]> Git Repo - linux.git/blame - fs/ceph/mds_client.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
[linux.git] / fs / ceph / mds_client.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
3d14c5d2 2#include <linux/ceph/ceph_debug.h>
2f2dc053 3
496e5955 4#include <linux/fs.h>
2f2dc053 5#include <linux/wait.h>
5a0e3ad6 6#include <linux/slab.h>
54008399 7#include <linux/gfp.h>
2f2dc053 8#include <linux/sched.h>
3d14c5d2
YS		 9#include <linux/debugfs.h>
10#include <linux/seq_file.h>
3e0708b9 11#include <linux/ratelimit.h>
9ba1e224 12#include <linux/bits.h>
70c94820 13#include <linux/ktime.h>
2f2dc053 14
2f2dc053 15#include "super.h"
3d14c5d2
YS		 16#include "mds_client.h"
17
1fe60e51 18#include <linux/ceph/ceph_features.h>
3d14c5d2
YS		 19#include <linux/ceph/messenger.h>
20#include <linux/ceph/decode.h>
21#include <linux/ceph/pagelist.h>
22#include <linux/ceph/auth.h>
23#include <linux/ceph/debugfs.h>
2f2dc053 24
81c5a148
YZ		 25#define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
26
2f2dc053
SW		 27/*
28 * A cluster of MDS (metadata server) daemons is responsible for
29 * managing the file system namespace (the directory hierarchy and
30 * inodes) and for coordinating shared access to storage. Metadata is
31 * partitioning hierarchically across a number of servers, and that
32 * partition varies over time as the cluster adjusts the distribution
33 * in order to balance load.
34 *
35 * The MDS client is primarily responsible to managing synchronous
36 * metadata requests for operations like open, unlink, and so forth.
37 * If there is a MDS failure, we find out about it when we (possibly
38 * request and) receive a new MDS map, and can resubmit affected
39 * requests.
40 *
41 * For the most part, though, we take advantage of a lossless
42 * communications channel to the MDS, and do not need to worry about
43 * timing out or resubmitting requests.
44 *
45 * We maintain a stateful "session" with each MDS we interact with.
46 * Within each session, we sent periodic heartbeat messages to ensure
47 * any capabilities or leases we have been issues remain valid. If
48 * the session times out and goes stale, our leases and capabilities
49 * are no longer valid.
50 */
51
20cb34ae 52struct ceph_reconnect_state {
81c5a148
YZ		 53 struct ceph_mds_session *session;
54 int nr_caps, nr_realms;
20cb34ae 55 struct ceph_pagelist *pagelist;
121f22a1 56 unsigned msg_version;
81c5a148 57 bool allow_multi;
20cb34ae
SW		 58};
59
2f2dc053
SW		 60static void __wake_requests(struct ceph_mds_client *mdsc,
61 struct list_head *head);
e3ec8d68 62static void ceph_cap_release_work(struct work_struct *work);
37c4efc1 63static void ceph_cap_reclaim_work(struct work_struct *work);
2f2dc053 64
9e32789f 65static const struct ceph_connection_operations mds_con_ops;
2f2dc053
SW		 66
67
68/*
69 * mds reply parsing
70 */
71
b37fe1f9
YZ		 72static int parse_reply_info_quota(void **p, void *end,
73 struct ceph_mds_reply_info_in *info)
74{
75 u8 struct_v, struct_compat;
76 u32 struct_len;
77
78 ceph_decode_8_safe(p, end, struct_v, bad);
79 ceph_decode_8_safe(p, end, struct_compat, bad);
80 /* struct_v is expected to be >= 1. we only
81 * understand encoding with struct_compat == 1. */
82 if (!struct_v || struct_compat != 1)
83 goto bad;
84 ceph_decode_32_safe(p, end, struct_len, bad);
85 ceph_decode_need(p, end, struct_len, bad);
86 end = *p + struct_len;
87 ceph_decode_64_safe(p, end, info->max_bytes, bad);
88 ceph_decode_64_safe(p, end, info->max_files, bad);
89 *p = end;
90 return 0;
91bad:
92 return -EIO;
93}
94
2f2dc053
SW		 95/*
96 * parse individual inode info
97 */
98static int parse_reply_info_in(void **p, void *end,
14303d20 99 struct ceph_mds_reply_info_in *info,
12b4629a 100 u64 features)
2f2dc053 101{
b37fe1f9
YZ		 102 int err = 0;
103 u8 struct_v = 0;
2f2dc053 104
b37fe1f9
YZ		 105 if (features == (u64)-1) {
106 u32 struct_len;
107 u8 struct_compat;
108 ceph_decode_8_safe(p, end, struct_v, bad);
109 ceph_decode_8_safe(p, end, struct_compat, bad);
110 /* struct_v is expected to be >= 1. we only understand
111 * encoding with struct_compat == 1. */
112 if (!struct_v || struct_compat != 1)
113 goto bad;
114 ceph_decode_32_safe(p, end, struct_len, bad);
115 ceph_decode_need(p, end, struct_len, bad);
116 end = *p + struct_len;
117 }
118
119 ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
2f2dc053
SW		 120 info->in = *p;
121 *p += sizeof(struct ceph_mds_reply_inode) +
122 sizeof(*info->in->fragtree.splits) *
123 le32_to_cpu(info->in->fragtree.nsplits);
124
125 ceph_decode_32_safe(p, end, info->symlink_len, bad);
126 ceph_decode_need(p, end, info->symlink_len, bad);
127 info->symlink = *p;
128 *p += info->symlink_len;
129
23c625ce
ID		 130 ceph_decode_copy_safe(p, end, &info->dir_layout,
131 sizeof(info->dir_layout), bad);
2f2dc053
SW		 132 ceph_decode_32_safe(p, end, info->xattr_len, bad);
133 ceph_decode_need(p, end, info->xattr_len, bad);
134 info->xattr_data = *p;
135 *p += info->xattr_len;
fb01d1f8 136
b37fe1f9
YZ		 137 if (features == (u64)-1) {
138 /* inline data */
fb01d1f8
YZ		 139 ceph_decode_64_safe(p, end, info->inline_version, bad);
140 ceph_decode_32_safe(p, end, info->inline_len, bad);
141 ceph_decode_need(p, end, info->inline_len, bad);
142 info->inline_data = *p;
143 *p += info->inline_len;
b37fe1f9
YZ		 144 /* quota */
145 err = parse_reply_info_quota(p, end, info);
146 if (err < 0)
147 goto out_bad;
148 /* pool namespace */
149 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
150 if (info->pool_ns_len > 0) {
151 ceph_decode_need(p, end, info->pool_ns_len, bad);
152 info->pool_ns_data = *p;
153 *p += info->pool_ns_len;
154 }
245ce991
JL		 155
156 /* btime */
157 ceph_decode_need(p, end, sizeof(info->btime), bad);
158 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
159
160 /* change attribute */
a35ead31 161 ceph_decode_64_safe(p, end, info->change_attr, bad);
fb01d1f8 162
08796873
YZ		 163 /* dir pin */
164 if (struct_v >= 2) {
165 ceph_decode_32_safe(p, end, info->dir_pin, bad);
166 } else {
167 info->dir_pin = -ENODATA;
168 }
169
193e7b37
DD		 170 /* snapshot birth time, remains zero for v<=2 */
171 if (struct_v >= 3) {
172 ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
173 ceph_decode_copy(p, &info->snap_btime,
174 sizeof(info->snap_btime));
175 } else {
176 memset(&info->snap_btime, 0, sizeof(info->snap_btime));
177 }
178
b37fe1f9
YZ		 179 *p = end;
180 } else {
181 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
182 ceph_decode_64_safe(p, end, info->inline_version, bad);
183 ceph_decode_32_safe(p, end, info->inline_len, bad);
184 ceph_decode_need(p, end, info->inline_len, bad);
185 info->inline_data = *p;
186 *p += info->inline_len;
187 } else
188 info->inline_version = CEPH_INLINE_NONE;
189
190 if (features & CEPH_FEATURE_MDS_QUOTA) {
191 err = parse_reply_info_quota(p, end, info);
192 if (err < 0)
193 goto out_bad;
194 } else {
195 info->max_bytes = 0;
196 info->max_files = 0;
197 }
198
199 info->pool_ns_len = 0;
200 info->pool_ns_data = NULL;
201 if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
202 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
203 if (info->pool_ns_len > 0) {
204 ceph_decode_need(p, end, info->pool_ns_len, bad);
205 info->pool_ns_data = *p;
206 *p += info->pool_ns_len;
207 }
208 }
08796873 209
245ce991
JL		 210 if (features & CEPH_FEATURE_FS_BTIME) {
211 ceph_decode_need(p, end, sizeof(info->btime), bad);
212 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
a35ead31 213 ceph_decode_64_safe(p, end, info->change_attr, bad);
245ce991
JL		 214 }
215
08796873 216 info->dir_pin = -ENODATA;
193e7b37 217 /* info->snap_btime remains zero */
b37fe1f9
YZ		 218 }
219 return 0;
220bad:
221 err = -EIO;
222out_bad:
223 return err;
224}
225
226static int parse_reply_info_dir(void **p, void *end,
227 struct ceph_mds_reply_dirfrag **dirfrag,
228 u64 features)
229{
230 if (features == (u64)-1) {
fb18a575
LH		 231 u8 struct_v, struct_compat;
232 u32 struct_len;
fb18a575
LH		 233 ceph_decode_8_safe(p, end, struct_v, bad);
234 ceph_decode_8_safe(p, end, struct_compat, bad);
b37fe1f9
YZ		 235 /* struct_v is expected to be >= 1. we only understand
236 * encoding whose struct_compat == 1. */
237 if (!struct_v || struct_compat != 1)
fb18a575
LH		 238 goto bad;
239 ceph_decode_32_safe(p, end, struct_len, bad);
240 ceph_decode_need(p, end, struct_len, bad);
b37fe1f9 241 end = *p + struct_len;
fb18a575
LH		 242 }
243
b37fe1f9
YZ		 244 ceph_decode_need(p, end, sizeof(**dirfrag), bad);
245 *dirfrag = *p;
246 *p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
247 if (unlikely(*p > end))
248 goto bad;
249 if (features == (u64)-1)
250 *p = end;
251 return 0;
252bad:
253 return -EIO;
254}
255
256static int parse_reply_info_lease(void **p, void *end,
257 struct ceph_mds_reply_lease **lease,
258 u64 features)
259{
260 if (features == (u64)-1) {
261 u8 struct_v, struct_compat;
262 u32 struct_len;
263 ceph_decode_8_safe(p, end, struct_v, bad);
264 ceph_decode_8_safe(p, end, struct_compat, bad);
265 /* struct_v is expected to be >= 1. we only understand
266 * encoding whose struct_compat == 1. */
267 if (!struct_v || struct_compat != 1)
268 goto bad;
269 ceph_decode_32_safe(p, end, struct_len, bad);
270 ceph_decode_need(p, end, struct_len, bad);
271 end = *p + struct_len;
5ea5c5e0
YZ		 272 }
273
b37fe1f9
YZ		 274 ceph_decode_need(p, end, sizeof(**lease), bad);
275 *lease = *p;
276 *p += sizeof(**lease);
277 if (features == (u64)-1)
278 *p = end;
2f2dc053
SW		 279 return 0;
280bad:
b37fe1f9 281 return -EIO;
2f2dc053
SW		 282}
283
284/*
285 * parse a normal reply, which may contain a (dir+)dentry and/or a
286 * target inode.
287 */
288static int parse_reply_info_trace(void **p, void *end,
14303d20 289 struct ceph_mds_reply_info_parsed *info,
12b4629a 290 u64 features)
2f2dc053
SW		 291{
292 int err;
293
294 if (info->head->is_dentry) {
14303d20 295 err = parse_reply_info_in(p, end, &info->diri, features);
2f2dc053
SW		 296 if (err < 0)
297 goto out_bad;
298
b37fe1f9
YZ		 299 err = parse_reply_info_dir(p, end, &info->dirfrag, features);
300 if (err < 0)
301 goto out_bad;
2f2dc053
SW		 302
303 ceph_decode_32_safe(p, end, info->dname_len, bad);
304 ceph_decode_need(p, end, info->dname_len, bad);
305 info->dname = *p;
306 *p += info->dname_len;
b37fe1f9
YZ		 307
308 err = parse_reply_info_lease(p, end, &info->dlease, features);
309 if (err < 0)
310 goto out_bad;
2f2dc053
SW		 311 }
312
313 if (info->head->is_target) {
14303d20 314 err = parse_reply_info_in(p, end, &info->targeti, features);
2f2dc053
SW		 315 if (err < 0)
316 goto out_bad;
317 }
318
319 if (unlikely(*p != end))
320 goto bad;
321 return 0;
322
323bad:
324 err = -EIO;
325out_bad:
326 pr_err("problem parsing mds trace %d\n", err);
327 return err;
328}
329
330/*
331 * parse readdir results
332 */
b37fe1f9 333static int parse_reply_info_readdir(void **p, void *end,
14303d20 334 struct ceph_mds_reply_info_parsed *info,
12b4629a 335 u64 features)
2f2dc053
SW		 336{
337 u32 num, i = 0;
338 int err;
339
b37fe1f9
YZ		 340 err = parse_reply_info_dir(p, end, &info->dir_dir, features);
341 if (err < 0)
342 goto out_bad;
2f2dc053
SW		 343
344 ceph_decode_need(p, end, sizeof(num) + 2, bad);
c89136ea 345 num = ceph_decode_32(p);
956d39d6
YZ		 346 {
347 u16 flags = ceph_decode_16(p);
348 info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
349 info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
f3c4ebe6 350 info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
79162547 351 info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
956d39d6 352 }
2f2dc053
SW		 353 if (num == 0)
354 goto done;
355
2a5beea3
YZ		 356 BUG_ON(!info->dir_entries);
357 if ((unsigned long)(info->dir_entries + num) >
358 (unsigned long)info->dir_entries + info->dir_buf_size) {
54008399
YZ		 359 pr_err("dir contents are larger than expected\n");
360 WARN_ON(1);
361 goto bad;
362 }
2f2dc053 363
54008399 364 info->dir_nr = num;
2f2dc053 365 while (num) {
2a5beea3 366 struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
2f2dc053 367 /* dentry */
b37fe1f9 368 ceph_decode_32_safe(p, end, rde->name_len, bad);
2a5beea3
YZ		 369 ceph_decode_need(p, end, rde->name_len, bad);
370 rde->name = *p;
371 *p += rde->name_len;
372 dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);
2f2dc053 373
b37fe1f9
YZ		 374 /* dentry lease */
375 err = parse_reply_info_lease(p, end, &rde->lease, features);
376 if (err)
377 goto out_bad;
2f2dc053 378 /* inode */
2a5beea3 379 err = parse_reply_info_in(p, end, &rde->inode, features);
2f2dc053
SW		 380 if (err < 0)
381 goto out_bad;
8974eebd
YZ		 382 /* ceph_readdir_prepopulate() will update it */
383 rde->offset = 0;
2f2dc053
SW		 384 i++;
385 num--;
386 }
387
388done:
1d3f8723
JL		 389 /* Skip over any unrecognized fields */
390 *p = end;
2f2dc053
SW		 391 return 0;
392
393bad:
394 err = -EIO;
395out_bad:
396 pr_err("problem parsing dir contents %d\n", err);
397 return err;
398}
399
25933abd
HS		 400/*
401 * parse fcntl F_GETLK results
402 */
403static int parse_reply_info_filelock(void **p, void *end,
14303d20 404 struct ceph_mds_reply_info_parsed *info,
12b4629a 405 u64 features)
25933abd
HS		 406{
407 if (*p + sizeof(*info->filelock_reply) > end)
408 goto bad;
409
410 info->filelock_reply = *p;
25933abd 411
1d3f8723
JL		 412 /* Skip over any unrecognized fields */
413 *p = end;
25933abd 414 return 0;
25933abd
HS		 415bad:
416 return -EIO;
417}
418
d4846487
JL		 419
420#if BITS_PER_LONG == 64
421
422#define DELEGATED_INO_AVAILABLE xa_mk_value(1)
423
424static int ceph_parse_deleg_inos(void **p, void *end,
425 struct ceph_mds_session *s)
426{
427 u32 sets;
428
429 ceph_decode_32_safe(p, end, sets, bad);
430 dout("got %u sets of delegated inodes\n", sets);
431 while (sets--) {
432 u64 start, len, ino;
433
434 ceph_decode_64_safe(p, end, start, bad);
435 ceph_decode_64_safe(p, end, len, bad);
436 while (len--) {
437 int err = xa_insert(&s->s_delegated_inos, ino = start++,
438 DELEGATED_INO_AVAILABLE,
439 GFP_KERNEL);
440 if (!err) {
441 dout("added delegated inode 0x%llx\n",
442 start - 1);
443 } else if (err == -EBUSY) {
444 pr_warn("ceph: MDS delegated inode 0x%llx more than once.\n",
445 start - 1);
446 } else {
447 return err;
448 }
449 }
450 }
451 return 0;
452bad:
453 return -EIO;
454}
455
456u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
457{
458 unsigned long ino;
459 void *val;
460
461 xa_for_each(&s->s_delegated_inos, ino, val) {
462 val = xa_erase(&s->s_delegated_inos, ino);
463 if (val == DELEGATED_INO_AVAILABLE)
464 return ino;
465 }
466 return 0;
467}
468
469int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
470{
471 return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,
472 GFP_KERNEL);
473}
474#else /* BITS_PER_LONG == 64 */
475/*
476 * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
477 * ignore delegated_inos on 32 bit arch. Maybe eventually add xarrays for top
478 * and bottom words?
479 */
480static int ceph_parse_deleg_inos(void **p, void *end,
481 struct ceph_mds_session *s)
482{
483 u32 sets;
484
485 ceph_decode_32_safe(p, end, sets, bad);
486 if (sets)
487 ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);
488 return 0;
489bad:
490 return -EIO;
491}
492
493u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
494{
495 return 0;
496}
497
498int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
499{
500 return 0;
501}
502#endif /* BITS_PER_LONG == 64 */
503
6e8575fa
SL		 504/*
505 * parse create results
506 */
507static int parse_reply_info_create(void **p, void *end,
508 struct ceph_mds_reply_info_parsed *info,
d4846487 509 u64 features, struct ceph_mds_session *s)
6e8575fa 510{
d4846487
JL		 511 int ret;
512
b37fe1f9
YZ		 513 if (features == (u64)-1 ||
514 (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
6e8575fa 515 if (*p == end) {
d4846487 516 /* Malformed reply? */
6e8575fa 517 info->has_create_ino = false;
d4846487
JL		 518 } else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {
519 u8 struct_v, struct_compat;
520 u32 len;
521
6e8575fa 522 info->has_create_ino = true;
d4846487
JL		 523 ceph_decode_8_safe(p, end, struct_v, bad);
524 ceph_decode_8_safe(p, end, struct_compat, bad);
525 ceph_decode_32_safe(p, end, len, bad);
526 ceph_decode_64_safe(p, end, info->ino, bad);
527 ret = ceph_parse_deleg_inos(p, end, s);
528 if (ret)
529 return ret;
530 } else {
531 /* legacy */
1d3f8723 532 ceph_decode_64_safe(p, end, info->ino, bad);
d4846487 533 info->has_create_ino = true;
6e8575fa 534 }
1d3f8723
JL		 535 } else {
536 if (*p != end)
537 goto bad;
6e8575fa
SL		 538 }
539
1d3f8723
JL		 540 /* Skip over any unrecognized fields */
541 *p = end;
6e8575fa 542 return 0;
6e8575fa
SL		 543bad:
544 return -EIO;
545}
546
25933abd
HS		 547/*
548 * parse extra results
549 */
550static int parse_reply_info_extra(void **p, void *end,
14303d20 551 struct ceph_mds_reply_info_parsed *info,
d4846487 552 u64 features, struct ceph_mds_session *s)
25933abd 553{
6df8c9d8
JL		 554 u32 op = le32_to_cpu(info->head->op);
555
556 if (op == CEPH_MDS_OP_GETFILELOCK)
14303d20 557 return parse_reply_info_filelock(p, end, info, features);
6df8c9d8 558 else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
b37fe1f9 559 return parse_reply_info_readdir(p, end, info, features);
6df8c9d8 560 else if (op == CEPH_MDS_OP_CREATE)
d4846487 561 return parse_reply_info_create(p, end, info, features, s);
6e8575fa
SL		 562 else
563 return -EIO;
25933abd
HS		 564}
565
2f2dc053
SW		 566/*
567 * parse entire mds reply
568 */
d4846487 569static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg,
14303d20 570 struct ceph_mds_reply_info_parsed *info,
12b4629a 571 u64 features)
2f2dc053
SW		 572{
573 void *p, *end;
574 u32 len;
575 int err;
576
577 info->head = msg->front.iov_base;
578 p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
579 end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
580
581 /* trace */
582 ceph_decode_32_safe(&p, end, len, bad);
583 if (len > 0) {
32852a81 584 ceph_decode_need(&p, end, len, bad);
14303d20 585 err = parse_reply_info_trace(&p, p+len, info, features);
2f2dc053
SW		 586 if (err < 0)
587 goto out_bad;
588 }
589
25933abd 590 /* extra */
2f2dc053
SW		 591 ceph_decode_32_safe(&p, end, len, bad);
592 if (len > 0) {
32852a81 593 ceph_decode_need(&p, end, len, bad);
d4846487 594 err = parse_reply_info_extra(&p, p+len, info, features, s);
2f2dc053
SW		 595 if (err < 0)
596 goto out_bad;
597 }
598
599 /* snap blob */
600 ceph_decode_32_safe(&p, end, len, bad);
601 info->snapblob_len = len;
602 info->snapblob = p;
603 p += len;
604
605 if (p != end)
606 goto bad;
607 return 0;
608
609bad:
610 err = -EIO;
611out_bad:
612 pr_err("mds parse_reply err %d\n", err);
613 return err;
614}
615
616static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
617{
2a5beea3 618 if (!info->dir_entries)
54008399 619 return;
2a5beea3 620 free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
2f2dc053
SW		 621}
622
623
624/*
625 * sessions
626 */
a687ecaf 627const char *ceph_session_state_name(int s)
2f2dc053
SW		 628{
629 switch (s) {
630 case CEPH_MDS_SESSION_NEW: return "new";
631 case CEPH_MDS_SESSION_OPENING: return "opening";
632 case CEPH_MDS_SESSION_OPEN: return "open";
633 case CEPH_MDS_SESSION_HUNG: return "hung";
634 case CEPH_MDS_SESSION_CLOSING: return "closing";
4d681c2f 635 case CEPH_MDS_SESSION_CLOSED: return "closed";
44ca18f2 636 case CEPH_MDS_SESSION_RESTARTING: return "restarting";
2f2dc053 637 case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
fcff415c 638 case CEPH_MDS_SESSION_REJECTED: return "rejected";
2f2dc053
SW		 639 default: return "???";
640 }
641}
642
5b3248c6 643struct ceph_mds_session *ceph_get_mds_session(struct ceph_mds_session *s)
2f2dc053 644{
3997c01d 645 if (refcount_inc_not_zero(&s->s_ref)) {
2f2dc053 646 dout("mdsc get_session %p %d -> %d\n", s,
3997c01d 647 refcount_read(&s->s_ref)-1, refcount_read(&s->s_ref));
2f2dc053
SW		 648 return s;
649 } else {
4c069a58 650 dout("mdsc get_session %p 0 -- FAIL\n", s);
2f2dc053
SW		 651 return NULL;
652 }
653}
654
655void ceph_put_mds_session(struct ceph_mds_session *s)
656{
657 dout("mdsc put_session %p %d -> %d\n", s,
3997c01d
ER		 658 refcount_read(&s->s_ref), refcount_read(&s->s_ref)-1);
659 if (refcount_dec_and_test(&s->s_ref)) {
6c4a1915 660 if (s->s_auth.authorizer)
6c1ea260 661 ceph_auth_destroy_authorizer(s->s_auth.authorizer);
88828190 662 WARN_ON(mutex_is_locked(&s->s_mutex));
d4846487 663 xa_destroy(&s->s_delegated_inos);
2f2dc053 664 kfree(s);
4e7a5dcd 665 }
2f2dc053
SW		 666}
667
668/*
669 * called under mdsc->mutex
670 */
671struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
672 int mds)
673{
d37b1d99 674 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
2f2dc053 675 return NULL;
5b3248c6 676 return ceph_get_mds_session(mdsc->sessions[mds]);
2f2dc053
SW		 677}
678
679static bool __have_session(struct ceph_mds_client *mdsc, int mds)
680{
98cfda81 681 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
2f2dc053 682 return false;
98cfda81
CX		 683 else
684 return true;
2f2dc053
SW		 685}
686
2600d2dd
SW		 687static int __verify_registered_session(struct ceph_mds_client *mdsc,
688 struct ceph_mds_session *s)
689{
690 if (s->s_mds >= mdsc->max_sessions ||
691 mdsc->sessions[s->s_mds] != s)
692 return -ENOENT;
693 return 0;
694}
695
2f2dc053
SW		 696/*
697 * create+register a new session for given mds.
698 * called under mdsc->mutex.
