xref: /linux/fs/nfs/nfs4proc.c (revision 840ef6c78e6a2f694b578ecb9063241c992aaa9e)
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 "nfs40.h"
71 #include "nfs42.h"
72 
73 #include "nfs4trace.h"
74 
75 #define NFSDBG_FACILITY		NFSDBG_PROC
76 
77 #define NFS4_BITMASK_SZ		3
78 
79 #define NFS4_POLL_RETRY_MIN	(HZ/10)
80 #define NFS4_POLL_RETRY_MAX	(15*HZ)
81 
82 /* file attributes which can be mapped to nfs attributes */
83 #define NFS4_VALID_ATTRS (ATTR_MODE \
84 	| ATTR_UID \
85 	| ATTR_GID \
86 	| ATTR_SIZE \
87 	| ATTR_ATIME \
88 	| ATTR_MTIME \
89 	| ATTR_CTIME \
90 	| ATTR_ATIME_SET \
91 	| ATTR_MTIME_SET)
92 
93 struct nfs4_opendata;
94 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
95 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
96 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
97 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
98 			      struct nfs_fattr *fattr, struct inode *inode);
99 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
100 			    struct nfs_fattr *fattr, struct iattr *sattr,
101 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel);
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 *, const nfs4_stateid *,
107 			      const struct cred *);
108 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
109 			      const struct cred *, bool);
110 
111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
112 static inline struct nfs4_label *
113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
114 	struct iattr *sattr, struct nfs4_label *label)
115 {
116 	struct lsm_context shim;
117 	int err;
118 
119 	if (label == NULL)
120 		return NULL;
121 
122 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
123 		return NULL;
124 
125 	label->lfs = 0;
126 	label->pi = 0;
127 	label->len = 0;
128 	label->label = NULL;
129 
130 	err = security_dentry_init_security(dentry, sattr->ia_mode,
131 				&dentry->d_name, NULL, &shim);
132 	if (err)
133 		return NULL;
134 
135 	label->lsmid = shim.id;
136 	label->label = shim.context;
137 	label->len = shim.len;
138 	return label;
139 }
140 static inline void
141 nfs4_label_release_security(struct nfs4_label *label)
142 {
143 	struct lsm_context shim;
144 
145 	if (label) {
146 		shim.context = label->label;
147 		shim.len = label->len;
148 		shim.id = label->lsmid;
149 		security_release_secctx(&shim);
150 	}
151 }
152 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
153 {
154 	if (label)
155 		return server->attr_bitmask;
156 
157 	return server->attr_bitmask_nl;
158 }
159 #else
160 static inline struct nfs4_label *
161 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
162 	struct iattr *sattr, struct nfs4_label *l)
163 { return NULL; }
164 static inline void
165 nfs4_label_release_security(struct nfs4_label *label)
166 { return; }
167 static inline u32 *
168 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
169 { return server->attr_bitmask; }
170 #endif
171 
172 /* Prevent leaks of NFSv4 errors into userland */
173 static int nfs4_map_errors(int err)
174 {
175 	if (err >= -1000)
176 		return err;
177 	switch (err) {
178 	case -NFS4ERR_RESOURCE:
179 	case -NFS4ERR_LAYOUTTRYLATER:
180 	case -NFS4ERR_RECALLCONFLICT:
181 	case -NFS4ERR_RETURNCONFLICT:
182 		return -EREMOTEIO;
183 	case -NFS4ERR_WRONGSEC:
184 	case -NFS4ERR_WRONG_CRED:
185 		return -EPERM;
186 	case -NFS4ERR_BADOWNER:
187 	case -NFS4ERR_BADNAME:
188 		return -EINVAL;
189 	case -NFS4ERR_SHARE_DENIED:
190 		return -EACCES;
191 	case -NFS4ERR_MINOR_VERS_MISMATCH:
192 		return -EPROTONOSUPPORT;
193 	case -NFS4ERR_FILE_OPEN:
194 		return -EBUSY;
195 	case -NFS4ERR_NOT_SAME:
196 		return -ENOTSYNC;
197 	case -ENETDOWN:
198 	case -ENETUNREACH:
199 		break;
200 	default:
201 		dprintk("%s could not handle NFSv4 error %d\n",
202 				__func__, -err);
203 		break;
204 	}
205 	return -EIO;
206 }
207 
208 /*
209  * This is our standard bitmap for GETATTR requests.
210  */
211 const u32 nfs4_fattr_bitmap[3] = {
212 	FATTR4_WORD0_TYPE
213 	| FATTR4_WORD0_CHANGE
214 	| FATTR4_WORD0_SIZE
215 	| FATTR4_WORD0_FSID
216 	| FATTR4_WORD0_FILEID,
217 	FATTR4_WORD1_MODE
218 	| FATTR4_WORD1_NUMLINKS
219 	| FATTR4_WORD1_OWNER
220 	| FATTR4_WORD1_OWNER_GROUP
221 	| FATTR4_WORD1_RAWDEV
222 	| FATTR4_WORD1_SPACE_USED
223 	| FATTR4_WORD1_TIME_ACCESS
224 	| FATTR4_WORD1_TIME_CREATE
225 	| FATTR4_WORD1_TIME_METADATA
226 	| FATTR4_WORD1_TIME_MODIFY
227 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
228 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
229 	FATTR4_WORD2_SECURITY_LABEL
230 #endif
231 };
232 
233 static const u32 nfs4_pnfs_open_bitmap[3] = {
234 	FATTR4_WORD0_TYPE
235 	| FATTR4_WORD0_CHANGE
236 	| FATTR4_WORD0_SIZE
237 	| FATTR4_WORD0_FSID
238 	| FATTR4_WORD0_FILEID,
239 	FATTR4_WORD1_MODE
240 	| FATTR4_WORD1_NUMLINKS
241 	| FATTR4_WORD1_OWNER
242 	| FATTR4_WORD1_OWNER_GROUP
243 	| FATTR4_WORD1_RAWDEV
244 	| FATTR4_WORD1_SPACE_USED
245 	| FATTR4_WORD1_TIME_ACCESS
246 	| FATTR4_WORD1_TIME_CREATE
247 	| FATTR4_WORD1_TIME_METADATA
248 	| FATTR4_WORD1_TIME_MODIFY,
249 	FATTR4_WORD2_MDSTHRESHOLD
250 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
251 	| FATTR4_WORD2_SECURITY_LABEL
252 #endif
253 };
254 
255 static const u32 nfs4_open_noattr_bitmap[3] = {
256 	FATTR4_WORD0_TYPE
257 	| FATTR4_WORD0_FILEID,
258 };
259 
260 const u32 nfs4_statfs_bitmap[3] = {
261 	FATTR4_WORD0_FILES_AVAIL
262 	| FATTR4_WORD0_FILES_FREE
263 	| FATTR4_WORD0_FILES_TOTAL,
264 	FATTR4_WORD1_SPACE_AVAIL
265 	| FATTR4_WORD1_SPACE_FREE
266 	| FATTR4_WORD1_SPACE_TOTAL
267 };
268 
269 const u32 nfs4_pathconf_bitmap[3] = {
270 	FATTR4_WORD0_MAXLINK
271 	| FATTR4_WORD0_MAXNAME,
272 	0
273 };
274 
275 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
276 			| FATTR4_WORD0_MAXREAD
277 			| FATTR4_WORD0_MAXWRITE
278 			| FATTR4_WORD0_LEASE_TIME,
279 			FATTR4_WORD1_TIME_DELTA
280 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
281 			FATTR4_WORD2_LAYOUT_BLKSIZE
282 			| FATTR4_WORD2_CLONE_BLKSIZE
283 			| FATTR4_WORD2_CHANGE_ATTR_TYPE
284 			| FATTR4_WORD2_XATTR_SUPPORT
285 };
286 
287 const u32 nfs4_fs_locations_bitmap[3] = {
288 	FATTR4_WORD0_CHANGE
289 	| FATTR4_WORD0_SIZE
290 	| FATTR4_WORD0_FSID
291 	| FATTR4_WORD0_FILEID
292 	| FATTR4_WORD0_FS_LOCATIONS,
293 	FATTR4_WORD1_OWNER
294 	| FATTR4_WORD1_OWNER_GROUP
295 	| FATTR4_WORD1_RAWDEV
296 	| FATTR4_WORD1_SPACE_USED
297 	| FATTR4_WORD1_TIME_ACCESS
298 	| FATTR4_WORD1_TIME_METADATA
299 	| FATTR4_WORD1_TIME_MODIFY
300 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
301 };
302 
303 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
304 				    struct inode *inode, unsigned long flags)
305 {
306 	unsigned long cache_validity;
307 
308 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
309 	if (!inode || !nfs_have_read_or_write_delegation(inode))
310 		return;
311 
312 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
313 
314 	/* Remove the attributes over which we have full control */
315 	dst[1] &= ~FATTR4_WORD1_RAWDEV;
316 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
317 		dst[0] &= ~FATTR4_WORD0_SIZE;
318 
319 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
320 		dst[0] &= ~FATTR4_WORD0_CHANGE;
321 
322 	if (!(cache_validity & NFS_INO_INVALID_MODE))
323 		dst[1] &= ~FATTR4_WORD1_MODE;
324 	if (!(cache_validity & NFS_INO_INVALID_OTHER))
325 		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
326 
327 	if (!(cache_validity & NFS_INO_INVALID_BTIME))
328 		dst[1] &= ~FATTR4_WORD1_TIME_CREATE;
329 
330 	if (nfs_have_delegated_mtime(inode)) {
331 		if (!(cache_validity & NFS_INO_INVALID_ATIME))
332 			dst[1] &= ~(FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET);
333 		if (!(cache_validity & NFS_INO_INVALID_MTIME))
334 			dst[1] &= ~(FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET);
335 		if (!(cache_validity & NFS_INO_INVALID_CTIME))
336 			dst[1] &= ~(FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY_SET);
337 	} else if (nfs_have_delegated_atime(inode)) {
338 		if (!(cache_validity & NFS_INO_INVALID_ATIME))
339 			dst[1] &= ~(FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET);
340 	}
341 }
342 
343 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
344 		struct nfs4_readdir_arg *readdir)
345 {
346 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
347 	__be32 *start, *p;
348 
349 	if (cookie > 2) {
350 		readdir->cookie = cookie;
351 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
352 		return;
353 	}
354 
355 	readdir->cookie = 0;
356 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
357 	if (cookie == 2)
358 		return;
359 
360 	/*
361 	 * NFSv4 servers do not return entries for '.' and '..'
362 	 * Therefore, we fake these entries here.  We let '.'
363 	 * have cookie 0 and '..' have cookie 1.  Note that
364 	 * when talking to the server, we always send cookie 0
365 	 * instead of 1 or 2.
366 	 */
367 	start = p = kmap_atomic(*readdir->pages);
368 
369 	if (cookie == 0) {
370 		*p++ = xdr_one;                                  /* next */
371 		*p++ = xdr_zero;                   /* cookie, first word */
372 		*p++ = xdr_one;                   /* cookie, second word */
373 		*p++ = xdr_one;                             /* entry len */
374 		memcpy(p, ".\0\0\0", 4);                        /* entry */
375 		p++;
376 		*p++ = xdr_one;                         /* bitmap length */
377 		*p++ = htonl(attrs);                           /* bitmap */
378 		*p++ = htonl(12);             /* attribute buffer length */
379 		*p++ = htonl(NF4DIR);
380 		p = xdr_encode_hyper(p, d_inode(dentry)->i_ino);
381 	}
382 
383 	*p++ = xdr_one;                                  /* next */
384 	*p++ = xdr_zero;                   /* cookie, first word */
385 	*p++ = xdr_two;                   /* cookie, second word */
386 	*p++ = xdr_two;                             /* entry len */
387 	memcpy(p, "..\0\0", 4);                         /* entry */
388 	p++;
389 	*p++ = xdr_one;                         /* bitmap length */
390 	*p++ = htonl(attrs);                           /* bitmap */
391 	*p++ = htonl(12);             /* attribute buffer length */
392 	*p++ = htonl(NF4DIR);
393 	spin_lock(&dentry->d_lock);
394 	p = xdr_encode_hyper(p, d_inode(dentry->d_parent)->i_ino);
395 	spin_unlock(&dentry->d_lock);
396 
397 	readdir->pgbase = (char *)p - (char *)start;
398 	readdir->count -= readdir->pgbase;
399 	kunmap_atomic(start);
400 }
401 
402 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
403 {
404 	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
405 		fattr->pre_change_attr = version;
406 		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
407 	}
408 }
409 
410 static void nfs4_test_and_free_stateid(struct nfs_server *server,
411 		nfs4_stateid *stateid,
412 		const struct cred *cred)
413 {
414 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
415 
416 	ops->test_and_free_expired(server, stateid, cred);
417 }
418 
419 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
420 		nfs4_stateid *stateid,
421 		const struct cred *cred)
422 {
423 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
424 	nfs4_test_and_free_stateid(server, stateid, cred);
425 }
426 
427 static void nfs4_free_revoked_stateid(struct nfs_server *server,
428 		const nfs4_stateid *stateid,
429 		const struct cred *cred)
430 {
431 	nfs4_stateid tmp;
432 
433 	nfs4_stateid_copy(&tmp, stateid);
434 	__nfs4_free_revoked_stateid(server, &tmp, cred);
435 }
436 
437 static long nfs4_update_delay(long *timeout)
438 {
439 	long ret;
440 	if (!timeout)
441 		return NFS4_POLL_RETRY_MAX;
442 	if (*timeout <= 0)
443 		*timeout = NFS4_POLL_RETRY_MIN;
444 	if (*timeout > NFS4_POLL_RETRY_MAX)
445 		*timeout = NFS4_POLL_RETRY_MAX;
446 	ret = *timeout;
447 	*timeout <<= 1;
448 	return ret;
449 }
450 
451 static int nfs4_delay_killable(long *timeout)
452 {
453 	might_sleep();
454 
455 	if (unlikely(nfs_current_task_exiting()))
456 		return -EINTR;
457 	__set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
458 	schedule_timeout(nfs4_update_delay(timeout));
459 	if (!__fatal_signal_pending(current))
460 		return 0;
461 	return -EINTR;
462 }
463 
464 static int nfs4_delay_interruptible(long *timeout)
465 {
466 	might_sleep();
467 
468 	if (unlikely(nfs_current_task_exiting()))
469 		return -EINTR;
470 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
471 	schedule_timeout(nfs4_update_delay(timeout));
472 	if (!signal_pending(current))
473 		return 0;
474 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
475 }
476 
477 static int nfs4_delay(long *timeout, bool interruptible)
478 {
479 	if (interruptible)
480 		return nfs4_delay_interruptible(timeout);
481 	return nfs4_delay_killable(timeout);
482 }
483 
484 static const nfs4_stateid *
485 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
486 {
487 	if (!stateid)
488 		return NULL;
489 	switch (stateid->type) {
490 	case NFS4_OPEN_STATEID_TYPE:
491 	case NFS4_LOCK_STATEID_TYPE:
492 	case NFS4_DELEGATION_STATEID_TYPE:
493 		return stateid;
494 	default:
495 		break;
496 	}
497 	return NULL;
498 }
499 
500 /* This is the error handling routine for processes that are allowed
501  * to sleep.
502  */
503 static int nfs4_do_handle_exception(struct nfs_server *server,
504 		int errorcode, struct nfs4_exception *exception)
505 {
506 	struct nfs_client *clp = server->nfs_client;
507 	struct nfs4_state *state = exception->state;
508 	const nfs4_stateid *stateid;
509 	struct inode *inode = exception->inode;
510 	int ret = errorcode;
511 
512 	exception->delay = 0;
513 	exception->recovering = 0;
514 	exception->retry = 0;
515 
516 	stateid = nfs4_recoverable_stateid(exception->stateid);
517 	if (stateid == NULL && state != NULL)
518 		stateid = nfs4_recoverable_stateid(&state->stateid);
519 
520 	switch(errorcode) {
521 		case 0:
522 			return 0;
523 		case -NFS4ERR_BADHANDLE:
524 		case -ESTALE:
525 			if (inode != NULL && S_ISREG(inode->i_mode))
526 				pnfs_destroy_layout(NFS_I(inode));
527 			break;
528 		case -NFS4ERR_DELEG_REVOKED:
529 		case -NFS4ERR_ADMIN_REVOKED:
530 		case -NFS4ERR_EXPIRED:
531 		case -NFS4ERR_BAD_STATEID:
532 		case -NFS4ERR_PARTNER_NO_AUTH:
533 			if (inode != NULL && stateid != NULL) {
534 				nfs_inode_find_state_and_recover(inode,
535 						stateid);
536 				goto wait_on_recovery;
537 			}
538 			fallthrough;
539 		case -NFS4ERR_OPENMODE:
540 			if (inode) {
541 				int err;
542 
543 				err = nfs_async_inode_return_delegation(inode,
544 						stateid);
545 				if (err == 0)
546 					goto wait_on_recovery;
547 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
548 					exception->retry = 1;
549 					break;
550 				}
551 			}
552 			if (state == NULL)
553 				break;
554 			ret = nfs4_schedule_stateid_recovery(server, state);
555 			if (ret < 0)
556 				break;
557 			goto wait_on_recovery;
558 		case -NFS4ERR_STALE_STATEID:
559 		case -NFS4ERR_STALE_CLIENTID:
560 			nfs4_schedule_lease_recovery(clp);
561 			goto wait_on_recovery;
562 		case -NFS4ERR_MOVED:
563 			ret = nfs4_schedule_migration_recovery(server);
564 			if (ret < 0)
565 				break;
566 			goto wait_on_recovery;
567 		case -NFS4ERR_LEASE_MOVED:
568 			nfs4_schedule_lease_moved_recovery(clp);
569 			goto wait_on_recovery;
570 		case -NFS4ERR_BADSESSION:
571 		case -NFS4ERR_BADSLOT:
572 		case -NFS4ERR_BAD_HIGH_SLOT:
573 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
574 		case -NFS4ERR_DEADSESSION:
575 		case -NFS4ERR_SEQ_FALSE_RETRY:
576 		case -NFS4ERR_SEQ_MISORDERED:
577 			/* Handled in nfs41_sequence_process() */
578 			goto wait_on_recovery;
579 		case -NFS4ERR_FILE_OPEN:
580 			if (exception->timeout > HZ) {
581 				/* We have retried a decent amount, time to
582 				 * fail
583 				 */
584 				ret = -EBUSY;
585 				break;
586 			}
587 			fallthrough;
588 		case -NFS4ERR_DELAY:
589 			nfs_inc_server_stats(server, NFSIOS_DELAY);
590 			fallthrough;
591 		case -NFS4ERR_GRACE:
592 		case -NFS4ERR_LAYOUTTRYLATER:
593 		case -NFS4ERR_RECALLCONFLICT:
594 		case -NFS4ERR_RETURNCONFLICT:
595 			exception->delay = 1;
596 			return 0;
597 
598 		case -NFS4ERR_RETRY_UNCACHED_REP:
599 		case -NFS4ERR_OLD_STATEID:
600 			exception->retry = 1;
601 			break;
602 		case -NFS4ERR_BADOWNER:
603 			/* The following works around a Linux server bug! */
604 		case -NFS4ERR_BADNAME:
605 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
606 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
607 				exception->retry = 1;
608 				printk(KERN_WARNING "NFS: v4 server %s "
609 						"does not accept raw "
610 						"uid/gids. "
611 						"Reenabling the idmapper.\n",
612 						server->nfs_client->cl_hostname);
613 			}
614 	}
615 	/* We failed to handle the error */
616 	return nfs4_map_errors(ret);
617 wait_on_recovery:
618 	exception->recovering = 1;
619 	return 0;
620 }
621 
622 /*
623  * Track the number of NFS4ERR_DELAY related retransmissions and return
624  * EAGAIN if the 'softerr' mount option is set, and we've exceeded the limit
625  * set by 'nfs_delay_retrans'.
626  */
627 static int nfs4_exception_should_retrans(const struct nfs_server *server,
628 					 struct nfs4_exception *exception)
629 {
630 	if (server->flags & NFS_MOUNT_SOFTERR && nfs_delay_retrans >= 0) {
631 		if (exception->retrans++ >= (unsigned short)nfs_delay_retrans)
632 			return -EAGAIN;
633 	}
634 	return 0;
635 }
636 
637 /* This is the error handling routine for processes that are allowed
638  * to sleep.
639  */
640 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
641 {
642 	struct nfs_client *clp = server->nfs_client;
643 	int ret;
644 
645 	ret = nfs4_do_handle_exception(server, errorcode, exception);
646 	if (exception->delay) {
647 		int ret2 = nfs4_exception_should_retrans(server, exception);
648 		if (ret2 < 0) {
649 			exception->retry = 0;
650 			return ret2;
651 		}
652 		ret = nfs4_delay(&exception->timeout,
653 				exception->interruptible);
654 		goto out_retry;
655 	}
656 	if (exception->recovering) {
657 		if (exception->task_is_privileged)
658 			return -EDEADLOCK;
659 		ret = nfs4_wait_clnt_recover(clp);
660 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
661 			return -EIO;
662 		goto out_retry;
663 	}
664 	return ret;
665 out_retry:
666 	if (ret == 0)
667 		exception->retry = 1;
668 	return ret;
669 }
670 
671 static int
672 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
673 		int errorcode, struct nfs4_exception *exception)
674 {
675 	struct nfs_client *clp = server->nfs_client;
676 	int ret;
677 
678 	if ((task->tk_rpc_status == -ENETDOWN ||
679 	     task->tk_rpc_status == -ENETUNREACH) &&
680 	    task->tk_flags & RPC_TASK_NETUNREACH_FATAL) {
681 		exception->delay = 0;
682 		exception->recovering = 0;
683 		exception->retry = 0;
684 		return -EIO;
685 	}
686 
687 	ret = nfs4_do_handle_exception(server, errorcode, exception);
688 	if (exception->delay) {
689 		int ret2 = nfs4_exception_should_retrans(server, exception);
690 		if (ret2 < 0) {
691 			exception->retry = 0;
692 			return ret2;
693 		}
694 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
695 		goto out_retry;
696 	}
697 	if (exception->recovering) {
698 		if (exception->task_is_privileged)
699 			return -EDEADLOCK;
700 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
701 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
702 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
703 		goto out_retry;
704 	}
705 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
706 		ret = -EIO;
707 	return ret;
708 out_retry:
709 	if (ret == 0) {
710 		exception->retry = 1;
711 		/*
712 		 * For NFS4ERR_MOVED, the client transport will need to
713 		 * be recomputed after migration recovery has completed.
714 		 */
715 		if (errorcode == -NFS4ERR_MOVED)
716 			rpc_task_release_transport(task);
717 	}
718 	return ret;
719 }
720 
721 int
722 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
723 			struct nfs4_state *state, long *timeout)
724 {
725 	struct nfs4_exception exception = {
726 		.state = state,
727 	};
728 
729 	if (task->tk_status >= 0)
730 		return 0;
731 	if (timeout)
732 		exception.timeout = *timeout;
733 	task->tk_status = nfs4_async_handle_exception(task, server,
734 			task->tk_status,
735 			&exception);
736 	if (exception.delay && timeout)
737 		*timeout = exception.timeout;
738 	if (exception.retry)
739 		return -EAGAIN;
740 	return 0;
741 }
742 
743 /*
744  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
745  * or 'false' otherwise.
746  */
747 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
748 {
749 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
750 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
751 }
752 
753 void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
754 {
755 	spin_lock(&clp->cl_lock);
756 	if (time_before(clp->cl_last_renewal,timestamp))
757 		clp->cl_last_renewal = timestamp;
758 	spin_unlock(&clp->cl_lock);
759 }
760 
761 void renew_lease(const struct nfs_server *server, unsigned long timestamp)
762 {
763 	struct nfs_client *clp = server->nfs_client;
764 
765 	if (!nfs4_has_session(clp))
766 		do_renew_lease(clp, timestamp);
767 }
768 
769 void nfs4_init_sequence(struct nfs_client *clp,
770 			struct nfs4_sequence_args *args,
771 			struct nfs4_sequence_res *res, int cache_reply,
772 			int privileged)
773 {
774 	args->sa_slot = NULL;
775 	args->sa_cache_this = cache_reply;
776 	args->sa_privileged = privileged;
777 
778 	res->sr_slot = NULL;
779 	res->sr_slot_ops = clp->cl_mvops->sequence_slot_ops;
780 }
781 
782 static void nfs41_release_slot(struct nfs4_slot *slot)
783 {
784 	struct nfs4_session *session;
785 	struct nfs4_slot_table *tbl;
786 	bool send_new_highest_used_slotid = false;
787 
788 	if (!slot)
789 		return;
790 	tbl = slot->table;
791 	session = tbl->session;
792 
793 	/* Bump the slot sequence number */
794 	if (slot->seq_done)
795 		slot->seq_nr++;
796 	slot->seq_done = 0;
797 
798 	spin_lock(&tbl->slot_tbl_lock);
799 	/* Be nice to the server: try to ensure that the last transmitted
800 	 * value for highest_user_slotid <= target_highest_slotid
801 	 */
802 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
803 		send_new_highest_used_slotid = true;
804 
805 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
806 		send_new_highest_used_slotid = false;
807 		goto out_unlock;
808 	}
809 	nfs4_free_slot(tbl, slot);
810 
811 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
812 		send_new_highest_used_slotid = false;
813 out_unlock:
814 	spin_unlock(&tbl->slot_tbl_lock);
815 	if (send_new_highest_used_slotid)
816 		nfs41_notify_server(session->clp);
817 	if (waitqueue_active(&tbl->slot_waitq))
818 		wake_up_all(&tbl->slot_waitq);
819 }
820 
821 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
822 {
823 	nfs41_release_slot(res->sr_slot);
824 	res->sr_slot = NULL;
825 }
826 
827 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
828 		u32 seqnr)
829 {
830 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
831 		slot->seq_nr_highest_sent = seqnr;
832 }
833 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
834 {
835 	nfs4_slot_sequence_record_sent(slot, seqnr);
836 	slot->seq_nr_last_acked = seqnr;
837 }
838 
839 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
840 				struct nfs4_slot *slot)
841 {
842 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
843 	if (!IS_ERR(task))
844 		rpc_put_task_async(task);
845 }
846 
847 static int nfs41_sequence_process(struct rpc_task *task,
848 		struct nfs4_sequence_res *res)
849 {
850 	struct nfs4_session *session;
851 	struct nfs4_slot *slot = res->sr_slot;
852 	struct nfs_client *clp;
853 	int status;
854 	int ret = 1;
855 
856 	if (slot == NULL)
857 		goto out_noaction;
858 	/* don't increment the sequence number if the task wasn't sent */
859 	if (!RPC_WAS_SENT(task) || slot->seq_done)
860 		goto out;
861 
862 	session = slot->table->session;
863 	clp = session->clp;
864 
865 	trace_nfs4_sequence_done(session, res);
866 
867 	status = res->sr_status;
868 	if (task->tk_status == -NFS4ERR_DEADSESSION)
869 		status = -NFS4ERR_DEADSESSION;
870 
871 	/* Check the SEQUENCE operation status */
872 	switch (status) {
873 	case 0:
874 		/* Mark this sequence number as having been acked */
875 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
876 		/* Update the slot's sequence and clientid lease timer */
877 		slot->seq_done = 1;
878 		do_renew_lease(clp, res->sr_timestamp);
879 		/* Check sequence flags */
880 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
881 				!!slot->privileged);
882 		nfs41_update_target_slotid(slot->table, slot, res);
883 		break;
884 	case 1:
885 		/*
886 		 * sr_status remains 1 if an RPC level error occurred.
887 		 * The server may or may not have processed the sequence
888 		 * operation..
889 		 */
890 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
891 		slot->seq_done = 1;
892 		goto out;
893 	case -NFS4ERR_DELAY:
894 		/* The server detected a resend of the RPC call and
895 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
896 		 * of RFC5661.
897 		 */
898 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
899 			__func__,
900 			slot->slot_nr,
901 			slot->seq_nr);
902 		goto out_retry;
903 	case -NFS4ERR_RETRY_UNCACHED_REP:
904 	case -NFS4ERR_SEQ_FALSE_RETRY:
905 		/*
906 		 * The server thinks we tried to replay a request.
907 		 * Retry the call after bumping the sequence ID.
908 		 */
909 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
910 		goto retry_new_seq;
911 	case -NFS4ERR_BADSLOT:
912 		/*
913 		 * The slot id we used was probably retired. Try again
914 		 * using a different slot id.
915 		 */
916 		if (slot->slot_nr < slot->table->target_highest_slotid)
917 			goto session_recover;
918 		goto retry_nowait;
919 	case -NFS4ERR_SEQ_MISORDERED:
920 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
921 		/*
922 		 * Were one or more calls using this slot interrupted?
923 		 * If the server never received the request, then our
924 		 * transmitted slot sequence number may be too high. However,
925 		 * if the server did receive the request then it might
926 		 * accidentally give us a reply with a mismatched operation.
927 		 * We can sort this out by sending a lone sequence operation
928 		 * to the server on the same slot.
929 		 */
930 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
931 			slot->seq_nr--;
932 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
933 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
934 				res->sr_slot = NULL;
935 			}
936 			goto retry_nowait;
937 		}
938 		/*
939 		 * RFC5661:
940 		 * A retry might be sent while the original request is
941 		 * still in progress on the replier. The replier SHOULD
942 		 * deal with the issue by returning NFS4ERR_DELAY as the
943 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
944 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
945 		 *
946 		 * Restart the search after a delay.
947 		 */
948 		slot->seq_nr = slot->seq_nr_highest_sent;
949 		goto out_retry;
950 	case -NFS4ERR_BADSESSION:
951 	case -NFS4ERR_DEADSESSION:
952 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
953 		goto session_recover;
954 	default:
955 		/* Just update the slot sequence no. */
956 		slot->seq_done = 1;
957 	}
958 out:
959 	/* The session may be reset by one of the error handlers. */
960 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
961 out_noaction:
962 	return ret;
963 session_recover:
964 	set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
965 	nfs4_schedule_session_recovery(session, status);
966 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
967 	nfs41_sequence_free_slot(res);
968 	goto out;
969 retry_new_seq:
970 	++slot->seq_nr;
971 retry_nowait:
972 	if (rpc_restart_call_prepare(task)) {
973 		nfs41_sequence_free_slot(res);
974 		task->tk_status = 0;
975 		ret = 0;
976 	}
977 	goto out;
978 out_retry:
979 	if (!rpc_restart_call(task))
980 		goto out;
981 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
982 	return 0;
983 }
984 
985 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
986 {
987 	if (!nfs41_sequence_process(task, res))
988 		return 0;
989 	if (res->sr_slot != NULL)
990 		nfs41_sequence_free_slot(res);
991 	return 1;
992 
993 }
994 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
995 
996 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
997 {
998 	struct nfs4_call_sync_data *data = calldata;
999 
1000 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
1001 
1002 	nfs4_setup_sequence(data->seq_server->nfs_client,
1003 			    data->seq_args, data->seq_res, task);
1004 }
1005 
1006 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
1007 {
1008 	struct nfs4_call_sync_data *data = calldata;
1009 
1010 	nfs41_sequence_done(task, data->seq_res);
1011 }
1012 
1013 static const struct rpc_call_ops nfs41_call_sync_ops = {
1014 	.rpc_call_prepare = nfs41_call_sync_prepare,
1015 	.rpc_call_done = nfs41_call_sync_done,
1016 };
1017 
1018 
1019 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1020 {
1021 	res->sr_timestamp = jiffies;
1022 	res->sr_status_flags = 0;
1023 	res->sr_status = 1;
1024 }
1025 
1026 static
1027 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1028 		struct nfs4_sequence_res *res,
1029 		struct nfs4_slot *slot)
1030 {
1031 	if (!slot)
1032 		return;
1033 	slot->privileged = args->sa_privileged ? 1 : 0;
1034 	args->sa_slot = slot;
1035 
1036 	res->sr_slot = slot;
1037 }
1038 
1039 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1040 {
1041 	if (res->sr_slot != NULL)
1042 		res->sr_slot_ops->free_slot(res);
1043 }
1044 
1045 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1046 {
1047 	if (res->sr_slot == NULL)
1048 		return 1;
1049 	return res->sr_slot_ops->process(task, res);
1050 }
1051 
1052 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
1053 {
1054 	if (res->sr_slot == NULL)
1055 		return 1;
1056 	return res->sr_slot_ops->done(task, res);
1057 }
1058 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1059 
1060 
1061 int nfs4_setup_sequence(struct nfs_client *client,
1062 			struct nfs4_sequence_args *args,
1063 			struct nfs4_sequence_res *res,
1064 			struct rpc_task *task)
1065 {
1066 	struct nfs4_session *session = nfs4_get_session(client);
1067 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1068 	struct nfs4_slot *slot;
1069 
1070 	/* slot already allocated? */
1071 	if (res->sr_slot != NULL)
1072 		goto out_start;
1073 
1074 	if (session)
1075 		tbl = &session->fc_slot_table;
1076 
1077 	spin_lock(&tbl->slot_tbl_lock);
1078 	/* The state manager will wait until the slot table is empty */
1079 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1080 		goto out_sleep;
1081 
1082 	slot = nfs4_alloc_slot(tbl);
1083 	if (IS_ERR(slot)) {
1084 		if (slot == ERR_PTR(-ENOMEM))
1085 			goto out_sleep_timeout;
1086 		goto out_sleep;
1087 	}
1088 	spin_unlock(&tbl->slot_tbl_lock);
1089 
1090 	nfs4_sequence_attach_slot(args, res, slot);
1091 
1092 	trace_nfs4_setup_sequence(session, args);
1093 out_start:
1094 	nfs41_sequence_res_init(res);
1095 	rpc_call_start(task);
1096 	return 0;
1097 out_sleep_timeout:
1098 	/* Try again in 1/4 second */
1099 	if (args->sa_privileged)
1100 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1101 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1102 	else
1103 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1104 				NULL, jiffies + (HZ >> 2));
1105 	spin_unlock(&tbl->slot_tbl_lock);
1106 	return -EAGAIN;
1107 out_sleep:
1108 	if (args->sa_privileged)
1109 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1110 				RPC_PRIORITY_PRIVILEGED);
1111 	else
1112 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1113 	spin_unlock(&tbl->slot_tbl_lock);
1114 	return -EAGAIN;
1115 }
1116 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1117 
1118 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1119 {
1120 	int ret;
1121 	struct rpc_task *task;
1122 
1123 	task = rpc_run_task(task_setup);
1124 	if (IS_ERR(task))
1125 		return PTR_ERR(task);
1126 
1127 	ret = task->tk_status;
1128 	rpc_put_task(task);
1129 	return ret;
1130 }
1131 
1132 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1133 			     struct nfs_server *server,
1134 			     struct rpc_message *msg,
1135 			     struct nfs4_sequence_args *args,
1136 			     struct nfs4_sequence_res *res,
1137 			     unsigned short task_flags)
1138 {
1139 	struct nfs_client *clp = server->nfs_client;
1140 	struct nfs4_call_sync_data data = {
1141 		.seq_server = server,
1142 		.seq_args = args,
1143 		.seq_res = res,
1144 	};
1145 	struct rpc_task_setup task_setup = {
1146 		.rpc_client = clnt,
1147 		.rpc_message = msg,
1148 		.callback_ops = clp->cl_mvops->call_sync_ops,
1149 		.callback_data = &data,
1150 		.flags = task_flags,
1151 	};
1152 
1153 	return nfs4_call_sync_custom(&task_setup);
1154 }
1155 
1156 int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1157 			    struct nfs_server *server,
1158 			    struct rpc_message *msg,
1159 			    struct nfs4_sequence_args *args,
1160 			    struct nfs4_sequence_res *res)
1161 {
1162 	unsigned short task_flags = 0;
1163 
1164 	if (server->caps & NFS_CAP_MOVEABLE)
1165 		task_flags = RPC_TASK_MOVEABLE;
1166 	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1167 }
1168 
1169 
1170 int nfs4_call_sync(struct rpc_clnt *clnt,
1171 		   struct nfs_server *server,
1172 		   struct rpc_message *msg,
1173 		   struct nfs4_sequence_args *args,
1174 		   struct nfs4_sequence_res *res,
1175 		   int cache_reply)
1176 {
1177 	nfs4_init_sequence(server->nfs_client, args, res, cache_reply, 0);
1178 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1179 }
1180 
1181 static void
1182 nfs4_inc_nlink_locked(struct inode *inode)
1183 {
1184 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1185 					     NFS_INO_INVALID_CTIME |
1186 					     NFS_INO_INVALID_NLINK);
1187 	inc_nlink(inode);
1188 }
1189 
1190 static void
1191 nfs4_inc_nlink(struct inode *inode)
1192 {
1193 	spin_lock(&inode->i_lock);
1194 	nfs4_inc_nlink_locked(inode);
1195 	spin_unlock(&inode->i_lock);
1196 }
1197 
1198 static void
1199 nfs4_dec_nlink_locked(struct inode *inode)
1200 {
1201 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1202 					     NFS_INO_INVALID_CTIME |
1203 					     NFS_INO_INVALID_NLINK);
1204 	drop_nlink(inode);
1205 }
1206 
1207 static void
1208 nfs4_update_changeattr_locked(struct inode *inode,
1209 		struct nfs4_change_info *cinfo,
1210 		unsigned long timestamp, unsigned long cache_validity)
1211 {
1212 	struct nfs_inode *nfsi = NFS_I(inode);
1213 	u64 change_attr = inode_peek_iversion_raw(inode);
1214 
1215 	if (!nfs_have_delegated_mtime(inode))
1216 		cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1217 	if (S_ISDIR(inode->i_mode))
1218 		cache_validity |= NFS_INO_INVALID_DATA;
1219 
1220 	switch (NFS_SERVER(inode)->change_attr_type) {
1221 	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1222 		if (cinfo->after == change_attr)
1223 			goto out;
1224 		break;
1225 	default:
1226 		if ((s64)(change_attr - cinfo->after) >= 0)
1227 			goto out;
1228 	}
1229 
1230 	inode_set_iversion_raw(inode, cinfo->after);
1231 	if (!cinfo->atomic || cinfo->before != change_attr) {
1232 		if (S_ISDIR(inode->i_mode))
1233 			nfs_force_lookup_revalidate(inode);
1234 
1235 		if (!nfs_have_delegated_attributes(inode))
1236 			cache_validity |=
1237 				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1238 				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1239 				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1240 				NFS_INO_INVALID_MODE | NFS_INO_INVALID_BTIME |
1241 				NFS_INO_INVALID_XATTR;
1242 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1243 	}
1244 	nfsi->attrtimeo_timestamp = jiffies;
1245 	nfsi->read_cache_jiffies = timestamp;
1246 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1247 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1248 out:
1249 	nfs_set_cache_invalid(inode, cache_validity);
1250 }
1251 
1252 void
1253 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1254 		unsigned long timestamp, unsigned long cache_validity)
1255 {
1256 	spin_lock(&dir->i_lock);
1257 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1258 	spin_unlock(&dir->i_lock);
1259 }
1260 
1261 struct nfs4_open_createattrs {
1262 	struct nfs4_label *label;
1263 	struct iattr *sattr;
1264 	const __u32 verf[2];
1265 };
1266 
1267 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1268 		int err, struct nfs4_exception *exception)
1269 {
1270 	if (err != -EINVAL)
1271 		return false;
1272 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1273 		return false;
1274 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1275 	exception->retry = 1;
1276 	return true;
1277 }
1278 
1279 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1280 {
1281 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1282 }
1283 
1284 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1285 {
1286 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1287 
1288 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1289 }
1290 
1291 static u32
1292 nfs4_fmode_to_share_access(fmode_t fmode)
1293 {
1294 	u32 res = 0;
1295 
1296 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1297 	case FMODE_READ:
1298 		res = NFS4_SHARE_ACCESS_READ;
1299 		break;
1300 	case FMODE_WRITE:
1301 		res = NFS4_SHARE_ACCESS_WRITE;
1302 		break;
1303 	case FMODE_READ|FMODE_WRITE:
1304 		res = NFS4_SHARE_ACCESS_BOTH;
1305 	}
1306 	return res;
1307 }
1308 
1309 static u32
1310 nfs4_map_atomic_open_share(struct nfs_server *server,
1311 		fmode_t fmode, int openflags)
1312 {
1313 	u32 res = nfs4_fmode_to_share_access(fmode);
1314 
1315 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1316 		goto out;
1317 	/* Want no delegation if we're using O_DIRECT */
1318 	if (openflags & O_DIRECT) {
1319 		res |= NFS4_SHARE_WANT_NO_DELEG;
1320 		goto out;
1321 	}
1322 	/* res |= NFS4_SHARE_WANT_NO_PREFERENCE; */
1323 	if (server->caps & NFS_CAP_DELEGTIME)
1324 		res |= NFS4_SHARE_WANT_DELEG_TIMESTAMPS;
1325 	if (server->caps & NFS_CAP_OPEN_XOR)
1326 		res |= NFS4_SHARE_WANT_OPEN_XOR_DELEGATION;
1327 out:
1328 	return res;
1329 }
1330 
1331 static enum open_claim_type4
1332 nfs4_map_atomic_open_claim(struct nfs_server *server,
1333 		enum open_claim_type4 claim)
1334 {
1335 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1336 		return claim;
1337 	switch (claim) {
1338 	default:
1339 		return claim;
1340 	case NFS4_OPEN_CLAIM_FH:
1341 		return NFS4_OPEN_CLAIM_NULL;
1342 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1343 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1344 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1345 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1346 	}
1347 }
1348 
1349 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1350 {
1351 	p->o_res.f_attr = &p->f_attr;
1352 	p->o_res.seqid = p->o_arg.seqid;
1353 	p->c_res.seqid = p->c_arg.seqid;
1354 	p->o_res.server = p->o_arg.server;
1355 	p->o_res.access_request = p->o_arg.access;
1356 	nfs_fattr_init(&p->f_attr);
1357 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1358 }
1359 
1360 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1361 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1362 		const struct nfs4_open_createattrs *c,
1363 		enum open_claim_type4 claim,
1364 		gfp_t gfp_mask)
1365 {
1366 	struct dentry *parent = dget_parent(dentry);
1367 	struct inode *dir = d_inode(parent);
1368 	struct nfs_server *server = NFS_SERVER(dir);
1369 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1370 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1371 	struct nfs4_opendata *p;
1372 
1373 	p = kzalloc_obj(*p, gfp_mask);
1374 	if (p == NULL)
1375 		goto err;
1376 
1377 	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1378 	if (IS_ERR(p->f_attr.label))
1379 		goto err_free_p;
1380 
1381 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1382 	if (IS_ERR(p->a_label))
1383 		goto err_free_f;
1384 
1385 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1386 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1387 	if (IS_ERR(p->o_arg.seqid))
1388 		goto err_free_label;
1389 	nfs_sb_active(dentry->d_sb);
1390 	p->dentry = dget(dentry);
1391 	p->dir = parent;
1392 	p->owner = sp;
1393 	atomic_inc(&sp->so_count);
1394 	p->o_arg.open_flags = flags;
1395 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1396 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1397 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1398 			fmode, flags);
1399 	if (flags & O_CREAT) {
1400 		p->o_arg.umask = current_umask();
1401 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1402 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1403 			p->o_arg.u.attrs = &p->attrs;
1404 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1405 
1406 			memcpy(p->o_arg.u.verifier.data, c->verf,
1407 					sizeof(p->o_arg.u.verifier.data));
1408 		}
1409 	}
1410 	/* ask server to check for all possible rights as results
1411 	 * are cached */
1412 	switch (p->o_arg.claim) {
1413 	default:
1414 		break;
1415 	case NFS4_OPEN_CLAIM_NULL:
1416 	case NFS4_OPEN_CLAIM_FH:
1417 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1418 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1419 				  NFS4_ACCESS_EXECUTE |
1420 				  nfs_access_xattr_mask(server);
1421 	}
1422 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1423 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1424 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1425 	p->o_arg.name = &dentry->d_name;
1426 	p->o_arg.server = server;
1427 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1428 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1429 	switch (p->o_arg.claim) {
1430 	case NFS4_OPEN_CLAIM_NULL:
1431 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1432 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1433 		p->o_arg.fh = NFS_FH(dir);
1434 		break;
1435 	case NFS4_OPEN_CLAIM_PREVIOUS:
1436 	case NFS4_OPEN_CLAIM_FH:
1437 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1438 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1439 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1440 	}
1441 	p->c_arg.fh = &p->o_res.fh;
1442 	p->c_arg.stateid = &p->o_res.stateid;
1443 	p->c_arg.seqid = p->o_arg.seqid;
1444 	nfs4_init_opendata_res(p);
1445 	kref_init(&p->kref);
1446 	return p;
1447 
1448 err_free_label:
1449 	nfs4_label_free(p->a_label);
1450 err_free_f:
1451 	nfs4_label_free(p->f_attr.label);
1452 err_free_p:
1453 	kfree(p);
1454 err:
1455 	dput(parent);
1456 	return NULL;
1457 }
1458 
1459 static void nfs4_opendata_free(struct kref *kref)
1460 {
1461 	struct nfs4_opendata *p = container_of(kref,
1462 			struct nfs4_opendata, kref);
1463 	struct super_block *sb = p->dentry->d_sb;
1464 
1465 	nfs4_lgopen_release(p->lgp);
1466 	nfs_free_seqid(p->o_arg.seqid);
1467 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1468 	if (p->state != NULL)
1469 		nfs4_put_open_state(p->state);
1470 	nfs4_put_state_owner(p->owner);
1471 
1472 	nfs4_label_free(p->a_label);
1473 	nfs4_label_free(p->f_attr.label);
1474 
1475 	dput(p->dir);
1476 	dput(p->dentry);
1477 	nfs_sb_deactive(sb);
1478 	nfs_fattr_free_names(&p->f_attr);
1479 	kfree(p->f_attr.mdsthreshold);
1480 	kfree(p);
1481 }
1482 
1483 static void nfs4_opendata_put(struct nfs4_opendata *p)
1484 {
1485 	if (p != NULL)
1486 		kref_put(&p->kref, nfs4_opendata_free);
1487 }
1488 
1489 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1490 		fmode_t fmode)
1491 {
1492 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1493 	case FMODE_READ|FMODE_WRITE:
1494 		return state->n_rdwr != 0;
1495 	case FMODE_WRITE:
1496 		return state->n_wronly != 0;
1497 	case FMODE_READ:
1498 		return state->n_rdonly != 0;
1499 	}
1500 	WARN_ON_ONCE(1);
1501 	return false;
1502 }
1503 
1504 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1505 		int open_mode, enum open_claim_type4 claim)
1506 {
1507 	int ret = 0;
1508 
1509 	if (open_mode & (O_EXCL|O_TRUNC))
1510 		goto out;
1511 	switch (claim) {
1512 	case NFS4_OPEN_CLAIM_NULL:
1513 	case NFS4_OPEN_CLAIM_FH:
1514 		goto out;
1515 	default:
1516 		break;
1517 	}
1518 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1519 		case FMODE_READ:
1520 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1521 				&& state->n_rdonly != 0;
1522 			break;
1523 		case FMODE_WRITE:
1524 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1525 				&& state->n_wronly != 0;
1526 			break;
1527 		case FMODE_READ|FMODE_WRITE:
1528 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1529 				&& state->n_rdwr != 0;
1530 	}
1531 out:
1532 	return ret;
1533 }
1534 
1535 static bool can_open_delegated(const struct inode *inode, fmode_t fmode,
1536 		enum open_claim_type4 claim, nfs4_stateid *stateid)
1537 {
1538 	struct nfs_delegation *delegation;
1539 	bool ret = false;
1540 
1541 	delegation = nfs4_get_valid_delegation(inode);
1542 	if (!delegation)
1543 		return false;
1544 	if ((delegation->type & fmode) != fmode)
1545 		goto out_put_delegation;
1546 
1547 	switch (claim) {
1548 	case NFS4_OPEN_CLAIM_PREVIOUS:
1549 		if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1550 			break;
1551 		fallthrough;
1552 	case NFS4_OPEN_CLAIM_NULL:
1553 	case NFS4_OPEN_CLAIM_FH:
1554 		nfs_mark_delegation_referenced(delegation);
1555 		/* Save the delegation stateid */
1556 		if (stateid)
1557 			nfs4_stateid_copy(stateid, &delegation->stateid);
1558 		ret = true;
1559 		break;
1560 	default:
1561 		break;
1562 	}
1563 
1564 out_put_delegation:
1565 	nfs_put_delegation(delegation);
1566 	return ret;
1567 }
1568 
1569 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1570 {
1571 	switch (fmode) {
1572 		case FMODE_WRITE:
1573 			state->n_wronly++;
1574 			break;
1575 		case FMODE_READ:
1576 			state->n_rdonly++;
1577 			break;
1578 		case FMODE_READ|FMODE_WRITE:
1579 			state->n_rdwr++;
1580 	}
1581 	nfs4_state_set_mode_locked(state, state->state | fmode);
1582 }
1583 
1584 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1585 {
1586 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1587 		return true;
1588 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1589 		return true;
1590 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1591 		return true;
1592 	return false;
1593 }
1594 
1595 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1596 {
1597 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1598 		wake_up_all(&state->waitq);
1599 }
1600 
1601 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1602 {
1603 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1604 	bool need_recover = false;
1605 
1606 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1607 		need_recover = true;
1608 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1609 		need_recover = true;
1610 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1611 		need_recover = true;
1612 	if (need_recover)
1613 		nfs4_state_mark_reclaim_nograce(clp, state);
1614 }
1615 
1616 /*
1617  * Check for whether or not the caller may update the open stateid
1618  * to the value passed in by stateid.
