xref: /linux/fs/nfs/nfs4proc.c (revision 7354eb7f1558466e92e926802d36e69e42938ea9)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "sysfs.h"
67 #include "nfs4idmap.h"
68 #include "nfs4session.h"
69 #include "fscache.h"
70 #include "nfs42.h"
71 
72 #include "nfs4trace.h"
73 
74 #define NFSDBG_FACILITY		NFSDBG_PROC
75 
76 #define NFS4_BITMASK_SZ		3
77 
78 #define NFS4_POLL_RETRY_MIN	(HZ/10)
79 #define NFS4_POLL_RETRY_MAX	(15*HZ)
80 
81 /* file attributes which can be mapped to nfs attributes */
82 #define NFS4_VALID_ATTRS (ATTR_MODE \
83 	| ATTR_UID \
84 	| ATTR_GID \
85 	| ATTR_SIZE \
86 	| ATTR_ATIME \
87 	| ATTR_MTIME \
88 	| ATTR_CTIME \
89 	| ATTR_ATIME_SET \
90 	| ATTR_MTIME_SET)
91 
92 struct nfs4_opendata;
93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
97 			      struct nfs_fattr *fattr, struct inode *inode);
98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
99 			    struct nfs_fattr *fattr, struct iattr *sattr,
100 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel);
101 #ifdef CONFIG_NFS_V4_1
102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
103 		const struct cred *cred,
104 		struct nfs4_slot *slot,
105 		bool is_privileged);
106 static int nfs41_test_stateid(struct nfs_server *, const nfs4_stateid *,
107 			      const struct cred *);
108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
109 			      const struct cred *, bool);
110 #endif
111 
112 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
113 static inline struct nfs4_label *
114 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
115 	struct iattr *sattr, struct nfs4_label *label)
116 {
117 	int err;
118 
119 	if (label == NULL)
120 		return NULL;
121 
122 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
123 		return NULL;
124 
125 	label->lfs = 0;
126 	label->pi = 0;
127 	label->len = 0;
128 	label->label = NULL;
129 
130 	err = security_dentry_init_security(dentry, sattr->ia_mode,
131 				&dentry->d_name, NULL,
132 				(void **)&label->label, &label->len);
133 	if (err == 0)
134 		return label;
135 
136 	return NULL;
137 }
138 static inline void
139 nfs4_label_release_security(struct nfs4_label *label)
140 {
141 	if (label)
142 		security_release_secctx(label->label, label->len);
143 }
144 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
145 {
146 	if (label)
147 		return server->attr_bitmask;
148 
149 	return server->attr_bitmask_nl;
150 }
151 #else
152 static inline struct nfs4_label *
153 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
154 	struct iattr *sattr, struct nfs4_label *l)
155 { return NULL; }
156 static inline void
157 nfs4_label_release_security(struct nfs4_label *label)
158 { return; }
159 static inline u32 *
160 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
161 { return server->attr_bitmask; }
162 #endif
163 
164 /* Prevent leaks of NFSv4 errors into userland */
165 static int nfs4_map_errors(int err)
166 {
167 	if (err >= -1000)
168 		return err;
169 	switch (err) {
170 	case -NFS4ERR_RESOURCE:
171 	case -NFS4ERR_LAYOUTTRYLATER:
172 	case -NFS4ERR_RECALLCONFLICT:
173 	case -NFS4ERR_RETURNCONFLICT:
174 		return -EREMOTEIO;
175 	case -NFS4ERR_WRONGSEC:
176 	case -NFS4ERR_WRONG_CRED:
177 		return -EPERM;
178 	case -NFS4ERR_BADOWNER:
179 	case -NFS4ERR_BADNAME:
180 		return -EINVAL;
181 	case -NFS4ERR_SHARE_DENIED:
182 		return -EACCES;
183 	case -NFS4ERR_MINOR_VERS_MISMATCH:
184 		return -EPROTONOSUPPORT;
185 	case -NFS4ERR_FILE_OPEN:
186 		return -EBUSY;
187 	case -NFS4ERR_NOT_SAME:
188 		return -ENOTSYNC;
189 	default:
190 		dprintk("%s could not handle NFSv4 error %d\n",
191 				__func__, -err);
192 		break;
193 	}
194 	return -EIO;
195 }
196 
197 /*
198  * This is our standard bitmap for GETATTR requests.
199  */
200 const u32 nfs4_fattr_bitmap[3] = {
201 	FATTR4_WORD0_TYPE
202 	| FATTR4_WORD0_CHANGE
203 	| FATTR4_WORD0_SIZE
204 	| FATTR4_WORD0_FSID
205 	| FATTR4_WORD0_FILEID,
206 	FATTR4_WORD1_MODE
207 	| FATTR4_WORD1_NUMLINKS
208 	| FATTR4_WORD1_OWNER
209 	| FATTR4_WORD1_OWNER_GROUP
210 	| FATTR4_WORD1_RAWDEV
211 	| FATTR4_WORD1_SPACE_USED
212 	| FATTR4_WORD1_TIME_ACCESS
213 	| FATTR4_WORD1_TIME_METADATA
214 	| FATTR4_WORD1_TIME_MODIFY
215 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
216 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
217 	FATTR4_WORD2_SECURITY_LABEL
218 #endif
219 };
220 
221 static const u32 nfs4_pnfs_open_bitmap[3] = {
222 	FATTR4_WORD0_TYPE
223 	| FATTR4_WORD0_CHANGE
224 	| FATTR4_WORD0_SIZE
225 	| FATTR4_WORD0_FSID
226 	| FATTR4_WORD0_FILEID,
227 	FATTR4_WORD1_MODE
228 	| FATTR4_WORD1_NUMLINKS
229 	| FATTR4_WORD1_OWNER
230 	| FATTR4_WORD1_OWNER_GROUP
231 	| FATTR4_WORD1_RAWDEV
232 	| FATTR4_WORD1_SPACE_USED
233 	| FATTR4_WORD1_TIME_ACCESS
234 	| FATTR4_WORD1_TIME_METADATA
235 	| FATTR4_WORD1_TIME_MODIFY,
236 	FATTR4_WORD2_MDSTHRESHOLD
237 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
238 	| FATTR4_WORD2_SECURITY_LABEL
239 #endif
240 };
241 
242 static const u32 nfs4_open_noattr_bitmap[3] = {
243 	FATTR4_WORD0_TYPE
244 	| FATTR4_WORD0_FILEID,
245 };
246 
247 const u32 nfs4_statfs_bitmap[3] = {
248 	FATTR4_WORD0_FILES_AVAIL
249 	| FATTR4_WORD0_FILES_FREE
250 	| FATTR4_WORD0_FILES_TOTAL,
251 	FATTR4_WORD1_SPACE_AVAIL
252 	| FATTR4_WORD1_SPACE_FREE
253 	| FATTR4_WORD1_SPACE_TOTAL
254 };
255 
256 const u32 nfs4_pathconf_bitmap[3] = {
257 	FATTR4_WORD0_MAXLINK
258 	| FATTR4_WORD0_MAXNAME,
259 	0
260 };
261 
262 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
263 			| FATTR4_WORD0_MAXREAD
264 			| FATTR4_WORD0_MAXWRITE
265 			| FATTR4_WORD0_LEASE_TIME,
266 			FATTR4_WORD1_TIME_DELTA
267 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
268 			FATTR4_WORD2_LAYOUT_BLKSIZE
269 			| FATTR4_WORD2_CLONE_BLKSIZE
270 			| FATTR4_WORD2_CHANGE_ATTR_TYPE
271 			| FATTR4_WORD2_XATTR_SUPPORT
272 };
273 
274 const u32 nfs4_fs_locations_bitmap[3] = {
275 	FATTR4_WORD0_CHANGE
276 	| FATTR4_WORD0_SIZE
277 	| FATTR4_WORD0_FSID
278 	| FATTR4_WORD0_FILEID
279 	| FATTR4_WORD0_FS_LOCATIONS,
280 	FATTR4_WORD1_OWNER
281 	| FATTR4_WORD1_OWNER_GROUP
282 	| FATTR4_WORD1_RAWDEV
283 	| FATTR4_WORD1_SPACE_USED
284 	| FATTR4_WORD1_TIME_ACCESS
285 	| FATTR4_WORD1_TIME_METADATA
286 	| FATTR4_WORD1_TIME_MODIFY
287 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
288 };
289 
290 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
291 				    struct inode *inode, unsigned long flags)
292 {
293 	unsigned long cache_validity;
294 
295 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
296 	if (!inode || !nfs_have_read_or_write_delegation(inode))
297 		return;
298 
299 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
300 
301 	/* Remove the attributes over which we have full control */
302 	dst[1] &= ~FATTR4_WORD1_RAWDEV;
303 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
304 		dst[0] &= ~FATTR4_WORD0_SIZE;
305 
306 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
307 		dst[0] &= ~FATTR4_WORD0_CHANGE;
308 
309 	if (!(cache_validity & NFS_INO_INVALID_MODE))
310 		dst[1] &= ~FATTR4_WORD1_MODE;
311 	if (!(cache_validity & NFS_INO_INVALID_OTHER))
312 		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
313 
314 	if (nfs_have_delegated_mtime(inode)) {
315 		if (!(cache_validity & NFS_INO_INVALID_ATIME))
316 			dst[1] &= ~FATTR4_WORD1_TIME_ACCESS;
317 		if (!(cache_validity & NFS_INO_INVALID_MTIME))
318 			dst[1] &= ~FATTR4_WORD1_TIME_MODIFY;
319 		if (!(cache_validity & NFS_INO_INVALID_CTIME))
320 			dst[1] &= ~FATTR4_WORD1_TIME_METADATA;
321 	} else if (nfs_have_delegated_atime(inode)) {
322 		if (!(cache_validity & NFS_INO_INVALID_ATIME))
323 			dst[1] &= ~FATTR4_WORD1_TIME_ACCESS;
324 	}
325 }
326 
327 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
328 		struct nfs4_readdir_arg *readdir)
329 {
330 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
331 	__be32 *start, *p;
332 
333 	if (cookie > 2) {
334 		readdir->cookie = cookie;
335 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
336 		return;
337 	}
338 
339 	readdir->cookie = 0;
340 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
341 	if (cookie == 2)
342 		return;
343 
344 	/*
345 	 * NFSv4 servers do not return entries for '.' and '..'
346 	 * Therefore, we fake these entries here.  We let '.'
347 	 * have cookie 0 and '..' have cookie 1.  Note that
348 	 * when talking to the server, we always send cookie 0
349 	 * instead of 1 or 2.
350 	 */
351 	start = p = kmap_atomic(*readdir->pages);
352 
353 	if (cookie == 0) {
354 		*p++ = xdr_one;                                  /* next */
355 		*p++ = xdr_zero;                   /* cookie, first word */
356 		*p++ = xdr_one;                   /* cookie, second word */
357 		*p++ = xdr_one;                             /* entry len */
358 		memcpy(p, ".\0\0\0", 4);                        /* entry */
359 		p++;
360 		*p++ = xdr_one;                         /* bitmap length */
361 		*p++ = htonl(attrs);                           /* bitmap */
362 		*p++ = htonl(12);             /* attribute buffer length */
363 		*p++ = htonl(NF4DIR);
364 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
365 	}
366 
367 	*p++ = xdr_one;                                  /* next */
368 	*p++ = xdr_zero;                   /* cookie, first word */
369 	*p++ = xdr_two;                   /* cookie, second word */
370 	*p++ = xdr_two;                             /* entry len */
371 	memcpy(p, "..\0\0", 4);                         /* entry */
372 	p++;
373 	*p++ = xdr_one;                         /* bitmap length */
374 	*p++ = htonl(attrs);                           /* bitmap */
375 	*p++ = htonl(12);             /* attribute buffer length */
376 	*p++ = htonl(NF4DIR);
377 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
378 
379 	readdir->pgbase = (char *)p - (char *)start;
380 	readdir->count -= readdir->pgbase;
381 	kunmap_atomic(start);
382 }
383 
384 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
385 {
386 	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
387 		fattr->pre_change_attr = version;
388 		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
389 	}
390 }
391 
392 static void nfs4_test_and_free_stateid(struct nfs_server *server,
393 		nfs4_stateid *stateid,
394 		const struct cred *cred)
395 {
396 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
397 
398 	ops->test_and_free_expired(server, stateid, cred);
399 }
400 
401 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
402 		nfs4_stateid *stateid,
403 		const struct cred *cred)
404 {
405 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
406 	nfs4_test_and_free_stateid(server, stateid, cred);
407 }
408 
409 static void nfs4_free_revoked_stateid(struct nfs_server *server,
410 		const nfs4_stateid *stateid,
411 		const struct cred *cred)
412 {
413 	nfs4_stateid tmp;
414 
415 	nfs4_stateid_copy(&tmp, stateid);
416 	__nfs4_free_revoked_stateid(server, &tmp, cred);
417 }
418 
419 static long nfs4_update_delay(long *timeout)
420 {
421 	long ret;
422 	if (!timeout)
423 		return NFS4_POLL_RETRY_MAX;
424 	if (*timeout <= 0)
425 		*timeout = NFS4_POLL_RETRY_MIN;
426 	if (*timeout > NFS4_POLL_RETRY_MAX)
427 		*timeout = NFS4_POLL_RETRY_MAX;
428 	ret = *timeout;
429 	*timeout <<= 1;
430 	return ret;
431 }
432 
433 static int nfs4_delay_killable(long *timeout)
434 {
435 	might_sleep();
436 
437 	__set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
438 	schedule_timeout(nfs4_update_delay(timeout));
439 	if (!__fatal_signal_pending(current))
440 		return 0;
441 	return -EINTR;
442 }
443 
444 static int nfs4_delay_interruptible(long *timeout)
445 {
446 	might_sleep();
447 
448 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
449 	schedule_timeout(nfs4_update_delay(timeout));
450 	if (!signal_pending(current))
451 		return 0;
452 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
453 }
454 
455 static int nfs4_delay(long *timeout, bool interruptible)
456 {
457 	if (interruptible)
458 		return nfs4_delay_interruptible(timeout);
459 	return nfs4_delay_killable(timeout);
460 }
461 
462 static const nfs4_stateid *
463 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
464 {
465 	if (!stateid)
466 		return NULL;
467 	switch (stateid->type) {
468 	case NFS4_OPEN_STATEID_TYPE:
469 	case NFS4_LOCK_STATEID_TYPE:
470 	case NFS4_DELEGATION_STATEID_TYPE:
471 		return stateid;
472 	default:
473 		break;
474 	}
475 	return NULL;
476 }
477 
478 /* This is the error handling routine for processes that are allowed
479  * to sleep.
480  */
481 static int nfs4_do_handle_exception(struct nfs_server *server,
482 		int errorcode, struct nfs4_exception *exception)
483 {
484 	struct nfs_client *clp = server->nfs_client;
485 	struct nfs4_state *state = exception->state;
486 	const nfs4_stateid *stateid;
487 	struct inode *inode = exception->inode;
488 	int ret = errorcode;
489 
490 	exception->delay = 0;
491 	exception->recovering = 0;
492 	exception->retry = 0;
493 
494 	stateid = nfs4_recoverable_stateid(exception->stateid);
495 	if (stateid == NULL && state != NULL)
496 		stateid = nfs4_recoverable_stateid(&state->stateid);
497 
498 	switch(errorcode) {
499 		case 0:
500 			return 0;
501 		case -NFS4ERR_BADHANDLE:
502 		case -ESTALE:
503 			if (inode != NULL && S_ISREG(inode->i_mode))
504 				pnfs_destroy_layout(NFS_I(inode));
505 			break;
506 		case -NFS4ERR_DELEG_REVOKED:
507 		case -NFS4ERR_ADMIN_REVOKED:
508 		case -NFS4ERR_EXPIRED:
509 		case -NFS4ERR_BAD_STATEID:
510 		case -NFS4ERR_PARTNER_NO_AUTH:
511 			if (inode != NULL && stateid != NULL) {
512 				nfs_inode_find_state_and_recover(inode,
513 						stateid);
514 				goto wait_on_recovery;
515 			}
516 			fallthrough;
517 		case -NFS4ERR_OPENMODE:
518 			if (inode) {
519 				int err;
520 
521 				err = nfs_async_inode_return_delegation(inode,
522 						stateid);
523 				if (err == 0)
524 					goto wait_on_recovery;
525 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
526 					exception->retry = 1;
527 					break;
528 				}
529 			}
530 			if (state == NULL)
531 				break;
532 			ret = nfs4_schedule_stateid_recovery(server, state);
533 			if (ret < 0)
534 				break;
535 			goto wait_on_recovery;
536 		case -NFS4ERR_STALE_STATEID:
537 		case -NFS4ERR_STALE_CLIENTID:
538 			nfs4_schedule_lease_recovery(clp);
539 			goto wait_on_recovery;
540 		case -NFS4ERR_MOVED:
541 			ret = nfs4_schedule_migration_recovery(server);
542 			if (ret < 0)
543 				break;
544 			goto wait_on_recovery;
545 		case -NFS4ERR_LEASE_MOVED:
546 			nfs4_schedule_lease_moved_recovery(clp);
547 			goto wait_on_recovery;
548 #if defined(CONFIG_NFS_V4_1)
549 		case -NFS4ERR_BADSESSION:
550 		case -NFS4ERR_BADSLOT:
551 		case -NFS4ERR_BAD_HIGH_SLOT:
552 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
553 		case -NFS4ERR_DEADSESSION:
554 		case -NFS4ERR_SEQ_FALSE_RETRY:
555 		case -NFS4ERR_SEQ_MISORDERED:
556 			/* Handled in nfs41_sequence_process() */
557 			goto wait_on_recovery;
558 #endif /* defined(CONFIG_NFS_V4_1) */
559 		case -NFS4ERR_FILE_OPEN:
560 			if (exception->timeout > HZ) {
561 				/* We have retried a decent amount, time to
562 				 * fail
563 				 */
564 				ret = -EBUSY;
565 				break;
566 			}
567 			fallthrough;
568 		case -NFS4ERR_DELAY:
569 			nfs_inc_server_stats(server, NFSIOS_DELAY);
570 			fallthrough;
571 		case -NFS4ERR_GRACE:
572 		case -NFS4ERR_LAYOUTTRYLATER:
573 		case -NFS4ERR_RECALLCONFLICT:
574 		case -NFS4ERR_RETURNCONFLICT:
575 			exception->delay = 1;
576 			return 0;
577 
578 		case -NFS4ERR_RETRY_UNCACHED_REP:
579 		case -NFS4ERR_OLD_STATEID:
580 			exception->retry = 1;
581 			break;
582 		case -NFS4ERR_BADOWNER:
583 			/* The following works around a Linux server bug! */
584 		case -NFS4ERR_BADNAME:
585 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
586 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
587 				exception->retry = 1;
588 				printk(KERN_WARNING "NFS: v4 server %s "
589 						"does not accept raw "
590 						"uid/gids. "
591 						"Reenabling the idmapper.\n",
592 						server->nfs_client->cl_hostname);
593 			}
594 	}
595 	/* We failed to handle the error */
596 	return nfs4_map_errors(ret);
597 wait_on_recovery:
598 	exception->recovering = 1;
599 	return 0;
600 }
601 
602 /*
603  * Track the number of NFS4ERR_DELAY related retransmissions and return
604  * EAGAIN if the 'softerr' mount option is set, and we've exceeded the limit
605  * set by 'nfs_delay_retrans'.
606  */
607 static int nfs4_exception_should_retrans(const struct nfs_server *server,
608 					 struct nfs4_exception *exception)
609 {
610 	if (server->flags & NFS_MOUNT_SOFTERR && nfs_delay_retrans >= 0) {
611 		if (exception->retrans++ >= (unsigned short)nfs_delay_retrans)
612 			return -EAGAIN;
613 	}
614 	return 0;
615 }
616 
617 /* This is the error handling routine for processes that are allowed
618  * to sleep.
619  */
620 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
621 {
622 	struct nfs_client *clp = server->nfs_client;
623 	int ret;
624 
625 	ret = nfs4_do_handle_exception(server, errorcode, exception);
626 	if (exception->delay) {
627 		int ret2 = nfs4_exception_should_retrans(server, exception);
628 		if (ret2 < 0) {
629 			exception->retry = 0;
630 			return ret2;
631 		}
632 		ret = nfs4_delay(&exception->timeout,
633 				exception->interruptible);
634 		goto out_retry;
635 	}
636 	if (exception->recovering) {
637 		if (exception->task_is_privileged)
638 			return -EDEADLOCK;
639 		ret = nfs4_wait_clnt_recover(clp);
640 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
641 			return -EIO;
642 		goto out_retry;
643 	}
644 	return ret;
645 out_retry:
646 	if (ret == 0)
647 		exception->retry = 1;
648 	return ret;
649 }
650 
651 static int
652 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
653 		int errorcode, struct nfs4_exception *exception)
654 {
655 	struct nfs_client *clp = server->nfs_client;
656 	int ret;
657 
658 	ret = nfs4_do_handle_exception(server, errorcode, exception);
659 	if (exception->delay) {
660 		int ret2 = nfs4_exception_should_retrans(server, exception);
661 		if (ret2 < 0) {
662 			exception->retry = 0;
663 			return ret2;
664 		}
665 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
666 		goto out_retry;
667 	}
668 	if (exception->recovering) {
669 		if (exception->task_is_privileged)
670 			return -EDEADLOCK;
671 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
672 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
673 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
674 		goto out_retry;
675 	}
676 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
677 		ret = -EIO;
678 	return ret;
679 out_retry:
680 	if (ret == 0) {
681 		exception->retry = 1;
682 		/*
683 		 * For NFS4ERR_MOVED, the client transport will need to
684 		 * be recomputed after migration recovery has completed.
685 		 */
686 		if (errorcode == -NFS4ERR_MOVED)
687 			rpc_task_release_transport(task);
688 	}
689 	return ret;
690 }
691 
692 int
693 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
694 			struct nfs4_state *state, long *timeout)
695 {
696 	struct nfs4_exception exception = {
697 		.state = state,
698 	};
699 
700 	if (task->tk_status >= 0)
701 		return 0;
702 	if (timeout)
703 		exception.timeout = *timeout;
704 	task->tk_status = nfs4_async_handle_exception(task, server,
705 			task->tk_status,
706 			&exception);
707 	if (exception.delay && timeout)
708 		*timeout = exception.timeout;
709 	if (exception.retry)
710 		return -EAGAIN;
711 	return 0;
712 }
713 
714 /*
715  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
716  * or 'false' otherwise.
717  */
718 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
719 {
720 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
721 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
722 }
723 
724 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
725 {
726 	spin_lock(&clp->cl_lock);
727 	if (time_before(clp->cl_last_renewal,timestamp))
728 		clp->cl_last_renewal = timestamp;
729 	spin_unlock(&clp->cl_lock);
730 }
731 
732 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
733 {
734 	struct nfs_client *clp = server->nfs_client;
735 
736 	if (!nfs4_has_session(clp))
737 		do_renew_lease(clp, timestamp);
738 }
739 
740 struct nfs4_call_sync_data {
741 	const struct nfs_server *seq_server;
742 	struct nfs4_sequence_args *seq_args;
743 	struct nfs4_sequence_res *seq_res;
744 };
745 
746 void nfs4_init_sequence(struct nfs4_sequence_args *args,
747 			struct nfs4_sequence_res *res, int cache_reply,
748 			int privileged)
749 {
750 	args->sa_slot = NULL;
751 	args->sa_cache_this = cache_reply;
752 	args->sa_privileged = privileged;
753 
754 	res->sr_slot = NULL;
755 }
756 
757 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
758 {
759 	struct nfs4_slot *slot = res->sr_slot;
760 	struct nfs4_slot_table *tbl;
761 
762 	tbl = slot->table;
763 	spin_lock(&tbl->slot_tbl_lock);
764 	if (!nfs41_wake_and_assign_slot(tbl, slot))
765 		nfs4_free_slot(tbl, slot);
766 	spin_unlock(&tbl->slot_tbl_lock);
767 
768 	res->sr_slot = NULL;
769 }
770 
771 static int nfs40_sequence_done(struct rpc_task *task,
772 			       struct nfs4_sequence_res *res)
773 {
774 	if (res->sr_slot != NULL)
775 		nfs40_sequence_free_slot(res);
776 	return 1;
777 }
778 
779 #if defined(CONFIG_NFS_V4_1)
780 
781 static void nfs41_release_slot(struct nfs4_slot *slot)
782 {
783 	struct nfs4_session *session;
784 	struct nfs4_slot_table *tbl;
785 	bool send_new_highest_used_slotid = false;
786 
787 	if (!slot)
788 		return;
789 	tbl = slot->table;
790 	session = tbl->session;
791 
792 	/* Bump the slot sequence number */
793 	if (slot->seq_done)
794 		slot->seq_nr++;
795 	slot->seq_done = 0;
796 
797 	spin_lock(&tbl->slot_tbl_lock);
798 	/* Be nice to the server: try to ensure that the last transmitted
799 	 * value for highest_user_slotid <= target_highest_slotid
800 	 */
801 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
802 		send_new_highest_used_slotid = true;
803 
804 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
805 		send_new_highest_used_slotid = false;
806 		goto out_unlock;
807 	}
808 	nfs4_free_slot(tbl, slot);
809 
810 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
811 		send_new_highest_used_slotid = false;
812 out_unlock:
813 	spin_unlock(&tbl->slot_tbl_lock);
814 	if (send_new_highest_used_slotid)
815 		nfs41_notify_server(session->clp);
816 	if (waitqueue_active(&tbl->slot_waitq))
817 		wake_up_all(&tbl->slot_waitq);
818 }
819 
820 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
821 {
822 	nfs41_release_slot(res->sr_slot);
823 	res->sr_slot = NULL;
824 }
825 
826 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
827 		u32 seqnr)
828 {
829 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
830 		slot->seq_nr_highest_sent = seqnr;
831 }
832 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
833 {
834 	nfs4_slot_sequence_record_sent(slot, seqnr);
835 	slot->seq_nr_last_acked = seqnr;
836 }
837 
838 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
839 				struct nfs4_slot *slot)
840 {
841 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
842 	if (!IS_ERR(task))
843 		rpc_put_task_async(task);
844 }
845 
846 static int nfs41_sequence_process(struct rpc_task *task,
847 		struct nfs4_sequence_res *res)
848 {
849 	struct nfs4_session *session;
850 	struct nfs4_slot *slot = res->sr_slot;
851 	struct nfs_client *clp;
852 	int status;
853 	int ret = 1;
854 
855 	if (slot == NULL)
856 		goto out_noaction;
857 	/* don't increment the sequence number if the task wasn't sent */
858 	if (!RPC_WAS_SENT(task) || slot->seq_done)
859 		goto out;
860 
861 	session = slot->table->session;
862 	clp = session->clp;
863 
864 	trace_nfs4_sequence_done(session, res);
865 
866 	status = res->sr_status;
867 	if (task->tk_status == -NFS4ERR_DEADSESSION)
868 		status = -NFS4ERR_DEADSESSION;
869 
870 	/* Check the SEQUENCE operation status */
871 	switch (status) {
872 	case 0:
873 		/* Mark this sequence number as having been acked */
874 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
875 		/* Update the slot's sequence and clientid lease timer */
876 		slot->seq_done = 1;
877 		do_renew_lease(clp, res->sr_timestamp);
878 		/* Check sequence flags */
879 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
880 				!!slot->privileged);
881 		nfs41_update_target_slotid(slot->table, slot, res);
882 		break;
883 	case 1:
884 		/*
885 		 * sr_status remains 1 if an RPC level error occurred.
886 		 * The server may or may not have processed the sequence
887 		 * operation..
888 		 */
889 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
890 		slot->seq_done = 1;
891 		goto out;
892 	case -NFS4ERR_DELAY:
893 		/* The server detected a resend of the RPC call and
894 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
895 		 * of RFC5661.
896 		 */
897 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
898 			__func__,
899 			slot->slot_nr,
900 			slot->seq_nr);
901 		goto out_retry;
902 	case -NFS4ERR_RETRY_UNCACHED_REP:
903 	case -NFS4ERR_SEQ_FALSE_RETRY:
904 		/*
905 		 * The server thinks we tried to replay a request.
906 		 * Retry the call after bumping the sequence ID.
907 		 */
908 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
909 		goto retry_new_seq;
910 	case -NFS4ERR_BADSLOT:
911 		/*
912 		 * The slot id we used was probably retired. Try again
913 		 * using a different slot id.
914 		 */
915 		if (slot->slot_nr < slot->table->target_highest_slotid)
916 			goto session_recover;
917 		goto retry_nowait;
918 	case -NFS4ERR_SEQ_MISORDERED:
919 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
920 		/*
921 		 * Were one or more calls using this slot interrupted?
922 		 * If the server never received the request, then our
923 		 * transmitted slot sequence number may be too high. However,
924 		 * if the server did receive the request then it might
925 		 * accidentally give us a reply with a mismatched operation.
926 		 * We can sort this out by sending a lone sequence operation
927 		 * to the server on the same slot.
928 		 */
929 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
930 			slot->seq_nr--;
931 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
932 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
933 				res->sr_slot = NULL;
934 			}
935 			goto retry_nowait;
936 		}
937 		/*
938 		 * RFC5661:
939 		 * A retry might be sent while the original request is
940 		 * still in progress on the replier. The replier SHOULD
941 		 * deal with the issue by returning NFS4ERR_DELAY as the
942 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
943 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
944 		 *
945 		 * Restart the search after a delay.
946 		 */
947 		slot->seq_nr = slot->seq_nr_highest_sent;
948 		goto out_retry;
949 	case -NFS4ERR_BADSESSION:
950 	case -NFS4ERR_DEADSESSION:
951 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
952 		goto session_recover;
953 	default:
954 		/* Just update the slot sequence no. */
955 		slot->seq_done = 1;
956 	}
957 out:
958 	/* The session may be reset by one of the error handlers. */
959 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
960 out_noaction:
961 	return ret;
962 session_recover:
963 	set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
964 	nfs4_schedule_session_recovery(session, status);
965 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
966 	nfs41_sequence_free_slot(res);
967 	goto out;
968 retry_new_seq:
969 	++slot->seq_nr;
970 retry_nowait:
971 	if (rpc_restart_call_prepare(task)) {
972 		nfs41_sequence_free_slot(res);
973 		task->tk_status = 0;
974 		ret = 0;
975 	}
976 	goto out;
977 out_retry:
978 	if (!rpc_restart_call(task))
979 		goto out;
980 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
981 	return 0;
982 }
983 
984 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
985 {
986 	if (!nfs41_sequence_process(task, res))
987 		return 0;
988 	if (res->sr_slot != NULL)
989 		nfs41_sequence_free_slot(res);
990 	return 1;
991 
992 }
993 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
994 
995 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
996 {
997 	if (res->sr_slot == NULL)
998 		return 1;
999 	if (res->sr_slot->table->session != NULL)
1000 		return nfs41_sequence_process(task, res);
1001 	return nfs40_sequence_done(task, res);
1002 }
1003 
1004 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1005 {
1006 	if (res->sr_slot != NULL) {
1007 		if (res->sr_slot->table->session != NULL)
1008 			nfs41_sequence_free_slot(res);
1009 		else
1010 			nfs40_sequence_free_slot(res);
1011 	}
1012 }
1013 
1014 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
1015 {
1016 	if (res->sr_slot == NULL)
1017 		return 1;
1018 	if (!res->sr_slot->table->session)
1019 		return nfs40_sequence_done(task, res);
1020 	return nfs41_sequence_done(task, res);
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1023 
1024 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
1025 {
1026 	struct nfs4_call_sync_data *data = calldata;
1027 
1028 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
1029 
1030 	nfs4_setup_sequence(data->seq_server->nfs_client,
1031 			    data->seq_args, data->seq_res, task);
1032 }
1033 
1034 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
1035 {
1036 	struct nfs4_call_sync_data *data = calldata;
1037 
1038 	nfs41_sequence_done(task, data->seq_res);
1039 }
1040 
1041 static const struct rpc_call_ops nfs41_call_sync_ops = {
1042 	.rpc_call_prepare = nfs41_call_sync_prepare,
1043 	.rpc_call_done = nfs41_call_sync_done,
1044 };
1045 
1046 #else	/* !CONFIG_NFS_V4_1 */
1047 
1048 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1049 {
1050 	return nfs40_sequence_done(task, res);
1051 }
1052 
1053 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1054 {
1055 	if (res->sr_slot != NULL)
1056 		nfs40_sequence_free_slot(res);
1057 }
1058 
1059 int nfs4_sequence_done(struct rpc_task *task,
1060 		       struct nfs4_sequence_res *res)
1061 {
1062 	return nfs40_sequence_done(task, res);
1063 }
1064 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1065 
1066 #endif	/* !CONFIG_NFS_V4_1 */
1067 
1068 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1069 {
1070 	res->sr_timestamp = jiffies;
1071 	res->sr_status_flags = 0;
1072 	res->sr_status = 1;
1073 }
1074 
1075 static
1076 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1077 		struct nfs4_sequence_res *res,
1078 		struct nfs4_slot *slot)
1079 {
1080 	if (!slot)
1081 		return;
1082 	slot->privileged = args->sa_privileged ? 1 : 0;
1083 	args->sa_slot = slot;
1084 
1085 	res->sr_slot = slot;
1086 }
1087 
1088 int nfs4_setup_sequence(struct nfs_client *client,
1089 			struct nfs4_sequence_args *args,
1090 			struct nfs4_sequence_res *res,
1091 			struct rpc_task *task)
1092 {
1093 	struct nfs4_session *session = nfs4_get_session(client);
1094 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1095 	struct nfs4_slot *slot;
1096 
1097 	/* slot already allocated? */
1098 	if (res->sr_slot != NULL)
1099 		goto out_start;
1100 
1101 	if (session)
1102 		tbl = &session->fc_slot_table;
1103 
1104 	spin_lock(&tbl->slot_tbl_lock);
1105 	/* The state manager will wait until the slot table is empty */
1106 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1107 		goto out_sleep;
1108 
1109 	slot = nfs4_alloc_slot(tbl);
1110 	if (IS_ERR(slot)) {
1111 		if (slot == ERR_PTR(-ENOMEM))
1112 			goto out_sleep_timeout;
1113 		goto out_sleep;
1114 	}
1115 	spin_unlock(&tbl->slot_tbl_lock);
1116 
1117 	nfs4_sequence_attach_slot(args, res, slot);
1118 
1119 	trace_nfs4_setup_sequence(session, args);
1120 out_start:
1121 	nfs41_sequence_res_init(res);
1122 	rpc_call_start(task);
1123 	return 0;
1124 out_sleep_timeout:
1125 	/* Try again in 1/4 second */
1126 	if (args->sa_privileged)
1127 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1128 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1129 	else
1130 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1131 				NULL, jiffies + (HZ >> 2));
1132 	spin_unlock(&tbl->slot_tbl_lock);
1133 	return -EAGAIN;
1134 out_sleep:
1135 	if (args->sa_privileged)
1136 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1137 				RPC_PRIORITY_PRIVILEGED);
1138 	else
1139 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1140 	spin_unlock(&tbl->slot_tbl_lock);
1141 	return -EAGAIN;
1142 }
1143 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1144 
1145 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1146 {
1147 	struct nfs4_call_sync_data *data = calldata;
1148 	nfs4_setup_sequence(data->seq_server->nfs_client,
1149 				data->seq_args, data->seq_res, task);
1150 }
1151 
1152 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1153 {
1154 	struct nfs4_call_sync_data *data = calldata;
1155 	nfs4_sequence_done(task, data->seq_res);
1156 }
1157 
1158 static const struct rpc_call_ops nfs40_call_sync_ops = {
1159 	.rpc_call_prepare = nfs40_call_sync_prepare,
1160 	.rpc_call_done = nfs40_call_sync_done,
1161 };
1162 
1163 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1164 {
1165 	int ret;
1166 	struct rpc_task *task;
1167 
1168 	task = rpc_run_task(task_setup);
1169 	if (IS_ERR(task))
1170 		return PTR_ERR(task);
1171 
1172 	ret = task->tk_status;
1173 	rpc_put_task(task);
1174 	return ret;
1175 }
1176 
1177 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1178 			     struct nfs_server *server,
1179 			     struct rpc_message *msg,
1180 			     struct nfs4_sequence_args *args,
1181 			     struct nfs4_sequence_res *res,
1182 			     unsigned short task_flags)
1183 {
1184 	struct nfs_client *clp = server->nfs_client;
1185 	struct nfs4_call_sync_data data = {
1186 		.seq_server = server,
1187 		.seq_args = args,
1188 		.seq_res = res,
1189 	};
1190 	struct rpc_task_setup task_setup = {
1191 		.rpc_client = clnt,
1192 		.rpc_message = msg,
1193 		.callback_ops = clp->cl_mvops->call_sync_ops,
1194 		.callback_data = &data,
1195 		.flags = task_flags,
1196 	};
1197 
1198 	return nfs4_call_sync_custom(&task_setup);
1199 }
1200 
1201 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1202 				   struct nfs_server *server,
1203 				   struct rpc_message *msg,
1204 				   struct nfs4_sequence_args *args,
1205 				   struct nfs4_sequence_res *res)
1206 {
1207 	unsigned short task_flags = 0;
1208 
1209 	if (server->caps & NFS_CAP_MOVEABLE)
1210 		task_flags = RPC_TASK_MOVEABLE;
1211 	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1212 }
1213 
1214 
1215 int nfs4_call_sync(struct rpc_clnt *clnt,
1216 		   struct nfs_server *server,
1217 		   struct rpc_message *msg,
1218 		   struct nfs4_sequence_args *args,
1219 		   struct nfs4_sequence_res *res,
1220 		   int cache_reply)
1221 {
1222 	nfs4_init_sequence(args, res, cache_reply, 0);
1223 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1224 }
1225 
1226 static void
1227 nfs4_inc_nlink_locked(struct inode *inode)
1228 {
1229 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1230 					     NFS_INO_INVALID_CTIME |
1231 					     NFS_INO_INVALID_NLINK);
1232 	inc_nlink(inode);
1233 }
1234 
1235 static void
1236 nfs4_inc_nlink(struct inode *inode)
1237 {
1238 	spin_lock(&inode->i_lock);
1239 	nfs4_inc_nlink_locked(inode);
1240 	spin_unlock(&inode->i_lock);
1241 }
1242 
1243 static void
1244 nfs4_dec_nlink_locked(struct inode *inode)
1245 {
1246 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1247 					     NFS_INO_INVALID_CTIME |
1248 					     NFS_INO_INVALID_NLINK);
1249 	drop_nlink(inode);
1250 }
1251 
1252 static void
1253 nfs4_update_changeattr_locked(struct inode *inode,
1254 		struct nfs4_change_info *cinfo,
1255 		unsigned long timestamp, unsigned long cache_validity)
1256 {
1257 	struct nfs_inode *nfsi = NFS_I(inode);
1258 	u64 change_attr = inode_peek_iversion_raw(inode);
1259 
1260 	if (!nfs_have_delegated_mtime(inode))
1261 		cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1262 	if (S_ISDIR(inode->i_mode))
1263 		cache_validity |= NFS_INO_INVALID_DATA;
1264 
1265 	switch (NFS_SERVER(inode)->change_attr_type) {
1266 	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1267 		if (cinfo->after == change_attr)
1268 			goto out;
1269 		break;
1270 	default:
1271 		if ((s64)(change_attr - cinfo->after) >= 0)
1272 			goto out;
1273 	}
1274 
1275 	inode_set_iversion_raw(inode, cinfo->after);
1276 	if (!cinfo->atomic || cinfo->before != change_attr) {
1277 		if (S_ISDIR(inode->i_mode))
1278 			nfs_force_lookup_revalidate(inode);
1279 
1280 		if (!nfs_have_delegated_attributes(inode))
1281 			cache_validity |=
1282 				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1283 				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1284 				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1285 				NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1286 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1287 	}
1288 	nfsi->attrtimeo_timestamp = jiffies;
1289 	nfsi->read_cache_jiffies = timestamp;
1290 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1291 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1292 out:
1293 	nfs_set_cache_invalid(inode, cache_validity);
1294 }
1295 
1296 void
1297 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1298 		unsigned long timestamp, unsigned long cache_validity)
1299 {
1300 	spin_lock(&dir->i_lock);
1301 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1302 	spin_unlock(&dir->i_lock);
1303 }
1304 
1305 struct nfs4_open_createattrs {
1306 	struct nfs4_label *label;
1307 	struct iattr *sattr;
1308 	const __u32 verf[2];
1309 };
1310 
1311 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1312 		int err, struct nfs4_exception *exception)
1313 {
1314 	if (err != -EINVAL)
1315 		return false;
1316 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1317 		return false;
1318 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1319 	exception->retry = 1;
1320 	return true;
1321 }
1322 
1323 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1324 {
1325 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1326 }
1327 
1328 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1329 {
1330 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1331 
1332 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1333 }
1334 
1335 static u32
1336 nfs4_fmode_to_share_access(fmode_t fmode)
1337 {
1338 	u32 res = 0;
1339 
1340 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1341 	case FMODE_READ:
1342 		res = NFS4_SHARE_ACCESS_READ;
1343 		break;
1344 	case FMODE_WRITE:
1345 		res = NFS4_SHARE_ACCESS_WRITE;
1346 		break;
1347 	case FMODE_READ|FMODE_WRITE:
1348 		res = NFS4_SHARE_ACCESS_BOTH;
1349 	}
1350 	return res;
1351 }
1352 
1353 static u32
1354 nfs4_map_atomic_open_share(struct nfs_server *server,
1355 		fmode_t fmode, int openflags)
1356 {
1357 	u32 res = nfs4_fmode_to_share_access(fmode);
1358 
1359 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1360 		goto out;
1361 	/* Want no delegation if we're using O_DIRECT */
1362 	if (openflags & O_DIRECT) {
1363 		res |= NFS4_SHARE_WANT_NO_DELEG;
1364 		goto out;
1365 	}
1366 	/* res |= NFS4_SHARE_WANT_NO_PREFERENCE; */
1367 	if (server->caps & NFS_CAP_DELEGTIME)
1368 		res |= NFS4_SHARE_WANT_DELEG_TIMESTAMPS;
1369 	if (server->caps & NFS_CAP_OPEN_XOR)
1370 		res |= NFS4_SHARE_WANT_OPEN_XOR_DELEGATION;
1371 out:
1372 	return res;
1373 }
1374 
1375 static enum open_claim_type4
1376 nfs4_map_atomic_open_claim(struct nfs_server *server,
1377 		enum open_claim_type4 claim)
1378 {
1379 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1380 		return claim;
1381 	switch (claim) {
1382 	default:
1383 		return claim;
1384 	case NFS4_OPEN_CLAIM_FH:
1385 		return NFS4_OPEN_CLAIM_NULL;
1386 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1387 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1388 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1389 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1390 	}
1391 }
1392 
1393 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1394 {
1395 	p->o_res.f_attr = &p->f_attr;
1396 	p->o_res.seqid = p->o_arg.seqid;
1397 	p->c_res.seqid = p->c_arg.seqid;
1398 	p->o_res.server = p->o_arg.server;
1399 	p->o_res.access_request = p->o_arg.access;
1400 	nfs_fattr_init(&p->f_attr);
1401 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1402 }
1403 
1404 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1405 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1406 		const struct nfs4_open_createattrs *c,
1407 		enum open_claim_type4 claim,
1408 		gfp_t gfp_mask)
1409 {
1410 	struct dentry *parent = dget_parent(dentry);
1411 	struct inode *dir = d_inode(parent);
1412 	struct nfs_server *server = NFS_SERVER(dir);
1413 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1414 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1415 	struct nfs4_opendata *p;
1416 
1417 	p = kzalloc(sizeof(*p), gfp_mask);
1418 	if (p == NULL)
1419 		goto err;
1420 
1421 	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1422 	if (IS_ERR(p->f_attr.label))
1423 		goto err_free_p;
1424 
1425 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1426 	if (IS_ERR(p->a_label))
1427 		goto err_free_f;
1428 
1429 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1430 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1431 	if (IS_ERR(p->o_arg.seqid))
1432 		goto err_free_label;
1433 	nfs_sb_active(dentry->d_sb);
1434 	p->dentry = dget(dentry);
1435 	p->dir = parent;
1436 	p->owner = sp;
1437 	atomic_inc(&sp->so_count);
1438 	p->o_arg.open_flags = flags;
1439 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1440 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1441 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1442 			fmode, flags);
1443 	if (flags & O_CREAT) {
1444 		p->o_arg.umask = current_umask();
1445 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1446 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1447 			p->o_arg.u.attrs = &p->attrs;
1448 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1449 
1450 			memcpy(p->o_arg.u.verifier.data, c->verf,
1451 					sizeof(p->o_arg.u.verifier.data));
1452 		}
1453 	}
1454 	/* ask server to check for all possible rights as results
1455 	 * are cached */
1456 	switch (p->o_arg.claim) {
1457 	default:
1458 		break;
1459 	case NFS4_OPEN_CLAIM_NULL:
1460 	case NFS4_OPEN_CLAIM_FH:
1461 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1462 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1463 				  NFS4_ACCESS_EXECUTE |
1464 				  nfs_access_xattr_mask(server);
1465 	}
1466 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1467 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1468 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1469 	p->o_arg.name = &dentry->d_name;
1470 	p->o_arg.server = server;
1471 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1472 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1473 	switch (p->o_arg.claim) {
1474 	case NFS4_OPEN_CLAIM_NULL:
1475 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1476 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1477 		p->o_arg.fh = NFS_FH(dir);
1478 		break;
1479 	case NFS4_OPEN_CLAIM_PREVIOUS:
1480 	case NFS4_OPEN_CLAIM_FH:
1481 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1482 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1483 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1484 	}
1485 	p->c_arg.fh = &p->o_res.fh;
1486 	p->c_arg.stateid = &p->o_res.stateid;
1487 	p->c_arg.seqid = p->o_arg.seqid;
1488 	nfs4_init_opendata_res(p);
1489 	kref_init(&p->kref);
1490 	return p;
1491 
1492 err_free_label:
1493 	nfs4_label_free(p->a_label);
1494 err_free_f:
1495 	nfs4_label_free(p->f_attr.label);
1496 err_free_p:
1497 	kfree(p);
1498 err:
1499 	dput(parent);
1500 	return NULL;
1501 }
1502 
1503 static void nfs4_opendata_free(struct kref *kref)
1504 {
1505 	struct nfs4_opendata *p = container_of(kref,
1506 			struct nfs4_opendata, kref);
1507 	struct super_block *sb = p->dentry->d_sb;
1508 
1509 	nfs4_lgopen_release(p->lgp);
1510 	nfs_free_seqid(p->o_arg.seqid);
1511 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1512 	if (p->state != NULL)
1513 		nfs4_put_open_state(p->state);
1514 	nfs4_put_state_owner(p->owner);
1515 
1516 	nfs4_label_free(p->a_label);
1517 	nfs4_label_free(p->f_attr.label);
1518 
1519 	dput(p->dir);
1520 	dput(p->dentry);
1521 	nfs_sb_deactive(sb);
1522 	nfs_fattr_free_names(&p->f_attr);
1523 	kfree(p->f_attr.mdsthreshold);
1524 	kfree(p);
1525 }
1526 
1527 static void nfs4_opendata_put(struct nfs4_opendata *p)
1528 {
1529 	if (p != NULL)
1530 		kref_put(&p->kref, nfs4_opendata_free);
1531 }
1532 
1533 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1534 		fmode_t fmode)
1535 {
1536 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1537 	case FMODE_READ|FMODE_WRITE:
1538 		return state->n_rdwr != 0;
1539 	case FMODE_WRITE:
1540 		return state->n_wronly != 0;
1541 	case FMODE_READ:
1542 		return state->n_rdonly != 0;
1543 	}
1544 	WARN_ON_ONCE(1);
1545 	return false;
1546 }
1547 
1548 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1549 		int open_mode, enum open_claim_type4 claim)
1550 {
1551 	int ret = 0;
1552 
1553 	if (open_mode & (O_EXCL|O_TRUNC))
1554 		goto out;
1555 	switch (claim) {
1556 	case NFS4_OPEN_CLAIM_NULL:
1557 	case NFS4_OPEN_CLAIM_FH:
1558 		goto out;
1559 	default:
1560 		break;
1561 	}
1562 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1563 		case FMODE_READ:
1564 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1565 				&& state->n_rdonly != 0;
1566 			break;
1567 		case FMODE_WRITE:
1568 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1569 				&& state->n_wronly != 0;
1570 			break;
1571 		case FMODE_READ|FMODE_WRITE:
1572 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1573 				&& state->n_rdwr != 0;
1574 	}
1575 out:
1576 	return ret;
1577 }
1578 
1579 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1580 		enum open_claim_type4 claim)
1581 {
1582 	if (delegation == NULL)
1583 		return 0;
1584 	if ((delegation->type & fmode) != fmode)
1585 		return 0;
1586 	switch (claim) {
1587 	case NFS4_OPEN_CLAIM_NULL:
1588 	case NFS4_OPEN_CLAIM_FH:
1589 		break;
1590 	case NFS4_OPEN_CLAIM_PREVIOUS:
1591 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1592 			break;
1593 		fallthrough;
1594 	default:
1595 		return 0;
1596 	}
1597 	nfs_mark_delegation_referenced(delegation);
1598 	return 1;
1599 }
1600 
1601 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1602 {
1603 	switch (fmode) {
1604 		case FMODE_WRITE:
1605 			state->n_wronly++;
1606 			break;
1607 		case FMODE_READ:
1608 			state->n_rdonly++;
1609 			break;
1610 		case FMODE_READ|FMODE_WRITE:
1611 			state->n_rdwr++;
1612 	}
1613 	nfs4_state_set_mode_locked(state, state->state | fmode);
1614 }
1615 
1616 #ifdef CONFIG_NFS_V4_1
1617 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1618 {
1619 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1620 		return true;
1621 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1622 		return true;
1623 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1624 		return true;
1625 	return false;
1626 }
1627 #endif /* CONFIG_NFS_V4_1 */
1628 
1629 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1630 {
1631 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1632 		wake_up_all(&state->waitq);
1633 }
1634 
1635 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1636 {
1637 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1638 	bool need_recover = false;
1639 
1640 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1641 		need_recover = true;
1642 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1643 		need_recover = true;
1644 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1645 		need_recover = true;
1646 	if (need_recover)
1647 		nfs4_state_mark_reclaim_nograce(clp, state);
1648 }
1649 
1650 /*
1651  * Check for whether or not the caller may update the open stateid
1652  * to the value passed in by stateid.
