]>
Commit | Line | Data |
---|---|---|
3d14c5d2 | 1 | #include <linux/ceph/ceph_debug.h> |
ba75bb98 | 2 | |
3d14c5d2 | 3 | #include <linux/module.h> |
ba75bb98 | 4 | #include <linux/types.h> |
5a0e3ad6 | 5 | #include <linux/slab.h> |
ba75bb98 SW |
6 | #include <linux/random.h> |
7 | #include <linux/sched.h> | |
8 | ||
3d14c5d2 YS |
9 | #include <linux/ceph/mon_client.h> |
10 | #include <linux/ceph/libceph.h> | |
ab434b60 | 11 | #include <linux/ceph/debugfs.h> |
3d14c5d2 | 12 | #include <linux/ceph/decode.h> |
3d14c5d2 | 13 | #include <linux/ceph/auth.h> |
ba75bb98 SW |
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 | ||
9e32789f | 33 | static const struct ceph_connection_operations mon_con_ops; |
ba75bb98 | 34 | |
9bd2e6f8 SW |
35 | static int __validate_auth(struct ceph_mon_client *monc); |
36 | ||
ba75bb98 SW |
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; | |
4e7a5dcd SW |
47 | u32 len; |
48 | ||
49 | ceph_decode_32_safe(&p, end, len, bad); | |
50 | ceph_decode_need(&p, end, len, bad); | |
ba75bb98 SW |
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)); | |
c89136ea | 58 | epoch = ceph_decode_32(&p); |
ba75bb98 | 59 | |
c89136ea | 60 | num_mon = ceph_decode_32(&p); |
ba75bb98 SW |
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])); | |
63f2d211 SW |
72 | for (i = 0; i < num_mon; i++) |
73 | ceph_decode_addr(&m->mon_inst[i].addr); | |
ba75bb98 | 74 | |
ba75bb98 SW |
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, | |
3d14c5d2 | 79 | ceph_pr_addr(&m->mon_inst[i].addr.in_addr)); |
ba75bb98 SW |
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++) | |
103e2d3a | 96 | if (memcmp(addr, &m->mon_inst[i].addr, sizeof(*addr)) == 0) |
ba75bb98 SW |
97 | return 1; |
98 | return 0; | |
99 | } | |
100 | ||
5ce6e9db SW |
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); | |
97069001 | 109 | ceph_con_revoke(monc->con, monc->m_auth); |
5ce6e9db SW |
110 | ceph_msg_get(monc->m_auth); /* keep our ref */ |
111 | ceph_con_send(monc->con, monc->m_auth); | |
112 | } | |
113 | ||
ba75bb98 SW |
114 | /* |
115 | * Close monitor session, if any. | |
116 | */ | |
117 | static void __close_session(struct ceph_mon_client *monc) | |
118 | { | |
f6a2f5be SW |
119 | dout("__close_session closing mon%d\n", monc->cur_mon); |
120 | ceph_con_revoke(monc->con, 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); | |
ba75bb98 SW |
125 | } |
126 | ||
127 | /* | |
128 | * Open a session with a (new) monitor. | |
129 | */ | |
130 | static int __open_session(struct ceph_mon_client *monc) | |
131 | { | |
132 | char r; | |
4e7a5dcd | 133 | int ret; |
ba75bb98 SW |
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 | monc->con->peer_name.type = CEPH_ENTITY_TYPE_MON; | |
146 | monc->con->peer_name.num = cpu_to_le64(monc->cur_mon); | |
147 | ceph_con_open(monc->con, | |
148 | &monc->monmap->mon_inst[monc->cur_mon].addr); | |
4e7a5dcd SW |
149 | |
150 | /* initiatiate authentication handshake */ | |
151 | ret = ceph_auth_build_hello(monc->auth, | |
152 | monc->m_auth->front.iov_base, | |
153 | monc->m_auth->front_max); | |
5ce6e9db | 154 | __send_prepared_auth_request(monc, ret); |
ba75bb98 SW |
155 | } else { |
156 | dout("open_session mon%d already open\n", monc->cur_mon); | |
157 | } | |
158 | return 0; | |
159 | } | |
160 | ||
161 | static bool __sub_expired(struct ceph_mon_client *monc) | |
162 | { | |
163 | return time_after_eq(jiffies, monc->sub_renew_after); | |
164 | } | |
165 | ||
166 | /* | |
167 | * Reschedule delayed work timer. | |
168 | */ | |
169 | static void __schedule_delayed(struct ceph_mon_client *monc) | |
170 | { | |
171 | unsigned delay; | |
172 | ||
4e7a5dcd | 173 | if (monc->cur_mon < 0 || __sub_expired(monc)) |
