xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 36fe4655)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "sysfs.h"
67 #include "nfs4idmap.h"
68 #include "nfs4session.h"
69 #include "fscache.h"
70 #include "nfs42.h"
71 
72 #include "nfs4trace.h"
73 
74 #ifdef CONFIG_NFS_V4_2
75 #include "nfs42.h"
76 #endif /* CONFIG_NFS_V4_2 */
77 
78 #define NFSDBG_FACILITY		NFSDBG_PROC
79 
80 #define NFS4_BITMASK_SZ		3
81 
82 #define NFS4_POLL_RETRY_MIN	(HZ/10)
83 #define NFS4_POLL_RETRY_MAX	(15*HZ)
84 
85 /* file attributes which can be mapped to nfs attributes */
86 #define NFS4_VALID_ATTRS (ATTR_MODE \
87 	| ATTR_UID \
88 	| ATTR_GID \
89 	| ATTR_SIZE \
90 	| ATTR_ATIME \
91 	| ATTR_MTIME \
92 	| ATTR_CTIME \
93 	| ATTR_ATIME_SET \
94 	| ATTR_MTIME_SET)
95 
96 struct nfs4_opendata;
97 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
98 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
99 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
100 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
101 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
102 			    struct nfs_fattr *fattr, struct iattr *sattr,
103 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel,
104 			    struct nfs4_label *olabel);
105 #ifdef CONFIG_NFS_V4_1
106 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
107 		const struct cred *cred,
108 		struct nfs4_slot *slot,
109 		bool is_privileged);
110 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
111 		const struct cred *);
112 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
113 		const struct cred *, bool);
114 #endif
115 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
116 		struct nfs_server *server,
117 		struct nfs4_label *label);
118 
119 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
120 static inline struct nfs4_label *
121 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
122 	struct iattr *sattr, struct nfs4_label *label)
123 {
124 	int err;
125 
126 	if (label == NULL)
127 		return NULL;
128 
129 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
130 		return NULL;
131 
132 	err = security_dentry_init_security(dentry, sattr->ia_mode,
133 				&dentry->d_name, (void **)&label->label, &label->len);
134 	if (err == 0)
135 		return label;
136 
137 	return NULL;
138 }
139 static inline void
140 nfs4_label_release_security(struct nfs4_label *label)
141 {
142 	if (label)
143 		security_release_secctx(label->label, label->len);
144 }
145 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
146 {
147 	if (label)
148 		return server->attr_bitmask;
149 
150 	return server->attr_bitmask_nl;
151 }
152 #else
153 static inline struct nfs4_label *
154 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
155 	struct iattr *sattr, struct nfs4_label *l)
156 { return NULL; }
157 static inline void
158 nfs4_label_release_security(struct nfs4_label *label)
159 { return; }
160 static inline u32 *
161 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
162 { return server->attr_bitmask; }
163 #endif
164 
165 /* Prevent leaks of NFSv4 errors into userland */
166 static int nfs4_map_errors(int err)
167 {
168 	if (err >= -1000)
169 		return err;
170 	switch (err) {
171 	case -NFS4ERR_RESOURCE:
172 	case -NFS4ERR_LAYOUTTRYLATER:
173 	case -NFS4ERR_RECALLCONFLICT:
174 		return -EREMOTEIO;
175 	case -NFS4ERR_WRONGSEC:
176 	case -NFS4ERR_WRONG_CRED:
177 		return -EPERM;
178 	case -NFS4ERR_BADOWNER:
179 	case -NFS4ERR_BADNAME:
180 		return -EINVAL;
181 	case -NFS4ERR_SHARE_DENIED:
182 		return -EACCES;
183 	case -NFS4ERR_MINOR_VERS_MISMATCH:
184 		return -EPROTONOSUPPORT;
185 	case -NFS4ERR_FILE_OPEN:
186 		return -EBUSY;
187 	case -NFS4ERR_NOT_SAME:
188 		return -ENOTSYNC;
189 	default:
190 		dprintk("%s could not handle NFSv4 error %d\n",
191 				__func__, -err);
192 		break;
193 	}
194 	return -EIO;
195 }
196 
197 /*
198  * This is our standard bitmap for GETATTR requests.
199  */
200 const u32 nfs4_fattr_bitmap[3] = {
201 	FATTR4_WORD0_TYPE
202 	| FATTR4_WORD0_CHANGE
203 	| FATTR4_WORD0_SIZE
204 	| FATTR4_WORD0_FSID
205 	| FATTR4_WORD0_FILEID,
206 	FATTR4_WORD1_MODE
207 	| FATTR4_WORD1_NUMLINKS
208 	| FATTR4_WORD1_OWNER
209 	| FATTR4_WORD1_OWNER_GROUP
210 	| FATTR4_WORD1_RAWDEV
211 	| FATTR4_WORD1_SPACE_USED
212 	| FATTR4_WORD1_TIME_ACCESS
213 	| FATTR4_WORD1_TIME_METADATA
214 	| FATTR4_WORD1_TIME_MODIFY
215 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
216 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
217 	FATTR4_WORD2_SECURITY_LABEL
218 #endif
219 };
220 
221 static const u32 nfs4_pnfs_open_bitmap[3] = {
222 	FATTR4_WORD0_TYPE
223 	| FATTR4_WORD0_CHANGE
224 	| FATTR4_WORD0_SIZE
225 	| FATTR4_WORD0_FSID
226 	| FATTR4_WORD0_FILEID,
227 	FATTR4_WORD1_MODE
228 	| FATTR4_WORD1_NUMLINKS
229 	| FATTR4_WORD1_OWNER
230 	| FATTR4_WORD1_OWNER_GROUP
231 	| FATTR4_WORD1_RAWDEV
232 	| FATTR4_WORD1_SPACE_USED
233 	| FATTR4_WORD1_TIME_ACCESS
234 	| FATTR4_WORD1_TIME_METADATA
235 	| FATTR4_WORD1_TIME_MODIFY,
236 	FATTR4_WORD2_MDSTHRESHOLD
237 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
238 	| FATTR4_WORD2_SECURITY_LABEL
239 #endif
240 };
241 
242 static const u32 nfs4_open_noattr_bitmap[3] = {
243 	FATTR4_WORD0_TYPE
244 	| FATTR4_WORD0_FILEID,
245 };
246 
247 const u32 nfs4_statfs_bitmap[3] = {
248 	FATTR4_WORD0_FILES_AVAIL
249 	| FATTR4_WORD0_FILES_FREE
250 	| FATTR4_WORD0_FILES_TOTAL,
251 	FATTR4_WORD1_SPACE_AVAIL
252 	| FATTR4_WORD1_SPACE_FREE
253 	| FATTR4_WORD1_SPACE_TOTAL
254 };
255 
256 const u32 nfs4_pathconf_bitmap[3] = {
257 	FATTR4_WORD0_MAXLINK
258 	| FATTR4_WORD0_MAXNAME,
259 	0
260 };
261 
262 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
263 			| FATTR4_WORD0_MAXREAD
264 			| FATTR4_WORD0_MAXWRITE
265 			| FATTR4_WORD0_LEASE_TIME,
266 			FATTR4_WORD1_TIME_DELTA
267 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
268 			FATTR4_WORD2_LAYOUT_BLKSIZE
269 			| FATTR4_WORD2_CLONE_BLKSIZE
270 			| FATTR4_WORD2_XATTR_SUPPORT
271 };
272 
273 const u32 nfs4_fs_locations_bitmap[3] = {
274 	FATTR4_WORD0_CHANGE
275 	| FATTR4_WORD0_SIZE
276 	| FATTR4_WORD0_FSID
277 	| FATTR4_WORD0_FILEID
278 	| FATTR4_WORD0_FS_LOCATIONS,
279 	FATTR4_WORD1_OWNER
280 	| FATTR4_WORD1_OWNER_GROUP
281 	| FATTR4_WORD1_RAWDEV
282 	| FATTR4_WORD1_SPACE_USED
283 	| FATTR4_WORD1_TIME_ACCESS
284 	| FATTR4_WORD1_TIME_METADATA
285 	| FATTR4_WORD1_TIME_MODIFY
286 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
287 };
288 
289 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
290 		struct inode *inode)
291 {
292 	unsigned long cache_validity;
293 
294 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
295 	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
296 		return;
297 
298 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
299 	if (!(cache_validity & NFS_INO_REVAL_FORCED))
300 		cache_validity &= ~(NFS_INO_INVALID_CHANGE
301 				| NFS_INO_INVALID_SIZE);
302 
303 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
304 		dst[0] &= ~FATTR4_WORD0_SIZE;
305 
306 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
307 		dst[0] &= ~FATTR4_WORD0_CHANGE;
308 }
309 
310 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
311 		const __u32 *src, struct inode *inode)
312 {
313 	nfs4_bitmap_copy_adjust(dst, src, inode);
314 }
315 
316 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
317 		struct nfs4_readdir_arg *readdir)
318 {
319 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
320 	__be32 *start, *p;
321 
322 	if (cookie > 2) {
323 		readdir->cookie = cookie;
324 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
325 		return;
326 	}
327 
328 	readdir->cookie = 0;
329 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
330 	if (cookie == 2)
331 		return;
332 
333 	/*
334 	 * NFSv4 servers do not return entries for '.' and '..'
335 	 * Therefore, we fake these entries here.  We let '.'
336 	 * have cookie 0 and '..' have cookie 1.  Note that
337 	 * when talking to the server, we always send cookie 0
338 	 * instead of 1 or 2.
339 	 */
340 	start = p = kmap_atomic(*readdir->pages);
341 
342 	if (cookie == 0) {
343 		*p++ = xdr_one;                                  /* next */
344 		*p++ = xdr_zero;                   /* cookie, first word */
345 		*p++ = xdr_one;                   /* cookie, second word */
346 		*p++ = xdr_one;                             /* entry len */
347 		memcpy(p, ".\0\0\0", 4);                        /* entry */
348 		p++;
349 		*p++ = xdr_one;                         /* bitmap length */
350 		*p++ = htonl(attrs);                           /* bitmap */
351 		*p++ = htonl(12);             /* attribute buffer length */
352 		*p++ = htonl(NF4DIR);
353 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
354 	}
355 
356 	*p++ = xdr_one;                                  /* next */
357 	*p++ = xdr_zero;                   /* cookie, first word */
358 	*p++ = xdr_two;                   /* cookie, second word */
359 	*p++ = xdr_two;                             /* entry len */
360 	memcpy(p, "..\0\0", 4);                         /* entry */
361 	p++;
362 	*p++ = xdr_one;                         /* bitmap length */
363 	*p++ = htonl(attrs);                           /* bitmap */
364 	*p++ = htonl(12);             /* attribute buffer length */
365 	*p++ = htonl(NF4DIR);
366 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
367 
368 	readdir->pgbase = (char *)p - (char *)start;
369 	readdir->count -= readdir->pgbase;
370 	kunmap_atomic(start);
371 }
372 
373 static void nfs4_test_and_free_stateid(struct nfs_server *server,
374 		nfs4_stateid *stateid,
375 		const struct cred *cred)
376 {
377 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
378 
379 	ops->test_and_free_expired(server, stateid, cred);
380 }
381 
382 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
383 		nfs4_stateid *stateid,
384 		const struct cred *cred)
385 {
386 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
387 	nfs4_test_and_free_stateid(server, stateid, cred);
388 }
389 
390 static void nfs4_free_revoked_stateid(struct nfs_server *server,
391 		const nfs4_stateid *stateid,
392 		const struct cred *cred)
393 {
394 	nfs4_stateid tmp;
395 
396 	nfs4_stateid_copy(&tmp, stateid);
397 	__nfs4_free_revoked_stateid(server, &tmp, cred);
398 }
399 
400 static long nfs4_update_delay(long *timeout)
401 {
402 	long ret;
403 	if (!timeout)
404 		return NFS4_POLL_RETRY_MAX;
405 	if (*timeout <= 0)
406 		*timeout = NFS4_POLL_RETRY_MIN;
407 	if (*timeout > NFS4_POLL_RETRY_MAX)
408 		*timeout = NFS4_POLL_RETRY_MAX;
409 	ret = *timeout;
410 	*timeout <<= 1;
411 	return ret;
412 }
413 
414 static int nfs4_delay_killable(long *timeout)
415 {
416 	might_sleep();
417 
418 	freezable_schedule_timeout_killable_unsafe(
419 		nfs4_update_delay(timeout));
420 	if (!__fatal_signal_pending(current))
421 		return 0;
422 	return -EINTR;
423 }
424 
425 static int nfs4_delay_interruptible(long *timeout)
426 {
427 	might_sleep();
428 
429 	freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
430 	if (!signal_pending(current))
431 		return 0;
432 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
433 }
434 
435 static int nfs4_delay(long *timeout, bool interruptible)
436 {
437 	if (interruptible)
438 		return nfs4_delay_interruptible(timeout);
439 	return nfs4_delay_killable(timeout);
440 }
441 
442 static const nfs4_stateid *
443 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
444 {
445 	if (!stateid)
446 		return NULL;
447 	switch (stateid->type) {
448 	case NFS4_OPEN_STATEID_TYPE:
449 	case NFS4_LOCK_STATEID_TYPE:
450 	case NFS4_DELEGATION_STATEID_TYPE:
451 		return stateid;
452 	default:
453 		break;
454 	}
455 	return NULL;
456 }
457 
458 /* This is the error handling routine for processes that are allowed
459  * to sleep.
460  */
461 static int nfs4_do_handle_exception(struct nfs_server *server,
462 		int errorcode, struct nfs4_exception *exception)
463 {
464 	struct nfs_client *clp = server->nfs_client;
465 	struct nfs4_state *state = exception->state;
466 	const nfs4_stateid *stateid;
467 	struct inode *inode = exception->inode;
468 	int ret = errorcode;
469 
470 	exception->delay = 0;
471 	exception->recovering = 0;
472 	exception->retry = 0;
473 
474 	stateid = nfs4_recoverable_stateid(exception->stateid);
475 	if (stateid == NULL && state != NULL)
476 		stateid = nfs4_recoverable_stateid(&state->stateid);
477 
478 	switch(errorcode) {
479 		case 0:
480 			return 0;
481 		case -NFS4ERR_BADHANDLE:
482 		case -ESTALE:
483 			if (inode != NULL && S_ISREG(inode->i_mode))
484 				pnfs_destroy_layout(NFS_I(inode));
485 			break;
486 		case -NFS4ERR_DELEG_REVOKED:
487 		case -NFS4ERR_ADMIN_REVOKED:
488 		case -NFS4ERR_EXPIRED:
489 		case -NFS4ERR_BAD_STATEID:
490 		case -NFS4ERR_PARTNER_NO_AUTH:
491 			if (inode != NULL && stateid != NULL) {
492 				nfs_inode_find_state_and_recover(inode,
493 						stateid);
494 				goto wait_on_recovery;
495 			}
496 			fallthrough;
497 		case -NFS4ERR_OPENMODE:
498 			if (inode) {
499 				int err;
500 
501 				err = nfs_async_inode_return_delegation(inode,
502 						stateid);
503 				if (err == 0)
504 					goto wait_on_recovery;
505 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
506 					exception->retry = 1;
507 					break;
508 				}
509 			}
510 			if (state == NULL)
511 				break;
512 			ret = nfs4_schedule_stateid_recovery(server, state);
513 			if (ret < 0)
514 				break;
515 			goto wait_on_recovery;
516 		case -NFS4ERR_STALE_STATEID:
517 		case -NFS4ERR_STALE_CLIENTID:
518 			nfs4_schedule_lease_recovery(clp);
519 			goto wait_on_recovery;
520 		case -NFS4ERR_MOVED:
521 			ret = nfs4_schedule_migration_recovery(server);
522 			if (ret < 0)
523 				break;
524 			goto wait_on_recovery;
525 		case -NFS4ERR_LEASE_MOVED:
526 			nfs4_schedule_lease_moved_recovery(clp);
527 			goto wait_on_recovery;
528 #if defined(CONFIG_NFS_V4_1)
529 		case -NFS4ERR_BADSESSION:
530 		case -NFS4ERR_BADSLOT:
531 		case -NFS4ERR_BAD_HIGH_SLOT:
532 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
533 		case -NFS4ERR_DEADSESSION:
534 		case -NFS4ERR_SEQ_FALSE_RETRY:
535 		case -NFS4ERR_SEQ_MISORDERED:
536 			/* Handled in nfs41_sequence_process() */
537 			goto wait_on_recovery;
538 #endif /* defined(CONFIG_NFS_V4_1) */
539 		case -NFS4ERR_FILE_OPEN:
540 			if (exception->timeout > HZ) {
541 				/* We have retried a decent amount, time to
542 				 * fail
543 				 */
544 				ret = -EBUSY;
545 				break;
546 			}
547 			fallthrough;
548 		case -NFS4ERR_DELAY:
549 			nfs_inc_server_stats(server, NFSIOS_DELAY);
550 			fallthrough;
551 		case -NFS4ERR_GRACE:
552 		case -NFS4ERR_LAYOUTTRYLATER:
553 		case -NFS4ERR_RECALLCONFLICT:
554 			exception->delay = 1;
555 			return 0;
556 
557 		case -NFS4ERR_RETRY_UNCACHED_REP:
558 		case -NFS4ERR_OLD_STATEID:
559 			exception->retry = 1;
560 			break;
561 		case -NFS4ERR_BADOWNER:
562 			/* The following works around a Linux server bug! */
563 		case -NFS4ERR_BADNAME:
564 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
565 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
566 				exception->retry = 1;
567 				printk(KERN_WARNING "NFS: v4 server %s "
568 						"does not accept raw "
569 						"uid/gids. "
570 						"Reenabling the idmapper.\n",
571 						server->nfs_client->cl_hostname);
572 			}
573 	}
574 	/* We failed to handle the error */
575 	return nfs4_map_errors(ret);
576 wait_on_recovery:
577 	exception->recovering = 1;
578 	return 0;
579 }
580 
581 /* This is the error handling routine for processes that are allowed
582  * to sleep.
583  */
584 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
585 {
586 	struct nfs_client *clp = server->nfs_client;
587 	int ret;
588 
589 	ret = nfs4_do_handle_exception(server, errorcode, exception);
590 	if (exception->delay) {
591 		ret = nfs4_delay(&exception->timeout,
592 				exception->interruptible);
593 		goto out_retry;
594 	}
595 	if (exception->recovering) {
596 		ret = nfs4_wait_clnt_recover(clp);
597 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
598 			return -EIO;
599 		goto out_retry;
600 	}
601 	return ret;
602 out_retry:
603 	if (ret == 0)
604 		exception->retry = 1;
605 	return ret;
606 }
607 
608 static int
609 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
610 		int errorcode, struct nfs4_exception *exception)
611 {
612 	struct nfs_client *clp = server->nfs_client;
613 	int ret;
614 
615 	ret = nfs4_do_handle_exception(server, errorcode, exception);
616 	if (exception->delay) {
617 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
618 		goto out_retry;
619 	}
620 	if (exception->recovering) {
621 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
622 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
623 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
624 		goto out_retry;
625 	}
626 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
627 		ret = -EIO;
628 	return ret;
629 out_retry:
630 	if (ret == 0) {
631 		exception->retry = 1;
632 		/*
633 		 * For NFS4ERR_MOVED, the client transport will need to
634 		 * be recomputed after migration recovery has completed.
635 		 */
636 		if (errorcode == -NFS4ERR_MOVED)
637 			rpc_task_release_transport(task);
638 	}
639 	return ret;
640 }
641 
642 int
643 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
644 			struct nfs4_state *state, long *timeout)
645 {
646 	struct nfs4_exception exception = {
647 		.state = state,
648 	};
649 
650 	if (task->tk_status >= 0)
651 		return 0;
652 	if (timeout)
653 		exception.timeout = *timeout;
654 	task->tk_status = nfs4_async_handle_exception(task, server,
655 			task->tk_status,
656 			&exception);
657 	if (exception.delay && timeout)
658 		*timeout = exception.timeout;
659 	if (exception.retry)
660 		return -EAGAIN;
661 	return 0;
662 }
663 
664 /*
665  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
666  * or 'false' otherwise.
667  */
668 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
669 {
670 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
671 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
672 }
673 
674 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
675 {
676 	spin_lock(&clp->cl_lock);
677 	if (time_before(clp->cl_last_renewal,timestamp))
678 		clp->cl_last_renewal = timestamp;
679 	spin_unlock(&clp->cl_lock);
680 }
681 
682 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
683 {
684 	struct nfs_client *clp = server->nfs_client;
685 
686 	if (!nfs4_has_session(clp))
687 		do_renew_lease(clp, timestamp);
688 }
689 
690 struct nfs4_call_sync_data {
691 	const struct nfs_server *seq_server;
692 	struct nfs4_sequence_args *seq_args;
693 	struct nfs4_sequence_res *seq_res;
694 };
695 
696 void nfs4_init_sequence(struct nfs4_sequence_args *args,
697 			struct nfs4_sequence_res *res, int cache_reply,
698 			int privileged)
699 {
700 	args->sa_slot = NULL;
701 	args->sa_cache_this = cache_reply;
702 	args->sa_privileged = privileged;
703 
704 	res->sr_slot = NULL;
705 }
706 
707 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
708 {
709 	struct nfs4_slot *slot = res->sr_slot;
710 	struct nfs4_slot_table *tbl;
711 
712 	tbl = slot->table;
713 	spin_lock(&tbl->slot_tbl_lock);
714 	if (!nfs41_wake_and_assign_slot(tbl, slot))
715 		nfs4_free_slot(tbl, slot);
716 	spin_unlock(&tbl->slot_tbl_lock);
717 
718 	res->sr_slot = NULL;
719 }
720 
721 static int nfs40_sequence_done(struct rpc_task *task,
722 			       struct nfs4_sequence_res *res)
723 {
724 	if (res->sr_slot != NULL)
725 		nfs40_sequence_free_slot(res);
726 	return 1;
727 }
728 
729 #if defined(CONFIG_NFS_V4_1)
730 
731 static void nfs41_release_slot(struct nfs4_slot *slot)
732 {
733 	struct nfs4_session *session;
734 	struct nfs4_slot_table *tbl;
735 	bool send_new_highest_used_slotid = false;
736 
737 	if (!slot)
738 		return;
739 	tbl = slot->table;
740 	session = tbl->session;
741 
742 	/* Bump the slot sequence number */
743 	if (slot->seq_done)
744 		slot->seq_nr++;
745 	slot->seq_done = 0;
746 
747 	spin_lock(&tbl->slot_tbl_lock);
748 	/* Be nice to the server: try to ensure that the last transmitted
749 	 * value for highest_user_slotid <= target_highest_slotid
750 	 */
751 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
752 		send_new_highest_used_slotid = true;
753 
754 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
755 		send_new_highest_used_slotid = false;
756 		goto out_unlock;
757 	}
758 	nfs4_free_slot(tbl, slot);
759 
760 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
761 		send_new_highest_used_slotid = false;
762 out_unlock:
763 	spin_unlock(&tbl->slot_tbl_lock);
764 	if (send_new_highest_used_slotid)
765 		nfs41_notify_server(session->clp);
766 	if (waitqueue_active(&tbl->slot_waitq))
767 		wake_up_all(&tbl->slot_waitq);
768 }
769 
770 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
771 {
772 	nfs41_release_slot(res->sr_slot);
773 	res->sr_slot = NULL;
774 }
775 
776 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
777 		u32 seqnr)
778 {
779 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
780 		slot->seq_nr_highest_sent = seqnr;
781 }
782 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
783 		u32 seqnr)
784 {
785 	slot->seq_nr_highest_sent = seqnr;
786 	slot->seq_nr_last_acked = seqnr;
787 }
788 
789 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
790 				struct nfs4_slot *slot)
791 {
792 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
793 	if (!IS_ERR(task))
794 		rpc_put_task_async(task);
795 }
796 
797 static int nfs41_sequence_process(struct rpc_task *task,
798 		struct nfs4_sequence_res *res)
799 {
800 	struct nfs4_session *session;
801 	struct nfs4_slot *slot = res->sr_slot;
802 	struct nfs_client *clp;
803 	int status;
804 	int ret = 1;
805 
806 	if (slot == NULL)
807 		goto out_noaction;
808 	/* don't increment the sequence number if the task wasn't sent */
809 	if (!RPC_WAS_SENT(task) || slot->seq_done)
810 		goto out;
811 
812 	session = slot->table->session;
813 	clp = session->clp;
814 
815 	trace_nfs4_sequence_done(session, res);
816 
817 	status = res->sr_status;
818 	if (task->tk_status == -NFS4ERR_DEADSESSION)
819 		status = -NFS4ERR_DEADSESSION;
820 
821 	/* Check the SEQUENCE operation status */
822 	switch (status) {
823 	case 0:
824 		/* Mark this sequence number as having been acked */
825 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
826 		/* Update the slot's sequence and clientid lease timer */
827 		slot->seq_done = 1;
828 		do_renew_lease(clp, res->sr_timestamp);
829 		/* Check sequence flags */
830 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
831 				!!slot->privileged);
832 		nfs41_update_target_slotid(slot->table, slot, res);
833 		break;
834 	case 1:
835 		/*
836 		 * sr_status remains 1 if an RPC level error occurred.
837 		 * The server may or may not have processed the sequence
838 		 * operation..
839 		 */
840 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
841 		slot->seq_done = 1;
842 		goto out;
843 	case -NFS4ERR_DELAY:
844 		/* The server detected a resend of the RPC call and
845 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
846 		 * of RFC5661.
847 		 */
848 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
849 			__func__,
850 			slot->slot_nr,
851 			slot->seq_nr);
852 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
853 		goto out_retry;
854 	case -NFS4ERR_RETRY_UNCACHED_REP:
855 	case -NFS4ERR_SEQ_FALSE_RETRY:
856 		/*
857 		 * The server thinks we tried to replay a request.
858 		 * Retry the call after bumping the sequence ID.
859 		 */
860 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
861 		goto retry_new_seq;
862 	case -NFS4ERR_BADSLOT:
863 		/*
864 		 * The slot id we used was probably retired. Try again
865 		 * using a different slot id.
866 		 */
867 		if (slot->slot_nr < slot->table->target_highest_slotid)
868 			goto session_recover;
869 		goto retry_nowait;
870 	case -NFS4ERR_SEQ_MISORDERED:
871 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
872 		/*
873 		 * Were one or more calls using this slot interrupted?
874 		 * If the server never received the request, then our
875 		 * transmitted slot sequence number may be too high. However,
876 		 * if the server did receive the request then it might
877 		 * accidentally give us a reply with a mismatched operation.
878 		 * We can sort this out by sending a lone sequence operation
879 		 * to the server on the same slot.
880 		 */
881 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
882 			slot->seq_nr--;
883 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
884 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
885 				res->sr_slot = NULL;
886 			}
887 			goto retry_nowait;
888 		}
889 		/*
890 		 * RFC5661:
891 		 * A retry might be sent while the original request is
892 		 * still in progress on the replier. The replier SHOULD
893 		 * deal with the issue by returning NFS4ERR_DELAY as the
894 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
895 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
896 		 *
897 		 * Restart the search after a delay.
898 		 */
899 		slot->seq_nr = slot->seq_nr_highest_sent;
900 		goto out_retry;
901 	case -NFS4ERR_BADSESSION:
902 	case -NFS4ERR_DEADSESSION:
903 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
904 		goto session_recover;
905 	default:
906 		/* Just update the slot sequence no. */
907 		slot->seq_done = 1;
908 	}
909 out:
910 	/* The session may be reset by one of the error handlers. */
911 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
912 out_noaction:
913 	return ret;
914 session_recover:
915 	nfs4_schedule_session_recovery(session, status);
916 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
917 	nfs41_sequence_free_slot(res);
918 	goto out;
919 retry_new_seq:
920 	++slot->seq_nr;
921 retry_nowait:
922 	if (rpc_restart_call_prepare(task)) {
923 		nfs41_sequence_free_slot(res);
924 		task->tk_status = 0;
925 		ret = 0;
926 	}
927 	goto out;
928 out_retry:
929 	if (!rpc_restart_call(task))
930 		goto out;
931 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
932 	return 0;
933 }
934 
935 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
936 {
937 	if (!nfs41_sequence_process(task, res))
938 		return 0;
939 	if (res->sr_slot != NULL)
940 		nfs41_sequence_free_slot(res);
941 	return 1;
942 
943 }
944 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
945 
946 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
947 {
948 	if (res->sr_slot == NULL)
949 		return 1;
950 	if (res->sr_slot->table->session != NULL)
951 		return nfs41_sequence_process(task, res);
952 	return nfs40_sequence_done(task, res);
953 }
954 
955 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
956 {
957 	if (res->sr_slot != NULL) {
958 		if (res->sr_slot->table->session != NULL)
959 			nfs41_sequence_free_slot(res);
960 		else
961 			nfs40_sequence_free_slot(res);
962 	}
963 }
964 
965 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
966 {
967 	if (res->sr_slot == NULL)
968 		return 1;
969 	if (!res->sr_slot->table->session)
970 		return nfs40_sequence_done(task, res);
971 	return nfs41_sequence_done(task, res);
972 }
973 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
974 
975 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
976 {
977 	struct nfs4_call_sync_data *data = calldata;
978 
979 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
980 
981 	nfs4_setup_sequence(data->seq_server->nfs_client,
982 			    data->seq_args, data->seq_res, task);
983 }
984 
985 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
986 {
987 	struct nfs4_call_sync_data *data = calldata;
988 
989 	nfs41_sequence_done(task, data->seq_res);
990 }
991 
992 static const struct rpc_call_ops nfs41_call_sync_ops = {
993 	.rpc_call_prepare = nfs41_call_sync_prepare,
994 	.rpc_call_done = nfs41_call_sync_done,
995 };
996 
997 #else	/* !CONFIG_NFS_V4_1 */
998 
999 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1000 {
1001 	return nfs40_sequence_done(task, res);
1002 }
1003 
1004 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1005 {
1006 	if (res->sr_slot != NULL)
1007 		nfs40_sequence_free_slot(res);
1008 }
1009 
1010 int nfs4_sequence_done(struct rpc_task *task,
1011 		       struct nfs4_sequence_res *res)
1012 {
1013 	return nfs40_sequence_done(task, res);
1014 }
1015 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1016 
1017 #endif	/* !CONFIG_NFS_V4_1 */
1018 
1019 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1020 {
1021 	res->sr_timestamp = jiffies;
1022 	res->sr_status_flags = 0;
1023 	res->sr_status = 1;
1024 }
1025 
1026 static
1027 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1028 		struct nfs4_sequence_res *res,
1029 		struct nfs4_slot *slot)
1030 {
1031 	if (!slot)
1032 		return;
1033 	slot->privileged = args->sa_privileged ? 1 : 0;
1034 	args->sa_slot = slot;
1035 
1036 	res->sr_slot = slot;
1037 }
1038 
1039 int nfs4_setup_sequence(struct nfs_client *client,
1040 			struct nfs4_sequence_args *args,
1041 			struct nfs4_sequence_res *res,
1042 			struct rpc_task *task)
1043 {
1044 	struct nfs4_session *session = nfs4_get_session(client);
1045 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1046 	struct nfs4_slot *slot;
1047 
1048 	/* slot already allocated? */
1049 	if (res->sr_slot != NULL)
1050 		goto out_start;
1051 
1052 	if (session)
1053 		tbl = &session->fc_slot_table;
1054 
1055 	spin_lock(&tbl->slot_tbl_lock);
1056 	/* The state manager will wait until the slot table is empty */
1057 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1058 		goto out_sleep;
1059 
1060 	slot = nfs4_alloc_slot(tbl);
1061 	if (IS_ERR(slot)) {
1062 		if (slot == ERR_PTR(-ENOMEM))
1063 			goto out_sleep_timeout;
1064 		goto out_sleep;
1065 	}
1066 	spin_unlock(&tbl->slot_tbl_lock);
1067 
1068 	nfs4_sequence_attach_slot(args, res, slot);
1069 
1070 	trace_nfs4_setup_sequence(session, args);
1071 out_start:
1072 	nfs41_sequence_res_init(res);
1073 	rpc_call_start(task);
1074 	return 0;
1075 out_sleep_timeout:
1076 	/* Try again in 1/4 second */
1077 	if (args->sa_privileged)
1078 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1079 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1080 	else
1081 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1082 				NULL, jiffies + (HZ >> 2));
1083 	spin_unlock(&tbl->slot_tbl_lock);
1084 	return -EAGAIN;
1085 out_sleep:
1086 	if (args->sa_privileged)
1087 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1088 				RPC_PRIORITY_PRIVILEGED);
1089 	else
1090 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1091 	spin_unlock(&tbl->slot_tbl_lock);
1092 	return -EAGAIN;
1093 }
1094 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1095 
1096 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1097 {
1098 	struct nfs4_call_sync_data *data = calldata;
1099 	nfs4_setup_sequence(data->seq_server->nfs_client,
1100 				data->seq_args, data->seq_res, task);
1101 }
1102 
1103 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1104 {
1105 	struct nfs4_call_sync_data *data = calldata;
1106 	nfs4_sequence_done(task, data->seq_res);
1107 }
1108 
1109 static const struct rpc_call_ops nfs40_call_sync_ops = {
1110 	.rpc_call_prepare = nfs40_call_sync_prepare,
1111 	.rpc_call_done = nfs40_call_sync_done,
1112 };
1113 
1114 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1115 {
1116 	int ret;
1117 	struct rpc_task *task;
1118 
1119 	task = rpc_run_task(task_setup);
1120 	if (IS_ERR(task))
1121 		return PTR_ERR(task);
1122 
1123 	ret = task->tk_status;
1124 	rpc_put_task(task);
1125 	return ret;
1126 }
1127 
1128 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1129 			     struct nfs_server *server,
1130 			     struct rpc_message *msg,
1131 			     struct nfs4_sequence_args *args,
1132 			     struct nfs4_sequence_res *res,
1133 			     unsigned short task_flags)
1134 {
1135 	struct nfs_client *clp = server->nfs_client;
1136 	struct nfs4_call_sync_data data = {
1137 		.seq_server = server,
1138 		.seq_args = args,
1139 		.seq_res = res,
1140 	};
1141 	struct rpc_task_setup task_setup = {
1142 		.rpc_client = clnt,
1143 		.rpc_message = msg,
1144 		.callback_ops = clp->cl_mvops->call_sync_ops,
1145 		.callback_data = &data,
1146 		.flags = task_flags,
1147 	};
1148 
1149 	return nfs4_call_sync_custom(&task_setup);
1150 }
1151 
1152 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1153 				   struct nfs_server *server,
1154 				   struct rpc_message *msg,
1155 				   struct nfs4_sequence_args *args,
1156 				   struct nfs4_sequence_res *res)
1157 {
1158 	return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1159 }
1160 
1161 
1162 int nfs4_call_sync(struct rpc_clnt *clnt,
1163 		   struct nfs_server *server,
1164 		   struct rpc_message *msg,
1165 		   struct nfs4_sequence_args *args,
1166 		   struct nfs4_sequence_res *res,
1167 		   int cache_reply)
1168 {
1169 	nfs4_init_sequence(args, res, cache_reply, 0);
1170 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1171 }
1172 
1173 static void
1174 nfs4_inc_nlink_locked(struct inode *inode)
1175 {
1176 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1177 	inc_nlink(inode);
1178 }
1179 
1180 static void
1181 nfs4_dec_nlink_locked(struct inode *inode)
1182 {
1183 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1184 	drop_nlink(inode);
1185 }
1186 
1187 static void
1188 nfs4_update_changeattr_locked(struct inode *inode,
1189 		struct nfs4_change_info *cinfo,
1190 		unsigned long timestamp, unsigned long cache_validity)
1191 {
1192 	struct nfs_inode *nfsi = NFS_I(inode);
1193 
1194 	nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1195 		| NFS_INO_INVALID_MTIME
1196 		| cache_validity;
1197 
1198 	if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) {
1199 		nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1200 		nfsi->attrtimeo_timestamp = jiffies;
1201 	} else {
1202 		if (S_ISDIR(inode->i_mode)) {
1203 			nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1204 			nfs_force_lookup_revalidate(inode);
1205 		} else {
1206 			if (!NFS_PROTO(inode)->have_delegation(inode,
1207 							       FMODE_READ))
1208 				nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1209 		}
1210 
1211 		if (cinfo->before != inode_peek_iversion_raw(inode))
1212 			nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1213 						NFS_INO_INVALID_ACL |
1214 						NFS_INO_INVALID_XATTR;
1215 	}
1216 	inode_set_iversion_raw(inode, cinfo->after);
1217 	nfsi->read_cache_jiffies = timestamp;
1218 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1219 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1220 
1221 	if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
1222 		nfs_fscache_invalidate(inode);
1223 }
1224 
1225 void
1226 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1227 		unsigned long timestamp, unsigned long cache_validity)
1228 {
1229 	spin_lock(&dir->i_lock);
1230 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1231 	spin_unlock(&dir->i_lock);
1232 }
1233 
1234 struct nfs4_open_createattrs {
1235 	struct nfs4_label *label;
1236 	struct iattr *sattr;
1237 	const __u32 verf[2];
1238 };
1239 
1240 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1241 		int err, struct nfs4_exception *exception)
1242 {
1243 	if (err != -EINVAL)
1244 		return false;
1245 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1246 		return false;
1247 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1248 	exception->retry = 1;
1249 	return true;
1250 }
1251 
1252 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1253 {
1254 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1255 }
1256 
1257 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1258 {
1259 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1260 
1261 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1262 }
1263 
1264 static u32
1265 nfs4_map_atomic_open_share(struct nfs_server *server,
1266 		fmode_t fmode, int openflags)
1267 {
1268 	u32 res = 0;
1269 
1270 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1271 	case FMODE_READ:
1272 		res = NFS4_SHARE_ACCESS_READ;
1273 		break;
1274 	case FMODE_WRITE:
1275 		res = NFS4_SHARE_ACCESS_WRITE;
1276 		break;
1277 	case FMODE_READ|FMODE_WRITE:
1278 		res = NFS4_SHARE_ACCESS_BOTH;
1279 	}
1280 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1281 		goto out;
1282 	/* Want no delegation if we're using O_DIRECT */
1283 	if (openflags & O_DIRECT)
1284 		res |= NFS4_SHARE_WANT_NO_DELEG;
1285 out:
1286 	return res;
1287 }
1288 
1289 static enum open_claim_type4
1290 nfs4_map_atomic_open_claim(struct nfs_server *server,
1291 		enum open_claim_type4 claim)
1292 {
1293 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1294 		return claim;
1295 	switch (claim) {
1296 	default:
1297 		return claim;
1298 	case NFS4_OPEN_CLAIM_FH:
1299 		return NFS4_OPEN_CLAIM_NULL;
1300 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1301 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1302 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1303 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1304 	}
1305 }
1306 
1307 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1308 {
1309 	p->o_res.f_attr = &p->f_attr;
1310 	p->o_res.f_label = p->f_label;
1311 	p->o_res.seqid = p->o_arg.seqid;
1312 	p->c_res.seqid = p->c_arg.seqid;
1313 	p->o_res.server = p->o_arg.server;
1314 	p->o_res.access_request = p->o_arg.access;
1315 	nfs_fattr_init(&p->f_attr);
1316 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1317 }
1318 
1319 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1320 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1321 		const struct nfs4_open_createattrs *c,
1322 		enum open_claim_type4 claim,
1323 		gfp_t gfp_mask)
1324 {
1325 	struct dentry *parent = dget_parent(dentry);
1326 	struct inode *dir = d_inode(parent);
1327 	struct nfs_server *server = NFS_SERVER(dir);
1328 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1329 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1330 	struct nfs4_opendata *p;
1331 
1332 	p = kzalloc(sizeof(*p), gfp_mask);
1333 	if (p == NULL)
1334 		goto err;
1335 
1336 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1337 	if (IS_ERR(p->f_label))
1338 		goto err_free_p;
1339 
1340 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1341 	if (IS_ERR(p->a_label))
1342 		goto err_free_f;
1343 
1344 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1345 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1346 	if (IS_ERR(p->o_arg.seqid))
1347 		goto err_free_label;
1348 	nfs_sb_active(dentry->d_sb);
1349 	p->dentry = dget(dentry);
1350 	p->dir = parent;
1351 	p->owner = sp;
1352 	atomic_inc(&sp->so_count);
1353 	p->o_arg.open_flags = flags;
1354 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1355 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1356 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1357 			fmode, flags);
1358 	if (flags & O_CREAT) {
1359 		p->o_arg.umask = current_umask();
1360 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1361 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1362 			p->o_arg.u.attrs = &p->attrs;
1363 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1364 
1365 			memcpy(p->o_arg.u.verifier.data, c->verf,
1366 					sizeof(p->o_arg.u.verifier.data));
1367 		}
1368 	}
1369 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1370 	 * will return permission denied for all bits until close */
1371 	if (!(flags & O_EXCL)) {
1372 		/* ask server to check for all possible rights as results
1373 		 * are cached */
1374 		switch (p->o_arg.claim) {
1375 		default:
1376 			break;
1377 		case NFS4_OPEN_CLAIM_NULL:
1378 		case NFS4_OPEN_CLAIM_FH:
1379 			p->o_arg.access = NFS4_ACCESS_READ |
1380 				NFS4_ACCESS_MODIFY |
1381 				NFS4_ACCESS_EXTEND |
1382 				NFS4_ACCESS_EXECUTE;
1383 #ifdef CONFIG_NFS_V4_2
1384 			if (server->caps & NFS_CAP_XATTR)
1385 				p->o_arg.access |= NFS4_ACCESS_XAREAD |
1386 				    NFS4_ACCESS_XAWRITE |
1387 				    NFS4_ACCESS_XALIST;
1388 #endif
1389 		}
1390 	}
1391 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1392 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1393 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1394 	p->o_arg.name = &dentry->d_name;
1395 	p->o_arg.server = server;
1396 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1397 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1398 	switch (p->o_arg.claim) {
1399 	case NFS4_OPEN_CLAIM_NULL:
1400 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1401 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1402 		p->o_arg.fh = NFS_FH(dir);
1403 		break;
1404 	case NFS4_OPEN_CLAIM_PREVIOUS:
1405 	case NFS4_OPEN_CLAIM_FH:
1406 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1407 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1408 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1409 	}
1410 	p->c_arg.fh = &p->o_res.fh;
1411 	p->c_arg.stateid = &p->o_res.stateid;
1412 	p->c_arg.seqid = p->o_arg.seqid;
1413 	nfs4_init_opendata_res(p);
1414 	kref_init(&p->kref);
1415 	return p;
1416 
1417 err_free_label:
1418 	nfs4_label_free(p->a_label);
1419 err_free_f:
1420 	nfs4_label_free(p->f_label);
1421 err_free_p:
1422 	kfree(p);
1423 err:
1424 	dput(parent);
1425 	return NULL;
1426 }
1427 
1428 static void nfs4_opendata_free(struct kref *kref)
1429 {
1430 	struct nfs4_opendata *p = container_of(kref,
1431 			struct nfs4_opendata, kref);
1432 	struct super_block *sb = p->dentry->d_sb;
1433 
1434 	nfs4_lgopen_release(p->lgp);
1435 	nfs_free_seqid(p->o_arg.seqid);
1436 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1437 	if (p->state != NULL)
1438 		nfs4_put_open_state(p->state);
1439 	nfs4_put_state_owner(p->owner);
1440 
1441 	nfs4_label_free(p->a_label);
1442 	nfs4_label_free(p->f_label);
1443 
1444 	dput(p->dir);
1445 	dput(p->dentry);
1446 	nfs_sb_deactive(sb);
1447 	nfs_fattr_free_names(&p->f_attr);
1448 	kfree(p->f_attr.mdsthreshold);
1449 	kfree(p);
1450 }
1451 
1452 static void nfs4_opendata_put(struct nfs4_opendata *p)
1453 {
1454 	if (p != NULL)
1455 		kref_put(&p->kref, nfs4_opendata_free);
1456 }
1457 
1458 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1459 		fmode_t fmode)
1460 {
1461 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1462 	case FMODE_READ|FMODE_WRITE:
1463 		return state->n_rdwr != 0;
1464 	case FMODE_WRITE:
1465 		return state->n_wronly != 0;
1466 	case FMODE_READ:
1467 		return state->n_rdonly != 0;
1468 	}
1469 	WARN_ON_ONCE(1);
1470 	return false;
1471 }
1472 
1473 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1474 		int open_mode, enum open_claim_type4 claim)
1475 {
1476 	int ret = 0;
1477 
1478 	if (open_mode & (O_EXCL|O_TRUNC))
1479 		goto out;
1480 	switch (claim) {
1481 	case NFS4_OPEN_CLAIM_NULL:
1482 	case NFS4_OPEN_CLAIM_FH:
1483 		goto out;
1484 	default:
1485 		break;
1486 	}
1487 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1488 		case FMODE_READ:
1489 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1490 				&& state->n_rdonly != 0;
1491 			break;
1492 		case FMODE_WRITE:
1493 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1494 				&& state->n_wronly != 0;
1495 			break;
1496 		case FMODE_READ|FMODE_WRITE:
1497 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1498 				&& state->n_rdwr != 0;
1499 	}
1500 out:
1501 	return ret;
1502 }
1503 
1504 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1505 		enum open_claim_type4 claim)
1506 {
1507 	if (delegation == NULL)
1508 		return 0;
1509 	if ((delegation->type & fmode) != fmode)
1510 		return 0;
1511 	switch (claim) {
1512 	case NFS4_OPEN_CLAIM_NULL:
1513 	case NFS4_OPEN_CLAIM_FH:
1514 		break;
1515 	case NFS4_OPEN_CLAIM_PREVIOUS:
1516 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1517 			break;
1518 		fallthrough;
1519 	default:
1520 		return 0;
1521 	}
1522 	nfs_mark_delegation_referenced(delegation);
1523 	return 1;
1524 }
1525 
1526 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1527 {
1528 	switch (fmode) {
1529 		case FMODE_WRITE:
1530 			state->n_wronly++;
1531 			break;
1532 		case FMODE_READ:
1533 			state->n_rdonly++;
1534 			break;
1535 		case FMODE_READ|FMODE_WRITE:
1536 			state->n_rdwr++;
1537 	}
1538 	nfs4_state_set_mode_locked(state, state->state | fmode);
1539 }
1540 
1541 #ifdef CONFIG_NFS_V4_1
1542 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1543 {
1544 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1545 		return true;
1546 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1547 		return true;
1548 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1549 		return true;
1550 	return false;
1551 }
1552 #endif /* CONFIG_NFS_V4_1 */
1553 
1554 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1555 {
1556 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1557 		wake_up_all(&state->waitq);
1558 }
1559 
1560 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1561 {
1562 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1563 	bool need_recover = false;
1564 
1565 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1566 		need_recover = true;
1567 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1568 		need_recover = true;
1569 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1570 		need_recover = true;
1571 	if (need_recover)
1572 		nfs4_state_mark_reclaim_nograce(clp, state);
1573 }
1574 
1575 /*
1576  * Check for whether or not the caller may update the open stateid
1577  * to the value passed in by stateid.
