1 #include "ceph_debug.h"
3 #include <linux/types.h>
4 #include <linux/random.h>
5 #include <linux/sched.h>
7 #include "mon_client.h"
13 * Interact with Ceph monitor cluster. Handle requests for new map
14 * versions, and periodically resend as needed. Also implement
15 * statfs() and umount().
17 * A small cluster of Ceph "monitors" are responsible for managing critical
18 * cluster configuration and state information. An odd number (e.g., 3, 5)
19 * of cmon daemons use a modified version of the Paxos part-time parliament
20 * algorithm to manage the MDS map (mds cluster membership), OSD map, and
21 * list of clients who have mounted the file system.
23 * We maintain an open, active session with a monitor at all times in order to
24 * receive timely MDSMap updates. We periodically send a keepalive byte on the
25 * TCP socket to ensure we detect a failure. If the connection does break, we
26 * randomly hunt for a new monitor. Once the connection is reestablished, we
27 * resend any outstanding requests.
30 const static struct ceph_connection_operations mon_con_ops;
32 static int __validate_auth(struct ceph_mon_client *monc);
35 * Decode a monmap blob (e.g., during mount).
37 struct ceph_monmap *ceph_monmap_decode(void *p, void *end)
39 struct ceph_monmap *m = NULL;
41 struct ceph_fsid fsid;
46 ceph_decode_32_safe(&p, end, len, bad);
47 ceph_decode_need(&p, end, len, bad);
49 dout("monmap_decode %p %p len %d\n", p, end, (int)(end-p));
51 ceph_decode_16_safe(&p, end, version, bad);
53 ceph_decode_need(&p, end, sizeof(fsid) + 2*sizeof(u32), bad);
54 ceph_decode_copy(&p, &fsid, sizeof(fsid));
55 epoch = ceph_decode_32(&p);
57 num_mon = ceph_decode_32(&p);
58 ceph_decode_need(&p, end, num_mon*sizeof(m->mon_inst[0]), bad);
60 if (num_mon >= CEPH_MAX_MON)
62 m = kmalloc(sizeof(*m) + sizeof(m->mon_inst[0])*num_mon, GFP_NOFS);
64 return ERR_PTR(-ENOMEM);
68 ceph_decode_copy(&p, m->mon_inst, num_mon*sizeof(m->mon_inst[0]));
69 for (i = 0; i < num_mon; i++)
70 ceph_decode_addr(&m->mon_inst[i].addr);
72 dout("monmap_decode epoch %d, num_mon %d\n", m->epoch,
74 for (i = 0; i < m->num_mon; i++)
75 dout("monmap_decode mon%d is %s\n", i,
76 pr_addr(&m->mon_inst[i].addr.in_addr));
80 dout("monmap_decode failed with %d\n", err);
86 * return true if *addr is included in the monmap.
88 int ceph_monmap_contains(struct ceph_monmap *m, struct ceph_entity_addr *addr)
92 for (i = 0; i < m->num_mon; i++)
93 if (memcmp(addr, &m->mon_inst[i].addr, sizeof(*addr)) == 0)
99 * Send an auth request.
101 static void __send_prepared_auth_request(struct ceph_mon_client *monc, int len)
103 monc->pending_auth = 1;
104 monc->m_auth->front.iov_len = len;
105 monc->m_auth->hdr.front_len = cpu_to_le32(len);
106 ceph_msg_get(monc->m_auth); /* keep our ref */
107 ceph_con_send(monc->con, monc->m_auth);
111 * Close monitor session, if any.
113 static void __close_session(struct ceph_mon_client *monc)
116 dout("__close_session closing mon%d\n", monc->cur_mon);
117 ceph_con_revoke(monc->con, monc->m_auth);
118 ceph_con_close(monc->con);
120 monc->pending_auth = 0;
121 ceph_auth_reset(monc->auth);
126 * Open a session with a (new) monitor.