699 */
700static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
701 int mds)
702{
703 struct ceph_mds_session *s;
704
b38c9eb4 705 if (mds >= mdsc->mdsmap->possible_max_rank)
c338c07c
NY		 706 return ERR_PTR(-EINVAL);
707
2f2dc053 708 s = kzalloc(sizeof(*s), GFP_NOFS);
4736b009
DC		 709 if (!s)
710 return ERR_PTR(-ENOMEM);
47474d0b
CX		 711
712 if (mds >= mdsc->max_sessions) {
713 int newmax = 1 << get_count_order(mds + 1);
714 struct ceph_mds_session **sa;
715
716 dout("%s: realloc to %d\n", __func__, newmax);
717 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
718 if (!sa)
719 goto fail_realloc;
720 if (mdsc->sessions) {
721 memcpy(sa, mdsc->sessions,
722 mdsc->max_sessions * sizeof(void *));
723 kfree(mdsc->sessions);
724 }
725 mdsc->sessions = sa;
726 mdsc->max_sessions = newmax;
727 }
728
729 dout("%s: mds%d\n", __func__, mds);
2f2dc053
SW		 730 s->s_mdsc = mdsc;
731 s->s_mds = mds;
732 s->s_state = CEPH_MDS_SESSION_NEW;
733 s->s_ttl = 0;
734 s->s_seq = 0;
735 mutex_init(&s->s_mutex);
736
b7a9e5dd 737 ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
2f2dc053 738
d8fb02ab 739 spin_lock_init(&s->s_gen_ttl_lock);
1e9c2eb6 740 s->s_cap_gen = 1;
1ce208a6 741 s->s_cap_ttl = jiffies - 1;
d8fb02ab
AE		 742
743 spin_lock_init(&s->s_cap_lock);
2f2dc053
SW		 744 s->s_renew_requested = 0;
745 s->s_renew_seq = 0;
746 INIT_LIST_HEAD(&s->s_caps);
747 s->s_nr_caps = 0;
3997c01d 748 refcount_set(&s->s_ref, 1);
2f2dc053
SW		 749 INIT_LIST_HEAD(&s->s_waiting);
750 INIT_LIST_HEAD(&s->s_unsafe);
d4846487 751 xa_init(&s->s_delegated_inos);
2f2dc053 752 s->s_num_cap_releases = 0;
99a9c273 753 s->s_cap_reconnect = 0;
7c1332b8 754 s->s_cap_iterator = NULL;
2f2dc053 755 INIT_LIST_HEAD(&s->s_cap_releases);
e3ec8d68
YZ		 756 INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
757
1cf03a68 758 INIT_LIST_HEAD(&s->s_cap_dirty);
2f2dc053 759 INIT_LIST_HEAD(&s->s_cap_flushing);
2f2dc053 760
2f2dc053 761 mdsc->sessions[mds] = s;
86d8f67b 762 atomic_inc(&mdsc->num_sessions);
3997c01d 763 refcount_inc(&s->s_ref); /* one ref to sessions[], one to caller */
42ce56e5 764
b7a9e5dd
SW		 765 ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
766 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
42ce56e5 767
2f2dc053 768 return s;
42ce56e5
SW		 769
770fail_realloc:
771 kfree(s);
772 return ERR_PTR(-ENOMEM);
2f2dc053
SW		 773}
774
775/*
776 * called under mdsc->mutex
777 */
2600d2dd 778static void __unregister_session(struct ceph_mds_client *mdsc,
42ce56e5 779 struct ceph_mds_session *s)
2f2dc053 780{
2600d2dd
SW		 781 dout("__unregister_session mds%d %p\n", s->s_mds, s);
782 BUG_ON(mdsc->sessions[s->s_mds] != s);
42ce56e5
SW		 783 mdsc->sessions[s->s_mds] = NULL;
784 ceph_con_close(&s->s_con);
785 ceph_put_mds_session(s);
86d8f67b 786 atomic_dec(&mdsc->num_sessions);
2f2dc053
SW		 787}
788
789/*
790 * drop session refs in request.
791 *
792 * should be last request ref, or hold mdsc->mutex
793 */
794static void put_request_session(struct ceph_mds_request *req)
795{
796 if (req->r_session) {
797 ceph_put_mds_session(req->r_session);
798 req->r_session = NULL;
799 }
800}
801
153c8e6b 802void ceph_mdsc_release_request(struct kref *kref)
2f2dc053 803{
153c8e6b
SW		 804 struct ceph_mds_request *req = container_of(kref,
805 struct ceph_mds_request,
806 r_kref);
e64f44a8 807 ceph_mdsc_release_dir_caps_no_check(req);
54008399 808 destroy_reply_info(&req->r_reply_info);
153c8e6b
SW		 809 if (req->r_request)
810 ceph_msg_put(req->r_request);
54008399 811 if (req->r_reply)
153c8e6b 812 ceph_msg_put(req->r_reply);
153c8e6b 813 if (req->r_inode) {
41b02e1f 814 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
3e1d0452
YZ		 815 /* avoid calling iput_final() in mds dispatch threads */
816 ceph_async_iput(req->r_inode);
153c8e6b 817 }
9c1c2b35 818 if (req->r_parent) {
3dd69aab 819 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
9c1c2b35
JL		 820 ceph_async_iput(req->r_parent);
821 }
3e1d0452 822 ceph_async_iput(req->r_target_inode);
153c8e6b
SW		 823 if (req->r_dentry)
824 dput(req->r_dentry);
844d87c3
SW		 825 if (req->r_old_dentry)
826 dput(req->r_old_dentry);
827 if (req->r_old_dentry_dir) {
41b02e1f
SW		 828 /*
829 * track (and drop pins for) r_old_dentry_dir
830 * separately, since r_old_dentry's d_parent may have
831 * changed between the dir mutex being dropped and
832 * this request being freed.
833 */
834 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
835 CEPH_CAP_PIN);
3e1d0452 836 ceph_async_iput(req->r_old_dentry_dir);
2f2dc053 837 }
153c8e6b
SW		 838 kfree(req->r_path1);
839 kfree(req->r_path2);
25e6bae3
YZ		 840 if (req->r_pagelist)
841 ceph_pagelist_release(req->r_pagelist);
153c8e6b 842 put_request_session(req);
37151668 843 ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
428138c9 844 WARN_ON_ONCE(!list_empty(&req->r_wait));
058daab7 845 kmem_cache_free(ceph_mds_request_cachep, req);
2f2dc053
SW		 846}
847
fcd00b68
ID		 848DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
849
2f2dc053
SW		 850/*
851 * lookup session, bump ref if found.
852 *
853 * called under mdsc->mutex.
854 */
fcd00b68
ID		 855static struct ceph_mds_request *
856lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
2f2dc053
SW		 857{
858 struct ceph_mds_request *req;
44ca18f2 859
fcd00b68
ID		 860 req = lookup_request(&mdsc->request_tree, tid);
861 if (req)
862 ceph_mdsc_get_request(req);
44ca18f2 863
fcd00b68 864 return req;
2f2dc053
SW		 865}
866
867/*
868 * Register an in-flight request, and assign a tid. Link to directory
869 * are modifying (if any).
870 *
871 * Called under mdsc->mutex.
872 */
873static void __register_request(struct ceph_mds_client *mdsc,
874 struct ceph_mds_request *req,
875 struct inode *dir)
876{
e30ee581
ZZ		 877 int ret = 0;
878
2f2dc053 879 req->r_tid = ++mdsc->last_tid;
e30ee581
ZZ		 880 if (req->r_num_caps) {
881 ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
882 req->r_num_caps);
883 if (ret < 0) {
884 pr_err("__register_request %p "
885 "failed to reserve caps: %d\n", req, ret);
886 /* set req->r_err to fail early from __do_request */
887 req->r_err = ret;
888 return;
889 }
890 }
2f2dc053
SW		 891 dout("__register_request %p tid %lld\n", req, req->r_tid);
892 ceph_mdsc_get_request(req);
fcd00b68 893 insert_request(&mdsc->request_tree, req);
2f2dc053 894
cb4276cc
SW		 895 req->r_uid = current_fsuid();
896 req->r_gid = current_fsgid();
897
e8a7b8b1
YZ		 898 if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
899 mdsc->oldest_tid = req->r_tid;
900
2f2dc053 901 if (dir) {
3db0a2fc
JL		 902 struct ceph_inode_info *ci = ceph_inode(dir);
903
3b663780 904 ihold(dir);
2f2dc053 905 req->r_unsafe_dir = dir;
3db0a2fc
JL		 906 spin_lock(&ci->i_unsafe_lock);
907 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
908 spin_unlock(&ci->i_unsafe_lock);
2f2dc053
SW		 909 }
910}
911
912static void __unregister_request(struct ceph_mds_client *mdsc,
913 struct ceph_mds_request *req)
914{
915 dout("__unregister_request %p tid %lld\n", req, req->r_tid);
e8a7b8b1 916
df963ea8
JL		 917 /* Never leave an unregistered request on an unsafe list! */
918 list_del_init(&req->r_unsafe_item);
919
e8a7b8b1
YZ		 920 if (req->r_tid == mdsc->oldest_tid) {
921 struct rb_node *p = rb_next(&req->r_node);
922 mdsc->oldest_tid = 0;
923 while (p) {
924 struct ceph_mds_request *next_req =
925 rb_entry(p, struct ceph_mds_request, r_node);
926 if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
927 mdsc->oldest_tid = next_req->r_tid;
928 break;
929 }
930 p = rb_next(p);
931 }
932 }
933
fcd00b68 934 erase_request(&mdsc->request_tree, req);
2f2dc053 935
3db0a2fc 936 if (req->r_unsafe_dir) {
2f2dc053 937 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
2f2dc053
SW		 938 spin_lock(&ci->i_unsafe_lock);
939 list_del_init(&req->r_unsafe_dir_item);
940 spin_unlock(&ci->i_unsafe_lock);
4c06ace8 941 }
bc2de10d
JL		 942 if (req->r_target_inode &&
943 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
68cd5b4b
YZ		 944 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
945 spin_lock(&ci->i_unsafe_lock);
946 list_del_init(&req->r_unsafe_target_item);
947 spin_unlock(&ci->i_unsafe_lock);
948 }
3b663780 949
4c06ace8 950 if (req->r_unsafe_dir) {
3e1d0452
YZ		 951 /* avoid calling iput_final() in mds dispatch threads */
952 ceph_async_iput(req->r_unsafe_dir);
3b663780 953 req->r_unsafe_dir = NULL;
2f2dc053 954 }
94aa8ae1 955
fc55d2c9
YZ		 956 complete_all(&req->r_safe_completion);
957
94aa8ae1 958 ceph_mdsc_put_request(req);
2f2dc053
SW		 959}
960
30c71233
JL		 961/*
962 * Walk back up the dentry tree until we hit a dentry representing a
963 * non-snapshot inode. We do this using the rcu_read_lock (which must be held
964 * when calling this) to ensure that the objects won't disappear while we're
965 * working with them. Once we hit a candidate dentry, we attempt to take a
966 * reference to it, and return that as the result.
967 */
f1075480
DC		 968static struct inode *get_nonsnap_parent(struct dentry *dentry)
969{
970 struct inode *inode = NULL;
30c71233
JL		 971
972 while (dentry && !IS_ROOT(dentry)) {
973 inode = d_inode_rcu(dentry);
974 if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
975 break;
976 dentry = dentry->d_parent;
977 }
978 if (inode)
979 inode = igrab(inode);
980 return inode;
981}
982
2f2dc053
SW		 983/*
984 * Choose mds to send request to next. If there is a hint set in the
985 * request (e.g., due to a prior forward hint from the mds), use that.
986 * Otherwise, consult frag tree and/or caps to identify the
987 * appropriate mds. If all else fails, choose randomly.
988 *
989 * Called under mdsc->mutex.
990 */
991static int __choose_mds(struct ceph_mds_client *mdsc,
c4853e97
XL		 992 struct ceph_mds_request *req,
993 bool *random)
2f2dc053
SW		 994{
995 struct inode *inode;
996 struct ceph_inode_info *ci;
997 struct ceph_cap *cap;
998 int mode = req->r_direct_mode;
999 int mds = -1;
1000 u32 hash = req->r_direct_hash;
bc2de10d 1001 bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
2f2dc053 1002
c4853e97
XL		 1003 if (random)
1004 *random = false;
1005
2f2dc053
SW		 1006 /*
1007 * is there a specific mds we should try? ignore hint if we have
1008 * no session and the mds is not up (active or recovering).
1009 */
1010 if (req->r_resend_mds >= 0 &&
1011 (__have_session(mdsc, req->r_resend_mds) ||
1012 ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
3c802092 1013 dout("%s using resend_mds mds%d\n", __func__,
2f2dc053
SW		 1014 req->r_resend_mds);
1015 return req->r_resend_mds;
1016 }
1017
1018 if (mode == USE_RANDOM_MDS)
1019 goto random;
1020
1021 inode = NULL;
1022 if (req->r_inode) {
5d37ca14
YZ		 1023 if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
1024 inode = req->r_inode;
1025 ihold(inode);
1026 } else {
38f340cc
YZ		 1027 /* req->r_dentry is non-null for LSSNAP request */
1028 rcu_read_lock();
1029 inode = get_nonsnap_parent(req->r_dentry);
1030 rcu_read_unlock();
3c802092 1031 dout("%s using snapdir's parent %p\n", __func__, inode);
5d37ca14 1032 }
38f340cc 1033 } else if (req->r_dentry) {
d79698da 1034 /* ignore race with rename; old or new d_parent is okay */
30c71233
JL		 1035 struct dentry *parent;
1036 struct inode *dir;
1037
1038 rcu_read_lock();
41883ba8 1039 parent = READ_ONCE(req->r_dentry->d_parent);
3dd69aab 1040 dir = req->r_parent ? : d_inode_rcu(parent);
eb6bb1c5 1041
30c71233
JL		 1042 if (!dir || dir->i_sb != mdsc->fsc->sb) {
1043 /* not this fs or parent went negative */
2b0143b5 1044 inode = d_inode(req->r_dentry);
30c71233
JL		 1045 if (inode)
1046 ihold(inode);
eb6bb1c5
SW		 1047 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
1048 /* direct snapped/virtual snapdir requests
1049 * based on parent dir inode */
30c71233 1050 inode = get_nonsnap_parent(parent);
3c802092 1051 dout("%s using nonsnap parent %p\n", __func__, inode);
ca18bede 1052 } else {
eb6bb1c5 1053 /* dentry target */
2b0143b5 1054 inode = d_inode(req->r_dentry);
ca18bede
YZ		 1055 if (!inode || mode == USE_AUTH_MDS) {
1056 /* dir + name */
30c71233 1057 inode = igrab(dir);
ca18bede
YZ		 1058 hash = ceph_dentry_hash(dir, req->r_dentry);
1059 is_hash = true;
30c71233
JL		 1060 } else {
1061 ihold(inode);
ca18bede 1062 }
2f2dc053 1063 }
30c71233 1064 rcu_read_unlock();
2f2dc053 1065 }
eb6bb1c5 1066
3c802092
XL		 1067 dout("%s %p is_hash=%d (0x%x) mode %d\n", __func__, inode, (int)is_hash,
1068 hash, mode);
2f2dc053
SW		 1069 if (!inode)
1070 goto random;
1071 ci = ceph_inode(inode);
1072
1073 if (is_hash && S_ISDIR(inode->i_mode)) {
1074 struct ceph_inode_frag frag;
1075 int found;
1076
1077 ceph_choose_frag(ci, hash, &frag, &found);
1078 if (found) {
1079 if (mode == USE_ANY_MDS && frag.ndist > 0) {
1080 u8 r;
1081
1082 /* choose a random replica */
1083 get_random_bytes(&r, 1);
1084 r %= frag.ndist;
1085 mds = frag.dist[r];
3c802092
XL		 1086 dout("%s %p %llx.%llx frag %u mds%d (%d/%d)\n",
1087 __func__, inode, ceph_vinop(inode),
1088 frag.frag, mds, (int)r, frag.ndist);
d66bbd44 1089 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
5d47648f
XL		 1090 CEPH_MDS_STATE_ACTIVE &&
1091 !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))
30c71233 1092 goto out;
2f2dc053
SW		 1093 }
1094
1095 /* since this file/dir wasn't known to be
1096 * replicated, then we want to look for the
1097 * authoritative mds. */
2f2dc053
SW		 1098 if (frag.mds >= 0) {
1099 /* choose auth mds */
1100 mds = frag.mds;
3c802092
XL		 1101 dout("%s %p %llx.%llx frag %u mds%d (auth)\n",
1102 __func__, inode, ceph_vinop(inode),
1103 frag.frag, mds);
d66bbd44 1104 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
5d47648f 1105 CEPH_MDS_STATE_ACTIVE) {
224c7b67 1106 if (!ceph_mdsmap_is_laggy(mdsc->mdsmap,
5d47648f
XL		 1107 mds))
1108 goto out;
1109 }
2f2dc053 1110 }
5d47648f 1111 mode = USE_AUTH_MDS;
2f2dc053
SW		 1112 }
1113 }
1114
be655596 1115 spin_lock(&ci->i_ceph_lock);
2f2dc053
SW		 1116 cap = NULL;
1117 if (mode == USE_AUTH_MDS)
1118 cap = ci->i_auth_cap;
1119 if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1120 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1121 if (!cap) {
be655596 1122 spin_unlock(&ci->i_ceph_lock);
3e1d0452 1123 ceph_async_iput(inode);
2f2dc053
SW		 1124 goto random;
1125 }
1126 mds = cap->session->s_mds;
3c802092 1127 dout("%s %p %llx.%llx mds%d (%scap %p)\n", __func__,
2f2dc053
SW		 1128 inode, ceph_vinop(inode), mds,
1129 cap == ci->i_auth_cap ? "auth " : "", cap);
be655596 1130 spin_unlock(&ci->i_ceph_lock);
30c71233 1131out:
3e1d0452
YZ		 1132 /* avoid calling iput_final() while holding mdsc->mutex or
1133 * in mds dispatch threads */
1134 ceph_async_iput(inode);
2f2dc053
SW		 1135 return mds;
1136
1137random:
c4853e97
XL		 1138 if (random)
1139 *random = true;
1140
2f2dc053 1141 mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
3c802092 1142 dout("%s chose random mds%d\n", __func__, mds);
2f2dc053
SW		 1143 return mds;
1144}
1145
1146
1147/*
1148 * session messages
1149 */
1150static struct ceph_msg *create_session_msg(u32 op, u64 seq)
1151{
1152 struct ceph_msg *msg;
1153 struct ceph_mds_session_head *h;
1154
b61c2763
SW		 1155 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1156 false);
a79832f2 1157 if (!msg) {
2f2dc053 1158 pr_err("create_session_msg ENOMEM creating msg\n");
a79832f2 1159 return NULL;
2f2dc053
SW		 1160 }
1161 h = msg->front.iov_base;
1162 h->op = cpu_to_le32(op);
1163 h->seq = cpu_to_le64(seq);
dbd0c8bf
JS		 1164
1165 return msg;
1166}
1167
9ba1e224
XL		 1168static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1169#define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
b682c6d4 1170static int encode_supported_features(void **p, void *end)
342ce182 1171{
9ba1e224 1172 static const size_t count = ARRAY_SIZE(feature_bits);
342ce182
YZ		 1173
1174 if (count > 0) {
1175 size_t i;
9ba1e224 1176 size_t size = FEATURE_BYTES(count);
342ce182 1177
b682c6d4
XL		 1178 if (WARN_ON_ONCE(*p + 4 + size > end))
1179 return -ERANGE;
1180
342ce182
YZ		 1181 ceph_encode_32(p, size);
1182 memset(*p, 0, size);
1183 for (i = 0; i < count; i++)
9ba1e224 1184 ((unsigned char*)(*p))[i / 8] |= BIT(feature_bits[i] % 8);
342ce182
YZ		 1185 *p += size;
1186 } else {
b682c6d4
XL		 1187 if (WARN_ON_ONCE(*p + 4 > end))
1188 return -ERANGE;
1189
342ce182
YZ		 1190 ceph_encode_32(p, 0);
1191 }
b682c6d4
XL		 1192
1193 return 0;
342ce182
YZ		 1194}
1195
3b4168dd
XL		 1196static const unsigned char metric_bits[] = CEPHFS_METRIC_SPEC_CLIENT_SUPPORTED;
1197#define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
1198static int encode_metric_spec(void **p, void *end)
1199{
1200 static const size_t count = ARRAY_SIZE(metric_bits);
1201
1202 /* header */
1203 if (WARN_ON_ONCE(*p + 2 > end))
1204 return -ERANGE;
1205
1206 ceph_encode_8(p, 1); /* version */
1207 ceph_encode_8(p, 1); /* compat */
1208
1209 if (count > 0) {
1210 size_t i;
1211 size_t size = METRIC_BYTES(count);
1212
1213 if (WARN_ON_ONCE(*p + 4 + 4 + size > end))
1214 return -ERANGE;
1215
1216 /* metric spec info length */
1217 ceph_encode_32(p, 4 + size);
1218
1219 /* metric spec */
1220 ceph_encode_32(p, size);
1221 memset(*p, 0, size);
1222 for (i = 0; i < count; i++)
1223 ((unsigned char *)(*p))[i / 8] |= BIT(metric_bits[i] % 8);
1224 *p += size;
1225 } else {
1226 if (WARN_ON_ONCE(*p + 4 + 4 > end))
1227 return -ERANGE;
1228
1229 /* metric spec info length */
1230 ceph_encode_32(p, 4);
1231 /* metric spec */
1232 ceph_encode_32(p, 0);
1233 }
1234
1235 return 0;
1236}
1237
dbd0c8bf
JS		 1238/*
1239 * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1240 * to include additional client metadata fields.
1241 */
1242static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1243{
1244 struct ceph_msg *msg;
1245 struct ceph_mds_session_head *h;
1246 int i = -1;
342ce182 1247 int extra_bytes = 0;
dbd0c8bf
JS		 1248 int metadata_key_count = 0;
1249 struct ceph_options *opt = mdsc->fsc->client->options;
3f384954 1250 struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
9ba1e224 1251 size_t size, count;
342ce182 1252 void *p, *end;
b682c6d4 1253 int ret;
dbd0c8bf 1254
a6a5ce4f 1255 const char* metadata[][2] = {
717e6f28
YZ		 1256 {"hostname", mdsc->nodename},
1257 {"kernel_version", init_utsname()->release},
3f384954
YZ		 1258 {"entity_id", opt->name ? : ""},
1259 {"root", fsopt->server_path ? : "/"},
dbd0c8bf
JS		 1260 {NULL, NULL}
1261 };
1262
1263 /* Calculate serialized length of metadata */
342ce182 1264 extra_bytes = 4; /* map length */
d37b1d99 1265 for (i = 0; metadata[i][0]; ++i) {
342ce182 1266 extra_bytes += 8 + strlen(metadata[i][0]) +
dbd0c8bf
JS		 1267 strlen(metadata[i][1]);
1268 metadata_key_count++;
1269 }
9ba1e224 1270
342ce182 1271 /* supported feature */
9ba1e224
XL		 1272 size = 0;
1273 count = ARRAY_SIZE(feature_bits);
1274 if (count > 0)
1275 size = FEATURE_BYTES(count);
1276 extra_bytes += 4 + size;
dbd0c8bf 1277
3b4168dd
XL		 1278 /* metric spec */
1279 size = 0;
1280 count = ARRAY_SIZE(metric_bits);
1281 if (count > 0)
1282 size = METRIC_BYTES(count);
1283 extra_bytes += 2 + 4 + 4 + size;
1284
dbd0c8bf 1285 /* Allocate the message */
342ce182 1286 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
dbd0c8bf
JS		 1287 GFP_NOFS, false);
1288 if (!msg) {
1289 pr_err("create_session_msg ENOMEM creating msg\n");
b682c6d4 1290 return ERR_PTR(-ENOMEM);
dbd0c8bf 1291 }
342ce182
YZ		 1292 p = msg->front.iov_base;
1293 end = p + msg->front.iov_len;
1294
1295 h = p;
dbd0c8bf
JS		 1296 h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1297 h->seq = cpu_to_le64(seq);
1298
1299 /*
1300 * Serialize client metadata into waiting buffer space, using
1301 * the format that userspace expects for map<string, string>
7cfa0313 1302 *
3b4168dd 1303 * ClientSession messages with metadata are v4
dbd0c8bf 1304 */
3b4168dd 1305 msg->hdr.version = cpu_to_le16(4);
7cfa0313 1306 msg->hdr.compat_version = cpu_to_le16(1);
dbd0c8bf
JS		 1307
1308 /* The write pointer, following the session_head structure */
342ce182 1309 p += sizeof(*h);
dbd0c8bf
JS		 1310
1311 /* Number of entries in the map */
1312 ceph_encode_32(&p, metadata_key_count);
1313
1314 /* Two length-prefixed strings for each entry in the map */
d37b1d99 1315 for (i = 0; metadata[i][0]; ++i) {
dbd0c8bf
JS		 1316 size_t const key_len = strlen(metadata[i][0]);
1317 size_t const val_len = strlen(metadata[i][1]);
1318
1319 ceph_encode_32(&p, key_len);
1320 memcpy(p, metadata[i][0], key_len);
1321 p += key_len;
1322 ceph_encode_32(&p, val_len);
1323 memcpy(p, metadata[i][1], val_len);
1324 p += val_len;
1325 }
1326
b682c6d4
XL		 1327 ret = encode_supported_features(&p, end);
1328 if (ret) {
1329 pr_err("encode_supported_features failed!\n");
1330 ceph_msg_put(msg);
1331 return ERR_PTR(ret);
1332 }
1333
3b4168dd
XL		 1334 ret = encode_metric_spec(&p, end);
1335 if (ret) {
1336 pr_err("encode_metric_spec failed!\n");
1337 ceph_msg_put(msg);
1338 return ERR_PTR(ret);
1339 }
1340
342ce182
YZ		 1341 msg->front.iov_len = p - msg->front.iov_base;
1342 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1343
2f2dc053
SW		 1344 return msg;
1345}
1346
1347/*
1348 * send session open request.
