xref: /linux/fs/ceph/mds_client.c (revision bb46122db730f42f3fc1d9d511b3d6ebe8375cdd)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3 
4 #include <linux/fs.h>
5 #include <linux/wait.h>
6 #include <linux/slab.h>
7 #include <linux/gfp.h>
8 #include <linux/sched.h>
9 #include <linux/debugfs.h>
10 #include <linux/seq_file.h>
11 #include <linux/ratelimit.h>
12 #include <linux/bits.h>
13 #include <linux/ktime.h>
14 #include <linux/bitmap.h>
15 #include <linux/mnt_idmapping.h>
16 
17 #include "super.h"
18 #include "mds_client.h"
19 #include "crypto.h"
20 
21 #include <linux/ceph/ceph_features.h>
22 #include <linux/ceph/messenger.h>
23 #include <linux/ceph/decode.h>
24 #include <linux/ceph/pagelist.h>
25 #include <linux/ceph/auth.h>
26 #include <linux/ceph/debugfs.h>
27 
28 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
29 
30 /*
31  * A cluster of MDS (metadata server) daemons is responsible for
32  * managing the file system namespace (the directory hierarchy and
33  * inodes) and for coordinating shared access to storage.  Metadata is
34  * partitioning hierarchically across a number of servers, and that
35  * partition varies over time as the cluster adjusts the distribution
36  * in order to balance load.
37  *
38  * The MDS client is primarily responsible to managing synchronous
39  * metadata requests for operations like open, unlink, and so forth.
40  * If there is a MDS failure, we find out about it when we (possibly
41  * request and) receive a new MDS map, and can resubmit affected
42  * requests.
43  *
44  * For the most part, though, we take advantage of a lossless
45  * communications channel to the MDS, and do not need to worry about
46  * timing out or resubmitting requests.
47  *
48  * We maintain a stateful "session" with each MDS we interact with.
49  * Within each session, we sent periodic heartbeat messages to ensure
50  * any capabilities or leases we have been issues remain valid.  If
51  * the session times out and goes stale, our leases and capabilities
52  * are no longer valid.
53  */
54 
55 struct ceph_reconnect_state {
56 	struct ceph_mds_session *session;
57 	int nr_caps, nr_realms;
58 	struct ceph_pagelist *pagelist;
59 	unsigned msg_version;
60 	bool allow_multi;
61 };
62 
63 static void __wake_requests(struct ceph_mds_client *mdsc,
64 			    struct list_head *head);
65 static void ceph_cap_release_work(struct work_struct *work);
66 static void ceph_cap_reclaim_work(struct work_struct *work);
67 
68 static const struct ceph_connection_operations mds_con_ops;
69 
70 
71 /*
72  * mds reply parsing
73  */
74 
75 static int parse_reply_info_quota(void **p, void *end,
76 				  struct ceph_mds_reply_info_in *info)
77 {
78 	u8 struct_v, struct_compat;
79 	u32 struct_len;
80 
81 	ceph_decode_8_safe(p, end, struct_v, bad);
82 	ceph_decode_8_safe(p, end, struct_compat, bad);
83 	/* struct_v is expected to be >= 1. we only
84 	 * understand encoding with struct_compat == 1. */
85 	if (!struct_v || struct_compat != 1)
86 		goto bad;
87 	ceph_decode_32_safe(p, end, struct_len, bad);
88 	ceph_decode_need(p, end, struct_len, bad);
89 	end = *p + struct_len;
90 	ceph_decode_64_safe(p, end, info->max_bytes, bad);
91 	ceph_decode_64_safe(p, end, info->max_files, bad);
92 	*p = end;
93 	return 0;
94 bad:
95 	return -EIO;
96 }
97 
98 /*
99  * parse individual inode info
100  */
101 static int parse_reply_info_in(void **p, void *end,
102 			       struct ceph_mds_reply_info_in *info,
103 			       u64 features)
104 {
105 	int err = 0;
106 	u8 struct_v = 0;
107 
108 	if (features == (u64)-1) {
109 		u32 struct_len;
110 		u8 struct_compat;
111 		ceph_decode_8_safe(p, end, struct_v, bad);
112 		ceph_decode_8_safe(p, end, struct_compat, bad);
113 		/* struct_v is expected to be >= 1. we only understand
114 		 * encoding with struct_compat == 1. */
115 		if (!struct_v || struct_compat != 1)
116 			goto bad;
117 		ceph_decode_32_safe(p, end, struct_len, bad);
118 		ceph_decode_need(p, end, struct_len, bad);
119 		end = *p + struct_len;
120 	}
121 
122 	ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
123 	info->in = *p;
124 	*p += sizeof(struct ceph_mds_reply_inode) +
125 		sizeof(*info->in->fragtree.splits) *
126 		le32_to_cpu(info->in->fragtree.nsplits);
127 
128 	ceph_decode_32_safe(p, end, info->symlink_len, bad);
129 	ceph_decode_need(p, end, info->symlink_len, bad);
130 	info->symlink = *p;
131 	*p += info->symlink_len;
132 
133 	ceph_decode_copy_safe(p, end, &info->dir_layout,
134 			      sizeof(info->dir_layout), bad);
135 	ceph_decode_32_safe(p, end, info->xattr_len, bad);
136 	ceph_decode_need(p, end, info->xattr_len, bad);
137 	info->xattr_data = *p;
138 	*p += info->xattr_len;
139 
140 	if (features == (u64)-1) {
141 		/* inline data */
142 		ceph_decode_64_safe(p, end, info->inline_version, bad);
143 		ceph_decode_32_safe(p, end, info->inline_len, bad);
144 		ceph_decode_need(p, end, info->inline_len, bad);
145 		info->inline_data = *p;
146 		*p += info->inline_len;
147 		/* quota */
148 		err = parse_reply_info_quota(p, end, info);
149 		if (err < 0)
150 			goto out_bad;
151 		/* pool namespace */
152 		ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
153 		if (info->pool_ns_len > 0) {
154 			ceph_decode_need(p, end, info->pool_ns_len, bad);
155 			info->pool_ns_data = *p;
156 			*p += info->pool_ns_len;
157 		}
158 
159 		/* btime */
160 		ceph_decode_need(p, end, sizeof(info->btime), bad);
161 		ceph_decode_copy(p, &info->btime, sizeof(info->btime));
162 
163 		/* change attribute */
164 		ceph_decode_64_safe(p, end, info->change_attr, bad);
165 
166 		/* dir pin */
167 		if (struct_v >= 2) {
168 			ceph_decode_32_safe(p, end, info->dir_pin, bad);
169 		} else {
170 			info->dir_pin = -ENODATA;
171 		}
172 
173 		/* snapshot birth time, remains zero for v<=2 */
174 		if (struct_v >= 3) {
175 			ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
176 			ceph_decode_copy(p, &info->snap_btime,
177 					 sizeof(info->snap_btime));
178 		} else {
179 			memset(&info->snap_btime, 0, sizeof(info->snap_btime));
180 		}
181 
182 		/* snapshot count, remains zero for v<=3 */
183 		if (struct_v >= 4) {
184 			ceph_decode_64_safe(p, end, info->rsnaps, bad);
185 		} else {
186 			info->rsnaps = 0;
187 		}
188 
189 		if (struct_v >= 5) {
190 			u32 alen;
191 
192 			ceph_decode_32_safe(p, end, alen, bad);
193 
194 			while (alen--) {
195 				u32 len;
196 
197 				/* key */
198 				ceph_decode_32_safe(p, end, len, bad);
199 				ceph_decode_skip_n(p, end, len, bad);
200 				/* value */
201 				ceph_decode_32_safe(p, end, len, bad);
202 				ceph_decode_skip_n(p, end, len, bad);
203 			}
204 		}
205 
206 		/* fscrypt flag -- ignore */
207 		if (struct_v >= 6)
208 			ceph_decode_skip_8(p, end, bad);
209 
210 		info->fscrypt_auth = NULL;
211 		info->fscrypt_auth_len = 0;
212 		info->fscrypt_file = NULL;
213 		info->fscrypt_file_len = 0;
214 		if (struct_v >= 7) {
215 			ceph_decode_32_safe(p, end, info->fscrypt_auth_len, bad);
216 			if (info->fscrypt_auth_len) {
217 				info->fscrypt_auth = kmalloc(info->fscrypt_auth_len,
218 							     GFP_KERNEL);
219 				if (!info->fscrypt_auth)
220 					return -ENOMEM;
221 				ceph_decode_copy_safe(p, end, info->fscrypt_auth,
222 						      info->fscrypt_auth_len, bad);
223 			}
224 			ceph_decode_32_safe(p, end, info->fscrypt_file_len, bad);
225 			if (info->fscrypt_file_len) {
226 				info->fscrypt_file = kmalloc(info->fscrypt_file_len,
227 							     GFP_KERNEL);
228 				if (!info->fscrypt_file)
229 					return -ENOMEM;
230 				ceph_decode_copy_safe(p, end, info->fscrypt_file,
231 						      info->fscrypt_file_len, bad);
232 			}
233 		}
234 		*p = end;
235 	} else {
236 		/* legacy (unversioned) struct */
237 		if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
238 			ceph_decode_64_safe(p, end, info->inline_version, bad);
239 			ceph_decode_32_safe(p, end, info->inline_len, bad);
240 			ceph_decode_need(p, end, info->inline_len, bad);
241 			info->inline_data = *p;
242 			*p += info->inline_len;
243 		} else
244 			info->inline_version = CEPH_INLINE_NONE;
245 
246 		if (features & CEPH_FEATURE_MDS_QUOTA) {
247 			err = parse_reply_info_quota(p, end, info);
248 			if (err < 0)
249 				goto out_bad;
250 		} else {
251 			info->max_bytes = 0;
252 			info->max_files = 0;
253 		}
254 
255 		info->pool_ns_len = 0;
256 		info->pool_ns_data = NULL;
257 		if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
258 			ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
259 			if (info->pool_ns_len > 0) {
260 				ceph_decode_need(p, end, info->pool_ns_len, bad);
261 				info->pool_ns_data = *p;
262 				*p += info->pool_ns_len;
263 			}
264 		}
265 
266 		if (features & CEPH_FEATURE_FS_BTIME) {
267 			ceph_decode_need(p, end, sizeof(info->btime), bad);
268 			ceph_decode_copy(p, &info->btime, sizeof(info->btime));
269 			ceph_decode_64_safe(p, end, info->change_attr, bad);
270 		}
271 
272 		info->dir_pin = -ENODATA;
273 		/* info->snap_btime and info->rsnaps remain zero */
274 	}
275 	return 0;
276 bad:
277 	err = -EIO;
278 out_bad:
279 	return err;
280 }
281 
282 static int parse_reply_info_dir(void **p, void *end,
283 				struct ceph_mds_reply_dirfrag **dirfrag,
284 				u64 features)
285 {
286 	if (features == (u64)-1) {
287 		u8 struct_v, struct_compat;
288 		u32 struct_len;
289 		ceph_decode_8_safe(p, end, struct_v, bad);
290 		ceph_decode_8_safe(p, end, struct_compat, bad);
291 		/* struct_v is expected to be >= 1. we only understand
292 		 * encoding whose struct_compat == 1. */
293 		if (!struct_v || struct_compat != 1)
294 			goto bad;
295 		ceph_decode_32_safe(p, end, struct_len, bad);
296 		ceph_decode_need(p, end, struct_len, bad);
297 		end = *p + struct_len;
298 	}
299 
300 	ceph_decode_need(p, end, sizeof(**dirfrag), bad);
301 	*dirfrag = *p;
302 	*p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
303 	if (unlikely(*p > end))
304 		goto bad;
305 	if (features == (u64)-1)
306 		*p = end;
307 	return 0;
308 bad:
309 	return -EIO;
310 }
311 
312 static int parse_reply_info_lease(void **p, void *end,
313 				  struct ceph_mds_reply_lease **lease,
314 				  u64 features, u32 *altname_len, u8 **altname)
315 {
316 	u8 struct_v;
317 	u32 struct_len;
318 	void *lend;
319 
320 	if (features == (u64)-1) {
321 		u8 struct_compat;
322 
323 		ceph_decode_8_safe(p, end, struct_v, bad);
324 		ceph_decode_8_safe(p, end, struct_compat, bad);
325 
326 		/* struct_v is expected to be >= 1. we only understand
327 		 * encoding whose struct_compat == 1. */
328 		if (!struct_v || struct_compat != 1)
329 			goto bad;
330 
331 		ceph_decode_32_safe(p, end, struct_len, bad);
332 	} else {
333 		struct_len = sizeof(**lease);
334 		*altname_len = 0;
335 		*altname = NULL;
336 	}
337 
338 	lend = *p + struct_len;
339 	ceph_decode_need(p, end, struct_len, bad);
340 	*lease = *p;
341 	*p += sizeof(**lease);
342 
343 	if (features == (u64)-1) {
344 		if (struct_v >= 2) {
345 			ceph_decode_32_safe(p, end, *altname_len, bad);
346 			ceph_decode_need(p, end, *altname_len, bad);
347 			*altname = *p;
348 			*p += *altname_len;
349 		} else {
350 			*altname = NULL;
351 			*altname_len = 0;
352 		}
353 	}
354 	*p = lend;
355 	return 0;
356 bad:
357 	return -EIO;
358 }
359 
360 /*
361  * parse a normal reply, which may contain a (dir+)dentry and/or a
362  * target inode.
363  */
364 static int parse_reply_info_trace(void **p, void *end,
365 				  struct ceph_mds_reply_info_parsed *info,
366 				  u64 features)
367 {
368 	int err;
369 
370 	if (info->head->is_dentry) {
371 		err = parse_reply_info_in(p, end, &info->diri, features);
372 		if (err < 0)
373 			goto out_bad;
374 
375 		err = parse_reply_info_dir(p, end, &info->dirfrag, features);
376 		if (err < 0)
377 			goto out_bad;
378 
379 		ceph_decode_32_safe(p, end, info->dname_len, bad);
380 		ceph_decode_need(p, end, info->dname_len, bad);
381 		info->dname = *p;
382 		*p += info->dname_len;
383 
384 		err = parse_reply_info_lease(p, end, &info->dlease, features,
385 					     &info->altname_len, &info->altname);
386 		if (err < 0)
387 			goto out_bad;
388 	}
389 
390 	if (info->head->is_target) {
391 		err = parse_reply_info_in(p, end, &info->targeti, features);
392 		if (err < 0)
393 			goto out_bad;
394 	}
395 
396 	if (unlikely(*p != end))
397 		goto bad;
398 	return 0;
399 
400 bad:
401 	err = -EIO;
402 out_bad:
403 	pr_err("problem parsing mds trace %d\n", err);
404 	return err;
405 }
406 
407 /*
408  * parse readdir results
409  */
410 static int parse_reply_info_readdir(void **p, void *end,
411 				    struct ceph_mds_request *req,
412 				    u64 features)
413 {
414 	struct ceph_mds_reply_info_parsed *info = &req->r_reply_info;
415 	struct ceph_client *cl = req->r_mdsc->fsc->client;
416 	u32 num, i = 0;
417 	int err;
418 
419 	err = parse_reply_info_dir(p, end, &info->dir_dir, features);
420 	if (err < 0)
421 		goto out_bad;
422 
423 	ceph_decode_need(p, end, sizeof(num) + 2, bad);
424 	num = ceph_decode_32(p);
425 	{
426 		u16 flags = ceph_decode_16(p);
427 		info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
428 		info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
429 		info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
430 		info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
431 	}
432 	if (num == 0)
433 		goto done;
434 
435 	BUG_ON(!info->dir_entries);
436 	if ((unsigned long)(info->dir_entries + num) >
437 	    (unsigned long)info->dir_entries + info->dir_buf_size) {
438 		pr_err_client(cl, "dir contents are larger than expected\n");
439 		WARN_ON(1);
440 		goto bad;
441 	}
442 
443 	info->dir_nr = num;
444 	while (num) {
445 		struct inode *inode = d_inode(req->r_dentry);
446 		struct ceph_inode_info *ci = ceph_inode(inode);
447 		struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
448 		struct fscrypt_str tname = FSTR_INIT(NULL, 0);
449 		struct fscrypt_str oname = FSTR_INIT(NULL, 0);
450 		struct ceph_fname fname;
451 		u32 altname_len, _name_len;
452 		u8 *altname, *_name;
453 
454 		/* dentry */
455 		ceph_decode_32_safe(p, end, _name_len, bad);
456 		ceph_decode_need(p, end, _name_len, bad);
457 		_name = *p;
458 		*p += _name_len;
459 		doutc(cl, "parsed dir dname '%.*s'\n", _name_len, _name);
460 
461 		if (info->hash_order)
462 			rde->raw_hash = ceph_str_hash(ci->i_dir_layout.dl_dir_hash,
463 						      _name, _name_len);
464 
465 		/* dentry lease */
466 		err = parse_reply_info_lease(p, end, &rde->lease, features,
467 					     &altname_len, &altname);
468 		if (err)
469 			goto out_bad;
470 
471 		/*
472 		 * Try to dencrypt the dentry names and update them
473 		 * in the ceph_mds_reply_dir_entry struct.
474 		 */
475 		fname.dir = inode;
476 		fname.name = _name;
477 		fname.name_len = _name_len;
478 		fname.ctext = altname;
479 		fname.ctext_len = altname_len;
480 		/*
481 		 * The _name_len maybe larger than altname_len, such as
482 		 * when the human readable name length is in range of
483 		 * (CEPH_NOHASH_NAME_MAX, CEPH_NOHASH_NAME_MAX + SHA256_DIGEST_SIZE),
484 		 * then the copy in ceph_fname_to_usr will corrupt the
485 		 * data if there has no encryption key.
486 		 *
487 		 * Just set the no_copy flag and then if there has no
488 		 * encryption key the oname.name will be assigned to
489 		 * _name always.
490 		 */
491 		fname.no_copy = true;
492 		if (altname_len == 0) {
493 			/*
494 			 * Set tname to _name, and this will be used
495 			 * to do the base64_decode in-place. It's
496 			 * safe because the decoded string should
497 			 * always be shorter, which is 3/4 of origin
498 			 * string.
499 			 */
500 			tname.name = _name;
501 
502 			/*
503 			 * Set oname to _name too, and this will be
504 			 * used to do the dencryption in-place.
505 			 */
506 			oname.name = _name;
507 			oname.len = _name_len;
508 		} else {
509 			/*
510 			 * This will do the decryption only in-place
511 			 * from altname cryptext directly.
512 			 */
513 			oname.name = altname;
514 			oname.len = altname_len;
515 		}
516 		rde->is_nokey = false;
517 		err = ceph_fname_to_usr(&fname, &tname, &oname, &rde->is_nokey);
518 		if (err) {
519 			pr_err_client(cl, "unable to decode %.*s, got %d\n",
520 				      _name_len, _name, err);
521 			goto out_bad;
522 		}
523 		rde->name = oname.name;
524 		rde->name_len = oname.len;
525 
526 		/* inode */
527 		err = parse_reply_info_in(p, end, &rde->inode, features);
528 		if (err < 0)
529 			goto out_bad;
530 		/* ceph_readdir_prepopulate() will update it */
531 		rde->offset = 0;
532 		i++;
533 		num--;
534 	}
535 
536 done:
537 	/* Skip over any unrecognized fields */
538 	*p = end;
539 	return 0;
540 
541 bad:
542 	err = -EIO;
543 out_bad:
544 	pr_err_client(cl, "problem parsing dir contents %d\n", err);
545 	return err;
546 }
547 
548 /*
549  * parse fcntl F_GETLK results
550  */
551 static int parse_reply_info_filelock(void **p, void *end,
552 				     struct ceph_mds_reply_info_parsed *info,
553 				     u64 features)
554 {
555 	if (*p + sizeof(*info->filelock_reply) > end)
556 		goto bad;
557 
558 	info->filelock_reply = *p;
559 
560 	/* Skip over any unrecognized fields */
561 	*p = end;
562 	return 0;
563 bad:
564 	return -EIO;
565 }
566 
567 
568 #if BITS_PER_LONG == 64
569 
570 #define DELEGATED_INO_AVAILABLE		xa_mk_value(1)
571 
572 static int ceph_parse_deleg_inos(void **p, void *end,
573 				 struct ceph_mds_session *s)
574 {
575 	struct ceph_client *cl = s->s_mdsc->fsc->client;
576 	u32 sets;
577 
578 	ceph_decode_32_safe(p, end, sets, bad);
579 	doutc(cl, "got %u sets of delegated inodes\n", sets);
580 	while (sets--) {
581 		u64 start, len;
582 
583 		ceph_decode_64_safe(p, end, start, bad);
584 		ceph_decode_64_safe(p, end, len, bad);
585 
586 		/* Don't accept a delegation of system inodes */
587 		if (start < CEPH_INO_SYSTEM_BASE) {
588 			pr_warn_ratelimited_client(cl,
589 				"ignoring reserved inode range delegation (start=0x%llx len=0x%llx)\n",
590 				start, len);
591 			continue;
592 		}
593 		while (len--) {
594 			int err = xa_insert(&s->s_delegated_inos, start++,
595 					    DELEGATED_INO_AVAILABLE,
596 					    GFP_KERNEL);
597 			if (!err) {
598 				doutc(cl, "added delegated inode 0x%llx\n", start - 1);
599 			} else if (err == -EBUSY) {
600 				pr_warn_client(cl,
601 					"MDS delegated inode 0x%llx more than once.\n",
602 					start - 1);
603 			} else {
604 				return err;
605 			}
606 		}
607 	}
608 	return 0;
609 bad:
610 	return -EIO;
611 }
612 
613 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
614 {
615 	unsigned long ino;
616 	void *val;
617 
618 	xa_for_each(&s->s_delegated_inos, ino, val) {
619 		val = xa_erase(&s->s_delegated_inos, ino);
620 		if (val == DELEGATED_INO_AVAILABLE)
621 			return ino;
622 	}
623 	return 0;
624 }
625 
626 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
627 {
628 	return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,
629 			 GFP_KERNEL);
630 }
631 #else /* BITS_PER_LONG == 64 */
632 /*
633  * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
634  * ignore delegated_inos on 32 bit arch. Maybe eventually add xarrays for top
635  * and bottom words?
636  */
637 static int ceph_parse_deleg_inos(void **p, void *end,
638 				 struct ceph_mds_session *s)
639 {
640 	u32 sets;
641 
642 	ceph_decode_32_safe(p, end, sets, bad);
643 	if (sets)
644 		ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);
645 	return 0;
646 bad:
647 	return -EIO;
648 }
649 
650 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
651 {
652 	return 0;
653 }
654 
655 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
656 {
657 	return 0;
658 }
659 #endif /* BITS_PER_LONG == 64 */
660 
661 /*
662  * parse create results
663  */
664 static int parse_reply_info_create(void **p, void *end,
665 				  struct ceph_mds_reply_info_parsed *info,
666 				  u64 features, struct ceph_mds_session *s)
667 {
668 	int ret;
669 
670 	if (features == (u64)-1 ||
671 	    (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
672 		if (*p == end) {
673 			/* Malformed reply? */
674 			info->has_create_ino = false;
675 		} else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {
676 			info->has_create_ino = true;
677 			/* struct_v, struct_compat, and len */
678 			ceph_decode_skip_n(p, end, 2 + sizeof(u32), bad);
679 			ceph_decode_64_safe(p, end, info->ino, bad);
680 			ret = ceph_parse_deleg_inos(p, end, s);
681 			if (ret)
682 				return ret;
683 		} else {
684 			/* legacy */
685 			ceph_decode_64_safe(p, end, info->ino, bad);
686 			info->has_create_ino = true;
687 		}
688 	} else {
689 		if (*p != end)
690 			goto bad;
691 	}
692 
693 	/* Skip over any unrecognized fields */
694 	*p = end;
695 	return 0;
696 bad:
697 	return -EIO;
698 }
699 
700 static int parse_reply_info_getvxattr(void **p, void *end,
701 				      struct ceph_mds_reply_info_parsed *info,
702 				      u64 features)
703 {
704 	u32 value_len;
705 
706 	ceph_decode_skip_8(p, end, bad); /* skip current version: 1 */
707 	ceph_decode_skip_8(p, end, bad); /* skip first version: 1 */
708 	ceph_decode_skip_32(p, end, bad); /* skip payload length */
709 
710 	ceph_decode_32_safe(p, end, value_len, bad);
711 
712 	if (value_len == end - *p) {
713 	  info->xattr_info.xattr_value = *p;
714 	  info->xattr_info.xattr_value_len = value_len;
715 	  *p = end;
716 	  return value_len;
717 	}
718 bad:
719 	return -EIO;
720 }
721 
722 /*
723  * parse extra results
724  */
725 static int parse_reply_info_extra(void **p, void *end,
726 				  struct ceph_mds_request *req,
727 				  u64 features, struct ceph_mds_session *s)
728 {
729 	struct ceph_mds_reply_info_parsed *info = &req->r_reply_info;
730 	u32 op = le32_to_cpu(info->head->op);
731 
732 	if (op == CEPH_MDS_OP_GETFILELOCK)
733 		return parse_reply_info_filelock(p, end, info, features);
734 	else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
735 		return parse_reply_info_readdir(p, end, req, features);
736 	else if (op == CEPH_MDS_OP_CREATE)
737 		return parse_reply_info_create(p, end, info, features, s);
738 	else if (op == CEPH_MDS_OP_GETVXATTR)
739 		return parse_reply_info_getvxattr(p, end, info, features);
740 	else
741 		return -EIO;
742 }
743 
744 /*
745  * parse entire mds reply
746  */
747 static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg,
748 			    struct ceph_mds_request *req, u64 features)
749 {
750 	struct ceph_mds_reply_info_parsed *info = &req->r_reply_info;
751 	struct ceph_client *cl = s->s_mdsc->fsc->client;
752 	void *p, *end;
753 	u32 len;
754 	int err;
755 
756 	info->head = msg->front.iov_base;
757 	p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
758 	end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
759 
760 	/* trace */
761 	ceph_decode_32_safe(&p, end, len, bad);
762 	if (len > 0) {
763 		ceph_decode_need(&p, end, len, bad);
764 		err = parse_reply_info_trace(&p, p+len, info, features);
765 		if (err < 0)
766 			goto out_bad;
767 	}
768 
769 	/* extra */
770 	ceph_decode_32_safe(&p, end, len, bad);
771 	if (len > 0) {
772 		ceph_decode_need(&p, end, len, bad);
773 		err = parse_reply_info_extra(&p, p+len, req, features, s);
774 		if (err < 0)
775 			goto out_bad;
776 	}
777 
778 	/* snap blob */
779 	ceph_decode_32_safe(&p, end, len, bad);
780 	info->snapblob_len = len;
781 	info->snapblob = p;
782 	p += len;
783 
784 	if (p != end)
785 		goto bad;
786 	return 0;
787 
788 bad:
789 	err = -EIO;
790 out_bad:
791 	pr_err_client(cl, "mds parse_reply err %d\n", err);
792 	ceph_msg_dump(msg);
793 	return err;
794 }
795 
796 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
797 {
798 	int i;
799 
800 	kfree(info->diri.fscrypt_auth);
801 	kfree(info->diri.fscrypt_file);
802 	kfree(info->targeti.fscrypt_auth);
803 	kfree(info->targeti.fscrypt_file);
804 	if (!info->dir_entries)
805 		return;
806 
807 	for (i = 0; i < info->dir_nr; i++) {
808 		struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
809 
810 		kfree(rde->inode.fscrypt_auth);
811 		kfree(rde->inode.fscrypt_file);
812 	}
813 	free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
814 }
815 
816 /*
817  * In async unlink case the kclient won't wait for the first reply
818  * from MDS and just drop all the links and unhash the dentry and then
819  * succeeds immediately.
820  *
821  * For any new create/link/rename,etc requests followed by using the
822  * same file names we must wait for the first reply of the inflight
823  * unlink request, or the MDS possibly will fail these following
824  * requests with -EEXIST if the inflight async unlink request was
825  * delayed for some reasons.
826  *
827  * And the worst case is that for the none async openc request it will
828  * successfully open the file if the CDentry hasn't been unlinked yet,
829  * but later the previous delayed async unlink request will remove the
830  * CDenty. That means the just created file is possiblly deleted later
831  * by accident.
832  *
833  * We need to wait for the inflight async unlink requests to finish
834  * when creating new files/directories by using the same file names.