1653  *
1654  * Note: This function relies heavily on the server implementing
1655  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1656  * correctly.
1657  * i.e. The stateid seqids have to be initialised to 1, and
1658  * are then incremented on every state transition.
1659  */
1660 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1661 		const nfs4_stateid *stateid)
1662 {
1663 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1664 		/* The common case - we're updating to a new sequence number */
1665 		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1666 			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1667 				return true;
1668 			return false;
1669 		}
1670 		/* The server returned a new stateid */
1671 	}
1672 	/* This is the first OPEN in this generation */
1673 	if (stateid->seqid == cpu_to_be32(1))
1674 		return true;
1675 	return false;
1676 }
1677 
1678 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1679 {
1680 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1681 		return;
1682 	if (state->n_wronly)
1683 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1684 	if (state->n_rdonly)
1685 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1686 	if (state->n_rdwr)
1687 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1688 	set_bit(NFS_OPEN_STATE, &state->flags);
1689 }
1690 
1691 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1692 		nfs4_stateid *stateid, fmode_t fmode)
1693 {
1694 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1695 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1696 	case FMODE_WRITE:
1697 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1698 		break;
1699 	case FMODE_READ:
1700 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1701 		break;
1702 	case 0:
1703 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1704 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1705 		clear_bit(NFS_OPEN_STATE, &state->flags);
1706 	}
1707 	if (stateid == NULL)
1708 		return;
1709 	/* Handle OPEN+OPEN_DOWNGRADE races */
1710 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1711 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1712 		nfs_resync_open_stateid_locked(state);
1713 		goto out;
1714 	}
1715 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1716 		nfs4_stateid_copy(&state->stateid, stateid);
1717 	nfs4_stateid_copy(&state->open_stateid, stateid);
1718 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1719 out:
1720 	nfs_state_log_update_open_stateid(state);
1721 }
1722 
1723 static void nfs_clear_open_stateid(struct nfs4_state *state,
1724 	nfs4_stateid *arg_stateid,
1725 	nfs4_stateid *stateid, fmode_t fmode)
1726 {
1727 	write_seqlock(&state->seqlock);
1728 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1729 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1730 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1731 	write_sequnlock(&state->seqlock);
1732 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1733 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1734 }
1735 
1736 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1737 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1738 	__must_hold(&state->owner->so_lock)
1739 	__must_hold(&state->seqlock)
1740 	__must_hold(RCU)
1741 
1742 {
1743 	DEFINE_WAIT(wait);
1744 	int status = 0;
1745 	for (;;) {
1746 
1747 		if (nfs_stateid_is_sequential(state, stateid))
1748 			break;
1749 
1750 		if (status)
1751 			break;
1752 		/* Rely on seqids for serialisation with NFSv4.0 */
1753 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1754 			break;
1755 
1756 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1757 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1758 		/*
1759 		 * Ensure we process the state changes in the same order
1760 		 * in which the server processed them by delaying the
1761 		 * update of the stateid until we are in sequence.
1762 		 */
1763 		write_sequnlock(&state->seqlock);
1764 		spin_unlock(&state->owner->so_lock);
1765 		rcu_read_unlock();
1766 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1767 
1768 		if (!fatal_signal_pending(current)) {
1769 			if (schedule_timeout(5*HZ) == 0)
1770 				status = -EAGAIN;
1771 			else
1772 				status = 0;
1773 		} else
1774 			status = -EINTR;
1775 		finish_wait(&state->waitq, &wait);
1776 		rcu_read_lock();
1777 		spin_lock(&state->owner->so_lock);
1778 		write_seqlock(&state->seqlock);
1779 	}
1780 
1781 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1782 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1783 		nfs4_stateid_copy(freeme, &state->open_stateid);
1784 		nfs_test_and_clear_all_open_stateid(state);
1785 	}
1786 
1787 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1788 		nfs4_stateid_copy(&state->stateid, stateid);
1789 	nfs4_stateid_copy(&state->open_stateid, stateid);
1790 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1791 	nfs_state_log_update_open_stateid(state);
1792 }
1793 
1794 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1795 		const nfs4_stateid *open_stateid,
1796 		fmode_t fmode,
1797 		nfs4_stateid *freeme)
1798 {
1799 	/*
1800 	 * Protect the call to nfs4_state_set_mode_locked and
1801 	 * serialise the stateid update
1802 	 */
1803 	write_seqlock(&state->seqlock);
1804 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1805 	switch (fmode) {
1806 	case FMODE_READ:
1807 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1808 		break;
1809 	case FMODE_WRITE:
1810 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1811 		break;
1812 	case FMODE_READ|FMODE_WRITE:
1813 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1814 	}
1815 	set_bit(NFS_OPEN_STATE, &state->flags);
1816 	write_sequnlock(&state->seqlock);
1817 }
1818 
1819 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1820 {
1821 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1822 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1823 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1824 	clear_bit(NFS_OPEN_STATE, &state->flags);
1825 }
1826 
1827 static void nfs_state_set_delegation(struct nfs4_state *state,
1828 		const nfs4_stateid *deleg_stateid,
1829 		fmode_t fmode)
1830 {
1831 	/*
1832 	 * Protect the call to nfs4_state_set_mode_locked and
1833 	 * serialise the stateid update
1834 	 */
1835 	write_seqlock(&state->seqlock);
1836 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1837 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1838 	write_sequnlock(&state->seqlock);
1839 }
1840 
1841 static void nfs_state_clear_delegation(struct nfs4_state *state)
1842 {
1843 	write_seqlock(&state->seqlock);
1844 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1845 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1846 	write_sequnlock(&state->seqlock);
1847 }
1848 
1849 int update_open_stateid(struct nfs4_state *state,
1850 		const nfs4_stateid *open_stateid,
1851 		const nfs4_stateid *delegation,
1852 		fmode_t fmode)
1853 {
1854 	struct nfs_server *server = NFS_SERVER(state->inode);
1855 	struct nfs_client *clp = server->nfs_client;
1856 	struct nfs_inode *nfsi = NFS_I(state->inode);
1857 	struct nfs_delegation *deleg_cur;
1858 	nfs4_stateid freeme = { };
1859 	int ret = 0;
1860 
1861 	fmode &= (FMODE_READ|FMODE_WRITE);
1862 
1863 	rcu_read_lock();
1864 	spin_lock(&state->owner->so_lock);
1865 	if (open_stateid != NULL) {
1866 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1867 		ret = 1;
1868 	}
1869 
1870 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1871 	if (deleg_cur == NULL)
1872 		goto no_delegation;
1873 
1874 	spin_lock(&deleg_cur->lock);
1875 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1876 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1877 	    (deleg_cur->type & fmode) != fmode)
1878 		goto no_delegation_unlock;
1879 
1880 	if (delegation == NULL)
1881 		delegation = &deleg_cur->stateid;
1882 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1883 		goto no_delegation_unlock;
1884 
1885 	nfs_mark_delegation_referenced(deleg_cur);
1886 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1887 	ret = 1;
1888 no_delegation_unlock:
1889 	spin_unlock(&deleg_cur->lock);
1890 no_delegation:
1891 	if (ret)
1892 		update_open_stateflags(state, fmode);
1893 	spin_unlock(&state->owner->so_lock);
1894 	rcu_read_unlock();
1895 
1896 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1897 		nfs4_schedule_state_manager(clp);
1898 	if (freeme.type != 0)
1899 		nfs4_test_and_free_stateid(server, &freeme,
1900 				state->owner->so_cred);
1901 
1902 	return ret;
1903 }
1904 
1905 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1906 		const nfs4_stateid *stateid)
1907 {
1908 	struct nfs4_state *state = lsp->ls_state;
1909 	bool ret = false;
1910 
1911 	spin_lock(&state->state_lock);
1912 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1913 		goto out_noupdate;
1914 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1915 		goto out_noupdate;
1916 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1917 	ret = true;
1918 out_noupdate:
1919 	spin_unlock(&state->state_lock);
1920 	return ret;
1921 }
1922 
1923 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1924 {
1925 	struct nfs_delegation *delegation;
1926 
1927 	fmode &= FMODE_READ|FMODE_WRITE;
1928 	rcu_read_lock();
1929 	delegation = nfs4_get_valid_delegation(inode);
1930 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1931 		rcu_read_unlock();
1932 		return;
1933 	}
1934 	rcu_read_unlock();
1935 	nfs4_inode_return_delegation(inode);
1936 }
1937 
1938 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1939 {
1940 	struct nfs4_state *state = opendata->state;
1941 	struct nfs_delegation *delegation;
1942 	int open_mode = opendata->o_arg.open_flags;
1943 	fmode_t fmode = opendata->o_arg.fmode;
1944 	enum open_claim_type4 claim = opendata->o_arg.claim;
1945 	nfs4_stateid stateid;
1946 	int ret = -EAGAIN;
1947 
1948 	for (;;) {
1949 		spin_lock(&state->owner->so_lock);
1950 		if (can_open_cached(state, fmode, open_mode, claim)) {
1951 			update_open_stateflags(state, fmode);
1952 			spin_unlock(&state->owner->so_lock);
1953 			goto out_return_state;
1954 		}
1955 		spin_unlock(&state->owner->so_lock);
1956 		rcu_read_lock();
1957 		delegation = nfs4_get_valid_delegation(state->inode);
1958 		if (!can_open_delegated(delegation, fmode, claim)) {
1959 			rcu_read_unlock();
1960 			break;
1961 		}
1962 		/* Save the delegation */
1963 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1964 		rcu_read_unlock();
1965 		nfs_release_seqid(opendata->o_arg.seqid);
1966 		if (!opendata->is_recover) {
1967 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1968 			if (ret != 0)
1969 				goto out;
1970 		}
1971 		ret = -EAGAIN;
1972 
1973 		/* Try to update the stateid using the delegation */
1974 		if (update_open_stateid(state, NULL, &stateid, fmode))
1975 			goto out_return_state;
1976 	}
1977 out:
1978 	return ERR_PTR(ret);
1979 out_return_state:
1980 	refcount_inc(&state->count);
1981 	return state;
1982 }
1983 
1984 static void
1985 nfs4_process_delegation(struct inode *inode, const struct cred *cred,
1986 			enum open_claim_type4 claim,
1987 			const struct nfs4_open_delegation *delegation)
1988 {
1989 	switch (delegation->open_delegation_type) {
1990 	case NFS4_OPEN_DELEGATE_READ:
1991 	case NFS4_OPEN_DELEGATE_WRITE:
1992 	case NFS4_OPEN_DELEGATE_READ_ATTRS_DELEG:
1993 	case NFS4_OPEN_DELEGATE_WRITE_ATTRS_DELEG:
1994 		break;
1995 	default:
1996 		return;
1997 	}
1998 	switch (claim) {
1999 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2000 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2001 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
2002 				   "returning a delegation for "
2003 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
2004 				   NFS_SERVER(inode)->nfs_client->cl_hostname);
2005 		break;
2006 	case NFS4_OPEN_CLAIM_PREVIOUS:
2007 		nfs_inode_reclaim_delegation(inode, cred, delegation->type,
2008 					     &delegation->stateid,
2009 					     delegation->pagemod_limit,
2010 					     delegation->open_delegation_type);
2011 		break;
2012 	default:
2013 		nfs_inode_set_delegation(inode, cred, delegation->type,
2014 					 &delegation->stateid,
2015 					 delegation->pagemod_limit,
2016 					 delegation->open_delegation_type);
2017 	}
2018 	if (delegation->do_recall)
2019 		nfs_async_inode_return_delegation(inode, &delegation->stateid);
2020 }
2021 
2022 /*
2023  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
2024  * and update the nfs4_state.
2025  */
2026 static struct nfs4_state *
2027 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
2028 {
2029 	struct inode *inode = data->state->inode;
2030 	struct nfs4_state *state = data->state;
2031 	int ret;
2032 
2033 	if (!data->rpc_done) {
2034 		if (data->rpc_status)
2035 			return ERR_PTR(data->rpc_status);
2036 		return nfs4_try_open_cached(data);
2037 	}
2038 
2039 	ret = nfs_refresh_inode(inode, &data->f_attr);
2040 	if (ret)
2041 		return ERR_PTR(ret);
2042 
2043 	nfs4_process_delegation(state->inode,
2044 				data->owner->so_cred,
2045 				data->o_arg.claim,
2046 				&data->o_res.delegation);
2047 
2048 	if (!(data->o_res.rflags & NFS4_OPEN_RESULT_NO_OPEN_STATEID)) {
2049 		if (!update_open_stateid(state, &data->o_res.stateid,
2050 					 NULL, data->o_arg.fmode))
2051 			return ERR_PTR(-EAGAIN);
2052 	} else if (!update_open_stateid(state, NULL, NULL, data->o_arg.fmode))
2053 		return ERR_PTR(-EAGAIN);
2054 	refcount_inc(&state->count);
2055 
2056 	return state;
2057 }
2058 
2059 static struct inode *
2060 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2061 {
2062 	struct inode *inode;
2063 
2064 	switch (data->o_arg.claim) {
2065 	case NFS4_OPEN_CLAIM_NULL:
2066 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2067 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2068 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2069 			return ERR_PTR(-EAGAIN);
2070 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2071 				&data->f_attr);
2072 		break;
2073 	default:
2074 		inode = d_inode(data->dentry);
2075 		ihold(inode);
2076 		nfs_refresh_inode(inode, &data->f_attr);
2077 	}
2078 	return inode;
2079 }
2080 
2081 static struct nfs4_state *
2082 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2083 {
2084 	struct nfs4_state *state;
2085 	struct inode *inode;
2086 
2087 	inode = nfs4_opendata_get_inode(data);
2088 	if (IS_ERR(inode))
2089 		return ERR_CAST(inode);
2090 	if (data->state != NULL && data->state->inode == inode) {
2091 		state = data->state;
2092 		refcount_inc(&state->count);
2093 	} else
2094 		state = nfs4_get_open_state(inode, data->owner);
2095 	iput(inode);
2096 	if (state == NULL)
2097 		state = ERR_PTR(-ENOMEM);
2098 	return state;
2099 }
2100 
2101 static struct nfs4_state *
2102 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2103 {
2104 	struct nfs4_state *state;
2105 
2106 	if (!data->rpc_done) {
2107 		state = nfs4_try_open_cached(data);
2108 		trace_nfs4_cached_open(data->state);
2109 		goto out;
2110 	}
2111 
2112 	state = nfs4_opendata_find_nfs4_state(data);
2113 	if (IS_ERR(state))
2114 		goto out;
2115 
2116 	nfs4_process_delegation(state->inode,
2117 				data->owner->so_cred,
2118 				data->o_arg.claim,
2119 				&data->o_res.delegation);
2120 
2121 	if (!(data->o_res.rflags & NFS4_OPEN_RESULT_NO_OPEN_STATEID)) {
2122 		if (!update_open_stateid(state, &data->o_res.stateid,
2123 					 NULL, data->o_arg.fmode)) {
2124 			nfs4_put_open_state(state);
2125 			state = ERR_PTR(-EAGAIN);
2126 		}
2127 	} else if (!update_open_stateid(state, NULL, NULL, data->o_arg.fmode)) {
2128 		nfs4_put_open_state(state);
2129 		state = ERR_PTR(-EAGAIN);
2130 	}
2131 out:
2132 	nfs_release_seqid(data->o_arg.seqid);
2133 	return state;
2134 }
2135 
2136 static struct nfs4_state *
2137 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2138 {
2139 	struct nfs4_state *ret;
2140 
2141 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2142 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2143 	else
2144 		ret = _nfs4_opendata_to_nfs4_state(data);
2145 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2146 	return ret;
2147 }
2148 
2149 static struct nfs_open_context *
2150 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2151 {
2152 	struct nfs_inode *nfsi = NFS_I(state->inode);
2153 	struct nfs_open_context *ctx;
2154 
2155 	rcu_read_lock();
2156 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2157 		if (ctx->state != state)
2158 			continue;
2159 		if ((ctx->mode & mode) != mode)
2160 			continue;
2161 		if (!get_nfs_open_context(ctx))
2162 			continue;
2163 		rcu_read_unlock();
2164 		return ctx;
2165 	}
2166 	rcu_read_unlock();
2167 	return ERR_PTR(-ENOENT);
2168 }
2169 
2170 static struct nfs_open_context *
2171 nfs4_state_find_open_context(struct nfs4_state *state)
2172 {
2173 	struct nfs_open_context *ctx;
2174 
2175 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2176 	if (!IS_ERR(ctx))
2177 		return ctx;
2178 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2179 	if (!IS_ERR(ctx))
2180 		return ctx;
2181 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2182 }
2183 
2184 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2185 		struct nfs4_state *state, enum open_claim_type4 claim)
2186 {
2187 	struct nfs4_opendata *opendata;
2188 
2189 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2190 			NULL, claim, GFP_NOFS);
2191 	if (opendata == NULL)
2192 		return ERR_PTR(-ENOMEM);
2193 	opendata->state = state;
2194 	refcount_inc(&state->count);
2195 	return opendata;
2196 }
2197 
2198 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2199 				    fmode_t fmode)
2200 {
2201 	struct nfs4_state *newstate;
2202 	struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2203 	int openflags = opendata->o_arg.open_flags;
2204 	int ret;
2205 
2206 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2207 		return 0;
2208 	opendata->o_arg.fmode = fmode;
2209 	opendata->o_arg.share_access =
2210 		nfs4_map_atomic_open_share(server, fmode, openflags);
2211 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2212 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2213 	nfs4_init_opendata_res(opendata);
2214 	ret = _nfs4_recover_proc_open(opendata);
2215 	if (ret != 0)
2216 		return ret;
2217 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2218 	if (IS_ERR(newstate))
2219 		return PTR_ERR(newstate);
2220 	if (newstate != opendata->state)
2221 		ret = -ESTALE;
2222 	nfs4_close_state(newstate, fmode);
2223 	return ret;
2224 }
2225 
2226 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2227 {
2228 	int ret;
2229 
2230 	/* memory barrier prior to reading state->n_* */
2231 	smp_rmb();
2232 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2233 	if (ret != 0)
2234 		return ret;
2235 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2236 	if (ret != 0)
2237 		return ret;
2238 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2239 	if (ret != 0)
2240 		return ret;
2241 	/*
2242 	 * We may have performed cached opens for all three recoveries.
2243 	 * Check if we need to update the current stateid.
2244 	 */
2245 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2246 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2247 		write_seqlock(&state->seqlock);
2248 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2249 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2250 		write_sequnlock(&state->seqlock);
2251 	}
2252 	return 0;
2253 }
2254 
2255 /*
2256  * OPEN_RECLAIM:
2257  * 	reclaim state on the server after a reboot.
2258  */
2259 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2260 {
2261 	struct nfs_delegation *delegation;
2262 	struct nfs4_opendata *opendata;
2263 	u32 delegation_type = NFS4_OPEN_DELEGATE_NONE;
2264 	int status;
2265 
2266 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2267 			NFS4_OPEN_CLAIM_PREVIOUS);
2268 	if (IS_ERR(opendata))
2269 		return PTR_ERR(opendata);
2270 	rcu_read_lock();
2271 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2272 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0) {
2273 		switch(delegation->type) {
2274 		case FMODE_READ:
2275 			delegation_type = NFS4_OPEN_DELEGATE_READ;
2276 			if (test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags))
2277 				delegation_type = NFS4_OPEN_DELEGATE_READ_ATTRS_DELEG;
2278 			break;
2279 		case FMODE_WRITE:
2280 		case FMODE_READ|FMODE_WRITE:
2281 			delegation_type = NFS4_OPEN_DELEGATE_WRITE;
2282 			if (test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags))
2283 				delegation_type = NFS4_OPEN_DELEGATE_WRITE_ATTRS_DELEG;
2284 		}
2285 	}
2286 	rcu_read_unlock();
2287 	opendata->o_arg.u.delegation_type = delegation_type;
2288 	status = nfs4_open_recover(opendata, state);
2289 	nfs4_opendata_put(opendata);
2290 	return status;
2291 }
2292 
2293 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2294 {
2295 	struct nfs_server *server = NFS_SERVER(state->inode);
2296 	struct nfs4_exception exception = { };
2297 	int err;
2298 	do {
2299 		err = _nfs4_do_open_reclaim(ctx, state);
2300 		trace_nfs4_open_reclaim(ctx, 0, err);
2301 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2302 			continue;
2303 		if (err != -NFS4ERR_DELAY)
2304 			break;
2305 		nfs4_handle_exception(server, err, &exception);
2306 	} while (exception.retry);
2307 	return err;
2308 }
2309 
2310 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2311 {
2312 	struct nfs_open_context *ctx;
2313 	int ret;
2314 
2315 	ctx = nfs4_state_find_open_context(state);
2316 	if (IS_ERR(ctx))
2317 		return -EAGAIN;
2318 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2319 	nfs_state_clear_open_state_flags(state);
2320 	ret = nfs4_do_open_reclaim(ctx, state);
2321 	put_nfs_open_context(ctx);
2322 	return ret;
2323 }
2324 
2325 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)
2326 {
2327 	switch (err) {
2328 		default:
2329 			printk(KERN_ERR "NFS: %s: unhandled error "
2330 					"%d.\n", __func__, err);
2331 			fallthrough;
2332 		case 0:
2333 		case -ENOENT:
2334 		case -EAGAIN:
2335 		case -ESTALE:
2336 		case -ETIMEDOUT:
2337 			break;
2338 		case -NFS4ERR_BADSESSION:
2339 		case -NFS4ERR_BADSLOT:
2340 		case -NFS4ERR_BAD_HIGH_SLOT:
2341 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2342 		case -NFS4ERR_DEADSESSION:
2343 			return -EAGAIN;
2344 		case -NFS4ERR_STALE_CLIENTID:
2345 		case -NFS4ERR_STALE_STATEID:
2346 			/* Don't recall a delegation if it was lost */
2347 			nfs4_schedule_lease_recovery(server->nfs_client);
2348 			return -EAGAIN;
2349 		case -NFS4ERR_MOVED:
2350 			nfs4_schedule_migration_recovery(server);
2351 			return -EAGAIN;
2352 		case -NFS4ERR_LEASE_MOVED:
2353 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2354 			return -EAGAIN;
2355 		case -NFS4ERR_DELEG_REVOKED:
2356 		case -NFS4ERR_ADMIN_REVOKED:
2357 		case -NFS4ERR_EXPIRED:
2358 		case -NFS4ERR_BAD_STATEID:
2359 		case -NFS4ERR_OPENMODE:
2360 			nfs_inode_find_state_and_recover(state->inode,
2361 					stateid);
2362 			nfs4_schedule_stateid_recovery(server, state);
2363 			return -EAGAIN;
2364 		case -NFS4ERR_DELAY:
2365 		case -NFS4ERR_GRACE:
2366 			ssleep(1);
2367 			return -EAGAIN;
2368 		case -ENOMEM:
2369 		case -NFS4ERR_DENIED:
2370 			if (fl) {
2371 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2372 				if (lsp)
2373 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2374 			}
2375 			return 0;
2376 	}
2377 	return err;
2378 }
2379 
2380 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2381 		struct nfs4_state *state, const nfs4_stateid *stateid)
2382 {
2383 	struct nfs_server *server = NFS_SERVER(state->inode);
2384 	struct nfs4_opendata *opendata;
2385 	int err = 0;
2386 
2387 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2388 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2389 	if (IS_ERR(opendata))
2390 		return PTR_ERR(opendata);
2391 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2392 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2393 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2394 		if (err)
2395 			goto out;
2396 	}
2397 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2398 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2399 		if (err)
2400 			goto out;
2401 	}
2402 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2403 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2404 		if (err)
2405 			goto out;
2406 	}
2407 	nfs_state_clear_delegation(state);
2408 out:
2409 	nfs4_opendata_put(opendata);
2410 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2411 }
2412 
2413 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2414 {
2415 	struct nfs4_opendata *data = calldata;
2416 
2417 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2418 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2419 }
2420 
2421 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2422 {
2423 	struct nfs4_opendata *data = calldata;
2424 
2425 	nfs40_sequence_done(task, &data->c_res.seq_res);
2426 
2427 	data->rpc_status = task->tk_status;
2428 	if (data->rpc_status == 0) {
2429 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2430 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2431 		renew_lease(data->o_res.server, data->timestamp);
2432 		data->rpc_done = true;
2433 	}
2434 }
2435 
2436 static void nfs4_open_confirm_release(void *calldata)
2437 {
2438 	struct nfs4_opendata *data = calldata;
2439 	struct nfs4_state *state = NULL;
2440 
2441 	/* If this request hasn't been cancelled, do nothing */
2442 	if (!data->cancelled)
2443 		goto out_free;
2444 	/* In case of error, no cleanup! */
2445 	if (!data->rpc_done)
2446 		goto out_free;
2447 	state = nfs4_opendata_to_nfs4_state(data);
2448 	if (!IS_ERR(state))
2449 		nfs4_close_state(state, data->o_arg.fmode);
2450 out_free:
2451 	nfs4_opendata_put(data);
2452 }
2453 
2454 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2455 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2456 	.rpc_call_done = nfs4_open_confirm_done,
2457 	.rpc_release = nfs4_open_confirm_release,
2458 };
2459 
2460 /*
2461  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2462  */
2463 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2464 {
2465 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2466 	struct rpc_task *task;
2467 	struct  rpc_message msg = {
2468 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2469 		.rpc_argp = &data->c_arg,
2470 		.rpc_resp = &data->c_res,
2471 		.rpc_cred = data->owner->so_cred,
2472 	};
2473 	struct rpc_task_setup task_setup_data = {
2474 		.rpc_client = server->client,
2475 		.rpc_message = &msg,
2476 		.callback_ops = &nfs4_open_confirm_ops,
2477 		.callback_data = data,
2478 		.workqueue = nfsiod_workqueue,
2479 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2480 	};
2481 	int status;
2482 
2483 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2484 				data->is_recover);
2485 	kref_get(&data->kref);
2486 	data->rpc_done = false;
2487 	data->rpc_status = 0;
2488 	data->timestamp = jiffies;
2489 	task = rpc_run_task(&task_setup_data);
2490 	if (IS_ERR(task))
2491 		return PTR_ERR(task);
2492 	status = rpc_wait_for_completion_task(task);
2493 	if (status != 0) {
2494 		data->cancelled = true;
2495 		smp_wmb();
2496 	} else
2497 		status = data->rpc_status;
2498 	rpc_put_task(task);
2499 	return status;
2500 }
2501 
2502 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2503 {
2504 	struct nfs4_opendata *data = calldata;
2505 	struct nfs4_state_owner *sp = data->owner;
2506 	struct nfs_client *clp = sp->so_server->nfs_client;
2507 	enum open_claim_type4 claim = data->o_arg.claim;
2508 
2509 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2510 		goto out_wait;
2511 	/*
2512 	 * Check if we still need to send an OPEN call, or if we can use
2513 	 * a delegation instead.
2514 	 */
2515 	if (data->state != NULL) {
2516 		struct nfs_delegation *delegation;
2517 
2518 		if (can_open_cached(data->state, data->o_arg.fmode,
2519 					data->o_arg.open_flags, claim))
2520 			goto out_no_action;
2521 		rcu_read_lock();
2522 		delegation = nfs4_get_valid_delegation(data->state->inode);
2523 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2524 			goto unlock_no_action;
2525 		rcu_read_unlock();
2526 	}
2527 	/* Update client id. */
2528 	data->o_arg.clientid = clp->cl_clientid;
2529 	switch (claim) {
2530 	default:
2531 		break;
2532 	case NFS4_OPEN_CLAIM_PREVIOUS:
2533 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2534 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2535 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2536 		fallthrough;
2537 	case NFS4_OPEN_CLAIM_FH:
2538 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2539 	}
2540 	data->timestamp = jiffies;
2541 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2542 				&data->o_arg.seq_args,
2543 				&data->o_res.seq_res,
2544 				task) != 0)
2545 		nfs_release_seqid(data->o_arg.seqid);
2546 
2547 	/* Set the create mode (note dependency on the session type) */
2548 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2549 	if (data->o_arg.open_flags & O_EXCL) {
2550 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2551 		if (clp->cl_mvops->minor_version == 0) {
2552 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2553 			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2554 			 * because ACCESS will return permission denied for
2555 			 * all bits until close */
2556 			data->o_res.access_request = data->o_arg.access = 0;
2557 		} else if (nfs4_has_persistent_session(clp))
2558 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2559 	}
2560 	return;
2561 unlock_no_action:
2562 	trace_nfs4_cached_open(data->state);
2563 	rcu_read_unlock();
2564 out_no_action:
2565 	task->tk_action = NULL;
2566 out_wait:
2567 	nfs4_sequence_done(task, &data->o_res.seq_res);
2568 }
2569 
2570 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2571 {
2572 	struct nfs4_opendata *data = calldata;
2573 
2574 	data->rpc_status = task->tk_status;
2575 
2576 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2577 		return;
2578 
2579 	if (task->tk_status == 0) {
2580 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2581 			switch (data->o_res.f_attr->mode & S_IFMT) {
2582 			case S_IFREG:
2583 				break;
2584 			case S_IFLNK:
2585 				data->rpc_status = -ELOOP;
2586 				break;
2587 			case S_IFDIR:
2588 				data->rpc_status = -EISDIR;
2589 				break;
2590 			default:
2591 				data->rpc_status = -ENOTDIR;
2592 			}
2593 		}
2594 		renew_lease(data->o_res.server, data->timestamp);
2595 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2596 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2597 	}
2598 	data->rpc_done = true;
2599 }
2600 
2601 static void nfs4_open_release(void *calldata)
2602 {
2603 	struct nfs4_opendata *data = calldata;
2604 	struct nfs4_state *state = NULL;
2605 
2606 	/* If this request hasn't been cancelled, do nothing */
2607 	if (!data->cancelled)
2608 		goto out_free;
2609 	/* In case of error, no cleanup! */
2610 	if (data->rpc_status != 0 || !data->rpc_done)
2611 		goto out_free;
2612 	/* In case we need an open_confirm, no cleanup! */
2613 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2614 		goto out_free;
2615 	state = nfs4_opendata_to_nfs4_state(data);
2616 	if (!IS_ERR(state))
2617 		nfs4_close_state(state, data->o_arg.fmode);
2618 out_free:
2619 	nfs4_opendata_put(data);
2620 }
2621 
2622 static const struct rpc_call_ops nfs4_open_ops = {
2623 	.rpc_call_prepare = nfs4_open_prepare,
2624 	.rpc_call_done = nfs4_open_done,
2625 	.rpc_release = nfs4_open_release,
2626 };
2627 
2628 static int nfs4_run_open_task(struct nfs4_opendata *data,
2629 			      struct nfs_open_context *ctx)
2630 {
2631 	struct inode *dir = d_inode(data->dir);
2632 	struct nfs_server *server = NFS_SERVER(dir);
2633 	struct nfs_openargs *o_arg = &data->o_arg;
2634 	struct nfs_openres *o_res = &data->o_res;
2635 	struct rpc_task *task;
2636 	struct rpc_message msg = {
2637 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2638 		.rpc_argp = o_arg,
2639 		.rpc_resp = o_res,
2640 		.rpc_cred = data->owner->so_cred,
2641 	};
2642 	struct rpc_task_setup task_setup_data = {
2643 		.rpc_client = server->client,
2644 		.rpc_message = &msg,
2645 		.callback_ops = &nfs4_open_ops,
2646 		.callback_data = data,
2647 		.workqueue = nfsiod_workqueue,
2648 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2649 	};
2650 	int status;
2651 
2652 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2653 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2654 
2655 	kref_get(&data->kref);
2656 	data->rpc_done = false;
2657 	data->rpc_status = 0;
2658 	data->cancelled = false;
2659 	data->is_recover = false;
2660 	if (!ctx) {
2661 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2662 		data->is_recover = true;
2663 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2664 	} else {
2665 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2666 		pnfs_lgopen_prepare(data, ctx);
2667 	}
2668 	task = rpc_run_task(&task_setup_data);
2669 	if (IS_ERR(task))
2670 		return PTR_ERR(task);
2671 	status = rpc_wait_for_completion_task(task);
2672 	if (status != 0) {
2673 		data->cancelled = true;
2674 		smp_wmb();
2675 	} else
2676 		status = data->rpc_status;
2677 	rpc_put_task(task);
2678 
2679 	return status;
2680 }
2681 
2682 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2683 {
2684 	struct inode *dir = d_inode(data->dir);
2685 	struct nfs_openres *o_res = &data->o_res;
2686 	int status;
2687 
2688 	status = nfs4_run_open_task(data, NULL);
2689 	if (status != 0 || !data->rpc_done)
2690 		return status;
2691 
2692 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2693 
2694 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2695 		status = _nfs4_proc_open_confirm(data);
2696 
2697 	return status;
2698 }
2699 
2700 /*
2701  * Additional permission checks in order to distinguish between an
2702  * open for read, and an open for execute. This works around the
2703  * fact that NFSv4 OPEN treats read and execute permissions as being
2704  * the same.
2705  * Note that in the non-execute case, we want to turn off permission
2706  * checking if we just created a new file (POSIX open() semantics).