1619  *
1620  * Note: This function relies heavily on the server implementing
1621  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1622  * correctly.
1623  * i.e. The stateid seqids have to be initialised to 1, and
1624  * are then incremented on every state transition.
1625  */
1626 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1627 		const nfs4_stateid *stateid)
1628 {
1629 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1630 		/* The common case - we're updating to a new sequence number */
1631 		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1632 			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1633 				return true;
1634 			return false;
1635 		}
1636 		/* The server returned a new stateid */
1637 	}
1638 	/* This is the first OPEN in this generation */
1639 	if (stateid->seqid == cpu_to_be32(1))
1640 		return true;
1641 	return false;
1642 }
1643 
1644 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1645 {
1646 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1647 		return;
1648 	if (state->n_wronly)
1649 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1650 	if (state->n_rdonly)
1651 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1652 	if (state->n_rdwr)
1653 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1654 	set_bit(NFS_OPEN_STATE, &state->flags);
1655 }
1656 
1657 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1658 		nfs4_stateid *stateid, fmode_t fmode)
1659 {
1660 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1661 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1662 	case FMODE_WRITE:
1663 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1664 		break;
1665 	case FMODE_READ:
1666 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1667 		break;
1668 	case 0:
1669 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1670 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1671 		clear_bit(NFS_OPEN_STATE, &state->flags);
1672 	}
1673 	if (stateid == NULL)
1674 		return;
1675 	/* Handle OPEN+OPEN_DOWNGRADE races */
1676 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1677 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1678 		nfs_resync_open_stateid_locked(state);
1679 		goto out;
1680 	}
1681 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1682 		nfs4_stateid_copy(&state->stateid, stateid);
1683 	nfs4_stateid_copy(&state->open_stateid, stateid);
1684 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1685 out:
1686 	nfs_state_log_update_open_stateid(state);
1687 }
1688 
1689 static void nfs_clear_open_stateid(struct nfs4_state *state,
1690 	nfs4_stateid *arg_stateid,
1691 	nfs4_stateid *stateid, fmode_t fmode)
1692 {
1693 	write_seqlock(&state->seqlock);
1694 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1695 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1696 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1697 	write_sequnlock(&state->seqlock);
1698 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1699 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1700 }
1701 
1702 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1703 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1704 	__must_hold(&state->owner->so_lock)
1705 	__must_hold(&state->seqlock)
1706 	__must_hold(RCU)
1707 
1708 {
1709 	DEFINE_WAIT(wait);
1710 	int status = 0;
1711 	for (;;) {
1712 
1713 		if (nfs_stateid_is_sequential(state, stateid))
1714 			break;
1715 
1716 		if (status) {
1717 			if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1718 			    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1719 				trace_nfs4_open_stateid_update_skip(state->inode,
1720 								    stateid, status);
1721 				return;
1722 			} else {
1723 				break;
1724 			}
1725 		}
1726 
1727 		/* Rely on seqids for serialisation with NFSv4.0 */
1728 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1729 			break;
1730 
1731 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1732 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1733 		/*
1734 		 * Ensure we process the state changes in the same order
1735 		 * in which the server processed them by delaying the
1736 		 * update of the stateid until we are in sequence.
1737 		 */
1738 		write_sequnlock(&state->seqlock);
1739 		spin_unlock(&state->owner->so_lock);
1740 		rcu_read_unlock();
1741 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1742 
1743 		if (!fatal_signal_pending(current) &&
1744 		    !nfs_current_task_exiting()) {
1745 			if (schedule_timeout(5*HZ) == 0)
1746 				status = -EAGAIN;
1747 			else
1748 				status = 0;
1749 		} else
1750 			status = -EINTR;
1751 		finish_wait(&state->waitq, &wait);
1752 		rcu_read_lock();
1753 		spin_lock(&state->owner->so_lock);
1754 		write_seqlock(&state->seqlock);
1755 	}
1756 
1757 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1758 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1759 		nfs4_stateid_copy(freeme, &state->open_stateid);
1760 		nfs_test_and_clear_all_open_stateid(state);
1761 	}
1762 
1763 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1764 		nfs4_stateid_copy(&state->stateid, stateid);
1765 	nfs4_stateid_copy(&state->open_stateid, stateid);
1766 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1767 	nfs_state_log_update_open_stateid(state);
1768 }
1769 
1770 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1771 		const nfs4_stateid *open_stateid,
1772 		fmode_t fmode,
1773 		nfs4_stateid *freeme)
1774 {
1775 	/*
1776 	 * Protect the call to nfs4_state_set_mode_locked and
1777 	 * serialise the stateid update
1778 	 */
1779 	write_seqlock(&state->seqlock);
1780 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1781 	switch (fmode) {
1782 	case FMODE_READ:
1783 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1784 		break;
1785 	case FMODE_WRITE:
1786 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1787 		break;
1788 	case FMODE_READ|FMODE_WRITE:
1789 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1790 	}
1791 	set_bit(NFS_OPEN_STATE, &state->flags);
1792 	write_sequnlock(&state->seqlock);
1793 }
1794 
1795 void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1796 {
1797 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1798 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1799 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1800 	clear_bit(NFS_OPEN_STATE, &state->flags);
1801 }
1802 
1803 static void nfs_state_set_delegation(struct nfs4_state *state,
1804 		const nfs4_stateid *deleg_stateid,
1805 		fmode_t fmode)
1806 {
1807 	/*
1808 	 * Protect the call to nfs4_state_set_mode_locked and
1809 	 * serialise the stateid update
1810 	 */
1811 	write_seqlock(&state->seqlock);
1812 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1813 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1814 	write_sequnlock(&state->seqlock);
1815 }
1816 
1817 void nfs_state_clear_delegation(struct nfs4_state *state)
1818 {
1819 	write_seqlock(&state->seqlock);
1820 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1821 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1822 	write_sequnlock(&state->seqlock);
1823 }
1824 
1825 int update_open_stateid(struct nfs4_state *state,
1826 		const nfs4_stateid *open_stateid,
1827 		const nfs4_stateid *delegation,
1828 		fmode_t fmode)
1829 {
1830 	struct nfs_server *server = NFS_SERVER(state->inode);
1831 	struct nfs_client *clp = server->nfs_client;
1832 	struct nfs_delegation *deleg_cur;
1833 	nfs4_stateid freeme = { };
1834 	int ret = 0;
1835 
1836 	fmode &= (FMODE_READ|FMODE_WRITE);
1837 
1838 	spin_lock(&state->owner->so_lock);
1839 	if (open_stateid != NULL) {
1840 		rcu_read_lock();
1841 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1842 		rcu_read_unlock();
1843 		ret = 1;
1844 	}
1845 
1846 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1847 	if (deleg_cur == NULL)
1848 		goto no_delegation;
1849 
1850 	spin_lock(&deleg_cur->lock);
1851 	if (!deleg_cur->inode ||
1852 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1853 	    (deleg_cur->type & fmode) != fmode)
1854 		goto no_delegation_unlock;
1855 
1856 	if (delegation == NULL)
1857 		delegation = &deleg_cur->stateid;
1858 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1859 		goto no_delegation_unlock;
1860 
1861 	nfs_mark_delegation_referenced(deleg_cur);
1862 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1863 	ret = 1;
1864 no_delegation_unlock:
1865 	spin_unlock(&deleg_cur->lock);
1866 	nfs_put_delegation(deleg_cur);
1867 no_delegation:
1868 	if (ret)
1869 		update_open_stateflags(state, fmode);
1870 	spin_unlock(&state->owner->so_lock);
1871 
1872 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1873 		nfs4_schedule_state_manager(clp);
1874 	if (freeme.type != 0)
1875 		nfs4_test_and_free_stateid(server, &freeme,
1876 				state->owner->so_cred);
1877 
1878 	return ret;
1879 }
1880 
1881 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1882 		const nfs4_stateid *stateid)
1883 {
1884 	struct nfs4_state *state = lsp->ls_state;
1885 	bool ret = false;
1886 
1887 	spin_lock(&state->state_lock);
1888 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1889 		goto out_noupdate;
1890 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1891 		goto out_noupdate;
1892 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1893 	ret = true;
1894 out_noupdate:
1895 	spin_unlock(&state->state_lock);
1896 	return ret;
1897 }
1898 
1899 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1900 {
1901 	struct nfs_delegation *delegation;
1902 
1903 	fmode &= FMODE_READ|FMODE_WRITE;
1904 	delegation = nfs4_get_valid_delegation(inode);
1905 	if (!delegation)
1906 		return;
1907 	if ((delegation->type & fmode) != fmode)
1908 		nfs4_inode_return_delegation(inode);
1909 	nfs_put_delegation(delegation);
1910 }
1911 
1912 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1913 {
1914 	struct nfs4_state *state = opendata->state;
1915 	int open_mode = opendata->o_arg.open_flags;
1916 	fmode_t fmode = opendata->o_arg.fmode;
1917 	enum open_claim_type4 claim = opendata->o_arg.claim;
1918 	nfs4_stateid stateid;
1919 	int ret = -EAGAIN;
1920 
1921 	for (;;) {
1922 		spin_lock(&state->owner->so_lock);
1923 		if (can_open_cached(state, fmode, open_mode, claim)) {
1924 			update_open_stateflags(state, fmode);
1925 			spin_unlock(&state->owner->so_lock);
1926 			goto out_return_state;
1927 		}
1928 		spin_unlock(&state->owner->so_lock);
1929 
1930 		if (!can_open_delegated(state->inode, fmode, claim, &stateid))
1931 			break;
1932 
1933 		nfs_release_seqid(opendata->o_arg.seqid);
1934 		if (!opendata->is_recover) {
1935 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1936 			if (ret != 0)
1937 				goto out;
1938 		}
1939 		ret = -EAGAIN;
1940 
1941 		/* Try to update the stateid using the delegation */
1942 		if (update_open_stateid(state, NULL, &stateid, fmode))
1943 			goto out_return_state;
1944 	}
1945 out:
1946 	return ERR_PTR(ret);
1947 out_return_state:
1948 	refcount_inc(&state->count);
1949 	return state;
1950 }
1951 
1952 static void
1953 nfs4_process_delegation(struct inode *inode, const struct cred *cred,
1954 			enum open_claim_type4 claim,
1955 			const struct nfs4_open_delegation *delegation)
1956 {
1957 	switch (delegation->open_delegation_type) {
1958 	case NFS4_OPEN_DELEGATE_READ:
1959 	case NFS4_OPEN_DELEGATE_WRITE:
1960 	case NFS4_OPEN_DELEGATE_READ_ATTRS_DELEG:
1961 	case NFS4_OPEN_DELEGATE_WRITE_ATTRS_DELEG:
1962 		break;
1963 	default:
1964 		return;
1965 	}
1966 	switch (claim) {
1967 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1968 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1969 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1970 				   "returning a delegation for "
1971 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1972 				   NFS_SERVER(inode)->nfs_client->cl_hostname);
1973 		break;
1974 	case NFS4_OPEN_CLAIM_PREVIOUS:
1975 		nfs_inode_reclaim_delegation(inode, cred, delegation->type,
1976 					     &delegation->stateid,
1977 					     delegation->pagemod_limit,
1978 					     delegation->open_delegation_type);
1979 		break;
1980 	default:
1981 		nfs_inode_set_delegation(inode, cred, delegation->type,
1982 					 &delegation->stateid,
1983 					 delegation->pagemod_limit,
1984 					 delegation->open_delegation_type);
1985 	}
1986 	if (delegation->do_recall)
1987 		nfs_async_inode_return_delegation(inode, &delegation->stateid);
1988 }
1989 
1990 /*
1991  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1992  * and update the nfs4_state.
1993  */
1994 static struct nfs4_state *
1995 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1996 {
1997 	struct inode *inode = data->state->inode;
1998 	struct nfs4_state *state = data->state;
1999 	int ret;
2000 
2001 	if (!data->rpc_done) {
2002 		if (data->rpc_status)
2003 			return ERR_PTR(data->rpc_status);
2004 		return nfs4_try_open_cached(data);
2005 	}
2006 
2007 	ret = nfs_refresh_inode(inode, &data->f_attr);
2008 	if (ret)
2009 		return ERR_PTR(ret);
2010 
2011 	nfs4_process_delegation(state->inode,
2012 				data->owner->so_cred,
2013 				data->o_arg.claim,
2014 				&data->o_res.delegation);
2015 
2016 	if (!(data->o_res.rflags & NFS4_OPEN_RESULT_NO_OPEN_STATEID)) {
2017 		if (!update_open_stateid(state, &data->o_res.stateid,
2018 					 NULL, data->o_arg.fmode))
2019 			return ERR_PTR(-EAGAIN);
2020 	} else if (!update_open_stateid(state, NULL, NULL, data->o_arg.fmode))
2021 		return ERR_PTR(-EAGAIN);
2022 	refcount_inc(&state->count);
2023 
2024 	return state;
2025 }
2026 
2027 static struct inode *
2028 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2029 {
2030 	struct inode *inode;
2031 
2032 	switch (data->o_arg.claim) {
2033 	case NFS4_OPEN_CLAIM_NULL:
2034 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2035 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2036 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2037 			return ERR_PTR(-EAGAIN);
2038 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2039 				&data->f_attr);
2040 		break;
2041 	default:
2042 		inode = d_inode(data->dentry);
2043 		ihold(inode);
2044 		nfs_refresh_inode(inode, &data->f_attr);
2045 	}
2046 	return inode;
2047 }
2048 
2049 static struct nfs4_state *
2050 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2051 {
2052 	struct nfs4_state *state;
2053 	struct inode *inode;
2054 
2055 	inode = nfs4_opendata_get_inode(data);
2056 	if (IS_ERR(inode))
2057 		return ERR_CAST(inode);
2058 	if (data->state != NULL && data->state->inode == inode) {
2059 		state = data->state;
2060 		refcount_inc(&state->count);
2061 	} else
2062 		state = nfs4_get_open_state(inode, data->owner);
2063 	iput(inode);
2064 	if (state == NULL)
2065 		state = ERR_PTR(-ENOMEM);
2066 	return state;
2067 }
2068 
2069 static struct nfs4_state *
2070 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2071 {
2072 	struct nfs4_state *state;
2073 
2074 	if (!data->rpc_done) {
2075 		state = nfs4_try_open_cached(data);
2076 		trace_nfs4_cached_open(data->state);
2077 		goto out;
2078 	}
2079 
2080 	state = nfs4_opendata_find_nfs4_state(data);
2081 	if (IS_ERR(state))
2082 		goto out;
2083 
2084 	nfs4_process_delegation(state->inode,
2085 				data->owner->so_cred,
2086 				data->o_arg.claim,
2087 				&data->o_res.delegation);
2088 
2089 	if (!(data->o_res.rflags & NFS4_OPEN_RESULT_NO_OPEN_STATEID)) {
2090 		if (!update_open_stateid(state, &data->o_res.stateid,
2091 					 NULL, data->o_arg.fmode)) {
2092 			nfs4_put_open_state(state);
2093 			state = ERR_PTR(-EAGAIN);
2094 		}
2095 	} else if (!update_open_stateid(state, NULL, NULL, data->o_arg.fmode)) {
2096 		nfs4_put_open_state(state);
2097 		state = ERR_PTR(-EAGAIN);
2098 	}
2099 out:
2100 	nfs_release_seqid(data->o_arg.seqid);
2101 	return state;
2102 }
2103 
2104 static struct nfs4_state *
2105 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2106 {
2107 	struct nfs4_state *ret;
2108 
2109 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2110 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2111 	else
2112 		ret = _nfs4_opendata_to_nfs4_state(data);
2113 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2114 	return ret;
2115 }
2116 
2117 static struct nfs_open_context *
2118 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2119 {
2120 	struct nfs_inode *nfsi = NFS_I(state->inode);
2121 	struct nfs_open_context *ctx;
2122 
2123 	rcu_read_lock();
2124 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2125 		if (ctx->state != state)
2126 			continue;
2127 		if ((ctx->mode & mode) != mode)
2128 			continue;
2129 		if (!get_nfs_open_context(ctx))
2130 			continue;
2131 		rcu_read_unlock();
2132 		return ctx;
2133 	}
2134 	rcu_read_unlock();
2135 	return ERR_PTR(-ENOENT);
2136 }
2137 
2138 static struct nfs_open_context *
2139 nfs4_state_find_open_context(struct nfs4_state *state)
2140 {
2141 	struct nfs_open_context *ctx;
2142 
2143 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2144 	if (!IS_ERR(ctx))
2145 		return ctx;
2146 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2147 	if (!IS_ERR(ctx))
2148 		return ctx;
2149 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2150 }
2151 
2152 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2153 		struct nfs4_state *state, enum open_claim_type4 claim)
2154 {
2155 	struct nfs4_opendata *opendata;
2156 
2157 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2158 			NULL, claim, GFP_NOFS);
2159 	if (opendata == NULL)
2160 		return ERR_PTR(-ENOMEM);
2161 	opendata->state = state;
2162 	refcount_inc(&state->count);
2163 	return opendata;
2164 }
2165 
2166 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2167 				    fmode_t fmode)
2168 {
2169 	struct nfs4_state *newstate;
2170 	struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2171 	int openflags = opendata->o_arg.open_flags;
2172 	int ret;
2173 
2174 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2175 		return 0;
2176 	opendata->o_arg.fmode = fmode;
2177 	opendata->o_arg.share_access =
2178 		nfs4_map_atomic_open_share(server, fmode, openflags);
2179 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2180 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2181 	nfs4_init_opendata_res(opendata);
2182 	ret = _nfs4_recover_proc_open(opendata);
2183 	if (ret != 0)
2184 		return ret;
2185 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2186 	if (IS_ERR(newstate))
2187 		return PTR_ERR(newstate);
2188 	if (newstate != opendata->state)
2189 		ret = -ESTALE;
2190 	nfs4_close_state(newstate, fmode);
2191 	return ret;
2192 }
2193 
2194 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2195 {
2196 	int ret;
2197 
2198 	/* memory barrier prior to reading state->n_* */
2199 	smp_rmb();
2200 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2201 	if (ret != 0)
2202 		return ret;
2203 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2204 	if (ret != 0)
2205 		return ret;
2206 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2207 	if (ret != 0)
2208 		return ret;
2209 	/*
2210 	 * We may have performed cached opens for all three recoveries.
2211 	 * Check if we need to update the current stateid.
2212 	 */
2213 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2214 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2215 		write_seqlock(&state->seqlock);
2216 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2217 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2218 		write_sequnlock(&state->seqlock);
2219 	}
2220 	return 0;
2221 }
2222 
2223 /*
2224  * OPEN_RECLAIM:
2225  * 	reclaim state on the server after a reboot.
2226  */
2227 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2228 {
2229 	struct nfs_delegation *delegation;
2230 	struct nfs4_opendata *opendata;
2231 	u32 delegation_type = NFS4_OPEN_DELEGATE_NONE;
2232 	int status;
2233 
2234 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2235 			NFS4_OPEN_CLAIM_PREVIOUS);
2236 	if (IS_ERR(opendata))
2237 		return PTR_ERR(opendata);
2238 	rcu_read_lock();
2239 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2240 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0) {
2241 		switch(delegation->type) {
2242 		case FMODE_READ:
2243 			delegation_type = NFS4_OPEN_DELEGATE_READ;
2244 			if (test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags))
2245 				delegation_type = NFS4_OPEN_DELEGATE_READ_ATTRS_DELEG;
2246 			break;
2247 		case FMODE_WRITE:
2248 		case FMODE_READ|FMODE_WRITE:
2249 			delegation_type = NFS4_OPEN_DELEGATE_WRITE;
2250 			if (test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags))
2251 				delegation_type = NFS4_OPEN_DELEGATE_WRITE_ATTRS_DELEG;
2252 		}
2253 	}
2254 	rcu_read_unlock();
2255 	opendata->o_arg.u.delegation_type = delegation_type;
2256 	status = nfs4_open_recover(opendata, state);
2257 	nfs4_opendata_put(opendata);
2258 	return status;
2259 }
2260 
2261 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2262 {
2263 	struct nfs_server *server = NFS_SERVER(state->inode);
2264 	struct nfs4_exception exception = { };
2265 	int err;
2266 	do {
2267 		err = _nfs4_do_open_reclaim(ctx, state);
2268 		trace_nfs4_open_reclaim(ctx, 0, err);
2269 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2270 			continue;
2271 		if (err != -NFS4ERR_DELAY)
2272 			break;
2273 		nfs4_handle_exception(server, err, &exception);
2274 	} while (exception.retry);
2275 	return err;
2276 }
2277 
2278 int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2279 {
2280 	struct nfs_open_context *ctx;
2281 	int ret;
2282 
2283 	ctx = nfs4_state_find_open_context(state);
2284 	if (IS_ERR(ctx))
2285 		return -EAGAIN;
2286 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2287 	nfs_state_clear_open_state_flags(state);
2288 	ret = nfs4_do_open_reclaim(ctx, state);
2289 	put_nfs_open_context(ctx);
2290 	return ret;
2291 }
2292 
2293 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)
2294 {
2295 	switch (err) {
2296 		default:
2297 			printk(KERN_ERR "NFS: %s: unhandled error "
2298 					"%d.\n", __func__, err);
2299 			fallthrough;
2300 		case 0:
2301 		case -ENOENT:
2302 		case -EAGAIN:
2303 		case -ESTALE:
2304 		case -ETIMEDOUT:
2305 			break;
2306 		case -NFS4ERR_BADSESSION:
2307 		case -NFS4ERR_BADSLOT:
2308 		case -NFS4ERR_BAD_HIGH_SLOT:
2309 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2310 		case -NFS4ERR_DEADSESSION:
2311 			return -EAGAIN;
2312 		case -NFS4ERR_STALE_CLIENTID:
2313 		case -NFS4ERR_STALE_STATEID:
2314 			/* Don't recall a delegation if it was lost */
2315 			nfs4_schedule_lease_recovery(server->nfs_client);
2316 			return -EAGAIN;
2317 		case -NFS4ERR_MOVED:
2318 			nfs4_schedule_migration_recovery(server);
2319 			return -EAGAIN;
2320 		case -NFS4ERR_LEASE_MOVED:
2321 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2322 			return -EAGAIN;
2323 		case -NFS4ERR_DELEG_REVOKED:
2324 		case -NFS4ERR_ADMIN_REVOKED:
2325 		case -NFS4ERR_EXPIRED:
2326 		case -NFS4ERR_BAD_STATEID:
2327 		case -NFS4ERR_OPENMODE:
2328 			nfs_inode_find_state_and_recover(state->inode,
2329 					stateid);
2330 			nfs4_schedule_stateid_recovery(server, state);
2331 			return -EAGAIN;
2332 		case -NFS4ERR_DELAY:
2333 		case -NFS4ERR_GRACE:
2334 			ssleep(1);
2335 			return -EAGAIN;
2336 		case -ENOMEM:
2337 		case -NFS4ERR_DENIED:
2338 			if (fl) {
2339 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2340 				if (lsp)
2341 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2342 			}
2343 			return 0;
2344 	}
2345 	return err;
2346 }
2347 
2348 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2349 		struct nfs4_state *state, const nfs4_stateid *stateid)
2350 {
2351 	struct nfs_server *server = NFS_SERVER(state->inode);
2352 	struct nfs4_opendata *opendata;
2353 	int err = 0;
2354 
2355 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2356 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2357 	if (IS_ERR(opendata))
2358 		return PTR_ERR(opendata);
2359 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2360 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2361 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2362 		if (err)
2363 			goto out;
2364 	}
2365 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2366 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2367 		if (err)
2368 			goto out;
2369 	}
2370 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2371 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2372 		if (err)
2373 			goto out;
2374 	}
2375 	nfs_state_clear_delegation(state);
2376 out:
2377 	nfs4_opendata_put(opendata);
2378 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2379 }
2380 
2381 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2382 {
2383 	struct nfs4_opendata *data = calldata;
2384 
2385 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2386 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2387 }
2388 
2389 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2390 {
2391 	struct nfs4_opendata *data = calldata;
2392 
2393 	data->c_res.seq_res.sr_slot_ops->done(task, &data->c_res.seq_res);
2394 
2395 	data->rpc_status = task->tk_status;
2396 	if (data->rpc_status == 0) {
2397 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2398 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2399 		renew_lease(data->o_res.server, data->timestamp);
2400 		data->rpc_done = true;
2401 	}
2402 }
2403 
2404 static void nfs4_open_confirm_release(void *calldata)
2405 {
2406 	struct nfs4_opendata *data = calldata;
2407 	struct nfs4_state *state = NULL;
2408 
2409 	/* If this request hasn't been cancelled, do nothing */
2410 	if (!data->cancelled)
2411 		goto out_free;
2412 	/* In case of error, no cleanup! */
2413 	if (!data->rpc_done)
2414 		goto out_free;
2415 	state = nfs4_opendata_to_nfs4_state(data);
2416 	if (!IS_ERR(state))
2417 		nfs4_close_state(state, data->o_arg.fmode);
2418 out_free:
2419 	nfs4_opendata_put(data);
2420 }
2421 
2422 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2423 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2424 	.rpc_call_done = nfs4_open_confirm_done,
2425 	.rpc_release = nfs4_open_confirm_release,
2426 };
2427 
2428 /*
2429  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2430  */
2431 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2432 {
2433 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2434 	struct rpc_task *task;
2435 	struct  rpc_message msg = {
2436 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2437 		.rpc_argp = &data->c_arg,
2438 		.rpc_resp = &data->c_res,
2439 		.rpc_cred = data->owner->so_cred,
2440 	};
2441 	struct rpc_task_setup task_setup_data = {
2442 		.rpc_client = server->client,
2443 		.rpc_message = &msg,
2444 		.callback_ops = &nfs4_open_confirm_ops,
2445 		.callback_data = data,
2446 		.workqueue = nfsiod_workqueue,
2447 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2448 	};
2449 	int status;
2450 
2451 	nfs4_init_sequence(server->nfs_client, &data->c_arg.seq_args,
2452 			   &data->c_res.seq_res, 1, data->is_recover);
2453 	kref_get(&data->kref);
2454 	data->rpc_done = false;
2455 	data->rpc_status = 0;
2456 	data->timestamp = jiffies;
2457 	task = rpc_run_task(&task_setup_data);
2458 	if (IS_ERR(task))
2459 		return PTR_ERR(task);
2460 	status = rpc_wait_for_completion_task(task);
2461 	if (status != 0) {
2462 		data->cancelled = true;
2463 		smp_wmb();
2464 	} else
2465 		status = data->rpc_status;
2466 	rpc_put_task(task);
2467 	return status;
2468 }
2469 
2470 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2471 {
2472 	struct nfs4_opendata *data = calldata;
2473 	struct nfs4_state_owner *sp = data->owner;
2474 	struct nfs_client *clp = sp->so_server->nfs_client;
2475 	enum open_claim_type4 claim = data->o_arg.claim;
2476 
2477 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2478 		goto out_wait;
2479 	/*
2480 	 * Check if we still need to send an OPEN call, or if we can use
2481 	 * a delegation instead.
2482 	 */
2483 	if (data->state != NULL) {
2484 		if (can_open_cached(data->state, data->o_arg.fmode,
2485 					data->o_arg.open_flags, claim))
2486 			goto out_no_action;
2487 		if (can_open_delegated(data->state->inode, data->o_arg.fmode,
2488 				claim, NULL)) {
2489 			trace_nfs4_cached_open(data->state);
2490 			goto out_no_action;
2491 		}
2492 	}
2493 	/* Update client id. */
2494 	data->o_arg.clientid = clp->cl_clientid;
2495 	switch (claim) {
2496 	default:
2497 		break;
2498 	case NFS4_OPEN_CLAIM_PREVIOUS:
2499 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2500 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2501 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2502 		fallthrough;
2503 	case NFS4_OPEN_CLAIM_FH:
2504 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2505 	}
2506 	data->timestamp = jiffies;
2507 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2508 				&data->o_arg.seq_args,
2509 				&data->o_res.seq_res,
2510 				task) != 0)
2511 		nfs_release_seqid(data->o_arg.seqid);
2512 
2513 	/* Set the create mode (note dependency on the session type) */
2514 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2515 	if (data->o_arg.open_flags & O_EXCL) {
2516 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2517 		if (clp->cl_mvops->minor_version == 0) {
2518 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2519 			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2520 			 * because ACCESS will return permission denied for
2521 			 * all bits until close */
2522 			data->o_res.access_request = data->o_arg.access = 0;
2523 		} else if (nfs4_has_persistent_session(clp))
2524 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2525 	}
2526 	return;
2527 
2528 out_no_action:
2529 	task->tk_action = NULL;
2530 out_wait:
2531 	nfs4_sequence_done(task, &data->o_res.seq_res);
2532 }
2533 
2534 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2535 {
2536 	struct nfs4_opendata *data = calldata;
2537 
2538 	data->rpc_status = task->tk_status;
2539 
2540 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2541 		return;
2542 
2543 	if (task->tk_status == 0) {
2544 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2545 			switch (data->o_res.f_attr->mode & S_IFMT) {
2546 			case S_IFREG:
2547 				break;
2548 			case S_IFLNK:
2549 				data->rpc_status = -ELOOP;
2550 				break;
2551 			case S_IFDIR:
2552 				data->rpc_status = -EISDIR;
2553 				break;
2554 			default:
2555 				data->rpc_status = -ENOTDIR;
2556 			}
2557 		}
2558 		renew_lease(data->o_res.server, data->timestamp);
2559 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2560 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2561 	}
2562 	data->rpc_done = true;
2563 }
2564 
2565 static void nfs4_open_release(void *calldata)
2566 {
2567 	struct nfs4_opendata *data = calldata;
2568 	struct nfs4_state *state = NULL;
2569 
2570 	/* In case of error, no cleanup! */
2571 	if (data->rpc_status != 0 || !data->rpc_done) {
2572 		nfs_release_seqid(data->o_arg.seqid);
2573 		goto out_free;
2574 	}
2575 	/* If this request hasn't been cancelled, do nothing */
2576 	if (!data->cancelled)
2577 		goto out_free;
2578 	/* In case we need an open_confirm, no cleanup! */
2579 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2580 		goto out_free;
2581 	state = nfs4_opendata_to_nfs4_state(data);
2582 	if (!IS_ERR(state))
2583 		nfs4_close_state(state, data->o_arg.fmode);
2584 out_free:
2585 	nfs4_opendata_put(data);
2586 }
2587 
2588 static const struct rpc_call_ops nfs4_open_ops = {
2589 	.rpc_call_prepare = nfs4_open_prepare,
2590 	.rpc_call_done = nfs4_open_done,
2591 	.rpc_release = nfs4_open_release,
2592 };
2593 
2594 static int nfs4_run_open_task(struct nfs4_opendata *data,
2595 			      struct nfs_open_context *ctx)
2596 {
2597 	struct inode *dir = d_inode(data->dir);
2598 	struct nfs_server *server = NFS_SERVER(dir);
2599 	struct nfs_client *clp = server->nfs_client;
2600 	struct nfs_openargs *o_arg = &data->o_arg;
2601 	struct nfs_openres *o_res = &data->o_res;
2602 	struct rpc_task *task;
2603 	struct rpc_message msg = {
2604 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2605 		.rpc_argp = o_arg,
2606 		.rpc_resp = o_res,
2607 		.rpc_cred = data->owner->so_cred,
2608 	};
2609 	struct rpc_task_setup task_setup_data = {
2610 		.rpc_client = server->client,
2611 		.rpc_message = &msg,
2612 		.callback_ops = &nfs4_open_ops,
2613 		.callback_data = data,
2614 		.workqueue = nfsiod_workqueue,
2615 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2616 	};
2617 	int status;
2618 
2619 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2620 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2621 
2622 	kref_get(&data->kref);
2623 	data->rpc_done = false;
2624 	data->rpc_status = 0;
2625 	data->cancelled = false;
2626 	data->is_recover = false;
2627 	if (!ctx) {
2628 		nfs4_init_sequence(clp, &o_arg->seq_args, &o_res->seq_res, 1, 1);
2629 		data->is_recover = true;
2630 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2631 	} else {
2632 		nfs4_init_sequence(clp, &o_arg->seq_args, &o_res->seq_res, 1, 0);
2633 		pnfs_lgopen_prepare(data, ctx);
2634 	}
2635 	task = rpc_run_task(&task_setup_data);
2636 	if (IS_ERR(task))
2637 		return PTR_ERR(task);
2638 	status = rpc_wait_for_completion_task(task);
2639 	if (status != 0) {
2640 		data->cancelled = true;
2641 		smp_wmb();
2642 	} else
2643 		status = data->rpc_status;
2644 	rpc_put_task(task);
2645 
2646 	return status;
2647 }
2648 
2649 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2650 {
2651 	struct inode *dir = d_inode(data->dir);
2652 	struct nfs_openres *o_res = &data->o_res;
2653 	int status;
2654 
2655 	status = nfs4_run_open_task(data, NULL);
2656 	if (status != 0 || !data->rpc_done)
2657 		return status;
2658 
2659 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2660 
2661 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2662 		status = _nfs4_proc_open_confirm(data);
2663 
2664 	return status;
2665 }
2666 
2667 /*
2668  * Additional permission checks in order to distinguish between an
2669  * open for read, and an open for execute. This works around the
2670  * fact that NFSv4 OPEN treats read and execute permissions as being
2671  * the same.
2672  * Note that in the non-execute case, we want to turn off permission
2673  * checking if we just created a new file (POSIX open() semantics).