835  */
836 int ceph_wait_on_conflict_unlink(struct dentry *dentry)
837 {
838 	struct ceph_fs_client *fsc = ceph_sb_to_fs_client(dentry->d_sb);
839 	struct ceph_client *cl = fsc->client;
840 	struct dentry *pdentry = dentry->d_parent;
841 	struct dentry *udentry, *found = NULL;
842 	struct ceph_dentry_info *di;
843 	struct qstr dname;
844 	u32 hash = dentry->d_name.hash;
845 	int err;
846 
847 	dname.name = dentry->d_name.name;
848 	dname.len = dentry->d_name.len;
849 
850 	rcu_read_lock();
851 	hash_for_each_possible_rcu(fsc->async_unlink_conflict, di,
852 				   hnode, hash) {
853 		udentry = di->dentry;
854 
855 		spin_lock(&udentry->d_lock);
856 		if (udentry->d_name.hash != hash)
857 			goto next;
858 		if (unlikely(udentry->d_parent != pdentry))
859 			goto next;
860 		if (!hash_hashed(&di->hnode))
861 			goto next;
862 
863 		if (!test_bit(CEPH_DENTRY_ASYNC_UNLINK_BIT, &di->flags))
864 			pr_warn_client(cl, "dentry %p:%pd async unlink bit is not set\n",
865 				       dentry, dentry);
866 
867 		if (!d_same_name(udentry, pdentry, &dname))
868 			goto next;
869 
870 		found = dget_dlock(udentry);
871 		spin_unlock(&udentry->d_lock);
872 		break;
873 next:
874 		spin_unlock(&udentry->d_lock);
875 	}
876 	rcu_read_unlock();
877 
878 	if (likely(!found))
879 		return 0;
880 
881 	doutc(cl, "dentry %p:%pd conflict with old %p:%pd\n", dentry, dentry,
882 	      found, found);
883 
884 	err = wait_on_bit(&di->flags, CEPH_DENTRY_ASYNC_UNLINK_BIT,
885 			  TASK_KILLABLE);
886 	dput(found);
887 	return err;
888 }
889 
890 
891 /*
892  * sessions
893  */
894 const char *ceph_session_state_name(int s)
895 {
896 	switch (s) {
897 	case CEPH_MDS_SESSION_NEW: return "new";
898 	case CEPH_MDS_SESSION_OPENING: return "opening";
899 	case CEPH_MDS_SESSION_OPEN: return "open";
900 	case CEPH_MDS_SESSION_HUNG: return "hung";
901 	case CEPH_MDS_SESSION_CLOSING: return "closing";
902 	case CEPH_MDS_SESSION_CLOSED: return "closed";
903 	case CEPH_MDS_SESSION_RESTARTING: return "restarting";
904 	case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
905 	case CEPH_MDS_SESSION_REJECTED: return "rejected";
906 	default: return "???";
907 	}
908 }
909 
910 struct ceph_mds_session *ceph_get_mds_session(struct ceph_mds_session *s)
911 {
912 	if (refcount_inc_not_zero(&s->s_ref))
913 		return s;
914 	return NULL;
915 }
916 
917 void ceph_put_mds_session(struct ceph_mds_session *s)
918 {
919 	if (IS_ERR_OR_NULL(s))
920 		return;
921 
922 	if (refcount_dec_and_test(&s->s_ref)) {
923 		if (s->s_auth.authorizer)
924 			ceph_auth_destroy_authorizer(s->s_auth.authorizer);
925 		WARN_ON(mutex_is_locked(&s->s_mutex));
926 		xa_destroy(&s->s_delegated_inos);
927 		kfree(s);
928 	}
929 }
930 
931 /*
932  * called under mdsc->mutex
933  */
934 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
935 						   int mds)
936 {
937 	if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
938 		return NULL;
939 	return ceph_get_mds_session(mdsc->sessions[mds]);
940 }
941 
942 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
943 {
944 	if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
945 		return false;
946 	else
947 		return true;
948 }
949 
950 static int __verify_registered_session(struct ceph_mds_client *mdsc,
951 				       struct ceph_mds_session *s)
952 {
953 	if (s->s_mds >= mdsc->max_sessions ||
954 	    mdsc->sessions[s->s_mds] != s)
955 		return -ENOENT;
956 	return 0;
957 }
958 
959 /*
960  * create+register a new session for given mds.
961  * called under mdsc->mutex.
962  */
963 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
964 						 int mds)
965 {
966 	struct ceph_client *cl = mdsc->fsc->client;
967 	struct ceph_mds_session *s;
968 
969 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO)
970 		return ERR_PTR(-EIO);
971 
972 	if (mds >= mdsc->mdsmap->possible_max_rank)
973 		return ERR_PTR(-EINVAL);
974 
975 	s = kzalloc(sizeof(*s), GFP_NOFS);
976 	if (!s)
977 		return ERR_PTR(-ENOMEM);
978 
979 	if (mds >= mdsc->max_sessions) {
980 		int newmax = 1 << get_count_order(mds + 1);
981 		struct ceph_mds_session **sa;
982 
983 		doutc(cl, "realloc to %d\n", newmax);
984 		sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
985 		if (!sa)
986 			goto fail_realloc;
987 		if (mdsc->sessions) {
988 			memcpy(sa, mdsc->sessions,
989 			       mdsc->max_sessions * sizeof(void *));
990 			kfree(mdsc->sessions);
991 		}
992 		mdsc->sessions = sa;
993 		mdsc->max_sessions = newmax;
994 	}
995 
996 	doutc(cl, "mds%d\n", mds);
997 	s->s_mdsc = mdsc;
998 	s->s_mds = mds;
999 	s->s_state = CEPH_MDS_SESSION_NEW;
1000 	mutex_init(&s->s_mutex);
1001 
1002 	ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
1003 
1004 	atomic_set(&s->s_cap_gen, 1);
1005 	s->s_cap_ttl = jiffies - 1;
1006 
1007 	spin_lock_init(&s->s_cap_lock);
1008 	INIT_LIST_HEAD(&s->s_caps);
1009 	refcount_set(&s->s_ref, 1);
1010 	INIT_LIST_HEAD(&s->s_waiting);
1011 	INIT_LIST_HEAD(&s->s_unsafe);
1012 	xa_init(&s->s_delegated_inos);
1013 	INIT_LIST_HEAD(&s->s_cap_releases);
1014 	INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
1015 
1016 	INIT_LIST_HEAD(&s->s_cap_dirty);
1017 	INIT_LIST_HEAD(&s->s_cap_flushing);
1018 
1019 	mdsc->sessions[mds] = s;
1020 	atomic_inc(&mdsc->num_sessions);
1021 	refcount_inc(&s->s_ref);  /* one ref to sessions[], one to caller */
1022 
1023 	ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
1024 		      ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
1025 
1026 	return s;
1027 
1028 fail_realloc:
1029 	kfree(s);
1030 	return ERR_PTR(-ENOMEM);
1031 }
1032 
1033 /*
1034  * called under mdsc->mutex
1035  */
1036 static void __unregister_session(struct ceph_mds_client *mdsc,
1037 			       struct ceph_mds_session *s)
1038 {
1039 	doutc(mdsc->fsc->client, "mds%d %p\n", s->s_mds, s);
1040 	BUG_ON(mdsc->sessions[s->s_mds] != s);
1041 	mdsc->sessions[s->s_mds] = NULL;
1042 	ceph_con_close(&s->s_con);
1043 	ceph_put_mds_session(s);
1044 	atomic_dec(&mdsc->num_sessions);
1045 }
1046 
1047 /*
1048  * drop session refs in request.
1049  *
1050  * should be last request ref, or hold mdsc->mutex
1051  */
1052 static void put_request_session(struct ceph_mds_request *req)
1053 {
1054 	if (req->r_session) {
1055 		ceph_put_mds_session(req->r_session);
1056 		req->r_session = NULL;
1057 	}
1058 }
1059 
1060 void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc,
1061 				void (*cb)(struct ceph_mds_session *),
1062 				bool check_state)
1063 {
1064 	int mds;
1065 
1066 	mutex_lock(&mdsc->mutex);
1067 	for (mds = 0; mds < mdsc->max_sessions; ++mds) {
1068 		struct ceph_mds_session *s;
1069 
1070 		s = __ceph_lookup_mds_session(mdsc, mds);
1071 		if (!s)
1072 			continue;
1073 
1074 		if (check_state && !check_session_state(s)) {
1075 			ceph_put_mds_session(s);
1076 			continue;
1077 		}
1078 
1079 		mutex_unlock(&mdsc->mutex);
1080 		cb(s);
1081 		ceph_put_mds_session(s);
1082 		mutex_lock(&mdsc->mutex);
1083 	}
1084 	mutex_unlock(&mdsc->mutex);
1085 }
1086 
1087 void ceph_mdsc_release_request(struct kref *kref)
1088 {
1089 	struct ceph_mds_request *req = container_of(kref,
1090 						    struct ceph_mds_request,
1091 						    r_kref);
1092 	ceph_mdsc_release_dir_caps_no_check(req);
1093 	destroy_reply_info(&req->r_reply_info);
1094 	if (req->r_request)
1095 		ceph_msg_put(req->r_request);
1096 	if (req->r_reply)
1097 		ceph_msg_put(req->r_reply);
1098 	if (req->r_inode) {
1099 		ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
1100 		iput(req->r_inode);
1101 	}
1102 	if (req->r_parent) {
1103 		ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
1104 		iput(req->r_parent);
1105 	}
1106 	iput(req->r_target_inode);
1107 	iput(req->r_new_inode);
1108 	if (req->r_dentry)
1109 		dput(req->r_dentry);
1110 	if (req->r_old_dentry)
1111 		dput(req->r_old_dentry);
1112 	if (req->r_old_dentry_dir) {
1113 		/*
1114 		 * track (and drop pins for) r_old_dentry_dir
1115 		 * separately, since r_old_dentry's d_parent may have
1116 		 * changed between the dir mutex being dropped and
1117 		 * this request being freed.
1118 		 */
1119 		ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
1120 				  CEPH_CAP_PIN);
1121 		iput(req->r_old_dentry_dir);
1122 	}
1123 	kfree(req->r_path1);
1124 	kfree(req->r_path2);
1125 	put_cred(req->r_cred);
1126 	if (req->r_mnt_idmap)
1127 		mnt_idmap_put(req->r_mnt_idmap);
1128 	if (req->r_pagelist)
1129 		ceph_pagelist_release(req->r_pagelist);
1130 	kfree(req->r_fscrypt_auth);
1131 	kfree(req->r_altname);
1132 	put_request_session(req);
1133 	ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
1134 	WARN_ON_ONCE(!list_empty(&req->r_wait));
1135 	kmem_cache_free(ceph_mds_request_cachep, req);
1136 }
1137 
1138 DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
1139 
1140 /*
1141  * lookup session, bump ref if found.
1142  *
1143  * called under mdsc->mutex.
1144  */
1145 static struct ceph_mds_request *
1146 lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
1147 {
1148 	struct ceph_mds_request *req;
1149 
1150 	req = lookup_request(&mdsc->request_tree, tid);
1151 	if (req)
1152 		ceph_mdsc_get_request(req);
1153 
1154 	return req;
1155 }
1156 
1157 /*
1158  * Register an in-flight request, and assign a tid.  Link to directory
1159  * are modifying (if any).
1160  *
1161  * Called under mdsc->mutex.
1162  */
1163 static void __register_request(struct ceph_mds_client *mdsc,
1164 			       struct ceph_mds_request *req,
1165 			       struct inode *dir)
1166 {
1167 	struct ceph_client *cl = mdsc->fsc->client;
1168 	int ret = 0;
1169 
1170 	req->r_tid = ++mdsc->last_tid;
1171 	if (req->r_num_caps) {
1172 		ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
1173 					req->r_num_caps);
1174 		if (ret < 0) {
1175 			pr_err_client(cl, "%p failed to reserve caps: %d\n",
1176 				      req, ret);
1177 			/* set req->r_err to fail early from __do_request */
1178 			req->r_err = ret;
1179 			return;
1180 		}
1181 	}
1182 	doutc(cl, "%p tid %lld\n", req, req->r_tid);
1183 	ceph_mdsc_get_request(req);
1184 	insert_request(&mdsc->request_tree, req);
1185 
1186 	req->r_cred = get_current_cred();
1187 	if (!req->r_mnt_idmap)
1188 		req->r_mnt_idmap = &nop_mnt_idmap;
1189 
1190 	if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
1191 		mdsc->oldest_tid = req->r_tid;
1192 
1193 	if (dir) {
1194 		struct ceph_inode_info *ci = ceph_inode(dir);
1195 
1196 		ihold(dir);
1197 		req->r_unsafe_dir = dir;
1198 		spin_lock(&ci->i_unsafe_lock);
1199 		list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
1200 		spin_unlock(&ci->i_unsafe_lock);
1201 	}
1202 }
1203 
1204 static void __unregister_request(struct ceph_mds_client *mdsc,
1205 				 struct ceph_mds_request *req)
1206 {
1207 	doutc(mdsc->fsc->client, "%p tid %lld\n", req, req->r_tid);
1208 
1209 	/* Never leave an unregistered request on an unsafe list! */
1210 	list_del_init(&req->r_unsafe_item);
1211 
1212 	if (req->r_tid == mdsc->oldest_tid) {
1213 		struct rb_node *p = rb_next(&req->r_node);
1214 		mdsc->oldest_tid = 0;
1215 		while (p) {
1216 			struct ceph_mds_request *next_req =
1217 				rb_entry(p, struct ceph_mds_request, r_node);
1218 			if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
1219 				mdsc->oldest_tid = next_req->r_tid;
1220 				break;
1221 			}
1222 			p = rb_next(p);
1223 		}
1224 	}
1225 
1226 	erase_request(&mdsc->request_tree, req);
1227 
1228 	if (req->r_unsafe_dir) {
1229 		struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
1230 		spin_lock(&ci->i_unsafe_lock);
1231 		list_del_init(&req->r_unsafe_dir_item);
1232 		spin_unlock(&ci->i_unsafe_lock);
1233 	}
1234 	if (req->r_target_inode &&
1235 	    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
1236 		struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
1237 		spin_lock(&ci->i_unsafe_lock);
1238 		list_del_init(&req->r_unsafe_target_item);
1239 		spin_unlock(&ci->i_unsafe_lock);
1240 	}
1241 
1242 	if (req->r_unsafe_dir) {
1243 		iput(req->r_unsafe_dir);
1244 		req->r_unsafe_dir = NULL;
1245 	}
1246 
1247 	complete_all(&req->r_safe_completion);
1248 
1249 	ceph_mdsc_put_request(req);
1250 }
1251 
1252 /*
1253  * Walk back up the dentry tree until we hit a dentry representing a
1254  * non-snapshot inode. We do this using the rcu_read_lock (which must be held
1255  * when calling this) to ensure that the objects won't disappear while we're
1256  * working with them. Once we hit a candidate dentry, we attempt to take a
1257  * reference to it, and return that as the result.
1258  */
1259 static struct inode *get_nonsnap_parent(struct dentry *dentry)
1260 {
1261 	struct inode *inode = NULL;
1262 
1263 	while (dentry && !IS_ROOT(dentry)) {
1264 		inode = d_inode_rcu(dentry);
1265 		if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
1266 			break;
1267 		dentry = dentry->d_parent;
1268 	}
1269 	if (inode)
1270 		inode = igrab(inode);
1271 	return inode;
1272 }
1273 
1274 /*
1275  * Choose mds to send request to next.  If there is a hint set in the
1276  * request (e.g., due to a prior forward hint from the mds), use that.
1277  * Otherwise, consult frag tree and/or caps to identify the
1278  * appropriate mds.  If all else fails, choose randomly.
1279  *
1280  * Called under mdsc->mutex.
1281  */
1282 static int __choose_mds(struct ceph_mds_client *mdsc,
1283 			struct ceph_mds_request *req,
1284 			bool *random)
1285 {
1286 	struct inode *inode;
1287 	struct ceph_inode_info *ci;
1288 	struct ceph_cap *cap;
1289 	int mode = req->r_direct_mode;
1290 	int mds = -1;
1291 	u32 hash = req->r_direct_hash;
1292 	bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
1293 	struct ceph_client *cl = mdsc->fsc->client;
1294 
1295 	if (random)
1296 		*random = false;
1297 
1298 	/*
1299 	 * is there a specific mds we should try?  ignore hint if we have
1300 	 * no session and the mds is not up (active or recovering).
1301 	 */
1302 	if (req->r_resend_mds >= 0 &&
1303 	    (__have_session(mdsc, req->r_resend_mds) ||
1304 	     ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
1305 		doutc(cl, "using resend_mds mds%d\n", req->r_resend_mds);
1306 		return req->r_resend_mds;
1307 	}
1308 
1309 	if (mode == USE_RANDOM_MDS)
1310 		goto random;
1311 
1312 	inode = NULL;
1313 	if (req->r_inode) {
1314 		if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
1315 			inode = req->r_inode;
1316 			ihold(inode);
1317 		} else {
1318 			/* req->r_dentry is non-null for LSSNAP request */
1319 			rcu_read_lock();
1320 			inode = get_nonsnap_parent(req->r_dentry);
1321 			rcu_read_unlock();
1322 			doutc(cl, "using snapdir's parent %p %llx.%llx\n",
1323 			      inode, ceph_vinop(inode));
1324 		}
1325 	} else if (req->r_dentry) {
1326 		/* ignore race with rename; old or new d_parent is okay */
1327 		struct dentry *parent;
1328 		struct inode *dir;
1329 
1330 		rcu_read_lock();
1331 		parent = READ_ONCE(req->r_dentry->d_parent);
1332 		dir = req->r_parent ? : d_inode_rcu(parent);
1333 
1334 		if (!dir || dir->i_sb != mdsc->fsc->sb) {
1335 			/*  not this fs or parent went negative */
1336 			inode = d_inode(req->r_dentry);
1337 			if (inode)
1338 				ihold(inode);
1339 		} else if (ceph_snap(dir) != CEPH_NOSNAP) {
1340 			/* direct snapped/virtual snapdir requests
1341 			 * based on parent dir inode */
1342 			inode = get_nonsnap_parent(parent);
1343 			doutc(cl, "using nonsnap parent %p %llx.%llx\n",
1344 			      inode, ceph_vinop(inode));
1345 		} else {
1346 			/* dentry target */
1347 			inode = d_inode(req->r_dentry);
1348 			if (!inode || mode == USE_AUTH_MDS) {
1349 				/* dir + name */
1350 				inode = igrab(dir);
1351 				hash = ceph_dentry_hash(dir, req->r_dentry);
1352 				is_hash = true;
1353 			} else {
1354 				ihold(inode);
1355 			}
1356 		}
1357 		rcu_read_unlock();
1358 	}
1359 
1360 	if (!inode)
1361 		goto random;
1362 
1363 	doutc(cl, "%p %llx.%llx is_hash=%d (0x%x) mode %d\n", inode,
1364 	      ceph_vinop(inode), (int)is_hash, hash, mode);
1365 	ci = ceph_inode(inode);
1366 
1367 	if (is_hash && S_ISDIR(inode->i_mode)) {
1368 		struct ceph_inode_frag frag;
1369 		int found;
1370 
1371 		ceph_choose_frag(ci, hash, &frag, &found);
1372 		if (found) {
1373 			if (mode == USE_ANY_MDS && frag.ndist > 0) {
1374 				u8 r;
1375 
1376 				/* choose a random replica */
1377 				get_random_bytes(&r, 1);
1378 				r %= frag.ndist;
1379 				mds = frag.dist[r];
1380 				doutc(cl, "%p %llx.%llx frag %u mds%d (%d/%d)\n",
1381 				      inode, ceph_vinop(inode), frag.frag,
1382 				      mds, (int)r, frag.ndist);
1383 				if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1384 				    CEPH_MDS_STATE_ACTIVE &&
1385 				    !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))
1386 					goto out;
1387 			}
1388 
1389 			/* since this file/dir wasn't known to be
1390 			 * replicated, then we want to look for the
1391 			 * authoritative mds. */
1392 			if (frag.mds >= 0) {
1393 				/* choose auth mds */
1394 				mds = frag.mds;
1395 				doutc(cl, "%p %llx.%llx frag %u mds%d (auth)\n",
1396 				      inode, ceph_vinop(inode), frag.frag, mds);
1397 				if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1398 				    CEPH_MDS_STATE_ACTIVE) {
1399 					if (!ceph_mdsmap_is_laggy(mdsc->mdsmap,
1400 								  mds))
1401 						goto out;
1402 				}
1403 			}
1404 			mode = USE_AUTH_MDS;
1405 		}
1406 	}
1407 
1408 	spin_lock(&ci->i_ceph_lock);
1409 	cap = NULL;
1410 	if (mode == USE_AUTH_MDS)
1411 		cap = ci->i_auth_cap;
1412 	if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1413 		cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1414 	if (!cap) {
1415 		spin_unlock(&ci->i_ceph_lock);
1416 		iput(inode);
1417 		goto random;
1418 	}
1419 	mds = cap->session->s_mds;
1420 	doutc(cl, "%p %llx.%llx mds%d (%scap %p)\n", inode,
1421 	      ceph_vinop(inode), mds,
1422 	      cap == ci->i_auth_cap ? "auth " : "", cap);
1423 	spin_unlock(&ci->i_ceph_lock);
1424 out:
1425 	iput(inode);
1426 	return mds;
1427 
1428 random:
1429 	if (random)
1430 		*random = true;
1431 
1432 	mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1433 	doutc(cl, "chose random mds%d\n", mds);
1434 	return mds;
1435 }
1436 
1437 
1438 /*
1439  * session messages
1440  */
1441 struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq)
1442 {
1443 	struct ceph_msg *msg;
1444 	struct ceph_mds_session_head *h;
1445 
1446 	msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1447 			   false);
1448 	if (!msg) {
1449 		pr_err("ENOMEM creating session %s msg\n",
1450 		       ceph_session_op_name(op));
1451 		return NULL;
1452 	}
1453 	h = msg->front.iov_base;
1454 	h->op = cpu_to_le32(op);
1455 	h->seq = cpu_to_le64(seq);
1456 
1457 	return msg;
1458 }
1459 
1460 static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1461 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
1462 static int encode_supported_features(void **p, void *end)
1463 {
1464 	static const size_t count = ARRAY_SIZE(feature_bits);
1465 
1466 	if (count > 0) {
1467 		size_t i;
1468 		size_t size = FEATURE_BYTES(count);
1469 		unsigned long bit;
1470 
1471 		if (WARN_ON_ONCE(*p + 4 + size > end))
1472 			return -ERANGE;
1473 
1474 		ceph_encode_32(p, size);
1475 		memset(*p, 0, size);
1476 		for (i = 0; i < count; i++) {
1477 			bit = feature_bits[i];
1478 			((unsigned char *)(*p))[bit / 8] |= BIT(bit % 8);
1479 		}
1480 		*p += size;
1481 	} else {
1482 		if (WARN_ON_ONCE(*p + 4 > end))
1483 			return -ERANGE;
1484 
1485 		ceph_encode_32(p, 0);
1486 	}
1487 
1488 	return 0;
1489 }
1490 
1491 static const unsigned char metric_bits[] = CEPHFS_METRIC_SPEC_CLIENT_SUPPORTED;
1492 #define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
1493 static int encode_metric_spec(void **p, void *end)
1494 {
1495 	static const size_t count = ARRAY_SIZE(metric_bits);
1496 
1497 	/* header */
1498 	if (WARN_ON_ONCE(*p + 2 > end))
1499 		return -ERANGE;
1500 
1501 	ceph_encode_8(p, 1); /* version */
1502 	ceph_encode_8(p, 1); /* compat */
1503 
1504 	if (count > 0) {
1505 		size_t i;
1506 		size_t size = METRIC_BYTES(count);
1507 
1508 		if (WARN_ON_ONCE(*p + 4 + 4 + size > end))
1509 			return -ERANGE;
1510 
1511 		/* metric spec info length */
1512 		ceph_encode_32(p, 4 + size);
1513 
1514 		/* metric spec */
1515 		ceph_encode_32(p, size);
1516 		memset(*p, 0, size);
1517 		for (i = 0; i < count; i++)
1518 			((unsigned char *)(*p))[i / 8] |= BIT(metric_bits[i] % 8);
1519 		*p += size;
1520 	} else {
1521 		if (WARN_ON_ONCE(*p + 4 + 4 > end))
1522 			return -ERANGE;
1523 
1524 		/* metric spec info length */
1525 		ceph_encode_32(p, 4);
1526 		/* metric spec */
1527 		ceph_encode_32(p, 0);
1528 	}
1529 
1530 	return 0;
1531 }
1532 
1533 /*
1534  * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1535  * to include additional client metadata fields.
1536  */
1537 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1538 {
1539 	struct ceph_msg *msg;
1540 	struct ceph_mds_session_head *h;
1541 	int i;
1542 	int extra_bytes = 0;
1543 	int metadata_key_count = 0;
1544 	struct ceph_options *opt = mdsc->fsc->client->options;
1545 	struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1546 	struct ceph_client *cl = mdsc->fsc->client;
1547 	size_t size, count;
1548 	void *p, *end;
1549 	int ret;
1550 
1551 	const char* metadata[][2] = {
1552 		{"hostname", mdsc->nodename},
1553 		{"kernel_version", init_utsname()->release},
1554 		{"entity_id", opt->name ? : ""},
1555 		{"root", fsopt->server_path ? : "/"},
1556 		{NULL, NULL}
1557 	};
1558 
1559 	/* Calculate serialized length of metadata */
1560 	extra_bytes = 4;  /* map length */
1561 	for (i = 0; metadata[i][0]; ++i) {
1562 		extra_bytes += 8 + strlen(metadata[i][0]) +
1563 			strlen(metadata[i][1]);
1564 		metadata_key_count++;
1565 	}
1566 
1567 	/* supported feature */
1568 	size = 0;
1569 	count = ARRAY_SIZE(feature_bits);
1570 	if (count > 0)
1571 		size = FEATURE_BYTES(count);
1572 	extra_bytes += 4 + size;
1573 
1574 	/* metric spec */
1575 	size = 0;
1576 	count = ARRAY_SIZE(metric_bits);
1577 	if (count > 0)
1578 		size = METRIC_BYTES(count);
1579 	extra_bytes += 2 + 4 + 4 + size;
1580 
1581 	/* Allocate the message */
1582 	msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1583 			   GFP_NOFS, false);
1584 	if (!msg) {
1585 		pr_err_client(cl, "ENOMEM creating session open msg\n");
1586 		return ERR_PTR(-ENOMEM);
1587 	}
1588 	p = msg->front.iov_base;
1589 	end = p + msg->front.iov_len;
1590 
1591 	h = p;
1592 	h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1593 	h->seq = cpu_to_le64(seq);
1594 
1595 	/*
1596 	 * Serialize client metadata into waiting buffer space, using
1597 	 * the format that userspace expects for map<string, string>
1598 	 *
1599 	 * ClientSession messages with metadata are v4
1600 	 */
1601 	msg->hdr.version = cpu_to_le16(4);
1602 	msg->hdr.compat_version = cpu_to_le16(1);
1603 
1604 	/* The write pointer, following the session_head structure */
1605 	p += sizeof(*h);
1606 
1607 	/* Number of entries in the map */
1608 	ceph_encode_32(&p, metadata_key_count);
1609 
1610 	/* Two length-prefixed strings for each entry in the map */
1611 	for (i = 0; metadata[i][0]; ++i) {
1612 		size_t const key_len = strlen(metadata[i][0]);
1613 		size_t const val_len = strlen(metadata[i][1]);
1614 
1615 		ceph_encode_32(&p, key_len);
1616 		memcpy(p, metadata[i][0], key_len);
1617 		p += key_len;
1618 		ceph_encode_32(&p, val_len);
1619 		memcpy(p, metadata[i][1], val_len);
1620 		p += val_len;
1621 	}
1622 
1623 	ret = encode_supported_features(&p, end);
1624 	if (ret) {
1625 		pr_err_client(cl, "encode_supported_features failed!\n");
1626 		ceph_msg_put(msg);
1627 		return ERR_PTR(ret);
1628 	}
1629 
1630 	ret = encode_metric_spec(&p, end);
1631 	if (ret) {
1632 		pr_err_client(cl, "encode_metric_spec failed!\n");
1633 		ceph_msg_put(msg);
1634 		return ERR_PTR(ret);
1635 	}
1636 
1637 	msg->front.iov_len = p - msg->front.iov_base;
1638 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1639 
1640 	return msg;
1641 }
1642 
1643 /*
1644  * send session open request.
