projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ceph: whitespace cleanup
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ceph / mon_client.c
CommitLineData
ba75bb98
SW		 1#include "ceph_debug.h"
2
3#include <linux/types.h>
4#include <linux/random.h>
5#include <linux/sched.h>
6
7#include "mon_client.h"
8#include "super.h"
4e7a5dcd 9#include "auth.h"
ba75bb98
SW		 10#include "decode.h"
11
12/*
13 * Interact with Ceph monitor cluster. Handle requests for new map
14 * versions, and periodically resend as needed. Also implement
15 * statfs() and umount().
16 *
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.
22 *
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.
28 */
29
30const static struct ceph_connection_operations mon_con_ops;
31
32/*
33 * Decode a monmap blob (e.g., during mount).
34 */
35struct ceph_monmap *ceph_monmap_decode(void *p, void *end)
36{
37 struct ceph_monmap *m = NULL;
38 int i, err = -EINVAL;
39 struct ceph_fsid fsid;
40 u32 epoch, num_mon;
41 u16 version;
4e7a5dcd
SW		 42 u32 len;
43
44 ceph_decode_32_safe(&p, end, len, bad);
45 ceph_decode_need(&p, end, len, bad);
ba75bb98
SW		 46
47 dout("monmap_decode %p %p len %d\n", p, end, (int)(end-p));
48
49 ceph_decode_16_safe(&p, end, version, bad);
50
51 ceph_decode_need(&p, end, sizeof(fsid) + 2*sizeof(u32), bad);
52 ceph_decode_copy(&p, &fsid, sizeof(fsid));
c89136ea 53 epoch = ceph_decode_32(&p);
ba75bb98 54
c89136ea 55 num_mon = ceph_decode_32(&p);
ba75bb98
SW		 56 ceph_decode_need(&p, end, num_mon*sizeof(m->mon_inst[0]), bad);
57
58 if (num_mon >= CEPH_MAX_MON)
59 goto bad;
60 m = kmalloc(sizeof(*m) + sizeof(m->mon_inst[0])*num_mon, GFP_NOFS);
61 if (m == NULL)
62 return ERR_PTR(-ENOMEM);
63 m->fsid = fsid;
64 m->epoch = epoch;
65 m->num_mon = num_mon;
66 ceph_decode_copy(&p, m->mon_inst, num_mon*sizeof(m->mon_inst[0]));
63f2d211
SW		 67 for (i = 0; i < num_mon; i++)
68 ceph_decode_addr(&m->mon_inst[i].addr);
ba75bb98 69
ba75bb98
SW		 70 dout("monmap_decode epoch %d, num_mon %d\n", m->epoch,
71 m->num_mon);
72 for (i = 0; i < m->num_mon; i++)
73 dout("monmap_decode mon%d is %s\n", i,
74 pr_addr(&m->mon_inst[i].addr.in_addr));
75 return m;
76
77bad:
78 dout("monmap_decode failed with %d\n", err);
79 kfree(m);
80 return ERR_PTR(err);
81}
82
83/*
84 * return true if *addr is included in the monmap.
85 */
86int ceph_monmap_contains(struct ceph_monmap *m, struct ceph_entity_addr *addr)
87{
88 int i;
89
90 for (i = 0; i < m->num_mon; i++)
91 if (ceph_entity_addr_equal(addr, &m->mon_inst[i].addr))
92 return 1;
93 return 0;
94}
95
96/*
97 * Close monitor session, if any.
98 */
99static void __close_session(struct ceph_mon_client *monc)
100{
101 if (monc->con) {
102 dout("__close_session closing mon%d\n", monc->cur_mon);
4e7a5dcd 103 ceph_con_revoke(monc->con, monc->m_auth);
ba75bb98
SW		 104 ceph_con_close(monc->con);
105 monc->cur_mon = -1;
4e7a5dcd 106 ceph_auth_reset(monc->auth);
ba75bb98
SW		 107 }
108}
109
110/*
111 * Open a session with a (new) monitor.