2674  */
2675 static int nfs4_opendata_access(const struct cred *cred,
2676 				struct nfs4_opendata *opendata,
2677 				struct nfs4_state *state, fmode_t fmode)
2678 {
2679 	struct nfs_access_entry cache;
2680 	u32 mask, flags;
2681 
2682 	/* access call failed or for some reason the server doesn't
2683 	 * support any access modes -- defer access call until later */
2684 	if (opendata->o_res.access_supported == 0)
2685 		return 0;
2686 
2687 	mask = 0;
2688 	if (fmode & FMODE_EXEC) {
2689 		/* ONLY check for exec rights */
2690 		if (S_ISDIR(state->inode->i_mode))
2691 			mask = NFS4_ACCESS_LOOKUP;
2692 		else
2693 			mask = NFS4_ACCESS_EXECUTE;
2694 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2695 		mask = NFS4_ACCESS_READ;
2696 
2697 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2698 	nfs_access_add_cache(state->inode, &cache, cred);
2699 
2700 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2701 	if ((mask & ~cache.mask & flags) == 0)
2702 		return 0;
2703 
2704 	return -EACCES;
2705 }
2706 
2707 /*
2708  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2709  */
2710 static int _nfs4_proc_open(struct nfs4_opendata *data,
2711 			   struct nfs_open_context *ctx)
2712 {
2713 	struct inode *dir = d_inode(data->dir);
2714 	struct nfs_server *server = NFS_SERVER(dir);
2715 	struct nfs_openargs *o_arg = &data->o_arg;
2716 	struct nfs_openres *o_res = &data->o_res;
2717 	int status;
2718 
2719 	status = nfs4_run_open_task(data, ctx);
2720 	if (!data->rpc_done)
2721 		return status;
2722 	if (status != 0) {
2723 		if (status == -NFS4ERR_BADNAME &&
2724 				!(o_arg->open_flags & O_CREAT))
2725 			return -ENOENT;
2726 		return status;
2727 	}
2728 
2729 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2730 
2731 	if (o_arg->open_flags & O_CREAT) {
2732 		if (o_arg->open_flags & O_EXCL)
2733 			data->file_created = true;
2734 		else if (o_res->cinfo.before != o_res->cinfo.after)
2735 			data->file_created = true;
2736 		if (data->file_created ||
2737 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2738 			nfs4_update_changeattr(dir, &o_res->cinfo,
2739 					o_res->f_attr->time_start,
2740 					NFS_INO_INVALID_DATA);
2741 	}
2742 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2743 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2744 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2745 		status = _nfs4_proc_open_confirm(data);
2746 		if (status != 0)
2747 			return status;
2748 	}
2749 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2750 		struct nfs_fh *fh = &o_res->fh;
2751 
2752 		nfs4_sequence_free_slot(&o_res->seq_res);
2753 		if (o_arg->claim == NFS4_OPEN_CLAIM_FH)
2754 			fh = NFS_FH(d_inode(data->dentry));
2755 		nfs4_proc_getattr(server, fh, o_res->f_attr, NULL);
2756 	}
2757 	return 0;
2758 }
2759 
2760 /*
2761  * OPEN_EXPIRED:
2762  * 	reclaim state on the server after a network partition.
2763  * 	Assumes caller holds the appropriate lock
2764  */
2765 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2766 {
2767 	struct nfs4_opendata *opendata;
2768 	int ret;
2769 
2770 	opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2771 	if (IS_ERR(opendata))
2772 		return PTR_ERR(opendata);
2773 	/*
2774 	 * We're not recovering a delegation, so ask for no delegation.
2775 	 * Otherwise the recovery thread could deadlock with an outstanding
2776 	 * delegation return.
2777 	 */
2778 	opendata->o_arg.open_flags = O_DIRECT;
2779 	ret = nfs4_open_recover(opendata, state);
2780 	if (ret == -ESTALE)
2781 		d_drop(ctx->dentry);
2782 	nfs4_opendata_put(opendata);
2783 	return ret;
2784 }
2785 
2786 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2787 {
2788 	struct nfs_server *server = NFS_SERVER(state->inode);
2789 	struct nfs4_exception exception = { };
2790 	int err;
2791 
2792 	do {
2793 		err = _nfs4_open_expired(ctx, state);
2794 		trace_nfs4_open_expired(ctx, 0, err);
2795 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2796 			continue;
2797 		switch (err) {
2798 		default:
2799 			goto out;
2800 		case -NFS4ERR_GRACE:
2801 		case -NFS4ERR_DELAY:
2802 			nfs4_handle_exception(server, err, &exception);
2803 			err = 0;
2804 		}
2805 	} while (exception.retry);
2806 out:
2807 	return err;
2808 }
2809 
2810 int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2811 {
2812 	struct nfs_open_context *ctx;
2813 	int ret;
2814 
2815 	ctx = nfs4_state_find_open_context(state);
2816 	if (IS_ERR(ctx))
2817 		return -EAGAIN;
2818 	ret = nfs4_do_open_expired(ctx, state);
2819 	put_nfs_open_context(ctx);
2820 	return ret;
2821 }
2822 
2823 void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2824 		const nfs4_stateid *stateid)
2825 {
2826 	nfs_remove_bad_delegation(state->inode, stateid);
2827 	nfs_state_clear_delegation(state);
2828 }
2829 
2830 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2831 					       nfs4_stateid *stateid, const struct cred *cred)
2832 {
2833 	int status;
2834 
2835 	switch (stateid->type) {
2836 	default:
2837 		break;
2838 	case NFS4_INVALID_STATEID_TYPE:
2839 	case NFS4_SPECIAL_STATEID_TYPE:
2840 	case NFS4_FREED_STATEID_TYPE:
2841 		return -NFS4ERR_BAD_STATEID;
2842 	case NFS4_REVOKED_STATEID_TYPE:
2843 		goto out_free;
2844 	}
2845 
2846 	status = nfs41_test_stateid(server, stateid, cred);
2847 	switch (status) {
2848 	case -NFS4ERR_EXPIRED:
2849 	case -NFS4ERR_ADMIN_REVOKED:
2850 	case -NFS4ERR_DELEG_REVOKED:
2851 		break;
2852 	default:
2853 		return status;
2854 	}
2855 out_free:
2856 	/* Ack the revoked state to the server */
2857 	nfs41_free_stateid(server, stateid, cred, true);
2858 	return -NFS4ERR_EXPIRED;
2859 }
2860 
2861 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2862 {
2863 	struct nfs_server *server = NFS_SERVER(state->inode);
2864 	nfs4_stateid stateid;
2865 	struct nfs_delegation *delegation;
2866 	const struct cred *cred = NULL;
2867 	int status, ret = NFS_OK;
2868 
2869 	/* Get the delegation credential for use by test/free_stateid */
2870 	rcu_read_lock();
2871 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2872 	if (delegation == NULL) {
2873 		rcu_read_unlock();
2874 		nfs_state_clear_delegation(state);
2875 		return NFS_OK;
2876 	}
2877 
2878 	spin_lock(&delegation->lock);
2879 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2880 
2881 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2882 				&delegation->flags)) {
2883 		spin_unlock(&delegation->lock);
2884 		rcu_read_unlock();
2885 		return NFS_OK;
2886 	}
2887 
2888 	if (delegation->cred)
2889 		cred = get_cred(delegation->cred);
2890 	spin_unlock(&delegation->lock);
2891 	rcu_read_unlock();
2892 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2893 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2894 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2895 		nfs_finish_clear_delegation_stateid(state, &stateid);
2896 	else
2897 		ret = status;
2898 
2899 	put_cred(cred);
2900 	return ret;
2901 }
2902 
2903 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2904 {
2905 	nfs4_stateid tmp;
2906 
2907 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2908 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2909 				&tmp, NULL) &&
2910 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2911 		nfs_state_set_delegation(state, &tmp, state->state);
2912 	else
2913 		nfs_state_clear_delegation(state);
2914 }
2915 
2916 /**
2917  * nfs41_check_expired_locks - possibly free a lock stateid
2918  *
2919  * @state: NFSv4 state for an inode
2920  *
2921  * Returns NFS_OK if recovery for this stateid is now finished.
2922  * Otherwise a negative NFS4ERR value is returned.
2923  */
2924 static int nfs41_check_expired_locks(struct nfs4_state *state)
2925 {
2926 	int status, ret = NFS_OK;
2927 	struct nfs4_lock_state *lsp, *prev = NULL;
2928 	struct nfs_server *server = NFS_SERVER(state->inode);
2929 
2930 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2931 		goto out;
2932 
2933 	spin_lock(&state->state_lock);
2934 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2935 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2936 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2937 
2938 			refcount_inc(&lsp->ls_count);
2939 			spin_unlock(&state->state_lock);
2940 
2941 			nfs4_put_lock_state(prev);
2942 			prev = lsp;
2943 
2944 			status = nfs41_test_and_free_expired_stateid(server,
2945 					&lsp->ls_stateid,
2946 					cred);
2947 			trace_nfs4_test_lock_stateid(state, lsp, status);
2948 			if (status == -NFS4ERR_EXPIRED ||
2949 			    status == -NFS4ERR_BAD_STATEID) {
2950 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2951 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2952 				if (!recover_lost_locks)
2953 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2954 			} else if (status != NFS_OK) {
2955 				ret = status;
2956 				nfs4_put_lock_state(prev);
2957 				goto out;
2958 			}
2959 			spin_lock(&state->state_lock);
2960 		}
2961 	}
2962 	spin_unlock(&state->state_lock);
2963 	nfs4_put_lock_state(prev);
2964 out:
2965 	return ret;
2966 }
2967 
2968 /**
2969  * nfs41_check_open_stateid - possibly free an open stateid
2970  *
2971  * @state: NFSv4 state for an inode
2972  *
2973  * Returns NFS_OK if recovery for this stateid is now finished.
2974  * Otherwise a negative NFS4ERR value is returned.
2975  */
2976 static int nfs41_check_open_stateid(struct nfs4_state *state)
2977 {
2978 	struct nfs_server *server = NFS_SERVER(state->inode);
2979 	nfs4_stateid *stateid = &state->open_stateid;
2980 	const struct cred *cred = state->owner->so_cred;
2981 	int status;
2982 
2983 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2984 		return -NFS4ERR_BAD_STATEID;
2985 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2986 	trace_nfs4_test_open_stateid(state, NULL, status);
2987 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2988 		nfs_state_clear_open_state_flags(state);
2989 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2990 		return status;
2991 	}
2992 	if (nfs_open_stateid_recover_openmode(state))
2993 		return -NFS4ERR_OPENMODE;
2994 	return NFS_OK;
2995 }
2996 
2997 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2998 {
2999 	int status;
3000 
3001 	status = nfs41_check_delegation_stateid(state);
3002 	if (status != NFS_OK)
3003 		return status;
3004 	nfs41_delegation_recover_stateid(state);
3005 
3006 	status = nfs41_check_expired_locks(state);
3007 	if (status != NFS_OK)
3008 		return status;
3009 	status = nfs41_check_open_stateid(state);
3010 	if (status != NFS_OK)
3011 		status = nfs4_open_expired(sp, state);
3012 	return status;
3013 }
3014 
3015 /*
3016  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
3017  * fields corresponding to attributes that were used to store the verifier.
3018  * Make sure we clobber those fields in the later setattr call
3019  */
3020 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
3021 				struct iattr *sattr, struct nfs4_label **label)
3022 {
3023 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
3024 	__u32 attrset[3];
3025 	unsigned ret;
3026 	unsigned i;
3027 
3028 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3029 		attrset[i] = opendata->o_res.attrset[i];
3030 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3031 			attrset[i] &= ~bitmask[i];
3032 	}
3033 
3034 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3035 		sattr->ia_valid : 0;
3036 
3037 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3038 		if (sattr->ia_valid & ATTR_ATIME_SET)
3039 			ret |= ATTR_ATIME_SET;
3040 		else
3041 			ret |= ATTR_ATIME;
3042 	}
3043 
3044 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3045 		if (sattr->ia_valid & ATTR_MTIME_SET)
3046 			ret |= ATTR_MTIME_SET;
3047 		else
3048 			ret |= ATTR_MTIME;
3049 	}
3050 
3051 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3052 		*label = NULL;
3053 	return ret;
3054 }
3055 
3056 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3057 		struct nfs_open_context *ctx)
3058 {
3059 	struct nfs4_state_owner *sp = opendata->owner;
3060 	struct nfs_server *server = sp->so_server;
3061 	struct dentry *dentry;
3062 	struct nfs4_state *state;
3063 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3064 	struct inode *dir = d_inode(opendata->dir);
3065 	unsigned long dir_verifier;
3066 	int ret;
3067 
3068 	dir_verifier = nfs_save_change_attribute(dir);
3069 
3070 	ret = _nfs4_proc_open(opendata, ctx);
3071 	if (ret != 0)
3072 		goto out;
3073 
3074 	state = _nfs4_opendata_to_nfs4_state(opendata);
3075 	ret = PTR_ERR(state);
3076 	if (IS_ERR(state))
3077 		goto out;
3078 	ctx->state = state;
3079 	if (server->caps & NFS_CAP_POSIX_LOCK)
3080 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3081 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3082 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3083 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3084 		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3085 
3086 	switch(opendata->o_arg.claim) {
3087 	default:
3088 		break;
3089 	case NFS4_OPEN_CLAIM_NULL:
3090 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3091 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3092 		if (!opendata->rpc_done)
3093 			break;
3094 		if (opendata->o_res.delegation.type != 0)
3095 			dir_verifier = nfs_save_change_attribute(dir);
3096 	}
3097 
3098 	dentry = opendata->dentry;
3099 	nfs_set_verifier(dentry, dir_verifier);
3100 	if (d_really_is_negative(dentry)) {
3101 		struct dentry *alias;
3102 		d_drop(dentry);
3103 		alias = d_splice_alias(igrab(state->inode), dentry);
3104 		/* d_splice_alias() can't fail here - it's a non-directory */
3105 		if (alias) {
3106 			dput(ctx->dentry);
3107 			nfs_set_verifier(alias, dir_verifier);
3108 			ctx->dentry = dentry = alias;
3109 		}
3110 	}
3111 
3112 	/* Parse layoutget results before we check for access */
3113 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3114 
3115 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3116 	if (ret != 0)
3117 		goto out;
3118 
3119 	if (d_inode(dentry) == state->inode)
3120 		nfs_inode_attach_open_context(ctx);
3121 
3122 out:
3123 	if (!opendata->cancelled) {
3124 		if (opendata->lgp) {
3125 			nfs4_lgopen_release(opendata->lgp);
3126 			opendata->lgp = NULL;
3127 		}
3128 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3129 	}
3130 	return ret;
3131 }
3132 
3133 /*
3134  * Returns a referenced nfs4_state
3135  */
3136 static int _nfs4_do_open(struct inode *dir,
3137 			struct nfs_open_context *ctx,
3138 			int flags,
3139 			const struct nfs4_open_createattrs *c,
3140 			int *opened)
3141 {
3142 	struct nfs4_state_owner  *sp;
3143 	struct nfs4_state     *state = NULL;
3144 	struct nfs_server       *server = NFS_SERVER(dir);
3145 	struct nfs4_opendata *opendata;
3146 	struct dentry *dentry = ctx->dentry;
3147 	const struct cred *cred = ctx->cred;
3148 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3149 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3150 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3151 	struct iattr *sattr = c->sattr;
3152 	struct nfs4_label *label = c->label;
3153 	int status;
3154 
3155 	/* Protect against reboot recovery conflicts */
3156 	status = -ENOMEM;
3157 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3158 	if (sp == NULL) {
3159 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3160 		goto out_err;
3161 	}
3162 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3163 	if (status != 0)
3164 		goto err_put_state_owner;
3165 	if (d_really_is_positive(dentry))
3166 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3167 	status = -ENOMEM;
3168 	if (d_really_is_positive(dentry))
3169 		claim = NFS4_OPEN_CLAIM_FH;
3170 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3171 			c, claim, GFP_KERNEL);
3172 	if (opendata == NULL)
3173 		goto err_put_state_owner;
3174 
3175 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3176 		if (!opendata->f_attr.mdsthreshold) {
3177 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3178 			if (!opendata->f_attr.mdsthreshold)
3179 				goto err_opendata_put;
3180 		}
3181 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3182 	}
3183 	if (d_really_is_positive(dentry))
3184 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3185 
3186 	status = _nfs4_open_and_get_state(opendata, ctx);
3187 	if (status != 0)
3188 		goto err_opendata_put;
3189 	state = ctx->state;
3190 
3191 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3192 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3193 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3194 		/*
3195 		 * send create attributes which was not set by open
3196 		 * with an extra setattr.
3197 		 */
3198 		if (attrs || label) {
3199 			unsigned ia_old = sattr->ia_valid;
3200 
3201 			sattr->ia_valid = attrs;
3202 			nfs_fattr_init(opendata->o_res.f_attr);
3203 			status = nfs4_do_setattr(state->inode, cred,
3204 					opendata->o_res.f_attr, sattr,
3205 					ctx, label);
3206 			if (status == 0) {
3207 				nfs_setattr_update_inode(state->inode, sattr,
3208 						opendata->o_res.f_attr);
3209 				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3210 			}
3211 			sattr->ia_valid = ia_old;
3212 		}
3213 	}
3214 	if (opened && opendata->file_created)
3215 		*opened = 1;
3216 
3217 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3218 		*ctx_th = opendata->f_attr.mdsthreshold;
3219 		opendata->f_attr.mdsthreshold = NULL;
3220 	}
3221 
3222 	nfs4_opendata_put(opendata);
3223 	nfs4_put_state_owner(sp);
3224 	return 0;
3225 err_opendata_put:
3226 	nfs4_opendata_put(opendata);
3227 err_put_state_owner:
3228 	nfs4_put_state_owner(sp);
3229 out_err:
3230 	return status;
3231 }
3232 
3233 
3234 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3235 					struct nfs_open_context *ctx,
3236 					int flags,
3237 					struct iattr *sattr,
3238 					struct nfs4_label *label,
3239 					int *opened)
3240 {
3241 	struct nfs_server *server = NFS_SERVER(dir);
3242 	struct nfs4_exception exception = {
3243 		.interruptible = true,
3244 	};
3245 	struct nfs4_state *res;
3246 	struct nfs4_open_createattrs c = {
3247 		.label = label,
3248 		.sattr = sattr,
3249 		.verf = {
3250 			[0] = (__u32)jiffies,
3251 			[1] = (__u32)current->pid,
3252 		},
3253 	};
3254 	int status;
3255 
3256 	do {
3257 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3258 		res = ctx->state;
3259 		trace_nfs4_open_file(ctx, flags, status);
3260 		if (status == 0)
3261 			break;
3262 		/* NOTE: BAD_SEQID means the server and client disagree about the
3263 		 * book-keeping w.r.t. state-changing operations
3264 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3265 		 * It is actually a sign of a bug on the client or on the server.
3266 		 *
3267 		 * If we receive a BAD_SEQID error in the particular case of
3268 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3269 		 * have unhashed the old state_owner for us, and that we can
3270 		 * therefore safely retry using a new one. We should still warn
3271 		 * the user though...
3272 		 */
3273 		if (status == -NFS4ERR_BAD_SEQID) {
3274 			pr_warn_ratelimited("NFS: v4 server %s "
3275 					" returned a bad sequence-id error!\n",
3276 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3277 			exception.retry = 1;
3278 			continue;
3279 		}
3280 		/*
3281 		 * BAD_STATEID on OPEN means that the server cancelled our
3282 		 * state before it received the OPEN_CONFIRM.
3283 		 * Recover by retrying the request as per the discussion
3284 		 * on Page 181 of RFC3530.
3285 		 */
3286 		if (status == -NFS4ERR_BAD_STATEID) {
3287 			exception.retry = 1;
3288 			continue;
3289 		}
3290 		if (status == -NFS4ERR_EXPIRED) {
3291 			nfs4_schedule_lease_recovery(server->nfs_client);
3292 			exception.retry = 1;
3293 			continue;
3294 		}
3295 		if (status == -EAGAIN) {
3296 			/* We must have found a delegation */
3297 			exception.retry = 1;
3298 			continue;
3299 		}
3300 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3301 			continue;
3302 		res = ERR_PTR(nfs4_handle_exception(server,
3303 					status, &exception));
3304 	} while (exception.retry);
3305 	return res;
3306 }
3307 
3308 static int _nfs4_do_setattr(struct inode *inode,
3309 			    struct nfs_setattrargs *arg,
3310 			    struct nfs_setattrres *res,
3311 			    const struct cred *cred,
3312 			    struct nfs_open_context *ctx)
3313 {
3314 	struct nfs_server *server = NFS_SERVER(inode);
3315 	struct rpc_message msg = {
3316 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3317 		.rpc_argp	= arg,
3318 		.rpc_resp	= res,
3319 		.rpc_cred	= cred,
3320 	};
3321 	const struct cred *delegation_cred = NULL;
3322 	unsigned long timestamp = jiffies;
3323 	bool truncate;
3324 	int status;
3325 
3326 	nfs_fattr_init(res->fattr);
3327 
3328 	/* Servers should only apply open mode checks for file size changes */
3329 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3330 	if (!truncate) {
3331 		nfs4_inode_make_writeable(inode);
3332 		goto zero_stateid;
3333 	}
3334 
3335 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3336 		/* Use that stateid */
3337 	} else if (ctx != NULL && ctx->state) {
3338 		struct nfs_lock_context *l_ctx;
3339 		if (!nfs4_valid_open_stateid(ctx->state))
3340 			return -EBADF;
3341 		l_ctx = nfs_get_lock_context(ctx);
3342 		if (IS_ERR(l_ctx))
3343 			return PTR_ERR(l_ctx);
3344 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3345 						&arg->stateid, &delegation_cred);
3346 		nfs_put_lock_context(l_ctx);
3347 		if (status == -EIO)
3348 			return -EBADF;
3349 		else if (status == -EAGAIN)
3350 			goto zero_stateid;
3351 	} else {
3352 zero_stateid:
3353 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3354 	}
3355 	if (delegation_cred)
3356 		msg.rpc_cred = delegation_cred;
3357 
3358 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3359 
3360 	put_cred(delegation_cred);
3361 	if (status == 0 && ctx != NULL)
3362 		renew_lease(server, timestamp);
3363 	trace_nfs4_setattr(inode, &arg->stateid, status);
3364 	return status;
3365 }
3366 
3367 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3368 			   struct nfs_fattr *fattr, struct iattr *sattr,
3369 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3370 {
3371 	struct nfs_server *server = NFS_SERVER(inode);
3372 	__u32 bitmask[NFS4_BITMASK_SZ];
3373 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3374 	struct nfs_setattrargs	arg = {
3375 		.fh		= NFS_FH(inode),
3376 		.iap		= sattr,
3377 		.server		= server,
3378 		.bitmask = bitmask,
3379 		.label		= ilabel,
3380 	};
3381 	struct nfs_setattrres  res = {
3382 		.fattr		= fattr,
3383 		.server		= server,
3384 	};
3385 	struct nfs4_exception exception = {
3386 		.state = state,
3387 		.inode = inode,
3388 		.stateid = &arg.stateid,
3389 	};
3390 	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE |
3391 				     NFS_INO_INVALID_CTIME;
3392 	int err;
3393 
3394 	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3395 		adjust_flags |= NFS_INO_INVALID_MODE;
3396 	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3397 		adjust_flags |= NFS_INO_INVALID_OTHER;
3398 	if (sattr->ia_valid & ATTR_ATIME)
3399 		adjust_flags |= NFS_INO_INVALID_ATIME;
3400 	if (sattr->ia_valid & ATTR_MTIME)
3401 		adjust_flags |= NFS_INO_INVALID_MTIME;
3402 
3403 	do {
3404 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3405 					inode, adjust_flags);
3406 
3407 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3408 		switch (err) {
3409 		case -NFS4ERR_OPENMODE:
3410 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3411 				pr_warn_once("NFSv4: server %s is incorrectly "
3412 						"applying open mode checks to "
3413 						"a SETATTR that is not "
3414 						"changing file size.\n",
3415 						server->nfs_client->cl_hostname);
3416 			}
3417 			if (state && !(state->state & FMODE_WRITE)) {
3418 				err = -EBADF;
3419 				if (sattr->ia_valid & ATTR_OPEN)
3420 					err = -EACCES;
3421 				goto out;
3422 			}
3423 		}
3424 		err = nfs4_handle_exception(server, err, &exception);
3425 	} while (exception.retry);
3426 out:
3427 	return err;
3428 }
3429 
3430 static bool
3431 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3432 {
3433 	if (inode == NULL || !nfs_have_layout(inode))
3434 		return false;
3435 
3436 	return pnfs_wait_on_layoutreturn(inode, task);
3437 }
3438 
3439 /*
3440  * Update the seqid of an open stateid
3441  */
3442 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3443 		struct nfs4_state *state)
3444 {
3445 	__be32 seqid_open;
3446 	u32 dst_seqid;
3447 	int seq;
3448 
3449 	for (;;) {
3450 		if (!nfs4_valid_open_stateid(state))
3451 			break;
3452 		seq = read_seqbegin(&state->seqlock);
3453 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3454 			nfs4_stateid_copy(dst, &state->open_stateid);
3455 			if (read_seqretry(&state->seqlock, seq))
3456 				continue;
3457 			break;
3458 		}
3459 		seqid_open = state->open_stateid.seqid;
3460 		if (read_seqretry(&state->seqlock, seq))
3461 			continue;
3462 
3463 		dst_seqid = be32_to_cpu(dst->seqid);
3464 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3465 			dst->seqid = seqid_open;
3466 		break;
3467 	}
3468 }
3469 
3470 /*
3471  * Update the seqid of an open stateid after receiving
3472  * NFS4ERR_OLD_STATEID
3473  */
3474 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3475 		struct nfs4_state *state)
3476 {
3477 	__be32 seqid_open;
3478 	u32 dst_seqid;
3479 	bool ret;
3480 	int seq, status = -EAGAIN;
3481 	DEFINE_WAIT(wait);
3482 
3483 	for (;;) {
3484 		ret = false;
3485 		if (!nfs4_valid_open_stateid(state))
3486 			break;
3487 		seq = read_seqbegin(&state->seqlock);
3488 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3489 			if (read_seqretry(&state->seqlock, seq))
3490 				continue;
3491 			break;
3492 		}
3493 
3494 		write_seqlock(&state->seqlock);
3495 		seqid_open = state->open_stateid.seqid;
3496 
3497 		dst_seqid = be32_to_cpu(dst->seqid);
3498 
3499 		/* Did another OPEN bump the state's seqid?  try again: */
3500 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3501 			dst->seqid = seqid_open;
3502 			write_sequnlock(&state->seqlock);
3503 			ret = true;
3504 			break;
3505 		}
3506 
3507 		/* server says we're behind but we haven't seen the update yet */
3508 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3509 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3510 		write_sequnlock(&state->seqlock);
3511 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3512 
3513 		if (fatal_signal_pending(current) || nfs_current_task_exiting())
3514 			status = -EINTR;
3515 		else
3516 			if (schedule_timeout(5*HZ) != 0)
3517 				status = 0;
3518 
3519 		finish_wait(&state->waitq, &wait);
3520 
3521 		if (!status)
3522 			continue;
3523 		if (status == -EINTR)
3524 			break;
3525 
3526 		/* we slept the whole 5 seconds, we must have lost a seqid */
3527 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3528 		ret = true;
3529 		break;
3530 	}
3531 
3532 	return ret;
3533 }
3534 
3535 struct nfs4_closedata {
3536 	struct inode *inode;
3537 	struct nfs4_state *state;
3538 	struct nfs_closeargs arg;
3539 	struct nfs_closeres res;
3540 	struct {
3541 		struct nfs4_layoutreturn_args arg;
3542 		struct nfs4_layoutreturn_res res;
3543 		struct nfs4_xdr_opaque_data ld_private;
3544 		u32 roc_barrier;
3545 		bool roc;
3546 	} lr;
3547 	struct nfs_fattr fattr;
3548 	unsigned long timestamp;
3549 	unsigned short retrans;
3550 };
3551 
3552 static void nfs4_free_closedata(void *data)
3553 {
3554 	struct nfs4_closedata *calldata = data;
3555 	struct nfs4_state_owner *sp = calldata->state->owner;
3556 	struct super_block *sb = calldata->state->inode->i_sb;
3557 
3558 	if (calldata->lr.roc)
3559 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3560 				calldata->res.lr_ret);
3561 	nfs4_put_open_state(calldata->state);
3562 	nfs_free_seqid(calldata->arg.seqid);
3563 	nfs4_put_state_owner(sp);
3564 	nfs_sb_deactive(sb);
3565 	kfree(calldata);
3566 }
3567 
3568 static void nfs4_close_done(struct rpc_task *task, void *data)
3569 {
3570 	struct nfs4_closedata *calldata = data;
3571 	struct nfs4_state *state = calldata->state;
3572 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3573 	nfs4_stateid *res_stateid = NULL;
3574 	struct nfs4_exception exception = {
3575 		.state = state,
3576 		.inode = calldata->inode,
3577 		.stateid = &calldata->arg.stateid,
3578 		.retrans = calldata->retrans,
3579 	};
3580 
3581 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3582 		return;
3583 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3584 
3585 	/* Handle Layoutreturn errors */
3586 	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3587 			  &calldata->res.lr_ret) == -EAGAIN)
3588 		goto out_restart;
3589 
3590 	/* hmm. we are done with the inode, and in the process of freeing
3591 	 * the state_owner. we keep this around to process errors
3592 	 */
3593 	switch (task->tk_status) {
3594 		case 0:
3595 			res_stateid = &calldata->res.stateid;
3596 			renew_lease(server, calldata->timestamp);
3597 			break;
3598 		case -NFS4ERR_ACCESS:
3599 			if (calldata->arg.bitmask != NULL) {
3600 				calldata->arg.bitmask = NULL;
3601 				calldata->res.fattr = NULL;
3602 				goto out_restart;
3603 
3604 			}
3605 			break;
3606 		case -NFS4ERR_OLD_STATEID:
3607 			/* Did we race with OPEN? */
3608 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3609 						state))
3610 				goto out_restart;
3611 			goto out_release;
3612 		case -NFS4ERR_ADMIN_REVOKED:
3613 		case -NFS4ERR_STALE_STATEID:
3614 		case -NFS4ERR_EXPIRED:
3615 			nfs4_free_revoked_stateid(server,
3616 					&calldata->arg.stateid,
3617 					task->tk_msg.rpc_cred);
3618 			fallthrough;
3619 		case -NFS4ERR_BAD_STATEID:
3620 			if (calldata->arg.fmode == 0)
3621 				break;
3622 			fallthrough;
3623 		default:
3624 			task->tk_status = nfs4_async_handle_exception(task,
3625 					server, task->tk_status, &exception);
3626 			calldata->retrans = exception.retrans;
3627 			if (exception.retry)
3628 				goto out_restart;
3629 	}
3630 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3631 			res_stateid, calldata->arg.fmode);
3632 out_release:
3633 	task->tk_status = 0;
3634 	nfs_release_seqid(calldata->arg.seqid);
3635 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3636 	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3637 	return;
3638 out_restart:
3639 	task->tk_status = 0;
3640 	rpc_restart_call_prepare(task);
3641 	goto out_release;
3642 }
3643 
3644 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3645 {
3646 	struct nfs4_closedata *calldata = data;
3647 	struct nfs4_state *state = calldata->state;
3648 	struct inode *inode = calldata->inode;
3649 	struct nfs_server *server = NFS_SERVER(inode);
3650 	struct pnfs_layout_hdr *lo;
3651 	bool is_rdonly, is_wronly, is_rdwr;
3652 	int call_close = 0;
3653 
3654 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3655 		goto out_wait;
3656 
3657 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3658 	spin_lock(&state->owner->so_lock);
3659 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3660 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3661 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3662 	/* Calculate the change in open mode */
3663 	calldata->arg.fmode = 0;
3664 	if (state->n_rdwr == 0) {
3665 		if (state->n_rdonly == 0)
3666 			call_close |= is_rdonly;
3667 		else if (is_rdonly)
3668 			calldata->arg.fmode |= FMODE_READ;
3669 		if (state->n_wronly == 0)
3670 			call_close |= is_wronly;
3671 		else if (is_wronly)
3672 			calldata->arg.fmode |= FMODE_WRITE;
3673 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3674 			call_close |= is_rdwr;
3675 	} else if (is_rdwr)
3676 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3677 
3678 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3679 	if (!nfs4_valid_open_stateid(state))
3680 		call_close = 0;
3681 	spin_unlock(&state->owner->so_lock);
3682 
3683 	if (!call_close) {
3684 		/* Note: exit _without_ calling nfs4_close_done */
3685 		goto out_no_action;
3686 	}
3687 
3688 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3689 		nfs_release_seqid(calldata->arg.seqid);
3690 		goto out_wait;
3691 	}
3692 
3693 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3694 	if (lo && !pnfs_layout_is_valid(lo)) {
3695 		calldata->arg.lr_args = NULL;
3696 		calldata->res.lr_res = NULL;
3697 	}
3698 
3699 	if (calldata->arg.fmode == 0)
3700 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3701 
3702 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3703 		/* Close-to-open cache consistency revalidation */
3704 		if (!nfs4_have_delegation(inode, FMODE_READ, 0)) {
3705 			nfs4_bitmask_set(calldata->arg.bitmask_store,
3706 					 server->cache_consistency_bitmask,
3707 					 inode, 0);
3708 			calldata->arg.bitmask = calldata->arg.bitmask_store;
3709 		} else
3710 			calldata->arg.bitmask = NULL;
3711 	}
3712 
3713 	calldata->arg.share_access =
3714 		nfs4_fmode_to_share_access(calldata->arg.fmode);
3715 
3716 	if (calldata->res.fattr == NULL)
3717 		calldata->arg.bitmask = NULL;
3718 	else if (calldata->arg.bitmask == NULL)
3719 		calldata->res.fattr = NULL;
3720 	calldata->timestamp = jiffies;
3721 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3722 				&calldata->arg.seq_args,
3723 				&calldata->res.seq_res,
3724 				task) != 0)
3725 		nfs_release_seqid(calldata->arg.seqid);
3726 	return;
3727 out_no_action:
3728 	task->tk_action = NULL;
3729 out_wait:
3730 	nfs4_sequence_done(task, &calldata->res.seq_res);
3731 }
3732 
3733 static const struct rpc_call_ops nfs4_close_ops = {
3734 	.rpc_call_prepare = nfs4_close_prepare,
3735 	.rpc_call_done = nfs4_close_done,
3736 	.rpc_release = nfs4_free_closedata,
3737 };
3738 
3739 /*
3740  * It is possible for data to be read/written from a mem-mapped file
3741  * after the sys_close call (which hits the vfs layer as a flush).
3742  * This means that we can't safely call nfsv4 close on a file until
3743  * the inode is cleared. This in turn means that we are not good
3744  * NFSv4 citizens - we do not indicate to the server to update the file's
3745  * share state even when we are done with one of the three share
3746  * stateid's in the inode.
3747  *
3748  * NOTE: Caller must be holding the sp->so_owner semaphore!
3749  */
3750 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3751 {
3752 	struct nfs_server *server = NFS_SERVER(state->inode);
3753 	struct nfs_client *clp = server->nfs_client;
3754 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3755 	struct nfs4_closedata *calldata;
3756 	struct nfs4_state_owner *sp = state->owner;
3757 	struct rpc_task *task;
3758 	struct rpc_message msg = {
3759 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3760 		.rpc_cred = state->owner->so_cred,
3761 	};
3762 	struct rpc_task_setup task_setup_data = {
3763 		.rpc_client = server->client,
3764 		.rpc_message = &msg,
3765 		.callback_ops = &nfs4_close_ops,
3766 		.workqueue = nfsiod_workqueue,
3767 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3768 	};
3769 	int status = -ENOMEM;
3770 
3771 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3772 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3773 
3774 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_CLEANUP,
3775 		&task_setup_data.rpc_client, &msg);
3776 
3777 	calldata = kzalloc_obj(*calldata, gfp_mask);
3778 	if (calldata == NULL)
3779 		goto out;
3780 	nfs4_init_sequence(clp, &calldata->arg.seq_args,
3781 			   &calldata->res.seq_res, 1, 0);
3782 	calldata->inode = state->inode;
3783 	calldata->state = state;
3784 	calldata->arg.fh = NFS_FH(state->inode);
3785 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3786 		goto out_free_calldata;
3787 	/* Serialization for the sequence id */
3788 	alloc_seqid = clp->cl_mvops->alloc_seqid;
3789 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3790 	if (IS_ERR(calldata->arg.seqid))
3791 		goto out_free_calldata;
3792 	nfs_fattr_init(&calldata->fattr);
3793 	calldata->arg.fmode = 0;
3794 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3795 	calldata->res.fattr = &calldata->fattr;
3796 	calldata->res.seqid = calldata->arg.seqid;
3797 	calldata->res.server = server;
3798 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3799 	calldata->lr.roc = pnfs_roc(state->inode, &calldata->lr.arg,
3800 				    &calldata->lr.res, msg.rpc_cred, wait);
3801 	if (calldata->lr.roc) {
3802 		calldata->arg.lr_args = &calldata->lr.arg;
3803 		calldata->res.lr_res = &calldata->lr.res;
3804 	}
3805 	nfs_sb_active(calldata->inode->i_sb);
3806 
3807 	msg.rpc_argp = &calldata->arg;
3808 	msg.rpc_resp = &calldata->res;
3809 	task_setup_data.callback_data = calldata;
3810 	task = rpc_run_task(&task_setup_data);
3811 	if (IS_ERR(task))
3812 		return PTR_ERR(task);
3813 	status = 0;
3814 	if (wait)
3815 		status = rpc_wait_for_completion_task(task);
3816 	rpc_put_task(task);
3817 	return status;
3818 out_free_calldata:
3819 	kfree(calldata);
3820 out:
3821 	nfs4_put_open_state(state);
3822 	nfs4_put_state_owner(sp);
3823 	return status;
3824 }
3825 
3826 static struct inode *
3827 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3828 		int open_flags, struct iattr *attr, int *opened)
3829 {
3830 	struct nfs4_state *state;
3831 	struct nfs4_label l, *label;
3832 
3833 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3834 
3835 	/* Protect against concurrent sillydeletes */
3836 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3837 
3838 	nfs4_label_release_security(label);
3839 
3840 	if (IS_ERR(state))
3841 		return ERR_CAST(state);
3842 	return state->inode;
3843 }
3844 
3845 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3846 {
3847 	struct dentry *dentry = ctx->dentry;
3848 	if (ctx->state == NULL)
3849 		return;
3850 	if (dentry->d_flags & DCACHE_NFSFS_RENAMED)
3851 		nfs4_inode_set_return_delegation_on_close(d_inode(dentry));
3852 	if (is_sync)
3853 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3854 	else
3855 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3856 }
3857 
3858 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3859 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3860 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_OPEN_ARGUMENTS - 1UL)
3861 
3862 #define FATTR4_WORD2_NFS42_TIME_DELEG_MASK \
3863 	(FATTR4_WORD2_TIME_DELEG_MODIFY|FATTR4_WORD2_TIME_DELEG_ACCESS)
3864 static bool nfs4_server_delegtime_capable(struct nfs4_server_caps_res *res)
3865 {
3866 	u32 share_access_want = res->open_caps.oa_share_access_want[0];
3867 	u32 attr_bitmask = res->attr_bitmask[2];
3868 
3869 	return (share_access_want & NFS4_SHARE_WANT_DELEG_TIMESTAMPS) &&
3870 	       ((attr_bitmask & FATTR4_WORD2_NFS42_TIME_DELEG_MASK) ==
3871 					FATTR4_WORD2_NFS42_TIME_DELEG_MASK);
3872 }
3873 
3874 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3875 {
3876 	u32 minorversion = server->nfs_client->cl_minorversion;
3877 	u32 bitmask[3] = {
3878 		[0] = FATTR4_WORD0_SUPPORTED_ATTRS,
3879 	};
3880 	struct nfs4_server_caps_arg args = {
3881 		.fhandle = fhandle,
3882 		.bitmask = bitmask,
3883 	};
3884 	struct nfs4_server_caps_res res = {};
3885 	struct rpc_message msg = {
3886 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3887 		.rpc_argp = &args,
3888 		.rpc_resp = &res,
3889 	};
3890 	int status;
3891 	int i;
3892 
3893 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3894 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3895 		     FATTR4_WORD0_LINK_SUPPORT |
3896 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3897 		     FATTR4_WORD0_ACLSUPPORT |
3898 		     FATTR4_WORD0_CASE_INSENSITIVE |
3899 		     FATTR4_WORD0_CASE_PRESERVING;
3900 	if (minorversion)
3901 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3902 	if (minorversion > 1)
3903 		bitmask[2] |= FATTR4_WORD2_OPEN_ARGUMENTS;
3904 
3905 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3906 	if (status == 0) {
3907 		bitmask[0] = (FATTR4_WORD0_SUPPORTED_ATTRS |
3908 			      FATTR4_WORD0_FH_EXPIRE_TYPE |
3909 			      FATTR4_WORD0_LINK_SUPPORT |
3910 			      FATTR4_WORD0_SYMLINK_SUPPORT |
3911 			      FATTR4_WORD0_ACLSUPPORT |
3912 			      FATTR4_WORD0_CASE_INSENSITIVE |
3913 			      FATTR4_WORD0_CASE_PRESERVING) &
3914 			     res.attr_bitmask[0];
3915 		/* Sanity check the server answers */
3916 		switch (minorversion) {
3917 		case 0:
3918 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3919 			res.attr_bitmask[2] = 0;
3920 			break;
3921 		case 1:
3922 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3923 			bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT &
3924 				     res.attr_bitmask[2];
3925 			break;
3926 		case 2:
3927 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3928 			bitmask[2] = (FATTR4_WORD2_SUPPATTR_EXCLCREAT |
3929 				      FATTR4_WORD2_OPEN_ARGUMENTS) &
3930 				     res.attr_bitmask[2];
3931 		}
3932 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3933 		server->caps &=
3934 			~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS | NFS_CAP_SYMLINKS |
3935 			  NFS_CAP_SECURITY_LABEL | NFS_CAP_FS_LOCATIONS |
3936 			  NFS_CAP_OPEN_XOR | NFS_CAP_DELEGTIME |
3937 			  NFS_CAP_CASE_INSENSITIVE | NFS_CAP_CASE_NONPRESERVING);
3938 		server->fattr_valid = NFS_ATTR_FATTR_V4;
3939 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3940 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3941 			server->caps |= NFS_CAP_ACLS;
3942 		if (res.has_links != 0)
3943 			server->caps |= NFS_CAP_HARDLINKS;
3944 		if (res.has_symlinks != 0)
3945 			server->caps |= NFS_CAP_SYMLINKS;
3946 		if (res.case_insensitive)
3947 			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3948 		if ((res.attr_bitmask[0] & FATTR4_WORD0_CASE_PRESERVING) &&
3949 		    !res.case_preserving)
3950 			server->caps |= NFS_CAP_CASE_NONPRESERVING;
3951 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3952 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3953 			server->caps |= NFS_CAP_SECURITY_LABEL;
3954 #endif
3955 		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3956 			server->caps |= NFS_CAP_FS_LOCATIONS;
3957 		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3958 			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3959 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3960 			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3961 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3962 			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3963 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3964 			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3965 				NFS_ATTR_FATTR_OWNER_NAME);
3966 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3967 			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3968 				NFS_ATTR_FATTR_GROUP_NAME);
3969 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3970 			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3971 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3972 			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3973 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3974 			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3975 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3976 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3977 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3978 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3979 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_CREATE))
3980 			server->fattr_valid &= ~NFS_ATTR_FATTR_BTIME;
3981 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3982 				sizeof(server->attr_bitmask));
3983 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3984 
3985 		if (res.open_caps.oa_share_access_want[0] &
3986 		    NFS4_SHARE_WANT_OPEN_XOR_DELEGATION)
3987 			server->caps |= NFS_CAP_OPEN_XOR;
3988 		if (nfs4_server_delegtime_capable(&res))
3989 			server->caps |= NFS_CAP_DELEGTIME;
3990 
3991 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3992 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3993 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3994 		server->cache_consistency_bitmask[2] = 0;
3995 
3996 		/* Avoid a regression due to buggy server */
3997 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3998 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3999 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
4000 			sizeof(server->exclcreat_bitmask));
4001 
4002 		server->acl_bitmask = res.acl_bitmask;
4003 		server->fh_expire_type = res.fh_expire_type;
4004 	}
4005 
4006 	return status;
4007 }
4008 
4009 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
4010 {
4011 	struct nfs4_exception exception = {
4012 		.interruptible = true,
4013 	};
4014 	int err;
4015 
4016 	do {
4017 		err = nfs4_handle_exception(server,
4018 				_nfs4_server_capabilities(server, fhandle),
4019 				&exception);
4020 	} while (exception.retry);
4021 	return err;
4022 }
4023 
4024 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
4025 					  struct nfs_client *clp,
4026 					  struct nfs_server *server)
4027 {
4028 	int i;
4029 
4030 	for (i = 0; i < location->nservers; i++) {
4031 		struct nfs4_string *srv_loc = &location->servers[i];
4032 		struct sockaddr_storage addr;
4033 		size_t addrlen;
4034 		struct xprt_create xprt_args = {
4035 			.ident = 0,
4036 			.net = clp->cl_net,
4037 		};
4038 		struct nfs4_add_xprt_data xprtdata = {
4039 			.clp = clp,
4040 		};
4041 		struct rpc_add_xprt_test rpcdata = {
4042 			.add_xprt_test = clp->cl_mvops->session_trunk,
4043 			.data = &xprtdata,
4044 		};
4045 		char *servername = NULL;
4046 
4047 		if (!srv_loc->len)
4048 			continue;
4049 
4050 		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
4051 						&addr, sizeof(addr),
4052 						clp->cl_net, server->port);
4053 		if (!addrlen)
4054 			return;
4055 		xprt_args.dstaddr = (struct sockaddr *)&addr;
4056 		xprt_args.addrlen = addrlen;
4057 		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
4058 		if (!servername)
4059 			return;
4060 		memcpy(servername, srv_loc->data, srv_loc->len);
4061 		servername[srv_loc->len] = '\0';
4062 		xprt_args.servername = servername;
4063 
4064 		xprtdata.cred = nfs4_get_clid_cred(clp);
4065 		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
4066 				  rpc_clnt_setup_test_and_add_xprt,
4067 				  &rpcdata);
4068 		if (xprtdata.cred)
4069 			put_cred(xprtdata.cred);
4070 		kfree(servername);
4071 	}
4072 }
4073 
4074 static bool _is_same_nfs4_pathname(struct nfs4_pathname *path1,
4075 				   struct nfs4_pathname *path2)
4076 {
4077 	int i;
4078 
4079 	if (path1->ncomponents != path2->ncomponents)
4080 		return false;
4081 	for (i = 0; i < path1->ncomponents; i++) {
4082 		if (path1->components[i].len != path2->components[i].len)
4083 			return false;
4084 		if (memcmp(path1->components[i].data, path2->components[i].data,
4085 				path1->components[i].len))
4086 			return false;
4087 	}
4088 	return true;
4089 }
4090 
4091 static int _nfs4_discover_trunking(struct nfs_server *server,
4092 				   struct nfs_fh *fhandle)
4093 {
4094 	struct nfs4_fs_locations *locations = NULL;
4095 	struct page *page;
4096 	const struct cred *cred;
4097 	struct nfs_client *clp = server->nfs_client;
4098 	const struct nfs4_state_maintenance_ops *ops =
4099 		clp->cl_mvops->state_renewal_ops;
4100 	int status = -ENOMEM, i;
4101 
4102 	cred = ops->get_state_renewal_cred(clp);
4103 	if (cred == NULL) {
4104 		cred = nfs4_get_clid_cred(clp);
4105 		if (cred == NULL)
4106 			return -ENOKEY;
4107 	}
4108 
4109 	page = alloc_page(GFP_KERNEL);
4110 	if (!page)
4111 		goto out_put_cred;
4112 	locations = kmalloc_obj(struct nfs4_fs_locations);
4113 	if (!locations)
4114 		goto out_free;
4115 	locations->fattr = nfs_alloc_fattr();
4116 	if (!locations->fattr)
4117 		goto out_free_2;
4118 
4119 	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4120 					 cred);
4121 	if (status)
4122 		goto out_free_3;
4123 
4124 	for (i = 0; i < locations->nlocations; i++) {
4125 		if (!_is_same_nfs4_pathname(&locations->fs_path,
4126 					&locations->locations[i].rootpath))
4127 			continue;
4128 		test_fs_location_for_trunking(&locations->locations[i], clp,
4129 					      server);
4130 	}
4131 out_free_3:
4132 	kfree(locations->fattr);
4133 out_free_2:
4134 	kfree(locations);
4135 out_free:
4136 	__free_page(page);
4137 out_put_cred:
4138 	put_cred(cred);
4139 	return status;
4140 }
4141 
4142 static int nfs4_discover_trunking(struct nfs_server *server,
4143 				  struct nfs_fh *fhandle)
4144 {
4145 	struct nfs4_exception exception = {
4146 		.interruptible = true,
4147 	};
4148 	struct nfs_client *clp = server->nfs_client;
4149 	int err = 0;
4150 
4151 	if (!nfs4_has_session(clp))
4152 		goto out;
4153 	do {
4154 		err = nfs4_handle_exception(server,
4155 				_nfs4_discover_trunking(server, fhandle),
4156 				&exception);
4157 	} while (exception.retry);
4158 out:
4159 	return err;
4160 }
4161 
4162 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4163 			     struct nfs_fattr *fattr)
4164 {
4165 	u32 bitmask[3] = {
4166 		[0] = FATTR4_WORD0_TYPE | FATTR4_WORD0_CHANGE |
4167 		      FATTR4_WORD0_SIZE | FATTR4_WORD0_FSID,
4168 	};
4169 	struct nfs4_lookup_root_arg args = {
4170 		.bitmask = bitmask,
4171 	};
4172 	struct nfs4_lookup_res res = {
4173 		.server = server,
4174 		.fattr = fattr,
4175 		.fh = fhandle,
4176 	};
4177 	struct rpc_message msg = {
4178 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4179 		.rpc_argp = &args,
4180 		.rpc_resp = &res,
4181 	};
4182 
4183 	nfs_fattr_init(fattr);
4184 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4185 }
4186 
4187 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4188 			    struct nfs_fattr *fattr)
4189 {
4190 	struct nfs4_exception exception = {
4191 		.interruptible = true,
4192 	};
4193 	int err;
4194 	do {
4195 		err = _nfs4_lookup_root(server, fhandle, fattr);
4196 		trace_nfs4_lookup_root(server, fhandle, fattr, err);
4197 		switch (err) {
4198 		case 0:
4199 		case -NFS4ERR_WRONGSEC:
4200 			goto out;
4201 		default:
4202 			err = nfs4_handle_exception(server, err, &exception);
4203 		}
4204 	} while (exception.retry);
4205 out:
4206 	return err;
4207 }
4208 
4209 static int nfs4_lookup_root_sec(struct nfs_server *server,
4210 				struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4211 				rpc_authflavor_t flavor)
4212 {
4213 	struct rpc_auth_create_args auth_args = {
4214 		.pseudoflavor = flavor,
4215 	};
4216 	struct rpc_auth *auth;
4217 
4218 	auth = rpcauth_create(&auth_args, server->client);
4219 	if (IS_ERR(auth))
4220 		return -EACCES;
4221 	return nfs4_lookup_root(server, fhandle, fattr);
4222 }
4223 
4224 /*
4225  * Retry pseudoroot lookup with various security flavors.  We do this when:
4226  *
4227  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4228  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4229  *
4230  * Returns zero on success, or a negative NFS4ERR value, or a
4231  * negative errno value.