1349 *
1350 * called under mdsc->mutex
1351 */
1352static int __open_session(struct ceph_mds_client *mdsc,
1353 struct ceph_mds_session *session)
1354{
1355 struct ceph_msg *msg;
1356 int mstate;
1357 int mds = session->s_mds;
2f2dc053
SW		 1358
1359 /* wait for mds to go active? */
1360 mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1361 dout("open_session to mds%d (%s)\n", mds,
1362 ceph_mds_state_name(mstate));
1363 session->s_state = CEPH_MDS_SESSION_OPENING;
1364 session->s_renew_requested = jiffies;
1365
1366 /* send connect message */
dbd0c8bf 1367 msg = create_session_open_msg(mdsc, session->s_seq);
b682c6d4
XL		 1368 if (IS_ERR(msg))
1369 return PTR_ERR(msg);
2f2dc053 1370 ceph_con_send(&session->s_con, msg);
2f2dc053
SW		 1371 return 0;
1372}
1373
ed0552a1
SW		 1374/*
1375 * open sessions for any export targets for the given mds
1376 *
1377 * called under mdsc->mutex
1378 */
5d72d13c
YZ		 1379static struct ceph_mds_session *
1380__open_export_target_session(struct ceph_mds_client *mdsc, int target)
1381{
1382 struct ceph_mds_session *session;
b682c6d4 1383 int ret;
5d72d13c
YZ		 1384
1385 session = __ceph_lookup_mds_session(mdsc, target);
1386 if (!session) {
1387 session = register_session(mdsc, target);
1388 if (IS_ERR(session))
1389 return session;
1390 }
1391 if (session->s_state == CEPH_MDS_SESSION_NEW ||
b682c6d4
XL		 1392 session->s_state == CEPH_MDS_SESSION_CLOSING) {
1393 ret = __open_session(mdsc, session);
1394 if (ret)
1395 return ERR_PTR(ret);
1396 }
5d72d13c
YZ		 1397
1398 return session;
1399}
1400
1401struct ceph_mds_session *
1402ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1403{
1404 struct ceph_mds_session *session;
1405
1406 dout("open_export_target_session to mds%d\n", target);
1407
1408 mutex_lock(&mdsc->mutex);
1409 session = __open_export_target_session(mdsc, target);
1410 mutex_unlock(&mdsc->mutex);
1411
1412 return session;
1413}
1414
ed0552a1
SW		 1415static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1416 struct ceph_mds_session *session)
1417{
1418 struct ceph_mds_info *mi;
1419 struct ceph_mds_session *ts;
1420 int i, mds = session->s_mds;
ed0552a1 1421
b38c9eb4 1422 if (mds >= mdsc->mdsmap->possible_max_rank)
ed0552a1 1423 return;
5d72d13c 1424
ed0552a1
SW		 1425 mi = &mdsc->mdsmap->m_info[mds];
1426 dout("open_export_target_sessions for mds%d (%d targets)\n",
1427 session->s_mds, mi->num_export_targets);
1428
1429 for (i = 0; i < mi->num_export_targets; i++) {
5d72d13c
YZ		 1430 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1431 if (!IS_ERR(ts))
1432 ceph_put_mds_session(ts);
ed0552a1
SW		 1433 }
1434}
1435
154f42c2
SW		 1436void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1437 struct ceph_mds_session *session)
1438{
1439 mutex_lock(&mdsc->mutex);
1440 __open_export_target_sessions(mdsc, session);
1441 mutex_unlock(&mdsc->mutex);
1442}
1443
2f2dc053
SW		 1444/*
1445 * session caps
1446 */
1447
c8a96a31
JL		 1448static void detach_cap_releases(struct ceph_mds_session *session,
1449 struct list_head *target)
2f2dc053 1450{
c8a96a31
JL		 1451 lockdep_assert_held(&session->s_cap_lock);
1452
1453 list_splice_init(&session->s_cap_releases, target);
745a8e3b 1454 session->s_num_cap_releases = 0;
c8a96a31
JL		 1455 dout("dispose_cap_releases mds%d\n", session->s_mds);
1456}
2f2dc053 1457
c8a96a31
JL		 1458static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1459 struct list_head *dispose)
1460{
1461 while (!list_empty(dispose)) {
745a8e3b
YZ		 1462 struct ceph_cap *cap;
1463 /* zero out the in-progress message */
c8a96a31 1464 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
745a8e3b
YZ		 1465 list_del(&cap->session_caps);
1466 ceph_put_cap(mdsc, cap);
2f2dc053 1467 }
2f2dc053
SW		 1468}
1469
1c841a96
YZ		 1470static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1471 struct ceph_mds_session *session)
1472{
1473 struct ceph_mds_request *req;
1474 struct rb_node *p;
f4b97866 1475 struct ceph_inode_info *ci;
1c841a96
YZ		 1476
1477 dout("cleanup_session_requests mds%d\n", session->s_mds);
1478 mutex_lock(&mdsc->mutex);
1479 while (!list_empty(&session->s_unsafe)) {
1480 req = list_first_entry(&session->s_unsafe,
1481 struct ceph_mds_request, r_unsafe_item);
3e0708b9
YZ		 1482 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1483 req->r_tid);
f4b97866
YZ		 1484 if (req->r_target_inode) {
1485 /* dropping unsafe change of inode's attributes */
1486 ci = ceph_inode(req->r_target_inode);
1487 errseq_set(&ci->i_meta_err, -EIO);
1488 }
1489 if (req->r_unsafe_dir) {
1490 /* dropping unsafe directory operation */
1491 ci = ceph_inode(req->r_unsafe_dir);
1492 errseq_set(&ci->i_meta_err, -EIO);
1493 }
1c841a96
YZ		 1494 __unregister_request(mdsc, req);
1495 }
1496 /* zero r_attempts, so kick_requests() will re-send requests */
1497 p = rb_first(&mdsc->request_tree);
1498 while (p) {
1499 req = rb_entry(p, struct ceph_mds_request, r_node);
1500 p = rb_next(p);
1501 if (req->r_session &&
1502 req->r_session->s_mds == session->s_mds)
1503 req->r_attempts = 0;
1504 }
1505 mutex_unlock(&mdsc->mutex);
1506}
1507
2f2dc053 1508/*
f818a736
SW		 1509 * Helper to safely iterate over all caps associated with a session, with
1510 * special care taken to handle a racing __ceph_remove_cap().
2f2dc053 1511 *
f818a736 1512 * Caller must hold session s_mutex.
2f2dc053 1513 */
f5d77269
JL		 1514int ceph_iterate_session_caps(struct ceph_mds_session *session,
1515 int (*cb)(struct inode *, struct ceph_cap *,
1516 void *), void *arg)
2f2dc053 1517{
7c1332b8
SW		 1518 struct list_head *p;
1519 struct ceph_cap *cap;
1520 struct inode *inode, *last_inode = NULL;
1521 struct ceph_cap *old_cap = NULL;
2f2dc053
SW		 1522 int ret;
1523
1524 dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1525 spin_lock(&session->s_cap_lock);
7c1332b8
SW		 1526 p = session->s_caps.next;
1527 while (p != &session->s_caps) {
1528 cap = list_entry(p, struct ceph_cap, session_caps);
2f2dc053 1529 inode = igrab(&cap->ci->vfs_inode);
7c1332b8
SW		 1530 if (!inode) {
1531 p = p->next;
2f2dc053 1532 continue;
7c1332b8
SW		 1533 }
1534 session->s_cap_iterator = cap;
2f2dc053 1535 spin_unlock(&session->s_cap_lock);
7c1332b8
SW		 1536
1537 if (last_inode) {
3e1d0452
YZ		 1538 /* avoid calling iput_final() while holding
1539 * s_mutex or in mds dispatch threads */
1540 ceph_async_iput(last_inode);
7c1332b8
SW		 1541 last_inode = NULL;
1542 }
1543 if (old_cap) {
37151668 1544 ceph_put_cap(session->s_mdsc, old_cap);
7c1332b8
SW		 1545 old_cap = NULL;
1546 }
1547
2f2dc053 1548 ret = cb(inode, cap, arg);
7c1332b8
SW		 1549 last_inode = inode;
1550
2f2dc053 1551 spin_lock(&session->s_cap_lock);
7c1332b8 1552 p = p->next;
d37b1d99 1553 if (!cap->ci) {
7c1332b8
SW		 1554 dout("iterate_session_caps finishing cap %p removal\n",
1555 cap);
1556 BUG_ON(cap->session != session);
745a8e3b 1557 cap->session = NULL;
7c1332b8
SW		 1558 list_del_init(&cap->session_caps);
1559 session->s_nr_caps--;
4f1d756d 1560 atomic64_dec(&session->s_mdsc->metric.total_caps);
e3ec8d68
YZ		 1561 if (cap->queue_release)
1562 __ceph_queue_cap_release(session, cap);
1563 else
745a8e3b 1564 old_cap = cap; /* put_cap it w/o locks held */
7c1332b8 1565 }
5dacf091
SW		 1566 if (ret < 0)
1567 goto out;
2f2dc053 1568 }
5dacf091
SW		 1569 ret = 0;
1570out:
7c1332b8 1571 session->s_cap_iterator = NULL;
2f2dc053 1572 spin_unlock(&session->s_cap_lock);
7c1332b8 1573
3e1d0452 1574 ceph_async_iput(last_inode);
7c1332b8 1575 if (old_cap)
37151668 1576 ceph_put_cap(session->s_mdsc, old_cap);
7c1332b8 1577
5dacf091 1578 return ret;
2f2dc053
SW		 1579}
1580
1581static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
6c99f254 1582 void *arg)
2f2dc053 1583{
6c93df5d 1584 struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
2f2dc053 1585 struct ceph_inode_info *ci = ceph_inode(inode);
553adfd9 1586 LIST_HEAD(to_remove);
f4b97866 1587 bool dirty_dropped = false;
6c93df5d 1588 bool invalidate = false;
6c99f254 1589
2f2dc053
SW		 1590 dout("removing cap %p, ci is %p, inode is %p\n",
1591 cap, ci, &ci->vfs_inode);
be655596 1592 spin_lock(&ci->i_ceph_lock);
a096b09a 1593 __ceph_remove_cap(cap, false);
571ade33 1594 if (!ci->i_auth_cap) {
553adfd9 1595 struct ceph_cap_flush *cf;
6c93df5d 1596 struct ceph_mds_client *mdsc = fsc->mdsc;
6c99f254 1597
d468e729
YZ		 1598 if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
1599 if (inode->i_data.nrpages > 0)
1600 invalidate = true;
1601 if (ci->i_wrbuffer_ref > 0)
1602 mapping_set_error(&inode->i_data, -EIO);
1603 }
6c93df5d 1604
e4500b5e
YZ		 1605 while (!list_empty(&ci->i_cap_flush_list)) {
1606 cf = list_first_entry(&ci->i_cap_flush_list,
1607 struct ceph_cap_flush, i_list);
8cdcc07d 1608 list_move(&cf->i_list, &to_remove);
553adfd9
YZ		 1609 }
1610
6c99f254 1611 spin_lock(&mdsc->cap_dirty_lock);
8310b089 1612
e4500b5e
YZ		 1613 list_for_each_entry(cf, &to_remove, i_list)
1614 list_del(&cf->g_list);
8310b089 1615
6c99f254 1616 if (!list_empty(&ci->i_dirty_item)) {
3e0708b9
YZ		 1617 pr_warn_ratelimited(
1618 " dropping dirty %s state for %p %lld\n",
6c99f254
SW		 1619 ceph_cap_string(ci->i_dirty_caps),
1620 inode, ceph_ino(inode));
1621 ci->i_dirty_caps = 0;
1622 list_del_init(&ci->i_dirty_item);
f4b97866 1623 dirty_dropped = true;
6c99f254
SW		 1624 }
1625 if (!list_empty(&ci->i_flushing_item)) {
3e0708b9
YZ		 1626 pr_warn_ratelimited(
1627 " dropping dirty+flushing %s state for %p %lld\n",
6c99f254
SW		 1628 ceph_cap_string(ci->i_flushing_caps),
1629 inode, ceph_ino(inode));
1630 ci->i_flushing_caps = 0;
1631 list_del_init(&ci->i_flushing_item);
1632 mdsc->num_cap_flushing--;
f4b97866 1633 dirty_dropped = true;
6c99f254 1634 }
6c99f254 1635 spin_unlock(&mdsc->cap_dirty_lock);
553adfd9 1636
f4b97866
YZ		 1637 if (dirty_dropped) {
1638 errseq_set(&ci->i_meta_err, -EIO);
1639
1640 if (ci->i_wrbuffer_ref_head == 0 &&
1641 ci->i_wr_ref == 0 &&
1642 ci->i_dirty_caps == 0 &&
1643 ci->i_flushing_caps == 0) {
1644 ceph_put_snap_context(ci->i_head_snapc);
1645 ci->i_head_snapc = NULL;
1646 }
1647 }
1648
b3f8d68f
YZ		 1649 if (atomic_read(&ci->i_filelock_ref) > 0) {
1650 /* make further file lock syscall return -EIO */
1651 ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
1652 pr_warn_ratelimited(" dropping file locks for %p %lld\n",
1653 inode, ceph_ino(inode));
1654 }
1655
f66fd9f0 1656 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
e4500b5e 1657 list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);
f66fd9f0
YZ		 1658 ci->i_prealloc_cap_flush = NULL;
1659 }
6c99f254 1660 }
be655596 1661 spin_unlock(&ci->i_ceph_lock);
553adfd9
YZ		 1662 while (!list_empty(&to_remove)) {
1663 struct ceph_cap_flush *cf;
1664 cf = list_first_entry(&to_remove,
e4500b5e
YZ		 1665 struct ceph_cap_flush, i_list);
1666 list_del(&cf->i_list);
f66fd9f0 1667 ceph_free_cap_flush(cf);
553adfd9 1668 }
77310320
YZ		 1669
1670 wake_up_all(&ci->i_cap_wq);
6c93df5d
YZ		 1671 if (invalidate)
1672 ceph_queue_invalidate(inode);
f4b97866 1673 if (dirty_dropped)
6c99f254 1674 iput(inode);
2f2dc053
SW		 1675 return 0;
1676}
1677
1678/*
1679 * caller must hold session s_mutex
1680 */
1681static void remove_session_caps(struct ceph_mds_session *session)
1682{
6c93df5d
YZ		 1683 struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1684 struct super_block *sb = fsc->sb;
c8a96a31
JL		 1685 LIST_HEAD(dispose);
1686
2f2dc053 1687 dout("remove_session_caps on %p\n", session);
f5d77269 1688 ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
6f60f889 1689
c8799fc4
YZ		 1690 wake_up_all(&fsc->mdsc->cap_flushing_wq);
1691
6f60f889
YZ		 1692 spin_lock(&session->s_cap_lock);
1693 if (session->s_nr_caps > 0) {
6f60f889
YZ		 1694 struct inode *inode;
1695 struct ceph_cap *cap, *prev = NULL;
1696 struct ceph_vino vino;
1697 /*
1698 * iterate_session_caps() skips inodes that are being
1699 * deleted, we need to wait until deletions are complete.
1700 * __wait_on_freeing_inode() is designed for the job,
1701 * but it is not exported, so use lookup inode function
1702 * to access it.
1703 */
1704 while (!list_empty(&session->s_caps)) {
1705 cap = list_entry(session->s_caps.next,
1706 struct ceph_cap, session_caps);
1707 if (cap == prev)
1708 break;
1709 prev = cap;
1710 vino = cap->ci->i_vino;
1711 spin_unlock(&session->s_cap_lock);
1712
ed284c49 1713 inode = ceph_find_inode(sb, vino);
3e1d0452
YZ		 1714 /* avoid calling iput_final() while holding s_mutex */
1715 ceph_async_iput(inode);
6f60f889
YZ		 1716
1717 spin_lock(&session->s_cap_lock);
1718 }
1719 }
745a8e3b
YZ		 1720
1721 // drop cap expires and unlock s_cap_lock
c8a96a31 1722 detach_cap_releases(session, &dispose);
6f60f889 1723
2f2dc053 1724 BUG_ON(session->s_nr_caps > 0);
6c99f254 1725 BUG_ON(!list_empty(&session->s_cap_flushing));
c8a96a31
JL		 1726 spin_unlock(&session->s_cap_lock);
1727 dispose_cap_releases(session->s_mdsc, &dispose);
2f2dc053
SW		 1728}
1729
d2f8bb27
YZ		 1730enum {
1731 RECONNECT,
1732 RENEWCAPS,
1733 FORCE_RO,
1734};
1735
2f2dc053
SW		 1736/*
1737 * wake up any threads waiting on this session's caps. if the cap is
1738 * old (didn't get renewed on the client reconnect), remove it now.
1739 *
1740 * caller must hold s_mutex.
1741 */
1742static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1743 void *arg)
1744{
0dc2570f 1745 struct ceph_inode_info *ci = ceph_inode(inode);
d2f8bb27 1746 unsigned long ev = (unsigned long)arg;
0dc2570f 1747
d2f8bb27 1748 if (ev == RECONNECT) {
be655596 1749 spin_lock(&ci->i_ceph_lock);
0dc2570f
SW		 1750 ci->i_wanted_max_size = 0;
1751 ci->i_requested_max_size = 0;
be655596 1752 spin_unlock(&ci->i_ceph_lock);
d2f8bb27
YZ		 1753 } else if (ev == RENEWCAPS) {
1754 if (cap->cap_gen < cap->session->s_cap_gen) {
1755 /* mds did not re-issue stale cap */
1756 spin_lock(&ci->i_ceph_lock);
1757 cap->issued = cap->implemented = CEPH_CAP_PIN;
d2f8bb27
YZ		 1758 spin_unlock(&ci->i_ceph_lock);
1759 }
1760 } else if (ev == FORCE_RO) {
0dc2570f 1761 }
e5360309 1762 wake_up_all(&ci->i_cap_wq);
2f2dc053
SW		 1763 return 0;
1764}
1765
d2f8bb27 1766static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
2f2dc053
SW		 1767{
1768 dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
f5d77269
JL		 1769 ceph_iterate_session_caps(session, wake_up_session_cb,
1770 (void *)(unsigned long)ev);
2f2dc053
SW		 1771}
1772
1773/*
1774 * Send periodic message to MDS renewing all currently held caps. The
1775 * ack will reset the expiration for all caps from this session.
1776 *
1777 * caller holds s_mutex
1778 */
1779static int send_renew_caps(struct ceph_mds_client *mdsc,
1780 struct ceph_mds_session *session)
1781{
1782 struct ceph_msg *msg;
1783 int state;
1784
1785 if (time_after_eq(jiffies, session->s_cap_ttl) &&
1786 time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1787 pr_info("mds%d caps stale\n", session->s_mds);
e4cb4cb8 1788 session->s_renew_requested = jiffies;
2f2dc053
SW		 1789
1790 /* do not try to renew caps until a recovering mds has reconnected
1791 * with its clients. */
1792 state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1793 if (state < CEPH_MDS_STATE_RECONNECT) {
1794 dout("send_renew_caps ignoring mds%d (%s)\n",
1795 session->s_mds, ceph_mds_state_name(state));
1796 return 0;
1797 }
1798
1799 dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1800 ceph_mds_state_name(state));
2f2dc053
SW		 1801 msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1802 ++session->s_renew_seq);
a79832f2
SW		 1803 if (!msg)
1804 return -ENOMEM;
2f2dc053
SW		 1805 ceph_con_send(&session->s_con, msg);
1806 return 0;
1807}
1808
186e4f7a
YZ		 1809static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1810 struct ceph_mds_session *session, u64 seq)
1811{
1812 struct ceph_msg *msg;
1813
1814 dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
a687ecaf 1815 session->s_mds, ceph_session_state_name(session->s_state), seq);
186e4f7a
YZ		 1816 msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1817 if (!msg)
1818 return -ENOMEM;
1819 ceph_con_send(&session->s_con, msg);
1820 return 0;
1821}
1822
1823
2f2dc053
SW		 1824/*
1825 * Note new cap ttl, and any transition from stale -> not stale (fresh?).
0dc2570f
SW		 1826 *
1827 * Called under session->s_mutex
2f2dc053
SW		 1828 */
1829static void renewed_caps(struct ceph_mds_client *mdsc,
1830 struct ceph_mds_session *session, int is_renew)
1831{
1832 int was_stale;
1833 int wake = 0;
1834
1835 spin_lock(&session->s_cap_lock);
1ce208a6 1836 was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
2f2dc053
SW		 1837
1838 session->s_cap_ttl = session->s_renew_requested +
1839 mdsc->mdsmap->m_session_timeout*HZ;
1840
1841 if (was_stale) {
1842 if (time_before(jiffies, session->s_cap_ttl)) {
1843 pr_info("mds%d caps renewed\n", session->s_mds);
1844 wake = 1;
1845 } else {
1846 pr_info("mds%d caps still stale\n", session->s_mds);
1847 }
1848 }
1849 dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1850 session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1851 time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1852 spin_unlock(&session->s_cap_lock);
1853
1854 if (wake)
d2f8bb27 1855 wake_up_session_caps(session, RENEWCAPS);
2f2dc053
SW		 1856}
1857
1858/*
1859 * send a session close request
1860 */
3e699bd8 1861static int request_close_session(struct ceph_mds_session *session)
2f2dc053
SW		 1862{
1863 struct ceph_msg *msg;
2f2dc053
SW		 1864
1865 dout("request_close_session mds%d state %s seq %lld\n",
a687ecaf 1866 session->s_mds, ceph_session_state_name(session->s_state),
2f2dc053
SW		 1867 session->s_seq);
1868 msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
a79832f2
SW		 1869 if (!msg)
1870 return -ENOMEM;
1871 ceph_con_send(&session->s_con, msg);
fcff415c 1872 return 1;
2f2dc053
SW		 1873}
1874
1875/*
1876 * Called with s_mutex held.
1877 */
1878static int __close_session(struct ceph_mds_client *mdsc,
1879 struct ceph_mds_session *session)
1880{
1881 if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1882 return 0;
1883 session->s_state = CEPH_MDS_SESSION_CLOSING;
3e699bd8 1884 return request_close_session(session);
2f2dc053
SW		 1885}
1886
040d7860
YZ		 1887static bool drop_negative_children(struct dentry *dentry)
1888{
1889 struct dentry *child;
1890 bool all_negative = true;
1891
1892 if (!d_is_dir(dentry))
1893 goto out;
1894
1895 spin_lock(&dentry->d_lock);
1896 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1897 if (d_really_is_positive(child)) {
1898 all_negative = false;
1899 break;
1900 }
1901 }
1902 spin_unlock(&dentry->d_lock);
1903
1904 if (all_negative)
1905 shrink_dcache_parent(dentry);
1906out:
1907 return all_negative;
1908}
1909
2f2dc053
SW		 1910/*
1911 * Trim old(er) caps.
1912 *
1913 * Because we can't cache an inode without one or more caps, we do
1914 * this indirectly: if a cap is unused, we prune its aliases, at which
1915 * point the inode will hopefully get dropped to.
1916 *
1917 * Yes, this is a bit sloppy. Our only real goal here is to respond to
1918 * memory pressure from the MDS, though, so it needn't be perfect.
1919 */
1920static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1921{
533a2818 1922 int *remaining = arg;
2f2dc053 1923 struct ceph_inode_info *ci = ceph_inode(inode);
979abfdd 1924 int used, wanted, oissued, mine;
2f2dc053 1925
533a2818 1926 if (*remaining <= 0)
2f2dc053
SW		 1927 return -1;
1928
be655596 1929 spin_lock(&ci->i_ceph_lock);
2f2dc053
SW		 1930 mine = cap->issued | cap->implemented;
1931 used = __ceph_caps_used(ci);
979abfdd 1932 wanted = __ceph_caps_file_wanted(ci);
2f2dc053
SW		 1933 oissued = __ceph_caps_issued_other(ci, cap);
1934
979abfdd 1935 dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
2f2dc053 1936 inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
979abfdd
YZ		 1937 ceph_cap_string(used), ceph_cap_string(wanted));
1938 if (cap == ci->i_auth_cap) {
622f3e25
YZ		 1939 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1940 !list_empty(&ci->i_cap_snaps))
979abfdd
YZ		 1941 goto out;
1942 if ((used | wanted) & CEPH_CAP_ANY_WR)
1943 goto out;
89aa5930
YZ		 1944 /* Note: it's possible that i_filelock_ref becomes non-zero
1945 * after dropping auth caps. It doesn't hurt because reply
1946 * of lock mds request will re-add auth caps. */
1947 if (atomic_read(&ci->i_filelock_ref) > 0)
1948 goto out;
979abfdd 1949 }
5e804ac4
YZ		 1950 /* The inode has cached pages, but it's no longer used.
1951 * we can safely drop it */
525d15e8
YZ		 1952 if (S_ISREG(inode->i_mode) &&
1953 wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
5e804ac4
YZ		 1954 !(oissued & CEPH_CAP_FILE_CACHE)) {
1955 used = 0;
1956 oissued = 0;
1957 }
979abfdd 1958 if ((used | wanted) & ~oissued & mine)
2f2dc053
SW		 1959 goto out; /* we need these caps */
1960
2f2dc053
SW		 1961 if (oissued) {
1962 /* we aren't the only cap.. just remove us */
a096b09a 1963 __ceph_remove_cap(cap, true);
533a2818 1964 (*remaining)--;
2f2dc053 1965 } else {
040d7860 1966 struct dentry *dentry;
5e804ac4 1967 /* try dropping referring dentries */
be655596 1968 spin_unlock(&ci->i_ceph_lock);
040d7860
YZ		 1969 dentry = d_find_any_alias(inode);
1970 if (dentry && drop_negative_children(dentry)) {
1971 int count;
1972 dput(dentry);
1973 d_prune_aliases(inode);
1974 count = atomic_read(&inode->i_count);
1975 if (count == 1)
533a2818 1976 (*remaining)--;
040d7860
YZ		 1977 dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1978 inode, cap, count);
1979 } else {
1980 dput(dentry);
1981 }
2f2dc053
SW		 1982 return 0;
1983 }
1984
1985out:
be655596 1986 spin_unlock(&ci->i_ceph_lock);
2f2dc053
SW		 1987 return 0;
1988}
1989
1990/*
1991 * Trim session cap count down to some max number.