128 static int __open_session(struct ceph_mon_client *monc)
133 if (monc->cur_mon < 0) {
134 get_random_bytes(&r, 1);
135 monc->cur_mon = r % monc->monmap->num_mon;
136 dout("open_session num=%d r=%d -> mon%d\n",
137 monc->monmap->num_mon, r, monc->cur_mon);
139 monc->sub_renew_after = jiffies; /* i.e., expired */
140 monc->want_next_osdmap = !!monc->want_next_osdmap;
142 dout("open_session mon%d opening\n", monc->cur_mon);
143 monc->con->peer_name.type = CEPH_ENTITY_TYPE_MON;
144 monc->con->peer_name.num = cpu_to_le64(monc->cur_mon);
145 ceph_con_open(monc->con,
146 &monc->monmap->mon_inst[monc->cur_mon].addr);
148 /* initiatiate authentication handshake */
149 ret = ceph_auth_build_hello(monc->auth,
150 monc->m_auth->front.iov_base,
151 monc->m_auth->front_max);
152 __send_prepared_auth_request(monc, ret);
154 dout("open_session mon%d already open\n", monc->cur_mon);
159 static bool __sub_expired(struct ceph_mon_client *monc)
161 return time_after_eq(jiffies, monc->sub_renew_after);
165 * Reschedule delayed work timer.
167 static void __schedule_delayed(struct ceph_mon_client *monc)
171 if (monc->cur_mon < 0 || __sub_expired(monc))
175 dout("__schedule_delayed after %u\n", delay);
176 schedule_delayed_work(&monc->delayed_work, delay);
180 * Send subscribe request for mdsmap and/or osdmap.
182 static void __send_subscribe(struct ceph_mon_client *monc)
184 dout("__send_subscribe sub_sent=%u exp=%u want_osd=%d\n",
185 (unsigned)monc->sub_sent, __sub_expired(monc),
186 monc->want_next_osdmap);
187 if ((__sub_expired(monc) && !monc->sub_sent) ||
188 monc->want_next_osdmap == 1) {
189 struct ceph_msg *msg;
190 struct ceph_mon_subscribe_item *i;
193 msg = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 96, 0, 0, NULL);
197 p = msg->front.iov_base;
198 end = p + msg->front.iov_len;
200 dout("__send_subscribe to 'mdsmap' %u+\n",
201 (unsigned)monc->have_mdsmap);
202 if (monc->want_next_osdmap) {
203 dout("__send_subscribe to 'osdmap' %u\n",
204 (unsigned)monc->have_osdmap);
205 ceph_encode_32(&p, 3);
206 ceph_encode_string(&p, end, "osdmap", 6);
208 i->have = cpu_to_le64(monc->have_osdmap);
211 monc->want_next_osdmap = 2; /* requested */
213 ceph_encode_32(&p, 2);
215 ceph_encode_string(&p, end, "mdsmap", 6);
217 i->have = cpu_to_le64(monc->have_mdsmap);
220 ceph_encode_string(&p, end, "monmap", 6);
226 msg->front.iov_len = p - msg->front.iov_base;
227 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
228 ceph_con_send(monc->con, msg);
230 monc->sub_sent = jiffies | 1; /* never 0 */
234 static void handle_subscribe_ack(struct ceph_mon_client *monc,
235 struct ceph_msg *msg)
238 struct ceph_mon_subscribe_ack *h = msg->front.iov_base;
240 if (msg->front.iov_len < sizeof(*h))
242 seconds = le32_to_cpu(h->duration);
244 mutex_lock(&monc->mutex);
246 pr_info("mon%d %s session established\n",
247 monc->cur_mon, pr_addr(&monc->con->peer_addr.in_addr));
248 monc->hunting = false;
250 dout("handle_subscribe_ack after %d seconds\n", seconds);
251 monc->sub_renew_after = monc->sub_sent + (seconds >> 1)*HZ - 1;
253 mutex_unlock(&monc->mutex);
256 pr_err("got corrupt subscribe-ack msg\n");
261 * Keep track of which maps we have
263 int ceph_monc_got_mdsmap(struct ceph_mon_client *monc, u32 got)
265 mutex_lock(&monc->mutex);
266 monc->have_mdsmap = got;
267 mutex_unlock(&monc->mutex);
271 int ceph_monc_got_osdmap(struct ceph_mon_client *monc, u32 got)
273 mutex_lock(&monc->mutex);
274 monc->have_osdmap = got;
275 monc->want_next_osdmap = 0;
276 mutex_unlock(&monc->mutex);
281 * Register interest in the next osdmap
283 void ceph_monc_request_next_osdmap(struct ceph_mon_client *monc)
285 dout("request_next_osdmap have %u\n", monc->have_osdmap);
286 mutex_lock(&monc->mutex);
287 if (!monc->want_next_osdmap)
288 monc->want_next_osdmap = 1;
289 if (monc->want_next_osdmap < 2)
290 __send_subscribe(monc);
291 mutex_unlock(&monc->mutex);
297 int ceph_monc_open_session(struct ceph_mon_client *monc)
300 monc->con = kmalloc(sizeof(*monc->con), GFP_KERNEL);
303 ceph_con_init(monc->client->msgr, monc->con);
304 monc->con->private = monc;
305 monc->con->ops = &mon_con_ops;
308 mutex_lock(&monc->mutex);
309 __open_session(monc);
310 __schedule_delayed(monc);
311 mutex_unlock(&monc->mutex);
316 * The monitor responds with mount ack indicate mount success. The
317 * included client ticket allows the client to talk to MDSs and OSDs.
