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 #include <linux/bits.h>
13 #include <linux/ktime.h>
14 #include <linux/bitmap.h>
15
16 #include "super.h"
17 #include "mds_client.h"
18
19 #include <linux/ceph/ceph_features.h>
20 #include <linux/ceph/messenger.h>
21 #include <linux/ceph/decode.h>
22 #include <linux/ceph/pagelist.h>
23 #include <linux/ceph/auth.h>
24 #include <linux/ceph/debugfs.h>
25
26 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
27
28 /*
29 * A cluster of MDS (metadata server) daemons is responsible for
30 * managing the file system namespace (the directory hierarchy and
31 * inodes) and for coordinating shared access to storage. Metadata is
32 * partitioning hierarchically across a number of servers, and that
33 * partition varies over time as the cluster adjusts the distribution
34 * in order to balance load.
35 *
36 * The MDS client is primarily responsible to managing synchronous
37 * metadata requests for operations like open, unlink, and so forth.
38 * If there is a MDS failure, we find out about it when we (possibly
39 * request and) receive a new MDS map, and can resubmit affected
40 * requests.
41 *
42 * For the most part, though, we take advantage of a lossless
43 * communications channel to the MDS, and do not need to worry about
44 * timing out or resubmitting requests.
45 *
46 * We maintain a stateful "session" with each MDS we interact with.
47 * Within each session, we sent periodic heartbeat messages to ensure
48 * any capabilities or leases we have been issues remain valid. If
49 * the session times out and goes stale, our leases and capabilities
50 * are no longer valid.
51 */
52
53 struct ceph_reconnect_state {
54 struct ceph_mds_session *session;
55 int nr_caps, nr_realms;
56 struct ceph_pagelist *pagelist;
57 unsigned msg_version;
58 bool allow_multi;
59 };
60
61 static void __wake_requests(struct ceph_mds_client *mdsc,
62 struct list_head *head);
63 static void ceph_cap_release_work(struct work_struct *work);
64 static void ceph_cap_reclaim_work(struct work_struct *work);
65
66 static const struct ceph_connection_operations mds_con_ops;
67
68
69 /*
70 * mds reply parsing
71 */
72
parse_reply_info_quota(void ** p,void * end,struct ceph_mds_reply_info_in * info)73 static int parse_reply_info_quota(void **p, void *end,
74 struct ceph_mds_reply_info_in *info)
75 {
76 u8 struct_v, struct_compat;
77 u32 struct_len;
78
79 ceph_decode_8_safe(p, end, struct_v, bad);
80 ceph_decode_8_safe(p, end, struct_compat, bad);
81 /* struct_v is expected to be >= 1. we only
82 * understand encoding with struct_compat == 1. */
83 if (!struct_v || struct_compat != 1)
84 goto bad;
85 ceph_decode_32_safe(p, end, struct_len, bad);
86 ceph_decode_need(p, end, struct_len, bad);
87 end = *p + struct_len;
88 ceph_decode_64_safe(p, end, info->max_bytes, bad);
89 ceph_decode_64_safe(p, end, info->max_files, bad);
90 *p = end;
91 return 0;
92 bad:
93 return -EIO;
94 }
95
96 /*
97 * parse individual inode info
98 */
parse_reply_info_in(void ** p,void * end,struct ceph_mds_reply_info_in * info,u64 features)99 static int parse_reply_info_in(void **p, void *end,
100 struct ceph_mds_reply_info_in *info,
101 u64 features)
102 {
103 int err = 0;
104 u8 struct_v = 0;
105
106 if (features == (u64)-1) {
107 u32 struct_len;
108 u8 struct_compat;
109 ceph_decode_8_safe(p, end, struct_v, bad);
110 ceph_decode_8_safe(p, end, struct_compat, bad);
111 /* struct_v is expected to be >= 1. we only understand
112 * encoding with struct_compat == 1. */
113 if (!struct_v || struct_compat != 1)
114 goto bad;
115 ceph_decode_32_safe(p, end, struct_len, bad);
116 ceph_decode_need(p, end, struct_len, bad);
117 end = *p + struct_len;
118 }
119
120 ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
121 info->in = *p;
122 *p += sizeof(struct ceph_mds_reply_inode) +
123 sizeof(*info->in->fragtree.splits) *
124 le32_to_cpu(info->in->fragtree.nsplits);
125
126 ceph_decode_32_safe(p, end, info->symlink_len, bad);
127 ceph_decode_need(p, end, info->symlink_len, bad);
128 info->symlink = *p;
129 *p += info->symlink_len;
130
131 ceph_decode_copy_safe(p, end, &info->dir_layout,
132 sizeof(info->dir_layout), bad);
133 ceph_decode_32_safe(p, end, info->xattr_len, bad);
134 ceph_decode_need(p, end, info->xattr_len, bad);
135 info->xattr_data = *p;
136 *p += info->xattr_len;
137
138 if (features == (u64)-1) {
139 /* inline data */
140 ceph_decode_64_safe(p, end, info->inline_version, bad);
141 ceph_decode_32_safe(p, end, info->inline_len, bad);
142 ceph_decode_need(p, end, info->inline_len, bad);
143 info->inline_data = *p;
144 *p += info->inline_len;
145 /* quota */
146 err = parse_reply_info_quota(p, end, info);
147 if (err < 0)
148 goto out_bad;
149 /* pool namespace */
150 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
151 if (info->pool_ns_len > 0) {
152 ceph_decode_need(p, end, info->pool_ns_len, bad);
153 info->pool_ns_data = *p;
154 *p += info->pool_ns_len;
155 }
156
157 /* btime */
158 ceph_decode_need(p, end, sizeof(info->btime), bad);
159 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
160
161 /* change attribute */
162 ceph_decode_64_safe(p, end, info->change_attr, bad);
163
164 /* dir pin */
165 if (struct_v >= 2) {
166 ceph_decode_32_safe(p, end, info->dir_pin, bad);
167 } else {
168 info->dir_pin = -ENODATA;
169 }
170
171 /* snapshot birth time, remains zero for v<=2 */
172 if (struct_v >= 3) {
173 ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
174 ceph_decode_copy(p, &info->snap_btime,
175 sizeof(info->snap_btime));
176 } else {
177 memset(&info->snap_btime, 0, sizeof(info->snap_btime));
178 }
179
180 /* snapshot count, remains zero for v<=3 */
181 if (struct_v >= 4) {
182 ceph_decode_64_safe(p, end, info->rsnaps, bad);
183 } else {
184 info->rsnaps = 0;
185 }
186
187 *p = end;
188 } else {
189 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
190 ceph_decode_64_safe(p, end, info->inline_version, bad);
191 ceph_decode_32_safe(p, end, info->inline_len, bad);
192 ceph_decode_need(p, end, info->inline_len, bad);
193 info->inline_data = *p;
194 *p += info->inline_len;
195 } else
196 info->inline_version = CEPH_INLINE_NONE;
197
198 if (features & CEPH_FEATURE_MDS_QUOTA) {
199 err = parse_reply_info_quota(p, end, info);
200 if (err < 0)
201 goto out_bad;
202 } else {
203 info->max_bytes = 0;
204 info->max_files = 0;
205 }
206
207 info->pool_ns_len = 0;
208 info->pool_ns_data = NULL;
209 if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
210 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
211 if (info->pool_ns_len > 0) {
212 ceph_decode_need(p, end, info->pool_ns_len, bad);
213 info->pool_ns_data = *p;
214 *p += info->pool_ns_len;
215 }
216 }
217
218 if (features & CEPH_FEATURE_FS_BTIME) {
219 ceph_decode_need(p, end, sizeof(info->btime), bad);
220 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
221 ceph_decode_64_safe(p, end, info->change_attr, bad);
222 }
223
224 info->dir_pin = -ENODATA;
225 /* info->snap_btime and info->rsnaps remain zero */
226 }
227 return 0;
228 bad:
229 err = -EIO;
230 out_bad:
231 return err;
232 }
233
parse_reply_info_dir(void ** p,void * end,struct ceph_mds_reply_dirfrag ** dirfrag,u64 features)234 static int parse_reply_info_dir(void **p, void *end,
235 struct ceph_mds_reply_dirfrag **dirfrag,
236 u64 features)
237 {
238 if (features == (u64)-1) {
239 u8 struct_v, struct_compat;
240 u32 struct_len;
241 ceph_decode_8_safe(p, end, struct_v, bad);
242 ceph_decode_8_safe(p, end, struct_compat, bad);
243 /* struct_v is expected to be >= 1. we only understand
244 * encoding whose struct_compat == 1. */
245 if (!struct_v || struct_compat != 1)
246 goto bad;
247 ceph_decode_32_safe(p, end, struct_len, bad);
248 ceph_decode_need(p, end, struct_len, bad);
249 end = *p + struct_len;
250 }
251
252 ceph_decode_need(p, end, sizeof(**dirfrag), bad);
253 *dirfrag = *p;
254 *p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
255 if (unlikely(*p > end))
256 goto bad;
257 if (features == (u64)-1)
258 *p = end;
259 return 0;
260 bad:
261 return -EIO;
262 }
263
parse_reply_info_lease(void ** p,void * end,struct ceph_mds_reply_lease ** lease,u64 features)264 static int parse_reply_info_lease(void **p, void *end,
265 struct ceph_mds_reply_lease **lease,
266 u64 features)
267 {
268 if (features == (u64)-1) {
269 u8 struct_v, struct_compat;
270 u32 struct_len;
271 ceph_decode_8_safe(p, end, struct_v, bad);
272 ceph_decode_8_safe(p, end, struct_compat, bad);
273 /* struct_v is expected to be >= 1. we only understand
274 * encoding whose struct_compat == 1. */
275 if (!struct_v || struct_compat != 1)
276 goto bad;
277 ceph_decode_32_safe(p, end, struct_len, bad);
278 ceph_decode_need(p, end, struct_len, bad);
279 end = *p + struct_len;
280 }
281
282 ceph_decode_need(p, end, sizeof(**lease), bad);
283 *lease = *p;
284 *p += sizeof(**lease);
285 if (features == (u64)-1)
286 *p = end;
287 return 0;
288 bad:
289 return -EIO;
290 }
291
292 /*
293 * parse a normal reply, which may contain a (dir+)dentry and/or a
294 * target inode.
295 */
parse_reply_info_trace(void ** p,void * end,struct ceph_mds_reply_info_parsed * info,u64 features)296 static int parse_reply_info_trace(void **p, void *end,
297 struct ceph_mds_reply_info_parsed *info,
298 u64 features)
299 {
300 int err;
301
302 if (info->head->is_dentry) {
303 err = parse_reply_info_in(p, end, &info->diri, features);
304 if (err < 0)
305 goto out_bad;
306
307 err = parse_reply_info_dir(p, end, &info->dirfrag, features);
308 if (err < 0)
309 goto out_bad;
310
311 ceph_decode_32_safe(p, end, info->dname_len, bad);
312 ceph_decode_need(p, end, info->dname_len, bad);
313 info->dname = *p;
314 *p += info->dname_len;
315
316 err = parse_reply_info_lease(p, end, &info->dlease, features);
317 if (err < 0)
318 goto out_bad;
319 }
320
321 if (info->head->is_target) {
322 err = parse_reply_info_in(p, end, &info->targeti, features);
323 if (err < 0)
324 goto out_bad;
325 }
326
327 if (unlikely(*p != end))
328 goto bad;
329 return 0;
330
331 bad:
332 err = -EIO;
333 out_bad:
334 pr_err("problem parsing mds trace %d\n", err);
335 return err;
336 }
337
338 /*
339 * parse readdir results
340 */
parse_reply_info_readdir(void ** p,void * end,struct ceph_mds_reply_info_parsed * info,u64 features)341 static int parse_reply_info_readdir(void **p, void *end,
342 struct ceph_mds_reply_info_parsed *info,
343 u64 features)
344 {
345 u32 num, i = 0;
346 int err;
347
348 err = parse_reply_info_dir(p, end, &info->dir_dir, features);
349 if (err < 0)
350 goto out_bad;
351
352 ceph_decode_need(p, end, sizeof(num) + 2, bad);
353 num = ceph_decode_32(p);
354 {
355 u16 flags = ceph_decode_16(p);
356 info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
357 info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
358 info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
359 info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
360 }
361 if (num == 0)
362 goto done;
363
364 BUG_ON(!info->dir_entries);
365 if ((unsigned long)(info->dir_entries + num) >
366 (unsigned long)info->dir_entries + info->dir_buf_size) {
367 pr_err("dir contents are larger than expected\n");
368 WARN_ON(1);
369 goto bad;
370 }
371
372 info->dir_nr = num;
373 while (num) {
374 struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
375 /* dentry */
376 ceph_decode_32_safe(p, end, rde->name_len, bad);
377 ceph_decode_need(p, end, rde->name_len, bad);
378 rde->name = *p;
379 *p += rde->name_len;
380 dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);
381
382 /* dentry lease */
383 err = parse_reply_info_lease(p, end, &rde->lease, features);
384 if (err)
385 goto out_bad;
386 /* inode */
387 err = parse_reply_info_in(p, end, &rde->inode, features);
388 if (err < 0)
389 goto out_bad;
390 /* ceph_readdir_prepopulate() will update it */
391 rde->offset = 0;
392 i++;
393 num--;
394 }
395
396 done:
397 /* Skip over any unrecognized fields */
398 *p = end;
399 return 0;
400
401 bad:
402 err = -EIO;
403 out_bad:
404 pr_err("problem parsing dir contents %d\n", err);
405 return err;
406 }
407
408 /*
409 * parse fcntl F_GETLK results
410 */
parse_reply_info_filelock(void ** p,void * end,struct ceph_mds_reply_info_parsed * info,u64 features)411 static int parse_reply_info_filelock(void **p, void *end,
412 struct ceph_mds_reply_info_parsed *info,
413 u64 features)
414 {
415 if (*p + sizeof(*info->filelock_reply) > end)
416 goto bad;
417
418 info->filelock_reply = *p;
419
420 /* Skip over any unrecognized fields */
421 *p = end;
422 return 0;
423 bad:
424 return -EIO;
425 }
426
427
428 #if BITS_PER_LONG == 64
429
430 #define DELEGATED_INO_AVAILABLE xa_mk_value(1)
431
ceph_parse_deleg_inos(void ** p,void * end,struct ceph_mds_session * s)432 static int ceph_parse_deleg_inos(void **p, void *end,
433 struct ceph_mds_session *s)
434 {
435 u32 sets;
436
437 ceph_decode_32_safe(p, end, sets, bad);
438 dout("got %u sets of delegated inodes\n", sets);
439 while (sets--) {
440 u64 start, len;
441
442 ceph_decode_64_safe(p, end, start, bad);
443 ceph_decode_64_safe(p, end, len, bad);
444
445 /* Don't accept a delegation of system inodes */
446 if (start < CEPH_INO_SYSTEM_BASE) {
447 pr_warn_ratelimited("ceph: ignoring reserved inode range delegation (start=0x%llx len=0x%llx)\n",
448 start, len);
449 continue;
450 }
451 while (len--) {
452 int err = xa_insert(&s->s_delegated_inos, start++,
453 DELEGATED_INO_AVAILABLE,
454 GFP_KERNEL);
455 if (!err) {
456 dout("added delegated inode 0x%llx\n",
457 start - 1);
458 } else if (err == -EBUSY) {
459 pr_warn("MDS delegated inode 0x%llx more than once.\n",
460 start - 1);
461 } else {
462 return err;
463 }
464 }
465 }
466 return 0;
467 bad:
468 return -EIO;
469 }
470
ceph_get_deleg_ino(struct ceph_mds_session * s)471 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
472 {
473 unsigned long ino;
474 void *val;
475
476 xa_for_each(&s->s_delegated_inos, ino, val) {
477 val = xa_erase(&s->s_delegated_inos, ino);
478 if (val == DELEGATED_INO_AVAILABLE)
479 return ino;
480 }
481 return 0;
482 }
483
ceph_restore_deleg_ino(struct ceph_mds_session * s,u64 ino)484 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
485 {
486 return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,
487 GFP_KERNEL);
488 }
489 #else /* BITS_PER_LONG == 64 */
490 /*
491 * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
492 * ignore delegated_inos on 32 bit arch. Maybe eventually add xarrays for top
493 * and bottom words?
494 */
ceph_parse_deleg_inos(void ** p,void * end,struct ceph_mds_session * s)495 static int ceph_parse_deleg_inos(void **p, void *end,
496 struct ceph_mds_session *s)
497 {
498 u32 sets;
499
500 ceph_decode_32_safe(p, end, sets, bad);
501 if (sets)
502 ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);
503 return 0;
504 bad:
505 return -EIO;
506 }
507
ceph_get_deleg_ino(struct ceph_mds_session * s)508 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
509 {
510 return 0;
511 }
512
ceph_restore_deleg_ino(struct ceph_mds_session * s,u64 ino)513 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
514 {
515 return 0;
516 }
517 #endif /* BITS_PER_LONG == 64 */
518
519 /*
520 * parse create results
521 */
parse_reply_info_create(void ** p,void * end,struct ceph_mds_reply_info_parsed * info,u64 features,struct ceph_mds_session * s)522 static int parse_reply_info_create(void **p, void *end,
523 struct ceph_mds_reply_info_parsed *info,
524 u64 features, struct ceph_mds_session *s)
525 {
526 int ret;
527
528 if (features == (u64)-1 ||
529 (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
530 if (*p == end) {
531 /* Malformed reply? */
532 info->has_create_ino = false;
533 } else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {
534 info->has_create_ino = true;
535 /* struct_v, struct_compat, and len */
536 ceph_decode_skip_n(p, end, 2 + sizeof(u32), bad);
537 ceph_decode_64_safe(p, end, info->ino, bad);
538 ret = ceph_parse_deleg_inos(p, end, s);
539 if (ret)
540 return ret;
541 } else {
542 /* legacy */
543 ceph_decode_64_safe(p, end, info->ino, bad);
544 info->has_create_ino = true;
545 }
546 } else {
547 if (*p != end)
548 goto bad;
549 }
550
551 /* Skip over any unrecognized fields */
552 *p = end;
553 return 0;
554 bad:
555 return -EIO;
556 }
557
parse_reply_info_getvxattr(void ** p,void * end,struct ceph_mds_reply_info_parsed * info,u64 features)558 static int parse_reply_info_getvxattr(void **p, void *end,
559 struct ceph_mds_reply_info_parsed *info,
560 u64 features)
561 {
562 u32 value_len;
563
564 ceph_decode_skip_8(p, end, bad); /* skip current version: 1 */
565 ceph_decode_skip_8(p, end, bad); /* skip first version: 1 */
566 ceph_decode_skip_32(p, end, bad); /* skip payload length */
567
568 ceph_decode_32_safe(p, end, value_len, bad);
569
570 if (value_len == end - *p) {
571 info->xattr_info.xattr_value = *p;
572 info->xattr_info.xattr_value_len = value_len;
573 *p = end;
574 return value_len;
575 }
576 bad:
577 return -EIO;
578 }
579
580 /*
581 * parse extra results
582 */
parse_reply_info_extra(void ** p,void * end,struct ceph_mds_reply_info_parsed * info,u64 features,struct ceph_mds_session * s)583 static int parse_reply_info_extra(void **p, void *end,
584 struct ceph_mds_reply_info_parsed *info,
585 u64 features, struct ceph_mds_session *s)
586 {
587 u32 op = le32_to_cpu(info->head->op);
588
589 if (op == CEPH_MDS_OP_GETFILELOCK)
590 return parse_reply_info_filelock(p, end, info, features);
591 else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
592 return parse_reply_info_readdir(p, end, info, features);
593 else if (op == CEPH_MDS_OP_CREATE)
594 return parse_reply_info_create(p, end, info, features, s);
595 else if (op == CEPH_MDS_OP_GETVXATTR)
596 return parse_reply_info_getvxattr(p, end, info, features);
597 else
598 return -EIO;
599 }
600
601 /*
602 * parse entire mds reply
603 */
parse_reply_info(struct ceph_mds_session * s,struct ceph_msg * msg,struct ceph_mds_reply_info_parsed * info,u64 features)604 static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg,
605 struct ceph_mds_reply_info_parsed *info,
606 u64 features)
607 {
608 void *p, *end;
609 u32 len;
610 int err;
611
612 info->head = msg->front.iov_base;
613 p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
614 end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
615
616 /* trace */
617 ceph_decode_32_safe(&p, end, len, bad);
618 if (len > 0) {
619 ceph_decode_need(&p, end, len, bad);
620 err = parse_reply_info_trace(&p, p+len, info, features);
621 if (err < 0)
622 goto out_bad;
623 }
624
625 /* extra */
626 ceph_decode_32_safe(&p, end, len, bad);
627 if (len > 0) {
628 ceph_decode_need(&p, end, len, bad);
629 err = parse_reply_info_extra(&p, p+len, info, features, s);
630 if (err < 0)
631 goto out_bad;
632 }
633
634 /* snap blob */
635 ceph_decode_32_safe(&p, end, len, bad);
636 info->snapblob_len = len;
637 info->snapblob = p;
638 p += len;
639
640 if (p != end)
641 goto bad;
642 return 0;
643
644 bad:
645 err = -EIO;
646 out_bad:
647 pr_err("mds parse_reply err %d\n", err);
648 return err;
649 }
650
destroy_reply_info(struct ceph_mds_reply_info_parsed * info)651 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
652 {
653 if (!info->dir_entries)
654 return;
655 free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
656 }
657
658 /*
659 * In async unlink case the kclient won't wait for the first reply
660 * from MDS and just drop all the links and unhash the dentry and then
661 * succeeds immediately.
662 *
663 * For any new create/link/rename,etc requests followed by using the
664 * same file names we must wait for the first reply of the inflight
665 * unlink request, or the MDS possibly will fail these following
666 * requests with -EEXIST if the inflight async unlink request was
667 * delayed for some reasons.
668 *
669 * And the worst case is that for the none async openc request it will
670 * successfully open the file if the CDentry hasn't been unlinked yet,
671 * but later the previous delayed async unlink request will remove the
672 * CDenty. That means the just created file is possiblly deleted later
673 * by accident.
674 *
675 * We need to wait for the inflight async unlink requests to finish
676 * when creating new files/directories by using the same file names.