1645  *
1646  * called under mdsc->mutex
1647  */
1648 static int __open_session(struct ceph_mds_client *mdsc,
1649 			  struct ceph_mds_session *session)
1650 {
1651 	struct ceph_msg *msg;
1652 	int mstate;
1653 	int mds = session->s_mds;
1654 
1655 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO)
1656 		return -EIO;
1657 
1658 	/* wait for mds to go active? */
1659 	mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1660 	doutc(mdsc->fsc->client, "open_session to mds%d (%s)\n", mds,
1661 	      ceph_mds_state_name(mstate));
1662 	session->s_state = CEPH_MDS_SESSION_OPENING;
1663 	session->s_renew_requested = jiffies;
1664 
1665 	/* send connect message */
1666 	msg = create_session_open_msg(mdsc, session->s_seq);
1667 	if (IS_ERR(msg))
1668 		return PTR_ERR(msg);
1669 	ceph_con_send(&session->s_con, msg);
1670 	return 0;
1671 }
1672 
1673 /*
1674  * open sessions for any export targets for the given mds
1675  *
1676  * called under mdsc->mutex
1677  */
1678 static struct ceph_mds_session *
1679 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1680 {
1681 	struct ceph_mds_session *session;
1682 	int ret;
1683 
1684 	session = __ceph_lookup_mds_session(mdsc, target);
1685 	if (!session) {
1686 		session = register_session(mdsc, target);
1687 		if (IS_ERR(session))
1688 			return session;
1689 	}
1690 	if (session->s_state == CEPH_MDS_SESSION_NEW ||
1691 	    session->s_state == CEPH_MDS_SESSION_CLOSING) {
1692 		ret = __open_session(mdsc, session);
1693 		if (ret)
1694 			return ERR_PTR(ret);
1695 	}
1696 
1697 	return session;
1698 }
1699 
1700 struct ceph_mds_session *
1701 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1702 {
1703 	struct ceph_mds_session *session;
1704 	struct ceph_client *cl = mdsc->fsc->client;
1705 
1706 	doutc(cl, "to mds%d\n", target);
1707 
1708 	mutex_lock(&mdsc->mutex);
1709 	session = __open_export_target_session(mdsc, target);
1710 	mutex_unlock(&mdsc->mutex);
1711 
1712 	return session;
1713 }
1714 
1715 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1716 					  struct ceph_mds_session *session)
1717 {
1718 	struct ceph_mds_info *mi;
1719 	struct ceph_mds_session *ts;
1720 	int i, mds = session->s_mds;
1721 	struct ceph_client *cl = mdsc->fsc->client;
1722 
1723 	if (mds >= mdsc->mdsmap->possible_max_rank)
1724 		return;
1725 
1726 	mi = &mdsc->mdsmap->m_info[mds];
1727 	doutc(cl, "for mds%d (%d targets)\n", session->s_mds,
1728 	      mi->num_export_targets);
1729 
1730 	for (i = 0; i < mi->num_export_targets; i++) {
1731 		ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1732 		ceph_put_mds_session(ts);
1733 	}
1734 }
1735 
1736 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1737 					   struct ceph_mds_session *session)
1738 {
1739 	mutex_lock(&mdsc->mutex);
1740 	__open_export_target_sessions(mdsc, session);
1741 	mutex_unlock(&mdsc->mutex);
1742 }
1743 
1744 /*
1745  * session caps
1746  */
1747 
1748 static void detach_cap_releases(struct ceph_mds_session *session,
1749 				struct list_head *target)
1750 {
1751 	struct ceph_client *cl = session->s_mdsc->fsc->client;
1752 
1753 	lockdep_assert_held(&session->s_cap_lock);
1754 
1755 	list_splice_init(&session->s_cap_releases, target);
1756 	session->s_num_cap_releases = 0;
1757 	doutc(cl, "mds%d\n", session->s_mds);
1758 }
1759 
1760 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1761 				 struct list_head *dispose)
1762 {
1763 	while (!list_empty(dispose)) {
1764 		struct ceph_cap *cap;
1765 		/* zero out the in-progress message */
1766 		cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1767 		list_del(&cap->session_caps);
1768 		ceph_put_cap(mdsc, cap);
1769 	}
1770 }
1771 
1772 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1773 				     struct ceph_mds_session *session)
1774 {
1775 	struct ceph_client *cl = mdsc->fsc->client;
1776 	struct ceph_mds_request *req;
1777 	struct rb_node *p;
1778 
1779 	doutc(cl, "mds%d\n", session->s_mds);
1780 	mutex_lock(&mdsc->mutex);
1781 	while (!list_empty(&session->s_unsafe)) {
1782 		req = list_first_entry(&session->s_unsafe,
1783 				       struct ceph_mds_request, r_unsafe_item);
1784 		pr_warn_ratelimited_client(cl, " dropping unsafe request %llu\n",
1785 					   req->r_tid);
1786 		if (req->r_target_inode)
1787 			mapping_set_error(req->r_target_inode->i_mapping, -EIO);
1788 		if (req->r_unsafe_dir)
1789 			mapping_set_error(req->r_unsafe_dir->i_mapping, -EIO);
1790 		__unregister_request(mdsc, req);
1791 	}
1792 	/* zero r_attempts, so kick_requests() will re-send requests */
1793 	p = rb_first(&mdsc->request_tree);
1794 	while (p) {
1795 		req = rb_entry(p, struct ceph_mds_request, r_node);
1796 		p = rb_next(p);
1797 		if (req->r_session &&
1798 		    req->r_session->s_mds == session->s_mds)
1799 			req->r_attempts = 0;
1800 	}
1801 	mutex_unlock(&mdsc->mutex);
1802 }
1803 
1804 /*
1805  * Helper to safely iterate over all caps associated with a session, with
1806  * special care taken to handle a racing __ceph_remove_cap().
1807  *
1808  * Caller must hold session s_mutex.
1809  */
1810 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1811 			      int (*cb)(struct inode *, int mds, void *),
1812 			      void *arg)
1813 {
1814 	struct ceph_client *cl = session->s_mdsc->fsc->client;
1815 	struct list_head *p;
1816 	struct ceph_cap *cap;
1817 	struct inode *inode, *last_inode = NULL;
1818 	struct ceph_cap *old_cap = NULL;
1819 	int ret;
1820 
1821 	doutc(cl, "%p mds%d\n", session, session->s_mds);
1822 	spin_lock(&session->s_cap_lock);
1823 	p = session->s_caps.next;
1824 	while (p != &session->s_caps) {
1825 		int mds;
1826 
1827 		cap = list_entry(p, struct ceph_cap, session_caps);
1828 		inode = igrab(&cap->ci->netfs.inode);
1829 		if (!inode) {
1830 			p = p->next;
1831 			continue;
1832 		}
1833 		session->s_cap_iterator = cap;
1834 		mds = cap->mds;
1835 		spin_unlock(&session->s_cap_lock);
1836 
1837 		if (last_inode) {
1838 			iput(last_inode);
1839 			last_inode = NULL;
1840 		}
1841 		if (old_cap) {
1842 			ceph_put_cap(session->s_mdsc, old_cap);
1843 			old_cap = NULL;
1844 		}
1845 
1846 		ret = cb(inode, mds, arg);
1847 		last_inode = inode;
1848 
1849 		spin_lock(&session->s_cap_lock);
1850 		p = p->next;
1851 		if (!cap->ci) {
1852 			doutc(cl, "finishing cap %p removal\n", cap);
1853 			BUG_ON(cap->session != session);
1854 			cap->session = NULL;
1855 			list_del_init(&cap->session_caps);
1856 			session->s_nr_caps--;
1857 			atomic64_dec(&session->s_mdsc->metric.total_caps);
1858 			if (cap->queue_release)
1859 				__ceph_queue_cap_release(session, cap);
1860 			else
1861 				old_cap = cap;  /* put_cap it w/o locks held */
1862 		}
1863 		if (ret < 0)
1864 			goto out;
1865 	}
1866 	ret = 0;
1867 out:
1868 	session->s_cap_iterator = NULL;
1869 	spin_unlock(&session->s_cap_lock);
1870 
1871 	iput(last_inode);
1872 	if (old_cap)
1873 		ceph_put_cap(session->s_mdsc, old_cap);
1874 
1875 	return ret;
1876 }
1877 
1878 static int remove_session_caps_cb(struct inode *inode, int mds, void *arg)
1879 {
1880 	struct ceph_inode_info *ci = ceph_inode(inode);
1881 	struct ceph_client *cl = ceph_inode_to_client(inode);
1882 	bool invalidate = false;
1883 	struct ceph_cap *cap;
1884 	int iputs = 0;
1885 
1886 	spin_lock(&ci->i_ceph_lock);
1887 	cap = __get_cap_for_mds(ci, mds);
1888 	if (cap) {
1889 		doutc(cl, " removing cap %p, ci is %p, inode is %p\n",
1890 		      cap, ci, &ci->netfs.inode);
1891 
1892 		iputs = ceph_purge_inode_cap(inode, cap, &invalidate);
1893 	}
1894 	spin_unlock(&ci->i_ceph_lock);
1895 
1896 	if (cap)
1897 		wake_up_all(&ci->i_cap_wq);
1898 	if (invalidate)
1899 		ceph_queue_invalidate(inode);
1900 	while (iputs--)
1901 		iput(inode);
1902 	return 0;
1903 }
1904 
1905 /*
1906  * caller must hold session s_mutex
1907  */
1908 static void remove_session_caps(struct ceph_mds_session *session)
1909 {
1910 	struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1911 	struct super_block *sb = fsc->sb;
1912 	LIST_HEAD(dispose);
1913 
1914 	doutc(fsc->client, "on %p\n", session);
1915 	ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1916 
1917 	wake_up_all(&fsc->mdsc->cap_flushing_wq);
1918 
1919 	spin_lock(&session->s_cap_lock);
1920 	if (session->s_nr_caps > 0) {
1921 		struct inode *inode;
1922 		struct ceph_cap *cap, *prev = NULL;
1923 		struct ceph_vino vino;
1924 		/*
1925 		 * iterate_session_caps() skips inodes that are being
1926 		 * deleted, we need to wait until deletions are complete.
1927 		 * __wait_on_freeing_inode() is designed for the job,
1928 		 * but it is not exported, so use lookup inode function
1929 		 * to access it.
1930 		 */
1931 		while (!list_empty(&session->s_caps)) {
1932 			cap = list_entry(session->s_caps.next,
1933 					 struct ceph_cap, session_caps);
1934 			if (cap == prev)
1935 				break;
1936 			prev = cap;
1937 			vino = cap->ci->i_vino;
1938 			spin_unlock(&session->s_cap_lock);
1939 
1940 			inode = ceph_find_inode(sb, vino);
1941 			iput(inode);
1942 
1943 			spin_lock(&session->s_cap_lock);
1944 		}
1945 	}
1946 
1947 	// drop cap expires and unlock s_cap_lock
1948 	detach_cap_releases(session, &dispose);
1949 
1950 	BUG_ON(session->s_nr_caps > 0);
1951 	BUG_ON(!list_empty(&session->s_cap_flushing));
1952 	spin_unlock(&session->s_cap_lock);
1953 	dispose_cap_releases(session->s_mdsc, &dispose);
1954 }
1955 
1956 enum {
1957 	RECONNECT,
1958 	RENEWCAPS,
1959 	FORCE_RO,
1960 };
1961 
1962 /*
1963  * wake up any threads waiting on this session's caps.  if the cap is
1964  * old (didn't get renewed on the client reconnect), remove it now.
1965  *
1966  * caller must hold s_mutex.
1967  */
1968 static int wake_up_session_cb(struct inode *inode, int mds, void *arg)
1969 {
1970 	struct ceph_inode_info *ci = ceph_inode(inode);
1971 	unsigned long ev = (unsigned long)arg;
1972 
1973 	if (ev == RECONNECT) {
1974 		spin_lock(&ci->i_ceph_lock);
1975 		ci->i_wanted_max_size = 0;
1976 		ci->i_requested_max_size = 0;
1977 		spin_unlock(&ci->i_ceph_lock);
1978 	} else if (ev == RENEWCAPS) {
1979 		struct ceph_cap *cap;
1980 
1981 		spin_lock(&ci->i_ceph_lock);
1982 		cap = __get_cap_for_mds(ci, mds);
1983 		/* mds did not re-issue stale cap */
1984 		if (cap && cap->cap_gen < atomic_read(&cap->session->s_cap_gen))
1985 			cap->issued = cap->implemented = CEPH_CAP_PIN;
1986 		spin_unlock(&ci->i_ceph_lock);
1987 	} else if (ev == FORCE_RO) {
1988 	}
1989 	wake_up_all(&ci->i_cap_wq);
1990 	return 0;
1991 }
1992 
1993 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1994 {
1995 	struct ceph_client *cl = session->s_mdsc->fsc->client;
1996 
1997 	doutc(cl, "session %p mds%d\n", session, session->s_mds);
1998 	ceph_iterate_session_caps(session, wake_up_session_cb,
1999 				  (void *)(unsigned long)ev);
2000 }
2001 
2002 /*
2003  * Send periodic message to MDS renewing all currently held caps.  The
2004  * ack will reset the expiration for all caps from this session.
2005  *
2006  * caller holds s_mutex
2007  */
2008 static int send_renew_caps(struct ceph_mds_client *mdsc,
2009 			   struct ceph_mds_session *session)
2010 {
2011 	struct ceph_client *cl = mdsc->fsc->client;
2012 	struct ceph_msg *msg;
2013 	int state;
2014 
2015 	if (time_after_eq(jiffies, session->s_cap_ttl) &&
2016 	    time_after_eq(session->s_cap_ttl, session->s_renew_requested))
2017 		pr_info_client(cl, "mds%d caps stale\n", session->s_mds);
2018 	session->s_renew_requested = jiffies;
2019 
2020 	/* do not try to renew caps until a recovering mds has reconnected
2021 	 * with its clients. */
2022 	state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
2023 	if (state < CEPH_MDS_STATE_RECONNECT) {
2024 		doutc(cl, "ignoring mds%d (%s)\n", session->s_mds,
2025 		      ceph_mds_state_name(state));
2026 		return 0;
2027 	}
2028 
2029 	doutc(cl, "to mds%d (%s)\n", session->s_mds,
2030 	      ceph_mds_state_name(state));
2031 	msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
2032 				      ++session->s_renew_seq);
2033 	if (!msg)
2034 		return -ENOMEM;
2035 	ceph_con_send(&session->s_con, msg);
2036 	return 0;
2037 }
2038 
2039 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
2040 			     struct ceph_mds_session *session, u64 seq)
2041 {
2042 	struct ceph_client *cl = mdsc->fsc->client;
2043 	struct ceph_msg *msg;
2044 
2045 	doutc(cl, "to mds%d (%s)s seq %lld\n", session->s_mds,
2046 	      ceph_session_state_name(session->s_state), seq);
2047 	msg = ceph_create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
2048 	if (!msg)
2049 		return -ENOMEM;
2050 	ceph_con_send(&session->s_con, msg);
2051 	return 0;
2052 }
2053 
2054 
2055 /*
2056  * Note new cap ttl, and any transition from stale -> not stale (fresh?).
2057  *
2058  * Called under session->s_mutex
2059  */
2060 static void renewed_caps(struct ceph_mds_client *mdsc,
2061 			 struct ceph_mds_session *session, int is_renew)
2062 {
2063 	struct ceph_client *cl = mdsc->fsc->client;
2064 	int was_stale;
2065 	int wake = 0;
2066 
2067 	spin_lock(&session->s_cap_lock);
2068 	was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
2069 
2070 	session->s_cap_ttl = session->s_renew_requested +
2071 		mdsc->mdsmap->m_session_timeout*HZ;
2072 
2073 	if (was_stale) {
2074 		if (time_before(jiffies, session->s_cap_ttl)) {
2075 			pr_info_client(cl, "mds%d caps renewed\n",
2076 				       session->s_mds);
2077 			wake = 1;
2078 		} else {
2079 			pr_info_client(cl, "mds%d caps still stale\n",
2080 				       session->s_mds);
2081 		}
2082 	}
2083 	doutc(cl, "mds%d ttl now %lu, was %s, now %s\n", session->s_mds,
2084 	      session->s_cap_ttl, was_stale ? "stale" : "fresh",
2085 	      time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
2086 	spin_unlock(&session->s_cap_lock);
2087 
2088 	if (wake)
2089 		wake_up_session_caps(session, RENEWCAPS);
2090 }
2091 
2092 /*
2093  * send a session close request
2094  */
2095 static int request_close_session(struct ceph_mds_session *session)
2096 {
2097 	struct ceph_client *cl = session->s_mdsc->fsc->client;
2098 	struct ceph_msg *msg;
2099 
2100 	doutc(cl, "mds%d state %s seq %lld\n", session->s_mds,
2101 	      ceph_session_state_name(session->s_state), session->s_seq);
2102 	msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_CLOSE,
2103 				      session->s_seq);
2104 	if (!msg)
2105 		return -ENOMEM;
2106 	ceph_con_send(&session->s_con, msg);
2107 	return 1;
2108 }
2109 
2110 /*
2111  * Called with s_mutex held.
2112  */
2113 static int __close_session(struct ceph_mds_client *mdsc,
2114 			 struct ceph_mds_session *session)
2115 {
2116 	if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
2117 		return 0;
2118 	session->s_state = CEPH_MDS_SESSION_CLOSING;
2119 	return request_close_session(session);
2120 }
2121 
2122 static bool drop_negative_children(struct dentry *dentry)
2123 {
2124 	struct dentry *child;
2125 	bool all_negative = true;
2126 
2127 	if (!d_is_dir(dentry))
2128 		goto out;
2129 
2130 	spin_lock(&dentry->d_lock);
2131 	list_for_each_entry(child, &dentry->d_subdirs, d_child) {
2132 		if (d_really_is_positive(child)) {
2133 			all_negative = false;
2134 			break;
2135 		}
2136 	}
2137 	spin_unlock(&dentry->d_lock);
2138 
2139 	if (all_negative)
2140 		shrink_dcache_parent(dentry);
2141 out:
2142 	return all_negative;
2143 }
2144 
2145 /*
2146  * Trim old(er) caps.
2147  *
2148  * Because we can't cache an inode without one or more caps, we do
2149  * this indirectly: if a cap is unused, we prune its aliases, at which
2150  * point the inode will hopefully get dropped to.
2151  *
2152  * Yes, this is a bit sloppy.  Our only real goal here is to respond to
2153  * memory pressure from the MDS, though, so it needn't be perfect.
2154  */
2155 static int trim_caps_cb(struct inode *inode, int mds, void *arg)
2156 {
2157 	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
2158 	struct ceph_client *cl = mdsc->fsc->client;
2159 	int *remaining = arg;
2160 	struct ceph_inode_info *ci = ceph_inode(inode);
2161 	int used, wanted, oissued, mine;
2162 	struct ceph_cap *cap;
2163 
2164 	if (*remaining <= 0)
2165 		return -1;
2166 
2167 	spin_lock(&ci->i_ceph_lock);
2168 	cap = __get_cap_for_mds(ci, mds);
2169 	if (!cap) {
2170 		spin_unlock(&ci->i_ceph_lock);
2171 		return 0;
2172 	}
2173 	mine = cap->issued | cap->implemented;
2174 	used = __ceph_caps_used(ci);
2175 	wanted = __ceph_caps_file_wanted(ci);
2176 	oissued = __ceph_caps_issued_other(ci, cap);
2177 
2178 	doutc(cl, "%p %llx.%llx cap %p mine %s oissued %s used %s wanted %s\n",
2179 	      inode, ceph_vinop(inode), cap, ceph_cap_string(mine),
2180 	      ceph_cap_string(oissued), ceph_cap_string(used),
2181 	      ceph_cap_string(wanted));
2182 	if (cap == ci->i_auth_cap) {
2183 		if (ci->i_dirty_caps || ci->i_flushing_caps ||
2184 		    !list_empty(&ci->i_cap_snaps))
2185 			goto out;
2186 		if ((used | wanted) & CEPH_CAP_ANY_WR)
2187 			goto out;
2188 		/* Note: it's possible that i_filelock_ref becomes non-zero
2189 		 * after dropping auth caps. It doesn't hurt because reply
2190 		 * of lock mds request will re-add auth caps. */
2191 		if (atomic_read(&ci->i_filelock_ref) > 0)
2192 			goto out;
2193 	}
2194 	/* The inode has cached pages, but it's no longer used.
2195 	 * we can safely drop it */
2196 	if (S_ISREG(inode->i_mode) &&
2197 	    wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
2198 	    !(oissued & CEPH_CAP_FILE_CACHE)) {
2199 	  used = 0;
2200 	  oissued = 0;
2201 	}
2202 	if ((used | wanted) & ~oissued & mine)
2203 		goto out;   /* we need these caps */
2204 
2205 	if (oissued) {
2206 		/* we aren't the only cap.. just remove us */
2207 		ceph_remove_cap(mdsc, cap, true);
2208 		(*remaining)--;
2209 	} else {
2210 		struct dentry *dentry;
2211 		/* try dropping referring dentries */
2212 		spin_unlock(&ci->i_ceph_lock);
2213 		dentry = d_find_any_alias(inode);
2214 		if (dentry && drop_negative_children(dentry)) {
2215 			int count;
2216 			dput(dentry);
2217 			d_prune_aliases(inode);
2218 			count = atomic_read(&inode->i_count);
2219 			if (count == 1)
2220 				(*remaining)--;
2221 			doutc(cl, "%p %llx.%llx cap %p pruned, count now %d\n",
2222 			      inode, ceph_vinop(inode), cap, count);
2223 		} else {
2224 			dput(dentry);
2225 		}
2226 		return 0;
2227 	}
2228 
2229 out:
2230 	spin_unlock(&ci->i_ceph_lock);
2231 	return 0;
2232 }
2233 
2234 /*
2235  * Trim session cap count down to some max number.
2236  */
2237 int ceph_trim_caps(struct ceph_mds_client *mdsc,
2238 		   struct ceph_mds_session *session,
2239 		   int max_caps)
2240 {
2241 	struct ceph_client *cl = mdsc->fsc->client;
2242 	int trim_caps = session->s_nr_caps - max_caps;
2243 
2244 	doutc(cl, "mds%d start: %d / %d, trim %d\n", session->s_mds,
2245 	      session->s_nr_caps, max_caps, trim_caps);
2246 	if (trim_caps > 0) {
2247 		int remaining = trim_caps;
2248 
2249 		ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
2250 		doutc(cl, "mds%d done: %d / %d, trimmed %d\n",
2251 		      session->s_mds, session->s_nr_caps, max_caps,
2252 		      trim_caps - remaining);
2253 	}
2254 
2255 	ceph_flush_cap_releases(mdsc, session);
2256 	return 0;
2257 }
2258 
2259 static int check_caps_flush(struct ceph_mds_client *mdsc,
2260 			    u64 want_flush_tid)
2261 {
2262 	struct ceph_client *cl = mdsc->fsc->client;
2263 	int ret = 1;
2264 
2265 	spin_lock(&mdsc->cap_dirty_lock);
2266 	if (!list_empty(&mdsc->cap_flush_list)) {
2267 		struct ceph_cap_flush *cf =
2268 			list_first_entry(&mdsc->cap_flush_list,
2269 					 struct ceph_cap_flush, g_list);
2270 		if (cf->tid <= want_flush_tid) {
2271 			doutc(cl, "still flushing tid %llu <= %llu\n",
2272 			      cf->tid, want_flush_tid);
2273 			ret = 0;
2274 		}
2275 	}
2276 	spin_unlock(&mdsc->cap_dirty_lock);
2277 	return ret;
2278 }
2279 
2280 /*
2281  * flush all dirty inode data to disk.
2282  *
2283  * returns true if we've flushed through want_flush_tid
2284  */
2285 static void wait_caps_flush(struct ceph_mds_client *mdsc,
2286 			    u64 want_flush_tid)
2287 {
2288 	struct ceph_client *cl = mdsc->fsc->client;
2289 
2290 	doutc(cl, "want %llu\n", want_flush_tid);
2291 
2292 	wait_event(mdsc->cap_flushing_wq,
2293 		   check_caps_flush(mdsc, want_flush_tid));
2294 
2295 	doutc(cl, "ok, flushed thru %llu\n", want_flush_tid);
2296 }
2297 
2298 /*
2299  * called under s_mutex
2300  */
2301 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
2302 				   struct ceph_mds_session *session)
2303 {
2304 	struct ceph_client *cl = mdsc->fsc->client;
2305 	struct ceph_msg *msg = NULL;
2306 	struct ceph_mds_cap_release *head;
2307 	struct ceph_mds_cap_item *item;
2308 	struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
2309 	struct ceph_cap *cap;
2310 	LIST_HEAD(tmp_list);
2311 	int num_cap_releases;
2312 	__le32	barrier, *cap_barrier;
2313 
2314 	down_read(&osdc->lock);
2315 	barrier = cpu_to_le32(osdc->epoch_barrier);
2316 	up_read(&osdc->lock);
2317 
2318 	spin_lock(&session->s_cap_lock);
2319 again:
2320 	list_splice_init(&session->s_cap_releases, &tmp_list);
2321 	num_cap_releases = session->s_num_cap_releases;
2322 	session->s_num_cap_releases = 0;
2323 	spin_unlock(&session->s_cap_lock);
2324 
2325 	while (!list_empty(&tmp_list)) {
2326 		if (!msg) {
2327 			msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
2328 					PAGE_SIZE, GFP_NOFS, false);
2329 			if (!msg)
2330 				goto out_err;
2331 			head = msg->front.iov_base;
2332 			head->num = cpu_to_le32(0);
2333 			msg->front.iov_len = sizeof(*head);
2334 
2335 			msg->hdr.version = cpu_to_le16(2);
2336 			msg->hdr.compat_version = cpu_to_le16(1);
2337 		}
2338 
2339 		cap = list_first_entry(&tmp_list, struct ceph_cap,
2340 					session_caps);
2341 		list_del(&cap->session_caps);
2342 		num_cap_releases--;
2343 
2344 		head = msg->front.iov_base;
2345 		put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2346 				   &head->num);
2347 		item = msg->front.iov_base + msg->front.iov_len;
2348 		item->ino = cpu_to_le64(cap->cap_ino);
2349 		item->cap_id = cpu_to_le64(cap->cap_id);
2350 		item->migrate_seq = cpu_to_le32(cap->mseq);
2351 		item->seq = cpu_to_le32(cap->issue_seq);
2352 		msg->front.iov_len += sizeof(*item);
2353 
2354 		ceph_put_cap(mdsc, cap);
2355 
2356 		if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
2357 			// Append cap_barrier field
2358 			cap_barrier = msg->front.iov_base + msg->front.iov_len;
2359 			*cap_barrier = barrier;
2360 			msg->front.iov_len += sizeof(*cap_barrier);
2361 
2362 			msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2363 			doutc(cl, "mds%d %p\n", session->s_mds, msg);
2364 			ceph_con_send(&session->s_con, msg);
2365 			msg = NULL;
2366 		}
2367 	}
2368 
2369 	BUG_ON(num_cap_releases != 0);
2370 
2371 	spin_lock(&session->s_cap_lock);
2372 	if (!list_empty(&session->s_cap_releases))
2373 		goto again;
2374 	spin_unlock(&session->s_cap_lock);
2375 
2376 	if (msg) {
2377 		// Append cap_barrier field
2378 		cap_barrier = msg->front.iov_base + msg->front.iov_len;
2379 		*cap_barrier = barrier;
2380 		msg->front.iov_len += sizeof(*cap_barrier);
2381 
2382 		msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2383 		doutc(cl, "mds%d %p\n", session->s_mds, msg);
2384 		ceph_con_send(&session->s_con, msg);
2385 	}
2386 	return;
2387 out_err:
2388 	pr_err_client(cl, "mds%d, failed to allocate message\n",
2389 		      session->s_mds);
2390 	spin_lock(&session->s_cap_lock);
2391 	list_splice(&tmp_list, &session->s_cap_releases);
2392 	session->s_num_cap_releases += num_cap_releases;
2393 	spin_unlock(&session->s_cap_lock);
2394 }
2395 
2396 static void ceph_cap_release_work(struct work_struct *work)
2397 {
2398 	struct ceph_mds_session *session =
2399 		container_of(work, struct ceph_mds_session, s_cap_release_work);
2400 
2401 	mutex_lock(&session->s_mutex);
2402 	if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2403 	    session->s_state == CEPH_MDS_SESSION_HUNG)
2404 		ceph_send_cap_releases(session->s_mdsc, session);
2405 	mutex_unlock(&session->s_mutex);
2406 	ceph_put_mds_session(session);
2407 }
2408 
2409 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2410 		             struct ceph_mds_session *session)
2411 {
2412 	struct ceph_client *cl = mdsc->fsc->client;
2413 	if (mdsc->stopping)
2414 		return;
2415 
2416 	ceph_get_mds_session(session);
2417 	if (queue_work(mdsc->fsc->cap_wq,
2418 		       &session->s_cap_release_work)) {
2419 		doutc(cl, "cap release work queued\n");
2420 	} else {
2421 		ceph_put_mds_session(session);
2422 		doutc(cl, "failed to queue cap release work\n");
2423 	}
2424 }
2425 
2426 /*
2427  * caller holds session->s_cap_lock
2428  */
2429 void __ceph_queue_cap_release(struct ceph_mds_session *session,
2430 			      struct ceph_cap *cap)
2431 {
2432 	list_add_tail(&cap->session_caps, &session->s_cap_releases);
2433 	session->s_num_cap_releases++;
2434 
2435 	if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2436 		ceph_flush_cap_releases(session->s_mdsc, session);
2437 }
2438 
2439 static void ceph_cap_reclaim_work(struct work_struct *work)
2440 {
2441 	struct ceph_mds_client *mdsc =
2442 		container_of(work, struct ceph_mds_client, cap_reclaim_work);
2443 	int ret = ceph_trim_dentries(mdsc);
2444 	if (ret == -EAGAIN)
2445 		ceph_queue_cap_reclaim_work(mdsc);
2446 }
2447 
2448 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2449 {
2450 	struct ceph_client *cl = mdsc->fsc->client;
2451 	if (mdsc->stopping)
2452 		return;
2453 
2454         if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2455                 doutc(cl, "caps reclaim work queued\n");
2456         } else {
2457                 doutc(cl, "failed to queue caps release work\n");
2458         }
2459 }
2460 
2461 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2462 {
2463 	int val;
2464 	if (!nr)
2465 		return;
2466 	val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2467 	if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
2468 		atomic_set(&mdsc->cap_reclaim_pending, 0);
2469 		ceph_queue_cap_reclaim_work(mdsc);
2470 	}
2471 }
2472 
2473 /*
2474  * requests
2475  */
2476 
2477 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2478 				    struct inode *dir)
2479 {
2480 	struct ceph_inode_info *ci = ceph_inode(dir);
2481 	struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2482 	struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2483 	size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2484 	unsigned int num_entries;
2485 	int order;
2486 
2487 	spin_lock(&ci->i_ceph_lock);
2488 	num_entries = ci->i_files + ci->i_subdirs;
2489 	spin_unlock(&ci->i_ceph_lock);
2490 	num_entries = max(num_entries, 1U);
2491 	num_entries = min(num_entries, opt->max_readdir);
2492 
2493 	order = get_order(size * num_entries);
2494 	while (order >= 0) {
2495 		rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2496 							     __GFP_NOWARN |
2497 							     __GFP_ZERO,
2498 							     order);
2499 		if (rinfo->dir_entries)
2500 			break;
2501 		order--;
2502 	}
2503 	if (!rinfo->dir_entries)
2504 		return -ENOMEM;
2505 
2506 	num_entries = (PAGE_SIZE << order) / size;
2507 	num_entries = min(num_entries, opt->max_readdir);
2508 
2509 	rinfo->dir_buf_size = PAGE_SIZE << order;
2510 	req->r_num_caps = num_entries + 1;
2511 	req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2512 	req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2513 	return 0;
2514 }
2515 
2516 /*
2517  * Create an mds request.
