xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision e818bd08)
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 	err = security_dentry_init_security(dentry, sattr->ia_mode,
126 				&dentry->d_name, NULL,
127 				(void **)&label->label, &label->len);
128 	if (err == 0)
129 		return label;
130 
131 	return NULL;
132 }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 {
136 	if (label)
137 		security_release_secctx(label->label, label->len);
138 }
139 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
140 {
141 	if (label)
142 		return server->attr_bitmask;
143 
144 	return server->attr_bitmask_nl;
145 }
146 #else
147 static inline struct nfs4_label *
148 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
149 	struct iattr *sattr, struct nfs4_label *l)
150 { return NULL; }
151 static inline void
152 nfs4_label_release_security(struct nfs4_label *label)
153 { return; }
154 static inline u32 *
155 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
156 { return server->attr_bitmask; }
157 #endif
158 
159 /* Prevent leaks of NFSv4 errors into userland */
160 static int nfs4_map_errors(int err)
161 {
162 	if (err >= -1000)
163 		return err;
164 	switch (err) {
165 	case -NFS4ERR_RESOURCE:
166 	case -NFS4ERR_LAYOUTTRYLATER:
167 	case -NFS4ERR_RECALLCONFLICT:
168 		return -EREMOTEIO;
169 	case -NFS4ERR_WRONGSEC:
170 	case -NFS4ERR_WRONG_CRED:
171 		return -EPERM;
172 	case -NFS4ERR_BADOWNER:
173 	case -NFS4ERR_BADNAME:
174 		return -EINVAL;
175 	case -NFS4ERR_SHARE_DENIED:
176 		return -EACCES;
177 	case -NFS4ERR_MINOR_VERS_MISMATCH:
178 		return -EPROTONOSUPPORT;
179 	case -NFS4ERR_FILE_OPEN:
180 		return -EBUSY;
181 	case -NFS4ERR_NOT_SAME:
182 		return -ENOTSYNC;
183 	default:
184 		dprintk("%s could not handle NFSv4 error %d\n",
185 				__func__, -err);
186 		break;
187 	}
188 	return -EIO;
189 }
190 
191 /*
192  * This is our standard bitmap for GETATTR requests.
193  */
194 const u32 nfs4_fattr_bitmap[3] = {
195 	FATTR4_WORD0_TYPE
196 	| FATTR4_WORD0_CHANGE
197 	| FATTR4_WORD0_SIZE
198 	| FATTR4_WORD0_FSID
199 	| FATTR4_WORD0_FILEID,
200 	FATTR4_WORD1_MODE
201 	| FATTR4_WORD1_NUMLINKS
202 	| FATTR4_WORD1_OWNER
203 	| FATTR4_WORD1_OWNER_GROUP
204 	| FATTR4_WORD1_RAWDEV
205 	| FATTR4_WORD1_SPACE_USED
206 	| FATTR4_WORD1_TIME_ACCESS
207 	| FATTR4_WORD1_TIME_METADATA
208 	| FATTR4_WORD1_TIME_MODIFY
209 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
210 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
211 	FATTR4_WORD2_SECURITY_LABEL
212 #endif
213 };
214 
215 static const u32 nfs4_pnfs_open_bitmap[3] = {
216 	FATTR4_WORD0_TYPE
217 	| FATTR4_WORD0_CHANGE
218 	| FATTR4_WORD0_SIZE
219 	| FATTR4_WORD0_FSID
220 	| FATTR4_WORD0_FILEID,
221 	FATTR4_WORD1_MODE
222 	| FATTR4_WORD1_NUMLINKS
223 	| FATTR4_WORD1_OWNER
224 	| FATTR4_WORD1_OWNER_GROUP
225 	| FATTR4_WORD1_RAWDEV
226 	| FATTR4_WORD1_SPACE_USED
227 	| FATTR4_WORD1_TIME_ACCESS
228 	| FATTR4_WORD1_TIME_METADATA
229 	| FATTR4_WORD1_TIME_MODIFY,
230 	FATTR4_WORD2_MDSTHRESHOLD
231 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
232 	| FATTR4_WORD2_SECURITY_LABEL
233 #endif
234 };
235 
236 static const u32 nfs4_open_noattr_bitmap[3] = {
237 	FATTR4_WORD0_TYPE
238 	| FATTR4_WORD0_FILEID,
239 };
240 
241 const u32 nfs4_statfs_bitmap[3] = {
242 	FATTR4_WORD0_FILES_AVAIL
243 	| FATTR4_WORD0_FILES_FREE
244 	| FATTR4_WORD0_FILES_TOTAL,
245 	FATTR4_WORD1_SPACE_AVAIL
246 	| FATTR4_WORD1_SPACE_FREE
247 	| FATTR4_WORD1_SPACE_TOTAL
248 };
249 
250 const u32 nfs4_pathconf_bitmap[3] = {
251 	FATTR4_WORD0_MAXLINK
252 	| FATTR4_WORD0_MAXNAME,
253 	0
254 };
255 
256 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
257 			| FATTR4_WORD0_MAXREAD
258 			| FATTR4_WORD0_MAXWRITE
259 			| FATTR4_WORD0_LEASE_TIME,
260 			FATTR4_WORD1_TIME_DELTA
261 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
262 			FATTR4_WORD2_LAYOUT_BLKSIZE
263 			| FATTR4_WORD2_CLONE_BLKSIZE
264 			| FATTR4_WORD2_CHANGE_ATTR_TYPE
265 			| FATTR4_WORD2_XATTR_SUPPORT
266 };
267 
268 const u32 nfs4_fs_locations_bitmap[3] = {
269 	FATTR4_WORD0_CHANGE
270 	| FATTR4_WORD0_SIZE
271 	| FATTR4_WORD0_FSID
272 	| FATTR4_WORD0_FILEID
273 	| FATTR4_WORD0_FS_LOCATIONS,
274 	FATTR4_WORD1_OWNER
275 	| FATTR4_WORD1_OWNER_GROUP
276 	| FATTR4_WORD1_RAWDEV
277 	| FATTR4_WORD1_SPACE_USED
278 	| FATTR4_WORD1_TIME_ACCESS
279 	| FATTR4_WORD1_TIME_METADATA
280 	| FATTR4_WORD1_TIME_MODIFY
281 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
282 };
283 
284 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
285 				    struct inode *inode, unsigned long flags)
286 {
287 	unsigned long cache_validity;
288 
289 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
290 	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
291 		return;
292 
293 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
294 
295 	/* Remove the attributes over which we have full control */
296 	dst[1] &= ~FATTR4_WORD1_RAWDEV;
297 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
298 		dst[0] &= ~FATTR4_WORD0_SIZE;
299 
300 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
301 		dst[0] &= ~FATTR4_WORD0_CHANGE;
302 
303 	if (!(cache_validity & NFS_INO_INVALID_MODE))
304 		dst[1] &= ~FATTR4_WORD1_MODE;
305 	if (!(cache_validity & NFS_INO_INVALID_OTHER))
306 		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
307 }
308 
309 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
310 		struct nfs4_readdir_arg *readdir)
311 {
312 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
313 	__be32 *start, *p;
314 
315 	if (cookie > 2) {
316 		readdir->cookie = cookie;
317 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
318 		return;
319 	}
320 
321 	readdir->cookie = 0;
322 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
323 	if (cookie == 2)
324 		return;
325 
326 	/*
327 	 * NFSv4 servers do not return entries for '.' and '..'
328 	 * Therefore, we fake these entries here.  We let '.'
329 	 * have cookie 0 and '..' have cookie 1.  Note that
330 	 * when talking to the server, we always send cookie 0
331 	 * instead of 1 or 2.
332 	 */
333 	start = p = kmap_atomic(*readdir->pages);
334 
335 	if (cookie == 0) {
336 		*p++ = xdr_one;                                  /* next */
337 		*p++ = xdr_zero;                   /* cookie, first word */
338 		*p++ = xdr_one;                   /* cookie, second word */
339 		*p++ = xdr_one;                             /* entry len */
340 		memcpy(p, ".\0\0\0", 4);                        /* entry */
341 		p++;
342 		*p++ = xdr_one;                         /* bitmap length */
343 		*p++ = htonl(attrs);                           /* bitmap */
344 		*p++ = htonl(12);             /* attribute buffer length */
345 		*p++ = htonl(NF4DIR);
346 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
347 	}
348 
349 	*p++ = xdr_one;                                  /* next */
350 	*p++ = xdr_zero;                   /* cookie, first word */
351 	*p++ = xdr_two;                   /* cookie, second word */
352 	*p++ = xdr_two;                             /* entry len */
353 	memcpy(p, "..\0\0", 4);                         /* entry */
354 	p++;
355 	*p++ = xdr_one;                         /* bitmap length */
356 	*p++ = htonl(attrs);                           /* bitmap */
357 	*p++ = htonl(12);             /* attribute buffer length */
358 	*p++ = htonl(NF4DIR);
359 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
360 
361 	readdir->pgbase = (char *)p - (char *)start;
362 	readdir->count -= readdir->pgbase;
363 	kunmap_atomic(start);
364 }
365 
366 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
367 {
368 	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
369 		fattr->pre_change_attr = version;
370 		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
371 	}
372 }
373 
374 static void nfs4_test_and_free_stateid(struct nfs_server *server,
375 		nfs4_stateid *stateid,
376 		const struct cred *cred)
377 {
378 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
379 
380 	ops->test_and_free_expired(server, stateid, cred);
381 }
382 
383 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
384 		nfs4_stateid *stateid,
385 		const struct cred *cred)
386 {
387 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
388 	nfs4_test_and_free_stateid(server, stateid, cred);
389 }
390 
391 static void nfs4_free_revoked_stateid(struct nfs_server *server,
392 		const nfs4_stateid *stateid,
393 		const struct cred *cred)
394 {
395 	nfs4_stateid tmp;
396 
397 	nfs4_stateid_copy(&tmp, stateid);
398 	__nfs4_free_revoked_stateid(server, &tmp, cred);
399 }
400 
401 static long nfs4_update_delay(long *timeout)
402 {
403 	long ret;
404 	if (!timeout)
405 		return NFS4_POLL_RETRY_MAX;
406 	if (*timeout <= 0)
407 		*timeout = NFS4_POLL_RETRY_MIN;
408 	if (*timeout > NFS4_POLL_RETRY_MAX)
409 		*timeout = NFS4_POLL_RETRY_MAX;
410 	ret = *timeout;
411 	*timeout <<= 1;
412 	return ret;
413 }
414 
415 static int nfs4_delay_killable(long *timeout)
416 {
417 	might_sleep();
418 
419 	freezable_schedule_timeout_killable_unsafe(
420 		nfs4_update_delay(timeout));
421 	if (!__fatal_signal_pending(current))
422 		return 0;
423 	return -EINTR;
424 }
425 
426 static int nfs4_delay_interruptible(long *timeout)
427 {
428 	might_sleep();
429 
430 	freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
431 	if (!signal_pending(current))
432 		return 0;
433 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
434 }
435 
436 static int nfs4_delay(long *timeout, bool interruptible)
437 {
438 	if (interruptible)
439 		return nfs4_delay_interruptible(timeout);
440 	return nfs4_delay_killable(timeout);
441 }
442 
443 static const nfs4_stateid *
444 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
445 {
446 	if (!stateid)
447 		return NULL;
448 	switch (stateid->type) {
449 	case NFS4_OPEN_STATEID_TYPE:
450 	case NFS4_LOCK_STATEID_TYPE:
451 	case NFS4_DELEGATION_STATEID_TYPE:
452 		return stateid;
453 	default:
454 		break;
455 	}
456 	return NULL;
457 }
458 
459 /* This is the error handling routine for processes that are allowed
460  * to sleep.
461  */
462 static int nfs4_do_handle_exception(struct nfs_server *server,
463 		int errorcode, struct nfs4_exception *exception)
464 {
465 	struct nfs_client *clp = server->nfs_client;
466 	struct nfs4_state *state = exception->state;
467 	const nfs4_stateid *stateid;
468 	struct inode *inode = exception->inode;
469 	int ret = errorcode;
470 
471 	exception->delay = 0;
472 	exception->recovering = 0;
473 	exception->retry = 0;
474 
475 	stateid = nfs4_recoverable_stateid(exception->stateid);
476 	if (stateid == NULL && state != NULL)
477 		stateid = nfs4_recoverable_stateid(&state->stateid);
478 
479 	switch(errorcode) {
480 		case 0:
481 			return 0;
482 		case -NFS4ERR_BADHANDLE:
483 		case -ESTALE:
484 			if (inode != NULL && S_ISREG(inode->i_mode))
485 				pnfs_destroy_layout(NFS_I(inode));
486 			break;
487 		case -NFS4ERR_DELEG_REVOKED:
488 		case -NFS4ERR_ADMIN_REVOKED:
489 		case -NFS4ERR_EXPIRED:
490 		case -NFS4ERR_BAD_STATEID:
491 		case -NFS4ERR_PARTNER_NO_AUTH:
492 			if (inode != NULL && stateid != NULL) {
493 				nfs_inode_find_state_and_recover(inode,
494 						stateid);
495 				goto wait_on_recovery;
496 			}
497 			fallthrough;
498 		case -NFS4ERR_OPENMODE:
499 			if (inode) {
500 				int err;
501 
502 				err = nfs_async_inode_return_delegation(inode,
503 						stateid);
504 				if (err == 0)
505 					goto wait_on_recovery;
506 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
507 					exception->retry = 1;
508 					break;
509 				}
510 			}
511 			if (state == NULL)
512 				break;
513 			ret = nfs4_schedule_stateid_recovery(server, state);
514 			if (ret < 0)
515 				break;
516 			goto wait_on_recovery;
517 		case -NFS4ERR_STALE_STATEID:
518 		case -NFS4ERR_STALE_CLIENTID:
519 			nfs4_schedule_lease_recovery(clp);
520 			goto wait_on_recovery;
521 		case -NFS4ERR_MOVED:
522 			ret = nfs4_schedule_migration_recovery(server);
523 			if (ret < 0)
524 				break;
525 			goto wait_on_recovery;
526 		case -NFS4ERR_LEASE_MOVED:
527 			nfs4_schedule_lease_moved_recovery(clp);
528 			goto wait_on_recovery;
529 #if defined(CONFIG_NFS_V4_1)
530 		case -NFS4ERR_BADSESSION:
531 		case -NFS4ERR_BADSLOT:
532 		case -NFS4ERR_BAD_HIGH_SLOT:
533 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
534 		case -NFS4ERR_DEADSESSION:
535 		case -NFS4ERR_SEQ_FALSE_RETRY:
536 		case -NFS4ERR_SEQ_MISORDERED:
537 			/* Handled in nfs41_sequence_process() */
538 			goto wait_on_recovery;
539 #endif /* defined(CONFIG_NFS_V4_1) */
540 		case -NFS4ERR_FILE_OPEN:
541 			if (exception->timeout > HZ) {
542 				/* We have retried a decent amount, time to
543 				 * fail
544 				 */
545 				ret = -EBUSY;
546 				break;
547 			}
548 			fallthrough;
549 		case -NFS4ERR_DELAY:
550 			nfs_inc_server_stats(server, NFSIOS_DELAY);
551 			fallthrough;
552 		case -NFS4ERR_GRACE:
553 		case -NFS4ERR_LAYOUTTRYLATER:
554 		case -NFS4ERR_RECALLCONFLICT:
555 			exception->delay = 1;
556 			return 0;
557 
558 		case -NFS4ERR_RETRY_UNCACHED_REP:
559 		case -NFS4ERR_OLD_STATEID:
560 			exception->retry = 1;
561 			break;
562 		case -NFS4ERR_BADOWNER:
563 			/* The following works around a Linux server bug! */
564 		case -NFS4ERR_BADNAME:
565 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
566 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
567 				exception->retry = 1;
568 				printk(KERN_WARNING "NFS: v4 server %s "
569 						"does not accept raw "
570 						"uid/gids. "
571 						"Reenabling the idmapper.\n",
572 						server->nfs_client->cl_hostname);
573 			}
574 	}
575 	/* We failed to handle the error */
576 	return nfs4_map_errors(ret);
577 wait_on_recovery:
578 	exception->recovering = 1;
579 	return 0;
580 }
581 
582 /* This is the error handling routine for processes that are allowed
583  * to sleep.
584  */
585 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
586 {
587 	struct nfs_client *clp = server->nfs_client;
588 	int ret;
589 
590 	ret = nfs4_do_handle_exception(server, errorcode, exception);
591 	if (exception->delay) {
592 		ret = nfs4_delay(&exception->timeout,
593 				exception->interruptible);
594 		goto out_retry;
595 	}
596 	if (exception->recovering) {
597 		if (exception->task_is_privileged)
598 			return -EDEADLOCK;
599 		ret = nfs4_wait_clnt_recover(clp);
600 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
601 			return -EIO;
602 		goto out_retry;
603 	}
604 	return ret;
605 out_retry:
606 	if (ret == 0)
607 		exception->retry = 1;
608 	return ret;
609 }
610 
611 static int
612 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
613 		int errorcode, struct nfs4_exception *exception)
614 {
615 	struct nfs_client *clp = server->nfs_client;
616 	int ret;
617 
618 	ret = nfs4_do_handle_exception(server, errorcode, exception);
619 	if (exception->delay) {
620 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
621 		goto out_retry;
622 	}
623 	if (exception->recovering) {
624 		if (exception->task_is_privileged)
625 			return -EDEADLOCK;
626 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
627 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
628 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
629 		goto out_retry;
630 	}
631 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
632 		ret = -EIO;
633 	return ret;
634 out_retry:
635 	if (ret == 0) {
636 		exception->retry = 1;
637 		/*
638 		 * For NFS4ERR_MOVED, the client transport will need to
639 		 * be recomputed after migration recovery has completed.
640 		 */
641 		if (errorcode == -NFS4ERR_MOVED)
642 			rpc_task_release_transport(task);
643 	}
644 	return ret;
645 }
646 
647 int
648 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
649 			struct nfs4_state *state, long *timeout)
650 {
651 	struct nfs4_exception exception = {
652 		.state = state,
653 	};
654 
655 	if (task->tk_status >= 0)
656 		return 0;
657 	if (timeout)
658 		exception.timeout = *timeout;
659 	task->tk_status = nfs4_async_handle_exception(task, server,
660 			task->tk_status,
661 			&exception);
662 	if (exception.delay && timeout)
663 		*timeout = exception.timeout;
664 	if (exception.retry)
665 		return -EAGAIN;
666 	return 0;
667 }
668 
669 /*
670  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
671  * or 'false' otherwise.
672  */
673 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
674 {
675 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
676 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
677 }
678 
679 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
680 {
681 	spin_lock(&clp->cl_lock);
682 	if (time_before(clp->cl_last_renewal,timestamp))
683 		clp->cl_last_renewal = timestamp;
684 	spin_unlock(&clp->cl_lock);
685 }
686 
687 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
688 {
689 	struct nfs_client *clp = server->nfs_client;
690 
691 	if (!nfs4_has_session(clp))
692 		do_renew_lease(clp, timestamp);
693 }
694 
695 struct nfs4_call_sync_data {
696 	const struct nfs_server *seq_server;
697 	struct nfs4_sequence_args *seq_args;
698 	struct nfs4_sequence_res *seq_res;
699 };
700 
701 void nfs4_init_sequence(struct nfs4_sequence_args *args,
702 			struct nfs4_sequence_res *res, int cache_reply,
703 			int privileged)
704 {
705 	args->sa_slot = NULL;
706 	args->sa_cache_this = cache_reply;
707 	args->sa_privileged = privileged;
708 
709 	res->sr_slot = NULL;
710 }
711 
712 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
713 {
714 	struct nfs4_slot *slot = res->sr_slot;
715 	struct nfs4_slot_table *tbl;
716 
717 	tbl = slot->table;
718 	spin_lock(&tbl->slot_tbl_lock);
719 	if (!nfs41_wake_and_assign_slot(tbl, slot))
720 		nfs4_free_slot(tbl, slot);
721 	spin_unlock(&tbl->slot_tbl_lock);
722 
723 	res->sr_slot = NULL;
724 }
725 
726 static int nfs40_sequence_done(struct rpc_task *task,
727 			       struct nfs4_sequence_res *res)
728 {
729 	if (res->sr_slot != NULL)
730 		nfs40_sequence_free_slot(res);
731 	return 1;
732 }
733 
734 #if defined(CONFIG_NFS_V4_1)
735 
736 static void nfs41_release_slot(struct nfs4_slot *slot)
737 {
738 	struct nfs4_session *session;
739 	struct nfs4_slot_table *tbl;
740 	bool send_new_highest_used_slotid = false;
741 
742 	if (!slot)
743 		return;
744 	tbl = slot->table;
745 	session = tbl->session;
746 
747 	/* Bump the slot sequence number */
748 	if (slot->seq_done)
749 		slot->seq_nr++;
750 	slot->seq_done = 0;
751 
752 	spin_lock(&tbl->slot_tbl_lock);
753 	/* Be nice to the server: try to ensure that the last transmitted
754 	 * value for highest_user_slotid <= target_highest_slotid
755 	 */
756 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
757 		send_new_highest_used_slotid = true;
758 
759 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
760 		send_new_highest_used_slotid = false;
761 		goto out_unlock;
762 	}
763 	nfs4_free_slot(tbl, slot);
764 
765 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
766 		send_new_highest_used_slotid = false;
767 out_unlock:
768 	spin_unlock(&tbl->slot_tbl_lock);
769 	if (send_new_highest_used_slotid)
770 		nfs41_notify_server(session->clp);
771 	if (waitqueue_active(&tbl->slot_waitq))
772 		wake_up_all(&tbl->slot_waitq);
773 }
774 
775 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
776 {
777 	nfs41_release_slot(res->sr_slot);
778 	res->sr_slot = NULL;
779 }
780 
781 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
782 		u32 seqnr)
783 {
784 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
785 		slot->seq_nr_highest_sent = seqnr;
786 }
787 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
788 {
789 	nfs4_slot_sequence_record_sent(slot, seqnr);
790 	slot->seq_nr_last_acked = seqnr;
791 }
792 
793 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
794 				struct nfs4_slot *slot)
795 {
796 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
797 	if (!IS_ERR(task))
798 		rpc_put_task_async(task);
799 }
800 
801 static int nfs41_sequence_process(struct rpc_task *task,
802 		struct nfs4_sequence_res *res)
803 {
804 	struct nfs4_session *session;
805 	struct nfs4_slot *slot = res->sr_slot;
806 	struct nfs_client *clp;
807 	int status;
808 	int ret = 1;
809 
810 	if (slot == NULL)
811 		goto out_noaction;
812 	/* don't increment the sequence number if the task wasn't sent */
813 	if (!RPC_WAS_SENT(task) || slot->seq_done)
814 		goto out;
815 
816 	session = slot->table->session;
817 	clp = session->clp;
818 
819 	trace_nfs4_sequence_done(session, res);
820 
821 	status = res->sr_status;
822 	if (task->tk_status == -NFS4ERR_DEADSESSION)
823 		status = -NFS4ERR_DEADSESSION;
824 
825 	/* Check the SEQUENCE operation status */
826 	switch (status) {
827 	case 0:
828 		/* Mark this sequence number as having been acked */
829 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
830 		/* Update the slot's sequence and clientid lease timer */
831 		slot->seq_done = 1;
832 		do_renew_lease(clp, res->sr_timestamp);
833 		/* Check sequence flags */
834 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
835 				!!slot->privileged);
836 		nfs41_update_target_slotid(slot->table, slot, res);
837 		break;
838 	case 1:
839 		/*
840 		 * sr_status remains 1 if an RPC level error occurred.
841 		 * The server may or may not have processed the sequence
842 		 * operation..
843 		 */
844 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
845 		slot->seq_done = 1;
846 		goto out;
847 	case -NFS4ERR_DELAY:
848 		/* The server detected a resend of the RPC call and
849 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
850 		 * of RFC5661.
851 		 */
852 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
853 			__func__,
854 			slot->slot_nr,
855 			slot->seq_nr);
856 		goto out_retry;
857 	case -NFS4ERR_RETRY_UNCACHED_REP:
858 	case -NFS4ERR_SEQ_FALSE_RETRY:
859 		/*
860 		 * The server thinks we tried to replay a request.
861 		 * Retry the call after bumping the sequence ID.
862 		 */
863 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
864 		goto retry_new_seq;
865 	case -NFS4ERR_BADSLOT:
866 		/*
867 		 * The slot id we used was probably retired. Try again
868 		 * using a different slot id.
869 		 */
870 		if (slot->slot_nr < slot->table->target_highest_slotid)
871 			goto session_recover;
872 		goto retry_nowait;
873 	case -NFS4ERR_SEQ_MISORDERED:
874 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
875 		/*
876 		 * Were one or more calls using this slot interrupted?
877 		 * If the server never received the request, then our
878 		 * transmitted slot sequence number may be too high. However,
879 		 * if the server did receive the request then it might
880 		 * accidentally give us a reply with a mismatched operation.
881 		 * We can sort this out by sending a lone sequence operation
882 		 * to the server on the same slot.
883 		 */
884 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
885 			slot->seq_nr--;
886 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
887 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
888 				res->sr_slot = NULL;
889 			}
890 			goto retry_nowait;
891 		}
892 		/*
893 		 * RFC5661:
894 		 * A retry might be sent while the original request is
895 		 * still in progress on the replier. The replier SHOULD
896 		 * deal with the issue by returning NFS4ERR_DELAY as the
897 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
898 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
899 		 *
900 		 * Restart the search after a delay.
901 		 */
902 		slot->seq_nr = slot->seq_nr_highest_sent;
903 		goto out_retry;
904 	case -NFS4ERR_BADSESSION:
905 	case -NFS4ERR_DEADSESSION:
906 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
907 		goto session_recover;
908 	default:
909 		/* Just update the slot sequence no. */
910 		slot->seq_done = 1;
911 	}
912 out:
913 	/* The session may be reset by one of the error handlers. */
914 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
915 out_noaction:
916 	return ret;
917 session_recover:
918 	nfs4_schedule_session_recovery(session, status);
919 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
920 	nfs41_sequence_free_slot(res);
921 	goto out;
922 retry_new_seq:
923 	++slot->seq_nr;
924 retry_nowait:
925 	if (rpc_restart_call_prepare(task)) {
926 		nfs41_sequence_free_slot(res);
927 		task->tk_status = 0;
928 		ret = 0;
929 	}
930 	goto out;
931 out_retry:
932 	if (!rpc_restart_call(task))
933 		goto out;
934 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
935 	return 0;
936 }
937 
938 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
939 {
940 	if (!nfs41_sequence_process(task, res))
941 		return 0;
942 	if (res->sr_slot != NULL)
943 		nfs41_sequence_free_slot(res);
944 	return 1;
945 
946 }
947 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
948 
949 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
950 {
951 	if (res->sr_slot == NULL)
952 		return 1;
953 	if (res->sr_slot->table->session != NULL)
954 		return nfs41_sequence_process(task, res);
955 	return nfs40_sequence_done(task, res);
956 }
957 
958 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
959 {
960 	if (res->sr_slot != NULL) {
961 		if (res->sr_slot->table->session != NULL)
962 			nfs41_sequence_free_slot(res);
963 		else
964 			nfs40_sequence_free_slot(res);
965 	}
966 }
967 
968 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
969 {
970 	if (res->sr_slot == NULL)
971 		return 1;
972 	if (!res->sr_slot->table->session)
973 		return nfs40_sequence_done(task, res);
974 	return nfs41_sequence_done(task, res);
975 }
976 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
977 
978 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
979 {
980 	struct nfs4_call_sync_data *data = calldata;
981 
982 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
983 
984 	nfs4_setup_sequence(data->seq_server->nfs_client,
985 			    data->seq_args, data->seq_res, task);
986 }
987 
988 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
989 {
990 	struct nfs4_call_sync_data *data = calldata;
991 
992 	nfs41_sequence_done(task, data->seq_res);
993 }
994 
995 static const struct rpc_call_ops nfs41_call_sync_ops = {
996 	.rpc_call_prepare = nfs41_call_sync_prepare,
997 	.rpc_call_done = nfs41_call_sync_done,
998 };
999 
1000 #else	/* !CONFIG_NFS_V4_1 */
1001 
1002 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1003 {
1004 	return nfs40_sequence_done(task, res);
1005 }
1006 
1007 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1008 {
1009 	if (res->sr_slot != NULL)
1010 		nfs40_sequence_free_slot(res);
1011 }
1012 
1013 int nfs4_sequence_done(struct rpc_task *task,
1014 		       struct nfs4_sequence_res *res)
1015 {
1016 	return nfs40_sequence_done(task, res);
1017 }
1018 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1019 
1020 #endif	/* !CONFIG_NFS_V4_1 */
1021 
1022 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1023 {
1024 	res->sr_timestamp = jiffies;
1025 	res->sr_status_flags = 0;
1026 	res->sr_status = 1;
1027 }
1028 
1029 static
1030 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1031 		struct nfs4_sequence_res *res,
1032 		struct nfs4_slot *slot)
1033 {
1034 	if (!slot)
1035 		return;
1036 	slot->privileged = args->sa_privileged ? 1 : 0;
1037 	args->sa_slot = slot;
1038 
1039 	res->sr_slot = slot;
1040 }
1041 
1042 int nfs4_setup_sequence(struct nfs_client *client,
1043 			struct nfs4_sequence_args *args,
1044 			struct nfs4_sequence_res *res,
1045 			struct rpc_task *task)
1046 {
1047 	struct nfs4_session *session = nfs4_get_session(client);
1048 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1049 	struct nfs4_slot *slot;
1050 
1051 	/* slot already allocated? */
1052 	if (res->sr_slot != NULL)
1053 		goto out_start;
1054 
1055 	if (session)
1056 		tbl = &session->fc_slot_table;
1057 
1058 	spin_lock(&tbl->slot_tbl_lock);
1059 	/* The state manager will wait until the slot table is empty */
1060 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1061 		goto out_sleep;
1062 
1063 	slot = nfs4_alloc_slot(tbl);
1064 	if (IS_ERR(slot)) {
1065 		if (slot == ERR_PTR(-ENOMEM))
1066 			goto out_sleep_timeout;
1067 		goto out_sleep;
1068 	}
1069 	spin_unlock(&tbl->slot_tbl_lock);
1070 
1071 	nfs4_sequence_attach_slot(args, res, slot);
1072 
1073 	trace_nfs4_setup_sequence(session, args);
1074 out_start:
1075 	nfs41_sequence_res_init(res);
1076 	rpc_call_start(task);
1077 	return 0;
1078 out_sleep_timeout:
1079 	/* Try again in 1/4 second */
1080 	if (args->sa_privileged)
1081 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1082 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1083 	else
1084 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1085 				NULL, jiffies + (HZ >> 2));
1086 	spin_unlock(&tbl->slot_tbl_lock);
1087 	return -EAGAIN;
1088 out_sleep:
1089 	if (args->sa_privileged)
1090 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1091 				RPC_PRIORITY_PRIVILEGED);
1092 	else
1093 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1094 	spin_unlock(&tbl->slot_tbl_lock);
1095 	return -EAGAIN;
1096 }
1097 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1098 
1099 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1100 {
1101 	struct nfs4_call_sync_data *data = calldata;
1102 	nfs4_setup_sequence(data->seq_server->nfs_client,
1103 				data->seq_args, data->seq_res, task);
1104 }
1105 
1106 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1107 {
1108 	struct nfs4_call_sync_data *data = calldata;
1109 	nfs4_sequence_done(task, data->seq_res);
1110 }
1111 
1112 static const struct rpc_call_ops nfs40_call_sync_ops = {
1113 	.rpc_call_prepare = nfs40_call_sync_prepare,
1114 	.rpc_call_done = nfs40_call_sync_done,
1115 };
1116 
1117 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1118 {
1119 	int ret;
1120 	struct rpc_task *task;
1121 
1122 	task = rpc_run_task(task_setup);
1123 	if (IS_ERR(task))
1124 		return PTR_ERR(task);
1125 
1126 	ret = task->tk_status;
1127 	rpc_put_task(task);
1128 	return ret;
1129 }
1130 
1131 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1132 			     struct nfs_server *server,
1133 			     struct rpc_message *msg,
1134 			     struct nfs4_sequence_args *args,
1135 			     struct nfs4_sequence_res *res,
1136 			     unsigned short task_flags)
1137 {
1138 	struct nfs_client *clp = server->nfs_client;
1139 	struct nfs4_call_sync_data data = {
1140 		.seq_server = server,
1141 		.seq_args = args,
1142 		.seq_res = res,
1143 	};
1144 	struct rpc_task_setup task_setup = {
1145 		.rpc_client = clnt,
1146 		.rpc_message = msg,
1147 		.callback_ops = clp->cl_mvops->call_sync_ops,
1148 		.callback_data = &data,
1149 		.flags = task_flags,
1150 	};
1151 
1152 	return nfs4_call_sync_custom(&task_setup);
1153 }
1154 
1155 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1156 				   struct nfs_server *server,
1157 				   struct rpc_message *msg,
1158 				   struct nfs4_sequence_args *args,
1159 				   struct nfs4_sequence_res *res)
1160 {
1161 	unsigned short task_flags = 0;
1162 
1163 	if (server->caps & NFS_CAP_MOVEABLE)
1164 		task_flags = RPC_TASK_MOVEABLE;
1165 	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1166 }
1167 
1168 
1169 int nfs4_call_sync(struct rpc_clnt *clnt,
1170 		   struct nfs_server *server,
1171 		   struct rpc_message *msg,
1172 		   struct nfs4_sequence_args *args,
1173 		   struct nfs4_sequence_res *res,
1174 		   int cache_reply)
1175 {
1176 	nfs4_init_sequence(args, res, cache_reply, 0);
1177 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1178 }
1179 
1180 static void
1181 nfs4_inc_nlink_locked(struct inode *inode)
1182 {
1183 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1184 					     NFS_INO_INVALID_CTIME |
1185 					     NFS_INO_INVALID_NLINK);
1186 	inc_nlink(inode);
1187 }
1188 
1189 static void
1190 nfs4_inc_nlink(struct inode *inode)
1191 {
1192 	spin_lock(&inode->i_lock);
1193 	nfs4_inc_nlink_locked(inode);
1194 	spin_unlock(&inode->i_lock);
1195 }
1196 
1197 static void
1198 nfs4_dec_nlink_locked(struct inode *inode)
1199 {
1200 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1201 					     NFS_INO_INVALID_CTIME |
1202 					     NFS_INO_INVALID_NLINK);
1203 	drop_nlink(inode);
1204 }
1205 
1206 static void
1207 nfs4_update_changeattr_locked(struct inode *inode,
1208 		struct nfs4_change_info *cinfo,
1209 		unsigned long timestamp, unsigned long cache_validity)
1210 {
1211 	struct nfs_inode *nfsi = NFS_I(inode);
1212 	u64 change_attr = inode_peek_iversion_raw(inode);
1213 
1214 	cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1215 	if (S_ISDIR(inode->i_mode))
1216 		cache_validity |= NFS_INO_INVALID_DATA;
1217 
1218 	switch (NFS_SERVER(inode)->change_attr_type) {
1219 	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1220 		if (cinfo->after == change_attr)
1221 			goto out;
1222 		break;
1223 	default:
1224 		if ((s64)(change_attr - cinfo->after) >= 0)
1225 			goto out;
1226 	}
1227 
1228 	inode_set_iversion_raw(inode, cinfo->after);
1229 	if (!cinfo->atomic || cinfo->before != change_attr) {
1230 		if (S_ISDIR(inode->i_mode))
1231 			nfs_force_lookup_revalidate(inode);
1232 
1233 		if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1234 			cache_validity |=
1235 				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1236 				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1237 				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1238 				NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1239 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1240 	}
1241 	nfsi->attrtimeo_timestamp = jiffies;
1242 	nfsi->read_cache_jiffies = timestamp;
1243 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1244 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1245 out:
1246 	nfs_set_cache_invalid(inode, cache_validity);
1247 }
1248 
1249 void
1250 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1251 		unsigned long timestamp, unsigned long cache_validity)
1252 {
1253 	spin_lock(&dir->i_lock);
1254 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1255 	spin_unlock(&dir->i_lock);
1256 }
1257 
1258 struct nfs4_open_createattrs {
1259 	struct nfs4_label *label;
1260 	struct iattr *sattr;
1261 	const __u32 verf[2];
1262 };
1263 
1264 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1265 		int err, struct nfs4_exception *exception)
1266 {
1267 	if (err != -EINVAL)
1268 		return false;
1269 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1270 		return false;
1271 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1272 	exception->retry = 1;
1273 	return true;
1274 }
1275 
1276 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1277 {
1278 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1279 }
1280 
1281 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1282 {
1283 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1284 
1285 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1286 }
1287 
1288 static u32
1289 nfs4_map_atomic_open_share(struct nfs_server *server,
1290 		fmode_t fmode, int openflags)
1291 {
1292 	u32 res = 0;
1293 
1294 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1295 	case FMODE_READ:
1296 		res = NFS4_SHARE_ACCESS_READ;
1297 		break;
1298 	case FMODE_WRITE:
1299 		res = NFS4_SHARE_ACCESS_WRITE;
1300 		break;
1301 	case FMODE_READ|FMODE_WRITE:
1302 		res = NFS4_SHARE_ACCESS_BOTH;
1303 	}
1304 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1305 		goto out;
1306 	/* Want no delegation if we're using O_DIRECT */
1307 	if (openflags & O_DIRECT)
1308 		res |= NFS4_SHARE_WANT_NO_DELEG;
1309 out:
1310 	return res;
1311 }
1312 
1313 static enum open_claim_type4
1314 nfs4_map_atomic_open_claim(struct nfs_server *server,
1315 		enum open_claim_type4 claim)
1316 {
1317 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1318 		return claim;
1319 	switch (claim) {
1320 	default:
1321 		return claim;
1322 	case NFS4_OPEN_CLAIM_FH:
1323 		return NFS4_OPEN_CLAIM_NULL;
1324 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1325 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1326 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1327 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1328 	}
1329 }
1330 
1331 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1332 {
1333 	p->o_res.f_attr = &p->f_attr;
1334 	p->o_res.seqid = p->o_arg.seqid;
1335 	p->c_res.seqid = p->c_arg.seqid;
1336 	p->o_res.server = p->o_arg.server;
1337 	p->o_res.access_request = p->o_arg.access;
1338 	nfs_fattr_init(&p->f_attr);
1339 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1340 }
1341 
1342 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1343 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1344 		const struct nfs4_open_createattrs *c,
1345 		enum open_claim_type4 claim,
1346 		gfp_t gfp_mask)
1347 {
1348 	struct dentry *parent = dget_parent(dentry);
1349 	struct inode *dir = d_inode(parent);
1350 	struct nfs_server *server = NFS_SERVER(dir);
1351 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1352 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1353 	struct nfs4_opendata *p;
1354 
1355 	p = kzalloc(sizeof(*p), gfp_mask);
1356 	if (p == NULL)
1357 		goto err;
1358 
1359 	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1360 	if (IS_ERR(p->f_attr.label))
1361 		goto err_free_p;
1362 
1363 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1364 	if (IS_ERR(p->a_label))
1365 		goto err_free_f;
1366 
1367 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1368 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1369 	if (IS_ERR(p->o_arg.seqid))
1370 		goto err_free_label;
1371 	nfs_sb_active(dentry->d_sb);
1372 	p->dentry = dget(dentry);
1373 	p->dir = parent;
1374 	p->owner = sp;
1375 	atomic_inc(&sp->so_count);
1376 	p->o_arg.open_flags = flags;
1377 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1378 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1379 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1380 			fmode, flags);
1381 	if (flags & O_CREAT) {
1382 		p->o_arg.umask = current_umask();
1383 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1384 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1385 			p->o_arg.u.attrs = &p->attrs;
1386 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1387 
1388 			memcpy(p->o_arg.u.verifier.data, c->verf,
1389 					sizeof(p->o_arg.u.verifier.data));
1390 		}
1391 	}
1392 	/* ask server to check for all possible rights as results
1393 	 * are cached */
1394 	switch (p->o_arg.claim) {
1395 	default:
1396 		break;
1397 	case NFS4_OPEN_CLAIM_NULL:
1398 	case NFS4_OPEN_CLAIM_FH:
1399 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1400 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1401 				  NFS4_ACCESS_EXECUTE |
1402 				  nfs_access_xattr_mask(server);
1403 	}
1404 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1405 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1406 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1407 	p->o_arg.name = &dentry->d_name;
1408 	p->o_arg.server = server;
1409 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1410 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1411 	switch (p->o_arg.claim) {
1412 	case NFS4_OPEN_CLAIM_NULL:
1413 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1414 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1415 		p->o_arg.fh = NFS_FH(dir);
1416 		break;
1417 	case NFS4_OPEN_CLAIM_PREVIOUS:
1418 	case NFS4_OPEN_CLAIM_FH:
1419 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1420 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1421 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1422 	}
1423 	p->c_arg.fh = &p->o_res.fh;
1424 	p->c_arg.stateid = &p->o_res.stateid;
1425 	p->c_arg.seqid = p->o_arg.seqid;
1426 	nfs4_init_opendata_res(p);
1427 	kref_init(&p->kref);
1428 	return p;
1429 
1430 err_free_label:
1431 	nfs4_label_free(p->a_label);
1432 err_free_f:
1433 	nfs4_label_free(p->f_attr.label);
1434 err_free_p:
1435 	kfree(p);
1436 err:
1437 	dput(parent);
1438 	return NULL;
1439 }
1440 
1441 static void nfs4_opendata_free(struct kref *kref)
1442 {
1443 	struct nfs4_opendata *p = container_of(kref,
1444 			struct nfs4_opendata, kref);
1445 	struct super_block *sb = p->dentry->d_sb;
1446 
1447 	nfs4_lgopen_release(p->lgp);
1448 	nfs_free_seqid(p->o_arg.seqid);
1449 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1450 	if (p->state != NULL)
1451 		nfs4_put_open_state(p->state);
1452 	nfs4_put_state_owner(p->owner);
1453 
1454 	nfs4_label_free(p->a_label);
1455 	nfs4_label_free(p->f_attr.label);
1456 
1457 	dput(p->dir);
1458 	dput(p->dentry);
1459 	nfs_sb_deactive(sb);
1460 	nfs_fattr_free_names(&p->f_attr);
1461 	kfree(p->f_attr.mdsthreshold);
1462 	kfree(p);
1463 }
1464 
1465 static void nfs4_opendata_put(struct nfs4_opendata *p)
1466 {
1467 	if (p != NULL)
1468 		kref_put(&p->kref, nfs4_opendata_free);
1469 }
1470 
1471 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1472 		fmode_t fmode)
1473 {
1474 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1475 	case FMODE_READ|FMODE_WRITE:
1476 		return state->n_rdwr != 0;
1477 	case FMODE_WRITE:
1478 		return state->n_wronly != 0;
1479 	case FMODE_READ:
1480 		return state->n_rdonly != 0;
1481 	}
1482 	WARN_ON_ONCE(1);
1483 	return false;
1484 }
1485 
1486 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1487 		int open_mode, enum open_claim_type4 claim)
1488 {
1489 	int ret = 0;
1490 
1491 	if (open_mode & (O_EXCL|O_TRUNC))
1492 		goto out;
1493 	switch (claim) {
1494 	case NFS4_OPEN_CLAIM_NULL:
1495 	case NFS4_OPEN_CLAIM_FH:
1496 		goto out;
1497 	default:
1498 		break;
1499 	}
1500 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1501 		case FMODE_READ:
1502 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1503 				&& state->n_rdonly != 0;
1504 			break;
1505 		case FMODE_WRITE:
1506 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1507 				&& state->n_wronly != 0;
1508 			break;
1509 		case FMODE_READ|FMODE_WRITE:
1510 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1511 				&& state->n_rdwr != 0;
1512 	}
1513 out:
1514 	return ret;
1515 }
1516 
1517 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1518 		enum open_claim_type4 claim)
1519 {
1520 	if (delegation == NULL)
1521 		return 0;
1522 	if ((delegation->type & fmode) != fmode)
1523 		return 0;
1524 	switch (claim) {
1525 	case NFS4_OPEN_CLAIM_NULL:
1526 	case NFS4_OPEN_CLAIM_FH:
1527 		break;
1528 	case NFS4_OPEN_CLAIM_PREVIOUS:
1529 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1530 			break;
1531 		fallthrough;
1532 	default:
1533 		return 0;
1534 	}
1535 	nfs_mark_delegation_referenced(delegation);
1536 	return 1;
1537 }
1538 
1539 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1540 {
1541 	switch (fmode) {
1542 		case FMODE_WRITE:
1543 			state->n_wronly++;
1544 			break;
1545 		case FMODE_READ:
1546 			state->n_rdonly++;
1547 			break;
1548 		case FMODE_READ|FMODE_WRITE:
1549 			state->n_rdwr++;
1550 	}
1551 	nfs4_state_set_mode_locked(state, state->state | fmode);
1552 }
1553 
1554 #ifdef CONFIG_NFS_V4_1
1555 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1556 {
1557 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1558 		return true;
1559 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1560 		return true;
1561 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1562 		return true;
1563 	return false;
1564 }
1565 #endif /* CONFIG_NFS_V4_1 */
1566 
1567 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1568 {
1569 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1570 		wake_up_all(&state->waitq);
1571 }
1572 
1573 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1574 {
1575 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1576 	bool need_recover = false;
1577 
1578 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1579 		need_recover = true;
1580 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1581 		need_recover = true;
1582 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1583 		need_recover = true;
1584 	if (need_recover)
1585 		nfs4_state_mark_reclaim_nograce(clp, state);
1586 }
1587 
1588 /*
1589  * Check for whether or not the caller may update the open stateid
1590  * to the value passed in by stateid.