1578  *
1579  * Note: This function relies heavily on the server implementing
1580  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1581  * correctly.
1582  * i.e. The stateid seqids have to be initialised to 1, and
1583  * are then incremented on every state transition.
1584  */
1585 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1586 		const nfs4_stateid *stateid)
1587 {
1588 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1589 		/* The common case - we're updating to a new sequence number */
1590 		if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1591 			nfs4_stateid_is_next(&state->open_stateid, stateid)) {
1592 			return true;
1593 		}
1594 	} else {
1595 		/* This is the first OPEN in this generation */
1596 		if (stateid->seqid == cpu_to_be32(1))
1597 			return true;
1598 	}
1599 	return false;
1600 }
1601 
1602 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1603 {
1604 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1605 		return;
1606 	if (state->n_wronly)
1607 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1608 	if (state->n_rdonly)
1609 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1610 	if (state->n_rdwr)
1611 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1612 	set_bit(NFS_OPEN_STATE, &state->flags);
1613 }
1614 
1615 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1616 		nfs4_stateid *stateid, fmode_t fmode)
1617 {
1618 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1619 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1620 	case FMODE_WRITE:
1621 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1622 		break;
1623 	case FMODE_READ:
1624 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1625 		break;
1626 	case 0:
1627 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1628 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1629 		clear_bit(NFS_OPEN_STATE, &state->flags);
1630 	}
1631 	if (stateid == NULL)
1632 		return;
1633 	/* Handle OPEN+OPEN_DOWNGRADE races */
1634 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1635 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1636 		nfs_resync_open_stateid_locked(state);
1637 		goto out;
1638 	}
1639 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1640 		nfs4_stateid_copy(&state->stateid, stateid);
1641 	nfs4_stateid_copy(&state->open_stateid, stateid);
1642 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1643 out:
1644 	nfs_state_log_update_open_stateid(state);
1645 }
1646 
1647 static void nfs_clear_open_stateid(struct nfs4_state *state,
1648 	nfs4_stateid *arg_stateid,
1649 	nfs4_stateid *stateid, fmode_t fmode)
1650 {
1651 	write_seqlock(&state->seqlock);
1652 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1653 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1654 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1655 	write_sequnlock(&state->seqlock);
1656 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1657 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1658 }
1659 
1660 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1661 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1662 	__must_hold(&state->owner->so_lock)
1663 	__must_hold(&state->seqlock)
1664 	__must_hold(RCU)
1665 
1666 {
1667 	DEFINE_WAIT(wait);
1668 	int status = 0;
1669 	for (;;) {
1670 
1671 		if (nfs_stateid_is_sequential(state, stateid))
1672 			break;
1673 
1674 		if (status)
1675 			break;
1676 		/* Rely on seqids for serialisation with NFSv4.0 */
1677 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1678 			break;
1679 
1680 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1681 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1682 		/*
1683 		 * Ensure we process the state changes in the same order
1684 		 * in which the server processed them by delaying the
1685 		 * update of the stateid until we are in sequence.
1686 		 */
1687 		write_sequnlock(&state->seqlock);
1688 		spin_unlock(&state->owner->so_lock);
1689 		rcu_read_unlock();
1690 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1691 
1692 		if (!signal_pending(current)) {
1693 			if (schedule_timeout(5*HZ) == 0)
1694 				status = -EAGAIN;
1695 			else
1696 				status = 0;
1697 		} else
1698 			status = -EINTR;
1699 		finish_wait(&state->waitq, &wait);
1700 		rcu_read_lock();
1701 		spin_lock(&state->owner->so_lock);
1702 		write_seqlock(&state->seqlock);
1703 	}
1704 
1705 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1706 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1707 		nfs4_stateid_copy(freeme, &state->open_stateid);
1708 		nfs_test_and_clear_all_open_stateid(state);
1709 	}
1710 
1711 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1712 		nfs4_stateid_copy(&state->stateid, stateid);
1713 	nfs4_stateid_copy(&state->open_stateid, stateid);
1714 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1715 	nfs_state_log_update_open_stateid(state);
1716 }
1717 
1718 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1719 		const nfs4_stateid *open_stateid,
1720 		fmode_t fmode,
1721 		nfs4_stateid *freeme)
1722 {
1723 	/*
1724 	 * Protect the call to nfs4_state_set_mode_locked and
1725 	 * serialise the stateid update
1726 	 */
1727 	write_seqlock(&state->seqlock);
1728 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1729 	switch (fmode) {
1730 	case FMODE_READ:
1731 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1732 		break;
1733 	case FMODE_WRITE:
1734 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1735 		break;
1736 	case FMODE_READ|FMODE_WRITE:
1737 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1738 	}
1739 	set_bit(NFS_OPEN_STATE, &state->flags);
1740 	write_sequnlock(&state->seqlock);
1741 }
1742 
1743 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1744 {
1745 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1746 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1747 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1748 	clear_bit(NFS_OPEN_STATE, &state->flags);
1749 }
1750 
1751 static void nfs_state_set_delegation(struct nfs4_state *state,
1752 		const nfs4_stateid *deleg_stateid,
1753 		fmode_t fmode)
1754 {
1755 	/*
1756 	 * Protect the call to nfs4_state_set_mode_locked and
1757 	 * serialise the stateid update
1758 	 */
1759 	write_seqlock(&state->seqlock);
1760 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1761 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1762 	write_sequnlock(&state->seqlock);
1763 }
1764 
1765 static void nfs_state_clear_delegation(struct nfs4_state *state)
1766 {
1767 	write_seqlock(&state->seqlock);
1768 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1769 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1770 	write_sequnlock(&state->seqlock);
1771 }
1772 
1773 int update_open_stateid(struct nfs4_state *state,
1774 		const nfs4_stateid *open_stateid,
1775 		const nfs4_stateid *delegation,
1776 		fmode_t fmode)
1777 {
1778 	struct nfs_server *server = NFS_SERVER(state->inode);
1779 	struct nfs_client *clp = server->nfs_client;
1780 	struct nfs_inode *nfsi = NFS_I(state->inode);
1781 	struct nfs_delegation *deleg_cur;
1782 	nfs4_stateid freeme = { };
1783 	int ret = 0;
1784 
1785 	fmode &= (FMODE_READ|FMODE_WRITE);
1786 
1787 	rcu_read_lock();
1788 	spin_lock(&state->owner->so_lock);
1789 	if (open_stateid != NULL) {
1790 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1791 		ret = 1;
1792 	}
1793 
1794 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1795 	if (deleg_cur == NULL)
1796 		goto no_delegation;
1797 
1798 	spin_lock(&deleg_cur->lock);
1799 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1800 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1801 	    (deleg_cur->type & fmode) != fmode)
1802 		goto no_delegation_unlock;
1803 
1804 	if (delegation == NULL)
1805 		delegation = &deleg_cur->stateid;
1806 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1807 		goto no_delegation_unlock;
1808 
1809 	nfs_mark_delegation_referenced(deleg_cur);
1810 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1811 	ret = 1;
1812 no_delegation_unlock:
1813 	spin_unlock(&deleg_cur->lock);
1814 no_delegation:
1815 	if (ret)
1816 		update_open_stateflags(state, fmode);
1817 	spin_unlock(&state->owner->so_lock);
1818 	rcu_read_unlock();
1819 
1820 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1821 		nfs4_schedule_state_manager(clp);
1822 	if (freeme.type != 0)
1823 		nfs4_test_and_free_stateid(server, &freeme,
1824 				state->owner->so_cred);
1825 
1826 	return ret;
1827 }
1828 
1829 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1830 		const nfs4_stateid *stateid)
1831 {
1832 	struct nfs4_state *state = lsp->ls_state;
1833 	bool ret = false;
1834 
1835 	spin_lock(&state->state_lock);
1836 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1837 		goto out_noupdate;
1838 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1839 		goto out_noupdate;
1840 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1841 	ret = true;
1842 out_noupdate:
1843 	spin_unlock(&state->state_lock);
1844 	return ret;
1845 }
1846 
1847 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1848 {
1849 	struct nfs_delegation *delegation;
1850 
1851 	fmode &= FMODE_READ|FMODE_WRITE;
1852 	rcu_read_lock();
1853 	delegation = nfs4_get_valid_delegation(inode);
1854 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1855 		rcu_read_unlock();
1856 		return;
1857 	}
1858 	rcu_read_unlock();
1859 	nfs4_inode_return_delegation(inode);
1860 }
1861 
1862 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1863 {
1864 	struct nfs4_state *state = opendata->state;
1865 	struct nfs_delegation *delegation;
1866 	int open_mode = opendata->o_arg.open_flags;
1867 	fmode_t fmode = opendata->o_arg.fmode;
1868 	enum open_claim_type4 claim = opendata->o_arg.claim;
1869 	nfs4_stateid stateid;
1870 	int ret = -EAGAIN;
1871 
1872 	for (;;) {
1873 		spin_lock(&state->owner->so_lock);
1874 		if (can_open_cached(state, fmode, open_mode, claim)) {
1875 			update_open_stateflags(state, fmode);
1876 			spin_unlock(&state->owner->so_lock);
1877 			goto out_return_state;
1878 		}
1879 		spin_unlock(&state->owner->so_lock);
1880 		rcu_read_lock();
1881 		delegation = nfs4_get_valid_delegation(state->inode);
1882 		if (!can_open_delegated(delegation, fmode, claim)) {
1883 			rcu_read_unlock();
1884 			break;
1885 		}
1886 		/* Save the delegation */
1887 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1888 		rcu_read_unlock();
1889 		nfs_release_seqid(opendata->o_arg.seqid);
1890 		if (!opendata->is_recover) {
1891 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1892 			if (ret != 0)
1893 				goto out;
1894 		}
1895 		ret = -EAGAIN;
1896 
1897 		/* Try to update the stateid using the delegation */
1898 		if (update_open_stateid(state, NULL, &stateid, fmode))
1899 			goto out_return_state;
1900 	}
1901 out:
1902 	return ERR_PTR(ret);
1903 out_return_state:
1904 	refcount_inc(&state->count);
1905 	return state;
1906 }
1907 
1908 static void
1909 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1910 {
1911 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1912 	struct nfs_delegation *delegation;
1913 	int delegation_flags = 0;
1914 
1915 	rcu_read_lock();
1916 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1917 	if (delegation)
1918 		delegation_flags = delegation->flags;
1919 	rcu_read_unlock();
1920 	switch (data->o_arg.claim) {
1921 	default:
1922 		break;
1923 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1924 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1925 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1926 				   "returning a delegation for "
1927 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1928 				   clp->cl_hostname);
1929 		return;
1930 	}
1931 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1932 		nfs_inode_set_delegation(state->inode,
1933 				data->owner->so_cred,
1934 				data->o_res.delegation_type,
1935 				&data->o_res.delegation,
1936 				data->o_res.pagemod_limit);
1937 	else
1938 		nfs_inode_reclaim_delegation(state->inode,
1939 				data->owner->so_cred,
1940 				data->o_res.delegation_type,
1941 				&data->o_res.delegation,
1942 				data->o_res.pagemod_limit);
1943 
1944 	if (data->o_res.do_recall)
1945 		nfs_async_inode_return_delegation(state->inode,
1946 						  &data->o_res.delegation);
1947 }
1948 
1949 /*
1950  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1951  * and update the nfs4_state.
1952  */
1953 static struct nfs4_state *
1954 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1955 {
1956 	struct inode *inode = data->state->inode;
1957 	struct nfs4_state *state = data->state;
1958 	int ret;
1959 
1960 	if (!data->rpc_done) {
1961 		if (data->rpc_status)
1962 			return ERR_PTR(data->rpc_status);
1963 		/* cached opens have already been processed */
1964 		goto update;
1965 	}
1966 
1967 	ret = nfs_refresh_inode(inode, &data->f_attr);
1968 	if (ret)
1969 		return ERR_PTR(ret);
1970 
1971 	if (data->o_res.delegation_type != 0)
1972 		nfs4_opendata_check_deleg(data, state);
1973 update:
1974 	if (!update_open_stateid(state, &data->o_res.stateid,
1975 				NULL, data->o_arg.fmode))
1976 		return ERR_PTR(-EAGAIN);
1977 	refcount_inc(&state->count);
1978 
1979 	return state;
1980 }
1981 
1982 static struct inode *
1983 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1984 {
1985 	struct inode *inode;
1986 
1987 	switch (data->o_arg.claim) {
1988 	case NFS4_OPEN_CLAIM_NULL:
1989 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1990 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1991 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1992 			return ERR_PTR(-EAGAIN);
1993 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1994 				&data->f_attr, data->f_label);
1995 		break;
1996 	default:
1997 		inode = d_inode(data->dentry);
1998 		ihold(inode);
1999 		nfs_refresh_inode(inode, &data->f_attr);
2000 	}
2001 	return inode;
2002 }
2003 
2004 static struct nfs4_state *
2005 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2006 {
2007 	struct nfs4_state *state;
2008 	struct inode *inode;
2009 
2010 	inode = nfs4_opendata_get_inode(data);
2011 	if (IS_ERR(inode))
2012 		return ERR_CAST(inode);
2013 	if (data->state != NULL && data->state->inode == inode) {
2014 		state = data->state;
2015 		refcount_inc(&state->count);
2016 	} else
2017 		state = nfs4_get_open_state(inode, data->owner);
2018 	iput(inode);
2019 	if (state == NULL)
2020 		state = ERR_PTR(-ENOMEM);
2021 	return state;
2022 }
2023 
2024 static struct nfs4_state *
2025 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2026 {
2027 	struct nfs4_state *state;
2028 
2029 	if (!data->rpc_done) {
2030 		state = nfs4_try_open_cached(data);
2031 		trace_nfs4_cached_open(data->state);
2032 		goto out;
2033 	}
2034 
2035 	state = nfs4_opendata_find_nfs4_state(data);
2036 	if (IS_ERR(state))
2037 		goto out;
2038 
2039 	if (data->o_res.delegation_type != 0)
2040 		nfs4_opendata_check_deleg(data, state);
2041 	if (!update_open_stateid(state, &data->o_res.stateid,
2042 				NULL, data->o_arg.fmode)) {
2043 		nfs4_put_open_state(state);
2044 		state = ERR_PTR(-EAGAIN);
2045 	}
2046 out:
2047 	nfs_release_seqid(data->o_arg.seqid);
2048 	return state;
2049 }
2050 
2051 static struct nfs4_state *
2052 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2053 {
2054 	struct nfs4_state *ret;
2055 
2056 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2057 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2058 	else
2059 		ret = _nfs4_opendata_to_nfs4_state(data);
2060 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2061 	return ret;
2062 }
2063 
2064 static struct nfs_open_context *
2065 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2066 {
2067 	struct nfs_inode *nfsi = NFS_I(state->inode);
2068 	struct nfs_open_context *ctx;
2069 
2070 	rcu_read_lock();
2071 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2072 		if (ctx->state != state)
2073 			continue;
2074 		if ((ctx->mode & mode) != mode)
2075 			continue;
2076 		if (!get_nfs_open_context(ctx))
2077 			continue;
2078 		rcu_read_unlock();
2079 		return ctx;
2080 	}
2081 	rcu_read_unlock();
2082 	return ERR_PTR(-ENOENT);
2083 }
2084 
2085 static struct nfs_open_context *
2086 nfs4_state_find_open_context(struct nfs4_state *state)
2087 {
2088 	struct nfs_open_context *ctx;
2089 
2090 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2091 	if (!IS_ERR(ctx))
2092 		return ctx;
2093 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2094 	if (!IS_ERR(ctx))
2095 		return ctx;
2096 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2097 }
2098 
2099 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2100 		struct nfs4_state *state, enum open_claim_type4 claim)
2101 {
2102 	struct nfs4_opendata *opendata;
2103 
2104 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2105 			NULL, claim, GFP_NOFS);
2106 	if (opendata == NULL)
2107 		return ERR_PTR(-ENOMEM);
2108 	opendata->state = state;
2109 	refcount_inc(&state->count);
2110 	return opendata;
2111 }
2112 
2113 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2114 		fmode_t fmode)
2115 {
2116 	struct nfs4_state *newstate;
2117 	int ret;
2118 
2119 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2120 		return 0;
2121 	opendata->o_arg.open_flags = 0;
2122 	opendata->o_arg.fmode = fmode;
2123 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2124 			NFS_SB(opendata->dentry->d_sb),
2125 			fmode, 0);
2126 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2127 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2128 	nfs4_init_opendata_res(opendata);
2129 	ret = _nfs4_recover_proc_open(opendata);
2130 	if (ret != 0)
2131 		return ret;
2132 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2133 	if (IS_ERR(newstate))
2134 		return PTR_ERR(newstate);
2135 	if (newstate != opendata->state)
2136 		ret = -ESTALE;
2137 	nfs4_close_state(newstate, fmode);
2138 	return ret;
2139 }
2140 
2141 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2142 {
2143 	int ret;
2144 
2145 	/* memory barrier prior to reading state->n_* */
2146 	smp_rmb();
2147 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2148 	if (ret != 0)
2149 		return ret;
2150 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2151 	if (ret != 0)
2152 		return ret;
2153 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2154 	if (ret != 0)
2155 		return ret;
2156 	/*
2157 	 * We may have performed cached opens for all three recoveries.
2158 	 * Check if we need to update the current stateid.
2159 	 */
2160 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2161 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2162 		write_seqlock(&state->seqlock);
2163 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2164 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2165 		write_sequnlock(&state->seqlock);
2166 	}
2167 	return 0;
2168 }
2169 
2170 /*
2171  * OPEN_RECLAIM:
2172  * 	reclaim state on the server after a reboot.
2173  */
2174 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2175 {
2176 	struct nfs_delegation *delegation;
2177 	struct nfs4_opendata *opendata;
2178 	fmode_t delegation_type = 0;
2179 	int status;
2180 
2181 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2182 			NFS4_OPEN_CLAIM_PREVIOUS);
2183 	if (IS_ERR(opendata))
2184 		return PTR_ERR(opendata);
2185 	rcu_read_lock();
2186 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2187 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2188 		delegation_type = delegation->type;
2189 	rcu_read_unlock();
2190 	opendata->o_arg.u.delegation_type = delegation_type;
2191 	status = nfs4_open_recover(opendata, state);
2192 	nfs4_opendata_put(opendata);
2193 	return status;
2194 }
2195 
2196 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2197 {
2198 	struct nfs_server *server = NFS_SERVER(state->inode);
2199 	struct nfs4_exception exception = { };
2200 	int err;
2201 	do {
2202 		err = _nfs4_do_open_reclaim(ctx, state);
2203 		trace_nfs4_open_reclaim(ctx, 0, err);
2204 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2205 			continue;
2206 		if (err != -NFS4ERR_DELAY)
2207 			break;
2208 		nfs4_handle_exception(server, err, &exception);
2209 	} while (exception.retry);
2210 	return err;
2211 }
2212 
2213 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2214 {
2215 	struct nfs_open_context *ctx;
2216 	int ret;
2217 
2218 	ctx = nfs4_state_find_open_context(state);
2219 	if (IS_ERR(ctx))
2220 		return -EAGAIN;
2221 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2222 	nfs_state_clear_open_state_flags(state);
2223 	ret = nfs4_do_open_reclaim(ctx, state);
2224 	put_nfs_open_context(ctx);
2225 	return ret;
2226 }
2227 
2228 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2229 {
2230 	switch (err) {
2231 		default:
2232 			printk(KERN_ERR "NFS: %s: unhandled error "
2233 					"%d.\n", __func__, err);
2234 		case 0:
2235 		case -ENOENT:
2236 		case -EAGAIN:
2237 		case -ESTALE:
2238 		case -ETIMEDOUT:
2239 			break;
2240 		case -NFS4ERR_BADSESSION:
2241 		case -NFS4ERR_BADSLOT:
2242 		case -NFS4ERR_BAD_HIGH_SLOT:
2243 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2244 		case -NFS4ERR_DEADSESSION:
2245 			return -EAGAIN;
2246 		case -NFS4ERR_STALE_CLIENTID:
2247 		case -NFS4ERR_STALE_STATEID:
2248 			/* Don't recall a delegation if it was lost */
2249 			nfs4_schedule_lease_recovery(server->nfs_client);
2250 			return -EAGAIN;
2251 		case -NFS4ERR_MOVED:
2252 			nfs4_schedule_migration_recovery(server);
2253 			return -EAGAIN;
2254 		case -NFS4ERR_LEASE_MOVED:
2255 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2256 			return -EAGAIN;
2257 		case -NFS4ERR_DELEG_REVOKED:
2258 		case -NFS4ERR_ADMIN_REVOKED:
2259 		case -NFS4ERR_EXPIRED:
2260 		case -NFS4ERR_BAD_STATEID:
2261 		case -NFS4ERR_OPENMODE:
2262 			nfs_inode_find_state_and_recover(state->inode,
2263 					stateid);
2264 			nfs4_schedule_stateid_recovery(server, state);
2265 			return -EAGAIN;
2266 		case -NFS4ERR_DELAY:
2267 		case -NFS4ERR_GRACE:
2268 			ssleep(1);
2269 			return -EAGAIN;
2270 		case -ENOMEM:
2271 		case -NFS4ERR_DENIED:
2272 			if (fl) {
2273 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2274 				if (lsp)
2275 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2276 			}
2277 			return 0;
2278 	}
2279 	return err;
2280 }
2281 
2282 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2283 		struct nfs4_state *state, const nfs4_stateid *stateid)
2284 {
2285 	struct nfs_server *server = NFS_SERVER(state->inode);
2286 	struct nfs4_opendata *opendata;
2287 	int err = 0;
2288 
2289 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2290 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2291 	if (IS_ERR(opendata))
2292 		return PTR_ERR(opendata);
2293 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2294 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2295 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2296 		if (err)
2297 			goto out;
2298 	}
2299 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2300 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2301 		if (err)
2302 			goto out;
2303 	}
2304 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2305 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2306 		if (err)
2307 			goto out;
2308 	}
2309 	nfs_state_clear_delegation(state);
2310 out:
2311 	nfs4_opendata_put(opendata);
2312 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2313 }
2314 
2315 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2316 {
2317 	struct nfs4_opendata *data = calldata;
2318 
2319 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2320 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2321 }
2322 
2323 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2324 {
2325 	struct nfs4_opendata *data = calldata;
2326 
2327 	nfs40_sequence_done(task, &data->c_res.seq_res);
2328 
2329 	data->rpc_status = task->tk_status;
2330 	if (data->rpc_status == 0) {
2331 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2332 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2333 		renew_lease(data->o_res.server, data->timestamp);
2334 		data->rpc_done = true;
2335 	}
2336 }
2337 
2338 static void nfs4_open_confirm_release(void *calldata)
2339 {
2340 	struct nfs4_opendata *data = calldata;
2341 	struct nfs4_state *state = NULL;
2342 
2343 	/* If this request hasn't been cancelled, do nothing */
2344 	if (!data->cancelled)
2345 		goto out_free;
2346 	/* In case of error, no cleanup! */
2347 	if (!data->rpc_done)
2348 		goto out_free;
2349 	state = nfs4_opendata_to_nfs4_state(data);
2350 	if (!IS_ERR(state))
2351 		nfs4_close_state(state, data->o_arg.fmode);
2352 out_free:
2353 	nfs4_opendata_put(data);
2354 }
2355 
2356 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2357 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2358 	.rpc_call_done = nfs4_open_confirm_done,
2359 	.rpc_release = nfs4_open_confirm_release,
2360 };
2361 
2362 /*
2363  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2364  */
2365 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2366 {
2367 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2368 	struct rpc_task *task;
2369 	struct  rpc_message msg = {
2370 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2371 		.rpc_argp = &data->c_arg,
2372 		.rpc_resp = &data->c_res,
2373 		.rpc_cred = data->owner->so_cred,
2374 	};
2375 	struct rpc_task_setup task_setup_data = {
2376 		.rpc_client = server->client,
2377 		.rpc_message = &msg,
2378 		.callback_ops = &nfs4_open_confirm_ops,
2379 		.callback_data = data,
2380 		.workqueue = nfsiod_workqueue,
2381 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2382 	};
2383 	int status;
2384 
2385 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2386 				data->is_recover);
2387 	kref_get(&data->kref);
2388 	data->rpc_done = false;
2389 	data->rpc_status = 0;
2390 	data->timestamp = jiffies;
2391 	task = rpc_run_task(&task_setup_data);
2392 	if (IS_ERR(task))
2393 		return PTR_ERR(task);
2394 	status = rpc_wait_for_completion_task(task);
2395 	if (status != 0) {
2396 		data->cancelled = true;
2397 		smp_wmb();
2398 	} else
2399 		status = data->rpc_status;
2400 	rpc_put_task(task);
2401 	return status;
2402 }
2403 
2404 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2405 {
2406 	struct nfs4_opendata *data = calldata;
2407 	struct nfs4_state_owner *sp = data->owner;
2408 	struct nfs_client *clp = sp->so_server->nfs_client;
2409 	enum open_claim_type4 claim = data->o_arg.claim;
2410 
2411 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2412 		goto out_wait;
2413 	/*
2414 	 * Check if we still need to send an OPEN call, or if we can use
2415 	 * a delegation instead.
2416 	 */
2417 	if (data->state != NULL) {
2418 		struct nfs_delegation *delegation;
2419 
2420 		if (can_open_cached(data->state, data->o_arg.fmode,
2421 					data->o_arg.open_flags, claim))
2422 			goto out_no_action;
2423 		rcu_read_lock();
2424 		delegation = nfs4_get_valid_delegation(data->state->inode);
2425 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2426 			goto unlock_no_action;
2427 		rcu_read_unlock();
2428 	}
2429 	/* Update client id. */
2430 	data->o_arg.clientid = clp->cl_clientid;
2431 	switch (claim) {
2432 	default:
2433 		break;
2434 	case NFS4_OPEN_CLAIM_PREVIOUS:
2435 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2436 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2437 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2438 		fallthrough;
2439 	case NFS4_OPEN_CLAIM_FH:
2440 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2441 	}
2442 	data->timestamp = jiffies;
2443 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2444 				&data->o_arg.seq_args,
2445 				&data->o_res.seq_res,
2446 				task) != 0)
2447 		nfs_release_seqid(data->o_arg.seqid);
2448 
2449 	/* Set the create mode (note dependency on the session type) */
2450 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2451 	if (data->o_arg.open_flags & O_EXCL) {
2452 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2453 		if (nfs4_has_persistent_session(clp))
2454 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2455 		else if (clp->cl_mvops->minor_version > 0)
2456 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2457 	}
2458 	return;
2459 unlock_no_action:
2460 	trace_nfs4_cached_open(data->state);
2461 	rcu_read_unlock();
2462 out_no_action:
2463 	task->tk_action = NULL;
2464 out_wait:
2465 	nfs4_sequence_done(task, &data->o_res.seq_res);
2466 }
2467 
2468 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2469 {
2470 	struct nfs4_opendata *data = calldata;
2471 
2472 	data->rpc_status = task->tk_status;
2473 
2474 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2475 		return;
2476 
2477 	if (task->tk_status == 0) {
2478 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2479 			switch (data->o_res.f_attr->mode & S_IFMT) {
2480 			case S_IFREG:
2481 				break;
2482 			case S_IFLNK:
2483 				data->rpc_status = -ELOOP;
2484 				break;
2485 			case S_IFDIR:
2486 				data->rpc_status = -EISDIR;
2487 				break;
2488 			default:
2489 				data->rpc_status = -ENOTDIR;
2490 			}
2491 		}
2492 		renew_lease(data->o_res.server, data->timestamp);
2493 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2494 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2495 	}
2496 	data->rpc_done = true;
2497 }
2498 
2499 static void nfs4_open_release(void *calldata)
2500 {
2501 	struct nfs4_opendata *data = calldata;
2502 	struct nfs4_state *state = NULL;
2503 
2504 	/* If this request hasn't been cancelled, do nothing */
2505 	if (!data->cancelled)
2506 		goto out_free;
2507 	/* In case of error, no cleanup! */
2508 	if (data->rpc_status != 0 || !data->rpc_done)
2509 		goto out_free;
2510 	/* In case we need an open_confirm, no cleanup! */
2511 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2512 		goto out_free;
2513 	state = nfs4_opendata_to_nfs4_state(data);
2514 	if (!IS_ERR(state))
2515 		nfs4_close_state(state, data->o_arg.fmode);
2516 out_free:
2517 	nfs4_opendata_put(data);
2518 }
2519 
2520 static const struct rpc_call_ops nfs4_open_ops = {
2521 	.rpc_call_prepare = nfs4_open_prepare,
2522 	.rpc_call_done = nfs4_open_done,
2523 	.rpc_release = nfs4_open_release,
2524 };
2525 
2526 static int nfs4_run_open_task(struct nfs4_opendata *data,
2527 			      struct nfs_open_context *ctx)
2528 {
2529 	struct inode *dir = d_inode(data->dir);
2530 	struct nfs_server *server = NFS_SERVER(dir);
2531 	struct nfs_openargs *o_arg = &data->o_arg;
2532 	struct nfs_openres *o_res = &data->o_res;
2533 	struct rpc_task *task;
2534 	struct rpc_message msg = {
2535 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2536 		.rpc_argp = o_arg,
2537 		.rpc_resp = o_res,
2538 		.rpc_cred = data->owner->so_cred,
2539 	};
2540 	struct rpc_task_setup task_setup_data = {
2541 		.rpc_client = server->client,
2542 		.rpc_message = &msg,
2543 		.callback_ops = &nfs4_open_ops,
2544 		.callback_data = data,
2545 		.workqueue = nfsiod_workqueue,
2546 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2547 	};
2548 	int status;
2549 
2550 	kref_get(&data->kref);
2551 	data->rpc_done = false;
2552 	data->rpc_status = 0;
2553 	data->cancelled = false;
2554 	data->is_recover = false;
2555 	if (!ctx) {
2556 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2557 		data->is_recover = true;
2558 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2559 	} else {
2560 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2561 		pnfs_lgopen_prepare(data, ctx);
2562 	}
2563 	task = rpc_run_task(&task_setup_data);
2564 	if (IS_ERR(task))
2565 		return PTR_ERR(task);
2566 	status = rpc_wait_for_completion_task(task);
2567 	if (status != 0) {
2568 		data->cancelled = true;
2569 		smp_wmb();
2570 	} else
2571 		status = data->rpc_status;
2572 	rpc_put_task(task);
2573 
2574 	return status;
2575 }
2576 
2577 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2578 {
2579 	struct inode *dir = d_inode(data->dir);
2580 	struct nfs_openres *o_res = &data->o_res;
2581 	int status;
2582 
2583 	status = nfs4_run_open_task(data, NULL);
2584 	if (status != 0 || !data->rpc_done)
2585 		return status;
2586 
2587 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2588 
2589 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2590 		status = _nfs4_proc_open_confirm(data);
2591 
2592 	return status;
2593 }
2594 
2595 /*
2596  * Additional permission checks in order to distinguish between an
2597  * open for read, and an open for execute. This works around the
2598  * fact that NFSv4 OPEN treats read and execute permissions as being
2599  * the same.
2600  * Note that in the non-execute case, we want to turn off permission
2601  * checking if we just created a new file (POSIX open() semantics).
2602  */
2603 static int nfs4_opendata_access(const struct cred *cred,
2604 				struct nfs4_opendata *opendata,
2605 				struct nfs4_state *state, fmode_t fmode,
2606 				int openflags)
2607 {
2608 	struct nfs_access_entry cache;
2609 	u32 mask, flags;
2610 
2611 	/* access call failed or for some reason the server doesn't
2612 	 * support any access modes -- defer access call until later */
2613 	if (opendata->o_res.access_supported == 0)
2614 		return 0;
2615 
2616 	mask = 0;
2617 	/*
2618 	 * Use openflags to check for exec, because fmode won't
2619 	 * always have FMODE_EXEC set when file open for exec.
