xref: /linux/net/ceph/mon_client.c (revision 25aee3debe0464f6c680173041fa3de30ec9ff54)
1 #include <linux/ceph/ceph_debug.h>
2 
3 #include <linux/module.h>
4 #include <linux/types.h>
5 #include <linux/slab.h>
6 #include <linux/random.h>
7 #include <linux/sched.h>
8 
9 #include <linux/ceph/mon_client.h>
10 #include <linux/ceph/libceph.h>
11 #include <linux/ceph/debugfs.h>
12 #include <linux/ceph/decode.h>
13 #include <linux/ceph/auth.h>
14 
15 /*
16  * Interact with Ceph monitor cluster.  Handle requests for new map
17  * versions, and periodically resend as needed.  Also implement
18  * statfs() and umount().
19  *
20  * A small cluster of Ceph "monitors" are responsible for managing critical
21  * cluster configuration and state information.  An odd number (e.g., 3, 5)
22  * of cmon daemons use a modified version of the Paxos part-time parliament
23  * algorithm to manage the MDS map (mds cluster membership), OSD map, and
24  * list of clients who have mounted the file system.
25  *
26  * We maintain an open, active session with a monitor at all times in order to
27  * receive timely MDSMap updates.  We periodically send a keepalive byte on the
28  * TCP socket to ensure we detect a failure.  If the connection does break, we
29  * randomly hunt for a new monitor.  Once the connection is reestablished, we
30  * resend any outstanding requests.
31  */
32 
33 static const struct ceph_connection_operations mon_con_ops;
34 
35 static int __validate_auth(struct ceph_mon_client *monc);
36 
37 /*
38  * Decode a monmap blob (e.g., during mount).
39  */
40 struct ceph_monmap *ceph_monmap_decode(void *p, void *end)
41 {
42 	struct ceph_monmap *m = NULL;
43 	int i, err = -EINVAL;
44 	struct ceph_fsid fsid;
45 	u32 epoch, num_mon;
46 	u16 version;
47 	u32 len;
48 
49 	ceph_decode_32_safe(&p, end, len, bad);
50 	ceph_decode_need(&p, end, len, bad);
51 
52 	dout("monmap_decode %p %p len %d\n", p, end, (int)(end-p));
53 
54 	ceph_decode_16_safe(&p, end, version, bad);
55 
56 	ceph_decode_need(&p, end, sizeof(fsid) + 2*sizeof(u32), bad);
57 	ceph_decode_copy(&p, &fsid, sizeof(fsid));
58 	epoch = ceph_decode_32(&p);
59 
60 	num_mon = ceph_decode_32(&p);
61 	ceph_decode_need(&p, end, num_mon*sizeof(m->mon_inst[0]), bad);
62 
63 	if (num_mon >= CEPH_MAX_MON)
64 		goto bad;
65 	m = kmalloc(sizeof(*m) + sizeof(m->mon_inst[0])*num_mon, GFP_NOFS);
66 	if (m == NULL)
67 		return ERR_PTR(-ENOMEM);
68 	m->fsid = fsid;
69 	m->epoch = epoch;
70 	m->num_mon = num_mon;
71 	ceph_decode_copy(&p, m->mon_inst, num_mon*sizeof(m->mon_inst[0]));
72 	for (i = 0; i < num_mon; i++)
73 		ceph_decode_addr(&m->mon_inst[i].addr);
74 
75 	dout("monmap_decode epoch %d, num_mon %d\n", m->epoch,
76 	     m->num_mon);
77 	for (i = 0; i < m->num_mon; i++)
78 		dout("monmap_decode  mon%d is %s\n", i,
79 		     ceph_pr_addr(&m->mon_inst[i].addr.in_addr));
80 	return m;
81 
82 bad:
83 	dout("monmap_decode failed with %d\n", err);
84 	kfree(m);
85 	return ERR_PTR(err);
86 }
87 
88 /*
89  * return true if *addr is included in the monmap.
90  */
91 int ceph_monmap_contains(struct ceph_monmap *m, struct ceph_entity_addr *addr)
92 {
93 	int i;
94 
95 	for (i = 0; i < m->num_mon; i++)
96 		if (memcmp(addr, &m->mon_inst[i].addr, sizeof(*addr)) == 0)
97 			return 1;
98 	return 0;
99 }
100 
101 /*
102  * Send an auth request.
103  */
104 static void __send_prepared_auth_request(struct ceph_mon_client *monc, int len)
105 {
106 	monc->pending_auth = 1;
107 	monc->m_auth->front.iov_len = len;
108 	monc->m_auth->hdr.front_len = cpu_to_le32(len);
109 	ceph_msg_revoke(monc->m_auth);
110 	ceph_msg_get(monc->m_auth);  /* keep our ref */
111 	ceph_con_send(&monc->con, monc->m_auth);
112 }
113 
114 /*
115  * Close monitor session, if any.