1992 */
e30ee581
ZZ		 1993int ceph_trim_caps(struct ceph_mds_client *mdsc,
1994 struct ceph_mds_session *session,
1995 int max_caps)
2f2dc053
SW		 1996{
1997 int trim_caps = session->s_nr_caps - max_caps;
1998
1999 dout("trim_caps mds%d start: %d / %d, trim %d\n",
2000 session->s_mds, session->s_nr_caps, max_caps, trim_caps);
2001 if (trim_caps > 0) {
533a2818
JL		 2002 int remaining = trim_caps;
2003
2004 ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
2f2dc053
SW		 2005 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
2006 session->s_mds, session->s_nr_caps, max_caps,
533a2818 2007 trim_caps - remaining);
2f2dc053 2008 }
a56371d9 2009
e3ec8d68 2010 ceph_flush_cap_releases(mdsc, session);
2f2dc053
SW		 2011 return 0;
2012}
2013
8310b089
YZ		 2014static int check_caps_flush(struct ceph_mds_client *mdsc,
2015 u64 want_flush_tid)
2016{
8310b089
YZ		 2017 int ret = 1;
2018
2019 spin_lock(&mdsc->cap_dirty_lock);
e4500b5e
YZ		 2020 if (!list_empty(&mdsc->cap_flush_list)) {
2021 struct ceph_cap_flush *cf =
2022 list_first_entry(&mdsc->cap_flush_list,
2023 struct ceph_cap_flush, g_list);
2024 if (cf->tid <= want_flush_tid) {
2025 dout("check_caps_flush still flushing tid "
2026 "%llu <= %llu\n", cf->tid, want_flush_tid);
2027 ret = 0;
2028 }
8310b089
YZ		 2029 }
2030 spin_unlock(&mdsc->cap_dirty_lock);
2031 return ret;
d3383a8e
YZ		 2032}
2033
2f2dc053
SW		 2034/*
2035 * flush all dirty inode data to disk.
2036 *
8310b089 2037 * returns true if we've flushed through want_flush_tid
2f2dc053 2038 */
affbc19a 2039static void wait_caps_flush(struct ceph_mds_client *mdsc,
0e294387 2040 u64 want_flush_tid)
2f2dc053 2041{
0e294387 2042 dout("check_caps_flush want %llu\n", want_flush_tid);
8310b089
YZ		 2043
2044 wait_event(mdsc->cap_flushing_wq,
2045 check_caps_flush(mdsc, want_flush_tid));
2046
2047 dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
2f2dc053
SW		 2048}
2049
2050/*
2051 * called under s_mutex
2052 */
e3ec8d68
YZ		 2053static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
2054 struct ceph_mds_session *session)
2f2dc053 2055{
745a8e3b
YZ		 2056 struct ceph_msg *msg = NULL;
2057 struct ceph_mds_cap_release *head;
2058 struct ceph_mds_cap_item *item;
92475f05 2059 struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
745a8e3b
YZ		 2060 struct ceph_cap *cap;
2061 LIST_HEAD(tmp_list);
2062 int num_cap_releases;
92475f05
JL		 2063 __le32 barrier, *cap_barrier;
2064
2065 down_read(&osdc->lock);
2066 barrier = cpu_to_le32(osdc->epoch_barrier);
2067 up_read(&osdc->lock);
2f2dc053 2068
0f8605f2 2069 spin_lock(&session->s_cap_lock);
745a8e3b
YZ		 2070again:
2071 list_splice_init(&session->s_cap_releases, &tmp_list);
2072 num_cap_releases = session->s_num_cap_releases;
2073 session->s_num_cap_releases = 0;
2f2dc053 2074 spin_unlock(&session->s_cap_lock);
e01a5946 2075
745a8e3b
YZ		 2076 while (!list_empty(&tmp_list)) {
2077 if (!msg) {
2078 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
09cbfeaf 2079 PAGE_SIZE, GFP_NOFS, false);
745a8e3b
YZ		 2080 if (!msg)
2081 goto out_err;
2082 head = msg->front.iov_base;
2083 head->num = cpu_to_le32(0);
2084 msg->front.iov_len = sizeof(*head);
92475f05
JL		 2085
2086 msg->hdr.version = cpu_to_le16(2);
2087 msg->hdr.compat_version = cpu_to_le16(1);
745a8e3b 2088 }
92475f05 2089
745a8e3b
YZ		 2090 cap = list_first_entry(&tmp_list, struct ceph_cap,
2091 session_caps);
2092 list_del(&cap->session_caps);
2093 num_cap_releases--;
e01a5946 2094
00bd8edb 2095 head = msg->front.iov_base;
4198aba4
JL		 2096 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2097 &head->num);
745a8e3b
YZ		 2098 item = msg->front.iov_base + msg->front.iov_len;
2099 item->ino = cpu_to_le64(cap->cap_ino);
2100 item->cap_id = cpu_to_le64(cap->cap_id);
2101 item->migrate_seq = cpu_to_le32(cap->mseq);
2102 item->seq = cpu_to_le32(cap->issue_seq);
2103 msg->front.iov_len += sizeof(*item);
2104
2105 ceph_put_cap(mdsc, cap);
2106
2107 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
92475f05
JL		 2108 // Append cap_barrier field
2109 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2110 *cap_barrier = barrier;
2111 msg->front.iov_len += sizeof(*cap_barrier);
2112
745a8e3b
YZ		 2113 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2114 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2115 ceph_con_send(&session->s_con, msg);
2116 msg = NULL;
2117 }
00bd8edb 2118 }
e01a5946 2119
745a8e3b 2120 BUG_ON(num_cap_releases != 0);
e01a5946 2121
745a8e3b
YZ		 2122 spin_lock(&session->s_cap_lock);
2123 if (!list_empty(&session->s_cap_releases))
2124 goto again;
2125 spin_unlock(&session->s_cap_lock);
2126
2127 if (msg) {
92475f05
JL		 2128 // Append cap_barrier field
2129 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2130 *cap_barrier = barrier;
2131 msg->front.iov_len += sizeof(*cap_barrier);
2132
745a8e3b
YZ		 2133 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2134 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2135 ceph_con_send(&session->s_con, msg);
e01a5946 2136 }
745a8e3b
YZ		 2137 return;
2138out_err:
2139 pr_err("send_cap_releases mds%d, failed to allocate message\n",
2140 session->s_mds);
2141 spin_lock(&session->s_cap_lock);
2142 list_splice(&tmp_list, &session->s_cap_releases);
2143 session->s_num_cap_releases += num_cap_releases;
2144 spin_unlock(&session->s_cap_lock);
e01a5946
SW		 2145}
2146
e3ec8d68
YZ		 2147static void ceph_cap_release_work(struct work_struct *work)
2148{
2149 struct ceph_mds_session *session =
2150 container_of(work, struct ceph_mds_session, s_cap_release_work);
2151
2152 mutex_lock(&session->s_mutex);
2153 if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2154 session->s_state == CEPH_MDS_SESSION_HUNG)
2155 ceph_send_cap_releases(session->s_mdsc, session);
2156 mutex_unlock(&session->s_mutex);
2157 ceph_put_mds_session(session);
2158}
2159
2160void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2161 struct ceph_mds_session *session)
2162{
2163 if (mdsc->stopping)
2164 return;
2165
5b3248c6 2166 ceph_get_mds_session(session);
e3ec8d68
YZ		 2167 if (queue_work(mdsc->fsc->cap_wq,
2168 &session->s_cap_release_work)) {
2169 dout("cap release work queued\n");
2170 } else {
2171 ceph_put_mds_session(session);
2172 dout("failed to queue cap release work\n");
2173 }
2174}
2175
2176/*
2177 * caller holds session->s_cap_lock
2178 */
2179void __ceph_queue_cap_release(struct ceph_mds_session *session,
2180 struct ceph_cap *cap)
2181{
2182 list_add_tail(&cap->session_caps, &session->s_cap_releases);
2183 session->s_num_cap_releases++;
2184
2185 if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2186 ceph_flush_cap_releases(session->s_mdsc, session);
2187}
2188
37c4efc1
YZ		 2189static void ceph_cap_reclaim_work(struct work_struct *work)
2190{
2191 struct ceph_mds_client *mdsc =
2192 container_of(work, struct ceph_mds_client, cap_reclaim_work);
2193 int ret = ceph_trim_dentries(mdsc);
2194 if (ret == -EAGAIN)
2195 ceph_queue_cap_reclaim_work(mdsc);
2196}
2197
2198void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2199{
2200 if (mdsc->stopping)
2201 return;
2202
2203 if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2204 dout("caps reclaim work queued\n");
2205 } else {
2206 dout("failed to queue caps release work\n");
2207 }
2208}
2209
fe33032d
YZ		 2210void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2211{
2212 int val;
2213 if (!nr)
2214 return;
2215 val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
bba1560b 2216 if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
fe33032d
YZ		 2217 atomic_set(&mdsc->cap_reclaim_pending, 0);
2218 ceph_queue_cap_reclaim_work(mdsc);
2219 }
2220}
2221
2f2dc053
SW		 2222/*
2223 * requests
2224 */
2225
54008399
YZ		 2226int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2227 struct inode *dir)
2228{
2229 struct ceph_inode_info *ci = ceph_inode(dir);
2230 struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2231 struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2a5beea3 2232 size_t size = sizeof(struct ceph_mds_reply_dir_entry);
ad8c28a9
JL		 2233 unsigned int num_entries;
2234 int order;
54008399
YZ		 2235
2236 spin_lock(&ci->i_ceph_lock);
2237 num_entries = ci->i_files + ci->i_subdirs;
2238 spin_unlock(&ci->i_ceph_lock);
ad8c28a9 2239 num_entries = max(num_entries, 1U);
54008399
YZ		 2240 num_entries = min(num_entries, opt->max_readdir);
2241
2242 order = get_order(size * num_entries);
2243 while (order >= 0) {
2a5beea3
YZ		 2244 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2245 __GFP_NOWARN,
2246 order);
2247 if (rinfo->dir_entries)
54008399
YZ		 2248 break;
2249 order--;
2250 }
2a5beea3 2251 if (!rinfo->dir_entries)
54008399
YZ		 2252 return -ENOMEM;
2253
2254 num_entries = (PAGE_SIZE << order) / size;
2255 num_entries = min(num_entries, opt->max_readdir);
2256
2257 rinfo->dir_buf_size = PAGE_SIZE << order;
2258 req->r_num_caps = num_entries + 1;
2259 req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2260 req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2261 return 0;
2262}
2263
2f2dc053
SW		 2264/*
2265 * Create an mds request.
2266 */
2267struct ceph_mds_request *
2268ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2269{
058daab7 2270 struct ceph_mds_request *req;
2f2dc053 2271
058daab7 2272 req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2f2dc053
SW		 2273 if (!req)
2274 return ERR_PTR(-ENOMEM);
2275
b4556396 2276 mutex_init(&req->r_fill_mutex);
37151668 2277 req->r_mdsc = mdsc;
2f2dc053 2278 req->r_started = jiffies;
70c94820 2279 req->r_start_latency = ktime_get();
2f2dc053
SW		 2280 req->r_resend_mds = -1;
2281 INIT_LIST_HEAD(&req->r_unsafe_dir_item);
68cd5b4b 2282 INIT_LIST_HEAD(&req->r_unsafe_target_item);
2f2dc053 2283 req->r_fmode = -1;
153c8e6b 2284 kref_init(&req->r_kref);
fcd00b68 2285 RB_CLEAR_NODE(&req->r_node);
2f2dc053
SW		 2286 INIT_LIST_HEAD(&req->r_wait);
2287 init_completion(&req->r_completion);
2288 init_completion(&req->r_safe_completion);
2289 INIT_LIST_HEAD(&req->r_unsafe_item);
2290
668c9a61 2291 ktime_get_coarse_real_ts64(&req->r_stamp);
b8e69066 2292
2f2dc053
SW		 2293 req->r_op = op;
2294 req->r_direct_mode = mode;
2295 return req;
2296}
2297
2298/*
44ca18f2 2299 * return oldest (lowest) request, tid in request tree, 0 if none.
2f2dc053
SW		 2300 *
2301 * called under mdsc->mutex.
2302 */
44ca18f2
SW		 2303static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2304{
2305 if (RB_EMPTY_ROOT(&mdsc->request_tree))
2306 return NULL;
2307 return rb_entry(rb_first(&mdsc->request_tree),
2308 struct ceph_mds_request, r_node);
2309}
2310
e8a7b8b1 2311static inline u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2f2dc053 2312{
e8a7b8b1 2313 return mdsc->oldest_tid;
2f2dc053
SW		 2314}
2315
2316/*
2317 * Build a dentry's path. Allocate on heap; caller must kfree. Based
2318 * on build_path_from_dentry in fs/cifs/dir.c.
2319 *
2320 * If @stop_on_nosnap, generate path relative to the first non-snapped
2321 * inode.
2322 *
2323 * Encode hidden .snap dirs as a double /, i.e.
2324 * foo/.snap/bar -> foo//bar
2325 */
69a10fb3 2326char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2f2dc053
SW		 2327 int stop_on_nosnap)
2328{
2329 struct dentry *temp;
2330 char *path;
f77f21bb 2331 int pos;
1b71fe2e 2332 unsigned seq;
69a10fb3 2333 u64 base;
2f2dc053 2334
d37b1d99 2335 if (!dentry)
2f2dc053
SW		 2336 return ERR_PTR(-EINVAL);
2337
f77f21bb 2338 path = __getname();
d37b1d99 2339 if (!path)
2f2dc053 2340 return ERR_PTR(-ENOMEM);
f77f21bb
JL		 2341retry:
2342 pos = PATH_MAX - 1;
2343 path[pos] = '\0';
2344
2345 seq = read_seqbegin(&rename_lock);
1b71fe2e 2346 rcu_read_lock();
f77f21bb
JL		 2347 temp = dentry;
2348 for (;;) {
1b71fe2e 2349 struct inode *inode;
2f2dc053 2350
1b71fe2e 2351 spin_lock(&temp->d_lock);
2b0143b5 2352 inode = d_inode(temp);
2f2dc053 2353 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
104648ad 2354 dout("build_path path+%d: %p SNAPDIR\n",
2f2dc053 2355 pos, temp);
d6b8bd67 2356 } else if (stop_on_nosnap && inode && dentry != temp &&
2f2dc053 2357 ceph_snap(inode) == CEPH_NOSNAP) {
9d5a09e6 2358 spin_unlock(&temp->d_lock);
d6b8bd67 2359 pos++; /* get rid of any prepended '/' */
2f2dc053
SW		 2360 break;
2361 } else {
2362 pos -= temp->d_name.len;
1b71fe2e
AV		 2363 if (pos < 0) {
2364 spin_unlock(&temp->d_lock);
2f2dc053 2365 break;
1b71fe2e 2366 }
f77f21bb 2367 memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2f2dc053 2368 }
1b71fe2e 2369 spin_unlock(&temp->d_lock);
41883ba8 2370 temp = READ_ONCE(temp->d_parent);
f77f21bb
JL		 2371
2372 /* Are we at the root? */
2373 if (IS_ROOT(temp))
2374 break;
2375
2376 /* Are we out of buffer? */
2377 if (--pos < 0)
2378 break;
2379
2380 path[pos] = '/';
2f2dc053 2381 }
69a10fb3 2382 base = ceph_ino(d_inode(temp));
1b71fe2e 2383 rcu_read_unlock();
f5946bcc
JL		 2384
2385 if (read_seqretry(&rename_lock, seq))
2386 goto retry;
2387
2388 if (pos < 0) {
2389 /*
2390 * A rename didn't occur, but somehow we didn't end up where
2391 * we thought we would. Throw a warning and try again.
2392 */
2393 pr_warn("build_path did not end path lookup where "
2394 "expected, pos is %d\n", pos);
2f2dc053
SW		 2395 goto retry;
2396 }
2397
69a10fb3 2398 *pbase = base;
f77f21bb 2399 *plen = PATH_MAX - 1 - pos;
104648ad 2400 dout("build_path on %p %d built %llx '%.*s'\n",
f77f21bb
JL		 2401 dentry, d_count(dentry), base, *plen, path + pos);
2402 return path + pos;
2f2dc053
SW		 2403}
2404
fd36a717 2405static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2f2dc053 2406 const char **ppath, int *ppathlen, u64 *pino,
1bcb3440 2407 bool *pfreepath, bool parent_locked)
2f2dc053
SW		 2408{
2409 char *path;
2410
c6b0b656 2411 rcu_read_lock();
fd36a717
JL		 2412 if (!dir)
2413 dir = d_inode_rcu(dentry->d_parent);
964fff74 2414 if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
c6b0b656
JL		 2415 *pino = ceph_ino(dir);
2416 rcu_read_unlock();
964fff74
JL		 2417 *ppath = dentry->d_name.name;
2418 *ppathlen = dentry->d_name.len;
2f2dc053
SW		 2419 return 0;
2420 }
c6b0b656 2421 rcu_read_unlock();
2f2dc053
SW		 2422 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2423 if (IS_ERR(path))
2424 return PTR_ERR(path);
2425 *ppath = path;
1bcb3440 2426 *pfreepath = true;
2f2dc053
SW		 2427 return 0;
2428}
2429
2430static int build_inode_path(struct inode *inode,
2431 const char **ppath, int *ppathlen, u64 *pino,
1bcb3440 2432 bool *pfreepath)
2f2dc053
SW		 2433{
2434 struct dentry *dentry;
2435 char *path;
2436
2437 if (ceph_snap(inode) == CEPH_NOSNAP) {
2438 *pino = ceph_ino(inode);
2439 *ppathlen = 0;
2440 return 0;
2441 }
2442 dentry = d_find_alias(inode);
2443 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2444 dput(dentry);
2445 if (IS_ERR(path))
2446 return PTR_ERR(path);
2447 *ppath = path;
1bcb3440 2448 *pfreepath = true;
2f2dc053
SW		 2449 return 0;
2450}
2451
2452/*
2453 * request arguments may be specified via an inode *, a dentry *, or
2454 * an explicit ino+path.
2455 */
2456static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
fd36a717
JL		 2457 struct inode *rdiri, const char *rpath,
2458 u64 rino, const char **ppath, int *pathlen,
1bcb3440 2459 u64 *ino, bool *freepath, bool parent_locked)
2f2dc053
SW		 2460{
2461 int r = 0;
2462
2463 if (rinode) {
2464 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2465 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2466 ceph_snap(rinode));
2467 } else if (rdentry) {
fd36a717 2468 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
1bcb3440 2469 freepath, parent_locked);
2f2dc053
SW		 2470 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2471 *ppath);
795858db 2472 } else if (rpath || rino) {
2f2dc053
SW		 2473 *ino = rino;
2474 *ppath = rpath;
b000056a 2475 *pathlen = rpath ? strlen(rpath) : 0;
2f2dc053
SW		 2476 dout(" path %.*s\n", *pathlen, rpath);
2477 }
2478
2479 return r;
2480}
2481
2482/*
2483 * called under mdsc->mutex
2484 */
2485static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
2486 struct ceph_mds_request *req,
6e6f0923 2487 int mds, bool drop_cap_releases)
2f2dc053
SW		 2488{
2489 struct ceph_msg *msg;
2490 struct ceph_mds_request_head *head;
2491 const char *path1 = NULL;
2492 const char *path2 = NULL;
2493 u64 ino1 = 0, ino2 = 0;
2494 int pathlen1 = 0, pathlen2 = 0;
1bcb3440 2495 bool freepath1 = false, freepath2 = false;
2f2dc053
SW		 2496 int len;
2497 u16 releases;
2498 void *p, *end;
2499 int ret;
2500
2501 ret = set_request_path_attr(req->r_inode, req->r_dentry,
3dd69aab 2502 req->r_parent, req->r_path1, req->r_ino1.ino,
1bcb3440
JL		 2503 &path1, &pathlen1, &ino1, &freepath1,
2504 test_bit(CEPH_MDS_R_PARENT_LOCKED,
2505 &req->r_req_flags));
2f2dc053
SW		 2506 if (ret < 0) {
2507 msg = ERR_PTR(ret);
2508 goto out;
2509 }
2510
1bcb3440 2511 /* If r_old_dentry is set, then assume that its parent is locked */
2f2dc053 2512 ret = set_request_path_attr(NULL, req->r_old_dentry,
fd36a717 2513 req->r_old_dentry_dir,
2f2dc053 2514 req->r_path2, req->r_ino2.ino,
1bcb3440 2515 &path2, &pathlen2, &ino2, &freepath2, true);
2f2dc053
SW		 2516 if (ret < 0) {
2517 msg = ERR_PTR(ret);
2518 goto out_free1;
2519 }
2520
2521 len = sizeof(*head) +
b8e69066 2522 pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
777d738a 2523 sizeof(struct ceph_timespec);
2f2dc053
SW		 2524
2525 /* calculate (max) length for cap releases */
2526 len += sizeof(struct ceph_mds_request_release) *
2527 (!!req->r_inode_drop + !!req->r_dentry_drop +
2528 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2529 if (req->r_dentry_drop)
c1dfc277 2530 len += pathlen1;
2f2dc053 2531 if (req->r_old_dentry_drop)
c1dfc277 2532 len += pathlen2;
2f2dc053 2533
0d9c1ab3 2534 msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
a79832f2
SW		 2535 if (!msg) {
2536 msg = ERR_PTR(-ENOMEM);
2f2dc053 2537 goto out_free2;
a79832f2 2538 }
2f2dc053 2539
7cfa0313 2540 msg->hdr.version = cpu_to_le16(2);
6df058c0
SW		 2541 msg->hdr.tid = cpu_to_le64(req->r_tid);
2542
2f2dc053
SW		 2543 head = msg->front.iov_base;
2544 p = msg->front.iov_base + sizeof(*head);
2545 end = msg->front.iov_base + msg->front.iov_len;
2546
2547 head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2548 head->op = cpu_to_le32(req->r_op);
ff3d0046
EB		 2549 head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));
2550 head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));
6deb8008 2551 head->ino = cpu_to_le64(req->r_deleg_ino);
2f2dc053
SW		 2552 head->args = req->r_args;
2553
2554 ceph_encode_filepath(&p, end, ino1, path1);
2555 ceph_encode_filepath(&p, end, ino2, path2);
2556
e979cf50
SW		 2557 /* make note of release offset, in case we need to replay */
2558 req->r_request_release_offset = p - msg->front.iov_base;
2559
2f2dc053
SW		 2560 /* cap releases */
2561 releases = 0;
2562 if (req->r_inode_drop)
2563 releases += ceph_encode_inode_release(&p,
2b0143b5 2564 req->r_inode ? req->r_inode : d_inode(req->r_dentry),
719a2514
YZ		 2565 mds, req->r_inode_drop, req->r_inode_unless,
2566 req->r_op == CEPH_MDS_OP_READDIR);
2f2dc053
SW		 2567 if (req->r_dentry_drop)
2568 releases += ceph_encode_dentry_release(&p, req->r_dentry,
3dd69aab 2569 req->r_parent, mds, req->r_dentry_drop,
ca6c8ae0 2570 req->r_dentry_unless);
2f2dc053
SW		 2571 if (req->r_old_dentry_drop)
2572 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
ca6c8ae0
JL		 2573 req->r_old_dentry_dir, mds,
2574 req->r_old_dentry_drop,
2575 req->r_old_dentry_unless);
2f2dc053
SW		 2576 if (req->r_old_inode_drop)
2577 releases += ceph_encode_inode_release(&p,
2b0143b5 2578 d_inode(req->r_old_dentry),
2f2dc053 2579 mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
6e6f0923
YZ		 2580
2581 if (drop_cap_releases) {
2582 releases = 0;
2583 p = msg->front.iov_base + req->r_request_release_offset;
2584 }
2585
2f2dc053
SW		 2586 head->num_releases = cpu_to_le16(releases);
2587
b8e69066 2588 /* time stamp */
1f041a89
YZ		 2589 {
2590 struct ceph_timespec ts;
0ed1e90a 2591 ceph_encode_timespec64(&ts, &req->r_stamp);
1f041a89
YZ		 2592 ceph_encode_copy(&p, &ts, sizeof(ts));
2593 }
b8e69066 2594
b682c6d4
XL		 2595 if (WARN_ON_ONCE(p > end)) {
2596 ceph_msg_put(msg);
2597 msg = ERR_PTR(-ERANGE);
2598 goto out_free2;
2599 }
2600
2f2dc053
SW		 2601 msg->front.iov_len = p - msg->front.iov_base;
2602 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2603
25e6bae3
YZ		 2604 if (req->r_pagelist) {
2605 struct ceph_pagelist *pagelist = req->r_pagelist;
25e6bae3
YZ		 2606 ceph_msg_data_add_pagelist(msg, pagelist);
2607 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2608 } else {
2609 msg->hdr.data_len = 0;
ebf18f47 2610 }
02afca6c 2611
2f2dc053
SW		 2612 msg->hdr.data_off = cpu_to_le16(0);
2613
2614out_free2:
2615 if (freepath2)
f77f21bb 2616 ceph_mdsc_free_path((char *)path2, pathlen2);
2f2dc053
SW		 2617out_free1:
2618 if (freepath1)
f77f21bb 2619 ceph_mdsc_free_path((char *)path1, pathlen1);
2f2dc053
SW		 2620out:
2621 return msg;
2622}
2623
2624/*
2625 * called under mdsc->mutex if error, under no mutex if
2626 * success.