2620 	 */
2621 	if (openflags & __FMODE_EXEC) {
2622 		/* ONLY check for exec rights */
2623 		if (S_ISDIR(state->inode->i_mode))
2624 			mask = NFS4_ACCESS_LOOKUP;
2625 		else
2626 			mask = NFS4_ACCESS_EXECUTE;
2627 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2628 		mask = NFS4_ACCESS_READ;
2629 
2630 	cache.cred = cred;
2631 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2632 	nfs_access_add_cache(state->inode, &cache);
2633 
2634 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2635 	if ((mask & ~cache.mask & flags) == 0)
2636 		return 0;
2637 
2638 	return -EACCES;
2639 }
2640 
2641 /*
2642  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2643  */
2644 static int _nfs4_proc_open(struct nfs4_opendata *data,
2645 			   struct nfs_open_context *ctx)
2646 {
2647 	struct inode *dir = d_inode(data->dir);
2648 	struct nfs_server *server = NFS_SERVER(dir);
2649 	struct nfs_openargs *o_arg = &data->o_arg;
2650 	struct nfs_openres *o_res = &data->o_res;
2651 	int status;
2652 
2653 	status = nfs4_run_open_task(data, ctx);
2654 	if (!data->rpc_done)
2655 		return status;
2656 	if (status != 0) {
2657 		if (status == -NFS4ERR_BADNAME &&
2658 				!(o_arg->open_flags & O_CREAT))
2659 			return -ENOENT;
2660 		return status;
2661 	}
2662 
2663 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2664 
2665 	if (o_arg->open_flags & O_CREAT) {
2666 		if (o_arg->open_flags & O_EXCL)
2667 			data->file_created = true;
2668 		else if (o_res->cinfo.before != o_res->cinfo.after)
2669 			data->file_created = true;
2670 		if (data->file_created ||
2671 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2672 			nfs4_update_changeattr(dir, &o_res->cinfo,
2673 					o_res->f_attr->time_start,
2674 					NFS_INO_INVALID_DATA);
2675 	}
2676 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2677 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2678 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2679 		status = _nfs4_proc_open_confirm(data);
2680 		if (status != 0)
2681 			return status;
2682 	}
2683 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2684 		nfs4_sequence_free_slot(&o_res->seq_res);
2685 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2686 				o_res->f_label, NULL);
2687 	}
2688 	return 0;
2689 }
2690 
2691 /*
2692  * OPEN_EXPIRED:
2693  * 	reclaim state on the server after a network partition.
2694  * 	Assumes caller holds the appropriate lock
2695  */
2696 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2697 {
2698 	struct nfs4_opendata *opendata;
2699 	int ret;
2700 
2701 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2702 			NFS4_OPEN_CLAIM_FH);
2703 	if (IS_ERR(opendata))
2704 		return PTR_ERR(opendata);
2705 	ret = nfs4_open_recover(opendata, state);
2706 	if (ret == -ESTALE)
2707 		d_drop(ctx->dentry);
2708 	nfs4_opendata_put(opendata);
2709 	return ret;
2710 }
2711 
2712 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2713 {
2714 	struct nfs_server *server = NFS_SERVER(state->inode);
2715 	struct nfs4_exception exception = { };
2716 	int err;
2717 
2718 	do {
2719 		err = _nfs4_open_expired(ctx, state);
2720 		trace_nfs4_open_expired(ctx, 0, err);
2721 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2722 			continue;
2723 		switch (err) {
2724 		default:
2725 			goto out;
2726 		case -NFS4ERR_GRACE:
2727 		case -NFS4ERR_DELAY:
2728 			nfs4_handle_exception(server, err, &exception);
2729 			err = 0;
2730 		}
2731 	} while (exception.retry);
2732 out:
2733 	return err;
2734 }
2735 
2736 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2737 {
2738 	struct nfs_open_context *ctx;
2739 	int ret;
2740 
2741 	ctx = nfs4_state_find_open_context(state);
2742 	if (IS_ERR(ctx))
2743 		return -EAGAIN;
2744 	ret = nfs4_do_open_expired(ctx, state);
2745 	put_nfs_open_context(ctx);
2746 	return ret;
2747 }
2748 
2749 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2750 		const nfs4_stateid *stateid)
2751 {
2752 	nfs_remove_bad_delegation(state->inode, stateid);
2753 	nfs_state_clear_delegation(state);
2754 }
2755 
2756 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2757 {
2758 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2759 		nfs_finish_clear_delegation_stateid(state, NULL);
2760 }
2761 
2762 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2763 {
2764 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2765 	nfs40_clear_delegation_stateid(state);
2766 	nfs_state_clear_open_state_flags(state);
2767 	return nfs4_open_expired(sp, state);
2768 }
2769 
2770 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2771 		nfs4_stateid *stateid,
2772 		const struct cred *cred)
2773 {
2774 	return -NFS4ERR_BAD_STATEID;
2775 }
2776 
2777 #if defined(CONFIG_NFS_V4_1)
2778 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2779 		nfs4_stateid *stateid,
2780 		const struct cred *cred)
2781 {
2782 	int status;
2783 
2784 	switch (stateid->type) {
2785 	default:
2786 		break;
2787 	case NFS4_INVALID_STATEID_TYPE:
2788 	case NFS4_SPECIAL_STATEID_TYPE:
2789 		return -NFS4ERR_BAD_STATEID;
2790 	case NFS4_REVOKED_STATEID_TYPE:
2791 		goto out_free;
2792 	}
2793 
2794 	status = nfs41_test_stateid(server, stateid, cred);
2795 	switch (status) {
2796 	case -NFS4ERR_EXPIRED:
2797 	case -NFS4ERR_ADMIN_REVOKED:
2798 	case -NFS4ERR_DELEG_REVOKED:
2799 		break;
2800 	default:
2801 		return status;
2802 	}
2803 out_free:
2804 	/* Ack the revoked state to the server */
2805 	nfs41_free_stateid(server, stateid, cred, true);
2806 	return -NFS4ERR_EXPIRED;
2807 }
2808 
2809 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2810 {
2811 	struct nfs_server *server = NFS_SERVER(state->inode);
2812 	nfs4_stateid stateid;
2813 	struct nfs_delegation *delegation;
2814 	const struct cred *cred = NULL;
2815 	int status, ret = NFS_OK;
2816 
2817 	/* Get the delegation credential for use by test/free_stateid */
2818 	rcu_read_lock();
2819 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2820 	if (delegation == NULL) {
2821 		rcu_read_unlock();
2822 		nfs_state_clear_delegation(state);
2823 		return NFS_OK;
2824 	}
2825 
2826 	spin_lock(&delegation->lock);
2827 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2828 
2829 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2830 				&delegation->flags)) {
2831 		spin_unlock(&delegation->lock);
2832 		rcu_read_unlock();
2833 		return NFS_OK;
2834 	}
2835 
2836 	if (delegation->cred)
2837 		cred = get_cred(delegation->cred);
2838 	spin_unlock(&delegation->lock);
2839 	rcu_read_unlock();
2840 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2841 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2842 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2843 		nfs_finish_clear_delegation_stateid(state, &stateid);
2844 	else
2845 		ret = status;
2846 
2847 	put_cred(cred);
2848 	return ret;
2849 }
2850 
2851 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2852 {
2853 	nfs4_stateid tmp;
2854 
2855 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2856 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2857 				&tmp, NULL) &&
2858 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2859 		nfs_state_set_delegation(state, &tmp, state->state);
2860 	else
2861 		nfs_state_clear_delegation(state);
2862 }
2863 
2864 /**
2865  * nfs41_check_expired_locks - possibly free a lock stateid
2866  *
2867  * @state: NFSv4 state for an inode
2868  *
2869  * Returns NFS_OK if recovery for this stateid is now finished.
2870  * Otherwise a negative NFS4ERR value is returned.
2871  */
2872 static int nfs41_check_expired_locks(struct nfs4_state *state)
2873 {
2874 	int status, ret = NFS_OK;
2875 	struct nfs4_lock_state *lsp, *prev = NULL;
2876 	struct nfs_server *server = NFS_SERVER(state->inode);
2877 
2878 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2879 		goto out;
2880 
2881 	spin_lock(&state->state_lock);
2882 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2883 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2884 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2885 
2886 			refcount_inc(&lsp->ls_count);
2887 			spin_unlock(&state->state_lock);
2888 
2889 			nfs4_put_lock_state(prev);
2890 			prev = lsp;
2891 
2892 			status = nfs41_test_and_free_expired_stateid(server,
2893 					&lsp->ls_stateid,
2894 					cred);
2895 			trace_nfs4_test_lock_stateid(state, lsp, status);
2896 			if (status == -NFS4ERR_EXPIRED ||
2897 			    status == -NFS4ERR_BAD_STATEID) {
2898 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2899 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2900 				if (!recover_lost_locks)
2901 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2902 			} else if (status != NFS_OK) {
2903 				ret = status;
2904 				nfs4_put_lock_state(prev);
2905 				goto out;
2906 			}
2907 			spin_lock(&state->state_lock);
2908 		}
2909 	}
2910 	spin_unlock(&state->state_lock);
2911 	nfs4_put_lock_state(prev);
2912 out:
2913 	return ret;
2914 }
2915 
2916 /**
2917  * nfs41_check_open_stateid - possibly free an open stateid
2918  *
2919  * @state: NFSv4 state for an inode
2920  *
2921  * Returns NFS_OK if recovery for this stateid is now finished.
2922  * Otherwise a negative NFS4ERR value is returned.
2923  */
2924 static int nfs41_check_open_stateid(struct nfs4_state *state)
2925 {
2926 	struct nfs_server *server = NFS_SERVER(state->inode);
2927 	nfs4_stateid *stateid = &state->open_stateid;
2928 	const struct cred *cred = state->owner->so_cred;
2929 	int status;
2930 
2931 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2932 		return -NFS4ERR_BAD_STATEID;
2933 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2934 	trace_nfs4_test_open_stateid(state, NULL, status);
2935 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2936 		nfs_state_clear_open_state_flags(state);
2937 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2938 		return status;
2939 	}
2940 	if (nfs_open_stateid_recover_openmode(state))
2941 		return -NFS4ERR_OPENMODE;
2942 	return NFS_OK;
2943 }
2944 
2945 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2946 {
2947 	int status;
2948 
2949 	status = nfs41_check_delegation_stateid(state);
2950 	if (status != NFS_OK)
2951 		return status;
2952 	nfs41_delegation_recover_stateid(state);
2953 
2954 	status = nfs41_check_expired_locks(state);
2955 	if (status != NFS_OK)
2956 		return status;
2957 	status = nfs41_check_open_stateid(state);
2958 	if (status != NFS_OK)
2959 		status = nfs4_open_expired(sp, state);
2960 	return status;
2961 }
2962 #endif
2963 
2964 /*
2965  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2966  * fields corresponding to attributes that were used to store the verifier.
2967  * Make sure we clobber those fields in the later setattr call
2968  */
2969 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2970 				struct iattr *sattr, struct nfs4_label **label)
2971 {
2972 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2973 	__u32 attrset[3];
2974 	unsigned ret;
2975 	unsigned i;
2976 
2977 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2978 		attrset[i] = opendata->o_res.attrset[i];
2979 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2980 			attrset[i] &= ~bitmask[i];
2981 	}
2982 
2983 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2984 		sattr->ia_valid : 0;
2985 
2986 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2987 		if (sattr->ia_valid & ATTR_ATIME_SET)
2988 			ret |= ATTR_ATIME_SET;
2989 		else
2990 			ret |= ATTR_ATIME;
2991 	}
2992 
2993 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2994 		if (sattr->ia_valid & ATTR_MTIME_SET)
2995 			ret |= ATTR_MTIME_SET;
2996 		else
2997 			ret |= ATTR_MTIME;
2998 	}
2999 
3000 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3001 		*label = NULL;
3002 	return ret;
3003 }
3004 
3005 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3006 		int flags, struct nfs_open_context *ctx)
3007 {
3008 	struct nfs4_state_owner *sp = opendata->owner;
3009 	struct nfs_server *server = sp->so_server;
3010 	struct dentry *dentry;
3011 	struct nfs4_state *state;
3012 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3013 	struct inode *dir = d_inode(opendata->dir);
3014 	unsigned long dir_verifier;
3015 	unsigned int seq;
3016 	int ret;
3017 
3018 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3019 	dir_verifier = nfs_save_change_attribute(dir);
3020 
3021 	ret = _nfs4_proc_open(opendata, ctx);
3022 	if (ret != 0)
3023 		goto out;
3024 
3025 	state = _nfs4_opendata_to_nfs4_state(opendata);
3026 	ret = PTR_ERR(state);
3027 	if (IS_ERR(state))
3028 		goto out;
3029 	ctx->state = state;
3030 	if (server->caps & NFS_CAP_POSIX_LOCK)
3031 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3032 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3033 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3034 
3035 	dentry = opendata->dentry;
3036 	if (d_really_is_negative(dentry)) {
3037 		struct dentry *alias;
3038 		d_drop(dentry);
3039 		alias = d_exact_alias(dentry, state->inode);
3040 		if (!alias)
3041 			alias = d_splice_alias(igrab(state->inode), dentry);
3042 		/* d_splice_alias() can't fail here - it's a non-directory */
3043 		if (alias) {
3044 			dput(ctx->dentry);
3045 			ctx->dentry = dentry = alias;
3046 		}
3047 	}
3048 
3049 	switch(opendata->o_arg.claim) {
3050 	default:
3051 		break;
3052 	case NFS4_OPEN_CLAIM_NULL:
3053 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3054 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3055 		if (!opendata->rpc_done)
3056 			break;
3057 		if (opendata->o_res.delegation_type != 0)
3058 			dir_verifier = nfs_save_change_attribute(dir);
3059 		nfs_set_verifier(dentry, dir_verifier);
3060 	}
3061 
3062 	/* Parse layoutget results before we check for access */
3063 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3064 
3065 	ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3066 			acc_mode, flags);
3067 	if (ret != 0)
3068 		goto out;
3069 
3070 	if (d_inode(dentry) == state->inode) {
3071 		nfs_inode_attach_open_context(ctx);
3072 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3073 			nfs4_schedule_stateid_recovery(server, state);
3074 	}
3075 
3076 out:
3077 	if (!opendata->cancelled)
3078 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3079 	return ret;
3080 }
3081 
3082 /*
3083  * Returns a referenced nfs4_state
3084  */
3085 static int _nfs4_do_open(struct inode *dir,
3086 			struct nfs_open_context *ctx,
3087 			int flags,
3088 			const struct nfs4_open_createattrs *c,
3089 			int *opened)
3090 {
3091 	struct nfs4_state_owner  *sp;
3092 	struct nfs4_state     *state = NULL;
3093 	struct nfs_server       *server = NFS_SERVER(dir);
3094 	struct nfs4_opendata *opendata;
3095 	struct dentry *dentry = ctx->dentry;
3096 	const struct cred *cred = ctx->cred;
3097 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3098 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3099 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3100 	struct iattr *sattr = c->sattr;
3101 	struct nfs4_label *label = c->label;
3102 	struct nfs4_label *olabel = NULL;
3103 	int status;
3104 
3105 	/* Protect against reboot recovery conflicts */
3106 	status = -ENOMEM;
3107 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3108 	if (sp == NULL) {
3109 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3110 		goto out_err;
3111 	}
3112 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3113 	if (status != 0)
3114 		goto err_put_state_owner;
3115 	if (d_really_is_positive(dentry))
3116 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3117 	status = -ENOMEM;
3118 	if (d_really_is_positive(dentry))
3119 		claim = NFS4_OPEN_CLAIM_FH;
3120 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3121 			c, claim, GFP_KERNEL);
3122 	if (opendata == NULL)
3123 		goto err_put_state_owner;
3124 
3125 	if (label) {
3126 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
3127 		if (IS_ERR(olabel)) {
3128 			status = PTR_ERR(olabel);
3129 			goto err_opendata_put;
3130 		}
3131 	}
3132 
3133 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3134 		if (!opendata->f_attr.mdsthreshold) {
3135 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3136 			if (!opendata->f_attr.mdsthreshold)
3137 				goto err_free_label;
3138 		}
3139 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3140 	}
3141 	if (d_really_is_positive(dentry))
3142 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3143 
3144 	status = _nfs4_open_and_get_state(opendata, flags, ctx);
3145 	if (status != 0)
3146 		goto err_free_label;
3147 	state = ctx->state;
3148 
3149 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3150 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3151 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3152 		/*
3153 		 * send create attributes which was not set by open
3154 		 * with an extra setattr.
3155 		 */
3156 		if (attrs || label) {
3157 			unsigned ia_old = sattr->ia_valid;
3158 
3159 			sattr->ia_valid = attrs;
3160 			nfs_fattr_init(opendata->o_res.f_attr);
3161 			status = nfs4_do_setattr(state->inode, cred,
3162 					opendata->o_res.f_attr, sattr,
3163 					ctx, label, olabel);
3164 			if (status == 0) {
3165 				nfs_setattr_update_inode(state->inode, sattr,
3166 						opendata->o_res.f_attr);
3167 				nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3168 			}
3169 			sattr->ia_valid = ia_old;
3170 		}
3171 	}
3172 	if (opened && opendata->file_created)
3173 		*opened = 1;
3174 
3175 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3176 		*ctx_th = opendata->f_attr.mdsthreshold;
3177 		opendata->f_attr.mdsthreshold = NULL;
3178 	}
3179 
3180 	nfs4_label_free(olabel);
3181 
3182 	nfs4_opendata_put(opendata);
3183 	nfs4_put_state_owner(sp);
3184 	return 0;
3185 err_free_label:
3186 	nfs4_label_free(olabel);
3187 err_opendata_put:
3188 	nfs4_opendata_put(opendata);
3189 err_put_state_owner:
3190 	nfs4_put_state_owner(sp);
3191 out_err:
3192 	return status;
3193 }
3194 
3195 
3196 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3197 					struct nfs_open_context *ctx,
3198 					int flags,
3199 					struct iattr *sattr,
3200 					struct nfs4_label *label,
3201 					int *opened)
3202 {
3203 	struct nfs_server *server = NFS_SERVER(dir);
3204 	struct nfs4_exception exception = {
3205 		.interruptible = true,
3206 	};
3207 	struct nfs4_state *res;
3208 	struct nfs4_open_createattrs c = {
3209 		.label = label,
3210 		.sattr = sattr,
3211 		.verf = {
3212 			[0] = (__u32)jiffies,
3213 			[1] = (__u32)current->pid,
3214 		},
3215 	};
3216 	int status;
3217 
3218 	do {
3219 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3220 		res = ctx->state;
3221 		trace_nfs4_open_file(ctx, flags, status);
3222 		if (status == 0)
3223 			break;
3224 		/* NOTE: BAD_SEQID means the server and client disagree about the
3225 		 * book-keeping w.r.t. state-changing operations
3226 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3227 		 * It is actually a sign of a bug on the client or on the server.
3228 		 *
3229 		 * If we receive a BAD_SEQID error in the particular case of
3230 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3231 		 * have unhashed the old state_owner for us, and that we can
3232 		 * therefore safely retry using a new one. We should still warn
3233 		 * the user though...
3234 		 */
3235 		if (status == -NFS4ERR_BAD_SEQID) {
3236 			pr_warn_ratelimited("NFS: v4 server %s "
3237 					" returned a bad sequence-id error!\n",
3238 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3239 			exception.retry = 1;
3240 			continue;
3241 		}
3242 		/*
3243 		 * BAD_STATEID on OPEN means that the server cancelled our
3244 		 * state before it received the OPEN_CONFIRM.
3245 		 * Recover by retrying the request as per the discussion
3246 		 * on Page 181 of RFC3530.
3247 		 */
3248 		if (status == -NFS4ERR_BAD_STATEID) {
3249 			exception.retry = 1;
3250 			continue;
3251 		}
3252 		if (status == -NFS4ERR_EXPIRED) {
3253 			nfs4_schedule_lease_recovery(server->nfs_client);
3254 			exception.retry = 1;
3255 			continue;
3256 		}
3257 		if (status == -EAGAIN) {
3258 			/* We must have found a delegation */
3259 			exception.retry = 1;
3260 			continue;
3261 		}
3262 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3263 			continue;
3264 		res = ERR_PTR(nfs4_handle_exception(server,
3265 					status, &exception));
3266 	} while (exception.retry);
3267 	return res;
3268 }
3269 
3270 static int _nfs4_do_setattr(struct inode *inode,
3271 			    struct nfs_setattrargs *arg,
3272 			    struct nfs_setattrres *res,
3273 			    const struct cred *cred,
3274 			    struct nfs_open_context *ctx)
3275 {
3276 	struct nfs_server *server = NFS_SERVER(inode);
3277 	struct rpc_message msg = {
3278 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3279 		.rpc_argp	= arg,
3280 		.rpc_resp	= res,
3281 		.rpc_cred	= cred,
3282 	};
3283 	const struct cred *delegation_cred = NULL;
3284 	unsigned long timestamp = jiffies;
3285 	bool truncate;
3286 	int status;
3287 
3288 	nfs_fattr_init(res->fattr);
3289 
3290 	/* Servers should only apply open mode checks for file size changes */
3291 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3292 	if (!truncate) {
3293 		nfs4_inode_make_writeable(inode);
3294 		goto zero_stateid;
3295 	}
3296 
3297 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3298 		/* Use that stateid */
3299 	} else if (ctx != NULL && ctx->state) {
3300 		struct nfs_lock_context *l_ctx;
3301 		if (!nfs4_valid_open_stateid(ctx->state))
3302 			return -EBADF;
3303 		l_ctx = nfs_get_lock_context(ctx);
3304 		if (IS_ERR(l_ctx))
3305 			return PTR_ERR(l_ctx);
3306 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3307 						&arg->stateid, &delegation_cred);
3308 		nfs_put_lock_context(l_ctx);
3309 		if (status == -EIO)
3310 			return -EBADF;
3311 		else if (status == -EAGAIN)
3312 			goto zero_stateid;
3313 	} else {
3314 zero_stateid:
3315 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3316 	}
3317 	if (delegation_cred)
3318 		msg.rpc_cred = delegation_cred;
3319 
3320 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3321 
3322 	put_cred(delegation_cred);
3323 	if (status == 0 && ctx != NULL)
3324 		renew_lease(server, timestamp);
3325 	trace_nfs4_setattr(inode, &arg->stateid, status);
3326 	return status;
3327 }
3328 
3329 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3330 			   struct nfs_fattr *fattr, struct iattr *sattr,
3331 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3332 			   struct nfs4_label *olabel)
3333 {
3334 	struct nfs_server *server = NFS_SERVER(inode);
3335 	__u32 bitmask[NFS4_BITMASK_SZ];
3336 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3337 	struct nfs_setattrargs	arg = {
3338 		.fh		= NFS_FH(inode),
3339 		.iap		= sattr,
3340 		.server		= server,
3341 		.bitmask = bitmask,
3342 		.label		= ilabel,
3343 	};
3344 	struct nfs_setattrres  res = {
3345 		.fattr		= fattr,
3346 		.label		= olabel,
3347 		.server		= server,
3348 	};
3349 	struct nfs4_exception exception = {
3350 		.state = state,
3351 		.inode = inode,
3352 		.stateid = &arg.stateid,
3353 	};
3354 	int err;
3355 
3356 	do {
3357 		nfs4_bitmap_copy_adjust_setattr(bitmask,
3358 				nfs4_bitmask(server, olabel),
3359 				inode);
3360 
3361 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3362 		switch (err) {
3363 		case -NFS4ERR_OPENMODE:
3364 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3365 				pr_warn_once("NFSv4: server %s is incorrectly "
3366 						"applying open mode checks to "
3367 						"a SETATTR that is not "
3368 						"changing file size.\n",
3369 						server->nfs_client->cl_hostname);
3370 			}
3371 			if (state && !(state->state & FMODE_WRITE)) {
3372 				err = -EBADF;
3373 				if (sattr->ia_valid & ATTR_OPEN)
3374 					err = -EACCES;
3375 				goto out;
3376 			}
3377 		}
3378 		err = nfs4_handle_exception(server, err, &exception);
3379 	} while (exception.retry);
3380 out:
3381 	return err;
3382 }
3383 
3384 static bool
3385 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3386 {
3387 	if (inode == NULL || !nfs_have_layout(inode))
3388 		return false;
3389 
3390 	return pnfs_wait_on_layoutreturn(inode, task);
3391 }
3392 
3393 /*
3394  * Update the seqid of an open stateid
3395  */
3396 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3397 		struct nfs4_state *state)
3398 {
3399 	__be32 seqid_open;
3400 	u32 dst_seqid;
3401 	int seq;
3402 
3403 	for (;;) {
3404 		if (!nfs4_valid_open_stateid(state))
3405 			break;
3406 		seq = read_seqbegin(&state->seqlock);
3407 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3408 			nfs4_stateid_copy(dst, &state->open_stateid);
3409 			if (read_seqretry(&state->seqlock, seq))
3410 				continue;
3411 			break;
3412 		}
3413 		seqid_open = state->open_stateid.seqid;
3414 		if (read_seqretry(&state->seqlock, seq))
3415 			continue;
3416 
3417 		dst_seqid = be32_to_cpu(dst->seqid);
3418 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3419 			dst->seqid = seqid_open;
3420 		break;
3421 	}
3422 }
3423 
3424 /*
3425  * Update the seqid of an open stateid after receiving
3426  * NFS4ERR_OLD_STATEID
3427  */
3428 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3429 		struct nfs4_state *state)
3430 {
3431 	__be32 seqid_open;
3432 	u32 dst_seqid;
3433 	bool ret;
3434 	int seq, status = -EAGAIN;
3435 	DEFINE_WAIT(wait);
3436 
3437 	for (;;) {
3438 		ret = false;
3439 		if (!nfs4_valid_open_stateid(state))
3440 			break;
3441 		seq = read_seqbegin(&state->seqlock);
3442 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3443 			if (read_seqretry(&state->seqlock, seq))
3444 				continue;
3445 			break;
3446 		}
3447 
3448 		write_seqlock(&state->seqlock);
3449 		seqid_open = state->open_stateid.seqid;
3450 
3451 		dst_seqid = be32_to_cpu(dst->seqid);
3452 
3453 		/* Did another OPEN bump the state's seqid?  try again: */
3454 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3455 			dst->seqid = seqid_open;
3456 			write_sequnlock(&state->seqlock);
3457 			ret = true;
3458 			break;
3459 		}
3460 
3461 		/* server says we're behind but we haven't seen the update yet */
3462 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3463 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3464 		write_sequnlock(&state->seqlock);
3465 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3466 
3467 		if (signal_pending(current))
3468 			status = -EINTR;
3469 		else
3470 			if (schedule_timeout(5*HZ) != 0)
3471 				status = 0;
3472 
3473 		finish_wait(&state->waitq, &wait);
3474 
3475 		if (!status)
3476 			continue;
3477 		if (status == -EINTR)
3478 			break;
3479 
3480 		/* we slept the whole 5 seconds, we must have lost a seqid */
3481 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3482 		ret = true;
3483 		break;
3484 	}
3485 
3486 	return ret;
3487 }
3488 
3489 struct nfs4_closedata {
3490 	struct inode *inode;
3491 	struct nfs4_state *state;
3492 	struct nfs_closeargs arg;
3493 	struct nfs_closeres res;
3494 	struct {
3495 		struct nfs4_layoutreturn_args arg;
3496 		struct nfs4_layoutreturn_res res;
3497 		struct nfs4_xdr_opaque_data ld_private;
3498 		u32 roc_barrier;
3499 		bool roc;
3500 	} lr;
3501 	struct nfs_fattr fattr;
3502 	unsigned long timestamp;
3503 };
3504 
3505 static void nfs4_free_closedata(void *data)
3506 {
3507 	struct nfs4_closedata *calldata = data;
3508 	struct nfs4_state_owner *sp = calldata->state->owner;
3509 	struct super_block *sb = calldata->state->inode->i_sb;
3510 
3511 	if (calldata->lr.roc)
3512 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3513 				calldata->res.lr_ret);
3514 	nfs4_put_open_state(calldata->state);
3515 	nfs_free_seqid(calldata->arg.seqid);
3516 	nfs4_put_state_owner(sp);
3517 	nfs_sb_deactive(sb);
3518 	kfree(calldata);
3519 }
3520 
3521 static void nfs4_close_done(struct rpc_task *task, void *data)
3522 {
3523 	struct nfs4_closedata *calldata = data;
3524 	struct nfs4_state *state = calldata->state;
3525 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3526 	nfs4_stateid *res_stateid = NULL;
3527 	struct nfs4_exception exception = {
3528 		.state = state,
3529 		.inode = calldata->inode,
3530 		.stateid = &calldata->arg.stateid,
3531 	};
3532 
3533 	dprintk("%s: begin!\n", __func__);
3534 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3535 		return;
3536 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3537 
3538 	/* Handle Layoutreturn errors */
3539 	if (pnfs_roc_done(task, calldata->inode,
3540 				&calldata->arg.lr_args,
3541 				&calldata->res.lr_res,
3542 				&calldata->res.lr_ret) == -EAGAIN)
3543 		goto out_restart;
3544 
3545 	/* hmm. we are done with the inode, and in the process of freeing
3546 	 * the state_owner. we keep this around to process errors
3547 	 */
3548 	switch (task->tk_status) {
3549 		case 0:
3550 			res_stateid = &calldata->res.stateid;
3551 			renew_lease(server, calldata->timestamp);
3552 			break;
3553 		case -NFS4ERR_ACCESS:
3554 			if (calldata->arg.bitmask != NULL) {
3555 				calldata->arg.bitmask = NULL;
3556 				calldata->res.fattr = NULL;
3557 				goto out_restart;
3558 
3559 			}
3560 			break;
3561 		case -NFS4ERR_OLD_STATEID:
3562 			/* Did we race with OPEN? */
3563 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3564 						state))
3565 				goto out_restart;
3566 			goto out_release;
3567 		case -NFS4ERR_ADMIN_REVOKED:
3568 		case -NFS4ERR_STALE_STATEID:
3569 		case -NFS4ERR_EXPIRED:
3570 			nfs4_free_revoked_stateid(server,
3571 					&calldata->arg.stateid,
3572 					task->tk_msg.rpc_cred);
3573 			fallthrough;
3574 		case -NFS4ERR_BAD_STATEID:
3575 			if (calldata->arg.fmode == 0)
3576 				break;
3577 			fallthrough;
3578 		default:
3579 			task->tk_status = nfs4_async_handle_exception(task,
3580 					server, task->tk_status, &exception);
3581 			if (exception.retry)
3582 				goto out_restart;
3583 	}
3584 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3585 			res_stateid, calldata->arg.fmode);
3586 out_release:
3587 	task->tk_status = 0;
3588 	nfs_release_seqid(calldata->arg.seqid);
3589 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3590 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3591 	return;
3592 out_restart:
3593 	task->tk_status = 0;
3594 	rpc_restart_call_prepare(task);
3595 	goto out_release;
3596 }
3597 
3598 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3599 {
3600 	struct nfs4_closedata *calldata = data;
3601 	struct nfs4_state *state = calldata->state;
3602 	struct inode *inode = calldata->inode;
3603 	struct pnfs_layout_hdr *lo;
3604 	bool is_rdonly, is_wronly, is_rdwr;
3605 	int call_close = 0;
3606 
3607 	dprintk("%s: begin!\n", __func__);
3608 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3609 		goto out_wait;
3610 
3611 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3612 	spin_lock(&state->owner->so_lock);
3613 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3614 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3615 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3616 	/* Calculate the change in open mode */
3617 	calldata->arg.fmode = 0;
3618 	if (state->n_rdwr == 0) {
3619 		if (state->n_rdonly == 0)
3620 			call_close |= is_rdonly;
3621 		else if (is_rdonly)
3622 			calldata->arg.fmode |= FMODE_READ;
3623 		if (state->n_wronly == 0)
3624 			call_close |= is_wronly;
3625 		else if (is_wronly)
3626 			calldata->arg.fmode |= FMODE_WRITE;
3627 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3628 			call_close |= is_rdwr;
3629 	} else if (is_rdwr)
3630 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3631 
3632 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3633 	if (!nfs4_valid_open_stateid(state))
3634 		call_close = 0;
3635 	spin_unlock(&state->owner->so_lock);
3636 
3637 	if (!call_close) {
3638 		/* Note: exit _without_ calling nfs4_close_done */
3639 		goto out_no_action;
3640 	}
3641 
3642 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3643 		nfs_release_seqid(calldata->arg.seqid);
3644 		goto out_wait;
3645 	}
3646 
3647 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3648 	if (lo && !pnfs_layout_is_valid(lo)) {
3649 		calldata->arg.lr_args = NULL;
3650 		calldata->res.lr_res = NULL;
3651 	}
3652 
3653 	if (calldata->arg.fmode == 0)
3654 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3655 
3656 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3657 		/* Close-to-open cache consistency revalidation */
3658 		if (!nfs4_have_delegation(inode, FMODE_READ)) {
3659 			calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3660 			nfs4_bitmask_adjust(calldata->arg.bitmask, inode, NFS_SERVER(inode), NULL);
3661 		} else
3662 			calldata->arg.bitmask = NULL;
3663 	}
3664 
3665 	calldata->arg.share_access =
3666 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3667 				calldata->arg.fmode, 0);
3668 
3669 	if (calldata->res.fattr == NULL)
3670 		calldata->arg.bitmask = NULL;
3671 	else if (calldata->arg.bitmask == NULL)
3672 		calldata->res.fattr = NULL;
3673 	calldata->timestamp = jiffies;
3674 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3675 				&calldata->arg.seq_args,
3676 				&calldata->res.seq_res,
3677 				task) != 0)
3678 		nfs_release_seqid(calldata->arg.seqid);
3679 	dprintk("%s: done!\n", __func__);
3680 	return;
3681 out_no_action:
3682 	task->tk_action = NULL;
3683 out_wait:
3684 	nfs4_sequence_done(task, &calldata->res.seq_res);
3685 }
3686 
3687 static const struct rpc_call_ops nfs4_close_ops = {
3688 	.rpc_call_prepare = nfs4_close_prepare,
3689 	.rpc_call_done = nfs4_close_done,
3690 	.rpc_release = nfs4_free_closedata,
3691 };
3692 
3693 /*
3694  * It is possible for data to be read/written from a mem-mapped file
3695  * after the sys_close call (which hits the vfs layer as a flush).
3696  * This means that we can't safely call nfsv4 close on a file until
3697  * the inode is cleared. This in turn means that we are not good
3698  * NFSv4 citizens - we do not indicate to the server to update the file's
3699  * share state even when we are done with one of the three share
3700  * stateid's in the inode.
3701  *
3702  * NOTE: Caller must be holding the sp->so_owner semaphore!
3703  */
3704 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3705 {
3706 	struct nfs_server *server = NFS_SERVER(state->inode);
3707 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3708 	struct nfs4_closedata *calldata;
3709 	struct nfs4_state_owner *sp = state->owner;
3710 	struct rpc_task *task;
3711 	struct rpc_message msg = {
3712 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3713 		.rpc_cred = state->owner->so_cred,
3714 	};
3715 	struct rpc_task_setup task_setup_data = {
3716 		.rpc_client = server->client,
3717 		.rpc_message = &msg,
3718 		.callback_ops = &nfs4_close_ops,
3719 		.workqueue = nfsiod_workqueue,
3720 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3721 	};
3722 	int status = -ENOMEM;
3723 
3724 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3725 		&task_setup_data.rpc_client, &msg);
3726 
3727 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3728 	if (calldata == NULL)
3729 		goto out;
3730 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3731 	calldata->inode = state->inode;
3732 	calldata->state = state;
3733 	calldata->arg.fh = NFS_FH(state->inode);
3734 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3735 		goto out_free_calldata;
3736 	/* Serialization for the sequence id */
3737 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3738 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3739 	if (IS_ERR(calldata->arg.seqid))
3740 		goto out_free_calldata;
3741 	nfs_fattr_init(&calldata->fattr);
3742 	calldata->arg.fmode = 0;
3743 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3744 	calldata->res.fattr = &calldata->fattr;
3745 	calldata->res.seqid = calldata->arg.seqid;
3746 	calldata->res.server = server;
3747 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3748 	calldata->lr.roc = pnfs_roc(state->inode,
3749 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3750 	if (calldata->lr.roc) {
3751 		calldata->arg.lr_args = &calldata->lr.arg;
3752 		calldata->res.lr_res = &calldata->lr.res;
3753 	}
3754 	nfs_sb_active(calldata->inode->i_sb);
3755 
3756 	msg.rpc_argp = &calldata->arg;
3757 	msg.rpc_resp = &calldata->res;
3758 	task_setup_data.callback_data = calldata;
3759 	task = rpc_run_task(&task_setup_data);
3760 	if (IS_ERR(task))
3761 		return PTR_ERR(task);
3762 	status = 0;
3763 	if (wait)
3764 		status = rpc_wait_for_completion_task(task);
3765 	rpc_put_task(task);
3766 	return status;
3767 out_free_calldata:
3768 	kfree(calldata);
3769 out:
3770 	nfs4_put_open_state(state);
3771 	nfs4_put_state_owner(sp);
3772 	return status;
3773 }
3774 
3775 static struct inode *
3776 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3777 		int open_flags, struct iattr *attr, int *opened)
3778 {
3779 	struct nfs4_state *state;
3780 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3781 
3782 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3783 
3784 	/* Protect against concurrent sillydeletes */
3785 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3786 
3787 	nfs4_label_release_security(label);
3788 
3789 	if (IS_ERR(state))
3790 		return ERR_CAST(state);
3791 	return state->inode;
3792 }
3793 
3794 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3795 {
3796 	if (ctx->state == NULL)
3797 		return;
3798 	if (is_sync)
3799 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3800 	else
3801 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3802 }
3803 
3804 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3805 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3806 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3807 
3808 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3809 {
3810 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3811 	struct nfs4_server_caps_arg args = {
3812 		.fhandle = fhandle,
3813 		.bitmask = bitmask,
3814 	};
3815 	struct nfs4_server_caps_res res = {};
3816 	struct rpc_message msg = {
3817 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3818 		.rpc_argp = &args,
3819 		.rpc_resp = &res,
3820 	};
3821 	int status;
3822 	int i;
3823 
3824 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3825 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3826 		     FATTR4_WORD0_LINK_SUPPORT |
3827 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3828 		     FATTR4_WORD0_ACLSUPPORT;
3829 	if (minorversion)
3830 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3831 
3832 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3833 	if (status == 0) {
3834 		/* Sanity check the server answers */
3835 		switch (minorversion) {
3836 		case 0:
3837 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3838 			res.attr_bitmask[2] = 0;
3839 			break;
3840 		case 1:
3841 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3842 			break;
3843 		case 2:
3844 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3845 		}
3846 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3847 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3848 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3849 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3850 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3851 				NFS_CAP_CTIME|NFS_CAP_MTIME|
3852 				NFS_CAP_SECURITY_LABEL);
3853 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3854 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3855 			server->caps |= NFS_CAP_ACLS;
3856 		if (res.has_links != 0)
3857 			server->caps |= NFS_CAP_HARDLINKS;
3858 		if (res.has_symlinks != 0)
3859 			server->caps |= NFS_CAP_SYMLINKS;
3860 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3861 			server->caps |= NFS_CAP_FILEID;
3862 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3863 			server->caps |= NFS_CAP_MODE;
3864 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3865 			server->caps |= NFS_CAP_NLINK;
3866 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3867 			server->caps |= NFS_CAP_OWNER;
3868 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3869 			server->caps |= NFS_CAP_OWNER_GROUP;
3870 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3871 			server->caps |= NFS_CAP_ATIME;
3872 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3873 			server->caps |= NFS_CAP_CTIME;
3874 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3875 			server->caps |= NFS_CAP_MTIME;
3876 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3877 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3878 			server->caps |= NFS_CAP_SECURITY_LABEL;
3879 #endif
3880 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3881 				sizeof(server->attr_bitmask));
3882 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3883 
3884 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3885 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3886 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3887 		server->cache_consistency_bitmask[2] = 0;
3888 
3889 		/* Avoid a regression due to buggy server */
3890 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3891 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3892 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3893 			sizeof(server->exclcreat_bitmask));
3894 
3895 		server->acl_bitmask = res.acl_bitmask;
3896 		server->fh_expire_type = res.fh_expire_type;
3897 	}
3898 
3899 	return status;
3900 }
3901 
3902 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3903 {
3904 	struct nfs4_exception exception = {
3905 		.interruptible = true,
3906 	};
3907 	int err;
3908 	do {
3909 		err = nfs4_handle_exception(server,
3910 				_nfs4_server_capabilities(server, fhandle),
3911 				&exception);
3912 	} while (exception.retry);
3913 	return err;
3914 }
3915 
3916 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3917 		struct nfs_fsinfo *info)
3918 {
3919 	u32 bitmask[3];
3920 	struct nfs4_lookup_root_arg args = {
3921 		.bitmask = bitmask,
3922 	};
3923 	struct nfs4_lookup_res res = {
3924 		.server = server,
3925 		.fattr = info->fattr,
3926 		.fh = fhandle,
3927 	};
3928 	struct rpc_message msg = {
3929 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3930 		.rpc_argp = &args,
3931 		.rpc_resp = &res,
3932 	};
3933 
3934 	bitmask[0] = nfs4_fattr_bitmap[0];
3935 	bitmask[1] = nfs4_fattr_bitmap[1];
3936 	/*
3937 	 * Process the label in the upcoming getfattr
3938 	 */
3939 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3940 
3941 	nfs_fattr_init(info->fattr);
3942 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3943 }
3944 
3945 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3946 		struct nfs_fsinfo *info)
3947 {
3948 	struct nfs4_exception exception = {
3949 		.interruptible = true,
3950 	};
3951 	int err;
3952 	do {
3953 		err = _nfs4_lookup_root(server, fhandle, info);
3954 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3955 		switch (err) {
3956 		case 0:
3957 		case -NFS4ERR_WRONGSEC:
3958 			goto out;
3959 		default:
3960 			err = nfs4_handle_exception(server, err, &exception);
3961 		}
3962 	} while (exception.retry);
3963 out:
3964 	return err;
3965 }
3966 
3967 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3968 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3969 {
3970 	struct rpc_auth_create_args auth_args = {
3971 		.pseudoflavor = flavor,
3972 	};
3973 	struct rpc_auth *auth;
3974 
3975 	auth = rpcauth_create(&auth_args, server->client);
3976 	if (IS_ERR(auth))
3977 		return -EACCES;
3978 	return nfs4_lookup_root(server, fhandle, info);
3979 }
3980 
3981 /*
3982  * Retry pseudoroot lookup with various security flavors.  We do this when:
3983  *
3984  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3985  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3986  *
3987  * Returns zero on success, or a negative NFS4ERR value, or a
3988  * negative errno value.