116  */
117 static void __close_session(struct ceph_mon_client *monc)
118 {
119 	dout("__close_session closing mon%d\n", monc->cur_mon);
120 	ceph_msg_revoke(monc->m_auth);
121 	ceph_msg_revoke_incoming(monc->m_auth_reply);
122 	ceph_msg_revoke(monc->m_subscribe);
123 	ceph_msg_revoke_incoming(monc->m_subscribe_ack);
124 	ceph_con_close(&monc->con);
125 	monc->cur_mon = -1;
126 	monc->pending_auth = 0;
127 	ceph_auth_reset(monc->auth);
128 }
129 
130 /*
131  * Open a session with a (new) monitor.
132  */
133 static int __open_session(struct ceph_mon_client *monc)
134 {
135 	char r;
136 	int ret;
137 
138 	if (monc->cur_mon < 0) {
139 		get_random_bytes(&r, 1);
140 		monc->cur_mon = r % monc->monmap->num_mon;
141 		dout("open_session num=%d r=%d -> mon%d\n",
142 		     monc->monmap->num_mon, r, monc->cur_mon);
143 		monc->sub_sent = 0;
144 		monc->sub_renew_after = jiffies;  /* i.e., expired */
145 		monc->want_next_osdmap = !!monc->want_next_osdmap;
146 
147 		dout("open_session mon%d opening\n", monc->cur_mon);
148 		ceph_con_open(&monc->con,
149 			      CEPH_ENTITY_TYPE_MON, monc->cur_mon,
150 			      &monc->monmap->mon_inst[monc->cur_mon].addr);
151 
152 		/* initiatiate authentication handshake */
153 		ret = ceph_auth_build_hello(monc->auth,
154 					    monc->m_auth->front.iov_base,
155 					    monc->m_auth->front_max);
156 		__send_prepared_auth_request(monc, ret);
157 	} else {
158 		dout("open_session mon%d already open\n", monc->cur_mon);
159 	}
160 	return 0;
161 }
162 
163 static bool __sub_expired(struct ceph_mon_client *monc)
164 {
165 	return time_after_eq(jiffies, monc->sub_renew_after);
166 }
167 
168 /*
169  * Reschedule delayed work timer.
170  */
171 static void __schedule_delayed(struct ceph_mon_client *monc)
172 {
173 	unsigned int delay;
174 
175 	if (monc->cur_mon < 0 || __sub_expired(monc))
176 		delay = 10 * HZ;
177 	else
178 		delay = 20 * HZ;
179 	dout("__schedule_delayed after %u\n", delay);
180 	schedule_delayed_work(&monc->delayed_work, delay);
181 }
182 
183 /*
184  * Send subscribe request for mdsmap and/or osdmap.
185  */
186 static void __send_subscribe(struct ceph_mon_client *monc)
187 {
188 	dout("__send_subscribe sub_sent=%u exp=%u want_osd=%d\n",
189 	     (unsigned int)monc->sub_sent, __sub_expired(monc),
190 	     monc->want_next_osdmap);
191 	if ((__sub_expired(monc) && !monc->sub_sent) ||
192 	    monc->want_next_osdmap == 1) {
193 		struct ceph_msg *msg = monc->m_subscribe;
194 		struct ceph_mon_subscribe_item *i;
195 		void *p, *end;
196 		int num;
197 
198 		p = msg->front.iov_base;
199 		end = p + msg->front_max;
200 
201 		num = 1 + !!monc->want_next_osdmap + !!monc->want_mdsmap;
202 		ceph_encode_32(&p, num);
203 
204 		if (monc->want_next_osdmap) {
205 			dout("__send_subscribe to 'osdmap' %u\n",
206 			     (unsigned int)monc->have_osdmap);
207 			ceph_encode_string(&p, end, "osdmap", 6);
208 			i = p;
209 			i->have = cpu_to_le64(monc->have_osdmap);
210 			i->onetime = 1;
211 			p += sizeof(*i);
212 			monc->want_next_osdmap = 2;  /* requested */
213 		}
214 		if (monc->want_mdsmap) {
215 			dout("__send_subscribe to 'mdsmap' %u+\n",
216 			     (unsigned int)monc->have_mdsmap);
217 			ceph_encode_string(&p, end, "mdsmap", 6);
218 			i = p;
219 			i->have = cpu_to_le64(monc->have_mdsmap);
220 			i->onetime = 0;
221 			p += sizeof(*i);
222 		}
223 		ceph_encode_string(&p, end, "monmap", 6);
224 		i = p;
225 		i->have = 0;
226 		i->onetime = 0;
227 		p += sizeof(*i);
228 
229 		msg->front.iov_len = p - msg->front.iov_base;
230 		msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
231 		ceph_msg_revoke(msg);
232 		ceph_con_send(&monc->con, ceph_msg_get(msg));
233 
234 		monc->sub_sent = jiffies | 1;  /* never 0 */
235 	}
236 }
237 
