xref: /linux/net/ceph/mon_client.c (revision 00498b994113a871a556f7ff24a4cf8a00611700)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3 
4 #include <linux/module.h>
5 #include <linux/types.h>
6 #include <linux/slab.h>
7 #include <linux/random.h>
8 #include <linux/sched.h>
9 
10 #include <linux/ceph/ceph_features.h>
11 #include <linux/ceph/mon_client.h>
12 #include <linux/ceph/libceph.h>
13 #include <linux/ceph/debugfs.h>
14 #include <linux/ceph/decode.h>
15 #include <linux/ceph/auth.h>
16 
17 /*
18  * Interact with Ceph monitor cluster.  Handle requests for new map
19  * versions, and periodically resend as needed.  Also implement
20  * statfs() and umount().
21  *
22  * A small cluster of Ceph "monitors" are responsible for managing critical
23  * cluster configuration and state information.  An odd number (e.g., 3, 5)
24  * of cmon daemons use a modified version of the Paxos part-time parliament
25  * algorithm to manage the MDS map (mds cluster membership), OSD map, and
26  * list of clients who have mounted the file system.
27  *
28  * We maintain an open, active session with a monitor at all times in order to
29  * receive timely MDSMap updates.  We periodically send a keepalive byte on the
30  * TCP socket to ensure we detect a failure.  If the connection does break, we
31  * randomly hunt for a new monitor.  Once the connection is reestablished, we
32  * resend any outstanding requests.
33  */
34 
35 static const struct ceph_connection_operations mon_con_ops;
36 
37 static int __validate_auth(struct ceph_mon_client *monc);
38 
39 static int decode_mon_info(void **p, void *end, bool msgr2,
40 			   struct ceph_entity_addr *addr)
41 {
42 	void *mon_info_end;
43 	u32 struct_len;
44 	u8 struct_v;
45 	int ret;
46 
47 	ret = ceph_start_decoding(p, end, 1, "mon_info_t", &struct_v,
48 				  &struct_len);
49 	if (ret)
50 		return ret;
51 
52 	mon_info_end = *p + struct_len;
53 	ceph_decode_skip_string(p, end, e_inval);  /* skip mon name */
54 	ret = ceph_decode_entity_addrvec(p, end, msgr2, addr);
55 	if (ret)
56 		return ret;
57 
58 	*p = mon_info_end;
59 	return 0;
60 
61 e_inval:
62 	return -EINVAL;
63 }
64 
65 /*
66  * Decode a monmap blob (e.g., during mount).
67  *
68  * Assume MonMap v3 (i.e. encoding with MONNAMES and MONENC).
69  */
70 static struct ceph_monmap *ceph_monmap_decode(void **p, void *end, bool msgr2)
71 {
72 	struct ceph_monmap *monmap = NULL;
73 	struct ceph_fsid fsid;
74 	u32 struct_len;
75 	int blob_len;
76 	int num_mon;
77 	u8 struct_v;
78 	u32 epoch;
79 	int ret;
80 	int i;
81 
82 	ceph_decode_32_safe(p, end, blob_len, e_inval);
83 	ceph_decode_need(p, end, blob_len, e_inval);
84 
85 	ret = ceph_start_decoding(p, end, 6, "monmap", &struct_v, &struct_len);
86 	if (ret)
87 		goto fail;
88 
89 	dout("%s struct_v %d\n", __func__, struct_v);
90 	ceph_decode_copy_safe(p, end, &fsid, sizeof(fsid), e_inval);
91 	ceph_decode_32_safe(p, end, epoch, e_inval);
92 	if (struct_v >= 6) {
93 		u32 feat_struct_len;
94 		u8 feat_struct_v;
95 
96 		*p += sizeof(struct ceph_timespec);  /* skip last_changed */
97 		*p += sizeof(struct ceph_timespec);  /* skip created */
98 
99 		ret = ceph_start_decoding(p, end, 1, "mon_feature_t",
100 					  &feat_struct_v, &feat_struct_len);
101 		if (ret)
102 			goto fail;
103 
104 		*p += feat_struct_len;  /* skip persistent_features */
105 
106 		ret = ceph_start_decoding(p, end, 1, "mon_feature_t",
107 					  &feat_struct_v, &feat_struct_len);
108 		if (ret)
109 			goto fail;
110 
111 		*p += feat_struct_len;  /* skip optional_features */
112 	}
113 	ceph_decode_32_safe(p, end, num_mon, e_inval);
114 
115 	dout("%s fsid %pU epoch %u num_mon %d\n", __func__, &fsid, epoch,
116 	     num_mon);
117 	if (num_mon > CEPH_MAX_MON)
118 		goto e_inval;
119 
120 	monmap = kmalloc(struct_size(monmap, mon_inst, num_mon), GFP_NOIO);
121 	if (!monmap) {
122 		ret = -ENOMEM;
123 		goto fail;
124 	}
125 	monmap->fsid = fsid;
126 	monmap->epoch = epoch;
127 	monmap->num_mon = num_mon;
128 
129 	/* legacy_mon_addr map or mon_info map */
130 	for (i = 0; i < num_mon; i++) {
131 		struct ceph_entity_inst *inst = &monmap->mon_inst[i];
132 
133 		ceph_decode_skip_string(p, end, e_inval);  /* skip mon name */
134 		inst->name.type = CEPH_ENTITY_TYPE_MON;
135 		inst->name.num = cpu_to_le64(i);
136 
137 		if (struct_v >= 6)
138 			ret = decode_mon_info(p, end, msgr2, &inst->addr);
139 		else
140 			ret = ceph_decode_entity_addr(p, end, &inst->addr);
141 		if (ret)
142 			goto fail;
143 
144 		dout("%s mon%d addr %s\n", __func__, i,
145 		     ceph_pr_addr(&inst->addr));
146 	}
147 
148 	return monmap;
149 
150 e_inval:
151 	ret = -EINVAL;
152 fail:
153 	kfree(monmap);
154 	return ERR_PTR(ret);
155 }
156 
157 /*
158  * return true if *addr is included in the monmap.
159  */
160 int ceph_monmap_contains(struct ceph_monmap *m, struct ceph_entity_addr *addr)
161 {
162 	int i;
163 
164 	for (i = 0; i < m->num_mon; i++) {
165 		if (ceph_addr_equal_no_type(addr, &m->mon_inst[i].addr))
166 			return 1;
167 	}
168 
169 	return 0;
170 }
171 
172 /*
173  * Send an auth request.
174  */
175 static void __send_prepared_auth_request(struct ceph_mon_client *monc, int len)
176 {
177 	monc->pending_auth = 1;
178 	monc->m_auth->front.iov_len = len;
179 	monc->m_auth->hdr.front_len = cpu_to_le32(len);
180 	ceph_msg_revoke(monc->m_auth);
181 	ceph_msg_get(monc->m_auth);  /* keep our ref */
182 	ceph_con_send(&monc->con, monc->m_auth);
183 }
184 
185 /*
186  * Close monitor session, if any.