677 */
ceph_wait_on_conflict_unlink(struct dentry * dentry)678 int ceph_wait_on_conflict_unlink(struct dentry *dentry)
679 {
680 struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
681 struct dentry *pdentry = dentry->d_parent;
682 struct dentry *udentry, *found = NULL;
683 struct ceph_dentry_info *di;
684 struct qstr dname;
685 u32 hash = dentry->d_name.hash;
686 int err;
687
688 dname.name = dentry->d_name.name;
689 dname.len = dentry->d_name.len;
690
691 rcu_read_lock();
692 hash_for_each_possible_rcu(fsc->async_unlink_conflict, di,
693 hnode, hash) {
694 udentry = di->dentry;
695
696 spin_lock(&udentry->d_lock);
697 if (udentry->d_name.hash != hash)
698 goto next;
699 if (unlikely(udentry->d_parent != pdentry))
700 goto next;
701 if (!hash_hashed(&di->hnode))
702 goto next;
703
704 if (!test_bit(CEPH_DENTRY_ASYNC_UNLINK_BIT, &di->flags))
705 pr_warn("%s dentry %p:%pd async unlink bit is not set\n",
706 __func__, dentry, dentry);
707
708 if (!d_same_name(udentry, pdentry, &dname))
709 goto next;
710
711 spin_unlock(&udentry->d_lock);
712 found = dget(udentry);
713 break;
714 next:
715 spin_unlock(&udentry->d_lock);
716 }
717 rcu_read_unlock();
718
719 if (likely(!found))
720 return 0;
721
722 dout("%s dentry %p:%pd conflict with old %p:%pd\n", __func__,
723 dentry, dentry, found, found);
724
725 err = wait_on_bit(&di->flags, CEPH_DENTRY_ASYNC_UNLINK_BIT,
726 TASK_KILLABLE);
727 dput(found);
728 return err;
729 }
730
731
732 /*
733 * sessions
734 */
ceph_session_state_name(int s)735 const char *ceph_session_state_name(int s)
736 {
737 switch (s) {
738 case CEPH_MDS_SESSION_NEW: return "new";
739 case CEPH_MDS_SESSION_OPENING: return "opening";
740 case CEPH_MDS_SESSION_OPEN: return "open";
741 case CEPH_MDS_SESSION_HUNG: return "hung";
742 case CEPH_MDS_SESSION_CLOSING: return "closing";
743 case CEPH_MDS_SESSION_CLOSED: return "closed";
744 case CEPH_MDS_SESSION_RESTARTING: return "restarting";
745 case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
746 case CEPH_MDS_SESSION_REJECTED: return "rejected";
747 default: return "???";
748 }
749 }
750
ceph_get_mds_session(struct ceph_mds_session * s)751 struct ceph_mds_session *ceph_get_mds_session(struct ceph_mds_session *s)
752 {
753 if (refcount_inc_not_zero(&s->s_ref))
754 return s;
755 return NULL;
756 }
757
ceph_put_mds_session(struct ceph_mds_session * s)758 void ceph_put_mds_session(struct ceph_mds_session *s)
759 {
760 if (IS_ERR_OR_NULL(s))
761 return;
762
763 if (refcount_dec_and_test(&s->s_ref)) {
764 if (s->s_auth.authorizer)
765 ceph_auth_destroy_authorizer(s->s_auth.authorizer);
766 WARN_ON(mutex_is_locked(&s->s_mutex));
767 xa_destroy(&s->s_delegated_inos);
768 kfree(s);
769 }
770 }
771
772 /*
773 * called under mdsc->mutex
774 */
__ceph_lookup_mds_session(struct ceph_mds_client * mdsc,int mds)775 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
776 int mds)
777 {
778 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
779 return NULL;
780 return ceph_get_mds_session(mdsc->sessions[mds]);
781 }
782
__have_session(struct ceph_mds_client * mdsc,int mds)783 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
784 {
785 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
786 return false;
787 else
788 return true;
789 }
790
__verify_registered_session(struct ceph_mds_client * mdsc,struct ceph_mds_session * s)791 static int __verify_registered_session(struct ceph_mds_client *mdsc,
792 struct ceph_mds_session *s)
793 {
794 if (s->s_mds >= mdsc->max_sessions ||
795 mdsc->sessions[s->s_mds] != s)
796 return -ENOENT;
797 return 0;
798 }
799
800 /*
801 * create+register a new session for given mds.
802 * called under mdsc->mutex.
803 */
register_session(struct ceph_mds_client * mdsc,int mds)804 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
805 int mds)
806 {
807 struct ceph_mds_session *s;
808
809 if (mds >= mdsc->mdsmap->possible_max_rank)
810 return ERR_PTR(-EINVAL);
811
812 s = kzalloc(sizeof(*s), GFP_NOFS);
813 if (!s)
814 return ERR_PTR(-ENOMEM);
815
816 if (mds >= mdsc->max_sessions) {
817 int newmax = 1 << get_count_order(mds + 1);
818 struct ceph_mds_session **sa;
819
820 dout("%s: realloc to %d\n", __func__, newmax);
821 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
822 if (!sa)
823 goto fail_realloc;
824 if (mdsc->sessions) {
825 memcpy(sa, mdsc->sessions,
826 mdsc->max_sessions * sizeof(void *));
827 kfree(mdsc->sessions);
828 }
829 mdsc->sessions = sa;
830 mdsc->max_sessions = newmax;
831 }
832
833 dout("%s: mds%d\n", __func__, mds);
834 s->s_mdsc = mdsc;
835 s->s_mds = mds;
836 s->s_state = CEPH_MDS_SESSION_NEW;
837 mutex_init(&s->s_mutex);
838
839 ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
840
841 atomic_set(&s->s_cap_gen, 1);
842 s->s_cap_ttl = jiffies - 1;
843
844 spin_lock_init(&s->s_cap_lock);
845 INIT_LIST_HEAD(&s->s_caps);
846 refcount_set(&s->s_ref, 1);
847 INIT_LIST_HEAD(&s->s_waiting);
848 INIT_LIST_HEAD(&s->s_unsafe);
849 xa_init(&s->s_delegated_inos);
850 INIT_LIST_HEAD(&s->s_cap_releases);
851 INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
852
853 INIT_LIST_HEAD(&s->s_cap_dirty);
854 INIT_LIST_HEAD(&s->s_cap_flushing);
855
856 mdsc->sessions[mds] = s;
857 atomic_inc(&mdsc->num_sessions);
858 refcount_inc(&s->s_ref); /* one ref to sessions[], one to caller */
859
860 ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
861 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
862
863 return s;
864
865 fail_realloc:
866 kfree(s);
867 return ERR_PTR(-ENOMEM);
868 }
869
870 /*
871 * called under mdsc->mutex
872 */
__unregister_session(struct ceph_mds_client * mdsc,struct ceph_mds_session * s)873 static void __unregister_session(struct ceph_mds_client *mdsc,
874 struct ceph_mds_session *s)
875 {
876 dout("__unregister_session mds%d %p\n", s->s_mds, s);
877 BUG_ON(mdsc->sessions[s->s_mds] != s);
878 mdsc->sessions[s->s_mds] = NULL;
879 ceph_con_close(&s->s_con);
880 ceph_put_mds_session(s);
881 atomic_dec(&mdsc->num_sessions);
882 }
883
884 /*
885 * drop session refs in request.
886 *
887 * should be last request ref, or hold mdsc->mutex
888 */
put_request_session(struct ceph_mds_request * req)889 static void put_request_session(struct ceph_mds_request *req)
890 {
891 if (req->r_session) {
892 ceph_put_mds_session(req->r_session);
893 req->r_session = NULL;
894 }
895 }
896
ceph_mdsc_iterate_sessions(struct ceph_mds_client * mdsc,void (* cb)(struct ceph_mds_session *),bool check_state)897 void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc,
898 void (*cb)(struct ceph_mds_session *),
899 bool check_state)
900 {
901 int mds;
902
903 mutex_lock(&mdsc->mutex);
904 for (mds = 0; mds < mdsc->max_sessions; ++mds) {
905 struct ceph_mds_session *s;
906
907 s = __ceph_lookup_mds_session(mdsc, mds);
908 if (!s)
909 continue;
910
911 if (check_state && !check_session_state(s)) {
912 ceph_put_mds_session(s);
913 continue;
914 }
915
916 mutex_unlock(&mdsc->mutex);
917 cb(s);
918 ceph_put_mds_session(s);
919 mutex_lock(&mdsc->mutex);
920 }
921 mutex_unlock(&mdsc->mutex);
922 }
923
ceph_mdsc_release_request(struct kref * kref)924 void ceph_mdsc_release_request(struct kref *kref)
925 {
926 struct ceph_mds_request *req = container_of(kref,
927 struct ceph_mds_request,
928 r_kref);
929 ceph_mdsc_release_dir_caps_no_check(req);
930 destroy_reply_info(&req->r_reply_info);
931 if (req->r_request)
932 ceph_msg_put(req->r_request);
933 if (req->r_reply)
934 ceph_msg_put(req->r_reply);
935 if (req->r_inode) {
936 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
937 iput(req->r_inode);
938 }
939 if (req->r_parent) {
940 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
941 iput(req->r_parent);
942 }
943 iput(req->r_target_inode);
944 if (req->r_dentry)
945 dput(req->r_dentry);
946 if (req->r_old_dentry)
947 dput(req->r_old_dentry);
948 if (req->r_old_dentry_dir) {
949 /*
950 * track (and drop pins for) r_old_dentry_dir
951 * separately, since r_old_dentry's d_parent may have
952 * changed between the dir mutex being dropped and
953 * this request being freed.
954 */
955 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
956 CEPH_CAP_PIN);
957 iput(req->r_old_dentry_dir);
958 }
959 kfree(req->r_path1);
960 kfree(req->r_path2);
961 put_cred(req->r_cred);
962 if (req->r_pagelist)
963 ceph_pagelist_release(req->r_pagelist);
964 put_request_session(req);
965 ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
966 WARN_ON_ONCE(!list_empty(&req->r_wait));
967 kmem_cache_free(ceph_mds_request_cachep, req);
968 }
969
DEFINE_RB_FUNCS(request,struct ceph_mds_request,r_tid,r_node)970 DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
971
972 /*
973 * lookup session, bump ref if found.
974 *
975 * called under mdsc->mutex.
976 */
977 static struct ceph_mds_request *
978 lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
979 {
980 struct ceph_mds_request *req;
981
982 req = lookup_request(&mdsc->request_tree, tid);
983 if (req)
984 ceph_mdsc_get_request(req);
985
986 return req;
987 }
988
989 /*
990 * Register an in-flight request, and assign a tid. Link to directory
991 * are modifying (if any).
992 *
993 * Called under mdsc->mutex.
994 */
__register_request(struct ceph_mds_client * mdsc,struct ceph_mds_request * req,struct inode * dir)995 static void __register_request(struct ceph_mds_client *mdsc,
996 struct ceph_mds_request *req,
997 struct inode *dir)
998 {
999 int ret = 0;
1000
1001 req->r_tid = ++mdsc->last_tid;
1002 if (req->r_num_caps) {
1003 ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
1004 req->r_num_caps);
1005 if (ret < 0) {
1006 pr_err("__register_request %p "
1007 "failed to reserve caps: %d\n", req, ret);
1008 /* set req->r_err to fail early from __do_request */
1009 req->r_err = ret;
1010 return;
1011 }
1012 }
1013 dout("__register_request %p tid %lld\n", req, req->r_tid);
1014 ceph_mdsc_get_request(req);
1015 insert_request(&mdsc->request_tree, req);
1016
1017 req->r_cred = get_current_cred();
1018
1019 if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
1020 mdsc->oldest_tid = req->r_tid;
1021
1022 if (dir) {
1023 struct ceph_inode_info *ci = ceph_inode(dir);
1024
1025 ihold(dir);
1026 req->r_unsafe_dir = dir;
1027 spin_lock(&ci->i_unsafe_lock);
1028 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
1029 spin_unlock(&ci->i_unsafe_lock);
1030 }
1031 }
1032
__unregister_request(struct ceph_mds_client * mdsc,struct ceph_mds_request * req)1033 static void __unregister_request(struct ceph_mds_client *mdsc,
1034 struct ceph_mds_request *req)
1035 {
1036 dout("__unregister_request %p tid %lld\n", req, req->r_tid);
1037
1038 /* Never leave an unregistered request on an unsafe list! */
1039 list_del_init(&req->r_unsafe_item);
1040
1041 if (req->r_tid == mdsc->oldest_tid) {
1042 struct rb_node *p = rb_next(&req->r_node);
1043 mdsc->oldest_tid = 0;
1044 while (p) {
1045 struct ceph_mds_request *next_req =
1046 rb_entry(p, struct ceph_mds_request, r_node);
1047 if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
1048 mdsc->oldest_tid = next_req->r_tid;
1049 break;
1050 }
1051 p = rb_next(p);
1052 }
1053 }
1054
1055 erase_request(&mdsc->request_tree, req);
1056
1057 if (req->r_unsafe_dir) {
1058 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
1059 spin_lock(&ci->i_unsafe_lock);
1060 list_del_init(&req->r_unsafe_dir_item);
1061 spin_unlock(&ci->i_unsafe_lock);
1062 }
1063 if (req->r_target_inode &&
1064 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
1065 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
1066 spin_lock(&ci->i_unsafe_lock);
1067 list_del_init(&req->r_unsafe_target_item);
1068 spin_unlock(&ci->i_unsafe_lock);
1069 }
1070
1071 if (req->r_unsafe_dir) {
1072 iput(req->r_unsafe_dir);
1073 req->r_unsafe_dir = NULL;
1074 }
1075
1076 complete_all(&req->r_safe_completion);
1077
1078 ceph_mdsc_put_request(req);
1079 }
1080
1081 /*
1082 * Walk back up the dentry tree until we hit a dentry representing a
1083 * non-snapshot inode. We do this using the rcu_read_lock (which must be held
1084 * when calling this) to ensure that the objects won't disappear while we're
1085 * working with them. Once we hit a candidate dentry, we attempt to take a
1086 * reference to it, and return that as the result.
1087 */
get_nonsnap_parent(struct dentry * dentry)1088 static struct inode *get_nonsnap_parent(struct dentry *dentry)
1089 {
1090 struct inode *inode = NULL;
1091
1092 while (dentry && !IS_ROOT(dentry)) {
1093 inode = d_inode_rcu(dentry);
1094 if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
1095 break;
1096 dentry = dentry->d_parent;
1097 }
1098 if (inode)
1099 inode = igrab(inode);
1100 return inode;
1101 }
1102
1103 /*
1104 * Choose mds to send request to next. If there is a hint set in the
1105 * request (e.g., due to a prior forward hint from the mds), use that.
1106 * Otherwise, consult frag tree and/or caps to identify the
1107 * appropriate mds. If all else fails, choose randomly.
1108 *
1109 * Called under mdsc->mutex.
1110 */
__choose_mds(struct ceph_mds_client * mdsc,struct ceph_mds_request * req,bool * random)1111 static int __choose_mds(struct ceph_mds_client *mdsc,
1112 struct ceph_mds_request *req,
1113 bool *random)
1114 {
1115 struct inode *inode;
1116 struct ceph_inode_info *ci;
1117 struct ceph_cap *cap;
1118 int mode = req->r_direct_mode;
1119 int mds = -1;
1120 u32 hash = req->r_direct_hash;
1121 bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
1122
1123 if (random)
1124 *random = false;
1125
1126 /*
1127 * is there a specific mds we should try? ignore hint if we have
1128 * no session and the mds is not up (active or recovering).
1129 */
1130 if (req->r_resend_mds >= 0 &&
1131 (__have_session(mdsc, req->r_resend_mds) ||
1132 ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
1133 dout("%s using resend_mds mds%d\n", __func__,
1134 req->r_resend_mds);
1135 return req->r_resend_mds;
1136 }
1137
1138 if (mode == USE_RANDOM_MDS)
1139 goto random;
1140
1141 inode = NULL;
1142 if (req->r_inode) {
1143 if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
1144 inode = req->r_inode;
1145 ihold(inode);
1146 } else {
1147 /* req->r_dentry is non-null for LSSNAP request */
1148 rcu_read_lock();
1149 inode = get_nonsnap_parent(req->r_dentry);
1150 rcu_read_unlock();
1151 dout("%s using snapdir's parent %p\n", __func__, inode);
1152 }
1153 } else if (req->r_dentry) {
1154 /* ignore race with rename; old or new d_parent is okay */
1155 struct dentry *parent;
1156 struct inode *dir;
1157
1158 rcu_read_lock();
1159 parent = READ_ONCE(req->r_dentry->d_parent);
1160 dir = req->r_parent ? : d_inode_rcu(parent);
1161
1162 if (!dir || dir->i_sb != mdsc->fsc->sb) {
1163 /* not this fs or parent went negative */
1164 inode = d_inode(req->r_dentry);
1165 if (inode)
1166 ihold(inode);
1167 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
1168 /* direct snapped/virtual snapdir requests
1169 * based on parent dir inode */
1170 inode = get_nonsnap_parent(parent);
1171 dout("%s using nonsnap parent %p\n", __func__, inode);
1172 } else {
1173 /* dentry target */
1174 inode = d_inode(req->r_dentry);
1175 if (!inode || mode == USE_AUTH_MDS) {
1176 /* dir + name */
1177 inode = igrab(dir);
1178 hash = ceph_dentry_hash(dir, req->r_dentry);
1179 is_hash = true;
1180 } else {
1181 ihold(inode);
1182 }
1183 }
1184 rcu_read_unlock();
1185 }
1186
1187 dout("%s %p is_hash=%d (0x%x) mode %d\n", __func__, inode, (int)is_hash,
1188 hash, mode);
1189 if (!inode)
1190 goto random;
1191 ci = ceph_inode(inode);
1192
1193 if (is_hash && S_ISDIR(inode->i_mode)) {
1194 struct ceph_inode_frag frag;
1195 int found;
1196
1197 ceph_choose_frag(ci, hash, &frag, &found);
1198 if (found) {
1199 if (mode == USE_ANY_MDS && frag.ndist > 0) {
1200 u8 r;
1201
1202 /* choose a random replica */
1203 get_random_bytes(&r, 1);
1204 r %= frag.ndist;
1205 mds = frag.dist[r];
1206 dout("%s %p %llx.%llx frag %u mds%d (%d/%d)\n",
1207 __func__, inode, ceph_vinop(inode),
1208 frag.frag, mds, (int)r, frag.ndist);
1209 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1210 CEPH_MDS_STATE_ACTIVE &&
1211 !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))
1212 goto out;
1213 }
1214
1215 /* since this file/dir wasn't known to be
1216 * replicated, then we want to look for the
1217 * authoritative mds. */
1218 if (frag.mds >= 0) {
1219 /* choose auth mds */
1220 mds = frag.mds;
1221 dout("%s %p %llx.%llx frag %u mds%d (auth)\n",
1222 __func__, inode, ceph_vinop(inode),
1223 frag.frag, mds);
1224 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1225 CEPH_MDS_STATE_ACTIVE) {
1226 if (!ceph_mdsmap_is_laggy(mdsc->mdsmap,
1227 mds))
1228 goto out;
1229 }
1230 }
1231 mode = USE_AUTH_MDS;
1232 }
1233 }
1234
1235 spin_lock(&ci->i_ceph_lock);
1236 cap = NULL;
1237 if (mode == USE_AUTH_MDS)
1238 cap = ci->i_auth_cap;
1239 if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1240 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1241 if (!cap) {
1242 spin_unlock(&ci->i_ceph_lock);
1243 iput(inode);
1244 goto random;
1245 }
1246 mds = cap->session->s_mds;
1247 dout("%s %p %llx.%llx mds%d (%scap %p)\n", __func__,
1248 inode, ceph_vinop(inode), mds,
1249 cap == ci->i_auth_cap ? "auth " : "", cap);
1250 spin_unlock(&ci->i_ceph_lock);
1251 out:
1252 iput(inode);
1253 return mds;
1254
1255 random:
1256 if (random)
1257 *random = true;
1258
1259 mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1260 dout("%s chose random mds%d\n", __func__, mds);
1261 return mds;
1262 }
1263
1264
1265 /*
1266 * session messages
1267 */
ceph_create_session_msg(u32 op,u64 seq)1268 struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq)
1269 {
1270 struct ceph_msg *msg;
1271 struct ceph_mds_session_head *h;
1272
1273 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1274 false);
1275 if (!msg) {
1276 pr_err("ENOMEM creating session %s msg\n",
1277 ceph_session_op_name(op));
1278 return NULL;
1279 }
1280 h = msg->front.iov_base;
1281 h->op = cpu_to_le32(op);
1282 h->seq = cpu_to_le64(seq);
1283
1284 return msg;
1285 }
1286
1287 static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1288 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
encode_supported_features(void ** p,void * end)1289 static int encode_supported_features(void **p, void *end)
1290 {
1291 static const size_t count = ARRAY_SIZE(feature_bits);
1292
1293 if (count > 0) {
1294 size_t i;
1295 size_t size = FEATURE_BYTES(count);
1296 unsigned long bit;
1297
1298 if (WARN_ON_ONCE(*p + 4 + size > end))
1299 return -ERANGE;
1300
1301 ceph_encode_32(p, size);
1302 memset(*p, 0, size);
1303 for (i = 0; i < count; i++) {
1304 bit = feature_bits[i];
1305 ((unsigned char *)(*p))[bit / 8] |= BIT(bit % 8);
1306 }
1307 *p += size;
1308 } else {
1309 if (WARN_ON_ONCE(*p + 4 > end))
1310 return -ERANGE;
1311
1312 ceph_encode_32(p, 0);
1313 }
1314
1315 return 0;
1316 }
1317
1318 static const unsigned char metric_bits[] = CEPHFS_METRIC_SPEC_CLIENT_SUPPORTED;
1319 #define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
encode_metric_spec(void ** p,void * end)1320 static int encode_metric_spec(void **p, void *end)
1321 {
1322 static const size_t count = ARRAY_SIZE(metric_bits);
1323
1324 /* header */
1325 if (WARN_ON_ONCE(*p + 2 > end))
1326 return -ERANGE;
1327
1328 ceph_encode_8(p, 1); /* version */
1329 ceph_encode_8(p, 1); /* compat */
1330
1331 if (count > 0) {
1332 size_t i;
1333 size_t size = METRIC_BYTES(count);
1334
1335 if (WARN_ON_ONCE(*p + 4 + 4 + size > end))
1336 return -ERANGE;
1337
1338 /* metric spec info length */
1339 ceph_encode_32(p, 4 + size);
1340
1341 /* metric spec */
1342 ceph_encode_32(p, size);
1343 memset(*p, 0, size);
1344 for (i = 0; i < count; i++)
1345 ((unsigned char *)(*p))[i / 8] |= BIT(metric_bits[i] % 8);
1346 *p += size;
1347 } else {
1348 if (WARN_ON_ONCE(*p + 4 + 4 > end))
1349 return -ERANGE;
1350
1351 /* metric spec info length */
1352 ceph_encode_32(p, 4);
1353 /* metric spec */
1354 ceph_encode_32(p, 0);
1355 }
1356
1357 return 0;
1358 }
1359
1360 /*
1361 * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1362 * to include additional client metadata fields.
1363 */
create_session_open_msg(struct ceph_mds_client * mdsc,u64 seq)1364 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1365 {
1366 struct ceph_msg *msg;
1367 struct ceph_mds_session_head *h;
1368 int i;
1369 int extra_bytes = 0;
1370 int metadata_key_count = 0;
1371 struct ceph_options *opt = mdsc->fsc->client->options;
1372 struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1373 size_t size, count;
1374 void *p, *end;
1375 int ret;
1376
1377 const char* metadata[][2] = {
1378 {"hostname", mdsc->nodename},
1379 {"kernel_version", init_utsname()->release},
1380 {"entity_id", opt->name ? : ""},
1381 {"root", fsopt->server_path ? : "/"},
1382 {NULL, NULL}
1383 };
1384
1385 /* Calculate serialized length of metadata */
1386 extra_bytes = 4; /* map length */
1387 for (i = 0; metadata[i][0]; ++i) {
1388 extra_bytes += 8 + strlen(metadata[i][0]) +
1389 strlen(metadata[i][1]);
1390 metadata_key_count++;
1391 }
1392
1393 /* supported feature */
1394 size = 0;
1395 count = ARRAY_SIZE(feature_bits);
1396 if (count > 0)
1397 size = FEATURE_BYTES(count);
1398 extra_bytes += 4 + size;
1399
1400 /* metric spec */
1401 size = 0;
1402 count = ARRAY_SIZE(metric_bits);
1403 if (count > 0)
1404 size = METRIC_BYTES(count);
1405 extra_bytes += 2 + 4 + 4 + size;
1406
1407 /* Allocate the message */
1408 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1409 GFP_NOFS, false);
1410 if (!msg) {
1411 pr_err("ENOMEM creating session open msg\n");
1412 return ERR_PTR(-ENOMEM);
1413 }
1414 p = msg->front.iov_base;
1415 end = p + msg->front.iov_len;
1416
1417 h = p;
1418 h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1419 h->seq = cpu_to_le64(seq);
1420
1421 /*
1422 * Serialize client metadata into waiting buffer space, using
1423 * the format that userspace expects for map<string, string>
1424 *
1425 * ClientSession messages with metadata are v4
1426 */
1427 msg->hdr.version = cpu_to_le16(4);
1428 msg->hdr.compat_version = cpu_to_le16(1);
1429
1430 /* The write pointer, following the session_head structure */
1431 p += sizeof(*h);
1432
1433 /* Number of entries in the map */
1434 ceph_encode_32(&p, metadata_key_count);
1435
1436 /* Two length-prefixed strings for each entry in the map */
1437 for (i = 0; metadata[i][0]; ++i) {
1438 size_t const key_len = strlen(metadata[i][0]);
1439 size_t const val_len = strlen(metadata[i][1]);
1440
1441 ceph_encode_32(&p, key_len);
1442 memcpy(p, metadata[i][0], key_len);
1443 p += key_len;
1444 ceph_encode_32(&p, val_len);
1445 memcpy(p, metadata[i][1], val_len);
1446 p += val_len;
1447 }
1448
1449 ret = encode_supported_features(&p, end);
1450 if (ret) {
1451 pr_err("encode_supported_features failed!\n");
1452 ceph_msg_put(msg);
1453 return ERR_PTR(ret);
1454 }
1455
1456 ret = encode_metric_spec(&p, end);
1457 if (ret) {
1458 pr_err("encode_metric_spec failed!\n");
1459 ceph_msg_put(msg);
1460 return ERR_PTR(ret);
1461 }
1462
1463 msg->front.iov_len = p - msg->front.iov_base;
1464 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1465
1466 return msg;
1467 }
1468
1469 /*
1470 * send session open request.
