xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision a0936e9e)
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 		return -EREMOTEIO;
174 	case -NFS4ERR_WRONGSEC:
175 	case -NFS4ERR_WRONG_CRED:
176 		return -EPERM;
177 	case -NFS4ERR_BADOWNER:
178 	case -NFS4ERR_BADNAME:
179 		return -EINVAL;
180 	case -NFS4ERR_SHARE_DENIED:
181 		return -EACCES;
182 	case -NFS4ERR_MINOR_VERS_MISMATCH:
183 		return -EPROTONOSUPPORT;
184 	case -NFS4ERR_FILE_OPEN:
185 		return -EBUSY;
186 	case -NFS4ERR_NOT_SAME:
187 		return -ENOTSYNC;
188 	default:
189 		dprintk("%s could not handle NFSv4 error %d\n",
190 				__func__, -err);
191 		break;
192 	}
193 	return -EIO;
194 }
195 
196 /*
197  * This is our standard bitmap for GETATTR requests.
198  */
199 const u32 nfs4_fattr_bitmap[3] = {
200 	FATTR4_WORD0_TYPE
201 	| FATTR4_WORD0_CHANGE
202 	| FATTR4_WORD0_SIZE
203 	| FATTR4_WORD0_FSID
204 	| FATTR4_WORD0_FILEID,
205 	FATTR4_WORD1_MODE
206 	| FATTR4_WORD1_NUMLINKS
207 	| FATTR4_WORD1_OWNER
208 	| FATTR4_WORD1_OWNER_GROUP
209 	| FATTR4_WORD1_RAWDEV
210 	| FATTR4_WORD1_SPACE_USED
211 	| FATTR4_WORD1_TIME_ACCESS
212 	| FATTR4_WORD1_TIME_METADATA
213 	| FATTR4_WORD1_TIME_MODIFY
214 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
215 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
216 	FATTR4_WORD2_SECURITY_LABEL
217 #endif
218 };
219 
220 static const u32 nfs4_pnfs_open_bitmap[3] = {
221 	FATTR4_WORD0_TYPE
222 	| FATTR4_WORD0_CHANGE
223 	| FATTR4_WORD0_SIZE
224 	| FATTR4_WORD0_FSID
225 	| FATTR4_WORD0_FILEID,
226 	FATTR4_WORD1_MODE
227 	| FATTR4_WORD1_NUMLINKS
228 	| FATTR4_WORD1_OWNER
229 	| FATTR4_WORD1_OWNER_GROUP
230 	| FATTR4_WORD1_RAWDEV
231 	| FATTR4_WORD1_SPACE_USED
232 	| FATTR4_WORD1_TIME_ACCESS
233 	| FATTR4_WORD1_TIME_METADATA
234 	| FATTR4_WORD1_TIME_MODIFY,
235 	FATTR4_WORD2_MDSTHRESHOLD
236 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
237 	| FATTR4_WORD2_SECURITY_LABEL
238 #endif
239 };
240 
241 static const u32 nfs4_open_noattr_bitmap[3] = {
242 	FATTR4_WORD0_TYPE
243 	| FATTR4_WORD0_FILEID,
244 };
245 
246 const u32 nfs4_statfs_bitmap[3] = {
247 	FATTR4_WORD0_FILES_AVAIL
248 	| FATTR4_WORD0_FILES_FREE
249 	| FATTR4_WORD0_FILES_TOTAL,
250 	FATTR4_WORD1_SPACE_AVAIL
251 	| FATTR4_WORD1_SPACE_FREE
252 	| FATTR4_WORD1_SPACE_TOTAL
253 };
254 
255 const u32 nfs4_pathconf_bitmap[3] = {
256 	FATTR4_WORD0_MAXLINK
257 	| FATTR4_WORD0_MAXNAME,
258 	0
259 };
260 
261 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
262 			| FATTR4_WORD0_MAXREAD
263 			| FATTR4_WORD0_MAXWRITE
264 			| FATTR4_WORD0_LEASE_TIME,
265 			FATTR4_WORD1_TIME_DELTA
266 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
267 			FATTR4_WORD2_LAYOUT_BLKSIZE
268 			| FATTR4_WORD2_CLONE_BLKSIZE
269 			| FATTR4_WORD2_CHANGE_ATTR_TYPE
270 			| FATTR4_WORD2_XATTR_SUPPORT
271 };
272 
273 const u32 nfs4_fs_locations_bitmap[3] = {
274 	FATTR4_WORD0_CHANGE
275 	| FATTR4_WORD0_SIZE
276 	| FATTR4_WORD0_FSID
277 	| FATTR4_WORD0_FILEID
278 	| FATTR4_WORD0_FS_LOCATIONS,
279 	FATTR4_WORD1_OWNER
280 	| FATTR4_WORD1_OWNER_GROUP
281 	| FATTR4_WORD1_RAWDEV
282 	| FATTR4_WORD1_SPACE_USED
283 	| FATTR4_WORD1_TIME_ACCESS
284 	| FATTR4_WORD1_TIME_METADATA
285 	| FATTR4_WORD1_TIME_MODIFY
286 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
287 };
288 
289 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
290 				    struct inode *inode, unsigned long flags)
291 {
292 	unsigned long cache_validity;
293 
294 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
295 	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
296 		return;
297 
298 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
299 
300 	/* Remove the attributes over which we have full control */
301 	dst[1] &= ~FATTR4_WORD1_RAWDEV;
302 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
303 		dst[0] &= ~FATTR4_WORD0_SIZE;
304 
305 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
306 		dst[0] &= ~FATTR4_WORD0_CHANGE;
307 
308 	if (!(cache_validity & NFS_INO_INVALID_MODE))
309 		dst[1] &= ~FATTR4_WORD1_MODE;
310 	if (!(cache_validity & NFS_INO_INVALID_OTHER))
311 		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
312 }
313 
314 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
315 		struct nfs4_readdir_arg *readdir)
316 {
317 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
318 	__be32 *start, *p;
319 
320 	if (cookie > 2) {
321 		readdir->cookie = cookie;
322 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
323 		return;
324 	}
325 
326 	readdir->cookie = 0;
327 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
328 	if (cookie == 2)
329 		return;
330 
331 	/*
332 	 * NFSv4 servers do not return entries for '.' and '..'
333 	 * Therefore, we fake these entries here.  We let '.'
334 	 * have cookie 0 and '..' have cookie 1.  Note that
335 	 * when talking to the server, we always send cookie 0
336 	 * instead of 1 or 2.
337 	 */
338 	start = p = kmap_atomic(*readdir->pages);
339 
340 	if (cookie == 0) {
341 		*p++ = xdr_one;                                  /* next */
342 		*p++ = xdr_zero;                   /* cookie, first word */
343 		*p++ = xdr_one;                   /* cookie, second word */
344 		*p++ = xdr_one;                             /* entry len */
345 		memcpy(p, ".\0\0\0", 4);                        /* entry */
346 		p++;
347 		*p++ = xdr_one;                         /* bitmap length */
348 		*p++ = htonl(attrs);                           /* bitmap */
349 		*p++ = htonl(12);             /* attribute buffer length */
350 		*p++ = htonl(NF4DIR);
351 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
352 	}
353 
354 	*p++ = xdr_one;                                  /* next */
355 	*p++ = xdr_zero;                   /* cookie, first word */
356 	*p++ = xdr_two;                   /* cookie, second word */
357 	*p++ = xdr_two;                             /* entry len */
358 	memcpy(p, "..\0\0", 4);                         /* entry */
359 	p++;
360 	*p++ = xdr_one;                         /* bitmap length */
361 	*p++ = htonl(attrs);                           /* bitmap */
362 	*p++ = htonl(12);             /* attribute buffer length */
363 	*p++ = htonl(NF4DIR);
364 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
365 
366 	readdir->pgbase = (char *)p - (char *)start;
367 	readdir->count -= readdir->pgbase;
368 	kunmap_atomic(start);
369 }
370 
371 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
372 {
373 	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
374 		fattr->pre_change_attr = version;
375 		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
376 	}
377 }
378 
379 static void nfs4_test_and_free_stateid(struct nfs_server *server,
380 		nfs4_stateid *stateid,
381 		const struct cred *cred)
382 {
383 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
384 
385 	ops->test_and_free_expired(server, stateid, cred);
386 }
387 
388 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
389 		nfs4_stateid *stateid,
390 		const struct cred *cred)
391 {
392 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
393 	nfs4_test_and_free_stateid(server, stateid, cred);
394 }
395 
396 static void nfs4_free_revoked_stateid(struct nfs_server *server,
397 		const nfs4_stateid *stateid,
398 		const struct cred *cred)
399 {
400 	nfs4_stateid tmp;
401 
402 	nfs4_stateid_copy(&tmp, stateid);
403 	__nfs4_free_revoked_stateid(server, &tmp, cred);
404 }
405 
406 static long nfs4_update_delay(long *timeout)
407 {
408 	long ret;
409 	if (!timeout)
410 		return NFS4_POLL_RETRY_MAX;
411 	if (*timeout <= 0)
412 		*timeout = NFS4_POLL_RETRY_MIN;
413 	if (*timeout > NFS4_POLL_RETRY_MAX)
414 		*timeout = NFS4_POLL_RETRY_MAX;
415 	ret = *timeout;
416 	*timeout <<= 1;
417 	return ret;
418 }
419 
420 static int nfs4_delay_killable(long *timeout)
421 {
422 	might_sleep();
423 
424 	__set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
425 	schedule_timeout(nfs4_update_delay(timeout));
426 	if (!__fatal_signal_pending(current))
427 		return 0;
428 	return -EINTR;
429 }
430 
431 static int nfs4_delay_interruptible(long *timeout)
432 {
433 	might_sleep();
434 
435 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
436 	schedule_timeout(nfs4_update_delay(timeout));
437 	if (!signal_pending(current))
438 		return 0;
439 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
440 }
441 
442 static int nfs4_delay(long *timeout, bool interruptible)
443 {
444 	if (interruptible)
445 		return nfs4_delay_interruptible(timeout);
446 	return nfs4_delay_killable(timeout);
447 }
448 
449 static const nfs4_stateid *
450 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
451 {
452 	if (!stateid)
453 		return NULL;
454 	switch (stateid->type) {
455 	case NFS4_OPEN_STATEID_TYPE:
456 	case NFS4_LOCK_STATEID_TYPE:
457 	case NFS4_DELEGATION_STATEID_TYPE:
458 		return stateid;
459 	default:
460 		break;
461 	}
462 	return NULL;
463 }
464 
465 /* This is the error handling routine for processes that are allowed
466  * to sleep.
467  */
468 static int nfs4_do_handle_exception(struct nfs_server *server,
469 		int errorcode, struct nfs4_exception *exception)
470 {
471 	struct nfs_client *clp = server->nfs_client;
472 	struct nfs4_state *state = exception->state;
473 	const nfs4_stateid *stateid;
474 	struct inode *inode = exception->inode;
475 	int ret = errorcode;
476 
477 	exception->delay = 0;
478 	exception->recovering = 0;
479 	exception->retry = 0;
480 
481 	stateid = nfs4_recoverable_stateid(exception->stateid);
482 	if (stateid == NULL && state != NULL)
483 		stateid = nfs4_recoverable_stateid(&state->stateid);
484 
485 	switch(errorcode) {
486 		case 0:
487 			return 0;
488 		case -NFS4ERR_BADHANDLE:
489 		case -ESTALE:
490 			if (inode != NULL && S_ISREG(inode->i_mode))
491 				pnfs_destroy_layout(NFS_I(inode));
492 			break;
493 		case -NFS4ERR_DELEG_REVOKED:
494 		case -NFS4ERR_ADMIN_REVOKED:
495 		case -NFS4ERR_EXPIRED:
496 		case -NFS4ERR_BAD_STATEID:
497 		case -NFS4ERR_PARTNER_NO_AUTH:
498 			if (inode != NULL && stateid != NULL) {
499 				nfs_inode_find_state_and_recover(inode,
500 						stateid);
501 				goto wait_on_recovery;
502 			}
503 			fallthrough;
504 		case -NFS4ERR_OPENMODE:
505 			if (inode) {
506 				int err;
507 
508 				err = nfs_async_inode_return_delegation(inode,
509 						stateid);
510 				if (err == 0)
511 					goto wait_on_recovery;
512 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
513 					exception->retry = 1;
514 					break;
515 				}
516 			}
517 			if (state == NULL)
518 				break;
519 			ret = nfs4_schedule_stateid_recovery(server, state);
520 			if (ret < 0)
521 				break;
522 			goto wait_on_recovery;
523 		case -NFS4ERR_STALE_STATEID:
524 		case -NFS4ERR_STALE_CLIENTID:
525 			nfs4_schedule_lease_recovery(clp);
526 			goto wait_on_recovery;
527 		case -NFS4ERR_MOVED:
528 			ret = nfs4_schedule_migration_recovery(server);
529 			if (ret < 0)
530 				break;
531 			goto wait_on_recovery;
532 		case -NFS4ERR_LEASE_MOVED:
533 			nfs4_schedule_lease_moved_recovery(clp);
534 			goto wait_on_recovery;
535 #if defined(CONFIG_NFS_V4_1)
536 		case -NFS4ERR_BADSESSION:
537 		case -NFS4ERR_BADSLOT:
538 		case -NFS4ERR_BAD_HIGH_SLOT:
539 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
540 		case -NFS4ERR_DEADSESSION:
541 		case -NFS4ERR_SEQ_FALSE_RETRY:
542 		case -NFS4ERR_SEQ_MISORDERED:
543 			/* Handled in nfs41_sequence_process() */
544 			goto wait_on_recovery;
545 #endif /* defined(CONFIG_NFS_V4_1) */
546 		case -NFS4ERR_FILE_OPEN:
547 			if (exception->timeout > HZ) {
548 				/* We have retried a decent amount, time to
549 				 * fail
550 				 */
551 				ret = -EBUSY;
552 				break;
553 			}
554 			fallthrough;
555 		case -NFS4ERR_DELAY:
556 			nfs_inc_server_stats(server, NFSIOS_DELAY);
557 			fallthrough;
558 		case -NFS4ERR_GRACE:
559 		case -NFS4ERR_LAYOUTTRYLATER:
560 		case -NFS4ERR_RECALLCONFLICT:
561 			exception->delay = 1;
562 			return 0;
563 
564 		case -NFS4ERR_RETRY_UNCACHED_REP:
565 		case -NFS4ERR_OLD_STATEID:
566 			exception->retry = 1;
567 			break;
568 		case -NFS4ERR_BADOWNER:
569 			/* The following works around a Linux server bug! */
570 		case -NFS4ERR_BADNAME:
571 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
572 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
573 				exception->retry = 1;
574 				printk(KERN_WARNING "NFS: v4 server %s "
575 						"does not accept raw "
576 						"uid/gids. "
577 						"Reenabling the idmapper.\n",
578 						server->nfs_client->cl_hostname);
579 			}
580 	}
581 	/* We failed to handle the error */
582 	return nfs4_map_errors(ret);
583 wait_on_recovery:
584 	exception->recovering = 1;
585 	return 0;
586 }
587 
588 /* This is the error handling routine for processes that are allowed
589  * to sleep.
590  */
591 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
592 {
593 	struct nfs_client *clp = server->nfs_client;
594 	int ret;
595 
596 	ret = nfs4_do_handle_exception(server, errorcode, exception);
597 	if (exception->delay) {
598 		ret = nfs4_delay(&exception->timeout,
599 				exception->interruptible);
600 		goto out_retry;
601 	}
602 	if (exception->recovering) {
603 		if (exception->task_is_privileged)
604 			return -EDEADLOCK;
605 		ret = nfs4_wait_clnt_recover(clp);
606 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
607 			return -EIO;
608 		goto out_retry;
609 	}
610 	return ret;
611 out_retry:
612 	if (ret == 0)
613 		exception->retry = 1;
614 	return ret;
615 }
616 
617 static int
618 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
619 		int errorcode, struct nfs4_exception *exception)
620 {
621 	struct nfs_client *clp = server->nfs_client;
622 	int ret;
623 
624 	ret = nfs4_do_handle_exception(server, errorcode, exception);
625 	if (exception->delay) {
626 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
627 		goto out_retry;
628 	}
629 	if (exception->recovering) {
630 		if (exception->task_is_privileged)
631 			return -EDEADLOCK;
632 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
633 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
634 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
635 		goto out_retry;
636 	}
637 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
638 		ret = -EIO;
639 	return ret;
640 out_retry:
641 	if (ret == 0) {
642 		exception->retry = 1;
643 		/*
644 		 * For NFS4ERR_MOVED, the client transport will need to
645 		 * be recomputed after migration recovery has completed.
646 		 */
647 		if (errorcode == -NFS4ERR_MOVED)
648 			rpc_task_release_transport(task);
649 	}
650 	return ret;
651 }
652 
653 int
654 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
655 			struct nfs4_state *state, long *timeout)
656 {
657 	struct nfs4_exception exception = {
658 		.state = state,
659 	};
660 
661 	if (task->tk_status >= 0)
662 		return 0;
663 	if (timeout)
664 		exception.timeout = *timeout;
665 	task->tk_status = nfs4_async_handle_exception(task, server,
666 			task->tk_status,
667 			&exception);
668 	if (exception.delay && timeout)
669 		*timeout = exception.timeout;
670 	if (exception.retry)
671 		return -EAGAIN;
672 	return 0;
673 }
674 
675 /*
676  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
677  * or 'false' otherwise.
678  */
679 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
680 {
681 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
682 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
683 }
684 
685 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
686 {
687 	spin_lock(&clp->cl_lock);
688 	if (time_before(clp->cl_last_renewal,timestamp))
689 		clp->cl_last_renewal = timestamp;
690 	spin_unlock(&clp->cl_lock);
691 }
692 
693 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
694 {
695 	struct nfs_client *clp = server->nfs_client;
696 
697 	if (!nfs4_has_session(clp))
698 		do_renew_lease(clp, timestamp);
699 }
700 
701 struct nfs4_call_sync_data {
702 	const struct nfs_server *seq_server;
703 	struct nfs4_sequence_args *seq_args;
704 	struct nfs4_sequence_res *seq_res;
705 };
706 
707 void nfs4_init_sequence(struct nfs4_sequence_args *args,
708 			struct nfs4_sequence_res *res, int cache_reply,
709 			int privileged)
710 {
711 	args->sa_slot = NULL;
712 	args->sa_cache_this = cache_reply;
713 	args->sa_privileged = privileged;
714 
715 	res->sr_slot = NULL;
716 }
717 
718 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
719 {
720 	struct nfs4_slot *slot = res->sr_slot;
721 	struct nfs4_slot_table *tbl;
722 
723 	tbl = slot->table;
724 	spin_lock(&tbl->slot_tbl_lock);
725 	if (!nfs41_wake_and_assign_slot(tbl, slot))
726 		nfs4_free_slot(tbl, slot);
727 	spin_unlock(&tbl->slot_tbl_lock);
728 
729 	res->sr_slot = NULL;
730 }
731 
732 static int nfs40_sequence_done(struct rpc_task *task,
733 			       struct nfs4_sequence_res *res)
734 {
735 	if (res->sr_slot != NULL)
736 		nfs40_sequence_free_slot(res);
737 	return 1;
738 }
739 
740 #if defined(CONFIG_NFS_V4_1)
741 
742 static void nfs41_release_slot(struct nfs4_slot *slot)
743 {
744 	struct nfs4_session *session;
745 	struct nfs4_slot_table *tbl;
746 	bool send_new_highest_used_slotid = false;
747 
748 	if (!slot)
749 		return;
750 	tbl = slot->table;
751 	session = tbl->session;
752 
753 	/* Bump the slot sequence number */
754 	if (slot->seq_done)
755 		slot->seq_nr++;
756 	slot->seq_done = 0;
757 
758 	spin_lock(&tbl->slot_tbl_lock);
759 	/* Be nice to the server: try to ensure that the last transmitted
760 	 * value for highest_user_slotid <= target_highest_slotid
761 	 */
762 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
763 		send_new_highest_used_slotid = true;
764 
765 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
766 		send_new_highest_used_slotid = false;
767 		goto out_unlock;
768 	}
769 	nfs4_free_slot(tbl, slot);
770 
771 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
772 		send_new_highest_used_slotid = false;
773 out_unlock:
774 	spin_unlock(&tbl->slot_tbl_lock);
775 	if (send_new_highest_used_slotid)
776 		nfs41_notify_server(session->clp);
777 	if (waitqueue_active(&tbl->slot_waitq))
778 		wake_up_all(&tbl->slot_waitq);
779 }
780 
781 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
782 {
783 	nfs41_release_slot(res->sr_slot);
784 	res->sr_slot = NULL;
785 }
786 
787 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
788 		u32 seqnr)
789 {
790 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
791 		slot->seq_nr_highest_sent = seqnr;
792 }
793 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
794 {
795 	nfs4_slot_sequence_record_sent(slot, seqnr);
796 	slot->seq_nr_last_acked = seqnr;
797 }
798 
799 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
800 				struct nfs4_slot *slot)
801 {
802 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
803 	if (!IS_ERR(task))
804 		rpc_put_task_async(task);
805 }
806 
807 static int nfs41_sequence_process(struct rpc_task *task,
808 		struct nfs4_sequence_res *res)
809 {
810 	struct nfs4_session *session;
811 	struct nfs4_slot *slot = res->sr_slot;
812 	struct nfs_client *clp;
813 	int status;
814 	int ret = 1;
815 
816 	if (slot == NULL)
817 		goto out_noaction;
818 	/* don't increment the sequence number if the task wasn't sent */
819 	if (!RPC_WAS_SENT(task) || slot->seq_done)
820 		goto out;
821 
822 	session = slot->table->session;
823 	clp = session->clp;
824 
825 	trace_nfs4_sequence_done(session, res);
826 
827 	status = res->sr_status;
828 	if (task->tk_status == -NFS4ERR_DEADSESSION)
829 		status = -NFS4ERR_DEADSESSION;
830 
831 	/* Check the SEQUENCE operation status */
832 	switch (status) {
833 	case 0:
834 		/* Mark this sequence number as having been acked */
835 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
836 		/* Update the slot's sequence and clientid lease timer */
837 		slot->seq_done = 1;
838 		do_renew_lease(clp, res->sr_timestamp);
839 		/* Check sequence flags */
840 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
841 				!!slot->privileged);
842 		nfs41_update_target_slotid(slot->table, slot, res);
843 		break;
844 	case 1:
845 		/*
846 		 * sr_status remains 1 if an RPC level error occurred.
847 		 * The server may or may not have processed the sequence
848 		 * operation..
849 		 */
850 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
851 		slot->seq_done = 1;
852 		goto out;
853 	case -NFS4ERR_DELAY:
854 		/* The server detected a resend of the RPC call and
855 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
856 		 * of RFC5661.
857 		 */
858 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
859 			__func__,
860 			slot->slot_nr,
861 			slot->seq_nr);
862 		goto out_retry;
863 	case -NFS4ERR_RETRY_UNCACHED_REP:
864 	case -NFS4ERR_SEQ_FALSE_RETRY:
865 		/*
866 		 * The server thinks we tried to replay a request.
867 		 * Retry the call after bumping the sequence ID.
868 		 */
869 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
870 		goto retry_new_seq;
871 	case -NFS4ERR_BADSLOT:
872 		/*
873 		 * The slot id we used was probably retired. Try again
874 		 * using a different slot id.
875 		 */
876 		if (slot->slot_nr < slot->table->target_highest_slotid)
877 			goto session_recover;
878 		goto retry_nowait;
879 	case -NFS4ERR_SEQ_MISORDERED:
880 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
881 		/*
882 		 * Were one or more calls using this slot interrupted?
883 		 * If the server never received the request, then our
884 		 * transmitted slot sequence number may be too high. However,
885 		 * if the server did receive the request then it might
886 		 * accidentally give us a reply with a mismatched operation.
887 		 * We can sort this out by sending a lone sequence operation
888 		 * to the server on the same slot.
889 		 */
890 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
891 			slot->seq_nr--;
892 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
893 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
894 				res->sr_slot = NULL;
895 			}
896 			goto retry_nowait;
897 		}
898 		/*
899 		 * RFC5661:
900 		 * A retry might be sent while the original request is
901 		 * still in progress on the replier. The replier SHOULD
902 		 * deal with the issue by returning NFS4ERR_DELAY as the
903 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
904 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
905 		 *
906 		 * Restart the search after a delay.
907 		 */
908 		slot->seq_nr = slot->seq_nr_highest_sent;
909 		goto out_retry;
910 	case -NFS4ERR_BADSESSION:
911 	case -NFS4ERR_DEADSESSION:
912 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
913 		goto session_recover;
914 	default:
915 		/* Just update the slot sequence no. */
916 		slot->seq_done = 1;
917 	}
918 out:
919 	/* The session may be reset by one of the error handlers. */
920 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
921 out_noaction:
922 	return ret;
923 session_recover:
924 	nfs4_schedule_session_recovery(session, status);
925 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
926 	nfs41_sequence_free_slot(res);
927 	goto out;
928 retry_new_seq:
929 	++slot->seq_nr;
930 retry_nowait:
931 	if (rpc_restart_call_prepare(task)) {
932 		nfs41_sequence_free_slot(res);
933 		task->tk_status = 0;
934 		ret = 0;
935 	}
936 	goto out;
937 out_retry:
938 	if (!rpc_restart_call(task))
939 		goto out;
940 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
941 	return 0;
942 }
943 
944 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
945 {
946 	if (!nfs41_sequence_process(task, res))
947 		return 0;
948 	if (res->sr_slot != NULL)
949 		nfs41_sequence_free_slot(res);
950 	return 1;
951 
952 }
953 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
954 
955 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
956 {
957 	if (res->sr_slot == NULL)
958 		return 1;
959 	if (res->sr_slot->table->session != NULL)
960 		return nfs41_sequence_process(task, res);
961 	return nfs40_sequence_done(task, res);
962 }
963 
964 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
965 {
966 	if (res->sr_slot != NULL) {
967 		if (res->sr_slot->table->session != NULL)
968 			nfs41_sequence_free_slot(res);
969 		else
970 			nfs40_sequence_free_slot(res);
971 	}
972 }
973 
974 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
975 {
976 	if (res->sr_slot == NULL)
977 		return 1;
978 	if (!res->sr_slot->table->session)
979 		return nfs40_sequence_done(task, res);
980 	return nfs41_sequence_done(task, res);
981 }
982 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
983 
984 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
985 {
986 	struct nfs4_call_sync_data *data = calldata;
987 
988 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
989 
990 	nfs4_setup_sequence(data->seq_server->nfs_client,
991 			    data->seq_args, data->seq_res, task);
992 }
993 
994 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
995 {
996 	struct nfs4_call_sync_data *data = calldata;
997 
998 	nfs41_sequence_done(task, data->seq_res);
999 }
1000 
1001 static const struct rpc_call_ops nfs41_call_sync_ops = {
1002 	.rpc_call_prepare = nfs41_call_sync_prepare,
1003 	.rpc_call_done = nfs41_call_sync_done,
1004 };
1005 
1006 #else	/* !CONFIG_NFS_V4_1 */
1007 
1008 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1009 {
1010 	return nfs40_sequence_done(task, res);
1011 }
1012 
1013 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1014 {
1015 	if (res->sr_slot != NULL)
1016 		nfs40_sequence_free_slot(res);
1017 }
1018 
1019 int nfs4_sequence_done(struct rpc_task *task,
1020 		       struct nfs4_sequence_res *res)
1021 {
1022 	return nfs40_sequence_done(task, res);
1023 }
1024 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1025 
1026 #endif	/* !CONFIG_NFS_V4_1 */
1027 
1028 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1029 {
1030 	res->sr_timestamp = jiffies;
1031 	res->sr_status_flags = 0;
1032 	res->sr_status = 1;
1033 }
1034 
1035 static
1036 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1037 		struct nfs4_sequence_res *res,
1038 		struct nfs4_slot *slot)
1039 {
1040 	if (!slot)
1041 		return;
1042 	slot->privileged = args->sa_privileged ? 1 : 0;
1043 	args->sa_slot = slot;
1044 
1045 	res->sr_slot = slot;
1046 }
1047 
1048 int nfs4_setup_sequence(struct nfs_client *client,
1049 			struct nfs4_sequence_args *args,
1050 			struct nfs4_sequence_res *res,
1051 			struct rpc_task *task)
1052 {
1053 	struct nfs4_session *session = nfs4_get_session(client);
1054 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1055 	struct nfs4_slot *slot;
1056 
1057 	/* slot already allocated? */
1058 	if (res->sr_slot != NULL)
1059 		goto out_start;
1060 
1061 	if (session)
1062 		tbl = &session->fc_slot_table;
1063 
1064 	spin_lock(&tbl->slot_tbl_lock);
1065 	/* The state manager will wait until the slot table is empty */
1066 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1067 		goto out_sleep;
1068 
1069 	slot = nfs4_alloc_slot(tbl);
1070 	if (IS_ERR(slot)) {
1071 		if (slot == ERR_PTR(-ENOMEM))
1072 			goto out_sleep_timeout;
1073 		goto out_sleep;
1074 	}
1075 	spin_unlock(&tbl->slot_tbl_lock);
1076 
1077 	nfs4_sequence_attach_slot(args, res, slot);
1078 
1079 	trace_nfs4_setup_sequence(session, args);
1080 out_start:
1081 	nfs41_sequence_res_init(res);
1082 	rpc_call_start(task);
1083 	return 0;
1084 out_sleep_timeout:
1085 	/* Try again in 1/4 second */
1086 	if (args->sa_privileged)
1087 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1088 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1089 	else
1090 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1091 				NULL, jiffies + (HZ >> 2));
1092 	spin_unlock(&tbl->slot_tbl_lock);
1093 	return -EAGAIN;
1094 out_sleep:
1095 	if (args->sa_privileged)
1096 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1097 				RPC_PRIORITY_PRIVILEGED);
1098 	else
1099 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1100 	spin_unlock(&tbl->slot_tbl_lock);
1101 	return -EAGAIN;
1102 }
1103 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1104 
1105 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1106 {
1107 	struct nfs4_call_sync_data *data = calldata;
1108 	nfs4_setup_sequence(data->seq_server->nfs_client,
1109 				data->seq_args, data->seq_res, task);
1110 }
1111 
1112 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1113 {
1114 	struct nfs4_call_sync_data *data = calldata;
1115 	nfs4_sequence_done(task, data->seq_res);
1116 }
1117 
1118 static const struct rpc_call_ops nfs40_call_sync_ops = {
1119 	.rpc_call_prepare = nfs40_call_sync_prepare,
1120 	.rpc_call_done = nfs40_call_sync_done,
1121 };
1122 
1123 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1124 {
1125 	int ret;
1126 	struct rpc_task *task;
1127 
1128 	task = rpc_run_task(task_setup);
1129 	if (IS_ERR(task))
1130 		return PTR_ERR(task);
1131 
1132 	ret = task->tk_status;
1133 	rpc_put_task(task);
1134 	return ret;
1135 }
1136 
1137 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1138 			     struct nfs_server *server,
1139 			     struct rpc_message *msg,
1140 			     struct nfs4_sequence_args *args,
1141 			     struct nfs4_sequence_res *res,
1142 			     unsigned short task_flags)
1143 {
1144 	struct nfs_client *clp = server->nfs_client;
1145 	struct nfs4_call_sync_data data = {
1146 		.seq_server = server,
1147 		.seq_args = args,
1148 		.seq_res = res,
1149 	};
1150 	struct rpc_task_setup task_setup = {
1151 		.rpc_client = clnt,
1152 		.rpc_message = msg,
1153 		.callback_ops = clp->cl_mvops->call_sync_ops,
1154 		.callback_data = &data,
1155 		.flags = task_flags,
1156 	};
1157 
1158 	return nfs4_call_sync_custom(&task_setup);
1159 }
1160 
1161 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1162 				   struct nfs_server *server,
1163 				   struct rpc_message *msg,
1164 				   struct nfs4_sequence_args *args,
1165 				   struct nfs4_sequence_res *res)
1166 {
1167 	unsigned short task_flags = 0;
1168 
1169 	if (server->caps & NFS_CAP_MOVEABLE)
1170 		task_flags = RPC_TASK_MOVEABLE;
1171 	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1172 }
1173 
1174 
1175 int nfs4_call_sync(struct rpc_clnt *clnt,
1176 		   struct nfs_server *server,
1177 		   struct rpc_message *msg,
1178 		   struct nfs4_sequence_args *args,
1179 		   struct nfs4_sequence_res *res,
1180 		   int cache_reply)
1181 {
1182 	nfs4_init_sequence(args, res, cache_reply, 0);
1183 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1184 }
1185 
1186 static void
1187 nfs4_inc_nlink_locked(struct inode *inode)
1188 {
1189 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1190 					     NFS_INO_INVALID_CTIME |
1191 					     NFS_INO_INVALID_NLINK);
1192 	inc_nlink(inode);
1193 }
1194 
1195 static void
1196 nfs4_inc_nlink(struct inode *inode)
1197 {
1198 	spin_lock(&inode->i_lock);
1199 	nfs4_inc_nlink_locked(inode);
1200 	spin_unlock(&inode->i_lock);
1201 }
1202 
1203 static void
1204 nfs4_dec_nlink_locked(struct inode *inode)
1205 {
1206 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1207 					     NFS_INO_INVALID_CTIME |
1208 					     NFS_INO_INVALID_NLINK);
1209 	drop_nlink(inode);
1210 }
1211 
1212 static void
1213 nfs4_update_changeattr_locked(struct inode *inode,
1214 		struct nfs4_change_info *cinfo,
1215 		unsigned long timestamp, unsigned long cache_validity)
1216 {
1217 	struct nfs_inode *nfsi = NFS_I(inode);
1218 	u64 change_attr = inode_peek_iversion_raw(inode);
1219 
1220 	cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1221 	if (S_ISDIR(inode->i_mode))
1222 		cache_validity |= NFS_INO_INVALID_DATA;
1223 
1224 	switch (NFS_SERVER(inode)->change_attr_type) {
1225 	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1226 		if (cinfo->after == change_attr)
1227 			goto out;
1228 		break;
1229 	default:
1230 		if ((s64)(change_attr - cinfo->after) >= 0)
1231 			goto out;
1232 	}
1233 
1234 	inode_set_iversion_raw(inode, cinfo->after);
1235 	if (!cinfo->atomic || cinfo->before != change_attr) {
1236 		if (S_ISDIR(inode->i_mode))
1237 			nfs_force_lookup_revalidate(inode);
1238 
1239 		if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1240 			cache_validity |=
1241 				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1242 				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1243 				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1244 				NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1245 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1246 	}
1247 	nfsi->attrtimeo_timestamp = jiffies;
1248 	nfsi->read_cache_jiffies = timestamp;
1249 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1250 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1251 out:
1252 	nfs_set_cache_invalid(inode, cache_validity);
1253 }
1254 
1255 void
1256 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1257 		unsigned long timestamp, unsigned long cache_validity)
1258 {
1259 	spin_lock(&dir->i_lock);
1260 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1261 	spin_unlock(&dir->i_lock);
1262 }
1263 
1264 struct nfs4_open_createattrs {
1265 	struct nfs4_label *label;
1266 	struct iattr *sattr;
1267 	const __u32 verf[2];
1268 };
1269 
1270 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1271 		int err, struct nfs4_exception *exception)
1272 {
1273 	if (err != -EINVAL)
1274 		return false;
1275 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1276 		return false;
1277 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1278 	exception->retry = 1;
1279 	return true;
1280 }
1281 
1282 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1283 {
1284 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1285 }
1286 
1287 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1288 {
1289 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1290 
1291 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1292 }
1293 
1294 static u32
1295 nfs4_map_atomic_open_share(struct nfs_server *server,
1296 		fmode_t fmode, int openflags)
1297 {
1298 	u32 res = 0;
1299 
1300 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1301 	case FMODE_READ:
1302 		res = NFS4_SHARE_ACCESS_READ;
1303 		break;
1304 	case FMODE_WRITE:
1305 		res = NFS4_SHARE_ACCESS_WRITE;
1306 		break;
1307 	case FMODE_READ|FMODE_WRITE:
1308 		res = NFS4_SHARE_ACCESS_BOTH;
1309 	}
1310 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1311 		goto out;
1312 	/* Want no delegation if we're using O_DIRECT */
1313 	if (openflags & O_DIRECT)
1314 		res |= NFS4_SHARE_WANT_NO_DELEG;
1315 out:
1316 	return res;
1317 }
1318 
1319 static enum open_claim_type4
1320 nfs4_map_atomic_open_claim(struct nfs_server *server,
1321 		enum open_claim_type4 claim)
1322 {
1323 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1324 		return claim;
1325 	switch (claim) {
1326 	default:
1327 		return claim;
1328 	case NFS4_OPEN_CLAIM_FH:
1329 		return NFS4_OPEN_CLAIM_NULL;
1330 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1331 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1332 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1333 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1334 	}
1335 }
1336 
1337 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1338 {
1339 	p->o_res.f_attr = &p->f_attr;
1340 	p->o_res.seqid = p->o_arg.seqid;
1341 	p->c_res.seqid = p->c_arg.seqid;
1342 	p->o_res.server = p->o_arg.server;
1343 	p->o_res.access_request = p->o_arg.access;
1344 	nfs_fattr_init(&p->f_attr);
1345 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1346 }
1347 
1348 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1349 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1350 		const struct nfs4_open_createattrs *c,
1351 		enum open_claim_type4 claim,
1352 		gfp_t gfp_mask)
1353 {
1354 	struct dentry *parent = dget_parent(dentry);
1355 	struct inode *dir = d_inode(parent);
1356 	struct nfs_server *server = NFS_SERVER(dir);
1357 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1358 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1359 	struct nfs4_opendata *p;
1360 
1361 	p = kzalloc(sizeof(*p), gfp_mask);
1362 	if (p == NULL)
1363 		goto err;
1364 
1365 	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1366 	if (IS_ERR(p->f_attr.label))
1367 		goto err_free_p;
1368 
1369 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1370 	if (IS_ERR(p->a_label))
1371 		goto err_free_f;
1372 
1373 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1374 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1375 	if (IS_ERR(p->o_arg.seqid))
1376 		goto err_free_label;
1377 	nfs_sb_active(dentry->d_sb);
1378 	p->dentry = dget(dentry);
1379 	p->dir = parent;
1380 	p->owner = sp;
1381 	atomic_inc(&sp->so_count);
1382 	p->o_arg.open_flags = flags;
1383 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1384 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1385 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1386 			fmode, flags);
1387 	if (flags & O_CREAT) {
1388 		p->o_arg.umask = current_umask();
1389 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1390 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1391 			p->o_arg.u.attrs = &p->attrs;
1392 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1393 
1394 			memcpy(p->o_arg.u.verifier.data, c->verf,
1395 					sizeof(p->o_arg.u.verifier.data));
1396 		}
1397 	}
1398 	/* ask server to check for all possible rights as results
1399 	 * are cached */
1400 	switch (p->o_arg.claim) {
1401 	default:
1402 		break;
1403 	case NFS4_OPEN_CLAIM_NULL:
1404 	case NFS4_OPEN_CLAIM_FH:
1405 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1406 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1407 				  NFS4_ACCESS_EXECUTE |
1408 				  nfs_access_xattr_mask(server);
1409 	}
1410 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1411 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1412 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1413 	p->o_arg.name = &dentry->d_name;
1414 	p->o_arg.server = server;
1415 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1416 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1417 	switch (p->o_arg.claim) {
1418 	case NFS4_OPEN_CLAIM_NULL:
1419 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1420 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1421 		p->o_arg.fh = NFS_FH(dir);
1422 		break;
1423 	case NFS4_OPEN_CLAIM_PREVIOUS:
1424 	case NFS4_OPEN_CLAIM_FH:
1425 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1426 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1427 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1428 	}
1429 	p->c_arg.fh = &p->o_res.fh;
1430 	p->c_arg.stateid = &p->o_res.stateid;
1431 	p->c_arg.seqid = p->o_arg.seqid;
1432 	nfs4_init_opendata_res(p);
1433 	kref_init(&p->kref);
1434 	return p;
1435 
1436 err_free_label:
1437 	nfs4_label_free(p->a_label);
1438 err_free_f:
1439 	nfs4_label_free(p->f_attr.label);
1440 err_free_p:
1441 	kfree(p);
1442 err:
1443 	dput(parent);
1444 	return NULL;
1445 }
1446 
1447 static void nfs4_opendata_free(struct kref *kref)
1448 {
1449 	struct nfs4_opendata *p = container_of(kref,
1450 			struct nfs4_opendata, kref);
1451 	struct super_block *sb = p->dentry->d_sb;
1452 
1453 	nfs4_lgopen_release(p->lgp);
1454 	nfs_free_seqid(p->o_arg.seqid);
1455 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1456 	if (p->state != NULL)
1457 		nfs4_put_open_state(p->state);
1458 	nfs4_put_state_owner(p->owner);
1459 
1460 	nfs4_label_free(p->a_label);
1461 	nfs4_label_free(p->f_attr.label);
1462 
1463 	dput(p->dir);
1464 	dput(p->dentry);
1465 	nfs_sb_deactive(sb);
1466 	nfs_fattr_free_names(&p->f_attr);
1467 	kfree(p->f_attr.mdsthreshold);
1468 	kfree(p);
1469 }
1470 
1471 static void nfs4_opendata_put(struct nfs4_opendata *p)
1472 {
1473 	if (p != NULL)
1474 		kref_put(&p->kref, nfs4_opendata_free);
1475 }
1476 
1477 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1478 		fmode_t fmode)
1479 {
1480 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1481 	case FMODE_READ|FMODE_WRITE:
1482 		return state->n_rdwr != 0;
1483 	case FMODE_WRITE:
1484 		return state->n_wronly != 0;
1485 	case FMODE_READ:
1486 		return state->n_rdonly != 0;
1487 	}
1488 	WARN_ON_ONCE(1);
1489 	return false;
1490 }
1491 
1492 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1493 		int open_mode, enum open_claim_type4 claim)
1494 {
1495 	int ret = 0;
1496 
1497 	if (open_mode & (O_EXCL|O_TRUNC))
1498 		goto out;
1499 	switch (claim) {
1500 	case NFS4_OPEN_CLAIM_NULL:
1501 	case NFS4_OPEN_CLAIM_FH:
1502 		goto out;
1503 	default:
1504 		break;
1505 	}
1506 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1507 		case FMODE_READ:
1508 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1509 				&& state->n_rdonly != 0;
1510 			break;
1511 		case FMODE_WRITE:
1512 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1513 				&& state->n_wronly != 0;
1514 			break;
1515 		case FMODE_READ|FMODE_WRITE:
1516 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1517 				&& state->n_rdwr != 0;
1518 	}
1519 out:
1520 	return ret;
1521 }
1522 
1523 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1524 		enum open_claim_type4 claim)
1525 {
1526 	if (delegation == NULL)
1527 		return 0;
1528 	if ((delegation->type & fmode) != fmode)
1529 		return 0;
1530 	switch (claim) {
1531 	case NFS4_OPEN_CLAIM_NULL:
1532 	case NFS4_OPEN_CLAIM_FH:
1533 		break;
1534 	case NFS4_OPEN_CLAIM_PREVIOUS:
1535 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1536 			break;
1537 		fallthrough;
1538 	default:
1539 		return 0;
1540 	}
1541 	nfs_mark_delegation_referenced(delegation);
1542 	return 1;
1543 }
1544 
1545 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1546 {
1547 	switch (fmode) {
1548 		case FMODE_WRITE:
1549 			state->n_wronly++;
1550 			break;
1551 		case FMODE_READ:
1552 			state->n_rdonly++;
1553 			break;
1554 		case FMODE_READ|FMODE_WRITE:
1555 			state->n_rdwr++;
1556 	}
1557 	nfs4_state_set_mode_locked(state, state->state | fmode);
1558 }
1559 
1560 #ifdef CONFIG_NFS_V4_1
1561 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1562 {
1563 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1564 		return true;
1565 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1566 		return true;
1567 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1568 		return true;
1569 	return false;
1570 }
1571 #endif /* CONFIG_NFS_V4_1 */
1572 
1573 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1574 {
1575 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1576 		wake_up_all(&state->waitq);
1577 }
1578 
1579 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1580 {
1581 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1582 	bool need_recover = false;
1583 
1584 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1585 		need_recover = true;
1586 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1587 		need_recover = true;
1588 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1589 		need_recover = true;
1590 	if (need_recover)
1591 		nfs4_state_mark_reclaim_nograce(clp, state);
1592 }
1593 
1594 /*
1595  * Check for whether or not the caller may update the open stateid
1596  * to the value passed in by stateid.
