xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision cbc4daca)
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 #define NFSDBG_FACILITY		NFSDBG_PROC
75 
76 #define NFS4_BITMASK_SZ		3
77 
78 #define NFS4_POLL_RETRY_MIN	(HZ/10)
79 #define NFS4_POLL_RETRY_MAX	(15*HZ)
80 
81 /* file attributes which can be mapped to nfs attributes */
82 #define NFS4_VALID_ATTRS (ATTR_MODE \
83 	| ATTR_UID \
84 	| ATTR_GID \
85 	| ATTR_SIZE \
86 	| ATTR_ATIME \
87 	| ATTR_MTIME \
88 	| ATTR_CTIME \
89 	| ATTR_ATIME_SET \
90 	| ATTR_MTIME_SET)
91 
92 struct nfs4_opendata;
93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
97 			      struct nfs_fattr *fattr, struct inode *inode);
98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
99 			    struct nfs_fattr *fattr, struct iattr *sattr,
100 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel);
101 #ifdef CONFIG_NFS_V4_1
102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
103 		const struct cred *cred,
104 		struct nfs4_slot *slot,
105 		bool is_privileged);
106 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
107 		const struct cred *);
108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
109 		const struct cred *, bool);
110 #endif
111 
112 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
113 static inline struct nfs4_label *
114 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
115 	struct iattr *sattr, struct nfs4_label *label)
116 {
117 	int err;
118 
119 	if (label == NULL)
120 		return NULL;
121 
122 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
123 		return NULL;
124 
125 	label->lfs = 0;
126 	label->pi = 0;
127 	label->len = 0;
128 	label->label = NULL;
129 
130 	err = security_dentry_init_security(dentry, sattr->ia_mode,
131 				&dentry->d_name, NULL,
132 				(void **)&label->label, &label->len);
133 	if (err == 0)
134 		return label;
135 
136 	return NULL;
137 }
138 static inline void
139 nfs4_label_release_security(struct nfs4_label *label)
140 {
141 	if (label)
142 		security_release_secctx(label->label, label->len);
143 }
144 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
145 {
146 	if (label)
147 		return server->attr_bitmask;
148 
149 	return server->attr_bitmask_nl;
150 }
151 #else
152 static inline struct nfs4_label *
153 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
154 	struct iattr *sattr, struct nfs4_label *l)
155 { return NULL; }
156 static inline void
157 nfs4_label_release_security(struct nfs4_label *label)
158 { return; }
159 static inline u32 *
160 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
161 { return server->attr_bitmask; }
162 #endif
163 
164 /* Prevent leaks of NFSv4 errors into userland */
165 static int nfs4_map_errors(int err)
166 {
167 	if (err >= -1000)
168 		return err;
169 	switch (err) {
170 	case -NFS4ERR_RESOURCE:
171 	case -NFS4ERR_LAYOUTTRYLATER:
172 	case -NFS4ERR_RECALLCONFLICT:
173 	case -NFS4ERR_RETURNCONFLICT:
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_CHANGE_ATTR_TYPE
271 			| FATTR4_WORD2_XATTR_SUPPORT
272 };
273 
274 const u32 nfs4_fs_locations_bitmap[3] = {
275 	FATTR4_WORD0_CHANGE
276 	| FATTR4_WORD0_SIZE
277 	| FATTR4_WORD0_FSID
278 	| FATTR4_WORD0_FILEID
279 	| FATTR4_WORD0_FS_LOCATIONS,
280 	FATTR4_WORD1_OWNER
281 	| FATTR4_WORD1_OWNER_GROUP
282 	| FATTR4_WORD1_RAWDEV
283 	| FATTR4_WORD1_SPACE_USED
284 	| FATTR4_WORD1_TIME_ACCESS
285 	| FATTR4_WORD1_TIME_METADATA
286 	| FATTR4_WORD1_TIME_MODIFY
287 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
288 };
289 
290 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
291 				    struct inode *inode, unsigned long flags)
292 {
293 	unsigned long cache_validity;
294 
295 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
296 	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
297 		return;
298 
299 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
300 
301 	/* Remove the attributes over which we have full control */
302 	dst[1] &= ~FATTR4_WORD1_RAWDEV;
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 	if (!(cache_validity & NFS_INO_INVALID_MODE))
310 		dst[1] &= ~FATTR4_WORD1_MODE;
311 	if (!(cache_validity & NFS_INO_INVALID_OTHER))
312 		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
313 }
314 
315 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
316 		struct nfs4_readdir_arg *readdir)
317 {
318 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
319 	__be32 *start, *p;
320 
321 	if (cookie > 2) {
322 		readdir->cookie = cookie;
323 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
324 		return;
325 	}
326 
327 	readdir->cookie = 0;
328 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
329 	if (cookie == 2)
330 		return;
331 
332 	/*
333 	 * NFSv4 servers do not return entries for '.' and '..'
334 	 * Therefore, we fake these entries here.  We let '.'
335 	 * have cookie 0 and '..' have cookie 1.  Note that
336 	 * when talking to the server, we always send cookie 0
337 	 * instead of 1 or 2.
338 	 */
339 	start = p = kmap_atomic(*readdir->pages);
340 
341 	if (cookie == 0) {
342 		*p++ = xdr_one;                                  /* next */
343 		*p++ = xdr_zero;                   /* cookie, first word */
344 		*p++ = xdr_one;                   /* cookie, second word */
345 		*p++ = xdr_one;                             /* entry len */
346 		memcpy(p, ".\0\0\0", 4);                        /* entry */
347 		p++;
348 		*p++ = xdr_one;                         /* bitmap length */
349 		*p++ = htonl(attrs);                           /* bitmap */
350 		*p++ = htonl(12);             /* attribute buffer length */
351 		*p++ = htonl(NF4DIR);
352 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
353 	}
354 
355 	*p++ = xdr_one;                                  /* next */
356 	*p++ = xdr_zero;                   /* cookie, first word */
357 	*p++ = xdr_two;                   /* cookie, second word */
358 	*p++ = xdr_two;                             /* entry len */
359 	memcpy(p, "..\0\0", 4);                         /* entry */
360 	p++;
361 	*p++ = xdr_one;                         /* bitmap length */
362 	*p++ = htonl(attrs);                           /* bitmap */
363 	*p++ = htonl(12);             /* attribute buffer length */
364 	*p++ = htonl(NF4DIR);
365 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
366 
367 	readdir->pgbase = (char *)p - (char *)start;
368 	readdir->count -= readdir->pgbase;
369 	kunmap_atomic(start);
370 }
371 
372 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
373 {
374 	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
375 		fattr->pre_change_attr = version;
376 		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
377 	}
378 }
379 
380 static void nfs4_test_and_free_stateid(struct nfs_server *server,
381 		nfs4_stateid *stateid,
382 		const struct cred *cred)
383 {
384 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
385 
386 	ops->test_and_free_expired(server, stateid, cred);
387 }
388 
389 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
390 		nfs4_stateid *stateid,
391 		const struct cred *cred)
392 {
393 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
394 	nfs4_test_and_free_stateid(server, stateid, cred);
395 }
396 
397 static void nfs4_free_revoked_stateid(struct nfs_server *server,
398 		const nfs4_stateid *stateid,
399 		const struct cred *cred)
400 {
401 	nfs4_stateid tmp;
402 
403 	nfs4_stateid_copy(&tmp, stateid);
404 	__nfs4_free_revoked_stateid(server, &tmp, cred);
405 }
406 
407 static long nfs4_update_delay(long *timeout)
408 {
409 	long ret;
410 	if (!timeout)
411 		return NFS4_POLL_RETRY_MAX;
412 	if (*timeout <= 0)
413 		*timeout = NFS4_POLL_RETRY_MIN;
414 	if (*timeout > NFS4_POLL_RETRY_MAX)
415 		*timeout = NFS4_POLL_RETRY_MAX;
416 	ret = *timeout;
417 	*timeout <<= 1;
418 	return ret;
419 }
420 
421 static int nfs4_delay_killable(long *timeout)
422 {
423 	might_sleep();
424 
425 	__set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
426 	schedule_timeout(nfs4_update_delay(timeout));
427 	if (!__fatal_signal_pending(current))
428 		return 0;
429 	return -EINTR;
430 }
431 
432 static int nfs4_delay_interruptible(long *timeout)
433 {
434 	might_sleep();
435 
436 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
437 	schedule_timeout(nfs4_update_delay(timeout));
438 	if (!signal_pending(current))
439 		return 0;
440 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
441 }
442 
443 static int nfs4_delay(long *timeout, bool interruptible)
444 {
445 	if (interruptible)
446 		return nfs4_delay_interruptible(timeout);
447 	return nfs4_delay_killable(timeout);
448 }
449 
450 static const nfs4_stateid *
451 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
452 {
453 	if (!stateid)
454 		return NULL;
455 	switch (stateid->type) {
456 	case NFS4_OPEN_STATEID_TYPE:
457 	case NFS4_LOCK_STATEID_TYPE:
458 	case NFS4_DELEGATION_STATEID_TYPE:
459 		return stateid;
460 	default:
461 		break;
462 	}
463 	return NULL;
464 }
465 
466 /* This is the error handling routine for processes that are allowed
467  * to sleep.
468  */
469 static int nfs4_do_handle_exception(struct nfs_server *server,
470 		int errorcode, struct nfs4_exception *exception)
471 {
472 	struct nfs_client *clp = server->nfs_client;
473 	struct nfs4_state *state = exception->state;
474 	const nfs4_stateid *stateid;
475 	struct inode *inode = exception->inode;
476 	int ret = errorcode;
477 
478 	exception->delay = 0;
479 	exception->recovering = 0;
480 	exception->retry = 0;
481 
482 	stateid = nfs4_recoverable_stateid(exception->stateid);
483 	if (stateid == NULL && state != NULL)
484 		stateid = nfs4_recoverable_stateid(&state->stateid);
485 
486 	switch(errorcode) {
487 		case 0:
488 			return 0;
489 		case -NFS4ERR_BADHANDLE:
490 		case -ESTALE:
491 			if (inode != NULL && S_ISREG(inode->i_mode))
492 				pnfs_destroy_layout(NFS_I(inode));
493 			break;
494 		case -NFS4ERR_DELEG_REVOKED:
495 		case -NFS4ERR_ADMIN_REVOKED:
496 		case -NFS4ERR_EXPIRED:
497 		case -NFS4ERR_BAD_STATEID:
498 		case -NFS4ERR_PARTNER_NO_AUTH:
499 			if (inode != NULL && stateid != NULL) {
500 				nfs_inode_find_state_and_recover(inode,
501 						stateid);
502 				goto wait_on_recovery;
503 			}
504 			fallthrough;
505 		case -NFS4ERR_OPENMODE:
506 			if (inode) {
507 				int err;
508 
509 				err = nfs_async_inode_return_delegation(inode,
510 						stateid);
511 				if (err == 0)
512 					goto wait_on_recovery;
513 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
514 					exception->retry = 1;
515 					break;
516 				}
517 			}
518 			if (state == NULL)
519 				break;
520 			ret = nfs4_schedule_stateid_recovery(server, state);
521 			if (ret < 0)
522 				break;
523 			goto wait_on_recovery;
524 		case -NFS4ERR_STALE_STATEID:
525 		case -NFS4ERR_STALE_CLIENTID:
526 			nfs4_schedule_lease_recovery(clp);
527 			goto wait_on_recovery;
528 		case -NFS4ERR_MOVED:
529 			ret = nfs4_schedule_migration_recovery(server);
530 			if (ret < 0)
531 				break;
532 			goto wait_on_recovery;
533 		case -NFS4ERR_LEASE_MOVED:
534 			nfs4_schedule_lease_moved_recovery(clp);
535 			goto wait_on_recovery;
536 #if defined(CONFIG_NFS_V4_1)
537 		case -NFS4ERR_BADSESSION:
538 		case -NFS4ERR_BADSLOT:
539 		case -NFS4ERR_BAD_HIGH_SLOT:
540 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
541 		case -NFS4ERR_DEADSESSION:
542 		case -NFS4ERR_SEQ_FALSE_RETRY:
543 		case -NFS4ERR_SEQ_MISORDERED:
544 			/* Handled in nfs41_sequence_process() */
545 			goto wait_on_recovery;
546 #endif /* defined(CONFIG_NFS_V4_1) */
547 		case -NFS4ERR_FILE_OPEN:
548 			if (exception->timeout > HZ) {
549 				/* We have retried a decent amount, time to
550 				 * fail
551 				 */
552 				ret = -EBUSY;
553 				break;
554 			}
555 			fallthrough;
556 		case -NFS4ERR_DELAY:
557 			nfs_inc_server_stats(server, NFSIOS_DELAY);
558 			fallthrough;
559 		case -NFS4ERR_GRACE:
560 		case -NFS4ERR_LAYOUTTRYLATER:
561 		case -NFS4ERR_RECALLCONFLICT:
562 		case -NFS4ERR_RETURNCONFLICT:
563 			exception->delay = 1;
564 			return 0;
565 
566 		case -NFS4ERR_RETRY_UNCACHED_REP:
567 		case -NFS4ERR_OLD_STATEID:
568 			exception->retry = 1;
569 			break;
570 		case -NFS4ERR_BADOWNER:
571 			/* The following works around a Linux server bug! */
572 		case -NFS4ERR_BADNAME:
573 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
574 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
575 				exception->retry = 1;
576 				printk(KERN_WARNING "NFS: v4 server %s "
577 						"does not accept raw "
578 						"uid/gids. "
579 						"Reenabling the idmapper.\n",
580 						server->nfs_client->cl_hostname);
581 			}
582 	}
583 	/* We failed to handle the error */
584 	return nfs4_map_errors(ret);
585 wait_on_recovery:
586 	exception->recovering = 1;
587 	return 0;
588 }
589 
590 /* This is the error handling routine for processes that are allowed
591  * to sleep.
592  */
593 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
594 {
595 	struct nfs_client *clp = server->nfs_client;
596 	int ret;
597 
598 	ret = nfs4_do_handle_exception(server, errorcode, exception);
599 	if (exception->delay) {
600 		ret = nfs4_delay(&exception->timeout,
601 				exception->interruptible);
602 		goto out_retry;
603 	}
604 	if (exception->recovering) {
605 		if (exception->task_is_privileged)
606 			return -EDEADLOCK;
607 		ret = nfs4_wait_clnt_recover(clp);
608 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
609 			return -EIO;
610 		goto out_retry;
611 	}
612 	return ret;
613 out_retry:
614 	if (ret == 0)
615 		exception->retry = 1;
616 	return ret;
617 }
618 
619 static int
620 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
621 		int errorcode, struct nfs4_exception *exception)
622 {
623 	struct nfs_client *clp = server->nfs_client;
624 	int ret;
625 
626 	ret = nfs4_do_handle_exception(server, errorcode, exception);
627 	if (exception->delay) {
628 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
629 		goto out_retry;
630 	}
631 	if (exception->recovering) {
632 		if (exception->task_is_privileged)
633 			return -EDEADLOCK;
634 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
635 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
636 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
637 		goto out_retry;
638 	}
639 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
640 		ret = -EIO;
641 	return ret;
642 out_retry:
643 	if (ret == 0) {
644 		exception->retry = 1;
645 		/*
646 		 * For NFS4ERR_MOVED, the client transport will need to
647 		 * be recomputed after migration recovery has completed.
648 		 */
649 		if (errorcode == -NFS4ERR_MOVED)
650 			rpc_task_release_transport(task);
651 	}
652 	return ret;
653 }
654 
655 int
656 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
657 			struct nfs4_state *state, long *timeout)
658 {
659 	struct nfs4_exception exception = {
660 		.state = state,
661 	};
662 
663 	if (task->tk_status >= 0)
664 		return 0;
665 	if (timeout)
666 		exception.timeout = *timeout;
667 	task->tk_status = nfs4_async_handle_exception(task, server,
668 			task->tk_status,
669 			&exception);
670 	if (exception.delay && timeout)
671 		*timeout = exception.timeout;
672 	if (exception.retry)
673 		return -EAGAIN;
674 	return 0;
675 }
676 
677 /*
678  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
679  * or 'false' otherwise.
680  */
681 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
682 {
683 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
684 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
685 }
686 
687 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
688 {
689 	spin_lock(&clp->cl_lock);
690 	if (time_before(clp->cl_last_renewal,timestamp))
691 		clp->cl_last_renewal = timestamp;
692 	spin_unlock(&clp->cl_lock);
693 }
694 
695 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
696 {
697 	struct nfs_client *clp = server->nfs_client;
698 
699 	if (!nfs4_has_session(clp))
700 		do_renew_lease(clp, timestamp);
701 }
702 
703 struct nfs4_call_sync_data {
704 	const struct nfs_server *seq_server;
705 	struct nfs4_sequence_args *seq_args;
706 	struct nfs4_sequence_res *seq_res;
707 };
708 
709 void nfs4_init_sequence(struct nfs4_sequence_args *args,
710 			struct nfs4_sequence_res *res, int cache_reply,
711 			int privileged)
712 {
713 	args->sa_slot = NULL;
714 	args->sa_cache_this = cache_reply;
715 	args->sa_privileged = privileged;
716 
717 	res->sr_slot = NULL;
718 }
719 
720 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
721 {
722 	struct nfs4_slot *slot = res->sr_slot;
723 	struct nfs4_slot_table *tbl;
724 
725 	tbl = slot->table;
726 	spin_lock(&tbl->slot_tbl_lock);
727 	if (!nfs41_wake_and_assign_slot(tbl, slot))
728 		nfs4_free_slot(tbl, slot);
729 	spin_unlock(&tbl->slot_tbl_lock);
730 
731 	res->sr_slot = NULL;
732 }
733 
734 static int nfs40_sequence_done(struct rpc_task *task,
735 			       struct nfs4_sequence_res *res)
736 {
737 	if (res->sr_slot != NULL)
738 		nfs40_sequence_free_slot(res);
739 	return 1;
740 }
741 
742 #if defined(CONFIG_NFS_V4_1)
743 
744 static void nfs41_release_slot(struct nfs4_slot *slot)
745 {
746 	struct nfs4_session *session;
747 	struct nfs4_slot_table *tbl;
748 	bool send_new_highest_used_slotid = false;
749 
750 	if (!slot)
751 		return;
752 	tbl = slot->table;
753 	session = tbl->session;
754 
755 	/* Bump the slot sequence number */
756 	if (slot->seq_done)
757 		slot->seq_nr++;
758 	slot->seq_done = 0;
759 
760 	spin_lock(&tbl->slot_tbl_lock);
761 	/* Be nice to the server: try to ensure that the last transmitted
762 	 * value for highest_user_slotid <= target_highest_slotid
763 	 */
764 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
765 		send_new_highest_used_slotid = true;
766 
767 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
768 		send_new_highest_used_slotid = false;
769 		goto out_unlock;
770 	}
771 	nfs4_free_slot(tbl, slot);
772 
773 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
774 		send_new_highest_used_slotid = false;
775 out_unlock:
776 	spin_unlock(&tbl->slot_tbl_lock);
777 	if (send_new_highest_used_slotid)
778 		nfs41_notify_server(session->clp);
779 	if (waitqueue_active(&tbl->slot_waitq))
780 		wake_up_all(&tbl->slot_waitq);
781 }
782 
783 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
784 {
785 	nfs41_release_slot(res->sr_slot);
786 	res->sr_slot = NULL;
787 }
788 
789 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
790 		u32 seqnr)
791 {
792 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
793 		slot->seq_nr_highest_sent = seqnr;
794 }
795 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
796 {
797 	nfs4_slot_sequence_record_sent(slot, seqnr);
798 	slot->seq_nr_last_acked = seqnr;
799 }
800 
801 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
802 				struct nfs4_slot *slot)
803 {
804 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
805 	if (!IS_ERR(task))
806 		rpc_put_task_async(task);
807 }
808 
809 static int nfs41_sequence_process(struct rpc_task *task,
810 		struct nfs4_sequence_res *res)
811 {
812 	struct nfs4_session *session;
813 	struct nfs4_slot *slot = res->sr_slot;
814 	struct nfs_client *clp;
815 	int status;
816 	int ret = 1;
817 
818 	if (slot == NULL)
819 		goto out_noaction;
820 	/* don't increment the sequence number if the task wasn't sent */
821 	if (!RPC_WAS_SENT(task) || slot->seq_done)
822 		goto out;
823 
824 	session = slot->table->session;
825 	clp = session->clp;
826 
827 	trace_nfs4_sequence_done(session, res);
828 
829 	status = res->sr_status;
830 	if (task->tk_status == -NFS4ERR_DEADSESSION)
831 		status = -NFS4ERR_DEADSESSION;
832 
833 	/* Check the SEQUENCE operation status */
834 	switch (status) {
835 	case 0:
836 		/* Mark this sequence number as having been acked */
837 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
838 		/* Update the slot's sequence and clientid lease timer */
839 		slot->seq_done = 1;
840 		do_renew_lease(clp, res->sr_timestamp);
841 		/* Check sequence flags */
842 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
843 				!!slot->privileged);
844 		nfs41_update_target_slotid(slot->table, slot, res);
845 		break;
846 	case 1:
847 		/*
848 		 * sr_status remains 1 if an RPC level error occurred.
849 		 * The server may or may not have processed the sequence
850 		 * operation..
851 		 */
852 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
853 		slot->seq_done = 1;
854 		goto out;
855 	case -NFS4ERR_DELAY:
856 		/* The server detected a resend of the RPC call and
857 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
858 		 * of RFC5661.
859 		 */
860 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
861 			__func__,
862 			slot->slot_nr,
863 			slot->seq_nr);
864 		goto out_retry;
865 	case -NFS4ERR_RETRY_UNCACHED_REP:
866 	case -NFS4ERR_SEQ_FALSE_RETRY:
867 		/*
868 		 * The server thinks we tried to replay a request.
869 		 * Retry the call after bumping the sequence ID.
870 		 */
871 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
872 		goto retry_new_seq;
873 	case -NFS4ERR_BADSLOT:
874 		/*
875 		 * The slot id we used was probably retired. Try again
876 		 * using a different slot id.
877 		 */
878 		if (slot->slot_nr < slot->table->target_highest_slotid)
879 			goto session_recover;
880 		goto retry_nowait;
881 	case -NFS4ERR_SEQ_MISORDERED:
882 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
883 		/*
884 		 * Were one or more calls using this slot interrupted?
885 		 * If the server never received the request, then our
886 		 * transmitted slot sequence number may be too high. However,
887 		 * if the server did receive the request then it might
888 		 * accidentally give us a reply with a mismatched operation.
889 		 * We can sort this out by sending a lone sequence operation
890 		 * to the server on the same slot.
891 		 */
892 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
893 			slot->seq_nr--;
894 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
895 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
896 				res->sr_slot = NULL;
897 			}
898 			goto retry_nowait;
899 		}
900 		/*
901 		 * RFC5661:
902 		 * A retry might be sent while the original request is
903 		 * still in progress on the replier. The replier SHOULD
904 		 * deal with the issue by returning NFS4ERR_DELAY as the
905 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
906 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
907 		 *
908 		 * Restart the search after a delay.
909 		 */
910 		slot->seq_nr = slot->seq_nr_highest_sent;
911 		goto out_retry;
912 	case -NFS4ERR_BADSESSION:
913 	case -NFS4ERR_DEADSESSION:
914 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
915 		goto session_recover;
916 	default:
917 		/* Just update the slot sequence no. */
918 		slot->seq_done = 1;
919 	}
920 out:
921 	/* The session may be reset by one of the error handlers. */
922 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
923 out_noaction:
924 	return ret;
925 session_recover:
926 	set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
927 	nfs4_schedule_session_recovery(session, status);
928 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
929 	nfs41_sequence_free_slot(res);
930 	goto out;
931 retry_new_seq:
932 	++slot->seq_nr;
933 retry_nowait:
934 	if (rpc_restart_call_prepare(task)) {
935 		nfs41_sequence_free_slot(res);
936 		task->tk_status = 0;
937 		ret = 0;
938 	}
939 	goto out;
940 out_retry:
941 	if (!rpc_restart_call(task))
942 		goto out;
943 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
944 	return 0;
945 }
946 
947 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
948 {
949 	if (!nfs41_sequence_process(task, res))
950 		return 0;
951 	if (res->sr_slot != NULL)
952 		nfs41_sequence_free_slot(res);
953 	return 1;
954 
955 }
956 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
957 
958 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
959 {
960 	if (res->sr_slot == NULL)
961 		return 1;
962 	if (res->sr_slot->table->session != NULL)
963 		return nfs41_sequence_process(task, res);
964 	return nfs40_sequence_done(task, res);
965 }
966 
967 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
968 {
969 	if (res->sr_slot != NULL) {
970 		if (res->sr_slot->table->session != NULL)
971 			nfs41_sequence_free_slot(res);
972 		else
973 			nfs40_sequence_free_slot(res);
974 	}
975 }
976 
977 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
978 {
979 	if (res->sr_slot == NULL)
980 		return 1;
981 	if (!res->sr_slot->table->session)
982 		return nfs40_sequence_done(task, res);
983 	return nfs41_sequence_done(task, res);
984 }
985 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
986 
987 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
988 {
989 	struct nfs4_call_sync_data *data = calldata;
990 
991 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
992 
993 	nfs4_setup_sequence(data->seq_server->nfs_client,
994 			    data->seq_args, data->seq_res, task);
995 }
996 
997 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
998 {
999 	struct nfs4_call_sync_data *data = calldata;
1000 
1001 	nfs41_sequence_done(task, data->seq_res);
1002 }
1003 
1004 static const struct rpc_call_ops nfs41_call_sync_ops = {
1005 	.rpc_call_prepare = nfs41_call_sync_prepare,
1006 	.rpc_call_done = nfs41_call_sync_done,
1007 };
1008 
1009 #else	/* !CONFIG_NFS_V4_1 */
1010 
1011 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1012 {
1013 	return nfs40_sequence_done(task, res);
1014 }
1015 
1016 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1017 {
1018 	if (res->sr_slot != NULL)
1019 		nfs40_sequence_free_slot(res);
1020 }
1021 
1022 int nfs4_sequence_done(struct rpc_task *task,
1023 		       struct nfs4_sequence_res *res)
1024 {
1025 	return nfs40_sequence_done(task, res);
1026 }
1027 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1028 
1029 #endif	/* !CONFIG_NFS_V4_1 */
1030 
1031 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1032 {
1033 	res->sr_timestamp = jiffies;
1034 	res->sr_status_flags = 0;
1035 	res->sr_status = 1;
1036 }
1037 
1038 static
1039 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1040 		struct nfs4_sequence_res *res,
1041 		struct nfs4_slot *slot)
1042 {
1043 	if (!slot)
1044 		return;
1045 	slot->privileged = args->sa_privileged ? 1 : 0;
1046 	args->sa_slot = slot;
1047 
1048 	res->sr_slot = slot;
1049 }
1050 
1051 int nfs4_setup_sequence(struct nfs_client *client,
1052 			struct nfs4_sequence_args *args,
1053 			struct nfs4_sequence_res *res,
1054 			struct rpc_task *task)
1055 {
1056 	struct nfs4_session *session = nfs4_get_session(client);
1057 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1058 	struct nfs4_slot *slot;
1059 
1060 	/* slot already allocated? */
1061 	if (res->sr_slot != NULL)
1062 		goto out_start;
1063 
1064 	if (session)
1065 		tbl = &session->fc_slot_table;
1066 
1067 	spin_lock(&tbl->slot_tbl_lock);
1068 	/* The state manager will wait until the slot table is empty */
1069 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1070 		goto out_sleep;
1071 
1072 	slot = nfs4_alloc_slot(tbl);
1073 	if (IS_ERR(slot)) {
1074 		if (slot == ERR_PTR(-ENOMEM))
1075 			goto out_sleep_timeout;
1076 		goto out_sleep;
1077 	}
1078 	spin_unlock(&tbl->slot_tbl_lock);
1079 
1080 	nfs4_sequence_attach_slot(args, res, slot);
1081 
1082 	trace_nfs4_setup_sequence(session, args);
1083 out_start:
1084 	nfs41_sequence_res_init(res);
1085 	rpc_call_start(task);
1086 	return 0;
1087 out_sleep_timeout:
1088 	/* Try again in 1/4 second */
1089 	if (args->sa_privileged)
1090 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1091 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1092 	else
1093 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1094 				NULL, jiffies + (HZ >> 2));
1095 	spin_unlock(&tbl->slot_tbl_lock);
1096 	return -EAGAIN;
1097 out_sleep:
1098 	if (args->sa_privileged)
1099 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1100 				RPC_PRIORITY_PRIVILEGED);
1101 	else
1102 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1103 	spin_unlock(&tbl->slot_tbl_lock);
1104 	return -EAGAIN;
1105 }
1106 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1107 
1108 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1109 {
1110 	struct nfs4_call_sync_data *data = calldata;
1111 	nfs4_setup_sequence(data->seq_server->nfs_client,
1112 				data->seq_args, data->seq_res, task);
1113 }
1114 
1115 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1116 {
1117 	struct nfs4_call_sync_data *data = calldata;
1118 	nfs4_sequence_done(task, data->seq_res);
1119 }
1120 
1121 static const struct rpc_call_ops nfs40_call_sync_ops = {
1122 	.rpc_call_prepare = nfs40_call_sync_prepare,
1123 	.rpc_call_done = nfs40_call_sync_done,
1124 };
1125 
1126 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1127 {
1128 	int ret;
1129 	struct rpc_task *task;
1130 
1131 	task = rpc_run_task(task_setup);
1132 	if (IS_ERR(task))
1133 		return PTR_ERR(task);
1134 
1135 	ret = task->tk_status;
1136 	rpc_put_task(task);
1137 	return ret;
1138 }
1139 
1140 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1141 			     struct nfs_server *server,
1142 			     struct rpc_message *msg,
1143 			     struct nfs4_sequence_args *args,
1144 			     struct nfs4_sequence_res *res,
1145 			     unsigned short task_flags)
1146 {
1147 	struct nfs_client *clp = server->nfs_client;
1148 	struct nfs4_call_sync_data data = {
1149 		.seq_server = server,
1150 		.seq_args = args,
1151 		.seq_res = res,
1152 	};
1153 	struct rpc_task_setup task_setup = {
1154 		.rpc_client = clnt,
1155 		.rpc_message = msg,
1156 		.callback_ops = clp->cl_mvops->call_sync_ops,
1157 		.callback_data = &data,
1158 		.flags = task_flags,
1159 	};
1160 
1161 	return nfs4_call_sync_custom(&task_setup);
1162 }
1163 
1164 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1165 				   struct nfs_server *server,
1166 				   struct rpc_message *msg,
1167 				   struct nfs4_sequence_args *args,
1168 				   struct nfs4_sequence_res *res)
1169 {
1170 	unsigned short task_flags = 0;
1171 
1172 	if (server->caps & NFS_CAP_MOVEABLE)
1173 		task_flags = RPC_TASK_MOVEABLE;
1174 	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1175 }
1176 
1177 
1178 int nfs4_call_sync(struct rpc_clnt *clnt,
1179 		   struct nfs_server *server,
1180 		   struct rpc_message *msg,
1181 		   struct nfs4_sequence_args *args,
1182 		   struct nfs4_sequence_res *res,
1183 		   int cache_reply)
1184 {
1185 	nfs4_init_sequence(args, res, cache_reply, 0);
1186 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1187 }
1188 
1189 static void
1190 nfs4_inc_nlink_locked(struct inode *inode)
1191 {
1192 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1193 					     NFS_INO_INVALID_CTIME |
1194 					     NFS_INO_INVALID_NLINK);
1195 	inc_nlink(inode);
1196 }
1197 
1198 static void
1199 nfs4_inc_nlink(struct inode *inode)
1200 {
1201 	spin_lock(&inode->i_lock);
1202 	nfs4_inc_nlink_locked(inode);
1203 	spin_unlock(&inode->i_lock);
1204 }
1205 
1206 static void
1207 nfs4_dec_nlink_locked(struct inode *inode)
1208 {
1209 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1210 					     NFS_INO_INVALID_CTIME |
1211 					     NFS_INO_INVALID_NLINK);
1212 	drop_nlink(inode);
1213 }
1214 
1215 static void
1216 nfs4_update_changeattr_locked(struct inode *inode,
1217 		struct nfs4_change_info *cinfo,
1218 		unsigned long timestamp, unsigned long cache_validity)
1219 {
1220 	struct nfs_inode *nfsi = NFS_I(inode);
1221 	u64 change_attr = inode_peek_iversion_raw(inode);
1222 
1223 	cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1224 	if (S_ISDIR(inode->i_mode))
1225 		cache_validity |= NFS_INO_INVALID_DATA;
1226 
1227 	switch (NFS_SERVER(inode)->change_attr_type) {
1228 	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1229 		if (cinfo->after == change_attr)
1230 			goto out;
1231 		break;
1232 	default:
1233 		if ((s64)(change_attr - cinfo->after) >= 0)
1234 			goto out;
1235 	}
1236 
1237 	inode_set_iversion_raw(inode, cinfo->after);
1238 	if (!cinfo->atomic || cinfo->before != change_attr) {
1239 		if (S_ISDIR(inode->i_mode))
1240 			nfs_force_lookup_revalidate(inode);
1241 
1242 		if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1243 			cache_validity |=
1244 				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1245 				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1246 				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1247 				NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1248 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1249 	}
1250 	nfsi->attrtimeo_timestamp = jiffies;
1251 	nfsi->read_cache_jiffies = timestamp;
1252 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1253 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1254 out:
1255 	nfs_set_cache_invalid(inode, cache_validity);
1256 }
1257 
1258 void
1259 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1260 		unsigned long timestamp, unsigned long cache_validity)
1261 {
1262 	spin_lock(&dir->i_lock);
1263 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1264 	spin_unlock(&dir->i_lock);
1265 }
1266 
1267 struct nfs4_open_createattrs {
1268 	struct nfs4_label *label;
1269 	struct iattr *sattr;
1270 	const __u32 verf[2];
1271 };
1272 
1273 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1274 		int err, struct nfs4_exception *exception)
1275 {
1276 	if (err != -EINVAL)
1277 		return false;
1278 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1279 		return false;
1280 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1281 	exception->retry = 1;
1282 	return true;
1283 }
1284 
1285 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1286 {
1287 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1288 }
1289 
1290 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1291 {
1292 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1293 
1294 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1295 }
1296 
1297 static u32
1298 nfs4_map_atomic_open_share(struct nfs_server *server,
1299 		fmode_t fmode, int openflags)
1300 {
1301 	u32 res = 0;
1302 
1303 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1304 	case FMODE_READ:
1305 		res = NFS4_SHARE_ACCESS_READ;
1306 		break;
1307 	case FMODE_WRITE:
1308 		res = NFS4_SHARE_ACCESS_WRITE;
1309 		break;
1310 	case FMODE_READ|FMODE_WRITE:
1311 		res = NFS4_SHARE_ACCESS_BOTH;
1312 	}
1313 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1314 		goto out;
1315 	/* Want no delegation if we're using O_DIRECT */
1316 	if (openflags & O_DIRECT)
1317 		res |= NFS4_SHARE_WANT_NO_DELEG;
1318 out:
1319 	return res;
1320 }
1321 
1322 static enum open_claim_type4
1323 nfs4_map_atomic_open_claim(struct nfs_server *server,
1324 		enum open_claim_type4 claim)
1325 {
1326 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1327 		return claim;
1328 	switch (claim) {
1329 	default:
1330 		return claim;
1331 	case NFS4_OPEN_CLAIM_FH:
1332 		return NFS4_OPEN_CLAIM_NULL;
1333 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1334 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1335 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1336 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1337 	}
1338 }
1339 
1340 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1341 {
1342 	p->o_res.f_attr = &p->f_attr;
1343 	p->o_res.seqid = p->o_arg.seqid;
1344 	p->c_res.seqid = p->c_arg.seqid;
1345 	p->o_res.server = p->o_arg.server;
1346 	p->o_res.access_request = p->o_arg.access;
1347 	nfs_fattr_init(&p->f_attr);
1348 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1349 }
1350 
1351 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1352 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1353 		const struct nfs4_open_createattrs *c,
1354 		enum open_claim_type4 claim,
1355 		gfp_t gfp_mask)
1356 {
1357 	struct dentry *parent = dget_parent(dentry);
1358 	struct inode *dir = d_inode(parent);
1359 	struct nfs_server *server = NFS_SERVER(dir);
1360 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1361 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1362 	struct nfs4_opendata *p;
1363 
1364 	p = kzalloc(sizeof(*p), gfp_mask);
1365 	if (p == NULL)
1366 		goto err;
1367 
1368 	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1369 	if (IS_ERR(p->f_attr.label))
1370 		goto err_free_p;
1371 
1372 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1373 	if (IS_ERR(p->a_label))
1374 		goto err_free_f;
1375 
1376 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1377 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1378 	if (IS_ERR(p->o_arg.seqid))
1379 		goto err_free_label;
1380 	nfs_sb_active(dentry->d_sb);
1381 	p->dentry = dget(dentry);
1382 	p->dir = parent;
1383 	p->owner = sp;
1384 	atomic_inc(&sp->so_count);
1385 	p->o_arg.open_flags = flags;
1386 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1387 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1388 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1389 			fmode, flags);
1390 	if (flags & O_CREAT) {
1391 		p->o_arg.umask = current_umask();
1392 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1393 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1394 			p->o_arg.u.attrs = &p->attrs;
1395 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1396 
1397 			memcpy(p->o_arg.u.verifier.data, c->verf,
1398 					sizeof(p->o_arg.u.verifier.data));
1399 		}
1400 	}
1401 	/* ask server to check for all possible rights as results
1402 	 * are cached */
1403 	switch (p->o_arg.claim) {
1404 	default:
1405 		break;
1406 	case NFS4_OPEN_CLAIM_NULL:
1407 	case NFS4_OPEN_CLAIM_FH:
1408 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1409 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1410 				  NFS4_ACCESS_EXECUTE |
1411 				  nfs_access_xattr_mask(server);
1412 	}
1413 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1414 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1415 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1416 	p->o_arg.name = &dentry->d_name;
1417 	p->o_arg.server = server;
1418 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1419 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1420 	switch (p->o_arg.claim) {
1421 	case NFS4_OPEN_CLAIM_NULL:
1422 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1423 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1424 		p->o_arg.fh = NFS_FH(dir);
1425 		break;
1426 	case NFS4_OPEN_CLAIM_PREVIOUS:
1427 	case NFS4_OPEN_CLAIM_FH:
1428 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1429 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1430 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1431 	}
1432 	p->c_arg.fh = &p->o_res.fh;
1433 	p->c_arg.stateid = &p->o_res.stateid;
1434 	p->c_arg.seqid = p->o_arg.seqid;
1435 	nfs4_init_opendata_res(p);
1436 	kref_init(&p->kref);
1437 	return p;
1438 
1439 err_free_label:
1440 	nfs4_label_free(p->a_label);
1441 err_free_f:
1442 	nfs4_label_free(p->f_attr.label);
1443 err_free_p:
1444 	kfree(p);
1445 err:
1446 	dput(parent);
1447 	return NULL;
1448 }
1449 
1450 static void nfs4_opendata_free(struct kref *kref)
1451 {
1452 	struct nfs4_opendata *p = container_of(kref,
1453 			struct nfs4_opendata, kref);
1454 	struct super_block *sb = p->dentry->d_sb;
1455 
1456 	nfs4_lgopen_release(p->lgp);
1457 	nfs_free_seqid(p->o_arg.seqid);
1458 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1459 	if (p->state != NULL)
1460 		nfs4_put_open_state(p->state);
1461 	nfs4_put_state_owner(p->owner);
1462 
1463 	nfs4_label_free(p->a_label);
1464 	nfs4_label_free(p->f_attr.label);
1465 
1466 	dput(p->dir);
1467 	dput(p->dentry);
1468 	nfs_sb_deactive(sb);
1469 	nfs_fattr_free_names(&p->f_attr);
1470 	kfree(p->f_attr.mdsthreshold);
1471 	kfree(p);
1472 }
1473 
1474 static void nfs4_opendata_put(struct nfs4_opendata *p)
1475 {
1476 	if (p != NULL)
1477 		kref_put(&p->kref, nfs4_opendata_free);
1478 }
1479 
1480 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1481 		fmode_t fmode)
1482 {
1483 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1484 	case FMODE_READ|FMODE_WRITE:
1485 		return state->n_rdwr != 0;
1486 	case FMODE_WRITE:
1487 		return state->n_wronly != 0;
1488 	case FMODE_READ:
1489 		return state->n_rdonly != 0;
1490 	}
1491 	WARN_ON_ONCE(1);
1492 	return false;
1493 }
1494 
1495 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1496 		int open_mode, enum open_claim_type4 claim)
1497 {
1498 	int ret = 0;
1499 
1500 	if (open_mode & (O_EXCL|O_TRUNC))
1501 		goto out;
1502 	switch (claim) {
1503 	case NFS4_OPEN_CLAIM_NULL:
1504 	case NFS4_OPEN_CLAIM_FH:
1505 		goto out;
1506 	default:
1507 		break;
1508 	}
1509 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1510 		case FMODE_READ:
1511 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1512 				&& state->n_rdonly != 0;
1513 			break;
1514 		case FMODE_WRITE:
1515 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1516 				&& state->n_wronly != 0;
1517 			break;
1518 		case FMODE_READ|FMODE_WRITE:
1519 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1520 				&& state->n_rdwr != 0;
1521 	}
1522 out:
1523 	return ret;
1524 }
1525 
1526 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1527 		enum open_claim_type4 claim)
1528 {
1529 	if (delegation == NULL)
1530 		return 0;
1531 	if ((delegation->type & fmode) != fmode)
1532 		return 0;
1533 	switch (claim) {
1534 	case NFS4_OPEN_CLAIM_NULL:
1535 	case NFS4_OPEN_CLAIM_FH:
1536 		break;
1537 	case NFS4_OPEN_CLAIM_PREVIOUS:
1538 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1539 			break;
1540 		fallthrough;
1541 	default:
1542 		return 0;
1543 	}
1544 	nfs_mark_delegation_referenced(delegation);
1545 	return 1;
1546 }
1547 
1548 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1549 {
1550 	switch (fmode) {
1551 		case FMODE_WRITE:
1552 			state->n_wronly++;
1553 			break;
1554 		case FMODE_READ:
1555 			state->n_rdonly++;
1556 			break;
1557 		case FMODE_READ|FMODE_WRITE:
1558 			state->n_rdwr++;
1559 	}
1560 	nfs4_state_set_mode_locked(state, state->state | fmode);
1561 }
1562 
1563 #ifdef CONFIG_NFS_V4_1
1564 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1565 {
1566 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1567 		return true;
1568 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1569 		return true;
1570 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1571 		return true;
1572 	return false;
1573 }
1574 #endif /* CONFIG_NFS_V4_1 */
1575 
1576 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1577 {
1578 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1579 		wake_up_all(&state->waitq);
1580 }
1581 
1582 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1583 {
1584 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1585 	bool need_recover = false;
1586 
1587 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1588 		need_recover = true;
1589 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1590 		need_recover = true;
1591 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1592 		need_recover = true;
1593 	if (need_recover)
1594 		nfs4_state_mark_reclaim_nograce(clp, state);
1595 }
1596 
1597 /*
1598  * Check for whether or not the caller may update the open stateid
1599  * to the value passed in by stateid.