4232  */
4233 int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4234 			      struct nfs_fattr *fattr)
4235 {
4236 	/* Per 3530bis 15.33.5 */
4237 	static const rpc_authflavor_t flav_array[] = {
4238 		RPC_AUTH_GSS_KRB5P,
4239 		RPC_AUTH_GSS_KRB5I,
4240 		RPC_AUTH_GSS_KRB5,
4241 		RPC_AUTH_UNIX,			/* courtesy */
4242 		RPC_AUTH_NULL,
4243 	};
4244 	int status = -EPERM;
4245 	size_t i;
4246 
4247 	if (server->auth_info.flavor_len > 0) {
4248 		/* try each flavor specified by user */
4249 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4250 			status = nfs4_lookup_root_sec(
4251 				server, fhandle, fattr,
4252 				server->auth_info.flavors[i]);
4253 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4254 				continue;
4255 			break;
4256 		}
4257 	} else {
4258 		/* no flavors specified by user, try default list */
4259 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4260 			status = nfs4_lookup_root_sec(server, fhandle, fattr,
4261 						      flav_array[i]);
4262 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4263 				continue;
4264 			break;
4265 		}
4266 	}
4267 
4268 	/*
4269 	 * -EACCES could mean that the user doesn't have correct permissions
4270 	 * to access the mount.  It could also mean that we tried to mount
4271 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4272 	 * existing mount programs don't handle -EACCES very well so it should
4273 	 * be mapped to -EPERM instead.
4274 	 */
4275 	if (status == -EACCES)
4276 		status = -EPERM;
4277 	return status;
4278 }
4279 
4280 /**
4281  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4282  * @server: initialized nfs_server handle
4283  * @fhandle: we fill in the pseudo-fs root file handle
4284  * @fattr: we fill in a bare bones struct fattr
4285  * @auth_probe: probe the auth flavours
4286  *
4287  * Returns zero on success, or a negative errno.
4288  */
4289 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4290 			 struct nfs_fattr *fattr, bool auth_probe)
4291 {
4292 	int status = 0;
4293 
4294 	if (!auth_probe)
4295 		status = nfs4_lookup_root(server, fhandle, fattr);
4296 
4297 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4298 		status = server->nfs_client->cl_mvops->find_root_sec(
4299 			server, fhandle, fattr);
4300 
4301 	return nfs4_map_errors(status);
4302 }
4303 
4304 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4305 			      struct nfs_fsinfo *info)
4306 {
4307 	int error;
4308 	struct nfs_fattr *fattr = info->fattr;
4309 
4310 	error = nfs4_server_capabilities(server, mntfh);
4311 	if (error < 0) {
4312 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4313 		return error;
4314 	}
4315 
4316 	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4317 	if (error < 0) {
4318 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4319 		goto out;
4320 	}
4321 
4322 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4323 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4324 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4325 
4326 out:
4327 	return error;
4328 }
4329 
4330 /*
4331  * Get locations and (maybe) other attributes of a referral.
4332  * Note that we'll actually follow the referral later when
4333  * we detect fsid mismatch in inode revalidation
4334  */
4335 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4336 			     const struct qstr *name, struct nfs_fattr *fattr,
4337 			     struct nfs_fh *fhandle)
4338 {
4339 	int status = -ENOMEM;
4340 	struct page *page = NULL;
4341 	struct nfs4_fs_locations *locations = NULL;
4342 
4343 	page = alloc_page(GFP_KERNEL);
4344 	if (page == NULL)
4345 		goto out;
4346 	locations = kmalloc_obj(struct nfs4_fs_locations);
4347 	if (locations == NULL)
4348 		goto out;
4349 
4350 	locations->fattr = fattr;
4351 
4352 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4353 	if (status != 0)
4354 		goto out;
4355 
4356 	/*
4357 	 * If the fsid didn't change, this is a migration event, not a
4358 	 * referral.  Cause us to drop into the exception handler, which
4359 	 * will kick off migration recovery.
4360 	 */
4361 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4362 		dprintk("%s: server did not return a different fsid for"
4363 			" a referral at %s\n", __func__, name->name);
4364 		status = -NFS4ERR_MOVED;
4365 		goto out;
4366 	}
4367 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4368 	nfs_fixup_referral_attributes(fattr);
4369 	memset(fhandle, 0, sizeof(struct nfs_fh));
4370 out:
4371 	if (page)
4372 		__free_page(page);
4373 	kfree(locations);
4374 	return status;
4375 }
4376 
4377 static bool should_request_dir_deleg(struct inode *inode)
4378 {
4379 	if (!directory_delegations)
4380 		return false;
4381 	if (!inode)
4382 		return false;
4383 	if (!S_ISDIR(inode->i_mode))
4384 		return false;
4385 	if (!nfs_server_capable(inode, NFS_CAP_DIR_DELEG))
4386 		return false;
4387 	if (!test_and_clear_bit(NFS_INO_REQ_DIR_DELEG, &(NFS_I(inode)->flags)))
4388 		return false;
4389 	if (nfs4_have_delegation(inode, FMODE_READ, 0))
4390 		return false;
4391 	return true;
4392 }
4393 
4394 static void nfs4_call_getattr_prepare(struct rpc_task *task, void *calldata)
4395 {
4396 	struct nfs4_call_sync_data *data = calldata;
4397 	nfs4_setup_sequence(data->seq_server->nfs_client, data->seq_args,
4398 			    data->seq_res, task);
4399 }
4400 
4401 static void nfs4_call_getattr_done(struct rpc_task *task, void *calldata)
4402 {
4403 	struct nfs4_call_sync_data *data = calldata;
4404 
4405 	nfs4_sequence_process(task, data->seq_res);
4406 }
4407 
4408 static const struct rpc_call_ops nfs4_call_getattr_ops = {
4409 	.rpc_call_prepare = nfs4_call_getattr_prepare,
4410 	.rpc_call_done = nfs4_call_getattr_done,
4411 };
4412 
4413 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4414 				struct nfs_fattr *fattr, struct inode *inode)
4415 {
4416 	__u32 bitmask[NFS4_BITMASK_SZ];
4417 	struct nfs4_getattr_arg args = {
4418 		.fh = fhandle,
4419 		.bitmask = bitmask,
4420 	};
4421 	struct nfs4_getattr_res res = {
4422 		.fattr = fattr,
4423 		.server = server,
4424 	};
4425 	struct rpc_message msg = {
4426 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4427 		.rpc_argp = &args,
4428 		.rpc_resp = &res,
4429 	};
4430 	struct nfs4_call_sync_data data = {
4431 		.seq_server = server,
4432 		.seq_args = &args.seq_args,
4433 		.seq_res = &res.seq_res,
4434 	};
4435 	struct rpc_task_setup task_setup = {
4436 		.rpc_client = server->client,
4437 		.rpc_message = &msg,
4438 		.callback_ops = &nfs4_call_getattr_ops,
4439 		.callback_data = &data,
4440 	};
4441 	struct nfs_client *clp = server->nfs_client;
4442 	struct nfs4_gdd_res gdd_res;
4443 	int status;
4444 
4445 	if (nfs4_has_session(clp))
4446 		task_setup.flags = RPC_TASK_MOVEABLE;
4447 
4448 	/* Is this is an attribute revalidation, subject to softreval? */
4449 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4450 		task_setup.flags |= RPC_TASK_TIMEOUT;
4451 
4452 	args.get_dir_deleg = should_request_dir_deleg(inode);
4453 	if (args.get_dir_deleg)
4454 		res.gdd_res = &gdd_res;
4455 
4456 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4457 	nfs_fattr_init(fattr);
4458 	nfs4_init_sequence(clp, &args.seq_args, &res.seq_res, 0, 0);
4459 
4460 	status = nfs4_call_sync_custom(&task_setup);
4461 
4462 	if (args.get_dir_deleg) {
4463 		switch (status) {
4464 		case 0:
4465 			if (gdd_res.status != GDD4_OK)
4466 				break;
4467 			nfs_inode_set_delegation(inode, current_cred(),
4468 						 FMODE_READ, &gdd_res.deleg, 0,
4469 						 NFS4_OPEN_DELEGATE_READ);
4470 			break;
4471 		case -ENOTSUPP:
4472 		case -EOPNOTSUPP:
4473 			server->caps &= ~NFS_CAP_DIR_DELEG;
4474 			break;
4475 		case -NFS4ERR_INVAL:
4476 		case -NFS4ERR_IO:
4477 		case -NFS4ERR_DIRDELEG_UNAVAIL:
4478 		case -NFS4ERR_NOTDIR:
4479 			clear_bit(NFS_INO_REQ_DIR_DELEG, &(NFS_I(inode)->flags));
4480 			status = -EAGAIN;
4481 		}
4482 	}
4483 
4484 	nfs4_sequence_free_slot(&res.seq_res);
4485 	return status;
4486 }
4487 
4488 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4489 				struct nfs_fattr *fattr, struct inode *inode)
4490 {
4491 	struct nfs4_exception exception = {
4492 		.interruptible = true,
4493 	};
4494 	int err;
4495 	do {
4496 		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4497 		trace_nfs4_getattr(server, fhandle, fattr, err);
4498 		switch (err) {
4499 		default:
4500 			err = nfs4_handle_exception(server, err, &exception);
4501 			break;
4502 		case -EAGAIN:
4503 		case -ENOTSUPP:
4504 		case -EOPNOTSUPP:
4505 			exception.retry = true;
4506 		}
4507 	} while (exception.retry);
4508 	return err;
4509 }
4510 
4511 /*
4512  * The file is not closed if it is opened due to the a request to change
4513  * the size of the file. The open call will not be needed once the
4514  * VFS layer lookup-intents are implemented.
4515  *
4516  * Close is called when the inode is destroyed.
4517  * If we haven't opened the file for O_WRONLY, we
4518  * need to in the size_change case to obtain a stateid.
4519  *
4520  * Got race?
4521  * Because OPEN is always done by name in nfsv4, it is
4522  * possible that we opened a different file by the same
4523  * name.  We can recognize this race condition, but we
4524  * can't do anything about it besides returning an error.
4525  *
4526  * This will be fixed with VFS changes (lookup-intent).
4527  */
4528 static int
4529 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4530 		  struct iattr *sattr)
4531 {
4532 	struct inode *inode = d_inode(dentry);
4533 	const struct cred *cred = NULL;
4534 	struct nfs_open_context *ctx = NULL;
4535 	int status;
4536 
4537 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4538 	    sattr->ia_valid & ATTR_SIZE &&
4539 	    sattr->ia_size < i_size_read(inode))
4540 		pnfs_commit_and_return_layout(inode);
4541 
4542 	nfs_fattr_init(fattr);
4543 
4544 	/* Deal with open(O_TRUNC) */
4545 	if (sattr->ia_valid & ATTR_OPEN)
4546 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4547 
4548 	/* Optimization: if the end result is no change, don't RPC */
4549 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4550 		return 0;
4551 
4552 	/* Search for an existing open(O_WRITE) file */
4553 	if (sattr->ia_valid & ATTR_FILE) {
4554 
4555 		ctx = nfs_file_open_context(sattr->ia_file);
4556 		if (ctx)
4557 			cred = ctx->cred;
4558 	}
4559 
4560 	/* Return any delegations if we're going to change ACLs */
4561 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4562 		nfs4_inode_make_writeable(inode);
4563 
4564 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4565 	if (status == 0) {
4566 		nfs_setattr_update_inode(inode, sattr, fattr);
4567 		nfs_setsecurity(inode, fattr);
4568 	}
4569 	return status;
4570 }
4571 
4572 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4573 		struct dentry *dentry, const struct qstr *name,
4574 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4575 {
4576 	struct nfs_server *server = NFS_SERVER(dir);
4577 	int		       status;
4578 	struct nfs4_lookup_arg args = {
4579 		.bitmask = server->attr_bitmask,
4580 		.dir_fh = NFS_FH(dir),
4581 		.name = name,
4582 	};
4583 	struct nfs4_lookup_res res = {
4584 		.server = server,
4585 		.fattr = fattr,
4586 		.fh = fhandle,
4587 	};
4588 	struct rpc_message msg = {
4589 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4590 		.rpc_argp = &args,
4591 		.rpc_resp = &res,
4592 	};
4593 	unsigned short task_flags = 0;
4594 
4595 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4596 		task_flags = RPC_TASK_MOVEABLE;
4597 
4598 	/* Is this is an attribute revalidation, subject to softreval? */
4599 	if (nfs_lookup_is_soft_revalidate(dentry))
4600 		task_flags |= RPC_TASK_TIMEOUT;
4601 
4602 	args.bitmask = nfs4_bitmask(server, fattr->label);
4603 
4604 	nfs_fattr_init(fattr);
4605 
4606 	dprintk("NFS call  lookup %pd2\n", dentry);
4607 	nfs4_init_sequence(server->nfs_client, &args.seq_args, &res.seq_res, 0, 0);
4608 	status = nfs4_do_call_sync(clnt, server, &msg,
4609 			&args.seq_args, &res.seq_res, task_flags);
4610 	dprintk("NFS reply lookup: %d\n", status);
4611 	return status;
4612 }
4613 
4614 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4615 {
4616 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4617 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4618 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4619 	fattr->nlink = 2;
4620 }
4621 
4622 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4623 				   struct dentry *dentry, const struct qstr *name,
4624 				   struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4625 {
4626 	struct nfs4_exception exception = {
4627 		.interruptible = true,
4628 	};
4629 	struct rpc_clnt *client = *clnt;
4630 	int err;
4631 	do {
4632 		err = _nfs4_proc_lookup(client, dir, dentry, name, fhandle, fattr);
4633 		trace_nfs4_lookup(dir, name, err);
4634 		switch (err) {
4635 		case -NFS4ERR_BADNAME:
4636 			err = -ENOENT;
4637 			goto out;
4638 		case -NFS4ERR_MOVED:
4639 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4640 			if (err == -NFS4ERR_MOVED)
4641 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4642 			goto out;
4643 		case -NFS4ERR_WRONGSEC:
4644 			err = -EPERM;
4645 			if (client != *clnt)
4646 				goto out;
4647 			client = nfs4_negotiate_security(client, dir, name);
4648 			if (IS_ERR(client))
4649 				return PTR_ERR(client);
4650 
4651 			exception.retry = 1;
4652 			break;
4653 		default:
4654 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4655 		}
4656 	} while (exception.retry);
4657 
4658 out:
4659 	if (err == 0)
4660 		*clnt = client;
4661 	else if (client != *clnt)
4662 		rpc_shutdown_client(client);
4663 
4664 	return err;
4665 }
4666 
4667 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry, const struct qstr *name,
4668 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4669 {
4670 	int status;
4671 	struct rpc_clnt *client = NFS_CLIENT(dir);
4672 
4673 	status = nfs4_proc_lookup_common(&client, dir, dentry, name, fhandle, fattr);
4674 	if (client != NFS_CLIENT(dir)) {
4675 		rpc_shutdown_client(client);
4676 		nfs_fixup_secinfo_attributes(fattr);
4677 	}
4678 	return status;
4679 }
4680 
4681 struct rpc_clnt *
4682 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4683 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4684 {
4685 	struct rpc_clnt *client = NFS_CLIENT(dir);
4686 	int status;
4687 
4688 	status = nfs4_proc_lookup_common(&client, dir, dentry, &dentry->d_name,
4689 					 fhandle, fattr);
4690 	if (status < 0)
4691 		return ERR_PTR(status);
4692 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4693 }
4694 
4695 static int _nfs4_proc_lookupp(struct inode *inode,
4696 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4697 {
4698 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4699 	struct nfs_server *server = NFS_SERVER(inode);
4700 	int		       status;
4701 	struct nfs4_lookupp_arg args = {
4702 		.bitmask = server->attr_bitmask,
4703 		.fh = NFS_FH(inode),
4704 	};
4705 	struct nfs4_lookupp_res res = {
4706 		.server = server,
4707 		.fattr = fattr,
4708 		.fh = fhandle,
4709 	};
4710 	struct rpc_message msg = {
4711 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4712 		.rpc_argp = &args,
4713 		.rpc_resp = &res,
4714 	};
4715 	unsigned short task_flags = 0;
4716 
4717 	if (server->flags & NFS_MOUNT_SOFTREVAL)
4718 		task_flags |= RPC_TASK_TIMEOUT;
4719 	if (server->caps & NFS_CAP_MOVEABLE)
4720 		task_flags |= RPC_TASK_MOVEABLE;
4721 
4722 	args.bitmask = nfs4_bitmask(server, fattr->label);
4723 
4724 	nfs_fattr_init(fattr);
4725 	nfs4_init_sequence(server->nfs_client, &args.seq_args, &res.seq_res, 0, 0);
4726 
4727 	dprintk("NFS call  lookupp ino=0x%llx\n", inode->i_ino);
4728 	status = nfs4_do_call_sync(clnt, server, &msg, &args.seq_args,
4729 				   &res.seq_res, task_flags);
4730 	dprintk("NFS reply lookupp: %d\n", status);
4731 	return status;
4732 }
4733 
4734 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4735 			     struct nfs_fattr *fattr)
4736 {
4737 	struct nfs4_exception exception = {
4738 		.interruptible = true,
4739 	};
4740 	int err;
4741 	do {
4742 		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4743 		trace_nfs4_lookupp(inode, err);
4744 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4745 				&exception);
4746 	} while (exception.retry);
4747 	return err;
4748 }
4749 
4750 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4751 			     const struct cred *cred)
4752 {
4753 	struct nfs_server *server = NFS_SERVER(inode);
4754 	struct nfs4_accessargs args = {
4755 		.fh = NFS_FH(inode),
4756 		.access = entry->mask,
4757 	};
4758 	struct nfs4_accessres res = {
4759 		.server = server,
4760 	};
4761 	struct rpc_message msg = {
4762 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4763 		.rpc_argp = &args,
4764 		.rpc_resp = &res,
4765 		.rpc_cred = cred,
4766 	};
4767 	int status = 0;
4768 
4769 	if (!nfs4_have_delegation(inode, FMODE_READ, 0)) {
4770 		nfs_request_directory_delegation(inode);
4771 		res.fattr = nfs_alloc_fattr();
4772 		if (res.fattr == NULL)
4773 			return -ENOMEM;
4774 		args.bitmask = server->cache_consistency_bitmask;
4775 	}
4776 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4777 	if (!status) {
4778 		nfs_access_set_mask(entry, res.access);
4779 		if (res.fattr)
4780 			nfs_refresh_inode(inode, res.fattr);
4781 	}
4782 	nfs_free_fattr(res.fattr);
4783 	return status;
4784 }
4785 
4786 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4787 			    const struct cred *cred)
4788 {
4789 	struct nfs4_exception exception = {
4790 		.interruptible = true,
4791 	};
4792 	int err;
4793 	do {
4794 		err = _nfs4_proc_access(inode, entry, cred);
4795 		trace_nfs4_access(inode, err);
4796 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4797 				&exception);
4798 	} while (exception.retry);
4799 	return err;
4800 }
4801 
4802 /*
4803  * TODO: For the time being, we don't try to get any attributes
4804  * along with any of the zero-copy operations READ, READDIR,
4805  * READLINK, WRITE.
4806  *
4807  * In the case of the first three, we want to put the GETATTR
4808  * after the read-type operation -- this is because it is hard
4809  * to predict the length of a GETATTR response in v4, and thus
4810  * align the READ data correctly.  This means that the GETATTR
4811  * may end up partially falling into the page cache, and we should
4812  * shift it into the 'tail' of the xdr_buf before processing.
4813  * To do this efficiently, we need to know the total length
4814  * of data received, which doesn't seem to be available outside
4815  * of the RPC layer.
4816  *
4817  * In the case of WRITE, we also want to put the GETATTR after
4818  * the operation -- in this case because we want to make sure
4819  * we get the post-operation mtime and size.
4820  *
4821  * Both of these changes to the XDR layer would in fact be quite
4822  * minor, but I decided to leave them for a subsequent patch.
4823  */
4824 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4825 		unsigned int pgbase, unsigned int pglen)
4826 {
4827 	struct nfs4_readlink args = {
4828 		.fh       = NFS_FH(inode),
4829 		.pgbase	  = pgbase,
4830 		.pglen    = pglen,
4831 		.pages    = &page,
4832 	};
4833 	struct nfs4_readlink_res res;
4834 	struct rpc_message msg = {
4835 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4836 		.rpc_argp = &args,
4837 		.rpc_resp = &res,
4838 	};
4839 
4840 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4841 }
4842 
4843 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4844 		unsigned int pgbase, unsigned int pglen)
4845 {
4846 	struct nfs4_exception exception = {
4847 		.interruptible = true,
4848 	};
4849 	int err;
4850 	do {
4851 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4852 		trace_nfs4_readlink(inode, err);
4853 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4854 				&exception);
4855 	} while (exception.retry);
4856 	return err;
4857 }
4858 
4859 /*
4860  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4861  */
4862 static int
4863 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4864 		 int flags)
4865 {
4866 	struct nfs_server *server = NFS_SERVER(dir);
4867 	struct nfs4_label l, *ilabel;
4868 	struct nfs_open_context *ctx;
4869 	struct nfs4_state *state;
4870 	int status = 0;
4871 
4872 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4873 	if (IS_ERR(ctx))
4874 		return PTR_ERR(ctx);
4875 
4876 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4877 
4878 	nfs_request_directory_delegation(dir);
4879 
4880 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4881 		sattr->ia_mode &= ~current_umask();
4882 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4883 	if (IS_ERR(state)) {
4884 		status = PTR_ERR(state);
4885 		goto out;
4886 	}
4887 out:
4888 	nfs4_label_release_security(ilabel);
4889 	put_nfs_open_context(ctx);
4890 	return status;
4891 }
4892 
4893 static int
4894 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4895 {
4896 	struct nfs_server *server = NFS_SERVER(dir);
4897 	struct nfs_removeargs args = {
4898 		.fh = NFS_FH(dir),
4899 		.name = *name,
4900 	};
4901 	struct nfs_removeres res = {
4902 		.server = server,
4903 	};
4904 	struct rpc_message msg = {
4905 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4906 		.rpc_argp = &args,
4907 		.rpc_resp = &res,
4908 	};
4909 	unsigned long timestamp = jiffies;
4910 	int status;
4911 
4912 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4913 	if (status == 0) {
4914 		spin_lock(&dir->i_lock);
4915 		/* Removing a directory decrements nlink in the parent */
4916 		if (ftype == NF4DIR && dir->i_nlink > 2)
4917 			nfs4_dec_nlink_locked(dir);
4918 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4919 					      NFS_INO_INVALID_DATA);
4920 		spin_unlock(&dir->i_lock);
4921 	}
4922 	return status;
4923 }
4924 
4925 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4926 {
4927 	struct nfs4_exception exception = {
4928 		.interruptible = true,
4929 	};
4930 	struct inode *inode = d_inode(dentry);
4931 	int err;
4932 
4933 	if (inode) {
4934 		if (inode->i_nlink == 1)
4935 			nfs4_inode_return_delegation(inode);
4936 		else
4937 			nfs4_inode_make_writeable(inode);
4938 	}
4939 	do {
4940 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4941 		trace_nfs4_remove(dir, &dentry->d_name, err);
4942 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4943 				&exception);
4944 	} while (exception.retry);
4945 	return err;
4946 }
4947 
4948 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4949 {
4950 	struct nfs4_exception exception = {
4951 		.interruptible = true,
4952 	};
4953 	int err;
4954 
4955 	do {
4956 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4957 		trace_nfs4_remove(dir, name, err);
4958 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4959 				&exception);
4960 	} while (exception.retry);
4961 	return err;
4962 }
4963 
4964 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4965 		struct dentry *dentry,
4966 		struct inode *inode)
4967 {
4968 	struct nfs_removeargs *args = msg->rpc_argp;
4969 	struct nfs_removeres *res = msg->rpc_resp;
4970 	struct nfs_server *server = NFS_SB(dentry->d_sb);
4971 
4972 	res->server = server;
4973 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4974 	nfs4_init_sequence(server->nfs_client, &args->seq_args,
4975 			   &res->seq_res, 1, 0);
4976 
4977 	nfs_fattr_init(res->dir_attr);
4978 	nfs_request_directory_delegation(d_inode(dentry->d_parent));
4979 
4980 	if (inode) {
4981 		nfs4_inode_return_delegation(inode);
4982 		nfs_d_prune_case_insensitive_aliases(inode);
4983 	}
4984 }
4985 
4986 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4987 {
4988 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4989 			&data->args.seq_args,
4990 			&data->res.seq_res,
4991 			task);
4992 }
4993 
4994 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4995 {
4996 	struct nfs_unlinkdata *data = task->tk_calldata;
4997 	struct nfs_removeres *res = &data->res;
4998 
4999 	if (!nfs4_sequence_done(task, &res->seq_res))
5000 		return 0;
5001 	if (nfs4_async_handle_error(task, res->server, NULL,
5002 				    &data->timeout) == -EAGAIN)
5003 		return 0;
5004 	if (task->tk_status == 0)
5005 		nfs4_update_changeattr(dir, &res->cinfo,
5006 				res->dir_attr->time_start,
5007 				NFS_INO_INVALID_DATA);
5008 	return 1;
5009 }
5010 
5011 static void nfs4_proc_rename_setup(struct rpc_message *msg,
5012 		struct dentry *old_dentry,
5013 		struct dentry *new_dentry,
5014 		struct inode *same_parent)
5015 {
5016 	struct nfs_server *server = NFS_SB(old_dentry->d_sb);
5017 	struct nfs_renameargs *arg = msg->rpc_argp;
5018 	struct nfs_renameres *res = msg->rpc_resp;
5019 	struct inode *old_inode = d_inode(old_dentry);
5020 	struct inode *new_inode = d_inode(new_dentry);
5021 
5022 	if (old_inode)
5023 		nfs4_inode_make_writeable(old_inode);
5024 	if (new_inode)
5025 		nfs4_inode_return_delegation(new_inode);
5026 	if (same_parent)
5027 		nfs_request_directory_delegation(same_parent);
5028 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
5029 	res->server = server;
5030 	nfs4_init_sequence(server->nfs_client, &arg->seq_args,
5031 			   &res->seq_res, 1, 0);
5032 }
5033 
5034 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
5035 {
5036 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
5037 			&data->args.seq_args,
5038 			&data->res.seq_res,
5039 			task);
5040 }
5041 
5042 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
5043 				 struct inode *new_dir)
5044 {
5045 	struct nfs_renamedata *data = task->tk_calldata;
5046 	struct nfs_renameres *res = &data->res;
5047 
5048 	if (!nfs4_sequence_done(task, &res->seq_res))
5049 		return 0;
5050 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
5051 		return 0;
5052 
5053 	if (task->tk_status == 0) {
5054 		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
5055 		if (new_dir != old_dir) {
5056 			/* Note: If we moved a directory, nlink will change */
5057 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
5058 					res->old_fattr->time_start,
5059 					NFS_INO_INVALID_NLINK |
5060 					    NFS_INO_INVALID_DATA);
5061 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
5062 					res->new_fattr->time_start,
5063 					NFS_INO_INVALID_NLINK |
5064 					    NFS_INO_INVALID_DATA);
5065 			nfs_update_delegated_mtime(new_dir);
5066 		} else
5067 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
5068 					res->old_fattr->time_start,
5069 					NFS_INO_INVALID_DATA);
5070 	}
5071 	return 1;
5072 }
5073 
5074 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
5075 {
5076 	struct nfs_server *server = NFS_SERVER(inode);
5077 	__u32 bitmask[NFS4_BITMASK_SZ];
5078 	struct nfs4_link_arg arg = {
5079 		.fh     = NFS_FH(inode),
5080 		.dir_fh = NFS_FH(dir),
5081 		.name   = name,
5082 		.bitmask = bitmask,
5083 	};
5084 	struct nfs4_link_res res = {
5085 		.server = server,
5086 	};
5087 	struct rpc_message msg = {
5088 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
5089 		.rpc_argp = &arg,
5090 		.rpc_resp = &res,
5091 	};
5092 	int status = -ENOMEM;
5093 
5094 	res.fattr = nfs_alloc_fattr_with_label(server);
5095 	if (res.fattr == NULL)
5096 		goto out;
5097 
5098 	nfs4_inode_make_writeable(inode);
5099 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label),
5100 				inode,
5101 				NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME);
5102 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5103 	if (!status) {
5104 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
5105 				       NFS_INO_INVALID_DATA);
5106 		nfs4_inc_nlink(inode);
5107 		status = nfs_post_op_update_inode(inode, res.fattr);
5108 		if (!status)
5109 			nfs_setsecurity(inode, res.fattr);
5110 	}
5111 
5112 out:
5113 	nfs_free_fattr(res.fattr);
5114 	return status;
5115 }
5116 
5117 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
5118 {
5119 	struct nfs4_exception exception = {
5120 		.interruptible = true,
5121 	};
5122 	int err;
5123 	do {
5124 		err = nfs4_handle_exception(NFS_SERVER(inode),
5125 				_nfs4_proc_link(inode, dir, name),
5126 				&exception);
5127 	} while (exception.retry);
5128 	return err;
5129 }
5130 
5131 struct nfs4_createdata {
5132 	struct rpc_message msg;
5133 	struct nfs4_create_arg arg;
5134 	struct nfs4_create_res res;
5135 	struct nfs_fh fh;
5136 	struct nfs_fattr fattr;
5137 };
5138 
5139 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
5140 		const struct qstr *name, struct iattr *sattr, u32 ftype)
5141 {
5142 	struct nfs4_createdata *data;
5143 
5144 	data = kzalloc_obj(*data);
5145 	if (data != NULL) {
5146 		struct nfs_server *server = NFS_SERVER(dir);
5147 
5148 		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
5149 		if (IS_ERR(data->fattr.label))
5150 			goto out_free;
5151 
5152 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
5153 		data->msg.rpc_argp = &data->arg;
5154 		data->msg.rpc_resp = &data->res;
5155 		data->arg.dir_fh = NFS_FH(dir);
5156 		data->arg.server = server;
5157 		data->arg.name = name;
5158 		data->arg.attrs = sattr;
5159 		data->arg.ftype = ftype;
5160 		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
5161 		data->arg.umask = current_umask();
5162 		data->res.server = server;
5163 		data->res.fh = &data->fh;
5164 		data->res.fattr = &data->fattr;
5165 		nfs_fattr_init(data->res.fattr);
5166 	}
5167 	return data;
5168 out_free:
5169 	kfree(data);
5170 	return NULL;
5171 }
5172 
5173 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
5174 {
5175 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
5176 				    &data->arg.seq_args, &data->res.seq_res, 1);
5177 	if (status == 0) {
5178 		spin_lock(&dir->i_lock);
5179 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
5180 					      data->res.fattr->time_start,
5181 					      NFS_INO_INVALID_DATA);
5182 		spin_unlock(&dir->i_lock);
5183 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
5184 	}
5185 	return status;
5186 }
5187 
5188 static struct dentry *nfs4_do_mkdir(struct inode *dir, struct dentry *dentry,
5189 				    struct nfs4_createdata *data, int *statusp)
5190 {
5191 	struct dentry *ret;
5192 
5193 	*statusp = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
5194 				    &data->arg.seq_args, &data->res.seq_res, 1);
5195 
5196 	if (*statusp)
5197 		return NULL;
5198 
5199 	spin_lock(&dir->i_lock);
5200 	/* Creating a directory bumps nlink in the parent */
5201 	nfs4_inc_nlink_locked(dir);
5202 	nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
5203 				      data->res.fattr->time_start,
5204 				      NFS_INO_INVALID_DATA);
5205 	spin_unlock(&dir->i_lock);
5206 	ret = nfs_add_or_obtain(dentry, data->res.fh, data->res.fattr);
5207 	if (!IS_ERR(ret))
5208 		return ret;
5209 	*statusp = PTR_ERR(ret);
5210 	return NULL;
5211 }
5212 
5213 static void nfs4_free_createdata(struct nfs4_createdata *data)
5214 {
5215 	nfs4_label_free(data->fattr.label);
5216 	kfree(data);
5217 }
5218 
5219 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5220 		struct folio *folio, unsigned int len, struct iattr *sattr,
5221 		struct nfs4_label *label)
5222 {
5223 	struct page *page = &folio->page;
5224 	struct nfs4_createdata *data;
5225 	int status = -ENAMETOOLONG;
5226 
5227 	if (len > NFS4_MAXPATHLEN)
5228 		goto out;
5229 
5230 	status = -ENOMEM;
5231 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5232 	if (data == NULL)
5233 		goto out;
5234 
5235 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5236 	data->arg.u.symlink.pages = &page;
5237 	data->arg.u.symlink.len = len;
5238 	data->arg.label = label;
5239 
5240 	status = nfs4_do_create(dir, dentry, data);
5241 
5242 	nfs4_free_createdata(data);
5243 out:
5244 	return status;
5245 }
5246 
5247 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5248 		struct folio *folio, unsigned int len, struct iattr *sattr)
5249 {
5250 	struct nfs4_exception exception = {
5251 		.interruptible = true,
5252 	};
5253 	struct nfs4_label l, *label;
5254 	int err;
5255 
5256 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5257 
5258 	do {
5259 		err = _nfs4_proc_symlink(dir, dentry, folio, len, sattr, label);
5260 		trace_nfs4_symlink(dir, &dentry->d_name, err);
5261 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5262 				&exception);
5263 	} while (exception.retry);
5264 
5265 	nfs4_label_release_security(label);
5266 	return err;
5267 }
5268 
5269 static struct dentry *_nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5270 				       struct iattr *sattr,
5271 				       struct nfs4_label *label, int *statusp)
5272 {
5273 	struct nfs4_createdata *data;
5274 	struct dentry *ret = NULL;
5275 
5276 	*statusp = -ENOMEM;
5277 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5278 	if (data == NULL)
5279 		goto out;
5280 
5281 	data->arg.label = label;
5282 	ret = nfs4_do_mkdir(dir, dentry, data, statusp);
5283 
5284 	nfs4_free_createdata(data);
5285 out:
5286 	return ret;
5287 }
5288 
5289 static struct dentry *nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5290 				      struct iattr *sattr)
5291 {
5292 	struct nfs_server *server = NFS_SERVER(dir);
5293 	struct nfs4_exception exception = {
5294 		.interruptible = true,
5295 	};
5296 	struct nfs4_label l, *label;
5297 	struct dentry *alias;
5298 	int err;
5299 
5300 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5301 
5302 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5303 		sattr->ia_mode &= ~current_umask();
5304 	do {
5305 		alias = _nfs4_proc_mkdir(dir, dentry, sattr, label, &err);
5306 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5307 		err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
5308 		if (err)
5309 			alias = ERR_PTR(err);
5310 	} while (exception.retry);
5311 	nfs4_label_release_security(label);
5312 
5313 	return alias;
5314 }
5315 
5316 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5317 			      struct nfs_readdir_res *nr_res)
5318 {
5319 	struct inode		*dir = d_inode(nr_arg->dentry);
5320 	struct nfs_server	*server = NFS_SERVER(dir);
5321 	struct nfs4_readdir_arg args = {
5322 		.fh = NFS_FH(dir),
5323 		.pages = nr_arg->pages,
5324 		.pgbase = 0,
5325 		.count = nr_arg->page_len,
5326 		.plus = nr_arg->plus,
5327 	};
5328 	struct nfs4_readdir_res res;
5329 	struct rpc_message msg = {
5330 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5331 		.rpc_argp = &args,
5332 		.rpc_resp = &res,
5333 		.rpc_cred = nr_arg->cred,
5334 	};
5335 	int			status;
5336 
5337 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5338 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5339 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5340 		args.bitmask = server->attr_bitmask_nl;
5341 	else
5342 		args.bitmask = server->attr_bitmask;
5343 
5344 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5345 	res.pgbase = args.pgbase;
5346 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5347 			&res.seq_res, 0);
5348 	if (status >= 0) {
5349 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5350 		status += args.pgbase;