2707  */
2708 static int nfs4_opendata_access(const struct cred *cred,
2709 				struct nfs4_opendata *opendata,
2710 				struct nfs4_state *state, fmode_t fmode)
2711 {
2712 	struct nfs_access_entry cache;
2713 	u32 mask, flags;
2714 
2715 	/* access call failed or for some reason the server doesn't
2716 	 * support any access modes -- defer access call until later */
2717 	if (opendata->o_res.access_supported == 0)
2718 		return 0;
2719 
2720 	mask = 0;
2721 	if (fmode & FMODE_EXEC) {
2722 		/* ONLY check for exec rights */
2723 		if (S_ISDIR(state->inode->i_mode))
2724 			mask = NFS4_ACCESS_LOOKUP;
2725 		else
2726 			mask = NFS4_ACCESS_EXECUTE;
2727 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2728 		mask = NFS4_ACCESS_READ;
2729 
2730 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2731 	nfs_access_add_cache(state->inode, &cache, cred);
2732 
2733 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2734 	if ((mask & ~cache.mask & flags) == 0)
2735 		return 0;
2736 
2737 	return -EACCES;
2738 }
2739 
2740 /*
2741  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2742  */
2743 static int _nfs4_proc_open(struct nfs4_opendata *data,
2744 			   struct nfs_open_context *ctx)
2745 {
2746 	struct inode *dir = d_inode(data->dir);
2747 	struct nfs_server *server = NFS_SERVER(dir);
2748 	struct nfs_openargs *o_arg = &data->o_arg;
2749 	struct nfs_openres *o_res = &data->o_res;
2750 	int status;
2751 
2752 	status = nfs4_run_open_task(data, ctx);
2753 	if (!data->rpc_done)
2754 		return status;
2755 	if (status != 0) {
2756 		if (status == -NFS4ERR_BADNAME &&
2757 				!(o_arg->open_flags & O_CREAT))
2758 			return -ENOENT;
2759 		return status;
2760 	}
2761 
2762 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2763 
2764 	if (o_arg->open_flags & O_CREAT) {
2765 		if (o_arg->open_flags & O_EXCL)
2766 			data->file_created = true;
2767 		else if (o_res->cinfo.before != o_res->cinfo.after)
2768 			data->file_created = true;
2769 		if (data->file_created ||
2770 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2771 			nfs4_update_changeattr(dir, &o_res->cinfo,
2772 					o_res->f_attr->time_start,
2773 					NFS_INO_INVALID_DATA);
2774 	}
2775 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2776 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2777 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2778 		status = _nfs4_proc_open_confirm(data);
2779 		if (status != 0)
2780 			return status;
2781 	}
2782 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2783 		struct nfs_fh *fh = &o_res->fh;
2784 
2785 		nfs4_sequence_free_slot(&o_res->seq_res);
2786 		if (o_arg->claim == NFS4_OPEN_CLAIM_FH)
2787 			fh = NFS_FH(d_inode(data->dentry));
2788 		nfs4_proc_getattr(server, fh, o_res->f_attr, NULL);
2789 	}
2790 	return 0;
2791 }
2792 
2793 /*
2794  * OPEN_EXPIRED:
2795  * 	reclaim state on the server after a network partition.
2796  * 	Assumes caller holds the appropriate lock
2797  */
2798 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2799 {
2800 	struct nfs4_opendata *opendata;
2801 	int ret;
2802 
2803 	opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2804 	if (IS_ERR(opendata))
2805 		return PTR_ERR(opendata);
2806 	/*
2807 	 * We're not recovering a delegation, so ask for no delegation.
2808 	 * Otherwise the recovery thread could deadlock with an outstanding
2809 	 * delegation return.
2810 	 */
2811 	opendata->o_arg.open_flags = O_DIRECT;
2812 	ret = nfs4_open_recover(opendata, state);
2813 	if (ret == -ESTALE)
2814 		d_drop(ctx->dentry);
2815 	nfs4_opendata_put(opendata);
2816 	return ret;
2817 }
2818 
2819 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2820 {
2821 	struct nfs_server *server = NFS_SERVER(state->inode);
2822 	struct nfs4_exception exception = { };
2823 	int err;
2824 
2825 	do {
2826 		err = _nfs4_open_expired(ctx, state);
2827 		trace_nfs4_open_expired(ctx, 0, err);
2828 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2829 			continue;
2830 		switch (err) {
2831 		default:
2832 			goto out;
2833 		case -NFS4ERR_GRACE:
2834 		case -NFS4ERR_DELAY:
2835 			nfs4_handle_exception(server, err, &exception);
2836 			err = 0;
2837 		}
2838 	} while (exception.retry);
2839 out:
2840 	return err;
2841 }
2842 
2843 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2844 {
2845 	struct nfs_open_context *ctx;
2846 	int ret;
2847 
2848 	ctx = nfs4_state_find_open_context(state);
2849 	if (IS_ERR(ctx))
2850 		return -EAGAIN;
2851 	ret = nfs4_do_open_expired(ctx, state);
2852 	put_nfs_open_context(ctx);
2853 	return ret;
2854 }
2855 
2856 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2857 		const nfs4_stateid *stateid)
2858 {
2859 	nfs_remove_bad_delegation(state->inode, stateid);
2860 	nfs_state_clear_delegation(state);
2861 }
2862 
2863 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2864 {
2865 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2866 		nfs_finish_clear_delegation_stateid(state, NULL);
2867 }
2868 
2869 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2870 {
2871 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2872 	nfs40_clear_delegation_stateid(state);
2873 	nfs_state_clear_open_state_flags(state);
2874 	return nfs4_open_expired(sp, state);
2875 }
2876 
2877 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2878 					       const nfs4_stateid *stateid,
2879 					       const struct cred *cred)
2880 {
2881 	return -NFS4ERR_BAD_STATEID;
2882 }
2883 
2884 #if defined(CONFIG_NFS_V4_1)
2885 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2886 					       const nfs4_stateid *stateid,
2887 					       const struct cred *cred)
2888 {
2889 	int status;
2890 
2891 	switch (stateid->type) {
2892 	default:
2893 		break;
2894 	case NFS4_INVALID_STATEID_TYPE:
2895 	case NFS4_SPECIAL_STATEID_TYPE:
2896 		return -NFS4ERR_BAD_STATEID;
2897 	case NFS4_REVOKED_STATEID_TYPE:
2898 		goto out_free;
2899 	}
2900 
2901 	status = nfs41_test_stateid(server, stateid, cred);
2902 	switch (status) {
2903 	case -NFS4ERR_EXPIRED:
2904 	case -NFS4ERR_ADMIN_REVOKED:
2905 	case -NFS4ERR_DELEG_REVOKED:
2906 		break;
2907 	default:
2908 		return status;
2909 	}
2910 out_free:
2911 	/* Ack the revoked state to the server */
2912 	nfs41_free_stateid(server, stateid, cred, true);
2913 	return -NFS4ERR_EXPIRED;
2914 }
2915 
2916 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2917 {
2918 	struct nfs_server *server = NFS_SERVER(state->inode);
2919 	nfs4_stateid stateid;
2920 	struct nfs_delegation *delegation;
2921 	const struct cred *cred = NULL;
2922 	int status, ret = NFS_OK;
2923 
2924 	/* Get the delegation credential for use by test/free_stateid */
2925 	rcu_read_lock();
2926 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2927 	if (delegation == NULL) {
2928 		rcu_read_unlock();
2929 		nfs_state_clear_delegation(state);
2930 		return NFS_OK;
2931 	}
2932 
2933 	spin_lock(&delegation->lock);
2934 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2935 
2936 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2937 				&delegation->flags)) {
2938 		spin_unlock(&delegation->lock);
2939 		rcu_read_unlock();
2940 		return NFS_OK;
2941 	}
2942 
2943 	if (delegation->cred)
2944 		cred = get_cred(delegation->cred);
2945 	spin_unlock(&delegation->lock);
2946 	rcu_read_unlock();
2947 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2948 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2949 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2950 		nfs_finish_clear_delegation_stateid(state, &stateid);
2951 	else
2952 		ret = status;
2953 
2954 	put_cred(cred);
2955 	return ret;
2956 }
2957 
2958 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2959 {
2960 	nfs4_stateid tmp;
2961 
2962 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2963 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2964 				&tmp, NULL) &&
2965 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2966 		nfs_state_set_delegation(state, &tmp, state->state);
2967 	else
2968 		nfs_state_clear_delegation(state);
2969 }
2970 
2971 /**
2972  * nfs41_check_expired_locks - possibly free a lock stateid
2973  *
2974  * @state: NFSv4 state for an inode
2975  *
2976  * Returns NFS_OK if recovery for this stateid is now finished.
2977  * Otherwise a negative NFS4ERR value is returned.
2978  */
2979 static int nfs41_check_expired_locks(struct nfs4_state *state)
2980 {
2981 	int status, ret = NFS_OK;
2982 	struct nfs4_lock_state *lsp, *prev = NULL;
2983 	struct nfs_server *server = NFS_SERVER(state->inode);
2984 
2985 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2986 		goto out;
2987 
2988 	spin_lock(&state->state_lock);
2989 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2990 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2991 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2992 
2993 			refcount_inc(&lsp->ls_count);
2994 			spin_unlock(&state->state_lock);
2995 
2996 			nfs4_put_lock_state(prev);
2997 			prev = lsp;
2998 
2999 			status = nfs41_test_and_free_expired_stateid(server,
3000 					&lsp->ls_stateid,
3001 					cred);
3002 			trace_nfs4_test_lock_stateid(state, lsp, status);
3003 			if (status == -NFS4ERR_EXPIRED ||
3004 			    status == -NFS4ERR_BAD_STATEID) {
3005 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
3006 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
3007 				if (!recover_lost_locks)
3008 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
3009 			} else if (status != NFS_OK) {
3010 				ret = status;
3011 				nfs4_put_lock_state(prev);
3012 				goto out;
3013 			}
3014 			spin_lock(&state->state_lock);
3015 		}
3016 	}
3017 	spin_unlock(&state->state_lock);
3018 	nfs4_put_lock_state(prev);
3019 out:
3020 	return ret;
3021 }
3022 
3023 /**
3024  * nfs41_check_open_stateid - possibly free an open stateid
3025  *
3026  * @state: NFSv4 state for an inode
3027  *
3028  * Returns NFS_OK if recovery for this stateid is now finished.
3029  * Otherwise a negative NFS4ERR value is returned.
3030  */
3031 static int nfs41_check_open_stateid(struct nfs4_state *state)
3032 {
3033 	struct nfs_server *server = NFS_SERVER(state->inode);
3034 	nfs4_stateid *stateid = &state->open_stateid;
3035 	const struct cred *cred = state->owner->so_cred;
3036 	int status;
3037 
3038 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3039 		return -NFS4ERR_BAD_STATEID;
3040 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
3041 	trace_nfs4_test_open_stateid(state, NULL, status);
3042 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
3043 		nfs_state_clear_open_state_flags(state);
3044 		stateid->type = NFS4_INVALID_STATEID_TYPE;
3045 		return status;
3046 	}
3047 	if (nfs_open_stateid_recover_openmode(state))
3048 		return -NFS4ERR_OPENMODE;
3049 	return NFS_OK;
3050 }
3051 
3052 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
3053 {
3054 	int status;
3055 
3056 	status = nfs41_check_delegation_stateid(state);
3057 	if (status != NFS_OK)
3058 		return status;
3059 	nfs41_delegation_recover_stateid(state);
3060 
3061 	status = nfs41_check_expired_locks(state);
3062 	if (status != NFS_OK)
3063 		return status;
3064 	status = nfs41_check_open_stateid(state);
3065 	if (status != NFS_OK)
3066 		status = nfs4_open_expired(sp, state);
3067 	return status;
3068 }
3069 #endif
3070 
3071 /*
3072  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
3073  * fields corresponding to attributes that were used to store the verifier.
3074  * Make sure we clobber those fields in the later setattr call
3075  */
3076 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
3077 				struct iattr *sattr, struct nfs4_label **label)
3078 {
3079 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
3080 	__u32 attrset[3];
3081 	unsigned ret;
3082 	unsigned i;
3083 
3084 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3085 		attrset[i] = opendata->o_res.attrset[i];
3086 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3087 			attrset[i] &= ~bitmask[i];
3088 	}
3089 
3090 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3091 		sattr->ia_valid : 0;
3092 
3093 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3094 		if (sattr->ia_valid & ATTR_ATIME_SET)
3095 			ret |= ATTR_ATIME_SET;
3096 		else
3097 			ret |= ATTR_ATIME;
3098 	}
3099 
3100 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3101 		if (sattr->ia_valid & ATTR_MTIME_SET)
3102 			ret |= ATTR_MTIME_SET;
3103 		else
3104 			ret |= ATTR_MTIME;
3105 	}
3106 
3107 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3108 		*label = NULL;
3109 	return ret;
3110 }
3111 
3112 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3113 		struct nfs_open_context *ctx)
3114 {
3115 	struct nfs4_state_owner *sp = opendata->owner;
3116 	struct nfs_server *server = sp->so_server;
3117 	struct dentry *dentry;
3118 	struct nfs4_state *state;
3119 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3120 	struct inode *dir = d_inode(opendata->dir);
3121 	unsigned long dir_verifier;
3122 	int ret;
3123 
3124 	dir_verifier = nfs_save_change_attribute(dir);
3125 
3126 	ret = _nfs4_proc_open(opendata, ctx);
3127 	if (ret != 0)
3128 		goto out;
3129 
3130 	state = _nfs4_opendata_to_nfs4_state(opendata);
3131 	ret = PTR_ERR(state);
3132 	if (IS_ERR(state))
3133 		goto out;
3134 	ctx->state = state;
3135 	if (server->caps & NFS_CAP_POSIX_LOCK)
3136 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3137 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3138 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3139 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3140 		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3141 
3142 	dentry = opendata->dentry;
3143 	if (d_really_is_negative(dentry)) {
3144 		struct dentry *alias;
3145 		d_drop(dentry);
3146 		alias = d_exact_alias(dentry, state->inode);
3147 		if (!alias)
3148 			alias = d_splice_alias(igrab(state->inode), dentry);
3149 		/* d_splice_alias() can't fail here - it's a non-directory */
3150 		if (alias) {
3151 			dput(ctx->dentry);
3152 			ctx->dentry = dentry = alias;
3153 		}
3154 	}
3155 
3156 	switch(opendata->o_arg.claim) {
3157 	default:
3158 		break;
3159 	case NFS4_OPEN_CLAIM_NULL:
3160 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3161 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3162 		if (!opendata->rpc_done)
3163 			break;
3164 		if (opendata->o_res.delegation.type != 0)
3165 			dir_verifier = nfs_save_change_attribute(dir);
3166 		nfs_set_verifier(dentry, dir_verifier);
3167 	}
3168 
3169 	/* Parse layoutget results before we check for access */
3170 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3171 
3172 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3173 	if (ret != 0)
3174 		goto out;
3175 
3176 	if (d_inode(dentry) == state->inode)
3177 		nfs_inode_attach_open_context(ctx);
3178 
3179 out:
3180 	if (!opendata->cancelled) {
3181 		if (opendata->lgp) {
3182 			nfs4_lgopen_release(opendata->lgp);
3183 			opendata->lgp = NULL;
3184 		}
3185 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3186 	}
3187 	return ret;
3188 }
3189 
3190 /*
3191  * Returns a referenced nfs4_state
3192  */
3193 static int _nfs4_do_open(struct inode *dir,
3194 			struct nfs_open_context *ctx,
3195 			int flags,
3196 			const struct nfs4_open_createattrs *c,
3197 			int *opened)
3198 {
3199 	struct nfs4_state_owner  *sp;
3200 	struct nfs4_state     *state = NULL;
3201 	struct nfs_server       *server = NFS_SERVER(dir);
3202 	struct nfs4_opendata *opendata;
3203 	struct dentry *dentry = ctx->dentry;
3204 	const struct cred *cred = ctx->cred;
3205 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3206 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3207 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3208 	struct iattr *sattr = c->sattr;
3209 	struct nfs4_label *label = c->label;
3210 	int status;
3211 
3212 	/* Protect against reboot recovery conflicts */
3213 	status = -ENOMEM;
3214 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3215 	if (sp == NULL) {
3216 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3217 		goto out_err;
3218 	}
3219 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3220 	if (status != 0)
3221 		goto err_put_state_owner;
3222 	if (d_really_is_positive(dentry))
3223 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3224 	status = -ENOMEM;
3225 	if (d_really_is_positive(dentry))
3226 		claim = NFS4_OPEN_CLAIM_FH;
3227 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3228 			c, claim, GFP_KERNEL);
3229 	if (opendata == NULL)
3230 		goto err_put_state_owner;
3231 
3232 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3233 		if (!opendata->f_attr.mdsthreshold) {
3234 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3235 			if (!opendata->f_attr.mdsthreshold)
3236 				goto err_opendata_put;
3237 		}
3238 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3239 	}
3240 	if (d_really_is_positive(dentry))
3241 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3242 
3243 	status = _nfs4_open_and_get_state(opendata, ctx);
3244 	if (status != 0)
3245 		goto err_opendata_put;
3246 	state = ctx->state;
3247 
3248 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3249 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3250 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3251 		/*
3252 		 * send create attributes which was not set by open
3253 		 * with an extra setattr.
3254 		 */
3255 		if (attrs || label) {
3256 			unsigned ia_old = sattr->ia_valid;
3257 
3258 			sattr->ia_valid = attrs;
3259 			nfs_fattr_init(opendata->o_res.f_attr);
3260 			status = nfs4_do_setattr(state->inode, cred,
3261 					opendata->o_res.f_attr, sattr,
3262 					ctx, label);
3263 			if (status == 0) {
3264 				nfs_setattr_update_inode(state->inode, sattr,
3265 						opendata->o_res.f_attr);
3266 				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3267 			}
3268 			sattr->ia_valid = ia_old;
3269 		}
3270 	}
3271 	if (opened && opendata->file_created)
3272 		*opened = 1;
3273 
3274 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3275 		*ctx_th = opendata->f_attr.mdsthreshold;
3276 		opendata->f_attr.mdsthreshold = NULL;
3277 	}
3278 
3279 	nfs4_opendata_put(opendata);
3280 	nfs4_put_state_owner(sp);
3281 	return 0;
3282 err_opendata_put:
3283 	nfs4_opendata_put(opendata);
3284 err_put_state_owner:
3285 	nfs4_put_state_owner(sp);
3286 out_err:
3287 	return status;
3288 }
3289 
3290 
3291 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3292 					struct nfs_open_context *ctx,
3293 					int flags,
3294 					struct iattr *sattr,
3295 					struct nfs4_label *label,
3296 					int *opened)
3297 {
3298 	struct nfs_server *server = NFS_SERVER(dir);
3299 	struct nfs4_exception exception = {
3300 		.interruptible = true,
3301 	};
3302 	struct nfs4_state *res;
3303 	struct nfs4_open_createattrs c = {
3304 		.label = label,
3305 		.sattr = sattr,
3306 		.verf = {
3307 			[0] = (__u32)jiffies,
3308 			[1] = (__u32)current->pid,
3309 		},
3310 	};
3311 	int status;
3312 
3313 	do {
3314 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3315 		res = ctx->state;
3316 		trace_nfs4_open_file(ctx, flags, status);
3317 		if (status == 0)
3318 			break;
3319 		/* NOTE: BAD_SEQID means the server and client disagree about the
3320 		 * book-keeping w.r.t. state-changing operations
3321 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3322 		 * It is actually a sign of a bug on the client or on the server.
3323 		 *
3324 		 * If we receive a BAD_SEQID error in the particular case of
3325 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3326 		 * have unhashed the old state_owner for us, and that we can
3327 		 * therefore safely retry using a new one. We should still warn
3328 		 * the user though...
3329 		 */
3330 		if (status == -NFS4ERR_BAD_SEQID) {
3331 			pr_warn_ratelimited("NFS: v4 server %s "
3332 					" returned a bad sequence-id error!\n",
3333 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3334 			exception.retry = 1;
3335 			continue;
3336 		}
3337 		/*
3338 		 * BAD_STATEID on OPEN means that the server cancelled our
3339 		 * state before it received the OPEN_CONFIRM.
3340 		 * Recover by retrying the request as per the discussion
3341 		 * on Page 181 of RFC3530.
3342 		 */
3343 		if (status == -NFS4ERR_BAD_STATEID) {
3344 			exception.retry = 1;
3345 			continue;
3346 		}
3347 		if (status == -NFS4ERR_EXPIRED) {
3348 			nfs4_schedule_lease_recovery(server->nfs_client);
3349 			exception.retry = 1;
3350 			continue;
3351 		}
3352 		if (status == -EAGAIN) {
3353 			/* We must have found a delegation */
3354 			exception.retry = 1;
3355 			continue;
3356 		}
3357 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3358 			continue;
3359 		res = ERR_PTR(nfs4_handle_exception(server,
3360 					status, &exception));
3361 	} while (exception.retry);
3362 	return res;
3363 }
3364 
3365 static int _nfs4_do_setattr(struct inode *inode,
3366 			    struct nfs_setattrargs *arg,
3367 			    struct nfs_setattrres *res,
3368 			    const struct cred *cred,
3369 			    struct nfs_open_context *ctx)
3370 {
3371 	struct nfs_server *server = NFS_SERVER(inode);
3372 	struct rpc_message msg = {
3373 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3374 		.rpc_argp	= arg,
3375 		.rpc_resp	= res,
3376 		.rpc_cred	= cred,
3377 	};
3378 	const struct cred *delegation_cred = NULL;
3379 	unsigned long timestamp = jiffies;
3380 	bool truncate;
3381 	int status;
3382 
3383 	nfs_fattr_init(res->fattr);
3384 
3385 	/* Servers should only apply open mode checks for file size changes */
3386 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3387 	if (!truncate) {
3388 		nfs4_inode_make_writeable(inode);
3389 		goto zero_stateid;
3390 	}
3391 
3392 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3393 		/* Use that stateid */
3394 	} else if (ctx != NULL && ctx->state) {
3395 		struct nfs_lock_context *l_ctx;
3396 		if (!nfs4_valid_open_stateid(ctx->state))
3397 			return -EBADF;
3398 		l_ctx = nfs_get_lock_context(ctx);
3399 		if (IS_ERR(l_ctx))
3400 			return PTR_ERR(l_ctx);
3401 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3402 						&arg->stateid, &delegation_cred);
3403 		nfs_put_lock_context(l_ctx);
3404 		if (status == -EIO)
3405 			return -EBADF;
3406 		else if (status == -EAGAIN)
3407 			goto zero_stateid;
3408 	} else {
3409 zero_stateid:
3410 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3411 	}
3412 	if (delegation_cred)
3413 		msg.rpc_cred = delegation_cred;
3414 
3415 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3416 
3417 	put_cred(delegation_cred);
3418 	if (status == 0 && ctx != NULL)
3419 		renew_lease(server, timestamp);
3420 	trace_nfs4_setattr(inode, &arg->stateid, status);
3421 	return status;
3422 }
3423 
3424 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3425 			   struct nfs_fattr *fattr, struct iattr *sattr,
3426 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3427 {
3428 	struct nfs_server *server = NFS_SERVER(inode);
3429 	__u32 bitmask[NFS4_BITMASK_SZ];
3430 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3431 	struct nfs_setattrargs	arg = {
3432 		.fh		= NFS_FH(inode),
3433 		.iap		= sattr,
3434 		.server		= server,
3435 		.bitmask = bitmask,
3436 		.label		= ilabel,
3437 	};
3438 	struct nfs_setattrres  res = {
3439 		.fattr		= fattr,
3440 		.server		= server,
3441 	};
3442 	struct nfs4_exception exception = {
3443 		.state = state,
3444 		.inode = inode,
3445 		.stateid = &arg.stateid,
3446 	};
3447 	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE |
3448 				     NFS_INO_INVALID_CTIME;
3449 	int err;
3450 
3451 	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3452 		adjust_flags |= NFS_INO_INVALID_MODE;
3453 	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3454 		adjust_flags |= NFS_INO_INVALID_OTHER;
3455 
3456 	do {
3457 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3458 					inode, adjust_flags);
3459 
3460 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3461 		switch (err) {
3462 		case -NFS4ERR_OPENMODE:
3463 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3464 				pr_warn_once("NFSv4: server %s is incorrectly "
3465 						"applying open mode checks to "
3466 						"a SETATTR that is not "
3467 						"changing file size.\n",
3468 						server->nfs_client->cl_hostname);
3469 			}
3470 			if (state && !(state->state & FMODE_WRITE)) {
3471 				err = -EBADF;
3472 				if (sattr->ia_valid & ATTR_OPEN)
3473 					err = -EACCES;
3474 				goto out;
3475 			}
3476 		}
3477 		err = nfs4_handle_exception(server, err, &exception);
3478 	} while (exception.retry);
3479 out:
3480 	return err;
3481 }
3482 
3483 static bool
3484 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3485 {
3486 	if (inode == NULL || !nfs_have_layout(inode))
3487 		return false;
3488 
3489 	return pnfs_wait_on_layoutreturn(inode, task);
3490 }
3491 
3492 /*
3493  * Update the seqid of an open stateid
3494  */
3495 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3496 		struct nfs4_state *state)
3497 {
3498 	__be32 seqid_open;
3499 	u32 dst_seqid;
3500 	int seq;
3501 
3502 	for (;;) {
3503 		if (!nfs4_valid_open_stateid(state))
3504 			break;
3505 		seq = read_seqbegin(&state->seqlock);
3506 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3507 			nfs4_stateid_copy(dst, &state->open_stateid);
3508 			if (read_seqretry(&state->seqlock, seq))
3509 				continue;
3510 			break;
3511 		}
3512 		seqid_open = state->open_stateid.seqid;
3513 		if (read_seqretry(&state->seqlock, seq))
3514 			continue;
3515 
3516 		dst_seqid = be32_to_cpu(dst->seqid);
3517 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3518 			dst->seqid = seqid_open;
3519 		break;
3520 	}
3521 }
3522 
3523 /*
3524  * Update the seqid of an open stateid after receiving
3525  * NFS4ERR_OLD_STATEID
3526  */
3527 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3528 		struct nfs4_state *state)
3529 {
3530 	__be32 seqid_open;
3531 	u32 dst_seqid;
3532 	bool ret;
3533 	int seq, status = -EAGAIN;
3534 	DEFINE_WAIT(wait);
3535 
3536 	for (;;) {
3537 		ret = false;
3538 		if (!nfs4_valid_open_stateid(state))
3539 			break;
3540 		seq = read_seqbegin(&state->seqlock);
3541 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3542 			if (read_seqretry(&state->seqlock, seq))
3543 				continue;
3544 			break;
3545 		}
3546 
3547 		write_seqlock(&state->seqlock);
3548 		seqid_open = state->open_stateid.seqid;
3549 
3550 		dst_seqid = be32_to_cpu(dst->seqid);
3551 
3552 		/* Did another OPEN bump the state's seqid?  try again: */
3553 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3554 			dst->seqid = seqid_open;
3555 			write_sequnlock(&state->seqlock);
3556 			ret = true;
3557 			break;
3558 		}
3559 
3560 		/* server says we're behind but we haven't seen the update yet */
3561 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3562 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3563 		write_sequnlock(&state->seqlock);
3564 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3565 
3566 		if (fatal_signal_pending(current))
3567 			status = -EINTR;
3568 		else
3569 			if (schedule_timeout(5*HZ) != 0)
3570 				status = 0;
3571 
3572 		finish_wait(&state->waitq, &wait);
3573 
3574 		if (!status)
3575 			continue;
3576 		if (status == -EINTR)
3577 			break;
3578 
3579 		/* we slept the whole 5 seconds, we must have lost a seqid */
3580 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3581 		ret = true;
3582 		break;
3583 	}
3584 
3585 	return ret;
3586 }
3587 
3588 struct nfs4_closedata {
3589 	struct inode *inode;
3590 	struct nfs4_state *state;
3591 	struct nfs_closeargs arg;
3592 	struct nfs_closeres res;
3593 	struct {
3594 		struct nfs4_layoutreturn_args arg;
3595 		struct nfs4_layoutreturn_res res;
3596 		struct nfs4_xdr_opaque_data ld_private;
3597 		u32 roc_barrier;
3598 		bool roc;
3599 	} lr;
3600 	struct nfs_fattr fattr;
3601 	unsigned long timestamp;
3602 };
3603 
3604 static void nfs4_free_closedata(void *data)
3605 {
3606 	struct nfs4_closedata *calldata = data;
3607 	struct nfs4_state_owner *sp = calldata->state->owner;
3608 	struct super_block *sb = calldata->state->inode->i_sb;
3609 
3610 	if (calldata->lr.roc)
3611 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3612 				calldata->res.lr_ret);
3613 	nfs4_put_open_state(calldata->state);
3614 	nfs_free_seqid(calldata->arg.seqid);
3615 	nfs4_put_state_owner(sp);
3616 	nfs_sb_deactive(sb);
3617 	kfree(calldata);
3618 }
3619 
3620 static void nfs4_close_done(struct rpc_task *task, void *data)
3621 {
3622 	struct nfs4_closedata *calldata = data;
3623 	struct nfs4_state *state = calldata->state;
3624 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3625 	nfs4_stateid *res_stateid = NULL;
3626 	struct nfs4_exception exception = {
3627 		.state = state,
3628 		.inode = calldata->inode,
3629 		.stateid = &calldata->arg.stateid,
3630 	};
3631 
3632 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3633 		return;
3634 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3635 
3636 	/* Handle Layoutreturn errors */
3637 	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3638 			  &calldata->res.lr_ret) == -EAGAIN)
3639 		goto out_restart;
3640 
3641 	/* hmm. we are done with the inode, and in the process of freeing
3642 	 * the state_owner. we keep this around to process errors
3643 	 */
3644 	switch (task->tk_status) {
3645 		case 0:
3646 			res_stateid = &calldata->res.stateid;
3647 			renew_lease(server, calldata->timestamp);
3648 			break;
3649 		case -NFS4ERR_ACCESS:
3650 			if (calldata->arg.bitmask != NULL) {
3651 				calldata->arg.bitmask = NULL;
3652 				calldata->res.fattr = NULL;
3653 				goto out_restart;
3654 
3655 			}
3656 			break;
3657 		case -NFS4ERR_OLD_STATEID:
3658 			/* Did we race with OPEN? */
3659 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3660 						state))
3661 				goto out_restart;
3662 			goto out_release;
3663 		case -NFS4ERR_ADMIN_REVOKED:
3664 		case -NFS4ERR_STALE_STATEID:
3665 		case -NFS4ERR_EXPIRED:
3666 			nfs4_free_revoked_stateid(server,
3667 					&calldata->arg.stateid,
3668 					task->tk_msg.rpc_cred);
3669 			fallthrough;
3670 		case -NFS4ERR_BAD_STATEID:
3671 			if (calldata->arg.fmode == 0)
3672 				break;
3673 			fallthrough;
3674 		default:
3675 			task->tk_status = nfs4_async_handle_exception(task,
3676 					server, task->tk_status, &exception);
3677 			if (exception.retry)
3678 				goto out_restart;
3679 	}
3680 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3681 			res_stateid, calldata->arg.fmode);
3682 out_release:
3683 	task->tk_status = 0;
3684 	nfs_release_seqid(calldata->arg.seqid);
3685 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3686 	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3687 	return;
3688 out_restart:
3689 	task->tk_status = 0;
3690 	rpc_restart_call_prepare(task);
3691 	goto out_release;
3692 }
3693 
3694 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3695 {
3696 	struct nfs4_closedata *calldata = data;
3697 	struct nfs4_state *state = calldata->state;
3698 	struct inode *inode = calldata->inode;
3699 	struct nfs_server *server = NFS_SERVER(inode);
3700 	struct pnfs_layout_hdr *lo;
3701 	bool is_rdonly, is_wronly, is_rdwr;
3702 	int call_close = 0;
3703 
3704 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3705 		goto out_wait;
3706 
3707 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3708 	spin_lock(&state->owner->so_lock);
3709 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3710 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3711 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3712 	/* Calculate the change in open mode */
3713 	calldata->arg.fmode = 0;
3714 	if (state->n_rdwr == 0) {
3715 		if (state->n_rdonly == 0)
3716 			call_close |= is_rdonly;
3717 		else if (is_rdonly)
3718 			calldata->arg.fmode |= FMODE_READ;
3719 		if (state->n_wronly == 0)
3720 			call_close |= is_wronly;
3721 		else if (is_wronly)
3722 			calldata->arg.fmode |= FMODE_WRITE;
3723 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3724 			call_close |= is_rdwr;
3725 	} else if (is_rdwr)
3726 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3727 
3728 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3729 	if (!nfs4_valid_open_stateid(state))
3730 		call_close = 0;
3731 	spin_unlock(&state->owner->so_lock);
3732 
3733 	if (!call_close) {
3734 		/* Note: exit _without_ calling nfs4_close_done */
3735 		goto out_no_action;
3736 	}
3737 
3738 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3739 		nfs_release_seqid(calldata->arg.seqid);
3740 		goto out_wait;
3741 	}
3742 
3743 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3744 	if (lo && !pnfs_layout_is_valid(lo)) {
3745 		calldata->arg.lr_args = NULL;
3746 		calldata->res.lr_res = NULL;
3747 	}
3748 
3749 	if (calldata->arg.fmode == 0)
3750 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3751 
3752 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3753 		/* Close-to-open cache consistency revalidation */
3754 		if (!nfs4_have_delegation(inode, FMODE_READ, 0)) {
3755 			nfs4_bitmask_set(calldata->arg.bitmask_store,
3756 					 server->cache_consistency_bitmask,
3757 					 inode, 0);
3758 			calldata->arg.bitmask = calldata->arg.bitmask_store;
3759 		} else
3760 			calldata->arg.bitmask = NULL;
3761 	}
3762 
3763 	calldata->arg.share_access =
3764 		nfs4_fmode_to_share_access(calldata->arg.fmode);
3765 
3766 	if (calldata->res.fattr == NULL)
3767 		calldata->arg.bitmask = NULL;
3768 	else if (calldata->arg.bitmask == NULL)
3769 		calldata->res.fattr = NULL;
3770 	calldata->timestamp = jiffies;
3771 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3772 				&calldata->arg.seq_args,
3773 				&calldata->res.seq_res,
3774 				task) != 0)
3775 		nfs_release_seqid(calldata->arg.seqid);
3776 	return;
3777 out_no_action:
3778 	task->tk_action = NULL;
3779 out_wait:
3780 	nfs4_sequence_done(task, &calldata->res.seq_res);
3781 }
3782 
3783 static const struct rpc_call_ops nfs4_close_ops = {
3784 	.rpc_call_prepare = nfs4_close_prepare,
3785 	.rpc_call_done = nfs4_close_done,
3786 	.rpc_release = nfs4_free_closedata,
3787 };
3788 
3789 /*
3790  * It is possible for data to be read/written from a mem-mapped file
3791  * after the sys_close call (which hits the vfs layer as a flush).
3792  * This means that we can't safely call nfsv4 close on a file until
3793  * the inode is cleared. This in turn means that we are not good
3794  * NFSv4 citizens - we do not indicate to the server to update the file's
3795  * share state even when we are done with one of the three share
3796  * stateid's in the inode.
3797  *
3798  * NOTE: Caller must be holding the sp->so_owner semaphore!
3799  */
3800 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3801 {
3802 	struct nfs_server *server = NFS_SERVER(state->inode);
3803 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3804 	struct nfs4_closedata *calldata;
3805 	struct nfs4_state_owner *sp = state->owner;
3806 	struct rpc_task *task;
3807 	struct rpc_message msg = {
3808 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3809 		.rpc_cred = state->owner->so_cred,
3810 	};
3811 	struct rpc_task_setup task_setup_data = {
3812 		.rpc_client = server->client,
3813 		.rpc_message = &msg,
3814 		.callback_ops = &nfs4_close_ops,
3815 		.workqueue = nfsiod_workqueue,
3816 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3817 	};
3818 	int status = -ENOMEM;
3819 
3820 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3821 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3822 
3823 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3824 		&task_setup_data.rpc_client, &msg);
3825 
3826 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3827 	if (calldata == NULL)
3828 		goto out;
3829 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3830 	calldata->inode = state->inode;
3831 	calldata->state = state;
3832 	calldata->arg.fh = NFS_FH(state->inode);
3833 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3834 		goto out_free_calldata;
3835 	/* Serialization for the sequence id */
3836 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3837 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3838 	if (IS_ERR(calldata->arg.seqid))
3839 		goto out_free_calldata;
3840 	nfs_fattr_init(&calldata->fattr);
3841 	calldata->arg.fmode = 0;
3842 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3843 	calldata->res.fattr = &calldata->fattr;
3844 	calldata->res.seqid = calldata->arg.seqid;
3845 	calldata->res.server = server;
3846 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3847 	calldata->lr.roc = pnfs_roc(state->inode,
3848 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3849 	if (calldata->lr.roc) {
3850 		calldata->arg.lr_args = &calldata->lr.arg;
3851 		calldata->res.lr_res = &calldata->lr.res;
3852 	}
3853 	nfs_sb_active(calldata->inode->i_sb);
3854 
3855 	msg.rpc_argp = &calldata->arg;
3856 	msg.rpc_resp = &calldata->res;
3857 	task_setup_data.callback_data = calldata;
3858 	task = rpc_run_task(&task_setup_data);
3859 	if (IS_ERR(task))
3860 		return PTR_ERR(task);
3861 	status = 0;
3862 	if (wait)
3863 		status = rpc_wait_for_completion_task(task);
3864 	rpc_put_task(task);
3865 	return status;
3866 out_free_calldata:
3867 	kfree(calldata);
3868 out:
3869 	nfs4_put_open_state(state);
3870 	nfs4_put_state_owner(sp);
3871 	return status;
3872 }
3873 
3874 static struct inode *
3875 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3876 		int open_flags, struct iattr *attr, int *opened)
3877 {
3878 	struct nfs4_state *state;
3879 	struct nfs4_label l, *label;
3880 
3881 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3882 
3883 	/* Protect against concurrent sillydeletes */
3884 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3885 
3886 	nfs4_label_release_security(label);
3887 
3888 	if (IS_ERR(state))
3889 		return ERR_CAST(state);
3890 	return state->inode;
3891 }
3892 
3893 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3894 {
3895 	if (ctx->state == NULL)
3896 		return;
3897 	if (is_sync)
3898 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3899 	else
3900 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3901 }
3902 
3903 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3904 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3905 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_OPEN_ARGUMENTS - 1UL)
3906 
3907 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3908 {
3909 	u32 minorversion = server->nfs_client->cl_minorversion;
3910 	u32 bitmask[3] = {
3911 		[0] = FATTR4_WORD0_SUPPORTED_ATTRS,
3912 	};
3913 	struct nfs4_server_caps_arg args = {
3914 		.fhandle = fhandle,
3915 		.bitmask = bitmask,
3916 	};
3917 	struct nfs4_server_caps_res res = {};
3918 	struct rpc_message msg = {
3919 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3920 		.rpc_argp = &args,
3921 		.rpc_resp = &res,
3922 	};
3923 	int status;
3924 	int i;
3925 
3926 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3927 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3928 		     FATTR4_WORD0_LINK_SUPPORT |
3929 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3930 		     FATTR4_WORD0_ACLSUPPORT |
3931 		     FATTR4_WORD0_CASE_INSENSITIVE |
3932 		     FATTR4_WORD0_CASE_PRESERVING;
3933 	if (minorversion)
3934 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3935 
3936 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3937 	if (status == 0) {
3938 		bitmask[0] = (FATTR4_WORD0_SUPPORTED_ATTRS |
3939 			      FATTR4_WORD0_FH_EXPIRE_TYPE |
3940 			      FATTR4_WORD0_LINK_SUPPORT |
3941 			      FATTR4_WORD0_SYMLINK_SUPPORT |
3942 			      FATTR4_WORD0_ACLSUPPORT |
3943 			      FATTR4_WORD0_CASE_INSENSITIVE |
3944 			      FATTR4_WORD0_CASE_PRESERVING) &
3945 			     res.attr_bitmask[0];
3946 		/* Sanity check the server answers */
3947 		switch (minorversion) {
3948 		case 0:
3949 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3950 			res.attr_bitmask[2] = 0;
3951 			break;
3952 		case 1:
3953 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3954 			bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT &
3955 				     res.attr_bitmask[2];
3956 			break;
3957 		case 2:
3958 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3959 			bitmask[2] = (FATTR4_WORD2_SUPPATTR_EXCLCREAT |
3960 				      FATTR4_WORD2_OPEN_ARGUMENTS) &
3961 				     res.attr_bitmask[2];
3962 		}
3963 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3964 		server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3965 				  NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3966 		server->fattr_valid = NFS_ATTR_FATTR_V4;
3967 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3968 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3969 			server->caps |= NFS_CAP_ACLS;
3970 		if (res.has_links != 0)
3971 			server->caps |= NFS_CAP_HARDLINKS;
3972 		if (res.has_symlinks != 0)
3973 			server->caps |= NFS_CAP_SYMLINKS;
3974 		if (res.case_insensitive)
3975 			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3976 		if (res.case_preserving)
3977 			server->caps |= NFS_CAP_CASE_PRESERVING;
3978 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3979 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3980 			server->caps |= NFS_CAP_SECURITY_LABEL;
3981 #endif
3982 		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3983 			server->caps |= NFS_CAP_FS_LOCATIONS;
3984 		if (res.attr_bitmask[2] & FATTR4_WORD2_TIME_DELEG_MODIFY)
3985 			server->caps |= NFS_CAP_DELEGTIME;
3986 		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3987 			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3988 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3989 			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3990 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3991 			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3992 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3993 			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3994 				NFS_ATTR_FATTR_OWNER_NAME);
3995 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3996 			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3997 				NFS_ATTR_FATTR_GROUP_NAME);
3998 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3999 			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
4000 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
4001 			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
4002 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
4003 			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
4004 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
4005 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
4006 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
4007 				sizeof(server->attr_bitmask));
4008 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
4009 
4010 		if (res.open_caps.oa_share_access_want[0] &
4011 		    NFS4_SHARE_WANT_OPEN_XOR_DELEGATION)
4012 			server->caps |= NFS_CAP_OPEN_XOR;
4013 
4014 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
4015 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
4016 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
4017 		server->cache_consistency_bitmask[2] = 0;
4018 
4019 		/* Avoid a regression due to buggy server */
4020 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
4021 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
4022 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
4023 			sizeof(server->exclcreat_bitmask));
4024 
4025 		server->acl_bitmask = res.acl_bitmask;
4026 		server->fh_expire_type = res.fh_expire_type;
4027 	}
4028 
4029 	return status;
4030 }
4031 
4032 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
4033 {
4034 	struct nfs4_exception exception = {
4035 		.interruptible = true,
4036 	};
4037 	int err;
4038 
4039 	nfs4_server_set_init_caps(server);
4040 	do {
4041 		err = nfs4_handle_exception(server,
4042 				_nfs4_server_capabilities(server, fhandle),
4043 				&exception);
4044 	} while (exception.retry);
4045 	return err;
4046 }
4047 
4048 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
4049 					  struct nfs_client *clp,
4050 					  struct nfs_server *server)
4051 {
4052 	int i;
4053 
4054 	for (i = 0; i < location->nservers; i++) {
4055 		struct nfs4_string *srv_loc = &location->servers[i];
4056 		struct sockaddr_storage addr;
4057 		size_t addrlen;
4058 		struct xprt_create xprt_args = {
4059 			.ident = 0,
4060 			.net = clp->cl_net,
4061 		};
4062 		struct nfs4_add_xprt_data xprtdata = {
4063 			.clp = clp,
4064 		};
4065 		struct rpc_add_xprt_test rpcdata = {
4066 			.add_xprt_test = clp->cl_mvops->session_trunk,
4067 			.data = &xprtdata,
4068 		};
4069 		char *servername = NULL;
4070 
4071 		if (!srv_loc->len)
4072 			continue;
4073 
4074 		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
4075 						&addr, sizeof(addr),
4076 						clp->cl_net, server->port);
4077 		if (!addrlen)
4078 			return;
4079 		xprt_args.dstaddr = (struct sockaddr *)&addr;
4080 		xprt_args.addrlen = addrlen;
4081 		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
4082 		if (!servername)
4083 			return;
4084 		memcpy(servername, srv_loc->data, srv_loc->len);
4085 		servername[srv_loc->len] = '\0';
4086 		xprt_args.servername = servername;
4087 
4088 		xprtdata.cred = nfs4_get_clid_cred(clp);
4089 		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
4090 				  rpc_clnt_setup_test_and_add_xprt,
4091 				  &rpcdata);
4092 		if (xprtdata.cred)
4093 			put_cred(xprtdata.cred);
4094 		kfree(servername);
4095 	}
4096 }
4097 
4098 static bool _is_same_nfs4_pathname(struct nfs4_pathname *path1,
4099 				   struct nfs4_pathname *path2)
4100 {
4101 	int i;
4102 
4103 	if (path1->ncomponents != path2->ncomponents)
4104 		return false;
4105 	for (i = 0; i < path1->ncomponents; i++) {
4106 		if (path1->components[i].len != path2->components[i].len)
4107 			return false;
4108 		if (memcmp(path1->components[i].data, path2->components[i].data,
4109 				path1->components[i].len))
4110 			return false;
4111 	}
4112 	return true;
4113 }
4114 
4115 static int _nfs4_discover_trunking(struct nfs_server *server,
4116 				   struct nfs_fh *fhandle)
4117 {
4118 	struct nfs4_fs_locations *locations = NULL;
4119 	struct page *page;
4120 	const struct cred *cred;
4121 	struct nfs_client *clp = server->nfs_client;
4122 	const struct nfs4_state_maintenance_ops *ops =
4123 		clp->cl_mvops->state_renewal_ops;
4124 	int status = -ENOMEM, i;
4125 
4126 	cred = ops->get_state_renewal_cred(clp);
4127 	if (cred == NULL) {
4128 		cred = nfs4_get_clid_cred(clp);
4129 		if (cred == NULL)
4130 			return -ENOKEY;
4131 	}
4132 
4133 	page = alloc_page(GFP_KERNEL);
4134 	if (!page)
4135 		goto out_put_cred;
4136 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4137 	if (!locations)
4138 		goto out_free;
4139 	locations->fattr = nfs_alloc_fattr();
4140 	if (!locations->fattr)
4141 		goto out_free_2;
4142 
4143 	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4144 					 cred);
4145 	if (status)
4146 		goto out_free_3;
4147 
4148 	for (i = 0; i < locations->nlocations; i++) {
4149 		if (!_is_same_nfs4_pathname(&locations->fs_path,
4150 					&locations->locations[i].rootpath))
4151 			continue;
4152 		test_fs_location_for_trunking(&locations->locations[i], clp,
4153 					      server);
4154 	}
4155 out_free_3:
4156 	kfree(locations->fattr);
4157 out_free_2:
4158 	kfree(locations);
4159 out_free:
4160 	__free_page(page);
4161 out_put_cred:
4162 	put_cred(cred);
4163 	return status;
4164 }
4165 
4166 static int nfs4_discover_trunking(struct nfs_server *server,
4167 				  struct nfs_fh *fhandle)
4168 {
4169 	struct nfs4_exception exception = {
4170 		.interruptible = true,
4171 	};
4172 	struct nfs_client *clp = server->nfs_client;
4173 	int err = 0;
4174 
4175 	if (!nfs4_has_session(clp))
4176 		goto out;
4177 	do {
4178 		err = nfs4_handle_exception(server,
4179 				_nfs4_discover_trunking(server, fhandle),
4180 				&exception);
4181 	} while (exception.retry);
4182 out:
4183 	return err;
4184 }
4185 
4186 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4187 		struct nfs_fsinfo *info)
4188 {
4189 	u32 bitmask[3];
4190 	struct nfs4_lookup_root_arg args = {
4191 		.bitmask = bitmask,
4192 	};
4193 	struct nfs4_lookup_res res = {
4194 		.server = server,
4195 		.fattr = info->fattr,
4196 		.fh = fhandle,
4197 	};
4198 	struct rpc_message msg = {
4199 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4200 		.rpc_argp = &args,
4201 		.rpc_resp = &res,
4202 	};
4203 
4204 	bitmask[0] = nfs4_fattr_bitmap[0];
4205 	bitmask[1] = nfs4_fattr_bitmap[1];
4206 	/*
4207 	 * Process the label in the upcoming getfattr
4208 	 */
4209 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4210 
4211 	nfs_fattr_init(info->fattr);
4212 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4213 }
4214 
4215 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4216 		struct nfs_fsinfo *info)
4217 {
4218 	struct nfs4_exception exception = {
4219 		.interruptible = true,
4220 	};
4221 	int err;
4222 	do {
4223 		err = _nfs4_lookup_root(server, fhandle, info);
4224 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4225 		switch (err) {
4226 		case 0:
4227 		case -NFS4ERR_WRONGSEC:
4228 			goto out;
4229 		default:
4230 			err = nfs4_handle_exception(server, err, &exception);
4231 		}
4232 	} while (exception.retry);
4233 out:
4234 	return err;
4235 }
4236 
4237 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4238 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4239 {
4240 	struct rpc_auth_create_args auth_args = {
4241 		.pseudoflavor = flavor,
4242 	};
4243 	struct rpc_auth *auth;
4244 
4245 	auth = rpcauth_create(&auth_args, server->client);
4246 	if (IS_ERR(auth))
4247 		return -EACCES;
4248 	return nfs4_lookup_root(server, fhandle, info);
4249 }
4250 
4251 /*
4252  * Retry pseudoroot lookup with various security flavors.  We do this when:
4253  *
4254  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4255  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4256  *
4257  * Returns zero on success, or a negative NFS4ERR value, or a
4258  * negative errno value.
