mon_client.c revision 735a72ef952d42a256f79ae3e6dc1c17a45c041b
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
33static const struct ceph_connection_operations mon_con_ops;
34
35static int __validate_auth(struct ceph_mon_client *monc);
36
37/*
38 * Decode a monmap blob (e.g., during mount).
39 */
40struct 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
82bad:
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 */
91int 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 */
104static 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 */
117static 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_con_close(&monc->con);
122	monc->cur_mon = -1;
123	monc->pending_auth = 0;
124	ceph_auth_reset(monc->auth);
125}
126
127/*
128 * Open a session with a (new) monitor.
129 */
130static int __open_session(struct ceph_mon_client *monc)
131{
132	char r;
133	int ret;
134
135	if (monc->cur_mon < 0) {
136		get_random_bytes(&r, 1);
137		monc->cur_mon = r % monc->monmap->num_mon;
138		dout("open_session num=%d r=%d -> mon%d\n",
139		     monc->monmap->num_mon, r, monc->cur_mon);
140		monc->sub_sent = 0;
141		monc->sub_renew_after = jiffies;  /* i.e., expired */
142		monc->want_next_osdmap = !!monc->want_next_osdmap;
143
144		dout("open_session mon%d opening\n", monc->cur_mon);
145		ceph_con_open(&monc->con,
146			      CEPH_ENTITY_TYPE_MON, monc->cur_mon,
147			      &monc->monmap->mon_inst[monc->cur_mon].addr);
148
149		/* initiatiate authentication handshake */
150		ret = ceph_auth_build_hello(monc->auth,
151					    monc->m_auth->front.iov_base,
152					    monc->m_auth->front_max);
153		__send_prepared_auth_request(monc, ret);
154	} else {
155		dout("open_session mon%d already open\n", monc->cur_mon);
156	}
157	return 0;
158}
159
160static bool __sub_expired(struct ceph_mon_client *monc)
161{
162	return time_after_eq(jiffies, monc->sub_renew_after);
163}
164
165/*
166 * Reschedule delayed work timer.
167 */
168static void __schedule_delayed(struct ceph_mon_client *monc)
169{
170	unsigned int delay;
171
172	if (monc->cur_mon < 0 || __sub_expired(monc))
173		delay = 10 * HZ;
174	else
175		delay = 20 * HZ;
176	dout("__schedule_delayed after %u\n", delay);
177	schedule_delayed_work(&monc->delayed_work, delay);
178}
179
180/*
181 * Send subscribe request for mdsmap and/or osdmap.
182 */
183static void __send_subscribe(struct ceph_mon_client *monc)
184{
185	dout("__send_subscribe sub_sent=%u exp=%u want_osd=%d\n",
186	     (unsigned int)monc->sub_sent, __sub_expired(monc),
187	     monc->want_next_osdmap);
188	if ((__sub_expired(monc) && !monc->sub_sent) ||
189	    monc->want_next_osdmap == 1) {
190		struct ceph_msg *msg = monc->m_subscribe;
191		struct ceph_mon_subscribe_item *i;
192		void *p, *end;
193		int num;
194
195		p = msg->front.iov_base;
196		end = p + msg->front_max;
197
198		num = 1 + !!monc->want_next_osdmap + !!monc->want_mdsmap;
199		ceph_encode_32(&p, num);
200
201		if (monc->want_next_osdmap) {
202			dout("__send_subscribe to 'osdmap' %u\n",