2518  */
2519 struct ceph_mds_request *
2520 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2521 {
2522 	struct ceph_mds_request *req;
2523 
2524 	req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2525 	if (!req)
2526 		return ERR_PTR(-ENOMEM);
2527 
2528 	mutex_init(&req->r_fill_mutex);
2529 	req->r_mdsc = mdsc;
2530 	req->r_started = jiffies;
2531 	req->r_start_latency = ktime_get();
2532 	req->r_resend_mds = -1;
2533 	INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2534 	INIT_LIST_HEAD(&req->r_unsafe_target_item);
2535 	req->r_fmode = -1;
2536 	req->r_feature_needed = -1;
2537 	kref_init(&req->r_kref);
2538 	RB_CLEAR_NODE(&req->r_node);
2539 	INIT_LIST_HEAD(&req->r_wait);
2540 	init_completion(&req->r_completion);
2541 	init_completion(&req->r_safe_completion);
2542 	INIT_LIST_HEAD(&req->r_unsafe_item);
2543 
2544 	ktime_get_coarse_real_ts64(&req->r_stamp);
2545 
2546 	req->r_op = op;
2547 	req->r_direct_mode = mode;
2548 	return req;
2549 }
2550 
2551 /*
2552  * return oldest (lowest) request, tid in request tree, 0 if none.
2553  *
2554  * called under mdsc->mutex.
2555  */
2556 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2557 {
2558 	if (RB_EMPTY_ROOT(&mdsc->request_tree))
2559 		return NULL;
2560 	return rb_entry(rb_first(&mdsc->request_tree),
2561 			struct ceph_mds_request, r_node);
2562 }
2563 
2564 static inline  u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2565 {
2566 	return mdsc->oldest_tid;
2567 }
2568 
2569 #if IS_ENABLED(CONFIG_FS_ENCRYPTION)
2570 static u8 *get_fscrypt_altname(const struct ceph_mds_request *req, u32 *plen)
2571 {
2572 	struct inode *dir = req->r_parent;
2573 	struct dentry *dentry = req->r_dentry;
2574 	u8 *cryptbuf = NULL;
2575 	u32 len = 0;
2576 	int ret = 0;
2577 
2578 	/* only encode if we have parent and dentry */
2579 	if (!dir || !dentry)
2580 		goto success;
2581 
2582 	/* No-op unless this is encrypted */
2583 	if (!IS_ENCRYPTED(dir))
2584 		goto success;
2585 
2586 	ret = ceph_fscrypt_prepare_readdir(dir);
2587 	if (ret < 0)
2588 		return ERR_PTR(ret);
2589 
2590 	/* No key? Just ignore it. */
2591 	if (!fscrypt_has_encryption_key(dir))
2592 		goto success;
2593 
2594 	if (!fscrypt_fname_encrypted_size(dir, dentry->d_name.len, NAME_MAX,
2595 					  &len)) {
2596 		WARN_ON_ONCE(1);
2597 		return ERR_PTR(-ENAMETOOLONG);
2598 	}
2599 
2600 	/* No need to append altname if name is short enough */
2601 	if (len <= CEPH_NOHASH_NAME_MAX) {
2602 		len = 0;
2603 		goto success;
2604 	}
2605 
2606 	cryptbuf = kmalloc(len, GFP_KERNEL);
2607 	if (!cryptbuf)
2608 		return ERR_PTR(-ENOMEM);
2609 
2610 	ret = fscrypt_fname_encrypt(dir, &dentry->d_name, cryptbuf, len);
2611 	if (ret) {
2612 		kfree(cryptbuf);
2613 		return ERR_PTR(ret);
2614 	}
2615 success:
2616 	*plen = len;
2617 	return cryptbuf;
2618 }
2619 #else
2620 static u8 *get_fscrypt_altname(const struct ceph_mds_request *req, u32 *plen)
2621 {
2622 	*plen = 0;
2623 	return NULL;
2624 }
2625 #endif
2626 
2627 /**
2628  * ceph_mdsc_build_path - build a path string to a given dentry
2629  * @mdsc: mds client
2630  * @dentry: dentry to which path should be built
2631  * @plen: returned length of string
2632  * @pbase: returned base inode number
2633  * @for_wire: is this path going to be sent to the MDS?
2634  *
2635  * Build a string that represents the path to the dentry. This is mostly called
2636  * for two different purposes:
2637  *
2638  * 1) we need to build a path string to send to the MDS (for_wire == true)
2639  * 2) we need a path string for local presentation (e.g. debugfs)
2640  *    (for_wire == false)
2641  *
2642  * The path is built in reverse, starting with the dentry. Walk back up toward
2643  * the root, building the path until the first non-snapped inode is reached
2644  * (for_wire) or the root inode is reached (!for_wire).
2645  *
2646  * Encode hidden .snap dirs as a double /, i.e.
2647  *   foo/.snap/bar -> foo//bar
2648  */
2649 char *ceph_mdsc_build_path(struct ceph_mds_client *mdsc, struct dentry *dentry,
2650 			   int *plen, u64 *pbase, int for_wire)
2651 {
2652 	struct ceph_client *cl = mdsc->fsc->client;
2653 	struct dentry *cur;
2654 	struct inode *inode;
2655 	char *path;
2656 	int pos;
2657 	unsigned seq;
2658 	u64 base;
2659 
2660 	if (!dentry)
2661 		return ERR_PTR(-EINVAL);
2662 
2663 	path = __getname();
2664 	if (!path)
2665 		return ERR_PTR(-ENOMEM);
2666 retry:
2667 	pos = PATH_MAX - 1;
2668 	path[pos] = '\0';
2669 
2670 	seq = read_seqbegin(&rename_lock);
2671 	cur = dget(dentry);
2672 	for (;;) {
2673 		struct dentry *parent;
2674 
2675 		spin_lock(&cur->d_lock);
2676 		inode = d_inode(cur);
2677 		if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2678 			doutc(cl, "path+%d: %p SNAPDIR\n", pos, cur);
2679 			spin_unlock(&cur->d_lock);
2680 			parent = dget_parent(cur);
2681 		} else if (for_wire && inode && dentry != cur &&
2682 			   ceph_snap(inode) == CEPH_NOSNAP) {
2683 			spin_unlock(&cur->d_lock);
2684 			pos++; /* get rid of any prepended '/' */
2685 			break;
2686 		} else if (!for_wire || !IS_ENCRYPTED(d_inode(cur->d_parent))) {
2687 			pos -= cur->d_name.len;
2688 			if (pos < 0) {
2689 				spin_unlock(&cur->d_lock);
2690 				break;
2691 			}
2692 			memcpy(path + pos, cur->d_name.name, cur->d_name.len);
2693 			spin_unlock(&cur->d_lock);
2694 			parent = dget_parent(cur);
2695 		} else {
2696 			int len, ret;
2697 			char buf[NAME_MAX];
2698 
2699 			/*
2700 			 * Proactively copy name into buf, in case we need to
2701 			 * present it as-is.
2702 			 */
2703 			memcpy(buf, cur->d_name.name, cur->d_name.len);
2704 			len = cur->d_name.len;
2705 			spin_unlock(&cur->d_lock);
2706 			parent = dget_parent(cur);
2707 
2708 			ret = ceph_fscrypt_prepare_readdir(d_inode(parent));
2709 			if (ret < 0) {
2710 				dput(parent);
2711 				dput(cur);
2712 				return ERR_PTR(ret);
2713 			}
2714 
2715 			if (fscrypt_has_encryption_key(d_inode(parent))) {
2716 				len = ceph_encode_encrypted_fname(d_inode(parent),
2717 								  cur, buf);
2718 				if (len < 0) {
2719 					dput(parent);
2720 					dput(cur);
2721 					return ERR_PTR(len);
2722 				}
2723 			}
2724 			pos -= len;
2725 			if (pos < 0) {
2726 				dput(parent);
2727 				break;
2728 			}
2729 			memcpy(path + pos, buf, len);
2730 		}
2731 		dput(cur);
2732 		cur = parent;
2733 
2734 		/* Are we at the root? */
2735 		if (IS_ROOT(cur))
2736 			break;
2737 
2738 		/* Are we out of buffer? */
2739 		if (--pos < 0)
2740 			break;
2741 
2742 		path[pos] = '/';
2743 	}
2744 	inode = d_inode(cur);
2745 	base = inode ? ceph_ino(inode) : 0;
2746 	dput(cur);
2747 
2748 	if (read_seqretry(&rename_lock, seq))
2749 		goto retry;
2750 
2751 	if (pos < 0) {
2752 		/*
2753 		 * A rename didn't occur, but somehow we didn't end up where
2754 		 * we thought we would. Throw a warning and try again.
2755 		 */
2756 		pr_warn_client(cl, "did not end path lookup where expected (pos = %d)\n",
2757 			       pos);
2758 		goto retry;
2759 	}
2760 
2761 	*pbase = base;
2762 	*plen = PATH_MAX - 1 - pos;
2763 	doutc(cl, "on %p %d built %llx '%.*s'\n", dentry, d_count(dentry),
2764 	      base, *plen, path + pos);
2765 	return path + pos;
2766 }
2767 
2768 static int build_dentry_path(struct ceph_mds_client *mdsc, struct dentry *dentry,
2769 			     struct inode *dir, const char **ppath, int *ppathlen,
2770 			     u64 *pino, bool *pfreepath, bool parent_locked)
2771 {
2772 	char *path;
2773 
2774 	rcu_read_lock();
2775 	if (!dir)
2776 		dir = d_inode_rcu(dentry->d_parent);
2777 	if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP &&
2778 	    !IS_ENCRYPTED(dir)) {
2779 		*pino = ceph_ino(dir);
2780 		rcu_read_unlock();
2781 		*ppath = dentry->d_name.name;
2782 		*ppathlen = dentry->d_name.len;
2783 		return 0;
2784 	}
2785 	rcu_read_unlock();
2786 	path = ceph_mdsc_build_path(mdsc, dentry, ppathlen, pino, 1);
2787 	if (IS_ERR(path))
2788 		return PTR_ERR(path);
2789 	*ppath = path;
2790 	*pfreepath = true;
2791 	return 0;
2792 }
2793 
2794 static int build_inode_path(struct inode *inode,
2795 			    const char **ppath, int *ppathlen, u64 *pino,
2796 			    bool *pfreepath)
2797 {
2798 	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
2799 	struct dentry *dentry;
2800 	char *path;
2801 
2802 	if (ceph_snap(inode) == CEPH_NOSNAP) {
2803 		*pino = ceph_ino(inode);
2804 		*ppathlen = 0;
2805 		return 0;
2806 	}
2807 	dentry = d_find_alias(inode);
2808 	path = ceph_mdsc_build_path(mdsc, dentry, ppathlen, pino, 1);
2809 	dput(dentry);
2810 	if (IS_ERR(path))
2811 		return PTR_ERR(path);
2812 	*ppath = path;
2813 	*pfreepath = true;
2814 	return 0;
2815 }
2816 
2817 /*
2818  * request arguments may be specified via an inode *, a dentry *, or
2819  * an explicit ino+path.
2820  */
2821 static int set_request_path_attr(struct ceph_mds_client *mdsc, struct inode *rinode,
2822 				 struct dentry *rdentry, struct inode *rdiri,
2823 				 const char *rpath, u64 rino, const char **ppath,
2824 				 int *pathlen, u64 *ino, bool *freepath,
2825 				 bool parent_locked)
2826 {
2827 	struct ceph_client *cl = mdsc->fsc->client;
2828 	int r = 0;
2829 
2830 	if (rinode) {
2831 		r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2832 		doutc(cl, " inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2833 		      ceph_snap(rinode));
2834 	} else if (rdentry) {
2835 		r = build_dentry_path(mdsc, rdentry, rdiri, ppath, pathlen, ino,
2836 					freepath, parent_locked);
2837 		doutc(cl, " dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen, *ppath);
2838 	} else if (rpath || rino) {
2839 		*ino = rino;
2840 		*ppath = rpath;
2841 		*pathlen = rpath ? strlen(rpath) : 0;
2842 		doutc(cl, " path %.*s\n", *pathlen, rpath);
2843 	}
2844 
2845 	return r;
2846 }
2847 
2848 static void encode_mclientrequest_tail(void **p,
2849 				       const struct ceph_mds_request *req)
2850 {
2851 	struct ceph_timespec ts;
2852 	int i;
2853 
2854 	ceph_encode_timespec64(&ts, &req->r_stamp);
2855 	ceph_encode_copy(p, &ts, sizeof(ts));
2856 
2857 	/* v4: gid_list */
2858 	ceph_encode_32(p, req->r_cred->group_info->ngroups);
2859 	for (i = 0; i < req->r_cred->group_info->ngroups; i++)
2860 		ceph_encode_64(p, from_kgid(&init_user_ns,
2861 					    req->r_cred->group_info->gid[i]));
2862 
2863 	/* v5: altname */
2864 	ceph_encode_32(p, req->r_altname_len);
2865 	ceph_encode_copy(p, req->r_altname, req->r_altname_len);
2866 
2867 	/* v6: fscrypt_auth and fscrypt_file */
2868 	if (req->r_fscrypt_auth) {
2869 		u32 authlen = ceph_fscrypt_auth_len(req->r_fscrypt_auth);
2870 
2871 		ceph_encode_32(p, authlen);
2872 		ceph_encode_copy(p, req->r_fscrypt_auth, authlen);
2873 	} else {
2874 		ceph_encode_32(p, 0);
2875 	}
2876 	if (test_bit(CEPH_MDS_R_FSCRYPT_FILE, &req->r_req_flags)) {
2877 		ceph_encode_32(p, sizeof(__le64));
2878 		ceph_encode_64(p, req->r_fscrypt_file);
2879 	} else {
2880 		ceph_encode_32(p, 0);
2881 	}
2882 }
2883 
2884 static inline u16 mds_supported_head_version(struct ceph_mds_session *session)
2885 {
2886 	if (!test_bit(CEPHFS_FEATURE_32BITS_RETRY_FWD, &session->s_features))
2887 		return 1;
2888 
2889 	if (!test_bit(CEPHFS_FEATURE_HAS_OWNER_UIDGID, &session->s_features))
2890 		return 2;
2891 
2892 	return CEPH_MDS_REQUEST_HEAD_VERSION;
2893 }
2894 
2895 static struct ceph_mds_request_head_legacy *
2896 find_legacy_request_head(void *p, u64 features)
2897 {
2898 	bool legacy = !(features & CEPH_FEATURE_FS_BTIME);
2899 	struct ceph_mds_request_head_old *ohead;
2900 
2901 	if (legacy)
2902 		return (struct ceph_mds_request_head_legacy *)p;
2903 	ohead = (struct ceph_mds_request_head_old *)p;
2904 	return (struct ceph_mds_request_head_legacy *)&ohead->oldest_client_tid;
2905 }
2906 
2907 /*
2908  * called under mdsc->mutex
2909  */
2910 static struct ceph_msg *create_request_message(struct ceph_mds_session *session,
2911 					       struct ceph_mds_request *req,
2912 					       bool drop_cap_releases)
2913 {
2914 	int mds = session->s_mds;
2915 	struct ceph_mds_client *mdsc = session->s_mdsc;
2916 	struct ceph_client *cl = mdsc->fsc->client;
2917 	struct ceph_msg *msg;
2918 	struct ceph_mds_request_head_legacy *lhead;
2919 	const char *path1 = NULL;
2920 	const char *path2 = NULL;
2921 	u64 ino1 = 0, ino2 = 0;
2922 	int pathlen1 = 0, pathlen2 = 0;
2923 	bool freepath1 = false, freepath2 = false;
2924 	struct dentry *old_dentry = NULL;
2925 	int len;
2926 	u16 releases;
2927 	void *p, *end;
2928 	int ret;
2929 	bool legacy = !(session->s_con.peer_features & CEPH_FEATURE_FS_BTIME);
2930 	u16 request_head_version = mds_supported_head_version(session);
2931 	kuid_t caller_fsuid = req->r_cred->fsuid;
2932 	kgid_t caller_fsgid = req->r_cred->fsgid;
2933 
2934 	ret = set_request_path_attr(mdsc, req->r_inode, req->r_dentry,
2935 			      req->r_parent, req->r_path1, req->r_ino1.ino,
2936 			      &path1, &pathlen1, &ino1, &freepath1,
2937 			      test_bit(CEPH_MDS_R_PARENT_LOCKED,
2938 					&req->r_req_flags));
2939 	if (ret < 0) {
2940 		msg = ERR_PTR(ret);
2941 		goto out;
2942 	}
2943 
2944 	/* If r_old_dentry is set, then assume that its parent is locked */
2945 	if (req->r_old_dentry &&
2946 	    !(req->r_old_dentry->d_flags & DCACHE_DISCONNECTED))
2947 		old_dentry = req->r_old_dentry;
2948 	ret = set_request_path_attr(mdsc, NULL, old_dentry,
2949 			      req->r_old_dentry_dir,
2950 			      req->r_path2, req->r_ino2.ino,
2951 			      &path2, &pathlen2, &ino2, &freepath2, true);
2952 	if (ret < 0) {
2953 		msg = ERR_PTR(ret);
2954 		goto out_free1;
2955 	}
2956 
2957 	req->r_altname = get_fscrypt_altname(req, &req->r_altname_len);
2958 	if (IS_ERR(req->r_altname)) {
2959 		msg = ERR_CAST(req->r_altname);
2960 		req->r_altname = NULL;
2961 		goto out_free2;
2962 	}
2963 
2964 	/*
2965 	 * For old cephs without supporting the 32bit retry/fwd feature
2966 	 * it will copy the raw memories directly when decoding the
2967 	 * requests. While new cephs will decode the head depending the
2968 	 * version member, so we need to make sure it will be compatible
2969 	 * with them both.
2970 	 */
2971 	if (legacy)
2972 		len = sizeof(struct ceph_mds_request_head_legacy);
2973 	else if (request_head_version == 1)
2974 		len = sizeof(struct ceph_mds_request_head_old);
2975 	else if (request_head_version == 2)
2976 		len = offsetofend(struct ceph_mds_request_head, ext_num_fwd);
2977 	else
2978 		len = sizeof(struct ceph_mds_request_head);
2979 
2980 	/* filepaths */
2981 	len += 2 * (1 + sizeof(u32) + sizeof(u64));
2982 	len += pathlen1 + pathlen2;
2983 
2984 	/* cap releases */
2985 	len += sizeof(struct ceph_mds_request_release) *
2986 		(!!req->r_inode_drop + !!req->r_dentry_drop +
2987 		 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2988 
2989 	if (req->r_dentry_drop)
2990 		len += pathlen1;
2991 	if (req->r_old_dentry_drop)
2992 		len += pathlen2;
2993 
2994 	/* MClientRequest tail */
2995 
2996 	/* req->r_stamp */
2997 	len += sizeof(struct ceph_timespec);
2998 
2999 	/* gid list */
3000 	len += sizeof(u32) + (sizeof(u64) * req->r_cred->group_info->ngroups);
3001 
3002 	/* alternate name */
3003 	len += sizeof(u32) + req->r_altname_len;
3004 
3005 	/* fscrypt_auth */
3006 	len += sizeof(u32); // fscrypt_auth
3007 	if (req->r_fscrypt_auth)
3008 		len += ceph_fscrypt_auth_len(req->r_fscrypt_auth);
3009 
3010 	/* fscrypt_file */
3011 	len += sizeof(u32);
3012 	if (test_bit(CEPH_MDS_R_FSCRYPT_FILE, &req->r_req_flags))
3013 		len += sizeof(__le64);
3014 
3015 	msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
3016 	if (!msg) {
3017 		msg = ERR_PTR(-ENOMEM);
3018 		goto out_free2;
3019 	}
3020 
3021 	msg->hdr.tid = cpu_to_le64(req->r_tid);
3022 
3023 	lhead = find_legacy_request_head(msg->front.iov_base,
3024 					 session->s_con.peer_features);
3025 
3026 	if ((req->r_mnt_idmap != &nop_mnt_idmap) &&
3027 	    !test_bit(CEPHFS_FEATURE_HAS_OWNER_UIDGID, &session->s_features)) {
3028 		WARN_ON_ONCE(!IS_CEPH_MDS_OP_NEWINODE(req->r_op));
3029 
3030 		if (enable_unsafe_idmap) {
3031 			pr_warn_once_client(cl,
3032 				"idmapped mount is used and CEPHFS_FEATURE_HAS_OWNER_UIDGID"
3033 				" is not supported by MDS. UID/GID-based restrictions may"
3034 				" not work properly.\n");
3035 
3036 			caller_fsuid = from_vfsuid(req->r_mnt_idmap, &init_user_ns,
3037 						   VFSUIDT_INIT(req->r_cred->fsuid));
3038 			caller_fsgid = from_vfsgid(req->r_mnt_idmap, &init_user_ns,
3039 						   VFSGIDT_INIT(req->r_cred->fsgid));
3040 		} else {
3041 			pr_err_ratelimited_client(cl,
3042 				"idmapped mount is used and CEPHFS_FEATURE_HAS_OWNER_UIDGID"
3043 				" is not supported by MDS. Fail request with -EIO.\n");
3044 
3045 			ret = -EIO;
3046 			goto out_err;
3047 		}
3048 	}
3049 
3050 	/*
3051 	 * The ceph_mds_request_head_legacy didn't contain a version field, and
3052 	 * one was added when we moved the message version from 3->4.