238 static void handle_subscribe_ack(struct ceph_mon_client *monc,
239 				 struct ceph_msg *msg)
240 {
241 	unsigned int seconds;
242 	struct ceph_mon_subscribe_ack *h = msg->front.iov_base;
243 
244 	if (msg->front.iov_len < sizeof(*h))
245 		goto bad;
246 	seconds = le32_to_cpu(h->duration);
247 
248 	mutex_lock(&monc->mutex);
249 	if (monc->hunting) {
250 		pr_info("mon%d %s session established\n",
251 			monc->cur_mon,
252 			ceph_pr_addr(&monc->con.peer_addr.in_addr));
253 		monc->hunting = false;
254 	}
255 	dout("handle_subscribe_ack after %d seconds\n", seconds);
256 	monc->sub_renew_after = monc->sub_sent + (seconds >> 1)*HZ - 1;
257 	monc->sub_sent = 0;
258 	mutex_unlock(&monc->mutex);
259 	return;
260 bad:
261 	pr_err("got corrupt subscribe-ack msg\n");
262 	ceph_msg_dump(msg);
263 }
264 
265 /*
266  * Keep track of which maps we have
267  */
268 int ceph_monc_got_mdsmap(struct ceph_mon_client *monc, u32 got)
269 {
270 	mutex_lock(&monc->mutex);
271 	monc->have_mdsmap = got;
272 	mutex_unlock(&monc->mutex);
273 	return 0;
274 }
275 EXPORT_SYMBOL(ceph_monc_got_mdsmap);
276 
277 int ceph_monc_got_osdmap(struct ceph_mon_client *monc, u32 got)
278 {
279 	mutex_lock(&monc->mutex);
280 	monc->have_osdmap = got;
281 	monc->want_next_osdmap = 0;
282 	mutex_unlock(&monc->mutex);
283 	return 0;
284 }
285 
286 /*
287  * Register interest in the next osdmap
288  */
289 void ceph_monc_request_next_osdmap(struct ceph_mon_client *monc)
290 {
291 	dout("request_next_osdmap have %u\n", monc->have_osdmap);
292 	mutex_lock(&monc->mutex);
293 	if (!monc->want_next_osdmap)
294 		monc->want_next_osdmap = 1;
295 	if (monc->want_next_osdmap < 2)
296 		__send_subscribe(monc);
297 	mutex_unlock(&monc->mutex);
298 }
299 
300 /*
301  *
302  */
303 int ceph_monc_open_session(struct ceph_mon_client *monc)
304 {
305 	mutex_lock(&monc->mutex);
306 	__open_session(monc);
307 	__schedule_delayed(monc);
308 	mutex_unlock(&monc->mutex);
309 	return 0;
310 }
311 EXPORT_SYMBOL(ceph_monc_open_session);
312 
313 /*
314  * The monitor responds with mount ack indicate mount success.  The
315  * included client ticket allows the client to talk to MDSs and OSDs.
316  */
317 static void ceph_monc_handle_map(struct ceph_mon_client *monc,
318 				 struct ceph_msg *msg)
319 {
320 	struct ceph_client *client = monc->client;
321 	struct ceph_monmap *monmap = NULL, *old = monc->monmap;
322 	void *p, *end;
323 
324 	mutex_lock(&monc->mutex);
325 
326 	dout("handle_monmap\n");
327 	p = msg->front.iov_base;
328 	end = p + msg->front.iov_len;
329 
330 	monmap = ceph_monmap_decode(p, end);
331 	if (IS_ERR(monmap)) {
332 		pr_err("problem decoding monmap, %d\n",
333 		       (int)PTR_ERR(monmap));
334 		goto out;
335 	}
336 
337 	if (ceph_check_fsid(monc->client, &monmap->fsid) < 0) {
338 		kfree(monmap);
339 		goto out;
340 	}
341 
342 	client->monc.monmap = monmap;
343 	kfree(old);
344 
345 	if (!client->have_fsid) {
346 		client->have_fsid = true;
347 		mutex_unlock(&monc->mutex);
348 		/*
349 		 * do debugfs initialization without mutex to avoid
350 		 * creating a locking dependency
351 		 */
352 		ceph_debugfs_client_init(client);
353 		goto out_unlocked;
354 	}
355 out:
356 	mutex_unlock(&monc->mutex);
357 out_unlocked:
358 	wake_up_all(&client->auth_wq);
359 }
360 
361 /*
362  * generic requests (e.g., statfs, poolop)
363  */
364 static struct ceph_mon_generic_request *__lookup_generic_req(
365 	struct ceph_mon_client *monc, u64 tid)
366 {
367 	struct ceph_mon_generic_request *req;
368 	struct rb_node *n = monc->generic_request_tree.rb_node;
369 
370 	while (n) {
371 		req = rb_entry(n, struct ceph_mon_generic_request, node);
372 		if (tid < req->tid)
373 			n = n->rb_left;
374 		else if (tid > req->tid)
375 			n = n->rb_right;
376 		else
377 			return req;