1591  *
1592  * Note: This function relies heavily on the server implementing
1593  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1594  * correctly.
1595  * i.e. The stateid seqids have to be initialised to 1, and
1596  * are then incremented on every state transition.
1597  */
1598 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1599 		const nfs4_stateid *stateid)
1600 {
1601 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1602 		/* The common case - we're updating to a new sequence number */
1603 		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1604 			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1605 				return true;
1606 			return false;
1607 		}
1608 		/* The server returned a new stateid */
1609 	}
1610 	/* This is the first OPEN in this generation */
1611 	if (stateid->seqid == cpu_to_be32(1))
1612 		return true;
1613 	return false;
1614 }
1615 
1616 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1617 {
1618 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1619 		return;
1620 	if (state->n_wronly)
1621 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1622 	if (state->n_rdonly)
1623 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1624 	if (state->n_rdwr)
1625 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1626 	set_bit(NFS_OPEN_STATE, &state->flags);
1627 }
1628 
1629 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1630 		nfs4_stateid *stateid, fmode_t fmode)
1631 {
1632 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1633 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1634 	case FMODE_WRITE:
1635 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1636 		break;
1637 	case FMODE_READ:
1638 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1639 		break;
1640 	case 0:
1641 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1642 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1643 		clear_bit(NFS_OPEN_STATE, &state->flags);
1644 	}
1645 	if (stateid == NULL)
1646 		return;
1647 	/* Handle OPEN+OPEN_DOWNGRADE races */
1648 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1649 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1650 		nfs_resync_open_stateid_locked(state);
1651 		goto out;
1652 	}
1653 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1654 		nfs4_stateid_copy(&state->stateid, stateid);
1655 	nfs4_stateid_copy(&state->open_stateid, stateid);
1656 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1657 out:
1658 	nfs_state_log_update_open_stateid(state);
1659 }
1660 
1661 static void nfs_clear_open_stateid(struct nfs4_state *state,
1662 	nfs4_stateid *arg_stateid,
1663 	nfs4_stateid *stateid, fmode_t fmode)
1664 {
1665 	write_seqlock(&state->seqlock);
1666 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1667 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1668 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1669 	write_sequnlock(&state->seqlock);
1670 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1671 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1672 }
1673 
1674 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1675 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1676 	__must_hold(&state->owner->so_lock)
1677 	__must_hold(&state->seqlock)
1678 	__must_hold(RCU)
1679 
1680 {
1681 	DEFINE_WAIT(wait);
1682 	int status = 0;
1683 	for (;;) {
1684 
1685 		if (nfs_stateid_is_sequential(state, stateid))
1686 			break;
1687 
1688 		if (status)
1689 			break;
1690 		/* Rely on seqids for serialisation with NFSv4.0 */
1691 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1692 			break;
1693 
1694 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1695 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1696 		/*
1697 		 * Ensure we process the state changes in the same order
1698 		 * in which the server processed them by delaying the
1699 		 * update of the stateid until we are in sequence.
1700 		 */
1701 		write_sequnlock(&state->seqlock);
1702 		spin_unlock(&state->owner->so_lock);
1703 		rcu_read_unlock();
1704 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1705 
1706 		if (!fatal_signal_pending(current)) {
1707 			if (schedule_timeout(5*HZ) == 0)
1708 				status = -EAGAIN;
1709 			else
1710 				status = 0;
1711 		} else
1712 			status = -EINTR;
1713 		finish_wait(&state->waitq, &wait);
1714 		rcu_read_lock();
1715 		spin_lock(&state->owner->so_lock);
1716 		write_seqlock(&state->seqlock);
1717 	}
1718 
1719 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1720 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1721 		nfs4_stateid_copy(freeme, &state->open_stateid);
1722 		nfs_test_and_clear_all_open_stateid(state);
1723 	}
1724 
1725 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1726 		nfs4_stateid_copy(&state->stateid, stateid);
1727 	nfs4_stateid_copy(&state->open_stateid, stateid);
1728 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1729 	nfs_state_log_update_open_stateid(state);
1730 }
1731 
1732 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1733 		const nfs4_stateid *open_stateid,
1734 		fmode_t fmode,
1735 		nfs4_stateid *freeme)
1736 {
1737 	/*
1738 	 * Protect the call to nfs4_state_set_mode_locked and
1739 	 * serialise the stateid update
1740 	 */
1741 	write_seqlock(&state->seqlock);
1742 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1743 	switch (fmode) {
1744 	case FMODE_READ:
1745 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1746 		break;
1747 	case FMODE_WRITE:
1748 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1749 		break;
1750 	case FMODE_READ|FMODE_WRITE:
1751 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1752 	}
1753 	set_bit(NFS_OPEN_STATE, &state->flags);
1754 	write_sequnlock(&state->seqlock);
1755 }
1756 
1757 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1758 {
1759 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1760 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1761 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1762 	clear_bit(NFS_OPEN_STATE, &state->flags);
1763 }
1764 
1765 static void nfs_state_set_delegation(struct nfs4_state *state,
1766 		const nfs4_stateid *deleg_stateid,
1767 		fmode_t fmode)
1768 {
1769 	/*
1770 	 * Protect the call to nfs4_state_set_mode_locked and
1771 	 * serialise the stateid update
1772 	 */
1773 	write_seqlock(&state->seqlock);
1774 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1775 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1776 	write_sequnlock(&state->seqlock);
1777 }
1778 
1779 static void nfs_state_clear_delegation(struct nfs4_state *state)
1780 {
1781 	write_seqlock(&state->seqlock);
1782 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1783 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1784 	write_sequnlock(&state->seqlock);
1785 }
1786 
1787 int update_open_stateid(struct nfs4_state *state,
1788 		const nfs4_stateid *open_stateid,
1789 		const nfs4_stateid *delegation,
1790 		fmode_t fmode)
1791 {
1792 	struct nfs_server *server = NFS_SERVER(state->inode);
1793 	struct nfs_client *clp = server->nfs_client;
1794 	struct nfs_inode *nfsi = NFS_I(state->inode);
1795 	struct nfs_delegation *deleg_cur;
1796 	nfs4_stateid freeme = { };
1797 	int ret = 0;
1798 
1799 	fmode &= (FMODE_READ|FMODE_WRITE);
1800 
1801 	rcu_read_lock();
1802 	spin_lock(&state->owner->so_lock);
1803 	if (open_stateid != NULL) {
1804 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1805 		ret = 1;
1806 	}
1807 
1808 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1809 	if (deleg_cur == NULL)
1810 		goto no_delegation;
1811 
1812 	spin_lock(&deleg_cur->lock);
1813 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1814 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1815 	    (deleg_cur->type & fmode) != fmode)
1816 		goto no_delegation_unlock;
1817 
1818 	if (delegation == NULL)
1819 		delegation = &deleg_cur->stateid;
1820 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1821 		goto no_delegation_unlock;
1822 
1823 	nfs_mark_delegation_referenced(deleg_cur);
1824 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1825 	ret = 1;
1826 no_delegation_unlock:
1827 	spin_unlock(&deleg_cur->lock);
1828 no_delegation:
1829 	if (ret)
1830 		update_open_stateflags(state, fmode);
1831 	spin_unlock(&state->owner->so_lock);
1832 	rcu_read_unlock();
1833 
1834 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1835 		nfs4_schedule_state_manager(clp);
1836 	if (freeme.type != 0)
1837 		nfs4_test_and_free_stateid(server, &freeme,
1838 				state->owner->so_cred);
1839 
1840 	return ret;
1841 }
1842 
1843 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1844 		const nfs4_stateid *stateid)
1845 {
1846 	struct nfs4_state *state = lsp->ls_state;
1847 	bool ret = false;
1848 
1849 	spin_lock(&state->state_lock);
1850 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1851 		goto out_noupdate;
1852 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1853 		goto out_noupdate;
1854 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1855 	ret = true;
1856 out_noupdate:
1857 	spin_unlock(&state->state_lock);
1858 	return ret;
1859 }
1860 
1861 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1862 {
1863 	struct nfs_delegation *delegation;
1864 
1865 	fmode &= FMODE_READ|FMODE_WRITE;
1866 	rcu_read_lock();
1867 	delegation = nfs4_get_valid_delegation(inode);
1868 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1869 		rcu_read_unlock();
1870 		return;
1871 	}
1872 	rcu_read_unlock();
1873 	nfs4_inode_return_delegation(inode);
1874 }
1875 
1876 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1877 {
1878 	struct nfs4_state *state = opendata->state;
1879 	struct nfs_delegation *delegation;
1880 	int open_mode = opendata->o_arg.open_flags;
1881 	fmode_t fmode = opendata->o_arg.fmode;
1882 	enum open_claim_type4 claim = opendata->o_arg.claim;
1883 	nfs4_stateid stateid;
1884 	int ret = -EAGAIN;
1885 
1886 	for (;;) {
1887 		spin_lock(&state->owner->so_lock);
1888 		if (can_open_cached(state, fmode, open_mode, claim)) {
1889 			update_open_stateflags(state, fmode);
1890 			spin_unlock(&state->owner->so_lock);
1891 			goto out_return_state;
1892 		}
1893 		spin_unlock(&state->owner->so_lock);
1894 		rcu_read_lock();
1895 		delegation = nfs4_get_valid_delegation(state->inode);
1896 		if (!can_open_delegated(delegation, fmode, claim)) {
1897 			rcu_read_unlock();
1898 			break;
1899 		}
1900 		/* Save the delegation */
1901 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1902 		rcu_read_unlock();
1903 		nfs_release_seqid(opendata->o_arg.seqid);
1904 		if (!opendata->is_recover) {
1905 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1906 			if (ret != 0)
1907 				goto out;
1908 		}
1909 		ret = -EAGAIN;
1910 
1911 		/* Try to update the stateid using the delegation */
1912 		if (update_open_stateid(state, NULL, &stateid, fmode))
1913 			goto out_return_state;
1914 	}
1915 out:
1916 	return ERR_PTR(ret);
1917 out_return_state:
1918 	refcount_inc(&state->count);
1919 	return state;
1920 }
1921 
1922 static void
1923 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1924 {
1925 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1926 	struct nfs_delegation *delegation;
1927 	int delegation_flags = 0;
1928 
1929 	rcu_read_lock();
1930 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1931 	if (delegation)
1932 		delegation_flags = delegation->flags;
1933 	rcu_read_unlock();
1934 	switch (data->o_arg.claim) {
1935 	default:
1936 		break;
1937 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1938 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1939 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1940 				   "returning a delegation for "
1941 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1942 				   clp->cl_hostname);
1943 		return;
1944 	}
1945 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1946 		nfs_inode_set_delegation(state->inode,
1947 				data->owner->so_cred,
1948 				data->o_res.delegation_type,
1949 				&data->o_res.delegation,
1950 				data->o_res.pagemod_limit);
1951 	else
1952 		nfs_inode_reclaim_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 
1958 	if (data->o_res.do_recall)
1959 		nfs_async_inode_return_delegation(state->inode,
1960 						  &data->o_res.delegation);
1961 }
1962 
1963 /*
1964  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1965  * and update the nfs4_state.
1966  */
1967 static struct nfs4_state *
1968 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1969 {
1970 	struct inode *inode = data->state->inode;
1971 	struct nfs4_state *state = data->state;
1972 	int ret;
1973 
1974 	if (!data->rpc_done) {
1975 		if (data->rpc_status)
1976 			return ERR_PTR(data->rpc_status);
1977 		/* cached opens have already been processed */
1978 		goto update;
1979 	}
1980 
1981 	ret = nfs_refresh_inode(inode, &data->f_attr);
1982 	if (ret)
1983 		return ERR_PTR(ret);
1984 
1985 	if (data->o_res.delegation_type != 0)
1986 		nfs4_opendata_check_deleg(data, state);
1987 update:
1988 	if (!update_open_stateid(state, &data->o_res.stateid,
1989 				NULL, data->o_arg.fmode))
1990 		return ERR_PTR(-EAGAIN);
1991 	refcount_inc(&state->count);
1992 
1993 	return state;
1994 }
1995 
1996 static struct inode *
1997 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1998 {
1999 	struct inode *inode;
2000 
2001 	switch (data->o_arg.claim) {
2002 	case NFS4_OPEN_CLAIM_NULL:
2003 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2004 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2005 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2006 			return ERR_PTR(-EAGAIN);
2007 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2008 				&data->f_attr);
2009 		break;
2010 	default:
2011 		inode = d_inode(data->dentry);
2012 		ihold(inode);
2013 		nfs_refresh_inode(inode, &data->f_attr);
2014 	}
2015 	return inode;
2016 }
2017 
2018 static struct nfs4_state *
2019 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2020 {
2021 	struct nfs4_state *state;
2022 	struct inode *inode;
2023 
2024 	inode = nfs4_opendata_get_inode(data);
2025 	if (IS_ERR(inode))
2026 		return ERR_CAST(inode);
2027 	if (data->state != NULL && data->state->inode == inode) {
2028 		state = data->state;
2029 		refcount_inc(&state->count);
2030 	} else
2031 		state = nfs4_get_open_state(inode, data->owner);
2032 	iput(inode);
2033 	if (state == NULL)
2034 		state = ERR_PTR(-ENOMEM);
2035 	return state;
2036 }
2037 
2038 static struct nfs4_state *
2039 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2040 {
2041 	struct nfs4_state *state;
2042 
2043 	if (!data->rpc_done) {
2044 		state = nfs4_try_open_cached(data);
2045 		trace_nfs4_cached_open(data->state);
2046 		goto out;
2047 	}
2048 
2049 	state = nfs4_opendata_find_nfs4_state(data);
2050 	if (IS_ERR(state))
2051 		goto out;
2052 
2053 	if (data->o_res.delegation_type != 0)
2054 		nfs4_opendata_check_deleg(data, state);
2055 	if (!update_open_stateid(state, &data->o_res.stateid,
2056 				NULL, data->o_arg.fmode)) {
2057 		nfs4_put_open_state(state);
2058 		state = ERR_PTR(-EAGAIN);
2059 	}
2060 out:
2061 	nfs_release_seqid(data->o_arg.seqid);
2062 	return state;
2063 }
2064 
2065 static struct nfs4_state *
2066 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2067 {
2068 	struct nfs4_state *ret;
2069 
2070 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2071 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2072 	else
2073 		ret = _nfs4_opendata_to_nfs4_state(data);
2074 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2075 	return ret;
2076 }
2077 
2078 static struct nfs_open_context *
2079 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2080 {
2081 	struct nfs_inode *nfsi = NFS_I(state->inode);
2082 	struct nfs_open_context *ctx;
2083 
2084 	rcu_read_lock();
2085 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2086 		if (ctx->state != state)
2087 			continue;
2088 		if ((ctx->mode & mode) != mode)
2089 			continue;
2090 		if (!get_nfs_open_context(ctx))
2091 			continue;
2092 		rcu_read_unlock();
2093 		return ctx;
2094 	}
2095 	rcu_read_unlock();
2096 	return ERR_PTR(-ENOENT);
2097 }
2098 
2099 static struct nfs_open_context *
2100 nfs4_state_find_open_context(struct nfs4_state *state)
2101 {
2102 	struct nfs_open_context *ctx;
2103 
2104 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2105 	if (!IS_ERR(ctx))
2106 		return ctx;
2107 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2108 	if (!IS_ERR(ctx))
2109 		return ctx;
2110 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2111 }
2112 
2113 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2114 		struct nfs4_state *state, enum open_claim_type4 claim)
2115 {
2116 	struct nfs4_opendata *opendata;
2117 
2118 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2119 			NULL, claim, GFP_NOFS);
2120 	if (opendata == NULL)
2121 		return ERR_PTR(-ENOMEM);
2122 	opendata->state = state;
2123 	refcount_inc(&state->count);
2124 	return opendata;
2125 }
2126 
2127 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2128 		fmode_t fmode)
2129 {
2130 	struct nfs4_state *newstate;
2131 	int ret;
2132 
2133 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2134 		return 0;
2135 	opendata->o_arg.open_flags = 0;
2136 	opendata->o_arg.fmode = fmode;
2137 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2138 			NFS_SB(opendata->dentry->d_sb),
2139 			fmode, 0);
2140 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2141 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2142 	nfs4_init_opendata_res(opendata);
2143 	ret = _nfs4_recover_proc_open(opendata);
2144 	if (ret != 0)
2145 		return ret;
2146 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2147 	if (IS_ERR(newstate))
2148 		return PTR_ERR(newstate);
2149 	if (newstate != opendata->state)
2150 		ret = -ESTALE;
2151 	nfs4_close_state(newstate, fmode);
2152 	return ret;
2153 }
2154 
2155 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2156 {
2157 	int ret;
2158 
2159 	/* memory barrier prior to reading state->n_* */
2160 	smp_rmb();
2161 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2162 	if (ret != 0)
2163 		return ret;
2164 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2165 	if (ret != 0)
2166 		return ret;
2167 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2168 	if (ret != 0)
2169 		return ret;
2170 	/*
2171 	 * We may have performed cached opens for all three recoveries.
2172 	 * Check if we need to update the current stateid.
2173 	 */
2174 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2175 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2176 		write_seqlock(&state->seqlock);
2177 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2178 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2179 		write_sequnlock(&state->seqlock);
2180 	}
2181 	return 0;
2182 }
2183 
2184 /*
2185  * OPEN_RECLAIM:
2186  * 	reclaim state on the server after a reboot.
2187  */
2188 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2189 {
2190 	struct nfs_delegation *delegation;
2191 	struct nfs4_opendata *opendata;
2192 	fmode_t delegation_type = 0;
2193 	int status;
2194 
2195 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2196 			NFS4_OPEN_CLAIM_PREVIOUS);
2197 	if (IS_ERR(opendata))
2198 		return PTR_ERR(opendata);
2199 	rcu_read_lock();
2200 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2201 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2202 		delegation_type = delegation->type;
2203 	rcu_read_unlock();
2204 	opendata->o_arg.u.delegation_type = delegation_type;
2205 	status = nfs4_open_recover(opendata, state);
2206 	nfs4_opendata_put(opendata);
2207 	return status;
2208 }
2209 
2210 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2211 {
2212 	struct nfs_server *server = NFS_SERVER(state->inode);
2213 	struct nfs4_exception exception = { };
2214 	int err;
2215 	do {
2216 		err = _nfs4_do_open_reclaim(ctx, state);
2217 		trace_nfs4_open_reclaim(ctx, 0, err);
2218 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2219 			continue;
2220 		if (err != -NFS4ERR_DELAY)
2221 			break;
2222 		nfs4_handle_exception(server, err, &exception);
2223 	} while (exception.retry);
2224 	return err;
2225 }
2226 
2227 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2228 {
2229 	struct nfs_open_context *ctx;
2230 	int ret;
2231 
2232 	ctx = nfs4_state_find_open_context(state);
2233 	if (IS_ERR(ctx))
2234 		return -EAGAIN;
2235 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2236 	nfs_state_clear_open_state_flags(state);
2237 	ret = nfs4_do_open_reclaim(ctx, state);
2238 	put_nfs_open_context(ctx);
2239 	return ret;
2240 }
2241 
2242 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)
2243 {
2244 	switch (err) {
2245 		default:
2246 			printk(KERN_ERR "NFS: %s: unhandled error "
2247 					"%d.\n", __func__, err);
2248 			fallthrough;
2249 		case 0:
2250 		case -ENOENT:
2251 		case -EAGAIN:
2252 		case -ESTALE:
2253 		case -ETIMEDOUT:
2254 			break;
2255 		case -NFS4ERR_BADSESSION:
2256 		case -NFS4ERR_BADSLOT:
2257 		case -NFS4ERR_BAD_HIGH_SLOT:
2258 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2259 		case -NFS4ERR_DEADSESSION:
2260 			return -EAGAIN;
2261 		case -NFS4ERR_STALE_CLIENTID:
2262 		case -NFS4ERR_STALE_STATEID:
2263 			/* Don't recall a delegation if it was lost */
2264 			nfs4_schedule_lease_recovery(server->nfs_client);
2265 			return -EAGAIN;
2266 		case -NFS4ERR_MOVED:
2267 			nfs4_schedule_migration_recovery(server);
2268 			return -EAGAIN;
2269 		case -NFS4ERR_LEASE_MOVED:
2270 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2271 			return -EAGAIN;
2272 		case -NFS4ERR_DELEG_REVOKED:
2273 		case -NFS4ERR_ADMIN_REVOKED:
2274 		case -NFS4ERR_EXPIRED:
2275 		case -NFS4ERR_BAD_STATEID:
2276 		case -NFS4ERR_OPENMODE:
2277 			nfs_inode_find_state_and_recover(state->inode,
2278 					stateid);
2279 			nfs4_schedule_stateid_recovery(server, state);
2280 			return -EAGAIN;
2281 		case -NFS4ERR_DELAY:
2282 		case -NFS4ERR_GRACE:
2283 			ssleep(1);
2284 			return -EAGAIN;
2285 		case -ENOMEM:
2286 		case -NFS4ERR_DENIED:
2287 			if (fl) {
2288 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2289 				if (lsp)
2290 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2291 			}
2292 			return 0;
2293 	}
2294 	return err;
2295 }
2296 
2297 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2298 		struct nfs4_state *state, const nfs4_stateid *stateid)
2299 {
2300 	struct nfs_server *server = NFS_SERVER(state->inode);
2301 	struct nfs4_opendata *opendata;
2302 	int err = 0;
2303 
2304 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2305 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2306 	if (IS_ERR(opendata))
2307 		return PTR_ERR(opendata);
2308 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2309 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2310 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2311 		if (err)
2312 			goto out;
2313 	}
2314 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2315 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2316 		if (err)
2317 			goto out;
2318 	}
2319 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2320 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2321 		if (err)
2322 			goto out;
2323 	}
2324 	nfs_state_clear_delegation(state);
2325 out:
2326 	nfs4_opendata_put(opendata);
2327 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2328 }
2329 
2330 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2331 {
2332 	struct nfs4_opendata *data = calldata;
2333 
2334 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2335 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2336 }
2337 
2338 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2339 {
2340 	struct nfs4_opendata *data = calldata;
2341 
2342 	nfs40_sequence_done(task, &data->c_res.seq_res);
2343 
2344 	data->rpc_status = task->tk_status;
2345 	if (data->rpc_status == 0) {
2346 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2347 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2348 		renew_lease(data->o_res.server, data->timestamp);
2349 		data->rpc_done = true;
2350 	}
2351 }
2352 
2353 static void nfs4_open_confirm_release(void *calldata)
2354 {
2355 	struct nfs4_opendata *data = calldata;
2356 	struct nfs4_state *state = NULL;
2357 
2358 	/* If this request hasn't been cancelled, do nothing */
2359 	if (!data->cancelled)
2360 		goto out_free;
2361 	/* In case of error, no cleanup! */
2362 	if (!data->rpc_done)
2363 		goto out_free;
2364 	state = nfs4_opendata_to_nfs4_state(data);
2365 	if (!IS_ERR(state))
2366 		nfs4_close_state(state, data->o_arg.fmode);
2367 out_free:
2368 	nfs4_opendata_put(data);
2369 }
2370 
2371 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2372 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2373 	.rpc_call_done = nfs4_open_confirm_done,
2374 	.rpc_release = nfs4_open_confirm_release,
2375 };
2376 
2377 /*
2378  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2379  */
2380 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2381 {
2382 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2383 	struct rpc_task *task;
2384 	struct  rpc_message msg = {
2385 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2386 		.rpc_argp = &data->c_arg,
2387 		.rpc_resp = &data->c_res,
2388 		.rpc_cred = data->owner->so_cred,
2389 	};
2390 	struct rpc_task_setup task_setup_data = {
2391 		.rpc_client = server->client,
2392 		.rpc_message = &msg,
2393 		.callback_ops = &nfs4_open_confirm_ops,
2394 		.callback_data = data,
2395 		.workqueue = nfsiod_workqueue,
2396 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2397 	};
2398 	int status;
2399 
2400 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2401 				data->is_recover);
2402 	kref_get(&data->kref);
2403 	data->rpc_done = false;
2404 	data->rpc_status = 0;
2405 	data->timestamp = jiffies;
2406 	task = rpc_run_task(&task_setup_data);
2407 	if (IS_ERR(task))
2408 		return PTR_ERR(task);
2409 	status = rpc_wait_for_completion_task(task);
2410 	if (status != 0) {
2411 		data->cancelled = true;
2412 		smp_wmb();
2413 	} else
2414 		status = data->rpc_status;
2415 	rpc_put_task(task);
2416 	return status;
2417 }
2418 
2419 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2420 {
2421 	struct nfs4_opendata *data = calldata;
2422 	struct nfs4_state_owner *sp = data->owner;
2423 	struct nfs_client *clp = sp->so_server->nfs_client;
2424 	enum open_claim_type4 claim = data->o_arg.claim;
2425 
2426 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2427 		goto out_wait;
2428 	/*
2429 	 * Check if we still need to send an OPEN call, or if we can use
2430 	 * a delegation instead.
2431 	 */
2432 	if (data->state != NULL) {
2433 		struct nfs_delegation *delegation;
2434 
2435 		if (can_open_cached(data->state, data->o_arg.fmode,
2436 					data->o_arg.open_flags, claim))
2437 			goto out_no_action;
2438 		rcu_read_lock();
2439 		delegation = nfs4_get_valid_delegation(data->state->inode);
2440 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2441 			goto unlock_no_action;
2442 		rcu_read_unlock();
2443 	}
2444 	/* Update client id. */
2445 	data->o_arg.clientid = clp->cl_clientid;
2446 	switch (claim) {
2447 	default:
2448 		break;
2449 	case NFS4_OPEN_CLAIM_PREVIOUS:
2450 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2451 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2452 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2453 		fallthrough;
2454 	case NFS4_OPEN_CLAIM_FH:
2455 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2456 	}
2457 	data->timestamp = jiffies;
2458 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2459 				&data->o_arg.seq_args,
2460 				&data->o_res.seq_res,
2461 				task) != 0)
2462 		nfs_release_seqid(data->o_arg.seqid);
2463 
2464 	/* Set the create mode (note dependency on the session type) */
2465 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2466 	if (data->o_arg.open_flags & O_EXCL) {
2467 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2468 		if (clp->cl_mvops->minor_version == 0) {
2469 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2470 			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2471 			 * because ACCESS will return permission denied for
2472 			 * all bits until close */
2473 			data->o_res.access_request = data->o_arg.access = 0;
2474 		} else if (nfs4_has_persistent_session(clp))
2475 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2476 	}
2477 	return;
2478 unlock_no_action:
2479 	trace_nfs4_cached_open(data->state);
2480 	rcu_read_unlock();
2481 out_no_action:
2482 	task->tk_action = NULL;
2483 out_wait:
2484 	nfs4_sequence_done(task, &data->o_res.seq_res);
2485 }
2486 
2487 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2488 {
2489 	struct nfs4_opendata *data = calldata;
2490 
2491 	data->rpc_status = task->tk_status;
2492 
2493 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2494 		return;
2495 
2496 	if (task->tk_status == 0) {
2497 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2498 			switch (data->o_res.f_attr->mode & S_IFMT) {
2499 			case S_IFREG:
2500 				break;
2501 			case S_IFLNK:
2502 				data->rpc_status = -ELOOP;
2503 				break;
2504 			case S_IFDIR:
2505 				data->rpc_status = -EISDIR;
2506 				break;
2507 			default:
2508 				data->rpc_status = -ENOTDIR;
2509 			}
2510 		}
2511 		renew_lease(data->o_res.server, data->timestamp);
2512 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2513 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2514 	}
2515 	data->rpc_done = true;
2516 }
2517 
2518 static void nfs4_open_release(void *calldata)
2519 {
2520 	struct nfs4_opendata *data = calldata;
2521 	struct nfs4_state *state = NULL;
2522 
2523 	/* If this request hasn't been cancelled, do nothing */
2524 	if (!data->cancelled)
2525 		goto out_free;
2526 	/* In case of error, no cleanup! */
2527 	if (data->rpc_status != 0 || !data->rpc_done)
2528 		goto out_free;
2529 	/* In case we need an open_confirm, no cleanup! */
2530 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2531 		goto out_free;
2532 	state = nfs4_opendata_to_nfs4_state(data);
2533 	if (!IS_ERR(state))
2534 		nfs4_close_state(state, data->o_arg.fmode);
2535 out_free:
2536 	nfs4_opendata_put(data);
2537 }
2538 
2539 static const struct rpc_call_ops nfs4_open_ops = {
2540 	.rpc_call_prepare = nfs4_open_prepare,
2541 	.rpc_call_done = nfs4_open_done,
2542 	.rpc_release = nfs4_open_release,
2543 };
2544 
2545 static int nfs4_run_open_task(struct nfs4_opendata *data,
2546 			      struct nfs_open_context *ctx)
2547 {
2548 	struct inode *dir = d_inode(data->dir);
2549 	struct nfs_server *server = NFS_SERVER(dir);
2550 	struct nfs_openargs *o_arg = &data->o_arg;
2551 	struct nfs_openres *o_res = &data->o_res;
2552 	struct rpc_task *task;
2553 	struct rpc_message msg = {
2554 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2555 		.rpc_argp = o_arg,
2556 		.rpc_resp = o_res,
2557 		.rpc_cred = data->owner->so_cred,
2558 	};
2559 	struct rpc_task_setup task_setup_data = {
2560 		.rpc_client = server->client,
2561 		.rpc_message = &msg,
2562 		.callback_ops = &nfs4_open_ops,
2563 		.callback_data = data,
2564 		.workqueue = nfsiod_workqueue,
2565 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2566 	};
2567 	int status;
2568 
2569 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2570 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2571 
2572 	kref_get(&data->kref);
2573 	data->rpc_done = false;
2574 	data->rpc_status = 0;
2575 	data->cancelled = false;
2576 	data->is_recover = false;
2577 	if (!ctx) {
2578 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2579 		data->is_recover = true;
2580 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2581 	} else {
2582 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2583 		pnfs_lgopen_prepare(data, ctx);
2584 	}
2585 	task = rpc_run_task(&task_setup_data);
2586 	if (IS_ERR(task))
2587 		return PTR_ERR(task);
2588 	status = rpc_wait_for_completion_task(task);
2589 	if (status != 0) {
2590 		data->cancelled = true;
2591 		smp_wmb();
2592 	} else
2593 		status = data->rpc_status;
2594 	rpc_put_task(task);
2595 
2596 	return status;
2597 }
2598 
2599 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2600 {
2601 	struct inode *dir = d_inode(data->dir);
2602 	struct nfs_openres *o_res = &data->o_res;
2603 	int status;
2604 
2605 	status = nfs4_run_open_task(data, NULL);
2606 	if (status != 0 || !data->rpc_done)
2607 		return status;
2608 
2609 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2610 
2611 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2612 		status = _nfs4_proc_open_confirm(data);
2613 
2614 	return status;
2615 }
2616 
2617 /*
2618  * Additional permission checks in order to distinguish between an
2619  * open for read, and an open for execute. This works around the
2620  * fact that NFSv4 OPEN treats read and execute permissions as being
2621  * the same.
2622  * Note that in the non-execute case, we want to turn off permission
2623  * checking if we just created a new file (POSIX open() semantics).
2624  */
2625 static int nfs4_opendata_access(const struct cred *cred,
2626 				struct nfs4_opendata *opendata,
2627 				struct nfs4_state *state, fmode_t fmode,
2628 				int openflags)
2629 {
2630 	struct nfs_access_entry cache;
2631 	u32 mask, flags;
2632 
2633 	/* access call failed or for some reason the server doesn't
2634 	 * support any access modes -- defer access call until later */
2635 	if (opendata->o_res.access_supported == 0)
2636 		return 0;
2637 
2638 	mask = 0;
2639 	/*
2640 	 * Use openflags to check for exec, because fmode won't
2641 	 * always have FMODE_EXEC set when file open for exec.
2642 	 */
2643 	if (openflags & __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,
2722 			NFS4_OPEN_CLAIM_FH);
2723 	if (IS_ERR(opendata))
2724 		return PTR_ERR(opendata);
2725 	ret = nfs4_open_recover(opendata, state);
2726 	if (ret == -ESTALE)
2727 		d_drop(ctx->dentry);
2728 	nfs4_opendata_put(opendata);
2729 	return ret;
2730 }
2731 
2732 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2733 {
2734 	struct nfs_server *server = NFS_SERVER(state->inode);
2735 	struct nfs4_exception exception = { };
2736 	int err;
2737 
2738 	do {
2739 		err = _nfs4_open_expired(ctx, state);
2740 		trace_nfs4_open_expired(ctx, 0, err);
2741 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2742 			continue;
2743 		switch (err) {
2744 		default:
2745 			goto out;
2746 		case -NFS4ERR_GRACE:
2747 		case -NFS4ERR_DELAY:
2748 			nfs4_handle_exception(server, err, &exception);
2749 			err = 0;
2750 		}
2751 	} while (exception.retry);
2752 out:
2753 	return err;
2754 }
2755 
2756 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2757 {
2758 	struct nfs_open_context *ctx;
2759 	int ret;
2760 
2761 	ctx = nfs4_state_find_open_context(state);
2762 	if (IS_ERR(ctx))
2763 		return -EAGAIN;
2764 	ret = nfs4_do_open_expired(ctx, state);
2765 	put_nfs_open_context(ctx);
2766 	return ret;
2767 }
2768 
2769 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2770 		const nfs4_stateid *stateid)
2771 {
2772 	nfs_remove_bad_delegation(state->inode, stateid);
2773 	nfs_state_clear_delegation(state);
2774 }
2775 
2776 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2777 {
2778 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2779 		nfs_finish_clear_delegation_stateid(state, NULL);
2780 }
2781 
2782 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2783 {
2784 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2785 	nfs40_clear_delegation_stateid(state);
2786 	nfs_state_clear_open_state_flags(state);
2787 	return nfs4_open_expired(sp, state);
2788 }
2789 
2790 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2791 		nfs4_stateid *stateid,
2792 		const struct cred *cred)
2793 {
2794 	return -NFS4ERR_BAD_STATEID;
2795 }
2796 
2797 #if defined(CONFIG_NFS_V4_1)
2798 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2799 		nfs4_stateid *stateid,
2800 		const struct cred *cred)
2801 {
2802 	int status;
2803 
2804 	switch (stateid->type) {
2805 	default:
2806 		break;
2807 	case NFS4_INVALID_STATEID_TYPE:
2808 	case NFS4_SPECIAL_STATEID_TYPE:
2809 		return -NFS4ERR_BAD_STATEID;
2810 	case NFS4_REVOKED_STATEID_TYPE:
2811 		goto out_free;
2812 	}
2813 
2814 	status = nfs41_test_stateid(server, stateid, cred);
2815 	switch (status) {
2816 	case -NFS4ERR_EXPIRED:
2817 	case -NFS4ERR_ADMIN_REVOKED:
2818 	case -NFS4ERR_DELEG_REVOKED:
2819 		break;
2820 	default:
2821 		return status;
2822 	}
2823 out_free:
2824 	/* Ack the revoked state to the server */
2825 	nfs41_free_stateid(server, stateid, cred, true);
2826 	return -NFS4ERR_EXPIRED;
2827 }
2828 
2829 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2830 {
2831 	struct nfs_server *server = NFS_SERVER(state->inode);
2832 	nfs4_stateid stateid;
2833 	struct nfs_delegation *delegation;
2834 	const struct cred *cred = NULL;
2835 	int status, ret = NFS_OK;
2836 
2837 	/* Get the delegation credential for use by test/free_stateid */
2838 	rcu_read_lock();
2839 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2840 	if (delegation == NULL) {
2841 		rcu_read_unlock();
2842 		nfs_state_clear_delegation(state);
2843 		return NFS_OK;
2844 	}
2845 
2846 	spin_lock(&delegation->lock);
2847 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2848 
2849 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2850 				&delegation->flags)) {
2851 		spin_unlock(&delegation->lock);
2852 		rcu_read_unlock();
2853 		return NFS_OK;
2854 	}
2855 
2856 	if (delegation->cred)
2857 		cred = get_cred(delegation->cred);
2858 	spin_unlock(&delegation->lock);
2859 	rcu_read_unlock();
2860 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2861 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2862 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2863 		nfs_finish_clear_delegation_stateid(state, &stateid);
2864 	else
2865 		ret = status;
2866 
2867 	put_cred(cred);
2868 	return ret;
2869 }
2870 
2871 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2872 {
2873 	nfs4_stateid tmp;
2874 
2875 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2876 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2877 				&tmp, NULL) &&
2878 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2879 		nfs_state_set_delegation(state, &tmp, state->state);
2880 	else
2881 		nfs_state_clear_delegation(state);
2882 }
2883 
2884 /**
2885  * nfs41_check_expired_locks - possibly free a lock stateid
2886  *
2887  * @state: NFSv4 state for an inode
2888  *
2889  * Returns NFS_OK if recovery for this stateid is now finished.