5351 	}
5352 
5353 	nfs_invalidate_atime(dir);
5354 
5355 	dprintk("%s: returns %d\n", __func__, status);
5356 	return status;
5357 }
5358 
5359 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5360 			     struct nfs_readdir_res *res)
5361 {
5362 	struct nfs4_exception exception = {
5363 		.interruptible = true,
5364 	};
5365 	int err;
5366 	do {
5367 		err = _nfs4_proc_readdir(arg, res);
5368 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5369 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5370 					    err, &exception);
5371 	} while (exception.retry);
5372 	return err;
5373 }
5374 
5375 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5376 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5377 {
5378 	struct nfs4_createdata *data;
5379 	int mode = sattr->ia_mode;
5380 	int status = -ENOMEM;
5381 
5382 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5383 	if (data == NULL)
5384 		goto out;
5385 
5386 	if (S_ISFIFO(mode))
5387 		data->arg.ftype = NF4FIFO;
5388 	else if (S_ISBLK(mode)) {
5389 		data->arg.ftype = NF4BLK;
5390 		data->arg.u.device.specdata1 = MAJOR(rdev);
5391 		data->arg.u.device.specdata2 = MINOR(rdev);
5392 	}
5393 	else if (S_ISCHR(mode)) {
5394 		data->arg.ftype = NF4CHR;
5395 		data->arg.u.device.specdata1 = MAJOR(rdev);
5396 		data->arg.u.device.specdata2 = MINOR(rdev);
5397 	} else if (!S_ISSOCK(mode)) {
5398 		status = -EINVAL;
5399 		goto out_free;
5400 	}
5401 
5402 	data->arg.label = label;
5403 	status = nfs4_do_create(dir, dentry, data);
5404 out_free:
5405 	nfs4_free_createdata(data);
5406 out:
5407 	return status;
5408 }
5409 
5410 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5411 		struct iattr *sattr, dev_t rdev)
5412 {
5413 	struct nfs_server *server = NFS_SERVER(dir);
5414 	struct nfs4_exception exception = {
5415 		.interruptible = true,
5416 	};
5417 	struct nfs4_label l, *label;
5418 	int err;
5419 
5420 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5421 
5422 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5423 		sattr->ia_mode &= ~current_umask();
5424 	do {
5425 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5426 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5427 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5428 				&exception);
5429 	} while (exception.retry);
5430 
5431 	nfs4_label_release_security(label);
5432 
5433 	return err;
5434 }
5435 
5436 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5437 		 struct nfs_fsstat *fsstat)
5438 {
5439 	struct nfs4_statfs_arg args = {
5440 		.fh = fhandle,
5441 		.bitmask = server->attr_bitmask,
5442 	};
5443 	struct nfs4_statfs_res res = {
5444 		.fsstat = fsstat,
5445 	};
5446 	struct rpc_message msg = {
5447 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5448 		.rpc_argp = &args,
5449 		.rpc_resp = &res,
5450 	};
5451 
5452 	nfs_fattr_init(fsstat->fattr);
5453 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5454 }
5455 
5456 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5457 {
5458 	struct nfs4_exception exception = {
5459 		.interruptible = true,
5460 	};
5461 	int err;
5462 	do {
5463 		err = nfs4_handle_exception(server,
5464 				_nfs4_proc_statfs(server, fhandle, fsstat),
5465 				&exception);
5466 	} while (exception.retry);
5467 	return err;
5468 }
5469 
5470 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5471 		struct nfs_fsinfo *fsinfo)
5472 {
5473 	struct nfs4_fsinfo_arg args = {
5474 		.fh = fhandle,
5475 		.bitmask = server->attr_bitmask,
5476 	};
5477 	struct nfs4_fsinfo_res res = {
5478 		.fsinfo = fsinfo,
5479 	};
5480 	struct rpc_message msg = {
5481 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5482 		.rpc_argp = &args,
5483 		.rpc_resp = &res,
5484 	};
5485 
5486 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5487 }
5488 
5489 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5490 {
5491 	struct nfs4_exception exception = {
5492 		.interruptible = true,
5493 	};
5494 	int err;
5495 
5496 	do {
5497 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5498 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5499 		if (err == 0) {
5500 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time);
5501 			break;
5502 		}
5503 		err = nfs4_handle_exception(server, err, &exception);
5504 	} while (exception.retry);
5505 	return err;
5506 }
5507 
5508 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5509 {
5510 	int error;
5511 
5512 	nfs_fattr_init(fsinfo->fattr);
5513 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5514 	if (error == 0) {
5515 		/* block layout checks this! */
5516 		server->pnfs_blksize = fsinfo->blksize;
5517 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5518 	}
5519 
5520 	return error;
5521 }
5522 
5523 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5524 		struct nfs_pathconf *pathconf)
5525 {
5526 	struct nfs4_pathconf_arg args = {
5527 		.fh = fhandle,
5528 		.bitmask = server->attr_bitmask,
5529 	};
5530 	struct nfs4_pathconf_res res = {
5531 		.pathconf = pathconf,
5532 	};
5533 	struct rpc_message msg = {
5534 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5535 		.rpc_argp = &args,
5536 		.rpc_resp = &res,
5537 	};
5538 
5539 	/* None of the pathconf attributes are mandatory to implement */
5540 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5541 		memset(pathconf, 0, sizeof(*pathconf));
5542 		return 0;
5543 	}
5544 
5545 	nfs_fattr_init(pathconf->fattr);
5546 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5547 }
5548 
5549 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5550 		struct nfs_pathconf *pathconf)
5551 {
5552 	struct nfs4_exception exception = {
5553 		.interruptible = true,
5554 	};
5555 	int err;
5556 
5557 	do {
5558 		err = nfs4_handle_exception(server,
5559 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5560 				&exception);
5561 	} while (exception.retry);
5562 	return err;
5563 }
5564 
5565 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5566 		const struct nfs_open_context *ctx,
5567 		const struct nfs_lock_context *l_ctx,
5568 		fmode_t fmode)
5569 {
5570 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5571 }
5572 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5573 
5574 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5575 		const struct nfs_open_context *ctx,
5576 		const struct nfs_lock_context *l_ctx,
5577 		fmode_t fmode)
5578 {
5579 	nfs4_stateid _current_stateid;
5580 
5581 	/* If the current stateid represents a lost lock, then exit */
5582 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5583 		return true;
5584 	return nfs4_stateid_match(stateid, &_current_stateid);
5585 }
5586 
5587 static bool nfs4_error_stateid_expired(int err)
5588 {
5589 	switch (err) {
5590 	case -NFS4ERR_DELEG_REVOKED:
5591 	case -NFS4ERR_ADMIN_REVOKED:
5592 	case -NFS4ERR_BAD_STATEID:
5593 	case -NFS4ERR_STALE_STATEID:
5594 	case -NFS4ERR_OLD_STATEID:
5595 	case -NFS4ERR_OPENMODE:
5596 	case -NFS4ERR_EXPIRED:
5597 		return true;
5598 	}
5599 	return false;
5600 }
5601 
5602 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5603 {
5604 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5605 
5606 	trace_nfs4_read(hdr, task->tk_status);
5607 	if (task->tk_status < 0) {
5608 		struct nfs4_exception exception = {
5609 			.inode = hdr->inode,
5610 			.state = hdr->args.context->state,
5611 			.stateid = &hdr->args.stateid,
5612 			.retrans = hdr->retrans,
5613 		};
5614 		task->tk_status = nfs4_async_handle_exception(task,
5615 				server, task->tk_status, &exception);
5616 		hdr->retrans = exception.retrans;
5617 		if (exception.retry) {
5618 			rpc_restart_call_prepare(task);
5619 			return -EAGAIN;
5620 		}
5621 	}
5622 
5623 	if (task->tk_status > 0)
5624 		renew_lease(server, hdr->timestamp);
5625 	return 0;
5626 }
5627 
5628 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5629 		struct nfs_pgio_args *args)
5630 {
5631 
5632 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5633 		nfs4_stateid_is_current(&args->stateid,
5634 				args->context,
5635 				args->lock_context,
5636 				FMODE_READ))
5637 		return false;
5638 	rpc_restart_call_prepare(task);
5639 	return true;
5640 }
5641 
5642 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5643 					 struct nfs_pgio_header *hdr)
5644 {
5645 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5646 	struct rpc_message *msg = &task->tk_msg;
5647 
5648 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5649 	    task->tk_status == -ENOTSUPP) {
5650 		server->caps &= ~NFS_CAP_READ_PLUS;
5651 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5652 		rpc_restart_call_prepare(task);
5653 		return true;
5654 	}
5655 	return false;
5656 }
5657 
5658 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5659 {
5660 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5661 		return -EAGAIN;
5662 	if (nfs4_read_stateid_changed(task, &hdr->args))
5663 		return -EAGAIN;
5664 	if (nfs4_read_plus_not_supported(task, hdr))
5665 		return -EAGAIN;
5666 	if (task->tk_status > 0)
5667 		nfs_invalidate_atime(hdr->inode);
5668 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5669 				    nfs4_read_done_cb(task, hdr);
5670 }
5671 
5672 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5673 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5674 				    struct rpc_message *msg)
5675 {
5676 	/* Note: We don't use READ_PLUS with pNFS yet */
5677 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5678 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5679 		return nfs_read_alloc_scratch(hdr, READ_PLUS_SCRATCH_SIZE);
5680 	}
5681 	return false;
5682 }
5683 #else
5684 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5685 				    struct rpc_message *msg)
5686 {
5687 	return false;
5688 }
5689 #endif /* CONFIG_NFS_V4_2 */
5690 
5691 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5692 				 struct rpc_message *msg)
5693 {
5694 	hdr->timestamp   = jiffies;
5695 	if (!hdr->pgio_done_cb)
5696 		hdr->pgio_done_cb = nfs4_read_done_cb;
5697 	if (!nfs42_read_plus_support(hdr, msg))
5698 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5699 	nfs4_init_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5700 			   &hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5701 }
5702 
5703 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5704 				      struct nfs_pgio_header *hdr)
5705 {
5706 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5707 			&hdr->args.seq_args,
5708 			&hdr->res.seq_res,
5709 			task))
5710 		return 0;
5711 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5712 				hdr->args.lock_context,
5713 				hdr->rw_mode) == -EIO)
5714 		return -EIO;
5715 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5716 		return -EIO;
5717 	return 0;
5718 }
5719 
5720 static int nfs4_write_done_cb(struct rpc_task *task,
5721 			      struct nfs_pgio_header *hdr)
5722 {
5723 	struct inode *inode = hdr->inode;
5724 
5725 	trace_nfs4_write(hdr, task->tk_status);
5726 	if (task->tk_status < 0) {
5727 		struct nfs4_exception exception = {
5728 			.inode = hdr->inode,
5729 			.state = hdr->args.context->state,
5730 			.stateid = &hdr->args.stateid,
5731 			.retrans = hdr->retrans,
5732 		};
5733 		task->tk_status = nfs4_async_handle_exception(task,
5734 				NFS_SERVER(inode), task->tk_status,
5735 				&exception);
5736 		hdr->retrans = exception.retrans;
5737 		if (exception.retry) {
5738 			rpc_restart_call_prepare(task);
5739 			return -EAGAIN;
5740 		}
5741 	}
5742 	if (task->tk_status >= 0) {
5743 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5744 		nfs_writeback_update_inode(hdr);
5745 	}
5746 	return 0;
5747 }
5748 
5749 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5750 		struct nfs_pgio_args *args)
5751 {
5752 
5753 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5754 		nfs4_stateid_is_current(&args->stateid,
5755 				args->context,
5756 				args->lock_context,
5757 				FMODE_WRITE))
5758 		return false;
5759 	rpc_restart_call_prepare(task);
5760 	return true;
5761 }
5762 
5763 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5764 {
5765 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5766 		return -EAGAIN;
5767 	if (nfs4_write_stateid_changed(task, &hdr->args))
5768 		return -EAGAIN;
5769 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5770 		nfs4_write_done_cb(task, hdr);
5771 }
5772 
5773 static
5774 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5775 {
5776 	/* Don't request attributes for pNFS or O_DIRECT writes */
5777 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5778 		return false;
5779 	/* Otherwise, request attributes if and only if we don't hold
5780 	 * a delegation
5781 	 */
5782 	return nfs4_have_delegation(hdr->inode, FMODE_READ, 0) == 0;
5783 }
5784 
5785 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5786 		      struct inode *inode, unsigned long cache_validity)
5787 {
5788 	struct nfs_server *server = NFS_SERVER(inode);
5789 	unsigned int i;
5790 
5791 	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5792 	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5793 
5794 	if (cache_validity & NFS_INO_INVALID_CHANGE)
5795 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5796 	if (cache_validity & NFS_INO_INVALID_ATIME)
5797 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5798 	if (cache_validity & NFS_INO_INVALID_MODE)
5799 		bitmask[1] |= FATTR4_WORD1_MODE;
5800 	if (cache_validity & NFS_INO_INVALID_OTHER)
5801 		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5802 	if (cache_validity & NFS_INO_INVALID_NLINK)
5803 		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5804 	if (cache_validity & NFS_INO_INVALID_CTIME)
5805 		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5806 	if (cache_validity & NFS_INO_INVALID_MTIME)
5807 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5808 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5809 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5810 	if (cache_validity & NFS_INO_INVALID_BTIME)
5811 		bitmask[1] |= FATTR4_WORD1_TIME_CREATE;
5812 
5813 	if (cache_validity & NFS_INO_INVALID_SIZE)
5814 		bitmask[0] |= FATTR4_WORD0_SIZE;
5815 
5816 	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5817 		bitmask[i] &= server->attr_bitmask[i];
5818 }
5819 
5820 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5821 				  struct rpc_message *msg,
5822 				  struct rpc_clnt **clnt)
5823 {
5824 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5825 
5826 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5827 		hdr->args.bitmask = NULL;
5828 		hdr->res.fattr = NULL;
5829 	} else {
5830 		nfs4_bitmask_set(hdr->args.bitmask_store,
5831 				 server->cache_consistency_bitmask,
5832 				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5833 		hdr->args.bitmask = hdr->args.bitmask_store;
5834 	}
5835 
5836 	if (!hdr->pgio_done_cb)
5837 		hdr->pgio_done_cb = nfs4_write_done_cb;
5838 	hdr->res.server = server;
5839 	hdr->timestamp   = jiffies;
5840 
5841 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5842 	nfs4_init_sequence(server->nfs_client, &hdr->args.seq_args,
5843 			   &hdr->res.seq_res, 0, 0);
5844 	nfs4_state_protect_write(hdr->ds_clp ? hdr->ds_clp : server->nfs_client, clnt, msg, hdr);
5845 }
5846 
5847 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5848 {
5849 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5850 			&data->args.seq_args,
5851 			&data->res.seq_res,
5852 			task);
5853 }
5854 
5855 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5856 {
5857 	struct inode *inode = data->inode;
5858 
5859 	trace_nfs4_commit(data, task->tk_status);
5860 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5861 				    NULL, NULL) == -EAGAIN) {
5862 		rpc_restart_call_prepare(task);
5863 		return -EAGAIN;
5864 	}
5865 	return 0;
5866 }
5867 
5868 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5869 {
5870 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5871 		return -EAGAIN;
5872 	return data->commit_done_cb(task, data);
5873 }
5874 
5875 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5876 				   struct rpc_clnt **clnt)
5877 {
5878 	struct nfs_server *server = NFS_SERVER(data->inode);
5879 
5880 	if (data->commit_done_cb == NULL)
5881 		data->commit_done_cb = nfs4_commit_done_cb;
5882 	data->res.server = server;
5883 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5884 	nfs4_init_sequence(server->nfs_client, &data->args.seq_args,
5885 			   &data->res.seq_res, 1, 0);
5886 	nfs4_state_protect(data->ds_clp ? data->ds_clp : server->nfs_client,
5887 			NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5888 }
5889 
5890 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5891 				struct nfs_commitres *res)
5892 {
5893 	struct inode *dst_inode = file_inode(dst);
5894 	struct nfs_server *server = NFS_SERVER(dst_inode);
5895 	struct rpc_message msg = {
5896 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5897 		.rpc_argp = args,
5898 		.rpc_resp = res,
5899 	};
5900 
5901 	args->fh = NFS_FH(dst_inode);
5902 	return nfs4_call_sync(server->client, server, &msg,
5903 			&args->seq_args, &res->seq_res, 1);
5904 }
5905 
5906 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5907 {
5908 	struct nfs_commitargs args = {
5909 		.offset = offset,
5910 		.count = count,
5911 	};
5912 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5913 	struct nfs4_exception exception = { };
5914 	int status;
5915 
5916 	do {
5917 		status = _nfs4_proc_commit(dst, &args, res);
5918 		status = nfs4_handle_exception(dst_server, status, &exception);
5919 	} while (exception.retry);
5920 
5921 	return status;
5922 }
5923 
5924 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5925 				      enum nfs4_acl_type type)
5926 {
5927 	switch (type) {
5928 	default:
5929 		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5930 	case NFS4ACL_DACL:
5931 		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5932 	case NFS4ACL_SACL:
5933 		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5934 	}
5935 }
5936 
5937 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5938  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5939  * the stack.
5940  */
5941 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5942 
5943 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5944 		struct page **pages)
5945 {
5946 	struct page *newpage, **spages;
5947 	int rc = 0;
5948 	size_t len;
5949 	spages = pages;
5950 
5951 	do {
5952 		len = min_t(size_t, PAGE_SIZE, buflen);
5953 		newpage = alloc_page(GFP_KERNEL);
5954 
5955 		if (newpage == NULL)
5956 			goto unwind;
5957 		memcpy(page_address(newpage), buf, len);
5958 		buf += len;
5959 		buflen -= len;
5960 		*pages++ = newpage;
5961 		rc++;
5962 	} while (buflen != 0);
5963 
5964 	return rc;
5965 
5966 unwind:
5967 	for(; rc > 0; rc--)
5968 		__free_page(spages[rc-1]);
5969 	return -ENOMEM;
5970 }
5971 
5972 struct nfs4_cached_acl {
5973 	enum nfs4_acl_type type;
5974 	int cached;
5975 	size_t len;
5976 	char data[];
5977 };
5978 
5979 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5980 {
5981 	struct nfs_inode *nfsi = NFS_I(inode);
5982 
5983 	spin_lock(&inode->i_lock);
5984 	kfree(nfsi->nfs4_acl);
5985 	nfsi->nfs4_acl = acl;
5986 	spin_unlock(&inode->i_lock);
5987 }
5988 
5989 static void nfs4_zap_acl_attr(struct inode *inode)
5990 {
5991 	nfs4_set_cached_acl(inode, NULL);
5992 }
5993 
5994 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5995 				    size_t buflen, enum nfs4_acl_type type)
5996 {
5997 	struct nfs_inode *nfsi = NFS_I(inode);
5998 	struct nfs4_cached_acl *acl;
5999 	int ret = -ENOENT;
6000 
6001 	spin_lock(&inode->i_lock);
6002 	acl = nfsi->nfs4_acl;
6003 	if (acl == NULL)
6004 		goto out;
6005 	if (acl->type != type)
6006 		goto out;
6007 	if (buf == NULL) /* user is just asking for length */
6008 		goto out_len;
6009 	if (acl->cached == 0)
6010 		goto out;
6011 	ret = -ERANGE; /* see getxattr(2) man page */
6012 	if (acl->len > buflen)
6013 		goto out;
6014 	memcpy(buf, acl->data, acl->len);
6015 out_len:
6016 	ret = acl->len;
6017 out:
6018 	spin_unlock(&inode->i_lock);
6019 	return ret;
6020 }
6021 
6022 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
6023 				  size_t pgbase, size_t acl_len,
6024 				  enum nfs4_acl_type type)
6025 {
6026 	struct nfs4_cached_acl *acl;
6027 	size_t buflen = sizeof(*acl) + acl_len;
6028 
6029 	if (buflen <= PAGE_SIZE) {
6030 		acl = kmalloc(buflen, GFP_KERNEL);
6031 		if (acl == NULL)
6032 			goto out;
6033 		acl->cached = 1;
6034 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
6035 	} else {
6036 		acl = kmalloc_obj(*acl);
6037 		if (acl == NULL)
6038 			goto out;
6039 		acl->cached = 0;
6040 	}
6041 	acl->type = type;
6042 	acl->len = acl_len;
6043 out:
6044 	nfs4_set_cached_acl(inode, acl);
6045 }
6046 
6047 /*
6048  * The getxattr API returns the required buffer length when called with a
6049  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
6050  * the required buf.  On a NULL buf, we send a page of data to the server
6051  * guessing that the ACL request can be serviced by a page. If so, we cache
6052  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
6053  * the cache. If not so, we throw away the page, and cache the required
6054  * length. The next getxattr call will then produce another round trip to
6055  * the server, this time with the input buf of the required size.
6056  */
6057 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
6058 				       size_t buflen, enum nfs4_acl_type type)
6059 {
6060 	struct page **pages;
6061 	struct nfs_getaclargs args = {
6062 		.fh = NFS_FH(inode),
6063 		.acl_type = type,
6064 		.acl_len = buflen,
6065 	};
6066 	struct nfs_getaclres res = {
6067 		.acl_type = type,
6068 		.acl_len = buflen,
6069 	};
6070 	struct rpc_message msg = {
6071 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
6072 		.rpc_argp = &args,
6073 		.rpc_resp = &res,
6074 	};
6075 	unsigned int npages;
6076 	int ret = -ENOMEM, i;
6077 	struct nfs_server *server = NFS_SERVER(inode);
6078 
6079 	if (buflen == 0)
6080 		buflen = server->rsize;
6081 
6082 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
6083 	pages = kmalloc_objs(struct page *, npages);
6084 	if (!pages)
6085 		return -ENOMEM;
6086 
6087 	args.acl_pages = pages;
6088 
6089 	for (i = 0; i < npages; i++) {
6090 		pages[i] = alloc_page(GFP_KERNEL);
6091 		if (!pages[i])
6092 			goto out_free;
6093 	}
6094 
6095 	/* for decoding across pages */
6096 	res.acl_scratch = folio_alloc(GFP_KERNEL, 0);
6097 	if (!res.acl_scratch)
6098 		goto out_free;
6099 
6100 	args.acl_len = npages * PAGE_SIZE;
6101 
6102 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
6103 		__func__, buf, buflen, npages, args.acl_len);
6104 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
6105 			     &msg, &args.seq_args, &res.seq_res, 0);
6106 	if (ret)
6107 		goto out_free;
6108 
6109 	/* Handle the case where the passed-in buffer is too short */
6110 	if (res.acl_flags & NFS4_ACL_TRUNC) {
6111 		/* Did the user only issue a request for the acl length? */
6112 		if (buf == NULL)
6113 			goto out_ok;
6114 		ret = -ERANGE;
6115 		goto out_free;
6116 	}
6117 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
6118 			      type);
6119 	if (buf) {
6120 		if (res.acl_len > buflen) {
6121 			ret = -ERANGE;
6122 			goto out_free;
6123 		}
6124 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
6125 	}
6126 out_ok:
6127 	ret = res.acl_len;
6128 out_free:
6129 	while (--i >= 0)
6130 		__free_page(pages[i]);
6131 	if (res.acl_scratch)
6132 		folio_put(res.acl_scratch);
6133 	kfree(pages);
6134 	return ret;
6135 }
6136 
6137 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6138 				     size_t buflen, enum nfs4_acl_type type)
6139 {
6140 	struct nfs4_exception exception = {
6141 		.interruptible = true,
6142 	};
6143 	ssize_t ret;
6144 	do {
6145 		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6146 		trace_nfs4_get_acl(inode, ret);
6147 		if (ret >= 0)
6148 			break;
6149 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6150 	} while (exception.retry);
6151 	return ret;
6152 }
6153 
6154 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6155 				 enum nfs4_acl_type type)
6156 {
6157 	struct nfs_server *server = NFS_SERVER(inode);
6158 	int ret;
6159 
6160 	if (unlikely(NFS_FH(inode)->size == 0))
6161 		return -ENODATA;
6162 	if (!nfs4_server_supports_acls(server, type))
6163 		return -EOPNOTSUPP;
6164 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6165 	if (ret < 0)
6166 		return ret;
6167 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6168 		nfs_zap_acl_cache(inode);
6169 	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6170 	if (ret != -ENOENT)
6171 		/* -ENOENT is returned if there is no ACL or if there is an ACL
6172 		 * but no cached acl data, just the acl length */
6173 		return ret;
6174 	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6175 }
6176 
6177 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6178 			       size_t buflen, enum nfs4_acl_type type)
6179 {
6180 	struct nfs_server *server = NFS_SERVER(inode);
6181 	struct page *pages[NFS4ACL_MAXPAGES];
6182 	struct nfs_setaclargs arg = {
6183 		.fh = NFS_FH(inode),
6184 		.acl_type = type,
6185 		.acl_len = buflen,
6186 		.acl_pages = pages,
6187 	};
6188 	struct nfs_setaclres res;
6189 	struct rpc_message msg = {
6190 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6191 		.rpc_argp	= &arg,
6192 		.rpc_resp	= &res,
6193 	};
6194 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6195 	int ret, i;
6196 
6197 	/* You can't remove system.nfs4_acl: */
6198 	if (buflen == 0)
6199 		return -EINVAL;
6200 	if (!nfs4_server_supports_acls(server, type))
6201 		return -EOPNOTSUPP;
6202 	if (npages > ARRAY_SIZE(pages))
6203 		return -ERANGE;
6204 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6205 	if (i < 0)
6206 		return i;
6207 	nfs4_inode_make_writeable(inode);
6208 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6209 
6210 	/*
6211 	 * Free each page after tx, so the only ref left is
6212 	 * held by the network stack
6213 	 */
6214 	for (; i > 0; i--)
6215 		put_page(pages[i-1]);
6216 
6217 	/*
6218 	 * Acl update can result in inode attribute update.
6219 	 * so mark the attribute cache invalid.
6220 	 */
6221 	spin_lock(&inode->i_lock);
6222 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6223 					     NFS_INO_INVALID_CTIME |
6224 					     NFS_INO_REVAL_FORCED);
6225 	spin_unlock(&inode->i_lock);
6226 	nfs_access_zap_cache(inode);
6227 	nfs_zap_acl_cache(inode);
6228 	return ret;
6229 }
6230 
6231 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6232 			     size_t buflen, enum nfs4_acl_type type)
6233 {
6234 	struct nfs4_exception exception = { };
6235 	int err;
6236 
6237 	if (unlikely(NFS_FH(inode)->size == 0))
6238 		return -ENODATA;
6239 	do {
6240 		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6241 		trace_nfs4_set_acl(inode, err);
6242 		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6243 			/*
6244 			 * no need to retry since the kernel
6245 			 * isn't involved in encoding the ACEs.
6246 			 */
6247 			err = -EINVAL;
6248 			break;
6249 		}
6250 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6251 				&exception);
6252 	} while (exception.retry);
6253 	return err;
6254 }
6255 
6256 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6257 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6258 					size_t buflen)
6259 {
6260 	struct nfs_server *server = NFS_SERVER(inode);
6261 	struct nfs4_label label = {0, 0, 0, buflen, buf};
6262 
6263 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6264 	struct nfs_fattr fattr = {
6265 		.label = &label,
6266 	};
6267 	struct nfs4_getattr_arg arg = {
6268 		.fh		= NFS_FH(inode),
6269 		.bitmask	= bitmask,
6270 	};
6271 	struct nfs4_getattr_res res = {
6272 		.fattr		= &fattr,
6273 		.server		= server,
6274 	};
6275 	struct rpc_message msg = {
6276 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6277 		.rpc_argp	= &arg,
6278 		.rpc_resp	= &res,
6279 	};
6280 	int ret;
6281 
6282 	nfs_fattr_init(&fattr);
6283 
6284 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6285 	if (ret)
6286 		return ret;
6287 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6288 		return -ENOENT;
6289 	return label.len;
6290 }
6291 
6292 static int nfs4_get_security_label(struct inode *inode, void *buf,
6293 					size_t buflen)
6294 {
6295 	struct nfs4_exception exception = {
6296 		.interruptible = true,
6297 	};
6298 	int err;
6299 
6300 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6301 		return -EOPNOTSUPP;
6302 
6303 	do {
6304 		err = _nfs4_get_security_label(inode, buf, buflen);
6305 		trace_nfs4_get_security_label(inode, err);
6306 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6307 				&exception);
6308 	} while (exception.retry);
6309 	return err;
6310 }
6311 
6312 static int _nfs4_do_set_security_label(struct inode *inode,
6313 		struct nfs4_label *ilabel,
6314 		struct nfs_fattr *fattr)
6315 {
6316 
6317 	struct iattr sattr = {0};
6318 	struct nfs_server *server = NFS_SERVER(inode);
6319 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6320 	struct nfs_setattrargs arg = {
6321 		.fh		= NFS_FH(inode),
6322 		.iap		= &sattr,
6323 		.server		= server,
6324 		.bitmask	= bitmask,
6325 		.label		= ilabel,
6326 	};
6327 	struct nfs_setattrres res = {
6328 		.fattr		= fattr,
6329 		.server		= server,
6330 	};
6331 	struct rpc_message msg = {
6332 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6333 		.rpc_argp	= &arg,
6334 		.rpc_resp	= &res,
6335 	};
6336 	int status;
6337 
6338 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6339 
6340 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6341 	if (status)
6342 		dprintk("%s failed: %d\n", __func__, status);
6343 
6344 	return status;
6345 }
6346 
6347 static int nfs4_do_set_security_label(struct inode *inode,
6348 		struct nfs4_label *ilabel,
6349 		struct nfs_fattr *fattr)
6350 {
6351 	struct nfs4_exception exception = { };
6352 	int err;
6353 
6354 	do {
6355 		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6356 		trace_nfs4_set_security_label(inode, err);
6357 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6358 				&exception);
6359 	} while (exception.retry);
6360 	return err;
6361 }
6362 
6363 static int
6364 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6365 {
6366 	struct nfs4_label ilabel = {0, 0, 0, buflen, (char *)buf };
6367 	struct nfs_fattr *fattr;
6368 	int status;
6369 
6370 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6371 		return -EOPNOTSUPP;
6372 
6373 	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6374 	if (fattr == NULL)
6375 		return -ENOMEM;
6376 
6377 	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6378 	if (status == 0)
6379 		nfs_setsecurity(inode, fattr);
6380 
6381 	nfs_free_fattr(fattr);
6382 	return status;
6383 }
6384 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6385 
6386 
6387 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6388 				    nfs4_verifier *bootverf)
6389 {
6390 	__be32 verf[2];
6391 
6392 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6393 		/* An impossible timestamp guarantees this value
6394 		 * will never match a generated boot time. */
6395 		verf[0] = cpu_to_be32(U32_MAX);
6396 		verf[1] = cpu_to_be32(U32_MAX);
6397 	} else {
6398 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6399 		u64 ns = ktime_to_ns(nn->boot_time);
6400 
6401 		verf[0] = cpu_to_be32(ns >> 32);
6402 		verf[1] = cpu_to_be32(ns);
6403 	}
6404 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6405 }
6406 
6407 static size_t
6408 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6409 {
6410 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6411 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6412 	const char *id;
6413 
6414 	buf[0] = '\0';
6415 
6416 	if (nn_clp) {
6417 		rcu_read_lock();
6418 		id = rcu_dereference(nn_clp->identifier);
6419 		if (id)
6420 			strscpy(buf, id, buflen);
6421 		rcu_read_unlock();
6422 	}
6423 
6424 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6425 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6426 
6427 	return strlen(buf);
6428 }
6429 
6430 static int
6431 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6432 {
6433 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6434 	size_t buflen;
6435 	size_t len;
6436 	char *str;
6437 
6438 	if (clp->cl_owner_id != NULL)
6439 		return 0;
6440 
6441 	rcu_read_lock();
6442 	len = 14 +
6443 		strlen(clp->cl_rpcclient->cl_nodename) +
6444 		1 +
6445 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6446 		1;
6447 	rcu_read_unlock();
6448 
6449 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6450 	if (buflen)
6451 		len += buflen + 1;
6452 
6453 	if (len > NFS4_OPAQUE_LIMIT + 1)
6454 		return -EINVAL;
6455 
6456 	/*
6457 	 * Since this string is allocated at mount time, and held until the
6458 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6459 	 * about a memory-reclaim deadlock.
6460 	 */
6461 	str = kmalloc(len, GFP_KERNEL);
6462 	if (!str)
6463 		return -ENOMEM;
6464 
6465 	rcu_read_lock();
6466 	if (buflen)
6467 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6468 			  clp->cl_rpcclient->cl_nodename, buf,
6469 			  rpc_peeraddr2str(clp->cl_rpcclient,
6470 					   RPC_DISPLAY_ADDR));
6471 	else
6472 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6473 			  clp->cl_rpcclient->cl_nodename,
6474 			  rpc_peeraddr2str(clp->cl_rpcclient,
6475 					   RPC_DISPLAY_ADDR));
6476 	rcu_read_unlock();
6477 
6478 	clp->cl_owner_id = str;
6479 	return 0;
6480 }
6481 
6482 static int
6483 nfs4_init_uniform_client_string(struct nfs_client *clp)
6484 {
6485 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6486 	size_t buflen;
6487 	size_t len;
6488 	char *str;
6489 
6490 	if (clp->cl_owner_id != NULL)
6491 		return 0;
6492 
6493 	len = 10 + 10 + 1 + 10 + 1 +
6494 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6495 
6496 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6497 	if (buflen)
6498 		len += buflen + 1;
6499 
6500 	if (len > NFS4_OPAQUE_LIMIT + 1)
6501 		return -EINVAL;
6502 
6503 	/*
6504 	 * Since this string is allocated at mount time, and held until the
6505 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6506 	 * about a memory-reclaim deadlock.
6507 	 */
6508 	str = kmalloc(len, GFP_KERNEL);
6509 	if (!str)
6510 		return -ENOMEM;
6511 
6512 	if (buflen)
6513 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6514 			  clp->rpc_ops->version, clp->cl_minorversion,
6515 			  buf, clp->cl_rpcclient->cl_nodename);
6516 	else
6517 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6518 			  clp->rpc_ops->version, clp->cl_minorversion,
6519 			  clp->cl_rpcclient->cl_nodename);
6520 	clp->cl_owner_id = str;
6521 	return 0;
6522 }
6523 
6524 /*
6525  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6526  * services.  Advertise one based on the address family of the
6527  * clientaddr.
