xref: /openbmc/linux/net/ceph/mon_client.c (revision a1117495)
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 	 * Queue a keepalive to ensure that in case of an early fault
261 	 * the messenger doesn't put us into STANDBY state and instead
262 	 * retries.  This also ensures that our timestamp is valid by
263 	 * the time we finish hunting and delayed_work() checks it.
264 	 */
265 	ceph_con_keepalive(&monc->con);
266 	if (ceph_msgr2(monc->client)) {
267 		monc->pending_auth = 1;
268 		return;
269 	}
270 
271 	/* initiate authentication handshake */
272 	ret = ceph_auth_build_hello(monc->auth,
273 				    monc->m_auth->front.iov_base,
274 				    monc->m_auth->front_alloc_len);
275 	BUG_ON(ret <= 0);
276 	__send_prepared_auth_request(monc, ret);
277 }
278 
279 static void reopen_session(struct ceph_mon_client *monc)
280 {
281 	if (!monc->hunting)
282 		pr_info("mon%d %s session lost, hunting for new mon\n",
283 		    monc->cur_mon, ceph_pr_addr(&monc->con.peer_addr));
284 
285 	__close_session(monc);
286 	__open_session(monc);
287 }
288 
289 void ceph_monc_reopen_session(struct ceph_mon_client *monc)
290 {
291 	mutex_lock(&monc->mutex);
292 	reopen_session(monc);
293 	mutex_unlock(&monc->mutex);
294 }
295 
296 static void un_backoff(struct ceph_mon_client *monc)
297 {
298 	monc->hunt_mult /= 2; /* reduce by 50% */
299 	if (monc->hunt_mult < 1)
300 		monc->hunt_mult = 1;
301 	dout("%s hunt_mult now %d\n", __func__, monc->hunt_mult);
302 }
303 
304 /*
305  * Reschedule delayed work timer.
306  */
307 static void __schedule_delayed(struct ceph_mon_client *monc)
308 {
309 	unsigned long delay;
310 
311 	if (monc->hunting)
312 		delay = CEPH_MONC_HUNT_INTERVAL * monc->hunt_mult;
313 	else
314 		delay = CEPH_MONC_PING_INTERVAL;
315 
316 	dout("__schedule_delayed after %lu\n", delay);
317 	mod_delayed_work(system_wq, &monc->delayed_work,
318 			 round_jiffies_relative(delay));
319 }
320 
321 const char *ceph_sub_str[] = {
322 	[CEPH_SUB_MONMAP] = "monmap",
323 	[CEPH_SUB_OSDMAP] = "osdmap",
324 	[CEPH_SUB_FSMAP]  = "fsmap.user",
325 	[CEPH_SUB_MDSMAP] = "mdsmap",
326 };
327 
328 /*
329  * Send subscribe request for one or more maps, according to
330  * monc->subs.
331  */
332 static void __send_subscribe(struct ceph_mon_client *monc)
333 {
334 	struct ceph_msg *msg = monc->m_subscribe;
335 	void *p = msg->front.iov_base;
336 	void *const end = p + msg->front_alloc_len;
337 	int num = 0;
338 	int i;
339 
340 	dout("%s sent %lu\n", __func__, monc->sub_renew_sent);
341 
342 	BUG_ON(monc->cur_mon < 0);
343 
344 	if (!monc->sub_renew_sent)
345 		monc->sub_renew_sent = jiffies | 1; /* never 0 */
346 
347 	msg->hdr.version = cpu_to_le16(2);
348 
349 	for (i = 0; i < ARRAY_SIZE(monc->subs); i++) {
350 		if (monc->subs[i].want)
351 			num++;
352 	}
353 	BUG_ON(num < 1); /* monmap sub is always there */
354 	ceph_encode_32(&p, num);
355 	for (i = 0; i < ARRAY_SIZE(monc->subs); i++) {
356 		char buf[32];
357 		int len;
358 
359 		if (!monc->subs[i].want)
360 			continue;
361 
362 		len = sprintf(buf, "%s", ceph_sub_str[i]);
363 		if (i == CEPH_SUB_MDSMAP &&
364 		    monc->fs_cluster_id != CEPH_FS_CLUSTER_ID_NONE)
365 			len += sprintf(buf + len, ".%d", monc->fs_cluster_id);
366 
367 		dout("%s %s start %llu flags 0x%x\n", __func__, buf,
368 		     le64_to_cpu(monc->subs[i].item.start),
369 		     monc->subs[i].item.flags);
370 		ceph_encode_string(&p, end, buf, len);
371 		memcpy(p, &monc->subs[i].item, sizeof(monc->subs[i].item));
372 		p += sizeof(monc->subs[i].item);
373 	}
374 
375 	BUG_ON(p > end);
376 	msg->front.iov_len = p - msg->front.iov_base;
377 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
378 	ceph_msg_revoke(msg);
379 	ceph_con_send(&monc->con, ceph_msg_get(msg));
380 }
381 
382 static void handle_subscribe_ack(struct ceph_mon_client *monc,
383 				 struct ceph_msg *msg)
384 {
385 	unsigned int seconds;
386 	struct ceph_mon_subscribe_ack *h = msg->front.iov_base;
387 
388 	if (msg->front.iov_len < sizeof(*h))
389 		goto bad;
390 	seconds = le32_to_cpu(h->duration);
391 
392 	mutex_lock(&monc->mutex);
393 	if (monc->sub_renew_sent) {
394 		/*
395 		 * This is only needed for legacy (infernalis or older)
396 		 * MONs -- see delayed_work().