203			     (unsigned int)monc->have_osdmap);
204			ceph_encode_string(&p, end, "osdmap", 6);
205			i = p;
206			i->have = cpu_to_le64(monc->have_osdmap);
207			i->onetime = 1;
208			p += sizeof(*i);
209			monc->want_next_osdmap = 2;  /* requested */
210		}
211		if (monc->want_mdsmap) {
212			dout("__send_subscribe to 'mdsmap' %u+\n",
213			     (unsigned int)monc->have_mdsmap);
214			ceph_encode_string(&p, end, "mdsmap", 6);
215			i = p;
216			i->have = cpu_to_le64(monc->have_mdsmap);
217			i->onetime = 0;
218			p += sizeof(*i);
219		}
220		ceph_encode_string(&p, end, "monmap", 6);
221		i = p;
222		i->have = 0;
223		i->onetime = 0;
224		p += sizeof(*i);
225
226		msg->front.iov_len = p - msg->front.iov_base;
227		msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
228		ceph_msg_revoke(msg);
229		ceph_con_send(&monc->con, ceph_msg_get(msg));
230
231		monc->sub_sent = jiffies | 1;  /* never 0 */
232	}
233}
234
235static void handle_subscribe_ack(struct ceph_mon_client *monc,
236				 struct ceph_msg *msg)
237{
238	unsigned int seconds;
239	struct ceph_mon_subscribe_ack *h = msg->front.iov_base;
240
241	if (msg->front.iov_len < sizeof(*h))
242		goto bad;
243	seconds = le32_to_cpu(h->duration);
244
245	mutex_lock(&monc->mutex);
246	if (monc->hunting) {
247		pr_info("mon%d %s session established\n",
248			monc->cur_mon,
249			ceph_pr_addr(&monc->con.peer_addr.in_addr));
250		monc->hunting = false;
251	}
252	dout("handle_subscribe_ack after %d seconds\n", seconds);
253	monc->sub_renew_after = monc->sub_sent + (seconds >> 1)*HZ - 1;
254	monc->sub_sent = 0;
255	mutex_unlock(&monc->mutex);
256	return;
257bad:
258	pr_err("got corrupt subscribe-ack msg\n");
259	ceph_msg_dump(msg);
260}
261
262/*
263 * Keep track of which maps we have
264 */
265int ceph_monc_got_mdsmap(struct ceph_mon_client *monc, u32 got)
266{
267	mutex_lock(&monc->mutex);
268	monc->have_mdsmap = got;
269	mutex_unlock(&monc->mutex);
270	return 0;
271}
272EXPORT_SYMBOL(ceph_monc_got_mdsmap);
273
274int ceph_monc_got_osdmap(struct ceph_mon_client *monc, u32 got)
275{
276	mutex_lock(&monc->mutex);
277	monc->have_osdmap = got;
278	monc->want_next_osdmap = 0;
279	mutex_unlock(&monc->mutex);
280	return 0;
281}
282
283/*
284 * Register interest in the next osdmap
285 */
286void ceph_monc_request_next_osdmap(struct ceph_mon_client *monc)
287{
288	dout("request_next_osdmap have %u\n", monc->have_osdmap);
289	mutex_lock(&monc->mutex);
290	if (!monc->want_next_osdmap)
291		monc->want_next_osdmap = 1;
292	if (monc->want_next_osdmap < 2)
293		__send_subscribe(monc);
294	mutex_unlock(&monc->mutex);
295}
296
297/*
298 *
299 */
300int ceph_monc_open_session(struct ceph_mon_client *monc)
301{
302	mutex_lock(&monc->mutex);
303	__open_session(monc);
304	__schedule_delayed(monc);
305	mutex_unlock(&monc->mutex);
306	return 0;
307}
308EXPORT_SYMBOL(ceph_monc_open_session);
309
310/*
311 * The monitor responds with mount ack indicate mount success.  The
312 * included client ticket allows the client to talk to MDSs and OSDs.