112 */
113static int __open_session(struct ceph_mon_client *monc)
114{
115 char r;
4e7a5dcd 116 int ret;
ba75bb98
SW		 117
118 if (monc->cur_mon < 0) {
119 get_random_bytes(&r, 1);
120 monc->cur_mon = r % monc->monmap->num_mon;
121 dout("open_session num=%d r=%d -> mon%d\n",
122 monc->monmap->num_mon, r, monc->cur_mon);
123 monc->sub_sent = 0;
124 monc->sub_renew_after = jiffies; /* i.e., expired */
125 monc->want_next_osdmap = !!monc->want_next_osdmap;
126
127 dout("open_session mon%d opening\n", monc->cur_mon);
128 monc->con->peer_name.type = CEPH_ENTITY_TYPE_MON;
129 monc->con->peer_name.num = cpu_to_le64(monc->cur_mon);
130 ceph_con_open(monc->con,
131 &monc->monmap->mon_inst[monc->cur_mon].addr);
4e7a5dcd
SW		 132
133 /* initiatiate authentication handshake */
134 ret = ceph_auth_build_hello(monc->auth,
135 monc->m_auth->front.iov_base,
136 monc->m_auth->front_max);
137 monc->m_auth->front.iov_len = ret;
138 monc->m_auth->hdr.front_len = cpu_to_le32(ret);
139 ceph_msg_get(monc->m_auth); /* keep our ref */
140 ceph_con_send(monc->con, monc->m_auth);
ba75bb98
SW		 141 } else {
142 dout("open_session mon%d already open\n", monc->cur_mon);
143 }
144 return 0;
145}
146
147static bool __sub_expired(struct ceph_mon_client *monc)
148{
149 return time_after_eq(jiffies, monc->sub_renew_after);
150}
151
152/*
153 * Reschedule delayed work timer.
154 */
155static void __schedule_delayed(struct ceph_mon_client *monc)
156{
157 unsigned delay;
158
4e7a5dcd 159 if (monc->cur_mon < 0 || __sub_expired(monc))
ba75bb98
SW		 160 delay = 10 * HZ;
161 else
162 delay = 20 * HZ;
163 dout("__schedule_delayed after %u\n", delay);
164 schedule_delayed_work(&monc->delayed_work, delay);
165}
166
167/*
168 * Send subscribe request for mdsmap and/or osdmap.
169 */
170static void __send_subscribe(struct ceph_mon_client *monc)
171{
172 dout("__send_subscribe sub_sent=%u exp=%u want_osd=%d\n",
173 (unsigned)monc->sub_sent, __sub_expired(monc),
174 monc->want_next_osdmap);
175 if ((__sub_expired(monc) && !monc->sub_sent) ||
176 monc->want_next_osdmap == 1) {
177 struct ceph_msg *msg;
178 struct ceph_mon_subscribe_item *i;
179 void *p, *end;
180
4e7a5dcd 181 msg = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 96, 0, 0, NULL);
ba75bb98
SW		 182 if (!msg)
183 return;
184
185 p = msg->front.iov_base;
186 end = p + msg->front.iov_len;
187
188 dout("__send_subscribe to 'mdsmap' %u+\n",
189 (unsigned)monc->have_mdsmap);
190 if (monc->want_next_osdmap) {
191 dout("__send_subscribe to 'osdmap' %u\n",
192 (unsigned)monc->have_osdmap);
4e7a5dcd 193 ceph_encode_32(&p, 3);
ba75bb98
SW		 194 ceph_encode_string(&p, end, "osdmap", 6);
195 i = p;
196 i->have = cpu_to_le64(monc->have_osdmap);
197 i->onetime = 1;
198 p += sizeof(*i);
199 monc->want_next_osdmap = 2; /* requested */
200 } else {
4e7a5dcd 201 ceph_encode_32(&p, 2);
ba75bb98
SW		 202 }