397 		 */
398 		monc->sub_renew_after = monc->sub_renew_sent +
399 					    (seconds >> 1) * HZ - 1;
400 		dout("%s sent %lu duration %d renew after %lu\n", __func__,
401 		     monc->sub_renew_sent, seconds, monc->sub_renew_after);
402 		monc->sub_renew_sent = 0;
403 	} else {
404 		dout("%s sent %lu renew after %lu, ignoring\n", __func__,
405 		     monc->sub_renew_sent, monc->sub_renew_after);
406 	}
407 	mutex_unlock(&monc->mutex);
408 	return;
409 bad:
410 	pr_err("got corrupt subscribe-ack msg\n");
411 	ceph_msg_dump(msg);
412 }
413 
414 /*
415  * Register interest in a map
416  *
417  * @sub: one of CEPH_SUB_*
418  * @epoch: X for "every map since X", or 0 for "just the latest"
419  */
420 static bool __ceph_monc_want_map(struct ceph_mon_client *monc, int sub,
421 				 u32 epoch, bool continuous)
422 {
423 	__le64 start = cpu_to_le64(epoch);
424 	u8 flags = !continuous ? CEPH_SUBSCRIBE_ONETIME : 0;
425 
426 	dout("%s %s epoch %u continuous %d\n", __func__, ceph_sub_str[sub],
427 	     epoch, continuous);
428 
429 	if (monc->subs[sub].want &&
430 	    monc->subs[sub].item.start == start &&
431 	    monc->subs[sub].item.flags == flags)
432 		return false;
433 
434 	monc->subs[sub].item.start = start;
435 	monc->subs[sub].item.flags = flags;
436 	monc->subs[sub].want = true;
437 
438 	return true;
439 }
440 
441 bool ceph_monc_want_map(struct ceph_mon_client *monc, int sub, u32 epoch,
442 			bool continuous)
443 {
444 	bool need_request;
445 
446 	mutex_lock(&monc->mutex);
447 	need_request = __ceph_monc_want_map(monc, sub, epoch, continuous);
448 	mutex_unlock(&monc->mutex);
449 
450 	return need_request;
451 }
452 EXPORT_SYMBOL(ceph_monc_want_map);
453 
454 /*
455  * Keep track of which maps we have
456  *
457  * @sub: one of CEPH_SUB_*
458  */
459 static void __ceph_monc_got_map(struct ceph_mon_client *monc, int sub,
460 				u32 epoch)
461 {
462 	dout("%s %s epoch %u\n", __func__, ceph_sub_str[sub], epoch);
463 
464 	if (monc->subs[sub].want) {
465 		if (monc->subs[sub].item.flags & CEPH_SUBSCRIBE_ONETIME)
466 			monc->subs[sub].want = false;
467 		else
468 			monc->subs[sub].item.start = cpu_to_le64(epoch + 1);
469 	}
470 
471 	monc->subs[sub].have = epoch;
472 }
473 
474 void ceph_monc_got_map(struct ceph_mon_client *monc, int sub, u32 epoch)
475 {
476 	mutex_lock(&monc->mutex);
477 	__ceph_monc_got_map(monc, sub, epoch);
478 	mutex_unlock(&monc->mutex);
479 }
480 EXPORT_SYMBOL(ceph_monc_got_map);
481 
482 void ceph_monc_renew_subs(struct ceph_mon_client *monc)
483 {
484 	mutex_lock(&monc->mutex);
485 	__send_subscribe(monc);
486 	mutex_unlock(&monc->mutex);
487 }
488 EXPORT_SYMBOL(ceph_monc_renew_subs);
489 
490 /*
491  * Wait for an osdmap with a given epoch.
492  *
493  * @epoch: epoch to wait for
494  * @timeout: in jiffies, 0 means "wait forever"
495  */
496 int ceph_monc_wait_osdmap(struct ceph_mon_client *monc, u32 epoch,
497 			  unsigned long timeout)
498 {
499 	unsigned long started = jiffies;
500 	long ret;
501 
502 	mutex_lock(&monc->mutex);
503 	while (monc->subs[CEPH_SUB_OSDMAP].have < epoch) {
504 		mutex_unlock(&monc->mutex);
505 
506 		if (timeout && time_after_eq(jiffies, started + timeout))
507 			return -ETIMEDOUT;
508 
509 		ret = wait_event_interruptible_timeout(monc->client->auth_wq,
510 				     monc->subs[CEPH_SUB_OSDMAP].have >= epoch,
511 				     ceph_timeout_jiffies(timeout));
512 		if (ret < 0)
513 			return ret;
514 
515 		mutex_lock(&monc->mutex);
516 	}
517 
518 	mutex_unlock(&monc->mutex);
519 	return 0;
520 }
521 EXPORT_SYMBOL(ceph_monc_wait_osdmap);
522 
523 /*
524  * Open a session with a random monitor.  Request monmap and osdmap,
525  * which are waited upon in __ceph_open_session().
526  */
527 int ceph_monc_open_session(struct ceph_mon_client *monc)
528 {
529 	mutex_lock(&monc->mutex);
530 	__ceph_monc_want_map(monc, CEPH_SUB_MONMAP, 0, true);
531 	__ceph_monc_want_map(monc, CEPH_SUB_OSDMAP, 0, false);
532 	__open_session(monc);
533 	__schedule_delayed(monc);
534 	mutex_unlock(&monc->mutex);
535 	return 0;
536 }
537 EXPORT_SYMBOL(ceph_monc_open_session);
538 
539 static void ceph_monc_handle_map(struct ceph_mon_client *monc,
540 				 struct ceph_msg *msg)
541 {
542 	struct ceph_client *client = monc->client;
543 	struct ceph_monmap *monmap;
544 	void *p, *end;
545 
546 	mutex_lock(&monc->mutex);
547 
548 	dout("handle_monmap\n");
549 	p = msg->front.iov_base;
550 	end = p + msg->front.iov_len;
551 
552 	monmap = ceph_monmap_decode(&p, end, ceph_msgr2(client));
553 	if (IS_ERR(monmap)) {
554 		pr_err("problem decoding monmap, %d\n",
555 		       (int)PTR_ERR(monmap));
556 		ceph_msg_dump(msg);
557 		goto out;
558 	}
559 
560 	if (ceph_check_fsid(client, &monmap->fsid) < 0) {
561 		kfree(monmap);
562 		goto out;
563 	}
564 
565 	kfree(monc->monmap);
566 	monc->monmap = monmap;
567 
568 	__ceph_monc_got_map(monc, CEPH_SUB_MONMAP, monc->monmap->epoch);
569 	client->have_fsid = true;
570 
571 out:
572 	mutex_unlock(&monc->mutex);
573 	wake_up_all(&client->auth_wq);
574 }
575 
576 /*
577  * generic requests (currently statfs, mon_get_version)
578  */
579 DEFINE_RB_FUNCS(generic_request, struct ceph_mon_generic_request, tid, node)
580 