6528  */
6529 static unsigned int
6530 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6531 {
6532 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6533 		return scnprintf(buf, len, "tcp6");
6534 	else
6535 		return scnprintf(buf, len, "tcp");
6536 }
6537 
6538 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6539 {
6540 	struct nfs4_setclientid *sc = calldata;
6541 
6542 	if (task->tk_status == 0)
6543 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6544 }
6545 
6546 static const struct rpc_call_ops nfs4_setclientid_ops = {
6547 	.rpc_call_done = nfs4_setclientid_done,
6548 };
6549 
6550 /**
6551  * nfs4_proc_setclientid - Negotiate client ID
6552  * @clp: state data structure
6553  * @program: RPC program for NFSv4 callback service
6554  * @port: IP port number for NFS4 callback service
6555  * @cred: credential to use for this call
6556  * @res: where to place the result
6557  *
6558  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6559  */
6560 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6561 		unsigned short port, const struct cred *cred,
6562 		struct nfs4_setclientid_res *res)
6563 {
6564 	nfs4_verifier sc_verifier;
6565 	struct nfs4_setclientid setclientid = {
6566 		.sc_verifier = &sc_verifier,
6567 		.sc_prog = program,
6568 		.sc_clnt = clp,
6569 	};
6570 	struct rpc_message msg = {
6571 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6572 		.rpc_argp = &setclientid,
6573 		.rpc_resp = res,
6574 		.rpc_cred = cred,
6575 	};
6576 	struct rpc_task_setup task_setup_data = {
6577 		.rpc_client = clp->cl_rpcclient,
6578 		.rpc_message = &msg,
6579 		.callback_ops = &nfs4_setclientid_ops,
6580 		.callback_data = &setclientid,
6581 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6582 	};
6583 	unsigned long now = jiffies;
6584 	int status;
6585 
6586 	/* nfs_client_id4 */
6587 	nfs4_init_boot_verifier(clp, &sc_verifier);
6588 
6589 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6590 		status = nfs4_init_uniform_client_string(clp);
6591 	else
6592 		status = nfs4_init_nonuniform_client_string(clp);
6593 
6594 	if (status)
6595 		goto out;
6596 
6597 	/* cb_client4 */
6598 	setclientid.sc_netid_len =
6599 				nfs4_init_callback_netid(clp,
6600 						setclientid.sc_netid,
6601 						sizeof(setclientid.sc_netid));
6602 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6603 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6604 				clp->cl_ipaddr, port >> 8, port & 255);
6605 
6606 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6607 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6608 		clp->cl_owner_id);
6609 
6610 	status = nfs4_call_sync_custom(&task_setup_data);
6611 	if (setclientid.sc_cred) {
6612 		kfree(clp->cl_acceptor);
6613 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6614 		put_rpccred(setclientid.sc_cred);
6615 	}
6616 
6617 	if (status == 0)
6618 		do_renew_lease(clp, now);
6619 out:
6620 	trace_nfs4_setclientid(clp, status);
6621 	dprintk("NFS reply setclientid: %d\n", status);
6622 	return status;
6623 }
6624 
6625 /**
6626  * nfs4_proc_setclientid_confirm - Confirm client ID
6627  * @clp: state data structure
6628  * @arg: result of a previous SETCLIENTID
6629  * @cred: credential to use for this call
6630  *
6631  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6632  */
6633 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6634 		struct nfs4_setclientid_res *arg,
6635 		const struct cred *cred)
6636 {
6637 	struct rpc_message msg = {
6638 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6639 		.rpc_argp = arg,
6640 		.rpc_cred = cred,
6641 	};
6642 	int status;
6643 
6644 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6645 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6646 		clp->cl_clientid);
6647 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6648 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6649 	trace_nfs4_setclientid_confirm(clp, status);
6650 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6651 	return status;
6652 }
6653 
6654 struct nfs4_delegreturndata {
6655 	struct nfs4_delegreturnargs args;
6656 	struct nfs4_delegreturnres res;
6657 	struct nfs_fh fh;
6658 	nfs4_stateid stateid;
6659 	unsigned long timestamp;
6660 	unsigned short retrans;
6661 	struct {
6662 		struct nfs4_layoutreturn_args arg;
6663 		struct nfs4_layoutreturn_res res;
6664 		struct nfs4_xdr_opaque_data ld_private;
6665 		u32 roc_barrier;
6666 		bool roc;
6667 	} lr;
6668 	struct nfs4_delegattr sattr;
6669 	struct nfs_fattr fattr;
6670 	int rpc_status;
6671 	struct inode *inode;
6672 };
6673 
6674 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6675 {
6676 	struct nfs4_delegreturndata *data = calldata;
6677 	struct nfs4_exception exception = {
6678 		.inode = data->inode,
6679 		.stateid = &data->stateid,
6680 		.task_is_privileged = data->args.seq_args.sa_privileged,
6681 		.retrans = data->retrans,
6682 	};
6683 
6684 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6685 		return;
6686 
6687 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6688 
6689 	/* Handle Layoutreturn errors */
6690 	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6691 			  &data->res.lr_ret) == -EAGAIN)
6692 		goto out_restart;
6693 
6694 	if (data->args.sattr_args && task->tk_status != 0) {
6695 		switch(data->res.sattr_ret) {
6696 		case 0:
6697 			data->args.sattr_args = NULL;
6698 			data->res.sattr_res = false;
6699 			break;
6700 		case -NFS4ERR_ADMIN_REVOKED:
6701 		case -NFS4ERR_DELEG_REVOKED:
6702 		case -NFS4ERR_EXPIRED:
6703 		case -NFS4ERR_BAD_STATEID:
6704 			/* Let the main handler below do stateid recovery */
6705 			break;
6706 		case -NFS4ERR_OLD_STATEID:
6707 			if (nfs4_refresh_delegation_stateid(&data->stateid,
6708 						data->inode))
6709 				goto out_restart;
6710 			fallthrough;
6711 		default:
6712 			data->args.sattr_args = NULL;
6713 			data->res.sattr_res = false;
6714 			goto out_restart;
6715 		}
6716 	}
6717 
6718 	switch (task->tk_status) {
6719 	case 0:
6720 		renew_lease(data->res.server, data->timestamp);
6721 		break;
6722 	case -NFS4ERR_ADMIN_REVOKED:
6723 	case -NFS4ERR_DELEG_REVOKED:
6724 	case -NFS4ERR_EXPIRED:
6725 		nfs4_free_revoked_stateid(data->res.server,
6726 				data->args.stateid,
6727 				task->tk_msg.rpc_cred);
6728 		fallthrough;
6729 	case -NFS4ERR_BAD_STATEID:
6730 	case -NFS4ERR_STALE_STATEID:
6731 	case -ETIMEDOUT:
6732 		task->tk_status = 0;
6733 		break;
6734 	case -NFS4ERR_OLD_STATEID:
6735 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6736 			nfs4_stateid_seqid_inc(&data->stateid);
6737 		if (data->args.bitmask) {
6738 			data->args.bitmask = NULL;
6739 			data->res.fattr = NULL;
6740 		}
6741 		goto out_restart;
6742 	case -NFS4ERR_ACCESS:
6743 		if (data->args.bitmask) {
6744 			data->args.bitmask = NULL;
6745 			data->res.fattr = NULL;
6746 			goto out_restart;
6747 		}
6748 		fallthrough;
6749 	default:
6750 		task->tk_status = nfs4_async_handle_exception(task,
6751 				data->res.server, task->tk_status,
6752 				&exception);
6753 		data->retrans = exception.retrans;
6754 		if (exception.retry)
6755 			goto out_restart;
6756 	}
6757 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6758 	data->rpc_status = task->tk_status;
6759 	return;
6760 out_restart:
6761 	task->tk_status = 0;
6762 	rpc_restart_call_prepare(task);
6763 }
6764 
6765 static void nfs4_delegreturn_release(void *calldata)
6766 {
6767 	struct nfs4_delegreturndata *data = calldata;
6768 	struct inode *inode = data->inode;
6769 
6770 	if (data->lr.roc)
6771 		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6772 				 data->res.lr_ret);
6773 	if (inode) {
6774 		nfs4_fattr_set_prechange(&data->fattr,
6775 					 inode_peek_iversion_raw(inode));
6776 		nfs_refresh_inode(inode, &data->fattr);
6777 		nfs_iput_and_deactive(inode);
6778 	}
6779 	kfree(calldata);
6780 }
6781 
6782 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6783 {
6784 	struct nfs4_delegreturndata *d_data;
6785 	struct pnfs_layout_hdr *lo;
6786 
6787 	d_data = data;
6788 
6789 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6790 		nfs4_sequence_done(task, &d_data->res.seq_res);
6791 		return;
6792 	}
6793 
6794 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6795 	if (lo && !pnfs_layout_is_valid(lo)) {
6796 		d_data->args.lr_args = NULL;
6797 		d_data->res.lr_res = NULL;
6798 	}
6799 
6800 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6801 			&d_data->args.seq_args,
6802 			&d_data->res.seq_res,
6803 			task);
6804 }
6805 
6806 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6807 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6808 	.rpc_call_done = nfs4_delegreturn_done,
6809 	.rpc_release = nfs4_delegreturn_release,
6810 };
6811 
6812 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
6813 				  const nfs4_stateid *stateid,
6814 				  struct nfs_delegation *delegation,
6815 				  int issync)
6816 {
6817 	struct nfs4_delegreturndata *data;
6818 	struct nfs_server *server = NFS_SERVER(inode);
6819 	struct rpc_task *task;
6820 	struct rpc_message msg = {
6821 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6822 		.rpc_cred = cred,
6823 	};
6824 	struct rpc_task_setup task_setup_data = {
6825 		.rpc_client = server->client,
6826 		.rpc_message = &msg,
6827 		.callback_ops = &nfs4_delegreturn_ops,
6828 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6829 	};
6830 	int status = 0;
6831 
6832 	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6833 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6834 
6835 	data = kzalloc_obj(*data);
6836 	if (data == NULL)
6837 		return -ENOMEM;
6838 
6839 	nfs4_state_protect(server->nfs_client,
6840 			NFS_SP4_MACH_CRED_CLEANUP,
6841 			&task_setup_data.rpc_client, &msg);
6842 
6843 	data->args.fhandle = &data->fh;
6844 	data->args.stateid = &data->stateid;
6845 	nfs4_bitmask_set(data->args.bitmask_store,
6846 			 server->cache_consistency_bitmask, inode, 0);
6847 	data->args.bitmask = data->args.bitmask_store;
6848 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6849 	nfs4_stateid_copy(&data->stateid, stateid);
6850 	data->res.fattr = &data->fattr;
6851 	data->res.server = server;
6852 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6853 	data->lr.arg.ld_private = &data->lr.ld_private;
6854 	nfs_fattr_init(data->res.fattr);
6855 	data->timestamp = jiffies;
6856 	data->rpc_status = 0;
6857 	data->inode = nfs_igrab_and_active(inode);
6858 	if (data->inode || issync) {
6859 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6860 					cred, issync);
6861 		if (data->lr.roc) {
6862 			data->args.lr_args = &data->lr.arg;
6863 			data->res.lr_res = &data->lr.res;
6864 		}
6865 	}
6866 
6867 	if (delegation &&
6868 	    test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags)) {
6869 		if (delegation->type & FMODE_READ) {
6870 			data->sattr.atime = inode_get_atime(inode);
6871 			data->sattr.atime_set = true;
6872 		}
6873 		if (delegation->type & FMODE_WRITE) {
6874 			data->sattr.mtime = inode_get_mtime(inode);
6875 			data->sattr.mtime_set = true;
6876 		}
6877 		data->args.sattr_args = &data->sattr;
6878 		data->res.sattr_res = true;
6879 	}
6880 
6881 	nfs4_init_sequence(server->nfs_client, &data->args.seq_args,
6882 			   &data->res.seq_res, 1, !data->inode ? 1 : 0);
6883 
6884 	task_setup_data.callback_data = data;
6885 	msg.rpc_argp = &data->args;
6886 	msg.rpc_resp = &data->res;
6887 	task = rpc_run_task(&task_setup_data);
6888 	if (IS_ERR(task))
6889 		return PTR_ERR(task);
6890 	if (!issync)
6891 		goto out;
6892 	status = rpc_wait_for_completion_task(task);
6893 	if (status != 0)
6894 		goto out;
6895 	status = data->rpc_status;
6896 out:
6897 	rpc_put_task(task);
6898 	return status;
6899 }
6900 
6901 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
6902 			  const nfs4_stateid *stateid,
6903 			  struct nfs_delegation *delegation, int issync)
6904 {
6905 	struct nfs_server *server = NFS_SERVER(inode);
6906 	struct nfs4_exception exception = { };
6907 	int err;
6908 	do {
6909 		err = _nfs4_proc_delegreturn(inode, cred, stateid,
6910 					     delegation, issync);
6911 		trace_nfs4_delegreturn(inode, stateid, err);
6912 		switch (err) {
6913 			case -NFS4ERR_STALE_STATEID:
6914 			case -NFS4ERR_EXPIRED:
6915 			case 0:
6916 				return 0;
6917 		}
6918 		err = nfs4_handle_exception(server, err, &exception);
6919 	} while (exception.retry);
6920 	return err;
6921 }
6922 
6923 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6924 {
6925 	struct inode *inode = state->inode;
6926 	struct nfs_server *server = NFS_SERVER(inode);
6927 	struct nfs_client *clp = server->nfs_client;
6928 	struct nfs_lockt_args arg = {
6929 		.fh = NFS_FH(inode),
6930 		.fl = request,
6931 	};
6932 	struct nfs_lockt_res res = {
6933 		.denied = request,
6934 	};
6935 	struct rpc_message msg = {
6936 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6937 		.rpc_argp	= &arg,
6938 		.rpc_resp	= &res,
6939 		.rpc_cred	= state->owner->so_cred,
6940 	};
6941 	struct nfs4_lock_state *lsp;
6942 	int status;
6943 
6944 	arg.lock_owner.clientid = clp->cl_clientid;
6945 	status = nfs4_set_lock_state(state, request);
6946 	if (status != 0)
6947 		goto out;
6948 	lsp = request->fl_u.nfs4_fl.owner;
6949 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6950 	arg.lock_owner.s_dev = server->s_dev;
6951 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6952 	switch (status) {
6953 		case 0:
6954 			request->c.flc_type = F_UNLCK;
6955 			break;
6956 		case -NFS4ERR_DENIED:
6957 			status = 0;
6958 	}
6959 	request->fl_ops->fl_release_private(request);
6960 	request->fl_ops = NULL;
6961 out:
6962 	return status;
6963 }
6964 
6965 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6966 {
6967 	struct nfs4_exception exception = {
6968 		.interruptible = true,
6969 	};
6970 	int err;
6971 
6972 	do {
6973 		err = _nfs4_proc_getlk(state, cmd, request);
6974 		trace_nfs4_get_lock(request, state, cmd, err);
6975 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6976 				&exception);
6977 	} while (exception.retry);
6978 	return err;
6979 }
6980 
6981 /*
6982  * Update the seqid of a lock stateid after receiving
6983  * NFS4ERR_OLD_STATEID
6984  */
6985 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6986 		struct nfs4_lock_state *lsp)
6987 {
6988 	struct nfs4_state *state = lsp->ls_state;
6989 	bool ret = false;
6990 
6991 	spin_lock(&state->state_lock);
6992 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6993 		goto out;
6994 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6995 		nfs4_stateid_seqid_inc(dst);
6996 	else
6997 		dst->seqid = lsp->ls_stateid.seqid;
6998 	ret = true;
6999 out:
7000 	spin_unlock(&state->state_lock);
7001 	return ret;
7002 }
7003 
7004 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
7005 		struct nfs4_lock_state *lsp)
7006 {
7007 	struct nfs4_state *state = lsp->ls_state;
7008 	bool ret;
7009 
7010 	spin_lock(&state->state_lock);
7011 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
7012 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
7013 	spin_unlock(&state->state_lock);
7014 	return ret;
7015 }
7016 
7017 struct nfs4_unlockdata {
7018 	struct nfs_locku_args arg;
7019 	struct nfs_locku_res res;
7020 	struct nfs4_lock_state *lsp;
7021 	struct nfs_open_context *ctx;
7022 	struct nfs_lock_context *l_ctx;
7023 	struct file_lock fl;
7024 	struct nfs_server *server;
7025 	unsigned long timestamp;
7026 	unsigned short retrans;
7027 };
7028 
7029 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
7030 		struct nfs_open_context *ctx,
7031 		struct nfs4_lock_state *lsp,
7032 		struct nfs_seqid *seqid)
7033 {
7034 	struct nfs4_unlockdata *p;
7035 	struct nfs4_state *state = lsp->ls_state;
7036 	struct inode *inode = state->inode;
7037 	struct nfs_lock_context *l_ctx;
7038 
7039 	p = kzalloc_obj(*p);
7040 	if (p == NULL)
7041 		return NULL;
7042 	l_ctx = nfs_get_lock_context(ctx);
7043 	if (!IS_ERR(l_ctx)) {
7044 		p->l_ctx = l_ctx;
7045 	} else {
7046 		kfree(p);
7047 		return NULL;
7048 	}
7049 	p->arg.fh = NFS_FH(inode);
7050 	p->arg.fl = &p->fl;
7051 	p->arg.seqid = seqid;
7052 	p->res.seqid = seqid;
7053 	p->lsp = lsp;
7054 	/* Ensure we don't close file until we're done freeing locks! */
7055 	p->ctx = get_nfs_open_context(ctx);
7056 	locks_init_lock(&p->fl);
7057 	locks_copy_lock(&p->fl, fl);
7058 	p->server = NFS_SERVER(inode);
7059 	spin_lock(&state->state_lock);
7060 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
7061 	spin_unlock(&state->state_lock);
7062 	return p;
7063 }
7064 
7065 static void nfs4_locku_release_calldata(void *data)
7066 {
7067 	struct nfs4_unlockdata *calldata = data;
7068 	nfs_free_seqid(calldata->arg.seqid);
7069 	nfs4_put_lock_state(calldata->lsp);
7070 	nfs_put_lock_context(calldata->l_ctx);
7071 	put_nfs_open_context(calldata->ctx);
7072 	kfree(calldata);
7073 }
7074 
7075 static void nfs4_locku_done(struct rpc_task *task, void *data)
7076 {
7077 	struct nfs4_unlockdata *calldata = data;
7078 	struct nfs4_exception exception = {
7079 		.inode = calldata->lsp->ls_state->inode,
7080 		.stateid = &calldata->arg.stateid,
7081 		.retrans = calldata->retrans,
7082 	};
7083 
7084 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
7085 		return;
7086 	switch (task->tk_status) {
7087 		case 0:
7088 			renew_lease(calldata->server, calldata->timestamp);
7089 			if (nfs4_update_lock_stateid(calldata->lsp,
7090 					&calldata->res.stateid))
7091 				break;
7092 			fallthrough;
7093 		case -NFS4ERR_ADMIN_REVOKED:
7094 		case -NFS4ERR_EXPIRED:
7095 			nfs4_free_revoked_stateid(calldata->server,
7096 					&calldata->arg.stateid,
7097 					task->tk_msg.rpc_cred);
7098 			fallthrough;
7099 		case -NFS4ERR_BAD_STATEID:
7100 		case -NFS4ERR_STALE_STATEID:
7101 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
7102 						calldata->lsp))
7103 				rpc_restart_call_prepare(task);
7104 			break;
7105 		case -NFS4ERR_OLD_STATEID:
7106 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
7107 						calldata->lsp))
7108 				rpc_restart_call_prepare(task);
7109 			break;
7110 		default:
7111 			task->tk_status = nfs4_async_handle_exception(task,
7112 					calldata->server, task->tk_status,
7113 					&exception);
7114 			calldata->retrans = exception.retrans;
7115 			if (exception.retry)
7116 				rpc_restart_call_prepare(task);
7117 	}
7118 	nfs_release_seqid(calldata->arg.seqid);
7119 }
7120 
7121 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
7122 {
7123 	struct nfs4_unlockdata *calldata = data;
7124 
7125 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
7126 		nfs_async_iocounter_wait(task, calldata->l_ctx))
7127 		return;
7128 
7129 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
7130 		goto out_wait;
7131 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
7132 		/* Note: exit _without_ running nfs4_locku_done */
7133 		goto out_no_action;
7134 	}
7135 	calldata->timestamp = jiffies;
7136 	if (nfs4_setup_sequence(calldata->server->nfs_client,
7137 				&calldata->arg.seq_args,
7138 				&calldata->res.seq_res,
7139 				task) != 0)
7140 		nfs_release_seqid(calldata->arg.seqid);
7141 	return;
7142 out_no_action:
7143 	task->tk_action = NULL;
7144 out_wait:
7145 	nfs4_sequence_done(task, &calldata->res.seq_res);
7146 }
7147 
7148 static const struct rpc_call_ops nfs4_locku_ops = {
7149 	.rpc_call_prepare = nfs4_locku_prepare,
7150 	.rpc_call_done = nfs4_locku_done,
7151 	.rpc_release = nfs4_locku_release_calldata,
7152 };
7153 
7154 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
7155 		struct nfs_open_context *ctx,
7156 		struct nfs4_lock_state *lsp,
7157 		struct nfs_seqid *seqid)
7158 {
7159 	struct nfs4_unlockdata *data;
7160 	struct nfs_client *clp = NFS_SERVER(lsp->ls_state->inode)->nfs_client;
7161 	struct rpc_message msg = {
7162 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
7163 		.rpc_cred = ctx->cred,
7164 	};
7165 	struct rpc_task_setup task_setup_data = {
7166 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
7167 		.rpc_message = &msg,
7168 		.callback_ops = &nfs4_locku_ops,
7169 		.workqueue = nfsiod_workqueue,
7170 		.flags = RPC_TASK_ASYNC,
7171 	};
7172 
7173 	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
7174 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7175 
7176 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_CLEANUP,
7177 			   &task_setup_data.rpc_client, &msg);
7178 
7179 	/* Ensure this is an unlock - when canceling a lock, the
7180 	 * canceled lock is passed in, and it won't be an unlock.
7181 	 */
7182 	fl->c.flc_type = F_UNLCK;
7183 	if (fl->c.flc_flags & FL_CLOSE)
7184 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
7185 
7186 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
7187 	if (data == NULL) {
7188 		nfs_free_seqid(seqid);
7189 		return ERR_PTR(-ENOMEM);
7190 	}
7191 
7192 	nfs4_init_sequence(clp, &data->arg.seq_args, &data->res.seq_res, 1, 0);
7193 	msg.rpc_argp = &data->arg;
7194 	msg.rpc_resp = &data->res;
7195 	task_setup_data.callback_data = data;
7196 	return rpc_run_task(&task_setup_data);
7197 }
7198 
7199 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7200 {
7201 	struct inode *inode = state->inode;
7202 	struct nfs4_state_owner *sp = state->owner;
7203 	struct nfs_inode *nfsi = NFS_I(inode);
7204 	struct nfs_seqid *seqid;
7205 	struct nfs4_lock_state *lsp;
7206 	struct rpc_task *task;
7207 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7208 	int status = 0;
7209 	unsigned char saved_flags = request->c.flc_flags;
7210 
7211 	status = nfs4_set_lock_state(state, request);
7212 	/* Unlock _before_ we do the RPC call */
7213 	request->c.flc_flags |= FL_EXISTS;
7214 	/* Exclude nfs_delegation_claim_locks() */
7215 	mutex_lock(&sp->so_delegreturn_mutex);
7216 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7217 	down_read(&nfsi->rwsem);
7218 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7219 		up_read(&nfsi->rwsem);
7220 		mutex_unlock(&sp->so_delegreturn_mutex);
7221 		goto out;
7222 	}
7223 	lsp = request->fl_u.nfs4_fl.owner;
7224 	set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7225 	up_read(&nfsi->rwsem);
7226 	mutex_unlock(&sp->so_delegreturn_mutex);
7227 	if (status != 0)
7228 		goto out;
7229 	/* Is this a delegated lock? */
7230 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7231 		goto out;
7232 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7233 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7234 	status = -ENOMEM;
7235 	if (IS_ERR(seqid))
7236 		goto out;
7237 	task = nfs4_do_unlck(request,
7238 			     nfs_file_open_context(request->c.flc_file),
7239 			     lsp, seqid);
7240 	status = PTR_ERR(task);
7241 	if (IS_ERR(task))
7242 		goto out;
7243 	status = rpc_wait_for_completion_task(task);
7244 	rpc_put_task(task);
7245 out:
7246 	request->c.flc_flags = saved_flags;
7247 	trace_nfs4_unlock(request, state, F_SETLK, status);
7248 	return status;
7249 }
7250 
7251 struct nfs4_lockdata {
7252 	struct nfs_lock_args arg;
7253 	struct nfs_lock_res res;
7254 	struct nfs4_lock_state *lsp;
7255 	struct nfs_open_context *ctx;
7256 	struct file_lock fl;
7257 	unsigned long timestamp;
7258 	int rpc_status;
7259 	int cancelled;
7260 	struct nfs_server *server;
7261 };
7262 
7263 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7264 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7265 		gfp_t gfp_mask)
7266 {
7267 	struct nfs4_lockdata *p;
7268 	struct inode *inode = lsp->ls_state->inode;
7269 	struct nfs_server *server = NFS_SERVER(inode);
7270 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7271 
7272 	p = kzalloc_obj(*p, gfp_mask);
7273 	if (p == NULL)
7274 		return NULL;
7275 
7276 	p->arg.fh = NFS_FH(inode);
7277 	p->arg.fl = &p->fl;
7278 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7279 	if (IS_ERR(p->arg.open_seqid))
7280 		goto out_free;
7281 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7282 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7283 	if (IS_ERR(p->arg.lock_seqid))
7284 		goto out_free_seqid;
7285 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7286 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7287 	p->arg.lock_owner.s_dev = server->s_dev;
7288 	p->res.lock_seqid = p->arg.lock_seqid;
7289 	p->lsp = lsp;
7290 	p->server = server;
7291 	p->ctx = get_nfs_open_context(ctx);
7292 	locks_init_lock(&p->fl);
7293 	locks_copy_lock(&p->fl, fl);
7294 	return p;
7295 out_free_seqid:
7296 	nfs_free_seqid(p->arg.open_seqid);
7297 out_free:
7298 	kfree(p);
7299 	return NULL;
7300 }
7301 
7302 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7303 {
7304 	struct nfs4_lockdata *data = calldata;
7305 	struct nfs4_state *state = data->lsp->ls_state;
7306 
7307 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7308 		goto out_wait;
7309 	/* Do we need to do an open_to_lock_owner? */
7310 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7311 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7312 			goto out_release_lock_seqid;
7313 		}
7314 		nfs4_stateid_copy(&data->arg.open_stateid,
7315 				&state->open_stateid);
7316 		data->arg.new_lock_owner = 1;
7317 		data->res.open_seqid = data->arg.open_seqid;
7318 	} else {
7319 		data->arg.new_lock_owner = 0;
7320 		nfs4_stateid_copy(&data->arg.lock_stateid,
7321 				&data->lsp->ls_stateid);
7322 	}
7323 	if (!nfs4_valid_open_stateid(state)) {
7324 		data->rpc_status = -EBADF;
7325 		task->tk_action = NULL;
7326 		goto out_release_open_seqid;
7327 	}
7328 	data->timestamp = jiffies;
7329 	if (nfs4_setup_sequence(data->server->nfs_client,
7330 				&data->arg.seq_args,
7331 				&data->res.seq_res,
7332 				task) == 0)
7333 		return;
7334 out_release_open_seqid:
7335 	nfs_release_seqid(data->arg.open_seqid);
7336 out_release_lock_seqid:
7337 	nfs_release_seqid(data->arg.lock_seqid);
7338 out_wait:
7339 	nfs4_sequence_done(task, &data->res.seq_res);
7340 	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7341 }
7342 
7343 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7344 {
7345 	struct nfs4_lockdata *data = calldata;
7346 	struct nfs4_lock_state *lsp = data->lsp;
7347 
7348 	if (!nfs4_sequence_done(task, &data->res.seq_res))
7349 		return;
7350 
7351 	data->rpc_status = task->tk_status;
7352 	switch (task->tk_status) {
7353 	case 0:
7354 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7355 				data->timestamp);
7356 		if (data->arg.new_lock_owner != 0) {
7357 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7358 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7359 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7360 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7361 			goto out_restart;
7362 		break;
7363 	case -NFS4ERR_OLD_STATEID:
7364 		if (data->arg.new_lock_owner != 0 &&
7365 			nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7366 					lsp->ls_state))
7367 			goto out_restart;
7368 		if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7369 			goto out_restart;
7370 		fallthrough;
7371 	case -NFS4ERR_BAD_STATEID:
7372 	case -NFS4ERR_STALE_STATEID:
7373 	case -NFS4ERR_EXPIRED:
7374 		if (data->arg.new_lock_owner != 0) {
7375 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7376 						&lsp->ls_state->open_stateid))
7377 				goto out_restart;
7378 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7379 						&lsp->ls_stateid))
7380 				goto out_restart;
7381 	}
7382 out_done:
7383 	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7384 	return;
7385 out_restart:
7386 	if (!data->cancelled)
7387 		rpc_restart_call_prepare(task);
7388 	goto out_done;
7389 }
7390 
7391 static void nfs4_lock_release(void *calldata)
7392 {
7393 	struct nfs4_lockdata *data = calldata;
7394 
7395 	nfs_free_seqid(data->arg.open_seqid);
7396 	if (data->cancelled && data->rpc_status == 0) {
7397 		struct rpc_task *task;
7398 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7399 				data->arg.lock_seqid);
7400 		if (!IS_ERR(task))
7401 			rpc_put_task_async(task);
7402 		dprintk("%s: cancelling lock!\n", __func__);
7403 	} else
7404 		nfs_free_seqid(data->arg.lock_seqid);
7405 	nfs4_put_lock_state(data->lsp);
7406 	put_nfs_open_context(data->ctx);
7407 	kfree(data);
7408 }
7409 
7410 static const struct rpc_call_ops nfs4_lock_ops = {
7411 	.rpc_call_prepare = nfs4_lock_prepare,
7412 	.rpc_call_done = nfs4_lock_done,
7413 	.rpc_release = nfs4_lock_release,
7414 };
7415 
7416 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7417 {
7418 	switch (error) {
7419 	case -NFS4ERR_ADMIN_REVOKED:
7420 	case -NFS4ERR_EXPIRED:
7421 	case -NFS4ERR_BAD_STATEID:
7422 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7423 		if (new_lock_owner != 0 ||
7424 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7425 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7426 		break;
7427 	case -NFS4ERR_STALE_STATEID:
7428 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7429 		nfs4_schedule_lease_recovery(server->nfs_client);
7430 	}
7431 }
7432 
7433 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7434 {
7435 	struct nfs4_lockdata *data;
7436 	struct rpc_task *task;
7437 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
7438 	struct rpc_message msg = {
7439 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7440 		.rpc_cred = state->owner->so_cred,
7441 	};
7442 	struct rpc_task_setup task_setup_data = {
7443 		.rpc_client = NFS_CLIENT(state->inode),
7444 		.rpc_message = &msg,
7445 		.callback_ops = &nfs4_lock_ops,
7446 		.workqueue = nfsiod_workqueue,
7447 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7448 	};
7449 	int ret;
7450 
7451 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7452 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7453 
7454 	data = nfs4_alloc_lockdata(fl,
7455 				   nfs_file_open_context(fl->c.flc_file),
7456 				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7457 	if (data == NULL)
7458 		return -ENOMEM;
7459 	if (IS_SETLKW(cmd))
7460 		data->arg.block = 1;
7461 	nfs4_init_sequence(clp, &data->arg.seq_args, &data->res.seq_res, 1,
7462 				recovery_type > NFS_LOCK_NEW);
7463 	msg.rpc_argp = &data->arg;
7464 	msg.rpc_resp = &data->res;
7465 	task_setup_data.callback_data = data;
7466 
7467 	if (recovery_type == NFS_LOCK_RECLAIM)
7468 		data->arg.reclaim = NFS_LOCK_RECLAIM;
7469 
7470 	task = rpc_run_task(&task_setup_data);
7471 	if (IS_ERR(task))
7472 		return PTR_ERR(task);
7473 	ret = rpc_wait_for_completion_task(task);
7474 	if (ret == 0) {
7475 		ret = data->rpc_status;
7476 		if (ret)
7477 			nfs4_handle_setlk_error(data->server, data->lsp,
7478 					data->arg.new_lock_owner, ret);
7479 	} else
7480 		data->cancelled = true;
7481 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7482 	rpc_put_task(task);
7483 	dprintk("%s: ret = %d\n", __func__, ret);
7484 	return ret;
7485 }
7486 
7487 int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7488 {
7489 	struct nfs_server *server = NFS_SERVER(state->inode);
7490 	struct nfs4_exception exception = {
7491 		.inode = state->inode,
7492 	};
7493 	int err;
7494 
7495 	do {
7496 		/* Cache the lock if possible... */
7497 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7498 			return 0;
7499 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7500 		if (err != -NFS4ERR_DELAY)
7501 			break;
7502 		nfs4_handle_exception(server, err, &exception);
7503 	} while (exception.retry);
7504 	return err;
7505 }
7506 
7507 int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7508 {
7509 	struct nfs_server *server = NFS_SERVER(state->inode);
7510 	struct nfs4_exception exception = {
7511 		.inode = state->inode,
7512 	};
7513 	int err;
7514 
7515 	err = nfs4_set_lock_state(state, request);
7516 	if (err != 0)
7517 		return err;
7518 	if (!recover_lost_locks) {
7519 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7520 		return 0;
7521 	}
7522 	do {
7523 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7524 			return 0;
7525 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7526 		switch (err) {
7527 		default:
7528 			goto out;
7529 		case -NFS4ERR_GRACE:
7530 		case -NFS4ERR_DELAY:
7531 			nfs4_handle_exception(server, err, &exception);
7532 			err = 0;
7533 		}
7534 	} while (exception.retry);
7535 out:
7536 	return err;
7537 }
7538 
7539 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7540 {
7541 	struct nfs4_lock_state *lsp;
7542 	int status;
7543 
7544 	status = nfs4_set_lock_state(state, request);
7545 	if (status != 0)
7546 		return status;
7547 	lsp = request->fl_u.nfs4_fl.owner;
7548 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7549 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7550 		return 0;
7551 	return nfs4_lock_expired(state, request);
7552 }
7553 
7554 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7555 {
7556 	struct nfs_inode *nfsi = NFS_I(state->inode);
7557 	struct nfs4_state_owner *sp = state->owner;
7558 	unsigned char flags = request->c.flc_flags;
7559 	int status;
7560 
7561 	request->c.flc_flags |= FL_ACCESS;
7562 	status = locks_lock_inode_wait(state->inode, request);
7563 	if (status < 0)
7564 		goto out;
7565 	mutex_lock(&sp->so_delegreturn_mutex);
7566 	down_read(&nfsi->rwsem);
7567 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7568 		/* Yes: cache locks! */
7569 		/* ...but avoid races with delegation recall... */
7570 		request->c.flc_flags = flags & ~FL_SLEEP;
7571 		status = locks_lock_inode_wait(state->inode, request);
7572 		up_read(&nfsi->rwsem);
7573 		mutex_unlock(&sp->so_delegreturn_mutex);
7574 		goto out;
7575 	}
7576 	up_read(&nfsi->rwsem);
7577 	mutex_unlock(&sp->so_delegreturn_mutex);
7578 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7579 	if (status)
7580 		goto out;
7581 
7582 	down_read(&nfsi->rwsem);
7583 	request->c.flc_flags &= ~(FL_SLEEP | FL_ACCESS);
7584 	status = locks_lock_inode_wait(state->inode, request);
7585 	up_read(&nfsi->rwsem);
7586 out:
7587 	request->c.flc_flags = flags;
7588 	return status;
7589 }
7590 
7591 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7592 {
7593 	struct nfs4_exception exception = {
7594 		.state = state,
7595 		.inode = state->inode,
7596 		.interruptible = true,
7597 	};
7598 	int err;
7599 
7600 	do {
7601 		err = _nfs4_proc_setlk(state, cmd, request);
7602 		if (err == -NFS4ERR_DENIED)
7603 			err = -EAGAIN;
7604 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7605 				err, &exception);
7606 	} while (exception.retry);
7607 	return err;
7608 }
7609 
7610 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7611 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7612 
7613 static int
7614 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7615 			struct file_lock *request)
7616 {
7617 	int		status = -ERESTARTSYS;
7618 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7619 
7620 	while(!signalled()) {
7621 		status = nfs4_proc_setlk(state, cmd, request);
7622 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7623 			break;
7624 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7625 		schedule_timeout(timeout);
7626 		timeout *= 2;
7627 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7628 		status = -ERESTARTSYS;
7629 	}
7630 	return status;
7631 }
7632 
7633 struct nfs4_lock_waiter {
7634 	struct inode		*inode;
7635 	struct nfs_lowner	owner;
7636 	wait_queue_entry_t	wait;
7637 };
7638 
7639 static int
7640 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7641 {
7642 	struct nfs4_lock_waiter	*waiter	=
7643 		container_of(wait, struct nfs4_lock_waiter, wait);
7644 
7645 	/* NULL key means to wake up everyone */
7646 	if (key) {
7647 		struct cb_notify_lock_args	*cbnl = key;
7648 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7649 						*wowner = &waiter->owner;
7650 
7651 		/* Only wake if the callback was for the same owner. */
7652 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7653 			return 0;
7654 
7655 		/* Make sure it's for the right inode */
7656 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7657 			return 0;
7658 	}
7659 
7660 	return woken_wake_function(wait, mode, flags, key);
7661 }
7662 
7663 static int
7664 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7665 {
7666 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7667 	struct nfs_server *server = NFS_SERVER(state->inode);
7668 	struct nfs_client *clp = server->nfs_client;
7669 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7670 	struct nfs4_lock_waiter waiter = {
7671 		.inode = state->inode,
7672 		.owner = { .clientid = clp->cl_clientid,
7673 			   .id = lsp->ls_seqid.owner_id,
7674 			   .s_dev = server->s_dev },
7675 	};
7676 	int status;
7677 
7678 	/* Don't bother with waitqueue if we don't expect a callback */
7679 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7680 		return nfs4_retry_setlk_simple(state, cmd, request);
7681 
7682 	init_wait(&waiter.wait);
7683 	waiter.wait.func = nfs4_wake_lock_waiter;
7684 	add_wait_queue(q, &waiter.wait);
7685 
7686 	do {
7687 		status = nfs4_proc_setlk(state, cmd, request);
7688 		if (status != -EAGAIN || IS_SETLK(cmd))
7689 			break;
7690 
7691 		status = -ERESTARTSYS;
7692 		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7693 			   NFS4_LOCK_MAXTIMEOUT);
7694 	} while (!signalled());
7695 
7696 	remove_wait_queue(q, &waiter.wait);
7697 
7698 	return status;
7699 }
7700 
7701 static int
7702 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7703 {
7704 	struct nfs_open_context *ctx;
7705 	struct nfs4_state *state;
7706 	int status;
7707 
7708 	/* verify open state */
7709 	ctx = nfs_file_open_context(filp);
7710 	state = ctx->state;
7711 
7712 	if (IS_GETLK(cmd)) {
7713 		if (state != NULL)
7714 			return nfs4_proc_getlk(state, F_GETLK, request);
7715 		return 0;
7716 	}
7717 
7718 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7719 		return -EINVAL;
7720 
7721 	if (lock_is_unlock(request)) {
7722 		if (state != NULL)
7723 			return nfs4_proc_unlck(state, cmd, request);
7724 		return 0;
7725 	}
7726 
7727 	if (state == NULL)
7728 		return -ENOLCK;
7729 
7730 	if ((request->c.flc_flags & FL_POSIX) &&
7731 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7732 		return -ENOLCK;
7733 
7734 	/*
7735 	 * Don't rely on the VFS having checked the file open mode,
7736 	 * since it won't do this for flock() locks.
7737 	 */
7738 	switch (request->c.flc_type) {
7739 	case F_RDLCK:
7740 		if (!(filp->f_mode & FMODE_READ))
7741 			return -EBADF;
7742 		break;
7743 	case F_WRLCK:
7744 		if (!(filp->f_mode & FMODE_WRITE))
7745 			return -EBADF;
7746 	}
7747 
7748 	status = nfs4_set_lock_state(state, request);
7749 	if (status != 0)
7750 		return status;
7751 
7752 	return nfs4_retry_setlk(state, cmd, request);
7753 }
7754 
7755 static int nfs4_delete_lease(struct file *file, void **priv)
7756 {
7757 	return generic_setlease(file, F_UNLCK, NULL, priv);
7758 }
7759 
7760 static int nfs4_add_lease(struct file *file, int arg, struct file_lease **lease,
7761 			  void **priv)
7762 {
7763 	struct inode *inode = file_inode(file);
7764 	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7765 	int ret;
7766 
7767 	/* No delegation, no lease */
7768 	if (!nfs4_have_delegation(inode, type, 0))
7769 		return -EAGAIN;
7770 	ret = generic_setlease(file, arg, lease, priv);
7771 	if (ret || nfs4_have_delegation(inode, type, 0))
7772 		return ret;
7773 	/* We raced with a delegation return */
7774 	nfs4_delete_lease(file, priv);
7775 	return -EAGAIN;
7776 }
7777 
7778 int nfs4_proc_setlease(struct file *file, int arg, struct file_lease **lease,
7779 		       void **priv)
7780 {
7781 	switch (arg) {
7782 	case F_RDLCK:
7783 	case F_WRLCK:
7784 		return nfs4_add_lease(file, arg, lease, priv);
7785 	case F_UNLCK:
7786 		return nfs4_delete_lease(file, priv);
7787 	default:
7788 		return -EINVAL;
7789 	}
7790 }
7791 
7792 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7793 {
7794 	struct nfs_server *server = NFS_SERVER(state->inode);
7795 	int err;
7796 
7797 	err = nfs4_set_lock_state(state, fl);
7798 	if (err != 0)
7799 		return err;
7800 	do {
7801 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7802 		if (err != -NFS4ERR_DELAY && err != -NFS4ERR_GRACE)
7803 			break;
7804 		ssleep(1);
7805 	} while (err == -NFS4ERR_DELAY || err == -NFSERR_GRACE);
7806 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7807 }
7808 
7809 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7810 
7811 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7812 				   struct mnt_idmap *idmap,
7813 				   struct dentry *unused, struct inode *inode,
7814 				   const char *key, const void *buf,
7815 				   size_t buflen, int flags)
7816 {
7817 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7818 }
7819 
7820 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7821 				   struct dentry *unused, struct inode *inode,
7822 				   const char *key, void *buf, size_t buflen)
7823 {
7824 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7825 }
7826 
7827 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7828 {
7829 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7830 }
7831 
7832 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7833 
7834 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7835 				    struct mnt_idmap *idmap,
7836 				    struct dentry *unused, struct inode *inode,
7837 				    const char *key, const void *buf,
7838 				    size_t buflen, int flags)
7839 {
7840 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7841 }
7842 
7843 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7844 				    struct dentry *unused, struct inode *inode,
7845 				    const char *key, void *buf, size_t buflen)
7846 {
7847 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7848 }
7849 
7850 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7851 {
7852 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7853 }
7854 
7855 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7856 
7857 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7858 				    struct mnt_idmap *idmap,
7859 				    struct dentry *unused, struct inode *inode,
7860 				    const char *key, const void *buf,
7861 				    size_t buflen, int flags)
7862 {
7863 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7864 }
7865 
7866 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7867 				    struct dentry *unused, struct inode *inode,
7868 				    const char *key, void *buf, size_t buflen)
7869 {
7870 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7871 }
7872 
7873 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7874 {
7875 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7876 }
7877 
7878 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7879 
7880 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7881 				     struct mnt_idmap *idmap,
7882 				     struct dentry *unused, struct inode *inode,
7883 				     const char *key, const void *buf,
7884 				     size_t buflen, int flags)
7885 {
7886 	if (security_ismaclabel(key))
7887 		return nfs4_set_security_label(inode, buf, buflen);
7888 
7889 	return -EOPNOTSUPP;
7890 }
7891 
7892 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7893 				     struct dentry *unused, struct inode *inode,
7894 				     const char *key, void *buf, size_t buflen)
7895 {
7896 	if (security_ismaclabel(key))
7897 		return nfs4_get_security_label(inode, buf, buflen);
7898 	return -EOPNOTSUPP;
7899 }
7900 
7901 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7902 	.prefix = XATTR_SECURITY_PREFIX,
7903 	.get	= nfs4_xattr_get_nfs4_label,
7904 	.set	= nfs4_xattr_set_nfs4_label,
7905 };
7906 
7907 #endif
7908 
7909 #ifdef CONFIG_NFS_V4_2
7910 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7911 				    struct mnt_idmap *idmap,
7912 				    struct dentry *unused, struct inode *inode,
7913 				    const char *key, const void *buf,
7914 				    size_t buflen, int flags)
7915 {
7916 	u32 mask;
7917 	int ret;
7918 
7919 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7920 		return -EOPNOTSUPP;
7921 
7922 	/*
7923 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7924 	 * flags right now. Handling of xattr operations use the normal
7925 	 * file read/write permissions.
7926 	 *
7927 	 * Just in case the server has other ideas (which RFC 8276 allows),
7928 	 * do a cached access check for the XA* flags to possibly avoid
7929 	 * doing an RPC and getting EACCES back.
7930 	 */
7931 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7932 		if (!(mask & NFS_ACCESS_XAWRITE))
7933 			return -EACCES;
7934 	}
7935 
7936 	if (buf == NULL) {
7937 		ret = nfs42_proc_removexattr(inode, key);
7938 		if (!ret)
7939 			nfs4_xattr_cache_remove(inode, key);
7940 	} else {
7941 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7942 		if (!ret)
7943 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7944 	}
7945 
7946 	return ret;
7947 }
7948 
7949 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7950 				    struct dentry *unused, struct inode *inode,
7951 				    const char *key, void *buf, size_t buflen)
7952 {
7953 	u32 mask;
7954 	ssize_t ret;
7955 
7956 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7957 		return -EOPNOTSUPP;
7958 
7959 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7960 		if (!(mask & NFS_ACCESS_XAREAD))
7961 			return -EACCES;
7962 	}
7963 
7964 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7965 	if (ret)
7966 		return ret;
7967 
7968 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7969 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7970 		return ret;
7971 
7972 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7973 
7974 	return ret;
7975 }
7976 
7977 static ssize_t
7978 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7979 {
7980 	u64 cookie;
7981 	bool eof;
7982 	ssize_t ret, size;
7983 	char *buf;
7984 	size_t buflen;
7985 	u32 mask;
7986 
7987 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7988 		return 0;
7989 
7990 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7991 		if (!(mask & NFS_ACCESS_XALIST))
7992 			return 0;
7993 	}
7994 
7995 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7996 	if (ret)
7997 		return ret;
7998 
7999 	ret = nfs4_xattr_cache_list(inode, list, list_len);
8000 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
8001 		return ret;
8002 
8003 	cookie = 0;
8004 	eof = false;
8005 	buflen = list_len ? list_len : XATTR_LIST_MAX;
8006 	buf = list_len ? list : NULL;
8007 	size = 0;
8008 
8009 	while (!eof) {
8010 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
8011 		    &cookie, &eof);
8012 		if (ret < 0)
8013 			return ret;
8014 
8015 		if (list_len) {
8016 			buf += ret;
8017 			buflen -= ret;
8018 		}
8019 		size += ret;
8020 	}
8021 
8022 	if (list_len)
8023 		nfs4_xattr_cache_set_list(inode, list, size);
8024 
8025 	return size;
8026 }
8027 
8028 #else
8029 
8030 static ssize_t
8031 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
8032 {
8033 	return 0;
8034 }
8035 #endif /* CONFIG_NFS_V4_2 */
8036 
8037 /*
8038  * nfs_fhget will use either the mounted_on_fileid or the fileid
8039  */
8040 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
8041 {
8042 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
8043 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
8044 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
8045 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
8046 		return;
8047 
8048 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
8049 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
8050 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
8051 	fattr->nlink = 2;
8052 }
8053 
8054 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8055 				   const struct qstr *name,
8056 				   struct nfs4_fs_locations *fs_locations,
8057 				   struct page *page)
8058 {
8059 	struct nfs_server *server = NFS_SERVER(dir);
8060 	u32 bitmask[3];
8061 	struct nfs4_fs_locations_arg args = {
8062 		.dir_fh = NFS_FH(dir),
8063 		.name = name,
8064 		.page = page,
8065 		.bitmask = bitmask,
8066 	};
8067 	struct nfs4_fs_locations_res res = {
8068 		.fs_locations = fs_locations,
8069 	};
8070 	struct rpc_message msg = {
8071 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8072 		.rpc_argp = &args,
8073 		.rpc_resp = &res,
8074 	};
8075 	int status;
8076 
8077 	dprintk("%s: start\n", __func__);
8078 
8079 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
8080 	bitmask[1] = nfs4_fattr_bitmap[1];
8081 
8082 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
8083 	 * is not supported */
8084 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
8085 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
8086 	else
8087 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8088 
8089 	nfs_fattr_init(fs_locations->fattr);
8090 	fs_locations->server = server;
8091 	fs_locations->nlocations = 0;
8092 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8093 	dprintk("%s: returned status = %d\n", __func__, status);
8094 	return status;
8095 }
8096 
8097 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8098 			   const struct qstr *name,
8099 			   struct nfs4_fs_locations *fs_locations,
8100 			   struct page *page)
8101 {
8102 	struct nfs4_exception exception = {
8103 		.interruptible = true,
8104 	};
8105 	int err;
8106 	do {
8107 		err = _nfs4_proc_fs_locations(client, dir, name,
8108 				fs_locations, page);
8109 		trace_nfs4_get_fs_locations(dir, name, err);
8110 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8111 				&exception);
8112 	} while (exception.retry);
8113 	return err;
8114 }
8115 
8116 /*
8117  * This operation also signals the server that this client is
8118  * performing migration recovery.  The server can stop asserting
8119  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8120  * performing this operation is identified in the SEQUENCE
8121  * operation in this compound.