ba75bb98 SW |
174 | delay = 10 * HZ; |
175 | else | |
176 | delay = 20 * HZ; | |
177 | dout("__schedule_delayed after %u\n", delay); | |
178 | schedule_delayed_work(&monc->delayed_work, delay); | |
179 | } | |
180 | ||
181 | /* | |
182 | * Send subscribe request for mdsmap and/or osdmap. | |
183 | */ | |
184 | static void __send_subscribe(struct ceph_mon_client *monc) | |
185 | { | |
186 | dout("__send_subscribe sub_sent=%u exp=%u want_osd=%d\n", | |
187 | (unsigned)monc->sub_sent, __sub_expired(monc), | |
188 | monc->want_next_osdmap); | |
189 | if ((__sub_expired(monc) && !monc->sub_sent) || | |
190 | monc->want_next_osdmap == 1) { | |
240ed68e | 191 | struct ceph_msg *msg = monc->m_subscribe; |
ba75bb98 SW |
192 | struct ceph_mon_subscribe_item *i; |
193 | void *p, *end; | |
3d14c5d2 | 194 | int num; |
ba75bb98 | 195 | |
ba75bb98 | 196 | p = msg->front.iov_base; |
240ed68e | 197 | end = p + msg->front_max; |
ba75bb98 | 198 | |
3d14c5d2 YS |
199 | num = 1 + !!monc->want_next_osdmap + !!monc->want_mdsmap; |
200 | ceph_encode_32(&p, num); | |
201 | ||
ba75bb98 SW |
202 | if (monc->want_next_osdmap) { |
203 | dout("__send_subscribe to 'osdmap' %u\n", | |
204 | (unsigned)monc->have_osdmap); | |
ba75bb98 SW |
205 | ceph_encode_string(&p, end, "osdmap", 6); |
206 | i = p; | |
207 | i->have = cpu_to_le64(monc->have_osdmap); | |
208 | i->onetime = 1; | |
209 | p += sizeof(*i); | |
210 | monc->want_next_osdmap = 2; /* requested */ | |
ba75bb98 | 211 | } |
3d14c5d2 YS |
212 | if (monc->want_mdsmap) { |
213 | dout("__send_subscribe to 'mdsmap' %u+\n", | |
214 | (unsigned)monc->have_mdsmap); | |
215 | ceph_encode_string(&p, end, "mdsmap", 6); | |
216 | i = p; | |
217 | i->have = cpu_to_le64(monc->have_mdsmap); | |
218 | i->onetime = 0; | |
219 | p += sizeof(*i); | |
220 | } | |
4e7a5dcd SW |
221 | ceph_encode_string(&p, end, "monmap", 6); |
222 | i = p; | |
223 | i->have = 0; | |
224 | i->onetime = 0; | |
225 | p += sizeof(*i); | |
ba75bb98 SW |
226 | |
227 | msg->front.iov_len = p - msg->front.iov_base; | |
228 | msg->hdr.front_len = cpu_to_le32(msg->front.iov_len); | |
240ed68e SW |
229 | ceph_con_revoke(monc->con, msg); |
230 | ceph_con_send(monc->con, ceph_msg_get(msg)); | |
ba75bb98 SW |
231 | |
232 | monc->sub_sent = jiffies | 1; /* never 0 */ | |
233 | } | |
234 | } | |
235 | ||
236 | static void handle_subscribe_ack(struct ceph_mon_client *monc, | |
237 | struct ceph_msg *msg) | |
238 | { | |
239 | unsigned seconds; | |
07bd10fb SW |
240 | struct ceph_mon_subscribe_ack *h = msg->front.iov_base; |
241 | ||
242 | if (msg->front.iov_len < sizeof(*h)) | |
243 | goto bad; | |
244 | seconds = le32_to_cpu(h->duration); | |
ba75bb98 | 245 | |
ba75bb98 SW |
246 | mutex_lock(&monc->mutex); |
247 | if (monc->hunting) { | |
248 | pr_info("mon%d %s session established\n", | |
3d14c5d2 YS |
249 | monc->cur_mon, |
250 | ceph_pr_addr(&monc->con->peer_addr.in_addr)); | |
ba75bb98 SW |
251 | monc->hunting = false; |
252 | } | |
253 | dout("handle_subscribe_ack after %d seconds\n", seconds); | |
0656d11b | 254 | monc->sub_renew_after = monc->sub_sent + (seconds >> 1)*HZ - 1; |
ba75bb98 SW |
255 | monc->sub_sent = 0; |
256 | mutex_unlock(&monc->mutex); | |
257 | return; | |
258 | bad: | |
259 | pr_err("got corrupt subscribe-ack msg\n"); | |
9ec7cab1 | 260 | ceph_msg_dump(msg); |
ba75bb98 SW |
261 | } |
262 | ||
263 | /* | |
264 | * Keep track of which maps we have | |
265 | */ | |
266 | int ceph_monc_got_mdsmap(struct ceph_mon_client *monc, u32 got) | |
267 | { | |
268 | mutex_lock(&monc->mutex); | |
269 | monc->have_mdsmap = got; | |
270 | mutex_unlock(&monc->mutex); | |
271 | return 0; | |
272 | } | |
3d14c5d2 | 273 | EXPORT_SYMBOL(ceph_monc_got_mdsmap); |
