xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 379e4adf)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "sysfs.h"
67 #include "nfs4idmap.h"
68 #include "nfs4session.h"
69 #include "fscache.h"
70 #include "nfs42.h"
71 
72 #include "nfs4trace.h"
73 
74 #define NFSDBG_FACILITY		NFSDBG_PROC
75 
76 #define NFS4_BITMASK_SZ		3
77 
78 #define NFS4_POLL_RETRY_MIN	(HZ/10)
79 #define NFS4_POLL_RETRY_MAX	(15*HZ)
80 
81 /* file attributes which can be mapped to nfs attributes */
82 #define NFS4_VALID_ATTRS (ATTR_MODE \
83 	| ATTR_UID \
84 	| ATTR_GID \
85 	| ATTR_SIZE \
86 	| ATTR_ATIME \
87 	| ATTR_MTIME \
88 	| ATTR_CTIME \
89 	| ATTR_ATIME_SET \
90 	| ATTR_MTIME_SET)
91 
92 struct nfs4_opendata;
93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
97 			      struct nfs_fattr *fattr, struct inode *inode);
98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
99 			    struct nfs_fattr *fattr, struct iattr *sattr,
100 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel);
101 #ifdef CONFIG_NFS_V4_1
102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
103 		const struct cred *cred,
104 		struct nfs4_slot *slot,
105 		bool is_privileged);
106 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
107 		const struct cred *);
108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
109 		const struct cred *, bool);
110 #endif
111 
112 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
113 static inline struct nfs4_label *
114 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
115 	struct iattr *sattr, struct nfs4_label *label)
116 {
117 	int err;
118 
119 	if (label == NULL)
120 		return NULL;
121 
122 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
123 		return NULL;
124 
125 	label->lfs = 0;
126 	label->pi = 0;
127 	label->len = 0;
128 	label->label = NULL;
129 
130 	err = security_dentry_init_security(dentry, sattr->ia_mode,
131 				&dentry->d_name, NULL,
132 				(void **)&label->label, &label->len);
133 	if (err == 0)
134 		return label;
135 
136 	return NULL;
137 }
138 static inline void
139 nfs4_label_release_security(struct nfs4_label *label)
140 {
141 	if (label)
142 		security_release_secctx(label->label, label->len);
143 }
144 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
145 {
146 	if (label)
147 		return server->attr_bitmask;
148 
149 	return server->attr_bitmask_nl;
150 }
151 #else
152 static inline struct nfs4_label *
153 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
154 	struct iattr *sattr, struct nfs4_label *l)
155 { return NULL; }
156 static inline void
157 nfs4_label_release_security(struct nfs4_label *label)
158 { return; }
159 static inline u32 *
160 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
161 { return server->attr_bitmask; }
162 #endif
163 
164 /* Prevent leaks of NFSv4 errors into userland */
165 static int nfs4_map_errors(int err)
166 {
167 	if (err >= -1000)
168 		return err;
169 	switch (err) {
170 	case -NFS4ERR_RESOURCE:
171 	case -NFS4ERR_LAYOUTTRYLATER:
172 	case -NFS4ERR_RECALLCONFLICT:
173 		return -EREMOTEIO;
174 	case -NFS4ERR_WRONGSEC:
175 	case -NFS4ERR_WRONG_CRED:
176 		return -EPERM;
177 	case -NFS4ERR_BADOWNER:
178 	case -NFS4ERR_BADNAME:
179 		return -EINVAL;
180 	case -NFS4ERR_SHARE_DENIED:
181 		return -EACCES;
182 	case -NFS4ERR_MINOR_VERS_MISMATCH:
183 		return -EPROTONOSUPPORT;
184 	case -NFS4ERR_FILE_OPEN:
185 		return -EBUSY;
186 	case -NFS4ERR_NOT_SAME:
187 		return -ENOTSYNC;
188 	default:
189 		dprintk("%s could not handle NFSv4 error %d\n",
190 				__func__, -err);
191 		break;
192 	}
193 	return -EIO;
194 }
195 
196 /*
197  * This is our standard bitmap for GETATTR requests.
