?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.2.1
[linux/fpc-iii.git] / fs / ceph / mds_client.c
blob6aa07af67603ada211f49268d3845ea62b625720
1 #include <linux/ceph/ceph_debug.h>
2
3 #include <linux/fs.h>
4 #include <linux/wait.h>
5 #include <linux/slab.h>
6 #include <linux/gfp.h>
7 #include <linux/sched.h>
8 #include <linux/debugfs.h>
9 #include <linux/seq_file.h>
10 #include <linux/utsname.h>
11 #include <linux/ratelimit.h>
12
13 #include "super.h"
14 #include "mds_client.h"
15
16 #include <linux/ceph/ceph_features.h>
17 #include <linux/ceph/messenger.h>
18 #include <linux/ceph/decode.h>
19 #include <linux/ceph/pagelist.h>
20 #include <linux/ceph/auth.h>
21 #include <linux/ceph/debugfs.h>
22
23 /*
24 * A cluster of MDS (metadata server) daemons is responsible for
25 * managing the file system namespace (the directory hierarchy and
26 * inodes) and for coordinating shared access to storage. Metadata is
27 * partitioning hierarchically across a number of servers, and that
28 * partition varies over time as the cluster adjusts the distribution
29 * in order to balance load.
30 *
31 * The MDS client is primarily responsible to managing synchronous
32 * metadata requests for operations like open, unlink, and so forth.
33 * If there is a MDS failure, we find out about it when we (possibly
34 * request and) receive a new MDS map, and can resubmit affected
35 * requests.
36 *
37 * For the most part, though, we take advantage of a lossless
38 * communications channel to the MDS, and do not need to worry about
39 * timing out or resubmitting requests.
40 *
41 * We maintain a stateful "session" with each MDS we interact with.
42 * Within each session, we sent periodic heartbeat messages to ensure
43 * any capabilities or leases we have been issues remain valid. If
44 * the session times out and goes stale, our leases and capabilities
45 * are no longer valid.
46 */
47
48 struct ceph_reconnect_state {
49 int nr_caps;
50 struct ceph_pagelist *pagelist;
51 bool flock;
52 };
53
54 static void __wake_requests(struct ceph_mds_client *mdsc,
55 struct list_head *head);
56
57 static const struct ceph_connection_operations mds_con_ops;
58
59
60 /*
61 * mds reply parsing
62 */
63
64 /*
65 * parse individual inode info
66 */
67 static int parse_reply_info_in(void **p, void *end,
68 struct ceph_mds_reply_info_in *info,
69 u64 features)
70 {
71 int err = -EIO;
72
73 info->in = *p;
74 *p += sizeof(struct ceph_mds_reply_inode) +
75 sizeof(*info->in->fragtree.splits) *
76 le32_to_cpu(info->in->fragtree.nsplits);
77
78 ceph_decode_32_safe(p, end, info->symlink_len, bad);
79 ceph_decode_need(p, end, info->symlink_len, bad);
80 info->symlink = *p;
81 *p += info->symlink_len;
82
83 if (features & CEPH_FEATURE_DIRLAYOUTHASH)
84 ceph_decode_copy_safe(p, end, &info->dir_layout,
85 sizeof(info->dir_layout), bad);
86 else
87 memset(&info->dir_layout, 0, sizeof(info->dir_layout));
88
89 ceph_decode_32_safe(p, end, info->xattr_len, bad);
90 ceph_decode_need(p, end, info->xattr_len, bad);
91 info->xattr_data = *p;
92 *p += info->xattr_len;
93
94 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
95 ceph_decode_64_safe(p, end, info->inline_version, bad);
96 ceph_decode_32_safe(p, end, info->inline_len, bad);
97 ceph_decode_need(p, end, info->inline_len, bad);
98 info->inline_data = *p;
99 *p += info->inline_len;
100 } else
101 info->inline_version = CEPH_INLINE_NONE;
102
103 return 0;
104 bad:
105 return err;
106 }
107
108 /*
109 * parse a normal reply, which may contain a (dir+)dentry and/or a
110 * target inode.
111 */
112 static int parse_reply_info_trace(void **p, void *end,
113 struct ceph_mds_reply_info_parsed *info,
114 u64 features)
115 {
116 int err;
117
118 if (info->head->is_dentry) {
119 err = parse_reply_info_in(p, end, &info->diri, features);
120 if (err < 0)
121 goto out_bad;
122
123 if (unlikely(*p + sizeof(*info->dirfrag) > end))
124 goto bad;
125 info->dirfrag = *p;
126 *p += sizeof(*info->dirfrag) +
127 sizeof(u32)*le32_to_cpu(info->dirfrag->ndist);
128 if (unlikely(*p > end))
129 goto bad;
130
131 ceph_decode_32_safe(p, end, info->dname_len, bad);
132 ceph_decode_need(p, end, info->dname_len, bad);
133 info->dname = *p;
134 *p += info->dname_len;
135 info->dlease = *p;
136 *p += sizeof(*info->dlease);
137 }
138
139 if (info->head->is_target) {
140 err = parse_reply_info_in(p, end, &info->targeti, features);
141 if (err < 0)
142 goto out_bad;
143 }
144
145 if (unlikely(*p != end))
146 goto bad;
147 return 0;
148
149 bad:
150 err = -EIO;
151 out_bad:
152 pr_err("problem parsing mds trace %d\n", err);
153 return err;
154 }
155
156 /*
157 * parse readdir results
158 */
159 static int parse_reply_info_dir(void **p, void *end,
160 struct ceph_mds_reply_info_parsed *info,
161 u64 features)
162 {
163 u32 num, i = 0;
164 int err;
165
166 info->dir_dir = *p;
167 if (*p + sizeof(*info->dir_dir) > end)
168 goto bad;
169 *p += sizeof(*info->dir_dir) +
170 sizeof(u32)*le32_to_cpu(info->dir_dir->ndist);
171 if (*p > end)
172 goto bad;
173
174 ceph_decode_need(p, end, sizeof(num) + 2, bad);
175 num = ceph_decode_32(p);
176 info->dir_end = ceph_decode_8(p);
177 info->dir_complete = ceph_decode_8(p);
178 if (num == 0)
179 goto done;
180
181 BUG_ON(!info->dir_in);
182 info->dir_dname = (void *)(info->dir_in + num);
183 info->dir_dname_len = (void *)(info->dir_dname + num);
184 info->dir_dlease = (void *)(info->dir_dname_len + num);
185 if ((unsigned long)(info->dir_dlease + num) >
186 (unsigned long)info->dir_in + info->dir_buf_size) {
187 pr_err("dir contents are larger than expected\n");
188 WARN_ON(1);
189 goto bad;
190 }
191
192 info->dir_nr = num;
193 while (num) {
194 /* dentry */
195 ceph_decode_need(p, end, sizeof(u32)*2, bad);
196 info->dir_dname_len[i] = ceph_decode_32(p);
197 ceph_decode_need(p, end, info->dir_dname_len[i], bad);
198 info->dir_dname[i] = *p;
199 *p += info->dir_dname_len[i];
200 dout("parsed dir dname '%.*s'\n", info->dir_dname_len[i],
201 info->dir_dname[i]);
202 info->dir_dlease[i] = *p;
203 *p += sizeof(struct ceph_mds_reply_lease);
204
205 /* inode */
206 err = parse_reply_info_in(p, end, &info->dir_in[i], features);
207 if (err < 0)
208 goto out_bad;
209 i++;
210 num--;
211 }
212
213 done:
214 if (*p != end)
215 goto bad;
216 return 0;
217
218 bad:
219 err = -EIO;
220 out_bad:
221 pr_err("problem parsing dir contents %d\n", err);
222 return err;
223 }
224
225 /*
226 * parse fcntl F_GETLK results
227 */
228 static int parse_reply_info_filelock(void **p, void *end,
229 struct ceph_mds_reply_info_parsed *info,
230 u64 features)
231 {
232 if (*p + sizeof(*info->filelock_reply) > end)
233 goto bad;
234
235 info->filelock_reply = *p;
236 *p += sizeof(*info->filelock_reply);
237
238 if (unlikely(*p != end))
239 goto bad;
240 return 0;
241
242 bad:
243 return -EIO;
244 }
245
246 /*
247 * parse create results
248 */
249 static int parse_reply_info_create(void **p, void *end,
250 struct ceph_mds_reply_info_parsed *info,
251 u64 features)
252 {
253 if (features & CEPH_FEATURE_REPLY_CREATE_INODE) {
254 if (*p == end) {
255 info->has_create_ino = false;
256 } else {
257 info->has_create_ino = true;
258 info->ino = ceph_decode_64(p);
259 }
260 }
261
262 if (unlikely(*p != end))
263 goto bad;
264 return 0;
265
266 bad:
267 return -EIO;
268 }
269
270 /*
271 * parse extra results
272 */
273 static int parse_reply_info_extra(void **p, void *end,
274 struct ceph_mds_reply_info_parsed *info,
275 u64 features)
276 {
277 if (info->head->op == CEPH_MDS_OP_GETFILELOCK)
278 return parse_reply_info_filelock(p, end, info, features);
279 else if (info->head->op == CEPH_MDS_OP_READDIR ||
280 info->head->op == CEPH_MDS_OP_LSSNAP)
281 return parse_reply_info_dir(p, end, info, features);
282 else if (info->head->op == CEPH_MDS_OP_CREATE)
283 return parse_reply_info_create(p, end, info, features);
284 else
285 return -EIO;
286 }
287
288 /*
289 * parse entire mds reply
290 */
291 static int parse_reply_info(struct ceph_msg *msg,
292 struct ceph_mds_reply_info_parsed *info,
293 u64 features)
294 {
295 void *p, *end;
296 u32 len;
297 int err;
298
299 info->head = msg->front.iov_base;
300 p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
301 end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
302
303 /* trace */
304 ceph_decode_32_safe(&p, end, len, bad);
305 if (len > 0) {
306 ceph_decode_need(&p, end, len, bad);
307 err = parse_reply_info_trace(&p, p+len, info, features);
308 if (err < 0)
309 goto out_bad;
310 }
311
312 /* extra */
313 ceph_decode_32_safe(&p, end, len, bad);
314 if (len > 0) {
315 ceph_decode_need(&p, end, len, bad);
316 err = parse_reply_info_extra(&p, p+len, info, features);
317 if (err < 0)
318 goto out_bad;
319 }
320
321 /* snap blob */
322 ceph_decode_32_safe(&p, end, len, bad);
323 info->snapblob_len = len;
324 info->snapblob = p;
325 p += len;
326
327 if (p != end)
328 goto bad;
329 return 0;
330
331 bad:
332 err = -EIO;
333 out_bad:
334 pr_err("mds parse_reply err %d\n", err);
335 return err;
336 }
337
338 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
339 {
340 if (!info->dir_in)
341 return;
342 free_pages((unsigned long)info->dir_in, get_order(info->dir_buf_size));
343 }
344
345
346 /*
347 * sessions
348 */
349 const char *ceph_session_state_name(int s)
350 {
351 switch (s) {
352 case CEPH_MDS_SESSION_NEW: return "new";
353 case CEPH_MDS_SESSION_OPENING: return "opening";
354 case CEPH_MDS_SESSION_OPEN: return "open";
355 case CEPH_MDS_SESSION_HUNG: return "hung";
356 case CEPH_MDS_SESSION_CLOSING: return "closing";
357 case CEPH_MDS_SESSION_RESTARTING: return "restarting";
358 case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
359 default: return "???";
360 }
361 }
362
363 static struct ceph_mds_session *get_session(struct ceph_mds_session *s)
364 {
365 if (atomic_inc_not_zero(&s->s_ref)) {
366 dout("mdsc get_session %p %d -> %d\n", s,
367 atomic_read(&s->s_ref)-1, atomic_read(&s->s_ref));
368 return s;
369 } else {
370 dout("mdsc get_session %p 0 -- FAIL", s);
371 return NULL;
372 }
373 }
374
375 void ceph_put_mds_session(struct ceph_mds_session *s)
376 {
377 dout("mdsc put_session %p %d -> %d\n", s,
378 atomic_read(&s->s_ref), atomic_read(&s->s_ref)-1);
379 if (atomic_dec_and_test(&s->s_ref)) {
380 if (s->s_auth.authorizer)
381 ceph_auth_destroy_authorizer(
382 s->s_mdsc->fsc->client->monc.auth,
383 s->s_auth.authorizer);
384 kfree(s);
385 }
386 }
387
388 /*
389 * called under mdsc->mutex
390 */
391 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
392 int mds)
393 {
394 struct ceph_mds_session *session;
395
396 if (mds >= mdsc->max_sessions || mdsc->sessions[mds] == NULL)
397 return NULL;
398 session = mdsc->sessions[mds];
399 dout("lookup_mds_session %p %d\n", session,
400 atomic_read(&session->s_ref));
401 get_session(session);
402 return session;
403 }
404
405 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
406 {
407 if (mds >= mdsc->max_sessions)
408 return false;
409 return mdsc->sessions[mds];
410 }
411
412 static int __verify_registered_session(struct ceph_mds_client *mdsc,
413 struct ceph_mds_session *s)
414 {
415 if (s->s_mds >= mdsc->max_sessions ||
416 mdsc->sessions[s->s_mds] != s)
417 return -ENOENT;
418 return 0;
419 }
420
421 /*
422 * create+register a new session for given mds.
423 * called under mdsc->mutex.
424 */
425 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
426 int mds)
427 {
428 struct ceph_mds_session *s;
429
430 if (mds >= mdsc->mdsmap->m_max_mds)
431 return ERR_PTR(-EINVAL);
432
433 s = kzalloc(sizeof(*s), GFP_NOFS);
434 if (!s)
435 return ERR_PTR(-ENOMEM);
436 s->s_mdsc = mdsc;
437 s->s_mds = mds;
438 s->s_state = CEPH_MDS_SESSION_NEW;
439 s->s_ttl = 0;
440 s->s_seq = 0;
441 mutex_init(&s->s_mutex);
442
443 ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
444
445 spin_lock_init(&s->s_gen_ttl_lock);
446 s->s_cap_gen = 0;
447 s->s_cap_ttl = jiffies - 1;
448
449 spin_lock_init(&s->s_cap_lock);
450 s->s_renew_requested = 0;
451 s->s_renew_seq = 0;
452 INIT_LIST_HEAD(&s->s_caps);
453 s->s_nr_caps = 0;
454 s->s_trim_caps = 0;
455 atomic_set(&s->s_ref, 1);
456 INIT_LIST_HEAD(&s->s_waiting);
457 INIT_LIST_HEAD(&s->s_unsafe);
458 s->s_num_cap_releases = 0;
459 s->s_cap_reconnect = 0;
460 s->s_cap_iterator = NULL;
461 INIT_LIST_HEAD(&s->s_cap_releases);
462 INIT_LIST_HEAD(&s->s_cap_flushing);
463 INIT_LIST_HEAD(&s->s_cap_snaps_flushing);
464
465 dout("register_session mds%d\n", mds);
466 if (mds >= mdsc->max_sessions) {
467 int newmax = 1 << get_count_order(mds+1);
468 struct ceph_mds_session **sa;
469
470 dout("register_session realloc to %d\n", newmax);
471 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
472 if (sa == NULL)
473 goto fail_realloc;
474 if (mdsc->sessions) {
475 memcpy(sa, mdsc->sessions,
476 mdsc->max_sessions * sizeof(void *));
477 kfree(mdsc->sessions);
478 }
479 mdsc->sessions = sa;
480 mdsc->max_sessions = newmax;
481 }
482 mdsc->sessions[mds] = s;
483 atomic_inc(&mdsc->num_sessions);
484 atomic_inc(&s->s_ref); /* one ref to sessions[], one to caller */
485
486 ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
487 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
488
489 return s;
490
491 fail_realloc:
492 kfree(s);
493 return ERR_PTR(-ENOMEM);
494 }
495
496 /*
497 * called under mdsc->mutex
498 */
499 static void __unregister_session(struct ceph_mds_client *mdsc,
500 struct ceph_mds_session *s)
501 {
502 dout("__unregister_session mds%d %p\n", s->s_mds, s);
503 BUG_ON(mdsc->sessions[s->s_mds] != s);
504 mdsc->sessions[s->s_mds] = NULL;
505 ceph_con_close(&s->s_con);
506 ceph_put_mds_session(s);
507 atomic_dec(&mdsc->num_sessions);
508 }
509
510 /*
511 * drop session refs in request.
