/ceph/libceph.h>
13 #include <linux/ceph/messenger.h>
14 
15 /* static tag bytes (protocol control messages) */
16 static char tag_msg = CEPH_MSGR_TAG_MSG;
17 static char tag_ack = CEPH_MSGR_TAG_ACK;
18 static char tag_keepalive = CEPH_MSGR_TAG_KEEPALIVE;
19 static char tag_keepalive2 = CEPH_MSGR_TAG_KEEPALIVE2;
20 
21 /*
22  * If @buf is NULL, discard up to @len bytes.
23  */
24 static int ceph_tcp_recvmsg(struct socket *sock, void *buf, size_t len)
25 {
26 	struct kvec iov = {buf, len};
27 	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
28 	int r;
29 
30 	if (!buf)
31 		msg.msg_flags |= MSG_TRUNC;
32 
33 	iov_iter_kvec(&msg.msg_iter, ITER_DEST, &iov, 1, len);
34 	r = sock_recvmsg(sock, &msg, msg.msg_flags);
35 	if (r == -EAGAIN)
36 		r = 0;
37 	return r;
38 }
39 
40 static int ceph_tcp_recvpage(struct socket *sock, struct page *page,
41 		     int page_offset, size_t length)
42 {
43 	struct bio_vec bvec;
44 	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
45 	int r;
46 
47 	BUG_ON(page_offset + length > PAGE_SIZE);
48 	bvec_set_page(&bvec, page, length, page_offset);
49 	iov_iter_bvec(&msg.msg_iter, ITER_DEST, &bvec, 1, length);
50 	r = sock_recvmsg(sock, &msg, msg.msg_flags);
51 	if (r == -EAGAIN)
52 		r = 0;
53 	return r;
54 }
55 
56 /*
57  * write something.  @more is true if caller will be sending more data
58  * shortly.
59  */
60 static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov,
61 			    size_t kvlen, size_t len, bool more)
62 {
63 	struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL };
64 	int r;
65 
66 	if (more)
67 		msg.msg_flags |= MSG_MORE;
68 	else
69 		msg.msg_flags |= MSG_EOR;  /* superfluous, but what the hell */
70 
71 	r = kernel_sendmsg(sock, &msg, iov, kvlen, len);
72 	if (r == -EAGAIN)
73 		r = 0;
74 	return r;
75 }
76 
77 /*
78  * @more: either or both of MSG_MORE and MSG_SENDPAGE_NOTLAST
79  */
80 static int ceph_tcp_sendpage(struct socket *sock, struct page *page,
81 			     int offset, size_t size, int more)
82 {
83 	ssize_t (*sendpage)(struct socket *sock, struct page *page,
84 			    int offset, size_t size, int flags);
85 	int flags = MSG_DONTWAIT | MSG_NOSIGNAL | more;
86 	int ret;
87 
88 	/*
89 	 * sendpage cannot properly handle pages with page_count == 0,
90 	 * we need to fall back to sendmsg if that's the case.
91 	 *
92 	 * Same goes for slab pages: skb_can_coalesce() allows
93 	 * coalescing neighboring slab objects into a single frag which
94 	 * triggers one of hardened usercopy checks.
95 	 */
96 	if (sendpage_ok(page))
97 		sendpage = sock->ops->sendpage;
98 	else
99 		sendpage = sock_no_sendpage;
100 
101 	ret = sendpage(sock, page, offset, size, flags);
102 	if (ret == -EAGAIN)
103 		ret = 0;
104 
105 	return ret;
106 }
107 
108 static void con_out_kvec_reset(struct ceph_connection *con)
109 {
110 	BUG_ON(con->v1.out_skip);
111 
112 	con->v1.out_kvec_left = 0;
113 	con->v1.out_kvec_bytes = 0;
114 	con->v1.out_kvec_cur = &con->v1.out_kvec[0];
115 }
116 
117 static void con_out_kvec_add(struct ceph_connection *con,
118 				size_t size, void *data)
119 {
120 	int index = con->v1.out_kvec_left;
121 
122 	BUG_ON(con->v1.out_skip);
123 	BUG_ON(index >= ARRAY_SIZE(con->v1.out_kvec));
124 
125 	con->v1.out_kvec[index].iov_len = size;
126 	con->v1.out_kvec[index].iov_base = data;
127 	con->v1.out_kvec_left++;
128 	con->v1.out_kvec_bytes += size;
129 }
130 
131 /*
132  * Chop off a kvec from the end.  Return residual number of bytes for
133  * that kvec, i.e. how many bytes would have been written if the kvec
134  * hadn't been nuked.
135  */
136 static int con_out_kvec_skip(struct ceph_connection *con)
137 {
138 	int skip = 0;
139 
140 	if (con->v1.out_kvec_bytes > 0) {
141 		skip = con->v1.out_kvec_cur[con->v1.out_kvec_left - 1].iov_len;
142 		BUG_ON(con->v1.out_kvec_bytes < skip);
143 		BUG_ON(!con->v1.out_kvec_left);
144 		con->v1.out_kvec_bytes -= skip;
145 		con->v1.out_kvec_left--;
146 	}
147 
148 	return skip;
149 }
150 
151 static size_t sizeof_footer(struct ceph_connection *con)
152 {
153 	return (con->peer_features & CEPH_FEATURE_MSG_AUTH) ?
154 	    sizeof(struct ceph_msg_footer) :
155 	    sizeof(struct ceph_msg_footer_old);
156 }
157 
158 static void prepare_message_data(struct ceph_msg *msg, u32 data_len)
159 {
160 	/* Initialize data cursor */
161 
162 	ceph_msg_data_cursor_init(&msg->cursor, msg, data_len);
163 }
164 
165 /*
166  * Prepare footer for currently outgoing message, and finish things
167  * off.  Assumes out_kvec* are already valid.. we just add on to the end.
168  */
169 static void prepare_write_message_footer(struct ceph_connection *con)
170 {
171 	struct ceph_msg *m = con->out_msg;
172 
173 	m->footer.flags |= CEPH_MSG_FOOTER_COMPLETE;
174 
175 	dout("prepare_write_message_footer %p\n", con);
176 	con_out_kvec_add(con, sizeof_footer(con), &m->footer);
177 	if (con->peer_features & CEPH_FEATURE_MSG_AUTH) {
178 		if (con->ops->sign_message)
179 			con->ops->sign_message(m);
180 		else
181 			m->footer.sig = 0;
182 	} else {
183 		m->old_footer.flags = m->footer.flags;
184 	}
185 	con->v1.out_more = m->more_to_follow;
186 	con->v1.out_msg_done = true;
187 }
188 
189 /*
190  * Prepare headers for the next outgoing message.