187  */
188 static void __close_session(struct ceph_mon_client *monc)
189 {
190 	dout("__close_session closing mon%d\n", monc->cur_mon);
191 	ceph_msg_revoke(monc->m_auth);
192 	ceph_msg_revoke_incoming(monc->m_auth_reply);
193 	ceph_msg_revoke(monc->m_subscribe);
194 	ceph_msg_revoke_incoming(monc->m_subscribe_ack);
195 	ceph_con_close(&monc->con);
196 
197 	monc->pending_auth = 0;
198 	ceph_auth_reset(monc->auth);
199 }
200 
201 /*
202  * Pick a new monitor at random and set cur_mon.  If we are repicking
203  * (i.e. cur_mon is already set), be sure to pick a different one.
204  */
205 static void pick_new_mon(struct ceph_mon_client *monc)
206 {
207 	int old_mon = monc->cur_mon;
208 
209 	BUG_ON(monc->monmap->num_mon < 1);
210 
211 	if (monc->monmap->num_mon == 1) {
212 		monc->cur_mon = 0;
213 	} else {
214 		int max = monc->monmap->num_mon;
215 		int o = -1;
216 		int n;
217 
218 		if (monc->cur_mon >= 0) {
219 			if (monc->cur_mon < monc->monmap->num_mon)
220 				o = monc->cur_mon;
221 			if (o >= 0)
222 				max--;
223 		}
224 
225 		n = prandom_u32() % max;
226 		if (o >= 0 && n >= o)
227 			n++;
228 
229 		monc->cur_mon = n;
230 	}
231 
232 	dout("%s mon%d -> mon%d out of %d mons\n", __func__, old_mon,
233 	     monc->cur_mon, monc->monmap->num_mon);
234 }
235 
236 /*
237  * Open a session with a new monitor.
238  */
239 static void __open_session(struct ceph_mon_client *monc)
240 {
241 	int ret;
242 
243 	pick_new_mon(monc);
244 
245 	monc->hunting = true;
246 	if (monc->had_a_connection) {
247 		monc->hunt_mult *= CEPH_MONC_HUNT_BACKOFF;
248 		if (monc->hunt_mult > CEPH_MONC_HUNT_MAX_MULT)
249 			monc->hunt_mult = CEPH_MONC_HUNT_MAX_MULT;
250 	}
251 
252 	monc->sub_renew_after = jiffies; /* i.e., expired */
253 	monc->sub_renew_sent = 0;
254 
255 	dout("%s opening mon%d\n", __func__, monc->cur_mon);
256 	ceph_con_open(&monc->con, CEPH_ENTITY_TYPE_MON, monc->cur_mon,
257 		      &monc->monmap->mon_inst[monc->cur_mon].addr);
258 
259 	/*
260 	 * send an initial keepalive to ensure our timestamp is valid
261 	 * by the time we are in an OPENED state
262 	 */
263 	ceph_con_keepalive(&monc->con);
264 
265 	/* initiate authentication handshake */
266 	ret = ceph_auth_build_hello(monc->auth,
267 				    monc->m_auth->front.iov_base,
268 				    monc->m_auth->front_alloc_len);
269 	BUG_ON(ret <= 0);
270 	__send_prepared_auth_request(monc, ret);
271 }
272 
273 static void reopen_session(struct ceph_mon_client *monc)
274 {
275 	if (!monc->hunting)
276 		pr_info("mon%d %s session lost, hunting for new mon\n",
277 		    monc->cur_mon, ceph_pr_addr(&monc->con.peer_addr));
278 
279 	__close_session(monc);
280 	__open_session(monc);
281 }
282 
283 void ceph_monc_reopen_session(struct ceph_mon_client *monc)
284 {
285 	mutex_lock(&monc->mutex);
286 	reopen_session(monc);
287 	mutex_unlock(&monc->mutex);
288 }
289 
290 static void un_backoff(struct ceph_mon_client *monc)
291 {
292 	monc->hunt_mult /= 2; /* reduce by 50% */
293 	if (monc->hunt_mult < 1)
294 		monc->hunt_mult = 1;
295 	dout("%s hunt_mult now %d\n", __func__, monc->hunt_mult);
296 }
297 
298 /*
299  * Reschedule delayed work timer.
300  */
301 static void __schedule_delayed(struct ceph_mon_client *monc)
302 {
303 	unsigned long delay;
304 
305 	if (monc->hunting)
306 		delay = CEPH_MONC_HUNT_INTERVAL * monc->hunt_mult;
307 	else
308 		delay = CEPH_MONC_PING_INTERVAL;
309 
310 	dout("__schedule_delayed after %lu\n", delay);
311 	mod_delayed_work(system_wq, &monc->delayed_work,
312 			 round_jiffies_relative(delay));
313 }
314 
315 const char *ceph_sub_str[] = {
316 	[CEPH_SUB_MONMAP] = "monmap",
317 	[CEPH_SUB_OSDMAP] = "osdmap",
318 	[CEPH_SUB_FSMAP]  = "fsmap.user",
319 	[CEPH_SUB_MDSMAP] = "mdsmap",
320 };
321 
322 /*
323  * Send subscribe request for one or more maps, according to
324  * monc->subs.
325  */
326 static void __send_subscribe(struct ceph_mon_client *monc)
327 {
328 	struct ceph_msg *msg = monc->m_subscribe;
329 	void *p = msg->front.iov_base;
330 	void *const end = p + msg->front_alloc_len;
331 	int num = 0;
332 	int i;
333 
334 	dout("%s sent %lu\n", __func__, monc->sub_renew_sent);
335 
336 	BUG_ON(monc->cur_mon < 0);
337 
338 	if (!monc->sub_renew_sent)
339 		monc->sub_renew_sent = jiffies | 1; /* never 0 */
340 
341 	msg->hdr.version = cpu_to_le16(2);
342 
343 	for (i = 0; i < ARRAY_SIZE(monc->subs); i++) {
344 		if (monc->subs[i].want)
345 			num++;
346 	}
347 	BUG_ON(num < 1); /* monmap sub is always there */
348 	ceph_encode_32(&p, num);
349 	for (i = 0; i < ARRAY_SIZE(monc->subs); i++) {
350 		char buf[32];
351 		int len;
352 
353 		if (!monc->subs[i].want)
354 			continue;
355 
356 		len = sprintf(buf, "%s", ceph_sub_str[i]);
357 		if (i == CEPH_SUB_MDSMAP &&
358 		    monc->fs_cluster_id != CEPH_FS_CLUSTER_ID_NONE)
359 			len += sprintf(buf + len, ".%d", monc->fs_cluster_id);
360 
361 		dout("%s %s start %llu flags 0x%x\n", __func__, buf,
362 		     le64_to_cpu(monc->subs[i].item.start),
363 		     monc->subs[i].item.flags);
364 		ceph_encode_string(&p, end, buf, len);
365 		memcpy(p, &monc->subs[i].item, sizeof(monc->subs[i].item));
366 		p += sizeof(monc->subs[i].item);
367 	}
368 
369 	BUG_ON(p > end);
370 	msg->front.iov_len = p - msg->front.iov_base;
371 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
372 	ceph_msg_revoke(msg);
373 	ceph_con_send(&monc->con, ceph_msg_get(msg));
374 }
375 
376 static void handle_subscribe_ack(struct ceph_mon_client *monc,
377 				 struct ceph_msg *msg)
378 {
379 	unsigned int seconds;
380 	struct ceph_mon_subscribe_ack *h = msg->front.iov_base;
381 
382 	if (msg->front.iov_len < sizeof(*h))
383 		goto bad;
384 	seconds = le32_to_cpu(h->duration);
385 
386 	mutex_lock(&monc->mutex);
387 	if (monc->sub_renew_sent) {
388 		/*
389 		 * This is only needed for legacy (infernalis or older)
390 		 * MONs -- see delayed_work().