203 ceph_encode_string(&p, end, "mdsmap", 6);
204 i = p;
205 i->have = cpu_to_le64(monc->have_mdsmap);
206 i->onetime = 0;
207 p += sizeof(*i);
4e7a5dcd
SW		 208 ceph_encode_string(&p, end, "monmap", 6);
209 i = p;
210 i->have = 0;
211 i->onetime = 0;
212 p += sizeof(*i);
ba75bb98
SW		 213
214 msg->front.iov_len = p - msg->front.iov_base;
215 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
216 ceph_con_send(monc->con, msg);
217
218 monc->sub_sent = jiffies | 1; /* never 0 */
219 }
220}
221
222static void handle_subscribe_ack(struct ceph_mon_client *monc,
223 struct ceph_msg *msg)
224{
225 unsigned seconds;
07bd10fb
SW		 226 struct ceph_mon_subscribe_ack *h = msg->front.iov_base;
227
228 if (msg->front.iov_len < sizeof(*h))
229 goto bad;
230 seconds = le32_to_cpu(h->duration);
ba75bb98 231
ba75bb98
SW		 232 mutex_lock(&monc->mutex);
233 if (monc->hunting) {
234 pr_info("mon%d %s session established\n",
235 monc->cur_mon, pr_addr(&monc->con->peer_addr.in_addr));
236 monc->hunting = false;
237 }
238 dout("handle_subscribe_ack after %d seconds\n", seconds);
0656d11b 239 monc->sub_renew_after = monc->sub_sent + (seconds >> 1)*HZ - 1;
ba75bb98
SW		 240 monc->sub_sent = 0;
241 mutex_unlock(&monc->mutex);
242 return;
243bad:
244 pr_err("got corrupt subscribe-ack msg\n");
245}
246
247/*
248 * Keep track of which maps we have
249 */
250int ceph_monc_got_mdsmap(struct ceph_mon_client *monc, u32 got)
251{
252 mutex_lock(&monc->mutex);
253 monc->have_mdsmap = got;
254 mutex_unlock(&monc->mutex);
255 return 0;
256}
257
258int ceph_monc_got_osdmap(struct ceph_mon_client *monc, u32 got)
259{
260 mutex_lock(&monc->mutex);
261 monc->have_osdmap = got;
262 monc->want_next_osdmap = 0;
263 mutex_unlock(&monc->mutex);
264 return 0;
265}
266
267/*
268 * Register interest in the next osdmap
269 */
270void ceph_monc_request_next_osdmap(struct ceph_mon_client *monc)
271{
272 dout("request_next_osdmap have %u\n", monc->have_osdmap);
273 mutex_lock(&monc->mutex);
274 if (!monc->want_next_osdmap)
275 monc->want_next_osdmap = 1;
276 if (monc->want_next_osdmap < 2)
277 __send_subscribe(monc);
278 mutex_unlock(&monc->mutex);
279}
280
4e7a5dcd 281/*
50b885b9 282 *
4e7a5dcd
SW		 283 */
284int ceph_monc_open_session(struct ceph_mon_client *monc)
ba75bb98
SW		 285{
286 if (!monc->con) {
287 monc->con = kmalloc(sizeof(*monc->con), GFP_KERNEL);
288 if (!monc->con)
289 return -ENOMEM;
290 ceph_con_init(monc->client->msgr, monc->con);
291 monc->con->private = monc;
292 monc->con->ops = &mon_con_ops;
293 }
294
295 mutex_lock(&monc->mutex);
4e7a5dcd 296 __open_session(monc);
ba75bb98
SW		 297 __schedule_delayed(monc);
298 mutex_unlock(&monc->mutex);
299 return 0;
300}
301
4e7a5dcd
SW		 302/*
303 * The monitor responds with mount ack indicate mount success. The
304 * included client ticket allows the client to talk to MDSs and OSDs.