1597  *
1598  * Note: This function relies heavily on the server implementing
1599  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1600  * correctly.
1601  * i.e. The stateid seqids have to be initialised to 1, and
1602  * are then incremented on every state transition.
1603  */
1604 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1605 		const nfs4_stateid *stateid)
1606 {
1607 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1608 		/* The common case - we're updating to a new sequence number */
1609 		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1610 			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1611 				return true;
1612 			return false;
1613 		}
1614 		/* The server returned a new stateid */
1615 	}
1616 	/* This is the first OPEN in this generation */
1617 	if (stateid->seqid == cpu_to_be32(1))
1618 		return true;
1619 	return false;
1620 }
1621 
1622 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1623 {
1624 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1625 		return;
1626 	if (state->n_wronly)
1627 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1628 	if (state->n_rdonly)
1629 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1630 	if (state->n_rdwr)
1631 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1632 	set_bit(NFS_OPEN_STATE, &state->flags);
1633 }
1634 
1635 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1636 		nfs4_stateid *stateid, fmode_t fmode)
1637 {
1638 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1639 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1640 	case FMODE_WRITE:
1641 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1642 		break;
1643 	case FMODE_READ:
1644 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1645 		break;
1646 	case 0:
1647 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1648 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1649 		clear_bit(NFS_OPEN_STATE, &state->flags);
1650 	}
1651 	if (stateid == NULL)
1652 		return;
1653 	/* Handle OPEN+OPEN_DOWNGRADE races */
1654 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1655 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1656 		nfs_resync_open_stateid_locked(state);
1657 		goto out;
1658 	}
1659 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1660 		nfs4_stateid_copy(&state->stateid, stateid);
1661 	nfs4_stateid_copy(&state->open_stateid, stateid);
1662 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1663 out:
1664 	nfs_state_log_update_open_stateid(state);
1665 }
1666 
1667 static void nfs_clear_open_stateid(struct nfs4_state *state,
1668 	nfs4_stateid *arg_stateid,
1669 	nfs4_stateid *stateid, fmode_t fmode)
1670 {
1671 	write_seqlock(&state->seqlock);
1672 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1673 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1674 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1675 	write_sequnlock(&state->seqlock);
1676 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1677 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1678 }
1679 
1680 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1681 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1682 	__must_hold(&state->owner->so_lock)
1683 	__must_hold(&state->seqlock)
1684 	__must_hold(RCU)
1685 
1686 {
1687 	DEFINE_WAIT(wait);
1688 	int status = 0;
1689 	for (;;) {
1690 
1691 		if (nfs_stateid_is_sequential(state, stateid))
1692 			break;
1693 
1694 		if (status)
1695 			break;
1696 		/* Rely on seqids for serialisation with NFSv4.0 */
1697 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1698 			break;
1699 
1700 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1701 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1702 		/*
1703 		 * Ensure we process the state changes in the same order
1704 		 * in which the server processed them by delaying the
1705 		 * update of the stateid until we are in sequence.
1706 		 */
1707 		write_sequnlock(&state->seqlock);
1708 		spin_unlock(&state->owner->so_lock);
1709 		rcu_read_unlock();
1710 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1711 
1712 		if (!fatal_signal_pending(current)) {
1713 			if (schedule_timeout(5*HZ) == 0)
1714 				status = -EAGAIN;
1715 			else
1716 				status = 0;
1717 		} else
1718 			status = -EINTR;
1719 		finish_wait(&state->waitq, &wait);
1720 		rcu_read_lock();
1721 		spin_lock(&state->owner->so_lock);
1722 		write_seqlock(&state->seqlock);
1723 	}
1724 
1725 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1726 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1727 		nfs4_stateid_copy(freeme, &state->open_stateid);
1728 		nfs_test_and_clear_all_open_stateid(state);
1729 	}
1730 
1731 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1732 		nfs4_stateid_copy(&state->stateid, stateid);
1733 	nfs4_stateid_copy(&state->open_stateid, stateid);
1734 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1735 	nfs_state_log_update_open_stateid(state);
1736 }
1737 
1738 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1739 		const nfs4_stateid *open_stateid,
1740 		fmode_t fmode,
1741 		nfs4_stateid *freeme)
1742 {
1743 	/*
1744 	 * Protect the call to nfs4_state_set_mode_locked and
1745 	 * serialise the stateid update
1746 	 */
1747 	write_seqlock(&state->seqlock);
1748 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1749 	switch (fmode) {
1750 	case FMODE_READ:
1751 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1752 		break;
1753 	case FMODE_WRITE:
1754 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1755 		break;
1756 	case FMODE_READ|FMODE_WRITE:
1757 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1758 	}
1759 	set_bit(NFS_OPEN_STATE, &state->flags);
1760 	write_sequnlock(&state->seqlock);
1761 }
1762 
1763 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1764 {
1765 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1766 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1767 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1768 	clear_bit(NFS_OPEN_STATE, &state->flags);
1769 }
1770 
1771 static void nfs_state_set_delegation(struct nfs4_state *state,
1772 		const nfs4_stateid *deleg_stateid,
1773 		fmode_t fmode)
1774 {
1775 	/*
1776 	 * Protect the call to nfs4_state_set_mode_locked and
1777 	 * serialise the stateid update
1778 	 */
1779 	write_seqlock(&state->seqlock);
1780 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1781 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1782 	write_sequnlock(&state->seqlock);
1783 }
1784 
1785 static void nfs_state_clear_delegation(struct nfs4_state *state)
1786 {
1787 	write_seqlock(&state->seqlock);
1788 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1789 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1790 	write_sequnlock(&state->seqlock);
1791 }
1792 
1793 int update_open_stateid(struct nfs4_state *state,
1794 		const nfs4_stateid *open_stateid,
1795 		const nfs4_stateid *delegation,
1796 		fmode_t fmode)
1797 {
1798 	struct nfs_server *server = NFS_SERVER(state->inode);
1799 	struct nfs_client *clp = server->nfs_client;
1800 	struct nfs_inode *nfsi = NFS_I(state->inode);
1801 	struct nfs_delegation *deleg_cur;
1802 	nfs4_stateid freeme = { };
1803 	int ret = 0;
1804 
1805 	fmode &= (FMODE_READ|FMODE_WRITE);
1806 
1807 	rcu_read_lock();
1808 	spin_lock(&state->owner->so_lock);
1809 	if (open_stateid != NULL) {
1810 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1811 		ret = 1;
1812 	}
1813 
1814 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1815 	if (deleg_cur == NULL)
1816 		goto no_delegation;
1817 
1818 	spin_lock(&deleg_cur->lock);
1819 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1820 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1821 	    (deleg_cur->type & fmode) != fmode)
1822 		goto no_delegation_unlock;
1823 
1824 	if (delegation == NULL)
1825 		delegation = &deleg_cur->stateid;
1826 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1827 		goto no_delegation_unlock;
1828 
1829 	nfs_mark_delegation_referenced(deleg_cur);
1830 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1831 	ret = 1;
1832 no_delegation_unlock:
1833 	spin_unlock(&deleg_cur->lock);
1834 no_delegation:
1835 	if (ret)
1836 		update_open_stateflags(state, fmode);
1837 	spin_unlock(&state->owner->so_lock);
1838 	rcu_read_unlock();
1839 
1840 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1841 		nfs4_schedule_state_manager(clp);
1842 	if (freeme.type != 0)
1843 		nfs4_test_and_free_stateid(server, &freeme,
1844 				state->owner->so_cred);
1845 
1846 	return ret;
1847 }
1848 
1849 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1850 		const nfs4_stateid *stateid)
1851 {
1852 	struct nfs4_state *state = lsp->ls_state;
1853 	bool ret = false;
1854 
1855 	spin_lock(&state->state_lock);
1856 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1857 		goto out_noupdate;
1858 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1859 		goto out_noupdate;
1860 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1861 	ret = true;
1862 out_noupdate:
1863 	spin_unlock(&state->state_lock);
1864 	return ret;
1865 }
1866 
1867 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1868 {
1869 	struct nfs_delegation *delegation;
1870 
1871 	fmode &= FMODE_READ|FMODE_WRITE;
1872 	rcu_read_lock();
1873 	delegation = nfs4_get_valid_delegation(inode);
1874 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1875 		rcu_read_unlock();
1876 		return;
1877 	}
1878 	rcu_read_unlock();
1879 	nfs4_inode_return_delegation(inode);
1880 }
1881 
1882 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1883 {
1884 	struct nfs4_state *state = opendata->state;
1885 	struct nfs_delegation *delegation;
1886 	int open_mode = opendata->o_arg.open_flags;
1887 	fmode_t fmode = opendata->o_arg.fmode;
1888 	enum open_claim_type4 claim = opendata->o_arg.claim;
1889 	nfs4_stateid stateid;
1890 	int ret = -EAGAIN;
1891 
1892 	for (;;) {
1893 		spin_lock(&state->owner->so_lock);
1894 		if (can_open_cached(state, fmode, open_mode, claim)) {
1895 			update_open_stateflags(state, fmode);
1896 			spin_unlock(&state->owner->so_lock);
1897 			goto out_return_state;
1898 		}
1899 		spin_unlock(&state->owner->so_lock);
1900 		rcu_read_lock();
1901 		delegation = nfs4_get_valid_delegation(state->inode);
1902 		if (!can_open_delegated(delegation, fmode, claim)) {
1903 			rcu_read_unlock();
1904 			break;
1905 		}
1906 		/* Save the delegation */
1907 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1908 		rcu_read_unlock();
1909 		nfs_release_seqid(opendata->o_arg.seqid);
1910 		if (!opendata->is_recover) {
1911 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1912 			if (ret != 0)
1913 				goto out;
1914 		}
1915 		ret = -EAGAIN;
1916 
1917 		/* Try to update the stateid using the delegation */
1918 		if (update_open_stateid(state, NULL, &stateid, fmode))
1919 			goto out_return_state;
1920 	}
1921 out:
1922 	return ERR_PTR(ret);
1923 out_return_state:
1924 	refcount_inc(&state->count);
1925 	return state;
1926 }
1927 
1928 static void
1929 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1930 {
1931 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1932 	struct nfs_delegation *delegation;
1933 	int delegation_flags = 0;
1934 
1935 	rcu_read_lock();
1936 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1937 	if (delegation)
1938 		delegation_flags = delegation->flags;
1939 	rcu_read_unlock();
1940 	switch (data->o_arg.claim) {
1941 	default:
1942 		break;
1943 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1944 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1945 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1946 				   "returning a delegation for "
1947 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1948 				   clp->cl_hostname);
1949 		return;
1950 	}
1951 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1952 		nfs_inode_set_delegation(state->inode,
1953 				data->owner->so_cred,
1954 				data->o_res.delegation_type,
1955 				&data->o_res.delegation,
1956 				data->o_res.pagemod_limit);
1957 	else
1958 		nfs_inode_reclaim_delegation(state->inode,
1959 				data->owner->so_cred,
1960 				data->o_res.delegation_type,
1961 				&data->o_res.delegation,
1962 				data->o_res.pagemod_limit);
1963 
1964 	if (data->o_res.do_recall)
1965 		nfs_async_inode_return_delegation(state->inode,
1966 						  &data->o_res.delegation);
1967 }
1968 
1969 /*
1970  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1971  * and update the nfs4_state.
1972  */
1973 static struct nfs4_state *
1974 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1975 {
1976 	struct inode *inode = data->state->inode;
1977 	struct nfs4_state *state = data->state;
1978 	int ret;
1979 
1980 	if (!data->rpc_done) {
1981 		if (data->rpc_status)
1982 			return ERR_PTR(data->rpc_status);
1983 		return nfs4_try_open_cached(data);
1984 	}
1985 
1986 	ret = nfs_refresh_inode(inode, &data->f_attr);
1987 	if (ret)
1988 		return ERR_PTR(ret);
1989 
1990 	if (data->o_res.delegation_type != 0)
1991 		nfs4_opendata_check_deleg(data, state);
1992 
1993 	if (!update_open_stateid(state, &data->o_res.stateid,
1994 				NULL, data->o_arg.fmode))
1995 		return ERR_PTR(-EAGAIN);
1996 	refcount_inc(&state->count);
1997 
1998 	return state;
1999 }
2000 
2001 static struct inode *
2002 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2003 {
2004 	struct inode *inode;
2005 
2006 	switch (data->o_arg.claim) {
2007 	case NFS4_OPEN_CLAIM_NULL:
2008 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2009 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2010 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2011 			return ERR_PTR(-EAGAIN);
2012 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2013 				&data->f_attr);
2014 		break;
2015 	default:
2016 		inode = d_inode(data->dentry);
2017 		ihold(inode);
2018 		nfs_refresh_inode(inode, &data->f_attr);
2019 	}
2020 	return inode;
2021 }
2022 
2023 static struct nfs4_state *
2024 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2025 {
2026 	struct nfs4_state *state;
2027 	struct inode *inode;
2028 
2029 	inode = nfs4_opendata_get_inode(data);
2030 	if (IS_ERR(inode))
2031 		return ERR_CAST(inode);
2032 	if (data->state != NULL && data->state->inode == inode) {
2033 		state = data->state;
2034 		refcount_inc(&state->count);
2035 	} else
2036 		state = nfs4_get_open_state(inode, data->owner);
2037 	iput(inode);
2038 	if (state == NULL)
2039 		state = ERR_PTR(-ENOMEM);
2040 	return state;
2041 }
2042 
2043 static struct nfs4_state *
2044 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2045 {
2046 	struct nfs4_state *state;
2047 
2048 	if (!data->rpc_done) {
2049 		state = nfs4_try_open_cached(data);
2050 		trace_nfs4_cached_open(data->state);
2051 		goto out;
2052 	}
2053 
2054 	state = nfs4_opendata_find_nfs4_state(data);
2055 	if (IS_ERR(state))
2056 		goto out;
2057 
2058 	if (data->o_res.delegation_type != 0)
2059 		nfs4_opendata_check_deleg(data, state);
2060 	if (!update_open_stateid(state, &data->o_res.stateid,
2061 				NULL, data->o_arg.fmode)) {
2062 		nfs4_put_open_state(state);
2063 		state = ERR_PTR(-EAGAIN);
2064 	}
2065 out:
2066 	nfs_release_seqid(data->o_arg.seqid);
2067 	return state;
2068 }
2069 
2070 static struct nfs4_state *
2071 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2072 {
2073 	struct nfs4_state *ret;
2074 
2075 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2076 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2077 	else
2078 		ret = _nfs4_opendata_to_nfs4_state(data);
2079 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2080 	return ret;
2081 }
2082 
2083 static struct nfs_open_context *
2084 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2085 {
2086 	struct nfs_inode *nfsi = NFS_I(state->inode);
2087 	struct nfs_open_context *ctx;
2088 
2089 	rcu_read_lock();
2090 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2091 		if (ctx->state != state)
2092 			continue;
2093 		if ((ctx->mode & mode) != mode)
2094 			continue;
2095 		if (!get_nfs_open_context(ctx))
2096 			continue;
2097 		rcu_read_unlock();
2098 		return ctx;
2099 	}
2100 	rcu_read_unlock();
2101 	return ERR_PTR(-ENOENT);
2102 }
2103 
2104 static struct nfs_open_context *
2105 nfs4_state_find_open_context(struct nfs4_state *state)
2106 {
2107 	struct nfs_open_context *ctx;
2108 
2109 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2110 	if (!IS_ERR(ctx))
2111 		return ctx;
2112 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2113 	if (!IS_ERR(ctx))
2114 		return ctx;
2115 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2116 }
2117 
2118 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2119 		struct nfs4_state *state, enum open_claim_type4 claim)
2120 {
2121 	struct nfs4_opendata *opendata;
2122 
2123 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2124 			NULL, claim, GFP_NOFS);
2125 	if (opendata == NULL)
2126 		return ERR_PTR(-ENOMEM);
2127 	opendata->state = state;
2128 	refcount_inc(&state->count);
2129 	return opendata;
2130 }
2131 
2132 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2133 				    fmode_t fmode)
2134 {
2135 	struct nfs4_state *newstate;
2136 	struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2137 	int openflags = opendata->o_arg.open_flags;
2138 	int ret;
2139 
2140 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2141 		return 0;
2142 	opendata->o_arg.fmode = fmode;
2143 	opendata->o_arg.share_access =
2144 		nfs4_map_atomic_open_share(server, fmode, openflags);
2145 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2146 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2147 	nfs4_init_opendata_res(opendata);
2148 	ret = _nfs4_recover_proc_open(opendata);
2149 	if (ret != 0)
2150 		return ret;
2151 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2152 	if (IS_ERR(newstate))
2153 		return PTR_ERR(newstate);
2154 	if (newstate != opendata->state)
2155 		ret = -ESTALE;
2156 	nfs4_close_state(newstate, fmode);
2157 	return ret;
2158 }
2159 
2160 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2161 {
2162 	int ret;
2163 
2164 	/* memory barrier prior to reading state->n_* */
2165 	smp_rmb();
2166 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2167 	if (ret != 0)
2168 		return ret;
2169 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2170 	if (ret != 0)
2171 		return ret;
2172 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2173 	if (ret != 0)
2174 		return ret;
2175 	/*
2176 	 * We may have performed cached opens for all three recoveries.
2177 	 * Check if we need to update the current stateid.
2178 	 */
2179 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2180 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2181 		write_seqlock(&state->seqlock);
2182 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2183 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2184 		write_sequnlock(&state->seqlock);
2185 	}
2186 	return 0;
2187 }
2188 
2189 /*
2190  * OPEN_RECLAIM:
2191  * 	reclaim state on the server after a reboot.
2192  */
2193 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2194 {
2195 	struct nfs_delegation *delegation;
2196 	struct nfs4_opendata *opendata;
2197 	fmode_t delegation_type = 0;
2198 	int status;
2199 
2200 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2201 			NFS4_OPEN_CLAIM_PREVIOUS);
2202 	if (IS_ERR(opendata))
2203 		return PTR_ERR(opendata);
2204 	rcu_read_lock();
2205 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2206 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2207 		delegation_type = delegation->type;
2208 	rcu_read_unlock();
2209 	opendata->o_arg.u.delegation_type = delegation_type;
2210 	status = nfs4_open_recover(opendata, state);
2211 	nfs4_opendata_put(opendata);
2212 	return status;
2213 }
2214 
2215 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2216 {
2217 	struct nfs_server *server = NFS_SERVER(state->inode);
2218 	struct nfs4_exception exception = { };
2219 	int err;
2220 	do {
2221 		err = _nfs4_do_open_reclaim(ctx, state);
2222 		trace_nfs4_open_reclaim(ctx, 0, err);
2223 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2224 			continue;
2225 		if (err != -NFS4ERR_DELAY)
2226 			break;
2227 		nfs4_handle_exception(server, err, &exception);
2228 	} while (exception.retry);
2229 	return err;
2230 }
2231 
2232 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2233 {
2234 	struct nfs_open_context *ctx;
2235 	int ret;
2236 
2237 	ctx = nfs4_state_find_open_context(state);
2238 	if (IS_ERR(ctx))
2239 		return -EAGAIN;
2240 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2241 	nfs_state_clear_open_state_flags(state);
2242 	ret = nfs4_do_open_reclaim(ctx, state);
2243 	put_nfs_open_context(ctx);
2244 	return ret;
2245 }
2246 
2247 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)
2248 {
2249 	switch (err) {
2250 		default:
2251 			printk(KERN_ERR "NFS: %s: unhandled error "
2252 					"%d.\n", __func__, err);
2253 			fallthrough;
2254 		case 0:
2255 		case -ENOENT:
2256 		case -EAGAIN:
2257 		case -ESTALE:
2258 		case -ETIMEDOUT:
2259 			break;
2260 		case -NFS4ERR_BADSESSION:
2261 		case -NFS4ERR_BADSLOT:
2262 		case -NFS4ERR_BAD_HIGH_SLOT:
2263 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2264 		case -NFS4ERR_DEADSESSION:
2265 			return -EAGAIN;
2266 		case -NFS4ERR_STALE_CLIENTID:
2267 		case -NFS4ERR_STALE_STATEID:
2268 			/* Don't recall a delegation if it was lost */
2269 			nfs4_schedule_lease_recovery(server->nfs_client);
2270 			return -EAGAIN;
2271 		case -NFS4ERR_MOVED:
2272 			nfs4_schedule_migration_recovery(server);
2273 			return -EAGAIN;
2274 		case -NFS4ERR_LEASE_MOVED:
2275 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2276 			return -EAGAIN;
2277 		case -NFS4ERR_DELEG_REVOKED:
2278 		case -NFS4ERR_ADMIN_REVOKED:
2279 		case -NFS4ERR_EXPIRED:
2280 		case -NFS4ERR_BAD_STATEID:
2281 		case -NFS4ERR_OPENMODE:
2282 			nfs_inode_find_state_and_recover(state->inode,
2283 					stateid);
2284 			nfs4_schedule_stateid_recovery(server, state);
2285 			return -EAGAIN;
2286 		case -NFS4ERR_DELAY:
2287 		case -NFS4ERR_GRACE:
2288 			ssleep(1);
2289 			return -EAGAIN;
2290 		case -ENOMEM:
2291 		case -NFS4ERR_DENIED:
2292 			if (fl) {
2293 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2294 				if (lsp)
2295 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2296 			}
2297 			return 0;
2298 	}
2299 	return err;
2300 }
2301 
2302 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2303 		struct nfs4_state *state, const nfs4_stateid *stateid)
2304 {
2305 	struct nfs_server *server = NFS_SERVER(state->inode);
2306 	struct nfs4_opendata *opendata;
2307 	int err = 0;
2308 
2309 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2310 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2311 	if (IS_ERR(opendata))
2312 		return PTR_ERR(opendata);
2313 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2314 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2315 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2316 		if (err)
2317 			goto out;
2318 	}
2319 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2320 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2321 		if (err)
2322 			goto out;
2323 	}
2324 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2325 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2326 		if (err)
2327 			goto out;
2328 	}
2329 	nfs_state_clear_delegation(state);
2330 out:
2331 	nfs4_opendata_put(opendata);
2332 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2333 }
2334 
2335 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2336 {
2337 	struct nfs4_opendata *data = calldata;
2338 
2339 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2340 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2341 }
2342 
2343 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2344 {
2345 	struct nfs4_opendata *data = calldata;
2346 
2347 	nfs40_sequence_done(task, &data->c_res.seq_res);
2348 
2349 	data->rpc_status = task->tk_status;
2350 	if (data->rpc_status == 0) {
2351 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2352 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2353 		renew_lease(data->o_res.server, data->timestamp);
2354 		data->rpc_done = true;
2355 	}
2356 }
2357 
2358 static void nfs4_open_confirm_release(void *calldata)
2359 {
2360 	struct nfs4_opendata *data = calldata;
2361 	struct nfs4_state *state = NULL;
2362 
2363 	/* If this request hasn't been cancelled, do nothing */
2364 	if (!data->cancelled)
2365 		goto out_free;
2366 	/* In case of error, no cleanup! */
2367 	if (!data->rpc_done)
2368 		goto out_free;
2369 	state = nfs4_opendata_to_nfs4_state(data);
2370 	if (!IS_ERR(state))
2371 		nfs4_close_state(state, data->o_arg.fmode);
2372 out_free:
2373 	nfs4_opendata_put(data);
2374 }
2375 
2376 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2377 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2378 	.rpc_call_done = nfs4_open_confirm_done,
2379 	.rpc_release = nfs4_open_confirm_release,
2380 };
2381 
2382 /*
2383  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2384  */
2385 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2386 {
2387 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2388 	struct rpc_task *task;
2389 	struct  rpc_message msg = {
2390 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2391 		.rpc_argp = &data->c_arg,
2392 		.rpc_resp = &data->c_res,
2393 		.rpc_cred = data->owner->so_cred,
2394 	};
2395 	struct rpc_task_setup task_setup_data = {
2396 		.rpc_client = server->client,
2397 		.rpc_message = &msg,
2398 		.callback_ops = &nfs4_open_confirm_ops,
2399 		.callback_data = data,
2400 		.workqueue = nfsiod_workqueue,
2401 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2402 	};
2403 	int status;
2404 
2405 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2406 				data->is_recover);
2407 	kref_get(&data->kref);
2408 	data->rpc_done = false;
2409 	data->rpc_status = 0;
2410 	data->timestamp = jiffies;
2411 	task = rpc_run_task(&task_setup_data);
2412 	if (IS_ERR(task))
2413 		return PTR_ERR(task);
2414 	status = rpc_wait_for_completion_task(task);
2415 	if (status != 0) {
2416 		data->cancelled = true;
2417 		smp_wmb();
2418 	} else
2419 		status = data->rpc_status;
2420 	rpc_put_task(task);
2421 	return status;
2422 }
2423 
2424 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2425 {
2426 	struct nfs4_opendata *data = calldata;
2427 	struct nfs4_state_owner *sp = data->owner;
2428 	struct nfs_client *clp = sp->so_server->nfs_client;
2429 	enum open_claim_type4 claim = data->o_arg.claim;
2430 
2431 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2432 		goto out_wait;
2433 	/*
2434 	 * Check if we still need to send an OPEN call, or if we can use
2435 	 * a delegation instead.
2436 	 */
2437 	if (data->state != NULL) {
2438 		struct nfs_delegation *delegation;
2439 
2440 		if (can_open_cached(data->state, data->o_arg.fmode,
2441 					data->o_arg.open_flags, claim))
2442 			goto out_no_action;
2443 		rcu_read_lock();
2444 		delegation = nfs4_get_valid_delegation(data->state->inode);
2445 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2446 			goto unlock_no_action;
2447 		rcu_read_unlock();
2448 	}
2449 	/* Update client id. */
2450 	data->o_arg.clientid = clp->cl_clientid;
2451 	switch (claim) {
2452 	default:
2453 		break;
2454 	case NFS4_OPEN_CLAIM_PREVIOUS:
2455 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2456 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2457 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2458 		fallthrough;
2459 	case NFS4_OPEN_CLAIM_FH:
2460 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2461 	}
2462 	data->timestamp = jiffies;
2463 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2464 				&data->o_arg.seq_args,
2465 				&data->o_res.seq_res,
2466 				task) != 0)
2467 		nfs_release_seqid(data->o_arg.seqid);
2468 
2469 	/* Set the create mode (note dependency on the session type) */
2470 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2471 	if (data->o_arg.open_flags & O_EXCL) {
2472 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2473 		if (clp->cl_mvops->minor_version == 0) {
2474 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2475 			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2476 			 * because ACCESS will return permission denied for
2477 			 * all bits until close */
2478 			data->o_res.access_request = data->o_arg.access = 0;
2479 		} else if (nfs4_has_persistent_session(clp))
2480 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2481 	}
2482 	return;
2483 unlock_no_action:
2484 	trace_nfs4_cached_open(data->state);
2485 	rcu_read_unlock();
2486 out_no_action:
2487 	task->tk_action = NULL;
2488 out_wait:
2489 	nfs4_sequence_done(task, &data->o_res.seq_res);
2490 }
2491 
2492 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2493 {
2494 	struct nfs4_opendata *data = calldata;
2495 
2496 	data->rpc_status = task->tk_status;
2497 
2498 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2499 		return;
2500 
2501 	if (task->tk_status == 0) {
2502 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2503 			switch (data->o_res.f_attr->mode & S_IFMT) {
2504 			case S_IFREG:
2505 				break;
2506 			case S_IFLNK:
2507 				data->rpc_status = -ELOOP;
2508 				break;
2509 			case S_IFDIR:
2510 				data->rpc_status = -EISDIR;
2511 				break;
2512 			default:
2513 				data->rpc_status = -ENOTDIR;
2514 			}
2515 		}
2516 		renew_lease(data->o_res.server, data->timestamp);
2517 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2518 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2519 	}
2520 	data->rpc_done = true;
2521 }
2522 
2523 static void nfs4_open_release(void *calldata)
2524 {
2525 	struct nfs4_opendata *data = calldata;
2526 	struct nfs4_state *state = NULL;
2527 
2528 	/* If this request hasn't been cancelled, do nothing */
2529 	if (!data->cancelled)
2530 		goto out_free;
2531 	/* In case of error, no cleanup! */
2532 	if (data->rpc_status != 0 || !data->rpc_done)
2533 		goto out_free;
2534 	/* In case we need an open_confirm, no cleanup! */
2535 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2536 		goto out_free;
2537 	state = nfs4_opendata_to_nfs4_state(data);
2538 	if (!IS_ERR(state))
2539 		nfs4_close_state(state, data->o_arg.fmode);
2540 out_free:
2541 	nfs4_opendata_put(data);
2542 }
2543 
2544 static const struct rpc_call_ops nfs4_open_ops = {
2545 	.rpc_call_prepare = nfs4_open_prepare,
2546 	.rpc_call_done = nfs4_open_done,
2547 	.rpc_release = nfs4_open_release,
2548 };
2549 
2550 static int nfs4_run_open_task(struct nfs4_opendata *data,
2551 			      struct nfs_open_context *ctx)
2552 {
2553 	struct inode *dir = d_inode(data->dir);
2554 	struct nfs_server *server = NFS_SERVER(dir);
2555 	struct nfs_openargs *o_arg = &data->o_arg;
2556 	struct nfs_openres *o_res = &data->o_res;
2557 	struct rpc_task *task;
2558 	struct rpc_message msg = {
2559 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2560 		.rpc_argp = o_arg,
2561 		.rpc_resp = o_res,
2562 		.rpc_cred = data->owner->so_cred,
2563 	};
2564 	struct rpc_task_setup task_setup_data = {
2565 		.rpc_client = server->client,
2566 		.rpc_message = &msg,
2567 		.callback_ops = &nfs4_open_ops,
2568 		.callback_data = data,
2569 		.workqueue = nfsiod_workqueue,
2570 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2571 	};
2572 	int status;
2573 
2574 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2575 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2576 
2577 	kref_get(&data->kref);
2578 	data->rpc_done = false;
2579 	data->rpc_status = 0;
2580 	data->cancelled = false;
2581 	data->is_recover = false;
2582 	if (!ctx) {
2583 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2584 		data->is_recover = true;
2585 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2586 	} else {
2587 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2588 		pnfs_lgopen_prepare(data, ctx);
2589 	}
2590 	task = rpc_run_task(&task_setup_data);
2591 	if (IS_ERR(task))
2592 		return PTR_ERR(task);
2593 	status = rpc_wait_for_completion_task(task);
2594 	if (status != 0) {
2595 		data->cancelled = true;
2596 		smp_wmb();
2597 	} else
2598 		status = data->rpc_status;
2599 	rpc_put_task(task);
2600 
2601 	return status;
2602 }
2603 
2604 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2605 {
2606 	struct inode *dir = d_inode(data->dir);
2607 	struct nfs_openres *o_res = &data->o_res;
2608 	int status;
2609 
2610 	status = nfs4_run_open_task(data, NULL);
2611 	if (status != 0 || !data->rpc_done)
2612 		return status;
2613 
2614 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2615 
2616 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2617 		status = _nfs4_proc_open_confirm(data);
2618 
2619 	return status;
2620 }
2621 
2622 /*
2623  * Additional permission checks in order to distinguish between an
2624  * open for read, and an open for execute. This works around the
2625  * fact that NFSv4 OPEN treats read and execute permissions as being
2626  * the same.
2627  * Note that in the non-execute case, we want to turn off permission
2628  * checking if we just created a new file (POSIX open() semantics).
2629  */
2630 static int nfs4_opendata_access(const struct cred *cred,
2631 				struct nfs4_opendata *opendata,
2632 				struct nfs4_state *state, fmode_t fmode)
2633 {
2634 	struct nfs_access_entry cache;
2635 	u32 mask, flags;
2636 
2637 	/* access call failed or for some reason the server doesn't
2638 	 * support any access modes -- defer access call until later */
2639 	if (opendata->o_res.access_supported == 0)
2640 		return 0;
2641 
2642 	mask = 0;
2643 	if (fmode & FMODE_EXEC) {
2644 		/* ONLY check for exec rights */
2645 		if (S_ISDIR(state->inode->i_mode))
2646 			mask = NFS4_ACCESS_LOOKUP;
2647 		else
2648 			mask = NFS4_ACCESS_EXECUTE;
2649 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2650 		mask = NFS4_ACCESS_READ;
2651 
2652 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2653 	nfs_access_add_cache(state->inode, &cache, cred);
2654 
2655 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2656 	if ((mask & ~cache.mask & flags) == 0)
2657 		return 0;
2658 
2659 	return -EACCES;
2660 }
2661 
2662 /*
2663  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2664  */
2665 static int _nfs4_proc_open(struct nfs4_opendata *data,
2666 			   struct nfs_open_context *ctx)
2667 {
2668 	struct inode *dir = d_inode(data->dir);
2669 	struct nfs_server *server = NFS_SERVER(dir);
2670 	struct nfs_openargs *o_arg = &data->o_arg;
2671 	struct nfs_openres *o_res = &data->o_res;
2672 	int status;
2673 
2674 	status = nfs4_run_open_task(data, ctx);
2675 	if (!data->rpc_done)
2676 		return status;
2677 	if (status != 0) {
2678 		if (status == -NFS4ERR_BADNAME &&
2679 				!(o_arg->open_flags & O_CREAT))
2680 			return -ENOENT;
2681 		return status;
2682 	}
2683 
2684 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2685 
2686 	if (o_arg->open_flags & O_CREAT) {
2687 		if (o_arg->open_flags & O_EXCL)
2688 			data->file_created = true;
2689 		else if (o_res->cinfo.before != o_res->cinfo.after)
2690 			data->file_created = true;
2691 		if (data->file_created ||
2692 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2693 			nfs4_update_changeattr(dir, &o_res->cinfo,
2694 					o_res->f_attr->time_start,
2695 					NFS_INO_INVALID_DATA);
2696 	}
2697 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2698 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2699 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2700 		status = _nfs4_proc_open_confirm(data);
2701 		if (status != 0)
2702 			return status;
2703 	}
2704 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2705 		nfs4_sequence_free_slot(&o_res->seq_res);
2706 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL);
2707 	}
2708 	return 0;
2709 }
2710 
2711 /*
2712  * OPEN_EXPIRED:
2713  * 	reclaim state on the server after a network partition.
2714  * 	Assumes caller holds the appropriate lock
2715  */
2716 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2717 {
2718 	struct nfs4_opendata *opendata;
2719 	int ret;
2720 
2721 	opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2722 	if (IS_ERR(opendata))
2723 		return PTR_ERR(opendata);
2724 	/*
2725 	 * We're not recovering a delegation, so ask for no delegation.
2726 	 * Otherwise the recovery thread could deadlock with an outstanding
2727 	 * delegation return.