313 */
314static void ceph_monc_handle_map(struct ceph_mon_client *monc,
315				 struct ceph_msg *msg)
316{
317	struct ceph_client *client = monc->client;
318	struct ceph_monmap *monmap = NULL, *old = monc->monmap;
319	void *p, *end;
320
321	mutex_lock(&monc->mutex);
322
323	dout("handle_monmap\n");
324	p = msg->front.iov_base;
325	end = p + msg->front.iov_len;
326
327	monmap = ceph_monmap_decode(p, end);
328	if (IS_ERR(monmap)) {
329		pr_err("problem decoding monmap, %d\n",
330		       (int)PTR_ERR(monmap));
331		goto out;
332	}
333
334	if (ceph_check_fsid(monc->client, &monmap->fsid) < 0) {
335		kfree(monmap);
336		goto out;
337	}
338
339	client->monc.monmap = monmap;
340	kfree(old);
341
342	if (!client->have_fsid) {
343		client->have_fsid = true;
344		mutex_unlock(&monc->mutex);
345		/*
346		 * do debugfs initialization without mutex to avoid
347		 * creating a locking dependency
348		 */
349		ceph_debugfs_client_init(client);
350		goto out_unlocked;
351	}
352out:
353	mutex_unlock(&monc->mutex);
354out_unlocked:
355	wake_up_all(&client->auth_wq);
356}
357
358/*
359 * generic requests (e.g., statfs, poolop)
360 */
361static struct ceph_mon_generic_request *__lookup_generic_req(
362	struct ceph_mon_client *monc, u64 tid)
363{
364	struct ceph_mon_generic_request *req;
365	struct rb_node *n = monc->generic_request_tree.rb_node;
366
367	while (n) {
368		req = rb_entry(n, struct ceph_mon_generic_request, node);
369		if (tid < req->tid)
370			n = n->rb_left;
371		else if (tid > req->tid)
372			n = n->rb_right;
373		else
374			return req;
375	}
376	return NULL;
377}
378
379static void __insert_generic_request(struct ceph_mon_client *monc,
380			    struct ceph_mon_generic_request *new)
381{
382	struct rb_node **p = &monc->generic_request_tree.rb_node;
383	struct rb_node *parent = NULL;
384	struct ceph_mon_generic_request *req = NULL;
385
386	while (*p) {
387		parent = *p;
388		req = rb_entry(parent, struct ceph_mon_generic_request, node);
389		if (new->tid < req->tid)
390			p = &(*p)->rb_left;
391		else if (new->tid > req->tid)
392			p = &(*p)->rb_right;
393		else
394			BUG();
395	}
396
397	rb_link_node(&new->node, parent, p);
398	rb_insert_color(&new->node, &monc->generic_request_tree);
399}
400
401static void release_generic_request(struct kref *kref)
402{
403	struct ceph_mon_generic_request *req =
404		container_of(kref, struct ceph_mon_generic_request, kref);
405
406	if (req->reply)
407		ceph_msg_put(req->reply);
408	if (req->request)
409		ceph_msg_put(req->request);
410
411	kfree(req);
412}
413
414static void put_generic_request(struct ceph_mon_generic_request *req)
415{
416	kref_put(&req->kref, release_generic_request);
417}
418
419static void get_generic_request(struct ceph_mon_generic_request *req)
420{
421	kref_get(&req->kref);
422}
423
424static struct ceph_msg *get_generic_reply(struct ceph_connection *con,
425					 struct ceph_msg_header *hdr,
426					 int *skip)
427{
428	struct ceph_mon_client *monc = con->private;
429	struct ceph_mon_generic_request *req;
430	u64 tid = le64_to_cpu(hdr->tid);
431	struct ceph_msg *m;
432
433	mutex_lock(&monc->mutex);
434	req = __lookup_generic_req(monc, tid);
435	if (!req) {
436		dout("get_generic_reply %lld dne\n", tid);
437		*skip = 1;
438		m = NULL;
439	} else {
440		dout("get_generic_reply %lld got %p\n", tid, req->reply);
441		*skip = 0;
442		m = ceph_msg_get(req->reply);
443		/*
444		 * we don't need to track the connection reading into
445		 * this reply because we only have one open connection
446		 * at a time, ever.