2627 */
2628static void complete_request(struct ceph_mds_client *mdsc,
2629 struct ceph_mds_request *req)
2630{
70c94820
XL		 2631 req->r_end_latency = ktime_get();
2632
2f2dc053
SW		 2633 if (req->r_callback)
2634 req->r_callback(mdsc, req);
111c7081 2635 complete_all(&req->r_completion);
2f2dc053
SW		 2636}
2637
2638/*
2639 * called under mdsc->mutex
2640 */
2641static int __prepare_send_request(struct ceph_mds_client *mdsc,
2642 struct ceph_mds_request *req,
6e6f0923 2643 int mds, bool drop_cap_releases)
2f2dc053
SW		 2644{
2645 struct ceph_mds_request_head *rhead;
2646 struct ceph_msg *msg;
2647 int flags = 0;
2648
2f2dc053 2649 req->r_attempts++;
e55b71f8
GF		 2650 if (req->r_inode) {
2651 struct ceph_cap *cap =
2652 ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2653
2654 if (cap)
2655 req->r_sent_on_mseq = cap->mseq;
2656 else
2657 req->r_sent_on_mseq = -1;
2658 }
2f2dc053
SW		 2659 dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2660 req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2661
bc2de10d 2662 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
c5c9a0bf 2663 void *p;
01a92f17
SW		 2664 /*
2665 * Replay. Do not regenerate message (and rebuild
2666 * paths, etc.); just use the original message.
2667 * Rebuilding paths will break for renames because
2668 * d_move mangles the src name.
2669 */
2670 msg = req->r_request;
2671 rhead = msg->front.iov_base;
2672
2673 flags = le32_to_cpu(rhead->flags);
2674 flags |= CEPH_MDS_FLAG_REPLAY;
2675 rhead->flags = cpu_to_le32(flags);
2676
2677 if (req->r_target_inode)
2678 rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2679
2680 rhead->num_retry = req->r_attempts - 1;
e979cf50
SW		 2681
2682 /* remove cap/dentry releases from message */
2683 rhead->num_releases = 0;
c5c9a0bf
YZ		 2684
2685 /* time stamp */
2686 p = msg->front.iov_base + req->r_request_release_offset;
1f041a89
YZ		 2687 {
2688 struct ceph_timespec ts;
0ed1e90a 2689 ceph_encode_timespec64(&ts, &req->r_stamp);
1f041a89
YZ		 2690 ceph_encode_copy(&p, &ts, sizeof(ts));
2691 }
c5c9a0bf
YZ		 2692
2693 msg->front.iov_len = p - msg->front.iov_base;
2694 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
01a92f17
SW		 2695 return 0;
2696 }
2697
2f2dc053
SW		 2698 if (req->r_request) {
2699 ceph_msg_put(req->r_request);
2700 req->r_request = NULL;
2701 }
6e6f0923 2702 msg = create_request_message(mdsc, req, mds, drop_cap_releases);
2f2dc053 2703 if (IS_ERR(msg)) {
e1518c7c 2704 req->r_err = PTR_ERR(msg);
a79832f2 2705 return PTR_ERR(msg);
2f2dc053
SW		 2706 }
2707 req->r_request = msg;
2708
2709 rhead = msg->front.iov_base;
2f2dc053 2710 rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
bc2de10d 2711 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2f2dc053 2712 flags |= CEPH_MDS_FLAG_REPLAY;
3bb48b41
JL		 2713 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
2714 flags |= CEPH_MDS_FLAG_ASYNC;
3dd69aab 2715 if (req->r_parent)
2f2dc053
SW		 2716 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2717 rhead->flags = cpu_to_le32(flags);
2718 rhead->num_fwd = req->r_num_fwd;
2719 rhead->num_retry = req->r_attempts - 1;
2720
3dd69aab 2721 dout(" r_parent = %p\n", req->r_parent);
2f2dc053
SW		 2722 return 0;
2723}
2724
9cf54563
XL		 2725/*
2726 * called under mdsc->mutex
2727 */
2728static int __send_request(struct ceph_mds_client *mdsc,
2729 struct ceph_mds_session *session,
2730 struct ceph_mds_request *req,
2731 bool drop_cap_releases)
2732{
2733 int err;
2734
2735 err = __prepare_send_request(mdsc, req, session->s_mds,
2736 drop_cap_releases);
2737 if (!err) {
2738 ceph_msg_get(req->r_request);
2739 ceph_con_send(&session->s_con, req->r_request);
2740 }
2741
2742 return err;
2743}
2744
2f2dc053
SW		 2745/*
2746 * send request, or put it on the appropriate wait list.
2747 */
d5548492 2748static void __do_request(struct ceph_mds_client *mdsc,
2f2dc053
SW		 2749 struct ceph_mds_request *req)
2750{
2751 struct ceph_mds_session *session = NULL;
2752 int mds = -1;
48fec5d0 2753 int err = 0;
c4853e97 2754 bool random;
2f2dc053 2755
bc2de10d
JL		 2756 if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2757 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
eb1b8af3 2758 __unregister_request(mdsc, req);
d5548492 2759 return;
eb1b8af3 2760 }
2f2dc053
SW		 2761
2762 if (req->r_timeout &&
2763 time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2764 dout("do_request timed out\n");
8ccf7fcc 2765 err = -ETIMEDOUT;
2f2dc053
SW		 2766 goto finish;
2767 }
52953d55 2768 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
48fec5d0
YZ		 2769 dout("do_request forced umount\n");
2770 err = -EIO;
2771 goto finish;
2772 }
52953d55 2773 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
e9e427f0
YZ		 2774 if (mdsc->mdsmap_err) {
2775 err = mdsc->mdsmap_err;
2776 dout("do_request mdsmap err %d\n", err);
2777 goto finish;
2778 }
cc8e8342
YZ		 2779 if (mdsc->mdsmap->m_epoch == 0) {
2780 dout("do_request no mdsmap, waiting for map\n");
2781 list_add(&req->r_wait, &mdsc->waiting_for_map);
d5548492 2782 return;
cc8e8342 2783 }
e9e427f0
YZ		 2784 if (!(mdsc->fsc->mount_options->flags &
2785 CEPH_MOUNT_OPT_MOUNTWAIT) &&
2786 !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
97820058 2787 err = -EHOSTUNREACH;
e9e427f0
YZ		 2788 goto finish;
2789 }
2790 }
2f2dc053 2791
dc69e2e9
SW		 2792 put_request_session(req);
2793
c4853e97 2794 mds = __choose_mds(mdsc, req, &random);
2f2dc053
SW		 2795 if (mds < 0 ||
2796 ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
3bb48b41
JL		 2797 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2798 err = -EJUKEBOX;
2799 goto finish;
2800 }
2f2dc053
SW		 2801 dout("do_request no mds or not active, waiting for map\n");
2802 list_add(&req->r_wait, &mdsc->waiting_for_map);
d5548492 2803 return;
2f2dc053
SW		 2804 }
2805
2806 /* get, open session */
2807 session = __ceph_lookup_mds_session(mdsc, mds);
9c423956 2808 if (!session) {
2f2dc053 2809 session = register_session(mdsc, mds);
9c423956
SW		 2810 if (IS_ERR(session)) {
2811 err = PTR_ERR(session);
2812 goto finish;
2813 }
2814 }
5b3248c6 2815 req->r_session = ceph_get_mds_session(session);
dc69e2e9 2816
2f2dc053 2817 dout("do_request mds%d session %p state %s\n", mds, session,
a687ecaf 2818 ceph_session_state_name(session->s_state));
2f2dc053
SW		 2819 if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2820 session->s_state != CEPH_MDS_SESSION_HUNG) {
fcff415c
YZ		 2821 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2822 err = -EACCES;
2823 goto out_session;
2824 }
3bb48b41
JL		 2825 /*
2826 * We cannot queue async requests since the caps and delegated
2827 * inodes are bound to the session. Just return -EJUKEBOX and
2828 * let the caller retry a sync request in that case.
2829 */
2830 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2831 err = -EJUKEBOX;
2832 goto out_session;
2833 }
2f2dc053 2834 if (session->s_state == CEPH_MDS_SESSION_NEW ||
c4853e97 2835 session->s_state == CEPH_MDS_SESSION_CLOSING) {
b682c6d4
XL		 2836 err = __open_session(mdsc, session);
2837 if (err)
2838 goto out_session;
c4853e97
XL		 2839 /* retry the same mds later */
2840 if (random)
2841 req->r_resend_mds = mds;
2842 }
2f2dc053
SW		 2843 list_add(&req->r_wait, &session->s_waiting);
2844 goto out_session;
2845 }
2846
2847 /* send request */
2f2dc053
SW		 2848 req->r_resend_mds = -1; /* forget any previous mds hint */
2849
2850 if (req->r_request_started == 0) /* note request start time */
2851 req->r_request_started = jiffies;
2852
9cf54563 2853 err = __send_request(mdsc, session, req, false);
2f2dc053
SW		 2854
2855out_session:
2856 ceph_put_mds_session(session);
48fec5d0
YZ		 2857finish:
2858 if (err) {
2859 dout("__do_request early error %d\n", err);
2860 req->r_err = err;
2861 complete_request(mdsc, req);
2862 __unregister_request(mdsc, req);
2863 }
d5548492 2864 return;
2f2dc053
SW		 2865}
2866
2867/*
2868 * called under mdsc->mutex
2869 */
2870static void __wake_requests(struct ceph_mds_client *mdsc,
2871 struct list_head *head)
2872{
ed75ec2c
YZ		 2873 struct ceph_mds_request *req;
2874 LIST_HEAD(tmp_list);
2875
2876 list_splice_init(head, &tmp_list);
2f2dc053 2877
ed75ec2c
YZ		 2878 while (!list_empty(&tmp_list)) {
2879 req = list_entry(tmp_list.next,
2880 struct ceph_mds_request, r_wait);
2f2dc053 2881 list_del_init(&req->r_wait);
7971bd92 2882 dout(" wake request %p tid %llu\n", req, req->r_tid);
2f2dc053
SW		 2883 __do_request(mdsc, req);
2884 }
2885}
2886
2887/*
2888 * Wake up threads with requests pending for @mds, so that they can
29790f26 2889 * resubmit their requests to a possibly different mds.
2f2dc053 2890 */
29790f26 2891static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2f2dc053 2892{
44ca18f2 2893 struct ceph_mds_request *req;
282c1052 2894 struct rb_node *p = rb_first(&mdsc->request_tree);
2f2dc053
SW		 2895
2896 dout("kick_requests mds%d\n", mds);
282c1052 2897 while (p) {
44ca18f2 2898 req = rb_entry(p, struct ceph_mds_request, r_node);
282c1052 2899 p = rb_next(p);
bc2de10d 2900 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
44ca18f2 2901 continue;
3de22be6
YZ		 2902 if (req->r_attempts > 0)
2903 continue; /* only new requests */
44ca18f2
SW		 2904 if (req->r_session &&
2905 req->r_session->s_mds == mds) {
2906 dout(" kicking tid %llu\n", req->r_tid);
03974e81 2907 list_del_init(&req->r_wait);
44ca18f2 2908 __do_request(mdsc, req);
2f2dc053
SW		 2909 }
2910 }
2911}
2912
86bda539 2913int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
2f2dc053
SW		 2914 struct ceph_mds_request *req)
2915{
891f3f5a 2916 int err = 0;
86bda539
JL		 2917
2918 /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
2919 if (req->r_inode)
2920 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
9c1c2b35 2921 if (req->r_parent) {
719a2514
YZ		 2922 struct ceph_inode_info *ci = ceph_inode(req->r_parent);
2923 int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
2924 CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
2925 spin_lock(&ci->i_ceph_lock);
2926 ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
2927 __ceph_touch_fmode(ci, mdsc, fmode);
2928 spin_unlock(&ci->i_ceph_lock);
9c1c2b35
JL		 2929 ihold(req->r_parent);
2930 }
86bda539
JL		 2931 if (req->r_old_dentry_dir)
2932 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2933 CEPH_CAP_PIN);
2934
891f3f5a
JL		 2935 if (req->r_inode) {
2936 err = ceph_wait_on_async_create(req->r_inode);
2937 if (err) {
2938 dout("%s: wait for async create returned: %d\n",
2939 __func__, err);
2940 return err;
2941 }
2942 }
2943
2944 if (!err && req->r_old_inode) {
2945 err = ceph_wait_on_async_create(req->r_old_inode);
2946 if (err) {
2947 dout("%s: wait for async create returned: %d\n",
2948 __func__, err);
2949 return err;
2950 }
2951 }
2952
86bda539 2953 dout("submit_request on %p for inode %p\n", req, dir);
2f2dc053 2954 mutex_lock(&mdsc->mutex);
86bda539 2955 __register_request(mdsc, req, dir);
2f2dc053 2956 __do_request(mdsc, req);
86bda539 2957 err = req->r_err;
2f2dc053 2958 mutex_unlock(&mdsc->mutex);
86bda539 2959 return err;
2f2dc053
SW		 2960}
2961
8340f22c
JL		 2962static int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
2963 struct ceph_mds_request *req)
2f2dc053
SW		 2964{
2965 int err;
2966
e1518c7c 2967 /* wait */
e1518c7c 2968 dout("do_request waiting\n");
5be73034 2969 if (!req->r_timeout && req->r_wait_for_completion) {
9280be24 2970 err = req->r_wait_for_completion(mdsc, req);
e1518c7c 2971 } else {
5be73034
ID		 2972 long timeleft = wait_for_completion_killable_timeout(
2973 &req->r_completion,
2974 ceph_timeout_jiffies(req->r_timeout));
2975 if (timeleft > 0)
2976 err = 0;
2977 else if (!timeleft)
8ccf7fcc 2978 err = -ETIMEDOUT; /* timed out */
5be73034
ID		 2979 else
2980 err = timeleft; /* killed */
e1518c7c
SW		 2981 }
2982 dout("do_request waited, got %d\n", err);
2983 mutex_lock(&mdsc->mutex);
5b1daecd 2984
e1518c7c 2985 /* only abort if we didn't race with a real reply */
bc2de10d 2986 if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
e1518c7c
SW		 2987 err = le32_to_cpu(req->r_reply_info.head->result);
2988 } else if (err < 0) {
2989 dout("aborted request %lld with %d\n", req->r_tid, err);
b4556396
SW		 2990
2991 /*
2992 * ensure we aren't running concurrently with
2993 * ceph_fill_trace or ceph_readdir_prepopulate, which
2994 * rely on locks (dir mutex) held by our caller.
2995 */
2996 mutex_lock(&req->r_fill_mutex);
e1518c7c 2997 req->r_err = err;
bc2de10d 2998 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
b4556396 2999 mutex_unlock(&req->r_fill_mutex);
5b1daecd 3000
3dd69aab 3001 if (req->r_parent &&
167c9e35
SW		 3002 (req->r_op & CEPH_MDS_OP_WRITE))
3003 ceph_invalidate_dir_request(req);
2f2dc053 3004 } else {
e1518c7c 3005 err = req->r_err;
2f2dc053 3006 }
2f2dc053 3007
e1518c7c 3008 mutex_unlock(&mdsc->mutex);
8340f22c
JL		 3009 return err;
3010}
3011
3012/*
3013 * Synchrously perform an mds request. Take care of all of the
3014 * session setup, forwarding, retry details.
3015 */
3016int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
3017 struct inode *dir,
3018 struct ceph_mds_request *req)
3019{
3020 int err;
3021
3022 dout("do_request on %p\n", req);
3023
3024 /* issue */
3025 err = ceph_mdsc_submit_request(mdsc, dir, req);
3026 if (!err)
3027 err = ceph_mdsc_wait_request(mdsc, req);
2f2dc053
SW		 3028 dout("do_request %p done, result %d\n", req, err);
3029 return err;
3030}
3031
167c9e35 3032/*
2f276c51 3033 * Invalidate dir's completeness, dentry lease state on an aborted MDS
167c9e35
SW		 3034 * namespace request.
3035 */
3036void ceph_invalidate_dir_request(struct ceph_mds_request *req)
3037{
8d8f371c
YZ		 3038 struct inode *dir = req->r_parent;
3039 struct inode *old_dir = req->r_old_dentry_dir;
167c9e35 3040
8d8f371c 3041 dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
167c9e35 3042
8d8f371c
YZ		 3043 ceph_dir_clear_complete(dir);
3044 if (old_dir)
3045 ceph_dir_clear_complete(old_dir);
167c9e35
SW		 3046 if (req->r_dentry)
3047 ceph_invalidate_dentry_lease(req->r_dentry);
3048 if (req->r_old_dentry)
3049 ceph_invalidate_dentry_lease(req->r_old_dentry);
3050}
3051
2f2dc053
SW		 3052/*
3053 * Handle mds reply.
3054 *
3055 * We take the session mutex and parse and process the reply immediately.
3056 * This preserves the logical ordering of replies, capabilities, etc., sent
3057 * by the MDS as they are applied to our local cache.
3058 */
3059static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
3060{
3061 struct ceph_mds_client *mdsc = session->s_mdsc;
3062 struct ceph_mds_request *req;
3063 struct ceph_mds_reply_head *head = msg->front.iov_base;
3064 struct ceph_mds_reply_info_parsed *rinfo; /* parsed reply info */
982d6011 3065 struct ceph_snap_realm *realm;
2f2dc053
SW		 3066 u64 tid;
3067 int err, result;
2600d2dd 3068 int mds = session->s_mds;
2f2dc053 3069
2f2dc053
SW		 3070 if (msg->front.iov_len < sizeof(*head)) {
3071 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
9ec7cab1 3072 ceph_msg_dump(msg);
2f2dc053
SW		 3073 return;
3074 }
3075
3076 /* get request, session */
6df058c0 3077 tid = le64_to_cpu(msg->hdr.tid);
2f2dc053 3078 mutex_lock(&mdsc->mutex);
fcd00b68 3079 req = lookup_get_request(mdsc, tid);
2f2dc053
SW		 3080 if (!req) {
3081 dout("handle_reply on unknown tid %llu\n", tid);
3082 mutex_unlock(&mdsc->mutex);
3083 return;
3084 }
3085 dout("handle_reply %p\n", req);
2f2dc053
SW		 3086
3087 /* correct session? */
d96d6049 3088 if (req->r_session != session) {
2f2dc053
SW		 3089 pr_err("mdsc_handle_reply got %llu on session mds%d"
3090 " not mds%d\n", tid, session->s_mds,
3091 req->r_session ? req->r_session->s_mds : -1);
3092 mutex_unlock(&mdsc->mutex);
3093 goto out;
3094 }
3095
3096 /* dup? */
bc2de10d
JL		 3097 if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3098 (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
f3ae1b97 3099 pr_warn("got a dup %s reply on %llu from mds%d\n",
2f2dc053
SW		 3100 head->safe ? "safe" : "unsafe", tid, mds);
3101 mutex_unlock(&mdsc->mutex);
3102 goto out;
3103 }
bc2de10d 3104 if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
f3ae1b97 3105 pr_warn("got unsafe after safe on %llu from mds%d\n",
85792d0d
SW		 3106 tid, mds);
3107 mutex_unlock(&mdsc->mutex);
3108 goto out;
3109 }
2f2dc053
SW		 3110
3111 result = le32_to_cpu(head->result);
3112
3113 /*
e55b71f8
GF		 3114 * Handle an ESTALE
3115 * if we're not talking to the authority, send to them
3116 * if the authority has changed while we weren't looking,
3117 * send to new authority
3118 * Otherwise we just have to return an ESTALE
2f2dc053
SW		 3119 */
3120 if (result == -ESTALE) {
4c069a58 3121 dout("got ESTALE on request %llu\n", req->r_tid);
51da8e8c 3122 req->r_resend_mds = -1;
ca18bede 3123 if (req->r_direct_mode != USE_AUTH_MDS) {
4c069a58 3124 dout("not using auth, setting for that now\n");
e55b71f8 3125 req->r_direct_mode = USE_AUTH_MDS;
2f2dc053
SW		 3126 __do_request(mdsc, req);
3127 mutex_unlock(&mdsc->mutex);
3128 goto out;
e55b71f8 3129 } else {
c4853e97 3130 int mds = __choose_mds(mdsc, req, NULL);
ca18bede 3131 if (mds >= 0 && mds != req->r_session->s_mds) {
4c069a58 3132 dout("but auth changed, so resending\n");
e55b71f8
GF		 3133 __do_request(mdsc, req);
3134 mutex_unlock(&mdsc->mutex);
3135 goto out;
3136 }
2f2dc053 3137 }
4c069a58 3138 dout("have to return ESTALE on request %llu\n", req->r_tid);
2f2dc053
SW		 3139 }
3140
e55b71f8 3141
2f2dc053 3142 if (head->safe) {
bc2de10d 3143 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
2f2dc053 3144 __unregister_request(mdsc, req);
2f2dc053 3145
07edc057
XL		 3146 /* last request during umount? */
3147 if (mdsc->stopping && !__get_oldest_req(mdsc))
3148 complete_all(&mdsc->safe_umount_waiters);
3149
bc2de10d 3150 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2f2dc053
SW		 3151 /*
3152 * We already handled the unsafe response, now do the
3153 * cleanup. No need to examine the response; the MDS
3154 * doesn't include any result info in the safe
3155 * response. And even if it did, there is nothing
3156 * useful we could do with a revised return value.
3157 */
3158 dout("got safe reply %llu, mds%d\n", tid, mds);
2f2dc053 3159
2f2dc053
SW		 3160 mutex_unlock(&mdsc->mutex);
3161 goto out;
3162 }
e1518c7c 3163 } else {
bc2de10d 3164 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
2f2dc053
SW		 3165 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3166 }
3167
3168 dout("handle_reply tid %lld result %d\n", tid, result);
3169 rinfo = &req->r_reply_info;
b37fe1f9 3170 if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
d4846487 3171 err = parse_reply_info(session, msg, rinfo, (u64)-1);
b37fe1f9 3172 else
d4846487 3173 err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);
2f2dc053
SW		 3174 mutex_unlock(&mdsc->mutex);
3175
3176 mutex_lock(&session->s_mutex);
3177 if (err < 0) {
25933abd 3178 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
9ec7cab1 3179 ceph_msg_dump(msg);
2f2dc053
SW		 3180 goto out_err;
3181 }
3182
3183 /* snap trace */
982d6011 3184 realm = NULL;
2f2dc053
SW		 3185 if (rinfo->snapblob_len) {
3186 down_write(&mdsc->snap_rwsem);
3187 ceph_update_snap_trace(mdsc, rinfo->snapblob,
982d6011
YZ		 3188 rinfo->snapblob + rinfo->snapblob_len,
3189 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3190 &realm);
2f2dc053
SW		 3191 downgrade_write(&mdsc->snap_rwsem);
3192 } else {
3193 down_read(&mdsc->snap_rwsem);
3194 }
3195
3196 /* insert trace into our cache */
b4556396 3197 mutex_lock(&req->r_fill_mutex);
315f2408 3198 current->journal_info = req;
f5a03b08 3199 err = ceph_fill_trace(mdsc->fsc->sb, req);
2f2dc053 3200 if (err == 0) {
6e8575fa 3201 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
81c6aea5 3202 req->r_op == CEPH_MDS_OP_LSSNAP))
2f2dc053 3203 ceph_readdir_prepopulate(req, req->r_session);
2f2dc053 3204 }
315f2408 3205 current->journal_info = NULL;
b4556396 3206 mutex_unlock(&req->r_fill_mutex);
2f2dc053
SW		 3207
3208 up_read(&mdsc->snap_rwsem);
982d6011
YZ		 3209 if (realm)
3210 ceph_put_snap_realm(mdsc, realm);
68cd5b4b 3211
fe33032d
YZ		 3212 if (err == 0) {
3213 if (req->r_target_inode &&
3214 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3215 struct ceph_inode_info *ci =
3216 ceph_inode(req->r_target_inode);
3217 spin_lock(&ci->i_unsafe_lock);
3218 list_add_tail(&req->r_unsafe_target_item,
3219 &ci->i_unsafe_iops);
3220 spin_unlock(&ci->i_unsafe_lock);
3221 }
3222
3223 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
68cd5b4b 3224 }
2f2dc053 3225out_err:
e1518c7c 3226 mutex_lock(&mdsc->mutex);
bc2de10d 3227 if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
e1518c7c
SW		 3228 if (err) {
3229 req->r_err = err;
3230 } else {
5fdb1389 3231 req->r_reply = ceph_msg_get(msg);
bc2de10d 3232 set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
e1518c7c 3233 }
2f2dc053 3234 } else {
e1518c7c 3235 dout("reply arrived after request %lld was aborted\n", tid);
2f2dc053 3236 }
e1518c7c 3237 mutex_unlock(&mdsc->mutex);
2f2dc053 3238
2f2dc053
SW		 3239 mutex_unlock(&session->s_mutex);
3240
3241 /* kick calling process */
3242 complete_request(mdsc, req);
70c94820
XL		 3243
3244 ceph_update_metadata_latency(&mdsc->metric, req->r_start_latency,
3245 req->r_end_latency, err);
2f2dc053
SW		 3246out:
3247 ceph_mdsc_put_request(req);
3248 return;
3249}
3250
3251
3252
3253/*
3254 * handle mds notification that our request has been forwarded.