1600  *
1601  * Note: This function relies heavily on the server implementing
1602  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1603  * correctly.
1604  * i.e. The stateid seqids have to be initialised to 1, and
1605  * are then incremented on every state transition.
1606  */
1607 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1608 		const nfs4_stateid *stateid)
1609 {
1610 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1611 		/* The common case - we're updating to a new sequence number */
1612 		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1613 			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1614 				return true;
1615 			return false;
1616 		}
1617 		/* The server returned a new stateid */
1618 	}
1619 	/* This is the first OPEN in this generation */
1620 	if (stateid->seqid == cpu_to_be32(1))
1621 		return true;
1622 	return false;
1623 }
1624 
1625 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1626 {
1627 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1628 		return;
1629 	if (state->n_wronly)
1630 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1631 	if (state->n_rdonly)
1632 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1633 	if (state->n_rdwr)
1634 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1635 	set_bit(NFS_OPEN_STATE, &state->flags);
1636 }
1637 
1638 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1639 		nfs4_stateid *stateid, fmode_t fmode)
1640 {
1641 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1642 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1643 	case FMODE_WRITE:
1644 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1645 		break;
1646 	case FMODE_READ:
1647 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1648 		break;
1649 	case 0:
1650 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1651 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1652 		clear_bit(NFS_OPEN_STATE, &state->flags);
1653 	}
1654 	if (stateid == NULL)
1655 		return;
1656 	/* Handle OPEN+OPEN_DOWNGRADE races */
1657 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1658 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1659 		nfs_resync_open_stateid_locked(state);
1660 		goto out;
1661 	}
1662 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1663 		nfs4_stateid_copy(&state->stateid, stateid);
1664 	nfs4_stateid_copy(&state->open_stateid, stateid);
1665 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1666 out:
1667 	nfs_state_log_update_open_stateid(state);
1668 }
1669 
1670 static void nfs_clear_open_stateid(struct nfs4_state *state,
1671 	nfs4_stateid *arg_stateid,
1672 	nfs4_stateid *stateid, fmode_t fmode)
1673 {
1674 	write_seqlock(&state->seqlock);
1675 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1676 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1677 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1678 	write_sequnlock(&state->seqlock);
1679 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1680 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1681 }
1682 
1683 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1684 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1685 	__must_hold(&state->owner->so_lock)
1686 	__must_hold(&state->seqlock)
1687 	__must_hold(RCU)
1688 
1689 {
1690 	DEFINE_WAIT(wait);
1691 	int status = 0;
1692 	for (;;) {
1693 
1694 		if (nfs_stateid_is_sequential(state, stateid))
1695 			break;
1696 
1697 		if (status)
1698 			break;
1699 		/* Rely on seqids for serialisation with NFSv4.0 */
1700 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1701 			break;
1702 
1703 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1704 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1705 		/*
1706 		 * Ensure we process the state changes in the same order
1707 		 * in which the server processed them by delaying the
1708 		 * update of the stateid until we are in sequence.
1709 		 */
1710 		write_sequnlock(&state->seqlock);
1711 		spin_unlock(&state->owner->so_lock);
1712 		rcu_read_unlock();
1713 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1714 
1715 		if (!fatal_signal_pending(current)) {
1716 			if (schedule_timeout(5*HZ) == 0)
1717 				status = -EAGAIN;
1718 			else
1719 				status = 0;
1720 		} else
1721 			status = -EINTR;
1722 		finish_wait(&state->waitq, &wait);
1723 		rcu_read_lock();
1724 		spin_lock(&state->owner->so_lock);
1725 		write_seqlock(&state->seqlock);
1726 	}
1727 
1728 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1729 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1730 		nfs4_stateid_copy(freeme, &state->open_stateid);
1731 		nfs_test_and_clear_all_open_stateid(state);
1732 	}
1733 
1734 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1735 		nfs4_stateid_copy(&state->stateid, stateid);
1736 	nfs4_stateid_copy(&state->open_stateid, stateid);
1737 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1738 	nfs_state_log_update_open_stateid(state);
1739 }
1740 
1741 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1742 		const nfs4_stateid *open_stateid,
1743 		fmode_t fmode,
1744 		nfs4_stateid *freeme)
1745 {
1746 	/*
1747 	 * Protect the call to nfs4_state_set_mode_locked and
1748 	 * serialise the stateid update
1749 	 */
1750 	write_seqlock(&state->seqlock);
1751 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1752 	switch (fmode) {
1753 	case FMODE_READ:
1754 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1755 		break;
1756 	case FMODE_WRITE:
1757 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1758 		break;
1759 	case FMODE_READ|FMODE_WRITE:
1760 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1761 	}
1762 	set_bit(NFS_OPEN_STATE, &state->flags);
1763 	write_sequnlock(&state->seqlock);
1764 }
1765 
1766 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1767 {
1768 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1769 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1770 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1771 	clear_bit(NFS_OPEN_STATE, &state->flags);
1772 }
1773 
1774 static void nfs_state_set_delegation(struct nfs4_state *state,
1775 		const nfs4_stateid *deleg_stateid,
1776 		fmode_t fmode)
1777 {
1778 	/*
1779 	 * Protect the call to nfs4_state_set_mode_locked and
1780 	 * serialise the stateid update
1781 	 */
1782 	write_seqlock(&state->seqlock);
1783 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1784 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1785 	write_sequnlock(&state->seqlock);
1786 }
1787 
1788 static void nfs_state_clear_delegation(struct nfs4_state *state)
1789 {
1790 	write_seqlock(&state->seqlock);
1791 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1792 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1793 	write_sequnlock(&state->seqlock);
1794 }
1795 
1796 int update_open_stateid(struct nfs4_state *state,
1797 		const nfs4_stateid *open_stateid,
1798 		const nfs4_stateid *delegation,
1799 		fmode_t fmode)
1800 {
1801 	struct nfs_server *server = NFS_SERVER(state->inode);
1802 	struct nfs_client *clp = server->nfs_client;
1803 	struct nfs_inode *nfsi = NFS_I(state->inode);
1804 	struct nfs_delegation *deleg_cur;
1805 	nfs4_stateid freeme = { };
1806 	int ret = 0;
1807 
1808 	fmode &= (FMODE_READ|FMODE_WRITE);
1809 
1810 	rcu_read_lock();
1811 	spin_lock(&state->owner->so_lock);
1812 	if (open_stateid != NULL) {
1813 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1814 		ret = 1;
1815 	}
1816 
1817 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1818 	if (deleg_cur == NULL)
1819 		goto no_delegation;
1820 
1821 	spin_lock(&deleg_cur->lock);
1822 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1823 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1824 	    (deleg_cur->type & fmode) != fmode)
1825 		goto no_delegation_unlock;
1826 
1827 	if (delegation == NULL)
1828 		delegation = &deleg_cur->stateid;
1829 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1830 		goto no_delegation_unlock;
1831 
1832 	nfs_mark_delegation_referenced(deleg_cur);
1833 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1834 	ret = 1;
1835 no_delegation_unlock:
1836 	spin_unlock(&deleg_cur->lock);
1837 no_delegation:
1838 	if (ret)
1839 		update_open_stateflags(state, fmode);
1840 	spin_unlock(&state->owner->so_lock);
1841 	rcu_read_unlock();
1842 
1843 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1844 		nfs4_schedule_state_manager(clp);
1845 	if (freeme.type != 0)
1846 		nfs4_test_and_free_stateid(server, &freeme,
1847 				state->owner->so_cred);
1848 
1849 	return ret;
1850 }
1851 
1852 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1853 		const nfs4_stateid *stateid)
1854 {
1855 	struct nfs4_state *state = lsp->ls_state;
1856 	bool ret = false;
1857 
1858 	spin_lock(&state->state_lock);
1859 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1860 		goto out_noupdate;
1861 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1862 		goto out_noupdate;
1863 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1864 	ret = true;
1865 out_noupdate:
1866 	spin_unlock(&state->state_lock);
1867 	return ret;
1868 }
1869 
1870 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1871 {
1872 	struct nfs_delegation *delegation;
1873 
1874 	fmode &= FMODE_READ|FMODE_WRITE;
1875 	rcu_read_lock();
1876 	delegation = nfs4_get_valid_delegation(inode);
1877 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1878 		rcu_read_unlock();
1879 		return;
1880 	}
1881 	rcu_read_unlock();
1882 	nfs4_inode_return_delegation(inode);
1883 }
1884 
1885 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1886 {
1887 	struct nfs4_state *state = opendata->state;
1888 	struct nfs_delegation *delegation;
1889 	int open_mode = opendata->o_arg.open_flags;
1890 	fmode_t fmode = opendata->o_arg.fmode;
1891 	enum open_claim_type4 claim = opendata->o_arg.claim;
1892 	nfs4_stateid stateid;
1893 	int ret = -EAGAIN;
1894 
1895 	for (;;) {
1896 		spin_lock(&state->owner->so_lock);
1897 		if (can_open_cached(state, fmode, open_mode, claim)) {
1898 			update_open_stateflags(state, fmode);
1899 			spin_unlock(&state->owner->so_lock);
1900 			goto out_return_state;
1901 		}
1902 		spin_unlock(&state->owner->so_lock);
1903 		rcu_read_lock();
1904 		delegation = nfs4_get_valid_delegation(state->inode);
1905 		if (!can_open_delegated(delegation, fmode, claim)) {
1906 			rcu_read_unlock();
1907 			break;
1908 		}
1909 		/* Save the delegation */
1910 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1911 		rcu_read_unlock();
1912 		nfs_release_seqid(opendata->o_arg.seqid);
1913 		if (!opendata->is_recover) {
1914 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1915 			if (ret != 0)
1916 				goto out;
1917 		}
1918 		ret = -EAGAIN;
1919 
1920 		/* Try to update the stateid using the delegation */
1921 		if (update_open_stateid(state, NULL, &stateid, fmode))
1922 			goto out_return_state;
1923 	}
1924 out:
1925 	return ERR_PTR(ret);
1926 out_return_state:
1927 	refcount_inc(&state->count);
1928 	return state;
1929 }
1930 
1931 static void
1932 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1933 {
1934 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1935 	struct nfs_delegation *delegation;
1936 	int delegation_flags = 0;
1937 
1938 	rcu_read_lock();
1939 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1940 	if (delegation)
1941 		delegation_flags = delegation->flags;
1942 	rcu_read_unlock();
1943 	switch (data->o_arg.claim) {
1944 	default:
1945 		break;
1946 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1947 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1948 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1949 				   "returning a delegation for "
1950 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1951 				   clp->cl_hostname);
1952 		return;
1953 	}
1954 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1955 		nfs_inode_set_delegation(state->inode,
1956 				data->owner->so_cred,
1957 				data->o_res.delegation_type,
1958 				&data->o_res.delegation,
1959 				data->o_res.pagemod_limit);
1960 	else
1961 		nfs_inode_reclaim_delegation(state->inode,
1962 				data->owner->so_cred,
1963 				data->o_res.delegation_type,
1964 				&data->o_res.delegation,
1965 				data->o_res.pagemod_limit);
1966 
1967 	if (data->o_res.do_recall)
1968 		nfs_async_inode_return_delegation(state->inode,
1969 						  &data->o_res.delegation);
1970 }
1971 
1972 /*
1973  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1974  * and update the nfs4_state.
1975  */
1976 static struct nfs4_state *
1977 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1978 {
1979 	struct inode *inode = data->state->inode;
1980 	struct nfs4_state *state = data->state;
1981 	int ret;
1982 
1983 	if (!data->rpc_done) {
1984 		if (data->rpc_status)
1985 			return ERR_PTR(data->rpc_status);
1986 		return nfs4_try_open_cached(data);
1987 	}
1988 
1989 	ret = nfs_refresh_inode(inode, &data->f_attr);
1990 	if (ret)
1991 		return ERR_PTR(ret);
1992 
1993 	if (data->o_res.delegation_type != 0)
1994 		nfs4_opendata_check_deleg(data, state);
1995 
1996 	if (!update_open_stateid(state, &data->o_res.stateid,
1997 				NULL, data->o_arg.fmode))
1998 		return ERR_PTR(-EAGAIN);
1999 	refcount_inc(&state->count);
2000 
2001 	return state;
2002 }
2003 
2004 static struct inode *
2005 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2006 {
2007 	struct inode *inode;
2008 
2009 	switch (data->o_arg.claim) {
2010 	case NFS4_OPEN_CLAIM_NULL:
2011 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2012 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2013 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2014 			return ERR_PTR(-EAGAIN);
2015 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2016 				&data->f_attr);
2017 		break;
2018 	default:
2019 		inode = d_inode(data->dentry);
2020 		ihold(inode);
2021 		nfs_refresh_inode(inode, &data->f_attr);
2022 	}
2023 	return inode;
2024 }
2025 
2026 static struct nfs4_state *
2027 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2028 {
2029 	struct nfs4_state *state;
2030 	struct inode *inode;
2031 
2032 	inode = nfs4_opendata_get_inode(data);
2033 	if (IS_ERR(inode))
2034 		return ERR_CAST(inode);
2035 	if (data->state != NULL && data->state->inode == inode) {
2036 		state = data->state;
2037 		refcount_inc(&state->count);
2038 	} else
2039 		state = nfs4_get_open_state(inode, data->owner);
2040 	iput(inode);
2041 	if (state == NULL)
2042 		state = ERR_PTR(-ENOMEM);
2043 	return state;
2044 }
2045 
2046 static struct nfs4_state *
2047 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2048 {
2049 	struct nfs4_state *state;
2050 
2051 	if (!data->rpc_done) {
2052 		state = nfs4_try_open_cached(data);
2053 		trace_nfs4_cached_open(data->state);
2054 		goto out;
2055 	}
2056 
2057 	state = nfs4_opendata_find_nfs4_state(data);
2058 	if (IS_ERR(state))
2059 		goto out;
2060 
2061 	if (data->o_res.delegation_type != 0)
2062 		nfs4_opendata_check_deleg(data, state);
2063 	if (!update_open_stateid(state, &data->o_res.stateid,
2064 				NULL, data->o_arg.fmode)) {
2065 		nfs4_put_open_state(state);
2066 		state = ERR_PTR(-EAGAIN);
2067 	}
2068 out:
2069 	nfs_release_seqid(data->o_arg.seqid);
2070 	return state;
2071 }
2072 
2073 static struct nfs4_state *
2074 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2075 {
2076 	struct nfs4_state *ret;
2077 
2078 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2079 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2080 	else
2081 		ret = _nfs4_opendata_to_nfs4_state(data);
2082 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2083 	return ret;
2084 }
2085 
2086 static struct nfs_open_context *
2087 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2088 {
2089 	struct nfs_inode *nfsi = NFS_I(state->inode);
2090 	struct nfs_open_context *ctx;
2091 
2092 	rcu_read_lock();
2093 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2094 		if (ctx->state != state)
2095 			continue;
2096 		if ((ctx->mode & mode) != mode)
2097 			continue;
2098 		if (!get_nfs_open_context(ctx))
2099 			continue;
2100 		rcu_read_unlock();
2101 		return ctx;
2102 	}
2103 	rcu_read_unlock();
2104 	return ERR_PTR(-ENOENT);
2105 }
2106 
2107 static struct nfs_open_context *
2108 nfs4_state_find_open_context(struct nfs4_state *state)
2109 {
2110 	struct nfs_open_context *ctx;
2111 
2112 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2113 	if (!IS_ERR(ctx))
2114 		return ctx;
2115 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2116 	if (!IS_ERR(ctx))
2117 		return ctx;
2118 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2119 }
2120 
2121 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2122 		struct nfs4_state *state, enum open_claim_type4 claim)
2123 {
2124 	struct nfs4_opendata *opendata;
2125 
2126 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2127 			NULL, claim, GFP_NOFS);
2128 	if (opendata == NULL)
2129 		return ERR_PTR(-ENOMEM);
2130 	opendata->state = state;
2131 	refcount_inc(&state->count);
2132 	return opendata;
2133 }
2134 
2135 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2136 				    fmode_t fmode)
2137 {
2138 	struct nfs4_state *newstate;
2139 	struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2140 	int openflags = opendata->o_arg.open_flags;
2141 	int ret;
2142 
2143 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2144 		return 0;
2145 	opendata->o_arg.fmode = fmode;
2146 	opendata->o_arg.share_access =
2147 		nfs4_map_atomic_open_share(server, fmode, openflags);
2148 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2149 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2150 	nfs4_init_opendata_res(opendata);
2151 	ret = _nfs4_recover_proc_open(opendata);
2152 	if (ret != 0)
2153 		return ret;
2154 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2155 	if (IS_ERR(newstate))
2156 		return PTR_ERR(newstate);
2157 	if (newstate != opendata->state)
2158 		ret = -ESTALE;
2159 	nfs4_close_state(newstate, fmode);
2160 	return ret;
2161 }
2162 
2163 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2164 {
2165 	int ret;
2166 
2167 	/* memory barrier prior to reading state->n_* */
2168 	smp_rmb();
2169 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2170 	if (ret != 0)
2171 		return ret;
2172 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2173 	if (ret != 0)
2174 		return ret;
2175 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2176 	if (ret != 0)
2177 		return ret;
2178 	/*
2179 	 * We may have performed cached opens for all three recoveries.
2180 	 * Check if we need to update the current stateid.
2181 	 */
2182 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2183 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2184 		write_seqlock(&state->seqlock);
2185 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2186 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2187 		write_sequnlock(&state->seqlock);
2188 	}
2189 	return 0;
2190 }
2191 
2192 /*
2193  * OPEN_RECLAIM:
2194  * 	reclaim state on the server after a reboot.
2195  */
2196 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2197 {
2198 	struct nfs_delegation *delegation;
2199 	struct nfs4_opendata *opendata;
2200 	fmode_t delegation_type = 0;
2201 	int status;
2202 
2203 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2204 			NFS4_OPEN_CLAIM_PREVIOUS);
2205 	if (IS_ERR(opendata))
2206 		return PTR_ERR(opendata);
2207 	rcu_read_lock();
2208 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2209 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2210 		delegation_type = delegation->type;
2211 	rcu_read_unlock();
2212 	opendata->o_arg.u.delegation_type = delegation_type;
2213 	status = nfs4_open_recover(opendata, state);
2214 	nfs4_opendata_put(opendata);
2215 	return status;
2216 }
2217 
2218 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2219 {
2220 	struct nfs_server *server = NFS_SERVER(state->inode);
2221 	struct nfs4_exception exception = { };
2222 	int err;
2223 	do {
2224 		err = _nfs4_do_open_reclaim(ctx, state);
2225 		trace_nfs4_open_reclaim(ctx, 0, err);
2226 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2227 			continue;
2228 		if (err != -NFS4ERR_DELAY)
2229 			break;
2230 		nfs4_handle_exception(server, err, &exception);
2231 	} while (exception.retry);
2232 	return err;
2233 }
2234 
2235 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2236 {
2237 	struct nfs_open_context *ctx;
2238 	int ret;
2239 
2240 	ctx = nfs4_state_find_open_context(state);
2241 	if (IS_ERR(ctx))
2242 		return -EAGAIN;
2243 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2244 	nfs_state_clear_open_state_flags(state);
2245 	ret = nfs4_do_open_reclaim(ctx, state);
2246 	put_nfs_open_context(ctx);
2247 	return ret;
2248 }
2249 
2250 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)
2251 {
2252 	switch (err) {
2253 		default:
2254 			printk(KERN_ERR "NFS: %s: unhandled error "
2255 					"%d.\n", __func__, err);
2256 			fallthrough;
2257 		case 0:
2258 		case -ENOENT:
2259 		case -EAGAIN:
2260 		case -ESTALE:
2261 		case -ETIMEDOUT:
2262 			break;
2263 		case -NFS4ERR_BADSESSION:
2264 		case -NFS4ERR_BADSLOT:
2265 		case -NFS4ERR_BAD_HIGH_SLOT:
2266 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2267 		case -NFS4ERR_DEADSESSION:
2268 			return -EAGAIN;
2269 		case -NFS4ERR_STALE_CLIENTID:
2270 		case -NFS4ERR_STALE_STATEID:
2271 			/* Don't recall a delegation if it was lost */
2272 			nfs4_schedule_lease_recovery(server->nfs_client);
2273 			return -EAGAIN;
2274 		case -NFS4ERR_MOVED:
2275 			nfs4_schedule_migration_recovery(server);
2276 			return -EAGAIN;
2277 		case -NFS4ERR_LEASE_MOVED:
2278 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2279 			return -EAGAIN;
2280 		case -NFS4ERR_DELEG_REVOKED:
2281 		case -NFS4ERR_ADMIN_REVOKED:
2282 		case -NFS4ERR_EXPIRED:
2283 		case -NFS4ERR_BAD_STATEID:
2284 		case -NFS4ERR_OPENMODE:
2285 			nfs_inode_find_state_and_recover(state->inode,
2286 					stateid);
2287 			nfs4_schedule_stateid_recovery(server, state);
2288 			return -EAGAIN;
2289 		case -NFS4ERR_DELAY:
2290 		case -NFS4ERR_GRACE:
2291 			ssleep(1);
2292 			return -EAGAIN;
2293 		case -ENOMEM:
2294 		case -NFS4ERR_DENIED:
2295 			if (fl) {
2296 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2297 				if (lsp)
2298 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2299 			}
2300 			return 0;
2301 	}
2302 	return err;
2303 }
2304 
2305 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2306 		struct nfs4_state *state, const nfs4_stateid *stateid)
2307 {
2308 	struct nfs_server *server = NFS_SERVER(state->inode);
2309 	struct nfs4_opendata *opendata;
2310 	int err = 0;
2311 
2312 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2313 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2314 	if (IS_ERR(opendata))
2315 		return PTR_ERR(opendata);
2316 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2317 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2318 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2319 		if (err)
2320 			goto out;
2321 	}
2322 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2323 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2324 		if (err)
2325 			goto out;
2326 	}
2327 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2328 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2329 		if (err)
2330 			goto out;
2331 	}
2332 	nfs_state_clear_delegation(state);
2333 out:
2334 	nfs4_opendata_put(opendata);
2335 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2336 }
2337 
2338 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2339 {
2340 	struct nfs4_opendata *data = calldata;
2341 
2342 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2343 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2344 }
2345 
2346 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2347 {
2348 	struct nfs4_opendata *data = calldata;
2349 
2350 	nfs40_sequence_done(task, &data->c_res.seq_res);
2351 
2352 	data->rpc_status = task->tk_status;
2353 	if (data->rpc_status == 0) {
2354 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2355 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2356 		renew_lease(data->o_res.server, data->timestamp);
2357 		data->rpc_done = true;
2358 	}
2359 }
2360 
2361 static void nfs4_open_confirm_release(void *calldata)
2362 {
2363 	struct nfs4_opendata *data = calldata;
2364 	struct nfs4_state *state = NULL;
2365 
2366 	/* If this request hasn't been cancelled, do nothing */
2367 	if (!data->cancelled)
2368 		goto out_free;
2369 	/* In case of error, no cleanup! */
2370 	if (!data->rpc_done)
2371 		goto out_free;
2372 	state = nfs4_opendata_to_nfs4_state(data);
2373 	if (!IS_ERR(state))
2374 		nfs4_close_state(state, data->o_arg.fmode);
2375 out_free:
2376 	nfs4_opendata_put(data);
2377 }
2378 
2379 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2380 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2381 	.rpc_call_done = nfs4_open_confirm_done,
2382 	.rpc_release = nfs4_open_confirm_release,
2383 };
2384 
2385 /*
2386  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2387  */
2388 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2389 {
2390 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2391 	struct rpc_task *task;
2392 	struct  rpc_message msg = {
2393 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2394 		.rpc_argp = &data->c_arg,
2395 		.rpc_resp = &data->c_res,
2396 		.rpc_cred = data->owner->so_cred,
2397 	};
2398 	struct rpc_task_setup task_setup_data = {
2399 		.rpc_client = server->client,
2400 		.rpc_message = &msg,
2401 		.callback_ops = &nfs4_open_confirm_ops,
2402 		.callback_data = data,
2403 		.workqueue = nfsiod_workqueue,
2404 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2405 	};
2406 	int status;
2407 
2408 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2409 				data->is_recover);
2410 	kref_get(&data->kref);
2411 	data->rpc_done = false;
2412 	data->rpc_status = 0;
2413 	data->timestamp = jiffies;
2414 	task = rpc_run_task(&task_setup_data);
2415 	if (IS_ERR(task))
2416 		return PTR_ERR(task);
2417 	status = rpc_wait_for_completion_task(task);
2418 	if (status != 0) {
2419 		data->cancelled = true;
2420 		smp_wmb();
2421 	} else
2422 		status = data->rpc_status;
2423 	rpc_put_task(task);
2424 	return status;
2425 }
2426 
2427 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2428 {
2429 	struct nfs4_opendata *data = calldata;
2430 	struct nfs4_state_owner *sp = data->owner;
2431 	struct nfs_client *clp = sp->so_server->nfs_client;
2432 	enum open_claim_type4 claim = data->o_arg.claim;
2433 
2434 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2435 		goto out_wait;
2436 	/*
2437 	 * Check if we still need to send an OPEN call, or if we can use
2438 	 * a delegation instead.
2439 	 */
2440 	if (data->state != NULL) {
2441 		struct nfs_delegation *delegation;
2442 
2443 		if (can_open_cached(data->state, data->o_arg.fmode,
2444 					data->o_arg.open_flags, claim))
2445 			goto out_no_action;
2446 		rcu_read_lock();
2447 		delegation = nfs4_get_valid_delegation(data->state->inode);
2448 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2449 			goto unlock_no_action;
2450 		rcu_read_unlock();
2451 	}
2452 	/* Update client id. */
2453 	data->o_arg.clientid = clp->cl_clientid;
2454 	switch (claim) {
2455 	default:
2456 		break;
2457 	case NFS4_OPEN_CLAIM_PREVIOUS:
2458 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2459 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2460 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2461 		fallthrough;
2462 	case NFS4_OPEN_CLAIM_FH:
2463 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2464 	}
2465 	data->timestamp = jiffies;
2466 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2467 				&data->o_arg.seq_args,
2468 				&data->o_res.seq_res,
2469 				task) != 0)
2470 		nfs_release_seqid(data->o_arg.seqid);
2471 
2472 	/* Set the create mode (note dependency on the session type) */
2473 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2474 	if (data->o_arg.open_flags & O_EXCL) {
2475 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2476 		if (clp->cl_mvops->minor_version == 0) {
2477 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2478 			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2479 			 * because ACCESS will return permission denied for
2480 			 * all bits until close */
2481 			data->o_res.access_request = data->o_arg.access = 0;
2482 		} else if (nfs4_has_persistent_session(clp))
2483 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2484 	}
2485 	return;
2486 unlock_no_action:
2487 	trace_nfs4_cached_open(data->state);
2488 	rcu_read_unlock();
2489 out_no_action:
2490 	task->tk_action = NULL;
2491 out_wait:
2492 	nfs4_sequence_done(task, &data->o_res.seq_res);
2493 }
2494 
2495 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2496 {
2497 	struct nfs4_opendata *data = calldata;
2498 
2499 	data->rpc_status = task->tk_status;
2500 
2501 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2502 		return;
2503 
2504 	if (task->tk_status == 0) {
2505 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2506 			switch (data->o_res.f_attr->mode & S_IFMT) {
2507 			case S_IFREG:
2508 				break;
2509 			case S_IFLNK:
2510 				data->rpc_status = -ELOOP;
2511 				break;
2512 			case S_IFDIR:
2513 				data->rpc_status = -EISDIR;
2514 				break;
2515 			default:
2516 				data->rpc_status = -ENOTDIR;
2517 			}
2518 		}
2519 		renew_lease(data->o_res.server, data->timestamp);
2520 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2521 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2522 	}
2523 	data->rpc_done = true;
2524 }
2525 
2526 static void nfs4_open_release(void *calldata)
2527 {
2528 	struct nfs4_opendata *data = calldata;
2529 	struct nfs4_state *state = NULL;
2530 
2531 	/* If this request hasn't been cancelled, do nothing */
2532 	if (!data->cancelled)
2533 		goto out_free;
2534 	/* In case of error, no cleanup! */
2535 	if (data->rpc_status != 0 || !data->rpc_done)
2536 		goto out_free;
2537 	/* In case we need an open_confirm, no cleanup! */
2538 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2539 		goto out_free;
2540 	state = nfs4_opendata_to_nfs4_state(data);
2541 	if (!IS_ERR(state))
2542 		nfs4_close_state(state, data->o_arg.fmode);
2543 out_free:
2544 	nfs4_opendata_put(data);
2545 }
2546 
2547 static const struct rpc_call_ops nfs4_open_ops = {
2548 	.rpc_call_prepare = nfs4_open_prepare,
2549 	.rpc_call_done = nfs4_open_done,
2550 	.rpc_release = nfs4_open_release,
2551 };
2552 
2553 static int nfs4_run_open_task(struct nfs4_opendata *data,
2554 			      struct nfs_open_context *ctx)
2555 {
2556 	struct inode *dir = d_inode(data->dir);
2557 	struct nfs_server *server = NFS_SERVER(dir);
2558 	struct nfs_openargs *o_arg = &data->o_arg;
2559 	struct nfs_openres *o_res = &data->o_res;
2560 	struct rpc_task *task;
2561 	struct rpc_message msg = {
2562 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2563 		.rpc_argp = o_arg,
2564 		.rpc_resp = o_res,
2565 		.rpc_cred = data->owner->so_cred,
2566 	};
2567 	struct rpc_task_setup task_setup_data = {
2568 		.rpc_client = server->client,
2569 		.rpc_message = &msg,
2570 		.callback_ops = &nfs4_open_ops,
2571 		.callback_data = data,
2572 		.workqueue = nfsiod_workqueue,
2573 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2574 	};
2575 	int status;
2576 
2577 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2578 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2579 
2580 	kref_get(&data->kref);
2581 	data->rpc_done = false;
2582 	data->rpc_status = 0;
2583 	data->cancelled = false;
2584 	data->is_recover = false;
2585 	if (!ctx) {
2586 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2587 		data->is_recover = true;
2588 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2589 	} else {
2590 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2591 		pnfs_lgopen_prepare(data, ctx);
2592 	}
2593 	task = rpc_run_task(&task_setup_data);
2594 	if (IS_ERR(task))
2595 		return PTR_ERR(task);
2596 	status = rpc_wait_for_completion_task(task);
2597 	if (status != 0) {
2598 		data->cancelled = true;
2599 		smp_wmb();
2600 	} else
2601 		status = data->rpc_status;
2602 	rpc_put_task(task);
2603 
2604 	return status;
2605 }
2606 
2607 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2608 {
2609 	struct inode *dir = d_inode(data->dir);
2610 	struct nfs_openres *o_res = &data->o_res;
2611 	int status;
2612 
2613 	status = nfs4_run_open_task(data, NULL);
2614 	if (status != 0 || !data->rpc_done)
2615 		return status;
2616 
2617 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2618 
2619 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2620 		status = _nfs4_proc_open_confirm(data);
2621 
2622 	return status;
2623 }
2624 
2625 /*
2626  * Additional permission checks in order to distinguish between an
2627  * open for read, and an open for execute. This works around the
2628  * fact that NFSv4 OPEN treats read and execute permissions as being
2629  * the same.
2630  * Note that in the non-execute case, we want to turn off permission
2631  * checking if we just created a new file (POSIX open() semantics).
2632  */
2633 static int nfs4_opendata_access(const struct cred *cred,
2634 				struct nfs4_opendata *opendata,
2635 				struct nfs4_state *state, fmode_t fmode)
2636 {
2637 	struct nfs_access_entry cache;
2638 	u32 mask, flags;
2639 
2640 	/* access call failed or for some reason the server doesn't
2641 	 * support any access modes -- defer access call until later */
2642 	if (opendata->o_res.access_supported == 0)
2643 		return 0;
2644 
2645 	mask = 0;
2646 	if (fmode & FMODE_EXEC) {
2647 		/* ONLY check for exec rights */
2648 		if (S_ISDIR(state->inode->i_mode))
2649 			mask = NFS4_ACCESS_LOOKUP;
2650 		else
2651 			mask = NFS4_ACCESS_EXECUTE;
2652 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2653 		mask = NFS4_ACCESS_READ;
2654 
2655 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2656 	nfs_access_add_cache(state->inode, &cache, cred);
2657 
2658 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2659 	if ((mask & ~cache.mask & flags) == 0)
2660 		return 0;
2661 
2662 	return -EACCES;
2663 }
2664 
2665 /*
2666  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2667  */
2668 static int _nfs4_proc_open(struct nfs4_opendata *data,
2669 			   struct nfs_open_context *ctx)
2670 {
2671 	struct inode *dir = d_inode(data->dir);
2672 	struct nfs_server *server = NFS_SERVER(dir);
2673 	struct nfs_openargs *o_arg = &data->o_arg;
2674 	struct nfs_openres *o_res = &data->o_res;
2675 	int status;
2676 
2677 	status = nfs4_run_open_task(data, ctx);
2678 	if (!data->rpc_done)
2679 		return status;
2680 	if (status != 0) {
2681 		if (status == -NFS4ERR_BADNAME &&
2682 				!(o_arg->open_flags & O_CREAT))
2683 			return -ENOENT;
2684 		return status;
2685 	}
2686 
2687 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2688 
2689 	if (o_arg->open_flags & O_CREAT) {
2690 		if (o_arg->open_flags & O_EXCL)
2691 			data->file_created = true;
2692 		else if (o_res->cinfo.before != o_res->cinfo.after)
2693 			data->file_created = true;
2694 		if (data->file_created ||
2695 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2696 			nfs4_update_changeattr(dir, &o_res->cinfo,
2697 					o_res->f_attr->time_start,
2698 					NFS_INO_INVALID_DATA);
2699 	}
2700 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2701 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2702 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2703 		status = _nfs4_proc_open_confirm(data);
2704 		if (status != 0)
2705 			return status;
2706 	}
2707 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2708 		struct nfs_fh *fh = &o_res->fh;
2709 
2710 		nfs4_sequence_free_slot(&o_res->seq_res);
2711 		if (o_arg->claim == NFS4_OPEN_CLAIM_FH)
2712 			fh = NFS_FH(d_inode(data->dentry));
2713 		nfs4_proc_getattr(server, fh, o_res->f_attr, NULL);
2714 	}
2715 	return 0;
2716 }
2717 
2718 /*
2719  * OPEN_EXPIRED:
2720  * 	reclaim state on the server after a network partition.
2721  * 	Assumes caller holds the appropriate lock
2722  */
2723 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2724 {
2725 	struct nfs4_opendata *opendata;
2726 	int ret;
2727 
2728 	opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2729 	if (IS_ERR(opendata))
2730 		return PTR_ERR(opendata);
2731 	/*
2732 	 * We're not recovering a delegation, so ask for no delegation.
2733 	 * Otherwise the recovery thread could deadlock with an outstanding
2734 	 * delegation return.
2735 	 */
2736 	opendata->o_arg.open_flags = O_DIRECT;
2737 	ret = nfs4_open_recover(opendata, state);
2738 	if (ret == -ESTALE)
2739 		d_drop(ctx->dentry);
2740 	nfs4_opendata_put(opendata);
2741 	return ret;
2742 }
2743 
2744 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2745 {
2746 	struct nfs_server *server = NFS_SERVER(state->inode);
2747 	struct nfs4_exception exception = { };
2748 	int err;
2749 
2750 	do {
2751 		err = _nfs4_open_expired(ctx, state);
2752 		trace_nfs4_open_expired(ctx, 0, err);
2753 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2754 			continue;
2755 		switch (err) {
2756 		default:
2757 			goto out;
2758 		case -NFS4ERR_GRACE:
2759 		case -NFS4ERR_DELAY:
2760 			nfs4_handle_exception(server, err, &exception);
2761 			err = 0;
2762 		}
2763 	} while (exception.retry);
2764 out:
2765 	return err;
2766 }
2767 
2768 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2769 {
2770 	struct nfs_open_context *ctx;
2771 	int ret;
2772 
2773 	ctx = nfs4_state_find_open_context(state);
2774 	if (IS_ERR(ctx))
2775 		return -EAGAIN;
2776 	ret = nfs4_do_open_expired(ctx, state);
2777 	put_nfs_open_context(ctx);
2778 	return ret;
2779 }
2780 
2781 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2782 		const nfs4_stateid *stateid)
2783 {
2784 	nfs_remove_bad_delegation(state->inode, stateid);
2785 	nfs_state_clear_delegation(state);
2786 }
2787 
2788 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2789 {
2790 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2791 		nfs_finish_clear_delegation_stateid(state, NULL);
2792 }
2793 
2794 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2795 {
2796 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2797 	nfs40_clear_delegation_stateid(state);
2798 	nfs_state_clear_open_state_flags(state);
2799 	return nfs4_open_expired(sp, state);
2800 }
2801 
2802 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2803 		nfs4_stateid *stateid,
2804 		const struct cred *cred)
2805 {
2806 	return -NFS4ERR_BAD_STATEID;
2807 }
2808 
2809 #if defined(CONFIG_NFS_V4_1)
2810 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2811 		nfs4_stateid *stateid,
2812 		const struct cred *cred)
2813 {
2814 	int status;
2815 
2816 	switch (stateid->type) {
2817 	default:
2818 		break;
2819 	case NFS4_INVALID_STATEID_TYPE:
2820 	case NFS4_SPECIAL_STATEID_TYPE:
2821 		return -NFS4ERR_BAD_STATEID;
2822 	case NFS4_REVOKED_STATEID_TYPE:
2823 		goto out_free;
2824 	}
2825 
2826 	status = nfs41_test_stateid(server, stateid, cred);
2827 	switch (status) {
2828 	case -NFS4ERR_EXPIRED:
2829 	case -NFS4ERR_ADMIN_REVOKED:
2830 	case -NFS4ERR_DELEG_REVOKED:
2831 		break;
2832 	default:
2833 		return status;
2834 	}
2835 out_free:
2836 	/* Ack the revoked state to the server */
2837 	nfs41_free_stateid(server, stateid, cred, true);
2838 	return -NFS4ERR_EXPIRED;
2839 }
2840 
2841 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2842 {
2843 	struct nfs_server *server = NFS_SERVER(state->inode);
2844 	nfs4_stateid stateid;
2845 	struct nfs_delegation *delegation;
2846 	const struct cred *cred = NULL;
2847 	int status, ret = NFS_OK;
2848 
2849 	/* Get the delegation credential for use by test/free_stateid */
2850 	rcu_read_lock();
2851 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2852 	if (delegation == NULL) {
2853 		rcu_read_unlock();
2854 		nfs_state_clear_delegation(state);
2855 		return NFS_OK;
2856 	}
2857 
2858 	spin_lock(&delegation->lock);
2859 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2860 
2861 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2862 				&delegation->flags)) {
2863 		spin_unlock(&delegation->lock);
2864 		rcu_read_unlock();
2865 		return NFS_OK;
2866 	}
2867 
2868 	if (delegation->cred)
2869 		cred = get_cred(delegation->cred);
2870 	spin_unlock(&delegation->lock);
2871 	rcu_read_unlock();
2872 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2873 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2874 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2875 		nfs_finish_clear_delegation_stateid(state, &stateid);
2876 	else
2877 		ret = status;
2878 
2879 	put_cred(cred);
2880 	return ret;
2881 }
2882 
2883 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2884 {
2885 	nfs4_stateid tmp;
2886 
2887 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2888 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2889 				&tmp, NULL) &&
2890 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2891 		nfs_state_set_delegation(state, &tmp, state->state);
2892 	else
2893 		nfs_state_clear_delegation(state);
2894 }
2895 
2896 /**
2897  * nfs41_check_expired_locks - possibly free a lock stateid
2898  *
2899  * @state: NFSv4 state for an inode
2900  *
2901  * Returns NFS_OK if recovery for this stateid is now finished.