2728 	 */
2729 	opendata->o_arg.open_flags = O_DIRECT;
2730 	ret = nfs4_open_recover(opendata, state);
2731 	if (ret == -ESTALE)
2732 		d_drop(ctx->dentry);
2733 	nfs4_opendata_put(opendata);
2734 	return ret;
2735 }
2736 
2737 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2738 {
2739 	struct nfs_server *server = NFS_SERVER(state->inode);
2740 	struct nfs4_exception exception = { };
2741 	int err;
2742 
2743 	do {
2744 		err = _nfs4_open_expired(ctx, state);
2745 		trace_nfs4_open_expired(ctx, 0, err);
2746 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2747 			continue;
2748 		switch (err) {
2749 		default:
2750 			goto out;
2751 		case -NFS4ERR_GRACE:
2752 		case -NFS4ERR_DELAY:
2753 			nfs4_handle_exception(server, err, &exception);
2754 			err = 0;
2755 		}
2756 	} while (exception.retry);
2757 out:
2758 	return err;
2759 }
2760 
2761 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2762 {
2763 	struct nfs_open_context *ctx;
2764 	int ret;
2765 
2766 	ctx = nfs4_state_find_open_context(state);
2767 	if (IS_ERR(ctx))
2768 		return -EAGAIN;
2769 	ret = nfs4_do_open_expired(ctx, state);
2770 	put_nfs_open_context(ctx);
2771 	return ret;
2772 }
2773 
2774 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2775 		const nfs4_stateid *stateid)
2776 {
2777 	nfs_remove_bad_delegation(state->inode, stateid);
2778 	nfs_state_clear_delegation(state);
2779 }
2780 
2781 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2782 {
2783 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2784 		nfs_finish_clear_delegation_stateid(state, NULL);
2785 }
2786 
2787 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2788 {
2789 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2790 	nfs40_clear_delegation_stateid(state);
2791 	nfs_state_clear_open_state_flags(state);
2792 	return nfs4_open_expired(sp, state);
2793 }
2794 
2795 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2796 		nfs4_stateid *stateid,
2797 		const struct cred *cred)
2798 {
2799 	return -NFS4ERR_BAD_STATEID;
2800 }
2801 
2802 #if defined(CONFIG_NFS_V4_1)
2803 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2804 		nfs4_stateid *stateid,
2805 		const struct cred *cred)
2806 {
2807 	int status;
2808 
2809 	switch (stateid->type) {
2810 	default:
2811 		break;
2812 	case NFS4_INVALID_STATEID_TYPE:
2813 	case NFS4_SPECIAL_STATEID_TYPE:
2814 		return -NFS4ERR_BAD_STATEID;
2815 	case NFS4_REVOKED_STATEID_TYPE:
2816 		goto out_free;
2817 	}
2818 
2819 	status = nfs41_test_stateid(server, stateid, cred);
2820 	switch (status) {
2821 	case -NFS4ERR_EXPIRED:
2822 	case -NFS4ERR_ADMIN_REVOKED:
2823 	case -NFS4ERR_DELEG_REVOKED:
2824 		break;
2825 	default:
2826 		return status;
2827 	}
2828 out_free:
2829 	/* Ack the revoked state to the server */
2830 	nfs41_free_stateid(server, stateid, cred, true);
2831 	return -NFS4ERR_EXPIRED;
2832 }
2833 
2834 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2835 {
2836 	struct nfs_server *server = NFS_SERVER(state->inode);
2837 	nfs4_stateid stateid;
2838 	struct nfs_delegation *delegation;
2839 	const struct cred *cred = NULL;
2840 	int status, ret = NFS_OK;
2841 
2842 	/* Get the delegation credential for use by test/free_stateid */
2843 	rcu_read_lock();
2844 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2845 	if (delegation == NULL) {
2846 		rcu_read_unlock();
2847 		nfs_state_clear_delegation(state);
2848 		return NFS_OK;
2849 	}
2850 
2851 	spin_lock(&delegation->lock);
2852 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2853 
2854 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2855 				&delegation->flags)) {
2856 		spin_unlock(&delegation->lock);
2857 		rcu_read_unlock();
2858 		return NFS_OK;
2859 	}
2860 
2861 	if (delegation->cred)
2862 		cred = get_cred(delegation->cred);
2863 	spin_unlock(&delegation->lock);
2864 	rcu_read_unlock();
2865 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2866 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2867 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2868 		nfs_finish_clear_delegation_stateid(state, &stateid);
2869 	else
2870 		ret = status;
2871 
2872 	put_cred(cred);
2873 	return ret;
2874 }
2875 
2876 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2877 {
2878 	nfs4_stateid tmp;
2879 
2880 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2881 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2882 				&tmp, NULL) &&
2883 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2884 		nfs_state_set_delegation(state, &tmp, state->state);
2885 	else
2886 		nfs_state_clear_delegation(state);
2887 }
2888 
2889 /**
2890  * nfs41_check_expired_locks - possibly free a lock stateid
2891  *
2892  * @state: NFSv4 state for an inode
2893  *
2894  * Returns NFS_OK if recovery for this stateid is now finished.
2895  * Otherwise a negative NFS4ERR value is returned.
2896  */
2897 static int nfs41_check_expired_locks(struct nfs4_state *state)
2898 {
2899 	int status, ret = NFS_OK;
2900 	struct nfs4_lock_state *lsp, *prev = NULL;
2901 	struct nfs_server *server = NFS_SERVER(state->inode);
2902 
2903 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2904 		goto out;
2905 
2906 	spin_lock(&state->state_lock);
2907 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2908 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2909 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2910 
2911 			refcount_inc(&lsp->ls_count);
2912 			spin_unlock(&state->state_lock);
2913 
2914 			nfs4_put_lock_state(prev);
2915 			prev = lsp;
2916 
2917 			status = nfs41_test_and_free_expired_stateid(server,
2918 					&lsp->ls_stateid,
2919 					cred);
2920 			trace_nfs4_test_lock_stateid(state, lsp, status);
2921 			if (status == -NFS4ERR_EXPIRED ||
2922 			    status == -NFS4ERR_BAD_STATEID) {
2923 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2924 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2925 				if (!recover_lost_locks)
2926 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2927 			} else if (status != NFS_OK) {
2928 				ret = status;
2929 				nfs4_put_lock_state(prev);
2930 				goto out;
2931 			}
2932 			spin_lock(&state->state_lock);
2933 		}
2934 	}
2935 	spin_unlock(&state->state_lock);
2936 	nfs4_put_lock_state(prev);
2937 out:
2938 	return ret;
2939 }
2940 
2941 /**
2942  * nfs41_check_open_stateid - possibly free an open stateid
2943  *
2944  * @state: NFSv4 state for an inode
2945  *
2946  * Returns NFS_OK if recovery for this stateid is now finished.
2947  * Otherwise a negative NFS4ERR value is returned.
2948  */
2949 static int nfs41_check_open_stateid(struct nfs4_state *state)
2950 {
2951 	struct nfs_server *server = NFS_SERVER(state->inode);
2952 	nfs4_stateid *stateid = &state->open_stateid;
2953 	const struct cred *cred = state->owner->so_cred;
2954 	int status;
2955 
2956 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2957 		return -NFS4ERR_BAD_STATEID;
2958 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2959 	trace_nfs4_test_open_stateid(state, NULL, status);
2960 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2961 		nfs_state_clear_open_state_flags(state);
2962 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2963 		return status;
2964 	}
2965 	if (nfs_open_stateid_recover_openmode(state))
2966 		return -NFS4ERR_OPENMODE;
2967 	return NFS_OK;
2968 }
2969 
2970 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2971 {
2972 	int status;
2973 
2974 	status = nfs41_check_delegation_stateid(state);
2975 	if (status != NFS_OK)
2976 		return status;
2977 	nfs41_delegation_recover_stateid(state);
2978 
2979 	status = nfs41_check_expired_locks(state);
2980 	if (status != NFS_OK)
2981 		return status;
2982 	status = nfs41_check_open_stateid(state);
2983 	if (status != NFS_OK)
2984 		status = nfs4_open_expired(sp, state);
2985 	return status;
2986 }
2987 #endif
2988 
2989 /*
2990  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2991  * fields corresponding to attributes that were used to store the verifier.
2992  * Make sure we clobber those fields in the later setattr call
2993  */
2994 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2995 				struct iattr *sattr, struct nfs4_label **label)
2996 {
2997 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2998 	__u32 attrset[3];
2999 	unsigned ret;
3000 	unsigned i;
3001 
3002 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3003 		attrset[i] = opendata->o_res.attrset[i];
3004 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3005 			attrset[i] &= ~bitmask[i];
3006 	}
3007 
3008 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3009 		sattr->ia_valid : 0;
3010 
3011 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3012 		if (sattr->ia_valid & ATTR_ATIME_SET)
3013 			ret |= ATTR_ATIME_SET;
3014 		else
3015 			ret |= ATTR_ATIME;
3016 	}
3017 
3018 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3019 		if (sattr->ia_valid & ATTR_MTIME_SET)
3020 			ret |= ATTR_MTIME_SET;
3021 		else
3022 			ret |= ATTR_MTIME;
3023 	}
3024 
3025 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3026 		*label = NULL;
3027 	return ret;
3028 }
3029 
3030 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3031 		struct nfs_open_context *ctx)
3032 {
3033 	struct nfs4_state_owner *sp = opendata->owner;
3034 	struct nfs_server *server = sp->so_server;
3035 	struct dentry *dentry;
3036 	struct nfs4_state *state;
3037 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3038 	struct inode *dir = d_inode(opendata->dir);
3039 	unsigned long dir_verifier;
3040 	unsigned int seq;
3041 	int ret;
3042 
3043 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3044 	dir_verifier = nfs_save_change_attribute(dir);
3045 
3046 	ret = _nfs4_proc_open(opendata, ctx);
3047 	if (ret != 0)
3048 		goto out;
3049 
3050 	state = _nfs4_opendata_to_nfs4_state(opendata);
3051 	ret = PTR_ERR(state);
3052 	if (IS_ERR(state))
3053 		goto out;
3054 	ctx->state = state;
3055 	if (server->caps & NFS_CAP_POSIX_LOCK)
3056 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3057 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3058 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3059 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3060 		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3061 
3062 	dentry = opendata->dentry;
3063 	if (d_really_is_negative(dentry)) {
3064 		struct dentry *alias;
3065 		d_drop(dentry);
3066 		alias = d_exact_alias(dentry, state->inode);
3067 		if (!alias)
3068 			alias = d_splice_alias(igrab(state->inode), dentry);
3069 		/* d_splice_alias() can't fail here - it's a non-directory */
3070 		if (alias) {
3071 			dput(ctx->dentry);
3072 			ctx->dentry = dentry = alias;
3073 		}
3074 	}
3075 
3076 	switch(opendata->o_arg.claim) {
3077 	default:
3078 		break;
3079 	case NFS4_OPEN_CLAIM_NULL:
3080 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3081 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3082 		if (!opendata->rpc_done)
3083 			break;
3084 		if (opendata->o_res.delegation_type != 0)
3085 			dir_verifier = nfs_save_change_attribute(dir);
3086 		nfs_set_verifier(dentry, dir_verifier);
3087 	}
3088 
3089 	/* Parse layoutget results before we check for access */
3090 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3091 
3092 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3093 	if (ret != 0)
3094 		goto out;
3095 
3096 	if (d_inode(dentry) == state->inode) {
3097 		nfs_inode_attach_open_context(ctx);
3098 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3099 			nfs4_schedule_stateid_recovery(server, state);
3100 	}
3101 
3102 out:
3103 	if (!opendata->cancelled) {
3104 		if (opendata->lgp) {
3105 			nfs4_lgopen_release(opendata->lgp);
3106 			opendata->lgp = NULL;
3107 		}
3108 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3109 	}
3110 	return ret;
3111 }
3112 
3113 /*
3114  * Returns a referenced nfs4_state
3115  */
3116 static int _nfs4_do_open(struct inode *dir,
3117 			struct nfs_open_context *ctx,
3118 			int flags,
3119 			const struct nfs4_open_createattrs *c,
3120 			int *opened)
3121 {
3122 	struct nfs4_state_owner  *sp;
3123 	struct nfs4_state     *state = NULL;
3124 	struct nfs_server       *server = NFS_SERVER(dir);
3125 	struct nfs4_opendata *opendata;
3126 	struct dentry *dentry = ctx->dentry;
3127 	const struct cred *cred = ctx->cred;
3128 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3129 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3130 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3131 	struct iattr *sattr = c->sattr;
3132 	struct nfs4_label *label = c->label;
3133 	int status;
3134 
3135 	/* Protect against reboot recovery conflicts */
3136 	status = -ENOMEM;
3137 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3138 	if (sp == NULL) {
3139 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3140 		goto out_err;
3141 	}
3142 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3143 	if (status != 0)
3144 		goto err_put_state_owner;
3145 	if (d_really_is_positive(dentry))
3146 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3147 	status = -ENOMEM;
3148 	if (d_really_is_positive(dentry))
3149 		claim = NFS4_OPEN_CLAIM_FH;
3150 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3151 			c, claim, GFP_KERNEL);
3152 	if (opendata == NULL)
3153 		goto err_put_state_owner;
3154 
3155 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3156 		if (!opendata->f_attr.mdsthreshold) {
3157 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3158 			if (!opendata->f_attr.mdsthreshold)
3159 				goto err_opendata_put;
3160 		}
3161 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3162 	}
3163 	if (d_really_is_positive(dentry))
3164 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3165 
3166 	status = _nfs4_open_and_get_state(opendata, ctx);
3167 	if (status != 0)
3168 		goto err_opendata_put;
3169 	state = ctx->state;
3170 
3171 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3172 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3173 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3174 		/*
3175 		 * send create attributes which was not set by open
3176 		 * with an extra setattr.
3177 		 */
3178 		if (attrs || label) {
3179 			unsigned ia_old = sattr->ia_valid;
3180 
3181 			sattr->ia_valid = attrs;
3182 			nfs_fattr_init(opendata->o_res.f_attr);
3183 			status = nfs4_do_setattr(state->inode, cred,
3184 					opendata->o_res.f_attr, sattr,
3185 					ctx, label);
3186 			if (status == 0) {
3187 				nfs_setattr_update_inode(state->inode, sattr,
3188 						opendata->o_res.f_attr);
3189 				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3190 			}
3191 			sattr->ia_valid = ia_old;
3192 		}
3193 	}
3194 	if (opened && opendata->file_created)
3195 		*opened = 1;
3196 
3197 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3198 		*ctx_th = opendata->f_attr.mdsthreshold;
3199 		opendata->f_attr.mdsthreshold = NULL;
3200 	}
3201 
3202 	nfs4_opendata_put(opendata);
3203 	nfs4_put_state_owner(sp);
3204 	return 0;
3205 err_opendata_put:
3206 	nfs4_opendata_put(opendata);
3207 err_put_state_owner:
3208 	nfs4_put_state_owner(sp);
3209 out_err:
3210 	return status;
3211 }
3212 
3213 
3214 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3215 					struct nfs_open_context *ctx,
3216 					int flags,
3217 					struct iattr *sattr,
3218 					struct nfs4_label *label,
3219 					int *opened)
3220 {
3221 	struct nfs_server *server = NFS_SERVER(dir);
3222 	struct nfs4_exception exception = {
3223 		.interruptible = true,
3224 	};
3225 	struct nfs4_state *res;
3226 	struct nfs4_open_createattrs c = {
3227 		.label = label,
3228 		.sattr = sattr,
3229 		.verf = {
3230 			[0] = (__u32)jiffies,
3231 			[1] = (__u32)current->pid,
3232 		},
3233 	};
3234 	int status;
3235 
3236 	do {
3237 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3238 		res = ctx->state;
3239 		trace_nfs4_open_file(ctx, flags, status);
3240 		if (status == 0)
3241 			break;
3242 		/* NOTE: BAD_SEQID means the server and client disagree about the
3243 		 * book-keeping w.r.t. state-changing operations
3244 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3245 		 * It is actually a sign of a bug on the client or on the server.
3246 		 *
3247 		 * If we receive a BAD_SEQID error in the particular case of
3248 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3249 		 * have unhashed the old state_owner for us, and that we can
3250 		 * therefore safely retry using a new one. We should still warn
3251 		 * the user though...
3252 		 */
3253 		if (status == -NFS4ERR_BAD_SEQID) {
3254 			pr_warn_ratelimited("NFS: v4 server %s "
3255 					" returned a bad sequence-id error!\n",
3256 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3257 			exception.retry = 1;
3258 			continue;
3259 		}
3260 		/*
3261 		 * BAD_STATEID on OPEN means that the server cancelled our
3262 		 * state before it received the OPEN_CONFIRM.
3263 		 * Recover by retrying the request as per the discussion
3264 		 * on Page 181 of RFC3530.
3265 		 */
3266 		if (status == -NFS4ERR_BAD_STATEID) {
3267 			exception.retry = 1;
3268 			continue;
3269 		}
3270 		if (status == -NFS4ERR_EXPIRED) {
3271 			nfs4_schedule_lease_recovery(server->nfs_client);
3272 			exception.retry = 1;
3273 			continue;
3274 		}
3275 		if (status == -EAGAIN) {
3276 			/* We must have found a delegation */
3277 			exception.retry = 1;
3278 			continue;
3279 		}
3280 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3281 			continue;
3282 		res = ERR_PTR(nfs4_handle_exception(server,
3283 					status, &exception));
3284 	} while (exception.retry);
3285 	return res;
3286 }
3287 
3288 static int _nfs4_do_setattr(struct inode *inode,
3289 			    struct nfs_setattrargs *arg,
3290 			    struct nfs_setattrres *res,
3291 			    const struct cred *cred,
3292 			    struct nfs_open_context *ctx)
3293 {
3294 	struct nfs_server *server = NFS_SERVER(inode);
3295 	struct rpc_message msg = {
3296 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3297 		.rpc_argp	= arg,
3298 		.rpc_resp	= res,
3299 		.rpc_cred	= cred,
3300 	};
3301 	const struct cred *delegation_cred = NULL;
3302 	unsigned long timestamp = jiffies;
3303 	bool truncate;
3304 	int status;
3305 
3306 	nfs_fattr_init(res->fattr);
3307 
3308 	/* Servers should only apply open mode checks for file size changes */
3309 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3310 	if (!truncate) {
3311 		nfs4_inode_make_writeable(inode);
3312 		goto zero_stateid;
3313 	}
3314 
3315 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3316 		/* Use that stateid */
3317 	} else if (ctx != NULL && ctx->state) {
3318 		struct nfs_lock_context *l_ctx;
3319 		if (!nfs4_valid_open_stateid(ctx->state))
3320 			return -EBADF;
3321 		l_ctx = nfs_get_lock_context(ctx);
3322 		if (IS_ERR(l_ctx))
3323 			return PTR_ERR(l_ctx);
3324 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3325 						&arg->stateid, &delegation_cred);
3326 		nfs_put_lock_context(l_ctx);
3327 		if (status == -EIO)
3328 			return -EBADF;
3329 		else if (status == -EAGAIN)
3330 			goto zero_stateid;
3331 	} else {
3332 zero_stateid:
3333 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3334 	}
3335 	if (delegation_cred)
3336 		msg.rpc_cred = delegation_cred;
3337 
3338 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3339 
3340 	put_cred(delegation_cred);
3341 	if (status == 0 && ctx != NULL)
3342 		renew_lease(server, timestamp);
3343 	trace_nfs4_setattr(inode, &arg->stateid, status);
3344 	return status;
3345 }
3346 
3347 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3348 			   struct nfs_fattr *fattr, struct iattr *sattr,
3349 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3350 {
3351 	struct nfs_server *server = NFS_SERVER(inode);
3352 	__u32 bitmask[NFS4_BITMASK_SZ];
3353 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3354 	struct nfs_setattrargs	arg = {
3355 		.fh		= NFS_FH(inode),
3356 		.iap		= sattr,
3357 		.server		= server,
3358 		.bitmask = bitmask,
3359 		.label		= ilabel,
3360 	};
3361 	struct nfs_setattrres  res = {
3362 		.fattr		= fattr,
3363 		.server		= server,
3364 	};
3365 	struct nfs4_exception exception = {
3366 		.state = state,
3367 		.inode = inode,
3368 		.stateid = &arg.stateid,
3369 	};
3370 	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3371 	int err;
3372 
3373 	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3374 		adjust_flags |= NFS_INO_INVALID_MODE;
3375 	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3376 		adjust_flags |= NFS_INO_INVALID_OTHER;
3377 
3378 	do {
3379 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3380 					inode, adjust_flags);
3381 
3382 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3383 		switch (err) {
3384 		case -NFS4ERR_OPENMODE:
3385 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3386 				pr_warn_once("NFSv4: server %s is incorrectly "
3387 						"applying open mode checks to "
3388 						"a SETATTR that is not "
3389 						"changing file size.\n",
3390 						server->nfs_client->cl_hostname);
3391 			}
3392 			if (state && !(state->state & FMODE_WRITE)) {
3393 				err = -EBADF;
3394 				if (sattr->ia_valid & ATTR_OPEN)
3395 					err = -EACCES;
3396 				goto out;
3397 			}
3398 		}
3399 		err = nfs4_handle_exception(server, err, &exception);
3400 	} while (exception.retry);
3401 out:
3402 	return err;
3403 }
3404 
3405 static bool
3406 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3407 {
3408 	if (inode == NULL || !nfs_have_layout(inode))
3409 		return false;
3410 
3411 	return pnfs_wait_on_layoutreturn(inode, task);
3412 }
3413 
3414 /*
3415  * Update the seqid of an open stateid
3416  */
3417 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3418 		struct nfs4_state *state)
3419 {
3420 	__be32 seqid_open;
3421 	u32 dst_seqid;
3422 	int seq;
3423 
3424 	for (;;) {
3425 		if (!nfs4_valid_open_stateid(state))
3426 			break;
3427 		seq = read_seqbegin(&state->seqlock);
3428 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3429 			nfs4_stateid_copy(dst, &state->open_stateid);
3430 			if (read_seqretry(&state->seqlock, seq))
3431 				continue;
3432 			break;
3433 		}
3434 		seqid_open = state->open_stateid.seqid;
3435 		if (read_seqretry(&state->seqlock, seq))
3436 			continue;
3437 
3438 		dst_seqid = be32_to_cpu(dst->seqid);
3439 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3440 			dst->seqid = seqid_open;
3441 		break;
3442 	}
3443 }
3444 
3445 /*
3446  * Update the seqid of an open stateid after receiving
3447  * NFS4ERR_OLD_STATEID
3448  */
3449 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3450 		struct nfs4_state *state)
3451 {
3452 	__be32 seqid_open;
3453 	u32 dst_seqid;
3454 	bool ret;
3455 	int seq, status = -EAGAIN;
3456 	DEFINE_WAIT(wait);
3457 
3458 	for (;;) {
3459 		ret = false;
3460 		if (!nfs4_valid_open_stateid(state))
3461 			break;
3462 		seq = read_seqbegin(&state->seqlock);
3463 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3464 			if (read_seqretry(&state->seqlock, seq))
3465 				continue;
3466 			break;
3467 		}
3468 
3469 		write_seqlock(&state->seqlock);
3470 		seqid_open = state->open_stateid.seqid;
3471 
3472 		dst_seqid = be32_to_cpu(dst->seqid);
3473 
3474 		/* Did another OPEN bump the state's seqid?  try again: */
3475 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3476 			dst->seqid = seqid_open;
3477 			write_sequnlock(&state->seqlock);
3478 			ret = true;
3479 			break;
3480 		}
3481 
3482 		/* server says we're behind but we haven't seen the update yet */
3483 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3484 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3485 		write_sequnlock(&state->seqlock);
3486 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3487 
3488 		if (fatal_signal_pending(current))
3489 			status = -EINTR;
3490 		else
3491 			if (schedule_timeout(5*HZ) != 0)
3492 				status = 0;
3493 
3494 		finish_wait(&state->waitq, &wait);
3495 
3496 		if (!status)
3497 			continue;
3498 		if (status == -EINTR)
3499 			break;
3500 
3501 		/* we slept the whole 5 seconds, we must have lost a seqid */
3502 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3503 		ret = true;
3504 		break;
3505 	}
3506 
3507 	return ret;
3508 }
3509 
3510 struct nfs4_closedata {
3511 	struct inode *inode;
3512 	struct nfs4_state *state;
3513 	struct nfs_closeargs arg;
3514 	struct nfs_closeres res;
3515 	struct {
3516 		struct nfs4_layoutreturn_args arg;
3517 		struct nfs4_layoutreturn_res res;
3518 		struct nfs4_xdr_opaque_data ld_private;
3519 		u32 roc_barrier;
3520 		bool roc;
3521 	} lr;
3522 	struct nfs_fattr fattr;
3523 	unsigned long timestamp;
3524 };
3525 
3526 static void nfs4_free_closedata(void *data)
3527 {
3528 	struct nfs4_closedata *calldata = data;
3529 	struct nfs4_state_owner *sp = calldata->state->owner;
3530 	struct super_block *sb = calldata->state->inode->i_sb;
3531 
3532 	if (calldata->lr.roc)
3533 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3534 				calldata->res.lr_ret);
3535 	nfs4_put_open_state(calldata->state);
3536 	nfs_free_seqid(calldata->arg.seqid);
3537 	nfs4_put_state_owner(sp);
3538 	nfs_sb_deactive(sb);
3539 	kfree(calldata);
3540 }
3541 
3542 static void nfs4_close_done(struct rpc_task *task, void *data)
3543 {
3544 	struct nfs4_closedata *calldata = data;
3545 	struct nfs4_state *state = calldata->state;
3546 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3547 	nfs4_stateid *res_stateid = NULL;
3548 	struct nfs4_exception exception = {
3549 		.state = state,
3550 		.inode = calldata->inode,
3551 		.stateid = &calldata->arg.stateid,
3552 	};
3553 
3554 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3555 		return;
3556 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3557 
3558 	/* Handle Layoutreturn errors */
3559 	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3560 			  &calldata->res.lr_ret) == -EAGAIN)
3561 		goto out_restart;
3562 
3563 	/* hmm. we are done with the inode, and in the process of freeing
3564 	 * the state_owner. we keep this around to process errors
3565 	 */
3566 	switch (task->tk_status) {
3567 		case 0:
3568 			res_stateid = &calldata->res.stateid;
3569 			renew_lease(server, calldata->timestamp);
3570 			break;
3571 		case -NFS4ERR_ACCESS:
3572 			if (calldata->arg.bitmask != NULL) {
3573 				calldata->arg.bitmask = NULL;
3574 				calldata->res.fattr = NULL;
3575 				goto out_restart;
3576 
3577 			}
3578 			break;
3579 		case -NFS4ERR_OLD_STATEID:
3580 			/* Did we race with OPEN? */
3581 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3582 						state))
3583 				goto out_restart;
3584 			goto out_release;
3585 		case -NFS4ERR_ADMIN_REVOKED:
3586 		case -NFS4ERR_STALE_STATEID:
3587 		case -NFS4ERR_EXPIRED:
3588 			nfs4_free_revoked_stateid(server,
3589 					&calldata->arg.stateid,
3590 					task->tk_msg.rpc_cred);
3591 			fallthrough;
3592 		case -NFS4ERR_BAD_STATEID:
3593 			if (calldata->arg.fmode == 0)
3594 				break;
3595 			fallthrough;
3596 		default:
3597 			task->tk_status = nfs4_async_handle_exception(task,
3598 					server, task->tk_status, &exception);
3599 			if (exception.retry)
3600 				goto out_restart;
3601 	}
3602 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3603 			res_stateid, calldata->arg.fmode);
3604 out_release:
3605 	task->tk_status = 0;
3606 	nfs_release_seqid(calldata->arg.seqid);
3607 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3608 	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3609 	return;
3610 out_restart:
3611 	task->tk_status = 0;
3612 	rpc_restart_call_prepare(task);
3613 	goto out_release;
3614 }
3615 
3616 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3617 {
3618 	struct nfs4_closedata *calldata = data;
3619 	struct nfs4_state *state = calldata->state;
3620 	struct inode *inode = calldata->inode;
3621 	struct nfs_server *server = NFS_SERVER(inode);
3622 	struct pnfs_layout_hdr *lo;
3623 	bool is_rdonly, is_wronly, is_rdwr;
3624 	int call_close = 0;
3625 
3626 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3627 		goto out_wait;
3628 
3629 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3630 	spin_lock(&state->owner->so_lock);
3631 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3632 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3633 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3634 	/* Calculate the change in open mode */
3635 	calldata->arg.fmode = 0;
3636 	if (state->n_rdwr == 0) {
3637 		if (state->n_rdonly == 0)
3638 			call_close |= is_rdonly;
3639 		else if (is_rdonly)
3640 			calldata->arg.fmode |= FMODE_READ;
3641 		if (state->n_wronly == 0)
3642 			call_close |= is_wronly;
3643 		else if (is_wronly)
3644 			calldata->arg.fmode |= FMODE_WRITE;
3645 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3646 			call_close |= is_rdwr;
3647 	} else if (is_rdwr)
3648 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3649 
3650 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3651 	if (!nfs4_valid_open_stateid(state))
3652 		call_close = 0;
3653 	spin_unlock(&state->owner->so_lock);
3654 
3655 	if (!call_close) {
3656 		/* Note: exit _without_ calling nfs4_close_done */
3657 		goto out_no_action;
3658 	}
3659 
3660 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3661 		nfs_release_seqid(calldata->arg.seqid);
3662 		goto out_wait;
3663 	}
3664 
3665 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3666 	if (lo && !pnfs_layout_is_valid(lo)) {
3667 		calldata->arg.lr_args = NULL;
3668 		calldata->res.lr_res = NULL;
3669 	}
3670 
3671 	if (calldata->arg.fmode == 0)
3672 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3673 
3674 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3675 		/* Close-to-open cache consistency revalidation */
3676 		if (!nfs4_have_delegation(inode, FMODE_READ)) {
3677 			nfs4_bitmask_set(calldata->arg.bitmask_store,
3678 					 server->cache_consistency_bitmask,
3679 					 inode, 0);
3680 			calldata->arg.bitmask = calldata->arg.bitmask_store;
3681 		} else
3682 			calldata->arg.bitmask = NULL;
3683 	}
3684 
3685 	calldata->arg.share_access =
3686 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3687 				calldata->arg.fmode, 0);
3688 
3689 	if (calldata->res.fattr == NULL)
3690 		calldata->arg.bitmask = NULL;
3691 	else if (calldata->arg.bitmask == NULL)
3692 		calldata->res.fattr = NULL;
3693 	calldata->timestamp = jiffies;
3694 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3695 				&calldata->arg.seq_args,
3696 				&calldata->res.seq_res,
3697 				task) != 0)
3698 		nfs_release_seqid(calldata->arg.seqid);
3699 	return;
3700 out_no_action:
3701 	task->tk_action = NULL;
3702 out_wait:
3703 	nfs4_sequence_done(task, &calldata->res.seq_res);
3704 }
3705 
3706 static const struct rpc_call_ops nfs4_close_ops = {
3707 	.rpc_call_prepare = nfs4_close_prepare,
3708 	.rpc_call_done = nfs4_close_done,
3709 	.rpc_release = nfs4_free_closedata,
3710 };
3711 
3712 /*
3713  * It is possible for data to be read/written from a mem-mapped file
3714  * after the sys_close call (which hits the vfs layer as a flush).
3715  * This means that we can't safely call nfsv4 close on a file until
3716  * the inode is cleared. This in turn means that we are not good
3717  * NFSv4 citizens - we do not indicate to the server to update the file's
3718  * share state even when we are done with one of the three share
3719  * stateid's in the inode.
3720  *
3721  * NOTE: Caller must be holding the sp->so_owner semaphore!
3722  */
3723 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3724 {
3725 	struct nfs_server *server = NFS_SERVER(state->inode);
3726 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3727 	struct nfs4_closedata *calldata;
3728 	struct nfs4_state_owner *sp = state->owner;
3729 	struct rpc_task *task;
3730 	struct rpc_message msg = {
3731 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3732 		.rpc_cred = state->owner->so_cred,
3733 	};
3734 	struct rpc_task_setup task_setup_data = {
3735 		.rpc_client = server->client,
3736 		.rpc_message = &msg,
3737 		.callback_ops = &nfs4_close_ops,
3738 		.workqueue = nfsiod_workqueue,
3739 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3740 	};
3741 	int status = -ENOMEM;
3742 
3743 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3744 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3745 
3746 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3747 		&task_setup_data.rpc_client, &msg);
3748 
3749 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3750 	if (calldata == NULL)
3751 		goto out;
3752 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3753 	calldata->inode = state->inode;
3754 	calldata->state = state;
3755 	calldata->arg.fh = NFS_FH(state->inode);
3756 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3757 		goto out_free_calldata;
3758 	/* Serialization for the sequence id */
3759 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3760 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3761 	if (IS_ERR(calldata->arg.seqid))
3762 		goto out_free_calldata;
3763 	nfs_fattr_init(&calldata->fattr);
3764 	calldata->arg.fmode = 0;
3765 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3766 	calldata->res.fattr = &calldata->fattr;
3767 	calldata->res.seqid = calldata->arg.seqid;
3768 	calldata->res.server = server;
3769 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3770 	calldata->lr.roc = pnfs_roc(state->inode,
3771 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3772 	if (calldata->lr.roc) {
3773 		calldata->arg.lr_args = &calldata->lr.arg;
3774 		calldata->res.lr_res = &calldata->lr.res;
3775 	}
3776 	nfs_sb_active(calldata->inode->i_sb);
3777 
3778 	msg.rpc_argp = &calldata->arg;
3779 	msg.rpc_resp = &calldata->res;
3780 	task_setup_data.callback_data = calldata;
3781 	task = rpc_run_task(&task_setup_data);
3782 	if (IS_ERR(task))
3783 		return PTR_ERR(task);
3784 	status = 0;
3785 	if (wait)
3786 		status = rpc_wait_for_completion_task(task);
3787 	rpc_put_task(task);
3788 	return status;
3789 out_free_calldata:
3790 	kfree(calldata);
3791 out:
3792 	nfs4_put_open_state(state);
3793 	nfs4_put_state_owner(sp);
3794 	return status;
3795 }
3796 
3797 static struct inode *
3798 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3799 		int open_flags, struct iattr *attr, int *opened)
3800 {
3801 	struct nfs4_state *state;
3802 	struct nfs4_label l, *label;
3803 
3804 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3805 
3806 	/* Protect against concurrent sillydeletes */
3807 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3808 
3809 	nfs4_label_release_security(label);
3810 
3811 	if (IS_ERR(state))
3812 		return ERR_CAST(state);
3813 	return state->inode;
3814 }
3815 
3816 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3817 {
3818 	if (ctx->state == NULL)
3819 		return;
3820 	if (is_sync)
3821 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3822 	else
3823 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3824 }
3825 
3826 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3827 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3828 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3829 
3830 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3831 {
3832 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3833 	struct nfs4_server_caps_arg args = {
3834 		.fhandle = fhandle,
3835 		.bitmask = bitmask,
3836 	};
3837 	struct nfs4_server_caps_res res = {};
3838 	struct rpc_message msg = {
3839 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3840 		.rpc_argp = &args,
3841 		.rpc_resp = &res,
3842 	};
3843 	int status;
3844 	int i;
3845 
3846 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3847 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3848 		     FATTR4_WORD0_LINK_SUPPORT |
3849 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3850 		     FATTR4_WORD0_ACLSUPPORT |
3851 		     FATTR4_WORD0_CASE_INSENSITIVE |
3852 		     FATTR4_WORD0_CASE_PRESERVING;
3853 	if (minorversion)
3854 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3855 
3856 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3857 	if (status == 0) {
3858 		/* Sanity check the server answers */
3859 		switch (minorversion) {
3860 		case 0:
3861 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3862 			res.attr_bitmask[2] = 0;
3863 			break;
3864 		case 1:
3865 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3866 			break;
3867 		case 2:
3868 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3869 		}
3870 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3871 		server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3872 				  NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3873 		server->fattr_valid = NFS_ATTR_FATTR_V4;
3874 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3875 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3876 			server->caps |= NFS_CAP_ACLS;
3877 		if (res.has_links != 0)
3878 			server->caps |= NFS_CAP_HARDLINKS;
3879 		if (res.has_symlinks != 0)
3880 			server->caps |= NFS_CAP_SYMLINKS;
3881 		if (res.case_insensitive)
3882 			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3883 		if (res.case_preserving)
3884 			server->caps |= NFS_CAP_CASE_PRESERVING;
3885 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3886 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3887 			server->caps |= NFS_CAP_SECURITY_LABEL;
3888 #endif
3889 		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3890 			server->caps |= NFS_CAP_FS_LOCATIONS;
3891 		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3892 			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3893 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3894 			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3895 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3896 			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3897 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3898 			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3899 				NFS_ATTR_FATTR_OWNER_NAME);
3900 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3901 			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3902 				NFS_ATTR_FATTR_GROUP_NAME);
3903 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3904 			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3905 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3906 			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3907 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3908 			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3909 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3910 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3911 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3912 				sizeof(server->attr_bitmask));
3913 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3914 
3915 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3916 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3917 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3918 		server->cache_consistency_bitmask[2] = 0;
3919 
3920 		/* Avoid a regression due to buggy server */
3921 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3922 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3923 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3924 			sizeof(server->exclcreat_bitmask));
3925 
3926 		server->acl_bitmask = res.acl_bitmask;
3927 		server->fh_expire_type = res.fh_expire_type;
3928 	}
3929 
3930 	return status;
3931 }
3932 
3933 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3934 {
3935 	struct nfs4_exception exception = {
3936 		.interruptible = true,
3937 	};
3938 	int err;
3939 
3940 	nfs4_server_set_init_caps(server);
3941 	do {
3942 		err = nfs4_handle_exception(server,
3943 				_nfs4_server_capabilities(server, fhandle),
3944 				&exception);
3945 	} while (exception.retry);
3946 	return err;
3947 }
3948 
3949 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3950 					  struct nfs_client *clp,
3951 					  struct nfs_server *server)
3952 {
3953 	int i;
3954 
3955 	for (i = 0; i < location->nservers; i++) {
3956 		struct nfs4_string *srv_loc = &location->servers[i];
3957 		struct sockaddr_storage addr;
3958 		size_t addrlen;
3959 		struct xprt_create xprt_args = {
3960 			.ident = 0,
3961 			.net = clp->cl_net,
3962 		};
3963 		struct nfs4_add_xprt_data xprtdata = {
3964 			.clp = clp,
3965 		};
3966 		struct rpc_add_xprt_test rpcdata = {
3967 			.add_xprt_test = clp->cl_mvops->session_trunk,
3968 			.data = &xprtdata,
3969 		};
3970 		char *servername = NULL;
3971 
3972 		if (!srv_loc->len)
3973 			continue;
3974 
3975 		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3976 						&addr, sizeof(addr),
3977 						clp->cl_net, server->port);
3978 		if (!addrlen)
3979 			return;
3980 		xprt_args.dstaddr = (struct sockaddr *)&addr;
3981 		xprt_args.addrlen = addrlen;
3982 		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3983 		if (!servername)
3984 			return;
3985 		memcpy(servername, srv_loc->data, srv_loc->len);
3986 		servername[srv_loc->len] = '\0';
3987 		xprt_args.servername = servername;
3988 
3989 		xprtdata.cred = nfs4_get_clid_cred(clp);
3990 		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3991 				  rpc_clnt_setup_test_and_add_xprt,
3992 				  &rpcdata);
3993 		if (xprtdata.cred)
3994 			put_cred(xprtdata.cred);
3995 		kfree(servername);
3996 	}
3997 }
3998 
3999 static int _nfs4_discover_trunking(struct nfs_server *server,
4000 				   struct nfs_fh *fhandle)
4001 {
4002 	struct nfs4_fs_locations *locations = NULL;
4003 	struct page *page;
4004 	const struct cred *cred;
4005 	struct nfs_client *clp = server->nfs_client;
4006 	const struct nfs4_state_maintenance_ops *ops =
4007 		clp->cl_mvops->state_renewal_ops;
4008 	int status = -ENOMEM, i;
4009 
4010 	cred = ops->get_state_renewal_cred(clp);
4011 	if (cred == NULL) {
4012 		cred = nfs4_get_clid_cred(clp);
4013 		if (cred == NULL)
4014 			return -ENOKEY;
4015 	}
4016 
4017 	page = alloc_page(GFP_KERNEL);
4018 	if (!page)
4019 		goto out_put_cred;
4020 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4021 	if (!locations)
4022 		goto out_free;
4023 	locations->fattr = nfs_alloc_fattr();
4024 	if (!locations->fattr)
4025 		goto out_free_2;
4026 
4027 	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4028 					 cred);
4029 	if (status)
4030 		goto out_free_3;
4031 
4032 	for (i = 0; i < locations->nlocations; i++)
4033 		test_fs_location_for_trunking(&locations->locations[i], clp,
4034 					      server);
4035 out_free_3:
4036 	kfree(locations->fattr);
4037 out_free_2:
4038 	kfree(locations);
4039 out_free:
4040 	__free_page(page);
4041 out_put_cred:
4042 	put_cred(cred);
4043 	return status;
4044 }
4045 
4046 static int nfs4_discover_trunking(struct nfs_server *server,
4047 				  struct nfs_fh *fhandle)
4048 {
4049 	struct nfs4_exception exception = {
4050 		.interruptible = true,
4051 	};
4052 	struct nfs_client *clp = server->nfs_client;
4053 	int err = 0;
4054 
4055 	if (!nfs4_has_session(clp))
4056 		goto out;
4057 	do {
4058 		err = nfs4_handle_exception(server,
4059 				_nfs4_discover_trunking(server, fhandle),
4060 				&exception);
4061 	} while (exception.retry);
4062 out:
4063 	return err;
4064 }
4065 
4066 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4067 		struct nfs_fsinfo *info)
4068 {
4069 	u32 bitmask[3];
4070 	struct nfs4_lookup_root_arg args = {
4071 		.bitmask = bitmask,
4072 	};
4073 	struct nfs4_lookup_res res = {
4074 		.server = server,
4075 		.fattr = info->fattr,
4076 		.fh = fhandle,
4077 	};
4078 	struct rpc_message msg = {
4079 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4080 		.rpc_argp = &args,
4081 		.rpc_resp = &res,
4082 	};
4083 
4084 	bitmask[0] = nfs4_fattr_bitmap[0];
4085 	bitmask[1] = nfs4_fattr_bitmap[1];
4086 	/*
4087 	 * Process the label in the upcoming getfattr
4088 	 */
4089 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4090 
4091 	nfs_fattr_init(info->fattr);
4092 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4093 }
4094 
4095 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4096 		struct nfs_fsinfo *info)
4097 {
4098 	struct nfs4_exception exception = {
4099 		.interruptible = true,
4100 	};
4101 	int err;
4102 	do {
4103 		err = _nfs4_lookup_root(server, fhandle, info);
4104 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4105 		switch (err) {
4106 		case 0:
4107 		case -NFS4ERR_WRONGSEC:
4108 			goto out;
4109 		default:
4110 			err = nfs4_handle_exception(server, err, &exception);
4111 		}
4112 	} while (exception.retry);
4113 out:
4114 	return err;
4115 }
4116 
4117 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4118 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4119 {
4120 	struct rpc_auth_create_args auth_args = {
4121 		.pseudoflavor = flavor,
4122 	};
4123 	struct rpc_auth *auth;
4124 
4125 	auth = rpcauth_create(&auth_args, server->client);
4126 	if (IS_ERR(auth))
4127 		return -EACCES;
4128 	return nfs4_lookup_root(server, fhandle, info);
4129 }
4130 
4131 /*
4132  * Retry pseudoroot lookup with various security flavors.  We do this when:
4133  *
4134  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4135  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4136  *
4137  * Returns zero on success, or a negative NFS4ERR value, or a
4138  * negative errno value.