3989  */
3990 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3991 			      struct nfs_fsinfo *info)
3992 {
3993 	/* Per 3530bis 15.33.5 */
3994 	static const rpc_authflavor_t flav_array[] = {
3995 		RPC_AUTH_GSS_KRB5P,
3996 		RPC_AUTH_GSS_KRB5I,
3997 		RPC_AUTH_GSS_KRB5,
3998 		RPC_AUTH_UNIX,			/* courtesy */
3999 		RPC_AUTH_NULL,
4000 	};
4001 	int status = -EPERM;
4002 	size_t i;
4003 
4004 	if (server->auth_info.flavor_len > 0) {
4005 		/* try each flavor specified by user */
4006 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4007 			status = nfs4_lookup_root_sec(server, fhandle, info,
4008 						server->auth_info.flavors[i]);
4009 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4010 				continue;
4011 			break;
4012 		}
4013 	} else {
4014 		/* no flavors specified by user, try default list */
4015 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4016 			status = nfs4_lookup_root_sec(server, fhandle, info,
4017 						      flav_array[i]);
4018 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4019 				continue;
4020 			break;
4021 		}
4022 	}
4023 
4024 	/*
4025 	 * -EACCES could mean that the user doesn't have correct permissions
4026 	 * to access the mount.  It could also mean that we tried to mount
4027 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4028 	 * existing mount programs don't handle -EACCES very well so it should
4029 	 * be mapped to -EPERM instead.
4030 	 */
4031 	if (status == -EACCES)
4032 		status = -EPERM;
4033 	return status;
4034 }
4035 
4036 /**
4037  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4038  * @server: initialized nfs_server handle
4039  * @fhandle: we fill in the pseudo-fs root file handle
4040  * @info: we fill in an FSINFO struct
4041  * @auth_probe: probe the auth flavours
4042  *
4043  * Returns zero on success, or a negative errno.
4044  */
4045 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4046 			 struct nfs_fsinfo *info,
4047 			 bool auth_probe)
4048 {
4049 	int status = 0;
4050 
4051 	if (!auth_probe)
4052 		status = nfs4_lookup_root(server, fhandle, info);
4053 
4054 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4055 		status = server->nfs_client->cl_mvops->find_root_sec(server,
4056 				fhandle, info);
4057 
4058 	if (status == 0)
4059 		status = nfs4_server_capabilities(server, fhandle);
4060 	if (status == 0)
4061 		status = nfs4_do_fsinfo(server, fhandle, info);
4062 
4063 	return nfs4_map_errors(status);
4064 }
4065 
4066 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4067 			      struct nfs_fsinfo *info)
4068 {
4069 	int error;
4070 	struct nfs_fattr *fattr = info->fattr;
4071 	struct nfs4_label *label = fattr->label;
4072 
4073 	error = nfs4_server_capabilities(server, mntfh);
4074 	if (error < 0) {
4075 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4076 		return error;
4077 	}
4078 
4079 	error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4080 	if (error < 0) {
4081 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4082 		goto out;
4083 	}
4084 
4085 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4086 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4087 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4088 
4089 out:
4090 	return error;
4091 }
4092 
4093 /*
4094  * Get locations and (maybe) other attributes of a referral.
4095  * Note that we'll actually follow the referral later when
4096  * we detect fsid mismatch in inode revalidation
4097  */
4098 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4099 			     const struct qstr *name, struct nfs_fattr *fattr,
4100 			     struct nfs_fh *fhandle)
4101 {
4102 	int status = -ENOMEM;
4103 	struct page *page = NULL;
4104 	struct nfs4_fs_locations *locations = NULL;
4105 
4106 	page = alloc_page(GFP_KERNEL);
4107 	if (page == NULL)
4108 		goto out;
4109 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4110 	if (locations == NULL)
4111 		goto out;
4112 
4113 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4114 	if (status != 0)
4115 		goto out;
4116 
4117 	/*
4118 	 * If the fsid didn't change, this is a migration event, not a
4119 	 * referral.  Cause us to drop into the exception handler, which
4120 	 * will kick off migration recovery.
4121 	 */
4122 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4123 		dprintk("%s: server did not return a different fsid for"
4124 			" a referral at %s\n", __func__, name->name);
4125 		status = -NFS4ERR_MOVED;
4126 		goto out;
4127 	}
4128 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4129 	nfs_fixup_referral_attributes(&locations->fattr);
4130 
4131 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
4132 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4133 	memset(fhandle, 0, sizeof(struct nfs_fh));
4134 out:
4135 	if (page)
4136 		__free_page(page);
4137 	kfree(locations);
4138 	return status;
4139 }
4140 
4141 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4142 				struct nfs_fattr *fattr, struct nfs4_label *label,
4143 				struct inode *inode)
4144 {
4145 	__u32 bitmask[NFS4_BITMASK_SZ];
4146 	struct nfs4_getattr_arg args = {
4147 		.fh = fhandle,
4148 		.bitmask = bitmask,
4149 	};
4150 	struct nfs4_getattr_res res = {
4151 		.fattr = fattr,
4152 		.label = label,
4153 		.server = server,
4154 	};
4155 	struct rpc_message msg = {
4156 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4157 		.rpc_argp = &args,
4158 		.rpc_resp = &res,
4159 	};
4160 	unsigned short task_flags = 0;
4161 
4162 	/* Is this is an attribute revalidation, subject to softreval? */
4163 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4164 		task_flags |= RPC_TASK_TIMEOUT;
4165 
4166 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4167 
4168 	nfs_fattr_init(fattr);
4169 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4170 	return nfs4_do_call_sync(server->client, server, &msg,
4171 			&args.seq_args, &res.seq_res, task_flags);
4172 }
4173 
4174 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4175 				struct nfs_fattr *fattr, struct nfs4_label *label,
4176 				struct inode *inode)
4177 {
4178 	struct nfs4_exception exception = {
4179 		.interruptible = true,
4180 	};
4181 	int err;
4182 	do {
4183 		err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4184 		trace_nfs4_getattr(server, fhandle, fattr, err);
4185 		err = nfs4_handle_exception(server, err,
4186 				&exception);
4187 	} while (exception.retry);
4188 	return err;
4189 }
4190 
4191 /*
4192  * The file is not closed if it is opened due to the a request to change
4193  * the size of the file. The open call will not be needed once the
4194  * VFS layer lookup-intents are implemented.
4195  *
4196  * Close is called when the inode is destroyed.
4197  * If we haven't opened the file for O_WRONLY, we
4198  * need to in the size_change case to obtain a stateid.
4199  *
4200  * Got race?
4201  * Because OPEN is always done by name in nfsv4, it is
4202  * possible that we opened a different file by the same
4203  * name.  We can recognize this race condition, but we
4204  * can't do anything about it besides returning an error.
4205  *
4206  * This will be fixed with VFS changes (lookup-intent).
4207  */
4208 static int
4209 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4210 		  struct iattr *sattr)
4211 {
4212 	struct inode *inode = d_inode(dentry);
4213 	const struct cred *cred = NULL;
4214 	struct nfs_open_context *ctx = NULL;
4215 	struct nfs4_label *label = NULL;
4216 	int status;
4217 
4218 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4219 	    sattr->ia_valid & ATTR_SIZE &&
4220 	    sattr->ia_size < i_size_read(inode))
4221 		pnfs_commit_and_return_layout(inode);
4222 
4223 	nfs_fattr_init(fattr);
4224 
4225 	/* Deal with open(O_TRUNC) */
4226 	if (sattr->ia_valid & ATTR_OPEN)
4227 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4228 
4229 	/* Optimization: if the end result is no change, don't RPC */
4230 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4231 		return 0;
4232 
4233 	/* Search for an existing open(O_WRITE) file */
4234 	if (sattr->ia_valid & ATTR_FILE) {
4235 
4236 		ctx = nfs_file_open_context(sattr->ia_file);
4237 		if (ctx)
4238 			cred = ctx->cred;
4239 	}
4240 
4241 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4242 	if (IS_ERR(label))
4243 		return PTR_ERR(label);
4244 
4245 	/* Return any delegations if we're going to change ACLs */
4246 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4247 		nfs4_inode_make_writeable(inode);
4248 
4249 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4250 	if (status == 0) {
4251 		nfs_setattr_update_inode(inode, sattr, fattr);
4252 		nfs_setsecurity(inode, fattr, label);
4253 	}
4254 	nfs4_label_free(label);
4255 	return status;
4256 }
4257 
4258 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4259 		struct dentry *dentry, struct nfs_fh *fhandle,
4260 		struct nfs_fattr *fattr, struct nfs4_label *label)
4261 {
4262 	struct nfs_server *server = NFS_SERVER(dir);
4263 	int		       status;
4264 	struct nfs4_lookup_arg args = {
4265 		.bitmask = server->attr_bitmask,
4266 		.dir_fh = NFS_FH(dir),
4267 		.name = &dentry->d_name,
4268 	};
4269 	struct nfs4_lookup_res res = {
4270 		.server = server,
4271 		.fattr = fattr,
4272 		.label = label,
4273 		.fh = fhandle,
4274 	};
4275 	struct rpc_message msg = {
4276 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4277 		.rpc_argp = &args,
4278 		.rpc_resp = &res,
4279 	};
4280 	unsigned short task_flags = 0;
4281 
4282 	/* Is this is an attribute revalidation, subject to softreval? */
4283 	if (nfs_lookup_is_soft_revalidate(dentry))
4284 		task_flags |= RPC_TASK_TIMEOUT;
4285 
4286 	args.bitmask = nfs4_bitmask(server, label);
4287 
4288 	nfs_fattr_init(fattr);
4289 
4290 	dprintk("NFS call  lookup %pd2\n", dentry);
4291 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4292 	status = nfs4_do_call_sync(clnt, server, &msg,
4293 			&args.seq_args, &res.seq_res, task_flags);
4294 	dprintk("NFS reply lookup: %d\n", status);
4295 	return status;
4296 }
4297 
4298 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4299 {
4300 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4301 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4302 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4303 	fattr->nlink = 2;
4304 }
4305 
4306 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4307 				   struct dentry *dentry, struct nfs_fh *fhandle,
4308 				   struct nfs_fattr *fattr, struct nfs4_label *label)
4309 {
4310 	struct nfs4_exception exception = {
4311 		.interruptible = true,
4312 	};
4313 	struct rpc_clnt *client = *clnt;
4314 	const struct qstr *name = &dentry->d_name;
4315 	int err;
4316 	do {
4317 		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4318 		trace_nfs4_lookup(dir, name, err);
4319 		switch (err) {
4320 		case -NFS4ERR_BADNAME:
4321 			err = -ENOENT;
4322 			goto out;
4323 		case -NFS4ERR_MOVED:
4324 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4325 			if (err == -NFS4ERR_MOVED)
4326 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4327 			goto out;
4328 		case -NFS4ERR_WRONGSEC:
4329 			err = -EPERM;
4330 			if (client != *clnt)
4331 				goto out;
4332 			client = nfs4_negotiate_security(client, dir, name);
4333 			if (IS_ERR(client))
4334 				return PTR_ERR(client);
4335 
4336 			exception.retry = 1;
4337 			break;
4338 		default:
4339 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4340 		}
4341 	} while (exception.retry);
4342 
4343 out:
4344 	if (err == 0)
4345 		*clnt = client;
4346 	else if (client != *clnt)
4347 		rpc_shutdown_client(client);
4348 
4349 	return err;
4350 }
4351 
4352 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4353 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4354 			    struct nfs4_label *label)
4355 {
4356 	int status;
4357 	struct rpc_clnt *client = NFS_CLIENT(dir);
4358 
4359 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4360 	if (client != NFS_CLIENT(dir)) {
4361 		rpc_shutdown_client(client);
4362 		nfs_fixup_secinfo_attributes(fattr);
4363 	}
4364 	return status;
4365 }
4366 
4367 struct rpc_clnt *
4368 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4369 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4370 {
4371 	struct rpc_clnt *client = NFS_CLIENT(dir);
4372 	int status;
4373 
4374 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4375 	if (status < 0)
4376 		return ERR_PTR(status);
4377 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4378 }
4379 
4380 static int _nfs4_proc_lookupp(struct inode *inode,
4381 		struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4382 		struct nfs4_label *label)
4383 {
4384 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4385 	struct nfs_server *server = NFS_SERVER(inode);
4386 	int		       status;
4387 	struct nfs4_lookupp_arg args = {
4388 		.bitmask = server->attr_bitmask,
4389 		.fh = NFS_FH(inode),
4390 	};
4391 	struct nfs4_lookupp_res res = {
4392 		.server = server,
4393 		.fattr = fattr,
4394 		.label = label,
4395 		.fh = fhandle,
4396 	};
4397 	struct rpc_message msg = {
4398 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4399 		.rpc_argp = &args,
4400 		.rpc_resp = &res,
4401 	};
4402 	unsigned short task_flags = 0;
4403 
4404 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4405 		task_flags |= RPC_TASK_TIMEOUT;
4406 
4407 	args.bitmask = nfs4_bitmask(server, label);
4408 
4409 	nfs_fattr_init(fattr);
4410 
4411 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4412 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4413 				&res.seq_res, task_flags);
4414 	dprintk("NFS reply lookupp: %d\n", status);
4415 	return status;
4416 }
4417 
4418 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4419 			     struct nfs_fattr *fattr, struct nfs4_label *label)
4420 {
4421 	struct nfs4_exception exception = {
4422 		.interruptible = true,
4423 	};
4424 	int err;
4425 	do {
4426 		err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4427 		trace_nfs4_lookupp(inode, err);
4428 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4429 				&exception);
4430 	} while (exception.retry);
4431 	return err;
4432 }
4433 
4434 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4435 {
4436 	struct nfs_server *server = NFS_SERVER(inode);
4437 	struct nfs4_accessargs args = {
4438 		.fh = NFS_FH(inode),
4439 		.access = entry->mask,
4440 	};
4441 	struct nfs4_accessres res = {
4442 		.server = server,
4443 	};
4444 	struct rpc_message msg = {
4445 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4446 		.rpc_argp = &args,
4447 		.rpc_resp = &res,
4448 		.rpc_cred = entry->cred,
4449 	};
4450 	int status = 0;
4451 
4452 	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4453 		res.fattr = nfs_alloc_fattr();
4454 		if (res.fattr == NULL)
4455 			return -ENOMEM;
4456 		args.bitmask = server->cache_consistency_bitmask;
4457 	}
4458 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4459 	if (!status) {
4460 		nfs_access_set_mask(entry, res.access);
4461 		if (res.fattr)
4462 			nfs_refresh_inode(inode, res.fattr);
4463 	}
4464 	nfs_free_fattr(res.fattr);
4465 	return status;
4466 }
4467 
4468 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4469 {
4470 	struct nfs4_exception exception = {
4471 		.interruptible = true,
4472 	};
4473 	int err;
4474 	do {
4475 		err = _nfs4_proc_access(inode, entry);
4476 		trace_nfs4_access(inode, err);
4477 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4478 				&exception);
4479 	} while (exception.retry);
4480 	return err;
4481 }
4482 
4483 /*
4484  * TODO: For the time being, we don't try to get any attributes
4485  * along with any of the zero-copy operations READ, READDIR,
4486  * READLINK, WRITE.
4487  *
4488  * In the case of the first three, we want to put the GETATTR
4489  * after the read-type operation -- this is because it is hard
4490  * to predict the length of a GETATTR response in v4, and thus
4491  * align the READ data correctly.  This means that the GETATTR
4492  * may end up partially falling into the page cache, and we should
4493  * shift it into the 'tail' of the xdr_buf before processing.
4494  * To do this efficiently, we need to know the total length
4495  * of data received, which doesn't seem to be available outside
4496  * of the RPC layer.
4497  *
4498  * In the case of WRITE, we also want to put the GETATTR after
4499  * the operation -- in this case because we want to make sure
4500  * we get the post-operation mtime and size.
4501  *
4502  * Both of these changes to the XDR layer would in fact be quite
4503  * minor, but I decided to leave them for a subsequent patch.
4504  */
4505 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4506 		unsigned int pgbase, unsigned int pglen)
4507 {
4508 	struct nfs4_readlink args = {
4509 		.fh       = NFS_FH(inode),
4510 		.pgbase	  = pgbase,
4511 		.pglen    = pglen,
4512 		.pages    = &page,
4513 	};
4514 	struct nfs4_readlink_res res;
4515 	struct rpc_message msg = {
4516 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4517 		.rpc_argp = &args,
4518 		.rpc_resp = &res,
4519 	};
4520 
4521 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4522 }
4523 
4524 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4525 		unsigned int pgbase, unsigned int pglen)
4526 {
4527 	struct nfs4_exception exception = {
4528 		.interruptible = true,
4529 	};
4530 	int err;
4531 	do {
4532 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4533 		trace_nfs4_readlink(inode, err);
4534 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4535 				&exception);
4536 	} while (exception.retry);
4537 	return err;
4538 }
4539 
4540 /*
4541  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4542  */
4543 static int
4544 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4545 		 int flags)
4546 {
4547 	struct nfs_server *server = NFS_SERVER(dir);
4548 	struct nfs4_label l, *ilabel = NULL;
4549 	struct nfs_open_context *ctx;
4550 	struct nfs4_state *state;
4551 	int status = 0;
4552 
4553 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4554 	if (IS_ERR(ctx))
4555 		return PTR_ERR(ctx);
4556 
4557 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4558 
4559 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4560 		sattr->ia_mode &= ~current_umask();
4561 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4562 	if (IS_ERR(state)) {
4563 		status = PTR_ERR(state);
4564 		goto out;
4565 	}
4566 out:
4567 	nfs4_label_release_security(ilabel);
4568 	put_nfs_open_context(ctx);
4569 	return status;
4570 }
4571 
4572 static int
4573 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4574 {
4575 	struct nfs_server *server = NFS_SERVER(dir);
4576 	struct nfs_removeargs args = {
4577 		.fh = NFS_FH(dir),
4578 		.name = *name,
4579 	};
4580 	struct nfs_removeres res = {
4581 		.server = server,
4582 	};
4583 	struct rpc_message msg = {
4584 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4585 		.rpc_argp = &args,
4586 		.rpc_resp = &res,
4587 	};
4588 	unsigned long timestamp = jiffies;
4589 	int status;
4590 
4591 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4592 	if (status == 0) {
4593 		spin_lock(&dir->i_lock);
4594 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4595 					      NFS_INO_INVALID_DATA);
4596 		/* Removing a directory decrements nlink in the parent */
4597 		if (ftype == NF4DIR && dir->i_nlink > 2)
4598 			nfs4_dec_nlink_locked(dir);
4599 		spin_unlock(&dir->i_lock);
4600 	}
4601 	return status;
4602 }
4603 
4604 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4605 {
4606 	struct nfs4_exception exception = {
4607 		.interruptible = true,
4608 	};
4609 	struct inode *inode = d_inode(dentry);
4610 	int err;
4611 
4612 	if (inode) {
4613 		if (inode->i_nlink == 1)
4614 			nfs4_inode_return_delegation(inode);
4615 		else
4616 			nfs4_inode_make_writeable(inode);
4617 	}
4618 	do {
4619 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4620 		trace_nfs4_remove(dir, &dentry->d_name, err);
4621 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4622 				&exception);
4623 	} while (exception.retry);
4624 	return err;
4625 }
4626 
4627 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4628 {
4629 	struct nfs4_exception exception = {
4630 		.interruptible = true,
4631 	};
4632 	int err;
4633 
4634 	do {
4635 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4636 		trace_nfs4_remove(dir, name, err);
4637 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4638 				&exception);
4639 	} while (exception.retry);
4640 	return err;
4641 }
4642 
4643 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4644 		struct dentry *dentry,
4645 		struct inode *inode)
4646 {
4647 	struct nfs_removeargs *args = msg->rpc_argp;
4648 	struct nfs_removeres *res = msg->rpc_resp;
4649 
4650 	res->server = NFS_SB(dentry->d_sb);
4651 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4652 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4653 
4654 	nfs_fattr_init(res->dir_attr);
4655 
4656 	if (inode)
4657 		nfs4_inode_return_delegation(inode);
4658 }
4659 
4660 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4661 {
4662 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4663 			&data->args.seq_args,
4664 			&data->res.seq_res,
4665 			task);
4666 }
4667 
4668 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4669 {
4670 	struct nfs_unlinkdata *data = task->tk_calldata;
4671 	struct nfs_removeres *res = &data->res;
4672 
4673 	if (!nfs4_sequence_done(task, &res->seq_res))
4674 		return 0;
4675 	if (nfs4_async_handle_error(task, res->server, NULL,
4676 				    &data->timeout) == -EAGAIN)
4677 		return 0;
4678 	if (task->tk_status == 0)
4679 		nfs4_update_changeattr(dir, &res->cinfo,
4680 				res->dir_attr->time_start,
4681 				NFS_INO_INVALID_DATA);
4682 	return 1;
4683 }
4684 
4685 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4686 		struct dentry *old_dentry,
4687 		struct dentry *new_dentry)
4688 {
4689 	struct nfs_renameargs *arg = msg->rpc_argp;
4690 	struct nfs_renameres *res = msg->rpc_resp;
4691 	struct inode *old_inode = d_inode(old_dentry);
4692 	struct inode *new_inode = d_inode(new_dentry);
4693 
4694 	if (old_inode)
4695 		nfs4_inode_make_writeable(old_inode);
4696 	if (new_inode)
4697 		nfs4_inode_return_delegation(new_inode);
4698 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4699 	res->server = NFS_SB(old_dentry->d_sb);
4700 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4701 }
4702 
4703 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4704 {
4705 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4706 			&data->args.seq_args,
4707 			&data->res.seq_res,
4708 			task);
4709 }
4710 
4711 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4712 				 struct inode *new_dir)
4713 {
4714 	struct nfs_renamedata *data = task->tk_calldata;
4715 	struct nfs_renameres *res = &data->res;
4716 
4717 	if (!nfs4_sequence_done(task, &res->seq_res))
4718 		return 0;
4719 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4720 		return 0;
4721 
4722 	if (task->tk_status == 0) {
4723 		if (new_dir != old_dir) {
4724 			/* Note: If we moved a directory, nlink will change */
4725 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4726 					res->old_fattr->time_start,
4727 					NFS_INO_INVALID_OTHER |
4728 					    NFS_INO_INVALID_DATA);
4729 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4730 					res->new_fattr->time_start,
4731 					NFS_INO_INVALID_OTHER |
4732 					    NFS_INO_INVALID_DATA);
4733 		} else
4734 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4735 					res->old_fattr->time_start,
4736 					NFS_INO_INVALID_DATA);
4737 	}
4738 	return 1;
4739 }
4740 
4741 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4742 {
4743 	struct nfs_server *server = NFS_SERVER(inode);
4744 	__u32 bitmask[NFS4_BITMASK_SZ];
4745 	struct nfs4_link_arg arg = {
4746 		.fh     = NFS_FH(inode),
4747 		.dir_fh = NFS_FH(dir),
4748 		.name   = name,
4749 		.bitmask = bitmask,
4750 	};
4751 	struct nfs4_link_res res = {
4752 		.server = server,
4753 		.label = NULL,
4754 	};
4755 	struct rpc_message msg = {
4756 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4757 		.rpc_argp = &arg,
4758 		.rpc_resp = &res,
4759 	};
4760 	int status = -ENOMEM;
4761 
4762 	res.fattr = nfs_alloc_fattr();
4763 	if (res.fattr == NULL)
4764 		goto out;
4765 
4766 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
4767 	if (IS_ERR(res.label)) {
4768 		status = PTR_ERR(res.label);
4769 		goto out;
4770 	}
4771 
4772 	nfs4_inode_make_writeable(inode);
4773 	nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4774 
4775 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4776 	if (!status) {
4777 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4778 				       NFS_INO_INVALID_DATA);
4779 		status = nfs_post_op_update_inode(inode, res.fattr);
4780 		if (!status)
4781 			nfs_setsecurity(inode, res.fattr, res.label);
4782 	}
4783 
4784 
4785 	nfs4_label_free(res.label);
4786 
4787 out:
4788 	nfs_free_fattr(res.fattr);
4789 	return status;
4790 }
4791 
4792 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4793 {
4794 	struct nfs4_exception exception = {
4795 		.interruptible = true,
4796 	};
4797 	int err;
4798 	do {
4799 		err = nfs4_handle_exception(NFS_SERVER(inode),
4800 				_nfs4_proc_link(inode, dir, name),
4801 				&exception);
4802 	} while (exception.retry);
4803 	return err;
4804 }
4805 
4806 struct nfs4_createdata {
4807 	struct rpc_message msg;
4808 	struct nfs4_create_arg arg;
4809 	struct nfs4_create_res res;
4810 	struct nfs_fh fh;
4811 	struct nfs_fattr fattr;
4812 	struct nfs4_label *label;
4813 };
4814 
4815 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4816 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4817 {
4818 	struct nfs4_createdata *data;
4819 
4820 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4821 	if (data != NULL) {
4822 		struct nfs_server *server = NFS_SERVER(dir);
4823 
4824 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
4825 		if (IS_ERR(data->label))
4826 			goto out_free;
4827 
4828 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4829 		data->msg.rpc_argp = &data->arg;
4830 		data->msg.rpc_resp = &data->res;
4831 		data->arg.dir_fh = NFS_FH(dir);
4832 		data->arg.server = server;
4833 		data->arg.name = name;
4834 		data->arg.attrs = sattr;
4835 		data->arg.ftype = ftype;
4836 		data->arg.bitmask = nfs4_bitmask(server, data->label);
4837 		data->arg.umask = current_umask();
4838 		data->res.server = server;
4839 		data->res.fh = &data->fh;
4840 		data->res.fattr = &data->fattr;
4841 		data->res.label = data->label;
4842 		nfs_fattr_init(data->res.fattr);
4843 	}
4844 	return data;
4845 out_free:
4846 	kfree(data);
4847 	return NULL;
4848 }
4849 
4850 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4851 {
4852 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4853 				    &data->arg.seq_args, &data->res.seq_res, 1);
4854 	if (status == 0) {
4855 		spin_lock(&dir->i_lock);
4856 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4857 				data->res.fattr->time_start,
4858 				NFS_INO_INVALID_DATA);
4859 		/* Creating a directory bumps nlink in the parent */
4860 		if (data->arg.ftype == NF4DIR)
4861 			nfs4_inc_nlink_locked(dir);
4862 		spin_unlock(&dir->i_lock);
4863 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4864 	}
4865 	return status;
4866 }
4867 
4868 static void nfs4_free_createdata(struct nfs4_createdata *data)
4869 {
4870 	nfs4_label_free(data->label);
4871 	kfree(data);
4872 }
4873 
4874 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4875 		struct page *page, unsigned int len, struct iattr *sattr,
4876 		struct nfs4_label *label)
4877 {
4878 	struct nfs4_createdata *data;
4879 	int status = -ENAMETOOLONG;
4880 
4881 	if (len > NFS4_MAXPATHLEN)
4882 		goto out;
4883 
4884 	status = -ENOMEM;
4885 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4886 	if (data == NULL)
4887 		goto out;
4888 
4889 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4890 	data->arg.u.symlink.pages = &page;
4891 	data->arg.u.symlink.len = len;
4892 	data->arg.label = label;
4893 
4894 	status = nfs4_do_create(dir, dentry, data);
4895 
4896 	nfs4_free_createdata(data);
4897 out:
4898 	return status;
4899 }
4900 
4901 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4902 		struct page *page, unsigned int len, struct iattr *sattr)
4903 {
4904 	struct nfs4_exception exception = {
4905 		.interruptible = true,
4906 	};
4907 	struct nfs4_label l, *label = NULL;
4908 	int err;
4909 
4910 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4911 
4912 	do {
4913 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4914 		trace_nfs4_symlink(dir, &dentry->d_name, err);
4915 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4916 				&exception);
4917 	} while (exception.retry);
4918 
4919 	nfs4_label_release_security(label);
4920 	return err;
4921 }
4922 
4923 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4924 		struct iattr *sattr, struct nfs4_label *label)
4925 {
4926 	struct nfs4_createdata *data;
4927 	int status = -ENOMEM;
4928 
4929 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4930 	if (data == NULL)
4931 		goto out;
4932 
4933 	data->arg.label = label;
4934 	status = nfs4_do_create(dir, dentry, data);
4935 
4936 	nfs4_free_createdata(data);
4937 out:
4938 	return status;
4939 }
4940 
4941 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4942 		struct iattr *sattr)
4943 {
4944 	struct nfs_server *server = NFS_SERVER(dir);
4945 	struct nfs4_exception exception = {
4946 		.interruptible = true,
4947 	};
4948 	struct nfs4_label l, *label = NULL;
4949 	int err;
4950 
4951 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4952 
4953 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4954 		sattr->ia_mode &= ~current_umask();
4955 	do {
4956 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4957 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
4958 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4959 				&exception);
4960 	} while (exception.retry);
4961 	nfs4_label_release_security(label);
4962 
4963 	return err;
4964 }
4965 
4966 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
4967 			      struct nfs_readdir_res *nr_res)
4968 {
4969 	struct inode		*dir = d_inode(nr_arg->dentry);
4970 	struct nfs_server	*server = NFS_SERVER(dir);
4971 	struct nfs4_readdir_arg args = {
4972 		.fh = NFS_FH(dir),
4973 		.pages = nr_arg->pages,
4974 		.pgbase = 0,
4975 		.count = nr_arg->page_len,
4976 		.plus = nr_arg->plus,
4977 	};
4978 	struct nfs4_readdir_res res;
4979 	struct rpc_message msg = {
4980 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4981 		.rpc_argp = &args,
4982 		.rpc_resp = &res,
4983 		.rpc_cred = nr_arg->cred,
4984 	};
4985 	int			status;
4986 
4987 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
4988 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
4989 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
4990 		args.bitmask = server->attr_bitmask_nl;
4991 	else
4992 		args.bitmask = server->attr_bitmask;
4993 
4994 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
4995 	res.pgbase = args.pgbase;
4996 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
4997 			&res.seq_res, 0);
4998 	if (status >= 0) {
4999 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5000 		status += args.pgbase;
5001 	}
5002 
5003 	nfs_invalidate_atime(dir);
5004 
5005 	dprintk("%s: returns %d\n", __func__, status);
5006 	return status;
5007 }
5008 
5009 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5010 			     struct nfs_readdir_res *res)
5011 {
5012 	struct nfs4_exception exception = {
5013 		.interruptible = true,
5014 	};
5015 	int err;
5016 	do {
5017 		err = _nfs4_proc_readdir(arg, res);
5018 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5019 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5020 					    err, &exception);
5021 	} while (exception.retry);
5022 	return err;
5023 }
5024 
5025 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5026 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5027 {
5028 	struct nfs4_createdata *data;
5029 	int mode = sattr->ia_mode;
5030 	int status = -ENOMEM;
5031 
5032 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5033 	if (data == NULL)
5034 		goto out;
5035 
5036 	if (S_ISFIFO(mode))
5037 		data->arg.ftype = NF4FIFO;
5038 	else if (S_ISBLK(mode)) {
5039 		data->arg.ftype = NF4BLK;
5040 		data->arg.u.device.specdata1 = MAJOR(rdev);
5041 		data->arg.u.device.specdata2 = MINOR(rdev);
5042 	}
5043 	else if (S_ISCHR(mode)) {
5044 		data->arg.ftype = NF4CHR;
5045 		data->arg.u.device.specdata1 = MAJOR(rdev);
5046 		data->arg.u.device.specdata2 = MINOR(rdev);
5047 	} else if (!S_ISSOCK(mode)) {
5048 		status = -EINVAL;
5049 		goto out_free;
5050 	}
5051 
5052 	data->arg.label = label;
5053 	status = nfs4_do_create(dir, dentry, data);
5054 out_free:
5055 	nfs4_free_createdata(data);
5056 out:
5057 	return status;
5058 }
5059 
5060 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5061 		struct iattr *sattr, dev_t rdev)
5062 {
5063 	struct nfs_server *server = NFS_SERVER(dir);
5064 	struct nfs4_exception exception = {
5065 		.interruptible = true,
5066 	};
5067 	struct nfs4_label l, *label = NULL;
5068 	int err;
5069 
5070 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5071 
5072 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5073 		sattr->ia_mode &= ~current_umask();
5074 	do {
5075 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5076 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5077 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5078 				&exception);
5079 	} while (exception.retry);
5080 
5081 	nfs4_label_release_security(label);
5082 
5083 	return err;
5084 }
5085 
5086 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5087 		 struct nfs_fsstat *fsstat)
5088 {
5089 	struct nfs4_statfs_arg args = {
5090 		.fh = fhandle,
5091 		.bitmask = server->attr_bitmask,
5092 	};
5093 	struct nfs4_statfs_res res = {
5094 		.fsstat = fsstat,
5095 	};
5096 	struct rpc_message msg = {
5097 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5098 		.rpc_argp = &args,
5099 		.rpc_resp = &res,
5100 	};
5101 
5102 	nfs_fattr_init(fsstat->fattr);
5103 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5104 }
5105 
5106 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5107 {
5108 	struct nfs4_exception exception = {
5109 		.interruptible = true,
5110 	};
5111 	int err;
5112 	do {
5113 		err = nfs4_handle_exception(server,
5114 				_nfs4_proc_statfs(server, fhandle, fsstat),
5115 				&exception);
5116 	} while (exception.retry);
5117 	return err;
5118 }
5119 
5120 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5121 		struct nfs_fsinfo *fsinfo)
5122 {
5123 	struct nfs4_fsinfo_arg args = {
5124 		.fh = fhandle,
5125 		.bitmask = server->attr_bitmask,
5126 	};
5127 	struct nfs4_fsinfo_res res = {
5128 		.fsinfo = fsinfo,
5129 	};
5130 	struct rpc_message msg = {
5131 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5132 		.rpc_argp = &args,
5133 		.rpc_resp = &res,
5134 	};
5135 
5136 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5137 }
5138 
5139 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5140 {
5141 	struct nfs4_exception exception = {
5142 		.interruptible = true,
5143 	};
5144 	int err;
5145 
5146 	do {
5147 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5148 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5149 		if (err == 0) {
5150 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5151 			break;
5152 		}
5153 		err = nfs4_handle_exception(server, err, &exception);
5154 	} while (exception.retry);
5155 	return err;
5156 }
5157 
5158 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5159 {
5160 	int error;
5161 
5162 	nfs_fattr_init(fsinfo->fattr);
5163 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5164 	if (error == 0) {
5165 		/* block layout checks this! */
5166 		server->pnfs_blksize = fsinfo->blksize;
5167 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5168 	}
5169 
5170 	return error;
5171 }
5172 
5173 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5174 		struct nfs_pathconf *pathconf)
5175 {
5176 	struct nfs4_pathconf_arg args = {
5177 		.fh = fhandle,
5178 		.bitmask = server->attr_bitmask,
5179 	};
5180 	struct nfs4_pathconf_res res = {
5181 		.pathconf = pathconf,
5182 	};
5183 	struct rpc_message msg = {
5184 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5185 		.rpc_argp = &args,
5186 		.rpc_resp = &res,
5187 	};
5188 
5189 	/* None of the pathconf attributes are mandatory to implement */
5190 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5191 		memset(pathconf, 0, sizeof(*pathconf));
5192 		return 0;
5193 	}
5194 
5195 	nfs_fattr_init(pathconf->fattr);
5196 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5197 }
5198 
5199 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5200 		struct nfs_pathconf *pathconf)
5201 {
5202 	struct nfs4_exception exception = {
5203 		.interruptible = true,
5204 	};
5205 	int err;
5206 
5207 	do {
5208 		err = nfs4_handle_exception(server,
5209 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5210 				&exception);
5211 	} while (exception.retry);
5212 	return err;
5213 }
5214 
5215 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5216 		const struct nfs_open_context *ctx,
5217 		const struct nfs_lock_context *l_ctx,
5218 		fmode_t fmode)
5219 {
5220 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5221 }
5222 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5223 
5224 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5225 		const struct nfs_open_context *ctx,
5226 		const struct nfs_lock_context *l_ctx,
5227 		fmode_t fmode)
5228 {
5229 	nfs4_stateid _current_stateid;
5230 
5231 	/* If the current stateid represents a lost lock, then exit */
5232 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5233 		return true;
5234 	return nfs4_stateid_match(stateid, &_current_stateid);
5235 }
5236 
5237 static bool nfs4_error_stateid_expired(int err)
5238 {
5239 	switch (err) {
5240 	case -NFS4ERR_DELEG_REVOKED:
5241 	case -NFS4ERR_ADMIN_REVOKED:
5242 	case -NFS4ERR_BAD_STATEID:
5243 	case -NFS4ERR_STALE_STATEID:
5244 	case -NFS4ERR_OLD_STATEID:
5245 	case -NFS4ERR_OPENMODE:
5246 	case -NFS4ERR_EXPIRED:
5247 		return true;
5248 	}
5249 	return false;
5250 }
5251 
5252 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5253 {
5254 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5255 
5256 	trace_nfs4_read(hdr, task->tk_status);
5257 	if (task->tk_status < 0) {
5258 		struct nfs4_exception exception = {
5259 			.inode = hdr->inode,
5260 			.state = hdr->args.context->state,
5261 			.stateid = &hdr->args.stateid,
5262 		};
5263 		task->tk_status = nfs4_async_handle_exception(task,
5264 				server, task->tk_status, &exception);
5265 		if (exception.retry) {
5266 			rpc_restart_call_prepare(task);
5267 			return -EAGAIN;
5268 		}
5269 	}
5270 
5271 	if (task->tk_status > 0)
5272 		renew_lease(server, hdr->timestamp);
5273 	return 0;
5274 }
5275 
5276 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5277 		struct nfs_pgio_args *args)
5278 {
5279 
5280 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5281 		nfs4_stateid_is_current(&args->stateid,
5282 				args->context,
5283 				args->lock_context,
5284 				FMODE_READ))
5285 		return false;
5286 	rpc_restart_call_prepare(task);
5287 	return true;
5288 }
5289 
5290 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5291 					 struct nfs_pgio_header *hdr)
5292 {
5293 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5294 	struct rpc_message *msg = &task->tk_msg;
5295 
5296 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5297 	    server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5298 		server->caps &= ~NFS_CAP_READ_PLUS;
5299 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5300 		rpc_restart_call_prepare(task);
5301 		return true;
5302 	}
5303 	return false;
5304 }
5305 
5306 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5307 {
5308 	dprintk("--> %s\n", __func__);
5309 
5310 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5311 		return -EAGAIN;
5312 	if (nfs4_read_stateid_changed(task, &hdr->args))
5313 		return -EAGAIN;
5314 	if (nfs4_read_plus_not_supported(task, hdr))
5315 		return -EAGAIN;
5316 	if (task->tk_status > 0)
5317 		nfs_invalidate_atime(hdr->inode);
5318 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5319 				    nfs4_read_done_cb(task, hdr);
5320 }
5321 
5322 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5323 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5324 				    struct rpc_message *msg)
5325 {
5326 	/* Note: We don't use READ_PLUS with pNFS yet */
5327 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5328 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5329 }
5330 #else
5331 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5332 				    struct rpc_message *msg)
5333 {
5334 }
5335 #endif /* CONFIG_NFS_V4_2 */
5336 
5337 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5338 				 struct rpc_message *msg)
5339 {
5340 	hdr->timestamp   = jiffies;
5341 	if (!hdr->pgio_done_cb)
5342 		hdr->pgio_done_cb = nfs4_read_done_cb;
5343 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5344 	nfs42_read_plus_support(hdr, msg);
5345 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5346 }
5347 
5348 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5349 				      struct nfs_pgio_header *hdr)
5350 {
5351 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5352 			&hdr->args.seq_args,
5353 			&hdr->res.seq_res,
5354 			task))
5355 		return 0;
5356 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5357 				hdr->args.lock_context,
5358 				hdr->rw_mode) == -EIO)
5359 		return -EIO;
5360 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5361 		return -EIO;
5362 	return 0;
5363 }
5364 
5365 static int nfs4_write_done_cb(struct rpc_task *task,
5366 			      struct nfs_pgio_header *hdr)
5367 {
5368 	struct inode *inode = hdr->inode;
5369 
5370 	trace_nfs4_write(hdr, task->tk_status);
5371 	if (task->tk_status < 0) {
5372 		struct nfs4_exception exception = {
5373 			.inode = hdr->inode,
5374 			.state = hdr->args.context->state,
5375 			.stateid = &hdr->args.stateid,
5376 		};
5377 		task->tk_status = nfs4_async_handle_exception(task,
5378 				NFS_SERVER(inode), task->tk_status,
5379 				&exception);
5380 		if (exception.retry) {
5381 			rpc_restart_call_prepare(task);
5382 			return -EAGAIN;
5383 		}
5384 	}
5385 	if (task->tk_status >= 0) {
5386 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5387 		nfs_writeback_update_inode(hdr);
5388 	}
5389 	return 0;
5390 }
5391 
5392 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5393 		struct nfs_pgio_args *args)
5394 {
5395 
5396 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5397 		nfs4_stateid_is_current(&args->stateid,
5398 				args->context,
5399 				args->lock_context,
5400 				FMODE_WRITE))
5401 		return false;
5402 	rpc_restart_call_prepare(task);
5403 	return true;
5404 }
5405 
5406 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5407 {
5408 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5409 		return -EAGAIN;
5410 	if (nfs4_write_stateid_changed(task, &hdr->args))
5411 		return -EAGAIN;
5412 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5413 		nfs4_write_done_cb(task, hdr);
5414 }
5415 
5416 static
5417 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5418 {
5419 	/* Don't request attributes for pNFS or O_DIRECT writes */
5420 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5421 		return false;
5422 	/* Otherwise, request attributes if and only if we don't hold
5423 	 * a delegation
5424 	 */
5425 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5426 }
5427 
5428 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
5429 				struct nfs_server *server,
5430 				struct nfs4_label *label)
5431 {
5432 
5433 	unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5434 
5435 	if ((cache_validity & NFS_INO_INVALID_DATA) ||
5436 		(cache_validity & NFS_INO_REVAL_PAGECACHE) ||
5437 		(cache_validity & NFS_INO_REVAL_FORCED) ||
5438 		(cache_validity & NFS_INO_INVALID_OTHER))
5439 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
5440 
5441 	if (cache_validity & NFS_INO_INVALID_ATIME)
5442 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5443 	if (cache_validity & NFS_INO_INVALID_ACCESS)
5444 		bitmask[0] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER |
5445 				FATTR4_WORD1_OWNER_GROUP;
5446 	if (cache_validity & NFS_INO_INVALID_ACL)
5447 		bitmask[0] |= FATTR4_WORD0_ACL;
5448 	if (cache_validity & NFS_INO_INVALID_LABEL)
5449 		bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5450 	if (cache_validity & NFS_INO_INVALID_CTIME)
5451 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5452 	if (cache_validity & NFS_INO_INVALID_MTIME)
5453 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5454 	if (cache_validity & NFS_INO_INVALID_SIZE)
5455 		bitmask[0] |= FATTR4_WORD0_SIZE;
5456 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5457 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5458 }
5459 
5460 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5461 				  struct rpc_message *msg,
5462 				  struct rpc_clnt **clnt)
5463 {
5464 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5465 
5466 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5467 		hdr->args.bitmask = NULL;
5468 		hdr->res.fattr = NULL;
5469 	} else {
5470 		hdr->args.bitmask = server->cache_consistency_bitmask;
5471 		nfs4_bitmask_adjust(hdr->args.bitmask, hdr->inode, server, NULL);
5472 	}
5473 
5474 	if (!hdr->pgio_done_cb)
5475 		hdr->pgio_done_cb = nfs4_write_done_cb;
5476 	hdr->res.server = server;
5477 	hdr->timestamp   = jiffies;
5478 
5479 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5480 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5481 	nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5482 }
5483 
5484 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5485 {
5486 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5487 			&data->args.seq_args,
5488 			&data->res.seq_res,
5489 			task);
5490 }
5491 
5492 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5493 {
5494 	struct inode *inode = data->inode;
5495 
5496 	trace_nfs4_commit(data, task->tk_status);
5497 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5498 				    NULL, NULL) == -EAGAIN) {
5499 		rpc_restart_call_prepare(task);
5500 		return -EAGAIN;
5501 	}
5502 	return 0;
5503 }
5504 
5505 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5506 {
5507 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5508 		return -EAGAIN;
5509 	return data->commit_done_cb(task, data);
5510 }
5511 
5512 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5513 				   struct rpc_clnt **clnt)
5514 {
5515 	struct nfs_server *server = NFS_SERVER(data->inode);
5516 
5517 	if (data->commit_done_cb == NULL)
5518 		data->commit_done_cb = nfs4_commit_done_cb;
5519 	data->res.server = server;
5520 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5521 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5522 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5523 }
5524 
5525 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5526 				struct nfs_commitres *res)
5527 {
5528 	struct inode *dst_inode = file_inode(dst);
5529 	struct nfs_server *server = NFS_SERVER(dst_inode);
5530 	struct rpc_message msg = {
5531 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5532 		.rpc_argp = args,
5533 		.rpc_resp = res,
5534 	};
5535 
5536 	args->fh = NFS_FH(dst_inode);
5537 	return nfs4_call_sync(server->client, server, &msg,
5538 			&args->seq_args, &res->seq_res, 1);
5539 }
5540 
5541 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5542 {
5543 	struct nfs_commitargs args = {
5544 		.offset = offset,
5545 		.count = count,
5546 	};
5547 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5548 	struct nfs4_exception exception = { };
5549 	int status;
5550 
5551 	do {
5552 		status = _nfs4_proc_commit(dst, &args, res);
5553 		status = nfs4_handle_exception(dst_server, status, &exception);
5554 	} while (exception.retry);
5555 
5556 	return status;
5557 }
5558 
5559 struct nfs4_renewdata {
5560 	struct nfs_client	*client;
5561 	unsigned long		timestamp;
5562 };
5563 
5564 /*
5565  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5566  * standalone procedure for queueing an asynchronous RENEW.