2890  * Otherwise a negative NFS4ERR value is returned.
2891  */
2892 static int nfs41_check_expired_locks(struct nfs4_state *state)
2893 {
2894 	int status, ret = NFS_OK;
2895 	struct nfs4_lock_state *lsp, *prev = NULL;
2896 	struct nfs_server *server = NFS_SERVER(state->inode);
2897 
2898 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2899 		goto out;
2900 
2901 	spin_lock(&state->state_lock);
2902 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2903 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2904 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2905 
2906 			refcount_inc(&lsp->ls_count);
2907 			spin_unlock(&state->state_lock);
2908 
2909 			nfs4_put_lock_state(prev);
2910 			prev = lsp;
2911 
2912 			status = nfs41_test_and_free_expired_stateid(server,
2913 					&lsp->ls_stateid,
2914 					cred);
2915 			trace_nfs4_test_lock_stateid(state, lsp, status);
2916 			if (status == -NFS4ERR_EXPIRED ||
2917 			    status == -NFS4ERR_BAD_STATEID) {
2918 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2919 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2920 				if (!recover_lost_locks)
2921 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2922 			} else if (status != NFS_OK) {
2923 				ret = status;
2924 				nfs4_put_lock_state(prev);
2925 				goto out;
2926 			}
2927 			spin_lock(&state->state_lock);
2928 		}
2929 	}
2930 	spin_unlock(&state->state_lock);
2931 	nfs4_put_lock_state(prev);
2932 out:
2933 	return ret;
2934 }
2935 
2936 /**
2937  * nfs41_check_open_stateid - possibly free an open stateid
2938  *
2939  * @state: NFSv4 state for an inode
2940  *
2941  * Returns NFS_OK if recovery for this stateid is now finished.
2942  * Otherwise a negative NFS4ERR value is returned.
2943  */
2944 static int nfs41_check_open_stateid(struct nfs4_state *state)
2945 {
2946 	struct nfs_server *server = NFS_SERVER(state->inode);
2947 	nfs4_stateid *stateid = &state->open_stateid;
2948 	const struct cred *cred = state->owner->so_cred;
2949 	int status;
2950 
2951 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2952 		return -NFS4ERR_BAD_STATEID;
2953 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2954 	trace_nfs4_test_open_stateid(state, NULL, status);
2955 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2956 		nfs_state_clear_open_state_flags(state);
2957 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2958 		return status;
2959 	}
2960 	if (nfs_open_stateid_recover_openmode(state))
2961 		return -NFS4ERR_OPENMODE;
2962 	return NFS_OK;
2963 }
2964 
2965 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2966 {
2967 	int status;
2968 
2969 	status = nfs41_check_delegation_stateid(state);
2970 	if (status != NFS_OK)
2971 		return status;
2972 	nfs41_delegation_recover_stateid(state);
2973 
2974 	status = nfs41_check_expired_locks(state);
2975 	if (status != NFS_OK)
2976 		return status;
2977 	status = nfs41_check_open_stateid(state);
2978 	if (status != NFS_OK)
2979 		status = nfs4_open_expired(sp, state);
2980 	return status;
2981 }
2982 #endif
2983 
2984 /*
2985  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2986  * fields corresponding to attributes that were used to store the verifier.
2987  * Make sure we clobber those fields in the later setattr call
2988  */
2989 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2990 				struct iattr *sattr, struct nfs4_label **label)
2991 {
2992 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2993 	__u32 attrset[3];
2994 	unsigned ret;
2995 	unsigned i;
2996 
2997 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2998 		attrset[i] = opendata->o_res.attrset[i];
2999 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3000 			attrset[i] &= ~bitmask[i];
3001 	}
3002 
3003 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3004 		sattr->ia_valid : 0;
3005 
3006 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3007 		if (sattr->ia_valid & ATTR_ATIME_SET)
3008 			ret |= ATTR_ATIME_SET;
3009 		else
3010 			ret |= ATTR_ATIME;
3011 	}
3012 
3013 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3014 		if (sattr->ia_valid & ATTR_MTIME_SET)
3015 			ret |= ATTR_MTIME_SET;
3016 		else
3017 			ret |= ATTR_MTIME;
3018 	}
3019 
3020 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3021 		*label = NULL;
3022 	return ret;
3023 }
3024 
3025 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3026 		int flags, struct nfs_open_context *ctx)
3027 {
3028 	struct nfs4_state_owner *sp = opendata->owner;
3029 	struct nfs_server *server = sp->so_server;
3030 	struct dentry *dentry;
3031 	struct nfs4_state *state;
3032 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3033 	struct inode *dir = d_inode(opendata->dir);
3034 	unsigned long dir_verifier;
3035 	unsigned int seq;
3036 	int ret;
3037 
3038 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3039 	dir_verifier = nfs_save_change_attribute(dir);
3040 
3041 	ret = _nfs4_proc_open(opendata, ctx);
3042 	if (ret != 0)
3043 		goto out;
3044 
3045 	state = _nfs4_opendata_to_nfs4_state(opendata);
3046 	ret = PTR_ERR(state);
3047 	if (IS_ERR(state))
3048 		goto out;
3049 	ctx->state = state;
3050 	if (server->caps & NFS_CAP_POSIX_LOCK)
3051 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3052 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3053 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3054 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3055 		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3056 
3057 	dentry = opendata->dentry;
3058 	if (d_really_is_negative(dentry)) {
3059 		struct dentry *alias;
3060 		d_drop(dentry);
3061 		alias = d_exact_alias(dentry, state->inode);
3062 		if (!alias)
3063 			alias = d_splice_alias(igrab(state->inode), dentry);
3064 		/* d_splice_alias() can't fail here - it's a non-directory */
3065 		if (alias) {
3066 			dput(ctx->dentry);
3067 			ctx->dentry = dentry = alias;
3068 		}
3069 	}
3070 
3071 	switch(opendata->o_arg.claim) {
3072 	default:
3073 		break;
3074 	case NFS4_OPEN_CLAIM_NULL:
3075 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3076 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3077 		if (!opendata->rpc_done)
3078 			break;
3079 		if (opendata->o_res.delegation_type != 0)
3080 			dir_verifier = nfs_save_change_attribute(dir);
3081 		nfs_set_verifier(dentry, dir_verifier);
3082 	}
3083 
3084 	/* Parse layoutget results before we check for access */
3085 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3086 
3087 	ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3088 			acc_mode, flags);
3089 	if (ret != 0)
3090 		goto out;
3091 
3092 	if (d_inode(dentry) == state->inode) {
3093 		nfs_inode_attach_open_context(ctx);
3094 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3095 			nfs4_schedule_stateid_recovery(server, state);
3096 	}
3097 
3098 out:
3099 	if (opendata->lgp) {
3100 		nfs4_lgopen_release(opendata->lgp);
3101 		opendata->lgp = NULL;
3102 	}
3103 	if (!opendata->cancelled)
3104 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3105 	return ret;
3106 }
3107 
3108 /*
3109  * Returns a referenced nfs4_state
3110  */
3111 static int _nfs4_do_open(struct inode *dir,
3112 			struct nfs_open_context *ctx,
3113 			int flags,
3114 			const struct nfs4_open_createattrs *c,
3115 			int *opened)
3116 {
3117 	struct nfs4_state_owner  *sp;
3118 	struct nfs4_state     *state = NULL;
3119 	struct nfs_server       *server = NFS_SERVER(dir);
3120 	struct nfs4_opendata *opendata;
3121 	struct dentry *dentry = ctx->dentry;
3122 	const struct cred *cred = ctx->cred;
3123 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3124 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3125 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3126 	struct iattr *sattr = c->sattr;
3127 	struct nfs4_label *label = c->label;
3128 	int status;
3129 
3130 	/* Protect against reboot recovery conflicts */
3131 	status = -ENOMEM;
3132 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3133 	if (sp == NULL) {
3134 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3135 		goto out_err;
3136 	}
3137 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3138 	if (status != 0)
3139 		goto err_put_state_owner;
3140 	if (d_really_is_positive(dentry))
3141 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3142 	status = -ENOMEM;
3143 	if (d_really_is_positive(dentry))
3144 		claim = NFS4_OPEN_CLAIM_FH;
3145 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3146 			c, claim, GFP_KERNEL);
3147 	if (opendata == NULL)
3148 		goto err_put_state_owner;
3149 
3150 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3151 		if (!opendata->f_attr.mdsthreshold) {
3152 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3153 			if (!opendata->f_attr.mdsthreshold)
3154 				goto err_opendata_put;
3155 		}
3156 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3157 	}
3158 	if (d_really_is_positive(dentry))
3159 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3160 
3161 	status = _nfs4_open_and_get_state(opendata, flags, ctx);
3162 	if (status != 0)
3163 		goto err_opendata_put;
3164 	state = ctx->state;
3165 
3166 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3167 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3168 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3169 		/*
3170 		 * send create attributes which was not set by open
3171 		 * with an extra setattr.
3172 		 */
3173 		if (attrs || label) {
3174 			unsigned ia_old = sattr->ia_valid;
3175 
3176 			sattr->ia_valid = attrs;
3177 			nfs_fattr_init(opendata->o_res.f_attr);
3178 			status = nfs4_do_setattr(state->inode, cred,
3179 					opendata->o_res.f_attr, sattr,
3180 					ctx, label);
3181 			if (status == 0) {
3182 				nfs_setattr_update_inode(state->inode, sattr,
3183 						opendata->o_res.f_attr);
3184 				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3185 			}
3186 			sattr->ia_valid = ia_old;
3187 		}
3188 	}
3189 	if (opened && opendata->file_created)
3190 		*opened = 1;
3191 
3192 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3193 		*ctx_th = opendata->f_attr.mdsthreshold;
3194 		opendata->f_attr.mdsthreshold = NULL;
3195 	}
3196 
3197 	nfs4_opendata_put(opendata);
3198 	nfs4_put_state_owner(sp);
3199 	return 0;
3200 err_opendata_put:
3201 	nfs4_opendata_put(opendata);
3202 err_put_state_owner:
3203 	nfs4_put_state_owner(sp);
3204 out_err:
3205 	return status;
3206 }
3207 
3208 
3209 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3210 					struct nfs_open_context *ctx,
3211 					int flags,
3212 					struct iattr *sattr,
3213 					struct nfs4_label *label,
3214 					int *opened)
3215 {
3216 	struct nfs_server *server = NFS_SERVER(dir);
3217 	struct nfs4_exception exception = {
3218 		.interruptible = true,
3219 	};
3220 	struct nfs4_state *res;
3221 	struct nfs4_open_createattrs c = {
3222 		.label = label,
3223 		.sattr = sattr,
3224 		.verf = {
3225 			[0] = (__u32)jiffies,
3226 			[1] = (__u32)current->pid,
3227 		},
3228 	};
3229 	int status;
3230 
3231 	do {
3232 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3233 		res = ctx->state;
3234 		trace_nfs4_open_file(ctx, flags, status);
3235 		if (status == 0)
3236 			break;
3237 		/* NOTE: BAD_SEQID means the server and client disagree about the
3238 		 * book-keeping w.r.t. state-changing operations
3239 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3240 		 * It is actually a sign of a bug on the client or on the server.
3241 		 *
3242 		 * If we receive a BAD_SEQID error in the particular case of
3243 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3244 		 * have unhashed the old state_owner for us, and that we can
3245 		 * therefore safely retry using a new one. We should still warn
3246 		 * the user though...
3247 		 */
3248 		if (status == -NFS4ERR_BAD_SEQID) {
3249 			pr_warn_ratelimited("NFS: v4 server %s "
3250 					" returned a bad sequence-id error!\n",
3251 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3252 			exception.retry = 1;
3253 			continue;
3254 		}
3255 		/*
3256 		 * BAD_STATEID on OPEN means that the server cancelled our
3257 		 * state before it received the OPEN_CONFIRM.
3258 		 * Recover by retrying the request as per the discussion
3259 		 * on Page 181 of RFC3530.
3260 		 */
3261 		if (status == -NFS4ERR_BAD_STATEID) {
3262 			exception.retry = 1;
3263 			continue;
3264 		}
3265 		if (status == -NFS4ERR_EXPIRED) {
3266 			nfs4_schedule_lease_recovery(server->nfs_client);
3267 			exception.retry = 1;
3268 			continue;
3269 		}
3270 		if (status == -EAGAIN) {
3271 			/* We must have found a delegation */
3272 			exception.retry = 1;
3273 			continue;
3274 		}
3275 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3276 			continue;
3277 		res = ERR_PTR(nfs4_handle_exception(server,
3278 					status, &exception));
3279 	} while (exception.retry);
3280 	return res;
3281 }
3282 
3283 static int _nfs4_do_setattr(struct inode *inode,
3284 			    struct nfs_setattrargs *arg,
3285 			    struct nfs_setattrres *res,
3286 			    const struct cred *cred,
3287 			    struct nfs_open_context *ctx)
3288 {
3289 	struct nfs_server *server = NFS_SERVER(inode);
3290 	struct rpc_message msg = {
3291 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3292 		.rpc_argp	= arg,
3293 		.rpc_resp	= res,
3294 		.rpc_cred	= cred,
3295 	};
3296 	const struct cred *delegation_cred = NULL;
3297 	unsigned long timestamp = jiffies;
3298 	bool truncate;
3299 	int status;
3300 
3301 	nfs_fattr_init(res->fattr);
3302 
3303 	/* Servers should only apply open mode checks for file size changes */
3304 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3305 	if (!truncate) {
3306 		nfs4_inode_make_writeable(inode);
3307 		goto zero_stateid;
3308 	}
3309 
3310 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3311 		/* Use that stateid */
3312 	} else if (ctx != NULL && ctx->state) {
3313 		struct nfs_lock_context *l_ctx;
3314 		if (!nfs4_valid_open_stateid(ctx->state))
3315 			return -EBADF;
3316 		l_ctx = nfs_get_lock_context(ctx);
3317 		if (IS_ERR(l_ctx))
3318 			return PTR_ERR(l_ctx);
3319 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3320 						&arg->stateid, &delegation_cred);
3321 		nfs_put_lock_context(l_ctx);
3322 		if (status == -EIO)
3323 			return -EBADF;
3324 		else if (status == -EAGAIN)
3325 			goto zero_stateid;
3326 	} else {
3327 zero_stateid:
3328 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3329 	}
3330 	if (delegation_cred)
3331 		msg.rpc_cred = delegation_cred;
3332 
3333 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3334 
3335 	put_cred(delegation_cred);
3336 	if (status == 0 && ctx != NULL)
3337 		renew_lease(server, timestamp);
3338 	trace_nfs4_setattr(inode, &arg->stateid, status);
3339 	return status;
3340 }
3341 
3342 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3343 			   struct nfs_fattr *fattr, struct iattr *sattr,
3344 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3345 {
3346 	struct nfs_server *server = NFS_SERVER(inode);
3347 	__u32 bitmask[NFS4_BITMASK_SZ];
3348 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3349 	struct nfs_setattrargs	arg = {
3350 		.fh		= NFS_FH(inode),
3351 		.iap		= sattr,
3352 		.server		= server,
3353 		.bitmask = bitmask,
3354 		.label		= ilabel,
3355 	};
3356 	struct nfs_setattrres  res = {
3357 		.fattr		= fattr,
3358 		.server		= server,
3359 	};
3360 	struct nfs4_exception exception = {
3361 		.state = state,
3362 		.inode = inode,
3363 		.stateid = &arg.stateid,
3364 	};
3365 	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3366 	int err;
3367 
3368 	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3369 		adjust_flags |= NFS_INO_INVALID_MODE;
3370 	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3371 		adjust_flags |= NFS_INO_INVALID_OTHER;
3372 
3373 	do {
3374 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3375 					inode, adjust_flags);
3376 
3377 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3378 		switch (err) {
3379 		case -NFS4ERR_OPENMODE:
3380 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3381 				pr_warn_once("NFSv4: server %s is incorrectly "
3382 						"applying open mode checks to "
3383 						"a SETATTR that is not "
3384 						"changing file size.\n",
3385 						server->nfs_client->cl_hostname);
3386 			}
3387 			if (state && !(state->state & FMODE_WRITE)) {
3388 				err = -EBADF;
3389 				if (sattr->ia_valid & ATTR_OPEN)
3390 					err = -EACCES;
3391 				goto out;
3392 			}
3393 		}
3394 		err = nfs4_handle_exception(server, err, &exception);
3395 	} while (exception.retry);
3396 out:
3397 	return err;
3398 }
3399 
3400 static bool
3401 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3402 {
3403 	if (inode == NULL || !nfs_have_layout(inode))
3404 		return false;
3405 
3406 	return pnfs_wait_on_layoutreturn(inode, task);
3407 }
3408 
3409 /*
3410  * Update the seqid of an open stateid
3411  */
3412 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3413 		struct nfs4_state *state)
3414 {
3415 	__be32 seqid_open;
3416 	u32 dst_seqid;
3417 	int seq;
3418 
3419 	for (;;) {
3420 		if (!nfs4_valid_open_stateid(state))
3421 			break;
3422 		seq = read_seqbegin(&state->seqlock);
3423 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3424 			nfs4_stateid_copy(dst, &state->open_stateid);
3425 			if (read_seqretry(&state->seqlock, seq))
3426 				continue;
3427 			break;
3428 		}
3429 		seqid_open = state->open_stateid.seqid;
3430 		if (read_seqretry(&state->seqlock, seq))
3431 			continue;
3432 
3433 		dst_seqid = be32_to_cpu(dst->seqid);
3434 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3435 			dst->seqid = seqid_open;
3436 		break;
3437 	}
3438 }
3439 
3440 /*
3441  * Update the seqid of an open stateid after receiving
3442  * NFS4ERR_OLD_STATEID
3443  */
3444 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3445 		struct nfs4_state *state)
3446 {
3447 	__be32 seqid_open;
3448 	u32 dst_seqid;
3449 	bool ret;
3450 	int seq, status = -EAGAIN;
3451 	DEFINE_WAIT(wait);
3452 
3453 	for (;;) {
3454 		ret = false;
3455 		if (!nfs4_valid_open_stateid(state))
3456 			break;
3457 		seq = read_seqbegin(&state->seqlock);
3458 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3459 			if (read_seqretry(&state->seqlock, seq))
3460 				continue;
3461 			break;
3462 		}
3463 
3464 		write_seqlock(&state->seqlock);
3465 		seqid_open = state->open_stateid.seqid;
3466 
3467 		dst_seqid = be32_to_cpu(dst->seqid);
3468 
3469 		/* Did another OPEN bump the state's seqid?  try again: */
3470 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3471 			dst->seqid = seqid_open;
3472 			write_sequnlock(&state->seqlock);
3473 			ret = true;
3474 			break;
3475 		}
3476 
3477 		/* server says we're behind but we haven't seen the update yet */
3478 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3479 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3480 		write_sequnlock(&state->seqlock);
3481 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3482 
3483 		if (fatal_signal_pending(current))
3484 			status = -EINTR;
3485 		else
3486 			if (schedule_timeout(5*HZ) != 0)
3487 				status = 0;
3488 
3489 		finish_wait(&state->waitq, &wait);
3490 
3491 		if (!status)
3492 			continue;
3493 		if (status == -EINTR)
3494 			break;
3495 
3496 		/* we slept the whole 5 seconds, we must have lost a seqid */
3497 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3498 		ret = true;
3499 		break;
3500 	}
3501 
3502 	return ret;
3503 }
3504 
3505 struct nfs4_closedata {
3506 	struct inode *inode;
3507 	struct nfs4_state *state;
3508 	struct nfs_closeargs arg;
3509 	struct nfs_closeres res;
3510 	struct {
3511 		struct nfs4_layoutreturn_args arg;
3512 		struct nfs4_layoutreturn_res res;
3513 		struct nfs4_xdr_opaque_data ld_private;
3514 		u32 roc_barrier;
3515 		bool roc;
3516 	} lr;
3517 	struct nfs_fattr fattr;
3518 	unsigned long timestamp;
3519 };
3520 
3521 static void nfs4_free_closedata(void *data)
3522 {
3523 	struct nfs4_closedata *calldata = data;
3524 	struct nfs4_state_owner *sp = calldata->state->owner;
3525 	struct super_block *sb = calldata->state->inode->i_sb;
3526 
3527 	if (calldata->lr.roc)
3528 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3529 				calldata->res.lr_ret);
3530 	nfs4_put_open_state(calldata->state);
3531 	nfs_free_seqid(calldata->arg.seqid);
3532 	nfs4_put_state_owner(sp);
3533 	nfs_sb_deactive(sb);
3534 	kfree(calldata);
3535 }
3536 
3537 static void nfs4_close_done(struct rpc_task *task, void *data)
3538 {
3539 	struct nfs4_closedata *calldata = data;
3540 	struct nfs4_state *state = calldata->state;
3541 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3542 	nfs4_stateid *res_stateid = NULL;
3543 	struct nfs4_exception exception = {
3544 		.state = state,
3545 		.inode = calldata->inode,
3546 		.stateid = &calldata->arg.stateid,
3547 	};
3548 
3549 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3550 		return;
3551 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3552 
3553 	/* Handle Layoutreturn errors */
3554 	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3555 			  &calldata->res.lr_ret) == -EAGAIN)
3556 		goto out_restart;
3557 
3558 	/* hmm. we are done with the inode, and in the process of freeing
3559 	 * the state_owner. we keep this around to process errors
3560 	 */
3561 	switch (task->tk_status) {
3562 		case 0:
3563 			res_stateid = &calldata->res.stateid;
3564 			renew_lease(server, calldata->timestamp);
3565 			break;
3566 		case -NFS4ERR_ACCESS:
3567 			if (calldata->arg.bitmask != NULL) {
3568 				calldata->arg.bitmask = NULL;
3569 				calldata->res.fattr = NULL;
3570 				goto out_restart;
3571 
3572 			}
3573 			break;
3574 		case -NFS4ERR_OLD_STATEID:
3575 			/* Did we race with OPEN? */
3576 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3577 						state))
3578 				goto out_restart;
3579 			goto out_release;
3580 		case -NFS4ERR_ADMIN_REVOKED:
3581 		case -NFS4ERR_STALE_STATEID:
3582 		case -NFS4ERR_EXPIRED:
3583 			nfs4_free_revoked_stateid(server,
3584 					&calldata->arg.stateid,
3585 					task->tk_msg.rpc_cred);
3586 			fallthrough;
3587 		case -NFS4ERR_BAD_STATEID:
3588 			if (calldata->arg.fmode == 0)
3589 				break;
3590 			fallthrough;
3591 		default:
3592 			task->tk_status = nfs4_async_handle_exception(task,
3593 					server, task->tk_status, &exception);
3594 			if (exception.retry)
3595 				goto out_restart;
3596 	}
3597 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3598 			res_stateid, calldata->arg.fmode);
3599 out_release:
3600 	task->tk_status = 0;
3601 	nfs_release_seqid(calldata->arg.seqid);
3602 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3603 	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3604 	return;
3605 out_restart:
3606 	task->tk_status = 0;
3607 	rpc_restart_call_prepare(task);
3608 	goto out_release;
3609 }
3610 
3611 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3612 {
3613 	struct nfs4_closedata *calldata = data;
3614 	struct nfs4_state *state = calldata->state;
3615 	struct inode *inode = calldata->inode;
3616 	struct nfs_server *server = NFS_SERVER(inode);
3617 	struct pnfs_layout_hdr *lo;
3618 	bool is_rdonly, is_wronly, is_rdwr;
3619 	int call_close = 0;
3620 
3621 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3622 		goto out_wait;
3623 
3624 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3625 	spin_lock(&state->owner->so_lock);
3626 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3627 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3628 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3629 	/* Calculate the change in open mode */
3630 	calldata->arg.fmode = 0;
3631 	if (state->n_rdwr == 0) {
3632 		if (state->n_rdonly == 0)
3633 			call_close |= is_rdonly;
3634 		else if (is_rdonly)
3635 			calldata->arg.fmode |= FMODE_READ;
3636 		if (state->n_wronly == 0)
3637 			call_close |= is_wronly;
3638 		else if (is_wronly)
3639 			calldata->arg.fmode |= FMODE_WRITE;
3640 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3641 			call_close |= is_rdwr;
3642 	} else if (is_rdwr)
3643 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3644 
3645 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3646 	if (!nfs4_valid_open_stateid(state))
3647 		call_close = 0;
3648 	spin_unlock(&state->owner->so_lock);
3649 
3650 	if (!call_close) {
3651 		/* Note: exit _without_ calling nfs4_close_done */
3652 		goto out_no_action;
3653 	}
3654 
3655 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3656 		nfs_release_seqid(calldata->arg.seqid);
3657 		goto out_wait;
3658 	}
3659 
3660 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3661 	if (lo && !pnfs_layout_is_valid(lo)) {
3662 		calldata->arg.lr_args = NULL;
3663 		calldata->res.lr_res = NULL;
3664 	}
3665 
3666 	if (calldata->arg.fmode == 0)
3667 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3668 
3669 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3670 		/* Close-to-open cache consistency revalidation */
3671 		if (!nfs4_have_delegation(inode, FMODE_READ)) {
3672 			nfs4_bitmask_set(calldata->arg.bitmask_store,
3673 					 server->cache_consistency_bitmask,
3674 					 inode, 0);
3675 			calldata->arg.bitmask = calldata->arg.bitmask_store;
3676 		} else
3677 			calldata->arg.bitmask = NULL;
3678 	}
3679 
3680 	calldata->arg.share_access =
3681 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3682 				calldata->arg.fmode, 0);
3683 
3684 	if (calldata->res.fattr == NULL)
3685 		calldata->arg.bitmask = NULL;
3686 	else if (calldata->arg.bitmask == NULL)
3687 		calldata->res.fattr = NULL;
3688 	calldata->timestamp = jiffies;
3689 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3690 				&calldata->arg.seq_args,
3691 				&calldata->res.seq_res,
3692 				task) != 0)
3693 		nfs_release_seqid(calldata->arg.seqid);
3694 	return;
3695 out_no_action:
3696 	task->tk_action = NULL;
3697 out_wait:
3698 	nfs4_sequence_done(task, &calldata->res.seq_res);
3699 }
3700 
3701 static const struct rpc_call_ops nfs4_close_ops = {
3702 	.rpc_call_prepare = nfs4_close_prepare,
3703 	.rpc_call_done = nfs4_close_done,
3704 	.rpc_release = nfs4_free_closedata,
3705 };
3706 
3707 /*
3708  * It is possible for data to be read/written from a mem-mapped file
3709  * after the sys_close call (which hits the vfs layer as a flush).
3710  * This means that we can't safely call nfsv4 close on a file until
3711  * the inode is cleared. This in turn means that we are not good
3712  * NFSv4 citizens - we do not indicate to the server to update the file's
3713  * share state even when we are done with one of the three share
3714  * stateid's in the inode.
3715  *
3716  * NOTE: Caller must be holding the sp->so_owner semaphore!
3717  */
3718 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3719 {
3720 	struct nfs_server *server = NFS_SERVER(state->inode);
3721 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3722 	struct nfs4_closedata *calldata;
3723 	struct nfs4_state_owner *sp = state->owner;
3724 	struct rpc_task *task;
3725 	struct rpc_message msg = {
3726 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3727 		.rpc_cred = state->owner->so_cred,
3728 	};
3729 	struct rpc_task_setup task_setup_data = {
3730 		.rpc_client = server->client,
3731 		.rpc_message = &msg,
3732 		.callback_ops = &nfs4_close_ops,
3733 		.workqueue = nfsiod_workqueue,
3734 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3735 	};
3736 	int status = -ENOMEM;
3737 
3738 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3739 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3740 
3741 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3742 		&task_setup_data.rpc_client, &msg);
3743 
3744 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3745 	if (calldata == NULL)
3746 		goto out;
3747 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3748 	calldata->inode = state->inode;
3749 	calldata->state = state;
3750 	calldata->arg.fh = NFS_FH(state->inode);
3751 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3752 		goto out_free_calldata;
3753 	/* Serialization for the sequence id */
3754 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3755 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3756 	if (IS_ERR(calldata->arg.seqid))
3757 		goto out_free_calldata;
3758 	nfs_fattr_init(&calldata->fattr);
3759 	calldata->arg.fmode = 0;
3760 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3761 	calldata->res.fattr = &calldata->fattr;
3762 	calldata->res.seqid = calldata->arg.seqid;
3763 	calldata->res.server = server;
3764 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3765 	calldata->lr.roc = pnfs_roc(state->inode,
3766 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3767 	if (calldata->lr.roc) {
3768 		calldata->arg.lr_args = &calldata->lr.arg;
3769 		calldata->res.lr_res = &calldata->lr.res;
3770 	}
3771 	nfs_sb_active(calldata->inode->i_sb);
3772 
3773 	msg.rpc_argp = &calldata->arg;
3774 	msg.rpc_resp = &calldata->res;
3775 	task_setup_data.callback_data = calldata;
3776 	task = rpc_run_task(&task_setup_data);
3777 	if (IS_ERR(task))
3778 		return PTR_ERR(task);
3779 	status = 0;
3780 	if (wait)
3781 		status = rpc_wait_for_completion_task(task);
3782 	rpc_put_task(task);
3783 	return status;
3784 out_free_calldata:
3785 	kfree(calldata);
3786 out:
3787 	nfs4_put_open_state(state);
3788 	nfs4_put_state_owner(sp);
3789 	return status;
3790 }
3791 
3792 static struct inode *
3793 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3794 		int open_flags, struct iattr *attr, int *opened)
3795 {
3796 	struct nfs4_state *state;
3797 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3798 
3799 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3800 
3801 	/* Protect against concurrent sillydeletes */
3802 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3803 
3804 	nfs4_label_release_security(label);
3805 
3806 	if (IS_ERR(state))
3807 		return ERR_CAST(state);
3808 	return state->inode;
3809 }
3810 
3811 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3812 {
3813 	if (ctx->state == NULL)
3814 		return;
3815 	if (is_sync)
3816 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3817 	else
3818 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3819 }
3820 
3821 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3822 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3823 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3824 
3825 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3826 {
3827 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3828 	struct nfs4_server_caps_arg args = {
3829 		.fhandle = fhandle,
3830 		.bitmask = bitmask,
3831 	};
3832 	struct nfs4_server_caps_res res = {};
3833 	struct rpc_message msg = {
3834 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3835 		.rpc_argp = &args,
3836 		.rpc_resp = &res,
3837 	};
3838 	int status;
3839 	int i;
3840 
3841 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3842 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3843 		     FATTR4_WORD0_LINK_SUPPORT |
3844 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3845 		     FATTR4_WORD0_ACLSUPPORT |
3846 		     FATTR4_WORD0_CASE_INSENSITIVE |
3847 		     FATTR4_WORD0_CASE_PRESERVING;
3848 	if (minorversion)
3849 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3850 
3851 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3852 	if (status == 0) {
3853 		/* Sanity check the server answers */
3854 		switch (minorversion) {
3855 		case 0:
3856 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3857 			res.attr_bitmask[2] = 0;
3858 			break;
3859 		case 1:
3860 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3861 			break;
3862 		case 2:
3863 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3864 		}
3865 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3866 		server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3867 				  NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3868 		server->fattr_valid = NFS_ATTR_FATTR_V4;
3869 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3870 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3871 			server->caps |= NFS_CAP_ACLS;
3872 		if (res.has_links != 0)
3873 			server->caps |= NFS_CAP_HARDLINKS;
3874 		if (res.has_symlinks != 0)
3875 			server->caps |= NFS_CAP_SYMLINKS;
3876 		if (res.case_insensitive)
3877 			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3878 		if (res.case_preserving)
3879 			server->caps |= NFS_CAP_CASE_PRESERVING;
3880 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3881 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3882 			server->caps |= NFS_CAP_SECURITY_LABEL;
3883 #endif
3884 		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3885 			server->caps |= NFS_CAP_FS_LOCATIONS;
3886 		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3887 			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3888 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3889 			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3890 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3891 			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3892 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3893 			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3894 				NFS_ATTR_FATTR_OWNER_NAME);
3895 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3896 			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3897 				NFS_ATTR_FATTR_GROUP_NAME);
3898 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3899 			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3900 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3901 			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3902 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3903 			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3904 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3905 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3906 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3907 				sizeof(server->attr_bitmask));
3908 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3909 
3910 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3911 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3912 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3913 		server->cache_consistency_bitmask[2] = 0;
3914 
3915 		/* Avoid a regression due to buggy server */
3916 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3917 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3918 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3919 			sizeof(server->exclcreat_bitmask));
3920 
3921 		server->acl_bitmask = res.acl_bitmask;
3922 		server->fh_expire_type = res.fh_expire_type;
3923 	}
3924 
3925 	return status;
3926 }
3927 
3928 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3929 {
3930 	struct nfs4_exception exception = {
3931 		.interruptible = true,
3932 	};
3933 	int err;
3934 
3935 	nfs4_server_set_init_caps(server);
3936 	do {
3937 		err = nfs4_handle_exception(server,
3938 				_nfs4_server_capabilities(server, fhandle),
3939 				&exception);
3940 	} while (exception.retry);
3941 	return err;
3942 }
3943 
3944 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3945 					  struct nfs_client *clp,
3946 					  struct nfs_server *server)
3947 {
3948 	int i;
3949 
3950 	for (i = 0; i < location->nservers; i++) {
3951 		struct nfs4_string *srv_loc = &location->servers[i];
3952 		struct sockaddr addr;
3953 		size_t addrlen;
3954 		struct xprt_create xprt_args = {
3955 			.ident = 0,
3956 			.net = clp->cl_net,
3957 		};
3958 		struct nfs4_add_xprt_data xprtdata = {
3959 			.clp = clp,
3960 		};
3961 		struct rpc_add_xprt_test rpcdata = {
3962 			.add_xprt_test = clp->cl_mvops->session_trunk,
3963 			.data = &xprtdata,
3964 		};
3965 		char *servername = NULL;
3966 
3967 		if (!srv_loc->len)
3968 			continue;
3969 
3970 		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3971 						&addr, sizeof(addr),
3972 						clp->cl_net, server->port);
3973 		if (!addrlen)
3974 			return;
3975 		xprt_args.dstaddr = &addr;
3976 		xprt_args.addrlen = addrlen;
3977 		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3978 		if (!servername)
3979 			return;
3980 		memcpy(servername, srv_loc->data, srv_loc->len);
3981 		servername[srv_loc->len] = '\0';
3982 		xprt_args.servername = servername;
3983 
3984 		xprtdata.cred = nfs4_get_clid_cred(clp);
3985 		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3986 				  rpc_clnt_setup_test_and_add_xprt,
3987 				  &rpcdata);
3988 		if (xprtdata.cred)
3989 			put_cred(xprtdata.cred);
3990 		kfree(servername);
3991 	}
3992 }
3993 
3994 static int _nfs4_discover_trunking(struct nfs_server *server,
3995 				   struct nfs_fh *fhandle)
3996 {
3997 	struct nfs4_fs_locations *locations = NULL;
3998 	struct page *page;
3999 	const struct cred *cred;
4000 	struct nfs_client *clp = server->nfs_client;
4001 	const struct nfs4_state_maintenance_ops *ops =
4002 		clp->cl_mvops->state_renewal_ops;
4003 	int status = -ENOMEM, i;
4004 
4005 	cred = ops->get_state_renewal_cred(clp);
4006 	if (cred == NULL) {
4007 		cred = nfs4_get_clid_cred(clp);
4008 		if (cred == NULL)
4009 			return -ENOKEY;
4010 	}
4011 
4012 	page = alloc_page(GFP_KERNEL);
4013 	if (!page)
4014 		return -ENOMEM;
4015 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4016 	if (!locations)
4017 		goto out_free;
4018 	locations->fattr = nfs_alloc_fattr();
4019 	if (!locations->fattr)
4020 		goto out_free_2;
4021 
4022 	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4023 					 cred);
4024 	if (status)
4025 		goto out_free_3;
4026 
4027 	for (i = 0; i < locations->nlocations; i++)
4028 		test_fs_location_for_trunking(&locations->locations[i], clp,
4029 					      server);
4030 out_free_3:
4031 	kfree(locations->fattr);
4032 out_free_2:
4033 	kfree(locations);
4034 out_free:
4035 	__free_page(page);
4036 	return status;
4037 }
4038 
4039 static int nfs4_discover_trunking(struct nfs_server *server,
4040 				  struct nfs_fh *fhandle)
4041 {
4042 	struct nfs4_exception exception = {
4043 		.interruptible = true,
4044 	};
4045 	struct nfs_client *clp = server->nfs_client;
4046 	int err = 0;
4047 
4048 	if (!nfs4_has_session(clp))
4049 		goto out;
4050 	do {
4051 		err = nfs4_handle_exception(server,
4052 				_nfs4_discover_trunking(server, fhandle),
4053 				&exception);
4054 	} while (exception.retry);
4055 out:
4056 	return err;
4057 }
4058 
4059 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4060 		struct nfs_fsinfo *info)
4061 {
4062 	u32 bitmask[3];
4063 	struct nfs4_lookup_root_arg args = {
4064 		.bitmask = bitmask,
4065 	};
4066 	struct nfs4_lookup_res res = {
4067 		.server = server,
4068 		.fattr = info->fattr,
4069 		.fh = fhandle,
4070 	};
4071 	struct rpc_message msg = {
4072 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4073 		.rpc_argp = &args,
4074 		.rpc_resp = &res,
4075 	};
4076 
4077 	bitmask[0] = nfs4_fattr_bitmap[0];
4078 	bitmask[1] = nfs4_fattr_bitmap[1];
4079 	/*
4080 	 * Process the label in the upcoming getfattr
4081 	 */
4082 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4083 
4084 	nfs_fattr_init(info->fattr);
4085 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4086 }
4087 
4088 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4089 		struct nfs_fsinfo *info)
4090 {
4091 	struct nfs4_exception exception = {
4092 		.interruptible = true,
4093 	};
4094 	int err;
4095 	do {
4096 		err = _nfs4_lookup_root(server, fhandle, info);
4097 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4098 		switch (err) {
4099 		case 0:
4100 		case -NFS4ERR_WRONGSEC:
4101 			goto out;
4102 		default:
4103 			err = nfs4_handle_exception(server, err, &exception);
4104 		}
4105 	} while (exception.retry);
4106 out:
4107 	return err;
4108 }
4109 
4110 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4111 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4112 {
4113 	struct rpc_auth_create_args auth_args = {
4114 		.pseudoflavor = flavor,
4115 	};
4116 	struct rpc_auth *auth;
4117 
4118 	auth = rpcauth_create(&auth_args, server->client);
4119 	if (IS_ERR(auth))
4120 		return -EACCES;
4121 	return nfs4_lookup_root(server, fhandle, info);
4122 }
4123 
4124 /*
4125  * Retry pseudoroot lookup with various security flavors.  We do this when:
4126  *
4127  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4128  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4129  *
4130  * Returns zero on success, or a negative NFS4ERR value, or a
4131  * negative errno value.