198  */
199 const u32 nfs4_fattr_bitmap[3] = {
200 	FATTR4_WORD0_TYPE
201 	| FATTR4_WORD0_CHANGE
202 	| FATTR4_WORD0_SIZE
203 	| FATTR4_WORD0_FSID
204 	| FATTR4_WORD0_FILEID,
205 	FATTR4_WORD1_MODE
206 	| FATTR4_WORD1_NUMLINKS
207 	| FATTR4_WORD1_OWNER
208 	| FATTR4_WORD1_OWNER_GROUP
209 	| FATTR4_WORD1_RAWDEV
210 	| FATTR4_WORD1_SPACE_USED
211 	| FATTR4_WORD1_TIME_ACCESS
212 	| FATTR4_WORD1_TIME_METADATA
213 	| FATTR4_WORD1_TIME_MODIFY
214 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
215 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
216 	FATTR4_WORD2_SECURITY_LABEL
217 #endif
218 };
219 
220 static const u32 nfs4_pnfs_open_bitmap[3] = {
221 	FATTR4_WORD0_TYPE
222 	| FATTR4_WORD0_CHANGE
223 	| FATTR4_WORD0_SIZE
224 	| FATTR4_WORD0_FSID
225 	| FATTR4_WORD0_FILEID,
226 	FATTR4_WORD1_MODE
227 	| FATTR4_WORD1_NUMLINKS
228 	| FATTR4_WORD1_OWNER
229 	| FATTR4_WORD1_OWNER_GROUP
230 	| FATTR4_WORD1_RAWDEV
231 	| FATTR4_WORD1_SPACE_USED
232 	| FATTR4_WORD1_TIME_ACCESS
233 	| FATTR4_WORD1_TIME_METADATA
234 	| FATTR4_WORD1_TIME_MODIFY,
235 	FATTR4_WORD2_MDSTHRESHOLD
236 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
237 	| FATTR4_WORD2_SECURITY_LABEL
238 #endif
239 };
240 
241 static const u32 nfs4_open_noattr_bitmap[3] = {
242 	FATTR4_WORD0_TYPE
243 	| FATTR4_WORD0_FILEID,
244 };
245 
246 const u32 nfs4_statfs_bitmap[3] = {
247 	FATTR4_WORD0_FILES_AVAIL
248 	| FATTR4_WORD0_FILES_FREE
249 	| FATTR4_WORD0_FILES_TOTAL,
250 	FATTR4_WORD1_SPACE_AVAIL
251 	| FATTR4_WORD1_SPACE_FREE
252 	| FATTR4_WORD1_SPACE_TOTAL
253 };
254 
255 const u32 nfs4_pathconf_bitmap[3] = {
256 	FATTR4_WORD0_MAXLINK
257 	| FATTR4_WORD0_MAXNAME,
258 	0
259 };
260 
261 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
262 			| FATTR4_WORD0_MAXREAD
263 			| FATTR4_WORD0_MAXWRITE
264 			| FATTR4_WORD0_LEASE_TIME,
265 			FATTR4_WORD1_TIME_DELTA
266 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
267 			FATTR4_WORD2_LAYOUT_BLKSIZE
268 			| FATTR4_WORD2_CLONE_BLKSIZE
269 			| FATTR4_WORD2_CHANGE_ATTR_TYPE
270 			| FATTR4_WORD2_XATTR_SUPPORT
271 };
272 
273 const u32 nfs4_fs_locations_bitmap[3] = {
274 	FATTR4_WORD0_CHANGE
275 	| FATTR4_WORD0_SIZE
276 	| FATTR4_WORD0_FSID
277 	| FATTR4_WORD0_FILEID
278 	| FATTR4_WORD0_FS_LOCATIONS,
279 	FATTR4_WORD1_OWNER
280 	| FATTR4_WORD1_OWNER_GROUP
281 	| FATTR4_WORD1_RAWDEV
282 	| FATTR4_WORD1_SPACE_USED
283 	| FATTR4_WORD1_TIME_ACCESS
284 	| FATTR4_WORD1_TIME_METADATA
285 	| FATTR4_WORD1_TIME_MODIFY
286 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
287 };
288 
289 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
290 				    struct inode *inode, unsigned long flags)
291 {
292 	unsigned long cache_validity;
293 
294 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
295 	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
296 		return;
297 
298 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
299 
300 	/* Remove the attributes over which we have full control */
301 	dst[1] &= ~FATTR4_WORD1_RAWDEV;
302 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
303 		dst[0] &= ~FATTR4_WORD0_SIZE;
304 
305 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
306 		dst[0] &= ~FATTR4_WORD0_CHANGE;
307 
308 	if (!(cache_validity & NFS_INO_INVALID_MODE))
309 		dst[1] &= ~FATTR4_WORD1_MODE;
310 	if (!(cache_validity & NFS_INO_INVALID_OTHER))
311 		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
312 }
313 
314 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
315 		struct nfs4_readdir_arg *readdir)
316 {
317 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
318 	__be32 *start, *p;
319 
320 	if (cookie > 2) {
321 		readdir->cookie = cookie;
322 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
323 		return;
324 	}
325 
326 	readdir->cookie = 0;
327 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
328 	if (cookie == 2)
329 		return;
330 
331 	/*
332 	 * NFSv4 servers do not return entries for '.' and '..'
333 	 * Therefore, we fake these entries here.  We let '.'
334 	 * have cookie 0 and '..' have cookie 1.  Note that
335 	 * when talking to the server, we always send cookie 0
336 	 * instead of 1 or 2.