319 static void ceph_monc_handle_map(struct ceph_mon_client *monc,
320 struct ceph_msg *msg)
322 struct ceph_client *client = monc->client;
323 struct ceph_monmap *monmap = NULL, *old = monc->monmap;
326 mutex_lock(&monc->mutex);
328 dout("handle_monmap\n");
329 p = msg->front.iov_base;
330 end = p + msg->front.iov_len;
332 monmap = ceph_monmap_decode(p, end);
333 if (IS_ERR(monmap)) {
334 pr_err("problem decoding monmap, %d\n",
335 (int)PTR_ERR(monmap));
339 if (ceph_check_fsid(monc->client, &monmap->fsid) < 0) {
344 client->monc.monmap = monmap;
348 mutex_unlock(&monc->mutex);
349 wake_up(&client->auth_wq);
355 static struct ceph_mon_statfs_request *__lookup_statfs(
356 struct ceph_mon_client *monc, u64 tid)
358 struct ceph_mon_statfs_request *req;
359 struct rb_node *n = monc->statfs_request_tree.rb_node;
362 req = rb_entry(n, struct ceph_mon_statfs_request, node);
365 else if (tid > req->tid)
373 static void __insert_statfs(struct ceph_mon_client *monc,
374 struct ceph_mon_statfs_request *new)
376 struct rb_node **p = &monc->statfs_request_tree.rb_node;
377 struct rb_node *parent = NULL;
378 struct ceph_mon_statfs_request *req = NULL;
382 req = rb_entry(parent, struct ceph_mon_statfs_request, node);
383 if (new->tid < req->tid)
385 else if (new->tid > req->tid)
391 rb_link_node(&new->node, parent, p);
392 rb_insert_color(&new->node, &monc->statfs_request_tree);
395 static void handle_statfs_reply(struct ceph_mon_client *monc,
396 struct ceph_msg *msg)
398 struct ceph_mon_statfs_request *req;
399 struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
402 if (msg->front.iov_len != sizeof(*reply))
404 tid = le64_to_cpu(msg->hdr.tid);
405 dout("handle_statfs_reply %p tid %llu\n", msg, tid);
407 mutex_lock(&monc->mutex);
408 req = __lookup_statfs(monc, tid);
410 *req->buf = reply->st;
413 mutex_unlock(&monc->mutex);
415 complete(&req->completion);
419 pr_err("corrupt statfs reply, no tid\n");
424 * (re)send a statfs request
426 static int send_statfs(struct ceph_mon_client *monc,
427 struct ceph_mon_statfs_request *req)
429 struct ceph_msg *msg;
430 struct ceph_mon_statfs *h;
432 dout("send_statfs tid %llu\n", req->tid);
433 msg = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), 0, 0, NULL);
437 msg->hdr.tid = cpu_to_le64(req->tid);
438 h = msg->front.iov_base;
439 h->monhdr.have_version = 0;
440 h->monhdr.session_mon = cpu_to_le16(-1);
441 h->monhdr.session_mon_tid = 0;
442 h->fsid = monc->monmap->fsid;
443 ceph_con_send(monc->con, msg);
448 * Do a synchronous statfs().
450 int ceph_monc_do_statfs(struct ceph_mon_client *monc, struct ceph_statfs *buf)
452 struct ceph_mon_statfs_request req;
456 init_completion(&req.completion);
458 /* allocate memory for reply */
459 err = ceph_msgpool_resv(&monc->msgpool_statfs_reply, 1);
463 /* register request */
464 mutex_lock(&monc->mutex);
465 req.tid = ++monc->last_tid;
466 req.last_attempt = jiffies;
467 req.delay = BASE_DELAY_INTERVAL;
468 __insert_statfs(monc, &req);
469 monc->num_statfs_requests++;
470 mutex_unlock(&monc->mutex);
472 /* send request and wait */
473 err = send_statfs(monc, &req);
475 err = wait_for_completion_interruptible(&req.completion);
477 mutex_lock(&monc->mutex);
478 rb_erase(&req.node, &monc->statfs_request_tree);
479 monc->num_statfs_requests--;
480 ceph_msgpool_resv(&monc->msgpool_statfs_reply, -1);
481 mutex_unlock(&monc->mutex);
489 * Resend pending statfs requests.