581 static void release_generic_request(struct kref *kref)
582 {
583 	struct ceph_mon_generic_request *req =
584 		container_of(kref, struct ceph_mon_generic_request, kref);
585 
586 	dout("%s greq %p request %p reply %p\n", __func__, req, req->request,
587 	     req->reply);
588 	WARN_ON(!RB_EMPTY_NODE(&req->node));
589 
590 	if (req->reply)
591 		ceph_msg_put(req->reply);
592 	if (req->request)
593 		ceph_msg_put(req->request);
594 
595 	kfree(req);
596 }
597 
598 static void put_generic_request(struct ceph_mon_generic_request *req)
599 {
600 	if (req)
601 		kref_put(&req->kref, release_generic_request);
602 }
603 
604 static void get_generic_request(struct ceph_mon_generic_request *req)
605 {
606 	kref_get(&req->kref);
607 }
608 
609 static struct ceph_mon_generic_request *
610 alloc_generic_request(struct ceph_mon_client *monc, gfp_t gfp)
611 {
612 	struct ceph_mon_generic_request *req;
613 
614 	req = kzalloc(sizeof(*req), gfp);
615 	if (!req)
616 		return NULL;
617 
618 	req->monc = monc;
619 	kref_init(&req->kref);
620 	RB_CLEAR_NODE(&req->node);
621 	init_completion(&req->completion);
622 
623 	dout("%s greq %p\n", __func__, req);
624 	return req;
625 }
626 
627 static void register_generic_request(struct ceph_mon_generic_request *req)
628 {
629 	struct ceph_mon_client *monc = req->monc;
630 
631 	WARN_ON(req->tid);
632 
633 	get_generic_request(req);
634 	req->tid = ++monc->last_tid;
635 	insert_generic_request(&monc->generic_request_tree, req);
636 }
637 
638 static void send_generic_request(struct ceph_mon_client *monc,
639 				 struct ceph_mon_generic_request *req)
640 {
641 	WARN_ON(!req->tid);
642 
643 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
644 	req->request->hdr.tid = cpu_to_le64(req->tid);
645 	ceph_con_send(&monc->con, ceph_msg_get(req->request));
646 }
647 
648 static void __finish_generic_request(struct ceph_mon_generic_request *req)
649 {
650 	struct ceph_mon_client *monc = req->monc;
651 
652 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
653 	erase_generic_request(&monc->generic_request_tree, req);
654 
655 	ceph_msg_revoke(req->request);
656 	ceph_msg_revoke_incoming(req->reply);
657 }
658 
659 static void finish_generic_request(struct ceph_mon_generic_request *req)
660 {
661 	__finish_generic_request(req);
662 	put_generic_request(req);
663 }
664 
665 static void complete_generic_request(struct ceph_mon_generic_request *req)
666 {
667 	if (req->complete_cb)
668 		req->complete_cb(req);
669 	else
670 		complete_all(&req->completion);
671 	put_generic_request(req);
672 }
673 
674 static void cancel_generic_request(struct ceph_mon_generic_request *req)
675 {
676 	struct ceph_mon_client *monc = req->monc;
677 	struct ceph_mon_generic_request *lookup_req;
678 
679 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
680 
681 	mutex_lock(&monc->mutex);
682 	lookup_req = lookup_generic_request(&monc->generic_request_tree,
683 					    req->tid);
684 	if (lookup_req) {
685 		WARN_ON(lookup_req != req);
686 		finish_generic_request(req);
687 	}
688 
689 	mutex_unlock(&monc->mutex);
690 }
691 
692 static int wait_generic_request(struct ceph_mon_generic_request *req)
693 {
694 	int ret;
695 
696 	dout("%s greq %p tid %llu\n", __func__, req, req->tid);
697 	ret = wait_for_completion_interruptible(&req->completion);
698 	if (ret)
699 		cancel_generic_request(req);
700 	else
701 		ret = req->result; /* completed */
702 
703 	return ret;
704 }
705 
706 static struct ceph_msg *get_generic_reply(struct ceph_connection *con,
707 					 struct ceph_msg_header *hdr,
708 					 int *skip)
709 {
710 	struct ceph_mon_client *monc = con->private;
711 	struct ceph_mon_generic_request *req;
712 	u64 tid = le64_to_cpu(hdr->tid);
713 	struct ceph_msg *m;
714 
715 	mutex_lock(&monc->mutex);
716 	req = lookup_generic_request(&monc->generic_request_tree, tid);
717 	if (!req) {
718 		dout("get_generic_reply %lld dne\n", tid);
719 		*skip = 1;
720 		m = NULL;
721 	} else {
722 		dout("get_generic_reply %lld got %p\n", tid, req->reply);
723 		*skip = 0;
724 		m = ceph_msg_get(req->reply);
725 		/*
726 		 * we don't need to track the connection reading into
727 		 * this reply because we only have one open connection
728 		 * at a time, ever.
729 		 */
730 	}
731 	mutex_unlock(&monc->mutex);
732 	return m;
733 }
734 
735 /*
736  * statfs
737  */
738 static void handle_statfs_reply(struct ceph_mon_client *monc,
739 				struct ceph_msg *msg)
740 {
741 	struct ceph_mon_generic_request *req;
742 	struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
743 	u64 tid = le64_to_cpu(msg->hdr.tid);
744 
745 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
746 
747 	if (msg->front.iov_len != sizeof(*reply))
748 		goto bad;
749 
750 	mutex_lock(&monc->mutex);
751 	req = lookup_generic_request(&monc->generic_request_tree, tid);
752 	if (!req) {
753 		mutex_unlock(&monc->mutex);
754 		return;
755 	}
756 
757 	req->result = 0;
758 	*req->u.st = reply->st; /* struct */
759 	__finish_generic_request(req);
760 	mutex_unlock(&monc->mutex);
761 
762 	complete_generic_request(req);
763 	return;
764 
765 bad:
766 	pr_err("corrupt statfs reply, tid %llu\n", tid);
767 	ceph_msg_dump(msg);
768 }
769 
770 /*
771  * Do a synchronous statfs().