4132  */
4133 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4134 			      struct nfs_fsinfo *info)
4135 {
4136 	/* Per 3530bis 15.33.5 */
4137 	static const rpc_authflavor_t flav_array[] = {
4138 		RPC_AUTH_GSS_KRB5P,
4139 		RPC_AUTH_GSS_KRB5I,
4140 		RPC_AUTH_GSS_KRB5,
4141 		RPC_AUTH_UNIX,			/* courtesy */
4142 		RPC_AUTH_NULL,
4143 	};
4144 	int status = -EPERM;
4145 	size_t i;
4146 
4147 	if (server->auth_info.flavor_len > 0) {
4148 		/* try each flavor specified by user */
4149 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4150 			status = nfs4_lookup_root_sec(server, fhandle, info,
4151 						server->auth_info.flavors[i]);
4152 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4153 				continue;
4154 			break;
4155 		}
4156 	} else {
4157 		/* no flavors specified by user, try default list */
4158 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4159 			status = nfs4_lookup_root_sec(server, fhandle, info,
4160 						      flav_array[i]);
4161 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4162 				continue;
4163 			break;
4164 		}
4165 	}
4166 
4167 	/*
4168 	 * -EACCES could mean that the user doesn't have correct permissions
4169 	 * to access the mount.  It could also mean that we tried to mount
4170 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4171 	 * existing mount programs don't handle -EACCES very well so it should
4172 	 * be mapped to -EPERM instead.
4173 	 */
4174 	if (status == -EACCES)
4175 		status = -EPERM;
4176 	return status;
4177 }
4178 
4179 /**
4180  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4181  * @server: initialized nfs_server handle
4182  * @fhandle: we fill in the pseudo-fs root file handle
4183  * @info: we fill in an FSINFO struct
4184  * @auth_probe: probe the auth flavours
4185  *
4186  * Returns zero on success, or a negative errno.
4187  */
4188 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4189 			 struct nfs_fsinfo *info,
4190 			 bool auth_probe)
4191 {
4192 	int status = 0;
4193 
4194 	if (!auth_probe)
4195 		status = nfs4_lookup_root(server, fhandle, info);
4196 
4197 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4198 		status = server->nfs_client->cl_mvops->find_root_sec(server,
4199 				fhandle, info);
4200 
4201 	if (status == 0)
4202 		status = nfs4_server_capabilities(server, fhandle);
4203 	if (status == 0)
4204 		status = nfs4_do_fsinfo(server, fhandle, info);
4205 
4206 	return nfs4_map_errors(status);
4207 }
4208 
4209 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4210 			      struct nfs_fsinfo *info)
4211 {
4212 	int error;
4213 	struct nfs_fattr *fattr = info->fattr;
4214 
4215 	error = nfs4_server_capabilities(server, mntfh);
4216 	if (error < 0) {
4217 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4218 		return error;
4219 	}
4220 
4221 	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4222 	if (error < 0) {
4223 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4224 		goto out;
4225 	}
4226 
4227 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4228 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4229 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4230 
4231 out:
4232 	return error;
4233 }
4234 
4235 /*
4236  * Get locations and (maybe) other attributes of a referral.
4237  * Note that we'll actually follow the referral later when
4238  * we detect fsid mismatch in inode revalidation
4239  */
4240 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4241 			     const struct qstr *name, struct nfs_fattr *fattr,
4242 			     struct nfs_fh *fhandle)
4243 {
4244 	int status = -ENOMEM;
4245 	struct page *page = NULL;
4246 	struct nfs4_fs_locations *locations = NULL;
4247 
4248 	page = alloc_page(GFP_KERNEL);
4249 	if (page == NULL)
4250 		goto out;
4251 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4252 	if (locations == NULL)
4253 		goto out;
4254 
4255 	locations->fattr = fattr;
4256 
4257 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4258 	if (status != 0)
4259 		goto out;
4260 
4261 	/*
4262 	 * If the fsid didn't change, this is a migration event, not a
4263 	 * referral.  Cause us to drop into the exception handler, which
4264 	 * will kick off migration recovery.
4265 	 */
4266 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4267 		dprintk("%s: server did not return a different fsid for"
4268 			" a referral at %s\n", __func__, name->name);
4269 		status = -NFS4ERR_MOVED;
4270 		goto out;
4271 	}
4272 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4273 	nfs_fixup_referral_attributes(fattr);
4274 	memset(fhandle, 0, sizeof(struct nfs_fh));
4275 out:
4276 	if (page)
4277 		__free_page(page);
4278 	kfree(locations);
4279 	return status;
4280 }
4281 
4282 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4283 				struct nfs_fattr *fattr, struct inode *inode)
4284 {
4285 	__u32 bitmask[NFS4_BITMASK_SZ];
4286 	struct nfs4_getattr_arg args = {
4287 		.fh = fhandle,
4288 		.bitmask = bitmask,
4289 	};
4290 	struct nfs4_getattr_res res = {
4291 		.fattr = fattr,
4292 		.server = server,
4293 	};
4294 	struct rpc_message msg = {
4295 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4296 		.rpc_argp = &args,
4297 		.rpc_resp = &res,
4298 	};
4299 	unsigned short task_flags = 0;
4300 
4301 	if (nfs4_has_session(server->nfs_client))
4302 		task_flags = RPC_TASK_MOVEABLE;
4303 
4304 	/* Is this is an attribute revalidation, subject to softreval? */
4305 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4306 		task_flags |= RPC_TASK_TIMEOUT;
4307 
4308 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4309 	nfs_fattr_init(fattr);
4310 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4311 	return nfs4_do_call_sync(server->client, server, &msg,
4312 			&args.seq_args, &res.seq_res, task_flags);
4313 }
4314 
4315 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4316 				struct nfs_fattr *fattr, struct inode *inode)
4317 {
4318 	struct nfs4_exception exception = {
4319 		.interruptible = true,
4320 	};
4321 	int err;
4322 	do {
4323 		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4324 		trace_nfs4_getattr(server, fhandle, fattr, err);
4325 		err = nfs4_handle_exception(server, err,
4326 				&exception);
4327 	} while (exception.retry);
4328 	return err;
4329 }
4330 
4331 /*
4332  * The file is not closed if it is opened due to the a request to change
4333  * the size of the file. The open call will not be needed once the
4334  * VFS layer lookup-intents are implemented.
4335  *
4336  * Close is called when the inode is destroyed.
4337  * If we haven't opened the file for O_WRONLY, we
4338  * need to in the size_change case to obtain a stateid.
4339  *
4340  * Got race?
4341  * Because OPEN is always done by name in nfsv4, it is
4342  * possible that we opened a different file by the same
4343  * name.  We can recognize this race condition, but we
4344  * can't do anything about it besides returning an error.
4345  *
4346  * This will be fixed with VFS changes (lookup-intent).
4347  */
4348 static int
4349 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4350 		  struct iattr *sattr)
4351 {
4352 	struct inode *inode = d_inode(dentry);
4353 	const struct cred *cred = NULL;
4354 	struct nfs_open_context *ctx = NULL;
4355 	int status;
4356 
4357 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4358 	    sattr->ia_valid & ATTR_SIZE &&
4359 	    sattr->ia_size < i_size_read(inode))
4360 		pnfs_commit_and_return_layout(inode);
4361 
4362 	nfs_fattr_init(fattr);
4363 
4364 	/* Deal with open(O_TRUNC) */
4365 	if (sattr->ia_valid & ATTR_OPEN)
4366 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4367 
4368 	/* Optimization: if the end result is no change, don't RPC */
4369 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4370 		return 0;
4371 
4372 	/* Search for an existing open(O_WRITE) file */
4373 	if (sattr->ia_valid & ATTR_FILE) {
4374 
4375 		ctx = nfs_file_open_context(sattr->ia_file);
4376 		if (ctx)
4377 			cred = ctx->cred;
4378 	}
4379 
4380 	/* Return any delegations if we're going to change ACLs */
4381 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4382 		nfs4_inode_make_writeable(inode);
4383 
4384 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4385 	if (status == 0) {
4386 		nfs_setattr_update_inode(inode, sattr, fattr);
4387 		nfs_setsecurity(inode, fattr);
4388 	}
4389 	return status;
4390 }
4391 
4392 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4393 		struct dentry *dentry, struct nfs_fh *fhandle,
4394 		struct nfs_fattr *fattr)
4395 {
4396 	struct nfs_server *server = NFS_SERVER(dir);
4397 	int		       status;
4398 	struct nfs4_lookup_arg args = {
4399 		.bitmask = server->attr_bitmask,
4400 		.dir_fh = NFS_FH(dir),
4401 		.name = &dentry->d_name,
4402 	};
4403 	struct nfs4_lookup_res res = {
4404 		.server = server,
4405 		.fattr = fattr,
4406 		.fh = fhandle,
4407 	};
4408 	struct rpc_message msg = {
4409 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4410 		.rpc_argp = &args,
4411 		.rpc_resp = &res,
4412 	};
4413 	unsigned short task_flags = 0;
4414 
4415 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4416 		task_flags = RPC_TASK_MOVEABLE;
4417 
4418 	/* Is this is an attribute revalidation, subject to softreval? */
4419 	if (nfs_lookup_is_soft_revalidate(dentry))
4420 		task_flags |= RPC_TASK_TIMEOUT;
4421 
4422 	args.bitmask = nfs4_bitmask(server, fattr->label);
4423 
4424 	nfs_fattr_init(fattr);
4425 
4426 	dprintk("NFS call  lookup %pd2\n", dentry);
4427 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4428 	status = nfs4_do_call_sync(clnt, server, &msg,
4429 			&args.seq_args, &res.seq_res, task_flags);
4430 	dprintk("NFS reply lookup: %d\n", status);
4431 	return status;
4432 }
4433 
4434 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4435 {
4436 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4437 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4438 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4439 	fattr->nlink = 2;
4440 }
4441 
4442 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4443 				   struct dentry *dentry, struct nfs_fh *fhandle,
4444 				   struct nfs_fattr *fattr)
4445 {
4446 	struct nfs4_exception exception = {
4447 		.interruptible = true,
4448 	};
4449 	struct rpc_clnt *client = *clnt;
4450 	const struct qstr *name = &dentry->d_name;
4451 	int err;
4452 	do {
4453 		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4454 		trace_nfs4_lookup(dir, name, err);
4455 		switch (err) {
4456 		case -NFS4ERR_BADNAME:
4457 			err = -ENOENT;
4458 			goto out;
4459 		case -NFS4ERR_MOVED:
4460 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4461 			if (err == -NFS4ERR_MOVED)
4462 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4463 			goto out;
4464 		case -NFS4ERR_WRONGSEC:
4465 			err = -EPERM;
4466 			if (client != *clnt)
4467 				goto out;
4468 			client = nfs4_negotiate_security(client, dir, name);
4469 			if (IS_ERR(client))
4470 				return PTR_ERR(client);
4471 
4472 			exception.retry = 1;
4473 			break;
4474 		default:
4475 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4476 		}
4477 	} while (exception.retry);
4478 
4479 out:
4480 	if (err == 0)
4481 		*clnt = client;
4482 	else if (client != *clnt)
4483 		rpc_shutdown_client(client);
4484 
4485 	return err;
4486 }
4487 
4488 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4489 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4490 {
4491 	int status;
4492 	struct rpc_clnt *client = NFS_CLIENT(dir);
4493 
4494 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4495 	if (client != NFS_CLIENT(dir)) {
4496 		rpc_shutdown_client(client);
4497 		nfs_fixup_secinfo_attributes(fattr);
4498 	}
4499 	return status;
4500 }
4501 
4502 struct rpc_clnt *
4503 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4504 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4505 {
4506 	struct rpc_clnt *client = NFS_CLIENT(dir);
4507 	int status;
4508 
4509 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4510 	if (status < 0)
4511 		return ERR_PTR(status);
4512 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4513 }
4514 
4515 static int _nfs4_proc_lookupp(struct inode *inode,
4516 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4517 {
4518 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4519 	struct nfs_server *server = NFS_SERVER(inode);
4520 	int		       status;
4521 	struct nfs4_lookupp_arg args = {
4522 		.bitmask = server->attr_bitmask,
4523 		.fh = NFS_FH(inode),
4524 	};
4525 	struct nfs4_lookupp_res res = {
4526 		.server = server,
4527 		.fattr = fattr,
4528 		.fh = fhandle,
4529 	};
4530 	struct rpc_message msg = {
4531 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4532 		.rpc_argp = &args,
4533 		.rpc_resp = &res,
4534 	};
4535 	unsigned short task_flags = 0;
4536 
4537 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4538 		task_flags |= RPC_TASK_TIMEOUT;
4539 
4540 	args.bitmask = nfs4_bitmask(server, fattr->label);
4541 
4542 	nfs_fattr_init(fattr);
4543 
4544 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4545 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4546 				&res.seq_res, task_flags);
4547 	dprintk("NFS reply lookupp: %d\n", status);
4548 	return status;
4549 }
4550 
4551 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4552 			     struct nfs_fattr *fattr)
4553 {
4554 	struct nfs4_exception exception = {
4555 		.interruptible = true,
4556 	};
4557 	int err;
4558 	do {
4559 		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4560 		trace_nfs4_lookupp(inode, err);
4561 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4562 				&exception);
4563 	} while (exception.retry);
4564 	return err;
4565 }
4566 
4567 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4568 			     const struct cred *cred)
4569 {
4570 	struct nfs_server *server = NFS_SERVER(inode);
4571 	struct nfs4_accessargs args = {
4572 		.fh = NFS_FH(inode),
4573 		.access = entry->mask,
4574 	};
4575 	struct nfs4_accessres res = {
4576 		.server = server,
4577 	};
4578 	struct rpc_message msg = {
4579 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4580 		.rpc_argp = &args,
4581 		.rpc_resp = &res,
4582 		.rpc_cred = cred,
4583 	};
4584 	int status = 0;
4585 
4586 	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4587 		res.fattr = nfs_alloc_fattr();
4588 		if (res.fattr == NULL)
4589 			return -ENOMEM;
4590 		args.bitmask = server->cache_consistency_bitmask;
4591 	}
4592 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4593 	if (!status) {
4594 		nfs_access_set_mask(entry, res.access);
4595 		if (res.fattr)
4596 			nfs_refresh_inode(inode, res.fattr);
4597 	}
4598 	nfs_free_fattr(res.fattr);
4599 	return status;
4600 }
4601 
4602 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4603 			    const struct cred *cred)
4604 {
4605 	struct nfs4_exception exception = {
4606 		.interruptible = true,
4607 	};
4608 	int err;
4609 	do {
4610 		err = _nfs4_proc_access(inode, entry, cred);
4611 		trace_nfs4_access(inode, err);
4612 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4613 				&exception);
4614 	} while (exception.retry);
4615 	return err;
4616 }
4617 
4618 /*
4619  * TODO: For the time being, we don't try to get any attributes
4620  * along with any of the zero-copy operations READ, READDIR,
4621  * READLINK, WRITE.
4622  *
4623  * In the case of the first three, we want to put the GETATTR
4624  * after the read-type operation -- this is because it is hard
4625  * to predict the length of a GETATTR response in v4, and thus
4626  * align the READ data correctly.  This means that the GETATTR
4627  * may end up partially falling into the page cache, and we should
4628  * shift it into the 'tail' of the xdr_buf before processing.
4629  * To do this efficiently, we need to know the total length
4630  * of data received, which doesn't seem to be available outside
4631  * of the RPC layer.
4632  *
4633  * In the case of WRITE, we also want to put the GETATTR after
4634  * the operation -- in this case because we want to make sure
4635  * we get the post-operation mtime and size.
4636  *
4637  * Both of these changes to the XDR layer would in fact be quite
4638  * minor, but I decided to leave them for a subsequent patch.
4639  */
4640 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4641 		unsigned int pgbase, unsigned int pglen)
4642 {
4643 	struct nfs4_readlink args = {
4644 		.fh       = NFS_FH(inode),
4645 		.pgbase	  = pgbase,
4646 		.pglen    = pglen,
4647 		.pages    = &page,
4648 	};
4649 	struct nfs4_readlink_res res;
4650 	struct rpc_message msg = {
4651 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4652 		.rpc_argp = &args,
4653 		.rpc_resp = &res,
4654 	};
4655 
4656 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4657 }
4658 
4659 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4660 		unsigned int pgbase, unsigned int pglen)
4661 {
4662 	struct nfs4_exception exception = {
4663 		.interruptible = true,
4664 	};
4665 	int err;
4666 	do {
4667 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4668 		trace_nfs4_readlink(inode, err);
4669 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4670 				&exception);
4671 	} while (exception.retry);
4672 	return err;
4673 }
4674 
4675 /*
4676  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4677  */
4678 static int
4679 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4680 		 int flags)
4681 {
4682 	struct nfs_server *server = NFS_SERVER(dir);
4683 	struct nfs4_label l, *ilabel = NULL;
4684 	struct nfs_open_context *ctx;
4685 	struct nfs4_state *state;
4686 	int status = 0;
4687 
4688 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4689 	if (IS_ERR(ctx))
4690 		return PTR_ERR(ctx);
4691 
4692 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4693 
4694 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4695 		sattr->ia_mode &= ~current_umask();
4696 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4697 	if (IS_ERR(state)) {
4698 		status = PTR_ERR(state);
4699 		goto out;
4700 	}
4701 out:
4702 	nfs4_label_release_security(ilabel);
4703 	put_nfs_open_context(ctx);
4704 	return status;
4705 }
4706 
4707 static int
4708 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4709 {
4710 	struct nfs_server *server = NFS_SERVER(dir);
4711 	struct nfs_removeargs args = {
4712 		.fh = NFS_FH(dir),
4713 		.name = *name,
4714 	};
4715 	struct nfs_removeres res = {
4716 		.server = server,
4717 	};
4718 	struct rpc_message msg = {
4719 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4720 		.rpc_argp = &args,
4721 		.rpc_resp = &res,
4722 	};
4723 	unsigned long timestamp = jiffies;
4724 	int status;
4725 
4726 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4727 	if (status == 0) {
4728 		spin_lock(&dir->i_lock);
4729 		/* Removing a directory decrements nlink in the parent */
4730 		if (ftype == NF4DIR && dir->i_nlink > 2)
4731 			nfs4_dec_nlink_locked(dir);
4732 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4733 					      NFS_INO_INVALID_DATA);
4734 		spin_unlock(&dir->i_lock);
4735 	}
4736 	return status;
4737 }
4738 
4739 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4740 {
4741 	struct nfs4_exception exception = {
4742 		.interruptible = true,
4743 	};
4744 	struct inode *inode = d_inode(dentry);
4745 	int err;
4746 
4747 	if (inode) {
4748 		if (inode->i_nlink == 1)
4749 			nfs4_inode_return_delegation(inode);
4750 		else
4751 			nfs4_inode_make_writeable(inode);
4752 	}
4753 	do {
4754 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4755 		trace_nfs4_remove(dir, &dentry->d_name, err);
4756 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4757 				&exception);
4758 	} while (exception.retry);
4759 	return err;
4760 }
4761 
4762 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4763 {
4764 	struct nfs4_exception exception = {
4765 		.interruptible = true,
4766 	};
4767 	int err;
4768 
4769 	do {
4770 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4771 		trace_nfs4_remove(dir, name, err);
4772 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4773 				&exception);
4774 	} while (exception.retry);
4775 	return err;
4776 }
4777 
4778 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4779 		struct dentry *dentry,
4780 		struct inode *inode)
4781 {
4782 	struct nfs_removeargs *args = msg->rpc_argp;
4783 	struct nfs_removeres *res = msg->rpc_resp;
4784 
4785 	res->server = NFS_SB(dentry->d_sb);
4786 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4787 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4788 
4789 	nfs_fattr_init(res->dir_attr);
4790 
4791 	if (inode) {
4792 		nfs4_inode_return_delegation(inode);
4793 		nfs_d_prune_case_insensitive_aliases(inode);
4794 	}
4795 }
4796 
4797 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4798 {
4799 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4800 			&data->args.seq_args,
4801 			&data->res.seq_res,
4802 			task);
4803 }
4804 
4805 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4806 {
4807 	struct nfs_unlinkdata *data = task->tk_calldata;
4808 	struct nfs_removeres *res = &data->res;
4809 
4810 	if (!nfs4_sequence_done(task, &res->seq_res))
4811 		return 0;
4812 	if (nfs4_async_handle_error(task, res->server, NULL,
4813 				    &data->timeout) == -EAGAIN)
4814 		return 0;
4815 	if (task->tk_status == 0)
4816 		nfs4_update_changeattr(dir, &res->cinfo,
4817 				res->dir_attr->time_start,
4818 				NFS_INO_INVALID_DATA);
4819 	return 1;
4820 }
4821 
4822 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4823 		struct dentry *old_dentry,
4824 		struct dentry *new_dentry)
4825 {
4826 	struct nfs_renameargs *arg = msg->rpc_argp;
4827 	struct nfs_renameres *res = msg->rpc_resp;
4828 	struct inode *old_inode = d_inode(old_dentry);
4829 	struct inode *new_inode = d_inode(new_dentry);
4830 
4831 	if (old_inode)
4832 		nfs4_inode_make_writeable(old_inode);
4833 	if (new_inode)
4834 		nfs4_inode_return_delegation(new_inode);
4835 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4836 	res->server = NFS_SB(old_dentry->d_sb);
4837 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4838 }
4839 
4840 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4841 {
4842 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4843 			&data->args.seq_args,
4844 			&data->res.seq_res,
4845 			task);
4846 }
4847 
4848 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4849 				 struct inode *new_dir)
4850 {
4851 	struct nfs_renamedata *data = task->tk_calldata;
4852 	struct nfs_renameres *res = &data->res;
4853 
4854 	if (!nfs4_sequence_done(task, &res->seq_res))
4855 		return 0;
4856 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4857 		return 0;
4858 
4859 	if (task->tk_status == 0) {
4860 		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4861 		if (new_dir != old_dir) {
4862 			/* Note: If we moved a directory, nlink will change */
4863 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4864 					res->old_fattr->time_start,
4865 					NFS_INO_INVALID_NLINK |
4866 					    NFS_INO_INVALID_DATA);
4867 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4868 					res->new_fattr->time_start,
4869 					NFS_INO_INVALID_NLINK |
4870 					    NFS_INO_INVALID_DATA);
4871 		} else
4872 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4873 					res->old_fattr->time_start,
4874 					NFS_INO_INVALID_DATA);
4875 	}
4876 	return 1;
4877 }
4878 
4879 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4880 {
4881 	struct nfs_server *server = NFS_SERVER(inode);
4882 	__u32 bitmask[NFS4_BITMASK_SZ];
4883 	struct nfs4_link_arg arg = {
4884 		.fh     = NFS_FH(inode),
4885 		.dir_fh = NFS_FH(dir),
4886 		.name   = name,
4887 		.bitmask = bitmask,
4888 	};
4889 	struct nfs4_link_res res = {
4890 		.server = server,
4891 	};
4892 	struct rpc_message msg = {
4893 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4894 		.rpc_argp = &arg,
4895 		.rpc_resp = &res,
4896 	};
4897 	int status = -ENOMEM;
4898 
4899 	res.fattr = nfs_alloc_fattr_with_label(server);
4900 	if (res.fattr == NULL)
4901 		goto out;
4902 
4903 	nfs4_inode_make_writeable(inode);
4904 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4905 				NFS_INO_INVALID_CHANGE);
4906 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4907 	if (!status) {
4908 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4909 				       NFS_INO_INVALID_DATA);
4910 		nfs4_inc_nlink(inode);
4911 		status = nfs_post_op_update_inode(inode, res.fattr);
4912 		if (!status)
4913 			nfs_setsecurity(inode, res.fattr);
4914 	}
4915 
4916 out:
4917 	nfs_free_fattr(res.fattr);
4918 	return status;
4919 }
4920 
4921 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4922 {
4923 	struct nfs4_exception exception = {
4924 		.interruptible = true,
4925 	};
4926 	int err;
4927 	do {
4928 		err = nfs4_handle_exception(NFS_SERVER(inode),
4929 				_nfs4_proc_link(inode, dir, name),
4930 				&exception);
4931 	} while (exception.retry);
4932 	return err;
4933 }
4934 
4935 struct nfs4_createdata {
4936 	struct rpc_message msg;
4937 	struct nfs4_create_arg arg;
4938 	struct nfs4_create_res res;
4939 	struct nfs_fh fh;
4940 	struct nfs_fattr fattr;
4941 };
4942 
4943 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4944 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4945 {
4946 	struct nfs4_createdata *data;
4947 
4948 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4949 	if (data != NULL) {
4950 		struct nfs_server *server = NFS_SERVER(dir);
4951 
4952 		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4953 		if (IS_ERR(data->fattr.label))
4954 			goto out_free;
4955 
4956 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4957 		data->msg.rpc_argp = &data->arg;
4958 		data->msg.rpc_resp = &data->res;
4959 		data->arg.dir_fh = NFS_FH(dir);
4960 		data->arg.server = server;
4961 		data->arg.name = name;
4962 		data->arg.attrs = sattr;
4963 		data->arg.ftype = ftype;
4964 		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4965 		data->arg.umask = current_umask();
4966 		data->res.server = server;
4967 		data->res.fh = &data->fh;
4968 		data->res.fattr = &data->fattr;
4969 		nfs_fattr_init(data->res.fattr);
4970 	}
4971 	return data;
4972 out_free:
4973 	kfree(data);
4974 	return NULL;
4975 }
4976 
4977 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4978 {
4979 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4980 				    &data->arg.seq_args, &data->res.seq_res, 1);
4981 	if (status == 0) {
4982 		spin_lock(&dir->i_lock);
4983 		/* Creating a directory bumps nlink in the parent */
4984 		if (data->arg.ftype == NF4DIR)
4985 			nfs4_inc_nlink_locked(dir);
4986 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4987 					      data->res.fattr->time_start,
4988 					      NFS_INO_INVALID_DATA);
4989 		spin_unlock(&dir->i_lock);
4990 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4991 	}
4992 	return status;
4993 }
4994 
4995 static void nfs4_free_createdata(struct nfs4_createdata *data)
4996 {
4997 	nfs4_label_free(data->fattr.label);
4998 	kfree(data);
4999 }
5000 
5001 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5002 		struct page *page, unsigned int len, struct iattr *sattr,
5003 		struct nfs4_label *label)
5004 {
5005 	struct nfs4_createdata *data;
5006 	int status = -ENAMETOOLONG;
5007 
5008 	if (len > NFS4_MAXPATHLEN)
5009 		goto out;
5010 
5011 	status = -ENOMEM;
5012 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5013 	if (data == NULL)
5014 		goto out;
5015 
5016 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5017 	data->arg.u.symlink.pages = &page;
5018 	data->arg.u.symlink.len = len;
5019 	data->arg.label = label;
5020 
5021 	status = nfs4_do_create(dir, dentry, data);
5022 
5023 	nfs4_free_createdata(data);
5024 out:
5025 	return status;
5026 }
5027 
5028 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5029 		struct page *page, unsigned int len, struct iattr *sattr)
5030 {
5031 	struct nfs4_exception exception = {
5032 		.interruptible = true,
5033 	};
5034 	struct nfs4_label l, *label = NULL;
5035 	int err;
5036 
5037 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5038 
5039 	do {
5040 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5041 		trace_nfs4_symlink(dir, &dentry->d_name, err);
5042 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5043 				&exception);
5044 	} while (exception.retry);
5045 
5046 	nfs4_label_release_security(label);
5047 	return err;
5048 }
5049 
5050 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5051 		struct iattr *sattr, struct nfs4_label *label)
5052 {
5053 	struct nfs4_createdata *data;
5054 	int status = -ENOMEM;
5055 
5056 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5057 	if (data == NULL)
5058 		goto out;
5059 
5060 	data->arg.label = label;
5061 	status = nfs4_do_create(dir, dentry, data);
5062 
5063 	nfs4_free_createdata(data);
5064 out:
5065 	return status;
5066 }
5067 
5068 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5069 		struct iattr *sattr)
5070 {
5071 	struct nfs_server *server = NFS_SERVER(dir);
5072 	struct nfs4_exception exception = {
5073 		.interruptible = true,
5074 	};
5075 	struct nfs4_label l, *label = NULL;
5076 	int err;
5077 
5078 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5079 
5080 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5081 		sattr->ia_mode &= ~current_umask();
5082 	do {
5083 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5084 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5085 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5086 				&exception);
5087 	} while (exception.retry);
5088 	nfs4_label_release_security(label);
5089 
5090 	return err;
5091 }
5092 
5093 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5094 			      struct nfs_readdir_res *nr_res)
5095 {
5096 	struct inode		*dir = d_inode(nr_arg->dentry);
5097 	struct nfs_server	*server = NFS_SERVER(dir);
5098 	struct nfs4_readdir_arg args = {
5099 		.fh = NFS_FH(dir),
5100 		.pages = nr_arg->pages,
5101 		.pgbase = 0,
5102 		.count = nr_arg->page_len,
5103 		.plus = nr_arg->plus,
5104 	};
5105 	struct nfs4_readdir_res res;
5106 	struct rpc_message msg = {
5107 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5108 		.rpc_argp = &args,
5109 		.rpc_resp = &res,
5110 		.rpc_cred = nr_arg->cred,
5111 	};
5112 	int			status;
5113 
5114 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5115 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5116 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5117 		args.bitmask = server->attr_bitmask_nl;
5118 	else
5119 		args.bitmask = server->attr_bitmask;
5120 
5121 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5122 	res.pgbase = args.pgbase;
5123 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5124 			&res.seq_res, 0);
5125 	if (status >= 0) {
5126 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5127 		status += args.pgbase;
5128 	}
5129 
5130 	nfs_invalidate_atime(dir);
5131 
5132 	dprintk("%s: returns %d\n", __func__, status);
5133 	return status;
5134 }
5135 
5136 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5137 			     struct nfs_readdir_res *res)
5138 {
5139 	struct nfs4_exception exception = {
5140 		.interruptible = true,
5141 	};
5142 	int err;
5143 	do {
5144 		err = _nfs4_proc_readdir(arg, res);
5145 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5146 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5147 					    err, &exception);
5148 	} while (exception.retry);
5149 	return err;
5150 }
5151 
5152 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5153 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5154 {
5155 	struct nfs4_createdata *data;
5156 	int mode = sattr->ia_mode;
5157 	int status = -ENOMEM;
5158 
5159 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5160 	if (data == NULL)
5161 		goto out;
5162 
5163 	if (S_ISFIFO(mode))
5164 		data->arg.ftype = NF4FIFO;
5165 	else if (S_ISBLK(mode)) {
5166 		data->arg.ftype = NF4BLK;
5167 		data->arg.u.device.specdata1 = MAJOR(rdev);
5168 		data->arg.u.device.specdata2 = MINOR(rdev);
5169 	}
5170 	else if (S_ISCHR(mode)) {
5171 		data->arg.ftype = NF4CHR;
5172 		data->arg.u.device.specdata1 = MAJOR(rdev);
5173 		data->arg.u.device.specdata2 = MINOR(rdev);
5174 	} else if (!S_ISSOCK(mode)) {
5175 		status = -EINVAL;
5176 		goto out_free;
5177 	}
5178 
5179 	data->arg.label = label;
5180 	status = nfs4_do_create(dir, dentry, data);
5181 out_free:
5182 	nfs4_free_createdata(data);
5183 out:
5184 	return status;
5185 }
5186 
5187 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5188 		struct iattr *sattr, dev_t rdev)
5189 {
5190 	struct nfs_server *server = NFS_SERVER(dir);
5191 	struct nfs4_exception exception = {
5192 		.interruptible = true,
5193 	};
5194 	struct nfs4_label l, *label = NULL;
5195 	int err;
5196 
5197 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5198 
5199 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5200 		sattr->ia_mode &= ~current_umask();
5201 	do {
5202 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5203 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5204 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5205 				&exception);
5206 	} while (exception.retry);
5207 
5208 	nfs4_label_release_security(label);
5209 
5210 	return err;
5211 }
5212 
5213 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5214 		 struct nfs_fsstat *fsstat)
5215 {
5216 	struct nfs4_statfs_arg args = {
5217 		.fh = fhandle,
5218 		.bitmask = server->attr_bitmask,
5219 	};
5220 	struct nfs4_statfs_res res = {
5221 		.fsstat = fsstat,
5222 	};
5223 	struct rpc_message msg = {
5224 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5225 		.rpc_argp = &args,
5226 		.rpc_resp = &res,
5227 	};
5228 
5229 	nfs_fattr_init(fsstat->fattr);
5230 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5231 }
5232 
5233 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5234 {
5235 	struct nfs4_exception exception = {
5236 		.interruptible = true,
5237 	};
5238 	int err;
5239 	do {
5240 		err = nfs4_handle_exception(server,
5241 				_nfs4_proc_statfs(server, fhandle, fsstat),
5242 				&exception);
5243 	} while (exception.retry);
5244 	return err;
5245 }
5246 
5247 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5248 		struct nfs_fsinfo *fsinfo)
5249 {
5250 	struct nfs4_fsinfo_arg args = {
5251 		.fh = fhandle,
5252 		.bitmask = server->attr_bitmask,
5253 	};
5254 	struct nfs4_fsinfo_res res = {
5255 		.fsinfo = fsinfo,
5256 	};
5257 	struct rpc_message msg = {
5258 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5259 		.rpc_argp = &args,
5260 		.rpc_resp = &res,
5261 	};
5262 
5263 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5264 }
5265 
5266 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5267 {
5268 	struct nfs4_exception exception = {
5269 		.interruptible = true,
5270 	};
5271 	int err;
5272 
5273 	do {
5274 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5275 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5276 		if (err == 0) {
5277 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5278 			break;
5279 		}
5280 		err = nfs4_handle_exception(server, err, &exception);
5281 	} while (exception.retry);
5282 	return err;
5283 }
5284 
5285 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5286 {
5287 	int error;
5288 
5289 	nfs_fattr_init(fsinfo->fattr);
5290 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5291 	if (error == 0) {
5292 		/* block layout checks this! */
5293 		server->pnfs_blksize = fsinfo->blksize;
5294 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5295 	}
5296 
5297 	return error;
5298 }
5299 
5300 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5301 		struct nfs_pathconf *pathconf)
5302 {
5303 	struct nfs4_pathconf_arg args = {
5304 		.fh = fhandle,
5305 		.bitmask = server->attr_bitmask,
5306 	};
5307 	struct nfs4_pathconf_res res = {
5308 		.pathconf = pathconf,
5309 	};
5310 	struct rpc_message msg = {
5311 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5312 		.rpc_argp = &args,
5313 		.rpc_resp = &res,
5314 	};
5315 
5316 	/* None of the pathconf attributes are mandatory to implement */
5317 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5318 		memset(pathconf, 0, sizeof(*pathconf));
5319 		return 0;
5320 	}
5321 
5322 	nfs_fattr_init(pathconf->fattr);
5323 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5324 }
5325 
5326 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5327 		struct nfs_pathconf *pathconf)
5328 {
5329 	struct nfs4_exception exception = {
5330 		.interruptible = true,
5331 	};
5332 	int err;
5333 
5334 	do {
5335 		err = nfs4_handle_exception(server,
5336 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5337 				&exception);
5338 	} while (exception.retry);
5339 	return err;
5340 }
5341 
5342 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5343 		const struct nfs_open_context *ctx,
5344 		const struct nfs_lock_context *l_ctx,
5345 		fmode_t fmode)
5346 {
5347 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5348 }
5349 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5350 
5351 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5352 		const struct nfs_open_context *ctx,
5353 		const struct nfs_lock_context *l_ctx,
5354 		fmode_t fmode)
5355 {
5356 	nfs4_stateid _current_stateid;
5357 
5358 	/* If the current stateid represents a lost lock, then exit */
5359 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5360 		return true;
5361 	return nfs4_stateid_match(stateid, &_current_stateid);
5362 }
5363 
5364 static bool nfs4_error_stateid_expired(int err)
5365 {
5366 	switch (err) {
5367 	case -NFS4ERR_DELEG_REVOKED:
5368 	case -NFS4ERR_ADMIN_REVOKED:
5369 	case -NFS4ERR_BAD_STATEID:
5370 	case -NFS4ERR_STALE_STATEID:
5371 	case -NFS4ERR_OLD_STATEID:
5372 	case -NFS4ERR_OPENMODE:
5373 	case -NFS4ERR_EXPIRED:
5374 		return true;
5375 	}
5376 	return false;
5377 }
5378 
5379 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5380 {
5381 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5382 
5383 	trace_nfs4_read(hdr, task->tk_status);
5384 	if (task->tk_status < 0) {
5385 		struct nfs4_exception exception = {
5386 			.inode = hdr->inode,
5387 			.state = hdr->args.context->state,
5388 			.stateid = &hdr->args.stateid,
5389 		};
5390 		task->tk_status = nfs4_async_handle_exception(task,
5391 				server, task->tk_status, &exception);
5392 		if (exception.retry) {
5393 			rpc_restart_call_prepare(task);
5394 			return -EAGAIN;
5395 		}
5396 	}
5397 
5398 	if (task->tk_status > 0)
5399 		renew_lease(server, hdr->timestamp);
5400 	return 0;
5401 }
5402 
5403 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5404 		struct nfs_pgio_args *args)
5405 {
5406 
5407 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5408 		nfs4_stateid_is_current(&args->stateid,
5409 				args->context,
5410 				args->lock_context,
5411 				FMODE_READ))
5412 		return false;
5413 	rpc_restart_call_prepare(task);
5414 	return true;
5415 }
5416 
5417 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5418 					 struct nfs_pgio_header *hdr)
5419 {
5420 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5421 	struct rpc_message *msg = &task->tk_msg;
5422 
5423 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5424 	    server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5425 		server->caps &= ~NFS_CAP_READ_PLUS;
5426 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5427 		rpc_restart_call_prepare(task);
5428 		return true;
5429 	}
5430 	return false;
5431 }
5432 
5433 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5434 {
5435 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5436 		return -EAGAIN;
5437 	if (nfs4_read_stateid_changed(task, &hdr->args))
5438 		return -EAGAIN;
5439 	if (nfs4_read_plus_not_supported(task, hdr))
5440 		return -EAGAIN;
5441 	if (task->tk_status > 0)
5442 		nfs_invalidate_atime(hdr->inode);
5443 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5444 				    nfs4_read_done_cb(task, hdr);
5445 }
5446 
5447 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5448 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5449 				    struct rpc_message *msg)
5450 {
5451 	/* Note: We don't use READ_PLUS with pNFS yet */
5452 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5453 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5454 }
5455 #else
5456 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5457 				    struct rpc_message *msg)
5458 {
5459 }
5460 #endif /* CONFIG_NFS_V4_2 */
5461 
5462 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5463 				 struct rpc_message *msg)
5464 {
5465 	hdr->timestamp   = jiffies;
5466 	if (!hdr->pgio_done_cb)
5467 		hdr->pgio_done_cb = nfs4_read_done_cb;
5468 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5469 	nfs42_read_plus_support(hdr, msg);
5470 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5471 }
5472 
5473 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5474 				      struct nfs_pgio_header *hdr)
5475 {
5476 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5477 			&hdr->args.seq_args,
5478 			&hdr->res.seq_res,
5479 			task))
5480 		return 0;
5481 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5482 				hdr->args.lock_context,
5483 				hdr->rw_mode) == -EIO)
5484 		return -EIO;
5485 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5486 		return -EIO;
5487 	return 0;
5488 }
5489 
5490 static int nfs4_write_done_cb(struct rpc_task *task,
5491 			      struct nfs_pgio_header *hdr)
5492 {
5493 	struct inode *inode = hdr->inode;
5494 
5495 	trace_nfs4_write(hdr, task->tk_status);
5496 	if (task->tk_status < 0) {
5497 		struct nfs4_exception exception = {
5498 			.inode = hdr->inode,
5499 			.state = hdr->args.context->state,
5500 			.stateid = &hdr->args.stateid,
5501 		};
5502 		task->tk_status = nfs4_async_handle_exception(task,
5503 				NFS_SERVER(inode), task->tk_status,
5504 				&exception);
5505 		if (exception.retry) {
5506 			rpc_restart_call_prepare(task);
5507 			return -EAGAIN;
5508 		}
5509 	}
5510 	if (task->tk_status >= 0) {
5511 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5512 		nfs_writeback_update_inode(hdr);
5513 	}
5514 	return 0;
5515 }
5516 
5517 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5518 		struct nfs_pgio_args *args)
5519 {
5520 
5521 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5522 		nfs4_stateid_is_current(&args->stateid,
5523 				args->context,
5524 				args->lock_context,
5525 				FMODE_WRITE))
5526 		return false;
5527 	rpc_restart_call_prepare(task);
5528 	return true;
5529 }
5530 
5531 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5532 {
5533 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5534 		return -EAGAIN;
5535 	if (nfs4_write_stateid_changed(task, &hdr->args))
5536 		return -EAGAIN;
5537 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5538 		nfs4_write_done_cb(task, hdr);
5539 }
5540 
5541 static
5542 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5543 {
5544 	/* Don't request attributes for pNFS or O_DIRECT writes */
5545 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5546 		return false;
5547 	/* Otherwise, request attributes if and only if we don't hold
5548 	 * a delegation
5549 	 */
5550 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5551 }
5552 
5553 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5554 		      struct inode *inode, unsigned long cache_validity)
5555 {
5556 	struct nfs_server *server = NFS_SERVER(inode);
5557 	unsigned int i;
5558 
5559 	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5560 	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5561 
5562 	if (cache_validity & NFS_INO_INVALID_CHANGE)
5563 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5564 	if (cache_validity & NFS_INO_INVALID_ATIME)
5565 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5566 	if (cache_validity & NFS_INO_INVALID_MODE)
5567 		bitmask[1] |= FATTR4_WORD1_MODE;
5568 	if (cache_validity & NFS_INO_INVALID_OTHER)
5569 		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5570 	if (cache_validity & NFS_INO_INVALID_NLINK)
5571 		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5572 	if (cache_validity & NFS_INO_INVALID_CTIME)
5573 		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5574 	if (cache_validity & NFS_INO_INVALID_MTIME)
5575 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5576 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5577 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5578 
5579 	if (cache_validity & NFS_INO_INVALID_SIZE)
5580 		bitmask[0] |= FATTR4_WORD0_SIZE;
5581 
5582 	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5583 		bitmask[i] &= server->attr_bitmask[i];
5584 }
5585 
5586 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5587 				  struct rpc_message *msg,
5588 				  struct rpc_clnt **clnt)
5589 {
5590 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5591 
5592 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5593 		hdr->args.bitmask = NULL;
5594 		hdr->res.fattr = NULL;
5595 	} else {
5596 		nfs4_bitmask_set(hdr->args.bitmask_store,
5597 				 server->cache_consistency_bitmask,
5598 				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5599 		hdr->args.bitmask = hdr->args.bitmask_store;
5600 	}
5601 
5602 	if (!hdr->pgio_done_cb)
5603 		hdr->pgio_done_cb = nfs4_write_done_cb;
5604 	hdr->res.server = server;
5605 	hdr->timestamp   = jiffies;
5606 
5607 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5608 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5609 	nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5610 }
5611 
5612 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5613 {
5614 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5615 			&data->args.seq_args,
5616 			&data->res.seq_res,
5617 			task);
5618 }
5619 
5620 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5621 {
5622 	struct inode *inode = data->inode;
5623 
5624 	trace_nfs4_commit(data, task->tk_status);
5625 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5626 				    NULL, NULL) == -EAGAIN) {
5627 		rpc_restart_call_prepare(task);
5628 		return -EAGAIN;
5629 	}
5630 	return 0;
5631 }
5632 
5633 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5634 {
5635 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5636 		return -EAGAIN;
5637 	return data->commit_done_cb(task, data);
5638 }
5639 
5640 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5641 				   struct rpc_clnt **clnt)
5642 {
5643 	struct nfs_server *server = NFS_SERVER(data->inode);
5644 
5645 	if (data->commit_done_cb == NULL)
5646 		data->commit_done_cb = nfs4_commit_done_cb;
5647 	data->res.server = server;
5648 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5649 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5650 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5651 }
5652 
5653 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5654 				struct nfs_commitres *res)
5655 {
5656 	struct inode *dst_inode = file_inode(dst);
5657 	struct nfs_server *server = NFS_SERVER(dst_inode);
5658 	struct rpc_message msg = {
5659 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5660 		.rpc_argp = args,
5661 		.rpc_resp = res,
5662 	};
5663 
5664 	args->fh = NFS_FH(dst_inode);
5665 	return nfs4_call_sync(server->client, server, &msg,
5666 			&args->seq_args, &res->seq_res, 1);
5667 }
5668 
5669 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5670 {
5671 	struct nfs_commitargs args = {
5672 		.offset = offset,
5673 		.count = count,
5674 	};
5675 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5676 	struct nfs4_exception exception = { };
5677 	int status;
5678 
5679 	do {
5680 		status = _nfs4_proc_commit(dst, &args, res);
5681 		status = nfs4_handle_exception(dst_server, status, &exception);
5682 	} while (exception.retry);
5683 
5684 	return status;
5685 }
5686 
5687 struct nfs4_renewdata {
5688 	struct nfs_client	*client;
5689 	unsigned long		timestamp;
5690 };
5691 
5692 /*
5693  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5694  * standalone procedure for queueing an asynchronous RENEW.