191  */
192 static void prepare_write_message(struct ceph_connection *con)
193 {
194 	struct ceph_msg *m;
195 	u32 crc;
196 
197 	con_out_kvec_reset(con);
198 	con->v1.out_msg_done = false;
199 
200 	/* Sneak an ack in there first?  If we can get it into the same
201 	 * TCP packet that's a good thing. */
202 	if (con->in_seq > con->in_seq_acked) {
203 		con->in_seq_acked = con->in_seq;
204 		con_out_kvec_add(con, sizeof (tag_ack), &tag_ack);
205 		con->v1.out_temp_ack = cpu_to_le64(con->in_seq_acked);
206 		con_out_kvec_add(con, sizeof(con->v1.out_temp_ack),
207 			&con->v1.out_temp_ack);
208 	}
209 
210 	ceph_con_get_out_msg(con);
211 	m = con->out_msg;
212 
213 	dout("prepare_write_message %p seq %lld type %d len %d+%d+%zd\n",
214 	     m, con->out_seq, le16_to_cpu(m->hdr.type),
215 	     le32_to_cpu(m->hdr.front_len), le32_to_cpu(m->hdr.middle_len),
216 	     m->data_length);
217 	WARN_ON(m->front.iov_len != le32_to_cpu(m->hdr.front_len));
218 	WARN_ON(m->data_length != le32_to_cpu(m->hdr.data_len));
219 
220 	/* tag + hdr + front + middle */
221 	con_out_kvec_add(con, sizeof (tag_msg), &tag_msg);
222 	con_out_kvec_add(con, sizeof(con->v1.out_hdr), &con->v1.out_hdr);
223 	con_out_kvec_add(con, m->front.iov_len, m->front.iov_base);
224 
225 	if (m->middle)
226 		con_out_kvec_add(con, m->middle->vec.iov_len,
227 			m->middle->vec.iov_base);
228 
229 	/* fill in hdr crc and finalize hdr */
230 	crc = crc32c(0, &m->hdr, offsetof(struct ceph_msg_header, crc));
231 	con->out_msg->hdr.crc = cpu_to_le32(crc);
232 	memcpy(&con->v1.out_hdr, &con->out_msg->hdr, sizeof(con->v1.out_hdr));
233 
234 	/* fill in front and middle crc, footer */
235 	crc = crc32c(0, m->front.iov_base, m->front.iov_len);
236 	con->out_msg->footer.front_crc = cpu_to_le32(crc);
237 	if (m->middle) {
238 		crc = crc32c(0, m->middle->vec.iov_base,
239 				m->middle->vec.iov_len);
240 		con->out_msg->footer.middle_crc = cpu_to_le32(crc);
241 	} else
242 		con->out_msg->footer.middle_crc = 0;
243 	dout("%s front_crc %u middle_crc %u\n", __func__,
244 	     le32_to_cpu(con->out_msg->footer.front_crc),
245 	     le32_to_cpu(con->out_msg->footer.middle_crc));
246 	con->out_msg->footer.flags = 0;
247 
248 	/* is there a data payload? */
249 	con->out_msg->footer.data_crc = 0;
250 	if (m->data_length) {
251 		prepare_message_data(con->out_msg, m->data_length);
252 		con->v1.out_more = 1;  /* data + footer will follow */
253 	} else {
254 		/* no, queue up footer too and be done */
255 		prepare_write_message_footer(con);
256 	}
257 
258 	ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
259 }
260 
261 /*
262  * Prepare an ack.
263  */
264 static void prepare_write_ack(struct ceph_connection *con)
265 {
266 	dout("prepare_write_ack %p %llu -> %llu\n", con,
267 	     con->in_seq_acked, con->in_seq);
268 	con->in_seq_acked = con->in_seq;
269 
270 	con_out_kvec_reset(con);
271 
272 	con_out_kvec_add(con, sizeof (tag_ack), &tag_ack);
273 
274 	con->v1.out_temp_ack = cpu_to_le64(con->in_seq_acked);
275 	con_out_kvec_add(con, sizeof(con->v1.out_temp_ack),
276 			 &con->v1.out_temp_ack);
277 
278 	con->v1.out_more = 1;  /* more will follow.. eventually.. */
279 	ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
280 }
281 
282 /*
283  * Prepare to share the seq during handshake
284  */
285 static void prepare_write_seq(struct ceph_connection *con)
286 {
287 	dout("prepare_write_seq %p %llu -> %llu\n", con,
288 	     con->in_seq_acked, con->in_seq);
289 	con->in_seq_acked = con->in_seq;
290 
291 	con_out_kvec_reset(con);
292 
293 	con->v1.out_temp_ack = cpu_to_le64(con->in_seq_acked);
294 	con_out_kvec_add(con, sizeof(con->v1.out_temp_ack),
295 			 &con->v1.out_temp_ack);
296 
297 	ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
298 }
299 
300 /*
301  * Prepare to write keepalive byte.
302  */
303 static void prepare_write_keepalive(struct ceph_connection *con)
304 {
305 	dout("prepare_write_keepalive %p\n", con);
306 	con_out_kvec_reset(con);
307 	if (con->peer_features & CEPH_FEATURE_MSGR_KEEPALIVE2) {
308 		struct timespec64 now;
309 
310 		ktime_get_real_ts64(&now);
311 		con_out_kvec_add(con, sizeof(tag_keepalive2), &tag_keepalive2);
312 		ceph_encode_timespec64(&con->v1.out_temp_keepalive2, &now);
313 		con_out_kvec_add(con, sizeof(con->v1.out_temp_keepalive2),
314 				 &con->v1.out_temp_keepalive2);
315 	} else {
316 		con_out_kvec_add(con, sizeof(tag_keepalive), &tag_keepalive);
317 	}
318 	ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
319 }
320 
321 /*
322  * Connection negotiation.
323  */
324 
325 static int get_connect_authorizer(struct ceph_connection *con)
326 {
327 	struct ceph_auth_handshake *auth;
328 	int auth_proto;
329 
330 	if (!con->ops->get_authorizer) {
331 		con->v1.auth = NULL;
332 		con->v1.out_connect.authorizer_protocol = CEPH_AUTH_UNKNOWN;
333 		con->v1.out_connect.authorizer_len = 0;
334 		return 0;
335 	}
336 
337 	auth = con->ops->get_authorizer(con, &auth_proto, con->v1.auth_retry);
338 	if (IS_ERR(auth))
339 		return PTR_ERR(auth);
340 
341 	con->v1.auth = auth;
342 	con->v1.out_connect.authorizer_protocol = cpu_to_le32(auth_proto);
343 	con->v1.out_connect.authorizer_len =
344 		cpu_to_le32(auth->authorizer_buf_len);
345 	return 0;
346 }
347 
348 /*
349  * We connected to a peer and are saying hello.
350  */
351 static void prepare_write_banner(struct ceph_connection *con)
352 {
353 	con_out_kvec_add(con, strlen(CEPH_BANNER), CEPH_BANNER);
354 	con_out_kvec_add(con, sizeof (con->msgr->my_enc_addr),
355 					&con->msgr->my_enc_addr);
356 
357 	con->v1.out_more = 0;
358 	ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
359 }
360 
361 static void __prepare_write_connect(struct ceph_connection *con)
362 {
363 	con_out_kvec_add(con, sizeof(con->v1.out_connect),
364 			 &con->v1.out_connect);
365 	if (con->v1.auth)
366 		con_out_kvec_add(con, con->v1.auth->authorizer_buf_len,
367 				 con->v1.auth->authorizer_buf);
368 
369 	con->v1.out_more = 0;
370 	ceph_con_flag_set(con, CEPH_CON_F_WRITE_PENDING);
371 }
372 
373 static int prepare_write_connect(struct ceph_connection *con)
374 {
375 	unsigned int global_seq = ceph_get_global_seq(con->msgr, 0);
376 	int proto;
377 	int ret;
378 
379 	switch (con->peer_name.type) {
380 	case CEPH_ENTITY_TYPE_MON:
381 		proto = CEPH_MONC_PROTOCOL;
382 		break;
383 	case CEPH_ENTITY_TYPE_OSD:
384 		proto = CEPH_OSDC_PROTOCOL;
385 		break;
386 	case CEPH_ENTITY_TYPE_MDS:
387 		proto = CEPH_MDSC_PROTOCOL;
388 		break;
389 	default:
390 		BUG();
391 	}
392 
393 	dout("prepare_write_connect %p cseq=%d gseq=%d proto=%d\n", con,
394 	     con->v1.connect_seq, global_seq, proto);
395 
396 	con->v1.out_connect.features =
397 		cpu_to_le64(from_msgr(con->msgr)->supported_features);
398 	con->v1.out_connect.host_type = cpu_to_le32(CEPH_ENTITY_TYPE_CLIENT);
399 	con->v1.out_connect.connect_seq = cpu_to_le32(con->v1.connect_seq);
400 	con->v1.out_connect.global_seq = cpu_to_le32(global_seq);
401 	con->v1.out_connect.protocol_version = cpu_to_le32(proto);
402 	con->v1.out_connect.flags = 0;
403 
404 	ret = get_connect_authorizer(con);
405 	if (ret)
406 		return ret;
407 
408 	__prepare_write_connect(con);
409 	return 0;
410 }
411 
412 /*
413  * write as much of pending kvecs to the socket as we can.