2902  * Otherwise a negative NFS4ERR value is returned.
2903  */
2904 static int nfs41_check_expired_locks(struct nfs4_state *state)
2905 {
2906 	int status, ret = NFS_OK;
2907 	struct nfs4_lock_state *lsp, *prev = NULL;
2908 	struct nfs_server *server = NFS_SERVER(state->inode);
2909 
2910 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2911 		goto out;
2912 
2913 	spin_lock(&state->state_lock);
2914 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2915 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2916 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2917 
2918 			refcount_inc(&lsp->ls_count);
2919 			spin_unlock(&state->state_lock);
2920 
2921 			nfs4_put_lock_state(prev);
2922 			prev = lsp;
2923 
2924 			status = nfs41_test_and_free_expired_stateid(server,
2925 					&lsp->ls_stateid,
2926 					cred);
2927 			trace_nfs4_test_lock_stateid(state, lsp, status);
2928 			if (status == -NFS4ERR_EXPIRED ||
2929 			    status == -NFS4ERR_BAD_STATEID) {
2930 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2931 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2932 				if (!recover_lost_locks)
2933 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2934 			} else if (status != NFS_OK) {
2935 				ret = status;
2936 				nfs4_put_lock_state(prev);
2937 				goto out;
2938 			}
2939 			spin_lock(&state->state_lock);
2940 		}
2941 	}
2942 	spin_unlock(&state->state_lock);
2943 	nfs4_put_lock_state(prev);
2944 out:
2945 	return ret;
2946 }
2947 
2948 /**
2949  * nfs41_check_open_stateid - possibly free an open stateid
2950  *
2951  * @state: NFSv4 state for an inode
2952  *
2953  * Returns NFS_OK if recovery for this stateid is now finished.
2954  * Otherwise a negative NFS4ERR value is returned.
2955  */
2956 static int nfs41_check_open_stateid(struct nfs4_state *state)
2957 {
2958 	struct nfs_server *server = NFS_SERVER(state->inode);
2959 	nfs4_stateid *stateid = &state->open_stateid;
2960 	const struct cred *cred = state->owner->so_cred;
2961 	int status;
2962 
2963 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2964 		return -NFS4ERR_BAD_STATEID;
2965 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2966 	trace_nfs4_test_open_stateid(state, NULL, status);
2967 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2968 		nfs_state_clear_open_state_flags(state);
2969 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2970 		return status;
2971 	}
2972 	if (nfs_open_stateid_recover_openmode(state))
2973 		return -NFS4ERR_OPENMODE;
2974 	return NFS_OK;
2975 }
2976 
2977 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2978 {
2979 	int status;
2980 
2981 	status = nfs41_check_delegation_stateid(state);
2982 	if (status != NFS_OK)
2983 		return status;
2984 	nfs41_delegation_recover_stateid(state);
2985 
2986 	status = nfs41_check_expired_locks(state);
2987 	if (status != NFS_OK)
2988 		return status;
2989 	status = nfs41_check_open_stateid(state);
2990 	if (status != NFS_OK)
2991 		status = nfs4_open_expired(sp, state);
2992 	return status;
2993 }
2994 #endif
2995 
2996 /*
2997  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2998  * fields corresponding to attributes that were used to store the verifier.
2999  * Make sure we clobber those fields in the later setattr call
3000  */
3001 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
3002 				struct iattr *sattr, struct nfs4_label **label)
3003 {
3004 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
3005 	__u32 attrset[3];
3006 	unsigned ret;
3007 	unsigned i;
3008 
3009 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3010 		attrset[i] = opendata->o_res.attrset[i];
3011 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3012 			attrset[i] &= ~bitmask[i];
3013 	}
3014 
3015 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3016 		sattr->ia_valid : 0;
3017 
3018 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3019 		if (sattr->ia_valid & ATTR_ATIME_SET)
3020 			ret |= ATTR_ATIME_SET;
3021 		else
3022 			ret |= ATTR_ATIME;
3023 	}
3024 
3025 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3026 		if (sattr->ia_valid & ATTR_MTIME_SET)
3027 			ret |= ATTR_MTIME_SET;
3028 		else
3029 			ret |= ATTR_MTIME;
3030 	}
3031 
3032 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3033 		*label = NULL;
3034 	return ret;
3035 }
3036 
3037 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3038 		struct nfs_open_context *ctx)
3039 {
3040 	struct nfs4_state_owner *sp = opendata->owner;
3041 	struct nfs_server *server = sp->so_server;
3042 	struct dentry *dentry;
3043 	struct nfs4_state *state;
3044 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3045 	struct inode *dir = d_inode(opendata->dir);
3046 	unsigned long dir_verifier;
3047 	unsigned int seq;
3048 	int ret;
3049 
3050 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3051 	dir_verifier = nfs_save_change_attribute(dir);
3052 
3053 	ret = _nfs4_proc_open(opendata, ctx);
3054 	if (ret != 0)
3055 		goto out;
3056 
3057 	state = _nfs4_opendata_to_nfs4_state(opendata);
3058 	ret = PTR_ERR(state);
3059 	if (IS_ERR(state))
3060 		goto out;
3061 	ctx->state = state;
3062 	if (server->caps & NFS_CAP_POSIX_LOCK)
3063 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3064 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3065 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3066 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3067 		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3068 
3069 	dentry = opendata->dentry;
3070 	if (d_really_is_negative(dentry)) {
3071 		struct dentry *alias;
3072 		d_drop(dentry);
3073 		alias = d_exact_alias(dentry, state->inode);
3074 		if (!alias)
3075 			alias = d_splice_alias(igrab(state->inode), dentry);
3076 		/* d_splice_alias() can't fail here - it's a non-directory */
3077 		if (alias) {
3078 			dput(ctx->dentry);
3079 			ctx->dentry = dentry = alias;
3080 		}
3081 	}
3082 
3083 	switch(opendata->o_arg.claim) {
3084 	default:
3085 		break;
3086 	case NFS4_OPEN_CLAIM_NULL:
3087 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3088 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3089 		if (!opendata->rpc_done)
3090 			break;
3091 		if (opendata->o_res.delegation_type != 0)
3092 			dir_verifier = nfs_save_change_attribute(dir);
3093 		nfs_set_verifier(dentry, dir_verifier);
3094 	}
3095 
3096 	/* Parse layoutget results before we check for access */
3097 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3098 
3099 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3100 	if (ret != 0)
3101 		goto out;
3102 
3103 	if (d_inode(dentry) == state->inode) {
3104 		nfs_inode_attach_open_context(ctx);
3105 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3106 			nfs4_schedule_stateid_recovery(server, state);
3107 	}
3108 
3109 out:
3110 	if (!opendata->cancelled) {
3111 		if (opendata->lgp) {
3112 			nfs4_lgopen_release(opendata->lgp);
3113 			opendata->lgp = NULL;
3114 		}
3115 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3116 	}
3117 	return ret;
3118 }
3119 
3120 /*
3121  * Returns a referenced nfs4_state
3122  */
3123 static int _nfs4_do_open(struct inode *dir,
3124 			struct nfs_open_context *ctx,
3125 			int flags,
3126 			const struct nfs4_open_createattrs *c,
3127 			int *opened)
3128 {
3129 	struct nfs4_state_owner  *sp;
3130 	struct nfs4_state     *state = NULL;
3131 	struct nfs_server       *server = NFS_SERVER(dir);
3132 	struct nfs4_opendata *opendata;
3133 	struct dentry *dentry = ctx->dentry;
3134 	const struct cred *cred = ctx->cred;
3135 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3136 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3137 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3138 	struct iattr *sattr = c->sattr;
3139 	struct nfs4_label *label = c->label;
3140 	int status;
3141 
3142 	/* Protect against reboot recovery conflicts */
3143 	status = -ENOMEM;
3144 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3145 	if (sp == NULL) {
3146 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3147 		goto out_err;
3148 	}
3149 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3150 	if (status != 0)
3151 		goto err_put_state_owner;
3152 	if (d_really_is_positive(dentry))
3153 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3154 	status = -ENOMEM;
3155 	if (d_really_is_positive(dentry))
3156 		claim = NFS4_OPEN_CLAIM_FH;
3157 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3158 			c, claim, GFP_KERNEL);
3159 	if (opendata == NULL)
3160 		goto err_put_state_owner;
3161 
3162 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3163 		if (!opendata->f_attr.mdsthreshold) {
3164 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3165 			if (!opendata->f_attr.mdsthreshold)
3166 				goto err_opendata_put;
3167 		}
3168 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3169 	}
3170 	if (d_really_is_positive(dentry))
3171 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3172 
3173 	status = _nfs4_open_and_get_state(opendata, ctx);
3174 	if (status != 0)
3175 		goto err_opendata_put;
3176 	state = ctx->state;
3177 
3178 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3179 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3180 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3181 		/*
3182 		 * send create attributes which was not set by open
3183 		 * with an extra setattr.
3184 		 */
3185 		if (attrs || label) {
3186 			unsigned ia_old = sattr->ia_valid;
3187 
3188 			sattr->ia_valid = attrs;
3189 			nfs_fattr_init(opendata->o_res.f_attr);
3190 			status = nfs4_do_setattr(state->inode, cred,
3191 					opendata->o_res.f_attr, sattr,
3192 					ctx, label);
3193 			if (status == 0) {
3194 				nfs_setattr_update_inode(state->inode, sattr,
3195 						opendata->o_res.f_attr);
3196 				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3197 			}
3198 			sattr->ia_valid = ia_old;
3199 		}
3200 	}
3201 	if (opened && opendata->file_created)
3202 		*opened = 1;
3203 
3204 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3205 		*ctx_th = opendata->f_attr.mdsthreshold;
3206 		opendata->f_attr.mdsthreshold = NULL;
3207 	}
3208 
3209 	nfs4_opendata_put(opendata);
3210 	nfs4_put_state_owner(sp);
3211 	return 0;
3212 err_opendata_put:
3213 	nfs4_opendata_put(opendata);
3214 err_put_state_owner:
3215 	nfs4_put_state_owner(sp);
3216 out_err:
3217 	return status;
3218 }
3219 
3220 
3221 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3222 					struct nfs_open_context *ctx,
3223 					int flags,
3224 					struct iattr *sattr,
3225 					struct nfs4_label *label,
3226 					int *opened)
3227 {
3228 	struct nfs_server *server = NFS_SERVER(dir);
3229 	struct nfs4_exception exception = {
3230 		.interruptible = true,
3231 	};
3232 	struct nfs4_state *res;
3233 	struct nfs4_open_createattrs c = {
3234 		.label = label,
3235 		.sattr = sattr,
3236 		.verf = {
3237 			[0] = (__u32)jiffies,
3238 			[1] = (__u32)current->pid,
3239 		},
3240 	};
3241 	int status;
3242 
3243 	do {
3244 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3245 		res = ctx->state;
3246 		trace_nfs4_open_file(ctx, flags, status);
3247 		if (status == 0)
3248 			break;
3249 		/* NOTE: BAD_SEQID means the server and client disagree about the
3250 		 * book-keeping w.r.t. state-changing operations
3251 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3252 		 * It is actually a sign of a bug on the client or on the server.
3253 		 *
3254 		 * If we receive a BAD_SEQID error in the particular case of
3255 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3256 		 * have unhashed the old state_owner for us, and that we can
3257 		 * therefore safely retry using a new one. We should still warn
3258 		 * the user though...
3259 		 */
3260 		if (status == -NFS4ERR_BAD_SEQID) {
3261 			pr_warn_ratelimited("NFS: v4 server %s "
3262 					" returned a bad sequence-id error!\n",
3263 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3264 			exception.retry = 1;
3265 			continue;
3266 		}
3267 		/*
3268 		 * BAD_STATEID on OPEN means that the server cancelled our
3269 		 * state before it received the OPEN_CONFIRM.
3270 		 * Recover by retrying the request as per the discussion
3271 		 * on Page 181 of RFC3530.
3272 		 */
3273 		if (status == -NFS4ERR_BAD_STATEID) {
3274 			exception.retry = 1;
3275 			continue;
3276 		}
3277 		if (status == -NFS4ERR_EXPIRED) {
3278 			nfs4_schedule_lease_recovery(server->nfs_client);
3279 			exception.retry = 1;
3280 			continue;
3281 		}
3282 		if (status == -EAGAIN) {
3283 			/* We must have found a delegation */
3284 			exception.retry = 1;
3285 			continue;
3286 		}
3287 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3288 			continue;
3289 		res = ERR_PTR(nfs4_handle_exception(server,
3290 					status, &exception));
3291 	} while (exception.retry);
3292 	return res;
3293 }
3294 
3295 static int _nfs4_do_setattr(struct inode *inode,
3296 			    struct nfs_setattrargs *arg,
3297 			    struct nfs_setattrres *res,
3298 			    const struct cred *cred,
3299 			    struct nfs_open_context *ctx)
3300 {
3301 	struct nfs_server *server = NFS_SERVER(inode);
3302 	struct rpc_message msg = {
3303 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3304 		.rpc_argp	= arg,
3305 		.rpc_resp	= res,
3306 		.rpc_cred	= cred,
3307 	};
3308 	const struct cred *delegation_cred = NULL;
3309 	unsigned long timestamp = jiffies;
3310 	bool truncate;
3311 	int status;
3312 
3313 	nfs_fattr_init(res->fattr);
3314 
3315 	/* Servers should only apply open mode checks for file size changes */
3316 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3317 	if (!truncate) {
3318 		nfs4_inode_make_writeable(inode);
3319 		goto zero_stateid;
3320 	}
3321 
3322 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3323 		/* Use that stateid */
3324 	} else if (ctx != NULL && ctx->state) {
3325 		struct nfs_lock_context *l_ctx;
3326 		if (!nfs4_valid_open_stateid(ctx->state))
3327 			return -EBADF;
3328 		l_ctx = nfs_get_lock_context(ctx);
3329 		if (IS_ERR(l_ctx))
3330 			return PTR_ERR(l_ctx);
3331 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3332 						&arg->stateid, &delegation_cred);
3333 		nfs_put_lock_context(l_ctx);
3334 		if (status == -EIO)
3335 			return -EBADF;
3336 		else if (status == -EAGAIN)
3337 			goto zero_stateid;
3338 	} else {
3339 zero_stateid:
3340 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3341 	}
3342 	if (delegation_cred)
3343 		msg.rpc_cred = delegation_cred;
3344 
3345 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3346 
3347 	put_cred(delegation_cred);
3348 	if (status == 0 && ctx != NULL)
3349 		renew_lease(server, timestamp);
3350 	trace_nfs4_setattr(inode, &arg->stateid, status);
3351 	return status;
3352 }
3353 
3354 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3355 			   struct nfs_fattr *fattr, struct iattr *sattr,
3356 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3357 {
3358 	struct nfs_server *server = NFS_SERVER(inode);
3359 	__u32 bitmask[NFS4_BITMASK_SZ];
3360 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3361 	struct nfs_setattrargs	arg = {
3362 		.fh		= NFS_FH(inode),
3363 		.iap		= sattr,
3364 		.server		= server,
3365 		.bitmask = bitmask,
3366 		.label		= ilabel,
3367 	};
3368 	struct nfs_setattrres  res = {
3369 		.fattr		= fattr,
3370 		.server		= server,
3371 	};
3372 	struct nfs4_exception exception = {
3373 		.state = state,
3374 		.inode = inode,
3375 		.stateid = &arg.stateid,
3376 	};
3377 	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3378 	int err;
3379 
3380 	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3381 		adjust_flags |= NFS_INO_INVALID_MODE;
3382 	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3383 		adjust_flags |= NFS_INO_INVALID_OTHER;
3384 
3385 	do {
3386 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3387 					inode, adjust_flags);
3388 
3389 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3390 		switch (err) {
3391 		case -NFS4ERR_OPENMODE:
3392 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3393 				pr_warn_once("NFSv4: server %s is incorrectly "
3394 						"applying open mode checks to "
3395 						"a SETATTR that is not "
3396 						"changing file size.\n",
3397 						server->nfs_client->cl_hostname);
3398 			}
3399 			if (state && !(state->state & FMODE_WRITE)) {
3400 				err = -EBADF;
3401 				if (sattr->ia_valid & ATTR_OPEN)
3402 					err = -EACCES;
3403 				goto out;
3404 			}
3405 		}
3406 		err = nfs4_handle_exception(server, err, &exception);
3407 	} while (exception.retry);
3408 out:
3409 	return err;
3410 }
3411 
3412 static bool
3413 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3414 {
3415 	if (inode == NULL || !nfs_have_layout(inode))
3416 		return false;
3417 
3418 	return pnfs_wait_on_layoutreturn(inode, task);
3419 }
3420 
3421 /*
3422  * Update the seqid of an open stateid
3423  */
3424 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3425 		struct nfs4_state *state)
3426 {
3427 	__be32 seqid_open;
3428 	u32 dst_seqid;
3429 	int seq;
3430 
3431 	for (;;) {
3432 		if (!nfs4_valid_open_stateid(state))
3433 			break;
3434 		seq = read_seqbegin(&state->seqlock);
3435 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3436 			nfs4_stateid_copy(dst, &state->open_stateid);
3437 			if (read_seqretry(&state->seqlock, seq))
3438 				continue;
3439 			break;
3440 		}
3441 		seqid_open = state->open_stateid.seqid;
3442 		if (read_seqretry(&state->seqlock, seq))
3443 			continue;
3444 
3445 		dst_seqid = be32_to_cpu(dst->seqid);
3446 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3447 			dst->seqid = seqid_open;
3448 		break;
3449 	}
3450 }
3451 
3452 /*
3453  * Update the seqid of an open stateid after receiving
3454  * NFS4ERR_OLD_STATEID
3455  */
3456 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3457 		struct nfs4_state *state)
3458 {
3459 	__be32 seqid_open;
3460 	u32 dst_seqid;
3461 	bool ret;
3462 	int seq, status = -EAGAIN;
3463 	DEFINE_WAIT(wait);
3464 
3465 	for (;;) {
3466 		ret = false;
3467 		if (!nfs4_valid_open_stateid(state))
3468 			break;
3469 		seq = read_seqbegin(&state->seqlock);
3470 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3471 			if (read_seqretry(&state->seqlock, seq))
3472 				continue;
3473 			break;
3474 		}
3475 
3476 		write_seqlock(&state->seqlock);
3477 		seqid_open = state->open_stateid.seqid;
3478 
3479 		dst_seqid = be32_to_cpu(dst->seqid);
3480 
3481 		/* Did another OPEN bump the state's seqid?  try again: */
3482 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3483 			dst->seqid = seqid_open;
3484 			write_sequnlock(&state->seqlock);
3485 			ret = true;
3486 			break;
3487 		}
3488 
3489 		/* server says we're behind but we haven't seen the update yet */
3490 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3491 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3492 		write_sequnlock(&state->seqlock);
3493 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3494 
3495 		if (fatal_signal_pending(current))
3496 			status = -EINTR;
3497 		else
3498 			if (schedule_timeout(5*HZ) != 0)
3499 				status = 0;
3500 
3501 		finish_wait(&state->waitq, &wait);
3502 
3503 		if (!status)
3504 			continue;
3505 		if (status == -EINTR)
3506 			break;
3507 
3508 		/* we slept the whole 5 seconds, we must have lost a seqid */
3509 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3510 		ret = true;
3511 		break;
3512 	}
3513 
3514 	return ret;
3515 }
3516 
3517 struct nfs4_closedata {
3518 	struct inode *inode;
3519 	struct nfs4_state *state;
3520 	struct nfs_closeargs arg;
3521 	struct nfs_closeres res;
3522 	struct {
3523 		struct nfs4_layoutreturn_args arg;
3524 		struct nfs4_layoutreturn_res res;
3525 		struct nfs4_xdr_opaque_data ld_private;
3526 		u32 roc_barrier;
3527 		bool roc;
3528 	} lr;
3529 	struct nfs_fattr fattr;
3530 	unsigned long timestamp;
3531 };
3532 
3533 static void nfs4_free_closedata(void *data)
3534 {
3535 	struct nfs4_closedata *calldata = data;
3536 	struct nfs4_state_owner *sp = calldata->state->owner;
3537 	struct super_block *sb = calldata->state->inode->i_sb;
3538 
3539 	if (calldata->lr.roc)
3540 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3541 				calldata->res.lr_ret);
3542 	nfs4_put_open_state(calldata->state);
3543 	nfs_free_seqid(calldata->arg.seqid);
3544 	nfs4_put_state_owner(sp);
3545 	nfs_sb_deactive(sb);
3546 	kfree(calldata);
3547 }
3548 
3549 static void nfs4_close_done(struct rpc_task *task, void *data)
3550 {
3551 	struct nfs4_closedata *calldata = data;
3552 	struct nfs4_state *state = calldata->state;
3553 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3554 	nfs4_stateid *res_stateid = NULL;
3555 	struct nfs4_exception exception = {
3556 		.state = state,
3557 		.inode = calldata->inode,
3558 		.stateid = &calldata->arg.stateid,
3559 	};
3560 
3561 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3562 		return;
3563 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3564 
3565 	/* Handle Layoutreturn errors */
3566 	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3567 			  &calldata->res.lr_ret) == -EAGAIN)
3568 		goto out_restart;
3569 
3570 	/* hmm. we are done with the inode, and in the process of freeing
3571 	 * the state_owner. we keep this around to process errors
3572 	 */
3573 	switch (task->tk_status) {
3574 		case 0:
3575 			res_stateid = &calldata->res.stateid;
3576 			renew_lease(server, calldata->timestamp);
3577 			break;
3578 		case -NFS4ERR_ACCESS:
3579 			if (calldata->arg.bitmask != NULL) {
3580 				calldata->arg.bitmask = NULL;
3581 				calldata->res.fattr = NULL;
3582 				goto out_restart;
3583 
3584 			}
3585 			break;
3586 		case -NFS4ERR_OLD_STATEID:
3587 			/* Did we race with OPEN? */
3588 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3589 						state))
3590 				goto out_restart;
3591 			goto out_release;
3592 		case -NFS4ERR_ADMIN_REVOKED:
3593 		case -NFS4ERR_STALE_STATEID:
3594 		case -NFS4ERR_EXPIRED:
3595 			nfs4_free_revoked_stateid(server,
3596 					&calldata->arg.stateid,
3597 					task->tk_msg.rpc_cred);
3598 			fallthrough;
3599 		case -NFS4ERR_BAD_STATEID:
3600 			if (calldata->arg.fmode == 0)
3601 				break;
3602 			fallthrough;
3603 		default:
3604 			task->tk_status = nfs4_async_handle_exception(task,
3605 					server, task->tk_status, &exception);
3606 			if (exception.retry)
3607 				goto out_restart;
3608 	}
3609 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3610 			res_stateid, calldata->arg.fmode);
3611 out_release:
3612 	task->tk_status = 0;
3613 	nfs_release_seqid(calldata->arg.seqid);
3614 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3615 	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3616 	return;
3617 out_restart:
3618 	task->tk_status = 0;
3619 	rpc_restart_call_prepare(task);
3620 	goto out_release;
3621 }
3622 
3623 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3624 {
3625 	struct nfs4_closedata *calldata = data;
3626 	struct nfs4_state *state = calldata->state;
3627 	struct inode *inode = calldata->inode;
3628 	struct nfs_server *server = NFS_SERVER(inode);
3629 	struct pnfs_layout_hdr *lo;
3630 	bool is_rdonly, is_wronly, is_rdwr;
3631 	int call_close = 0;
3632 
3633 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3634 		goto out_wait;
3635 
3636 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3637 	spin_lock(&state->owner->so_lock);
3638 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3639 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3640 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3641 	/* Calculate the change in open mode */
3642 	calldata->arg.fmode = 0;
3643 	if (state->n_rdwr == 0) {
3644 		if (state->n_rdonly == 0)
3645 			call_close |= is_rdonly;
3646 		else if (is_rdonly)
3647 			calldata->arg.fmode |= FMODE_READ;
3648 		if (state->n_wronly == 0)
3649 			call_close |= is_wronly;
3650 		else if (is_wronly)
3651 			calldata->arg.fmode |= FMODE_WRITE;
3652 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3653 			call_close |= is_rdwr;
3654 	} else if (is_rdwr)
3655 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3656 
3657 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3658 	if (!nfs4_valid_open_stateid(state))
3659 		call_close = 0;
3660 	spin_unlock(&state->owner->so_lock);
3661 
3662 	if (!call_close) {
3663 		/* Note: exit _without_ calling nfs4_close_done */
3664 		goto out_no_action;
3665 	}
3666 
3667 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3668 		nfs_release_seqid(calldata->arg.seqid);
3669 		goto out_wait;
3670 	}
3671 
3672 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3673 	if (lo && !pnfs_layout_is_valid(lo)) {
3674 		calldata->arg.lr_args = NULL;
3675 		calldata->res.lr_res = NULL;
3676 	}
3677 
3678 	if (calldata->arg.fmode == 0)
3679 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3680 
3681 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3682 		/* Close-to-open cache consistency revalidation */
3683 		if (!nfs4_have_delegation(inode, FMODE_READ)) {
3684 			nfs4_bitmask_set(calldata->arg.bitmask_store,
3685 					 server->cache_consistency_bitmask,
3686 					 inode, 0);
3687 			calldata->arg.bitmask = calldata->arg.bitmask_store;
3688 		} else
3689 			calldata->arg.bitmask = NULL;
3690 	}
3691 
3692 	calldata->arg.share_access =
3693 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3694 				calldata->arg.fmode, 0);
3695 
3696 	if (calldata->res.fattr == NULL)
3697 		calldata->arg.bitmask = NULL;
3698 	else if (calldata->arg.bitmask == NULL)
3699 		calldata->res.fattr = NULL;
3700 	calldata->timestamp = jiffies;
3701 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3702 				&calldata->arg.seq_args,
3703 				&calldata->res.seq_res,
3704 				task) != 0)
3705 		nfs_release_seqid(calldata->arg.seqid);
3706 	return;
3707 out_no_action:
3708 	task->tk_action = NULL;
3709 out_wait:
3710 	nfs4_sequence_done(task, &calldata->res.seq_res);
3711 }
3712 
3713 static const struct rpc_call_ops nfs4_close_ops = {
3714 	.rpc_call_prepare = nfs4_close_prepare,
3715 	.rpc_call_done = nfs4_close_done,
3716 	.rpc_release = nfs4_free_closedata,
3717 };
3718 
3719 /*
3720  * It is possible for data to be read/written from a mem-mapped file
3721  * after the sys_close call (which hits the vfs layer as a flush).
3722  * This means that we can't safely call nfsv4 close on a file until
3723  * the inode is cleared. This in turn means that we are not good
3724  * NFSv4 citizens - we do not indicate to the server to update the file's
3725  * share state even when we are done with one of the three share
3726  * stateid's in the inode.
3727  *
3728  * NOTE: Caller must be holding the sp->so_owner semaphore!
3729  */
3730 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3731 {
3732 	struct nfs_server *server = NFS_SERVER(state->inode);
3733 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3734 	struct nfs4_closedata *calldata;
3735 	struct nfs4_state_owner *sp = state->owner;
3736 	struct rpc_task *task;
3737 	struct rpc_message msg = {
3738 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3739 		.rpc_cred = state->owner->so_cred,
3740 	};
3741 	struct rpc_task_setup task_setup_data = {
3742 		.rpc_client = server->client,
3743 		.rpc_message = &msg,
3744 		.callback_ops = &nfs4_close_ops,
3745 		.workqueue = nfsiod_workqueue,
3746 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3747 	};
3748 	int status = -ENOMEM;
3749 
3750 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3751 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3752 
3753 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3754 		&task_setup_data.rpc_client, &msg);
3755 
3756 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3757 	if (calldata == NULL)
3758 		goto out;
3759 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3760 	calldata->inode = state->inode;
3761 	calldata->state = state;
3762 	calldata->arg.fh = NFS_FH(state->inode);
3763 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3764 		goto out_free_calldata;
3765 	/* Serialization for the sequence id */
3766 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3767 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3768 	if (IS_ERR(calldata->arg.seqid))
3769 		goto out_free_calldata;
3770 	nfs_fattr_init(&calldata->fattr);
3771 	calldata->arg.fmode = 0;
3772 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3773 	calldata->res.fattr = &calldata->fattr;
3774 	calldata->res.seqid = calldata->arg.seqid;
3775 	calldata->res.server = server;
3776 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3777 	calldata->lr.roc = pnfs_roc(state->inode,
3778 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3779 	if (calldata->lr.roc) {
3780 		calldata->arg.lr_args = &calldata->lr.arg;
3781 		calldata->res.lr_res = &calldata->lr.res;
3782 	}
3783 	nfs_sb_active(calldata->inode->i_sb);
3784 
3785 	msg.rpc_argp = &calldata->arg;
3786 	msg.rpc_resp = &calldata->res;
3787 	task_setup_data.callback_data = calldata;
3788 	task = rpc_run_task(&task_setup_data);
3789 	if (IS_ERR(task))
3790 		return PTR_ERR(task);
3791 	status = 0;
3792 	if (wait)
3793 		status = rpc_wait_for_completion_task(task);
3794 	rpc_put_task(task);
3795 	return status;
3796 out_free_calldata:
3797 	kfree(calldata);
3798 out:
3799 	nfs4_put_open_state(state);
3800 	nfs4_put_state_owner(sp);
3801 	return status;
3802 }
3803 
3804 static struct inode *
3805 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3806 		int open_flags, struct iattr *attr, int *opened)
3807 {
3808 	struct nfs4_state *state;
3809 	struct nfs4_label l, *label;
3810 
3811 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3812 
3813 	/* Protect against concurrent sillydeletes */
3814 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3815 
3816 	nfs4_label_release_security(label);
3817 
3818 	if (IS_ERR(state))
3819 		return ERR_CAST(state);
3820 	return state->inode;
3821 }
3822 
3823 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3824 {
3825 	if (ctx->state == NULL)
3826 		return;
3827 	if (is_sync)
3828 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3829 	else
3830 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3831 }
3832 
3833 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3834 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3835 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3836 
3837 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3838 {
3839 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3840 	struct nfs4_server_caps_arg args = {
3841 		.fhandle = fhandle,
3842 		.bitmask = bitmask,
3843 	};
3844 	struct nfs4_server_caps_res res = {};
3845 	struct rpc_message msg = {
3846 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3847 		.rpc_argp = &args,
3848 		.rpc_resp = &res,
3849 	};
3850 	int status;
3851 	int i;
3852 
3853 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3854 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3855 		     FATTR4_WORD0_LINK_SUPPORT |
3856 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3857 		     FATTR4_WORD0_ACLSUPPORT |
3858 		     FATTR4_WORD0_CASE_INSENSITIVE |
3859 		     FATTR4_WORD0_CASE_PRESERVING;
3860 	if (minorversion)
3861 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3862 
3863 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3864 	if (status == 0) {
3865 		/* Sanity check the server answers */
3866 		switch (minorversion) {
3867 		case 0:
3868 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3869 			res.attr_bitmask[2] = 0;
3870 			break;
3871 		case 1:
3872 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3873 			break;
3874 		case 2:
3875 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3876 		}
3877 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3878 		server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3879 				  NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3880 		server->fattr_valid = NFS_ATTR_FATTR_V4;
3881 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3882 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3883 			server->caps |= NFS_CAP_ACLS;
3884 		if (res.has_links != 0)
3885 			server->caps |= NFS_CAP_HARDLINKS;
3886 		if (res.has_symlinks != 0)
3887 			server->caps |= NFS_CAP_SYMLINKS;
3888 		if (res.case_insensitive)
3889 			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3890 		if (res.case_preserving)
3891 			server->caps |= NFS_CAP_CASE_PRESERVING;
3892 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3893 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3894 			server->caps |= NFS_CAP_SECURITY_LABEL;
3895 #endif
3896 		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3897 			server->caps |= NFS_CAP_FS_LOCATIONS;
3898 		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3899 			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3900 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3901 			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3902 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3903 			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3904 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3905 			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3906 				NFS_ATTR_FATTR_OWNER_NAME);
3907 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3908 			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3909 				NFS_ATTR_FATTR_GROUP_NAME);
3910 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3911 			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3912 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3913 			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3914 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3915 			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3916 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3917 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3918 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3919 				sizeof(server->attr_bitmask));
3920 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3921 
3922 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3923 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3924 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3925 		server->cache_consistency_bitmask[2] = 0;
3926 
3927 		/* Avoid a regression due to buggy server */
3928 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3929 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3930 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3931 			sizeof(server->exclcreat_bitmask));
3932 
3933 		server->acl_bitmask = res.acl_bitmask;
3934 		server->fh_expire_type = res.fh_expire_type;
3935 	}
3936 
3937 	return status;
3938 }
3939 
3940 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3941 {
3942 	struct nfs4_exception exception = {
3943 		.interruptible = true,
3944 	};
3945 	int err;
3946 
3947 	nfs4_server_set_init_caps(server);
3948 	do {
3949 		err = nfs4_handle_exception(server,
3950 				_nfs4_server_capabilities(server, fhandle),
3951 				&exception);
3952 	} while (exception.retry);
3953 	return err;
3954 }
3955 
3956 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3957 					  struct nfs_client *clp,
3958 					  struct nfs_server *server)
3959 {
3960 	int i;
3961 
3962 	for (i = 0; i < location->nservers; i++) {
3963 		struct nfs4_string *srv_loc = &location->servers[i];
3964 		struct sockaddr_storage addr;
3965 		size_t addrlen;
3966 		struct xprt_create xprt_args = {
3967 			.ident = 0,
3968 			.net = clp->cl_net,
3969 		};
3970 		struct nfs4_add_xprt_data xprtdata = {
3971 			.clp = clp,
3972 		};
3973 		struct rpc_add_xprt_test rpcdata = {
3974 			.add_xprt_test = clp->cl_mvops->session_trunk,
3975 			.data = &xprtdata,
3976 		};
3977 		char *servername = NULL;
3978 
3979 		if (!srv_loc->len)
3980 			continue;
3981 
3982 		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3983 						&addr, sizeof(addr),
3984 						clp->cl_net, server->port);
3985 		if (!addrlen)
3986 			return;
3987 		xprt_args.dstaddr = (struct sockaddr *)&addr;
3988 		xprt_args.addrlen = addrlen;
3989 		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3990 		if (!servername)
3991 			return;
3992 		memcpy(servername, srv_loc->data, srv_loc->len);
3993 		servername[srv_loc->len] = '\0';
3994 		xprt_args.servername = servername;
3995 
3996 		xprtdata.cred = nfs4_get_clid_cred(clp);
3997 		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3998 				  rpc_clnt_setup_test_and_add_xprt,
3999 				  &rpcdata);
4000 		if (xprtdata.cred)
4001 			put_cred(xprtdata.cred);
4002 		kfree(servername);
4003 	}
4004 }
4005 
4006 static int _nfs4_discover_trunking(struct nfs_server *server,
4007 				   struct nfs_fh *fhandle)
4008 {
4009 	struct nfs4_fs_locations *locations = NULL;
4010 	struct page *page;
4011 	const struct cred *cred;
4012 	struct nfs_client *clp = server->nfs_client;
4013 	const struct nfs4_state_maintenance_ops *ops =
4014 		clp->cl_mvops->state_renewal_ops;
4015 	int status = -ENOMEM, i;
4016 
4017 	cred = ops->get_state_renewal_cred(clp);
4018 	if (cred == NULL) {
4019 		cred = nfs4_get_clid_cred(clp);
4020 		if (cred == NULL)
4021 			return -ENOKEY;
4022 	}
4023 
4024 	page = alloc_page(GFP_KERNEL);
4025 	if (!page)
4026 		goto out_put_cred;
4027 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4028 	if (!locations)
4029 		goto out_free;
4030 	locations->fattr = nfs_alloc_fattr();
4031 	if (!locations->fattr)
4032 		goto out_free_2;
4033 
4034 	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4035 					 cred);
4036 	if (status)
4037 		goto out_free_3;
4038 
4039 	for (i = 0; i < locations->nlocations; i++)
4040 		test_fs_location_for_trunking(&locations->locations[i], clp,
4041 					      server);
4042 out_free_3:
4043 	kfree(locations->fattr);
4044 out_free_2:
4045 	kfree(locations);
4046 out_free:
4047 	__free_page(page);
4048 out_put_cred:
4049 	put_cred(cred);
4050 	return status;
4051 }
4052 
4053 static int nfs4_discover_trunking(struct nfs_server *server,
4054 				  struct nfs_fh *fhandle)
4055 {
4056 	struct nfs4_exception exception = {
4057 		.interruptible = true,
4058 	};
4059 	struct nfs_client *clp = server->nfs_client;
4060 	int err = 0;
4061 
4062 	if (!nfs4_has_session(clp))
4063 		goto out;
4064 	do {
4065 		err = nfs4_handle_exception(server,
4066 				_nfs4_discover_trunking(server, fhandle),
4067 				&exception);
4068 	} while (exception.retry);
4069 out:
4070 	return err;
4071 }
4072 
4073 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4074 		struct nfs_fsinfo *info)
4075 {
4076 	u32 bitmask[3];
4077 	struct nfs4_lookup_root_arg args = {
4078 		.bitmask = bitmask,
4079 	};
4080 	struct nfs4_lookup_res res = {
4081 		.server = server,
4082 		.fattr = info->fattr,
4083 		.fh = fhandle,
4084 	};
4085 	struct rpc_message msg = {
4086 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4087 		.rpc_argp = &args,
4088 		.rpc_resp = &res,
4089 	};
4090 
4091 	bitmask[0] = nfs4_fattr_bitmap[0];
4092 	bitmask[1] = nfs4_fattr_bitmap[1];
4093 	/*
4094 	 * Process the label in the upcoming getfattr
4095 	 */
4096 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4097 
4098 	nfs_fattr_init(info->fattr);
4099 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4100 }
4101 
4102 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4103 		struct nfs_fsinfo *info)
4104 {
4105 	struct nfs4_exception exception = {
4106 		.interruptible = true,
4107 	};
4108 	int err;
4109 	do {
4110 		err = _nfs4_lookup_root(server, fhandle, info);
4111 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4112 		switch (err) {
4113 		case 0:
4114 		case -NFS4ERR_WRONGSEC:
4115 			goto out;
4116 		default:
4117 			err = nfs4_handle_exception(server, err, &exception);
4118 		}
4119 	} while (exception.retry);
4120 out:
4121 	return err;
4122 }
4123 
4124 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4125 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4126 {
4127 	struct rpc_auth_create_args auth_args = {
4128 		.pseudoflavor = flavor,
4129 	};
4130 	struct rpc_auth *auth;
4131 
4132 	auth = rpcauth_create(&auth_args, server->client);
4133 	if (IS_ERR(auth))
4134 		return -EACCES;
4135 	return nfs4_lookup_root(server, fhandle, info);
4136 }
4137 
4138 /*
4139  * Retry pseudoroot lookup with various security flavors.  We do this when:
4140  *
4141  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4142  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4143  *
4144  * Returns zero on success, or a negative NFS4ERR value, or a
4145  * negative errno value.