5695  */
5696 static void nfs4_renew_release(void *calldata)
5697 {
5698 	struct nfs4_renewdata *data = calldata;
5699 	struct nfs_client *clp = data->client;
5700 
5701 	if (refcount_read(&clp->cl_count) > 1)
5702 		nfs4_schedule_state_renewal(clp);
5703 	nfs_put_client(clp);
5704 	kfree(data);
5705 }
5706 
5707 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5708 {
5709 	struct nfs4_renewdata *data = calldata;
5710 	struct nfs_client *clp = data->client;
5711 	unsigned long timestamp = data->timestamp;
5712 
5713 	trace_nfs4_renew_async(clp, task->tk_status);
5714 	switch (task->tk_status) {
5715 	case 0:
5716 		break;
5717 	case -NFS4ERR_LEASE_MOVED:
5718 		nfs4_schedule_lease_moved_recovery(clp);
5719 		break;
5720 	default:
5721 		/* Unless we're shutting down, schedule state recovery! */
5722 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5723 			return;
5724 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5725 			nfs4_schedule_lease_recovery(clp);
5726 			return;
5727 		}
5728 		nfs4_schedule_path_down_recovery(clp);
5729 	}
5730 	do_renew_lease(clp, timestamp);
5731 }
5732 
5733 static const struct rpc_call_ops nfs4_renew_ops = {
5734 	.rpc_call_done = nfs4_renew_done,
5735 	.rpc_release = nfs4_renew_release,
5736 };
5737 
5738 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5739 {
5740 	struct rpc_message msg = {
5741 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5742 		.rpc_argp	= clp,
5743 		.rpc_cred	= cred,
5744 	};
5745 	struct nfs4_renewdata *data;
5746 
5747 	if (renew_flags == 0)
5748 		return 0;
5749 	if (!refcount_inc_not_zero(&clp->cl_count))
5750 		return -EIO;
5751 	data = kmalloc(sizeof(*data), GFP_NOFS);
5752 	if (data == NULL) {
5753 		nfs_put_client(clp);
5754 		return -ENOMEM;
5755 	}
5756 	data->client = clp;
5757 	data->timestamp = jiffies;
5758 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5759 			&nfs4_renew_ops, data);
5760 }
5761 
5762 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5763 {
5764 	struct rpc_message msg = {
5765 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5766 		.rpc_argp	= clp,
5767 		.rpc_cred	= cred,
5768 	};
5769 	unsigned long now = jiffies;
5770 	int status;
5771 
5772 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5773 	if (status < 0)
5774 		return status;
5775 	do_renew_lease(clp, now);
5776 	return 0;
5777 }
5778 
5779 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5780 				      enum nfs4_acl_type type)
5781 {
5782 	switch (type) {
5783 	default:
5784 		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5785 	case NFS4ACL_DACL:
5786 		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5787 	case NFS4ACL_SACL:
5788 		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5789 	}
5790 }
5791 
5792 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5793  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5794  * the stack.
5795  */
5796 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5797 
5798 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5799 		struct page **pages)
5800 {
5801 	struct page *newpage, **spages;
5802 	int rc = 0;
5803 	size_t len;
5804 	spages = pages;
5805 
5806 	do {
5807 		len = min_t(size_t, PAGE_SIZE, buflen);
5808 		newpage = alloc_page(GFP_KERNEL);
5809 
5810 		if (newpage == NULL)
5811 			goto unwind;
5812 		memcpy(page_address(newpage), buf, len);
5813 		buf += len;
5814 		buflen -= len;
5815 		*pages++ = newpage;
5816 		rc++;
5817 	} while (buflen != 0);
5818 
5819 	return rc;
5820 
5821 unwind:
5822 	for(; rc > 0; rc--)
5823 		__free_page(spages[rc-1]);
5824 	return -ENOMEM;
5825 }
5826 
5827 struct nfs4_cached_acl {
5828 	enum nfs4_acl_type type;
5829 	int cached;
5830 	size_t len;
5831 	char data[];
5832 };
5833 
5834 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5835 {
5836 	struct nfs_inode *nfsi = NFS_I(inode);
5837 
5838 	spin_lock(&inode->i_lock);
5839 	kfree(nfsi->nfs4_acl);
5840 	nfsi->nfs4_acl = acl;
5841 	spin_unlock(&inode->i_lock);
5842 }
5843 
5844 static void nfs4_zap_acl_attr(struct inode *inode)
5845 {
5846 	nfs4_set_cached_acl(inode, NULL);
5847 }
5848 
5849 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5850 				    size_t buflen, enum nfs4_acl_type type)
5851 {
5852 	struct nfs_inode *nfsi = NFS_I(inode);
5853 	struct nfs4_cached_acl *acl;
5854 	int ret = -ENOENT;
5855 
5856 	spin_lock(&inode->i_lock);
5857 	acl = nfsi->nfs4_acl;
5858 	if (acl == NULL)
5859 		goto out;
5860 	if (acl->type != type)
5861 		goto out;
5862 	if (buf == NULL) /* user is just asking for length */
5863 		goto out_len;
5864 	if (acl->cached == 0)
5865 		goto out;
5866 	ret = -ERANGE; /* see getxattr(2) man page */
5867 	if (acl->len > buflen)
5868 		goto out;
5869 	memcpy(buf, acl->data, acl->len);
5870 out_len:
5871 	ret = acl->len;
5872 out:
5873 	spin_unlock(&inode->i_lock);
5874 	return ret;
5875 }
5876 
5877 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5878 				  size_t pgbase, size_t acl_len,
5879 				  enum nfs4_acl_type type)
5880 {
5881 	struct nfs4_cached_acl *acl;
5882 	size_t buflen = sizeof(*acl) + acl_len;
5883 
5884 	if (buflen <= PAGE_SIZE) {
5885 		acl = kmalloc(buflen, GFP_KERNEL);
5886 		if (acl == NULL)
5887 			goto out;
5888 		acl->cached = 1;
5889 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5890 	} else {
5891 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5892 		if (acl == NULL)
5893 			goto out;
5894 		acl->cached = 0;
5895 	}
5896 	acl->type = type;
5897 	acl->len = acl_len;
5898 out:
5899 	nfs4_set_cached_acl(inode, acl);
5900 }
5901 
5902 /*
5903  * The getxattr API returns the required buffer length when called with a
5904  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5905  * the required buf.  On a NULL buf, we send a page of data to the server
5906  * guessing that the ACL request can be serviced by a page. If so, we cache
5907  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5908  * the cache. If not so, we throw away the page, and cache the required
5909  * length. The next getxattr call will then produce another round trip to
5910  * the server, this time with the input buf of the required size.
5911  */
5912 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5913 				       size_t buflen, enum nfs4_acl_type type)
5914 {
5915 	struct page **pages;
5916 	struct nfs_getaclargs args = {
5917 		.fh = NFS_FH(inode),
5918 		.acl_type = type,
5919 		.acl_len = buflen,
5920 	};
5921 	struct nfs_getaclres res = {
5922 		.acl_type = type,
5923 		.acl_len = buflen,
5924 	};
5925 	struct rpc_message msg = {
5926 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5927 		.rpc_argp = &args,
5928 		.rpc_resp = &res,
5929 	};
5930 	unsigned int npages;
5931 	int ret = -ENOMEM, i;
5932 	struct nfs_server *server = NFS_SERVER(inode);
5933 
5934 	if (buflen == 0)
5935 		buflen = server->rsize;
5936 
5937 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5938 	pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5939 	if (!pages)
5940 		return -ENOMEM;
5941 
5942 	args.acl_pages = pages;
5943 
5944 	for (i = 0; i < npages; i++) {
5945 		pages[i] = alloc_page(GFP_KERNEL);
5946 		if (!pages[i])
5947 			goto out_free;
5948 	}
5949 
5950 	/* for decoding across pages */
5951 	res.acl_scratch = alloc_page(GFP_KERNEL);
5952 	if (!res.acl_scratch)
5953 		goto out_free;
5954 
5955 	args.acl_len = npages * PAGE_SIZE;
5956 
5957 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5958 		__func__, buf, buflen, npages, args.acl_len);
5959 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5960 			     &msg, &args.seq_args, &res.seq_res, 0);
5961 	if (ret)
5962 		goto out_free;
5963 
5964 	/* Handle the case where the passed-in buffer is too short */
5965 	if (res.acl_flags & NFS4_ACL_TRUNC) {
5966 		/* Did the user only issue a request for the acl length? */
5967 		if (buf == NULL)
5968 			goto out_ok;
5969 		ret = -ERANGE;
5970 		goto out_free;
5971 	}
5972 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5973 			      type);
5974 	if (buf) {
5975 		if (res.acl_len > buflen) {
5976 			ret = -ERANGE;
5977 			goto out_free;
5978 		}
5979 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5980 	}
5981 out_ok:
5982 	ret = res.acl_len;
5983 out_free:
5984 	for (i = 0; i < npages; i++)
5985 		if (pages[i])
5986 			__free_page(pages[i]);
5987 	if (res.acl_scratch)
5988 		__free_page(res.acl_scratch);
5989 	kfree(pages);
5990 	return ret;
5991 }
5992 
5993 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
5994 				     size_t buflen, enum nfs4_acl_type type)
5995 {
5996 	struct nfs4_exception exception = {
5997 		.interruptible = true,
5998 	};
5999 	ssize_t ret;
6000 	do {
6001 		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6002 		trace_nfs4_get_acl(inode, ret);
6003 		if (ret >= 0)
6004 			break;
6005 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6006 	} while (exception.retry);
6007 	return ret;
6008 }
6009 
6010 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6011 				 enum nfs4_acl_type type)
6012 {
6013 	struct nfs_server *server = NFS_SERVER(inode);
6014 	int ret;
6015 
6016 	if (!nfs4_server_supports_acls(server, type))
6017 		return -EOPNOTSUPP;
6018 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6019 	if (ret < 0)
6020 		return ret;
6021 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6022 		nfs_zap_acl_cache(inode);
6023 	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6024 	if (ret != -ENOENT)
6025 		/* -ENOENT is returned if there is no ACL or if there is an ACL
6026 		 * but no cached acl data, just the acl length */
6027 		return ret;
6028 	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6029 }
6030 
6031 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6032 			       size_t buflen, enum nfs4_acl_type type)
6033 {
6034 	struct nfs_server *server = NFS_SERVER(inode);
6035 	struct page *pages[NFS4ACL_MAXPAGES];
6036 	struct nfs_setaclargs arg = {
6037 		.fh = NFS_FH(inode),
6038 		.acl_type = type,
6039 		.acl_len = buflen,
6040 		.acl_pages = pages,
6041 	};
6042 	struct nfs_setaclres res;
6043 	struct rpc_message msg = {
6044 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6045 		.rpc_argp	= &arg,
6046 		.rpc_resp	= &res,
6047 	};
6048 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6049 	int ret, i;
6050 
6051 	/* You can't remove system.nfs4_acl: */
6052 	if (buflen == 0)
6053 		return -EINVAL;
6054 	if (!nfs4_server_supports_acls(server, type))
6055 		return -EOPNOTSUPP;
6056 	if (npages > ARRAY_SIZE(pages))
6057 		return -ERANGE;
6058 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6059 	if (i < 0)
6060 		return i;
6061 	nfs4_inode_make_writeable(inode);
6062 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6063 
6064 	/*
6065 	 * Free each page after tx, so the only ref left is
6066 	 * held by the network stack
6067 	 */
6068 	for (; i > 0; i--)
6069 		put_page(pages[i-1]);
6070 
6071 	/*
6072 	 * Acl update can result in inode attribute update.
6073 	 * so mark the attribute cache invalid.
6074 	 */
6075 	spin_lock(&inode->i_lock);
6076 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6077 					     NFS_INO_INVALID_CTIME |
6078 					     NFS_INO_REVAL_FORCED);
6079 	spin_unlock(&inode->i_lock);
6080 	nfs_access_zap_cache(inode);
6081 	nfs_zap_acl_cache(inode);
6082 	return ret;
6083 }
6084 
6085 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6086 			     size_t buflen, enum nfs4_acl_type type)
6087 {
6088 	struct nfs4_exception exception = { };
6089 	int err;
6090 	do {
6091 		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6092 		trace_nfs4_set_acl(inode, err);
6093 		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6094 			/*
6095 			 * no need to retry since the kernel
6096 			 * isn't involved in encoding the ACEs.
6097 			 */
6098 			err = -EINVAL;
6099 			break;
6100 		}
6101 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6102 				&exception);
6103 	} while (exception.retry);
6104 	return err;
6105 }
6106 
6107 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6108 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6109 					size_t buflen)
6110 {
6111 	struct nfs_server *server = NFS_SERVER(inode);
6112 	struct nfs4_label label = {0, 0, buflen, buf};
6113 
6114 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6115 	struct nfs_fattr fattr = {
6116 		.label = &label,
6117 	};
6118 	struct nfs4_getattr_arg arg = {
6119 		.fh		= NFS_FH(inode),
6120 		.bitmask	= bitmask,
6121 	};
6122 	struct nfs4_getattr_res res = {
6123 		.fattr		= &fattr,
6124 		.server		= server,
6125 	};
6126 	struct rpc_message msg = {
6127 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6128 		.rpc_argp	= &arg,
6129 		.rpc_resp	= &res,
6130 	};
6131 	int ret;
6132 
6133 	nfs_fattr_init(&fattr);
6134 
6135 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6136 	if (ret)
6137 		return ret;
6138 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6139 		return -ENOENT;
6140 	return label.len;
6141 }
6142 
6143 static int nfs4_get_security_label(struct inode *inode, void *buf,
6144 					size_t buflen)
6145 {
6146 	struct nfs4_exception exception = {
6147 		.interruptible = true,
6148 	};
6149 	int err;
6150 
6151 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6152 		return -EOPNOTSUPP;
6153 
6154 	do {
6155 		err = _nfs4_get_security_label(inode, buf, buflen);
6156 		trace_nfs4_get_security_label(inode, err);
6157 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6158 				&exception);
6159 	} while (exception.retry);
6160 	return err;
6161 }
6162 
6163 static int _nfs4_do_set_security_label(struct inode *inode,
6164 		struct nfs4_label *ilabel,
6165 		struct nfs_fattr *fattr)
6166 {
6167 
6168 	struct iattr sattr = {0};
6169 	struct nfs_server *server = NFS_SERVER(inode);
6170 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6171 	struct nfs_setattrargs arg = {
6172 		.fh		= NFS_FH(inode),
6173 		.iap		= &sattr,
6174 		.server		= server,
6175 		.bitmask	= bitmask,
6176 		.label		= ilabel,
6177 	};
6178 	struct nfs_setattrres res = {
6179 		.fattr		= fattr,
6180 		.server		= server,
6181 	};
6182 	struct rpc_message msg = {
6183 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6184 		.rpc_argp	= &arg,
6185 		.rpc_resp	= &res,
6186 	};
6187 	int status;
6188 
6189 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6190 
6191 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6192 	if (status)
6193 		dprintk("%s failed: %d\n", __func__, status);
6194 
6195 	return status;
6196 }
6197 
6198 static int nfs4_do_set_security_label(struct inode *inode,
6199 		struct nfs4_label *ilabel,
6200 		struct nfs_fattr *fattr)
6201 {
6202 	struct nfs4_exception exception = { };
6203 	int err;
6204 
6205 	do {
6206 		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6207 		trace_nfs4_set_security_label(inode, err);
6208 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6209 				&exception);
6210 	} while (exception.retry);
6211 	return err;
6212 }
6213 
6214 static int
6215 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6216 {
6217 	struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6218 	struct nfs_fattr *fattr;
6219 	int status;
6220 
6221 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6222 		return -EOPNOTSUPP;
6223 
6224 	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6225 	if (fattr == NULL)
6226 		return -ENOMEM;
6227 
6228 	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6229 	if (status == 0)
6230 		nfs_setsecurity(inode, fattr);
6231 
6232 	return status;
6233 }
6234 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6235 
6236 
6237 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6238 				    nfs4_verifier *bootverf)
6239 {
6240 	__be32 verf[2];
6241 
6242 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6243 		/* An impossible timestamp guarantees this value
6244 		 * will never match a generated boot time. */
6245 		verf[0] = cpu_to_be32(U32_MAX);
6246 		verf[1] = cpu_to_be32(U32_MAX);
6247 	} else {
6248 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6249 		u64 ns = ktime_to_ns(nn->boot_time);
6250 
6251 		verf[0] = cpu_to_be32(ns >> 32);
6252 		verf[1] = cpu_to_be32(ns);
6253 	}
6254 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6255 }
6256 
6257 static size_t
6258 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6259 {
6260 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6261 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6262 	const char *id;
6263 
6264 	buf[0] = '\0';
6265 
6266 	if (nn_clp) {
6267 		rcu_read_lock();
6268 		id = rcu_dereference(nn_clp->identifier);
6269 		if (id)
6270 			strscpy(buf, id, buflen);
6271 		rcu_read_unlock();
6272 	}
6273 
6274 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6275 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6276 
6277 	return strlen(buf);
6278 }
6279 
6280 static int
6281 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6282 {
6283 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6284 	size_t buflen;
6285 	size_t len;
6286 	char *str;
6287 
6288 	if (clp->cl_owner_id != NULL)
6289 		return 0;
6290 
6291 	rcu_read_lock();
6292 	len = 14 +
6293 		strlen(clp->cl_rpcclient->cl_nodename) +
6294 		1 +
6295 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6296 		1;
6297 	rcu_read_unlock();
6298 
6299 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6300 	if (buflen)
6301 		len += buflen + 1;
6302 
6303 	if (len > NFS4_OPAQUE_LIMIT + 1)
6304 		return -EINVAL;
6305 
6306 	/*
6307 	 * Since this string is allocated at mount time, and held until the
6308 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6309 	 * about a memory-reclaim deadlock.
6310 	 */
6311 	str = kmalloc(len, GFP_KERNEL);
6312 	if (!str)
6313 		return -ENOMEM;
6314 
6315 	rcu_read_lock();
6316 	if (buflen)
6317 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6318 			  clp->cl_rpcclient->cl_nodename, buf,
6319 			  rpc_peeraddr2str(clp->cl_rpcclient,
6320 					   RPC_DISPLAY_ADDR));
6321 	else
6322 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6323 			  clp->cl_rpcclient->cl_nodename,
6324 			  rpc_peeraddr2str(clp->cl_rpcclient,
6325 					   RPC_DISPLAY_ADDR));
6326 	rcu_read_unlock();
6327 
6328 	clp->cl_owner_id = str;
6329 	return 0;
6330 }
6331 
6332 static int
6333 nfs4_init_uniform_client_string(struct nfs_client *clp)
6334 {
6335 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6336 	size_t buflen;
6337 	size_t len;
6338 	char *str;
6339 
6340 	if (clp->cl_owner_id != NULL)
6341 		return 0;
6342 
6343 	len = 10 + 10 + 1 + 10 + 1 +
6344 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6345 
6346 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6347 	if (buflen)
6348 		len += buflen + 1;
6349 
6350 	if (len > NFS4_OPAQUE_LIMIT + 1)
6351 		return -EINVAL;
6352 
6353 	/*
6354 	 * Since this string is allocated at mount time, and held until the
6355 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6356 	 * about a memory-reclaim deadlock.
6357 	 */
6358 	str = kmalloc(len, GFP_KERNEL);
6359 	if (!str)
6360 		return -ENOMEM;
6361 
6362 	if (buflen)
6363 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6364 			  clp->rpc_ops->version, clp->cl_minorversion,
6365 			  buf, clp->cl_rpcclient->cl_nodename);
6366 	else
6367 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6368 			  clp->rpc_ops->version, clp->cl_minorversion,
6369 			  clp->cl_rpcclient->cl_nodename);
6370 	clp->cl_owner_id = str;
6371 	return 0;
6372 }
6373 
6374 /*
6375  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6376  * services.  Advertise one based on the address family of the
6377  * clientaddr.
6378  */
6379 static unsigned int
6380 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6381 {
6382 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6383 		return scnprintf(buf, len, "tcp6");
6384 	else
6385 		return scnprintf(buf, len, "tcp");
6386 }
6387 
6388 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6389 {
6390 	struct nfs4_setclientid *sc = calldata;
6391 
6392 	if (task->tk_status == 0)
6393 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6394 }
6395 
6396 static const struct rpc_call_ops nfs4_setclientid_ops = {
6397 	.rpc_call_done = nfs4_setclientid_done,
6398 };
6399 
6400 /**
6401  * nfs4_proc_setclientid - Negotiate client ID
6402  * @clp: state data structure
6403  * @program: RPC program for NFSv4 callback service
6404  * @port: IP port number for NFS4 callback service
6405  * @cred: credential to use for this call
6406  * @res: where to place the result
6407  *
6408  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6409  */
6410 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6411 		unsigned short port, const struct cred *cred,
6412 		struct nfs4_setclientid_res *res)
6413 {
6414 	nfs4_verifier sc_verifier;
6415 	struct nfs4_setclientid setclientid = {
6416 		.sc_verifier = &sc_verifier,
6417 		.sc_prog = program,
6418 		.sc_clnt = clp,
6419 	};
6420 	struct rpc_message msg = {
6421 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6422 		.rpc_argp = &setclientid,
6423 		.rpc_resp = res,
6424 		.rpc_cred = cred,
6425 	};
6426 	struct rpc_task_setup task_setup_data = {
6427 		.rpc_client = clp->cl_rpcclient,
6428 		.rpc_message = &msg,
6429 		.callback_ops = &nfs4_setclientid_ops,
6430 		.callback_data = &setclientid,
6431 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6432 	};
6433 	unsigned long now = jiffies;
6434 	int status;
6435 
6436 	/* nfs_client_id4 */
6437 	nfs4_init_boot_verifier(clp, &sc_verifier);
6438 
6439 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6440 		status = nfs4_init_uniform_client_string(clp);
6441 	else
6442 		status = nfs4_init_nonuniform_client_string(clp);
6443 
6444 	if (status)
6445 		goto out;
6446 
6447 	/* cb_client4 */
6448 	setclientid.sc_netid_len =
6449 				nfs4_init_callback_netid(clp,
6450 						setclientid.sc_netid,
6451 						sizeof(setclientid.sc_netid));
6452 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6453 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6454 				clp->cl_ipaddr, port >> 8, port & 255);
6455 
6456 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6457 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6458 		clp->cl_owner_id);
6459 
6460 	status = nfs4_call_sync_custom(&task_setup_data);
6461 	if (setclientid.sc_cred) {
6462 		kfree(clp->cl_acceptor);
6463 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6464 		put_rpccred(setclientid.sc_cred);
6465 	}
6466 
6467 	if (status == 0)
6468 		do_renew_lease(clp, now);
6469 out:
6470 	trace_nfs4_setclientid(clp, status);
6471 	dprintk("NFS reply setclientid: %d\n", status);
6472 	return status;
6473 }
6474 
6475 /**
6476  * nfs4_proc_setclientid_confirm - Confirm client ID
6477  * @clp: state data structure
6478  * @arg: result of a previous SETCLIENTID
6479  * @cred: credential to use for this call
6480  *
6481  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6482  */
6483 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6484 		struct nfs4_setclientid_res *arg,
6485 		const struct cred *cred)
6486 {
6487 	struct rpc_message msg = {
6488 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6489 		.rpc_argp = arg,
6490 		.rpc_cred = cred,
6491 	};
6492 	int status;
6493 
6494 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6495 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6496 		clp->cl_clientid);
6497 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6498 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6499 	trace_nfs4_setclientid_confirm(clp, status);
6500 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6501 	return status;
6502 }
6503 
6504 struct nfs4_delegreturndata {
6505 	struct nfs4_delegreturnargs args;
6506 	struct nfs4_delegreturnres res;
6507 	struct nfs_fh fh;
6508 	nfs4_stateid stateid;
6509 	unsigned long timestamp;
6510 	struct {
6511 		struct nfs4_layoutreturn_args arg;
6512 		struct nfs4_layoutreturn_res res;
6513 		struct nfs4_xdr_opaque_data ld_private;
6514 		u32 roc_barrier;
6515 		bool roc;
6516 	} lr;
6517 	struct nfs_fattr fattr;
6518 	int rpc_status;
6519 	struct inode *inode;
6520 };
6521 
6522 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6523 {
6524 	struct nfs4_delegreturndata *data = calldata;
6525 	struct nfs4_exception exception = {
6526 		.inode = data->inode,
6527 		.stateid = &data->stateid,
6528 		.task_is_privileged = data->args.seq_args.sa_privileged,
6529 	};
6530 
6531 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6532 		return;
6533 
6534 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6535 
6536 	/* Handle Layoutreturn errors */
6537 	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6538 			  &data->res.lr_ret) == -EAGAIN)
6539 		goto out_restart;
6540 
6541 	switch (task->tk_status) {
6542 	case 0:
6543 		renew_lease(data->res.server, data->timestamp);
6544 		break;
6545 	case -NFS4ERR_ADMIN_REVOKED:
6546 	case -NFS4ERR_DELEG_REVOKED:
6547 	case -NFS4ERR_EXPIRED:
6548 		nfs4_free_revoked_stateid(data->res.server,
6549 				data->args.stateid,
6550 				task->tk_msg.rpc_cred);
6551 		fallthrough;
6552 	case -NFS4ERR_BAD_STATEID:
6553 	case -NFS4ERR_STALE_STATEID:
6554 	case -ETIMEDOUT:
6555 		task->tk_status = 0;
6556 		break;
6557 	case -NFS4ERR_OLD_STATEID:
6558 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6559 			nfs4_stateid_seqid_inc(&data->stateid);
6560 		if (data->args.bitmask) {
6561 			data->args.bitmask = NULL;
6562 			data->res.fattr = NULL;
6563 		}
6564 		goto out_restart;
6565 	case -NFS4ERR_ACCESS:
6566 		if (data->args.bitmask) {
6567 			data->args.bitmask = NULL;
6568 			data->res.fattr = NULL;
6569 			goto out_restart;
6570 		}
6571 		fallthrough;
6572 	default:
6573 		task->tk_status = nfs4_async_handle_exception(task,
6574 				data->res.server, task->tk_status,
6575 				&exception);
6576 		if (exception.retry)
6577 			goto out_restart;
6578 	}
6579 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6580 	data->rpc_status = task->tk_status;
6581 	return;
6582 out_restart:
6583 	task->tk_status = 0;
6584 	rpc_restart_call_prepare(task);
6585 }
6586 
6587 static void nfs4_delegreturn_release(void *calldata)
6588 {
6589 	struct nfs4_delegreturndata *data = calldata;
6590 	struct inode *inode = data->inode;
6591 
6592 	if (data->lr.roc)
6593 		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6594 				 data->res.lr_ret);
6595 	if (inode) {
6596 		nfs4_fattr_set_prechange(&data->fattr,
6597 					 inode_peek_iversion_raw(inode));
6598 		nfs_refresh_inode(inode, &data->fattr);
6599 		nfs_iput_and_deactive(inode);
6600 	}
6601 	kfree(calldata);
6602 }
6603 
6604 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6605 {
6606 	struct nfs4_delegreturndata *d_data;
6607 	struct pnfs_layout_hdr *lo;
6608 
6609 	d_data = (struct nfs4_delegreturndata *)data;
6610 
6611 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6612 		nfs4_sequence_done(task, &d_data->res.seq_res);
6613 		return;
6614 	}
6615 
6616 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6617 	if (lo && !pnfs_layout_is_valid(lo)) {
6618 		d_data->args.lr_args = NULL;
6619 		d_data->res.lr_res = NULL;
6620 	}
6621 
6622 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6623 			&d_data->args.seq_args,
6624 			&d_data->res.seq_res,
6625 			task);
6626 }
6627 
6628 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6629 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6630 	.rpc_call_done = nfs4_delegreturn_done,
6631 	.rpc_release = nfs4_delegreturn_release,
6632 };
6633 
6634 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6635 {
6636 	struct nfs4_delegreturndata *data;
6637 	struct nfs_server *server = NFS_SERVER(inode);
6638 	struct rpc_task *task;
6639 	struct rpc_message msg = {
6640 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6641 		.rpc_cred = cred,
6642 	};
6643 	struct rpc_task_setup task_setup_data = {
6644 		.rpc_client = server->client,
6645 		.rpc_message = &msg,
6646 		.callback_ops = &nfs4_delegreturn_ops,
6647 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6648 	};
6649 	int status = 0;
6650 
6651 	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6652 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6653 
6654 	data = kzalloc(sizeof(*data), GFP_KERNEL);
6655 	if (data == NULL)
6656 		return -ENOMEM;
6657 
6658 	nfs4_state_protect(server->nfs_client,
6659 			NFS_SP4_MACH_CRED_CLEANUP,
6660 			&task_setup_data.rpc_client, &msg);
6661 
6662 	data->args.fhandle = &data->fh;
6663 	data->args.stateid = &data->stateid;
6664 	nfs4_bitmask_set(data->args.bitmask_store,
6665 			 server->cache_consistency_bitmask, inode, 0);
6666 	data->args.bitmask = data->args.bitmask_store;
6667 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6668 	nfs4_stateid_copy(&data->stateid, stateid);
6669 	data->res.fattr = &data->fattr;
6670 	data->res.server = server;
6671 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6672 	data->lr.arg.ld_private = &data->lr.ld_private;
6673 	nfs_fattr_init(data->res.fattr);
6674 	data->timestamp = jiffies;
6675 	data->rpc_status = 0;
6676 	data->inode = nfs_igrab_and_active(inode);
6677 	if (data->inode || issync) {
6678 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6679 					cred);
6680 		if (data->lr.roc) {
6681 			data->args.lr_args = &data->lr.arg;
6682 			data->res.lr_res = &data->lr.res;
6683 		}
6684 	}
6685 
6686 	if (!data->inode)
6687 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6688 				   1);
6689 	else
6690 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6691 				   0);
6692 	task_setup_data.callback_data = data;
6693 	msg.rpc_argp = &data->args;
6694 	msg.rpc_resp = &data->res;
6695 	task = rpc_run_task(&task_setup_data);
6696 	if (IS_ERR(task))
6697 		return PTR_ERR(task);
6698 	if (!issync)
6699 		goto out;
6700 	status = rpc_wait_for_completion_task(task);
6701 	if (status != 0)
6702 		goto out;
6703 	status = data->rpc_status;
6704 out:
6705 	rpc_put_task(task);
6706 	return status;
6707 }
6708 
6709 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6710 {
6711 	struct nfs_server *server = NFS_SERVER(inode);
6712 	struct nfs4_exception exception = { };
6713 	int err;
6714 	do {
6715 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6716 		trace_nfs4_delegreturn(inode, stateid, err);
6717 		switch (err) {
6718 			case -NFS4ERR_STALE_STATEID:
6719 			case -NFS4ERR_EXPIRED:
6720 			case 0:
6721 				return 0;
6722 		}
6723 		err = nfs4_handle_exception(server, err, &exception);
6724 	} while (exception.retry);
6725 	return err;
6726 }
6727 
6728 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6729 {
6730 	struct inode *inode = state->inode;
6731 	struct nfs_server *server = NFS_SERVER(inode);
6732 	struct nfs_client *clp = server->nfs_client;
6733 	struct nfs_lockt_args arg = {
6734 		.fh = NFS_FH(inode),
6735 		.fl = request,
6736 	};
6737 	struct nfs_lockt_res res = {
6738 		.denied = request,
6739 	};
6740 	struct rpc_message msg = {
6741 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6742 		.rpc_argp	= &arg,
6743 		.rpc_resp	= &res,
6744 		.rpc_cred	= state->owner->so_cred,
6745 	};
6746 	struct nfs4_lock_state *lsp;
6747 	int status;
6748 
6749 	arg.lock_owner.clientid = clp->cl_clientid;
6750 	status = nfs4_set_lock_state(state, request);
6751 	if (status != 0)
6752 		goto out;
6753 	lsp = request->fl_u.nfs4_fl.owner;
6754 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6755 	arg.lock_owner.s_dev = server->s_dev;
6756 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6757 	switch (status) {
6758 		case 0:
6759 			request->fl_type = F_UNLCK;
6760 			break;
6761 		case -NFS4ERR_DENIED:
6762 			status = 0;
6763 	}
6764 	request->fl_ops->fl_release_private(request);
6765 	request->fl_ops = NULL;
6766 out:
6767 	return status;
6768 }
6769 
6770 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6771 {
6772 	struct nfs4_exception exception = {
6773 		.interruptible = true,
6774 	};
6775 	int err;
6776 
6777 	do {
6778 		err = _nfs4_proc_getlk(state, cmd, request);
6779 		trace_nfs4_get_lock(request, state, cmd, err);
6780 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6781 				&exception);
6782 	} while (exception.retry);
6783 	return err;
6784 }
6785 
6786 /*
6787  * Update the seqid of a lock stateid after receiving
6788  * NFS4ERR_OLD_STATEID
6789  */
6790 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6791 		struct nfs4_lock_state *lsp)
6792 {
6793 	struct nfs4_state *state = lsp->ls_state;
6794 	bool ret = false;
6795 
6796 	spin_lock(&state->state_lock);
6797 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6798 		goto out;
6799 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6800 		nfs4_stateid_seqid_inc(dst);
6801 	else
6802 		dst->seqid = lsp->ls_stateid.seqid;
6803 	ret = true;
6804 out:
6805 	spin_unlock(&state->state_lock);
6806 	return ret;
6807 }
6808 
6809 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6810 		struct nfs4_lock_state *lsp)
6811 {
6812 	struct nfs4_state *state = lsp->ls_state;
6813 	bool ret;
6814 
6815 	spin_lock(&state->state_lock);
6816 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6817 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6818 	spin_unlock(&state->state_lock);
6819 	return ret;
6820 }
6821 
6822 struct nfs4_unlockdata {
6823 	struct nfs_locku_args arg;
6824 	struct nfs_locku_res res;
6825 	struct nfs4_lock_state *lsp;
6826 	struct nfs_open_context *ctx;
6827 	struct nfs_lock_context *l_ctx;
6828 	struct file_lock fl;
6829 	struct nfs_server *server;
6830 	unsigned long timestamp;
6831 };
6832 
6833 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6834 		struct nfs_open_context *ctx,
6835 		struct nfs4_lock_state *lsp,
6836 		struct nfs_seqid *seqid)
6837 {
6838 	struct nfs4_unlockdata *p;
6839 	struct nfs4_state *state = lsp->ls_state;
6840 	struct inode *inode = state->inode;
6841 
6842 	p = kzalloc(sizeof(*p), GFP_KERNEL);
6843 	if (p == NULL)
6844 		return NULL;
6845 	p->arg.fh = NFS_FH(inode);
6846 	p->arg.fl = &p->fl;
6847 	p->arg.seqid = seqid;
6848 	p->res.seqid = seqid;
6849 	p->lsp = lsp;
6850 	/* Ensure we don't close file until we're done freeing locks! */
6851 	p->ctx = get_nfs_open_context(ctx);
6852 	p->l_ctx = nfs_get_lock_context(ctx);
6853 	locks_init_lock(&p->fl);
6854 	locks_copy_lock(&p->fl, fl);
6855 	p->server = NFS_SERVER(inode);
6856 	spin_lock(&state->state_lock);
6857 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6858 	spin_unlock(&state->state_lock);
6859 	return p;
6860 }
6861 
6862 static void nfs4_locku_release_calldata(void *data)
6863 {
6864 	struct nfs4_unlockdata *calldata = data;
6865 	nfs_free_seqid(calldata->arg.seqid);
6866 	nfs4_put_lock_state(calldata->lsp);
6867 	nfs_put_lock_context(calldata->l_ctx);
6868 	put_nfs_open_context(calldata->ctx);
6869 	kfree(calldata);
6870 }
6871 
6872 static void nfs4_locku_done(struct rpc_task *task, void *data)
6873 {
6874 	struct nfs4_unlockdata *calldata = data;
6875 	struct nfs4_exception exception = {
6876 		.inode = calldata->lsp->ls_state->inode,
6877 		.stateid = &calldata->arg.stateid,
6878 	};
6879 
6880 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6881 		return;
6882 	switch (task->tk_status) {
6883 		case 0:
6884 			renew_lease(calldata->server, calldata->timestamp);
6885 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6886 			if (nfs4_update_lock_stateid(calldata->lsp,
6887 					&calldata->res.stateid))
6888 				break;
6889 			fallthrough;
6890 		case -NFS4ERR_ADMIN_REVOKED:
6891 		case -NFS4ERR_EXPIRED:
6892 			nfs4_free_revoked_stateid(calldata->server,
6893 					&calldata->arg.stateid,
6894 					task->tk_msg.rpc_cred);
6895 			fallthrough;
6896 		case -NFS4ERR_BAD_STATEID:
6897 		case -NFS4ERR_STALE_STATEID:
6898 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6899 						calldata->lsp))
6900 				rpc_restart_call_prepare(task);
6901 			break;
6902 		case -NFS4ERR_OLD_STATEID:
6903 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6904 						calldata->lsp))
6905 				rpc_restart_call_prepare(task);
6906 			break;
6907 		default:
6908 			task->tk_status = nfs4_async_handle_exception(task,
6909 					calldata->server, task->tk_status,
6910 					&exception);
6911 			if (exception.retry)
6912 				rpc_restart_call_prepare(task);
6913 	}
6914 	nfs_release_seqid(calldata->arg.seqid);
6915 }
6916 
6917 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6918 {
6919 	struct nfs4_unlockdata *calldata = data;
6920 
6921 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6922 		nfs_async_iocounter_wait(task, calldata->l_ctx))
6923 		return;
6924 
6925 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6926 		goto out_wait;
6927 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6928 		/* Note: exit _without_ running nfs4_locku_done */
6929 		goto out_no_action;
6930 	}
6931 	calldata->timestamp = jiffies;
6932 	if (nfs4_setup_sequence(calldata->server->nfs_client,
6933 				&calldata->arg.seq_args,
6934 				&calldata->res.seq_res,
6935 				task) != 0)
6936 		nfs_release_seqid(calldata->arg.seqid);
6937 	return;
6938 out_no_action:
6939 	task->tk_action = NULL;
6940 out_wait:
6941 	nfs4_sequence_done(task, &calldata->res.seq_res);
6942 }
6943 
6944 static const struct rpc_call_ops nfs4_locku_ops = {
6945 	.rpc_call_prepare = nfs4_locku_prepare,
6946 	.rpc_call_done = nfs4_locku_done,
6947 	.rpc_release = nfs4_locku_release_calldata,
6948 };
6949 
6950 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6951 		struct nfs_open_context *ctx,
6952 		struct nfs4_lock_state *lsp,
6953 		struct nfs_seqid *seqid)
6954 {
6955 	struct nfs4_unlockdata *data;
6956 	struct rpc_message msg = {
6957 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6958 		.rpc_cred = ctx->cred,
6959 	};
6960 	struct rpc_task_setup task_setup_data = {
6961 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6962 		.rpc_message = &msg,
6963 		.callback_ops = &nfs4_locku_ops,
6964 		.workqueue = nfsiod_workqueue,
6965 		.flags = RPC_TASK_ASYNC,
6966 	};
6967 
6968 	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6969 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6970 
6971 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6972 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6973 
6974 	/* Ensure this is an unlock - when canceling a lock, the
6975 	 * canceled lock is passed in, and it won't be an unlock.