491 static void __resend_statfs(struct ceph_mon_client *monc)
493 struct ceph_mon_statfs_request *req;
496 for (p = rb_first(&monc->statfs_request_tree); p; p = rb_next(p)) {
497 req = rb_entry(p, struct ceph_mon_statfs_request, node);
498 send_statfs(monc, req);
503 * Delayed work. If we haven't mounted yet, retry. Otherwise,
504 * renew/retry subscription as needed (in case it is timing out, or we
505 * got an ENOMEM). And keep the monitor connection alive.
507 static void delayed_work(struct work_struct *work)
509 struct ceph_mon_client *monc =
510 container_of(work, struct ceph_mon_client, delayed_work.work);
512 dout("monc delayed_work\n");
513 mutex_lock(&monc->mutex);
515 __close_session(monc);
516 __open_session(monc); /* continue hunting */
518 ceph_con_keepalive(monc->con);
520 __validate_auth(monc);
522 if (monc->auth->ops->is_authenticated(monc->auth))
523 __send_subscribe(monc);
525 __schedule_delayed(monc);
526 mutex_unlock(&monc->mutex);
530 * On startup, we build a temporary monmap populated with the IPs
531 * provided by mount(2).
533 static int build_initial_monmap(struct ceph_mon_client *monc)
535 struct ceph_mount_args *args = monc->client->mount_args;
536 struct ceph_entity_addr *mon_addr = args->mon_addr;
537 int num_mon = args->num_mon;
540 /* build initial monmap */
541 monc->monmap = kzalloc(sizeof(*monc->monmap) +
542 num_mon*sizeof(monc->monmap->mon_inst[0]),
546 for (i = 0; i < num_mon; i++) {
547 monc->monmap->mon_inst[i].addr = mon_addr[i];
548 monc->monmap->mon_inst[i].addr.nonce = 0;
549 monc->monmap->mon_inst[i].name.type =
550 CEPH_ENTITY_TYPE_MON;
551 monc->monmap->mon_inst[i].name.num = cpu_to_le64(i);
553 monc->monmap->num_mon = num_mon;
554 monc->have_fsid = false;
556 /* release addr memory */
557 kfree(args->mon_addr);
558 args->mon_addr = NULL;
563 int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl)
568 memset(monc, 0, sizeof(*monc));
571 mutex_init(&monc->mutex);
573 err = build_initial_monmap(monc);
580 monc->auth = ceph_auth_init(cl->mount_args->name,
581 cl->mount_args->secret);
582 if (IS_ERR(monc->auth))
583 return PTR_ERR(monc->auth);
584 monc->auth->want_keys =
585 CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON |
586 CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS;
589 err = ceph_msgpool_init(&monc->msgpool_subscribe_ack,
590 sizeof(struct ceph_mon_subscribe_ack), 1, false);
593 err = ceph_msgpool_init(&monc->msgpool_statfs_reply,
594 sizeof(struct ceph_mon_statfs_reply), 0, false);
597 err = ceph_msgpool_init(&monc->msgpool_auth_reply, 4096, 1, false);
601 monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, 0, 0, NULL);
602 monc->pending_auth = 0;
603 if (IS_ERR(monc->m_auth)) {
604 err = PTR_ERR(monc->m_auth);
610 monc->hunting = true;
611 monc->sub_renew_after = jiffies;
614 INIT_DELAYED_WORK(&monc->delayed_work, delayed_work);
615 monc->statfs_request_tree = RB_ROOT;
616 monc->num_statfs_requests = 0;
619 monc->have_mdsmap = 0;
620 monc->have_osdmap = 0;
621 monc->want_next_osdmap = 1;
625 ceph_msgpool_destroy(&monc->msgpool_auth_reply);
627 ceph_msgpool_destroy(&monc->msgpool_subscribe_ack);
629 ceph_msgpool_destroy(&monc->msgpool_statfs_reply);
636 void ceph_monc_stop(struct ceph_mon_client *monc)
639 cancel_delayed_work_sync(&monc->delayed_work);
641 mutex_lock(&monc->mutex);
642 __close_session(monc);
644 monc->con->private = NULL;
645 monc->con->ops->put(monc->con);
648 mutex_unlock(&monc->mutex);
650 ceph_auth_destroy(monc->auth);
652 ceph_msg_put(monc->m_auth);
653 ceph_msgpool_destroy(&monc->msgpool_subscribe_ack);