4146  */
4147 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4148 			      struct nfs_fsinfo *info)
4149 {
4150 	/* Per 3530bis 15.33.5 */
4151 	static const rpc_authflavor_t flav_array[] = {
4152 		RPC_AUTH_GSS_KRB5P,
4153 		RPC_AUTH_GSS_KRB5I,
4154 		RPC_AUTH_GSS_KRB5,
4155 		RPC_AUTH_UNIX,			/* courtesy */
4156 		RPC_AUTH_NULL,
4157 	};
4158 	int status = -EPERM;
4159 	size_t i;
4160 
4161 	if (server->auth_info.flavor_len > 0) {
4162 		/* try each flavor specified by user */
4163 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4164 			status = nfs4_lookup_root_sec(server, fhandle, info,
4165 						server->auth_info.flavors[i]);
4166 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4167 				continue;
4168 			break;
4169 		}
4170 	} else {
4171 		/* no flavors specified by user, try default list */
4172 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4173 			status = nfs4_lookup_root_sec(server, fhandle, info,
4174 						      flav_array[i]);
4175 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4176 				continue;
4177 			break;
4178 		}
4179 	}
4180 
4181 	/*
4182 	 * -EACCES could mean that the user doesn't have correct permissions
4183 	 * to access the mount.  It could also mean that we tried to mount
4184 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4185 	 * existing mount programs don't handle -EACCES very well so it should
4186 	 * be mapped to -EPERM instead.
4187 	 */
4188 	if (status == -EACCES)
4189 		status = -EPERM;
4190 	return status;
4191 }
4192 
4193 /**
4194  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4195  * @server: initialized nfs_server handle
4196  * @fhandle: we fill in the pseudo-fs root file handle
4197  * @info: we fill in an FSINFO struct
4198  * @auth_probe: probe the auth flavours
4199  *
4200  * Returns zero on success, or a negative errno.
4201  */
4202 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4203 			 struct nfs_fsinfo *info,
4204 			 bool auth_probe)
4205 {
4206 	int status = 0;
4207 
4208 	if (!auth_probe)
4209 		status = nfs4_lookup_root(server, fhandle, info);
4210 
4211 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4212 		status = server->nfs_client->cl_mvops->find_root_sec(server,
4213 				fhandle, info);
4214 
4215 	if (status == 0)
4216 		status = nfs4_server_capabilities(server, fhandle);
4217 	if (status == 0)
4218 		status = nfs4_do_fsinfo(server, fhandle, info);
4219 
4220 	return nfs4_map_errors(status);
4221 }
4222 
4223 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4224 			      struct nfs_fsinfo *info)
4225 {
4226 	int error;
4227 	struct nfs_fattr *fattr = info->fattr;
4228 
4229 	error = nfs4_server_capabilities(server, mntfh);
4230 	if (error < 0) {
4231 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4232 		return error;
4233 	}
4234 
4235 	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4236 	if (error < 0) {
4237 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4238 		goto out;
4239 	}
4240 
4241 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4242 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4243 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4244 
4245 out:
4246 	return error;
4247 }
4248 
4249 /*
4250  * Get locations and (maybe) other attributes of a referral.
4251  * Note that we'll actually follow the referral later when
4252  * we detect fsid mismatch in inode revalidation
4253  */
4254 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4255 			     const struct qstr *name, struct nfs_fattr *fattr,
4256 			     struct nfs_fh *fhandle)
4257 {
4258 	int status = -ENOMEM;
4259 	struct page *page = NULL;
4260 	struct nfs4_fs_locations *locations = NULL;
4261 
4262 	page = alloc_page(GFP_KERNEL);
4263 	if (page == NULL)
4264 		goto out;
4265 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4266 	if (locations == NULL)
4267 		goto out;
4268 
4269 	locations->fattr = fattr;
4270 
4271 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4272 	if (status != 0)
4273 		goto out;
4274 
4275 	/*
4276 	 * If the fsid didn't change, this is a migration event, not a
4277 	 * referral.  Cause us to drop into the exception handler, which
4278 	 * will kick off migration recovery.
4279 	 */
4280 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4281 		dprintk("%s: server did not return a different fsid for"
4282 			" a referral at %s\n", __func__, name->name);
4283 		status = -NFS4ERR_MOVED;
4284 		goto out;
4285 	}
4286 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4287 	nfs_fixup_referral_attributes(fattr);
4288 	memset(fhandle, 0, sizeof(struct nfs_fh));
4289 out:
4290 	if (page)
4291 		__free_page(page);
4292 	kfree(locations);
4293 	return status;
4294 }
4295 
4296 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4297 				struct nfs_fattr *fattr, struct inode *inode)
4298 {
4299 	__u32 bitmask[NFS4_BITMASK_SZ];
4300 	struct nfs4_getattr_arg args = {
4301 		.fh = fhandle,
4302 		.bitmask = bitmask,
4303 	};
4304 	struct nfs4_getattr_res res = {
4305 		.fattr = fattr,
4306 		.server = server,
4307 	};
4308 	struct rpc_message msg = {
4309 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4310 		.rpc_argp = &args,
4311 		.rpc_resp = &res,
4312 	};
4313 	unsigned short task_flags = 0;
4314 
4315 	if (nfs4_has_session(server->nfs_client))
4316 		task_flags = RPC_TASK_MOVEABLE;
4317 
4318 	/* Is this is an attribute revalidation, subject to softreval? */
4319 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4320 		task_flags |= RPC_TASK_TIMEOUT;
4321 
4322 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4323 	nfs_fattr_init(fattr);
4324 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4325 	return nfs4_do_call_sync(server->client, server, &msg,
4326 			&args.seq_args, &res.seq_res, task_flags);
4327 }
4328 
4329 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4330 				struct nfs_fattr *fattr, struct inode *inode)
4331 {
4332 	struct nfs4_exception exception = {
4333 		.interruptible = true,
4334 	};
4335 	int err;
4336 	do {
4337 		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4338 		trace_nfs4_getattr(server, fhandle, fattr, err);
4339 		err = nfs4_handle_exception(server, err,
4340 				&exception);
4341 	} while (exception.retry);
4342 	return err;
4343 }
4344 
4345 /*
4346  * The file is not closed if it is opened due to the a request to change
4347  * the size of the file. The open call will not be needed once the
4348  * VFS layer lookup-intents are implemented.
4349  *
4350  * Close is called when the inode is destroyed.
4351  * If we haven't opened the file for O_WRONLY, we
4352  * need to in the size_change case to obtain a stateid.
4353  *
4354  * Got race?
4355  * Because OPEN is always done by name in nfsv4, it is
4356  * possible that we opened a different file by the same
4357  * name.  We can recognize this race condition, but we
4358  * can't do anything about it besides returning an error.
4359  *
4360  * This will be fixed with VFS changes (lookup-intent).
4361  */
4362 static int
4363 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4364 		  struct iattr *sattr)
4365 {
4366 	struct inode *inode = d_inode(dentry);
4367 	const struct cred *cred = NULL;
4368 	struct nfs_open_context *ctx = NULL;
4369 	int status;
4370 
4371 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4372 	    sattr->ia_valid & ATTR_SIZE &&
4373 	    sattr->ia_size < i_size_read(inode))
4374 		pnfs_commit_and_return_layout(inode);
4375 
4376 	nfs_fattr_init(fattr);
4377 
4378 	/* Deal with open(O_TRUNC) */
4379 	if (sattr->ia_valid & ATTR_OPEN)
4380 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4381 
4382 	/* Optimization: if the end result is no change, don't RPC */
4383 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4384 		return 0;
4385 
4386 	/* Search for an existing open(O_WRITE) file */
4387 	if (sattr->ia_valid & ATTR_FILE) {
4388 
4389 		ctx = nfs_file_open_context(sattr->ia_file);
4390 		if (ctx)
4391 			cred = ctx->cred;
4392 	}
4393 
4394 	/* Return any delegations if we're going to change ACLs */
4395 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4396 		nfs4_inode_make_writeable(inode);
4397 
4398 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4399 	if (status == 0) {
4400 		nfs_setattr_update_inode(inode, sattr, fattr);
4401 		nfs_setsecurity(inode, fattr);
4402 	}
4403 	return status;
4404 }
4405 
4406 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4407 		struct dentry *dentry, struct nfs_fh *fhandle,
4408 		struct nfs_fattr *fattr)
4409 {
4410 	struct nfs_server *server = NFS_SERVER(dir);
4411 	int		       status;
4412 	struct nfs4_lookup_arg args = {
4413 		.bitmask = server->attr_bitmask,
4414 		.dir_fh = NFS_FH(dir),
4415 		.name = &dentry->d_name,
4416 	};
4417 	struct nfs4_lookup_res res = {
4418 		.server = server,
4419 		.fattr = fattr,
4420 		.fh = fhandle,
4421 	};
4422 	struct rpc_message msg = {
4423 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4424 		.rpc_argp = &args,
4425 		.rpc_resp = &res,
4426 	};
4427 	unsigned short task_flags = 0;
4428 
4429 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4430 		task_flags = RPC_TASK_MOVEABLE;
4431 
4432 	/* Is this is an attribute revalidation, subject to softreval? */
4433 	if (nfs_lookup_is_soft_revalidate(dentry))
4434 		task_flags |= RPC_TASK_TIMEOUT;
4435 
4436 	args.bitmask = nfs4_bitmask(server, fattr->label);
4437 
4438 	nfs_fattr_init(fattr);
4439 
4440 	dprintk("NFS call  lookup %pd2\n", dentry);
4441 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4442 	status = nfs4_do_call_sync(clnt, server, &msg,
4443 			&args.seq_args, &res.seq_res, task_flags);
4444 	dprintk("NFS reply lookup: %d\n", status);
4445 	return status;
4446 }
4447 
4448 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4449 {
4450 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4451 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4452 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4453 	fattr->nlink = 2;
4454 }
4455 
4456 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4457 				   struct dentry *dentry, struct nfs_fh *fhandle,
4458 				   struct nfs_fattr *fattr)
4459 {
4460 	struct nfs4_exception exception = {
4461 		.interruptible = true,
4462 	};
4463 	struct rpc_clnt *client = *clnt;
4464 	const struct qstr *name = &dentry->d_name;
4465 	int err;
4466 	do {
4467 		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4468 		trace_nfs4_lookup(dir, name, err);
4469 		switch (err) {
4470 		case -NFS4ERR_BADNAME:
4471 			err = -ENOENT;
4472 			goto out;
4473 		case -NFS4ERR_MOVED:
4474 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4475 			if (err == -NFS4ERR_MOVED)
4476 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4477 			goto out;
4478 		case -NFS4ERR_WRONGSEC:
4479 			err = -EPERM;
4480 			if (client != *clnt)
4481 				goto out;
4482 			client = nfs4_negotiate_security(client, dir, name);
4483 			if (IS_ERR(client))
4484 				return PTR_ERR(client);
4485 
4486 			exception.retry = 1;
4487 			break;
4488 		default:
4489 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4490 		}
4491 	} while (exception.retry);
4492 
4493 out:
4494 	if (err == 0)
4495 		*clnt = client;
4496 	else if (client != *clnt)
4497 		rpc_shutdown_client(client);
4498 
4499 	return err;
4500 }
4501 
4502 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4503 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4504 {
4505 	int status;
4506 	struct rpc_clnt *client = NFS_CLIENT(dir);
4507 
4508 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4509 	if (client != NFS_CLIENT(dir)) {
4510 		rpc_shutdown_client(client);
4511 		nfs_fixup_secinfo_attributes(fattr);
4512 	}
4513 	return status;
4514 }
4515 
4516 struct rpc_clnt *
4517 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4518 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4519 {
4520 	struct rpc_clnt *client = NFS_CLIENT(dir);
4521 	int status;
4522 
4523 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4524 	if (status < 0)
4525 		return ERR_PTR(status);
4526 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4527 }
4528 
4529 static int _nfs4_proc_lookupp(struct inode *inode,
4530 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4531 {
4532 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4533 	struct nfs_server *server = NFS_SERVER(inode);
4534 	int		       status;
4535 	struct nfs4_lookupp_arg args = {
4536 		.bitmask = server->attr_bitmask,
4537 		.fh = NFS_FH(inode),
4538 	};
4539 	struct nfs4_lookupp_res res = {
4540 		.server = server,
4541 		.fattr = fattr,
4542 		.fh = fhandle,
4543 	};
4544 	struct rpc_message msg = {
4545 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4546 		.rpc_argp = &args,
4547 		.rpc_resp = &res,
4548 	};
4549 	unsigned short task_flags = 0;
4550 
4551 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4552 		task_flags |= RPC_TASK_TIMEOUT;
4553 
4554 	args.bitmask = nfs4_bitmask(server, fattr->label);
4555 
4556 	nfs_fattr_init(fattr);
4557 
4558 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4559 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4560 				&res.seq_res, task_flags);
4561 	dprintk("NFS reply lookupp: %d\n", status);
4562 	return status;
4563 }
4564 
4565 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4566 			     struct nfs_fattr *fattr)
4567 {
4568 	struct nfs4_exception exception = {
4569 		.interruptible = true,
4570 	};
4571 	int err;
4572 	do {
4573 		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4574 		trace_nfs4_lookupp(inode, err);
4575 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4576 				&exception);
4577 	} while (exception.retry);
4578 	return err;
4579 }
4580 
4581 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4582 			     const struct cred *cred)
4583 {
4584 	struct nfs_server *server = NFS_SERVER(inode);
4585 	struct nfs4_accessargs args = {
4586 		.fh = NFS_FH(inode),
4587 		.access = entry->mask,
4588 	};
4589 	struct nfs4_accessres res = {
4590 		.server = server,
4591 	};
4592 	struct rpc_message msg = {
4593 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4594 		.rpc_argp = &args,
4595 		.rpc_resp = &res,
4596 		.rpc_cred = cred,
4597 	};
4598 	int status = 0;
4599 
4600 	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4601 		res.fattr = nfs_alloc_fattr();
4602 		if (res.fattr == NULL)
4603 			return -ENOMEM;
4604 		args.bitmask = server->cache_consistency_bitmask;
4605 	}
4606 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4607 	if (!status) {
4608 		nfs_access_set_mask(entry, res.access);
4609 		if (res.fattr)
4610 			nfs_refresh_inode(inode, res.fattr);
4611 	}
4612 	nfs_free_fattr(res.fattr);
4613 	return status;
4614 }
4615 
4616 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4617 			    const struct cred *cred)
4618 {
4619 	struct nfs4_exception exception = {
4620 		.interruptible = true,
4621 	};
4622 	int err;
4623 	do {
4624 		err = _nfs4_proc_access(inode, entry, cred);
4625 		trace_nfs4_access(inode, err);
4626 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4627 				&exception);
4628 	} while (exception.retry);
4629 	return err;
4630 }
4631 
4632 /*
4633  * TODO: For the time being, we don't try to get any attributes
4634  * along with any of the zero-copy operations READ, READDIR,
4635  * READLINK, WRITE.
4636  *
4637  * In the case of the first three, we want to put the GETATTR
4638  * after the read-type operation -- this is because it is hard
4639  * to predict the length of a GETATTR response in v4, and thus
4640  * align the READ data correctly.  This means that the GETATTR
4641  * may end up partially falling into the page cache, and we should
4642  * shift it into the 'tail' of the xdr_buf before processing.
4643  * To do this efficiently, we need to know the total length
4644  * of data received, which doesn't seem to be available outside
4645  * of the RPC layer.
4646  *
4647  * In the case of WRITE, we also want to put the GETATTR after
4648  * the operation -- in this case because we want to make sure
4649  * we get the post-operation mtime and size.
4650  *
4651  * Both of these changes to the XDR layer would in fact be quite
4652  * minor, but I decided to leave them for a subsequent patch.
4653  */
4654 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4655 		unsigned int pgbase, unsigned int pglen)
4656 {
4657 	struct nfs4_readlink args = {
4658 		.fh       = NFS_FH(inode),
4659 		.pgbase	  = pgbase,
4660 		.pglen    = pglen,
4661 		.pages    = &page,
4662 	};
4663 	struct nfs4_readlink_res res;
4664 	struct rpc_message msg = {
4665 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4666 		.rpc_argp = &args,
4667 		.rpc_resp = &res,
4668 	};
4669 
4670 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4671 }
4672 
4673 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4674 		unsigned int pgbase, unsigned int pglen)
4675 {
4676 	struct nfs4_exception exception = {
4677 		.interruptible = true,
4678 	};
4679 	int err;
4680 	do {
4681 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4682 		trace_nfs4_readlink(inode, err);
4683 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4684 				&exception);
4685 	} while (exception.retry);
4686 	return err;
4687 }
4688 
4689 /*
4690  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4691  */
4692 static int
4693 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4694 		 int flags)
4695 {
4696 	struct nfs_server *server = NFS_SERVER(dir);
4697 	struct nfs4_label l, *ilabel;
4698 	struct nfs_open_context *ctx;
4699 	struct nfs4_state *state;
4700 	int status = 0;
4701 
4702 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4703 	if (IS_ERR(ctx))
4704 		return PTR_ERR(ctx);
4705 
4706 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4707 
4708 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4709 		sattr->ia_mode &= ~current_umask();
4710 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4711 	if (IS_ERR(state)) {
4712 		status = PTR_ERR(state);
4713 		goto out;
4714 	}
4715 out:
4716 	nfs4_label_release_security(ilabel);
4717 	put_nfs_open_context(ctx);
4718 	return status;
4719 }
4720 
4721 static int
4722 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4723 {
4724 	struct nfs_server *server = NFS_SERVER(dir);
4725 	struct nfs_removeargs args = {
4726 		.fh = NFS_FH(dir),
4727 		.name = *name,
4728 	};
4729 	struct nfs_removeres res = {
4730 		.server = server,
4731 	};
4732 	struct rpc_message msg = {
4733 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4734 		.rpc_argp = &args,
4735 		.rpc_resp = &res,
4736 	};
4737 	unsigned long timestamp = jiffies;
4738 	int status;
4739 
4740 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4741 	if (status == 0) {
4742 		spin_lock(&dir->i_lock);
4743 		/* Removing a directory decrements nlink in the parent */
4744 		if (ftype == NF4DIR && dir->i_nlink > 2)
4745 			nfs4_dec_nlink_locked(dir);
4746 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4747 					      NFS_INO_INVALID_DATA);
4748 		spin_unlock(&dir->i_lock);
4749 	}
4750 	return status;
4751 }
4752 
4753 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4754 {
4755 	struct nfs4_exception exception = {
4756 		.interruptible = true,
4757 	};
4758 	struct inode *inode = d_inode(dentry);
4759 	int err;
4760 
4761 	if (inode) {
4762 		if (inode->i_nlink == 1)
4763 			nfs4_inode_return_delegation(inode);
4764 		else
4765 			nfs4_inode_make_writeable(inode);
4766 	}
4767 	do {
4768 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4769 		trace_nfs4_remove(dir, &dentry->d_name, err);
4770 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4771 				&exception);
4772 	} while (exception.retry);
4773 	return err;
4774 }
4775 
4776 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4777 {
4778 	struct nfs4_exception exception = {
4779 		.interruptible = true,
4780 	};
4781 	int err;
4782 
4783 	do {
4784 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4785 		trace_nfs4_remove(dir, name, err);
4786 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4787 				&exception);
4788 	} while (exception.retry);
4789 	return err;
4790 }
4791 
4792 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4793 		struct dentry *dentry,
4794 		struct inode *inode)
4795 {
4796 	struct nfs_removeargs *args = msg->rpc_argp;
4797 	struct nfs_removeres *res = msg->rpc_resp;
4798 
4799 	res->server = NFS_SB(dentry->d_sb);
4800 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4801 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4802 
4803 	nfs_fattr_init(res->dir_attr);
4804 
4805 	if (inode) {
4806 		nfs4_inode_return_delegation(inode);
4807 		nfs_d_prune_case_insensitive_aliases(inode);
4808 	}
4809 }
4810 
4811 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4812 {
4813 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4814 			&data->args.seq_args,
4815 			&data->res.seq_res,
4816 			task);
4817 }
4818 
4819 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4820 {
4821 	struct nfs_unlinkdata *data = task->tk_calldata;
4822 	struct nfs_removeres *res = &data->res;
4823 
4824 	if (!nfs4_sequence_done(task, &res->seq_res))
4825 		return 0;
4826 	if (nfs4_async_handle_error(task, res->server, NULL,
4827 				    &data->timeout) == -EAGAIN)
4828 		return 0;
4829 	if (task->tk_status == 0)
4830 		nfs4_update_changeattr(dir, &res->cinfo,
4831 				res->dir_attr->time_start,
4832 				NFS_INO_INVALID_DATA);
4833 	return 1;
4834 }
4835 
4836 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4837 		struct dentry *old_dentry,
4838 		struct dentry *new_dentry)
4839 {
4840 	struct nfs_renameargs *arg = msg->rpc_argp;
4841 	struct nfs_renameres *res = msg->rpc_resp;
4842 	struct inode *old_inode = d_inode(old_dentry);
4843 	struct inode *new_inode = d_inode(new_dentry);
4844 
4845 	if (old_inode)
4846 		nfs4_inode_make_writeable(old_inode);
4847 	if (new_inode)
4848 		nfs4_inode_return_delegation(new_inode);
4849 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4850 	res->server = NFS_SB(old_dentry->d_sb);
4851 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4852 }
4853 
4854 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4855 {
4856 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4857 			&data->args.seq_args,
4858 			&data->res.seq_res,
4859 			task);
4860 }
4861 
4862 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4863 				 struct inode *new_dir)
4864 {
4865 	struct nfs_renamedata *data = task->tk_calldata;
4866 	struct nfs_renameres *res = &data->res;
4867 
4868 	if (!nfs4_sequence_done(task, &res->seq_res))
4869 		return 0;
4870 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4871 		return 0;
4872 
4873 	if (task->tk_status == 0) {
4874 		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4875 		if (new_dir != old_dir) {
4876 			/* Note: If we moved a directory, nlink will change */
4877 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4878 					res->old_fattr->time_start,
4879 					NFS_INO_INVALID_NLINK |
4880 					    NFS_INO_INVALID_DATA);
4881 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4882 					res->new_fattr->time_start,
4883 					NFS_INO_INVALID_NLINK |
4884 					    NFS_INO_INVALID_DATA);
4885 		} else
4886 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4887 					res->old_fattr->time_start,
4888 					NFS_INO_INVALID_DATA);
4889 	}
4890 	return 1;
4891 }
4892 
4893 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4894 {
4895 	struct nfs_server *server = NFS_SERVER(inode);
4896 	__u32 bitmask[NFS4_BITMASK_SZ];
4897 	struct nfs4_link_arg arg = {
4898 		.fh     = NFS_FH(inode),
4899 		.dir_fh = NFS_FH(dir),
4900 		.name   = name,
4901 		.bitmask = bitmask,
4902 	};
4903 	struct nfs4_link_res res = {
4904 		.server = server,
4905 	};
4906 	struct rpc_message msg = {
4907 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4908 		.rpc_argp = &arg,
4909 		.rpc_resp = &res,
4910 	};
4911 	int status = -ENOMEM;
4912 
4913 	res.fattr = nfs_alloc_fattr_with_label(server);
4914 	if (res.fattr == NULL)
4915 		goto out;
4916 
4917 	nfs4_inode_make_writeable(inode);
4918 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4919 				NFS_INO_INVALID_CHANGE);
4920 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4921 	if (!status) {
4922 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4923 				       NFS_INO_INVALID_DATA);
4924 		nfs4_inc_nlink(inode);
4925 		status = nfs_post_op_update_inode(inode, res.fattr);
4926 		if (!status)
4927 			nfs_setsecurity(inode, res.fattr);
4928 	}
4929 
4930 out:
4931 	nfs_free_fattr(res.fattr);
4932 	return status;
4933 }
4934 
4935 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4936 {
4937 	struct nfs4_exception exception = {
4938 		.interruptible = true,
4939 	};
4940 	int err;
4941 	do {
4942 		err = nfs4_handle_exception(NFS_SERVER(inode),
4943 				_nfs4_proc_link(inode, dir, name),
4944 				&exception);
4945 	} while (exception.retry);
4946 	return err;
4947 }
4948 
4949 struct nfs4_createdata {
4950 	struct rpc_message msg;
4951 	struct nfs4_create_arg arg;
4952 	struct nfs4_create_res res;
4953 	struct nfs_fh fh;
4954 	struct nfs_fattr fattr;
4955 };
4956 
4957 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4958 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4959 {
4960 	struct nfs4_createdata *data;
4961 
4962 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4963 	if (data != NULL) {
4964 		struct nfs_server *server = NFS_SERVER(dir);
4965 
4966 		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4967 		if (IS_ERR(data->fattr.label))
4968 			goto out_free;
4969 
4970 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4971 		data->msg.rpc_argp = &data->arg;
4972 		data->msg.rpc_resp = &data->res;
4973 		data->arg.dir_fh = NFS_FH(dir);
4974 		data->arg.server = server;
4975 		data->arg.name = name;
4976 		data->arg.attrs = sattr;
4977 		data->arg.ftype = ftype;
4978 		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4979 		data->arg.umask = current_umask();
4980 		data->res.server = server;
4981 		data->res.fh = &data->fh;
4982 		data->res.fattr = &data->fattr;
4983 		nfs_fattr_init(data->res.fattr);
4984 	}
4985 	return data;
4986 out_free:
4987 	kfree(data);
4988 	return NULL;
4989 }
4990 
4991 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4992 {
4993 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4994 				    &data->arg.seq_args, &data->res.seq_res, 1);
4995 	if (status == 0) {
4996 		spin_lock(&dir->i_lock);
4997 		/* Creating a directory bumps nlink in the parent */
4998 		if (data->arg.ftype == NF4DIR)
4999 			nfs4_inc_nlink_locked(dir);
5000 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
5001 					      data->res.fattr->time_start,
5002 					      NFS_INO_INVALID_DATA);
5003 		spin_unlock(&dir->i_lock);
5004 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
5005 	}
5006 	return status;
5007 }
5008 
5009 static void nfs4_free_createdata(struct nfs4_createdata *data)
5010 {
5011 	nfs4_label_free(data->fattr.label);
5012 	kfree(data);
5013 }
5014 
5015 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5016 		struct page *page, unsigned int len, struct iattr *sattr,
5017 		struct nfs4_label *label)
5018 {
5019 	struct nfs4_createdata *data;
5020 	int status = -ENAMETOOLONG;
5021 
5022 	if (len > NFS4_MAXPATHLEN)
5023 		goto out;
5024 
5025 	status = -ENOMEM;
5026 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5027 	if (data == NULL)
5028 		goto out;
5029 
5030 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5031 	data->arg.u.symlink.pages = &page;
5032 	data->arg.u.symlink.len = len;
5033 	data->arg.label = label;
5034 
5035 	status = nfs4_do_create(dir, dentry, data);
5036 
5037 	nfs4_free_createdata(data);
5038 out:
5039 	return status;
5040 }
5041 
5042 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5043 		struct page *page, unsigned int len, struct iattr *sattr)
5044 {
5045 	struct nfs4_exception exception = {
5046 		.interruptible = true,
5047 	};
5048 	struct nfs4_label l, *label;
5049 	int err;
5050 
5051 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5052 
5053 	do {
5054 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5055 		trace_nfs4_symlink(dir, &dentry->d_name, err);
5056 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5057 				&exception);
5058 	} while (exception.retry);
5059 
5060 	nfs4_label_release_security(label);
5061 	return err;
5062 }
5063 
5064 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5065 		struct iattr *sattr, struct nfs4_label *label)
5066 {
5067 	struct nfs4_createdata *data;
5068 	int status = -ENOMEM;
5069 
5070 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5071 	if (data == NULL)
5072 		goto out;
5073 
5074 	data->arg.label = label;
5075 	status = nfs4_do_create(dir, dentry, data);
5076 
5077 	nfs4_free_createdata(data);
5078 out:
5079 	return status;
5080 }
5081 
5082 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5083 		struct iattr *sattr)
5084 {
5085 	struct nfs_server *server = NFS_SERVER(dir);
5086 	struct nfs4_exception exception = {
5087 		.interruptible = true,
5088 	};
5089 	struct nfs4_label l, *label;
5090 	int err;
5091 
5092 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5093 
5094 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5095 		sattr->ia_mode &= ~current_umask();
5096 	do {
5097 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5098 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5099 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5100 				&exception);
5101 	} while (exception.retry);
5102 	nfs4_label_release_security(label);
5103 
5104 	return err;
5105 }
5106 
5107 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5108 			      struct nfs_readdir_res *nr_res)
5109 {
5110 	struct inode		*dir = d_inode(nr_arg->dentry);
5111 	struct nfs_server	*server = NFS_SERVER(dir);
5112 	struct nfs4_readdir_arg args = {
5113 		.fh = NFS_FH(dir),
5114 		.pages = nr_arg->pages,
5115 		.pgbase = 0,
5116 		.count = nr_arg->page_len,
5117 		.plus = nr_arg->plus,
5118 	};
5119 	struct nfs4_readdir_res res;
5120 	struct rpc_message msg = {
5121 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5122 		.rpc_argp = &args,
5123 		.rpc_resp = &res,
5124 		.rpc_cred = nr_arg->cred,
5125 	};
5126 	int			status;
5127 
5128 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5129 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5130 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5131 		args.bitmask = server->attr_bitmask_nl;
5132 	else
5133 		args.bitmask = server->attr_bitmask;
5134 
5135 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5136 	res.pgbase = args.pgbase;
5137 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5138 			&res.seq_res, 0);
5139 	if (status >= 0) {
5140 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5141 		status += args.pgbase;
5142 	}
5143 
5144 	nfs_invalidate_atime(dir);
5145 
5146 	dprintk("%s: returns %d\n", __func__, status);
5147 	return status;
5148 }
5149 
5150 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5151 			     struct nfs_readdir_res *res)
5152 {
5153 	struct nfs4_exception exception = {
5154 		.interruptible = true,
5155 	};
5156 	int err;
5157 	do {
5158 		err = _nfs4_proc_readdir(arg, res);
5159 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5160 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5161 					    err, &exception);
5162 	} while (exception.retry);
5163 	return err;
5164 }
5165 
5166 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5167 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5168 {
5169 	struct nfs4_createdata *data;
5170 	int mode = sattr->ia_mode;
5171 	int status = -ENOMEM;
5172 
5173 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5174 	if (data == NULL)
5175 		goto out;
5176 
5177 	if (S_ISFIFO(mode))
5178 		data->arg.ftype = NF4FIFO;
5179 	else if (S_ISBLK(mode)) {
5180 		data->arg.ftype = NF4BLK;
5181 		data->arg.u.device.specdata1 = MAJOR(rdev);
5182 		data->arg.u.device.specdata2 = MINOR(rdev);
5183 	}
5184 	else if (S_ISCHR(mode)) {
5185 		data->arg.ftype = NF4CHR;
5186 		data->arg.u.device.specdata1 = MAJOR(rdev);
5187 		data->arg.u.device.specdata2 = MINOR(rdev);
5188 	} else if (!S_ISSOCK(mode)) {
5189 		status = -EINVAL;
5190 		goto out_free;
5191 	}
5192 
5193 	data->arg.label = label;
5194 	status = nfs4_do_create(dir, dentry, data);
5195 out_free:
5196 	nfs4_free_createdata(data);
5197 out:
5198 	return status;
5199 }
5200 
5201 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5202 		struct iattr *sattr, dev_t rdev)
5203 {
5204 	struct nfs_server *server = NFS_SERVER(dir);
5205 	struct nfs4_exception exception = {
5206 		.interruptible = true,
5207 	};
5208 	struct nfs4_label l, *label;
5209 	int err;
5210 
5211 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5212 
5213 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5214 		sattr->ia_mode &= ~current_umask();
5215 	do {
5216 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5217 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5218 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5219 				&exception);
5220 	} while (exception.retry);
5221 
5222 	nfs4_label_release_security(label);
5223 
5224 	return err;
5225 }
5226 
5227 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5228 		 struct nfs_fsstat *fsstat)
5229 {
5230 	struct nfs4_statfs_arg args = {
5231 		.fh = fhandle,
5232 		.bitmask = server->attr_bitmask,
5233 	};
5234 	struct nfs4_statfs_res res = {
5235 		.fsstat = fsstat,
5236 	};
5237 	struct rpc_message msg = {
5238 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5239 		.rpc_argp = &args,
5240 		.rpc_resp = &res,
5241 	};
5242 
5243 	nfs_fattr_init(fsstat->fattr);
5244 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5245 }
5246 
5247 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5248 {
5249 	struct nfs4_exception exception = {
5250 		.interruptible = true,
5251 	};
5252 	int err;
5253 	do {
5254 		err = nfs4_handle_exception(server,
5255 				_nfs4_proc_statfs(server, fhandle, fsstat),
5256 				&exception);
5257 	} while (exception.retry);
5258 	return err;
5259 }
5260 
5261 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5262 		struct nfs_fsinfo *fsinfo)
5263 {
5264 	struct nfs4_fsinfo_arg args = {
5265 		.fh = fhandle,
5266 		.bitmask = server->attr_bitmask,
5267 	};
5268 	struct nfs4_fsinfo_res res = {
5269 		.fsinfo = fsinfo,
5270 	};
5271 	struct rpc_message msg = {
5272 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5273 		.rpc_argp = &args,
5274 		.rpc_resp = &res,
5275 	};
5276 
5277 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5278 }
5279 
5280 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5281 {
5282 	struct nfs4_exception exception = {
5283 		.interruptible = true,
5284 	};
5285 	int err;
5286 
5287 	do {
5288 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5289 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5290 		if (err == 0) {
5291 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5292 			break;
5293 		}
5294 		err = nfs4_handle_exception(server, err, &exception);
5295 	} while (exception.retry);
5296 	return err;
5297 }
5298 
5299 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5300 {
5301 	int error;
5302 
5303 	nfs_fattr_init(fsinfo->fattr);
5304 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5305 	if (error == 0) {
5306 		/* block layout checks this! */
5307 		server->pnfs_blksize = fsinfo->blksize;
5308 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5309 	}
5310 
5311 	return error;
5312 }
5313 
5314 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5315 		struct nfs_pathconf *pathconf)
5316 {
5317 	struct nfs4_pathconf_arg args = {
5318 		.fh = fhandle,
5319 		.bitmask = server->attr_bitmask,
5320 	};
5321 	struct nfs4_pathconf_res res = {
5322 		.pathconf = pathconf,
5323 	};
5324 	struct rpc_message msg = {
5325 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5326 		.rpc_argp = &args,
5327 		.rpc_resp = &res,
5328 	};
5329 
5330 	/* None of the pathconf attributes are mandatory to implement */
5331 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5332 		memset(pathconf, 0, sizeof(*pathconf));
5333 		return 0;
5334 	}
5335 
5336 	nfs_fattr_init(pathconf->fattr);
5337 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5338 }
5339 
5340 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5341 		struct nfs_pathconf *pathconf)
5342 {
5343 	struct nfs4_exception exception = {
5344 		.interruptible = true,
5345 	};
5346 	int err;
5347 
5348 	do {
5349 		err = nfs4_handle_exception(server,
5350 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5351 				&exception);
5352 	} while (exception.retry);
5353 	return err;
5354 }
5355 
5356 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5357 		const struct nfs_open_context *ctx,
5358 		const struct nfs_lock_context *l_ctx,
5359 		fmode_t fmode)
5360 {
5361 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5362 }
5363 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5364 
5365 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5366 		const struct nfs_open_context *ctx,
5367 		const struct nfs_lock_context *l_ctx,
5368 		fmode_t fmode)
5369 {
5370 	nfs4_stateid _current_stateid;
5371 
5372 	/* If the current stateid represents a lost lock, then exit */
5373 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5374 		return true;
5375 	return nfs4_stateid_match(stateid, &_current_stateid);
5376 }
5377 
5378 static bool nfs4_error_stateid_expired(int err)
5379 {
5380 	switch (err) {
5381 	case -NFS4ERR_DELEG_REVOKED:
5382 	case -NFS4ERR_ADMIN_REVOKED:
5383 	case -NFS4ERR_BAD_STATEID:
5384 	case -NFS4ERR_STALE_STATEID:
5385 	case -NFS4ERR_OLD_STATEID:
5386 	case -NFS4ERR_OPENMODE:
5387 	case -NFS4ERR_EXPIRED:
5388 		return true;
5389 	}
5390 	return false;
5391 }
5392 
5393 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5394 {
5395 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5396 
5397 	trace_nfs4_read(hdr, task->tk_status);
5398 	if (task->tk_status < 0) {
5399 		struct nfs4_exception exception = {
5400 			.inode = hdr->inode,
5401 			.state = hdr->args.context->state,
5402 			.stateid = &hdr->args.stateid,
5403 		};
5404 		task->tk_status = nfs4_async_handle_exception(task,
5405 				server, task->tk_status, &exception);
5406 		if (exception.retry) {
5407 			rpc_restart_call_prepare(task);
5408 			return -EAGAIN;
5409 		}
5410 	}
5411 
5412 	if (task->tk_status > 0)
5413 		renew_lease(server, hdr->timestamp);
5414 	return 0;
5415 }
5416 
5417 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5418 		struct nfs_pgio_args *args)
5419 {
5420 
5421 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5422 		nfs4_stateid_is_current(&args->stateid,
5423 				args->context,
5424 				args->lock_context,
5425 				FMODE_READ))
5426 		return false;
5427 	rpc_restart_call_prepare(task);
5428 	return true;
5429 }
5430 
5431 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5432 					 struct nfs_pgio_header *hdr)
5433 {
5434 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5435 	struct rpc_message *msg = &task->tk_msg;
5436 
5437 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5438 	    task->tk_status == -ENOTSUPP) {
5439 		server->caps &= ~NFS_CAP_READ_PLUS;
5440 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5441 		rpc_restart_call_prepare(task);
5442 		return true;
5443 	}
5444 	return false;
5445 }
5446 
5447 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5448 {
5449 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5450 		return -EAGAIN;
5451 	if (nfs4_read_stateid_changed(task, &hdr->args))
5452 		return -EAGAIN;
5453 	if (nfs4_read_plus_not_supported(task, hdr))
5454 		return -EAGAIN;
5455 	if (task->tk_status > 0)
5456 		nfs_invalidate_atime(hdr->inode);
5457 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5458 				    nfs4_read_done_cb(task, hdr);
5459 }
5460 
5461 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5462 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5463 				    struct rpc_message *msg)
5464 {
5465 	/* Note: We don't use READ_PLUS with pNFS yet */
5466 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5467 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5468 		return nfs_read_alloc_scratch(hdr, READ_PLUS_SCRATCH_SIZE);
5469 	}
5470 	return false;
5471 }
5472 #else
5473 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5474 				    struct rpc_message *msg)
5475 {
5476 	return false;
5477 }
5478 #endif /* CONFIG_NFS_V4_2 */
5479 
5480 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5481 				 struct rpc_message *msg)
5482 {
5483 	hdr->timestamp   = jiffies;
5484 	if (!hdr->pgio_done_cb)
5485 		hdr->pgio_done_cb = nfs4_read_done_cb;
5486 	if (!nfs42_read_plus_support(hdr, msg))
5487 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5488 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5489 }
5490 
5491 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5492 				      struct nfs_pgio_header *hdr)
5493 {
5494 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5495 			&hdr->args.seq_args,
5496 			&hdr->res.seq_res,
5497 			task))
5498 		return 0;
5499 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5500 				hdr->args.lock_context,
5501 				hdr->rw_mode) == -EIO)
5502 		return -EIO;
5503 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5504 		return -EIO;
5505 	return 0;
5506 }
5507 
5508 static int nfs4_write_done_cb(struct rpc_task *task,
5509 			      struct nfs_pgio_header *hdr)
5510 {
5511 	struct inode *inode = hdr->inode;
5512 
5513 	trace_nfs4_write(hdr, task->tk_status);
5514 	if (task->tk_status < 0) {
5515 		struct nfs4_exception exception = {
5516 			.inode = hdr->inode,
5517 			.state = hdr->args.context->state,
5518 			.stateid = &hdr->args.stateid,
5519 		};
5520 		task->tk_status = nfs4_async_handle_exception(task,
5521 				NFS_SERVER(inode), task->tk_status,
5522 				&exception);
5523 		if (exception.retry) {
5524 			rpc_restart_call_prepare(task);
5525 			return -EAGAIN;
5526 		}
5527 	}
5528 	if (task->tk_status >= 0) {
5529 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5530 		nfs_writeback_update_inode(hdr);
5531 	}
5532 	return 0;
5533 }
5534 
5535 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5536 		struct nfs_pgio_args *args)
5537 {
5538 
5539 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5540 		nfs4_stateid_is_current(&args->stateid,
5541 				args->context,
5542 				args->lock_context,
5543 				FMODE_WRITE))
5544 		return false;
5545 	rpc_restart_call_prepare(task);
5546 	return true;
5547 }
5548 
5549 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5550 {
5551 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5552 		return -EAGAIN;
5553 	if (nfs4_write_stateid_changed(task, &hdr->args))
5554 		return -EAGAIN;
5555 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5556 		nfs4_write_done_cb(task, hdr);
5557 }
5558 
5559 static
5560 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5561 {
5562 	/* Don't request attributes for pNFS or O_DIRECT writes */
5563 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5564 		return false;
5565 	/* Otherwise, request attributes if and only if we don't hold
5566 	 * a delegation
5567 	 */
5568 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5569 }
5570 
5571 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5572 		      struct inode *inode, unsigned long cache_validity)
5573 {
5574 	struct nfs_server *server = NFS_SERVER(inode);
5575 	unsigned int i;
5576 
5577 	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5578 	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5579 
5580 	if (cache_validity & NFS_INO_INVALID_CHANGE)
5581 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5582 	if (cache_validity & NFS_INO_INVALID_ATIME)
5583 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5584 	if (cache_validity & NFS_INO_INVALID_MODE)
5585 		bitmask[1] |= FATTR4_WORD1_MODE;
5586 	if (cache_validity & NFS_INO_INVALID_OTHER)
5587 		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5588 	if (cache_validity & NFS_INO_INVALID_NLINK)
5589 		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5590 	if (cache_validity & NFS_INO_INVALID_CTIME)
5591 		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5592 	if (cache_validity & NFS_INO_INVALID_MTIME)
5593 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5594 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5595 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5596 
5597 	if (cache_validity & NFS_INO_INVALID_SIZE)
5598 		bitmask[0] |= FATTR4_WORD0_SIZE;
5599 
5600 	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5601 		bitmask[i] &= server->attr_bitmask[i];
5602 }
5603 
5604 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5605 				  struct rpc_message *msg,
5606 				  struct rpc_clnt **clnt)
5607 {
5608 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5609 
5610 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5611 		hdr->args.bitmask = NULL;
5612 		hdr->res.fattr = NULL;
5613 	} else {
5614 		nfs4_bitmask_set(hdr->args.bitmask_store,
5615 				 server->cache_consistency_bitmask,
5616 				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5617 		hdr->args.bitmask = hdr->args.bitmask_store;
5618 	}
5619 
5620 	if (!hdr->pgio_done_cb)
5621 		hdr->pgio_done_cb = nfs4_write_done_cb;
5622 	hdr->res.server = server;
5623 	hdr->timestamp   = jiffies;
5624 
5625 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5626 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5627 	nfs4_state_protect_write(hdr->ds_clp ? hdr->ds_clp : server->nfs_client, clnt, msg, hdr);
5628 }
5629 
5630 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5631 {
5632 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5633 			&data->args.seq_args,
5634 			&data->res.seq_res,
5635 			task);
5636 }
5637 
5638 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5639 {
5640 	struct inode *inode = data->inode;
5641 
5642 	trace_nfs4_commit(data, task->tk_status);
5643 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5644 				    NULL, NULL) == -EAGAIN) {
5645 		rpc_restart_call_prepare(task);
5646 		return -EAGAIN;
5647 	}
5648 	return 0;
5649 }
5650 
5651 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5652 {
5653 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5654 		return -EAGAIN;
5655 	return data->commit_done_cb(task, data);
5656 }
5657 
5658 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5659 				   struct rpc_clnt **clnt)
5660 {
5661 	struct nfs_server *server = NFS_SERVER(data->inode);
5662 
5663 	if (data->commit_done_cb == NULL)
5664 		data->commit_done_cb = nfs4_commit_done_cb;
5665 	data->res.server = server;
5666 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5667 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5668 	nfs4_state_protect(data->ds_clp ? data->ds_clp : server->nfs_client,
5669 			NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5670 }
5671 
5672 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5673 				struct nfs_commitres *res)
5674 {
5675 	struct inode *dst_inode = file_inode(dst);
5676 	struct nfs_server *server = NFS_SERVER(dst_inode);
5677 	struct rpc_message msg = {
5678 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5679 		.rpc_argp = args,
5680 		.rpc_resp = res,
5681 	};
5682 
5683 	args->fh = NFS_FH(dst_inode);
5684 	return nfs4_call_sync(server->client, server, &msg,
5685 			&args->seq_args, &res->seq_res, 1);
5686 }
5687 
5688 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5689 {
5690 	struct nfs_commitargs args = {
5691 		.offset = offset,
5692 		.count = count,
5693 	};
5694 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5695 	struct nfs4_exception exception = { };
5696 	int status;
5697 
5698 	do {
5699 		status = _nfs4_proc_commit(dst, &args, res);
5700 		status = nfs4_handle_exception(dst_server, status, &exception);
5701 	} while (exception.retry);
5702 
5703 	return status;
5704 }
5705 
5706 struct nfs4_renewdata {
5707 	struct nfs_client	*client;
5708 	unsigned long		timestamp;
5709 };
5710 
5711 /*
5712  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5713  * standalone procedure for queueing an asynchronous RENEW.