337 	 */
338 	start = p = kmap_atomic(*readdir->pages);
339 
340 	if (cookie == 0) {
341 		*p++ = xdr_one;                                  /* next */
342 		*p++ = xdr_zero;                   /* cookie, first word */
343 		*p++ = xdr_one;                   /* cookie, second word */
344 		*p++ = xdr_one;                             /* entry len */
345 		memcpy(p, ".\0\0\0", 4);                        /* entry */
346 		p++;
347 		*p++ = xdr_one;                         /* bitmap length */
348 		*p++ = htonl(attrs);                           /* bitmap */
349 		*p++ = htonl(12);             /* attribute buffer length */
350 		*p++ = htonl(NF4DIR);
351 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
352 	}
353 
354 	*p++ = xdr_one;                                  /* next */
355 	*p++ = xdr_zero;                   /* cookie, first word */
356 	*p++ = xdr_two;                   /* cookie, second word */
357 	*p++ = xdr_two;                             /* entry len */
358 	memcpy(p, "..\0\0", 4);                         /* entry */
359 	p++;
360 	*p++ = xdr_one;                         /* bitmap length */
361 	*p++ = htonl(attrs);                           /* bitmap */
362 	*p++ = htonl(12);             /* attribute buffer length */
363 	*p++ = htonl(NF4DIR);
364 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
365 
366 	readdir->pgbase = (char *)p - (char *)start;
367 	readdir->count -= readdir->pgbase;
368 	kunmap_atomic(start);
369 }
370 
371 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
372 {
373 	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
374 		fattr->pre_change_attr = version;
375 		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
376 	}
377 }
378 
379 static void nfs4_test_and_free_stateid(struct nfs_server *server,
380 		nfs4_stateid *stateid,
381 		const struct cred *cred)
382 {
383 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
384 
385 	ops->test_and_free_expired(server, stateid, cred);
386 }
387 
388 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
389 		nfs4_stateid *stateid,
390 		const struct cred *cred)
391 {
392 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
393 	nfs4_test_and_free_stateid(server, stateid, cred);
394 }
395 
396 static void nfs4_free_revoked_stateid(struct nfs_server *server,
397 		const nfs4_stateid *stateid,
398 		const struct cred *cred)
399 {
400 	nfs4_stateid tmp;
401 
402 	nfs4_stateid_copy(&tmp, stateid);
403 	__nfs4_free_revoked_stateid(server, &tmp, cred);
404 }
405 
406 static long nfs4_update_delay(long *timeout)
407 {
408 	long ret;
409 	if (!timeout)
410 		return NFS4_POLL_RETRY_MAX;
411 	if (*timeout <= 0)
412 		*timeout = NFS4_POLL_RETRY_MIN;
413 	if (*timeout > NFS4_POLL_RETRY_MAX)
414 		*timeout = NFS4_POLL_RETRY_MAX;
415 	ret = *timeout;
416 	*timeout <<= 1;
417 	return ret;
418 }
419 
420 static int nfs4_delay_killable(long *timeout)
421 {
422 	might_sleep();
423 
424 	__set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
425 	schedule_timeout(nfs4_update_delay(timeout));
426 	if (!__fatal_signal_pending(current))
427 		return 0;
428 	return -EINTR;
429 }
430 
431 static int nfs4_delay_interruptible(long *timeout)
432 {
433 	might_sleep();
434 
435 	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
436 	schedule_timeout(nfs4_update_delay(timeout));
437 	if (!signal_pending(current))
438 		return 0;
439 	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
440 }
441 
442 static int nfs4_delay(long *timeout, bool interruptible)
443 {
444 	if (interruptible)
445 		return nfs4_delay_interruptible(timeout);
446 	return nfs4_delay_killable(timeout);
447 }
448 
449 static const nfs4_stateid *
450 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
451 {
452 	if (!stateid)
453 		return NULL;
454 	switch (stateid->type) {
455 	case NFS4_OPEN_STATEID_TYPE:
456 	case NFS4_LOCK_STATEID_TYPE:
457 	case NFS4_DELEGATION_STATEID_TYPE:
458 		return stateid;
459 	default:
460 		break;
461 	}
462 	return NULL;
463 }
464 
465 /* This is the error handling routine for processes that are allowed
466  * to sleep.