3053 	 */
3054 	if (legacy) {
3055 		msg->hdr.version = cpu_to_le16(3);
3056 		p = msg->front.iov_base + sizeof(*lhead);
3057 	} else if (request_head_version == 1) {
3058 		struct ceph_mds_request_head_old *ohead = msg->front.iov_base;
3059 
3060 		msg->hdr.version = cpu_to_le16(4);
3061 		ohead->version = cpu_to_le16(1);
3062 		p = msg->front.iov_base + sizeof(*ohead);
3063 	} else if (request_head_version == 2) {
3064 		struct ceph_mds_request_head *nhead = msg->front.iov_base;
3065 
3066 		msg->hdr.version = cpu_to_le16(6);
3067 		nhead->version = cpu_to_le16(2);
3068 
3069 		p = msg->front.iov_base + offsetofend(struct ceph_mds_request_head, ext_num_fwd);
3070 	} else {
3071 		struct ceph_mds_request_head *nhead = msg->front.iov_base;
3072 		kuid_t owner_fsuid;
3073 		kgid_t owner_fsgid;
3074 
3075 		msg->hdr.version = cpu_to_le16(6);
3076 		nhead->version = cpu_to_le16(CEPH_MDS_REQUEST_HEAD_VERSION);
3077 		nhead->struct_len = cpu_to_le32(sizeof(struct ceph_mds_request_head));
3078 
3079 		if (IS_CEPH_MDS_OP_NEWINODE(req->r_op)) {
3080 			owner_fsuid = from_vfsuid(req->r_mnt_idmap, &init_user_ns,
3081 						VFSUIDT_INIT(req->r_cred->fsuid));
3082 			owner_fsgid = from_vfsgid(req->r_mnt_idmap, &init_user_ns,
3083 						VFSGIDT_INIT(req->r_cred->fsgid));
3084 			nhead->owner_uid = cpu_to_le32(from_kuid(&init_user_ns, owner_fsuid));
3085 			nhead->owner_gid = cpu_to_le32(from_kgid(&init_user_ns, owner_fsgid));
3086 		} else {
3087 			nhead->owner_uid = cpu_to_le32(-1);
3088 			nhead->owner_gid = cpu_to_le32(-1);
3089 		}
3090 
3091 		p = msg->front.iov_base + sizeof(*nhead);
3092 	}
3093 
3094 	end = msg->front.iov_base + msg->front.iov_len;
3095 
3096 	lhead->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
3097 	lhead->op = cpu_to_le32(req->r_op);
3098 	lhead->caller_uid = cpu_to_le32(from_kuid(&init_user_ns,
3099 						  caller_fsuid));
3100 	lhead->caller_gid = cpu_to_le32(from_kgid(&init_user_ns,
3101 						  caller_fsgid));
3102 	lhead->ino = cpu_to_le64(req->r_deleg_ino);
3103 	lhead->args = req->r_args;
3104 
3105 	ceph_encode_filepath(&p, end, ino1, path1);
3106 	ceph_encode_filepath(&p, end, ino2, path2);
3107 
3108 	/* make note of release offset, in case we need to replay */
3109 	req->r_request_release_offset = p - msg->front.iov_base;
3110 
3111 	/* cap releases */
3112 	releases = 0;
3113 	if (req->r_inode_drop)
3114 		releases += ceph_encode_inode_release(&p,
3115 		      req->r_inode ? req->r_inode : d_inode(req->r_dentry),
3116 		      mds, req->r_inode_drop, req->r_inode_unless,
3117 		      req->r_op == CEPH_MDS_OP_READDIR);
3118 	if (req->r_dentry_drop) {
3119 		ret = ceph_encode_dentry_release(&p, req->r_dentry,
3120 				req->r_parent, mds, req->r_dentry_drop,
3121 				req->r_dentry_unless);
3122 		if (ret < 0)
3123 			goto out_err;
3124 		releases += ret;
3125 	}
3126 	if (req->r_old_dentry_drop) {
3127 		ret = ceph_encode_dentry_release(&p, req->r_old_dentry,
3128 				req->r_old_dentry_dir, mds,
3129 				req->r_old_dentry_drop,
3130 				req->r_old_dentry_unless);
3131 		if (ret < 0)
3132 			goto out_err;
3133 		releases += ret;
3134 	}
3135 	if (req->r_old_inode_drop)
3136 		releases += ceph_encode_inode_release(&p,
3137 		      d_inode(req->r_old_dentry),
3138 		      mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
3139 
3140 	if (drop_cap_releases) {
3141 		releases = 0;
3142 		p = msg->front.iov_base + req->r_request_release_offset;
3143 	}
3144 
3145 	lhead->num_releases = cpu_to_le16(releases);
3146 
3147 	encode_mclientrequest_tail(&p, req);
3148 
3149 	if (WARN_ON_ONCE(p > end)) {
3150 		ceph_msg_put(msg);
3151 		msg = ERR_PTR(-ERANGE);
3152 		goto out_free2;
3153 	}
3154 
3155 	msg->front.iov_len = p - msg->front.iov_base;
3156 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
3157 
3158 	if (req->r_pagelist) {
3159 		struct ceph_pagelist *pagelist = req->r_pagelist;
3160 		ceph_msg_data_add_pagelist(msg, pagelist);
3161 		msg->hdr.data_len = cpu_to_le32(pagelist->length);
3162 	} else {
3163 		msg->hdr.data_len = 0;
3164 	}
3165 
3166 	msg->hdr.data_off = cpu_to_le16(0);
3167 
3168 out_free2:
3169 	if (freepath2)
3170 		ceph_mdsc_free_path((char *)path2, pathlen2);
3171 out_free1:
3172 	if (freepath1)
3173 		ceph_mdsc_free_path((char *)path1, pathlen1);
3174 out:
3175 	return msg;
3176 out_err:
3177 	ceph_msg_put(msg);
3178 	msg = ERR_PTR(ret);
3179 	goto out_free2;
3180 }
3181 
3182 /*
3183  * called under mdsc->mutex if error, under no mutex if
3184  * success.
3185  */
3186 static void complete_request(struct ceph_mds_client *mdsc,
3187 			     struct ceph_mds_request *req)
3188 {
3189 	req->r_end_latency = ktime_get();
3190 
3191 	if (req->r_callback)
3192 		req->r_callback(mdsc, req);
3193 	complete_all(&req->r_completion);
3194 }
3195 
3196 /*
3197  * called under mdsc->mutex
3198  */
3199 static int __prepare_send_request(struct ceph_mds_session *session,
3200 				  struct ceph_mds_request *req,
3201 				  bool drop_cap_releases)
3202 {
3203 	int mds = session->s_mds;
3204 	struct ceph_mds_client *mdsc = session->s_mdsc;
3205 	struct ceph_client *cl = mdsc->fsc->client;
3206 	struct ceph_mds_request_head_legacy *lhead;
3207 	struct ceph_mds_request_head *nhead;
3208 	struct ceph_msg *msg;
3209 	int flags = 0, old_max_retry;
3210 	bool old_version = !test_bit(CEPHFS_FEATURE_32BITS_RETRY_FWD,
3211 				     &session->s_features);
3212 
3213 	/*
3214 	 * Avoid inifinite retrying after overflow. The client will
3215 	 * increase the retry count and if the MDS is old version,
3216 	 * so we limit to retry at most 256 times.
3217 	 */
3218 	if (req->r_attempts) {
3219 	       old_max_retry = sizeof_field(struct ceph_mds_request_head_old,
3220 					    num_retry);
3221 	       old_max_retry = 1 << (old_max_retry * BITS_PER_BYTE);
3222 	       if ((old_version && req->r_attempts >= old_max_retry) ||
3223 		   ((uint32_t)req->r_attempts >= U32_MAX)) {
3224 			pr_warn_ratelimited_client(cl, "request tid %llu seq overflow\n",
3225 						   req->r_tid);
3226 			return -EMULTIHOP;
3227 	       }
3228 	}
3229 
3230 	req->r_attempts++;
3231 	if (req->r_inode) {
3232 		struct ceph_cap *cap =
3233 			ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
3234 
3235 		if (cap)
3236 			req->r_sent_on_mseq = cap->mseq;
3237 		else
3238 			req->r_sent_on_mseq = -1;
3239 	}
3240 	doutc(cl, "%p tid %lld %s (attempt %d)\n", req, req->r_tid,
3241 	      ceph_mds_op_name(req->r_op), req->r_attempts);
3242 
3243 	if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3244 		void *p;
3245 
3246 		/*
3247 		 * Replay.  Do not regenerate message (and rebuild
3248 		 * paths, etc.); just use the original message.
3249 		 * Rebuilding paths will break for renames because
3250 		 * d_move mangles the src name.
3251 		 */
3252 		msg = req->r_request;
3253 		lhead = find_legacy_request_head(msg->front.iov_base,
3254 						 session->s_con.peer_features);
3255 
3256 		flags = le32_to_cpu(lhead->flags);
3257 		flags |= CEPH_MDS_FLAG_REPLAY;
3258 		lhead->flags = cpu_to_le32(flags);
3259 
3260 		if (req->r_target_inode)
3261 			lhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
3262 
3263 		lhead->num_retry = req->r_attempts - 1;
3264 		if (!old_version) {
3265 			nhead = (struct ceph_mds_request_head*)msg->front.iov_base;
3266 			nhead->ext_num_retry = cpu_to_le32(req->r_attempts - 1);
3267 		}
3268 
3269 		/* remove cap/dentry releases from message */
3270 		lhead->num_releases = 0;
3271 
3272 		p = msg->front.iov_base + req->r_request_release_offset;
3273 		encode_mclientrequest_tail(&p, req);
3274 
3275 		msg->front.iov_len = p - msg->front.iov_base;
3276 		msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
3277 		return 0;
3278 	}
3279 
3280 	if (req->r_request) {
3281 		ceph_msg_put(req->r_request);
3282 		req->r_request = NULL;
3283 	}
3284 	msg = create_request_message(session, req, drop_cap_releases);
3285 	if (IS_ERR(msg)) {
3286 		req->r_err = PTR_ERR(msg);
3287 		return PTR_ERR(msg);
3288 	}
3289 	req->r_request = msg;
3290 
3291 	lhead = find_legacy_request_head(msg->front.iov_base,
3292 					 session->s_con.peer_features);
3293 	lhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
3294 	if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3295 		flags |= CEPH_MDS_FLAG_REPLAY;
3296 	if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
3297 		flags |= CEPH_MDS_FLAG_ASYNC;
3298 	if (req->r_parent)
3299 		flags |= CEPH_MDS_FLAG_WANT_DENTRY;
3300 	lhead->flags = cpu_to_le32(flags);
3301 	lhead->num_fwd = req->r_num_fwd;
3302 	lhead->num_retry = req->r_attempts - 1;
3303 	if (!old_version) {
3304 		nhead = (struct ceph_mds_request_head*)msg->front.iov_base;
3305 		nhead->ext_num_fwd = cpu_to_le32(req->r_num_fwd);
3306 		nhead->ext_num_retry = cpu_to_le32(req->r_attempts - 1);
3307 	}
3308 
3309 	doutc(cl, " r_parent = %p\n", req->r_parent);
3310 	return 0;
3311 }
3312 
3313 /*
3314  * called under mdsc->mutex
3315  */
3316 static int __send_request(struct ceph_mds_session *session,
3317 			  struct ceph_mds_request *req,
3318 			  bool drop_cap_releases)
3319 {
3320 	int err;
3321 
3322 	err = __prepare_send_request(session, req, drop_cap_releases);
3323 	if (!err) {
3324 		ceph_msg_get(req->r_request);
3325 		ceph_con_send(&session->s_con, req->r_request);
3326 	}
3327 
3328 	return err;
3329 }
3330 
3331 /*
3332  * send request, or put it on the appropriate wait list.
3333  */
3334 static void __do_request(struct ceph_mds_client *mdsc,
3335 			struct ceph_mds_request *req)
3336 {
3337 	struct ceph_client *cl = mdsc->fsc->client;
3338 	struct ceph_mds_session *session = NULL;
3339 	int mds = -1;
3340 	int err = 0;
3341 	bool random;
3342 
3343 	if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
3344 		if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
3345 			__unregister_request(mdsc, req);
3346 		return;
3347 	}
3348 
3349 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_FENCE_IO) {
3350 		doutc(cl, "metadata corrupted\n");
3351 		err = -EIO;
3352 		goto finish;
3353 	}
3354 	if (req->r_timeout &&
3355 	    time_after_eq(jiffies, req->r_started + req->r_timeout)) {
3356 		doutc(cl, "timed out\n");
3357 		err = -ETIMEDOUT;
3358 		goto finish;
3359 	}
3360 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
3361 		doutc(cl, "forced umount\n");
3362 		err = -EIO;
3363 		goto finish;
3364 	}
3365 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
3366 		if (mdsc->mdsmap_err) {
3367 			err = mdsc->mdsmap_err;
3368 			doutc(cl, "mdsmap err %d\n", err);
3369 			goto finish;
3370 		}
3371 		if (mdsc->mdsmap->m_epoch == 0) {
3372 			doutc(cl, "no mdsmap, waiting for map\n");
3373 			list_add(&req->r_wait, &mdsc->waiting_for_map);
3374 			return;
3375 		}
3376 		if (!(mdsc->fsc->mount_options->flags &
3377 		      CEPH_MOUNT_OPT_MOUNTWAIT) &&
3378 		    !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
3379 			err = -EHOSTUNREACH;
3380 			goto finish;
3381 		}
3382 	}
3383 
3384 	put_request_session(req);
3385 
3386 	mds = __choose_mds(mdsc, req, &random);
3387 	if (mds < 0 ||
3388 	    ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
3389 		if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
3390 			err = -EJUKEBOX;
3391 			goto finish;
3392 		}
3393 		doutc(cl, "no mds or not active, waiting for map\n");
3394 		list_add(&req->r_wait, &mdsc->waiting_for_map);
3395 		return;
3396 	}
3397 
3398 	/* get, open session */
3399 	session = __ceph_lookup_mds_session(mdsc, mds);
3400 	if (!session) {
3401 		session = register_session(mdsc, mds);
3402 		if (IS_ERR(session)) {
3403 			err = PTR_ERR(session);
3404 			goto finish;
3405 		}
3406 	}
3407 	req->r_session = ceph_get_mds_session(session);
3408 
3409 	doutc(cl, "mds%d session %p state %s\n", mds, session,
3410 	      ceph_session_state_name(session->s_state));
3411 
3412 	/*
3413 	 * The old ceph will crash the MDSs when see unknown OPs
3414 	 */
3415 	if (req->r_feature_needed > 0 &&
3416 	    !test_bit(req->r_feature_needed, &session->s_features)) {
3417 		err = -EOPNOTSUPP;
3418 		goto out_session;
3419 	}
3420 
3421 	if (session->s_state != CEPH_MDS_SESSION_OPEN &&
3422 	    session->s_state != CEPH_MDS_SESSION_HUNG) {
3423 		/*
3424 		 * We cannot queue async requests since the caps and delegated
3425 		 * inodes are bound to the session. Just return -EJUKEBOX and
3426 		 * let the caller retry a sync request in that case.
3427 		 */
3428 		if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
3429 			err = -EJUKEBOX;
3430 			goto out_session;
3431 		}
3432 
3433 		/*
3434 		 * If the session has been REJECTED, then return a hard error,
3435 		 * unless it's a CLEANRECOVER mount, in which case we'll queue
3436 		 * it to the mdsc queue.
3437 		 */
3438 		if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
3439 			if (ceph_test_mount_opt(mdsc->fsc, CLEANRECOVER))
3440 				list_add(&req->r_wait, &mdsc->waiting_for_map);
3441 			else
3442 				err = -EACCES;
3443 			goto out_session;
3444 		}
3445 
3446 		if (session->s_state == CEPH_MDS_SESSION_NEW ||
3447 		    session->s_state == CEPH_MDS_SESSION_CLOSING) {
3448 			err = __open_session(mdsc, session);
3449 			if (err)
3450 				goto out_session;
3451 			/* retry the same mds later */
3452 			if (random)
3453 				req->r_resend_mds = mds;
3454 		}
3455 		list_add(&req->r_wait, &session->s_waiting);
3456 		goto out_session;
3457 	}
3458 
3459 	/* send request */
3460 	req->r_resend_mds = -1;   /* forget any previous mds hint */
3461 
3462 	if (req->r_request_started == 0)   /* note request start time */
3463 		req->r_request_started = jiffies;
3464 
3465 	/*
3466 	 * For async create we will choose the auth MDS of frag in parent
3467 	 * directory to send the request and ususally this works fine, but
3468 	 * if the migrated the dirtory to another MDS before it could handle
3469 	 * it the request will be forwarded.
3470 	 *
3471 	 * And then the auth cap will be changed.
3472 	 */
3473 	if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags) && req->r_num_fwd) {
3474 		struct ceph_dentry_info *di = ceph_dentry(req->r_dentry);
3475 		struct ceph_inode_info *ci;
3476 		struct ceph_cap *cap;
3477 
3478 		/*
3479 		 * The request maybe handled very fast and the new inode
3480 		 * hasn't been linked to the dentry yet. We need to wait
3481 		 * for the ceph_finish_async_create(), which shouldn't be
3482 		 * stuck too long or fail in thoery, to finish when forwarding
3483 		 * the request.
3484 		 */
3485 		if (!d_inode(req->r_dentry)) {
3486 			err = wait_on_bit(&di->flags, CEPH_DENTRY_ASYNC_CREATE_BIT,
3487 					  TASK_KILLABLE);
3488 			if (err) {
3489 				mutex_lock(&req->r_fill_mutex);
3490 				set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3491 				mutex_unlock(&req->r_fill_mutex);
3492 				goto out_session;
3493 			}
3494 		}
3495 
3496 		ci = ceph_inode(d_inode(req->r_dentry));
3497 
3498 		spin_lock(&ci->i_ceph_lock);
3499 		cap = ci->i_auth_cap;
3500 		if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE && mds != cap->mds) {
3501 			doutc(cl, "session changed for auth cap %d -> %d\n",
3502 			      cap->session->s_mds, session->s_mds);
3503 
3504 			/* Remove the auth cap from old session */
3505 			spin_lock(&cap->session->s_cap_lock);
3506 			cap->session->s_nr_caps--;
3507 			list_del_init(&cap->session_caps);
3508 			spin_unlock(&cap->session->s_cap_lock);
3509 
3510 			/* Add the auth cap to the new session */
3511 			cap->mds = mds;
3512 			cap->session = session;
3513 			spin_lock(&session->s_cap_lock);
3514 			session->s_nr_caps++;
3515 			list_add_tail(&cap->session_caps, &session->s_caps);
3516 			spin_unlock(&session->s_cap_lock);
3517 
3518 			change_auth_cap_ses(ci, session);
3519 		}
3520 		spin_unlock(&ci->i_ceph_lock);
3521 	}
3522 
3523 	err = __send_request(session, req, false);
3524 
3525 out_session:
3526 	ceph_put_mds_session(session);
3527 finish:
3528 	if (err) {
3529 		doutc(cl, "early error %d\n", err);
3530 		req->r_err = err;
3531 		complete_request(mdsc, req);
3532 		__unregister_request(mdsc, req);
3533 	}
3534 	return;
3535 }
3536 
3537 /*
3538  * called under mdsc->mutex
3539  */
3540 static void __wake_requests(struct ceph_mds_client *mdsc,
3541 			    struct list_head *head)
3542 {
3543 	struct ceph_client *cl = mdsc->fsc->client;
3544 	struct ceph_mds_request *req;
3545 	LIST_HEAD(tmp_list);
3546 
3547 	list_splice_init(head, &tmp_list);
3548 
3549 	while (!list_empty(&tmp_list)) {
3550 		req = list_entry(tmp_list.next,
3551 				 struct ceph_mds_request, r_wait);
3552 		list_del_init(&req->r_wait);
3553 		doutc(cl, " wake request %p tid %llu\n", req,
3554 		      req->r_tid);
3555 		__do_request(mdsc, req);
3556 	}
3557 }
3558 
3559 /*
3560  * Wake up threads with requests pending for @mds, so that they can
3561  * resubmit their requests to a possibly different mds.
3562  */
3563 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
3564 {
3565 	struct ceph_client *cl = mdsc->fsc->client;
3566 	struct ceph_mds_request *req;
3567 	struct rb_node *p = rb_first(&mdsc->request_tree);
3568 
3569 	doutc(cl, "kick_requests mds%d\n", mds);
3570 	while (p) {
3571 		req = rb_entry(p, struct ceph_mds_request, r_node);
3572 		p = rb_next(p);
3573 		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3574 			continue;
3575 		if (req->r_attempts > 0)
3576 			continue; /* only new requests */
3577 		if (req->r_session &&
3578 		    req->r_session->s_mds == mds) {
3579 			doutc(cl, " kicking tid %llu\n", req->r_tid);
3580 			list_del_init(&req->r_wait);
3581 			__do_request(mdsc, req);
3582 		}
3583 	}
3584 }
3585 
3586 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
3587 			      struct ceph_mds_request *req)
3588 {
3589 	struct ceph_client *cl = mdsc->fsc->client;
3590 	int err = 0;
3591 
3592 	/* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
3593 	if (req->r_inode)
3594 		ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
3595 	if (req->r_parent) {
3596 		struct ceph_inode_info *ci = ceph_inode(req->r_parent);
3597 		int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
3598 			    CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
3599 		spin_lock(&ci->i_ceph_lock);
3600 		ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
3601 		__ceph_touch_fmode(ci, mdsc, fmode);
3602 		spin_unlock(&ci->i_ceph_lock);
3603 	}
3604 	if (req->r_old_dentry_dir)
3605 		ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
3606 				  CEPH_CAP_PIN);
3607 
3608 	if (req->r_inode) {
3609 		err = ceph_wait_on_async_create(req->r_inode);
3610 		if (err) {
3611 			doutc(cl, "wait for async create returned: %d\n", err);
3612 			return err;
3613 		}
3614 	}
3615 
3616 	if (!err && req->r_old_inode) {
3617 		err = ceph_wait_on_async_create(req->r_old_inode);
3618 		if (err) {
3619 			doutc(cl, "wait for async create returned: %d\n", err);
3620 			return err;
3621 		}
3622 	}
3623 
3624 	doutc(cl, "submit_request on %p for inode %p\n", req, dir);
3625 	mutex_lock(&mdsc->mutex);
3626 	__register_request(mdsc, req, dir);
3627 	__do_request(mdsc, req);
3628 	err = req->r_err;
3629 	mutex_unlock(&mdsc->mutex);
3630 	return err;
3631 }
3632 
3633 int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
3634 			   struct ceph_mds_request *req,
3635 			   ceph_mds_request_wait_callback_t wait_func)
3636 {
3637 	struct ceph_client *cl = mdsc->fsc->client;
3638 	int err;
3639 
3640 	/* wait */
3641 	doutc(cl, "do_request waiting\n");
3642 	if (wait_func) {
3643 		err = wait_func(mdsc, req);
3644 	} else {
3645 		long timeleft = wait_for_completion_killable_timeout(
3646 					&req->r_completion,
3647 					ceph_timeout_jiffies(req->r_timeout));
3648 		if (timeleft > 0)
3649 			err = 0;
3650 		else if (!timeleft)
3651 			err = -ETIMEDOUT;  /* timed out */
3652 		else
3653 			err = timeleft;  /* killed */
3654 	}
3655 	doutc(cl, "do_request waited, got %d\n", err);
3656 	mutex_lock(&mdsc->mutex);
3657 
3658 	/* only abort if we didn't race with a real reply */
3659 	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
3660 		err = le32_to_cpu(req->r_reply_info.head->result);
3661 	} else if (err < 0) {
3662 		doutc(cl, "aborted request %lld with %d\n", req->r_tid, err);
3663 
3664 		/*
3665 		 * ensure we aren't running concurrently with
3666 		 * ceph_fill_trace or ceph_readdir_prepopulate, which
3667 		 * rely on locks (dir mutex) held by our caller.
3668 		 */
3669 		mutex_lock(&req->r_fill_mutex);
3670 		req->r_err = err;
3671 		set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3672 		mutex_unlock(&req->r_fill_mutex);
3673 
3674 		if (req->r_parent &&
3675 		    (req->r_op & CEPH_MDS_OP_WRITE))
3676 			ceph_invalidate_dir_request(req);
3677 	} else {
3678 		err = req->r_err;
3679 	}
3680 
3681 	mutex_unlock(&mdsc->mutex);
3682 	return err;
3683 }
3684 
3685 /*
3686  * Synchrously perform an mds request.  Take care of all of the
3687  * session setup, forwarding, retry details.
3688  */
3689 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
3690 			 struct inode *dir,
3691 			 struct ceph_mds_request *req)
3692 {
3693 	struct ceph_client *cl = mdsc->fsc->client;
3694 	int err;
3695 
3696 	doutc(cl, "do_request on %p\n", req);
3697 
3698 	/* issue */
3699 	err = ceph_mdsc_submit_request(mdsc, dir, req);
3700 	if (!err)
3701 		err = ceph_mdsc_wait_request(mdsc, req, NULL);
3702 	doutc(cl, "do_request %p done, result %d\n", req, err);
3703 	return err;
3704 }
3705 
3706 /*
3707  * Invalidate dir's completeness, dentry lease state on an aborted MDS
3708  * namespace request.
3709  */
3710 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
3711 {
3712 	struct inode *dir = req->r_parent;
3713 	struct inode *old_dir = req->r_old_dentry_dir;
3714 	struct ceph_client *cl = req->r_mdsc->fsc->client;
3715 
3716 	doutc(cl, "invalidate_dir_request %p %p (complete, lease(s))\n",
3717 	      dir, old_dir);
3718 
3719 	ceph_dir_clear_complete(dir);
3720 	if (old_dir)
3721 		ceph_dir_clear_complete(old_dir);
3722 	if (req->r_dentry)
3723 		ceph_invalidate_dentry_lease(req->r_dentry);
3724 	if (req->r_old_dentry)
3725 		ceph_invalidate_dentry_lease(req->r_old_dentry);
3726 }
3727 
3728 /*
3729  * Handle mds reply.
3730  *
3731  * We take the session mutex and parse and process the reply immediately.
3732  * This preserves the logical ordering of replies, capabilities, etc., sent
3733  * by the MDS as they are applied to our local cache.
3734  */
3735 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
3736 {
3737 	struct ceph_mds_client *mdsc = session->s_mdsc;
3738 	struct ceph_client *cl = mdsc->fsc->client;
3739 	struct ceph_mds_request *req;
3740 	struct ceph_mds_reply_head *head = msg->front.iov_base;
3741 	struct ceph_mds_reply_info_parsed *rinfo;  /* parsed reply info */
3742 	struct ceph_snap_realm *realm;
3743 	u64 tid;
3744 	int err, result;
3745 	int mds = session->s_mds;
3746 	bool close_sessions = false;
3747 
3748 	if (msg->front.iov_len < sizeof(*head)) {
3749 		pr_err_client(cl, "got corrupt (short) reply\n");
3750 		ceph_msg_dump(msg);
3751 		return;
3752 	}
3753 
3754 	/* get request, session */
3755 	tid = le64_to_cpu(msg->hdr.tid);
3756 	mutex_lock(&mdsc->mutex);
3757 	req = lookup_get_request(mdsc, tid);
3758 	if (!req) {
3759 		doutc(cl, "on unknown tid %llu\n", tid);
3760 		mutex_unlock(&mdsc->mutex);
3761 		return;
3762 	}
3763 	doutc(cl, "handle_reply %p\n", req);
3764 
3765 	/* correct session? */
3766 	if (req->r_session != session) {
3767 		pr_err_client(cl, "got %llu on session mds%d not mds%d\n",
3768 			      tid, session->s_mds,
3769 			      req->r_session ? req->r_session->s_mds : -1);
3770 		mutex_unlock(&mdsc->mutex);
3771 		goto out;
3772 	}
3773 
3774 	/* dup? */
3775 	if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3776 	    (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
3777 		pr_warn_client(cl, "got a dup %s reply on %llu from mds%d\n",
3778 			       head->safe ? "safe" : "unsafe", tid, mds);
3779 		mutex_unlock(&mdsc->mutex);
3780 		goto out;
3781 	}
3782 	if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
3783 		pr_warn_client(cl, "got unsafe after safe on %llu from mds%d\n",
3784 			       tid, mds);
3785 		mutex_unlock(&mdsc->mutex);
3786 		goto out;
3787 	}
3788 
3789 	result = le32_to_cpu(head->result);
3790 
3791 	if (head->safe) {
3792 		set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
3793 		__unregister_request(mdsc, req);
3794 
3795 		/* last request during umount? */
3796 		if (mdsc->stopping && !__get_oldest_req(mdsc))
3797 			complete_all(&mdsc->safe_umount_waiters);
3798 
3799 		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3800 			/*
3801 			 * We already handled the unsafe response, now do the
3802 			 * cleanup.  No need to examine the response; the MDS
3803 			 * doesn't include any result info in the safe
3804 			 * response.  And even if it did, there is nothing
3805 			 * useful we could do with a revised return value.