512 *
513 * should be last request ref, or hold mdsc->mutex
514 */
515 static void put_request_session(struct ceph_mds_request *req)
516 {
517 if (req->r_session) {
518 ceph_put_mds_session(req->r_session);
519 req->r_session = NULL;
520 }
521 }
522
523 void ceph_mdsc_release_request(struct kref *kref)
524 {
525 struct ceph_mds_request *req = container_of(kref,
526 struct ceph_mds_request,
527 r_kref);
528 destroy_reply_info(&req->r_reply_info);
529 if (req->r_request)
530 ceph_msg_put(req->r_request);
531 if (req->r_reply)
532 ceph_msg_put(req->r_reply);
533 if (req->r_inode) {
534 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
535 iput(req->r_inode);
536 }
537 if (req->r_locked_dir)
538 ceph_put_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
539 iput(req->r_target_inode);
540 if (req->r_dentry)
541 dput(req->r_dentry);
542 if (req->r_old_dentry)
543 dput(req->r_old_dentry);
544 if (req->r_old_dentry_dir) {
545 /*
546 * track (and drop pins for) r_old_dentry_dir
547 * separately, since r_old_dentry's d_parent may have
548 * changed between the dir mutex being dropped and
549 * this request being freed.
550 */
551 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
552 CEPH_CAP_PIN);
553 iput(req->r_old_dentry_dir);
554 }
555 kfree(req->r_path1);
556 kfree(req->r_path2);
557 if (req->r_pagelist)
558 ceph_pagelist_release(req->r_pagelist);
559 put_request_session(req);
560 ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
561 kfree(req);
562 }
563
564 /*
565 * lookup session, bump ref if found.
566 *
567 * called under mdsc->mutex.
568 */
569 static struct ceph_mds_request *__lookup_request(struct ceph_mds_client *mdsc,
570 u64 tid)
571 {
572 struct ceph_mds_request *req;
573 struct rb_node *n = mdsc->request_tree.rb_node;
574
575 while (n) {
576 req = rb_entry(n, struct ceph_mds_request, r_node);
577 if (tid < req->r_tid)
578 n = n->rb_left;
579 else if (tid > req->r_tid)
580 n = n->rb_right;
581 else {
582 ceph_mdsc_get_request(req);
583 return req;
584 }
585 }
586 return NULL;
587 }
588
589 static void __insert_request(struct ceph_mds_client *mdsc,
590 struct ceph_mds_request *new)
591 {
592 struct rb_node **p = &mdsc->request_tree.rb_node;
593 struct rb_node *parent = NULL;
594 struct ceph_mds_request *req = NULL;
595
596 while (*p) {
597 parent = *p;
598 req = rb_entry(parent, struct ceph_mds_request, r_node);
599 if (new->r_tid < req->r_tid)
600 p = &(*p)->rb_left;
601 else if (new->r_tid > req->r_tid)
602 p = &(*p)->rb_right;
603 else
604 BUG();
605 }
606
607 rb_link_node(&new->r_node, parent, p);
608 rb_insert_color(&new->r_node, &mdsc->request_tree);
609 }
610
611 /*
612 * Register an in-flight request, and assign a tid. Link to directory
613 * are modifying (if any).
614 *
615 * Called under mdsc->mutex.
616 */
617 static void __register_request(struct ceph_mds_client *mdsc,
618 struct ceph_mds_request *req,
619 struct inode *dir)
620 {
621 req->r_tid = ++mdsc->last_tid;
622 if (req->r_num_caps)
623 ceph_reserve_caps(mdsc, &req->r_caps_reservation,
624 req->r_num_caps);
625 dout("__register_request %p tid %lld\n", req, req->r_tid);
626 ceph_mdsc_get_request(req);
627 __insert_request(mdsc, req);
628
629 req->r_uid = current_fsuid();
630 req->r_gid = current_fsgid();
631
632 if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
633 mdsc->oldest_tid = req->r_tid;
634
635 if (dir) {
636 struct ceph_inode_info *ci = ceph_inode(dir);
637
638 ihold(dir);
639 spin_lock(&ci->i_unsafe_lock);
640 req->r_unsafe_dir = dir;
641 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
642 spin_unlock(&ci->i_unsafe_lock);
643 }
644 }
645
646 static void __unregister_request(struct ceph_mds_client *mdsc,
647 struct ceph_mds_request *req)
648 {
649 dout("__unregister_request %p tid %lld\n", req, req->r_tid);
650
651 if (req->r_tid == mdsc->oldest_tid) {
652 struct rb_node *p = rb_next(&req->r_node);
653 mdsc->oldest_tid = 0;
654 while (p) {
655 struct ceph_mds_request *next_req =
656 rb_entry(p, struct ceph_mds_request, r_node);
657 if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
658 mdsc->oldest_tid = next_req->r_tid;
659 break;
660 }
661 p = rb_next(p);
662 }
663 }
664
665 rb_erase(&req->r_node, &mdsc->request_tree);
666 RB_CLEAR_NODE(&req->r_node);
667
668 if (req->r_unsafe_dir) {
669 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
670
671 spin_lock(&ci->i_unsafe_lock);
672 list_del_init(&req->r_unsafe_dir_item);
673 spin_unlock(&ci->i_unsafe_lock);
674
675 iput(req->r_unsafe_dir);
676 req->r_unsafe_dir = NULL;
677 }
678
679 complete_all(&req->r_safe_completion);
680
681 ceph_mdsc_put_request(req);
682 }
683
684 /*
685 * Choose mds to send request to next. If there is a hint set in the
686 * request (e.g., due to a prior forward hint from the mds), use that.
687 * Otherwise, consult frag tree and/or caps to identify the
688 * appropriate mds. If all else fails, choose randomly.
689 *
690 * Called under mdsc->mutex.
691 */
692 static struct dentry *get_nonsnap_parent(struct dentry *dentry)
693 {
694 /*
695 * we don't need to worry about protecting the d_parent access
696 * here because we never renaming inside the snapped namespace
697 * except to resplice to another snapdir, and either the old or new
698 * result is a valid result.
699 */
700 while (!IS_ROOT(dentry) && ceph_snap(d_inode(dentry)) != CEPH_NOSNAP)
701 dentry = dentry->d_parent;
702 return dentry;
703 }
704
705 static int __choose_mds(struct ceph_mds_client *mdsc,
706 struct ceph_mds_request *req)
707 {
708 struct inode *inode;
709 struct ceph_inode_info *ci;
710 struct ceph_cap *cap;
711 int mode = req->r_direct_mode;
712 int mds = -1;
713 u32 hash = req->r_direct_hash;
714 bool is_hash = req->r_direct_is_hash;
715
716 /*
717 * is there a specific mds we should try? ignore hint if we have
718 * no session and the mds is not up (active or recovering).
719 */
720 if (req->r_resend_mds >= 0 &&
721 (__have_session(mdsc, req->r_resend_mds) ||
722 ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
723 dout("choose_mds using resend_mds mds%d\n",
724 req->r_resend_mds);
725 return req->r_resend_mds;
726 }
727
728 if (mode == USE_RANDOM_MDS)
729 goto random;
730
731 inode = NULL;
732 if (req->r_inode) {
733 inode = req->r_inode;
734 } else if (req->r_dentry) {
735 /* ignore race with rename; old or new d_parent is okay */
736 struct dentry *parent = req->r_dentry->d_parent;
737 struct inode *dir = d_inode(parent);
738
739 if (dir->i_sb != mdsc->fsc->sb) {
740 /* not this fs! */
741 inode = d_inode(req->r_dentry);
742 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
743 /* direct snapped/virtual snapdir requests
744 * based on parent dir inode */
745 struct dentry *dn = get_nonsnap_parent(parent);
746 inode = d_inode(dn);
747 dout("__choose_mds using nonsnap parent %p\n", inode);
748 } else {
749 /* dentry target */
750 inode = d_inode(req->r_dentry);
751 if (!inode || mode == USE_AUTH_MDS) {
752 /* dir + name */
753 inode = dir;
754 hash = ceph_dentry_hash(dir, req->r_dentry);
755 is_hash = true;
756 }
757 }
758 }
759
760 dout("__choose_mds %p is_hash=%d (%d) mode %d\n", inode, (int)is_hash,
761 (int)hash, mode);
762 if (!inode)
763 goto random;
764 ci = ceph_inode(inode);
765
766 if (is_hash && S_ISDIR(inode->i_mode)) {
767 struct ceph_inode_frag frag;
768 int found;
769
770 ceph_choose_frag(ci, hash, &frag, &found);
771 if (found) {
772 if (mode == USE_ANY_MDS && frag.ndist > 0) {
773 u8 r;
774
775 /* choose a random replica */
776 get_random_bytes(&r, 1);
777 r %= frag.ndist;
778 mds = frag.dist[r];
779 dout("choose_mds %p %llx.%llx "
780 "frag %u mds%d (%d/%d)\n",
781 inode, ceph_vinop(inode),
782 frag.frag, mds,
783 (int)r, frag.ndist);
784 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
785 CEPH_MDS_STATE_ACTIVE)
786 return mds;
787 }
788
789 /* since this file/dir wasn't known to be
790 * replicated, then we want to look for the
791 * authoritative mds. */
792 mode = USE_AUTH_MDS;
793 if (frag.mds >= 0) {
794 /* choose auth mds */
795 mds = frag.mds;
796 dout("choose_mds %p %llx.%llx "
797 "frag %u mds%d (auth)\n",
798 inode, ceph_vinop(inode), frag.frag, mds);
799 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
800 CEPH_MDS_STATE_ACTIVE)
801 return mds;
802 }
803 }
804 }
805
806 spin_lock(&ci->i_ceph_lock);
807 cap = NULL;
808 if (mode == USE_AUTH_MDS)
809 cap = ci->i_auth_cap;
810 if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
811 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
812 if (!cap) {
813 spin_unlock(&ci->i_ceph_lock);
814 goto random;
815 }
816 mds = cap->session->s_mds;
817 dout("choose_mds %p %llx.%llx mds%d (%scap %p)\n",
818 inode, ceph_vinop(inode), mds,
819 cap == ci->i_auth_cap ? "auth " : "", cap);
820 spin_unlock(&ci->i_ceph_lock);
821 return mds;
822
823 random:
824 mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
825 dout("choose_mds chose random mds%d\n", mds);
826 return mds;
827 }
828
829
830 /*
831 * session messages
832 */
833 static struct ceph_msg *create_session_msg(u32 op, u64 seq)
834 {
835 struct ceph_msg *msg;
836 struct ceph_mds_session_head *h;
837
838 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
839 false);
840 if (!msg) {
841 pr_err("create_session_msg ENOMEM creating msg\n");
842 return NULL;
843 }
844 h = msg->front.iov_base;
845 h->op = cpu_to_le32(op);
846 h->seq = cpu_to_le64(seq);
847
848 return msg;
849 }
850
851 /*
852 * session message, specialization for CEPH_SESSION_REQUEST_OPEN
853 * to include additional client metadata fields.
854 */
855 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
856 {
857 struct ceph_msg *msg;
858 struct ceph_mds_session_head *h;
859 int i = -1;
860 int metadata_bytes = 0;
861 int metadata_key_count = 0;
862 struct ceph_options *opt = mdsc->fsc->client->options;
863 void *p;
864
865 const char* metadata[][2] = {
866 {"hostname", utsname()->nodename},
867 {"kernel_version", utsname()->release},
868 {"entity_id", opt->name ? opt->name : ""},
869 {NULL, NULL}
870 };
871
872 /* Calculate serialized length of metadata */
873 metadata_bytes = 4; /* map length */
874 for (i = 0; metadata[i][0] != NULL; ++i) {
875 metadata_bytes += 8 + strlen(metadata[i][0]) +
876 strlen(metadata[i][1]);
877 metadata_key_count++;
878 }
879
880 /* Allocate the message */
881 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + metadata_bytes,
882 GFP_NOFS, false);
883 if (!msg) {
884 pr_err("create_session_msg ENOMEM creating msg\n");
885 return NULL;
886 }
887 h = msg->front.iov_base;
888 h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
889 h->seq = cpu_to_le64(seq);
890
891 /*
892 * Serialize client metadata into waiting buffer space, using
893 * the format that userspace expects for map<string, string>
894 *
895 * ClientSession messages with metadata are v2
896 */
897 msg->hdr.version = cpu_to_le16(2);
898 msg->hdr.compat_version = cpu_to_le16(1);
899
900 /* The write pointer, following the session_head structure */
901 p = msg->front.iov_base + sizeof(*h);
902
903 /* Number of entries in the map */
904 ceph_encode_32(&p, metadata_key_count);
905
906 /* Two length-prefixed strings for each entry in the map */
907 for (i = 0; metadata[i][0] != NULL; ++i) {
908 size_t const key_len = strlen(metadata[i][0]);
909 size_t const val_len = strlen(metadata[i][1]);
910
911 ceph_encode_32(&p, key_len);
912 memcpy(p, metadata[i][0], key_len);
913 p += key_len;
914 ceph_encode_32(&p, val_len);
915 memcpy(p, metadata[i][1], val_len);
916 p += val_len;
917 }
918
919 return msg;
920 }
921
922 /*
923 * send session open request.
924 *
925 * called under mdsc->mutex
926 */
927 static int __open_session(struct ceph_mds_client *mdsc,
928 struct ceph_mds_session *session)
929 {
930 struct ceph_msg *msg;
931 int mstate;
932 int mds = session->s_mds;
933
934 /* wait for mds to go active? */
935 mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
936 dout("open_session to mds%d (%s)\n", mds,
937 ceph_mds_state_name(mstate));
938 session->s_state = CEPH_MDS_SESSION_OPENING;
939 session->s_renew_requested = jiffies;
940
941 /* send connect message */
942 msg = create_session_open_msg(mdsc, session->s_seq);
943 if (!msg)
944 return -ENOMEM;
945 ceph_con_send(&session->s_con, msg);
946 return 0;
947 }
948
949 /*
950 * open sessions for any export targets for the given mds
951 *
952 * called under mdsc->mutex
953 */
954 static struct ceph_mds_session *
955 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
956 {
957 struct ceph_mds_session *session;
958
959 session = __ceph_lookup_mds_session(mdsc, target);
960 if (!session) {
961 session = register_session(mdsc, target);
962 if (IS_ERR(session))
963 return session;
964 }
965 if (session->s_state == CEPH_MDS_SESSION_NEW ||
966 session->s_state == CEPH_MDS_SESSION_CLOSING)
967 __open_session(mdsc, session);
968
969 return session;
970 }
971
972 struct ceph_mds_session *
973 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
974 {
975 struct ceph_mds_session *session;
976
977 dout("open_export_target_session to mds%d\n", target);
978
979 mutex_lock(&mdsc->mutex);
980 session = __open_export_target_session(mdsc, target);
981 mutex_unlock(&mdsc->mutex);
982
983 return session;
984 }
985
986 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
987 struct ceph_mds_session *session)
988 {
989 struct ceph_mds_info *mi;
990 struct ceph_mds_session *ts;
991 int i, mds = session->s_mds;
992
993 if (mds >= mdsc->mdsmap->m_max_mds)
994 return;
995
996 mi = &mdsc->mdsmap->m_info[mds];
997 dout("open_export_target_sessions for mds%d (%d targets)\n",
998 session->s_mds, mi->num_export_targets);
999
1000 for (i = 0; i < mi->num_export_targets; i++) {
1001 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1002 if (!IS_ERR(ts))
1003 ceph_put_mds_session(ts);
1004 }
1005 }
1006
1007 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1008 struct ceph_mds_session *session)
1009 {
1010 mutex_lock(&mdsc->mutex);
1011 __open_export_target_sessions(mdsc, session);
1012 mutex_unlock(&mdsc->mutex);
1013 }
1014
1015 /*
1016 * session caps
1017 */
1018
1019 /* caller holds s_cap_lock, we drop it */
1020 static void cleanup_cap_releases(struct ceph_mds_client *mdsc,
1021 struct ceph_mds_session *session)
1022 __releases(session->s_cap_lock)
1023 {
1024 LIST_HEAD(tmp_list);
1025 list_splice_init(&session->s_cap_releases, &tmp_list);
1026 session->s_num_cap_releases = 0;
1027 spin_unlock(&session->s_cap_lock);
1028
1029 dout("cleanup_cap_releases mds%d\n", session->s_mds);
1030 while (!list_empty(&tmp_list)) {
1031 struct ceph_cap *cap;
1032 /* zero out the in-progress message */
1033 cap = list_first_entry(&tmp_list,
1034 struct ceph_cap, session_caps);
1035 list_del(&cap->session_caps);
1036 ceph_put_cap(mdsc, cap);
1037 }
1038 }
1039
1040 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1041 struct ceph_mds_session *session)
1042 {
1043 struct ceph_mds_request *req;
1044 struct rb_node *p;
1045
1046 dout("cleanup_session_requests mds%d\n", session->s_mds);
1047 mutex_lock(&mdsc->mutex);
1048 while (!list_empty(&session->s_unsafe)) {
1049 req = list_first_entry(&session->s_unsafe,
1050 struct ceph_mds_request, r_unsafe_item);
1051 list_del_init(&req->r_unsafe_item);
1052 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1053 req->r_tid);
1054 __unregister_request(mdsc, req);
1055 }
1056 /* zero r_attempts, so kick_requests() will re-send requests */
1057 p = rb_first(&mdsc->request_tree);
1058 while (p) {
1059 req = rb_entry(p, struct ceph_mds_request, r_node);
1060 p = rb_next(p);
1061 if (req->r_session &&
1062 req->r_session->s_mds == session->s_mds)
1063 req->r_attempts = 0;
1064 }
1065 mutex_unlock(&mdsc->mutex);
1066 }
1067
1068 /*
1069 * Helper to safely iterate over all caps associated with a session, with
1070 * special care taken to handle a racing __ceph_remove_cap().