378 	}
379 	return NULL;
380 }
381 
382 static void __insert_generic_request(struct ceph_mon_client *monc,
383 			    struct ceph_mon_generic_request *new)
384 {
385 	struct rb_node **p = &monc->generic_request_tree.rb_node;
386 	struct rb_node *parent = NULL;
387 	struct ceph_mon_generic_request *req = NULL;
388 
389 	while (*p) {
390 		parent = *p;
391 		req = rb_entry(parent, struct ceph_mon_generic_request, node);
392 		if (new->tid < req->tid)
393 			p = &(*p)->rb_left;
394 		else if (new->tid > req->tid)
395 			p = &(*p)->rb_right;
396 		else
397 			BUG();
398 	}
399 
400 	rb_link_node(&new->node, parent, p);
401 	rb_insert_color(&new->node, &monc->generic_request_tree);
402 }
403 
404 static void release_generic_request(struct kref *kref)
405 {
406 	struct ceph_mon_generic_request *req =
407 		container_of(kref, struct ceph_mon_generic_request, kref);
408 
409 	if (req->reply)
410 		ceph_msg_put(req->reply);
411 	if (req->request)
412 		ceph_msg_put(req->request);
413 
414 	kfree(req);
415 }
416 
417 static void put_generic_request(struct ceph_mon_generic_request *req)
418 {
419 	kref_put(&req->kref, release_generic_request);
420 }
421 
422 static void get_generic_request(struct ceph_mon_generic_request *req)
423 {
424 	kref_get(&req->kref);
425 }
426 
427 static struct ceph_msg *get_generic_reply(struct ceph_connection *con,
428 					 struct ceph_msg_header *hdr,
429 					 int *skip)
430 {
431 	struct ceph_mon_client *monc = con->private;
432 	struct ceph_mon_generic_request *req;
433 	u64 tid = le64_to_cpu(hdr->tid);
434 	struct ceph_msg *m;
435 
436 	mutex_lock(&monc->mutex);
437 	req = __lookup_generic_req(monc, tid);
438 	if (!req) {
439 		dout("get_generic_reply %lld dne\n", tid);
440 		*skip = 1;
441 		m = NULL;
442 	} else {
443 		dout("get_generic_reply %lld got %p\n", tid, req->reply);
444 		*skip = 0;
445 		m = ceph_msg_get(req->reply);
446 		/*
447 		 * we don't need to track the connection reading into
448 		 * this reply because we only have one open connection
449 		 * at a time, ever.
450 		 */
451 	}
452 	mutex_unlock(&monc->mutex);
453 	return m;
454 }
455 
456 static int do_generic_request(struct ceph_mon_client *monc,
457 			      struct ceph_mon_generic_request *req)
458 {
459 	int err;
460 
461 	/* register request */
462 	mutex_lock(&monc->mutex);
463 	req->tid = ++monc->last_tid;
464 	req->request->hdr.tid = cpu_to_le64(req->tid);
465 	__insert_generic_request(monc, req);
466 	monc->num_generic_requests++;
467 	ceph_con_send(&monc->con, ceph_msg_get(req->request));
468 	mutex_unlock(&monc->mutex);
469 
470 	err = wait_for_completion_interruptible(&req->completion);
471 
472 	mutex_lock(&monc->mutex);
473 	rb_erase(&req->node, &monc->generic_request_tree);
474 	monc->num_generic_requests--;
475 	mutex_unlock(&monc->mutex);
476 
477 	if (!err)
478 		err = req->result;
479 	return err;
480 }
481 
482 /*
483  * statfs
484  */
485 static void handle_statfs_reply(struct ceph_mon_client *monc,
486 				struct ceph_msg *msg)
487 {
488 	struct ceph_mon_generic_request *req;
489 	struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
490 	u64 tid = le64_to_cpu(msg->hdr.tid);
491 
492 	if (msg->front.iov_len != sizeof(*reply))
493 		goto bad;
494 	dout("handle_statfs_reply %p tid %llu\n", msg, tid);
495 
496 	mutex_lock(&monc->mutex);
497 	req = __lookup_generic_req(monc, tid);
498 	if (req) {
499 		*(struct ceph_statfs *)req->buf = reply->st;
500 		req->result = 0;
501 		get_generic_request(req);
502 	}
503 	mutex_unlock(&monc->mutex);
504 	if (req) {
505 		complete_all(&req->completion);
506 		put_generic_request(req);
507 	}
508 	return;
509 
510 bad:
511 	pr_err("corrupt generic reply, tid %llu\n", tid);
512 	ceph_msg_dump(msg);
513 }
514 
515 /*
516  * Do a synchronous statfs().