1471 *
1472 * called under mdsc->mutex
1473 */
__open_session(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)1474 static int __open_session(struct ceph_mds_client *mdsc,
1475 struct ceph_mds_session *session)
1476 {
1477 struct ceph_msg *msg;
1478 int mstate;
1479 int mds = session->s_mds;
1480
1481 /* wait for mds to go active? */
1482 mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1483 dout("open_session to mds%d (%s)\n", mds,
1484 ceph_mds_state_name(mstate));
1485 session->s_state = CEPH_MDS_SESSION_OPENING;
1486 session->s_renew_requested = jiffies;
1487
1488 /* send connect message */
1489 msg = create_session_open_msg(mdsc, session->s_seq);
1490 if (IS_ERR(msg))
1491 return PTR_ERR(msg);
1492 ceph_con_send(&session->s_con, msg);
1493 return 0;
1494 }
1495
1496 /*
1497 * open sessions for any export targets for the given mds
1498 *
1499 * called under mdsc->mutex
1500 */
1501 static struct ceph_mds_session *
__open_export_target_session(struct ceph_mds_client * mdsc,int target)1502 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1503 {
1504 struct ceph_mds_session *session;
1505 int ret;
1506
1507 session = __ceph_lookup_mds_session(mdsc, target);
1508 if (!session) {
1509 session = register_session(mdsc, target);
1510 if (IS_ERR(session))
1511 return session;
1512 }
1513 if (session->s_state == CEPH_MDS_SESSION_NEW ||
1514 session->s_state == CEPH_MDS_SESSION_CLOSING) {
1515 ret = __open_session(mdsc, session);
1516 if (ret)
1517 return ERR_PTR(ret);
1518 }
1519
1520 return session;
1521 }
1522
1523 struct ceph_mds_session *
ceph_mdsc_open_export_target_session(struct ceph_mds_client * mdsc,int target)1524 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1525 {
1526 struct ceph_mds_session *session;
1527
1528 dout("open_export_target_session to mds%d\n", target);
1529
1530 mutex_lock(&mdsc->mutex);
1531 session = __open_export_target_session(mdsc, target);
1532 mutex_unlock(&mdsc->mutex);
1533
1534 return session;
1535 }
1536
__open_export_target_sessions(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)1537 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1538 struct ceph_mds_session *session)
1539 {
1540 struct ceph_mds_info *mi;
1541 struct ceph_mds_session *ts;
1542 int i, mds = session->s_mds;
1543
1544 if (mds >= mdsc->mdsmap->possible_max_rank)
1545 return;
1546
1547 mi = &mdsc->mdsmap->m_info[mds];
1548 dout("open_export_target_sessions for mds%d (%d targets)\n",
1549 session->s_mds, mi->num_export_targets);
1550
1551 for (i = 0; i < mi->num_export_targets; i++) {
1552 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1553 ceph_put_mds_session(ts);
1554 }
1555 }
1556
ceph_mdsc_open_export_target_sessions(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)1557 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1558 struct ceph_mds_session *session)
1559 {
1560 mutex_lock(&mdsc->mutex);
1561 __open_export_target_sessions(mdsc, session);
1562 mutex_unlock(&mdsc->mutex);
1563 }
1564
1565 /*
1566 * session caps
1567 */
1568
detach_cap_releases(struct ceph_mds_session * session,struct list_head * target)1569 static void detach_cap_releases(struct ceph_mds_session *session,
1570 struct list_head *target)
1571 {
1572 lockdep_assert_held(&session->s_cap_lock);
1573
1574 list_splice_init(&session->s_cap_releases, target);
1575 session->s_num_cap_releases = 0;
1576 dout("dispose_cap_releases mds%d\n", session->s_mds);
1577 }
1578
dispose_cap_releases(struct ceph_mds_client * mdsc,struct list_head * dispose)1579 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1580 struct list_head *dispose)
1581 {
1582 while (!list_empty(dispose)) {
1583 struct ceph_cap *cap;
1584 /* zero out the in-progress message */
1585 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1586 list_del(&cap->session_caps);
1587 ceph_put_cap(mdsc, cap);
1588 }
1589 }
1590
cleanup_session_requests(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)1591 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1592 struct ceph_mds_session *session)
1593 {
1594 struct ceph_mds_request *req;
1595 struct rb_node *p;
1596
1597 dout("cleanup_session_requests mds%d\n", session->s_mds);
1598 mutex_lock(&mdsc->mutex);
1599 while (!list_empty(&session->s_unsafe)) {
1600 req = list_first_entry(&session->s_unsafe,
1601 struct ceph_mds_request, r_unsafe_item);
1602 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1603 req->r_tid);
1604 if (req->r_target_inode)
1605 mapping_set_error(req->r_target_inode->i_mapping, -EIO);
1606 if (req->r_unsafe_dir)
1607 mapping_set_error(req->r_unsafe_dir->i_mapping, -EIO);
1608 __unregister_request(mdsc, req);
1609 }
1610 /* zero r_attempts, so kick_requests() will re-send requests */
1611 p = rb_first(&mdsc->request_tree);
1612 while (p) {
1613 req = rb_entry(p, struct ceph_mds_request, r_node);
1614 p = rb_next(p);
1615 if (req->r_session &&
1616 req->r_session->s_mds == session->s_mds)
1617 req->r_attempts = 0;
1618 }
1619 mutex_unlock(&mdsc->mutex);
1620 }
1621
1622 /*
1623 * Helper to safely iterate over all caps associated with a session, with
1624 * special care taken to handle a racing __ceph_remove_cap().
1625 *
1626 * Caller must hold session s_mutex.
1627 */
ceph_iterate_session_caps(struct ceph_mds_session * session,int (* cb)(struct inode *,struct ceph_cap *,void *),void * arg)1628 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1629 int (*cb)(struct inode *, struct ceph_cap *,
1630 void *), void *arg)
1631 {
1632 struct list_head *p;
1633 struct ceph_cap *cap;
1634 struct inode *inode, *last_inode = NULL;
1635 struct ceph_cap *old_cap = NULL;
1636 int ret;
1637
1638 dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1639 spin_lock(&session->s_cap_lock);
1640 p = session->s_caps.next;
1641 while (p != &session->s_caps) {
1642 cap = list_entry(p, struct ceph_cap, session_caps);
1643 inode = igrab(&cap->ci->netfs.inode);
1644 if (!inode) {
1645 p = p->next;
1646 continue;
1647 }
1648 session->s_cap_iterator = cap;
1649 spin_unlock(&session->s_cap_lock);
1650
1651 if (last_inode) {
1652 iput(last_inode);
1653 last_inode = NULL;
1654 }
1655 if (old_cap) {
1656 ceph_put_cap(session->s_mdsc, old_cap);
1657 old_cap = NULL;
1658 }
1659
1660 ret = cb(inode, cap, arg);
1661 last_inode = inode;
1662
1663 spin_lock(&session->s_cap_lock);
1664 p = p->next;
1665 if (!cap->ci) {
1666 dout("iterate_session_caps finishing cap %p removal\n",
1667 cap);
1668 BUG_ON(cap->session != session);
1669 cap->session = NULL;
1670 list_del_init(&cap->session_caps);
1671 session->s_nr_caps--;
1672 atomic64_dec(&session->s_mdsc->metric.total_caps);
1673 if (cap->queue_release)
1674 __ceph_queue_cap_release(session, cap);
1675 else
1676 old_cap = cap; /* put_cap it w/o locks held */
1677 }
1678 if (ret < 0)
1679 goto out;
1680 }
1681 ret = 0;
1682 out:
1683 session->s_cap_iterator = NULL;
1684 spin_unlock(&session->s_cap_lock);
1685
1686 iput(last_inode);
1687 if (old_cap)
1688 ceph_put_cap(session->s_mdsc, old_cap);
1689
1690 return ret;
1691 }
1692
remove_session_caps_cb(struct inode * inode,struct ceph_cap * cap,void * arg)1693 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1694 void *arg)
1695 {
1696 struct ceph_inode_info *ci = ceph_inode(inode);
1697 bool invalidate = false;
1698 int iputs;
1699
1700 dout("removing cap %p, ci is %p, inode is %p\n",
1701 cap, ci, &ci->netfs.inode);
1702 spin_lock(&ci->i_ceph_lock);
1703 iputs = ceph_purge_inode_cap(inode, cap, &invalidate);
1704 spin_unlock(&ci->i_ceph_lock);
1705
1706 wake_up_all(&ci->i_cap_wq);
1707 if (invalidate)
1708 ceph_queue_invalidate(inode);
1709 while (iputs--)
1710 iput(inode);
1711 return 0;
1712 }
1713
1714 /*
1715 * caller must hold session s_mutex
1716 */
remove_session_caps(struct ceph_mds_session * session)1717 static void remove_session_caps(struct ceph_mds_session *session)
1718 {
1719 struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1720 struct super_block *sb = fsc->sb;
1721 LIST_HEAD(dispose);
1722
1723 dout("remove_session_caps on %p\n", session);
1724 ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1725
1726 wake_up_all(&fsc->mdsc->cap_flushing_wq);
1727
1728 spin_lock(&session->s_cap_lock);
1729 if (session->s_nr_caps > 0) {
1730 struct inode *inode;
1731 struct ceph_cap *cap, *prev = NULL;
1732 struct ceph_vino vino;
1733 /*
1734 * iterate_session_caps() skips inodes that are being
1735 * deleted, we need to wait until deletions are complete.
1736 * __wait_on_freeing_inode() is designed for the job,
1737 * but it is not exported, so use lookup inode function
1738 * to access it.
1739 */
1740 while (!list_empty(&session->s_caps)) {
1741 cap = list_entry(session->s_caps.next,
1742 struct ceph_cap, session_caps);
1743 if (cap == prev)
1744 break;
1745 prev = cap;
1746 vino = cap->ci->i_vino;
1747 spin_unlock(&session->s_cap_lock);
1748
1749 inode = ceph_find_inode(sb, vino);
1750 iput(inode);
1751
1752 spin_lock(&session->s_cap_lock);
1753 }
1754 }
1755
1756 // drop cap expires and unlock s_cap_lock
1757 detach_cap_releases(session, &dispose);
1758
1759 BUG_ON(session->s_nr_caps > 0);
1760 BUG_ON(!list_empty(&session->s_cap_flushing));
1761 spin_unlock(&session->s_cap_lock);
1762 dispose_cap_releases(session->s_mdsc, &dispose);
1763 }
1764
1765 enum {
1766 RECONNECT,
1767 RENEWCAPS,
1768 FORCE_RO,
1769 };
1770
1771 /*
1772 * wake up any threads waiting on this session's caps. if the cap is
1773 * old (didn't get renewed on the client reconnect), remove it now.
1774 *
1775 * caller must hold s_mutex.
1776 */
wake_up_session_cb(struct inode * inode,struct ceph_cap * cap,void * arg)1777 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1778 void *arg)
1779 {
1780 struct ceph_inode_info *ci = ceph_inode(inode);
1781 unsigned long ev = (unsigned long)arg;
1782
1783 if (ev == RECONNECT) {
1784 spin_lock(&ci->i_ceph_lock);
1785 ci->i_wanted_max_size = 0;
1786 ci->i_requested_max_size = 0;
1787 spin_unlock(&ci->i_ceph_lock);
1788 } else if (ev == RENEWCAPS) {
1789 if (cap->cap_gen < atomic_read(&cap->session->s_cap_gen)) {
1790 /* mds did not re-issue stale cap */
1791 spin_lock(&ci->i_ceph_lock);
1792 cap->issued = cap->implemented = CEPH_CAP_PIN;
1793 spin_unlock(&ci->i_ceph_lock);
1794 }
1795 } else if (ev == FORCE_RO) {
1796 }
1797 wake_up_all(&ci->i_cap_wq);
1798 return 0;
1799 }
1800
wake_up_session_caps(struct ceph_mds_session * session,int ev)1801 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1802 {
1803 dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1804 ceph_iterate_session_caps(session, wake_up_session_cb,
1805 (void *)(unsigned long)ev);
1806 }
1807
1808 /*
1809 * Send periodic message to MDS renewing all currently held caps. The
1810 * ack will reset the expiration for all caps from this session.
1811 *
1812 * caller holds s_mutex
1813 */
send_renew_caps(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)1814 static int send_renew_caps(struct ceph_mds_client *mdsc,
1815 struct ceph_mds_session *session)
1816 {
1817 struct ceph_msg *msg;
1818 int state;
1819
1820 if (time_after_eq(jiffies, session->s_cap_ttl) &&
1821 time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1822 pr_info("mds%d caps stale\n", session->s_mds);
1823 session->s_renew_requested = jiffies;
1824
1825 /* do not try to renew caps until a recovering mds has reconnected
1826 * with its clients. */
1827 state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1828 if (state < CEPH_MDS_STATE_RECONNECT) {
1829 dout("send_renew_caps ignoring mds%d (%s)\n",
1830 session->s_mds, ceph_mds_state_name(state));
1831 return 0;
1832 }
1833
1834 dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1835 ceph_mds_state_name(state));
1836 msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1837 ++session->s_renew_seq);
1838 if (!msg)
1839 return -ENOMEM;
1840 ceph_con_send(&session->s_con, msg);
1841 return 0;
1842 }
1843
send_flushmsg_ack(struct ceph_mds_client * mdsc,struct ceph_mds_session * session,u64 seq)1844 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1845 struct ceph_mds_session *session, u64 seq)
1846 {
1847 struct ceph_msg *msg;
1848
1849 dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1850 session->s_mds, ceph_session_state_name(session->s_state), seq);
1851 msg = ceph_create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1852 if (!msg)
1853 return -ENOMEM;
1854 ceph_con_send(&session->s_con, msg);
1855 return 0;
1856 }
1857
1858
1859 /*
1860 * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1861 *
1862 * Called under session->s_mutex
1863 */
renewed_caps(struct ceph_mds_client * mdsc,struct ceph_mds_session * session,int is_renew)1864 static void renewed_caps(struct ceph_mds_client *mdsc,
1865 struct ceph_mds_session *session, int is_renew)
1866 {
1867 int was_stale;
1868 int wake = 0;
1869
1870 spin_lock(&session->s_cap_lock);
1871 was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1872
1873 session->s_cap_ttl = session->s_renew_requested +
1874 mdsc->mdsmap->m_session_timeout*HZ;
1875
1876 if (was_stale) {
1877 if (time_before(jiffies, session->s_cap_ttl)) {
1878 pr_info("mds%d caps renewed\n", session->s_mds);
1879 wake = 1;
1880 } else {
1881 pr_info("mds%d caps still stale\n", session->s_mds);
1882 }
1883 }
1884 dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1885 session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1886 time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1887 spin_unlock(&session->s_cap_lock);
1888
1889 if (wake)
1890 wake_up_session_caps(session, RENEWCAPS);
1891 }
1892
1893 /*
1894 * send a session close request
1895 */
request_close_session(struct ceph_mds_session * session)1896 static int request_close_session(struct ceph_mds_session *session)
1897 {
1898 struct ceph_msg *msg;
1899
1900 dout("request_close_session mds%d state %s seq %lld\n",
1901 session->s_mds, ceph_session_state_name(session->s_state),
1902 session->s_seq);
1903 msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_CLOSE,
1904 session->s_seq);
1905 if (!msg)
1906 return -ENOMEM;
1907 ceph_con_send(&session->s_con, msg);
1908 return 1;
1909 }
1910
1911 /*
1912 * Called with s_mutex held.
1913 */
__close_session(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)1914 static int __close_session(struct ceph_mds_client *mdsc,
1915 struct ceph_mds_session *session)
1916 {
1917 if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1918 return 0;
1919 session->s_state = CEPH_MDS_SESSION_CLOSING;
1920 return request_close_session(session);
1921 }
1922
drop_negative_children(struct dentry * dentry)1923 static bool drop_negative_children(struct dentry *dentry)
1924 {
1925 struct dentry *child;
1926 bool all_negative = true;
1927
1928 if (!d_is_dir(dentry))
1929 goto out;
1930
1931 spin_lock(&dentry->d_lock);
1932 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1933 if (d_really_is_positive(child)) {
1934 all_negative = false;
1935 break;
1936 }
1937 }
1938 spin_unlock(&dentry->d_lock);
1939
1940 if (all_negative)
1941 shrink_dcache_parent(dentry);
1942 out:
1943 return all_negative;
1944 }
1945
1946 /*
1947 * Trim old(er) caps.
1948 *
1949 * Because we can't cache an inode without one or more caps, we do
1950 * this indirectly: if a cap is unused, we prune its aliases, at which
1951 * point the inode will hopefully get dropped to.
1952 *
1953 * Yes, this is a bit sloppy. Our only real goal here is to respond to
1954 * memory pressure from the MDS, though, so it needn't be perfect.
1955 */
trim_caps_cb(struct inode * inode,struct ceph_cap * cap,void * arg)1956 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1957 {
1958 int *remaining = arg;
1959 struct ceph_inode_info *ci = ceph_inode(inode);
1960 int used, wanted, oissued, mine;
1961
1962 if (*remaining <= 0)
1963 return -1;
1964
1965 spin_lock(&ci->i_ceph_lock);
1966 mine = cap->issued | cap->implemented;
1967 used = __ceph_caps_used(ci);
1968 wanted = __ceph_caps_file_wanted(ci);
1969 oissued = __ceph_caps_issued_other(ci, cap);
1970
1971 dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1972 inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1973 ceph_cap_string(used), ceph_cap_string(wanted));
1974 if (cap == ci->i_auth_cap) {
1975 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1976 !list_empty(&ci->i_cap_snaps))
1977 goto out;
1978 if ((used | wanted) & CEPH_CAP_ANY_WR)
1979 goto out;
1980 /* Note: it's possible that i_filelock_ref becomes non-zero
1981 * after dropping auth caps. It doesn't hurt because reply
1982 * of lock mds request will re-add auth caps. */
1983 if (atomic_read(&ci->i_filelock_ref) > 0)
1984 goto out;
1985 }
1986 /* The inode has cached pages, but it's no longer used.
1987 * we can safely drop it */
1988 if (S_ISREG(inode->i_mode) &&
1989 wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
1990 !(oissued & CEPH_CAP_FILE_CACHE)) {
1991 used = 0;
1992 oissued = 0;
1993 }
1994 if ((used | wanted) & ~oissued & mine)
1995 goto out; /* we need these caps */
1996
1997 if (oissued) {
1998 /* we aren't the only cap.. just remove us */
1999 ceph_remove_cap(cap, true);
2000 (*remaining)--;
2001 } else {
2002 struct dentry *dentry;
2003 /* try dropping referring dentries */
2004 spin_unlock(&ci->i_ceph_lock);
2005 dentry = d_find_any_alias(inode);
2006 if (dentry && drop_negative_children(dentry)) {
2007 int count;
2008 dput(dentry);
2009 d_prune_aliases(inode);
2010 count = atomic_read(&inode->i_count);
2011 if (count == 1)
2012 (*remaining)--;
2013 dout("trim_caps_cb %p cap %p pruned, count now %d\n",
2014 inode, cap, count);
2015 } else {
2016 dput(dentry);
2017 }
2018 return 0;
2019 }
2020
2021 out:
2022 spin_unlock(&ci->i_ceph_lock);
2023 return 0;
2024 }
2025
2026 /*
2027 * Trim session cap count down to some max number.
2028 */
ceph_trim_caps(struct ceph_mds_client * mdsc,struct ceph_mds_session * session,int max_caps)2029 int ceph_trim_caps(struct ceph_mds_client *mdsc,
2030 struct ceph_mds_session *session,
2031 int max_caps)
2032 {
2033 int trim_caps = session->s_nr_caps - max_caps;
2034
2035 dout("trim_caps mds%d start: %d / %d, trim %d\n",
2036 session->s_mds, session->s_nr_caps, max_caps, trim_caps);
2037 if (trim_caps > 0) {
2038 int remaining = trim_caps;
2039
2040 ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
2041 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
2042 session->s_mds, session->s_nr_caps, max_caps,
2043 trim_caps - remaining);
2044 }
2045
2046 ceph_flush_cap_releases(mdsc, session);
2047 return 0;
2048 }
2049
check_caps_flush(struct ceph_mds_client * mdsc,u64 want_flush_tid)2050 static int check_caps_flush(struct ceph_mds_client *mdsc,
2051 u64 want_flush_tid)
2052 {
2053 int ret = 1;
2054
2055 spin_lock(&mdsc->cap_dirty_lock);
2056 if (!list_empty(&mdsc->cap_flush_list)) {
2057 struct ceph_cap_flush *cf =
2058 list_first_entry(&mdsc->cap_flush_list,
2059 struct ceph_cap_flush, g_list);
2060 if (cf->tid <= want_flush_tid) {
2061 dout("check_caps_flush still flushing tid "
2062 "%llu <= %llu\n", cf->tid, want_flush_tid);
2063 ret = 0;
2064 }
2065 }
2066 spin_unlock(&mdsc->cap_dirty_lock);
2067 return ret;
2068 }
2069
2070 /*
2071 * flush all dirty inode data to disk.