5567  */
5568 static void nfs4_renew_release(void *calldata)
5569 {
5570 	struct nfs4_renewdata *data = calldata;
5571 	struct nfs_client *clp = data->client;
5572 
5573 	if (refcount_read(&clp->cl_count) > 1)
5574 		nfs4_schedule_state_renewal(clp);
5575 	nfs_put_client(clp);
5576 	kfree(data);
5577 }
5578 
5579 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5580 {
5581 	struct nfs4_renewdata *data = calldata;
5582 	struct nfs_client *clp = data->client;
5583 	unsigned long timestamp = data->timestamp;
5584 
5585 	trace_nfs4_renew_async(clp, task->tk_status);
5586 	switch (task->tk_status) {
5587 	case 0:
5588 		break;
5589 	case -NFS4ERR_LEASE_MOVED:
5590 		nfs4_schedule_lease_moved_recovery(clp);
5591 		break;
5592 	default:
5593 		/* Unless we're shutting down, schedule state recovery! */
5594 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5595 			return;
5596 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5597 			nfs4_schedule_lease_recovery(clp);
5598 			return;
5599 		}
5600 		nfs4_schedule_path_down_recovery(clp);
5601 	}
5602 	do_renew_lease(clp, timestamp);
5603 }
5604 
5605 static const struct rpc_call_ops nfs4_renew_ops = {
5606 	.rpc_call_done = nfs4_renew_done,
5607 	.rpc_release = nfs4_renew_release,
5608 };
5609 
5610 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5611 {
5612 	struct rpc_message msg = {
5613 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5614 		.rpc_argp	= clp,
5615 		.rpc_cred	= cred,
5616 	};
5617 	struct nfs4_renewdata *data;
5618 
5619 	if (renew_flags == 0)
5620 		return 0;
5621 	if (!refcount_inc_not_zero(&clp->cl_count))
5622 		return -EIO;
5623 	data = kmalloc(sizeof(*data), GFP_NOFS);
5624 	if (data == NULL) {
5625 		nfs_put_client(clp);
5626 		return -ENOMEM;
5627 	}
5628 	data->client = clp;
5629 	data->timestamp = jiffies;
5630 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5631 			&nfs4_renew_ops, data);
5632 }
5633 
5634 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5635 {
5636 	struct rpc_message msg = {
5637 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5638 		.rpc_argp	= clp,
5639 		.rpc_cred	= cred,
5640 	};
5641 	unsigned long now = jiffies;
5642 	int status;
5643 
5644 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5645 	if (status < 0)
5646 		return status;
5647 	do_renew_lease(clp, now);
5648 	return 0;
5649 }
5650 
5651 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5652 {
5653 	return server->caps & NFS_CAP_ACLS;
5654 }
5655 
5656 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5657  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5658  * the stack.
5659  */
5660 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5661 
5662 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5663 		struct page **pages)
5664 {
5665 	struct page *newpage, **spages;
5666 	int rc = 0;
5667 	size_t len;
5668 	spages = pages;
5669 
5670 	do {
5671 		len = min_t(size_t, PAGE_SIZE, buflen);
5672 		newpage = alloc_page(GFP_KERNEL);
5673 
5674 		if (newpage == NULL)
5675 			goto unwind;
5676 		memcpy(page_address(newpage), buf, len);
5677 		buf += len;
5678 		buflen -= len;
5679 		*pages++ = newpage;
5680 		rc++;
5681 	} while (buflen != 0);
5682 
5683 	return rc;
5684 
5685 unwind:
5686 	for(; rc > 0; rc--)
5687 		__free_page(spages[rc-1]);
5688 	return -ENOMEM;
5689 }
5690 
5691 struct nfs4_cached_acl {
5692 	int cached;
5693 	size_t len;
5694 	char data[];
5695 };
5696 
5697 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5698 {
5699 	struct nfs_inode *nfsi = NFS_I(inode);
5700 
5701 	spin_lock(&inode->i_lock);
5702 	kfree(nfsi->nfs4_acl);
5703 	nfsi->nfs4_acl = acl;
5704 	spin_unlock(&inode->i_lock);
5705 }
5706 
5707 static void nfs4_zap_acl_attr(struct inode *inode)
5708 {
5709 	nfs4_set_cached_acl(inode, NULL);
5710 }
5711 
5712 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5713 {
5714 	struct nfs_inode *nfsi = NFS_I(inode);
5715 	struct nfs4_cached_acl *acl;
5716 	int ret = -ENOENT;
5717 
5718 	spin_lock(&inode->i_lock);
5719 	acl = nfsi->nfs4_acl;
5720 	if (acl == NULL)
5721 		goto out;
5722 	if (buf == NULL) /* user is just asking for length */
5723 		goto out_len;
5724 	if (acl->cached == 0)
5725 		goto out;
5726 	ret = -ERANGE; /* see getxattr(2) man page */
5727 	if (acl->len > buflen)
5728 		goto out;
5729 	memcpy(buf, acl->data, acl->len);
5730 out_len:
5731 	ret = acl->len;
5732 out:
5733 	spin_unlock(&inode->i_lock);
5734 	return ret;
5735 }
5736 
5737 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5738 {
5739 	struct nfs4_cached_acl *acl;
5740 	size_t buflen = sizeof(*acl) + acl_len;
5741 
5742 	if (buflen <= PAGE_SIZE) {
5743 		acl = kmalloc(buflen, GFP_KERNEL);
5744 		if (acl == NULL)
5745 			goto out;
5746 		acl->cached = 1;
5747 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5748 	} else {
5749 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5750 		if (acl == NULL)
5751 			goto out;
5752 		acl->cached = 0;
5753 	}
5754 	acl->len = acl_len;
5755 out:
5756 	nfs4_set_cached_acl(inode, acl);
5757 }
5758 
5759 /*
5760  * The getxattr API returns the required buffer length when called with a
5761  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5762  * the required buf.  On a NULL buf, we send a page of data to the server
5763  * guessing that the ACL request can be serviced by a page. If so, we cache
5764  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5765  * the cache. If not so, we throw away the page, and cache the required
5766  * length. The next getxattr call will then produce another round trip to
5767  * the server, this time with the input buf of the required size.
5768  */
5769 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5770 {
5771 	struct page **pages;
5772 	struct nfs_getaclargs args = {
5773 		.fh = NFS_FH(inode),
5774 		.acl_len = buflen,
5775 	};
5776 	struct nfs_getaclres res = {
5777 		.acl_len = buflen,
5778 	};
5779 	struct rpc_message msg = {
5780 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5781 		.rpc_argp = &args,
5782 		.rpc_resp = &res,
5783 	};
5784 	unsigned int npages;
5785 	int ret = -ENOMEM, i;
5786 	struct nfs_server *server = NFS_SERVER(inode);
5787 
5788 	if (buflen == 0)
5789 		buflen = server->rsize;
5790 
5791 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5792 	pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5793 	if (!pages)
5794 		return -ENOMEM;
5795 
5796 	args.acl_pages = pages;
5797 
5798 	for (i = 0; i < npages; i++) {
5799 		pages[i] = alloc_page(GFP_KERNEL);
5800 		if (!pages[i])
5801 			goto out_free;
5802 	}
5803 
5804 	/* for decoding across pages */
5805 	res.acl_scratch = alloc_page(GFP_KERNEL);
5806 	if (!res.acl_scratch)
5807 		goto out_free;
5808 
5809 	args.acl_len = npages * PAGE_SIZE;
5810 
5811 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5812 		__func__, buf, buflen, npages, args.acl_len);
5813 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5814 			     &msg, &args.seq_args, &res.seq_res, 0);
5815 	if (ret)
5816 		goto out_free;
5817 
5818 	/* Handle the case where the passed-in buffer is too short */
5819 	if (res.acl_flags & NFS4_ACL_TRUNC) {
5820 		/* Did the user only issue a request for the acl length? */
5821 		if (buf == NULL)
5822 			goto out_ok;
5823 		ret = -ERANGE;
5824 		goto out_free;
5825 	}
5826 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5827 	if (buf) {
5828 		if (res.acl_len > buflen) {
5829 			ret = -ERANGE;
5830 			goto out_free;
5831 		}
5832 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5833 	}
5834 out_ok:
5835 	ret = res.acl_len;
5836 out_free:
5837 	for (i = 0; i < npages; i++)
5838 		if (pages[i])
5839 			__free_page(pages[i]);
5840 	if (res.acl_scratch)
5841 		__free_page(res.acl_scratch);
5842 	kfree(pages);
5843 	return ret;
5844 }
5845 
5846 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5847 {
5848 	struct nfs4_exception exception = {
5849 		.interruptible = true,
5850 	};
5851 	ssize_t ret;
5852 	do {
5853 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5854 		trace_nfs4_get_acl(inode, ret);
5855 		if (ret >= 0)
5856 			break;
5857 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5858 	} while (exception.retry);
5859 	return ret;
5860 }
5861 
5862 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5863 {
5864 	struct nfs_server *server = NFS_SERVER(inode);
5865 	int ret;
5866 
5867 	if (!nfs4_server_supports_acls(server))
5868 		return -EOPNOTSUPP;
5869 	ret = nfs_revalidate_inode(server, inode);
5870 	if (ret < 0)
5871 		return ret;
5872 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5873 		nfs_zap_acl_cache(inode);
5874 	ret = nfs4_read_cached_acl(inode, buf, buflen);
5875 	if (ret != -ENOENT)
5876 		/* -ENOENT is returned if there is no ACL or if there is an ACL
5877 		 * but no cached acl data, just the acl length */
5878 		return ret;
5879 	return nfs4_get_acl_uncached(inode, buf, buflen);
5880 }
5881 
5882 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5883 {
5884 	struct nfs_server *server = NFS_SERVER(inode);
5885 	struct page *pages[NFS4ACL_MAXPAGES];
5886 	struct nfs_setaclargs arg = {
5887 		.fh		= NFS_FH(inode),
5888 		.acl_pages	= pages,
5889 		.acl_len	= buflen,
5890 	};
5891 	struct nfs_setaclres res;
5892 	struct rpc_message msg = {
5893 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5894 		.rpc_argp	= &arg,
5895 		.rpc_resp	= &res,
5896 	};
5897 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5898 	int ret, i;
5899 
5900 	if (!nfs4_server_supports_acls(server))
5901 		return -EOPNOTSUPP;
5902 	if (npages > ARRAY_SIZE(pages))
5903 		return -ERANGE;
5904 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5905 	if (i < 0)
5906 		return i;
5907 	nfs4_inode_make_writeable(inode);
5908 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5909 
5910 	/*
5911 	 * Free each page after tx, so the only ref left is
5912 	 * held by the network stack
5913 	 */
5914 	for (; i > 0; i--)
5915 		put_page(pages[i-1]);
5916 
5917 	/*
5918 	 * Acl update can result in inode attribute update.
5919 	 * so mark the attribute cache invalid.
5920 	 */
5921 	spin_lock(&inode->i_lock);
5922 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5923 		| NFS_INO_INVALID_CTIME
5924 		| NFS_INO_REVAL_FORCED;
5925 	spin_unlock(&inode->i_lock);
5926 	nfs_access_zap_cache(inode);
5927 	nfs_zap_acl_cache(inode);
5928 	return ret;
5929 }
5930 
5931 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5932 {
5933 	struct nfs4_exception exception = { };
5934 	int err;
5935 	do {
5936 		err = __nfs4_proc_set_acl(inode, buf, buflen);
5937 		trace_nfs4_set_acl(inode, err);
5938 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5939 				&exception);
5940 	} while (exception.retry);
5941 	return err;
5942 }
5943 
5944 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5945 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5946 					size_t buflen)
5947 {
5948 	struct nfs_server *server = NFS_SERVER(inode);
5949 	struct nfs_fattr fattr;
5950 	struct nfs4_label label = {0, 0, buflen, buf};
5951 
5952 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5953 	struct nfs4_getattr_arg arg = {
5954 		.fh		= NFS_FH(inode),
5955 		.bitmask	= bitmask,
5956 	};
5957 	struct nfs4_getattr_res res = {
5958 		.fattr		= &fattr,
5959 		.label		= &label,
5960 		.server		= server,
5961 	};
5962 	struct rpc_message msg = {
5963 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5964 		.rpc_argp	= &arg,
5965 		.rpc_resp	= &res,
5966 	};
5967 	int ret;
5968 
5969 	nfs_fattr_init(&fattr);
5970 
5971 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5972 	if (ret)
5973 		return ret;
5974 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5975 		return -ENOENT;
5976 	return 0;
5977 }
5978 
5979 static int nfs4_get_security_label(struct inode *inode, void *buf,
5980 					size_t buflen)
5981 {
5982 	struct nfs4_exception exception = {
5983 		.interruptible = true,
5984 	};
5985 	int err;
5986 
5987 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5988 		return -EOPNOTSUPP;
5989 
5990 	do {
5991 		err = _nfs4_get_security_label(inode, buf, buflen);
5992 		trace_nfs4_get_security_label(inode, err);
5993 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5994 				&exception);
5995 	} while (exception.retry);
5996 	return err;
5997 }
5998 
5999 static int _nfs4_do_set_security_label(struct inode *inode,
6000 		struct nfs4_label *ilabel,
6001 		struct nfs_fattr *fattr,
6002 		struct nfs4_label *olabel)
6003 {
6004 
6005 	struct iattr sattr = {0};
6006 	struct nfs_server *server = NFS_SERVER(inode);
6007 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6008 	struct nfs_setattrargs arg = {
6009 		.fh		= NFS_FH(inode),
6010 		.iap		= &sattr,
6011 		.server		= server,
6012 		.bitmask	= bitmask,
6013 		.label		= ilabel,
6014 	};
6015 	struct nfs_setattrres res = {
6016 		.fattr		= fattr,
6017 		.label		= olabel,
6018 		.server		= server,
6019 	};
6020 	struct rpc_message msg = {
6021 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6022 		.rpc_argp	= &arg,
6023 		.rpc_resp	= &res,
6024 	};
6025 	int status;
6026 
6027 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6028 
6029 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6030 	if (status)
6031 		dprintk("%s failed: %d\n", __func__, status);
6032 
6033 	return status;
6034 }
6035 
6036 static int nfs4_do_set_security_label(struct inode *inode,
6037 		struct nfs4_label *ilabel,
6038 		struct nfs_fattr *fattr,
6039 		struct nfs4_label *olabel)
6040 {
6041 	struct nfs4_exception exception = { };
6042 	int err;
6043 
6044 	do {
6045 		err = _nfs4_do_set_security_label(inode, ilabel,
6046 				fattr, olabel);
6047 		trace_nfs4_set_security_label(inode, err);
6048 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6049 				&exception);
6050 	} while (exception.retry);
6051 	return err;
6052 }
6053 
6054 static int
6055 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6056 {
6057 	struct nfs4_label ilabel, *olabel = NULL;
6058 	struct nfs_fattr fattr;
6059 	int status;
6060 
6061 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6062 		return -EOPNOTSUPP;
6063 
6064 	nfs_fattr_init(&fattr);
6065 
6066 	ilabel.pi = 0;
6067 	ilabel.lfs = 0;
6068 	ilabel.label = (char *)buf;
6069 	ilabel.len = buflen;
6070 
6071 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6072 	if (IS_ERR(olabel)) {
6073 		status = -PTR_ERR(olabel);
6074 		goto out;
6075 	}
6076 
6077 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6078 	if (status == 0)
6079 		nfs_setsecurity(inode, &fattr, olabel);
6080 
6081 	nfs4_label_free(olabel);
6082 out:
6083 	return status;
6084 }
6085 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6086 
6087 
6088 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6089 				    nfs4_verifier *bootverf)
6090 {
6091 	__be32 verf[2];
6092 
6093 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6094 		/* An impossible timestamp guarantees this value
6095 		 * will never match a generated boot time. */
6096 		verf[0] = cpu_to_be32(U32_MAX);
6097 		verf[1] = cpu_to_be32(U32_MAX);
6098 	} else {
6099 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6100 		u64 ns = ktime_to_ns(nn->boot_time);
6101 
6102 		verf[0] = cpu_to_be32(ns >> 32);
6103 		verf[1] = cpu_to_be32(ns);
6104 	}
6105 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6106 }
6107 
6108 static size_t
6109 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6110 {
6111 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6112 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6113 	const char *id;
6114 
6115 	buf[0] = '\0';
6116 
6117 	if (nn_clp) {
6118 		rcu_read_lock();
6119 		id = rcu_dereference(nn_clp->identifier);
6120 		if (id)
6121 			strscpy(buf, id, buflen);
6122 		rcu_read_unlock();
6123 	}
6124 
6125 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6126 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6127 
6128 	return strlen(buf);
6129 }
6130 
6131 static int
6132 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6133 {
6134 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6135 	size_t buflen;
6136 	size_t len;
6137 	char *str;
6138 
6139 	if (clp->cl_owner_id != NULL)
6140 		return 0;
6141 
6142 	rcu_read_lock();
6143 	len = 14 +
6144 		strlen(clp->cl_rpcclient->cl_nodename) +
6145 		1 +
6146 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6147 		1;
6148 	rcu_read_unlock();
6149 
6150 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6151 	if (buflen)
6152 		len += buflen + 1;
6153 
6154 	if (len > NFS4_OPAQUE_LIMIT + 1)
6155 		return -EINVAL;
6156 
6157 	/*
6158 	 * Since this string is allocated at mount time, and held until the
6159 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6160 	 * about a memory-reclaim deadlock.
6161 	 */
6162 	str = kmalloc(len, GFP_KERNEL);
6163 	if (!str)
6164 		return -ENOMEM;
6165 
6166 	rcu_read_lock();
6167 	if (buflen)
6168 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6169 			  clp->cl_rpcclient->cl_nodename, buf,
6170 			  rpc_peeraddr2str(clp->cl_rpcclient,
6171 					   RPC_DISPLAY_ADDR));
6172 	else
6173 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6174 			  clp->cl_rpcclient->cl_nodename,
6175 			  rpc_peeraddr2str(clp->cl_rpcclient,
6176 					   RPC_DISPLAY_ADDR));
6177 	rcu_read_unlock();
6178 
6179 	clp->cl_owner_id = str;
6180 	return 0;
6181 }
6182 
6183 static int
6184 nfs4_init_uniform_client_string(struct nfs_client *clp)
6185 {
6186 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6187 	size_t buflen;
6188 	size_t len;
6189 	char *str;
6190 
6191 	if (clp->cl_owner_id != NULL)
6192 		return 0;
6193 
6194 	len = 10 + 10 + 1 + 10 + 1 +
6195 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6196 
6197 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6198 	if (buflen)
6199 		len += buflen + 1;
6200 
6201 	if (len > NFS4_OPAQUE_LIMIT + 1)
6202 		return -EINVAL;
6203 
6204 	/*
6205 	 * Since this string is allocated at mount time, and held until the
6206 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6207 	 * about a memory-reclaim deadlock.
6208 	 */
6209 	str = kmalloc(len, GFP_KERNEL);
6210 	if (!str)
6211 		return -ENOMEM;
6212 
6213 	if (buflen)
6214 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6215 			  clp->rpc_ops->version, clp->cl_minorversion,
6216 			  buf, clp->cl_rpcclient->cl_nodename);
6217 	else
6218 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6219 			  clp->rpc_ops->version, clp->cl_minorversion,
6220 			  clp->cl_rpcclient->cl_nodename);
6221 	clp->cl_owner_id = str;
6222 	return 0;
6223 }
6224 
6225 /*
6226  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6227  * services.  Advertise one based on the address family of the
6228  * clientaddr.
6229  */
6230 static unsigned int
6231 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6232 {
6233 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6234 		return scnprintf(buf, len, "tcp6");
6235 	else
6236 		return scnprintf(buf, len, "tcp");
6237 }
6238 
6239 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6240 {
6241 	struct nfs4_setclientid *sc = calldata;
6242 
6243 	if (task->tk_status == 0)
6244 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6245 }
6246 
6247 static const struct rpc_call_ops nfs4_setclientid_ops = {
6248 	.rpc_call_done = nfs4_setclientid_done,
6249 };
6250 
6251 /**
6252  * nfs4_proc_setclientid - Negotiate client ID
6253  * @clp: state data structure
6254  * @program: RPC program for NFSv4 callback service
6255  * @port: IP port number for NFS4 callback service
6256  * @cred: credential to use for this call
6257  * @res: where to place the result
6258  *
6259  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6260  */
6261 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6262 		unsigned short port, const struct cred *cred,
6263 		struct nfs4_setclientid_res *res)
6264 {
6265 	nfs4_verifier sc_verifier;
6266 	struct nfs4_setclientid setclientid = {
6267 		.sc_verifier = &sc_verifier,
6268 		.sc_prog = program,
6269 		.sc_clnt = clp,
6270 	};
6271 	struct rpc_message msg = {
6272 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6273 		.rpc_argp = &setclientid,
6274 		.rpc_resp = res,
6275 		.rpc_cred = cred,
6276 	};
6277 	struct rpc_task_setup task_setup_data = {
6278 		.rpc_client = clp->cl_rpcclient,
6279 		.rpc_message = &msg,
6280 		.callback_ops = &nfs4_setclientid_ops,
6281 		.callback_data = &setclientid,
6282 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6283 	};
6284 	unsigned long now = jiffies;
6285 	int status;
6286 
6287 	/* nfs_client_id4 */
6288 	nfs4_init_boot_verifier(clp, &sc_verifier);
6289 
6290 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6291 		status = nfs4_init_uniform_client_string(clp);
6292 	else
6293 		status = nfs4_init_nonuniform_client_string(clp);
6294 
6295 	if (status)
6296 		goto out;
6297 
6298 	/* cb_client4 */
6299 	setclientid.sc_netid_len =
6300 				nfs4_init_callback_netid(clp,
6301 						setclientid.sc_netid,
6302 						sizeof(setclientid.sc_netid));
6303 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6304 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6305 				clp->cl_ipaddr, port >> 8, port & 255);
6306 
6307 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6308 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6309 		clp->cl_owner_id);
6310 
6311 	status = nfs4_call_sync_custom(&task_setup_data);
6312 	if (setclientid.sc_cred) {
6313 		kfree(clp->cl_acceptor);
6314 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6315 		put_rpccred(setclientid.sc_cred);
6316 	}
6317 
6318 	if (status == 0)
6319 		do_renew_lease(clp, now);
6320 out:
6321 	trace_nfs4_setclientid(clp, status);
6322 	dprintk("NFS reply setclientid: %d\n", status);
6323 	return status;
6324 }
6325 
6326 /**
6327  * nfs4_proc_setclientid_confirm - Confirm client ID
6328  * @clp: state data structure
6329  * @arg: result of a previous SETCLIENTID
6330  * @cred: credential to use for this call
6331  *
6332  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6333  */
6334 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6335 		struct nfs4_setclientid_res *arg,
6336 		const struct cred *cred)
6337 {
6338 	struct rpc_message msg = {
6339 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6340 		.rpc_argp = arg,
6341 		.rpc_cred = cred,
6342 	};
6343 	int status;
6344 
6345 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6346 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6347 		clp->cl_clientid);
6348 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6349 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6350 	trace_nfs4_setclientid_confirm(clp, status);
6351 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6352 	return status;
6353 }
6354 
6355 struct nfs4_delegreturndata {
6356 	struct nfs4_delegreturnargs args;
6357 	struct nfs4_delegreturnres res;
6358 	struct nfs_fh fh;
6359 	nfs4_stateid stateid;
6360 	unsigned long timestamp;
6361 	struct {
6362 		struct nfs4_layoutreturn_args arg;
6363 		struct nfs4_layoutreturn_res res;
6364 		struct nfs4_xdr_opaque_data ld_private;
6365 		u32 roc_barrier;
6366 		bool roc;
6367 	} lr;
6368 	struct nfs_fattr fattr;
6369 	int rpc_status;
6370 	struct inode *inode;
6371 };
6372 
6373 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6374 {
6375 	struct nfs4_delegreturndata *data = calldata;
6376 	struct nfs4_exception exception = {
6377 		.inode = data->inode,
6378 		.stateid = &data->stateid,
6379 	};
6380 
6381 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6382 		return;
6383 
6384 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6385 
6386 	/* Handle Layoutreturn errors */
6387 	if (pnfs_roc_done(task, data->inode,
6388 				&data->args.lr_args,
6389 				&data->res.lr_res,
6390 				&data->res.lr_ret) == -EAGAIN)
6391 		goto out_restart;
6392 
6393 	switch (task->tk_status) {
6394 	case 0:
6395 		renew_lease(data->res.server, data->timestamp);
6396 		break;
6397 	case -NFS4ERR_ADMIN_REVOKED:
6398 	case -NFS4ERR_DELEG_REVOKED:
6399 	case -NFS4ERR_EXPIRED:
6400 		nfs4_free_revoked_stateid(data->res.server,
6401 				data->args.stateid,
6402 				task->tk_msg.rpc_cred);
6403 		fallthrough;
6404 	case -NFS4ERR_BAD_STATEID:
6405 	case -NFS4ERR_STALE_STATEID:
6406 	case -ETIMEDOUT:
6407 		task->tk_status = 0;
6408 		break;
6409 	case -NFS4ERR_OLD_STATEID:
6410 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6411 			nfs4_stateid_seqid_inc(&data->stateid);
6412 		if (data->args.bitmask) {
6413 			data->args.bitmask = NULL;
6414 			data->res.fattr = NULL;
6415 		}
6416 		goto out_restart;
6417 	case -NFS4ERR_ACCESS:
6418 		if (data->args.bitmask) {
6419 			data->args.bitmask = NULL;
6420 			data->res.fattr = NULL;
6421 			goto out_restart;
6422 		}
6423 		fallthrough;
6424 	default:
6425 		task->tk_status = nfs4_async_handle_exception(task,
6426 				data->res.server, task->tk_status,
6427 				&exception);
6428 		if (exception.retry)
6429 			goto out_restart;
6430 	}
6431 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6432 	data->rpc_status = task->tk_status;
6433 	return;
6434 out_restart:
6435 	task->tk_status = 0;
6436 	rpc_restart_call_prepare(task);
6437 }
6438 
6439 static void nfs4_delegreturn_release(void *calldata)
6440 {
6441 	struct nfs4_delegreturndata *data = calldata;
6442 	struct inode *inode = data->inode;
6443 
6444 	if (inode) {
6445 		if (data->lr.roc)
6446 			pnfs_roc_release(&data->lr.arg, &data->lr.res,
6447 					data->res.lr_ret);
6448 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6449 		nfs_iput_and_deactive(inode);
6450 	}
6451 	kfree(calldata);
6452 }
6453 
6454 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6455 {
6456 	struct nfs4_delegreturndata *d_data;
6457 	struct pnfs_layout_hdr *lo;
6458 
6459 	d_data = (struct nfs4_delegreturndata *)data;
6460 
6461 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6462 		nfs4_sequence_done(task, &d_data->res.seq_res);
6463 		return;
6464 	}
6465 
6466 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6467 	if (lo && !pnfs_layout_is_valid(lo)) {
6468 		d_data->args.lr_args = NULL;
6469 		d_data->res.lr_res = NULL;
6470 	}
6471 
6472 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6473 			&d_data->args.seq_args,
6474 			&d_data->res.seq_res,
6475 			task);
6476 }
6477 
6478 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6479 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6480 	.rpc_call_done = nfs4_delegreturn_done,
6481 	.rpc_release = nfs4_delegreturn_release,
6482 };
6483 
6484 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6485 {
6486 	struct nfs4_delegreturndata *data;
6487 	struct nfs_server *server = NFS_SERVER(inode);
6488 	struct rpc_task *task;
6489 	struct rpc_message msg = {
6490 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6491 		.rpc_cred = cred,
6492 	};
6493 	struct rpc_task_setup task_setup_data = {
6494 		.rpc_client = server->client,
6495 		.rpc_message = &msg,
6496 		.callback_ops = &nfs4_delegreturn_ops,
6497 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6498 	};
6499 	int status = 0;
6500 
6501 	data = kzalloc(sizeof(*data), GFP_NOFS);
6502 	if (data == NULL)
6503 		return -ENOMEM;
6504 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6505 
6506 	nfs4_state_protect(server->nfs_client,
6507 			NFS_SP4_MACH_CRED_CLEANUP,
6508 			&task_setup_data.rpc_client, &msg);
6509 
6510 	data->args.fhandle = &data->fh;
6511 	data->args.stateid = &data->stateid;
6512 	data->args.bitmask = server->cache_consistency_bitmask;
6513 	nfs4_bitmask_adjust(data->args.bitmask, inode, server, NULL);
6514 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6515 	nfs4_stateid_copy(&data->stateid, stateid);
6516 	data->res.fattr = &data->fattr;
6517 	data->res.server = server;
6518 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6519 	data->lr.arg.ld_private = &data->lr.ld_private;
6520 	nfs_fattr_init(data->res.fattr);
6521 	data->timestamp = jiffies;
6522 	data->rpc_status = 0;
6523 	data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6524 	data->inode = nfs_igrab_and_active(inode);
6525 	if (data->inode) {
6526 		if (data->lr.roc) {
6527 			data->args.lr_args = &data->lr.arg;
6528 			data->res.lr_res = &data->lr.res;
6529 		}
6530 	} else if (data->lr.roc) {
6531 		pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6532 		data->lr.roc = false;
6533 	}
6534 
6535 	task_setup_data.callback_data = data;
6536 	msg.rpc_argp = &data->args;
6537 	msg.rpc_resp = &data->res;
6538 	task = rpc_run_task(&task_setup_data);
6539 	if (IS_ERR(task))
6540 		return PTR_ERR(task);
6541 	if (!issync)
6542 		goto out;
6543 	status = rpc_wait_for_completion_task(task);
6544 	if (status != 0)
6545 		goto out;
6546 	status = data->rpc_status;
6547 out:
6548 	rpc_put_task(task);
6549 	return status;
6550 }
6551 
6552 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6553 {
6554 	struct nfs_server *server = NFS_SERVER(inode);
6555 	struct nfs4_exception exception = { };
6556 	int err;
6557 	do {
6558 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6559 		trace_nfs4_delegreturn(inode, stateid, err);
6560 		switch (err) {
6561 			case -NFS4ERR_STALE_STATEID:
6562 			case -NFS4ERR_EXPIRED:
6563 			case 0:
6564 				return 0;
6565 		}
6566 		err = nfs4_handle_exception(server, err, &exception);
6567 	} while (exception.retry);
6568 	return err;
6569 }
6570 
6571 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6572 {
6573 	struct inode *inode = state->inode;
6574 	struct nfs_server *server = NFS_SERVER(inode);
6575 	struct nfs_client *clp = server->nfs_client;
6576 	struct nfs_lockt_args arg = {
6577 		.fh = NFS_FH(inode),
6578 		.fl = request,
6579 	};
6580 	struct nfs_lockt_res res = {
6581 		.denied = request,
6582 	};
6583 	struct rpc_message msg = {
6584 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6585 		.rpc_argp	= &arg,
6586 		.rpc_resp	= &res,
6587 		.rpc_cred	= state->owner->so_cred,
6588 	};
6589 	struct nfs4_lock_state *lsp;
6590 	int status;
6591 
6592 	arg.lock_owner.clientid = clp->cl_clientid;
6593 	status = nfs4_set_lock_state(state, request);
6594 	if (status != 0)
6595 		goto out;
6596 	lsp = request->fl_u.nfs4_fl.owner;
6597 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6598 	arg.lock_owner.s_dev = server->s_dev;
6599 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6600 	switch (status) {
6601 		case 0:
6602 			request->fl_type = F_UNLCK;
6603 			break;
6604 		case -NFS4ERR_DENIED:
6605 			status = 0;
6606 	}
6607 	request->fl_ops->fl_release_private(request);
6608 	request->fl_ops = NULL;
6609 out:
6610 	return status;
6611 }
6612 
6613 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6614 {
6615 	struct nfs4_exception exception = {
6616 		.interruptible = true,
6617 	};
6618 	int err;
6619 
6620 	do {
6621 		err = _nfs4_proc_getlk(state, cmd, request);
6622 		trace_nfs4_get_lock(request, state, cmd, err);
6623 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6624 				&exception);
6625 	} while (exception.retry);
6626 	return err;
6627 }
6628 
6629 /*
6630  * Update the seqid of a lock stateid after receiving
6631  * NFS4ERR_OLD_STATEID
6632  */
6633 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6634 		struct nfs4_lock_state *lsp)
6635 {
6636 	struct nfs4_state *state = lsp->ls_state;
6637 	bool ret = false;
6638 
6639 	spin_lock(&state->state_lock);
6640 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6641 		goto out;
6642 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6643 		nfs4_stateid_seqid_inc(dst);
6644 	else
6645 		dst->seqid = lsp->ls_stateid.seqid;
6646 	ret = true;
6647 out:
6648 	spin_unlock(&state->state_lock);
6649 	return ret;
6650 }
6651 
6652 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6653 		struct nfs4_lock_state *lsp)
6654 {
6655 	struct nfs4_state *state = lsp->ls_state;
6656 	bool ret;
6657 
6658 	spin_lock(&state->state_lock);
6659 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6660 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6661 	spin_unlock(&state->state_lock);
6662 	return ret;
6663 }
6664 
6665 struct nfs4_unlockdata {
6666 	struct nfs_locku_args arg;
6667 	struct nfs_locku_res res;
6668 	struct nfs4_lock_state *lsp;
6669 	struct nfs_open_context *ctx;
6670 	struct nfs_lock_context *l_ctx;
6671 	struct file_lock fl;
6672 	struct nfs_server *server;
6673 	unsigned long timestamp;
6674 };
6675 
6676 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6677 		struct nfs_open_context *ctx,
6678 		struct nfs4_lock_state *lsp,
6679 		struct nfs_seqid *seqid)
6680 {
6681 	struct nfs4_unlockdata *p;
6682 	struct nfs4_state *state = lsp->ls_state;
6683 	struct inode *inode = state->inode;
6684 
6685 	p = kzalloc(sizeof(*p), GFP_NOFS);
6686 	if (p == NULL)
6687 		return NULL;
6688 	p->arg.fh = NFS_FH(inode);
6689 	p->arg.fl = &p->fl;
6690 	p->arg.seqid = seqid;
6691 	p->res.seqid = seqid;
6692 	p->lsp = lsp;
6693 	/* Ensure we don't close file until we're done freeing locks! */
6694 	p->ctx = get_nfs_open_context(ctx);
6695 	p->l_ctx = nfs_get_lock_context(ctx);
6696 	locks_init_lock(&p->fl);
6697 	locks_copy_lock(&p->fl, fl);
6698 	p->server = NFS_SERVER(inode);
6699 	spin_lock(&state->state_lock);
6700 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6701 	spin_unlock(&state->state_lock);
6702 	return p;
6703 }
6704 
6705 static void nfs4_locku_release_calldata(void *data)
6706 {
6707 	struct nfs4_unlockdata *calldata = data;
6708 	nfs_free_seqid(calldata->arg.seqid);
6709 	nfs4_put_lock_state(calldata->lsp);
6710 	nfs_put_lock_context(calldata->l_ctx);
6711 	put_nfs_open_context(calldata->ctx);
6712 	kfree(calldata);
6713 }
6714 
6715 static void nfs4_locku_done(struct rpc_task *task, void *data)
6716 {
6717 	struct nfs4_unlockdata *calldata = data;
6718 	struct nfs4_exception exception = {
6719 		.inode = calldata->lsp->ls_state->inode,
6720 		.stateid = &calldata->arg.stateid,
6721 	};
6722 
6723 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6724 		return;
6725 	switch (task->tk_status) {
6726 		case 0:
6727 			renew_lease(calldata->server, calldata->timestamp);
6728 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6729 			if (nfs4_update_lock_stateid(calldata->lsp,
6730 					&calldata->res.stateid))
6731 				break;
6732 			fallthrough;
6733 		case -NFS4ERR_ADMIN_REVOKED:
6734 		case -NFS4ERR_EXPIRED:
6735 			nfs4_free_revoked_stateid(calldata->server,
6736 					&calldata->arg.stateid,
6737 					task->tk_msg.rpc_cred);
6738 			fallthrough;
6739 		case -NFS4ERR_BAD_STATEID:
6740 		case -NFS4ERR_STALE_STATEID:
6741 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6742 						calldata->lsp))
6743 				rpc_restart_call_prepare(task);
6744 			break;
6745 		case -NFS4ERR_OLD_STATEID:
6746 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6747 						calldata->lsp))
6748 				rpc_restart_call_prepare(task);
6749 			break;
6750 		default:
6751 			task->tk_status = nfs4_async_handle_exception(task,
6752 					calldata->server, task->tk_status,
6753 					&exception);
6754 			if (exception.retry)
6755 				rpc_restart_call_prepare(task);
6756 	}
6757 	nfs_release_seqid(calldata->arg.seqid);
6758 }
6759 
6760 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6761 {
6762 	struct nfs4_unlockdata *calldata = data;
6763 
6764 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6765 		nfs_async_iocounter_wait(task, calldata->l_ctx))
6766 		return;
6767 
6768 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6769 		goto out_wait;
6770 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6771 		/* Note: exit _without_ running nfs4_locku_done */
6772 		goto out_no_action;
6773 	}
6774 	calldata->timestamp = jiffies;
6775 	if (nfs4_setup_sequence(calldata->server->nfs_client,
6776 				&calldata->arg.seq_args,
6777 				&calldata->res.seq_res,
6778 				task) != 0)
6779 		nfs_release_seqid(calldata->arg.seqid);
6780 	return;
6781 out_no_action:
6782 	task->tk_action = NULL;
6783 out_wait:
6784 	nfs4_sequence_done(task, &calldata->res.seq_res);
6785 }
6786 
6787 static const struct rpc_call_ops nfs4_locku_ops = {
6788 	.rpc_call_prepare = nfs4_locku_prepare,
6789 	.rpc_call_done = nfs4_locku_done,
6790 	.rpc_release = nfs4_locku_release_calldata,
6791 };
6792 
6793 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6794 		struct nfs_open_context *ctx,
6795 		struct nfs4_lock_state *lsp,
6796 		struct nfs_seqid *seqid)
6797 {
6798 	struct nfs4_unlockdata *data;
6799 	struct rpc_message msg = {
6800 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6801 		.rpc_cred = ctx->cred,
6802 	};
6803 	struct rpc_task_setup task_setup_data = {
6804 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6805 		.rpc_message = &msg,
6806 		.callback_ops = &nfs4_locku_ops,
6807 		.workqueue = nfsiod_workqueue,
6808 		.flags = RPC_TASK_ASYNC,
6809 	};
6810 
6811 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6812 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6813 
6814 	/* Ensure this is an unlock - when canceling a lock, the
6815 	 * canceled lock is passed in, and it won't be an unlock.