391 		 */
392 		monc->sub_renew_after = monc->sub_renew_sent +
393 					    (seconds >> 1) * HZ - 1;
394 		dout("%s sent %lu duration %d renew after %lu\n", __func__,
395 		     monc->sub_renew_sent, seconds, monc->sub_renew_after);
396 		monc->sub_renew_sent = 0;
397 	} else {
398 		dout("%s sent %lu renew after %lu, ignoring\n", __func__,
399 		     monc->sub_renew_sent, monc->sub_renew_after);
400 	}
401 	mutex_unlock(&monc->mutex);
402 	return;
403 bad:
404 	pr_err("got corrupt subscribe-ack msg\n");
405 	ceph_msg_dump(msg);
406 }
407 
408 /*
409  * Register interest in a map
410  *
411  * @sub: one of CEPH_SUB_*
412  * @epoch: X for "every map since X", or 0 for "just the latest"
413  */
414 static bool __ceph_monc_want_map(struct ceph_mon_client *monc, int sub,
415 				 u32 epoch, bool continuous)
416 {
417 	__le64 start = cpu_to_le64(epoch);
418 	u8 flags = !continuous ? CEPH_SUBSCRIBE_ONETIME : 0;
419 
420 	dout("%s %s epoch %u continuous %d\n", __func__, ceph_sub_str[sub],
421 	     epoch, continuous);
422 
423 	if (monc->subs[sub].want &&
424 	    monc->subs[sub].item.start == start &&
425 	    monc->subs[sub].item.flags == flags)
426 		return false;
427 
428 	monc->subs[sub].item.start = start;
429 	monc->subs[sub].item.flags = flags;
430 	monc->subs[sub].want = true;
431 
432 	return true;
433 }
434 
435 bool ceph_monc_want_map(struct ceph_mon_client *monc, int sub, u32 epoch,
436 			bool continuous)
437 {
438 	bool need_request;
439 
440 	mutex_lock(&monc->mutex);
441 	need_request = __ceph_monc_want_map(monc, sub, epoch, continuous);
442 	mutex_unlock(&monc->mutex);
443 
444 	return need_request;
445 }
446 EXPORT_SYMBOL(ceph_monc_want_map);
447 
448 /*
449  * Keep track of which maps we have
450  *
451  * @sub: one of CEPH_SUB_*
452  */
453 static void __ceph_monc_got_map(struct ceph_mon_client *monc, int sub,
454 				u32 epoch)
455 {
456 	dout("%s %s epoch %u\n", __func__, ceph_sub_str[sub], epoch);
457 
458 	if (monc->subs[sub].want) {
459 		if (monc->subs[sub].item.flags & CEPH_SUBSCRIBE_ONETIME)
460 			monc->subs[sub].want = false;
461 		else
462 			monc->subs[sub].item.start = cpu_to_le64(epoch + 1);
463 	}
464 
465 	monc->subs[sub].have = epoch;
466 }
467 
468 void ceph_monc_got_map(struct ceph_mon_client *monc, int sub, u32 epoch)
469 {
470 	mutex_lock(&monc->mutex);
471 	__ceph_monc_got_map(monc, sub, epoch);
472 	mutex_unlock(&monc->mutex);
473 }
474 EXPORT_SYMBOL(ceph_monc_got_map);
475 
476 void ceph_monc_renew_subs(struct ceph_mon_client *monc)
477 {
478 	mutex_lock(&monc->mutex);
479 	__send_subscribe(monc);
480 	mutex_unlock(&monc->mutex);
481 }
482 EXPORT_SYMBOL(ceph_monc_renew_subs);
483 
484 /*
485  * Wait for an osdmap with a given epoch.
486  *
487  * @epoch: epoch to wait for
488  * @timeout: in jiffies, 0 means "wait forever"
489  */
490 int ceph_monc_wait_osdmap(struct ceph_mon_client *monc, u32 epoch,
491 			  unsigned long timeout)
492 {
493 	unsigned long started = jiffies;
494 	long ret;
495 
496 	mutex_lock(&monc->mutex);
497 	while (monc->subs[CEPH_SUB_OSDMAP].have < epoch) {
498 		mutex_unlock(&monc->mutex);
499 
500 		if (timeout && time_after_eq(jiffies, started + timeout))
501 			return -ETIMEDOUT;
502 
503 		ret = wait_event_interruptible_timeout(monc->client->auth_wq,
504 				     monc->subs[CEPH_SUB_OSDMAP].have >= epoch,
505 				     ceph_timeout_jiffies(timeout));
506 		if (ret < 0)
507 			return ret;
508 
509 		mutex_lock(&monc->mutex);
510 	}
511 
512 	mutex_unlock(&monc->mutex);
513 	return 0;
514 }
515 EXPORT_SYMBOL(ceph_monc_wait_osdmap);
516 
517 /*
518  * Open a session with a random monitor.  Request monmap and osdmap,
519  * which are waited upon in __ceph_open_session().