2072 *
2073 * returns true if we've flushed through want_flush_tid
2074 */
wait_caps_flush(struct ceph_mds_client * mdsc,u64 want_flush_tid)2075 static void wait_caps_flush(struct ceph_mds_client *mdsc,
2076 u64 want_flush_tid)
2077 {
2078 dout("check_caps_flush want %llu\n", want_flush_tid);
2079
2080 wait_event(mdsc->cap_flushing_wq,
2081 check_caps_flush(mdsc, want_flush_tid));
2082
2083 dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
2084 }
2085
2086 /*
2087 * called under s_mutex
2088 */
ceph_send_cap_releases(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)2089 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
2090 struct ceph_mds_session *session)
2091 {
2092 struct ceph_msg *msg = NULL;
2093 struct ceph_mds_cap_release *head;
2094 struct ceph_mds_cap_item *item;
2095 struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
2096 struct ceph_cap *cap;
2097 LIST_HEAD(tmp_list);
2098 int num_cap_releases;
2099 __le32 barrier, *cap_barrier;
2100
2101 down_read(&osdc->lock);
2102 barrier = cpu_to_le32(osdc->epoch_barrier);
2103 up_read(&osdc->lock);
2104
2105 spin_lock(&session->s_cap_lock);
2106 again:
2107 list_splice_init(&session->s_cap_releases, &tmp_list);
2108 num_cap_releases = session->s_num_cap_releases;
2109 session->s_num_cap_releases = 0;
2110 spin_unlock(&session->s_cap_lock);
2111
2112 while (!list_empty(&tmp_list)) {
2113 if (!msg) {
2114 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
2115 PAGE_SIZE, GFP_NOFS, false);
2116 if (!msg)
2117 goto out_err;
2118 head = msg->front.iov_base;
2119 head->num = cpu_to_le32(0);
2120 msg->front.iov_len = sizeof(*head);
2121
2122 msg->hdr.version = cpu_to_le16(2);
2123 msg->hdr.compat_version = cpu_to_le16(1);
2124 }
2125
2126 cap = list_first_entry(&tmp_list, struct ceph_cap,
2127 session_caps);
2128 list_del(&cap->session_caps);
2129 num_cap_releases--;
2130
2131 head = msg->front.iov_base;
2132 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2133 &head->num);
2134 item = msg->front.iov_base + msg->front.iov_len;
2135 item->ino = cpu_to_le64(cap->cap_ino);
2136 item->cap_id = cpu_to_le64(cap->cap_id);
2137 item->migrate_seq = cpu_to_le32(cap->mseq);
2138 item->seq = cpu_to_le32(cap->issue_seq);
2139 msg->front.iov_len += sizeof(*item);
2140
2141 ceph_put_cap(mdsc, cap);
2142
2143 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
2144 // Append cap_barrier field
2145 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2146 *cap_barrier = barrier;
2147 msg->front.iov_len += sizeof(*cap_barrier);
2148
2149 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2150 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2151 ceph_con_send(&session->s_con, msg);
2152 msg = NULL;
2153 }
2154 }
2155
2156 BUG_ON(num_cap_releases != 0);
2157
2158 spin_lock(&session->s_cap_lock);
2159 if (!list_empty(&session->s_cap_releases))
2160 goto again;
2161 spin_unlock(&session->s_cap_lock);
2162
2163 if (msg) {
2164 // Append cap_barrier field
2165 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2166 *cap_barrier = barrier;
2167 msg->front.iov_len += sizeof(*cap_barrier);
2168
2169 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2170 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2171 ceph_con_send(&session->s_con, msg);
2172 }
2173 return;
2174 out_err:
2175 pr_err("send_cap_releases mds%d, failed to allocate message\n",
2176 session->s_mds);
2177 spin_lock(&session->s_cap_lock);
2178 list_splice(&tmp_list, &session->s_cap_releases);
2179 session->s_num_cap_releases += num_cap_releases;
2180 spin_unlock(&session->s_cap_lock);
2181 }
2182
ceph_cap_release_work(struct work_struct * work)2183 static void ceph_cap_release_work(struct work_struct *work)
2184 {
2185 struct ceph_mds_session *session =
2186 container_of(work, struct ceph_mds_session, s_cap_release_work);
2187
2188 mutex_lock(&session->s_mutex);
2189 if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2190 session->s_state == CEPH_MDS_SESSION_HUNG)
2191 ceph_send_cap_releases(session->s_mdsc, session);
2192 mutex_unlock(&session->s_mutex);
2193 ceph_put_mds_session(session);
2194 }
2195
ceph_flush_cap_releases(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)2196 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2197 struct ceph_mds_session *session)
2198 {
2199 if (mdsc->stopping)
2200 return;
2201
2202 ceph_get_mds_session(session);
2203 if (queue_work(mdsc->fsc->cap_wq,
2204 &session->s_cap_release_work)) {
2205 dout("cap release work queued\n");
2206 } else {
2207 ceph_put_mds_session(session);
2208 dout("failed to queue cap release work\n");
2209 }
2210 }
2211
2212 /*
2213 * caller holds session->s_cap_lock
2214 */
__ceph_queue_cap_release(struct ceph_mds_session * session,struct ceph_cap * cap)2215 void __ceph_queue_cap_release(struct ceph_mds_session *session,
2216 struct ceph_cap *cap)
2217 {
2218 list_add_tail(&cap->session_caps, &session->s_cap_releases);
2219 session->s_num_cap_releases++;
2220
2221 if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2222 ceph_flush_cap_releases(session->s_mdsc, session);
2223 }
2224
ceph_cap_reclaim_work(struct work_struct * work)2225 static void ceph_cap_reclaim_work(struct work_struct *work)
2226 {
2227 struct ceph_mds_client *mdsc =
2228 container_of(work, struct ceph_mds_client, cap_reclaim_work);
2229 int ret = ceph_trim_dentries(mdsc);
2230 if (ret == -EAGAIN)
2231 ceph_queue_cap_reclaim_work(mdsc);
2232 }
2233
ceph_queue_cap_reclaim_work(struct ceph_mds_client * mdsc)2234 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2235 {
2236 if (mdsc->stopping)
2237 return;
2238
2239 if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2240 dout("caps reclaim work queued\n");
2241 } else {
2242 dout("failed to queue caps release work\n");
2243 }
2244 }
2245
ceph_reclaim_caps_nr(struct ceph_mds_client * mdsc,int nr)2246 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2247 {
2248 int val;
2249 if (!nr)
2250 return;
2251 val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2252 if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
2253 atomic_set(&mdsc->cap_reclaim_pending, 0);
2254 ceph_queue_cap_reclaim_work(mdsc);
2255 }
2256 }
2257
2258 /*
2259 * requests
2260 */
2261
ceph_alloc_readdir_reply_buffer(struct ceph_mds_request * req,struct inode * dir)2262 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2263 struct inode *dir)
2264 {
2265 struct ceph_inode_info *ci = ceph_inode(dir);
2266 struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2267 struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2268 size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2269 unsigned int num_entries;
2270 int order;
2271
2272 spin_lock(&ci->i_ceph_lock);
2273 num_entries = ci->i_files + ci->i_subdirs;
2274 spin_unlock(&ci->i_ceph_lock);
2275 num_entries = max(num_entries, 1U);
2276 num_entries = min(num_entries, opt->max_readdir);
2277
2278 order = get_order(size * num_entries);
2279 while (order >= 0) {
2280 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2281 __GFP_NOWARN |
2282 __GFP_ZERO,
2283 order);
2284 if (rinfo->dir_entries)
2285 break;
2286 order--;
2287 }
2288 if (!rinfo->dir_entries)
2289 return -ENOMEM;
2290
2291 num_entries = (PAGE_SIZE << order) / size;
2292 num_entries = min(num_entries, opt->max_readdir);
2293
2294 rinfo->dir_buf_size = PAGE_SIZE << order;
2295 req->r_num_caps = num_entries + 1;
2296 req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2297 req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2298 return 0;
2299 }
2300
2301 /*
2302 * Create an mds request.
2303 */
2304 struct ceph_mds_request *
ceph_mdsc_create_request(struct ceph_mds_client * mdsc,int op,int mode)2305 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2306 {
2307 struct ceph_mds_request *req;
2308
2309 req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2310 if (!req)
2311 return ERR_PTR(-ENOMEM);
2312
2313 mutex_init(&req->r_fill_mutex);
2314 req->r_mdsc = mdsc;
2315 req->r_started = jiffies;
2316 req->r_start_latency = ktime_get();
2317 req->r_resend_mds = -1;
2318 INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2319 INIT_LIST_HEAD(&req->r_unsafe_target_item);
2320 req->r_fmode = -1;
2321 req->r_feature_needed = -1;
2322 kref_init(&req->r_kref);
2323 RB_CLEAR_NODE(&req->r_node);
2324 INIT_LIST_HEAD(&req->r_wait);
2325 init_completion(&req->r_completion);
2326 init_completion(&req->r_safe_completion);
2327 INIT_LIST_HEAD(&req->r_unsafe_item);
2328
2329 ktime_get_coarse_real_ts64(&req->r_stamp);
2330
2331 req->r_op = op;
2332 req->r_direct_mode = mode;
2333 return req;
2334 }
2335
2336 /*
2337 * return oldest (lowest) request, tid in request tree, 0 if none.
2338 *
2339 * called under mdsc->mutex.
2340 */
__get_oldest_req(struct ceph_mds_client * mdsc)2341 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2342 {
2343 if (RB_EMPTY_ROOT(&mdsc->request_tree))
2344 return NULL;
2345 return rb_entry(rb_first(&mdsc->request_tree),
2346 struct ceph_mds_request, r_node);
2347 }
2348
__get_oldest_tid(struct ceph_mds_client * mdsc)2349 static inline u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2350 {
2351 return mdsc->oldest_tid;
2352 }
2353
2354 /*
2355 * Build a dentry's path. Allocate on heap; caller must kfree. Based
2356 * on build_path_from_dentry in fs/cifs/dir.c.
2357 *
2358 * If @stop_on_nosnap, generate path relative to the first non-snapped
2359 * inode.
2360 *
2361 * Encode hidden .snap dirs as a double /, i.e.
2362 * foo/.snap/bar -> foo//bar
2363 */
ceph_mdsc_build_path(struct dentry * dentry,int * plen,u64 * pbase,int stop_on_nosnap)2364 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2365 int stop_on_nosnap)
2366 {
2367 struct dentry *temp;
2368 char *path;
2369 int pos;
2370 unsigned seq;
2371 u64 base;
2372
2373 if (!dentry)
2374 return ERR_PTR(-EINVAL);
2375
2376 path = __getname();
2377 if (!path)
2378 return ERR_PTR(-ENOMEM);
2379 retry:
2380 pos = PATH_MAX - 1;
2381 path[pos] = '\0';
2382
2383 seq = read_seqbegin(&rename_lock);
2384 rcu_read_lock();
2385 temp = dentry;
2386 for (;;) {
2387 struct inode *inode;
2388
2389 spin_lock(&temp->d_lock);
2390 inode = d_inode(temp);
2391 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2392 dout("build_path path+%d: %p SNAPDIR\n",
2393 pos, temp);
2394 } else if (stop_on_nosnap && inode && dentry != temp &&
2395 ceph_snap(inode) == CEPH_NOSNAP) {
2396 spin_unlock(&temp->d_lock);
2397 pos++; /* get rid of any prepended '/' */
2398 break;
2399 } else {
2400 pos -= temp->d_name.len;
2401 if (pos < 0) {
2402 spin_unlock(&temp->d_lock);
2403 break;
2404 }
2405 memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2406 }
2407 spin_unlock(&temp->d_lock);
2408 temp = READ_ONCE(temp->d_parent);
2409
2410 /* Are we at the root? */
2411 if (IS_ROOT(temp))
2412 break;
2413
2414 /* Are we out of buffer? */
2415 if (--pos < 0)
2416 break;
2417
2418 path[pos] = '/';
2419 }
2420 base = ceph_ino(d_inode(temp));
2421 rcu_read_unlock();
2422
2423 if (read_seqretry(&rename_lock, seq))
2424 goto retry;
2425
2426 if (pos < 0) {
2427 /*
2428 * A rename didn't occur, but somehow we didn't end up where
2429 * we thought we would. Throw a warning and try again.
2430 */
2431 pr_warn("build_path did not end path lookup where "
2432 "expected, pos is %d\n", pos);
2433 goto retry;
2434 }
2435
2436 *pbase = base;
2437 *plen = PATH_MAX - 1 - pos;
2438 dout("build_path on %p %d built %llx '%.*s'\n",
2439 dentry, d_count(dentry), base, *plen, path + pos);
2440 return path + pos;
2441 }
2442
build_dentry_path(struct dentry * dentry,struct inode * dir,const char ** ppath,int * ppathlen,u64 * pino,bool * pfreepath,bool parent_locked)2443 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2444 const char **ppath, int *ppathlen, u64 *pino,
2445 bool *pfreepath, bool parent_locked)
2446 {
2447 char *path;
2448
2449 rcu_read_lock();
2450 if (!dir)
2451 dir = d_inode_rcu(dentry->d_parent);
2452 if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
2453 *pino = ceph_ino(dir);
2454 rcu_read_unlock();
2455 *ppath = dentry->d_name.name;
2456 *ppathlen = dentry->d_name.len;
2457 return 0;
2458 }
2459 rcu_read_unlock();
2460 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2461 if (IS_ERR(path))
2462 return PTR_ERR(path);
2463 *ppath = path;
2464 *pfreepath = true;
2465 return 0;
2466 }
2467
build_inode_path(struct inode * inode,const char ** ppath,int * ppathlen,u64 * pino,bool * pfreepath)2468 static int build_inode_path(struct inode *inode,
2469 const char **ppath, int *ppathlen, u64 *pino,
2470 bool *pfreepath)
2471 {
2472 struct dentry *dentry;
2473 char *path;
2474
2475 if (ceph_snap(inode) == CEPH_NOSNAP) {
2476 *pino = ceph_ino(inode);
2477 *ppathlen = 0;
2478 return 0;
2479 }
2480 dentry = d_find_alias(inode);
2481 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2482 dput(dentry);
2483 if (IS_ERR(path))
2484 return PTR_ERR(path);
2485 *ppath = path;
2486 *pfreepath = true;
2487 return 0;
2488 }
2489
2490 /*
2491 * request arguments may be specified via an inode *, a dentry *, or
2492 * an explicit ino+path.
2493 */
set_request_path_attr(struct inode * rinode,struct dentry * rdentry,struct inode * rdiri,const char * rpath,u64 rino,const char ** ppath,int * pathlen,u64 * ino,bool * freepath,bool parent_locked)2494 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
2495 struct inode *rdiri, const char *rpath,
2496 u64 rino, const char **ppath, int *pathlen,
2497 u64 *ino, bool *freepath, bool parent_locked)
2498 {
2499 int r = 0;
2500
2501 if (rinode) {
2502 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2503 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2504 ceph_snap(rinode));
2505 } else if (rdentry) {
2506 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
2507 freepath, parent_locked);
2508 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2509 *ppath);
2510 } else if (rpath || rino) {
2511 *ino = rino;
2512 *ppath = rpath;
2513 *pathlen = rpath ? strlen(rpath) : 0;
2514 dout(" path %.*s\n", *pathlen, rpath);
2515 }
2516
2517 return r;
2518 }
2519
encode_timestamp_and_gids(void ** p,const struct ceph_mds_request * req)2520 static void encode_timestamp_and_gids(void **p,
2521 const struct ceph_mds_request *req)
2522 {
2523 struct ceph_timespec ts;
2524 int i;
2525
2526 ceph_encode_timespec64(&ts, &req->r_stamp);
2527 ceph_encode_copy(p, &ts, sizeof(ts));
2528
2529 /* gid_list */
2530 ceph_encode_32(p, req->r_cred->group_info->ngroups);
2531 for (i = 0; i < req->r_cred->group_info->ngroups; i++)
2532 ceph_encode_64(p, from_kgid(&init_user_ns,
2533 req->r_cred->group_info->gid[i]));
2534 }
2535
2536 /*
2537 * called under mdsc->mutex
2538 */
create_request_message(struct ceph_mds_session * session,struct ceph_mds_request * req,bool drop_cap_releases)2539 static struct ceph_msg *create_request_message(struct ceph_mds_session *session,
2540 struct ceph_mds_request *req,
2541 bool drop_cap_releases)
2542 {
2543 int mds = session->s_mds;
2544 struct ceph_mds_client *mdsc = session->s_mdsc;
2545 struct ceph_msg *msg;
2546 struct ceph_mds_request_head_old *head;
2547 const char *path1 = NULL;
2548 const char *path2 = NULL;
2549 u64 ino1 = 0, ino2 = 0;
2550 int pathlen1 = 0, pathlen2 = 0;
2551 bool freepath1 = false, freepath2 = false;
2552 int len;
2553 u16 releases;
2554 void *p, *end;
2555 int ret;
2556 bool legacy = !(session->s_con.peer_features & CEPH_FEATURE_FS_BTIME);
2557
2558 ret = set_request_path_attr(req->r_inode, req->r_dentry,
2559 req->r_parent, req->r_path1, req->r_ino1.ino,
2560 &path1, &pathlen1, &ino1, &freepath1,
2561 test_bit(CEPH_MDS_R_PARENT_LOCKED,
2562 &req->r_req_flags));
2563 if (ret < 0) {
2564 msg = ERR_PTR(ret);
2565 goto out;
2566 }
2567
2568 /* If r_old_dentry is set, then assume that its parent is locked */
2569 ret = set_request_path_attr(NULL, req->r_old_dentry,
2570 req->r_old_dentry_dir,
2571 req->r_path2, req->r_ino2.ino,
2572 &path2, &pathlen2, &ino2, &freepath2, true);
2573 if (ret < 0) {
2574 msg = ERR_PTR(ret);
2575 goto out_free1;
2576 }
2577
2578 len = legacy ? sizeof(*head) : sizeof(struct ceph_mds_request_head);
2579 len += pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
2580 sizeof(struct ceph_timespec);
2581 len += sizeof(u32) + (sizeof(u64) * req->r_cred->group_info->ngroups);
2582
2583 /* calculate (max) length for cap releases */
2584 len += sizeof(struct ceph_mds_request_release) *
2585 (!!req->r_inode_drop + !!req->r_dentry_drop +
2586 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2587
2588 if (req->r_dentry_drop)
2589 len += pathlen1;
2590 if (req->r_old_dentry_drop)
2591 len += pathlen2;
2592
2593 msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
2594 if (!msg) {
2595 msg = ERR_PTR(-ENOMEM);
2596 goto out_free2;
2597 }
2598
2599 msg->hdr.tid = cpu_to_le64(req->r_tid);
2600
2601 /*
2602 * The old ceph_mds_request_head didn't contain a version field, and
2603 * one was added when we moved the message version from 3->4.
2604 */
2605 if (legacy) {
2606 msg->hdr.version = cpu_to_le16(3);
2607 head = msg->front.iov_base;
2608 p = msg->front.iov_base + sizeof(*head);
2609 } else {
2610 struct ceph_mds_request_head *new_head = msg->front.iov_base;
2611
2612 msg->hdr.version = cpu_to_le16(4);
2613 new_head->version = cpu_to_le16(CEPH_MDS_REQUEST_HEAD_VERSION);
2614 head = (struct ceph_mds_request_head_old *)&new_head->oldest_client_tid;
2615 p = msg->front.iov_base + sizeof(*new_head);
2616 }
2617
2618 end = msg->front.iov_base + msg->front.iov_len;
2619
2620 head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2621 head->op = cpu_to_le32(req->r_op);
2622 head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns,
2623 req->r_cred->fsuid));
2624 head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns,
2625 req->r_cred->fsgid));
2626 head->ino = cpu_to_le64(req->r_deleg_ino);
2627 head->args = req->r_args;
2628
2629 ceph_encode_filepath(&p, end, ino1, path1);
2630 ceph_encode_filepath(&p, end, ino2, path2);
2631
2632 /* make note of release offset, in case we need to replay */
2633 req->r_request_release_offset = p - msg->front.iov_base;
2634
2635 /* cap releases */
2636 releases = 0;
2637 if (req->r_inode_drop)
2638 releases += ceph_encode_inode_release(&p,
2639 req->r_inode ? req->r_inode : d_inode(req->r_dentry),
2640 mds, req->r_inode_drop, req->r_inode_unless,
2641 req->r_op == CEPH_MDS_OP_READDIR);
2642 if (req->r_dentry_drop)
2643 releases += ceph_encode_dentry_release(&p, req->r_dentry,
2644 req->r_parent, mds, req->r_dentry_drop,
2645 req->r_dentry_unless);
2646 if (req->r_old_dentry_drop)
2647 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
2648 req->r_old_dentry_dir, mds,
2649 req->r_old_dentry_drop,
2650 req->r_old_dentry_unless);
2651 if (req->r_old_inode_drop)
2652 releases += ceph_encode_inode_release(&p,
2653 d_inode(req->r_old_dentry),
2654 mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
2655
2656 if (drop_cap_releases) {
2657 releases = 0;
2658 p = msg->front.iov_base + req->r_request_release_offset;
2659 }
2660
2661 head->num_releases = cpu_to_le16(releases);
2662
2663 encode_timestamp_and_gids(&p, req);
2664
2665 if (WARN_ON_ONCE(p > end)) {
2666 ceph_msg_put(msg);
2667 msg = ERR_PTR(-ERANGE);
2668 goto out_free2;
2669 }
2670
2671 msg->front.iov_len = p - msg->front.iov_base;
2672 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2673
2674 if (req->r_pagelist) {
2675 struct ceph_pagelist *pagelist = req->r_pagelist;
2676 ceph_msg_data_add_pagelist(msg, pagelist);
2677 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2678 } else {
2679 msg->hdr.data_len = 0;
2680 }
2681
2682 msg->hdr.data_off = cpu_to_le16(0);
2683
2684 out_free2:
2685 if (freepath2)
2686 ceph_mdsc_free_path((char *)path2, pathlen2);
2687 out_free1:
2688 if (freepath1)
2689 ceph_mdsc_free_path((char *)path1, pathlen1);
2690 out:
2691 return msg;
2692 }
2693
2694 /*
2695 * called under mdsc->mutex if error, under no mutex if
2696 * success.
2697 */
complete_request(struct ceph_mds_client * mdsc,struct ceph_mds_request * req)2698 static void complete_request(struct ceph_mds_client *mdsc,
2699 struct ceph_mds_request *req)
2700 {
2701 req->r_end_latency = ktime_get();
2702
2703 if (req->r_callback)
2704 req->r_callback(mdsc, req);
2705 complete_all(&req->r_completion);
2706 }
2707
2708 static struct ceph_mds_request_head_old *
find_old_request_head(void * p,u64 features)2709 find_old_request_head(void *p, u64 features)
2710 {
2711 bool legacy = !(features & CEPH_FEATURE_FS_BTIME);
2712 struct ceph_mds_request_head *new_head;
2713
2714 if (legacy)
2715 return (struct ceph_mds_request_head_old *)p;
2716 new_head = (struct ceph_mds_request_head *)p;
2717 return (struct ceph_mds_request_head_old *)&new_head->oldest_client_tid;
2718 }
2719
2720 /*
2721 * called under mdsc->mutex
2722 */
__prepare_send_request(struct ceph_mds_session * session,struct ceph_mds_request * req,bool drop_cap_releases)2723 static int __prepare_send_request(struct ceph_mds_session *session,
2724 struct ceph_mds_request *req,
2725 bool drop_cap_releases)
2726 {
2727 int mds = session->s_mds;
2728 struct ceph_mds_client *mdsc = session->s_mdsc;
2729 struct ceph_mds_request_head_old *rhead;
2730 struct ceph_msg *msg;
2731 int flags = 0, max_retry;
2732
2733 /*
2734 * The type of 'r_attempts' in kernel 'ceph_mds_request'
2735 * is 'int', while in 'ceph_mds_request_head' the type of
2736 * 'num_retry' is '__u8'. So in case the request retries
2737 * exceeding 256 times, the MDS will receive a incorrect
2738 * retry seq.
2739 *
2740 * In this case it's ususally a bug in MDS and continue
2741 * retrying the request makes no sense.
2742 *
2743 * In future this could be fixed in ceph code, so avoid
2744 * using the hardcode here.
2745 */
2746 max_retry = sizeof_field(struct ceph_mds_request_head, num_retry);
2747 max_retry = 1 << (max_retry * BITS_PER_BYTE);
2748 if (req->r_attempts >= max_retry) {
2749 pr_warn_ratelimited("%s request tid %llu seq overflow\n",
2750 __func__, req->r_tid);
2751 return -EMULTIHOP;
2752 }
2753
2754 req->r_attempts++;
2755 if (req->r_inode) {
2756 struct ceph_cap *cap =
2757 ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2758
2759 if (cap)
2760 req->r_sent_on_mseq = cap->mseq;
2761 else
2762 req->r_sent_on_mseq = -1;
2763 }
2764 dout("%s %p tid %lld %s (attempt %d)\n", __func__, req,
2765 req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2766
2767 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2768 void *p;
2769
2770 /*
2771 * Replay. Do not regenerate message (and rebuild
2772 * paths, etc.); just use the original message.
2773 * Rebuilding paths will break for renames because
2774 * d_move mangles the src name.
2775 */
2776 msg = req->r_request;
2777 rhead = find_old_request_head(msg->front.iov_base,
2778 session->s_con.peer_features);
2779
2780 flags = le32_to_cpu(rhead->flags);
2781 flags |= CEPH_MDS_FLAG_REPLAY;
2782 rhead->flags = cpu_to_le32(flags);
2783
2784 if (req->r_target_inode)
2785 rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2786
2787 rhead->num_retry = req->r_attempts - 1;
2788
2789 /* remove cap/dentry releases from message */
2790 rhead->num_releases = 0;
2791
2792 p = msg->front.iov_base + req->r_request_release_offset;
2793 encode_timestamp_and_gids(&p, req);
2794
2795 msg->front.iov_len = p - msg->front.iov_base;
2796 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2797 return 0;
2798 }
2799
2800 if (req->r_request) {
2801 ceph_msg_put(req->r_request);
2802 req->r_request = NULL;
2803 }
2804 msg = create_request_message(session, req, drop_cap_releases);
2805 if (IS_ERR(msg)) {
2806 req->r_err = PTR_ERR(msg);
2807 return PTR_ERR(msg);
2808 }
2809 req->r_request = msg;
2810
2811 rhead = find_old_request_head(msg->front.iov_base,
2812 session->s_con.peer_features);
2813 rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2814 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2815 flags |= CEPH_MDS_FLAG_REPLAY;
2816 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
2817 flags |= CEPH_MDS_FLAG_ASYNC;
2818 if (req->r_parent)
2819 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2820 rhead->flags = cpu_to_le32(flags);
2821 rhead->num_fwd = req->r_num_fwd;
2822 rhead->num_retry = req->r_attempts - 1;
2823
2824 dout(" r_parent = %p\n", req->r_parent);
2825 return 0;
2826 }
2827
2828 /*
2829 * called under mdsc->mutex
2830 */
__send_request(struct ceph_mds_session * session,struct ceph_mds_request * req,bool drop_cap_releases)2831 static int __send_request(struct ceph_mds_session *session,
2832 struct ceph_mds_request *req,
2833 bool drop_cap_releases)
2834 {
2835 int err;
2836
2837 err = __prepare_send_request(session, req, drop_cap_releases);
2838 if (!err) {
2839 ceph_msg_get(req->r_request);
2840 ceph_con_send(&session->s_con, req->r_request);
2841 }
2842
2843 return err;
2844 }
2845
2846 /*
2847 * send request, or put it on the appropriate wait list.