6976 	 */
6977 	fl->fl_type = F_UNLCK;
6978 	if (fl->fl_flags & FL_CLOSE)
6979 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6980 
6981 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6982 	if (data == NULL) {
6983 		nfs_free_seqid(seqid);
6984 		return ERR_PTR(-ENOMEM);
6985 	}
6986 
6987 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6988 	msg.rpc_argp = &data->arg;
6989 	msg.rpc_resp = &data->res;
6990 	task_setup_data.callback_data = data;
6991 	return rpc_run_task(&task_setup_data);
6992 }
6993 
6994 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6995 {
6996 	struct inode *inode = state->inode;
6997 	struct nfs4_state_owner *sp = state->owner;
6998 	struct nfs_inode *nfsi = NFS_I(inode);
6999 	struct nfs_seqid *seqid;
7000 	struct nfs4_lock_state *lsp;
7001 	struct rpc_task *task;
7002 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7003 	int status = 0;
7004 	unsigned char fl_flags = request->fl_flags;
7005 
7006 	status = nfs4_set_lock_state(state, request);
7007 	/* Unlock _before_ we do the RPC call */
7008 	request->fl_flags |= FL_EXISTS;
7009 	/* Exclude nfs_delegation_claim_locks() */
7010 	mutex_lock(&sp->so_delegreturn_mutex);
7011 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7012 	down_read(&nfsi->rwsem);
7013 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7014 		up_read(&nfsi->rwsem);
7015 		mutex_unlock(&sp->so_delegreturn_mutex);
7016 		goto out;
7017 	}
7018 	up_read(&nfsi->rwsem);
7019 	mutex_unlock(&sp->so_delegreturn_mutex);
7020 	if (status != 0)
7021 		goto out;
7022 	/* Is this a delegated lock? */
7023 	lsp = request->fl_u.nfs4_fl.owner;
7024 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7025 		goto out;
7026 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7027 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7028 	status = -ENOMEM;
7029 	if (IS_ERR(seqid))
7030 		goto out;
7031 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7032 	status = PTR_ERR(task);
7033 	if (IS_ERR(task))
7034 		goto out;
7035 	status = rpc_wait_for_completion_task(task);
7036 	rpc_put_task(task);
7037 out:
7038 	request->fl_flags = fl_flags;
7039 	trace_nfs4_unlock(request, state, F_SETLK, status);
7040 	return status;
7041 }
7042 
7043 struct nfs4_lockdata {
7044 	struct nfs_lock_args arg;
7045 	struct nfs_lock_res res;
7046 	struct nfs4_lock_state *lsp;
7047 	struct nfs_open_context *ctx;
7048 	struct file_lock fl;
7049 	unsigned long timestamp;
7050 	int rpc_status;
7051 	int cancelled;
7052 	struct nfs_server *server;
7053 };
7054 
7055 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7056 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7057 		gfp_t gfp_mask)
7058 {
7059 	struct nfs4_lockdata *p;
7060 	struct inode *inode = lsp->ls_state->inode;
7061 	struct nfs_server *server = NFS_SERVER(inode);
7062 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7063 
7064 	p = kzalloc(sizeof(*p), gfp_mask);
7065 	if (p == NULL)
7066 		return NULL;
7067 
7068 	p->arg.fh = NFS_FH(inode);
7069 	p->arg.fl = &p->fl;
7070 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7071 	if (IS_ERR(p->arg.open_seqid))
7072 		goto out_free;
7073 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7074 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7075 	if (IS_ERR(p->arg.lock_seqid))
7076 		goto out_free_seqid;
7077 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7078 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7079 	p->arg.lock_owner.s_dev = server->s_dev;
7080 	p->res.lock_seqid = p->arg.lock_seqid;
7081 	p->lsp = lsp;
7082 	p->server = server;
7083 	p->ctx = get_nfs_open_context(ctx);
7084 	locks_init_lock(&p->fl);
7085 	locks_copy_lock(&p->fl, fl);
7086 	return p;
7087 out_free_seqid:
7088 	nfs_free_seqid(p->arg.open_seqid);
7089 out_free:
7090 	kfree(p);
7091 	return NULL;
7092 }
7093 
7094 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7095 {
7096 	struct nfs4_lockdata *data = calldata;
7097 	struct nfs4_state *state = data->lsp->ls_state;
7098 
7099 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7100 		goto out_wait;
7101 	/* Do we need to do an open_to_lock_owner? */
7102 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7103 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7104 			goto out_release_lock_seqid;
7105 		}
7106 		nfs4_stateid_copy(&data->arg.open_stateid,
7107 				&state->open_stateid);
7108 		data->arg.new_lock_owner = 1;
7109 		data->res.open_seqid = data->arg.open_seqid;
7110 	} else {
7111 		data->arg.new_lock_owner = 0;
7112 		nfs4_stateid_copy(&data->arg.lock_stateid,
7113 				&data->lsp->ls_stateid);
7114 	}
7115 	if (!nfs4_valid_open_stateid(state)) {
7116 		data->rpc_status = -EBADF;
7117 		task->tk_action = NULL;
7118 		goto out_release_open_seqid;
7119 	}
7120 	data->timestamp = jiffies;
7121 	if (nfs4_setup_sequence(data->server->nfs_client,
7122 				&data->arg.seq_args,
7123 				&data->res.seq_res,
7124 				task) == 0)
7125 		return;
7126 out_release_open_seqid:
7127 	nfs_release_seqid(data->arg.open_seqid);
7128 out_release_lock_seqid:
7129 	nfs_release_seqid(data->arg.lock_seqid);
7130 out_wait:
7131 	nfs4_sequence_done(task, &data->res.seq_res);
7132 	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7133 }
7134 
7135 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7136 {
7137 	struct nfs4_lockdata *data = calldata;
7138 	struct nfs4_lock_state *lsp = data->lsp;
7139 
7140 	if (!nfs4_sequence_done(task, &data->res.seq_res))
7141 		return;
7142 
7143 	data->rpc_status = task->tk_status;
7144 	switch (task->tk_status) {
7145 	case 0:
7146 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7147 				data->timestamp);
7148 		if (data->arg.new_lock && !data->cancelled) {
7149 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7150 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7151 				goto out_restart;
7152 		}
7153 		if (data->arg.new_lock_owner != 0) {
7154 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7155 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7156 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7157 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7158 			goto out_restart;
7159 		break;
7160 	case -NFS4ERR_BAD_STATEID:
7161 	case -NFS4ERR_OLD_STATEID:
7162 	case -NFS4ERR_STALE_STATEID:
7163 	case -NFS4ERR_EXPIRED:
7164 		if (data->arg.new_lock_owner != 0) {
7165 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7166 						&lsp->ls_state->open_stateid))
7167 				goto out_restart;
7168 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7169 						&lsp->ls_stateid))
7170 				goto out_restart;
7171 	}
7172 out_done:
7173 	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7174 	return;
7175 out_restart:
7176 	if (!data->cancelled)
7177 		rpc_restart_call_prepare(task);
7178 	goto out_done;
7179 }
7180 
7181 static void nfs4_lock_release(void *calldata)
7182 {
7183 	struct nfs4_lockdata *data = calldata;
7184 
7185 	nfs_free_seqid(data->arg.open_seqid);
7186 	if (data->cancelled && data->rpc_status == 0) {
7187 		struct rpc_task *task;
7188 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7189 				data->arg.lock_seqid);
7190 		if (!IS_ERR(task))
7191 			rpc_put_task_async(task);
7192 		dprintk("%s: cancelling lock!\n", __func__);
7193 	} else
7194 		nfs_free_seqid(data->arg.lock_seqid);
7195 	nfs4_put_lock_state(data->lsp);
7196 	put_nfs_open_context(data->ctx);
7197 	kfree(data);
7198 }
7199 
7200 static const struct rpc_call_ops nfs4_lock_ops = {
7201 	.rpc_call_prepare = nfs4_lock_prepare,
7202 	.rpc_call_done = nfs4_lock_done,
7203 	.rpc_release = nfs4_lock_release,
7204 };
7205 
7206 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7207 {
7208 	switch (error) {
7209 	case -NFS4ERR_ADMIN_REVOKED:
7210 	case -NFS4ERR_EXPIRED:
7211 	case -NFS4ERR_BAD_STATEID:
7212 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7213 		if (new_lock_owner != 0 ||
7214 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7215 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7216 		break;
7217 	case -NFS4ERR_STALE_STATEID:
7218 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7219 		nfs4_schedule_lease_recovery(server->nfs_client);
7220 	}
7221 }
7222 
7223 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7224 {
7225 	struct nfs4_lockdata *data;
7226 	struct rpc_task *task;
7227 	struct rpc_message msg = {
7228 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7229 		.rpc_cred = state->owner->so_cred,
7230 	};
7231 	struct rpc_task_setup task_setup_data = {
7232 		.rpc_client = NFS_CLIENT(state->inode),
7233 		.rpc_message = &msg,
7234 		.callback_ops = &nfs4_lock_ops,
7235 		.workqueue = nfsiod_workqueue,
7236 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7237 	};
7238 	int ret;
7239 
7240 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7241 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7242 
7243 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7244 				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7245 	if (data == NULL)
7246 		return -ENOMEM;
7247 	if (IS_SETLKW(cmd))
7248 		data->arg.block = 1;
7249 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7250 				recovery_type > NFS_LOCK_NEW);
7251 	msg.rpc_argp = &data->arg;
7252 	msg.rpc_resp = &data->res;
7253 	task_setup_data.callback_data = data;
7254 	if (recovery_type > NFS_LOCK_NEW) {
7255 		if (recovery_type == NFS_LOCK_RECLAIM)
7256 			data->arg.reclaim = NFS_LOCK_RECLAIM;
7257 	} else
7258 		data->arg.new_lock = 1;
7259 	task = rpc_run_task(&task_setup_data);
7260 	if (IS_ERR(task))
7261 		return PTR_ERR(task);
7262 	ret = rpc_wait_for_completion_task(task);
7263 	if (ret == 0) {
7264 		ret = data->rpc_status;
7265 		if (ret)
7266 			nfs4_handle_setlk_error(data->server, data->lsp,
7267 					data->arg.new_lock_owner, ret);
7268 	} else
7269 		data->cancelled = true;
7270 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7271 	rpc_put_task(task);
7272 	dprintk("%s: ret = %d\n", __func__, ret);
7273 	return ret;
7274 }
7275 
7276 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7277 {
7278 	struct nfs_server *server = NFS_SERVER(state->inode);
7279 	struct nfs4_exception exception = {
7280 		.inode = state->inode,
7281 	};
7282 	int err;
7283 
7284 	do {
7285 		/* Cache the lock if possible... */
7286 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7287 			return 0;
7288 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7289 		if (err != -NFS4ERR_DELAY)
7290 			break;
7291 		nfs4_handle_exception(server, err, &exception);
7292 	} while (exception.retry);
7293 	return err;
7294 }
7295 
7296 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7297 {
7298 	struct nfs_server *server = NFS_SERVER(state->inode);
7299 	struct nfs4_exception exception = {
7300 		.inode = state->inode,
7301 	};
7302 	int err;
7303 
7304 	err = nfs4_set_lock_state(state, request);
7305 	if (err != 0)
7306 		return err;
7307 	if (!recover_lost_locks) {
7308 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7309 		return 0;
7310 	}
7311 	do {
7312 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7313 			return 0;
7314 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7315 		switch (err) {
7316 		default:
7317 			goto out;
7318 		case -NFS4ERR_GRACE:
7319 		case -NFS4ERR_DELAY:
7320 			nfs4_handle_exception(server, err, &exception);
7321 			err = 0;
7322 		}
7323 	} while (exception.retry);
7324 out:
7325 	return err;
7326 }
7327 
7328 #if defined(CONFIG_NFS_V4_1)
7329 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7330 {
7331 	struct nfs4_lock_state *lsp;
7332 	int status;
7333 
7334 	status = nfs4_set_lock_state(state, request);
7335 	if (status != 0)
7336 		return status;
7337 	lsp = request->fl_u.nfs4_fl.owner;
7338 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7339 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7340 		return 0;
7341 	return nfs4_lock_expired(state, request);
7342 }
7343 #endif
7344 
7345 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7346 {
7347 	struct nfs_inode *nfsi = NFS_I(state->inode);
7348 	struct nfs4_state_owner *sp = state->owner;
7349 	unsigned char fl_flags = request->fl_flags;
7350 	int status;
7351 
7352 	request->fl_flags |= FL_ACCESS;
7353 	status = locks_lock_inode_wait(state->inode, request);
7354 	if (status < 0)
7355 		goto out;
7356 	mutex_lock(&sp->so_delegreturn_mutex);
7357 	down_read(&nfsi->rwsem);
7358 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7359 		/* Yes: cache locks! */
7360 		/* ...but avoid races with delegation recall... */
7361 		request->fl_flags = fl_flags & ~FL_SLEEP;
7362 		status = locks_lock_inode_wait(state->inode, request);
7363 		up_read(&nfsi->rwsem);
7364 		mutex_unlock(&sp->so_delegreturn_mutex);
7365 		goto out;
7366 	}
7367 	up_read(&nfsi->rwsem);
7368 	mutex_unlock(&sp->so_delegreturn_mutex);
7369 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7370 out:
7371 	request->fl_flags = fl_flags;
7372 	return status;
7373 }
7374 
7375 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7376 {
7377 	struct nfs4_exception exception = {
7378 		.state = state,
7379 		.inode = state->inode,
7380 		.interruptible = true,
7381 	};
7382 	int err;
7383 
7384 	do {
7385 		err = _nfs4_proc_setlk(state, cmd, request);
7386 		if (err == -NFS4ERR_DENIED)
7387 			err = -EAGAIN;
7388 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7389 				err, &exception);
7390 	} while (exception.retry);
7391 	return err;
7392 }
7393 
7394 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7395 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7396 
7397 static int
7398 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7399 			struct file_lock *request)
7400 {
7401 	int		status = -ERESTARTSYS;
7402 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7403 
7404 	while(!signalled()) {
7405 		status = nfs4_proc_setlk(state, cmd, request);
7406 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7407 			break;
7408 		freezable_schedule_timeout_interruptible(timeout);
7409 		timeout *= 2;
7410 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7411 		status = -ERESTARTSYS;
7412 	}
7413 	return status;
7414 }
7415 
7416 #ifdef CONFIG_NFS_V4_1
7417 struct nfs4_lock_waiter {
7418 	struct inode		*inode;
7419 	struct nfs_lowner	owner;
7420 	wait_queue_entry_t	wait;
7421 };
7422 
7423 static int
7424 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7425 {
7426 	struct nfs4_lock_waiter	*waiter	=
7427 		container_of(wait, struct nfs4_lock_waiter, wait);
7428 
7429 	/* NULL key means to wake up everyone */
7430 	if (key) {
7431 		struct cb_notify_lock_args	*cbnl = key;
7432 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7433 						*wowner = &waiter->owner;
7434 
7435 		/* Only wake if the callback was for the same owner. */
7436 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7437 			return 0;
7438 
7439 		/* Make sure it's for the right inode */
7440 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7441 			return 0;
7442 	}
7443 
7444 	return woken_wake_function(wait, mode, flags, key);
7445 }
7446 
7447 static int
7448 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7449 {
7450 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7451 	struct nfs_server *server = NFS_SERVER(state->inode);
7452 	struct nfs_client *clp = server->nfs_client;
7453 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7454 	struct nfs4_lock_waiter waiter = {
7455 		.inode = state->inode,
7456 		.owner = { .clientid = clp->cl_clientid,
7457 			   .id = lsp->ls_seqid.owner_id,
7458 			   .s_dev = server->s_dev },
7459 	};
7460 	int status;
7461 
7462 	/* Don't bother with waitqueue if we don't expect a callback */
7463 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7464 		return nfs4_retry_setlk_simple(state, cmd, request);
7465 
7466 	init_wait(&waiter.wait);
7467 	waiter.wait.func = nfs4_wake_lock_waiter;
7468 	add_wait_queue(q, &waiter.wait);
7469 
7470 	do {
7471 		status = nfs4_proc_setlk(state, cmd, request);
7472 		if (status != -EAGAIN || IS_SETLK(cmd))
7473 			break;
7474 
7475 		status = -ERESTARTSYS;
7476 		freezer_do_not_count();
7477 		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7478 			   NFS4_LOCK_MAXTIMEOUT);
7479 		freezer_count();
7480 	} while (!signalled());
7481 
7482 	remove_wait_queue(q, &waiter.wait);
7483 
7484 	return status;
7485 }
7486 #else /* !CONFIG_NFS_V4_1 */
7487 static inline int
7488 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7489 {
7490 	return nfs4_retry_setlk_simple(state, cmd, request);
7491 }
7492 #endif
7493 
7494 static int
7495 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7496 {
7497 	struct nfs_open_context *ctx;
7498 	struct nfs4_state *state;
7499 	int status;
7500 
7501 	/* verify open state */
7502 	ctx = nfs_file_open_context(filp);
7503 	state = ctx->state;
7504 
7505 	if (IS_GETLK(cmd)) {
7506 		if (state != NULL)
7507 			return nfs4_proc_getlk(state, F_GETLK, request);
7508 		return 0;
7509 	}
7510 
7511 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7512 		return -EINVAL;
7513 
7514 	if (request->fl_type == F_UNLCK) {
7515 		if (state != NULL)
7516 			return nfs4_proc_unlck(state, cmd, request);
7517 		return 0;
7518 	}
7519 
7520 	if (state == NULL)
7521 		return -ENOLCK;
7522 
7523 	if ((request->fl_flags & FL_POSIX) &&
7524 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7525 		return -ENOLCK;
7526 
7527 	/*
7528 	 * Don't rely on the VFS having checked the file open mode,
7529 	 * since it won't do this for flock() locks.
7530 	 */
7531 	switch (request->fl_type) {
7532 	case F_RDLCK:
7533 		if (!(filp->f_mode & FMODE_READ))
7534 			return -EBADF;
7535 		break;
7536 	case F_WRLCK:
7537 		if (!(filp->f_mode & FMODE_WRITE))
7538 			return -EBADF;
7539 	}
7540 
7541 	status = nfs4_set_lock_state(state, request);
7542 	if (status != 0)
7543 		return status;
7544 
7545 	return nfs4_retry_setlk(state, cmd, request);
7546 }
7547 
7548 static int nfs4_delete_lease(struct file *file, void **priv)
7549 {
7550 	return generic_setlease(file, F_UNLCK, NULL, priv);
7551 }
7552 
7553 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7554 			  void **priv)
7555 {
7556 	struct inode *inode = file_inode(file);
7557 	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7558 	int ret;
7559 
7560 	/* No delegation, no lease */
7561 	if (!nfs4_have_delegation(inode, type))
7562 		return -EAGAIN;
7563 	ret = generic_setlease(file, arg, lease, priv);
7564 	if (ret || nfs4_have_delegation(inode, type))
7565 		return ret;
7566 	/* We raced with a delegation return */
7567 	nfs4_delete_lease(file, priv);
7568 	return -EAGAIN;
7569 }
7570 
7571 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7572 		       void **priv)
7573 {
7574 	switch (arg) {
7575 	case F_RDLCK:
7576 	case F_WRLCK:
7577 		return nfs4_add_lease(file, arg, lease, priv);
7578 	case F_UNLCK:
7579 		return nfs4_delete_lease(file, priv);
7580 	default:
7581 		return -EINVAL;
7582 	}
7583 }
7584 
7585 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7586 {
7587 	struct nfs_server *server = NFS_SERVER(state->inode);
7588 	int err;
7589 
7590 	err = nfs4_set_lock_state(state, fl);
7591 	if (err != 0)
7592 		return err;
7593 	do {
7594 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7595 		if (err != -NFS4ERR_DELAY)
7596 			break;
7597 		ssleep(1);
7598 	} while (err == -NFS4ERR_DELAY);
7599 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7600 }
7601 
7602 struct nfs_release_lockowner_data {
7603 	struct nfs4_lock_state *lsp;
7604 	struct nfs_server *server;
7605 	struct nfs_release_lockowner_args args;
7606 	struct nfs_release_lockowner_res res;
7607 	unsigned long timestamp;
7608 };
7609 
7610 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7611 {
7612 	struct nfs_release_lockowner_data *data = calldata;
7613 	struct nfs_server *server = data->server;
7614 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7615 			   &data->res.seq_res, task);
7616 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7617 	data->timestamp = jiffies;
7618 }
7619 
7620 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7621 {
7622 	struct nfs_release_lockowner_data *data = calldata;
7623 	struct nfs_server *server = data->server;
7624 
7625 	nfs40_sequence_done(task, &data->res.seq_res);
7626 
7627 	switch (task->tk_status) {
7628 	case 0:
7629 		renew_lease(server, data->timestamp);
7630 		break;
7631 	case -NFS4ERR_STALE_CLIENTID:
7632 	case -NFS4ERR_EXPIRED:
7633 		nfs4_schedule_lease_recovery(server->nfs_client);
7634 		break;
7635 	case -NFS4ERR_LEASE_MOVED:
7636 	case -NFS4ERR_DELAY:
7637 		if (nfs4_async_handle_error(task, server,
7638 					    NULL, NULL) == -EAGAIN)
7639 			rpc_restart_call_prepare(task);
7640 	}
7641 }
7642 
7643 static void nfs4_release_lockowner_release(void *calldata)
7644 {
7645 	struct nfs_release_lockowner_data *data = calldata;
7646 	nfs4_free_lock_state(data->server, data->lsp);
7647 	kfree(calldata);
7648 }
7649 
7650 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7651 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7652 	.rpc_call_done = nfs4_release_lockowner_done,
7653 	.rpc_release = nfs4_release_lockowner_release,
7654 };
7655 
7656 static void
7657 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7658 {
7659 	struct nfs_release_lockowner_data *data;
7660 	struct rpc_message msg = {
7661 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7662 	};
7663 
7664 	if (server->nfs_client->cl_mvops->minor_version != 0)
7665 		return;
7666 
7667 	data = kmalloc(sizeof(*data), GFP_KERNEL);
7668 	if (!data)
7669 		return;
7670 	data->lsp = lsp;
7671 	data->server = server;
7672 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7673 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7674 	data->args.lock_owner.s_dev = server->s_dev;
7675 
7676 	msg.rpc_argp = &data->args;
7677 	msg.rpc_resp = &data->res;
7678 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7679 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7680 }
7681 
7682 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7683 
7684 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7685 				   struct user_namespace *mnt_userns,
7686 				   struct dentry *unused, struct inode *inode,
7687 				   const char *key, const void *buf,
7688 				   size_t buflen, int flags)
7689 {
7690 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7691 }
7692 
7693 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7694 				   struct dentry *unused, struct inode *inode,
7695 				   const char *key, void *buf, size_t buflen)
7696 {
7697 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7698 }
7699 
7700 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7701 {
7702 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7703 }
7704 
7705 #if defined(CONFIG_NFS_V4_1)
7706 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7707 
7708 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7709 				    struct user_namespace *mnt_userns,
7710 				    struct dentry *unused, struct inode *inode,
7711 				    const char *key, const void *buf,
7712 				    size_t buflen, int flags)
7713 {
7714 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7715 }
7716 
7717 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7718 				    struct dentry *unused, struct inode *inode,
7719 				    const char *key, void *buf, size_t buflen)
7720 {
7721 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7722 }
7723 
7724 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7725 {
7726 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7727 }
7728 
7729 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7730 
7731 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7732 				    struct user_namespace *mnt_userns,
7733 				    struct dentry *unused, struct inode *inode,
7734 				    const char *key, const void *buf,
7735 				    size_t buflen, int flags)
7736 {
7737 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7738 }
7739 
7740 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7741 				    struct dentry *unused, struct inode *inode,
7742 				    const char *key, void *buf, size_t buflen)
7743 {
7744 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7745 }
7746 
7747 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7748 {
7749 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7750 }
7751 
7752 #endif
7753 
7754 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7755 
7756 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7757 				     struct user_namespace *mnt_userns,
7758 				     struct dentry *unused, struct inode *inode,
7759 				     const char *key, const void *buf,
7760 				     size_t buflen, int flags)
7761 {
7762 	if (security_ismaclabel(key))
7763 		return nfs4_set_security_label(inode, buf, buflen);
7764 
7765 	return -EOPNOTSUPP;
7766 }
7767 
7768 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7769 				     struct dentry *unused, struct inode *inode,
7770 				     const char *key, void *buf, size_t buflen)
7771 {
7772 	if (security_ismaclabel(key))
7773 		return nfs4_get_security_label(inode, buf, buflen);
7774 	return -EOPNOTSUPP;
7775 }
7776 
7777 static ssize_t
7778 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7779 {
7780 	int len = 0;
7781 
7782 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7783 		len = security_inode_listsecurity(inode, list, list_len);
7784 		if (len >= 0 && list_len && len > list_len)
7785 			return -ERANGE;
7786 	}
7787 	return len;
7788 }
7789 
7790 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7791 	.prefix = XATTR_SECURITY_PREFIX,
7792 	.get	= nfs4_xattr_get_nfs4_label,
7793 	.set	= nfs4_xattr_set_nfs4_label,
7794 };
7795 
7796 #else
7797 
7798 static ssize_t
7799 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7800 {
7801 	return 0;
7802 }
7803 
7804 #endif
7805 
7806 #ifdef CONFIG_NFS_V4_2
7807 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7808 				    struct user_namespace *mnt_userns,
7809 				    struct dentry *unused, struct inode *inode,
7810 				    const char *key, const void *buf,
7811 				    size_t buflen, int flags)
7812 {
7813 	u32 mask;
7814 	int ret;
7815 
7816 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7817 		return -EOPNOTSUPP;
7818 
7819 	/*
7820 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7821 	 * flags right now. Handling of xattr operations use the normal
7822 	 * file read/write permissions.
7823 	 *
7824 	 * Just in case the server has other ideas (which RFC 8276 allows),
7825 	 * do a cached access check for the XA* flags to possibly avoid
7826 	 * doing an RPC and getting EACCES back.
7827 	 */
7828 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7829 		if (!(mask & NFS_ACCESS_XAWRITE))
7830 			return -EACCES;
7831 	}
7832 
7833 	if (buf == NULL) {
7834 		ret = nfs42_proc_removexattr(inode, key);
7835 		if (!ret)
7836 			nfs4_xattr_cache_remove(inode, key);
7837 	} else {
7838 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7839 		if (!ret)
7840 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7841 	}
7842 
7843 	return ret;
7844 }
7845 
7846 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7847 				    struct dentry *unused, struct inode *inode,
7848 				    const char *key, void *buf, size_t buflen)
7849 {
7850 	u32 mask;
7851 	ssize_t ret;
7852 
7853 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7854 		return -EOPNOTSUPP;
7855 
7856 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7857 		if (!(mask & NFS_ACCESS_XAREAD))
7858 			return -EACCES;
7859 	}
7860 
7861 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7862 	if (ret)
7863 		return ret;
7864 
7865 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7866 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7867 		return ret;
7868 
7869 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7870 
7871 	return ret;
7872 }
7873 
7874 static ssize_t
7875 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7876 {
7877 	u64 cookie;
7878 	bool eof;
7879 	ssize_t ret, size;
7880 	char *buf;
7881 	size_t buflen;
7882 	u32 mask;
7883 
7884 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7885 		return 0;
7886 
7887 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7888 		if (!(mask & NFS_ACCESS_XALIST))
7889 			return 0;
7890 	}
7891 
7892 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7893 	if (ret)
7894 		return ret;
7895 
7896 	ret = nfs4_xattr_cache_list(inode, list, list_len);
7897 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7898 		return ret;
7899 
7900 	cookie = 0;
7901 	eof = false;
7902 	buflen = list_len ? list_len : XATTR_LIST_MAX;
7903 	buf = list_len ? list : NULL;
7904 	size = 0;
7905 
7906 	while (!eof) {
7907 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
7908 		    &cookie, &eof);
7909 		if (ret < 0)
7910 			return ret;
7911 
7912 		if (list_len) {
7913 			buf += ret;
7914 			buflen -= ret;
7915 		}
7916 		size += ret;
7917 	}
7918 
7919 	if (list_len)
7920 		nfs4_xattr_cache_set_list(inode, list, size);
7921 
7922 	return size;
7923 }
7924 
7925 #else
7926 
7927 static ssize_t
7928 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7929 {
7930 	return 0;
7931 }
7932 #endif /* CONFIG_NFS_V4_2 */
7933 
7934 /*
7935  * nfs_fhget will use either the mounted_on_fileid or the fileid
7936  */
7937 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7938 {
7939 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7940 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7941 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7942 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7943 		return;
7944 
7945 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7946 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7947 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7948 	fattr->nlink = 2;
7949 }
7950 
7951 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7952 				   const struct qstr *name,
7953 				   struct nfs4_fs_locations *fs_locations,
7954 				   struct page *page)
7955 {
7956 	struct nfs_server *server = NFS_SERVER(dir);
7957 	u32 bitmask[3];
7958 	struct nfs4_fs_locations_arg args = {
7959 		.dir_fh = NFS_FH(dir),
7960 		.name = name,
7961 		.page = page,
7962 		.bitmask = bitmask,
7963 	};
7964 	struct nfs4_fs_locations_res res = {
7965 		.fs_locations = fs_locations,
7966 	};
7967 	struct rpc_message msg = {
7968 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7969 		.rpc_argp = &args,
7970 		.rpc_resp = &res,
7971 	};
7972 	int status;
7973 
7974 	dprintk("%s: start\n", __func__);
7975 
7976 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7977 	bitmask[1] = nfs4_fattr_bitmap[1];
7978 
7979 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
7980 	 * is not supported */
7981 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7982 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
7983 	else
7984 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7985 
7986 	nfs_fattr_init(fs_locations->fattr);
7987 	fs_locations->server = server;
7988 	fs_locations->nlocations = 0;
7989 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7990 	dprintk("%s: returned status = %d\n", __func__, status);
7991 	return status;
7992 }
7993 
7994 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7995 			   const struct qstr *name,
7996 			   struct nfs4_fs_locations *fs_locations,
7997 			   struct page *page)
7998 {
7999 	struct nfs4_exception exception = {
8000 		.interruptible = true,
8001 	};
8002 	int err;
8003 	do {
8004 		err = _nfs4_proc_fs_locations(client, dir, name,
8005 				fs_locations, page);
8006 		trace_nfs4_get_fs_locations(dir, name, err);
8007 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8008 				&exception);
8009 	} while (exception.retry);
8010 	return err;
8011 }
8012 
8013 /*
8014  * This operation also signals the server that this client is
8015  * performing migration recovery.  The server can stop returning
8016  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8017  * appended to this compound to identify the client ID which is
8018  * performing recovery.
8019  */
8020 static int _nfs40_proc_get_locations(struct nfs_server *server,
8021 				     struct nfs_fh *fhandle,
8022 				     struct nfs4_fs_locations *locations,
8023 				     struct page *page, const struct cred *cred)
8024 {
8025 	struct rpc_clnt *clnt = server->client;
8026 	u32 bitmask[2] = {
8027 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8028 	};
8029 	struct nfs4_fs_locations_arg args = {
8030 		.clientid	= server->nfs_client->cl_clientid,
8031 		.fh		= fhandle,
8032 		.page		= page,
8033 		.bitmask	= bitmask,
8034 		.migration	= 1,		/* skip LOOKUP */
8035 		.renew		= 1,		/* append RENEW */
8036 	};
8037 	struct nfs4_fs_locations_res res = {
8038 		.fs_locations	= locations,
8039 		.migration	= 1,
8040 		.renew		= 1,
8041 	};
8042 	struct rpc_message msg = {
8043 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8044 		.rpc_argp	= &args,
8045 		.rpc_resp	= &res,
8046 		.rpc_cred	= cred,
8047 	};
8048 	unsigned long now = jiffies;
8049 	int status;
8050 
8051 	nfs_fattr_init(locations->fattr);
8052 	locations->server = server;
8053 	locations->nlocations = 0;
8054 
8055 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8056 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8057 					&args.seq_args, &res.seq_res);
8058 	if (status)
8059 		return status;
8060 
8061 	renew_lease(server, now);
8062 	return 0;
8063 }
8064 
8065 #ifdef CONFIG_NFS_V4_1
8066 
8067 /*
8068  * This operation also signals the server that this client is
8069  * performing migration recovery.  The server can stop asserting
8070  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8071  * performing this operation is identified in the SEQUENCE
8072  * operation in this compound.
8073  *
8074  * When the client supports GETATTR(fs_locations_info), it can
8075  * be plumbed in here.
8076  */
8077 static int _nfs41_proc_get_locations(struct nfs_server *server,
8078 				     struct nfs_fh *fhandle,
8079 				     struct nfs4_fs_locations *locations,
8080 				     struct page *page, const struct cred *cred)
8081 {
8082 	struct rpc_clnt *clnt = server->client;
8083 	u32 bitmask[2] = {
8084 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8085 	};
8086 	struct nfs4_fs_locations_arg args = {
8087 		.fh		= fhandle,
8088 		.page		= page,
8089 		.bitmask	= bitmask,
8090 		.migration	= 1,		/* skip LOOKUP */
8091 	};
8092 	struct nfs4_fs_locations_res res = {
8093 		.fs_locations	= locations,
8094 		.migration	= 1,
8095 	};
8096 	struct rpc_message msg = {
8097 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8098 		.rpc_argp	= &args,
8099 		.rpc_resp	= &res,
8100 		.rpc_cred	= cred,
8101 	};
8102 	struct nfs4_call_sync_data data = {
8103 		.seq_server = server,
8104 		.seq_args = &args.seq_args,
8105 		.seq_res = &res.seq_res,
8106 	};
8107 	struct rpc_task_setup task_setup_data = {
8108 		.rpc_client = clnt,
8109 		.rpc_message = &msg,
8110 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8111 		.callback_data = &data,
8112 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8113 	};
8114 	int status;
8115 
8116 	nfs_fattr_init(locations->fattr);
8117 	locations->server = server;
8118 	locations->nlocations = 0;
8119 
8120 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8121 	status = nfs4_call_sync_custom(&task_setup_data);
8122 	if (status == NFS4_OK &&
8123 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8124 		status = -NFS4ERR_LEASE_MOVED;
8125 	return status;
8126 }
8127 
8128 #endif	/* CONFIG_NFS_V4_1 */
8129 
8130 /**
8131  * nfs4_proc_get_locations - discover locations for a migrated FSID
8132  * @server: pointer to nfs_server to process
8133  * @fhandle: pointer to the kernel NFS client file handle
8134  * @locations: result of query
8135  * @page: buffer
8136  * @cred: credential to use for this operation
8137  *
8138  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8139  * operation failed, or a negative errno if a local error occurred.