520  */
521 int ceph_monc_open_session(struct ceph_mon_client *monc)
522 {
523 	mutex_lock(&monc->mutex);
524 	__ceph_monc_want_map(monc, CEPH_SUB_MONMAP, 0, true);
525 	__ceph_monc_want_map(monc, CEPH_SUB_OSDMAP, 0, false);
526 	__open_session(monc);
527 	__schedule_delayed(monc);
528 	mutex_unlock(&monc->mutex);
529 	return 0;
530 }
531 EXPORT_SYMBOL(ceph_monc_open_session);
532 
533 static void ceph_monc_handle_map(struct ceph_mon_client *monc,
534 				 struct ceph_msg *msg)
535 {
536 	struct ceph_client *client = monc->client;
537 	struct ceph_monmap *monmap;
538 	void *p, *end;
539 
540 	mutex_lock(&monc->mutex);
541 
542 	dout("handle_monmap\n");
543 	p = msg->front.iov_base;
544 	end = p + msg->front.iov_len;
545 
546 	monmap = ceph_monmap_decode(&p, end, false);
547 	if (IS_ERR(monmap)) {
548 		pr_err("problem decoding monmap, %d\n",
549 		       (int)PTR_ERR(monmap));
550 		ceph_msg_dump(msg);
551 		goto out;
552 	}
553 
554 	if (ceph_check_fsid(client, &monmap->fsid) < 0) {
555 		kfree(monmap);
556 		goto out;
557 	}
558 
559 	kfree(monc->monmap);
560 	monc->monmap = monmap;
561 
562 	__ceph_monc_got_map(monc, CEPH_SUB_MONMAP, monc->monmap->epoch);
563 	client->have_fsid = true;
564 
565 out:
566 	mutex_unlock(&monc->mutex);
567 	wake_up_all(&client->auth_wq);
568 }
569 
570 /*
571  * generic requests (currently statfs, mon_get_version)
572  */
573 DEFINE_RB_FUNCS(generic_request, struct ceph_mon_generic_request, tid, node)
574 
575 static void release_generic_request(struct kref *kref)
576 {
577 	struct ceph_mon_generic_request *req =
578 		container_of(kref, struct ceph_mon_generic_request, kref);
579 
580 	dout("%s greq %p request %p reply %p\n", __func__, req, req->request,
581 	     req->reply);
582 	WARN_ON(!RB_EMPTY_NODE(&req->node));
583 
584 	if (req->reply)
585 		ceph_msg_put(req->reply);
586 	if (req->request)
587 		ceph_msg_put(req->request);
588 
589 	kfree(req);
590 }
591 
592 static void put_generic_request(struct ceph_mon_generic_request *req)
593 {
594 	if (req)
595 		kref_put(&req->kref, release_generic_request);
596 }
597 
598 static void get_generic_request(struct ceph_mon_generic_request *req)
599 {
600 	kref_get(&req->kref);
601 }
602 
603 static struct ceph_mon_generic_request *
604 alloc_generic_request(struct ceph_mon_client *monc, gfp_t gfp)
605 {
606 	struct ceph_mon_generic_request *req;
607 
608 	req = kzalloc(sizeof(*req), gfp);
609 	if (!req)
610 		return NULL;
611 
612 	req->monc = monc;
613 	kref_init(&req->kref);
614 	RB_CLEAR_NODE(&req->node);
615 	init_completion(&req->completion);
616 
617 	dout("%s greq %p\n", __func__, req);
618 	return req;
619 }
620 
621 static void register_generic_request(struct ceph_mon_generic_request *req)
622 {
623 	struct ceph_mon_client *monc = req->monc;
624 
625 	WARN_ON(req->tid);
626 
627 	get_generic_request(req);
628 	req->tid = ++monc->last_tid;
629 	insert_generic_request(&monc->generic_request_tree, req);
630 }
631 
632 static void send_generic_request(struct ceph_mon_client *monc,
633 				 struct ceph_mon_generic_request *req)
634 {
635 	WARN_ON(!req->tid);
636 
637 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
638 	req->request->hdr.tid = cpu_to_le64(req->tid);
639 	ceph_con_send(&monc->con, ceph_msg_get(req->request));
640 }
641 
642 static void __finish_generic_request(struct ceph_mon_generic_request *req)
643 {
644 	struct ceph_mon_client *monc = req->monc;
645 
646 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
647 	erase_generic_request(&monc->generic_request_tree, req);
648 
649 	ceph_msg_revoke(req->request);
650 	ceph_msg_revoke_incoming(req->reply);
651 }
652 
653 static void finish_generic_request(struct ceph_mon_generic_request *req)
654 {
655 	__finish_generic_request(req);
656 	put_generic_request(req);
657 }
658 
659 static void complete_generic_request(struct ceph_mon_generic_request *req)
660 {
661 	if (req->complete_cb)
662 		req->complete_cb(req);
663 	else
664 		complete_all(&req->completion);
665 	put_generic_request(req);
666 }
667 
668 static void cancel_generic_request(struct ceph_mon_generic_request *req)
669 {
670 	struct ceph_mon_client *monc = req->monc;
671 	struct ceph_mon_generic_request *lookup_req;
672 
673 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
674 
675 	mutex_lock(&monc->mutex);
676 	lookup_req = lookup_generic_request(&monc->generic_request_tree,
677 					    req->tid);
678 	if (lookup_req) {
679 		WARN_ON(lookup_req != req);
680 		finish_generic_request(req);
681 	}
682 
683 	mutex_unlock(&monc->mutex);
684 }
685 
686 static int wait_generic_request(struct ceph_mon_generic_request *req)
687 {
688 	int ret;
689 
690 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
691 	ret = wait_for_completion_interruptible(&req->completion);
692 	if (ret)
693 		cancel_generic_request(req);
694 	else
695 		ret = req->result; /* completed */
696 
697 	return ret;
698 }
699 
700 static struct ceph_msg *get_generic_reply(struct ceph_connection *con,
701 					 struct ceph_msg_header *hdr,
702 					 int *skip)
703 {
704 	struct ceph_mon_client *monc = con->private;
705 	struct ceph_mon_generic_request *req;
706 	u64 tid = le64_to_cpu(hdr->tid);
707 	struct ceph_msg *m;
708 
709 	mutex_lock(&monc->mutex);
710 	req = lookup_generic_request(&monc->generic_request_tree, tid);
711 	if (!req) {
712 		dout("get_generic_reply %lld dne\n", tid);
713 		*skip = 1;
714 		m = NULL;
715 	} else {
716 		dout("get_generic_reply %lld got %p\n", tid, req->reply);
717 		*skip = 0;
718 		m = ceph_msg_get(req->reply);
719 		/*
720 		 * we don't need to track the connection reading into
721 		 * this reply because we only have one open connection
722 		 * at a time, ever.
723 		 */
724 	}
725 	mutex_unlock(&monc->mutex);
726 	return m;
727 }
728 
729 /*
730  * statfs
731  */
732 static void handle_statfs_reply(struct ceph_mon_client *monc,
733 				struct ceph_msg *msg)
734 {
735 	struct ceph_mon_generic_request *req;
736 	struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
737 	u64 tid = le64_to_cpu(msg->hdr.tid);
738 
739 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
740 
741 	if (msg->front.iov_len != sizeof(*reply))
742 		goto bad;
743 
744 	mutex_lock(&monc->mutex);
745 	req = lookup_generic_request(&monc->generic_request_tree, tid);
746 	if (!req) {
747 		mutex_unlock(&monc->mutex);
748 		return;
749 	}
750 
751 	req->result = 0;
752 	*req->u.st = reply->st; /* struct */
753 	__finish_generic_request(req);
754 	mutex_unlock(&monc->mutex);
755 
756 	complete_generic_request(req);
757 	return;
758 
759 bad:
760 	pr_err("corrupt statfs reply, tid %llu\n", tid);
761 	ceph_msg_dump(msg);
762 }
763 
764 /*
765  * Do a synchronous statfs().
