?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'icc-5.4-rc1' of https://git.linaro.org/people/georgi.djakov/linux into...
[linux/fpc-iii.git] / fs / ceph / mds_client.c
blob920e9f048bd8f4b38e26da4cb53ce5d9fb6bcda4
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3
4 #include <linux/fs.h>
5 #include <linux/wait.h>
6 #include <linux/slab.h>
7 #include <linux/gfp.h>
8 #include <linux/sched.h>
9 #include <linux/debugfs.h>
10 #include <linux/seq_file.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 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
24
25 /*
26 * A cluster of MDS (metadata server) daemons is responsible for
27 * managing the file system namespace (the directory hierarchy and
28 * inodes) and for coordinating shared access to storage. Metadata is
29 * partitioning hierarchically across a number of servers, and that
30 * partition varies over time as the cluster adjusts the distribution
31 * in order to balance load.
32 *
33 * The MDS client is primarily responsible to managing synchronous
34 * metadata requests for operations like open, unlink, and so forth.
35 * If there is a MDS failure, we find out about it when we (possibly
36 * request and) receive a new MDS map, and can resubmit affected
37 * requests.
38 *
39 * For the most part, though, we take advantage of a lossless
40 * communications channel to the MDS, and do not need to worry about
41 * timing out or resubmitting requests.
42 *
43 * We maintain a stateful "session" with each MDS we interact with.
44 * Within each session, we sent periodic heartbeat messages to ensure
45 * any capabilities or leases we have been issues remain valid. If
46 * the session times out and goes stale, our leases and capabilities
47 * are no longer valid.
48 */
49
50 struct ceph_reconnect_state {
51 struct ceph_mds_session *session;
52 int nr_caps, nr_realms;
53 struct ceph_pagelist *pagelist;
54 unsigned msg_version;
55 bool allow_multi;
56 };
57
58 static void __wake_requests(struct ceph_mds_client *mdsc,
59 struct list_head *head);
60 static void ceph_cap_release_work(struct work_struct *work);
61 static void ceph_cap_reclaim_work(struct work_struct *work);
62
63 static const struct ceph_connection_operations mds_con_ops;
64
65
66 /*
67 * mds reply parsing
68 */
69
70 static int parse_reply_info_quota(void **p, void *end,
71 struct ceph_mds_reply_info_in *info)
72 {
73 u8 struct_v, struct_compat;
74 u32 struct_len;
75
76 ceph_decode_8_safe(p, end, struct_v, bad);
77 ceph_decode_8_safe(p, end, struct_compat, bad);
78 /* struct_v is expected to be >= 1. we only
79 * understand encoding with struct_compat == 1. */
80 if (!struct_v || struct_compat != 1)
81 goto bad;
82 ceph_decode_32_safe(p, end, struct_len, bad);
83 ceph_decode_need(p, end, struct_len, bad);
84 end = *p + struct_len;
85 ceph_decode_64_safe(p, end, info->max_bytes, bad);
86 ceph_decode_64_safe(p, end, info->max_files, bad);
87 *p = end;
88 return 0;
89 bad:
90 return -EIO;
91 }
92
93 /*
94 * parse individual inode info
95 */
96 static int parse_reply_info_in(void **p, void *end,
97 struct ceph_mds_reply_info_in *info,
98 u64 features)
99 {
100 int err = 0;
101 u8 struct_v = 0;
102
103 if (features == (u64)-1) {
104 u32 struct_len;
105 u8 struct_compat;
106 ceph_decode_8_safe(p, end, struct_v, bad);
107 ceph_decode_8_safe(p, end, struct_compat, bad);
108 /* struct_v is expected to be >= 1. we only understand
109 * encoding with struct_compat == 1. */
110 if (!struct_v || struct_compat != 1)
111 goto bad;
112 ceph_decode_32_safe(p, end, struct_len, bad);
113 ceph_decode_need(p, end, struct_len, bad);
114 end = *p + struct_len;
115 }
116
117 ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
118 info->in = *p;
119 *p += sizeof(struct ceph_mds_reply_inode) +
120 sizeof(*info->in->fragtree.splits) *
121 le32_to_cpu(info->in->fragtree.nsplits);
122
123 ceph_decode_32_safe(p, end, info->symlink_len, bad);
124 ceph_decode_need(p, end, info->symlink_len, bad);
125 info->symlink = *p;
126 *p += info->symlink_len;
127
128 ceph_decode_copy_safe(p, end, &info->dir_layout,
129 sizeof(info->dir_layout), bad);
130 ceph_decode_32_safe(p, end, info->xattr_len, bad);
131 ceph_decode_need(p, end, info->xattr_len, bad);
132 info->xattr_data = *p;
133 *p += info->xattr_len;
134
135 if (features == (u64)-1) {
136 /* inline data */
137 ceph_decode_64_safe(p, end, info->inline_version, bad);
138 ceph_decode_32_safe(p, end, info->inline_len, bad);
139 ceph_decode_need(p, end, info->inline_len, bad);
140 info->inline_data = *p;
141 *p += info->inline_len;
142 /* quota */
143 err = parse_reply_info_quota(p, end, info);
144 if (err < 0)
145 goto out_bad;
146 /* pool namespace */
147 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
148 if (info->pool_ns_len > 0) {
149 ceph_decode_need(p, end, info->pool_ns_len, bad);
150 info->pool_ns_data = *p;
151 *p += info->pool_ns_len;
152 }
153
154 /* btime */
155 ceph_decode_need(p, end, sizeof(info->btime), bad);
156 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
157
158 /* change attribute */
159 ceph_decode_64_safe(p, end, info->change_attr, bad);
160
161 /* dir pin */
162 if (struct_v >= 2) {
163 ceph_decode_32_safe(p, end, info->dir_pin, bad);
164 } else {
165 info->dir_pin = -ENODATA;
166 }
167
168 /* snapshot birth time, remains zero for v<=2 */
169 if (struct_v >= 3) {
170 ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
171 ceph_decode_copy(p, &info->snap_btime,
172 sizeof(info->snap_btime));
173 } else {
174 memset(&info->snap_btime, 0, sizeof(info->snap_btime));
175 }
176
177 *p = end;
178 } else {
179 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
180 ceph_decode_64_safe(p, end, info->inline_version, bad);
181 ceph_decode_32_safe(p, end, info->inline_len, bad);
182 ceph_decode_need(p, end, info->inline_len, bad);
183 info->inline_data = *p;
184 *p += info->inline_len;
185 } else
186 info->inline_version = CEPH_INLINE_NONE;
187
188 if (features & CEPH_FEATURE_MDS_QUOTA) {
189 err = parse_reply_info_quota(p, end, info);
190 if (err < 0)
191 goto out_bad;
192 } else {
193 info->max_bytes = 0;
194 info->max_files = 0;
195 }
196
197 info->pool_ns_len = 0;
198 info->pool_ns_data = NULL;
199 if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
200 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
201 if (info->pool_ns_len > 0) {
202 ceph_decode_need(p, end, info->pool_ns_len, bad);
203 info->pool_ns_data = *p;
204 *p += info->pool_ns_len;
205 }
206 }
207
208 if (features & CEPH_FEATURE_FS_BTIME) {
209 ceph_decode_need(p, end, sizeof(info->btime), bad);
210 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
211 ceph_decode_64_safe(p, end, info->change_attr, bad);
212 }
213
214 info->dir_pin = -ENODATA;
215 /* info->snap_btime remains zero */
216 }
217 return 0;
218 bad:
219 err = -EIO;
220 out_bad:
221 return err;
222 }
223
224 static int parse_reply_info_dir(void **p, void *end,
225 struct ceph_mds_reply_dirfrag **dirfrag,
226 u64 features)
227 {
228 if (features == (u64)-1) {
229 u8 struct_v, struct_compat;
230 u32 struct_len;
231 ceph_decode_8_safe(p, end, struct_v, bad);
232 ceph_decode_8_safe(p, end, struct_compat, bad);
233 /* struct_v is expected to be >= 1. we only understand
234 * encoding whose struct_compat == 1. */
235 if (!struct_v || struct_compat != 1)
236 goto bad;
237 ceph_decode_32_safe(p, end, struct_len, bad);
238 ceph_decode_need(p, end, struct_len, bad);
239 end = *p + struct_len;
240 }
241
242 ceph_decode_need(p, end, sizeof(**dirfrag), bad);
243 *dirfrag = *p;
244 *p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
245 if (unlikely(*p > end))
246 goto bad;
247 if (features == (u64)-1)
248 *p = end;
249 return 0;
250 bad:
251 return -EIO;
252 }
253
254 static int parse_reply_info_lease(void **p, void *end,
255 struct ceph_mds_reply_lease **lease,
256 u64 features)
257 {
258 if (features == (u64)-1) {
259 u8 struct_v, struct_compat;
260 u32 struct_len;
261 ceph_decode_8_safe(p, end, struct_v, bad);
262 ceph_decode_8_safe(p, end, struct_compat, bad);
263 /* struct_v is expected to be >= 1. we only understand
264 * encoding whose struct_compat == 1. */
265 if (!struct_v || struct_compat != 1)
266 goto bad;
267 ceph_decode_32_safe(p, end, struct_len, bad);
268 ceph_decode_need(p, end, struct_len, bad);
269 end = *p + struct_len;
270 }
271
272 ceph_decode_need(p, end, sizeof(**lease), bad);
273 *lease = *p;
274 *p += sizeof(**lease);
275 if (features == (u64)-1)
276 *p = end;
277 return 0;
278 bad:
279 return -EIO;
280 }
281
282 /*
283 * parse a normal reply, which may contain a (dir+)dentry and/or a
284 * target inode.
285 */
286 static int parse_reply_info_trace(void **p, void *end,
287 struct ceph_mds_reply_info_parsed *info,
288 u64 features)
289 {
290 int err;
291
292 if (info->head->is_dentry) {
293 err = parse_reply_info_in(p, end, &info->diri, features);
294 if (err < 0)
295 goto out_bad;
296
297 err = parse_reply_info_dir(p, end, &info->dirfrag, features);
298 if (err < 0)
299 goto out_bad;
300
301 ceph_decode_32_safe(p, end, info->dname_len, bad);
302 ceph_decode_need(p, end, info->dname_len, bad);
303 info->dname = *p;
304 *p += info->dname_len;
305
306 err = parse_reply_info_lease(p, end, &info->dlease, features);
307 if (err < 0)
308 goto out_bad;
309 }
310
311 if (info->head->is_target) {
312 err = parse_reply_info_in(p, end, &info->targeti, features);
313 if (err < 0)
314 goto out_bad;
315 }
316
317 if (unlikely(*p != end))
318 goto bad;
319 return 0;
320
321 bad:
322 err = -EIO;
323 out_bad:
324 pr_err("problem parsing mds trace %d\n", err);
325 return err;
326 }
327
328 /*
329 * parse readdir results
330 */
331 static int parse_reply_info_readdir(void **p, void *end,
332 struct ceph_mds_reply_info_parsed *info,
333 u64 features)
334 {
335 u32 num, i = 0;
336 int err;
337
338 err = parse_reply_info_dir(p, end, &info->dir_dir, features);
339 if (err < 0)
340 goto out_bad;
341
342 ceph_decode_need(p, end, sizeof(num) + 2, bad);
343 num = ceph_decode_32(p);
344 {
345 u16 flags = ceph_decode_16(p);
346 info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
347 info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
348 info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
349 info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
350 }
351 if (num == 0)
352 goto done;
353
354 BUG_ON(!info->dir_entries);
355 if ((unsigned long)(info->dir_entries + num) >
356 (unsigned long)info->dir_entries + info->dir_buf_size) {
357 pr_err("dir contents are larger than expected\n");
358 WARN_ON(1);
359 goto bad;
360 }
361
362 info->dir_nr = num;
363 while (num) {
364 struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
365 /* dentry */
366 ceph_decode_32_safe(p, end, rde->name_len, bad);
367 ceph_decode_need(p, end, rde->name_len, bad);
368 rde->name = *p;
369 *p += rde->name_len;
370 dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);
371
372 /* dentry lease */
373 err = parse_reply_info_lease(p, end, &rde->lease, features);
374 if (err)
375 goto out_bad;
376 /* inode */
377 err = parse_reply_info_in(p, end, &rde->inode, features);
378 if (err < 0)
379 goto out_bad;
380 /* ceph_readdir_prepopulate() will update it */
381 rde->offset = 0;
382 i++;
383 num--;
384 }
385
386 done:
387 if (*p != end)
388 goto bad;
389 return 0;
390
391 bad:
392 err = -EIO;
393 out_bad:
394 pr_err("problem parsing dir contents %d\n", err);
395 return err;
396 }
397
398 /*
399 * parse fcntl F_GETLK results
400 */
401 static int parse_reply_info_filelock(void **p, void *end,
402 struct ceph_mds_reply_info_parsed *info,
403 u64 features)
404 {
405 if (*p + sizeof(*info->filelock_reply) > end)
406 goto bad;
407
408 info->filelock_reply = *p;
409 *p += sizeof(*info->filelock_reply);
410
411 if (unlikely(*p != end))
412 goto bad;
413 return 0;
414
415 bad:
416 return -EIO;
417 }
418
419 /*
420 * parse create results
421 */
422 static int parse_reply_info_create(void **p, void *end,
423 struct ceph_mds_reply_info_parsed *info,
424 u64 features)
425 {
426 if (features == (u64)-1 ||
427 (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
428 if (*p == end) {
429 info->has_create_ino = false;
430 } else {
431 info->has_create_ino = true;
432 info->ino = ceph_decode_64(p);
433 }
434 }
435
436 if (unlikely(*p != end))
437 goto bad;
438 return 0;
439
440 bad:
441 return -EIO;
442 }
443
444 /*
445 * parse extra results
446 */
447 static int parse_reply_info_extra(void **p, void *end,
448 struct ceph_mds_reply_info_parsed *info,
449 u64 features)
450 {
451 u32 op = le32_to_cpu(info->head->op);
452
453 if (op == CEPH_MDS_OP_GETFILELOCK)
454 return parse_reply_info_filelock(p, end, info, features);
455 else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
456 return parse_reply_info_readdir(p, end, info, features);
457 else if (op == CEPH_MDS_OP_CREATE)
458 return parse_reply_info_create(p, end, info, features);
459 else
460 return -EIO;
461 }
462
463 /*
464 * parse entire mds reply
465 */
466 static int parse_reply_info(struct ceph_msg *msg,
467 struct ceph_mds_reply_info_parsed *info,
468 u64 features)
469 {
470 void *p, *end;
471 u32 len;
472 int err;
473
474 info->head = msg->front.iov_base;
475 p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
476 end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
477
478 /* trace */
479 ceph_decode_32_safe(&p, end, len, bad);
480 if (len > 0) {
481 ceph_decode_need(&p, end, len, bad);
482 err = parse_reply_info_trace(&p, p+len, info, features);
483 if (err < 0)
484 goto out_bad;
485 }
486
487 /* extra */
488 ceph_decode_32_safe(&p, end, len, bad);
489 if (len > 0) {
490 ceph_decode_need(&p, end, len, bad);
491 err = parse_reply_info_extra(&p, p+len, info, features);
492 if (err < 0)
493 goto out_bad;
494 }
495
496 /* snap blob */
497 ceph_decode_32_safe(&p, end, len, bad);
498 info->snapblob_len = len;
499 info->snapblob = p;
500 p += len;
501
502 if (p != end)
503 goto bad;
504 return 0;
505
506 bad:
507 err = -EIO;
508 out_bad:
509 pr_err("mds parse_reply err %d\n", err);
510 return err;
511 }
512
513 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
514 {
515 if (!info->dir_entries)
516 return;
517 free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
518 }
519
520
521 /*
522 * sessions
523 */
524 const char *ceph_session_state_name(int s)
525 {
526 switch (s) {
527 case CEPH_MDS_SESSION_NEW: return "new";
528 case CEPH_MDS_SESSION_OPENING: return "opening";
529 case CEPH_MDS_SESSION_OPEN: return "open";
530 case CEPH_MDS_SESSION_HUNG: return "hung";
531 case CEPH_MDS_SESSION_CLOSING: return "closing";
532 case CEPH_MDS_SESSION_RESTARTING: return "restarting";
533 case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
534 case CEPH_MDS_SESSION_REJECTED: return "rejected";
535 default: return "???";
536 }
537 }
538
539 static struct ceph_mds_session *get_session(struct ceph_mds_session *s)
540 {
541 if (refcount_inc_not_zero(&s->s_ref)) {
542 dout("mdsc get_session %p %d -> %d\n", s,
543 refcount_read(&s->s_ref)-1, refcount_read(&s->s_ref));
544 return s;
545 } else {
546 dout("mdsc get_session %p 0 -- FAIL\n", s);
547 return NULL;
548 }
549 }
550
551 void ceph_put_mds_session(struct ceph_mds_session *s)
552 {
553 dout("mdsc put_session %p %d -> %d\n", s,
554 refcount_read(&s->s_ref), refcount_read(&s->s_ref)-1);
555 if (refcount_dec_and_test(&s->s_ref)) {
556 if (s->s_auth.authorizer)
557 ceph_auth_destroy_authorizer(s->s_auth.authorizer);
558 kfree(s);
559 }
560 }
561
562 /*
563 * called under mdsc->mutex
564 */
565 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
566 int mds)
567 {
568 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
569 return NULL;
570 return get_session(mdsc->sessions[mds]);
571 }
572
573 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
574 {
575 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
576 return false;
577 else
578 return true;
579 }
580
581 static int __verify_registered_session(struct ceph_mds_client *mdsc,
582 struct ceph_mds_session *s)
583 {
584 if (s->s_mds >= mdsc->max_sessions ||
585 mdsc->sessions[s->s_mds] != s)
586 return -ENOENT;
587 return 0;
588 }
589
590 /*
591 * create+register a new session for given mds.
592 * called under mdsc->mutex.
593 */
594 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
595 int mds)
596 {
597 struct ceph_mds_session *s;
598
599 if (mds >= mdsc->mdsmap->m_num_mds)
600 return ERR_PTR(-EINVAL);
601
602 s = kzalloc(sizeof(*s), GFP_NOFS);
603 if (!s)
604 return ERR_PTR(-ENOMEM);
605
606 if (mds >= mdsc->max_sessions) {
607 int newmax = 1 << get_count_order(mds + 1);
608 struct ceph_mds_session **sa;
609
610 dout("%s: realloc to %d\n", __func__, newmax);
611 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
612 if (!sa)
613 goto fail_realloc;
614 if (mdsc->sessions) {
615 memcpy(sa, mdsc->sessions,
616 mdsc->max_sessions * sizeof(void *));
617 kfree(mdsc->sessions);
618 }
619 mdsc->sessions = sa;
620 mdsc->max_sessions = newmax;
621 }
622
623 dout("%s: mds%d\n", __func__, mds);
624 s->s_mdsc = mdsc;
625 s->s_mds = mds;
626 s->s_state = CEPH_MDS_SESSION_NEW;
627 s->s_ttl = 0;
628 s->s_seq = 0;
629 mutex_init(&s->s_mutex);
630
631 ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
632
633 spin_lock_init(&s->s_gen_ttl_lock);
634 s->s_cap_gen = 1;
635 s->s_cap_ttl = jiffies - 1;
636
637 spin_lock_init(&s->s_cap_lock);
638 s->s_renew_requested = 0;
639 s->s_renew_seq = 0;
640 INIT_LIST_HEAD(&s->s_caps);
641 s->s_nr_caps = 0;
642 s->s_trim_caps = 0;
643 refcount_set(&s->s_ref, 1);
644 INIT_LIST_HEAD(&s->s_waiting);
645 INIT_LIST_HEAD(&s->s_unsafe);
646 s->s_num_cap_releases = 0;
647 s->s_cap_reconnect = 0;
648 s->s_cap_iterator = NULL;
649 INIT_LIST_HEAD(&s->s_cap_releases);
650 INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
651
652 INIT_LIST_HEAD(&s->s_cap_flushing);
653
654 mdsc->sessions[mds] = s;
655 atomic_inc(&mdsc->num_sessions);
656 refcount_inc(&s->s_ref); /* one ref to sessions[], one to caller */
657
658 ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
659 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
660
661 return s;
662
663 fail_realloc:
664 kfree(s);
665 return ERR_PTR(-ENOMEM);
666 }
667
668 /*
669 * called under mdsc->mutex
670 */
671 static void __unregister_session(struct ceph_mds_client *mdsc,
672 struct ceph_mds_session *s)
673 {
674 dout("__unregister_session mds%d %p\n", s->s_mds, s);
675 BUG_ON(mdsc->sessions[s->s_mds] != s);
676 mdsc->sessions[s->s_mds] = NULL;
677 s->s_state = 0;
678 ceph_con_close(&s->s_con);
679 ceph_put_mds_session(s);
680 atomic_dec(&mdsc->num_sessions);
681 }
682
683 /*
684 * drop session refs in request.