ba75bb98 SW |
274 | |
275 | int ceph_monc_got_osdmap(struct ceph_mon_client *monc, u32 got) | |
276 | { | |
277 | mutex_lock(&monc->mutex); | |
278 | monc->have_osdmap = got; | |
279 | monc->want_next_osdmap = 0; | |
280 | mutex_unlock(&monc->mutex); | |
281 | return 0; | |
282 | } | |
283 | ||
284 | /* | |
285 | * Register interest in the next osdmap | |
286 | */ | |
287 | void ceph_monc_request_next_osdmap(struct ceph_mon_client *monc) | |
288 | { | |
289 | dout("request_next_osdmap have %u\n", monc->have_osdmap); | |
290 | mutex_lock(&monc->mutex); | |
291 | if (!monc->want_next_osdmap) | |
292 | monc->want_next_osdmap = 1; | |
293 | if (monc->want_next_osdmap < 2) | |
294 | __send_subscribe(monc); | |
295 | mutex_unlock(&monc->mutex); | |
296 | } | |
297 | ||
4e7a5dcd | 298 | /* |
50b885b9 | 299 | * |
4e7a5dcd SW |
300 | */ |
301 | int ceph_monc_open_session(struct ceph_mon_client *monc) | |
ba75bb98 | 302 | { |
ba75bb98 | 303 | mutex_lock(&monc->mutex); |
4e7a5dcd | 304 | __open_session(monc); |
ba75bb98 SW |
305 | __schedule_delayed(monc); |
306 | mutex_unlock(&monc->mutex); | |
307 | return 0; | |
308 | } | |
3d14c5d2 | 309 | EXPORT_SYMBOL(ceph_monc_open_session); |
ba75bb98 | 310 | |
4e7a5dcd SW |
311 | /* |
312 | * The monitor responds with mount ack indicate mount success. The | |
313 | * included client ticket allows the client to talk to MDSs and OSDs. | |
314 | */ | |
0743304d SW |
315 | static void ceph_monc_handle_map(struct ceph_mon_client *monc, |
316 | struct ceph_msg *msg) | |
4e7a5dcd SW |
317 | { |
318 | struct ceph_client *client = monc->client; | |
319 | struct ceph_monmap *monmap = NULL, *old = monc->monmap; | |
320 | void *p, *end; | |
321 | ||
322 | mutex_lock(&monc->mutex); | |
323 | ||
324 | dout("handle_monmap\n"); | |
325 | p = msg->front.iov_base; | |
326 | end = p + msg->front.iov_len; | |
327 | ||
328 | monmap = ceph_monmap_decode(p, end); | |
329 | if (IS_ERR(monmap)) { | |
330 | pr_err("problem decoding monmap, %d\n", | |
331 | (int)PTR_ERR(monmap)); | |
d4a780ce | 332 | goto out; |
4e7a5dcd | 333 | } |
0743304d SW |
334 | |
335 | if (ceph_check_fsid(monc->client, &monmap->fsid) < 0) { | |
4e7a5dcd | 336 | kfree(monmap); |
d4a780ce | 337 | goto out; |
4e7a5dcd SW |
338 | } |
339 | ||
340 | client->monc.monmap = monmap; | |
4e7a5dcd SW |
341 | kfree(old); |
342 | ||
ab434b60 SW |
343 | if (!client->have_fsid) { |
344 | client->have_fsid = true; | |
345 | mutex_unlock(&monc->mutex); | |
346 | /* | |
347 | * do debugfs initialization without mutex to avoid | |
348 | * creating a locking dependency | |
349 | */ | |
350 | ceph_debugfs_client_init(client); | |
351 | goto out_unlocked; | |
352 | } | |
d4a780ce | 353 | out: |
4e7a5dcd | 354 | mutex_unlock(&monc->mutex); |
ab434b60 | 355 | out_unlocked: |
03066f23 | 356 | wake_up_all(&client->auth_wq); |
4e7a5dcd SW |
357 | } |
358 | ||
ba75bb98 | 359 | /* |
e56fa10e | 360 | * generic requests (e.g., statfs, poolop) |
ba75bb98 | 361 | */ |
f8c76f6f | 362 | static struct ceph_mon_generic_request *__lookup_generic_req( |
85ff03f6 SW |
363 | struct ceph_mon_client *monc, u64 tid) |
364 | { | |
f8c76f6f YS |
365 | struct ceph_mon_generic_request *req; |
366 | struct rb_node *n = monc->generic_request_tree.rb_node; | |
85ff03f6 SW |
367 | |
368 | while (n) { | |
f8c76f6f | 369 | req = rb_entry(n, struct ceph_mon_generic_request, node); |
85ff03f6 SW |
370 | if (tid < req->tid) |
371 | n = n->rb_left; | |
372 | else if (tid > req->tid) | |
373 | n = n->rb_right; | |
374 | else | |
375 | return req; | |
376 | } | |
377 | return NULL; | |
378 | } | |
379 | ||
f8c76f6f YS |
380 | static void __insert_generic_request(struct ceph_mon_client *monc, |