766  */
767 int ceph_monc_do_statfs(struct ceph_mon_client *monc, u64 data_pool,
768 			struct ceph_statfs *buf)
769 {
770 	struct ceph_mon_generic_request *req;
771 	struct ceph_mon_statfs *h;
772 	int ret = -ENOMEM;
773 
774 	req = alloc_generic_request(monc, GFP_NOFS);
775 	if (!req)
776 		goto out;
777 
778 	req->request = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), GFP_NOFS,
779 				    true);
780 	if (!req->request)
781 		goto out;
782 
783 	req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 64, GFP_NOFS, true);
784 	if (!req->reply)
785 		goto out;
786 
787 	req->u.st = buf;
788 	req->request->hdr.version = cpu_to_le16(2);
789 
790 	mutex_lock(&monc->mutex);
791 	register_generic_request(req);
792 	/* fill out request */
793 	h = req->request->front.iov_base;
794 	h->monhdr.have_version = 0;
795 	h->monhdr.session_mon = cpu_to_le16(-1);
796 	h->monhdr.session_mon_tid = 0;
797 	h->fsid = monc->monmap->fsid;
798 	h->contains_data_pool = (data_pool != CEPH_NOPOOL);
799 	h->data_pool = cpu_to_le64(data_pool);
800 	send_generic_request(monc, req);
801 	mutex_unlock(&monc->mutex);
802 
803 	ret = wait_generic_request(req);
804 out:
805 	put_generic_request(req);
806 	return ret;
807 }
808 EXPORT_SYMBOL(ceph_monc_do_statfs);
809 
810 static void handle_get_version_reply(struct ceph_mon_client *monc,
811 				     struct ceph_msg *msg)
812 {
813 	struct ceph_mon_generic_request *req;
814 	u64 tid = le64_to_cpu(msg->hdr.tid);
815 	void *p = msg->front.iov_base;
816 	void *end = p + msg->front_alloc_len;
817 	u64 handle;
818 
819 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
820 
821 	ceph_decode_need(&p, end, 2*sizeof(u64), bad);
822 	handle = ceph_decode_64(&p);
823 	if (tid != 0 && tid != handle)
824 		goto bad;
825 
826 	mutex_lock(&monc->mutex);
827 	req = lookup_generic_request(&monc->generic_request_tree, handle);
828 	if (!req) {
829 		mutex_unlock(&monc->mutex);
830 		return;
831 	}
832 
833 	req->result = 0;
834 	req->u.newest = ceph_decode_64(&p);
835 	__finish_generic_request(req);
836 	mutex_unlock(&monc->mutex);
837 
838 	complete_generic_request(req);
839 	return;
840 
841 bad:
842 	pr_err("corrupt mon_get_version reply, tid %llu\n", tid);
843 	ceph_msg_dump(msg);
844 }
845 
846 static struct ceph_mon_generic_request *
847 __ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
848 			ceph_monc_callback_t cb, u64 private_data)
849 {
850 	struct ceph_mon_generic_request *req;
851 
852 	req = alloc_generic_request(monc, GFP_NOIO);
853 	if (!req)
854 		goto err_put_req;
855 
856 	req->request = ceph_msg_new(CEPH_MSG_MON_GET_VERSION,
857 				    sizeof(u64) + sizeof(u32) + strlen(what),
858 				    GFP_NOIO, true);
859 	if (!req->request)
860 		goto err_put_req;
861 
862 	req->reply = ceph_msg_new(CEPH_MSG_MON_GET_VERSION_REPLY, 32, GFP_NOIO,
863 				  true);
864 	if (!req->reply)
865 		goto err_put_req;
866 
867 	req->complete_cb = cb;
868 	req->private_data = private_data;
869 
870 	mutex_lock(&monc->mutex);
871 	register_generic_request(req);
872 	{
873 		void *p = req->request->front.iov_base;
874 		void *const end = p + req->request->front_alloc_len;
875 
876 		ceph_encode_64(&p, req->tid); /* handle */
877 		ceph_encode_string(&p, end, what, strlen(what));
878 		WARN_ON(p != end);
879 	}
880 	send_generic_request(monc, req);
881 	mutex_unlock(&monc->mutex);
882 
883 	return req;
884 
885 err_put_req:
886 	put_generic_request(req);
887 	return ERR_PTR(-ENOMEM);
888 }
889 
890 /*
891  * Send MMonGetVersion and wait for the reply.
892  *
893  * @what: one of "mdsmap", "osdmap" or "monmap"
894  */
895 int ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
896 			  u64 *newest)
897 {
898 	struct ceph_mon_generic_request *req;
899 	int ret;
900 
901 	req = __ceph_monc_get_version(monc, what, NULL, 0);
902 	if (IS_ERR(req))
903 		return PTR_ERR(req);
904 
905 	ret = wait_generic_request(req);
906 	if (!ret)
907 		*newest = req->u.newest;
908 
909 	put_generic_request(req);
910 	return ret;
911 }
912 EXPORT_SYMBOL(ceph_monc_get_version);
913 
914 /*
915  * Send MMonGetVersion,
916  *
917  * @what: one of "mdsmap", "osdmap" or "monmap"
918  */
919 int ceph_monc_get_version_async(struct ceph_mon_client *monc, const char *what,
920 				ceph_monc_callback_t cb, u64 private_data)
921 {
922 	struct ceph_mon_generic_request *req;
923 
924 	req = __ceph_monc_get_version(monc, what, cb, private_data);
925 	if (IS_ERR(req))
926 		return PTR_ERR(req);
927 
928 	put_generic_request(req);
929 	return 0;
930 }
931 EXPORT_SYMBOL(ceph_monc_get_version_async);
932 
933 static void handle_command_ack(struct ceph_mon_client *monc,
934 			       struct ceph_msg *msg)
935 {
936 	struct ceph_mon_generic_request *req;
937 	void *p = msg->front.iov_base;
938 	void *const end = p + msg->front_alloc_len;
939 	u64 tid = le64_to_cpu(msg->hdr.tid);
940 
941 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
942 
943 	ceph_decode_need(&p, end, sizeof(struct ceph_mon_request_header) +
944 							    sizeof(u32), bad);
945 	p += sizeof(struct ceph_mon_request_header);
946 
947 	mutex_lock(&monc->mutex);
948 	req = lookup_generic_request(&monc->generic_request_tree, tid);
949 	if (!req) {
950 		mutex_unlock(&monc->mutex);
951 		return;
952 	}
953 
954 	req->result = ceph_decode_32(&p);
955 	__finish_generic_request(req);
956 	mutex_unlock(&monc->mutex);
957 
958 	complete_generic_request(req);
959 	return;
960 
961 bad:
962 	pr_err("corrupt mon_command ack, tid %llu\n", tid);
963 	ceph_msg_dump(msg);
964 }
965 
966 static __printf(2, 0)
967 int do_mon_command_vargs(struct ceph_mon_client *monc, const char *fmt,
968 			 va_list ap)
969 {
970 	struct ceph_mon_generic_request *req;
971 	struct ceph_mon_command *h;
972 	int ret = -ENOMEM;
973 	int len;
974 
975 	req = alloc_generic_request(monc, GFP_NOIO);
976 	if (!req)
977 		goto out;
978 
979 	req->request = ceph_msg_new(CEPH_MSG_MON_COMMAND, 256, GFP_NOIO, true);
980 	if (!req->request)
981 		goto out;
982 
983 	req->reply = ceph_msg_new(CEPH_MSG_MON_COMMAND_ACK, 512, GFP_NOIO,
984 				  true);
985 	if (!req->reply)
986 		goto out;
987 
988 	mutex_lock(&monc->mutex);
989 	register_generic_request(req);
990 	h = req->request->front.iov_base;
991 	h->monhdr.have_version = 0;
992 	h->monhdr.session_mon = cpu_to_le16(-1);
993 	h->monhdr.session_mon_tid = 0;
994 	h->fsid = monc->monmap->fsid;
995 	h->num_strs = cpu_to_le32(1);
996 	len = vsprintf(h->str, fmt, ap);
997 	h->str_len = cpu_to_le32(len);
998 	send_generic_request(monc, req);
999 	mutex_unlock(&monc->mutex);
1000 
1001 	ret = wait_generic_request(req);
1002 out:
1003 	put_generic_request(req);
1004 	return ret;
1005 }
1006 
1007 static __printf(2, 3)
1008 int do_mon_command(struct ceph_mon_client *monc, const char *fmt, ...)