5714  */
5715 static void nfs4_renew_release(void *calldata)
5716 {
5717 	struct nfs4_renewdata *data = calldata;
5718 	struct nfs_client *clp = data->client;
5719 
5720 	if (refcount_read(&clp->cl_count) > 1)
5721 		nfs4_schedule_state_renewal(clp);
5722 	nfs_put_client(clp);
5723 	kfree(data);
5724 }
5725 
5726 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5727 {
5728 	struct nfs4_renewdata *data = calldata;
5729 	struct nfs_client *clp = data->client;
5730 	unsigned long timestamp = data->timestamp;
5731 
5732 	trace_nfs4_renew_async(clp, task->tk_status);
5733 	switch (task->tk_status) {
5734 	case 0:
5735 		break;
5736 	case -NFS4ERR_LEASE_MOVED:
5737 		nfs4_schedule_lease_moved_recovery(clp);
5738 		break;
5739 	default:
5740 		/* Unless we're shutting down, schedule state recovery! */
5741 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5742 			return;
5743 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5744 			nfs4_schedule_lease_recovery(clp);
5745 			return;
5746 		}
5747 		nfs4_schedule_path_down_recovery(clp);
5748 	}
5749 	do_renew_lease(clp, timestamp);
5750 }
5751 
5752 static const struct rpc_call_ops nfs4_renew_ops = {
5753 	.rpc_call_done = nfs4_renew_done,
5754 	.rpc_release = nfs4_renew_release,
5755 };
5756 
5757 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5758 {
5759 	struct rpc_message msg = {
5760 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5761 		.rpc_argp	= clp,
5762 		.rpc_cred	= cred,
5763 	};
5764 	struct nfs4_renewdata *data;
5765 
5766 	if (renew_flags == 0)
5767 		return 0;
5768 	if (!refcount_inc_not_zero(&clp->cl_count))
5769 		return -EIO;
5770 	data = kmalloc(sizeof(*data), GFP_NOFS);
5771 	if (data == NULL) {
5772 		nfs_put_client(clp);
5773 		return -ENOMEM;
5774 	}
5775 	data->client = clp;
5776 	data->timestamp = jiffies;
5777 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5778 			&nfs4_renew_ops, data);
5779 }
5780 
5781 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5782 {
5783 	struct rpc_message msg = {
5784 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5785 		.rpc_argp	= clp,
5786 		.rpc_cred	= cred,
5787 	};
5788 	unsigned long now = jiffies;
5789 	int status;
5790 
5791 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5792 	if (status < 0)
5793 		return status;
5794 	do_renew_lease(clp, now);
5795 	return 0;
5796 }
5797 
5798 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5799 				      enum nfs4_acl_type type)
5800 {
5801 	switch (type) {
5802 	default:
5803 		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5804 	case NFS4ACL_DACL:
5805 		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5806 	case NFS4ACL_SACL:
5807 		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5808 	}
5809 }
5810 
5811 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5812  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5813  * the stack.
5814  */
5815 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5816 
5817 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5818 		struct page **pages)
5819 {
5820 	struct page *newpage, **spages;
5821 	int rc = 0;
5822 	size_t len;
5823 	spages = pages;
5824 
5825 	do {
5826 		len = min_t(size_t, PAGE_SIZE, buflen);
5827 		newpage = alloc_page(GFP_KERNEL);
5828 
5829 		if (newpage == NULL)
5830 			goto unwind;
5831 		memcpy(page_address(newpage), buf, len);
5832 		buf += len;
5833 		buflen -= len;
5834 		*pages++ = newpage;
5835 		rc++;
5836 	} while (buflen != 0);
5837 
5838 	return rc;
5839 
5840 unwind:
5841 	for(; rc > 0; rc--)
5842 		__free_page(spages[rc-1]);
5843 	return -ENOMEM;
5844 }
5845 
5846 struct nfs4_cached_acl {
5847 	enum nfs4_acl_type type;
5848 	int cached;
5849 	size_t len;
5850 	char data[];
5851 };
5852 
5853 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5854 {
5855 	struct nfs_inode *nfsi = NFS_I(inode);
5856 
5857 	spin_lock(&inode->i_lock);
5858 	kfree(nfsi->nfs4_acl);
5859 	nfsi->nfs4_acl = acl;
5860 	spin_unlock(&inode->i_lock);
5861 }
5862 
5863 static void nfs4_zap_acl_attr(struct inode *inode)
5864 {
5865 	nfs4_set_cached_acl(inode, NULL);
5866 }
5867 
5868 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5869 				    size_t buflen, enum nfs4_acl_type type)
5870 {
5871 	struct nfs_inode *nfsi = NFS_I(inode);
5872 	struct nfs4_cached_acl *acl;
5873 	int ret = -ENOENT;
5874 
5875 	spin_lock(&inode->i_lock);
5876 	acl = nfsi->nfs4_acl;
5877 	if (acl == NULL)
5878 		goto out;
5879 	if (acl->type != type)
5880 		goto out;
5881 	if (buf == NULL) /* user is just asking for length */
5882 		goto out_len;
5883 	if (acl->cached == 0)
5884 		goto out;
5885 	ret = -ERANGE; /* see getxattr(2) man page */
5886 	if (acl->len > buflen)
5887 		goto out;
5888 	memcpy(buf, acl->data, acl->len);
5889 out_len:
5890 	ret = acl->len;
5891 out:
5892 	spin_unlock(&inode->i_lock);
5893 	return ret;
5894 }
5895 
5896 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5897 				  size_t pgbase, size_t acl_len,
5898 				  enum nfs4_acl_type type)
5899 {
5900 	struct nfs4_cached_acl *acl;
5901 	size_t buflen = sizeof(*acl) + acl_len;
5902 
5903 	if (buflen <= PAGE_SIZE) {
5904 		acl = kmalloc(buflen, GFP_KERNEL);
5905 		if (acl == NULL)
5906 			goto out;
5907 		acl->cached = 1;
5908 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5909 	} else {
5910 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5911 		if (acl == NULL)
5912 			goto out;
5913 		acl->cached = 0;
5914 	}
5915 	acl->type = type;
5916 	acl->len = acl_len;
5917 out:
5918 	nfs4_set_cached_acl(inode, acl);
5919 }
5920 
5921 /*
5922  * The getxattr API returns the required buffer length when called with a
5923  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5924  * the required buf.  On a NULL buf, we send a page of data to the server
5925  * guessing that the ACL request can be serviced by a page. If so, we cache
5926  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5927  * the cache. If not so, we throw away the page, and cache the required
5928  * length. The next getxattr call will then produce another round trip to
5929  * the server, this time with the input buf of the required size.
5930  */
5931 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5932 				       size_t buflen, enum nfs4_acl_type type)
5933 {
5934 	struct page **pages;
5935 	struct nfs_getaclargs args = {
5936 		.fh = NFS_FH(inode),
5937 		.acl_type = type,
5938 		.acl_len = buflen,
5939 	};
5940 	struct nfs_getaclres res = {
5941 		.acl_type = type,
5942 		.acl_len = buflen,
5943 	};
5944 	struct rpc_message msg = {
5945 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5946 		.rpc_argp = &args,
5947 		.rpc_resp = &res,
5948 	};
5949 	unsigned int npages;
5950 	int ret = -ENOMEM, i;
5951 	struct nfs_server *server = NFS_SERVER(inode);
5952 
5953 	if (buflen == 0)
5954 		buflen = server->rsize;
5955 
5956 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5957 	pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5958 	if (!pages)
5959 		return -ENOMEM;
5960 
5961 	args.acl_pages = pages;
5962 
5963 	for (i = 0; i < npages; i++) {
5964 		pages[i] = alloc_page(GFP_KERNEL);
5965 		if (!pages[i])
5966 			goto out_free;
5967 	}
5968 
5969 	/* for decoding across pages */
5970 	res.acl_scratch = alloc_page(GFP_KERNEL);
5971 	if (!res.acl_scratch)
5972 		goto out_free;
5973 
5974 	args.acl_len = npages * PAGE_SIZE;
5975 
5976 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5977 		__func__, buf, buflen, npages, args.acl_len);
5978 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5979 			     &msg, &args.seq_args, &res.seq_res, 0);
5980 	if (ret)
5981 		goto out_free;
5982 
5983 	/* Handle the case where the passed-in buffer is too short */
5984 	if (res.acl_flags & NFS4_ACL_TRUNC) {
5985 		/* Did the user only issue a request for the acl length? */
5986 		if (buf == NULL)
5987 			goto out_ok;
5988 		ret = -ERANGE;
5989 		goto out_free;
5990 	}
5991 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5992 			      type);
5993 	if (buf) {
5994 		if (res.acl_len > buflen) {
5995 			ret = -ERANGE;
5996 			goto out_free;
5997 		}
5998 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5999 	}
6000 out_ok:
6001 	ret = res.acl_len;
6002 out_free:
6003 	while (--i >= 0)
6004 		__free_page(pages[i]);
6005 	if (res.acl_scratch)
6006 		__free_page(res.acl_scratch);
6007 	kfree(pages);
6008 	return ret;
6009 }
6010 
6011 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6012 				     size_t buflen, enum nfs4_acl_type type)
6013 {
6014 	struct nfs4_exception exception = {
6015 		.interruptible = true,
6016 	};
6017 	ssize_t ret;
6018 	do {
6019 		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6020 		trace_nfs4_get_acl(inode, ret);
6021 		if (ret >= 0)
6022 			break;
6023 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6024 	} while (exception.retry);
6025 	return ret;
6026 }
6027 
6028 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6029 				 enum nfs4_acl_type type)
6030 {
6031 	struct nfs_server *server = NFS_SERVER(inode);
6032 	int ret;
6033 
6034 	if (!nfs4_server_supports_acls(server, type))
6035 		return -EOPNOTSUPP;
6036 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6037 	if (ret < 0)
6038 		return ret;
6039 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6040 		nfs_zap_acl_cache(inode);
6041 	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6042 	if (ret != -ENOENT)
6043 		/* -ENOENT is returned if there is no ACL or if there is an ACL
6044 		 * but no cached acl data, just the acl length */
6045 		return ret;
6046 	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6047 }
6048 
6049 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6050 			       size_t buflen, enum nfs4_acl_type type)
6051 {
6052 	struct nfs_server *server = NFS_SERVER(inode);
6053 	struct page *pages[NFS4ACL_MAXPAGES];
6054 	struct nfs_setaclargs arg = {
6055 		.fh = NFS_FH(inode),
6056 		.acl_type = type,
6057 		.acl_len = buflen,
6058 		.acl_pages = pages,
6059 	};
6060 	struct nfs_setaclres res;
6061 	struct rpc_message msg = {
6062 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6063 		.rpc_argp	= &arg,
6064 		.rpc_resp	= &res,
6065 	};
6066 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6067 	int ret, i;
6068 
6069 	/* You can't remove system.nfs4_acl: */
6070 	if (buflen == 0)
6071 		return -EINVAL;
6072 	if (!nfs4_server_supports_acls(server, type))
6073 		return -EOPNOTSUPP;
6074 	if (npages > ARRAY_SIZE(pages))
6075 		return -ERANGE;
6076 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6077 	if (i < 0)
6078 		return i;
6079 	nfs4_inode_make_writeable(inode);
6080 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6081 
6082 	/*
6083 	 * Free each page after tx, so the only ref left is
6084 	 * held by the network stack
6085 	 */
6086 	for (; i > 0; i--)
6087 		put_page(pages[i-1]);
6088 
6089 	/*
6090 	 * Acl update can result in inode attribute update.
6091 	 * so mark the attribute cache invalid.
6092 	 */
6093 	spin_lock(&inode->i_lock);
6094 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6095 					     NFS_INO_INVALID_CTIME |
6096 					     NFS_INO_REVAL_FORCED);
6097 	spin_unlock(&inode->i_lock);
6098 	nfs_access_zap_cache(inode);
6099 	nfs_zap_acl_cache(inode);
6100 	return ret;
6101 }
6102 
6103 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6104 			     size_t buflen, enum nfs4_acl_type type)
6105 {
6106 	struct nfs4_exception exception = { };
6107 	int err;
6108 	do {
6109 		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6110 		trace_nfs4_set_acl(inode, err);
6111 		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6112 			/*
6113 			 * no need to retry since the kernel
6114 			 * isn't involved in encoding the ACEs.
6115 			 */
6116 			err = -EINVAL;
6117 			break;
6118 		}
6119 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6120 				&exception);
6121 	} while (exception.retry);
6122 	return err;
6123 }
6124 
6125 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6126 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6127 					size_t buflen)
6128 {
6129 	struct nfs_server *server = NFS_SERVER(inode);
6130 	struct nfs4_label label = {0, 0, buflen, buf};
6131 
6132 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6133 	struct nfs_fattr fattr = {
6134 		.label = &label,
6135 	};
6136 	struct nfs4_getattr_arg arg = {
6137 		.fh		= NFS_FH(inode),
6138 		.bitmask	= bitmask,
6139 	};
6140 	struct nfs4_getattr_res res = {
6141 		.fattr		= &fattr,
6142 		.server		= server,
6143 	};
6144 	struct rpc_message msg = {
6145 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6146 		.rpc_argp	= &arg,
6147 		.rpc_resp	= &res,
6148 	};
6149 	int ret;
6150 
6151 	nfs_fattr_init(&fattr);
6152 
6153 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6154 	if (ret)
6155 		return ret;
6156 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6157 		return -ENOENT;
6158 	return label.len;
6159 }
6160 
6161 static int nfs4_get_security_label(struct inode *inode, void *buf,
6162 					size_t buflen)
6163 {
6164 	struct nfs4_exception exception = {
6165 		.interruptible = true,
6166 	};
6167 	int err;
6168 
6169 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6170 		return -EOPNOTSUPP;
6171 
6172 	do {
6173 		err = _nfs4_get_security_label(inode, buf, buflen);
6174 		trace_nfs4_get_security_label(inode, err);
6175 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6176 				&exception);
6177 	} while (exception.retry);
6178 	return err;
6179 }
6180 
6181 static int _nfs4_do_set_security_label(struct inode *inode,
6182 		struct nfs4_label *ilabel,
6183 		struct nfs_fattr *fattr)
6184 {
6185 
6186 	struct iattr sattr = {0};
6187 	struct nfs_server *server = NFS_SERVER(inode);
6188 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6189 	struct nfs_setattrargs arg = {
6190 		.fh		= NFS_FH(inode),
6191 		.iap		= &sattr,
6192 		.server		= server,
6193 		.bitmask	= bitmask,
6194 		.label		= ilabel,
6195 	};
6196 	struct nfs_setattrres res = {
6197 		.fattr		= fattr,
6198 		.server		= server,
6199 	};
6200 	struct rpc_message msg = {
6201 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6202 		.rpc_argp	= &arg,
6203 		.rpc_resp	= &res,
6204 	};
6205 	int status;
6206 
6207 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6208 
6209 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6210 	if (status)
6211 		dprintk("%s failed: %d\n", __func__, status);
6212 
6213 	return status;
6214 }
6215 
6216 static int nfs4_do_set_security_label(struct inode *inode,
6217 		struct nfs4_label *ilabel,
6218 		struct nfs_fattr *fattr)
6219 {
6220 	struct nfs4_exception exception = { };
6221 	int err;
6222 
6223 	do {
6224 		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6225 		trace_nfs4_set_security_label(inode, err);
6226 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6227 				&exception);
6228 	} while (exception.retry);
6229 	return err;
6230 }
6231 
6232 static int
6233 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6234 {
6235 	struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6236 	struct nfs_fattr *fattr;
6237 	int status;
6238 
6239 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6240 		return -EOPNOTSUPP;
6241 
6242 	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6243 	if (fattr == NULL)
6244 		return -ENOMEM;
6245 
6246 	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6247 	if (status == 0)
6248 		nfs_setsecurity(inode, fattr);
6249 
6250 	return status;
6251 }
6252 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6253 
6254 
6255 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6256 				    nfs4_verifier *bootverf)
6257 {
6258 	__be32 verf[2];
6259 
6260 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6261 		/* An impossible timestamp guarantees this value
6262 		 * will never match a generated boot time. */
6263 		verf[0] = cpu_to_be32(U32_MAX);
6264 		verf[1] = cpu_to_be32(U32_MAX);
6265 	} else {
6266 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6267 		u64 ns = ktime_to_ns(nn->boot_time);
6268 
6269 		verf[0] = cpu_to_be32(ns >> 32);
6270 		verf[1] = cpu_to_be32(ns);
6271 	}
6272 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6273 }
6274 
6275 static size_t
6276 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6277 {
6278 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6279 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6280 	const char *id;
6281 
6282 	buf[0] = '\0';
6283 
6284 	if (nn_clp) {
6285 		rcu_read_lock();
6286 		id = rcu_dereference(nn_clp->identifier);
6287 		if (id)
6288 			strscpy(buf, id, buflen);
6289 		rcu_read_unlock();
6290 	}
6291 
6292 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6293 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6294 
6295 	return strlen(buf);
6296 }
6297 
6298 static int
6299 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6300 {
6301 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6302 	size_t buflen;
6303 	size_t len;
6304 	char *str;
6305 
6306 	if (clp->cl_owner_id != NULL)
6307 		return 0;
6308 
6309 	rcu_read_lock();
6310 	len = 14 +
6311 		strlen(clp->cl_rpcclient->cl_nodename) +
6312 		1 +
6313 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6314 		1;
6315 	rcu_read_unlock();
6316 
6317 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6318 	if (buflen)
6319 		len += buflen + 1;
6320 
6321 	if (len > NFS4_OPAQUE_LIMIT + 1)
6322 		return -EINVAL;
6323 
6324 	/*
6325 	 * Since this string is allocated at mount time, and held until the
6326 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6327 	 * about a memory-reclaim deadlock.
6328 	 */
6329 	str = kmalloc(len, GFP_KERNEL);
6330 	if (!str)
6331 		return -ENOMEM;
6332 
6333 	rcu_read_lock();
6334 	if (buflen)
6335 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6336 			  clp->cl_rpcclient->cl_nodename, buf,
6337 			  rpc_peeraddr2str(clp->cl_rpcclient,
6338 					   RPC_DISPLAY_ADDR));
6339 	else
6340 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6341 			  clp->cl_rpcclient->cl_nodename,
6342 			  rpc_peeraddr2str(clp->cl_rpcclient,
6343 					   RPC_DISPLAY_ADDR));
6344 	rcu_read_unlock();
6345 
6346 	clp->cl_owner_id = str;
6347 	return 0;
6348 }
6349 
6350 static int
6351 nfs4_init_uniform_client_string(struct nfs_client *clp)
6352 {
6353 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6354 	size_t buflen;
6355 	size_t len;
6356 	char *str;
6357 
6358 	if (clp->cl_owner_id != NULL)
6359 		return 0;
6360 
6361 	len = 10 + 10 + 1 + 10 + 1 +
6362 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6363 
6364 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6365 	if (buflen)
6366 		len += buflen + 1;
6367 
6368 	if (len > NFS4_OPAQUE_LIMIT + 1)
6369 		return -EINVAL;
6370 
6371 	/*
6372 	 * Since this string is allocated at mount time, and held until the
6373 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6374 	 * about a memory-reclaim deadlock.
6375 	 */
6376 	str = kmalloc(len, GFP_KERNEL);
6377 	if (!str)
6378 		return -ENOMEM;
6379 
6380 	if (buflen)
6381 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6382 			  clp->rpc_ops->version, clp->cl_minorversion,
6383 			  buf, clp->cl_rpcclient->cl_nodename);
6384 	else
6385 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6386 			  clp->rpc_ops->version, clp->cl_minorversion,
6387 			  clp->cl_rpcclient->cl_nodename);
6388 	clp->cl_owner_id = str;
6389 	return 0;
6390 }
6391 
6392 /*
6393  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6394  * services.  Advertise one based on the address family of the
6395  * clientaddr.
6396  */
6397 static unsigned int
6398 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6399 {
6400 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6401 		return scnprintf(buf, len, "tcp6");
6402 	else
6403 		return scnprintf(buf, len, "tcp");
6404 }
6405 
6406 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6407 {
6408 	struct nfs4_setclientid *sc = calldata;
6409 
6410 	if (task->tk_status == 0)
6411 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6412 }
6413 
6414 static const struct rpc_call_ops nfs4_setclientid_ops = {
6415 	.rpc_call_done = nfs4_setclientid_done,
6416 };
6417 
6418 /**
6419  * nfs4_proc_setclientid - Negotiate client ID
6420  * @clp: state data structure
6421  * @program: RPC program for NFSv4 callback service
6422  * @port: IP port number for NFS4 callback service
6423  * @cred: credential to use for this call
6424  * @res: where to place the result
6425  *
6426  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6427  */
6428 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6429 		unsigned short port, const struct cred *cred,
6430 		struct nfs4_setclientid_res *res)
6431 {
6432 	nfs4_verifier sc_verifier;
6433 	struct nfs4_setclientid setclientid = {
6434 		.sc_verifier = &sc_verifier,
6435 		.sc_prog = program,
6436 		.sc_clnt = clp,
6437 	};
6438 	struct rpc_message msg = {
6439 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6440 		.rpc_argp = &setclientid,
6441 		.rpc_resp = res,
6442 		.rpc_cred = cred,
6443 	};
6444 	struct rpc_task_setup task_setup_data = {
6445 		.rpc_client = clp->cl_rpcclient,
6446 		.rpc_message = &msg,
6447 		.callback_ops = &nfs4_setclientid_ops,
6448 		.callback_data = &setclientid,
6449 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6450 	};
6451 	unsigned long now = jiffies;
6452 	int status;
6453 
6454 	/* nfs_client_id4 */
6455 	nfs4_init_boot_verifier(clp, &sc_verifier);
6456 
6457 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6458 		status = nfs4_init_uniform_client_string(clp);
6459 	else
6460 		status = nfs4_init_nonuniform_client_string(clp);
6461 
6462 	if (status)
6463 		goto out;
6464 
6465 	/* cb_client4 */
6466 	setclientid.sc_netid_len =
6467 				nfs4_init_callback_netid(clp,
6468 						setclientid.sc_netid,
6469 						sizeof(setclientid.sc_netid));
6470 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6471 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6472 				clp->cl_ipaddr, port >> 8, port & 255);
6473 
6474 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6475 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6476 		clp->cl_owner_id);
6477 
6478 	status = nfs4_call_sync_custom(&task_setup_data);
6479 	if (setclientid.sc_cred) {
6480 		kfree(clp->cl_acceptor);
6481 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6482 		put_rpccred(setclientid.sc_cred);
6483 	}
6484 
6485 	if (status == 0)
6486 		do_renew_lease(clp, now);
6487 out:
6488 	trace_nfs4_setclientid(clp, status);
6489 	dprintk("NFS reply setclientid: %d\n", status);
6490 	return status;
6491 }
6492 
6493 /**
6494  * nfs4_proc_setclientid_confirm - Confirm client ID
6495  * @clp: state data structure
6496  * @arg: result of a previous SETCLIENTID
6497  * @cred: credential to use for this call
6498  *
6499  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6500  */
6501 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6502 		struct nfs4_setclientid_res *arg,
6503 		const struct cred *cred)
6504 {
6505 	struct rpc_message msg = {
6506 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6507 		.rpc_argp = arg,
6508 		.rpc_cred = cred,
6509 	};
6510 	int status;
6511 
6512 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6513 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6514 		clp->cl_clientid);
6515 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6516 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6517 	trace_nfs4_setclientid_confirm(clp, status);
6518 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6519 	return status;
6520 }
6521 
6522 struct nfs4_delegreturndata {
6523 	struct nfs4_delegreturnargs args;
6524 	struct nfs4_delegreturnres res;
6525 	struct nfs_fh fh;
6526 	nfs4_stateid stateid;
6527 	unsigned long timestamp;
6528 	struct {
6529 		struct nfs4_layoutreturn_args arg;
6530 		struct nfs4_layoutreturn_res res;
6531 		struct nfs4_xdr_opaque_data ld_private;
6532 		u32 roc_barrier;
6533 		bool roc;
6534 	} lr;
6535 	struct nfs_fattr fattr;
6536 	int rpc_status;
6537 	struct inode *inode;
6538 };
6539 
6540 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6541 {
6542 	struct nfs4_delegreturndata *data = calldata;
6543 	struct nfs4_exception exception = {
6544 		.inode = data->inode,
6545 		.stateid = &data->stateid,
6546 		.task_is_privileged = data->args.seq_args.sa_privileged,
6547 	};
6548 
6549 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6550 		return;
6551 
6552 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6553 
6554 	/* Handle Layoutreturn errors */
6555 	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6556 			  &data->res.lr_ret) == -EAGAIN)
6557 		goto out_restart;
6558 
6559 	switch (task->tk_status) {
6560 	case 0:
6561 		renew_lease(data->res.server, data->timestamp);
6562 		break;
6563 	case -NFS4ERR_ADMIN_REVOKED:
6564 	case -NFS4ERR_DELEG_REVOKED:
6565 	case -NFS4ERR_EXPIRED:
6566 		nfs4_free_revoked_stateid(data->res.server,
6567 				data->args.stateid,
6568 				task->tk_msg.rpc_cred);
6569 		fallthrough;
6570 	case -NFS4ERR_BAD_STATEID:
6571 	case -NFS4ERR_STALE_STATEID:
6572 	case -ETIMEDOUT:
6573 		task->tk_status = 0;
6574 		break;
6575 	case -NFS4ERR_OLD_STATEID:
6576 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6577 			nfs4_stateid_seqid_inc(&data->stateid);
6578 		if (data->args.bitmask) {
6579 			data->args.bitmask = NULL;
6580 			data->res.fattr = NULL;
6581 		}
6582 		goto out_restart;
6583 	case -NFS4ERR_ACCESS:
6584 		if (data->args.bitmask) {
6585 			data->args.bitmask = NULL;
6586 			data->res.fattr = NULL;
6587 			goto out_restart;
6588 		}
6589 		fallthrough;
6590 	default:
6591 		task->tk_status = nfs4_async_handle_exception(task,
6592 				data->res.server, task->tk_status,
6593 				&exception);
6594 		if (exception.retry)
6595 			goto out_restart;
6596 	}
6597 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6598 	data->rpc_status = task->tk_status;
6599 	return;
6600 out_restart:
6601 	task->tk_status = 0;
6602 	rpc_restart_call_prepare(task);
6603 }
6604 
6605 static void nfs4_delegreturn_release(void *calldata)
6606 {
6607 	struct nfs4_delegreturndata *data = calldata;
6608 	struct inode *inode = data->inode;
6609 
6610 	if (data->lr.roc)
6611 		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6612 				 data->res.lr_ret);
6613 	if (inode) {
6614 		nfs4_fattr_set_prechange(&data->fattr,
6615 					 inode_peek_iversion_raw(inode));
6616 		nfs_refresh_inode(inode, &data->fattr);
6617 		nfs_iput_and_deactive(inode);
6618 	}
6619 	kfree(calldata);
6620 }
6621 
6622 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6623 {
6624 	struct nfs4_delegreturndata *d_data;
6625 	struct pnfs_layout_hdr *lo;
6626 
6627 	d_data = data;
6628 
6629 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6630 		nfs4_sequence_done(task, &d_data->res.seq_res);
6631 		return;
6632 	}
6633 
6634 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6635 	if (lo && !pnfs_layout_is_valid(lo)) {
6636 		d_data->args.lr_args = NULL;
6637 		d_data->res.lr_res = NULL;
6638 	}
6639 
6640 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6641 			&d_data->args.seq_args,
6642 			&d_data->res.seq_res,
6643 			task);
6644 }
6645 
6646 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6647 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6648 	.rpc_call_done = nfs4_delegreturn_done,
6649 	.rpc_release = nfs4_delegreturn_release,
6650 };
6651 
6652 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6653 {
6654 	struct nfs4_delegreturndata *data;
6655 	struct nfs_server *server = NFS_SERVER(inode);
6656 	struct rpc_task *task;
6657 	struct rpc_message msg = {
6658 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6659 		.rpc_cred = cred,
6660 	};
6661 	struct rpc_task_setup task_setup_data = {
6662 		.rpc_client = server->client,
6663 		.rpc_message = &msg,
6664 		.callback_ops = &nfs4_delegreturn_ops,
6665 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6666 	};
6667 	int status = 0;
6668 
6669 	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6670 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6671 
6672 	data = kzalloc(sizeof(*data), GFP_KERNEL);
6673 	if (data == NULL)
6674 		return -ENOMEM;
6675 
6676 	nfs4_state_protect(server->nfs_client,
6677 			NFS_SP4_MACH_CRED_CLEANUP,
6678 			&task_setup_data.rpc_client, &msg);
6679 
6680 	data->args.fhandle = &data->fh;
6681 	data->args.stateid = &data->stateid;
6682 	nfs4_bitmask_set(data->args.bitmask_store,
6683 			 server->cache_consistency_bitmask, inode, 0);
6684 	data->args.bitmask = data->args.bitmask_store;
6685 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6686 	nfs4_stateid_copy(&data->stateid, stateid);
6687 	data->res.fattr = &data->fattr;
6688 	data->res.server = server;
6689 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6690 	data->lr.arg.ld_private = &data->lr.ld_private;
6691 	nfs_fattr_init(data->res.fattr);
6692 	data->timestamp = jiffies;
6693 	data->rpc_status = 0;
6694 	data->inode = nfs_igrab_and_active(inode);
6695 	if (data->inode || issync) {
6696 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6697 					cred);
6698 		if (data->lr.roc) {
6699 			data->args.lr_args = &data->lr.arg;
6700 			data->res.lr_res = &data->lr.res;
6701 		}
6702 	}
6703 
6704 	if (!data->inode)
6705 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6706 				   1);
6707 	else
6708 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6709 				   0);
6710 	task_setup_data.callback_data = data;
6711 	msg.rpc_argp = &data->args;
6712 	msg.rpc_resp = &data->res;
6713 	task = rpc_run_task(&task_setup_data);
6714 	if (IS_ERR(task))
6715 		return PTR_ERR(task);
6716 	if (!issync)
6717 		goto out;
6718 	status = rpc_wait_for_completion_task(task);
6719 	if (status != 0)
6720 		goto out;
6721 	status = data->rpc_status;
6722 out:
6723 	rpc_put_task(task);
6724 	return status;
6725 }
6726 
6727 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6728 {
6729 	struct nfs_server *server = NFS_SERVER(inode);
6730 	struct nfs4_exception exception = { };
6731 	int err;
6732 	do {
6733 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6734 		trace_nfs4_delegreturn(inode, stateid, err);
6735 		switch (err) {
6736 			case -NFS4ERR_STALE_STATEID:
6737 			case -NFS4ERR_EXPIRED:
6738 			case 0:
6739 				return 0;
6740 		}
6741 		err = nfs4_handle_exception(server, err, &exception);
6742 	} while (exception.retry);
6743 	return err;
6744 }
6745 
6746 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6747 {
6748 	struct inode *inode = state->inode;
6749 	struct nfs_server *server = NFS_SERVER(inode);
6750 	struct nfs_client *clp = server->nfs_client;
6751 	struct nfs_lockt_args arg = {
6752 		.fh = NFS_FH(inode),
6753 		.fl = request,
6754 	};
6755 	struct nfs_lockt_res res = {
6756 		.denied = request,
6757 	};
6758 	struct rpc_message msg = {
6759 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6760 		.rpc_argp	= &arg,
6761 		.rpc_resp	= &res,
6762 		.rpc_cred	= state->owner->so_cred,
6763 	};
6764 	struct nfs4_lock_state *lsp;
6765 	int status;
6766 
6767 	arg.lock_owner.clientid = clp->cl_clientid;
6768 	status = nfs4_set_lock_state(state, request);
6769 	if (status != 0)
6770 		goto out;
6771 	lsp = request->fl_u.nfs4_fl.owner;
6772 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6773 	arg.lock_owner.s_dev = server->s_dev;
6774 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6775 	switch (status) {
6776 		case 0:
6777 			request->fl_type = F_UNLCK;
6778 			break;
6779 		case -NFS4ERR_DENIED:
6780 			status = 0;
6781 	}
6782 	request->fl_ops->fl_release_private(request);
6783 	request->fl_ops = NULL;
6784 out:
6785 	return status;
6786 }
6787 
6788 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6789 {
6790 	struct nfs4_exception exception = {
6791 		.interruptible = true,
6792 	};
6793 	int err;
6794 
6795 	do {
6796 		err = _nfs4_proc_getlk(state, cmd, request);
6797 		trace_nfs4_get_lock(request, state, cmd, err);
6798 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6799 				&exception);
6800 	} while (exception.retry);
6801 	return err;
6802 }
6803 
6804 /*
6805  * Update the seqid of a lock stateid after receiving
6806  * NFS4ERR_OLD_STATEID
6807  */
6808 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6809 		struct nfs4_lock_state *lsp)
6810 {
6811 	struct nfs4_state *state = lsp->ls_state;
6812 	bool ret = false;
6813 
6814 	spin_lock(&state->state_lock);
6815 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6816 		goto out;
6817 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6818 		nfs4_stateid_seqid_inc(dst);
6819 	else
6820 		dst->seqid = lsp->ls_stateid.seqid;
6821 	ret = true;
6822 out:
6823 	spin_unlock(&state->state_lock);
6824 	return ret;
6825 }
6826 
6827 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6828 		struct nfs4_lock_state *lsp)
6829 {
6830 	struct nfs4_state *state = lsp->ls_state;
6831 	bool ret;
6832 
6833 	spin_lock(&state->state_lock);
6834 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6835 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6836 	spin_unlock(&state->state_lock);
6837 	return ret;
6838 }
6839 
6840 struct nfs4_unlockdata {
6841 	struct nfs_locku_args arg;
6842 	struct nfs_locku_res res;
6843 	struct nfs4_lock_state *lsp;
6844 	struct nfs_open_context *ctx;
6845 	struct nfs_lock_context *l_ctx;
6846 	struct file_lock fl;
6847 	struct nfs_server *server;
6848 	unsigned long timestamp;
6849 };
6850 
6851 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6852 		struct nfs_open_context *ctx,
6853 		struct nfs4_lock_state *lsp,
6854 		struct nfs_seqid *seqid)
6855 {
6856 	struct nfs4_unlockdata *p;
6857 	struct nfs4_state *state = lsp->ls_state;
6858 	struct inode *inode = state->inode;
6859 
6860 	p = kzalloc(sizeof(*p), GFP_KERNEL);
6861 	if (p == NULL)
6862 		return NULL;
6863 	p->arg.fh = NFS_FH(inode);
6864 	p->arg.fl = &p->fl;
6865 	p->arg.seqid = seqid;
6866 	p->res.seqid = seqid;
6867 	p->lsp = lsp;
6868 	/* Ensure we don't close file until we're done freeing locks! */
6869 	p->ctx = get_nfs_open_context(ctx);
6870 	p->l_ctx = nfs_get_lock_context(ctx);
6871 	locks_init_lock(&p->fl);
6872 	locks_copy_lock(&p->fl, fl);
6873 	p->server = NFS_SERVER(inode);
6874 	spin_lock(&state->state_lock);
6875 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6876 	spin_unlock(&state->state_lock);
6877 	return p;
6878 }
6879 
6880 static void nfs4_locku_release_calldata(void *data)
6881 {
6882 	struct nfs4_unlockdata *calldata = data;
6883 	nfs_free_seqid(calldata->arg.seqid);
6884 	nfs4_put_lock_state(calldata->lsp);
6885 	nfs_put_lock_context(calldata->l_ctx);
6886 	put_nfs_open_context(calldata->ctx);
6887 	kfree(calldata);
6888 }
6889 
6890 static void nfs4_locku_done(struct rpc_task *task, void *data)
6891 {
6892 	struct nfs4_unlockdata *calldata = data;
6893 	struct nfs4_exception exception = {
6894 		.inode = calldata->lsp->ls_state->inode,
6895 		.stateid = &calldata->arg.stateid,
6896 	};
6897 
6898 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6899 		return;
6900 	switch (task->tk_status) {
6901 		case 0:
6902 			renew_lease(calldata->server, calldata->timestamp);
6903 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6904 			if (nfs4_update_lock_stateid(calldata->lsp,
6905 					&calldata->res.stateid))
6906 				break;
6907 			fallthrough;
6908 		case -NFS4ERR_ADMIN_REVOKED:
6909 		case -NFS4ERR_EXPIRED:
6910 			nfs4_free_revoked_stateid(calldata->server,
6911 					&calldata->arg.stateid,
6912 					task->tk_msg.rpc_cred);
6913 			fallthrough;
6914 		case -NFS4ERR_BAD_STATEID:
6915 		case -NFS4ERR_STALE_STATEID:
6916 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6917 						calldata->lsp))
6918 				rpc_restart_call_prepare(task);
6919 			break;
6920 		case -NFS4ERR_OLD_STATEID:
6921 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6922 						calldata->lsp))
6923 				rpc_restart_call_prepare(task);
6924 			break;
6925 		default:
6926 			task->tk_status = nfs4_async_handle_exception(task,
6927 					calldata->server, task->tk_status,
6928 					&exception);
6929 			if (exception.retry)
6930 				rpc_restart_call_prepare(task);
6931 	}
6932 	nfs_release_seqid(calldata->arg.seqid);
6933 }
6934 
6935 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6936 {
6937 	struct nfs4_unlockdata *calldata = data;
6938 
6939 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6940 		nfs_async_iocounter_wait(task, calldata->l_ctx))
6941 		return;
6942 
6943 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6944 		goto out_wait;
6945 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6946 		/* Note: exit _without_ running nfs4_locku_done */
6947 		goto out_no_action;
6948 	}
6949 	calldata->timestamp = jiffies;
6950 	if (nfs4_setup_sequence(calldata->server->nfs_client,
6951 				&calldata->arg.seq_args,
6952 				&calldata->res.seq_res,
6953 				task) != 0)
6954 		nfs_release_seqid(calldata->arg.seqid);
6955 	return;
6956 out_no_action:
6957 	task->tk_action = NULL;
6958 out_wait:
6959 	nfs4_sequence_done(task, &calldata->res.seq_res);
6960 }
6961 
6962 static const struct rpc_call_ops nfs4_locku_ops = {
6963 	.rpc_call_prepare = nfs4_locku_prepare,
6964 	.rpc_call_done = nfs4_locku_done,
6965 	.rpc_release = nfs4_locku_release_calldata,
6966 };
6967 
6968 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6969 		struct nfs_open_context *ctx,
6970 		struct nfs4_lock_state *lsp,
6971 		struct nfs_seqid *seqid)
6972 {
6973 	struct nfs4_unlockdata *data;
6974 	struct rpc_message msg = {
6975 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6976 		.rpc_cred = ctx->cred,
6977 	};
6978 	struct rpc_task_setup task_setup_data = {
6979 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6980 		.rpc_message = &msg,
6981 		.callback_ops = &nfs4_locku_ops,
6982 		.workqueue = nfsiod_workqueue,
6983 		.flags = RPC_TASK_ASYNC,
6984 	};
6985 
6986 	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6987 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6988 
6989 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6990 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6991 
6992 	/* Ensure this is an unlock - when canceling a lock, the
6993 	 * canceled lock is passed in, and it won't be an unlock.