305 */
0743304d
SW		 306static void ceph_monc_handle_map(struct ceph_mon_client *monc,
307 struct ceph_msg *msg)
4e7a5dcd
SW		 308{
309 struct ceph_client *client = monc->client;
310 struct ceph_monmap *monmap = NULL, *old = monc->monmap;
311 void *p, *end;
312
313 mutex_lock(&monc->mutex);
314
315 dout("handle_monmap\n");
316 p = msg->front.iov_base;
317 end = p + msg->front.iov_len;
318
319 monmap = ceph_monmap_decode(p, end);
320 if (IS_ERR(monmap)) {
321 pr_err("problem decoding monmap, %d\n",
322 (int)PTR_ERR(monmap));
323 return;
324 }
0743304d
SW		 325
326 if (ceph_check_fsid(monc->client, &monmap->fsid) < 0) {
4e7a5dcd
SW		 327 kfree(monmap);
328 return;
329 }
330
331 client->monc.monmap = monmap;
4e7a5dcd
SW		 332 kfree(old);
333
334 mutex_unlock(&monc->mutex);
335 wake_up(&client->mount_wq);
336}
337
ba75bb98
SW		 338/*
339 * statfs
340 */
341static void handle_statfs_reply(struct ceph_mon_client *monc,
342 struct ceph_msg *msg)
343{
344 struct ceph_mon_statfs_request *req;
345 struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
346 u64 tid;
347
348 if (msg->front.iov_len != sizeof(*reply))
349 goto bad;
350 tid = le64_to_cpu(reply->tid);
351 dout("handle_statfs_reply %p tid %llu\n", msg, tid);
352
353 mutex_lock(&monc->mutex);
354 req = radix_tree_lookup(&monc->statfs_request_tree, tid);
355 if (req) {
356 *req->buf = reply->st;
357 req->result = 0;
358 }
359 mutex_unlock(&monc->mutex);
360 if (req)
361 complete(&req->completion);
362 return;
363
364bad:
365 pr_err("corrupt statfs reply, no tid\n");
366}
367
368/*
369 * (re)send a statfs request
370 */
371static int send_statfs(struct ceph_mon_client *monc,
372 struct ceph_mon_statfs_request *req)
373{
374 struct ceph_msg *msg;
375 struct ceph_mon_statfs *h;
ba75bb98
SW		 376
377 dout("send_statfs tid %llu\n", req->tid);
ba75bb98
SW		 378 msg = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), 0, 0, NULL);
379 if (IS_ERR(msg))
380 return PTR_ERR(msg);
381 req->request = msg;
382 h = msg->front.iov_base;
13e38c8a
SW		 383 h->monhdr.have_version = 0;
384 h->monhdr.session_mon = cpu_to_le16(-1);
385 h->monhdr.session_mon_tid = 0;
ba75bb98
SW		 386 h->fsid = monc->monmap->fsid;
387 h->tid = cpu_to_le64(req->tid);
388 ceph_con_send(monc->con, msg);
389 return 0;
390}
391
392/*
393 * Do a synchronous statfs().
394 */
395int ceph_monc_do_statfs(struct ceph_mon_client *monc, struct ceph_statfs *buf)
396{
397 struct ceph_mon_statfs_request req;
398 int err;
399
400 req.buf = buf;
401 init_completion(&req.completion);
402
403 /* allocate memory for reply */
404 err = ceph_msgpool_resv(&monc->msgpool_statfs_reply, 1);
405 if (err)
406 return err;
407
408 /* register request */
409 mutex_lock(&monc->mutex);
410 req.tid = ++monc->last_tid;
411 req.last_attempt = jiffies;
412 req.delay = BASE_DELAY_INTERVAL;
413 if (radix_tree_insert(&monc->statfs_request_tree, req.tid, &req) < 0) {
414 mutex_unlock(&monc->mutex);
415 pr_err("ENOMEM in do_statfs\n");
416 return -ENOMEM;
417 }
418 monc->num_statfs_requests++;
419 mutex_unlock(&monc->mutex);
420
421 /* send request and wait */
422 err = send_statfs(monc, &req);
423 if (!err)
424 err = wait_for_completion_interruptible(&req.completion);
425
426 mutex_lock(&monc->mutex);
427 radix_tree_delete(&monc->statfs_request_tree, req.tid);
428 monc->num_statfs_requests--;
429 ceph_msgpool_resv(&monc->msgpool_statfs_reply, -1);
430 mutex_unlock(&monc->mutex);
431
432 if (!err)
433 err = req.result;
434 return err;
435}
436
437/*
438 * Resend pending statfs requests.