654 ceph_msgpool_destroy(&monc->msgpool_statfs_reply);
655 ceph_msgpool_destroy(&monc->msgpool_auth_reply);
660 static void handle_auth_reply(struct ceph_mon_client *monc,
661 struct ceph_msg *msg)
665 mutex_lock(&monc->mutex);
666 monc->pending_auth = 0;
667 ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base,
669 monc->m_auth->front.iov_base,
670 monc->m_auth->front_max);
672 monc->client->auth_err = ret;
673 wake_up(&monc->client->auth_wq);
674 } else if (ret > 0) {
675 __send_prepared_auth_request(monc, ret);
676 } else if (monc->auth->ops->is_authenticated(monc->auth)) {
677 dout("authenticated, starting session\n");
679 monc->client->msgr->inst.name.type = CEPH_ENTITY_TYPE_CLIENT;
680 monc->client->msgr->inst.name.num = monc->auth->global_id;
682 __send_subscribe(monc);
683 __resend_statfs(monc);
685 mutex_unlock(&monc->mutex);
688 static int __validate_auth(struct ceph_mon_client *monc)
692 if (monc->pending_auth)
695 ret = ceph_build_auth(monc->auth, monc->m_auth->front.iov_base,
696 monc->m_auth->front_max);
698 return ret; /* either an error, or no need to authenticate */
699 __send_prepared_auth_request(monc, ret);
703 int ceph_monc_validate_auth(struct ceph_mon_client *monc)
707 mutex_lock(&monc->mutex);
708 ret = __validate_auth(monc);
709 mutex_unlock(&monc->mutex);
714 * handle incoming message
716 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
718 struct ceph_mon_client *monc = con->private;
719 int type = le16_to_cpu(msg->hdr.type);
725 case CEPH_MSG_AUTH_REPLY:
726 handle_auth_reply(monc, msg);
729 case CEPH_MSG_MON_SUBSCRIBE_ACK:
730 handle_subscribe_ack(monc, msg);
733 case CEPH_MSG_STATFS_REPLY:
734 handle_statfs_reply(monc, msg);
737 case CEPH_MSG_MON_MAP:
738 ceph_monc_handle_map(monc, msg);
741 case CEPH_MSG_MDS_MAP:
742 ceph_mdsc_handle_map(&monc->client->mdsc, msg);
745 case CEPH_MSG_OSD_MAP:
746 ceph_osdc_handle_map(&monc->client->osdc, msg);
750 pr_err("received unknown message type %d %s\n", type,
751 ceph_msg_type_name(type));
757 * Allocate memory for incoming message
759 static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con,
760 struct ceph_msg_header *hdr,
763 struct ceph_mon_client *monc = con->private;
764 int type = le16_to_cpu(hdr->type);
765 int front_len = le32_to_cpu(hdr->front_len);
766 struct ceph_msg *m = NULL;
771 case CEPH_MSG_MON_SUBSCRIBE_ACK:
772 m = ceph_msgpool_get(&monc->msgpool_subscribe_ack, front_len);
774 case CEPH_MSG_STATFS_REPLY:
775 m = ceph_msgpool_get(&monc->msgpool_statfs_reply, front_len);
777 case CEPH_MSG_AUTH_REPLY:
778 m = ceph_msgpool_get(&monc->msgpool_auth_reply, front_len);
780 case CEPH_MSG_MON_MAP:
781 case CEPH_MSG_MDS_MAP:
782 case CEPH_MSG_OSD_MAP:
783 m = ceph_msg_new(type, front_len, 0, 0, NULL);
788 pr_info("alloc_msg unknown type %d\n", type);
795 * If the monitor connection resets, pick a new monitor and resubmit
796 * any pending requests.
798 static void mon_fault(struct ceph_connection *con)
800 struct ceph_mon_client *monc = con->private;
806 mutex_lock(&monc->mutex);
810 if (monc->con && !monc->hunting)
811 pr_info("mon%d %s session lost, "
812 "hunting for new mon\n", monc->cur_mon,
813 pr_addr(&monc->con->peer_addr.in_addr));
815 __close_session(monc);
816 if (!monc->hunting) {
818 monc->hunting = true;
819 __open_session(monc);
821 /* already hunting, let's wait a bit */
822 __schedule_delayed(monc);
825 mutex_unlock(&monc->mutex);
828 const static struct ceph_connection_operations mon_con_ops = {
831 .dispatch = dispatch,
833 .alloc_msg = mon_alloc_msg,