685 *
686 * should be last request ref, or hold mdsc->mutex
687 */
688 static void put_request_session(struct ceph_mds_request *req)
689 {
690 if (req->r_session) {
691 ceph_put_mds_session(req->r_session);
692 req->r_session = NULL;
693 }
694 }
695
696 void ceph_mdsc_release_request(struct kref *kref)
697 {
698 struct ceph_mds_request *req = container_of(kref,
699 struct ceph_mds_request,
700 r_kref);
701 destroy_reply_info(&req->r_reply_info);
702 if (req->r_request)
703 ceph_msg_put(req->r_request);
704 if (req->r_reply)
705 ceph_msg_put(req->r_reply);
706 if (req->r_inode) {
707 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
708 /* avoid calling iput_final() in mds dispatch threads */
709 ceph_async_iput(req->r_inode);
710 }
711 if (req->r_parent)
712 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
713 ceph_async_iput(req->r_target_inode);
714 if (req->r_dentry)
715 dput(req->r_dentry);
716 if (req->r_old_dentry)
717 dput(req->r_old_dentry);
718 if (req->r_old_dentry_dir) {
719 /*
720 * track (and drop pins for) r_old_dentry_dir
721 * separately, since r_old_dentry's d_parent may have
722 * changed between the dir mutex being dropped and
723 * this request being freed.
724 */
725 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
726 CEPH_CAP_PIN);
727 ceph_async_iput(req->r_old_dentry_dir);
728 }
729 kfree(req->r_path1);
730 kfree(req->r_path2);
731 if (req->r_pagelist)
732 ceph_pagelist_release(req->r_pagelist);
733 put_request_session(req);
734 ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
735 WARN_ON_ONCE(!list_empty(&req->r_wait));
736 kfree(req);
737 }
738
739 DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
740
741 /*
742 * lookup session, bump ref if found.
743 *
744 * called under mdsc->mutex.
745 */
746 static struct ceph_mds_request *
747 lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
748 {
749 struct ceph_mds_request *req;
750
751 req = lookup_request(&mdsc->request_tree, tid);
752 if (req)
753 ceph_mdsc_get_request(req);
754
755 return req;
756 }
757
758 /*
759 * Register an in-flight request, and assign a tid. Link to directory
760 * are modifying (if any).
761 *
762 * Called under mdsc->mutex.
763 */
764 static void __register_request(struct ceph_mds_client *mdsc,
765 struct ceph_mds_request *req,
766 struct inode *dir)
767 {
768 int ret = 0;
769
770 req->r_tid = ++mdsc->last_tid;
771 if (req->r_num_caps) {
772 ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
773 req->r_num_caps);
774 if (ret < 0) {
775 pr_err("__register_request %p "
776 "failed to reserve caps: %d\n", req, ret);
777 /* set req->r_err to fail early from __do_request */
778 req->r_err = ret;
779 return;
780 }
781 }
782 dout("__register_request %p tid %lld\n", req, req->r_tid);
783 ceph_mdsc_get_request(req);
784 insert_request(&mdsc->request_tree, req);
785
786 req->r_uid = current_fsuid();
787 req->r_gid = current_fsgid();
788
789 if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
790 mdsc->oldest_tid = req->r_tid;
791
792 if (dir) {
793 ihold(dir);
794 req->r_unsafe_dir = dir;
795 }
796 }
797
798 static void __unregister_request(struct ceph_mds_client *mdsc,
799 struct ceph_mds_request *req)
800 {
801 dout("__unregister_request %p tid %lld\n", req, req->r_tid);
802
803 /* Never leave an unregistered request on an unsafe list! */
804 list_del_init(&req->r_unsafe_item);
805
806 if (req->r_tid == mdsc->oldest_tid) {
807 struct rb_node *p = rb_next(&req->r_node);
808 mdsc->oldest_tid = 0;
809 while (p) {
810 struct ceph_mds_request *next_req =
811 rb_entry(p, struct ceph_mds_request, r_node);
812 if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
813 mdsc->oldest_tid = next_req->r_tid;
814 break;
815 }
816 p = rb_next(p);
817 }
818 }
819
820 erase_request(&mdsc->request_tree, req);
821
822 if (req->r_unsafe_dir &&
823 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
824 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
825 spin_lock(&ci->i_unsafe_lock);
826 list_del_init(&req->r_unsafe_dir_item);
827 spin_unlock(&ci->i_unsafe_lock);
828 }
829 if (req->r_target_inode &&
830 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
831 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
832 spin_lock(&ci->i_unsafe_lock);
833 list_del_init(&req->r_unsafe_target_item);
834 spin_unlock(&ci->i_unsafe_lock);
835 }
836
837 if (req->r_unsafe_dir) {
838 /* avoid calling iput_final() in mds dispatch threads */
839 ceph_async_iput(req->r_unsafe_dir);
840 req->r_unsafe_dir = NULL;
841 }
842
843 complete_all(&req->r_safe_completion);
844
845 ceph_mdsc_put_request(req);
846 }
847
848 /*
849 * Walk back up the dentry tree until we hit a dentry representing a
850 * non-snapshot inode. We do this using the rcu_read_lock (which must be held
851 * when calling this) to ensure that the objects won't disappear while we're
852 * working with them. Once we hit a candidate dentry, we attempt to take a
853 * reference to it, and return that as the result.
854 */
855 static struct inode *get_nonsnap_parent(struct dentry *dentry)
856 {
857 struct inode *inode = NULL;
858
859 while (dentry && !IS_ROOT(dentry)) {
860 inode = d_inode_rcu(dentry);
861 if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
862 break;
863 dentry = dentry->d_parent;
864 }
865 if (inode)
866 inode = igrab(inode);
867 return inode;
868 }
869
870 /*
871 * Choose mds to send request to next. If there is a hint set in the
872 * request (e.g., due to a prior forward hint from the mds), use that.
873 * Otherwise, consult frag tree and/or caps to identify the
874 * appropriate mds. If all else fails, choose randomly.
875 *
876 * Called under mdsc->mutex.
877 */
878 static int __choose_mds(struct ceph_mds_client *mdsc,
879 struct ceph_mds_request *req)
880 {
881 struct inode *inode;
882 struct ceph_inode_info *ci;
883 struct ceph_cap *cap;
884 int mode = req->r_direct_mode;
885 int mds = -1;
886 u32 hash = req->r_direct_hash;
887 bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
888
889 /*
890 * is there a specific mds we should try? ignore hint if we have
891 * no session and the mds is not up (active or recovering).
892 */
893 if (req->r_resend_mds >= 0 &&
894 (__have_session(mdsc, req->r_resend_mds) ||
895 ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
896 dout("choose_mds using resend_mds mds%d\n",
897 req->r_resend_mds);
898 return req->r_resend_mds;
899 }
900
901 if (mode == USE_RANDOM_MDS)
902 goto random;
903
904 inode = NULL;
905 if (req->r_inode) {
906 if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
907 inode = req->r_inode;
908 ihold(inode);
909 } else {
910 /* req->r_dentry is non-null for LSSNAP request */
911 rcu_read_lock();
912 inode = get_nonsnap_parent(req->r_dentry);
913 rcu_read_unlock();
914 dout("__choose_mds using snapdir's parent %p\n", inode);
915 }
916 } else if (req->r_dentry) {
917 /* ignore race with rename; old or new d_parent is okay */
918 struct dentry *parent;
919 struct inode *dir;
920
921 rcu_read_lock();
922 parent = READ_ONCE(req->r_dentry->d_parent);
923 dir = req->r_parent ? : d_inode_rcu(parent);
924
925 if (!dir || dir->i_sb != mdsc->fsc->sb) {
926 /* not this fs or parent went negative */
927 inode = d_inode(req->r_dentry);
928 if (inode)
929 ihold(inode);
930 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
931 /* direct snapped/virtual snapdir requests
932 * based on parent dir inode */
933 inode = get_nonsnap_parent(parent);
934 dout("__choose_mds using nonsnap parent %p\n", inode);
935 } else {
936 /* dentry target */
937 inode = d_inode(req->r_dentry);
938 if (!inode || mode == USE_AUTH_MDS) {
939 /* dir + name */
940 inode = igrab(dir);
941 hash = ceph_dentry_hash(dir, req->r_dentry);
942 is_hash = true;
943 } else {
944 ihold(inode);
945 }
946 }
947 rcu_read_unlock();
948 }
949
950 dout("__choose_mds %p is_hash=%d (%d) mode %d\n", inode, (int)is_hash,
951 (int)hash, mode);
952 if (!inode)
953 goto random;
954 ci = ceph_inode(inode);
955
956 if (is_hash && S_ISDIR(inode->i_mode)) {
957 struct ceph_inode_frag frag;
958 int found;
959
960 ceph_choose_frag(ci, hash, &frag, &found);
961 if (found) {
962 if (mode == USE_ANY_MDS && frag.ndist > 0) {
963 u8 r;
964
965 /* choose a random replica */
966 get_random_bytes(&r, 1);
967 r %= frag.ndist;
968 mds = frag.dist[r];
969 dout("choose_mds %p %llx.%llx "
970 "frag %u mds%d (%d/%d)\n",
971 inode, ceph_vinop(inode),
972 frag.frag, mds,
973 (int)r, frag.ndist);
974 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
975 CEPH_MDS_STATE_ACTIVE)
976 goto out;
977 }
978
979 /* since this file/dir wasn't known to be
980 * replicated, then we want to look for the
981 * authoritative mds. */
982 mode = USE_AUTH_MDS;
983 if (frag.mds >= 0) {
984 /* choose auth mds */
985 mds = frag.mds;
986 dout("choose_mds %p %llx.%llx "
987 "frag %u mds%d (auth)\n",
988 inode, ceph_vinop(inode), frag.frag, mds);
989 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
990 CEPH_MDS_STATE_ACTIVE)
991 goto out;
992 }
993 }
994 }
995
996 spin_lock(&ci->i_ceph_lock);
997 cap = NULL;
998 if (mode == USE_AUTH_MDS)
999 cap = ci->i_auth_cap;
1000 if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1001 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1002 if (!cap) {
1003 spin_unlock(&ci->i_ceph_lock);
1004 ceph_async_iput(inode);
1005 goto random;
1006 }
1007 mds = cap->session->s_mds;
1008 dout("choose_mds %p %llx.%llx mds%d (%scap %p)\n",
1009 inode, ceph_vinop(inode), mds,
1010 cap == ci->i_auth_cap ? "auth " : "", cap);
1011 spin_unlock(&ci->i_ceph_lock);
1012 out:
1013 /* avoid calling iput_final() while holding mdsc->mutex or
1014 * in mds dispatch threads */
1015 ceph_async_iput(inode);
1016 return mds;
1017
1018 random:
1019 mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1020 dout("choose_mds chose random mds%d\n", mds);
1021 return mds;
1022 }
1023
1024
1025 /*
1026 * session messages
1027 */
1028 static struct ceph_msg *create_session_msg(u32 op, u64 seq)
1029 {
1030 struct ceph_msg *msg;
1031 struct ceph_mds_session_head *h;
1032
1033 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1034 false);
1035 if (!msg) {
1036 pr_err("create_session_msg ENOMEM creating msg\n");
1037 return NULL;
1038 }
1039 h = msg->front.iov_base;
1040 h->op = cpu_to_le32(op);
1041 h->seq = cpu_to_le64(seq);
1042
1043 return msg;
1044 }
1045
1046 static void encode_supported_features(void **p, void *end)
1047 {
1048 static const unsigned char bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1049 static const size_t count = ARRAY_SIZE(bits);
1050
1051 if (count > 0) {
1052 size_t i;
1053 size_t size = ((size_t)bits[count - 1] + 64) / 64 * 8;
1054
1055 BUG_ON(*p + 4 + size > end);
1056 ceph_encode_32(p, size);
1057 memset(*p, 0, size);
1058 for (i = 0; i < count; i++)
1059 ((unsigned char*)(*p))[i / 8] |= 1 << (bits[i] % 8);
1060 *p += size;
1061 } else {
1062 BUG_ON(*p + 4 > end);
1063 ceph_encode_32(p, 0);
1064 }
1065 }
1066
1067 /*
1068 * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1069 * to include additional client metadata fields.
1070 */
1071 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1072 {
1073 struct ceph_msg *msg;
1074 struct ceph_mds_session_head *h;
1075 int i = -1;
1076 int extra_bytes = 0;
1077 int metadata_key_count = 0;
1078 struct ceph_options *opt = mdsc->fsc->client->options;
1079 struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1080 void *p, *end;
1081
1082 const char* metadata[][2] = {
1083 {"hostname", mdsc->nodename},
1084 {"kernel_version", init_utsname()->release},
1085 {"entity_id", opt->name ? : ""},
1086 {"root", fsopt->server_path ? : "/"},
1087 {NULL, NULL}
1088 };
1089
1090 /* Calculate serialized length of metadata */
1091 extra_bytes = 4; /* map length */
1092 for (i = 0; metadata[i][0]; ++i) {
1093 extra_bytes += 8 + strlen(metadata[i][0]) +
1094 strlen(metadata[i][1]);
1095 metadata_key_count++;
1096 }
1097 /* supported feature */
1098 extra_bytes += 4 + 8;
1099
1100 /* Allocate the message */
1101 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1102 GFP_NOFS, false);
1103 if (!msg) {
1104 pr_err("create_session_msg ENOMEM creating msg\n");
1105 return NULL;
1106 }
1107 p = msg->front.iov_base;
1108 end = p + msg->front.iov_len;
1109
1110 h = p;
1111 h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1112 h->seq = cpu_to_le64(seq);
1113
1114 /*
1115 * Serialize client metadata into waiting buffer space, using
1116 * the format that userspace expects for map<string, string>
1117 *
1118 * ClientSession messages with metadata are v2
1119 */
1120 msg->hdr.version = cpu_to_le16(3);
1121 msg->hdr.compat_version = cpu_to_le16(1);
1122
1123 /* The write pointer, following the session_head structure */
1124 p += sizeof(*h);
1125
1126 /* Number of entries in the map */
1127 ceph_encode_32(&p, metadata_key_count);
1128
1129 /* Two length-prefixed strings for each entry in the map */
1130 for (i = 0; metadata[i][0]; ++i) {
1131 size_t const key_len = strlen(metadata[i][0]);
1132 size_t const val_len = strlen(metadata[i][1]);
1133
1134 ceph_encode_32(&p, key_len);
1135 memcpy(p, metadata[i][0], key_len);
1136 p += key_len;
1137 ceph_encode_32(&p, val_len);
1138 memcpy(p, metadata[i][1], val_len);
1139 p += val_len;
1140 }
1141
1142 encode_supported_features(&p, end);
1143 msg->front.iov_len = p - msg->front.iov_base;
1144 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1145
1146 return msg;
1147 }
1148
1149 /*
1150 * send session open request.
1151 *
1152 * called under mdsc->mutex
1153 */
1154 static int __open_session(struct ceph_mds_client *mdsc,
1155 struct ceph_mds_session *session)
1156 {
1157 struct ceph_msg *msg;
1158 int mstate;
1159 int mds = session->s_mds;
1160
1161 /* wait for mds to go active? */
1162 mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1163 dout("open_session to mds%d (%s)\n", mds,
1164 ceph_mds_state_name(mstate));
1165 session->s_state = CEPH_MDS_SESSION_OPENING;
1166 session->s_renew_requested = jiffies;
1167
1168 /* send connect message */
1169 msg = create_session_open_msg(mdsc, session->s_seq);
1170 if (!msg)
1171 return -ENOMEM;
1172 ceph_con_send(&session->s_con, msg);
1173 return 0;
1174 }
1175
1176 /*
1177 * open sessions for any export targets for the given mds
1178 *
1179 * called under mdsc->mutex
1180 */
1181 static struct ceph_mds_session *
1182 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1183 {
1184 struct ceph_mds_session *session;
1185
1186 session = __ceph_lookup_mds_session(mdsc, target);
1187 if (!session) {
1188 session = register_session(mdsc, target);
1189 if (IS_ERR(session))
1190 return session;
1191 }
1192 if (session->s_state == CEPH_MDS_SESSION_NEW ||
1193 session->s_state == CEPH_MDS_SESSION_CLOSING)
1194 __open_session(mdsc, session);
1195
1196 return session;
1197 }
1198
1199 struct ceph_mds_session *
1200 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1201 {
1202 struct ceph_mds_session *session;
1203
1204 dout("open_export_target_session to mds%d\n", target);
1205
1206 mutex_lock(&mdsc->mutex);
1207 session = __open_export_target_session(mdsc, target);
1208 mutex_unlock(&mdsc->mutex);
1209
1210 return session;
1211 }
1212
1213 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1214 struct ceph_mds_session *session)
1215 {
1216 struct ceph_mds_info *mi;
1217 struct ceph_mds_session *ts;
1218 int i, mds = session->s_mds;
1219
1220 if (mds >= mdsc->mdsmap->m_num_mds)
1221 return;
1222
1223 mi = &mdsc->mdsmap->m_info[mds];
1224 dout("open_export_target_sessions for mds%d (%d targets)\n",
1225 session->s_mds, mi->num_export_targets);
1226
1227 for (i = 0; i < mi->num_export_targets; i++) {
1228 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1229 if (!IS_ERR(ts))
1230 ceph_put_mds_session(ts);
1231 }
1232 }
1233
1234 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1235 struct ceph_mds_session *session)
1236 {
1237 mutex_lock(&mdsc->mutex);
1238 __open_export_target_sessions(mdsc, session);
1239 mutex_unlock(&mdsc->mutex);
1240 }
1241
1242 /*
1243 * session caps
1244 */
1245
1246 static void detach_cap_releases(struct ceph_mds_session *session,
1247 struct list_head *target)
1248 {
1249 lockdep_assert_held(&session->s_cap_lock);
1250
1251 list_splice_init(&session->s_cap_releases, target);
1252 session->s_num_cap_releases = 0;
1253 dout("dispose_cap_releases mds%d\n", session->s_mds);
1254 }
1255
1256 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1257 struct list_head *dispose)
1258 {
1259 while (!list_empty(dispose)) {
1260 struct ceph_cap *cap;
1261 /* zero out the in-progress message */
1262 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1263 list_del(&cap->session_caps);
1264 ceph_put_cap(mdsc, cap);
1265 }
1266 }
1267
1268 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1269 struct ceph_mds_session *session)
1270 {
1271 struct ceph_mds_request *req;
1272 struct rb_node *p;
1273
1274 dout("cleanup_session_requests mds%d\n", session->s_mds);
1275 mutex_lock(&mdsc->mutex);
1276 while (!list_empty(&session->s_unsafe)) {
1277 req = list_first_entry(&session->s_unsafe,
1278 struct ceph_mds_request, r_unsafe_item);
1279 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1280 req->r_tid);
1281 __unregister_request(mdsc, req);
1282 }
1283 /* zero r_attempts, so kick_requests() will re-send requests */
1284 p = rb_first(&mdsc->request_tree);
1285 while (p) {
1286 req = rb_entry(p, struct ceph_mds_request, r_node);
1287 p = rb_next(p);
1288 if (req->r_session &&
1289 req->r_session->s_mds == session->s_mds)
1290 req->r_attempts = 0;
1291 }
1292 mutex_unlock(&mdsc->mutex);
1293 }
1294
1295 /*
1296 * Helper to safely iterate over all caps associated with a session, with
1297 * special care taken to handle a racing __ceph_remove_cap().