447		 */
448	}
449	mutex_unlock(&monc->mutex);
450	return m;
451}
452
453static int do_generic_request(struct ceph_mon_client *monc,
454			      struct ceph_mon_generic_request *req)
455{
456	int err;
457
458	/* register request */
459	mutex_lock(&monc->mutex);
460	req->tid = ++monc->last_tid;
461	req->request->hdr.tid = cpu_to_le64(req->tid);
462	__insert_generic_request(monc, req);
463	monc->num_generic_requests++;
464	ceph_con_send(&monc->con, ceph_msg_get(req->request));
465	mutex_unlock(&monc->mutex);
466
467	err = wait_for_completion_interruptible(&req->completion);
468
469	mutex_lock(&monc->mutex);
470	rb_erase(&req->node, &monc->generic_request_tree);
471	monc->num_generic_requests--;
472	mutex_unlock(&monc->mutex);
473
474	if (!err)
475		err = req->result;
476	return err;
477}
478
479/*
480 * statfs
481 */
482static void handle_statfs_reply(struct ceph_mon_client *monc,
483				struct ceph_msg *msg)
484{
485	struct ceph_mon_generic_request *req;
486	struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
487	u64 tid = le64_to_cpu(msg->hdr.tid);
488
489	if (msg->front.iov_len != sizeof(*reply))
490		goto bad;
491	dout("handle_statfs_reply %p tid %llu\n", msg, tid);
492
493	mutex_lock(&monc->mutex);
494	req = __lookup_generic_req(monc, tid);
495	if (req) {
496		*(struct ceph_statfs *)req->buf = reply->st;
497		req->result = 0;
498		get_generic_request(req);
499	}
500	mutex_unlock(&monc->mutex);
501	if (req) {
502		complete_all(&req->completion);
503		put_generic_request(req);
504	}
505	return;
506
507bad:
508	pr_err("corrupt generic reply, tid %llu\n", tid);
509	ceph_msg_dump(msg);
510}
511
512/*
513 * Do a synchronous statfs().
514 */
515int ceph_monc_do_statfs(struct ceph_mon_client *monc, struct ceph_statfs *buf)
516{
517	struct ceph_mon_generic_request *req;
518	struct ceph_mon_statfs *h;
519	int err;
520
521	req = kzalloc(sizeof(*req), GFP_NOFS);
522	if (!req)
523		return -ENOMEM;
524
525	kref_init(&req->kref);
526	req->buf = buf;
527	req->buf_len = sizeof(*buf);
528	init_completion(&req->completion);
529
530	err = -ENOMEM;
531	req->request = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), GFP_NOFS,
532				    true);
533	if (!req->request)
534		goto out;
535	req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 1024, GFP_NOFS,
536				  true);
537	if (!req->reply)
538		goto out;
539
540	/* fill out request */
541	h = req->request->front.iov_base;
542	h->monhdr.have_version = 0;
543	h->monhdr.session_mon = cpu_to_le16(-1);
544	h->monhdr.session_mon_tid = 0;
545	h->fsid = monc->monmap->fsid;
546
547	err = do_generic_request(monc, req);
548
549out:
550	kref_put(&req->kref, release_generic_request);
551	return err;
552}
553EXPORT_SYMBOL(ceph_monc_do_statfs);
554
555/*
556 * pool ops
557 */
558static int get_poolop_reply_buf(const char *src, size_t src_len,
559				char *dst, size_t dst_len)
560{
561	u32 buf_len;
562
563	if (src_len != sizeof(u32) + dst_len)
564		return -EINVAL;
565
566	buf_len = le32_to_cpu(*(u32 *)src);
567	if (buf_len != dst_len)
568		return -EINVAL;
569
570	memcpy(dst, src + sizeof(u32), dst_len);
571	return 0;
572}
573
574static void handle_poolop_reply(struct ceph_mon_client *monc,
575				struct ceph_msg *msg)
576{
577	struct ceph_mon_generic_request *req;
578	struct ceph_mon_poolop_reply *reply = msg->front.iov_base;
579	u64 tid = le64_to_cpu(msg->hdr.tid);
580
581	if (msg->front.iov_len < sizeof(*reply))
582		goto bad;
583	dout("handle_poolop_reply %p tid %llu\n", msg, tid);
584
585	mutex_lock(&monc->mutex);
586	req = __lookup_generic_req(monc, tid);
587	if (req) {
588		if (req->buf_len &&
589		    get_poolop_reply_buf(msg->front.iov_base + sizeof(*reply),
590				     msg->front.iov_len - sizeof(*reply),
591				     req->buf, req->buf_len) < 0) {
592			mutex_unlock(&monc->mutex);
593			goto bad;
594		}
595		req->result = le32_to_cpu(reply->reply_code);
596		get_generic_request(req);
597	}
598	mutex_unlock(&monc->mutex);
599	if (req) {
600		complete(&req->completion);
601		put_generic_request(req);
602	}
603	return;
604
605bad:
606	pr_err("corrupt generic reply, tid %llu\n", tid);
607	ceph_msg_dump(msg);
608}
609
610/*
611 * Do a synchronous pool op.