517  */
518 int ceph_monc_do_statfs(struct ceph_mon_client *monc, struct ceph_statfs *buf)
519 {
520 	struct ceph_mon_generic_request *req;
521 	struct ceph_mon_statfs *h;
522 	int err;
523 
524 	req = kzalloc(sizeof(*req), GFP_NOFS);
525 	if (!req)
526 		return -ENOMEM;
527 
528 	kref_init(&req->kref);
529 	req->buf = buf;
530 	req->buf_len = sizeof(*buf);
531 	init_completion(&req->completion);
532 
533 	err = -ENOMEM;
534 	req->request = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), GFP_NOFS,
535 				    true);
536 	if (!req->request)
537 		goto out;
538 	req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 1024, GFP_NOFS,
539 				  true);
540 	if (!req->reply)
541 		goto out;
542 
543 	/* fill out request */
544 	h = req->request->front.iov_base;
545 	h->monhdr.have_version = 0;
546 	h->monhdr.session_mon = cpu_to_le16(-1);
547 	h->monhdr.session_mon_tid = 0;
548 	h->fsid = monc->monmap->fsid;
549 
550 	err = do_generic_request(monc, req);
551 
552 out:
553 	kref_put(&req->kref, release_generic_request);
554 	return err;
555 }
556 EXPORT_SYMBOL(ceph_monc_do_statfs);
557 
558 /*
559  * pool ops
560  */
561 static int get_poolop_reply_buf(const char *src, size_t src_len,
562 				char *dst, size_t dst_len)
563 {
564 	u32 buf_len;
565 
566 	if (src_len != sizeof(u32) + dst_len)
567 		return -EINVAL;
568 
569 	buf_len = le32_to_cpu(*(u32 *)src);
570 	if (buf_len != dst_len)
571 		return -EINVAL;
572 
573 	memcpy(dst, src + sizeof(u32), dst_len);
574 	return 0;
575 }
576 
577 static void handle_poolop_reply(struct ceph_mon_client *monc,
578 				struct ceph_msg *msg)
579 {
580 	struct ceph_mon_generic_request *req;
581 	struct ceph_mon_poolop_reply *reply = msg->front.iov_base;
582 	u64 tid = le64_to_cpu(msg->hdr.tid);
583 
584 	if (msg->front.iov_len < sizeof(*reply))
585 		goto bad;
586 	dout("handle_poolop_reply %p tid %llu\n", msg, tid);
587 
588 	mutex_lock(&monc->mutex);
589 	req = __lookup_generic_req(monc, tid);
590 	if (req) {
591 		if (req->buf_len &&
592 		    get_poolop_reply_buf(msg->front.iov_base + sizeof(*reply),
593 				     msg->front.iov_len - sizeof(*reply),
594 				     req->buf, req->buf_len) < 0) {
595 			mutex_unlock(&monc->mutex);
596 			goto bad;
597 		}
598 		req->result = le32_to_cpu(reply->reply_code);
599 		get_generic_request(req);
600 	}
601 	mutex_unlock(&monc->mutex);
602 	if (req) {
603 		complete(&req->completion);
604 		put_generic_request(req);
605 	}
606 	return;
607 
608 bad:
609 	pr_err("corrupt generic reply, tid %llu\n", tid);
610 	ceph_msg_dump(msg);
611 }
612 
613 /*
614  * Do a synchronous pool op.
615  */
616 int ceph_monc_do_poolop(struct ceph_mon_client *monc, u32 op,
617 			u32 pool, u64 snapid,
618 			char *buf, int len)
619 {
620 	struct ceph_mon_generic_request *req;
621 	struct ceph_mon_poolop *h;
622 	int err;
623 
624 	req = kzalloc(sizeof(*req), GFP_NOFS);
625 	if (!req)
626 		return -ENOMEM;
627 
628 	kref_init(&req->kref);
629 	req->buf = buf;
630 	req->buf_len = len;
631 	init_completion(&req->completion);
632 
633 	err = -ENOMEM;
634 	req->request = ceph_msg_new(CEPH_MSG_POOLOP, sizeof(*h), GFP_NOFS,
635 				    true);
636 	if (!req->request)
637 		goto out;
638 	req->reply = ceph_msg_new(CEPH_MSG_POOLOP_REPLY, 1024, GFP_NOFS,
639 				  true);
640 	if (!req->reply)
641 		goto out;
642 
643 	/* fill out request */
644 	req->request->hdr.version = cpu_to_le16(2);
645 	h = req->request->front.iov_base;
646 	h->monhdr.have_version = 0;
647 	h->monhdr.session_mon = cpu_to_le16(-1);
648 	h->monhdr.session_mon_tid = 0;
649 	h->fsid = monc->monmap->fsid;
650 	h->pool = cpu_to_le32(pool);
651 	h->op = cpu_to_le32(op);
652 	h->auid = 0;
653 	h->snapid = cpu_to_le64(snapid);
654 	h->name_len = 0;
655 
656 	err = do_generic_request(monc, req);
657 
658 out:
659 	kref_put(&req->kref, release_generic_request);
660 	return err;
661 }
662 
663 int ceph_monc_create_snapid(struct ceph_mon_client *monc,
664 			    u32 pool, u64 *snapid)
665 {
666 	return ceph_monc_do_poolop(monc,  POOL_OP_CREATE_UNMANAGED_SNAP,
667 				   pool, 0, (char *)snapid, sizeof(*snapid));
668 
669 }
670 EXPORT_SYMBOL(ceph_monc_create_snapid);
671 
672 int ceph_monc_delete_snapid(struct ceph_mon_client *monc,
673 			    u32 pool, u64 snapid)
674 {
675 	return ceph_monc_do_poolop(monc,  POOL_OP_CREATE_UNMANAGED_SNAP,
676 				   pool, snapid, 0, 0);
677 
678 }
679 
680 /*
681  * Resend pending generic requests.