3255 */
2600d2dd
SW		 3256static void handle_forward(struct ceph_mds_client *mdsc,
3257 struct ceph_mds_session *session,
3258 struct ceph_msg *msg)
2f2dc053
SW		 3259{
3260 struct ceph_mds_request *req;
a1ea787c 3261 u64 tid = le64_to_cpu(msg->hdr.tid);
2f2dc053
SW		 3262 u32 next_mds;
3263 u32 fwd_seq;
2f2dc053
SW		 3264 int err = -EINVAL;
3265 void *p = msg->front.iov_base;
3266 void *end = p + msg->front.iov_len;
2f2dc053 3267
a1ea787c 3268 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 3269 next_mds = ceph_decode_32(&p);
3270 fwd_seq = ceph_decode_32(&p);
2f2dc053
SW		 3271
3272 mutex_lock(&mdsc->mutex);
fcd00b68 3273 req = lookup_get_request(mdsc, tid);
2f2dc053 3274 if (!req) {
2a8e5e36 3275 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
2f2dc053
SW		 3276 goto out; /* dup reply? */
3277 }
3278
bc2de10d 3279 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
2a8e5e36
SW		 3280 dout("forward tid %llu aborted, unregistering\n", tid);
3281 __unregister_request(mdsc, req);
3282 } else if (fwd_seq <= req->r_num_fwd) {
3283 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
2f2dc053
SW		 3284 tid, next_mds, req->r_num_fwd, fwd_seq);
3285 } else {
3286 /* resend. forward race not possible; mds would drop */
2a8e5e36
SW		 3287 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3288 BUG_ON(req->r_err);
bc2de10d 3289 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3de22be6 3290 req->r_attempts = 0;
2f2dc053
SW		 3291 req->r_num_fwd = fwd_seq;
3292 req->r_resend_mds = next_mds;
3293 put_request_session(req);
3294 __do_request(mdsc, req);
3295 }
3296 ceph_mdsc_put_request(req);
3297out:
3298 mutex_unlock(&mdsc->mutex);
3299 return;
3300
3301bad:
3302 pr_err("mdsc_handle_forward decode error err=%d\n", err);
3303}
3304
131d7eb4 3305static int __decode_session_metadata(void **p, void *end,
0b98acd6 3306 bool *blocklisted)
84bf3950
YZ		 3307{
3308 /* map<string,string> */
3309 u32 n;
131d7eb4 3310 bool err_str;
84bf3950
YZ		 3311 ceph_decode_32_safe(p, end, n, bad);
3312 while (n-- > 0) {
3313 u32 len;
3314 ceph_decode_32_safe(p, end, len, bad);
3315 ceph_decode_need(p, end, len, bad);
131d7eb4 3316 err_str = !strncmp(*p, "error_string", len);
84bf3950
YZ		 3317 *p += len;
3318 ceph_decode_32_safe(p, end, len, bad);
3319 ceph_decode_need(p, end, len, bad);
4bb926e8
ID		 3320 /*
3321 * Match "blocklisted (blacklisted)" from newer MDSes,
3322 * or "blacklisted" from older MDSes.
3323 */
131d7eb4 3324 if (err_str && strnstr(*p, "blacklisted", len))
0b98acd6 3325 *blocklisted = true;
84bf3950
YZ		 3326 *p += len;
3327 }
3328 return 0;
3329bad:
3330 return -1;
3331}
3332
2f2dc053
SW		 3333/*
3334 * handle a mds session control message
3335 */
3336static void handle_session(struct ceph_mds_session *session,
3337 struct ceph_msg *msg)
3338{
3339 struct ceph_mds_client *mdsc = session->s_mdsc;
84bf3950
YZ		 3340 int mds = session->s_mds;
3341 int msg_version = le16_to_cpu(msg->hdr.version);
3342 void *p = msg->front.iov_base;
3343 void *end = p + msg->front.iov_len;
3344 struct ceph_mds_session_head *h;
2f2dc053 3345 u32 op;
0fa82633 3346 u64 seq, features = 0;
2f2dc053 3347 int wake = 0;
0b98acd6 3348 bool blocklisted = false;
2f2dc053 3349
2f2dc053 3350 /* decode */
84bf3950
YZ		 3351 ceph_decode_need(&p, end, sizeof(*h), bad);
3352 h = p;
3353 p += sizeof(*h);
3354
2f2dc053
SW		 3355 op = le32_to_cpu(h->op);
3356 seq = le64_to_cpu(h->seq);
3357
84bf3950
YZ		 3358 if (msg_version >= 3) {
3359 u32 len;
3360 /* version >= 2, metadata */
0b98acd6 3361 if (__decode_session_metadata(&p, end, &blocklisted) < 0)
84bf3950
YZ		 3362 goto bad;
3363 /* version >= 3, feature bits */
3364 ceph_decode_32_safe(&p, end, len, bad);
02e37571
JL		 3365 if (len) {
3366 ceph_decode_64_safe(&p, end, features, bad);
3367 p += len - sizeof(features);
3368 }
84bf3950
YZ		 3369 }
3370
2f2dc053 3371 mutex_lock(&mdsc->mutex);
0a07fc8c 3372 if (op == CEPH_SESSION_CLOSE) {
5b3248c6 3373 ceph_get_mds_session(session);
2600d2dd 3374 __unregister_session(mdsc, session);
0a07fc8c 3375 }
2f2dc053
SW		 3376 /* FIXME: this ttl calculation is generous */
3377 session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3378 mutex_unlock(&mdsc->mutex);
3379
3380 mutex_lock(&session->s_mutex);
3381
3382 dout("handle_session mds%d %s %p state %s seq %llu\n",
3383 mds, ceph_session_op_name(op), session,
a687ecaf 3384 ceph_session_state_name(session->s_state), seq);
2f2dc053
SW		 3385
3386 if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3387 session->s_state = CEPH_MDS_SESSION_OPEN;
3388 pr_info("mds%d came back\n", session->s_mds);
3389 }
3390
3391 switch (op) {
3392 case CEPH_SESSION_OPEN:
29790f26
SW		 3393 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3394 pr_info("mds%d reconnect success\n", session->s_mds);
2f2dc053 3395 session->s_state = CEPH_MDS_SESSION_OPEN;
84bf3950 3396 session->s_features = features;
2f2dc053 3397 renewed_caps(mdsc, session, 0);
18f473b3
XL		 3398 if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT, &session->s_features))
3399 metric_schedule_delayed(&mdsc->metric);
2f2dc053
SW		 3400 wake = 1;
3401 if (mdsc->stopping)
3402 __close_session(mdsc, session);
3403 break;
3404
3405 case CEPH_SESSION_RENEWCAPS:
3406 if (session->s_renew_seq == seq)
3407 renewed_caps(mdsc, session, 1);
3408 break;
3409
3410 case CEPH_SESSION_CLOSE:
29790f26
SW		 3411 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3412 pr_info("mds%d reconnect denied\n", session->s_mds);
4d681c2f 3413 session->s_state = CEPH_MDS_SESSION_CLOSED;
1c841a96 3414 cleanup_session_requests(mdsc, session);
2f2dc053 3415 remove_session_caps(session);
656e4382 3416 wake = 2; /* for good measure */
f3c60c59 3417 wake_up_all(&mdsc->session_close_wq);
2f2dc053
SW		 3418 break;
3419
3420 case CEPH_SESSION_STALE:
3421 pr_info("mds%d caps went stale, renewing\n",
3422 session->s_mds);
d8fb02ab 3423 spin_lock(&session->s_gen_ttl_lock);
2f2dc053 3424 session->s_cap_gen++;
1ce208a6 3425 session->s_cap_ttl = jiffies - 1;
d8fb02ab 3426 spin_unlock(&session->s_gen_ttl_lock);
2f2dc053
SW		 3427 send_renew_caps(mdsc, session);
3428 break;
3429
3430 case CEPH_SESSION_RECALL_STATE:
e30ee581 3431 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
2f2dc053
SW		 3432 break;
3433
186e4f7a
YZ		 3434 case CEPH_SESSION_FLUSHMSG:
3435 send_flushmsg_ack(mdsc, session, seq);
3436 break;
3437
03f4fcb0
YZ		 3438 case CEPH_SESSION_FORCE_RO:
3439 dout("force_session_readonly %p\n", session);
3440 spin_lock(&session->s_cap_lock);
3441 session->s_readonly = true;
3442 spin_unlock(&session->s_cap_lock);
d2f8bb27 3443 wake_up_session_caps(session, FORCE_RO);
03f4fcb0
YZ		 3444 break;
3445
fcff415c
YZ		 3446 case CEPH_SESSION_REJECT:
3447 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3448 pr_info("mds%d rejected session\n", session->s_mds);
3449 session->s_state = CEPH_MDS_SESSION_REJECTED;
3450 cleanup_session_requests(mdsc, session);
3451 remove_session_caps(session);
0b98acd6
ID		 3452 if (blocklisted)
3453 mdsc->fsc->blocklisted = true;
fcff415c
YZ		 3454 wake = 2; /* for good measure */
3455 break;
3456
2f2dc053
SW		 3457 default:
3458 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3459 WARN_ON(1);
3460 }
3461
3462 mutex_unlock(&session->s_mutex);
3463 if (wake) {
3464 mutex_lock(&mdsc->mutex);
3465 __wake_requests(mdsc, &session->s_waiting);
656e4382
YZ		 3466 if (wake == 2)
3467 kick_requests(mdsc, mds);
2f2dc053
SW		 3468 mutex_unlock(&mdsc->mutex);
3469 }
0a07fc8c
YZ		 3470 if (op == CEPH_SESSION_CLOSE)
3471 ceph_put_mds_session(session);
2f2dc053
SW		 3472 return;
3473
3474bad:
3475 pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3476 (int)msg->front.iov_len);
9ec7cab1 3477 ceph_msg_dump(msg);
2f2dc053
SW		 3478 return;
3479}
3480
a25949b9
JL		 3481void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
3482{
3483 int dcaps;
3484
3485 dcaps = xchg(&req->r_dir_caps, 0);
3486 if (dcaps) {
3487 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3488 ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
3489 }
3490}
3491
e64f44a8
XL		 3492void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
3493{
3494 int dcaps;
3495
3496 dcaps = xchg(&req->r_dir_caps, 0);
3497 if (dcaps) {
3498 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3499 ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
3500 dcaps);
3501 }
3502}
3503
2f2dc053
SW		 3504/*
3505 * called under session->mutex.
3506 */
3507static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3508 struct ceph_mds_session *session)
3509{
3510 struct ceph_mds_request *req, *nreq;
3de22be6 3511 struct rb_node *p;
2f2dc053
SW		 3512
3513 dout("replay_unsafe_requests mds%d\n", session->s_mds);
3514
3515 mutex_lock(&mdsc->mutex);
9cf54563
XL		 3516 list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
3517 __send_request(mdsc, session, req, true);
3de22be6
YZ		 3518
3519 /*
3520 * also re-send old requests when MDS enters reconnect stage. So that MDS
3521 * can process completed request in clientreplay stage.
3522 */
3523 p = rb_first(&mdsc->request_tree);
3524 while (p) {
3525 req = rb_entry(p, struct ceph_mds_request, r_node);
3526 p = rb_next(p);
bc2de10d 3527 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3de22be6
YZ		 3528 continue;
3529 if (req->r_attempts == 0)
3530 continue; /* only old requests */
a25949b9
JL		 3531 if (!req->r_session)
3532 continue;
3533 if (req->r_session->s_mds != session->s_mds)
3534 continue;
3535
e64f44a8 3536 ceph_mdsc_release_dir_caps_no_check(req);
a25949b9
JL		 3537
3538 __send_request(mdsc, session, req, true);
3de22be6 3539 }
2f2dc053
SW		 3540 mutex_unlock(&mdsc->mutex);
3541}
3542
81c5a148
YZ		 3543static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3544{
3545 struct ceph_msg *reply;
3546 struct ceph_pagelist *_pagelist;
3547 struct page *page;
3548 __le32 *addr;
3549 int err = -ENOMEM;
3550
3551 if (!recon_state->allow_multi)
3552 return -ENOSPC;
3553
3554 /* can't handle message that contains both caps and realm */
3555 BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3556
3557 /* pre-allocate new pagelist */
3558 _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3559 if (!_pagelist)
3560 return -ENOMEM;
3561
3562 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3563 if (!reply)
3564 goto fail_msg;
3565
3566 /* placeholder for nr_caps */
3567 err = ceph_pagelist_encode_32(_pagelist, 0);
3568 if (err < 0)
3569 goto fail;
3570
3571 if (recon_state->nr_caps) {
3572 /* currently encoding caps */
3573 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3574 if (err)
3575 goto fail;
3576 } else {
3577 /* placeholder for nr_realms (currently encoding relams) */
3578 err = ceph_pagelist_encode_32(_pagelist, 0);
3579 if (err < 0)
3580 goto fail;
3581 }
3582
3583 err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3584 if (err)
3585 goto fail;
3586
3587 page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3588 addr = kmap_atomic(page);
3589 if (recon_state->nr_caps) {
3590 /* currently encoding caps */
3591 *addr = cpu_to_le32(recon_state->nr_caps);
3592 } else {
3593 /* currently encoding relams */
3594 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3595 }
3596 kunmap_atomic(addr);
3597
3598 reply->hdr.version = cpu_to_le16(5);
3599 reply->hdr.compat_version = cpu_to_le16(4);
3600
3601 reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3602 ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3603
3604 ceph_con_send(&recon_state->session->s_con, reply);
3605 ceph_pagelist_release(recon_state->pagelist);
3606
3607 recon_state->pagelist = _pagelist;
3608 recon_state->nr_caps = 0;
3609 recon_state->nr_realms = 0;
3610 recon_state->msg_version = 5;
3611 return 0;
3612fail:
3613 ceph_msg_put(reply);
3614fail_msg:
3615 ceph_pagelist_release(_pagelist);
3616 return err;
3617}
3618
a33f6432
YZ		 3619static struct dentry* d_find_primary(struct inode *inode)
3620{
3621 struct dentry *alias, *dn = NULL;
3622
3623 if (hlist_empty(&inode->i_dentry))
3624 return NULL;
3625
3626 spin_lock(&inode->i_lock);
3627 if (hlist_empty(&inode->i_dentry))
3628 goto out_unlock;
3629
3630 if (S_ISDIR(inode->i_mode)) {
3631 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
3632 if (!IS_ROOT(alias))
3633 dn = dget(alias);
3634 goto out_unlock;
3635 }
3636
3637 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
3638 spin_lock(&alias->d_lock);
3639 if (!d_unhashed(alias) &&
3640 (ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {
3641 dn = dget_dlock(alias);
3642 }
3643 spin_unlock(&alias->d_lock);
3644 if (dn)
3645 break;
3646 }
3647out_unlock:
3648 spin_unlock(&inode->i_lock);
3649 return dn;
3650}
3651
2f2dc053
SW		 3652/*
3653 * Encode information about a cap for a reconnect with the MDS.
3654 */
a25949b9 3655static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
2f2dc053
SW		 3656 void *arg)
3657{
20cb34ae
SW		 3658 union {
3659 struct ceph_mds_cap_reconnect v2;
3660 struct ceph_mds_cap_reconnect_v1 v1;
3661 } rec;
b3f8d68f 3662 struct ceph_inode_info *ci = cap->ci;
20cb34ae
SW		 3663 struct ceph_reconnect_state *recon_state = arg;
3664 struct ceph_pagelist *pagelist = recon_state->pagelist;
a33f6432
YZ		 3665 struct dentry *dentry;
3666 char *path;
3667 int pathlen, err;
3668 u64 pathbase;
3469ed0d 3669 u64 snap_follows;
2f2dc053 3670
2f2dc053
SW		 3671 dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3672 inode, ceph_vinop(inode), cap, cap->cap_id,
3673 ceph_cap_string(cap->issued));
2f2dc053 3674
a33f6432
YZ		 3675 dentry = d_find_primary(inode);
3676 if (dentry) {
3677 /* set pathbase to parent dir when msg_version >= 2 */
3678 path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase,
3679 recon_state->msg_version >= 2);
3680 dput(dentry);
3681 if (IS_ERR(path)) {
3682 err = PTR_ERR(path);
3683 goto out_err;
3684 }
3685 } else {
3686 path = NULL;
3687 pathlen = 0;
3688 pathbase = 0;
3689 }
3690
be655596 3691 spin_lock(&ci->i_ceph_lock);
2f2dc053
SW		 3692 cap->seq = 0; /* reset cap seq */
3693 cap->issue_seq = 0; /* and issue_seq */
667ca05c 3694 cap->mseq = 0; /* and migrate_seq */
99a9c273 3695 cap->cap_gen = cap->session->s_cap_gen;
20cb34ae 3696
a25949b9 3697 /* These are lost when the session goes away */
785892fe
JL		 3698 if (S_ISDIR(inode->i_mode)) {
3699 if (cap->issued & CEPH_CAP_DIR_CREATE) {
3700 ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
3701 memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
3702 }
a25949b9 3703 cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
785892fe 3704 }
a25949b9 3705
121f22a1 3706 if (recon_state->msg_version >= 2) {
20cb34ae
SW		 3707 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3708 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3709 rec.v2.issued = cpu_to_le32(cap->issued);
3710 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
a33f6432 3711 rec.v2.pathbase = cpu_to_le64(pathbase);
ec1dff25
JL		 3712 rec.v2.flock_len = (__force __le32)
3713 ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
20cb34ae
SW		 3714 } else {
3715 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3716 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3717 rec.v1.issued = cpu_to_le32(cap->issued);
3718 rec.v1.size = cpu_to_le64(inode->i_size);
9bbeab41
AB		 3719 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3720 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
20cb34ae 3721 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
a33f6432 3722 rec.v1.pathbase = cpu_to_le64(pathbase);
20cb34ae 3723 }
3469ed0d
YZ		 3724
3725 if (list_empty(&ci->i_cap_snaps)) {
92776fd2 3726 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3469ed0d
YZ		 3727 } else {
3728 struct ceph_cap_snap *capsnap =
3729 list_first_entry(&ci->i_cap_snaps,
3730 struct ceph_cap_snap, ci_item);
3731 snap_follows = capsnap->follows;
20cb34ae 3732 }
be655596 3733 spin_unlock(&ci->i_ceph_lock);
2f2dc053 3734
121f22a1 3735 if (recon_state->msg_version >= 2) {
40819f6f 3736 int num_fcntl_locks, num_flock_locks;
4deb14a2 3737 struct ceph_filelock *flocks = NULL;
81c5a148 3738 size_t struct_len, total_len = sizeof(u64);
121f22a1 3739 u8 struct_v = 0;
39be95e9
JS		 3740
3741encode_again:
b3f8d68f
YZ		 3742 if (rec.v2.flock_len) {
3743 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3744 } else {
3745 num_fcntl_locks = 0;
3746 num_flock_locks = 0;
3747 }
4deb14a2 3748 if (num_fcntl_locks + num_flock_locks > 0) {
6da2ec56
KC		 3749 flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3750 sizeof(struct ceph_filelock),
3751 GFP_NOFS);
4deb14a2
YZ		 3752 if (!flocks) {
3753 err = -ENOMEM;
5ccedf1c 3754 goto out_err;
4deb14a2
YZ		 3755 }
3756 err = ceph_encode_locks_to_buffer(inode, flocks,
3757 num_fcntl_locks,
3758 num_flock_locks);
3759 if (err) {
3760 kfree(flocks);
3761 flocks = NULL;
3762 if (err == -ENOSPC)
3763 goto encode_again;
5ccedf1c 3764 goto out_err;
4deb14a2
YZ		 3765 }
3766 } else {
39be95e9 3767 kfree(flocks);
4deb14a2 3768 flocks = NULL;
39be95e9 3769 }
121f22a1
YZ		 3770
3771 if (recon_state->msg_version >= 3) {
3772 /* version, compat_version and struct_len */
81c5a148 3773 total_len += 2 * sizeof(u8) + sizeof(u32);
3469ed0d 3774 struct_v = 2;
121f22a1 3775 }
39be95e9
JS		 3776 /*
3777 * number of encoded locks is stable, so copy to pagelist
3778 */
121f22a1
YZ		 3779 struct_len = 2 * sizeof(u32) +
3780 (num_fcntl_locks + num_flock_locks) *
3781 sizeof(struct ceph_filelock);
3782 rec.v2.flock_len = cpu_to_le32(struct_len);
3783
a33f6432 3784 struct_len += sizeof(u32) + pathlen + sizeof(rec.v2);
121f22a1 3785
3469ed0d
YZ		 3786 if (struct_v >= 2)
3787 struct_len += sizeof(u64); /* snap_follows */
3788
121f22a1 3789 total_len += struct_len;
81c5a148
YZ		 3790
3791 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
3792 err = send_reconnect_partial(recon_state);
3793 if (err)
3794 goto out_freeflocks;
3795 pagelist = recon_state->pagelist;
5ccedf1c 3796 }
121f22a1 3797
81c5a148
YZ		 3798 err = ceph_pagelist_reserve(pagelist, total_len);
3799 if (err)
3800 goto out_freeflocks;
3801
3802 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
5ccedf1c
YZ		 3803 if (recon_state->msg_version >= 3) {
3804 ceph_pagelist_encode_8(pagelist, struct_v);
3805 ceph_pagelist_encode_8(pagelist, 1);
3806 ceph_pagelist_encode_32(pagelist, struct_len);
121f22a1 3807 }
a33f6432 3808 ceph_pagelist_encode_string(pagelist, path, pathlen);
5ccedf1c
YZ		 3809 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
3810 ceph_locks_to_pagelist(flocks, pagelist,
3811 num_fcntl_locks, num_flock_locks);
3812 if (struct_v >= 2)
3813 ceph_pagelist_encode_64(pagelist, snap_follows);
81c5a148 3814out_freeflocks:
39be95e9 3815 kfree(flocks);
3612abbd 3816 } else {
5ccedf1c 3817 err = ceph_pagelist_reserve(pagelist,
81c5a148
YZ		 3818 sizeof(u64) + sizeof(u32) +
3819 pathlen + sizeof(rec.v1));
a33f6432
YZ		 3820 if (err)
3821 goto out_err;
5ccedf1c 3822
81c5a148 3823 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
5ccedf1c
YZ		 3824 ceph_pagelist_encode_string(pagelist, path, pathlen);
3825 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
40819f6f 3826 }
44c99757 3827
5ccedf1c 3828out_err:
a33f6432
YZ		 3829 ceph_mdsc_free_path(path, pathlen);
3830 if (!err)
81c5a148
YZ		 3831 recon_state->nr_caps++;
3832 return err;
3833}
3834
3835static int encode_snap_realms(struct ceph_mds_client *mdsc,
3836 struct ceph_reconnect_state *recon_state)
3837{
3838 struct rb_node *p;
3839 struct ceph_pagelist *pagelist = recon_state->pagelist;
3840 int err = 0;
3841
3842 if (recon_state->msg_version >= 4) {
3843 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
3844 if (err < 0)
3845 goto fail;
3846 }
3847
3848 /*
3849 * snaprealms. we provide mds with the ino, seq (version), and
3850 * parent for all of our realms. If the mds has any newer info,
3851 * it will tell us.
3852 */
3853 for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
3854 struct ceph_snap_realm *realm =
3855 rb_entry(p, struct ceph_snap_realm, node);
3856 struct ceph_mds_snaprealm_reconnect sr_rec;
3857
3858 if (recon_state->msg_version >= 4) {
3859 size_t need = sizeof(u8) * 2 + sizeof(u32) +
3860 sizeof(sr_rec);
3861
3862 if (pagelist->length + need > RECONNECT_MAX_SIZE) {
3863 err = send_reconnect_partial(recon_state);
3864 if (err)
3865 goto fail;
3866 pagelist = recon_state->pagelist;
3867 }
3868
3869 err = ceph_pagelist_reserve(pagelist, need);
3870 if (err)
3871 goto fail;
3872
3873 ceph_pagelist_encode_8(pagelist, 1);
3874 ceph_pagelist_encode_8(pagelist, 1);
3875 ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
3876 }
3877
3878 dout(" adding snap realm %llx seq %lld parent %llx\n",
3879 realm->ino, realm->seq, realm->parent_ino);
3880 sr_rec.ino = cpu_to_le64(realm->ino);
3881 sr_rec.seq = cpu_to_le64(realm->seq);
3882 sr_rec.parent = cpu_to_le64(realm->parent_ino);
3883
3884 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
3885 if (err)
3886 goto fail;
3887
3888 recon_state->nr_realms++;
3889 }
3890fail:
93cea5be 3891 return err;
2f2dc053
SW		 3892}
3893
3894
3895/*
3896 * If an MDS fails and recovers, clients need to reconnect in order to
3897 * reestablish shared state. This includes all caps issued through
3898 * this session _and_ the snap_realm hierarchy. Because it's not
3899 * clear which snap realms the mds cares about, we send everything we
3900 * know about.. that ensures we'll then get any new info the
3901 * recovering MDS might have.
3902 *
3903 * This is a relatively heavyweight operation, but it's rare.
2f2dc053 3904 */
34b6c855
SW		 3905static void send_mds_reconnect(struct ceph_mds_client *mdsc,
3906 struct ceph_mds_session *session)
2f2dc053 3907{
2f2dc053 3908 struct ceph_msg *reply;
34b6c855 3909 int mds = session->s_mds;
9abf82b8 3910 int err = -ENOMEM;
81c5a148
YZ		 3911 struct ceph_reconnect_state recon_state = {
3912 .session = session,
3913 };
c8a96a31 3914 LIST_HEAD(dispose);
2f2dc053 3915
34b6c855 3916 pr_info("mds%d reconnect start\n", mds);
2f2dc053 3917
81c5a148
YZ		 3918 recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
3919 if (!recon_state.pagelist)
93cea5be 3920 goto fail_nopagelist;
93cea5be 3921
0d9c1ab3 3922 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
a79832f2 3923 if (!reply)
93cea5be 3924 goto fail_nomsg;
93cea5be 3925
d4846487
JL		 3926 xa_destroy(&session->s_delegated_inos);
3927
34b6c855
SW		 3928 mutex_lock(&session->s_mutex);
3929 session->s_state = CEPH_MDS_SESSION_RECONNECTING;
3930 session->s_seq = 0;
2f2dc053 3931
2f2dc053 3932 dout("session %p state %s\n", session,
a687ecaf 3933 ceph_session_state_name(session->s_state));
2f2dc053 3934
99a9c273
YZ		 3935 spin_lock(&session->s_gen_ttl_lock);
3936 session->s_cap_gen++;
3937 spin_unlock(&session->s_gen_ttl_lock);
3938
3939 spin_lock(&session->s_cap_lock);
03f4fcb0
YZ		 3940 /* don't know if session is readonly */
3941 session->s_readonly = 0;
99a9c273
YZ		 3942 /*
3943 * notify __ceph_remove_cap() that we are composing cap reconnect.