612 */
613int ceph_monc_do_poolop(struct ceph_mon_client *monc, u32 op,
614			u32 pool, u64 snapid,
615			char *buf, int len)
616{
617	struct ceph_mon_generic_request *req;
618	struct ceph_mon_poolop *h;
619	int err;
620
621	req = kzalloc(sizeof(*req), GFP_NOFS);
622	if (!req)
623		return -ENOMEM;
624
625	kref_init(&req->kref);
626	req->buf = buf;
627	req->buf_len = len;
628	init_completion(&req->completion);
629
630	err = -ENOMEM;
631	req->request = ceph_msg_new(CEPH_MSG_POOLOP, sizeof(*h), GFP_NOFS,
632				    true);
633	if (!req->request)
634		goto out;
635	req->reply = ceph_msg_new(CEPH_MSG_POOLOP_REPLY, 1024, GFP_NOFS,
636				  true);
637	if (!req->reply)
638		goto out;
639
640	/* fill out request */
641	req->request->hdr.version = cpu_to_le16(2);
642	h = req->request->front.iov_base;
643	h->monhdr.have_version = 0;
644	h->monhdr.session_mon = cpu_to_le16(-1);
645	h->monhdr.session_mon_tid = 0;
646	h->fsid = monc->monmap->fsid;
647	h->pool = cpu_to_le32(pool);
648	h->op = cpu_to_le32(op);
649	h->auid = 0;
650	h->snapid = cpu_to_le64(snapid);
651	h->name_len = 0;
652
653	err = do_generic_request(monc, req);
654
655out:
656	kref_put(&req->kref, release_generic_request);
657	return err;
658}
659
660int ceph_monc_create_snapid(struct ceph_mon_client *monc,
661			    u32 pool, u64 *snapid)
662{
663	return ceph_monc_do_poolop(monc,  POOL_OP_CREATE_UNMANAGED_SNAP,
664				   pool, 0, (char *)snapid, sizeof(*snapid));
665
666}
667EXPORT_SYMBOL(ceph_monc_create_snapid);
668
669int ceph_monc_delete_snapid(struct ceph_mon_client *monc,
670			    u32 pool, u64 snapid)
671{
672	return ceph_monc_do_poolop(monc,  POOL_OP_CREATE_UNMANAGED_SNAP,
673				   pool, snapid, 0, 0);
674
675}
676
677/*
678 * Resend pending generic requests.
679 */
680static void __resend_generic_request(struct ceph_mon_client *monc)
681{
682	struct ceph_mon_generic_request *req;
683	struct rb_node *p;
684
685	for (p = rb_first(&monc->generic_request_tree); p; p = rb_next(p)) {
686		req = rb_entry(p, struct ceph_mon_generic_request, node);
687		ceph_msg_revoke(req->request);
688		ceph_con_send(&monc->con, ceph_msg_get(req->request));
689	}
690}
691
692/*
693 * Delayed work.  If we haven't mounted yet, retry.  Otherwise,
694 * renew/retry subscription as needed (in case it is timing out, or we
695 * got an ENOMEM).  And keep the monitor connection alive.