682  */
683 static void __resend_generic_request(struct ceph_mon_client *monc)
684 {
685 	struct ceph_mon_generic_request *req;
686 	struct rb_node *p;
687 
688 	for (p = rb_first(&monc->generic_request_tree); p; p = rb_next(p)) {
689 		req = rb_entry(p, struct ceph_mon_generic_request, node);
690 		ceph_msg_revoke(req->request);
691 		ceph_msg_revoke_incoming(req->reply);
692 		ceph_con_send(&monc->con, ceph_msg_get(req->request));
693 	}
694 }
695 
696 /*
697  * Delayed work.  If we haven't mounted yet, retry.  Otherwise,
698  * renew/retry subscription as needed (in case it is timing out, or we
699  * got an ENOMEM).  And keep the monitor connection alive.
700  */
701 static void delayed_work(struct work_struct *work)
702 {
703 	struct ceph_mon_client *monc =
704 		container_of(work, struct ceph_mon_client, delayed_work.work);
705 
706 	dout("monc delayed_work\n");
707 	mutex_lock(&monc->mutex);
708 	if (monc->hunting) {
709 		__close_session(monc);
710 		__open_session(monc);  /* continue hunting */
711 	} else {
712 		ceph_con_keepalive(&monc->con);
713 
714 		__validate_auth(monc);
715 
716 		if (monc->auth->ops->is_authenticated(monc->auth))
717 			__send_subscribe(monc);
718 	}
719 	__schedule_delayed(monc);
720 	mutex_unlock(&monc->mutex);
721 }
722 
723 /*
724  * On startup, we build a temporary monmap populated with the IPs
725  * provided by mount(2).
726  */
727 static int build_initial_monmap(struct ceph_mon_client *monc)
728 {
729 	struct ceph_options *opt = monc->client->options;
730 	struct ceph_entity_addr *mon_addr = opt->mon_addr;
731 	int num_mon = opt->num_mon;
732 	int i;
733 
734 	/* build initial monmap */
735 	monc->monmap = kzalloc(sizeof(*monc->monmap) +
736 			       num_mon*sizeof(monc->monmap->mon_inst[0]),
737 			       GFP_KERNEL);
738 	if (!monc->monmap)
739 		return -ENOMEM;
740 	for (i = 0; i < num_mon; i++) {
741 		monc->monmap->mon_inst[i].addr = mon_addr[i];
742 		monc->monmap->mon_inst[i].addr.nonce = 0;
743 		monc->monmap->mon_inst[i].name.type =
744 			CEPH_ENTITY_TYPE_MON;
745 		monc->monmap->mon_inst[i].name.num = cpu_to_le64(i);
746 	}
747 	monc->monmap->num_mon = num_mon;
748 	monc->have_fsid = false;
749 	return 0;
750 }
751 
752 int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl)
753 {
754 	int err = 0;
755 
756 	dout("init\n");
757 	memset(monc, 0, sizeof(*monc));
758 	monc->client = cl;
759 	monc->monmap = NULL;
760 	mutex_init(&monc->mutex);
761 
762 	err = build_initial_monmap(monc);
763 	if (err)
764 		goto out;
765 
766 	/* connection */
767 	/* authentication */
768 	monc->auth = ceph_auth_init(cl->options->name,
769 				    cl->options->key);
770 	if (IS_ERR(monc->auth)) {
771 		err = PTR_ERR(monc->auth);
772 		goto out_monmap;
773 	}
774 	monc->auth->want_keys =
775 		CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON |
776 		CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS;
777 
778 	/* msgs */
779 	err = -ENOMEM;
780 	monc->m_subscribe_ack = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE_ACK,
781 				     sizeof(struct ceph_mon_subscribe_ack),
782 				     GFP_NOFS, true);
783 	if (!monc->m_subscribe_ack)
784 		goto out_auth;
785 
786 	monc->m_subscribe = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 96, GFP_NOFS,
787 					 true);
788 	if (!monc->m_subscribe)
789 		goto out_subscribe_ack;
790 
791 	monc->m_auth_reply = ceph_msg_new(CEPH_MSG_AUTH_REPLY, 4096, GFP_NOFS,
792 					  true);
793 	if (!monc->m_auth_reply)
794 		goto out_subscribe;
795 
796 	monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, GFP_NOFS, true);