381 | struct ceph_mon_generic_request *new) | |
85ff03f6 | 382 | { |
f8c76f6f | 383 | struct rb_node **p = &monc->generic_request_tree.rb_node; |
85ff03f6 | 384 | struct rb_node *parent = NULL; |
f8c76f6f | 385 | struct ceph_mon_generic_request *req = NULL; |
85ff03f6 SW |
386 | |
387 | while (*p) { | |
388 | parent = *p; | |
f8c76f6f | 389 | req = rb_entry(parent, struct ceph_mon_generic_request, node); |
85ff03f6 SW |
390 | if (new->tid < req->tid) |
391 | p = &(*p)->rb_left; | |
392 | else if (new->tid > req->tid) | |
393 | p = &(*p)->rb_right; | |
394 | else | |
395 | BUG(); | |
396 | } | |
397 | ||
398 | rb_link_node(&new->node, parent, p); | |
f8c76f6f | 399 | rb_insert_color(&new->node, &monc->generic_request_tree); |
85ff03f6 SW |
400 | } |
401 | ||
f8c76f6f | 402 | static void release_generic_request(struct kref *kref) |
3143edd3 | 403 | { |
f8c76f6f YS |
404 | struct ceph_mon_generic_request *req = |
405 | container_of(kref, struct ceph_mon_generic_request, kref); | |
3143edd3 SW |
406 | |
407 | if (req->reply) | |
408 | ceph_msg_put(req->reply); | |
409 | if (req->request) | |
410 | ceph_msg_put(req->request); | |
20547567 YS |
411 | |
412 | kfree(req); | |
3143edd3 SW |
413 | } |
414 | ||
f8c76f6f | 415 | static void put_generic_request(struct ceph_mon_generic_request *req) |
3143edd3 | 416 | { |
f8c76f6f | 417 | kref_put(&req->kref, release_generic_request); |
3143edd3 SW |
418 | } |
419 | ||
f8c76f6f | 420 | static void get_generic_request(struct ceph_mon_generic_request *req) |
3143edd3 SW |
421 | { |
422 | kref_get(&req->kref); | |
423 | } | |
424 | ||
f8c76f6f | 425 | static struct ceph_msg *get_generic_reply(struct ceph_connection *con, |
3143edd3 SW |
426 | struct ceph_msg_header *hdr, |
427 | int *skip) | |
428 | { | |
429 | struct ceph_mon_client *monc = con->private; | |
f8c76f6f | 430 | struct ceph_mon_generic_request *req; |
3143edd3 SW |
431 | u64 tid = le64_to_cpu(hdr->tid); |
432 | struct ceph_msg *m; | |
433 | ||
434 | mutex_lock(&monc->mutex); | |
f8c76f6f | 435 | req = __lookup_generic_req(monc, tid); |
3143edd3 | 436 | if (!req) { |
f8c76f6f | 437 | dout("get_generic_reply %lld dne\n", tid); |
3143edd3 SW |
438 | *skip = 1; |
439 | m = NULL; | |
440 | } else { | |
f8c76f6f | 441 | dout("get_generic_reply %lld got %p\n", tid, req->reply); |
3143edd3 SW |
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 | ||
e56fa10e YS |
453 | static 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 | */ | |
ba75bb98 SW |
482 | static void handle_statfs_reply(struct ceph_mon_client *monc, |
483 | struct ceph_msg *msg) | |
484 | { | |
f8c76f6f | 485 | struct ceph_mon_generic_request *req; |
ba75bb98 | 486 | struct ceph_mon_statfs_reply *reply = msg->front.iov_base; |
3143edd3 | 487 | u64 tid = le64_to_cpu(msg->hdr.tid); |
ba75bb98 SW |
488 | |
489 | if (msg->front.iov_len != sizeof(*reply)) | |
490 | goto bad; | |
ba75bb98 SW |
491 | dout("handle_statfs_reply %p tid %llu\n", msg, tid); |
492 | ||
493 | mutex_lock(&monc->mutex); | |
f8c76f6f | 494 | req = __lookup_generic_req(monc, tid); |
ba75bb98 | 495 | if (req) { |
f8c76f6f | 496 | *(struct ceph_statfs *)req->buf = reply->st; |
ba75bb98 | 497 | req->result = 0; |
f8c76f6f | 498 | get_generic_request(req); |
ba75bb98 SW |
499 | } |
500 | mutex_unlock(&monc->mutex); | |
3143edd3 | 501 | if (req) { |
03066f23 | 502 | complete_all(&req->completion); |
f8c76f6f | 503 | put_generic_request(req); |
3143edd3 | 504 | } |
ba75bb98 SW |
505 | return; |
506 | ||
507 | bad: | |
e56fa10e | 508 | pr_err("corrupt generic reply, tid %llu\n", tid); |
9ec7cab1 | 509 | ceph_msg_dump(msg); |
ba75bb98 SW |
510 | } |
511 | ||
512 | /* | |
3143edd3 | 513 | * Do a synchronous statfs(). |
ba75bb98 | 514 | */ |