6994 	 */
6995 	fl->fl_type = F_UNLCK;
6996 	if (fl->fl_flags & FL_CLOSE)
6997 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6998 
6999 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
7000 	if (data == NULL) {
7001 		nfs_free_seqid(seqid);
7002 		return ERR_PTR(-ENOMEM);
7003 	}
7004 
7005 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
7006 	msg.rpc_argp = &data->arg;
7007 	msg.rpc_resp = &data->res;
7008 	task_setup_data.callback_data = data;
7009 	return rpc_run_task(&task_setup_data);
7010 }
7011 
7012 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7013 {
7014 	struct inode *inode = state->inode;
7015 	struct nfs4_state_owner *sp = state->owner;
7016 	struct nfs_inode *nfsi = NFS_I(inode);
7017 	struct nfs_seqid *seqid;
7018 	struct nfs4_lock_state *lsp;
7019 	struct rpc_task *task;
7020 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7021 	int status = 0;
7022 	unsigned char fl_flags = request->fl_flags;
7023 
7024 	status = nfs4_set_lock_state(state, request);
7025 	/* Unlock _before_ we do the RPC call */
7026 	request->fl_flags |= FL_EXISTS;
7027 	/* Exclude nfs_delegation_claim_locks() */
7028 	mutex_lock(&sp->so_delegreturn_mutex);
7029 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7030 	down_read(&nfsi->rwsem);
7031 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7032 		up_read(&nfsi->rwsem);
7033 		mutex_unlock(&sp->so_delegreturn_mutex);
7034 		goto out;
7035 	}
7036 	lsp = request->fl_u.nfs4_fl.owner;
7037 	set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7038 	up_read(&nfsi->rwsem);
7039 	mutex_unlock(&sp->so_delegreturn_mutex);
7040 	if (status != 0)
7041 		goto out;
7042 	/* Is this a delegated lock? */
7043 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7044 		goto out;
7045 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7046 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7047 	status = -ENOMEM;
7048 	if (IS_ERR(seqid))
7049 		goto out;
7050 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7051 	status = PTR_ERR(task);
7052 	if (IS_ERR(task))
7053 		goto out;
7054 	status = rpc_wait_for_completion_task(task);
7055 	rpc_put_task(task);
7056 out:
7057 	request->fl_flags = fl_flags;
7058 	trace_nfs4_unlock(request, state, F_SETLK, status);
7059 	return status;
7060 }
7061 
7062 struct nfs4_lockdata {
7063 	struct nfs_lock_args arg;
7064 	struct nfs_lock_res res;
7065 	struct nfs4_lock_state *lsp;
7066 	struct nfs_open_context *ctx;
7067 	struct file_lock fl;
7068 	unsigned long timestamp;
7069 	int rpc_status;
7070 	int cancelled;
7071 	struct nfs_server *server;
7072 };
7073 
7074 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7075 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7076 		gfp_t gfp_mask)
7077 {
7078 	struct nfs4_lockdata *p;
7079 	struct inode *inode = lsp->ls_state->inode;
7080 	struct nfs_server *server = NFS_SERVER(inode);
7081 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7082 
7083 	p = kzalloc(sizeof(*p), gfp_mask);
7084 	if (p == NULL)
7085 		return NULL;
7086 
7087 	p->arg.fh = NFS_FH(inode);
7088 	p->arg.fl = &p->fl;
7089 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7090 	if (IS_ERR(p->arg.open_seqid))
7091 		goto out_free;
7092 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7093 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7094 	if (IS_ERR(p->arg.lock_seqid))
7095 		goto out_free_seqid;
7096 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7097 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7098 	p->arg.lock_owner.s_dev = server->s_dev;
7099 	p->res.lock_seqid = p->arg.lock_seqid;
7100 	p->lsp = lsp;
7101 	p->server = server;
7102 	p->ctx = get_nfs_open_context(ctx);
7103 	locks_init_lock(&p->fl);
7104 	locks_copy_lock(&p->fl, fl);
7105 	return p;
7106 out_free_seqid:
7107 	nfs_free_seqid(p->arg.open_seqid);
7108 out_free:
7109 	kfree(p);
7110 	return NULL;
7111 }
7112 
7113 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7114 {
7115 	struct nfs4_lockdata *data = calldata;
7116 	struct nfs4_state *state = data->lsp->ls_state;
7117 
7118 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7119 		goto out_wait;
7120 	/* Do we need to do an open_to_lock_owner? */
7121 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7122 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7123 			goto out_release_lock_seqid;
7124 		}
7125 		nfs4_stateid_copy(&data->arg.open_stateid,
7126 				&state->open_stateid);
7127 		data->arg.new_lock_owner = 1;
7128 		data->res.open_seqid = data->arg.open_seqid;
7129 	} else {
7130 		data->arg.new_lock_owner = 0;
7131 		nfs4_stateid_copy(&data->arg.lock_stateid,
7132 				&data->lsp->ls_stateid);
7133 	}
7134 	if (!nfs4_valid_open_stateid(state)) {
7135 		data->rpc_status = -EBADF;
7136 		task->tk_action = NULL;
7137 		goto out_release_open_seqid;
7138 	}
7139 	data->timestamp = jiffies;
7140 	if (nfs4_setup_sequence(data->server->nfs_client,
7141 				&data->arg.seq_args,
7142 				&data->res.seq_res,
7143 				task) == 0)
7144 		return;
7145 out_release_open_seqid:
7146 	nfs_release_seqid(data->arg.open_seqid);
7147 out_release_lock_seqid:
7148 	nfs_release_seqid(data->arg.lock_seqid);
7149 out_wait:
7150 	nfs4_sequence_done(task, &data->res.seq_res);
7151 	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7152 }
7153 
7154 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7155 {
7156 	struct nfs4_lockdata *data = calldata;
7157 	struct nfs4_lock_state *lsp = data->lsp;
7158 
7159 	if (!nfs4_sequence_done(task, &data->res.seq_res))
7160 		return;
7161 
7162 	data->rpc_status = task->tk_status;
7163 	switch (task->tk_status) {
7164 	case 0:
7165 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7166 				data->timestamp);
7167 		if (data->arg.new_lock && !data->cancelled) {
7168 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7169 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7170 				goto out_restart;
7171 		}
7172 		if (data->arg.new_lock_owner != 0) {
7173 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7174 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7175 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7176 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7177 			goto out_restart;
7178 		break;
7179 	case -NFS4ERR_OLD_STATEID:
7180 		if (data->arg.new_lock_owner != 0 &&
7181 			nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7182 					lsp->ls_state))
7183 			goto out_restart;
7184 		if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7185 			goto out_restart;
7186 		fallthrough;
7187 	case -NFS4ERR_BAD_STATEID:
7188 	case -NFS4ERR_STALE_STATEID:
7189 	case -NFS4ERR_EXPIRED:
7190 		if (data->arg.new_lock_owner != 0) {
7191 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7192 						&lsp->ls_state->open_stateid))
7193 				goto out_restart;
7194 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7195 						&lsp->ls_stateid))
7196 				goto out_restart;
7197 	}
7198 out_done:
7199 	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7200 	return;
7201 out_restart:
7202 	if (!data->cancelled)
7203 		rpc_restart_call_prepare(task);
7204 	goto out_done;
7205 }
7206 
7207 static void nfs4_lock_release(void *calldata)
7208 {
7209 	struct nfs4_lockdata *data = calldata;
7210 
7211 	nfs_free_seqid(data->arg.open_seqid);
7212 	if (data->cancelled && data->rpc_status == 0) {
7213 		struct rpc_task *task;
7214 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7215 				data->arg.lock_seqid);
7216 		if (!IS_ERR(task))
7217 			rpc_put_task_async(task);
7218 		dprintk("%s: cancelling lock!\n", __func__);
7219 	} else
7220 		nfs_free_seqid(data->arg.lock_seqid);
7221 	nfs4_put_lock_state(data->lsp);
7222 	put_nfs_open_context(data->ctx);
7223 	kfree(data);
7224 }
7225 
7226 static const struct rpc_call_ops nfs4_lock_ops = {
7227 	.rpc_call_prepare = nfs4_lock_prepare,
7228 	.rpc_call_done = nfs4_lock_done,
7229 	.rpc_release = nfs4_lock_release,
7230 };
7231 
7232 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7233 {
7234 	switch (error) {
7235 	case -NFS4ERR_ADMIN_REVOKED:
7236 	case -NFS4ERR_EXPIRED:
7237 	case -NFS4ERR_BAD_STATEID:
7238 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7239 		if (new_lock_owner != 0 ||
7240 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7241 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7242 		break;
7243 	case -NFS4ERR_STALE_STATEID:
7244 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7245 		nfs4_schedule_lease_recovery(server->nfs_client);
7246 	}
7247 }
7248 
7249 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7250 {
7251 	struct nfs4_lockdata *data;
7252 	struct rpc_task *task;
7253 	struct rpc_message msg = {
7254 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7255 		.rpc_cred = state->owner->so_cred,
7256 	};
7257 	struct rpc_task_setup task_setup_data = {
7258 		.rpc_client = NFS_CLIENT(state->inode),
7259 		.rpc_message = &msg,
7260 		.callback_ops = &nfs4_lock_ops,
7261 		.workqueue = nfsiod_workqueue,
7262 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7263 	};
7264 	int ret;
7265 
7266 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7267 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7268 
7269 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7270 				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7271 	if (data == NULL)
7272 		return -ENOMEM;
7273 	if (IS_SETLKW(cmd))
7274 		data->arg.block = 1;
7275 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7276 				recovery_type > NFS_LOCK_NEW);
7277 	msg.rpc_argp = &data->arg;
7278 	msg.rpc_resp = &data->res;
7279 	task_setup_data.callback_data = data;
7280 	if (recovery_type > NFS_LOCK_NEW) {
7281 		if (recovery_type == NFS_LOCK_RECLAIM)
7282 			data->arg.reclaim = NFS_LOCK_RECLAIM;
7283 	} else
7284 		data->arg.new_lock = 1;
7285 	task = rpc_run_task(&task_setup_data);
7286 	if (IS_ERR(task))
7287 		return PTR_ERR(task);
7288 	ret = rpc_wait_for_completion_task(task);
7289 	if (ret == 0) {
7290 		ret = data->rpc_status;
7291 		if (ret)
7292 			nfs4_handle_setlk_error(data->server, data->lsp,
7293 					data->arg.new_lock_owner, ret);
7294 	} else
7295 		data->cancelled = true;
7296 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7297 	rpc_put_task(task);
7298 	dprintk("%s: ret = %d\n", __func__, ret);
7299 	return ret;
7300 }
7301 
7302 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7303 {
7304 	struct nfs_server *server = NFS_SERVER(state->inode);
7305 	struct nfs4_exception exception = {
7306 		.inode = state->inode,
7307 	};
7308 	int err;
7309 
7310 	do {
7311 		/* Cache the lock if possible... */
7312 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7313 			return 0;
7314 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7315 		if (err != -NFS4ERR_DELAY)
7316 			break;
7317 		nfs4_handle_exception(server, err, &exception);
7318 	} while (exception.retry);
7319 	return err;
7320 }
7321 
7322 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7323 {
7324 	struct nfs_server *server = NFS_SERVER(state->inode);
7325 	struct nfs4_exception exception = {
7326 		.inode = state->inode,
7327 	};
7328 	int err;
7329 
7330 	err = nfs4_set_lock_state(state, request);
7331 	if (err != 0)
7332 		return err;
7333 	if (!recover_lost_locks) {
7334 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7335 		return 0;
7336 	}
7337 	do {
7338 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7339 			return 0;
7340 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7341 		switch (err) {
7342 		default:
7343 			goto out;
7344 		case -NFS4ERR_GRACE:
7345 		case -NFS4ERR_DELAY:
7346 			nfs4_handle_exception(server, err, &exception);
7347 			err = 0;
7348 		}
7349 	} while (exception.retry);
7350 out:
7351 	return err;
7352 }
7353 
7354 #if defined(CONFIG_NFS_V4_1)
7355 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7356 {
7357 	struct nfs4_lock_state *lsp;
7358 	int status;
7359 
7360 	status = nfs4_set_lock_state(state, request);
7361 	if (status != 0)
7362 		return status;
7363 	lsp = request->fl_u.nfs4_fl.owner;
7364 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7365 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7366 		return 0;
7367 	return nfs4_lock_expired(state, request);
7368 }
7369 #endif
7370 
7371 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7372 {
7373 	struct nfs_inode *nfsi = NFS_I(state->inode);
7374 	struct nfs4_state_owner *sp = state->owner;
7375 	unsigned char fl_flags = request->fl_flags;
7376 	int status;
7377 
7378 	request->fl_flags |= FL_ACCESS;
7379 	status = locks_lock_inode_wait(state->inode, request);
7380 	if (status < 0)
7381 		goto out;
7382 	mutex_lock(&sp->so_delegreturn_mutex);
7383 	down_read(&nfsi->rwsem);
7384 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7385 		/* Yes: cache locks! */
7386 		/* ...but avoid races with delegation recall... */
7387 		request->fl_flags = fl_flags & ~FL_SLEEP;
7388 		status = locks_lock_inode_wait(state->inode, request);
7389 		up_read(&nfsi->rwsem);
7390 		mutex_unlock(&sp->so_delegreturn_mutex);
7391 		goto out;
7392 	}
7393 	up_read(&nfsi->rwsem);
7394 	mutex_unlock(&sp->so_delegreturn_mutex);
7395 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7396 out:
7397 	request->fl_flags = fl_flags;
7398 	return status;
7399 }
7400 
7401 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7402 {
7403 	struct nfs4_exception exception = {
7404 		.state = state,
7405 		.inode = state->inode,
7406 		.interruptible = true,
7407 	};
7408 	int err;
7409 
7410 	do {
7411 		err = _nfs4_proc_setlk(state, cmd, request);
7412 		if (err == -NFS4ERR_DENIED)
7413 			err = -EAGAIN;
7414 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7415 				err, &exception);
7416 	} while (exception.retry);
7417 	return err;
7418 }
7419 
7420 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7421 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7422 
7423 static int
7424 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7425 			struct file_lock *request)
7426 {
7427 	int		status = -ERESTARTSYS;
7428 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7429 
7430 	while(!signalled()) {
7431 		status = nfs4_proc_setlk(state, cmd, request);
7432 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7433 			break;
7434 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7435 		schedule_timeout(timeout);
7436 		timeout *= 2;
7437 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7438 		status = -ERESTARTSYS;
7439 	}
7440 	return status;
7441 }
7442 
7443 #ifdef CONFIG_NFS_V4_1
7444 struct nfs4_lock_waiter {
7445 	struct inode		*inode;
7446 	struct nfs_lowner	owner;
7447 	wait_queue_entry_t	wait;
7448 };
7449 
7450 static int
7451 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7452 {
7453 	struct nfs4_lock_waiter	*waiter	=
7454 		container_of(wait, struct nfs4_lock_waiter, wait);
7455 
7456 	/* NULL key means to wake up everyone */
7457 	if (key) {
7458 		struct cb_notify_lock_args	*cbnl = key;
7459 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7460 						*wowner = &waiter->owner;
7461 
7462 		/* Only wake if the callback was for the same owner. */
7463 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7464 			return 0;
7465 
7466 		/* Make sure it's for the right inode */
7467 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7468 			return 0;
7469 	}
7470 
7471 	return woken_wake_function(wait, mode, flags, key);
7472 }
7473 
7474 static int
7475 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7476 {
7477 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7478 	struct nfs_server *server = NFS_SERVER(state->inode);
7479 	struct nfs_client *clp = server->nfs_client;
7480 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7481 	struct nfs4_lock_waiter waiter = {
7482 		.inode = state->inode,
7483 		.owner = { .clientid = clp->cl_clientid,
7484 			   .id = lsp->ls_seqid.owner_id,
7485 			   .s_dev = server->s_dev },
7486 	};
7487 	int status;
7488 
7489 	/* Don't bother with waitqueue if we don't expect a callback */
7490 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7491 		return nfs4_retry_setlk_simple(state, cmd, request);
7492 
7493 	init_wait(&waiter.wait);
7494 	waiter.wait.func = nfs4_wake_lock_waiter;
7495 	add_wait_queue(q, &waiter.wait);
7496 
7497 	do {
7498 		status = nfs4_proc_setlk(state, cmd, request);
7499 		if (status != -EAGAIN || IS_SETLK(cmd))
7500 			break;
7501 
7502 		status = -ERESTARTSYS;
7503 		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7504 			   NFS4_LOCK_MAXTIMEOUT);
7505 	} while (!signalled());
7506 
7507 	remove_wait_queue(q, &waiter.wait);
7508 
7509 	return status;
7510 }
7511 #else /* !CONFIG_NFS_V4_1 */
7512 static inline int
7513 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7514 {
7515 	return nfs4_retry_setlk_simple(state, cmd, request);
7516 }
7517 #endif
7518 
7519 static int
7520 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7521 {
7522 	struct nfs_open_context *ctx;
7523 	struct nfs4_state *state;
7524 	int status;
7525 
7526 	/* verify open state */
7527 	ctx = nfs_file_open_context(filp);
7528 	state = ctx->state;
7529 
7530 	if (IS_GETLK(cmd)) {
7531 		if (state != NULL)
7532 			return nfs4_proc_getlk(state, F_GETLK, request);
7533 		return 0;
7534 	}
7535 
7536 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7537 		return -EINVAL;
7538 
7539 	if (request->fl_type == F_UNLCK) {
7540 		if (state != NULL)
7541 			return nfs4_proc_unlck(state, cmd, request);
7542 		return 0;
7543 	}
7544 
7545 	if (state == NULL)
7546 		return -ENOLCK;
7547 
7548 	if ((request->fl_flags & FL_POSIX) &&
7549 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7550 		return -ENOLCK;
7551 
7552 	/*
7553 	 * Don't rely on the VFS having checked the file open mode,
7554 	 * since it won't do this for flock() locks.
7555 	 */
7556 	switch (request->fl_type) {
7557 	case F_RDLCK:
7558 		if (!(filp->f_mode & FMODE_READ))
7559 			return -EBADF;
7560 		break;
7561 	case F_WRLCK:
7562 		if (!(filp->f_mode & FMODE_WRITE))
7563 			return -EBADF;
7564 	}
7565 
7566 	status = nfs4_set_lock_state(state, request);
7567 	if (status != 0)
7568 		return status;
7569 
7570 	return nfs4_retry_setlk(state, cmd, request);
7571 }
7572 
7573 static int nfs4_delete_lease(struct file *file, void **priv)
7574 {
7575 	return generic_setlease(file, F_UNLCK, NULL, priv);
7576 }
7577 
7578 static int nfs4_add_lease(struct file *file, int arg, struct file_lock **lease,
7579 			  void **priv)
7580 {
7581 	struct inode *inode = file_inode(file);
7582 	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7583 	int ret;
7584 
7585 	/* No delegation, no lease */
7586 	if (!nfs4_have_delegation(inode, type))
7587 		return -EAGAIN;
7588 	ret = generic_setlease(file, arg, lease, priv);
7589 	if (ret || nfs4_have_delegation(inode, type))
7590 		return ret;
7591 	/* We raced with a delegation return */
7592 	nfs4_delete_lease(file, priv);
7593 	return -EAGAIN;
7594 }
7595 
7596 int nfs4_proc_setlease(struct file *file, int arg, struct file_lock **lease,
7597 		       void **priv)
7598 {
7599 	switch (arg) {
7600 	case F_RDLCK:
7601 	case F_WRLCK:
7602 		return nfs4_add_lease(file, arg, lease, priv);
7603 	case F_UNLCK:
7604 		return nfs4_delete_lease(file, priv);
7605 	default:
7606 		return -EINVAL;
7607 	}
7608 }
7609 
7610 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7611 {
7612 	struct nfs_server *server = NFS_SERVER(state->inode);
7613 	int err;
7614 
7615 	err = nfs4_set_lock_state(state, fl);
7616 	if (err != 0)
7617 		return err;
7618 	do {
7619 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7620 		if (err != -NFS4ERR_DELAY)
7621 			break;
7622 		ssleep(1);
7623 	} while (err == -NFS4ERR_DELAY);
7624 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7625 }
7626 
7627 struct nfs_release_lockowner_data {
7628 	struct nfs4_lock_state *lsp;
7629 	struct nfs_server *server;
7630 	struct nfs_release_lockowner_args args;
7631 	struct nfs_release_lockowner_res res;
7632 	unsigned long timestamp;
7633 };
7634 
7635 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7636 {
7637 	struct nfs_release_lockowner_data *data = calldata;
7638 	struct nfs_server *server = data->server;
7639 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7640 			   &data->res.seq_res, task);
7641 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7642 	data->timestamp = jiffies;
7643 }
7644 
7645 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7646 {
7647 	struct nfs_release_lockowner_data *data = calldata;
7648 	struct nfs_server *server = data->server;
7649 
7650 	nfs40_sequence_done(task, &data->res.seq_res);
7651 
7652 	switch (task->tk_status) {
7653 	case 0:
7654 		renew_lease(server, data->timestamp);
7655 		break;
7656 	case -NFS4ERR_STALE_CLIENTID:
7657 	case -NFS4ERR_EXPIRED:
7658 		nfs4_schedule_lease_recovery(server->nfs_client);
7659 		break;
7660 	case -NFS4ERR_LEASE_MOVED:
7661 	case -NFS4ERR_DELAY:
7662 		if (nfs4_async_handle_error(task, server,
7663 					    NULL, NULL) == -EAGAIN)
7664 			rpc_restart_call_prepare(task);
7665 	}
7666 }
7667 
7668 static void nfs4_release_lockowner_release(void *calldata)
7669 {
7670 	struct nfs_release_lockowner_data *data = calldata;
7671 	nfs4_free_lock_state(data->server, data->lsp);
7672 	kfree(calldata);
7673 }
7674 
7675 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7676 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7677 	.rpc_call_done = nfs4_release_lockowner_done,
7678 	.rpc_release = nfs4_release_lockowner_release,
7679 };
7680 
7681 static void
7682 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7683 {
7684 	struct nfs_release_lockowner_data *data;
7685 	struct rpc_message msg = {
7686 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7687 	};
7688 
7689 	if (server->nfs_client->cl_mvops->minor_version != 0)
7690 		return;
7691 
7692 	data = kmalloc(sizeof(*data), GFP_KERNEL);
7693 	if (!data)
7694 		return;
7695 	data->lsp = lsp;
7696 	data->server = server;
7697 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7698 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7699 	data->args.lock_owner.s_dev = server->s_dev;
7700 
7701 	msg.rpc_argp = &data->args;
7702 	msg.rpc_resp = &data->res;
7703 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7704 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7705 }
7706 
7707 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7708 
7709 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7710 				   struct mnt_idmap *idmap,
7711 				   struct dentry *unused, struct inode *inode,
7712 				   const char *key, const void *buf,
7713 				   size_t buflen, int flags)
7714 {
7715 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7716 }
7717 
7718 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7719 				   struct dentry *unused, struct inode *inode,
7720 				   const char *key, void *buf, size_t buflen)
7721 {
7722 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7723 }
7724 
7725 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7726 {
7727 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7728 }
7729 
7730 #if defined(CONFIG_NFS_V4_1)
7731 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7732 
7733 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7734 				    struct mnt_idmap *idmap,
7735 				    struct dentry *unused, struct inode *inode,
7736 				    const char *key, const void *buf,
7737 				    size_t buflen, int flags)
7738 {
7739 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7740 }
7741 
7742 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7743 				    struct dentry *unused, struct inode *inode,
7744 				    const char *key, void *buf, size_t buflen)
7745 {
7746 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7747 }
7748 
7749 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7750 {
7751 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7752 }
7753 
7754 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7755 
7756 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7757 				    struct mnt_idmap *idmap,
7758 				    struct dentry *unused, struct inode *inode,
7759 				    const char *key, const void *buf,
7760 				    size_t buflen, int flags)
7761 {
7762 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7763 }
7764 
7765 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7766 				    struct dentry *unused, struct inode *inode,
7767 				    const char *key, void *buf, size_t buflen)
7768 {
7769 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7770 }
7771 
7772 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7773 {
7774 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7775 }
7776 
7777 #endif
7778 
7779 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7780 
7781 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7782 				     struct mnt_idmap *idmap,
7783 				     struct dentry *unused, struct inode *inode,
7784 				     const char *key, const void *buf,
7785 				     size_t buflen, int flags)
7786 {
7787 	if (security_ismaclabel(key))
7788 		return nfs4_set_security_label(inode, buf, buflen);
7789 
7790 	return -EOPNOTSUPP;
7791 }
7792 
7793 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7794 				     struct dentry *unused, struct inode *inode,
7795 				     const char *key, void *buf, size_t buflen)
7796 {
7797 	if (security_ismaclabel(key))
7798 		return nfs4_get_security_label(inode, buf, buflen);
7799 	return -EOPNOTSUPP;
7800 }
7801 
7802 static ssize_t
7803 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7804 {
7805 	int len = 0;
7806 
7807 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7808 		len = security_inode_listsecurity(inode, list, list_len);
7809 		if (len >= 0 && list_len && len > list_len)
7810 			return -ERANGE;
7811 	}
7812 	return len;
7813 }
7814 
7815 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7816 	.prefix = XATTR_SECURITY_PREFIX,
7817 	.get	= nfs4_xattr_get_nfs4_label,
7818 	.set	= nfs4_xattr_set_nfs4_label,
7819 };
7820 
7821 #else
7822 
7823 static ssize_t
7824 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7825 {
7826 	return 0;
7827 }
7828 
7829 #endif
7830 
7831 #ifdef CONFIG_NFS_V4_2
7832 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7833 				    struct mnt_idmap *idmap,
7834 				    struct dentry *unused, struct inode *inode,
7835 				    const char *key, const void *buf,
7836 				    size_t buflen, int flags)
7837 {
7838 	u32 mask;
7839 	int ret;
7840 
7841 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7842 		return -EOPNOTSUPP;
7843 
7844 	/*
7845 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7846 	 * flags right now. Handling of xattr operations use the normal
7847 	 * file read/write permissions.
7848 	 *
7849 	 * Just in case the server has other ideas (which RFC 8276 allows),
7850 	 * do a cached access check for the XA* flags to possibly avoid
7851 	 * doing an RPC and getting EACCES back.
7852 	 */
7853 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7854 		if (!(mask & NFS_ACCESS_XAWRITE))
7855 			return -EACCES;
7856 	}
7857 
7858 	if (buf == NULL) {
7859 		ret = nfs42_proc_removexattr(inode, key);
7860 		if (!ret)
7861 			nfs4_xattr_cache_remove(inode, key);
7862 	} else {
7863 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7864 		if (!ret)
7865 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7866 	}
7867 
7868 	return ret;
7869 }
7870 
7871 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7872 				    struct dentry *unused, struct inode *inode,
7873 				    const char *key, void *buf, size_t buflen)
7874 {
7875 	u32 mask;
7876 	ssize_t ret;
7877 
7878 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7879 		return -EOPNOTSUPP;
7880 
7881 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7882 		if (!(mask & NFS_ACCESS_XAREAD))
7883 			return -EACCES;
7884 	}
7885 
7886 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7887 	if (ret)
7888 		return ret;
7889 
7890 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7891 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7892 		return ret;
7893 
7894 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7895 
7896 	return ret;
7897 }
7898 
7899 static ssize_t
7900 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7901 {
7902 	u64 cookie;
7903 	bool eof;
7904 	ssize_t ret, size;
7905 	char *buf;
7906 	size_t buflen;
7907 	u32 mask;
7908 
7909 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7910 		return 0;
7911 
7912 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7913 		if (!(mask & NFS_ACCESS_XALIST))
7914 			return 0;
7915 	}
7916 
7917 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7918 	if (ret)
7919 		return ret;
7920 
7921 	ret = nfs4_xattr_cache_list(inode, list, list_len);
7922 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7923 		return ret;
7924 
7925 	cookie = 0;
7926 	eof = false;
7927 	buflen = list_len ? list_len : XATTR_LIST_MAX;
7928 	buf = list_len ? list : NULL;
7929 	size = 0;
7930 
7931 	while (!eof) {
7932 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
7933 		    &cookie, &eof);
7934 		if (ret < 0)
7935 			return ret;
7936 
7937 		if (list_len) {
7938 			buf += ret;
7939 			buflen -= ret;
7940 		}
7941 		size += ret;
7942 	}
7943 
7944 	if (list_len)
7945 		nfs4_xattr_cache_set_list(inode, list, size);
7946 
7947 	return size;
7948 }
7949 
7950 #else
7951 
7952 static ssize_t
7953 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7954 {
7955 	return 0;
7956 }
7957 #endif /* CONFIG_NFS_V4_2 */
7958 
7959 /*
7960  * nfs_fhget will use either the mounted_on_fileid or the fileid
7961  */
7962 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7963 {
7964 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7965 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7966 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7967 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7968 		return;
7969 
7970 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7971 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7972 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7973 	fattr->nlink = 2;
7974 }
7975 
7976 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7977 				   const struct qstr *name,
7978 				   struct nfs4_fs_locations *fs_locations,
7979 				   struct page *page)
7980 {
7981 	struct nfs_server *server = NFS_SERVER(dir);
7982 	u32 bitmask[3];
7983 	struct nfs4_fs_locations_arg args = {
7984 		.dir_fh = NFS_FH(dir),
7985 		.name = name,
7986 		.page = page,
7987 		.bitmask = bitmask,
7988 	};
7989 	struct nfs4_fs_locations_res res = {
7990 		.fs_locations = fs_locations,
7991 	};
7992 	struct rpc_message msg = {
7993 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7994 		.rpc_argp = &args,
7995 		.rpc_resp = &res,
7996 	};
7997 	int status;
7998 
7999 	dprintk("%s: start\n", __func__);
8000 
8001 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
8002 	bitmask[1] = nfs4_fattr_bitmap[1];
8003 
8004 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
8005 	 * is not supported */
8006 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
8007 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
8008 	else
8009 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8010 
8011 	nfs_fattr_init(fs_locations->fattr);
8012 	fs_locations->server = server;
8013 	fs_locations->nlocations = 0;
8014 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8015 	dprintk("%s: returned status = %d\n", __func__, status);
8016 	return status;
8017 }
8018 
8019 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8020 			   const struct qstr *name,
8021 			   struct nfs4_fs_locations *fs_locations,
8022 			   struct page *page)
8023 {
8024 	struct nfs4_exception exception = {
8025 		.interruptible = true,
8026 	};
8027 	int err;
8028 	do {
8029 		err = _nfs4_proc_fs_locations(client, dir, name,
8030 				fs_locations, page);
8031 		trace_nfs4_get_fs_locations(dir, name, err);
8032 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8033 				&exception);
8034 	} while (exception.retry);
8035 	return err;
8036 }
8037 
8038 /*
8039  * This operation also signals the server that this client is
8040  * performing migration recovery.  The server can stop returning
8041  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8042  * appended to this compound to identify the client ID which is
8043  * performing recovery.
8044  */
8045 static int _nfs40_proc_get_locations(struct nfs_server *server,
8046 				     struct nfs_fh *fhandle,
8047 				     struct nfs4_fs_locations *locations,
8048 				     struct page *page, const struct cred *cred)
8049 {
8050 	struct rpc_clnt *clnt = server->client;
8051 	u32 bitmask[2] = {
8052 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8053 	};
8054 	struct nfs4_fs_locations_arg args = {
8055 		.clientid	= server->nfs_client->cl_clientid,
8056 		.fh		= fhandle,
8057 		.page		= page,
8058 		.bitmask	= bitmask,
8059 		.migration	= 1,		/* skip LOOKUP */
8060 		.renew		= 1,		/* append RENEW */
8061 	};
8062 	struct nfs4_fs_locations_res res = {
8063 		.fs_locations	= locations,
8064 		.migration	= 1,
8065 		.renew		= 1,
8066 	};
8067 	struct rpc_message msg = {
8068 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8069 		.rpc_argp	= &args,
8070 		.rpc_resp	= &res,
8071 		.rpc_cred	= cred,
8072 	};
8073 	unsigned long now = jiffies;
8074 	int status;
8075 
8076 	nfs_fattr_init(locations->fattr);
8077 	locations->server = server;
8078 	locations->nlocations = 0;
8079 
8080 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8081 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8082 					&args.seq_args, &res.seq_res);
8083 	if (status)
8084 		return status;
8085 
8086 	renew_lease(server, now);
8087 	return 0;
8088 }
8089 
8090 #ifdef CONFIG_NFS_V4_1
8091 
8092 /*
8093  * This operation also signals the server that this client is
8094  * performing migration recovery.  The server can stop asserting
8095  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8096  * performing this operation is identified in the SEQUENCE
8097  * operation in this compound.
8098  *
8099  * When the client supports GETATTR(fs_locations_info), it can
8100  * be plumbed in here.
8101  */
8102 static int _nfs41_proc_get_locations(struct nfs_server *server,
8103 				     struct nfs_fh *fhandle,
8104 				     struct nfs4_fs_locations *locations,
8105 				     struct page *page, const struct cred *cred)
8106 {
8107 	struct rpc_clnt *clnt = server->client;
8108 	u32 bitmask[2] = {
8109 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8110 	};
8111 	struct nfs4_fs_locations_arg args = {
8112 		.fh		= fhandle,
8113 		.page		= page,
8114 		.bitmask	= bitmask,
8115 		.migration	= 1,		/* skip LOOKUP */
8116 	};
8117 	struct nfs4_fs_locations_res res = {
8118 		.fs_locations	= locations,
8119 		.migration	= 1,
8120 	};
8121 	struct rpc_message msg = {
8122 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8123 		.rpc_argp	= &args,
8124 		.rpc_resp	= &res,
8125 		.rpc_cred	= cred,
8126 	};
8127 	struct nfs4_call_sync_data data = {
8128 		.seq_server = server,
8129 		.seq_args = &args.seq_args,
8130 		.seq_res = &res.seq_res,
8131 	};
8132 	struct rpc_task_setup task_setup_data = {
8133 		.rpc_client = clnt,
8134 		.rpc_message = &msg,
8135 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8136 		.callback_data = &data,
8137 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8138 	};
8139 	int status;
8140 
8141 	nfs_fattr_init(locations->fattr);
8142 	locations->server = server;
8143 	locations->nlocations = 0;
8144 
8145 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8146 	status = nfs4_call_sync_custom(&task_setup_data);
8147 	if (status == NFS4_OK &&
8148 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8149 		status = -NFS4ERR_LEASE_MOVED;
8150 	return status;
8151 }
8152 
8153 #endif	/* CONFIG_NFS_V4_1 */
8154 
8155 /**
8156  * nfs4_proc_get_locations - discover locations for a migrated FSID
8157  * @server: pointer to nfs_server to process
8158  * @fhandle: pointer to the kernel NFS client file handle
8159  * @locations: result of query
8160  * @page: buffer
8161  * @cred: credential to use for this operation
8162  *
8163  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8164  * operation failed, or a negative errno if a local error occurred.
8165  *
8166  * On success, "locations" is filled in, but if the server has
8167  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8168  * asserted.
8169  *
8170  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8171  * from this client that require migration recovery.