2848 */
__do_request(struct ceph_mds_client * mdsc,struct ceph_mds_request * req)2849 static void __do_request(struct ceph_mds_client *mdsc,
2850 struct ceph_mds_request *req)
2851 {
2852 struct ceph_mds_session *session = NULL;
2853 int mds = -1;
2854 int err = 0;
2855 bool random;
2856
2857 if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2858 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
2859 __unregister_request(mdsc, req);
2860 return;
2861 }
2862
2863 if (req->r_timeout &&
2864 time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2865 dout("do_request timed out\n");
2866 err = -ETIMEDOUT;
2867 goto finish;
2868 }
2869 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2870 dout("do_request forced umount\n");
2871 err = -EIO;
2872 goto finish;
2873 }
2874 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
2875 if (mdsc->mdsmap_err) {
2876 err = mdsc->mdsmap_err;
2877 dout("do_request mdsmap err %d\n", err);
2878 goto finish;
2879 }
2880 if (mdsc->mdsmap->m_epoch == 0) {
2881 dout("do_request no mdsmap, waiting for map\n");
2882 list_add(&req->r_wait, &mdsc->waiting_for_map);
2883 return;
2884 }
2885 if (!(mdsc->fsc->mount_options->flags &
2886 CEPH_MOUNT_OPT_MOUNTWAIT) &&
2887 !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
2888 err = -EHOSTUNREACH;
2889 goto finish;
2890 }
2891 }
2892
2893 put_request_session(req);
2894
2895 mds = __choose_mds(mdsc, req, &random);
2896 if (mds < 0 ||
2897 ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2898 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2899 err = -EJUKEBOX;
2900 goto finish;
2901 }
2902 dout("do_request no mds or not active, waiting for map\n");
2903 list_add(&req->r_wait, &mdsc->waiting_for_map);
2904 return;
2905 }
2906
2907 /* get, open session */
2908 session = __ceph_lookup_mds_session(mdsc, mds);
2909 if (!session) {
2910 session = register_session(mdsc, mds);
2911 if (IS_ERR(session)) {
2912 err = PTR_ERR(session);
2913 goto finish;
2914 }
2915 }
2916 req->r_session = ceph_get_mds_session(session);
2917
2918 dout("do_request mds%d session %p state %s\n", mds, session,
2919 ceph_session_state_name(session->s_state));
2920
2921 /*
2922 * The old ceph will crash the MDSs when see unknown OPs
2923 */
2924 if (req->r_feature_needed > 0 &&
2925 !test_bit(req->r_feature_needed, &session->s_features)) {
2926 err = -EOPNOTSUPP;
2927 goto out_session;
2928 }
2929
2930 if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2931 session->s_state != CEPH_MDS_SESSION_HUNG) {
2932 /*
2933 * We cannot queue async requests since the caps and delegated
2934 * inodes are bound to the session. Just return -EJUKEBOX and
2935 * let the caller retry a sync request in that case.
2936 */
2937 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2938 err = -EJUKEBOX;
2939 goto out_session;
2940 }
2941
2942 /*
2943 * If the session has been REJECTED, then return a hard error,
2944 * unless it's a CLEANRECOVER mount, in which case we'll queue
2945 * it to the mdsc queue.
2946 */
2947 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2948 if (ceph_test_mount_opt(mdsc->fsc, CLEANRECOVER))
2949 list_add(&req->r_wait, &mdsc->waiting_for_map);
2950 else
2951 err = -EACCES;
2952 goto out_session;
2953 }
2954
2955 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2956 session->s_state == CEPH_MDS_SESSION_CLOSING) {
2957 err = __open_session(mdsc, session);
2958 if (err)
2959 goto out_session;
2960 /* retry the same mds later */
2961 if (random)
2962 req->r_resend_mds = mds;
2963 }
2964 list_add(&req->r_wait, &session->s_waiting);
2965 goto out_session;
2966 }
2967
2968 /* send request */
2969 req->r_resend_mds = -1; /* forget any previous mds hint */
2970
2971 if (req->r_request_started == 0) /* note request start time */
2972 req->r_request_started = jiffies;
2973
2974 /*
2975 * For async create we will choose the auth MDS of frag in parent
2976 * directory to send the request and ususally this works fine, but
2977 * if the migrated the dirtory to another MDS before it could handle
2978 * it the request will be forwarded.
2979 *
2980 * And then the auth cap will be changed.
2981 */
2982 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags) && req->r_num_fwd) {
2983 struct ceph_dentry_info *di = ceph_dentry(req->r_dentry);
2984 struct ceph_inode_info *ci;
2985 struct ceph_cap *cap;
2986
2987 /*
2988 * The request maybe handled very fast and the new inode
2989 * hasn't been linked to the dentry yet. We need to wait
2990 * for the ceph_finish_async_create(), which shouldn't be
2991 * stuck too long or fail in thoery, to finish when forwarding
2992 * the request.
2993 */
2994 if (!d_inode(req->r_dentry)) {
2995 err = wait_on_bit(&di->flags, CEPH_DENTRY_ASYNC_CREATE_BIT,
2996 TASK_KILLABLE);
2997 if (err) {
2998 mutex_lock(&req->r_fill_mutex);
2999 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3000 mutex_unlock(&req->r_fill_mutex);
3001 goto out_session;
3002 }
3003 }
3004
3005 ci = ceph_inode(d_inode(req->r_dentry));
3006
3007 spin_lock(&ci->i_ceph_lock);
3008 cap = ci->i_auth_cap;
3009 if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE && mds != cap->mds) {
3010 dout("do_request session changed for auth cap %d -> %d\n",
3011 cap->session->s_mds, session->s_mds);
3012
3013 /* Remove the auth cap from old session */
3014 spin_lock(&cap->session->s_cap_lock);
3015 cap->session->s_nr_caps--;
3016 list_del_init(&cap->session_caps);
3017 spin_unlock(&cap->session->s_cap_lock);
3018
3019 /* Add the auth cap to the new session */
3020 cap->mds = mds;
3021 cap->session = session;
3022 spin_lock(&session->s_cap_lock);
3023 session->s_nr_caps++;
3024 list_add_tail(&cap->session_caps, &session->s_caps);
3025 spin_unlock(&session->s_cap_lock);
3026
3027 change_auth_cap_ses(ci, session);
3028 }
3029 spin_unlock(&ci->i_ceph_lock);
3030 }
3031
3032 err = __send_request(session, req, false);
3033
3034 out_session:
3035 ceph_put_mds_session(session);
3036 finish:
3037 if (err) {
3038 dout("__do_request early error %d\n", err);
3039 req->r_err = err;
3040 complete_request(mdsc, req);
3041 __unregister_request(mdsc, req);
3042 }
3043 return;
3044 }
3045
3046 /*
3047 * called under mdsc->mutex
3048 */
__wake_requests(struct ceph_mds_client * mdsc,struct list_head * head)3049 static void __wake_requests(struct ceph_mds_client *mdsc,
3050 struct list_head *head)
3051 {
3052 struct ceph_mds_request *req;
3053 LIST_HEAD(tmp_list);
3054
3055 list_splice_init(head, &tmp_list);
3056
3057 while (!list_empty(&tmp_list)) {
3058 req = list_entry(tmp_list.next,
3059 struct ceph_mds_request, r_wait);
3060 list_del_init(&req->r_wait);
3061 dout(" wake request %p tid %llu\n", req, req->r_tid);
3062 __do_request(mdsc, req);
3063 }
3064 }
3065
3066 /*
3067 * Wake up threads with requests pending for @mds, so that they can
3068 * resubmit their requests to a possibly different mds.
3069 */
kick_requests(struct ceph_mds_client * mdsc,int mds)3070 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
3071 {
3072 struct ceph_mds_request *req;
3073 struct rb_node *p = rb_first(&mdsc->request_tree);
3074
3075 dout("kick_requests mds%d\n", mds);
3076 while (p) {
3077 req = rb_entry(p, struct ceph_mds_request, r_node);
3078 p = rb_next(p);
3079 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3080 continue;
3081 if (req->r_attempts > 0)
3082 continue; /* only new requests */
3083 if (req->r_session &&
3084 req->r_session->s_mds == mds) {
3085 dout(" kicking tid %llu\n", req->r_tid);
3086 list_del_init(&req->r_wait);
3087 __do_request(mdsc, req);
3088 }
3089 }
3090 }
3091
ceph_mdsc_submit_request(struct ceph_mds_client * mdsc,struct inode * dir,struct ceph_mds_request * req)3092 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
3093 struct ceph_mds_request *req)
3094 {
3095 int err = 0;
3096
3097 /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
3098 if (req->r_inode)
3099 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
3100 if (req->r_parent) {
3101 struct ceph_inode_info *ci = ceph_inode(req->r_parent);
3102 int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
3103 CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
3104 spin_lock(&ci->i_ceph_lock);
3105 ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
3106 __ceph_touch_fmode(ci, mdsc, fmode);
3107 spin_unlock(&ci->i_ceph_lock);
3108 }
3109 if (req->r_old_dentry_dir)
3110 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
3111 CEPH_CAP_PIN);
3112
3113 if (req->r_inode) {
3114 err = ceph_wait_on_async_create(req->r_inode);
3115 if (err) {
3116 dout("%s: wait for async create returned: %d\n",
3117 __func__, err);
3118 return err;
3119 }
3120 }
3121
3122 if (!err && req->r_old_inode) {
3123 err = ceph_wait_on_async_create(req->r_old_inode);
3124 if (err) {
3125 dout("%s: wait for async create returned: %d\n",
3126 __func__, err);
3127 return err;
3128 }
3129 }
3130
3131 dout("submit_request on %p for inode %p\n", req, dir);
3132 mutex_lock(&mdsc->mutex);
3133 __register_request(mdsc, req, dir);
3134 __do_request(mdsc, req);
3135 err = req->r_err;
3136 mutex_unlock(&mdsc->mutex);
3137 return err;
3138 }
3139
ceph_mdsc_wait_request(struct ceph_mds_client * mdsc,struct ceph_mds_request * req,ceph_mds_request_wait_callback_t wait_func)3140 int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
3141 struct ceph_mds_request *req,
3142 ceph_mds_request_wait_callback_t wait_func)
3143 {
3144 int err;
3145
3146 /* wait */
3147 dout("do_request waiting\n");
3148 if (wait_func) {
3149 err = wait_func(mdsc, req);
3150 } else {
3151 long timeleft = wait_for_completion_killable_timeout(
3152 &req->r_completion,
3153 ceph_timeout_jiffies(req->r_timeout));
3154 if (timeleft > 0)
3155 err = 0;
3156 else if (!timeleft)
3157 err = -ETIMEDOUT; /* timed out */
3158 else
3159 err = timeleft; /* killed */
3160 }
3161 dout("do_request waited, got %d\n", err);
3162 mutex_lock(&mdsc->mutex);
3163
3164 /* only abort if we didn't race with a real reply */
3165 if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
3166 err = le32_to_cpu(req->r_reply_info.head->result);
3167 } else if (err < 0) {
3168 dout("aborted request %lld with %d\n", req->r_tid, err);
3169
3170 /*
3171 * ensure we aren't running concurrently with
3172 * ceph_fill_trace or ceph_readdir_prepopulate, which
3173 * rely on locks (dir mutex) held by our caller.
3174 */
3175 mutex_lock(&req->r_fill_mutex);
3176 req->r_err = err;
3177 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3178 mutex_unlock(&req->r_fill_mutex);
3179
3180 if (req->r_parent &&
3181 (req->r_op & CEPH_MDS_OP_WRITE))
3182 ceph_invalidate_dir_request(req);
3183 } else {
3184 err = req->r_err;
3185 }
3186
3187 mutex_unlock(&mdsc->mutex);
3188 return err;
3189 }
3190
3191 /*
3192 * Synchrously perform an mds request. Take care of all of the
3193 * session setup, forwarding, retry details.
3194 */
ceph_mdsc_do_request(struct ceph_mds_client * mdsc,struct inode * dir,struct ceph_mds_request * req)3195 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
3196 struct inode *dir,
3197 struct ceph_mds_request *req)
3198 {
3199 int err;
3200
3201 dout("do_request on %p\n", req);
3202
3203 /* issue */
3204 err = ceph_mdsc_submit_request(mdsc, dir, req);
3205 if (!err)
3206 err = ceph_mdsc_wait_request(mdsc, req, NULL);
3207 dout("do_request %p done, result %d\n", req, err);
3208 return err;
3209 }
3210
3211 /*
3212 * Invalidate dir's completeness, dentry lease state on an aborted MDS
3213 * namespace request.
3214 */
ceph_invalidate_dir_request(struct ceph_mds_request * req)3215 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
3216 {
3217 struct inode *dir = req->r_parent;
3218 struct inode *old_dir = req->r_old_dentry_dir;
3219
3220 dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
3221
3222 ceph_dir_clear_complete(dir);
3223 if (old_dir)
3224 ceph_dir_clear_complete(old_dir);
3225 if (req->r_dentry)
3226 ceph_invalidate_dentry_lease(req->r_dentry);
3227 if (req->r_old_dentry)
3228 ceph_invalidate_dentry_lease(req->r_old_dentry);
3229 }
3230
3231 /*
3232 * Handle mds reply.
3233 *
3234 * We take the session mutex and parse and process the reply immediately.
3235 * This preserves the logical ordering of replies, capabilities, etc., sent
3236 * by the MDS as they are applied to our local cache.
3237 */
handle_reply(struct ceph_mds_session * session,struct ceph_msg * msg)3238 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
3239 {
3240 struct ceph_mds_client *mdsc = session->s_mdsc;
3241 struct ceph_mds_request *req;
3242 struct ceph_mds_reply_head *head = msg->front.iov_base;
3243 struct ceph_mds_reply_info_parsed *rinfo; /* parsed reply info */
3244 struct ceph_snap_realm *realm;
3245 u64 tid;
3246 int err, result;
3247 int mds = session->s_mds;
3248
3249 if (msg->front.iov_len < sizeof(*head)) {
3250 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
3251 ceph_msg_dump(msg);
3252 return;
3253 }
3254
3255 /* get request, session */
3256 tid = le64_to_cpu(msg->hdr.tid);
3257 mutex_lock(&mdsc->mutex);
3258 req = lookup_get_request(mdsc, tid);
3259 if (!req) {
3260 dout("handle_reply on unknown tid %llu\n", tid);
3261 mutex_unlock(&mdsc->mutex);
3262 return;
3263 }
3264 dout("handle_reply %p\n", req);
3265
3266 /* correct session? */
3267 if (req->r_session != session) {
3268 pr_err("mdsc_handle_reply got %llu on session mds%d"
3269 " not mds%d\n", tid, session->s_mds,
3270 req->r_session ? req->r_session->s_mds : -1);
3271 mutex_unlock(&mdsc->mutex);
3272 goto out;
3273 }
3274
3275 /* dup? */
3276 if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3277 (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
3278 pr_warn("got a dup %s reply on %llu from mds%d\n",
3279 head->safe ? "safe" : "unsafe", tid, mds);
3280 mutex_unlock(&mdsc->mutex);
3281 goto out;
3282 }
3283 if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
3284 pr_warn("got unsafe after safe on %llu from mds%d\n",
3285 tid, mds);
3286 mutex_unlock(&mdsc->mutex);
3287 goto out;
3288 }
3289
3290 result = le32_to_cpu(head->result);
3291
3292 if (head->safe) {
3293 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
3294 __unregister_request(mdsc, req);
3295
3296 /* last request during umount? */
3297 if (mdsc->stopping && !__get_oldest_req(mdsc))
3298 complete_all(&mdsc->safe_umount_waiters);
3299
3300 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3301 /*
3302 * We already handled the unsafe response, now do the
3303 * cleanup. No need to examine the response; the MDS
3304 * doesn't include any result info in the safe
3305 * response. And even if it did, there is nothing
3306 * useful we could do with a revised return value.
3307 */
3308 dout("got safe reply %llu, mds%d\n", tid, mds);
3309
3310 mutex_unlock(&mdsc->mutex);
3311 goto out;
3312 }
3313 } else {
3314 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
3315 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3316 }
3317
3318 dout("handle_reply tid %lld result %d\n", tid, result);
3319 rinfo = &req->r_reply_info;
3320 if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
3321 err = parse_reply_info(session, msg, rinfo, (u64)-1);
3322 else
3323 err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);
3324 mutex_unlock(&mdsc->mutex);
3325
3326 /* Must find target inode outside of mutexes to avoid deadlocks */
3327 if ((err >= 0) && rinfo->head->is_target) {
3328 struct inode *in;
3329 struct ceph_vino tvino = {
3330 .ino = le64_to_cpu(rinfo->targeti.in->ino),
3331 .snap = le64_to_cpu(rinfo->targeti.in->snapid)
3332 };
3333
3334 in = ceph_get_inode(mdsc->fsc->sb, tvino);
3335 if (IS_ERR(in)) {
3336 err = PTR_ERR(in);
3337 mutex_lock(&session->s_mutex);
3338 goto out_err;
3339 }
3340 req->r_target_inode = in;
3341 }
3342
3343 mutex_lock(&session->s_mutex);
3344 if (err < 0) {
3345 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
3346 ceph_msg_dump(msg);
3347 goto out_err;
3348 }
3349
3350 /* snap trace */
3351 realm = NULL;
3352 if (rinfo->snapblob_len) {
3353 down_write(&mdsc->snap_rwsem);
3354 ceph_update_snap_trace(mdsc, rinfo->snapblob,
3355 rinfo->snapblob + rinfo->snapblob_len,
3356 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3357 &realm);
3358 downgrade_write(&mdsc->snap_rwsem);
3359 } else {
3360 down_read(&mdsc->snap_rwsem);
3361 }
3362
3363 /* insert trace into our cache */
3364 mutex_lock(&req->r_fill_mutex);
3365 current->journal_info = req;
3366 err = ceph_fill_trace(mdsc->fsc->sb, req);
3367 if (err == 0) {
3368 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
3369 req->r_op == CEPH_MDS_OP_LSSNAP))
3370 ceph_readdir_prepopulate(req, req->r_session);
3371 }
3372 current->journal_info = NULL;
3373 mutex_unlock(&req->r_fill_mutex);
3374
3375 up_read(&mdsc->snap_rwsem);
3376 if (realm)
3377 ceph_put_snap_realm(mdsc, realm);
3378
3379 if (err == 0) {
3380 if (req->r_target_inode &&
3381 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3382 struct ceph_inode_info *ci =
3383 ceph_inode(req->r_target_inode);
3384 spin_lock(&ci->i_unsafe_lock);
3385 list_add_tail(&req->r_unsafe_target_item,
3386 &ci->i_unsafe_iops);
3387 spin_unlock(&ci->i_unsafe_lock);
3388 }
3389
3390 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
3391 }
3392 out_err:
3393 mutex_lock(&mdsc->mutex);
3394 if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3395 if (err) {
3396 req->r_err = err;
3397 } else {
3398 req->r_reply = ceph_msg_get(msg);
3399 set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
3400 }
3401 } else {
3402 dout("reply arrived after request %lld was aborted\n", tid);
3403 }
3404 mutex_unlock(&mdsc->mutex);
3405
3406 mutex_unlock(&session->s_mutex);
3407
3408 /* kick calling process */
3409 complete_request(mdsc, req);
3410
3411 ceph_update_metadata_metrics(&mdsc->metric, req->r_start_latency,
3412 req->r_end_latency, err);
3413 out:
3414 ceph_mdsc_put_request(req);
3415 return;
3416 }
3417
3418
3419
3420 /*
3421 * handle mds notification that our request has been forwarded.
3422 */
handle_forward(struct ceph_mds_client * mdsc,struct ceph_mds_session * session,struct ceph_msg * msg)3423 static void handle_forward(struct ceph_mds_client *mdsc,
3424 struct ceph_mds_session *session,
3425 struct ceph_msg *msg)
3426 {
3427 struct ceph_mds_request *req;
3428 u64 tid = le64_to_cpu(msg->hdr.tid);
3429 u32 next_mds;
3430 u32 fwd_seq;
3431 int err = -EINVAL;
3432 void *p = msg->front.iov_base;
3433 void *end = p + msg->front.iov_len;
3434 bool aborted = false;
3435
3436 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3437 next_mds = ceph_decode_32(&p);
3438 fwd_seq = ceph_decode_32(&p);
3439
3440 mutex_lock(&mdsc->mutex);
3441 req = lookup_get_request(mdsc, tid);
3442 if (!req) {
3443 mutex_unlock(&mdsc->mutex);
3444 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
3445 return; /* dup reply? */
3446 }
3447
3448 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3449 dout("forward tid %llu aborted, unregistering\n", tid);
3450 __unregister_request(mdsc, req);
3451 } else if (fwd_seq <= req->r_num_fwd) {
3452 /*
3453 * The type of 'num_fwd' in ceph 'MClientRequestForward'
3454 * is 'int32_t', while in 'ceph_mds_request_head' the
3455 * type is '__u8'. So in case the request bounces between
3456 * MDSes exceeding 256 times, the client will get stuck.
3457 *
3458 * In this case it's ususally a bug in MDS and continue
3459 * bouncing the request makes no sense.
3460 *
3461 * In future this could be fixed in ceph code, so avoid
3462 * using the hardcode here.
3463 */
3464 int max = sizeof_field(struct ceph_mds_request_head, num_fwd);
3465 max = 1 << (max * BITS_PER_BYTE);
3466 if (req->r_num_fwd >= max) {
3467 mutex_lock(&req->r_fill_mutex);
3468 req->r_err = -EMULTIHOP;
3469 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3470 mutex_unlock(&req->r_fill_mutex);
3471 aborted = true;
3472 pr_warn_ratelimited("forward tid %llu seq overflow\n",
3473 tid);
3474 } else {
3475 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
3476 tid, next_mds, req->r_num_fwd, fwd_seq);
3477 }
3478 } else {
3479 /* resend. forward race not possible; mds would drop */
3480 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3481 BUG_ON(req->r_err);
3482 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3483 req->r_attempts = 0;
3484 req->r_num_fwd = fwd_seq;
3485 req->r_resend_mds = next_mds;
3486 put_request_session(req);
3487 __do_request(mdsc, req);
3488 }
3489 mutex_unlock(&mdsc->mutex);
3490
3491 /* kick calling process */
3492 if (aborted)
3493 complete_request(mdsc, req);
3494 ceph_mdsc_put_request(req);
3495 return;
3496
3497 bad:
3498 pr_err("mdsc_handle_forward decode error err=%d\n", err);
3499 }
3500
__decode_session_metadata(void ** p,void * end,bool * blocklisted)3501 static int __decode_session_metadata(void **p, void *end,
3502 bool *blocklisted)
3503 {
3504 /* map<string,string> */
3505 u32 n;
3506 bool err_str;
3507 ceph_decode_32_safe(p, end, n, bad);
3508 while (n-- > 0) {
3509 u32 len;
3510 ceph_decode_32_safe(p, end, len, bad);
3511 ceph_decode_need(p, end, len, bad);
3512 err_str = !strncmp(*p, "error_string", len);
3513 *p += len;
3514 ceph_decode_32_safe(p, end, len, bad);
3515 ceph_decode_need(p, end, len, bad);
3516 /*
3517 * Match "blocklisted (blacklisted)" from newer MDSes,
3518 * or "blacklisted" from older MDSes.