3806 			 */
3807 			doutc(cl, "got safe reply %llu, mds%d\n", tid, mds);
3808 
3809 			mutex_unlock(&mdsc->mutex);
3810 			goto out;
3811 		}
3812 	} else {
3813 		set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
3814 		list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3815 	}
3816 
3817 	doutc(cl, "tid %lld result %d\n", tid, result);
3818 	if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
3819 		err = parse_reply_info(session, msg, req, (u64)-1);
3820 	else
3821 		err = parse_reply_info(session, msg, req,
3822 				       session->s_con.peer_features);
3823 	mutex_unlock(&mdsc->mutex);
3824 
3825 	/* Must find target inode outside of mutexes to avoid deadlocks */
3826 	rinfo = &req->r_reply_info;
3827 	if ((err >= 0) && rinfo->head->is_target) {
3828 		struct inode *in = xchg(&req->r_new_inode, NULL);
3829 		struct ceph_vino tvino = {
3830 			.ino  = le64_to_cpu(rinfo->targeti.in->ino),
3831 			.snap = le64_to_cpu(rinfo->targeti.in->snapid)
3832 		};
3833 
3834 		/*
3835 		 * If we ended up opening an existing inode, discard
3836 		 * r_new_inode
3837 		 */
3838 		if (req->r_op == CEPH_MDS_OP_CREATE &&
3839 		    !req->r_reply_info.has_create_ino) {
3840 			/* This should never happen on an async create */
3841 			WARN_ON_ONCE(req->r_deleg_ino);
3842 			iput(in);
3843 			in = NULL;
3844 		}
3845 
3846 		in = ceph_get_inode(mdsc->fsc->sb, tvino, in);
3847 		if (IS_ERR(in)) {
3848 			err = PTR_ERR(in);
3849 			mutex_lock(&session->s_mutex);
3850 			goto out_err;
3851 		}
3852 		req->r_target_inode = in;
3853 	}
3854 
3855 	mutex_lock(&session->s_mutex);
3856 	if (err < 0) {
3857 		pr_err_client(cl, "got corrupt reply mds%d(tid:%lld)\n",
3858 			      mds, tid);
3859 		ceph_msg_dump(msg);
3860 		goto out_err;
3861 	}
3862 
3863 	/* snap trace */
3864 	realm = NULL;
3865 	if (rinfo->snapblob_len) {
3866 		down_write(&mdsc->snap_rwsem);
3867 		err = ceph_update_snap_trace(mdsc, rinfo->snapblob,
3868 				rinfo->snapblob + rinfo->snapblob_len,
3869 				le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3870 				&realm);
3871 		if (err) {
3872 			up_write(&mdsc->snap_rwsem);
3873 			close_sessions = true;
3874 			if (err == -EIO)
3875 				ceph_msg_dump(msg);
3876 			goto out_err;
3877 		}
3878 		downgrade_write(&mdsc->snap_rwsem);
3879 	} else {
3880 		down_read(&mdsc->snap_rwsem);
3881 	}
3882 
3883 	/* insert trace into our cache */
3884 	mutex_lock(&req->r_fill_mutex);
3885 	current->journal_info = req;
3886 	err = ceph_fill_trace(mdsc->fsc->sb, req);
3887 	if (err == 0) {
3888 		if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
3889 				    req->r_op == CEPH_MDS_OP_LSSNAP))
3890 			err = ceph_readdir_prepopulate(req, req->r_session);
3891 	}
3892 	current->journal_info = NULL;
3893 	mutex_unlock(&req->r_fill_mutex);
3894 
3895 	up_read(&mdsc->snap_rwsem);
3896 	if (realm)
3897 		ceph_put_snap_realm(mdsc, realm);
3898 
3899 	if (err == 0) {
3900 		if (req->r_target_inode &&
3901 		    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3902 			struct ceph_inode_info *ci =
3903 				ceph_inode(req->r_target_inode);
3904 			spin_lock(&ci->i_unsafe_lock);
3905 			list_add_tail(&req->r_unsafe_target_item,
3906 				      &ci->i_unsafe_iops);
3907 			spin_unlock(&ci->i_unsafe_lock);
3908 		}
3909 
3910 		ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
3911 	}
3912 out_err:
3913 	mutex_lock(&mdsc->mutex);
3914 	if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3915 		if (err) {
3916 			req->r_err = err;
3917 		} else {
3918 			req->r_reply =  ceph_msg_get(msg);
3919 			set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
3920 		}
3921 	} else {
3922 		doutc(cl, "reply arrived after request %lld was aborted\n", tid);
3923 	}
3924 	mutex_unlock(&mdsc->mutex);
3925 
3926 	mutex_unlock(&session->s_mutex);
3927 
3928 	/* kick calling process */
3929 	complete_request(mdsc, req);
3930 
3931 	ceph_update_metadata_metrics(&mdsc->metric, req->r_start_latency,
3932 				     req->r_end_latency, err);
3933 out:
3934 	ceph_mdsc_put_request(req);
3935 
3936 	/* Defer closing the sessions after s_mutex lock being released */
3937 	if (close_sessions)
3938 		ceph_mdsc_close_sessions(mdsc);
3939 	return;
3940 }
3941 
3942 
3943 
3944 /*
3945  * handle mds notification that our request has been forwarded.
3946  */
3947 static void handle_forward(struct ceph_mds_client *mdsc,
3948 			   struct ceph_mds_session *session,
3949 			   struct ceph_msg *msg)
3950 {
3951 	struct ceph_client *cl = mdsc->fsc->client;
3952 	struct ceph_mds_request *req;
3953 	u64 tid = le64_to_cpu(msg->hdr.tid);
3954 	u32 next_mds;
3955 	u32 fwd_seq;
3956 	int err = -EINVAL;
3957 	void *p = msg->front.iov_base;
3958 	void *end = p + msg->front.iov_len;
3959 	bool aborted = false;
3960 
3961 	ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3962 	next_mds = ceph_decode_32(&p);
3963 	fwd_seq = ceph_decode_32(&p);
3964 
3965 	mutex_lock(&mdsc->mutex);
3966 	req = lookup_get_request(mdsc, tid);
3967 	if (!req) {
3968 		mutex_unlock(&mdsc->mutex);
3969 		doutc(cl, "forward tid %llu to mds%d - req dne\n", tid, next_mds);
3970 		return;  /* dup reply? */
3971 	}
3972 
3973 	if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3974 		doutc(cl, "forward tid %llu aborted, unregistering\n", tid);
3975 		__unregister_request(mdsc, req);
3976 	} else if (fwd_seq <= req->r_num_fwd || (uint32_t)fwd_seq >= U32_MAX) {
3977 		/*
3978 		 * Avoid inifinite retrying after overflow.
3979 		 *
3980 		 * The MDS will increase the fwd count and in client side
3981 		 * if the num_fwd is less than the one saved in request
3982 		 * that means the MDS is an old version and overflowed of
3983 		 * 8 bits.
3984 		 */
3985 		mutex_lock(&req->r_fill_mutex);
3986 		req->r_err = -EMULTIHOP;
3987 		set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3988 		mutex_unlock(&req->r_fill_mutex);
3989 		aborted = true;
3990 		pr_warn_ratelimited_client(cl, "forward tid %llu seq overflow\n",
3991 					   tid);
3992 	} else {
3993 		/* resend. forward race not possible; mds would drop */
3994 		doutc(cl, "forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3995 		BUG_ON(req->r_err);
3996 		BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3997 		req->r_attempts = 0;
3998 		req->r_num_fwd = fwd_seq;
3999 		req->r_resend_mds = next_mds;
4000 		put_request_session(req);
4001 		__do_request(mdsc, req);
4002 	}
4003 	mutex_unlock(&mdsc->mutex);
4004 
4005 	/* kick calling process */
4006 	if (aborted)
4007 		complete_request(mdsc, req);
4008 	ceph_mdsc_put_request(req);
4009 	return;
4010 
4011 bad:
4012 	pr_err_client(cl, "decode error err=%d\n", err);
4013 	ceph_msg_dump(msg);
4014 }
4015 
4016 static int __decode_session_metadata(void **p, void *end,
4017 				     bool *blocklisted)
4018 {
4019 	/* map<string,string> */
4020 	u32 n;
4021 	bool err_str;
4022 	ceph_decode_32_safe(p, end, n, bad);
4023 	while (n-- > 0) {
4024 		u32 len;
4025 		ceph_decode_32_safe(p, end, len, bad);
4026 		ceph_decode_need(p, end, len, bad);
4027 		err_str = !strncmp(*p, "error_string", len);
4028 		*p += len;
4029 		ceph_decode_32_safe(p, end, len, bad);
4030 		ceph_decode_need(p, end, len, bad);
4031 		/*
4032 		 * Match "blocklisted (blacklisted)" from newer MDSes,
4033 		 * or "blacklisted" from older MDSes.
4034 		 */
4035 		if (err_str && strnstr(*p, "blacklisted", len))
4036 			*blocklisted = true;
4037 		*p += len;
4038 	}
4039 	return 0;
4040 bad:
4041 	return -1;
4042 }
4043 
4044 /*
4045  * handle a mds session control message
4046  */
4047 static void handle_session(struct ceph_mds_session *session,
4048 			   struct ceph_msg *msg)
4049 {
4050 	struct ceph_mds_client *mdsc = session->s_mdsc;
4051 	struct ceph_client *cl = mdsc->fsc->client;
4052 	int mds = session->s_mds;
4053 	int msg_version = le16_to_cpu(msg->hdr.version);
4054 	void *p = msg->front.iov_base;
4055 	void *end = p + msg->front.iov_len;
4056 	struct ceph_mds_session_head *h;
4057 	u32 op;
4058 	u64 seq, features = 0;
4059 	int wake = 0;
4060 	bool blocklisted = false;
4061 
4062 	/* decode */
4063 	ceph_decode_need(&p, end, sizeof(*h), bad);
4064 	h = p;
4065 	p += sizeof(*h);
4066 
4067 	op = le32_to_cpu(h->op);
4068 	seq = le64_to_cpu(h->seq);
4069 
4070 	if (msg_version >= 3) {
4071 		u32 len;
4072 		/* version >= 2 and < 5, decode metadata, skip otherwise
4073 		 * as it's handled via flags.
4074 		 */
4075 		if (msg_version >= 5)
4076 			ceph_decode_skip_map(&p, end, string, string, bad);
4077 		else if (__decode_session_metadata(&p, end, &blocklisted) < 0)
4078 			goto bad;
4079 
4080 		/* version >= 3, feature bits */
4081 		ceph_decode_32_safe(&p, end, len, bad);
4082 		if (len) {
4083 			ceph_decode_64_safe(&p, end, features, bad);
4084 			p += len - sizeof(features);
4085 		}
4086 	}
4087 
4088 	if (msg_version >= 5) {
4089 		u32 flags, len;
4090 
4091 		/* version >= 4 */
4092 		ceph_decode_skip_16(&p, end, bad); /* struct_v, struct_cv */
4093 		ceph_decode_32_safe(&p, end, len, bad); /* len */
4094 		ceph_decode_skip_n(&p, end, len, bad); /* metric_spec */
4095 
4096 		/* version >= 5, flags   */
4097 		ceph_decode_32_safe(&p, end, flags, bad);
4098 		if (flags & CEPH_SESSION_BLOCKLISTED) {
4099 			pr_warn_client(cl, "mds%d session blocklisted\n",
4100 				       session->s_mds);
4101 			blocklisted = true;
4102 		}
4103 	}
4104 
4105 	mutex_lock(&mdsc->mutex);
4106 	if (op == CEPH_SESSION_CLOSE) {
4107 		ceph_get_mds_session(session);
4108 		__unregister_session(mdsc, session);
4109 	}
4110 	/* FIXME: this ttl calculation is generous */
4111 	session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
4112 	mutex_unlock(&mdsc->mutex);
4113 
4114 	mutex_lock(&session->s_mutex);
4115 
4116 	doutc(cl, "mds%d %s %p state %s seq %llu\n", mds,
4117 	      ceph_session_op_name(op), session,
4118 	      ceph_session_state_name(session->s_state), seq);
4119 
4120 	if (session->s_state == CEPH_MDS_SESSION_HUNG) {
4121 		session->s_state = CEPH_MDS_SESSION_OPEN;
4122 		pr_info_client(cl, "mds%d came back\n", session->s_mds);
4123 	}
4124 
4125 	switch (op) {
4126 	case CEPH_SESSION_OPEN:
4127 		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
4128 			pr_info_client(cl, "mds%d reconnect success\n",
4129 				       session->s_mds);
4130 
4131 		if (session->s_state == CEPH_MDS_SESSION_OPEN) {
4132 			pr_notice_client(cl, "mds%d is already opened\n",
4133 					 session->s_mds);
4134 		} else {
4135 			session->s_state = CEPH_MDS_SESSION_OPEN;
4136 			session->s_features = features;
4137 			renewed_caps(mdsc, session, 0);
4138 			if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT,
4139 				     &session->s_features))
4140 				metric_schedule_delayed(&mdsc->metric);
4141 		}
4142 
4143 		/*
4144 		 * The connection maybe broken and the session in client
4145 		 * side has been reinitialized, need to update the seq
4146 		 * anyway.
4147 		 */
4148 		if (!session->s_seq && seq)
4149 			session->s_seq = seq;
4150 
4151 		wake = 1;
4152 		if (mdsc->stopping)
4153 			__close_session(mdsc, session);
4154 		break;
4155 
4156 	case CEPH_SESSION_RENEWCAPS:
4157 		if (session->s_renew_seq == seq)
4158 			renewed_caps(mdsc, session, 1);
4159 		break;
4160 
4161 	case CEPH_SESSION_CLOSE:
4162 		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
4163 			pr_info_client(cl, "mds%d reconnect denied\n",
4164 				       session->s_mds);
4165 		session->s_state = CEPH_MDS_SESSION_CLOSED;
4166 		cleanup_session_requests(mdsc, session);
4167 		remove_session_caps(session);
4168 		wake = 2; /* for good measure */
4169 		wake_up_all(&mdsc->session_close_wq);
4170 		break;
4171 
4172 	case CEPH_SESSION_STALE:
4173 		pr_info_client(cl, "mds%d caps went stale, renewing\n",
4174 			       session->s_mds);
4175 		atomic_inc(&session->s_cap_gen);
4176 		session->s_cap_ttl = jiffies - 1;
4177 		send_renew_caps(mdsc, session);
4178 		break;
4179 
4180 	case CEPH_SESSION_RECALL_STATE:
4181 		ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
4182 		break;
4183 
4184 	case CEPH_SESSION_FLUSHMSG:
4185 		/* flush cap releases */
4186 		spin_lock(&session->s_cap_lock);
4187 		if (session->s_num_cap_releases)
4188 			ceph_flush_cap_releases(mdsc, session);
4189 		spin_unlock(&session->s_cap_lock);
4190 
4191 		send_flushmsg_ack(mdsc, session, seq);
4192 		break;
4193 
4194 	case CEPH_SESSION_FORCE_RO:
4195 		doutc(cl, "force_session_readonly %p\n", session);
4196 		spin_lock(&session->s_cap_lock);
4197 		session->s_readonly = true;
4198 		spin_unlock(&session->s_cap_lock);
4199 		wake_up_session_caps(session, FORCE_RO);
4200 		break;
4201 
4202 	case CEPH_SESSION_REJECT:
4203 		WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
4204 		pr_info_client(cl, "mds%d rejected session\n",
4205 			       session->s_mds);
4206 		session->s_state = CEPH_MDS_SESSION_REJECTED;
4207 		cleanup_session_requests(mdsc, session);
4208 		remove_session_caps(session);
4209 		if (blocklisted)
4210 			mdsc->fsc->blocklisted = true;
4211 		wake = 2; /* for good measure */
4212 		break;
4213 
4214 	default:
4215 		pr_err_client(cl, "bad op %d mds%d\n", op, mds);
4216 		WARN_ON(1);
4217 	}
4218 
4219 	mutex_unlock(&session->s_mutex);
4220 	if (wake) {
4221 		mutex_lock(&mdsc->mutex);
4222 		__wake_requests(mdsc, &session->s_waiting);
4223 		if (wake == 2)
4224 			kick_requests(mdsc, mds);
4225 		mutex_unlock(&mdsc->mutex);
4226 	}
4227 	if (op == CEPH_SESSION_CLOSE)
4228 		ceph_put_mds_session(session);
4229 	return;
4230 
4231 bad:
4232 	pr_err_client(cl, "corrupt message mds%d len %d\n", mds,
4233 		      (int)msg->front.iov_len);
4234 	ceph_msg_dump(msg);
4235 	return;
4236 }
4237 
4238 void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
4239 {
4240 	struct ceph_client *cl = req->r_mdsc->fsc->client;
4241 	int dcaps;
4242 
4243 	dcaps = xchg(&req->r_dir_caps, 0);
4244 	if (dcaps) {
4245 		doutc(cl, "releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
4246 		ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
4247 	}
4248 }
4249 
4250 void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
4251 {
4252 	struct ceph_client *cl = req->r_mdsc->fsc->client;
4253 	int dcaps;
4254 
4255 	dcaps = xchg(&req->r_dir_caps, 0);
4256 	if (dcaps) {
4257 		doutc(cl, "releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
4258 		ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
4259 						dcaps);
4260 	}
4261 }
4262 
4263 /*
4264  * called under session->mutex.
4265  */
4266 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
4267 				   struct ceph_mds_session *session)
4268 {
4269 	struct ceph_mds_request *req, *nreq;
4270 	struct rb_node *p;
4271 
4272 	doutc(mdsc->fsc->client, "mds%d\n", session->s_mds);
4273 
4274 	mutex_lock(&mdsc->mutex);
4275 	list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
4276 		__send_request(session, req, true);
4277 
4278 	/*
4279 	 * also re-send old requests when MDS enters reconnect stage. So that MDS
4280 	 * can process completed request in clientreplay stage.
4281 	 */
4282 	p = rb_first(&mdsc->request_tree);
4283 	while (p) {
4284 		req = rb_entry(p, struct ceph_mds_request, r_node);
4285 		p = rb_next(p);
4286 		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
4287 			continue;
4288 		if (req->r_attempts == 0)
4289 			continue; /* only old requests */
4290 		if (!req->r_session)
4291 			continue;
4292 		if (req->r_session->s_mds != session->s_mds)
4293 			continue;
4294 
4295 		ceph_mdsc_release_dir_caps_no_check(req);
4296 
4297 		__send_request(session, req, true);
4298 	}
4299 	mutex_unlock(&mdsc->mutex);
4300 }
4301 
4302 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
4303 {
4304 	struct ceph_msg *reply;
4305 	struct ceph_pagelist *_pagelist;
4306 	struct page *page;
4307 	__le32 *addr;
4308 	int err = -ENOMEM;
4309 
4310 	if (!recon_state->allow_multi)
4311 		return -ENOSPC;
4312 
4313 	/* can't handle message that contains both caps and realm */
4314 	BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
4315 
4316 	/* pre-allocate new pagelist */
4317 	_pagelist = ceph_pagelist_alloc(GFP_NOFS);
4318 	if (!_pagelist)
4319 		return -ENOMEM;
4320 
4321 	reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
4322 	if (!reply)
4323 		goto fail_msg;
4324 
4325 	/* placeholder for nr_caps */
4326 	err = ceph_pagelist_encode_32(_pagelist, 0);
4327 	if (err < 0)
4328 		goto fail;
4329 
4330 	if (recon_state->nr_caps) {
4331 		/* currently encoding caps */
4332 		err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
4333 		if (err)
4334 			goto fail;
4335 	} else {
4336 		/* placeholder for nr_realms (currently encoding relams) */
4337 		err = ceph_pagelist_encode_32(_pagelist, 0);
4338 		if (err < 0)
4339 			goto fail;
4340 	}
4341 
4342 	err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
4343 	if (err)
4344 		goto fail;
4345 
4346 	page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
4347 	addr = kmap_atomic(page);
4348 	if (recon_state->nr_caps) {
4349 		/* currently encoding caps */
4350 		*addr = cpu_to_le32(recon_state->nr_caps);
4351 	} else {
4352 		/* currently encoding relams */
4353 		*(addr + 1) = cpu_to_le32(recon_state->nr_realms);
4354 	}
4355 	kunmap_atomic(addr);
4356 
4357 	reply->hdr.version = cpu_to_le16(5);
4358 	reply->hdr.compat_version = cpu_to_le16(4);
4359 
4360 	reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
4361 	ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
4362 
4363 	ceph_con_send(&recon_state->session->s_con, reply);
4364 	ceph_pagelist_release(recon_state->pagelist);
4365 
4366 	recon_state->pagelist = _pagelist;
4367 	recon_state->nr_caps = 0;
4368 	recon_state->nr_realms = 0;
4369 	recon_state->msg_version = 5;
4370 	return 0;
4371 fail:
4372 	ceph_msg_put(reply);
4373 fail_msg:
4374 	ceph_pagelist_release(_pagelist);
4375 	return err;
4376 }
4377 
4378 static struct dentry* d_find_primary(struct inode *inode)
4379 {
4380 	struct dentry *alias, *dn = NULL;
4381 
4382 	if (hlist_empty(&inode->i_dentry))
4383 		return NULL;
4384 
4385 	spin_lock(&inode->i_lock);
4386 	if (hlist_empty(&inode->i_dentry))
4387 		goto out_unlock;
4388 
4389 	if (S_ISDIR(inode->i_mode)) {
4390 		alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
4391 		if (!IS_ROOT(alias))
4392 			dn = dget(alias);
4393 		goto out_unlock;
4394 	}
4395 
4396 	hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
4397 		spin_lock(&alias->d_lock);
4398 		if (!d_unhashed(alias) &&
4399 		    (ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {
4400 			dn = dget_dlock(alias);
4401 		}
4402 		spin_unlock(&alias->d_lock);
4403 		if (dn)
4404 			break;
4405 	}
4406 out_unlock:
4407 	spin_unlock(&inode->i_lock);
4408 	return dn;
4409 }
4410 
4411 /*
4412  * Encode information about a cap for a reconnect with the MDS.
4413  */
4414 static int reconnect_caps_cb(struct inode *inode, int mds, void *arg)
4415 {
4416 	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
4417 	struct ceph_client *cl = ceph_inode_to_client(inode);
4418 	union {
4419 		struct ceph_mds_cap_reconnect v2;
4420 		struct ceph_mds_cap_reconnect_v1 v1;
4421 	} rec;
4422 	struct ceph_inode_info *ci = ceph_inode(inode);
4423 	struct ceph_reconnect_state *recon_state = arg;
4424 	struct ceph_pagelist *pagelist = recon_state->pagelist;
4425 	struct dentry *dentry;
4426 	struct ceph_cap *cap;
4427 	char *path;
4428 	int pathlen = 0, err;
4429 	u64 pathbase;
4430 	u64 snap_follows;
4431 
4432 	dentry = d_find_primary(inode);
4433 	if (dentry) {
4434 		/* set pathbase to parent dir when msg_version >= 2 */
4435 		path = ceph_mdsc_build_path(mdsc, dentry, &pathlen, &pathbase,
4436 					    recon_state->msg_version >= 2);
4437 		dput(dentry);
4438 		if (IS_ERR(path)) {
4439 			err = PTR_ERR(path);
4440 			goto out_err;
4441 		}
4442 	} else {
4443 		path = NULL;
4444 		pathbase = 0;
4445 	}
4446 
4447 	spin_lock(&ci->i_ceph_lock);
4448 	cap = __get_cap_for_mds(ci, mds);
4449 	if (!cap) {
4450 		spin_unlock(&ci->i_ceph_lock);
4451 		err = 0;
4452 		goto out_err;
4453 	}
4454 	doutc(cl, " adding %p ino %llx.%llx cap %p %lld %s\n", inode,
4455 	      ceph_vinop(inode), cap, cap->cap_id,
4456 	      ceph_cap_string(cap->issued));
4457 
4458 	cap->seq = 0;        /* reset cap seq */
4459 	cap->issue_seq = 0;  /* and issue_seq */
4460 	cap->mseq = 0;       /* and migrate_seq */
4461 	cap->cap_gen = atomic_read(&cap->session->s_cap_gen);
4462 
4463 	/* These are lost when the session goes away */
4464 	if (S_ISDIR(inode->i_mode)) {
4465 		if (cap->issued & CEPH_CAP_DIR_CREATE) {
4466 			ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
4467 			memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
4468 		}
4469 		cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
4470 	}
4471 
4472 	if (recon_state->msg_version >= 2) {
4473 		rec.v2.cap_id = cpu_to_le64(cap->cap_id);
4474 		rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
4475 		rec.v2.issued = cpu_to_le32(cap->issued);
4476 		rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
4477 		rec.v2.pathbase = cpu_to_le64(pathbase);
4478 		rec.v2.flock_len = (__force __le32)
4479 			((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
4480 	} else {
4481 		struct timespec64 ts;
4482 
4483 		rec.v1.cap_id = cpu_to_le64(cap->cap_id);
4484 		rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
4485 		rec.v1.issued = cpu_to_le32(cap->issued);
4486 		rec.v1.size = cpu_to_le64(i_size_read(inode));
4487 		ts = inode_get_mtime(inode);
4488 		ceph_encode_timespec64(&rec.v1.mtime, &ts);
4489 		ts = inode_get_atime(inode);
4490 		ceph_encode_timespec64(&rec.v1.atime, &ts);
4491 		rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
4492 		rec.v1.pathbase = cpu_to_le64(pathbase);
4493 	}
4494 
4495 	if (list_empty(&ci->i_cap_snaps)) {
4496 		snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
4497 	} else {
4498 		struct ceph_cap_snap *capsnap =
4499 			list_first_entry(&ci->i_cap_snaps,
4500 					 struct ceph_cap_snap, ci_item);
4501 		snap_follows = capsnap->follows;
4502 	}
4503 	spin_unlock(&ci->i_ceph_lock);
4504 
4505 	if (recon_state->msg_version >= 2) {
4506 		int num_fcntl_locks, num_flock_locks;
4507 		struct ceph_filelock *flocks = NULL;
4508 		size_t struct_len, total_len = sizeof(u64);
4509 		u8 struct_v = 0;
4510 
4511 encode_again:
4512 		if (rec.v2.flock_len) {
4513 			ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
4514 		} else {
4515 			num_fcntl_locks = 0;
4516 			num_flock_locks = 0;
4517 		}
4518 		if (num_fcntl_locks + num_flock_locks > 0) {
4519 			flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
4520 					       sizeof(struct ceph_filelock),
4521 					       GFP_NOFS);
4522 			if (!flocks) {
4523 				err = -ENOMEM;
4524 				goto out_err;
4525 			}
4526 			err = ceph_encode_locks_to_buffer(inode, flocks,
4527 							  num_fcntl_locks,
4528 							  num_flock_locks);
4529 			if (err) {
4530 				kfree(flocks);
4531 				flocks = NULL;
4532 				if (err == -ENOSPC)
4533 					goto encode_again;
4534 				goto out_err;
4535 			}
4536 		} else {
4537 			kfree(flocks);
4538 			flocks = NULL;
4539 		}
4540 
4541 		if (recon_state->msg_version >= 3) {
4542 			/* version, compat_version and struct_len */
4543 			total_len += 2 * sizeof(u8) + sizeof(u32);
4544 			struct_v = 2;
4545 		}
4546 		/*
4547 		 * number of encoded locks is stable, so copy to pagelist
4548 		 */
4549 		struct_len = 2 * sizeof(u32) +
4550 			    (num_fcntl_locks + num_flock_locks) *
4551 			    sizeof(struct ceph_filelock);
4552 		rec.v2.flock_len = cpu_to_le32(struct_len);
4553 
4554 		struct_len += sizeof(u32) + pathlen + sizeof(rec.v2);
4555 
4556 		if (struct_v >= 2)
4557 			struct_len += sizeof(u64); /* snap_follows */
4558 
4559 		total_len += struct_len;
4560 
4561 		if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
4562 			err = send_reconnect_partial(recon_state);
4563 			if (err)
4564 				goto out_freeflocks;
4565 			pagelist = recon_state->pagelist;
4566 		}
4567 
4568 		err = ceph_pagelist_reserve(pagelist, total_len);
4569 		if (err)
4570 			goto out_freeflocks;
4571 
4572 		ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
4573 		if (recon_state->msg_version >= 3) {
4574 			ceph_pagelist_encode_8(pagelist, struct_v);
4575 			ceph_pagelist_encode_8(pagelist, 1);
4576 			ceph_pagelist_encode_32(pagelist, struct_len);
4577 		}
4578 		ceph_pagelist_encode_string(pagelist, path, pathlen);
4579 		ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
4580 		ceph_locks_to_pagelist(flocks, pagelist,
4581 				       num_fcntl_locks, num_flock_locks);
4582 		if (struct_v >= 2)
4583 			ceph_pagelist_encode_64(pagelist, snap_follows);
4584 out_freeflocks:
4585 		kfree(flocks);
4586 	} else {
4587 		err = ceph_pagelist_reserve(pagelist,
4588 					    sizeof(u64) + sizeof(u32) +
4589 					    pathlen + sizeof(rec.v1));
4590 		if (err)
4591 			goto out_err;
4592 
4593 		ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
4594 		ceph_pagelist_encode_string(pagelist, path, pathlen);
4595 		ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
4596 	}
4597 
4598 out_err:
4599 	ceph_mdsc_free_path(path, pathlen);
4600 	if (!err)
4601 		recon_state->nr_caps++;
4602 	return err;
4603 }
4604 
4605 static int encode_snap_realms(struct ceph_mds_client *mdsc,
4606 			      struct ceph_reconnect_state *recon_state)
4607 {
4608 	struct rb_node *p;
4609 	struct ceph_pagelist *pagelist = recon_state->pagelist;
4610 	struct ceph_client *cl = mdsc->fsc->client;
4611 	int err = 0;
4612 
4613 	if (recon_state->msg_version >= 4) {
4614 		err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
4615 		if (err < 0)
4616 			goto fail;
4617 	}
4618 
4619 	/*
4620 	 * snaprealms.  we provide mds with the ino, seq (version), and
4621 	 * parent for all of our realms.  If the mds has any newer info,
4622 	 * it will tell us.