4259  */
4260 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4261 			      struct nfs_fsinfo *info)
4262 {
4263 	/* Per 3530bis 15.33.5 */
4264 	static const rpc_authflavor_t flav_array[] = {
4265 		RPC_AUTH_GSS_KRB5P,
4266 		RPC_AUTH_GSS_KRB5I,
4267 		RPC_AUTH_GSS_KRB5,
4268 		RPC_AUTH_UNIX,			/* courtesy */
4269 		RPC_AUTH_NULL,
4270 	};
4271 	int status = -EPERM;
4272 	size_t i;
4273 
4274 	if (server->auth_info.flavor_len > 0) {
4275 		/* try each flavor specified by user */
4276 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4277 			status = nfs4_lookup_root_sec(server, fhandle, info,
4278 						server->auth_info.flavors[i]);
4279 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4280 				continue;
4281 			break;
4282 		}
4283 	} else {
4284 		/* no flavors specified by user, try default list */
4285 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4286 			status = nfs4_lookup_root_sec(server, fhandle, info,
4287 						      flav_array[i]);
4288 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4289 				continue;
4290 			break;
4291 		}
4292 	}
4293 
4294 	/*
4295 	 * -EACCES could mean that the user doesn't have correct permissions
4296 	 * to access the mount.  It could also mean that we tried to mount
4297 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4298 	 * existing mount programs don't handle -EACCES very well so it should
4299 	 * be mapped to -EPERM instead.
4300 	 */
4301 	if (status == -EACCES)
4302 		status = -EPERM;
4303 	return status;
4304 }
4305 
4306 /**
4307  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4308  * @server: initialized nfs_server handle
4309  * @fhandle: we fill in the pseudo-fs root file handle
4310  * @info: we fill in an FSINFO struct
4311  * @auth_probe: probe the auth flavours
4312  *
4313  * Returns zero on success, or a negative errno.
4314  */
4315 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4316 			 struct nfs_fsinfo *info,
4317 			 bool auth_probe)
4318 {
4319 	int status = 0;
4320 
4321 	if (!auth_probe)
4322 		status = nfs4_lookup_root(server, fhandle, info);
4323 
4324 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4325 		status = server->nfs_client->cl_mvops->find_root_sec(server,
4326 				fhandle, info);
4327 
4328 	if (status == 0)
4329 		status = nfs4_server_capabilities(server, fhandle);
4330 	if (status == 0)
4331 		status = nfs4_do_fsinfo(server, fhandle, info);
4332 
4333 	return nfs4_map_errors(status);
4334 }
4335 
4336 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4337 			      struct nfs_fsinfo *info)
4338 {
4339 	int error;
4340 	struct nfs_fattr *fattr = info->fattr;
4341 
4342 	error = nfs4_server_capabilities(server, mntfh);
4343 	if (error < 0) {
4344 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4345 		return error;
4346 	}
4347 
4348 	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4349 	if (error < 0) {
4350 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4351 		goto out;
4352 	}
4353 
4354 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4355 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4356 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4357 
4358 out:
4359 	return error;
4360 }
4361 
4362 /*
4363  * Get locations and (maybe) other attributes of a referral.
4364  * Note that we'll actually follow the referral later when
4365  * we detect fsid mismatch in inode revalidation
4366  */
4367 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4368 			     const struct qstr *name, struct nfs_fattr *fattr,
4369 			     struct nfs_fh *fhandle)
4370 {
4371 	int status = -ENOMEM;
4372 	struct page *page = NULL;
4373 	struct nfs4_fs_locations *locations = NULL;
4374 
4375 	page = alloc_page(GFP_KERNEL);
4376 	if (page == NULL)
4377 		goto out;
4378 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4379 	if (locations == NULL)
4380 		goto out;
4381 
4382 	locations->fattr = fattr;
4383 
4384 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4385 	if (status != 0)
4386 		goto out;
4387 
4388 	/*
4389 	 * If the fsid didn't change, this is a migration event, not a
4390 	 * referral.  Cause us to drop into the exception handler, which
4391 	 * will kick off migration recovery.
4392 	 */
4393 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4394 		dprintk("%s: server did not return a different fsid for"
4395 			" a referral at %s\n", __func__, name->name);
4396 		status = -NFS4ERR_MOVED;
4397 		goto out;
4398 	}
4399 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4400 	nfs_fixup_referral_attributes(fattr);
4401 	memset(fhandle, 0, sizeof(struct nfs_fh));
4402 out:
4403 	if (page)
4404 		__free_page(page);
4405 	kfree(locations);
4406 	return status;
4407 }
4408 
4409 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4410 				struct nfs_fattr *fattr, struct inode *inode)
4411 {
4412 	__u32 bitmask[NFS4_BITMASK_SZ];
4413 	struct nfs4_getattr_arg args = {
4414 		.fh = fhandle,
4415 		.bitmask = bitmask,
4416 	};
4417 	struct nfs4_getattr_res res = {
4418 		.fattr = fattr,
4419 		.server = server,
4420 	};
4421 	struct rpc_message msg = {
4422 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4423 		.rpc_argp = &args,
4424 		.rpc_resp = &res,
4425 	};
4426 	unsigned short task_flags = 0;
4427 
4428 	if (nfs4_has_session(server->nfs_client))
4429 		task_flags = RPC_TASK_MOVEABLE;
4430 
4431 	/* Is this is an attribute revalidation, subject to softreval? */
4432 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4433 		task_flags |= RPC_TASK_TIMEOUT;
4434 
4435 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4436 	nfs_fattr_init(fattr);
4437 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4438 	return nfs4_do_call_sync(server->client, server, &msg,
4439 			&args.seq_args, &res.seq_res, task_flags);
4440 }
4441 
4442 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4443 				struct nfs_fattr *fattr, struct inode *inode)
4444 {
4445 	struct nfs4_exception exception = {
4446 		.interruptible = true,
4447 	};
4448 	int err;
4449 	do {
4450 		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4451 		trace_nfs4_getattr(server, fhandle, fattr, err);
4452 		err = nfs4_handle_exception(server, err,
4453 				&exception);
4454 	} while (exception.retry);
4455 	return err;
4456 }
4457 
4458 /*
4459  * The file is not closed if it is opened due to the a request to change
4460  * the size of the file. The open call will not be needed once the
4461  * VFS layer lookup-intents are implemented.
4462  *
4463  * Close is called when the inode is destroyed.
4464  * If we haven't opened the file for O_WRONLY, we
4465  * need to in the size_change case to obtain a stateid.
4466  *
4467  * Got race?
4468  * Because OPEN is always done by name in nfsv4, it is
4469  * possible that we opened a different file by the same
4470  * name.  We can recognize this race condition, but we
4471  * can't do anything about it besides returning an error.
4472  *
4473  * This will be fixed with VFS changes (lookup-intent).
4474  */
4475 static int
4476 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4477 		  struct iattr *sattr)
4478 {
4479 	struct inode *inode = d_inode(dentry);
4480 	const struct cred *cred = NULL;
4481 	struct nfs_open_context *ctx = NULL;
4482 	int status;
4483 
4484 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4485 	    sattr->ia_valid & ATTR_SIZE &&
4486 	    sattr->ia_size < i_size_read(inode))
4487 		pnfs_commit_and_return_layout(inode);
4488 
4489 	nfs_fattr_init(fattr);
4490 
4491 	/* Deal with open(O_TRUNC) */
4492 	if (sattr->ia_valid & ATTR_OPEN)
4493 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4494 
4495 	/* Optimization: if the end result is no change, don't RPC */
4496 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4497 		return 0;
4498 
4499 	/* Search for an existing open(O_WRITE) file */
4500 	if (sattr->ia_valid & ATTR_FILE) {
4501 
4502 		ctx = nfs_file_open_context(sattr->ia_file);
4503 		if (ctx)
4504 			cred = ctx->cred;
4505 	}
4506 
4507 	/* Return any delegations if we're going to change ACLs */
4508 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4509 		nfs4_inode_make_writeable(inode);
4510 
4511 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4512 	if (status == 0) {
4513 		nfs_setattr_update_inode(inode, sattr, fattr);
4514 		nfs_setsecurity(inode, fattr);
4515 	}
4516 	return status;
4517 }
4518 
4519 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4520 		struct dentry *dentry, struct nfs_fh *fhandle,
4521 		struct nfs_fattr *fattr)
4522 {
4523 	struct nfs_server *server = NFS_SERVER(dir);
4524 	int		       status;
4525 	struct nfs4_lookup_arg args = {
4526 		.bitmask = server->attr_bitmask,
4527 		.dir_fh = NFS_FH(dir),
4528 		.name = &dentry->d_name,
4529 	};
4530 	struct nfs4_lookup_res res = {
4531 		.server = server,
4532 		.fattr = fattr,
4533 		.fh = fhandle,
4534 	};
4535 	struct rpc_message msg = {
4536 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4537 		.rpc_argp = &args,
4538 		.rpc_resp = &res,
4539 	};
4540 	unsigned short task_flags = 0;
4541 
4542 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4543 		task_flags = RPC_TASK_MOVEABLE;
4544 
4545 	/* Is this is an attribute revalidation, subject to softreval? */
4546 	if (nfs_lookup_is_soft_revalidate(dentry))
4547 		task_flags |= RPC_TASK_TIMEOUT;
4548 
4549 	args.bitmask = nfs4_bitmask(server, fattr->label);
4550 
4551 	nfs_fattr_init(fattr);
4552 
4553 	dprintk("NFS call  lookup %pd2\n", dentry);
4554 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4555 	status = nfs4_do_call_sync(clnt, server, &msg,
4556 			&args.seq_args, &res.seq_res, task_flags);
4557 	dprintk("NFS reply lookup: %d\n", status);
4558 	return status;
4559 }
4560 
4561 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4562 {
4563 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4564 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4565 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4566 	fattr->nlink = 2;
4567 }
4568 
4569 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4570 				   struct dentry *dentry, struct nfs_fh *fhandle,
4571 				   struct nfs_fattr *fattr)
4572 {
4573 	struct nfs4_exception exception = {
4574 		.interruptible = true,
4575 	};
4576 	struct rpc_clnt *client = *clnt;
4577 	const struct qstr *name = &dentry->d_name;
4578 	int err;
4579 	do {
4580 		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4581 		trace_nfs4_lookup(dir, name, err);
4582 		switch (err) {
4583 		case -NFS4ERR_BADNAME:
4584 			err = -ENOENT;
4585 			goto out;
4586 		case -NFS4ERR_MOVED:
4587 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4588 			if (err == -NFS4ERR_MOVED)
4589 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4590 			goto out;
4591 		case -NFS4ERR_WRONGSEC:
4592 			err = -EPERM;
4593 			if (client != *clnt)
4594 				goto out;
4595 			client = nfs4_negotiate_security(client, dir, name);
4596 			if (IS_ERR(client))
4597 				return PTR_ERR(client);
4598 
4599 			exception.retry = 1;
4600 			break;
4601 		default:
4602 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4603 		}
4604 	} while (exception.retry);
4605 
4606 out:
4607 	if (err == 0)
4608 		*clnt = client;
4609 	else if (client != *clnt)
4610 		rpc_shutdown_client(client);
4611 
4612 	return err;
4613 }
4614 
4615 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4616 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4617 {
4618 	int status;
4619 	struct rpc_clnt *client = NFS_CLIENT(dir);
4620 
4621 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4622 	if (client != NFS_CLIENT(dir)) {
4623 		rpc_shutdown_client(client);
4624 		nfs_fixup_secinfo_attributes(fattr);
4625 	}
4626 	return status;
4627 }
4628 
4629 struct rpc_clnt *
4630 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4631 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4632 {
4633 	struct rpc_clnt *client = NFS_CLIENT(dir);
4634 	int status;
4635 
4636 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4637 	if (status < 0)
4638 		return ERR_PTR(status);
4639 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4640 }
4641 
4642 static int _nfs4_proc_lookupp(struct inode *inode,
4643 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4644 {
4645 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4646 	struct nfs_server *server = NFS_SERVER(inode);
4647 	int		       status;
4648 	struct nfs4_lookupp_arg args = {
4649 		.bitmask = server->attr_bitmask,
4650 		.fh = NFS_FH(inode),
4651 	};
4652 	struct nfs4_lookupp_res res = {
4653 		.server = server,
4654 		.fattr = fattr,
4655 		.fh = fhandle,
4656 	};
4657 	struct rpc_message msg = {
4658 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4659 		.rpc_argp = &args,
4660 		.rpc_resp = &res,
4661 	};
4662 	unsigned short task_flags = 0;
4663 
4664 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4665 		task_flags |= RPC_TASK_TIMEOUT;
4666 
4667 	args.bitmask = nfs4_bitmask(server, fattr->label);
4668 
4669 	nfs_fattr_init(fattr);
4670 
4671 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4672 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4673 				&res.seq_res, task_flags);
4674 	dprintk("NFS reply lookupp: %d\n", status);
4675 	return status;
4676 }
4677 
4678 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4679 			     struct nfs_fattr *fattr)
4680 {
4681 	struct nfs4_exception exception = {
4682 		.interruptible = true,
4683 	};
4684 	int err;
4685 	do {
4686 		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4687 		trace_nfs4_lookupp(inode, err);
4688 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4689 				&exception);
4690 	} while (exception.retry);
4691 	return err;
4692 }
4693 
4694 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4695 			     const struct cred *cred)
4696 {
4697 	struct nfs_server *server = NFS_SERVER(inode);
4698 	struct nfs4_accessargs args = {
4699 		.fh = NFS_FH(inode),
4700 		.access = entry->mask,
4701 	};
4702 	struct nfs4_accessres res = {
4703 		.server = server,
4704 	};
4705 	struct rpc_message msg = {
4706 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4707 		.rpc_argp = &args,
4708 		.rpc_resp = &res,
4709 		.rpc_cred = cred,
4710 	};
4711 	int status = 0;
4712 
4713 	if (!nfs4_have_delegation(inode, FMODE_READ, 0)) {
4714 		res.fattr = nfs_alloc_fattr();
4715 		if (res.fattr == NULL)
4716 			return -ENOMEM;
4717 		args.bitmask = server->cache_consistency_bitmask;
4718 	}
4719 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4720 	if (!status) {
4721 		nfs_access_set_mask(entry, res.access);
4722 		if (res.fattr)
4723 			nfs_refresh_inode(inode, res.fattr);
4724 	}
4725 	nfs_free_fattr(res.fattr);
4726 	return status;
4727 }
4728 
4729 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4730 			    const struct cred *cred)
4731 {
4732 	struct nfs4_exception exception = {
4733 		.interruptible = true,
4734 	};
4735 	int err;
4736 	do {
4737 		err = _nfs4_proc_access(inode, entry, cred);
4738 		trace_nfs4_access(inode, err);
4739 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4740 				&exception);
4741 	} while (exception.retry);
4742 	return err;
4743 }
4744 
4745 /*
4746  * TODO: For the time being, we don't try to get any attributes
4747  * along with any of the zero-copy operations READ, READDIR,
4748  * READLINK, WRITE.
4749  *
4750  * In the case of the first three, we want to put the GETATTR
4751  * after the read-type operation -- this is because it is hard
4752  * to predict the length of a GETATTR response in v4, and thus
4753  * align the READ data correctly.  This means that the GETATTR
4754  * may end up partially falling into the page cache, and we should
4755  * shift it into the 'tail' of the xdr_buf before processing.
4756  * To do this efficiently, we need to know the total length
4757  * of data received, which doesn't seem to be available outside
4758  * of the RPC layer.
4759  *
4760  * In the case of WRITE, we also want to put the GETATTR after
4761  * the operation -- in this case because we want to make sure
4762  * we get the post-operation mtime and size.
4763  *
4764  * Both of these changes to the XDR layer would in fact be quite
4765  * minor, but I decided to leave them for a subsequent patch.
4766  */
4767 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4768 		unsigned int pgbase, unsigned int pglen)
4769 {
4770 	struct nfs4_readlink args = {
4771 		.fh       = NFS_FH(inode),
4772 		.pgbase	  = pgbase,
4773 		.pglen    = pglen,
4774 		.pages    = &page,
4775 	};
4776 	struct nfs4_readlink_res res;
4777 	struct rpc_message msg = {
4778 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4779 		.rpc_argp = &args,
4780 		.rpc_resp = &res,
4781 	};
4782 
4783 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4784 }
4785 
4786 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4787 		unsigned int pgbase, unsigned int pglen)
4788 {
4789 	struct nfs4_exception exception = {
4790 		.interruptible = true,
4791 	};
4792 	int err;
4793 	do {
4794 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4795 		trace_nfs4_readlink(inode, err);
4796 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4797 				&exception);
4798 	} while (exception.retry);
4799 	return err;
4800 }
4801 
4802 /*
4803  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4804  */
4805 static int
4806 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4807 		 int flags)
4808 {
4809 	struct nfs_server *server = NFS_SERVER(dir);
4810 	struct nfs4_label l, *ilabel;
4811 	struct nfs_open_context *ctx;
4812 	struct nfs4_state *state;
4813 	int status = 0;
4814 
4815 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4816 	if (IS_ERR(ctx))
4817 		return PTR_ERR(ctx);
4818 
4819 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4820 
4821 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4822 		sattr->ia_mode &= ~current_umask();
4823 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4824 	if (IS_ERR(state)) {
4825 		status = PTR_ERR(state);
4826 		goto out;
4827 	}
4828 out:
4829 	nfs4_label_release_security(ilabel);
4830 	put_nfs_open_context(ctx);
4831 	return status;
4832 }
4833 
4834 static int
4835 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4836 {
4837 	struct nfs_server *server = NFS_SERVER(dir);
4838 	struct nfs_removeargs args = {
4839 		.fh = NFS_FH(dir),
4840 		.name = *name,
4841 	};
4842 	struct nfs_removeres res = {
4843 		.server = server,
4844 	};
4845 	struct rpc_message msg = {
4846 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4847 		.rpc_argp = &args,
4848 		.rpc_resp = &res,
4849 	};
4850 	unsigned long timestamp = jiffies;
4851 	int status;
4852 
4853 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4854 	if (status == 0) {
4855 		spin_lock(&dir->i_lock);
4856 		/* Removing a directory decrements nlink in the parent */
4857 		if (ftype == NF4DIR && dir->i_nlink > 2)
4858 			nfs4_dec_nlink_locked(dir);
4859 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4860 					      NFS_INO_INVALID_DATA);
4861 		spin_unlock(&dir->i_lock);
4862 	}
4863 	return status;
4864 }
4865 
4866 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4867 {
4868 	struct nfs4_exception exception = {
4869 		.interruptible = true,
4870 	};
4871 	struct inode *inode = d_inode(dentry);
4872 	int err;
4873 
4874 	if (inode) {
4875 		if (inode->i_nlink == 1)
4876 			nfs4_inode_return_delegation(inode);
4877 		else
4878 			nfs4_inode_make_writeable(inode);
4879 	}
4880 	do {
4881 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4882 		trace_nfs4_remove(dir, &dentry->d_name, err);
4883 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4884 				&exception);
4885 	} while (exception.retry);
4886 	return err;
4887 }
4888 
4889 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4890 {
4891 	struct nfs4_exception exception = {
4892 		.interruptible = true,
4893 	};
4894 	int err;
4895 
4896 	do {
4897 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4898 		trace_nfs4_remove(dir, name, err);
4899 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4900 				&exception);
4901 	} while (exception.retry);
4902 	return err;
4903 }
4904 
4905 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4906 		struct dentry *dentry,
4907 		struct inode *inode)
4908 {
4909 	struct nfs_removeargs *args = msg->rpc_argp;
4910 	struct nfs_removeres *res = msg->rpc_resp;
4911 
4912 	res->server = NFS_SB(dentry->d_sb);
4913 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4914 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4915 
4916 	nfs_fattr_init(res->dir_attr);
4917 
4918 	if (inode) {
4919 		nfs4_inode_return_delegation(inode);
4920 		nfs_d_prune_case_insensitive_aliases(inode);
4921 	}
4922 }
4923 
4924 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4925 {
4926 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4927 			&data->args.seq_args,
4928 			&data->res.seq_res,
4929 			task);
4930 }
4931 
4932 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4933 {
4934 	struct nfs_unlinkdata *data = task->tk_calldata;
4935 	struct nfs_removeres *res = &data->res;
4936 
4937 	if (!nfs4_sequence_done(task, &res->seq_res))
4938 		return 0;
4939 	if (nfs4_async_handle_error(task, res->server, NULL,
4940 				    &data->timeout) == -EAGAIN)
4941 		return 0;
4942 	if (task->tk_status == 0)
4943 		nfs4_update_changeattr(dir, &res->cinfo,
4944 				res->dir_attr->time_start,
4945 				NFS_INO_INVALID_DATA);
4946 	return 1;
4947 }
4948 
4949 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4950 		struct dentry *old_dentry,
4951 		struct dentry *new_dentry)
4952 {
4953 	struct nfs_renameargs *arg = msg->rpc_argp;
4954 	struct nfs_renameres *res = msg->rpc_resp;
4955 	struct inode *old_inode = d_inode(old_dentry);
4956 	struct inode *new_inode = d_inode(new_dentry);
4957 
4958 	if (old_inode)
4959 		nfs4_inode_make_writeable(old_inode);
4960 	if (new_inode)
4961 		nfs4_inode_return_delegation(new_inode);
4962 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4963 	res->server = NFS_SB(old_dentry->d_sb);
4964 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4965 }
4966 
4967 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4968 {
4969 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4970 			&data->args.seq_args,
4971 			&data->res.seq_res,
4972 			task);
4973 }
4974 
4975 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4976 				 struct inode *new_dir)
4977 {
4978 	struct nfs_renamedata *data = task->tk_calldata;
4979 	struct nfs_renameres *res = &data->res;
4980 
4981 	if (!nfs4_sequence_done(task, &res->seq_res))
4982 		return 0;
4983 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4984 		return 0;
4985 
4986 	if (task->tk_status == 0) {
4987 		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4988 		if (new_dir != old_dir) {
4989 			/* Note: If we moved a directory, nlink will change */
4990 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4991 					res->old_fattr->time_start,
4992 					NFS_INO_INVALID_NLINK |
4993 					    NFS_INO_INVALID_DATA);
4994 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4995 					res->new_fattr->time_start,
4996 					NFS_INO_INVALID_NLINK |
4997 					    NFS_INO_INVALID_DATA);
4998 		} else
4999 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
5000 					res->old_fattr->time_start,
5001 					NFS_INO_INVALID_DATA);
5002 	}
5003 	return 1;
5004 }
5005 
5006 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
5007 {
5008 	struct nfs_server *server = NFS_SERVER(inode);
5009 	__u32 bitmask[NFS4_BITMASK_SZ];
5010 	struct nfs4_link_arg arg = {
5011 		.fh     = NFS_FH(inode),
5012 		.dir_fh = NFS_FH(dir),
5013 		.name   = name,
5014 		.bitmask = bitmask,
5015 	};
5016 	struct nfs4_link_res res = {
5017 		.server = server,
5018 	};
5019 	struct rpc_message msg = {
5020 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
5021 		.rpc_argp = &arg,
5022 		.rpc_resp = &res,
5023 	};
5024 	int status = -ENOMEM;
5025 
5026 	res.fattr = nfs_alloc_fattr_with_label(server);
5027 	if (res.fattr == NULL)
5028 		goto out;
5029 
5030 	nfs4_inode_make_writeable(inode);
5031 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label),
5032 				inode,
5033 				NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME);
5034 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5035 	if (!status) {
5036 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
5037 				       NFS_INO_INVALID_DATA);
5038 		nfs4_inc_nlink(inode);
5039 		status = nfs_post_op_update_inode(inode, res.fattr);
5040 		if (!status)
5041 			nfs_setsecurity(inode, res.fattr);
5042 	}
5043 
5044 out:
5045 	nfs_free_fattr(res.fattr);
5046 	return status;
5047 }
5048 
5049 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
5050 {
5051 	struct nfs4_exception exception = {
5052 		.interruptible = true,
5053 	};
5054 	int err;
5055 	do {
5056 		err = nfs4_handle_exception(NFS_SERVER(inode),
5057 				_nfs4_proc_link(inode, dir, name),
5058 				&exception);
5059 	} while (exception.retry);
5060 	return err;
5061 }
5062 
5063 struct nfs4_createdata {
5064 	struct rpc_message msg;
5065 	struct nfs4_create_arg arg;
5066 	struct nfs4_create_res res;
5067 	struct nfs_fh fh;
5068 	struct nfs_fattr fattr;
5069 };
5070 
5071 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
5072 		const struct qstr *name, struct iattr *sattr, u32 ftype)
5073 {
5074 	struct nfs4_createdata *data;
5075 
5076 	data = kzalloc(sizeof(*data), GFP_KERNEL);
5077 	if (data != NULL) {
5078 		struct nfs_server *server = NFS_SERVER(dir);
5079 
5080 		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
5081 		if (IS_ERR(data->fattr.label))
5082 			goto out_free;
5083 
5084 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
5085 		data->msg.rpc_argp = &data->arg;
5086 		data->msg.rpc_resp = &data->res;
5087 		data->arg.dir_fh = NFS_FH(dir);
5088 		data->arg.server = server;
5089 		data->arg.name = name;
5090 		data->arg.attrs = sattr;
5091 		data->arg.ftype = ftype;
5092 		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
5093 		data->arg.umask = current_umask();
5094 		data->res.server = server;
5095 		data->res.fh = &data->fh;
5096 		data->res.fattr = &data->fattr;
5097 		nfs_fattr_init(data->res.fattr);
5098 	}
5099 	return data;
5100 out_free:
5101 	kfree(data);
5102 	return NULL;
5103 }
5104 
5105 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
5106 {
5107 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
5108 				    &data->arg.seq_args, &data->res.seq_res, 1);
5109 	if (status == 0) {
5110 		spin_lock(&dir->i_lock);
5111 		/* Creating a directory bumps nlink in the parent */
5112 		if (data->arg.ftype == NF4DIR)
5113 			nfs4_inc_nlink_locked(dir);
5114 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
5115 					      data->res.fattr->time_start,
5116 					      NFS_INO_INVALID_DATA);
5117 		spin_unlock(&dir->i_lock);
5118 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
5119 	}
5120 	return status;
5121 }
5122 
5123 static void nfs4_free_createdata(struct nfs4_createdata *data)
5124 {
5125 	nfs4_label_free(data->fattr.label);
5126 	kfree(data);
5127 }
5128 
5129 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5130 		struct folio *folio, unsigned int len, struct iattr *sattr,
5131 		struct nfs4_label *label)
5132 {
5133 	struct page *page = &folio->page;
5134 	struct nfs4_createdata *data;
5135 	int status = -ENAMETOOLONG;
5136 
5137 	if (len > NFS4_MAXPATHLEN)
5138 		goto out;
5139 
5140 	status = -ENOMEM;
5141 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5142 	if (data == NULL)
5143 		goto out;
5144 
5145 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5146 	data->arg.u.symlink.pages = &page;
5147 	data->arg.u.symlink.len = len;
5148 	data->arg.label = label;
5149 
5150 	status = nfs4_do_create(dir, dentry, data);
5151 
5152 	nfs4_free_createdata(data);
5153 out:
5154 	return status;
5155 }
5156 
5157 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5158 		struct folio *folio, unsigned int len, struct iattr *sattr)
5159 {
5160 	struct nfs4_exception exception = {
5161 		.interruptible = true,
5162 	};
5163 	struct nfs4_label l, *label;
5164 	int err;
5165 
5166 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5167 
5168 	do {
5169 		err = _nfs4_proc_symlink(dir, dentry, folio, len, sattr, label);
5170 		trace_nfs4_symlink(dir, &dentry->d_name, err);
5171 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5172 				&exception);
5173 	} while (exception.retry);
5174 
5175 	nfs4_label_release_security(label);
5176 	return err;
5177 }
5178 
5179 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5180 		struct iattr *sattr, struct nfs4_label *label)
5181 {
5182 	struct nfs4_createdata *data;
5183 	int status = -ENOMEM;
5184 
5185 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5186 	if (data == NULL)
5187 		goto out;
5188 
5189 	data->arg.label = label;
5190 	status = nfs4_do_create(dir, dentry, data);
5191 
5192 	nfs4_free_createdata(data);
5193 out:
5194 	return status;
5195 }
5196 
5197 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5198 		struct iattr *sattr)
5199 {
5200 	struct nfs_server *server = NFS_SERVER(dir);
5201 	struct nfs4_exception exception = {
5202 		.interruptible = true,
5203 	};
5204 	struct nfs4_label l, *label;
5205 	int err;
5206 
5207 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5208 
5209 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5210 		sattr->ia_mode &= ~current_umask();
5211 	do {
5212 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5213 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5214 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5215 				&exception);
5216 	} while (exception.retry);
5217 	nfs4_label_release_security(label);
5218 
5219 	return err;
5220 }
5221 
5222 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5223 			      struct nfs_readdir_res *nr_res)
5224 {
5225 	struct inode		*dir = d_inode(nr_arg->dentry);
5226 	struct nfs_server	*server = NFS_SERVER(dir);
5227 	struct nfs4_readdir_arg args = {
5228 		.fh = NFS_FH(dir),
5229 		.pages = nr_arg->pages,
5230 		.pgbase = 0,
5231 		.count = nr_arg->page_len,
5232 		.plus = nr_arg->plus,
5233 	};
5234 	struct nfs4_readdir_res res;
5235 	struct rpc_message msg = {
5236 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5237 		.rpc_argp = &args,
5238 		.rpc_resp = &res,
5239 		.rpc_cred = nr_arg->cred,
5240 	};
5241 	int			status;
5242 
5243 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5244 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5245 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5246 		args.bitmask = server->attr_bitmask_nl;
5247 	else
5248 		args.bitmask = server->attr_bitmask;
5249 
5250 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5251 	res.pgbase = args.pgbase;
5252 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5253 			&res.seq_res, 0);
5254 	if (status >= 0) {
5255 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5256 		status += args.pgbase;
5257 	}
5258 
5259 	nfs_invalidate_atime(dir);
5260 
5261 	dprintk("%s: returns %d\n", __func__, status);
5262 	return status;
5263 }
5264 
5265 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5266 			     struct nfs_readdir_res *res)
5267 {
5268 	struct nfs4_exception exception = {
5269 		.interruptible = true,
5270 	};
5271 	int err;
5272 	do {
5273 		err = _nfs4_proc_readdir(arg, res);
5274 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5275 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5276 					    err, &exception);
5277 	} while (exception.retry);
5278 	return err;
5279 }
5280 
5281 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5282 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5283 {
5284 	struct nfs4_createdata *data;
5285 	int mode = sattr->ia_mode;
5286 	int status = -ENOMEM;
5287 
5288 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5289 	if (data == NULL)
5290 		goto out;
5291 
5292 	if (S_ISFIFO(mode))
5293 		data->arg.ftype = NF4FIFO;
5294 	else if (S_ISBLK(mode)) {
5295 		data->arg.ftype = NF4BLK;
5296 		data->arg.u.device.specdata1 = MAJOR(rdev);
5297 		data->arg.u.device.specdata2 = MINOR(rdev);
5298 	}
5299 	else if (S_ISCHR(mode)) {
5300 		data->arg.ftype = NF4CHR;
5301 		data->arg.u.device.specdata1 = MAJOR(rdev);
5302 		data->arg.u.device.specdata2 = MINOR(rdev);
5303 	} else if (!S_ISSOCK(mode)) {
5304 		status = -EINVAL;
5305 		goto out_free;
5306 	}
5307 
5308 	data->arg.label = label;
5309 	status = nfs4_do_create(dir, dentry, data);
5310 out_free:
5311 	nfs4_free_createdata(data);
5312 out:
5313 	return status;
5314 }
5315 
5316 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5317 		struct iattr *sattr, dev_t rdev)
5318 {
5319 	struct nfs_server *server = NFS_SERVER(dir);
5320 	struct nfs4_exception exception = {
5321 		.interruptible = true,
5322 	};
5323 	struct nfs4_label l, *label;
5324 	int err;
5325 
5326 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5327 
5328 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5329 		sattr->ia_mode &= ~current_umask();
5330 	do {
5331 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5332 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5333 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5334 				&exception);
5335 	} while (exception.retry);
5336 
5337 	nfs4_label_release_security(label);
5338 
5339 	return err;
5340 }
5341 
5342 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5343 		 struct nfs_fsstat *fsstat)
5344 {
5345 	struct nfs4_statfs_arg args = {
5346 		.fh = fhandle,
5347 		.bitmask = server->attr_bitmask,
5348 	};
5349 	struct nfs4_statfs_res res = {
5350 		.fsstat = fsstat,
5351 	};
5352 	struct rpc_message msg = {
5353 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5354 		.rpc_argp = &args,
5355 		.rpc_resp = &res,
5356 	};
5357 
5358 	nfs_fattr_init(fsstat->fattr);
5359 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5360 }
5361 
5362 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5363 {
5364 	struct nfs4_exception exception = {
5365 		.interruptible = true,
5366 	};
5367 	int err;
5368 	do {
5369 		err = nfs4_handle_exception(server,
5370 				_nfs4_proc_statfs(server, fhandle, fsstat),
5371 				&exception);
5372 	} while (exception.retry);
5373 	return err;
5374 }
5375 
5376 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5377 		struct nfs_fsinfo *fsinfo)
5378 {
5379 	struct nfs4_fsinfo_arg args = {
5380 		.fh = fhandle,
5381 		.bitmask = server->attr_bitmask,
5382 	};
5383 	struct nfs4_fsinfo_res res = {
5384 		.fsinfo = fsinfo,
5385 	};
5386 	struct rpc_message msg = {
5387 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5388 		.rpc_argp = &args,
5389 		.rpc_resp = &res,
5390 	};
5391 
5392 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5393 }
5394 
5395 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5396 {
5397 	struct nfs4_exception exception = {
5398 		.interruptible = true,
5399 	};
5400 	int err;
5401 
5402 	do {
5403 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5404 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5405 		if (err == 0) {
5406 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5407 			break;
5408 		}
5409 		err = nfs4_handle_exception(server, err, &exception);
5410 	} while (exception.retry);
5411 	return err;
5412 }
5413 
5414 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5415 {
5416 	int error;
5417 
5418 	nfs_fattr_init(fsinfo->fattr);
5419 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5420 	if (error == 0) {
5421 		/* block layout checks this! */
5422 		server->pnfs_blksize = fsinfo->blksize;
5423 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5424 	}
5425 
5426 	return error;
5427 }
5428 
5429 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5430 		struct nfs_pathconf *pathconf)
5431 {
5432 	struct nfs4_pathconf_arg args = {
5433 		.fh = fhandle,
5434 		.bitmask = server->attr_bitmask,
5435 	};
5436 	struct nfs4_pathconf_res res = {
5437 		.pathconf = pathconf,
5438 	};
5439 	struct rpc_message msg = {
5440 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5441 		.rpc_argp = &args,
5442 		.rpc_resp = &res,
5443 	};
5444 
5445 	/* None of the pathconf attributes are mandatory to implement */
5446 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5447 		memset(pathconf, 0, sizeof(*pathconf));
5448 		return 0;
5449 	}
5450 
5451 	nfs_fattr_init(pathconf->fattr);
5452 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5453 }
5454 
5455 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5456 		struct nfs_pathconf *pathconf)
5457 {
5458 	struct nfs4_exception exception = {
5459 		.interruptible = true,
5460 	};
5461 	int err;
5462 
5463 	do {
5464 		err = nfs4_handle_exception(server,
5465 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5466 				&exception);
5467 	} while (exception.retry);
5468 	return err;
5469 }
5470 
5471 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5472 		const struct nfs_open_context *ctx,
5473 		const struct nfs_lock_context *l_ctx,
5474 		fmode_t fmode)
5475 {
5476 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5477 }
5478 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5479 
5480 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5481 		const struct nfs_open_context *ctx,
5482 		const struct nfs_lock_context *l_ctx,
5483 		fmode_t fmode)
5484 {
5485 	nfs4_stateid _current_stateid;
5486 
5487 	/* If the current stateid represents a lost lock, then exit */
5488 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5489 		return true;
5490 	return nfs4_stateid_match(stateid, &_current_stateid);
5491 }
5492 
5493 static bool nfs4_error_stateid_expired(int err)
5494 {
5495 	switch (err) {
5496 	case -NFS4ERR_DELEG_REVOKED:
5497 	case -NFS4ERR_ADMIN_REVOKED:
5498 	case -NFS4ERR_BAD_STATEID:
5499 	case -NFS4ERR_STALE_STATEID:
5500 	case -NFS4ERR_OLD_STATEID:
5501 	case -NFS4ERR_OPENMODE:
5502 	case -NFS4ERR_EXPIRED:
5503 		return true;
5504 	}
5505 	return false;
5506 }
5507 
5508 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5509 {
5510 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5511 
5512 	trace_nfs4_read(hdr, task->tk_status);
5513 	if (task->tk_status < 0) {
5514 		struct nfs4_exception exception = {
5515 			.inode = hdr->inode,
5516 			.state = hdr->args.context->state,
5517 			.stateid = &hdr->args.stateid,
5518 		};
5519 		task->tk_status = nfs4_async_handle_exception(task,
5520 				server, task->tk_status, &exception);
5521 		if (exception.retry) {
5522 			rpc_restart_call_prepare(task);
5523 			return -EAGAIN;
5524 		}
5525 	}
5526 
5527 	if (task->tk_status > 0)
5528 		renew_lease(server, hdr->timestamp);
5529 	return 0;
5530 }
5531 
5532 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5533 		struct nfs_pgio_args *args)
5534 {
5535 
5536 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5537 		nfs4_stateid_is_current(&args->stateid,
5538 				args->context,
5539 				args->lock_context,
5540 				FMODE_READ))
5541 		return false;
5542 	rpc_restart_call_prepare(task);
5543 	return true;
5544 }
5545 
5546 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5547 					 struct nfs_pgio_header *hdr)
5548 {
5549 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5550 	struct rpc_message *msg = &task->tk_msg;
5551 
5552 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5553 	    task->tk_status == -ENOTSUPP) {
5554 		server->caps &= ~NFS_CAP_READ_PLUS;
5555 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5556 		rpc_restart_call_prepare(task);
5557 		return true;
5558 	}
5559 	return false;
5560 }
5561 
5562 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5563 {
5564 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5565 		return -EAGAIN;
5566 	if (nfs4_read_stateid_changed(task, &hdr->args))
5567 		return -EAGAIN;
5568 	if (nfs4_read_plus_not_supported(task, hdr))
5569 		return -EAGAIN;
5570 	if (task->tk_status > 0)
5571 		nfs_invalidate_atime(hdr->inode);
5572 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5573 				    nfs4_read_done_cb(task, hdr);
5574 }
5575 
5576 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5577 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5578 				    struct rpc_message *msg)
5579 {
5580 	/* Note: We don't use READ_PLUS with pNFS yet */
5581 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5582 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5583 		return nfs_read_alloc_scratch(hdr, READ_PLUS_SCRATCH_SIZE);
5584 	}
5585 	return false;
5586 }
5587 #else
5588 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5589 				    struct rpc_message *msg)
5590 {
5591 	return false;
5592 }
5593 #endif /* CONFIG_NFS_V4_2 */
5594 
5595 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5596 				 struct rpc_message *msg)
5597 {
5598 	hdr->timestamp   = jiffies;
5599 	if (!hdr->pgio_done_cb)
5600 		hdr->pgio_done_cb = nfs4_read_done_cb;
5601 	if (!nfs42_read_plus_support(hdr, msg))
5602 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5603 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5604 }
5605 
5606 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5607 				      struct nfs_pgio_header *hdr)
5608 {
5609 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5610 			&hdr->args.seq_args,
5611 			&hdr->res.seq_res,
5612 			task))
5613 		return 0;
5614 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5615 				hdr->args.lock_context,
5616 				hdr->rw_mode) == -EIO)
5617 		return -EIO;
5618 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5619 		return -EIO;
5620 	return 0;
5621 }
5622 
5623 static int nfs4_write_done_cb(struct rpc_task *task,
5624 			      struct nfs_pgio_header *hdr)
5625 {
5626 	struct inode *inode = hdr->inode;
5627 
5628 	trace_nfs4_write(hdr, task->tk_status);
5629 	if (task->tk_status < 0) {
5630 		struct nfs4_exception exception = {
5631 			.inode = hdr->inode,
5632 			.state = hdr->args.context->state,
5633 			.stateid = &hdr->args.stateid,
5634 		};
5635 		task->tk_status = nfs4_async_handle_exception(task,
5636 				NFS_SERVER(inode), task->tk_status,
5637 				&exception);
5638 		if (exception.retry) {
5639 			rpc_restart_call_prepare(task);
5640 			return -EAGAIN;
5641 		}
5642 	}
5643 	if (task->tk_status >= 0) {
5644 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5645 		nfs_writeback_update_inode(hdr);
5646 	}
5647 	return 0;
5648 }
5649 
5650 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5651 		struct nfs_pgio_args *args)
5652 {
5653 
5654 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5655 		nfs4_stateid_is_current(&args->stateid,
5656 				args->context,
5657 				args->lock_context,
5658 				FMODE_WRITE))
5659 		return false;
5660 	rpc_restart_call_prepare(task);
5661 	return true;
5662 }
5663 
5664 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5665 {
5666 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5667 		return -EAGAIN;
5668 	if (nfs4_write_stateid_changed(task, &hdr->args))
5669 		return -EAGAIN;
5670 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5671 		nfs4_write_done_cb(task, hdr);
5672 }
5673 
5674 static
5675 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5676 {
5677 	/* Don't request attributes for pNFS or O_DIRECT writes */
5678 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5679 		return false;
5680 	/* Otherwise, request attributes if and only if we don't hold
5681 	 * a delegation
5682 	 */
5683 	return nfs4_have_delegation(hdr->inode, FMODE_READ, 0) == 0;
5684 }
5685 
5686 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5687 		      struct inode *inode, unsigned long cache_validity)
5688 {
5689 	struct nfs_server *server = NFS_SERVER(inode);
5690 	unsigned int i;
5691 
5692 	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5693 	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5694 
5695 	if (cache_validity & NFS_INO_INVALID_CHANGE)
5696 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5697 	if (cache_validity & NFS_INO_INVALID_ATIME)
5698 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5699 	if (cache_validity & NFS_INO_INVALID_MODE)
5700 		bitmask[1] |= FATTR4_WORD1_MODE;
5701 	if (cache_validity & NFS_INO_INVALID_OTHER)
5702 		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5703 	if (cache_validity & NFS_INO_INVALID_NLINK)
5704 		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5705 	if (cache_validity & NFS_INO_INVALID_CTIME)
5706 		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5707 	if (cache_validity & NFS_INO_INVALID_MTIME)
5708 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5709 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5710 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5711 
5712 	if (cache_validity & NFS_INO_INVALID_SIZE)
5713 		bitmask[0] |= FATTR4_WORD0_SIZE;
5714 
5715 	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5716 		bitmask[i] &= server->attr_bitmask[i];
5717 }
5718 
5719 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5720 				  struct rpc_message *msg,
5721 				  struct rpc_clnt **clnt)
5722 {
5723 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5724 
5725 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5726 		hdr->args.bitmask = NULL;
5727 		hdr->res.fattr = NULL;
5728 	} else {
5729 		nfs4_bitmask_set(hdr->args.bitmask_store,
5730 				 server->cache_consistency_bitmask,
5731 				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5732 		hdr->args.bitmask = hdr->args.bitmask_store;
5733 	}
5734 
5735 	if (!hdr->pgio_done_cb)
5736 		hdr->pgio_done_cb = nfs4_write_done_cb;
5737 	hdr->res.server = server;
5738 	hdr->timestamp   = jiffies;
5739 
5740 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5741 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5742 	nfs4_state_protect_write(hdr->ds_clp ? hdr->ds_clp : server->nfs_client, clnt, msg, hdr);
5743 }
5744 
5745 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5746 {
5747 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5748 			&data->args.seq_args,
5749 			&data->res.seq_res,
5750 			task);
5751 }
5752 
5753 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5754 {
5755 	struct inode *inode = data->inode;
5756 
5757 	trace_nfs4_commit(data, task->tk_status);
5758 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5759 				    NULL, NULL) == -EAGAIN) {
5760 		rpc_restart_call_prepare(task);
5761 		return -EAGAIN;
5762 	}
5763 	return 0;
5764 }
5765 
5766 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5767 {
5768 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5769 		return -EAGAIN;
5770 	return data->commit_done_cb(task, data);
5771 }
5772 
5773 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5774 				   struct rpc_clnt **clnt)
5775 {
5776 	struct nfs_server *server = NFS_SERVER(data->inode);
5777 
5778 	if (data->commit_done_cb == NULL)
5779 		data->commit_done_cb = nfs4_commit_done_cb;
5780 	data->res.server = server;
5781 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5782 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5783 	nfs4_state_protect(data->ds_clp ? data->ds_clp : server->nfs_client,
5784 			NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5785 }
5786 
5787 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5788 				struct nfs_commitres *res)
5789 {
5790 	struct inode *dst_inode = file_inode(dst);
5791 	struct nfs_server *server = NFS_SERVER(dst_inode);
5792 	struct rpc_message msg = {
5793 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5794 		.rpc_argp = args,
5795 		.rpc_resp = res,
5796 	};
5797 
5798 	args->fh = NFS_FH(dst_inode);
5799 	return nfs4_call_sync(server->client, server, &msg,
5800 			&args->seq_args, &res->seq_res, 1);
5801 }
5802 
5803 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5804 {
5805 	struct nfs_commitargs args = {
5806 		.offset = offset,
5807 		.count = count,
5808 	};
5809 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5810 	struct nfs4_exception exception = { };
5811 	int status;
5812 
5813 	do {
5814 		status = _nfs4_proc_commit(dst, &args, res);
5815 		status = nfs4_handle_exception(dst_server, status, &exception);
5816 	} while (exception.retry);
5817 
5818 	return status;
5819 }
5820 
5821 struct nfs4_renewdata {
5822 	struct nfs_client	*client;
5823 	unsigned long		timestamp;
5824 };
5825 
5826 /*
5827  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5828  * standalone procedure for queueing an asynchronous RENEW.