797 	monc->pending_auth = 0;
798 	if (!monc->m_auth)
799 		goto out_auth_reply;
800 
801 	ceph_con_init(&monc->con, monc, &mon_con_ops,
802 		      &monc->client->msgr);
803 
804 	monc->cur_mon = -1;
805 	monc->hunting = true;
806 	monc->sub_renew_after = jiffies;
807 	monc->sub_sent = 0;
808 
809 	INIT_DELAYED_WORK(&monc->delayed_work, delayed_work);
810 	monc->generic_request_tree = RB_ROOT;
811 	monc->num_generic_requests = 0;
812 	monc->last_tid = 0;
813 
814 	monc->have_mdsmap = 0;
815 	monc->have_osdmap = 0;
816 	monc->want_next_osdmap = 1;
817 	return 0;
818 
819 out_auth_reply:
820 	ceph_msg_put(monc->m_auth_reply);
821 out_subscribe:
822 	ceph_msg_put(monc->m_subscribe);
823 out_subscribe_ack:
824 	ceph_msg_put(monc->m_subscribe_ack);
825 out_auth:
826 	ceph_auth_destroy(monc->auth);
827 out_monmap:
828 	kfree(monc->monmap);
829 out:
830 	return err;
831 }
832 EXPORT_SYMBOL(ceph_monc_init);
833 
834 void ceph_monc_stop(struct ceph_mon_client *monc)
835 {
836 	dout("stop\n");
837 	cancel_delayed_work_sync(&monc->delayed_work);
838 
839 	mutex_lock(&monc->mutex);
840 	__close_session(monc);
841 
842 	mutex_unlock(&monc->mutex);
843 
844 	/*
845 	 * flush msgr queue before we destroy ourselves to ensure that:
846 	 *  - any work that references our embedded con is finished.
847 	 *  - any osd_client or other work that may reference an authorizer
848 	 *    finishes before we shut down the auth subsystem.
849 	 */
850 	ceph_msgr_flush();
851 
852 	ceph_auth_destroy(monc->auth);
853 
854 	ceph_msg_put(monc->m_auth);
855 	ceph_msg_put(monc->m_auth_reply);
856 	ceph_msg_put(monc->m_subscribe);
857 	ceph_msg_put(monc->m_subscribe_ack);
858 
859 	kfree(monc->monmap);
860 }
861 EXPORT_SYMBOL(ceph_monc_stop);
862 
863 static void handle_auth_reply(struct ceph_mon_client *monc,
864 			      struct ceph_msg *msg)
865 {
866 	int ret;
867 	int was_auth = 0;
868 
869 	mutex_lock(&monc->mutex);
870 	if (monc->auth->ops)
871 		was_auth = monc->auth->ops->is_authenticated(monc->auth);
872 	monc->pending_auth = 0;
873 	ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base,
874 				     msg->front.iov_len,
875 				     monc->m_auth->front.iov_base,
876 				     monc->m_auth->front_max);
877 	if (ret < 0) {
878 		monc->client->auth_err = ret;
879 		wake_up_all(&monc->client->auth_wq);
880 	} else if (ret > 0) {
881 		__send_prepared_auth_request(monc, ret);
882 	} else if (!was_auth && monc->auth->ops->is_authenticated(monc->auth)) {
883 		dout("authenticated, starting session\n");
884 
885 		monc->client->msgr.inst.name.type = CEPH_ENTITY_TYPE_CLIENT;
886 		monc->client->msgr.inst.name.num =
887 					cpu_to_le64(monc->auth->global_id);
888 
889 		__send_subscribe(monc);
890 		__resend_generic_request(monc);
891 	}
892 	mutex_unlock(&monc->mutex);
893 }
894 
895 static int __validate_auth(struct ceph_mon_client *monc)
896 {
897 	int ret;
898 
899 	if (monc->pending_auth)
900 		return 0;
901 
902 	ret = ceph_build_auth(monc->auth, monc->m_auth->front.iov_base,
903 			      monc->m_auth->front_max);
904 	if (ret <= 0)
905 		return ret; /* either an error, or no need to authenticate */
906 	__send_prepared_auth_request(monc, ret);
907 	return 0;
908 }
909 
910 int ceph_monc_validate_auth(struct ceph_mon_client *monc)
911 {
912 	int ret;
913 
914 	mutex_lock(&monc->mutex);
915 	ret = __validate_auth(monc);
916 	mutex_unlock(&monc->mutex);
917 	return ret;
918 }
919 EXPORT_SYMBOL(ceph_monc_validate_auth);
920 
921 /*
922  * handle incoming message
923  */
924 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
925 {