696 */
697static void delayed_work(struct work_struct *work)
698{
699	struct ceph_mon_client *monc =
700		container_of(work, struct ceph_mon_client, delayed_work.work);
701
702	dout("monc delayed_work\n");
703	mutex_lock(&monc->mutex);
704	if (monc->hunting) {
705		__close_session(monc);
706		__open_session(monc);  /* continue hunting */
707	} else {
708		ceph_con_keepalive(&monc->con);
709
710		__validate_auth(monc);
711
712		if (monc->auth->ops->is_authenticated(monc->auth))
713			__send_subscribe(monc);
714	}
715	__schedule_delayed(monc);
716	mutex_unlock(&monc->mutex);
717}
718
719/*
720 * On startup, we build a temporary monmap populated with the IPs
721 * provided by mount(2).
722 */
723static int build_initial_monmap(struct ceph_mon_client *monc)
724{
725	struct ceph_options *opt = monc->client->options;
726	struct ceph_entity_addr *mon_addr = opt->mon_addr;
727	int num_mon = opt->num_mon;
728	int i;
729
730	/* build initial monmap */
731	monc->monmap = kzalloc(sizeof(*monc->monmap) +
732			       num_mon*sizeof(monc->monmap->mon_inst[0]),
733			       GFP_KERNEL);
734	if (!monc->monmap)
735		return -ENOMEM;
736	for (i = 0; i < num_mon; i++) {
737		monc->monmap->mon_inst[i].addr = mon_addr[i];
738		monc->monmap->mon_inst[i].addr.nonce = 0;
739		monc->monmap->mon_inst[i].name.type =
740			CEPH_ENTITY_TYPE_MON;
741		monc->monmap->mon_inst[i].name.num = cpu_to_le64(i);
742	}
743	monc->monmap->num_mon = num_mon;
744	monc->have_fsid = false;
745	return 0;
746}
747
748int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl)
749{
750	int err = 0;
751
752	dout("init\n");
753	memset(monc, 0, sizeof(*monc));
754	monc->client = cl;
755	monc->monmap = NULL;
756	mutex_init(&monc->mutex);
757
758	err = build_initial_monmap(monc);
759	if (err)
760		goto out;
761
762	/* connection */
763	/* authentication */
764	monc->auth = ceph_auth_init(cl->options->name,
765				    cl->options->key);
766	if (IS_ERR(monc->auth)) {
767		err = PTR_ERR(monc->auth);
768		goto out_monmap;
769	}
770	monc->auth->want_keys =
771		CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON |
772		CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS;
773
774	/* msgs */
775	err = -ENOMEM;
776	monc->m_subscribe_ack = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE_ACK,
777				     sizeof(struct ceph_mon_subscribe_ack),
778				     GFP_NOFS, true);
779	if (!monc->m_subscribe_ack)
780		goto out_auth;
781
782	monc->m_subscribe = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 96, GFP_NOFS,
783					 true);
784	if (!monc->m_subscribe)
785		goto out_subscribe_ack;
786
787	monc->m_auth_reply = ceph_msg_new(CEPH_MSG_AUTH_REPLY, 4096, GFP_NOFS,
788					  true);
789	if (!monc->m_auth_reply)
790		goto out_subscribe;
791
792	monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, GFP_NOFS, true);
793	monc->pending_auth = 0;
794	if (!monc->m_auth)
795		goto out_auth_reply;
796
797	ceph_con_init(&monc->con, monc, &mon_con_ops,
798		      &monc->client->msgr);
799
800	monc->cur_mon = -1;
801	monc->hunting = true;
802	monc->sub_renew_after = jiffies;
803	monc->sub_sent = 0;
804
805	INIT_DELAYED_WORK(&monc->delayed_work, delayed_work);
806	monc->generic_request_tree = RB_ROOT;
807	monc->num_generic_requests = 0;
808	monc->last_tid = 0;
809
810	monc->have_mdsmap = 0;
811	monc->have_osdmap = 0;
812	monc->want_next_osdmap = 1;
813	return 0;
814
815out_auth_reply:
816	ceph_msg_put(monc->m_auth_reply);
817out_subscribe:
818	ceph_msg_put(monc->m_subscribe);
819out_subscribe_ack:
820	ceph_msg_put(monc->m_subscribe_ack);
821out_auth:
822	ceph_auth_destroy(monc->auth);
823out_monmap:
824	kfree(monc->monmap);
825out:
826	return err;
827}
828EXPORT_SYMBOL(ceph_monc_init);
829
830void ceph_monc_stop(struct ceph_mon_client *monc)
831{
832	dout("stop\n");
833	cancel_delayed_work_sync(&monc->delayed_work);
834
835	mutex_lock(&monc->mutex);
836	__close_session(monc);
837
838	mutex_unlock(&monc->mutex);
839
840	/*
841	 * flush msgr queue before we destroy ourselves to ensure that:
842	 *  - any work that references our embedded con is finished.