6816 	 */
6817 	fl->fl_type = F_UNLCK;
6818 	if (fl->fl_flags & FL_CLOSE)
6819 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6820 
6821 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6822 	if (data == NULL) {
6823 		nfs_free_seqid(seqid);
6824 		return ERR_PTR(-ENOMEM);
6825 	}
6826 
6827 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6828 	msg.rpc_argp = &data->arg;
6829 	msg.rpc_resp = &data->res;
6830 	task_setup_data.callback_data = data;
6831 	return rpc_run_task(&task_setup_data);
6832 }
6833 
6834 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6835 {
6836 	struct inode *inode = state->inode;
6837 	struct nfs4_state_owner *sp = state->owner;
6838 	struct nfs_inode *nfsi = NFS_I(inode);
6839 	struct nfs_seqid *seqid;
6840 	struct nfs4_lock_state *lsp;
6841 	struct rpc_task *task;
6842 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6843 	int status = 0;
6844 	unsigned char fl_flags = request->fl_flags;
6845 
6846 	status = nfs4_set_lock_state(state, request);
6847 	/* Unlock _before_ we do the RPC call */
6848 	request->fl_flags |= FL_EXISTS;
6849 	/* Exclude nfs_delegation_claim_locks() */
6850 	mutex_lock(&sp->so_delegreturn_mutex);
6851 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6852 	down_read(&nfsi->rwsem);
6853 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6854 		up_read(&nfsi->rwsem);
6855 		mutex_unlock(&sp->so_delegreturn_mutex);
6856 		goto out;
6857 	}
6858 	up_read(&nfsi->rwsem);
6859 	mutex_unlock(&sp->so_delegreturn_mutex);
6860 	if (status != 0)
6861 		goto out;
6862 	/* Is this a delegated lock? */
6863 	lsp = request->fl_u.nfs4_fl.owner;
6864 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6865 		goto out;
6866 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6867 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6868 	status = -ENOMEM;
6869 	if (IS_ERR(seqid))
6870 		goto out;
6871 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6872 	status = PTR_ERR(task);
6873 	if (IS_ERR(task))
6874 		goto out;
6875 	status = rpc_wait_for_completion_task(task);
6876 	rpc_put_task(task);
6877 out:
6878 	request->fl_flags = fl_flags;
6879 	trace_nfs4_unlock(request, state, F_SETLK, status);
6880 	return status;
6881 }
6882 
6883 struct nfs4_lockdata {
6884 	struct nfs_lock_args arg;
6885 	struct nfs_lock_res res;
6886 	struct nfs4_lock_state *lsp;
6887 	struct nfs_open_context *ctx;
6888 	struct file_lock fl;
6889 	unsigned long timestamp;
6890 	int rpc_status;
6891 	int cancelled;
6892 	struct nfs_server *server;
6893 };
6894 
6895 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6896 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6897 		gfp_t gfp_mask)
6898 {
6899 	struct nfs4_lockdata *p;
6900 	struct inode *inode = lsp->ls_state->inode;
6901 	struct nfs_server *server = NFS_SERVER(inode);
6902 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6903 
6904 	p = kzalloc(sizeof(*p), gfp_mask);
6905 	if (p == NULL)
6906 		return NULL;
6907 
6908 	p->arg.fh = NFS_FH(inode);
6909 	p->arg.fl = &p->fl;
6910 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6911 	if (IS_ERR(p->arg.open_seqid))
6912 		goto out_free;
6913 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6914 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6915 	if (IS_ERR(p->arg.lock_seqid))
6916 		goto out_free_seqid;
6917 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6918 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6919 	p->arg.lock_owner.s_dev = server->s_dev;
6920 	p->res.lock_seqid = p->arg.lock_seqid;
6921 	p->lsp = lsp;
6922 	p->server = server;
6923 	p->ctx = get_nfs_open_context(ctx);
6924 	locks_init_lock(&p->fl);
6925 	locks_copy_lock(&p->fl, fl);
6926 	return p;
6927 out_free_seqid:
6928 	nfs_free_seqid(p->arg.open_seqid);
6929 out_free:
6930 	kfree(p);
6931 	return NULL;
6932 }
6933 
6934 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6935 {
6936 	struct nfs4_lockdata *data = calldata;
6937 	struct nfs4_state *state = data->lsp->ls_state;
6938 
6939 	dprintk("%s: begin!\n", __func__);
6940 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6941 		goto out_wait;
6942 	/* Do we need to do an open_to_lock_owner? */
6943 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6944 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6945 			goto out_release_lock_seqid;
6946 		}
6947 		nfs4_stateid_copy(&data->arg.open_stateid,
6948 				&state->open_stateid);
6949 		data->arg.new_lock_owner = 1;
6950 		data->res.open_seqid = data->arg.open_seqid;
6951 	} else {
6952 		data->arg.new_lock_owner = 0;
6953 		nfs4_stateid_copy(&data->arg.lock_stateid,
6954 				&data->lsp->ls_stateid);
6955 	}
6956 	if (!nfs4_valid_open_stateid(state)) {
6957 		data->rpc_status = -EBADF;
6958 		task->tk_action = NULL;
6959 		goto out_release_open_seqid;
6960 	}
6961 	data->timestamp = jiffies;
6962 	if (nfs4_setup_sequence(data->server->nfs_client,
6963 				&data->arg.seq_args,
6964 				&data->res.seq_res,
6965 				task) == 0)
6966 		return;
6967 out_release_open_seqid:
6968 	nfs_release_seqid(data->arg.open_seqid);
6969 out_release_lock_seqid:
6970 	nfs_release_seqid(data->arg.lock_seqid);
6971 out_wait:
6972 	nfs4_sequence_done(task, &data->res.seq_res);
6973 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6974 }
6975 
6976 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6977 {
6978 	struct nfs4_lockdata *data = calldata;
6979 	struct nfs4_lock_state *lsp = data->lsp;
6980 
6981 	dprintk("%s: begin!\n", __func__);
6982 
6983 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6984 		return;
6985 
6986 	data->rpc_status = task->tk_status;
6987 	switch (task->tk_status) {
6988 	case 0:
6989 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6990 				data->timestamp);
6991 		if (data->arg.new_lock && !data->cancelled) {
6992 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6993 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6994 				goto out_restart;
6995 		}
6996 		if (data->arg.new_lock_owner != 0) {
6997 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
6998 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6999 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7000 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7001 			goto out_restart;
7002 		break;
7003 	case -NFS4ERR_BAD_STATEID:
7004 	case -NFS4ERR_OLD_STATEID:
7005 	case -NFS4ERR_STALE_STATEID:
7006 	case -NFS4ERR_EXPIRED:
7007 		if (data->arg.new_lock_owner != 0) {
7008 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7009 						&lsp->ls_state->open_stateid))
7010 				goto out_restart;
7011 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7012 						&lsp->ls_stateid))
7013 				goto out_restart;
7014 	}
7015 out_done:
7016 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7017 	return;
7018 out_restart:
7019 	if (!data->cancelled)
7020 		rpc_restart_call_prepare(task);
7021 	goto out_done;
7022 }
7023 
7024 static void nfs4_lock_release(void *calldata)
7025 {
7026 	struct nfs4_lockdata *data = calldata;
7027 
7028 	dprintk("%s: begin!\n", __func__);
7029 	nfs_free_seqid(data->arg.open_seqid);
7030 	if (data->cancelled && data->rpc_status == 0) {
7031 		struct rpc_task *task;
7032 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7033 				data->arg.lock_seqid);
7034 		if (!IS_ERR(task))
7035 			rpc_put_task_async(task);
7036 		dprintk("%s: cancelling lock!\n", __func__);
7037 	} else
7038 		nfs_free_seqid(data->arg.lock_seqid);
7039 	nfs4_put_lock_state(data->lsp);
7040 	put_nfs_open_context(data->ctx);
7041 	kfree(data);
7042 	dprintk("%s: done!\n", __func__);
7043 }
7044 
7045 static const struct rpc_call_ops nfs4_lock_ops = {
7046 	.rpc_call_prepare = nfs4_lock_prepare,
7047 	.rpc_call_done = nfs4_lock_done,
7048 	.rpc_release = nfs4_lock_release,
7049 };
7050 
7051 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7052 {
7053 	switch (error) {
7054 	case -NFS4ERR_ADMIN_REVOKED:
7055 	case -NFS4ERR_EXPIRED:
7056 	case -NFS4ERR_BAD_STATEID:
7057 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7058 		if (new_lock_owner != 0 ||
7059 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7060 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7061 		break;
7062 	case -NFS4ERR_STALE_STATEID:
7063 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7064 		nfs4_schedule_lease_recovery(server->nfs_client);
7065 	}
7066 }
7067 
7068 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7069 {
7070 	struct nfs4_lockdata *data;
7071 	struct rpc_task *task;
7072 	struct rpc_message msg = {
7073 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7074 		.rpc_cred = state->owner->so_cred,
7075 	};
7076 	struct rpc_task_setup task_setup_data = {
7077 		.rpc_client = NFS_CLIENT(state->inode),
7078 		.rpc_message = &msg,
7079 		.callback_ops = &nfs4_lock_ops,
7080 		.workqueue = nfsiod_workqueue,
7081 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7082 	};
7083 	int ret;
7084 
7085 	dprintk("%s: begin!\n", __func__);
7086 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7087 			fl->fl_u.nfs4_fl.owner,
7088 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7089 	if (data == NULL)
7090 		return -ENOMEM;
7091 	if (IS_SETLKW(cmd))
7092 		data->arg.block = 1;
7093 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7094 				recovery_type > NFS_LOCK_NEW);
7095 	msg.rpc_argp = &data->arg;
7096 	msg.rpc_resp = &data->res;
7097 	task_setup_data.callback_data = data;
7098 	if (recovery_type > NFS_LOCK_NEW) {
7099 		if (recovery_type == NFS_LOCK_RECLAIM)
7100 			data->arg.reclaim = NFS_LOCK_RECLAIM;
7101 	} else
7102 		data->arg.new_lock = 1;
7103 	task = rpc_run_task(&task_setup_data);
7104 	if (IS_ERR(task))
7105 		return PTR_ERR(task);
7106 	ret = rpc_wait_for_completion_task(task);
7107 	if (ret == 0) {
7108 		ret = data->rpc_status;
7109 		if (ret)
7110 			nfs4_handle_setlk_error(data->server, data->lsp,
7111 					data->arg.new_lock_owner, ret);
7112 	} else
7113 		data->cancelled = true;
7114 	rpc_put_task(task);
7115 	dprintk("%s: done, ret = %d!\n", __func__, ret);
7116 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7117 	return ret;
7118 }
7119 
7120 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7121 {
7122 	struct nfs_server *server = NFS_SERVER(state->inode);
7123 	struct nfs4_exception exception = {
7124 		.inode = state->inode,
7125 	};
7126 	int err;
7127 
7128 	do {
7129 		/* Cache the lock if possible... */
7130 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7131 			return 0;
7132 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7133 		if (err != -NFS4ERR_DELAY)
7134 			break;
7135 		nfs4_handle_exception(server, err, &exception);
7136 	} while (exception.retry);
7137 	return err;
7138 }
7139 
7140 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7141 {
7142 	struct nfs_server *server = NFS_SERVER(state->inode);
7143 	struct nfs4_exception exception = {
7144 		.inode = state->inode,
7145 	};
7146 	int err;
7147 
7148 	err = nfs4_set_lock_state(state, request);
7149 	if (err != 0)
7150 		return err;
7151 	if (!recover_lost_locks) {
7152 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7153 		return 0;
7154 	}
7155 	do {
7156 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7157 			return 0;
7158 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7159 		switch (err) {
7160 		default:
7161 			goto out;
7162 		case -NFS4ERR_GRACE:
7163 		case -NFS4ERR_DELAY:
7164 			nfs4_handle_exception(server, err, &exception);
7165 			err = 0;
7166 		}
7167 	} while (exception.retry);
7168 out:
7169 	return err;
7170 }
7171 
7172 #if defined(CONFIG_NFS_V4_1)
7173 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7174 {
7175 	struct nfs4_lock_state *lsp;
7176 	int status;
7177 
7178 	status = nfs4_set_lock_state(state, request);
7179 	if (status != 0)
7180 		return status;
7181 	lsp = request->fl_u.nfs4_fl.owner;
7182 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7183 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7184 		return 0;
7185 	return nfs4_lock_expired(state, request);
7186 }
7187 #endif
7188 
7189 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7190 {
7191 	struct nfs_inode *nfsi = NFS_I(state->inode);
7192 	struct nfs4_state_owner *sp = state->owner;
7193 	unsigned char fl_flags = request->fl_flags;
7194 	int status;
7195 
7196 	request->fl_flags |= FL_ACCESS;
7197 	status = locks_lock_inode_wait(state->inode, request);
7198 	if (status < 0)
7199 		goto out;
7200 	mutex_lock(&sp->so_delegreturn_mutex);
7201 	down_read(&nfsi->rwsem);
7202 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7203 		/* Yes: cache locks! */
7204 		/* ...but avoid races with delegation recall... */
7205 		request->fl_flags = fl_flags & ~FL_SLEEP;
7206 		status = locks_lock_inode_wait(state->inode, request);
7207 		up_read(&nfsi->rwsem);
7208 		mutex_unlock(&sp->so_delegreturn_mutex);
7209 		goto out;
7210 	}
7211 	up_read(&nfsi->rwsem);
7212 	mutex_unlock(&sp->so_delegreturn_mutex);
7213 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7214 out:
7215 	request->fl_flags = fl_flags;
7216 	return status;
7217 }
7218 
7219 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7220 {
7221 	struct nfs4_exception exception = {
7222 		.state = state,
7223 		.inode = state->inode,
7224 		.interruptible = true,
7225 	};
7226 	int err;
7227 
7228 	do {
7229 		err = _nfs4_proc_setlk(state, cmd, request);
7230 		if (err == -NFS4ERR_DENIED)
7231 			err = -EAGAIN;
7232 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7233 				err, &exception);
7234 	} while (exception.retry);
7235 	return err;
7236 }
7237 
7238 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7239 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7240 
7241 static int
7242 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7243 			struct file_lock *request)
7244 {
7245 	int		status = -ERESTARTSYS;
7246 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7247 
7248 	while(!signalled()) {
7249 		status = nfs4_proc_setlk(state, cmd, request);
7250 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7251 			break;
7252 		freezable_schedule_timeout_interruptible(timeout);
7253 		timeout *= 2;
7254 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7255 		status = -ERESTARTSYS;
7256 	}
7257 	return status;
7258 }
7259 
7260 #ifdef CONFIG_NFS_V4_1
7261 struct nfs4_lock_waiter {
7262 	struct task_struct	*task;
7263 	struct inode		*inode;
7264 	struct nfs_lowner	*owner;
7265 };
7266 
7267 static int
7268 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7269 {
7270 	int ret;
7271 	struct nfs4_lock_waiter	*waiter	= wait->private;
7272 
7273 	/* NULL key means to wake up everyone */
7274 	if (key) {
7275 		struct cb_notify_lock_args	*cbnl = key;
7276 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7277 						*wowner = waiter->owner;
7278 
7279 		/* Only wake if the callback was for the same owner. */
7280 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7281 			return 0;
7282 
7283 		/* Make sure it's for the right inode */
7284 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7285 			return 0;
7286 	}
7287 
7288 	/* override "private" so we can use default_wake_function */
7289 	wait->private = waiter->task;
7290 	ret = woken_wake_function(wait, mode, flags, key);
7291 	if (ret)
7292 		list_del_init(&wait->entry);
7293 	wait->private = waiter;
7294 	return ret;
7295 }
7296 
7297 static int
7298 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7299 {
7300 	int status = -ERESTARTSYS;
7301 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7302 	struct nfs_server *server = NFS_SERVER(state->inode);
7303 	struct nfs_client *clp = server->nfs_client;
7304 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7305 	struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7306 				    .id = lsp->ls_seqid.owner_id,
7307 				    .s_dev = server->s_dev };
7308 	struct nfs4_lock_waiter waiter = { .task  = current,
7309 					   .inode = state->inode,
7310 					   .owner = &owner};
7311 	wait_queue_entry_t wait;
7312 
7313 	/* Don't bother with waitqueue if we don't expect a callback */
7314 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7315 		return nfs4_retry_setlk_simple(state, cmd, request);
7316 
7317 	init_wait(&wait);
7318 	wait.private = &waiter;
7319 	wait.func = nfs4_wake_lock_waiter;
7320 
7321 	while(!signalled()) {
7322 		add_wait_queue(q, &wait);
7323 		status = nfs4_proc_setlk(state, cmd, request);
7324 		if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7325 			finish_wait(q, &wait);
7326 			break;
7327 		}
7328 
7329 		status = -ERESTARTSYS;
7330 		freezer_do_not_count();
7331 		wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7332 		freezer_count();
7333 		finish_wait(q, &wait);
7334 	}
7335 
7336 	return status;
7337 }
7338 #else /* !CONFIG_NFS_V4_1 */
7339 static inline int
7340 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7341 {
7342 	return nfs4_retry_setlk_simple(state, cmd, request);
7343 }
7344 #endif
7345 
7346 static int
7347 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7348 {
7349 	struct nfs_open_context *ctx;
7350 	struct nfs4_state *state;
7351 	int status;
7352 
7353 	/* verify open state */
7354 	ctx = nfs_file_open_context(filp);
7355 	state = ctx->state;
7356 
7357 	if (IS_GETLK(cmd)) {
7358 		if (state != NULL)
7359 			return nfs4_proc_getlk(state, F_GETLK, request);
7360 		return 0;
7361 	}
7362 
7363 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7364 		return -EINVAL;
7365 
7366 	if (request->fl_type == F_UNLCK) {
7367 		if (state != NULL)
7368 			return nfs4_proc_unlck(state, cmd, request);
7369 		return 0;
7370 	}
7371 
7372 	if (state == NULL)
7373 		return -ENOLCK;
7374 
7375 	if ((request->fl_flags & FL_POSIX) &&
7376 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7377 		return -ENOLCK;
7378 
7379 	/*
7380 	 * Don't rely on the VFS having checked the file open mode,
7381 	 * since it won't do this for flock() locks.
7382 	 */
7383 	switch (request->fl_type) {
7384 	case F_RDLCK:
7385 		if (!(filp->f_mode & FMODE_READ))
7386 			return -EBADF;
7387 		break;
7388 	case F_WRLCK:
7389 		if (!(filp->f_mode & FMODE_WRITE))
7390 			return -EBADF;
7391 	}
7392 
7393 	status = nfs4_set_lock_state(state, request);
7394 	if (status != 0)
7395 		return status;
7396 
7397 	return nfs4_retry_setlk(state, cmd, request);
7398 }
7399 
7400 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7401 {
7402 	struct nfs_server *server = NFS_SERVER(state->inode);
7403 	int err;
7404 
7405 	err = nfs4_set_lock_state(state, fl);
7406 	if (err != 0)
7407 		return err;
7408 	do {
7409 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7410 		if (err != -NFS4ERR_DELAY)
7411 			break;
7412 		ssleep(1);
7413 	} while (err == -NFS4ERR_DELAY);
7414 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7415 }
7416 
7417 struct nfs_release_lockowner_data {
7418 	struct nfs4_lock_state *lsp;
7419 	struct nfs_server *server;
7420 	struct nfs_release_lockowner_args args;
7421 	struct nfs_release_lockowner_res res;
7422 	unsigned long timestamp;
7423 };
7424 
7425 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7426 {
7427 	struct nfs_release_lockowner_data *data = calldata;
7428 	struct nfs_server *server = data->server;
7429 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7430 			   &data->res.seq_res, task);
7431 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7432 	data->timestamp = jiffies;
7433 }
7434 
7435 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7436 {
7437 	struct nfs_release_lockowner_data *data = calldata;
7438 	struct nfs_server *server = data->server;
7439 
7440 	nfs40_sequence_done(task, &data->res.seq_res);
7441 
7442 	switch (task->tk_status) {
7443 	case 0:
7444 		renew_lease(server, data->timestamp);
7445 		break;
7446 	case -NFS4ERR_STALE_CLIENTID:
7447 	case -NFS4ERR_EXPIRED:
7448 		nfs4_schedule_lease_recovery(server->nfs_client);
7449 		break;
7450 	case -NFS4ERR_LEASE_MOVED:
7451 	case -NFS4ERR_DELAY:
7452 		if (nfs4_async_handle_error(task, server,
7453 					    NULL, NULL) == -EAGAIN)
7454 			rpc_restart_call_prepare(task);
7455 	}
7456 }
7457 
7458 static void nfs4_release_lockowner_release(void *calldata)
7459 {
7460 	struct nfs_release_lockowner_data *data = calldata;
7461 	nfs4_free_lock_state(data->server, data->lsp);
7462 	kfree(calldata);
7463 }
7464 
7465 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7466 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7467 	.rpc_call_done = nfs4_release_lockowner_done,
7468 	.rpc_release = nfs4_release_lockowner_release,
7469 };
7470 
7471 static void
7472 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7473 {
7474 	struct nfs_release_lockowner_data *data;
7475 	struct rpc_message msg = {
7476 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7477 	};
7478 
7479 	if (server->nfs_client->cl_mvops->minor_version != 0)
7480 		return;
7481 
7482 	data = kmalloc(sizeof(*data), GFP_NOFS);
7483 	if (!data)
7484 		return;
7485 	data->lsp = lsp;
7486 	data->server = server;
7487 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7488 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7489 	data->args.lock_owner.s_dev = server->s_dev;
7490 
7491 	msg.rpc_argp = &data->args;
7492 	msg.rpc_resp = &data->res;
7493 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7494 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7495 }
7496 
7497 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7498 
7499 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7500 				   struct dentry *unused, struct inode *inode,
7501 				   const char *key, const void *buf,
7502 				   size_t buflen, int flags)
7503 {
7504 	return nfs4_proc_set_acl(inode, buf, buflen);
7505 }
7506 
7507 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7508 				   struct dentry *unused, struct inode *inode,
7509 				   const char *key, void *buf, size_t buflen)
7510 {
7511 	return nfs4_proc_get_acl(inode, buf, buflen);
7512 }
7513 
7514 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7515 {
7516 	return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7517 }
7518 
7519 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7520 
7521 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7522 				     struct dentry *unused, struct inode *inode,
7523 				     const char *key, const void *buf,
7524 				     size_t buflen, int flags)
7525 {
7526 	if (security_ismaclabel(key))
7527 		return nfs4_set_security_label(inode, buf, buflen);
7528 
7529 	return -EOPNOTSUPP;
7530 }
7531 
7532 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7533 				     struct dentry *unused, struct inode *inode,
7534 				     const char *key, void *buf, size_t buflen)
7535 {
7536 	if (security_ismaclabel(key))
7537 		return nfs4_get_security_label(inode, buf, buflen);
7538 	return -EOPNOTSUPP;
7539 }
7540 
7541 static ssize_t
7542 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7543 {
7544 	int len = 0;
7545 
7546 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7547 		len = security_inode_listsecurity(inode, list, list_len);
7548 		if (len >= 0 && list_len && len > list_len)
7549 			return -ERANGE;
7550 	}
7551 	return len;
7552 }
7553 
7554 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7555 	.prefix = XATTR_SECURITY_PREFIX,
7556 	.get	= nfs4_xattr_get_nfs4_label,
7557 	.set	= nfs4_xattr_set_nfs4_label,
7558 };
7559 
7560 #else
7561 
7562 static ssize_t
7563 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7564 {
7565 	return 0;
7566 }
7567 
7568 #endif
7569 
7570 #ifdef CONFIG_NFS_V4_2
7571 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7572 				    struct dentry *unused, struct inode *inode,
7573 				    const char *key, const void *buf,
7574 				    size_t buflen, int flags)
7575 {
7576 	struct nfs_access_entry cache;
7577 	int ret;
7578 
7579 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7580 		return -EOPNOTSUPP;
7581 
7582 	/*
7583 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7584 	 * flags right now. Handling of xattr operations use the normal
7585 	 * file read/write permissions.
7586 	 *
7587 	 * Just in case the server has other ideas (which RFC 8276 allows),
7588 	 * do a cached access check for the XA* flags to possibly avoid
7589 	 * doing an RPC and getting EACCES back.
7590 	 */
7591 	if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7592 		if (!(cache.mask & NFS_ACCESS_XAWRITE))
7593 			return -EACCES;
7594 	}
7595 
7596 	if (buf == NULL) {
7597 		ret = nfs42_proc_removexattr(inode, key);
7598 		if (!ret)
7599 			nfs4_xattr_cache_remove(inode, key);
7600 	} else {
7601 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7602 		if (!ret)
7603 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7604 	}
7605 
7606 	return ret;
7607 }
7608 
7609 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7610 				    struct dentry *unused, struct inode *inode,
7611 				    const char *key, void *buf, size_t buflen)
7612 {
7613 	struct nfs_access_entry cache;
7614 	ssize_t ret;
7615 
7616 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7617 		return -EOPNOTSUPP;
7618 
7619 	if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7620 		if (!(cache.mask & NFS_ACCESS_XAREAD))
7621 			return -EACCES;
7622 	}
7623 
7624 	ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7625 	if (ret)
7626 		return ret;
7627 
7628 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7629 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7630 		return ret;
7631 
7632 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7633 
7634 	return ret;
7635 }
7636 
7637 static ssize_t
7638 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7639 {
7640 	u64 cookie;
7641 	bool eof;
7642 	ssize_t ret, size;
7643 	char *buf;
7644 	size_t buflen;
7645 	struct nfs_access_entry cache;
7646 
7647 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7648 		return 0;
7649 
7650 	if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7651 		if (!(cache.mask & NFS_ACCESS_XALIST))
7652 			return 0;
7653 	}
7654 
7655 	ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7656 	if (ret)
7657 		return ret;
7658 
7659 	ret = nfs4_xattr_cache_list(inode, list, list_len);
7660 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7661 		return ret;
7662 
7663 	cookie = 0;
7664 	eof = false;
7665 	buflen = list_len ? list_len : XATTR_LIST_MAX;
7666 	buf = list_len ? list : NULL;
7667 	size = 0;
7668 
7669 	while (!eof) {
7670 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
7671 		    &cookie, &eof);
7672 		if (ret < 0)
7673 			return ret;
7674 
7675 		if (list_len) {
7676 			buf += ret;
7677 			buflen -= ret;
7678 		}
7679 		size += ret;
7680 	}
7681 
7682 	if (list_len)
7683 		nfs4_xattr_cache_set_list(inode, list, size);
7684 
7685 	return size;
7686 }
7687 
7688 #else
7689 
7690 static ssize_t
7691 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7692 {
7693 	return 0;
7694 }
7695 #endif /* CONFIG_NFS_V4_2 */
7696 
7697 /*
7698  * nfs_fhget will use either the mounted_on_fileid or the fileid
7699  */
7700 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7701 {
7702 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7703 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7704 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7705 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7706 		return;
7707 
7708 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7709 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7710 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7711 	fattr->nlink = 2;
7712 }
7713 
7714 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7715 				   const struct qstr *name,
7716 				   struct nfs4_fs_locations *fs_locations,
7717 				   struct page *page)
7718 {
7719 	struct nfs_server *server = NFS_SERVER(dir);
7720 	u32 bitmask[3];
7721 	struct nfs4_fs_locations_arg args = {
7722 		.dir_fh = NFS_FH(dir),
7723 		.name = name,
7724 		.page = page,
7725 		.bitmask = bitmask,
7726 	};
7727 	struct nfs4_fs_locations_res res = {
7728 		.fs_locations = fs_locations,
7729 	};
7730 	struct rpc_message msg = {
7731 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7732 		.rpc_argp = &args,
7733 		.rpc_resp = &res,
7734 	};
7735 	int status;
7736 
7737 	dprintk("%s: start\n", __func__);
7738 
7739 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7740 	bitmask[1] = nfs4_fattr_bitmap[1];
7741 
7742 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
7743 	 * is not supported */
7744 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7745 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
7746 	else
7747 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7748 
7749 	nfs_fattr_init(&fs_locations->fattr);
7750 	fs_locations->server = server;
7751 	fs_locations->nlocations = 0;
7752 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7753 	dprintk("%s: returned status = %d\n", __func__, status);
7754 	return status;
7755 }
7756 
7757 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7758 			   const struct qstr *name,
7759 			   struct nfs4_fs_locations *fs_locations,
7760 			   struct page *page)
7761 {
7762 	struct nfs4_exception exception = {
7763 		.interruptible = true,
7764 	};
7765 	int err;
7766 	do {
7767 		err = _nfs4_proc_fs_locations(client, dir, name,
7768 				fs_locations, page);
7769 		trace_nfs4_get_fs_locations(dir, name, err);
7770 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
7771 				&exception);
7772 	} while (exception.retry);
7773 	return err;
7774 }
7775 
7776 /*
7777  * This operation also signals the server that this client is
7778  * performing migration recovery.  The server can stop returning
7779  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7780  * appended to this compound to identify the client ID which is
7781  * performing recovery.
7782  */
7783 static int _nfs40_proc_get_locations(struct inode *inode,
7784 				     struct nfs4_fs_locations *locations,
7785 				     struct page *page, const struct cred *cred)
7786 {
7787 	struct nfs_server *server = NFS_SERVER(inode);
7788 	struct rpc_clnt *clnt = server->client;
7789 	u32 bitmask[2] = {
7790 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7791 	};
7792 	struct nfs4_fs_locations_arg args = {
7793 		.clientid	= server->nfs_client->cl_clientid,
7794 		.fh		= NFS_FH(inode),
7795 		.page		= page,
7796 		.bitmask	= bitmask,
7797 		.migration	= 1,		/* skip LOOKUP */
7798 		.renew		= 1,		/* append RENEW */
7799 	};
7800 	struct nfs4_fs_locations_res res = {
7801 		.fs_locations	= locations,
7802 		.migration	= 1,
7803 		.renew		= 1,
7804 	};
7805 	struct rpc_message msg = {
7806 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7807 		.rpc_argp	= &args,
7808 		.rpc_resp	= &res,
7809 		.rpc_cred	= cred,
7810 	};
7811 	unsigned long now = jiffies;
7812 	int status;
7813 
7814 	nfs_fattr_init(&locations->fattr);
7815 	locations->server = server;
7816 	locations->nlocations = 0;
7817 
7818 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7819 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7820 					&args.seq_args, &res.seq_res);
7821 	if (status)
7822 		return status;
7823 
7824 	renew_lease(server, now);
7825 	return 0;
7826 }
7827 
7828 #ifdef CONFIG_NFS_V4_1
7829 
7830 /*
7831  * This operation also signals the server that this client is
7832  * performing migration recovery.  The server can stop asserting
7833  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7834  * performing this operation is identified in the SEQUENCE
7835  * operation in this compound.
7836  *
7837  * When the client supports GETATTR(fs_locations_info), it can
7838  * be plumbed in here.
7839  */
7840 static int _nfs41_proc_get_locations(struct inode *inode,
7841 				     struct nfs4_fs_locations *locations,
7842 				     struct page *page, const struct cred *cred)
7843 {
7844 	struct nfs_server *server = NFS_SERVER(inode);
7845 	struct rpc_clnt *clnt = server->client;
7846 	u32 bitmask[2] = {
7847 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7848 	};
7849 	struct nfs4_fs_locations_arg args = {
7850 		.fh		= NFS_FH(inode),
7851 		.page		= page,
7852 		.bitmask	= bitmask,
7853 		.migration	= 1,		/* skip LOOKUP */
7854 	};
7855 	struct nfs4_fs_locations_res res = {
7856 		.fs_locations	= locations,
7857 		.migration	= 1,
7858 	};
7859 	struct rpc_message msg = {
7860 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7861 		.rpc_argp	= &args,
7862 		.rpc_resp	= &res,
7863 		.rpc_cred	= cred,
7864 	};
7865 	int status;
7866 
7867 	nfs_fattr_init(&locations->fattr);
7868 	locations->server = server;
7869 	locations->nlocations = 0;
7870 
7871 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7872 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7873 					&args.seq_args, &res.seq_res);
7874 	if (status == NFS4_OK &&
7875 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7876 		status = -NFS4ERR_LEASE_MOVED;
7877 	return status;
7878 }
7879 
7880 #endif	/* CONFIG_NFS_V4_1 */
7881 
7882 /**
7883  * nfs4_proc_get_locations - discover locations for a migrated FSID
7884  * @inode: inode on FSID that is migrating
7885  * @locations: result of query
7886  * @page: buffer
7887  * @cred: credential to use for this operation
7888  *
7889  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7890  * operation failed, or a negative errno if a local error occurred.