772  */
773 int ceph_monc_do_statfs(struct ceph_mon_client *monc, u64 data_pool,
774 			struct ceph_statfs *buf)
775 {
776 	struct ceph_mon_generic_request *req;
777 	struct ceph_mon_statfs *h;
778 	int ret = -ENOMEM;
779 
780 	req = alloc_generic_request(monc, GFP_NOFS);
781 	if (!req)
782 		goto out;
783 
784 	req->request = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), GFP_NOFS,
785 				    true);
786 	if (!req->request)
787 		goto out;
788 
789 	req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 64, GFP_NOFS, true);
790 	if (!req->reply)
791 		goto out;
792 
793 	req->u.st = buf;
794 	req->request->hdr.version = cpu_to_le16(2);
795 
796 	mutex_lock(&monc->mutex);
797 	register_generic_request(req);
798 	/* fill out request */
799 	h = req->request->front.iov_base;
800 	h->monhdr.have_version = 0;
801 	h->monhdr.session_mon = cpu_to_le16(-1);
802 	h->monhdr.session_mon_tid = 0;
803 	h->fsid = monc->monmap->fsid;
804 	h->contains_data_pool = (data_pool != CEPH_NOPOOL);
805 	h->data_pool = cpu_to_le64(data_pool);
806 	send_generic_request(monc, req);
807 	mutex_unlock(&monc->mutex);
808 
809 	ret = wait_generic_request(req);
810 out:
811 	put_generic_request(req);
812 	return ret;
813 }
814 EXPORT_SYMBOL(ceph_monc_do_statfs);
815 
816 static void handle_get_version_reply(struct ceph_mon_client *monc,
817 				     struct ceph_msg *msg)
818 {
819 	struct ceph_mon_generic_request *req;
820 	u64 tid = le64_to_cpu(msg->hdr.tid);
821 	void *p = msg->front.iov_base;
822 	void *end = p + msg->front_alloc_len;
823 	u64 handle;
824 
825 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
826 
827 	ceph_decode_need(&p, end, 2*sizeof(u64), bad);
828 	handle = ceph_decode_64(&p);
829 	if (tid != 0 && tid != handle)
830 		goto bad;
831 
832 	mutex_lock(&monc->mutex);
833 	req = lookup_generic_request(&monc->generic_request_tree, handle);
834 	if (!req) {
835 		mutex_unlock(&monc->mutex);
836 		return;
837 	}
838 
839 	req->result = 0;
840 	req->u.newest = ceph_decode_64(&p);
841 	__finish_generic_request(req);
842 	mutex_unlock(&monc->mutex);
843 
844 	complete_generic_request(req);
845 	return;
846 
847 bad:
848 	pr_err("corrupt mon_get_version reply, tid %llu\n", tid);
849 	ceph_msg_dump(msg);
850 }
851 
852 static struct ceph_mon_generic_request *
853 __ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
854 			ceph_monc_callback_t cb, u64 private_data)
855 {
856 	struct ceph_mon_generic_request *req;
857 
858 	req = alloc_generic_request(monc, GFP_NOIO);
859 	if (!req)
860 		goto err_put_req;
861 
862 	req->request = ceph_msg_new(CEPH_MSG_MON_GET_VERSION,
863 				    sizeof(u64) + sizeof(u32) + strlen(what),
864 				    GFP_NOIO, true);
865 	if (!req->request)
866 		goto err_put_req;
867 
868 	req->reply = ceph_msg_new(CEPH_MSG_MON_GET_VERSION_REPLY, 32, GFP_NOIO,
869 				  true);
870 	if (!req->reply)
871 		goto err_put_req;
872 
873 	req->complete_cb = cb;
874 	req->private_data = private_data;
875 
876 	mutex_lock(&monc->mutex);
877 	register_generic_request(req);
878 	{
879 		void *p = req->request->front.iov_base;
880 		void *const end = p + req->request->front_alloc_len;
881 
882 		ceph_encode_64(&p, req->tid); /* handle */
883 		ceph_encode_string(&p, end, what, strlen(what));
884 		WARN_ON(p != end);
885 	}
886 	send_generic_request(monc, req);
887 	mutex_unlock(&monc->mutex);
888 
889 	return req;
890 
891 err_put_req:
892 	put_generic_request(req);
893 	return ERR_PTR(-ENOMEM);
894 }
895 
896 /*
897  * Send MMonGetVersion and wait for the reply.
898  *
899  * @what: one of "mdsmap", "osdmap" or "monmap"
900  */
901 int ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
902 			  u64 *newest)
903 {
904 	struct ceph_mon_generic_request *req;
905 	int ret;
906 
907 	req = __ceph_monc_get_version(monc, what, NULL, 0);
908 	if (IS_ERR(req))
909 		return PTR_ERR(req);
910 
911 	ret = wait_generic_request(req);
912 	if (!ret)
913 		*newest = req->u.newest;
914 
915 	put_generic_request(req);
916 	return ret;
917 }
918 EXPORT_SYMBOL(ceph_monc_get_version);
919 
920 /*
921  * Send MMonGetVersion,
922  *
923  * @what: one of "mdsmap", "osdmap" or "monmap"
924  */
925 int ceph_monc_get_version_async(struct ceph_mon_client *monc, const char *what,
926 				ceph_monc_callback_t cb, u64 private_data)
927 {
928 	struct ceph_mon_generic_request *req;
929 
930 	req = __ceph_monc_get_version(monc, what, cb, private_data);
931 	if (IS_ERR(req))
932 		return PTR_ERR(req);
933 
934 	put_generic_request(req);
935 	return 0;
936 }
937 EXPORT_SYMBOL(ceph_monc_get_version_async);
938 
939 static void handle_command_ack(struct ceph_mon_client *monc,
940 			       struct ceph_msg *msg)
941 {
942 	struct ceph_mon_generic_request *req;
943 	void *p = msg->front.iov_base;
944 	void *const end = p + msg->front_alloc_len;
945 	u64 tid = le64_to_cpu(msg->hdr.tid);
946 
947 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
948 
949 	ceph_decode_need(&p, end, sizeof(struct ceph_mon_request_header) +
950 							    sizeof(u32), bad);
951 	p += sizeof(struct ceph_mon_request_header);
952 
953 	mutex_lock(&monc->mutex);