843	 *  - any osd_client or other work that may reference an authorizer
844	 *    finishes before we shut down the auth subsystem.
845	 */
846	ceph_msgr_flush();
847
848	ceph_auth_destroy(monc->auth);
849
850	ceph_msg_put(monc->m_auth);
851	ceph_msg_put(monc->m_auth_reply);
852	ceph_msg_put(monc->m_subscribe);
853	ceph_msg_put(monc->m_subscribe_ack);
854
855	kfree(monc->monmap);
856}
857EXPORT_SYMBOL(ceph_monc_stop);
858
859static void handle_auth_reply(struct ceph_mon_client *monc,
860			      struct ceph_msg *msg)
861{
862	int ret;
863	int was_auth = 0;
864
865	mutex_lock(&monc->mutex);
866	if (monc->auth->ops)
867		was_auth = monc->auth->ops->is_authenticated(monc->auth);
868	monc->pending_auth = 0;
869	ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base,
870				     msg->front.iov_len,
871				     monc->m_auth->front.iov_base,
872				     monc->m_auth->front_max);
873	if (ret < 0) {
874		monc->client->auth_err = ret;
875		wake_up_all(&monc->client->auth_wq);
876	} else if (ret > 0) {
877		__send_prepared_auth_request(monc, ret);
878	} else if (!was_auth && monc->auth->ops->is_authenticated(monc->auth)) {
879		dout("authenticated, starting session\n");
880
881		monc->client->msgr.inst.name.type = CEPH_ENTITY_TYPE_CLIENT;
882		monc->client->msgr.inst.name.num =
883					cpu_to_le64(monc->auth->global_id);
884
885		__send_subscribe(monc);
886		__resend_generic_request(monc);
887	}
888	mutex_unlock(&monc->mutex);
889}
890
891static int __validate_auth(struct ceph_mon_client *monc)
892{
893	int ret;
894
895	if (monc->pending_auth)
896		return 0;
897
898	ret = ceph_build_auth(monc->auth, monc->m_auth->front.iov_base,
899			      monc->m_auth->front_max);
900	if (ret <= 0)
901		return ret; /* either an error, or no need to authenticate */
902	__send_prepared_auth_request(monc, ret);
903	return 0;
904}
905
906int ceph_monc_validate_auth(struct ceph_mon_client *monc)
907{
908	int ret;
909
910	mutex_lock(&monc->mutex);
911	ret = __validate_auth(monc);
912	mutex_unlock(&monc->mutex);
913	return ret;
914}
915EXPORT_SYMBOL(ceph_monc_validate_auth);
916
917/*
918 * handle incoming message
919 */
920static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
921{
922	struct ceph_mon_client *monc = con->private;
923	int type = le16_to_cpu(msg->hdr.type);
924
925	if (!monc)
926		return;
927
928	switch (type) {
929	case CEPH_MSG_AUTH_REPLY:
930		handle_auth_reply(monc, msg);
931		break;
932
933	case CEPH_MSG_MON_SUBSCRIBE_ACK:
934		handle_subscribe_ack(monc, msg);
935		break;
936
937	case CEPH_MSG_STATFS_REPLY:
938		handle_statfs_reply(monc, msg);
939		break;
940
941	case CEPH_MSG_POOLOP_REPLY:
942		handle_poolop_reply(monc, msg);
943		break;
944
945	case CEPH_MSG_MON_MAP:
946		ceph_monc_handle_map(monc, msg);
947		break;
948
949	case CEPH_MSG_OSD_MAP:
950		ceph_osdc_handle_map(&monc->client->osdc, msg);
951		break;
952
953	default:
954		/* can the chained handler handle it? */
955		if (monc->client->extra_mon_dispatch &&
956		    monc->client->extra_mon_dispatch(monc->client, msg) == 0)
957			break;
958
959		pr_err("received unknown message type %d %s\n", type,
960		       ceph_msg_type_name(type));
961	}
962	ceph_msg_put(msg);
963}
964
965/*
966 * Allocate memory for incoming message
967 */
968static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con,
969				      struct ceph_msg_header *hdr,
970				      int *skip)
971{
972	struct ceph_mon_client *monc = con->private;
973	int type = le16_to_cpu(hdr->type);
974	int front_len = le32_to_cpu(hdr->front_len);
975	struct ceph_msg *m = NULL;
976
977	*skip = 0;
978
979	switch (type) {
980	case CEPH_MSG_MON_SUBSCRIBE_ACK:
981		m = ceph_msg_get(monc->m_subscribe_ack);
982		break;
983	case CEPH_MSG_POOLOP_REPLY:
984	case CEPH_MSG_STATFS_REPLY:
985		return get_generic_reply(con, hdr, skip);
986	case CEPH_MSG_AUTH_REPLY:
987		m = ceph_msg_get(monc->m_auth_reply);
988		break;
989	case CEPH_MSG_MON_MAP:
990	case CEPH_MSG_MDS_MAP:
991	case CEPH_MSG_OSD_MAP:
992		m = ceph_msg_new(type, front_len, GFP_NOFS, false);
993		if (!m)
994			return NULL;	/* ENOMEM--return skip == 0 */
995		break;
996	}
997
998	if (!m) {
999		pr_info("alloc_msg unknown type %d\n", type);
1000		*skip = 1;
1001	}
1002	return m;
1003}
1004
1005/*
1006 * If the monitor connection resets, pick a new monitor and resubmit
1007 * any pending requests.
1008 */
1009static void mon_fault(struct ceph_connection *con)
1010{
1011	struct ceph_mon_client *monc = con->private;
1012
1013	if (!monc)
1014		return;
1015
1016	dout("mon_fault\n");
1017	mutex_lock(&monc->mutex);
1018	if (!con->private)
1019		goto out;
1020
1021	if (!monc->hunting)
1022		pr_info("mon%d %s session lost, "
1023			"hunting for new mon\n", monc->cur_mon,
1024			ceph_pr_addr(&monc->con.peer_addr.in_addr));
1025
1026	__close_session(monc);
1027	if (!monc->hunting) {
1028		/* start hunting */
1029		monc->hunting = true;
1030		__open_session(monc);
1031	} else {
1032		/* already hunting, let's wait a bit */
1033		__schedule_delayed(monc);
1034	}
1035out:
1036	mutex_unlock(&monc->mutex);
1037}
1038
1039/*
1040 * We can ignore refcounting on the connection struct, as all references
1041 * will come from the messenger workqueue, which is drained prior to
1042 * mon_client destruction.
1043 */
1044static struct ceph_connection *con_get(struct ceph_connection *con)
1045{
1046	return con;
1047}
1048
1049static void con_put(struct ceph_connection *con)
1050{
1051}
1052
1053static const struct ceph_connection_operations mon_con_ops = {
1054	.get = con_get,
1055	.put = con_put,
1056	.dispatch = dispatch,
1057	.fault = mon_fault,
1058	.alloc_msg = mon_alloc_msg,
1059};
1060