3519 */
3520 if (err_str && strnstr(*p, "blacklisted", len))
3521 *blocklisted = true;
3522 *p += len;
3523 }
3524 return 0;
3525 bad:
3526 return -1;
3527 }
3528
3529 /*
3530 * handle a mds session control message
3531 */
handle_session(struct ceph_mds_session * session,struct ceph_msg * msg)3532 static void handle_session(struct ceph_mds_session *session,
3533 struct ceph_msg *msg)
3534 {
3535 struct ceph_mds_client *mdsc = session->s_mdsc;
3536 int mds = session->s_mds;
3537 int msg_version = le16_to_cpu(msg->hdr.version);
3538 void *p = msg->front.iov_base;
3539 void *end = p + msg->front.iov_len;
3540 struct ceph_mds_session_head *h;
3541 u32 op;
3542 u64 seq, features = 0;
3543 int wake = 0;
3544 bool blocklisted = false;
3545
3546 /* decode */
3547 ceph_decode_need(&p, end, sizeof(*h), bad);
3548 h = p;
3549 p += sizeof(*h);
3550
3551 op = le32_to_cpu(h->op);
3552 seq = le64_to_cpu(h->seq);
3553
3554 if (msg_version >= 3) {
3555 u32 len;
3556 /* version >= 2 and < 5, decode metadata, skip otherwise
3557 * as it's handled via flags.
3558 */
3559 if (msg_version >= 5)
3560 ceph_decode_skip_map(&p, end, string, string, bad);
3561 else if (__decode_session_metadata(&p, end, &blocklisted) < 0)
3562 goto bad;
3563
3564 /* version >= 3, feature bits */
3565 ceph_decode_32_safe(&p, end, len, bad);
3566 if (len) {
3567 ceph_decode_64_safe(&p, end, features, bad);
3568 p += len - sizeof(features);
3569 }
3570 }
3571
3572 if (msg_version >= 5) {
3573 u32 flags, len;
3574
3575 /* version >= 4 */
3576 ceph_decode_skip_16(&p, end, bad); /* struct_v, struct_cv */
3577 ceph_decode_32_safe(&p, end, len, bad); /* len */
3578 ceph_decode_skip_n(&p, end, len, bad); /* metric_spec */
3579
3580 /* version >= 5, flags */
3581 ceph_decode_32_safe(&p, end, flags, bad);
3582 if (flags & CEPH_SESSION_BLOCKLISTED) {
3583 pr_warn("mds%d session blocklisted\n", session->s_mds);
3584 blocklisted = true;
3585 }
3586 }
3587
3588 mutex_lock(&mdsc->mutex);
3589 if (op == CEPH_SESSION_CLOSE) {
3590 ceph_get_mds_session(session);
3591 __unregister_session(mdsc, session);
3592 }
3593 /* FIXME: this ttl calculation is generous */
3594 session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3595 mutex_unlock(&mdsc->mutex);
3596
3597 mutex_lock(&session->s_mutex);
3598
3599 dout("handle_session mds%d %s %p state %s seq %llu\n",
3600 mds, ceph_session_op_name(op), session,
3601 ceph_session_state_name(session->s_state), seq);
3602
3603 if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3604 session->s_state = CEPH_MDS_SESSION_OPEN;
3605 pr_info("mds%d came back\n", session->s_mds);
3606 }
3607
3608 switch (op) {
3609 case CEPH_SESSION_OPEN:
3610 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3611 pr_info("mds%d reconnect success\n", session->s_mds);
3612
3613 if (session->s_state == CEPH_MDS_SESSION_OPEN) {
3614 pr_notice("mds%d is already opened\n", session->s_mds);
3615 } else {
3616 session->s_state = CEPH_MDS_SESSION_OPEN;
3617 session->s_features = features;
3618 renewed_caps(mdsc, session, 0);
3619 if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT,
3620 &session->s_features))
3621 metric_schedule_delayed(&mdsc->metric);
3622 }
3623
3624 /*
3625 * The connection maybe broken and the session in client
3626 * side has been reinitialized, need to update the seq
3627 * anyway.
3628 */
3629 if (!session->s_seq && seq)
3630 session->s_seq = seq;
3631
3632 wake = 1;
3633 if (mdsc->stopping)
3634 __close_session(mdsc, session);
3635 break;
3636
3637 case CEPH_SESSION_RENEWCAPS:
3638 if (session->s_renew_seq == seq)
3639 renewed_caps(mdsc, session, 1);
3640 break;
3641
3642 case CEPH_SESSION_CLOSE:
3643 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3644 pr_info("mds%d reconnect denied\n", session->s_mds);
3645 session->s_state = CEPH_MDS_SESSION_CLOSED;
3646 cleanup_session_requests(mdsc, session);
3647 remove_session_caps(session);
3648 wake = 2; /* for good measure */
3649 wake_up_all(&mdsc->session_close_wq);
3650 break;
3651
3652 case CEPH_SESSION_STALE:
3653 pr_info("mds%d caps went stale, renewing\n",
3654 session->s_mds);
3655 atomic_inc(&session->s_cap_gen);
3656 session->s_cap_ttl = jiffies - 1;
3657 send_renew_caps(mdsc, session);
3658 break;
3659
3660 case CEPH_SESSION_RECALL_STATE:
3661 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
3662 break;
3663
3664 case CEPH_SESSION_FLUSHMSG:
3665 send_flushmsg_ack(mdsc, session, seq);
3666 break;
3667
3668 case CEPH_SESSION_FORCE_RO:
3669 dout("force_session_readonly %p\n", session);
3670 spin_lock(&session->s_cap_lock);
3671 session->s_readonly = true;
3672 spin_unlock(&session->s_cap_lock);
3673 wake_up_session_caps(session, FORCE_RO);
3674 break;
3675
3676 case CEPH_SESSION_REJECT:
3677 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3678 pr_info("mds%d rejected session\n", session->s_mds);
3679 session->s_state = CEPH_MDS_SESSION_REJECTED;
3680 cleanup_session_requests(mdsc, session);
3681 remove_session_caps(session);
3682 if (blocklisted)
3683 mdsc->fsc->blocklisted = true;
3684 wake = 2; /* for good measure */
3685 break;
3686
3687 default:
3688 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3689 WARN_ON(1);
3690 }
3691
3692 mutex_unlock(&session->s_mutex);
3693 if (wake) {
3694 mutex_lock(&mdsc->mutex);
3695 __wake_requests(mdsc, &session->s_waiting);
3696 if (wake == 2)
3697 kick_requests(mdsc, mds);
3698 mutex_unlock(&mdsc->mutex);
3699 }
3700 if (op == CEPH_SESSION_CLOSE)
3701 ceph_put_mds_session(session);
3702 return;
3703
3704 bad:
3705 pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3706 (int)msg->front.iov_len);
3707 ceph_msg_dump(msg);
3708 return;
3709 }
3710
ceph_mdsc_release_dir_caps(struct ceph_mds_request * req)3711 void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
3712 {
3713 int dcaps;
3714
3715 dcaps = xchg(&req->r_dir_caps, 0);
3716 if (dcaps) {
3717 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3718 ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
3719 }
3720 }
3721
ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request * req)3722 void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
3723 {
3724 int dcaps;
3725
3726 dcaps = xchg(&req->r_dir_caps, 0);
3727 if (dcaps) {
3728 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3729 ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
3730 dcaps);
3731 }
3732 }
3733
3734 /*
3735 * called under session->mutex.
3736 */
replay_unsafe_requests(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)3737 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3738 struct ceph_mds_session *session)
3739 {
3740 struct ceph_mds_request *req, *nreq;
3741 struct rb_node *p;
3742
3743 dout("replay_unsafe_requests mds%d\n", session->s_mds);
3744
3745 mutex_lock(&mdsc->mutex);
3746 list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
3747 __send_request(session, req, true);
3748
3749 /*
3750 * also re-send old requests when MDS enters reconnect stage. So that MDS
3751 * can process completed request in clientreplay stage.
3752 */
3753 p = rb_first(&mdsc->request_tree);
3754 while (p) {
3755 req = rb_entry(p, struct ceph_mds_request, r_node);
3756 p = rb_next(p);
3757 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3758 continue;
3759 if (req->r_attempts == 0)
3760 continue; /* only old requests */
3761 if (!req->r_session)
3762 continue;
3763 if (req->r_session->s_mds != session->s_mds)
3764 continue;
3765
3766 ceph_mdsc_release_dir_caps_no_check(req);
3767
3768 __send_request(session, req, true);
3769 }
3770 mutex_unlock(&mdsc->mutex);
3771 }
3772
send_reconnect_partial(struct ceph_reconnect_state * recon_state)3773 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3774 {
3775 struct ceph_msg *reply;
3776 struct ceph_pagelist *_pagelist;
3777 struct page *page;
3778 __le32 *addr;
3779 int err = -ENOMEM;
3780
3781 if (!recon_state->allow_multi)
3782 return -ENOSPC;
3783
3784 /* can't handle message that contains both caps and realm */
3785 BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3786
3787 /* pre-allocate new pagelist */
3788 _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3789 if (!_pagelist)
3790 return -ENOMEM;
3791
3792 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3793 if (!reply)
3794 goto fail_msg;
3795
3796 /* placeholder for nr_caps */
3797 err = ceph_pagelist_encode_32(_pagelist, 0);
3798 if (err < 0)
3799 goto fail;
3800
3801 if (recon_state->nr_caps) {
3802 /* currently encoding caps */
3803 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3804 if (err)
3805 goto fail;
3806 } else {
3807 /* placeholder for nr_realms (currently encoding relams) */
3808 err = ceph_pagelist_encode_32(_pagelist, 0);
3809 if (err < 0)
3810 goto fail;
3811 }
3812
3813 err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3814 if (err)
3815 goto fail;
3816
3817 page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3818 addr = kmap_atomic(page);
3819 if (recon_state->nr_caps) {
3820 /* currently encoding caps */
3821 *addr = cpu_to_le32(recon_state->nr_caps);
3822 } else {
3823 /* currently encoding relams */
3824 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3825 }
3826 kunmap_atomic(addr);
3827
3828 reply->hdr.version = cpu_to_le16(5);
3829 reply->hdr.compat_version = cpu_to_le16(4);
3830
3831 reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3832 ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3833
3834 ceph_con_send(&recon_state->session->s_con, reply);
3835 ceph_pagelist_release(recon_state->pagelist);
3836
3837 recon_state->pagelist = _pagelist;
3838 recon_state->nr_caps = 0;
3839 recon_state->nr_realms = 0;
3840 recon_state->msg_version = 5;
3841 return 0;
3842 fail:
3843 ceph_msg_put(reply);
3844 fail_msg:
3845 ceph_pagelist_release(_pagelist);
3846 return err;
3847 }
3848
d_find_primary(struct inode * inode)3849 static struct dentry* d_find_primary(struct inode *inode)
3850 {
3851 struct dentry *alias, *dn = NULL;
3852
3853 if (hlist_empty(&inode->i_dentry))
3854 return NULL;
3855
3856 spin_lock(&inode->i_lock);
3857 if (hlist_empty(&inode->i_dentry))
3858 goto out_unlock;
3859
3860 if (S_ISDIR(inode->i_mode)) {
3861 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
3862 if (!IS_ROOT(alias))
3863 dn = dget(alias);
3864 goto out_unlock;
3865 }
3866
3867 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
3868 spin_lock(&alias->d_lock);
3869 if (!d_unhashed(alias) &&
3870 (ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {
3871 dn = dget_dlock(alias);
3872 }
3873 spin_unlock(&alias->d_lock);
3874 if (dn)
3875 break;
3876 }
3877 out_unlock:
3878 spin_unlock(&inode->i_lock);
3879 return dn;
3880 }
3881
3882 /*
3883 * Encode information about a cap for a reconnect with the MDS.
3884 */
reconnect_caps_cb(struct inode * inode,struct ceph_cap * cap,void * arg)3885 static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
3886 void *arg)
3887 {
3888 union {
3889 struct ceph_mds_cap_reconnect v2;
3890 struct ceph_mds_cap_reconnect_v1 v1;
3891 } rec;
3892 struct ceph_inode_info *ci = cap->ci;
3893 struct ceph_reconnect_state *recon_state = arg;
3894 struct ceph_pagelist *pagelist = recon_state->pagelist;
3895 struct dentry *dentry;
3896 char *path;
3897 int pathlen = 0, err;
3898 u64 pathbase;
3899 u64 snap_follows;
3900
3901 dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3902 inode, ceph_vinop(inode), cap, cap->cap_id,
3903 ceph_cap_string(cap->issued));
3904
3905 dentry = d_find_primary(inode);
3906 if (dentry) {
3907 /* set pathbase to parent dir when msg_version >= 2 */
3908 path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase,
3909 recon_state->msg_version >= 2);
3910 dput(dentry);
3911 if (IS_ERR(path)) {
3912 err = PTR_ERR(path);
3913 goto out_err;
3914 }
3915 } else {
3916 path = NULL;
3917 pathbase = 0;
3918 }
3919
3920 spin_lock(&ci->i_ceph_lock);
3921 cap->seq = 0; /* reset cap seq */
3922 cap->issue_seq = 0; /* and issue_seq */
3923 cap->mseq = 0; /* and migrate_seq */
3924 cap->cap_gen = atomic_read(&cap->session->s_cap_gen);
3925
3926 /* These are lost when the session goes away */
3927 if (S_ISDIR(inode->i_mode)) {
3928 if (cap->issued & CEPH_CAP_DIR_CREATE) {
3929 ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
3930 memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
3931 }
3932 cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
3933 }
3934
3935 if (recon_state->msg_version >= 2) {
3936 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3937 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3938 rec.v2.issued = cpu_to_le32(cap->issued);
3939 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3940 rec.v2.pathbase = cpu_to_le64(pathbase);
3941 rec.v2.flock_len = (__force __le32)
3942 ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
3943 } else {
3944 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3945 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3946 rec.v1.issued = cpu_to_le32(cap->issued);
3947 rec.v1.size = cpu_to_le64(i_size_read(inode));
3948 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3949 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
3950 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3951 rec.v1.pathbase = cpu_to_le64(pathbase);
3952 }
3953
3954 if (list_empty(&ci->i_cap_snaps)) {
3955 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3956 } else {
3957 struct ceph_cap_snap *capsnap =
3958 list_first_entry(&ci->i_cap_snaps,
3959 struct ceph_cap_snap, ci_item);
3960 snap_follows = capsnap->follows;
3961 }
3962 spin_unlock(&ci->i_ceph_lock);
3963
3964 if (recon_state->msg_version >= 2) {
3965 int num_fcntl_locks, num_flock_locks;
3966 struct ceph_filelock *flocks = NULL;
3967 size_t struct_len, total_len = sizeof(u64);
3968 u8 struct_v = 0;
3969
3970 encode_again:
3971 if (rec.v2.flock_len) {
3972 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3973 } else {
3974 num_fcntl_locks = 0;
3975 num_flock_locks = 0;
3976 }
3977 if (num_fcntl_locks + num_flock_locks > 0) {
3978 flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3979 sizeof(struct ceph_filelock),
3980 GFP_NOFS);
3981 if (!flocks) {
3982 err = -ENOMEM;
3983 goto out_err;
3984 }
3985 err = ceph_encode_locks_to_buffer(inode, flocks,
3986 num_fcntl_locks,
3987 num_flock_locks);
3988 if (err) {
3989 kfree(flocks);
3990 flocks = NULL;
3991 if (err == -ENOSPC)
3992 goto encode_again;
3993 goto out_err;
3994 }
3995 } else {
3996 kfree(flocks);
3997 flocks = NULL;
3998 }
3999
4000 if (recon_state->msg_version >= 3) {
4001 /* version, compat_version and struct_len */
4002 total_len += 2 * sizeof(u8) + sizeof(u32);
4003 struct_v = 2;
4004 }
4005 /*
4006 * number of encoded locks is stable, so copy to pagelist
4007 */
4008 struct_len = 2 * sizeof(u32) +
4009 (num_fcntl_locks + num_flock_locks) *
4010 sizeof(struct ceph_filelock);
4011 rec.v2.flock_len = cpu_to_le32(struct_len);
4012
4013 struct_len += sizeof(u32) + pathlen + sizeof(rec.v2);
4014
4015 if (struct_v >= 2)
4016 struct_len += sizeof(u64); /* snap_follows */
4017
4018 total_len += struct_len;
4019
4020 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
4021 err = send_reconnect_partial(recon_state);
4022 if (err)
4023 goto out_freeflocks;
4024 pagelist = recon_state->pagelist;
4025 }
4026
4027 err = ceph_pagelist_reserve(pagelist, total_len);
4028 if (err)
4029 goto out_freeflocks;
4030
4031 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
4032 if (recon_state->msg_version >= 3) {
4033 ceph_pagelist_encode_8(pagelist, struct_v);
4034 ceph_pagelist_encode_8(pagelist, 1);
4035 ceph_pagelist_encode_32(pagelist, struct_len);
4036 }
4037 ceph_pagelist_encode_string(pagelist, path, pathlen);
4038 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
4039 ceph_locks_to_pagelist(flocks, pagelist,
4040 num_fcntl_locks, num_flock_locks);
4041 if (struct_v >= 2)
4042 ceph_pagelist_encode_64(pagelist, snap_follows);
4043 out_freeflocks:
4044 kfree(flocks);
4045 } else {
4046 err = ceph_pagelist_reserve(pagelist,
4047 sizeof(u64) + sizeof(u32) +
4048 pathlen + sizeof(rec.v1));
4049 if (err)
4050 goto out_err;
4051
4052 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
4053 ceph_pagelist_encode_string(pagelist, path, pathlen);
4054 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
4055 }
4056
4057 out_err:
4058 ceph_mdsc_free_path(path, pathlen);
4059 if (!err)
4060 recon_state->nr_caps++;
4061 return err;
4062 }
4063
encode_snap_realms(struct ceph_mds_client * mdsc,struct ceph_reconnect_state * recon_state)4064 static int encode_snap_realms(struct ceph_mds_client *mdsc,
4065 struct ceph_reconnect_state *recon_state)
4066 {
4067 struct rb_node *p;
4068 struct ceph_pagelist *pagelist = recon_state->pagelist;
4069 int err = 0;
4070
4071 if (recon_state->msg_version >= 4) {
4072 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
4073 if (err < 0)
4074 goto fail;
4075 }
4076
4077 /*
4078 * snaprealms. we provide mds with the ino, seq (version), and
4079 * parent for all of our realms. If the mds has any newer info,
4080 * it will tell us.
4081 */
4082 for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
4083 struct ceph_snap_realm *realm =
4084 rb_entry(p, struct ceph_snap_realm, node);
4085 struct ceph_mds_snaprealm_reconnect sr_rec;
4086
4087 if (recon_state->msg_version >= 4) {
4088 size_t need = sizeof(u8) * 2 + sizeof(u32) +
4089 sizeof(sr_rec);
4090
4091 if (pagelist->length + need > RECONNECT_MAX_SIZE) {
4092 err = send_reconnect_partial(recon_state);
4093 if (err)
4094 goto fail;
4095 pagelist = recon_state->pagelist;
4096 }
4097
4098 err = ceph_pagelist_reserve(pagelist, need);
4099 if (err)
4100 goto fail;
4101
4102 ceph_pagelist_encode_8(pagelist, 1);
4103 ceph_pagelist_encode_8(pagelist, 1);
4104 ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
4105 }
4106
4107 dout(" adding snap realm %llx seq %lld parent %llx\n",
4108 realm->ino, realm->seq, realm->parent_ino);
4109 sr_rec.ino = cpu_to_le64(realm->ino);
4110 sr_rec.seq = cpu_to_le64(realm->seq);
4111 sr_rec.parent = cpu_to_le64(realm->parent_ino);
4112
4113 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
4114 if (err)
4115 goto fail;
4116
4117 recon_state->nr_realms++;
4118 }
4119 fail:
4120 return err;
4121 }
4122
4123
4124 /*
4125 * If an MDS fails and recovers, clients need to reconnect in order to
4126 * reestablish shared state. This includes all caps issued through
4127 * this session _and_ the snap_realm hierarchy. Because it's not
4128 * clear which snap realms the mds cares about, we send everything we
4129 * know about.. that ensures we'll then get any new info the
4130 * recovering MDS might have.
4131 *
4132 * This is a relatively heavyweight operation, but it's rare.
4133 */
send_mds_reconnect(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)4134 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
4135 struct ceph_mds_session *session)
4136 {
4137 struct ceph_msg *reply;
4138 int mds = session->s_mds;
4139 int err = -ENOMEM;
4140 struct ceph_reconnect_state recon_state = {
4141 .session = session,
4142 };
4143 LIST_HEAD(dispose);
4144
4145 pr_info("mds%d reconnect start\n", mds);
4146
4147 recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
4148 if (!recon_state.pagelist)
4149 goto fail_nopagelist;
4150
4151 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
4152 if (!reply)
4153 goto fail_nomsg;
4154
4155 xa_destroy(&session->s_delegated_inos);
4156
4157 mutex_lock(&session->s_mutex);
4158 session->s_state = CEPH_MDS_SESSION_RECONNECTING;
4159 session->s_seq = 0;
4160
4161 dout("session %p state %s\n", session,
4162 ceph_session_state_name(session->s_state));
4163
4164 atomic_inc(&session->s_cap_gen);
4165
4166 spin_lock(&session->s_cap_lock);
4167 /* don't know if session is readonly */
4168 session->s_readonly = 0;
4169 /*
4170 * notify __ceph_remove_cap() that we are composing cap reconnect.
4171 * If a cap get released before being added to the cap reconnect,
4172 * __ceph_remove_cap() should skip queuing cap release.
4173 */
4174 session->s_cap_reconnect = 1;
4175 /* drop old cap expires; we're about to reestablish that state */
4176 detach_cap_releases(session, &dispose);
4177 spin_unlock(&session->s_cap_lock);
4178 dispose_cap_releases(mdsc, &dispose);
4179
4180 /* trim unused caps to reduce MDS's cache rejoin time */
4181 if (mdsc->fsc->sb->s_root)
4182 shrink_dcache_parent(mdsc->fsc->sb->s_root);
4183
4184 ceph_con_close(&session->s_con);
4185 ceph_con_open(&session->s_con,
4186 CEPH_ENTITY_TYPE_MDS, mds,
4187 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
4188
4189 /* replay unsafe requests */
4190 replay_unsafe_requests(mdsc, session);
4191
4192 ceph_early_kick_flushing_caps(mdsc, session);
4193
4194 down_read(&mdsc->snap_rwsem);
4195
4196 /* placeholder for nr_caps */
4197 err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
4198 if (err)
4199 goto fail;
4200
4201 if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
4202 recon_state.msg_version = 3;
4203 recon_state.allow_multi = true;
4204 } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
4205 recon_state.msg_version = 3;
4206 } else {
4207 recon_state.msg_version = 2;
4208 }
4209 /* trsaverse this session's caps */
4210 err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
4211
4212 spin_lock(&session->s_cap_lock);
4213 session->s_cap_reconnect = 0;
4214 spin_unlock(&session->s_cap_lock);
4215
4216 if (err < 0)
4217 goto fail;
4218
4219 /* check if all realms can be encoded into current message */
4220 if (mdsc->num_snap_realms) {
4221 size_t total_len =
4222 recon_state.pagelist->length +
4223 mdsc->num_snap_realms *
4224 sizeof(struct ceph_mds_snaprealm_reconnect);
4225 if (recon_state.msg_version >= 4) {
4226 /* number of realms */
4227 total_len += sizeof(u32);
4228 /* version, compat_version and struct_len */
4229 total_len += mdsc->num_snap_realms *
4230 (2 * sizeof(u8) + sizeof(u32));
4231 }
4232 if (total_len > RECONNECT_MAX_SIZE) {
4233 if (!recon_state.allow_multi) {
4234 err = -ENOSPC;
4235 goto fail;
4236 }
4237 if (recon_state.nr_caps) {
4238 err = send_reconnect_partial(&recon_state);
4239 if (err)
4240 goto fail;
4241 }
4242 recon_state.msg_version = 5;
4243 }
4244 }
4245
4246 err = encode_snap_realms(mdsc, &recon_state);
4247 if (err < 0)
4248 goto fail;
4249
4250 if (recon_state.msg_version >= 5) {
4251 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
4252 if (err < 0)
4253 goto fail;
4254 }
4255
4256 if (recon_state.nr_caps || recon_state.nr_realms) {
4257 struct page *page =
4258 list_first_entry(&recon_state.pagelist->head,
4259 struct page, lru);
4260 __le32 *addr = kmap_atomic(page);
4261 if (recon_state.nr_caps) {
4262 WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
4263 *addr = cpu_to_le32(recon_state.nr_caps);
4264 } else if (recon_state.msg_version >= 4) {
4265 *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
4266 }
4267 kunmap_atomic(addr);
4268 }
4269
4270 reply->hdr.version = cpu_to_le16(recon_state.msg_version);
4271 if (recon_state.msg_version >= 4)
4272 reply->hdr.compat_version = cpu_to_le16(4);
4273
4274 reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
4275 ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
4276
4277 ceph_con_send(&session->s_con, reply);
4278
4279 mutex_unlock(&session->s_mutex);
4280
4281 mutex_lock(&mdsc->mutex);
4282 __wake_requests(mdsc, &session->s_waiting);
4283 mutex_unlock(&mdsc->mutex);
4284
4285 up_read(&mdsc->snap_rwsem);
4286 ceph_pagelist_release(recon_state.pagelist);
4287 return;
4288
4289 fail:
4290 ceph_msg_put(reply);
4291 up_read(&mdsc->snap_rwsem);
4292 mutex_unlock(&session->s_mutex);
4293 fail_nomsg:
4294 ceph_pagelist_release(recon_state.pagelist);
4295 fail_nopagelist:
4296 pr_err("error %d preparing reconnect for mds%d\n", err, mds);
4297 return;
4298 }
4299
4300
4301 /*
4302 * compare old and new mdsmaps, kicking requests
4303 * and closing out old connections as necessary
4304 *
4305 * called under mdsc->mutex.