1071 *
1072 * Caller must hold session s_mutex.
1073 */
1074 static int iterate_session_caps(struct ceph_mds_session *session,
1075 int (*cb)(struct inode *, struct ceph_cap *,
1076 void *), void *arg)
1077 {
1078 struct list_head *p;
1079 struct ceph_cap *cap;
1080 struct inode *inode, *last_inode = NULL;
1081 struct ceph_cap *old_cap = NULL;
1082 int ret;
1083
1084 dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1085 spin_lock(&session->s_cap_lock);
1086 p = session->s_caps.next;
1087 while (p != &session->s_caps) {
1088 cap = list_entry(p, struct ceph_cap, session_caps);
1089 inode = igrab(&cap->ci->vfs_inode);
1090 if (!inode) {
1091 p = p->next;
1092 continue;
1093 }
1094 session->s_cap_iterator = cap;
1095 spin_unlock(&session->s_cap_lock);
1096
1097 if (last_inode) {
1098 iput(last_inode);
1099 last_inode = NULL;
1100 }
1101 if (old_cap) {
1102 ceph_put_cap(session->s_mdsc, old_cap);
1103 old_cap = NULL;
1104 }
1105
1106 ret = cb(inode, cap, arg);
1107 last_inode = inode;
1108
1109 spin_lock(&session->s_cap_lock);
1110 p = p->next;
1111 if (cap->ci == NULL) {
1112 dout("iterate_session_caps finishing cap %p removal\n",
1113 cap);
1114 BUG_ON(cap->session != session);
1115 cap->session = NULL;
1116 list_del_init(&cap->session_caps);
1117 session->s_nr_caps--;
1118 if (cap->queue_release) {
1119 list_add_tail(&cap->session_caps,
1120 &session->s_cap_releases);
1121 session->s_num_cap_releases++;
1122 } else {
1123 old_cap = cap; /* put_cap it w/o locks held */
1124 }
1125 }
1126 if (ret < 0)
1127 goto out;
1128 }
1129 ret = 0;
1130 out:
1131 session->s_cap_iterator = NULL;
1132 spin_unlock(&session->s_cap_lock);
1133
1134 iput(last_inode);
1135 if (old_cap)
1136 ceph_put_cap(session->s_mdsc, old_cap);
1137
1138 return ret;
1139 }
1140
1141 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1142 void *arg)
1143 {
1144 struct ceph_inode_info *ci = ceph_inode(inode);
1145 LIST_HEAD(to_remove);
1146 int drop = 0;
1147
1148 dout("removing cap %p, ci is %p, inode is %p\n",
1149 cap, ci, &ci->vfs_inode);
1150 spin_lock(&ci->i_ceph_lock);
1151 __ceph_remove_cap(cap, false);
1152 if (!ci->i_auth_cap) {
1153 struct ceph_cap_flush *cf;
1154 struct ceph_mds_client *mdsc =
1155 ceph_sb_to_client(inode->i_sb)->mdsc;
1156
1157 while (true) {
1158 struct rb_node *n = rb_first(&ci->i_cap_flush_tree);
1159 if (!n)
1160 break;
1161 cf = rb_entry(n, struct ceph_cap_flush, i_node);
1162 rb_erase(&cf->i_node, &ci->i_cap_flush_tree);
1163 list_add(&cf->list, &to_remove);
1164 }
1165
1166 spin_lock(&mdsc->cap_dirty_lock);
1167
1168 list_for_each_entry(cf, &to_remove, list)
1169 rb_erase(&cf->g_node, &mdsc->cap_flush_tree);
1170
1171 if (!list_empty(&ci->i_dirty_item)) {
1172 pr_warn_ratelimited(
1173 " dropping dirty %s state for %p %lld\n",
1174 ceph_cap_string(ci->i_dirty_caps),
1175 inode, ceph_ino(inode));
1176 ci->i_dirty_caps = 0;
1177 list_del_init(&ci->i_dirty_item);
1178 drop = 1;
1179 }
1180 if (!list_empty(&ci->i_flushing_item)) {
1181 pr_warn_ratelimited(
1182 " dropping dirty+flushing %s state for %p %lld\n",
1183 ceph_cap_string(ci->i_flushing_caps),
1184 inode, ceph_ino(inode));
1185 ci->i_flushing_caps = 0;
1186 list_del_init(&ci->i_flushing_item);
1187 mdsc->num_cap_flushing--;
1188 drop = 1;
1189 }
1190 spin_unlock(&mdsc->cap_dirty_lock);
1191
1192 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
1193 list_add(&ci->i_prealloc_cap_flush->list, &to_remove);
1194 ci->i_prealloc_cap_flush = NULL;
1195 }
1196 }
1197 spin_unlock(&ci->i_ceph_lock);
1198 while (!list_empty(&to_remove)) {
1199 struct ceph_cap_flush *cf;
1200 cf = list_first_entry(&to_remove,
1201 struct ceph_cap_flush, list);
1202 list_del(&cf->list);
1203 ceph_free_cap_flush(cf);
1204 }
1205 while (drop--)
1206 iput(inode);
1207 return 0;
1208 }
1209
1210 /*
1211 * caller must hold session s_mutex
1212 */
1213 static void remove_session_caps(struct ceph_mds_session *session)
1214 {
1215 dout("remove_session_caps on %p\n", session);
1216 iterate_session_caps(session, remove_session_caps_cb, NULL);
1217
1218 spin_lock(&session->s_cap_lock);
1219 if (session->s_nr_caps > 0) {
1220 struct super_block *sb = session->s_mdsc->fsc->sb;
1221 struct inode *inode;
1222 struct ceph_cap *cap, *prev = NULL;
1223 struct ceph_vino vino;
1224 /*
1225 * iterate_session_caps() skips inodes that are being
1226 * deleted, we need to wait until deletions are complete.
1227 * __wait_on_freeing_inode() is designed for the job,
1228 * but it is not exported, so use lookup inode function
1229 * to access it.
1230 */
1231 while (!list_empty(&session->s_caps)) {
1232 cap = list_entry(session->s_caps.next,
1233 struct ceph_cap, session_caps);
1234 if (cap == prev)
1235 break;
1236 prev = cap;
1237 vino = cap->ci->i_vino;
1238 spin_unlock(&session->s_cap_lock);
1239
1240 inode = ceph_find_inode(sb, vino);
1241 iput(inode);
1242
1243 spin_lock(&session->s_cap_lock);
1244 }
1245 }
1246
1247 // drop cap expires and unlock s_cap_lock
1248 cleanup_cap_releases(session->s_mdsc, session);
1249
1250 BUG_ON(session->s_nr_caps > 0);
1251 BUG_ON(!list_empty(&session->s_cap_flushing));
1252 }
1253
1254 /*
1255 * wake up any threads waiting on this session's caps. if the cap is
1256 * old (didn't get renewed on the client reconnect), remove it now.
1257 *
1258 * caller must hold s_mutex.
1259 */
1260 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1261 void *arg)
1262 {
1263 struct ceph_inode_info *ci = ceph_inode(inode);
1264
1265 wake_up_all(&ci->i_cap_wq);
1266 if (arg) {
1267 spin_lock(&ci->i_ceph_lock);
1268 ci->i_wanted_max_size = 0;
1269 ci->i_requested_max_size = 0;
1270 spin_unlock(&ci->i_ceph_lock);
1271 }
1272 return 0;
1273 }
1274
1275 static void wake_up_session_caps(struct ceph_mds_session *session,
1276 int reconnect)
1277 {
1278 dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1279 iterate_session_caps(session, wake_up_session_cb,
1280 (void *)(unsigned long)reconnect);
1281 }
1282
1283 /*
1284 * Send periodic message to MDS renewing all currently held caps. The
1285 * ack will reset the expiration for all caps from this session.
1286 *
1287 * caller holds s_mutex
1288 */
1289 static int send_renew_caps(struct ceph_mds_client *mdsc,
1290 struct ceph_mds_session *session)
1291 {
1292 struct ceph_msg *msg;
1293 int state;
1294
1295 if (time_after_eq(jiffies, session->s_cap_ttl) &&
1296 time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1297 pr_info("mds%d caps stale\n", session->s_mds);
1298 session->s_renew_requested = jiffies;
1299
1300 /* do not try to renew caps until a recovering mds has reconnected
1301 * with its clients. */
1302 state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1303 if (state < CEPH_MDS_STATE_RECONNECT) {
1304 dout("send_renew_caps ignoring mds%d (%s)\n",
1305 session->s_mds, ceph_mds_state_name(state));
1306 return 0;
1307 }
1308
1309 dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1310 ceph_mds_state_name(state));
1311 msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1312 ++session->s_renew_seq);
1313 if (!msg)
1314 return -ENOMEM;
1315 ceph_con_send(&session->s_con, msg);
1316 return 0;
1317 }
1318
1319 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1320 struct ceph_mds_session *session, u64 seq)
1321 {
1322 struct ceph_msg *msg;
1323
1324 dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1325 session->s_mds, ceph_session_state_name(session->s_state), seq);
1326 msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1327 if (!msg)
1328 return -ENOMEM;
1329 ceph_con_send(&session->s_con, msg);
1330 return 0;
1331 }
1332
1333
1334 /*
1335 * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1336 *
1337 * Called under session->s_mutex
1338 */
1339 static void renewed_caps(struct ceph_mds_client *mdsc,
1340 struct ceph_mds_session *session, int is_renew)
1341 {
1342 int was_stale;
1343 int wake = 0;
1344
1345 spin_lock(&session->s_cap_lock);
1346 was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1347
1348 session->s_cap_ttl = session->s_renew_requested +
1349 mdsc->mdsmap->m_session_timeout*HZ;
1350
1351 if (was_stale) {
1352 if (time_before(jiffies, session->s_cap_ttl)) {
1353 pr_info("mds%d caps renewed\n", session->s_mds);
1354 wake = 1;
1355 } else {
1356 pr_info("mds%d caps still stale\n", session->s_mds);
1357 }
1358 }
1359 dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1360 session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1361 time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1362 spin_unlock(&session->s_cap_lock);
1363
1364 if (wake)
1365 wake_up_session_caps(session, 0);
1366 }
1367
1368 /*
1369 * send a session close request
1370 */
1371 static int request_close_session(struct ceph_mds_client *mdsc,
1372 struct ceph_mds_session *session)
1373 {
1374 struct ceph_msg *msg;
1375
1376 dout("request_close_session mds%d state %s seq %lld\n",
1377 session->s_mds, ceph_session_state_name(session->s_state),
1378 session->s_seq);
1379 msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
1380 if (!msg)
1381 return -ENOMEM;
1382 ceph_con_send(&session->s_con, msg);
1383 return 0;
1384 }
1385
1386 /*
1387 * Called with s_mutex held.
1388 */
1389 static int __close_session(struct ceph_mds_client *mdsc,
1390 struct ceph_mds_session *session)
1391 {
1392 if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1393 return 0;
1394 session->s_state = CEPH_MDS_SESSION_CLOSING;
1395 return request_close_session(mdsc, session);
1396 }
1397
1398 /*
1399 * Trim old(er) caps.
1400 *
1401 * Because we can't cache an inode without one or more caps, we do
1402 * this indirectly: if a cap is unused, we prune its aliases, at which
1403 * point the inode will hopefully get dropped to.
1404 *
1405 * Yes, this is a bit sloppy. Our only real goal here is to respond to
1406 * memory pressure from the MDS, though, so it needn't be perfect.
1407 */
1408 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1409 {
1410 struct ceph_mds_session *session = arg;
1411 struct ceph_inode_info *ci = ceph_inode(inode);
1412 int used, wanted, oissued, mine;
1413
1414 if (session->s_trim_caps <= 0)
1415 return -1;
1416
1417 spin_lock(&ci->i_ceph_lock);
1418 mine = cap->issued | cap->implemented;
1419 used = __ceph_caps_used(ci);
1420 wanted = __ceph_caps_file_wanted(ci);
1421 oissued = __ceph_caps_issued_other(ci, cap);
1422
1423 dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1424 inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1425 ceph_cap_string(used), ceph_cap_string(wanted));
1426 if (cap == ci->i_auth_cap) {
1427 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1428 !list_empty(&ci->i_cap_snaps))
1429 goto out;
1430 if ((used | wanted) & CEPH_CAP_ANY_WR)
1431 goto out;
1432 }
1433 if ((used | wanted) & ~oissued & mine)
1434 goto out; /* we need these caps */
1435
1436 session->s_trim_caps--;
1437 if (oissued) {
1438 /* we aren't the only cap.. just remove us */
1439 __ceph_remove_cap(cap, true);
1440 } else {
1441 /* try to drop referring dentries */
1442 spin_unlock(&ci->i_ceph_lock);
1443 d_prune_aliases(inode);
1444 dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1445 inode, cap, atomic_read(&inode->i_count));
1446 return 0;
1447 }
1448
1449 out:
1450 spin_unlock(&ci->i_ceph_lock);
1451 return 0;
1452 }
1453
1454 /*
1455 * Trim session cap count down to some max number.
1456 */
1457 static int trim_caps(struct ceph_mds_client *mdsc,
1458 struct ceph_mds_session *session,
1459 int max_caps)
1460 {
1461 int trim_caps = session->s_nr_caps - max_caps;
1462
1463 dout("trim_caps mds%d start: %d / %d, trim %d\n",
1464 session->s_mds, session->s_nr_caps, max_caps, trim_caps);
1465 if (trim_caps > 0) {
1466 session->s_trim_caps = trim_caps;
1467 iterate_session_caps(session, trim_caps_cb, session);
1468 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
1469 session->s_mds, session->s_nr_caps, max_caps,
1470 trim_caps - session->s_trim_caps);
1471 session->s_trim_caps = 0;
1472 }
1473
1474 ceph_send_cap_releases(mdsc, session);
1475 return 0;
1476 }
1477
1478 static int check_capsnap_flush(struct ceph_inode_info *ci,
1479 u64 want_snap_seq)
1480 {
1481 int ret = 1;
1482 spin_lock(&ci->i_ceph_lock);
1483 if (want_snap_seq > 0 && !list_empty(&ci->i_cap_snaps)) {
1484 struct ceph_cap_snap *capsnap =
1485 list_first_entry(&ci->i_cap_snaps,
1486 struct ceph_cap_snap, ci_item);
1487 ret = capsnap->follows >= want_snap_seq;
1488 }
1489 spin_unlock(&ci->i_ceph_lock);
1490 return ret;
1491 }
1492
1493 static int check_caps_flush(struct ceph_mds_client *mdsc,
1494 u64 want_flush_tid)
1495 {
1496 struct rb_node *n;
1497 struct ceph_cap_flush *cf;
1498 int ret = 1;
1499
1500 spin_lock(&mdsc->cap_dirty_lock);
1501 n = rb_first(&mdsc->cap_flush_tree);
1502 cf = n ? rb_entry(n, struct ceph_cap_flush, g_node) : NULL;
1503 if (cf && cf->tid <= want_flush_tid) {
1504 dout("check_caps_flush still flushing tid %llu <= %llu\n",
1505 cf->tid, want_flush_tid);
1506 ret = 0;
1507 }
1508 spin_unlock(&mdsc->cap_dirty_lock);
1509 return ret;
1510 }
1511
1512 /*
1513 * flush all dirty inode data to disk.