414  *  1 -> done
415  *  0 -> socket full, but more to do
416  * <0 -> error
417  */
418 static int write_partial_kvec(struct ceph_connection *con)
419 {
420 	int ret;
421 
422 	dout("write_partial_kvec %p %d left\n", con, con->v1.out_kvec_bytes);
423 	while (con->v1.out_kvec_bytes > 0) {
424 		ret = ceph_tcp_sendmsg(con->sock, con->v1.out_kvec_cur,
425 				       con->v1.out_kvec_left,
426 				       con->v1.out_kvec_bytes,
427 				       con->v1.out_more);
428 		if (ret <= 0)
429 			goto out;
430 		con->v1.out_kvec_bytes -= ret;
431 		if (!con->v1.out_kvec_bytes)
432 			break;            /* done */
433 
434 		/* account for full iov entries consumed */
435 		while (ret >= con->v1.out_kvec_cur->iov_len) {
436 			BUG_ON(!con->v1.out_kvec_left);
437 			ret -= con->v1.out_kvec_cur->iov_len;
438 			con->v1.out_kvec_cur++;
439 			con->v1.out_kvec_left--;
440 		}
441 		/* and for a partially-consumed entry */
442 		if (ret) {
443 			con->v1.out_kvec_cur->iov_len -= ret;
444 			con->v1.out_kvec_cur->iov_base += ret;
445 		}
446 	}
447 	con->v1.out_kvec_left = 0;
448 	ret = 1;
449 out:
450 	dout("write_partial_kvec %p %d left in %d kvecs ret = %d\n", con,
451 	     con->v1.out_kvec_bytes, con->v1.out_kvec_left, ret);
452 	return ret;  /* done! */
453 }
454 
455 /*
456  * Write as much message data payload as we can.  If we finish, queue
457  * up the footer.
458  *  1 -> done, footer is now queued in out_kvec[].
459  *  0 -> socket full, but more to do
460  * <0 -> error
461  */
462 static int write_partial_message_data(struct ceph_connection *con)
463 {
464 	struct ceph_msg *msg = con->out_msg;
465 	struct ceph_msg_data_cursor *cursor = &msg->cursor;
466 	bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
467 	int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
468 	u32 crc;
469 
470 	dout("%s %p msg %p\n", __func__, con, msg);
471 
472 	if (!msg->num_data_items)
473 		return -EINVAL;
474 
475 	/*
476 	 * Iterate through each page that contains data to be
477 	 * written, and send as much as possible for each.
478 	 *
479 	 * If we are calculating the data crc (the default), we will
480 	 * need to map the page.  If we have no pages, they have
481 	 * been revoked, so use the zero page.
482 	 */
483 	crc = do_datacrc ? le32_to_cpu(msg->footer.data_crc) : 0;
484 	while (cursor->total_resid) {
485 		struct page *page;
486 		size_t page_offset;
487 		size_t length;
488 		int ret;
489 
490 		if (!cursor->resid) {
491 			ceph_msg_data_advance(cursor, 0);
492 			continue;
493 		}
494 
495 		page = ceph_msg_data_next(cursor, &page_offset, &length);
496 		if (length == cursor->total_resid)
497 			more = MSG_MORE;
498 		ret = ceph_tcp_sendpage(con->sock, page, page_offset, length,
499 					more);
500 		if (ret <= 0) {
501 			if (do_datacrc)
502 				msg->footer.data_crc = cpu_to_le32(crc);
503 
504 			return ret;
505 		}
506 		if (do_datacrc && cursor->need_crc)
507 			crc = ceph_crc32c_page(crc, page, page_offset, length);
508 		ceph_msg_data_advance(cursor, (size_t)ret);
509 	}
510 
511 	dout("%s %p msg %p done\n", __func__, con, msg);
512 
513 	/* prepare and queue up footer, too */
514 	if (do_datacrc)
515 		msg->footer.data_crc = cpu_to_le32(crc);
516 	else
517 		msg->footer.flags |= CEPH_MSG_FOOTER_NOCRC;
518 	con_out_kvec_reset(con);
519 	prepare_write_message_footer(con);
520 
521 	return 1;	/* must return > 0 to indicate success */
522 }
523 
524 /*
525  * write some zeros
526  */
527 static int write_partial_skip(struct ceph_connection *con)
528 {
529 	int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
530 	int ret;
531 
532 	dout("%s %p %d left\n", __func__, con, con->v1.out_skip);
533 	while (con->v1.out_skip > 0) {
534 		size_t size = min(con->v1.out_skip, (int)PAGE_SIZE);
535 
536 		if (size == con->v1.out_skip)
537 			more = MSG_MORE;
538 		ret = ceph_tcp_sendpage(con->sock, ceph_zero_page, 0, size,
539 					more);
540 		if (ret <= 0)
541 			goto out;
542 		con->v1.out_skip -= ret;
543 	}
544 	ret = 1;
545 out:
546 	return ret;
547 }
548 
549 /*
550  * Prepare to read connection handshake, or an ack.
551  */
552 static void prepare_read_banner(struct ceph_connection *con)
553 {
554 	dout("prepare_read_banner %p\n", con);
555 	con->v1.in_base_pos = 0;
556 }
557 
558 static void prepare_read_connect(struct ceph_connection *con)
559 {
560 	dout("prepare_read_connect %p\n", con);
561 	con->v1.in_base_pos = 0;
562 }
563 
564 static void prepare_read_ack(struct ceph_connection *con)
565 {
566 	dout("prepare_read_ack %p\n", con);
567 	con->v1.in_base_pos = 0;
568 }
569 
570 static void prepare_read_seq(struct ceph_connection *con)
571 {
572 	dout("prepare_read_seq %p\n", con);
573 	con->v1.in_base_pos = 0;
574 	con->v1.in_tag = CEPH_MSGR_TAG_SEQ;
575 }
576 
577 static void prepare_read_tag(struct ceph_connection *con)
578 {
579 	dout("prepare_read_tag %p\n", con);
580 	con->v1.in_base_pos = 0;
581 	con->v1.in_tag = CEPH_MSGR_TAG_READY;
582 }
583 
584 static void prepare_read_keepalive_ack(struct ceph_connection *con)
585 {
586 	dout("prepare_read_keepalive_ack %p\n", con);
587 	con->v1.in_base_pos = 0;
588 }
589 
590 /*
591  * Prepare to read a message.