4139  */
4140 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4141 			      struct nfs_fsinfo *info)
4142 {
4143 	/* Per 3530bis 15.33.5 */
4144 	static const rpc_authflavor_t flav_array[] = {
4145 		RPC_AUTH_GSS_KRB5P,
4146 		RPC_AUTH_GSS_KRB5I,
4147 		RPC_AUTH_GSS_KRB5,
4148 		RPC_AUTH_UNIX,			/* courtesy */
4149 		RPC_AUTH_NULL,
4150 	};
4151 	int status = -EPERM;
4152 	size_t i;
4153 
4154 	if (server->auth_info.flavor_len > 0) {
4155 		/* try each flavor specified by user */
4156 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4157 			status = nfs4_lookup_root_sec(server, fhandle, info,
4158 						server->auth_info.flavors[i]);
4159 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4160 				continue;
4161 			break;
4162 		}
4163 	} else {
4164 		/* no flavors specified by user, try default list */
4165 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4166 			status = nfs4_lookup_root_sec(server, fhandle, info,
4167 						      flav_array[i]);
4168 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4169 				continue;
4170 			break;
4171 		}
4172 	}
4173 
4174 	/*
4175 	 * -EACCES could mean that the user doesn't have correct permissions
4176 	 * to access the mount.  It could also mean that we tried to mount
4177 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4178 	 * existing mount programs don't handle -EACCES very well so it should
4179 	 * be mapped to -EPERM instead.
4180 	 */
4181 	if (status == -EACCES)
4182 		status = -EPERM;
4183 	return status;
4184 }
4185 
4186 /**
4187  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4188  * @server: initialized nfs_server handle
4189  * @fhandle: we fill in the pseudo-fs root file handle
4190  * @info: we fill in an FSINFO struct
4191  * @auth_probe: probe the auth flavours
4192  *
4193  * Returns zero on success, or a negative errno.
4194  */
4195 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4196 			 struct nfs_fsinfo *info,
4197 			 bool auth_probe)
4198 {
4199 	int status = 0;
4200 
4201 	if (!auth_probe)
4202 		status = nfs4_lookup_root(server, fhandle, info);
4203 
4204 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4205 		status = server->nfs_client->cl_mvops->find_root_sec(server,
4206 				fhandle, info);
4207 
4208 	if (status == 0)
4209 		status = nfs4_server_capabilities(server, fhandle);
4210 	if (status == 0)
4211 		status = nfs4_do_fsinfo(server, fhandle, info);
4212 
4213 	return nfs4_map_errors(status);
4214 }
4215 
4216 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4217 			      struct nfs_fsinfo *info)
4218 {
4219 	int error;
4220 	struct nfs_fattr *fattr = info->fattr;
4221 
4222 	error = nfs4_server_capabilities(server, mntfh);
4223 	if (error < 0) {
4224 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4225 		return error;
4226 	}
4227 
4228 	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4229 	if (error < 0) {
4230 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4231 		goto out;
4232 	}
4233 
4234 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4235 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4236 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4237 
4238 out:
4239 	return error;
4240 }
4241 
4242 /*
4243  * Get locations and (maybe) other attributes of a referral.
4244  * Note that we'll actually follow the referral later when
4245  * we detect fsid mismatch in inode revalidation
4246  */
4247 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4248 			     const struct qstr *name, struct nfs_fattr *fattr,
4249 			     struct nfs_fh *fhandle)
4250 {
4251 	int status = -ENOMEM;
4252 	struct page *page = NULL;
4253 	struct nfs4_fs_locations *locations = NULL;
4254 
4255 	page = alloc_page(GFP_KERNEL);
4256 	if (page == NULL)
4257 		goto out;
4258 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4259 	if (locations == NULL)
4260 		goto out;
4261 
4262 	locations->fattr = fattr;
4263 
4264 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4265 	if (status != 0)
4266 		goto out;
4267 
4268 	/*
4269 	 * If the fsid didn't change, this is a migration event, not a
4270 	 * referral.  Cause us to drop into the exception handler, which
4271 	 * will kick off migration recovery.
4272 	 */
4273 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4274 		dprintk("%s: server did not return a different fsid for"
4275 			" a referral at %s\n", __func__, name->name);
4276 		status = -NFS4ERR_MOVED;
4277 		goto out;
4278 	}
4279 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4280 	nfs_fixup_referral_attributes(fattr);
4281 	memset(fhandle, 0, sizeof(struct nfs_fh));
4282 out:
4283 	if (page)
4284 		__free_page(page);
4285 	kfree(locations);
4286 	return status;
4287 }
4288 
4289 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4290 				struct nfs_fattr *fattr, struct inode *inode)
4291 {
4292 	__u32 bitmask[NFS4_BITMASK_SZ];
4293 	struct nfs4_getattr_arg args = {
4294 		.fh = fhandle,
4295 		.bitmask = bitmask,
4296 	};
4297 	struct nfs4_getattr_res res = {
4298 		.fattr = fattr,
4299 		.server = server,
4300 	};
4301 	struct rpc_message msg = {
4302 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4303 		.rpc_argp = &args,
4304 		.rpc_resp = &res,
4305 	};
4306 	unsigned short task_flags = 0;
4307 
4308 	if (nfs4_has_session(server->nfs_client))
4309 		task_flags = RPC_TASK_MOVEABLE;
4310 
4311 	/* Is this is an attribute revalidation, subject to softreval? */
4312 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4313 		task_flags |= RPC_TASK_TIMEOUT;
4314 
4315 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4316 	nfs_fattr_init(fattr);
4317 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4318 	return nfs4_do_call_sync(server->client, server, &msg,
4319 			&args.seq_args, &res.seq_res, task_flags);
4320 }
4321 
4322 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4323 				struct nfs_fattr *fattr, struct inode *inode)
4324 {
4325 	struct nfs4_exception exception = {
4326 		.interruptible = true,
4327 	};
4328 	int err;
4329 	do {
4330 		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4331 		trace_nfs4_getattr(server, fhandle, fattr, err);
4332 		err = nfs4_handle_exception(server, err,
4333 				&exception);
4334 	} while (exception.retry);
4335 	return err;
4336 }
4337 
4338 /*
4339  * The file is not closed if it is opened due to the a request to change
4340  * the size of the file. The open call will not be needed once the
4341  * VFS layer lookup-intents are implemented.
4342  *
4343  * Close is called when the inode is destroyed.
4344  * If we haven't opened the file for O_WRONLY, we
4345  * need to in the size_change case to obtain a stateid.
4346  *
4347  * Got race?
4348  * Because OPEN is always done by name in nfsv4, it is
4349  * possible that we opened a different file by the same
4350  * name.  We can recognize this race condition, but we
4351  * can't do anything about it besides returning an error.
4352  *
4353  * This will be fixed with VFS changes (lookup-intent).
4354  */
4355 static int
4356 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4357 		  struct iattr *sattr)
4358 {
4359 	struct inode *inode = d_inode(dentry);
4360 	const struct cred *cred = NULL;
4361 	struct nfs_open_context *ctx = NULL;
4362 	int status;
4363 
4364 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4365 	    sattr->ia_valid & ATTR_SIZE &&
4366 	    sattr->ia_size < i_size_read(inode))
4367 		pnfs_commit_and_return_layout(inode);
4368 
4369 	nfs_fattr_init(fattr);
4370 
4371 	/* Deal with open(O_TRUNC) */
4372 	if (sattr->ia_valid & ATTR_OPEN)
4373 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4374 
4375 	/* Optimization: if the end result is no change, don't RPC */
4376 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4377 		return 0;
4378 
4379 	/* Search for an existing open(O_WRITE) file */
4380 	if (sattr->ia_valid & ATTR_FILE) {
4381 
4382 		ctx = nfs_file_open_context(sattr->ia_file);
4383 		if (ctx)
4384 			cred = ctx->cred;
4385 	}
4386 
4387 	/* Return any delegations if we're going to change ACLs */
4388 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4389 		nfs4_inode_make_writeable(inode);
4390 
4391 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4392 	if (status == 0) {
4393 		nfs_setattr_update_inode(inode, sattr, fattr);
4394 		nfs_setsecurity(inode, fattr);
4395 	}
4396 	return status;
4397 }
4398 
4399 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4400 		struct dentry *dentry, struct nfs_fh *fhandle,
4401 		struct nfs_fattr *fattr)
4402 {
4403 	struct nfs_server *server = NFS_SERVER(dir);
4404 	int		       status;
4405 	struct nfs4_lookup_arg args = {
4406 		.bitmask = server->attr_bitmask,
4407 		.dir_fh = NFS_FH(dir),
4408 		.name = &dentry->d_name,
4409 	};
4410 	struct nfs4_lookup_res res = {
4411 		.server = server,
4412 		.fattr = fattr,
4413 		.fh = fhandle,
4414 	};
4415 	struct rpc_message msg = {
4416 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4417 		.rpc_argp = &args,
4418 		.rpc_resp = &res,
4419 	};
4420 	unsigned short task_flags = 0;
4421 
4422 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4423 		task_flags = RPC_TASK_MOVEABLE;
4424 
4425 	/* Is this is an attribute revalidation, subject to softreval? */
4426 	if (nfs_lookup_is_soft_revalidate(dentry))
4427 		task_flags |= RPC_TASK_TIMEOUT;
4428 
4429 	args.bitmask = nfs4_bitmask(server, fattr->label);
4430 
4431 	nfs_fattr_init(fattr);
4432 
4433 	dprintk("NFS call  lookup %pd2\n", dentry);
4434 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4435 	status = nfs4_do_call_sync(clnt, server, &msg,
4436 			&args.seq_args, &res.seq_res, task_flags);
4437 	dprintk("NFS reply lookup: %d\n", status);
4438 	return status;
4439 }
4440 
4441 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4442 {
4443 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4444 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4445 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4446 	fattr->nlink = 2;
4447 }
4448 
4449 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4450 				   struct dentry *dentry, struct nfs_fh *fhandle,
4451 				   struct nfs_fattr *fattr)
4452 {
4453 	struct nfs4_exception exception = {
4454 		.interruptible = true,
4455 	};
4456 	struct rpc_clnt *client = *clnt;
4457 	const struct qstr *name = &dentry->d_name;
4458 	int err;
4459 	do {
4460 		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4461 		trace_nfs4_lookup(dir, name, err);
4462 		switch (err) {
4463 		case -NFS4ERR_BADNAME:
4464 			err = -ENOENT;
4465 			goto out;
4466 		case -NFS4ERR_MOVED:
4467 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4468 			if (err == -NFS4ERR_MOVED)
4469 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4470 			goto out;
4471 		case -NFS4ERR_WRONGSEC:
4472 			err = -EPERM;
4473 			if (client != *clnt)
4474 				goto out;
4475 			client = nfs4_negotiate_security(client, dir, name);
4476 			if (IS_ERR(client))
4477 				return PTR_ERR(client);
4478 
4479 			exception.retry = 1;
4480 			break;
4481 		default:
4482 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4483 		}
4484 	} while (exception.retry);
4485 
4486 out:
4487 	if (err == 0)
4488 		*clnt = client;
4489 	else if (client != *clnt)
4490 		rpc_shutdown_client(client);
4491 
4492 	return err;
4493 }
4494 
4495 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4496 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4497 {
4498 	int status;
4499 	struct rpc_clnt *client = NFS_CLIENT(dir);
4500 
4501 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4502 	if (client != NFS_CLIENT(dir)) {
4503 		rpc_shutdown_client(client);
4504 		nfs_fixup_secinfo_attributes(fattr);
4505 	}
4506 	return status;
4507 }
4508 
4509 struct rpc_clnt *
4510 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4511 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4512 {
4513 	struct rpc_clnt *client = NFS_CLIENT(dir);
4514 	int status;
4515 
4516 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4517 	if (status < 0)
4518 		return ERR_PTR(status);
4519 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4520 }
4521 
4522 static int _nfs4_proc_lookupp(struct inode *inode,
4523 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4524 {
4525 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4526 	struct nfs_server *server = NFS_SERVER(inode);
4527 	int		       status;
4528 	struct nfs4_lookupp_arg args = {
4529 		.bitmask = server->attr_bitmask,
4530 		.fh = NFS_FH(inode),
4531 	};
4532 	struct nfs4_lookupp_res res = {
4533 		.server = server,
4534 		.fattr = fattr,
4535 		.fh = fhandle,
4536 	};
4537 	struct rpc_message msg = {
4538 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4539 		.rpc_argp = &args,
4540 		.rpc_resp = &res,
4541 	};
4542 	unsigned short task_flags = 0;
4543 
4544 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4545 		task_flags |= RPC_TASK_TIMEOUT;
4546 
4547 	args.bitmask = nfs4_bitmask(server, fattr->label);
4548 
4549 	nfs_fattr_init(fattr);
4550 
4551 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4552 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4553 				&res.seq_res, task_flags);
4554 	dprintk("NFS reply lookupp: %d\n", status);
4555 	return status;
4556 }
4557 
4558 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4559 			     struct nfs_fattr *fattr)
4560 {
4561 	struct nfs4_exception exception = {
4562 		.interruptible = true,
4563 	};
4564 	int err;
4565 	do {
4566 		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4567 		trace_nfs4_lookupp(inode, err);
4568 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4569 				&exception);
4570 	} while (exception.retry);
4571 	return err;
4572 }
4573 
4574 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4575 			     const struct cred *cred)
4576 {
4577 	struct nfs_server *server = NFS_SERVER(inode);
4578 	struct nfs4_accessargs args = {
4579 		.fh = NFS_FH(inode),
4580 		.access = entry->mask,
4581 	};
4582 	struct nfs4_accessres res = {
4583 		.server = server,
4584 	};
4585 	struct rpc_message msg = {
4586 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4587 		.rpc_argp = &args,
4588 		.rpc_resp = &res,
4589 		.rpc_cred = cred,
4590 	};
4591 	int status = 0;
4592 
4593 	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4594 		res.fattr = nfs_alloc_fattr();
4595 		if (res.fattr == NULL)
4596 			return -ENOMEM;
4597 		args.bitmask = server->cache_consistency_bitmask;
4598 	}
4599 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4600 	if (!status) {
4601 		nfs_access_set_mask(entry, res.access);
4602 		if (res.fattr)
4603 			nfs_refresh_inode(inode, res.fattr);
4604 	}
4605 	nfs_free_fattr(res.fattr);
4606 	return status;
4607 }
4608 
4609 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4610 			    const struct cred *cred)
4611 {
4612 	struct nfs4_exception exception = {
4613 		.interruptible = true,
4614 	};
4615 	int err;
4616 	do {
4617 		err = _nfs4_proc_access(inode, entry, cred);
4618 		trace_nfs4_access(inode, err);
4619 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4620 				&exception);
4621 	} while (exception.retry);
4622 	return err;
4623 }
4624 
4625 /*
4626  * TODO: For the time being, we don't try to get any attributes
4627  * along with any of the zero-copy operations READ, READDIR,
4628  * READLINK, WRITE.
4629  *
4630  * In the case of the first three, we want to put the GETATTR
4631  * after the read-type operation -- this is because it is hard
4632  * to predict the length of a GETATTR response in v4, and thus
4633  * align the READ data correctly.  This means that the GETATTR
4634  * may end up partially falling into the page cache, and we should
4635  * shift it into the 'tail' of the xdr_buf before processing.
4636  * To do this efficiently, we need to know the total length
4637  * of data received, which doesn't seem to be available outside
4638  * of the RPC layer.
4639  *
4640  * In the case of WRITE, we also want to put the GETATTR after
4641  * the operation -- in this case because we want to make sure
4642  * we get the post-operation mtime and size.
4643  *
4644  * Both of these changes to the XDR layer would in fact be quite
4645  * minor, but I decided to leave them for a subsequent patch.
4646  */
4647 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4648 		unsigned int pgbase, unsigned int pglen)
4649 {
4650 	struct nfs4_readlink args = {
4651 		.fh       = NFS_FH(inode),
4652 		.pgbase	  = pgbase,
4653 		.pglen    = pglen,
4654 		.pages    = &page,
4655 	};
4656 	struct nfs4_readlink_res res;
4657 	struct rpc_message msg = {
4658 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4659 		.rpc_argp = &args,
4660 		.rpc_resp = &res,
4661 	};
4662 
4663 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4664 }
4665 
4666 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4667 		unsigned int pgbase, unsigned int pglen)
4668 {
4669 	struct nfs4_exception exception = {
4670 		.interruptible = true,
4671 	};
4672 	int err;
4673 	do {
4674 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4675 		trace_nfs4_readlink(inode, err);
4676 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4677 				&exception);
4678 	} while (exception.retry);
4679 	return err;
4680 }
4681 
4682 /*
4683  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4684  */
4685 static int
4686 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4687 		 int flags)
4688 {
4689 	struct nfs_server *server = NFS_SERVER(dir);
4690 	struct nfs4_label l, *ilabel;
4691 	struct nfs_open_context *ctx;
4692 	struct nfs4_state *state;
4693 	int status = 0;
4694 
4695 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4696 	if (IS_ERR(ctx))
4697 		return PTR_ERR(ctx);
4698 
4699 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4700 
4701 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4702 		sattr->ia_mode &= ~current_umask();
4703 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4704 	if (IS_ERR(state)) {
4705 		status = PTR_ERR(state);
4706 		goto out;
4707 	}
4708 out:
4709 	nfs4_label_release_security(ilabel);
4710 	put_nfs_open_context(ctx);
4711 	return status;
4712 }
4713 
4714 static int
4715 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4716 {
4717 	struct nfs_server *server = NFS_SERVER(dir);
4718 	struct nfs_removeargs args = {
4719 		.fh = NFS_FH(dir),
4720 		.name = *name,
4721 	};
4722 	struct nfs_removeres res = {
4723 		.server = server,
4724 	};
4725 	struct rpc_message msg = {
4726 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4727 		.rpc_argp = &args,
4728 		.rpc_resp = &res,
4729 	};
4730 	unsigned long timestamp = jiffies;
4731 	int status;
4732 
4733 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4734 	if (status == 0) {
4735 		spin_lock(&dir->i_lock);
4736 		/* Removing a directory decrements nlink in the parent */
4737 		if (ftype == NF4DIR && dir->i_nlink > 2)
4738 			nfs4_dec_nlink_locked(dir);
4739 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4740 					      NFS_INO_INVALID_DATA);
4741 		spin_unlock(&dir->i_lock);
4742 	}
4743 	return status;
4744 }
4745 
4746 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4747 {
4748 	struct nfs4_exception exception = {
4749 		.interruptible = true,
4750 	};
4751 	struct inode *inode = d_inode(dentry);
4752 	int err;
4753 
4754 	if (inode) {
4755 		if (inode->i_nlink == 1)
4756 			nfs4_inode_return_delegation(inode);
4757 		else
4758 			nfs4_inode_make_writeable(inode);
4759 	}
4760 	do {
4761 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4762 		trace_nfs4_remove(dir, &dentry->d_name, err);
4763 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4764 				&exception);
4765 	} while (exception.retry);
4766 	return err;
4767 }
4768 
4769 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4770 {
4771 	struct nfs4_exception exception = {
4772 		.interruptible = true,
4773 	};
4774 	int err;
4775 
4776 	do {
4777 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4778 		trace_nfs4_remove(dir, name, err);
4779 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4780 				&exception);
4781 	} while (exception.retry);
4782 	return err;
4783 }
4784 
4785 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4786 		struct dentry *dentry,
4787 		struct inode *inode)
4788 {
4789 	struct nfs_removeargs *args = msg->rpc_argp;
4790 	struct nfs_removeres *res = msg->rpc_resp;
4791 
4792 	res->server = NFS_SB(dentry->d_sb);
4793 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4794 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4795 
4796 	nfs_fattr_init(res->dir_attr);
4797 
4798 	if (inode) {
4799 		nfs4_inode_return_delegation(inode);
4800 		nfs_d_prune_case_insensitive_aliases(inode);
4801 	}
4802 }
4803 
4804 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4805 {
4806 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4807 			&data->args.seq_args,
4808 			&data->res.seq_res,
4809 			task);
4810 }
4811 
4812 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4813 {
4814 	struct nfs_unlinkdata *data = task->tk_calldata;
4815 	struct nfs_removeres *res = &data->res;
4816 
4817 	if (!nfs4_sequence_done(task, &res->seq_res))
4818 		return 0;
4819 	if (nfs4_async_handle_error(task, res->server, NULL,
4820 				    &data->timeout) == -EAGAIN)
4821 		return 0;
4822 	if (task->tk_status == 0)
4823 		nfs4_update_changeattr(dir, &res->cinfo,
4824 				res->dir_attr->time_start,
4825 				NFS_INO_INVALID_DATA);
4826 	return 1;
4827 }
4828 
4829 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4830 		struct dentry *old_dentry,
4831 		struct dentry *new_dentry)
4832 {
4833 	struct nfs_renameargs *arg = msg->rpc_argp;
4834 	struct nfs_renameres *res = msg->rpc_resp;
4835 	struct inode *old_inode = d_inode(old_dentry);
4836 	struct inode *new_inode = d_inode(new_dentry);
4837 
4838 	if (old_inode)
4839 		nfs4_inode_make_writeable(old_inode);
4840 	if (new_inode)
4841 		nfs4_inode_return_delegation(new_inode);
4842 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4843 	res->server = NFS_SB(old_dentry->d_sb);
4844 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4845 }
4846 
4847 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4848 {
4849 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4850 			&data->args.seq_args,
4851 			&data->res.seq_res,
4852 			task);
4853 }
4854 
4855 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4856 				 struct inode *new_dir)
4857 {
4858 	struct nfs_renamedata *data = task->tk_calldata;
4859 	struct nfs_renameres *res = &data->res;
4860 
4861 	if (!nfs4_sequence_done(task, &res->seq_res))
4862 		return 0;
4863 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4864 		return 0;
4865 
4866 	if (task->tk_status == 0) {
4867 		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4868 		if (new_dir != old_dir) {
4869 			/* Note: If we moved a directory, nlink will change */
4870 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4871 					res->old_fattr->time_start,
4872 					NFS_INO_INVALID_NLINK |
4873 					    NFS_INO_INVALID_DATA);
4874 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4875 					res->new_fattr->time_start,
4876 					NFS_INO_INVALID_NLINK |
4877 					    NFS_INO_INVALID_DATA);
4878 		} else
4879 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4880 					res->old_fattr->time_start,
4881 					NFS_INO_INVALID_DATA);
4882 	}
4883 	return 1;
4884 }
4885 
4886 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4887 {
4888 	struct nfs_server *server = NFS_SERVER(inode);
4889 	__u32 bitmask[NFS4_BITMASK_SZ];
4890 	struct nfs4_link_arg arg = {
4891 		.fh     = NFS_FH(inode),
4892 		.dir_fh = NFS_FH(dir),
4893 		.name   = name,
4894 		.bitmask = bitmask,
4895 	};
4896 	struct nfs4_link_res res = {
4897 		.server = server,
4898 	};
4899 	struct rpc_message msg = {
4900 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4901 		.rpc_argp = &arg,
4902 		.rpc_resp = &res,
4903 	};
4904 	int status = -ENOMEM;
4905 
4906 	res.fattr = nfs_alloc_fattr_with_label(server);
4907 	if (res.fattr == NULL)
4908 		goto out;
4909 
4910 	nfs4_inode_make_writeable(inode);
4911 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4912 				NFS_INO_INVALID_CHANGE);
4913 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4914 	if (!status) {
4915 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4916 				       NFS_INO_INVALID_DATA);
4917 		nfs4_inc_nlink(inode);
4918 		status = nfs_post_op_update_inode(inode, res.fattr);
4919 		if (!status)
4920 			nfs_setsecurity(inode, res.fattr);
4921 	}
4922 
4923 out:
4924 	nfs_free_fattr(res.fattr);
4925 	return status;
4926 }
4927 
4928 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4929 {
4930 	struct nfs4_exception exception = {
4931 		.interruptible = true,
4932 	};
4933 	int err;
4934 	do {
4935 		err = nfs4_handle_exception(NFS_SERVER(inode),
4936 				_nfs4_proc_link(inode, dir, name),
4937 				&exception);
4938 	} while (exception.retry);
4939 	return err;
4940 }
4941 
4942 struct nfs4_createdata {
4943 	struct rpc_message msg;
4944 	struct nfs4_create_arg arg;
4945 	struct nfs4_create_res res;
4946 	struct nfs_fh fh;
4947 	struct nfs_fattr fattr;
4948 };
4949 
4950 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4951 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4952 {
4953 	struct nfs4_createdata *data;
4954 
4955 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4956 	if (data != NULL) {
4957 		struct nfs_server *server = NFS_SERVER(dir);
4958 
4959 		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4960 		if (IS_ERR(data->fattr.label))
4961 			goto out_free;
4962 
4963 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4964 		data->msg.rpc_argp = &data->arg;
4965 		data->msg.rpc_resp = &data->res;
4966 		data->arg.dir_fh = NFS_FH(dir);
4967 		data->arg.server = server;
4968 		data->arg.name = name;
4969 		data->arg.attrs = sattr;
4970 		data->arg.ftype = ftype;
4971 		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4972 		data->arg.umask = current_umask();
4973 		data->res.server = server;
4974 		data->res.fh = &data->fh;
4975 		data->res.fattr = &data->fattr;
4976 		nfs_fattr_init(data->res.fattr);
4977 	}
4978 	return data;
4979 out_free:
4980 	kfree(data);
4981 	return NULL;
4982 }
4983 
4984 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4985 {
4986 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4987 				    &data->arg.seq_args, &data->res.seq_res, 1);
4988 	if (status == 0) {
4989 		spin_lock(&dir->i_lock);
4990 		/* Creating a directory bumps nlink in the parent */
4991 		if (data->arg.ftype == NF4DIR)
4992 			nfs4_inc_nlink_locked(dir);
4993 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4994 					      data->res.fattr->time_start,
4995 					      NFS_INO_INVALID_DATA);
4996 		spin_unlock(&dir->i_lock);
4997 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4998 	}
4999 	return status;
5000 }
5001 
5002 static void nfs4_free_createdata(struct nfs4_createdata *data)
5003 {
5004 	nfs4_label_free(data->fattr.label);
5005 	kfree(data);
5006 }
5007 
5008 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5009 		struct page *page, unsigned int len, struct iattr *sattr,
5010 		struct nfs4_label *label)
5011 {
5012 	struct nfs4_createdata *data;
5013 	int status = -ENAMETOOLONG;
5014 
5015 	if (len > NFS4_MAXPATHLEN)
5016 		goto out;
5017 
5018 	status = -ENOMEM;
5019 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5020 	if (data == NULL)
5021 		goto out;
5022 
5023 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5024 	data->arg.u.symlink.pages = &page;
5025 	data->arg.u.symlink.len = len;
5026 	data->arg.label = label;
5027 
5028 	status = nfs4_do_create(dir, dentry, data);
5029 
5030 	nfs4_free_createdata(data);
5031 out:
5032 	return status;
5033 }
5034 
5035 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5036 		struct page *page, unsigned int len, struct iattr *sattr)
5037 {
5038 	struct nfs4_exception exception = {
5039 		.interruptible = true,
5040 	};
5041 	struct nfs4_label l, *label;
5042 	int err;
5043 
5044 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5045 
5046 	do {
5047 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5048 		trace_nfs4_symlink(dir, &dentry->d_name, err);
5049 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5050 				&exception);
5051 	} while (exception.retry);
5052 
5053 	nfs4_label_release_security(label);
5054 	return err;
5055 }
5056 
5057 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5058 		struct iattr *sattr, struct nfs4_label *label)
5059 {
5060 	struct nfs4_createdata *data;
5061 	int status = -ENOMEM;
5062 
5063 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5064 	if (data == NULL)
5065 		goto out;
5066 
5067 	data->arg.label = label;
5068 	status = nfs4_do_create(dir, dentry, data);
5069 
5070 	nfs4_free_createdata(data);
5071 out:
5072 	return status;
5073 }
5074 
5075 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5076 		struct iattr *sattr)
5077 {
5078 	struct nfs_server *server = NFS_SERVER(dir);
5079 	struct nfs4_exception exception = {
5080 		.interruptible = true,
5081 	};
5082 	struct nfs4_label l, *label;
5083 	int err;
5084 
5085 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5086 
5087 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5088 		sattr->ia_mode &= ~current_umask();
5089 	do {
5090 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5091 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5092 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5093 				&exception);
5094 	} while (exception.retry);
5095 	nfs4_label_release_security(label);
5096 
5097 	return err;
5098 }
5099 
5100 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5101 			      struct nfs_readdir_res *nr_res)
5102 {
5103 	struct inode		*dir = d_inode(nr_arg->dentry);
5104 	struct nfs_server	*server = NFS_SERVER(dir);
5105 	struct nfs4_readdir_arg args = {
5106 		.fh = NFS_FH(dir),
5107 		.pages = nr_arg->pages,
5108 		.pgbase = 0,
5109 		.count = nr_arg->page_len,
5110 		.plus = nr_arg->plus,
5111 	};
5112 	struct nfs4_readdir_res res;
5113 	struct rpc_message msg = {
5114 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5115 		.rpc_argp = &args,
5116 		.rpc_resp = &res,
5117 		.rpc_cred = nr_arg->cred,
5118 	};
5119 	int			status;
5120 
5121 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5122 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5123 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5124 		args.bitmask = server->attr_bitmask_nl;
5125 	else
5126 		args.bitmask = server->attr_bitmask;
5127 
5128 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5129 	res.pgbase = args.pgbase;
5130 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5131 			&res.seq_res, 0);
5132 	if (status >= 0) {
5133 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5134 		status += args.pgbase;
5135 	}
5136 
5137 	nfs_invalidate_atime(dir);
5138 
5139 	dprintk("%s: returns %d\n", __func__, status);
5140 	return status;
5141 }
5142 
5143 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5144 			     struct nfs_readdir_res *res)
5145 {
5146 	struct nfs4_exception exception = {
5147 		.interruptible = true,
5148 	};
5149 	int err;
5150 	do {
5151 		err = _nfs4_proc_readdir(arg, res);
5152 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5153 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5154 					    err, &exception);
5155 	} while (exception.retry);
5156 	return err;
5157 }
5158 
5159 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5160 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5161 {
5162 	struct nfs4_createdata *data;
5163 	int mode = sattr->ia_mode;
5164 	int status = -ENOMEM;
5165 
5166 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5167 	if (data == NULL)
5168 		goto out;
5169 
5170 	if (S_ISFIFO(mode))
5171 		data->arg.ftype = NF4FIFO;
5172 	else if (S_ISBLK(mode)) {
5173 		data->arg.ftype = NF4BLK;
5174 		data->arg.u.device.specdata1 = MAJOR(rdev);
5175 		data->arg.u.device.specdata2 = MINOR(rdev);
5176 	}
5177 	else if (S_ISCHR(mode)) {
5178 		data->arg.ftype = NF4CHR;
5179 		data->arg.u.device.specdata1 = MAJOR(rdev);
5180 		data->arg.u.device.specdata2 = MINOR(rdev);
5181 	} else if (!S_ISSOCK(mode)) {
5182 		status = -EINVAL;
5183 		goto out_free;
5184 	}
5185 
5186 	data->arg.label = label;
5187 	status = nfs4_do_create(dir, dentry, data);
5188 out_free:
5189 	nfs4_free_createdata(data);
5190 out:
5191 	return status;
5192 }
5193 
5194 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5195 		struct iattr *sattr, dev_t rdev)
5196 {
5197 	struct nfs_server *server = NFS_SERVER(dir);
5198 	struct nfs4_exception exception = {
5199 		.interruptible = true,
5200 	};
5201 	struct nfs4_label l, *label;
5202 	int err;
5203 
5204 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5205 
5206 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5207 		sattr->ia_mode &= ~current_umask();
5208 	do {
5209 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5210 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5211 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5212 				&exception);
5213 	} while (exception.retry);
5214 
5215 	nfs4_label_release_security(label);
5216 
5217 	return err;
5218 }
5219 
5220 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5221 		 struct nfs_fsstat *fsstat)
5222 {
5223 	struct nfs4_statfs_arg args = {
5224 		.fh = fhandle,
5225 		.bitmask = server->attr_bitmask,
5226 	};
5227 	struct nfs4_statfs_res res = {
5228 		.fsstat = fsstat,
5229 	};
5230 	struct rpc_message msg = {
5231 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5232 		.rpc_argp = &args,
5233 		.rpc_resp = &res,
5234 	};
5235 
5236 	nfs_fattr_init(fsstat->fattr);
5237 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5238 }
5239 
5240 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5241 {
5242 	struct nfs4_exception exception = {
5243 		.interruptible = true,
5244 	};
5245 	int err;
5246 	do {
5247 		err = nfs4_handle_exception(server,
5248 				_nfs4_proc_statfs(server, fhandle, fsstat),
5249 				&exception);
5250 	} while (exception.retry);
5251 	return err;
5252 }
5253 
5254 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5255 		struct nfs_fsinfo *fsinfo)
5256 {
5257 	struct nfs4_fsinfo_arg args = {
5258 		.fh = fhandle,
5259 		.bitmask = server->attr_bitmask,
5260 	};
5261 	struct nfs4_fsinfo_res res = {
5262 		.fsinfo = fsinfo,
5263 	};
5264 	struct rpc_message msg = {
5265 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5266 		.rpc_argp = &args,
5267 		.rpc_resp = &res,
5268 	};
5269 
5270 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5271 }
5272 
5273 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5274 {
5275 	struct nfs4_exception exception = {
5276 		.interruptible = true,
5277 	};
5278 	int err;
5279 
5280 	do {
5281 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5282 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5283 		if (err == 0) {
5284 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5285 			break;
5286 		}
5287 		err = nfs4_handle_exception(server, err, &exception);
5288 	} while (exception.retry);
5289 	return err;
5290 }
5291 
5292 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5293 {
5294 	int error;
5295 
5296 	nfs_fattr_init(fsinfo->fattr);
5297 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5298 	if (error == 0) {
5299 		/* block layout checks this! */
5300 		server->pnfs_blksize = fsinfo->blksize;
5301 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5302 	}
5303 
5304 	return error;
5305 }
5306 
5307 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5308 		struct nfs_pathconf *pathconf)
5309 {
5310 	struct nfs4_pathconf_arg args = {
5311 		.fh = fhandle,
5312 		.bitmask = server->attr_bitmask,
5313 	};
5314 	struct nfs4_pathconf_res res = {
5315 		.pathconf = pathconf,
5316 	};
5317 	struct rpc_message msg = {
5318 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5319 		.rpc_argp = &args,
5320 		.rpc_resp = &res,
5321 	};
5322 
5323 	/* None of the pathconf attributes are mandatory to implement */
5324 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5325 		memset(pathconf, 0, sizeof(*pathconf));
5326 		return 0;
5327 	}
5328 
5329 	nfs_fattr_init(pathconf->fattr);
5330 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5331 }
5332 
5333 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5334 		struct nfs_pathconf *pathconf)
5335 {
5336 	struct nfs4_exception exception = {
5337 		.interruptible = true,
5338 	};
5339 	int err;
5340 
5341 	do {
5342 		err = nfs4_handle_exception(server,
5343 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5344 				&exception);
5345 	} while (exception.retry);
5346 	return err;
5347 }
5348 
5349 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5350 		const struct nfs_open_context *ctx,
5351 		const struct nfs_lock_context *l_ctx,
5352 		fmode_t fmode)
5353 {
5354 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5355 }
5356 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5357 
5358 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5359 		const struct nfs_open_context *ctx,
5360 		const struct nfs_lock_context *l_ctx,
5361 		fmode_t fmode)
5362 {
5363 	nfs4_stateid _current_stateid;
5364 
5365 	/* If the current stateid represents a lost lock, then exit */
5366 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5367 		return true;
5368 	return nfs4_stateid_match(stateid, &_current_stateid);
5369 }
5370 
5371 static bool nfs4_error_stateid_expired(int err)
5372 {
5373 	switch (err) {
5374 	case -NFS4ERR_DELEG_REVOKED:
5375 	case -NFS4ERR_ADMIN_REVOKED:
5376 	case -NFS4ERR_BAD_STATEID:
5377 	case -NFS4ERR_STALE_STATEID:
5378 	case -NFS4ERR_OLD_STATEID:
5379 	case -NFS4ERR_OPENMODE:
5380 	case -NFS4ERR_EXPIRED:
5381 		return true;
5382 	}
5383 	return false;
5384 }
5385 
5386 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5387 {
5388 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5389 
5390 	trace_nfs4_read(hdr, task->tk_status);
5391 	if (task->tk_status < 0) {
5392 		struct nfs4_exception exception = {
5393 			.inode = hdr->inode,
5394 			.state = hdr->args.context->state,
5395 			.stateid = &hdr->args.stateid,
5396 		};
5397 		task->tk_status = nfs4_async_handle_exception(task,
5398 				server, task->tk_status, &exception);
5399 		if (exception.retry) {
5400 			rpc_restart_call_prepare(task);
5401 			return -EAGAIN;
5402 		}
5403 	}
5404 
5405 	if (task->tk_status > 0)
5406 		renew_lease(server, hdr->timestamp);
5407 	return 0;
5408 }
5409 
5410 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5411 		struct nfs_pgio_args *args)
5412 {
5413 
5414 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5415 		nfs4_stateid_is_current(&args->stateid,
5416 				args->context,
5417 				args->lock_context,
5418 				FMODE_READ))
5419 		return false;
5420 	rpc_restart_call_prepare(task);
5421 	return true;
5422 }
5423 
5424 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5425 					 struct nfs_pgio_header *hdr)
5426 {
5427 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5428 	struct rpc_message *msg = &task->tk_msg;
5429 
5430 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5431 	    server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5432 		server->caps &= ~NFS_CAP_READ_PLUS;
5433 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5434 		rpc_restart_call_prepare(task);
5435 		return true;
5436 	}
5437 	return false;
5438 }
5439 
5440 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5441 {
5442 	if (hdr->res.scratch)
5443 		kfree(hdr->res.scratch);
5444 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5445 		return -EAGAIN;
5446 	if (nfs4_read_stateid_changed(task, &hdr->args))
5447 		return -EAGAIN;
5448 	if (nfs4_read_plus_not_supported(task, hdr))
5449 		return -EAGAIN;
5450 	if (task->tk_status > 0)
5451 		nfs_invalidate_atime(hdr->inode);
5452 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5453 				    nfs4_read_done_cb(task, hdr);
5454 }
5455 
5456 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5457 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5458 				    struct rpc_message *msg)
5459 {
5460 	/* Note: We don't use READ_PLUS with pNFS yet */
5461 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5462 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5463 		hdr->res.scratch = kmalloc(32, GFP_KERNEL);
5464 		return hdr->res.scratch != NULL;
5465 	}
5466 	return false;
5467 }
5468 #else
5469 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5470 				    struct rpc_message *msg)
5471 {
5472 	return false;
5473 }
5474 #endif /* CONFIG_NFS_V4_2 */
5475 
5476 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5477 				 struct rpc_message *msg)
5478 {
5479 	hdr->timestamp   = jiffies;
5480 	if (!hdr->pgio_done_cb)
5481 		hdr->pgio_done_cb = nfs4_read_done_cb;
5482 	if (!nfs42_read_plus_support(hdr, msg))
5483 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5484 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5485 }
5486 
5487 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5488 				      struct nfs_pgio_header *hdr)
5489 {
5490 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5491 			&hdr->args.seq_args,
5492 			&hdr->res.seq_res,
5493 			task))
5494 		return 0;
5495 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5496 				hdr->args.lock_context,
5497 				hdr->rw_mode) == -EIO)
5498 		return -EIO;
5499 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5500 		return -EIO;
5501 	return 0;
5502 }
5503 
5504 static int nfs4_write_done_cb(struct rpc_task *task,
5505 			      struct nfs_pgio_header *hdr)
5506 {
5507 	struct inode *inode = hdr->inode;
5508 
5509 	trace_nfs4_write(hdr, task->tk_status);
5510 	if (task->tk_status < 0) {
5511 		struct nfs4_exception exception = {
5512 			.inode = hdr->inode,
5513 			.state = hdr->args.context->state,
5514 			.stateid = &hdr->args.stateid,
5515 		};
5516 		task->tk_status = nfs4_async_handle_exception(task,
5517 				NFS_SERVER(inode), task->tk_status,
5518 				&exception);
5519 		if (exception.retry) {
5520 			rpc_restart_call_prepare(task);
5521 			return -EAGAIN;
5522 		}
5523 	}
5524 	if (task->tk_status >= 0) {
5525 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5526 		nfs_writeback_update_inode(hdr);
5527 	}
5528 	return 0;
5529 }
5530 
5531 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5532 		struct nfs_pgio_args *args)
5533 {
5534 
5535 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5536 		nfs4_stateid_is_current(&args->stateid,
5537 				args->context,
5538 				args->lock_context,
5539 				FMODE_WRITE))
5540 		return false;
5541 	rpc_restart_call_prepare(task);
5542 	return true;
5543 }
5544 
5545 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5546 {
5547 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5548 		return -EAGAIN;
5549 	if (nfs4_write_stateid_changed(task, &hdr->args))
5550 		return -EAGAIN;
5551 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5552 		nfs4_write_done_cb(task, hdr);
5553 }
5554 
5555 static
5556 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5557 {
5558 	/* Don't request attributes for pNFS or O_DIRECT writes */
5559 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5560 		return false;
5561 	/* Otherwise, request attributes if and only if we don't hold
5562 	 * a delegation
5563 	 */
5564 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5565 }
5566 
5567 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5568 		      struct inode *inode, unsigned long cache_validity)
5569 {
5570 	struct nfs_server *server = NFS_SERVER(inode);
5571 	unsigned int i;
5572 
5573 	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5574 	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5575 
5576 	if (cache_validity & NFS_INO_INVALID_CHANGE)
5577 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5578 	if (cache_validity & NFS_INO_INVALID_ATIME)
5579 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5580 	if (cache_validity & NFS_INO_INVALID_MODE)
5581 		bitmask[1] |= FATTR4_WORD1_MODE;
5582 	if (cache_validity & NFS_INO_INVALID_OTHER)
5583 		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5584 	if (cache_validity & NFS_INO_INVALID_NLINK)
5585 		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5586 	if (cache_validity & NFS_INO_INVALID_CTIME)
5587 		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5588 	if (cache_validity & NFS_INO_INVALID_MTIME)
5589 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5590 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5591 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5592 
5593 	if (cache_validity & NFS_INO_INVALID_SIZE)
5594 		bitmask[0] |= FATTR4_WORD0_SIZE;
5595 
5596 	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5597 		bitmask[i] &= server->attr_bitmask[i];
5598 }
5599 
5600 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5601 				  struct rpc_message *msg,
5602 				  struct rpc_clnt **clnt)
5603 {
5604 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5605 
5606 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5607 		hdr->args.bitmask = NULL;
5608 		hdr->res.fattr = NULL;
5609 	} else {
5610 		nfs4_bitmask_set(hdr->args.bitmask_store,
5611 				 server->cache_consistency_bitmask,
5612 				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5613 		hdr->args.bitmask = hdr->args.bitmask_store;
5614 	}
5615 
5616 	if (!hdr->pgio_done_cb)
5617 		hdr->pgio_done_cb = nfs4_write_done_cb;
5618 	hdr->res.server = server;
5619 	hdr->timestamp   = jiffies;
5620 
5621 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5622 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5623 	nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5624 }
5625 
5626 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5627 {
5628 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5629 			&data->args.seq_args,
5630 			&data->res.seq_res,
5631 			task);
5632 }
5633 
5634 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5635 {
5636 	struct inode *inode = data->inode;
5637 
5638 	trace_nfs4_commit(data, task->tk_status);
5639 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5640 				    NULL, NULL) == -EAGAIN) {
5641 		rpc_restart_call_prepare(task);
5642 		return -EAGAIN;
5643 	}
5644 	return 0;
5645 }
5646 
5647 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5648 {
5649 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5650 		return -EAGAIN;
5651 	return data->commit_done_cb(task, data);
5652 }
5653 
5654 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5655 				   struct rpc_clnt **clnt)
5656 {
5657 	struct nfs_server *server = NFS_SERVER(data->inode);
5658 
5659 	if (data->commit_done_cb == NULL)
5660 		data->commit_done_cb = nfs4_commit_done_cb;
5661 	data->res.server = server;
5662 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5663 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5664 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5665 }
5666 
5667 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5668 				struct nfs_commitres *res)
5669 {
5670 	struct inode *dst_inode = file_inode(dst);
5671 	struct nfs_server *server = NFS_SERVER(dst_inode);
5672 	struct rpc_message msg = {
5673 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5674 		.rpc_argp = args,
5675 		.rpc_resp = res,
5676 	};
5677 
5678 	args->fh = NFS_FH(dst_inode);
5679 	return nfs4_call_sync(server->client, server, &msg,
5680 			&args->seq_args, &res->seq_res, 1);
5681 }
5682 
5683 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5684 {
5685 	struct nfs_commitargs args = {
5686 		.offset = offset,
5687 		.count = count,
5688 	};
5689 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5690 	struct nfs4_exception exception = { };
5691 	int status;
5692 
5693 	do {
5694 		status = _nfs4_proc_commit(dst, &args, res);
5695 		status = nfs4_handle_exception(dst_server, status, &exception);
5696 	} while (exception.retry);
5697 
5698 	return status;
5699 }
5700 
5701 struct nfs4_renewdata {
5702 	struct nfs_client	*client;
5703 	unsigned long		timestamp;
5704 };
5705 
5706 /*
5707  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5708  * standalone procedure for queueing an asynchronous RENEW.