3944 * If a cap get released before being added to the cap reconnect,
3945 * __ceph_remove_cap() should skip queuing cap release.
3946 */
3947 session->s_cap_reconnect = 1;
e01a5946 3948 /* drop old cap expires; we're about to reestablish that state */
c8a96a31
JL		 3949 detach_cap_releases(session, &dispose);
3950 spin_unlock(&session->s_cap_lock);
3951 dispose_cap_releases(mdsc, &dispose);
e01a5946 3952
5d23371f 3953 /* trim unused caps to reduce MDS's cache rejoin time */
c0bd50e2
YZ		 3954 if (mdsc->fsc->sb->s_root)
3955 shrink_dcache_parent(mdsc->fsc->sb->s_root);
5d23371f
YZ		 3956
3957 ceph_con_close(&session->s_con);
3958 ceph_con_open(&session->s_con,
3959 CEPH_ENTITY_TYPE_MDS, mds,
3960 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
3961
3962 /* replay unsafe requests */
3963 replay_unsafe_requests(mdsc, session);
3964
81c5a148
YZ		 3965 ceph_early_kick_flushing_caps(mdsc, session);
3966
5d23371f
YZ		 3967 down_read(&mdsc->snap_rwsem);
3968
81c5a148
YZ		 3969 /* placeholder for nr_caps */
3970 err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
93cea5be
SW		 3971 if (err)
3972 goto fail;
20cb34ae 3973
81c5a148 3974 if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
121f22a1 3975 recon_state.msg_version = 3;
81c5a148
YZ		 3976 recon_state.allow_multi = true;
3977 } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
3978 recon_state.msg_version = 3;
3979 } else {
23c625ce 3980 recon_state.msg_version = 2;
81c5a148
YZ		 3981 }
3982 /* trsaverse this session's caps */
a25949b9 3983 err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
2f2dc053 3984
99a9c273
YZ		 3985 spin_lock(&session->s_cap_lock);
3986 session->s_cap_reconnect = 0;
3987 spin_unlock(&session->s_cap_lock);
3988
81c5a148
YZ		 3989 if (err < 0)
3990 goto fail;
2f2dc053 3991
81c5a148
YZ		 3992 /* check if all realms can be encoded into current message */
3993 if (mdsc->num_snap_realms) {
3994 size_t total_len =
3995 recon_state.pagelist->length +
3996 mdsc->num_snap_realms *
3997 sizeof(struct ceph_mds_snaprealm_reconnect);
3998 if (recon_state.msg_version >= 4) {
3999 /* number of realms */
4000 total_len += sizeof(u32);
4001 /* version, compat_version and struct_len */
4002 total_len += mdsc->num_snap_realms *
4003 (2 * sizeof(u8) + sizeof(u32));
4004 }
4005 if (total_len > RECONNECT_MAX_SIZE) {
4006 if (!recon_state.allow_multi) {
4007 err = -ENOSPC;
4008 goto fail;
4009 }
4010 if (recon_state.nr_caps) {
4011 err = send_reconnect_partial(&recon_state);
4012 if (err)
4013 goto fail;
4014 }
4015 recon_state.msg_version = 5;
4016 }
2f2dc053 4017 }
2f2dc053 4018
81c5a148
YZ		 4019 err = encode_snap_realms(mdsc, &recon_state);
4020 if (err < 0)
4021 goto fail;
4022
4023 if (recon_state.msg_version >= 5) {
4024 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
4025 if (err < 0)
4026 goto fail;
4027 }
44c99757 4028
81c5a148
YZ		 4029 if (recon_state.nr_caps || recon_state.nr_realms) {
4030 struct page *page =
4031 list_first_entry(&recon_state.pagelist->head,
4032 struct page, lru);
44c99757 4033 __le32 *addr = kmap_atomic(page);
81c5a148
YZ		 4034 if (recon_state.nr_caps) {
4035 WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
4036 *addr = cpu_to_le32(recon_state.nr_caps);
4037 } else if (recon_state.msg_version >= 4) {
4038 *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
4039 }
44c99757 4040 kunmap_atomic(addr);
ebf18f47 4041 }
44c99757 4042
81c5a148
YZ		 4043 reply->hdr.version = cpu_to_le16(recon_state.msg_version);
4044 if (recon_state.msg_version >= 4)
4045 reply->hdr.compat_version = cpu_to_le16(4);
e548e9b9 4046
81c5a148
YZ		 4047 reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
4048 ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
e548e9b9 4049
2f2dc053
SW		 4050 ceph_con_send(&session->s_con, reply);
4051
9abf82b8
SW		 4052 mutex_unlock(&session->s_mutex);
4053
4054 mutex_lock(&mdsc->mutex);
4055 __wake_requests(mdsc, &session->s_waiting);
4056 mutex_unlock(&mdsc->mutex);
4057
2f2dc053 4058 up_read(&mdsc->snap_rwsem);
81c5a148 4059 ceph_pagelist_release(recon_state.pagelist);
2f2dc053
SW		 4060 return;
4061
93cea5be 4062fail:
2f2dc053 4063 ceph_msg_put(reply);
9abf82b8
SW		 4064 up_read(&mdsc->snap_rwsem);
4065 mutex_unlock(&session->s_mutex);
93cea5be 4066fail_nomsg:
81c5a148 4067 ceph_pagelist_release(recon_state.pagelist);
93cea5be 4068fail_nopagelist:
9abf82b8 4069 pr_err("error %d preparing reconnect for mds%d\n", err, mds);
9abf82b8 4070 return;
2f2dc053
SW		 4071}
4072
4073
4074/*
4075 * compare old and new mdsmaps, kicking requests
4076 * and closing out old connections as necessary
4077 *
4078 * called under mdsc->mutex.
4079 */
4080static void check_new_map(struct ceph_mds_client *mdsc,
4081 struct ceph_mdsmap *newmap,
4082 struct ceph_mdsmap *oldmap)
4083{
4084 int i;
4085 int oldstate, newstate;
4086 struct ceph_mds_session *s;
4087
4088 dout("check_new_map new %u old %u\n",
4089 newmap->m_epoch, oldmap->m_epoch);
4090
b38c9eb4 4091 for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
d37b1d99 4092 if (!mdsc->sessions[i])
2f2dc053
SW		 4093 continue;
4094 s = mdsc->sessions[i];
4095 oldstate = ceph_mdsmap_get_state(oldmap, i);
4096 newstate = ceph_mdsmap_get_state(newmap, i);
4097
0deb01c9 4098 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
2f2dc053 4099 i, ceph_mds_state_name(oldstate),
0deb01c9 4100 ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
2f2dc053 4101 ceph_mds_state_name(newstate),
0deb01c9 4102 ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
a687ecaf 4103 ceph_session_state_name(s->s_state));
2f2dc053 4104
b38c9eb4 4105 if (i >= newmap->possible_max_rank) {
6f0f597b 4106 /* force close session for stopped mds */
5b3248c6 4107 ceph_get_mds_session(s);
6f0f597b
YZ		 4108 __unregister_session(mdsc, s);
4109 __wake_requests(mdsc, &s->s_waiting);
4110 mutex_unlock(&mdsc->mutex);
2827528d 4111
6f0f597b
YZ		 4112 mutex_lock(&s->s_mutex);
4113 cleanup_session_requests(mdsc, s);
4114 remove_session_caps(s);
4115 mutex_unlock(&s->s_mutex);
2827528d 4116
6f0f597b 4117 ceph_put_mds_session(s);
2827528d 4118
6f0f597b
YZ		 4119 mutex_lock(&mdsc->mutex);
4120 kick_requests(mdsc, i);
4121 continue;
4122 }
4123
4124 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4125 ceph_mdsmap_get_addr(newmap, i),
4126 sizeof(struct ceph_entity_addr))) {
4127 /* just close it */
4128 mutex_unlock(&mdsc->mutex);
4129 mutex_lock(&s->s_mutex);
4130 mutex_lock(&mdsc->mutex);
4131 ceph_con_close(&s->s_con);
4132 mutex_unlock(&s->s_mutex);
4133 s->s_state = CEPH_MDS_SESSION_RESTARTING;
2f2dc053
SW		 4134 } else if (oldstate == newstate) {
4135 continue; /* nothing new with this mds */
4136 }
4137
4138 /*
4139 * send reconnect?
4140 */
4141 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
34b6c855
SW		 4142 newstate >= CEPH_MDS_STATE_RECONNECT) {
4143 mutex_unlock(&mdsc->mutex);
4144 send_mds_reconnect(mdsc, s);
4145 mutex_lock(&mdsc->mutex);
4146 }
2f2dc053
SW		 4147
4148 /*
29790f26 4149 * kick request on any mds that has gone active.
2f2dc053
SW		 4150 */
4151 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4152 newstate >= CEPH_MDS_STATE_ACTIVE) {
29790f26
SW		 4153 if (oldstate != CEPH_MDS_STATE_CREATING &&
4154 oldstate != CEPH_MDS_STATE_STARTING)
4155 pr_info("mds%d recovery completed\n", s->s_mds);
4156 kick_requests(mdsc, i);
ea8412b2 4157 mutex_unlock(&mdsc->mutex);
829ad4db 4158 mutex_lock(&s->s_mutex);
ea8412b2 4159 mutex_lock(&mdsc->mutex);
2f2dc053 4160 ceph_kick_flushing_caps(mdsc, s);
829ad4db 4161 mutex_unlock(&s->s_mutex);
d2f8bb27 4162 wake_up_session_caps(s, RECONNECT);
2f2dc053
SW		 4163 }
4164 }
cb170a22 4165
b38c9eb4 4166 for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
cb170a22
SW		 4167 s = mdsc->sessions[i];
4168 if (!s)
4169 continue;
4170 if (!ceph_mdsmap_is_laggy(newmap, i))
4171 continue;
4172 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4173 s->s_state == CEPH_MDS_SESSION_HUNG ||
4174 s->s_state == CEPH_MDS_SESSION_CLOSING) {
4175 dout(" connecting to export targets of laggy mds%d\n",
4176 i);
4177 __open_export_target_sessions(mdsc, s);
4178 }
4179 }
2f2dc053
SW		 4180}
4181
4182
4183
4184/*
4185 * leases
4186 */
4187
4188/*
4189 * caller must hold session s_mutex, dentry->d_lock
4190 */
4191void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
4192{
4193 struct ceph_dentry_info *di = ceph_dentry(dentry);
4194
4195 ceph_put_mds_session(di->lease_session);
4196 di->lease_session = NULL;
4197}
4198
2600d2dd
SW		 4199static void handle_lease(struct ceph_mds_client *mdsc,
4200 struct ceph_mds_session *session,
4201 struct ceph_msg *msg)
2f2dc053 4202{
3d14c5d2 4203 struct super_block *sb = mdsc->fsc->sb;
2f2dc053 4204 struct inode *inode;
2f2dc053
SW		 4205 struct dentry *parent, *dentry;
4206 struct ceph_dentry_info *di;
2600d2dd 4207 int mds = session->s_mds;
2f2dc053 4208 struct ceph_mds_lease *h = msg->front.iov_base;
1e5ea23d 4209 u32 seq;
2f2dc053 4210 struct ceph_vino vino;
2f2dc053
SW		 4211 struct qstr dname;
4212 int release = 0;
4213
2f2dc053
SW		 4214 dout("handle_lease from mds%d\n", mds);
4215
4216 /* decode */
4217 if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
4218 goto bad;
4219 vino.ino = le64_to_cpu(h->ino);
4220 vino.snap = CEPH_NOSNAP;
1e5ea23d 4221 seq = le32_to_cpu(h->seq);
0fcf6c02
YZ		 4222 dname.len = get_unaligned_le32(h + 1);
4223 if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
2f2dc053 4224 goto bad;
0fcf6c02 4225 dname.name = (void *)(h + 1) + sizeof(u32);
2f2dc053 4226
2f2dc053
SW		 4227 /* lookup inode */
4228 inode = ceph_find_inode(sb, vino);
2f90b852
SW		 4229 dout("handle_lease %s, ino %llx %p %.*s\n",
4230 ceph_lease_op_name(h->action), vino.ino, inode,
1e5ea23d 4231 dname.len, dname.name);
6cd3bcad
YZ		 4232
4233 mutex_lock(&session->s_mutex);
62575e27 4234 inc_session_sequence(session);
6cd3bcad 4235
d37b1d99 4236 if (!inode) {
2f2dc053
SW		 4237 dout("handle_lease no inode %llx\n", vino.ino);
4238 goto release;
4239 }
2f2dc053
SW		 4240
4241 /* dentry */
4242 parent = d_find_alias(inode);
4243 if (!parent) {
4244 dout("no parent dentry on inode %p\n", inode);
4245 WARN_ON(1);
4246 goto release; /* hrm... */
4247 }
8387ff25 4248 dname.hash = full_name_hash(parent, dname.name, dname.len);
2f2dc053
SW		 4249 dentry = d_lookup(parent, &dname);
4250 dput(parent);
4251 if (!dentry)
4252 goto release;
4253
4254 spin_lock(&dentry->d_lock);
4255 di = ceph_dentry(dentry);
4256 switch (h->action) {
4257 case CEPH_MDS_LEASE_REVOKE:
3d8eb7a9 4258 if (di->lease_session == session) {
1e5ea23d
SW		 4259 if (ceph_seq_cmp(di->lease_seq, seq) > 0)
4260 h->seq = cpu_to_le32(di->lease_seq);
2f2dc053
SW		 4261 __ceph_mdsc_drop_dentry_lease(dentry);
4262 }
4263 release = 1;
4264 break;
4265
4266 case CEPH_MDS_LEASE_RENEW:
3d8eb7a9 4267 if (di->lease_session == session &&
2f2dc053
SW		 4268 di->lease_gen == session->s_cap_gen &&
4269 di->lease_renew_from &&
4270 di->lease_renew_after == 0) {
4271 unsigned long duration =
3563dbdd 4272 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
2f2dc053 4273
1e5ea23d 4274 di->lease_seq = seq;
9b16f03c 4275 di->time = di->lease_renew_from + duration;
2f2dc053
SW		 4276 di->lease_renew_after = di->lease_renew_from +
4277 (duration >> 1);
4278 di->lease_renew_from = 0;
4279 }
4280 break;
4281 }
4282 spin_unlock(&dentry->d_lock);
4283 dput(dentry);
4284
4285 if (!release)
4286 goto out;
4287
4288release:
4289 /* let's just reuse the same message */
4290 h->action = CEPH_MDS_LEASE_REVOKE_ACK;
4291 ceph_msg_get(msg);
4292 ceph_con_send(&session->s_con, msg);
4293
4294out:
2f2dc053 4295 mutex_unlock(&session->s_mutex);
3e1d0452
YZ		 4296 /* avoid calling iput_final() in mds dispatch threads */
4297 ceph_async_iput(inode);
2f2dc053
SW		 4298 return;
4299
4300bad:
4301 pr_err("corrupt lease message\n");
9ec7cab1 4302 ceph_msg_dump(msg);
2f2dc053
SW		 4303}
4304
4305void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
2f2dc053
SW		 4306 struct dentry *dentry, char action,
4307 u32 seq)
4308{
4309 struct ceph_msg *msg;
4310 struct ceph_mds_lease *lease;
8f2a98ef
YZ		 4311 struct inode *dir;
4312 int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
2f2dc053 4313
8f2a98ef
YZ		 4314 dout("lease_send_msg identry %p %s to mds%d\n",
4315 dentry, ceph_lease_op_name(action), session->s_mds);
2f2dc053 4316
b61c2763 4317 msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
a79832f2 4318 if (!msg)
2f2dc053
SW		 4319 return;
4320 lease = msg->front.iov_base;
4321 lease->action = action;
2f2dc053 4322 lease->seq = cpu_to_le32(seq);
2f2dc053 4323
8f2a98ef
YZ		 4324 spin_lock(&dentry->d_lock);
4325 dir = d_inode(dentry->d_parent);
4326 lease->ino = cpu_to_le64(ceph_ino(dir));
4327 lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
4328
4329 put_unaligned_le32(dentry->d_name.len, lease + 1);
4330 memcpy((void *)(lease + 1) + 4,
4331 dentry->d_name.name, dentry->d_name.len);
4332 spin_unlock(&dentry->d_lock);
2f2dc053
SW		 4333 /*
4334 * if this is a preemptive lease RELEASE, no need to
4335 * flush request stream, since the actual request will
4336 * soon follow.
4337 */
4338 msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
4339
4340 ceph_con_send(&session->s_con, msg);
4341}
4342
2f2dc053 4343/*
7aac453a 4344 * lock unlock sessions, to wait ongoing session activities
2f2dc053 4345 */
7aac453a 4346static void lock_unlock_sessions(struct ceph_mds_client *mdsc)
2f2dc053
SW		 4347{
4348 int i;
4349
2f2dc053
SW		 4350 mutex_lock(&mdsc->mutex);
4351 for (i = 0; i < mdsc->max_sessions; i++) {
4352 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4353 if (!s)
4354 continue;
4355 mutex_unlock(&mdsc->mutex);
4356 mutex_lock(&s->s_mutex);
4357 mutex_unlock(&s->s_mutex);
4358 ceph_put_mds_session(s);
4359 mutex_lock(&mdsc->mutex);
4360 }
4361 mutex_unlock(&mdsc->mutex);
4362}
4363
131d7eb4
YZ		 4364static void maybe_recover_session(struct ceph_mds_client *mdsc)
4365{
4366 struct ceph_fs_client *fsc = mdsc->fsc;
4367
4368 if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
4369 return;
2f2dc053 4370
131d7eb4
YZ		 4371 if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
4372 return;
4373
0b98acd6 4374 if (!READ_ONCE(fsc->blocklisted))
131d7eb4
YZ		 4375 return;
4376
4377 if (fsc->last_auto_reconnect &&
4378 time_before(jiffies, fsc->last_auto_reconnect + HZ * 60 * 30))
4379 return;
4380
0b98acd6 4381 pr_info("auto reconnect after blocklisted\n");
131d7eb4
YZ		 4382 fsc->last_auto_reconnect = jiffies;
4383 ceph_force_reconnect(fsc->sb);
4384}
2f2dc053 4385
3e699bd8
XL		 4386bool check_session_state(struct ceph_mds_session *s)
4387{
62575e27
JL		 4388 switch (s->s_state) {
4389 case CEPH_MDS_SESSION_OPEN:
4390 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
3e699bd8
XL		 4391 s->s_state = CEPH_MDS_SESSION_HUNG;
4392 pr_info("mds%d hung\n", s->s_mds);
4393 }
62575e27
JL		 4394 break;
4395 case CEPH_MDS_SESSION_CLOSING:
4396 /* Should never reach this when we're unmounting */
4397 WARN_ON_ONCE(true);
4398 fallthrough;
4399 case CEPH_MDS_SESSION_NEW:
4400 case CEPH_MDS_SESSION_RESTARTING:
4401 case CEPH_MDS_SESSION_CLOSED:
4402 case CEPH_MDS_SESSION_REJECTED:
3e699bd8 4403 return false;
62575e27 4404 }
3e699bd8
XL		 4405
4406 return true;
4407}
4408
62575e27
JL		 4409/*
4410 * If the sequence is incremented while we're waiting on a REQUEST_CLOSE reply,
4411 * then we need to retransmit that request.
4412 */
4413void inc_session_sequence(struct ceph_mds_session *s)
4414{
4415 lockdep_assert_held(&s->s_mutex);
4416
4417 s->s_seq++;
4418
4419 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4420 int ret;
4421
4422 dout("resending session close request for mds%d\n", s->s_mds);
4423 ret = request_close_session(s);
4424 if (ret < 0)
4425 pr_err("unable to close session to mds%d: %d\n",
4426 s->s_mds, ret);
4427 }
4428}
4429
2f2dc053
SW		 4430/*
4431 * delayed work -- periodically trim expired leases, renew caps with mds
4432 */
4433static void schedule_delayed(struct ceph_mds_client *mdsc)
4434{
4435 int delay = 5;
4436 unsigned hz = round_jiffies_relative(HZ * delay);
4437 schedule_delayed_work(&mdsc->delayed_work, hz);
4438}
4439
4440static void delayed_work(struct work_struct *work)
4441{
4442 int i;
4443 struct ceph_mds_client *mdsc =
4444 container_of(work, struct ceph_mds_client, delayed_work.work);
4445 int renew_interval;
4446 int renew_caps;
4447
4448 dout("mdsc delayed_work\n");
75c9627e 4449
fa996773
XL		 4450 if (mdsc->stopping)
4451 return;
4452
2f2dc053
SW		 4453 mutex_lock(&mdsc->mutex);
4454 renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4455 renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4456 mdsc->last_renew_caps);
4457 if (renew_caps)
4458 mdsc->last_renew_caps = jiffies;
4459
4460 for (i = 0; i < mdsc->max_sessions; i++) {
4461 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
d37b1d99 4462 if (!s)
2f2dc053 4463 continue;
3e699bd8
XL		 4464
4465 if (!check_session_state(s)) {
2f2dc053
SW		 4466 ceph_put_mds_session(s);
4467 continue;
4468 }
4469 mutex_unlock(&mdsc->mutex);
4470
4471 mutex_lock(&s->s_mutex);
4472 if (renew_caps)
4473 send_renew_caps(mdsc, s);
4474 else
4475 ceph_con_keepalive(&s->s_con);
aab53dd9
SW		 4476 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4477 s->s_state == CEPH_MDS_SESSION_HUNG)
3d7ded4d 4478 ceph_send_cap_releases(mdsc, s);
2f2dc053
SW		 4479 mutex_unlock(&s->s_mutex);
4480 ceph_put_mds_session(s);
4481
4482 mutex_lock(&mdsc->mutex);
4483 }
4484 mutex_unlock(&mdsc->mutex);
4485
37c4efc1
YZ		 4486 ceph_check_delayed_caps(mdsc);
4487
4488 ceph_queue_cap_reclaim_work(mdsc);
4489
4490 ceph_trim_snapid_map(mdsc);
4491
131d7eb4
YZ		 4492 maybe_recover_session(mdsc);
4493
2f2dc053
SW		 4494 schedule_delayed(mdsc);
4495}
4496
3d14c5d2 4497int ceph_mdsc_init(struct ceph_fs_client *fsc)
2f2dc053 4498
2f2dc053 4499{
3d14c5d2 4500 struct ceph_mds_client *mdsc;
f9009efa 4501 int err;
3d14c5d2
YS		 4502
4503 mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4504 if (!mdsc)
4505 return -ENOMEM;
4506 mdsc->fsc = fsc;
2f2dc053
SW		 4507 mutex_init(&mdsc->mutex);
4508 mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
d37b1d99 4509 if (!mdsc->mdsmap) {
f9009efa
XL		 4510 err = -ENOMEM;
4511 goto err_mdsc;
fb3101b6 4512 }
2d06eeb8 4513
2f2dc053 4514 init_completion(&mdsc->safe_umount_waiters);
f3c60c59 4515 init_waitqueue_head(&mdsc->session_close_wq);
2f2dc053
SW		 4516 INIT_LIST_HEAD(&mdsc->waiting_for_map);
4517 mdsc->sessions = NULL;
86d8f67b 4518 atomic_set(&mdsc->num_sessions, 0);
2f2dc053
SW		 4519 mdsc->max_sessions = 0;
4520 mdsc->stopping = 0;
d557c48d 4521 atomic64_set(&mdsc->quotarealms_count, 0);
0c44a8e0
LH		 4522 mdsc->quotarealms_inodes = RB_ROOT;
4523 mutex_init(&mdsc->quotarealms_inodes_mutex);
affbc19a 4524 mdsc->last_snap_seq = 0;
2f2dc053 4525 init_rwsem(&mdsc->snap_rwsem);
a105f00c 4526 mdsc->snap_realms = RB_ROOT;
2f2dc053 4527 INIT_LIST_HEAD(&mdsc->snap_empty);
81c5a148 4528 mdsc->num_snap_realms = 0;
2f2dc053
SW		 4529 spin_lock_init(&mdsc->snap_empty_lock);
4530 mdsc->last_tid = 0;
e8a7b8b1 4531 mdsc->oldest_tid = 0;
44ca18f2 4532 mdsc->request_tree = RB_ROOT;
2f2dc053
SW		 4533 INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4534 mdsc->last_renew_caps = jiffies;
4535 INIT_LIST_HEAD(&mdsc->cap_delay_list);
3a3430af 4536 INIT_LIST_HEAD(&mdsc->cap_wait_list);
2f2dc053
SW		 4537 spin_lock_init(&mdsc->cap_delay_lock);
4538 INIT_LIST_HEAD(&mdsc->snap_flush_list);
4539 spin_lock_init(&mdsc->snap_flush_lock);
553adfd9 4540 mdsc->last_cap_flush_tid = 1;
e4500b5e 4541 INIT_LIST_HEAD(&mdsc->cap_flush_list);
db354052 4542 INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
2f2dc053
SW		 4543 mdsc->num_cap_flushing = 0;
4544 spin_lock_init(&mdsc->cap_dirty_lock);
4545 init_waitqueue_head(&mdsc->cap_flushing_wq);
37c4efc1 4546 INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
fe33032d 4547 atomic_set(&mdsc->cap_reclaim_pending, 0);
f9009efa
XL		 4548 err = ceph_metric_init(&mdsc->metric);
4549 if (err)
4550 goto err_mdsmap;
37c4efc1
YZ		 4551
4552 spin_lock_init(&mdsc->dentry_list_lock);
4553 INIT_LIST_HEAD(&mdsc->dentry_leases);
4554 INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
2d06eeb8 4555
37151668 4556 ceph_caps_init(mdsc);
fe33032d 4557 ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
37151668 4558
75c9627e
YZ		 4559 spin_lock_init(&mdsc->snapid_map_lock);
4560 mdsc->snapid_map_tree = RB_ROOT;
4561 INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4562
10183a69
YZ		 4563 init_rwsem(&mdsc->pool_perm_rwsem);
4564 mdsc->pool_perm_tree = RB_ROOT;
4565
dfeb84d4
YZ		 4566 strscpy(mdsc->nodename, utsname()->nodename,
4567 sizeof(mdsc->nodename));
a7caa88f
XL		 4568
4569 fsc->mdsc = mdsc;
5f44f142 4570 return 0;
f9009efa
XL		 4571
4572err_mdsmap:
4573 kfree(mdsc->mdsmap);
4574err_mdsc:
4575 kfree(mdsc);
4576 return err;
2f2dc053
SW		 4577}
4578
4579/*
4580 * Wait for safe replies on open mds requests. If we time out, drop
4581 * all requests from the tree to avoid dangling dentry refs.