592  */
593 static int prepare_read_message(struct ceph_connection *con)
594 {
595 	dout("prepare_read_message %p\n", con);
596 	BUG_ON(con->in_msg != NULL);
597 	con->v1.in_base_pos = 0;
598 	con->in_front_crc = con->in_middle_crc = con->in_data_crc = 0;
599 	return 0;
600 }
601 
602 static int read_partial(struct ceph_connection *con,
603 			int end, int size, void *object)
604 {
605 	while (con->v1.in_base_pos < end) {
606 		int left = end - con->v1.in_base_pos;
607 		int have = size - left;
608 		int ret = ceph_tcp_recvmsg(con->sock, object + have, left);
609 		if (ret <= 0)
610 			return ret;
611 		con->v1.in_base_pos += ret;
612 	}
613 	return 1;
614 }
615 
616 /*
617  * Read all or part of the connect-side handshake on a new connection
618  */
619 static int read_partial_banner(struct ceph_connection *con)
620 {
621 	int size;
622 	int end;
623 	int ret;
624 
625 	dout("read_partial_banner %p at %d\n", con, con->v1.in_base_pos);
626 
627 	/* peer's banner */
628 	size = strlen(CEPH_BANNER);
629 	end = size;
630 	ret = read_partial(con, end, size, con->v1.in_banner);
631 	if (ret <= 0)
632 		goto out;
633 
634 	size = sizeof(con->v1.actual_peer_addr);
635 	end += size;
636 	ret = read_partial(con, end, size, &con->v1.actual_peer_addr);
637 	if (ret <= 0)
638 		goto out;
639 	ceph_decode_banner_addr(&con->v1.actual_peer_addr);
640 
641 	size = sizeof(con->v1.peer_addr_for_me);
642 	end += size;
643 	ret = read_partial(con, end, size, &con->v1.peer_addr_for_me);
644 	if (ret <= 0)
645 		goto out;
646 	ceph_decode_banner_addr(&con->v1.peer_addr_for_me);
647 
648 out:
649 	return ret;
650 }
651 
652 static int read_partial_connect(struct ceph_connection *con)
653 {
654 	int size;
655 	int end;
656 	int ret;
657 
658 	dout("read_partial_connect %p at %d\n", con, con->v1.in_base_pos);
659 
660 	size = sizeof(con->v1.in_reply);
661 	end = size;
662 	ret = read_partial(con, end, size, &con->v1.in_reply);
663 	if (ret <= 0)
664 		goto out;
665 
666 	if (con->v1.auth) {
667 		size = le32_to_cpu(con->v1.in_reply.authorizer_len);
668 		if (size > con->v1.auth->authorizer_reply_buf_len) {
669 			pr_err("authorizer reply too big: %d > %zu\n", size,
670 			       con->v1.auth->authorizer_reply_buf_len);
671 			ret = -EINVAL;
672 			goto out;
673 		}
674 
675 		end += size;
676 		ret = read_partial(con, end, size,
677 				   con->v1.auth->authorizer_reply_buf);
678 		if (ret <= 0)
679 			goto out;
680 	}
681 
682 	dout("read_partial_connect %p tag %d, con_seq = %u, g_seq = %u\n",
683 	     con, con->v1.in_reply.tag,
684 	     le32_to_cpu(con->v1.in_reply.connect_seq),
685 	     le32_to_cpu(con->v1.in_reply.global_seq));
686 out:
687 	return ret;
688 }
689 
690 /*
691  * Verify the hello banner looks okay.
692  */
693 static int verify_hello(struct ceph_connection *con)
694 {
695 	if (memcmp(con->v1.in_banner, CEPH_BANNER, strlen(CEPH_BANNER))) {
696 		pr_err("connect to %s got bad banner\n",
697 		       ceph_pr_addr(&con->peer_addr));
698 		con->error_msg = "protocol error, bad banner";
699 		return -1;
700 	}
701 	return 0;
702 }
703 
704 static int process_banner(struct ceph_connection *con)
705 {
706 	struct ceph_entity_addr *my_addr = &con->msgr->inst.addr;
707 
708 	dout("process_banner on %p\n", con);
709 
710 	if (verify_hello(con) < 0)
711 		return -1;
712 
713 	/*
714 	 * Make sure the other end is who we wanted.  note that the other
715 	 * end may not yet know their ip address, so if it's 0.0.0.0, give
716 	 * them the benefit of the doubt.
717 	 */
718 	if (memcmp(&con->peer_addr, &con->v1.actual_peer_addr,
719 		   sizeof(con->peer_addr)) != 0 &&
720 	    !(ceph_addr_is_blank(&con->v1.actual_peer_addr) &&
721 	      con->v1.actual_peer_addr.nonce == con->peer_addr.nonce)) {
722 		pr_warn("wrong peer, want %s/%u, got %s/%u\n",
723 			ceph_pr_addr(&con->peer_addr),
724 			le32_to_cpu(con->peer_addr.nonce),
725 			ceph_pr_addr(&con->v1.actual_peer_addr),
726 			le32_to_cpu(con->v1.actual_peer_addr.nonce));
727 		con->error_msg = "wrong peer at address";
728 		return -1;
729 	}
730 
731 	/*
732 	 * did we learn our address?
733 	 */
734 	if (ceph_addr_is_blank(my_addr)) {
735 		memcpy(&my_addr->in_addr,
736 		       &con->v1.peer_addr_for_me.in_addr,
737 		       sizeof(con->v1.peer_addr_for_me.in_addr));
738 		ceph_addr_set_port(my_addr, 0);
739 		ceph_encode_my_addr(con->msgr);
740 		dout("process_banner learned my addr is %s\n",
741 		     ceph_pr_addr(my_addr));
742 	}
743 
744 	return 0;
745 }
746 
747 static int process_connect(struct ceph_connection *con)
748 {
749 	u64 sup_feat = from_msgr(con->msgr)->supported_features;
750 	u64 req_feat = from_msgr(con->msgr)->required_features;
751 	u64 server_feat = le64_to_cpu(con->v1.in_reply.features);
752 	int ret;
753 
754 	dout("process_connect on %p tag %d\n", con, con->v1.in_tag);
755 
756 	if (con->v1.auth) {
757 		int len = le32_to_cpu(con->v1.in_reply.authorizer_len);
758 
759 		/*
760 		 * Any connection that defines ->get_authorizer()
761 		 * should also define ->add_authorizer_challenge() and
762 		 * ->verify_authorizer_reply().
763 		 *
764 		 * See get_connect_authorizer().