1009 {
1010 	va_list ap;
1011 	int ret;
1012 
1013 	va_start(ap, fmt);
1014 	ret = do_mon_command_vargs(monc, fmt, ap);
1015 	va_end(ap);
1016 	return ret;
1017 }
1018 
1019 int ceph_monc_blocklist_add(struct ceph_mon_client *monc,
1020 			    struct ceph_entity_addr *client_addr)
1021 {
1022 	int ret;
1023 
1024 	ret = do_mon_command(monc,
1025 			     "{ \"prefix\": \"osd blocklist\", \
1026 				\"blocklistop\": \"add\", \
1027 				\"addr\": \"%pISpc/%u\" }",
1028 			     &client_addr->in_addr,
1029 			     le32_to_cpu(client_addr->nonce));
1030 	if (ret == -EINVAL) {
1031 		/*
1032 		 * The monitor returns EINVAL on an unrecognized command.
1033 		 * Try the legacy command -- it is exactly the same except
1034 		 * for the name.
1035 		 */
1036 		ret = do_mon_command(monc,
1037 				     "{ \"prefix\": \"osd blacklist\", \
1038 					\"blacklistop\": \"add\", \
1039 					\"addr\": \"%pISpc/%u\" }",
1040 				     &client_addr->in_addr,
1041 				     le32_to_cpu(client_addr->nonce));
1042 	}
1043 	if (ret)
1044 		return ret;
1045 
1046 	/*
1047 	 * Make sure we have the osdmap that includes the blocklist
1048 	 * entry.  This is needed to ensure that the OSDs pick up the
1049 	 * new blocklist before processing any future requests from
1050 	 * this client.
1051 	 */
1052 	return ceph_wait_for_latest_osdmap(monc->client, 0);
1053 }
1054 EXPORT_SYMBOL(ceph_monc_blocklist_add);
1055 
1056 /*
1057  * Resend pending generic requests.
1058  */
1059 static void __resend_generic_request(struct ceph_mon_client *monc)
1060 {
1061 	struct ceph_mon_generic_request *req;
1062 	struct rb_node *p;
1063 
1064 	for (p = rb_first(&monc->generic_request_tree); p; p = rb_next(p)) {
1065 		req = rb_entry(p, struct ceph_mon_generic_request, node);
1066 		ceph_msg_revoke(req->request);
1067 		ceph_msg_revoke_incoming(req->reply);
1068 		ceph_con_send(&monc->con, ceph_msg_get(req->request));
1069 	}
1070 }
1071 
1072 /*
1073  * Delayed work.  If we haven't mounted yet, retry.  Otherwise,
1074  * renew/retry subscription as needed (in case it is timing out, or we
1075  * got an ENOMEM).  And keep the monitor connection alive.
1076  */
1077 static void delayed_work(struct work_struct *work)
1078 {
1079 	struct ceph_mon_client *monc =
1080 		container_of(work, struct ceph_mon_client, delayed_work.work);
1081 
1082 	dout("monc delayed_work\n");
1083 	mutex_lock(&monc->mutex);
1084 	if (monc->hunting) {
1085 		dout("%s continuing hunt\n", __func__);
1086 		reopen_session(monc);
1087 	} else {
1088 		int is_auth = ceph_auth_is_authenticated(monc->auth);
1089 		if (ceph_con_keepalive_expired(&monc->con,
1090 					       CEPH_MONC_PING_TIMEOUT)) {
1091 			dout("monc keepalive timeout\n");
1092 			is_auth = 0;
1093 			reopen_session(monc);
1094 		}
1095 
1096 		if (!monc->hunting) {
1097 			ceph_con_keepalive(&monc->con);
1098 			__validate_auth(monc);
1099 			un_backoff(monc);
1100 		}
1101 
1102 		if (is_auth &&
1103 		    !(monc->con.peer_features & CEPH_FEATURE_MON_STATEFUL_SUB)) {
1104 			unsigned long now = jiffies;
1105 
1106 			dout("%s renew subs? now %lu renew after %lu\n",
1107 			     __func__, now, monc->sub_renew_after);
1108 			if (time_after_eq(now, monc->sub_renew_after))
1109 				__send_subscribe(monc);
1110 		}
1111 	}
1112 	__schedule_delayed(monc);
1113 	mutex_unlock(&monc->mutex);
1114 }
1115 
1116 /*
1117  * On startup, we build a temporary monmap populated with the IPs
1118  * provided by mount(2).