7891  *
7892  * On success, "locations" is filled in, but if the server has
7893  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7894  * asserted.
7895  *
7896  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7897  * from this client that require migration recovery.
7898  */
7899 int nfs4_proc_get_locations(struct inode *inode,
7900 			    struct nfs4_fs_locations *locations,
7901 			    struct page *page, const struct cred *cred)
7902 {
7903 	struct nfs_server *server = NFS_SERVER(inode);
7904 	struct nfs_client *clp = server->nfs_client;
7905 	const struct nfs4_mig_recovery_ops *ops =
7906 					clp->cl_mvops->mig_recovery_ops;
7907 	struct nfs4_exception exception = {
7908 		.interruptible = true,
7909 	};
7910 	int status;
7911 
7912 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7913 		(unsigned long long)server->fsid.major,
7914 		(unsigned long long)server->fsid.minor,
7915 		clp->cl_hostname);
7916 	nfs_display_fhandle(NFS_FH(inode), __func__);
7917 
7918 	do {
7919 		status = ops->get_locations(inode, locations, page, cred);
7920 		if (status != -NFS4ERR_DELAY)
7921 			break;
7922 		nfs4_handle_exception(server, status, &exception);
7923 	} while (exception.retry);
7924 	return status;
7925 }
7926 
7927 /*
7928  * This operation also signals the server that this client is
7929  * performing "lease moved" recovery.  The server can stop
7930  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7931  * is appended to this compound to identify the client ID which is
7932  * performing recovery.
7933  */
7934 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7935 {
7936 	struct nfs_server *server = NFS_SERVER(inode);
7937 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7938 	struct rpc_clnt *clnt = server->client;
7939 	struct nfs4_fsid_present_arg args = {
7940 		.fh		= NFS_FH(inode),
7941 		.clientid	= clp->cl_clientid,
7942 		.renew		= 1,		/* append RENEW */
7943 	};
7944 	struct nfs4_fsid_present_res res = {
7945 		.renew		= 1,
7946 	};
7947 	struct rpc_message msg = {
7948 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7949 		.rpc_argp	= &args,
7950 		.rpc_resp	= &res,
7951 		.rpc_cred	= cred,
7952 	};
7953 	unsigned long now = jiffies;
7954 	int status;
7955 
7956 	res.fh = nfs_alloc_fhandle();
7957 	if (res.fh == NULL)
7958 		return -ENOMEM;
7959 
7960 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7961 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7962 						&args.seq_args, &res.seq_res);
7963 	nfs_free_fhandle(res.fh);
7964 	if (status)
7965 		return status;
7966 
7967 	do_renew_lease(clp, now);
7968 	return 0;
7969 }
7970 
7971 #ifdef CONFIG_NFS_V4_1
7972 
7973 /*
7974  * This operation also signals the server that this client is
7975  * performing "lease moved" recovery.  The server can stop asserting
7976  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7977  * this operation is identified in the SEQUENCE operation in this
7978  * compound.
7979  */
7980 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7981 {
7982 	struct nfs_server *server = NFS_SERVER(inode);
7983 	struct rpc_clnt *clnt = server->client;
7984 	struct nfs4_fsid_present_arg args = {
7985 		.fh		= NFS_FH(inode),
7986 	};
7987 	struct nfs4_fsid_present_res res = {
7988 	};
7989 	struct rpc_message msg = {
7990 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7991 		.rpc_argp	= &args,
7992 		.rpc_resp	= &res,
7993 		.rpc_cred	= cred,
7994 	};
7995 	int status;
7996 
7997 	res.fh = nfs_alloc_fhandle();
7998 	if (res.fh == NULL)
7999 		return -ENOMEM;
8000 
8001 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8002 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8003 						&args.seq_args, &res.seq_res);
8004 	nfs_free_fhandle(res.fh);
8005 	if (status == NFS4_OK &&
8006 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8007 		status = -NFS4ERR_LEASE_MOVED;
8008 	return status;
8009 }
8010 
8011 #endif	/* CONFIG_NFS_V4_1 */
8012 
8013 /**
8014  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8015  * @inode: inode on FSID to check
8016  * @cred: credential to use for this operation
8017  *
8018  * Server indicates whether the FSID is present, moved, or not
8019  * recognized.  This operation is necessary to clear a LEASE_MOVED
8020  * condition for this client ID.
8021  *
8022  * Returns NFS4_OK if the FSID is present on this server,
8023  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8024  *  NFS4ERR code if some error occurred on the server, or a
8025  *  negative errno if a local failure occurred.
8026  */
8027 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8028 {
8029 	struct nfs_server *server = NFS_SERVER(inode);
8030 	struct nfs_client *clp = server->nfs_client;
8031 	const struct nfs4_mig_recovery_ops *ops =
8032 					clp->cl_mvops->mig_recovery_ops;
8033 	struct nfs4_exception exception = {
8034 		.interruptible = true,
8035 	};
8036 	int status;
8037 
8038 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8039 		(unsigned long long)server->fsid.major,
8040 		(unsigned long long)server->fsid.minor,
8041 		clp->cl_hostname);
8042 	nfs_display_fhandle(NFS_FH(inode), __func__);
8043 
8044 	do {
8045 		status = ops->fsid_present(inode, cred);
8046 		if (status != -NFS4ERR_DELAY)
8047 			break;
8048 		nfs4_handle_exception(server, status, &exception);
8049 	} while (exception.retry);
8050 	return status;
8051 }
8052 
8053 /*
8054  * If 'use_integrity' is true and the state managment nfs_client
8055  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8056  * and the machine credential as per RFC3530bis and RFC5661 Security
8057  * Considerations sections. Otherwise, just use the user cred with the
8058  * filesystem's rpc_client.
8059  */
8060 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8061 {
8062 	int status;
8063 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8064 	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8065 	struct nfs4_secinfo_arg args = {
8066 		.dir_fh = NFS_FH(dir),
8067 		.name   = name,
8068 	};
8069 	struct nfs4_secinfo_res res = {
8070 		.flavors     = flavors,
8071 	};
8072 	struct rpc_message msg = {
8073 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8074 		.rpc_argp = &args,
8075 		.rpc_resp = &res,
8076 	};
8077 	struct nfs4_call_sync_data data = {
8078 		.seq_server = NFS_SERVER(dir),
8079 		.seq_args = &args.seq_args,
8080 		.seq_res = &res.seq_res,
8081 	};
8082 	struct rpc_task_setup task_setup = {
8083 		.rpc_client = clnt,
8084 		.rpc_message = &msg,
8085 		.callback_ops = clp->cl_mvops->call_sync_ops,
8086 		.callback_data = &data,
8087 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8088 	};
8089 	const struct cred *cred = NULL;
8090 
8091 	if (use_integrity) {
8092 		clnt = clp->cl_rpcclient;
8093 		task_setup.rpc_client = clnt;
8094 
8095 		cred = nfs4_get_clid_cred(clp);
8096 		msg.rpc_cred = cred;
8097 	}
8098 
8099 	dprintk("NFS call  secinfo %s\n", name->name);
8100 
8101 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8102 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8103 	status = nfs4_call_sync_custom(&task_setup);
8104 
8105 	dprintk("NFS reply  secinfo: %d\n", status);
8106 
8107 	put_cred(cred);
8108 	return status;
8109 }
8110 
8111 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8112 		      struct nfs4_secinfo_flavors *flavors)
8113 {
8114 	struct nfs4_exception exception = {
8115 		.interruptible = true,
8116 	};
8117 	int err;
8118 	do {
8119 		err = -NFS4ERR_WRONGSEC;
8120 
8121 		/* try to use integrity protection with machine cred */
8122 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8123 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8124 
8125 		/*
8126 		 * if unable to use integrity protection, or SECINFO with
8127 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8128 		 * disallowed by spec, but exists in deployed servers) use
8129 		 * the current filesystem's rpc_client and the user cred.
8130 		 */
8131 		if (err == -NFS4ERR_WRONGSEC)
8132 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8133 
8134 		trace_nfs4_secinfo(dir, name, err);
8135 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8136 				&exception);
8137 	} while (exception.retry);
8138 	return err;
8139 }
8140 
8141 #ifdef CONFIG_NFS_V4_1
8142 /*
8143  * Check the exchange flags returned by the server for invalid flags, having
8144  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8145  * DS flags set.
8146  */
8147 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8148 {
8149 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8150 		goto out_inval;
8151 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8152 		goto out_inval;
8153 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8154 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8155 		goto out_inval;
8156 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8157 		goto out_inval;
8158 	return NFS_OK;
8159 out_inval:
8160 	return -NFS4ERR_INVAL;
8161 }
8162 
8163 static bool
8164 nfs41_same_server_scope(struct nfs41_server_scope *a,
8165 			struct nfs41_server_scope *b)
8166 {
8167 	if (a->server_scope_sz != b->server_scope_sz)
8168 		return false;
8169 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8170 }
8171 
8172 static void
8173 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8174 {
8175 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8176 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8177 	struct nfs_client *clp = args->client;
8178 
8179 	switch (task->tk_status) {
8180 	case -NFS4ERR_BADSESSION:
8181 	case -NFS4ERR_DEADSESSION:
8182 		nfs4_schedule_session_recovery(clp->cl_session,
8183 				task->tk_status);
8184 	}
8185 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8186 			res->dir != NFS4_CDFS4_BOTH) {
8187 		rpc_task_close_connection(task);
8188 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8189 			rpc_restart_call(task);
8190 	}
8191 }
8192 
8193 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8194 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8195 };
8196 
8197 /*
8198  * nfs4_proc_bind_one_conn_to_session()
8199  *
8200  * The 4.1 client currently uses the same TCP connection for the
8201  * fore and backchannel.
8202  */
8203 static
8204 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8205 		struct rpc_xprt *xprt,
8206 		struct nfs_client *clp,
8207 		const struct cred *cred)
8208 {
8209 	int status;
8210 	struct nfs41_bind_conn_to_session_args args = {
8211 		.client = clp,
8212 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8213 		.retries = 0,
8214 	};
8215 	struct nfs41_bind_conn_to_session_res res;
8216 	struct rpc_message msg = {
8217 		.rpc_proc =
8218 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8219 		.rpc_argp = &args,
8220 		.rpc_resp = &res,
8221 		.rpc_cred = cred,
8222 	};
8223 	struct rpc_task_setup task_setup_data = {
8224 		.rpc_client = clnt,
8225 		.rpc_xprt = xprt,
8226 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8227 		.rpc_message = &msg,
8228 		.flags = RPC_TASK_TIMEOUT,
8229 	};
8230 	struct rpc_task *task;
8231 
8232 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8233 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8234 		args.dir = NFS4_CDFC4_FORE;
8235 
8236 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8237 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8238 		args.dir = NFS4_CDFC4_FORE;
8239 
8240 	task = rpc_run_task(&task_setup_data);
8241 	if (!IS_ERR(task)) {
8242 		status = task->tk_status;
8243 		rpc_put_task(task);
8244 	} else
8245 		status = PTR_ERR(task);
8246 	trace_nfs4_bind_conn_to_session(clp, status);
8247 	if (status == 0) {
8248 		if (memcmp(res.sessionid.data,
8249 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8250 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8251 			return -EIO;
8252 		}
8253 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8254 			dprintk("NFS: %s: Unexpected direction from server\n",
8255 				__func__);
8256 			return -EIO;
8257 		}
8258 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8259 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8260 				__func__);
8261 			return -EIO;
8262 		}
8263 	}
8264 
8265 	return status;
8266 }
8267 
8268 struct rpc_bind_conn_calldata {
8269 	struct nfs_client *clp;
8270 	const struct cred *cred;
8271 };
8272 
8273 static int
8274 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8275 		struct rpc_xprt *xprt,
8276 		void *calldata)
8277 {
8278 	struct rpc_bind_conn_calldata *p = calldata;
8279 
8280 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8281 }
8282 
8283 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8284 {
8285 	struct rpc_bind_conn_calldata data = {
8286 		.clp = clp,
8287 		.cred = cred,
8288 	};
8289 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8290 			nfs4_proc_bind_conn_to_session_callback, &data);
8291 }
8292 
8293 /*
8294  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8295  * and operations we'd like to see to enable certain features in the allow map
8296  */
8297 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8298 	.how = SP4_MACH_CRED,
8299 	.enforce.u.words = {
8300 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8301 		      1 << (OP_EXCHANGE_ID - 32) |
8302 		      1 << (OP_CREATE_SESSION - 32) |
8303 		      1 << (OP_DESTROY_SESSION - 32) |
8304 		      1 << (OP_DESTROY_CLIENTID - 32)
8305 	},
8306 	.allow.u.words = {
8307 		[0] = 1 << (OP_CLOSE) |
8308 		      1 << (OP_OPEN_DOWNGRADE) |
8309 		      1 << (OP_LOCKU) |
8310 		      1 << (OP_DELEGRETURN) |
8311 		      1 << (OP_COMMIT),
8312 		[1] = 1 << (OP_SECINFO - 32) |
8313 		      1 << (OP_SECINFO_NO_NAME - 32) |
8314 		      1 << (OP_LAYOUTRETURN - 32) |
8315 		      1 << (OP_TEST_STATEID - 32) |
8316 		      1 << (OP_FREE_STATEID - 32) |
8317 		      1 << (OP_WRITE - 32)
8318 	}
8319 };
8320 
8321 /*
8322  * Select the state protection mode for client `clp' given the server results
8323  * from exchange_id in `sp'.
8324  *
8325  * Returns 0 on success, negative errno otherwise.
8326  */
8327 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8328 				 struct nfs41_state_protection *sp)
8329 {
8330 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8331 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8332 		      1 << (OP_EXCHANGE_ID - 32) |
8333 		      1 << (OP_CREATE_SESSION - 32) |
8334 		      1 << (OP_DESTROY_SESSION - 32) |
8335 		      1 << (OP_DESTROY_CLIENTID - 32)
8336 	};
8337 	unsigned long flags = 0;
8338 	unsigned int i;
8339 	int ret = 0;
8340 
8341 	if (sp->how == SP4_MACH_CRED) {
8342 		/* Print state protect result */
8343 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8344 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8345 			if (test_bit(i, sp->enforce.u.longs))
8346 				dfprintk(MOUNT, "  enforce op %d\n", i);
8347 			if (test_bit(i, sp->allow.u.longs))
8348 				dfprintk(MOUNT, "  allow op %d\n", i);
8349 		}
8350 
8351 		/* make sure nothing is on enforce list that isn't supported */
8352 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8353 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8354 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8355 				ret = -EINVAL;
8356 				goto out;
8357 			}
8358 		}
8359 
8360 		/*
8361 		 * Minimal mode - state operations are allowed to use machine
8362 		 * credential.  Note this already happens by default, so the
8363 		 * client doesn't have to do anything more than the negotiation.
8364 		 *
8365 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8366 		 *       we're already using the machine cred for exchange_id
8367 		 *       and will never use a different cred.
8368 		 */
8369 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8370 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8371 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8372 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8373 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8374 			dfprintk(MOUNT, "  minimal mode enabled\n");
8375 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8376 		} else {
8377 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8378 			ret = -EINVAL;
8379 			goto out;
8380 		}
8381 
8382 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8383 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8384 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8385 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8386 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8387 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8388 		}
8389 
8390 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8391 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8392 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8393 		}
8394 
8395 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8396 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8397 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8398 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8399 		}
8400 
8401 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8402 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8403 			dfprintk(MOUNT, "  stateid mode enabled\n");
8404 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8405 		}
8406 
8407 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8408 			dfprintk(MOUNT, "  write mode enabled\n");
8409 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8410 		}
8411 
8412 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8413 			dfprintk(MOUNT, "  commit mode enabled\n");
8414 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8415 		}
8416 	}
8417 out:
8418 	clp->cl_sp4_flags = flags;
8419 	return ret;
8420 }
8421 
8422 struct nfs41_exchange_id_data {
8423 	struct nfs41_exchange_id_res res;
8424 	struct nfs41_exchange_id_args args;
8425 };
8426 
8427 static void nfs4_exchange_id_release(void *data)
8428 {
8429 	struct nfs41_exchange_id_data *cdata =
8430 					(struct nfs41_exchange_id_data *)data;
8431 
8432 	nfs_put_client(cdata->args.client);
8433 	kfree(cdata->res.impl_id);
8434 	kfree(cdata->res.server_scope);
8435 	kfree(cdata->res.server_owner);
8436 	kfree(cdata);
8437 }
8438 
8439 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8440 	.rpc_release = nfs4_exchange_id_release,
8441 };
8442 
8443 /*
8444  * _nfs4_proc_exchange_id()
8445  *
8446  * Wrapper for EXCHANGE_ID operation.
8447  */
8448 static struct rpc_task *
8449 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8450 			u32 sp4_how, struct rpc_xprt *xprt)
8451 {
8452 	struct rpc_message msg = {
8453 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8454 		.rpc_cred = cred,
8455 	};
8456 	struct rpc_task_setup task_setup_data = {
8457 		.rpc_client = clp->cl_rpcclient,
8458 		.callback_ops = &nfs4_exchange_id_call_ops,
8459 		.rpc_message = &msg,
8460 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8461 	};
8462 	struct nfs41_exchange_id_data *calldata;
8463 	int status;
8464 
8465 	if (!refcount_inc_not_zero(&clp->cl_count))
8466 		return ERR_PTR(-EIO);
8467 
8468 	status = -ENOMEM;
8469 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8470 	if (!calldata)
8471 		goto out;
8472 
8473 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8474 
8475 	status = nfs4_init_uniform_client_string(clp);
8476 	if (status)
8477 		goto out_calldata;
8478 
8479 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8480 						GFP_NOFS);
8481 	status = -ENOMEM;
8482 	if (unlikely(calldata->res.server_owner == NULL))
8483 		goto out_calldata;
8484 
8485 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8486 					GFP_NOFS);
8487 	if (unlikely(calldata->res.server_scope == NULL))
8488 		goto out_server_owner;
8489 
8490 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8491 	if (unlikely(calldata->res.impl_id == NULL))
8492 		goto out_server_scope;
8493 
8494 	switch (sp4_how) {
8495 	case SP4_NONE:
8496 		calldata->args.state_protect.how = SP4_NONE;
8497 		break;
8498 
8499 	case SP4_MACH_CRED:
8500 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8501 		break;
8502 
8503 	default:
8504 		/* unsupported! */
8505 		WARN_ON_ONCE(1);
8506 		status = -EINVAL;
8507 		goto out_impl_id;
8508 	}
8509 	if (xprt) {
8510 		task_setup_data.rpc_xprt = xprt;
8511 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8512 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8513 				sizeof(calldata->args.verifier.data));
8514 	}
8515 	calldata->args.client = clp;
8516 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8517 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8518 #ifdef CONFIG_NFS_V4_1_MIGRATION
8519 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8520 #endif
8521 	msg.rpc_argp = &calldata->args;
8522 	msg.rpc_resp = &calldata->res;
8523 	task_setup_data.callback_data = calldata;
8524 
8525 	return rpc_run_task(&task_setup_data);
8526 
8527 out_impl_id:
8528 	kfree(calldata->res.impl_id);
8529 out_server_scope:
8530 	kfree(calldata->res.server_scope);
8531 out_server_owner:
8532 	kfree(calldata->res.server_owner);
8533 out_calldata:
8534 	kfree(calldata);
8535 out:
8536 	nfs_put_client(clp);
8537 	return ERR_PTR(status);
8538 }
8539 
8540 /*
8541  * _nfs4_proc_exchange_id()
8542  *
8543  * Wrapper for EXCHANGE_ID operation.
8544  */
8545 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8546 			u32 sp4_how)
8547 {
8548 	struct rpc_task *task;
8549 	struct nfs41_exchange_id_args *argp;
8550 	struct nfs41_exchange_id_res *resp;
8551 	unsigned long now = jiffies;
8552 	int status;
8553 
8554 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8555 	if (IS_ERR(task))
8556 		return PTR_ERR(task);
8557 
8558 	argp = task->tk_msg.rpc_argp;
8559 	resp = task->tk_msg.rpc_resp;
8560 	status = task->tk_status;
8561 	if (status  != 0)
8562 		goto out;
8563 
8564 	status = nfs4_check_cl_exchange_flags(resp->flags,
8565 			clp->cl_mvops->minor_version);
8566 	if (status  != 0)
8567 		goto out;
8568 
8569 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8570 	if (status != 0)
8571 		goto out;
8572 
8573 	do_renew_lease(clp, now);
8574 
8575 	clp->cl_clientid = resp->clientid;
8576 	clp->cl_exchange_flags = resp->flags;
8577 	clp->cl_seqid = resp->seqid;
8578 	/* Client ID is not confirmed */
8579 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8580 		clear_bit(NFS4_SESSION_ESTABLISHED,
8581 			  &clp->cl_session->session_state);
8582 
8583 	if (clp->cl_serverscope != NULL &&
8584 	    !nfs41_same_server_scope(clp->cl_serverscope,
8585 				resp->server_scope)) {
8586 		dprintk("%s: server_scope mismatch detected\n",
8587 			__func__);
8588 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8589 	}
8590 
8591 	swap(clp->cl_serverowner, resp->server_owner);
8592 	swap(clp->cl_serverscope, resp->server_scope);
8593 	swap(clp->cl_implid, resp->impl_id);
8594 
8595 	/* Save the EXCHANGE_ID verifier session trunk tests */
8596 	memcpy(clp->cl_confirm.data, argp->verifier.data,
8597 	       sizeof(clp->cl_confirm.data));
8598 out:
8599 	trace_nfs4_exchange_id(clp, status);
8600 	rpc_put_task(task);
8601 	return status;
8602 }
8603 
8604 /*
8605  * nfs4_proc_exchange_id()
8606  *
8607  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8608  *
8609  * Since the clientid has expired, all compounds using sessions
8610  * associated with the stale clientid will be returning
8611  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8612  * be in some phase of session reset.
8613  *
8614  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8615  */
8616 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8617 {
8618 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8619 	int status;
8620 
8621 	/* try SP4_MACH_CRED if krb5i/p	*/
8622 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8623 	    authflavor == RPC_AUTH_GSS_KRB5P) {
8624 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8625 		if (!status)
8626 			return 0;
8627 	}
8628 
8629 	/* try SP4_NONE */
8630 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8631 }
8632 
8633 /**
8634  * nfs4_test_session_trunk
8635  *
8636  * This is an add_xprt_test() test function called from
8637  * rpc_clnt_setup_test_and_add_xprt.
8638  *
8639  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8640  * and is dereferrenced in nfs4_exchange_id_release
8641  *
8642  * Upon success, add the new transport to the rpc_clnt
8643  *
8644  * @clnt: struct rpc_clnt to get new transport
8645  * @xprt: the rpc_xprt to test
8646  * @data: call data for _nfs4_proc_exchange_id.
8647  */
8648 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8649 			    void *data)
8650 {
8651 	struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8652 	struct rpc_task *task;
8653 	int status;
8654 
8655 	u32 sp4_how;
8656 
8657 	dprintk("--> %s try %s\n", __func__,
8658 		xprt->address_strings[RPC_DISPLAY_ADDR]);
8659 
8660 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8661 
8662 	/* Test connection for session trunking. Async exchange_id call */
8663 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8664 	if (IS_ERR(task))
8665 		return;
8666 
8667 	status = task->tk_status;
8668 	if (status == 0)
8669 		status = nfs4_detect_session_trunking(adata->clp,
8670 				task->tk_msg.rpc_resp, xprt);
8671 
8672 	if (status == 0)
8673 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8674 
8675 	rpc_put_task(task);
8676 }
8677 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8678 
8679 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8680 		const struct cred *cred)
8681 {
8682 	struct rpc_message msg = {
8683 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8684 		.rpc_argp = clp,
8685 		.rpc_cred = cred,
8686 	};
8687 	int status;
8688 
8689 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
8690 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8691 	trace_nfs4_destroy_clientid(clp, status);
8692 	if (status)
8693 		dprintk("NFS: Got error %d from the server %s on "
8694 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
8695 	return status;
8696 }
8697 
8698 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8699 		const struct cred *cred)
8700 {
8701 	unsigned int loop;
8702 	int ret;
8703 
8704 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8705 		ret = _nfs4_proc_destroy_clientid(clp, cred);
8706 		switch (ret) {
8707 		case -NFS4ERR_DELAY:
8708 		case -NFS4ERR_CLIENTID_BUSY:
8709 			ssleep(1);
8710 			break;
8711 		default:
8712 			return ret;
8713 		}
8714 	}
8715 	return 0;
8716 }
8717 
8718 int nfs4_destroy_clientid(struct nfs_client *clp)
8719 {
8720 	const struct cred *cred;
8721 	int ret = 0;
8722 
8723 	if (clp->cl_mvops->minor_version < 1)
8724 		goto out;
8725 	if (clp->cl_exchange_flags == 0)
8726 		goto out;
8727 	if (clp->cl_preserve_clid)
8728 		goto out;
8729 	cred = nfs4_get_clid_cred(clp);
8730 	ret = nfs4_proc_destroy_clientid(clp, cred);
8731 	put_cred(cred);
8732 	switch (ret) {
8733 	case 0:
8734 	case -NFS4ERR_STALE_CLIENTID:
8735 		clp->cl_exchange_flags = 0;
8736 	}
8737 out:
8738 	return ret;
8739 }
8740 
8741 #endif /* CONFIG_NFS_V4_1 */
8742 
8743 struct nfs4_get_lease_time_data {
8744 	struct nfs4_get_lease_time_args *args;
8745 	struct nfs4_get_lease_time_res *res;
8746 	struct nfs_client *clp;
8747 };
8748 
8749 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8750 					void *calldata)
8751 {
8752 	struct nfs4_get_lease_time_data *data =
8753 			(struct nfs4_get_lease_time_data *)calldata;
8754 
8755 	dprintk("--> %s\n", __func__);
8756 	/* just setup sequence, do not trigger session recovery
8757 	   since we're invoked within one */
8758 	nfs4_setup_sequence(data->clp,
8759 			&data->args->la_seq_args,
8760 			&data->res->lr_seq_res,
8761 			task);
8762 	dprintk("<-- %s\n", __func__);
8763 }
8764 
8765 /*
8766  * Called from nfs4_state_manager thread for session setup, so don't recover
8767  * from sequence operation or clientid errors.
8768  */
8769 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8770 {
8771 	struct nfs4_get_lease_time_data *data =
8772 			(struct nfs4_get_lease_time_data *)calldata;
8773 
8774 	dprintk("--> %s\n", __func__);
8775 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8776 		return;
8777 	switch (task->tk_status) {
8778 	case -NFS4ERR_DELAY:
8779 	case -NFS4ERR_GRACE:
8780 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8781 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
8782 		task->tk_status = 0;
8783 		fallthrough;
8784 	case -NFS4ERR_RETRY_UNCACHED_REP:
8785 		rpc_restart_call_prepare(task);
8786 		return;
8787 	}
8788 	dprintk("<-- %s\n", __func__);
8789 }
8790 
8791 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8792 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
8793 	.rpc_call_done = nfs4_get_lease_time_done,
8794 };
8795 
8796 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8797 {
8798 	struct nfs4_get_lease_time_args args;
8799 	struct nfs4_get_lease_time_res res = {
8800 		.lr_fsinfo = fsinfo,
8801 	};
8802 	struct nfs4_get_lease_time_data data = {
8803 		.args = &args,
8804 		.res = &res,
8805 		.clp = clp,
8806 	};
8807 	struct rpc_message msg = {
8808 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8809 		.rpc_argp = &args,
8810 		.rpc_resp = &res,
8811 	};
8812 	struct rpc_task_setup task_setup = {
8813 		.rpc_client = clp->cl_rpcclient,
8814 		.rpc_message = &msg,
8815 		.callback_ops = &nfs4_get_lease_time_ops,
8816 		.callback_data = &data,
8817 		.flags = RPC_TASK_TIMEOUT,
8818 	};
8819 
8820 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8821 	return nfs4_call_sync_custom(&task_setup);
8822 }
8823 
8824 #ifdef CONFIG_NFS_V4_1
8825 
8826 /*
8827  * Initialize the values to be used by the client in CREATE_SESSION
8828  * If nfs4_init_session set the fore channel request and response sizes,
8829  * use them.
8830  *
8831  * Set the back channel max_resp_sz_cached to zero to force the client to
8832  * always set csa_cachethis to FALSE because the current implementation
8833  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8834  */
8835 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8836 				    struct rpc_clnt *clnt)
8837 {
8838 	unsigned int max_rqst_sz, max_resp_sz;
8839 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8840 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8841 
8842 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8843 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8844 
8845 	/* Fore channel attributes */
8846 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
8847 	args->fc_attrs.max_resp_sz = max_resp_sz;
8848 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
8849 	args->fc_attrs.max_reqs = max_session_slots;
8850 
8851 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8852 		"max_ops=%u max_reqs=%u\n",
8853 		__func__,
8854 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8855 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8856 
8857 	/* Back channel attributes */
8858 	args->bc_attrs.max_rqst_sz = max_bc_payload;
8859 	args->bc_attrs.max_resp_sz = max_bc_payload;
8860 	args->bc_attrs.max_resp_sz_cached = 0;
8861 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8862 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8863 	if (args->bc_attrs.max_reqs > max_bc_slots)
8864 		args->bc_attrs.max_reqs = max_bc_slots;
8865 
8866 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8867 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8868 		__func__,
8869 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8870 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8871 		args->bc_attrs.max_reqs);
8872 }
8873 
8874 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8875 		struct nfs41_create_session_res *res)
8876 {
8877 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
8878 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8879 
8880 	if (rcvd->max_resp_sz > sent->max_resp_sz)
8881 		return -EINVAL;
8882 	/*
8883 	 * Our requested max_ops is the minimum we need; we're not
8884 	 * prepared to break up compounds into smaller pieces than that.
8885 	 * So, no point even trying to continue if the server won't
8886 	 * cooperate:
8887 	 */
8888 	if (rcvd->max_ops < sent->max_ops)
8889 		return -EINVAL;
8890 	if (rcvd->max_reqs == 0)
8891 		return -EINVAL;
8892 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8893 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8894 	return 0;
8895 }
8896 
8897 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8898 		struct nfs41_create_session_res *res)
8899 {
8900 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
8901 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8902 
8903 	if (!(res->flags & SESSION4_BACK_CHAN))
8904 		goto out;
8905 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8906 		return -EINVAL;
8907 	if (rcvd->max_resp_sz < sent->max_resp_sz)
8908 		return -EINVAL;
8909 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8910 		return -EINVAL;
8911 	if (rcvd->max_ops > sent->max_ops)
8912 		return -EINVAL;
8913 	if (rcvd->max_reqs > sent->max_reqs)
8914 		return -EINVAL;
8915 out:
8916 	return 0;
8917 }
8918 
8919 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8920 				     struct nfs41_create_session_res *res)
8921 {
8922 	int ret;
8923 
8924 	ret = nfs4_verify_fore_channel_attrs(args, res);
8925 	if (ret)
8926 		return ret;
8927 	return nfs4_verify_back_channel_attrs(args, res);
8928 }
8929 
8930 static void nfs4_update_session(struct nfs4_session *session,
8931 		struct nfs41_create_session_res *res)
8932 {
8933 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8934 	/* Mark client id and session as being confirmed */
8935 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8936 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8937 	session->flags = res->flags;
8938 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8939 	if (res->flags & SESSION4_BACK_CHAN)
8940 		memcpy(&session->bc_attrs, &res->bc_attrs,
8941 				sizeof(session->bc_attrs));
8942 }
8943 
8944 static int _nfs4_proc_create_session(struct nfs_client *clp,
8945 		const struct cred *cred)
8946 {
8947 	struct nfs4_session *session = clp->cl_session;
8948 	struct nfs41_create_session_args args = {
8949 		.client = clp,
8950 		.clientid = clp->cl_clientid,
8951 		.seqid = clp->cl_seqid,
8952 		.cb_program = NFS4_CALLBACK,
8953 	};
8954 	struct nfs41_create_session_res res;
8955 
8956 	struct rpc_message msg = {
8957 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8958 		.rpc_argp = &args,
8959 		.rpc_resp = &res,
8960 		.rpc_cred = cred,
8961 	};
8962 	int status;
8963 
8964 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8965 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8966 
8967 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8968 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8969 	trace_nfs4_create_session(clp, status);
8970 
8971 	switch (status) {
8972 	case -NFS4ERR_STALE_CLIENTID:
8973 	case -NFS4ERR_DELAY:
8974 	case -ETIMEDOUT:
8975 	case -EACCES:
8976 	case -EAGAIN:
8977 		goto out;
8978 	}
8979 
8980 	clp->cl_seqid++;
8981 	if (!status) {
8982 		/* Verify the session's negotiated channel_attrs values */
8983 		status = nfs4_verify_channel_attrs(&args, &res);
8984 		/* Increment the clientid slot sequence id */
8985 		if (status)
8986 			goto out;
8987 		nfs4_update_session(session, &res);
8988 	}
8989 out:
8990 	return status;
8991 }
8992 
8993 /*
8994  * Issues a CREATE_SESSION operation to the server.
8995  * It is the responsibility of the caller to verify the session is
8996  * expired before calling this routine.
8997  */
8998 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8999 {
9000 	int status;
9001 	unsigned *ptr;
9002 	struct nfs4_session *session = clp->cl_session;
9003 
9004 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9005 
9006 	status = _nfs4_proc_create_session(clp, cred);
9007 	if (status)
9008 		goto out;
9009 
9010 	/* Init or reset the session slot tables */
9011 	status = nfs4_setup_session_slot_tables(session);
9012 	dprintk("slot table setup returned %d\n", status);
9013 	if (status)
9014 		goto out;
9015 
9016 	ptr = (unsigned *)&session->sess_id.data[0];
9017 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9018 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9019 out:
9020 	dprintk("<-- %s\n", __func__);
9021 	return status;
9022 }
9023 
9024 /*
9025  * Issue the over-the-wire RPC DESTROY_SESSION.
9026  * The caller must serialize access to this routine.
9027  */
9028 int nfs4_proc_destroy_session(struct nfs4_session *session,
9029 		const struct cred *cred)
9030 {
9031 	struct rpc_message msg = {
9032 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9033 		.rpc_argp = session,
9034 		.rpc_cred = cred,
9035 	};
9036 	int status = 0;
9037 
9038 	dprintk("--> nfs4_proc_destroy_session\n");
9039 
9040 	/* session is still being setup */
9041 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9042 		return 0;
9043 
9044 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9045 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9046 	trace_nfs4_destroy_session(session->clp, status);
9047 
9048 	if (status)
9049 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9050 			"Session has been destroyed regardless...\n", status);
9051 
9052 	dprintk("<-- nfs4_proc_destroy_session\n");
9053 	return status;
9054 }
9055 
9056 /*
9057  * Renew the cl_session lease.
9058  */
9059 struct nfs4_sequence_data {
9060 	struct nfs_client *clp;
9061 	struct nfs4_sequence_args args;
9062 	struct nfs4_sequence_res res;
9063 };
9064 
9065 static void nfs41_sequence_release(void *data)
9066 {
9067 	struct nfs4_sequence_data *calldata = data;
9068 	struct nfs_client *clp = calldata->clp;
9069 
9070 	if (refcount_read(&clp->cl_count) > 1)
9071 		nfs4_schedule_state_renewal(clp);
9072 	nfs_put_client(clp);
9073 	kfree(calldata);
9074 }
9075 
9076 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9077 {
9078 	switch(task->tk_status) {
9079 	case -NFS4ERR_DELAY:
9080 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9081 		return -EAGAIN;
9082 	default:
9083 		nfs4_schedule_lease_recovery(clp);
9084 	}
9085 	return 0;
9086 }
9087 
9088 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9089 {
9090 	struct nfs4_sequence_data *calldata = data;
9091 	struct nfs_client *clp = calldata->clp;
9092 
9093 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9094 		return;
9095 
9096 	trace_nfs4_sequence(clp, task->tk_status);
9097 	if (task->tk_status < 0) {
9098 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9099 		if (refcount_read(&clp->cl_count) == 1)
9100 			goto out;
9101 
9102 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9103 			rpc_restart_call_prepare(task);
9104 			return;
9105 		}
9106 	}
9107 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9108 out:
9109 	dprintk("<-- %s\n", __func__);
9110 }
9111 
9112 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9113 {
9114 	struct nfs4_sequence_data *calldata = data;
9115 	struct nfs_client *clp = calldata->clp;
9116 	struct nfs4_sequence_args *args;
9117 	struct nfs4_sequence_res *res;
9118 
9119 	args = task->tk_msg.rpc_argp;
9120 	res = task->tk_msg.rpc_resp;
9121 
9122 	nfs4_setup_sequence(clp, args, res, task);
9123 }
9124 
9125 static const struct rpc_call_ops nfs41_sequence_ops = {
9126 	.rpc_call_done = nfs41_sequence_call_done,
9127 	.rpc_call_prepare = nfs41_sequence_prepare,
9128 	.rpc_release = nfs41_sequence_release,
9129 };
9130 
9131 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9132 		const struct cred *cred,
9133 		struct nfs4_slot *slot,
9134 		bool is_privileged)
9135 {
9136 	struct nfs4_sequence_data *calldata;
9137 	struct rpc_message msg = {
9138 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9139 		.rpc_cred = cred,
9140 	};
9141 	struct rpc_task_setup task_setup_data = {
9142 		.rpc_client = clp->cl_rpcclient,
9143 		.rpc_message = &msg,
9144 		.callback_ops = &nfs41_sequence_ops,
9145 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
9146 	};
9147 	struct rpc_task *ret;
9148 
9149 	ret = ERR_PTR(-EIO);
9150 	if (!refcount_inc_not_zero(&clp->cl_count))
9151 		goto out_err;
9152 
9153 	ret = ERR_PTR(-ENOMEM);
9154 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9155 	if (calldata == NULL)
9156 		goto out_put_clp;
9157 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9158 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9159 	msg.rpc_argp = &calldata->args;
9160 	msg.rpc_resp = &calldata->res;
9161 	calldata->clp = clp;
9162 	task_setup_data.callback_data = calldata;
9163 
9164 	ret = rpc_run_task(&task_setup_data);
9165 	if (IS_ERR(ret))
9166 		goto out_err;
9167 	return ret;
9168 out_put_clp:
9169 	nfs_put_client(clp);
9170 out_err:
9171 	nfs41_release_slot(slot);
9172 	return ret;
9173 }
9174 
9175 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9176 {
9177 	struct rpc_task *task;
9178 	int ret = 0;
9179 
9180 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9181 		return -EAGAIN;
9182 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9183 	if (IS_ERR(task))
9184 		ret = PTR_ERR(task);
9185 	else
9186 		rpc_put_task_async(task);
9187 	dprintk("<-- %s status=%d\n", __func__, ret);
9188 	return ret;
9189 }
9190 
9191 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9192 {
9193 	struct rpc_task *task;
9194 	int ret;
9195 
9196 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9197 	if (IS_ERR(task)) {
9198 		ret = PTR_ERR(task);
9199 		goto out;
9200 	}
9201 	ret = rpc_wait_for_completion_task(task);
9202 	if (!ret)
9203 		ret = task->tk_status;
9204 	rpc_put_task(task);
9205 out:
9206 	dprintk("<-- %s status=%d\n", __func__, ret);
9207 	return ret;
9208 }
9209 
9210 struct nfs4_reclaim_complete_data {
9211 	struct nfs_client *clp;
9212 	struct nfs41_reclaim_complete_args arg;
9213 	struct nfs41_reclaim_complete_res res;
9214 };
9215 
9216 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9217 {
9218 	struct nfs4_reclaim_complete_data *calldata = data;
9219 
9220 	nfs4_setup_sequence(calldata->clp,
9221 			&calldata->arg.seq_args,
9222 			&calldata->res.seq_res,
9223 			task);
9224 }
9225 
9226 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9227 {
9228 	switch(task->tk_status) {
9229 	case 0:
9230 		wake_up_all(&clp->cl_lock_waitq);
9231 		fallthrough;
9232 	case -NFS4ERR_COMPLETE_ALREADY:
9233 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9234 		break;
9235 	case -NFS4ERR_DELAY:
9236 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9237 		fallthrough;
9238 	case -NFS4ERR_RETRY_UNCACHED_REP:
9239 		return -EAGAIN;
9240 	case -NFS4ERR_BADSESSION:
9241 	case -NFS4ERR_DEADSESSION:
9242 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9243 		break;
9244 	default:
9245 		nfs4_schedule_lease_recovery(clp);
9246 	}
9247 	return 0;
9248 }
9249 
9250 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9251 {
9252 	struct nfs4_reclaim_complete_data *calldata = data;
9253 	struct nfs_client *clp = calldata->clp;
9254 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9255 
9256 	dprintk("--> %s\n", __func__);
9257 	if (!nfs41_sequence_done(task, res))
9258 		return;
9259 
9260 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9261 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9262 		rpc_restart_call_prepare(task);
9263 		return;
9264 	}
9265 	dprintk("<-- %s\n", __func__);
9266 }
9267 
9268 static void nfs4_free_reclaim_complete_data(void *data)
9269 {
9270 	struct nfs4_reclaim_complete_data *calldata = data;
9271 
9272 	kfree(calldata);
9273 }
9274 
9275 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9276 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9277 	.rpc_call_done = nfs4_reclaim_complete_done,
9278 	.rpc_release = nfs4_free_reclaim_complete_data,
9279 };
9280 
9281 /*
9282  * Issue a global reclaim complete.