4623 	 */
4624 	for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
4625 		struct ceph_snap_realm *realm =
4626 		       rb_entry(p, struct ceph_snap_realm, node);
4627 		struct ceph_mds_snaprealm_reconnect sr_rec;
4628 
4629 		if (recon_state->msg_version >= 4) {
4630 			size_t need = sizeof(u8) * 2 + sizeof(u32) +
4631 				      sizeof(sr_rec);
4632 
4633 			if (pagelist->length + need > RECONNECT_MAX_SIZE) {
4634 				err = send_reconnect_partial(recon_state);
4635 				if (err)
4636 					goto fail;
4637 				pagelist = recon_state->pagelist;
4638 			}
4639 
4640 			err = ceph_pagelist_reserve(pagelist, need);
4641 			if (err)
4642 				goto fail;
4643 
4644 			ceph_pagelist_encode_8(pagelist, 1);
4645 			ceph_pagelist_encode_8(pagelist, 1);
4646 			ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
4647 		}
4648 
4649 		doutc(cl, " adding snap realm %llx seq %lld parent %llx\n",
4650 		      realm->ino, realm->seq, realm->parent_ino);
4651 		sr_rec.ino = cpu_to_le64(realm->ino);
4652 		sr_rec.seq = cpu_to_le64(realm->seq);
4653 		sr_rec.parent = cpu_to_le64(realm->parent_ino);
4654 
4655 		err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
4656 		if (err)
4657 			goto fail;
4658 
4659 		recon_state->nr_realms++;
4660 	}
4661 fail:
4662 	return err;
4663 }
4664 
4665 
4666 /*
4667  * If an MDS fails and recovers, clients need to reconnect in order to
4668  * reestablish shared state.  This includes all caps issued through
4669  * this session _and_ the snap_realm hierarchy.  Because it's not
4670  * clear which snap realms the mds cares about, we send everything we
4671  * know about.. that ensures we'll then get any new info the
4672  * recovering MDS might have.
4673  *
4674  * This is a relatively heavyweight operation, but it's rare.
4675  */
4676 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
4677 			       struct ceph_mds_session *session)
4678 {
4679 	struct ceph_client *cl = mdsc->fsc->client;
4680 	struct ceph_msg *reply;
4681 	int mds = session->s_mds;
4682 	int err = -ENOMEM;
4683 	struct ceph_reconnect_state recon_state = {
4684 		.session = session,
4685 	};
4686 	LIST_HEAD(dispose);
4687 
4688 	pr_info_client(cl, "mds%d reconnect start\n", mds);
4689 
4690 	recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
4691 	if (!recon_state.pagelist)
4692 		goto fail_nopagelist;
4693 
4694 	reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
4695 	if (!reply)
4696 		goto fail_nomsg;
4697 
4698 	xa_destroy(&session->s_delegated_inos);
4699 
4700 	mutex_lock(&session->s_mutex);
4701 	session->s_state = CEPH_MDS_SESSION_RECONNECTING;
4702 	session->s_seq = 0;
4703 
4704 	doutc(cl, "session %p state %s\n", session,
4705 	      ceph_session_state_name(session->s_state));
4706 
4707 	atomic_inc(&session->s_cap_gen);
4708 
4709 	spin_lock(&session->s_cap_lock);
4710 	/* don't know if session is readonly */
4711 	session->s_readonly = 0;
4712 	/*
4713 	 * notify __ceph_remove_cap() that we are composing cap reconnect.
4714 	 * If a cap get released before being added to the cap reconnect,
4715 	 * __ceph_remove_cap() should skip queuing cap release.
4716 	 */
4717 	session->s_cap_reconnect = 1;
4718 	/* drop old cap expires; we're about to reestablish that state */
4719 	detach_cap_releases(session, &dispose);
4720 	spin_unlock(&session->s_cap_lock);
4721 	dispose_cap_releases(mdsc, &dispose);
4722 
4723 	/* trim unused caps to reduce MDS's cache rejoin time */
4724 	if (mdsc->fsc->sb->s_root)
4725 		shrink_dcache_parent(mdsc->fsc->sb->s_root);
4726 
4727 	ceph_con_close(&session->s_con);
4728 	ceph_con_open(&session->s_con,
4729 		      CEPH_ENTITY_TYPE_MDS, mds,
4730 		      ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
4731 
4732 	/* replay unsafe requests */
4733 	replay_unsafe_requests(mdsc, session);
4734 
4735 	ceph_early_kick_flushing_caps(mdsc, session);
4736 
4737 	down_read(&mdsc->snap_rwsem);
4738 
4739 	/* placeholder for nr_caps */
4740 	err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
4741 	if (err)
4742 		goto fail;
4743 
4744 	if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
4745 		recon_state.msg_version = 3;
4746 		recon_state.allow_multi = true;
4747 	} else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
4748 		recon_state.msg_version = 3;
4749 	} else {
4750 		recon_state.msg_version = 2;
4751 	}
4752 	/* trsaverse this session's caps */
4753 	err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
4754 
4755 	spin_lock(&session->s_cap_lock);
4756 	session->s_cap_reconnect = 0;
4757 	spin_unlock(&session->s_cap_lock);
4758 
4759 	if (err < 0)
4760 		goto fail;
4761 
4762 	/* check if all realms can be encoded into current message */
4763 	if (mdsc->num_snap_realms) {
4764 		size_t total_len =
4765 			recon_state.pagelist->length +
4766 			mdsc->num_snap_realms *
4767 			sizeof(struct ceph_mds_snaprealm_reconnect);
4768 		if (recon_state.msg_version >= 4) {
4769 			/* number of realms */
4770 			total_len += sizeof(u32);
4771 			/* version, compat_version and struct_len */
4772 			total_len += mdsc->num_snap_realms *
4773 				     (2 * sizeof(u8) + sizeof(u32));
4774 		}
4775 		if (total_len > RECONNECT_MAX_SIZE) {
4776 			if (!recon_state.allow_multi) {
4777 				err = -ENOSPC;
4778 				goto fail;
4779 			}
4780 			if (recon_state.nr_caps) {
4781 				err = send_reconnect_partial(&recon_state);
4782 				if (err)
4783 					goto fail;
4784 			}
4785 			recon_state.msg_version = 5;
4786 		}
4787 	}
4788 
4789 	err = encode_snap_realms(mdsc, &recon_state);
4790 	if (err < 0)
4791 		goto fail;
4792 
4793 	if (recon_state.msg_version >= 5) {
4794 		err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
4795 		if (err < 0)
4796 			goto fail;
4797 	}
4798 
4799 	if (recon_state.nr_caps || recon_state.nr_realms) {
4800 		struct page *page =
4801 			list_first_entry(&recon_state.pagelist->head,
4802 					struct page, lru);
4803 		__le32 *addr = kmap_atomic(page);
4804 		if (recon_state.nr_caps) {
4805 			WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
4806 			*addr = cpu_to_le32(recon_state.nr_caps);
4807 		} else if (recon_state.msg_version >= 4) {
4808 			*(addr + 1) = cpu_to_le32(recon_state.nr_realms);
4809 		}
4810 		kunmap_atomic(addr);
4811 	}
4812 
4813 	reply->hdr.version = cpu_to_le16(recon_state.msg_version);
4814 	if (recon_state.msg_version >= 4)
4815 		reply->hdr.compat_version = cpu_to_le16(4);
4816 
4817 	reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
4818 	ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
4819 
4820 	ceph_con_send(&session->s_con, reply);
4821 
4822 	mutex_unlock(&session->s_mutex);
4823 
4824 	mutex_lock(&mdsc->mutex);
4825 	__wake_requests(mdsc, &session->s_waiting);
4826 	mutex_unlock(&mdsc->mutex);
4827 
4828 	up_read(&mdsc->snap_rwsem);
4829 	ceph_pagelist_release(recon_state.pagelist);
4830 	return;
4831 
4832 fail:
4833 	ceph_msg_put(reply);
4834 	up_read(&mdsc->snap_rwsem);
4835 	mutex_unlock(&session->s_mutex);
4836 fail_nomsg:
4837 	ceph_pagelist_release(recon_state.pagelist);
4838 fail_nopagelist:
4839 	pr_err_client(cl, "error %d preparing reconnect for mds%d\n",
4840 		      err, mds);
4841 	return;
4842 }
4843 
4844 
4845 /*
4846  * compare old and new mdsmaps, kicking requests
4847  * and closing out old connections as necessary
4848  *
4849  * called under mdsc->mutex.
4850  */
4851 static void check_new_map(struct ceph_mds_client *mdsc,
4852 			  struct ceph_mdsmap *newmap,
4853 			  struct ceph_mdsmap *oldmap)
4854 {
4855 	int i, j, err;
4856 	int oldstate, newstate;
4857 	struct ceph_mds_session *s;
4858 	unsigned long targets[DIV_ROUND_UP(CEPH_MAX_MDS, sizeof(unsigned long))] = {0};
4859 	struct ceph_client *cl = mdsc->fsc->client;
4860 
4861 	doutc(cl, "new %u old %u\n", newmap->m_epoch, oldmap->m_epoch);
4862 
4863 	if (newmap->m_info) {
4864 		for (i = 0; i < newmap->possible_max_rank; i++) {
4865 			for (j = 0; j < newmap->m_info[i].num_export_targets; j++)
4866 				set_bit(newmap->m_info[i].export_targets[j], targets);
4867 		}
4868 	}
4869 
4870 	for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4871 		if (!mdsc->sessions[i])
4872 			continue;
4873 		s = mdsc->sessions[i];
4874 		oldstate = ceph_mdsmap_get_state(oldmap, i);
4875 		newstate = ceph_mdsmap_get_state(newmap, i);
4876 
4877 		doutc(cl, "mds%d state %s%s -> %s%s (session %s)\n",
4878 		      i, ceph_mds_state_name(oldstate),
4879 		      ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
4880 		      ceph_mds_state_name(newstate),
4881 		      ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
4882 		      ceph_session_state_name(s->s_state));
4883 
4884 		if (i >= newmap->possible_max_rank) {
4885 			/* force close session for stopped mds */
4886 			ceph_get_mds_session(s);
4887 			__unregister_session(mdsc, s);
4888 			__wake_requests(mdsc, &s->s_waiting);
4889 			mutex_unlock(&mdsc->mutex);
4890 
4891 			mutex_lock(&s->s_mutex);
4892 			cleanup_session_requests(mdsc, s);
4893 			remove_session_caps(s);
4894 			mutex_unlock(&s->s_mutex);
4895 
4896 			ceph_put_mds_session(s);
4897 
4898 			mutex_lock(&mdsc->mutex);
4899 			kick_requests(mdsc, i);
4900 			continue;
4901 		}
4902 
4903 		if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4904 			   ceph_mdsmap_get_addr(newmap, i),
4905 			   sizeof(struct ceph_entity_addr))) {
4906 			/* just close it */
4907 			mutex_unlock(&mdsc->mutex);
4908 			mutex_lock(&s->s_mutex);
4909 			mutex_lock(&mdsc->mutex);
4910 			ceph_con_close(&s->s_con);
4911 			mutex_unlock(&s->s_mutex);
4912 			s->s_state = CEPH_MDS_SESSION_RESTARTING;
4913 		} else if (oldstate == newstate) {
4914 			continue;  /* nothing new with this mds */
4915 		}
4916 
4917 		/*
4918 		 * send reconnect?
4919 		 */
4920 		if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
4921 		    newstate >= CEPH_MDS_STATE_RECONNECT) {
4922 			mutex_unlock(&mdsc->mutex);
4923 			clear_bit(i, targets);
4924 			send_mds_reconnect(mdsc, s);
4925 			mutex_lock(&mdsc->mutex);
4926 		}
4927 
4928 		/*
4929 		 * kick request on any mds that has gone active.
4930 		 */
4931 		if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4932 		    newstate >= CEPH_MDS_STATE_ACTIVE) {
4933 			if (oldstate != CEPH_MDS_STATE_CREATING &&
4934 			    oldstate != CEPH_MDS_STATE_STARTING)
4935 				pr_info_client(cl, "mds%d recovery completed\n",
4936 					       s->s_mds);
4937 			kick_requests(mdsc, i);
4938 			mutex_unlock(&mdsc->mutex);
4939 			mutex_lock(&s->s_mutex);
4940 			mutex_lock(&mdsc->mutex);
4941 			ceph_kick_flushing_caps(mdsc, s);
4942 			mutex_unlock(&s->s_mutex);
4943 			wake_up_session_caps(s, RECONNECT);
4944 		}
4945 	}
4946 
4947 	/*
4948 	 * Only open and reconnect sessions that don't exist yet.
4949 	 */
4950 	for (i = 0; i < newmap->possible_max_rank; i++) {
4951 		/*
4952 		 * In case the import MDS is crashed just after
4953 		 * the EImportStart journal is flushed, so when
4954 		 * a standby MDS takes over it and is replaying
4955 		 * the EImportStart journal the new MDS daemon
4956 		 * will wait the client to reconnect it, but the
4957 		 * client may never register/open the session yet.
4958 		 *
4959 		 * Will try to reconnect that MDS daemon if the
4960 		 * rank number is in the export targets array and
4961 		 * is the up:reconnect state.
4962 		 */
4963 		newstate = ceph_mdsmap_get_state(newmap, i);
4964 		if (!test_bit(i, targets) || newstate != CEPH_MDS_STATE_RECONNECT)
4965 			continue;
4966 
4967 		/*
4968 		 * The session maybe registered and opened by some
4969 		 * requests which were choosing random MDSes during
4970 		 * the mdsc->mutex's unlock/lock gap below in rare
4971 		 * case. But the related MDS daemon will just queue
4972 		 * that requests and be still waiting for the client's
4973 		 * reconnection request in up:reconnect state.
4974 		 */
4975 		s = __ceph_lookup_mds_session(mdsc, i);
4976 		if (likely(!s)) {
4977 			s = __open_export_target_session(mdsc, i);
4978 			if (IS_ERR(s)) {
4979 				err = PTR_ERR(s);
4980 				pr_err_client(cl,
4981 					      "failed to open export target session, err %d\n",
4982 					      err);
4983 				continue;
4984 			}
4985 		}
4986 		doutc(cl, "send reconnect to export target mds.%d\n", i);
4987 		mutex_unlock(&mdsc->mutex);
4988 		send_mds_reconnect(mdsc, s);
4989 		ceph_put_mds_session(s);
4990 		mutex_lock(&mdsc->mutex);
4991 	}
4992 
4993 	for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4994 		s = mdsc->sessions[i];
4995 		if (!s)
4996 			continue;
4997 		if (!ceph_mdsmap_is_laggy(newmap, i))
4998 			continue;
4999 		if (s->s_state == CEPH_MDS_SESSION_OPEN ||
5000 		    s->s_state == CEPH_MDS_SESSION_HUNG ||
5001 		    s->s_state == CEPH_MDS_SESSION_CLOSING) {
5002 			doutc(cl, " connecting to export targets of laggy mds%d\n", i);
5003 			__open_export_target_sessions(mdsc, s);
5004 		}
5005 	}
5006 }
5007 
5008 
5009 
5010 /*
5011  * leases
5012  */
5013 
5014 /*
5015  * caller must hold session s_mutex, dentry->d_lock
5016  */
5017 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
5018 {
5019 	struct ceph_dentry_info *di = ceph_dentry(dentry);
5020 
5021 	ceph_put_mds_session(di->lease_session);
5022 	di->lease_session = NULL;
5023 }
5024 
5025 static void handle_lease(struct ceph_mds_client *mdsc,
5026 			 struct ceph_mds_session *session,
5027 			 struct ceph_msg *msg)
5028 {
5029 	struct ceph_client *cl = mdsc->fsc->client;
5030 	struct super_block *sb = mdsc->fsc->sb;
5031 	struct inode *inode;
5032 	struct dentry *parent, *dentry;
5033 	struct ceph_dentry_info *di;
5034 	int mds = session->s_mds;
5035 	struct ceph_mds_lease *h = msg->front.iov_base;
5036 	u32 seq;
5037 	struct ceph_vino vino;
5038 	struct qstr dname;
5039 	int release = 0;
5040 
5041 	doutc(cl, "from mds%d\n", mds);
5042 
5043 	if (!ceph_inc_mds_stopping_blocker(mdsc, session))
5044 		return;
5045 
5046 	/* decode */
5047 	if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
5048 		goto bad;
5049 	vino.ino = le64_to_cpu(h->ino);
5050 	vino.snap = CEPH_NOSNAP;
5051 	seq = le32_to_cpu(h->seq);
5052 	dname.len = get_unaligned_le32(h + 1);
5053 	if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
5054 		goto bad;
5055 	dname.name = (void *)(h + 1) + sizeof(u32);
5056 
5057 	/* lookup inode */
5058 	inode = ceph_find_inode(sb, vino);
5059 	doutc(cl, "%s, ino %llx %p %.*s\n", ceph_lease_op_name(h->action),
5060 	      vino.ino, inode, dname.len, dname.name);
5061 
5062 	mutex_lock(&session->s_mutex);
5063 	if (!inode) {
5064 		doutc(cl, "no inode %llx\n", vino.ino);
5065 		goto release;
5066 	}
5067 
5068 	/* dentry */
5069 	parent = d_find_alias(inode);
5070 	if (!parent) {
5071 		doutc(cl, "no parent dentry on inode %p\n", inode);
5072 		WARN_ON(1);
5073 		goto release;  /* hrm... */
5074 	}
5075 	dname.hash = full_name_hash(parent, dname.name, dname.len);
5076 	dentry = d_lookup(parent, &dname);
5077 	dput(parent);
5078 	if (!dentry)
5079 		goto release;
5080 
5081 	spin_lock(&dentry->d_lock);
5082 	di = ceph_dentry(dentry);
5083 	switch (h->action) {
5084 	case CEPH_MDS_LEASE_REVOKE:
5085 		if (di->lease_session == session) {
5086 			if (ceph_seq_cmp(di->lease_seq, seq) > 0)
5087 				h->seq = cpu_to_le32(di->lease_seq);
5088 			__ceph_mdsc_drop_dentry_lease(dentry);
5089 		}
5090 		release = 1;
5091 		break;
5092 
5093 	case CEPH_MDS_LEASE_RENEW:
5094 		if (di->lease_session == session &&
5095 		    di->lease_gen == atomic_read(&session->s_cap_gen) &&
5096 		    di->lease_renew_from &&
5097 		    di->lease_renew_after == 0) {
5098 			unsigned long duration =
5099 				msecs_to_jiffies(le32_to_cpu(h->duration_ms));
5100 
5101 			di->lease_seq = seq;
5102 			di->time = di->lease_renew_from + duration;
5103 			di->lease_renew_after = di->lease_renew_from +
5104 				(duration >> 1);
5105 			di->lease_renew_from = 0;
5106 		}
5107 		break;
5108 	}
5109 	spin_unlock(&dentry->d_lock);
5110 	dput(dentry);
5111 
5112 	if (!release)
5113 		goto out;
5114 
5115 release:
5116 	/* let's just reuse the same message */
5117 	h->action = CEPH_MDS_LEASE_REVOKE_ACK;
5118 	ceph_msg_get(msg);
5119 	ceph_con_send(&session->s_con, msg);
5120 
5121 out:
5122 	mutex_unlock(&session->s_mutex);
5123 	iput(inode);
5124 
5125 	ceph_dec_mds_stopping_blocker(mdsc);
5126 	return;
5127 
5128 bad:
5129 	ceph_dec_mds_stopping_blocker(mdsc);
5130 
5131 	pr_err_client(cl, "corrupt lease message\n");
5132 	ceph_msg_dump(msg);
5133 }
5134 
5135 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
5136 			      struct dentry *dentry, char action,
5137 			      u32 seq)
5138 {
5139 	struct ceph_client *cl = session->s_mdsc->fsc->client;
5140 	struct ceph_msg *msg;
5141 	struct ceph_mds_lease *lease;
5142 	struct inode *dir;
5143 	int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
5144 
5145 	doutc(cl, "identry %p %s to mds%d\n", dentry, ceph_lease_op_name(action),
5146 	      session->s_mds);
5147 
5148 	msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
5149 	if (!msg)
5150 		return;
5151 	lease = msg->front.iov_base;
5152 	lease->action = action;
5153 	lease->seq = cpu_to_le32(seq);
5154 
5155 	spin_lock(&dentry->d_lock);
5156 	dir = d_inode(dentry->d_parent);
5157 	lease->ino = cpu_to_le64(ceph_ino(dir));
5158 	lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
5159 
5160 	put_unaligned_le32(dentry->d_name.len, lease + 1);
5161 	memcpy((void *)(lease + 1) + 4,
5162 	       dentry->d_name.name, dentry->d_name.len);
5163 	spin_unlock(&dentry->d_lock);
5164 
5165 	ceph_con_send(&session->s_con, msg);
5166 }
5167 
5168 /*
5169  * lock unlock the session, to wait ongoing session activities
5170  */
5171 static void lock_unlock_session(struct ceph_mds_session *s)
5172 {
5173 	mutex_lock(&s->s_mutex);
5174 	mutex_unlock(&s->s_mutex);
5175 }
5176 
5177 static void maybe_recover_session(struct ceph_mds_client *mdsc)
5178 {
5179 	struct ceph_client *cl = mdsc->fsc->client;
5180 	struct ceph_fs_client *fsc = mdsc->fsc;
5181 
5182 	if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
5183 		return;
5184 
5185 	if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
5186 		return;
5187 
5188 	if (!READ_ONCE(fsc->blocklisted))
5189 		return;
5190 
5191 	pr_info_client(cl, "auto reconnect after blocklisted\n");
5192 	ceph_force_reconnect(fsc->sb);
5193 }
5194 
5195 bool check_session_state(struct ceph_mds_session *s)
5196 {
5197 	struct ceph_client *cl = s->s_mdsc->fsc->client;
5198 
5199 	switch (s->s_state) {
5200 	case CEPH_MDS_SESSION_OPEN:
5201 		if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
5202 			s->s_state = CEPH_MDS_SESSION_HUNG;
5203 			pr_info_client(cl, "mds%d hung\n", s->s_mds);
5204 		}
5205 		break;
5206 	case CEPH_MDS_SESSION_CLOSING:
5207 	case CEPH_MDS_SESSION_NEW:
5208 	case CEPH_MDS_SESSION_RESTARTING:
5209 	case CEPH_MDS_SESSION_CLOSED:
5210 	case CEPH_MDS_SESSION_REJECTED:
5211 		return false;
5212 	}
5213 
5214 	return true;
5215 }
5216 
5217 /*
5218  * If the sequence is incremented while we're waiting on a REQUEST_CLOSE reply,
5219  * then we need to retransmit that request.
5220  */
5221 void inc_session_sequence(struct ceph_mds_session *s)
5222 {
5223 	struct ceph_client *cl = s->s_mdsc->fsc->client;
5224 
5225 	lockdep_assert_held(&s->s_mutex);
5226 
5227 	s->s_seq++;
5228 
5229 	if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
5230 		int ret;
5231 
5232 		doutc(cl, "resending session close request for mds%d\n", s->s_mds);
5233 		ret = request_close_session(s);
5234 		if (ret < 0)
5235 			pr_err_client(cl, "unable to close session to mds%d: %d\n",
5236 				      s->s_mds, ret);
5237 	}
5238 }
5239 
5240 /*
5241  * delayed work -- periodically trim expired leases, renew caps with mds.  If
5242  * the @delay parameter is set to 0 or if it's more than 5 secs, the default
5243  * workqueue delay value of 5 secs will be used.
5244  */
5245 static void schedule_delayed(struct ceph_mds_client *mdsc, unsigned long delay)
5246 {
5247 	unsigned long max_delay = HZ * 5;
5248 
5249 	/* 5 secs default delay */
5250 	if (!delay || (delay > max_delay))
5251 		delay = max_delay;
5252 	schedule_delayed_work(&mdsc->delayed_work,
5253 			      round_jiffies_relative(delay));
5254 }
5255 
5256 static void delayed_work(struct work_struct *work)
5257 {
5258 	struct ceph_mds_client *mdsc =
5259 		container_of(work, struct ceph_mds_client, delayed_work.work);
5260 	unsigned long delay;
5261 	int renew_interval;
5262 	int renew_caps;
5263 	int i;
5264 
5265 	doutc(mdsc->fsc->client, "mdsc delayed_work\n");
5266 
5267 	if (mdsc->stopping >= CEPH_MDSC_STOPPING_FLUSHED)
5268 		return;
5269 
5270 	mutex_lock(&mdsc->mutex);
5271 	renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
5272 	renew_caps = time_after_eq(jiffies, HZ*renew_interval +
5273 				   mdsc->last_renew_caps);
5274 	if (renew_caps)
5275 		mdsc->last_renew_caps = jiffies;
5276 
5277 	for (i = 0; i < mdsc->max_sessions; i++) {
5278 		struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
5279 		if (!s)
5280 			continue;
5281 
5282 		if (!check_session_state(s)) {
5283 			ceph_put_mds_session(s);
5284 			continue;
5285 		}
5286 		mutex_unlock(&mdsc->mutex);
5287 
5288 		mutex_lock(&s->s_mutex);
5289 		if (renew_caps)
5290 			send_renew_caps(mdsc, s);
5291 		else
5292 			ceph_con_keepalive(&s->s_con);
5293 		if (s->s_state == CEPH_MDS_SESSION_OPEN ||
5294 		    s->s_state == CEPH_MDS_SESSION_HUNG)
5295 			ceph_send_cap_releases(mdsc, s);
5296 		mutex_unlock(&s->s_mutex);
5297 		ceph_put_mds_session(s);
5298 
5299 		mutex_lock(&mdsc->mutex);
5300 	}
5301 	mutex_unlock(&mdsc->mutex);
5302 
5303 	delay = ceph_check_delayed_caps(mdsc);
5304 
5305 	ceph_queue_cap_reclaim_work(mdsc);
5306 
5307 	ceph_trim_snapid_map(mdsc);
5308 
5309 	maybe_recover_session(mdsc);
5310 
5311 	schedule_delayed(mdsc, delay);
5312 }
5313 
5314 int ceph_mdsc_init(struct ceph_fs_client *fsc)
5315 
5316 {
5317 	struct ceph_mds_client *mdsc;
5318 	int err;
5319 
5320 	mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
5321 	if (!mdsc)
5322 		return -ENOMEM;
5323 	mdsc->fsc = fsc;
5324 	mutex_init(&mdsc->mutex);
5325 	mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
5326 	if (!mdsc->mdsmap) {
5327 		err = -ENOMEM;
5328 		goto err_mdsc;
5329 	}
5330 
5331 	init_completion(&mdsc->safe_umount_waiters);
5332 	spin_lock_init(&mdsc->stopping_lock);
5333 	atomic_set(&mdsc->stopping_blockers, 0);
5334 	init_completion(&mdsc->stopping_waiter);
5335 	init_waitqueue_head(&mdsc->session_close_wq);
5336 	INIT_LIST_HEAD(&mdsc->waiting_for_map);
5337 	mdsc->quotarealms_inodes = RB_ROOT;
5338 	mutex_init(&mdsc->quotarealms_inodes_mutex);
5339 	init_rwsem(&mdsc->snap_rwsem);
5340 	mdsc->snap_realms = RB_ROOT;
5341 	INIT_LIST_HEAD(&mdsc->snap_empty);
5342 	spin_lock_init(&mdsc->snap_empty_lock);
5343 	mdsc->request_tree = RB_ROOT;
5344 	INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
5345 	mdsc->last_renew_caps = jiffies;
5346 	INIT_LIST_HEAD(&mdsc->cap_delay_list);
5347 	INIT_LIST_HEAD(&mdsc->cap_wait_list);
5348 	spin_lock_init(&mdsc->cap_delay_lock);
5349 	INIT_LIST_HEAD(&mdsc->snap_flush_list);
5350 	spin_lock_init(&mdsc->snap_flush_lock);
5351 	mdsc->last_cap_flush_tid = 1;
5352 	INIT_LIST_HEAD(&mdsc->cap_flush_list);
5353 	INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
5354 	spin_lock_init(&mdsc->cap_dirty_lock);
5355 	init_waitqueue_head(&mdsc->cap_flushing_wq);
5356 	INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
5357 	err = ceph_metric_init(&mdsc->metric);
5358 	if (err)
5359 		goto err_mdsmap;
5360 
5361 	spin_lock_init(&mdsc->dentry_list_lock);
5362 	INIT_LIST_HEAD(&mdsc->dentry_leases);
5363 	INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
5364 
5365 	ceph_caps_init(mdsc);
5366 	ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
5367 
5368 	spin_lock_init(&mdsc->snapid_map_lock);
5369 	mdsc->snapid_map_tree = RB_ROOT;
5370 	INIT_LIST_HEAD(&mdsc->snapid_map_lru);
5371 
5372 	init_rwsem(&mdsc->pool_perm_rwsem);
5373 	mdsc->pool_perm_tree = RB_ROOT;
5374 
5375 	strscpy(mdsc->nodename, utsname()->nodename,
5376 		sizeof(mdsc->nodename));
5377 
5378 	fsc->mdsc = mdsc;
5379 	return 0;
5380 
5381 err_mdsmap:
5382 	kfree(mdsc->mdsmap);
5383 err_mdsc:
5384 	kfree(mdsc);
5385 	return err;
5386 }
5387 
5388 /*
5389  * Wait for safe replies on open mds requests.  If we time out, drop
5390  * all requests from the tree to avoid dangling dentry refs.