765 		 */
766 		if (con->v1.in_reply.tag ==
767 				CEPH_MSGR_TAG_CHALLENGE_AUTHORIZER) {
768 			ret = con->ops->add_authorizer_challenge(
769 				con, con->v1.auth->authorizer_reply_buf, len);
770 			if (ret < 0)
771 				return ret;
772 
773 			con_out_kvec_reset(con);
774 			__prepare_write_connect(con);
775 			prepare_read_connect(con);
776 			return 0;
777 		}
778 
779 		if (len) {
780 			ret = con->ops->verify_authorizer_reply(con);
781 			if (ret < 0) {
782 				con->error_msg = "bad authorize reply";
783 				return ret;
784 			}
785 		}
786 	}
787 
788 	switch (con->v1.in_reply.tag) {
789 	case CEPH_MSGR_TAG_FEATURES:
790 		pr_err("%s%lld %s feature set mismatch,"
791 		       " my %llx < server's %llx, missing %llx\n",
792 		       ENTITY_NAME(con->peer_name),
793 		       ceph_pr_addr(&con->peer_addr),
794 		       sup_feat, server_feat, server_feat & ~sup_feat);
795 		con->error_msg = "missing required protocol features";
796 		return -1;
797 
798 	case CEPH_MSGR_TAG_BADPROTOVER:
799 		pr_err("%s%lld %s protocol version mismatch,"
800 		       " my %d != server's %d\n",
801 		       ENTITY_NAME(con->peer_name),
802 		       ceph_pr_addr(&con->peer_addr),
803 		       le32_to_cpu(con->v1.out_connect.protocol_version),
804 		       le32_to_cpu(con->v1.in_reply.protocol_version));
805 		con->error_msg = "protocol version mismatch";
806 		return -1;
807 
808 	case CEPH_MSGR_TAG_BADAUTHORIZER:
809 		con->v1.auth_retry++;
810 		dout("process_connect %p got BADAUTHORIZER attempt %d\n", con,
811 		     con->v1.auth_retry);
812 		if (con->v1.auth_retry == 2) {
813 			con->error_msg = "connect authorization failure";
814 			return -1;
815 		}
816 		con_out_kvec_reset(con);
817 		ret = prepare_write_connect(con);
818 		if (ret < 0)
819 			return ret;
820 		prepare_read_connect(con);
821 		break;
822 
823 	case CEPH_MSGR_TAG_RESETSESSION:
824 		/*
825 		 * If we connected with a large connect_seq but the peer
826 		 * has no record of a session with us (no connection, or
827 		 * connect_seq == 0), they will send RESETSESION to indicate
828 		 * that they must have reset their session, and may have
829 		 * dropped messages.
830 		 */
831 		dout("process_connect got RESET peer seq %u\n",
832 		     le32_to_cpu(con->v1.in_reply.connect_seq));
833 		pr_info("%s%lld %s session reset\n",
834 			ENTITY_NAME(con->peer_name),
835 			ceph_pr_addr(&con->peer_addr));
836 		ceph_con_reset_session(con);
837 		con_out_kvec_reset(con);
838 		ret = prepare_write_connect(con);
839 		if (ret < 0)
840 			return ret;
841 		prepare_read_connect(con);
842 
843 		/* Tell ceph about it. */
844 		mutex_unlock(&con->mutex);
845 		if (con->ops->peer_reset)
846 			con->ops->peer_reset(con);
847 		mutex_lock(&con->mutex);
848 		if (con->state != CEPH_CON_S_V1_CONNECT_MSG)
849 			return -EAGAIN;
850 		break;
851 
852 	case CEPH_MSGR_TAG_RETRY_SESSION:
853 		/*
854 		 * If we sent a smaller connect_seq than the peer has, try
855 		 * again with a larger value.
856 		 */
857 		dout("process_connect got RETRY_SESSION my seq %u, peer %u\n",
858 		     le32_to_cpu(con->v1.out_connect.connect_seq),
859 		     le32_to_cpu(con->v1.in_reply.connect_seq));
860 		con->v1.connect_seq = le32_to_cpu(con->v1.in_reply.connect_seq);
861 		con_out_kvec_reset(con);
862 		ret = prepare_write_connect(con);
863 		if (ret < 0)
864 			return ret;
865 		prepare_read_connect(con);
866 		break;
867 
868 	case CEPH_MSGR_TAG_RETRY_GLOBAL:
869 		/*
870 		 * If we sent a smaller global_seq than the peer has, try
871 		 * again with a larger value.
872 		 */
873 		dout("process_connect got RETRY_GLOBAL my %u peer_gseq %u\n",
874 		     con->v1.peer_global_seq,
875 		     le32_to_cpu(con->v1.in_reply.global_seq));
876 		ceph_get_global_seq(con->msgr,
877 				    le32_to_cpu(con->v1.in_reply.global_seq));
878 		con_out_kvec_reset(con);
879 		ret = prepare_write_connect(con);
880 		if (ret < 0)
881 			return ret;
882 		prepare_read_connect(con);
883 		break;
884 
885 	case CEPH_MSGR_TAG_SEQ:
886 	case CEPH_MSGR_TAG_READY:
887 		if (req_feat & ~server_feat) {
888 			pr_err("%s%lld %s protocol feature mismatch,"
889 			       " my required %llx > server's %llx, need %llx\n",
890 			       ENTITY_NAME(con->peer_name),
891 			       ceph_pr_addr(&con->peer_addr),
892 			       req_feat, server_feat, req_feat & ~server_feat);
893 			con->error_msg = "missing required protocol features";
894 			return -1;
895 		}
896 
897 		WARN_ON(con->state != CEPH_CON_S_V1_CONNECT_MSG);
898 		con->state = CEPH_CON_S_OPEN;
899 		con->v1.auth_retry = 0;    /* we authenticated; clear flag */
900 		con->v1.peer_global_seq =
901 			le32_to_cpu(con->v1.in_reply.global_seq);
902 		con->v1.connect_seq++;
903 		con->peer_features = server_feat;
904 		dout("process_connect got READY gseq %d cseq %d (%d)\n",
905 		     con->v1.peer_global_seq,
906 		     le32_to_cpu(con->v1.in_reply.connect_seq),
907 		     con->v1.connect_seq);
908 		WARN_ON(con->v1.connect_seq !=
909 			le32_to_cpu(con->v1.in_reply.connect_seq));
910 
911 		if (con->v1.in_reply.flags & CEPH_MSG_CONNECT_LOSSY)
912 			ceph_con_flag_set(con, CEPH_CON_F_LOSSYTX);
913 
914 		con->delay = 0;      /* reset backoff memory */
915 
916 		if (con->v1.in_reply.tag == CEPH_MSGR_TAG_SEQ) {
917 			prepare_write_seq(con);
918 			prepare_read_seq(con);
919 		} else {
920 			prepare_read_tag(con);
921 		}
922 		break;
923 
924 	case CEPH_MSGR_TAG_WAIT:
925 		/*
926 		 * If there is a connection race (we are opening
927 		 * connections to each other), one of us may just have
928 		 * to WAIT.  This shouldn't happen if we are the
929 		 * client.
930 		 */
931 		con->error_msg = "protocol error, got WAIT as client";
932 		return -1;
933 
934 	default:
935 		con->error_msg = "protocol error, garbage tag during connect";
936 		return -1;
937 	}
938 	return 0;
939 }
940 
941 /*
942  * read (part of) an ack
943  */
944 static int read_partial_ack(struct ceph_connection *con)
945 {
946 	int size = sizeof(con->v1.in_temp_ack);
947 	int end = size;
948 
949 	return read_partial(con, end, size, &con->v1.in_temp_ack);
950 }
951 
952 /*
953  * We can finally discard anything that's been acked.