1514 *
1515 * returns true if we've flushed through want_flush_tid
1516 */
1517 static void wait_caps_flush(struct ceph_mds_client *mdsc,
1518 u64 want_flush_tid, u64 want_snap_seq)
1519 {
1520 int mds;
1521
1522 dout("check_caps_flush want %llu snap want %llu\n",
1523 want_flush_tid, want_snap_seq);
1524 mutex_lock(&mdsc->mutex);
1525 for (mds = 0; mds < mdsc->max_sessions; ) {
1526 struct ceph_mds_session *session = mdsc->sessions[mds];
1527 struct inode *inode = NULL;
1528
1529 if (!session) {
1530 mds++;
1531 continue;
1532 }
1533 get_session(session);
1534 mutex_unlock(&mdsc->mutex);
1535
1536 mutex_lock(&session->s_mutex);
1537 if (!list_empty(&session->s_cap_snaps_flushing)) {
1538 struct ceph_cap_snap *capsnap =
1539 list_first_entry(&session->s_cap_snaps_flushing,
1540 struct ceph_cap_snap,
1541 flushing_item);
1542 struct ceph_inode_info *ci = capsnap->ci;
1543 if (!check_capsnap_flush(ci, want_snap_seq)) {
1544 dout("check_cap_flush still flushing snap %p "
1545 "follows %lld <= %lld to mds%d\n",
1546 &ci->vfs_inode, capsnap->follows,
1547 want_snap_seq, mds);
1548 inode = igrab(&ci->vfs_inode);
1549 }
1550 }
1551 mutex_unlock(&session->s_mutex);
1552 ceph_put_mds_session(session);
1553
1554 if (inode) {
1555 wait_event(mdsc->cap_flushing_wq,
1556 check_capsnap_flush(ceph_inode(inode),
1557 want_snap_seq));
1558 iput(inode);
1559 } else {
1560 mds++;
1561 }
1562
1563 mutex_lock(&mdsc->mutex);
1564 }
1565 mutex_unlock(&mdsc->mutex);
1566
1567 wait_event(mdsc->cap_flushing_wq,
1568 check_caps_flush(mdsc, want_flush_tid));
1569
1570 dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
1571 }
1572
1573 /*
1574 * called under s_mutex
1575 */
1576 void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
1577 struct ceph_mds_session *session)
1578 {
1579 struct ceph_msg *msg = NULL;
1580 struct ceph_mds_cap_release *head;
1581 struct ceph_mds_cap_item *item;
1582 struct ceph_cap *cap;
1583 LIST_HEAD(tmp_list);
1584 int num_cap_releases;
1585
1586 spin_lock(&session->s_cap_lock);
1587 again:
1588 list_splice_init(&session->s_cap_releases, &tmp_list);
1589 num_cap_releases = session->s_num_cap_releases;
1590 session->s_num_cap_releases = 0;
1591 spin_unlock(&session->s_cap_lock);
1592
1593 while (!list_empty(&tmp_list)) {
1594 if (!msg) {
1595 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
1596 PAGE_CACHE_SIZE, GFP_NOFS, false);
1597 if (!msg)
1598 goto out_err;
1599 head = msg->front.iov_base;
1600 head->num = cpu_to_le32(0);
1601 msg->front.iov_len = sizeof(*head);
1602 }
1603 cap = list_first_entry(&tmp_list, struct ceph_cap,
1604 session_caps);
1605 list_del(&cap->session_caps);
1606 num_cap_releases--;
1607
1608 head = msg->front.iov_base;
1609 le32_add_cpu(&head->num, 1);
1610 item = msg->front.iov_base + msg->front.iov_len;
1611 item->ino = cpu_to_le64(cap->cap_ino);
1612 item->cap_id = cpu_to_le64(cap->cap_id);
1613 item->migrate_seq = cpu_to_le32(cap->mseq);
1614 item->seq = cpu_to_le32(cap->issue_seq);
1615 msg->front.iov_len += sizeof(*item);
1616
1617 ceph_put_cap(mdsc, cap);
1618
1619 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
1620 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1621 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
1622 ceph_con_send(&session->s_con, msg);
1623 msg = NULL;
1624 }
1625 }
1626
1627 BUG_ON(num_cap_releases != 0);
1628
1629 spin_lock(&session->s_cap_lock);
1630 if (!list_empty(&session->s_cap_releases))
1631 goto again;
1632 spin_unlock(&session->s_cap_lock);
1633
1634 if (msg) {
1635 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1636 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
1637 ceph_con_send(&session->s_con, msg);
1638 }
1639 return;
1640 out_err:
1641 pr_err("send_cap_releases mds%d, failed to allocate message\n",
1642 session->s_mds);
1643 spin_lock(&session->s_cap_lock);
1644 list_splice(&tmp_list, &session->s_cap_releases);
1645 session->s_num_cap_releases += num_cap_releases;
1646 spin_unlock(&session->s_cap_lock);
1647 }
1648
1649 /*
1650 * requests
1651 */
1652
1653 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
1654 struct inode *dir)
1655 {
1656 struct ceph_inode_info *ci = ceph_inode(dir);
1657 struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
1658 struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
1659 size_t size = sizeof(*rinfo->dir_in) + sizeof(*rinfo->dir_dname_len) +
1660 sizeof(*rinfo->dir_dname) + sizeof(*rinfo->dir_dlease);
1661 int order, num_entries;
1662
1663 spin_lock(&ci->i_ceph_lock);
1664 num_entries = ci->i_files + ci->i_subdirs;
1665 spin_unlock(&ci->i_ceph_lock);
1666 num_entries = max(num_entries, 1);
1667 num_entries = min(num_entries, opt->max_readdir);
1668
1669 order = get_order(size * num_entries);
1670 while (order >= 0) {
1671 rinfo->dir_in = (void*)__get_free_pages(GFP_KERNEL |
1672 __GFP_NOWARN,
1673 order);
1674 if (rinfo->dir_in)
1675 break;
1676 order--;
1677 }
1678 if (!rinfo->dir_in)
1679 return -ENOMEM;
1680
1681 num_entries = (PAGE_SIZE << order) / size;
1682 num_entries = min(num_entries, opt->max_readdir);
1683
1684 rinfo->dir_buf_size = PAGE_SIZE << order;
1685 req->r_num_caps = num_entries + 1;
1686 req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
1687 req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
1688 return 0;
1689 }
1690
1691 /*
1692 * Create an mds request.
1693 */
1694 struct ceph_mds_request *
1695 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
1696 {
1697 struct ceph_mds_request *req = kzalloc(sizeof(*req), GFP_NOFS);
1698
1699 if (!req)
1700 return ERR_PTR(-ENOMEM);
1701
1702 mutex_init(&req->r_fill_mutex);
1703 req->r_mdsc = mdsc;
1704 req->r_started = jiffies;
1705 req->r_resend_mds = -1;
1706 INIT_LIST_HEAD(&req->r_unsafe_dir_item);
1707 req->r_fmode = -1;
1708 kref_init(&req->r_kref);
1709 INIT_LIST_HEAD(&req->r_wait);
1710 init_completion(&req->r_completion);
1711 init_completion(&req->r_safe_completion);
1712 INIT_LIST_HEAD(&req->r_unsafe_item);
1713
1714 req->r_stamp = CURRENT_TIME;
1715
1716 req->r_op = op;
1717 req->r_direct_mode = mode;
1718 return req;
1719 }
1720
1721 /*
1722 * return oldest (lowest) request, tid in request tree, 0 if none.
1723 *
1724 * called under mdsc->mutex.
1725 */
1726 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
1727 {
1728 if (RB_EMPTY_ROOT(&mdsc->request_tree))
1729 return NULL;
1730 return rb_entry(rb_first(&mdsc->request_tree),
1731 struct ceph_mds_request, r_node);
1732 }
1733
1734 static inline u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
1735 {
1736 return mdsc->oldest_tid;
1737 }
1738
1739 /*
1740 * Build a dentry's path. Allocate on heap; caller must kfree. Based
1741 * on build_path_from_dentry in fs/cifs/dir.c.
1742 *
1743 * If @stop_on_nosnap, generate path relative to the first non-snapped
1744 * inode.
1745 *
1746 * Encode hidden .snap dirs as a double /, i.e.
1747 * foo/.snap/bar -> foo//bar
1748 */
1749 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *base,
1750 int stop_on_nosnap)
1751 {
1752 struct dentry *temp;
1753 char *path;
1754 int len, pos;
1755 unsigned seq;
1756
1757 if (dentry == NULL)
1758 return ERR_PTR(-EINVAL);
1759
1760 retry:
1761 len = 0;
1762 seq = read_seqbegin(&rename_lock);
1763 rcu_read_lock();
1764 for (temp = dentry; !IS_ROOT(temp);) {
1765 struct inode *inode = d_inode(temp);
1766 if (inode && ceph_snap(inode) == CEPH_SNAPDIR)
1767 len++; /* slash only */
1768 else if (stop_on_nosnap && inode &&
1769 ceph_snap(inode) == CEPH_NOSNAP)
1770 break;
1771 else
1772 len += 1 + temp->d_name.len;
1773 temp = temp->d_parent;
1774 }
1775 rcu_read_unlock();
1776 if (len)
1777 len--; /* no leading '/' */
1778
1779 path = kmalloc(len+1, GFP_NOFS);
1780 if (path == NULL)
1781 return ERR_PTR(-ENOMEM);
1782 pos = len;
1783 path[pos] = 0; /* trailing null */
1784 rcu_read_lock();
1785 for (temp = dentry; !IS_ROOT(temp) && pos != 0; ) {
1786 struct inode *inode;
1787
1788 spin_lock(&temp->d_lock);
1789 inode = d_inode(temp);
1790 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
1791 dout("build_path path+%d: %p SNAPDIR\n",
1792 pos, temp);
1793 } else if (stop_on_nosnap && inode &&
1794 ceph_snap(inode) == CEPH_NOSNAP) {
1795 spin_unlock(&temp->d_lock);
1796 break;
1797 } else {
1798 pos -= temp->d_name.len;
1799 if (pos < 0) {
1800 spin_unlock(&temp->d_lock);
1801 break;
1802 }
1803 strncpy(path + pos, temp->d_name.name,
1804 temp->d_name.len);
1805 }
1806 spin_unlock(&temp->d_lock);
1807 if (pos)
1808 path[--pos] = '/';
1809 temp = temp->d_parent;
1810 }
1811 rcu_read_unlock();
1812 if (pos != 0 || read_seqretry(&rename_lock, seq)) {
1813 pr_err("build_path did not end path lookup where "
1814 "expected, namelen is %d, pos is %d\n", len, pos);
1815 /* presumably this is only possible if racing with a
1816 rename of one of the parent directories (we can not
1817 lock the dentries above us to prevent this, but
1818 retrying should be harmless) */
1819 kfree(path);
1820 goto retry;
1821 }
1822
1823 *base = ceph_ino(d_inode(temp));
1824 *plen = len;
1825 dout("build_path on %p %d built %llx '%.*s'\n",
1826 dentry, d_count(dentry), *base, len, path);
1827 return path;
1828 }
1829
1830 static int build_dentry_path(struct dentry *dentry,
1831 const char **ppath, int *ppathlen, u64 *pino,
1832 int *pfreepath)
1833 {
1834 char *path;
1835
1836 if (ceph_snap(d_inode(dentry->d_parent)) == CEPH_NOSNAP) {
1837 *pino = ceph_ino(d_inode(dentry->d_parent));
1838 *ppath = dentry->d_name.name;
1839 *ppathlen = dentry->d_name.len;
1840 return 0;
1841 }
1842 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
1843 if (IS_ERR(path))
1844 return PTR_ERR(path);
1845 *ppath = path;
1846 *pfreepath = 1;
1847 return 0;
1848 }
1849
1850 static int build_inode_path(struct inode *inode,
1851 const char **ppath, int *ppathlen, u64 *pino,
1852 int *pfreepath)
1853 {
1854 struct dentry *dentry;
1855 char *path;
1856
1857 if (ceph_snap(inode) == CEPH_NOSNAP) {
1858 *pino = ceph_ino(inode);
1859 *ppathlen = 0;
1860 return 0;
1861 }
1862 dentry = d_find_alias(inode);
1863 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
1864 dput(dentry);
1865 if (IS_ERR(path))
1866 return PTR_ERR(path);
1867 *ppath = path;
1868 *pfreepath = 1;
1869 return 0;
1870 }
1871
1872 /*
1873 * request arguments may be specified via an inode *, a dentry *, or
1874 * an explicit ino+path.
1875 */
1876 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
1877 const char *rpath, u64 rino,
1878 const char **ppath, int *pathlen,
1879 u64 *ino, int *freepath)
1880 {
1881 int r = 0;
1882
1883 if (rinode) {
1884 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
1885 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
1886 ceph_snap(rinode));
1887 } else if (rdentry) {
1888 r = build_dentry_path(rdentry, ppath, pathlen, ino, freepath);
1889 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
1890 *ppath);
1891 } else if (rpath || rino) {
1892 *ino = rino;
1893 *ppath = rpath;
1894 *pathlen = rpath ? strlen(rpath) : 0;
1895 dout(" path %.*s\n", *pathlen, rpath);
1896 }
1897
1898 return r;
1899 }
1900
1901 /*
1902 * called under mdsc->mutex
1903 */
1904 static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
1905 struct ceph_mds_request *req,
1906 int mds, bool drop_cap_releases)
1907 {
1908 struct ceph_msg *msg;
1909 struct ceph_mds_request_head *head;
1910 const char *path1 = NULL;
1911 const char *path2 = NULL;
1912 u64 ino1 = 0, ino2 = 0;
1913 int pathlen1 = 0, pathlen2 = 0;
1914 int freepath1 = 0, freepath2 = 0;
1915 int len;
1916 u16 releases;
1917 void *p, *end;
1918 int ret;
1919
1920 ret = set_request_path_attr(req->r_inode, req->r_dentry,
1921 req->r_path1, req->r_ino1.ino,
1922 &path1, &pathlen1, &ino1, &freepath1);
1923 if (ret < 0) {
1924 msg = ERR_PTR(ret);
1925 goto out;
1926 }
1927
1928 ret = set_request_path_attr(NULL, req->r_old_dentry,
1929 req->r_path2, req->r_ino2.ino,
1930 &path2, &pathlen2, &ino2, &freepath2);
1931 if (ret < 0) {
1932 msg = ERR_PTR(ret);
1933 goto out_free1;
1934 }
1935
1936 len = sizeof(*head) +
1937 pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
1938 sizeof(struct timespec);
1939
1940 /* calculate (max) length for cap releases */
1941 len += sizeof(struct ceph_mds_request_release) *
1942 (!!req->r_inode_drop + !!req->r_dentry_drop +
1943 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
1944 if (req->r_dentry_drop)
1945 len += req->r_dentry->d_name.len;
1946 if (req->r_old_dentry_drop)
1947 len += req->r_old_dentry->d_name.len;
1948
1949 msg = ceph_msg_new(CEPH_MSG_CLIENT_REQUEST, len, GFP_NOFS, false);
1950 if (!msg) {
1951 msg = ERR_PTR(-ENOMEM);
1952 goto out_free2;
1953 }
1954
1955 msg->hdr.version = cpu_to_le16(2);
1956 msg->hdr.tid = cpu_to_le64(req->r_tid);
1957
1958 head = msg->front.iov_base;
1959 p = msg->front.iov_base + sizeof(*head);
1960 end = msg->front.iov_base + msg->front.iov_len;
1961
1962 head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
1963 head->op = cpu_to_le32(req->r_op);
1964 head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));
1965 head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));
1966 head->args = req->r_args;
1967
1968 ceph_encode_filepath(&p, end, ino1, path1);
1969 ceph_encode_filepath(&p, end, ino2, path2);
1970
1971 /* make note of release offset, in case we need to replay */
1972 req->r_request_release_offset = p - msg->front.iov_base;
1973
1974 /* cap releases */
1975 releases = 0;
1976 if (req->r_inode_drop)
1977 releases += ceph_encode_inode_release(&p,
1978 req->r_inode ? req->r_inode : d_inode(req->r_dentry),
1979 mds, req->r_inode_drop, req->r_inode_unless, 0);
1980 if (req->r_dentry_drop)
1981 releases += ceph_encode_dentry_release(&p, req->r_dentry,
1982 mds, req->r_dentry_drop, req->r_dentry_unless);
1983 if (req->r_old_dentry_drop)
1984 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
1985 mds, req->r_old_dentry_drop, req->r_old_dentry_unless);
1986 if (req->r_old_inode_drop)
1987 releases += ceph_encode_inode_release(&p,
1988 d_inode(req->r_old_dentry),
1989 mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
1990
1991 if (drop_cap_releases) {
1992 releases = 0;
1993 p = msg->front.iov_base + req->r_request_release_offset;
1994 }
1995
1996 head->num_releases = cpu_to_le16(releases);
1997
1998 /* time stamp */
1999 {
2000 struct ceph_timespec ts;
2001 ceph_encode_timespec(&ts, &req->r_stamp);
2002 ceph_encode_copy(&p, &ts, sizeof(ts));
2003 }
2004
2005 BUG_ON(p > end);
2006 msg->front.iov_len = p - msg->front.iov_base;
2007 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2008
2009 if (req->r_pagelist) {
2010 struct ceph_pagelist *pagelist = req->r_pagelist;
2011 atomic_inc(&pagelist->refcnt);
2012 ceph_msg_data_add_pagelist(msg, pagelist);
2013 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2014 } else {
2015 msg->hdr.data_len = 0;
2016 }
2017
2018 msg->hdr.data_off = cpu_to_le16(0);
2019
2020 out_free2:
2021 if (freepath2)
2022 kfree((char *)path2);
2023 out_free1:
2024 if (freepath1)
2025 kfree((char *)path1);
2026 out:
2027 return msg;
2028 }
2029
2030 /*
2031 * called under mdsc->mutex if error, under no mutex if
2032 * success.