1298 *
1299 * Caller must hold session s_mutex.
1300 */
1301 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1302 int (*cb)(struct inode *, struct ceph_cap *,
1303 void *), void *arg)
1304 {
1305 struct list_head *p;
1306 struct ceph_cap *cap;
1307 struct inode *inode, *last_inode = NULL;
1308 struct ceph_cap *old_cap = NULL;
1309 int ret;
1310
1311 dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1312 spin_lock(&session->s_cap_lock);
1313 p = session->s_caps.next;
1314 while (p != &session->s_caps) {
1315 cap = list_entry(p, struct ceph_cap, session_caps);
1316 inode = igrab(&cap->ci->vfs_inode);
1317 if (!inode) {
1318 p = p->next;
1319 continue;
1320 }
1321 session->s_cap_iterator = cap;
1322 spin_unlock(&session->s_cap_lock);
1323
1324 if (last_inode) {
1325 /* avoid calling iput_final() while holding
1326 * s_mutex or in mds dispatch threads */
1327 ceph_async_iput(last_inode);
1328 last_inode = NULL;
1329 }
1330 if (old_cap) {
1331 ceph_put_cap(session->s_mdsc, old_cap);
1332 old_cap = NULL;
1333 }
1334
1335 ret = cb(inode, cap, arg);
1336 last_inode = inode;
1337
1338 spin_lock(&session->s_cap_lock);
1339 p = p->next;
1340 if (!cap->ci) {
1341 dout("iterate_session_caps finishing cap %p removal\n",
1342 cap);
1343 BUG_ON(cap->session != session);
1344 cap->session = NULL;
1345 list_del_init(&cap->session_caps);
1346 session->s_nr_caps--;
1347 if (cap->queue_release)
1348 __ceph_queue_cap_release(session, cap);
1349 else
1350 old_cap = cap; /* put_cap it w/o locks held */
1351 }
1352 if (ret < 0)
1353 goto out;
1354 }
1355 ret = 0;
1356 out:
1357 session->s_cap_iterator = NULL;
1358 spin_unlock(&session->s_cap_lock);
1359
1360 ceph_async_iput(last_inode);
1361 if (old_cap)
1362 ceph_put_cap(session->s_mdsc, old_cap);
1363
1364 return ret;
1365 }
1366
1367 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1368 void *arg)
1369 {
1370 struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
1371 struct ceph_inode_info *ci = ceph_inode(inode);
1372 LIST_HEAD(to_remove);
1373 bool drop = false;
1374 bool invalidate = false;
1375
1376 dout("removing cap %p, ci is %p, inode is %p\n",
1377 cap, ci, &ci->vfs_inode);
1378 spin_lock(&ci->i_ceph_lock);
1379 if (cap->mds_wanted | cap->issued)
1380 ci->i_ceph_flags |= CEPH_I_CAP_DROPPED;
1381 __ceph_remove_cap(cap, false);
1382 if (!ci->i_auth_cap) {
1383 struct ceph_cap_flush *cf;
1384 struct ceph_mds_client *mdsc = fsc->mdsc;
1385
1386 if (ci->i_wrbuffer_ref > 0 &&
1387 READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
1388 invalidate = true;
1389
1390 while (!list_empty(&ci->i_cap_flush_list)) {
1391 cf = list_first_entry(&ci->i_cap_flush_list,
1392 struct ceph_cap_flush, i_list);
1393 list_move(&cf->i_list, &to_remove);
1394 }
1395
1396 spin_lock(&mdsc->cap_dirty_lock);
1397
1398 list_for_each_entry(cf, &to_remove, i_list)
1399 list_del(&cf->g_list);
1400
1401 if (!list_empty(&ci->i_dirty_item)) {
1402 pr_warn_ratelimited(
1403 " dropping dirty %s state for %p %lld\n",
1404 ceph_cap_string(ci->i_dirty_caps),
1405 inode, ceph_ino(inode));
1406 ci->i_dirty_caps = 0;
1407 list_del_init(&ci->i_dirty_item);
1408 drop = true;
1409 }
1410 if (!list_empty(&ci->i_flushing_item)) {
1411 pr_warn_ratelimited(
1412 " dropping dirty+flushing %s state for %p %lld\n",
1413 ceph_cap_string(ci->i_flushing_caps),
1414 inode, ceph_ino(inode));
1415 ci->i_flushing_caps = 0;
1416 list_del_init(&ci->i_flushing_item);
1417 mdsc->num_cap_flushing--;
1418 drop = true;
1419 }
1420 spin_unlock(&mdsc->cap_dirty_lock);
1421
1422 if (atomic_read(&ci->i_filelock_ref) > 0) {
1423 /* make further file lock syscall return -EIO */
1424 ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
1425 pr_warn_ratelimited(" dropping file locks for %p %lld\n",
1426 inode, ceph_ino(inode));
1427 }
1428
1429 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
1430 list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);
1431 ci->i_prealloc_cap_flush = NULL;
1432 }
1433
1434 if (drop &&
1435 ci->i_wrbuffer_ref_head == 0 &&
1436 ci->i_wr_ref == 0 &&
1437 ci->i_dirty_caps == 0 &&
1438 ci->i_flushing_caps == 0) {
1439 ceph_put_snap_context(ci->i_head_snapc);
1440 ci->i_head_snapc = NULL;
1441 }
1442 }
1443 spin_unlock(&ci->i_ceph_lock);
1444 while (!list_empty(&to_remove)) {
1445 struct ceph_cap_flush *cf;
1446 cf = list_first_entry(&to_remove,
1447 struct ceph_cap_flush, i_list);
1448 list_del(&cf->i_list);
1449 ceph_free_cap_flush(cf);
1450 }
1451
1452 wake_up_all(&ci->i_cap_wq);
1453 if (invalidate)
1454 ceph_queue_invalidate(inode);
1455 if (drop)
1456 iput(inode);
1457 return 0;
1458 }
1459
1460 /*
1461 * caller must hold session s_mutex
1462 */
1463 static void remove_session_caps(struct ceph_mds_session *session)
1464 {
1465 struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1466 struct super_block *sb = fsc->sb;
1467 LIST_HEAD(dispose);
1468
1469 dout("remove_session_caps on %p\n", session);
1470 ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1471
1472 wake_up_all(&fsc->mdsc->cap_flushing_wq);
1473
1474 spin_lock(&session->s_cap_lock);
1475 if (session->s_nr_caps > 0) {
1476 struct inode *inode;
1477 struct ceph_cap *cap, *prev = NULL;
1478 struct ceph_vino vino;
1479 /*
1480 * iterate_session_caps() skips inodes that are being
1481 * deleted, we need to wait until deletions are complete.
1482 * __wait_on_freeing_inode() is designed for the job,
1483 * but it is not exported, so use lookup inode function
1484 * to access it.
1485 */
1486 while (!list_empty(&session->s_caps)) {
1487 cap = list_entry(session->s_caps.next,
1488 struct ceph_cap, session_caps);
1489 if (cap == prev)
1490 break;
1491 prev = cap;
1492 vino = cap->ci->i_vino;
1493 spin_unlock(&session->s_cap_lock);
1494
1495 inode = ceph_find_inode(sb, vino);
1496 /* avoid calling iput_final() while holding s_mutex */
1497 ceph_async_iput(inode);
1498
1499 spin_lock(&session->s_cap_lock);
1500 }
1501 }
1502
1503 // drop cap expires and unlock s_cap_lock
1504 detach_cap_releases(session, &dispose);
1505
1506 BUG_ON(session->s_nr_caps > 0);
1507 BUG_ON(!list_empty(&session->s_cap_flushing));
1508 spin_unlock(&session->s_cap_lock);
1509 dispose_cap_releases(session->s_mdsc, &dispose);
1510 }
1511
1512 enum {
1513 RECONNECT,
1514 RENEWCAPS,
1515 FORCE_RO,
1516 };
1517
1518 /*
1519 * wake up any threads waiting on this session's caps. if the cap is
1520 * old (didn't get renewed on the client reconnect), remove it now.
1521 *
1522 * caller must hold s_mutex.
1523 */
1524 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1525 void *arg)
1526 {
1527 struct ceph_inode_info *ci = ceph_inode(inode);
1528 unsigned long ev = (unsigned long)arg;
1529
1530 if (ev == RECONNECT) {
1531 spin_lock(&ci->i_ceph_lock);
1532 ci->i_wanted_max_size = 0;
1533 ci->i_requested_max_size = 0;
1534 spin_unlock(&ci->i_ceph_lock);
1535 } else if (ev == RENEWCAPS) {
1536 if (cap->cap_gen < cap->session->s_cap_gen) {
1537 /* mds did not re-issue stale cap */
1538 spin_lock(&ci->i_ceph_lock);
1539 cap->issued = cap->implemented = CEPH_CAP_PIN;
1540 /* make sure mds knows what we want */
1541 if (__ceph_caps_file_wanted(ci) & ~cap->mds_wanted)
1542 ci->i_ceph_flags |= CEPH_I_CAP_DROPPED;
1543 spin_unlock(&ci->i_ceph_lock);
1544 }
1545 } else if (ev == FORCE_RO) {
1546 }
1547 wake_up_all(&ci->i_cap_wq);
1548 return 0;
1549 }
1550
1551 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1552 {
1553 dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1554 ceph_iterate_session_caps(session, wake_up_session_cb,
1555 (void *)(unsigned long)ev);
1556 }
1557
1558 /*
1559 * Send periodic message to MDS renewing all currently held caps. The
1560 * ack will reset the expiration for all caps from this session.
1561 *
1562 * caller holds s_mutex
1563 */
1564 static int send_renew_caps(struct ceph_mds_client *mdsc,
1565 struct ceph_mds_session *session)
1566 {
1567 struct ceph_msg *msg;
1568 int state;
1569
1570 if (time_after_eq(jiffies, session->s_cap_ttl) &&
1571 time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1572 pr_info("mds%d caps stale\n", session->s_mds);
1573 session->s_renew_requested = jiffies;
1574
1575 /* do not try to renew caps until a recovering mds has reconnected
1576 * with its clients. */
1577 state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1578 if (state < CEPH_MDS_STATE_RECONNECT) {
1579 dout("send_renew_caps ignoring mds%d (%s)\n",
1580 session->s_mds, ceph_mds_state_name(state));
1581 return 0;
1582 }
1583
1584 dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1585 ceph_mds_state_name(state));
1586 msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1587 ++session->s_renew_seq);
1588 if (!msg)
1589 return -ENOMEM;
1590 ceph_con_send(&session->s_con, msg);
1591 return 0;
1592 }
1593
1594 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1595 struct ceph_mds_session *session, u64 seq)
1596 {
1597 struct ceph_msg *msg;
1598
1599 dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1600 session->s_mds, ceph_session_state_name(session->s_state), seq);
1601 msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1602 if (!msg)
1603 return -ENOMEM;
1604 ceph_con_send(&session->s_con, msg);
1605 return 0;
1606 }
1607
1608
1609 /*
1610 * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1611 *
1612 * Called under session->s_mutex
1613 */
1614 static void renewed_caps(struct ceph_mds_client *mdsc,
1615 struct ceph_mds_session *session, int is_renew)
1616 {
1617 int was_stale;
1618 int wake = 0;
1619
1620 spin_lock(&session->s_cap_lock);
1621 was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1622
1623 session->s_cap_ttl = session->s_renew_requested +
1624 mdsc->mdsmap->m_session_timeout*HZ;
1625
1626 if (was_stale) {
1627 if (time_before(jiffies, session->s_cap_ttl)) {
1628 pr_info("mds%d caps renewed\n", session->s_mds);
1629 wake = 1;
1630 } else {
1631 pr_info("mds%d caps still stale\n", session->s_mds);
1632 }
1633 }
1634 dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1635 session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1636 time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1637 spin_unlock(&session->s_cap_lock);
1638
1639 if (wake)
1640 wake_up_session_caps(session, RENEWCAPS);
1641 }
1642
1643 /*
1644 * send a session close request
1645 */
1646 static int request_close_session(struct ceph_mds_client *mdsc,
1647 struct ceph_mds_session *session)
1648 {
1649 struct ceph_msg *msg;
1650
1651 dout("request_close_session mds%d state %s seq %lld\n",
1652 session->s_mds, ceph_session_state_name(session->s_state),
1653 session->s_seq);
1654 msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
1655 if (!msg)
1656 return -ENOMEM;
1657 ceph_con_send(&session->s_con, msg);
1658 return 1;
1659 }
1660
1661 /*
1662 * Called with s_mutex held.
1663 */
1664 static int __close_session(struct ceph_mds_client *mdsc,
1665 struct ceph_mds_session *session)
1666 {
1667 if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1668 return 0;
1669 session->s_state = CEPH_MDS_SESSION_CLOSING;
1670 return request_close_session(mdsc, session);
1671 }
1672
1673 static bool drop_negative_children(struct dentry *dentry)
1674 {
1675 struct dentry *child;
1676 bool all_negative = true;
1677
1678 if (!d_is_dir(dentry))
1679 goto out;
1680
1681 spin_lock(&dentry->d_lock);
1682 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1683 if (d_really_is_positive(child)) {
1684 all_negative = false;
1685 break;
1686 }
1687 }
1688 spin_unlock(&dentry->d_lock);
1689
1690 if (all_negative)
1691 shrink_dcache_parent(dentry);
1692 out:
1693 return all_negative;
1694 }
1695
1696 /*
1697 * Trim old(er) caps.
1698 *
1699 * Because we can't cache an inode without one or more caps, we do
1700 * this indirectly: if a cap is unused, we prune its aliases, at which
1701 * point the inode will hopefully get dropped to.
1702 *
1703 * Yes, this is a bit sloppy. Our only real goal here is to respond to
1704 * memory pressure from the MDS, though, so it needn't be perfect.
1705 */
1706 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1707 {
1708 struct ceph_mds_session *session = arg;
1709 struct ceph_inode_info *ci = ceph_inode(inode);
1710 int used, wanted, oissued, mine;
1711
1712 if (session->s_trim_caps <= 0)
1713 return -1;
1714
1715 spin_lock(&ci->i_ceph_lock);
1716 mine = cap->issued | cap->implemented;
1717 used = __ceph_caps_used(ci);
1718 wanted = __ceph_caps_file_wanted(ci);
1719 oissued = __ceph_caps_issued_other(ci, cap);
1720
1721 dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1722 inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1723 ceph_cap_string(used), ceph_cap_string(wanted));
1724 if (cap == ci->i_auth_cap) {
1725 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1726 !list_empty(&ci->i_cap_snaps))
1727 goto out;
1728 if ((used | wanted) & CEPH_CAP_ANY_WR)
1729 goto out;
1730 /* Note: it's possible that i_filelock_ref becomes non-zero
1731 * after dropping auth caps. It doesn't hurt because reply
1732 * of lock mds request will re-add auth caps. */
1733 if (atomic_read(&ci->i_filelock_ref) > 0)
1734 goto out;
1735 }
1736 /* The inode has cached pages, but it's no longer used.
1737 * we can safely drop it */
1738 if (wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
1739 !(oissued & CEPH_CAP_FILE_CACHE)) {
1740 used = 0;
1741 oissued = 0;
1742 }
1743 if ((used | wanted) & ~oissued & mine)
1744 goto out; /* we need these caps */
1745
1746 if (oissued) {
1747 /* we aren't the only cap.. just remove us */
1748 __ceph_remove_cap(cap, true);
1749 session->s_trim_caps--;
1750 } else {
1751 struct dentry *dentry;
1752 /* try dropping referring dentries */
1753 spin_unlock(&ci->i_ceph_lock);
1754 dentry = d_find_any_alias(inode);
1755 if (dentry && drop_negative_children(dentry)) {
1756 int count;
1757 dput(dentry);
1758 d_prune_aliases(inode);
1759 count = atomic_read(&inode->i_count);
1760 if (count == 1)
1761 session->s_trim_caps--;
1762 dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1763 inode, cap, count);
1764 } else {
1765 dput(dentry);
1766 }
1767 return 0;
1768 }
1769
1770 out:
1771 spin_unlock(&ci->i_ceph_lock);
1772 return 0;
1773 }
1774
1775 /*
1776 * Trim session cap count down to some max number.
1777 */
1778 int ceph_trim_caps(struct ceph_mds_client *mdsc,
1779 struct ceph_mds_session *session,
1780 int max_caps)
1781 {
1782 int trim_caps = session->s_nr_caps - max_caps;
1783
1784 dout("trim_caps mds%d start: %d / %d, trim %d\n",
1785 session->s_mds, session->s_nr_caps, max_caps, trim_caps);
1786 if (trim_caps > 0) {
1787 session->s_trim_caps = trim_caps;
1788 ceph_iterate_session_caps(session, trim_caps_cb, session);
1789 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
1790 session->s_mds, session->s_nr_caps, max_caps,
1791 trim_caps - session->s_trim_caps);
1792 session->s_trim_caps = 0;
1793 }
1794
1795 ceph_flush_cap_releases(mdsc, session);
1796 return 0;
1797 }
1798
1799 static int check_caps_flush(struct ceph_mds_client *mdsc,
1800 u64 want_flush_tid)
1801 {
1802 int ret = 1;
1803
1804 spin_lock(&mdsc->cap_dirty_lock);
1805 if (!list_empty(&mdsc->cap_flush_list)) {
1806 struct ceph_cap_flush *cf =
1807 list_first_entry(&mdsc->cap_flush_list,
1808 struct ceph_cap_flush, g_list);
1809 if (cf->tid <= want_flush_tid) {
1810 dout("check_caps_flush still flushing tid "
1811 "%llu <= %llu\n", cf->tid, want_flush_tid);
1812 ret = 0;
1813 }
1814 }
1815 spin_unlock(&mdsc->cap_dirty_lock);
1816 return ret;
1817 }
1818
1819 /*
1820 * flush all dirty inode data to disk.