954 	req = lookup_generic_request(&monc->generic_request_tree, tid);
955 	if (!req) {
956 		mutex_unlock(&monc->mutex);
957 		return;
958 	}
959 
960 	req->result = ceph_decode_32(&p);
961 	__finish_generic_request(req);
962 	mutex_unlock(&monc->mutex);
963 
964 	complete_generic_request(req);
965 	return;
966 
967 bad:
968 	pr_err("corrupt mon_command ack, tid %llu\n", tid);
969 	ceph_msg_dump(msg);
970 }
971 
972 static __printf(2, 0)
973 int do_mon_command_vargs(struct ceph_mon_client *monc, const char *fmt,
974 			 va_list ap)
975 {
976 	struct ceph_mon_generic_request *req;
977 	struct ceph_mon_command *h;
978 	int ret = -ENOMEM;
979 	int len;
980 
981 	req = alloc_generic_request(monc, GFP_NOIO);
982 	if (!req)
983 		goto out;
984 
985 	req->request = ceph_msg_new(CEPH_MSG_MON_COMMAND, 256, GFP_NOIO, true);
986 	if (!req->request)
987 		goto out;
988 
989 	req->reply = ceph_msg_new(CEPH_MSG_MON_COMMAND_ACK, 512, GFP_NOIO,
990 				  true);
991 	if (!req->reply)
992 		goto out;
993 
994 	mutex_lock(&monc->mutex);
995 	register_generic_request(req);
996 	h = req->request->front.iov_base;
997 	h->monhdr.have_version = 0;
998 	h->monhdr.session_mon = cpu_to_le16(-1);
999 	h->monhdr.session_mon_tid = 0;
1000 	h->fsid = monc->monmap->fsid;
1001 	h->num_strs = cpu_to_le32(1);
1002 	len = vsprintf(h->str, fmt, ap);
1003 	h->str_len = cpu_to_le32(len);
1004 	send_generic_request(monc, req);
1005 	mutex_unlock(&monc->mutex);
1006 
1007 	ret = wait_generic_request(req);
1008 out:
1009 	put_generic_request(req);
1010 	return ret;
1011 }
1012 
1013 static __printf(2, 3)
1014 int do_mon_command(struct ceph_mon_client *monc, const char *fmt, ...)
1015 {
1016 	va_list ap;
1017 	int ret;
1018 
1019 	va_start(ap, fmt);
1020 	ret = do_mon_command_vargs(monc, fmt, ap);
1021 	va_end(ap);
1022 	return ret;
1023 }
1024 
1025 int ceph_monc_blocklist_add(struct ceph_mon_client *monc,
1026 			    struct ceph_entity_addr *client_addr)
1027 {
1028 	int ret;
1029 
1030 	ret = do_mon_command(monc,
1031 			     "{ \"prefix\": \"osd blocklist\", \
1032 				\"blocklistop\": \"add\", \
1033 				\"addr\": \"%pISpc/%u\" }",
1034 			     &client_addr->in_addr,
1035 			     le32_to_cpu(client_addr->nonce));
1036 	if (ret == -EINVAL) {
1037 		/*
1038 		 * The monitor returns EINVAL on an unrecognized command.
1039 		 * Try the legacy command -- it is exactly the same except
1040 		 * for the name.
1041 		 */
1042 		ret = do_mon_command(monc,
1043 				     "{ \"prefix\": \"osd blacklist\", \
1044 					\"blacklistop\": \"add\", \
1045 					\"addr\": \"%pISpc/%u\" }",
1046 				     &client_addr->in_addr,
1047 				     le32_to_cpu(client_addr->nonce));
1048 	}
1049 	if (ret)
1050 		return ret;
1051 
1052 	/*
1053 	 * Make sure we have the osdmap that includes the blocklist
1054 	 * entry.  This is needed to ensure that the OSDs pick up the
1055 	 * new blocklist before processing any future requests from
1056 	 * this client.
1057 	 */
1058 	return ceph_wait_for_latest_osdmap(monc->client, 0);
1059 }
1060 EXPORT_SYMBOL(ceph_monc_blocklist_add);
1061 
1062 /*
1063  * Resend pending generic requests.
1064  */
1065 static void __resend_generic_request(struct ceph_mon_client *monc)
1066 {
1067 	struct ceph_mon_generic_request *req;
1068 	struct rb_node *p;
1069 
1070 	for (p = rb_first(&monc->generic_request_tree); p; p = rb_next(p)) {
1071 		req = rb_entry(p, struct ceph_mon_generic_request, node);
1072 		ceph_msg_revoke(req->request);
1073 		ceph_msg_revoke_incoming(req->reply);
1074 		ceph_con_send(&monc->con, ceph_msg_get(req->request));
1075 	}
1076 }
1077 
1078 /*
1079  * Delayed work.  If we haven't mounted yet, retry.  Otherwise,
1080  * renew/retry subscription as needed (in case it is timing out, or we
1081  * got an ENOMEM).  And keep the monitor connection alive.
1082  */
1083 static void delayed_work(struct work_struct *work)
1084 {
1085 	struct ceph_mon_client *monc =
1086 		container_of(work, struct ceph_mon_client, delayed_work.work);
1087 
1088 	dout("monc delayed_work\n");
1089 	mutex_lock(&monc->mutex);
1090 	if (monc->hunting) {
1091 		dout("%s continuing hunt\n", __func__);
1092 		reopen_session(monc);
1093 	} else {
1094 		int is_auth = ceph_auth_is_authenticated(monc->auth);
1095 		if (ceph_con_keepalive_expired(&monc->con,
1096 					       CEPH_MONC_PING_TIMEOUT)) {
1097 			dout("monc keepalive timeout\n");
1098 			is_auth = 0;
1099 			reopen_session(monc);
1100 		}
1101 
1102 		if (!monc->hunting) {
1103 			ceph_con_keepalive(&monc->con);
1104 			__validate_auth(monc);
1105 			un_backoff(monc);
1106 		}
1107 
1108 		if (is_auth &&
1109 		    !(monc->con.peer_features & CEPH_FEATURE_MON_STATEFUL_SUB)) {
1110 			unsigned long now = jiffies;
1111 
1112 			dout("%s renew subs? now %lu renew after %lu\n",
1113 			     __func__, now, monc->sub_renew_after);
1114 			if (time_after_eq(now, monc->sub_renew_after))
1115 				__send_subscribe(monc);
1116 		}
1117 	}
1118 	__schedule_delayed(monc);
1119 	mutex_unlock(&monc->mutex);
1120 }
1121 
1122 /*
1123  * On startup, we build a temporary monmap populated with the IPs
1124  * provided by mount(2).