467  */
468 static int nfs4_do_handle_exception(struct nfs_server *server,
469 		int errorcode, struct nfs4_exception *exception)
470 {
471 	struct nfs_client *clp = server->nfs_client;
472 	struct nfs4_state *state = exception->state;
473 	const nfs4_stateid *stateid;
474 	struct inode *inode = exception->inode;
475 	int ret = errorcode;
476 
477 	exception->delay = 0;
478 	exception->recovering = 0;
479 	exception->retry = 0;
480 
481 	stateid = nfs4_recoverable_stateid(exception->stateid);
482 	if (stateid == NULL && state != NULL)
483 		stateid = nfs4_recoverable_stateid(&state->stateid);
484 
485 	switch(errorcode) {
486 		case 0:
487 			return 0;
488 		case -NFS4ERR_BADHANDLE:
489 		case -ESTALE:
490 			if (inode != NULL && S_ISREG(inode->i_mode))
491 				pnfs_destroy_layout(NFS_I(inode));
492 			break;
493 		case -NFS4ERR_DELEG_REVOKED:
494 		case -NFS4ERR_ADMIN_REVOKED:
495 		case -NFS4ERR_EXPIRED:
496 		case -NFS4ERR_BAD_STATEID:
497 		case -NFS4ERR_PARTNER_NO_AUTH:
498 			if (inode != NULL && stateid != NULL) {
499 				nfs_inode_find_state_and_recover(inode,
500 						stateid);
501 				goto wait_on_recovery;
502 			}
503 			fallthrough;
504 		case -NFS4ERR_OPENMODE:
505 			if (inode) {
506 				int err;
507 
508 				err = nfs_async_inode_return_delegation(inode,
509 						stateid);
510 				if (err == 0)
511 					goto wait_on_recovery;
512 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
513 					exception->retry = 1;
514 					break;
515 				}
516 			}
517 			if (state == NULL)
518 				break;
519 			ret = nfs4_schedule_stateid_recovery(server, state);
520 			if (ret < 0)
521 				break;
522 			goto wait_on_recovery;
523 		case -NFS4ERR_STALE_STATEID:
524 		case -NFS4ERR_STALE_CLIENTID:
525 			nfs4_schedule_lease_recovery(clp);
526 			goto wait_on_recovery;
527 		case -NFS4ERR_MOVED:
528 			ret = nfs4_schedule_migration_recovery(server);
529 			if (ret < 0)
530 				break;
531 			goto wait_on_recovery;
532 		case -NFS4ERR_LEASE_MOVED:
533 			nfs4_schedule_lease_moved_recovery(clp);
534 			goto wait_on_recovery;
535 #if defined(CONFIG_NFS_V4_1)
536 		case -NFS4ERR_BADSESSION:
537 		case -NFS4ERR_BADSLOT:
538 		case -NFS4ERR_BAD_HIGH_SLOT:
539 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
540 		case -NFS4ERR_DEADSESSION:
541 		case -NFS4ERR_SEQ_FALSE_RETRY:
542 		case -NFS4ERR_SEQ_MISORDERED:
543 			/* Handled in nfs41_sequence_process() */
544 			goto wait_on_recovery;
545 #endif /* defined(CONFIG_NFS_V4_1) */
546 		case -NFS4ERR_FILE_OPEN:
547 			if (exception->timeout > HZ) {
548 				/* We have retried a decent amount, time to
549 				 * fail
550 				 */
551 				ret = -EBUSY;
552 				break;
553 			}
554 			fallthrough;
555 		case -NFS4ERR_DELAY:
556 			nfs_inc_server_stats(server, NFSIOS_DELAY);
557 			fallthrough;
558 		case -NFS4ERR_GRACE:
559 		case -NFS4ERR_LAYOUTTRYLATER:
560 		case -NFS4ERR_RECALLCONFLICT:
561 			exception->delay = 1;
562 			return 0;
563 
564 		case -NFS4ERR_RETRY_UNCACHED_REP:
565 		case -NFS4ERR_OLD_STATEID:
566 			exception->retry = 1;
567 			break;
568 		case -NFS4ERR_BADOWNER:
569 			/* The following works around a Linux server bug! */
570 		case -NFS4ERR_BADNAME:
571 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
572 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
573 				exception->retry = 1;
574 				printk(KERN_WARNING "NFS: v4 server %s "
575 						"does not accept raw "
576 						"uid/gids. "
577 						"Reenabling the idmapper.\n",
578 						server->nfs_client->cl_hostname);
579 			}
580 	}
581 	/* We failed to handle the error */
582 	return nfs4_map_errors(ret);
583 wait_on_recovery:
584 	exception->recovering = 1;
585 	return 0;
586 }
587 
588 /* This is the error handling routine for processes that are allowed
589  * to sleep.
590  */
591 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
592 {
593 	struct nfs_client *clp = server->nfs_client;
594 	int ret;
595 
596 	ret = nfs4_do_handle_exception(server, errorcode, exception);
597 	if (exception->delay) {
598 		ret = nfs4_delay(&exception->timeout,
599 				exception->interruptible);
600 		goto out_retry;
601 	}
602 	if (exception->recovering) {
603 		if (exception->task_is_privileged)
604 			return -EDEADLOCK;
605 		ret = nfs4_wait_clnt_recover(clp);
606 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
607 			return -EIO;
608 		goto out_retry;
609 	}
610 	return ret;
611 out_retry:
612 	if (ret == 0)
613 		exception->retry = 1;
614 	return ret;
615 }
616 
617 static int
618 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
619 		int errorcode, struct nfs4_exception *exception)
620 {
621 	struct nfs_client *clp = server->nfs_client;
622 	int ret;
623 
624 	ret = nfs4_do_handle_exception(server, errorcode, exception);
625 	if (exception->delay) {
626 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
627 		goto out_retry;
628 	}
629 	if (exception->recovering) {
630 		if (exception->task_is_privileged)
631 			return -EDEADLOCK;
632 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
633 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
634 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
635 		goto out_retry;
636 	}
637 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
638 		ret = -EIO;
639 	return ret;
640 out_retry:
641 	if (ret == 0) {
642 		exception->retry = 1;
643 		/*
644 		 * For NFS4ERR_MOVED, the client transport will need to
645 		 * be recomputed after migration recovery has completed.