1119  */
1120 static int build_initial_monmap(struct ceph_mon_client *monc)
1121 {
1122 	struct ceph_options *opt = monc->client->options;
1123 	struct ceph_entity_addr *mon_addr = opt->mon_addr;
1124 	int num_mon = opt->num_mon;
1125 	int i;
1126 
1127 	/* build initial monmap */
1128 	monc->monmap = kzalloc(struct_size(monc->monmap, mon_inst, num_mon),
1129 			       GFP_KERNEL);
1130 	if (!monc->monmap)
1131 		return -ENOMEM;
1132 	for (i = 0; i < num_mon; i++) {
1133 		monc->monmap->mon_inst[i].addr = mon_addr[i];
1134 		monc->monmap->mon_inst[i].addr.nonce = 0;
1135 		monc->monmap->mon_inst[i].name.type =
1136 			CEPH_ENTITY_TYPE_MON;
1137 		monc->monmap->mon_inst[i].name.num = cpu_to_le64(i);
1138 	}
1139 	monc->monmap->num_mon = num_mon;
1140 	return 0;
1141 }
1142 
1143 int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl)
1144 {
1145 	int err = 0;
1146 
1147 	dout("init\n");
1148 	memset(monc, 0, sizeof(*monc));
1149 	monc->client = cl;
1150 	monc->monmap = NULL;
1151 	mutex_init(&monc->mutex);
1152 
1153 	err = build_initial_monmap(monc);
1154 	if (err)
1155 		goto out;
1156 
1157 	/* connection */
1158 	/* authentication */
1159 	monc->auth = ceph_auth_init(cl->options->name, cl->options->key,
1160 				    cl->options->con_modes);
1161 	if (IS_ERR(monc->auth)) {
1162 		err = PTR_ERR(monc->auth);
1163 		goto out_monmap;
1164 	}
1165 	monc->auth->want_keys =
1166 		CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON |
1167 		CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS;
1168 
1169 	/* msgs */
1170 	err = -ENOMEM;
1171 	monc->m_subscribe_ack = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE_ACK,
1172 				     sizeof(struct ceph_mon_subscribe_ack),
1173 				     GFP_KERNEL, true);
1174 	if (!monc->m_subscribe_ack)
1175 		goto out_auth;
1176 
1177 	monc->m_subscribe = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 128,
1178 					 GFP_KERNEL, true);
1179 	if (!monc->m_subscribe)
1180 		goto out_subscribe_ack;
1181 
1182 	monc->m_auth_reply = ceph_msg_new(CEPH_MSG_AUTH_REPLY, 4096,
1183 					  GFP_KERNEL, true);
1184 	if (!monc->m_auth_reply)
1185 		goto out_subscribe;
1186 
1187 	monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, GFP_KERNEL, true);
1188 	monc->pending_auth = 0;
1189 	if (!monc->m_auth)
1190 		goto out_auth_reply;
1191 
1192 	ceph_con_init(&monc->con, monc, &mon_con_ops,
1193 		      &monc->client->msgr);
1194 
1195 	monc->cur_mon = -1;
1196 	monc->had_a_connection = false;
1197 	monc->hunt_mult = 1;
1198 
1199 	INIT_DELAYED_WORK(&monc->delayed_work, delayed_work);
1200 	monc->generic_request_tree = RB_ROOT;
1201 	monc->last_tid = 0;
1202 
1203 	monc->fs_cluster_id = CEPH_FS_CLUSTER_ID_NONE;
1204 
1205 	return 0;
1206 
1207 out_auth_reply:
1208 	ceph_msg_put(monc->m_auth_reply);
1209 out_subscribe:
1210 	ceph_msg_put(monc->m_subscribe);
1211 out_subscribe_ack:
1212 	ceph_msg_put(monc->m_subscribe_ack);
1213 out_auth:
1214 	ceph_auth_destroy(monc->auth);
1215 out_monmap:
1216 	kfree(monc->monmap);
1217 out:
1218 	return err;
1219 }
1220 EXPORT_SYMBOL(ceph_monc_init);
1221 
1222 void ceph_monc_stop(struct ceph_mon_client *monc)
1223 {
1224 	dout("stop\n");
1225 	cancel_delayed_work_sync(&monc->delayed_work);
1226 
1227 	mutex_lock(&monc->mutex);
1228 	__close_session(monc);
1229 	monc->cur_mon = -1;
1230 	mutex_unlock(&monc->mutex);
1231 
1232 	/*
1233 	 * flush msgr queue before we destroy ourselves to ensure that:
1234 	 *  - any work that references our embedded con is finished.
1235 	 *  - any osd_client or other work that may reference an authorizer
1236 	 *    finishes before we shut down the auth subsystem.
1237 	 */
1238 	ceph_msgr_flush();
1239 
1240 	ceph_auth_destroy(monc->auth);
1241 
1242 	WARN_ON(!RB_EMPTY_ROOT(&monc->generic_request_tree));
1243 
1244 	ceph_msg_put(monc->m_auth);
1245 	ceph_msg_put(monc->m_auth_reply);
1246 	ceph_msg_put(monc->m_subscribe);
1247 	ceph_msg_put(monc->m_subscribe_ack);
1248 
1249 	kfree(monc->monmap);
1250 }
1251 EXPORT_SYMBOL(ceph_monc_stop);
1252 
1253 static void finish_hunting(struct ceph_mon_client *monc)
1254 {
1255 	if (monc->hunting) {
1256 		dout("%s found mon%d\n", __func__, monc->cur_mon);
1257 		monc->hunting = false;
1258 		monc->had_a_connection = true;
1259 		un_backoff(monc);
1260 		__schedule_delayed(monc);
1261 	}
1262 }
1263 
1264 static void finish_auth(struct ceph_mon_client *monc, int auth_err,
1265 			bool was_authed)
1266 {
1267 	dout("%s auth_err %d was_authed %d\n", __func__, auth_err, was_authed);
1268 	WARN_ON(auth_err > 0);
1269 
1270 	monc->pending_auth = 0;
1271 	if (auth_err) {
1272 		monc->client->auth_err = auth_err;
1273 		wake_up_all(&monc->client->auth_wq);
1274 		return;
1275 	}
1276 
1277 	if (!was_authed && ceph_auth_is_authenticated(monc->auth)) {
1278 		dout("%s authenticated, starting session global_id %llu\n",
1279 		     __func__, monc->auth->global_id);
1280 
1281 		monc->client->msgr.inst.name.type = CEPH_ENTITY_TYPE_CLIENT;
1282 		monc->client->msgr.inst.name.num =
1283 					cpu_to_le64(monc->auth->global_id);
1284 
1285 		__send_subscribe(monc);
1286 		__resend_generic_request(monc);
1287 
1288 		pr_info("mon%d %s session established\n", monc->cur_mon,
1289 			ceph_pr_addr(&monc->con.peer_addr));
1290 	}
1291 }
1292 
1293 static void handle_auth_reply(struct ceph_mon_client *monc,
1294 			      struct ceph_msg *msg)
1295 {
1296 	bool was_authed;
1297 	int ret;
1298 
1299 	mutex_lock(&monc->mutex);