3143edd3 | 515 | int ceph_monc_do_statfs(struct ceph_mon_client *monc, struct ceph_statfs *buf) |
ba75bb98 | 516 | { |
f8c76f6f | 517 | struct ceph_mon_generic_request *req; |
ba75bb98 | 518 | struct ceph_mon_statfs *h; |
3143edd3 SW |
519 | int err; |
520 | ||
cffe7b6d | 521 | req = kzalloc(sizeof(*req), GFP_NOFS); |
3143edd3 SW |
522 | if (!req) |
523 | return -ENOMEM; | |
524 | ||
3143edd3 SW |
525 | kref_init(&req->kref); |
526 | req->buf = buf; | |
e56fa10e | 527 | req->buf_len = sizeof(*buf); |
3143edd3 | 528 | init_completion(&req->completion); |
ba75bb98 | 529 | |
a79832f2 | 530 | err = -ENOMEM; |
b61c2763 SW |
531 | req->request = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), GFP_NOFS, |
532 | true); | |
a79832f2 | 533 | if (!req->request) |
3143edd3 | 534 | goto out; |
b61c2763 SW |
535 | req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 1024, GFP_NOFS, |
536 | true); | |
a79832f2 | 537 | if (!req->reply) |
3143edd3 | 538 | goto out; |
3143edd3 SW |
539 | |
540 | /* fill out request */ | |
541 | h = req->request->front.iov_base; | |
13e38c8a SW |
542 | h->monhdr.have_version = 0; |
543 | h->monhdr.session_mon = cpu_to_le16(-1); | |
544 | h->monhdr.session_mon_tid = 0; | |
ba75bb98 | 545 | h->fsid = monc->monmap->fsid; |
ba75bb98 | 546 | |
e56fa10e | 547 | err = do_generic_request(monc, req); |
ba75bb98 | 548 | |
e56fa10e YS |
549 | out: |
550 | kref_put(&req->kref, release_generic_request); | |
551 | return err; | |
552 | } | |
3d14c5d2 | 553 | EXPORT_SYMBOL(ceph_monc_do_statfs); |
e56fa10e YS |
554 | |
555 | /* | |
556 | * pool ops | |
557 | */ | |
558 | static 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 | ||
574 | static 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); | |
ba75bb98 SW |
584 | |
585 | mutex_lock(&monc->mutex); | |
e56fa10e YS |
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 | } | |
ba75bb98 | 598 | mutex_unlock(&monc->mutex); |
e56fa10e YS |
599 | if (req) { |
600 | complete(&req->completion); | |
601 | put_generic_request(req); | |
602 | } | |
603 | return; | |
ba75bb98 | 604 | |
e56fa10e YS |
605 | bad: |
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 | */ | |
613 | int 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; | |
b61c2763 SW |
631 | req->request = ceph_msg_new(CEPH_MSG_POOLOP, sizeof(*h), GFP_NOFS, |
632 | true); | |
e56fa10e YS |
633 | if (!req->request) |
634 | goto out; | |
b61c2763 SW |
635 | req->reply = ceph_msg_new(CEPH_MSG_POOLOP_REPLY, 1024, GFP_NOFS, |
636 | true); | |
e56fa10e YS |
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); | |
3143edd3 SW |
654 | |
655 | out: | |
f8c76f6f | 656 | kref_put(&req->kref, release_generic_request); |
ba75bb98 SW |
657 | return err; |
658 | } | |
659 | ||
e56fa10e YS |
660 | int 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 | } | |
3d14c5d2 | 667 | EXPORT_SYMBOL(ceph_monc_create_snapid); |
e56fa10e YS |
668 | |
669 | int 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 | ||
ba75bb98 | 677 | /* |
e56fa10e | 678 | * Resend pending generic requests. |
ba75bb98 | 679 | */ |
f8c76f6f | 680 | static void __resend_generic_request(struct ceph_mon_client *monc) |
ba75bb98 | 681 | { |
f8c76f6f | 682 | struct ceph_mon_generic_request *req; |
85ff03f6 | 683 | struct rb_node *p; |
ba75bb98 | 684 | |
f8c76f6f YS |
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); | |
97069001 | 687 | ceph_con_revoke(monc->con, req->request); |
3143edd3 | 688 | ceph_con_send(monc->con, ceph_msg_get(req->request)); |
ba75bb98 SW |
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 | */ | |
697 | static 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); | |
4e7a5dcd SW |
704 | if (monc->hunting) { |