646 		 */
647 		if (errorcode == -NFS4ERR_MOVED)
648 			rpc_task_release_transport(task);
649 	}
650 	return ret;
651 }
652 
653 int
654 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
655 			struct nfs4_state *state, long *timeout)
656 {
657 	struct nfs4_exception exception = {
658 		.state = state,
659 	};
660 
661 	if (task->tk_status >= 0)
662 		return 0;
663 	if (timeout)
664 		exception.timeout = *timeout;
665 	task->tk_status = nfs4_async_handle_exception(task, server,
666 			task->tk_status,
667 			&exception);
668 	if (exception.delay && timeout)
669 		*timeout = exception.timeout;
670 	if (exception.retry)
671 		return -EAGAIN;
672 	return 0;
673 }
674 
675 /*
676  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
677  * or 'false' otherwise.
678  */
679 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
680 {
681 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
682 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
683 }
684 
685 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
686 {
687 	spin_lock(&clp->cl_lock);
688 	if (time_before(clp->cl_last_renewal,timestamp))
689 		clp->cl_last_renewal = timestamp;
690 	spin_unlock(&clp->cl_lock);
691 }
692 
693 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
694 {
695 	struct nfs_client *clp = server->nfs_client;
696 
697 	if (!nfs4_has_session(clp))
698 		do_renew_lease(clp, timestamp);
699 }
700 
701 struct nfs4_call_sync_data {
702 	const struct nfs_server *seq_server;
703 	struct nfs4_sequence_args *seq_args;
704 	struct nfs4_sequence_res *seq_res;
705 };
706 
707 void nfs4_init_sequence(struct nfs4_sequence_args *args,
708 			struct nfs4_sequence_res *res, int cache_reply,
709 			int privileged)
710 {
711 	args->sa_slot = NULL;
712 	args->sa_cache_this = cache_reply;
713 	args->sa_privileged = privileged;
714 
715 	res->sr_slot = NULL;
716 }
717 
718 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
719 {
720 	struct nfs4_slot *slot = res->sr_slot;
721 	struct nfs4_slot_table *tbl;
722 
723 	tbl = slot->table;
724 	spin_lock(&tbl->slot_tbl_lock);
725 	if (!nfs41_wake_and_assign_slot(tbl, slot))
726 		nfs4_free_slot(tbl, slot);
727 	spin_unlock(&tbl->slot_tbl_lock);
728 
729 	res->sr_slot = NULL;
730 }
731 
732 static int nfs40_sequence_done(struct rpc_task *task,
733 			       struct nfs4_sequence_res *res)
734 {
735 	if (res->sr_slot != NULL)
736 		nfs40_sequence_free_slot(res);
737 	return 1;
738 }
739 
740 #if defined(CONFIG_NFS_V4_1)
741 
742 static void nfs41_release_slot(struct nfs4_slot *slot)
743 {
744 	struct nfs4_session *session;
745 	struct nfs4_slot_table *tbl;
746 	bool send_new_highest_used_slotid = false;
747 
748 	if (!slot)
749 		return;
750 	tbl = slot->table;
751 	session = tbl->session;
752 
753 	/* Bump the slot sequence number */
754 	if (slot->seq_done)
755 		slot->seq_nr++;
756 	slot->seq_done = 0;
757 
758 	spin_lock(&tbl->slot_tbl_lock);
759 	/* Be nice to the server: try to ensure that the last transmitted
760 	 * value for highest_user_slotid <= target_highest_slotid
761 	 */
762 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
763 		send_new_highest_used_slotid = true;
764 
765 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
766 		send_new_highest_used_slotid = false;
767 		goto out_unlock;
768 	}
769 	nfs4_free_slot(tbl, slot);
770 
771 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
772 		send_new_highest_used_slotid = false;
773 out_unlock:
774 	spin_unlock(&tbl->slot_tbl_lock);
775 	if (send_new_highest_used_slotid)
776 		nfs41_notify_server(session->clp);
777 	if (waitqueue_active(&tbl->slot_waitq))
778 		wake_up_all(&tbl->slot_waitq);
779 }
780 
781 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
782 {
783 	nfs41_release_slot(res->sr_slot);
784 	res->sr_slot = NULL;
785 }
786 
787 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
788 		u32 seqnr)
789 {
790 	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
791 		slot->seq_nr_highest_sent = seqnr;
792 }
793 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
794 {
795 	nfs4_slot_sequence_record_sent(slot, seqnr);
796 	slot->seq_nr_last_acked = seqnr;
797 }
798 
799 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
800 				struct nfs4_slot *slot)
801 {
802 	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
803 	if (!IS_ERR(task))
804 		rpc_put_task_async(task);
805 }
806 
807 static int nfs41_sequence_process(struct rpc_task *task,
808 		struct nfs4_sequence_res *res)
809 {
810 	struct nfs4_session *session;
811 	struct nfs4_slot *slot = res->sr_slot;
812 	struct nfs_client *clp;
813 	int status;
814 	int ret = 1;
815 
816 	if (slot == NULL)
817 		goto out_noaction;
818 	/* don't increment the sequence number if the task wasn't sent */
819 	if (!RPC_WAS_SENT(task) || slot->seq_done)
820 		goto out;
821 
822 	session = slot->table->session;
823 	clp = session->clp;
824 
825 	trace_nfs4_sequence_done(session, res);
826 
827 	status = res->sr_status;
828 	if (task->tk_status == -NFS4ERR_DEADSESSION)
829 		status = -NFS4ERR_DEADSESSION;
830 
831 	/* Check the SEQUENCE operation status */
832 	switch (status) {
833 	case 0:
834 		/* Mark this sequence number as having been acked */
835 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
836 		/* Update the slot's sequence and clientid lease timer */
837 		slot->seq_done = 1;
838 		do_renew_lease(clp, res->sr_timestamp);
839 		/* Check sequence flags */
840 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
841 				!!slot->privileged);
842 		nfs41_update_target_slotid(slot->table, slot, res);
843 		break;
844 	case 1:
845 		/*
846 		 * sr_status remains 1 if an RPC level error occurred.