1300 	was_authed = ceph_auth_is_authenticated(monc->auth);
1301 	ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base,
1302 				     msg->front.iov_len,
1303 				     monc->m_auth->front.iov_base,
1304 				     monc->m_auth->front_alloc_len);
1305 	if (ret > 0) {
1306 		__send_prepared_auth_request(monc, ret);
1307 	} else {
1308 		finish_auth(monc, ret, was_authed);
1309 		finish_hunting(monc);
1310 	}
1311 	mutex_unlock(&monc->mutex);
1312 }
1313 
1314 static int __validate_auth(struct ceph_mon_client *monc)
1315 {
1316 	int ret;
1317 
1318 	if (monc->pending_auth)
1319 		return 0;
1320 
1321 	ret = ceph_build_auth(monc->auth, monc->m_auth->front.iov_base,
1322 			      monc->m_auth->front_alloc_len);
1323 	if (ret <= 0)
1324 		return ret; /* either an error, or no need to authenticate */
1325 	__send_prepared_auth_request(monc, ret);
1326 	return 0;
1327 }
1328 
1329 int ceph_monc_validate_auth(struct ceph_mon_client *monc)
1330 {
1331 	int ret;
1332 
1333 	mutex_lock(&monc->mutex);
1334 	ret = __validate_auth(monc);
1335 	mutex_unlock(&monc->mutex);
1336 	return ret;
1337 }
1338 EXPORT_SYMBOL(ceph_monc_validate_auth);
1339 
1340 /*
1341  * handle incoming message
1342  */
1343 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
1344 {
1345 	struct ceph_mon_client *monc = con->private;
1346 	int type = le16_to_cpu(msg->hdr.type);
1347 
1348 	switch (type) {
1349 	case CEPH_MSG_AUTH_REPLY:
1350 		handle_auth_reply(monc, msg);
1351 		break;
1352 
1353 	case CEPH_MSG_MON_SUBSCRIBE_ACK:
1354 		handle_subscribe_ack(monc, msg);
1355 		break;
1356 
1357 	case CEPH_MSG_STATFS_REPLY:
1358 		handle_statfs_reply(monc, msg);
1359 		break;
1360 
1361 	case CEPH_MSG_MON_GET_VERSION_REPLY:
1362 		handle_get_version_reply(monc, msg);
1363 		break;
1364 
1365 	case CEPH_MSG_MON_COMMAND_ACK:
1366 		handle_command_ack(monc, msg);
1367 		break;
1368 
1369 	case CEPH_MSG_MON_MAP:
1370 		ceph_monc_handle_map(monc, msg);
1371 		break;
1372 
1373 	case CEPH_MSG_OSD_MAP:
1374 		ceph_osdc_handle_map(&monc->client->osdc, msg);
1375 		break;
1376 
1377 	default:
1378 		/* can the chained handler handle it? */
1379 		if (monc->client->extra_mon_dispatch &&
1380 		    monc->client->extra_mon_dispatch(monc->client, msg) == 0)
1381 			break;
1382 
1383 		pr_err("received unknown message type %d %s\n", type,
1384 		       ceph_msg_type_name(type));
1385 	}
1386 	ceph_msg_put(msg);
1387 }
1388 
1389 /*
1390  * Allocate memory for incoming message
1391  */
1392 static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con,
1393 				      struct ceph_msg_header *hdr,
1394 				      int *skip)
1395 {
1396 	struct ceph_mon_client *monc = con->private;
1397 	int type = le16_to_cpu(hdr->type);
1398 	int front_len = le32_to_cpu(hdr->front_len);
1399 	struct ceph_msg *m = NULL;
1400 
1401 	*skip = 0;
1402 
1403 	switch (type) {
1404 	case CEPH_MSG_MON_SUBSCRIBE_ACK:
1405 		m = ceph_msg_get(monc->m_subscribe_ack);
1406 		break;
1407 	case CEPH_MSG_STATFS_REPLY:
1408 	case CEPH_MSG_MON_COMMAND_ACK:
1409 		return get_generic_reply(con, hdr, skip);
1410 	case CEPH_MSG_AUTH_REPLY:
1411 		m = ceph_msg_get(monc->m_auth_reply);
1412 		break;
1413 	case CEPH_MSG_MON_GET_VERSION_REPLY:
1414 		if (le64_to_cpu(hdr->tid) != 0)
1415 			return get_generic_reply(con, hdr, skip);
1416 
1417 		/*
1418 		 * Older OSDs don't set reply tid even if the orignal
1419 		 * request had a non-zero tid.  Work around this weirdness
1420 		 * by allocating a new message.
1421 		 */
1422 		fallthrough;
1423 	case CEPH_MSG_MON_MAP:
1424 	case CEPH_MSG_MDS_MAP:
1425 	case CEPH_MSG_OSD_MAP:
1426 	case CEPH_MSG_FS_MAP_USER:
1427 		m = ceph_msg_new(type, front_len, GFP_NOFS, false);
1428 		if (!m)
1429 			return NULL;	/* ENOMEM--return skip == 0 */
1430 		break;
1431 	}
1432 
1433 	if (!m) {
1434 		pr_info("alloc_msg unknown type %d\n", type);
1435 		*skip = 1;
1436 	} else if (front_len > m->front_alloc_len) {
1437 		pr_warn("mon_alloc_msg front %d > prealloc %d (%u#%llu)\n",
1438 			front_len, m->front_alloc_len,
1439 			(unsigned int)con->peer_name.type,
1440 			le64_to_cpu(con->peer_name.num));
1441 		ceph_msg_put(m);
1442 		m = ceph_msg_new(type, front_len, GFP_NOFS, false);
1443 	}
1444 
1445 	return m;
1446 }
1447 
1448 /*
1449  * If the monitor connection resets, pick a new monitor and resubmit
1450  * any pending requests.
1451  */
1452 static void mon_fault(struct ceph_connection *con)
1453 {
1454 	struct ceph_mon_client *monc = con->private;
1455 
1456 	mutex_lock(&monc->mutex);
1457 	dout("%s mon%d\n", __func__, monc->cur_mon);
1458 	if (monc->cur_mon >= 0) {
1459 		if (!monc->hunting) {
1460 			dout("%s hunting for new mon\n", __func__);
1461 			reopen_session(monc);
1462 			__schedule_delayed(monc);
1463 		} else {
1464 			dout("%s already hunting\n", __func__);
1465 		}
1466 	}
1467 	mutex_unlock(&monc->mutex);
1468 }
1469 
1470 /*
1471  * We can ignore refcounting on the connection struct, as all references
1472  * will come from the messenger workqueue, which is drained prior to
1473  * mon_client destruction.
1474  */
1475 static struct ceph_connection *con_get(struct ceph_connection *con)
1476 {
1477 	return con;
1478 }
1479 
1480 static void con_put(struct ceph_connection *con)
1481 {
1482 }
1483 
1484 static const struct ceph_connection_operations mon_con_ops = {
1485 	.get = con_get,
1486 	.put = con_put,
1487 	.dispatch = dispatch,
1488 	.fault = mon_fault,
1489 	.alloc_msg = mon_alloc_msg,
1490 };
1491