5829  */
5830 static void nfs4_renew_release(void *calldata)
5831 {
5832 	struct nfs4_renewdata *data = calldata;
5833 	struct nfs_client *clp = data->client;
5834 
5835 	if (refcount_read(&clp->cl_count) > 1)
5836 		nfs4_schedule_state_renewal(clp);
5837 	nfs_put_client(clp);
5838 	kfree(data);
5839 }
5840 
5841 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5842 {
5843 	struct nfs4_renewdata *data = calldata;
5844 	struct nfs_client *clp = data->client;
5845 	unsigned long timestamp = data->timestamp;
5846 
5847 	trace_nfs4_renew_async(clp, task->tk_status);
5848 	switch (task->tk_status) {
5849 	case 0:
5850 		break;
5851 	case -NFS4ERR_LEASE_MOVED:
5852 		nfs4_schedule_lease_moved_recovery(clp);
5853 		break;
5854 	default:
5855 		/* Unless we're shutting down, schedule state recovery! */
5856 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5857 			return;
5858 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5859 			nfs4_schedule_lease_recovery(clp);
5860 			return;
5861 		}
5862 		nfs4_schedule_path_down_recovery(clp);
5863 	}
5864 	do_renew_lease(clp, timestamp);
5865 }
5866 
5867 static const struct rpc_call_ops nfs4_renew_ops = {
5868 	.rpc_call_done = nfs4_renew_done,
5869 	.rpc_release = nfs4_renew_release,
5870 };
5871 
5872 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5873 {
5874 	struct rpc_message msg = {
5875 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5876 		.rpc_argp	= clp,
5877 		.rpc_cred	= cred,
5878 	};
5879 	struct nfs4_renewdata *data;
5880 
5881 	if (renew_flags == 0)
5882 		return 0;
5883 	if (!refcount_inc_not_zero(&clp->cl_count))
5884 		return -EIO;
5885 	data = kmalloc(sizeof(*data), GFP_NOFS);
5886 	if (data == NULL) {
5887 		nfs_put_client(clp);
5888 		return -ENOMEM;
5889 	}
5890 	data->client = clp;
5891 	data->timestamp = jiffies;
5892 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5893 			&nfs4_renew_ops, data);
5894 }
5895 
5896 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5897 {
5898 	struct rpc_message msg = {
5899 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5900 		.rpc_argp	= clp,
5901 		.rpc_cred	= cred,
5902 	};
5903 	unsigned long now = jiffies;
5904 	int status;
5905 
5906 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5907 	if (status < 0)
5908 		return status;
5909 	do_renew_lease(clp, now);
5910 	return 0;
5911 }
5912 
5913 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5914 				      enum nfs4_acl_type type)
5915 {
5916 	switch (type) {
5917 	default:
5918 		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5919 	case NFS4ACL_DACL:
5920 		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5921 	case NFS4ACL_SACL:
5922 		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5923 	}
5924 }
5925 
5926 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5927  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5928  * the stack.
5929  */
5930 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5931 
5932 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5933 		struct page **pages)
5934 {
5935 	struct page *newpage, **spages;
5936 	int rc = 0;
5937 	size_t len;
5938 	spages = pages;
5939 
5940 	do {
5941 		len = min_t(size_t, PAGE_SIZE, buflen);
5942 		newpage = alloc_page(GFP_KERNEL);
5943 
5944 		if (newpage == NULL)
5945 			goto unwind;
5946 		memcpy(page_address(newpage), buf, len);
5947 		buf += len;
5948 		buflen -= len;
5949 		*pages++ = newpage;
5950 		rc++;
5951 	} while (buflen != 0);
5952 
5953 	return rc;
5954 
5955 unwind:
5956 	for(; rc > 0; rc--)
5957 		__free_page(spages[rc-1]);
5958 	return -ENOMEM;
5959 }
5960 
5961 struct nfs4_cached_acl {
5962 	enum nfs4_acl_type type;
5963 	int cached;
5964 	size_t len;
5965 	char data[];
5966 };
5967 
5968 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5969 {
5970 	struct nfs_inode *nfsi = NFS_I(inode);
5971 
5972 	spin_lock(&inode->i_lock);
5973 	kfree(nfsi->nfs4_acl);
5974 	nfsi->nfs4_acl = acl;
5975 	spin_unlock(&inode->i_lock);
5976 }
5977 
5978 static void nfs4_zap_acl_attr(struct inode *inode)
5979 {
5980 	nfs4_set_cached_acl(inode, NULL);
5981 }
5982 
5983 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5984 				    size_t buflen, enum nfs4_acl_type type)
5985 {
5986 	struct nfs_inode *nfsi = NFS_I(inode);
5987 	struct nfs4_cached_acl *acl;
5988 	int ret = -ENOENT;
5989 
5990 	spin_lock(&inode->i_lock);
5991 	acl = nfsi->nfs4_acl;
5992 	if (acl == NULL)
5993 		goto out;
5994 	if (acl->type != type)
5995 		goto out;
5996 	if (buf == NULL) /* user is just asking for length */
5997 		goto out_len;
5998 	if (acl->cached == 0)
5999 		goto out;
6000 	ret = -ERANGE; /* see getxattr(2) man page */
6001 	if (acl->len > buflen)
6002 		goto out;
6003 	memcpy(buf, acl->data, acl->len);
6004 out_len:
6005 	ret = acl->len;
6006 out:
6007 	spin_unlock(&inode->i_lock);
6008 	return ret;
6009 }
6010 
6011 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
6012 				  size_t pgbase, size_t acl_len,
6013 				  enum nfs4_acl_type type)
6014 {
6015 	struct nfs4_cached_acl *acl;
6016 	size_t buflen = sizeof(*acl) + acl_len;
6017 
6018 	if (buflen <= PAGE_SIZE) {
6019 		acl = kmalloc(buflen, GFP_KERNEL);
6020 		if (acl == NULL)
6021 			goto out;
6022 		acl->cached = 1;
6023 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
6024 	} else {
6025 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
6026 		if (acl == NULL)
6027 			goto out;
6028 		acl->cached = 0;
6029 	}
6030 	acl->type = type;
6031 	acl->len = acl_len;
6032 out:
6033 	nfs4_set_cached_acl(inode, acl);
6034 }
6035 
6036 /*
6037  * The getxattr API returns the required buffer length when called with a
6038  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
6039  * the required buf.  On a NULL buf, we send a page of data to the server
6040  * guessing that the ACL request can be serviced by a page. If so, we cache
6041  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
6042  * the cache. If not so, we throw away the page, and cache the required
6043  * length. The next getxattr call will then produce another round trip to
6044  * the server, this time with the input buf of the required size.
6045  */
6046 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
6047 				       size_t buflen, enum nfs4_acl_type type)
6048 {
6049 	struct page **pages;
6050 	struct nfs_getaclargs args = {
6051 		.fh = NFS_FH(inode),
6052 		.acl_type = type,
6053 		.acl_len = buflen,
6054 	};
6055 	struct nfs_getaclres res = {
6056 		.acl_type = type,
6057 		.acl_len = buflen,
6058 	};
6059 	struct rpc_message msg = {
6060 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
6061 		.rpc_argp = &args,
6062 		.rpc_resp = &res,
6063 	};
6064 	unsigned int npages;
6065 	int ret = -ENOMEM, i;
6066 	struct nfs_server *server = NFS_SERVER(inode);
6067 
6068 	if (buflen == 0)
6069 		buflen = server->rsize;
6070 
6071 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
6072 	pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
6073 	if (!pages)
6074 		return -ENOMEM;
6075 
6076 	args.acl_pages = pages;
6077 
6078 	for (i = 0; i < npages; i++) {
6079 		pages[i] = alloc_page(GFP_KERNEL);
6080 		if (!pages[i])
6081 			goto out_free;
6082 	}
6083 
6084 	/* for decoding across pages */
6085 	res.acl_scratch = alloc_page(GFP_KERNEL);
6086 	if (!res.acl_scratch)
6087 		goto out_free;
6088 
6089 	args.acl_len = npages * PAGE_SIZE;
6090 
6091 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
6092 		__func__, buf, buflen, npages, args.acl_len);
6093 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
6094 			     &msg, &args.seq_args, &res.seq_res, 0);
6095 	if (ret)
6096 		goto out_free;
6097 
6098 	/* Handle the case where the passed-in buffer is too short */
6099 	if (res.acl_flags & NFS4_ACL_TRUNC) {
6100 		/* Did the user only issue a request for the acl length? */
6101 		if (buf == NULL)
6102 			goto out_ok;
6103 		ret = -ERANGE;
6104 		goto out_free;
6105 	}
6106 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
6107 			      type);
6108 	if (buf) {
6109 		if (res.acl_len > buflen) {
6110 			ret = -ERANGE;
6111 			goto out_free;
6112 		}
6113 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
6114 	}
6115 out_ok:
6116 	ret = res.acl_len;
6117 out_free:
6118 	while (--i >= 0)
6119 		__free_page(pages[i]);
6120 	if (res.acl_scratch)
6121 		__free_page(res.acl_scratch);
6122 	kfree(pages);
6123 	return ret;
6124 }
6125 
6126 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6127 				     size_t buflen, enum nfs4_acl_type type)
6128 {
6129 	struct nfs4_exception exception = {
6130 		.interruptible = true,
6131 	};
6132 	ssize_t ret;
6133 	do {
6134 		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6135 		trace_nfs4_get_acl(inode, ret);
6136 		if (ret >= 0)
6137 			break;
6138 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6139 	} while (exception.retry);
6140 	return ret;
6141 }
6142 
6143 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6144 				 enum nfs4_acl_type type)
6145 {
6146 	struct nfs_server *server = NFS_SERVER(inode);
6147 	int ret;
6148 
6149 	if (!nfs4_server_supports_acls(server, type))
6150 		return -EOPNOTSUPP;
6151 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6152 	if (ret < 0)
6153 		return ret;
6154 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6155 		nfs_zap_acl_cache(inode);
6156 	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6157 	if (ret != -ENOENT)
6158 		/* -ENOENT is returned if there is no ACL or if there is an ACL
6159 		 * but no cached acl data, just the acl length */
6160 		return ret;
6161 	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6162 }
6163 
6164 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6165 			       size_t buflen, enum nfs4_acl_type type)
6166 {
6167 	struct nfs_server *server = NFS_SERVER(inode);
6168 	struct page *pages[NFS4ACL_MAXPAGES];
6169 	struct nfs_setaclargs arg = {
6170 		.fh = NFS_FH(inode),
6171 		.acl_type = type,
6172 		.acl_len = buflen,
6173 		.acl_pages = pages,
6174 	};
6175 	struct nfs_setaclres res;
6176 	struct rpc_message msg = {
6177 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6178 		.rpc_argp	= &arg,
6179 		.rpc_resp	= &res,
6180 	};
6181 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6182 	int ret, i;
6183 
6184 	/* You can't remove system.nfs4_acl: */
6185 	if (buflen == 0)
6186 		return -EINVAL;
6187 	if (!nfs4_server_supports_acls(server, type))
6188 		return -EOPNOTSUPP;
6189 	if (npages > ARRAY_SIZE(pages))
6190 		return -ERANGE;
6191 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6192 	if (i < 0)
6193 		return i;
6194 	nfs4_inode_make_writeable(inode);
6195 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6196 
6197 	/*
6198 	 * Free each page after tx, so the only ref left is
6199 	 * held by the network stack
6200 	 */
6201 	for (; i > 0; i--)
6202 		put_page(pages[i-1]);
6203 
6204 	/*
6205 	 * Acl update can result in inode attribute update.
6206 	 * so mark the attribute cache invalid.
6207 	 */
6208 	spin_lock(&inode->i_lock);
6209 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6210 					     NFS_INO_INVALID_CTIME |
6211 					     NFS_INO_REVAL_FORCED);
6212 	spin_unlock(&inode->i_lock);
6213 	nfs_access_zap_cache(inode);
6214 	nfs_zap_acl_cache(inode);
6215 	return ret;
6216 }
6217 
6218 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6219 			     size_t buflen, enum nfs4_acl_type type)
6220 {
6221 	struct nfs4_exception exception = { };
6222 	int err;
6223 	do {
6224 		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6225 		trace_nfs4_set_acl(inode, err);
6226 		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6227 			/*
6228 			 * no need to retry since the kernel
6229 			 * isn't involved in encoding the ACEs.
6230 			 */
6231 			err = -EINVAL;
6232 			break;
6233 		}
6234 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6235 				&exception);
6236 	} while (exception.retry);
6237 	return err;
6238 }
6239 
6240 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6241 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6242 					size_t buflen)
6243 {
6244 	struct nfs_server *server = NFS_SERVER(inode);
6245 	struct nfs4_label label = {0, 0, buflen, buf};
6246 
6247 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6248 	struct nfs_fattr fattr = {
6249 		.label = &label,
6250 	};
6251 	struct nfs4_getattr_arg arg = {
6252 		.fh		= NFS_FH(inode),
6253 		.bitmask	= bitmask,
6254 	};
6255 	struct nfs4_getattr_res res = {
6256 		.fattr		= &fattr,
6257 		.server		= server,
6258 	};
6259 	struct rpc_message msg = {
6260 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6261 		.rpc_argp	= &arg,
6262 		.rpc_resp	= &res,
6263 	};
6264 	int ret;
6265 
6266 	nfs_fattr_init(&fattr);
6267 
6268 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6269 	if (ret)
6270 		return ret;
6271 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6272 		return -ENOENT;
6273 	return label.len;
6274 }
6275 
6276 static int nfs4_get_security_label(struct inode *inode, void *buf,
6277 					size_t buflen)
6278 {
6279 	struct nfs4_exception exception = {
6280 		.interruptible = true,
6281 	};
6282 	int err;
6283 
6284 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6285 		return -EOPNOTSUPP;
6286 
6287 	do {
6288 		err = _nfs4_get_security_label(inode, buf, buflen);
6289 		trace_nfs4_get_security_label(inode, err);
6290 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6291 				&exception);
6292 	} while (exception.retry);
6293 	return err;
6294 }
6295 
6296 static int _nfs4_do_set_security_label(struct inode *inode,
6297 		struct nfs4_label *ilabel,
6298 		struct nfs_fattr *fattr)
6299 {
6300 
6301 	struct iattr sattr = {0};
6302 	struct nfs_server *server = NFS_SERVER(inode);
6303 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6304 	struct nfs_setattrargs arg = {
6305 		.fh		= NFS_FH(inode),
6306 		.iap		= &sattr,
6307 		.server		= server,
6308 		.bitmask	= bitmask,
6309 		.label		= ilabel,
6310 	};
6311 	struct nfs_setattrres res = {
6312 		.fattr		= fattr,
6313 		.server		= server,
6314 	};
6315 	struct rpc_message msg = {
6316 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6317 		.rpc_argp	= &arg,
6318 		.rpc_resp	= &res,
6319 	};
6320 	int status;
6321 
6322 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6323 
6324 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6325 	if (status)
6326 		dprintk("%s failed: %d\n", __func__, status);
6327 
6328 	return status;
6329 }
6330 
6331 static int nfs4_do_set_security_label(struct inode *inode,
6332 		struct nfs4_label *ilabel,
6333 		struct nfs_fattr *fattr)
6334 {
6335 	struct nfs4_exception exception = { };
6336 	int err;
6337 
6338 	do {
6339 		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6340 		trace_nfs4_set_security_label(inode, err);
6341 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6342 				&exception);
6343 	} while (exception.retry);
6344 	return err;
6345 }
6346 
6347 static int
6348 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6349 {
6350 	struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6351 	struct nfs_fattr *fattr;
6352 	int status;
6353 
6354 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6355 		return -EOPNOTSUPP;
6356 
6357 	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6358 	if (fattr == NULL)
6359 		return -ENOMEM;
6360 
6361 	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6362 	if (status == 0)
6363 		nfs_setsecurity(inode, fattr);
6364 
6365 	nfs_free_fattr(fattr);
6366 	return status;
6367 }
6368 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6369 
6370 
6371 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6372 				    nfs4_verifier *bootverf)
6373 {
6374 	__be32 verf[2];
6375 
6376 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6377 		/* An impossible timestamp guarantees this value
6378 		 * will never match a generated boot time. */
6379 		verf[0] = cpu_to_be32(U32_MAX);
6380 		verf[1] = cpu_to_be32(U32_MAX);
6381 	} else {
6382 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6383 		u64 ns = ktime_to_ns(nn->boot_time);
6384 
6385 		verf[0] = cpu_to_be32(ns >> 32);
6386 		verf[1] = cpu_to_be32(ns);
6387 	}
6388 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6389 }
6390 
6391 static size_t
6392 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6393 {
6394 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6395 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6396 	const char *id;
6397 
6398 	buf[0] = '\0';
6399 
6400 	if (nn_clp) {
6401 		rcu_read_lock();
6402 		id = rcu_dereference(nn_clp->identifier);
6403 		if (id)
6404 			strscpy(buf, id, buflen);
6405 		rcu_read_unlock();
6406 	}
6407 
6408 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6409 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6410 
6411 	return strlen(buf);
6412 }
6413 
6414 static int
6415 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6416 {
6417 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6418 	size_t buflen;
6419 	size_t len;
6420 	char *str;
6421 
6422 	if (clp->cl_owner_id != NULL)
6423 		return 0;
6424 
6425 	rcu_read_lock();
6426 	len = 14 +
6427 		strlen(clp->cl_rpcclient->cl_nodename) +
6428 		1 +
6429 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6430 		1;
6431 	rcu_read_unlock();
6432 
6433 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6434 	if (buflen)
6435 		len += buflen + 1;
6436 
6437 	if (len > NFS4_OPAQUE_LIMIT + 1)
6438 		return -EINVAL;
6439 
6440 	/*
6441 	 * Since this string is allocated at mount time, and held until the
6442 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6443 	 * about a memory-reclaim deadlock.
6444 	 */
6445 	str = kmalloc(len, GFP_KERNEL);
6446 	if (!str)
6447 		return -ENOMEM;
6448 
6449 	rcu_read_lock();
6450 	if (buflen)
6451 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6452 			  clp->cl_rpcclient->cl_nodename, buf,
6453 			  rpc_peeraddr2str(clp->cl_rpcclient,
6454 					   RPC_DISPLAY_ADDR));
6455 	else
6456 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6457 			  clp->cl_rpcclient->cl_nodename,
6458 			  rpc_peeraddr2str(clp->cl_rpcclient,
6459 					   RPC_DISPLAY_ADDR));
6460 	rcu_read_unlock();
6461 
6462 	clp->cl_owner_id = str;
6463 	return 0;
6464 }
6465 
6466 static int
6467 nfs4_init_uniform_client_string(struct nfs_client *clp)
6468 {
6469 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6470 	size_t buflen;
6471 	size_t len;
6472 	char *str;
6473 
6474 	if (clp->cl_owner_id != NULL)
6475 		return 0;
6476 
6477 	len = 10 + 10 + 1 + 10 + 1 +
6478 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6479 
6480 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6481 	if (buflen)
6482 		len += buflen + 1;
6483 
6484 	if (len > NFS4_OPAQUE_LIMIT + 1)
6485 		return -EINVAL;
6486 
6487 	/*
6488 	 * Since this string is allocated at mount time, and held until the
6489 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6490 	 * about a memory-reclaim deadlock.
6491 	 */
6492 	str = kmalloc(len, GFP_KERNEL);
6493 	if (!str)
6494 		return -ENOMEM;
6495 
6496 	if (buflen)
6497 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6498 			  clp->rpc_ops->version, clp->cl_minorversion,
6499 			  buf, clp->cl_rpcclient->cl_nodename);
6500 	else
6501 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6502 			  clp->rpc_ops->version, clp->cl_minorversion,
6503 			  clp->cl_rpcclient->cl_nodename);
6504 	clp->cl_owner_id = str;
6505 	return 0;
6506 }
6507 
6508 /*
6509  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6510  * services.  Advertise one based on the address family of the
6511  * clientaddr.
6512  */
6513 static unsigned int
6514 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6515 {
6516 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6517 		return scnprintf(buf, len, "tcp6");
6518 	else
6519 		return scnprintf(buf, len, "tcp");
6520 }
6521 
6522 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6523 {
6524 	struct nfs4_setclientid *sc = calldata;
6525 
6526 	if (task->tk_status == 0)
6527 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6528 }
6529 
6530 static const struct rpc_call_ops nfs4_setclientid_ops = {
6531 	.rpc_call_done = nfs4_setclientid_done,
6532 };
6533 
6534 /**
6535  * nfs4_proc_setclientid - Negotiate client ID
6536  * @clp: state data structure
6537  * @program: RPC program for NFSv4 callback service
6538  * @port: IP port number for NFS4 callback service
6539  * @cred: credential to use for this call
6540  * @res: where to place the result
6541  *
6542  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6543  */
6544 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6545 		unsigned short port, const struct cred *cred,
6546 		struct nfs4_setclientid_res *res)
6547 {
6548 	nfs4_verifier sc_verifier;
6549 	struct nfs4_setclientid setclientid = {
6550 		.sc_verifier = &sc_verifier,
6551 		.sc_prog = program,
6552 		.sc_clnt = clp,
6553 	};
6554 	struct rpc_message msg = {
6555 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6556 		.rpc_argp = &setclientid,
6557 		.rpc_resp = res,
6558 		.rpc_cred = cred,
6559 	};
6560 	struct rpc_task_setup task_setup_data = {
6561 		.rpc_client = clp->cl_rpcclient,
6562 		.rpc_message = &msg,
6563 		.callback_ops = &nfs4_setclientid_ops,
6564 		.callback_data = &setclientid,
6565 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6566 	};
6567 	unsigned long now = jiffies;
6568 	int status;
6569 
6570 	/* nfs_client_id4 */
6571 	nfs4_init_boot_verifier(clp, &sc_verifier);
6572 
6573 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6574 		status = nfs4_init_uniform_client_string(clp);
6575 	else
6576 		status = nfs4_init_nonuniform_client_string(clp);
6577 
6578 	if (status)
6579 		goto out;
6580 
6581 	/* cb_client4 */
6582 	setclientid.sc_netid_len =
6583 				nfs4_init_callback_netid(clp,
6584 						setclientid.sc_netid,
6585 						sizeof(setclientid.sc_netid));
6586 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6587 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6588 				clp->cl_ipaddr, port >> 8, port & 255);
6589 
6590 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6591 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6592 		clp->cl_owner_id);
6593 
6594 	status = nfs4_call_sync_custom(&task_setup_data);
6595 	if (setclientid.sc_cred) {
6596 		kfree(clp->cl_acceptor);
6597 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6598 		put_rpccred(setclientid.sc_cred);
6599 	}
6600 
6601 	if (status == 0)
6602 		do_renew_lease(clp, now);
6603 out:
6604 	trace_nfs4_setclientid(clp, status);
6605 	dprintk("NFS reply setclientid: %d\n", status);
6606 	return status;
6607 }
6608 
6609 /**
6610  * nfs4_proc_setclientid_confirm - Confirm client ID
6611  * @clp: state data structure
6612  * @arg: result of a previous SETCLIENTID
6613  * @cred: credential to use for this call
6614  *
6615  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6616  */
6617 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6618 		struct nfs4_setclientid_res *arg,
6619 		const struct cred *cred)
6620 {
6621 	struct rpc_message msg = {
6622 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6623 		.rpc_argp = arg,
6624 		.rpc_cred = cred,
6625 	};
6626 	int status;
6627 
6628 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6629 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6630 		clp->cl_clientid);
6631 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6632 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6633 	trace_nfs4_setclientid_confirm(clp, status);
6634 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6635 	return status;
6636 }
6637 
6638 struct nfs4_delegreturndata {
6639 	struct nfs4_delegreturnargs args;
6640 	struct nfs4_delegreturnres res;
6641 	struct nfs_fh fh;
6642 	nfs4_stateid stateid;
6643 	unsigned long timestamp;
6644 	struct {
6645 		struct nfs4_layoutreturn_args arg;
6646 		struct nfs4_layoutreturn_res res;
6647 		struct nfs4_xdr_opaque_data ld_private;
6648 		u32 roc_barrier;
6649 		bool roc;
6650 	} lr;
6651 	struct nfs4_delegattr sattr;
6652 	struct nfs_fattr fattr;
6653 	int rpc_status;
6654 	struct inode *inode;
6655 };
6656 
6657 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6658 {
6659 	struct nfs4_delegreturndata *data = calldata;
6660 	struct nfs4_exception exception = {
6661 		.inode = data->inode,
6662 		.stateid = &data->stateid,
6663 		.task_is_privileged = data->args.seq_args.sa_privileged,
6664 	};
6665 
6666 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6667 		return;
6668 
6669 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6670 
6671 	/* Handle Layoutreturn errors */
6672 	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6673 			  &data->res.lr_ret) == -EAGAIN)
6674 		goto out_restart;
6675 
6676 	if (data->args.sattr_args && task->tk_status != 0) {
6677 		switch(data->res.sattr_ret) {
6678 		case 0:
6679 			data->args.sattr_args = NULL;
6680 			data->res.sattr_res = false;
6681 			break;
6682 		case -NFS4ERR_ADMIN_REVOKED:
6683 		case -NFS4ERR_DELEG_REVOKED:
6684 		case -NFS4ERR_EXPIRED:
6685 		case -NFS4ERR_BAD_STATEID:
6686 			/* Let the main handler below do stateid recovery */
6687 			break;
6688 		case -NFS4ERR_OLD_STATEID:
6689 			if (nfs4_refresh_delegation_stateid(&data->stateid,
6690 						data->inode))
6691 				goto out_restart;
6692 			fallthrough;
6693 		default:
6694 			data->args.sattr_args = NULL;
6695 			data->res.sattr_res = false;
6696 			goto out_restart;
6697 		}
6698 	}
6699 
6700 	switch (task->tk_status) {
6701 	case 0:
6702 		renew_lease(data->res.server, data->timestamp);
6703 		break;
6704 	case -NFS4ERR_ADMIN_REVOKED:
6705 	case -NFS4ERR_DELEG_REVOKED:
6706 	case -NFS4ERR_EXPIRED:
6707 		nfs4_free_revoked_stateid(data->res.server,
6708 				data->args.stateid,
6709 				task->tk_msg.rpc_cred);
6710 		fallthrough;
6711 	case -NFS4ERR_BAD_STATEID:
6712 	case -NFS4ERR_STALE_STATEID:
6713 	case -ETIMEDOUT:
6714 		task->tk_status = 0;
6715 		break;
6716 	case -NFS4ERR_OLD_STATEID:
6717 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6718 			nfs4_stateid_seqid_inc(&data->stateid);
6719 		if (data->args.bitmask) {
6720 			data->args.bitmask = NULL;
6721 			data->res.fattr = NULL;
6722 		}
6723 		goto out_restart;
6724 	case -NFS4ERR_ACCESS:
6725 		if (data->args.bitmask) {
6726 			data->args.bitmask = NULL;
6727 			data->res.fattr = NULL;
6728 			goto out_restart;
6729 		}
6730 		fallthrough;
6731 	default:
6732 		task->tk_status = nfs4_async_handle_exception(task,
6733 				data->res.server, task->tk_status,
6734 				&exception);
6735 		if (exception.retry)
6736 			goto out_restart;
6737 	}
6738 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6739 	data->rpc_status = task->tk_status;
6740 	return;
6741 out_restart:
6742 	task->tk_status = 0;
6743 	rpc_restart_call_prepare(task);
6744 }
6745 
6746 static void nfs4_delegreturn_release(void *calldata)
6747 {
6748 	struct nfs4_delegreturndata *data = calldata;
6749 	struct inode *inode = data->inode;
6750 
6751 	if (data->lr.roc)
6752 		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6753 				 data->res.lr_ret);
6754 	if (inode) {
6755 		nfs4_fattr_set_prechange(&data->fattr,
6756 					 inode_peek_iversion_raw(inode));
6757 		nfs_refresh_inode(inode, &data->fattr);
6758 		nfs_iput_and_deactive(inode);
6759 	}
6760 	kfree(calldata);
6761 }
6762 
6763 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6764 {
6765 	struct nfs4_delegreturndata *d_data;
6766 	struct pnfs_layout_hdr *lo;
6767 
6768 	d_data = data;
6769 
6770 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6771 		nfs4_sequence_done(task, &d_data->res.seq_res);
6772 		return;
6773 	}
6774 
6775 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6776 	if (lo && !pnfs_layout_is_valid(lo)) {
6777 		d_data->args.lr_args = NULL;
6778 		d_data->res.lr_res = NULL;
6779 	}
6780 
6781 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6782 			&d_data->args.seq_args,
6783 			&d_data->res.seq_res,
6784 			task);
6785 }
6786 
6787 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6788 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6789 	.rpc_call_done = nfs4_delegreturn_done,
6790 	.rpc_release = nfs4_delegreturn_release,
6791 };
6792 
6793 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
6794 				  const nfs4_stateid *stateid,
6795 				  struct nfs_delegation *delegation,
6796 				  int issync)
6797 {
6798 	struct nfs4_delegreturndata *data;
6799 	struct nfs_server *server = NFS_SERVER(inode);
6800 	struct rpc_task *task;
6801 	struct rpc_message msg = {
6802 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6803 		.rpc_cred = cred,
6804 	};
6805 	struct rpc_task_setup task_setup_data = {
6806 		.rpc_client = server->client,
6807 		.rpc_message = &msg,
6808 		.callback_ops = &nfs4_delegreturn_ops,
6809 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6810 	};
6811 	int status = 0;
6812 
6813 	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6814 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6815 
6816 	data = kzalloc(sizeof(*data), GFP_KERNEL);
6817 	if (data == NULL)
6818 		return -ENOMEM;
6819 
6820 	nfs4_state_protect(server->nfs_client,
6821 			NFS_SP4_MACH_CRED_CLEANUP,
6822 			&task_setup_data.rpc_client, &msg);
6823 
6824 	data->args.fhandle = &data->fh;
6825 	data->args.stateid = &data->stateid;
6826 	nfs4_bitmask_set(data->args.bitmask_store,
6827 			 server->cache_consistency_bitmask, inode, 0);
6828 	data->args.bitmask = data->args.bitmask_store;
6829 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6830 	nfs4_stateid_copy(&data->stateid, stateid);
6831 	data->res.fattr = &data->fattr;
6832 	data->res.server = server;
6833 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6834 	data->lr.arg.ld_private = &data->lr.ld_private;
6835 	nfs_fattr_init(data->res.fattr);
6836 	data->timestamp = jiffies;
6837 	data->rpc_status = 0;
6838 	data->inode = nfs_igrab_and_active(inode);
6839 	if (data->inode || issync) {
6840 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6841 					cred);
6842 		if (data->lr.roc) {
6843 			data->args.lr_args = &data->lr.arg;
6844 			data->res.lr_res = &data->lr.res;
6845 		}
6846 	}
6847 
6848 	if (delegation &&
6849 	    test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags)) {
6850 		if (delegation->type & FMODE_READ) {
6851 			data->sattr.atime = inode_get_atime(inode);
6852 			data->sattr.atime_set = true;
6853 		}
6854 		if (delegation->type & FMODE_WRITE) {
6855 			data->sattr.mtime = inode_get_mtime(inode);
6856 			data->sattr.mtime_set = true;
6857 		}
6858 		data->args.sattr_args = &data->sattr;
6859 		data->res.sattr_res = true;
6860 	}
6861 
6862 	if (!data->inode)
6863 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6864 				   1);
6865 	else
6866 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6867 				   0);
6868 
6869 	task_setup_data.callback_data = data;
6870 	msg.rpc_argp = &data->args;
6871 	msg.rpc_resp = &data->res;
6872 	task = rpc_run_task(&task_setup_data);
6873 	if (IS_ERR(task))
6874 		return PTR_ERR(task);
6875 	if (!issync)
6876 		goto out;
6877 	status = rpc_wait_for_completion_task(task);
6878 	if (status != 0)
6879 		goto out;
6880 	status = data->rpc_status;
6881 out:
6882 	rpc_put_task(task);
6883 	return status;
6884 }
6885 
6886 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
6887 			  const nfs4_stateid *stateid,
6888 			  struct nfs_delegation *delegation, int issync)
6889 {
6890 	struct nfs_server *server = NFS_SERVER(inode);
6891 	struct nfs4_exception exception = { };
6892 	int err;
6893 	do {
6894 		err = _nfs4_proc_delegreturn(inode, cred, stateid,
6895 					     delegation, issync);
6896 		trace_nfs4_delegreturn(inode, stateid, err);
6897 		switch (err) {
6898 			case -NFS4ERR_STALE_STATEID:
6899 			case -NFS4ERR_EXPIRED:
6900 			case 0:
6901 				return 0;
6902 		}
6903 		err = nfs4_handle_exception(server, err, &exception);
6904 	} while (exception.retry);
6905 	return err;
6906 }
6907 
6908 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6909 {
6910 	struct inode *inode = state->inode;
6911 	struct nfs_server *server = NFS_SERVER(inode);
6912 	struct nfs_client *clp = server->nfs_client;
6913 	struct nfs_lockt_args arg = {
6914 		.fh = NFS_FH(inode),
6915 		.fl = request,
6916 	};
6917 	struct nfs_lockt_res res = {
6918 		.denied = request,
6919 	};
6920 	struct rpc_message msg = {
6921 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6922 		.rpc_argp	= &arg,
6923 		.rpc_resp	= &res,
6924 		.rpc_cred	= state->owner->so_cred,
6925 	};
6926 	struct nfs4_lock_state *lsp;
6927 	int status;
6928 
6929 	arg.lock_owner.clientid = clp->cl_clientid;
6930 	status = nfs4_set_lock_state(state, request);
6931 	if (status != 0)
6932 		goto out;
6933 	lsp = request->fl_u.nfs4_fl.owner;
6934 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6935 	arg.lock_owner.s_dev = server->s_dev;
6936 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6937 	switch (status) {
6938 		case 0:
6939 			request->c.flc_type = F_UNLCK;
6940 			break;
6941 		case -NFS4ERR_DENIED:
6942 			status = 0;
6943 	}
6944 	request->fl_ops->fl_release_private(request);
6945 	request->fl_ops = NULL;
6946 out:
6947 	return status;
6948 }
6949 
6950 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6951 {
6952 	struct nfs4_exception exception = {
6953 		.interruptible = true,
6954 	};
6955 	int err;
6956 
6957 	do {
6958 		err = _nfs4_proc_getlk(state, cmd, request);
6959 		trace_nfs4_get_lock(request, state, cmd, err);
6960 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6961 				&exception);
6962 	} while (exception.retry);
6963 	return err;
6964 }
6965 
6966 /*
6967  * Update the seqid of a lock stateid after receiving
6968  * NFS4ERR_OLD_STATEID
6969  */
6970 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6971 		struct nfs4_lock_state *lsp)
6972 {
6973 	struct nfs4_state *state = lsp->ls_state;
6974 	bool ret = false;
6975 
6976 	spin_lock(&state->state_lock);
6977 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6978 		goto out;
6979 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6980 		nfs4_stateid_seqid_inc(dst);
6981 	else
6982 		dst->seqid = lsp->ls_stateid.seqid;
6983 	ret = true;
6984 out:
6985 	spin_unlock(&state->state_lock);
6986 	return ret;
6987 }
6988 
6989 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6990 		struct nfs4_lock_state *lsp)
6991 {
6992 	struct nfs4_state *state = lsp->ls_state;
6993 	bool ret;
6994 
6995 	spin_lock(&state->state_lock);
6996 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6997 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6998 	spin_unlock(&state->state_lock);
6999 	return ret;
7000 }
7001 
7002 struct nfs4_unlockdata {
7003 	struct nfs_locku_args arg;
7004 	struct nfs_locku_res res;
7005 	struct nfs4_lock_state *lsp;
7006 	struct nfs_open_context *ctx;
7007 	struct nfs_lock_context *l_ctx;
7008 	struct file_lock fl;
7009 	struct nfs_server *server;
7010 	unsigned long timestamp;
7011 };
7012 
7013 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
7014 		struct nfs_open_context *ctx,
7015 		struct nfs4_lock_state *lsp,
7016 		struct nfs_seqid *seqid)
7017 {
7018 	struct nfs4_unlockdata *p;
7019 	struct nfs4_state *state = lsp->ls_state;
7020 	struct inode *inode = state->inode;
7021 
7022 	p = kzalloc(sizeof(*p), GFP_KERNEL);
7023 	if (p == NULL)
7024 		return NULL;
7025 	p->arg.fh = NFS_FH(inode);
7026 	p->arg.fl = &p->fl;
7027 	p->arg.seqid = seqid;
7028 	p->res.seqid = seqid;
7029 	p->lsp = lsp;
7030 	/* Ensure we don't close file until we're done freeing locks! */
7031 	p->ctx = get_nfs_open_context(ctx);
7032 	p->l_ctx = nfs_get_lock_context(ctx);
7033 	locks_init_lock(&p->fl);
7034 	locks_copy_lock(&p->fl, fl);
7035 	p->server = NFS_SERVER(inode);
7036 	spin_lock(&state->state_lock);
7037 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
7038 	spin_unlock(&state->state_lock);
7039 	return p;
7040 }
7041 
7042 static void nfs4_locku_release_calldata(void *data)
7043 {
7044 	struct nfs4_unlockdata *calldata = data;
7045 	nfs_free_seqid(calldata->arg.seqid);
7046 	nfs4_put_lock_state(calldata->lsp);
7047 	nfs_put_lock_context(calldata->l_ctx);
7048 	put_nfs_open_context(calldata->ctx);
7049 	kfree(calldata);
7050 }
7051 
7052 static void nfs4_locku_done(struct rpc_task *task, void *data)
7053 {
7054 	struct nfs4_unlockdata *calldata = data;
7055 	struct nfs4_exception exception = {
7056 		.inode = calldata->lsp->ls_state->inode,
7057 		.stateid = &calldata->arg.stateid,
7058 	};
7059 
7060 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
7061 		return;
7062 	switch (task->tk_status) {
7063 		case 0:
7064 			renew_lease(calldata->server, calldata->timestamp);
7065 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
7066 			if (nfs4_update_lock_stateid(calldata->lsp,
7067 					&calldata->res.stateid))
7068 				break;
7069 			fallthrough;
7070 		case -NFS4ERR_ADMIN_REVOKED:
7071 		case -NFS4ERR_EXPIRED:
7072 			nfs4_free_revoked_stateid(calldata->server,
7073 					&calldata->arg.stateid,
7074 					task->tk_msg.rpc_cred);
7075 			fallthrough;
7076 		case -NFS4ERR_BAD_STATEID:
7077 		case -NFS4ERR_STALE_STATEID:
7078 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
7079 						calldata->lsp))
7080 				rpc_restart_call_prepare(task);
7081 			break;
7082 		case -NFS4ERR_OLD_STATEID:
7083 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
7084 						calldata->lsp))
7085 				rpc_restart_call_prepare(task);
7086 			break;
7087 		default:
7088 			task->tk_status = nfs4_async_handle_exception(task,
7089 					calldata->server, task->tk_status,
7090 					&exception);
7091 			if (exception.retry)
7092 				rpc_restart_call_prepare(task);
7093 	}
7094 	nfs_release_seqid(calldata->arg.seqid);
7095 }
7096 
7097 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
7098 {
7099 	struct nfs4_unlockdata *calldata = data;
7100 
7101 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
7102 		nfs_async_iocounter_wait(task, calldata->l_ctx))
7103 		return;
7104 
7105 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
7106 		goto out_wait;
7107 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
7108 		/* Note: exit _without_ running nfs4_locku_done */
7109 		goto out_no_action;
7110 	}
7111 	calldata->timestamp = jiffies;
7112 	if (nfs4_setup_sequence(calldata->server->nfs_client,
7113 				&calldata->arg.seq_args,
7114 				&calldata->res.seq_res,
7115 				task) != 0)
7116 		nfs_release_seqid(calldata->arg.seqid);
7117 	return;
7118 out_no_action:
7119 	task->tk_action = NULL;
7120 out_wait:
7121 	nfs4_sequence_done(task, &calldata->res.seq_res);
7122 }
7123 
7124 static const struct rpc_call_ops nfs4_locku_ops = {
7125 	.rpc_call_prepare = nfs4_locku_prepare,
7126 	.rpc_call_done = nfs4_locku_done,
7127 	.rpc_release = nfs4_locku_release_calldata,
7128 };
7129 
7130 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
7131 		struct nfs_open_context *ctx,
7132 		struct nfs4_lock_state *lsp,
7133 		struct nfs_seqid *seqid)
7134 {
7135 	struct nfs4_unlockdata *data;
7136 	struct rpc_message msg = {
7137 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
7138 		.rpc_cred = ctx->cred,
7139 	};
7140 	struct rpc_task_setup task_setup_data = {
7141 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
7142 		.rpc_message = &msg,
7143 		.callback_ops = &nfs4_locku_ops,
7144 		.workqueue = nfsiod_workqueue,
7145 		.flags = RPC_TASK_ASYNC,
7146 	};
7147 
7148 	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
7149 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7150 
7151 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
7152 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
7153 
7154 	/* Ensure this is an unlock - when canceling a lock, the
7155 	 * canceled lock is passed in, and it won't be an unlock.