1125  */
1126 static int build_initial_monmap(struct ceph_mon_client *monc)
1127 {
1128 	__le32 my_type = ceph_msgr2(monc->client) ?
1129 		CEPH_ENTITY_ADDR_TYPE_MSGR2 : CEPH_ENTITY_ADDR_TYPE_LEGACY;
1130 	struct ceph_options *opt = monc->client->options;
1131 	int num_mon = opt->num_mon;
1132 	int i;
1133 
1134 	/* build initial monmap */
1135 	monc->monmap = kzalloc(struct_size(monc->monmap, mon_inst, num_mon),
1136 			       GFP_KERNEL);
1137 	if (!monc->monmap)
1138 		return -ENOMEM;
1139 
1140 	for (i = 0; i < num_mon; i++) {
1141 		struct ceph_entity_inst *inst = &monc->monmap->mon_inst[i];
1142 
1143 		memcpy(&inst->addr.in_addr, &opt->mon_addr[i].in_addr,
1144 		       sizeof(inst->addr.in_addr));
1145 		inst->addr.type = my_type;
1146 		inst->addr.nonce = 0;
1147 		inst->name.type = CEPH_ENTITY_TYPE_MON;
1148 		inst->name.num = cpu_to_le64(i);
1149 	}
1150 	monc->monmap->num_mon = num_mon;
1151 	return 0;
1152 }
1153 
1154 int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl)
1155 {
1156 	int err;
1157 
1158 	dout("init\n");
1159 	memset(monc, 0, sizeof(*monc));
1160 	monc->client = cl;
1161 	mutex_init(&monc->mutex);
1162 
1163 	err = build_initial_monmap(monc);
1164 	if (err)
1165 		goto out;
1166 
1167 	/* connection */
1168 	/* authentication */
1169 	monc->auth = ceph_auth_init(cl->options->name, cl->options->key,
1170 				    cl->options->con_modes);
1171 	if (IS_ERR(monc->auth)) {
1172 		err = PTR_ERR(monc->auth);
1173 		goto out_monmap;
1174 	}
1175 	monc->auth->want_keys =
1176 		CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON |
1177 		CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS;
1178 
1179 	/* msgs */
1180 	err = -ENOMEM;
1181 	monc->m_subscribe_ack = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE_ACK,
1182 				     sizeof(struct ceph_mon_subscribe_ack),
1183 				     GFP_KERNEL, true);
1184 	if (!monc->m_subscribe_ack)
1185 		goto out_auth;
1186 
1187 	monc->m_subscribe = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 128,
1188 					 GFP_KERNEL, true);
1189 	if (!monc->m_subscribe)
1190 		goto out_subscribe_ack;
1191 
1192 	monc->m_auth_reply = ceph_msg_new(CEPH_MSG_AUTH_REPLY, 4096,
1193 					  GFP_KERNEL, true);
1194 	if (!monc->m_auth_reply)
1195 		goto out_subscribe;
1196 
1197 	monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, GFP_KERNEL, true);
1198 	monc->pending_auth = 0;
1199 	if (!monc->m_auth)
1200 		goto out_auth_reply;
1201 
1202 	ceph_con_init(&monc->con, monc, &mon_con_ops,
1203 		      &monc->client->msgr);
1204 
1205 	monc->cur_mon = -1;
1206 	monc->had_a_connection = false;
1207 	monc->hunt_mult = 1;
1208 
1209 	INIT_DELAYED_WORK(&monc->delayed_work, delayed_work);
1210 	monc->generic_request_tree = RB_ROOT;
1211 	monc->last_tid = 0;
1212 
1213 	monc->fs_cluster_id = CEPH_FS_CLUSTER_ID_NONE;
1214 
1215 	return 0;
1216 
1217 out_auth_reply:
1218 	ceph_msg_put(monc->m_auth_reply);
1219 out_subscribe:
1220 	ceph_msg_put(monc->m_subscribe);
1221 out_subscribe_ack:
1222 	ceph_msg_put(monc->m_subscribe_ack);
1223 out_auth:
1224 	ceph_auth_destroy(monc->auth);
1225 out_monmap:
1226 	kfree(monc->monmap);
1227 out:
1228 	return err;
1229 }
1230 EXPORT_SYMBOL(ceph_monc_init);
1231 
1232 void ceph_monc_stop(struct ceph_mon_client *monc)
1233 {
1234 	dout("stop\n");
1235 	cancel_delayed_work_sync(&monc->delayed_work);
1236 
1237 	mutex_lock(&monc->mutex);
1238 	__close_session(monc);
1239 	monc->cur_mon = -1;
1240 	mutex_unlock(&monc->mutex);
1241 
1242 	/*
1243 	 * flush msgr queue before we destroy ourselves to ensure that:
1244 	 *  - any work that references our embedded con is finished.
1245 	 *  - any osd_client or other work that may reference an authorizer
1246 	 *    finishes before we shut down the auth subsystem.