8122  *
8123  * When the client supports GETATTR(fs_locations_info), it can
8124  * be plumbed in here.
8125  */
8126 static int _nfs41_proc_get_locations(struct nfs_server *server,
8127 				     struct nfs_fh *fhandle,
8128 				     struct nfs4_fs_locations *locations,
8129 				     struct page *page, const struct cred *cred)
8130 {
8131 	struct rpc_clnt *clnt = server->client;
8132 	struct nfs_client *clp = server->nfs_client;
8133 	u32 bitmask[2] = {
8134 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8135 	};
8136 	struct nfs4_fs_locations_arg args = {
8137 		.fh		= fhandle,
8138 		.page		= page,
8139 		.bitmask	= bitmask,
8140 		.migration	= 1,		/* skip LOOKUP */
8141 	};
8142 	struct nfs4_fs_locations_res res = {
8143 		.fs_locations	= locations,
8144 		.migration	= 1,
8145 	};
8146 	struct rpc_message msg = {
8147 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8148 		.rpc_argp	= &args,
8149 		.rpc_resp	= &res,
8150 		.rpc_cred	= cred,
8151 	};
8152 	struct nfs4_call_sync_data data = {
8153 		.seq_server = server,
8154 		.seq_args = &args.seq_args,
8155 		.seq_res = &res.seq_res,
8156 	};
8157 	struct rpc_task_setup task_setup_data = {
8158 		.rpc_client = clnt,
8159 		.rpc_message = &msg,
8160 		.callback_ops = clp->cl_mvops->call_sync_ops,
8161 		.callback_data = &data,
8162 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8163 	};
8164 	int status;
8165 
8166 	nfs_fattr_init(locations->fattr);
8167 	locations->server = server;
8168 	locations->nlocations = 0;
8169 
8170 	nfs4_init_sequence(clp, &args.seq_args, &res.seq_res, 0, 1);
8171 	status = nfs4_call_sync_custom(&task_setup_data);
8172 	if (status == NFS4_OK &&
8173 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8174 		status = -NFS4ERR_LEASE_MOVED;
8175 	return status;
8176 }
8177 
8178 /**
8179  * nfs4_proc_get_locations - discover locations for a migrated FSID
8180  * @server: pointer to nfs_server to process
8181  * @fhandle: pointer to the kernel NFS client file handle
8182  * @locations: result of query
8183  * @page: buffer
8184  * @cred: credential to use for this operation
8185  *
8186  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8187  * operation failed, or a negative errno if a local error occurred.
8188  *
8189  * On success, "locations" is filled in, but if the server has
8190  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8191  * asserted.
8192  *
8193  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8194  * from this client that require migration recovery.
8195  */
8196 int nfs4_proc_get_locations(struct nfs_server *server,
8197 			    struct nfs_fh *fhandle,
8198 			    struct nfs4_fs_locations *locations,
8199 			    struct page *page, const struct cred *cred)
8200 {
8201 	struct nfs_client *clp = server->nfs_client;
8202 	const struct nfs4_mig_recovery_ops *ops =
8203 					clp->cl_mvops->mig_recovery_ops;
8204 	struct nfs4_exception exception = {
8205 		.interruptible = true,
8206 	};
8207 	int status;
8208 
8209 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8210 		(unsigned long long)server->fsid.major,
8211 		(unsigned long long)server->fsid.minor,
8212 		clp->cl_hostname);
8213 	nfs_display_fhandle(fhandle, __func__);
8214 
8215 	do {
8216 		status = ops->get_locations(server, fhandle, locations, page,
8217 					    cred);
8218 		if (status != -NFS4ERR_DELAY)
8219 			break;
8220 		nfs4_handle_exception(server, status, &exception);
8221 	} while (exception.retry);
8222 	return status;
8223 }
8224 
8225 /*
8226  * This operation also signals the server that this client is
8227  * performing "lease moved" recovery.  The server can stop asserting
8228  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8229  * this operation is identified in the SEQUENCE operation in this
8230  * compound.
8231  */
8232 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8233 {
8234 	struct nfs_server *server = NFS_SERVER(inode);
8235 	struct rpc_clnt *clnt = server->client;
8236 	struct nfs4_fsid_present_arg args = {
8237 		.fh		= NFS_FH(inode),
8238 	};
8239 	struct nfs4_fsid_present_res res = {
8240 	};
8241 	struct rpc_message msg = {
8242 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8243 		.rpc_argp	= &args,
8244 		.rpc_resp	= &res,
8245 		.rpc_cred	= cred,
8246 	};
8247 	int status;
8248 
8249 	res.fh = nfs_alloc_fhandle();
8250 	if (res.fh == NULL)
8251 		return -ENOMEM;
8252 
8253 	nfs4_init_sequence(server->nfs_client, &args.seq_args, &res.seq_res, 0, 1);
8254 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8255 						&args.seq_args, &res.seq_res);
8256 	nfs_free_fhandle(res.fh);
8257 	if (status == NFS4_OK &&
8258 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8259 		status = -NFS4ERR_LEASE_MOVED;
8260 	return status;
8261 }
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(clp, &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 /*
8392  * Check the exchange flags returned by the server for invalid flags, having
8393  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8394  * DS flags set.
8395  */
8396 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8397 {
8398 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8399 		goto out_inval;
8400 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8401 		goto out_inval;
8402 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8403 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8404 		goto out_inval;
8405 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8406 		goto out_inval;
8407 	return NFS_OK;
8408 out_inval:
8409 	return -NFS4ERR_INVAL;
8410 }
8411 
8412 static bool
8413 nfs41_same_server_scope(struct nfs41_server_scope *a,
8414 			struct nfs41_server_scope *b)
8415 {
8416 	if (a->server_scope_sz != b->server_scope_sz)
8417 		return false;
8418 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8419 }
8420 
8421 static void
8422 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8423 {
8424 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8425 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8426 	struct nfs_client *clp = args->client;
8427 
8428 	switch (task->tk_status) {
8429 	case -NFS4ERR_BADSESSION:
8430 	case -NFS4ERR_DEADSESSION:
8431 		nfs4_schedule_session_recovery(clp->cl_session,
8432 				task->tk_status);
8433 		return;
8434 	}
8435 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8436 			res->dir != NFS4_CDFS4_BOTH) {
8437 		rpc_task_close_connection(task);
8438 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8439 			rpc_restart_call(task);
8440 	}
8441 }
8442 
8443 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8444 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8445 };
8446 
8447 /*
8448  * nfs4_proc_bind_one_conn_to_session()
8449  *
8450  * The 4.1 client currently uses the same TCP connection for the
8451  * fore and backchannel.
8452  */
8453 static
8454 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8455 		struct rpc_xprt *xprt,
8456 		struct nfs_client *clp,
8457 		const struct cred *cred)
8458 {
8459 	int status;
8460 	struct nfs41_bind_conn_to_session_args args = {
8461 		.client = clp,
8462 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8463 		.retries = 0,
8464 	};
8465 	struct nfs41_bind_conn_to_session_res res;
8466 	struct rpc_message msg = {
8467 		.rpc_proc =
8468 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8469 		.rpc_argp = &args,
8470 		.rpc_resp = &res,
8471 		.rpc_cred = cred,
8472 	};
8473 	struct rpc_task_setup task_setup_data = {
8474 		.rpc_client = clnt,
8475 		.rpc_xprt = xprt,
8476 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8477 		.rpc_message = &msg,
8478 		.flags = RPC_TASK_TIMEOUT,
8479 	};
8480 	struct rpc_task *task;
8481 
8482 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8483 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8484 		args.dir = NFS4_CDFC4_FORE;
8485 
8486 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8487 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8488 		args.dir = NFS4_CDFC4_FORE;
8489 
8490 	task = rpc_run_task(&task_setup_data);
8491 	if (!IS_ERR(task)) {
8492 		status = task->tk_status;
8493 		rpc_put_task(task);
8494 	} else
8495 		status = PTR_ERR(task);
8496 	trace_nfs4_bind_conn_to_session(clp, status);
8497 	if (status == 0) {
8498 		if (memcmp(res.sessionid.data,
8499 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8500 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8501 			return -EIO;
8502 		}
8503 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8504 			dprintk("NFS: %s: Unexpected direction from server\n",
8505 				__func__);
8506 			return -EIO;
8507 		}
8508 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8509 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8510 				__func__);
8511 			return -EIO;
8512 		}
8513 	}
8514 
8515 	return status;
8516 }
8517 
8518 struct rpc_bind_conn_calldata {
8519 	struct nfs_client *clp;
8520 	const struct cred *cred;
8521 };
8522 
8523 static int
8524 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8525 		struct rpc_xprt *xprt,
8526 		void *calldata)
8527 {
8528 	struct rpc_bind_conn_calldata *p = calldata;
8529 
8530 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8531 }
8532 
8533 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8534 {
8535 	struct rpc_bind_conn_calldata data = {
8536 		.clp = clp,
8537 		.cred = cred,
8538 	};
8539 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8540 			nfs4_proc_bind_conn_to_session_callback, &data);
8541 }
8542 
8543 /*
8544  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8545  * and operations we'd like to see to enable certain features in the allow map
8546  */
8547 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8548 	.how = SP4_MACH_CRED,
8549 	.enforce.u.words = {
8550 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8551 		      1 << (OP_EXCHANGE_ID - 32) |
8552 		      1 << (OP_CREATE_SESSION - 32) |
8553 		      1 << (OP_DESTROY_SESSION - 32) |
8554 		      1 << (OP_DESTROY_CLIENTID - 32)
8555 	},
8556 	.allow.u.words = {
8557 		[0] = 1 << (OP_CLOSE) |
8558 		      1 << (OP_OPEN_DOWNGRADE) |
8559 		      1 << (OP_LOCKU) |
8560 		      1 << (OP_DELEGRETURN) |
8561 		      1 << (OP_COMMIT),
8562 		[1] = 1 << (OP_SECINFO - 32) |
8563 		      1 << (OP_SECINFO_NO_NAME - 32) |
8564 		      1 << (OP_LAYOUTRETURN - 32) |
8565 		      1 << (OP_TEST_STATEID - 32) |
8566 		      1 << (OP_FREE_STATEID - 32) |
8567 		      1 << (OP_WRITE - 32)
8568 	}
8569 };
8570 
8571 /*
8572  * Select the state protection mode for client `clp' given the server results
8573  * from exchange_id in `sp'.
8574  *
8575  * Returns 0 on success, negative errno otherwise.
8576  */
8577 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8578 				 struct nfs41_state_protection *sp)
8579 {
8580 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8581 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8582 		      1 << (OP_EXCHANGE_ID - 32) |
8583 		      1 << (OP_CREATE_SESSION - 32) |
8584 		      1 << (OP_DESTROY_SESSION - 32) |
8585 		      1 << (OP_DESTROY_CLIENTID - 32)
8586 	};
8587 	unsigned long flags = 0;
8588 	unsigned int i;
8589 	int ret = 0;
8590 
8591 	if (sp->how == SP4_MACH_CRED) {
8592 		/* Print state protect result */
8593 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8594 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8595 			if (test_bit(i, sp->enforce.u.longs))
8596 				dfprintk(MOUNT, "  enforce op %d\n", i);
8597 			if (test_bit(i, sp->allow.u.longs))
8598 				dfprintk(MOUNT, "  allow op %d\n", i);
8599 		}
8600 
8601 		/* make sure nothing is on enforce list that isn't supported */
8602 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8603 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8604 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8605 				ret = -EINVAL;
8606 				goto out;
8607 			}
8608 		}
8609 
8610 		/*
8611 		 * Minimal mode - state operations are allowed to use machine
8612 		 * credential.  Note this already happens by default, so the
8613 		 * client doesn't have to do anything more than the negotiation.
8614 		 *
8615 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8616 		 *       we're already using the machine cred for exchange_id
8617 		 *       and will never use a different cred.
8618 		 */
8619 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8620 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8621 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8622 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8623 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8624 			dfprintk(MOUNT, "  minimal mode enabled\n");
8625 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8626 		} else {
8627 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8628 			ret = -EINVAL;
8629 			goto out;
8630 		}
8631 
8632 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8633 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8634 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8635 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8636 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8637 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8638 		}
8639 
8640 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8641 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8642 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8643 		}
8644 
8645 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8646 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8647 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8648 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8649 		}
8650 
8651 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8652 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8653 			dfprintk(MOUNT, "  stateid mode enabled\n");
8654 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8655 		}
8656 
8657 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8658 			dfprintk(MOUNT, "  write mode enabled\n");
8659 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8660 		}
8661 
8662 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8663 			dfprintk(MOUNT, "  commit mode enabled\n");
8664 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8665 		}
8666 	}
8667 out:
8668 	clp->cl_sp4_flags = flags;
8669 	return ret;
8670 }
8671 
8672 struct nfs41_exchange_id_data {
8673 	struct nfs41_exchange_id_res res;
8674 	struct nfs41_exchange_id_args args;
8675 };
8676 
8677 static void nfs4_exchange_id_release(void *data)
8678 {
8679 	struct nfs41_exchange_id_data *cdata =
8680 					(struct nfs41_exchange_id_data *)data;
8681 
8682 	nfs_put_client(cdata->args.client);
8683 	kfree(cdata->res.impl_id);
8684 	kfree(cdata->res.server_scope);
8685 	kfree(cdata->res.server_owner);
8686 	kfree(cdata);
8687 }
8688 
8689 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8690 	.rpc_release = nfs4_exchange_id_release,
8691 };
8692 
8693 /*
8694  * _nfs4_proc_exchange_id()
8695  *
8696  * Wrapper for EXCHANGE_ID operation.
8697  */
8698 static struct rpc_task *
8699 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8700 			u32 sp4_how, struct rpc_xprt *xprt)
8701 {
8702 	struct rpc_message msg = {
8703 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8704 		.rpc_cred = cred,
8705 	};
8706 	struct rpc_task_setup task_setup_data = {
8707 		.rpc_client = clp->cl_rpcclient,
8708 		.callback_ops = &nfs4_exchange_id_call_ops,
8709 		.rpc_message = &msg,
8710 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8711 	};
8712 	struct nfs41_exchange_id_data *calldata;
8713 	int status;
8714 
8715 	if (!refcount_inc_not_zero(&clp->cl_count))
8716 		return ERR_PTR(-EIO);
8717 
8718 	status = -ENOMEM;
8719 	calldata = kzalloc_obj(*calldata, GFP_NOFS);
8720 	if (!calldata)
8721 		goto out;
8722 
8723 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8724 
8725 	status = nfs4_init_uniform_client_string(clp);
8726 	if (status)
8727 		goto out_calldata;
8728 
8729 	calldata->res.server_owner = kzalloc_obj(struct nfs41_server_owner,
8730 						 GFP_NOFS);
8731 	status = -ENOMEM;
8732 	if (unlikely(calldata->res.server_owner == NULL))
8733 		goto out_calldata;
8734 
8735 	calldata->res.server_scope = kzalloc_obj(struct nfs41_server_scope,
8736 						 GFP_NOFS);
8737 	if (unlikely(calldata->res.server_scope == NULL))
8738 		goto out_server_owner;
8739 
8740 	calldata->res.impl_id = kzalloc_obj(struct nfs41_impl_id, GFP_NOFS);
8741 	if (unlikely(calldata->res.impl_id == NULL))
8742 		goto out_server_scope;
8743 
8744 	switch (sp4_how) {
8745 	case SP4_NONE:
8746 		calldata->args.state_protect.how = SP4_NONE;
8747 		break;
8748 
8749 	case SP4_MACH_CRED:
8750 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8751 		break;
8752 
8753 	default:
8754 		/* unsupported! */
8755 		WARN_ON_ONCE(1);
8756 		status = -EINVAL;
8757 		goto out_impl_id;
8758 	}
8759 	if (xprt) {
8760 		task_setup_data.rpc_xprt = xprt;
8761 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8762 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8763 				sizeof(calldata->args.verifier.data));
8764 	}
8765 	calldata->args.client = clp;
8766 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8767 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8768 #ifdef CONFIG_NFS_V4_1_MIGRATION
8769 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8770 #endif
8771 	if (test_bit(NFS_CS_PNFS, &clp->cl_flags))
8772 		calldata->args.flags |= EXCHGID4_FLAG_USE_PNFS_DS;
8773 	msg.rpc_argp = &calldata->args;
8774 	msg.rpc_resp = &calldata->res;
8775 	task_setup_data.callback_data = calldata;
8776 
8777 	return rpc_run_task(&task_setup_data);
8778 
8779 out_impl_id:
8780 	kfree(calldata->res.impl_id);
8781 out_server_scope:
8782 	kfree(calldata->res.server_scope);
8783 out_server_owner:
8784 	kfree(calldata->res.server_owner);
8785 out_calldata:
8786 	kfree(calldata);
8787 out:
8788 	nfs_put_client(clp);
8789 	return ERR_PTR(status);
8790 }
8791 
8792 /*
8793  * _nfs4_proc_exchange_id()
8794  *
8795  * Wrapper for EXCHANGE_ID operation.
8796  */
8797 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8798 			u32 sp4_how)
8799 {
8800 	struct rpc_task *task;
8801 	struct nfs41_exchange_id_args *argp;
8802 	struct nfs41_exchange_id_res *resp;
8803 	unsigned long now = jiffies;
8804 	int status;
8805 
8806 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8807 	if (IS_ERR(task))
8808 		return PTR_ERR(task);
8809 
8810 	argp = task->tk_msg.rpc_argp;
8811 	resp = task->tk_msg.rpc_resp;
8812 	status = task->tk_status;
8813 	if (status  != 0)
8814 		goto out;
8815 
8816 	status = nfs4_check_cl_exchange_flags(resp->flags,
8817 			clp->cl_mvops->minor_version);
8818 	if (status  != 0)
8819 		goto out;
8820 
8821 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8822 	if (status != 0)
8823 		goto out;
8824 
8825 	do_renew_lease(clp, now);
8826 
8827 	clp->cl_clientid = resp->clientid;
8828 	clp->cl_exchange_flags = resp->flags;
8829 	clp->cl_seqid = resp->seqid;
8830 	/* Client ID is not confirmed */
8831 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8832 		clear_bit(NFS4_SESSION_ESTABLISHED,
8833 			  &clp->cl_session->session_state);
8834 
8835 	if (clp->cl_serverscope != NULL &&
8836 	    !nfs41_same_server_scope(clp->cl_serverscope,
8837 				resp->server_scope)) {
8838 		dprintk("%s: server_scope mismatch detected\n",
8839 			__func__);
8840 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8841 	}
8842 
8843 	swap(clp->cl_serverowner, resp->server_owner);
8844 	swap(clp->cl_serverscope, resp->server_scope);
8845 	swap(clp->cl_implid, resp->impl_id);
8846 
8847 	/* Save the EXCHANGE_ID verifier session trunk tests */
8848 	memcpy(clp->cl_confirm.data, argp->verifier.data,
8849 	       sizeof(clp->cl_confirm.data));
8850 out:
8851 	trace_nfs4_exchange_id(clp, status);
8852 	rpc_put_task(task);
8853 	return status;
8854 }
8855 
8856 /*
8857  * nfs4_proc_exchange_id()
8858  *
8859  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8860  *
8861  * Since the clientid has expired, all compounds using sessions
8862  * associated with the stale clientid will be returning
8863  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8864  * be in some phase of session reset.
8865  *
8866  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8867  */
8868 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8869 {
8870 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8871 	int status;
8872 
8873 	/* try SP4_MACH_CRED if krb5i/p	*/
8874 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8875 	    authflavor == RPC_AUTH_GSS_KRB5P) {
8876 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8877 		if (!status)
8878 			return 0;
8879 	}
8880 
8881 	/* try SP4_NONE */
8882 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8883 }
8884 
8885 /**
8886  * nfs4_test_session_trunk
8887  *
8888  * This is an add_xprt_test() test function called from
8889  * rpc_clnt_setup_test_and_add_xprt.
8890  *
8891  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8892  * and is dereferrenced in nfs4_exchange_id_release
8893  *
8894  * Upon success, add the new transport to the rpc_clnt
8895  *
8896  * @clnt: struct rpc_clnt to get new transport
8897  * @xprt: the rpc_xprt to test
8898  * @data: call data for _nfs4_proc_exchange_id.
8899  */
8900 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8901 			    void *data)
8902 {
8903 	struct nfs4_add_xprt_data *adata = data;
8904 	struct rpc_task *task;
8905 	int status;
8906 
8907 	u32 sp4_how;
8908 
8909 	dprintk("--> %s try %s\n", __func__,
8910 		xprt->address_strings[RPC_DISPLAY_ADDR]);
8911 
8912 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8913 
8914 try_again:
8915 	/* Test connection for session trunking. Async exchange_id call */
8916 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8917 	if (IS_ERR(task))
8918 		return;
8919 
8920 	status = task->tk_status;
8921 	if (status == 0) {
8922 		status = nfs4_detect_session_trunking(adata->clp,
8923 				task->tk_msg.rpc_resp, xprt);
8924 		trace_nfs4_trunked_exchange_id(adata->clp,
8925 			xprt->address_strings[RPC_DISPLAY_ADDR], status);
8926 	}
8927 	if (status == 0)
8928 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8929 	else if (status != -NFS4ERR_DELAY && rpc_clnt_xprt_switch_has_addr(clnt,
8930 				(struct sockaddr *)&xprt->addr))
8931 		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8932 
8933 	rpc_put_task(task);
8934 	if (status == -NFS4ERR_DELAY) {
8935 		ssleep(1);
8936 		goto try_again;
8937 	}
8938 }
8939 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8940 
8941 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8942 		const struct cred *cred)
8943 {
8944 	struct rpc_message msg = {
8945 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8946 		.rpc_argp = clp,
8947 		.rpc_cred = cred,
8948 	};
8949 	int status;
8950 
8951 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
8952 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8953 	trace_nfs4_destroy_clientid(clp, status);
8954 	if (status)
8955 		dprintk("NFS: Got error %d from the server %s on "
8956 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
8957 	return status;
8958 }
8959 
8960 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8961 		const struct cred *cred)
8962 {
8963 	unsigned int loop;
8964 	int ret;
8965 
8966 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8967 		ret = _nfs4_proc_destroy_clientid(clp, cred);
8968 		switch (ret) {
8969 		case -NFS4ERR_DELAY:
8970 		case -NFS4ERR_CLIENTID_BUSY:
8971 			ssleep(1);
8972 			break;
8973 		default:
8974 			return ret;
8975 		}
8976 	}
8977 	return 0;
8978 }
8979 
8980 int nfs4_destroy_clientid(struct nfs_client *clp)
8981 {
8982 	const struct cred *cred;
8983 	int ret = 0;
8984 
8985 	if (clp->cl_mvops->minor_version < 1)
8986 		goto out;
8987 	if (clp->cl_exchange_flags == 0)
8988 		goto out;
8989 	if (clp->cl_preserve_clid)
8990 		goto out;
8991 	cred = nfs4_get_clid_cred(clp);
8992 	ret = nfs4_proc_destroy_clientid(clp, cred);
8993 	put_cred(cred);
8994 	switch (ret) {
8995 	case 0:
8996 	case -NFS4ERR_STALE_CLIENTID:
8997 		clp->cl_exchange_flags = 0;
8998 	}
8999 out:
9000 	return ret;
9001 }
9002 
9003 struct nfs4_get_lease_time_data {
9004 	struct nfs4_get_lease_time_args *args;
9005 	struct nfs4_get_lease_time_res *res;
9006 	struct nfs_client *clp;
9007 };
9008 
9009 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9010 					void *calldata)
9011 {
9012 	struct nfs4_get_lease_time_data *data =
9013 			(struct nfs4_get_lease_time_data *)calldata;
9014 
9015 	/* just setup sequence, do not trigger session recovery
9016 	   since we're invoked within one */
9017 	nfs4_setup_sequence(data->clp,
9018 			&data->args->la_seq_args,
9019 			&data->res->lr_seq_res,
9020 			task);
9021 }
9022 
9023 /*
9024  * Called from nfs4_state_manager thread for session setup, so don't recover
9025  * from sequence operation or clientid errors.
9026  */
9027 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9028 {
9029 	struct nfs4_get_lease_time_data *data =
9030 			(struct nfs4_get_lease_time_data *)calldata;
9031 
9032 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9033 		return;
9034 	switch (task->tk_status) {
9035 	case -NFS4ERR_DELAY:
9036 	case -NFS4ERR_GRACE:
9037 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9038 		task->tk_status = 0;
9039 		fallthrough;
9040 	case -NFS4ERR_RETRY_UNCACHED_REP:
9041 		rpc_restart_call_prepare(task);
9042 		return;
9043 	}
9044 }
9045 
9046 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9047 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9048 	.rpc_call_done = nfs4_get_lease_time_done,
9049 };
9050 
9051 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9052 {
9053 	struct nfs4_get_lease_time_args args;
9054 	struct nfs4_get_lease_time_res res = {
9055 		.lr_fsinfo = fsinfo,
9056 	};
9057 	struct nfs4_get_lease_time_data data = {
9058 		.args = &args,
9059 		.res = &res,
9060 		.clp = clp,
9061 	};
9062 	struct rpc_message msg = {
9063 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9064 		.rpc_argp = &args,
9065 		.rpc_resp = &res,
9066 	};
9067 	struct rpc_task_setup task_setup = {
9068 		.rpc_client = clp->cl_rpcclient,
9069 		.rpc_message = &msg,
9070 		.callback_ops = &nfs4_get_lease_time_ops,
9071 		.callback_data = &data,
9072 		.flags = RPC_TASK_TIMEOUT,
9073 	};
9074 
9075 	nfs4_init_sequence(clp, &args.la_seq_args, &res.lr_seq_res, 0, 1);
9076 	return nfs4_call_sync_custom(&task_setup);
9077 }
9078 
9079 /*
9080  * Initialize the values to be used by the client in CREATE_SESSION
9081  * If nfs4_init_session set the fore channel request and response sizes,
9082  * use them.
9083  *
9084  * Set the back channel max_resp_sz_cached to zero to force the client to
9085  * always set csa_cachethis to FALSE because the current implementation
9086  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9087  */
9088 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9089 				    struct rpc_clnt *clnt)
9090 {
9091 	unsigned int max_rqst_sz, max_resp_sz;
9092 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9093 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9094 
9095 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9096 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9097 
9098 	/* Fore channel attributes */
9099 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9100 	args->fc_attrs.max_resp_sz = max_resp_sz;
9101 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9102 	args->fc_attrs.max_reqs = max_session_slots;
9103 
9104 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9105 		"max_ops=%u max_reqs=%u\n",
9106 		__func__,
9107 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9108 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9109 
9110 	/* Back channel attributes */
9111 	args->bc_attrs.max_rqst_sz = max_bc_payload;
9112 	args->bc_attrs.max_resp_sz = max_bc_payload;
9113 	args->bc_attrs.max_resp_sz_cached = 0;
9114 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9115 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9116 	if (args->bc_attrs.max_reqs > max_bc_slots)
9117 		args->bc_attrs.max_reqs = max_bc_slots;
9118 
9119 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9120 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9121 		__func__,
9122 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9123 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9124 		args->bc_attrs.max_reqs);
9125 }
9126 
9127 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9128 		struct nfs41_create_session_res *res)
9129 {
9130 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9131 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9132 
9133 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9134 		return -EINVAL;
9135 	/*
9136 	 * Our requested max_ops is the minimum we need; we're not
9137 	 * prepared to break up compounds into smaller pieces than that.
9138 	 * So, no point even trying to continue if the server won't
9139 	 * cooperate:
9140 	 */
9141 	if (rcvd->max_ops < sent->max_ops)
9142 		return -EINVAL;
9143 	if (rcvd->max_reqs == 0)
9144 		return -EINVAL;
9145 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9146 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9147 	return 0;
9148 }
9149 
9150 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9151 		struct nfs41_create_session_res *res)
9152 {
9153 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9154 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9155 
9156 	if (!(res->flags & SESSION4_BACK_CHAN))
9157 		goto out;
9158 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9159 		return -EINVAL;
9160 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9161 		return -EINVAL;
9162 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9163 		return -EINVAL;
9164 	if (rcvd->max_ops > sent->max_ops)
9165 		return -EINVAL;
9166 	if (rcvd->max_reqs > sent->max_reqs)
9167 		return -EINVAL;
9168 out:
9169 	return 0;
9170 }
9171 
9172 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9173 				     struct nfs41_create_session_res *res)
9174 {
9175 	int ret;
9176 
9177 	ret = nfs4_verify_fore_channel_attrs(args, res);
9178 	if (ret)
9179 		return ret;
9180 	return nfs4_verify_back_channel_attrs(args, res);
9181 }
9182 
9183 static void nfs4_update_session(struct nfs4_session *session,
9184 		struct nfs41_create_session_res *res)
9185 {
9186 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9187 	/* Mark client id and session as being confirmed */
9188 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9189 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9190 	session->flags = res->flags;
9191 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9192 	if (res->flags & SESSION4_BACK_CHAN)
9193 		memcpy(&session->bc_attrs, &res->bc_attrs,
9194 				sizeof(session->bc_attrs));
9195 }
9196 
9197 static int _nfs4_proc_create_session(struct nfs_client *clp,
9198 		const struct cred *cred)
9199 {
9200 	struct nfs4_session *session = clp->cl_session;
9201 	struct nfs41_create_session_args args = {
9202 		.client = clp,
9203 		.clientid = clp->cl_clientid,
9204 		.seqid = clp->cl_seqid,
9205 		.cb_program = NFS4_CALLBACK,
9206 	};
9207 	struct nfs41_create_session_res res;
9208 
9209 	struct rpc_message msg = {
9210 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9211 		.rpc_argp = &args,
9212 		.rpc_resp = &res,
9213 		.rpc_cred = cred,
9214 	};
9215 	int status;
9216 
9217 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9218 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9219 
9220 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9221 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9222 	trace_nfs4_create_session(clp, status);
9223 
9224 	switch (status) {
9225 	case -NFS4ERR_STALE_CLIENTID:
9226 	case -NFS4ERR_DELAY:
9227 	case -ETIMEDOUT:
9228 	case -EACCES:
9229 	case -EAGAIN:
9230 		goto out;
9231 	}
9232 
9233 	clp->cl_seqid++;
9234 	if (!status) {
9235 		/* Verify the session's negotiated channel_attrs values */
9236 		status = nfs4_verify_channel_attrs(&args, &res);
9237 		/* Increment the clientid slot sequence id */
9238 		if (status)
9239 			goto out;
9240 		nfs4_update_session(session, &res);
9241 	}
9242 out:
9243 	return status;
9244 }
9245 
9246 /*
9247  * Issues a CREATE_SESSION operation to the server.
9248  * It is the responsibility of the caller to verify the session is
9249  * expired before calling this routine.
9250  */
9251 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9252 {
9253 	int status;
9254 	unsigned *ptr;
9255 	struct nfs4_session *session = clp->cl_session;
9256 	struct nfs4_add_xprt_data xprtdata = {
9257 		.clp = clp,
9258 	};
9259 	struct rpc_add_xprt_test rpcdata = {
9260 		.add_xprt_test = clp->cl_mvops->session_trunk,
9261 		.data = &xprtdata,
9262 	};
9263 
9264 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9265 
9266 	status = _nfs4_proc_create_session(clp, cred);
9267 	if (status)
9268 		goto out;
9269 
9270 	/* Init or reset the session slot tables */
9271 	status = nfs4_setup_session_slot_tables(session);
9272 	dprintk("slot table setup returned %d\n", status);
9273 	if (status)
9274 		goto out;
9275 
9276 	ptr = (unsigned *)&session->sess_id.data[0];
9277 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9278 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9279 	rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9280 out:
9281 	return status;
9282 }
9283 
9284 /*
9285  * Issue the over-the-wire RPC DESTROY_SESSION.
9286  * The caller must serialize access to this routine.
9287  */
9288 int nfs4_proc_destroy_session(struct nfs4_session *session,
9289 		const struct cred *cred)
9290 {
9291 	struct rpc_message msg = {
9292 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9293 		.rpc_argp = session,
9294 		.rpc_cred = cred,
9295 	};
9296 	int status = 0;
9297 
9298 	/* session is still being setup */
9299 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9300 		return 0;
9301 
9302 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9303 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9304 	trace_nfs4_destroy_session(session->clp, status);
9305 
9306 	if (status)
9307 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9308 			"Session has been destroyed regardless...\n", status);
9309 	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9310 	return status;
9311 }
9312 
9313 /*
9314  * Renew the cl_session lease.
9315  */
9316 struct nfs4_sequence_data {
9317 	struct nfs_client *clp;
9318 	struct nfs4_sequence_args args;
9319 	struct nfs4_sequence_res res;
9320 };
9321 
9322 static void nfs41_sequence_release(void *data)
9323 {
9324 	struct nfs4_sequence_data *calldata = data;
9325 	struct nfs_client *clp = calldata->clp;
9326 
9327 	if (refcount_read(&clp->cl_count) > 1)
9328 		nfs4_schedule_state_renewal(clp);
9329 	nfs_put_client(clp);
9330 	kfree(calldata);
9331 }
9332 
9333 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9334 {
9335 	switch(task->tk_status) {
9336 	case -NFS4ERR_DELAY:
9337 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9338 		return -EAGAIN;
9339 	default:
9340 		nfs4_schedule_lease_recovery(clp);
9341 	}
9342 	return 0;
9343 }
9344 
9345 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9346 {
9347 	struct nfs4_sequence_data *calldata = data;
9348 	struct nfs_client *clp = calldata->clp;
9349 
9350 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9351 		return;
9352 
9353 	trace_nfs4_sequence(clp, task->tk_status);
9354 	if (task->tk_status < 0 && clp->cl_cons_state >= 0) {
9355 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9356 		if (refcount_read(&clp->cl_count) == 1)
9357 			return;
9358 
9359 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9360 			rpc_restart_call_prepare(task);
9361 			return;
9362 		}
9363 	}
9364 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9365 }
9366 
9367 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9368 {
9369 	struct nfs4_sequence_data *calldata = data;
9370 	struct nfs_client *clp = calldata->clp;
9371 	struct nfs4_sequence_args *args;
9372 	struct nfs4_sequence_res *res;
9373 
9374 	args = task->tk_msg.rpc_argp;
9375 	res = task->tk_msg.rpc_resp;
9376 
9377 	nfs4_setup_sequence(clp, args, res, task);
9378 }
9379 
9380 static const struct rpc_call_ops nfs41_sequence_ops = {
9381 	.rpc_call_done = nfs41_sequence_call_done,
9382 	.rpc_call_prepare = nfs41_sequence_prepare,
9383 	.rpc_release = nfs41_sequence_release,
9384 };
9385 
9386 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9387 		const struct cred *cred,
9388 		struct nfs4_slot *slot,
9389 		bool is_privileged)
9390 {
9391 	struct nfs4_sequence_data *calldata;
9392 	struct rpc_message msg = {
9393 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9394 		.rpc_cred = cred,
9395 	};
9396 	struct rpc_task_setup task_setup_data = {
9397 		.rpc_client = clp->cl_rpcclient,
9398 		.rpc_message = &msg,
9399 		.callback_ops = &nfs41_sequence_ops,
9400 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9401 	};
9402 	struct rpc_task *ret;
9403 
9404 	ret = ERR_PTR(-EIO);
9405 	if (!refcount_inc_not_zero(&clp->cl_count))
9406 		goto out_err;
9407 
9408 	ret = ERR_PTR(-ENOMEM);
9409 	calldata = kzalloc_obj(*calldata);
9410 	if (calldata == NULL)
9411 		goto out_put_clp;
9412 	nfs4_init_sequence(clp, &calldata->args, &calldata->res, 0, is_privileged);
9413 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9414 	msg.rpc_argp = &calldata->args;
9415 	msg.rpc_resp = &calldata->res;
9416 	calldata->clp = clp;
9417 	task_setup_data.callback_data = calldata;
9418 
9419 	ret = rpc_run_task(&task_setup_data);
9420 	if (IS_ERR(ret))
9421 		goto out_err;
9422 	return ret;
9423 out_put_clp:
9424 	nfs_put_client(clp);
9425 out_err:
9426 	nfs41_release_slot(slot);
9427 	return ret;
9428 }
9429 
9430 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9431 {
9432 	struct rpc_task *task;
9433 	int ret = 0;
9434 
9435 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9436 		return -EAGAIN;
9437 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9438 	if (IS_ERR(task))
9439 		ret = PTR_ERR(task);
9440 	else
9441 		rpc_put_task_async(task);
9442 	dprintk("<-- %s status=%d\n", __func__, ret);
9443 	return ret;
9444 }
9445 
9446 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9447 {
9448 	struct rpc_task *task;
9449 	int ret;
9450 
9451 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9452 	if (IS_ERR(task)) {
9453 		ret = PTR_ERR(task);
9454 		goto out;
9455 	}
9456 	ret = rpc_wait_for_completion_task(task);
9457 	if (!ret)
9458 		ret = task->tk_status;
9459 	rpc_put_task(task);
9460 out:
9461 	dprintk("<-- %s status=%d\n", __func__, ret);
9462 	return ret;
9463 }
9464 
9465 struct nfs4_reclaim_complete_data {
9466 	struct nfs_client *clp;
9467 	struct nfs41_reclaim_complete_args arg;
9468 	struct nfs41_reclaim_complete_res res;
9469 };
9470 
9471 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9472 {
9473 	struct nfs4_reclaim_complete_data *calldata = data;
9474 
9475 	nfs4_setup_sequence(calldata->clp,
9476 			&calldata->arg.seq_args,
9477 			&calldata->res.seq_res,
9478 			task);
9479 }
9480 
9481 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9482 {
9483 	switch(task->tk_status) {
9484 	case 0:
9485 		wake_up_all(&clp->cl_lock_waitq);
9486 		fallthrough;
9487 	case -NFS4ERR_COMPLETE_ALREADY:
9488 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9489 		break;
9490 	case -NFS4ERR_DELAY:
9491 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9492 		fallthrough;
9493 	case -NFS4ERR_RETRY_UNCACHED_REP:
9494 	case -EACCES:
9495 		dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9496 			__func__, task->tk_status, clp->cl_hostname);
9497 		return -EAGAIN;
9498 	case -NFS4ERR_BADSESSION:
9499 	case -NFS4ERR_DEADSESSION:
9500 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9501 		break;
9502 	default:
9503 		nfs4_schedule_lease_recovery(clp);
9504 	}
9505 	return 0;
9506 }
9507 
9508 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9509 {
9510 	struct nfs4_reclaim_complete_data *calldata = data;
9511 	struct nfs_client *clp = calldata->clp;
9512 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9513 
9514 	if (!nfs41_sequence_done(task, res))
9515 		return;
9516 
9517 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9518 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9519 		rpc_restart_call_prepare(task);
9520 		return;
9521 	}
9522 }
9523 
9524 static void nfs4_free_reclaim_complete_data(void *data)
9525 {
9526 	struct nfs4_reclaim_complete_data *calldata = data;
9527 
9528 	kfree(calldata);
9529 }
9530 
9531 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9532 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9533 	.rpc_call_done = nfs4_reclaim_complete_done,
9534 	.rpc_release = nfs4_free_reclaim_complete_data,
9535 };
9536 
9537 /*
9538  * Issue a global reclaim complete.
9539  */
9540 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9541 		const struct cred *cred)
9542 {
9543 	struct nfs4_reclaim_complete_data *calldata;
9544 	struct rpc_message msg = {
9545 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9546 		.rpc_cred = cred,
9547 	};
9548 	struct rpc_task_setup task_setup_data = {
9549 		.rpc_client = clp->cl_rpcclient,
9550 		.rpc_message = &msg,
9551 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9552 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9553 	};
9554 	int status = -ENOMEM;
9555 
9556 	calldata = kzalloc_obj(*calldata, GFP_NOFS);
9557 	if (calldata == NULL)
9558 		goto out;
9559 	calldata->clp = clp;
9560 	calldata->arg.one_fs = 0;
9561 
9562 	nfs4_init_sequence(clp, &calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9563 	msg.rpc_argp = &calldata->arg;
9564 	msg.rpc_resp = &calldata->res;
9565 	task_setup_data.callback_data = calldata;
9566 	status = nfs4_call_sync_custom(&task_setup_data);
9567 out:
9568 	dprintk("<-- %s status=%d\n", __func__, status);
9569 	return status;
9570 }
9571 
9572 static void
9573 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9574 {
9575 	struct nfs4_layoutget *lgp = calldata;
9576 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9577 
9578 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9579 				&lgp->res.seq_res, task);
9580 }
9581 
9582 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9583 {
9584 	struct nfs4_layoutget *lgp = calldata;
9585 
9586 	nfs41_sequence_process(task, &lgp->res.seq_res);
9587 }
9588 
9589 static int
9590 nfs4_layoutget_handle_exception(struct rpc_task *task,
9591 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9592 {
9593 	struct inode *inode = lgp->args.inode;
9594 	struct nfs_server *server = NFS_SERVER(inode);
9595 	struct pnfs_layout_hdr *lo = lgp->lo;
9596 	int nfs4err = task->tk_status;
9597 	int err, status = 0;
9598 	LIST_HEAD(head);
9599 
9600 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9601 
9602 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9603 
9604 	exception->state = NULL;
9605 	exception->stateid = NULL;
9606 
9607 	switch (nfs4err) {
9608 	case 0:
9609 		goto out;
9610 
9611 	/*
9612 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9613 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9614 	 * retry go inband.