9283  */
9284 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9285 		const struct cred *cred)
9286 {
9287 	struct nfs4_reclaim_complete_data *calldata;
9288 	struct rpc_message msg = {
9289 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9290 		.rpc_cred = cred,
9291 	};
9292 	struct rpc_task_setup task_setup_data = {
9293 		.rpc_client = clp->cl_rpcclient,
9294 		.rpc_message = &msg,
9295 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9296 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9297 	};
9298 	int status = -ENOMEM;
9299 
9300 	dprintk("--> %s\n", __func__);
9301 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9302 	if (calldata == NULL)
9303 		goto out;
9304 	calldata->clp = clp;
9305 	calldata->arg.one_fs = 0;
9306 
9307 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9308 	msg.rpc_argp = &calldata->arg;
9309 	msg.rpc_resp = &calldata->res;
9310 	task_setup_data.callback_data = calldata;
9311 	status = nfs4_call_sync_custom(&task_setup_data);
9312 out:
9313 	dprintk("<-- %s status=%d\n", __func__, status);
9314 	return status;
9315 }
9316 
9317 static void
9318 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9319 {
9320 	struct nfs4_layoutget *lgp = calldata;
9321 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9322 
9323 	dprintk("--> %s\n", __func__);
9324 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9325 				&lgp->res.seq_res, task);
9326 	dprintk("<-- %s\n", __func__);
9327 }
9328 
9329 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9330 {
9331 	struct nfs4_layoutget *lgp = calldata;
9332 
9333 	dprintk("--> %s\n", __func__);
9334 	nfs41_sequence_process(task, &lgp->res.seq_res);
9335 	dprintk("<-- %s\n", __func__);
9336 }
9337 
9338 static int
9339 nfs4_layoutget_handle_exception(struct rpc_task *task,
9340 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9341 {
9342 	struct inode *inode = lgp->args.inode;
9343 	struct nfs_server *server = NFS_SERVER(inode);
9344 	struct pnfs_layout_hdr *lo;
9345 	int nfs4err = task->tk_status;
9346 	int err, status = 0;
9347 	LIST_HEAD(head);
9348 
9349 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9350 
9351 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9352 
9353 	switch (nfs4err) {
9354 	case 0:
9355 		goto out;
9356 
9357 	/*
9358 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9359 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9360 	 * retry go inband.
9361 	 */
9362 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9363 		status = -ENODATA;
9364 		goto out;
9365 	/*
9366 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9367 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9368 	 */
9369 	case -NFS4ERR_BADLAYOUT:
9370 		status = -EOVERFLOW;
9371 		goto out;
9372 	/*
9373 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9374 	 * (or clients) writing to the same RAID stripe except when
9375 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9376 	 *
9377 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9378 	 * while, and then eventually give up.
9379 	 */
9380 	case -NFS4ERR_LAYOUTTRYLATER:
9381 		if (lgp->args.minlength == 0) {
9382 			status = -EOVERFLOW;
9383 			goto out;
9384 		}
9385 		status = -EBUSY;
9386 		break;
9387 	case -NFS4ERR_RECALLCONFLICT:
9388 		status = -ERECALLCONFLICT;
9389 		break;
9390 	case -NFS4ERR_DELEG_REVOKED:
9391 	case -NFS4ERR_ADMIN_REVOKED:
9392 	case -NFS4ERR_EXPIRED:
9393 	case -NFS4ERR_BAD_STATEID:
9394 		exception->timeout = 0;
9395 		spin_lock(&inode->i_lock);
9396 		lo = NFS_I(inode)->layout;
9397 		/* If the open stateid was bad, then recover it. */
9398 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9399 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9400 			spin_unlock(&inode->i_lock);
9401 			exception->state = lgp->args.ctx->state;
9402 			exception->stateid = &lgp->args.stateid;
9403 			break;
9404 		}
9405 
9406 		/*
9407 		 * Mark the bad layout state as invalid, then retry
9408 		 */
9409 		pnfs_mark_layout_stateid_invalid(lo, &head);
9410 		spin_unlock(&inode->i_lock);
9411 		nfs_commit_inode(inode, 0);
9412 		pnfs_free_lseg_list(&head);
9413 		status = -EAGAIN;
9414 		goto out;
9415 	}
9416 
9417 	err = nfs4_handle_exception(server, nfs4err, exception);
9418 	if (!status) {
9419 		if (exception->retry)
9420 			status = -EAGAIN;
9421 		else
9422 			status = err;
9423 	}
9424 out:
9425 	dprintk("<-- %s\n", __func__);
9426 	return status;
9427 }
9428 
9429 size_t max_response_pages(struct nfs_server *server)
9430 {
9431 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9432 	return nfs_page_array_len(0, max_resp_sz);
9433 }
9434 
9435 static void nfs4_layoutget_release(void *calldata)
9436 {
9437 	struct nfs4_layoutget *lgp = calldata;
9438 
9439 	dprintk("--> %s\n", __func__);
9440 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9441 	pnfs_layoutget_free(lgp);
9442 	dprintk("<-- %s\n", __func__);
9443 }
9444 
9445 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9446 	.rpc_call_prepare = nfs4_layoutget_prepare,
9447 	.rpc_call_done = nfs4_layoutget_done,
9448 	.rpc_release = nfs4_layoutget_release,
9449 };
9450 
9451 struct pnfs_layout_segment *
9452 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9453 {
9454 	struct inode *inode = lgp->args.inode;
9455 	struct nfs_server *server = NFS_SERVER(inode);
9456 	struct rpc_task *task;
9457 	struct rpc_message msg = {
9458 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9459 		.rpc_argp = &lgp->args,
9460 		.rpc_resp = &lgp->res,
9461 		.rpc_cred = lgp->cred,
9462 	};
9463 	struct rpc_task_setup task_setup_data = {
9464 		.rpc_client = server->client,
9465 		.rpc_message = &msg,
9466 		.callback_ops = &nfs4_layoutget_call_ops,
9467 		.callback_data = lgp,
9468 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9469 	};
9470 	struct pnfs_layout_segment *lseg = NULL;
9471 	struct nfs4_exception exception = {
9472 		.inode = inode,
9473 		.timeout = *timeout,
9474 	};
9475 	int status = 0;
9476 
9477 	dprintk("--> %s\n", __func__);
9478 
9479 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9480 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
9481 
9482 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9483 
9484 	task = rpc_run_task(&task_setup_data);
9485 
9486 	status = rpc_wait_for_completion_task(task);
9487 	if (status != 0)
9488 		goto out;
9489 
9490 	if (task->tk_status < 0) {
9491 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9492 		*timeout = exception.timeout;
9493 	} else if (lgp->res.layoutp->len == 0) {
9494 		status = -EAGAIN;
9495 		*timeout = nfs4_update_delay(&exception.timeout);
9496 	} else
9497 		lseg = pnfs_layout_process(lgp);
9498 out:
9499 	trace_nfs4_layoutget(lgp->args.ctx,
9500 			&lgp->args.range,
9501 			&lgp->res.range,
9502 			&lgp->res.stateid,
9503 			status);
9504 
9505 	rpc_put_task(task);
9506 	dprintk("<-- %s status=%d\n", __func__, status);
9507 	if (status)
9508 		return ERR_PTR(status);
9509 	return lseg;
9510 }
9511 
9512 static void
9513 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9514 {
9515 	struct nfs4_layoutreturn *lrp = calldata;
9516 
9517 	dprintk("--> %s\n", __func__);
9518 	nfs4_setup_sequence(lrp->clp,
9519 			&lrp->args.seq_args,
9520 			&lrp->res.seq_res,
9521 			task);
9522 	if (!pnfs_layout_is_valid(lrp->args.layout))
9523 		rpc_exit(task, 0);
9524 }
9525 
9526 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9527 {
9528 	struct nfs4_layoutreturn *lrp = calldata;
9529 	struct nfs_server *server;
9530 
9531 	dprintk("--> %s\n", __func__);
9532 
9533 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9534 		return;
9535 
9536 	/*
9537 	 * Was there an RPC level error? Assume the call succeeded,
9538 	 * and that we need to release the layout
9539 	 */
9540 	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9541 		lrp->res.lrs_present = 0;
9542 		return;
9543 	}
9544 
9545 	server = NFS_SERVER(lrp->args.inode);
9546 	switch (task->tk_status) {
9547 	case -NFS4ERR_OLD_STATEID:
9548 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9549 					&lrp->args.range,
9550 					lrp->args.inode))
9551 			goto out_restart;
9552 		fallthrough;
9553 	default:
9554 		task->tk_status = 0;
9555 		fallthrough;
9556 	case 0:
9557 		break;
9558 	case -NFS4ERR_DELAY:
9559 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9560 			break;
9561 		goto out_restart;
9562 	}
9563 	dprintk("<-- %s\n", __func__);
9564 	return;
9565 out_restart:
9566 	task->tk_status = 0;
9567 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9568 	rpc_restart_call_prepare(task);
9569 }
9570 
9571 static void nfs4_layoutreturn_release(void *calldata)
9572 {
9573 	struct nfs4_layoutreturn *lrp = calldata;
9574 	struct pnfs_layout_hdr *lo = lrp->args.layout;
9575 
9576 	dprintk("--> %s\n", __func__);
9577 	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9578 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9579 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9580 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9581 		lrp->ld_private.ops->free(&lrp->ld_private);
9582 	pnfs_put_layout_hdr(lrp->args.layout);
9583 	nfs_iput_and_deactive(lrp->inode);
9584 	put_cred(lrp->cred);
9585 	kfree(calldata);
9586 	dprintk("<-- %s\n", __func__);
9587 }
9588 
9589 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9590 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9591 	.rpc_call_done = nfs4_layoutreturn_done,
9592 	.rpc_release = nfs4_layoutreturn_release,
9593 };
9594 
9595 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9596 {
9597 	struct rpc_task *task;
9598 	struct rpc_message msg = {
9599 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9600 		.rpc_argp = &lrp->args,
9601 		.rpc_resp = &lrp->res,
9602 		.rpc_cred = lrp->cred,
9603 	};
9604 	struct rpc_task_setup task_setup_data = {
9605 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9606 		.rpc_message = &msg,
9607 		.callback_ops = &nfs4_layoutreturn_call_ops,
9608 		.callback_data = lrp,
9609 	};
9610 	int status = 0;
9611 
9612 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9613 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9614 			&task_setup_data.rpc_client, &msg);
9615 
9616 	dprintk("--> %s\n", __func__);
9617 	if (!sync) {
9618 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9619 		if (!lrp->inode) {
9620 			nfs4_layoutreturn_release(lrp);
9621 			return -EAGAIN;
9622 		}
9623 		task_setup_data.flags |= RPC_TASK_ASYNC;
9624 	}
9625 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9626 	task = rpc_run_task(&task_setup_data);
9627 	if (IS_ERR(task))
9628 		return PTR_ERR(task);
9629 	if (sync)
9630 		status = task->tk_status;
9631 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9632 	dprintk("<-- %s status=%d\n", __func__, status);
9633 	rpc_put_task(task);
9634 	return status;
9635 }
9636 
9637 static int
9638 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9639 		struct pnfs_device *pdev,
9640 		const struct cred *cred)
9641 {
9642 	struct nfs4_getdeviceinfo_args args = {
9643 		.pdev = pdev,
9644 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9645 			NOTIFY_DEVICEID4_DELETE,
9646 	};
9647 	struct nfs4_getdeviceinfo_res res = {
9648 		.pdev = pdev,
9649 	};
9650 	struct rpc_message msg = {
9651 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9652 		.rpc_argp = &args,
9653 		.rpc_resp = &res,
9654 		.rpc_cred = cred,
9655 	};
9656 	int status;
9657 
9658 	dprintk("--> %s\n", __func__);
9659 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9660 	if (res.notification & ~args.notify_types)
9661 		dprintk("%s: unsupported notification\n", __func__);
9662 	if (res.notification != args.notify_types)
9663 		pdev->nocache = 1;
9664 
9665 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9666 
9667 	dprintk("<-- %s status=%d\n", __func__, status);
9668 
9669 	return status;
9670 }
9671 
9672 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9673 		struct pnfs_device *pdev,
9674 		const struct cred *cred)
9675 {
9676 	struct nfs4_exception exception = { };
9677 	int err;
9678 
9679 	do {
9680 		err = nfs4_handle_exception(server,
9681 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9682 					&exception);
9683 	} while (exception.retry);
9684 	return err;
9685 }
9686 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9687 
9688 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9689 {
9690 	struct nfs4_layoutcommit_data *data = calldata;
9691 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9692 
9693 	nfs4_setup_sequence(server->nfs_client,
9694 			&data->args.seq_args,
9695 			&data->res.seq_res,
9696 			task);
9697 }
9698 
9699 static void
9700 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9701 {
9702 	struct nfs4_layoutcommit_data *data = calldata;
9703 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9704 
9705 	if (!nfs41_sequence_done(task, &data->res.seq_res))
9706 		return;
9707 
9708 	switch (task->tk_status) { /* Just ignore these failures */
9709 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9710 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9711 	case -NFS4ERR_BADLAYOUT:     /* no layout */
9712 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
9713 		task->tk_status = 0;
9714 	case 0:
9715 		break;
9716 	default:
9717 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9718 			rpc_restart_call_prepare(task);
9719 			return;
9720 		}
9721 	}
9722 }
9723 
9724 static void nfs4_layoutcommit_release(void *calldata)
9725 {
9726 	struct nfs4_layoutcommit_data *data = calldata;
9727 
9728 	pnfs_cleanup_layoutcommit(data);
9729 	nfs_post_op_update_inode_force_wcc(data->args.inode,
9730 					   data->res.fattr);
9731 	put_cred(data->cred);
9732 	nfs_iput_and_deactive(data->inode);
9733 	kfree(data);
9734 }
9735 
9736 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9737 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
9738 	.rpc_call_done = nfs4_layoutcommit_done,
9739 	.rpc_release = nfs4_layoutcommit_release,
9740 };
9741 
9742 int
9743 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9744 {
9745 	struct rpc_message msg = {
9746 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9747 		.rpc_argp = &data->args,
9748 		.rpc_resp = &data->res,
9749 		.rpc_cred = data->cred,
9750 	};
9751 	struct rpc_task_setup task_setup_data = {
9752 		.task = &data->task,
9753 		.rpc_client = NFS_CLIENT(data->args.inode),
9754 		.rpc_message = &msg,
9755 		.callback_ops = &nfs4_layoutcommit_ops,
9756 		.callback_data = data,
9757 	};
9758 	struct rpc_task *task;
9759 	int status = 0;
9760 
9761 	dprintk("NFS: initiating layoutcommit call. sync %d "
9762 		"lbw: %llu inode %lu\n", sync,
9763 		data->args.lastbytewritten,
9764 		data->args.inode->i_ino);
9765 
9766 	if (!sync) {
9767 		data->inode = nfs_igrab_and_active(data->args.inode);
9768 		if (data->inode == NULL) {
9769 			nfs4_layoutcommit_release(data);
9770 			return -EAGAIN;
9771 		}
9772 		task_setup_data.flags = RPC_TASK_ASYNC;
9773 	}
9774 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9775 	task = rpc_run_task(&task_setup_data);
9776 	if (IS_ERR(task))
9777 		return PTR_ERR(task);
9778 	if (sync)
9779 		status = task->tk_status;
9780 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9781 	dprintk("%s: status %d\n", __func__, status);
9782 	rpc_put_task(task);
9783 	return status;
9784 }
9785 
9786 /*
9787  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9788  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9789  */
9790 static int
9791 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9792 		    struct nfs_fsinfo *info,
9793 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9794 {
9795 	struct nfs41_secinfo_no_name_args args = {
9796 		.style = SECINFO_STYLE_CURRENT_FH,
9797 	};
9798 	struct nfs4_secinfo_res res = {
9799 		.flavors = flavors,
9800 	};
9801 	struct rpc_message msg = {
9802 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9803 		.rpc_argp = &args,
9804 		.rpc_resp = &res,
9805 	};
9806 	struct nfs4_call_sync_data data = {
9807 		.seq_server = server,
9808 		.seq_args = &args.seq_args,
9809 		.seq_res = &res.seq_res,
9810 	};
9811 	struct rpc_task_setup task_setup = {
9812 		.rpc_client = server->client,
9813 		.rpc_message = &msg,
9814 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9815 		.callback_data = &data,
9816 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9817 	};
9818 	const struct cred *cred = NULL;
9819 	int status;
9820 
9821 	if (use_integrity) {
9822 		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9823 
9824 		cred = nfs4_get_clid_cred(server->nfs_client);
9825 		msg.rpc_cred = cred;
9826 	}
9827 
9828 	dprintk("--> %s\n", __func__);
9829 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9830 	status = nfs4_call_sync_custom(&task_setup);
9831 	dprintk("<-- %s status=%d\n", __func__, status);
9832 
9833 	put_cred(cred);
9834 
9835 	return status;
9836 }
9837 
9838 static int
9839 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9840 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9841 {
9842 	struct nfs4_exception exception = {
9843 		.interruptible = true,
9844 	};
9845 	int err;
9846 	do {
9847 		/* first try using integrity protection */
9848 		err = -NFS4ERR_WRONGSEC;
9849 
9850 		/* try to use integrity protection with machine cred */
9851 		if (_nfs4_is_integrity_protected(server->nfs_client))
9852 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9853 							  flavors, true);
9854 
9855 		/*
9856 		 * if unable to use integrity protection, or SECINFO with
9857 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
9858 		 * disallowed by spec, but exists in deployed servers) use
9859 		 * the current filesystem's rpc_client and the user cred.
9860 		 */
9861 		if (err == -NFS4ERR_WRONGSEC)
9862 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9863 							  flavors, false);
9864 
9865 		switch (err) {
9866 		case 0:
9867 		case -NFS4ERR_WRONGSEC:
9868 		case -ENOTSUPP:
9869 			goto out;
9870 		default:
9871 			err = nfs4_handle_exception(server, err, &exception);
9872 		}
9873 	} while (exception.retry);
9874 out:
9875 	return err;
9876 }
9877 
9878 static int
9879 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9880 		    struct nfs_fsinfo *info)
9881 {
9882 	int err;
9883 	struct page *page;
9884 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9885 	struct nfs4_secinfo_flavors *flavors;
9886 	struct nfs4_secinfo4 *secinfo;
9887 	int i;
9888 
9889 	page = alloc_page(GFP_KERNEL);
9890 	if (!page) {
9891 		err = -ENOMEM;
9892 		goto out;
9893 	}
9894 
9895 	flavors = page_address(page);
9896 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9897 
9898 	/*
9899 	 * Fall back on "guess and check" method if
9900 	 * the server doesn't support SECINFO_NO_NAME
9901 	 */
9902 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9903 		err = nfs4_find_root_sec(server, fhandle, info);
9904 		goto out_freepage;
9905 	}
9906 	if (err)
9907 		goto out_freepage;
9908 
9909 	for (i = 0; i < flavors->num_flavors; i++) {
9910 		secinfo = &flavors->flavors[i];
9911 
9912 		switch (secinfo->flavor) {
9913 		case RPC_AUTH_NULL:
9914 		case RPC_AUTH_UNIX:
9915 		case RPC_AUTH_GSS:
9916 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9917 					&secinfo->flavor_info);
9918 			break;
9919 		default:
9920 			flavor = RPC_AUTH_MAXFLAVOR;
9921 			break;
9922 		}
9923 
9924 		if (!nfs_auth_info_match(&server->auth_info, flavor))
9925 			flavor = RPC_AUTH_MAXFLAVOR;
9926 
9927 		if (flavor != RPC_AUTH_MAXFLAVOR) {
9928 			err = nfs4_lookup_root_sec(server, fhandle,
9929 						   info, flavor);
9930 			if (!err)
9931 				break;
9932 		}
9933 	}
9934 
9935 	if (flavor == RPC_AUTH_MAXFLAVOR)
9936 		err = -EPERM;
9937 
9938 out_freepage:
9939 	put_page(page);
9940 	if (err == -EACCES)
9941 		return -EPERM;
9942 out:
9943 	return err;
9944 }
9945 
9946 static int _nfs41_test_stateid(struct nfs_server *server,
9947 		nfs4_stateid *stateid,
9948 		const struct cred *cred)
9949 {
9950 	int status;
9951 	struct nfs41_test_stateid_args args = {
9952 		.stateid = stateid,
9953 	};
9954 	struct nfs41_test_stateid_res res;
9955 	struct rpc_message msg = {
9956 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9957 		.rpc_argp = &args,
9958 		.rpc_resp = &res,
9959 		.rpc_cred = cred,
9960 	};
9961 	struct rpc_clnt *rpc_client = server->client;
9962 
9963 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9964 		&rpc_client, &msg);
9965 
9966 	dprintk("NFS call  test_stateid %p\n", stateid);
9967 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9968 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9969 			&args.seq_args, &res.seq_res);
9970 	if (status != NFS_OK) {
9971 		dprintk("NFS reply test_stateid: failed, %d\n", status);
9972 		return status;
9973 	}
9974 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9975 	return -res.status;
9976 }
9977 
9978 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9979 		int err, struct nfs4_exception *exception)
9980 {
9981 	exception->retry = 0;
9982 	switch(err) {
9983 	case -NFS4ERR_DELAY:
9984 	case -NFS4ERR_RETRY_UNCACHED_REP:
9985 		nfs4_handle_exception(server, err, exception);
9986 		break;
9987 	case -NFS4ERR_BADSESSION:
9988 	case -NFS4ERR_BADSLOT:
9989 	case -NFS4ERR_BAD_HIGH_SLOT:
9990 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9991 	case -NFS4ERR_DEADSESSION:
9992 		nfs4_do_handle_exception(server, err, exception);
9993 	}
9994 }
9995 
9996 /**
9997  * nfs41_test_stateid - perform a TEST_STATEID operation
9998  *
9999  * @server: server / transport on which to perform the operation
10000  * @stateid: state ID to test
10001  * @cred: credential
10002  *
10003  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10004  * Otherwise a negative NFS4ERR value is returned if the operation
10005  * failed or the state ID is not currently valid.
10006  */
10007 static int nfs41_test_stateid(struct nfs_server *server,
10008 		nfs4_stateid *stateid,
10009 		const struct cred *cred)
10010 {
10011 	struct nfs4_exception exception = {
10012 		.interruptible = true,
10013 	};
10014 	int err;
10015 	do {
10016 		err = _nfs41_test_stateid(server, stateid, cred);
10017 		nfs4_handle_delay_or_session_error(server, err, &exception);
10018 	} while (exception.retry);
10019 	return err;
10020 }
10021 
10022 struct nfs_free_stateid_data {
10023 	struct nfs_server *server;
10024 	struct nfs41_free_stateid_args args;
10025 	struct nfs41_free_stateid_res res;
10026 };
10027 
10028 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10029 {
10030 	struct nfs_free_stateid_data *data = calldata;
10031 	nfs4_setup_sequence(data->server->nfs_client,
10032 			&data->args.seq_args,
10033 			&data->res.seq_res,
10034 			task);
10035 }
10036 
10037 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10038 {
10039 	struct nfs_free_stateid_data *data = calldata;
10040 
10041 	nfs41_sequence_done(task, &data->res.seq_res);
10042 
10043 	switch (task->tk_status) {
10044 	case -NFS4ERR_DELAY:
10045 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10046 			rpc_restart_call_prepare(task);
10047 	}
10048 }
10049 
10050 static void nfs41_free_stateid_release(void *calldata)
10051 {
10052 	kfree(calldata);
10053 }
10054 
10055 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10056 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10057 	.rpc_call_done = nfs41_free_stateid_done,
10058 	.rpc_release = nfs41_free_stateid_release,
10059 };
10060 
10061 /**
10062  * nfs41_free_stateid - perform a FREE_STATEID operation
10063  *
10064  * @server: server / transport on which to perform the operation
10065  * @stateid: state ID to release
10066  * @cred: credential
10067  * @privileged: set to true if this call needs to be privileged
10068  *
10069  * Note: this function is always asynchronous.
10070  */
10071 static int nfs41_free_stateid(struct nfs_server *server,
10072 		const nfs4_stateid *stateid,
10073 		const struct cred *cred,
10074 		bool privileged)
10075 {
10076 	struct rpc_message msg = {
10077 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10078 		.rpc_cred = cred,
10079 	};
10080 	struct rpc_task_setup task_setup = {
10081 		.rpc_client = server->client,
10082 		.rpc_message = &msg,
10083 		.callback_ops = &nfs41_free_stateid_ops,
10084 		.flags = RPC_TASK_ASYNC,
10085 	};
10086 	struct nfs_free_stateid_data *data;
10087 	struct rpc_task *task;
10088 
10089 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10090 		&task_setup.rpc_client, &msg);
10091 
10092 	dprintk("NFS call  free_stateid %p\n", stateid);
10093 	data = kmalloc(sizeof(*data), GFP_NOFS);
10094 	if (!data)
10095 		return -ENOMEM;
10096 	data->server = server;
10097 	nfs4_stateid_copy(&data->args.stateid, stateid);
10098 
10099 	task_setup.callback_data = data;
10100 
10101 	msg.rpc_argp = &data->args;
10102 	msg.rpc_resp = &data->res;
10103 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10104 	task = rpc_run_task(&task_setup);
10105 	if (IS_ERR(task))
10106 		return PTR_ERR(task);
10107 	rpc_put_task(task);
10108 	return 0;
10109 }
10110 
10111 static void
10112 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10113 {
10114 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10115 
10116 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10117 	nfs4_free_lock_state(server, lsp);
10118 }
10119 
10120 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10121 		const nfs4_stateid *s2)
10122 {
10123 	if (s1->type != s2->type)
10124 		return false;
10125 
10126 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10127 		return false;
10128 
10129 	if (s1->seqid == s2->seqid)
10130 		return true;
10131 
10132 	return s1->seqid == 0 || s2->seqid == 0;
10133 }
10134 
10135 #endif /* CONFIG_NFS_V4_1 */
10136 
10137 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10138 		const nfs4_stateid *s2)
10139 {
10140 	return nfs4_stateid_match(s1, s2);
10141 }
10142 
10143 
10144 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10145 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10146 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10147 	.recover_open	= nfs4_open_reclaim,
10148 	.recover_lock	= nfs4_lock_reclaim,
10149 	.establish_clid = nfs4_init_clientid,
10150 	.detect_trunking = nfs40_discover_server_trunking,
10151 };
10152 
10153 #if defined(CONFIG_NFS_V4_1)
10154 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10155 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10156 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10157 	.recover_open	= nfs4_open_reclaim,
10158 	.recover_lock	= nfs4_lock_reclaim,
10159 	.establish_clid = nfs41_init_clientid,
10160 	.reclaim_complete = nfs41_proc_reclaim_complete,
10161 	.detect_trunking = nfs41_discover_server_trunking,
10162 };
10163 #endif /* CONFIG_NFS_V4_1 */
10164 
10165 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10166 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10167 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10168 	.recover_open	= nfs40_open_expired,
10169 	.recover_lock	= nfs4_lock_expired,
10170 	.establish_clid = nfs4_init_clientid,
10171 };
10172 
10173 #if defined(CONFIG_NFS_V4_1)
10174 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10175 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10176 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10177 	.recover_open	= nfs41_open_expired,
10178 	.recover_lock	= nfs41_lock_expired,
10179 	.establish_clid = nfs41_init_clientid,
10180 };
10181 #endif /* CONFIG_NFS_V4_1 */
10182 
10183 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10184 	.sched_state_renewal = nfs4_proc_async_renew,
10185 	.get_state_renewal_cred = nfs4_get_renew_cred,
10186 	.renew_lease = nfs4_proc_renew,
10187 };
10188 
10189 #if defined(CONFIG_NFS_V4_1)
10190 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10191 	.sched_state_renewal = nfs41_proc_async_sequence,
10192 	.get_state_renewal_cred = nfs4_get_machine_cred,
10193 	.renew_lease = nfs4_proc_sequence,
10194 };
10195 #endif
10196 
10197 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10198 	.get_locations = _nfs40_proc_get_locations,
10199 	.fsid_present = _nfs40_proc_fsid_present,
10200 };
10201 
10202 #if defined(CONFIG_NFS_V4_1)
10203 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10204 	.get_locations = _nfs41_proc_get_locations,
10205 	.fsid_present = _nfs41_proc_fsid_present,
10206 };
10207 #endif	/* CONFIG_NFS_V4_1 */
10208 
10209 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10210 	.minor_version = 0,
10211 	.init_caps = NFS_CAP_READDIRPLUS
10212 		| NFS_CAP_ATOMIC_OPEN
10213 		| NFS_CAP_POSIX_LOCK,
10214 	.init_client = nfs40_init_client,
10215 	.shutdown_client = nfs40_shutdown_client,
10216 	.match_stateid = nfs4_match_stateid,
10217 	.find_root_sec = nfs4_find_root_sec,
10218 	.free_lock_state = nfs4_release_lockowner,
10219 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10220 	.alloc_seqid = nfs_alloc_seqid,
10221 	.call_sync_ops = &nfs40_call_sync_ops,
10222 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10223 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10224 	.state_renewal_ops = &nfs40_state_renewal_ops,
10225 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10226 };
10227 
10228 #if defined(CONFIG_NFS_V4_1)
10229 static struct nfs_seqid *
10230 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10231 {
10232 	return NULL;
10233 }
10234 
10235 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10236 	.minor_version = 1,
10237 	.init_caps = NFS_CAP_READDIRPLUS
10238 		| NFS_CAP_ATOMIC_OPEN
10239 		| NFS_CAP_POSIX_LOCK
10240 		| NFS_CAP_STATEID_NFSV41
10241 		| NFS_CAP_ATOMIC_OPEN_V1
10242 		| NFS_CAP_LGOPEN,
10243 	.init_client = nfs41_init_client,
10244 	.shutdown_client = nfs41_shutdown_client,
10245 	.match_stateid = nfs41_match_stateid,
10246 	.find_root_sec = nfs41_find_root_sec,
10247 	.free_lock_state = nfs41_free_lock_state,
10248 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10249 	.alloc_seqid = nfs_alloc_no_seqid,
10250 	.session_trunk = nfs4_test_session_trunk,
10251 	.call_sync_ops = &nfs41_call_sync_ops,
10252 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10253 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10254 	.state_renewal_ops = &nfs41_state_renewal_ops,
10255 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10256 };
10257 #endif
10258 
10259 #if defined(CONFIG_NFS_V4_2)
10260 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10261 	.minor_version = 2,
10262 	.init_caps = NFS_CAP_READDIRPLUS
10263 		| NFS_CAP_ATOMIC_OPEN
10264 		| NFS_CAP_POSIX_LOCK
10265 		| NFS_CAP_STATEID_NFSV41
10266 		| NFS_CAP_ATOMIC_OPEN_V1
10267 		| NFS_CAP_LGOPEN
10268 		| NFS_CAP_ALLOCATE
10269 		| NFS_CAP_COPY
10270 		| NFS_CAP_OFFLOAD_CANCEL
10271 		| NFS_CAP_COPY_NOTIFY
10272 		| NFS_CAP_DEALLOCATE
10273 		| NFS_CAP_SEEK
10274 		| NFS_CAP_LAYOUTSTATS
10275 		| NFS_CAP_CLONE
10276 		| NFS_CAP_LAYOUTERROR
10277 		| NFS_CAP_READ_PLUS,
10278 	.init_client = nfs41_init_client,
10279 	.shutdown_client = nfs41_shutdown_client,
10280 	.match_stateid = nfs41_match_stateid,
10281 	.find_root_sec = nfs41_find_root_sec,
10282 	.free_lock_state = nfs41_free_lock_state,
10283 	.call_sync_ops = &nfs41_call_sync_ops,
10284 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10285 	.alloc_seqid = nfs_alloc_no_seqid,
10286 	.session_trunk = nfs4_test_session_trunk,
10287 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10288 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10289 	.state_renewal_ops = &nfs41_state_renewal_ops,
10290 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10291 };
10292 #endif
10293 
10294 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10295 	[0] = &nfs_v4_0_minor_ops,
10296 #if defined(CONFIG_NFS_V4_1)
10297 	[1] = &nfs_v4_1_minor_ops,
10298 #endif
10299 #if defined(CONFIG_NFS_V4_2)
10300 	[2] = &nfs_v4_2_minor_ops,
10301 #endif
10302 };
10303 
10304 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10305 {
10306 	ssize_t error, error2, error3;
10307 
10308 	error = generic_listxattr(dentry, list, size);
10309 	if (error < 0)
10310 		return error;
10311 	if (list) {
10312 		list += error;
10313 		size -= error;
10314 	}
10315 
10316 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10317 	if (error2 < 0)
10318 		return error2;
10319 
10320 	if (list) {
10321 		list += error2;
10322 		size -= error2;
10323 	}
10324 
10325 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10326 	if (error3 < 0)
10327 		return error3;
10328 
10329 	return error + error2 + error3;
10330 }
10331 
10332 static const struct inode_operations nfs4_dir_inode_operations = {
10333 	.create		= nfs_create,
10334 	.lookup		= nfs_lookup,
10335 	.atomic_open	= nfs_atomic_open,
10336 	.link		= nfs_link,
10337 	.unlink		= nfs_unlink,
10338 	.symlink	= nfs_symlink,
10339 	.mkdir		= nfs_mkdir,
10340 	.rmdir		= nfs_rmdir,
10341 	.mknod		= nfs_mknod,
10342 	.rename		= nfs_rename,
10343 	.permission	= nfs_permission,
10344 	.getattr	= nfs_getattr,
10345 	.setattr	= nfs_setattr,
10346 	.listxattr	= nfs4_listxattr,
10347 };
10348 
10349 static const struct inode_operations nfs4_file_inode_operations = {
10350 	.permission	= nfs_permission,
10351 	.getattr	= nfs_getattr,
10352 	.setattr	= nfs_setattr,
10353 	.listxattr	= nfs4_listxattr,
10354 };
10355 
10356 const struct nfs_rpc_ops nfs_v4_clientops = {
10357 	.version	= 4,			/* protocol version */
10358 	.dentry_ops	= &nfs4_dentry_operations,
10359 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10360 	.file_inode_ops	= &nfs4_file_inode_operations,
10361 	.file_ops	= &nfs4_file_operations,
10362 	.getroot	= nfs4_proc_get_root,
10363 	.submount	= nfs4_submount,
10364 	.try_get_tree	= nfs4_try_get_tree,
10365 	.getattr	= nfs4_proc_getattr,
10366 	.setattr	= nfs4_proc_setattr,
10367 	.lookup		= nfs4_proc_lookup,
10368 	.lookupp	= nfs4_proc_lookupp,
10369 	.access		= nfs4_proc_access,
10370 	.readlink	= nfs4_proc_readlink,
10371 	.create		= nfs4_proc_create,
10372 	.remove		= nfs4_proc_remove,
10373 	.unlink_setup	= nfs4_proc_unlink_setup,
10374 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10375 	.unlink_done	= nfs4_proc_unlink_done,
10376 	.rename_setup	= nfs4_proc_rename_setup,
10377 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10378 	.rename_done	= nfs4_proc_rename_done,
10379 	.link		= nfs4_proc_link,
10380 	.symlink	= nfs4_proc_symlink,
10381 	.mkdir		= nfs4_proc_mkdir,
10382 	.rmdir		= nfs4_proc_rmdir,
10383 	.readdir	= nfs4_proc_readdir,
10384 	.mknod		= nfs4_proc_mknod,
10385 	.statfs		= nfs4_proc_statfs,
10386 	.fsinfo		= nfs4_proc_fsinfo,
10387 	.pathconf	= nfs4_proc_pathconf,
10388 	.set_capabilities = nfs4_server_capabilities,
10389 	.decode_dirent	= nfs4_decode_dirent,
10390 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10391 	.read_setup	= nfs4_proc_read_setup,
10392 	.read_done	= nfs4_read_done,
10393 	.write_setup	= nfs4_proc_write_setup,
10394 	.write_done	= nfs4_write_done,
10395 	.commit_setup	= nfs4_proc_commit_setup,
10396 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10397 	.commit_done	= nfs4_commit_done,
10398 	.lock		= nfs4_proc_lock,
10399 	.clear_acl_cache = nfs4_zap_acl_attr,
10400 	.close_context  = nfs4_close_context,
10401 	.open_context	= nfs4_atomic_open,
10402 	.have_delegation = nfs4_have_delegation,
10403 	.alloc_client	= nfs4_alloc_client,
10404 	.init_client	= nfs4_init_client,
10405 	.free_client	= nfs4_free_client,
10406 	.create_server	= nfs4_create_server,
10407 	.clone_server	= nfs_clone_server,
10408 };
10409 
10410 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10411 	.name	= XATTR_NAME_NFSV4_ACL,
10412 	.list	= nfs4_xattr_list_nfs4_acl,
10413 	.get	= nfs4_xattr_get_nfs4_acl,
10414 	.set	= nfs4_xattr_set_nfs4_acl,
10415 };
10416 
10417 #ifdef CONFIG_NFS_V4_2
10418 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10419 	.prefix	= XATTR_USER_PREFIX,
10420 	.get	= nfs4_xattr_get_nfs4_user,
10421 	.set	= nfs4_xattr_set_nfs4_user,
10422 };
10423 #endif
10424 
10425 const struct xattr_handler *nfs4_xattr_handlers[] = {
10426 	&nfs4_xattr_nfs4_acl_handler,
10427 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10428 	&nfs4_xattr_nfs4_label_handler,
10429 #endif
10430 #ifdef CONFIG_NFS_V4_2
10431 	&nfs4_xattr_nfs4_user_handler,
10432 #endif
10433 	NULL
10434 };
10435 
10436 /*
10437  * Local variables:
10438  *  c-basic-offset: 8
10439  * End:
10440  */
10441