4582 */
4583static void wait_requests(struct ceph_mds_client *mdsc)
4584{
a319bf56 4585 struct ceph_options *opts = mdsc->fsc->client->options;
2f2dc053 4586 struct ceph_mds_request *req;
2f2dc053
SW		 4587
4588 mutex_lock(&mdsc->mutex);
44ca18f2 4589 if (__get_oldest_req(mdsc)) {
2f2dc053 4590 mutex_unlock(&mdsc->mutex);
44ca18f2 4591
2f2dc053
SW		 4592 dout("wait_requests waiting for requests\n");
4593 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
a319bf56 4594 ceph_timeout_jiffies(opts->mount_timeout));
2f2dc053
SW		 4595
4596 /* tear down remaining requests */
44ca18f2
SW		 4597 mutex_lock(&mdsc->mutex);
4598 while ((req = __get_oldest_req(mdsc))) {
2f2dc053
SW		 4599 dout("wait_requests timed out on tid %llu\n",
4600 req->r_tid);
428138c9 4601 list_del_init(&req->r_wait);
44ca18f2 4602 __unregister_request(mdsc, req);
2f2dc053
SW		 4603 }
4604 }
4605 mutex_unlock(&mdsc->mutex);
4606 dout("wait_requests done\n");
4607}
4608
4609/*
4610 * called before mount is ro, and before dentries are torn down.
4611 * (hmm, does this still race with new lookups?)
4612 */
4613void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4614{
4615 dout("pre_umount\n");
4616 mdsc->stopping = 1;
4617
7aac453a 4618 lock_unlock_sessions(mdsc);
afcdaea3 4619 ceph_flush_dirty_caps(mdsc);
2f2dc053 4620 wait_requests(mdsc);
17c688c3
SW		 4621
4622 /*
4623 * wait for reply handlers to drop their request refs and
4624 * their inode/dcache refs
4625 */
4626 ceph_msgr_flush();
0c44a8e0
LH		 4627
4628 ceph_cleanup_quotarealms_inodes(mdsc);
2f2dc053
SW		 4629}
4630
4631/*
4632 * wait for all write mds requests to flush.
4633 */
4634static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
4635{
80fc7314 4636 struct ceph_mds_request *req = NULL, *nextreq;
44ca18f2 4637 struct rb_node *n;
2f2dc053
SW		 4638
4639 mutex_lock(&mdsc->mutex);
4640 dout("wait_unsafe_requests want %lld\n", want_tid);
80fc7314 4641restart:
44ca18f2
SW		 4642 req = __get_oldest_req(mdsc);
4643 while (req && req->r_tid <= want_tid) {
80fc7314
SW		 4644 /* find next request */
4645 n = rb_next(&req->r_node);
4646 if (n)
4647 nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4648 else
4649 nextreq = NULL;
e8a7b8b1
YZ		 4650 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4651 (req->r_op & CEPH_MDS_OP_WRITE)) {
44ca18f2
SW		 4652 /* write op */
4653 ceph_mdsc_get_request(req);
80fc7314
SW		 4654 if (nextreq)
4655 ceph_mdsc_get_request(nextreq);
44ca18f2
SW		 4656 mutex_unlock(&mdsc->mutex);
4657 dout("wait_unsafe_requests wait on %llu (want %llu)\n",
4658 req->r_tid, want_tid);
4659 wait_for_completion(&req->r_safe_completion);
4660 mutex_lock(&mdsc->mutex);
44ca18f2 4661 ceph_mdsc_put_request(req);
80fc7314
SW		 4662 if (!nextreq)
4663 break; /* next dne before, so we're done! */
4664 if (RB_EMPTY_NODE(&nextreq->r_node)) {
4665 /* next request was removed from tree */
4666 ceph_mdsc_put_request(nextreq);
4667 goto restart;
4668 }
4669 ceph_mdsc_put_request(nextreq); /* won't go away */
44ca18f2 4670 }
80fc7314 4671 req = nextreq;
2f2dc053
SW		 4672 }
4673 mutex_unlock(&mdsc->mutex);
4674 dout("wait_unsafe_requests done\n");
4675}
4676
4677void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4678{
0e294387 4679 u64 want_tid, want_flush;
2f2dc053 4680
52953d55 4681 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
56b7cf95
SW		 4682 return;
4683
2f2dc053
SW		 4684 dout("sync\n");
4685 mutex_lock(&mdsc->mutex);
4686 want_tid = mdsc->last_tid;
2f2dc053 4687 mutex_unlock(&mdsc->mutex);
2f2dc053 4688
afcdaea3 4689 ceph_flush_dirty_caps(mdsc);
d3383a8e 4690 spin_lock(&mdsc->cap_dirty_lock);
8310b089 4691 want_flush = mdsc->last_cap_flush_tid;
c8799fc4
YZ		 4692 if (!list_empty(&mdsc->cap_flush_list)) {
4693 struct ceph_cap_flush *cf =
4694 list_last_entry(&mdsc->cap_flush_list,
4695 struct ceph_cap_flush, g_list);
4696 cf->wake = true;
4697 }
d3383a8e
YZ		 4698 spin_unlock(&mdsc->cap_dirty_lock);
4699
0e294387
YZ		 4700 dout("sync want tid %lld flush_seq %lld\n",
4701 want_tid, want_flush);
2f2dc053
SW		 4702
4703 wait_unsafe_requests(mdsc, want_tid);
0e294387 4704 wait_caps_flush(mdsc, want_flush);
2f2dc053
SW		 4705}
4706
f3c60c59
SW		 4707/*
4708 * true if all sessions are closed, or we force unmount
4709 */
fcff415c 4710static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
f3c60c59 4711{
52953d55 4712 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
f3c60c59 4713 return true;
fcff415c 4714 return atomic_read(&mdsc->num_sessions) <= skipped;
f3c60c59 4715}
2f2dc053
SW		 4716
4717/*
4718 * called after sb is ro.
4719 */
4720void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
4721{
a319bf56 4722 struct ceph_options *opts = mdsc->fsc->client->options;
2f2dc053
SW		 4723 struct ceph_mds_session *session;
4724 int i;
fcff415c 4725 int skipped = 0;
2f2dc053
SW		 4726
4727 dout("close_sessions\n");
4728
2f2dc053 4729 /* close sessions */
f3c60c59
SW		 4730 mutex_lock(&mdsc->mutex);
4731 for (i = 0; i < mdsc->max_sessions; i++) {
4732 session = __ceph_lookup_mds_session(mdsc, i);
4733 if (!session)
4734 continue;
2f2dc053 4735 mutex_unlock(&mdsc->mutex);
f3c60c59 4736 mutex_lock(&session->s_mutex);
fcff415c
YZ		 4737 if (__close_session(mdsc, session) <= 0)
4738 skipped++;
f3c60c59
SW		 4739 mutex_unlock(&session->s_mutex);
4740 ceph_put_mds_session(session);
2f2dc053
SW		 4741 mutex_lock(&mdsc->mutex);
4742 }
f3c60c59
SW		 4743 mutex_unlock(&mdsc->mutex);
4744
4745 dout("waiting for sessions to close\n");
fcff415c
YZ		 4746 wait_event_timeout(mdsc->session_close_wq,
4747 done_closing_sessions(mdsc, skipped),
a319bf56 4748 ceph_timeout_jiffies(opts->mount_timeout));
2f2dc053
SW		 4749
4750 /* tear down remaining sessions */
f3c60c59 4751 mutex_lock(&mdsc->mutex);
2f2dc053
SW		 4752 for (i = 0; i < mdsc->max_sessions; i++) {
4753 if (mdsc->sessions[i]) {
5b3248c6 4754 session = ceph_get_mds_session(mdsc->sessions[i]);
2600d2dd 4755 __unregister_session(mdsc, session);
2f2dc053
SW		 4756 mutex_unlock(&mdsc->mutex);
4757 mutex_lock(&session->s_mutex);
4758 remove_session_caps(session);
4759 mutex_unlock(&session->s_mutex);
4760 ceph_put_mds_session(session);
4761 mutex_lock(&mdsc->mutex);
4762 }
4763 }
2f2dc053 4764 WARN_ON(!list_empty(&mdsc->cap_delay_list));
2f2dc053
SW		 4765 mutex_unlock(&mdsc->mutex);
4766
75c9627e 4767 ceph_cleanup_snapid_map(mdsc);
2f2dc053
SW		 4768 ceph_cleanup_empty_realms(mdsc);
4769
37c4efc1 4770 cancel_work_sync(&mdsc->cap_reclaim_work);
2f2dc053
SW		 4771 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4772
4773 dout("stopped\n");
4774}
4775
48fec5d0
YZ		 4776void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
4777{
4778 struct ceph_mds_session *session;
4779 int mds;
4780
4781 dout("force umount\n");
4782
4783 mutex_lock(&mdsc->mutex);
4784 for (mds = 0; mds < mdsc->max_sessions; mds++) {
4785 session = __ceph_lookup_mds_session(mdsc, mds);
4786 if (!session)
4787 continue;
d468e729
YZ		 4788
4789 if (session->s_state == CEPH_MDS_SESSION_REJECTED)
4790 __unregister_session(mdsc, session);
4791 __wake_requests(mdsc, &session->s_waiting);
48fec5d0 4792 mutex_unlock(&mdsc->mutex);
d468e729 4793
48fec5d0
YZ		 4794 mutex_lock(&session->s_mutex);
4795 __close_session(mdsc, session);
4796 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
4797 cleanup_session_requests(mdsc, session);
4798 remove_session_caps(session);
4799 }
4800 mutex_unlock(&session->s_mutex);
4801 ceph_put_mds_session(session);
d468e729 4802
48fec5d0
YZ		 4803 mutex_lock(&mdsc->mutex);
4804 kick_requests(mdsc, mds);
4805 }
4806 __wake_requests(mdsc, &mdsc->waiting_for_map);
4807 mutex_unlock(&mdsc->mutex);
4808}
4809
3d14c5d2 4810static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
2f2dc053
SW		 4811{
4812 dout("stop\n");
fa996773
XL		 4813 /*
4814 * Make sure the delayed work stopped before releasing
4815 * the resources.
4816 *
4817 * Because the cancel_delayed_work_sync() will only
4818 * guarantee that the work finishes executing. But the
4819 * delayed work will re-arm itself again after that.
4820 */
4821 flush_delayed_work(&mdsc->delayed_work);
4822
2f2dc053
SW		 4823 if (mdsc->mdsmap)
4824 ceph_mdsmap_destroy(mdsc->mdsmap);
4825 kfree(mdsc->sessions);
37151668 4826 ceph_caps_finalize(mdsc);
10183a69 4827 ceph_pool_perm_destroy(mdsc);
2f2dc053
SW		 4828}
4829
3d14c5d2
YS		 4830void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
4831{
4832 struct ceph_mds_client *mdsc = fsc->mdsc;
ef550f6f 4833 dout("mdsc_destroy %p\n", mdsc);
ef550f6f 4834
50c55aec
CX		 4835 if (!mdsc)
4836 return;
4837
ef550f6f
SW		 4838 /* flush out any connection work with references to us */
4839 ceph_msgr_flush();
4840
62a65f36
YZ		 4841 ceph_mdsc_stop(mdsc);
4842
f9009efa
XL		 4843 ceph_metric_destroy(&mdsc->metric);
4844
18f473b3 4845 flush_delayed_work(&mdsc->metric.delayed_work);
3d14c5d2
YS		 4846 fsc->mdsc = NULL;
4847 kfree(mdsc);
ef550f6f 4848 dout("mdsc_destroy %p done\n", mdsc);
3d14c5d2
YS		 4849}
4850
430afbad
YZ		 4851void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4852{
4853 struct ceph_fs_client *fsc = mdsc->fsc;
4854 const char *mds_namespace = fsc->mount_options->mds_namespace;
4855 void *p = msg->front.iov_base;
4856 void *end = p + msg->front.iov_len;
4857 u32 epoch;
4858 u32 map_len;
4859 u32 num_fs;
4860 u32 mount_fscid = (u32)-1;
4861 u8 struct_v, struct_cv;
4862 int err = -EINVAL;
4863
4864 ceph_decode_need(&p, end, sizeof(u32), bad);
4865 epoch = ceph_decode_32(&p);
4866
4867 dout("handle_fsmap epoch %u\n", epoch);
4868
4869 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4870 struct_v = ceph_decode_8(&p);
4871 struct_cv = ceph_decode_8(&p);
4872 map_len = ceph_decode_32(&p);
4873
4874 ceph_decode_need(&p, end, sizeof(u32) * 3, bad);
4875 p += sizeof(u32) * 2; /* skip epoch and legacy_client_fscid */
4876
4877 num_fs = ceph_decode_32(&p);
4878 while (num_fs-- > 0) {
4879 void *info_p, *info_end;
4880 u32 info_len;
4881 u8 info_v, info_cv;
4882 u32 fscid, namelen;
4883
4884 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4885 info_v = ceph_decode_8(&p);
4886 info_cv = ceph_decode_8(&p);
4887 info_len = ceph_decode_32(&p);
4888 ceph_decode_need(&p, end, info_len, bad);
4889 info_p = p;
4890 info_end = p + info_len;
4891 p = info_end;
4892
4893 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
4894 fscid = ceph_decode_32(&info_p);
4895 namelen = ceph_decode_32(&info_p);
4896 ceph_decode_need(&info_p, info_end, namelen, bad);
4897
4898 if (mds_namespace &&
4899 strlen(mds_namespace) == namelen &&
4900 !strncmp(mds_namespace, (char *)info_p, namelen)) {
4901 mount_fscid = fscid;
4902 break;
4903 }
4904 }
4905
4906 ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
4907 if (mount_fscid != (u32)-1) {
4908 fsc->client->monc.fs_cluster_id = mount_fscid;
4909 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
4910 0, true);
4911 ceph_monc_renew_subs(&fsc->client->monc);
4912 } else {
4913 err = -ENOENT;
4914 goto err_out;
4915 }
4916 return;
76bd6ec4 4917
430afbad
YZ		 4918bad:
4919 pr_err("error decoding fsmap\n");
4920err_out:
4921 mutex_lock(&mdsc->mutex);
76bd6ec4 4922 mdsc->mdsmap_err = err;
430afbad
YZ		 4923 __wake_requests(mdsc, &mdsc->waiting_for_map);
4924 mutex_unlock(&mdsc->mutex);
430afbad 4925}
2f2dc053
SW		 4926
4927/*
4928 * handle mds map update.
4929 */
430afbad 4930void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
2f2dc053
SW		 4931{
4932 u32 epoch;
4933 u32 maplen;
4934 void *p = msg->front.iov_base;
4935 void *end = p + msg->front.iov_len;
4936 struct ceph_mdsmap *newmap, *oldmap;
4937 struct ceph_fsid fsid;
4938 int err = -EINVAL;
4939
4940 ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
4941 ceph_decode_copy(&p, &fsid, sizeof(fsid));
3d14c5d2 4942 if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
0743304d 4943 return;
c89136ea
SW		 4944 epoch = ceph_decode_32(&p);
4945 maplen = ceph_decode_32(&p);
2f2dc053
SW		 4946 dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
4947
4948 /* do we need it? */
2f2dc053
SW		 4949 mutex_lock(&mdsc->mutex);
4950 if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
4951 dout("handle_map epoch %u <= our %u\n",
4952 epoch, mdsc->mdsmap->m_epoch);
4953 mutex_unlock(&mdsc->mutex);
4954 return;
4955 }
4956
4957 newmap = ceph_mdsmap_decode(&p, end);
4958 if (IS_ERR(newmap)) {
4959 err = PTR_ERR(newmap);
4960 goto bad_unlock;
4961 }
4962
4963 /* swap into place */
4964 if (mdsc->mdsmap) {
4965 oldmap = mdsc->mdsmap;
4966 mdsc->mdsmap = newmap;
4967 check_new_map(mdsc, newmap, oldmap);
4968 ceph_mdsmap_destroy(oldmap);
4969 } else {
4970 mdsc->mdsmap = newmap; /* first mds map */
4971 }
719784ba
CX		 4972 mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
4973 MAX_LFS_FILESIZE);
2f2dc053
SW		 4974
4975 __wake_requests(mdsc, &mdsc->waiting_for_map);
82dcabad
ID		 4976 ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
4977 mdsc->mdsmap->m_epoch);
2f2dc053
SW		 4978
4979 mutex_unlock(&mdsc->mutex);
4980 schedule_delayed(mdsc);
4981 return;
4982
4983bad_unlock:
4984 mutex_unlock(&mdsc->mutex);
4985bad:
4986 pr_err("error decoding mdsmap %d\n", err);
4987 return;
4988}
4989
4990static struct ceph_connection *con_get(struct ceph_connection *con)
4991{
4992 struct ceph_mds_session *s = con->private;
4993
5b3248c6 4994 if (ceph_get_mds_session(s))
2f2dc053 4995 return con;
2f2dc053
SW		 4996 return NULL;
4997}
4998
4999static void con_put(struct ceph_connection *con)
5000{
5001 struct ceph_mds_session *s = con->private;
5002
2f2dc053
SW		 5003 ceph_put_mds_session(s);
5004}
5005
5006/*
5007 * if the client is unresponsive for long enough, the mds will kill
5008 * the session entirely.
5009 */
5010static void peer_reset(struct ceph_connection *con)
5011{
5012 struct ceph_mds_session *s = con->private;
7e70f0ed 5013 struct ceph_mds_client *mdsc = s->s_mdsc;
2f2dc053 5014
f3ae1b97 5015 pr_warn("mds%d closed our session\n", s->s_mds);
7e70f0ed 5016 send_mds_reconnect(mdsc, s);
2f2dc053
SW		 5017}
5018
5019static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5020{
5021 struct ceph_mds_session *s = con->private;
5022 struct ceph_mds_client *mdsc = s->s_mdsc;
5023 int type = le16_to_cpu(msg->hdr.type);
5024
2600d2dd
SW		 5025 mutex_lock(&mdsc->mutex);
5026 if (__verify_registered_session(mdsc, s) < 0) {
5027 mutex_unlock(&mdsc->mutex);
5028 goto out;
5029 }
5030 mutex_unlock(&mdsc->mutex);
5031
2f2dc053
SW		 5032 switch (type) {
5033 case CEPH_MSG_MDS_MAP:
430afbad
YZ		 5034 ceph_mdsc_handle_mdsmap(mdsc, msg);
5035 break;
5036 case CEPH_MSG_FS_MAP_USER:
5037 ceph_mdsc_handle_fsmap(mdsc, msg);
2f2dc053
SW		 5038 break;
5039 case CEPH_MSG_CLIENT_SESSION:
5040 handle_session(s, msg);
5041 break;
5042 case CEPH_MSG_CLIENT_REPLY:
5043 handle_reply(s, msg);
5044 break;
5045 case CEPH_MSG_CLIENT_REQUEST_FORWARD:
2600d2dd 5046 handle_forward(mdsc, s, msg);
2f2dc053
SW		 5047 break;
5048 case CEPH_MSG_CLIENT_CAPS:
5049 ceph_handle_caps(s, msg);
5050 break;
5051 case CEPH_MSG_CLIENT_SNAP:
2600d2dd 5052 ceph_handle_snap(mdsc, s, msg);
2f2dc053
SW		 5053 break;
5054 case CEPH_MSG_CLIENT_LEASE:
2600d2dd 5055 handle_lease(mdsc, s, msg);
2f2dc053 5056 break;
fb18a575
LH		 5057 case CEPH_MSG_CLIENT_QUOTA:
5058 ceph_handle_quota(mdsc, s, msg);
5059 break;
2f2dc053
SW		 5060
5061 default:
5062 pr_err("received unknown message type %d %s\n", type,
5063 ceph_msg_type_name(type));
5064 }
2600d2dd 5065out:
2f2dc053
SW		 5066 ceph_msg_put(msg);
5067}
5068
4e7a5dcd
SW		 5069/*
5070 * authentication
5071 */
a3530df3
AE		 5072
5073/*
5074 * Note: returned pointer is the address of a structure that's
5075 * managed separately. Caller must *not* attempt to free it.
5076 */
5077static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
8f43fb53 5078 int *proto, int force_new)
4e7a5dcd
SW		 5079{
5080 struct ceph_mds_session *s = con->private;
5081 struct ceph_mds_client *mdsc = s->s_mdsc;
3d14c5d2 5082 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
74f1869f 5083 struct ceph_auth_handshake *auth = &s->s_auth;
4e7a5dcd 5084
74f1869f 5085 if (force_new && auth->authorizer) {
6c1ea260 5086 ceph_auth_destroy_authorizer(auth->authorizer);
74f1869f 5087 auth->authorizer = NULL;
4e7a5dcd 5088 }
27859f97
SW		 5089 if (!auth->authorizer) {
5090 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
5091 auth);
0bed9b5c
SW		 5092 if (ret)
5093 return ERR_PTR(ret);
27859f97
SW		 5094 } else {
5095 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
5096 auth);
a255651d 5097 if (ret)
a3530df3 5098 return ERR_PTR(ret);
4e7a5dcd 5099 }
4e7a5dcd 5100 *proto = ac->protocol;
74f1869f 5101
a3530df3 5102 return auth;
4e7a5dcd
SW		 5103}
5104
6daca13d
ID		 5105static int add_authorizer_challenge(struct ceph_connection *con,
5106 void *challenge_buf, int challenge_buf_len)
5107{
5108 struct ceph_mds_session *s = con->private;
5109 struct ceph_mds_client *mdsc = s->s_mdsc;
5110 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5111
5112 return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
5113 challenge_buf, challenge_buf_len);
5114}
4e7a5dcd 5115
0dde5848 5116static int verify_authorizer_reply(struct ceph_connection *con)
4e7a5dcd
SW		 5117{
5118 struct ceph_mds_session *s = con->private;
5119 struct ceph_mds_client *mdsc = s->s_mdsc;
3d14c5d2 5120 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4e7a5dcd 5121
0dde5848 5122 return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer);
4e7a5dcd
SW		 5123}
5124
9bd2e6f8
SW		 5125static int invalidate_authorizer(struct ceph_connection *con)
5126{
5127 struct ceph_mds_session *s = con->private;
5128 struct ceph_mds_client *mdsc = s->s_mdsc;
3d14c5d2 5129 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
9bd2e6f8 5130
27859f97 5131 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
9bd2e6f8 5132
3d14c5d2 5133 return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
9bd2e6f8
SW		 5134}
5135
53ded495
AE		 5136static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
5137 struct ceph_msg_header *hdr, int *skip)
5138{
5139 struct ceph_msg *msg;
5140 int type = (int) le16_to_cpu(hdr->type);
5141 int front_len = (int) le32_to_cpu(hdr->front_len);
5142
5143 if (con->in_msg)
5144 return con->in_msg;
5145
5146 *skip = 0;
5147 msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
5148 if (!msg) {
5149 pr_err("unable to allocate msg type %d len %d\n",
5150 type, front_len);
5151 return NULL;
5152 }
53ded495
AE		 5153
5154 return msg;
5155}
5156
79dbd1ba 5157static int mds_sign_message(struct ceph_msg *msg)
33d07337 5158{
79dbd1ba 5159 struct ceph_mds_session *s = msg->con->private;
33d07337 5160 struct ceph_auth_handshake *auth = &s->s_auth;
79dbd1ba 5161
33d07337
YZ		 5162 return ceph_auth_sign_message(auth, msg);
5163}
5164
79dbd1ba 5165static int mds_check_message_signature(struct ceph_msg *msg)
33d07337 5166{
79dbd1ba 5167 struct ceph_mds_session *s = msg->con->private;
33d07337 5168 struct ceph_auth_handshake *auth = &s->s_auth;
79dbd1ba 5169
33d07337
YZ		 5170 return ceph_auth_check_message_signature(auth, msg);
5171}
5172
9e32789f 5173static const struct ceph_connection_operations mds_con_ops = {
2f2dc053
SW		 5174 .get = con_get,
5175 .put = con_put,
5176 .dispatch = dispatch,
4e7a5dcd 5177 .get_authorizer = get_authorizer,
6daca13d 5178 .add_authorizer_challenge = add_authorizer_challenge,
4e7a5dcd 5179 .verify_authorizer_reply = verify_authorizer_reply,
9bd2e6f8 5180 .invalidate_authorizer = invalidate_authorizer,
2f2dc053 5181 .peer_reset = peer_reset,
53ded495 5182 .alloc_msg = mds_alloc_msg,
79dbd1ba
ID		 5183 .sign_message = mds_sign_message,
5184 .check_message_signature = mds_check_message_signature,
2f2dc053
SW		 5185};
5186
2f2dc053 5187/* eof */
Empty description
This page took 1.413896 seconds and 4 git commands to generate.