5709  */
5710 static void nfs4_renew_release(void *calldata)
5711 {
5712 	struct nfs4_renewdata *data = calldata;
5713 	struct nfs_client *clp = data->client;
5714 
5715 	if (refcount_read(&clp->cl_count) > 1)
5716 		nfs4_schedule_state_renewal(clp);
5717 	nfs_put_client(clp);
5718 	kfree(data);
5719 }
5720 
5721 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5722 {
5723 	struct nfs4_renewdata *data = calldata;
5724 	struct nfs_client *clp = data->client;
5725 	unsigned long timestamp = data->timestamp;
5726 
5727 	trace_nfs4_renew_async(clp, task->tk_status);
5728 	switch (task->tk_status) {
5729 	case 0:
5730 		break;
5731 	case -NFS4ERR_LEASE_MOVED:
5732 		nfs4_schedule_lease_moved_recovery(clp);
5733 		break;
5734 	default:
5735 		/* Unless we're shutting down, schedule state recovery! */
5736 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5737 			return;
5738 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5739 			nfs4_schedule_lease_recovery(clp);
5740 			return;
5741 		}
5742 		nfs4_schedule_path_down_recovery(clp);
5743 	}
5744 	do_renew_lease(clp, timestamp);
5745 }
5746 
5747 static const struct rpc_call_ops nfs4_renew_ops = {
5748 	.rpc_call_done = nfs4_renew_done,
5749 	.rpc_release = nfs4_renew_release,
5750 };
5751 
5752 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5753 {
5754 	struct rpc_message msg = {
5755 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5756 		.rpc_argp	= clp,
5757 		.rpc_cred	= cred,
5758 	};
5759 	struct nfs4_renewdata *data;
5760 
5761 	if (renew_flags == 0)
5762 		return 0;
5763 	if (!refcount_inc_not_zero(&clp->cl_count))
5764 		return -EIO;
5765 	data = kmalloc(sizeof(*data), GFP_NOFS);
5766 	if (data == NULL) {
5767 		nfs_put_client(clp);
5768 		return -ENOMEM;
5769 	}
5770 	data->client = clp;
5771 	data->timestamp = jiffies;
5772 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5773 			&nfs4_renew_ops, data);
5774 }
5775 
5776 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5777 {
5778 	struct rpc_message msg = {
5779 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5780 		.rpc_argp	= clp,
5781 		.rpc_cred	= cred,
5782 	};
5783 	unsigned long now = jiffies;
5784 	int status;
5785 
5786 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5787 	if (status < 0)
5788 		return status;
5789 	do_renew_lease(clp, now);
5790 	return 0;
5791 }
5792 
5793 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5794 				      enum nfs4_acl_type type)
5795 {
5796 	switch (type) {
5797 	default:
5798 		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5799 	case NFS4ACL_DACL:
5800 		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5801 	case NFS4ACL_SACL:
5802 		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5803 	}
5804 }
5805 
5806 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5807  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5808  * the stack.
5809  */
5810 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5811 
5812 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5813 		struct page **pages)
5814 {
5815 	struct page *newpage, **spages;
5816 	int rc = 0;
5817 	size_t len;
5818 	spages = pages;
5819 
5820 	do {
5821 		len = min_t(size_t, PAGE_SIZE, buflen);
5822 		newpage = alloc_page(GFP_KERNEL);
5823 
5824 		if (newpage == NULL)
5825 			goto unwind;
5826 		memcpy(page_address(newpage), buf, len);
5827 		buf += len;
5828 		buflen -= len;
5829 		*pages++ = newpage;
5830 		rc++;
5831 	} while (buflen != 0);
5832 
5833 	return rc;
5834 
5835 unwind:
5836 	for(; rc > 0; rc--)
5837 		__free_page(spages[rc-1]);
5838 	return -ENOMEM;
5839 }
5840 
5841 struct nfs4_cached_acl {
5842 	enum nfs4_acl_type type;
5843 	int cached;
5844 	size_t len;
5845 	char data[];
5846 };
5847 
5848 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5849 {
5850 	struct nfs_inode *nfsi = NFS_I(inode);
5851 
5852 	spin_lock(&inode->i_lock);
5853 	kfree(nfsi->nfs4_acl);
5854 	nfsi->nfs4_acl = acl;
5855 	spin_unlock(&inode->i_lock);
5856 }
5857 
5858 static void nfs4_zap_acl_attr(struct inode *inode)
5859 {
5860 	nfs4_set_cached_acl(inode, NULL);
5861 }
5862 
5863 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5864 				    size_t buflen, enum nfs4_acl_type type)
5865 {
5866 	struct nfs_inode *nfsi = NFS_I(inode);
5867 	struct nfs4_cached_acl *acl;
5868 	int ret = -ENOENT;
5869 
5870 	spin_lock(&inode->i_lock);
5871 	acl = nfsi->nfs4_acl;
5872 	if (acl == NULL)
5873 		goto out;
5874 	if (acl->type != type)
5875 		goto out;
5876 	if (buf == NULL) /* user is just asking for length */
5877 		goto out_len;
5878 	if (acl->cached == 0)
5879 		goto out;
5880 	ret = -ERANGE; /* see getxattr(2) man page */
5881 	if (acl->len > buflen)
5882 		goto out;
5883 	memcpy(buf, acl->data, acl->len);
5884 out_len:
5885 	ret = acl->len;
5886 out:
5887 	spin_unlock(&inode->i_lock);
5888 	return ret;
5889 }
5890 
5891 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5892 				  size_t pgbase, size_t acl_len,
5893 				  enum nfs4_acl_type type)
5894 {
5895 	struct nfs4_cached_acl *acl;
5896 	size_t buflen = sizeof(*acl) + acl_len;
5897 
5898 	if (buflen <= PAGE_SIZE) {
5899 		acl = kmalloc(buflen, GFP_KERNEL);
5900 		if (acl == NULL)
5901 			goto out;
5902 		acl->cached = 1;
5903 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5904 	} else {
5905 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5906 		if (acl == NULL)
5907 			goto out;
5908 		acl->cached = 0;
5909 	}
5910 	acl->type = type;
5911 	acl->len = acl_len;
5912 out:
5913 	nfs4_set_cached_acl(inode, acl);
5914 }
5915 
5916 /*
5917  * The getxattr API returns the required buffer length when called with a
5918  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5919  * the required buf.  On a NULL buf, we send a page of data to the server
5920  * guessing that the ACL request can be serviced by a page. If so, we cache
5921  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5922  * the cache. If not so, we throw away the page, and cache the required
5923  * length. The next getxattr call will then produce another round trip to
5924  * the server, this time with the input buf of the required size.
5925  */
5926 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5927 				       size_t buflen, enum nfs4_acl_type type)
5928 {
5929 	struct page **pages;
5930 	struct nfs_getaclargs args = {
5931 		.fh = NFS_FH(inode),
5932 		.acl_type = type,
5933 		.acl_len = buflen,
5934 	};
5935 	struct nfs_getaclres res = {
5936 		.acl_type = type,
5937 		.acl_len = buflen,
5938 	};
5939 	struct rpc_message msg = {
5940 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5941 		.rpc_argp = &args,
5942 		.rpc_resp = &res,
5943 	};
5944 	unsigned int npages;
5945 	int ret = -ENOMEM, i;
5946 	struct nfs_server *server = NFS_SERVER(inode);
5947 
5948 	if (buflen == 0)
5949 		buflen = server->rsize;
5950 
5951 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5952 	pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5953 	if (!pages)
5954 		return -ENOMEM;
5955 
5956 	args.acl_pages = pages;
5957 
5958 	for (i = 0; i < npages; i++) {
5959 		pages[i] = alloc_page(GFP_KERNEL);
5960 		if (!pages[i])
5961 			goto out_free;
5962 	}
5963 
5964 	/* for decoding across pages */
5965 	res.acl_scratch = alloc_page(GFP_KERNEL);
5966 	if (!res.acl_scratch)
5967 		goto out_free;
5968 
5969 	args.acl_len = npages * PAGE_SIZE;
5970 
5971 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5972 		__func__, buf, buflen, npages, args.acl_len);
5973 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5974 			     &msg, &args.seq_args, &res.seq_res, 0);
5975 	if (ret)
5976 		goto out_free;
5977 
5978 	/* Handle the case where the passed-in buffer is too short */
5979 	if (res.acl_flags & NFS4_ACL_TRUNC) {
5980 		/* Did the user only issue a request for the acl length? */
5981 		if (buf == NULL)
5982 			goto out_ok;
5983 		ret = -ERANGE;
5984 		goto out_free;
5985 	}
5986 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5987 			      type);
5988 	if (buf) {
5989 		if (res.acl_len > buflen) {
5990 			ret = -ERANGE;
5991 			goto out_free;
5992 		}
5993 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5994 	}
5995 out_ok:
5996 	ret = res.acl_len;
5997 out_free:
5998 	for (i = 0; i < npages; i++)
5999 		if (pages[i])
6000 			__free_page(pages[i]);
6001 	if (res.acl_scratch)
6002 		__free_page(res.acl_scratch);
6003 	kfree(pages);
6004 	return ret;
6005 }
6006 
6007 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6008 				     size_t buflen, enum nfs4_acl_type type)
6009 {
6010 	struct nfs4_exception exception = {
6011 		.interruptible = true,
6012 	};
6013 	ssize_t ret;
6014 	do {
6015 		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6016 		trace_nfs4_get_acl(inode, ret);
6017 		if (ret >= 0)
6018 			break;
6019 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6020 	} while (exception.retry);
6021 	return ret;
6022 }
6023 
6024 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6025 				 enum nfs4_acl_type type)
6026 {
6027 	struct nfs_server *server = NFS_SERVER(inode);
6028 	int ret;
6029 
6030 	if (!nfs4_server_supports_acls(server, type))
6031 		return -EOPNOTSUPP;
6032 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6033 	if (ret < 0)
6034 		return ret;
6035 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6036 		nfs_zap_acl_cache(inode);
6037 	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6038 	if (ret != -ENOENT)
6039 		/* -ENOENT is returned if there is no ACL or if there is an ACL
6040 		 * but no cached acl data, just the acl length */
6041 		return ret;
6042 	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6043 }
6044 
6045 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6046 			       size_t buflen, enum nfs4_acl_type type)
6047 {
6048 	struct nfs_server *server = NFS_SERVER(inode);
6049 	struct page *pages[NFS4ACL_MAXPAGES];
6050 	struct nfs_setaclargs arg = {
6051 		.fh = NFS_FH(inode),
6052 		.acl_type = type,
6053 		.acl_len = buflen,
6054 		.acl_pages = pages,
6055 	};
6056 	struct nfs_setaclres res;
6057 	struct rpc_message msg = {
6058 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6059 		.rpc_argp	= &arg,
6060 		.rpc_resp	= &res,
6061 	};
6062 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6063 	int ret, i;
6064 
6065 	/* You can't remove system.nfs4_acl: */
6066 	if (buflen == 0)
6067 		return -EINVAL;
6068 	if (!nfs4_server_supports_acls(server, type))
6069 		return -EOPNOTSUPP;
6070 	if (npages > ARRAY_SIZE(pages))
6071 		return -ERANGE;
6072 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6073 	if (i < 0)
6074 		return i;
6075 	nfs4_inode_make_writeable(inode);
6076 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6077 
6078 	/*
6079 	 * Free each page after tx, so the only ref left is
6080 	 * held by the network stack
6081 	 */
6082 	for (; i > 0; i--)
6083 		put_page(pages[i-1]);
6084 
6085 	/*
6086 	 * Acl update can result in inode attribute update.
6087 	 * so mark the attribute cache invalid.
6088 	 */
6089 	spin_lock(&inode->i_lock);
6090 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6091 					     NFS_INO_INVALID_CTIME |
6092 					     NFS_INO_REVAL_FORCED);
6093 	spin_unlock(&inode->i_lock);
6094 	nfs_access_zap_cache(inode);
6095 	nfs_zap_acl_cache(inode);
6096 	return ret;
6097 }
6098 
6099 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6100 			     size_t buflen, enum nfs4_acl_type type)
6101 {
6102 	struct nfs4_exception exception = { };
6103 	int err;
6104 	do {
6105 		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6106 		trace_nfs4_set_acl(inode, err);
6107 		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6108 			/*
6109 			 * no need to retry since the kernel
6110 			 * isn't involved in encoding the ACEs.
6111 			 */
6112 			err = -EINVAL;
6113 			break;
6114 		}
6115 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6116 				&exception);
6117 	} while (exception.retry);
6118 	return err;
6119 }
6120 
6121 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6122 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6123 					size_t buflen)
6124 {
6125 	struct nfs_server *server = NFS_SERVER(inode);
6126 	struct nfs4_label label = {0, 0, buflen, buf};
6127 
6128 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6129 	struct nfs_fattr fattr = {
6130 		.label = &label,
6131 	};
6132 	struct nfs4_getattr_arg arg = {
6133 		.fh		= NFS_FH(inode),
6134 		.bitmask	= bitmask,
6135 	};
6136 	struct nfs4_getattr_res res = {
6137 		.fattr		= &fattr,
6138 		.server		= server,
6139 	};
6140 	struct rpc_message msg = {
6141 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6142 		.rpc_argp	= &arg,
6143 		.rpc_resp	= &res,
6144 	};
6145 	int ret;
6146 
6147 	nfs_fattr_init(&fattr);
6148 
6149 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6150 	if (ret)
6151 		return ret;
6152 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6153 		return -ENOENT;
6154 	return label.len;
6155 }
6156 
6157 static int nfs4_get_security_label(struct inode *inode, void *buf,
6158 					size_t buflen)
6159 {
6160 	struct nfs4_exception exception = {
6161 		.interruptible = true,
6162 	};
6163 	int err;
6164 
6165 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6166 		return -EOPNOTSUPP;
6167 
6168 	do {
6169 		err = _nfs4_get_security_label(inode, buf, buflen);
6170 		trace_nfs4_get_security_label(inode, err);
6171 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6172 				&exception);
6173 	} while (exception.retry);
6174 	return err;
6175 }
6176 
6177 static int _nfs4_do_set_security_label(struct inode *inode,
6178 		struct nfs4_label *ilabel,
6179 		struct nfs_fattr *fattr)
6180 {
6181 
6182 	struct iattr sattr = {0};
6183 	struct nfs_server *server = NFS_SERVER(inode);
6184 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6185 	struct nfs_setattrargs arg = {
6186 		.fh		= NFS_FH(inode),
6187 		.iap		= &sattr,
6188 		.server		= server,
6189 		.bitmask	= bitmask,
6190 		.label		= ilabel,
6191 	};
6192 	struct nfs_setattrres res = {
6193 		.fattr		= fattr,
6194 		.server		= server,
6195 	};
6196 	struct rpc_message msg = {
6197 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6198 		.rpc_argp	= &arg,
6199 		.rpc_resp	= &res,
6200 	};
6201 	int status;
6202 
6203 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6204 
6205 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6206 	if (status)
6207 		dprintk("%s failed: %d\n", __func__, status);
6208 
6209 	return status;
6210 }
6211 
6212 static int nfs4_do_set_security_label(struct inode *inode,
6213 		struct nfs4_label *ilabel,
6214 		struct nfs_fattr *fattr)
6215 {
6216 	struct nfs4_exception exception = { };
6217 	int err;
6218 
6219 	do {
6220 		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6221 		trace_nfs4_set_security_label(inode, err);
6222 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6223 				&exception);
6224 	} while (exception.retry);
6225 	return err;
6226 }
6227 
6228 static int
6229 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6230 {
6231 	struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6232 	struct nfs_fattr *fattr;
6233 	int status;
6234 
6235 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6236 		return -EOPNOTSUPP;
6237 
6238 	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6239 	if (fattr == NULL)
6240 		return -ENOMEM;
6241 
6242 	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6243 	if (status == 0)
6244 		nfs_setsecurity(inode, fattr);
6245 
6246 	return status;
6247 }
6248 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6249 
6250 
6251 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6252 				    nfs4_verifier *bootverf)
6253 {
6254 	__be32 verf[2];
6255 
6256 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6257 		/* An impossible timestamp guarantees this value
6258 		 * will never match a generated boot time. */
6259 		verf[0] = cpu_to_be32(U32_MAX);
6260 		verf[1] = cpu_to_be32(U32_MAX);
6261 	} else {
6262 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6263 		u64 ns = ktime_to_ns(nn->boot_time);
6264 
6265 		verf[0] = cpu_to_be32(ns >> 32);
6266 		verf[1] = cpu_to_be32(ns);
6267 	}
6268 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6269 }
6270 
6271 static size_t
6272 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6273 {
6274 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6275 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6276 	const char *id;
6277 
6278 	buf[0] = '\0';
6279 
6280 	if (nn_clp) {
6281 		rcu_read_lock();
6282 		id = rcu_dereference(nn_clp->identifier);
6283 		if (id)
6284 			strscpy(buf, id, buflen);
6285 		rcu_read_unlock();
6286 	}
6287 
6288 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6289 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6290 
6291 	return strlen(buf);
6292 }
6293 
6294 static int
6295 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6296 {
6297 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6298 	size_t buflen;
6299 	size_t len;
6300 	char *str;
6301 
6302 	if (clp->cl_owner_id != NULL)
6303 		return 0;
6304 
6305 	rcu_read_lock();
6306 	len = 14 +
6307 		strlen(clp->cl_rpcclient->cl_nodename) +
6308 		1 +
6309 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6310 		1;
6311 	rcu_read_unlock();
6312 
6313 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6314 	if (buflen)
6315 		len += buflen + 1;
6316 
6317 	if (len > NFS4_OPAQUE_LIMIT + 1)
6318 		return -EINVAL;
6319 
6320 	/*
6321 	 * Since this string is allocated at mount time, and held until the
6322 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6323 	 * about a memory-reclaim deadlock.
6324 	 */
6325 	str = kmalloc(len, GFP_KERNEL);
6326 	if (!str)
6327 		return -ENOMEM;
6328 
6329 	rcu_read_lock();
6330 	if (buflen)
6331 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6332 			  clp->cl_rpcclient->cl_nodename, buf,
6333 			  rpc_peeraddr2str(clp->cl_rpcclient,
6334 					   RPC_DISPLAY_ADDR));
6335 	else
6336 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6337 			  clp->cl_rpcclient->cl_nodename,
6338 			  rpc_peeraddr2str(clp->cl_rpcclient,
6339 					   RPC_DISPLAY_ADDR));
6340 	rcu_read_unlock();
6341 
6342 	clp->cl_owner_id = str;
6343 	return 0;
6344 }
6345 
6346 static int
6347 nfs4_init_uniform_client_string(struct nfs_client *clp)
6348 {
6349 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6350 	size_t buflen;
6351 	size_t len;
6352 	char *str;
6353 
6354 	if (clp->cl_owner_id != NULL)
6355 		return 0;
6356 
6357 	len = 10 + 10 + 1 + 10 + 1 +
6358 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6359 
6360 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6361 	if (buflen)
6362 		len += buflen + 1;
6363 
6364 	if (len > NFS4_OPAQUE_LIMIT + 1)
6365 		return -EINVAL;
6366 
6367 	/*
6368 	 * Since this string is allocated at mount time, and held until the
6369 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6370 	 * about a memory-reclaim deadlock.
6371 	 */
6372 	str = kmalloc(len, GFP_KERNEL);
6373 	if (!str)
6374 		return -ENOMEM;
6375 
6376 	if (buflen)
6377 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6378 			  clp->rpc_ops->version, clp->cl_minorversion,
6379 			  buf, clp->cl_rpcclient->cl_nodename);
6380 	else
6381 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6382 			  clp->rpc_ops->version, clp->cl_minorversion,
6383 			  clp->cl_rpcclient->cl_nodename);
6384 	clp->cl_owner_id = str;
6385 	return 0;
6386 }
6387 
6388 /*
6389  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6390  * services.  Advertise one based on the address family of the
6391  * clientaddr.
6392  */
6393 static unsigned int
6394 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6395 {
6396 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6397 		return scnprintf(buf, len, "tcp6");
6398 	else
6399 		return scnprintf(buf, len, "tcp");
6400 }
6401 
6402 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6403 {
6404 	struct nfs4_setclientid *sc = calldata;
6405 
6406 	if (task->tk_status == 0)
6407 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6408 }
6409 
6410 static const struct rpc_call_ops nfs4_setclientid_ops = {
6411 	.rpc_call_done = nfs4_setclientid_done,
6412 };
6413 
6414 /**
6415  * nfs4_proc_setclientid - Negotiate client ID
6416  * @clp: state data structure
6417  * @program: RPC program for NFSv4 callback service
6418  * @port: IP port number for NFS4 callback service
6419  * @cred: credential to use for this call
6420  * @res: where to place the result
6421  *
6422  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6423  */
6424 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6425 		unsigned short port, const struct cred *cred,
6426 		struct nfs4_setclientid_res *res)
6427 {
6428 	nfs4_verifier sc_verifier;
6429 	struct nfs4_setclientid setclientid = {
6430 		.sc_verifier = &sc_verifier,
6431 		.sc_prog = program,
6432 		.sc_clnt = clp,
6433 	};
6434 	struct rpc_message msg = {
6435 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6436 		.rpc_argp = &setclientid,
6437 		.rpc_resp = res,
6438 		.rpc_cred = cred,
6439 	};
6440 	struct rpc_task_setup task_setup_data = {
6441 		.rpc_client = clp->cl_rpcclient,
6442 		.rpc_message = &msg,
6443 		.callback_ops = &nfs4_setclientid_ops,
6444 		.callback_data = &setclientid,
6445 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6446 	};
6447 	unsigned long now = jiffies;
6448 	int status;
6449 
6450 	/* nfs_client_id4 */
6451 	nfs4_init_boot_verifier(clp, &sc_verifier);
6452 
6453 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6454 		status = nfs4_init_uniform_client_string(clp);
6455 	else
6456 		status = nfs4_init_nonuniform_client_string(clp);
6457 
6458 	if (status)
6459 		goto out;
6460 
6461 	/* cb_client4 */
6462 	setclientid.sc_netid_len =
6463 				nfs4_init_callback_netid(clp,
6464 						setclientid.sc_netid,
6465 						sizeof(setclientid.sc_netid));
6466 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6467 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6468 				clp->cl_ipaddr, port >> 8, port & 255);
6469 
6470 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6471 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6472 		clp->cl_owner_id);
6473 
6474 	status = nfs4_call_sync_custom(&task_setup_data);
6475 	if (setclientid.sc_cred) {
6476 		kfree(clp->cl_acceptor);
6477 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6478 		put_rpccred(setclientid.sc_cred);
6479 	}
6480 
6481 	if (status == 0)
6482 		do_renew_lease(clp, now);
6483 out:
6484 	trace_nfs4_setclientid(clp, status);
6485 	dprintk("NFS reply setclientid: %d\n", status);
6486 	return status;
6487 }
6488 
6489 /**
6490  * nfs4_proc_setclientid_confirm - Confirm client ID
6491  * @clp: state data structure
6492  * @arg: result of a previous SETCLIENTID
6493  * @cred: credential to use for this call
6494  *
6495  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6496  */
6497 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6498 		struct nfs4_setclientid_res *arg,
6499 		const struct cred *cred)
6500 {
6501 	struct rpc_message msg = {
6502 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6503 		.rpc_argp = arg,
6504 		.rpc_cred = cred,
6505 	};
6506 	int status;
6507 
6508 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6509 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6510 		clp->cl_clientid);
6511 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6512 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6513 	trace_nfs4_setclientid_confirm(clp, status);
6514 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6515 	return status;
6516 }
6517 
6518 struct nfs4_delegreturndata {
6519 	struct nfs4_delegreturnargs args;
6520 	struct nfs4_delegreturnres res;
6521 	struct nfs_fh fh;
6522 	nfs4_stateid stateid;
6523 	unsigned long timestamp;
6524 	struct {
6525 		struct nfs4_layoutreturn_args arg;
6526 		struct nfs4_layoutreturn_res res;
6527 		struct nfs4_xdr_opaque_data ld_private;
6528 		u32 roc_barrier;
6529 		bool roc;
6530 	} lr;
6531 	struct nfs_fattr fattr;
6532 	int rpc_status;
6533 	struct inode *inode;
6534 };
6535 
6536 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6537 {
6538 	struct nfs4_delegreturndata *data = calldata;
6539 	struct nfs4_exception exception = {
6540 		.inode = data->inode,
6541 		.stateid = &data->stateid,
6542 		.task_is_privileged = data->args.seq_args.sa_privileged,
6543 	};
6544 
6545 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6546 		return;
6547 
6548 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6549 
6550 	/* Handle Layoutreturn errors */
6551 	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6552 			  &data->res.lr_ret) == -EAGAIN)
6553 		goto out_restart;
6554 
6555 	switch (task->tk_status) {
6556 	case 0:
6557 		renew_lease(data->res.server, data->timestamp);
6558 		break;
6559 	case -NFS4ERR_ADMIN_REVOKED:
6560 	case -NFS4ERR_DELEG_REVOKED:
6561 	case -NFS4ERR_EXPIRED:
6562 		nfs4_free_revoked_stateid(data->res.server,
6563 				data->args.stateid,
6564 				task->tk_msg.rpc_cred);
6565 		fallthrough;
6566 	case -NFS4ERR_BAD_STATEID:
6567 	case -NFS4ERR_STALE_STATEID:
6568 	case -ETIMEDOUT:
6569 		task->tk_status = 0;
6570 		break;
6571 	case -NFS4ERR_OLD_STATEID:
6572 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6573 			nfs4_stateid_seqid_inc(&data->stateid);
6574 		if (data->args.bitmask) {
6575 			data->args.bitmask = NULL;
6576 			data->res.fattr = NULL;
6577 		}
6578 		goto out_restart;
6579 	case -NFS4ERR_ACCESS:
6580 		if (data->args.bitmask) {
6581 			data->args.bitmask = NULL;
6582 			data->res.fattr = NULL;
6583 			goto out_restart;
6584 		}
6585 		fallthrough;
6586 	default:
6587 		task->tk_status = nfs4_async_handle_exception(task,
6588 				data->res.server, task->tk_status,
6589 				&exception);
6590 		if (exception.retry)
6591 			goto out_restart;
6592 	}
6593 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6594 	data->rpc_status = task->tk_status;
6595 	return;
6596 out_restart:
6597 	task->tk_status = 0;
6598 	rpc_restart_call_prepare(task);
6599 }
6600 
6601 static void nfs4_delegreturn_release(void *calldata)
6602 {
6603 	struct nfs4_delegreturndata *data = calldata;
6604 	struct inode *inode = data->inode;
6605 
6606 	if (data->lr.roc)
6607 		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6608 				 data->res.lr_ret);
6609 	if (inode) {
6610 		nfs4_fattr_set_prechange(&data->fattr,
6611 					 inode_peek_iversion_raw(inode));
6612 		nfs_refresh_inode(inode, &data->fattr);
6613 		nfs_iput_and_deactive(inode);
6614 	}
6615 	kfree(calldata);
6616 }
6617 
6618 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6619 {
6620 	struct nfs4_delegreturndata *d_data;
6621 	struct pnfs_layout_hdr *lo;
6622 
6623 	d_data = data;
6624 
6625 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6626 		nfs4_sequence_done(task, &d_data->res.seq_res);
6627 		return;
6628 	}
6629 
6630 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6631 	if (lo && !pnfs_layout_is_valid(lo)) {
6632 		d_data->args.lr_args = NULL;
6633 		d_data->res.lr_res = NULL;
6634 	}
6635 
6636 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6637 			&d_data->args.seq_args,
6638 			&d_data->res.seq_res,
6639 			task);
6640 }
6641 
6642 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6643 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6644 	.rpc_call_done = nfs4_delegreturn_done,
6645 	.rpc_release = nfs4_delegreturn_release,
6646 };
6647 
6648 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6649 {
6650 	struct nfs4_delegreturndata *data;
6651 	struct nfs_server *server = NFS_SERVER(inode);
6652 	struct rpc_task *task;
6653 	struct rpc_message msg = {
6654 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6655 		.rpc_cred = cred,
6656 	};
6657 	struct rpc_task_setup task_setup_data = {
6658 		.rpc_client = server->client,
6659 		.rpc_message = &msg,
6660 		.callback_ops = &nfs4_delegreturn_ops,
6661 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6662 	};
6663 	int status = 0;
6664 
6665 	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6666 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6667 
6668 	data = kzalloc(sizeof(*data), GFP_KERNEL);
6669 	if (data == NULL)
6670 		return -ENOMEM;
6671 
6672 	nfs4_state_protect(server->nfs_client,
6673 			NFS_SP4_MACH_CRED_CLEANUP,
6674 			&task_setup_data.rpc_client, &msg);
6675 
6676 	data->args.fhandle = &data->fh;
6677 	data->args.stateid = &data->stateid;
6678 	nfs4_bitmask_set(data->args.bitmask_store,
6679 			 server->cache_consistency_bitmask, inode, 0);
6680 	data->args.bitmask = data->args.bitmask_store;
6681 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6682 	nfs4_stateid_copy(&data->stateid, stateid);
6683 	data->res.fattr = &data->fattr;
6684 	data->res.server = server;
6685 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6686 	data->lr.arg.ld_private = &data->lr.ld_private;
6687 	nfs_fattr_init(data->res.fattr);
6688 	data->timestamp = jiffies;
6689 	data->rpc_status = 0;
6690 	data->inode = nfs_igrab_and_active(inode);
6691 	if (data->inode || issync) {
6692 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6693 					cred);
6694 		if (data->lr.roc) {
6695 			data->args.lr_args = &data->lr.arg;
6696 			data->res.lr_res = &data->lr.res;
6697 		}
6698 	}
6699 
6700 	if (!data->inode)
6701 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6702 				   1);
6703 	else
6704 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6705 				   0);
6706 	task_setup_data.callback_data = data;
6707 	msg.rpc_argp = &data->args;
6708 	msg.rpc_resp = &data->res;
6709 	task = rpc_run_task(&task_setup_data);
6710 	if (IS_ERR(task))
6711 		return PTR_ERR(task);
6712 	if (!issync)
6713 		goto out;
6714 	status = rpc_wait_for_completion_task(task);
6715 	if (status != 0)
6716 		goto out;
6717 	status = data->rpc_status;
6718 out:
6719 	rpc_put_task(task);
6720 	return status;
6721 }
6722 
6723 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6724 {
6725 	struct nfs_server *server = NFS_SERVER(inode);
6726 	struct nfs4_exception exception = { };
6727 	int err;
6728 	do {
6729 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6730 		trace_nfs4_delegreturn(inode, stateid, err);
6731 		switch (err) {
6732 			case -NFS4ERR_STALE_STATEID:
6733 			case -NFS4ERR_EXPIRED:
6734 			case 0:
6735 				return 0;
6736 		}
6737 		err = nfs4_handle_exception(server, err, &exception);
6738 	} while (exception.retry);
6739 	return err;
6740 }
6741 
6742 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6743 {
6744 	struct inode *inode = state->inode;
6745 	struct nfs_server *server = NFS_SERVER(inode);
6746 	struct nfs_client *clp = server->nfs_client;
6747 	struct nfs_lockt_args arg = {
6748 		.fh = NFS_FH(inode),
6749 		.fl = request,
6750 	};
6751 	struct nfs_lockt_res res = {
6752 		.denied = request,
6753 	};
6754 	struct rpc_message msg = {
6755 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6756 		.rpc_argp	= &arg,
6757 		.rpc_resp	= &res,
6758 		.rpc_cred	= state->owner->so_cred,
6759 	};
6760 	struct nfs4_lock_state *lsp;
6761 	int status;
6762 
6763 	arg.lock_owner.clientid = clp->cl_clientid;
6764 	status = nfs4_set_lock_state(state, request);
6765 	if (status != 0)
6766 		goto out;
6767 	lsp = request->fl_u.nfs4_fl.owner;
6768 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6769 	arg.lock_owner.s_dev = server->s_dev;
6770 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6771 	switch (status) {
6772 		case 0:
6773 			request->fl_type = F_UNLCK;
6774 			break;
6775 		case -NFS4ERR_DENIED:
6776 			status = 0;
6777 	}
6778 	request->fl_ops->fl_release_private(request);
6779 	request->fl_ops = NULL;
6780 out:
6781 	return status;
6782 }
6783 
6784 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6785 {
6786 	struct nfs4_exception exception = {
6787 		.interruptible = true,
6788 	};
6789 	int err;
6790 
6791 	do {
6792 		err = _nfs4_proc_getlk(state, cmd, request);
6793 		trace_nfs4_get_lock(request, state, cmd, err);
6794 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6795 				&exception);
6796 	} while (exception.retry);
6797 	return err;
6798 }
6799 
6800 /*
6801  * Update the seqid of a lock stateid after receiving
6802  * NFS4ERR_OLD_STATEID
6803  */
6804 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6805 		struct nfs4_lock_state *lsp)
6806 {
6807 	struct nfs4_state *state = lsp->ls_state;
6808 	bool ret = false;
6809 
6810 	spin_lock(&state->state_lock);
6811 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6812 		goto out;
6813 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6814 		nfs4_stateid_seqid_inc(dst);
6815 	else
6816 		dst->seqid = lsp->ls_stateid.seqid;
6817 	ret = true;
6818 out:
6819 	spin_unlock(&state->state_lock);
6820 	return ret;
6821 }
6822 
6823 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6824 		struct nfs4_lock_state *lsp)
6825 {
6826 	struct nfs4_state *state = lsp->ls_state;
6827 	bool ret;
6828 
6829 	spin_lock(&state->state_lock);
6830 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6831 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6832 	spin_unlock(&state->state_lock);
6833 	return ret;
6834 }
6835 
6836 struct nfs4_unlockdata {
6837 	struct nfs_locku_args arg;
6838 	struct nfs_locku_res res;
6839 	struct nfs4_lock_state *lsp;
6840 	struct nfs_open_context *ctx;
6841 	struct nfs_lock_context *l_ctx;
6842 	struct file_lock fl;
6843 	struct nfs_server *server;
6844 	unsigned long timestamp;
6845 };
6846 
6847 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6848 		struct nfs_open_context *ctx,
6849 		struct nfs4_lock_state *lsp,
6850 		struct nfs_seqid *seqid)
6851 {
6852 	struct nfs4_unlockdata *p;
6853 	struct nfs4_state *state = lsp->ls_state;
6854 	struct inode *inode = state->inode;
6855 
6856 	p = kzalloc(sizeof(*p), GFP_KERNEL);
6857 	if (p == NULL)
6858 		return NULL;
6859 	p->arg.fh = NFS_FH(inode);
6860 	p->arg.fl = &p->fl;
6861 	p->arg.seqid = seqid;
6862 	p->res.seqid = seqid;
6863 	p->lsp = lsp;
6864 	/* Ensure we don't close file until we're done freeing locks! */
6865 	p->ctx = get_nfs_open_context(ctx);
6866 	p->l_ctx = nfs_get_lock_context(ctx);
6867 	locks_init_lock(&p->fl);
6868 	locks_copy_lock(&p->fl, fl);
6869 	p->server = NFS_SERVER(inode);
6870 	spin_lock(&state->state_lock);
6871 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6872 	spin_unlock(&state->state_lock);
6873 	return p;
6874 }
6875 
6876 static void nfs4_locku_release_calldata(void *data)
6877 {
6878 	struct nfs4_unlockdata *calldata = data;
6879 	nfs_free_seqid(calldata->arg.seqid);
6880 	nfs4_put_lock_state(calldata->lsp);
6881 	nfs_put_lock_context(calldata->l_ctx);
6882 	put_nfs_open_context(calldata->ctx);
6883 	kfree(calldata);
6884 }
6885 
6886 static void nfs4_locku_done(struct rpc_task *task, void *data)
6887 {
6888 	struct nfs4_unlockdata *calldata = data;
6889 	struct nfs4_exception exception = {
6890 		.inode = calldata->lsp->ls_state->inode,
6891 		.stateid = &calldata->arg.stateid,
6892 	};
6893 
6894 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6895 		return;
6896 	switch (task->tk_status) {
6897 		case 0:
6898 			renew_lease(calldata->server, calldata->timestamp);
6899 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6900 			if (nfs4_update_lock_stateid(calldata->lsp,
6901 					&calldata->res.stateid))
6902 				break;
6903 			fallthrough;
6904 		case -NFS4ERR_ADMIN_REVOKED:
6905 		case -NFS4ERR_EXPIRED:
6906 			nfs4_free_revoked_stateid(calldata->server,
6907 					&calldata->arg.stateid,
6908 					task->tk_msg.rpc_cred);
6909 			fallthrough;
6910 		case -NFS4ERR_BAD_STATEID:
6911 		case -NFS4ERR_STALE_STATEID:
6912 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6913 						calldata->lsp))
6914 				rpc_restart_call_prepare(task);
6915 			break;
6916 		case -NFS4ERR_OLD_STATEID:
6917 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6918 						calldata->lsp))
6919 				rpc_restart_call_prepare(task);
6920 			break;
6921 		default:
6922 			task->tk_status = nfs4_async_handle_exception(task,
6923 					calldata->server, task->tk_status,
6924 					&exception);
6925 			if (exception.retry)
6926 				rpc_restart_call_prepare(task);
6927 	}
6928 	nfs_release_seqid(calldata->arg.seqid);
6929 }
6930 
6931 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6932 {
6933 	struct nfs4_unlockdata *calldata = data;
6934 
6935 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6936 		nfs_async_iocounter_wait(task, calldata->l_ctx))
6937 		return;
6938 
6939 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6940 		goto out_wait;
6941 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6942 		/* Note: exit _without_ running nfs4_locku_done */
6943 		goto out_no_action;
6944 	}
6945 	calldata->timestamp = jiffies;
6946 	if (nfs4_setup_sequence(calldata->server->nfs_client,
6947 				&calldata->arg.seq_args,
6948 				&calldata->res.seq_res,
6949 				task) != 0)
6950 		nfs_release_seqid(calldata->arg.seqid);
6951 	return;
6952 out_no_action:
6953 	task->tk_action = NULL;
6954 out_wait:
6955 	nfs4_sequence_done(task, &calldata->res.seq_res);
6956 }
6957 
6958 static const struct rpc_call_ops nfs4_locku_ops = {
6959 	.rpc_call_prepare = nfs4_locku_prepare,
6960 	.rpc_call_done = nfs4_locku_done,
6961 	.rpc_release = nfs4_locku_release_calldata,
6962 };
6963 
6964 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6965 		struct nfs_open_context *ctx,
6966 		struct nfs4_lock_state *lsp,
6967 		struct nfs_seqid *seqid)
6968 {
6969 	struct nfs4_unlockdata *data;
6970 	struct rpc_message msg = {
6971 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6972 		.rpc_cred = ctx->cred,
6973 	};
6974 	struct rpc_task_setup task_setup_data = {
6975 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6976 		.rpc_message = &msg,
6977 		.callback_ops = &nfs4_locku_ops,
6978 		.workqueue = nfsiod_workqueue,
6979 		.flags = RPC_TASK_ASYNC,
6980 	};
6981 
6982 	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6983 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6984 
6985 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6986 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6987 
6988 	/* Ensure this is an unlock - when canceling a lock, the
6989 	 * canceled lock is passed in, and it won't be an unlock.