8140  *
8141  * On success, "locations" is filled in, but if the server has
8142  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8143  * asserted.
8144  *
8145  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8146  * from this client that require migration recovery.
8147  */
8148 int nfs4_proc_get_locations(struct nfs_server *server,
8149 			    struct nfs_fh *fhandle,
8150 			    struct nfs4_fs_locations *locations,
8151 			    struct page *page, const struct cred *cred)
8152 {
8153 	struct nfs_client *clp = server->nfs_client;
8154 	const struct nfs4_mig_recovery_ops *ops =
8155 					clp->cl_mvops->mig_recovery_ops;
8156 	struct nfs4_exception exception = {
8157 		.interruptible = true,
8158 	};
8159 	int status;
8160 
8161 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8162 		(unsigned long long)server->fsid.major,
8163 		(unsigned long long)server->fsid.minor,
8164 		clp->cl_hostname);
8165 	nfs_display_fhandle(fhandle, __func__);
8166 
8167 	do {
8168 		status = ops->get_locations(server, fhandle, locations, page,
8169 					    cred);
8170 		if (status != -NFS4ERR_DELAY)
8171 			break;
8172 		nfs4_handle_exception(server, status, &exception);
8173 	} while (exception.retry);
8174 	return status;
8175 }
8176 
8177 /*
8178  * This operation also signals the server that this client is
8179  * performing "lease moved" recovery.  The server can stop
8180  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8181  * is appended to this compound to identify the client ID which is
8182  * performing recovery.
8183  */
8184 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8185 {
8186 	struct nfs_server *server = NFS_SERVER(inode);
8187 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8188 	struct rpc_clnt *clnt = server->client;
8189 	struct nfs4_fsid_present_arg args = {
8190 		.fh		= NFS_FH(inode),
8191 		.clientid	= clp->cl_clientid,
8192 		.renew		= 1,		/* append RENEW */
8193 	};
8194 	struct nfs4_fsid_present_res res = {
8195 		.renew		= 1,
8196 	};
8197 	struct rpc_message msg = {
8198 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8199 		.rpc_argp	= &args,
8200 		.rpc_resp	= &res,
8201 		.rpc_cred	= cred,
8202 	};
8203 	unsigned long now = jiffies;
8204 	int status;
8205 
8206 	res.fh = nfs_alloc_fhandle();
8207 	if (res.fh == NULL)
8208 		return -ENOMEM;
8209 
8210 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8211 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8212 						&args.seq_args, &res.seq_res);
8213 	nfs_free_fhandle(res.fh);
8214 	if (status)
8215 		return status;
8216 
8217 	do_renew_lease(clp, now);
8218 	return 0;
8219 }
8220 
8221 #ifdef CONFIG_NFS_V4_1
8222 
8223 /*
8224  * This operation also signals the server that this client is
8225  * performing "lease moved" recovery.  The server can stop asserting
8226  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8227  * this operation is identified in the SEQUENCE operation in this
8228  * compound.
8229  */
8230 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8231 {
8232 	struct nfs_server *server = NFS_SERVER(inode);
8233 	struct rpc_clnt *clnt = server->client;
8234 	struct nfs4_fsid_present_arg args = {
8235 		.fh		= NFS_FH(inode),
8236 	};
8237 	struct nfs4_fsid_present_res res = {
8238 	};
8239 	struct rpc_message msg = {
8240 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8241 		.rpc_argp	= &args,
8242 		.rpc_resp	= &res,
8243 		.rpc_cred	= cred,
8244 	};
8245 	int status;
8246 
8247 	res.fh = nfs_alloc_fhandle();
8248 	if (res.fh == NULL)
8249 		return -ENOMEM;
8250 
8251 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8252 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8253 						&args.seq_args, &res.seq_res);
8254 	nfs_free_fhandle(res.fh);
8255 	if (status == NFS4_OK &&
8256 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8257 		status = -NFS4ERR_LEASE_MOVED;
8258 	return status;
8259 }
8260 
8261 #endif	/* CONFIG_NFS_V4_1 */
8262 
8263 /**
8264  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8265  * @inode: inode on FSID to check
8266  * @cred: credential to use for this operation
8267  *
8268  * Server indicates whether the FSID is present, moved, or not
8269  * recognized.  This operation is necessary to clear a LEASE_MOVED
8270  * condition for this client ID.
8271  *
8272  * Returns NFS4_OK if the FSID is present on this server,
8273  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8274  *  NFS4ERR code if some error occurred on the server, or a
8275  *  negative errno if a local failure occurred.
8276  */
8277 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8278 {
8279 	struct nfs_server *server = NFS_SERVER(inode);
8280 	struct nfs_client *clp = server->nfs_client;
8281 	const struct nfs4_mig_recovery_ops *ops =
8282 					clp->cl_mvops->mig_recovery_ops;
8283 	struct nfs4_exception exception = {
8284 		.interruptible = true,
8285 	};
8286 	int status;
8287 
8288 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8289 		(unsigned long long)server->fsid.major,
8290 		(unsigned long long)server->fsid.minor,
8291 		clp->cl_hostname);
8292 	nfs_display_fhandle(NFS_FH(inode), __func__);
8293 
8294 	do {
8295 		status = ops->fsid_present(inode, cred);
8296 		if (status != -NFS4ERR_DELAY)
8297 			break;
8298 		nfs4_handle_exception(server, status, &exception);
8299 	} while (exception.retry);
8300 	return status;
8301 }
8302 
8303 /*
8304  * If 'use_integrity' is true and the state managment nfs_client
8305  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8306  * and the machine credential as per RFC3530bis and RFC5661 Security
8307  * Considerations sections. Otherwise, just use the user cred with the
8308  * filesystem's rpc_client.
8309  */
8310 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8311 {
8312 	int status;
8313 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8314 	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8315 	struct nfs4_secinfo_arg args = {
8316 		.dir_fh = NFS_FH(dir),
8317 		.name   = name,
8318 	};
8319 	struct nfs4_secinfo_res res = {
8320 		.flavors     = flavors,
8321 	};
8322 	struct rpc_message msg = {
8323 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8324 		.rpc_argp = &args,
8325 		.rpc_resp = &res,
8326 	};
8327 	struct nfs4_call_sync_data data = {
8328 		.seq_server = NFS_SERVER(dir),
8329 		.seq_args = &args.seq_args,
8330 		.seq_res = &res.seq_res,
8331 	};
8332 	struct rpc_task_setup task_setup = {
8333 		.rpc_client = clnt,
8334 		.rpc_message = &msg,
8335 		.callback_ops = clp->cl_mvops->call_sync_ops,
8336 		.callback_data = &data,
8337 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8338 	};
8339 	const struct cred *cred = NULL;
8340 
8341 	if (use_integrity) {
8342 		clnt = clp->cl_rpcclient;
8343 		task_setup.rpc_client = clnt;
8344 
8345 		cred = nfs4_get_clid_cred(clp);
8346 		msg.rpc_cred = cred;
8347 	}
8348 
8349 	dprintk("NFS call  secinfo %s\n", name->name);
8350 
8351 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8352 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8353 	status = nfs4_call_sync_custom(&task_setup);
8354 
8355 	dprintk("NFS reply  secinfo: %d\n", status);
8356 
8357 	put_cred(cred);
8358 	return status;
8359 }
8360 
8361 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8362 		      struct nfs4_secinfo_flavors *flavors)
8363 {
8364 	struct nfs4_exception exception = {
8365 		.interruptible = true,
8366 	};
8367 	int err;
8368 	do {
8369 		err = -NFS4ERR_WRONGSEC;
8370 
8371 		/* try to use integrity protection with machine cred */
8372 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8373 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8374 
8375 		/*
8376 		 * if unable to use integrity protection, or SECINFO with
8377 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8378 		 * disallowed by spec, but exists in deployed servers) use
8379 		 * the current filesystem's rpc_client and the user cred.
8380 		 */
8381 		if (err == -NFS4ERR_WRONGSEC)
8382 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8383 
8384 		trace_nfs4_secinfo(dir, name, err);
8385 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8386 				&exception);
8387 	} while (exception.retry);
8388 	return err;
8389 }
8390 
8391 #ifdef CONFIG_NFS_V4_1
8392 /*
8393  * Check the exchange flags returned by the server for invalid flags, having
8394  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8395  * DS flags set.
8396  */
8397 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8398 {
8399 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8400 		goto out_inval;
8401 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8402 		goto out_inval;
8403 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8404 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8405 		goto out_inval;
8406 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8407 		goto out_inval;
8408 	return NFS_OK;
8409 out_inval:
8410 	return -NFS4ERR_INVAL;
8411 }
8412 
8413 static bool
8414 nfs41_same_server_scope(struct nfs41_server_scope *a,
8415 			struct nfs41_server_scope *b)
8416 {
8417 	if (a->server_scope_sz != b->server_scope_sz)
8418 		return false;
8419 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8420 }
8421 
8422 static void
8423 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8424 {
8425 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8426 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8427 	struct nfs_client *clp = args->client;
8428 
8429 	switch (task->tk_status) {
8430 	case -NFS4ERR_BADSESSION:
8431 	case -NFS4ERR_DEADSESSION:
8432 		nfs4_schedule_session_recovery(clp->cl_session,
8433 				task->tk_status);
8434 		return;
8435 	}
8436 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8437 			res->dir != NFS4_CDFS4_BOTH) {
8438 		rpc_task_close_connection(task);
8439 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8440 			rpc_restart_call(task);
8441 	}
8442 }
8443 
8444 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8445 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8446 };
8447 
8448 /*
8449  * nfs4_proc_bind_one_conn_to_session()
8450  *
8451  * The 4.1 client currently uses the same TCP connection for the
8452  * fore and backchannel.
8453  */
8454 static
8455 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8456 		struct rpc_xprt *xprt,
8457 		struct nfs_client *clp,
8458 		const struct cred *cred)
8459 {
8460 	int status;
8461 	struct nfs41_bind_conn_to_session_args args = {
8462 		.client = clp,
8463 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8464 		.retries = 0,
8465 	};
8466 	struct nfs41_bind_conn_to_session_res res;
8467 	struct rpc_message msg = {
8468 		.rpc_proc =
8469 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8470 		.rpc_argp = &args,
8471 		.rpc_resp = &res,
8472 		.rpc_cred = cred,
8473 	};
8474 	struct rpc_task_setup task_setup_data = {
8475 		.rpc_client = clnt,
8476 		.rpc_xprt = xprt,
8477 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8478 		.rpc_message = &msg,
8479 		.flags = RPC_TASK_TIMEOUT,
8480 	};
8481 	struct rpc_task *task;
8482 
8483 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8484 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8485 		args.dir = NFS4_CDFC4_FORE;
8486 
8487 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8488 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8489 		args.dir = NFS4_CDFC4_FORE;
8490 
8491 	task = rpc_run_task(&task_setup_data);
8492 	if (!IS_ERR(task)) {
8493 		status = task->tk_status;
8494 		rpc_put_task(task);
8495 	} else
8496 		status = PTR_ERR(task);
8497 	trace_nfs4_bind_conn_to_session(clp, status);
8498 	if (status == 0) {
8499 		if (memcmp(res.sessionid.data,
8500 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8501 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8502 			return -EIO;
8503 		}
8504 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8505 			dprintk("NFS: %s: Unexpected direction from server\n",
8506 				__func__);
8507 			return -EIO;
8508 		}
8509 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8510 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8511 				__func__);
8512 			return -EIO;
8513 		}
8514 	}
8515 
8516 	return status;
8517 }
8518 
8519 struct rpc_bind_conn_calldata {
8520 	struct nfs_client *clp;
8521 	const struct cred *cred;
8522 };
8523 
8524 static int
8525 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8526 		struct rpc_xprt *xprt,
8527 		void *calldata)
8528 {
8529 	struct rpc_bind_conn_calldata *p = calldata;
8530 
8531 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8532 }
8533 
8534 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8535 {
8536 	struct rpc_bind_conn_calldata data = {
8537 		.clp = clp,
8538 		.cred = cred,
8539 	};
8540 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8541 			nfs4_proc_bind_conn_to_session_callback, &data);
8542 }
8543 
8544 /*
8545  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8546  * and operations we'd like to see to enable certain features in the allow map
8547  */
8548 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8549 	.how = SP4_MACH_CRED,
8550 	.enforce.u.words = {
8551 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8552 		      1 << (OP_EXCHANGE_ID - 32) |
8553 		      1 << (OP_CREATE_SESSION - 32) |
8554 		      1 << (OP_DESTROY_SESSION - 32) |
8555 		      1 << (OP_DESTROY_CLIENTID - 32)
8556 	},
8557 	.allow.u.words = {
8558 		[0] = 1 << (OP_CLOSE) |
8559 		      1 << (OP_OPEN_DOWNGRADE) |
8560 		      1 << (OP_LOCKU) |
8561 		      1 << (OP_DELEGRETURN) |
8562 		      1 << (OP_COMMIT),
8563 		[1] = 1 << (OP_SECINFO - 32) |
8564 		      1 << (OP_SECINFO_NO_NAME - 32) |
8565 		      1 << (OP_LAYOUTRETURN - 32) |
8566 		      1 << (OP_TEST_STATEID - 32) |
8567 		      1 << (OP_FREE_STATEID - 32) |
8568 		      1 << (OP_WRITE - 32)
8569 	}
8570 };
8571 
8572 /*
8573  * Select the state protection mode for client `clp' given the server results
8574  * from exchange_id in `sp'.
8575  *
8576  * Returns 0 on success, negative errno otherwise.
8577  */
8578 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8579 				 struct nfs41_state_protection *sp)
8580 {
8581 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8582 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8583 		      1 << (OP_EXCHANGE_ID - 32) |
8584 		      1 << (OP_CREATE_SESSION - 32) |
8585 		      1 << (OP_DESTROY_SESSION - 32) |
8586 		      1 << (OP_DESTROY_CLIENTID - 32)
8587 	};
8588 	unsigned long flags = 0;
8589 	unsigned int i;
8590 	int ret = 0;
8591 
8592 	if (sp->how == SP4_MACH_CRED) {
8593 		/* Print state protect result */
8594 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8595 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8596 			if (test_bit(i, sp->enforce.u.longs))
8597 				dfprintk(MOUNT, "  enforce op %d\n", i);
8598 			if (test_bit(i, sp->allow.u.longs))
8599 				dfprintk(MOUNT, "  allow op %d\n", i);
8600 		}
8601 
8602 		/* make sure nothing is on enforce list that isn't supported */
8603 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8604 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8605 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8606 				ret = -EINVAL;
8607 				goto out;
8608 			}
8609 		}
8610 
8611 		/*
8612 		 * Minimal mode - state operations are allowed to use machine
8613 		 * credential.  Note this already happens by default, so the
8614 		 * client doesn't have to do anything more than the negotiation.
8615 		 *
8616 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8617 		 *       we're already using the machine cred for exchange_id
8618 		 *       and will never use a different cred.
8619 		 */
8620 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8621 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8622 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8623 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8624 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8625 			dfprintk(MOUNT, "  minimal mode enabled\n");
8626 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8627 		} else {
8628 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8629 			ret = -EINVAL;
8630 			goto out;
8631 		}
8632 
8633 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8634 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8635 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8636 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8637 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8638 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8639 		}
8640 
8641 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8642 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8643 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8644 		}
8645 
8646 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8647 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8648 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8649 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8650 		}
8651 
8652 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8653 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8654 			dfprintk(MOUNT, "  stateid mode enabled\n");
8655 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8656 		}
8657 
8658 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8659 			dfprintk(MOUNT, "  write mode enabled\n");
8660 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8661 		}
8662 
8663 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8664 			dfprintk(MOUNT, "  commit mode enabled\n");
8665 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8666 		}
8667 	}
8668 out:
8669 	clp->cl_sp4_flags = flags;
8670 	return ret;
8671 }
8672 
8673 struct nfs41_exchange_id_data {
8674 	struct nfs41_exchange_id_res res;
8675 	struct nfs41_exchange_id_args args;
8676 };
8677 
8678 static void nfs4_exchange_id_release(void *data)
8679 {
8680 	struct nfs41_exchange_id_data *cdata =
8681 					(struct nfs41_exchange_id_data *)data;
8682 
8683 	nfs_put_client(cdata->args.client);
8684 	kfree(cdata->res.impl_id);
8685 	kfree(cdata->res.server_scope);
8686 	kfree(cdata->res.server_owner);
8687 	kfree(cdata);
8688 }
8689 
8690 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8691 	.rpc_release = nfs4_exchange_id_release,
8692 };
8693 
8694 /*
8695  * _nfs4_proc_exchange_id()
8696  *
8697  * Wrapper for EXCHANGE_ID operation.
8698  */
8699 static struct rpc_task *
8700 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8701 			u32 sp4_how, struct rpc_xprt *xprt)
8702 {
8703 	struct rpc_message msg = {
8704 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8705 		.rpc_cred = cred,
8706 	};
8707 	struct rpc_task_setup task_setup_data = {
8708 		.rpc_client = clp->cl_rpcclient,
8709 		.callback_ops = &nfs4_exchange_id_call_ops,
8710 		.rpc_message = &msg,
8711 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8712 	};
8713 	struct nfs41_exchange_id_data *calldata;
8714 	int status;
8715 
8716 	if (!refcount_inc_not_zero(&clp->cl_count))
8717 		return ERR_PTR(-EIO);
8718 
8719 	status = -ENOMEM;
8720 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8721 	if (!calldata)
8722 		goto out;
8723 
8724 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8725 
8726 	status = nfs4_init_uniform_client_string(clp);
8727 	if (status)
8728 		goto out_calldata;
8729 
8730 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8731 						GFP_NOFS);
8732 	status = -ENOMEM;
8733 	if (unlikely(calldata->res.server_owner == NULL))
8734 		goto out_calldata;
8735 
8736 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8737 					GFP_NOFS);
8738 	if (unlikely(calldata->res.server_scope == NULL))
8739 		goto out_server_owner;
8740 
8741 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8742 	if (unlikely(calldata->res.impl_id == NULL))
8743 		goto out_server_scope;
8744 
8745 	switch (sp4_how) {
8746 	case SP4_NONE:
8747 		calldata->args.state_protect.how = SP4_NONE;
8748 		break;
8749 
8750 	case SP4_MACH_CRED:
8751 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8752 		break;
8753 
8754 	default:
8755 		/* unsupported! */
8756 		WARN_ON_ONCE(1);
8757 		status = -EINVAL;
8758 		goto out_impl_id;
8759 	}
8760 	if (xprt) {
8761 		task_setup_data.rpc_xprt = xprt;
8762 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8763 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8764 				sizeof(calldata->args.verifier.data));
8765 	}
8766 	calldata->args.client = clp;
8767 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8768 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8769 #ifdef CONFIG_NFS_V4_1_MIGRATION
8770 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8771 #endif
8772 	msg.rpc_argp = &calldata->args;
8773 	msg.rpc_resp = &calldata->res;
8774 	task_setup_data.callback_data = calldata;
8775 
8776 	return rpc_run_task(&task_setup_data);
8777 
8778 out_impl_id:
8779 	kfree(calldata->res.impl_id);
8780 out_server_scope:
8781 	kfree(calldata->res.server_scope);
8782 out_server_owner:
8783 	kfree(calldata->res.server_owner);
8784 out_calldata:
8785 	kfree(calldata);
8786 out:
8787 	nfs_put_client(clp);
8788 	return ERR_PTR(status);
8789 }
8790 
8791 /*
8792  * _nfs4_proc_exchange_id()
8793  *
8794  * Wrapper for EXCHANGE_ID operation.
8795  */
8796 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8797 			u32 sp4_how)
8798 {
8799 	struct rpc_task *task;
8800 	struct nfs41_exchange_id_args *argp;
8801 	struct nfs41_exchange_id_res *resp;
8802 	unsigned long now = jiffies;
8803 	int status;
8804 
8805 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8806 	if (IS_ERR(task))
8807 		return PTR_ERR(task);
8808 
8809 	argp = task->tk_msg.rpc_argp;
8810 	resp = task->tk_msg.rpc_resp;
8811 	status = task->tk_status;
8812 	if (status  != 0)
8813 		goto out;
8814 
8815 	status = nfs4_check_cl_exchange_flags(resp->flags,
8816 			clp->cl_mvops->minor_version);
8817 	if (status  != 0)
8818 		goto out;
8819 
8820 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8821 	if (status != 0)
8822 		goto out;
8823 
8824 	do_renew_lease(clp, now);
8825 
8826 	clp->cl_clientid = resp->clientid;
8827 	clp->cl_exchange_flags = resp->flags;
8828 	clp->cl_seqid = resp->seqid;
8829 	/* Client ID is not confirmed */
8830 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8831 		clear_bit(NFS4_SESSION_ESTABLISHED,
8832 			  &clp->cl_session->session_state);
8833 
8834 	if (clp->cl_serverscope != NULL &&
8835 	    !nfs41_same_server_scope(clp->cl_serverscope,
8836 				resp->server_scope)) {
8837 		dprintk("%s: server_scope mismatch detected\n",
8838 			__func__);
8839 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8840 	}
8841 
8842 	swap(clp->cl_serverowner, resp->server_owner);
8843 	swap(clp->cl_serverscope, resp->server_scope);
8844 	swap(clp->cl_implid, resp->impl_id);
8845 
8846 	/* Save the EXCHANGE_ID verifier session trunk tests */
8847 	memcpy(clp->cl_confirm.data, argp->verifier.data,
8848 	       sizeof(clp->cl_confirm.data));
8849 out:
8850 	trace_nfs4_exchange_id(clp, status);
8851 	rpc_put_task(task);
8852 	return status;
8853 }
8854 
8855 /*
8856  * nfs4_proc_exchange_id()
8857  *
8858  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8859  *
8860  * Since the clientid has expired, all compounds using sessions
8861  * associated with the stale clientid will be returning
8862  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8863  * be in some phase of session reset.
8864  *
8865  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8866  */
8867 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8868 {
8869 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8870 	int status;
8871 
8872 	/* try SP4_MACH_CRED if krb5i/p	*/
8873 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8874 	    authflavor == RPC_AUTH_GSS_KRB5P) {
8875 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8876 		if (!status)
8877 			return 0;
8878 	}
8879 
8880 	/* try SP4_NONE */
8881 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8882 }
8883 
8884 /**
8885  * nfs4_test_session_trunk
8886  *
8887  * This is an add_xprt_test() test function called from
8888  * rpc_clnt_setup_test_and_add_xprt.
8889  *
8890  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8891  * and is dereferrenced in nfs4_exchange_id_release
8892  *
8893  * Upon success, add the new transport to the rpc_clnt
8894  *
8895  * @clnt: struct rpc_clnt to get new transport
8896  * @xprt: the rpc_xprt to test
8897  * @data: call data for _nfs4_proc_exchange_id.
8898  */
8899 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8900 			    void *data)
8901 {
8902 	struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8903 	struct rpc_task *task;
8904 	int status;
8905 
8906 	u32 sp4_how;
8907 
8908 	dprintk("--> %s try %s\n", __func__,
8909 		xprt->address_strings[RPC_DISPLAY_ADDR]);
8910 
8911 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8912 
8913 	/* Test connection for session trunking. Async exchange_id call */
8914 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8915 	if (IS_ERR(task))
8916 		return;
8917 
8918 	status = task->tk_status;
8919 	if (status == 0)
8920 		status = nfs4_detect_session_trunking(adata->clp,
8921 				task->tk_msg.rpc_resp, xprt);
8922 
8923 	if (status == 0)
8924 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8925 	else if (rpc_clnt_xprt_switch_has_addr(clnt,
8926 				(struct sockaddr *)&xprt->addr))
8927 		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8928 
8929 	rpc_put_task(task);
8930 }
8931 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8932 
8933 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8934 		const struct cred *cred)
8935 {
8936 	struct rpc_message msg = {
8937 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8938 		.rpc_argp = clp,
8939 		.rpc_cred = cred,
8940 	};
8941 	int status;
8942 
8943 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
8944 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8945 	trace_nfs4_destroy_clientid(clp, status);
8946 	if (status)
8947 		dprintk("NFS: Got error %d from the server %s on "
8948 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
8949 	return status;
8950 }
8951 
8952 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8953 		const struct cred *cred)
8954 {
8955 	unsigned int loop;
8956 	int ret;
8957 
8958 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8959 		ret = _nfs4_proc_destroy_clientid(clp, cred);
8960 		switch (ret) {
8961 		case -NFS4ERR_DELAY:
8962 		case -NFS4ERR_CLIENTID_BUSY:
8963 			ssleep(1);
8964 			break;
8965 		default:
8966 			return ret;
8967 		}
8968 	}
8969 	return 0;
8970 }
8971 
8972 int nfs4_destroy_clientid(struct nfs_client *clp)
8973 {
8974 	const struct cred *cred;
8975 	int ret = 0;
8976 
8977 	if (clp->cl_mvops->minor_version < 1)
8978 		goto out;
8979 	if (clp->cl_exchange_flags == 0)
8980 		goto out;
8981 	if (clp->cl_preserve_clid)
8982 		goto out;
8983 	cred = nfs4_get_clid_cred(clp);
8984 	ret = nfs4_proc_destroy_clientid(clp, cred);
8985 	put_cred(cred);
8986 	switch (ret) {
8987 	case 0:
8988 	case -NFS4ERR_STALE_CLIENTID:
8989 		clp->cl_exchange_flags = 0;
8990 	}
8991 out:
8992 	return ret;
8993 }
8994 
8995 #endif /* CONFIG_NFS_V4_1 */
8996 
8997 struct nfs4_get_lease_time_data {
8998 	struct nfs4_get_lease_time_args *args;
8999 	struct nfs4_get_lease_time_res *res;
9000 	struct nfs_client *clp;
9001 };
9002 
9003 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9004 					void *calldata)
9005 {
9006 	struct nfs4_get_lease_time_data *data =
9007 			(struct nfs4_get_lease_time_data *)calldata;
9008 
9009 	/* just setup sequence, do not trigger session recovery
9010 	   since we're invoked within one */
9011 	nfs4_setup_sequence(data->clp,
9012 			&data->args->la_seq_args,
9013 			&data->res->lr_seq_res,
9014 			task);
9015 }
9016 
9017 /*
9018  * Called from nfs4_state_manager thread for session setup, so don't recover
9019  * from sequence operation or clientid errors.
9020  */
9021 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9022 {
9023 	struct nfs4_get_lease_time_data *data =
9024 			(struct nfs4_get_lease_time_data *)calldata;
9025 
9026 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9027 		return;
9028 	switch (task->tk_status) {
9029 	case -NFS4ERR_DELAY:
9030 	case -NFS4ERR_GRACE:
9031 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9032 		task->tk_status = 0;
9033 		fallthrough;
9034 	case -NFS4ERR_RETRY_UNCACHED_REP:
9035 		rpc_restart_call_prepare(task);
9036 		return;
9037 	}
9038 }
9039 
9040 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9041 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9042 	.rpc_call_done = nfs4_get_lease_time_done,
9043 };
9044 
9045 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9046 {
9047 	struct nfs4_get_lease_time_args args;
9048 	struct nfs4_get_lease_time_res res = {
9049 		.lr_fsinfo = fsinfo,
9050 	};
9051 	struct nfs4_get_lease_time_data data = {
9052 		.args = &args,
9053 		.res = &res,
9054 		.clp = clp,
9055 	};
9056 	struct rpc_message msg = {
9057 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9058 		.rpc_argp = &args,
9059 		.rpc_resp = &res,
9060 	};
9061 	struct rpc_task_setup task_setup = {
9062 		.rpc_client = clp->cl_rpcclient,
9063 		.rpc_message = &msg,
9064 		.callback_ops = &nfs4_get_lease_time_ops,
9065 		.callback_data = &data,
9066 		.flags = RPC_TASK_TIMEOUT,
9067 	};
9068 
9069 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9070 	return nfs4_call_sync_custom(&task_setup);
9071 }
9072 
9073 #ifdef CONFIG_NFS_V4_1
9074 
9075 /*
9076  * Initialize the values to be used by the client in CREATE_SESSION
9077  * If nfs4_init_session set the fore channel request and response sizes,
9078  * use them.
9079  *
9080  * Set the back channel max_resp_sz_cached to zero to force the client to
9081  * always set csa_cachethis to FALSE because the current implementation
9082  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9083  */
9084 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9085 				    struct rpc_clnt *clnt)
9086 {
9087 	unsigned int max_rqst_sz, max_resp_sz;
9088 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9089 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9090 
9091 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9092 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9093 
9094 	/* Fore channel attributes */
9095 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9096 	args->fc_attrs.max_resp_sz = max_resp_sz;
9097 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9098 	args->fc_attrs.max_reqs = max_session_slots;
9099 
9100 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9101 		"max_ops=%u max_reqs=%u\n",
9102 		__func__,
9103 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9104 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9105 
9106 	/* Back channel attributes */
9107 	args->bc_attrs.max_rqst_sz = max_bc_payload;
9108 	args->bc_attrs.max_resp_sz = max_bc_payload;
9109 	args->bc_attrs.max_resp_sz_cached = 0;
9110 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9111 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9112 	if (args->bc_attrs.max_reqs > max_bc_slots)
9113 		args->bc_attrs.max_reqs = max_bc_slots;
9114 
9115 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9116 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9117 		__func__,
9118 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9119 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9120 		args->bc_attrs.max_reqs);
9121 }
9122 
9123 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9124 		struct nfs41_create_session_res *res)
9125 {
9126 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9127 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9128 
9129 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9130 		return -EINVAL;
9131 	/*
9132 	 * Our requested max_ops is the minimum we need; we're not
9133 	 * prepared to break up compounds into smaller pieces than that.
9134 	 * So, no point even trying to continue if the server won't
9135 	 * cooperate:
9136 	 */
9137 	if (rcvd->max_ops < sent->max_ops)
9138 		return -EINVAL;
9139 	if (rcvd->max_reqs == 0)
9140 		return -EINVAL;
9141 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9142 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9143 	return 0;
9144 }
9145 
9146 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9147 		struct nfs41_create_session_res *res)
9148 {
9149 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9150 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9151 
9152 	if (!(res->flags & SESSION4_BACK_CHAN))
9153 		goto out;
9154 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9155 		return -EINVAL;
9156 	if (rcvd->max_resp_sz < sent->max_resp_sz)
9157 		return -EINVAL;
9158 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9159 		return -EINVAL;
9160 	if (rcvd->max_ops > sent->max_ops)
9161 		return -EINVAL;
9162 	if (rcvd->max_reqs > sent->max_reqs)
9163 		return -EINVAL;
9164 out:
9165 	return 0;
9166 }
9167 
9168 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9169 				     struct nfs41_create_session_res *res)
9170 {
9171 	int ret;
9172 
9173 	ret = nfs4_verify_fore_channel_attrs(args, res);
9174 	if (ret)
9175 		return ret;
9176 	return nfs4_verify_back_channel_attrs(args, res);
9177 }
9178 
9179 static void nfs4_update_session(struct nfs4_session *session,
9180 		struct nfs41_create_session_res *res)
9181 {
9182 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9183 	/* Mark client id and session as being confirmed */
9184 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9185 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9186 	session->flags = res->flags;
9187 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9188 	if (res->flags & SESSION4_BACK_CHAN)
9189 		memcpy(&session->bc_attrs, &res->bc_attrs,
9190 				sizeof(session->bc_attrs));
9191 }
9192 
9193 static int _nfs4_proc_create_session(struct nfs_client *clp,
9194 		const struct cred *cred)
9195 {
9196 	struct nfs4_session *session = clp->cl_session;
9197 	struct nfs41_create_session_args args = {
9198 		.client = clp,
9199 		.clientid = clp->cl_clientid,
9200 		.seqid = clp->cl_seqid,
9201 		.cb_program = NFS4_CALLBACK,
9202 	};
9203 	struct nfs41_create_session_res res;
9204 
9205 	struct rpc_message msg = {
9206 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9207 		.rpc_argp = &args,
9208 		.rpc_resp = &res,
9209 		.rpc_cred = cred,
9210 	};
9211 	int status;
9212 
9213 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9214 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9215 
9216 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9217 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9218 	trace_nfs4_create_session(clp, status);
9219 
9220 	switch (status) {
9221 	case -NFS4ERR_STALE_CLIENTID:
9222 	case -NFS4ERR_DELAY:
9223 	case -ETIMEDOUT:
9224 	case -EACCES:
9225 	case -EAGAIN:
9226 		goto out;
9227 	}
9228 
9229 	clp->cl_seqid++;
9230 	if (!status) {
9231 		/* Verify the session's negotiated channel_attrs values */
9232 		status = nfs4_verify_channel_attrs(&args, &res);
9233 		/* Increment the clientid slot sequence id */
9234 		if (status)
9235 			goto out;
9236 		nfs4_update_session(session, &res);
9237 	}
9238 out:
9239 	return status;
9240 }
9241 
9242 /*
9243  * Issues a CREATE_SESSION operation to the server.
9244  * It is the responsibility of the caller to verify the session is
9245  * expired before calling this routine.
9246  */
9247 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9248 {
9249 	int status;
9250 	unsigned *ptr;
9251 	struct nfs4_session *session = clp->cl_session;
9252 
9253 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9254 
9255 	status = _nfs4_proc_create_session(clp, cred);
9256 	if (status)
9257 		goto out;
9258 
9259 	/* Init or reset the session slot tables */
9260 	status = nfs4_setup_session_slot_tables(session);
9261 	dprintk("slot table setup returned %d\n", status);
9262 	if (status)
9263 		goto out;
9264 
9265 	ptr = (unsigned *)&session->sess_id.data[0];
9266 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9267 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9268 out:
9269 	return status;
9270 }
9271 
9272 /*
9273  * Issue the over-the-wire RPC DESTROY_SESSION.
9274  * The caller must serialize access to this routine.
9275  */
9276 int nfs4_proc_destroy_session(struct nfs4_session *session,
9277 		const struct cred *cred)
9278 {
9279 	struct rpc_message msg = {
9280 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9281 		.rpc_argp = session,
9282 		.rpc_cred = cred,
9283 	};
9284 	int status = 0;
9285 
9286 	/* session is still being setup */
9287 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9288 		return 0;
9289 
9290 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9291 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9292 	trace_nfs4_destroy_session(session->clp, status);
9293 
9294 	if (status)
9295 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9296 			"Session has been destroyed regardless...\n", status);
9297 	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9298 	return status;
9299 }
9300 
9301 /*
9302  * Renew the cl_session lease.
9303  */
9304 struct nfs4_sequence_data {
9305 	struct nfs_client *clp;
9306 	struct nfs4_sequence_args args;
9307 	struct nfs4_sequence_res res;
9308 };
9309 
9310 static void nfs41_sequence_release(void *data)
9311 {
9312 	struct nfs4_sequence_data *calldata = data;
9313 	struct nfs_client *clp = calldata->clp;
9314 
9315 	if (refcount_read(&clp->cl_count) > 1)
9316 		nfs4_schedule_state_renewal(clp);
9317 	nfs_put_client(clp);
9318 	kfree(calldata);
9319 }
9320 
9321 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9322 {
9323 	switch(task->tk_status) {
9324 	case -NFS4ERR_DELAY:
9325 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9326 		return -EAGAIN;
9327 	default:
9328 		nfs4_schedule_lease_recovery(clp);
9329 	}
9330 	return 0;
9331 }
9332 
9333 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9334 {
9335 	struct nfs4_sequence_data *calldata = data;
9336 	struct nfs_client *clp = calldata->clp;
9337 
9338 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9339 		return;
9340 
9341 	trace_nfs4_sequence(clp, task->tk_status);
9342 	if (task->tk_status < 0) {
9343 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9344 		if (refcount_read(&clp->cl_count) == 1)
9345 			return;
9346 
9347 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9348 			rpc_restart_call_prepare(task);
9349 			return;
9350 		}
9351 	}
9352 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9353 }
9354 
9355 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9356 {
9357 	struct nfs4_sequence_data *calldata = data;
9358 	struct nfs_client *clp = calldata->clp;
9359 	struct nfs4_sequence_args *args;
9360 	struct nfs4_sequence_res *res;
9361 
9362 	args = task->tk_msg.rpc_argp;
9363 	res = task->tk_msg.rpc_resp;
9364 
9365 	nfs4_setup_sequence(clp, args, res, task);
9366 }
9367 
9368 static const struct rpc_call_ops nfs41_sequence_ops = {
9369 	.rpc_call_done = nfs41_sequence_call_done,
9370 	.rpc_call_prepare = nfs41_sequence_prepare,
9371 	.rpc_release = nfs41_sequence_release,
9372 };
9373 
9374 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9375 		const struct cred *cred,
9376 		struct nfs4_slot *slot,
9377 		bool is_privileged)
9378 {
9379 	struct nfs4_sequence_data *calldata;
9380 	struct rpc_message msg = {
9381 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9382 		.rpc_cred = cred,
9383 	};
9384 	struct rpc_task_setup task_setup_data = {
9385 		.rpc_client = clp->cl_rpcclient,
9386 		.rpc_message = &msg,
9387 		.callback_ops = &nfs41_sequence_ops,
9388 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9389 	};
9390 	struct rpc_task *ret;
9391 
9392 	ret = ERR_PTR(-EIO);
9393 	if (!refcount_inc_not_zero(&clp->cl_count))
9394 		goto out_err;
9395 
9396 	ret = ERR_PTR(-ENOMEM);
9397 	calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9398 	if (calldata == NULL)
9399 		goto out_put_clp;
9400 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9401 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9402 	msg.rpc_argp = &calldata->args;
9403 	msg.rpc_resp = &calldata->res;
9404 	calldata->clp = clp;
9405 	task_setup_data.callback_data = calldata;
9406 
9407 	ret = rpc_run_task(&task_setup_data);
9408 	if (IS_ERR(ret))
9409 		goto out_err;
9410 	return ret;
9411 out_put_clp:
9412 	nfs_put_client(clp);
9413 out_err:
9414 	nfs41_release_slot(slot);
9415 	return ret;
9416 }
9417 
9418 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9419 {
9420 	struct rpc_task *task;
9421 	int ret = 0;
9422 
9423 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9424 		return -EAGAIN;
9425 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9426 	if (IS_ERR(task))
9427 		ret = PTR_ERR(task);
9428 	else
9429 		rpc_put_task_async(task);
9430 	dprintk("<-- %s status=%d\n", __func__, ret);
9431 	return ret;
9432 }
9433 
9434 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9435 {
9436 	struct rpc_task *task;
9437 	int ret;
9438 
9439 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9440 	if (IS_ERR(task)) {
9441 		ret = PTR_ERR(task);
9442 		goto out;
9443 	}
9444 	ret = rpc_wait_for_completion_task(task);
9445 	if (!ret)
9446 		ret = task->tk_status;
9447 	rpc_put_task(task);
9448 out:
9449 	dprintk("<-- %s status=%d\n", __func__, ret);
9450 	return ret;
9451 }
9452 
9453 struct nfs4_reclaim_complete_data {
9454 	struct nfs_client *clp;
9455 	struct nfs41_reclaim_complete_args arg;
9456 	struct nfs41_reclaim_complete_res res;
9457 };
9458 
9459 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9460 {
9461 	struct nfs4_reclaim_complete_data *calldata = data;
9462 
9463 	nfs4_setup_sequence(calldata->clp,
9464 			&calldata->arg.seq_args,
9465 			&calldata->res.seq_res,
9466 			task);
9467 }
9468 
9469 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9470 {
9471 	switch(task->tk_status) {
9472 	case 0:
9473 		wake_up_all(&clp->cl_lock_waitq);
9474 		fallthrough;
9475 	case -NFS4ERR_COMPLETE_ALREADY:
9476 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9477 		break;
9478 	case -NFS4ERR_DELAY:
9479 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9480 		fallthrough;
9481 	case -NFS4ERR_RETRY_UNCACHED_REP:
9482 		return -EAGAIN;
9483 	case -NFS4ERR_BADSESSION:
9484 	case -NFS4ERR_DEADSESSION:
9485 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9486 		break;
9487 	default:
9488 		nfs4_schedule_lease_recovery(clp);
9489 	}
9490 	return 0;
9491 }
9492 
9493 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9494 {
9495 	struct nfs4_reclaim_complete_data *calldata = data;
9496 	struct nfs_client *clp = calldata->clp;
9497 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9498 
9499 	if (!nfs41_sequence_done(task, res))
9500 		return;
9501 
9502 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9503 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9504 		rpc_restart_call_prepare(task);
9505 		return;
9506 	}
9507 }
9508 
9509 static void nfs4_free_reclaim_complete_data(void *data)
9510 {
9511 	struct nfs4_reclaim_complete_data *calldata = data;
9512 
9513 	kfree(calldata);
9514 }
9515 
9516 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9517 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9518 	.rpc_call_done = nfs4_reclaim_complete_done,
9519 	.rpc_release = nfs4_free_reclaim_complete_data,
9520 };
9521 
9522 /*
9523  * Issue a global reclaim complete.