1821 *
1822 * returns true if we've flushed through want_flush_tid
1823 */
1824 static void wait_caps_flush(struct ceph_mds_client *mdsc,
1825 u64 want_flush_tid)
1826 {
1827 dout("check_caps_flush want %llu\n", want_flush_tid);
1828
1829 wait_event(mdsc->cap_flushing_wq,
1830 check_caps_flush(mdsc, want_flush_tid));
1831
1832 dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
1833 }
1834
1835 /*
1836 * called under s_mutex
1837 */
1838 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
1839 struct ceph_mds_session *session)
1840 {
1841 struct ceph_msg *msg = NULL;
1842 struct ceph_mds_cap_release *head;
1843 struct ceph_mds_cap_item *item;
1844 struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
1845 struct ceph_cap *cap;
1846 LIST_HEAD(tmp_list);
1847 int num_cap_releases;
1848 __le32 barrier, *cap_barrier;
1849
1850 down_read(&osdc->lock);
1851 barrier = cpu_to_le32(osdc->epoch_barrier);
1852 up_read(&osdc->lock);
1853
1854 spin_lock(&session->s_cap_lock);
1855 again:
1856 list_splice_init(&session->s_cap_releases, &tmp_list);
1857 num_cap_releases = session->s_num_cap_releases;
1858 session->s_num_cap_releases = 0;
1859 spin_unlock(&session->s_cap_lock);
1860
1861 while (!list_empty(&tmp_list)) {
1862 if (!msg) {
1863 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
1864 PAGE_SIZE, GFP_NOFS, false);
1865 if (!msg)
1866 goto out_err;
1867 head = msg->front.iov_base;
1868 head->num = cpu_to_le32(0);
1869 msg->front.iov_len = sizeof(*head);
1870
1871 msg->hdr.version = cpu_to_le16(2);
1872 msg->hdr.compat_version = cpu_to_le16(1);
1873 }
1874
1875 cap = list_first_entry(&tmp_list, struct ceph_cap,
1876 session_caps);
1877 list_del(&cap->session_caps);
1878 num_cap_releases--;
1879
1880 head = msg->front.iov_base;
1881 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
1882 &head->num);
1883 item = msg->front.iov_base + msg->front.iov_len;
1884 item->ino = cpu_to_le64(cap->cap_ino);
1885 item->cap_id = cpu_to_le64(cap->cap_id);
1886 item->migrate_seq = cpu_to_le32(cap->mseq);
1887 item->seq = cpu_to_le32(cap->issue_seq);
1888 msg->front.iov_len += sizeof(*item);
1889
1890 ceph_put_cap(mdsc, cap);
1891
1892 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
1893 // Append cap_barrier field
1894 cap_barrier = msg->front.iov_base + msg->front.iov_len;
1895 *cap_barrier = barrier;
1896 msg->front.iov_len += sizeof(*cap_barrier);
1897
1898 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1899 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
1900 ceph_con_send(&session->s_con, msg);
1901 msg = NULL;
1902 }
1903 }
1904
1905 BUG_ON(num_cap_releases != 0);
1906
1907 spin_lock(&session->s_cap_lock);
1908 if (!list_empty(&session->s_cap_releases))
1909 goto again;
1910 spin_unlock(&session->s_cap_lock);
1911
1912 if (msg) {
1913 // Append cap_barrier field
1914 cap_barrier = msg->front.iov_base + msg->front.iov_len;
1915 *cap_barrier = barrier;
1916 msg->front.iov_len += sizeof(*cap_barrier);
1917
1918 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1919 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
1920 ceph_con_send(&session->s_con, msg);
1921 }
1922 return;
1923 out_err:
1924 pr_err("send_cap_releases mds%d, failed to allocate message\n",
1925 session->s_mds);
1926 spin_lock(&session->s_cap_lock);
1927 list_splice(&tmp_list, &session->s_cap_releases);
1928 session->s_num_cap_releases += num_cap_releases;
1929 spin_unlock(&session->s_cap_lock);
1930 }
1931
1932 static void ceph_cap_release_work(struct work_struct *work)
1933 {
1934 struct ceph_mds_session *session =
1935 container_of(work, struct ceph_mds_session, s_cap_release_work);
1936
1937 mutex_lock(&session->s_mutex);
1938 if (session->s_state == CEPH_MDS_SESSION_OPEN ||
1939 session->s_state == CEPH_MDS_SESSION_HUNG)
1940 ceph_send_cap_releases(session->s_mdsc, session);
1941 mutex_unlock(&session->s_mutex);
1942 ceph_put_mds_session(session);
1943 }
1944
1945 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
1946 struct ceph_mds_session *session)
1947 {
1948 if (mdsc->stopping)
1949 return;
1950
1951 get_session(session);
1952 if (queue_work(mdsc->fsc->cap_wq,
1953 &session->s_cap_release_work)) {
1954 dout("cap release work queued\n");
1955 } else {
1956 ceph_put_mds_session(session);
1957 dout("failed to queue cap release work\n");
1958 }
1959 }
1960
1961 /*
1962 * caller holds session->s_cap_lock
1963 */
1964 void __ceph_queue_cap_release(struct ceph_mds_session *session,
1965 struct ceph_cap *cap)
1966 {
1967 list_add_tail(&cap->session_caps, &session->s_cap_releases);
1968 session->s_num_cap_releases++;
1969
1970 if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
1971 ceph_flush_cap_releases(session->s_mdsc, session);
1972 }
1973
1974 static void ceph_cap_reclaim_work(struct work_struct *work)
1975 {
1976 struct ceph_mds_client *mdsc =
1977 container_of(work, struct ceph_mds_client, cap_reclaim_work);
1978 int ret = ceph_trim_dentries(mdsc);
1979 if (ret == -EAGAIN)
1980 ceph_queue_cap_reclaim_work(mdsc);
1981 }
1982
1983 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
1984 {
1985 if (mdsc->stopping)
1986 return;
1987
1988 if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
1989 dout("caps reclaim work queued\n");
1990 } else {
1991 dout("failed to queue caps release work\n");
1992 }
1993 }
1994
1995 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
1996 {
1997 int val;
1998 if (!nr)
1999 return;
2000 val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2001 if (!(val % CEPH_CAPS_PER_RELEASE)) {
2002 atomic_set(&mdsc->cap_reclaim_pending, 0);
2003 ceph_queue_cap_reclaim_work(mdsc);
2004 }
2005 }
2006
2007 /*
2008 * requests
2009 */
2010
2011 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2012 struct inode *dir)
2013 {
2014 struct ceph_inode_info *ci = ceph_inode(dir);
2015 struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2016 struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2017 size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2018 int order, num_entries;
2019
2020 spin_lock(&ci->i_ceph_lock);
2021 num_entries = ci->i_files + ci->i_subdirs;
2022 spin_unlock(&ci->i_ceph_lock);
2023 num_entries = max(num_entries, 1);
2024 num_entries = min(num_entries, opt->max_readdir);
2025
2026 order = get_order(size * num_entries);
2027 while (order >= 0) {
2028 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2029 __GFP_NOWARN,
2030 order);
2031 if (rinfo->dir_entries)
2032 break;
2033 order--;
2034 }
2035 if (!rinfo->dir_entries)
2036 return -ENOMEM;
2037
2038 num_entries = (PAGE_SIZE << order) / size;
2039 num_entries = min(num_entries, opt->max_readdir);
2040
2041 rinfo->dir_buf_size = PAGE_SIZE << order;
2042 req->r_num_caps = num_entries + 1;
2043 req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2044 req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2045 return 0;
2046 }
2047
2048 /*
2049 * Create an mds request.
2050 */
2051 struct ceph_mds_request *
2052 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2053 {
2054 struct ceph_mds_request *req = kzalloc(sizeof(*req), GFP_NOFS);
2055 struct timespec64 ts;
2056
2057 if (!req)
2058 return ERR_PTR(-ENOMEM);
2059
2060 mutex_init(&req->r_fill_mutex);
2061 req->r_mdsc = mdsc;
2062 req->r_started = jiffies;
2063 req->r_resend_mds = -1;
2064 INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2065 INIT_LIST_HEAD(&req->r_unsafe_target_item);
2066 req->r_fmode = -1;
2067 kref_init(&req->r_kref);
2068 RB_CLEAR_NODE(&req->r_node);
2069 INIT_LIST_HEAD(&req->r_wait);
2070 init_completion(&req->r_completion);
2071 init_completion(&req->r_safe_completion);
2072 INIT_LIST_HEAD(&req->r_unsafe_item);
2073
2074 ktime_get_coarse_real_ts64(&ts);
2075 req->r_stamp = timespec64_trunc(ts, mdsc->fsc->sb->s_time_gran);
2076
2077 req->r_op = op;
2078 req->r_direct_mode = mode;
2079 return req;
2080 }
2081
2082 /*
2083 * return oldest (lowest) request, tid in request tree, 0 if none.
2084 *
2085 * called under mdsc->mutex.
2086 */
2087 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2088 {
2089 if (RB_EMPTY_ROOT(&mdsc->request_tree))
2090 return NULL;
2091 return rb_entry(rb_first(&mdsc->request_tree),
2092 struct ceph_mds_request, r_node);
2093 }
2094
2095 static inline u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2096 {
2097 return mdsc->oldest_tid;
2098 }
2099
2100 /*
2101 * Build a dentry's path. Allocate on heap; caller must kfree. Based
2102 * on build_path_from_dentry in fs/cifs/dir.c.
2103 *
2104 * If @stop_on_nosnap, generate path relative to the first non-snapped
2105 * inode.
2106 *
2107 * Encode hidden .snap dirs as a double /, i.e.
2108 * foo/.snap/bar -> foo//bar
2109 */
2110 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2111 int stop_on_nosnap)
2112 {
2113 struct dentry *temp;
2114 char *path;
2115 int pos;
2116 unsigned seq;
2117 u64 base;
2118
2119 if (!dentry)
2120 return ERR_PTR(-EINVAL);
2121
2122 path = __getname();
2123 if (!path)
2124 return ERR_PTR(-ENOMEM);
2125 retry:
2126 pos = PATH_MAX - 1;
2127 path[pos] = '\0';
2128
2129 seq = read_seqbegin(&rename_lock);
2130 rcu_read_lock();
2131 temp = dentry;
2132 for (;;) {
2133 struct inode *inode;
2134
2135 spin_lock(&temp->d_lock);
2136 inode = d_inode(temp);
2137 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2138 dout("build_path path+%d: %p SNAPDIR\n",
2139 pos, temp);
2140 } else if (stop_on_nosnap && inode && dentry != temp &&
2141 ceph_snap(inode) == CEPH_NOSNAP) {
2142 spin_unlock(&temp->d_lock);
2143 pos++; /* get rid of any prepended '/' */
2144 break;
2145 } else {
2146 pos -= temp->d_name.len;
2147 if (pos < 0) {
2148 spin_unlock(&temp->d_lock);
2149 break;
2150 }
2151 memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2152 }
2153 spin_unlock(&temp->d_lock);
2154 temp = READ_ONCE(temp->d_parent);
2155
2156 /* Are we at the root? */
2157 if (IS_ROOT(temp))
2158 break;
2159
2160 /* Are we out of buffer? */
2161 if (--pos < 0)
2162 break;
2163
2164 path[pos] = '/';
2165 }
2166 base = ceph_ino(d_inode(temp));
2167 rcu_read_unlock();
2168 if (pos < 0 || read_seqretry(&rename_lock, seq)) {
2169 pr_err("build_path did not end path lookup where "
2170 "expected, pos is %d\n", pos);
2171 /* presumably this is only possible if racing with a
2172 rename of one of the parent directories (we can not
2173 lock the dentries above us to prevent this, but
2174 retrying should be harmless) */
2175 goto retry;
2176 }
2177
2178 *pbase = base;
2179 *plen = PATH_MAX - 1 - pos;
2180 dout("build_path on %p %d built %llx '%.*s'\n",
2181 dentry, d_count(dentry), base, *plen, path + pos);
2182 return path + pos;
2183 }
2184
2185 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2186 const char **ppath, int *ppathlen, u64 *pino,
2187 bool *pfreepath, bool parent_locked)
2188 {
2189 char *path;
2190
2191 rcu_read_lock();
2192 if (!dir)
2193 dir = d_inode_rcu(dentry->d_parent);
2194 if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
2195 *pino = ceph_ino(dir);
2196 rcu_read_unlock();
2197 *ppath = dentry->d_name.name;
2198 *ppathlen = dentry->d_name.len;
2199 return 0;
2200 }
2201 rcu_read_unlock();
2202 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2203 if (IS_ERR(path))
2204 return PTR_ERR(path);
2205 *ppath = path;
2206 *pfreepath = true;
2207 return 0;
2208 }
2209
2210 static int build_inode_path(struct inode *inode,
2211 const char **ppath, int *ppathlen, u64 *pino,
2212 bool *pfreepath)
2213 {
2214 struct dentry *dentry;
2215 char *path;
2216
2217 if (ceph_snap(inode) == CEPH_NOSNAP) {
2218 *pino = ceph_ino(inode);
2219 *ppathlen = 0;
2220 return 0;
2221 }
2222 dentry = d_find_alias(inode);
2223 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2224 dput(dentry);
2225 if (IS_ERR(path))
2226 return PTR_ERR(path);
2227 *ppath = path;
2228 *pfreepath = true;
2229 return 0;
2230 }
2231
2232 /*
2233 * request arguments may be specified via an inode *, a dentry *, or
2234 * an explicit ino+path.
2235 */
2236 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
2237 struct inode *rdiri, const char *rpath,
2238 u64 rino, const char **ppath, int *pathlen,
2239 u64 *ino, bool *freepath, bool parent_locked)
2240 {
2241 int r = 0;
2242
2243 if (rinode) {
2244 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2245 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2246 ceph_snap(rinode));
2247 } else if (rdentry) {
2248 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
2249 freepath, parent_locked);
2250 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2251 *ppath);
2252 } else if (rpath || rino) {
2253 *ino = rino;
2254 *ppath = rpath;
2255 *pathlen = rpath ? strlen(rpath) : 0;
2256 dout(" path %.*s\n", *pathlen, rpath);
2257 }
2258
2259 return r;
2260 }
2261
2262 /*
2263 * called under mdsc->mutex
2264 */
2265 static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
2266 struct ceph_mds_request *req,
2267 int mds, bool drop_cap_releases)
2268 {
2269 struct ceph_msg *msg;
2270 struct ceph_mds_request_head *head;
2271 const char *path1 = NULL;
2272 const char *path2 = NULL;
2273 u64 ino1 = 0, ino2 = 0;
2274 int pathlen1 = 0, pathlen2 = 0;
2275 bool freepath1 = false, freepath2 = false;
2276 int len;
2277 u16 releases;
2278 void *p, *end;
2279 int ret;
2280
2281 ret = set_request_path_attr(req->r_inode, req->r_dentry,
2282 req->r_parent, req->r_path1, req->r_ino1.ino,
2283 &path1, &pathlen1, &ino1, &freepath1,
2284 test_bit(CEPH_MDS_R_PARENT_LOCKED,
2285 &req->r_req_flags));
2286 if (ret < 0) {
2287 msg = ERR_PTR(ret);
2288 goto out;
2289 }
2290
2291 /* If r_old_dentry is set, then assume that its parent is locked */
2292 ret = set_request_path_attr(NULL, req->r_old_dentry,
2293 req->r_old_dentry_dir,
2294 req->r_path2, req->r_ino2.ino,
2295 &path2, &pathlen2, &ino2, &freepath2, true);
2296 if (ret < 0) {
2297 msg = ERR_PTR(ret);
2298 goto out_free1;
2299 }
2300
2301 len = sizeof(*head) +
2302 pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
2303 sizeof(struct ceph_timespec);
2304
2305 /* calculate (max) length for cap releases */
2306 len += sizeof(struct ceph_mds_request_release) *
2307 (!!req->r_inode_drop + !!req->r_dentry_drop +
2308 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2309 if (req->r_dentry_drop)
2310 len += pathlen1;
2311 if (req->r_old_dentry_drop)
2312 len += pathlen2;
2313
2314 msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
2315 if (!msg) {
2316 msg = ERR_PTR(-ENOMEM);
2317 goto out_free2;
2318 }
2319
2320 msg->hdr.version = cpu_to_le16(2);
2321 msg->hdr.tid = cpu_to_le64(req->r_tid);
2322
2323 head = msg->front.iov_base;
2324 p = msg->front.iov_base + sizeof(*head);
2325 end = msg->front.iov_base + msg->front.iov_len;
2326
2327 head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2328 head->op = cpu_to_le32(req->r_op);
2329 head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));
2330 head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));
2331 head->args = req->r_args;
2332
2333 ceph_encode_filepath(&p, end, ino1, path1);
2334 ceph_encode_filepath(&p, end, ino2, path2);
2335
2336 /* make note of release offset, in case we need to replay */
2337 req->r_request_release_offset = p - msg->front.iov_base;
2338
2339 /* cap releases */
2340 releases = 0;
2341 if (req->r_inode_drop)
2342 releases += ceph_encode_inode_release(&p,
2343 req->r_inode ? req->r_inode : d_inode(req->r_dentry),
2344 mds, req->r_inode_drop, req->r_inode_unless, 0);
2345 if (req->r_dentry_drop)
2346 releases += ceph_encode_dentry_release(&p, req->r_dentry,
2347 req->r_parent, mds, req->r_dentry_drop,
2348 req->r_dentry_unless);
2349 if (req->r_old_dentry_drop)
2350 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
2351 req->r_old_dentry_dir, mds,
2352 req->r_old_dentry_drop,
2353 req->r_old_dentry_unless);
2354 if (req->r_old_inode_drop)
2355 releases += ceph_encode_inode_release(&p,
2356 d_inode(req->r_old_dentry),
2357 mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
2358
2359 if (drop_cap_releases) {
2360 releases = 0;
2361 p = msg->front.iov_base + req->r_request_release_offset;
2362 }
2363
2364 head->num_releases = cpu_to_le16(releases);
2365
2366 /* time stamp */
2367 {
2368 struct ceph_timespec ts;
2369 ceph_encode_timespec64(&ts, &req->r_stamp);
2370 ceph_encode_copy(&p, &ts, sizeof(ts));
2371 }
2372
2373 BUG_ON(p > end);
2374 msg->front.iov_len = p - msg->front.iov_base;
2375 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2376
2377 if (req->r_pagelist) {
2378 struct ceph_pagelist *pagelist = req->r_pagelist;
2379 ceph_msg_data_add_pagelist(msg, pagelist);
2380 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2381 } else {
2382 msg->hdr.data_len = 0;
2383 }
2384
2385 msg->hdr.data_off = cpu_to_le16(0);
2386
2387 out_free2:
2388 if (freepath2)
2389 ceph_mdsc_free_path((char *)path2, pathlen2);
2390 out_free1:
2391 if (freepath1)
2392 ceph_mdsc_free_path((char *)path1, pathlen1);
2393 out:
2394 return msg;
2395 }
2396
2397 /*
2398 * called under mdsc->mutex if error, under no mutex if
2399 * success.
2400 */
2401 static void complete_request(struct ceph_mds_client *mdsc,
2402 struct ceph_mds_request *req)
2403 {
2404 if (req->r_callback)
2405 req->r_callback(mdsc, req);
2406 complete_all(&req->r_completion);
2407 }
2408
2409 /*
2410 * called under mdsc->mutex
2411 */
2412 static int __prepare_send_request(struct ceph_mds_client *mdsc,
2413 struct ceph_mds_request *req,
2414 int mds, bool drop_cap_releases)
2415 {
2416 struct ceph_mds_request_head *rhead;
2417 struct ceph_msg *msg;
2418 int flags = 0;
2419
2420 req->r_attempts++;
2421 if (req->r_inode) {
2422 struct ceph_cap *cap =
2423 ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2424
2425 if (cap)
2426 req->r_sent_on_mseq = cap->mseq;
2427 else
2428 req->r_sent_on_mseq = -1;
2429 }
2430 dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2431 req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2432
2433 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2434 void *p;
2435 /*
2436 * Replay. Do not regenerate message (and rebuild
2437 * paths, etc.); just use the original message.