847 		 * The server may or may not have processed the sequence
848 		 * operation..
849 		 */
850 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
851 		slot->seq_done = 1;
852 		goto out;
853 	case -NFS4ERR_DELAY:
854 		/* The server detected a resend of the RPC call and
855 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
856 		 * of RFC5661.
857 		 */
858 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
859 			__func__,
860 			slot->slot_nr,
861 			slot->seq_nr);
862 		goto out_retry;
863 	case -NFS4ERR_RETRY_UNCACHED_REP:
864 	case -NFS4ERR_SEQ_FALSE_RETRY:
865 		/*
866 		 * The server thinks we tried to replay a request.
867 		 * Retry the call after bumping the sequence ID.
868 		 */
869 		nfs4_slot_sequence_acked(slot, slot->seq_nr);
870 		goto retry_new_seq;
871 	case -NFS4ERR_BADSLOT:
872 		/*
873 		 * The slot id we used was probably retired. Try again
874 		 * using a different slot id.
875 		 */
876 		if (slot->slot_nr < slot->table->target_highest_slotid)
877 			goto session_recover;
878 		goto retry_nowait;
879 	case -NFS4ERR_SEQ_MISORDERED:
880 		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
881 		/*
882 		 * Were one or more calls using this slot interrupted?
883 		 * If the server never received the request, then our
884 		 * transmitted slot sequence number may be too high. However,
885 		 * if the server did receive the request then it might
886 		 * accidentally give us a reply with a mismatched operation.
887 		 * We can sort this out by sending a lone sequence operation
888 		 * to the server on the same slot.
889 		 */
890 		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
891 			slot->seq_nr--;
892 			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
893 				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
894 				res->sr_slot = NULL;
895 			}
896 			goto retry_nowait;
897 		}
898 		/*
899 		 * RFC5661:
900 		 * A retry might be sent while the original request is
901 		 * still in progress on the replier. The replier SHOULD
902 		 * deal with the issue by returning NFS4ERR_DELAY as the
903 		 * reply to SEQUENCE or CB_SEQUENCE operation, but
904 		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
905 		 *
906 		 * Restart the search after a delay.
907 		 */
908 		slot->seq_nr = slot->seq_nr_highest_sent;
909 		goto out_retry;
910 	case -NFS4ERR_BADSESSION:
911 	case -NFS4ERR_DEADSESSION:
912 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
913 		goto session_recover;
914 	default:
915 		/* Just update the slot sequence no. */
916 		slot->seq_done = 1;
917 	}
918 out:
919 	/* The session may be reset by one of the error handlers. */
920 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
921 out_noaction:
922 	return ret;
923 session_recover:
924 	set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
925 	nfs4_schedule_session_recovery(session, status);
926 	dprintk("%s ERROR: %d Reset session\n", __func__, status);
927 	nfs41_sequence_free_slot(res);
928 	goto out;
929 retry_new_seq:
930 	++slot->seq_nr;
931 retry_nowait:
932 	if (rpc_restart_call_prepare(task)) {
933 		nfs41_sequence_free_slot(res);
934 		task->tk_status = 0;
935 		ret = 0;
936 	}
937 	goto out;
938 out_retry:
939 	if (!rpc_restart_call(task))
940 		goto out;
941 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
942 	return 0;
943 }
944 
945 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
946 {
947 	if (!nfs41_sequence_process(task, res))
948 		return 0;
949 	if (res->sr_slot != NULL)
950 		nfs41_sequence_free_slot(res);
951 	return 1;
952 
953 }
954 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
955 
956 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
957 {
958 	if (res->sr_slot == NULL)
959 		return 1;
960 	if (res->sr_slot->table->session != NULL)
961 		return nfs41_sequence_process(task, res);
962 	return nfs40_sequence_done(task, res);
963 }
964 
965 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
966 {
967 	if (res->sr_slot != NULL) {
968 		if (res->sr_slot->table->session != NULL)
969 			nfs41_sequence_free_slot(res);
970 		else
971 			nfs40_sequence_free_slot(res);
972 	}
973 }
974 
975 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
976 {
977 	if (res->sr_slot == NULL)
978 		return 1;
979 	if (!res->sr_slot->table->session)
980 		return nfs40_sequence_done(task, res);
981 	return nfs41_sequence_done(task, res);
982 }
983 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
984 
985 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
986 {
987 	struct nfs4_call_sync_data *data = calldata;
988 
989 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
990 
991 	nfs4_setup_sequence(data->seq_server->nfs_client,