439 */
440static void __resend_statfs(struct ceph_mon_client *monc)
441{
442 u64 next_tid = 0;
443 int got;
444 int did = 0;
445 struct ceph_mon_statfs_request *req;
446
447 while (1) {
448 got = radix_tree_gang_lookup(&monc->statfs_request_tree,
449 (void **)&req,
450 next_tid, 1);
451 if (got == 0)
452 break;
453 did++;
454 next_tid = req->tid + 1;
455
456 send_statfs(monc, req);
457 }
458}
459
460/*
461 * Delayed work. If we haven't mounted yet, retry. Otherwise,
462 * renew/retry subscription as needed (in case it is timing out, or we
463 * got an ENOMEM). And keep the monitor connection alive.
464 */
465static void delayed_work(struct work_struct *work)
466{
467 struct ceph_mon_client *monc =
468 container_of(work, struct ceph_mon_client, delayed_work.work);
469
470 dout("monc delayed_work\n");
471 mutex_lock(&monc->mutex);
4e7a5dcd
SW		 472 if (monc->hunting) {
473 __close_session(monc);
474 __open_session(monc); /* continue hunting */
ba75bb98 475 } else {
4e7a5dcd
SW		 476 ceph_con_keepalive(monc->con);
477 if (monc->auth->ops->is_authenticated(monc->auth))
478 __send_subscribe(monc);
ba75bb98 479 }
ba75bb98
SW		 480 __schedule_delayed(monc);
481 mutex_unlock(&monc->mutex);
482}
483
6b805185
SW		 484/*
485 * On startup, we build a temporary monmap populated with the IPs
486 * provided by mount(2).
487 */
488static int build_initial_monmap(struct ceph_mon_client *monc)
489{
490 struct ceph_mount_args *args = monc->client->mount_args;
491 struct ceph_entity_addr *mon_addr = args->mon_addr;
492 int num_mon = args->num_mon;
493 int i;
494
495 /* build initial monmap */
496 monc->monmap = kzalloc(sizeof(*monc->monmap) +
497 num_mon*sizeof(monc->monmap->mon_inst[0]),
498 GFP_KERNEL);
499 if (!monc->monmap)
500 return -ENOMEM;
501 for (i = 0; i < num_mon; i++) {
502 monc->monmap->mon_inst[i].addr = mon_addr[i];
503 monc->monmap->mon_inst[i].addr.erank = 0;
504 monc->monmap->mon_inst[i].addr.nonce = 0;
505 monc->monmap->mon_inst[i].name.type =
506 CEPH_ENTITY_TYPE_MON;
507 monc->monmap->mon_inst[i].name.num = cpu_to_le64(i);
508 }
509 monc->monmap->num_mon = num_mon;
4e7a5dcd 510 monc->have_fsid = false;
6b805185
SW		 511
512 /* release addr memory */
513 kfree(args->mon_addr);
514 args->mon_addr = NULL;
515 args->num_mon = 0;
516 return 0;
517}
518
ba75bb98
SW		 519int ceph_monc_init(struct ceph_mon_client *monc, struct ceph_client *cl)
520{
521 int err = 0;
522
523 dout("init\n");
524 memset(monc, 0, sizeof(*monc));
525 monc->client = cl;
526 monc->monmap = NULL;
527 mutex_init(&monc->mutex);
528
6b805185
SW		 529 err = build_initial_monmap(monc);
530 if (err)
531 goto out;
532
ba75bb98
SW		 533 monc->con = NULL;
534
4e7a5dcd
SW		 535 /* authentication */
536 monc->auth = ceph_auth_init(cl->mount_args->name,
537 cl->mount_args->secret);
538 if (IS_ERR(monc->auth))
539 return PTR_ERR(monc->auth);
540 monc->auth->want_keys =
541 CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON |
542 CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS;