4306 */
check_new_map(struct ceph_mds_client * mdsc,struct ceph_mdsmap * newmap,struct ceph_mdsmap * oldmap)4307 static void check_new_map(struct ceph_mds_client *mdsc,
4308 struct ceph_mdsmap *newmap,
4309 struct ceph_mdsmap *oldmap)
4310 {
4311 int i, j, err;
4312 int oldstate, newstate;
4313 struct ceph_mds_session *s;
4314 unsigned long targets[DIV_ROUND_UP(CEPH_MAX_MDS, sizeof(unsigned long))] = {0};
4315
4316 dout("check_new_map new %u old %u\n",
4317 newmap->m_epoch, oldmap->m_epoch);
4318
4319 if (newmap->m_info) {
4320 for (i = 0; i < newmap->possible_max_rank; i++) {
4321 for (j = 0; j < newmap->m_info[i].num_export_targets; j++)
4322 set_bit(newmap->m_info[i].export_targets[j], targets);
4323 }
4324 }
4325
4326 for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4327 if (!mdsc->sessions[i])
4328 continue;
4329 s = mdsc->sessions[i];
4330 oldstate = ceph_mdsmap_get_state(oldmap, i);
4331 newstate = ceph_mdsmap_get_state(newmap, i);
4332
4333 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
4334 i, ceph_mds_state_name(oldstate),
4335 ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
4336 ceph_mds_state_name(newstate),
4337 ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
4338 ceph_session_state_name(s->s_state));
4339
4340 if (i >= newmap->possible_max_rank) {
4341 /* force close session for stopped mds */
4342 ceph_get_mds_session(s);
4343 __unregister_session(mdsc, s);
4344 __wake_requests(mdsc, &s->s_waiting);
4345 mutex_unlock(&mdsc->mutex);
4346
4347 mutex_lock(&s->s_mutex);
4348 cleanup_session_requests(mdsc, s);
4349 remove_session_caps(s);
4350 mutex_unlock(&s->s_mutex);
4351
4352 ceph_put_mds_session(s);
4353
4354 mutex_lock(&mdsc->mutex);
4355 kick_requests(mdsc, i);
4356 continue;
4357 }
4358
4359 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4360 ceph_mdsmap_get_addr(newmap, i),
4361 sizeof(struct ceph_entity_addr))) {
4362 /* just close it */
4363 mutex_unlock(&mdsc->mutex);
4364 mutex_lock(&s->s_mutex);
4365 mutex_lock(&mdsc->mutex);
4366 ceph_con_close(&s->s_con);
4367 mutex_unlock(&s->s_mutex);
4368 s->s_state = CEPH_MDS_SESSION_RESTARTING;
4369 } else if (oldstate == newstate) {
4370 continue; /* nothing new with this mds */
4371 }
4372
4373 /*
4374 * send reconnect?
4375 */
4376 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
4377 newstate >= CEPH_MDS_STATE_RECONNECT) {
4378 mutex_unlock(&mdsc->mutex);
4379 clear_bit(i, targets);
4380 send_mds_reconnect(mdsc, s);
4381 mutex_lock(&mdsc->mutex);
4382 }
4383
4384 /*
4385 * kick request on any mds that has gone active.
4386 */
4387 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4388 newstate >= CEPH_MDS_STATE_ACTIVE) {
4389 if (oldstate != CEPH_MDS_STATE_CREATING &&
4390 oldstate != CEPH_MDS_STATE_STARTING)
4391 pr_info("mds%d recovery completed\n", s->s_mds);
4392 kick_requests(mdsc, i);
4393 mutex_unlock(&mdsc->mutex);
4394 mutex_lock(&s->s_mutex);
4395 mutex_lock(&mdsc->mutex);
4396 ceph_kick_flushing_caps(mdsc, s);
4397 mutex_unlock(&s->s_mutex);
4398 wake_up_session_caps(s, RECONNECT);
4399 }
4400 }
4401
4402 /*
4403 * Only open and reconnect sessions that don't exist yet.
4404 */
4405 for (i = 0; i < newmap->possible_max_rank; i++) {
4406 /*
4407 * In case the import MDS is crashed just after
4408 * the EImportStart journal is flushed, so when
4409 * a standby MDS takes over it and is replaying
4410 * the EImportStart journal the new MDS daemon
4411 * will wait the client to reconnect it, but the
4412 * client may never register/open the session yet.
4413 *
4414 * Will try to reconnect that MDS daemon if the
4415 * rank number is in the export targets array and
4416 * is the up:reconnect state.
4417 */
4418 newstate = ceph_mdsmap_get_state(newmap, i);
4419 if (!test_bit(i, targets) || newstate != CEPH_MDS_STATE_RECONNECT)
4420 continue;
4421
4422 /*
4423 * The session maybe registered and opened by some
4424 * requests which were choosing random MDSes during
4425 * the mdsc->mutex's unlock/lock gap below in rare
4426 * case. But the related MDS daemon will just queue
4427 * that requests and be still waiting for the client's
4428 * reconnection request in up:reconnect state.
4429 */
4430 s = __ceph_lookup_mds_session(mdsc, i);
4431 if (likely(!s)) {
4432 s = __open_export_target_session(mdsc, i);
4433 if (IS_ERR(s)) {
4434 err = PTR_ERR(s);
4435 pr_err("failed to open export target session, err %d\n",
4436 err);
4437 continue;
4438 }
4439 }
4440 dout("send reconnect to export target mds.%d\n", i);
4441 mutex_unlock(&mdsc->mutex);
4442 send_mds_reconnect(mdsc, s);
4443 ceph_put_mds_session(s);
4444 mutex_lock(&mdsc->mutex);
4445 }
4446
4447 for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4448 s = mdsc->sessions[i];
4449 if (!s)
4450 continue;
4451 if (!ceph_mdsmap_is_laggy(newmap, i))
4452 continue;
4453 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4454 s->s_state == CEPH_MDS_SESSION_HUNG ||
4455 s->s_state == CEPH_MDS_SESSION_CLOSING) {
4456 dout(" connecting to export targets of laggy mds%d\n",
4457 i);
4458 __open_export_target_sessions(mdsc, s);
4459 }
4460 }
4461 }
4462
4463
4464
4465 /*
4466 * leases
4467 */
4468
4469 /*
4470 * caller must hold session s_mutex, dentry->d_lock
4471 */
__ceph_mdsc_drop_dentry_lease(struct dentry * dentry)4472 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
4473 {
4474 struct ceph_dentry_info *di = ceph_dentry(dentry);
4475
4476 ceph_put_mds_session(di->lease_session);
4477 di->lease_session = NULL;
4478 }
4479
handle_lease(struct ceph_mds_client * mdsc,struct ceph_mds_session * session,struct ceph_msg * msg)4480 static void handle_lease(struct ceph_mds_client *mdsc,
4481 struct ceph_mds_session *session,
4482 struct ceph_msg *msg)
4483 {
4484 struct super_block *sb = mdsc->fsc->sb;
4485 struct inode *inode;
4486 struct dentry *parent, *dentry;
4487 struct ceph_dentry_info *di;
4488 int mds = session->s_mds;
4489 struct ceph_mds_lease *h = msg->front.iov_base;
4490 u32 seq;
4491 struct ceph_vino vino;
4492 struct qstr dname;
4493 int release = 0;
4494
4495 dout("handle_lease from mds%d\n", mds);
4496
4497 /* decode */
4498 if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
4499 goto bad;
4500 vino.ino = le64_to_cpu(h->ino);
4501 vino.snap = CEPH_NOSNAP;
4502 seq = le32_to_cpu(h->seq);
4503 dname.len = get_unaligned_le32(h + 1);
4504 if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
4505 goto bad;
4506 dname.name = (void *)(h + 1) + sizeof(u32);
4507
4508 /* lookup inode */
4509 inode = ceph_find_inode(sb, vino);
4510 dout("handle_lease %s, ino %llx %p %.*s\n",
4511 ceph_lease_op_name(h->action), vino.ino, inode,
4512 dname.len, dname.name);
4513
4514 mutex_lock(&session->s_mutex);
4515 inc_session_sequence(session);
4516
4517 if (!inode) {
4518 dout("handle_lease no inode %llx\n", vino.ino);
4519 goto release;
4520 }
4521
4522 /* dentry */
4523 parent = d_find_alias(inode);
4524 if (!parent) {
4525 dout("no parent dentry on inode %p\n", inode);
4526 WARN_ON(1);
4527 goto release; /* hrm... */
4528 }
4529 dname.hash = full_name_hash(parent, dname.name, dname.len);
4530 dentry = d_lookup(parent, &dname);
4531 dput(parent);
4532 if (!dentry)
4533 goto release;
4534
4535 spin_lock(&dentry->d_lock);
4536 di = ceph_dentry(dentry);
4537 switch (h->action) {
4538 case CEPH_MDS_LEASE_REVOKE:
4539 if (di->lease_session == session) {
4540 if (ceph_seq_cmp(di->lease_seq, seq) > 0)
4541 h->seq = cpu_to_le32(di->lease_seq);
4542 __ceph_mdsc_drop_dentry_lease(dentry);
4543 }
4544 release = 1;
4545 break;
4546
4547 case CEPH_MDS_LEASE_RENEW:
4548 if (di->lease_session == session &&
4549 di->lease_gen == atomic_read(&session->s_cap_gen) &&
4550 di->lease_renew_from &&
4551 di->lease_renew_after == 0) {
4552 unsigned long duration =
4553 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
4554
4555 di->lease_seq = seq;
4556 di->time = di->lease_renew_from + duration;
4557 di->lease_renew_after = di->lease_renew_from +
4558 (duration >> 1);
4559 di->lease_renew_from = 0;
4560 }
4561 break;
4562 }
4563 spin_unlock(&dentry->d_lock);
4564 dput(dentry);
4565
4566 if (!release)
4567 goto out;
4568
4569 release:
4570 /* let's just reuse the same message */
4571 h->action = CEPH_MDS_LEASE_REVOKE_ACK;
4572 ceph_msg_get(msg);
4573 ceph_con_send(&session->s_con, msg);
4574
4575 out:
4576 mutex_unlock(&session->s_mutex);
4577 iput(inode);
4578 return;
4579
4580 bad:
4581 pr_err("corrupt lease message\n");
4582 ceph_msg_dump(msg);
4583 }
4584
ceph_mdsc_lease_send_msg(struct ceph_mds_session * session,struct dentry * dentry,char action,u32 seq)4585 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
4586 struct dentry *dentry, char action,
4587 u32 seq)
4588 {
4589 struct ceph_msg *msg;
4590 struct ceph_mds_lease *lease;
4591 struct inode *dir;
4592 int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
4593
4594 dout("lease_send_msg identry %p %s to mds%d\n",
4595 dentry, ceph_lease_op_name(action), session->s_mds);
4596
4597 msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
4598 if (!msg)
4599 return;
4600 lease = msg->front.iov_base;
4601 lease->action = action;
4602 lease->seq = cpu_to_le32(seq);
4603
4604 spin_lock(&dentry->d_lock);
4605 dir = d_inode(dentry->d_parent);
4606 lease->ino = cpu_to_le64(ceph_ino(dir));
4607 lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
4608
4609 put_unaligned_le32(dentry->d_name.len, lease + 1);
4610 memcpy((void *)(lease + 1) + 4,
4611 dentry->d_name.name, dentry->d_name.len);
4612 spin_unlock(&dentry->d_lock);
4613
4614 ceph_con_send(&session->s_con, msg);
4615 }
4616
4617 /*
4618 * lock unlock the session, to wait ongoing session activities
4619 */
lock_unlock_session(struct ceph_mds_session * s)4620 static void lock_unlock_session(struct ceph_mds_session *s)
4621 {
4622 mutex_lock(&s->s_mutex);
4623 mutex_unlock(&s->s_mutex);
4624 }
4625
maybe_recover_session(struct ceph_mds_client * mdsc)4626 static void maybe_recover_session(struct ceph_mds_client *mdsc)
4627 {
4628 struct ceph_fs_client *fsc = mdsc->fsc;
4629
4630 if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
4631 return;
4632
4633 if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
4634 return;
4635
4636 if (!READ_ONCE(fsc->blocklisted))
4637 return;
4638
4639 pr_info("auto reconnect after blocklisted\n");
4640 ceph_force_reconnect(fsc->sb);
4641 }
4642
check_session_state(struct ceph_mds_session * s)4643 bool check_session_state(struct ceph_mds_session *s)
4644 {
4645 switch (s->s_state) {
4646 case CEPH_MDS_SESSION_OPEN:
4647 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
4648 s->s_state = CEPH_MDS_SESSION_HUNG;
4649 pr_info("mds%d hung\n", s->s_mds);
4650 }
4651 break;
4652 case CEPH_MDS_SESSION_CLOSING:
4653 case CEPH_MDS_SESSION_NEW:
4654 case CEPH_MDS_SESSION_RESTARTING:
4655 case CEPH_MDS_SESSION_CLOSED:
4656 case CEPH_MDS_SESSION_REJECTED:
4657 return false;
4658 }
4659
4660 return true;
4661 }
4662
4663 /*
4664 * If the sequence is incremented while we're waiting on a REQUEST_CLOSE reply,
4665 * then we need to retransmit that request.
4666 */
inc_session_sequence(struct ceph_mds_session * s)4667 void inc_session_sequence(struct ceph_mds_session *s)
4668 {
4669 lockdep_assert_held(&s->s_mutex);
4670
4671 s->s_seq++;
4672
4673 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4674 int ret;
4675
4676 dout("resending session close request for mds%d\n", s->s_mds);
4677 ret = request_close_session(s);
4678 if (ret < 0)
4679 pr_err("unable to close session to mds%d: %d\n",
4680 s->s_mds, ret);
4681 }
4682 }
4683
4684 /*
4685 * delayed work -- periodically trim expired leases, renew caps with mds. If
4686 * the @delay parameter is set to 0 or if it's more than 5 secs, the default
4687 * workqueue delay value of 5 secs will be used.
4688 */
schedule_delayed(struct ceph_mds_client * mdsc,unsigned long delay)4689 static void schedule_delayed(struct ceph_mds_client *mdsc, unsigned long delay)
4690 {
4691 unsigned long max_delay = HZ * 5;
4692
4693 /* 5 secs default delay */
4694 if (!delay || (delay > max_delay))
4695 delay = max_delay;
4696 schedule_delayed_work(&mdsc->delayed_work,
4697 round_jiffies_relative(delay));
4698 }
4699
delayed_work(struct work_struct * work)4700 static void delayed_work(struct work_struct *work)
4701 {
4702 struct ceph_mds_client *mdsc =
4703 container_of(work, struct ceph_mds_client, delayed_work.work);
4704 unsigned long delay;
4705 int renew_interval;
4706 int renew_caps;
4707 int i;
4708
4709 dout("mdsc delayed_work\n");
4710
4711 if (mdsc->stopping)
4712 return;
4713
4714 mutex_lock(&mdsc->mutex);
4715 renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4716 renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4717 mdsc->last_renew_caps);
4718 if (renew_caps)
4719 mdsc->last_renew_caps = jiffies;
4720
4721 for (i = 0; i < mdsc->max_sessions; i++) {
4722 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4723 if (!s)
4724 continue;
4725
4726 if (!check_session_state(s)) {
4727 ceph_put_mds_session(s);
4728 continue;
4729 }
4730 mutex_unlock(&mdsc->mutex);
4731
4732 mutex_lock(&s->s_mutex);
4733 if (renew_caps)
4734 send_renew_caps(mdsc, s);
4735 else
4736 ceph_con_keepalive(&s->s_con);
4737 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4738 s->s_state == CEPH_MDS_SESSION_HUNG)
4739 ceph_send_cap_releases(mdsc, s);
4740 mutex_unlock(&s->s_mutex);
4741 ceph_put_mds_session(s);
4742
4743 mutex_lock(&mdsc->mutex);
4744 }
4745 mutex_unlock(&mdsc->mutex);
4746
4747 delay = ceph_check_delayed_caps(mdsc);
4748
4749 ceph_queue_cap_reclaim_work(mdsc);
4750
4751 ceph_trim_snapid_map(mdsc);
4752
4753 maybe_recover_session(mdsc);
4754
4755 schedule_delayed(mdsc, delay);
4756 }
4757
ceph_mdsc_init(struct ceph_fs_client * fsc)4758 int ceph_mdsc_init(struct ceph_fs_client *fsc)
4759
4760 {
4761 struct ceph_mds_client *mdsc;
4762 int err;
4763
4764 mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4765 if (!mdsc)
4766 return -ENOMEM;
4767 mdsc->fsc = fsc;
4768 mutex_init(&mdsc->mutex);
4769 mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
4770 if (!mdsc->mdsmap) {
4771 err = -ENOMEM;
4772 goto err_mdsc;
4773 }
4774
4775 init_completion(&mdsc->safe_umount_waiters);
4776 init_waitqueue_head(&mdsc->session_close_wq);
4777 INIT_LIST_HEAD(&mdsc->waiting_for_map);
4778 mdsc->quotarealms_inodes = RB_ROOT;
4779 mutex_init(&mdsc->quotarealms_inodes_mutex);
4780 init_rwsem(&mdsc->snap_rwsem);
4781 mdsc->snap_realms = RB_ROOT;
4782 INIT_LIST_HEAD(&mdsc->snap_empty);
4783 spin_lock_init(&mdsc->snap_empty_lock);
4784 mdsc->request_tree = RB_ROOT;
4785 INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4786 mdsc->last_renew_caps = jiffies;
4787 INIT_LIST_HEAD(&mdsc->cap_delay_list);
4788 INIT_LIST_HEAD(&mdsc->cap_wait_list);
4789 spin_lock_init(&mdsc->cap_delay_lock);
4790 INIT_LIST_HEAD(&mdsc->snap_flush_list);
4791 spin_lock_init(&mdsc->snap_flush_lock);
4792 mdsc->last_cap_flush_tid = 1;
4793 INIT_LIST_HEAD(&mdsc->cap_flush_list);
4794 INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
4795 spin_lock_init(&mdsc->cap_dirty_lock);
4796 init_waitqueue_head(&mdsc->cap_flushing_wq);
4797 INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
4798 err = ceph_metric_init(&mdsc->metric);
4799 if (err)
4800 goto err_mdsmap;
4801
4802 spin_lock_init(&mdsc->dentry_list_lock);
4803 INIT_LIST_HEAD(&mdsc->dentry_leases);
4804 INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
4805
4806 ceph_caps_init(mdsc);
4807 ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
4808
4809 spin_lock_init(&mdsc->snapid_map_lock);
4810 mdsc->snapid_map_tree = RB_ROOT;
4811 INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4812
4813 init_rwsem(&mdsc->pool_perm_rwsem);
4814 mdsc->pool_perm_tree = RB_ROOT;
4815
4816 strscpy(mdsc->nodename, utsname()->nodename,
4817 sizeof(mdsc->nodename));
4818
4819 fsc->mdsc = mdsc;
4820 return 0;
4821
4822 err_mdsmap:
4823 kfree(mdsc->mdsmap);
4824 err_mdsc:
4825 kfree(mdsc);
4826 return err;
4827 }
4828
4829 /*
4830 * Wait for safe replies on open mds requests. If we time out, drop
4831 * all requests from the tree to avoid dangling dentry refs.
4832 */
wait_requests(struct ceph_mds_client * mdsc)4833 static void wait_requests(struct ceph_mds_client *mdsc)
4834 {
4835 struct ceph_options *opts = mdsc->fsc->client->options;
4836 struct ceph_mds_request *req;
4837
4838 mutex_lock(&mdsc->mutex);
4839 if (__get_oldest_req(mdsc)) {
4840 mutex_unlock(&mdsc->mutex);
4841
4842 dout("wait_requests waiting for requests\n");
4843 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
4844 ceph_timeout_jiffies(opts->mount_timeout));
4845
4846 /* tear down remaining requests */
4847 mutex_lock(&mdsc->mutex);
4848 while ((req = __get_oldest_req(mdsc))) {
4849 dout("wait_requests timed out on tid %llu\n",
4850 req->r_tid);
4851 list_del_init(&req->r_wait);
4852 __unregister_request(mdsc, req);
4853 }
4854 }
4855 mutex_unlock(&mdsc->mutex);
4856 dout("wait_requests done\n");
4857 }
4858
send_flush_mdlog(struct ceph_mds_session * s)4859 void send_flush_mdlog(struct ceph_mds_session *s)
4860 {
4861 struct ceph_msg *msg;
4862
4863 /*
4864 * Pre-luminous MDS crashes when it sees an unknown session request
4865 */
4866 if (!CEPH_HAVE_FEATURE(s->s_con.peer_features, SERVER_LUMINOUS))
4867 return;
4868
4869 mutex_lock(&s->s_mutex);
4870 dout("request mdlog flush to mds%d (%s)s seq %lld\n", s->s_mds,
4871 ceph_session_state_name(s->s_state), s->s_seq);
4872 msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_FLUSH_MDLOG,
4873 s->s_seq);
4874 if (!msg) {
4875 pr_err("failed to request mdlog flush to mds%d (%s) seq %lld\n",
4876 s->s_mds, ceph_session_state_name(s->s_state), s->s_seq);
4877 } else {
4878 ceph_con_send(&s->s_con, msg);
4879 }
4880 mutex_unlock(&s->s_mutex);
4881 }
4882
4883 /*
4884 * called before mount is ro, and before dentries are torn down.
4885 * (hmm, does this still race with new lookups?)
4886 */
ceph_mdsc_pre_umount(struct ceph_mds_client * mdsc)4887 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4888 {
4889 dout("pre_umount\n");
4890 mdsc->stopping = 1;
4891
4892 ceph_mdsc_iterate_sessions(mdsc, send_flush_mdlog, true);
4893 ceph_mdsc_iterate_sessions(mdsc, lock_unlock_session, false);
4894 ceph_flush_dirty_caps(mdsc);
4895 wait_requests(mdsc);
4896
4897 /*
4898 * wait for reply handlers to drop their request refs and
4899 * their inode/dcache refs
4900 */
4901 ceph_msgr_flush();
4902
4903 ceph_cleanup_quotarealms_inodes(mdsc);
4904 }
4905
4906 /*
4907 * flush the mdlog and wait for all write mds requests to flush.