9615 	 */
9616 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9617 		status = -ENODATA;
9618 		goto out;
9619 	/*
9620 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9621 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9622 	 */
9623 	case -NFS4ERR_BADLAYOUT:
9624 		status = -EOVERFLOW;
9625 		goto out;
9626 	/*
9627 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9628 	 * (or clients) writing to the same RAID stripe except when
9629 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9630 	 *
9631 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9632 	 * while, and then eventually give up.
9633 	 */
9634 	case -NFS4ERR_LAYOUTTRYLATER:
9635 		if (lgp->args.minlength == 0) {
9636 			status = -EOVERFLOW;
9637 			goto out;
9638 		}
9639 		status = -EBUSY;
9640 		break;
9641 	case -NFS4ERR_RECALLCONFLICT:
9642 	case -NFS4ERR_RETURNCONFLICT:
9643 		status = -ERECALLCONFLICT;
9644 		break;
9645 	case -NFS4ERR_DELEG_REVOKED:
9646 	case -NFS4ERR_ADMIN_REVOKED:
9647 	case -NFS4ERR_EXPIRED:
9648 	case -NFS4ERR_BAD_STATEID:
9649 		exception->timeout = 0;
9650 		spin_lock(&inode->i_lock);
9651 		/* If the open stateid was bad, then recover it. */
9652 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9653 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9654 			spin_unlock(&inode->i_lock);
9655 			exception->state = lgp->args.ctx->state;
9656 			exception->stateid = &lgp->args.stateid;
9657 			break;
9658 		}
9659 
9660 		/*
9661 		 * Mark the bad layout state as invalid, then retry
9662 		 */
9663 		pnfs_mark_layout_stateid_invalid(lo, &head);
9664 		spin_unlock(&inode->i_lock);
9665 		nfs_commit_inode(inode, 0);
9666 		pnfs_free_lseg_list(&head);
9667 		status = -EAGAIN;
9668 		goto out;
9669 	}
9670 
9671 	err = nfs4_handle_exception(server, nfs4err, exception);
9672 	if (!status) {
9673 		if (exception->retry)
9674 			status = -EAGAIN;
9675 		else
9676 			status = err;
9677 	}
9678 out:
9679 	return status;
9680 }
9681 
9682 size_t max_response_pages(struct nfs_server *server)
9683 {
9684 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9685 	return nfs_page_array_len(0, max_resp_sz);
9686 }
9687 
9688 static void nfs4_layoutget_release(void *calldata)
9689 {
9690 	struct nfs4_layoutget *lgp = calldata;
9691 
9692 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9693 	pnfs_layoutget_free(lgp);
9694 }
9695 
9696 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9697 	.rpc_call_prepare = nfs4_layoutget_prepare,
9698 	.rpc_call_done = nfs4_layoutget_done,
9699 	.rpc_release = nfs4_layoutget_release,
9700 };
9701 
9702 struct pnfs_layout_segment *
9703 nfs4_proc_layoutget(struct nfs4_layoutget *lgp,
9704 		    struct nfs4_exception *exception)
9705 {
9706 	struct inode *inode = lgp->args.inode;
9707 	struct nfs_server *server = NFS_SERVER(inode);
9708 	struct rpc_task *task;
9709 	struct rpc_message msg = {
9710 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9711 		.rpc_argp = &lgp->args,
9712 		.rpc_resp = &lgp->res,
9713 		.rpc_cred = lgp->cred,
9714 	};
9715 	struct rpc_task_setup task_setup_data = {
9716 		.rpc_client = server->client,
9717 		.rpc_message = &msg,
9718 		.callback_ops = &nfs4_layoutget_call_ops,
9719 		.callback_data = lgp,
9720 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9721 			 RPC_TASK_MOVEABLE,
9722 	};
9723 	struct pnfs_layout_segment *lseg = NULL;
9724 	int status = 0;
9725 
9726 	nfs4_init_sequence(server->nfs_client, &lgp->args.seq_args,
9727 			   &lgp->res.seq_res, 0, 0);
9728 	exception->retry = 0;
9729 
9730 	task = rpc_run_task(&task_setup_data);
9731 	if (IS_ERR(task))
9732 		return ERR_CAST(task);
9733 
9734 	status = rpc_wait_for_completion_task(task);
9735 	if (status != 0)
9736 		goto out;
9737 
9738 	if (task->tk_status < 0) {
9739 		exception->retry = 1;
9740 		status = nfs4_layoutget_handle_exception(task, lgp, exception);
9741 	} else if (lgp->res.layoutp->len == 0) {
9742 		exception->retry = 1;
9743 		status = -EAGAIN;
9744 		nfs4_update_delay(&exception->timeout);
9745 	} else
9746 		lseg = pnfs_layout_process(lgp);
9747 out:
9748 	trace_nfs4_layoutget(lgp->args.ctx,
9749 			&lgp->args.range,
9750 			&lgp->res.range,
9751 			&lgp->res.stateid,
9752 			status);
9753 
9754 	rpc_put_task(task);
9755 	dprintk("<-- %s status=%d\n", __func__, status);
9756 	if (status)
9757 		return ERR_PTR(status);
9758 	return lseg;
9759 }
9760 
9761 static void
9762 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9763 {
9764 	struct nfs4_layoutreturn *lrp = calldata;
9765 
9766 	nfs4_setup_sequence(lrp->clp,
9767 			&lrp->args.seq_args,
9768 			&lrp->res.seq_res,
9769 			task);
9770 	if (!pnfs_layout_is_valid(lrp->args.layout))
9771 		rpc_exit(task, 0);
9772 }
9773 
9774 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9775 {
9776 	struct nfs4_layoutreturn *lrp = calldata;
9777 	struct nfs_server *server;
9778 
9779 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9780 		return;
9781 
9782 	if (task->tk_rpc_status < 0) {
9783 		switch (task->tk_rpc_status) {
9784 		case -EACCES:
9785 		case -EIO:
9786 		case -EKEYEXPIRED:
9787 		case -ERESTARTSYS:
9788 		case -EINTR:
9789 			lrp->rpc_status = 0;
9790 			break;
9791 		case -ENETDOWN:
9792 		case -ENETUNREACH:
9793 			if (task->tk_flags & RPC_TASK_NETUNREACH_FATAL)
9794 				lrp->rpc_status = 0;
9795 			else
9796 				lrp->rpc_status = -EAGAIN;
9797 			break;
9798 		default:
9799 			lrp->rpc_status = -EAGAIN;
9800 			break;
9801 		}
9802 		lrp->res.lrs_present = 0;
9803 		return;
9804 	}
9805 
9806 	server = NFS_SERVER(lrp->args.inode);
9807 	switch (task->tk_status) {
9808 	case -NFS4ERR_OLD_STATEID:
9809 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9810 					&lrp->args.range,
9811 					lrp->args.inode))
9812 			goto out_restart;
9813 		fallthrough;
9814 	default:
9815 		task->tk_status = 0;
9816 		lrp->res.lrs_present = 0;
9817 		fallthrough;
9818 	case 0:
9819 		break;
9820 	case -NFS4ERR_BADSESSION:
9821 	case -NFS4ERR_DEADSESSION:
9822 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9823 		nfs4_schedule_session_recovery(server->nfs_client->cl_session,
9824 					       task->tk_status);
9825 		lrp->res.lrs_present = 0;
9826 		lrp->rpc_status = -EAGAIN;
9827 		task->tk_status = 0;
9828 		break;
9829 	case -NFS4ERR_DELAY:
9830 		if (nfs4_async_handle_error(task, server, NULL, NULL) ==
9831 		    -EAGAIN)
9832 			goto out_restart;
9833 		lrp->res.lrs_present = 0;
9834 		break;
9835 	}
9836 	return;
9837 out_restart:
9838 	task->tk_status = 0;
9839 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9840 	rpc_restart_call_prepare(task);
9841 }
9842 
9843 static void nfs4_layoutreturn_release(void *calldata)
9844 {
9845 	struct nfs4_layoutreturn *lrp = calldata;
9846 	struct pnfs_layout_hdr *lo = lrp->args.layout;
9847 
9848 	if (lrp->rpc_status == 0 || !lrp->inode)
9849 		pnfs_layoutreturn_free_lsegs(
9850 			lo, &lrp->args.stateid, &lrp->args.range,
9851 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9852 	else
9853 		pnfs_layoutreturn_retry_later(lo, &lrp->args.stateid,
9854 					      &lrp->args.range);
9855 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9856 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9857 		lrp->ld_private.ops->free(&lrp->ld_private);
9858 	pnfs_put_layout_hdr(lrp->args.layout);
9859 	nfs_iput_and_deactive(lrp->inode);
9860 	put_cred(lrp->cred);
9861 	kfree(calldata);
9862 }
9863 
9864 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9865 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9866 	.rpc_call_done = nfs4_layoutreturn_done,
9867 	.rpc_release = nfs4_layoutreturn_release,
9868 };
9869 
9870 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, unsigned int flags)
9871 {
9872 	struct nfs_client *clp = NFS_SERVER(lrp->args.inode)->nfs_client;
9873 	struct rpc_task *task;
9874 	struct rpc_message msg = {
9875 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9876 		.rpc_argp = &lrp->args,
9877 		.rpc_resp = &lrp->res,
9878 		.rpc_cred = lrp->cred,
9879 	};
9880 	struct rpc_task_setup task_setup_data = {
9881 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9882 		.rpc_message = &msg,
9883 		.callback_ops = &nfs4_layoutreturn_call_ops,
9884 		.callback_data = lrp,
9885 		.flags = RPC_TASK_MOVEABLE,
9886 	};
9887 	int status = 0;
9888 
9889 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9890 			   &task_setup_data.rpc_client, &msg);
9891 
9892 	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9893 	if (flags & PNFS_FL_LAYOUTRETURN_ASYNC) {
9894 		if (!lrp->inode) {
9895 			nfs4_layoutreturn_release(lrp);
9896 			return -EAGAIN;
9897 		}
9898 		task_setup_data.flags |= RPC_TASK_ASYNC;
9899 	}
9900 	if (!lrp->inode)
9901 		flags |= PNFS_FL_LAYOUTRETURN_PRIVILEGED;
9902 
9903 	nfs4_init_sequence(clp, &lrp->args.seq_args, &lrp->res.seq_res, 1,
9904 			   flags & PNFS_FL_LAYOUTRETURN_PRIVILEGED ? 1 : 0);
9905 	task = rpc_run_task(&task_setup_data);
9906 	if (IS_ERR(task))
9907 		return PTR_ERR(task);
9908 	if (!(flags & PNFS_FL_LAYOUTRETURN_ASYNC))
9909 		status = task->tk_status;
9910 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9911 	dprintk("<-- %s status=%d\n", __func__, status);
9912 	rpc_put_task(task);
9913 	return status;
9914 }
9915 
9916 static int
9917 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9918 		struct pnfs_device *pdev,
9919 		const struct cred *cred)
9920 {
9921 	struct nfs4_getdeviceinfo_args args = {
9922 		.pdev = pdev,
9923 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9924 			NOTIFY_DEVICEID4_DELETE,
9925 	};
9926 	struct nfs4_getdeviceinfo_res res = {
9927 		.pdev = pdev,
9928 	};
9929 	struct rpc_message msg = {
9930 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9931 		.rpc_argp = &args,
9932 		.rpc_resp = &res,
9933 		.rpc_cred = cred,
9934 	};
9935 	int status;
9936 
9937 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9938 	if (res.notification & ~args.notify_types)
9939 		dprintk("%s: unsupported notification\n", __func__);
9940 	if (res.notification != args.notify_types)
9941 		pdev->nocache = 1;
9942 
9943 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9944 
9945 	dprintk("<-- %s status=%d\n", __func__, status);
9946 
9947 	return status;
9948 }
9949 
9950 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9951 		struct pnfs_device *pdev,
9952 		const struct cred *cred)
9953 {
9954 	struct nfs4_exception exception = { };
9955 	int err;
9956 
9957 	do {
9958 		err = nfs4_handle_exception(server,
9959 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9960 					&exception);
9961 	} while (exception.retry);
9962 	return err;
9963 }
9964 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9965 
9966 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9967 {
9968 	struct nfs4_layoutcommit_data *data = calldata;
9969 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9970 
9971 	nfs4_setup_sequence(server->nfs_client,
9972 			&data->args.seq_args,
9973 			&data->res.seq_res,
9974 			task);
9975 }
9976 
9977 static void
9978 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9979 {
9980 	struct nfs4_layoutcommit_data *data = calldata;
9981 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9982 
9983 	if (!nfs41_sequence_done(task, &data->res.seq_res))
9984 		return;
9985 
9986 	switch (task->tk_status) { /* Just ignore these failures */
9987 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9988 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9989 	case -NFS4ERR_BADLAYOUT:     /* no layout */
9990 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
9991 		task->tk_status = 0;
9992 		break;
9993 	case -NFS4ERR_OLD_STATEID: {
9994 		u32 old_seqid = be32_to_cpu(data->args.stateid.seqid);
9995 		struct pnfs_layout_range range = {
9996 			.iomode = IOMODE_ANY,
9997 			.offset = 0,
9998 			.length = NFS4_MAX_UINT64,
9999 		};
10000 
10001 		if (nfs4_layout_refresh_old_stateid(&data->args.stateid,
10002 						    &range,
10003 						    data->args.inode)) {
10004 			struct pnfs_layout_hdr *lo;
10005 
10006 			spin_lock(&data->args.inode->i_lock);
10007 			lo = NFS_I(data->args.inode)->layout;
10008 			if (lo && pnfs_layout_is_valid(lo) &&
10009 			    nfs4_stateid_match_other(&data->args.stateid,
10010 						     &lo->plh_stateid))
10011 				pnfs_set_layout_stateid(lo, &data->args.stateid,
10012 							NULL, false);
10013 			spin_unlock(&data->args.inode->i_lock);
10014 
10015 			dprintk("%s: refreshed OLD_STATEID inode %llu seq %u->%u\n",
10016 				__func__, data->args.inode->i_ino,
10017 				old_seqid,
10018 				be32_to_cpu(data->args.stateid.seqid));
10019 
10020 			rpc_restart_call_prepare(task);
10021 			return;
10022 		}
10023 		fallthrough;
10024 	}
10025 	case 0:
10026 		break;
10027 	default:
10028 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
10029 			rpc_restart_call_prepare(task);
10030 			return;
10031 		}
10032 	}
10033 }
10034 
10035 static void nfs4_layoutcommit_release(void *calldata)
10036 {
10037 	struct nfs4_layoutcommit_data *data = calldata;
10038 
10039 	pnfs_cleanup_layoutcommit(data);
10040 	nfs_post_op_update_inode_force_wcc(data->args.inode,
10041 					   data->res.fattr);
10042 	put_cred(data->cred);
10043 	nfs_iput_and_deactive(data->inode);
10044 	kfree(data);
10045 }
10046 
10047 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
10048 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
10049 	.rpc_call_done = nfs4_layoutcommit_done,
10050 	.rpc_release = nfs4_layoutcommit_release,
10051 };
10052 
10053 int
10054 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10055 {
10056 	struct rpc_message msg = {
10057 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10058 		.rpc_argp = &data->args,
10059 		.rpc_resp = &data->res,
10060 		.rpc_cred = data->cred,
10061 	};
10062 	struct rpc_task_setup task_setup_data = {
10063 		.task = &data->task,
10064 		.rpc_client = NFS_CLIENT(data->args.inode),
10065 		.rpc_message = &msg,
10066 		.callback_ops = &nfs4_layoutcommit_ops,
10067 		.callback_data = data,
10068 		.flags = RPC_TASK_MOVEABLE,
10069 	};
10070 	struct rpc_task *task;
10071 	int status = 0;
10072 
10073 	dprintk("NFS: initiating layoutcommit call. sync %d "
10074 		"lbw: %llu inode %llu\n", sync,
10075 		data->args.lastbytewritten,
10076 		data->args.inode->i_ino);
10077 
10078 	if (!sync) {
10079 		data->inode = nfs_igrab_and_active(data->args.inode);
10080 		if (data->inode == NULL) {
10081 			nfs4_layoutcommit_release(data);
10082 			return -EAGAIN;
10083 		}
10084 		task_setup_data.flags = RPC_TASK_ASYNC;
10085 	}
10086 	nfs4_init_sequence(NFS_SERVER(data->args.inode)->nfs_client,
10087 			   &data->args.seq_args, &data->res.seq_res, 1, 0);
10088 	task = rpc_run_task(&task_setup_data);
10089 	if (IS_ERR(task))
10090 		return PTR_ERR(task);
10091 	if (sync)
10092 		status = task->tk_status;
10093 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10094 	dprintk("%s: status %d\n", __func__, status);
10095 	rpc_put_task(task);
10096 	return status;
10097 }
10098 
10099 /*
10100  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10101  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10102  */
10103 static int _nfs41_proc_secinfo_no_name(struct nfs_server *server,
10104 				       struct nfs_fh *fhandle,
10105 				       struct nfs4_secinfo_flavors *flavors,
10106 				       bool use_integrity)
10107 {
10108 	struct nfs_client *clp = server->nfs_client;
10109 	struct nfs41_secinfo_no_name_args args = {
10110 		.style = SECINFO_STYLE_CURRENT_FH,
10111 	};
10112 	struct nfs4_secinfo_res res = {
10113 		.flavors = flavors,
10114 	};
10115 	struct rpc_message msg = {
10116 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10117 		.rpc_argp = &args,
10118 		.rpc_resp = &res,
10119 	};
10120 	struct nfs4_call_sync_data data = {
10121 		.seq_server = server,
10122 		.seq_args = &args.seq_args,
10123 		.seq_res = &res.seq_res,
10124 	};
10125 	struct rpc_task_setup task_setup = {
10126 		.rpc_client = server->client,
10127 		.rpc_message = &msg,
10128 		.callback_ops = clp->cl_mvops->call_sync_ops,
10129 		.callback_data = &data,
10130 		.flags = RPC_TASK_NO_ROUND_ROBIN,
10131 	};
10132 	const struct cred *cred = NULL;
10133 	int status;
10134 
10135 	if (use_integrity) {
10136 		task_setup.rpc_client = clp->cl_rpcclient;
10137 
10138 		cred = nfs4_get_clid_cred(clp);
10139 		msg.rpc_cred = cred;
10140 	}
10141 
10142 	nfs4_init_sequence(clp, &args.seq_args, &res.seq_res, 0, 0);
10143 	status = nfs4_call_sync_custom(&task_setup);
10144 	dprintk("<-- %s status=%d\n", __func__, status);
10145 
10146 	put_cred(cred);
10147 
10148 	return status;
10149 }
10150 
10151 static int nfs41_proc_secinfo_no_name(struct nfs_server *server,
10152 				      struct nfs_fh *fhandle,
10153 				      struct nfs4_secinfo_flavors *flavors)
10154 {
10155 	struct nfs4_exception exception = {
10156 		.interruptible = true,
10157 	};
10158 	int err;
10159 	do {
10160 		/* first try using integrity protection */
10161 		err = -NFS4ERR_WRONGSEC;
10162 
10163 		/* try to use integrity protection with machine cred */
10164 		if (_nfs4_is_integrity_protected(server->nfs_client))
10165 			err = _nfs41_proc_secinfo_no_name(server, fhandle,
10166 							  flavors, true);
10167 
10168 		/*
10169 		 * if unable to use integrity protection, or SECINFO with
10170 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10171 		 * disallowed by spec, but exists in deployed servers) use
10172 		 * the current filesystem's rpc_client and the user cred.
10173 		 */
10174 		if (err == -NFS4ERR_WRONGSEC)
10175 			err = _nfs41_proc_secinfo_no_name(server, fhandle,
10176 							  flavors, false);
10177 
10178 		switch (err) {
10179 		case 0:
10180 		case -NFS4ERR_WRONGSEC:
10181 		case -ENOTSUPP:
10182 			goto out;
10183 		default:
10184 			err = nfs4_handle_exception(server, err, &exception);
10185 		}
10186 	} while (exception.retry);
10187 out:
10188 	return err;
10189 }
10190 
10191 static int nfs41_find_root_sec(struct nfs_server *server,
10192 			       struct nfs_fh *fhandle, struct nfs_fattr *fattr)
10193 {
10194 	int err;
10195 	struct page *page;
10196 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10197 	struct nfs4_secinfo_flavors *flavors;
10198 	struct nfs4_secinfo4 *secinfo;
10199 	int i;
10200 
10201 	page = alloc_page(GFP_KERNEL);
10202 	if (!page) {
10203 		err = -ENOMEM;
10204 		goto out;
10205 	}
10206 
10207 	flavors = page_address(page);
10208 	err = nfs41_proc_secinfo_no_name(server, fhandle, flavors);
10209 
10210 	/*
10211 	 * Fall back on "guess and check" method if
10212 	 * the server doesn't support SECINFO_NO_NAME
10213 	 */
10214 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10215 		err = nfs4_find_root_sec(server, fhandle, fattr);
10216 		goto out_freepage;
10217 	}
10218 	if (err)
10219 		goto out_freepage;
10220 
10221 	for (i = 0; i < flavors->num_flavors; i++) {
10222 		secinfo = &flavors->flavors[i];
10223 
10224 		switch (secinfo->flavor) {
10225 		case RPC_AUTH_NULL:
10226 		case RPC_AUTH_UNIX:
10227 		case RPC_AUTH_GSS:
10228 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10229 					&secinfo->flavor_info);
10230 			break;
10231 		default:
10232 			flavor = RPC_AUTH_MAXFLAVOR;
10233 			break;
10234 		}
10235 
10236 		if (!nfs_auth_info_match(&server->auth_info, flavor))
10237 			flavor = RPC_AUTH_MAXFLAVOR;
10238 
10239 		if (flavor != RPC_AUTH_MAXFLAVOR) {
10240 			err = nfs4_lookup_root_sec(server, fhandle, fattr,
10241 						   flavor);
10242 			if (!err)
10243 				break;
10244 		}
10245 	}
10246 
10247 	if (flavor == RPC_AUTH_MAXFLAVOR)
10248 		err = -EPERM;
10249 
10250 out_freepage:
10251 	put_page(page);
10252 	if (err == -EACCES)
10253 		return -EPERM;
10254 out:
10255 	return err;
10256 }
10257 
10258 static int _nfs41_test_stateid(struct nfs_server *server,
10259 			       const nfs4_stateid *stateid,
10260 			       const struct cred *cred)
10261 {
10262 	int status;
10263 	struct nfs41_test_stateid_args args = {
10264 		.stateid = *stateid,
10265 	};
10266 	struct nfs41_test_stateid_res res;
10267 	struct rpc_message msg = {
10268 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10269 		.rpc_argp = &args,
10270 		.rpc_resp = &res,
10271 		.rpc_cred = cred,
10272 	};
10273 	struct rpc_clnt *rpc_client = server->client;
10274 	struct nfs_client *clp = server->nfs_client;
10275 
10276 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_STATEID, &rpc_client, &msg);
10277 
10278 	dprintk("NFS call  test_stateid %p\n", stateid);
10279 	nfs4_init_sequence(clp, &args.seq_args, &res.seq_res, 0, 1);
10280 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10281 			&args.seq_args, &res.seq_res);
10282 	if (status != NFS_OK) {
10283 		dprintk("NFS reply test_stateid: failed, %d\n", status);
10284 		return status;
10285 	}
10286 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10287 	return -res.status;
10288 }
10289 
10290 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10291 		int err, struct nfs4_exception *exception)
10292 {
10293 	exception->retry = 0;
10294 	switch(err) {
10295 	case -NFS4ERR_DELAY:
10296 	case -NFS4ERR_RETRY_UNCACHED_REP:
10297 		nfs4_handle_exception(server, err, exception);
10298 		break;
10299 	case -NFS4ERR_BADSESSION:
10300 	case -NFS4ERR_BADSLOT:
10301 	case -NFS4ERR_BAD_HIGH_SLOT:
10302 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10303 	case -NFS4ERR_DEADSESSION:
10304 		nfs4_do_handle_exception(server, err, exception);
10305 	}
10306 }
10307 
10308 /**
10309  * nfs41_test_stateid - perform a TEST_STATEID operation
10310  *
10311  * @server: server / transport on which to perform the operation
10312  * @stateid: state ID to test
10313  * @cred: credential
10314  *
10315  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10316  * Otherwise a negative NFS4ERR value is returned if the operation
10317  * failed or the state ID is not currently valid.
10318  */
10319 static int nfs41_test_stateid(struct nfs_server *server,
10320 			      const nfs4_stateid *stateid,
10321 			      const struct cred *cred)
10322 {
10323 	struct nfs4_exception exception = {
10324 		.interruptible = true,
10325 	};
10326 	int err;
10327 	do {
10328 		err = _nfs41_test_stateid(server, stateid, cred);
10329 		nfs4_handle_delay_or_session_error(server, err, &exception);
10330 	} while (exception.retry);
10331 	return err;
10332 }
10333 
10334 struct nfs_free_stateid_data {
10335 	struct nfs_server *server;
10336 	struct nfs41_free_stateid_args args;
10337 	struct nfs41_free_stateid_res res;
10338 };
10339 
10340 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10341 {
10342 	struct nfs_free_stateid_data *data = calldata;
10343 	nfs4_setup_sequence(data->server->nfs_client,
10344 			&data->args.seq_args,
10345 			&data->res.seq_res,
10346 			task);
10347 }
10348 
10349 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10350 {
10351 	struct nfs_free_stateid_data *data = calldata;
10352 
10353 	nfs41_sequence_done(task, &data->res.seq_res);
10354 
10355 	switch (task->tk_status) {
10356 	case -NFS4ERR_DELAY:
10357 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10358 			rpc_restart_call_prepare(task);
10359 	}
10360 }
10361 
10362 static void nfs41_free_stateid_release(void *calldata)
10363 {
10364 	struct nfs_free_stateid_data *data = calldata;
10365 	struct nfs_client *clp = data->server->nfs_client;
10366 
10367 	nfs_put_client(clp);
10368 	kfree(calldata);
10369 }
10370 
10371 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10372 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10373 	.rpc_call_done = nfs41_free_stateid_done,
10374 	.rpc_release = nfs41_free_stateid_release,
10375 };
10376 
10377 /**
10378  * nfs41_free_stateid - perform a FREE_STATEID operation
10379  *
10380  * @server: server / transport on which to perform the operation
10381  * @stateid: state ID to release
10382  * @cred: credential
10383  * @privileged: set to true if this call needs to be privileged
10384  *
10385  * Note: this function is always asynchronous.
10386  */
10387 static int nfs41_free_stateid(struct nfs_server *server,
10388 		nfs4_stateid *stateid,
10389 		const struct cred *cred,
10390 		bool privileged)
10391 {
10392 	struct rpc_message msg = {
10393 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10394 		.rpc_cred = cred,
10395 	};
10396 	struct rpc_task_setup task_setup = {
10397 		.rpc_client = server->client,
10398 		.rpc_message = &msg,
10399 		.callback_ops = &nfs41_free_stateid_ops,
10400 		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10401 	};
10402 	struct nfs_free_stateid_data *data;
10403 	struct rpc_task *task;
10404 	struct nfs_client *clp = server->nfs_client;
10405 
10406 	if (!refcount_inc_not_zero(&clp->cl_count))
10407 		return -EIO;
10408 
10409 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_STATEID,
10410 		&task_setup.rpc_client, &msg);
10411 
10412 	dprintk("NFS call  free_stateid %p\n", stateid);
10413 	data = kmalloc_obj(*data);
10414 	if (!data)
10415 		return -ENOMEM;
10416 	data->server = server;
10417 	nfs4_stateid_copy(&data->args.stateid, stateid);
10418 
10419 	task_setup.callback_data = data;
10420 
10421 	msg.rpc_argp = &data->args;
10422 	msg.rpc_resp = &data->res;
10423 	nfs4_init_sequence(clp, &data->args.seq_args, &data->res.seq_res, 1,
10424 			   privileged);
10425 	task = rpc_run_task(&task_setup);
10426 	if (IS_ERR(task))
10427 		return PTR_ERR(task);
10428 	rpc_put_task(task);
10429 	stateid->type = NFS4_FREED_STATEID_TYPE;
10430 	return 0;
10431 }
10432 
10433 static void
10434 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10435 {
10436 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10437 
10438 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10439 	nfs4_free_lock_state(server, lsp);
10440 }
10441 
10442 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10443 		const nfs4_stateid *s2)
10444 {
10445 	trace_nfs41_match_stateid(s1, s2);
10446 
10447 	if (s1->type != s2->type)
10448 		return false;
10449 
10450 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10451 		return false;
10452 
10453 	if (s1->seqid == s2->seqid)
10454 		return true;
10455 
10456 	return s1->seqid == 0 || s2->seqid == 0;
10457 }
10458 
10459 bool nfs4_match_stateid(const nfs4_stateid *s1,
10460 		const nfs4_stateid *s2)
10461 {
10462 	trace_nfs4_match_stateid(s1, s2);
10463 
10464 	return nfs4_stateid_match(s1, s2);
10465 }
10466 
10467 
10468 static const struct nfs4_sequence_slot_ops nfs41_sequence_slot_ops = {
10469 	.process = nfs41_sequence_process,
10470 	.done = nfs41_sequence_done,
10471 	.free_slot = nfs41_sequence_free_slot,
10472 };
10473 
10474 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10475 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10476 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10477 	.recover_open	= nfs4_open_reclaim,
10478 	.recover_lock	= nfs4_lock_reclaim,
10479 	.establish_clid = nfs41_init_clientid,
10480 	.reclaim_complete = nfs41_proc_reclaim_complete,
10481 	.detect_trunking = nfs41_discover_server_trunking,
10482 };
10483 
10484 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10485 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10486 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10487 	.recover_open	= nfs41_open_expired,
10488 	.recover_lock	= nfs41_lock_expired,
10489 	.establish_clid = nfs41_init_clientid,
10490 };
10491 
10492 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10493 	.sched_state_renewal = nfs41_proc_async_sequence,
10494 	.get_state_renewal_cred = nfs4_get_machine_cred,
10495 	.renew_lease = nfs4_proc_sequence,
10496 };
10497 
10498 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10499 	.get_locations = _nfs41_proc_get_locations,
10500 	.fsid_present = _nfs41_proc_fsid_present,
10501 };
10502 
10503 static struct nfs_seqid *
10504 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10505 {
10506 	return NULL;
10507 }
10508 
10509 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10510 	.minor_version = 1,
10511 	.init_caps = NFS_CAP_READDIRPLUS
10512 		| NFS_CAP_ATOMIC_OPEN
10513 		| NFS_CAP_DIR_DELEG
10514 		| NFS_CAP_POSIX_LOCK
10515 		| NFS_CAP_STATEID_NFSV41
10516 		| NFS_CAP_ATOMIC_OPEN_V1
10517 		| NFS_CAP_LGOPEN
10518 		| NFS_CAP_MOVEABLE,
10519 	.init_client = nfs41_init_client,
10520 	.shutdown_client = nfs41_shutdown_client,
10521 	.match_stateid = nfs41_match_stateid,
10522 	.find_root_sec = nfs41_find_root_sec,
10523 	.free_lock_state = nfs41_free_lock_state,
10524 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10525 	.alloc_seqid = nfs_alloc_no_seqid,
10526 	.session_trunk = nfs4_test_session_trunk,
10527 	.call_sync_ops = &nfs41_call_sync_ops,
10528 	.sequence_slot_ops = &nfs41_sequence_slot_ops,
10529 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10530 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10531 	.state_renewal_ops = &nfs41_state_renewal_ops,
10532 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10533 };
10534 
10535 #if defined(CONFIG_NFS_V4_2)
10536 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10537 	.minor_version = 2,
10538 	.init_caps = NFS_CAP_READDIRPLUS
10539 		| NFS_CAP_ATOMIC_OPEN
10540 		| NFS_CAP_DIR_DELEG
10541 		| NFS_CAP_POSIX_LOCK
10542 		| NFS_CAP_STATEID_NFSV41
10543 		| NFS_CAP_ATOMIC_OPEN_V1
10544 		| NFS_CAP_LGOPEN
10545 		| NFS_CAP_ALLOCATE
10546 		| NFS_CAP_COPY
10547 		| NFS_CAP_OFFLOAD_CANCEL
10548 		| NFS_CAP_COPY_NOTIFY
10549 		| NFS_CAP_DEALLOCATE
10550 		| NFS_CAP_ZERO_RANGE
10551 		| NFS_CAP_SEEK
10552 		| NFS_CAP_LAYOUTSTATS
10553 		| NFS_CAP_CLONE
10554 		| NFS_CAP_LAYOUTERROR
10555 		| NFS_CAP_READ_PLUS
10556 		| NFS_CAP_MOVEABLE
10557 		| NFS_CAP_OFFLOAD_STATUS,
10558 	.init_client = nfs41_init_client,
10559 	.shutdown_client = nfs41_shutdown_client,
10560 	.match_stateid = nfs41_match_stateid,
10561 	.find_root_sec = nfs41_find_root_sec,
10562 	.free_lock_state = nfs41_free_lock_state,
10563 	.call_sync_ops = &nfs41_call_sync_ops,
10564 	.sequence_slot_ops = &nfs41_sequence_slot_ops,
10565 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10566 	.alloc_seqid = nfs_alloc_no_seqid,
10567 	.session_trunk = nfs4_test_session_trunk,
10568 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10569 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10570 	.state_renewal_ops = &nfs41_state_renewal_ops,
10571 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10572 };
10573 #endif
10574 
10575 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10576 #if defined(CONFIG_NFS_V4_0)
10577 	[0] = &nfs_v4_0_minor_ops,
10578 #endif /* CONFIG_NFS_V4_0 */
10579 	[1] = &nfs_v4_1_minor_ops,
10580 #if defined(CONFIG_NFS_V4_2)
10581 	[2] = &nfs_v4_2_minor_ops,
10582 #endif
10583 };
10584 
10585 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10586 {
10587 	ssize_t error, error2, error3;
10588 	size_t left = size;
10589 
10590 	error = generic_listxattr(dentry, list, left);
10591 	if (error < 0)
10592 		return error;
10593 	if (list) {
10594 		list += error;
10595 		left -= error;
10596 	}
10597 
10598 	error2 = security_inode_listsecurity(d_inode(dentry), &list, &left);
10599 	if (error2 < 0)
10600 		return error2;
10601 	error2 = size - error - left;
10602 
10603 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, left);
10604 	if (error3 < 0)
10605 		return error3;
10606 
10607 	error += error2 + error3;
10608 	if (size && error > size)
10609 		return -ERANGE;
10610 	return error;
10611 }
10612 
10613 static void nfs4_enable_swap(struct inode *inode)
10614 {
10615 	/* The state manager thread must always be running.
10616 	 * It will notice the client is a swapper, and stay put.
10617 	 */
10618 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10619 
10620 	nfs4_schedule_state_manager(clp);
10621 }
10622 
10623 static void nfs4_disable_swap(struct inode *inode)
10624 {
10625 	/* The state manager thread will now exit once it is
10626 	 * woken.
10627 	 */
10628 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10629 
10630 	set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10631 	clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10632 	wake_up_var(&clp->cl_state);
10633 }
10634 
10635 static const struct inode_operations nfs4_dir_inode_operations = {
10636 	.create		= nfs_create,
10637 	.lookup		= nfs_lookup,
10638 	.atomic_open	= nfs_atomic_open,
10639 	.link		= nfs_link,
10640 	.unlink		= nfs_unlink,
10641 	.symlink	= nfs_symlink,
10642 	.mkdir		= nfs_mkdir,
10643 	.rmdir		= nfs_rmdir,
10644 	.mknod		= nfs_mknod,
10645 	.rename		= nfs_rename,
10646 	.permission	= nfs_permission,
10647 	.getattr	= nfs_getattr,
10648 	.setattr	= nfs_setattr,
10649 	.listxattr	= nfs4_listxattr,
10650 	.fileattr_get	= nfs_fileattr_get,
10651 };
10652 
10653 static const struct inode_operations nfs4_file_inode_operations = {
10654 	.permission	= nfs_permission,
10655 	.getattr	= nfs_getattr,
10656 	.setattr	= nfs_setattr,
10657 	.listxattr	= nfs4_listxattr,
10658 	.fileattr_get	= nfs_fileattr_get,
10659 };
10660 
10661 static struct nfs_server *nfs4_clone_server(struct nfs_server *source,
10662 		struct nfs_fh *fh, struct nfs_fattr *fattr,
10663 		rpc_authflavor_t flavor)
10664 {
10665 	struct nfs_server *server;
10666 	int error;
10667 
10668 	server = nfs_clone_server(source, fh, fattr, flavor);
10669 	if (IS_ERR(server))
10670 		return server;
10671 
10672 	nfs4_session_limit_rwsize(server);
10673 	nfs4_session_limit_xasize(server);
10674 
10675 	error = nfs4_delegation_hash_alloc(server);
10676 	if (error) {
10677 		nfs_free_server(server);
10678 		return ERR_PTR(error);
10679 	}
10680 
10681 	return server;
10682 }
10683 
10684 const struct nfs_rpc_ops nfs_v4_clientops = {
10685 	.version	= 4,			/* protocol version */
10686 	.dentry_ops	= &nfs4_dentry_operations,
10687 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10688 	.file_inode_ops	= &nfs4_file_inode_operations,
10689 	.file_ops	= &nfs4_file_operations,
10690 	.getroot	= nfs4_proc_get_root,
10691 	.submount	= nfs4_submount,
10692 	.try_get_tree	= nfs4_try_get_tree,
10693 	.getattr	= nfs4_proc_getattr,
10694 	.setattr	= nfs4_proc_setattr,
10695 	.lookup		= nfs4_proc_lookup,
10696 	.lookupp	= nfs4_proc_lookupp,
10697 	.access		= nfs4_proc_access,
10698 	.readlink	= nfs4_proc_readlink,
10699 	.create		= nfs4_proc_create,
10700 	.remove		= nfs4_proc_remove,
10701 	.unlink_setup	= nfs4_proc_unlink_setup,
10702 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10703 	.unlink_done	= nfs4_proc_unlink_done,
10704 	.rename_setup	= nfs4_proc_rename_setup,
10705 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10706 	.rename_done	= nfs4_proc_rename_done,
10707 	.link		= nfs4_proc_link,
10708 	.symlink	= nfs4_proc_symlink,
10709 	.mkdir		= nfs4_proc_mkdir,
10710 	.rmdir		= nfs4_proc_rmdir,
10711 	.readdir	= nfs4_proc_readdir,
10712 	.mknod		= nfs4_proc_mknod,
10713 	.statfs		= nfs4_proc_statfs,
10714 	.fsinfo		= nfs4_proc_fsinfo,
10715 	.pathconf	= nfs4_proc_pathconf,
10716 	.set_capabilities = nfs4_server_capabilities,
10717 	.decode_dirent	= nfs4_decode_dirent,
10718 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10719 	.read_setup	= nfs4_proc_read_setup,
10720 	.read_done	= nfs4_read_done,
10721 	.write_setup	= nfs4_proc_write_setup,
10722 	.write_done	= nfs4_write_done,
10723 	.commit_setup	= nfs4_proc_commit_setup,
10724 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10725 	.commit_done	= nfs4_commit_done,
10726 	.lock		= nfs4_proc_lock,
10727 	.clear_acl_cache = nfs4_zap_acl_attr,
10728 	.close_context  = nfs4_close_context,
10729 	.open_context	= nfs4_atomic_open,
10730 	.have_delegation = nfs4_have_delegation,
10731 	.return_delegation = nfs4_inode_return_delegation,
10732 	.alloc_client	= nfs4_alloc_client,
10733 	.init_client	= nfs4_init_client,
10734 	.free_client	= nfs4_free_client,
10735 	.create_server	= nfs4_create_server,
10736 	.clone_server	= nfs4_clone_server,
10737 	.discover_trunking = nfs4_discover_trunking,
10738 	.enable_swap	= nfs4_enable_swap,
10739 	.disable_swap	= nfs4_disable_swap,
10740 };
10741 
10742 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10743 	.name	= XATTR_NAME_NFSV4_ACL,
10744 	.list	= nfs4_xattr_list_nfs4_acl,
10745 	.get	= nfs4_xattr_get_nfs4_acl,
10746 	.set	= nfs4_xattr_set_nfs4_acl,
10747 };
10748 
10749 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10750 	.name	= XATTR_NAME_NFSV4_DACL,
10751 	.list	= nfs4_xattr_list_nfs4_dacl,
10752 	.get	= nfs4_xattr_get_nfs4_dacl,
10753 	.set	= nfs4_xattr_set_nfs4_dacl,
10754 };
10755 
10756 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10757 	.name	= XATTR_NAME_NFSV4_SACL,
10758 	.list	= nfs4_xattr_list_nfs4_sacl,
10759 	.get	= nfs4_xattr_get_nfs4_sacl,
10760 	.set	= nfs4_xattr_set_nfs4_sacl,
10761 };
10762 
10763 #ifdef CONFIG_NFS_V4_2
10764 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10765 	.prefix	= XATTR_USER_PREFIX,
10766 	.get	= nfs4_xattr_get_nfs4_user,
10767 	.set	= nfs4_xattr_set_nfs4_user,
10768 };
10769 #endif
10770 
10771 const struct xattr_handler * const nfs4_xattr_handlers[] = {
10772 	&nfs4_xattr_nfs4_acl_handler,
10773 	&nfs4_xattr_nfs4_dacl_handler,
10774 	&nfs4_xattr_nfs4_sacl_handler,
10775 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10776 	&nfs4_xattr_nfs4_label_handler,
10777 #endif
10778 #ifdef CONFIG_NFS_V4_2
10779 	&nfs4_xattr_nfs4_user_handler,
10780 #endif
10781 	NULL
10782 };
10783