7156 	 */
7157 	fl->c.flc_type = F_UNLCK;
7158 	if (fl->c.flc_flags & FL_CLOSE)
7159 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
7160 
7161 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
7162 	if (data == NULL) {
7163 		nfs_free_seqid(seqid);
7164 		return ERR_PTR(-ENOMEM);
7165 	}
7166 
7167 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
7168 	msg.rpc_argp = &data->arg;
7169 	msg.rpc_resp = &data->res;
7170 	task_setup_data.callback_data = data;
7171 	return rpc_run_task(&task_setup_data);
7172 }
7173 
7174 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7175 {
7176 	struct inode *inode = state->inode;
7177 	struct nfs4_state_owner *sp = state->owner;
7178 	struct nfs_inode *nfsi = NFS_I(inode);
7179 	struct nfs_seqid *seqid;
7180 	struct nfs4_lock_state *lsp;
7181 	struct rpc_task *task;
7182 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7183 	int status = 0;
7184 	unsigned char saved_flags = request->c.flc_flags;
7185 
7186 	status = nfs4_set_lock_state(state, request);
7187 	/* Unlock _before_ we do the RPC call */
7188 	request->c.flc_flags |= FL_EXISTS;
7189 	/* Exclude nfs_delegation_claim_locks() */
7190 	mutex_lock(&sp->so_delegreturn_mutex);
7191 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7192 	down_read(&nfsi->rwsem);
7193 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7194 		up_read(&nfsi->rwsem);
7195 		mutex_unlock(&sp->so_delegreturn_mutex);
7196 		goto out;
7197 	}
7198 	lsp = request->fl_u.nfs4_fl.owner;
7199 	set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7200 	up_read(&nfsi->rwsem);
7201 	mutex_unlock(&sp->so_delegreturn_mutex);
7202 	if (status != 0)
7203 		goto out;
7204 	/* Is this a delegated lock? */
7205 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7206 		goto out;
7207 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7208 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7209 	status = -ENOMEM;
7210 	if (IS_ERR(seqid))
7211 		goto out;
7212 	task = nfs4_do_unlck(request,
7213 			     nfs_file_open_context(request->c.flc_file),
7214 			     lsp, seqid);
7215 	status = PTR_ERR(task);
7216 	if (IS_ERR(task))
7217 		goto out;
7218 	status = rpc_wait_for_completion_task(task);
7219 	rpc_put_task(task);
7220 out:
7221 	request->c.flc_flags = saved_flags;
7222 	trace_nfs4_unlock(request, state, F_SETLK, status);
7223 	return status;
7224 }
7225 
7226 struct nfs4_lockdata {
7227 	struct nfs_lock_args arg;
7228 	struct nfs_lock_res res;
7229 	struct nfs4_lock_state *lsp;
7230 	struct nfs_open_context *ctx;
7231 	struct file_lock fl;
7232 	unsigned long timestamp;
7233 	int rpc_status;
7234 	int cancelled;
7235 	struct nfs_server *server;
7236 };
7237 
7238 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7239 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7240 		gfp_t gfp_mask)
7241 {
7242 	struct nfs4_lockdata *p;
7243 	struct inode *inode = lsp->ls_state->inode;
7244 	struct nfs_server *server = NFS_SERVER(inode);
7245 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7246 
7247 	p = kzalloc(sizeof(*p), gfp_mask);
7248 	if (p == NULL)
7249 		return NULL;
7250 
7251 	p->arg.fh = NFS_FH(inode);
7252 	p->arg.fl = &p->fl;
7253 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7254 	if (IS_ERR(p->arg.open_seqid))
7255 		goto out_free;
7256 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7257 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7258 	if (IS_ERR(p->arg.lock_seqid))
7259 		goto out_free_seqid;
7260 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7261 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7262 	p->arg.lock_owner.s_dev = server->s_dev;
7263 	p->res.lock_seqid = p->arg.lock_seqid;
7264 	p->lsp = lsp;
7265 	p->server = server;
7266 	p->ctx = get_nfs_open_context(ctx);
7267 	locks_init_lock(&p->fl);
7268 	locks_copy_lock(&p->fl, fl);
7269 	return p;
7270 out_free_seqid:
7271 	nfs_free_seqid(p->arg.open_seqid);
7272 out_free:
7273 	kfree(p);
7274 	return NULL;
7275 }
7276 
7277 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7278 {
7279 	struct nfs4_lockdata *data = calldata;
7280 	struct nfs4_state *state = data->lsp->ls_state;
7281 
7282 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7283 		goto out_wait;
7284 	/* Do we need to do an open_to_lock_owner? */
7285 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7286 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7287 			goto out_release_lock_seqid;
7288 		}
7289 		nfs4_stateid_copy(&data->arg.open_stateid,
7290 				&state->open_stateid);
7291 		data->arg.new_lock_owner = 1;
7292 		data->res.open_seqid = data->arg.open_seqid;
7293 	} else {
7294 		data->arg.new_lock_owner = 0;
7295 		nfs4_stateid_copy(&data->arg.lock_stateid,
7296 				&data->lsp->ls_stateid);
7297 	}
7298 	if (!nfs4_valid_open_stateid(state)) {
7299 		data->rpc_status = -EBADF;
7300 		task->tk_action = NULL;
7301 		goto out_release_open_seqid;
7302 	}
7303 	data->timestamp = jiffies;
7304 	if (nfs4_setup_sequence(data->server->nfs_client,
7305 				&data->arg.seq_args,
7306 				&data->res.seq_res,
7307 				task) == 0)
7308 		return;
7309 out_release_open_seqid:
7310 	nfs_release_seqid(data->arg.open_seqid);
7311 out_release_lock_seqid:
7312 	nfs_release_seqid(data->arg.lock_seqid);
7313 out_wait:
7314 	nfs4_sequence_done(task, &data->res.seq_res);
7315 	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7316 }
7317 
7318 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7319 {
7320 	struct nfs4_lockdata *data = calldata;
7321 	struct nfs4_lock_state *lsp = data->lsp;
7322 
7323 	if (!nfs4_sequence_done(task, &data->res.seq_res))
7324 		return;
7325 
7326 	data->rpc_status = task->tk_status;
7327 	switch (task->tk_status) {
7328 	case 0:
7329 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7330 				data->timestamp);
7331 		if (data->arg.new_lock && !data->cancelled) {
7332 			data->fl.c.flc_flags &= ~(FL_SLEEP | FL_ACCESS);
7333 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7334 				goto out_restart;
7335 		}
7336 		if (data->arg.new_lock_owner != 0) {
7337 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7338 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7339 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7340 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7341 			goto out_restart;
7342 		break;
7343 	case -NFS4ERR_OLD_STATEID:
7344 		if (data->arg.new_lock_owner != 0 &&
7345 			nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7346 					lsp->ls_state))
7347 			goto out_restart;
7348 		if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7349 			goto out_restart;
7350 		fallthrough;
7351 	case -NFS4ERR_BAD_STATEID:
7352 	case -NFS4ERR_STALE_STATEID:
7353 	case -NFS4ERR_EXPIRED:
7354 		if (data->arg.new_lock_owner != 0) {
7355 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7356 						&lsp->ls_state->open_stateid))
7357 				goto out_restart;
7358 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7359 						&lsp->ls_stateid))
7360 				goto out_restart;
7361 	}
7362 out_done:
7363 	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7364 	return;
7365 out_restart:
7366 	if (!data->cancelled)
7367 		rpc_restart_call_prepare(task);
7368 	goto out_done;
7369 }
7370 
7371 static void nfs4_lock_release(void *calldata)
7372 {
7373 	struct nfs4_lockdata *data = calldata;
7374 
7375 	nfs_free_seqid(data->arg.open_seqid);
7376 	if (data->cancelled && data->rpc_status == 0) {
7377 		struct rpc_task *task;
7378 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7379 				data->arg.lock_seqid);
7380 		if (!IS_ERR(task))
7381 			rpc_put_task_async(task);
7382 		dprintk("%s: cancelling lock!\n", __func__);
7383 	} else
7384 		nfs_free_seqid(data->arg.lock_seqid);
7385 	nfs4_put_lock_state(data->lsp);
7386 	put_nfs_open_context(data->ctx);
7387 	kfree(data);
7388 }
7389 
7390 static const struct rpc_call_ops nfs4_lock_ops = {
7391 	.rpc_call_prepare = nfs4_lock_prepare,
7392 	.rpc_call_done = nfs4_lock_done,
7393 	.rpc_release = nfs4_lock_release,
7394 };
7395 
7396 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7397 {
7398 	switch (error) {
7399 	case -NFS4ERR_ADMIN_REVOKED:
7400 	case -NFS4ERR_EXPIRED:
7401 	case -NFS4ERR_BAD_STATEID:
7402 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7403 		if (new_lock_owner != 0 ||
7404 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7405 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7406 		break;
7407 	case -NFS4ERR_STALE_STATEID:
7408 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7409 		nfs4_schedule_lease_recovery(server->nfs_client);
7410 	}
7411 }
7412 
7413 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7414 {
7415 	struct nfs4_lockdata *data;
7416 	struct rpc_task *task;
7417 	struct rpc_message msg = {
7418 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7419 		.rpc_cred = state->owner->so_cred,
7420 	};
7421 	struct rpc_task_setup task_setup_data = {
7422 		.rpc_client = NFS_CLIENT(state->inode),
7423 		.rpc_message = &msg,
7424 		.callback_ops = &nfs4_lock_ops,
7425 		.workqueue = nfsiod_workqueue,
7426 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7427 	};
7428 	int ret;
7429 
7430 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7431 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7432 
7433 	data = nfs4_alloc_lockdata(fl,
7434 				   nfs_file_open_context(fl->c.flc_file),
7435 				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7436 	if (data == NULL)
7437 		return -ENOMEM;
7438 	if (IS_SETLKW(cmd))
7439 		data->arg.block = 1;
7440 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7441 				recovery_type > NFS_LOCK_NEW);
7442 	msg.rpc_argp = &data->arg;
7443 	msg.rpc_resp = &data->res;
7444 	task_setup_data.callback_data = data;
7445 	if (recovery_type > NFS_LOCK_NEW) {
7446 		if (recovery_type == NFS_LOCK_RECLAIM)
7447 			data->arg.reclaim = NFS_LOCK_RECLAIM;
7448 	} else
7449 		data->arg.new_lock = 1;
7450 	task = rpc_run_task(&task_setup_data);
7451 	if (IS_ERR(task))
7452 		return PTR_ERR(task);
7453 	ret = rpc_wait_for_completion_task(task);
7454 	if (ret == 0) {
7455 		ret = data->rpc_status;
7456 		if (ret)
7457 			nfs4_handle_setlk_error(data->server, data->lsp,
7458 					data->arg.new_lock_owner, ret);
7459 	} else
7460 		data->cancelled = true;
7461 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7462 	rpc_put_task(task);
7463 	dprintk("%s: ret = %d\n", __func__, ret);
7464 	return ret;
7465 }
7466 
7467 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7468 {
7469 	struct nfs_server *server = NFS_SERVER(state->inode);
7470 	struct nfs4_exception exception = {
7471 		.inode = state->inode,
7472 	};
7473 	int err;
7474 
7475 	do {
7476 		/* Cache the lock if possible... */
7477 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7478 			return 0;
7479 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7480 		if (err != -NFS4ERR_DELAY)
7481 			break;
7482 		nfs4_handle_exception(server, err, &exception);
7483 	} while (exception.retry);
7484 	return err;
7485 }
7486 
7487 static int nfs4_lock_expired(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 	err = nfs4_set_lock_state(state, request);
7496 	if (err != 0)
7497 		return err;
7498 	if (!recover_lost_locks) {
7499 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7500 		return 0;
7501 	}
7502 	do {
7503 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7504 			return 0;
7505 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7506 		switch (err) {
7507 		default:
7508 			goto out;
7509 		case -NFS4ERR_GRACE:
7510 		case -NFS4ERR_DELAY:
7511 			nfs4_handle_exception(server, err, &exception);
7512 			err = 0;
7513 		}
7514 	} while (exception.retry);
7515 out:
7516 	return err;
7517 }
7518 
7519 #if defined(CONFIG_NFS_V4_1)
7520 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7521 {
7522 	struct nfs4_lock_state *lsp;
7523 	int status;
7524 
7525 	status = nfs4_set_lock_state(state, request);
7526 	if (status != 0)
7527 		return status;
7528 	lsp = request->fl_u.nfs4_fl.owner;
7529 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7530 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7531 		return 0;
7532 	return nfs4_lock_expired(state, request);
7533 }
7534 #endif
7535 
7536 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7537 {
7538 	struct nfs_inode *nfsi = NFS_I(state->inode);
7539 	struct nfs4_state_owner *sp = state->owner;
7540 	unsigned char flags = request->c.flc_flags;
7541 	int status;
7542 
7543 	request->c.flc_flags |= FL_ACCESS;
7544 	status = locks_lock_inode_wait(state->inode, request);
7545 	if (status < 0)
7546 		goto out;
7547 	mutex_lock(&sp->so_delegreturn_mutex);
7548 	down_read(&nfsi->rwsem);
7549 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7550 		/* Yes: cache locks! */
7551 		/* ...but avoid races with delegation recall... */
7552 		request->c.flc_flags = flags & ~FL_SLEEP;
7553 		status = locks_lock_inode_wait(state->inode, request);
7554 		up_read(&nfsi->rwsem);
7555 		mutex_unlock(&sp->so_delegreturn_mutex);
7556 		goto out;
7557 	}
7558 	up_read(&nfsi->rwsem);
7559 	mutex_unlock(&sp->so_delegreturn_mutex);
7560 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7561 out:
7562 	request->c.flc_flags = flags;
7563 	return status;
7564 }
7565 
7566 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7567 {
7568 	struct nfs4_exception exception = {
7569 		.state = state,
7570 		.inode = state->inode,
7571 		.interruptible = true,
7572 	};
7573 	int err;
7574 
7575 	do {
7576 		err = _nfs4_proc_setlk(state, cmd, request);
7577 		if (err == -NFS4ERR_DENIED)
7578 			err = -EAGAIN;
7579 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7580 				err, &exception);
7581 	} while (exception.retry);
7582 	return err;
7583 }
7584 
7585 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7586 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7587 
7588 static int
7589 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7590 			struct file_lock *request)
7591 {
7592 	int		status = -ERESTARTSYS;
7593 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7594 
7595 	while(!signalled()) {
7596 		status = nfs4_proc_setlk(state, cmd, request);
7597 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7598 			break;
7599 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7600 		schedule_timeout(timeout);
7601 		timeout *= 2;
7602 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7603 		status = -ERESTARTSYS;
7604 	}
7605 	return status;
7606 }
7607 
7608 #ifdef CONFIG_NFS_V4_1
7609 struct nfs4_lock_waiter {
7610 	struct inode		*inode;
7611 	struct nfs_lowner	owner;
7612 	wait_queue_entry_t	wait;
7613 };
7614 
7615 static int
7616 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7617 {
7618 	struct nfs4_lock_waiter	*waiter	=
7619 		container_of(wait, struct nfs4_lock_waiter, wait);
7620 
7621 	/* NULL key means to wake up everyone */
7622 	if (key) {
7623 		struct cb_notify_lock_args	*cbnl = key;
7624 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7625 						*wowner = &waiter->owner;
7626 
7627 		/* Only wake if the callback was for the same owner. */
7628 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7629 			return 0;
7630 
7631 		/* Make sure it's for the right inode */
7632 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7633 			return 0;
7634 	}
7635 
7636 	return woken_wake_function(wait, mode, flags, key);
7637 }
7638 
7639 static int
7640 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7641 {
7642 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7643 	struct nfs_server *server = NFS_SERVER(state->inode);
7644 	struct nfs_client *clp = server->nfs_client;
7645 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7646 	struct nfs4_lock_waiter waiter = {
7647 		.inode = state->inode,
7648 		.owner = { .clientid = clp->cl_clientid,
7649 			   .id = lsp->ls_seqid.owner_id,
7650 			   .s_dev = server->s_dev },
7651 	};
7652 	int status;
7653 
7654 	/* Don't bother with waitqueue if we don't expect a callback */
7655 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7656 		return nfs4_retry_setlk_simple(state, cmd, request);
7657 
7658 	init_wait(&waiter.wait);
7659 	waiter.wait.func = nfs4_wake_lock_waiter;
7660 	add_wait_queue(q, &waiter.wait);
7661 
7662 	do {
7663 		status = nfs4_proc_setlk(state, cmd, request);
7664 		if (status != -EAGAIN || IS_SETLK(cmd))
7665 			break;
7666 
7667 		status = -ERESTARTSYS;
7668 		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7669 			   NFS4_LOCK_MAXTIMEOUT);
7670 	} while (!signalled());
7671 
7672 	remove_wait_queue(q, &waiter.wait);
7673 
7674 	return status;
7675 }
7676 #else /* !CONFIG_NFS_V4_1 */
7677 static inline int
7678 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7679 {
7680 	return nfs4_retry_setlk_simple(state, cmd, request);
7681 }
7682 #endif
7683 
7684 static int
7685 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7686 {
7687 	struct nfs_open_context *ctx;
7688 	struct nfs4_state *state;
7689 	int status;
7690 
7691 	/* verify open state */
7692 	ctx = nfs_file_open_context(filp);
7693 	state = ctx->state;
7694 
7695 	if (IS_GETLK(cmd)) {
7696 		if (state != NULL)
7697 			return nfs4_proc_getlk(state, F_GETLK, request);
7698 		return 0;
7699 	}
7700 
7701 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7702 		return -EINVAL;
7703 
7704 	if (lock_is_unlock(request)) {
7705 		if (state != NULL)
7706 			return nfs4_proc_unlck(state, cmd, request);
7707 		return 0;
7708 	}
7709 
7710 	if (state == NULL)
7711 		return -ENOLCK;
7712 
7713 	if ((request->c.flc_flags & FL_POSIX) &&
7714 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7715 		return -ENOLCK;
7716 
7717 	/*
7718 	 * Don't rely on the VFS having checked the file open mode,
7719 	 * since it won't do this for flock() locks.
7720 	 */
7721 	switch (request->c.flc_type) {
7722 	case F_RDLCK:
7723 		if (!(filp->f_mode & FMODE_READ))
7724 			return -EBADF;
7725 		break;
7726 	case F_WRLCK:
7727 		if (!(filp->f_mode & FMODE_WRITE))
7728 			return -EBADF;
7729 	}
7730 
7731 	status = nfs4_set_lock_state(state, request);
7732 	if (status != 0)
7733 		return status;
7734 
7735 	return nfs4_retry_setlk(state, cmd, request);
7736 }
7737 
7738 static int nfs4_delete_lease(struct file *file, void **priv)
7739 {
7740 	return generic_setlease(file, F_UNLCK, NULL, priv);
7741 }
7742 
7743 static int nfs4_add_lease(struct file *file, int arg, struct file_lease **lease,
7744 			  void **priv)
7745 {
7746 	struct inode *inode = file_inode(file);
7747 	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7748 	int ret;
7749 
7750 	/* No delegation, no lease */
7751 	if (!nfs4_have_delegation(inode, type, 0))
7752 		return -EAGAIN;
7753 	ret = generic_setlease(file, arg, lease, priv);
7754 	if (ret || nfs4_have_delegation(inode, type, 0))
7755 		return ret;
7756 	/* We raced with a delegation return */
7757 	nfs4_delete_lease(file, priv);
7758 	return -EAGAIN;
7759 }
7760 
7761 int nfs4_proc_setlease(struct file *file, int arg, struct file_lease **lease,
7762 		       void **priv)
7763 {
7764 	switch (arg) {
7765 	case F_RDLCK:
7766 	case F_WRLCK:
7767 		return nfs4_add_lease(file, arg, lease, priv);
7768 	case F_UNLCK:
7769 		return nfs4_delete_lease(file, priv);
7770 	default:
7771 		return -EINVAL;
7772 	}
7773 }
7774 
7775 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7776 {
7777 	struct nfs_server *server = NFS_SERVER(state->inode);
7778 	int err;
7779 
7780 	err = nfs4_set_lock_state(state, fl);
7781 	if (err != 0)
7782 		return err;
7783 	do {
7784 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7785 		if (err != -NFS4ERR_DELAY)
7786 			break;
7787 		ssleep(1);
7788 	} while (err == -NFS4ERR_DELAY);
7789 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7790 }
7791 
7792 struct nfs_release_lockowner_data {
7793 	struct nfs4_lock_state *lsp;
7794 	struct nfs_server *server;
7795 	struct nfs_release_lockowner_args args;
7796 	struct nfs_release_lockowner_res res;
7797 	unsigned long timestamp;
7798 };
7799 
7800 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7801 {
7802 	struct nfs_release_lockowner_data *data = calldata;
7803 	struct nfs_server *server = data->server;
7804 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7805 			   &data->res.seq_res, task);
7806 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7807 	data->timestamp = jiffies;
7808 }
7809 
7810 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7811 {
7812 	struct nfs_release_lockowner_data *data = calldata;
7813 	struct nfs_server *server = data->server;
7814 
7815 	nfs40_sequence_done(task, &data->res.seq_res);
7816 
7817 	switch (task->tk_status) {
7818 	case 0:
7819 		renew_lease(server, data->timestamp);
7820 		break;
7821 	case -NFS4ERR_STALE_CLIENTID:
7822 	case -NFS4ERR_EXPIRED:
7823 		nfs4_schedule_lease_recovery(server->nfs_client);
7824 		break;
7825 	case -NFS4ERR_LEASE_MOVED:
7826 	case -NFS4ERR_DELAY:
7827 		if (nfs4_async_handle_error(task, server,
7828 					    NULL, NULL) == -EAGAIN)
7829 			rpc_restart_call_prepare(task);
7830 	}
7831 }
7832 
7833 static void nfs4_release_lockowner_release(void *calldata)
7834 {
7835 	struct nfs_release_lockowner_data *data = calldata;
7836 	nfs4_free_lock_state(data->server, data->lsp);
7837 	kfree(calldata);
7838 }
7839 
7840 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7841 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7842 	.rpc_call_done = nfs4_release_lockowner_done,
7843 	.rpc_release = nfs4_release_lockowner_release,
7844 };
7845 
7846 static void
7847 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7848 {
7849 	struct nfs_release_lockowner_data *data;
7850 	struct rpc_message msg = {
7851 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7852 	};
7853 
7854 	if (server->nfs_client->cl_mvops->minor_version != 0)
7855 		return;
7856 
7857 	data = kmalloc(sizeof(*data), GFP_KERNEL);
7858 	if (!data)
7859 		return;
7860 	data->lsp = lsp;
7861 	data->server = server;
7862 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7863 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7864 	data->args.lock_owner.s_dev = server->s_dev;
7865 
7866 	msg.rpc_argp = &data->args;
7867 	msg.rpc_resp = &data->res;
7868 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7869 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7870 }
7871 
7872 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7873 
7874 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7875 				   struct mnt_idmap *idmap,
7876 				   struct dentry *unused, struct inode *inode,
7877 				   const char *key, const void *buf,
7878 				   size_t buflen, int flags)
7879 {
7880 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7881 }
7882 
7883 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7884 				   struct dentry *unused, struct inode *inode,
7885 				   const char *key, void *buf, size_t buflen)
7886 {
7887 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7888 }
7889 
7890 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7891 {
7892 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7893 }
7894 
7895 #if defined(CONFIG_NFS_V4_1)
7896 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7897 
7898 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7899 				    struct mnt_idmap *idmap,
7900 				    struct dentry *unused, struct inode *inode,
7901 				    const char *key, const void *buf,
7902 				    size_t buflen, int flags)
7903 {
7904 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7905 }
7906 
7907 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7908 				    struct dentry *unused, struct inode *inode,
7909 				    const char *key, void *buf, size_t buflen)
7910 {
7911 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7912 }
7913 
7914 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7915 {
7916 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7917 }
7918 
7919 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7920 
7921 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7922 				    struct mnt_idmap *idmap,
7923 				    struct dentry *unused, struct inode *inode,
7924 				    const char *key, const void *buf,
7925 				    size_t buflen, int flags)
7926 {
7927 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7928 }
7929 
7930 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7931 				    struct dentry *unused, struct inode *inode,
7932 				    const char *key, void *buf, size_t buflen)
7933 {
7934 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7935 }
7936 
7937 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7938 {
7939 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7940 }
7941 
7942 #endif
7943 
7944 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7945 
7946 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7947 				     struct mnt_idmap *idmap,
7948 				     struct dentry *unused, struct inode *inode,
7949 				     const char *key, const void *buf,
7950 				     size_t buflen, int flags)
7951 {
7952 	if (security_ismaclabel(key))
7953 		return nfs4_set_security_label(inode, buf, buflen);
7954 
7955 	return -EOPNOTSUPP;
7956 }
7957 
7958 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7959 				     struct dentry *unused, struct inode *inode,
7960 				     const char *key, void *buf, size_t buflen)
7961 {
7962 	if (security_ismaclabel(key))
7963 		return nfs4_get_security_label(inode, buf, buflen);
7964 	return -EOPNOTSUPP;
7965 }
7966 
7967 static ssize_t
7968 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7969 {
7970 	int len = 0;
7971 
7972 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7973 		len = security_inode_listsecurity(inode, list, list_len);
7974 		if (len >= 0 && list_len && len > list_len)
7975 			return -ERANGE;
7976 	}
7977 	return len;
7978 }
7979 
7980 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7981 	.prefix = XATTR_SECURITY_PREFIX,
7982 	.get	= nfs4_xattr_get_nfs4_label,
7983 	.set	= nfs4_xattr_set_nfs4_label,
7984 };
7985 
7986 #else
7987 
7988 static ssize_t
7989 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7990 {
7991 	return 0;
7992 }
7993 
7994 #endif
7995 
7996 #ifdef CONFIG_NFS_V4_2
7997 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7998 				    struct mnt_idmap *idmap,
7999 				    struct dentry *unused, struct inode *inode,
8000 				    const char *key, const void *buf,
8001 				    size_t buflen, int flags)
8002 {
8003 	u32 mask;
8004 	int ret;
8005 
8006 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
8007 		return -EOPNOTSUPP;
8008 
8009 	/*
8010 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
8011 	 * flags right now. Handling of xattr operations use the normal
8012 	 * file read/write permissions.
8013 	 *
8014 	 * Just in case the server has other ideas (which RFC 8276 allows),
8015 	 * do a cached access check for the XA* flags to possibly avoid
8016 	 * doing an RPC and getting EACCES back.
8017 	 */
8018 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
8019 		if (!(mask & NFS_ACCESS_XAWRITE))
8020 			return -EACCES;
8021 	}
8022 
8023 	if (buf == NULL) {
8024 		ret = nfs42_proc_removexattr(inode, key);
8025 		if (!ret)
8026 			nfs4_xattr_cache_remove(inode, key);
8027 	} else {
8028 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
8029 		if (!ret)
8030 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
8031 	}
8032 
8033 	return ret;
8034 }
8035 
8036 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
8037 				    struct dentry *unused, struct inode *inode,
8038 				    const char *key, void *buf, size_t buflen)
8039 {
8040 	u32 mask;
8041 	ssize_t ret;
8042 
8043 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
8044 		return -EOPNOTSUPP;
8045 
8046 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
8047 		if (!(mask & NFS_ACCESS_XAREAD))
8048 			return -EACCES;
8049 	}
8050 
8051 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
8052 	if (ret)
8053 		return ret;
8054 
8055 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
8056 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
8057 		return ret;
8058 
8059 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
8060 
8061 	return ret;
8062 }
8063 
8064 static ssize_t
8065 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
8066 {
8067 	u64 cookie;
8068 	bool eof;
8069 	ssize_t ret, size;
8070 	char *buf;
8071 	size_t buflen;
8072 	u32 mask;
8073 
8074 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
8075 		return 0;
8076 
8077 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
8078 		if (!(mask & NFS_ACCESS_XALIST))
8079 			return 0;
8080 	}
8081 
8082 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
8083 	if (ret)
8084 		return ret;
8085 
8086 	ret = nfs4_xattr_cache_list(inode, list, list_len);
8087 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
8088 		return ret;
8089 
8090 	cookie = 0;
8091 	eof = false;
8092 	buflen = list_len ? list_len : XATTR_LIST_MAX;
8093 	buf = list_len ? list : NULL;
8094 	size = 0;
8095 
8096 	while (!eof) {
8097 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
8098 		    &cookie, &eof);
8099 		if (ret < 0)
8100 			return ret;
8101 
8102 		if (list_len) {
8103 			buf += ret;
8104 			buflen -= ret;
8105 		}
8106 		size += ret;
8107 	}
8108 
8109 	if (list_len)
8110 		nfs4_xattr_cache_set_list(inode, list, size);
8111 
8112 	return size;
8113 }
8114 
8115 #else
8116 
8117 static ssize_t
8118 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
8119 {
8120 	return 0;
8121 }
8122 #endif /* CONFIG_NFS_V4_2 */
8123 
8124 /*
8125  * nfs_fhget will use either the mounted_on_fileid or the fileid
8126  */
8127 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
8128 {
8129 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
8130 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
8131 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
8132 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
8133 		return;
8134 
8135 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
8136 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
8137 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
8138 	fattr->nlink = 2;
8139 }
8140 
8141 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8142 				   const struct qstr *name,
8143 				   struct nfs4_fs_locations *fs_locations,
8144 				   struct page *page)
8145 {
8146 	struct nfs_server *server = NFS_SERVER(dir);
8147 	u32 bitmask[3];
8148 	struct nfs4_fs_locations_arg args = {
8149 		.dir_fh = NFS_FH(dir),
8150 		.name = name,
8151 		.page = page,
8152 		.bitmask = bitmask,
8153 	};
8154 	struct nfs4_fs_locations_res res = {
8155 		.fs_locations = fs_locations,
8156 	};
8157 	struct rpc_message msg = {
8158 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8159 		.rpc_argp = &args,
8160 		.rpc_resp = &res,
8161 	};
8162 	int status;
8163 
8164 	dprintk("%s: start\n", __func__);
8165 
8166 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
8167 	bitmask[1] = nfs4_fattr_bitmap[1];
8168 
8169 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
8170 	 * is not supported */
8171 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
8172 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
8173 	else
8174 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8175 
8176 	nfs_fattr_init(fs_locations->fattr);
8177 	fs_locations->server = server;
8178 	fs_locations->nlocations = 0;
8179 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8180 	dprintk("%s: returned status = %d\n", __func__, status);
8181 	return status;
8182 }
8183 
8184 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8185 			   const struct qstr *name,
8186 			   struct nfs4_fs_locations *fs_locations,
8187 			   struct page *page)
8188 {
8189 	struct nfs4_exception exception = {
8190 		.interruptible = true,
8191 	};
8192 	int err;
8193 	do {
8194 		err = _nfs4_proc_fs_locations(client, dir, name,
8195 				fs_locations, page);
8196 		trace_nfs4_get_fs_locations(dir, name, err);
8197 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8198 				&exception);
8199 	} while (exception.retry);
8200 	return err;
8201 }
8202 
8203 /*
8204  * This operation also signals the server that this client is
8205  * performing migration recovery.  The server can stop returning
8206  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8207  * appended to this compound to identify the client ID which is
8208  * performing recovery.
8209  */
8210 static int _nfs40_proc_get_locations(struct nfs_server *server,
8211 				     struct nfs_fh *fhandle,
8212 				     struct nfs4_fs_locations *locations,
8213 				     struct page *page, const struct cred *cred)
8214 {
8215 	struct rpc_clnt *clnt = server->client;
8216 	u32 bitmask[2] = {
8217 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8218 	};
8219 	struct nfs4_fs_locations_arg args = {
8220 		.clientid	= server->nfs_client->cl_clientid,
8221 		.fh		= fhandle,
8222 		.page		= page,
8223 		.bitmask	= bitmask,
8224 		.migration	= 1,		/* skip LOOKUP */
8225 		.renew		= 1,		/* append RENEW */
8226 	};
8227 	struct nfs4_fs_locations_res res = {
8228 		.fs_locations	= locations,
8229 		.migration	= 1,
8230 		.renew		= 1,
8231 	};
8232 	struct rpc_message msg = {
8233 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8234 		.rpc_argp	= &args,
8235 		.rpc_resp	= &res,
8236 		.rpc_cred	= cred,
8237 	};
8238 	unsigned long now = jiffies;
8239 	int status;
8240 
8241 	nfs_fattr_init(locations->fattr);
8242 	locations->server = server;
8243 	locations->nlocations = 0;
8244 
8245 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8246 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8247 					&args.seq_args, &res.seq_res);
8248 	if (status)
8249 		return status;
8250 
8251 	renew_lease(server, now);
8252 	return 0;
8253 }
8254 
8255 #ifdef CONFIG_NFS_V4_1
8256 
8257 /*
8258  * This operation also signals the server that this client is
8259  * performing migration recovery.  The server can stop asserting
8260  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8261  * performing this operation is identified in the SEQUENCE
8262  * operation in this compound.
8263  *
8264  * When the client supports GETATTR(fs_locations_info), it can
8265  * be plumbed in here.