705 | __close_session(monc); | |
706 | __open_session(monc); /* continue hunting */ | |
ba75bb98 | 707 | } else { |
4e7a5dcd | 708 | ceph_con_keepalive(monc->con); |
9bd2e6f8 SW |
709 | |
710 | __validate_auth(monc); | |
711 | ||
4e7a5dcd SW |
712 | if (monc->auth->ops->is_authenticated(monc->auth)) |
713 | __send_subscribe(monc); | |
ba75bb98 | 714 | } |
ba75bb98 SW |
715 | __schedule_delayed(monc); |
716 | mutex_unlock(&monc->mutex); | |
717 | } | |
718 | ||
6b805185 SW |
719 | /* |
720 | * On startup, we build a temporary monmap populated with the IPs | |
721 | * provided by mount(2). | |
722 | */ | |
723 | static int build_initial_monmap(struct ceph_mon_client *monc) | |
724 | { | |
3d14c5d2 YS |
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; | |
6b805185 SW |
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]; | |
6b805185 SW |
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; | |
4e7a5dcd | 744 | monc->have_fsid = false; |
6b805185 SW |
745 | return 0; |
746 | } | |
747 | ||
ba75bb98 SW |
748 | int 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 | ||
6b805185 SW |
758 | err = build_initial_monmap(monc); |
759 | if (err) | |
760 | goto out; | |
761 | ||
f6a2f5be SW |
762 | /* connection */ |
763 | monc->con = kmalloc(sizeof(*monc->con), GFP_KERNEL); | |
764 | if (!monc->con) | |
765 | goto out_monmap; | |
15d9882c | 766 | ceph_con_init(&monc->client->msgr, monc->con); |
f6a2f5be SW |
767 | monc->con->private = monc; |
768 | monc->con->ops = &mon_con_ops; | |
ba75bb98 | 769 | |
4e7a5dcd | 770 | /* authentication */ |
3d14c5d2 | 771 | monc->auth = ceph_auth_init(cl->options->name, |
8323c3aa | 772 | cl->options->key); |
49d9224c NW |
773 | if (IS_ERR(monc->auth)) { |
774 | err = PTR_ERR(monc->auth); | |
775 | goto out_con; | |
776 | } | |
4e7a5dcd SW |
777 | monc->auth->want_keys = |
778 | CEPH_ENTITY_TYPE_AUTH | CEPH_ENTITY_TYPE_MON | | |
779 | CEPH_ENTITY_TYPE_OSD | CEPH_ENTITY_TYPE_MDS; | |
780 | ||
240ed68e | 781 | /* msgs */ |
a79832f2 | 782 | err = -ENOMEM; |
7c315c55 | 783 | monc->m_subscribe_ack = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE_ACK, |
34d23762 | 784 | sizeof(struct ceph_mon_subscribe_ack), |
b61c2763 | 785 | GFP_NOFS, true); |
a79832f2 | 786 | if (!monc->m_subscribe_ack) |
49d9224c | 787 | goto out_auth; |
6694d6b9 | 788 | |
b61c2763 SW |
789 | monc->m_subscribe = ceph_msg_new(CEPH_MSG_MON_SUBSCRIBE, 96, GFP_NOFS, |
790 | true); | |
240ed68e SW |
791 | if (!monc->m_subscribe) |
792 | goto out_subscribe_ack; | |
793 | ||
b61c2763 SW |
794 | monc->m_auth_reply = ceph_msg_new(CEPH_MSG_AUTH_REPLY, 4096, GFP_NOFS, |
795 | true); | |
a79832f2 | 796 | if (!monc->m_auth_reply) |
240ed68e | 797 | goto out_subscribe; |
4e7a5dcd | 798 | |
b61c2763 | 799 | monc->m_auth = ceph_msg_new(CEPH_MSG_AUTH, 4096, GFP_NOFS, true); |
9bd2e6f8 | 800 | monc->pending_auth = 0; |
a79832f2 | 801 | if (!monc->m_auth) |
6694d6b9 | 802 | goto out_auth_reply; |
ba75bb98 SW |
803 | |
804 | monc->cur_mon = -1; | |
4e7a5dcd | 805 | monc->hunting = true; |
ba75bb98 SW |
806 | monc->sub_renew_after = jiffies; |
807 | monc->sub_sent = 0; | |
808 | ||
809 | INIT_DELAYED_WORK(&monc->delayed_work, delayed_work); | |
f8c76f6f YS |
810 | monc->generic_request_tree = RB_ROOT; |
811 | monc->num_generic_requests = 0; | |
ba75bb98 SW |
812 | monc->last_tid = 0; |
813 | ||
814 | monc->have_mdsmap = 0; | |
815 | monc->have_osdmap = 0; | |
816 | monc->want_next_osdmap = 1; | |
4e7a5dcd SW |
817 | return 0; |
818 | ||
6694d6b9 SW |
819 | out_auth_reply: |
820 | ceph_msg_put(monc->m_auth_reply); | |
240ed68e SW |
821 | out_subscribe: |
822 | ceph_msg_put(monc->m_subscribe); | |
7c315c55 SW |
823 | out_subscribe_ack: |