992 			    data->seq_args, data->seq_res, task);
993 }
994 
995 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
996 {
997 	struct nfs4_call_sync_data *data = calldata;
998 
999 	nfs41_sequence_done(task, data->seq_res);
1000 }
1001 
1002 static const struct rpc_call_ops nfs41_call_sync_ops = {
1003 	.rpc_call_prepare = nfs41_call_sync_prepare,
1004 	.rpc_call_done = nfs41_call_sync_done,
1005 };
1006 
1007 #else	/* !CONFIG_NFS_V4_1 */
1008 
1009 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1010 {
1011 	return nfs40_sequence_done(task, res);
1012 }
1013 
1014 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1015 {
1016 	if (res->sr_slot != NULL)
1017 		nfs40_sequence_free_slot(res);
1018 }
1019 
1020 int nfs4_sequence_done(struct rpc_task *task,
1021 		       struct nfs4_sequence_res *res)
1022 {
1023 	return nfs40_sequence_done(task, res);
1024 }
1025 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1026 
1027 #endif	/* !CONFIG_NFS_V4_1 */
1028 
1029 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1030 {
1031 	res->sr_timestamp = jiffies;
1032 	res->sr_status_flags = 0;
1033 	res->sr_status = 1;
1034 }
1035 
1036 static
1037 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1038 		struct nfs4_sequence_res *res,
1039 		struct nfs4_slot *slot)
1040 {
1041 	if (!slot)
1042 		return;
1043 	slot->privileged = args->sa_privileged ? 1 : 0;
1044 	args->sa_slot = slot;
1045 
1046 	res->sr_slot = slot;
1047 }
1048 
1049 int nfs4_setup_sequence(struct nfs_client *client,
1050 			struct nfs4_sequence_args *args,
1051 			struct nfs4_sequence_res *res,
1052 			struct rpc_task *task)
1053 {
1054 	struct nfs4_session *session = nfs4_get_session(client);
1055 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1056 	struct nfs4_slot *slot;
1057 
1058 	/* slot already allocated? */
1059 	if (res->sr_slot != NULL)
1060 		goto out_start;
1061 
1062 	if (session)
1063 		tbl = &session->fc_slot_table;
1064 
1065 	spin_lock(&tbl->slot_tbl_lock);
1066 	/* The state manager will wait until the slot table is empty */
1067 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1068 		goto out_sleep;
1069 
1070 	slot = nfs4_alloc_slot(tbl);
1071 	if (IS_ERR(slot)) {
1072 		if (slot == ERR_PTR(-ENOMEM))
1073 			goto out_sleep_timeout;
1074 		goto out_sleep;
1075 	}
1076 	spin_unlock(&tbl->slot_tbl_lock);
1077 
1078 	nfs4_sequence_attach_slot(args, res, slot);
1079 
1080 	trace_nfs4_setup_sequence(session, args);
1081 out_start:
1082 	nfs41_sequence_res_init(res);
1083 	rpc_call_start(task);
1084 	return 0;
1085 out_sleep_timeout:
1086 	/* Try again in 1/4 second */
1087 	if (args->sa_privileged)
1088 		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1089 				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1090 	else
1091 		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1092 				NULL, jiffies + (HZ >> 2));
1093 	spin_unlock(&tbl->slot_tbl_lock);
1094 	return -EAGAIN;
1095 out_sleep:
1096 	if (args->sa_privileged)
1097 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1098 				RPC_PRIORITY_PRIVILEGED);
1099 	else
1100 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1101 	spin_unlock(&tbl->slot_tbl_lock);
1102 	return -EAGAIN;
1103 }
1104 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1105 
1106 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1107 {
1108 	struct nfs4_call_sync_data *data = calldata;
1109 	nfs4_setup_sequence(data->seq_server->nfs_client,
1110 				data->seq_args, data->seq_res, task);
1111 }
1112 
1113 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1114 {
1115 	struct nfs4_call_sync_data *data = calldata;
1116 	nfs4_sequence_done(task, data->seq_res);
1117 }
1118 
1119 static const struct rpc_call_ops nfs40_call_sync_ops = {
1120 	.rpc_call_prepare = nfs40_call_sync_prepare,
1121 	.rpc_call_done = nfs40_call_sync_done,
1122 };
1123 
1124 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1125 {
1126 	int ret;
1127 	struct rpc_task *task;
1128 
1129 	task = rpc_run_task(task_setup);
1130 	if (IS_ERR(task))
1131 		return PTR_ERR(task);
1132 
1133 	ret = task->tk_status;
1134 	rpc_put_task(task);
1135 	return ret;
1136 }
1137 
1138 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1139 			     struct nfs_server *server,
1140 			     struct rpc_message *msg,
1141 			     struct nfs4_sequence_args *args,
1142 			     struct nfs4_sequence_res *res,
1143 			     unsigned short task_flags)
1144 {
1145 	struct nfs_client *clp = server->nfs_client;
1146 	struct nfs4_call_sync_data data = {
1147 		.seq_server = server,
1148 		.seq_args = args,
1149 		.seq_res = res,
1150 	};
1151 	struct rpc_task_setup task_setup = {