2438 * Rebuilding paths will break for renames because
2439 * d_move mangles the src name.
2440 */
2441 msg = req->r_request;
2442 rhead = msg->front.iov_base;
2443
2444 flags = le32_to_cpu(rhead->flags);
2445 flags |= CEPH_MDS_FLAG_REPLAY;
2446 rhead->flags = cpu_to_le32(flags);
2447
2448 if (req->r_target_inode)
2449 rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2450
2451 rhead->num_retry = req->r_attempts - 1;
2452
2453 /* remove cap/dentry releases from message */
2454 rhead->num_releases = 0;
2455
2456 /* time stamp */
2457 p = msg->front.iov_base + req->r_request_release_offset;
2458 {
2459 struct ceph_timespec ts;
2460 ceph_encode_timespec64(&ts, &req->r_stamp);
2461 ceph_encode_copy(&p, &ts, sizeof(ts));
2462 }
2463
2464 msg->front.iov_len = p - msg->front.iov_base;
2465 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2466 return 0;
2467 }
2468
2469 if (req->r_request) {
2470 ceph_msg_put(req->r_request);
2471 req->r_request = NULL;
2472 }
2473 msg = create_request_message(mdsc, req, mds, drop_cap_releases);
2474 if (IS_ERR(msg)) {
2475 req->r_err = PTR_ERR(msg);
2476 return PTR_ERR(msg);
2477 }
2478 req->r_request = msg;
2479
2480 rhead = msg->front.iov_base;
2481 rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2482 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2483 flags |= CEPH_MDS_FLAG_REPLAY;
2484 if (req->r_parent)
2485 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2486 rhead->flags = cpu_to_le32(flags);
2487 rhead->num_fwd = req->r_num_fwd;
2488 rhead->num_retry = req->r_attempts - 1;
2489 rhead->ino = 0;
2490
2491 dout(" r_parent = %p\n", req->r_parent);
2492 return 0;
2493 }
2494
2495 /*
2496 * send request, or put it on the appropriate wait list.
2497 */
2498 static void __do_request(struct ceph_mds_client *mdsc,
2499 struct ceph_mds_request *req)
2500 {
2501 struct ceph_mds_session *session = NULL;
2502 int mds = -1;
2503 int err = 0;
2504
2505 if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2506 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
2507 __unregister_request(mdsc, req);
2508 return;
2509 }
2510
2511 if (req->r_timeout &&
2512 time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2513 dout("do_request timed out\n");
2514 err = -EIO;
2515 goto finish;
2516 }
2517 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2518 dout("do_request forced umount\n");
2519 err = -EIO;
2520 goto finish;
2521 }
2522 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
2523 if (mdsc->mdsmap_err) {
2524 err = mdsc->mdsmap_err;
2525 dout("do_request mdsmap err %d\n", err);
2526 goto finish;
2527 }
2528 if (mdsc->mdsmap->m_epoch == 0) {
2529 dout("do_request no mdsmap, waiting for map\n");
2530 list_add(&req->r_wait, &mdsc->waiting_for_map);
2531 return;
2532 }
2533 if (!(mdsc->fsc->mount_options->flags &
2534 CEPH_MOUNT_OPT_MOUNTWAIT) &&
2535 !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
2536 err = -ENOENT;
2537 pr_info("probably no mds server is up\n");
2538 goto finish;
2539 }
2540 }
2541
2542 put_request_session(req);
2543
2544 mds = __choose_mds(mdsc, req);
2545 if (mds < 0 ||
2546 ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2547 dout("do_request no mds or not active, waiting for map\n");
2548 list_add(&req->r_wait, &mdsc->waiting_for_map);
2549 return;
2550 }
2551
2552 /* get, open session */
2553 session = __ceph_lookup_mds_session(mdsc, mds);
2554 if (!session) {
2555 session = register_session(mdsc, mds);
2556 if (IS_ERR(session)) {
2557 err = PTR_ERR(session);
2558 goto finish;
2559 }
2560 }
2561 req->r_session = get_session(session);
2562
2563 dout("do_request mds%d session %p state %s\n", mds, session,
2564 ceph_session_state_name(session->s_state));
2565 if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2566 session->s_state != CEPH_MDS_SESSION_HUNG) {
2567 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2568 err = -EACCES;
2569 goto out_session;
2570 }
2571 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2572 session->s_state == CEPH_MDS_SESSION_CLOSING)
2573 __open_session(mdsc, session);
2574 list_add(&req->r_wait, &session->s_waiting);
2575 goto out_session;
2576 }
2577
2578 /* send request */
2579 req->r_resend_mds = -1; /* forget any previous mds hint */
2580
2581 if (req->r_request_started == 0) /* note request start time */
2582 req->r_request_started = jiffies;
2583
2584 err = __prepare_send_request(mdsc, req, mds, false);
2585 if (!err) {
2586 ceph_msg_get(req->r_request);
2587 ceph_con_send(&session->s_con, req->r_request);
2588 }
2589
2590 out_session:
2591 ceph_put_mds_session(session);
2592 finish:
2593 if (err) {
2594 dout("__do_request early error %d\n", err);
2595 req->r_err = err;
2596 complete_request(mdsc, req);
2597 __unregister_request(mdsc, req);
2598 }
2599 return;
2600 }
2601
2602 /*
2603 * called under mdsc->mutex
2604 */
2605 static void __wake_requests(struct ceph_mds_client *mdsc,
2606 struct list_head *head)
2607 {
2608 struct ceph_mds_request *req;
2609 LIST_HEAD(tmp_list);
2610
2611 list_splice_init(head, &tmp_list);
2612
2613 while (!list_empty(&tmp_list)) {
2614 req = list_entry(tmp_list.next,
2615 struct ceph_mds_request, r_wait);
2616 list_del_init(&req->r_wait);
2617 dout(" wake request %p tid %llu\n", req, req->r_tid);
2618 __do_request(mdsc, req);
2619 }
2620 }
2621
2622 /*
2623 * Wake up threads with requests pending for @mds, so that they can
2624 * resubmit their requests to a possibly different mds.
2625 */
2626 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2627 {
2628 struct ceph_mds_request *req;
2629 struct rb_node *p = rb_first(&mdsc->request_tree);
2630
2631 dout("kick_requests mds%d\n", mds);
2632 while (p) {
2633 req = rb_entry(p, struct ceph_mds_request, r_node);
2634 p = rb_next(p);
2635 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2636 continue;
2637 if (req->r_attempts > 0)
2638 continue; /* only new requests */
2639 if (req->r_session &&
2640 req->r_session->s_mds == mds) {
2641 dout(" kicking tid %llu\n", req->r_tid);
2642 list_del_init(&req->r_wait);
2643 __do_request(mdsc, req);
2644 }
2645 }
2646 }
2647
2648 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
2649 struct ceph_mds_request *req)
2650 {
2651 int err;
2652
2653 /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
2654 if (req->r_inode)
2655 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2656 if (req->r_parent)
2657 ceph_get_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
2658 if (req->r_old_dentry_dir)
2659 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2660 CEPH_CAP_PIN);
2661
2662 dout("submit_request on %p for inode %p\n", req, dir);
2663 mutex_lock(&mdsc->mutex);
2664 __register_request(mdsc, req, dir);
2665 __do_request(mdsc, req);
2666 err = req->r_err;
2667 mutex_unlock(&mdsc->mutex);
2668 return err;
2669 }
2670
2671 static int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
2672 struct ceph_mds_request *req)
2673 {
2674 int err;
2675
2676 /* wait */
2677 dout("do_request waiting\n");
2678 if (!req->r_timeout && req->r_wait_for_completion) {
2679 err = req->r_wait_for_completion(mdsc, req);
2680 } else {
2681 long timeleft = wait_for_completion_killable_timeout(
2682 &req->r_completion,
2683 ceph_timeout_jiffies(req->r_timeout));
2684 if (timeleft > 0)
2685 err = 0;
2686 else if (!timeleft)
2687 err = -EIO; /* timed out */
2688 else
2689 err = timeleft; /* killed */
2690 }
2691 dout("do_request waited, got %d\n", err);
2692 mutex_lock(&mdsc->mutex);
2693
2694 /* only abort if we didn't race with a real reply */
2695 if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2696 err = le32_to_cpu(req->r_reply_info.head->result);
2697 } else if (err < 0) {
2698 dout("aborted request %lld with %d\n", req->r_tid, err);
2699
2700 /*
2701 * ensure we aren't running concurrently with
2702 * ceph_fill_trace or ceph_readdir_prepopulate, which
2703 * rely on locks (dir mutex) held by our caller.
2704 */
2705 mutex_lock(&req->r_fill_mutex);
2706 req->r_err = err;
2707 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
2708 mutex_unlock(&req->r_fill_mutex);
2709
2710 if (req->r_parent &&
2711 (req->r_op & CEPH_MDS_OP_WRITE))
2712 ceph_invalidate_dir_request(req);
2713 } else {
2714 err = req->r_err;
2715 }
2716
2717 mutex_unlock(&mdsc->mutex);
2718 return err;
2719 }
2720
2721 /*
2722 * Synchrously perform an mds request. Take care of all of the
2723 * session setup, forwarding, retry details.
2724 */
2725 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
2726 struct inode *dir,
2727 struct ceph_mds_request *req)
2728 {
2729 int err;
2730
2731 dout("do_request on %p\n", req);
2732
2733 /* issue */
2734 err = ceph_mdsc_submit_request(mdsc, dir, req);
2735 if (!err)
2736 err = ceph_mdsc_wait_request(mdsc, req);
2737 dout("do_request %p done, result %d\n", req, err);
2738 return err;
2739 }
2740
2741 /*
2742 * Invalidate dir's completeness, dentry lease state on an aborted MDS
2743 * namespace request.
2744 */
2745 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
2746 {
2747 struct inode *dir = req->r_parent;
2748 struct inode *old_dir = req->r_old_dentry_dir;
2749
2750 dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
2751
2752 ceph_dir_clear_complete(dir);
2753 if (old_dir)
2754 ceph_dir_clear_complete(old_dir);
2755 if (req->r_dentry)
2756 ceph_invalidate_dentry_lease(req->r_dentry);
2757 if (req->r_old_dentry)
2758 ceph_invalidate_dentry_lease(req->r_old_dentry);
2759 }
2760
2761 /*
2762 * Handle mds reply.
2763 *
2764 * We take the session mutex and parse and process the reply immediately.
2765 * This preserves the logical ordering of replies, capabilities, etc., sent
2766 * by the MDS as they are applied to our local cache.
2767 */
2768 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
2769 {
2770 struct ceph_mds_client *mdsc = session->s_mdsc;
2771 struct ceph_mds_request *req;
2772 struct ceph_mds_reply_head *head = msg->front.iov_base;
2773 struct ceph_mds_reply_info_parsed *rinfo; /* parsed reply info */
2774 struct ceph_snap_realm *realm;
2775 u64 tid;
2776 int err, result;
2777 int mds = session->s_mds;
2778
2779 if (msg->front.iov_len < sizeof(*head)) {
2780 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
2781 ceph_msg_dump(msg);
2782 return;
2783 }
2784
2785 /* get request, session */
2786 tid = le64_to_cpu(msg->hdr.tid);
2787 mutex_lock(&mdsc->mutex);
2788 req = lookup_get_request(mdsc, tid);
2789 if (!req) {
2790 dout("handle_reply on unknown tid %llu\n", tid);
2791 mutex_unlock(&mdsc->mutex);
2792 return;
2793 }
2794 dout("handle_reply %p\n", req);
2795
2796 /* correct session? */
2797 if (req->r_session != session) {
2798 pr_err("mdsc_handle_reply got %llu on session mds%d"
2799 " not mds%d\n", tid, session->s_mds,
2800 req->r_session ? req->r_session->s_mds : -1);
2801 mutex_unlock(&mdsc->mutex);
2802 goto out;
2803 }
2804
2805 /* dup? */
2806 if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
2807 (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
2808 pr_warn("got a dup %s reply on %llu from mds%d\n",
2809 head->safe ? "safe" : "unsafe", tid, mds);
2810 mutex_unlock(&mdsc->mutex);
2811 goto out;
2812 }
2813 if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
2814 pr_warn("got unsafe after safe on %llu from mds%d\n",
2815 tid, mds);
2816 mutex_unlock(&mdsc->mutex);
2817 goto out;
2818 }
2819
2820 result = le32_to_cpu(head->result);
2821
2822 /*
2823 * Handle an ESTALE
2824 * if we're not talking to the authority, send to them
2825 * if the authority has changed while we weren't looking,
2826 * send to new authority
2827 * Otherwise we just have to return an ESTALE
2828 */
2829 if (result == -ESTALE) {
2830 dout("got ESTALE on request %llu\n", req->r_tid);
2831 req->r_resend_mds = -1;
2832 if (req->r_direct_mode != USE_AUTH_MDS) {
2833 dout("not using auth, setting for that now\n");
2834 req->r_direct_mode = USE_AUTH_MDS;
2835 __do_request(mdsc, req);
2836 mutex_unlock(&mdsc->mutex);
2837 goto out;
2838 } else {
2839 int mds = __choose_mds(mdsc, req);
2840 if (mds >= 0 && mds != req->r_session->s_mds) {
2841 dout("but auth changed, so resending\n");
2842 __do_request(mdsc, req);
2843 mutex_unlock(&mdsc->mutex);
2844 goto out;
2845 }
2846 }
2847 dout("have to return ESTALE on request %llu\n", req->r_tid);
2848 }
2849
2850
2851 if (head->safe) {
2852 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
2853 __unregister_request(mdsc, req);
2854
2855 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2856 /*
2857 * We already handled the unsafe response, now do the
2858 * cleanup. No need to examine the response; the MDS
2859 * doesn't include any result info in the safe
2860 * response. And even if it did, there is nothing
2861 * useful we could do with a revised return value.
2862 */
2863 dout("got safe reply %llu, mds%d\n", tid, mds);
2864
2865 /* last unsafe request during umount? */
2866 if (mdsc->stopping && !__get_oldest_req(mdsc))
2867 complete_all(&mdsc->safe_umount_waiters);
2868 mutex_unlock(&mdsc->mutex);
2869 goto out;
2870 }
2871 } else {
2872 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
2873 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
2874 if (req->r_unsafe_dir) {
2875 struct ceph_inode_info *ci =
2876 ceph_inode(req->r_unsafe_dir);
2877 spin_lock(&ci->i_unsafe_lock);
2878 list_add_tail(&req->r_unsafe_dir_item,
2879 &ci->i_unsafe_dirops);
2880 spin_unlock(&ci->i_unsafe_lock);
2881 }
2882 }
2883
2884 dout("handle_reply tid %lld result %d\n", tid, result);
2885 rinfo = &req->r_reply_info;
2886 if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
2887 err = parse_reply_info(msg, rinfo, (u64)-1);
2888 else
2889 err = parse_reply_info(msg, rinfo, session->s_con.peer_features);
2890 mutex_unlock(&mdsc->mutex);
2891
2892 mutex_lock(&session->s_mutex);
2893 if (err < 0) {
2894 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
2895 ceph_msg_dump(msg);
2896 goto out_err;
2897 }
2898
2899 /* snap trace */
2900 realm = NULL;
2901 if (rinfo->snapblob_len) {
2902 down_write(&mdsc->snap_rwsem);
2903 ceph_update_snap_trace(mdsc, rinfo->snapblob,
2904 rinfo->snapblob + rinfo->snapblob_len,
2905 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
2906 &realm);
2907 downgrade_write(&mdsc->snap_rwsem);
2908 } else {
2909 down_read(&mdsc->snap_rwsem);
2910 }
2911
2912 /* insert trace into our cache */
2913 mutex_lock(&req->r_fill_mutex);
2914 current->journal_info = req;
2915 err = ceph_fill_trace(mdsc->fsc->sb, req);
2916 if (err == 0) {
2917 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
2918 req->r_op == CEPH_MDS_OP_LSSNAP))
2919 ceph_readdir_prepopulate(req, req->r_session);
2920 }
2921 current->journal_info = NULL;
2922 mutex_unlock(&req->r_fill_mutex);
2923
2924 up_read(&mdsc->snap_rwsem);
2925 if (realm)
2926 ceph_put_snap_realm(mdsc, realm);
2927
2928 if (err == 0) {
2929 if (req->r_target_inode &&
2930 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2931 struct ceph_inode_info *ci =
2932 ceph_inode(req->r_target_inode);
2933 spin_lock(&ci->i_unsafe_lock);
2934 list_add_tail(&req->r_unsafe_target_item,
2935 &ci->i_unsafe_iops);
2936 spin_unlock(&ci->i_unsafe_lock);
2937 }
2938
2939 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
2940 }
2941 out_err:
2942 mutex_lock(&mdsc->mutex);
2943 if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
2944 if (err) {
2945 req->r_err = err;
2946 } else {
2947 req->r_reply = ceph_msg_get(msg);
2948 set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
2949 }
2950 } else {
2951 dout("reply arrived after request %lld was aborted\n", tid);
2952 }
2953 mutex_unlock(&mdsc->mutex);
2954
2955 mutex_unlock(&session->s_mutex);
2956
2957 /* kick calling process */
2958 complete_request(mdsc, req);
2959 out:
2960 ceph_mdsc_put_request(req);
2961 return;
2962 }
2963
2964
2965
2966 /*
2967 * handle mds notification that our request has been forwarded.