824 | ceph_msg_put(monc->m_subscribe_ack); | |
49d9224c NW |
825 | out_auth: |
826 | ceph_auth_destroy(monc->auth); | |
f6a2f5be SW |
827 | out_con: |
828 | monc->con->ops->put(monc->con); | |
4e7a5dcd SW |
829 | out_monmap: |
830 | kfree(monc->monmap); | |
ba75bb98 SW |
831 | out: |
832 | return err; | |
833 | } | |
3d14c5d2 | 834 | EXPORT_SYMBOL(ceph_monc_init); |
ba75bb98 SW |
835 | |
836 | void ceph_monc_stop(struct ceph_mon_client *monc) | |
837 | { | |
838 | dout("stop\n"); | |
839 | cancel_delayed_work_sync(&monc->delayed_work); | |
840 | ||
841 | mutex_lock(&monc->mutex); | |
842 | __close_session(monc); | |
f6a2f5be SW |
843 | |
844 | monc->con->private = NULL; | |
845 | monc->con->ops->put(monc->con); | |
846 | monc->con = NULL; | |
847 | ||
ba75bb98 SW |
848 | mutex_unlock(&monc->mutex); |
849 | ||
4e7a5dcd SW |
850 | ceph_auth_destroy(monc->auth); |
851 | ||
852 | ceph_msg_put(monc->m_auth); | |
6694d6b9 | 853 | ceph_msg_put(monc->m_auth_reply); |
240ed68e | 854 | ceph_msg_put(monc->m_subscribe); |
7c315c55 | 855 | ceph_msg_put(monc->m_subscribe_ack); |
ba75bb98 SW |
856 | |
857 | kfree(monc->monmap); | |
858 | } | |
3d14c5d2 | 859 | EXPORT_SYMBOL(ceph_monc_stop); |
ba75bb98 | 860 | |
4e7a5dcd SW |
861 | static void handle_auth_reply(struct ceph_mon_client *monc, |
862 | struct ceph_msg *msg) | |
863 | { | |
864 | int ret; | |
09c4d6a7 | 865 | int was_auth = 0; |
4e7a5dcd SW |
866 | |
867 | mutex_lock(&monc->mutex); | |
09c4d6a7 SW |
868 | if (monc->auth->ops) |
869 | was_auth = monc->auth->ops->is_authenticated(monc->auth); | |
9bd2e6f8 | 870 | monc->pending_auth = 0; |
4e7a5dcd SW |
871 | ret = ceph_handle_auth_reply(monc->auth, msg->front.iov_base, |
872 | msg->front.iov_len, | |
873 | monc->m_auth->front.iov_base, | |
874 | monc->m_auth->front_max); | |
875 | if (ret < 0) { | |
9bd2e6f8 | 876 | monc->client->auth_err = ret; |
03066f23 | 877 | wake_up_all(&monc->client->auth_wq); |
4e7a5dcd | 878 | } else if (ret > 0) { |
9bd2e6f8 | 879 | __send_prepared_auth_request(monc, ret); |
09c4d6a7 | 880 | } else if (!was_auth && monc->auth->ops->is_authenticated(monc->auth)) { |
4e7a5dcd | 881 | dout("authenticated, starting session\n"); |
0743304d | 882 | |
15d9882c AE |
883 | monc->client->msgr.inst.name.type = CEPH_ENTITY_TYPE_CLIENT; |
884 | monc->client->msgr.inst.name.num = | |
0cf5537b | 885 | cpu_to_le64(monc->auth->global_id); |
0743304d | 886 | |
4e7a5dcd | 887 | __send_subscribe(monc); |
f8c76f6f | 888 | __resend_generic_request(monc); |
4e7a5dcd SW |
889 | } |
890 | mutex_unlock(&monc->mutex); | |
891 | } | |
892 | ||
9bd2e6f8 SW |
893 | static int __validate_auth(struct ceph_mon_client *monc) |
894 | { | |
895 | int ret; | |
896 | ||
897 | if (monc->pending_auth) | |
898 | return 0; | |
899 | ||
900 | ret = ceph_build_auth(monc->auth, monc->m_auth->front.iov_base, | |
901 | monc->m_auth->front_max); | |
902 | if (ret <= 0) | |
903 | return ret; /* either an error, or no need to authenticate */ | |
904 | __send_prepared_auth_request(monc, ret); | |
905 | return 0; | |
906 | } | |
907 | ||
908 | int ceph_monc_validate_auth(struct ceph_mon_client *monc) | |
909 | { | |
910 | int ret; | |
911 | ||
912 | mutex_lock(&monc->mutex); | |
913 | ret = __validate_auth(monc); | |
914 | mutex_unlock(&monc->mutex); | |
915 | return ret; | |
916 | } | |
3d14c5d2 | 917 | EXPORT_SYMBOL(ceph_monc_validate_auth); |
9bd2e6f8 | 918 | |
ba75bb98 SW |
919 | /* |
920 | * handle incoming message | |
921 | */ | |
922 | static void dispatch(struct ceph_connection *con, struct ceph_msg *msg) | |
923 | { | |
924 | struct ceph_mon_client *monc = con->private; | |
925 | int type = le16_to_cpu(msg->hdr.type); | |
926 | ||
927 | if (!monc) | |
928 | return; | |