8172  */
8173 int nfs4_proc_get_locations(struct nfs_server *server,
8174 			    struct nfs_fh *fhandle,
8175 			    struct nfs4_fs_locations *locations,
8176 			    struct page *page, const struct cred *cred)
8177 {
8178 	struct nfs_client *clp = server->nfs_client;
8179 	const struct nfs4_mig_recovery_ops *ops =
8180 					clp->cl_mvops->mig_recovery_ops;
8181 	struct nfs4_exception exception = {
8182 		.interruptible = true,
8183 	};
8184 	int status;
8185 
8186 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8187 		(unsigned long long)server->fsid.major,
8188 		(unsigned long long)server->fsid.minor,
8189 		clp->cl_hostname);
8190 	nfs_display_fhandle(fhandle, __func__);
8191 
8192 	do {
8193 		status = ops->get_locations(server, fhandle, locations, page,
8194 					    cred);
8195 		if (status != -NFS4ERR_DELAY)
8196 			break;
8197 		nfs4_handle_exception(server, status, &exception);
8198 	} while (exception.retry);
8199 	return status;
8200 }
8201 
8202 /*
8203  * This operation also signals the server that this client is
8204  * performing "lease moved" recovery.  The server can stop
8205  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8206  * is appended to this compound to identify the client ID which is
8207  * performing recovery.
8208  */
8209 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8210 {
8211 	struct nfs_server *server = NFS_SERVER(inode);
8212 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8213 	struct rpc_clnt *clnt = server->client;
8214 	struct nfs4_fsid_present_arg args = {
8215 		.fh		= NFS_FH(inode),
8216 		.clientid	= clp->cl_clientid,
8217 		.renew		= 1,		/* append RENEW */
8218 	};
8219 	struct nfs4_fsid_present_res res = {
8220 		.renew		= 1,
8221 	};
8222 	struct rpc_message msg = {
8223 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8224 		.rpc_argp	= &args,
8225 		.rpc_resp	= &res,
8226 		.rpc_cred	= cred,
8227 	};
8228 	unsigned long now = jiffies;
8229 	int status;
8230 
8231 	res.fh = nfs_alloc_fhandle();
8232 	if (res.fh == NULL)
8233 		return -ENOMEM;
8234 
8235 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8236 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8237 						&args.seq_args, &res.seq_res);
8238 	nfs_free_fhandle(res.fh);
8239 	if (status)
8240 		return status;
8241 
8242 	do_renew_lease(clp, now);
8243 	return 0;
8244 }
8245 
8246 #ifdef CONFIG_NFS_V4_1
8247 
8248 /*
8249  * This operation also signals the server that this client is
8250  * performing "lease moved" recovery.  The server can stop asserting
8251  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8252  * this operation is identified in the SEQUENCE operation in this
8253  * compound.
8254  */
8255 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8256 {
8257 	struct nfs_server *server = NFS_SERVER(inode);
8258 	struct rpc_clnt *clnt = server->client;
8259 	struct nfs4_fsid_present_arg args = {
8260 		.fh		= NFS_FH(inode),
8261 	};
8262 	struct nfs4_fsid_present_res res = {
8263 	};
8264 	struct rpc_message msg = {
8265 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8266 		.rpc_argp	= &args,
8267 		.rpc_resp	= &res,
8268 		.rpc_cred	= cred,
8269 	};
8270 	int status;
8271 
8272 	res.fh = nfs_alloc_fhandle();
8273 	if (res.fh == NULL)
8274 		return -ENOMEM;
8275 
8276 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8277 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8278 						&args.seq_args, &res.seq_res);
8279 	nfs_free_fhandle(res.fh);
8280 	if (status == NFS4_OK &&
8281 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8282 		status = -NFS4ERR_LEASE_MOVED;
8283 	return status;
8284 }
8285 
8286 #endif	/* CONFIG_NFS_V4_1 */
8287 
8288 /**
8289  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8290  * @inode: inode on FSID to check
8291  * @cred: credential to use for this operation
8292  *
8293  * Server indicates whether the FSID is present, moved, or not
8294  * recognized.  This operation is necessary to clear a LEASE_MOVED
8295  * condition for this client ID.
8296  *
8297  * Returns NFS4_OK if the FSID is present on this server,
8298  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8299  *  NFS4ERR code if some error occurred on the server, or a
8300  *  negative errno if a local failure occurred.
8301  */
8302 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8303 {
8304 	struct nfs_server *server = NFS_SERVER(inode);
8305 	struct nfs_client *clp = server->nfs_client;
8306 	const struct nfs4_mig_recovery_ops *ops =
8307 					clp->cl_mvops->mig_recovery_ops;
8308 	struct nfs4_exception exception = {
8309 		.interruptible = true,
8310 	};
8311 	int status;
8312 
8313 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8314 		(unsigned long long)server->fsid.major,
8315 		(unsigned long long)server->fsid.minor,
8316 		clp->cl_hostname);
8317 	nfs_display_fhandle(NFS_FH(inode), __func__);
8318 
8319 	do {
8320 		status = ops->fsid_present(inode, cred);
8321 		if (status != -NFS4ERR_DELAY)
8322 			break;
8323 		nfs4_handle_exception(server, status, &exception);
8324 	} while (exception.retry);
8325 	return status;
8326 }
8327 
8328 /*
8329  * If 'use_integrity' is true and the state managment nfs_client
8330  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8331  * and the machine credential as per RFC3530bis and RFC5661 Security
8332  * Considerations sections. Otherwise, just use the user cred with the
8333  * filesystem's rpc_client.
8334  */
8335 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8336 {
8337 	int status;
8338 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8339 	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8340 	struct nfs4_secinfo_arg args = {
8341 		.dir_fh = NFS_FH(dir),
8342 		.name   = name,
8343 	};
8344 	struct nfs4_secinfo_res res = {
8345 		.flavors     = flavors,
8346 	};
8347 	struct rpc_message msg = {
8348 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8349 		.rpc_argp = &args,
8350 		.rpc_resp = &res,
8351 	};
8352 	struct nfs4_call_sync_data data = {
8353 		.seq_server = NFS_SERVER(dir),
8354 		.seq_args = &args.seq_args,
8355 		.seq_res = &res.seq_res,
8356 	};
8357 	struct rpc_task_setup task_setup = {
8358 		.rpc_client = clnt,
8359 		.rpc_message = &msg,
8360 		.callback_ops = clp->cl_mvops->call_sync_ops,
8361 		.callback_data = &data,
8362 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8363 	};
8364 	const struct cred *cred = NULL;
8365 
8366 	if (use_integrity) {
8367 		clnt = clp->cl_rpcclient;
8368 		task_setup.rpc_client = clnt;
8369 
8370 		cred = nfs4_get_clid_cred(clp);
8371 		msg.rpc_cred = cred;
8372 	}
8373 
8374 	dprintk("NFS call  secinfo %s\n", name->name);
8375 
8376 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8377 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8378 	status = nfs4_call_sync_custom(&task_setup);
8379 
8380 	dprintk("NFS reply  secinfo: %d\n", status);
8381 
8382 	put_cred(cred);
8383 	return status;
8384 }
8385 
8386 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8387 		      struct nfs4_secinfo_flavors *flavors)
8388 {
8389 	struct nfs4_exception exception = {
8390 		.interruptible = true,
8391 	};
8392 	int err;
8393 	do {
8394 		err = -NFS4ERR_WRONGSEC;
8395 
8396 		/* try to use integrity protection with machine cred */
8397 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8398 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8399 
8400 		/*
8401 		 * if unable to use integrity protection, or SECINFO with
8402 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8403 		 * disallowed by spec, but exists in deployed servers) use
8404 		 * the current filesystem's rpc_client and the user cred.
8405 		 */
8406 		if (err == -NFS4ERR_WRONGSEC)
8407 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8408 
8409 		trace_nfs4_secinfo(dir, name, err);
8410 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8411 				&exception);
8412 	} while (exception.retry);
8413 	return err;
8414 }
8415 
8416 #ifdef CONFIG_NFS_V4_1
8417 /*
8418  * Check the exchange flags returned by the server for invalid flags, having
8419  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8420  * DS flags set.
8421  */
8422 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8423 {
8424 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8425 		goto out_inval;
8426 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8427 		goto out_inval;
8428 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8429 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8430 		goto out_inval;
8431 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8432 		goto out_inval;
8433 	return NFS_OK;
8434 out_inval:
8435 	return -NFS4ERR_INVAL;
8436 }
8437 
8438 static bool
8439 nfs41_same_server_scope(struct nfs41_server_scope *a,
8440 			struct nfs41_server_scope *b)
8441 {
8442 	if (a->server_scope_sz != b->server_scope_sz)
8443 		return false;
8444 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8445 }
8446 
8447 static void
8448 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8449 {
8450 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8451 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8452 	struct nfs_client *clp = args->client;
8453 
8454 	switch (task->tk_status) {
8455 	case -NFS4ERR_BADSESSION:
8456 	case -NFS4ERR_DEADSESSION:
8457 		nfs4_schedule_session_recovery(clp->cl_session,
8458 				task->tk_status);
8459 		return;
8460 	}
8461 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8462 			res->dir != NFS4_CDFS4_BOTH) {
8463 		rpc_task_close_connection(task);
8464 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8465 			rpc_restart_call(task);
8466 	}
8467 }
8468 
8469 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8470 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8471 };
8472 
8473 /*
8474  * nfs4_proc_bind_one_conn_to_session()
8475  *
8476  * The 4.1 client currently uses the same TCP connection for the
8477  * fore and backchannel.
8478  */
8479 static
8480 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8481 		struct rpc_xprt *xprt,
8482 		struct nfs_client *clp,
8483 		const struct cred *cred)
8484 {
8485 	int status;
8486 	struct nfs41_bind_conn_to_session_args args = {
8487 		.client = clp,
8488 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8489 		.retries = 0,
8490 	};
8491 	struct nfs41_bind_conn_to_session_res res;
8492 	struct rpc_message msg = {
8493 		.rpc_proc =
8494 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8495 		.rpc_argp = &args,
8496 		.rpc_resp = &res,
8497 		.rpc_cred = cred,
8498 	};
8499 	struct rpc_task_setup task_setup_data = {
8500 		.rpc_client = clnt,
8501 		.rpc_xprt = xprt,
8502 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8503 		.rpc_message = &msg,
8504 		.flags = RPC_TASK_TIMEOUT,
8505 	};
8506 	struct rpc_task *task;
8507 
8508 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8509 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8510 		args.dir = NFS4_CDFC4_FORE;
8511 
8512 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8513 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8514 		args.dir = NFS4_CDFC4_FORE;
8515 
8516 	task = rpc_run_task(&task_setup_data);
8517 	if (!IS_ERR(task)) {
8518 		status = task->tk_status;
8519 		rpc_put_task(task);
8520 	} else
8521 		status = PTR_ERR(task);
8522 	trace_nfs4_bind_conn_to_session(clp, status);
8523 	if (status == 0) {
8524 		if (memcmp(res.sessionid.data,
8525 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8526 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8527 			return -EIO;
8528 		}
8529 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8530 			dprintk("NFS: %s: Unexpected direction from server\n",
8531 				__func__);
8532 			return -EIO;
8533 		}
8534 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8535 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8536 				__func__);
8537 			return -EIO;
8538 		}
8539 	}
8540 
8541 	return status;
8542 }
8543 
8544 struct rpc_bind_conn_calldata {
8545 	struct nfs_client *clp;
8546 	const struct cred *cred;
8547 };
8548 
8549 static int
8550 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8551 		struct rpc_xprt *xprt,
8552 		void *calldata)
8553 {
8554 	struct rpc_bind_conn_calldata *p = calldata;
8555 
8556 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8557 }
8558 
8559 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8560 {
8561 	struct rpc_bind_conn_calldata data = {
8562 		.clp = clp,
8563 		.cred = cred,
8564 	};
8565 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8566 			nfs4_proc_bind_conn_to_session_callback, &data);
8567 }
8568 
8569 /*
8570  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8571  * and operations we'd like to see to enable certain features in the allow map
8572  */
8573 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8574 	.how = SP4_MACH_CRED,
8575 	.enforce.u.words = {
8576 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8577 		      1 << (OP_EXCHANGE_ID - 32) |
8578 		      1 << (OP_CREATE_SESSION - 32) |
8579 		      1 << (OP_DESTROY_SESSION - 32) |
8580 		      1 << (OP_DESTROY_CLIENTID - 32)
8581 	},
8582 	.allow.u.words = {
8583 		[0] = 1 << (OP_CLOSE) |
8584 		      1 << (OP_OPEN_DOWNGRADE) |
8585 		      1 << (OP_LOCKU) |
8586 		      1 << (OP_DELEGRETURN) |
8587 		      1 << (OP_COMMIT),
8588 		[1] = 1 << (OP_SECINFO - 32) |
8589 		      1 << (OP_SECINFO_NO_NAME - 32) |
8590 		      1 << (OP_LAYOUTRETURN - 32) |
8591 		      1 << (OP_TEST_STATEID - 32) |
8592 		      1 << (OP_FREE_STATEID - 32) |
8593 		      1 << (OP_WRITE - 32)
8594 	}
8595 };
8596 
8597 /*
8598  * Select the state protection mode for client `clp' given the server results
8599  * from exchange_id in `sp'.
8600  *
8601  * Returns 0 on success, negative errno otherwise.
8602  */
8603 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8604 				 struct nfs41_state_protection *sp)
8605 {
8606 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8607 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8608 		      1 << (OP_EXCHANGE_ID - 32) |
8609 		      1 << (OP_CREATE_SESSION - 32) |
8610 		      1 << (OP_DESTROY_SESSION - 32) |
8611 		      1 << (OP_DESTROY_CLIENTID - 32)
8612 	};
8613 	unsigned long flags = 0;
8614 	unsigned int i;
8615 	int ret = 0;
8616 
8617 	if (sp->how == SP4_MACH_CRED) {
8618 		/* Print state protect result */
8619 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8620 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8621 			if (test_bit(i, sp->enforce.u.longs))
8622 				dfprintk(MOUNT, "  enforce op %d\n", i);
8623 			if (test_bit(i, sp->allow.u.longs))
8624 				dfprintk(MOUNT, "  allow op %d\n", i);
8625 		}
8626 
8627 		/* make sure nothing is on enforce list that isn't supported */
8628 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8629 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8630 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8631 				ret = -EINVAL;
8632 				goto out;
8633 			}
8634 		}
8635 
8636 		/*
8637 		 * Minimal mode - state operations are allowed to use machine
8638 		 * credential.  Note this already happens by default, so the
8639 		 * client doesn't have to do anything more than the negotiation.
8640 		 *
8641 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8642 		 *       we're already using the machine cred for exchange_id
8643 		 *       and will never use a different cred.
8644 		 */
8645 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8646 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8647 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8648 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8649 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8650 			dfprintk(MOUNT, "  minimal mode enabled\n");
8651 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8652 		} else {
8653 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8654 			ret = -EINVAL;
8655 			goto out;
8656 		}
8657 
8658 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8659 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8660 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8661 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8662 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8663 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8664 		}
8665 
8666 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8667 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8668 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8669 		}
8670 
8671 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8672 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8673 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8674 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8675 		}
8676 
8677 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8678 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8679 			dfprintk(MOUNT, "  stateid mode enabled\n");
8680 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8681 		}
8682 
8683 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8684 			dfprintk(MOUNT, "  write mode enabled\n");
8685 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8686 		}
8687 
8688 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8689 			dfprintk(MOUNT, "  commit mode enabled\n");
8690 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8691 		}
8692 	}
8693 out:
8694 	clp->cl_sp4_flags = flags;
8695 	return ret;
8696 }
8697 
8698 struct nfs41_exchange_id_data {
8699 	struct nfs41_exchange_id_res res;
8700 	struct nfs41_exchange_id_args args;
8701 };
8702 
8703 static void nfs4_exchange_id_release(void *data)
8704 {
8705 	struct nfs41_exchange_id_data *cdata =
8706 					(struct nfs41_exchange_id_data *)data;
8707 
8708 	nfs_put_client(cdata->args.client);
8709 	kfree(cdata->res.impl_id);
8710 	kfree(cdata->res.server_scope);
8711 	kfree(cdata->res.server_owner);
8712 	kfree(cdata);
8713 }
8714 
8715 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8716 	.rpc_release = nfs4_exchange_id_release,
8717 };
8718 
8719 /*
8720  * _nfs4_proc_exchange_id()
8721  *
8722  * Wrapper for EXCHANGE_ID operation.
8723  */
8724 static struct rpc_task *
8725 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8726 			u32 sp4_how, struct rpc_xprt *xprt)
8727 {
8728 	struct rpc_message msg = {
8729 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8730 		.rpc_cred = cred,
8731 	};
8732 	struct rpc_task_setup task_setup_data = {
8733 		.rpc_client = clp->cl_rpcclient,
8734 		.callback_ops = &nfs4_exchange_id_call_ops,
8735 		.rpc_message = &msg,
8736 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8737 	};
8738 	struct nfs41_exchange_id_data *calldata;
8739 	int status;
8740 
8741 	if (!refcount_inc_not_zero(&clp->cl_count))
8742 		return ERR_PTR(-EIO);
8743 
8744 	status = -ENOMEM;
8745 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8746 	if (!calldata)
8747 		goto out;
8748 
8749 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8750 
8751 	status = nfs4_init_uniform_client_string(clp);
8752 	if (status)
8753 		goto out_calldata;
8754 
8755 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8756 						GFP_NOFS);
8757 	status = -ENOMEM;
8758 	if (unlikely(calldata->res.server_owner == NULL))
8759 		goto out_calldata;
8760 
8761 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8762 					GFP_NOFS);
8763 	if (unlikely(calldata->res.server_scope == NULL))
8764 		goto out_server_owner;
8765 
8766 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8767 	if (unlikely(calldata->res.impl_id == NULL))
8768 		goto out_server_scope;
8769 
8770 	switch (sp4_how) {
8771 	case SP4_NONE:
8772 		calldata->args.state_protect.how = SP4_NONE;
8773 		break;
8774 
8775 	case SP4_MACH_CRED:
8776 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8777 		break;
8778 
8779 	default:
8780 		/* unsupported! */
8781 		WARN_ON_ONCE(1);
8782 		status = -EINVAL;
8783 		goto out_impl_id;
8784 	}
8785 	if (xprt) {
8786 		task_setup_data.rpc_xprt = xprt;
8787 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8788 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8789 				sizeof(calldata->args.verifier.data));
8790 	}
8791 	calldata->args.client = clp;
8792 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8793 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8794 #ifdef CONFIG_NFS_V4_1_MIGRATION
8795 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8796 #endif
8797 	if (test_bit(NFS_CS_DS, &clp->cl_flags))
8798 		calldata->args.flags |= EXCHGID4_FLAG_USE_PNFS_DS;
8799 	msg.rpc_argp = &calldata->args;
8800 	msg.rpc_resp = &calldata->res;
8801 	task_setup_data.callback_data = calldata;
8802 
8803 	return rpc_run_task(&task_setup_data);
8804 
8805 out_impl_id:
8806 	kfree(calldata->res.impl_id);
8807 out_server_scope:
8808 	kfree(calldata->res.server_scope);
8809 out_server_owner:
8810 	kfree(calldata->res.server_owner);
8811 out_calldata:
8812 	kfree(calldata);
8813 out:
8814 	nfs_put_client(clp);
8815 	return ERR_PTR(status);
8816 }
8817 
8818 /*
8819  * _nfs4_proc_exchange_id()
8820  *
8821  * Wrapper for EXCHANGE_ID operation.
8822  */
8823 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8824 			u32 sp4_how)
8825 {
8826 	struct rpc_task *task;
8827 	struct nfs41_exchange_id_args *argp;
8828 	struct nfs41_exchange_id_res *resp;
8829 	unsigned long now = jiffies;
8830 	int status;
8831 
8832 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8833 	if (IS_ERR(task))
8834 		return PTR_ERR(task);
8835 
8836 	argp = task->tk_msg.rpc_argp;
8837 	resp = task->tk_msg.rpc_resp;
8838 	status = task->tk_status;
8839 	if (status  != 0)
8840 		goto out;
8841 
8842 	status = nfs4_check_cl_exchange_flags(resp->flags,
8843 			clp->cl_mvops->minor_version);
8844 	if (status  != 0)
8845 		goto out;
8846 
8847 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8848 	if (status != 0)
8849 		goto out;
8850 
8851 	do_renew_lease(clp, now);
8852 
8853 	clp->cl_clientid = resp->clientid;
8854 	clp->cl_exchange_flags = resp->flags;
8855 	clp->cl_seqid = resp->seqid;
8856 	/* Client ID is not confirmed */
8857 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8858 		clear_bit(NFS4_SESSION_ESTABLISHED,
8859 			  &clp->cl_session->session_state);
8860 
8861 	if (clp->cl_serverscope != NULL &&
8862 	    !nfs41_same_server_scope(clp->cl_serverscope,
8863 				resp->server_scope)) {
8864 		dprintk("%s: server_scope mismatch detected\n",
8865 			__func__);
8866 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8867 	}
8868 
8869 	swap(clp->cl_serverowner, resp->server_owner);
8870 	swap(clp->cl_serverscope, resp->server_scope);
8871 	swap(clp->cl_implid, resp->impl_id);
8872 
8873 	/* Save the EXCHANGE_ID verifier session trunk tests */
8874 	memcpy(clp->cl_confirm.data, argp->verifier.data,
8875 	       sizeof(clp->cl_confirm.data));
8876 out:
8877 	trace_nfs4_exchange_id(clp, status);
8878 	rpc_put_task(task);
8879 	return status;
8880 }
8881 
8882 /*
8883  * nfs4_proc_exchange_id()
8884  *
8885  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8886  *
8887  * Since the clientid has expired, all compounds using sessions
8888  * associated with the stale clientid will be returning
8889  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8890  * be in some phase of session reset.
8891  *
8892  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8893  */
8894 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8895 {
8896 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8897 	int status;
8898 
8899 	/* try SP4_MACH_CRED if krb5i/p	*/
8900 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8901 	    authflavor == RPC_AUTH_GSS_KRB5P) {
8902 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8903 		if (!status)
8904 			return 0;
8905 	}
8906 
8907 	/* try SP4_NONE */
8908 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8909 }
8910 
8911 /**
8912  * nfs4_test_session_trunk
8913  *
8914  * This is an add_xprt_test() test function called from
8915  * rpc_clnt_setup_test_and_add_xprt.
8916  *
8917  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8918  * and is dereferrenced in nfs4_exchange_id_release
8919  *
8920  * Upon success, add the new transport to the rpc_clnt
8921  *
8922  * @clnt: struct rpc_clnt to get new transport
8923  * @xprt: the rpc_xprt to test
8924  * @data: call data for _nfs4_proc_exchange_id.
8925  */
8926 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8927 			    void *data)
8928 {
8929 	struct nfs4_add_xprt_data *adata = data;
8930 	struct rpc_task *task;
8931 	int status;
8932 
8933 	u32 sp4_how;
8934 
8935 	dprintk("--> %s try %s\n", __func__,
8936 		xprt->address_strings[RPC_DISPLAY_ADDR]);
8937 
8938 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8939 
8940 try_again:
8941 	/* Test connection for session trunking. Async exchange_id call */
8942 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8943 	if (IS_ERR(task))
8944 		return;
8945 
8946 	status = task->tk_status;
8947 	if (status == 0)
8948 		status = nfs4_detect_session_trunking(adata->clp,
8949 				task->tk_msg.rpc_resp, xprt);
8950 
8951 	if (status == 0)
8952 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8953 	else if (status != -NFS4ERR_DELAY && rpc_clnt_xprt_switch_has_addr(clnt,
8954 				(struct sockaddr *)&xprt->addr))
8955 		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8956 
8957 	rpc_put_task(task);
8958 	if (status == -NFS4ERR_DELAY) {
8959 		ssleep(1);
8960 		goto try_again;
8961 	}
8962 }
8963 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8964 
8965 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8966 		const struct cred *cred)
8967 {
8968 	struct rpc_message msg = {
8969 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8970 		.rpc_argp = clp,
8971 		.rpc_cred = cred,
8972 	};
8973 	int status;
8974 
8975 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
8976 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8977 	trace_nfs4_destroy_clientid(clp, status);
8978 	if (status)
8979 		dprintk("NFS: Got error %d from the server %s on "
8980 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
8981 	return status;
8982 }
8983 
8984 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8985 		const struct cred *cred)
8986 {
8987 	unsigned int loop;
8988 	int ret;
8989 
8990 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8991 		ret = _nfs4_proc_destroy_clientid(clp, cred);
8992 		switch (ret) {
8993 		case -NFS4ERR_DELAY:
8994 		case -NFS4ERR_CLIENTID_BUSY:
8995 			ssleep(1);
8996 			break;
8997 		default:
8998 			return ret;
8999 		}
9000 	}
9001 	return 0;
9002 }
9003 
9004 int nfs4_destroy_clientid(struct nfs_client *clp)
9005 {
9006 	const struct cred *cred;
9007 	int ret = 0;
9008 
9009 	if (clp->cl_mvops->minor_version < 1)
9010 		goto out;
9011 	if (clp->cl_exchange_flags == 0)
9012 		goto out;
9013 	if (clp->cl_preserve_clid)
9014 		goto out;
9015 	cred = nfs4_get_clid_cred(clp);
9016 	ret = nfs4_proc_destroy_clientid(clp, cred);
9017 	put_cred(cred);
9018 	switch (ret) {
9019 	case 0:
9020 	case -NFS4ERR_STALE_CLIENTID:
9021 		clp->cl_exchange_flags = 0;
9022 	}
9023 out:
9024 	return ret;
9025 }
9026 
9027 #endif /* CONFIG_NFS_V4_1 */
9028 
9029 struct nfs4_get_lease_time_data {
9030 	struct nfs4_get_lease_time_args *args;
9031 	struct nfs4_get_lease_time_res *res;
9032 	struct nfs_client *clp;
9033 };
9034 
9035 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9036 					void *calldata)
9037 {
9038 	struct nfs4_get_lease_time_data *data =
9039 			(struct nfs4_get_lease_time_data *)calldata;
9040 
9041 	/* just setup sequence, do not trigger session recovery
9042 	   since we're invoked within one */
9043 	nfs4_setup_sequence(data->clp,
9044 			&data->args->la_seq_args,
9045 			&data->res->lr_seq_res,
9046 			task);
9047 }
9048 
9049 /*
9050  * Called from nfs4_state_manager thread for session setup, so don't recover
9051  * from sequence operation or clientid errors.
9052  */
9053 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9054 {
9055 	struct nfs4_get_lease_time_data *data =
9056 			(struct nfs4_get_lease_time_data *)calldata;
9057 
9058 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9059 		return;
9060 	switch (task->tk_status) {
9061 	case -NFS4ERR_DELAY:
9062 	case -NFS4ERR_GRACE:
9063 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9064 		task->tk_status = 0;
9065 		fallthrough;
9066 	case -NFS4ERR_RETRY_UNCACHED_REP:
9067 		rpc_restart_call_prepare(task);
9068 		return;
9069 	}
9070 }
9071 
9072 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9073 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9074 	.rpc_call_done = nfs4_get_lease_time_done,
9075 };
9076 
9077 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9078 {
9079 	struct nfs4_get_lease_time_args args;
9080 	struct nfs4_get_lease_time_res res = {
9081 		.lr_fsinfo = fsinfo,
9082 	};
9083 	struct nfs4_get_lease_time_data data = {
9084 		.args = &args,
9085 		.res = &res,
9086 		.clp = clp,
9087 	};
9088 	struct rpc_message msg = {
9089 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9090 		.rpc_argp = &args,
9091 		.rpc_resp = &res,
9092 	};
9093 	struct rpc_task_setup task_setup = {
9094 		.rpc_client = clp->cl_rpcclient,
9095 		.rpc_message = &msg,
9096 		.callback_ops = &nfs4_get_lease_time_ops,
9097 		.callback_data = &data,
9098 		.flags = RPC_TASK_TIMEOUT,
9099 	};
9100 
9101 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9102 	return nfs4_call_sync_custom(&task_setup);
9103 }
9104 
9105 #ifdef CONFIG_NFS_V4_1
9106 
9107 /*
9108  * Initialize the values to be used by the client in CREATE_SESSION
9109  * If nfs4_init_session set the fore channel request and response sizes,
9110  * use them.
9111  *
9112  * Set the back channel max_resp_sz_cached to zero to force the client to
9113  * always set csa_cachethis to FALSE because the current implementation
9114  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9115  */
9116 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9117 				    struct rpc_clnt *clnt)
9118 {
9119 	unsigned int max_rqst_sz, max_resp_sz;
9120 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9121 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9122 
9123 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9124 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9125 
9126 	/* Fore channel attributes */
9127 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9128 	args->fc_attrs.max_resp_sz = max_resp_sz;
9129 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9130 	args->fc_attrs.max_reqs = max_session_slots;
9131 
9132 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9133 		"max_ops=%u max_reqs=%u\n",
9134 		__func__,
9135 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9136 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9137 
9138 	/* Back channel attributes */
9139 	args->bc_attrs.max_rqst_sz = max_bc_payload;
9140 	args->bc_attrs.max_resp_sz = max_bc_payload;
9141 	args->bc_attrs.max_resp_sz_cached = 0;
9142 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9143 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9144 	if (args->bc_attrs.max_reqs > max_bc_slots)
9145 		args->bc_attrs.max_reqs = max_bc_slots;
9146 
9147 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9148 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9149 		__func__,
9150 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9151 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9152 		args->bc_attrs.max_reqs);
9153 }
9154 
9155 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9156 		struct nfs41_create_session_res *res)
9157 {
9158 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9159 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9160 
9161 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9162 		return -EINVAL;
9163 	/*
9164 	 * Our requested max_ops is the minimum we need; we're not
9165 	 * prepared to break up compounds into smaller pieces than that.
9166 	 * So, no point even trying to continue if the server won't
9167 	 * cooperate:
9168 	 */
9169 	if (rcvd->max_ops < sent->max_ops)
9170 		return -EINVAL;
9171 	if (rcvd->max_reqs == 0)
9172 		return -EINVAL;
9173 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9174 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9175 	return 0;
9176 }
9177 
9178 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9179 		struct nfs41_create_session_res *res)
9180 {
9181 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9182 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9183 
9184 	if (!(res->flags & SESSION4_BACK_CHAN))
9185 		goto out;
9186 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9187 		return -EINVAL;
9188 	if (rcvd->max_resp_sz < sent->max_resp_sz)
9189 		return -EINVAL;
9190 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9191 		return -EINVAL;
9192 	if (rcvd->max_ops > sent->max_ops)
9193 		return -EINVAL;
9194 	if (rcvd->max_reqs > sent->max_reqs)
9195 		return -EINVAL;
9196 out:
9197 	return 0;
9198 }
9199 
9200 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9201 				     struct nfs41_create_session_res *res)
9202 {
9203 	int ret;
9204 
9205 	ret = nfs4_verify_fore_channel_attrs(args, res);
9206 	if (ret)
9207 		return ret;
9208 	return nfs4_verify_back_channel_attrs(args, res);
9209 }
9210 
9211 static void nfs4_update_session(struct nfs4_session *session,
9212 		struct nfs41_create_session_res *res)
9213 {
9214 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9215 	/* Mark client id and session as being confirmed */
9216 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9217 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9218 	session->flags = res->flags;
9219 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9220 	if (res->flags & SESSION4_BACK_CHAN)
9221 		memcpy(&session->bc_attrs, &res->bc_attrs,
9222 				sizeof(session->bc_attrs));
9223 }
9224 
9225 static int _nfs4_proc_create_session(struct nfs_client *clp,
9226 		const struct cred *cred)
9227 {
9228 	struct nfs4_session *session = clp->cl_session;
9229 	struct nfs41_create_session_args args = {
9230 		.client = clp,
9231 		.clientid = clp->cl_clientid,
9232 		.seqid = clp->cl_seqid,
9233 		.cb_program = NFS4_CALLBACK,
9234 	};
9235 	struct nfs41_create_session_res res;
9236 
9237 	struct rpc_message msg = {
9238 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9239 		.rpc_argp = &args,
9240 		.rpc_resp = &res,
9241 		.rpc_cred = cred,
9242 	};
9243 	int status;
9244 
9245 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9246 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9247 
9248 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9249 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9250 	trace_nfs4_create_session(clp, status);
9251 
9252 	switch (status) {
9253 	case -NFS4ERR_STALE_CLIENTID:
9254 	case -NFS4ERR_DELAY:
9255 	case -ETIMEDOUT:
9256 	case -EACCES:
9257 	case -EAGAIN:
9258 		goto out;
9259 	}
9260 
9261 	clp->cl_seqid++;
9262 	if (!status) {
9263 		/* Verify the session's negotiated channel_attrs values */
9264 		status = nfs4_verify_channel_attrs(&args, &res);
9265 		/* Increment the clientid slot sequence id */
9266 		if (status)
9267 			goto out;
9268 		nfs4_update_session(session, &res);
9269 	}
9270 out:
9271 	return status;
9272 }
9273 
9274 /*
9275  * Issues a CREATE_SESSION operation to the server.
9276  * It is the responsibility of the caller to verify the session is
9277  * expired before calling this routine.
9278  */
9279 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9280 {
9281 	int status;
9282 	unsigned *ptr;
9283 	struct nfs4_session *session = clp->cl_session;
9284 	struct nfs4_add_xprt_data xprtdata = {
9285 		.clp = clp,
9286 	};
9287 	struct rpc_add_xprt_test rpcdata = {
9288 		.add_xprt_test = clp->cl_mvops->session_trunk,
9289 		.data = &xprtdata,
9290 	};
9291 
9292 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9293 
9294 	status = _nfs4_proc_create_session(clp, cred);
9295 	if (status)
9296 		goto out;
9297 
9298 	/* Init or reset the session slot tables */
9299 	status = nfs4_setup_session_slot_tables(session);
9300 	dprintk("slot table setup returned %d\n", status);
9301 	if (status)
9302 		goto out;
9303 
9304 	ptr = (unsigned *)&session->sess_id.data[0];
9305 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9306 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9307 	rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9308 out:
9309 	return status;
9310 }
9311 
9312 /*
9313  * Issue the over-the-wire RPC DESTROY_SESSION.
9314  * The caller must serialize access to this routine.
9315  */
9316 int nfs4_proc_destroy_session(struct nfs4_session *session,
9317 		const struct cred *cred)
9318 {
9319 	struct rpc_message msg = {
9320 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9321 		.rpc_argp = session,
9322 		.rpc_cred = cred,
9323 	};
9324 	int status = 0;
9325 
9326 	/* session is still being setup */
9327 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9328 		return 0;
9329 
9330 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9331 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9332 	trace_nfs4_destroy_session(session->clp, status);
9333 
9334 	if (status)
9335 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9336 			"Session has been destroyed regardless...\n", status);
9337 	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9338 	return status;
9339 }
9340 
9341 /*
9342  * Renew the cl_session lease.
9343  */
9344 struct nfs4_sequence_data {
9345 	struct nfs_client *clp;
9346 	struct nfs4_sequence_args args;
9347 	struct nfs4_sequence_res res;
9348 };
9349 
9350 static void nfs41_sequence_release(void *data)
9351 {
9352 	struct nfs4_sequence_data *calldata = data;
9353 	struct nfs_client *clp = calldata->clp;
9354 
9355 	if (refcount_read(&clp->cl_count) > 1)
9356 		nfs4_schedule_state_renewal(clp);
9357 	nfs_put_client(clp);
9358 	kfree(calldata);
9359 }
9360 
9361 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9362 {
9363 	switch(task->tk_status) {
9364 	case -NFS4ERR_DELAY:
9365 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9366 		return -EAGAIN;
9367 	default:
9368 		nfs4_schedule_lease_recovery(clp);
9369 	}
9370 	return 0;
9371 }
9372 
9373 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9374 {
9375 	struct nfs4_sequence_data *calldata = data;
9376 	struct nfs_client *clp = calldata->clp;
9377 
9378 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9379 		return;
9380 
9381 	trace_nfs4_sequence(clp, task->tk_status);
9382 	if (task->tk_status < 0 && !task->tk_client->cl_shutdown) {
9383 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9384 		if (refcount_read(&clp->cl_count) == 1)
9385 			return;
9386 
9387 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9388 			rpc_restart_call_prepare(task);
9389 			return;
9390 		}
9391 	}
9392 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9393 }
9394 
9395 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9396 {
9397 	struct nfs4_sequence_data *calldata = data;
9398 	struct nfs_client *clp = calldata->clp;
9399 	struct nfs4_sequence_args *args;
9400 	struct nfs4_sequence_res *res;
9401 
9402 	args = task->tk_msg.rpc_argp;
9403 	res = task->tk_msg.rpc_resp;
9404 
9405 	nfs4_setup_sequence(clp, args, res, task);
9406 }
9407 
9408 static const struct rpc_call_ops nfs41_sequence_ops = {
9409 	.rpc_call_done = nfs41_sequence_call_done,
9410 	.rpc_call_prepare = nfs41_sequence_prepare,
9411 	.rpc_release = nfs41_sequence_release,
9412 };
9413 
9414 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9415 		const struct cred *cred,
9416 		struct nfs4_slot *slot,
9417 		bool is_privileged)
9418 {
9419 	struct nfs4_sequence_data *calldata;
9420 	struct rpc_message msg = {
9421 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9422 		.rpc_cred = cred,
9423 	};
9424 	struct rpc_task_setup task_setup_data = {
9425 		.rpc_client = clp->cl_rpcclient,
9426 		.rpc_message = &msg,
9427 		.callback_ops = &nfs41_sequence_ops,
9428 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9429 	};
9430 	struct rpc_task *ret;
9431 
9432 	ret = ERR_PTR(-EIO);
9433 	if (!refcount_inc_not_zero(&clp->cl_count))
9434 		goto out_err;
9435 
9436 	ret = ERR_PTR(-ENOMEM);
9437 	calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9438 	if (calldata == NULL)
9439 		goto out_put_clp;
9440 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9441 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9442 	msg.rpc_argp = &calldata->args;
9443 	msg.rpc_resp = &calldata->res;
9444 	calldata->clp = clp;
9445 	task_setup_data.callback_data = calldata;
9446 
9447 	ret = rpc_run_task(&task_setup_data);
9448 	if (IS_ERR(ret))
9449 		goto out_err;
9450 	return ret;
9451 out_put_clp:
9452 	nfs_put_client(clp);
9453 out_err:
9454 	nfs41_release_slot(slot);
9455 	return ret;
9456 }
9457 
9458 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9459 {
9460 	struct rpc_task *task;
9461 	int ret = 0;
9462 
9463 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9464 		return -EAGAIN;
9465 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9466 	if (IS_ERR(task))
9467 		ret = PTR_ERR(task);
9468 	else
9469 		rpc_put_task_async(task);
9470 	dprintk("<-- %s status=%d\n", __func__, ret);
9471 	return ret;
9472 }
9473 
9474 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9475 {
9476 	struct rpc_task *task;
9477 	int ret;
9478 
9479 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9480 	if (IS_ERR(task)) {
9481 		ret = PTR_ERR(task);
9482 		goto out;
9483 	}
9484 	ret = rpc_wait_for_completion_task(task);
9485 	if (!ret)
9486 		ret = task->tk_status;
9487 	rpc_put_task(task);
9488 out:
9489 	dprintk("<-- %s status=%d\n", __func__, ret);
9490 	return ret;
9491 }
9492 
9493 struct nfs4_reclaim_complete_data {
9494 	struct nfs_client *clp;
9495 	struct nfs41_reclaim_complete_args arg;
9496 	struct nfs41_reclaim_complete_res res;
9497 };
9498 
9499 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9500 {
9501 	struct nfs4_reclaim_complete_data *calldata = data;
9502 
9503 	nfs4_setup_sequence(calldata->clp,
9504 			&calldata->arg.seq_args,
9505 			&calldata->res.seq_res,
9506 			task);
9507 }
9508 
9509 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9510 {
9511 	switch(task->tk_status) {
9512 	case 0:
9513 		wake_up_all(&clp->cl_lock_waitq);
9514 		fallthrough;
9515 	case -NFS4ERR_COMPLETE_ALREADY:
9516 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9517 		break;
9518 	case -NFS4ERR_DELAY:
9519 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9520 		fallthrough;
9521 	case -NFS4ERR_RETRY_UNCACHED_REP:
9522 	case -EACCES:
9523 		dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9524 			__func__, task->tk_status, clp->cl_hostname);
9525 		return -EAGAIN;
9526 	case -NFS4ERR_BADSESSION:
9527 	case -NFS4ERR_DEADSESSION:
9528 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9529 		break;
9530 	default:
9531 		nfs4_schedule_lease_recovery(clp);
9532 	}
9533 	return 0;
9534 }
9535 
9536 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9537 {
9538 	struct nfs4_reclaim_complete_data *calldata = data;
9539 	struct nfs_client *clp = calldata->clp;
9540 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9541 
9542 	if (!nfs41_sequence_done(task, res))
9543 		return;
9544 
9545 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9546 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9547 		rpc_restart_call_prepare(task);
9548 		return;
9549 	}
9550 }
9551 
9552 static void nfs4_free_reclaim_complete_data(void *data)
9553 {
9554 	struct nfs4_reclaim_complete_data *calldata = data;
9555 
9556 	kfree(calldata);
9557 }
9558 
9559 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9560 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9561 	.rpc_call_done = nfs4_reclaim_complete_done,
9562 	.rpc_release = nfs4_free_reclaim_complete_data,
9563 };
9564 
9565 /*
9566  * Issue a global reclaim complete.