9524  */
9525 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9526 		const struct cred *cred)
9527 {
9528 	struct nfs4_reclaim_complete_data *calldata;
9529 	struct rpc_message msg = {
9530 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9531 		.rpc_cred = cred,
9532 	};
9533 	struct rpc_task_setup task_setup_data = {
9534 		.rpc_client = clp->cl_rpcclient,
9535 		.rpc_message = &msg,
9536 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9537 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9538 	};
9539 	int status = -ENOMEM;
9540 
9541 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9542 	if (calldata == NULL)
9543 		goto out;
9544 	calldata->clp = clp;
9545 	calldata->arg.one_fs = 0;
9546 
9547 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9548 	msg.rpc_argp = &calldata->arg;
9549 	msg.rpc_resp = &calldata->res;
9550 	task_setup_data.callback_data = calldata;
9551 	status = nfs4_call_sync_custom(&task_setup_data);
9552 out:
9553 	dprintk("<-- %s status=%d\n", __func__, status);
9554 	return status;
9555 }
9556 
9557 static void
9558 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9559 {
9560 	struct nfs4_layoutget *lgp = calldata;
9561 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9562 
9563 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9564 				&lgp->res.seq_res, task);
9565 }
9566 
9567 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9568 {
9569 	struct nfs4_layoutget *lgp = calldata;
9570 
9571 	nfs41_sequence_process(task, &lgp->res.seq_res);
9572 }
9573 
9574 static int
9575 nfs4_layoutget_handle_exception(struct rpc_task *task,
9576 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9577 {
9578 	struct inode *inode = lgp->args.inode;
9579 	struct nfs_server *server = NFS_SERVER(inode);
9580 	struct pnfs_layout_hdr *lo = lgp->lo;
9581 	int nfs4err = task->tk_status;
9582 	int err, status = 0;
9583 	LIST_HEAD(head);
9584 
9585 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9586 
9587 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9588 
9589 	switch (nfs4err) {
9590 	case 0:
9591 		goto out;
9592 
9593 	/*
9594 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9595 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9596 	 * retry go inband.
9597 	 */
9598 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9599 		status = -ENODATA;
9600 		goto out;
9601 	/*
9602 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9603 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9604 	 */
9605 	case -NFS4ERR_BADLAYOUT:
9606 		status = -EOVERFLOW;
9607 		goto out;
9608 	/*
9609 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9610 	 * (or clients) writing to the same RAID stripe except when
9611 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9612 	 *
9613 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9614 	 * while, and then eventually give up.
9615 	 */
9616 	case -NFS4ERR_LAYOUTTRYLATER:
9617 		if (lgp->args.minlength == 0) {
9618 			status = -EOVERFLOW;
9619 			goto out;
9620 		}
9621 		status = -EBUSY;
9622 		break;
9623 	case -NFS4ERR_RECALLCONFLICT:
9624 		status = -ERECALLCONFLICT;
9625 		break;
9626 	case -NFS4ERR_DELEG_REVOKED:
9627 	case -NFS4ERR_ADMIN_REVOKED:
9628 	case -NFS4ERR_EXPIRED:
9629 	case -NFS4ERR_BAD_STATEID:
9630 		exception->timeout = 0;
9631 		spin_lock(&inode->i_lock);
9632 		/* If the open stateid was bad, then recover it. */
9633 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9634 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9635 			spin_unlock(&inode->i_lock);
9636 			exception->state = lgp->args.ctx->state;
9637 			exception->stateid = &lgp->args.stateid;
9638 			break;
9639 		}
9640 
9641 		/*
9642 		 * Mark the bad layout state as invalid, then retry
9643 		 */
9644 		pnfs_mark_layout_stateid_invalid(lo, &head);
9645 		spin_unlock(&inode->i_lock);
9646 		nfs_commit_inode(inode, 0);
9647 		pnfs_free_lseg_list(&head);
9648 		status = -EAGAIN;
9649 		goto out;
9650 	}
9651 
9652 	err = nfs4_handle_exception(server, nfs4err, exception);
9653 	if (!status) {
9654 		if (exception->retry)
9655 			status = -EAGAIN;
9656 		else
9657 			status = err;
9658 	}
9659 out:
9660 	return status;
9661 }
9662 
9663 size_t max_response_pages(struct nfs_server *server)
9664 {
9665 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9666 	return nfs_page_array_len(0, max_resp_sz);
9667 }
9668 
9669 static void nfs4_layoutget_release(void *calldata)
9670 {
9671 	struct nfs4_layoutget *lgp = calldata;
9672 
9673 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9674 	pnfs_layoutget_free(lgp);
9675 }
9676 
9677 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9678 	.rpc_call_prepare = nfs4_layoutget_prepare,
9679 	.rpc_call_done = nfs4_layoutget_done,
9680 	.rpc_release = nfs4_layoutget_release,
9681 };
9682 
9683 struct pnfs_layout_segment *
9684 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9685 {
9686 	struct inode *inode = lgp->args.inode;
9687 	struct nfs_server *server = NFS_SERVER(inode);
9688 	struct rpc_task *task;
9689 	struct rpc_message msg = {
9690 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9691 		.rpc_argp = &lgp->args,
9692 		.rpc_resp = &lgp->res,
9693 		.rpc_cred = lgp->cred,
9694 	};
9695 	struct rpc_task_setup task_setup_data = {
9696 		.rpc_client = server->client,
9697 		.rpc_message = &msg,
9698 		.callback_ops = &nfs4_layoutget_call_ops,
9699 		.callback_data = lgp,
9700 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9701 			 RPC_TASK_MOVEABLE,
9702 	};
9703 	struct pnfs_layout_segment *lseg = NULL;
9704 	struct nfs4_exception exception = {
9705 		.inode = inode,
9706 		.timeout = *timeout,
9707 	};
9708 	int status = 0;
9709 
9710 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9711 
9712 	task = rpc_run_task(&task_setup_data);
9713 	if (IS_ERR(task))
9714 		return ERR_CAST(task);
9715 
9716 	status = rpc_wait_for_completion_task(task);
9717 	if (status != 0)
9718 		goto out;
9719 
9720 	if (task->tk_status < 0) {
9721 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9722 		*timeout = exception.timeout;
9723 	} else if (lgp->res.layoutp->len == 0) {
9724 		status = -EAGAIN;
9725 		*timeout = nfs4_update_delay(&exception.timeout);
9726 	} else
9727 		lseg = pnfs_layout_process(lgp);
9728 out:
9729 	trace_nfs4_layoutget(lgp->args.ctx,
9730 			&lgp->args.range,
9731 			&lgp->res.range,
9732 			&lgp->res.stateid,
9733 			status);
9734 
9735 	rpc_put_task(task);
9736 	dprintk("<-- %s status=%d\n", __func__, status);
9737 	if (status)
9738 		return ERR_PTR(status);
9739 	return lseg;
9740 }
9741 
9742 static void
9743 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9744 {
9745 	struct nfs4_layoutreturn *lrp = calldata;
9746 
9747 	nfs4_setup_sequence(lrp->clp,
9748 			&lrp->args.seq_args,
9749 			&lrp->res.seq_res,
9750 			task);
9751 	if (!pnfs_layout_is_valid(lrp->args.layout))
9752 		rpc_exit(task, 0);
9753 }
9754 
9755 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9756 {
9757 	struct nfs4_layoutreturn *lrp = calldata;
9758 	struct nfs_server *server;
9759 
9760 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9761 		return;
9762 
9763 	/*
9764 	 * Was there an RPC level error? Assume the call succeeded,
9765 	 * and that we need to release the layout
9766 	 */
9767 	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9768 		lrp->res.lrs_present = 0;
9769 		return;
9770 	}
9771 
9772 	server = NFS_SERVER(lrp->args.inode);
9773 	switch (task->tk_status) {
9774 	case -NFS4ERR_OLD_STATEID:
9775 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9776 					&lrp->args.range,
9777 					lrp->args.inode))
9778 			goto out_restart;
9779 		fallthrough;
9780 	default:
9781 		task->tk_status = 0;
9782 		fallthrough;
9783 	case 0:
9784 		break;
9785 	case -NFS4ERR_DELAY:
9786 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9787 			break;
9788 		goto out_restart;
9789 	}
9790 	return;
9791 out_restart:
9792 	task->tk_status = 0;
9793 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9794 	rpc_restart_call_prepare(task);
9795 }
9796 
9797 static void nfs4_layoutreturn_release(void *calldata)
9798 {
9799 	struct nfs4_layoutreturn *lrp = calldata;
9800 	struct pnfs_layout_hdr *lo = lrp->args.layout;
9801 
9802 	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9803 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9804 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9805 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9806 		lrp->ld_private.ops->free(&lrp->ld_private);
9807 	pnfs_put_layout_hdr(lrp->args.layout);
9808 	nfs_iput_and_deactive(lrp->inode);
9809 	put_cred(lrp->cred);
9810 	kfree(calldata);
9811 }
9812 
9813 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9814 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9815 	.rpc_call_done = nfs4_layoutreturn_done,
9816 	.rpc_release = nfs4_layoutreturn_release,
9817 };
9818 
9819 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9820 {
9821 	struct rpc_task *task;
9822 	struct rpc_message msg = {
9823 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9824 		.rpc_argp = &lrp->args,
9825 		.rpc_resp = &lrp->res,
9826 		.rpc_cred = lrp->cred,
9827 	};
9828 	struct rpc_task_setup task_setup_data = {
9829 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9830 		.rpc_message = &msg,
9831 		.callback_ops = &nfs4_layoutreturn_call_ops,
9832 		.callback_data = lrp,
9833 		.flags = RPC_TASK_MOVEABLE,
9834 	};
9835 	int status = 0;
9836 
9837 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9838 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9839 			&task_setup_data.rpc_client, &msg);
9840 
9841 	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9842 	if (!sync) {
9843 		if (!lrp->inode) {
9844 			nfs4_layoutreturn_release(lrp);
9845 			return -EAGAIN;
9846 		}
9847 		task_setup_data.flags |= RPC_TASK_ASYNC;
9848 	}
9849 	if (!lrp->inode)
9850 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9851 				   1);
9852 	else
9853 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9854 				   0);
9855 	task = rpc_run_task(&task_setup_data);
9856 	if (IS_ERR(task))
9857 		return PTR_ERR(task);
9858 	if (sync)
9859 		status = task->tk_status;
9860 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9861 	dprintk("<-- %s status=%d\n", __func__, status);
9862 	rpc_put_task(task);
9863 	return status;
9864 }
9865 
9866 static int
9867 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9868 		struct pnfs_device *pdev,
9869 		const struct cred *cred)
9870 {
9871 	struct nfs4_getdeviceinfo_args args = {
9872 		.pdev = pdev,
9873 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9874 			NOTIFY_DEVICEID4_DELETE,
9875 	};
9876 	struct nfs4_getdeviceinfo_res res = {
9877 		.pdev = pdev,
9878 	};
9879 	struct rpc_message msg = {
9880 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9881 		.rpc_argp = &args,
9882 		.rpc_resp = &res,
9883 		.rpc_cred = cred,
9884 	};
9885 	int status;
9886 
9887 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9888 	if (res.notification & ~args.notify_types)
9889 		dprintk("%s: unsupported notification\n", __func__);
9890 	if (res.notification != args.notify_types)
9891 		pdev->nocache = 1;
9892 
9893 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9894 
9895 	dprintk("<-- %s status=%d\n", __func__, status);
9896 
9897 	return status;
9898 }
9899 
9900 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9901 		struct pnfs_device *pdev,
9902 		const struct cred *cred)
9903 {
9904 	struct nfs4_exception exception = { };
9905 	int err;
9906 
9907 	do {
9908 		err = nfs4_handle_exception(server,
9909 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9910 					&exception);
9911 	} while (exception.retry);
9912 	return err;
9913 }
9914 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9915 
9916 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9917 {
9918 	struct nfs4_layoutcommit_data *data = calldata;
9919 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9920 
9921 	nfs4_setup_sequence(server->nfs_client,
9922 			&data->args.seq_args,
9923 			&data->res.seq_res,
9924 			task);
9925 }
9926 
9927 static void
9928 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9929 {
9930 	struct nfs4_layoutcommit_data *data = calldata;
9931 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9932 
9933 	if (!nfs41_sequence_done(task, &data->res.seq_res))
9934 		return;
9935 
9936 	switch (task->tk_status) { /* Just ignore these failures */
9937 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9938 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9939 	case -NFS4ERR_BADLAYOUT:     /* no layout */
9940 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
9941 		task->tk_status = 0;
9942 		break;
9943 	case 0:
9944 		break;
9945 	default:
9946 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9947 			rpc_restart_call_prepare(task);
9948 			return;
9949 		}
9950 	}
9951 }
9952 
9953 static void nfs4_layoutcommit_release(void *calldata)
9954 {
9955 	struct nfs4_layoutcommit_data *data = calldata;
9956 
9957 	pnfs_cleanup_layoutcommit(data);
9958 	nfs_post_op_update_inode_force_wcc(data->args.inode,
9959 					   data->res.fattr);
9960 	put_cred(data->cred);
9961 	nfs_iput_and_deactive(data->inode);
9962 	kfree(data);
9963 }
9964 
9965 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9966 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
9967 	.rpc_call_done = nfs4_layoutcommit_done,
9968 	.rpc_release = nfs4_layoutcommit_release,
9969 };
9970 
9971 int
9972 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9973 {
9974 	struct rpc_message msg = {
9975 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9976 		.rpc_argp = &data->args,
9977 		.rpc_resp = &data->res,
9978 		.rpc_cred = data->cred,
9979 	};
9980 	struct rpc_task_setup task_setup_data = {
9981 		.task = &data->task,
9982 		.rpc_client = NFS_CLIENT(data->args.inode),
9983 		.rpc_message = &msg,
9984 		.callback_ops = &nfs4_layoutcommit_ops,
9985 		.callback_data = data,
9986 		.flags = RPC_TASK_MOVEABLE,
9987 	};
9988 	struct rpc_task *task;
9989 	int status = 0;
9990 
9991 	dprintk("NFS: initiating layoutcommit call. sync %d "
9992 		"lbw: %llu inode %lu\n", sync,
9993 		data->args.lastbytewritten,
9994 		data->args.inode->i_ino);
9995 
9996 	if (!sync) {
9997 		data->inode = nfs_igrab_and_active(data->args.inode);
9998 		if (data->inode == NULL) {
9999 			nfs4_layoutcommit_release(data);
10000 			return -EAGAIN;
10001 		}
10002 		task_setup_data.flags = RPC_TASK_ASYNC;
10003 	}
10004 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10005 	task = rpc_run_task(&task_setup_data);
10006 	if (IS_ERR(task))
10007 		return PTR_ERR(task);
10008 	if (sync)
10009 		status = task->tk_status;
10010 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10011 	dprintk("%s: status %d\n", __func__, status);
10012 	rpc_put_task(task);
10013 	return status;
10014 }
10015 
10016 /*
10017  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10018  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10019  */
10020 static int
10021 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10022 		    struct nfs_fsinfo *info,
10023 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10024 {
10025 	struct nfs41_secinfo_no_name_args args = {
10026 		.style = SECINFO_STYLE_CURRENT_FH,
10027 	};
10028 	struct nfs4_secinfo_res res = {
10029 		.flavors = flavors,
10030 	};
10031 	struct rpc_message msg = {
10032 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10033 		.rpc_argp = &args,
10034 		.rpc_resp = &res,
10035 	};
10036 	struct nfs4_call_sync_data data = {
10037 		.seq_server = server,
10038 		.seq_args = &args.seq_args,
10039 		.seq_res = &res.seq_res,
10040 	};
10041 	struct rpc_task_setup task_setup = {
10042 		.rpc_client = server->client,
10043 		.rpc_message = &msg,
10044 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10045 		.callback_data = &data,
10046 		.flags = RPC_TASK_NO_ROUND_ROBIN,
10047 	};
10048 	const struct cred *cred = NULL;
10049 	int status;
10050 
10051 	if (use_integrity) {
10052 		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10053 
10054 		cred = nfs4_get_clid_cred(server->nfs_client);
10055 		msg.rpc_cred = cred;
10056 	}
10057 
10058 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10059 	status = nfs4_call_sync_custom(&task_setup);
10060 	dprintk("<-- %s status=%d\n", __func__, status);
10061 
10062 	put_cred(cred);
10063 
10064 	return status;
10065 }
10066 
10067 static int
10068 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10069 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10070 {
10071 	struct nfs4_exception exception = {
10072 		.interruptible = true,
10073 	};
10074 	int err;
10075 	do {
10076 		/* first try using integrity protection */
10077 		err = -NFS4ERR_WRONGSEC;
10078 
10079 		/* try to use integrity protection with machine cred */
10080 		if (_nfs4_is_integrity_protected(server->nfs_client))
10081 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10082 							  flavors, true);
10083 
10084 		/*
10085 		 * if unable to use integrity protection, or SECINFO with
10086 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10087 		 * disallowed by spec, but exists in deployed servers) use
10088 		 * the current filesystem's rpc_client and the user cred.
10089 		 */
10090 		if (err == -NFS4ERR_WRONGSEC)
10091 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10092 							  flavors, false);
10093 
10094 		switch (err) {
10095 		case 0:
10096 		case -NFS4ERR_WRONGSEC:
10097 		case -ENOTSUPP:
10098 			goto out;
10099 		default:
10100 			err = nfs4_handle_exception(server, err, &exception);
10101 		}
10102 	} while (exception.retry);
10103 out:
10104 	return err;
10105 }
10106 
10107 static int
10108 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10109 		    struct nfs_fsinfo *info)
10110 {
10111 	int err;
10112 	struct page *page;
10113 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10114 	struct nfs4_secinfo_flavors *flavors;
10115 	struct nfs4_secinfo4 *secinfo;
10116 	int i;
10117 
10118 	page = alloc_page(GFP_KERNEL);
10119 	if (!page) {
10120 		err = -ENOMEM;
10121 		goto out;
10122 	}
10123 
10124 	flavors = page_address(page);
10125 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10126 
10127 	/*
10128 	 * Fall back on "guess and check" method if
10129 	 * the server doesn't support SECINFO_NO_NAME
10130 	 */
10131 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10132 		err = nfs4_find_root_sec(server, fhandle, info);
10133 		goto out_freepage;
10134 	}
10135 	if (err)
10136 		goto out_freepage;
10137 
10138 	for (i = 0; i < flavors->num_flavors; i++) {
10139 		secinfo = &flavors->flavors[i];
10140 
10141 		switch (secinfo->flavor) {
10142 		case RPC_AUTH_NULL:
10143 		case RPC_AUTH_UNIX:
10144 		case RPC_AUTH_GSS:
10145 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10146 					&secinfo->flavor_info);
10147 			break;
10148 		default:
10149 			flavor = RPC_AUTH_MAXFLAVOR;
10150 			break;
10151 		}
10152 
10153 		if (!nfs_auth_info_match(&server->auth_info, flavor))
10154 			flavor = RPC_AUTH_MAXFLAVOR;
10155 
10156 		if (flavor != RPC_AUTH_MAXFLAVOR) {
10157 			err = nfs4_lookup_root_sec(server, fhandle,
10158 						   info, flavor);
10159 			if (!err)
10160 				break;
10161 		}
10162 	}
10163 
10164 	if (flavor == RPC_AUTH_MAXFLAVOR)
10165 		err = -EPERM;
10166 
10167 out_freepage:
10168 	put_page(page);
10169 	if (err == -EACCES)
10170 		return -EPERM;
10171 out:
10172 	return err;
10173 }
10174 
10175 static int _nfs41_test_stateid(struct nfs_server *server,
10176 		nfs4_stateid *stateid,
10177 		const struct cred *cred)
10178 {
10179 	int status;
10180 	struct nfs41_test_stateid_args args = {
10181 		.stateid = stateid,
10182 	};
10183 	struct nfs41_test_stateid_res res;
10184 	struct rpc_message msg = {
10185 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10186 		.rpc_argp = &args,
10187 		.rpc_resp = &res,
10188 		.rpc_cred = cred,
10189 	};
10190 	struct rpc_clnt *rpc_client = server->client;
10191 
10192 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10193 		&rpc_client, &msg);
10194 
10195 	dprintk("NFS call  test_stateid %p\n", stateid);
10196 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10197 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10198 			&args.seq_args, &res.seq_res);
10199 	if (status != NFS_OK) {
10200 		dprintk("NFS reply test_stateid: failed, %d\n", status);
10201 		return status;
10202 	}
10203 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10204 	return -res.status;
10205 }
10206 
10207 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10208 		int err, struct nfs4_exception *exception)
10209 {
10210 	exception->retry = 0;
10211 	switch(err) {
10212 	case -NFS4ERR_DELAY:
10213 	case -NFS4ERR_RETRY_UNCACHED_REP:
10214 		nfs4_handle_exception(server, err, exception);
10215 		break;
10216 	case -NFS4ERR_BADSESSION:
10217 	case -NFS4ERR_BADSLOT:
10218 	case -NFS4ERR_BAD_HIGH_SLOT:
10219 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10220 	case -NFS4ERR_DEADSESSION:
10221 		nfs4_do_handle_exception(server, err, exception);
10222 	}
10223 }
10224 
10225 /**
10226  * nfs41_test_stateid - perform a TEST_STATEID operation
10227  *
10228  * @server: server / transport on which to perform the operation
10229  * @stateid: state ID to test
10230  * @cred: credential
10231  *
10232  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10233  * Otherwise a negative NFS4ERR value is returned if the operation
10234  * failed or the state ID is not currently valid.
10235  */
10236 static int nfs41_test_stateid(struct nfs_server *server,
10237 		nfs4_stateid *stateid,
10238 		const struct cred *cred)
10239 {
10240 	struct nfs4_exception exception = {
10241 		.interruptible = true,
10242 	};
10243 	int err;
10244 	do {
10245 		err = _nfs41_test_stateid(server, stateid, cred);
10246 		nfs4_handle_delay_or_session_error(server, err, &exception);
10247 	} while (exception.retry);
10248 	return err;
10249 }
10250 
10251 struct nfs_free_stateid_data {
10252 	struct nfs_server *server;
10253 	struct nfs41_free_stateid_args args;
10254 	struct nfs41_free_stateid_res res;
10255 };
10256 
10257 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10258 {
10259 	struct nfs_free_stateid_data *data = calldata;
10260 	nfs4_setup_sequence(data->server->nfs_client,
10261 			&data->args.seq_args,
10262 			&data->res.seq_res,
10263 			task);
10264 }
10265 
10266 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10267 {
10268 	struct nfs_free_stateid_data *data = calldata;
10269 
10270 	nfs41_sequence_done(task, &data->res.seq_res);
10271 
10272 	switch (task->tk_status) {
10273 	case -NFS4ERR_DELAY:
10274 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10275 			rpc_restart_call_prepare(task);
10276 	}
10277 }
10278 
10279 static void nfs41_free_stateid_release(void *calldata)
10280 {
10281 	struct nfs_free_stateid_data *data = calldata;
10282 	struct nfs_client *clp = data->server->nfs_client;
10283 
10284 	nfs_put_client(clp);
10285 	kfree(calldata);
10286 }
10287 
10288 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10289 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10290 	.rpc_call_done = nfs41_free_stateid_done,
10291 	.rpc_release = nfs41_free_stateid_release,
10292 };
10293 
10294 /**
10295  * nfs41_free_stateid - perform a FREE_STATEID operation
10296  *
10297  * @server: server / transport on which to perform the operation
10298  * @stateid: state ID to release
10299  * @cred: credential
10300  * @privileged: set to true if this call needs to be privileged
10301  *
10302  * Note: this function is always asynchronous.
10303  */
10304 static int nfs41_free_stateid(struct nfs_server *server,
10305 		const nfs4_stateid *stateid,
10306 		const struct cred *cred,
10307 		bool privileged)
10308 {
10309 	struct rpc_message msg = {
10310 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10311 		.rpc_cred = cred,
10312 	};
10313 	struct rpc_task_setup task_setup = {
10314 		.rpc_client = server->client,
10315 		.rpc_message = &msg,
10316 		.callback_ops = &nfs41_free_stateid_ops,
10317 		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10318 	};
10319 	struct nfs_free_stateid_data *data;
10320 	struct rpc_task *task;
10321 	struct nfs_client *clp = server->nfs_client;
10322 
10323 	if (!refcount_inc_not_zero(&clp->cl_count))
10324 		return -EIO;
10325 
10326 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10327 		&task_setup.rpc_client, &msg);
10328 
10329 	dprintk("NFS call  free_stateid %p\n", stateid);
10330 	data = kmalloc(sizeof(*data), GFP_KERNEL);
10331 	if (!data)
10332 		return -ENOMEM;
10333 	data->server = server;
10334 	nfs4_stateid_copy(&data->args.stateid, stateid);
10335 
10336 	task_setup.callback_data = data;
10337 
10338 	msg.rpc_argp = &data->args;
10339 	msg.rpc_resp = &data->res;
10340 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10341 	task = rpc_run_task(&task_setup);
10342 	if (IS_ERR(task))
10343 		return PTR_ERR(task);
10344 	rpc_put_task(task);
10345 	return 0;
10346 }
10347 
10348 static void
10349 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10350 {
10351 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10352 
10353 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10354 	nfs4_free_lock_state(server, lsp);
10355 }
10356 
10357 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10358 		const nfs4_stateid *s2)
10359 {
10360 	if (s1->type != s2->type)
10361 		return false;
10362 
10363 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10364 		return false;
10365 
10366 	if (s1->seqid == s2->seqid)
10367 		return true;
10368 
10369 	return s1->seqid == 0 || s2->seqid == 0;
10370 }
10371 
10372 #endif /* CONFIG_NFS_V4_1 */
10373 
10374 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10375 		const nfs4_stateid *s2)
10376 {
10377 	return nfs4_stateid_match(s1, s2);
10378 }
10379 
10380 
10381 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10382 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10383 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10384 	.recover_open	= nfs4_open_reclaim,
10385 	.recover_lock	= nfs4_lock_reclaim,
10386 	.establish_clid = nfs4_init_clientid,
10387 	.detect_trunking = nfs40_discover_server_trunking,
10388 };
10389 
10390 #if defined(CONFIG_NFS_V4_1)
10391 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10392 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10393 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10394 	.recover_open	= nfs4_open_reclaim,
10395 	.recover_lock	= nfs4_lock_reclaim,
10396 	.establish_clid = nfs41_init_clientid,
10397 	.reclaim_complete = nfs41_proc_reclaim_complete,
10398 	.detect_trunking = nfs41_discover_server_trunking,
10399 };
10400 #endif /* CONFIG_NFS_V4_1 */
10401 
10402 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10403 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10404 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10405 	.recover_open	= nfs40_open_expired,
10406 	.recover_lock	= nfs4_lock_expired,
10407 	.establish_clid = nfs4_init_clientid,
10408 };
10409 
10410 #if defined(CONFIG_NFS_V4_1)
10411 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10412 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10413 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10414 	.recover_open	= nfs41_open_expired,
10415 	.recover_lock	= nfs41_lock_expired,
10416 	.establish_clid = nfs41_init_clientid,
10417 };
10418 #endif /* CONFIG_NFS_V4_1 */
10419 
10420 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10421 	.sched_state_renewal = nfs4_proc_async_renew,
10422 	.get_state_renewal_cred = nfs4_get_renew_cred,
10423 	.renew_lease = nfs4_proc_renew,
10424 };
10425 
10426 #if defined(CONFIG_NFS_V4_1)
10427 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10428 	.sched_state_renewal = nfs41_proc_async_sequence,
10429 	.get_state_renewal_cred = nfs4_get_machine_cred,
10430 	.renew_lease = nfs4_proc_sequence,
10431 };
10432 #endif
10433 
10434 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10435 	.get_locations = _nfs40_proc_get_locations,
10436 	.fsid_present = _nfs40_proc_fsid_present,
10437 };
10438 
10439 #if defined(CONFIG_NFS_V4_1)
10440 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10441 	.get_locations = _nfs41_proc_get_locations,
10442 	.fsid_present = _nfs41_proc_fsid_present,
10443 };
10444 #endif	/* CONFIG_NFS_V4_1 */
10445 
10446 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10447 	.minor_version = 0,
10448 	.init_caps = NFS_CAP_READDIRPLUS
10449 		| NFS_CAP_ATOMIC_OPEN
10450 		| NFS_CAP_POSIX_LOCK,
10451 	.init_client = nfs40_init_client,
10452 	.shutdown_client = nfs40_shutdown_client,
10453 	.match_stateid = nfs4_match_stateid,
10454 	.find_root_sec = nfs4_find_root_sec,
10455 	.free_lock_state = nfs4_release_lockowner,
10456 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10457 	.alloc_seqid = nfs_alloc_seqid,
10458 	.call_sync_ops = &nfs40_call_sync_ops,
10459 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10460 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10461 	.state_renewal_ops = &nfs40_state_renewal_ops,
10462 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10463 };
10464 
10465 #if defined(CONFIG_NFS_V4_1)
10466 static struct nfs_seqid *
10467 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10468 {
10469 	return NULL;
10470 }
10471 
10472 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10473 	.minor_version = 1,
10474 	.init_caps = NFS_CAP_READDIRPLUS
10475 		| NFS_CAP_ATOMIC_OPEN
10476 		| NFS_CAP_POSIX_LOCK
10477 		| NFS_CAP_STATEID_NFSV41
10478 		| NFS_CAP_ATOMIC_OPEN_V1
10479 		| NFS_CAP_LGOPEN
10480 		| NFS_CAP_MOVEABLE,
10481 	.init_client = nfs41_init_client,
10482 	.shutdown_client = nfs41_shutdown_client,
10483 	.match_stateid = nfs41_match_stateid,
10484 	.find_root_sec = nfs41_find_root_sec,
10485 	.free_lock_state = nfs41_free_lock_state,
10486 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10487 	.alloc_seqid = nfs_alloc_no_seqid,
10488 	.session_trunk = nfs4_test_session_trunk,
10489 	.call_sync_ops = &nfs41_call_sync_ops,
10490 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10491 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10492 	.state_renewal_ops = &nfs41_state_renewal_ops,
10493 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10494 };
10495 #endif
10496 
10497 #if defined(CONFIG_NFS_V4_2)
10498 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10499 	.minor_version = 2,
10500 	.init_caps = NFS_CAP_READDIRPLUS
10501 		| NFS_CAP_ATOMIC_OPEN
10502 		| NFS_CAP_POSIX_LOCK
10503 		| NFS_CAP_STATEID_NFSV41
10504 		| NFS_CAP_ATOMIC_OPEN_V1
10505 		| NFS_CAP_LGOPEN
10506 		| NFS_CAP_ALLOCATE
10507 		| NFS_CAP_COPY
10508 		| NFS_CAP_OFFLOAD_CANCEL
10509 		| NFS_CAP_COPY_NOTIFY
10510 		| NFS_CAP_DEALLOCATE
10511 		| NFS_CAP_SEEK
10512 		| NFS_CAP_LAYOUTSTATS
10513 		| NFS_CAP_CLONE
10514 		| NFS_CAP_LAYOUTERROR
10515 		| NFS_CAP_READ_PLUS
10516 		| NFS_CAP_MOVEABLE,
10517 	.init_client = nfs41_init_client,
10518 	.shutdown_client = nfs41_shutdown_client,
10519 	.match_stateid = nfs41_match_stateid,
10520 	.find_root_sec = nfs41_find_root_sec,
10521 	.free_lock_state = nfs41_free_lock_state,
10522 	.call_sync_ops = &nfs41_call_sync_ops,
10523 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10524 	.alloc_seqid = nfs_alloc_no_seqid,
10525 	.session_trunk = nfs4_test_session_trunk,
10526 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10527 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10528 	.state_renewal_ops = &nfs41_state_renewal_ops,
10529 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10530 };
10531 #endif
10532 
10533 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10534 	[0] = &nfs_v4_0_minor_ops,
10535 #if defined(CONFIG_NFS_V4_1)
10536 	[1] = &nfs_v4_1_minor_ops,
10537 #endif
10538 #if defined(CONFIG_NFS_V4_2)
10539 	[2] = &nfs_v4_2_minor_ops,
10540 #endif
10541 };
10542 
10543 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10544 {
10545 	ssize_t error, error2, error3;
10546 
10547 	error = generic_listxattr(dentry, list, size);
10548 	if (error < 0)
10549 		return error;
10550 	if (list) {
10551 		list += error;
10552 		size -= error;
10553 	}
10554 
10555 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10556 	if (error2 < 0)
10557 		return error2;
10558 
10559 	if (list) {
10560 		list += error2;
10561 		size -= error2;
10562 	}
10563 
10564 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10565 	if (error3 < 0)
10566 		return error3;
10567 
10568 	return error + error2 + error3;
10569 }
10570 
10571 static void nfs4_enable_swap(struct inode *inode)
10572 {
10573 	/* The state manager thread must always be running.
10574 	 * It will notice the client is a swapper, and stay put.
10575 	 */
10576 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10577 
10578 	nfs4_schedule_state_manager(clp);
10579 }
10580 
10581 static void nfs4_disable_swap(struct inode *inode)
10582 {
10583 	/* The state manager thread will now exit once it is
10584 	 * woken.
10585 	 */
10586 	wake_up_var(&NFS_SERVER(inode)->nfs_client->cl_state);
10587 }
10588 
10589 static const struct inode_operations nfs4_dir_inode_operations = {
10590 	.create		= nfs_create,
10591 	.lookup		= nfs_lookup,
10592 	.atomic_open	= nfs_atomic_open,
10593 	.link		= nfs_link,
10594 	.unlink		= nfs_unlink,
10595 	.symlink	= nfs_symlink,
10596 	.mkdir		= nfs_mkdir,
10597 	.rmdir		= nfs_rmdir,
10598 	.mknod		= nfs_mknod,
10599 	.rename		= nfs_rename,
10600 	.permission	= nfs_permission,
10601 	.getattr	= nfs_getattr,
10602 	.setattr	= nfs_setattr,
10603 	.listxattr	= nfs4_listxattr,
10604 };
10605 
10606 static const struct inode_operations nfs4_file_inode_operations = {
10607 	.permission	= nfs_permission,
10608 	.getattr	= nfs_getattr,
10609 	.setattr	= nfs_setattr,
10610 	.listxattr	= nfs4_listxattr,
10611 };
10612 
10613 const struct nfs_rpc_ops nfs_v4_clientops = {
10614 	.version	= 4,			/* protocol version */
10615 	.dentry_ops	= &nfs4_dentry_operations,
10616 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10617 	.file_inode_ops	= &nfs4_file_inode_operations,
10618 	.file_ops	= &nfs4_file_operations,
10619 	.getroot	= nfs4_proc_get_root,
10620 	.submount	= nfs4_submount,
10621 	.try_get_tree	= nfs4_try_get_tree,
10622 	.getattr	= nfs4_proc_getattr,
10623 	.setattr	= nfs4_proc_setattr,
10624 	.lookup		= nfs4_proc_lookup,
10625 	.lookupp	= nfs4_proc_lookupp,
10626 	.access		= nfs4_proc_access,
10627 	.readlink	= nfs4_proc_readlink,
10628 	.create		= nfs4_proc_create,
10629 	.remove		= nfs4_proc_remove,
10630 	.unlink_setup	= nfs4_proc_unlink_setup,
10631 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10632 	.unlink_done	= nfs4_proc_unlink_done,
10633 	.rename_setup	= nfs4_proc_rename_setup,
10634 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10635 	.rename_done	= nfs4_proc_rename_done,
10636 	.link		= nfs4_proc_link,
10637 	.symlink	= nfs4_proc_symlink,
10638 	.mkdir		= nfs4_proc_mkdir,
10639 	.rmdir		= nfs4_proc_rmdir,
10640 	.readdir	= nfs4_proc_readdir,
10641 	.mknod		= nfs4_proc_mknod,
10642 	.statfs		= nfs4_proc_statfs,
10643 	.fsinfo		= nfs4_proc_fsinfo,
10644 	.pathconf	= nfs4_proc_pathconf,
10645 	.set_capabilities = nfs4_server_capabilities,
10646 	.decode_dirent	= nfs4_decode_dirent,
10647 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10648 	.read_setup	= nfs4_proc_read_setup,
10649 	.read_done	= nfs4_read_done,
10650 	.write_setup	= nfs4_proc_write_setup,
10651 	.write_done	= nfs4_write_done,
10652 	.commit_setup	= nfs4_proc_commit_setup,
10653 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10654 	.commit_done	= nfs4_commit_done,
10655 	.lock		= nfs4_proc_lock,
10656 	.clear_acl_cache = nfs4_zap_acl_attr,
10657 	.close_context  = nfs4_close_context,
10658 	.open_context	= nfs4_atomic_open,
10659 	.have_delegation = nfs4_have_delegation,
10660 	.alloc_client	= nfs4_alloc_client,
10661 	.init_client	= nfs4_init_client,
10662 	.free_client	= nfs4_free_client,
10663 	.create_server	= nfs4_create_server,
10664 	.clone_server	= nfs_clone_server,
10665 	.discover_trunking = nfs4_discover_trunking,
10666 	.enable_swap	= nfs4_enable_swap,
10667 	.disable_swap	= nfs4_disable_swap,
10668 };
10669 
10670 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10671 	.name	= XATTR_NAME_NFSV4_ACL,
10672 	.list	= nfs4_xattr_list_nfs4_acl,
10673 	.get	= nfs4_xattr_get_nfs4_acl,
10674 	.set	= nfs4_xattr_set_nfs4_acl,
10675 };
10676 
10677 #if defined(CONFIG_NFS_V4_1)
10678 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10679 	.name	= XATTR_NAME_NFSV4_DACL,
10680 	.list	= nfs4_xattr_list_nfs4_dacl,
10681 	.get	= nfs4_xattr_get_nfs4_dacl,
10682 	.set	= nfs4_xattr_set_nfs4_dacl,
10683 };
10684 
10685 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10686 	.name	= XATTR_NAME_NFSV4_SACL,
10687 	.list	= nfs4_xattr_list_nfs4_sacl,
10688 	.get	= nfs4_xattr_get_nfs4_sacl,
10689 	.set	= nfs4_xattr_set_nfs4_sacl,
10690 };
10691 #endif
10692 
10693 #ifdef CONFIG_NFS_V4_2
10694 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10695 	.prefix	= XATTR_USER_PREFIX,
10696 	.get	= nfs4_xattr_get_nfs4_user,
10697 	.set	= nfs4_xattr_set_nfs4_user,
10698 };
10699 #endif
10700 
10701 const struct xattr_handler *nfs4_xattr_handlers[] = {
10702 	&nfs4_xattr_nfs4_acl_handler,
10703 #if defined(CONFIG_NFS_V4_1)
10704 	&nfs4_xattr_nfs4_dacl_handler,
10705 	&nfs4_xattr_nfs4_sacl_handler,
10706 #endif
10707 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10708 	&nfs4_xattr_nfs4_label_handler,
10709 #endif
10710 #ifdef CONFIG_NFS_V4_2
10711 	&nfs4_xattr_nfs4_user_handler,
10712 #endif
10713 	NULL
10714 };
10715