929 | ||
930 | switch (type) { | |
4e7a5dcd SW |
931 | case CEPH_MSG_AUTH_REPLY: |
932 | handle_auth_reply(monc, msg); | |
ba75bb98 SW |
933 | break; |
934 | ||
935 | case CEPH_MSG_MON_SUBSCRIBE_ACK: | |
936 | handle_subscribe_ack(monc, msg); | |
937 | break; | |
938 | ||
939 | case CEPH_MSG_STATFS_REPLY: | |
940 | handle_statfs_reply(monc, msg); | |
941 | break; | |
942 | ||
e56fa10e YS |
943 | case CEPH_MSG_POOLOP_REPLY: |
944 | handle_poolop_reply(monc, msg); | |
945 | break; | |
946 | ||
4e7a5dcd SW |
947 | case CEPH_MSG_MON_MAP: |
948 | ceph_monc_handle_map(monc, msg); | |
949 | break; | |
950 | ||
ba75bb98 SW |
951 | case CEPH_MSG_OSD_MAP: |
952 | ceph_osdc_handle_map(&monc->client->osdc, msg); | |
953 | break; | |
954 | ||
955 | default: | |
3d14c5d2 YS |
956 | /* can the chained handler handle it? */ |
957 | if (monc->client->extra_mon_dispatch && | |
958 | monc->client->extra_mon_dispatch(monc->client, msg) == 0) | |
959 | break; | |
960 | ||
ba75bb98 SW |
961 | pr_err("received unknown message type %d %s\n", type, |
962 | ceph_msg_type_name(type)); | |
963 | } | |
964 | ceph_msg_put(msg); | |
965 | } | |
966 | ||
967 | /* | |
968 | * Allocate memory for incoming message | |
969 | */ | |
970 | static struct ceph_msg *mon_alloc_msg(struct ceph_connection *con, | |
2450418c YS |
971 | struct ceph_msg_header *hdr, |
972 | int *skip) | |
ba75bb98 SW |
973 | { |
974 | struct ceph_mon_client *monc = con->private; | |
975 | int type = le16_to_cpu(hdr->type); | |
2450418c | 976 | int front_len = le32_to_cpu(hdr->front_len); |
5b3a4db3 | 977 | struct ceph_msg *m = NULL; |
ba75bb98 | 978 | |
2450418c | 979 | *skip = 0; |
0547a9b3 | 980 | |
ba75bb98 | 981 | switch (type) { |
ba75bb98 | 982 | case CEPH_MSG_MON_SUBSCRIBE_ACK: |
7c315c55 | 983 | m = ceph_msg_get(monc->m_subscribe_ack); |
2450418c | 984 | break; |
e56fa10e | 985 | case CEPH_MSG_POOLOP_REPLY: |
ba75bb98 | 986 | case CEPH_MSG_STATFS_REPLY: |
f8c76f6f | 987 | return get_generic_reply(con, hdr, skip); |
4e7a5dcd | 988 | case CEPH_MSG_AUTH_REPLY: |
6694d6b9 | 989 | m = ceph_msg_get(monc->m_auth_reply); |
2450418c | 990 | break; |
5b3a4db3 SW |
991 | case CEPH_MSG_MON_MAP: |
992 | case CEPH_MSG_MDS_MAP: | |
993 | case CEPH_MSG_OSD_MAP: | |
b61c2763 | 994 | m = ceph_msg_new(type, front_len, GFP_NOFS, false); |
5b3a4db3 | 995 | break; |
ba75bb98 | 996 | } |
2450418c | 997 | |
5b3a4db3 SW |
998 | if (!m) { |
999 | pr_info("alloc_msg unknown type %d\n", type); | |
2450418c | 1000 | *skip = 1; |
5b3a4db3 | 1001 | } |
2450418c | 1002 | return m; |
ba75bb98 SW |
1003 | } |
1004 | ||
1005 | /* | |
1006 | * If the monitor connection resets, pick a new monitor and resubmit | |
1007 | * any pending requests. | |
1008 | */ | |
1009 | static 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 | ||
f6a2f5be | 1021 | if (!monc->hunting) |
ba75bb98 SW |
1022 | pr_info("mon%d %s session lost, " |
1023 | "hunting for new mon\n", monc->cur_mon, | |
3d14c5d2 | 1024 | ceph_pr_addr(&monc->con->peer_addr.in_addr)); |
ba75bb98 SW |
1025 | |
1026 | __close_session(monc); | |
1027 | if (!monc->hunting) { | |
1028 | /* start hunting */ | |
1029 | monc->hunting = true; | |
4e7a5dcd | 1030 | __open_session(monc); |
ba75bb98 SW |
1031 | } else { |
1032 | /* already hunting, let's wait a bit */ | |
1033 | __schedule_delayed(monc); | |
1034 | } | |
1035 | out: | |
1036 | mutex_unlock(&monc->mutex); | |
1037 | } | |
1038 | ||
9e32789f | 1039 | static const struct ceph_connection_operations mon_con_ops = { |
ba75bb98 SW |
1040 | .get = ceph_con_get, |
1041 | .put = ceph_con_put, | |
1042 | .dispatch = dispatch, | |
1043 | .fault = mon_fault, | |
1044 | .alloc_msg = mon_alloc_msg, | |
ba75bb98 | 1045 | }; |