2033 */
2034 static void complete_request(struct ceph_mds_client *mdsc,
2035 struct ceph_mds_request *req)
2036 {
2037 if (req->r_callback)
2038 req->r_callback(mdsc, req);
2039 else
2040 complete_all(&req->r_completion);
2041 }
2042
2043 /*
2044 * called under mdsc->mutex
2045 */
2046 static int __prepare_send_request(struct ceph_mds_client *mdsc,
2047 struct ceph_mds_request *req,
2048 int mds, bool drop_cap_releases)
2049 {
2050 struct ceph_mds_request_head *rhead;
2051 struct ceph_msg *msg;
2052 int flags = 0;
2053
2054 req->r_attempts++;
2055 if (req->r_inode) {
2056 struct ceph_cap *cap =
2057 ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2058
2059 if (cap)
2060 req->r_sent_on_mseq = cap->mseq;
2061 else
2062 req->r_sent_on_mseq = -1;
2063 }
2064 dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2065 req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2066
2067 if (req->r_got_unsafe) {
2068 void *p;
2069 /*
2070 * Replay. Do not regenerate message (and rebuild
2071 * paths, etc.); just use the original message.
2072 * Rebuilding paths will break for renames because
2073 * d_move mangles the src name.
2074 */
2075 msg = req->r_request;
2076 rhead = msg->front.iov_base;
2077
2078 flags = le32_to_cpu(rhead->flags);
2079 flags |= CEPH_MDS_FLAG_REPLAY;
2080 rhead->flags = cpu_to_le32(flags);
2081
2082 if (req->r_target_inode)
2083 rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2084
2085 rhead->num_retry = req->r_attempts - 1;
2086
2087 /* remove cap/dentry releases from message */
2088 rhead->num_releases = 0;
2089
2090 /* time stamp */
2091 p = msg->front.iov_base + req->r_request_release_offset;
2092 {
2093 struct ceph_timespec ts;
2094 ceph_encode_timespec(&ts, &req->r_stamp);
2095 ceph_encode_copy(&p, &ts, sizeof(ts));
2096 }
2097
2098 msg->front.iov_len = p - msg->front.iov_base;
2099 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2100 return 0;
2101 }
2102
2103 if (req->r_request) {
2104 ceph_msg_put(req->r_request);
2105 req->r_request = NULL;
2106 }
2107 msg = create_request_message(mdsc, req, mds, drop_cap_releases);
2108 if (IS_ERR(msg)) {
2109 req->r_err = PTR_ERR(msg);
2110 complete_request(mdsc, req);
2111 return PTR_ERR(msg);
2112 }
2113 req->r_request = msg;
2114
2115 rhead = msg->front.iov_base;
2116 rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2117 if (req->r_got_unsafe)
2118 flags |= CEPH_MDS_FLAG_REPLAY;
2119 if (req->r_locked_dir)
2120 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2121 rhead->flags = cpu_to_le32(flags);
2122 rhead->num_fwd = req->r_num_fwd;
2123 rhead->num_retry = req->r_attempts - 1;
2124 rhead->ino = 0;
2125
2126 dout(" r_locked_dir = %p\n", req->r_locked_dir);
2127 return 0;
2128 }
2129
2130 /*
2131 * send request, or put it on the appropriate wait list.
2132 */
2133 static int __do_request(struct ceph_mds_client *mdsc,
2134 struct ceph_mds_request *req)
2135 {
2136 struct ceph_mds_session *session = NULL;
2137 int mds = -1;
2138 int err = -EAGAIN;
2139
2140 if (req->r_err || req->r_got_result) {
2141 if (req->r_aborted)
2142 __unregister_request(mdsc, req);
2143 goto out;
2144 }
2145
2146 if (req->r_timeout &&
2147 time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2148 dout("do_request timed out\n");
2149 err = -EIO;
2150 goto finish;
2151 }
2152
2153 put_request_session(req);
2154
2155 mds = __choose_mds(mdsc, req);
2156 if (mds < 0 ||
2157 ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2158 dout("do_request no mds or not active, waiting for map\n");
2159 list_add(&req->r_wait, &mdsc->waiting_for_map);
2160 goto out;
2161 }
2162
2163 /* get, open session */
2164 session = __ceph_lookup_mds_session(mdsc, mds);
2165 if (!session) {
2166 session = register_session(mdsc, mds);
2167 if (IS_ERR(session)) {
2168 err = PTR_ERR(session);
2169 goto finish;
2170 }
2171 }
2172 req->r_session = get_session(session);
2173
2174 dout("do_request mds%d session %p state %s\n", mds, session,
2175 ceph_session_state_name(session->s_state));
2176 if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2177 session->s_state != CEPH_MDS_SESSION_HUNG) {
2178 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2179 session->s_state == CEPH_MDS_SESSION_CLOSING)
2180 __open_session(mdsc, session);
2181 list_add(&req->r_wait, &session->s_waiting);
2182 goto out_session;
2183 }
2184
2185 /* send request */
2186 req->r_resend_mds = -1; /* forget any previous mds hint */
2187
2188 if (req->r_request_started == 0) /* note request start time */
2189 req->r_request_started = jiffies;
2190
2191 err = __prepare_send_request(mdsc, req, mds, false);
2192 if (!err) {
2193 ceph_msg_get(req->r_request);
2194 ceph_con_send(&session->s_con, req->r_request);
2195 }
2196
2197 out_session:
2198 ceph_put_mds_session(session);
2199 out:
2200 return err;
2201
2202 finish:
2203 req->r_err = err;
2204 complete_request(mdsc, req);
2205 goto out;
2206 }
2207
2208 /*
2209 * called under mdsc->mutex
2210 */
2211 static void __wake_requests(struct ceph_mds_client *mdsc,
2212 struct list_head *head)
2213 {
2214 struct ceph_mds_request *req;
2215 LIST_HEAD(tmp_list);
2216
2217 list_splice_init(head, &tmp_list);
2218
2219 while (!list_empty(&tmp_list)) {
2220 req = list_entry(tmp_list.next,
2221 struct ceph_mds_request, r_wait);
2222 list_del_init(&req->r_wait);
2223 dout(" wake request %p tid %llu\n", req, req->r_tid);
2224 __do_request(mdsc, req);
2225 }
2226 }
2227
2228 /*
2229 * Wake up threads with requests pending for @mds, so that they can
2230 * resubmit their requests to a possibly different mds.
2231 */
2232 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2233 {
2234 struct ceph_mds_request *req;
2235 struct rb_node *p = rb_first(&mdsc->request_tree);
2236
2237 dout("kick_requests mds%d\n", mds);
2238 while (p) {
2239 req = rb_entry(p, struct ceph_mds_request, r_node);
2240 p = rb_next(p);
2241 if (req->r_got_unsafe)
2242 continue;
2243 if (req->r_attempts > 0)
2244 continue; /* only new requests */
2245 if (req->r_session &&
2246 req->r_session->s_mds == mds) {
2247 dout(" kicking tid %llu\n", req->r_tid);
2248 list_del_init(&req->r_wait);
2249 __do_request(mdsc, req);
2250 }
2251 }
2252 }
2253
2254 void ceph_mdsc_submit_request(struct ceph_mds_client *mdsc,
2255 struct ceph_mds_request *req)
2256 {
2257 dout("submit_request on %p\n", req);
2258 mutex_lock(&mdsc->mutex);
2259 __register_request(mdsc, req, NULL);
2260 __do_request(mdsc, req);
2261 mutex_unlock(&mdsc->mutex);
2262 }
2263
2264 /*
2265 * Synchrously perform an mds request. Take care of all of the
2266 * session setup, forwarding, retry details.
2267 */
2268 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
2269 struct inode *dir,
2270 struct ceph_mds_request *req)
2271 {
2272 int err;
2273
2274 dout("do_request on %p\n", req);
2275
2276 /* take CAP_PIN refs for r_inode, r_locked_dir, r_old_dentry */
2277 if (req->r_inode)
2278 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2279 if (req->r_locked_dir)
2280 ceph_get_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
2281 if (req->r_old_dentry_dir)
2282 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2283 CEPH_CAP_PIN);
2284
2285 /* issue */
2286 mutex_lock(&mdsc->mutex);
2287 __register_request(mdsc, req, dir);
2288 __do_request(mdsc, req);
2289
2290 if (req->r_err) {
2291 err = req->r_err;
2292 __unregister_request(mdsc, req);
2293 dout("do_request early error %d\n", err);
2294 goto out;
2295 }
2296
2297 /* wait */
2298 mutex_unlock(&mdsc->mutex);
2299 dout("do_request waiting\n");
2300 if (!req->r_timeout && req->r_wait_for_completion) {
2301 err = req->r_wait_for_completion(mdsc, req);
2302 } else {
2303 long timeleft = wait_for_completion_killable_timeout(
2304 &req->r_completion,
2305 ceph_timeout_jiffies(req->r_timeout));
2306 if (timeleft > 0)
2307 err = 0;
2308 else if (!timeleft)
2309 err = -EIO; /* timed out */
2310 else
2311 err = timeleft; /* killed */
2312 }
2313 dout("do_request waited, got %d\n", err);
2314 mutex_lock(&mdsc->mutex);
2315
2316 /* only abort if we didn't race with a real reply */
2317 if (req->r_got_result) {
2318 err = le32_to_cpu(req->r_reply_info.head->result);
2319 } else if (err < 0) {
2320 dout("aborted request %lld with %d\n", req->r_tid, err);
2321
2322 /*
2323 * ensure we aren't running concurrently with
2324 * ceph_fill_trace or ceph_readdir_prepopulate, which
2325 * rely on locks (dir mutex) held by our caller.
2326 */
2327 mutex_lock(&req->r_fill_mutex);
2328 req->r_err = err;
2329 req->r_aborted = true;
2330 mutex_unlock(&req->r_fill_mutex);
2331
2332 if (req->r_locked_dir &&
2333 (req->r_op & CEPH_MDS_OP_WRITE))
2334 ceph_invalidate_dir_request(req);
2335 } else {
2336 err = req->r_err;
2337 }
2338
2339 out:
2340 mutex_unlock(&mdsc->mutex);
2341 dout("do_request %p done, result %d\n", req, err);
2342 return err;
2343 }
2344
2345 /*
2346 * Invalidate dir's completeness, dentry lease state on an aborted MDS
2347 * namespace request.
2348 */
2349 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
2350 {
2351 struct inode *inode = req->r_locked_dir;
2352
2353 dout("invalidate_dir_request %p (complete, lease(s))\n", inode);
2354
2355 ceph_dir_clear_complete(inode);
2356 if (req->r_dentry)
2357 ceph_invalidate_dentry_lease(req->r_dentry);
2358 if (req->r_old_dentry)
2359 ceph_invalidate_dentry_lease(req->r_old_dentry);
2360 }
2361
2362 /*
2363 * Handle mds reply.
2364 *
2365 * We take the session mutex and parse and process the reply immediately.
2366 * This preserves the logical ordering of replies, capabilities, etc., sent
2367 * by the MDS as they are applied to our local cache.
2368 */
2369 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
2370 {
2371 struct ceph_mds_client *mdsc = session->s_mdsc;
2372 struct ceph_mds_request *req;
2373 struct ceph_mds_reply_head *head = msg->front.iov_base;
2374 struct ceph_mds_reply_info_parsed *rinfo; /* parsed reply info */
2375 struct ceph_snap_realm *realm;
2376 u64 tid;
2377 int err, result;
2378 int mds = session->s_mds;
2379
2380 if (msg->front.iov_len < sizeof(*head)) {
2381 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
2382 ceph_msg_dump(msg);
2383 return;
2384 }
2385
2386 /* get request, session */
2387 tid = le64_to_cpu(msg->hdr.tid);
2388 mutex_lock(&mdsc->mutex);
2389 req = __lookup_request(mdsc, tid);
2390 if (!req) {
2391 dout("handle_reply on unknown tid %llu\n", tid);
2392 mutex_unlock(&mdsc->mutex);
2393 return;
2394 }
2395 dout("handle_reply %p\n", req);
2396
2397 /* correct session? */
2398 if (req->r_session != session) {
2399 pr_err("mdsc_handle_reply got %llu on session mds%d"
2400 " not mds%d\n", tid, session->s_mds,
2401 req->r_session ? req->r_session->s_mds : -1);
2402 mutex_unlock(&mdsc->mutex);
2403 goto out;
2404 }
2405
2406 /* dup? */
2407 if ((req->r_got_unsafe && !head->safe) ||
2408 (req->r_got_safe && head->safe)) {
2409 pr_warn("got a dup %s reply on %llu from mds%d\n",
2410 head->safe ? "safe" : "unsafe", tid, mds);
2411 mutex_unlock(&mdsc->mutex);
2412 goto out;
2413 }
2414 if (req->r_got_safe && !head->safe) {
2415 pr_warn("got unsafe after safe on %llu from mds%d\n",
2416 tid, mds);
2417 mutex_unlock(&mdsc->mutex);
2418 goto out;
2419 }
2420
2421 result = le32_to_cpu(head->result);
2422
2423 /*
2424 * Handle an ESTALE
2425 * if we're not talking to the authority, send to them
2426 * if the authority has changed while we weren't looking,
2427 * send to new authority
2428 * Otherwise we just have to return an ESTALE
2429 */
2430 if (result == -ESTALE) {
2431 dout("got ESTALE on request %llu", req->r_tid);
2432 req->r_resend_mds = -1;
2433 if (req->r_direct_mode != USE_AUTH_MDS) {
2434 dout("not using auth, setting for that now");
2435 req->r_direct_mode = USE_AUTH_MDS;
2436 __do_request(mdsc, req);
2437 mutex_unlock(&mdsc->mutex);
2438 goto out;
2439 } else {
2440 int mds = __choose_mds(mdsc, req);
2441 if (mds >= 0 && mds != req->r_session->s_mds) {
2442 dout("but auth changed, so resending");
2443 __do_request(mdsc, req);
2444 mutex_unlock(&mdsc->mutex);
2445 goto out;
2446 }
2447 }
2448 dout("have to return ESTALE on request %llu", req->r_tid);
2449 }
2450
2451
2452 if (head->safe) {
2453 req->r_got_safe = true;
2454 __unregister_request(mdsc, req);
2455
2456 if (req->r_got_unsafe) {
2457 /*
2458 * We already handled the unsafe response, now do the
2459 * cleanup. No need to examine the response; the MDS
2460 * doesn't include any result info in the safe
2461 * response. And even if it did, there is nothing
2462 * useful we could do with a revised return value.
2463 */
2464 dout("got safe reply %llu, mds%d\n", tid, mds);
2465 list_del_init(&req->r_unsafe_item);
2466
2467 /* last unsafe request during umount? */
2468 if (mdsc->stopping && !__get_oldest_req(mdsc))
2469 complete_all(&mdsc->safe_umount_waiters);
2470 mutex_unlock(&mdsc->mutex);
2471 goto out;
2472 }
2473 } else {
2474 req->r_got_unsafe = true;
2475 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
2476 }
2477
2478 dout("handle_reply tid %lld result %d\n", tid, result);
2479 rinfo = &req->r_reply_info;
2480 err = parse_reply_info(msg, rinfo, session->s_con.peer_features);
2481 mutex_unlock(&mdsc->mutex);
2482
2483 mutex_lock(&session->s_mutex);
2484 if (err < 0) {
2485 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
2486 ceph_msg_dump(msg);
2487 goto out_err;
2488 }
2489
2490 /* snap trace */
2491 realm = NULL;
2492 if (rinfo->snapblob_len) {
2493 down_write(&mdsc->snap_rwsem);
2494 ceph_update_snap_trace(mdsc, rinfo->snapblob,
2495 rinfo->snapblob + rinfo->snapblob_len,
2496 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
2497 &realm);
2498 downgrade_write(&mdsc->snap_rwsem);
2499 } else {
2500 down_read(&mdsc->snap_rwsem);
2501 }
2502
2503 /* insert trace into our cache */
2504 mutex_lock(&req->r_fill_mutex);
2505 err = ceph_fill_trace(mdsc->fsc->sb, req, req->r_session);
2506 if (err == 0) {
2507 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
2508 req->r_op == CEPH_MDS_OP_LSSNAP))
2509 ceph_readdir_prepopulate(req, req->r_session);
2510 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
2511 }
2512 mutex_unlock(&req->r_fill_mutex);
2513
2514 up_read(&mdsc->snap_rwsem);
2515 if (realm)
2516 ceph_put_snap_realm(mdsc, realm);
2517 out_err:
2518 mutex_lock(&mdsc->mutex);
2519 if (!req->r_aborted) {
2520 if (err) {
2521 req->r_err = err;
2522 } else {
2523 req->r_reply = msg;
2524 ceph_msg_get(msg);
2525 req->r_got_result = true;
2526 }
2527 } else {
2528 dout("reply arrived after request %lld was aborted\n", tid);
2529 }
2530 mutex_unlock(&mdsc->mutex);
2531
2532 mutex_unlock(&session->s_mutex);
2533
2534 /* kick calling process */
2535 complete_request(mdsc, req);
2536 out:
2537 ceph_mdsc_put_request(req);
2538 return;
2539 }
2540
2541
2542
2543 /*
2544 * handle mds notification that our request has been forwarded.