9567  */
9568 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9569 		const struct cred *cred)
9570 {
9571 	struct nfs4_reclaim_complete_data *calldata;
9572 	struct rpc_message msg = {
9573 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9574 		.rpc_cred = cred,
9575 	};
9576 	struct rpc_task_setup task_setup_data = {
9577 		.rpc_client = clp->cl_rpcclient,
9578 		.rpc_message = &msg,
9579 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9580 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9581 	};
9582 	int status = -ENOMEM;
9583 
9584 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9585 	if (calldata == NULL)
9586 		goto out;
9587 	calldata->clp = clp;
9588 	calldata->arg.one_fs = 0;
9589 
9590 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9591 	msg.rpc_argp = &calldata->arg;
9592 	msg.rpc_resp = &calldata->res;
9593 	task_setup_data.callback_data = calldata;
9594 	status = nfs4_call_sync_custom(&task_setup_data);
9595 out:
9596 	dprintk("<-- %s status=%d\n", __func__, status);
9597 	return status;
9598 }
9599 
9600 static void
9601 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9602 {
9603 	struct nfs4_layoutget *lgp = calldata;
9604 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9605 
9606 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9607 				&lgp->res.seq_res, task);
9608 }
9609 
9610 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9611 {
9612 	struct nfs4_layoutget *lgp = calldata;
9613 
9614 	nfs41_sequence_process(task, &lgp->res.seq_res);
9615 }
9616 
9617 static int
9618 nfs4_layoutget_handle_exception(struct rpc_task *task,
9619 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9620 {
9621 	struct inode *inode = lgp->args.inode;
9622 	struct nfs_server *server = NFS_SERVER(inode);
9623 	struct pnfs_layout_hdr *lo = lgp->lo;
9624 	int nfs4err = task->tk_status;
9625 	int err, status = 0;
9626 	LIST_HEAD(head);
9627 
9628 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9629 
9630 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9631 
9632 	switch (nfs4err) {
9633 	case 0:
9634 		goto out;
9635 
9636 	/*
9637 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9638 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9639 	 * retry go inband.
9640 	 */
9641 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9642 		status = -ENODATA;
9643 		goto out;
9644 	/*
9645 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9646 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9647 	 */
9648 	case -NFS4ERR_BADLAYOUT:
9649 		status = -EOVERFLOW;
9650 		goto out;
9651 	/*
9652 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9653 	 * (or clients) writing to the same RAID stripe except when
9654 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9655 	 *
9656 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9657 	 * while, and then eventually give up.
9658 	 */
9659 	case -NFS4ERR_LAYOUTTRYLATER:
9660 		if (lgp->args.minlength == 0) {
9661 			status = -EOVERFLOW;
9662 			goto out;
9663 		}
9664 		status = -EBUSY;
9665 		break;
9666 	case -NFS4ERR_RECALLCONFLICT:
9667 	case -NFS4ERR_RETURNCONFLICT:
9668 		status = -ERECALLCONFLICT;
9669 		break;
9670 	case -NFS4ERR_DELEG_REVOKED:
9671 	case -NFS4ERR_ADMIN_REVOKED:
9672 	case -NFS4ERR_EXPIRED:
9673 	case -NFS4ERR_BAD_STATEID:
9674 		exception->timeout = 0;
9675 		spin_lock(&inode->i_lock);
9676 		/* If the open stateid was bad, then recover it. */
9677 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9678 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9679 			spin_unlock(&inode->i_lock);
9680 			exception->state = lgp->args.ctx->state;
9681 			exception->stateid = &lgp->args.stateid;
9682 			break;
9683 		}
9684 
9685 		/*
9686 		 * Mark the bad layout state as invalid, then retry
9687 		 */
9688 		pnfs_mark_layout_stateid_invalid(lo, &head);
9689 		spin_unlock(&inode->i_lock);
9690 		nfs_commit_inode(inode, 0);
9691 		pnfs_free_lseg_list(&head);
9692 		status = -EAGAIN;
9693 		goto out;
9694 	}
9695 
9696 	err = nfs4_handle_exception(server, nfs4err, exception);
9697 	if (!status) {
9698 		if (exception->retry)
9699 			status = -EAGAIN;
9700 		else
9701 			status = err;
9702 	}
9703 out:
9704 	return status;
9705 }
9706 
9707 size_t max_response_pages(struct nfs_server *server)
9708 {
9709 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9710 	return nfs_page_array_len(0, max_resp_sz);
9711 }
9712 
9713 static void nfs4_layoutget_release(void *calldata)
9714 {
9715 	struct nfs4_layoutget *lgp = calldata;
9716 
9717 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9718 	pnfs_layoutget_free(lgp);
9719 }
9720 
9721 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9722 	.rpc_call_prepare = nfs4_layoutget_prepare,
9723 	.rpc_call_done = nfs4_layoutget_done,
9724 	.rpc_release = nfs4_layoutget_release,
9725 };
9726 
9727 struct pnfs_layout_segment *
9728 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9729 {
9730 	struct inode *inode = lgp->args.inode;
9731 	struct nfs_server *server = NFS_SERVER(inode);
9732 	struct rpc_task *task;
9733 	struct rpc_message msg = {
9734 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9735 		.rpc_argp = &lgp->args,
9736 		.rpc_resp = &lgp->res,
9737 		.rpc_cred = lgp->cred,
9738 	};
9739 	struct rpc_task_setup task_setup_data = {
9740 		.rpc_client = server->client,
9741 		.rpc_message = &msg,
9742 		.callback_ops = &nfs4_layoutget_call_ops,
9743 		.callback_data = lgp,
9744 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9745 			 RPC_TASK_MOVEABLE,
9746 	};
9747 	struct pnfs_layout_segment *lseg = NULL;
9748 	struct nfs4_exception exception = {
9749 		.inode = inode,
9750 		.timeout = *timeout,
9751 	};
9752 	int status = 0;
9753 
9754 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9755 
9756 	task = rpc_run_task(&task_setup_data);
9757 	if (IS_ERR(task))
9758 		return ERR_CAST(task);
9759 
9760 	status = rpc_wait_for_completion_task(task);
9761 	if (status != 0)
9762 		goto out;
9763 
9764 	if (task->tk_status < 0) {
9765 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9766 		*timeout = exception.timeout;
9767 	} else if (lgp->res.layoutp->len == 0) {
9768 		status = -EAGAIN;
9769 		*timeout = nfs4_update_delay(&exception.timeout);
9770 	} else
9771 		lseg = pnfs_layout_process(lgp);
9772 out:
9773 	trace_nfs4_layoutget(lgp->args.ctx,
9774 			&lgp->args.range,
9775 			&lgp->res.range,
9776 			&lgp->res.stateid,
9777 			status);
9778 
9779 	rpc_put_task(task);
9780 	dprintk("<-- %s status=%d\n", __func__, status);
9781 	if (status)
9782 		return ERR_PTR(status);
9783 	return lseg;
9784 }
9785 
9786 static void
9787 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9788 {
9789 	struct nfs4_layoutreturn *lrp = calldata;
9790 
9791 	nfs4_setup_sequence(lrp->clp,
9792 			&lrp->args.seq_args,
9793 			&lrp->res.seq_res,
9794 			task);
9795 	if (!pnfs_layout_is_valid(lrp->args.layout))
9796 		rpc_exit(task, 0);
9797 }
9798 
9799 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9800 {
9801 	struct nfs4_layoutreturn *lrp = calldata;
9802 	struct nfs_server *server;
9803 
9804 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9805 		return;
9806 
9807 	/*
9808 	 * Was there an RPC level error? Assume the call succeeded,
9809 	 * and that we need to release the layout
9810 	 */
9811 	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9812 		lrp->res.lrs_present = 0;
9813 		return;
9814 	}
9815 
9816 	server = NFS_SERVER(lrp->args.inode);
9817 	switch (task->tk_status) {
9818 	case -NFS4ERR_OLD_STATEID:
9819 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9820 					&lrp->args.range,
9821 					lrp->args.inode))
9822 			goto out_restart;
9823 		fallthrough;
9824 	default:
9825 		task->tk_status = 0;
9826 		fallthrough;
9827 	case 0:
9828 		break;
9829 	case -NFS4ERR_DELAY:
9830 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9831 			break;
9832 		goto out_restart;
9833 	}
9834 	return;
9835 out_restart:
9836 	task->tk_status = 0;
9837 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9838 	rpc_restart_call_prepare(task);
9839 }
9840 
9841 static void nfs4_layoutreturn_release(void *calldata)
9842 {
9843 	struct nfs4_layoutreturn *lrp = calldata;
9844 	struct pnfs_layout_hdr *lo = lrp->args.layout;
9845 
9846 	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9847 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9848 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9849 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9850 		lrp->ld_private.ops->free(&lrp->ld_private);
9851 	pnfs_put_layout_hdr(lrp->args.layout);
9852 	nfs_iput_and_deactive(lrp->inode);
9853 	put_cred(lrp->cred);
9854 	kfree(calldata);
9855 }
9856 
9857 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9858 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9859 	.rpc_call_done = nfs4_layoutreturn_done,
9860 	.rpc_release = nfs4_layoutreturn_release,
9861 };
9862 
9863 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9864 {
9865 	struct rpc_task *task;
9866 	struct rpc_message msg = {
9867 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9868 		.rpc_argp = &lrp->args,
9869 		.rpc_resp = &lrp->res,
9870 		.rpc_cred = lrp->cred,
9871 	};
9872 	struct rpc_task_setup task_setup_data = {
9873 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9874 		.rpc_message = &msg,
9875 		.callback_ops = &nfs4_layoutreturn_call_ops,
9876 		.callback_data = lrp,
9877 		.flags = RPC_TASK_MOVEABLE,
9878 	};
9879 	int status = 0;
9880 
9881 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9882 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9883 			&task_setup_data.rpc_client, &msg);
9884 
9885 	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9886 	if (!sync) {
9887 		if (!lrp->inode) {
9888 			nfs4_layoutreturn_release(lrp);
9889 			return -EAGAIN;
9890 		}
9891 		task_setup_data.flags |= RPC_TASK_ASYNC;
9892 	}
9893 	if (!lrp->inode)
9894 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9895 				   1);
9896 	else
9897 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9898 				   0);
9899 	task = rpc_run_task(&task_setup_data);
9900 	if (IS_ERR(task))
9901 		return PTR_ERR(task);
9902 	if (sync)
9903 		status = task->tk_status;
9904 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9905 	dprintk("<-- %s status=%d\n", __func__, status);
9906 	rpc_put_task(task);
9907 	return status;
9908 }
9909 
9910 static int
9911 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9912 		struct pnfs_device *pdev,
9913 		const struct cred *cred)
9914 {
9915 	struct nfs4_getdeviceinfo_args args = {
9916 		.pdev = pdev,
9917 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9918 			NOTIFY_DEVICEID4_DELETE,
9919 	};
9920 	struct nfs4_getdeviceinfo_res res = {
9921 		.pdev = pdev,
9922 	};
9923 	struct rpc_message msg = {
9924 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9925 		.rpc_argp = &args,
9926 		.rpc_resp = &res,
9927 		.rpc_cred = cred,
9928 	};
9929 	int status;
9930 
9931 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9932 	if (res.notification & ~args.notify_types)
9933 		dprintk("%s: unsupported notification\n", __func__);
9934 	if (res.notification != args.notify_types)
9935 		pdev->nocache = 1;
9936 
9937 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9938 
9939 	dprintk("<-- %s status=%d\n", __func__, status);
9940 
9941 	return status;
9942 }
9943 
9944 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9945 		struct pnfs_device *pdev,
9946 		const struct cred *cred)
9947 {
9948 	struct nfs4_exception exception = { };
9949 	int err;
9950 
9951 	do {
9952 		err = nfs4_handle_exception(server,
9953 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9954 					&exception);
9955 	} while (exception.retry);
9956 	return err;
9957 }
9958 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9959 
9960 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9961 {
9962 	struct nfs4_layoutcommit_data *data = calldata;
9963 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9964 
9965 	nfs4_setup_sequence(server->nfs_client,
9966 			&data->args.seq_args,
9967 			&data->res.seq_res,
9968 			task);
9969 }
9970 
9971 static void
9972 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9973 {
9974 	struct nfs4_layoutcommit_data *data = calldata;
9975 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9976 
9977 	if (!nfs41_sequence_done(task, &data->res.seq_res))
9978 		return;
9979 
9980 	switch (task->tk_status) { /* Just ignore these failures */
9981 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9982 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9983 	case -NFS4ERR_BADLAYOUT:     /* no layout */
9984 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
9985 		task->tk_status = 0;
9986 		break;
9987 	case 0:
9988 		break;
9989 	default:
9990 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9991 			rpc_restart_call_prepare(task);
9992 			return;
9993 		}
9994 	}
9995 }
9996 
9997 static void nfs4_layoutcommit_release(void *calldata)
9998 {
9999 	struct nfs4_layoutcommit_data *data = calldata;
10000 
10001 	pnfs_cleanup_layoutcommit(data);
10002 	nfs_post_op_update_inode_force_wcc(data->args.inode,
10003 					   data->res.fattr);
10004 	put_cred(data->cred);
10005 	nfs_iput_and_deactive(data->inode);
10006 	kfree(data);
10007 }
10008 
10009 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
10010 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
10011 	.rpc_call_done = nfs4_layoutcommit_done,
10012 	.rpc_release = nfs4_layoutcommit_release,
10013 };
10014 
10015 int
10016 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10017 {
10018 	struct rpc_message msg = {
10019 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10020 		.rpc_argp = &data->args,
10021 		.rpc_resp = &data->res,
10022 		.rpc_cred = data->cred,
10023 	};
10024 	struct rpc_task_setup task_setup_data = {
10025 		.task = &data->task,
10026 		.rpc_client = NFS_CLIENT(data->args.inode),
10027 		.rpc_message = &msg,
10028 		.callback_ops = &nfs4_layoutcommit_ops,
10029 		.callback_data = data,
10030 		.flags = RPC_TASK_MOVEABLE,
10031 	};
10032 	struct rpc_task *task;
10033 	int status = 0;
10034 
10035 	dprintk("NFS: initiating layoutcommit call. sync %d "
10036 		"lbw: %llu inode %lu\n", sync,
10037 		data->args.lastbytewritten,
10038 		data->args.inode->i_ino);
10039 
10040 	if (!sync) {
10041 		data->inode = nfs_igrab_and_active(data->args.inode);
10042 		if (data->inode == NULL) {
10043 			nfs4_layoutcommit_release(data);
10044 			return -EAGAIN;
10045 		}
10046 		task_setup_data.flags = RPC_TASK_ASYNC;
10047 	}
10048 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10049 	task = rpc_run_task(&task_setup_data);
10050 	if (IS_ERR(task))
10051 		return PTR_ERR(task);
10052 	if (sync)
10053 		status = task->tk_status;
10054 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10055 	dprintk("%s: status %d\n", __func__, status);
10056 	rpc_put_task(task);
10057 	return status;
10058 }
10059 
10060 /*
10061  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10062  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10063  */
10064 static int
10065 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10066 		    struct nfs_fsinfo *info,
10067 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10068 {
10069 	struct nfs41_secinfo_no_name_args args = {
10070 		.style = SECINFO_STYLE_CURRENT_FH,
10071 	};
10072 	struct nfs4_secinfo_res res = {
10073 		.flavors = flavors,
10074 	};
10075 	struct rpc_message msg = {
10076 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10077 		.rpc_argp = &args,
10078 		.rpc_resp = &res,
10079 	};
10080 	struct nfs4_call_sync_data data = {
10081 		.seq_server = server,
10082 		.seq_args = &args.seq_args,
10083 		.seq_res = &res.seq_res,
10084 	};
10085 	struct rpc_task_setup task_setup = {
10086 		.rpc_client = server->client,
10087 		.rpc_message = &msg,
10088 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10089 		.callback_data = &data,
10090 		.flags = RPC_TASK_NO_ROUND_ROBIN,
10091 	};
10092 	const struct cred *cred = NULL;
10093 	int status;
10094 
10095 	if (use_integrity) {
10096 		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10097 
10098 		cred = nfs4_get_clid_cred(server->nfs_client);
10099 		msg.rpc_cred = cred;
10100 	}
10101 
10102 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10103 	status = nfs4_call_sync_custom(&task_setup);
10104 	dprintk("<-- %s status=%d\n", __func__, status);
10105 
10106 	put_cred(cred);
10107 
10108 	return status;
10109 }
10110 
10111 static int
10112 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10113 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10114 {
10115 	struct nfs4_exception exception = {
10116 		.interruptible = true,
10117 	};
10118 	int err;
10119 	do {
10120 		/* first try using integrity protection */
10121 		err = -NFS4ERR_WRONGSEC;
10122 
10123 		/* try to use integrity protection with machine cred */
10124 		if (_nfs4_is_integrity_protected(server->nfs_client))
10125 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10126 							  flavors, true);
10127 
10128 		/*
10129 		 * if unable to use integrity protection, or SECINFO with
10130 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10131 		 * disallowed by spec, but exists in deployed servers) use
10132 		 * the current filesystem's rpc_client and the user cred.
10133 		 */
10134 		if (err == -NFS4ERR_WRONGSEC)
10135 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10136 							  flavors, false);
10137 
10138 		switch (err) {
10139 		case 0:
10140 		case -NFS4ERR_WRONGSEC:
10141 		case -ENOTSUPP:
10142 			goto out;
10143 		default:
10144 			err = nfs4_handle_exception(server, err, &exception);
10145 		}
10146 	} while (exception.retry);
10147 out:
10148 	return err;
10149 }
10150 
10151 static int
10152 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10153 		    struct nfs_fsinfo *info)
10154 {
10155 	int err;
10156 	struct page *page;
10157 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10158 	struct nfs4_secinfo_flavors *flavors;
10159 	struct nfs4_secinfo4 *secinfo;
10160 	int i;
10161 
10162 	page = alloc_page(GFP_KERNEL);
10163 	if (!page) {
10164 		err = -ENOMEM;
10165 		goto out;
10166 	}
10167 
10168 	flavors = page_address(page);
10169 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10170 
10171 	/*
10172 	 * Fall back on "guess and check" method if
10173 	 * the server doesn't support SECINFO_NO_NAME
10174 	 */
10175 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10176 		err = nfs4_find_root_sec(server, fhandle, info);
10177 		goto out_freepage;
10178 	}
10179 	if (err)
10180 		goto out_freepage;
10181 
10182 	for (i = 0; i < flavors->num_flavors; i++) {
10183 		secinfo = &flavors->flavors[i];
10184 
10185 		switch (secinfo->flavor) {
10186 		case RPC_AUTH_NULL:
10187 		case RPC_AUTH_UNIX:
10188 		case RPC_AUTH_GSS:
10189 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10190 					&secinfo->flavor_info);
10191 			break;
10192 		default:
10193 			flavor = RPC_AUTH_MAXFLAVOR;
10194 			break;
10195 		}
10196 
10197 		if (!nfs_auth_info_match(&server->auth_info, flavor))
10198 			flavor = RPC_AUTH_MAXFLAVOR;
10199 
10200 		if (flavor != RPC_AUTH_MAXFLAVOR) {
10201 			err = nfs4_lookup_root_sec(server, fhandle,
10202 						   info, flavor);
10203 			if (!err)
10204 				break;
10205 		}
10206 	}
10207 
10208 	if (flavor == RPC_AUTH_MAXFLAVOR)
10209 		err = -EPERM;
10210 
10211 out_freepage:
10212 	put_page(page);
10213 	if (err == -EACCES)
10214 		return -EPERM;
10215 out:
10216 	return err;
10217 }
10218 
10219 static int _nfs41_test_stateid(struct nfs_server *server,
10220 		nfs4_stateid *stateid,
10221 		const struct cred *cred)
10222 {
10223 	int status;
10224 	struct nfs41_test_stateid_args args = {
10225 		.stateid = stateid,
10226 	};
10227 	struct nfs41_test_stateid_res res;
10228 	struct rpc_message msg = {
10229 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10230 		.rpc_argp = &args,
10231 		.rpc_resp = &res,
10232 		.rpc_cred = cred,
10233 	};
10234 	struct rpc_clnt *rpc_client = server->client;
10235 
10236 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10237 		&rpc_client, &msg);
10238 
10239 	dprintk("NFS call  test_stateid %p\n", stateid);
10240 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10241 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10242 			&args.seq_args, &res.seq_res);
10243 	if (status != NFS_OK) {
10244 		dprintk("NFS reply test_stateid: failed, %d\n", status);
10245 		return status;
10246 	}
10247 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10248 	return -res.status;
10249 }
10250 
10251 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10252 		int err, struct nfs4_exception *exception)
10253 {
10254 	exception->retry = 0;
10255 	switch(err) {
10256 	case -NFS4ERR_DELAY:
10257 	case -NFS4ERR_RETRY_UNCACHED_REP:
10258 		nfs4_handle_exception(server, err, exception);
10259 		break;
10260 	case -NFS4ERR_BADSESSION:
10261 	case -NFS4ERR_BADSLOT:
10262 	case -NFS4ERR_BAD_HIGH_SLOT:
10263 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10264 	case -NFS4ERR_DEADSESSION:
10265 		nfs4_do_handle_exception(server, err, exception);
10266 	}
10267 }
10268 
10269 /**
10270  * nfs41_test_stateid - perform a TEST_STATEID operation
10271  *
10272  * @server: server / transport on which to perform the operation
10273  * @stateid: state ID to test
10274  * @cred: credential
10275  *
10276  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10277  * Otherwise a negative NFS4ERR value is returned if the operation
10278  * failed or the state ID is not currently valid.
10279  */
10280 static int nfs41_test_stateid(struct nfs_server *server,
10281 		nfs4_stateid *stateid,
10282 		const struct cred *cred)
10283 {
10284 	struct nfs4_exception exception = {
10285 		.interruptible = true,
10286 	};
10287 	int err;
10288 	do {
10289 		err = _nfs41_test_stateid(server, stateid, cred);
10290 		nfs4_handle_delay_or_session_error(server, err, &exception);
10291 	} while (exception.retry);
10292 	return err;
10293 }
10294 
10295 struct nfs_free_stateid_data {
10296 	struct nfs_server *server;
10297 	struct nfs41_free_stateid_args args;
10298 	struct nfs41_free_stateid_res res;
10299 };
10300 
10301 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10302 {
10303 	struct nfs_free_stateid_data *data = calldata;
10304 	nfs4_setup_sequence(data->server->nfs_client,
10305 			&data->args.seq_args,
10306 			&data->res.seq_res,
10307 			task);
10308 }
10309 
10310 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10311 {
10312 	struct nfs_free_stateid_data *data = calldata;
10313 
10314 	nfs41_sequence_done(task, &data->res.seq_res);
10315 
10316 	switch (task->tk_status) {
10317 	case -NFS4ERR_DELAY:
10318 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10319 			rpc_restart_call_prepare(task);
10320 	}
10321 }
10322 
10323 static void nfs41_free_stateid_release(void *calldata)
10324 {
10325 	struct nfs_free_stateid_data *data = calldata;
10326 	struct nfs_client *clp = data->server->nfs_client;
10327 
10328 	nfs_put_client(clp);
10329 	kfree(calldata);
10330 }
10331 
10332 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10333 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10334 	.rpc_call_done = nfs41_free_stateid_done,
10335 	.rpc_release = nfs41_free_stateid_release,
10336 };
10337 
10338 /**
10339  * nfs41_free_stateid - perform a FREE_STATEID operation
10340  *
10341  * @server: server / transport on which to perform the operation
10342  * @stateid: state ID to release
10343  * @cred: credential
10344  * @privileged: set to true if this call needs to be privileged
10345  *
10346  * Note: this function is always asynchronous.
10347  */
10348 static int nfs41_free_stateid(struct nfs_server *server,
10349 		const nfs4_stateid *stateid,
10350 		const struct cred *cred,
10351 		bool privileged)
10352 {
10353 	struct rpc_message msg = {
10354 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10355 		.rpc_cred = cred,
10356 	};
10357 	struct rpc_task_setup task_setup = {
10358 		.rpc_client = server->client,
10359 		.rpc_message = &msg,
10360 		.callback_ops = &nfs41_free_stateid_ops,
10361 		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10362 	};
10363 	struct nfs_free_stateid_data *data;
10364 	struct rpc_task *task;
10365 	struct nfs_client *clp = server->nfs_client;
10366 
10367 	if (!refcount_inc_not_zero(&clp->cl_count))
10368 		return -EIO;
10369 
10370 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10371 		&task_setup.rpc_client, &msg);
10372 
10373 	dprintk("NFS call  free_stateid %p\n", stateid);
10374 	data = kmalloc(sizeof(*data), GFP_KERNEL);
10375 	if (!data)
10376 		return -ENOMEM;
10377 	data->server = server;
10378 	nfs4_stateid_copy(&data->args.stateid, stateid);
10379 
10380 	task_setup.callback_data = data;
10381 
10382 	msg.rpc_argp = &data->args;
10383 	msg.rpc_resp = &data->res;
10384 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10385 	task = rpc_run_task(&task_setup);
10386 	if (IS_ERR(task))
10387 		return PTR_ERR(task);
10388 	rpc_put_task(task);
10389 	return 0;
10390 }
10391 
10392 static void
10393 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10394 {
10395 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10396 
10397 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10398 	nfs4_free_lock_state(server, lsp);
10399 }
10400 
10401 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10402 		const nfs4_stateid *s2)
10403 {
10404 	if (s1->type != s2->type)
10405 		return false;
10406 
10407 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10408 		return false;
10409 
10410 	if (s1->seqid == s2->seqid)
10411 		return true;
10412 
10413 	return s1->seqid == 0 || s2->seqid == 0;
10414 }
10415 
10416 #endif /* CONFIG_NFS_V4_1 */
10417 
10418 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10419 		const nfs4_stateid *s2)
10420 {
10421 	return nfs4_stateid_match(s1, s2);
10422 }
10423 
10424 
10425 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10426 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10427 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10428 	.recover_open	= nfs4_open_reclaim,
10429 	.recover_lock	= nfs4_lock_reclaim,
10430 	.establish_clid = nfs4_init_clientid,
10431 	.detect_trunking = nfs40_discover_server_trunking,
10432 };
10433 
10434 #if defined(CONFIG_NFS_V4_1)
10435 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10436 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10437 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10438 	.recover_open	= nfs4_open_reclaim,
10439 	.recover_lock	= nfs4_lock_reclaim,
10440 	.establish_clid = nfs41_init_clientid,
10441 	.reclaim_complete = nfs41_proc_reclaim_complete,
10442 	.detect_trunking = nfs41_discover_server_trunking,
10443 };
10444 #endif /* CONFIG_NFS_V4_1 */
10445 
10446 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10447 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10448 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10449 	.recover_open	= nfs40_open_expired,
10450 	.recover_lock	= nfs4_lock_expired,
10451 	.establish_clid = nfs4_init_clientid,
10452 };
10453 
10454 #if defined(CONFIG_NFS_V4_1)
10455 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10456 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10457 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10458 	.recover_open	= nfs41_open_expired,
10459 	.recover_lock	= nfs41_lock_expired,
10460 	.establish_clid = nfs41_init_clientid,
10461 };
10462 #endif /* CONFIG_NFS_V4_1 */
10463 
10464 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10465 	.sched_state_renewal = nfs4_proc_async_renew,
10466 	.get_state_renewal_cred = nfs4_get_renew_cred,
10467 	.renew_lease = nfs4_proc_renew,
10468 };
10469 
10470 #if defined(CONFIG_NFS_V4_1)
10471 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10472 	.sched_state_renewal = nfs41_proc_async_sequence,
10473 	.get_state_renewal_cred = nfs4_get_machine_cred,
10474 	.renew_lease = nfs4_proc_sequence,
10475 };
10476 #endif
10477 
10478 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10479 	.get_locations = _nfs40_proc_get_locations,
10480 	.fsid_present = _nfs40_proc_fsid_present,
10481 };
10482 
10483 #if defined(CONFIG_NFS_V4_1)
10484 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10485 	.get_locations = _nfs41_proc_get_locations,
10486 	.fsid_present = _nfs41_proc_fsid_present,
10487 };
10488 #endif	/* CONFIG_NFS_V4_1 */
10489 
10490 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10491 	.minor_version = 0,
10492 	.init_caps = NFS_CAP_READDIRPLUS
10493 		| NFS_CAP_ATOMIC_OPEN
10494 		| NFS_CAP_POSIX_LOCK,
10495 	.init_client = nfs40_init_client,
10496 	.shutdown_client = nfs40_shutdown_client,
10497 	.match_stateid = nfs4_match_stateid,
10498 	.find_root_sec = nfs4_find_root_sec,
10499 	.free_lock_state = nfs4_release_lockowner,
10500 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10501 	.alloc_seqid = nfs_alloc_seqid,
10502 	.call_sync_ops = &nfs40_call_sync_ops,
10503 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10504 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10505 	.state_renewal_ops = &nfs40_state_renewal_ops,
10506 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10507 };
10508 
10509 #if defined(CONFIG_NFS_V4_1)
10510 static struct nfs_seqid *
10511 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10512 {
10513 	return NULL;
10514 }
10515 
10516 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10517 	.minor_version = 1,
10518 	.init_caps = NFS_CAP_READDIRPLUS
10519 		| NFS_CAP_ATOMIC_OPEN
10520 		| NFS_CAP_POSIX_LOCK
10521 		| NFS_CAP_STATEID_NFSV41
10522 		| NFS_CAP_ATOMIC_OPEN_V1
10523 		| NFS_CAP_LGOPEN
10524 		| NFS_CAP_MOVEABLE,
10525 	.init_client = nfs41_init_client,
10526 	.shutdown_client = nfs41_shutdown_client,
10527 	.match_stateid = nfs41_match_stateid,
10528 	.find_root_sec = nfs41_find_root_sec,
10529 	.free_lock_state = nfs41_free_lock_state,
10530 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10531 	.alloc_seqid = nfs_alloc_no_seqid,
10532 	.session_trunk = nfs4_test_session_trunk,
10533 	.call_sync_ops = &nfs41_call_sync_ops,
10534 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10535 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10536 	.state_renewal_ops = &nfs41_state_renewal_ops,
10537 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10538 };
10539 #endif
10540 
10541 #if defined(CONFIG_NFS_V4_2)
10542 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10543 	.minor_version = 2,
10544 	.init_caps = NFS_CAP_READDIRPLUS
10545 		| NFS_CAP_ATOMIC_OPEN
10546 		| NFS_CAP_POSIX_LOCK
10547 		| NFS_CAP_STATEID_NFSV41
10548 		| NFS_CAP_ATOMIC_OPEN_V1
10549 		| NFS_CAP_LGOPEN
10550 		| NFS_CAP_ALLOCATE
10551 		| NFS_CAP_COPY
10552 		| NFS_CAP_OFFLOAD_CANCEL
10553 		| NFS_CAP_COPY_NOTIFY
10554 		| NFS_CAP_DEALLOCATE
10555 		| NFS_CAP_SEEK
10556 		| NFS_CAP_LAYOUTSTATS
10557 		| NFS_CAP_CLONE
10558 		| NFS_CAP_LAYOUTERROR
10559 		| NFS_CAP_READ_PLUS
10560 		| NFS_CAP_MOVEABLE,
10561 	.init_client = nfs41_init_client,
10562 	.shutdown_client = nfs41_shutdown_client,
10563 	.match_stateid = nfs41_match_stateid,
10564 	.find_root_sec = nfs41_find_root_sec,
10565 	.free_lock_state = nfs41_free_lock_state,
10566 	.call_sync_ops = &nfs41_call_sync_ops,
10567 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10568 	.alloc_seqid = nfs_alloc_no_seqid,
10569 	.session_trunk = nfs4_test_session_trunk,
10570 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10571 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10572 	.state_renewal_ops = &nfs41_state_renewal_ops,
10573 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10574 };
10575 #endif
10576 
10577 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10578 	[0] = &nfs_v4_0_minor_ops,
10579 #if defined(CONFIG_NFS_V4_1)
10580 	[1] = &nfs_v4_1_minor_ops,
10581 #endif
10582 #if defined(CONFIG_NFS_V4_2)
10583 	[2] = &nfs_v4_2_minor_ops,
10584 #endif
10585 };
10586 
10587 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10588 {
10589 	ssize_t error, error2, error3;
10590 	size_t left = size;
10591 
10592 	error = generic_listxattr(dentry, list, left);
10593 	if (error < 0)
10594 		return error;
10595 	if (list) {
10596 		list += error;
10597 		left -= error;
10598 	}
10599 
10600 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, left);
10601 	if (error2 < 0)
10602 		return error2;
10603 
10604 	if (list) {
10605 		list += error2;
10606 		left -= error2;
10607 	}
10608 
10609 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, left);
10610 	if (error3 < 0)
10611 		return error3;
10612 
10613 	error += error2 + error3;
10614 	if (size && error > size)
10615 		return -ERANGE;
10616 	return error;
10617 }
10618 
10619 static void nfs4_enable_swap(struct inode *inode)
10620 {
10621 	/* The state manager thread must always be running.
10622 	 * It will notice the client is a swapper, and stay put.
10623 	 */
10624 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10625 
10626 	nfs4_schedule_state_manager(clp);
10627 }
10628 
10629 static void nfs4_disable_swap(struct inode *inode)
10630 {
10631 	/* The state manager thread will now exit once it is
10632 	 * woken.
10633 	 */
10634 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10635 
10636 	set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10637 	clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10638 	wake_up_var(&clp->cl_state);
10639 }
10640 
10641 static const struct inode_operations nfs4_dir_inode_operations = {
10642 	.create		= nfs_create,
10643 	.lookup		= nfs_lookup,
10644 	.atomic_open	= nfs_atomic_open,
10645 	.link		= nfs_link,
10646 	.unlink		= nfs_unlink,
10647 	.symlink	= nfs_symlink,
10648 	.mkdir		= nfs_mkdir,
10649 	.rmdir		= nfs_rmdir,
10650 	.mknod		= nfs_mknod,
10651 	.rename		= nfs_rename,
10652 	.permission	= nfs_permission,
10653 	.getattr	= nfs_getattr,
10654 	.setattr	= nfs_setattr,
10655 	.listxattr	= nfs4_listxattr,
10656 };
10657 
10658 static const struct inode_operations nfs4_file_inode_operations = {
10659 	.permission	= nfs_permission,
10660 	.getattr	= nfs_getattr,
10661 	.setattr	= nfs_setattr,
10662 	.listxattr	= nfs4_listxattr,
10663 };
10664 
10665 const struct nfs_rpc_ops nfs_v4_clientops = {
10666 	.version	= 4,			/* protocol version */
10667 	.dentry_ops	= &nfs4_dentry_operations,
10668 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10669 	.file_inode_ops	= &nfs4_file_inode_operations,
10670 	.file_ops	= &nfs4_file_operations,
10671 	.getroot	= nfs4_proc_get_root,
10672 	.submount	= nfs4_submount,
10673 	.try_get_tree	= nfs4_try_get_tree,
10674 	.getattr	= nfs4_proc_getattr,
10675 	.setattr	= nfs4_proc_setattr,
10676 	.lookup		= nfs4_proc_lookup,
10677 	.lookupp	= nfs4_proc_lookupp,
10678 	.access		= nfs4_proc_access,
10679 	.readlink	= nfs4_proc_readlink,
10680 	.create		= nfs4_proc_create,
10681 	.remove		= nfs4_proc_remove,
10682 	.unlink_setup	= nfs4_proc_unlink_setup,
10683 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10684 	.unlink_done	= nfs4_proc_unlink_done,
10685 	.rename_setup	= nfs4_proc_rename_setup,
10686 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10687 	.rename_done	= nfs4_proc_rename_done,
10688 	.link		= nfs4_proc_link,
10689 	.symlink	= nfs4_proc_symlink,
10690 	.mkdir		= nfs4_proc_mkdir,
10691 	.rmdir		= nfs4_proc_rmdir,
10692 	.readdir	= nfs4_proc_readdir,
10693 	.mknod		= nfs4_proc_mknod,
10694 	.statfs		= nfs4_proc_statfs,
10695 	.fsinfo		= nfs4_proc_fsinfo,
10696 	.pathconf	= nfs4_proc_pathconf,
10697 	.set_capabilities = nfs4_server_capabilities,
10698 	.decode_dirent	= nfs4_decode_dirent,
10699 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10700 	.read_setup	= nfs4_proc_read_setup,
10701 	.read_done	= nfs4_read_done,
10702 	.write_setup	= nfs4_proc_write_setup,
10703 	.write_done	= nfs4_write_done,
10704 	.commit_setup	= nfs4_proc_commit_setup,
10705 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10706 	.commit_done	= nfs4_commit_done,
10707 	.lock		= nfs4_proc_lock,
10708 	.clear_acl_cache = nfs4_zap_acl_attr,
10709 	.close_context  = nfs4_close_context,
10710 	.open_context	= nfs4_atomic_open,
10711 	.have_delegation = nfs4_have_delegation,
10712 	.alloc_client	= nfs4_alloc_client,
10713 	.init_client	= nfs4_init_client,
10714 	.free_client	= nfs4_free_client,
10715 	.create_server	= nfs4_create_server,
10716 	.clone_server	= nfs_clone_server,
10717 	.discover_trunking = nfs4_discover_trunking,
10718 	.enable_swap	= nfs4_enable_swap,
10719 	.disable_swap	= nfs4_disable_swap,
10720 };
10721 
10722 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10723 	.name	= XATTR_NAME_NFSV4_ACL,
10724 	.list	= nfs4_xattr_list_nfs4_acl,
10725 	.get	= nfs4_xattr_get_nfs4_acl,
10726 	.set	= nfs4_xattr_set_nfs4_acl,
10727 };
10728 
10729 #if defined(CONFIG_NFS_V4_1)
10730 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10731 	.name	= XATTR_NAME_NFSV4_DACL,
10732 	.list	= nfs4_xattr_list_nfs4_dacl,
10733 	.get	= nfs4_xattr_get_nfs4_dacl,
10734 	.set	= nfs4_xattr_set_nfs4_dacl,
10735 };
10736 
10737 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10738 	.name	= XATTR_NAME_NFSV4_SACL,
10739 	.list	= nfs4_xattr_list_nfs4_sacl,
10740 	.get	= nfs4_xattr_get_nfs4_sacl,
10741 	.set	= nfs4_xattr_set_nfs4_sacl,
10742 };
10743 #endif
10744 
10745 #ifdef CONFIG_NFS_V4_2
10746 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10747 	.prefix	= XATTR_USER_PREFIX,
10748 	.get	= nfs4_xattr_get_nfs4_user,
10749 	.set	= nfs4_xattr_set_nfs4_user,
10750 };
10751 #endif
10752 
10753 const struct xattr_handler *nfs4_xattr_handlers[] = {
10754 	&nfs4_xattr_nfs4_acl_handler,
10755 #if defined(CONFIG_NFS_V4_1)
10756 	&nfs4_xattr_nfs4_dacl_handler,
10757 	&nfs4_xattr_nfs4_sacl_handler,
10758 #endif
10759 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10760 	&nfs4_xattr_nfs4_label_handler,
10761 #endif
10762 #ifdef CONFIG_NFS_V4_2
10763 	&nfs4_xattr_nfs4_user_handler,
10764 #endif
10765 	NULL
10766 };
10767