6990 	 */
6991 	fl->fl_type = F_UNLCK;
6992 	if (fl->fl_flags & FL_CLOSE)
6993 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6994 
6995 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6996 	if (data == NULL) {
6997 		nfs_free_seqid(seqid);
6998 		return ERR_PTR(-ENOMEM);
6999 	}
7000 
7001 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
7002 	msg.rpc_argp = &data->arg;
7003 	msg.rpc_resp = &data->res;
7004 	task_setup_data.callback_data = data;
7005 	return rpc_run_task(&task_setup_data);
7006 }
7007 
7008 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7009 {
7010 	struct inode *inode = state->inode;
7011 	struct nfs4_state_owner *sp = state->owner;
7012 	struct nfs_inode *nfsi = NFS_I(inode);
7013 	struct nfs_seqid *seqid;
7014 	struct nfs4_lock_state *lsp;
7015 	struct rpc_task *task;
7016 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7017 	int status = 0;
7018 	unsigned char fl_flags = request->fl_flags;
7019 
7020 	status = nfs4_set_lock_state(state, request);
7021 	/* Unlock _before_ we do the RPC call */
7022 	request->fl_flags |= FL_EXISTS;
7023 	/* Exclude nfs_delegation_claim_locks() */
7024 	mutex_lock(&sp->so_delegreturn_mutex);
7025 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7026 	down_read(&nfsi->rwsem);
7027 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7028 		up_read(&nfsi->rwsem);
7029 		mutex_unlock(&sp->so_delegreturn_mutex);
7030 		goto out;
7031 	}
7032 	lsp = request->fl_u.nfs4_fl.owner;
7033 	set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7034 	up_read(&nfsi->rwsem);
7035 	mutex_unlock(&sp->so_delegreturn_mutex);
7036 	if (status != 0)
7037 		goto out;
7038 	/* Is this a delegated lock? */
7039 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7040 		goto out;
7041 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7042 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7043 	status = -ENOMEM;
7044 	if (IS_ERR(seqid))
7045 		goto out;
7046 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7047 	status = PTR_ERR(task);
7048 	if (IS_ERR(task))
7049 		goto out;
7050 	status = rpc_wait_for_completion_task(task);
7051 	rpc_put_task(task);
7052 out:
7053 	request->fl_flags = fl_flags;
7054 	trace_nfs4_unlock(request, state, F_SETLK, status);
7055 	return status;
7056 }
7057 
7058 struct nfs4_lockdata {
7059 	struct nfs_lock_args arg;
7060 	struct nfs_lock_res res;
7061 	struct nfs4_lock_state *lsp;
7062 	struct nfs_open_context *ctx;
7063 	struct file_lock fl;
7064 	unsigned long timestamp;
7065 	int rpc_status;
7066 	int cancelled;
7067 	struct nfs_server *server;
7068 };
7069 
7070 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7071 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7072 		gfp_t gfp_mask)
7073 {
7074 	struct nfs4_lockdata *p;
7075 	struct inode *inode = lsp->ls_state->inode;
7076 	struct nfs_server *server = NFS_SERVER(inode);
7077 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7078 
7079 	p = kzalloc(sizeof(*p), gfp_mask);
7080 	if (p == NULL)
7081 		return NULL;
7082 
7083 	p->arg.fh = NFS_FH(inode);
7084 	p->arg.fl = &p->fl;
7085 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7086 	if (IS_ERR(p->arg.open_seqid))
7087 		goto out_free;
7088 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7089 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7090 	if (IS_ERR(p->arg.lock_seqid))
7091 		goto out_free_seqid;
7092 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7093 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7094 	p->arg.lock_owner.s_dev = server->s_dev;
7095 	p->res.lock_seqid = p->arg.lock_seqid;
7096 	p->lsp = lsp;
7097 	p->server = server;
7098 	p->ctx = get_nfs_open_context(ctx);
7099 	locks_init_lock(&p->fl);
7100 	locks_copy_lock(&p->fl, fl);
7101 	return p;
7102 out_free_seqid:
7103 	nfs_free_seqid(p->arg.open_seqid);
7104 out_free:
7105 	kfree(p);
7106 	return NULL;
7107 }
7108 
7109 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7110 {
7111 	struct nfs4_lockdata *data = calldata;
7112 	struct nfs4_state *state = data->lsp->ls_state;
7113 
7114 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7115 		goto out_wait;
7116 	/* Do we need to do an open_to_lock_owner? */
7117 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7118 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7119 			goto out_release_lock_seqid;
7120 		}
7121 		nfs4_stateid_copy(&data->arg.open_stateid,
7122 				&state->open_stateid);
7123 		data->arg.new_lock_owner = 1;
7124 		data->res.open_seqid = data->arg.open_seqid;
7125 	} else {
7126 		data->arg.new_lock_owner = 0;
7127 		nfs4_stateid_copy(&data->arg.lock_stateid,
7128 				&data->lsp->ls_stateid);
7129 	}
7130 	if (!nfs4_valid_open_stateid(state)) {
7131 		data->rpc_status = -EBADF;
7132 		task->tk_action = NULL;
7133 		goto out_release_open_seqid;
7134 	}
7135 	data->timestamp = jiffies;
7136 	if (nfs4_setup_sequence(data->server->nfs_client,
7137 				&data->arg.seq_args,
7138 				&data->res.seq_res,
7139 				task) == 0)
7140 		return;
7141 out_release_open_seqid:
7142 	nfs_release_seqid(data->arg.open_seqid);
7143 out_release_lock_seqid:
7144 	nfs_release_seqid(data->arg.lock_seqid);
7145 out_wait:
7146 	nfs4_sequence_done(task, &data->res.seq_res);
7147 	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7148 }
7149 
7150 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7151 {
7152 	struct nfs4_lockdata *data = calldata;
7153 	struct nfs4_lock_state *lsp = data->lsp;
7154 	struct nfs_server *server = NFS_SERVER(d_inode(data->ctx->dentry));
7155 
7156 	if (!nfs4_sequence_done(task, &data->res.seq_res))
7157 		return;
7158 
7159 	data->rpc_status = task->tk_status;
7160 	switch (task->tk_status) {
7161 	case 0:
7162 		renew_lease(server, data->timestamp);
7163 		if (data->arg.new_lock && !data->cancelled) {
7164 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7165 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7166 				goto out_restart;
7167 		}
7168 		if (data->arg.new_lock_owner != 0) {
7169 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7170 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7171 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7172 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7173 			goto out_restart;
7174 		break;
7175 	case -NFS4ERR_BAD_STATEID:
7176 	case -NFS4ERR_OLD_STATEID:
7177 	case -NFS4ERR_STALE_STATEID:
7178 	case -NFS4ERR_EXPIRED:
7179 		if (data->arg.new_lock_owner != 0) {
7180 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7181 						&lsp->ls_state->open_stateid))
7182 				goto out_restart;
7183 			else if (nfs4_async_handle_error(task, server, lsp->ls_state, NULL) == -EAGAIN)
7184 				goto out_restart;
7185 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7186 						&lsp->ls_stateid))
7187 				goto out_restart;
7188 	}
7189 out_done:
7190 	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7191 	return;
7192 out_restart:
7193 	if (!data->cancelled)
7194 		rpc_restart_call_prepare(task);
7195 	goto out_done;
7196 }
7197 
7198 static void nfs4_lock_release(void *calldata)
7199 {
7200 	struct nfs4_lockdata *data = calldata;
7201 
7202 	nfs_free_seqid(data->arg.open_seqid);
7203 	if (data->cancelled && data->rpc_status == 0) {
7204 		struct rpc_task *task;
7205 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7206 				data->arg.lock_seqid);
7207 		if (!IS_ERR(task))
7208 			rpc_put_task_async(task);
7209 		dprintk("%s: cancelling lock!\n", __func__);
7210 	} else
7211 		nfs_free_seqid(data->arg.lock_seqid);
7212 	nfs4_put_lock_state(data->lsp);
7213 	put_nfs_open_context(data->ctx);
7214 	kfree(data);
7215 }
7216 
7217 static const struct rpc_call_ops nfs4_lock_ops = {
7218 	.rpc_call_prepare = nfs4_lock_prepare,
7219 	.rpc_call_done = nfs4_lock_done,
7220 	.rpc_release = nfs4_lock_release,
7221 };
7222 
7223 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7224 {
7225 	switch (error) {
7226 	case -NFS4ERR_ADMIN_REVOKED:
7227 	case -NFS4ERR_EXPIRED:
7228 	case -NFS4ERR_BAD_STATEID:
7229 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7230 		if (new_lock_owner != 0 ||
7231 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7232 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7233 		break;
7234 	case -NFS4ERR_STALE_STATEID:
7235 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7236 		nfs4_schedule_lease_recovery(server->nfs_client);
7237 	}
7238 }
7239 
7240 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7241 {
7242 	struct nfs4_lockdata *data;
7243 	struct rpc_task *task;
7244 	struct rpc_message msg = {
7245 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7246 		.rpc_cred = state->owner->so_cred,
7247 	};
7248 	struct rpc_task_setup task_setup_data = {
7249 		.rpc_client = NFS_CLIENT(state->inode),
7250 		.rpc_message = &msg,
7251 		.callback_ops = &nfs4_lock_ops,
7252 		.workqueue = nfsiod_workqueue,
7253 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7254 	};
7255 	int ret;
7256 
7257 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7258 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7259 
7260 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7261 				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7262 	if (data == NULL)
7263 		return -ENOMEM;
7264 	if (IS_SETLKW(cmd))
7265 		data->arg.block = 1;
7266 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7267 				recovery_type > NFS_LOCK_NEW);
7268 	msg.rpc_argp = &data->arg;
7269 	msg.rpc_resp = &data->res;
7270 	task_setup_data.callback_data = data;
7271 	if (recovery_type > NFS_LOCK_NEW) {
7272 		if (recovery_type == NFS_LOCK_RECLAIM)
7273 			data->arg.reclaim = NFS_LOCK_RECLAIM;
7274 	} else
7275 		data->arg.new_lock = 1;
7276 	task = rpc_run_task(&task_setup_data);
7277 	if (IS_ERR(task))
7278 		return PTR_ERR(task);
7279 	ret = rpc_wait_for_completion_task(task);
7280 	if (ret == 0) {
7281 		ret = data->rpc_status;
7282 		if (ret)
7283 			nfs4_handle_setlk_error(data->server, data->lsp,
7284 					data->arg.new_lock_owner, ret);
7285 	} else
7286 		data->cancelled = true;
7287 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7288 	rpc_put_task(task);
7289 	dprintk("%s: ret = %d\n", __func__, ret);
7290 	return ret;
7291 }
7292 
7293 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7294 {
7295 	struct nfs_server *server = NFS_SERVER(state->inode);
7296 	struct nfs4_exception exception = {
7297 		.inode = state->inode,
7298 	};
7299 	int err;
7300 
7301 	do {
7302 		/* Cache the lock if possible... */
7303 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7304 			return 0;
7305 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7306 		if (err != -NFS4ERR_DELAY)
7307 			break;
7308 		nfs4_handle_exception(server, err, &exception);
7309 	} while (exception.retry);
7310 	return err;
7311 }
7312 
7313 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7314 {
7315 	struct nfs_server *server = NFS_SERVER(state->inode);
7316 	struct nfs4_exception exception = {
7317 		.inode = state->inode,
7318 	};
7319 	int err;
7320 
7321 	err = nfs4_set_lock_state(state, request);
7322 	if (err != 0)
7323 		return err;
7324 	if (!recover_lost_locks) {
7325 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7326 		return 0;
7327 	}
7328 	do {
7329 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7330 			return 0;
7331 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7332 		switch (err) {
7333 		default:
7334 			goto out;
7335 		case -NFS4ERR_GRACE:
7336 		case -NFS4ERR_DELAY:
7337 			nfs4_handle_exception(server, err, &exception);
7338 			err = 0;
7339 		}
7340 	} while (exception.retry);
7341 out:
7342 	return err;
7343 }
7344 
7345 #if defined(CONFIG_NFS_V4_1)
7346 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7347 {
7348 	struct nfs4_lock_state *lsp;
7349 	int status;
7350 
7351 	status = nfs4_set_lock_state(state, request);
7352 	if (status != 0)
7353 		return status;
7354 	lsp = request->fl_u.nfs4_fl.owner;
7355 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7356 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7357 		return 0;
7358 	return nfs4_lock_expired(state, request);
7359 }
7360 #endif
7361 
7362 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7363 {
7364 	struct nfs_inode *nfsi = NFS_I(state->inode);
7365 	struct nfs4_state_owner *sp = state->owner;
7366 	unsigned char fl_flags = request->fl_flags;
7367 	int status;
7368 
7369 	request->fl_flags |= FL_ACCESS;
7370 	status = locks_lock_inode_wait(state->inode, request);
7371 	if (status < 0)
7372 		goto out;
7373 	mutex_lock(&sp->so_delegreturn_mutex);
7374 	down_read(&nfsi->rwsem);
7375 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7376 		/* Yes: cache locks! */
7377 		/* ...but avoid races with delegation recall... */
7378 		request->fl_flags = fl_flags & ~FL_SLEEP;
7379 		status = locks_lock_inode_wait(state->inode, request);
7380 		up_read(&nfsi->rwsem);
7381 		mutex_unlock(&sp->so_delegreturn_mutex);
7382 		goto out;
7383 	}
7384 	up_read(&nfsi->rwsem);
7385 	mutex_unlock(&sp->so_delegreturn_mutex);
7386 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7387 out:
7388 	request->fl_flags = fl_flags;
7389 	return status;
7390 }
7391 
7392 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7393 {
7394 	struct nfs4_exception exception = {
7395 		.state = state,
7396 		.inode = state->inode,
7397 		.interruptible = true,
7398 	};
7399 	int err;
7400 
7401 	do {
7402 		err = _nfs4_proc_setlk(state, cmd, request);
7403 		if (err == -NFS4ERR_DENIED)
7404 			err = -EAGAIN;
7405 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7406 				err, &exception);
7407 	} while (exception.retry);
7408 	return err;
7409 }
7410 
7411 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7412 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7413 
7414 static int
7415 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7416 			struct file_lock *request)
7417 {
7418 	int		status = -ERESTARTSYS;
7419 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7420 
7421 	while(!signalled()) {
7422 		status = nfs4_proc_setlk(state, cmd, request);
7423 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7424 			break;
7425 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7426 		schedule_timeout(timeout);
7427 		timeout *= 2;
7428 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7429 		status = -ERESTARTSYS;
7430 	}
7431 	return status;
7432 }
7433 
7434 #ifdef CONFIG_NFS_V4_1
7435 struct nfs4_lock_waiter {
7436 	struct inode		*inode;
7437 	struct nfs_lowner	owner;
7438 	wait_queue_entry_t	wait;
7439 };
7440 
7441 static int
7442 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7443 {
7444 	struct nfs4_lock_waiter	*waiter	=
7445 		container_of(wait, struct nfs4_lock_waiter, wait);
7446 
7447 	/* NULL key means to wake up everyone */
7448 	if (key) {
7449 		struct cb_notify_lock_args	*cbnl = key;
7450 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7451 						*wowner = &waiter->owner;
7452 
7453 		/* Only wake if the callback was for the same owner. */
7454 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7455 			return 0;
7456 
7457 		/* Make sure it's for the right inode */
7458 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7459 			return 0;
7460 	}
7461 
7462 	return woken_wake_function(wait, mode, flags, key);
7463 }
7464 
7465 static int
7466 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7467 {
7468 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7469 	struct nfs_server *server = NFS_SERVER(state->inode);
7470 	struct nfs_client *clp = server->nfs_client;
7471 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7472 	struct nfs4_lock_waiter waiter = {
7473 		.inode = state->inode,
7474 		.owner = { .clientid = clp->cl_clientid,
7475 			   .id = lsp->ls_seqid.owner_id,
7476 			   .s_dev = server->s_dev },
7477 	};
7478 	int status;
7479 
7480 	/* Don't bother with waitqueue if we don't expect a callback */
7481 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7482 		return nfs4_retry_setlk_simple(state, cmd, request);
7483 
7484 	init_wait(&waiter.wait);
7485 	waiter.wait.func = nfs4_wake_lock_waiter;
7486 	add_wait_queue(q, &waiter.wait);
7487 
7488 	do {
7489 		status = nfs4_proc_setlk(state, cmd, request);
7490 		if (status != -EAGAIN || IS_SETLK(cmd))
7491 			break;
7492 
7493 		status = -ERESTARTSYS;
7494 		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7495 			   NFS4_LOCK_MAXTIMEOUT);
7496 	} while (!signalled());
7497 
7498 	remove_wait_queue(q, &waiter.wait);
7499 
7500 	return status;
7501 }
7502 #else /* !CONFIG_NFS_V4_1 */
7503 static inline int
7504 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7505 {
7506 	return nfs4_retry_setlk_simple(state, cmd, request);
7507 }
7508 #endif
7509 
7510 static int
7511 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7512 {
7513 	struct nfs_open_context *ctx;
7514 	struct nfs4_state *state;
7515 	int status;
7516 
7517 	/* verify open state */
7518 	ctx = nfs_file_open_context(filp);
7519 	state = ctx->state;
7520 
7521 	if (IS_GETLK(cmd)) {
7522 		if (state != NULL)
7523 			return nfs4_proc_getlk(state, F_GETLK, request);
7524 		return 0;
7525 	}
7526 
7527 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7528 		return -EINVAL;
7529 
7530 	if (request->fl_type == F_UNLCK) {
7531 		if (state != NULL)
7532 			return nfs4_proc_unlck(state, cmd, request);
7533 		return 0;
7534 	}
7535 
7536 	if (state == NULL)
7537 		return -ENOLCK;
7538 
7539 	if ((request->fl_flags & FL_POSIX) &&
7540 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7541 		return -ENOLCK;
7542 
7543 	/*
7544 	 * Don't rely on the VFS having checked the file open mode,
7545 	 * since it won't do this for flock() locks.
7546 	 */
7547 	switch (request->fl_type) {
7548 	case F_RDLCK:
7549 		if (!(filp->f_mode & FMODE_READ))
7550 			return -EBADF;
7551 		break;
7552 	case F_WRLCK:
7553 		if (!(filp->f_mode & FMODE_WRITE))
7554 			return -EBADF;
7555 	}
7556 
7557 	status = nfs4_set_lock_state(state, request);
7558 	if (status != 0)
7559 		return status;
7560 
7561 	return nfs4_retry_setlk(state, cmd, request);
7562 }
7563 
7564 static int nfs4_delete_lease(struct file *file, void **priv)
7565 {
7566 	return generic_setlease(file, F_UNLCK, NULL, priv);
7567 }
7568 
7569 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7570 			  void **priv)
7571 {
7572 	struct inode *inode = file_inode(file);
7573 	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7574 	int ret;
7575 
7576 	/* No delegation, no lease */
7577 	if (!nfs4_have_delegation(inode, type))
7578 		return -EAGAIN;
7579 	ret = generic_setlease(file, arg, lease, priv);
7580 	if (ret || nfs4_have_delegation(inode, type))
7581 		return ret;
7582 	/* We raced with a delegation return */
7583 	nfs4_delete_lease(file, priv);
7584 	return -EAGAIN;
7585 }
7586 
7587 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7588 		       void **priv)
7589 {
7590 	switch (arg) {
7591 	case F_RDLCK:
7592 	case F_WRLCK:
7593 		return nfs4_add_lease(file, arg, lease, priv);
7594 	case F_UNLCK:
7595 		return nfs4_delete_lease(file, priv);
7596 	default:
7597 		return -EINVAL;
7598 	}
7599 }
7600 
7601 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7602 {
7603 	struct nfs_server *server = NFS_SERVER(state->inode);
7604 	int err;
7605 
7606 	err = nfs4_set_lock_state(state, fl);
7607 	if (err != 0)
7608 		return err;
7609 	do {
7610 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7611 		if (err != -NFS4ERR_DELAY)
7612 			break;
7613 		ssleep(1);
7614 	} while (err == -NFS4ERR_DELAY);
7615 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7616 }
7617 
7618 struct nfs_release_lockowner_data {
7619 	struct nfs4_lock_state *lsp;
7620 	struct nfs_server *server;
7621 	struct nfs_release_lockowner_args args;
7622 	struct nfs_release_lockowner_res res;
7623 	unsigned long timestamp;
7624 };
7625 
7626 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7627 {
7628 	struct nfs_release_lockowner_data *data = calldata;
7629 	struct nfs_server *server = data->server;
7630 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7631 			   &data->res.seq_res, task);
7632 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7633 	data->timestamp = jiffies;
7634 }
7635 
7636 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7637 {
7638 	struct nfs_release_lockowner_data *data = calldata;
7639 	struct nfs_server *server = data->server;
7640 
7641 	nfs40_sequence_done(task, &data->res.seq_res);
7642 
7643 	switch (task->tk_status) {
7644 	case 0:
7645 		renew_lease(server, data->timestamp);
7646 		break;
7647 	case -NFS4ERR_STALE_CLIENTID:
7648 	case -NFS4ERR_EXPIRED:
7649 		nfs4_schedule_lease_recovery(server->nfs_client);
7650 		break;
7651 	case -NFS4ERR_LEASE_MOVED:
7652 	case -NFS4ERR_DELAY:
7653 		if (nfs4_async_handle_error(task, server,
7654 					    NULL, NULL) == -EAGAIN)
7655 			rpc_restart_call_prepare(task);
7656 	}
7657 }
7658 
7659 static void nfs4_release_lockowner_release(void *calldata)
7660 {
7661 	struct nfs_release_lockowner_data *data = calldata;
7662 	nfs4_free_lock_state(data->server, data->lsp);
7663 	kfree(calldata);
7664 }
7665 
7666 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7667 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7668 	.rpc_call_done = nfs4_release_lockowner_done,
7669 	.rpc_release = nfs4_release_lockowner_release,
7670 };
7671 
7672 static void
7673 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7674 {
7675 	struct nfs_release_lockowner_data *data;
7676 	struct rpc_message msg = {
7677 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7678 	};
7679 
7680 	if (server->nfs_client->cl_mvops->minor_version != 0)
7681 		return;
7682 
7683 	data = kmalloc(sizeof(*data), GFP_KERNEL);
7684 	if (!data)
7685 		return;
7686 	data->lsp = lsp;
7687 	data->server = server;
7688 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7689 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7690 	data->args.lock_owner.s_dev = server->s_dev;
7691 
7692 	msg.rpc_argp = &data->args;
7693 	msg.rpc_resp = &data->res;
7694 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7695 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7696 }
7697 
7698 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7699 
7700 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7701 				   struct mnt_idmap *idmap,
7702 				   struct dentry *unused, struct inode *inode,
7703 				   const char *key, const void *buf,
7704 				   size_t buflen, int flags)
7705 {
7706 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7707 }
7708 
7709 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7710 				   struct dentry *unused, struct inode *inode,
7711 				   const char *key, void *buf, size_t buflen)
7712 {
7713 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7714 }
7715 
7716 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7717 {
7718 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7719 }
7720 
7721 #if defined(CONFIG_NFS_V4_1)
7722 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7723 
7724 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7725 				    struct mnt_idmap *idmap,
7726 				    struct dentry *unused, struct inode *inode,
7727 				    const char *key, const void *buf,
7728 				    size_t buflen, int flags)
7729 {
7730 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7731 }
7732 
7733 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7734 				    struct dentry *unused, struct inode *inode,
7735 				    const char *key, void *buf, size_t buflen)
7736 {
7737 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7738 }
7739 
7740 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7741 {
7742 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7743 }
7744 
7745 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7746 
7747 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7748 				    struct mnt_idmap *idmap,
7749 				    struct dentry *unused, struct inode *inode,
7750 				    const char *key, const void *buf,
7751 				    size_t buflen, int flags)
7752 {
7753 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7754 }
7755 
7756 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7757 				    struct dentry *unused, struct inode *inode,
7758 				    const char *key, void *buf, size_t buflen)
7759 {
7760 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7761 }
7762 
7763 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7764 {
7765 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7766 }
7767 
7768 #endif
7769 
7770 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7771 
7772 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7773 				     struct mnt_idmap *idmap,
7774 				     struct dentry *unused, struct inode *inode,
7775 				     const char *key, const void *buf,
7776 				     size_t buflen, int flags)
7777 {
7778 	if (security_ismaclabel(key))
7779 		return nfs4_set_security_label(inode, buf, buflen);
7780 
7781 	return -EOPNOTSUPP;
7782 }
7783 
7784 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7785 				     struct dentry *unused, struct inode *inode,
7786 				     const char *key, void *buf, size_t buflen)
7787 {
7788 	if (security_ismaclabel(key))
7789 		return nfs4_get_security_label(inode, buf, buflen);
7790 	return -EOPNOTSUPP;
7791 }
7792 
7793 static ssize_t
7794 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7795 {
7796 	int len = 0;
7797 
7798 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7799 		len = security_inode_listsecurity(inode, list, list_len);
7800 		if (len >= 0 && list_len && len > list_len)
7801 			return -ERANGE;
7802 	}
7803 	return len;
7804 }
7805 
7806 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7807 	.prefix = XATTR_SECURITY_PREFIX,
7808 	.get	= nfs4_xattr_get_nfs4_label,
7809 	.set	= nfs4_xattr_set_nfs4_label,
7810 };
7811 
7812 #else
7813 
7814 static ssize_t
7815 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7816 {
7817 	return 0;
7818 }
7819 
7820 #endif
7821 
7822 #ifdef CONFIG_NFS_V4_2
7823 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7824 				    struct mnt_idmap *idmap,
7825 				    struct dentry *unused, struct inode *inode,
7826 				    const char *key, const void *buf,
7827 				    size_t buflen, int flags)
7828 {
7829 	u32 mask;
7830 	int ret;
7831 
7832 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7833 		return -EOPNOTSUPP;
7834 
7835 	/*
7836 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7837 	 * flags right now. Handling of xattr operations use the normal
7838 	 * file read/write permissions.
7839 	 *
7840 	 * Just in case the server has other ideas (which RFC 8276 allows),
7841 	 * do a cached access check for the XA* flags to possibly avoid
7842 	 * doing an RPC and getting EACCES back.
7843 	 */
7844 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7845 		if (!(mask & NFS_ACCESS_XAWRITE))
7846 			return -EACCES;
7847 	}
7848 
7849 	if (buf == NULL) {
7850 		ret = nfs42_proc_removexattr(inode, key);
7851 		if (!ret)
7852 			nfs4_xattr_cache_remove(inode, key);
7853 	} else {
7854 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7855 		if (!ret)
7856 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7857 	}
7858 
7859 	return ret;
7860 }
7861 
7862 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7863 				    struct dentry *unused, struct inode *inode,
7864 				    const char *key, void *buf, size_t buflen)
7865 {
7866 	u32 mask;
7867 	ssize_t ret;
7868 
7869 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7870 		return -EOPNOTSUPP;
7871 
7872 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7873 		if (!(mask & NFS_ACCESS_XAREAD))
7874 			return -EACCES;
7875 	}
7876 
7877 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7878 	if (ret)
7879 		return ret;
7880 
7881 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7882 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7883 		return ret;
7884 
7885 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7886 
7887 	return ret;
7888 }
7889 
7890 static ssize_t
7891 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7892 {
7893 	u64 cookie;
7894 	bool eof;
7895 	ssize_t ret, size;
7896 	char *buf;
7897 	size_t buflen;
7898 	u32 mask;
7899 
7900 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7901 		return 0;
7902 
7903 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7904 		if (!(mask & NFS_ACCESS_XALIST))
7905 			return 0;
7906 	}
7907 
7908 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7909 	if (ret)
7910 		return ret;
7911 
7912 	ret = nfs4_xattr_cache_list(inode, list, list_len);
7913 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7914 		return ret;
7915 
7916 	cookie = 0;
7917 	eof = false;
7918 	buflen = list_len ? list_len : XATTR_LIST_MAX;
7919 	buf = list_len ? list : NULL;
7920 	size = 0;
7921 
7922 	while (!eof) {
7923 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
7924 		    &cookie, &eof);
7925 		if (ret < 0)
7926 			return ret;
7927 
7928 		if (list_len) {
7929 			buf += ret;
7930 			buflen -= ret;
7931 		}
7932 		size += ret;
7933 	}
7934 
7935 	if (list_len)
7936 		nfs4_xattr_cache_set_list(inode, list, size);
7937 
7938 	return size;
7939 }
7940 
7941 #else
7942 
7943 static ssize_t
7944 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7945 {
7946 	return 0;
7947 }
7948 #endif /* CONFIG_NFS_V4_2 */
7949 
7950 /*
7951  * nfs_fhget will use either the mounted_on_fileid or the fileid
7952  */
7953 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7954 {
7955 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7956 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7957 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7958 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7959 		return;
7960 
7961 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7962 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7963 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7964 	fattr->nlink = 2;
7965 }
7966 
7967 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7968 				   const struct qstr *name,
7969 				   struct nfs4_fs_locations *fs_locations,
7970 				   struct page *page)
7971 {
7972 	struct nfs_server *server = NFS_SERVER(dir);
7973 	u32 bitmask[3];
7974 	struct nfs4_fs_locations_arg args = {
7975 		.dir_fh = NFS_FH(dir),
7976 		.name = name,
7977 		.page = page,
7978 		.bitmask = bitmask,
7979 	};
7980 	struct nfs4_fs_locations_res res = {
7981 		.fs_locations = fs_locations,
7982 	};
7983 	struct rpc_message msg = {
7984 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7985 		.rpc_argp = &args,
7986 		.rpc_resp = &res,
7987 	};
7988 	int status;
7989 
7990 	dprintk("%s: start\n", __func__);
7991 
7992 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7993 	bitmask[1] = nfs4_fattr_bitmap[1];
7994 
7995 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
7996 	 * is not supported */
7997 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7998 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
7999 	else
8000 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8001 
8002 	nfs_fattr_init(fs_locations->fattr);
8003 	fs_locations->server = server;
8004 	fs_locations->nlocations = 0;
8005 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8006 	dprintk("%s: returned status = %d\n", __func__, status);
8007 	return status;
8008 }
8009 
8010 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8011 			   const struct qstr *name,
8012 			   struct nfs4_fs_locations *fs_locations,
8013 			   struct page *page)
8014 {
8015 	struct nfs4_exception exception = {
8016 		.interruptible = true,
8017 	};
8018 	int err;
8019 	do {
8020 		err = _nfs4_proc_fs_locations(client, dir, name,
8021 				fs_locations, page);
8022 		trace_nfs4_get_fs_locations(dir, name, err);
8023 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8024 				&exception);
8025 	} while (exception.retry);
8026 	return err;
8027 }
8028 
8029 /*
8030  * This operation also signals the server that this client is
8031  * performing migration recovery.  The server can stop returning
8032  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8033  * appended to this compound to identify the client ID which is
8034  * performing recovery.
8035  */
8036 static int _nfs40_proc_get_locations(struct nfs_server *server,
8037 				     struct nfs_fh *fhandle,
8038 				     struct nfs4_fs_locations *locations,
8039 				     struct page *page, const struct cred *cred)
8040 {
8041 	struct rpc_clnt *clnt = server->client;
8042 	u32 bitmask[2] = {
8043 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8044 	};
8045 	struct nfs4_fs_locations_arg args = {
8046 		.clientid	= server->nfs_client->cl_clientid,
8047 		.fh		= fhandle,
8048 		.page		= page,
8049 		.bitmask	= bitmask,
8050 		.migration	= 1,		/* skip LOOKUP */
8051 		.renew		= 1,		/* append RENEW */
8052 	};
8053 	struct nfs4_fs_locations_res res = {
8054 		.fs_locations	= locations,
8055 		.migration	= 1,
8056 		.renew		= 1,
8057 	};
8058 	struct rpc_message msg = {
8059 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8060 		.rpc_argp	= &args,
8061 		.rpc_resp	= &res,
8062 		.rpc_cred	= cred,
8063 	};
8064 	unsigned long now = jiffies;
8065 	int status;
8066 
8067 	nfs_fattr_init(locations->fattr);
8068 	locations->server = server;
8069 	locations->nlocations = 0;
8070 
8071 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8072 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8073 					&args.seq_args, &res.seq_res);
8074 	if (status)
8075 		return status;
8076 
8077 	renew_lease(server, now);
8078 	return 0;
8079 }
8080 
8081 #ifdef CONFIG_NFS_V4_1
8082 
8083 /*
8084  * This operation also signals the server that this client is
8085  * performing migration recovery.  The server can stop asserting
8086  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8087  * performing this operation is identified in the SEQUENCE
8088  * operation in this compound.
8089  *
8090  * When the client supports GETATTR(fs_locations_info), it can
8091  * be plumbed in here.
8092  */
8093 static int _nfs41_proc_get_locations(struct nfs_server *server,
8094 				     struct nfs_fh *fhandle,
8095 				     struct nfs4_fs_locations *locations,
8096 				     struct page *page, const struct cred *cred)
8097 {
8098 	struct rpc_clnt *clnt = server->client;
8099 	u32 bitmask[2] = {
8100 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8101 	};
8102 	struct nfs4_fs_locations_arg args = {
8103 		.fh		= fhandle,
8104 		.page		= page,
8105 		.bitmask	= bitmask,
8106 		.migration	= 1,		/* skip LOOKUP */
8107 	};
8108 	struct nfs4_fs_locations_res res = {
8109 		.fs_locations	= locations,
8110 		.migration	= 1,
8111 	};
8112 	struct rpc_message msg = {
8113 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8114 		.rpc_argp	= &args,
8115 		.rpc_resp	= &res,
8116 		.rpc_cred	= cred,
8117 	};
8118 	struct nfs4_call_sync_data data = {
8119 		.seq_server = server,
8120 		.seq_args = &args.seq_args,
8121 		.seq_res = &res.seq_res,
8122 	};
8123 	struct rpc_task_setup task_setup_data = {
8124 		.rpc_client = clnt,
8125 		.rpc_message = &msg,
8126 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8127 		.callback_data = &data,
8128 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8129 	};
8130 	int status;
8131 
8132 	nfs_fattr_init(locations->fattr);
8133 	locations->server = server;
8134 	locations->nlocations = 0;
8135 
8136 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8137 	status = nfs4_call_sync_custom(&task_setup_data);
8138 	if (status == NFS4_OK &&
8139 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8140 		status = -NFS4ERR_LEASE_MOVED;
8141 	return status;
8142 }
8143 
8144 #endif	/* CONFIG_NFS_V4_1 */
8145 
8146 /**
8147  * nfs4_proc_get_locations - discover locations for a migrated FSID
8148  * @server: pointer to nfs_server to process
8149  * @fhandle: pointer to the kernel NFS client file handle
8150  * @locations: result of query
8151  * @page: buffer
8152  * @cred: credential to use for this operation
8153  *
8154  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8155  * operation failed, or a negative errno if a local error occurred.