4908 */
flush_mdlog_and_wait_mdsc_unsafe_requests(struct ceph_mds_client * mdsc,u64 want_tid)4909 static void flush_mdlog_and_wait_mdsc_unsafe_requests(struct ceph_mds_client *mdsc,
4910 u64 want_tid)
4911 {
4912 struct ceph_mds_request *req = NULL, *nextreq;
4913 struct ceph_mds_session *last_session = NULL;
4914 struct rb_node *n;
4915
4916 mutex_lock(&mdsc->mutex);
4917 dout("%s want %lld\n", __func__, want_tid);
4918 restart:
4919 req = __get_oldest_req(mdsc);
4920 while (req && req->r_tid <= want_tid) {
4921 /* find next request */
4922 n = rb_next(&req->r_node);
4923 if (n)
4924 nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4925 else
4926 nextreq = NULL;
4927 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4928 (req->r_op & CEPH_MDS_OP_WRITE)) {
4929 struct ceph_mds_session *s = req->r_session;
4930
4931 if (!s) {
4932 req = nextreq;
4933 continue;
4934 }
4935
4936 /* write op */
4937 ceph_mdsc_get_request(req);
4938 if (nextreq)
4939 ceph_mdsc_get_request(nextreq);
4940 s = ceph_get_mds_session(s);
4941 mutex_unlock(&mdsc->mutex);
4942
4943 /* send flush mdlog request to MDS */
4944 if (last_session != s) {
4945 send_flush_mdlog(s);
4946 ceph_put_mds_session(last_session);
4947 last_session = s;
4948 } else {
4949 ceph_put_mds_session(s);
4950 }
4951 dout("%s wait on %llu (want %llu)\n", __func__,
4952 req->r_tid, want_tid);
4953 wait_for_completion(&req->r_safe_completion);
4954
4955 mutex_lock(&mdsc->mutex);
4956 ceph_mdsc_put_request(req);
4957 if (!nextreq)
4958 break; /* next dne before, so we're done! */
4959 if (RB_EMPTY_NODE(&nextreq->r_node)) {
4960 /* next request was removed from tree */
4961 ceph_mdsc_put_request(nextreq);
4962 goto restart;
4963 }
4964 ceph_mdsc_put_request(nextreq); /* won't go away */
4965 }
4966 req = nextreq;
4967 }
4968 mutex_unlock(&mdsc->mutex);
4969 ceph_put_mds_session(last_session);
4970 dout("%s done\n", __func__);
4971 }
4972
ceph_mdsc_sync(struct ceph_mds_client * mdsc)4973 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4974 {
4975 u64 want_tid, want_flush;
4976
4977 if (READ_ONCE(mdsc->fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN)
4978 return;
4979
4980 dout("sync\n");
4981 mutex_lock(&mdsc->mutex);
4982 want_tid = mdsc->last_tid;
4983 mutex_unlock(&mdsc->mutex);
4984
4985 ceph_flush_dirty_caps(mdsc);
4986 spin_lock(&mdsc->cap_dirty_lock);
4987 want_flush = mdsc->last_cap_flush_tid;
4988 if (!list_empty(&mdsc->cap_flush_list)) {
4989 struct ceph_cap_flush *cf =
4990 list_last_entry(&mdsc->cap_flush_list,
4991 struct ceph_cap_flush, g_list);
4992 cf->wake = true;
4993 }
4994 spin_unlock(&mdsc->cap_dirty_lock);
4995
4996 dout("sync want tid %lld flush_seq %lld\n",
4997 want_tid, want_flush);
4998
4999 flush_mdlog_and_wait_mdsc_unsafe_requests(mdsc, want_tid);
5000 wait_caps_flush(mdsc, want_flush);
5001 }
5002
5003 /*
5004 * true if all sessions are closed, or we force unmount
5005 */
done_closing_sessions(struct ceph_mds_client * mdsc,int skipped)5006 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
5007 {
5008 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
5009 return true;
5010 return atomic_read(&mdsc->num_sessions) <= skipped;
5011 }
5012
5013 /*
5014 * called after sb is ro.
5015 */
ceph_mdsc_close_sessions(struct ceph_mds_client * mdsc)5016 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
5017 {
5018 struct ceph_options *opts = mdsc->fsc->client->options;
5019 struct ceph_mds_session *session;
5020 int i;
5021 int skipped = 0;
5022
5023 dout("close_sessions\n");
5024
5025 /* close sessions */
5026 mutex_lock(&mdsc->mutex);
5027 for (i = 0; i < mdsc->max_sessions; i++) {
5028 session = __ceph_lookup_mds_session(mdsc, i);
5029 if (!session)
5030 continue;
5031 mutex_unlock(&mdsc->mutex);
5032 mutex_lock(&session->s_mutex);
5033 if (__close_session(mdsc, session) <= 0)
5034 skipped++;
5035 mutex_unlock(&session->s_mutex);
5036 ceph_put_mds_session(session);
5037 mutex_lock(&mdsc->mutex);
5038 }
5039 mutex_unlock(&mdsc->mutex);
5040
5041 dout("waiting for sessions to close\n");
5042 wait_event_timeout(mdsc->session_close_wq,
5043 done_closing_sessions(mdsc, skipped),
5044 ceph_timeout_jiffies(opts->mount_timeout));
5045
5046 /* tear down remaining sessions */
5047 mutex_lock(&mdsc->mutex);
5048 for (i = 0; i < mdsc->max_sessions; i++) {
5049 if (mdsc->sessions[i]) {
5050 session = ceph_get_mds_session(mdsc->sessions[i]);
5051 __unregister_session(mdsc, session);
5052 mutex_unlock(&mdsc->mutex);
5053 mutex_lock(&session->s_mutex);
5054 remove_session_caps(session);
5055 mutex_unlock(&session->s_mutex);
5056 ceph_put_mds_session(session);
5057 mutex_lock(&mdsc->mutex);
5058 }
5059 }
5060 WARN_ON(!list_empty(&mdsc->cap_delay_list));
5061 mutex_unlock(&mdsc->mutex);
5062
5063 ceph_cleanup_snapid_map(mdsc);
5064 ceph_cleanup_global_and_empty_realms(mdsc);
5065
5066 cancel_work_sync(&mdsc->cap_reclaim_work);
5067 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
5068
5069 dout("stopped\n");
5070 }
5071
ceph_mdsc_force_umount(struct ceph_mds_client * mdsc)5072 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
5073 {
5074 struct ceph_mds_session *session;
5075 int mds;
5076
5077 dout("force umount\n");
5078
5079 mutex_lock(&mdsc->mutex);
5080 for (mds = 0; mds < mdsc->max_sessions; mds++) {
5081 session = __ceph_lookup_mds_session(mdsc, mds);
5082 if (!session)
5083 continue;
5084
5085 if (session->s_state == CEPH_MDS_SESSION_REJECTED)
5086 __unregister_session(mdsc, session);
5087 __wake_requests(mdsc, &session->s_waiting);
5088 mutex_unlock(&mdsc->mutex);
5089
5090 mutex_lock(&session->s_mutex);
5091 __close_session(mdsc, session);
5092 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
5093 cleanup_session_requests(mdsc, session);
5094 remove_session_caps(session);
5095 }
5096 mutex_unlock(&session->s_mutex);
5097 ceph_put_mds_session(session);
5098
5099 mutex_lock(&mdsc->mutex);
5100 kick_requests(mdsc, mds);
5101 }
5102 __wake_requests(mdsc, &mdsc->waiting_for_map);
5103 mutex_unlock(&mdsc->mutex);
5104 }
5105
ceph_mdsc_stop(struct ceph_mds_client * mdsc)5106 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
5107 {
5108 dout("stop\n");
5109 /*
5110 * Make sure the delayed work stopped before releasing
5111 * the resources.
5112 *
5113 * Because the cancel_delayed_work_sync() will only
5114 * guarantee that the work finishes executing. But the
5115 * delayed work will re-arm itself again after that.
5116 */
5117 flush_delayed_work(&mdsc->delayed_work);
5118
5119 if (mdsc->mdsmap)
5120 ceph_mdsmap_destroy(mdsc->mdsmap);
5121 kfree(mdsc->sessions);
5122 ceph_caps_finalize(mdsc);
5123 ceph_pool_perm_destroy(mdsc);
5124 }
5125
ceph_mdsc_destroy(struct ceph_fs_client * fsc)5126 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
5127 {
5128 struct ceph_mds_client *mdsc = fsc->mdsc;
5129 dout("mdsc_destroy %p\n", mdsc);
5130
5131 if (!mdsc)
5132 return;
5133
5134 /* flush out any connection work with references to us */
5135 ceph_msgr_flush();
5136
5137 ceph_mdsc_stop(mdsc);
5138
5139 ceph_metric_destroy(&mdsc->metric);
5140
5141 fsc->mdsc = NULL;
5142 kfree(mdsc);
5143 dout("mdsc_destroy %p done\n", mdsc);
5144 }
5145
ceph_mdsc_handle_fsmap(struct ceph_mds_client * mdsc,struct ceph_msg * msg)5146 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
5147 {
5148 struct ceph_fs_client *fsc = mdsc->fsc;
5149 const char *mds_namespace = fsc->mount_options->mds_namespace;
5150 void *p = msg->front.iov_base;
5151 void *end = p + msg->front.iov_len;
5152 u32 epoch;
5153 u32 num_fs;
5154 u32 mount_fscid = (u32)-1;
5155 int err = -EINVAL;
5156
5157 ceph_decode_need(&p, end, sizeof(u32), bad);
5158 epoch = ceph_decode_32(&p);
5159
5160 dout("handle_fsmap epoch %u\n", epoch);
5161
5162 /* struct_v, struct_cv, map_len, epoch, legacy_client_fscid */
5163 ceph_decode_skip_n(&p, end, 2 + sizeof(u32) * 3, bad);
5164
5165 ceph_decode_32_safe(&p, end, num_fs, bad);
5166 while (num_fs-- > 0) {
5167 void *info_p, *info_end;
5168 u32 info_len;
5169 u32 fscid, namelen;
5170
5171 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
5172 p += 2; // info_v, info_cv
5173 info_len = ceph_decode_32(&p);
5174 ceph_decode_need(&p, end, info_len, bad);
5175 info_p = p;
5176 info_end = p + info_len;
5177 p = info_end;
5178
5179 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
5180 fscid = ceph_decode_32(&info_p);
5181 namelen = ceph_decode_32(&info_p);
5182 ceph_decode_need(&info_p, info_end, namelen, bad);
5183
5184 if (mds_namespace &&
5185 strlen(mds_namespace) == namelen &&
5186 !strncmp(mds_namespace, (char *)info_p, namelen)) {
5187 mount_fscid = fscid;
5188 break;
5189 }
5190 }
5191
5192 ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
5193 if (mount_fscid != (u32)-1) {
5194 fsc->client->monc.fs_cluster_id = mount_fscid;
5195 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
5196 0, true);
5197 ceph_monc_renew_subs(&fsc->client->monc);
5198 } else {
5199 err = -ENOENT;
5200 goto err_out;
5201 }
5202 return;
5203
5204 bad:
5205 pr_err("error decoding fsmap %d. Shutting down mount.\n", err);
5206 ceph_umount_begin(mdsc->fsc->sb);
5207 err_out:
5208 mutex_lock(&mdsc->mutex);
5209 mdsc->mdsmap_err = err;
5210 __wake_requests(mdsc, &mdsc->waiting_for_map);
5211 mutex_unlock(&mdsc->mutex);
5212 }
5213
5214 /*
5215 * handle mds map update.
5216 */
ceph_mdsc_handle_mdsmap(struct ceph_mds_client * mdsc,struct ceph_msg * msg)5217 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
5218 {
5219 u32 epoch;
5220 u32 maplen;
5221 void *p = msg->front.iov_base;
5222 void *end = p + msg->front.iov_len;
5223 struct ceph_mdsmap *newmap, *oldmap;
5224 struct ceph_fsid fsid;
5225 int err = -EINVAL;
5226
5227 ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
5228 ceph_decode_copy(&p, &fsid, sizeof(fsid));
5229 if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
5230 return;
5231 epoch = ceph_decode_32(&p);
5232 maplen = ceph_decode_32(&p);
5233 dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
5234
5235 /* do we need it? */
5236 mutex_lock(&mdsc->mutex);
5237 if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
5238 dout("handle_map epoch %u <= our %u\n",
5239 epoch, mdsc->mdsmap->m_epoch);
5240 mutex_unlock(&mdsc->mutex);
5241 return;
5242 }
5243
5244 newmap = ceph_mdsmap_decode(&p, end, ceph_msgr2(mdsc->fsc->client));
5245 if (IS_ERR(newmap)) {
5246 err = PTR_ERR(newmap);
5247 goto bad_unlock;
5248 }
5249
5250 /* swap into place */
5251 if (mdsc->mdsmap) {
5252 oldmap = mdsc->mdsmap;
5253 mdsc->mdsmap = newmap;
5254 check_new_map(mdsc, newmap, oldmap);
5255 ceph_mdsmap_destroy(oldmap);
5256 } else {
5257 mdsc->mdsmap = newmap; /* first mds map */
5258 }
5259 mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
5260 MAX_LFS_FILESIZE);
5261
5262 __wake_requests(mdsc, &mdsc->waiting_for_map);
5263 ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
5264 mdsc->mdsmap->m_epoch);
5265
5266 mutex_unlock(&mdsc->mutex);
5267 schedule_delayed(mdsc, 0);
5268 return;
5269
5270 bad_unlock:
5271 mutex_unlock(&mdsc->mutex);
5272 bad:
5273 pr_err("error decoding mdsmap %d. Shutting down mount.\n", err);
5274 ceph_umount_begin(mdsc->fsc->sb);
5275 return;
5276 }
5277
mds_get_con(struct ceph_connection * con)5278 static struct ceph_connection *mds_get_con(struct ceph_connection *con)
5279 {
5280 struct ceph_mds_session *s = con->private;
5281
5282 if (ceph_get_mds_session(s))
5283 return con;
5284 return NULL;
5285 }
5286
mds_put_con(struct ceph_connection * con)5287 static void mds_put_con(struct ceph_connection *con)
5288 {
5289 struct ceph_mds_session *s = con->private;
5290
5291 ceph_put_mds_session(s);
5292 }
5293
5294 /*
5295 * if the client is unresponsive for long enough, the mds will kill
5296 * the session entirely.
5297 */
mds_peer_reset(struct ceph_connection * con)5298 static void mds_peer_reset(struct ceph_connection *con)
5299 {
5300 struct ceph_mds_session *s = con->private;
5301 struct ceph_mds_client *mdsc = s->s_mdsc;
5302
5303 pr_warn("mds%d closed our session\n", s->s_mds);
5304 send_mds_reconnect(mdsc, s);
5305 }
5306
mds_dispatch(struct ceph_connection * con,struct ceph_msg * msg)5307 static void mds_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5308 {
5309 struct ceph_mds_session *s = con->private;
5310 struct ceph_mds_client *mdsc = s->s_mdsc;
5311 int type = le16_to_cpu(msg->hdr.type);
5312
5313 mutex_lock(&mdsc->mutex);
5314 if (__verify_registered_session(mdsc, s) < 0) {
5315 mutex_unlock(&mdsc->mutex);
5316 goto out;
5317 }
5318 mutex_unlock(&mdsc->mutex);
5319
5320 switch (type) {
5321 case CEPH_MSG_MDS_MAP:
5322 ceph_mdsc_handle_mdsmap(mdsc, msg);
5323 break;
5324 case CEPH_MSG_FS_MAP_USER:
5325 ceph_mdsc_handle_fsmap(mdsc, msg);
5326 break;
5327 case CEPH_MSG_CLIENT_SESSION:
5328 handle_session(s, msg);
5329 break;
5330 case CEPH_MSG_CLIENT_REPLY:
5331 handle_reply(s, msg);
5332 break;
5333 case CEPH_MSG_CLIENT_REQUEST_FORWARD:
5334 handle_forward(mdsc, s, msg);
5335 break;
5336 case CEPH_MSG_CLIENT_CAPS:
5337 ceph_handle_caps(s, msg);
5338 break;
5339 case CEPH_MSG_CLIENT_SNAP:
5340 ceph_handle_snap(mdsc, s, msg);
5341 break;
5342 case CEPH_MSG_CLIENT_LEASE:
5343 handle_lease(mdsc, s, msg);
5344 break;
5345 case CEPH_MSG_CLIENT_QUOTA:
5346 ceph_handle_quota(mdsc, s, msg);
5347 break;
5348
5349 default:
5350 pr_err("received unknown message type %d %s\n", type,
5351 ceph_msg_type_name(type));
5352 }
5353 out:
5354 ceph_msg_put(msg);
5355 }
5356
5357 /*
5358 * authentication
5359 */
5360
5361 /*
5362 * Note: returned pointer is the address of a structure that's
5363 * managed separately. Caller must *not* attempt to free it.
5364 */
5365 static struct ceph_auth_handshake *
mds_get_authorizer(struct ceph_connection * con,int * proto,int force_new)5366 mds_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5367 {
5368 struct ceph_mds_session *s = con->private;
5369 struct ceph_mds_client *mdsc = s->s_mdsc;
5370 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5371 struct ceph_auth_handshake *auth = &s->s_auth;
5372 int ret;
5373
5374 ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5375 force_new, proto, NULL, NULL);
5376 if (ret)
5377 return ERR_PTR(ret);
5378
5379 return auth;
5380 }
5381
mds_add_authorizer_challenge(struct ceph_connection * con,void * challenge_buf,int challenge_buf_len)5382 static int mds_add_authorizer_challenge(struct ceph_connection *con,
5383 void *challenge_buf, int challenge_buf_len)
5384 {
5385 struct ceph_mds_session *s = con->private;
5386 struct ceph_mds_client *mdsc = s->s_mdsc;
5387 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5388
5389 return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
5390 challenge_buf, challenge_buf_len);
5391 }
5392
mds_verify_authorizer_reply(struct ceph_connection * con)5393 static int mds_verify_authorizer_reply(struct ceph_connection *con)
5394 {
5395 struct ceph_mds_session *s = con->private;
5396 struct ceph_mds_client *mdsc = s->s_mdsc;
5397 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5398 struct ceph_auth_handshake *auth = &s->s_auth;
5399
5400 return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5401 auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5402 NULL, NULL, NULL, NULL);
5403 }
5404
mds_invalidate_authorizer(struct ceph_connection * con)5405 static int mds_invalidate_authorizer(struct ceph_connection *con)
5406 {
5407 struct ceph_mds_session *s = con->private;
5408 struct ceph_mds_client *mdsc = s->s_mdsc;
5409 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5410
5411 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
5412
5413 return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
5414 }
5415
mds_get_auth_request(struct ceph_connection * con,void * buf,int * buf_len,void ** authorizer,int * authorizer_len)5416 static int mds_get_auth_request(struct ceph_connection *con,
5417 void *buf, int *buf_len,
5418 void **authorizer, int *authorizer_len)
5419 {
5420 struct ceph_mds_session *s = con->private;
5421 struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5422 struct ceph_auth_handshake *auth = &s->s_auth;
5423 int ret;
5424
5425 ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5426 buf, buf_len);
5427 if (ret)
5428 return ret;
5429
5430 *authorizer = auth->authorizer_buf;
5431 *authorizer_len = auth->authorizer_buf_len;
5432 return 0;
5433 }
5434
mds_handle_auth_reply_more(struct ceph_connection * con,void * reply,int reply_len,void * buf,int * buf_len,void ** authorizer,int * authorizer_len)5435 static int mds_handle_auth_reply_more(struct ceph_connection *con,
5436 void *reply, int reply_len,
5437 void *buf, int *buf_len,
5438 void **authorizer, int *authorizer_len)
5439 {
5440 struct ceph_mds_session *s = con->private;
5441 struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5442 struct ceph_auth_handshake *auth = &s->s_auth;
5443 int ret;
5444
5445 ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5446 buf, buf_len);
5447 if (ret)
5448 return ret;
5449
5450 *authorizer = auth->authorizer_buf;
5451 *authorizer_len = auth->authorizer_buf_len;
5452 return 0;
5453 }
5454
mds_handle_auth_done(struct ceph_connection * con,u64 global_id,void * reply,int reply_len,u8 * session_key,int * session_key_len,u8 * con_secret,int * con_secret_len)5455 static int mds_handle_auth_done(struct ceph_connection *con,
5456 u64 global_id, void *reply, int reply_len,
5457 u8 *session_key, int *session_key_len,
5458 u8 *con_secret, int *con_secret_len)
5459 {
5460 struct ceph_mds_session *s = con->private;
5461 struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5462 struct ceph_auth_handshake *auth = &s->s_auth;
5463
5464 return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5465 session_key, session_key_len,
5466 con_secret, con_secret_len);
5467 }
5468
mds_handle_auth_bad_method(struct ceph_connection * con,int used_proto,int result,const int * allowed_protos,int proto_cnt,const int * allowed_modes,int mode_cnt)5469 static int mds_handle_auth_bad_method(struct ceph_connection *con,
5470 int used_proto, int result,
5471 const int *allowed_protos, int proto_cnt,
5472 const int *allowed_modes, int mode_cnt)
5473 {
5474 struct ceph_mds_session *s = con->private;
5475 struct ceph_mon_client *monc = &s->s_mdsc->fsc->client->monc;
5476 int ret;
5477
5478 if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_MDS,
5479 used_proto, result,
5480 allowed_protos, proto_cnt,
5481 allowed_modes, mode_cnt)) {
5482 ret = ceph_monc_validate_auth(monc);
5483 if (ret)
5484 return ret;
5485 }
5486
5487 return -EACCES;
5488 }
5489
mds_alloc_msg(struct ceph_connection * con,struct ceph_msg_header * hdr,int * skip)5490 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
5491 struct ceph_msg_header *hdr, int *skip)
5492 {
5493 struct ceph_msg *msg;
5494 int type = (int) le16_to_cpu(hdr->type);
5495 int front_len = (int) le32_to_cpu(hdr->front_len);
5496
5497 if (con->in_msg)
5498 return con->in_msg;
5499
5500 *skip = 0;
5501 msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
5502 if (!msg) {
5503 pr_err("unable to allocate msg type %d len %d\n",
5504 type, front_len);
5505 return NULL;
5506 }
5507
5508 return msg;
5509 }
5510
mds_sign_message(struct ceph_msg * msg)5511 static int mds_sign_message(struct ceph_msg *msg)
5512 {
5513 struct ceph_mds_session *s = msg->con->private;
5514 struct ceph_auth_handshake *auth = &s->s_auth;
5515
5516 return ceph_auth_sign_message(auth, msg);
5517 }
5518
mds_check_message_signature(struct ceph_msg * msg)5519 static int mds_check_message_signature(struct ceph_msg *msg)
5520 {
5521 struct ceph_mds_session *s = msg->con->private;
5522 struct ceph_auth_handshake *auth = &s->s_auth;
5523
5524 return ceph_auth_check_message_signature(auth, msg);
5525 }
5526
5527 static const struct ceph_connection_operations mds_con_ops = {
5528 .get = mds_get_con,
5529 .put = mds_put_con,
5530 .alloc_msg = mds_alloc_msg,
5531 .dispatch = mds_dispatch,
5532 .peer_reset = mds_peer_reset,
5533 .get_authorizer = mds_get_authorizer,
5534 .add_authorizer_challenge = mds_add_authorizer_challenge,
5535 .verify_authorizer_reply = mds_verify_authorizer_reply,
5536 .invalidate_authorizer = mds_invalidate_authorizer,
5537 .sign_message = mds_sign_message,
5538 .check_message_signature = mds_check_message_signature,
5539 .get_auth_request = mds_get_auth_request,
5540 .handle_auth_reply_more = mds_handle_auth_reply_more,
5541 .handle_auth_done = mds_handle_auth_done,
5542 .handle_auth_bad_method = mds_handle_auth_bad_method,
5543 };
5544
5545 /* eof */
5546