926 	struct ceph_mon_client *monc = con->private;
927 	int type = le16_to_cpu(msg->hdr.type);
928 
929 	if (!monc)
930 		return;
931 
932 	switch (type) {
933 	case CEPH_MSG_AUTH_REPLY:
934 		handle_auth_reply(monc, msg);
935 		break;
936 
937 	case CEPH_MSG_MON_SUBSCRIBE_ACK:
938 		handle_subscribe_ack(monc, msg);
939 		break;
940 
941 	case CEPH_MSG_STATFS_REPLY:
942 		handle_statfs_reply(monc, msg);
943 		break;
944 
945 	case CEPH_MSG_POOLOP_REPLY:
946 		handle_poolop_reply(monc, msg);
947 		break;
948 
949 	case CEPH_MSG_MON_MAP:
950 		ceph_monc_handle_map(monc, msg);
951 		break;
952 
953 	case CEPH_MSG_OSD_MAP:
954 		ceph_osdc_handle_map(&monc->client->osdc, msg);
955 		break;
956 
957 	default:
958 		/* can the chained handler handle it? */
959 		if (monc->client->extra_mon_dispatch &&
960 		    monc->client->extra_mon_dispatch(monc->client, msg) == 0)
961 			break;
962 
963 		pr_err("received unknown message type %d %s\n", type,
964 		       ceph_msg_type_name(type));
965 	}
966 	ceph_msg_put(msg);
967 }
968 
969 /*
970  * Allocate memory for incoming message
971  */
972 static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con,
973 				      struct ceph_msg_header *hdr,
974 				      int *skip)
975 {
976 	struct ceph_mon_client *monc = con->private;
977 	int type = le16_to_cpu(hdr->type);
978 	int front_len = le32_to_cpu(hdr->front_len);
979 	struct ceph_msg *m = NULL;
980 
981 	*skip = 0;
982 
983 	switch (type) {
984 	case CEPH_MSG_MON_SUBSCRIBE_ACK:
985 		m = ceph_msg_get(monc->m_subscribe_ack);
986 		break;
987 	case CEPH_MSG_POOLOP_REPLY:
988 	case CEPH_MSG_STATFS_REPLY:
989 		return get_generic_reply(con, hdr, skip);
990 	case CEPH_MSG_AUTH_REPLY:
991 		m = ceph_msg_get(monc->m_auth_reply);
992 		break;
993 	case CEPH_MSG_MON_MAP:
994 	case CEPH_MSG_MDS_MAP:
995 	case CEPH_MSG_OSD_MAP:
996 		m = ceph_msg_new(type, front_len, GFP_NOFS, false);
997 		if (!m)
998 			return NULL;	/* ENOMEM--return skip == 0 */
999 		break;
1000 	}
1001 
1002 	if (!m) {
1003 		pr_info("alloc_msg unknown type %d\n", type);
1004 		*skip = 1;
1005 	}
1006 	return m;
1007 }
1008 
1009 /*
1010  * If the monitor connection resets, pick a new monitor and resubmit
1011  * any pending requests.
1012  */
1013 static void mon_fault(struct ceph_connection *con)
1014 {
1015 	struct ceph_mon_client *monc = con->private;
1016 
1017 	if (!monc)
1018 		return;
1019 
1020 	dout("mon_fault\n");
1021 	mutex_lock(&monc->mutex);
1022 	if (!con->private)
1023 		goto out;
1024 
1025 	if (!monc->hunting)
1026 		pr_info("mon%d %s session lost, "
1027 			"hunting for new mon\n", monc->cur_mon,
1028 			ceph_pr_addr(&monc->con.peer_addr.in_addr));
1029 
1030 	__close_session(monc);
1031 	if (!monc->hunting) {
1032 		/* start hunting */
1033 		monc->hunting = true;
1034 		__open_session(monc);
1035 	} else {
1036 		/* already hunting, let's wait a bit */
1037 		__schedule_delayed(monc);
1038 	}
1039 out:
1040 	mutex_unlock(&monc->mutex);
1041 }
1042 
1043 /*
1044  * We can ignore refcounting on the connection struct, as all references
1045  * will come from the messenger workqueue, which is drained prior to
1046  * mon_client destruction.
1047  */
1048 static struct ceph_connection *con_get(struct ceph_connection *con)
1049 {
1050 	return con;
1051 }
1052 
1053 static void con_put(struct ceph_connection *con)
1054 {
1055 }
1056 
1057 static const struct ceph_connection_operations mon_con_ops = {
1058 	.get = con_get,
1059 	.put = con_put,
1060 	.dispatch = dispatch,
1061 	.fault = mon_fault,
1062 	.alloc_msg = mon_alloc_msg,
1063 };
1064