2968 */
2969 static void handle_forward(struct ceph_mds_client *mdsc,
2970 struct ceph_mds_session *session,
2971 struct ceph_msg *msg)
2972 {
2973 struct ceph_mds_request *req;
2974 u64 tid = le64_to_cpu(msg->hdr.tid);
2975 u32 next_mds;
2976 u32 fwd_seq;
2977 int err = -EINVAL;
2978 void *p = msg->front.iov_base;
2979 void *end = p + msg->front.iov_len;
2980
2981 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
2982 next_mds = ceph_decode_32(&p);
2983 fwd_seq = ceph_decode_32(&p);
2984
2985 mutex_lock(&mdsc->mutex);
2986 req = lookup_get_request(mdsc, tid);
2987 if (!req) {
2988 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
2989 goto out; /* dup reply? */
2990 }
2991
2992 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
2993 dout("forward tid %llu aborted, unregistering\n", tid);
2994 __unregister_request(mdsc, req);
2995 } else if (fwd_seq <= req->r_num_fwd) {
2996 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
2997 tid, next_mds, req->r_num_fwd, fwd_seq);
2998 } else {
2999 /* resend. forward race not possible; mds would drop */
3000 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3001 BUG_ON(req->r_err);
3002 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3003 req->r_attempts = 0;
3004 req->r_num_fwd = fwd_seq;
3005 req->r_resend_mds = next_mds;
3006 put_request_session(req);
3007 __do_request(mdsc, req);
3008 }
3009 ceph_mdsc_put_request(req);
3010 out:
3011 mutex_unlock(&mdsc->mutex);
3012 return;
3013
3014 bad:
3015 pr_err("mdsc_handle_forward decode error err=%d\n", err);
3016 }
3017
3018 static int __decode_and_drop_session_metadata(void **p, void *end)
3019 {
3020 /* map<string,string> */
3021 u32 n;
3022 ceph_decode_32_safe(p, end, n, bad);
3023 while (n-- > 0) {
3024 u32 len;
3025 ceph_decode_32_safe(p, end, len, bad);
3026 ceph_decode_need(p, end, len, bad);
3027 *p += len;
3028 ceph_decode_32_safe(p, end, len, bad);
3029 ceph_decode_need(p, end, len, bad);
3030 *p += len;
3031 }
3032 return 0;
3033 bad:
3034 return -1;
3035 }
3036
3037 /*
3038 * handle a mds session control message
3039 */
3040 static void handle_session(struct ceph_mds_session *session,
3041 struct ceph_msg *msg)
3042 {
3043 struct ceph_mds_client *mdsc = session->s_mdsc;
3044 int mds = session->s_mds;
3045 int msg_version = le16_to_cpu(msg->hdr.version);
3046 void *p = msg->front.iov_base;
3047 void *end = p + msg->front.iov_len;
3048 struct ceph_mds_session_head *h;
3049 u32 op;
3050 u64 seq;
3051 unsigned long features = 0;
3052 int wake = 0;
3053
3054 /* decode */
3055 ceph_decode_need(&p, end, sizeof(*h), bad);
3056 h = p;
3057 p += sizeof(*h);
3058
3059 op = le32_to_cpu(h->op);
3060 seq = le64_to_cpu(h->seq);
3061
3062 if (msg_version >= 3) {
3063 u32 len;
3064 /* version >= 2, metadata */
3065 if (__decode_and_drop_session_metadata(&p, end) < 0)
3066 goto bad;
3067 /* version >= 3, feature bits */
3068 ceph_decode_32_safe(&p, end, len, bad);
3069 ceph_decode_need(&p, end, len, bad);
3070 memcpy(&features, p, min_t(size_t, len, sizeof(features)));
3071 p += len;
3072 }
3073
3074 mutex_lock(&mdsc->mutex);
3075 if (op == CEPH_SESSION_CLOSE) {
3076 get_session(session);
3077 __unregister_session(mdsc, session);
3078 }
3079 /* FIXME: this ttl calculation is generous */
3080 session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3081 mutex_unlock(&mdsc->mutex);
3082
3083 mutex_lock(&session->s_mutex);
3084
3085 dout("handle_session mds%d %s %p state %s seq %llu\n",
3086 mds, ceph_session_op_name(op), session,
3087 ceph_session_state_name(session->s_state), seq);
3088
3089 if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3090 session->s_state = CEPH_MDS_SESSION_OPEN;
3091 pr_info("mds%d came back\n", session->s_mds);
3092 }
3093
3094 switch (op) {
3095 case CEPH_SESSION_OPEN:
3096 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3097 pr_info("mds%d reconnect success\n", session->s_mds);
3098 session->s_state = CEPH_MDS_SESSION_OPEN;
3099 session->s_features = features;
3100 renewed_caps(mdsc, session, 0);
3101 wake = 1;
3102 if (mdsc->stopping)
3103 __close_session(mdsc, session);
3104 break;
3105
3106 case CEPH_SESSION_RENEWCAPS:
3107 if (session->s_renew_seq == seq)
3108 renewed_caps(mdsc, session, 1);
3109 break;
3110
3111 case CEPH_SESSION_CLOSE:
3112 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3113 pr_info("mds%d reconnect denied\n", session->s_mds);
3114 cleanup_session_requests(mdsc, session);
3115 remove_session_caps(session);
3116 wake = 2; /* for good measure */
3117 wake_up_all(&mdsc->session_close_wq);
3118 break;
3119
3120 case CEPH_SESSION_STALE:
3121 pr_info("mds%d caps went stale, renewing\n",
3122 session->s_mds);
3123 spin_lock(&session->s_gen_ttl_lock);
3124 session->s_cap_gen++;
3125 session->s_cap_ttl = jiffies - 1;
3126 spin_unlock(&session->s_gen_ttl_lock);
3127 send_renew_caps(mdsc, session);
3128 break;
3129
3130 case CEPH_SESSION_RECALL_STATE:
3131 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
3132 break;
3133
3134 case CEPH_SESSION_FLUSHMSG:
3135 send_flushmsg_ack(mdsc, session, seq);
3136 break;
3137
3138 case CEPH_SESSION_FORCE_RO:
3139 dout("force_session_readonly %p\n", session);
3140 spin_lock(&session->s_cap_lock);
3141 session->s_readonly = true;
3142 spin_unlock(&session->s_cap_lock);
3143 wake_up_session_caps(session, FORCE_RO);
3144 break;
3145
3146 case CEPH_SESSION_REJECT:
3147 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3148 pr_info("mds%d rejected session\n", session->s_mds);
3149 session->s_state = CEPH_MDS_SESSION_REJECTED;
3150 cleanup_session_requests(mdsc, session);
3151 remove_session_caps(session);
3152 wake = 2; /* for good measure */
3153 break;
3154
3155 default:
3156 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3157 WARN_ON(1);
3158 }
3159
3160 mutex_unlock(&session->s_mutex);
3161 if (wake) {
3162 mutex_lock(&mdsc->mutex);
3163 __wake_requests(mdsc, &session->s_waiting);
3164 if (wake == 2)
3165 kick_requests(mdsc, mds);
3166 mutex_unlock(&mdsc->mutex);
3167 }
3168 if (op == CEPH_SESSION_CLOSE)
3169 ceph_put_mds_session(session);
3170 return;
3171
3172 bad:
3173 pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3174 (int)msg->front.iov_len);
3175 ceph_msg_dump(msg);
3176 return;
3177 }
3178
3179
3180 /*
3181 * called under session->mutex.
3182 */
3183 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3184 struct ceph_mds_session *session)
3185 {
3186 struct ceph_mds_request *req, *nreq;
3187 struct rb_node *p;
3188 int err;
3189
3190 dout("replay_unsafe_requests mds%d\n", session->s_mds);
3191
3192 mutex_lock(&mdsc->mutex);
3193 list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item) {
3194 err = __prepare_send_request(mdsc, req, session->s_mds, true);
3195 if (!err) {
3196 ceph_msg_get(req->r_request);
3197 ceph_con_send(&session->s_con, req->r_request);
3198 }
3199 }
3200
3201 /*
3202 * also re-send old requests when MDS enters reconnect stage. So that MDS
3203 * can process completed request in clientreplay stage.
3204 */
3205 p = rb_first(&mdsc->request_tree);
3206 while (p) {
3207 req = rb_entry(p, struct ceph_mds_request, r_node);
3208 p = rb_next(p);
3209 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3210 continue;
3211 if (req->r_attempts == 0)
3212 continue; /* only old requests */
3213 if (req->r_session &&
3214 req->r_session->s_mds == session->s_mds) {
3215 err = __prepare_send_request(mdsc, req,
3216 session->s_mds, true);
3217 if (!err) {
3218 ceph_msg_get(req->r_request);
3219 ceph_con_send(&session->s_con, req->r_request);
3220 }
3221 }
3222 }
3223 mutex_unlock(&mdsc->mutex);
3224 }
3225
3226 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3227 {
3228 struct ceph_msg *reply;
3229 struct ceph_pagelist *_pagelist;
3230 struct page *page;
3231 __le32 *addr;
3232 int err = -ENOMEM;
3233
3234 if (!recon_state->allow_multi)
3235 return -ENOSPC;
3236
3237 /* can't handle message that contains both caps and realm */
3238 BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3239
3240 /* pre-allocate new pagelist */
3241 _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3242 if (!_pagelist)
3243 return -ENOMEM;
3244
3245 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3246 if (!reply)
3247 goto fail_msg;
3248
3249 /* placeholder for nr_caps */
3250 err = ceph_pagelist_encode_32(_pagelist, 0);
3251 if (err < 0)
3252 goto fail;
3253
3254 if (recon_state->nr_caps) {
3255 /* currently encoding caps */
3256 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3257 if (err)
3258 goto fail;
3259 } else {
3260 /* placeholder for nr_realms (currently encoding relams) */
3261 err = ceph_pagelist_encode_32(_pagelist, 0);
3262 if (err < 0)
3263 goto fail;
3264 }
3265
3266 err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3267 if (err)
3268 goto fail;
3269
3270 page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3271 addr = kmap_atomic(page);
3272 if (recon_state->nr_caps) {
3273 /* currently encoding caps */
3274 *addr = cpu_to_le32(recon_state->nr_caps);
3275 } else {
3276 /* currently encoding relams */
3277 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3278 }
3279 kunmap_atomic(addr);
3280
3281 reply->hdr.version = cpu_to_le16(5);
3282 reply->hdr.compat_version = cpu_to_le16(4);
3283
3284 reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3285 ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3286
3287 ceph_con_send(&recon_state->session->s_con, reply);
3288 ceph_pagelist_release(recon_state->pagelist);
3289
3290 recon_state->pagelist = _pagelist;
3291 recon_state->nr_caps = 0;
3292 recon_state->nr_realms = 0;
3293 recon_state->msg_version = 5;
3294 return 0;
3295 fail:
3296 ceph_msg_put(reply);
3297 fail_msg:
3298 ceph_pagelist_release(_pagelist);
3299 return err;
3300 }
3301
3302 /*
3303 * Encode information about a cap for a reconnect with the MDS.
3304 */
3305 static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
3306 void *arg)
3307 {
3308 union {
3309 struct ceph_mds_cap_reconnect v2;
3310 struct ceph_mds_cap_reconnect_v1 v1;
3311 } rec;
3312 struct ceph_inode_info *ci = cap->ci;
3313 struct ceph_reconnect_state *recon_state = arg;
3314 struct ceph_pagelist *pagelist = recon_state->pagelist;
3315 int err;
3316 u64 snap_follows;
3317
3318 dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3319 inode, ceph_vinop(inode), cap, cap->cap_id,
3320 ceph_cap_string(cap->issued));
3321
3322 spin_lock(&ci->i_ceph_lock);
3323 cap->seq = 0; /* reset cap seq */
3324 cap->issue_seq = 0; /* and issue_seq */
3325 cap->mseq = 0; /* and migrate_seq */
3326 cap->cap_gen = cap->session->s_cap_gen;
3327
3328 if (recon_state->msg_version >= 2) {
3329 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3330 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3331 rec.v2.issued = cpu_to_le32(cap->issued);
3332 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3333 rec.v2.pathbase = 0;
3334 rec.v2.flock_len = (__force __le32)
3335 ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
3336 } else {
3337 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3338 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3339 rec.v1.issued = cpu_to_le32(cap->issued);
3340 rec.v1.size = cpu_to_le64(inode->i_size);
3341 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3342 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
3343 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3344 rec.v1.pathbase = 0;
3345 }
3346
3347 if (list_empty(&ci->i_cap_snaps)) {
3348 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3349 } else {
3350 struct ceph_cap_snap *capsnap =
3351 list_first_entry(&ci->i_cap_snaps,
3352 struct ceph_cap_snap, ci_item);
3353 snap_follows = capsnap->follows;
3354 }
3355 spin_unlock(&ci->i_ceph_lock);
3356
3357 if (recon_state->msg_version >= 2) {
3358 int num_fcntl_locks, num_flock_locks;
3359 struct ceph_filelock *flocks = NULL;
3360 size_t struct_len, total_len = sizeof(u64);
3361 u8 struct_v = 0;
3362
3363 encode_again:
3364 if (rec.v2.flock_len) {
3365 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3366 } else {
3367 num_fcntl_locks = 0;
3368 num_flock_locks = 0;
3369 }
3370 if (num_fcntl_locks + num_flock_locks > 0) {
3371 flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3372 sizeof(struct ceph_filelock),
3373 GFP_NOFS);
3374 if (!flocks) {
3375 err = -ENOMEM;
3376 goto out_err;
3377 }
3378 err = ceph_encode_locks_to_buffer(inode, flocks,
3379 num_fcntl_locks,
3380 num_flock_locks);
3381 if (err) {
3382 kfree(flocks);
3383 flocks = NULL;
3384 if (err == -ENOSPC)
3385 goto encode_again;
3386 goto out_err;
3387 }
3388 } else {
3389 kfree(flocks);
3390 flocks = NULL;
3391 }
3392
3393 if (recon_state->msg_version >= 3) {
3394 /* version, compat_version and struct_len */
3395 total_len += 2 * sizeof(u8) + sizeof(u32);
3396 struct_v = 2;
3397 }
3398 /*
3399 * number of encoded locks is stable, so copy to pagelist
3400 */
3401 struct_len = 2 * sizeof(u32) +
3402 (num_fcntl_locks + num_flock_locks) *
3403 sizeof(struct ceph_filelock);
3404 rec.v2.flock_len = cpu_to_le32(struct_len);
3405
3406 struct_len += sizeof(u32) + sizeof(rec.v2);
3407
3408 if (struct_v >= 2)
3409 struct_len += sizeof(u64); /* snap_follows */
3410
3411 total_len += struct_len;
3412
3413 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
3414 err = send_reconnect_partial(recon_state);
3415 if (err)
3416 goto out_freeflocks;
3417 pagelist = recon_state->pagelist;
3418 }
3419
3420 err = ceph_pagelist_reserve(pagelist, total_len);
3421 if (err)
3422 goto out_freeflocks;
3423
3424 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3425 if (recon_state->msg_version >= 3) {
3426 ceph_pagelist_encode_8(pagelist, struct_v);
3427 ceph_pagelist_encode_8(pagelist, 1);
3428 ceph_pagelist_encode_32(pagelist, struct_len);
3429 }
3430 ceph_pagelist_encode_string(pagelist, NULL, 0);
3431 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
3432 ceph_locks_to_pagelist(flocks, pagelist,
3433 num_fcntl_locks, num_flock_locks);
3434 if (struct_v >= 2)
3435 ceph_pagelist_encode_64(pagelist, snap_follows);
3436 out_freeflocks:
3437 kfree(flocks);
3438 } else {
3439 u64 pathbase = 0;
3440 int pathlen = 0;
3441 char *path = NULL;
3442 struct dentry *dentry;
3443
3444 dentry = d_find_alias(inode);
3445 if (dentry) {
3446 path = ceph_mdsc_build_path(dentry,
3447 &pathlen, &pathbase, 0);
3448 dput(dentry);
3449 if (IS_ERR(path)) {
3450 err = PTR_ERR(path);
3451 goto out_err;
3452 }
3453 rec.v1.pathbase = cpu_to_le64(pathbase);
3454 }
3455
3456 err = ceph_pagelist_reserve(pagelist,
3457 sizeof(u64) + sizeof(u32) +
3458 pathlen + sizeof(rec.v1));
3459 if (err) {
3460 goto out_freepath;
3461 }
3462
3463 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3464 ceph_pagelist_encode_string(pagelist, path, pathlen);
3465 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
3466 out_freepath:
3467 ceph_mdsc_free_path(path, pathlen);
3468 }
3469
3470 out_err:
3471 if (err >= 0)
3472 recon_state->nr_caps++;
3473 return err;
3474 }
3475
3476 static int encode_snap_realms(struct ceph_mds_client *mdsc,
3477 struct ceph_reconnect_state *recon_state)
3478 {
3479 struct rb_node *p;
3480 struct ceph_pagelist *pagelist = recon_state->pagelist;
3481 int err = 0;
3482
3483 if (recon_state->msg_version >= 4) {
3484 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
3485 if (err < 0)
3486 goto fail;
3487 }
3488
3489 /*
3490 * snaprealms. we provide mds with the ino, seq (version), and
3491 * parent for all of our realms. If the mds has any newer info,
3492 * it will tell us.
3493 */
3494 for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
3495 struct ceph_snap_realm *realm =
3496 rb_entry(p, struct ceph_snap_realm, node);
3497 struct ceph_mds_snaprealm_reconnect sr_rec;
3498
3499 if (recon_state->msg_version >= 4) {
3500 size_t need = sizeof(u8) * 2 + sizeof(u32) +
3501 sizeof(sr_rec);
3502
3503 if (pagelist->length + need > RECONNECT_MAX_SIZE) {
3504 err = send_reconnect_partial(recon_state);
3505 if (err)
3506 goto fail;
3507 pagelist = recon_state->pagelist;
3508 }
3509
3510 err = ceph_pagelist_reserve(pagelist, need);
3511 if (err)
3512 goto fail;
3513
3514 ceph_pagelist_encode_8(pagelist, 1);
3515 ceph_pagelist_encode_8(pagelist, 1);
3516 ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
3517 }
3518
3519 dout(" adding snap realm %llx seq %lld parent %llx\n",
3520 realm->ino, realm->seq, realm->parent_ino);
3521 sr_rec.ino = cpu_to_le64(realm->ino);
3522 sr_rec.seq = cpu_to_le64(realm->seq);
3523 sr_rec.parent = cpu_to_le64(realm->parent_ino);
3524
3525 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
3526 if (err)
3527 goto fail;
3528
3529 recon_state->nr_realms++;
3530 }
3531 fail:
3532 return err;
3533 }
3534
3535
3536 /*
3537 * If an MDS fails and recovers, clients need to reconnect in order to
3538 * reestablish shared state. This includes all caps issued through
3539 * this session _and_ the snap_realm hierarchy. Because it's not
3540 * clear which snap realms the mds cares about, we send everything we
3541 * know about.. that ensures we'll then get any new info the
3542 * recovering MDS might have.
3543 *
3544 * This is a relatively heavyweight operation, but it's rare.
3545 *
3546 * called with mdsc->mutex held.
3547 */
3548 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
3549 struct ceph_mds_session *session)
3550 {
3551 struct ceph_msg *reply;
3552 int mds = session->s_mds;
3553 int err = -ENOMEM;
3554 struct ceph_reconnect_state recon_state = {
3555 .session = session,
3556 };
3557 LIST_HEAD(dispose);
3558
3559 pr_info("mds%d reconnect start\n", mds);
3560
3561 recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
3562 if (!recon_state.pagelist)
3563 goto fail_nopagelist;
3564
3565 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3566 if (!reply)
3567 goto fail_nomsg;
3568
3569 mutex_lock(&session->s_mutex);
3570 session->s_state = CEPH_MDS_SESSION_RECONNECTING;
3571 session->s_seq = 0;
3572
3573 dout("session %p state %s\n", session,
3574 ceph_session_state_name(session->s_state));
3575
3576 spin_lock(&session->s_gen_ttl_lock);
3577 session->s_cap_gen++;
3578 spin_unlock(&session->s_gen_ttl_lock);
3579
3580 spin_lock(&session->s_cap_lock);
3581 /* don't know if session is readonly */
3582 session->s_readonly = 0;
3583 /*
3584 * notify __ceph_remove_cap() that we are composing cap reconnect.
3585 * If a cap get released before being added to the cap reconnect,
3586 * __ceph_remove_cap() should skip queuing cap release.