543
ba75bb98 544 /* msg pools */
07bd10fb
SW		 545 err = ceph_msgpool_init(&monc->msgpool_subscribe_ack,
546 sizeof(struct ceph_mon_subscribe_ack), 1, false);
ba75bb98 547 if (err < 0)
4e7a5dcd 548 goto out_monmap;
ba75bb98
SW		 549 err = ceph_msgpool_init(&monc->msgpool_statfs_reply,
550 sizeof(struct ceph_mon_statfs_reply), 0, false);
551 if (err < 0)
4e7a5dcd
SW		 552 goto out_pool1;
553 err = ceph_msgpool_init(&monc->msgpool_auth_reply, 4096, 1, false);
554 if (err < 0)
555 goto out_pool2;
556
557 monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, 0, 0, NULL);
558 if (IS_ERR(monc->m_auth)) {
559 err = PTR_ERR(monc->m_auth);
560 monc->m_auth = NULL;
561 goto out_pool3;
562 }
ba75bb98
SW		 563
564 monc->cur_mon = -1;
4e7a5dcd 565 monc->hunting = true;
ba75bb98
SW		 566 monc->sub_renew_after = jiffies;
567 monc->sub_sent = 0;
568
569 INIT_DELAYED_WORK(&monc->delayed_work, delayed_work);
570 INIT_RADIX_TREE(&monc->statfs_request_tree, GFP_NOFS);
571 monc->num_statfs_requests = 0;
572 monc->last_tid = 0;
573
574 monc->have_mdsmap = 0;
575 monc->have_osdmap = 0;
576 monc->want_next_osdmap = 1;
4e7a5dcd
SW		 577 return 0;
578
579out_pool3:
580 ceph_msgpool_destroy(&monc->msgpool_auth_reply);
581out_pool2:
582 ceph_msgpool_destroy(&monc->msgpool_subscribe_ack);
583out_pool1:
584 ceph_msgpool_destroy(&monc->msgpool_statfs_reply);
585out_monmap:
586 kfree(monc->monmap);
ba75bb98
SW		 587out:
588 return err;
589}
590
591void ceph_monc_stop(struct ceph_mon_client *monc)
592{
593 dout("stop\n");
594 cancel_delayed_work_sync(&monc->delayed_work);
595
596 mutex_lock(&monc->mutex);
597 __close_session(monc);
598 if (monc->con) {
599 monc->con->private = NULL;
600 monc->con->ops->put(monc->con);
601 monc->con = NULL;
602 }
603 mutex_unlock(&monc->mutex);
604
4e7a5dcd
SW		 605 ceph_auth_destroy(monc->auth);
606
607 ceph_msg_put(monc->m_auth);
ba75bb98
SW		 608 ceph_msgpool_destroy(&monc->msgpool_subscribe_ack);
609 ceph_msgpool_destroy(&monc->msgpool_statfs_reply);
4e7a5dcd 610 ceph_msgpool_destroy(&monc->msgpool_auth_reply);
ba75bb98
SW		 611
612 kfree(monc->monmap);
613}
614
615
4e7a5dcd
SW		 616static void handle_auth_reply(struct ceph_mon_client *monc,
617 struct ceph_msg *msg)
618{
619 int ret;
620
621 mutex_lock(&monc->mutex);
622 ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base,
623 msg->front.iov_len,
624 monc->m_auth->front.iov_base,
625 monc->m_auth->front_max);
626 if (ret < 0) {
627 monc->client->mount_err = ret;
628 wake_up(&monc->client->mount_wq);
629 } else if (ret > 0) {
630 monc->m_auth->front.iov_len = ret;
631 monc->m_auth->hdr.front_len = cpu_to_le32(ret);
632 ceph_msg_get(monc->m_auth); /* keep our ref */
633 ceph_con_send(monc->con, monc->m_auth);
634 } else if (monc->auth->ops->is_authenticated(monc->auth)) {
635 dout("authenticated, starting session\n");
0743304d
SW		 636
637 monc->client->msgr->inst.name.type = CEPH_ENTITY_TYPE_CLIENT;
638 monc->client->msgr->inst.name.num = monc->auth->global_id;