2545 */
2546 static void handle_forward(struct ceph_mds_client *mdsc,
2547 struct ceph_mds_session *session,
2548 struct ceph_msg *msg)
2549 {
2550 struct ceph_mds_request *req;
2551 u64 tid = le64_to_cpu(msg->hdr.tid);
2552 u32 next_mds;
2553 u32 fwd_seq;
2554 int err = -EINVAL;
2555 void *p = msg->front.iov_base;
2556 void *end = p + msg->front.iov_len;
2557
2558 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
2559 next_mds = ceph_decode_32(&p);
2560 fwd_seq = ceph_decode_32(&p);
2561
2562 mutex_lock(&mdsc->mutex);
2563 req = __lookup_request(mdsc, tid);
2564 if (!req) {
2565 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
2566 goto out; /* dup reply? */
2567 }
2568
2569 if (req->r_aborted) {
2570 dout("forward tid %llu aborted, unregistering\n", tid);
2571 __unregister_request(mdsc, req);
2572 } else if (fwd_seq <= req->r_num_fwd) {
2573 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
2574 tid, next_mds, req->r_num_fwd, fwd_seq);
2575 } else {
2576 /* resend. forward race not possible; mds would drop */
2577 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
2578 BUG_ON(req->r_err);
2579 BUG_ON(req->r_got_result);
2580 req->r_attempts = 0;
2581 req->r_num_fwd = fwd_seq;
2582 req->r_resend_mds = next_mds;
2583 put_request_session(req);
2584 __do_request(mdsc, req);
2585 }
2586 ceph_mdsc_put_request(req);
2587 out:
2588 mutex_unlock(&mdsc->mutex);
2589 return;
2590
2591 bad:
2592 pr_err("mdsc_handle_forward decode error err=%d\n", err);
2593 }
2594
2595 /*
2596 * handle a mds session control message
2597 */
2598 static void handle_session(struct ceph_mds_session *session,
2599 struct ceph_msg *msg)
2600 {
2601 struct ceph_mds_client *mdsc = session->s_mdsc;
2602 u32 op;
2603 u64 seq;
2604 int mds = session->s_mds;
2605 struct ceph_mds_session_head *h = msg->front.iov_base;
2606 int wake = 0;
2607
2608 /* decode */
2609 if (msg->front.iov_len != sizeof(*h))
2610 goto bad;
2611 op = le32_to_cpu(h->op);
2612 seq = le64_to_cpu(h->seq);
2613
2614 mutex_lock(&mdsc->mutex);
2615 if (op == CEPH_SESSION_CLOSE)
2616 __unregister_session(mdsc, session);
2617 /* FIXME: this ttl calculation is generous */
2618 session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
2619 mutex_unlock(&mdsc->mutex);
2620
2621 mutex_lock(&session->s_mutex);
2622
2623 dout("handle_session mds%d %s %p state %s seq %llu\n",
2624 mds, ceph_session_op_name(op), session,
2625 ceph_session_state_name(session->s_state), seq);
2626
2627 if (session->s_state == CEPH_MDS_SESSION_HUNG) {
2628 session->s_state = CEPH_MDS_SESSION_OPEN;
2629 pr_info("mds%d came back\n", session->s_mds);
2630 }
2631
2632 switch (op) {
2633 case CEPH_SESSION_OPEN:
2634 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
2635 pr_info("mds%d reconnect success\n", session->s_mds);
2636 session->s_state = CEPH_MDS_SESSION_OPEN;
2637 renewed_caps(mdsc, session, 0);
2638 wake = 1;
2639 if (mdsc->stopping)
2640 __close_session(mdsc, session);
2641 break;
2642
2643 case CEPH_SESSION_RENEWCAPS:
2644 if (session->s_renew_seq == seq)
2645 renewed_caps(mdsc, session, 1);
2646 break;
2647
2648 case CEPH_SESSION_CLOSE:
2649 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
2650 pr_info("mds%d reconnect denied\n", session->s_mds);
2651 cleanup_session_requests(mdsc, session);
2652 remove_session_caps(session);
2653 wake = 2; /* for good measure */
2654 wake_up_all(&mdsc->session_close_wq);
2655 break;
2656
2657 case CEPH_SESSION_STALE:
2658 pr_info("mds%d caps went stale, renewing\n",
2659 session->s_mds);
2660 spin_lock(&session->s_gen_ttl_lock);
2661 session->s_cap_gen++;
2662 session->s_cap_ttl = jiffies - 1;
2663 spin_unlock(&session->s_gen_ttl_lock);
2664 send_renew_caps(mdsc, session);
2665 break;
2666
2667 case CEPH_SESSION_RECALL_STATE:
2668 trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
2669 break;
2670
2671 case CEPH_SESSION_FLUSHMSG:
2672 send_flushmsg_ack(mdsc, session, seq);
2673 break;
2674
2675 case CEPH_SESSION_FORCE_RO:
2676 dout("force_session_readonly %p\n", session);
2677 spin_lock(&session->s_cap_lock);
2678 session->s_readonly = true;
2679 spin_unlock(&session->s_cap_lock);
2680 wake_up_session_caps(session, 0);
2681 break;
2682
2683 default:
2684 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
2685 WARN_ON(1);
2686 }
2687
2688 mutex_unlock(&session->s_mutex);
2689 if (wake) {
2690 mutex_lock(&mdsc->mutex);
2691 __wake_requests(mdsc, &session->s_waiting);
2692 if (wake == 2)
2693 kick_requests(mdsc, mds);
2694 mutex_unlock(&mdsc->mutex);
2695 }
2696 return;
2697
2698 bad:
2699 pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
2700 (int)msg->front.iov_len);
2701 ceph_msg_dump(msg);
2702 return;
2703 }
2704
2705
2706 /*
2707 * called under session->mutex.
2708 */
2709 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
2710 struct ceph_mds_session *session)
2711 {
2712 struct ceph_mds_request *req, *nreq;
2713 struct rb_node *p;
2714 int err;
2715
2716 dout("replay_unsafe_requests mds%d\n", session->s_mds);
2717
2718 mutex_lock(&mdsc->mutex);
2719 list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item) {
2720 err = __prepare_send_request(mdsc, req, session->s_mds, true);
2721 if (!err) {
2722 ceph_msg_get(req->r_request);
2723 ceph_con_send(&session->s_con, req->r_request);
2724 }
2725 }
2726
2727 /*
2728 * also re-send old requests when MDS enters reconnect stage. So that MDS
2729 * can process completed request in clientreplay stage.
2730 */
2731 p = rb_first(&mdsc->request_tree);
2732 while (p) {
2733 req = rb_entry(p, struct ceph_mds_request, r_node);
2734 p = rb_next(p);
2735 if (req->r_got_unsafe)
2736 continue;
2737 if (req->r_attempts == 0)
2738 continue; /* only old requests */
2739 if (req->r_session &&
2740 req->r_session->s_mds == session->s_mds) {
2741 err = __prepare_send_request(mdsc, req,
2742 session->s_mds, true);
2743 if (!err) {
2744 ceph_msg_get(req->r_request);
2745 ceph_con_send(&session->s_con, req->r_request);
2746 }
2747 }
2748 }
2749 mutex_unlock(&mdsc->mutex);
2750 }
2751
2752 /*
2753 * Encode information about a cap for a reconnect with the MDS.
2754 */
2755 static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
2756 void *arg)
2757 {
2758 union {
2759 struct ceph_mds_cap_reconnect v2;
2760 struct ceph_mds_cap_reconnect_v1 v1;
2761 } rec;
2762 size_t reclen;
2763 struct ceph_inode_info *ci;
2764 struct ceph_reconnect_state *recon_state = arg;
2765 struct ceph_pagelist *pagelist = recon_state->pagelist;
2766 char *path;
2767 int pathlen, err;
2768 u64 pathbase;
2769 struct dentry *dentry;
2770
2771 ci = cap->ci;
2772
2773 dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
2774 inode, ceph_vinop(inode), cap, cap->cap_id,
2775 ceph_cap_string(cap->issued));
2776 err = ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
2777 if (err)
2778 return err;
2779
2780 dentry = d_find_alias(inode);
2781 if (dentry) {
2782 path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase, 0);
2783 if (IS_ERR(path)) {
2784 err = PTR_ERR(path);
2785 goto out_dput;
2786 }
2787 } else {
2788 path = NULL;
2789 pathlen = 0;
2790 }
2791 err = ceph_pagelist_encode_string(pagelist, path, pathlen);
2792 if (err)
2793 goto out_free;
2794
2795 spin_lock(&ci->i_ceph_lock);
2796 cap->seq = 0; /* reset cap seq */
2797 cap->issue_seq = 0; /* and issue_seq */
2798 cap->mseq = 0; /* and migrate_seq */
2799 cap->cap_gen = cap->session->s_cap_gen;
2800
2801 if (recon_state->flock) {
2802 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
2803 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
2804 rec.v2.issued = cpu_to_le32(cap->issued);
2805 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
2806 rec.v2.pathbase = cpu_to_le64(pathbase);
2807 rec.v2.flock_len = 0;
2808 reclen = sizeof(rec.v2);
2809 } else {
2810 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
2811 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
2812 rec.v1.issued = cpu_to_le32(cap->issued);
2813 rec.v1.size = cpu_to_le64(inode->i_size);
2814 ceph_encode_timespec(&rec.v1.mtime, &inode->i_mtime);
2815 ceph_encode_timespec(&rec.v1.atime, &inode->i_atime);
2816 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
2817 rec.v1.pathbase = cpu_to_le64(pathbase);
2818 reclen = sizeof(rec.v1);
2819 }
2820 spin_unlock(&ci->i_ceph_lock);
2821
2822 if (recon_state->flock) {
2823 int num_fcntl_locks, num_flock_locks;
2824 struct ceph_filelock *flocks;
2825
2826 encode_again:
2827 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
2828 flocks = kmalloc((num_fcntl_locks+num_flock_locks) *
2829 sizeof(struct ceph_filelock), GFP_NOFS);
2830 if (!flocks) {
2831 err = -ENOMEM;
2832 goto out_free;
2833 }
2834 err = ceph_encode_locks_to_buffer(inode, flocks,
2835 num_fcntl_locks,
2836 num_flock_locks);
2837 if (err) {
2838 kfree(flocks);
2839 if (err == -ENOSPC)
2840 goto encode_again;
2841 goto out_free;
2842 }
2843 /*
2844 * number of encoded locks is stable, so copy to pagelist
2845 */
2846 rec.v2.flock_len = cpu_to_le32(2*sizeof(u32) +
2847 (num_fcntl_locks+num_flock_locks) *
2848 sizeof(struct ceph_filelock));
2849 err = ceph_pagelist_append(pagelist, &rec, reclen);
2850 if (!err)
2851 err = ceph_locks_to_pagelist(flocks, pagelist,
2852 num_fcntl_locks,
2853 num_flock_locks);
2854 kfree(flocks);
2855 } else {
2856 err = ceph_pagelist_append(pagelist, &rec, reclen);
2857 }
2858
2859 recon_state->nr_caps++;
2860 out_free:
2861 kfree(path);
2862 out_dput:
2863 dput(dentry);
2864 return err;
2865 }
2866
2867
2868 /*
2869 * If an MDS fails and recovers, clients need to reconnect in order to
2870 * reestablish shared state. This includes all caps issued through
2871 * this session _and_ the snap_realm hierarchy. Because it's not
2872 * clear which snap realms the mds cares about, we send everything we
2873 * know about.. that ensures we'll then get any new info the
2874 * recovering MDS might have.
2875 *
2876 * This is a relatively heavyweight operation, but it's rare.
2877 *
2878 * called with mdsc->mutex held.
2879 */
2880 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
2881 struct ceph_mds_session *session)
2882 {
2883 struct ceph_msg *reply;
2884 struct rb_node *p;
2885 int mds = session->s_mds;
2886 int err = -ENOMEM;
2887 int s_nr_caps;
2888 struct ceph_pagelist *pagelist;
2889 struct ceph_reconnect_state recon_state;
2890
2891 pr_info("mds%d reconnect start\n", mds);
2892
2893 pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
2894 if (!pagelist)
2895 goto fail_nopagelist;
2896 ceph_pagelist_init(pagelist);
2897
2898 reply = ceph_msg_new(CEPH_MSG_CLIENT_RECONNECT, 0, GFP_NOFS, false);
2899 if (!reply)
2900 goto fail_nomsg;
2901
2902 mutex_lock(&session->s_mutex);
2903 session->s_state = CEPH_MDS_SESSION_RECONNECTING;
2904 session->s_seq = 0;
2905
2906 dout("session %p state %s\n", session,
2907 ceph_session_state_name(session->s_state));
2908
2909 spin_lock(&session->s_gen_ttl_lock);
2910 session->s_cap_gen++;
2911 spin_unlock(&session->s_gen_ttl_lock);
2912
2913 spin_lock(&session->s_cap_lock);
2914 /* don't know if session is readonly */
2915 session->s_readonly = 0;
2916 /*
2917 * notify __ceph_remove_cap() that we are composing cap reconnect.
2918 * If a cap get released before being added to the cap reconnect,
2919 * __ceph_remove_cap() should skip queuing cap release.
2920 */
2921 session->s_cap_reconnect = 1;
2922 /* drop old cap expires; we're about to reestablish that state */
2923 cleanup_cap_releases(mdsc, session);
2924
2925 /* trim unused caps to reduce MDS's cache rejoin time */
2926 if (mdsc->fsc->sb->s_root)
2927 shrink_dcache_parent(mdsc->fsc->sb->s_root);
2928
2929 ceph_con_close(&session->s_con);
2930 ceph_con_open(&session->s_con,
2931 CEPH_ENTITY_TYPE_MDS, mds,
2932 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
2933
2934 /* replay unsafe requests */
2935 replay_unsafe_requests(mdsc, session);
2936
2937 down_read(&mdsc->snap_rwsem);
2938
2939 /* traverse this session's caps */
2940 s_nr_caps = session->s_nr_caps;
2941 err = ceph_pagelist_encode_32(pagelist, s_nr_caps);
2942 if (err)
2943 goto fail;
2944
2945 recon_state.nr_caps = 0;
2946 recon_state.pagelist = pagelist;
2947 recon_state.flock = session->s_con.peer_features & CEPH_FEATURE_FLOCK;
2948 err = iterate_session_caps(session, encode_caps_cb, &recon_state);
2949 if (err < 0)
2950 goto fail;
2951
2952 spin_lock(&session->s_cap_lock);
2953 session->s_cap_reconnect = 0;
2954 spin_unlock(&session->s_cap_lock);
2955
2956 /*
2957 * snaprealms. we provide mds with the ino, seq (version), and
2958 * parent for all of our realms. If the mds has any newer info,
2959 * it will tell us.