954  */
955 static void process_ack(struct ceph_connection *con)
956 {
957 	u64 ack = le64_to_cpu(con->v1.in_temp_ack);
958 
959 	if (con->v1.in_tag == CEPH_MSGR_TAG_ACK)
960 		ceph_con_discard_sent(con, ack);
961 	else
962 		ceph_con_discard_requeued(con, ack);
963 
964 	prepare_read_tag(con);
965 }
966 
967 static int read_partial_message_section(struct ceph_connection *con,
968 					struct kvec *section,
969 					unsigned int sec_len, u32 *crc)
970 {
971 	int ret, left;
972 
973 	BUG_ON(!section);
974 
975 	while (section->iov_len < sec_len) {
976 		BUG_ON(section->iov_base == NULL);
977 		left = sec_len - section->iov_len;
978 		ret = ceph_tcp_recvmsg(con->sock, (char *)section->iov_base +
979 				       section->iov_len, left);
980 		if (ret <= 0)
981 			return ret;
982 		section->iov_len += ret;
983 	}
984 	if (section->iov_len == sec_len)
985 		*crc = crc32c(0, section->iov_base, section->iov_len);
986 
987 	return 1;
988 }
989 
990 static int read_partial_msg_data(struct ceph_connection *con)
991 {
992 	struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor;
993 	bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
994 	struct page *page;
995 	size_t page_offset;
996 	size_t length;
997 	u32 crc = 0;
998 	int ret;
999 
1000 	if (do_datacrc)
1001 		crc = con->in_data_crc;
1002 	while (cursor->total_resid) {
1003 		if (!cursor->resid) {
1004 			ceph_msg_data_advance(cursor, 0);
1005 			continue;
1006 		}
1007 
1008 		page = ceph_msg_data_next(cursor, &page_offset, &length);
1009 		ret = ceph_tcp_recvpage(con->sock, page, page_offset, length);
1010 		if (ret <= 0) {
1011 			if (do_datacrc)
1012 				con->in_data_crc = crc;
1013 
1014 			return ret;
1015 		}
1016 
1017 		if (do_datacrc)
1018 			crc = ceph_crc32c_page(crc, page, page_offset, ret);
1019 		ceph_msg_data_advance(cursor, (size_t)ret);
1020 	}
1021 	if (do_datacrc)
1022 		con->in_data_crc = crc;
1023 
1024 	return 1;	/* must return > 0 to indicate success */
1025 }
1026 
1027 static int read_partial_msg_data_bounce(struct ceph_connection *con)
1028 {
1029 	struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor;
1030 	struct page *page;
1031 	size_t off, len;
1032 	u32 crc;
1033 	int ret;
1034 
1035 	if (unlikely(!con->bounce_page)) {
1036 		con->bounce_page = alloc_page(GFP_NOIO);
1037 		if (!con->bounce_page) {
1038 			pr_err("failed to allocate bounce page\n");
1039 			return -ENOMEM;
1040 		}
1041 	}
1042 
1043 	crc = con->in_data_crc;
1044 	while (cursor->total_resid) {
1045 		if (!cursor->resid) {
1046 			ceph_msg_data_advance(cursor, 0);
1047 			continue;
1048 		}
1049 
1050 		page = ceph_msg_data_next(cursor, &off, &len);
1051 		ret = ceph_tcp_recvpage(con->sock, con->bounce_page, 0, len);
1052 		if (ret <= 0) {
1053 			con->in_data_crc = crc;
1054 			return ret;
1055 		}
1056 
1057 		crc = crc32c(crc, page_address(con->bounce_page), ret);
1058 		memcpy_to_page(page, off, page_address(con->bounce_page), ret);
1059 
1060 		ceph_msg_data_advance(cursor, ret);
1061 	}
1062 	con->in_data_crc = crc;
1063 
1064 	return 1;	/* must return > 0 to indicate success */
1065 }
1066 
1067 /*
1068  * read (part of) a message.
1069  */
1070 static int read_partial_message(struct ceph_connection *con)
1071 {
1072 	struct ceph_msg *m = con->in_msg;
1073 	int size;
1074 	int end;
1075 	int ret;
1076 	unsigned int front_len, middle_len, data_len;
1077 	bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
1078 	bool need_sign = (con->peer_features & CEPH_FEATURE_MSG_AUTH);
1079 	u64 seq;
1080 	u32 crc;
1081 
1082 	dout("read_partial_message con %p msg %p\n", con, m);
1083 
1084 	/* header */
1085 	size = sizeof(con->v1.in_hdr);
1086 	end = size;
1087 	ret = read_partial(con, end, size, &con->v1.in_hdr);
1088 	if (ret <= 0)
1089 		return ret;
1090 
1091 	crc = crc32c(0, &con->v1.in_hdr, offsetof(struct ceph_msg_header, crc));
1092 	if (cpu_to_le32(crc) != con->v1.in_hdr.crc) {
1093 		pr_err("read_partial_message bad hdr crc %u != expected %u\n",
1094 		       crc, con->v1.in_hdr.crc);
1095 		return -EBADMSG;
1096 	}
1097 
1098 	front_len = le32_to_cpu(con->v1.in_hdr.front_len);
1099 	if (front_len > CEPH_MSG_MAX_FRONT_LEN)
1100 		return -EIO;
1101 	middle_len = le32_to_cpu(con->v1.in_hdr.middle_len);
1102 	if (middle_len > CEPH_MSG_MAX_MIDDLE_LEN)
1103 		return -EIO;
1104 	data_len = le32_to_cpu(con->v1.in_hdr.data_len);
1105 	if (data_len > CEPH_MSG_MAX_DATA_LEN)
1106 		return -EIO;
1107 
1108 	/* verify seq# */
1109 	seq = le64_to_cpu(con->v1.in_hdr.seq);
1110 	if ((s64)seq - (s64)con->in_seq < 1) {
1111 		pr_info("skipping %s%lld %s seq %lld expected %lld\n",
1112 			ENTITY_NAME(con->peer_name),
1113 			ceph_pr_addr(&con->peer_addr),
1114 			seq, con->in_seq + 1);
1115 		con->v1.in_base_pos = -front_len - middle_len - data_len -
1116 				      sizeof_footer(con);
1117 		con->v1.in_tag = CEPH_MSGR_TAG_READY;
1118 		return 1;
1119 	} else if ((s64)seq - (s64)con->in_seq > 1) {
1120 		pr_err("read_partial_message bad seq %lld expected %lld\n",
1121 		       seq, con->in_seq + 1);
1122 		con->error_msg = "bad message sequence # for incoming message";
1123 		return -EBADE;
1124 	}
1125 
1126 	/* allocate message? */
1127 	if (!con->in_msg) {
1128 		int skip = 0;
1129 
1130 		dout("got hdr type %d front %d data %d\n", con->v1.in_hdr.type,
1131 		     front_len, data_len);
1132 		ret = ceph_con_in_msg_alloc(con, &con->v1.in_hdr, &skip);
1133 		if (ret < 0)
1134 			return ret;
1135 
1136 		BUG_ON((!con->in_msg) ^ skip);
1137 		if (skip) {
1138 			/* skip this message */
1139 			dout("alloc_msg said skip message\n");
1140 			con->v1.in_base_pos = -front_len - middle_len -
1141 					      data_len - sizeof_footer(con);
1142 			con->v1.in_tag = CEPH_MSGR_TAG_READY;
1143 			con->in_seq++;
1144 			return 1;
1145 		}
1146 
1147 		BUG_ON(!con->in_msg);
1148 		BUG_ON(con->in_msg->con != con);
1149 		m = con->in_msg;
1150 		m->front.iov_len = 0;    /* haven't read it yet */
1151 		if (m->middle)
1152 			m->middle->vec.iov_len = 0;
1153 
1154 		/* prepare for data payload, if any */
1155 
1156 		if (data_len)
1157 			prepare_message_data(con->in_msg, data_len);
1158 	}
1159 
1160 	/* front */
1161 	ret = read_partial_message_section(con, &m->front, front_len,
1162 					   &con->in_front_crc);
1163 	if (ret <= 0)
1164 		return ret;
1165 
1166 	/* middle */
1167 	if (m->middle) {
1168 		ret = read_partial_message_section(con, &m->middle->vec,