8266  */
8267 static int _nfs41_proc_get_locations(struct nfs_server *server,
8268 				     struct nfs_fh *fhandle,
8269 				     struct nfs4_fs_locations *locations,
8270 				     struct page *page, const struct cred *cred)
8271 {
8272 	struct rpc_clnt *clnt = server->client;
8273 	u32 bitmask[2] = {
8274 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8275 	};
8276 	struct nfs4_fs_locations_arg args = {
8277 		.fh		= fhandle,
8278 		.page		= page,
8279 		.bitmask	= bitmask,
8280 		.migration	= 1,		/* skip LOOKUP */
8281 	};
8282 	struct nfs4_fs_locations_res res = {
8283 		.fs_locations	= locations,
8284 		.migration	= 1,
8285 	};
8286 	struct rpc_message msg = {
8287 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8288 		.rpc_argp	= &args,
8289 		.rpc_resp	= &res,
8290 		.rpc_cred	= cred,
8291 	};
8292 	struct nfs4_call_sync_data data = {
8293 		.seq_server = server,
8294 		.seq_args = &args.seq_args,
8295 		.seq_res = &res.seq_res,
8296 	};
8297 	struct rpc_task_setup task_setup_data = {
8298 		.rpc_client = clnt,
8299 		.rpc_message = &msg,
8300 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8301 		.callback_data = &data,
8302 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8303 	};
8304 	int status;
8305 
8306 	nfs_fattr_init(locations->fattr);
8307 	locations->server = server;
8308 	locations->nlocations = 0;
8309 
8310 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8311 	status = nfs4_call_sync_custom(&task_setup_data);
8312 	if (status == NFS4_OK &&
8313 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8314 		status = -NFS4ERR_LEASE_MOVED;
8315 	return status;
8316 }
8317 
8318 #endif	/* CONFIG_NFS_V4_1 */
8319 
8320 /**
8321  * nfs4_proc_get_locations - discover locations for a migrated FSID
8322  * @server: pointer to nfs_server to process
8323  * @fhandle: pointer to the kernel NFS client file handle
8324  * @locations: result of query
8325  * @page: buffer
8326  * @cred: credential to use for this operation
8327  *
8328  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8329  * operation failed, or a negative errno if a local error occurred.
8330  *
8331  * On success, "locations" is filled in, but if the server has
8332  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8333  * asserted.
8334  *
8335  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8336  * from this client that require migration recovery.
8337  */
8338 int nfs4_proc_get_locations(struct nfs_server *server,
8339 			    struct nfs_fh *fhandle,
8340 			    struct nfs4_fs_locations *locations,
8341 			    struct page *page, const struct cred *cred)
8342 {
8343 	struct nfs_client *clp = server->nfs_client;
8344 	const struct nfs4_mig_recovery_ops *ops =
8345 					clp->cl_mvops->mig_recovery_ops;
8346 	struct nfs4_exception exception = {
8347 		.interruptible = true,
8348 	};
8349 	int status;
8350 
8351 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8352 		(unsigned long long)server->fsid.major,
8353 		(unsigned long long)server->fsid.minor,
8354 		clp->cl_hostname);
8355 	nfs_display_fhandle(fhandle, __func__);
8356 
8357 	do {
8358 		status = ops->get_locations(server, fhandle, locations, page,
8359 					    cred);
8360 		if (status != -NFS4ERR_DELAY)
8361 			break;
8362 		nfs4_handle_exception(server, status, &exception);
8363 	} while (exception.retry);
8364 	return status;
8365 }
8366 
8367 /*
8368  * This operation also signals the server that this client is
8369  * performing "lease moved" recovery.  The server can stop
8370  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8371  * is appended to this compound to identify the client ID which is
8372  * performing recovery.
8373  */
8374 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8375 {
8376 	struct nfs_server *server = NFS_SERVER(inode);
8377 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8378 	struct rpc_clnt *clnt = server->client;
8379 	struct nfs4_fsid_present_arg args = {
8380 		.fh		= NFS_FH(inode),
8381 		.clientid	= clp->cl_clientid,
8382 		.renew		= 1,		/* append RENEW */
8383 	};
8384 	struct nfs4_fsid_present_res res = {
8385 		.renew		= 1,
8386 	};
8387 	struct rpc_message msg = {
8388 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8389 		.rpc_argp	= &args,
8390 		.rpc_resp	= &res,
8391 		.rpc_cred	= cred,
8392 	};
8393 	unsigned long now = jiffies;
8394 	int status;
8395 
8396 	res.fh = nfs_alloc_fhandle();
8397 	if (res.fh == NULL)
8398 		return -ENOMEM;
8399 
8400 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8401 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8402 						&args.seq_args, &res.seq_res);
8403 	nfs_free_fhandle(res.fh);
8404 	if (status)
8405 		return status;
8406 
8407 	do_renew_lease(clp, now);
8408 	return 0;
8409 }
8410 
8411 #ifdef CONFIG_NFS_V4_1
8412 
8413 /*
8414  * This operation also signals the server that this client is
8415  * performing "lease moved" recovery.  The server can stop asserting
8416  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8417  * this operation is identified in the SEQUENCE operation in this
8418  * compound.
8419  */
8420 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8421 {
8422 	struct nfs_server *server = NFS_SERVER(inode);
8423 	struct rpc_clnt *clnt = server->client;
8424 	struct nfs4_fsid_present_arg args = {
8425 		.fh		= NFS_FH(inode),
8426 	};
8427 	struct nfs4_fsid_present_res res = {
8428 	};
8429 	struct rpc_message msg = {
8430 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8431 		.rpc_argp	= &args,
8432 		.rpc_resp	= &res,
8433 		.rpc_cred	= cred,
8434 	};
8435 	int status;
8436 
8437 	res.fh = nfs_alloc_fhandle();
8438 	if (res.fh == NULL)
8439 		return -ENOMEM;
8440 
8441 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8442 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8443 						&args.seq_args, &res.seq_res);
8444 	nfs_free_fhandle(res.fh);
8445 	if (status == NFS4_OK &&
8446 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8447 		status = -NFS4ERR_LEASE_MOVED;
8448 	return status;
8449 }
8450 
8451 #endif	/* CONFIG_NFS_V4_1 */
8452 
8453 /**
8454  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8455  * @inode: inode on FSID to check
8456  * @cred: credential to use for this operation
8457  *
8458  * Server indicates whether the FSID is present, moved, or not
8459  * recognized.  This operation is necessary to clear a LEASE_MOVED
8460  * condition for this client ID.
8461  *
8462  * Returns NFS4_OK if the FSID is present on this server,
8463  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8464  *  NFS4ERR code if some error occurred on the server, or a
8465  *  negative errno if a local failure occurred.
8466  */
8467 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8468 {
8469 	struct nfs_server *server = NFS_SERVER(inode);
8470 	struct nfs_client *clp = server->nfs_client;
8471 	const struct nfs4_mig_recovery_ops *ops =
8472 					clp->cl_mvops->mig_recovery_ops;
8473 	struct nfs4_exception exception = {
8474 		.interruptible = true,
8475 	};
8476 	int status;
8477 
8478 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8479 		(unsigned long long)server->fsid.major,
8480 		(unsigned long long)server->fsid.minor,
8481 		clp->cl_hostname);
8482 	nfs_display_fhandle(NFS_FH(inode), __func__);
8483 
8484 	do {
8485 		status = ops->fsid_present(inode, cred);
8486 		if (status != -NFS4ERR_DELAY)
8487 			break;
8488 		nfs4_handle_exception(server, status, &exception);
8489 	} while (exception.retry);
8490 	return status;
8491 }
8492 
8493 /*
8494  * If 'use_integrity' is true and the state managment nfs_client
8495  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8496  * and the machine credential as per RFC3530bis and RFC5661 Security
8497  * Considerations sections. Otherwise, just use the user cred with the
8498  * filesystem's rpc_client.
8499  */
8500 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8501 {
8502 	int status;
8503 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8504 	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8505 	struct nfs4_secinfo_arg args = {
8506 		.dir_fh = NFS_FH(dir),
8507 		.name   = name,
8508 	};
8509 	struct nfs4_secinfo_res res = {
8510 		.flavors     = flavors,
8511 	};
8512 	struct rpc_message msg = {
8513 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8514 		.rpc_argp = &args,
8515 		.rpc_resp = &res,
8516 	};
8517 	struct nfs4_call_sync_data data = {
8518 		.seq_server = NFS_SERVER(dir),
8519 		.seq_args = &args.seq_args,
8520 		.seq_res = &res.seq_res,
8521 	};
8522 	struct rpc_task_setup task_setup = {
8523 		.rpc_client = clnt,
8524 		.rpc_message = &msg,
8525 		.callback_ops = clp->cl_mvops->call_sync_ops,
8526 		.callback_data = &data,
8527 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8528 	};
8529 	const struct cred *cred = NULL;
8530 
8531 	if (use_integrity) {
8532 		clnt = clp->cl_rpcclient;
8533 		task_setup.rpc_client = clnt;
8534 
8535 		cred = nfs4_get_clid_cred(clp);
8536 		msg.rpc_cred = cred;
8537 	}
8538 
8539 	dprintk("NFS call  secinfo %s\n", name->name);
8540 
8541 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8542 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8543 	status = nfs4_call_sync_custom(&task_setup);
8544 
8545 	dprintk("NFS reply  secinfo: %d\n", status);
8546 
8547 	put_cred(cred);
8548 	return status;
8549 }
8550 
8551 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8552 		      struct nfs4_secinfo_flavors *flavors)
8553 {
8554 	struct nfs4_exception exception = {
8555 		.interruptible = true,
8556 	};
8557 	int err;
8558 	do {
8559 		err = -NFS4ERR_WRONGSEC;
8560 
8561 		/* try to use integrity protection with machine cred */
8562 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8563 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8564 
8565 		/*
8566 		 * if unable to use integrity protection, or SECINFO with
8567 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8568 		 * disallowed by spec, but exists in deployed servers) use
8569 		 * the current filesystem's rpc_client and the user cred.
8570 		 */
8571 		if (err == -NFS4ERR_WRONGSEC)
8572 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8573 
8574 		trace_nfs4_secinfo(dir, name, err);
8575 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8576 				&exception);
8577 	} while (exception.retry);
8578 	return err;
8579 }
8580 
8581 #ifdef CONFIG_NFS_V4_1
8582 /*
8583  * Check the exchange flags returned by the server for invalid flags, having
8584  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8585  * DS flags set.
8586  */
8587 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8588 {
8589 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8590 		goto out_inval;
8591 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8592 		goto out_inval;
8593 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8594 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8595 		goto out_inval;
8596 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8597 		goto out_inval;
8598 	return NFS_OK;
8599 out_inval:
8600 	return -NFS4ERR_INVAL;
8601 }
8602 
8603 static bool
8604 nfs41_same_server_scope(struct nfs41_server_scope *a,
8605 			struct nfs41_server_scope *b)
8606 {
8607 	if (a->server_scope_sz != b->server_scope_sz)
8608 		return false;
8609 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8610 }
8611 
8612 static void
8613 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8614 {
8615 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8616 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8617 	struct nfs_client *clp = args->client;
8618 
8619 	switch (task->tk_status) {
8620 	case -NFS4ERR_BADSESSION:
8621 	case -NFS4ERR_DEADSESSION:
8622 		nfs4_schedule_session_recovery(clp->cl_session,
8623 				task->tk_status);
8624 		return;
8625 	}
8626 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8627 			res->dir != NFS4_CDFS4_BOTH) {
8628 		rpc_task_close_connection(task);
8629 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8630 			rpc_restart_call(task);
8631 	}
8632 }
8633 
8634 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8635 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8636 };
8637 
8638 /*
8639  * nfs4_proc_bind_one_conn_to_session()
8640  *
8641  * The 4.1 client currently uses the same TCP connection for the
8642  * fore and backchannel.
8643  */
8644 static
8645 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8646 		struct rpc_xprt *xprt,
8647 		struct nfs_client *clp,
8648 		const struct cred *cred)
8649 {
8650 	int status;
8651 	struct nfs41_bind_conn_to_session_args args = {
8652 		.client = clp,
8653 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8654 		.retries = 0,
8655 	};
8656 	struct nfs41_bind_conn_to_session_res res;
8657 	struct rpc_message msg = {
8658 		.rpc_proc =
8659 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8660 		.rpc_argp = &args,
8661 		.rpc_resp = &res,
8662 		.rpc_cred = cred,
8663 	};
8664 	struct rpc_task_setup task_setup_data = {
8665 		.rpc_client = clnt,
8666 		.rpc_xprt = xprt,
8667 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8668 		.rpc_message = &msg,
8669 		.flags = RPC_TASK_TIMEOUT,
8670 	};
8671 	struct rpc_task *task;
8672 
8673 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8674 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8675 		args.dir = NFS4_CDFC4_FORE;
8676 
8677 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8678 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8679 		args.dir = NFS4_CDFC4_FORE;
8680 
8681 	task = rpc_run_task(&task_setup_data);
8682 	if (!IS_ERR(task)) {
8683 		status = task->tk_status;
8684 		rpc_put_task(task);
8685 	} else
8686 		status = PTR_ERR(task);
8687 	trace_nfs4_bind_conn_to_session(clp, status);
8688 	if (status == 0) {
8689 		if (memcmp(res.sessionid.data,
8690 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8691 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8692 			return -EIO;
8693 		}
8694 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8695 			dprintk("NFS: %s: Unexpected direction from server\n",
8696 				__func__);
8697 			return -EIO;
8698 		}
8699 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8700 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8701 				__func__);
8702 			return -EIO;
8703 		}
8704 	}
8705 
8706 	return status;
8707 }
8708 
8709 struct rpc_bind_conn_calldata {
8710 	struct nfs_client *clp;
8711 	const struct cred *cred;
8712 };
8713 
8714 static int
8715 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8716 		struct rpc_xprt *xprt,
8717 		void *calldata)
8718 {
8719 	struct rpc_bind_conn_calldata *p = calldata;
8720 
8721 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8722 }
8723 
8724 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8725 {
8726 	struct rpc_bind_conn_calldata data = {
8727 		.clp = clp,
8728 		.cred = cred,
8729 	};
8730 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8731 			nfs4_proc_bind_conn_to_session_callback, &data);
8732 }
8733 
8734 /*
8735  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8736  * and operations we'd like to see to enable certain features in the allow map
8737  */
8738 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8739 	.how = SP4_MACH_CRED,
8740 	.enforce.u.words = {
8741 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8742 		      1 << (OP_EXCHANGE_ID - 32) |
8743 		      1 << (OP_CREATE_SESSION - 32) |
8744 		      1 << (OP_DESTROY_SESSION - 32) |
8745 		      1 << (OP_DESTROY_CLIENTID - 32)
8746 	},
8747 	.allow.u.words = {
8748 		[0] = 1 << (OP_CLOSE) |
8749 		      1 << (OP_OPEN_DOWNGRADE) |
8750 		      1 << (OP_LOCKU) |
8751 		      1 << (OP_DELEGRETURN) |
8752 		      1 << (OP_COMMIT),
8753 		[1] = 1 << (OP_SECINFO - 32) |
8754 		      1 << (OP_SECINFO_NO_NAME - 32) |
8755 		      1 << (OP_LAYOUTRETURN - 32) |
8756 		      1 << (OP_TEST_STATEID - 32) |
8757 		      1 << (OP_FREE_STATEID - 32) |
8758 		      1 << (OP_WRITE - 32)
8759 	}
8760 };
8761 
8762 /*
8763  * Select the state protection mode for client `clp' given the server results
8764  * from exchange_id in `sp'.
8765  *
8766  * Returns 0 on success, negative errno otherwise.
8767  */
8768 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8769 				 struct nfs41_state_protection *sp)
8770 {
8771 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8772 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8773 		      1 << (OP_EXCHANGE_ID - 32) |
8774 		      1 << (OP_CREATE_SESSION - 32) |
8775 		      1 << (OP_DESTROY_SESSION - 32) |
8776 		      1 << (OP_DESTROY_CLIENTID - 32)
8777 	};
8778 	unsigned long flags = 0;
8779 	unsigned int i;
8780 	int ret = 0;
8781 
8782 	if (sp->how == SP4_MACH_CRED) {
8783 		/* Print state protect result */
8784 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8785 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8786 			if (test_bit(i, sp->enforce.u.longs))
8787 				dfprintk(MOUNT, "  enforce op %d\n", i);
8788 			if (test_bit(i, sp->allow.u.longs))
8789 				dfprintk(MOUNT, "  allow op %d\n", i);
8790 		}
8791 
8792 		/* make sure nothing is on enforce list that isn't supported */
8793 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8794 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8795 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8796 				ret = -EINVAL;
8797 				goto out;
8798 			}
8799 		}
8800 
8801 		/*
8802 		 * Minimal mode - state operations are allowed to use machine
8803 		 * credential.  Note this already happens by default, so the
8804 		 * client doesn't have to do anything more than the negotiation.
8805 		 *
8806 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8807 		 *       we're already using the machine cred for exchange_id
8808 		 *       and will never use a different cred.
8809 		 */
8810 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8811 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8812 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8813 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8814 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8815 			dfprintk(MOUNT, "  minimal mode enabled\n");
8816 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8817 		} else {
8818 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8819 			ret = -EINVAL;
8820 			goto out;
8821 		}
8822 
8823 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8824 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8825 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8826 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8827 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8828 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8829 		}
8830 
8831 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8832 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8833 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8834 		}
8835 
8836 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8837 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8838 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8839 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8840 		}
8841 
8842 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8843 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8844 			dfprintk(MOUNT, "  stateid mode enabled\n");
8845 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8846 		}
8847 
8848 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8849 			dfprintk(MOUNT, "  write mode enabled\n");
8850 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8851 		}
8852 
8853 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8854 			dfprintk(MOUNT, "  commit mode enabled\n");
8855 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8856 		}
8857 	}
8858 out:
8859 	clp->cl_sp4_flags = flags;
8860 	return ret;
8861 }
8862 
8863 struct nfs41_exchange_id_data {
8864 	struct nfs41_exchange_id_res res;
8865 	struct nfs41_exchange_id_args args;
8866 };
8867 
8868 static void nfs4_exchange_id_release(void *data)
8869 {
8870 	struct nfs41_exchange_id_data *cdata =
8871 					(struct nfs41_exchange_id_data *)data;
8872 
8873 	nfs_put_client(cdata->args.client);
8874 	kfree(cdata->res.impl_id);
8875 	kfree(cdata->res.server_scope);
8876 	kfree(cdata->res.server_owner);
8877 	kfree(cdata);
8878 }
8879 
8880 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8881 	.rpc_release = nfs4_exchange_id_release,
8882 };
8883 
8884 /*
8885  * _nfs4_proc_exchange_id()
8886  *
8887  * Wrapper for EXCHANGE_ID operation.
8888  */
8889 static struct rpc_task *
8890 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8891 			u32 sp4_how, struct rpc_xprt *xprt)
8892 {
8893 	struct rpc_message msg = {
8894 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8895 		.rpc_cred = cred,
8896 	};
8897 	struct rpc_task_setup task_setup_data = {
8898 		.rpc_client = clp->cl_rpcclient,
8899 		.callback_ops = &nfs4_exchange_id_call_ops,
8900 		.rpc_message = &msg,
8901 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8902 	};
8903 	struct nfs41_exchange_id_data *calldata;
8904 	int status;
8905 
8906 	if (!refcount_inc_not_zero(&clp->cl_count))
8907 		return ERR_PTR(-EIO);
8908 
8909 	status = -ENOMEM;
8910 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8911 	if (!calldata)
8912 		goto out;
8913 
8914 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8915 
8916 	status = nfs4_init_uniform_client_string(clp);
8917 	if (status)
8918 		goto out_calldata;
8919 
8920 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8921 						GFP_NOFS);
8922 	status = -ENOMEM;
8923 	if (unlikely(calldata->res.server_owner == NULL))
8924 		goto out_calldata;
8925 
8926 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8927 					GFP_NOFS);
8928 	if (unlikely(calldata->res.server_scope == NULL))
8929 		goto out_server_owner;
8930 
8931 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8932 	if (unlikely(calldata->res.impl_id == NULL))
8933 		goto out_server_scope;
8934 
8935 	switch (sp4_how) {
8936 	case SP4_NONE:
8937 		calldata->args.state_protect.how = SP4_NONE;
8938 		break;
8939 
8940 	case SP4_MACH_CRED:
8941 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8942 		break;
8943 
8944 	default:
8945 		/* unsupported! */
8946 		WARN_ON_ONCE(1);
8947 		status = -EINVAL;
8948 		goto out_impl_id;
8949 	}
8950 	if (xprt) {
8951 		task_setup_data.rpc_xprt = xprt;
8952 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8953 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8954 				sizeof(calldata->args.verifier.data));
8955 	}
8956 	calldata->args.client = clp;
8957 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8958 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8959 #ifdef CONFIG_NFS_V4_1_MIGRATION
8960 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8961 #endif
8962 	if (test_bit(NFS_CS_PNFS, &clp->cl_flags))
8963 		calldata->args.flags |= EXCHGID4_FLAG_USE_PNFS_DS;
8964 	msg.rpc_argp = &calldata->args;
8965 	msg.rpc_resp = &calldata->res;
8966 	task_setup_data.callback_data = calldata;
8967 
8968 	return rpc_run_task(&task_setup_data);
8969 
8970 out_impl_id:
8971 	kfree(calldata->res.impl_id);
8972 out_server_scope:
8973 	kfree(calldata->res.server_scope);
8974 out_server_owner:
8975 	kfree(calldata->res.server_owner);
8976 out_calldata:
8977 	kfree(calldata);
8978 out:
8979 	nfs_put_client(clp);
8980 	return ERR_PTR(status);
8981 }
8982 
8983 /*
8984  * _nfs4_proc_exchange_id()
8985  *
8986  * Wrapper for EXCHANGE_ID operation.
8987  */
8988 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8989 			u32 sp4_how)
8990 {
8991 	struct rpc_task *task;
8992 	struct nfs41_exchange_id_args *argp;
8993 	struct nfs41_exchange_id_res *resp;
8994 	unsigned long now = jiffies;
8995 	int status;
8996 
8997 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8998 	if (IS_ERR(task))
8999 		return PTR_ERR(task);
9000 
9001 	argp = task->tk_msg.rpc_argp;
9002 	resp = task->tk_msg.rpc_resp;
9003 	status = task->tk_status;
9004 	if (status  != 0)
9005 		goto out;
9006 
9007 	status = nfs4_check_cl_exchange_flags(resp->flags,
9008 			clp->cl_mvops->minor_version);
9009 	if (status  != 0)
9010 		goto out;
9011 
9012 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
9013 	if (status != 0)
9014 		goto out;
9015 
9016 	do_renew_lease(clp, now);
9017 
9018 	clp->cl_clientid = resp->clientid;
9019 	clp->cl_exchange_flags = resp->flags;
9020 	clp->cl_seqid = resp->seqid;
9021 	/* Client ID is not confirmed */
9022 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
9023 		clear_bit(NFS4_SESSION_ESTABLISHED,
9024 			  &clp->cl_session->session_state);
9025 
9026 	if (clp->cl_serverscope != NULL &&
9027 	    !nfs41_same_server_scope(clp->cl_serverscope,
9028 				resp->server_scope)) {
9029 		dprintk("%s: server_scope mismatch detected\n",
9030 			__func__);
9031 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
9032 	}
9033 
9034 	swap(clp->cl_serverowner, resp->server_owner);
9035 	swap(clp->cl_serverscope, resp->server_scope);
9036 	swap(clp->cl_implid, resp->impl_id);
9037 
9038 	/* Save the EXCHANGE_ID verifier session trunk tests */
9039 	memcpy(clp->cl_confirm.data, argp->verifier.data,
9040 	       sizeof(clp->cl_confirm.data));
9041 out:
9042 	trace_nfs4_exchange_id(clp, status);
9043 	rpc_put_task(task);
9044 	return status;
9045 }
9046 
9047 /*
9048  * nfs4_proc_exchange_id()
9049  *
9050  * Returns zero, a negative errno, or a negative NFS4ERR status code.
9051  *
9052  * Since the clientid has expired, all compounds using sessions
9053  * associated with the stale clientid will be returning
9054  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
9055  * be in some phase of session reset.
9056  *
9057  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
9058  */
9059 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
9060 {
9061 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
9062 	int status;
9063 
9064 	/* try SP4_MACH_CRED if krb5i/p	*/
9065 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
9066 	    authflavor == RPC_AUTH_GSS_KRB5P) {
9067 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
9068 		if (!status)
9069 			return 0;
9070 	}
9071 
9072 	/* try SP4_NONE */
9073 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
9074 }
9075 
9076 /**
9077  * nfs4_test_session_trunk
9078  *
9079  * This is an add_xprt_test() test function called from
9080  * rpc_clnt_setup_test_and_add_xprt.
9081  *
9082  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
9083  * and is dereferrenced in nfs4_exchange_id_release
9084  *
9085  * Upon success, add the new transport to the rpc_clnt
9086  *
9087  * @clnt: struct rpc_clnt to get new transport
9088  * @xprt: the rpc_xprt to test
9089  * @data: call data for _nfs4_proc_exchange_id.
9090  */
9091 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
9092 			    void *data)
9093 {
9094 	struct nfs4_add_xprt_data *adata = data;
9095 	struct rpc_task *task;
9096 	int status;
9097 
9098 	u32 sp4_how;
9099 
9100 	dprintk("--> %s try %s\n", __func__,
9101 		xprt->address_strings[RPC_DISPLAY_ADDR]);
9102 
9103 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
9104 
9105 try_again:
9106 	/* Test connection for session trunking. Async exchange_id call */
9107 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
9108 	if (IS_ERR(task))
9109 		return;
9110 
9111 	status = task->tk_status;
9112 	if (status == 0) {
9113 		status = nfs4_detect_session_trunking(adata->clp,
9114 				task->tk_msg.rpc_resp, xprt);
9115 		trace_nfs4_trunked_exchange_id(adata->clp,
9116 			xprt->address_strings[RPC_DISPLAY_ADDR], status);
9117 	}
9118 	if (status == 0)
9119 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
9120 	else if (status != -NFS4ERR_DELAY && rpc_clnt_xprt_switch_has_addr(clnt,
9121 				(struct sockaddr *)&xprt->addr))
9122 		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
9123 
9124 	rpc_put_task(task);
9125 	if (status == -NFS4ERR_DELAY) {
9126 		ssleep(1);
9127 		goto try_again;
9128 	}
9129 }
9130 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
9131 
9132 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
9133 		const struct cred *cred)
9134 {
9135 	struct rpc_message msg = {
9136 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
9137 		.rpc_argp = clp,
9138 		.rpc_cred = cred,
9139 	};
9140 	int status;
9141 
9142 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
9143 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9144 	trace_nfs4_destroy_clientid(clp, status);
9145 	if (status)
9146 		dprintk("NFS: Got error %d from the server %s on "
9147 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
9148 	return status;
9149 }
9150 
9151 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
9152 		const struct cred *cred)
9153 {
9154 	unsigned int loop;
9155 	int ret;
9156 
9157 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
9158 		ret = _nfs4_proc_destroy_clientid(clp, cred);
9159 		switch (ret) {
9160 		case -NFS4ERR_DELAY:
9161 		case -NFS4ERR_CLIENTID_BUSY:
9162 			ssleep(1);
9163 			break;
9164 		default:
9165 			return ret;
9166 		}
9167 	}
9168 	return 0;
9169 }
9170 
9171 int nfs4_destroy_clientid(struct nfs_client *clp)
9172 {
9173 	const struct cred *cred;
9174 	int ret = 0;
9175 
9176 	if (clp->cl_mvops->minor_version < 1)
9177 		goto out;
9178 	if (clp->cl_exchange_flags == 0)
9179 		goto out;
9180 	if (clp->cl_preserve_clid)
9181 		goto out;
9182 	cred = nfs4_get_clid_cred(clp);
9183 	ret = nfs4_proc_destroy_clientid(clp, cred);
9184 	put_cred(cred);
9185 	switch (ret) {
9186 	case 0:
9187 	case -NFS4ERR_STALE_CLIENTID:
9188 		clp->cl_exchange_flags = 0;
9189 	}
9190 out:
9191 	return ret;
9192 }
9193 
9194 #endif /* CONFIG_NFS_V4_1 */
9195 
9196 struct nfs4_get_lease_time_data {
9197 	struct nfs4_get_lease_time_args *args;
9198 	struct nfs4_get_lease_time_res *res;
9199 	struct nfs_client *clp;
9200 };
9201 
9202 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9203 					void *calldata)
9204 {
9205 	struct nfs4_get_lease_time_data *data =
9206 			(struct nfs4_get_lease_time_data *)calldata;
9207 
9208 	/* just setup sequence, do not trigger session recovery
9209 	   since we're invoked within one */
9210 	nfs4_setup_sequence(data->clp,
9211 			&data->args->la_seq_args,
9212 			&data->res->lr_seq_res,
9213 			task);
9214 }
9215 
9216 /*
9217  * Called from nfs4_state_manager thread for session setup, so don't recover
9218  * from sequence operation or clientid errors.
9219  */
9220 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9221 {
9222 	struct nfs4_get_lease_time_data *data =
9223 			(struct nfs4_get_lease_time_data *)calldata;
9224 
9225 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9226 		return;
9227 	switch (task->tk_status) {
9228 	case -NFS4ERR_DELAY:
9229 	case -NFS4ERR_GRACE:
9230 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9231 		task->tk_status = 0;
9232 		fallthrough;
9233 	case -NFS4ERR_RETRY_UNCACHED_REP:
9234 		rpc_restart_call_prepare(task);
9235 		return;
9236 	}
9237 }
9238 
9239 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9240 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9241 	.rpc_call_done = nfs4_get_lease_time_done,
9242 };
9243 
9244 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9245 {
9246 	struct nfs4_get_lease_time_args args;
9247 	struct nfs4_get_lease_time_res res = {
9248 		.lr_fsinfo = fsinfo,
9249 	};
9250 	struct nfs4_get_lease_time_data data = {
9251 		.args = &args,
9252 		.res = &res,
9253 		.clp = clp,
9254 	};
9255 	struct rpc_message msg = {
9256 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9257 		.rpc_argp = &args,
9258 		.rpc_resp = &res,
9259 	};
9260 	struct rpc_task_setup task_setup = {
9261 		.rpc_client = clp->cl_rpcclient,
9262 		.rpc_message = &msg,
9263 		.callback_ops = &nfs4_get_lease_time_ops,
9264 		.callback_data = &data,
9265 		.flags = RPC_TASK_TIMEOUT,
9266 	};
9267 
9268 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9269 	return nfs4_call_sync_custom(&task_setup);
9270 }
9271 
9272 #ifdef CONFIG_NFS_V4_1
9273 
9274 /*
9275  * Initialize the values to be used by the client in CREATE_SESSION
9276  * If nfs4_init_session set the fore channel request and response sizes,
9277  * use them.
9278  *
9279  * Set the back channel max_resp_sz_cached to zero to force the client to
9280  * always set csa_cachethis to FALSE because the current implementation
9281  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9282  */
9283 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9284 				    struct rpc_clnt *clnt)
9285 {
9286 	unsigned int max_rqst_sz, max_resp_sz;
9287 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9288 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9289 
9290 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9291 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9292 
9293 	/* Fore channel attributes */
9294 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9295 	args->fc_attrs.max_resp_sz = max_resp_sz;
9296 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9297 	args->fc_attrs.max_reqs = max_session_slots;
9298 
9299 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9300 		"max_ops=%u max_reqs=%u\n",
9301 		__func__,
9302 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9303 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9304 
9305 	/* Back channel attributes */
9306 	args->bc_attrs.max_rqst_sz = max_bc_payload;
9307 	args->bc_attrs.max_resp_sz = max_bc_payload;
9308 	args->bc_attrs.max_resp_sz_cached = 0;
9309 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9310 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9311 	if (args->bc_attrs.max_reqs > max_bc_slots)
9312 		args->bc_attrs.max_reqs = max_bc_slots;
9313 
9314 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9315 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9316 		__func__,
9317 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9318 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9319 		args->bc_attrs.max_reqs);
9320 }
9321 
9322 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9323 		struct nfs41_create_session_res *res)
9324 {
9325 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9326 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9327 
9328 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9329 		return -EINVAL;
9330 	/*
9331 	 * Our requested max_ops is the minimum we need; we're not
9332 	 * prepared to break up compounds into smaller pieces than that.
9333 	 * So, no point even trying to continue if the server won't
9334 	 * cooperate:
9335 	 */
9336 	if (rcvd->max_ops < sent->max_ops)
9337 		return -EINVAL;
9338 	if (rcvd->max_reqs == 0)
9339 		return -EINVAL;
9340 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9341 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9342 	return 0;
9343 }
9344 
9345 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9346 		struct nfs41_create_session_res *res)
9347 {
9348 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9349 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9350 
9351 	if (!(res->flags & SESSION4_BACK_CHAN))
9352 		goto out;
9353 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9354 		return -EINVAL;
9355 	if (rcvd->max_resp_sz < sent->max_resp_sz)
9356 		return -EINVAL;
9357 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9358 		return -EINVAL;
9359 	if (rcvd->max_ops > sent->max_ops)
9360 		return -EINVAL;
9361 	if (rcvd->max_reqs > sent->max_reqs)
9362 		return -EINVAL;
9363 out:
9364 	return 0;
9365 }
9366 
9367 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9368 				     struct nfs41_create_session_res *res)
9369 {
9370 	int ret;
9371 
9372 	ret = nfs4_verify_fore_channel_attrs(args, res);
9373 	if (ret)
9374 		return ret;
9375 	return nfs4_verify_back_channel_attrs(args, res);
9376 }
9377 
9378 static void nfs4_update_session(struct nfs4_session *session,
9379 		struct nfs41_create_session_res *res)
9380 {
9381 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9382 	/* Mark client id and session as being confirmed */
9383 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9384 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9385 	session->flags = res->flags;
9386 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9387 	if (res->flags & SESSION4_BACK_CHAN)
9388 		memcpy(&session->bc_attrs, &res->bc_attrs,
9389 				sizeof(session->bc_attrs));
9390 }
9391 
9392 static int _nfs4_proc_create_session(struct nfs_client *clp,
9393 		const struct cred *cred)
9394 {
9395 	struct nfs4_session *session = clp->cl_session;
9396 	struct nfs41_create_session_args args = {
9397 		.client = clp,
9398 		.clientid = clp->cl_clientid,
9399 		.seqid = clp->cl_seqid,
9400 		.cb_program = NFS4_CALLBACK,
9401 	};
9402 	struct nfs41_create_session_res res;
9403 
9404 	struct rpc_message msg = {
9405 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9406 		.rpc_argp = &args,
9407 		.rpc_resp = &res,
9408 		.rpc_cred = cred,
9409 	};
9410 	int status;
9411 
9412 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9413 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9414 
9415 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9416 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9417 	trace_nfs4_create_session(clp, status);
9418 
9419 	switch (status) {
9420 	case -NFS4ERR_STALE_CLIENTID:
9421 	case -NFS4ERR_DELAY:
9422 	case -ETIMEDOUT:
9423 	case -EACCES:
9424 	case -EAGAIN:
9425 		goto out;
9426 	}
9427 
9428 	clp->cl_seqid++;
9429 	if (!status) {
9430 		/* Verify the session's negotiated channel_attrs values */
9431 		status = nfs4_verify_channel_attrs(&args, &res);
9432 		/* Increment the clientid slot sequence id */
9433 		if (status)
9434 			goto out;
9435 		nfs4_update_session(session, &res);
9436 	}
9437 out:
9438 	return status;
9439 }
9440 
9441 /*
9442  * Issues a CREATE_SESSION operation to the server.
9443  * It is the responsibility of the caller to verify the session is
9444  * expired before calling this routine.
9445  */
9446 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9447 {
9448 	int status;
9449 	unsigned *ptr;
9450 	struct nfs4_session *session = clp->cl_session;
9451 	struct nfs4_add_xprt_data xprtdata = {
9452 		.clp = clp,
9453 	};
9454 	struct rpc_add_xprt_test rpcdata = {
9455 		.add_xprt_test = clp->cl_mvops->session_trunk,
9456 		.data = &xprtdata,
9457 	};
9458 
9459 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9460 
9461 	status = _nfs4_proc_create_session(clp, cred);
9462 	if (status)
9463 		goto out;
9464 
9465 	/* Init or reset the session slot tables */
9466 	status = nfs4_setup_session_slot_tables(session);
9467 	dprintk("slot table setup returned %d\n", status);
9468 	if (status)
9469 		goto out;
9470 
9471 	ptr = (unsigned *)&session->sess_id.data[0];
9472 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9473 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9474 	rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9475 out:
9476 	return status;
9477 }
9478 
9479 /*
9480  * Issue the over-the-wire RPC DESTROY_SESSION.
9481  * The caller must serialize access to this routine.
9482  */
9483 int nfs4_proc_destroy_session(struct nfs4_session *session,
9484 		const struct cred *cred)
9485 {
9486 	struct rpc_message msg = {
9487 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9488 		.rpc_argp = session,
9489 		.rpc_cred = cred,
9490 	};
9491 	int status = 0;
9492 
9493 	/* session is still being setup */
9494 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9495 		return 0;
9496 
9497 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9498 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9499 	trace_nfs4_destroy_session(session->clp, status);
9500 
9501 	if (status)
9502 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9503 			"Session has been destroyed regardless...\n", status);
9504 	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9505 	return status;
9506 }
9507 
9508 /*
9509  * Renew the cl_session lease.
9510  */
9511 struct nfs4_sequence_data {
9512 	struct nfs_client *clp;
9513 	struct nfs4_sequence_args args;
9514 	struct nfs4_sequence_res res;
9515 };
9516 
9517 static void nfs41_sequence_release(void *data)
9518 {
9519 	struct nfs4_sequence_data *calldata = data;
9520 	struct nfs_client *clp = calldata->clp;
9521 
9522 	if (refcount_read(&clp->cl_count) > 1)
9523 		nfs4_schedule_state_renewal(clp);
9524 	nfs_put_client(clp);
9525 	kfree(calldata);
9526 }
9527 
9528 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9529 {
9530 	switch(task->tk_status) {
9531 	case -NFS4ERR_DELAY:
9532 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9533 		return -EAGAIN;
9534 	default:
9535 		nfs4_schedule_lease_recovery(clp);
9536 	}
9537 	return 0;
9538 }
9539 
9540 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9541 {
9542 	struct nfs4_sequence_data *calldata = data;
9543 	struct nfs_client *clp = calldata->clp;
9544 
9545 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9546 		return;
9547 
9548 	trace_nfs4_sequence(clp, task->tk_status);
9549 	if (task->tk_status < 0 && !task->tk_client->cl_shutdown) {
9550 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9551 		if (refcount_read(&clp->cl_count) == 1)
9552 			return;
9553 
9554 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9555 			rpc_restart_call_prepare(task);
9556 			return;
9557 		}
9558 	}
9559 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9560 }
9561 
9562 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9563 {
9564 	struct nfs4_sequence_data *calldata = data;
9565 	struct nfs_client *clp = calldata->clp;
9566 	struct nfs4_sequence_args *args;
9567 	struct nfs4_sequence_res *res;
9568 
9569 	args = task->tk_msg.rpc_argp;
9570 	res = task->tk_msg.rpc_resp;
9571 
9572 	nfs4_setup_sequence(clp, args, res, task);
9573 }
9574 
9575 static const struct rpc_call_ops nfs41_sequence_ops = {
9576 	.rpc_call_done = nfs41_sequence_call_done,
9577 	.rpc_call_prepare = nfs41_sequence_prepare,
9578 	.rpc_release = nfs41_sequence_release,
9579 };
9580 
9581 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9582 		const struct cred *cred,
9583 		struct nfs4_slot *slot,
9584 		bool is_privileged)
9585 {
9586 	struct nfs4_sequence_data *calldata;
9587 	struct rpc_message msg = {
9588 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9589 		.rpc_cred = cred,
9590 	};
9591 	struct rpc_task_setup task_setup_data = {
9592 		.rpc_client = clp->cl_rpcclient,
9593 		.rpc_message = &msg,
9594 		.callback_ops = &nfs41_sequence_ops,
9595 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9596 	};
9597 	struct rpc_task *ret;
9598 
9599 	ret = ERR_PTR(-EIO);
9600 	if (!refcount_inc_not_zero(&clp->cl_count))
9601 		goto out_err;
9602 
9603 	ret = ERR_PTR(-ENOMEM);
9604 	calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9605 	if (calldata == NULL)
9606 		goto out_put_clp;
9607 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9608 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9609 	msg.rpc_argp = &calldata->args;
9610 	msg.rpc_resp = &calldata->res;
9611 	calldata->clp = clp;
9612 	task_setup_data.callback_data = calldata;
9613 
9614 	ret = rpc_run_task(&task_setup_data);
9615 	if (IS_ERR(ret))
9616 		goto out_err;
9617 	return ret;
9618 out_put_clp:
9619 	nfs_put_client(clp);
9620 out_err:
9621 	nfs41_release_slot(slot);
9622 	return ret;
9623 }
9624 
9625 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9626 {
9627 	struct rpc_task *task;
9628 	int ret = 0;
9629 
9630 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9631 		return -EAGAIN;
9632 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9633 	if (IS_ERR(task))
9634 		ret = PTR_ERR(task);
9635 	else
9636 		rpc_put_task_async(task);
9637 	dprintk("<-- %s status=%d\n", __func__, ret);
9638 	return ret;
9639 }
9640 
9641 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9642 {
9643 	struct rpc_task *task;
9644 	int ret;
9645 
9646 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9647 	if (IS_ERR(task)) {
9648 		ret = PTR_ERR(task);
9649 		goto out;
9650 	}
9651 	ret = rpc_wait_for_completion_task(task);
9652 	if (!ret)
9653 		ret = task->tk_status;
9654 	rpc_put_task(task);
9655 out:
9656 	dprintk("<-- %s status=%d\n", __func__, ret);
9657 	return ret;
9658 }
9659 
9660 struct nfs4_reclaim_complete_data {
9661 	struct nfs_client *clp;
9662 	struct nfs41_reclaim_complete_args arg;
9663 	struct nfs41_reclaim_complete_res res;
9664 };
9665 
9666 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9667 {
9668 	struct nfs4_reclaim_complete_data *calldata = data;
9669 
9670 	nfs4_setup_sequence(calldata->clp,
9671 			&calldata->arg.seq_args,
9672 			&calldata->res.seq_res,
9673 			task);
9674 }
9675 
9676 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9677 {
9678 	switch(task->tk_status) {
9679 	case 0:
9680 		wake_up_all(&clp->cl_lock_waitq);
9681 		fallthrough;
9682 	case -NFS4ERR_COMPLETE_ALREADY:
9683 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9684 		break;
9685 	case -NFS4ERR_DELAY:
9686 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9687 		fallthrough;
9688 	case -NFS4ERR_RETRY_UNCACHED_REP:
9689 	case -EACCES:
9690 		dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9691 			__func__, task->tk_status, clp->cl_hostname);
9692 		return -EAGAIN;
9693 	case -NFS4ERR_BADSESSION:
9694 	case -NFS4ERR_DEADSESSION:
9695 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9696 		break;
9697 	default:
9698 		nfs4_schedule_lease_recovery(clp);
9699 	}
9700 	return 0;
9701 }
9702 
9703 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9704 {
9705 	struct nfs4_reclaim_complete_data *calldata = data;
9706 	struct nfs_client *clp = calldata->clp;
9707 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9708 
9709 	if (!nfs41_sequence_done(task, res))
9710 		return;
9711 
9712 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9713 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9714 		rpc_restart_call_prepare(task);
9715 		return;
9716 	}
9717 }
9718 
9719 static void nfs4_free_reclaim_complete_data(void *data)
9720 {
9721 	struct nfs4_reclaim_complete_data *calldata = data;
9722 
9723 	kfree(calldata);
9724 }
9725 
9726 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9727 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9728 	.rpc_call_done = nfs4_reclaim_complete_done,
9729 	.rpc_release = nfs4_free_reclaim_complete_data,
9730 };
9731 
9732 /*
9733  * Issue a global reclaim complete.