8156  *
8157  * On success, "locations" is filled in, but if the server has
8158  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8159  * asserted.
8160  *
8161  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8162  * from this client that require migration recovery.
8163  */
8164 int nfs4_proc_get_locations(struct nfs_server *server,
8165 			    struct nfs_fh *fhandle,
8166 			    struct nfs4_fs_locations *locations,
8167 			    struct page *page, const struct cred *cred)
8168 {
8169 	struct nfs_client *clp = server->nfs_client;
8170 	const struct nfs4_mig_recovery_ops *ops =
8171 					clp->cl_mvops->mig_recovery_ops;
8172 	struct nfs4_exception exception = {
8173 		.interruptible = true,
8174 	};
8175 	int status;
8176 
8177 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8178 		(unsigned long long)server->fsid.major,
8179 		(unsigned long long)server->fsid.minor,
8180 		clp->cl_hostname);
8181 	nfs_display_fhandle(fhandle, __func__);
8182 
8183 	do {
8184 		status = ops->get_locations(server, fhandle, locations, page,
8185 					    cred);
8186 		if (status != -NFS4ERR_DELAY)
8187 			break;
8188 		nfs4_handle_exception(server, status, &exception);
8189 	} while (exception.retry);
8190 	return status;
8191 }
8192 
8193 /*
8194  * This operation also signals the server that this client is
8195  * performing "lease moved" recovery.  The server can stop
8196  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8197  * is appended to this compound to identify the client ID which is
8198  * performing recovery.
8199  */
8200 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8201 {
8202 	struct nfs_server *server = NFS_SERVER(inode);
8203 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8204 	struct rpc_clnt *clnt = server->client;
8205 	struct nfs4_fsid_present_arg args = {
8206 		.fh		= NFS_FH(inode),
8207 		.clientid	= clp->cl_clientid,
8208 		.renew		= 1,		/* append RENEW */
8209 	};
8210 	struct nfs4_fsid_present_res res = {
8211 		.renew		= 1,
8212 	};
8213 	struct rpc_message msg = {
8214 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8215 		.rpc_argp	= &args,
8216 		.rpc_resp	= &res,
8217 		.rpc_cred	= cred,
8218 	};
8219 	unsigned long now = jiffies;
8220 	int status;
8221 
8222 	res.fh = nfs_alloc_fhandle();
8223 	if (res.fh == NULL)
8224 		return -ENOMEM;
8225 
8226 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8227 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8228 						&args.seq_args, &res.seq_res);
8229 	nfs_free_fhandle(res.fh);
8230 	if (status)
8231 		return status;
8232 
8233 	do_renew_lease(clp, now);
8234 	return 0;
8235 }
8236 
8237 #ifdef CONFIG_NFS_V4_1
8238 
8239 /*
8240  * This operation also signals the server that this client is
8241  * performing "lease moved" recovery.  The server can stop asserting
8242  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8243  * this operation is identified in the SEQUENCE operation in this
8244  * compound.
8245  */
8246 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8247 {
8248 	struct nfs_server *server = NFS_SERVER(inode);
8249 	struct rpc_clnt *clnt = server->client;
8250 	struct nfs4_fsid_present_arg args = {
8251 		.fh		= NFS_FH(inode),
8252 	};
8253 	struct nfs4_fsid_present_res res = {
8254 	};
8255 	struct rpc_message msg = {
8256 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8257 		.rpc_argp	= &args,
8258 		.rpc_resp	= &res,
8259 		.rpc_cred	= cred,
8260 	};
8261 	int status;
8262 
8263 	res.fh = nfs_alloc_fhandle();
8264 	if (res.fh == NULL)
8265 		return -ENOMEM;
8266 
8267 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8268 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8269 						&args.seq_args, &res.seq_res);
8270 	nfs_free_fhandle(res.fh);
8271 	if (status == NFS4_OK &&
8272 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8273 		status = -NFS4ERR_LEASE_MOVED;
8274 	return status;
8275 }
8276 
8277 #endif	/* CONFIG_NFS_V4_1 */
8278 
8279 /**
8280  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8281  * @inode: inode on FSID to check
8282  * @cred: credential to use for this operation
8283  *
8284  * Server indicates whether the FSID is present, moved, or not
8285  * recognized.  This operation is necessary to clear a LEASE_MOVED
8286  * condition for this client ID.
8287  *
8288  * Returns NFS4_OK if the FSID is present on this server,
8289  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8290  *  NFS4ERR code if some error occurred on the server, or a
8291  *  negative errno if a local failure occurred.
8292  */
8293 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8294 {
8295 	struct nfs_server *server = NFS_SERVER(inode);
8296 	struct nfs_client *clp = server->nfs_client;
8297 	const struct nfs4_mig_recovery_ops *ops =
8298 					clp->cl_mvops->mig_recovery_ops;
8299 	struct nfs4_exception exception = {
8300 		.interruptible = true,
8301 	};
8302 	int status;
8303 
8304 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8305 		(unsigned long long)server->fsid.major,
8306 		(unsigned long long)server->fsid.minor,
8307 		clp->cl_hostname);
8308 	nfs_display_fhandle(NFS_FH(inode), __func__);
8309 
8310 	do {
8311 		status = ops->fsid_present(inode, cred);
8312 		if (status != -NFS4ERR_DELAY)
8313 			break;
8314 		nfs4_handle_exception(server, status, &exception);
8315 	} while (exception.retry);
8316 	return status;
8317 }
8318 
8319 /*
8320  * If 'use_integrity' is true and the state managment nfs_client
8321  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8322  * and the machine credential as per RFC3530bis and RFC5661 Security
8323  * Considerations sections. Otherwise, just use the user cred with the
8324  * filesystem's rpc_client.
8325  */
8326 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8327 {
8328 	int status;
8329 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8330 	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8331 	struct nfs4_secinfo_arg args = {
8332 		.dir_fh = NFS_FH(dir),
8333 		.name   = name,
8334 	};
8335 	struct nfs4_secinfo_res res = {
8336 		.flavors     = flavors,
8337 	};
8338 	struct rpc_message msg = {
8339 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8340 		.rpc_argp = &args,
8341 		.rpc_resp = &res,
8342 	};
8343 	struct nfs4_call_sync_data data = {
8344 		.seq_server = NFS_SERVER(dir),
8345 		.seq_args = &args.seq_args,
8346 		.seq_res = &res.seq_res,
8347 	};
8348 	struct rpc_task_setup task_setup = {
8349 		.rpc_client = clnt,
8350 		.rpc_message = &msg,
8351 		.callback_ops = clp->cl_mvops->call_sync_ops,
8352 		.callback_data = &data,
8353 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8354 	};
8355 	const struct cred *cred = NULL;
8356 
8357 	if (use_integrity) {
8358 		clnt = clp->cl_rpcclient;
8359 		task_setup.rpc_client = clnt;
8360 
8361 		cred = nfs4_get_clid_cred(clp);
8362 		msg.rpc_cred = cred;
8363 	}
8364 
8365 	dprintk("NFS call  secinfo %s\n", name->name);
8366 
8367 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8368 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8369 	status = nfs4_call_sync_custom(&task_setup);
8370 
8371 	dprintk("NFS reply  secinfo: %d\n", status);
8372 
8373 	put_cred(cred);
8374 	return status;
8375 }
8376 
8377 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8378 		      struct nfs4_secinfo_flavors *flavors)
8379 {
8380 	struct nfs4_exception exception = {
8381 		.interruptible = true,
8382 	};
8383 	int err;
8384 	do {
8385 		err = -NFS4ERR_WRONGSEC;
8386 
8387 		/* try to use integrity protection with machine cred */
8388 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8389 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8390 
8391 		/*
8392 		 * if unable to use integrity protection, or SECINFO with
8393 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8394 		 * disallowed by spec, but exists in deployed servers) use
8395 		 * the current filesystem's rpc_client and the user cred.
8396 		 */
8397 		if (err == -NFS4ERR_WRONGSEC)
8398 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8399 
8400 		trace_nfs4_secinfo(dir, name, err);
8401 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8402 				&exception);
8403 	} while (exception.retry);
8404 	return err;
8405 }
8406 
8407 #ifdef CONFIG_NFS_V4_1
8408 /*
8409  * Check the exchange flags returned by the server for invalid flags, having
8410  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8411  * DS flags set.
8412  */
8413 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8414 {
8415 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8416 		goto out_inval;
8417 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8418 		goto out_inval;
8419 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8420 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8421 		goto out_inval;
8422 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8423 		goto out_inval;
8424 	return NFS_OK;
8425 out_inval:
8426 	return -NFS4ERR_INVAL;
8427 }
8428 
8429 static bool
8430 nfs41_same_server_scope(struct nfs41_server_scope *a,
8431 			struct nfs41_server_scope *b)
8432 {
8433 	if (a->server_scope_sz != b->server_scope_sz)
8434 		return false;
8435 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8436 }
8437 
8438 static void
8439 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8440 {
8441 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8442 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8443 	struct nfs_client *clp = args->client;
8444 
8445 	switch (task->tk_status) {
8446 	case -NFS4ERR_BADSESSION:
8447 	case -NFS4ERR_DEADSESSION:
8448 		nfs4_schedule_session_recovery(clp->cl_session,
8449 				task->tk_status);
8450 		return;
8451 	}
8452 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8453 			res->dir != NFS4_CDFS4_BOTH) {
8454 		rpc_task_close_connection(task);
8455 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8456 			rpc_restart_call(task);
8457 	}
8458 }
8459 
8460 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8461 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8462 };
8463 
8464 /*
8465  * nfs4_proc_bind_one_conn_to_session()
8466  *
8467  * The 4.1 client currently uses the same TCP connection for the
8468  * fore and backchannel.
8469  */
8470 static
8471 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8472 		struct rpc_xprt *xprt,
8473 		struct nfs_client *clp,
8474 		const struct cred *cred)
8475 {
8476 	int status;
8477 	struct nfs41_bind_conn_to_session_args args = {
8478 		.client = clp,
8479 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8480 		.retries = 0,
8481 	};
8482 	struct nfs41_bind_conn_to_session_res res;
8483 	struct rpc_message msg = {
8484 		.rpc_proc =
8485 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8486 		.rpc_argp = &args,
8487 		.rpc_resp = &res,
8488 		.rpc_cred = cred,
8489 	};
8490 	struct rpc_task_setup task_setup_data = {
8491 		.rpc_client = clnt,
8492 		.rpc_xprt = xprt,
8493 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8494 		.rpc_message = &msg,
8495 		.flags = RPC_TASK_TIMEOUT,
8496 	};
8497 	struct rpc_task *task;
8498 
8499 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8500 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8501 		args.dir = NFS4_CDFC4_FORE;
8502 
8503 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8504 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8505 		args.dir = NFS4_CDFC4_FORE;
8506 
8507 	task = rpc_run_task(&task_setup_data);
8508 	if (!IS_ERR(task)) {
8509 		status = task->tk_status;
8510 		rpc_put_task(task);
8511 	} else
8512 		status = PTR_ERR(task);
8513 	trace_nfs4_bind_conn_to_session(clp, status);
8514 	if (status == 0) {
8515 		if (memcmp(res.sessionid.data,
8516 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8517 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8518 			return -EIO;
8519 		}
8520 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8521 			dprintk("NFS: %s: Unexpected direction from server\n",
8522 				__func__);
8523 			return -EIO;
8524 		}
8525 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8526 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8527 				__func__);
8528 			return -EIO;
8529 		}
8530 	}
8531 
8532 	return status;
8533 }
8534 
8535 struct rpc_bind_conn_calldata {
8536 	struct nfs_client *clp;
8537 	const struct cred *cred;
8538 };
8539 
8540 static int
8541 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8542 		struct rpc_xprt *xprt,
8543 		void *calldata)
8544 {
8545 	struct rpc_bind_conn_calldata *p = calldata;
8546 
8547 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8548 }
8549 
8550 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8551 {
8552 	struct rpc_bind_conn_calldata data = {
8553 		.clp = clp,
8554 		.cred = cred,
8555 	};
8556 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8557 			nfs4_proc_bind_conn_to_session_callback, &data);
8558 }
8559 
8560 /*
8561  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8562  * and operations we'd like to see to enable certain features in the allow map
8563  */
8564 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8565 	.how = SP4_MACH_CRED,
8566 	.enforce.u.words = {
8567 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8568 		      1 << (OP_EXCHANGE_ID - 32) |
8569 		      1 << (OP_CREATE_SESSION - 32) |
8570 		      1 << (OP_DESTROY_SESSION - 32) |
8571 		      1 << (OP_DESTROY_CLIENTID - 32)
8572 	},
8573 	.allow.u.words = {
8574 		[0] = 1 << (OP_CLOSE) |
8575 		      1 << (OP_OPEN_DOWNGRADE) |
8576 		      1 << (OP_LOCKU) |
8577 		      1 << (OP_DELEGRETURN) |
8578 		      1 << (OP_COMMIT),
8579 		[1] = 1 << (OP_SECINFO - 32) |
8580 		      1 << (OP_SECINFO_NO_NAME - 32) |
8581 		      1 << (OP_LAYOUTRETURN - 32) |
8582 		      1 << (OP_TEST_STATEID - 32) |
8583 		      1 << (OP_FREE_STATEID - 32) |
8584 		      1 << (OP_WRITE - 32)
8585 	}
8586 };
8587 
8588 /*
8589  * Select the state protection mode for client `clp' given the server results
8590  * from exchange_id in `sp'.
8591  *
8592  * Returns 0 on success, negative errno otherwise.
8593  */
8594 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8595 				 struct nfs41_state_protection *sp)
8596 {
8597 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8598 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8599 		      1 << (OP_EXCHANGE_ID - 32) |
8600 		      1 << (OP_CREATE_SESSION - 32) |
8601 		      1 << (OP_DESTROY_SESSION - 32) |
8602 		      1 << (OP_DESTROY_CLIENTID - 32)
8603 	};
8604 	unsigned long flags = 0;
8605 	unsigned int i;
8606 	int ret = 0;
8607 
8608 	if (sp->how == SP4_MACH_CRED) {
8609 		/* Print state protect result */
8610 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8611 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8612 			if (test_bit(i, sp->enforce.u.longs))
8613 				dfprintk(MOUNT, "  enforce op %d\n", i);
8614 			if (test_bit(i, sp->allow.u.longs))
8615 				dfprintk(MOUNT, "  allow op %d\n", i);
8616 		}
8617 
8618 		/* make sure nothing is on enforce list that isn't supported */
8619 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8620 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8621 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8622 				ret = -EINVAL;
8623 				goto out;
8624 			}
8625 		}
8626 
8627 		/*
8628 		 * Minimal mode - state operations are allowed to use machine
8629 		 * credential.  Note this already happens by default, so the
8630 		 * client doesn't have to do anything more than the negotiation.
8631 		 *
8632 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8633 		 *       we're already using the machine cred for exchange_id
8634 		 *       and will never use a different cred.
8635 		 */
8636 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8637 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8638 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8639 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8640 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8641 			dfprintk(MOUNT, "  minimal mode enabled\n");
8642 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8643 		} else {
8644 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8645 			ret = -EINVAL;
8646 			goto out;
8647 		}
8648 
8649 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8650 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8651 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8652 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8653 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8654 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8655 		}
8656 
8657 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8658 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8659 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8660 		}
8661 
8662 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8663 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8664 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8665 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8666 		}
8667 
8668 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8669 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8670 			dfprintk(MOUNT, "  stateid mode enabled\n");
8671 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8672 		}
8673 
8674 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8675 			dfprintk(MOUNT, "  write mode enabled\n");
8676 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8677 		}
8678 
8679 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8680 			dfprintk(MOUNT, "  commit mode enabled\n");
8681 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8682 		}
8683 	}
8684 out:
8685 	clp->cl_sp4_flags = flags;
8686 	return ret;
8687 }
8688 
8689 struct nfs41_exchange_id_data {
8690 	struct nfs41_exchange_id_res res;
8691 	struct nfs41_exchange_id_args args;
8692 };
8693 
8694 static void nfs4_exchange_id_release(void *data)
8695 {
8696 	struct nfs41_exchange_id_data *cdata =
8697 					(struct nfs41_exchange_id_data *)data;
8698 
8699 	nfs_put_client(cdata->args.client);
8700 	kfree(cdata->res.impl_id);
8701 	kfree(cdata->res.server_scope);
8702 	kfree(cdata->res.server_owner);
8703 	kfree(cdata);
8704 }
8705 
8706 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8707 	.rpc_release = nfs4_exchange_id_release,
8708 };
8709 
8710 /*
8711  * _nfs4_proc_exchange_id()
8712  *
8713  * Wrapper for EXCHANGE_ID operation.
8714  */
8715 static struct rpc_task *
8716 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8717 			u32 sp4_how, struct rpc_xprt *xprt)
8718 {
8719 	struct rpc_message msg = {
8720 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8721 		.rpc_cred = cred,
8722 	};
8723 	struct rpc_task_setup task_setup_data = {
8724 		.rpc_client = clp->cl_rpcclient,
8725 		.callback_ops = &nfs4_exchange_id_call_ops,
8726 		.rpc_message = &msg,
8727 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8728 	};
8729 	struct nfs41_exchange_id_data *calldata;
8730 	int status;
8731 
8732 	if (!refcount_inc_not_zero(&clp->cl_count))
8733 		return ERR_PTR(-EIO);
8734 
8735 	status = -ENOMEM;
8736 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8737 	if (!calldata)
8738 		goto out;
8739 
8740 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8741 
8742 	status = nfs4_init_uniform_client_string(clp);
8743 	if (status)
8744 		goto out_calldata;
8745 
8746 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8747 						GFP_NOFS);
8748 	status = -ENOMEM;
8749 	if (unlikely(calldata->res.server_owner == NULL))
8750 		goto out_calldata;
8751 
8752 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8753 					GFP_NOFS);
8754 	if (unlikely(calldata->res.server_scope == NULL))
8755 		goto out_server_owner;
8756 
8757 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8758 	if (unlikely(calldata->res.impl_id == NULL))
8759 		goto out_server_scope;
8760 
8761 	switch (sp4_how) {
8762 	case SP4_NONE:
8763 		calldata->args.state_protect.how = SP4_NONE;
8764 		break;
8765 
8766 	case SP4_MACH_CRED:
8767 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8768 		break;
8769 
8770 	default:
8771 		/* unsupported! */
8772 		WARN_ON_ONCE(1);
8773 		status = -EINVAL;
8774 		goto out_impl_id;
8775 	}
8776 	if (xprt) {
8777 		task_setup_data.rpc_xprt = xprt;
8778 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8779 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8780 				sizeof(calldata->args.verifier.data));
8781 	}
8782 	calldata->args.client = clp;
8783 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8784 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8785 #ifdef CONFIG_NFS_V4_1_MIGRATION
8786 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8787 #endif
8788 	msg.rpc_argp = &calldata->args;
8789 	msg.rpc_resp = &calldata->res;
8790 	task_setup_data.callback_data = calldata;
8791 
8792 	return rpc_run_task(&task_setup_data);
8793 
8794 out_impl_id:
8795 	kfree(calldata->res.impl_id);
8796 out_server_scope:
8797 	kfree(calldata->res.server_scope);
8798 out_server_owner:
8799 	kfree(calldata->res.server_owner);
8800 out_calldata:
8801 	kfree(calldata);
8802 out:
8803 	nfs_put_client(clp);
8804 	return ERR_PTR(status);
8805 }
8806 
8807 /*
8808  * _nfs4_proc_exchange_id()
8809  *
8810  * Wrapper for EXCHANGE_ID operation.
8811  */
8812 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8813 			u32 sp4_how)
8814 {
8815 	struct rpc_task *task;
8816 	struct nfs41_exchange_id_args *argp;
8817 	struct nfs41_exchange_id_res *resp;
8818 	unsigned long now = jiffies;
8819 	int status;
8820 
8821 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8822 	if (IS_ERR(task))
8823 		return PTR_ERR(task);
8824 
8825 	argp = task->tk_msg.rpc_argp;
8826 	resp = task->tk_msg.rpc_resp;
8827 	status = task->tk_status;
8828 	if (status  != 0)
8829 		goto out;
8830 
8831 	status = nfs4_check_cl_exchange_flags(resp->flags,
8832 			clp->cl_mvops->minor_version);
8833 	if (status  != 0)
8834 		goto out;
8835 
8836 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8837 	if (status != 0)
8838 		goto out;
8839 
8840 	do_renew_lease(clp, now);
8841 
8842 	clp->cl_clientid = resp->clientid;
8843 	clp->cl_exchange_flags = resp->flags;
8844 	clp->cl_seqid = resp->seqid;
8845 	/* Client ID is not confirmed */
8846 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8847 		clear_bit(NFS4_SESSION_ESTABLISHED,
8848 			  &clp->cl_session->session_state);
8849 
8850 	if (clp->cl_serverscope != NULL &&
8851 	    !nfs41_same_server_scope(clp->cl_serverscope,
8852 				resp->server_scope)) {
8853 		dprintk("%s: server_scope mismatch detected\n",
8854 			__func__);
8855 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8856 	}
8857 
8858 	swap(clp->cl_serverowner, resp->server_owner);
8859 	swap(clp->cl_serverscope, resp->server_scope);
8860 	swap(clp->cl_implid, resp->impl_id);
8861 
8862 	/* Save the EXCHANGE_ID verifier session trunk tests */
8863 	memcpy(clp->cl_confirm.data, argp->verifier.data,
8864 	       sizeof(clp->cl_confirm.data));
8865 out:
8866 	trace_nfs4_exchange_id(clp, status);
8867 	rpc_put_task(task);
8868 	return status;
8869 }
8870 
8871 /*
8872  * nfs4_proc_exchange_id()
8873  *
8874  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8875  *
8876  * Since the clientid has expired, all compounds using sessions
8877  * associated with the stale clientid will be returning
8878  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8879  * be in some phase of session reset.
8880  *
8881  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8882  */
8883 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8884 {
8885 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8886 	int status;
8887 
8888 	/* try SP4_MACH_CRED if krb5i/p	*/
8889 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8890 	    authflavor == RPC_AUTH_GSS_KRB5P) {
8891 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8892 		if (!status)
8893 			return 0;
8894 	}
8895 
8896 	/* try SP4_NONE */
8897 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8898 }
8899 
8900 /**
8901  * nfs4_test_session_trunk
8902  *
8903  * This is an add_xprt_test() test function called from
8904  * rpc_clnt_setup_test_and_add_xprt.
8905  *
8906  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8907  * and is dereferrenced in nfs4_exchange_id_release
8908  *
8909  * Upon success, add the new transport to the rpc_clnt
8910  *
8911  * @clnt: struct rpc_clnt to get new transport
8912  * @xprt: the rpc_xprt to test
8913  * @data: call data for _nfs4_proc_exchange_id.
8914  */
8915 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8916 			    void *data)
8917 {
8918 	struct nfs4_add_xprt_data *adata = data;
8919 	struct rpc_task *task;
8920 	int status;
8921 
8922 	u32 sp4_how;
8923 
8924 	dprintk("--> %s try %s\n", __func__,
8925 		xprt->address_strings[RPC_DISPLAY_ADDR]);
8926 
8927 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8928 
8929 	/* Test connection for session trunking. Async exchange_id call */
8930 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8931 	if (IS_ERR(task))
8932 		return;
8933 
8934 	status = task->tk_status;
8935 	if (status == 0)
8936 		status = nfs4_detect_session_trunking(adata->clp,
8937 				task->tk_msg.rpc_resp, xprt);
8938 
8939 	if (status == 0)
8940 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8941 	else if (rpc_clnt_xprt_switch_has_addr(clnt,
8942 				(struct sockaddr *)&xprt->addr))
8943 		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8944 
8945 	rpc_put_task(task);
8946 }
8947 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8948 
8949 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8950 		const struct cred *cred)
8951 {
8952 	struct rpc_message msg = {
8953 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8954 		.rpc_argp = clp,
8955 		.rpc_cred = cred,
8956 	};
8957 	int status;
8958 
8959 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
8960 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8961 	trace_nfs4_destroy_clientid(clp, status);
8962 	if (status)
8963 		dprintk("NFS: Got error %d from the server %s on "
8964 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
8965 	return status;
8966 }
8967 
8968 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8969 		const struct cred *cred)
8970 {
8971 	unsigned int loop;
8972 	int ret;
8973 
8974 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8975 		ret = _nfs4_proc_destroy_clientid(clp, cred);
8976 		switch (ret) {
8977 		case -NFS4ERR_DELAY:
8978 		case -NFS4ERR_CLIENTID_BUSY:
8979 			ssleep(1);
8980 			break;
8981 		default:
8982 			return ret;
8983 		}
8984 	}
8985 	return 0;
8986 }
8987 
8988 int nfs4_destroy_clientid(struct nfs_client *clp)
8989 {
8990 	const struct cred *cred;
8991 	int ret = 0;
8992 
8993 	if (clp->cl_mvops->minor_version < 1)
8994 		goto out;
8995 	if (clp->cl_exchange_flags == 0)
8996 		goto out;
8997 	if (clp->cl_preserve_clid)
8998 		goto out;
8999 	cred = nfs4_get_clid_cred(clp);
9000 	ret = nfs4_proc_destroy_clientid(clp, cred);
9001 	put_cred(cred);
9002 	switch (ret) {
9003 	case 0:
9004 	case -NFS4ERR_STALE_CLIENTID:
9005 		clp->cl_exchange_flags = 0;
9006 	}
9007 out:
9008 	return ret;
9009 }
9010 
9011 #endif /* CONFIG_NFS_V4_1 */
9012 
9013 struct nfs4_get_lease_time_data {
9014 	struct nfs4_get_lease_time_args *args;
9015 	struct nfs4_get_lease_time_res *res;
9016 	struct nfs_client *clp;
9017 };
9018 
9019 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9020 					void *calldata)
9021 {
9022 	struct nfs4_get_lease_time_data *data =
9023 			(struct nfs4_get_lease_time_data *)calldata;
9024 
9025 	/* just setup sequence, do not trigger session recovery
9026 	   since we're invoked within one */
9027 	nfs4_setup_sequence(data->clp,
9028 			&data->args->la_seq_args,
9029 			&data->res->lr_seq_res,
9030 			task);
9031 }
9032 
9033 /*
9034  * Called from nfs4_state_manager thread for session setup, so don't recover
9035  * from sequence operation or clientid errors.
9036  */
9037 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9038 {
9039 	struct nfs4_get_lease_time_data *data =
9040 			(struct nfs4_get_lease_time_data *)calldata;
9041 
9042 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9043 		return;
9044 	switch (task->tk_status) {
9045 	case -NFS4ERR_DELAY:
9046 	case -NFS4ERR_GRACE:
9047 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9048 		task->tk_status = 0;
9049 		fallthrough;
9050 	case -NFS4ERR_RETRY_UNCACHED_REP:
9051 		rpc_restart_call_prepare(task);
9052 		return;
9053 	}
9054 }
9055 
9056 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9057 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9058 	.rpc_call_done = nfs4_get_lease_time_done,
9059 };
9060 
9061 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9062 {
9063 	struct nfs4_get_lease_time_args args;
9064 	struct nfs4_get_lease_time_res res = {
9065 		.lr_fsinfo = fsinfo,
9066 	};
9067 	struct nfs4_get_lease_time_data data = {
9068 		.args = &args,
9069 		.res = &res,
9070 		.clp = clp,
9071 	};
9072 	struct rpc_message msg = {
9073 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9074 		.rpc_argp = &args,
9075 		.rpc_resp = &res,
9076 	};
9077 	struct rpc_task_setup task_setup = {
9078 		.rpc_client = clp->cl_rpcclient,
9079 		.rpc_message = &msg,
9080 		.callback_ops = &nfs4_get_lease_time_ops,
9081 		.callback_data = &data,
9082 		.flags = RPC_TASK_TIMEOUT,
9083 	};
9084 
9085 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9086 	return nfs4_call_sync_custom(&task_setup);
9087 }
9088 
9089 #ifdef CONFIG_NFS_V4_1
9090 
9091 /*
9092  * Initialize the values to be used by the client in CREATE_SESSION
9093  * If nfs4_init_session set the fore channel request and response sizes,
9094  * use them.
9095  *
9096  * Set the back channel max_resp_sz_cached to zero to force the client to
9097  * always set csa_cachethis to FALSE because the current implementation
9098  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9099  */
9100 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9101 				    struct rpc_clnt *clnt)
9102 {
9103 	unsigned int max_rqst_sz, max_resp_sz;
9104 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9105 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9106 
9107 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9108 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9109 
9110 	/* Fore channel attributes */
9111 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9112 	args->fc_attrs.max_resp_sz = max_resp_sz;
9113 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9114 	args->fc_attrs.max_reqs = max_session_slots;
9115 
9116 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9117 		"max_ops=%u max_reqs=%u\n",
9118 		__func__,
9119 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9120 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9121 
9122 	/* Back channel attributes */
9123 	args->bc_attrs.max_rqst_sz = max_bc_payload;
9124 	args->bc_attrs.max_resp_sz = max_bc_payload;
9125 	args->bc_attrs.max_resp_sz_cached = 0;
9126 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9127 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9128 	if (args->bc_attrs.max_reqs > max_bc_slots)
9129 		args->bc_attrs.max_reqs = max_bc_slots;
9130 
9131 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9132 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9133 		__func__,
9134 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9135 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9136 		args->bc_attrs.max_reqs);
9137 }
9138 
9139 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9140 		struct nfs41_create_session_res *res)
9141 {
9142 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9143 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9144 
9145 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9146 		return -EINVAL;
9147 	/*
9148 	 * Our requested max_ops is the minimum we need; we're not
9149 	 * prepared to break up compounds into smaller pieces than that.
9150 	 * So, no point even trying to continue if the server won't
9151 	 * cooperate:
9152 	 */
9153 	if (rcvd->max_ops < sent->max_ops)
9154 		return -EINVAL;
9155 	if (rcvd->max_reqs == 0)
9156 		return -EINVAL;
9157 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9158 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9159 	return 0;
9160 }
9161 
9162 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9163 		struct nfs41_create_session_res *res)
9164 {
9165 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9166 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9167 
9168 	if (!(res->flags & SESSION4_BACK_CHAN))
9169 		goto out;
9170 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9171 		return -EINVAL;
9172 	if (rcvd->max_resp_sz < sent->max_resp_sz)
9173 		return -EINVAL;
9174 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9175 		return -EINVAL;
9176 	if (rcvd->max_ops > sent->max_ops)
9177 		return -EINVAL;
9178 	if (rcvd->max_reqs > sent->max_reqs)
9179 		return -EINVAL;
9180 out:
9181 	return 0;
9182 }
9183 
9184 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9185 				     struct nfs41_create_session_res *res)
9186 {
9187 	int ret;
9188 
9189 	ret = nfs4_verify_fore_channel_attrs(args, res);
9190 	if (ret)
9191 		return ret;
9192 	return nfs4_verify_back_channel_attrs(args, res);
9193 }
9194 
9195 static void nfs4_update_session(struct nfs4_session *session,
9196 		struct nfs41_create_session_res *res)
9197 {
9198 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9199 	/* Mark client id and session as being confirmed */
9200 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9201 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9202 	session->flags = res->flags;
9203 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9204 	if (res->flags & SESSION4_BACK_CHAN)
9205 		memcpy(&session->bc_attrs, &res->bc_attrs,
9206 				sizeof(session->bc_attrs));
9207 }
9208 
9209 static int _nfs4_proc_create_session(struct nfs_client *clp,
9210 		const struct cred *cred)
9211 {
9212 	struct nfs4_session *session = clp->cl_session;
9213 	struct nfs41_create_session_args args = {
9214 		.client = clp,
9215 		.clientid = clp->cl_clientid,
9216 		.seqid = clp->cl_seqid,
9217 		.cb_program = NFS4_CALLBACK,
9218 	};
9219 	struct nfs41_create_session_res res;
9220 
9221 	struct rpc_message msg = {
9222 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9223 		.rpc_argp = &args,
9224 		.rpc_resp = &res,
9225 		.rpc_cred = cred,
9226 	};
9227 	int status;
9228 
9229 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9230 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9231 
9232 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9233 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9234 	trace_nfs4_create_session(clp, status);
9235 
9236 	switch (status) {
9237 	case -NFS4ERR_STALE_CLIENTID:
9238 	case -NFS4ERR_DELAY:
9239 	case -ETIMEDOUT:
9240 	case -EACCES:
9241 	case -EAGAIN:
9242 		goto out;
9243 	}
9244 
9245 	clp->cl_seqid++;
9246 	if (!status) {
9247 		/* Verify the session's negotiated channel_attrs values */
9248 		status = nfs4_verify_channel_attrs(&args, &res);
9249 		/* Increment the clientid slot sequence id */
9250 		if (status)
9251 			goto out;
9252 		nfs4_update_session(session, &res);
9253 	}
9254 out:
9255 	return status;
9256 }
9257 
9258 /*
9259  * Issues a CREATE_SESSION operation to the server.
9260  * It is the responsibility of the caller to verify the session is
9261  * expired before calling this routine.
9262  */
9263 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9264 {
9265 	int status;
9266 	unsigned *ptr;
9267 	struct nfs4_session *session = clp->cl_session;
9268 	struct nfs4_add_xprt_data xprtdata = {
9269 		.clp = clp,
9270 	};
9271 	struct rpc_add_xprt_test rpcdata = {
9272 		.add_xprt_test = clp->cl_mvops->session_trunk,
9273 		.data = &xprtdata,
9274 	};
9275 
9276 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9277 
9278 	status = _nfs4_proc_create_session(clp, cred);
9279 	if (status)
9280 		goto out;
9281 
9282 	/* Init or reset the session slot tables */
9283 	status = nfs4_setup_session_slot_tables(session);
9284 	dprintk("slot table setup returned %d\n", status);
9285 	if (status)
9286 		goto out;
9287 
9288 	ptr = (unsigned *)&session->sess_id.data[0];
9289 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9290 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9291 	rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9292 out:
9293 	return status;
9294 }
9295 
9296 /*
9297  * Issue the over-the-wire RPC DESTROY_SESSION.
9298  * The caller must serialize access to this routine.
9299  */
9300 int nfs4_proc_destroy_session(struct nfs4_session *session,
9301 		const struct cred *cred)
9302 {
9303 	struct rpc_message msg = {
9304 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9305 		.rpc_argp = session,
9306 		.rpc_cred = cred,
9307 	};
9308 	int status = 0;
9309 
9310 	/* session is still being setup */
9311 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9312 		return 0;
9313 
9314 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9315 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9316 	trace_nfs4_destroy_session(session->clp, status);
9317 
9318 	if (status)
9319 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9320 			"Session has been destroyed regardless...\n", status);
9321 	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9322 	return status;
9323 }
9324 
9325 /*
9326  * Renew the cl_session lease.
9327  */
9328 struct nfs4_sequence_data {
9329 	struct nfs_client *clp;
9330 	struct nfs4_sequence_args args;
9331 	struct nfs4_sequence_res res;
9332 };
9333 
9334 static void nfs41_sequence_release(void *data)
9335 {
9336 	struct nfs4_sequence_data *calldata = data;
9337 	struct nfs_client *clp = calldata->clp;
9338 
9339 	if (refcount_read(&clp->cl_count) > 1)
9340 		nfs4_schedule_state_renewal(clp);
9341 	nfs_put_client(clp);
9342 	kfree(calldata);
9343 }
9344 
9345 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9346 {
9347 	switch(task->tk_status) {
9348 	case -NFS4ERR_DELAY:
9349 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9350 		return -EAGAIN;
9351 	default:
9352 		nfs4_schedule_lease_recovery(clp);
9353 	}
9354 	return 0;
9355 }
9356 
9357 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9358 {
9359 	struct nfs4_sequence_data *calldata = data;
9360 	struct nfs_client *clp = calldata->clp;
9361 
9362 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9363 		return;
9364 
9365 	trace_nfs4_sequence(clp, task->tk_status);
9366 	if (task->tk_status < 0) {
9367 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9368 		if (refcount_read(&clp->cl_count) == 1)
9369 			return;
9370 
9371 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9372 			rpc_restart_call_prepare(task);
9373 			return;
9374 		}
9375 	}
9376 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9377 }
9378 
9379 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9380 {
9381 	struct nfs4_sequence_data *calldata = data;
9382 	struct nfs_client *clp = calldata->clp;
9383 	struct nfs4_sequence_args *args;
9384 	struct nfs4_sequence_res *res;
9385 
9386 	args = task->tk_msg.rpc_argp;
9387 	res = task->tk_msg.rpc_resp;
9388 
9389 	nfs4_setup_sequence(clp, args, res, task);
9390 }
9391 
9392 static const struct rpc_call_ops nfs41_sequence_ops = {
9393 	.rpc_call_done = nfs41_sequence_call_done,
9394 	.rpc_call_prepare = nfs41_sequence_prepare,
9395 	.rpc_release = nfs41_sequence_release,
9396 };
9397 
9398 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9399 		const struct cred *cred,
9400 		struct nfs4_slot *slot,
9401 		bool is_privileged)
9402 {
9403 	struct nfs4_sequence_data *calldata;
9404 	struct rpc_message msg = {
9405 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9406 		.rpc_cred = cred,
9407 	};
9408 	struct rpc_task_setup task_setup_data = {
9409 		.rpc_client = clp->cl_rpcclient,
9410 		.rpc_message = &msg,
9411 		.callback_ops = &nfs41_sequence_ops,
9412 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9413 	};
9414 	struct rpc_task *ret;
9415 
9416 	ret = ERR_PTR(-EIO);
9417 	if (!refcount_inc_not_zero(&clp->cl_count))
9418 		goto out_err;
9419 
9420 	ret = ERR_PTR(-ENOMEM);
9421 	calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9422 	if (calldata == NULL)
9423 		goto out_put_clp;
9424 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9425 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9426 	msg.rpc_argp = &calldata->args;
9427 	msg.rpc_resp = &calldata->res;
9428 	calldata->clp = clp;
9429 	task_setup_data.callback_data = calldata;
9430 
9431 	ret = rpc_run_task(&task_setup_data);
9432 	if (IS_ERR(ret))
9433 		goto out_err;
9434 	return ret;
9435 out_put_clp:
9436 	nfs_put_client(clp);
9437 out_err:
9438 	nfs41_release_slot(slot);
9439 	return ret;
9440 }
9441 
9442 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9443 {
9444 	struct rpc_task *task;
9445 	int ret = 0;
9446 
9447 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9448 		return -EAGAIN;
9449 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9450 	if (IS_ERR(task))
9451 		ret = PTR_ERR(task);
9452 	else
9453 		rpc_put_task_async(task);
9454 	dprintk("<-- %s status=%d\n", __func__, ret);
9455 	return ret;
9456 }
9457 
9458 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9459 {
9460 	struct rpc_task *task;
9461 	int ret;
9462 
9463 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9464 	if (IS_ERR(task)) {
9465 		ret = PTR_ERR(task);
9466 		goto out;
9467 	}
9468 	ret = rpc_wait_for_completion_task(task);
9469 	if (!ret)
9470 		ret = task->tk_status;
9471 	rpc_put_task(task);
9472 out:
9473 	dprintk("<-- %s status=%d\n", __func__, ret);
9474 	return ret;
9475 }
9476 
9477 struct nfs4_reclaim_complete_data {
9478 	struct nfs_client *clp;
9479 	struct nfs41_reclaim_complete_args arg;
9480 	struct nfs41_reclaim_complete_res res;
9481 };
9482 
9483 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9484 {
9485 	struct nfs4_reclaim_complete_data *calldata = data;
9486 
9487 	nfs4_setup_sequence(calldata->clp,
9488 			&calldata->arg.seq_args,
9489 			&calldata->res.seq_res,
9490 			task);
9491 }
9492 
9493 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9494 {
9495 	switch(task->tk_status) {
9496 	case 0:
9497 		wake_up_all(&clp->cl_lock_waitq);
9498 		fallthrough;
9499 	case -NFS4ERR_COMPLETE_ALREADY:
9500 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9501 		break;
9502 	case -NFS4ERR_DELAY:
9503 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9504 		fallthrough;
9505 	case -NFS4ERR_RETRY_UNCACHED_REP:
9506 	case -EACCES:
9507 		dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9508 			__func__, task->tk_status, clp->cl_hostname);
9509 		return -EAGAIN;
9510 	case -NFS4ERR_BADSESSION:
9511 	case -NFS4ERR_DEADSESSION:
9512 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9513 		break;
9514 	default:
9515 		nfs4_schedule_lease_recovery(clp);
9516 	}
9517 	return 0;
9518 }
9519 
9520 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9521 {
9522 	struct nfs4_reclaim_complete_data *calldata = data;
9523 	struct nfs_client *clp = calldata->clp;
9524 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9525 
9526 	if (!nfs41_sequence_done(task, res))
9527 		return;
9528 
9529 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9530 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9531 		rpc_restart_call_prepare(task);
9532 		return;
9533 	}
9534 }
9535 
9536 static void nfs4_free_reclaim_complete_data(void *data)
9537 {
9538 	struct nfs4_reclaim_complete_data *calldata = data;
9539 
9540 	kfree(calldata);
9541 }
9542 
9543 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9544 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9545 	.rpc_call_done = nfs4_reclaim_complete_done,
9546 	.rpc_release = nfs4_free_reclaim_complete_data,
9547 };
9548 
9549 /*
9550  * Issue a global reclaim complete.