1247 	 */
1248 	ceph_msgr_flush();
1249 
1250 	ceph_auth_destroy(monc->auth);
1251 
1252 	WARN_ON(!RB_EMPTY_ROOT(&monc->generic_request_tree));
1253 
1254 	ceph_msg_put(monc->m_auth);
1255 	ceph_msg_put(monc->m_auth_reply);
1256 	ceph_msg_put(monc->m_subscribe);
1257 	ceph_msg_put(monc->m_subscribe_ack);
1258 
1259 	kfree(monc->monmap);
1260 }
1261 EXPORT_SYMBOL(ceph_monc_stop);
1262 
1263 static void finish_hunting(struct ceph_mon_client *monc)
1264 {
1265 	if (monc->hunting) {
1266 		dout("%s found mon%d\n", __func__, monc->cur_mon);
1267 		monc->hunting = false;
1268 		monc->had_a_connection = true;
1269 		un_backoff(monc);
1270 		__schedule_delayed(monc);
1271 	}
1272 }
1273 
1274 static void finish_auth(struct ceph_mon_client *monc, int auth_err,
1275 			bool was_authed)
1276 {
1277 	dout("%s auth_err %d was_authed %d\n", __func__, auth_err, was_authed);
1278 	WARN_ON(auth_err > 0);
1279 
1280 	monc->pending_auth = 0;
1281 	if (auth_err) {
1282 		monc->client->auth_err = auth_err;
1283 		wake_up_all(&monc->client->auth_wq);
1284 		return;
1285 	}
1286 
1287 	if (!was_authed && ceph_auth_is_authenticated(monc->auth)) {
1288 		dout("%s authenticated, starting session global_id %llu\n",
1289 		     __func__, monc->auth->global_id);
1290 
1291 		monc->client->msgr.inst.name.type = CEPH_ENTITY_TYPE_CLIENT;
1292 		monc->client->msgr.inst.name.num =
1293 					cpu_to_le64(monc->auth->global_id);
1294 
1295 		__send_subscribe(monc);
1296 		__resend_generic_request(monc);
1297 
1298 		pr_info("mon%d %s session established\n", monc->cur_mon,
1299 			ceph_pr_addr(&monc->con.peer_addr));
1300 	}
1301 }
1302 
1303 static void handle_auth_reply(struct ceph_mon_client *monc,
1304 			      struct ceph_msg *msg)
1305 {
1306 	bool was_authed;
1307 	int ret;
1308 
1309 	mutex_lock(&monc->mutex);
1310 	was_authed = ceph_auth_is_authenticated(monc->auth);
1311 	ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base,
1312 				     msg->front.iov_len,
1313 				     monc->m_auth->front.iov_base,
1314 				     monc->m_auth->front_alloc_len);
1315 	if (ret > 0) {
1316 		__send_prepared_auth_request(monc, ret);
1317 	} else {
1318 		finish_auth(monc, ret, was_authed);
1319 		finish_hunting(monc);
1320 	}
1321 	mutex_unlock(&monc->mutex);
1322 }
1323 
1324 static int __validate_auth(struct ceph_mon_client *monc)
1325 {
1326 	int ret;
1327 
1328 	if (monc->pending_auth)
1329 		return 0;
1330 
1331 	ret = ceph_build_auth(monc->auth, monc->m_auth->front.iov_base,
1332 			      monc->m_auth->front_alloc_len);
1333 	if (ret <= 0)
1334 		return ret; /* either an error, or no need to authenticate */
1335 	__send_prepared_auth_request(monc, ret);
1336 	return 0;
1337 }
1338 
1339 int ceph_monc_validate_auth(struct ceph_mon_client *monc)
1340 {
1341 	int ret;
1342 
1343 	mutex_lock(&monc->mutex);
1344 	ret = __validate_auth(monc);
1345 	mutex_unlock(&monc->mutex);
1346 	return ret;
1347 }
1348 EXPORT_SYMBOL(ceph_monc_validate_auth);
1349 
1350 static int mon_get_auth_request(struct ceph_connection *con,
1351 				void *buf, int *buf_len,
1352 				void **authorizer, int *authorizer_len)
1353 {
1354 	struct ceph_mon_client *monc = con->private;
1355 	int ret;
1356 
1357 	mutex_lock(&monc->mutex);
1358 	ret = ceph_auth_get_request(monc->auth, buf, *buf_len);
1359 	mutex_unlock(&monc->mutex);
1360 	if (ret < 0)
1361 		return ret;
1362 
1363 	*buf_len = ret;
1364 	*authorizer = NULL;
1365 	*authorizer_len = 0;
1366 	return 0;
1367 }
1368 
1369 static int mon_handle_auth_reply_more(struct ceph_connection *con,
1370 				      void *reply, int reply_len,
1371 				      void *buf, int *buf_len,
1372 				      void **authorizer, int *authorizer_len)
1373 {
1374 	struct ceph_mon_client *monc = con->private;
1375 	int ret;
1376 
1377 	mutex_lock(&monc->mutex);
1378 	ret = ceph_auth_handle_reply_more(monc->auth, reply, reply_len,
1379 					  buf, *buf_len);
1380 	mutex_unlock(&monc->mutex);
1381 	if (ret < 0)
1382 		return ret;
1383 
1384 	*buf_len = ret;
1385 	*authorizer = NULL;
1386 	*authorizer_len = 0;
1387 	return 0;
1388 }
1389 
1390 static int mon_handle_auth_done(struct ceph_connection *con,
1391 				u64 global_id, void *reply, int reply_len,
1392 				u8 *session_key, int *session_key_len,
1393 				u8 *con_secret, int *con_secret_len)
1394 {
1395 	struct ceph_mon_client *monc = con->private;
1396 	bool was_authed;
1397 	int ret;
1398 
1399 	mutex_lock(&monc->mutex);
1400 	WARN_ON(!monc->hunting);
1401 	was_authed = ceph_auth_is_authenticated(monc->auth);
1402 	ret = ceph_auth_handle_reply_done(monc->auth, global_id,
1403 					  reply, reply_len,
1404 					  session_key, session_key_len,
1405 					  con_secret, con_secret_len);
1406 	finish_auth(monc, ret, was_authed);
1407 	if (!ret)
1408 		finish_hunting(monc);
1409 	mutex_unlock(&monc->mutex);
1410 	return 0;
1411 }
1412 
1413 static int mon_handle_auth_bad_method(struct ceph_connection *con,
1414 				      int used_proto, int result,
1415 				      const int *allowed_protos, int proto_cnt,
1416 				      const int *allowed_modes, int mode_cnt)