2960 */
2961 for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
2962 struct ceph_snap_realm *realm =
2963 rb_entry(p, struct ceph_snap_realm, node);
2964 struct ceph_mds_snaprealm_reconnect sr_rec;
2965
2966 dout(" adding snap realm %llx seq %lld parent %llx\n",
2967 realm->ino, realm->seq, realm->parent_ino);
2968 sr_rec.ino = cpu_to_le64(realm->ino);
2969 sr_rec.seq = cpu_to_le64(realm->seq);
2970 sr_rec.parent = cpu_to_le64(realm->parent_ino);
2971 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
2972 if (err)
2973 goto fail;
2974 }
2975
2976 if (recon_state.flock)
2977 reply->hdr.version = cpu_to_le16(2);
2978
2979 /* raced with cap release? */
2980 if (s_nr_caps != recon_state.nr_caps) {
2981 struct page *page = list_first_entry(&pagelist->head,
2982 struct page, lru);
2983 __le32 *addr = kmap_atomic(page);
2984 *addr = cpu_to_le32(recon_state.nr_caps);
2985 kunmap_atomic(addr);
2986 }
2987
2988 reply->hdr.data_len = cpu_to_le32(pagelist->length);
2989 ceph_msg_data_add_pagelist(reply, pagelist);
2990
2991 ceph_early_kick_flushing_caps(mdsc, session);
2992
2993 ceph_con_send(&session->s_con, reply);
2994
2995 mutex_unlock(&session->s_mutex);
2996
2997 mutex_lock(&mdsc->mutex);
2998 __wake_requests(mdsc, &session->s_waiting);
2999 mutex_unlock(&mdsc->mutex);
3000
3001 up_read(&mdsc->snap_rwsem);
3002 return;
3003
3004 fail:
3005 ceph_msg_put(reply);
3006 up_read(&mdsc->snap_rwsem);
3007 mutex_unlock(&session->s_mutex);
3008 fail_nomsg:
3009 ceph_pagelist_release(pagelist);
3010 fail_nopagelist:
3011 pr_err("error %d preparing reconnect for mds%d\n", err, mds);
3012 return;
3013 }
3014
3015
3016 /*
3017 * compare old and new mdsmaps, kicking requests
3018 * and closing out old connections as necessary
3019 *
3020 * called under mdsc->mutex.
3021 */
3022 static void check_new_map(struct ceph_mds_client *mdsc,
3023 struct ceph_mdsmap *newmap,
3024 struct ceph_mdsmap *oldmap)
3025 {
3026 int i;
3027 int oldstate, newstate;
3028 struct ceph_mds_session *s;
3029
3030 dout("check_new_map new %u old %u\n",
3031 newmap->m_epoch, oldmap->m_epoch);
3032
3033 for (i = 0; i < oldmap->m_max_mds && i < mdsc->max_sessions; i++) {
3034 if (mdsc->sessions[i] == NULL)
3035 continue;
3036 s = mdsc->sessions[i];
3037 oldstate = ceph_mdsmap_get_state(oldmap, i);
3038 newstate = ceph_mdsmap_get_state(newmap, i);
3039
3040 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
3041 i, ceph_mds_state_name(oldstate),
3042 ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
3043 ceph_mds_state_name(newstate),
3044 ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
3045 ceph_session_state_name(s->s_state));
3046
3047 if (i >= newmap->m_max_mds ||
3048 memcmp(ceph_mdsmap_get_addr(oldmap, i),
3049 ceph_mdsmap_get_addr(newmap, i),
3050 sizeof(struct ceph_entity_addr))) {
3051 if (s->s_state == CEPH_MDS_SESSION_OPENING) {
3052 /* the session never opened, just close it
3053 * out now */
3054 __wake_requests(mdsc, &s->s_waiting);
3055 __unregister_session(mdsc, s);
3056 } else {
3057 /* just close it */
3058 mutex_unlock(&mdsc->mutex);
3059 mutex_lock(&s->s_mutex);
3060 mutex_lock(&mdsc->mutex);
3061 ceph_con_close(&s->s_con);
3062 mutex_unlock(&s->s_mutex);
3063 s->s_state = CEPH_MDS_SESSION_RESTARTING;
3064 }
3065 } else if (oldstate == newstate) {
3066 continue; /* nothing new with this mds */
3067 }
3068
3069 /*
3070 * send reconnect?
3071 */
3072 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
3073 newstate >= CEPH_MDS_STATE_RECONNECT) {
3074 mutex_unlock(&mdsc->mutex);
3075 send_mds_reconnect(mdsc, s);
3076 mutex_lock(&mdsc->mutex);
3077 }
3078
3079 /*
3080 * kick request on any mds that has gone active.
3081 */
3082 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
3083 newstate >= CEPH_MDS_STATE_ACTIVE) {
3084 if (oldstate != CEPH_MDS_STATE_CREATING &&
3085 oldstate != CEPH_MDS_STATE_STARTING)
3086 pr_info("mds%d recovery completed\n", s->s_mds);
3087 kick_requests(mdsc, i);
3088 ceph_kick_flushing_caps(mdsc, s);
3089 wake_up_session_caps(s, 1);
3090 }
3091 }
3092
3093 for (i = 0; i < newmap->m_max_mds && i < mdsc->max_sessions; i++) {
3094 s = mdsc->sessions[i];
3095 if (!s)
3096 continue;
3097 if (!ceph_mdsmap_is_laggy(newmap, i))
3098 continue;
3099 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
3100 s->s_state == CEPH_MDS_SESSION_HUNG ||
3101 s->s_state == CEPH_MDS_SESSION_CLOSING) {
3102 dout(" connecting to export targets of laggy mds%d\n",
3103 i);
3104 __open_export_target_sessions(mdsc, s);
3105 }
3106 }
3107 }
3108
3109
3110
3111 /*
3112 * leases
3113 */
3114
3115 /*
3116 * caller must hold session s_mutex, dentry->d_lock
3117 */
3118 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
3119 {
3120 struct ceph_dentry_info *di = ceph_dentry(dentry);
3121
3122 ceph_put_mds_session(di->lease_session);
3123 di->lease_session = NULL;
3124 }
3125
3126 static void handle_lease(struct ceph_mds_client *mdsc,
3127 struct ceph_mds_session *session,
3128 struct ceph_msg *msg)
3129 {
3130 struct super_block *sb = mdsc->fsc->sb;
3131 struct inode *inode;
3132 struct dentry *parent, *dentry;
3133 struct ceph_dentry_info *di;
3134 int mds = session->s_mds;
3135 struct ceph_mds_lease *h = msg->front.iov_base;
3136 u32 seq;
3137 struct ceph_vino vino;
3138 struct qstr dname;
3139 int release = 0;
3140
3141 dout("handle_lease from mds%d\n", mds);
3142
3143 /* decode */
3144 if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
3145 goto bad;
3146 vino.ino = le64_to_cpu(h->ino);
3147 vino.snap = CEPH_NOSNAP;
3148 seq = le32_to_cpu(h->seq);
3149 dname.name = (void *)h + sizeof(*h) + sizeof(u32);
3150 dname.len = msg->front.iov_len - sizeof(*h) - sizeof(u32);
3151 if (dname.len != get_unaligned_le32(h+1))
3152 goto bad;
3153
3154 /* lookup inode */
3155 inode = ceph_find_inode(sb, vino);
3156 dout("handle_lease %s, ino %llx %p %.*s\n",
3157 ceph_lease_op_name(h->action), vino.ino, inode,
3158 dname.len, dname.name);
3159
3160 mutex_lock(&session->s_mutex);
3161 session->s_seq++;
3162
3163 if (inode == NULL) {
3164 dout("handle_lease no inode %llx\n", vino.ino);
3165 goto release;
3166 }
3167
3168 /* dentry */
3169 parent = d_find_alias(inode);
3170 if (!parent) {
3171 dout("no parent dentry on inode %p\n", inode);
3172 WARN_ON(1);
3173 goto release; /* hrm... */
3174 }
3175 dname.hash = full_name_hash(dname.name, dname.len);
3176 dentry = d_lookup(parent, &dname);
3177 dput(parent);
3178 if (!dentry)
3179 goto release;
3180
3181 spin_lock(&dentry->d_lock);
3182 di = ceph_dentry(dentry);
3183 switch (h->action) {
3184 case CEPH_MDS_LEASE_REVOKE:
3185 if (di->lease_session == session) {
3186 if (ceph_seq_cmp(di->lease_seq, seq) > 0)
3187 h->seq = cpu_to_le32(di->lease_seq);
3188 __ceph_mdsc_drop_dentry_lease(dentry);
3189 }
3190 release = 1;
3191 break;
3192
3193 case CEPH_MDS_LEASE_RENEW:
3194 if (di->lease_session == session &&
3195 di->lease_gen == session->s_cap_gen &&
3196 di->lease_renew_from &&
3197 di->lease_renew_after == 0) {
3198 unsigned long duration =
3199 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
3200
3201 di->lease_seq = seq;
3202 dentry->d_time = di->lease_renew_from + duration;
3203 di->lease_renew_after = di->lease_renew_from +
3204 (duration >> 1);
3205 di->lease_renew_from = 0;
3206 }
3207 break;
3208 }
3209 spin_unlock(&dentry->d_lock);
3210 dput(dentry);
3211
3212 if (!release)
3213 goto out;
3214
3215 release:
3216 /* let's just reuse the same message */
3217 h->action = CEPH_MDS_LEASE_REVOKE_ACK;
3218 ceph_msg_get(msg);
3219 ceph_con_send(&session->s_con, msg);
3220
3221 out:
3222 iput(inode);
3223 mutex_unlock(&session->s_mutex);
3224 return;
3225
3226 bad:
3227 pr_err("corrupt lease message\n");
3228 ceph_msg_dump(msg);
3229 }
3230
3231 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
3232 struct inode *inode,
3233 struct dentry *dentry, char action,
3234 u32 seq)
3235 {
3236 struct ceph_msg *msg;
3237 struct ceph_mds_lease *lease;
3238 int len = sizeof(*lease) + sizeof(u32);
3239 int dnamelen = 0;
3240
3241 dout("lease_send_msg inode %p dentry %p %s to mds%d\n",
3242 inode, dentry, ceph_lease_op_name(action), session->s_mds);
3243 dnamelen = dentry->d_name.len;
3244 len += dnamelen;
3245
3246 msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
3247 if (!msg)
3248 return;
3249 lease = msg->front.iov_base;
3250 lease->action = action;
3251 lease->ino = cpu_to_le64(ceph_vino(inode).ino);
3252 lease->first = lease->last = cpu_to_le64(ceph_vino(inode).snap);
3253 lease->seq = cpu_to_le32(seq);
3254 put_unaligned_le32(dnamelen, lease + 1);
3255 memcpy((void *)(lease + 1) + 4, dentry->d_name.name, dnamelen);
3256
3257 /*
3258 * if this is a preemptive lease RELEASE, no need to
3259 * flush request stream, since the actual request will
3260 * soon follow.
3261 */
3262 msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
3263
3264 ceph_con_send(&session->s_con, msg);
3265 }
3266
3267 /*
3268 * Preemptively release a lease we expect to invalidate anyway.
3269 * Pass @inode always, @dentry is optional.
3270 */
3271 void ceph_mdsc_lease_release(struct ceph_mds_client *mdsc, struct inode *inode,
3272 struct dentry *dentry)
3273 {
3274 struct ceph_dentry_info *di;
3275 struct ceph_mds_session *session;
3276 u32 seq;
3277
3278 BUG_ON(inode == NULL);
3279 BUG_ON(dentry == NULL);
3280
3281 /* is dentry lease valid? */
3282 spin_lock(&dentry->d_lock);
3283 di = ceph_dentry(dentry);
3284 if (!di || !di->lease_session ||
3285 di->lease_session->s_mds < 0 ||
3286 di->lease_gen != di->lease_session->s_cap_gen ||
3287 !time_before(jiffies, dentry->d_time)) {
3288 dout("lease_release inode %p dentry %p -- "
3289 "no lease\n",
3290 inode, dentry);
3291 spin_unlock(&dentry->d_lock);
3292 return;
3293 }
3294
3295 /* we do have a lease on this dentry; note mds and seq */
3296 session = ceph_get_mds_session(di->lease_session);
3297 seq = di->lease_seq;
3298 __ceph_mdsc_drop_dentry_lease(dentry);
3299 spin_unlock(&dentry->d_lock);
3300
3301 dout("lease_release inode %p dentry %p to mds%d\n",
3302 inode, dentry, session->s_mds);
3303 ceph_mdsc_lease_send_msg(session, inode, dentry,
3304 CEPH_MDS_LEASE_RELEASE, seq);
3305 ceph_put_mds_session(session);
3306 }
3307
3308 /*
3309 * drop all leases (and dentry refs) in preparation for umount
3310 */
3311 static void drop_leases(struct ceph_mds_client *mdsc)
3312 {
3313 int i;
3314
3315 dout("drop_leases\n");
3316 mutex_lock(&mdsc->mutex);
3317 for (i = 0; i < mdsc->max_sessions; i++) {
3318 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
3319 if (!s)
3320 continue;
3321 mutex_unlock(&mdsc->mutex);
3322 mutex_lock(&s->s_mutex);
3323 mutex_unlock(&s->s_mutex);
3324 ceph_put_mds_session(s);
3325 mutex_lock(&mdsc->mutex);
3326 }
3327 mutex_unlock(&mdsc->mutex);
3328 }
3329
3330
3331
3332 /*
3333 * delayed work -- periodically trim expired leases, renew caps with mds
3334 */
3335 static void schedule_delayed(struct ceph_mds_client *mdsc)
3336 {
3337 int delay = 5;
3338 unsigned hz = round_jiffies_relative(HZ * delay);
3339 schedule_delayed_work(&mdsc->delayed_work, hz);
3340 }
3341
3342 static void delayed_work(struct work_struct *work)
3343 {
3344 int i;
3345 struct ceph_mds_client *mdsc =
3346 container_of(work, struct ceph_mds_client, delayed_work.work);
3347 int renew_interval;
3348 int renew_caps;
3349
3350 dout("mdsc delayed_work\n");
3351 ceph_check_delayed_caps(mdsc);
3352
3353 mutex_lock(&mdsc->mutex);
3354 renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
3355 renew_caps = time_after_eq(jiffies, HZ*renew_interval +
3356 mdsc->last_renew_caps);
3357 if (renew_caps)
3358 mdsc->last_renew_caps = jiffies;
3359
3360 for (i = 0; i < mdsc->max_sessions; i++) {
3361 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
3362 if (s == NULL)
3363 continue;
3364 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
3365 dout("resending session close request for mds%d\n",
3366 s->s_mds);
3367 request_close_session(mdsc, s);
3368 ceph_put_mds_session(s);
3369 continue;
3370 }
3371 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
3372 if (s->s_state == CEPH_MDS_SESSION_OPEN) {
3373 s->s_state = CEPH_MDS_SESSION_HUNG;
3374 pr_info("mds%d hung\n", s->s_mds);
3375 }
3376 }
3377 if (s->s_state < CEPH_MDS_SESSION_OPEN) {
3378 /* this mds is failed or recovering, just wait */
3379 ceph_put_mds_session(s);
3380 continue;
3381 }
3382 mutex_unlock(&mdsc->mutex);
3383
3384 mutex_lock(&s->s_mutex);
3385 if (renew_caps)
3386 send_renew_caps(mdsc, s);
3387 else
3388 ceph_con_keepalive(&s->s_con);
3389 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
3390 s->s_state == CEPH_MDS_SESSION_HUNG)
3391 ceph_send_cap_releases(mdsc, s);
3392 mutex_unlock(&s->s_mutex);
3393 ceph_put_mds_session(s);
3394
3395 mutex_lock(&mdsc->mutex);
3396 }
3397 mutex_unlock(&mdsc->mutex);
3398
3399 schedule_delayed(mdsc);
3400 }
3401
3402 int ceph_mdsc_init(struct ceph_fs_client *fsc)
3403
3404 {
3405 struct ceph_mds_client *mdsc;
3406
3407 mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
3408 if (!mdsc)
3409 return -ENOMEM;
3410 mdsc->fsc = fsc;
3411 fsc->mdsc = mdsc;
3412 mutex_init(&mdsc->mutex);
3413 mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
3414 if (mdsc->mdsmap == NULL) {
3415 kfree(mdsc);
3416 return -ENOMEM;
3417 }
3418
3419 init_completion(&mdsc->safe_umount_waiters);
3420 init_waitqueue_head(&mdsc->session_close_wq);
3421 INIT_LIST_HEAD(&mdsc->waiting_for_map);
3422 mdsc->sessions = NULL;
3423 atomic_set(&mdsc->num_sessions, 0);
3424 mdsc->max_sessions = 0;
3425 mdsc->stopping = 0;
3426 mdsc->last_snap_seq = 0;
3427 init_rwsem(&mdsc->snap_rwsem);
3428 mdsc->snap_realms = RB_ROOT;
3429 INIT_LIST_HEAD(&mdsc->snap_empty);
3430 spin_lock_init(&mdsc->snap_empty_lock);
3431 mdsc->last_tid = 0;
3432 mdsc->oldest_tid = 0;
3433 mdsc->request_tree = RB_ROOT;
3434 INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
3435 mdsc->last_renew_caps = jiffies;
3436 INIT_LIST_HEAD(&mdsc->cap_delay_list);
3437 spin_lock_init(&mdsc->cap_delay_lock);
3438 INIT_LIST_HEAD(&mdsc->snap_flush_list);
3439 spin_lock_init(&mdsc->snap_flush_lock);
3440 mdsc->last_cap_flush_tid = 1;
3441 mdsc->cap_flush_tree = RB_ROOT;
3442 INIT_LIST_HEAD(&mdsc->cap_dirty);
3443 INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
3444 mdsc->num_cap_flushing = 0;
3445 spin_lock_init(&mdsc->cap_dirty_lock);
3446 init_waitqueue_head(&mdsc->cap_flushing_wq);
3447 spin_lock_init(&mdsc->dentry_lru_lock);
3448 INIT_LIST_HEAD(&mdsc->dentry_lru);
3449
3450 ceph_caps_init(mdsc);
3451 ceph_adjust_min_caps(mdsc, fsc->min_caps);
3452
3453 init_rwsem(&mdsc->pool_perm_rwsem);
3454 mdsc->pool_perm_tree = RB_ROOT;
3455
3456 return 0;
3457 }
3458
3459 /*
3460 * Wait for safe replies on open mds requests. If we time out, drop
3461 * all requests from the tree to avoid dangling dentry refs.