1169 						   middle_len,
1170 						   &con->in_middle_crc);
1171 		if (ret <= 0)
1172 			return ret;
1173 	}
1174 
1175 	/* (page) data */
1176 	if (data_len) {
1177 		if (!m->num_data_items)
1178 			return -EIO;
1179 
1180 		if (ceph_test_opt(from_msgr(con->msgr), RXBOUNCE))
1181 			ret = read_partial_msg_data_bounce(con);
1182 		else
1183 			ret = read_partial_msg_data(con);
1184 		if (ret <= 0)
1185 			return ret;
1186 	}
1187 
1188 	/* footer */
1189 	size = sizeof_footer(con);
1190 	end += size;
1191 	ret = read_partial(con, end, size, &m->footer);
1192 	if (ret <= 0)
1193 		return ret;
1194 
1195 	if (!need_sign) {
1196 		m->footer.flags = m->old_footer.flags;
1197 		m->footer.sig = 0;
1198 	}
1199 
1200 	dout("read_partial_message got msg %p %d (%u) + %d (%u) + %d (%u)\n",
1201 	     m, front_len, m->footer.front_crc, middle_len,
1202 	     m->footer.middle_crc, data_len, m->footer.data_crc);
1203 
1204 	/* crc ok? */
1205 	if (con->in_front_crc != le32_to_cpu(m->footer.front_crc)) {
1206 		pr_err("read_partial_message %p front crc %u != exp. %u\n",
1207 		       m, con->in_front_crc, m->footer.front_crc);
1208 		return -EBADMSG;
1209 	}
1210 	if (con->in_middle_crc != le32_to_cpu(m->footer.middle_crc)) {
1211 		pr_err("read_partial_message %p middle crc %u != exp %u\n",
1212 		       m, con->in_middle_crc, m->footer.middle_crc);
1213 		return -EBADMSG;
1214 	}
1215 	if (do_datacrc &&
1216 	    (m->footer.flags & CEPH_MSG_FOOTER_NOCRC) == 0 &&
1217 	    con->in_data_crc != le32_to_cpu(m->footer.data_crc)) {
1218 		pr_err("read_partial_message %p data crc %u != exp. %u\n", m,
1219 		       con->in_data_crc, le32_to_cpu(m->footer.data_crc));
1220 		return -EBADMSG;
1221 	}
1222 
1223 	if (need_sign && con->ops->check_message_signature &&
1224 	    con->ops->check_message_signature(m)) {
1225 		pr_err("read_partial_message %p signature check failed\n", m);
1226 		return -EBADMSG;
1227 	}
1228 
1229 	return 1; /* done! */
1230 }
1231 
1232 static int read_keepalive_ack(struct ceph_connection *con)
1233 {
1234 	struct ceph_timespec ceph_ts;
1235 	size_t size = sizeof(ceph_ts);
1236 	int ret = read_partial(con, size, size, &ceph_ts);
1237 	if (ret <= 0)
1238 		return ret;
1239 	ceph_decode_timespec64(&con->last_keepalive_ack, &ceph_ts);
1240 	prepare_read_tag(con);
1241 	return 1;
1242 }
1243 
1244 /*
1245  * Read what we can from the socket.
1246  */
1247 int ceph_con_v1_try_read(struct ceph_connection *con)
1248 {
1249 	int ret = -1;
1250 
1251 more:
1252 	dout("try_read start %p state %d\n", con, con->state);
1253 	if (con->state != CEPH_CON_S_V1_BANNER &&
1254 	    con->state != CEPH_CON_S_V1_CONNECT_MSG &&
1255 	    con->state != CEPH_CON_S_OPEN)
1256 		return 0;
1257 
1258 	BUG_ON(!con->sock);
1259 
1260 	dout("try_read tag %d in_base_pos %d\n", con->v1.in_tag,
1261 	     con->v1.in_base_pos);
1262 
1263 	if (con->state == CEPH_CON_S_V1_BANNER) {
1264 		ret = read_partial_banner(con);
1265 		if (ret <= 0)
1266 			goto out;
1267 		ret = process_banner(con);
1268 		if (ret < 0)
1269 			goto out;
1270 
1271 		con->state = CEPH_CON_S_V1_CONNECT_MSG;
1272 
1273 		/*
1274 		 * Received banner is good, exchange connection info.
1275 		 * Do not reset out_kvec, as sending our banner raced
1276 		 * with receiving peer banner after connect completed.
1277 		 */
1278 		ret = prepare_write_connect(con);
1279 		if (ret < 0)
1280 			goto out;
1281 		prepare_read_connect(con);
1282 
1283 		/* Send connection info before awaiting response */
1284 		goto out;
1285 	}
1286 
1287 	if (con->state == CEPH_CON_S_V1_CONNECT_MSG) {
1288 		ret = read_partial_connect(con);
1289 		if (ret <= 0)
1290 			goto out;
1291 		ret = process_connect(con);
1292 		if (ret < 0)
1293 			goto out;
1294 		goto more;
1295 	}
1296 
1297 	WARN_ON(con->state != CEPH_CON_S_OPEN);
1298 
1299 	if (con->v1.in_base_pos < 0) {
1300 		/*
1301 		 * skipping + discarding content.
1302 		 */
1303 		ret = ceph_tcp_recvmsg(con->sock, NULL, -con->v1.in_base_pos);
1304 		if (ret <= 0)
1305 			goto out;
1306 		dout("skipped %d / %d bytes\n", ret, -con->v1.in_base_pos);
1307 		con->v1.in_base_pos += ret;
1308 		if (con->v1.in_base_pos)
1309 			goto more;
1310 	}
1311 	if (con->v1.in_tag == CEPH_MSGR_TAG_READY) {
1312 		/*
1313 		 * what's next?
1314 		 */
1315 		ret = ceph_tcp_recvmsg(con->sock, &con->v1.in_tag, 1);
1316 		if (ret <= 0)
1317 			goto out;
1318 		dout("try_read got tag %d\n", con->v1.in_tag);
1319 		switch (con->v1.in_tag) {
1320 		case CEPH_MSGR_TAG_MSG:
1321 			prepare_read_message(con);
1322 			break;
1323 		case CEPH_MSGR_TAG_ACK:
1324 			prepare_read_ack(con);
1325 			break;
1326 		case CEPH_MSGR_TAG_KEEPALIVE2_ACK:
1327 			prepare_read_keepalive_ack(con);
1328 			break;
1329 		case CEPH_MSGR_TAG_CLOSE:
1330 			ceph_con_close_socket(con);
1331 			con->state = CEPH_CON_S_CLOSED;
1332 			goto out;
1333 		default:
1334 			goto bad_tag;
1335 		}
1336 	}
1337 	if (con->v1.in_tag == CEPH_MSGR_TAG_MSG) {
1338 		ret = read_partial_message(con);
1339 		if (ret <= 0) {
1340 			switch (ret) {
1341 			case -EBADMSG:
1342 				con->error_msg = "bad crc/signature";
1343 				fallthrough;
1344 			case -EBADE:
1345 				ret = -EIO;
1346 				break;
1347 			case -EIO:
1348 				con->error_msg = "io error";
1349 				break;
1350 			}
1351 			goto out;
1352 		}
1353 		if (con->v1.in_tag == CEPH_MSGR_TAG_READY)
1354 			goto more;
1355 		ceph_con_process_message(con);
1356 		if (con->state == CEPH_CON_S_OPEN)
1357 			prepare_read_tag(con);
1358 		goto more;
1359 	}
1360 	if (con->v1.in_tag == CEPH_MSGR_TAG_ACK ||
1361 	    con->v1.in_tag == CEPH_MSGR_TAG_SEQ) {
1362 		/*
1363 		 * the final handshake seq exchange is semantically
1364 		 * equivalent to an ACK
1365 		 */
1366 		ret = read_partial_ack(con);
1367 		if (ret <= 0)
1368 			goto out;
1369 		process_ack(con);
1370 		goto more;
1371 	}
1372 	if (con->v1.in_tag == CEPH_MSGR_TAG_KEEPALIVE2_ACK) {
1373 		ret = read_keepalive_ack(con);
1374 		if (ret <= 0)
1375 			goto out;
1376 		goto more;
1377 	}
1378 
1379 out:
1380 	dout("try_read done on %p ret %d\n", con, ret);
1381 	return ret;
1382 
1383 bad_tag:
1384 	pr_err("try_read bad tag %d\n", con->v1.in_tag);
1385 	con->error_msg = "protocol error, garbage tag";
1386 	ret = -1;
1387 	goto out;
1388 }
1389 
1390 /*
1391  * Write something to the socket.  Called in a worker thread when the
1392  * socket appears to be writeable and we have something ready to send.