9734  */
9735 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9736 		const struct cred *cred)
9737 {
9738 	struct nfs4_reclaim_complete_data *calldata;
9739 	struct rpc_message msg = {
9740 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9741 		.rpc_cred = cred,
9742 	};
9743 	struct rpc_task_setup task_setup_data = {
9744 		.rpc_client = clp->cl_rpcclient,
9745 		.rpc_message = &msg,
9746 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9747 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9748 	};
9749 	int status = -ENOMEM;
9750 
9751 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9752 	if (calldata == NULL)
9753 		goto out;
9754 	calldata->clp = clp;
9755 	calldata->arg.one_fs = 0;
9756 
9757 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9758 	msg.rpc_argp = &calldata->arg;
9759 	msg.rpc_resp = &calldata->res;
9760 	task_setup_data.callback_data = calldata;
9761 	status = nfs4_call_sync_custom(&task_setup_data);
9762 out:
9763 	dprintk("<-- %s status=%d\n", __func__, status);
9764 	return status;
9765 }
9766 
9767 static void
9768 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9769 {
9770 	struct nfs4_layoutget *lgp = calldata;
9771 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9772 
9773 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9774 				&lgp->res.seq_res, task);
9775 }
9776 
9777 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9778 {
9779 	struct nfs4_layoutget *lgp = calldata;
9780 
9781 	nfs41_sequence_process(task, &lgp->res.seq_res);
9782 }
9783 
9784 static int
9785 nfs4_layoutget_handle_exception(struct rpc_task *task,
9786 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9787 {
9788 	struct inode *inode = lgp->args.inode;
9789 	struct nfs_server *server = NFS_SERVER(inode);
9790 	struct pnfs_layout_hdr *lo = lgp->lo;
9791 	int nfs4err = task->tk_status;
9792 	int err, status = 0;
9793 	LIST_HEAD(head);
9794 
9795 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9796 
9797 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9798 
9799 	exception->state = NULL;
9800 	exception->stateid = NULL;
9801 
9802 	switch (nfs4err) {
9803 	case 0:
9804 		goto out;
9805 
9806 	/*
9807 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9808 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9809 	 * retry go inband.
9810 	 */
9811 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9812 		status = -ENODATA;
9813 		goto out;
9814 	/*
9815 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9816 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9817 	 */
9818 	case -NFS4ERR_BADLAYOUT:
9819 		status = -EOVERFLOW;
9820 		goto out;
9821 	/*
9822 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9823 	 * (or clients) writing to the same RAID stripe except when
9824 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9825 	 *
9826 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9827 	 * while, and then eventually give up.
9828 	 */
9829 	case -NFS4ERR_LAYOUTTRYLATER:
9830 		if (lgp->args.minlength == 0) {
9831 			status = -EOVERFLOW;
9832 			goto out;
9833 		}
9834 		status = -EBUSY;
9835 		break;
9836 	case -NFS4ERR_RECALLCONFLICT:
9837 	case -NFS4ERR_RETURNCONFLICT:
9838 		status = -ERECALLCONFLICT;
9839 		break;
9840 	case -NFS4ERR_DELEG_REVOKED:
9841 	case -NFS4ERR_ADMIN_REVOKED:
9842 	case -NFS4ERR_EXPIRED:
9843 	case -NFS4ERR_BAD_STATEID:
9844 		exception->timeout = 0;
9845 		spin_lock(&inode->i_lock);
9846 		/* If the open stateid was bad, then recover it. */
9847 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9848 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9849 			spin_unlock(&inode->i_lock);
9850 			exception->state = lgp->args.ctx->state;
9851 			exception->stateid = &lgp->args.stateid;
9852 			break;
9853 		}
9854 
9855 		/*
9856 		 * Mark the bad layout state as invalid, then retry
9857 		 */
9858 		pnfs_mark_layout_stateid_invalid(lo, &head);
9859 		spin_unlock(&inode->i_lock);
9860 		nfs_commit_inode(inode, 0);
9861 		pnfs_free_lseg_list(&head);
9862 		status = -EAGAIN;
9863 		goto out;
9864 	}
9865 
9866 	err = nfs4_handle_exception(server, nfs4err, exception);
9867 	if (!status) {
9868 		if (exception->retry)
9869 			status = -EAGAIN;
9870 		else
9871 			status = err;
9872 	}
9873 out:
9874 	return status;
9875 }
9876 
9877 size_t max_response_pages(struct nfs_server *server)
9878 {
9879 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9880 	return nfs_page_array_len(0, max_resp_sz);
9881 }
9882 
9883 static void nfs4_layoutget_release(void *calldata)
9884 {
9885 	struct nfs4_layoutget *lgp = calldata;
9886 
9887 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9888 	pnfs_layoutget_free(lgp);
9889 }
9890 
9891 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9892 	.rpc_call_prepare = nfs4_layoutget_prepare,
9893 	.rpc_call_done = nfs4_layoutget_done,
9894 	.rpc_release = nfs4_layoutget_release,
9895 };
9896 
9897 struct pnfs_layout_segment *
9898 nfs4_proc_layoutget(struct nfs4_layoutget *lgp,
9899 		    struct nfs4_exception *exception)
9900 {
9901 	struct inode *inode = lgp->args.inode;
9902 	struct nfs_server *server = NFS_SERVER(inode);
9903 	struct rpc_task *task;
9904 	struct rpc_message msg = {
9905 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9906 		.rpc_argp = &lgp->args,
9907 		.rpc_resp = &lgp->res,
9908 		.rpc_cred = lgp->cred,
9909 	};
9910 	struct rpc_task_setup task_setup_data = {
9911 		.rpc_client = server->client,
9912 		.rpc_message = &msg,
9913 		.callback_ops = &nfs4_layoutget_call_ops,
9914 		.callback_data = lgp,
9915 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9916 			 RPC_TASK_MOVEABLE,
9917 	};
9918 	struct pnfs_layout_segment *lseg = NULL;
9919 	int status = 0;
9920 
9921 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9922 	exception->retry = 0;
9923 
9924 	task = rpc_run_task(&task_setup_data);
9925 	if (IS_ERR(task))
9926 		return ERR_CAST(task);
9927 
9928 	status = rpc_wait_for_completion_task(task);
9929 	if (status != 0)
9930 		goto out;
9931 
9932 	if (task->tk_status < 0) {
9933 		exception->retry = 1;
9934 		status = nfs4_layoutget_handle_exception(task, lgp, exception);
9935 	} else if (lgp->res.layoutp->len == 0) {
9936 		exception->retry = 1;
9937 		status = -EAGAIN;
9938 		nfs4_update_delay(&exception->timeout);
9939 	} else
9940 		lseg = pnfs_layout_process(lgp);
9941 out:
9942 	trace_nfs4_layoutget(lgp->args.ctx,
9943 			&lgp->args.range,
9944 			&lgp->res.range,
9945 			&lgp->res.stateid,
9946 			status);
9947 
9948 	rpc_put_task(task);
9949 	dprintk("<-- %s status=%d\n", __func__, status);
9950 	if (status)
9951 		return ERR_PTR(status);
9952 	return lseg;
9953 }
9954 
9955 static void
9956 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9957 {
9958 	struct nfs4_layoutreturn *lrp = calldata;
9959 
9960 	nfs4_setup_sequence(lrp->clp,
9961 			&lrp->args.seq_args,
9962 			&lrp->res.seq_res,
9963 			task);
9964 	if (!pnfs_layout_is_valid(lrp->args.layout))
9965 		rpc_exit(task, 0);
9966 }
9967 
9968 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9969 {
9970 	struct nfs4_layoutreturn *lrp = calldata;
9971 	struct nfs_server *server;
9972 
9973 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9974 		return;
9975 
9976 	if (task->tk_rpc_status == -ETIMEDOUT) {
9977 		lrp->rpc_status = -EAGAIN;
9978 		lrp->res.lrs_present = 0;
9979 		return;
9980 	}
9981 	/*
9982 	 * Was there an RPC level error? Assume the call succeeded,
9983 	 * and that we need to release the layout
9984 	 */
9985 	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9986 		lrp->res.lrs_present = 0;
9987 		return;
9988 	}
9989 
9990 	server = NFS_SERVER(lrp->args.inode);
9991 	switch (task->tk_status) {
9992 	case -NFS4ERR_OLD_STATEID:
9993 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9994 					&lrp->args.range,
9995 					lrp->args.inode))
9996 			goto out_restart;
9997 		fallthrough;
9998 	default:
9999 		task->tk_status = 0;
10000 		fallthrough;
10001 	case 0:
10002 		break;
10003 	case -NFS4ERR_BADSESSION:
10004 	case -NFS4ERR_DEADSESSION:
10005 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10006 		nfs4_schedule_session_recovery(server->nfs_client->cl_session,
10007 					       task->tk_status);
10008 		lrp->res.lrs_present = 0;
10009 		lrp->rpc_status = -EAGAIN;
10010 		task->tk_status = 0;
10011 		break;
10012 	case -NFS4ERR_DELAY:
10013 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
10014 			break;
10015 		goto out_restart;
10016 	}
10017 	return;
10018 out_restart:
10019 	task->tk_status = 0;
10020 	nfs4_sequence_free_slot(&lrp->res.seq_res);
10021 	rpc_restart_call_prepare(task);
10022 }
10023 
10024 static void nfs4_layoutreturn_release(void *calldata)
10025 {
10026 	struct nfs4_layoutreturn *lrp = calldata;
10027 	struct pnfs_layout_hdr *lo = lrp->args.layout;
10028 
10029 	if (lrp->rpc_status == 0 || !lrp->inode)
10030 		pnfs_layoutreturn_free_lsegs(
10031 			lo, &lrp->args.stateid, &lrp->args.range,
10032 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
10033 	else
10034 		pnfs_layoutreturn_retry_later(lo, &lrp->args.stateid,
10035 					      &lrp->args.range);
10036 	nfs4_sequence_free_slot(&lrp->res.seq_res);
10037 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
10038 		lrp->ld_private.ops->free(&lrp->ld_private);
10039 	pnfs_put_layout_hdr(lrp->args.layout);
10040 	nfs_iput_and_deactive(lrp->inode);
10041 	put_cred(lrp->cred);
10042 	kfree(calldata);
10043 }
10044 
10045 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
10046 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
10047 	.rpc_call_done = nfs4_layoutreturn_done,
10048 	.rpc_release = nfs4_layoutreturn_release,
10049 };
10050 
10051 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, unsigned int flags)
10052 {
10053 	struct rpc_task *task;
10054 	struct rpc_message msg = {
10055 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
10056 		.rpc_argp = &lrp->args,
10057 		.rpc_resp = &lrp->res,
10058 		.rpc_cred = lrp->cred,
10059 	};
10060 	struct rpc_task_setup task_setup_data = {
10061 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
10062 		.rpc_message = &msg,
10063 		.callback_ops = &nfs4_layoutreturn_call_ops,
10064 		.callback_data = lrp,
10065 		.flags = RPC_TASK_MOVEABLE,
10066 	};
10067 	int status = 0;
10068 
10069 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
10070 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
10071 			&task_setup_data.rpc_client, &msg);
10072 
10073 	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
10074 	if (flags & PNFS_FL_LAYOUTRETURN_ASYNC) {
10075 		if (!lrp->inode) {
10076 			nfs4_layoutreturn_release(lrp);
10077 			return -EAGAIN;
10078 		}
10079 		task_setup_data.flags |= RPC_TASK_ASYNC;
10080 	}
10081 	if (!lrp->inode)
10082 		flags |= PNFS_FL_LAYOUTRETURN_PRIVILEGED;
10083 	if (flags & PNFS_FL_LAYOUTRETURN_PRIVILEGED)
10084 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
10085 				   1);
10086 	else
10087 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
10088 				   0);
10089 	task = rpc_run_task(&task_setup_data);
10090 	if (IS_ERR(task))
10091 		return PTR_ERR(task);
10092 	if (!(flags & PNFS_FL_LAYOUTRETURN_ASYNC))
10093 		status = task->tk_status;
10094 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
10095 	dprintk("<-- %s status=%d\n", __func__, status);
10096 	rpc_put_task(task);
10097 	return status;
10098 }
10099 
10100 static int
10101 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
10102 		struct pnfs_device *pdev,
10103 		const struct cred *cred)
10104 {
10105 	struct nfs4_getdeviceinfo_args args = {
10106 		.pdev = pdev,
10107 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
10108 			NOTIFY_DEVICEID4_DELETE,
10109 	};
10110 	struct nfs4_getdeviceinfo_res res = {
10111 		.pdev = pdev,
10112 	};
10113 	struct rpc_message msg = {
10114 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
10115 		.rpc_argp = &args,
10116 		.rpc_resp = &res,
10117 		.rpc_cred = cred,
10118 	};
10119 	int status;
10120 
10121 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
10122 	if (res.notification & ~args.notify_types)
10123 		dprintk("%s: unsupported notification\n", __func__);
10124 	if (res.notification != args.notify_types)
10125 		pdev->nocache = 1;
10126 
10127 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
10128 
10129 	dprintk("<-- %s status=%d\n", __func__, status);
10130 
10131 	return status;
10132 }
10133 
10134 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
10135 		struct pnfs_device *pdev,
10136 		const struct cred *cred)
10137 {
10138 	struct nfs4_exception exception = { };
10139 	int err;
10140 
10141 	do {
10142 		err = nfs4_handle_exception(server,
10143 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
10144 					&exception);
10145 	} while (exception.retry);
10146 	return err;
10147 }
10148 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
10149 
10150 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
10151 {
10152 	struct nfs4_layoutcommit_data *data = calldata;
10153 	struct nfs_server *server = NFS_SERVER(data->args.inode);
10154 
10155 	nfs4_setup_sequence(server->nfs_client,
10156 			&data->args.seq_args,
10157 			&data->res.seq_res,
10158 			task);
10159 }
10160 
10161 static void
10162 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
10163 {
10164 	struct nfs4_layoutcommit_data *data = calldata;
10165 	struct nfs_server *server = NFS_SERVER(data->args.inode);
10166 
10167 	if (!nfs41_sequence_done(task, &data->res.seq_res))
10168 		return;
10169 
10170 	switch (task->tk_status) { /* Just ignore these failures */
10171 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
10172 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
10173 	case -NFS4ERR_BADLAYOUT:     /* no layout */
10174 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
10175 		task->tk_status = 0;
10176 		break;
10177 	case 0:
10178 		break;
10179 	default:
10180 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
10181 			rpc_restart_call_prepare(task);
10182 			return;
10183 		}
10184 	}
10185 }
10186 
10187 static void nfs4_layoutcommit_release(void *calldata)
10188 {
10189 	struct nfs4_layoutcommit_data *data = calldata;
10190 
10191 	pnfs_cleanup_layoutcommit(data);
10192 	nfs_post_op_update_inode_force_wcc(data->args.inode,
10193 					   data->res.fattr);
10194 	put_cred(data->cred);
10195 	nfs_iput_and_deactive(data->inode);
10196 	kfree(data);
10197 }
10198 
10199 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
10200 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
10201 	.rpc_call_done = nfs4_layoutcommit_done,
10202 	.rpc_release = nfs4_layoutcommit_release,
10203 };
10204 
10205 int
10206 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10207 {
10208 	struct rpc_message msg = {
10209 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10210 		.rpc_argp = &data->args,
10211 		.rpc_resp = &data->res,
10212 		.rpc_cred = data->cred,
10213 	};
10214 	struct rpc_task_setup task_setup_data = {
10215 		.task = &data->task,
10216 		.rpc_client = NFS_CLIENT(data->args.inode),
10217 		.rpc_message = &msg,
10218 		.callback_ops = &nfs4_layoutcommit_ops,
10219 		.callback_data = data,
10220 		.flags = RPC_TASK_MOVEABLE,
10221 	};
10222 	struct rpc_task *task;
10223 	int status = 0;
10224 
10225 	dprintk("NFS: initiating layoutcommit call. sync %d "
10226 		"lbw: %llu inode %lu\n", sync,
10227 		data->args.lastbytewritten,
10228 		data->args.inode->i_ino);
10229 
10230 	if (!sync) {
10231 		data->inode = nfs_igrab_and_active(data->args.inode);
10232 		if (data->inode == NULL) {
10233 			nfs4_layoutcommit_release(data);
10234 			return -EAGAIN;
10235 		}
10236 		task_setup_data.flags = RPC_TASK_ASYNC;
10237 	}
10238 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10239 	task = rpc_run_task(&task_setup_data);
10240 	if (IS_ERR(task))
10241 		return PTR_ERR(task);
10242 	if (sync)
10243 		status = task->tk_status;
10244 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10245 	dprintk("%s: status %d\n", __func__, status);
10246 	rpc_put_task(task);
10247 	return status;
10248 }
10249 
10250 /*
10251  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10252  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10253  */
10254 static int
10255 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10256 		    struct nfs_fsinfo *info,
10257 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10258 {
10259 	struct nfs41_secinfo_no_name_args args = {
10260 		.style = SECINFO_STYLE_CURRENT_FH,
10261 	};
10262 	struct nfs4_secinfo_res res = {
10263 		.flavors = flavors,
10264 	};
10265 	struct rpc_message msg = {
10266 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10267 		.rpc_argp = &args,
10268 		.rpc_resp = &res,
10269 	};
10270 	struct nfs4_call_sync_data data = {
10271 		.seq_server = server,
10272 		.seq_args = &args.seq_args,
10273 		.seq_res = &res.seq_res,
10274 	};
10275 	struct rpc_task_setup task_setup = {
10276 		.rpc_client = server->client,
10277 		.rpc_message = &msg,
10278 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10279 		.callback_data = &data,
10280 		.flags = RPC_TASK_NO_ROUND_ROBIN,
10281 	};
10282 	const struct cred *cred = NULL;
10283 	int status;
10284 
10285 	if (use_integrity) {
10286 		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10287 
10288 		cred = nfs4_get_clid_cred(server->nfs_client);
10289 		msg.rpc_cred = cred;
10290 	}
10291 
10292 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10293 	status = nfs4_call_sync_custom(&task_setup);
10294 	dprintk("<-- %s status=%d\n", __func__, status);
10295 
10296 	put_cred(cred);
10297 
10298 	return status;
10299 }
10300 
10301 static int
10302 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10303 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10304 {
10305 	struct nfs4_exception exception = {
10306 		.interruptible = true,
10307 	};
10308 	int err;
10309 	do {
10310 		/* first try using integrity protection */
10311 		err = -NFS4ERR_WRONGSEC;
10312 
10313 		/* try to use integrity protection with machine cred */
10314 		if (_nfs4_is_integrity_protected(server->nfs_client))
10315 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10316 							  flavors, true);
10317 
10318 		/*
10319 		 * if unable to use integrity protection, or SECINFO with
10320 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10321 		 * disallowed by spec, but exists in deployed servers) use
10322 		 * the current filesystem's rpc_client and the user cred.
10323 		 */
10324 		if (err == -NFS4ERR_WRONGSEC)
10325 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10326 							  flavors, false);
10327 
10328 		switch (err) {
10329 		case 0:
10330 		case -NFS4ERR_WRONGSEC:
10331 		case -ENOTSUPP:
10332 			goto out;
10333 		default:
10334 			err = nfs4_handle_exception(server, err, &exception);
10335 		}
10336 	} while (exception.retry);
10337 out:
10338 	return err;
10339 }
10340 
10341 static int
10342 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10343 		    struct nfs_fsinfo *info)
10344 {
10345 	int err;
10346 	struct page *page;
10347 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10348 	struct nfs4_secinfo_flavors *flavors;
10349 	struct nfs4_secinfo4 *secinfo;
10350 	int i;
10351 
10352 	page = alloc_page(GFP_KERNEL);
10353 	if (!page) {
10354 		err = -ENOMEM;
10355 		goto out;
10356 	}
10357 
10358 	flavors = page_address(page);
10359 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10360 
10361 	/*
10362 	 * Fall back on "guess and check" method if
10363 	 * the server doesn't support SECINFO_NO_NAME
10364 	 */
10365 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10366 		err = nfs4_find_root_sec(server, fhandle, info);
10367 		goto out_freepage;
10368 	}
10369 	if (err)
10370 		goto out_freepage;
10371 
10372 	for (i = 0; i < flavors->num_flavors; i++) {
10373 		secinfo = &flavors->flavors[i];
10374 
10375 		switch (secinfo->flavor) {
10376 		case RPC_AUTH_NULL:
10377 		case RPC_AUTH_UNIX:
10378 		case RPC_AUTH_GSS:
10379 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10380 					&secinfo->flavor_info);
10381 			break;
10382 		default:
10383 			flavor = RPC_AUTH_MAXFLAVOR;
10384 			break;
10385 		}
10386 
10387 		if (!nfs_auth_info_match(&server->auth_info, flavor))
10388 			flavor = RPC_AUTH_MAXFLAVOR;
10389 
10390 		if (flavor != RPC_AUTH_MAXFLAVOR) {
10391 			err = nfs4_lookup_root_sec(server, fhandle,
10392 						   info, flavor);
10393 			if (!err)
10394 				break;
10395 		}
10396 	}
10397 
10398 	if (flavor == RPC_AUTH_MAXFLAVOR)
10399 		err = -EPERM;
10400 
10401 out_freepage:
10402 	put_page(page);
10403 	if (err == -EACCES)
10404 		return -EPERM;
10405 out:
10406 	return err;
10407 }
10408 
10409 static int _nfs41_test_stateid(struct nfs_server *server,
10410 			       const nfs4_stateid *stateid,
10411 			       const struct cred *cred)
10412 {
10413 	int status;
10414 	struct nfs41_test_stateid_args args = {
10415 		.stateid = *stateid,
10416 	};
10417 	struct nfs41_test_stateid_res res;
10418 	struct rpc_message msg = {
10419 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10420 		.rpc_argp = &args,
10421 		.rpc_resp = &res,
10422 		.rpc_cred = cred,
10423 	};
10424 	struct rpc_clnt *rpc_client = server->client;
10425 
10426 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10427 		&rpc_client, &msg);
10428 
10429 	dprintk("NFS call  test_stateid %p\n", stateid);
10430 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10431 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10432 			&args.seq_args, &res.seq_res);
10433 	if (status != NFS_OK) {
10434 		dprintk("NFS reply test_stateid: failed, %d\n", status);
10435 		return status;
10436 	}
10437 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10438 	return -res.status;
10439 }
10440 
10441 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10442 		int err, struct nfs4_exception *exception)
10443 {
10444 	exception->retry = 0;
10445 	switch(err) {
10446 	case -NFS4ERR_DELAY:
10447 	case -NFS4ERR_RETRY_UNCACHED_REP:
10448 		nfs4_handle_exception(server, err, exception);
10449 		break;
10450 	case -NFS4ERR_BADSESSION:
10451 	case -NFS4ERR_BADSLOT:
10452 	case -NFS4ERR_BAD_HIGH_SLOT:
10453 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10454 	case -NFS4ERR_DEADSESSION:
10455 		nfs4_do_handle_exception(server, err, exception);
10456 	}
10457 }
10458 
10459 /**
10460  * nfs41_test_stateid - perform a TEST_STATEID operation
10461  *
10462  * @server: server / transport on which to perform the operation
10463  * @stateid: state ID to test
10464  * @cred: credential
10465  *
10466  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10467  * Otherwise a negative NFS4ERR value is returned if the operation
10468  * failed or the state ID is not currently valid.
10469  */
10470 static int nfs41_test_stateid(struct nfs_server *server,
10471 			      const nfs4_stateid *stateid,
10472 			      const struct cred *cred)
10473 {
10474 	struct nfs4_exception exception = {
10475 		.interruptible = true,
10476 	};
10477 	int err;
10478 	do {
10479 		err = _nfs41_test_stateid(server, stateid, cred);
10480 		nfs4_handle_delay_or_session_error(server, err, &exception);
10481 	} while (exception.retry);
10482 	return err;
10483 }
10484 
10485 struct nfs_free_stateid_data {
10486 	struct nfs_server *server;
10487 	struct nfs41_free_stateid_args args;
10488 	struct nfs41_free_stateid_res res;
10489 };
10490 
10491 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10492 {
10493 	struct nfs_free_stateid_data *data = calldata;
10494 	nfs4_setup_sequence(data->server->nfs_client,
10495 			&data->args.seq_args,
10496 			&data->res.seq_res,
10497 			task);
10498 }
10499 
10500 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10501 {
10502 	struct nfs_free_stateid_data *data = calldata;
10503 
10504 	nfs41_sequence_done(task, &data->res.seq_res);
10505 
10506 	switch (task->tk_status) {
10507 	case -NFS4ERR_DELAY:
10508 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10509 			rpc_restart_call_prepare(task);
10510 	}
10511 }
10512 
10513 static void nfs41_free_stateid_release(void *calldata)
10514 {
10515 	struct nfs_free_stateid_data *data = calldata;
10516 	struct nfs_client *clp = data->server->nfs_client;
10517 
10518 	nfs_put_client(clp);
10519 	kfree(calldata);
10520 }
10521 
10522 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10523 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10524 	.rpc_call_done = nfs41_free_stateid_done,
10525 	.rpc_release = nfs41_free_stateid_release,
10526 };
10527 
10528 /**
10529  * nfs41_free_stateid - perform a FREE_STATEID operation
10530  *
10531  * @server: server / transport on which to perform the operation
10532  * @stateid: state ID to release
10533  * @cred: credential
10534  * @privileged: set to true if this call needs to be privileged
10535  *
10536  * Note: this function is always asynchronous.
10537  */
10538 static int nfs41_free_stateid(struct nfs_server *server,
10539 		const nfs4_stateid *stateid,
10540 		const struct cred *cred,
10541 		bool privileged)
10542 {
10543 	struct rpc_message msg = {
10544 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10545 		.rpc_cred = cred,
10546 	};
10547 	struct rpc_task_setup task_setup = {
10548 		.rpc_client = server->client,
10549 		.rpc_message = &msg,
10550 		.callback_ops = &nfs41_free_stateid_ops,
10551 		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10552 	};
10553 	struct nfs_free_stateid_data *data;
10554 	struct rpc_task *task;
10555 	struct nfs_client *clp = server->nfs_client;
10556 
10557 	if (!refcount_inc_not_zero(&clp->cl_count))
10558 		return -EIO;
10559 
10560 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10561 		&task_setup.rpc_client, &msg);
10562 
10563 	dprintk("NFS call  free_stateid %p\n", stateid);
10564 	data = kmalloc(sizeof(*data), GFP_KERNEL);
10565 	if (!data)
10566 		return -ENOMEM;
10567 	data->server = server;
10568 	nfs4_stateid_copy(&data->args.stateid, stateid);
10569 
10570 	task_setup.callback_data = data;
10571 
10572 	msg.rpc_argp = &data->args;
10573 	msg.rpc_resp = &data->res;
10574 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10575 	task = rpc_run_task(&task_setup);
10576 	if (IS_ERR(task))
10577 		return PTR_ERR(task);
10578 	rpc_put_task(task);
10579 	return 0;
10580 }
10581 
10582 static void
10583 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10584 {
10585 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10586 
10587 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10588 	nfs4_free_lock_state(server, lsp);
10589 }
10590 
10591 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10592 		const nfs4_stateid *s2)
10593 {
10594 	if (s1->type != s2->type)
10595 		return false;
10596 
10597 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10598 		return false;
10599 
10600 	if (s1->seqid == s2->seqid)
10601 		return true;
10602 
10603 	return s1->seqid == 0 || s2->seqid == 0;
10604 }
10605 
10606 #endif /* CONFIG_NFS_V4_1 */
10607 
10608 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10609 		const nfs4_stateid *s2)
10610 {
10611 	return nfs4_stateid_match(s1, s2);
10612 }
10613 
10614 
10615 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10616 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10617 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10618 	.recover_open	= nfs4_open_reclaim,
10619 	.recover_lock	= nfs4_lock_reclaim,
10620 	.establish_clid = nfs4_init_clientid,
10621 	.detect_trunking = nfs40_discover_server_trunking,
10622 };
10623 
10624 #if defined(CONFIG_NFS_V4_1)
10625 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10626 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10627 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10628 	.recover_open	= nfs4_open_reclaim,
10629 	.recover_lock	= nfs4_lock_reclaim,
10630 	.establish_clid = nfs41_init_clientid,
10631 	.reclaim_complete = nfs41_proc_reclaim_complete,
10632 	.detect_trunking = nfs41_discover_server_trunking,
10633 };
10634 #endif /* CONFIG_NFS_V4_1 */
10635 
10636 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10637 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10638 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10639 	.recover_open	= nfs40_open_expired,
10640 	.recover_lock	= nfs4_lock_expired,
10641 	.establish_clid = nfs4_init_clientid,
10642 };
10643 
10644 #if defined(CONFIG_NFS_V4_1)
10645 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10646 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10647 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10648 	.recover_open	= nfs41_open_expired,
10649 	.recover_lock	= nfs41_lock_expired,
10650 	.establish_clid = nfs41_init_clientid,
10651 };
10652 #endif /* CONFIG_NFS_V4_1 */
10653 
10654 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10655 	.sched_state_renewal = nfs4_proc_async_renew,
10656 	.get_state_renewal_cred = nfs4_get_renew_cred,
10657 	.renew_lease = nfs4_proc_renew,
10658 };
10659 
10660 #if defined(CONFIG_NFS_V4_1)
10661 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10662 	.sched_state_renewal = nfs41_proc_async_sequence,
10663 	.get_state_renewal_cred = nfs4_get_machine_cred,
10664 	.renew_lease = nfs4_proc_sequence,
10665 };
10666 #endif
10667 
10668 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10669 	.get_locations = _nfs40_proc_get_locations,
10670 	.fsid_present = _nfs40_proc_fsid_present,
10671 };
10672 
10673 #if defined(CONFIG_NFS_V4_1)
10674 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10675 	.get_locations = _nfs41_proc_get_locations,
10676 	.fsid_present = _nfs41_proc_fsid_present,
10677 };
10678 #endif	/* CONFIG_NFS_V4_1 */
10679 
10680 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10681 	.minor_version = 0,
10682 	.init_caps = NFS_CAP_READDIRPLUS
10683 		| NFS_CAP_ATOMIC_OPEN
10684 		| NFS_CAP_POSIX_LOCK,
10685 	.init_client = nfs40_init_client,
10686 	.shutdown_client = nfs40_shutdown_client,
10687 	.match_stateid = nfs4_match_stateid,
10688 	.find_root_sec = nfs4_find_root_sec,
10689 	.free_lock_state = nfs4_release_lockowner,
10690 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10691 	.alloc_seqid = nfs_alloc_seqid,
10692 	.call_sync_ops = &nfs40_call_sync_ops,
10693 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10694 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10695 	.state_renewal_ops = &nfs40_state_renewal_ops,
10696 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10697 };
10698 
10699 #if defined(CONFIG_NFS_V4_1)
10700 static struct nfs_seqid *
10701 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10702 {
10703 	return NULL;
10704 }
10705 
10706 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10707 	.minor_version = 1,
10708 	.init_caps = NFS_CAP_READDIRPLUS
10709 		| NFS_CAP_ATOMIC_OPEN
10710 		| NFS_CAP_POSIX_LOCK
10711 		| NFS_CAP_STATEID_NFSV41
10712 		| NFS_CAP_ATOMIC_OPEN_V1
10713 		| NFS_CAP_LGOPEN
10714 		| NFS_CAP_MOVEABLE,
10715 	.init_client = nfs41_init_client,
10716 	.shutdown_client = nfs41_shutdown_client,
10717 	.match_stateid = nfs41_match_stateid,
10718 	.find_root_sec = nfs41_find_root_sec,
10719 	.free_lock_state = nfs41_free_lock_state,
10720 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10721 	.alloc_seqid = nfs_alloc_no_seqid,
10722 	.session_trunk = nfs4_test_session_trunk,
10723 	.call_sync_ops = &nfs41_call_sync_ops,
10724 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10725 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10726 	.state_renewal_ops = &nfs41_state_renewal_ops,
10727 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10728 };
10729 #endif
10730 
10731 #if defined(CONFIG_NFS_V4_2)
10732 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10733 	.minor_version = 2,
10734 	.init_caps = NFS_CAP_READDIRPLUS
10735 		| NFS_CAP_ATOMIC_OPEN
10736 		| NFS_CAP_POSIX_LOCK
10737 		| NFS_CAP_STATEID_NFSV41
10738 		| NFS_CAP_ATOMIC_OPEN_V1
10739 		| NFS_CAP_LGOPEN
10740 		| NFS_CAP_ALLOCATE
10741 		| NFS_CAP_COPY
10742 		| NFS_CAP_OFFLOAD_CANCEL
10743 		| NFS_CAP_COPY_NOTIFY
10744 		| NFS_CAP_DEALLOCATE
10745 		| NFS_CAP_SEEK
10746 		| NFS_CAP_LAYOUTSTATS
10747 		| NFS_CAP_CLONE
10748 		| NFS_CAP_LAYOUTERROR
10749 		| NFS_CAP_READ_PLUS
10750 		| NFS_CAP_MOVEABLE,
10751 	.init_client = nfs41_init_client,
10752 	.shutdown_client = nfs41_shutdown_client,
10753 	.match_stateid = nfs41_match_stateid,
10754 	.find_root_sec = nfs41_find_root_sec,
10755 	.free_lock_state = nfs41_free_lock_state,
10756 	.call_sync_ops = &nfs41_call_sync_ops,
10757 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10758 	.alloc_seqid = nfs_alloc_no_seqid,
10759 	.session_trunk = nfs4_test_session_trunk,
10760 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10761 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10762 	.state_renewal_ops = &nfs41_state_renewal_ops,
10763 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10764 };
10765 #endif
10766 
10767 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10768 	[0] = &nfs_v4_0_minor_ops,
10769 #if defined(CONFIG_NFS_V4_1)
10770 	[1] = &nfs_v4_1_minor_ops,
10771 #endif
10772 #if defined(CONFIG_NFS_V4_2)
10773 	[2] = &nfs_v4_2_minor_ops,
10774 #endif
10775 };
10776 
10777 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10778 {
10779 	ssize_t error, error2, error3;
10780 	size_t left = size;
10781 
10782 	error = generic_listxattr(dentry, list, left);
10783 	if (error < 0)
10784 		return error;
10785 	if (list) {
10786 		list += error;
10787 		left -= error;
10788 	}
10789 
10790 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, left);
10791 	if (error2 < 0)
10792 		return error2;
10793 
10794 	if (list) {
10795 		list += error2;
10796 		left -= error2;
10797 	}
10798 
10799 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, left);
10800 	if (error3 < 0)
10801 		return error3;
10802 
10803 	error += error2 + error3;
10804 	if (size && error > size)
10805 		return -ERANGE;
10806 	return error;
10807 }
10808 
10809 static void nfs4_enable_swap(struct inode *inode)
10810 {
10811 	/* The state manager thread must always be running.
10812 	 * It will notice the client is a swapper, and stay put.
10813 	 */
10814 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10815 
10816 	nfs4_schedule_state_manager(clp);
10817 }
10818 
10819 static void nfs4_disable_swap(struct inode *inode)
10820 {
10821 	/* The state manager thread will now exit once it is
10822 	 * woken.
10823 	 */
10824 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10825 
10826 	set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10827 	clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10828 	wake_up_var(&clp->cl_state);
10829 }
10830 
10831 static const struct inode_operations nfs4_dir_inode_operations = {
10832 	.create		= nfs_create,
10833 	.lookup		= nfs_lookup,
10834 	.atomic_open	= nfs_atomic_open,
10835 	.link		= nfs_link,
10836 	.unlink		= nfs_unlink,
10837 	.symlink	= nfs_symlink,
10838 	.mkdir		= nfs_mkdir,
10839 	.rmdir		= nfs_rmdir,
10840 	.mknod		= nfs_mknod,
10841 	.rename		= nfs_rename,
10842 	.permission	= nfs_permission,
10843 	.getattr	= nfs_getattr,
10844 	.setattr	= nfs_setattr,
10845 	.listxattr	= nfs4_listxattr,
10846 };
10847 
10848 static const struct inode_operations nfs4_file_inode_operations = {
10849 	.permission	= nfs_permission,
10850 	.getattr	= nfs_getattr,
10851 	.setattr	= nfs_setattr,
10852 	.listxattr	= nfs4_listxattr,
10853 };
10854 
10855 const struct nfs_rpc_ops nfs_v4_clientops = {
10856 	.version	= 4,			/* protocol version */
10857 	.dentry_ops	= &nfs4_dentry_operations,
10858 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10859 	.file_inode_ops	= &nfs4_file_inode_operations,
10860 	.file_ops	= &nfs4_file_operations,
10861 	.getroot	= nfs4_proc_get_root,
10862 	.submount	= nfs4_submount,
10863 	.try_get_tree	= nfs4_try_get_tree,
10864 	.getattr	= nfs4_proc_getattr,
10865 	.setattr	= nfs4_proc_setattr,
10866 	.lookup		= nfs4_proc_lookup,
10867 	.lookupp	= nfs4_proc_lookupp,
10868 	.access		= nfs4_proc_access,
10869 	.readlink	= nfs4_proc_readlink,
10870 	.create		= nfs4_proc_create,
10871 	.remove		= nfs4_proc_remove,
10872 	.unlink_setup	= nfs4_proc_unlink_setup,
10873 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10874 	.unlink_done	= nfs4_proc_unlink_done,
10875 	.rename_setup	= nfs4_proc_rename_setup,
10876 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10877 	.rename_done	= nfs4_proc_rename_done,
10878 	.link		= nfs4_proc_link,
10879 	.symlink	= nfs4_proc_symlink,
10880 	.mkdir		= nfs4_proc_mkdir,
10881 	.rmdir		= nfs4_proc_rmdir,
10882 	.readdir	= nfs4_proc_readdir,
10883 	.mknod		= nfs4_proc_mknod,
10884 	.statfs		= nfs4_proc_statfs,
10885 	.fsinfo		= nfs4_proc_fsinfo,
10886 	.pathconf	= nfs4_proc_pathconf,
10887 	.set_capabilities = nfs4_server_capabilities,
10888 	.decode_dirent	= nfs4_decode_dirent,
10889 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10890 	.read_setup	= nfs4_proc_read_setup,
10891 	.read_done	= nfs4_read_done,
10892 	.write_setup	= nfs4_proc_write_setup,
10893 	.write_done	= nfs4_write_done,
10894 	.commit_setup	= nfs4_proc_commit_setup,
10895 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10896 	.commit_done	= nfs4_commit_done,
10897 	.lock		= nfs4_proc_lock,
10898 	.clear_acl_cache = nfs4_zap_acl_attr,
10899 	.close_context  = nfs4_close_context,
10900 	.open_context	= nfs4_atomic_open,
10901 	.have_delegation = nfs4_have_delegation,
10902 	.return_delegation = nfs4_inode_return_delegation,
10903 	.alloc_client	= nfs4_alloc_client,
10904 	.init_client	= nfs4_init_client,
10905 	.free_client	= nfs4_free_client,
10906 	.create_server	= nfs4_create_server,
10907 	.clone_server	= nfs_clone_server,
10908 	.discover_trunking = nfs4_discover_trunking,
10909 	.enable_swap	= nfs4_enable_swap,
10910 	.disable_swap	= nfs4_disable_swap,
10911 };
10912 
10913 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10914 	.name	= XATTR_NAME_NFSV4_ACL,
10915 	.list	= nfs4_xattr_list_nfs4_acl,
10916 	.get	= nfs4_xattr_get_nfs4_acl,
10917 	.set	= nfs4_xattr_set_nfs4_acl,
10918 };
10919 
10920 #if defined(CONFIG_NFS_V4_1)
10921 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10922 	.name	= XATTR_NAME_NFSV4_DACL,
10923 	.list	= nfs4_xattr_list_nfs4_dacl,
10924 	.get	= nfs4_xattr_get_nfs4_dacl,
10925 	.set	= nfs4_xattr_set_nfs4_dacl,
10926 };
10927 
10928 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10929 	.name	= XATTR_NAME_NFSV4_SACL,
10930 	.list	= nfs4_xattr_list_nfs4_sacl,
10931 	.get	= nfs4_xattr_get_nfs4_sacl,
10932 	.set	= nfs4_xattr_set_nfs4_sacl,
10933 };
10934 #endif
10935 
10936 #ifdef CONFIG_NFS_V4_2
10937 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10938 	.prefix	= XATTR_USER_PREFIX,
10939 	.get	= nfs4_xattr_get_nfs4_user,
10940 	.set	= nfs4_xattr_set_nfs4_user,
10941 };
10942 #endif
10943 
10944 const struct xattr_handler * const nfs4_xattr_handlers[] = {
10945 	&nfs4_xattr_nfs4_acl_handler,
10946 #if defined(CONFIG_NFS_V4_1)
10947 	&nfs4_xattr_nfs4_dacl_handler,
10948 	&nfs4_xattr_nfs4_sacl_handler,
10949 #endif
10950 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10951 	&nfs4_xattr_nfs4_label_handler,
10952 #endif
10953 #ifdef CONFIG_NFS_V4_2
10954 	&nfs4_xattr_nfs4_user_handler,
10955 #endif
10956 	NULL
10957 };
10958