3462 */
3463 static void wait_requests(struct ceph_mds_client *mdsc)
3464 {
3465 struct ceph_options *opts = mdsc->fsc->client->options;
3466 struct ceph_mds_request *req;
3467
3468 mutex_lock(&mdsc->mutex);
3469 if (__get_oldest_req(mdsc)) {
3470 mutex_unlock(&mdsc->mutex);
3471
3472 dout("wait_requests waiting for requests\n");
3473 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
3474 ceph_timeout_jiffies(opts->mount_timeout));
3475
3476 /* tear down remaining requests */
3477 mutex_lock(&mdsc->mutex);
3478 while ((req = __get_oldest_req(mdsc))) {
3479 dout("wait_requests timed out on tid %llu\n",
3480 req->r_tid);
3481 __unregister_request(mdsc, req);
3482 }
3483 }
3484 mutex_unlock(&mdsc->mutex);
3485 dout("wait_requests done\n");
3486 }
3487
3488 /*
3489 * called before mount is ro, and before dentries are torn down.
3490 * (hmm, does this still race with new lookups?)
3491 */
3492 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
3493 {
3494 dout("pre_umount\n");
3495 mdsc->stopping = 1;
3496
3497 drop_leases(mdsc);
3498 ceph_flush_dirty_caps(mdsc);
3499 wait_requests(mdsc);
3500
3501 /*
3502 * wait for reply handlers to drop their request refs and
3503 * their inode/dcache refs
3504 */
3505 ceph_msgr_flush();
3506 }
3507
3508 /*
3509 * wait for all write mds requests to flush.
3510 */
3511 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
3512 {
3513 struct ceph_mds_request *req = NULL, *nextreq;
3514 struct rb_node *n;
3515
3516 mutex_lock(&mdsc->mutex);
3517 dout("wait_unsafe_requests want %lld\n", want_tid);
3518 restart:
3519 req = __get_oldest_req(mdsc);
3520 while (req && req->r_tid <= want_tid) {
3521 /* find next request */
3522 n = rb_next(&req->r_node);
3523 if (n)
3524 nextreq = rb_entry(n, struct ceph_mds_request, r_node);
3525 else
3526 nextreq = NULL;
3527 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
3528 (req->r_op & CEPH_MDS_OP_WRITE)) {
3529 /* write op */
3530 ceph_mdsc_get_request(req);
3531 if (nextreq)
3532 ceph_mdsc_get_request(nextreq);
3533 mutex_unlock(&mdsc->mutex);
3534 dout("wait_unsafe_requests wait on %llu (want %llu)\n",
3535 req->r_tid, want_tid);
3536 wait_for_completion(&req->r_safe_completion);
3537 mutex_lock(&mdsc->mutex);
3538 ceph_mdsc_put_request(req);
3539 if (!nextreq)
3540 break; /* next dne before, so we're done! */
3541 if (RB_EMPTY_NODE(&nextreq->r_node)) {
3542 /* next request was removed from tree */
3543 ceph_mdsc_put_request(nextreq);
3544 goto restart;
3545 }
3546 ceph_mdsc_put_request(nextreq); /* won't go away */
3547 }
3548 req = nextreq;
3549 }
3550 mutex_unlock(&mdsc->mutex);
3551 dout("wait_unsafe_requests done\n");
3552 }
3553
3554 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
3555 {
3556 u64 want_tid, want_flush, want_snap;
3557
3558 if (mdsc->fsc->mount_state == CEPH_MOUNT_SHUTDOWN)
3559 return;
3560
3561 dout("sync\n");
3562 mutex_lock(&mdsc->mutex);
3563 want_tid = mdsc->last_tid;
3564 mutex_unlock(&mdsc->mutex);
3565
3566 ceph_flush_dirty_caps(mdsc);
3567 spin_lock(&mdsc->cap_dirty_lock);
3568 want_flush = mdsc->last_cap_flush_tid;
3569 spin_unlock(&mdsc->cap_dirty_lock);
3570
3571 down_read(&mdsc->snap_rwsem);
3572 want_snap = mdsc->last_snap_seq;
3573 up_read(&mdsc->snap_rwsem);
3574
3575 dout("sync want tid %lld flush_seq %lld snap_seq %lld\n",
3576 want_tid, want_flush, want_snap);
3577
3578 wait_unsafe_requests(mdsc, want_tid);
3579 wait_caps_flush(mdsc, want_flush, want_snap);
3580 }
3581
3582 /*
3583 * true if all sessions are closed, or we force unmount
3584 */
3585 static bool done_closing_sessions(struct ceph_mds_client *mdsc)
3586 {
3587 if (mdsc->fsc->mount_state == CEPH_MOUNT_SHUTDOWN)
3588 return true;
3589 return atomic_read(&mdsc->num_sessions) == 0;
3590 }
3591
3592 /*
3593 * called after sb is ro.
3594 */
3595 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
3596 {
3597 struct ceph_options *opts = mdsc->fsc->client->options;
3598 struct ceph_mds_session *session;
3599 int i;
3600
3601 dout("close_sessions\n");
3602
3603 /* close sessions */
3604 mutex_lock(&mdsc->mutex);
3605 for (i = 0; i < mdsc->max_sessions; i++) {
3606 session = __ceph_lookup_mds_session(mdsc, i);
3607 if (!session)
3608 continue;
3609 mutex_unlock(&mdsc->mutex);
3610 mutex_lock(&session->s_mutex);
3611 __close_session(mdsc, session);
3612 mutex_unlock(&session->s_mutex);
3613 ceph_put_mds_session(session);
3614 mutex_lock(&mdsc->mutex);
3615 }
3616 mutex_unlock(&mdsc->mutex);
3617
3618 dout("waiting for sessions to close\n");
3619 wait_event_timeout(mdsc->session_close_wq, done_closing_sessions(mdsc),
3620 ceph_timeout_jiffies(opts->mount_timeout));
3621
3622 /* tear down remaining sessions */
3623 mutex_lock(&mdsc->mutex);
3624 for (i = 0; i < mdsc->max_sessions; i++) {
3625 if (mdsc->sessions[i]) {
3626 session = get_session(mdsc->sessions[i]);
3627 __unregister_session(mdsc, session);
3628 mutex_unlock(&mdsc->mutex);
3629 mutex_lock(&session->s_mutex);
3630 remove_session_caps(session);
3631 mutex_unlock(&session->s_mutex);
3632 ceph_put_mds_session(session);
3633 mutex_lock(&mdsc->mutex);
3634 }
3635 }
3636 WARN_ON(!list_empty(&mdsc->cap_delay_list));
3637 mutex_unlock(&mdsc->mutex);
3638
3639 ceph_cleanup_empty_realms(mdsc);
3640
3641 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
3642
3643 dout("stopped\n");
3644 }
3645
3646 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
3647 {
3648 dout("stop\n");
3649 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
3650 if (mdsc->mdsmap)
3651 ceph_mdsmap_destroy(mdsc->mdsmap);
3652 kfree(mdsc->sessions);
3653 ceph_caps_finalize(mdsc);
3654 ceph_pool_perm_destroy(mdsc);
3655 }
3656
3657 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
3658 {
3659 struct ceph_mds_client *mdsc = fsc->mdsc;
3660
3661 dout("mdsc_destroy %p\n", mdsc);
3662 ceph_mdsc_stop(mdsc);
3663
3664 /* flush out any connection work with references to us */
3665 ceph_msgr_flush();
3666
3667 fsc->mdsc = NULL;
3668 kfree(mdsc);
3669 dout("mdsc_destroy %p done\n", mdsc);
3670 }
3671
3672
3673 /*
3674 * handle mds map update.
3675 */
3676 void ceph_mdsc_handle_map(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
3677 {
3678 u32 epoch;
3679 u32 maplen;
3680 void *p = msg->front.iov_base;
3681 void *end = p + msg->front.iov_len;
3682 struct ceph_mdsmap *newmap, *oldmap;
3683 struct ceph_fsid fsid;
3684 int err = -EINVAL;
3685
3686 ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
3687 ceph_decode_copy(&p, &fsid, sizeof(fsid));
3688 if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
3689 return;
3690 epoch = ceph_decode_32(&p);
3691 maplen = ceph_decode_32(&p);
3692 dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
3693
3694 /* do we need it? */
3695 ceph_monc_got_mdsmap(&mdsc->fsc->client->monc, epoch);
3696 mutex_lock(&mdsc->mutex);
3697 if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
3698 dout("handle_map epoch %u <= our %u\n",
3699 epoch, mdsc->mdsmap->m_epoch);
3700 mutex_unlock(&mdsc->mutex);
3701 return;
3702 }
3703
3704 newmap = ceph_mdsmap_decode(&p, end);
3705 if (IS_ERR(newmap)) {
3706 err = PTR_ERR(newmap);
3707 goto bad_unlock;
3708 }
3709
3710 /* swap into place */
3711 if (mdsc->mdsmap) {
3712 oldmap = mdsc->mdsmap;
3713 mdsc->mdsmap = newmap;
3714 check_new_map(mdsc, newmap, oldmap);
3715 ceph_mdsmap_destroy(oldmap);
3716 } else {
3717 mdsc->mdsmap = newmap; /* first mds map */
3718 }
3719 mdsc->fsc->sb->s_maxbytes = mdsc->mdsmap->m_max_file_size;
3720
3721 __wake_requests(mdsc, &mdsc->waiting_for_map);
3722
3723 mutex_unlock(&mdsc->mutex);
3724 schedule_delayed(mdsc);
3725 return;
3726
3727 bad_unlock:
3728 mutex_unlock(&mdsc->mutex);
3729 bad:
3730 pr_err("error decoding mdsmap %d\n", err);
3731 return;
3732 }
3733
3734 static struct ceph_connection *con_get(struct ceph_connection *con)
3735 {
3736 struct ceph_mds_session *s = con->private;
3737
3738 if (get_session(s)) {
3739 dout("mdsc con_get %p ok (%d)\n", s, atomic_read(&s->s_ref));
3740 return con;
3741 }
3742 dout("mdsc con_get %p FAIL\n", s);
3743 return NULL;
3744 }
3745
3746 static void con_put(struct ceph_connection *con)
3747 {
3748 struct ceph_mds_session *s = con->private;
3749
3750 dout("mdsc con_put %p (%d)\n", s, atomic_read(&s->s_ref) - 1);
3751 ceph_put_mds_session(s);
3752 }
3753
3754 /*
3755 * if the client is unresponsive for long enough, the mds will kill
3756 * the session entirely.
3757 */
3758 static void peer_reset(struct ceph_connection *con)
3759 {
3760 struct ceph_mds_session *s = con->private;
3761 struct ceph_mds_client *mdsc = s->s_mdsc;
3762
3763 pr_warn("mds%d closed our session\n", s->s_mds);
3764 send_mds_reconnect(mdsc, s);
3765 }
3766
3767 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
3768 {
3769 struct ceph_mds_session *s = con->private;
3770 struct ceph_mds_client *mdsc = s->s_mdsc;
3771 int type = le16_to_cpu(msg->hdr.type);
3772
3773 mutex_lock(&mdsc->mutex);
3774 if (__verify_registered_session(mdsc, s) < 0) {
3775 mutex_unlock(&mdsc->mutex);
3776 goto out;
3777 }
3778 mutex_unlock(&mdsc->mutex);
3779
3780 switch (type) {
3781 case CEPH_MSG_MDS_MAP:
3782 ceph_mdsc_handle_map(mdsc, msg);
3783 break;
3784 case CEPH_MSG_CLIENT_SESSION:
3785 handle_session(s, msg);
3786 break;
3787 case CEPH_MSG_CLIENT_REPLY:
3788 handle_reply(s, msg);
3789 break;
3790 case CEPH_MSG_CLIENT_REQUEST_FORWARD:
3791 handle_forward(mdsc, s, msg);
3792 break;
3793 case CEPH_MSG_CLIENT_CAPS:
3794 ceph_handle_caps(s, msg);
3795 break;
3796 case CEPH_MSG_CLIENT_SNAP:
3797 ceph_handle_snap(mdsc, s, msg);
3798 break;
3799 case CEPH_MSG_CLIENT_LEASE:
3800 handle_lease(mdsc, s, msg);
3801 break;
3802
3803 default:
3804 pr_err("received unknown message type %d %s\n", type,
3805 ceph_msg_type_name(type));
3806 }
3807 out:
3808 ceph_msg_put(msg);
3809 }
3810
3811 /*
3812 * authentication
3813 */
3814
3815 /*
3816 * Note: returned pointer is the address of a structure that's
3817 * managed separately. Caller must *not* attempt to free it.
3818 */
3819 static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
3820 int *proto, int force_new)
3821 {
3822 struct ceph_mds_session *s = con->private;
3823 struct ceph_mds_client *mdsc = s->s_mdsc;
3824 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
3825 struct ceph_auth_handshake *auth = &s->s_auth;
3826
3827 if (force_new && auth->authorizer) {
3828 ceph_auth_destroy_authorizer(ac, auth->authorizer);
3829 auth->authorizer = NULL;
3830 }
3831 if (!auth->authorizer) {
3832 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
3833 auth);
3834 if (ret)
3835 return ERR_PTR(ret);
3836 } else {
3837 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
3838 auth);
3839 if (ret)
3840 return ERR_PTR(ret);
3841 }
3842 *proto = ac->protocol;
3843
3844 return auth;
3845 }
3846
3847
3848 static int verify_authorizer_reply(struct ceph_connection *con, int len)
3849 {
3850 struct ceph_mds_session *s = con->private;
3851 struct ceph_mds_client *mdsc = s->s_mdsc;
3852 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
3853
3854 return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer, len);
3855 }
3856
3857 static int invalidate_authorizer(struct ceph_connection *con)
3858 {
3859 struct ceph_mds_session *s = con->private;
3860 struct ceph_mds_client *mdsc = s->s_mdsc;
3861 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
3862
3863 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
3864
3865 return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
3866 }
3867
3868 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
3869 struct ceph_msg_header *hdr, int *skip)
3870 {
3871 struct ceph_msg *msg;
3872 int type = (int) le16_to_cpu(hdr->type);
3873 int front_len = (int) le32_to_cpu(hdr->front_len);
3874
3875 if (con->in_msg)
3876 return con->in_msg;
3877
3878 *skip = 0;
3879 msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
3880 if (!msg) {
3881 pr_err("unable to allocate msg type %d len %d\n",
3882 type, front_len);
3883 return NULL;
3884 }
3885
3886 return msg;
3887 }
3888
3889 static int sign_message(struct ceph_connection *con, struct ceph_msg *msg)
3890 {
3891 struct ceph_mds_session *s = con->private;
3892 struct ceph_auth_handshake *auth = &s->s_auth;
3893 return ceph_auth_sign_message(auth, msg);
3894 }
3895
3896 static int check_message_signature(struct ceph_connection *con, struct ceph_msg *msg)
3897 {
3898 struct ceph_mds_session *s = con->private;
3899 struct ceph_auth_handshake *auth = &s->s_auth;
3900 return ceph_auth_check_message_signature(auth, msg);
3901 }
3902
3903 static const struct ceph_connection_operations mds_con_ops = {
3904 .get = con_get,
3905 .put = con_put,
3906 .dispatch = dispatch,
3907 .get_authorizer = get_authorizer,
3908 .verify_authorizer_reply = verify_authorizer_reply,
3909 .invalidate_authorizer = invalidate_authorizer,
3910 .peer_reset = peer_reset,
3911 .alloc_msg = mds_alloc_msg,
3912 .sign_message = sign_message,
3913 .check_message_signature = check_message_signature,
3914 };
3915
3916 /* eof */
Linux with added FPC-III support