3587 */
3588 session->s_cap_reconnect = 1;
3589 /* drop old cap expires; we're about to reestablish that state */
3590 detach_cap_releases(session, &dispose);
3591 spin_unlock(&session->s_cap_lock);
3592 dispose_cap_releases(mdsc, &dispose);
3593
3594 /* trim unused caps to reduce MDS's cache rejoin time */
3595 if (mdsc->fsc->sb->s_root)
3596 shrink_dcache_parent(mdsc->fsc->sb->s_root);
3597
3598 ceph_con_close(&session->s_con);
3599 ceph_con_open(&session->s_con,
3600 CEPH_ENTITY_TYPE_MDS, mds,
3601 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
3602
3603 /* replay unsafe requests */
3604 replay_unsafe_requests(mdsc, session);
3605
3606 ceph_early_kick_flushing_caps(mdsc, session);
3607
3608 down_read(&mdsc->snap_rwsem);
3609
3610 /* placeholder for nr_caps */
3611 err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
3612 if (err)
3613 goto fail;
3614
3615 if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
3616 recon_state.msg_version = 3;
3617 recon_state.allow_multi = true;
3618 } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
3619 recon_state.msg_version = 3;
3620 } else {
3621 recon_state.msg_version = 2;
3622 }
3623 /* trsaverse this session's caps */
3624 err = ceph_iterate_session_caps(session, encode_caps_cb, &recon_state);
3625
3626 spin_lock(&session->s_cap_lock);
3627 session->s_cap_reconnect = 0;
3628 spin_unlock(&session->s_cap_lock);
3629
3630 if (err < 0)
3631 goto fail;
3632
3633 /* check if all realms can be encoded into current message */
3634 if (mdsc->num_snap_realms) {
3635 size_t total_len =
3636 recon_state.pagelist->length +
3637 mdsc->num_snap_realms *
3638 sizeof(struct ceph_mds_snaprealm_reconnect);
3639 if (recon_state.msg_version >= 4) {
3640 /* number of realms */
3641 total_len += sizeof(u32);
3642 /* version, compat_version and struct_len */
3643 total_len += mdsc->num_snap_realms *
3644 (2 * sizeof(u8) + sizeof(u32));
3645 }
3646 if (total_len > RECONNECT_MAX_SIZE) {
3647 if (!recon_state.allow_multi) {
3648 err = -ENOSPC;
3649 goto fail;
3650 }
3651 if (recon_state.nr_caps) {
3652 err = send_reconnect_partial(&recon_state);
3653 if (err)
3654 goto fail;
3655 }
3656 recon_state.msg_version = 5;
3657 }
3658 }
3659
3660 err = encode_snap_realms(mdsc, &recon_state);
3661 if (err < 0)
3662 goto fail;
3663
3664 if (recon_state.msg_version >= 5) {
3665 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
3666 if (err < 0)
3667 goto fail;
3668 }
3669
3670 if (recon_state.nr_caps || recon_state.nr_realms) {
3671 struct page *page =
3672 list_first_entry(&recon_state.pagelist->head,
3673 struct page, lru);
3674 __le32 *addr = kmap_atomic(page);
3675 if (recon_state.nr_caps) {
3676 WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
3677 *addr = cpu_to_le32(recon_state.nr_caps);
3678 } else if (recon_state.msg_version >= 4) {
3679 *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
3680 }
3681 kunmap_atomic(addr);
3682 }
3683
3684 reply->hdr.version = cpu_to_le16(recon_state.msg_version);
3685 if (recon_state.msg_version >= 4)
3686 reply->hdr.compat_version = cpu_to_le16(4);
3687
3688 reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
3689 ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
3690
3691 ceph_con_send(&session->s_con, reply);
3692
3693 mutex_unlock(&session->s_mutex);
3694
3695 mutex_lock(&mdsc->mutex);
3696 __wake_requests(mdsc, &session->s_waiting);
3697 mutex_unlock(&mdsc->mutex);
3698
3699 up_read(&mdsc->snap_rwsem);
3700 ceph_pagelist_release(recon_state.pagelist);
3701 return;
3702
3703 fail:
3704 ceph_msg_put(reply);
3705 up_read(&mdsc->snap_rwsem);
3706 mutex_unlock(&session->s_mutex);
3707 fail_nomsg:
3708 ceph_pagelist_release(recon_state.pagelist);
3709 fail_nopagelist:
3710 pr_err("error %d preparing reconnect for mds%d\n", err, mds);
3711 return;
3712 }
3713
3714
3715 /*
3716 * compare old and new mdsmaps, kicking requests
3717 * and closing out old connections as necessary
3718 *
3719 * called under mdsc->mutex.
3720 */
3721 static void check_new_map(struct ceph_mds_client *mdsc,
3722 struct ceph_mdsmap *newmap,
3723 struct ceph_mdsmap *oldmap)
3724 {
3725 int i;
3726 int oldstate, newstate;
3727 struct ceph_mds_session *s;
3728
3729 dout("check_new_map new %u old %u\n",
3730 newmap->m_epoch, oldmap->m_epoch);
3731
3732 for (i = 0; i < oldmap->m_num_mds && i < mdsc->max_sessions; i++) {
3733 if (!mdsc->sessions[i])
3734 continue;
3735 s = mdsc->sessions[i];
3736 oldstate = ceph_mdsmap_get_state(oldmap, i);
3737 newstate = ceph_mdsmap_get_state(newmap, i);
3738
3739 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
3740 i, ceph_mds_state_name(oldstate),
3741 ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
3742 ceph_mds_state_name(newstate),
3743 ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
3744 ceph_session_state_name(s->s_state));
3745
3746 if (i >= newmap->m_num_mds) {
3747 /* force close session for stopped mds */
3748 get_session(s);
3749 __unregister_session(mdsc, s);
3750 __wake_requests(mdsc, &s->s_waiting);
3751 mutex_unlock(&mdsc->mutex);
3752
3753 mutex_lock(&s->s_mutex);
3754 cleanup_session_requests(mdsc, s);
3755 remove_session_caps(s);
3756 mutex_unlock(&s->s_mutex);
3757
3758 ceph_put_mds_session(s);
3759
3760 mutex_lock(&mdsc->mutex);
3761 kick_requests(mdsc, i);
3762 continue;
3763 }
3764
3765 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
3766 ceph_mdsmap_get_addr(newmap, i),
3767 sizeof(struct ceph_entity_addr))) {
3768 /* just close it */
3769 mutex_unlock(&mdsc->mutex);
3770 mutex_lock(&s->s_mutex);
3771 mutex_lock(&mdsc->mutex);
3772 ceph_con_close(&s->s_con);
3773 mutex_unlock(&s->s_mutex);
3774 s->s_state = CEPH_MDS_SESSION_RESTARTING;
3775 } else if (oldstate == newstate) {
3776 continue; /* nothing new with this mds */
3777 }
3778
3779 /*
3780 * send reconnect?
3781 */
3782 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
3783 newstate >= CEPH_MDS_STATE_RECONNECT) {
3784 mutex_unlock(&mdsc->mutex);
3785 send_mds_reconnect(mdsc, s);
3786 mutex_lock(&mdsc->mutex);
3787 }
3788
3789 /*
3790 * kick request on any mds that has gone active.
3791 */
3792 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
3793 newstate >= CEPH_MDS_STATE_ACTIVE) {
3794 if (oldstate != CEPH_MDS_STATE_CREATING &&
3795 oldstate != CEPH_MDS_STATE_STARTING)
3796 pr_info("mds%d recovery completed\n", s->s_mds);
3797 kick_requests(mdsc, i);
3798 ceph_kick_flushing_caps(mdsc, s);
3799 wake_up_session_caps(s, RECONNECT);
3800 }
3801 }
3802
3803 for (i = 0; i < newmap->m_num_mds && i < mdsc->max_sessions; i++) {
3804 s = mdsc->sessions[i];
3805 if (!s)
3806 continue;
3807 if (!ceph_mdsmap_is_laggy(newmap, i))
3808 continue;
3809 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
3810 s->s_state == CEPH_MDS_SESSION_HUNG ||
3811 s->s_state == CEPH_MDS_SESSION_CLOSING) {
3812 dout(" connecting to export targets of laggy mds%d\n",
3813 i);
3814 __open_export_target_sessions(mdsc, s);
3815 }
3816 }
3817 }
3818
3819
3820
3821 /*
3822 * leases
3823 */
3824
3825 /*
3826 * caller must hold session s_mutex, dentry->d_lock
3827 */
3828 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
3829 {
3830 struct ceph_dentry_info *di = ceph_dentry(dentry);
3831
3832 ceph_put_mds_session(di->lease_session);
3833 di->lease_session = NULL;
3834 }
3835
3836 static void handle_lease(struct ceph_mds_client *mdsc,
3837 struct ceph_mds_session *session,
3838 struct ceph_msg *msg)
3839 {
3840 struct super_block *sb = mdsc->fsc->sb;
3841 struct inode *inode;
3842 struct dentry *parent, *dentry;
3843 struct ceph_dentry_info *di;
3844 int mds = session->s_mds;
3845 struct ceph_mds_lease *h = msg->front.iov_base;
3846 u32 seq;
3847 struct ceph_vino vino;
3848 struct qstr dname;
3849 int release = 0;
3850
3851 dout("handle_lease from mds%d\n", mds);
3852
3853 /* decode */
3854 if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
3855 goto bad;
3856 vino.ino = le64_to_cpu(h->ino);
3857 vino.snap = CEPH_NOSNAP;
3858 seq = le32_to_cpu(h->seq);
3859 dname.len = get_unaligned_le32(h + 1);
3860 if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
3861 goto bad;
3862 dname.name = (void *)(h + 1) + sizeof(u32);
3863
3864 /* lookup inode */
3865 inode = ceph_find_inode(sb, vino);
3866 dout("handle_lease %s, ino %llx %p %.*s\n",
3867 ceph_lease_op_name(h->action), vino.ino, inode,
3868 dname.len, dname.name);
3869
3870 mutex_lock(&session->s_mutex);
3871 session->s_seq++;
3872
3873 if (!inode) {
3874 dout("handle_lease no inode %llx\n", vino.ino);
3875 goto release;
3876 }
3877
3878 /* dentry */
3879 parent = d_find_alias(inode);
3880 if (!parent) {
3881 dout("no parent dentry on inode %p\n", inode);
3882 WARN_ON(1);
3883 goto release; /* hrm... */
3884 }
3885 dname.hash = full_name_hash(parent, dname.name, dname.len);
3886 dentry = d_lookup(parent, &dname);
3887 dput(parent);
3888 if (!dentry)
3889 goto release;
3890
3891 spin_lock(&dentry->d_lock);
3892 di = ceph_dentry(dentry);
3893 switch (h->action) {
3894 case CEPH_MDS_LEASE_REVOKE:
3895 if (di->lease_session == session) {
3896 if (ceph_seq_cmp(di->lease_seq, seq) > 0)
3897 h->seq = cpu_to_le32(di->lease_seq);
3898 __ceph_mdsc_drop_dentry_lease(dentry);
3899 }
3900 release = 1;
3901 break;
3902
3903 case CEPH_MDS_LEASE_RENEW:
3904 if (di->lease_session == session &&
3905 di->lease_gen == session->s_cap_gen &&
3906 di->lease_renew_from &&
3907 di->lease_renew_after == 0) {
3908 unsigned long duration =
3909 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
3910
3911 di->lease_seq = seq;
3912 di->time = di->lease_renew_from + duration;
3913 di->lease_renew_after = di->lease_renew_from +
3914 (duration >> 1);
3915 di->lease_renew_from = 0;
3916 }
3917 break;
3918 }
3919 spin_unlock(&dentry->d_lock);
3920 dput(dentry);
3921
3922 if (!release)
3923 goto out;
3924
3925 release:
3926 /* let's just reuse the same message */
3927 h->action = CEPH_MDS_LEASE_REVOKE_ACK;
3928 ceph_msg_get(msg);
3929 ceph_con_send(&session->s_con, msg);
3930
3931 out:
3932 mutex_unlock(&session->s_mutex);
3933 /* avoid calling iput_final() in mds dispatch threads */
3934 ceph_async_iput(inode);
3935 return;
3936
3937 bad:
3938 pr_err("corrupt lease message\n");
3939 ceph_msg_dump(msg);
3940 }
3941
3942 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
3943 struct dentry *dentry, char action,
3944 u32 seq)
3945 {
3946 struct ceph_msg *msg;
3947 struct ceph_mds_lease *lease;
3948 struct inode *dir;
3949 int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
3950
3951 dout("lease_send_msg identry %p %s to mds%d\n",
3952 dentry, ceph_lease_op_name(action), session->s_mds);
3953
3954 msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
3955 if (!msg)
3956 return;
3957 lease = msg->front.iov_base;
3958 lease->action = action;
3959 lease->seq = cpu_to_le32(seq);
3960
3961 spin_lock(&dentry->d_lock);
3962 dir = d_inode(dentry->d_parent);
3963 lease->ino = cpu_to_le64(ceph_ino(dir));
3964 lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
3965
3966 put_unaligned_le32(dentry->d_name.len, lease + 1);
3967 memcpy((void *)(lease + 1) + 4,
3968 dentry->d_name.name, dentry->d_name.len);
3969 spin_unlock(&dentry->d_lock);
3970 /*
3971 * if this is a preemptive lease RELEASE, no need to
3972 * flush request stream, since the actual request will
3973 * soon follow.
3974 */
3975 msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
3976
3977 ceph_con_send(&session->s_con, msg);
3978 }
3979
3980 /*
3981 * lock unlock sessions, to wait ongoing session activities
3982 */
3983 static void lock_unlock_sessions(struct ceph_mds_client *mdsc)
3984 {
3985 int i;
3986
3987 mutex_lock(&mdsc->mutex);
3988 for (i = 0; i < mdsc->max_sessions; i++) {
3989 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
3990 if (!s)
3991 continue;
3992 mutex_unlock(&mdsc->mutex);
3993 mutex_lock(&s->s_mutex);
3994 mutex_unlock(&s->s_mutex);
3995 ceph_put_mds_session(s);
3996 mutex_lock(&mdsc->mutex);
3997 }
3998 mutex_unlock(&mdsc->mutex);
3999 }
4000
4001
4002
4003 /*
4004 * delayed work -- periodically trim expired leases, renew caps with mds
4005 */
4006 static void schedule_delayed(struct ceph_mds_client *mdsc)
4007 {
4008 int delay = 5;
4009 unsigned hz = round_jiffies_relative(HZ * delay);
4010 schedule_delayed_work(&mdsc->delayed_work, hz);
4011 }
4012
4013 static void delayed_work(struct work_struct *work)
4014 {
4015 int i;
4016 struct ceph_mds_client *mdsc =
4017 container_of(work, struct ceph_mds_client, delayed_work.work);
4018 int renew_interval;
4019 int renew_caps;
4020
4021 dout("mdsc delayed_work\n");
4022
4023 mutex_lock(&mdsc->mutex);
4024 renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4025 renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4026 mdsc->last_renew_caps);
4027 if (renew_caps)
4028 mdsc->last_renew_caps = jiffies;
4029
4030 for (i = 0; i < mdsc->max_sessions; i++) {
4031 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4032 if (!s)
4033 continue;
4034 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4035 dout("resending session close request for mds%d\n",
4036 s->s_mds);
4037 request_close_session(mdsc, s);
4038 ceph_put_mds_session(s);
4039 continue;
4040 }
4041 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
4042 if (s->s_state == CEPH_MDS_SESSION_OPEN) {
4043 s->s_state = CEPH_MDS_SESSION_HUNG;
4044 pr_info("mds%d hung\n", s->s_mds);
4045 }
4046 }
4047 if (s->s_state < CEPH_MDS_SESSION_OPEN) {
4048 /* this mds is failed or recovering, just wait */
4049 ceph_put_mds_session(s);
4050 continue;
4051 }
4052 mutex_unlock(&mdsc->mutex);
4053
4054 mutex_lock(&s->s_mutex);
4055 if (renew_caps)
4056 send_renew_caps(mdsc, s);
4057 else
4058 ceph_con_keepalive(&s->s_con);
4059 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4060 s->s_state == CEPH_MDS_SESSION_HUNG)
4061 ceph_send_cap_releases(mdsc, s);
4062 mutex_unlock(&s->s_mutex);
4063 ceph_put_mds_session(s);
4064
4065 mutex_lock(&mdsc->mutex);
4066 }
4067 mutex_unlock(&mdsc->mutex);
4068
4069 ceph_check_delayed_caps(mdsc);
4070
4071 ceph_queue_cap_reclaim_work(mdsc);
4072
4073 ceph_trim_snapid_map(mdsc);
4074
4075 schedule_delayed(mdsc);
4076 }
4077
4078 int ceph_mdsc_init(struct ceph_fs_client *fsc)
4079
4080 {
4081 struct ceph_mds_client *mdsc;
4082
4083 mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4084 if (!mdsc)
4085 return -ENOMEM;
4086 mdsc->fsc = fsc;
4087 mutex_init(&mdsc->mutex);
4088 mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
4089 if (!mdsc->mdsmap) {
4090 kfree(mdsc);
4091 return -ENOMEM;
4092 }
4093
4094 fsc->mdsc = mdsc;
4095 init_completion(&mdsc->safe_umount_waiters);
4096 init_waitqueue_head(&mdsc->session_close_wq);
4097 INIT_LIST_HEAD(&mdsc->waiting_for_map);
4098 mdsc->sessions = NULL;
4099 atomic_set(&mdsc->num_sessions, 0);
4100 mdsc->max_sessions = 0;
4101 mdsc->stopping = 0;
4102 atomic64_set(&mdsc->quotarealms_count, 0);
4103 mdsc->quotarealms_inodes = RB_ROOT;
4104 mutex_init(&mdsc->quotarealms_inodes_mutex);
4105 mdsc->last_snap_seq = 0;
4106 init_rwsem(&mdsc->snap_rwsem);
4107 mdsc->snap_realms = RB_ROOT;
4108 INIT_LIST_HEAD(&mdsc->snap_empty);
4109 mdsc->num_snap_realms = 0;
4110 spin_lock_init(&mdsc->snap_empty_lock);
4111 mdsc->last_tid = 0;
4112 mdsc->oldest_tid = 0;
4113 mdsc->request_tree = RB_ROOT;
4114 INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4115 mdsc->last_renew_caps = jiffies;
4116 INIT_LIST_HEAD(&mdsc->cap_delay_list);
4117 spin_lock_init(&mdsc->cap_delay_lock);
4118 INIT_LIST_HEAD(&mdsc->snap_flush_list);
4119 spin_lock_init(&mdsc->snap_flush_lock);
4120 mdsc->last_cap_flush_tid = 1;
4121 INIT_LIST_HEAD(&mdsc->cap_flush_list);
4122 INIT_LIST_HEAD(&mdsc->cap_dirty);
4123 INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
4124 mdsc->num_cap_flushing = 0;
4125 spin_lock_init(&mdsc->cap_dirty_lock);
4126 init_waitqueue_head(&mdsc->cap_flushing_wq);
4127 INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
4128 atomic_set(&mdsc->cap_reclaim_pending, 0);
4129
4130 spin_lock_init(&mdsc->dentry_list_lock);
4131 INIT_LIST_HEAD(&mdsc->dentry_leases);
4132 INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
4133
4134 ceph_caps_init(mdsc);
4135 ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
4136
4137 spin_lock_init(&mdsc->snapid_map_lock);
4138 mdsc->snapid_map_tree = RB_ROOT;
4139 INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4140
4141 init_rwsem(&mdsc->pool_perm_rwsem);
4142 mdsc->pool_perm_tree = RB_ROOT;
4143
4144 strscpy(mdsc->nodename, utsname()->nodename,
4145 sizeof(mdsc->nodename));
4146 return 0;
4147 }
4148
4149 /*
4150 * Wait for safe replies on open mds requests. If we time out, drop
4151 * all requests from the tree to avoid dangling dentry refs.