1393  */
1394 int ceph_con_v1_try_write(struct ceph_connection *con)
1395 {
1396 	int ret = 1;
1397 
1398 	dout("try_write start %p state %d\n", con, con->state);
1399 	if (con->state != CEPH_CON_S_PREOPEN &&
1400 	    con->state != CEPH_CON_S_V1_BANNER &&
1401 	    con->state != CEPH_CON_S_V1_CONNECT_MSG &&
1402 	    con->state != CEPH_CON_S_OPEN)
1403 		return 0;
1404 
1405 	/* open the socket first? */
1406 	if (con->state == CEPH_CON_S_PREOPEN) {
1407 		BUG_ON(con->sock);
1408 		con->state = CEPH_CON_S_V1_BANNER;
1409 
1410 		con_out_kvec_reset(con);
1411 		prepare_write_banner(con);
1412 		prepare_read_banner(con);
1413 
1414 		BUG_ON(con->in_msg);
1415 		con->v1.in_tag = CEPH_MSGR_TAG_READY;
1416 		dout("try_write initiating connect on %p new state %d\n",
1417 		     con, con->state);
1418 		ret = ceph_tcp_connect(con);
1419 		if (ret < 0) {
1420 			con->error_msg = "connect error";
1421 			goto out;
1422 		}
1423 	}
1424 
1425 more:
1426 	dout("try_write out_kvec_bytes %d\n", con->v1.out_kvec_bytes);
1427 	BUG_ON(!con->sock);
1428 
1429 	/* kvec data queued? */
1430 	if (con->v1.out_kvec_left) {
1431 		ret = write_partial_kvec(con);
1432 		if (ret <= 0)
1433 			goto out;
1434 	}
1435 	if (con->v1.out_skip) {
1436 		ret = write_partial_skip(con);
1437 		if (ret <= 0)
1438 			goto out;
1439 	}
1440 
1441 	/* msg pages? */
1442 	if (con->out_msg) {
1443 		if (con->v1.out_msg_done) {
1444 			ceph_msg_put(con->out_msg);
1445 			con->out_msg = NULL;   /* we're done with this one */
1446 			goto do_next;
1447 		}
1448 
1449 		ret = write_partial_message_data(con);
1450 		if (ret == 1)
1451 			goto more;  /* we need to send the footer, too! */
1452 		if (ret == 0)
1453 			goto out;
1454 		if (ret < 0) {
1455 			dout("try_write write_partial_message_data err %d\n",
1456 			     ret);
1457 			goto out;
1458 		}
1459 	}
1460 
1461 do_next:
1462 	if (con->state == CEPH_CON_S_OPEN) {
1463 		if (ceph_con_flag_test_and_clear(con,
1464 				CEPH_CON_F_KEEPALIVE_PENDING)) {
1465 			prepare_write_keepalive(con);
1466 			goto more;
1467 		}
1468 		/* is anything else pending? */
1469 		if (!list_empty(&con->out_queue)) {
1470 			prepare_write_message(con);
1471 			goto more;
1472 		}
1473 		if (con->in_seq > con->in_seq_acked) {
1474 			prepare_write_ack(con);
1475 			goto more;
1476 		}
1477 	}
1478 
1479 	/* Nothing to do! */
1480 	ceph_con_flag_clear(con, CEPH_CON_F_WRITE_PENDING);
1481 	dout("try_write nothing else to write.\n");
1482 	ret = 0;
1483 out:
1484 	dout("try_write done on %p ret %d\n", con, ret);
1485 	return ret;
1486 }
1487 
1488 void ceph_con_v1_revoke(struct ceph_connection *con)
1489 {
1490 	struct ceph_msg *msg = con->out_msg;
1491 
1492 	WARN_ON(con->v1.out_skip);
1493 	/* footer */
1494 	if (con->v1.out_msg_done) {
1495 		con->v1.out_skip += con_out_kvec_skip(con);
1496 	} else {
1497 		WARN_ON(!msg->data_length);
1498 		con->v1.out_skip += sizeof_footer(con);
1499 	}
1500 	/* data, middle, front */
1501 	if (msg->data_length)
1502 		con->v1.out_skip += msg->cursor.total_resid;
1503 	if (msg->middle)
1504 		con->v1.out_skip += con_out_kvec_skip(con);
1505 	con->v1.out_skip += con_out_kvec_skip(con);
1506 
1507 	dout("%s con %p out_kvec_bytes %d out_skip %d\n", __func__, con,
1508 	     con->v1.out_kvec_bytes, con->v1.out_skip);
1509 }
1510 
1511 void ceph_con_v1_revoke_incoming(struct ceph_connection *con)
1512 {
1513 	unsigned int front_len = le32_to_cpu(con->v1.in_hdr.front_len);
1514 	unsigned int middle_len = le32_to_cpu(con->v1.in_hdr.middle_len);
1515 	unsigned int data_len = le32_to_cpu(con->v1.in_hdr.data_len);
1516 
1517 	/* skip rest of message */
1518 	con->v1.in_base_pos = con->v1.in_base_pos -
1519 			sizeof(struct ceph_msg_header) -
1520 			front_len -
1521 			middle_len -
1522 			data_len -
1523 			sizeof(struct ceph_msg_footer);
1524 
1525 	con->v1.in_tag = CEPH_MSGR_TAG_READY;
1526 	con->in_seq++;
1527 
1528 	dout("%s con %p in_base_pos %d\n", __func__, con, con->v1.in_base_pos);
1529 }
1530 
1531 bool ceph_con_v1_opened(struct ceph_connection *con)
1532 {
1533 	return con->v1.connect_seq;
1534 }
1535 
1536 void ceph_con_v1_reset_session(struct ceph_connection *con)
1537 {
1538 	con->v1.connect_seq = 0;
1539 	con->v1.peer_global_seq = 0;
1540 }
1541 
1542 void ceph_con_v1_reset_protocol(struct ceph_connection *con)
1543 {
1544 	con->v1.out_skip = 0;
1545 }
1546