639
4e7a5dcd
SW		 640 __send_subscribe(monc);
641 __resend_statfs(monc);
642 }
643 mutex_unlock(&monc->mutex);
644}
645
ba75bb98
SW		 646/*
647 * handle incoming message
648 */
649static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
650{
651 struct ceph_mon_client *monc = con->private;
652 int type = le16_to_cpu(msg->hdr.type);
653
654 if (!monc)
655 return;
656
657 switch (type) {
4e7a5dcd
SW		 658 case CEPH_MSG_AUTH_REPLY:
659 handle_auth_reply(monc, msg);
ba75bb98
SW		 660 break;
661
662 case CEPH_MSG_MON_SUBSCRIBE_ACK:
663 handle_subscribe_ack(monc, msg);
664 break;
665
666 case CEPH_MSG_STATFS_REPLY:
667 handle_statfs_reply(monc, msg);
668 break;
669
4e7a5dcd
SW		 670 case CEPH_MSG_MON_MAP:
671 ceph_monc_handle_map(monc, msg);
672 break;
673
ba75bb98
SW		 674 case CEPH_MSG_MDS_MAP:
675 ceph_mdsc_handle_map(&monc->client->mdsc, msg);
676 break;
677
678 case CEPH_MSG_OSD_MAP:
679 ceph_osdc_handle_map(&monc->client->osdc, msg);
680 break;
681
682 default:
683 pr_err("received unknown message type %d %s\n", type,
684 ceph_msg_type_name(type));
685 }
686 ceph_msg_put(msg);
687}
688
689/*
690 * Allocate memory for incoming message
691 */
692static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con,
693 struct ceph_msg_header *hdr)
694{
695 struct ceph_mon_client *monc = con->private;
696 int type = le16_to_cpu(hdr->type);
8f3bc053 697 int front = le32_to_cpu(hdr->front_len);
ba75bb98
SW		 698
699 switch (type) {
ba75bb98 700 case CEPH_MSG_MON_SUBSCRIBE_ACK:
8f3bc053 701 return ceph_msgpool_get(&monc->msgpool_subscribe_ack, front);
ba75bb98 702 case CEPH_MSG_STATFS_REPLY:
8f3bc053 703 return ceph_msgpool_get(&monc->msgpool_statfs_reply, front);
4e7a5dcd
SW		 704 case CEPH_MSG_AUTH_REPLY:
705 return ceph_msgpool_get(&monc->msgpool_auth_reply, front);
ba75bb98
SW		 706 }
707 return ceph_alloc_msg(con, hdr);
708}
709
710/*
711 * If the monitor connection resets, pick a new monitor and resubmit
712 * any pending requests.
713 */
714static void mon_fault(struct ceph_connection *con)
715{
716 struct ceph_mon_client *monc = con->private;
717
718 if (!monc)
719 return;
720
721 dout("mon_fault\n");
722 mutex_lock(&monc->mutex);
723 if (!con->private)
724 goto out;
725
726 if (monc->con && !monc->hunting)
727 pr_info("mon%d %s session lost, "
728 "hunting for new mon\n", monc->cur_mon,
729 pr_addr(&monc->con->peer_addr.in_addr));
730
731 __close_session(monc);
732 if (!monc->hunting) {
733 /* start hunting */
734 monc->hunting = true;
4e7a5dcd 735 __open_session(monc);
ba75bb98
SW		 736 } else {
737 /* already hunting, let's wait a bit */
738 __schedule_delayed(monc);
739 }
740out:
741 mutex_unlock(&monc->mutex);
742}
743
744const static struct ceph_connection_operations mon_con_ops = {
745 .get = ceph_con_get,
746 .put = ceph_con_put,
747 .dispatch = dispatch,
748 .fault = mon_fault,
749 .alloc_msg = mon_alloc_msg,
750 .alloc_middle = ceph_alloc_middle,
751};
kernel source for telekom puls (alcatel/mediatek)