4152 */
4153 static void wait_requests(struct ceph_mds_client *mdsc)
4154 {
4155 struct ceph_options *opts = mdsc->fsc->client->options;
4156 struct ceph_mds_request *req;
4157
4158 mutex_lock(&mdsc->mutex);
4159 if (__get_oldest_req(mdsc)) {
4160 mutex_unlock(&mdsc->mutex);
4161
4162 dout("wait_requests waiting for requests\n");
4163 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
4164 ceph_timeout_jiffies(opts->mount_timeout));
4165
4166 /* tear down remaining requests */
4167 mutex_lock(&mdsc->mutex);
4168 while ((req = __get_oldest_req(mdsc))) {
4169 dout("wait_requests timed out on tid %llu\n",
4170 req->r_tid);
4171 list_del_init(&req->r_wait);
4172 __unregister_request(mdsc, req);
4173 }
4174 }
4175 mutex_unlock(&mdsc->mutex);
4176 dout("wait_requests done\n");
4177 }
4178
4179 /*
4180 * called before mount is ro, and before dentries are torn down.
4181 * (hmm, does this still race with new lookups?)
4182 */
4183 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4184 {
4185 dout("pre_umount\n");
4186 mdsc->stopping = 1;
4187
4188 lock_unlock_sessions(mdsc);
4189 ceph_flush_dirty_caps(mdsc);
4190 wait_requests(mdsc);
4191
4192 /*
4193 * wait for reply handlers to drop their request refs and
4194 * their inode/dcache refs
4195 */
4196 ceph_msgr_flush();
4197
4198 ceph_cleanup_quotarealms_inodes(mdsc);
4199 }
4200
4201 /*
4202 * wait for all write mds requests to flush.
4203 */
4204 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
4205 {
4206 struct ceph_mds_request *req = NULL, *nextreq;
4207 struct rb_node *n;
4208
4209 mutex_lock(&mdsc->mutex);
4210 dout("wait_unsafe_requests want %lld\n", want_tid);
4211 restart:
4212 req = __get_oldest_req(mdsc);
4213 while (req && req->r_tid <= want_tid) {
4214 /* find next request */
4215 n = rb_next(&req->r_node);
4216 if (n)
4217 nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4218 else
4219 nextreq = NULL;
4220 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4221 (req->r_op & CEPH_MDS_OP_WRITE)) {
4222 /* write op */
4223 ceph_mdsc_get_request(req);
4224 if (nextreq)
4225 ceph_mdsc_get_request(nextreq);
4226 mutex_unlock(&mdsc->mutex);
4227 dout("wait_unsafe_requests wait on %llu (want %llu)\n",
4228 req->r_tid, want_tid);
4229 wait_for_completion(&req->r_safe_completion);
4230 mutex_lock(&mdsc->mutex);
4231 ceph_mdsc_put_request(req);
4232 if (!nextreq)
4233 break; /* next dne before, so we're done! */
4234 if (RB_EMPTY_NODE(&nextreq->r_node)) {
4235 /* next request was removed from tree */
4236 ceph_mdsc_put_request(nextreq);
4237 goto restart;
4238 }
4239 ceph_mdsc_put_request(nextreq); /* won't go away */
4240 }
4241 req = nextreq;
4242 }
4243 mutex_unlock(&mdsc->mutex);
4244 dout("wait_unsafe_requests done\n");
4245 }
4246
4247 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4248 {
4249 u64 want_tid, want_flush;
4250
4251 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4252 return;
4253
4254 dout("sync\n");
4255 mutex_lock(&mdsc->mutex);
4256 want_tid = mdsc->last_tid;
4257 mutex_unlock(&mdsc->mutex);
4258
4259 ceph_flush_dirty_caps(mdsc);
4260 spin_lock(&mdsc->cap_dirty_lock);
4261 want_flush = mdsc->last_cap_flush_tid;
4262 if (!list_empty(&mdsc->cap_flush_list)) {
4263 struct ceph_cap_flush *cf =
4264 list_last_entry(&mdsc->cap_flush_list,
4265 struct ceph_cap_flush, g_list);
4266 cf->wake = true;
4267 }
4268 spin_unlock(&mdsc->cap_dirty_lock);
4269
4270 dout("sync want tid %lld flush_seq %lld\n",
4271 want_tid, want_flush);
4272
4273 wait_unsafe_requests(mdsc, want_tid);
4274 wait_caps_flush(mdsc, want_flush);
4275 }
4276
4277 /*
4278 * true if all sessions are closed, or we force unmount
4279 */
4280 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
4281 {
4282 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4283 return true;
4284 return atomic_read(&mdsc->num_sessions) <= skipped;
4285 }
4286
4287 /*
4288 * called after sb is ro.
4289 */
4290 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
4291 {
4292 struct ceph_options *opts = mdsc->fsc->client->options;
4293 struct ceph_mds_session *session;
4294 int i;
4295 int skipped = 0;
4296
4297 dout("close_sessions\n");
4298
4299 /* close sessions */
4300 mutex_lock(&mdsc->mutex);
4301 for (i = 0; i < mdsc->max_sessions; i++) {
4302 session = __ceph_lookup_mds_session(mdsc, i);
4303 if (!session)
4304 continue;
4305 mutex_unlock(&mdsc->mutex);
4306 mutex_lock(&session->s_mutex);
4307 if (__close_session(mdsc, session) <= 0)
4308 skipped++;
4309 mutex_unlock(&session->s_mutex);
4310 ceph_put_mds_session(session);
4311 mutex_lock(&mdsc->mutex);
4312 }
4313 mutex_unlock(&mdsc->mutex);
4314
4315 dout("waiting for sessions to close\n");
4316 wait_event_timeout(mdsc->session_close_wq,
4317 done_closing_sessions(mdsc, skipped),
4318 ceph_timeout_jiffies(opts->mount_timeout));
4319
4320 /* tear down remaining sessions */
4321 mutex_lock(&mdsc->mutex);
4322 for (i = 0; i < mdsc->max_sessions; i++) {
4323 if (mdsc->sessions[i]) {
4324 session = get_session(mdsc->sessions[i]);
4325 __unregister_session(mdsc, session);
4326 mutex_unlock(&mdsc->mutex);
4327 mutex_lock(&session->s_mutex);
4328 remove_session_caps(session);
4329 mutex_unlock(&session->s_mutex);
4330 ceph_put_mds_session(session);
4331 mutex_lock(&mdsc->mutex);
4332 }
4333 }
4334 WARN_ON(!list_empty(&mdsc->cap_delay_list));
4335 mutex_unlock(&mdsc->mutex);
4336
4337 ceph_cleanup_snapid_map(mdsc);
4338 ceph_cleanup_empty_realms(mdsc);
4339
4340 cancel_work_sync(&mdsc->cap_reclaim_work);
4341 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4342
4343 dout("stopped\n");
4344 }
4345
4346 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
4347 {
4348 struct ceph_mds_session *session;
4349 int mds;
4350
4351 dout("force umount\n");
4352
4353 mutex_lock(&mdsc->mutex);
4354 for (mds = 0; mds < mdsc->max_sessions; mds++) {
4355 session = __ceph_lookup_mds_session(mdsc, mds);
4356 if (!session)
4357 continue;
4358 mutex_unlock(&mdsc->mutex);
4359 mutex_lock(&session->s_mutex);
4360 __close_session(mdsc, session);
4361 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
4362 cleanup_session_requests(mdsc, session);
4363 remove_session_caps(session);
4364 }
4365 mutex_unlock(&session->s_mutex);
4366 ceph_put_mds_session(session);
4367 mutex_lock(&mdsc->mutex);
4368 kick_requests(mdsc, mds);
4369 }
4370 __wake_requests(mdsc, &mdsc->waiting_for_map);
4371 mutex_unlock(&mdsc->mutex);
4372 }
4373
4374 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
4375 {
4376 dout("stop\n");
4377 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4378 if (mdsc->mdsmap)
4379 ceph_mdsmap_destroy(mdsc->mdsmap);
4380 kfree(mdsc->sessions);
4381 ceph_caps_finalize(mdsc);
4382 ceph_pool_perm_destroy(mdsc);
4383 }
4384
4385 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
4386 {
4387 struct ceph_mds_client *mdsc = fsc->mdsc;
4388 dout("mdsc_destroy %p\n", mdsc);
4389
4390 if (!mdsc)
4391 return;
4392
4393 /* flush out any connection work with references to us */
4394 ceph_msgr_flush();
4395
4396 ceph_mdsc_stop(mdsc);
4397
4398 fsc->mdsc = NULL;
4399 kfree(mdsc);
4400 dout("mdsc_destroy %p done\n", mdsc);
4401 }
4402
4403 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4404 {
4405 struct ceph_fs_client *fsc = mdsc->fsc;
4406 const char *mds_namespace = fsc->mount_options->mds_namespace;
4407 void *p = msg->front.iov_base;
4408 void *end = p + msg->front.iov_len;
4409 u32 epoch;
4410 u32 map_len;
4411 u32 num_fs;
4412 u32 mount_fscid = (u32)-1;
4413 u8 struct_v, struct_cv;
4414 int err = -EINVAL;
4415
4416 ceph_decode_need(&p, end, sizeof(u32), bad);
4417 epoch = ceph_decode_32(&p);
4418
4419 dout("handle_fsmap epoch %u\n", epoch);
4420
4421 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4422 struct_v = ceph_decode_8(&p);
4423 struct_cv = ceph_decode_8(&p);
4424 map_len = ceph_decode_32(&p);
4425
4426 ceph_decode_need(&p, end, sizeof(u32) * 3, bad);
4427 p += sizeof(u32) * 2; /* skip epoch and legacy_client_fscid */
4428
4429 num_fs = ceph_decode_32(&p);
4430 while (num_fs-- > 0) {
4431 void *info_p, *info_end;
4432 u32 info_len;
4433 u8 info_v, info_cv;
4434 u32 fscid, namelen;
4435
4436 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4437 info_v = ceph_decode_8(&p);
4438 info_cv = ceph_decode_8(&p);
4439 info_len = ceph_decode_32(&p);
4440 ceph_decode_need(&p, end, info_len, bad);
4441 info_p = p;
4442 info_end = p + info_len;
4443 p = info_end;
4444
4445 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
4446 fscid = ceph_decode_32(&info_p);
4447 namelen = ceph_decode_32(&info_p);
4448 ceph_decode_need(&info_p, info_end, namelen, bad);
4449
4450 if (mds_namespace &&
4451 strlen(mds_namespace) == namelen &&
4452 !strncmp(mds_namespace, (char *)info_p, namelen)) {
4453 mount_fscid = fscid;
4454 break;
4455 }
4456 }
4457
4458 ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
4459 if (mount_fscid != (u32)-1) {
4460 fsc->client->monc.fs_cluster_id = mount_fscid;
4461 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
4462 0, true);
4463 ceph_monc_renew_subs(&fsc->client->monc);
4464 } else {
4465 err = -ENOENT;
4466 goto err_out;
4467 }
4468 return;
4469
4470 bad:
4471 pr_err("error decoding fsmap\n");
4472 err_out:
4473 mutex_lock(&mdsc->mutex);
4474 mdsc->mdsmap_err = err;
4475 __wake_requests(mdsc, &mdsc->waiting_for_map);
4476 mutex_unlock(&mdsc->mutex);
4477 }
4478
4479 /*
4480 * handle mds map update.
4481 */
4482 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4483 {
4484 u32 epoch;
4485 u32 maplen;
4486 void *p = msg->front.iov_base;
4487 void *end = p + msg->front.iov_len;
4488 struct ceph_mdsmap *newmap, *oldmap;
4489 struct ceph_fsid fsid;
4490 int err = -EINVAL;
4491
4492 ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
4493 ceph_decode_copy(&p, &fsid, sizeof(fsid));
4494 if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
4495 return;
4496 epoch = ceph_decode_32(&p);
4497 maplen = ceph_decode_32(&p);
4498 dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
4499
4500 /* do we need it? */
4501 mutex_lock(&mdsc->mutex);
4502 if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
4503 dout("handle_map epoch %u <= our %u\n",
4504 epoch, mdsc->mdsmap->m_epoch);
4505 mutex_unlock(&mdsc->mutex);
4506 return;
4507 }
4508
4509 newmap = ceph_mdsmap_decode(&p, end);
4510 if (IS_ERR(newmap)) {
4511 err = PTR_ERR(newmap);
4512 goto bad_unlock;
4513 }
4514
4515 /* swap into place */
4516 if (mdsc->mdsmap) {
4517 oldmap = mdsc->mdsmap;
4518 mdsc->mdsmap = newmap;
4519 check_new_map(mdsc, newmap, oldmap);
4520 ceph_mdsmap_destroy(oldmap);
4521 } else {
4522 mdsc->mdsmap = newmap; /* first mds map */
4523 }
4524 mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
4525 MAX_LFS_FILESIZE);
4526
4527 __wake_requests(mdsc, &mdsc->waiting_for_map);
4528 ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
4529 mdsc->mdsmap->m_epoch);
4530
4531 mutex_unlock(&mdsc->mutex);
4532 schedule_delayed(mdsc);
4533 return;
4534
4535 bad_unlock:
4536 mutex_unlock(&mdsc->mutex);
4537 bad:
4538 pr_err("error decoding mdsmap %d\n", err);
4539 return;
4540 }
4541
4542 static struct ceph_connection *con_get(struct ceph_connection *con)
4543 {
4544 struct ceph_mds_session *s = con->private;
4545
4546 if (get_session(s)) {
4547 dout("mdsc con_get %p ok (%d)\n", s, refcount_read(&s->s_ref));
4548 return con;
4549 }
4550 dout("mdsc con_get %p FAIL\n", s);
4551 return NULL;
4552 }
4553
4554 static void con_put(struct ceph_connection *con)
4555 {
4556 struct ceph_mds_session *s = con->private;
4557
4558 dout("mdsc con_put %p (%d)\n", s, refcount_read(&s->s_ref) - 1);
4559 ceph_put_mds_session(s);
4560 }
4561
4562 /*
4563 * if the client is unresponsive for long enough, the mds will kill
4564 * the session entirely.
4565 */
4566 static void peer_reset(struct ceph_connection *con)
4567 {
4568 struct ceph_mds_session *s = con->private;
4569 struct ceph_mds_client *mdsc = s->s_mdsc;
4570
4571 pr_warn("mds%d closed our session\n", s->s_mds);
4572 send_mds_reconnect(mdsc, s);
4573 }
4574
4575 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
4576 {
4577 struct ceph_mds_session *s = con->private;
4578 struct ceph_mds_client *mdsc = s->s_mdsc;
4579 int type = le16_to_cpu(msg->hdr.type);
4580
4581 mutex_lock(&mdsc->mutex);
4582 if (__verify_registered_session(mdsc, s) < 0) {
4583 mutex_unlock(&mdsc->mutex);
4584 goto out;
4585 }
4586 mutex_unlock(&mdsc->mutex);
4587
4588 switch (type) {
4589 case CEPH_MSG_MDS_MAP:
4590 ceph_mdsc_handle_mdsmap(mdsc, msg);
4591 break;
4592 case CEPH_MSG_FS_MAP_USER:
4593 ceph_mdsc_handle_fsmap(mdsc, msg);
4594 break;
4595 case CEPH_MSG_CLIENT_SESSION:
4596 handle_session(s, msg);
4597 break;
4598 case CEPH_MSG_CLIENT_REPLY:
4599 handle_reply(s, msg);
4600 break;
4601 case CEPH_MSG_CLIENT_REQUEST_FORWARD:
4602 handle_forward(mdsc, s, msg);
4603 break;
4604 case CEPH_MSG_CLIENT_CAPS:
4605 ceph_handle_caps(s, msg);
4606 break;
4607 case CEPH_MSG_CLIENT_SNAP:
4608 ceph_handle_snap(mdsc, s, msg);
4609 break;
4610 case CEPH_MSG_CLIENT_LEASE:
4611 handle_lease(mdsc, s, msg);
4612 break;
4613 case CEPH_MSG_CLIENT_QUOTA:
4614 ceph_handle_quota(mdsc, s, msg);
4615 break;
4616
4617 default:
4618 pr_err("received unknown message type %d %s\n", type,
4619 ceph_msg_type_name(type));
4620 }
4621 out:
4622 ceph_msg_put(msg);
4623 }
4624
4625 /*
4626 * authentication
4627 */
4628
4629 /*
4630 * Note: returned pointer is the address of a structure that's
4631 * managed separately. Caller must *not* attempt to free it.
4632 */
4633 static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
4634 int *proto, int force_new)
4635 {
4636 struct ceph_mds_session *s = con->private;
4637 struct ceph_mds_client *mdsc = s->s_mdsc;
4638 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4639 struct ceph_auth_handshake *auth = &s->s_auth;
4640
4641 if (force_new && auth->authorizer) {
4642 ceph_auth_destroy_authorizer(auth->authorizer);
4643 auth->authorizer = NULL;
4644 }
4645 if (!auth->authorizer) {
4646 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
4647 auth);
4648 if (ret)
4649 return ERR_PTR(ret);
4650 } else {
4651 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
4652 auth);
4653 if (ret)
4654 return ERR_PTR(ret);
4655 }
4656 *proto = ac->protocol;
4657
4658 return auth;
4659 }
4660
4661 static int add_authorizer_challenge(struct ceph_connection *con,
4662 void *challenge_buf, int challenge_buf_len)
4663 {
4664 struct ceph_mds_session *s = con->private;
4665 struct ceph_mds_client *mdsc = s->s_mdsc;
4666 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4667
4668 return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
4669 challenge_buf, challenge_buf_len);
4670 }
4671
4672 static int verify_authorizer_reply(struct ceph_connection *con)
4673 {
4674 struct ceph_mds_session *s = con->private;
4675 struct ceph_mds_client *mdsc = s->s_mdsc;
4676 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4677
4678 return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer);
4679 }
4680
4681 static int invalidate_authorizer(struct ceph_connection *con)
4682 {
4683 struct ceph_mds_session *s = con->private;
4684 struct ceph_mds_client *mdsc = s->s_mdsc;
4685 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4686
4687 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
4688
4689 return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
4690 }
4691
4692 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
4693 struct ceph_msg_header *hdr, int *skip)
4694 {
4695 struct ceph_msg *msg;
4696 int type = (int) le16_to_cpu(hdr->type);
4697 int front_len = (int) le32_to_cpu(hdr->front_len);
4698
4699 if (con->in_msg)
4700 return con->in_msg;
4701
4702 *skip = 0;
4703 msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
4704 if (!msg) {
4705 pr_err("unable to allocate msg type %d len %d\n",
4706 type, front_len);
4707 return NULL;
4708 }
4709
4710 return msg;
4711 }
4712
4713 static int mds_sign_message(struct ceph_msg *msg)
4714 {
4715 struct ceph_mds_session *s = msg->con->private;
4716 struct ceph_auth_handshake *auth = &s->s_auth;
4717
4718 return ceph_auth_sign_message(auth, msg);
4719 }
4720
4721 static int mds_check_message_signature(struct ceph_msg *msg)
4722 {
4723 struct ceph_mds_session *s = msg->con->private;
4724 struct ceph_auth_handshake *auth = &s->s_auth;
4725
4726 return ceph_auth_check_message_signature(auth, msg);
4727 }
4728
4729 static const struct ceph_connection_operations mds_con_ops = {
4730 .get = con_get,
4731 .put = con_put,
4732 .dispatch = dispatch,
4733 .get_authorizer = get_authorizer,
4734 .add_authorizer_challenge = add_authorizer_challenge,
4735 .verify_authorizer_reply = verify_authorizer_reply,
4736 .invalidate_authorizer = invalidate_authorizer,
4737 .peer_reset = peer_reset,
4738 .alloc_msg = mds_alloc_msg,
4739 .sign_message = mds_sign_message,
4740 .check_message_signature = mds_check_message_signature,
4741 };
4742
4743 /* eof */
Linux with added FPC-III support