1417 {
1418 	struct ceph_mon_client *monc = con->private;
1419 	bool was_authed;
1420 
1421 	mutex_lock(&monc->mutex);
1422 	WARN_ON(!monc->hunting);
1423 	was_authed = ceph_auth_is_authenticated(monc->auth);
1424 	ceph_auth_handle_bad_method(monc->auth, used_proto, result,
1425 				    allowed_protos, proto_cnt,
1426 				    allowed_modes, mode_cnt);
1427 	finish_auth(monc, -EACCES, was_authed);
1428 	mutex_unlock(&monc->mutex);
1429 	return 0;
1430 }
1431 
1432 /*
1433  * handle incoming message
1434  */
1435 static void mon_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
1436 {
1437 	struct ceph_mon_client *monc = con->private;
1438 	int type = le16_to_cpu(msg->hdr.type);
1439 
1440 	switch (type) {
1441 	case CEPH_MSG_AUTH_REPLY:
1442 		handle_auth_reply(monc, msg);
1443 		break;
1444 
1445 	case CEPH_MSG_MON_SUBSCRIBE_ACK:
1446 		handle_subscribe_ack(monc, msg);
1447 		break;
1448 
1449 	case CEPH_MSG_STATFS_REPLY:
1450 		handle_statfs_reply(monc, msg);
1451 		break;
1452 
1453 	case CEPH_MSG_MON_GET_VERSION_REPLY:
1454 		handle_get_version_reply(monc, msg);
1455 		break;
1456 
1457 	case CEPH_MSG_MON_COMMAND_ACK:
1458 		handle_command_ack(monc, msg);
1459 		break;
1460 
1461 	case CEPH_MSG_MON_MAP:
1462 		ceph_monc_handle_map(monc, msg);
1463 		break;
1464 
1465 	case CEPH_MSG_OSD_MAP:
1466 		ceph_osdc_handle_map(&monc->client->osdc, msg);
1467 		break;
1468 
1469 	default:
1470 		/* can the chained handler handle it? */
1471 		if (monc->client->extra_mon_dispatch &&
1472 		    monc->client->extra_mon_dispatch(monc->client, msg) == 0)
1473 			break;
1474 
1475 		pr_err("received unknown message type %d %s\n", type,
1476 		       ceph_msg_type_name(type));
1477 	}
1478 	ceph_msg_put(msg);
1479 }
1480 
1481 /*
1482  * Allocate memory for incoming message
1483  */
1484 static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con,
1485 				      struct ceph_msg_header *hdr,
1486 				      int *skip)
1487 {
1488 	struct ceph_mon_client *monc = con->private;
1489 	int type = le16_to_cpu(hdr->type);
1490 	int front_len = le32_to_cpu(hdr->front_len);
1491 	struct ceph_msg *m = NULL;
1492 
1493 	*skip = 0;
1494 
1495 	switch (type) {
1496 	case CEPH_MSG_MON_SUBSCRIBE_ACK:
1497 		m = ceph_msg_get(monc->m_subscribe_ack);
1498 		break;
1499 	case CEPH_MSG_STATFS_REPLY:
1500 	case CEPH_MSG_MON_COMMAND_ACK:
1501 		return get_generic_reply(con, hdr, skip);
1502 	case CEPH_MSG_AUTH_REPLY:
1503 		m = ceph_msg_get(monc->m_auth_reply);
1504 		break;
1505 	case CEPH_MSG_MON_GET_VERSION_REPLY:
1506 		if (le64_to_cpu(hdr->tid) != 0)
1507 			return get_generic_reply(con, hdr, skip);
1508 
1509 		/*
1510 		 * Older OSDs don't set reply tid even if the original
1511 		 * request had a non-zero tid.  Work around this weirdness
1512 		 * by allocating a new message.
1513 		 */
1514 		fallthrough;
1515 	case CEPH_MSG_MON_MAP:
1516 	case CEPH_MSG_MDS_MAP:
1517 	case CEPH_MSG_OSD_MAP:
1518 	case CEPH_MSG_FS_MAP_USER:
1519 		m = ceph_msg_new(type, front_len, GFP_NOFS, false);
1520 		if (!m)
1521 			return NULL;	/* ENOMEM--return skip == 0 */
1522 		break;
1523 	}
1524 
1525 	if (!m) {
1526 		pr_info("alloc_msg unknown type %d\n", type);
1527 		*skip = 1;
1528 	} else if (front_len > m->front_alloc_len) {
1529 		pr_warn("mon_alloc_msg front %d > prealloc %d (%u#%llu)\n",
1530 			front_len, m->front_alloc_len,
1531 			(unsigned int)con->peer_name.type,
1532 			le64_to_cpu(con->peer_name.num));
1533 		ceph_msg_put(m);
1534 		m = ceph_msg_new(type, front_len, GFP_NOFS, false);
1535 	}
1536 
1537 	return m;
1538 }
1539 
1540 /*
1541  * If the monitor connection resets, pick a new monitor and resubmit
1542  * any pending requests.
1543  */
1544 static void mon_fault(struct ceph_connection *con)
1545 {
1546 	struct ceph_mon_client *monc = con->private;
1547 
1548 	mutex_lock(&monc->mutex);
1549 	dout("%s mon%d\n", __func__, monc->cur_mon);
1550 	if (monc->cur_mon >= 0) {
1551 		if (!monc->hunting) {
1552 			dout("%s hunting for new mon\n", __func__);
1553 			reopen_session(monc);
1554 			__schedule_delayed(monc);
1555 		} else {
1556 			dout("%s already hunting\n", __func__);
1557 		}
1558 	}
1559 	mutex_unlock(&monc->mutex);
1560 }
1561 
1562 /*
1563  * We can ignore refcounting on the connection struct, as all references
1564  * will come from the messenger workqueue, which is drained prior to
1565  * mon_client destruction.
1566  */
1567 static struct ceph_connection *mon_get_con(struct ceph_connection *con)
1568 {
1569 	return con;
1570 }
1571 
1572 static void mon_put_con(struct ceph_connection *con)
1573 {
1574 }
1575 
1576 static const struct ceph_connection_operations mon_con_ops = {
1577 	.get = mon_get_con,
1578 	.put = mon_put_con,
1579 	.alloc_msg = mon_alloc_msg,
1580 	.dispatch = mon_dispatch,
1581 	.fault = mon_fault,
1582 	.get_auth_request = mon_get_auth_request,
1583 	.handle_auth_reply_more = mon_handle_auth_reply_more,
1584 	.handle_auth_done = mon_handle_auth_done,
1585 	.handle_auth_bad_method = mon_handle_auth_bad_method,
1586 };
1587