5391  */
5392 static void wait_requests(struct ceph_mds_client *mdsc)
5393 {
5394 	struct ceph_client *cl = mdsc->fsc->client;
5395 	struct ceph_options *opts = mdsc->fsc->client->options;
5396 	struct ceph_mds_request *req;
5397 
5398 	mutex_lock(&mdsc->mutex);
5399 	if (__get_oldest_req(mdsc)) {
5400 		mutex_unlock(&mdsc->mutex);
5401 
5402 		doutc(cl, "waiting for requests\n");
5403 		wait_for_completion_timeout(&mdsc->safe_umount_waiters,
5404 				    ceph_timeout_jiffies(opts->mount_timeout));
5405 
5406 		/* tear down remaining requests */
5407 		mutex_lock(&mdsc->mutex);
5408 		while ((req = __get_oldest_req(mdsc))) {
5409 			doutc(cl, "timed out on tid %llu\n", req->r_tid);
5410 			list_del_init(&req->r_wait);
5411 			__unregister_request(mdsc, req);
5412 		}
5413 	}
5414 	mutex_unlock(&mdsc->mutex);
5415 	doutc(cl, "done\n");
5416 }
5417 
5418 void send_flush_mdlog(struct ceph_mds_session *s)
5419 {
5420 	struct ceph_client *cl = s->s_mdsc->fsc->client;
5421 	struct ceph_msg *msg;
5422 
5423 	/*
5424 	 * Pre-luminous MDS crashes when it sees an unknown session request
5425 	 */
5426 	if (!CEPH_HAVE_FEATURE(s->s_con.peer_features, SERVER_LUMINOUS))
5427 		return;
5428 
5429 	mutex_lock(&s->s_mutex);
5430 	doutc(cl, "request mdlog flush to mds%d (%s)s seq %lld\n",
5431 	      s->s_mds, ceph_session_state_name(s->s_state), s->s_seq);
5432 	msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_FLUSH_MDLOG,
5433 				      s->s_seq);
5434 	if (!msg) {
5435 		pr_err_client(cl, "failed to request mdlog flush to mds%d (%s) seq %lld\n",
5436 			      s->s_mds, ceph_session_state_name(s->s_state), s->s_seq);
5437 	} else {
5438 		ceph_con_send(&s->s_con, msg);
5439 	}
5440 	mutex_unlock(&s->s_mutex);
5441 }
5442 
5443 /*
5444  * called before mount is ro, and before dentries are torn down.
5445  * (hmm, does this still race with new lookups?)
5446  */
5447 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
5448 {
5449 	doutc(mdsc->fsc->client, "begin\n");
5450 	mdsc->stopping = CEPH_MDSC_STOPPING_BEGIN;
5451 
5452 	ceph_mdsc_iterate_sessions(mdsc, send_flush_mdlog, true);
5453 	ceph_mdsc_iterate_sessions(mdsc, lock_unlock_session, false);
5454 	ceph_flush_dirty_caps(mdsc);
5455 	wait_requests(mdsc);
5456 
5457 	/*
5458 	 * wait for reply handlers to drop their request refs and
5459 	 * their inode/dcache refs
5460 	 */
5461 	ceph_msgr_flush();
5462 
5463 	ceph_cleanup_quotarealms_inodes(mdsc);
5464 	doutc(mdsc->fsc->client, "done\n");
5465 }
5466 
5467 /*
5468  * flush the mdlog and wait for all write mds requests to flush.
5469  */
5470 static void flush_mdlog_and_wait_mdsc_unsafe_requests(struct ceph_mds_client *mdsc,
5471 						 u64 want_tid)
5472 {
5473 	struct ceph_client *cl = mdsc->fsc->client;
5474 	struct ceph_mds_request *req = NULL, *nextreq;
5475 	struct ceph_mds_session *last_session = NULL;
5476 	struct rb_node *n;
5477 
5478 	mutex_lock(&mdsc->mutex);
5479 	doutc(cl, "want %lld\n", want_tid);
5480 restart:
5481 	req = __get_oldest_req(mdsc);
5482 	while (req && req->r_tid <= want_tid) {
5483 		/* find next request */
5484 		n = rb_next(&req->r_node);
5485 		if (n)
5486 			nextreq = rb_entry(n, struct ceph_mds_request, r_node);
5487 		else
5488 			nextreq = NULL;
5489 		if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
5490 		    (req->r_op & CEPH_MDS_OP_WRITE)) {
5491 			struct ceph_mds_session *s = req->r_session;
5492 
5493 			if (!s) {
5494 				req = nextreq;
5495 				continue;
5496 			}
5497 
5498 			/* write op */
5499 			ceph_mdsc_get_request(req);
5500 			if (nextreq)
5501 				ceph_mdsc_get_request(nextreq);
5502 			s = ceph_get_mds_session(s);
5503 			mutex_unlock(&mdsc->mutex);
5504 
5505 			/* send flush mdlog request to MDS */
5506 			if (last_session != s) {
5507 				send_flush_mdlog(s);
5508 				ceph_put_mds_session(last_session);
5509 				last_session = s;
5510 			} else {
5511 				ceph_put_mds_session(s);
5512 			}
5513 			doutc(cl, "wait on %llu (want %llu)\n",
5514 			      req->r_tid, want_tid);
5515 			wait_for_completion(&req->r_safe_completion);
5516 
5517 			mutex_lock(&mdsc->mutex);
5518 			ceph_mdsc_put_request(req);
5519 			if (!nextreq)
5520 				break;  /* next dne before, so we're done! */
5521 			if (RB_EMPTY_NODE(&nextreq->r_node)) {
5522 				/* next request was removed from tree */
5523 				ceph_mdsc_put_request(nextreq);
5524 				goto restart;
5525 			}
5526 			ceph_mdsc_put_request(nextreq);  /* won't go away */
5527 		}
5528 		req = nextreq;
5529 	}
5530 	mutex_unlock(&mdsc->mutex);
5531 	ceph_put_mds_session(last_session);
5532 	doutc(cl, "done\n");
5533 }
5534 
5535 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
5536 {
5537 	struct ceph_client *cl = mdsc->fsc->client;
5538 	u64 want_tid, want_flush;
5539 
5540 	if (READ_ONCE(mdsc->fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN)
5541 		return;
5542 
5543 	doutc(cl, "sync\n");
5544 	mutex_lock(&mdsc->mutex);
5545 	want_tid = mdsc->last_tid;
5546 	mutex_unlock(&mdsc->mutex);
5547 
5548 	ceph_flush_dirty_caps(mdsc);
5549 	spin_lock(&mdsc->cap_dirty_lock);
5550 	want_flush = mdsc->last_cap_flush_tid;
5551 	if (!list_empty(&mdsc->cap_flush_list)) {
5552 		struct ceph_cap_flush *cf =
5553 			list_last_entry(&mdsc->cap_flush_list,
5554 					struct ceph_cap_flush, g_list);
5555 		cf->wake = true;
5556 	}
5557 	spin_unlock(&mdsc->cap_dirty_lock);
5558 
5559 	doutc(cl, "sync want tid %lld flush_seq %lld\n", want_tid, want_flush);
5560 
5561 	flush_mdlog_and_wait_mdsc_unsafe_requests(mdsc, want_tid);
5562 	wait_caps_flush(mdsc, want_flush);
5563 }
5564 
5565 /*
5566  * true if all sessions are closed, or we force unmount
5567  */
5568 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
5569 {
5570 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
5571 		return true;
5572 	return atomic_read(&mdsc->num_sessions) <= skipped;
5573 }
5574 
5575 /*
5576  * called after sb is ro or when metadata corrupted.
5577  */
5578 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
5579 {
5580 	struct ceph_options *opts = mdsc->fsc->client->options;
5581 	struct ceph_client *cl = mdsc->fsc->client;
5582 	struct ceph_mds_session *session;
5583 	int i;
5584 	int skipped = 0;
5585 
5586 	doutc(cl, "begin\n");
5587 
5588 	/* close sessions */
5589 	mutex_lock(&mdsc->mutex);
5590 	for (i = 0; i < mdsc->max_sessions; i++) {
5591 		session = __ceph_lookup_mds_session(mdsc, i);
5592 		if (!session)
5593 			continue;
5594 		mutex_unlock(&mdsc->mutex);
5595 		mutex_lock(&session->s_mutex);
5596 		if (__close_session(mdsc, session) <= 0)
5597 			skipped++;
5598 		mutex_unlock(&session->s_mutex);
5599 		ceph_put_mds_session(session);
5600 		mutex_lock(&mdsc->mutex);
5601 	}
5602 	mutex_unlock(&mdsc->mutex);
5603 
5604 	doutc(cl, "waiting for sessions to close\n");
5605 	wait_event_timeout(mdsc->session_close_wq,
5606 			   done_closing_sessions(mdsc, skipped),
5607 			   ceph_timeout_jiffies(opts->mount_timeout));
5608 
5609 	/* tear down remaining sessions */
5610 	mutex_lock(&mdsc->mutex);
5611 	for (i = 0; i < mdsc->max_sessions; i++) {
5612 		if (mdsc->sessions[i]) {
5613 			session = ceph_get_mds_session(mdsc->sessions[i]);
5614 			__unregister_session(mdsc, session);
5615 			mutex_unlock(&mdsc->mutex);
5616 			mutex_lock(&session->s_mutex);
5617 			remove_session_caps(session);
5618 			mutex_unlock(&session->s_mutex);
5619 			ceph_put_mds_session(session);
5620 			mutex_lock(&mdsc->mutex);
5621 		}
5622 	}
5623 	WARN_ON(!list_empty(&mdsc->cap_delay_list));
5624 	mutex_unlock(&mdsc->mutex);
5625 
5626 	ceph_cleanup_snapid_map(mdsc);
5627 	ceph_cleanup_global_and_empty_realms(mdsc);
5628 
5629 	cancel_work_sync(&mdsc->cap_reclaim_work);
5630 	cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
5631 
5632 	doutc(cl, "done\n");
5633 }
5634 
5635 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
5636 {
5637 	struct ceph_mds_session *session;
5638 	int mds;
5639 
5640 	doutc(mdsc->fsc->client, "force umount\n");
5641 
5642 	mutex_lock(&mdsc->mutex);
5643 	for (mds = 0; mds < mdsc->max_sessions; mds++) {
5644 		session = __ceph_lookup_mds_session(mdsc, mds);
5645 		if (!session)
5646 			continue;
5647 
5648 		if (session->s_state == CEPH_MDS_SESSION_REJECTED)
5649 			__unregister_session(mdsc, session);
5650 		__wake_requests(mdsc, &session->s_waiting);
5651 		mutex_unlock(&mdsc->mutex);
5652 
5653 		mutex_lock(&session->s_mutex);
5654 		__close_session(mdsc, session);
5655 		if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
5656 			cleanup_session_requests(mdsc, session);
5657 			remove_session_caps(session);
5658 		}
5659 		mutex_unlock(&session->s_mutex);
5660 		ceph_put_mds_session(session);
5661 
5662 		mutex_lock(&mdsc->mutex);
5663 		kick_requests(mdsc, mds);
5664 	}
5665 	__wake_requests(mdsc, &mdsc->waiting_for_map);
5666 	mutex_unlock(&mdsc->mutex);
5667 }
5668 
5669 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
5670 {
5671 	doutc(mdsc->fsc->client, "stop\n");
5672 	/*
5673 	 * Make sure the delayed work stopped before releasing
5674 	 * the resources.
5675 	 *
5676 	 * Because the cancel_delayed_work_sync() will only
5677 	 * guarantee that the work finishes executing. But the
5678 	 * delayed work will re-arm itself again after that.
5679 	 */
5680 	flush_delayed_work(&mdsc->delayed_work);
5681 
5682 	if (mdsc->mdsmap)
5683 		ceph_mdsmap_destroy(mdsc->mdsmap);
5684 	kfree(mdsc->sessions);
5685 	ceph_caps_finalize(mdsc);
5686 	ceph_pool_perm_destroy(mdsc);
5687 }
5688 
5689 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
5690 {
5691 	struct ceph_mds_client *mdsc = fsc->mdsc;
5692 	doutc(fsc->client, "%p\n", mdsc);
5693 
5694 	if (!mdsc)
5695 		return;
5696 
5697 	/* flush out any connection work with references to us */
5698 	ceph_msgr_flush();
5699 
5700 	ceph_mdsc_stop(mdsc);
5701 
5702 	ceph_metric_destroy(&mdsc->metric);
5703 
5704 	fsc->mdsc = NULL;
5705 	kfree(mdsc);
5706 	doutc(fsc->client, "%p done\n", mdsc);
5707 }
5708 
5709 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
5710 {
5711 	struct ceph_fs_client *fsc = mdsc->fsc;
5712 	struct ceph_client *cl = fsc->client;
5713 	const char *mds_namespace = fsc->mount_options->mds_namespace;
5714 	void *p = msg->front.iov_base;
5715 	void *end = p + msg->front.iov_len;
5716 	u32 epoch;
5717 	u32 num_fs;
5718 	u32 mount_fscid = (u32)-1;
5719 	int err = -EINVAL;
5720 
5721 	ceph_decode_need(&p, end, sizeof(u32), bad);
5722 	epoch = ceph_decode_32(&p);
5723 
5724 	doutc(cl, "epoch %u\n", epoch);
5725 
5726 	/* struct_v, struct_cv, map_len, epoch, legacy_client_fscid */
5727 	ceph_decode_skip_n(&p, end, 2 + sizeof(u32) * 3, bad);
5728 
5729 	ceph_decode_32_safe(&p, end, num_fs, bad);
5730 	while (num_fs-- > 0) {
5731 		void *info_p, *info_end;
5732 		u32 info_len;
5733 		u32 fscid, namelen;
5734 
5735 		ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
5736 		p += 2;		// info_v, info_cv
5737 		info_len = ceph_decode_32(&p);
5738 		ceph_decode_need(&p, end, info_len, bad);
5739 		info_p = p;
5740 		info_end = p + info_len;
5741 		p = info_end;
5742 
5743 		ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
5744 		fscid = ceph_decode_32(&info_p);
5745 		namelen = ceph_decode_32(&info_p);
5746 		ceph_decode_need(&info_p, info_end, namelen, bad);
5747 
5748 		if (mds_namespace &&
5749 		    strlen(mds_namespace) == namelen &&
5750 		    !strncmp(mds_namespace, (char *)info_p, namelen)) {
5751 			mount_fscid = fscid;
5752 			break;
5753 		}
5754 	}
5755 
5756 	ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
5757 	if (mount_fscid != (u32)-1) {
5758 		fsc->client->monc.fs_cluster_id = mount_fscid;
5759 		ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
5760 				   0, true);
5761 		ceph_monc_renew_subs(&fsc->client->monc);
5762 	} else {
5763 		err = -ENOENT;
5764 		goto err_out;
5765 	}
5766 	return;
5767 
5768 bad:
5769 	pr_err_client(cl, "error decoding fsmap %d. Shutting down mount.\n",
5770 		      err);
5771 	ceph_umount_begin(mdsc->fsc->sb);
5772 	ceph_msg_dump(msg);
5773 err_out:
5774 	mutex_lock(&mdsc->mutex);
5775 	mdsc->mdsmap_err = err;
5776 	__wake_requests(mdsc, &mdsc->waiting_for_map);
5777 	mutex_unlock(&mdsc->mutex);
5778 }
5779 
5780 /*
5781  * handle mds map update.
5782  */
5783 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
5784 {
5785 	struct ceph_client *cl = mdsc->fsc->client;
5786 	u32 epoch;
5787 	u32 maplen;
5788 	void *p = msg->front.iov_base;
5789 	void *end = p + msg->front.iov_len;
5790 	struct ceph_mdsmap *newmap, *oldmap;
5791 	struct ceph_fsid fsid;
5792 	int err = -EINVAL;
5793 
5794 	ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
5795 	ceph_decode_copy(&p, &fsid, sizeof(fsid));
5796 	if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
5797 		return;
5798 	epoch = ceph_decode_32(&p);
5799 	maplen = ceph_decode_32(&p);
5800 	doutc(cl, "epoch %u len %d\n", epoch, (int)maplen);
5801 
5802 	/* do we need it? */
5803 	mutex_lock(&mdsc->mutex);
5804 	if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
5805 		doutc(cl, "epoch %u <= our %u\n", epoch, mdsc->mdsmap->m_epoch);
5806 		mutex_unlock(&mdsc->mutex);
5807 		return;
5808 	}
5809 
5810 	newmap = ceph_mdsmap_decode(mdsc, &p, end, ceph_msgr2(mdsc->fsc->client));
5811 	if (IS_ERR(newmap)) {
5812 		err = PTR_ERR(newmap);
5813 		goto bad_unlock;
5814 	}
5815 
5816 	/* swap into place */
5817 	if (mdsc->mdsmap) {
5818 		oldmap = mdsc->mdsmap;
5819 		mdsc->mdsmap = newmap;
5820 		check_new_map(mdsc, newmap, oldmap);
5821 		ceph_mdsmap_destroy(oldmap);
5822 	} else {
5823 		mdsc->mdsmap = newmap;  /* first mds map */
5824 	}
5825 	mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
5826 					MAX_LFS_FILESIZE);
5827 
5828 	__wake_requests(mdsc, &mdsc->waiting_for_map);
5829 	ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
5830 			  mdsc->mdsmap->m_epoch);
5831 
5832 	mutex_unlock(&mdsc->mutex);
5833 	schedule_delayed(mdsc, 0);
5834 	return;
5835 
5836 bad_unlock:
5837 	mutex_unlock(&mdsc->mutex);
5838 bad:
5839 	pr_err_client(cl, "error decoding mdsmap %d. Shutting down mount.\n",
5840 		      err);
5841 	ceph_umount_begin(mdsc->fsc->sb);
5842 	ceph_msg_dump(msg);
5843 	return;
5844 }
5845 
5846 static struct ceph_connection *mds_get_con(struct ceph_connection *con)
5847 {
5848 	struct ceph_mds_session *s = con->private;
5849 
5850 	if (ceph_get_mds_session(s))
5851 		return con;
5852 	return NULL;
5853 }
5854 
5855 static void mds_put_con(struct ceph_connection *con)
5856 {
5857 	struct ceph_mds_session *s = con->private;
5858 
5859 	ceph_put_mds_session(s);
5860 }
5861 
5862 /*
5863  * if the client is unresponsive for long enough, the mds will kill
5864  * the session entirely.
5865  */
5866 static void mds_peer_reset(struct ceph_connection *con)
5867 {
5868 	struct ceph_mds_session *s = con->private;
5869 	struct ceph_mds_client *mdsc = s->s_mdsc;
5870 
5871 	pr_warn_client(mdsc->fsc->client, "mds%d closed our session\n",
5872 		       s->s_mds);
5873 	if (READ_ONCE(mdsc->fsc->mount_state) != CEPH_MOUNT_FENCE_IO)
5874 		send_mds_reconnect(mdsc, s);
5875 }
5876 
5877 static void mds_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5878 {
5879 	struct ceph_mds_session *s = con->private;
5880 	struct ceph_mds_client *mdsc = s->s_mdsc;
5881 	struct ceph_client *cl = mdsc->fsc->client;
5882 	int type = le16_to_cpu(msg->hdr.type);
5883 
5884 	mutex_lock(&mdsc->mutex);
5885 	if (__verify_registered_session(mdsc, s) < 0) {
5886 		mutex_unlock(&mdsc->mutex);
5887 		goto out;
5888 	}
5889 	mutex_unlock(&mdsc->mutex);
5890 
5891 	switch (type) {
5892 	case CEPH_MSG_MDS_MAP:
5893 		ceph_mdsc_handle_mdsmap(mdsc, msg);
5894 		break;
5895 	case CEPH_MSG_FS_MAP_USER:
5896 		ceph_mdsc_handle_fsmap(mdsc, msg);
5897 		break;
5898 	case CEPH_MSG_CLIENT_SESSION:
5899 		handle_session(s, msg);
5900 		break;
5901 	case CEPH_MSG_CLIENT_REPLY:
5902 		handle_reply(s, msg);
5903 		break;
5904 	case CEPH_MSG_CLIENT_REQUEST_FORWARD:
5905 		handle_forward(mdsc, s, msg);
5906 		break;
5907 	case CEPH_MSG_CLIENT_CAPS:
5908 		ceph_handle_caps(s, msg);
5909 		break;
5910 	case CEPH_MSG_CLIENT_SNAP:
5911 		ceph_handle_snap(mdsc, s, msg);
5912 		break;
5913 	case CEPH_MSG_CLIENT_LEASE:
5914 		handle_lease(mdsc, s, msg);
5915 		break;
5916 	case CEPH_MSG_CLIENT_QUOTA:
5917 		ceph_handle_quota(mdsc, s, msg);
5918 		break;
5919 
5920 	default:
5921 		pr_err_client(cl, "received unknown message type %d %s\n",
5922 			      type, ceph_msg_type_name(type));
5923 	}
5924 out:
5925 	ceph_msg_put(msg);
5926 }
5927 
5928 /*
5929  * authentication
5930  */
5931 
5932 /*
5933  * Note: returned pointer is the address of a structure that's
5934  * managed separately.  Caller must *not* attempt to free it.
5935  */
5936 static struct ceph_auth_handshake *
5937 mds_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5938 {
5939 	struct ceph_mds_session *s = con->private;
5940 	struct ceph_mds_client *mdsc = s->s_mdsc;
5941 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5942 	struct ceph_auth_handshake *auth = &s->s_auth;
5943 	int ret;
5944 
5945 	ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5946 					 force_new, proto, NULL, NULL);
5947 	if (ret)
5948 		return ERR_PTR(ret);
5949 
5950 	return auth;
5951 }
5952 
5953 static int mds_add_authorizer_challenge(struct ceph_connection *con,
5954 				    void *challenge_buf, int challenge_buf_len)
5955 {
5956 	struct ceph_mds_session *s = con->private;
5957 	struct ceph_mds_client *mdsc = s->s_mdsc;
5958 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5959 
5960 	return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
5961 					    challenge_buf, challenge_buf_len);
5962 }
5963 
5964 static int mds_verify_authorizer_reply(struct ceph_connection *con)
5965 {
5966 	struct ceph_mds_session *s = con->private;
5967 	struct ceph_mds_client *mdsc = s->s_mdsc;
5968 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5969 	struct ceph_auth_handshake *auth = &s->s_auth;
5970 
5971 	return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5972 		auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5973 		NULL, NULL, NULL, NULL);
5974 }
5975 
5976 static int mds_invalidate_authorizer(struct ceph_connection *con)
5977 {
5978 	struct ceph_mds_session *s = con->private;
5979 	struct ceph_mds_client *mdsc = s->s_mdsc;
5980 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5981 
5982 	ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
5983 
5984 	return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
5985 }
5986 
5987 static int mds_get_auth_request(struct ceph_connection *con,
5988 				void *buf, int *buf_len,
5989 				void **authorizer, int *authorizer_len)
5990 {
5991 	struct ceph_mds_session *s = con->private;
5992 	struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5993 	struct ceph_auth_handshake *auth = &s->s_auth;
5994 	int ret;
5995 
5996 	ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5997 				       buf, buf_len);
5998 	if (ret)
5999 		return ret;
6000 
6001 	*authorizer = auth->authorizer_buf;
6002 	*authorizer_len = auth->authorizer_buf_len;
6003 	return 0;
6004 }
6005 
6006 static int mds_handle_auth_reply_more(struct ceph_connection *con,
6007 				      void *reply, int reply_len,
6008 				      void *buf, int *buf_len,
6009 				      void **authorizer, int *authorizer_len)
6010 {
6011 	struct ceph_mds_session *s = con->private;
6012 	struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
6013 	struct ceph_auth_handshake *auth = &s->s_auth;
6014 	int ret;
6015 
6016 	ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
6017 					      buf, buf_len);
6018 	if (ret)
6019 		return ret;
6020 
6021 	*authorizer = auth->authorizer_buf;
6022 	*authorizer_len = auth->authorizer_buf_len;
6023 	return 0;
6024 }
6025 
6026 static int mds_handle_auth_done(struct ceph_connection *con,
6027 				u64 global_id, void *reply, int reply_len,
6028 				u8 *session_key, int *session_key_len,
6029 				u8 *con_secret, int *con_secret_len)
6030 {
6031 	struct ceph_mds_session *s = con->private;
6032 	struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
6033 	struct ceph_auth_handshake *auth = &s->s_auth;
6034 
6035 	return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
6036 					       session_key, session_key_len,
6037 					       con_secret, con_secret_len);
6038 }
6039 
6040 static int mds_handle_auth_bad_method(struct ceph_connection *con,
6041 				      int used_proto, int result,
6042 				      const int *allowed_protos, int proto_cnt,
6043 				      const int *allowed_modes, int mode_cnt)
6044 {
6045 	struct ceph_mds_session *s = con->private;
6046 	struct ceph_mon_client *monc = &s->s_mdsc->fsc->client->monc;
6047 	int ret;
6048 
6049 	if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_MDS,
6050 					    used_proto, result,
6051 					    allowed_protos, proto_cnt,
6052 					    allowed_modes, mode_cnt)) {
6053 		ret = ceph_monc_validate_auth(monc);
6054 		if (ret)
6055 			return ret;
6056 	}
6057 
6058 	return -EACCES;
6059 }
6060 
6061 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
6062 				struct ceph_msg_header *hdr, int *skip)
6063 {
6064 	struct ceph_msg *msg;
6065 	int type = (int) le16_to_cpu(hdr->type);
6066 	int front_len = (int) le32_to_cpu(hdr->front_len);
6067 
6068 	if (con->in_msg)
6069 		return con->in_msg;
6070 
6071 	*skip = 0;
6072 	msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
6073 	if (!msg) {
6074 		pr_err("unable to allocate msg type %d len %d\n",
6075 		       type, front_len);
6076 		return NULL;
6077 	}
6078 
6079 	return msg;
6080 }
6081 
6082 static int mds_sign_message(struct ceph_msg *msg)
6083 {
6084        struct ceph_mds_session *s = msg->con->private;
6085        struct ceph_auth_handshake *auth = &s->s_auth;
6086 
6087        return ceph_auth_sign_message(auth, msg);
6088 }
6089 
6090 static int mds_check_message_signature(struct ceph_msg *msg)
6091 {
6092        struct ceph_mds_session *s = msg->con->private;
6093        struct ceph_auth_handshake *auth = &s->s_auth;
6094 
6095        return ceph_auth_check_message_signature(auth, msg);
6096 }
6097 
6098 static const struct ceph_connection_operations mds_con_ops = {
6099 	.get = mds_get_con,
6100 	.put = mds_put_con,
6101 	.alloc_msg = mds_alloc_msg,
6102 	.dispatch = mds_dispatch,
6103 	.peer_reset = mds_peer_reset,
6104 	.get_authorizer = mds_get_authorizer,
6105 	.add_authorizer_challenge = mds_add_authorizer_challenge,
6106 	.verify_authorizer_reply = mds_verify_authorizer_reply,
6107 	.invalidate_authorizer = mds_invalidate_authorizer,
6108 	.sign_message = mds_sign_message,
6109 	.check_message_signature = mds_check_message_signature,
6110 	.get_auth_request = mds_get_auth_request,
6111 	.handle_auth_reply_more = mds_handle_auth_reply_more,
6112 	.handle_auth_done = mds_handle_auth_done,
6113 	.handle_auth_bad_method = mds_handle_auth_bad_method,
6114 };
6115 
6116 /* eof */
6117