9551  */
9552 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9553 		const struct cred *cred)
9554 {
9555 	struct nfs4_reclaim_complete_data *calldata;
9556 	struct rpc_message msg = {
9557 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9558 		.rpc_cred = cred,
9559 	};
9560 	struct rpc_task_setup task_setup_data = {
9561 		.rpc_client = clp->cl_rpcclient,
9562 		.rpc_message = &msg,
9563 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9564 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9565 	};
9566 	int status = -ENOMEM;
9567 
9568 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9569 	if (calldata == NULL)
9570 		goto out;
9571 	calldata->clp = clp;
9572 	calldata->arg.one_fs = 0;
9573 
9574 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9575 	msg.rpc_argp = &calldata->arg;
9576 	msg.rpc_resp = &calldata->res;
9577 	task_setup_data.callback_data = calldata;
9578 	status = nfs4_call_sync_custom(&task_setup_data);
9579 out:
9580 	dprintk("<-- %s status=%d\n", __func__, status);
9581 	return status;
9582 }
9583 
9584 static void
9585 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9586 {
9587 	struct nfs4_layoutget *lgp = calldata;
9588 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9589 
9590 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9591 				&lgp->res.seq_res, task);
9592 }
9593 
9594 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9595 {
9596 	struct nfs4_layoutget *lgp = calldata;
9597 
9598 	nfs41_sequence_process(task, &lgp->res.seq_res);
9599 }
9600 
9601 static int
9602 nfs4_layoutget_handle_exception(struct rpc_task *task,
9603 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9604 {
9605 	struct inode *inode = lgp->args.inode;
9606 	struct nfs_server *server = NFS_SERVER(inode);
9607 	struct pnfs_layout_hdr *lo = lgp->lo;
9608 	int nfs4err = task->tk_status;
9609 	int err, status = 0;
9610 	LIST_HEAD(head);
9611 
9612 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9613 
9614 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9615 
9616 	switch (nfs4err) {
9617 	case 0:
9618 		goto out;
9619 
9620 	/*
9621 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9622 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9623 	 * retry go inband.
9624 	 */
9625 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9626 		status = -ENODATA;
9627 		goto out;
9628 	/*
9629 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9630 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9631 	 */
9632 	case -NFS4ERR_BADLAYOUT:
9633 		status = -EOVERFLOW;
9634 		goto out;
9635 	/*
9636 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9637 	 * (or clients) writing to the same RAID stripe except when
9638 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9639 	 *
9640 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9641 	 * while, and then eventually give up.
9642 	 */
9643 	case -NFS4ERR_LAYOUTTRYLATER:
9644 		if (lgp->args.minlength == 0) {
9645 			status = -EOVERFLOW;
9646 			goto out;
9647 		}
9648 		status = -EBUSY;
9649 		break;
9650 	case -NFS4ERR_RECALLCONFLICT:
9651 		status = -ERECALLCONFLICT;
9652 		break;
9653 	case -NFS4ERR_DELEG_REVOKED:
9654 	case -NFS4ERR_ADMIN_REVOKED:
9655 	case -NFS4ERR_EXPIRED:
9656 	case -NFS4ERR_BAD_STATEID:
9657 		exception->timeout = 0;
9658 		spin_lock(&inode->i_lock);
9659 		/* If the open stateid was bad, then recover it. */
9660 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9661 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9662 			spin_unlock(&inode->i_lock);
9663 			exception->state = lgp->args.ctx->state;
9664 			exception->stateid = &lgp->args.stateid;
9665 			break;
9666 		}
9667 
9668 		/*
9669 		 * Mark the bad layout state as invalid, then retry
9670 		 */
9671 		pnfs_mark_layout_stateid_invalid(lo, &head);
9672 		spin_unlock(&inode->i_lock);
9673 		nfs_commit_inode(inode, 0);
9674 		pnfs_free_lseg_list(&head);
9675 		status = -EAGAIN;
9676 		goto out;
9677 	}
9678 
9679 	err = nfs4_handle_exception(server, nfs4err, exception);
9680 	if (!status) {
9681 		if (exception->retry)
9682 			status = -EAGAIN;
9683 		else
9684 			status = err;
9685 	}
9686 out:
9687 	return status;
9688 }
9689 
9690 size_t max_response_pages(struct nfs_server *server)
9691 {
9692 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9693 	return nfs_page_array_len(0, max_resp_sz);
9694 }
9695 
9696 static void nfs4_layoutget_release(void *calldata)
9697 {
9698 	struct nfs4_layoutget *lgp = calldata;
9699 
9700 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9701 	pnfs_layoutget_free(lgp);
9702 }
9703 
9704 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9705 	.rpc_call_prepare = nfs4_layoutget_prepare,
9706 	.rpc_call_done = nfs4_layoutget_done,
9707 	.rpc_release = nfs4_layoutget_release,
9708 };
9709 
9710 struct pnfs_layout_segment *
9711 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9712 {
9713 	struct inode *inode = lgp->args.inode;
9714 	struct nfs_server *server = NFS_SERVER(inode);
9715 	struct rpc_task *task;
9716 	struct rpc_message msg = {
9717 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9718 		.rpc_argp = &lgp->args,
9719 		.rpc_resp = &lgp->res,
9720 		.rpc_cred = lgp->cred,
9721 	};
9722 	struct rpc_task_setup task_setup_data = {
9723 		.rpc_client = server->client,
9724 		.rpc_message = &msg,
9725 		.callback_ops = &nfs4_layoutget_call_ops,
9726 		.callback_data = lgp,
9727 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9728 			 RPC_TASK_MOVEABLE,
9729 	};
9730 	struct pnfs_layout_segment *lseg = NULL;
9731 	struct nfs4_exception exception = {
9732 		.inode = inode,
9733 		.timeout = *timeout,
9734 	};
9735 	int status = 0;
9736 
9737 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9738 
9739 	task = rpc_run_task(&task_setup_data);
9740 	if (IS_ERR(task))
9741 		return ERR_CAST(task);
9742 
9743 	status = rpc_wait_for_completion_task(task);
9744 	if (status != 0)
9745 		goto out;
9746 
9747 	if (task->tk_status < 0) {
9748 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9749 		*timeout = exception.timeout;
9750 	} else if (lgp->res.layoutp->len == 0) {
9751 		status = -EAGAIN;
9752 		*timeout = nfs4_update_delay(&exception.timeout);
9753 	} else
9754 		lseg = pnfs_layout_process(lgp);
9755 out:
9756 	trace_nfs4_layoutget(lgp->args.ctx,
9757 			&lgp->args.range,
9758 			&lgp->res.range,
9759 			&lgp->res.stateid,
9760 			status);
9761 
9762 	rpc_put_task(task);
9763 	dprintk("<-- %s status=%d\n", __func__, status);
9764 	if (status)
9765 		return ERR_PTR(status);
9766 	return lseg;
9767 }
9768 
9769 static void
9770 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9771 {
9772 	struct nfs4_layoutreturn *lrp = calldata;
9773 
9774 	nfs4_setup_sequence(lrp->clp,
9775 			&lrp->args.seq_args,
9776 			&lrp->res.seq_res,
9777 			task);
9778 	if (!pnfs_layout_is_valid(lrp->args.layout))
9779 		rpc_exit(task, 0);
9780 }
9781 
9782 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9783 {
9784 	struct nfs4_layoutreturn *lrp = calldata;
9785 	struct nfs_server *server;
9786 
9787 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9788 		return;
9789 
9790 	/*
9791 	 * Was there an RPC level error? Assume the call succeeded,
9792 	 * and that we need to release the layout
9793 	 */
9794 	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9795 		lrp->res.lrs_present = 0;
9796 		return;
9797 	}
9798 
9799 	server = NFS_SERVER(lrp->args.inode);
9800 	switch (task->tk_status) {
9801 	case -NFS4ERR_OLD_STATEID:
9802 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9803 					&lrp->args.range,
9804 					lrp->args.inode))
9805 			goto out_restart;
9806 		fallthrough;
9807 	default:
9808 		task->tk_status = 0;
9809 		fallthrough;
9810 	case 0:
9811 		break;
9812 	case -NFS4ERR_DELAY:
9813 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9814 			break;
9815 		goto out_restart;
9816 	}
9817 	return;
9818 out_restart:
9819 	task->tk_status = 0;
9820 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9821 	rpc_restart_call_prepare(task);
9822 }
9823 
9824 static void nfs4_layoutreturn_release(void *calldata)
9825 {
9826 	struct nfs4_layoutreturn *lrp = calldata;
9827 	struct pnfs_layout_hdr *lo = lrp->args.layout;
9828 
9829 	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9830 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9831 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9832 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9833 		lrp->ld_private.ops->free(&lrp->ld_private);
9834 	pnfs_put_layout_hdr(lrp->args.layout);
9835 	nfs_iput_and_deactive(lrp->inode);
9836 	put_cred(lrp->cred);
9837 	kfree(calldata);
9838 }
9839 
9840 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9841 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9842 	.rpc_call_done = nfs4_layoutreturn_done,
9843 	.rpc_release = nfs4_layoutreturn_release,
9844 };
9845 
9846 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9847 {
9848 	struct rpc_task *task;
9849 	struct rpc_message msg = {
9850 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9851 		.rpc_argp = &lrp->args,
9852 		.rpc_resp = &lrp->res,
9853 		.rpc_cred = lrp->cred,
9854 	};
9855 	struct rpc_task_setup task_setup_data = {
9856 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9857 		.rpc_message = &msg,
9858 		.callback_ops = &nfs4_layoutreturn_call_ops,
9859 		.callback_data = lrp,
9860 		.flags = RPC_TASK_MOVEABLE,
9861 	};
9862 	int status = 0;
9863 
9864 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9865 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9866 			&task_setup_data.rpc_client, &msg);
9867 
9868 	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9869 	if (!sync) {
9870 		if (!lrp->inode) {
9871 			nfs4_layoutreturn_release(lrp);
9872 			return -EAGAIN;
9873 		}
9874 		task_setup_data.flags |= RPC_TASK_ASYNC;
9875 	}
9876 	if (!lrp->inode)
9877 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9878 				   1);
9879 	else
9880 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9881 				   0);
9882 	task = rpc_run_task(&task_setup_data);
9883 	if (IS_ERR(task))
9884 		return PTR_ERR(task);
9885 	if (sync)
9886 		status = task->tk_status;
9887 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9888 	dprintk("<-- %s status=%d\n", __func__, status);
9889 	rpc_put_task(task);
9890 	return status;
9891 }
9892 
9893 static int
9894 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9895 		struct pnfs_device *pdev,
9896 		const struct cred *cred)
9897 {
9898 	struct nfs4_getdeviceinfo_args args = {
9899 		.pdev = pdev,
9900 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9901 			NOTIFY_DEVICEID4_DELETE,
9902 	};
9903 	struct nfs4_getdeviceinfo_res res = {
9904 		.pdev = pdev,
9905 	};
9906 	struct rpc_message msg = {
9907 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9908 		.rpc_argp = &args,
9909 		.rpc_resp = &res,
9910 		.rpc_cred = cred,
9911 	};
9912 	int status;
9913 
9914 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9915 	if (res.notification & ~args.notify_types)
9916 		dprintk("%s: unsupported notification\n", __func__);
9917 	if (res.notification != args.notify_types)
9918 		pdev->nocache = 1;
9919 
9920 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9921 
9922 	dprintk("<-- %s status=%d\n", __func__, status);
9923 
9924 	return status;
9925 }
9926 
9927 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9928 		struct pnfs_device *pdev,
9929 		const struct cred *cred)
9930 {
9931 	struct nfs4_exception exception = { };
9932 	int err;
9933 
9934 	do {
9935 		err = nfs4_handle_exception(server,
9936 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9937 					&exception);
9938 	} while (exception.retry);
9939 	return err;
9940 }
9941 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9942 
9943 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9944 {
9945 	struct nfs4_layoutcommit_data *data = calldata;
9946 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9947 
9948 	nfs4_setup_sequence(server->nfs_client,
9949 			&data->args.seq_args,
9950 			&data->res.seq_res,
9951 			task);
9952 }
9953 
9954 static void
9955 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9956 {
9957 	struct nfs4_layoutcommit_data *data = calldata;
9958 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9959 
9960 	if (!nfs41_sequence_done(task, &data->res.seq_res))
9961 		return;
9962 
9963 	switch (task->tk_status) { /* Just ignore these failures */
9964 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9965 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9966 	case -NFS4ERR_BADLAYOUT:     /* no layout */
9967 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
9968 		task->tk_status = 0;
9969 		break;
9970 	case 0:
9971 		break;
9972 	default:
9973 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9974 			rpc_restart_call_prepare(task);
9975 			return;
9976 		}
9977 	}
9978 }
9979 
9980 static void nfs4_layoutcommit_release(void *calldata)
9981 {
9982 	struct nfs4_layoutcommit_data *data = calldata;
9983 
9984 	pnfs_cleanup_layoutcommit(data);
9985 	nfs_post_op_update_inode_force_wcc(data->args.inode,
9986 					   data->res.fattr);
9987 	put_cred(data->cred);
9988 	nfs_iput_and_deactive(data->inode);
9989 	kfree(data);
9990 }
9991 
9992 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9993 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
9994 	.rpc_call_done = nfs4_layoutcommit_done,
9995 	.rpc_release = nfs4_layoutcommit_release,
9996 };
9997 
9998 int
9999 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10000 {
10001 	struct rpc_message msg = {
10002 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10003 		.rpc_argp = &data->args,
10004 		.rpc_resp = &data->res,
10005 		.rpc_cred = data->cred,
10006 	};
10007 	struct rpc_task_setup task_setup_data = {
10008 		.task = &data->task,
10009 		.rpc_client = NFS_CLIENT(data->args.inode),
10010 		.rpc_message = &msg,
10011 		.callback_ops = &nfs4_layoutcommit_ops,
10012 		.callback_data = data,
10013 		.flags = RPC_TASK_MOVEABLE,
10014 	};
10015 	struct rpc_task *task;
10016 	int status = 0;
10017 
10018 	dprintk("NFS: initiating layoutcommit call. sync %d "
10019 		"lbw: %llu inode %lu\n", sync,
10020 		data->args.lastbytewritten,
10021 		data->args.inode->i_ino);
10022 
10023 	if (!sync) {
10024 		data->inode = nfs_igrab_and_active(data->args.inode);
10025 		if (data->inode == NULL) {
10026 			nfs4_layoutcommit_release(data);
10027 			return -EAGAIN;
10028 		}
10029 		task_setup_data.flags = RPC_TASK_ASYNC;
10030 	}
10031 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10032 	task = rpc_run_task(&task_setup_data);
10033 	if (IS_ERR(task))
10034 		return PTR_ERR(task);
10035 	if (sync)
10036 		status = task->tk_status;
10037 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10038 	dprintk("%s: status %d\n", __func__, status);
10039 	rpc_put_task(task);
10040 	return status;
10041 }
10042 
10043 /*
10044  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10045  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10046  */
10047 static int
10048 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10049 		    struct nfs_fsinfo *info,
10050 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10051 {
10052 	struct nfs41_secinfo_no_name_args args = {
10053 		.style = SECINFO_STYLE_CURRENT_FH,
10054 	};
10055 	struct nfs4_secinfo_res res = {
10056 		.flavors = flavors,
10057 	};
10058 	struct rpc_message msg = {
10059 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10060 		.rpc_argp = &args,
10061 		.rpc_resp = &res,
10062 	};
10063 	struct nfs4_call_sync_data data = {
10064 		.seq_server = server,
10065 		.seq_args = &args.seq_args,
10066 		.seq_res = &res.seq_res,
10067 	};
10068 	struct rpc_task_setup task_setup = {
10069 		.rpc_client = server->client,
10070 		.rpc_message = &msg,
10071 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10072 		.callback_data = &data,
10073 		.flags = RPC_TASK_NO_ROUND_ROBIN,
10074 	};
10075 	const struct cred *cred = NULL;
10076 	int status;
10077 
10078 	if (use_integrity) {
10079 		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10080 
10081 		cred = nfs4_get_clid_cred(server->nfs_client);
10082 		msg.rpc_cred = cred;
10083 	}
10084 
10085 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10086 	status = nfs4_call_sync_custom(&task_setup);
10087 	dprintk("<-- %s status=%d\n", __func__, status);
10088 
10089 	put_cred(cred);
10090 
10091 	return status;
10092 }
10093 
10094 static int
10095 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10096 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10097 {
10098 	struct nfs4_exception exception = {
10099 		.interruptible = true,
10100 	};
10101 	int err;
10102 	do {
10103 		/* first try using integrity protection */
10104 		err = -NFS4ERR_WRONGSEC;
10105 
10106 		/* try to use integrity protection with machine cred */
10107 		if (_nfs4_is_integrity_protected(server->nfs_client))
10108 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10109 							  flavors, true);
10110 
10111 		/*
10112 		 * if unable to use integrity protection, or SECINFO with
10113 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10114 		 * disallowed by spec, but exists in deployed servers) use
10115 		 * the current filesystem's rpc_client and the user cred.
10116 		 */
10117 		if (err == -NFS4ERR_WRONGSEC)
10118 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10119 							  flavors, false);
10120 
10121 		switch (err) {
10122 		case 0:
10123 		case -NFS4ERR_WRONGSEC:
10124 		case -ENOTSUPP:
10125 			goto out;
10126 		default:
10127 			err = nfs4_handle_exception(server, err, &exception);
10128 		}
10129 	} while (exception.retry);
10130 out:
10131 	return err;
10132 }
10133 
10134 static int
10135 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10136 		    struct nfs_fsinfo *info)
10137 {
10138 	int err;
10139 	struct page *page;
10140 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10141 	struct nfs4_secinfo_flavors *flavors;
10142 	struct nfs4_secinfo4 *secinfo;
10143 	int i;
10144 
10145 	page = alloc_page(GFP_KERNEL);
10146 	if (!page) {
10147 		err = -ENOMEM;
10148 		goto out;
10149 	}
10150 
10151 	flavors = page_address(page);
10152 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10153 
10154 	/*
10155 	 * Fall back on "guess and check" method if
10156 	 * the server doesn't support SECINFO_NO_NAME
10157 	 */
10158 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10159 		err = nfs4_find_root_sec(server, fhandle, info);
10160 		goto out_freepage;
10161 	}
10162 	if (err)
10163 		goto out_freepage;
10164 
10165 	for (i = 0; i < flavors->num_flavors; i++) {
10166 		secinfo = &flavors->flavors[i];
10167 
10168 		switch (secinfo->flavor) {
10169 		case RPC_AUTH_NULL:
10170 		case RPC_AUTH_UNIX:
10171 		case RPC_AUTH_GSS:
10172 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10173 					&secinfo->flavor_info);
10174 			break;
10175 		default:
10176 			flavor = RPC_AUTH_MAXFLAVOR;
10177 			break;
10178 		}
10179 
10180 		if (!nfs_auth_info_match(&server->auth_info, flavor))
10181 			flavor = RPC_AUTH_MAXFLAVOR;
10182 
10183 		if (flavor != RPC_AUTH_MAXFLAVOR) {
10184 			err = nfs4_lookup_root_sec(server, fhandle,
10185 						   info, flavor);
10186 			if (!err)
10187 				break;
10188 		}
10189 	}
10190 
10191 	if (flavor == RPC_AUTH_MAXFLAVOR)
10192 		err = -EPERM;
10193 
10194 out_freepage:
10195 	put_page(page);
10196 	if (err == -EACCES)
10197 		return -EPERM;
10198 out:
10199 	return err;
10200 }
10201 
10202 static int _nfs41_test_stateid(struct nfs_server *server,
10203 		nfs4_stateid *stateid,
10204 		const struct cred *cred)
10205 {
10206 	int status;
10207 	struct nfs41_test_stateid_args args = {
10208 		.stateid = stateid,
10209 	};
10210 	struct nfs41_test_stateid_res res;
10211 	struct rpc_message msg = {
10212 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10213 		.rpc_argp = &args,
10214 		.rpc_resp = &res,
10215 		.rpc_cred = cred,
10216 	};
10217 	struct rpc_clnt *rpc_client = server->client;
10218 
10219 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10220 		&rpc_client, &msg);
10221 
10222 	dprintk("NFS call  test_stateid %p\n", stateid);
10223 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10224 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10225 			&args.seq_args, &res.seq_res);
10226 	if (status != NFS_OK) {
10227 		dprintk("NFS reply test_stateid: failed, %d\n", status);
10228 		return status;
10229 	}
10230 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10231 	return -res.status;
10232 }
10233 
10234 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10235 		int err, struct nfs4_exception *exception)
10236 {
10237 	exception->retry = 0;
10238 	switch(err) {
10239 	case -NFS4ERR_DELAY:
10240 	case -NFS4ERR_RETRY_UNCACHED_REP:
10241 		nfs4_handle_exception(server, err, exception);
10242 		break;
10243 	case -NFS4ERR_BADSESSION:
10244 	case -NFS4ERR_BADSLOT:
10245 	case -NFS4ERR_BAD_HIGH_SLOT:
10246 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10247 	case -NFS4ERR_DEADSESSION:
10248 		nfs4_do_handle_exception(server, err, exception);
10249 	}
10250 }
10251 
10252 /**
10253  * nfs41_test_stateid - perform a TEST_STATEID operation
10254  *
10255  * @server: server / transport on which to perform the operation
10256  * @stateid: state ID to test
10257  * @cred: credential
10258  *
10259  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10260  * Otherwise a negative NFS4ERR value is returned if the operation
10261  * failed or the state ID is not currently valid.
10262  */
10263 static int nfs41_test_stateid(struct nfs_server *server,
10264 		nfs4_stateid *stateid,
10265 		const struct cred *cred)
10266 {
10267 	struct nfs4_exception exception = {
10268 		.interruptible = true,
10269 	};
10270 	int err;
10271 	do {
10272 		err = _nfs41_test_stateid(server, stateid, cred);
10273 		nfs4_handle_delay_or_session_error(server, err, &exception);
10274 	} while (exception.retry);
10275 	return err;
10276 }
10277 
10278 struct nfs_free_stateid_data {
10279 	struct nfs_server *server;
10280 	struct nfs41_free_stateid_args args;
10281 	struct nfs41_free_stateid_res res;
10282 };
10283 
10284 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10285 {
10286 	struct nfs_free_stateid_data *data = calldata;
10287 	nfs4_setup_sequence(data->server->nfs_client,
10288 			&data->args.seq_args,
10289 			&data->res.seq_res,
10290 			task);
10291 }
10292 
10293 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10294 {
10295 	struct nfs_free_stateid_data *data = calldata;
10296 
10297 	nfs41_sequence_done(task, &data->res.seq_res);
10298 
10299 	switch (task->tk_status) {
10300 	case -NFS4ERR_DELAY:
10301 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10302 			rpc_restart_call_prepare(task);
10303 	}
10304 }
10305 
10306 static void nfs41_free_stateid_release(void *calldata)
10307 {
10308 	struct nfs_free_stateid_data *data = calldata;
10309 	struct nfs_client *clp = data->server->nfs_client;
10310 
10311 	nfs_put_client(clp);
10312 	kfree(calldata);
10313 }
10314 
10315 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10316 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10317 	.rpc_call_done = nfs41_free_stateid_done,
10318 	.rpc_release = nfs41_free_stateid_release,
10319 };
10320 
10321 /**
10322  * nfs41_free_stateid - perform a FREE_STATEID operation
10323  *
10324  * @server: server / transport on which to perform the operation
10325  * @stateid: state ID to release
10326  * @cred: credential
10327  * @privileged: set to true if this call needs to be privileged
10328  *
10329  * Note: this function is always asynchronous.
10330  */
10331 static int nfs41_free_stateid(struct nfs_server *server,
10332 		const nfs4_stateid *stateid,
10333 		const struct cred *cred,
10334 		bool privileged)
10335 {
10336 	struct rpc_message msg = {
10337 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10338 		.rpc_cred = cred,
10339 	};
10340 	struct rpc_task_setup task_setup = {
10341 		.rpc_client = server->client,
10342 		.rpc_message = &msg,
10343 		.callback_ops = &nfs41_free_stateid_ops,
10344 		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10345 	};
10346 	struct nfs_free_stateid_data *data;
10347 	struct rpc_task *task;
10348 	struct nfs_client *clp = server->nfs_client;
10349 
10350 	if (!refcount_inc_not_zero(&clp->cl_count))
10351 		return -EIO;
10352 
10353 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10354 		&task_setup.rpc_client, &msg);
10355 
10356 	dprintk("NFS call  free_stateid %p\n", stateid);
10357 	data = kmalloc(sizeof(*data), GFP_KERNEL);
10358 	if (!data)
10359 		return -ENOMEM;
10360 	data->server = server;
10361 	nfs4_stateid_copy(&data->args.stateid, stateid);
10362 
10363 	task_setup.callback_data = data;
10364 
10365 	msg.rpc_argp = &data->args;
10366 	msg.rpc_resp = &data->res;
10367 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10368 	task = rpc_run_task(&task_setup);
10369 	if (IS_ERR(task))
10370 		return PTR_ERR(task);
10371 	rpc_put_task(task);
10372 	return 0;
10373 }
10374 
10375 static void
10376 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10377 {
10378 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10379 
10380 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10381 	nfs4_free_lock_state(server, lsp);
10382 }
10383 
10384 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10385 		const nfs4_stateid *s2)
10386 {
10387 	if (s1->type != s2->type)
10388 		return false;
10389 
10390 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10391 		return false;
10392 
10393 	if (s1->seqid == s2->seqid)
10394 		return true;
10395 
10396 	return s1->seqid == 0 || s2->seqid == 0;
10397 }
10398 
10399 #endif /* CONFIG_NFS_V4_1 */
10400 
10401 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10402 		const nfs4_stateid *s2)
10403 {
10404 	return nfs4_stateid_match(s1, s2);
10405 }
10406 
10407 
10408 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10409 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10410 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10411 	.recover_open	= nfs4_open_reclaim,
10412 	.recover_lock	= nfs4_lock_reclaim,
10413 	.establish_clid = nfs4_init_clientid,
10414 	.detect_trunking = nfs40_discover_server_trunking,
10415 };
10416 
10417 #if defined(CONFIG_NFS_V4_1)
10418 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10419 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10420 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10421 	.recover_open	= nfs4_open_reclaim,
10422 	.recover_lock	= nfs4_lock_reclaim,
10423 	.establish_clid = nfs41_init_clientid,
10424 	.reclaim_complete = nfs41_proc_reclaim_complete,
10425 	.detect_trunking = nfs41_discover_server_trunking,
10426 };
10427 #endif /* CONFIG_NFS_V4_1 */
10428 
10429 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10430 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10431 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10432 	.recover_open	= nfs40_open_expired,
10433 	.recover_lock	= nfs4_lock_expired,
10434 	.establish_clid = nfs4_init_clientid,
10435 };
10436 
10437 #if defined(CONFIG_NFS_V4_1)
10438 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10439 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10440 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10441 	.recover_open	= nfs41_open_expired,
10442 	.recover_lock	= nfs41_lock_expired,
10443 	.establish_clid = nfs41_init_clientid,
10444 };
10445 #endif /* CONFIG_NFS_V4_1 */
10446 
10447 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10448 	.sched_state_renewal = nfs4_proc_async_renew,
10449 	.get_state_renewal_cred = nfs4_get_renew_cred,
10450 	.renew_lease = nfs4_proc_renew,
10451 };
10452 
10453 #if defined(CONFIG_NFS_V4_1)
10454 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10455 	.sched_state_renewal = nfs41_proc_async_sequence,
10456 	.get_state_renewal_cred = nfs4_get_machine_cred,
10457 	.renew_lease = nfs4_proc_sequence,
10458 };
10459 #endif
10460 
10461 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10462 	.get_locations = _nfs40_proc_get_locations,
10463 	.fsid_present = _nfs40_proc_fsid_present,
10464 };
10465 
10466 #if defined(CONFIG_NFS_V4_1)
10467 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10468 	.get_locations = _nfs41_proc_get_locations,
10469 	.fsid_present = _nfs41_proc_fsid_present,
10470 };
10471 #endif	/* CONFIG_NFS_V4_1 */
10472 
10473 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10474 	.minor_version = 0,
10475 	.init_caps = NFS_CAP_READDIRPLUS
10476 		| NFS_CAP_ATOMIC_OPEN
10477 		| NFS_CAP_POSIX_LOCK,
10478 	.init_client = nfs40_init_client,
10479 	.shutdown_client = nfs40_shutdown_client,
10480 	.match_stateid = nfs4_match_stateid,
10481 	.find_root_sec = nfs4_find_root_sec,
10482 	.free_lock_state = nfs4_release_lockowner,
10483 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10484 	.alloc_seqid = nfs_alloc_seqid,
10485 	.call_sync_ops = &nfs40_call_sync_ops,
10486 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10487 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10488 	.state_renewal_ops = &nfs40_state_renewal_ops,
10489 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10490 };
10491 
10492 #if defined(CONFIG_NFS_V4_1)
10493 static struct nfs_seqid *
10494 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10495 {
10496 	return NULL;
10497 }
10498 
10499 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10500 	.minor_version = 1,
10501 	.init_caps = NFS_CAP_READDIRPLUS
10502 		| NFS_CAP_ATOMIC_OPEN
10503 		| NFS_CAP_POSIX_LOCK
10504 		| NFS_CAP_STATEID_NFSV41
10505 		| NFS_CAP_ATOMIC_OPEN_V1
10506 		| NFS_CAP_LGOPEN
10507 		| NFS_CAP_MOVEABLE,
10508 	.init_client = nfs41_init_client,
10509 	.shutdown_client = nfs41_shutdown_client,
10510 	.match_stateid = nfs41_match_stateid,
10511 	.find_root_sec = nfs41_find_root_sec,
10512 	.free_lock_state = nfs41_free_lock_state,
10513 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10514 	.alloc_seqid = nfs_alloc_no_seqid,
10515 	.session_trunk = nfs4_test_session_trunk,
10516 	.call_sync_ops = &nfs41_call_sync_ops,
10517 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10518 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10519 	.state_renewal_ops = &nfs41_state_renewal_ops,
10520 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10521 };
10522 #endif
10523 
10524 #if defined(CONFIG_NFS_V4_2)
10525 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10526 	.minor_version = 2,
10527 	.init_caps = NFS_CAP_READDIRPLUS
10528 		| NFS_CAP_ATOMIC_OPEN
10529 		| NFS_CAP_POSIX_LOCK
10530 		| NFS_CAP_STATEID_NFSV41
10531 		| NFS_CAP_ATOMIC_OPEN_V1
10532 		| NFS_CAP_LGOPEN
10533 		| NFS_CAP_ALLOCATE
10534 		| NFS_CAP_COPY
10535 		| NFS_CAP_OFFLOAD_CANCEL
10536 		| NFS_CAP_COPY_NOTIFY
10537 		| NFS_CAP_DEALLOCATE
10538 		| NFS_CAP_SEEK
10539 		| NFS_CAP_LAYOUTSTATS
10540 		| NFS_CAP_CLONE
10541 		| NFS_CAP_LAYOUTERROR
10542 		| NFS_CAP_READ_PLUS
10543 		| NFS_CAP_MOVEABLE,
10544 	.init_client = nfs41_init_client,
10545 	.shutdown_client = nfs41_shutdown_client,
10546 	.match_stateid = nfs41_match_stateid,
10547 	.find_root_sec = nfs41_find_root_sec,
10548 	.free_lock_state = nfs41_free_lock_state,
10549 	.call_sync_ops = &nfs41_call_sync_ops,
10550 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10551 	.alloc_seqid = nfs_alloc_no_seqid,
10552 	.session_trunk = nfs4_test_session_trunk,
10553 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10554 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10555 	.state_renewal_ops = &nfs41_state_renewal_ops,
10556 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10557 };
10558 #endif
10559 
10560 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10561 	[0] = &nfs_v4_0_minor_ops,
10562 #if defined(CONFIG_NFS_V4_1)
10563 	[1] = &nfs_v4_1_minor_ops,
10564 #endif
10565 #if defined(CONFIG_NFS_V4_2)
10566 	[2] = &nfs_v4_2_minor_ops,
10567 #endif
10568 };
10569 
10570 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10571 {
10572 	ssize_t error, error2, error3;
10573 
10574 	error = generic_listxattr(dentry, list, size);
10575 	if (error < 0)
10576 		return error;
10577 	if (list) {
10578 		list += error;
10579 		size -= error;
10580 	}
10581 
10582 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10583 	if (error2 < 0)
10584 		return error2;
10585 
10586 	if (list) {
10587 		list += error2;
10588 		size -= error2;
10589 	}
10590 
10591 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10592 	if (error3 < 0)
10593 		return error3;
10594 
10595 	return error + error2 + error3;
10596 }
10597 
10598 static void nfs4_enable_swap(struct inode *inode)
10599 {
10600 	/* The state manager thread must always be running.
10601 	 * It will notice the client is a swapper, and stay put.
10602 	 */
10603 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10604 
10605 	nfs4_schedule_state_manager(clp);
10606 }
10607 
10608 static void nfs4_disable_swap(struct inode *inode)
10609 {
10610 	/* The state manager thread will now exit once it is
10611 	 * woken.
10612 	 */
10613 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10614 
10615 	nfs4_schedule_state_manager(clp);
10616 }
10617 
10618 static const struct inode_operations nfs4_dir_inode_operations = {
10619 	.create		= nfs_create,
10620 	.lookup		= nfs_lookup,
10621 	.atomic_open	= nfs_atomic_open,
10622 	.link		= nfs_link,
10623 	.unlink		= nfs_unlink,
10624 	.symlink	= nfs_symlink,
10625 	.mkdir		= nfs_mkdir,
10626 	.rmdir		= nfs_rmdir,
10627 	.mknod		= nfs_mknod,
10628 	.rename		= nfs_rename,
10629 	.permission	= nfs_permission,
10630 	.getattr	= nfs_getattr,
10631 	.setattr	= nfs_setattr,
10632 	.listxattr	= nfs4_listxattr,
10633 };
10634 
10635 static const struct inode_operations nfs4_file_inode_operations = {
10636 	.permission	= nfs_permission,
10637 	.getattr	= nfs_getattr,
10638 	.setattr	= nfs_setattr,
10639 	.listxattr	= nfs4_listxattr,
10640 };
10641 
10642 const struct nfs_rpc_ops nfs_v4_clientops = {
10643 	.version	= 4,			/* protocol version */
10644 	.dentry_ops	= &nfs4_dentry_operations,
10645 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10646 	.file_inode_ops	= &nfs4_file_inode_operations,
10647 	.file_ops	= &nfs4_file_operations,
10648 	.getroot	= nfs4_proc_get_root,
10649 	.submount	= nfs4_submount,
10650 	.try_get_tree	= nfs4_try_get_tree,
10651 	.getattr	= nfs4_proc_getattr,
10652 	.setattr	= nfs4_proc_setattr,
10653 	.lookup		= nfs4_proc_lookup,
10654 	.lookupp	= nfs4_proc_lookupp,
10655 	.access		= nfs4_proc_access,
10656 	.readlink	= nfs4_proc_readlink,
10657 	.create		= nfs4_proc_create,
10658 	.remove		= nfs4_proc_remove,
10659 	.unlink_setup	= nfs4_proc_unlink_setup,
10660 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10661 	.unlink_done	= nfs4_proc_unlink_done,
10662 	.rename_setup	= nfs4_proc_rename_setup,
10663 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10664 	.rename_done	= nfs4_proc_rename_done,
10665 	.link		= nfs4_proc_link,
10666 	.symlink	= nfs4_proc_symlink,
10667 	.mkdir		= nfs4_proc_mkdir,
10668 	.rmdir		= nfs4_proc_rmdir,
10669 	.readdir	= nfs4_proc_readdir,
10670 	.mknod		= nfs4_proc_mknod,
10671 	.statfs		= nfs4_proc_statfs,
10672 	.fsinfo		= nfs4_proc_fsinfo,
10673 	.pathconf	= nfs4_proc_pathconf,
10674 	.set_capabilities = nfs4_server_capabilities,
10675 	.decode_dirent	= nfs4_decode_dirent,
10676 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10677 	.read_setup	= nfs4_proc_read_setup,
10678 	.read_done	= nfs4_read_done,
10679 	.write_setup	= nfs4_proc_write_setup,
10680 	.write_done	= nfs4_write_done,
10681 	.commit_setup	= nfs4_proc_commit_setup,
10682 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10683 	.commit_done	= nfs4_commit_done,
10684 	.lock		= nfs4_proc_lock,
10685 	.clear_acl_cache = nfs4_zap_acl_attr,
10686 	.close_context  = nfs4_close_context,
10687 	.open_context	= nfs4_atomic_open,
10688 	.have_delegation = nfs4_have_delegation,
10689 	.alloc_client	= nfs4_alloc_client,
10690 	.init_client	= nfs4_init_client,
10691 	.free_client	= nfs4_free_client,
10692 	.create_server	= nfs4_create_server,
10693 	.clone_server	= nfs_clone_server,
10694 	.discover_trunking = nfs4_discover_trunking,
10695 	.enable_swap	= nfs4_enable_swap,
10696 	.disable_swap	= nfs4_disable_swap,
10697 };
10698 
10699 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10700 	.name	= XATTR_NAME_NFSV4_ACL,
10701 	.list	= nfs4_xattr_list_nfs4_acl,
10702 	.get	= nfs4_xattr_get_nfs4_acl,
10703 	.set	= nfs4_xattr_set_nfs4_acl,
10704 };
10705 
10706 #if defined(CONFIG_NFS_V4_1)
10707 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10708 	.name	= XATTR_NAME_NFSV4_DACL,
10709 	.list	= nfs4_xattr_list_nfs4_dacl,
10710 	.get	= nfs4_xattr_get_nfs4_dacl,
10711 	.set	= nfs4_xattr_set_nfs4_dacl,
10712 };
10713 
10714 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10715 	.name	= XATTR_NAME_NFSV4_SACL,
10716 	.list	= nfs4_xattr_list_nfs4_sacl,
10717 	.get	= nfs4_xattr_get_nfs4_sacl,
10718 	.set	= nfs4_xattr_set_nfs4_sacl,
10719 };
10720 #endif
10721 
10722 #ifdef CONFIG_NFS_V4_2
10723 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10724 	.prefix	= XATTR_USER_PREFIX,
10725 	.get	= nfs4_xattr_get_nfs4_user,
10726 	.set	= nfs4_xattr_set_nfs4_user,
10727 };
10728 #endif
10729 
10730 const struct xattr_handler *nfs4_xattr_handlers[] = {
10731 	&nfs4_xattr_nfs4_acl_handler,
10732 #if defined(CONFIG_NFS_V4_1)
10733 	&nfs4_xattr_nfs4_dacl_handler,
10734 	&nfs4_xattr_nfs4_sacl_handler,
10735 #endif
10736 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10737 	&nfs4_xattr_nfs4_label_handler,
10738 #endif
10739 #ifdef CONFIG_NFS_V4_2
10740 	&nfs4_xattr_nfs4_user_handler,
10741 #endif
10742 	NULL
10743 };
10744