1152 		.rpc_client = clnt,
1153 		.rpc_message = msg,
1154 		.callback_ops = clp->cl_mvops->call_sync_ops,
1155 		.callback_data = &data,
1156 		.flags = task_flags,
1157 	};
1158 
1159 	return nfs4_call_sync_custom(&task_setup);
1160 }
1161 
1162 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1163 				   struct nfs_server *server,
1164 				   struct rpc_message *msg,
1165 				   struct nfs4_sequence_args *args,
1166 				   struct nfs4_sequence_res *res)
1167 {
1168 	unsigned short task_flags = 0;
1169 
1170 	if (server->caps & NFS_CAP_MOVEABLE)
1171 		task_flags = RPC_TASK_MOVEABLE;
1172 	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1173 }
1174 
1175 
1176 int nfs4_call_sync(struct rpc_clnt *clnt,
1177 		   struct nfs_server *server,
1178 		   struct rpc_message *msg,
1179 		   struct nfs4_sequence_args *args,
1180 		   struct nfs4_sequence_res *res,
1181 		   int cache_reply)
1182 {
1183 	nfs4_init_sequence(args, res, cache_reply, 0);
1184 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1185 }
1186 
1187 static void
1188 nfs4_inc_nlink_locked(struct inode *inode)
1189 {
1190 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1191 					     NFS_INO_INVALID_CTIME |
1192 					     NFS_INO_INVALID_NLINK);
1193 	inc_nlink(inode);
1194 }
1195 
1196 static void
1197 nfs4_inc_nlink(struct inode *inode)
1198 {
1199 	spin_lock(&inode->i_lock);
1200 	nfs4_inc_nlink_locked(inode);
1201 	spin_unlock(&inode->i_lock);
1202 }
1203 
1204 static void
1205 nfs4_dec_nlink_locked(struct inode *inode)
1206 {
1207 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1208 					     NFS_INO_INVALID_CTIME |
1209 					     NFS_INO_INVALID_NLINK);
1210 	drop_nlink(inode);
1211 }
1212 
1213 static void
1214 nfs4_update_changeattr_locked(struct inode *inode,
1215 		struct nfs4_change_info *cinfo,
1216 		unsigned long timestamp, unsigned long cache_validity)
1217 {
1218 	struct nfs_inode *nfsi = NFS_I(inode);
1219 	u64 change_attr = inode_peek_iversion_raw(inode);
1220 
1221 	cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1222 	if (S_ISDIR(inode->i_mode))
1223 		cache_validity |= NFS_INO_INVALID_DATA;
1224 
1225 	switch (NFS_SERVER(inode)->change_attr_type) {
1226 	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1227 		if (cinfo->after == change_attr)
1228 			goto out;
1229 		break;
1230 	default:
1231 		if ((s64)(change_attr - cinfo->after) >= 0)
1232 			goto out;
1233 	}
1234 
1235 	inode_set_iversion_raw(inode, cinfo->after);
1236 	if (!cinfo->atomic || cinfo->before != change_attr) {
1237 		if (S_ISDIR(inode->i_mode))
1238 			nfs_force_lookup_revalidate(inode);
1239 
1240 		if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1241 			cache_validity |=
1242 				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1243 				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1244 				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1245 				NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1246 		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1247 	}
1248 	nfsi->attrtimeo_timestamp = jiffies;
1249 	nfsi->read_cache_jiffies = timestamp;
1250 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1251 	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1252 out:
1253 	nfs_set_cache_invalid(inode, cache_validity);
1254 }
1255 
1256 void
1257 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1258 		unsigned long timestamp, unsigned long cache_validity)
1259 {
1260 	spin_lock(&dir->i_lock);
1261 	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1262 	spin_unlock(&dir->i_lock);
1263 }
1264 
1265 struct nfs4_open_createattrs {
1266 	struct nfs4_label *label;
1267 	struct iattr *sattr;
1268 	const __u32 verf[2];
1269 };
1270 
1271 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1272 		int err, struct nfs4_exception *exception)
1273 {
1274 	if (err != -EINVAL)
1275 		return false;
1276 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1277 		return false;
1278 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1279 	exception->retry = 1;
1280 	return true;
1281 }
1282 
1283 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1284 {
1285 	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1286 }
1287 
1288 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1289 {
1290 	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1291 
1292 	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1293 }
1294 
1295 static u32
1296 nfs4_map_atomic_open_share(struct nfs_server *server,
1297 		fmode_t fmode, int openflags)
1298 {
1299 	u32 res = 0;
1300 
1301 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1302 	case FMODE_READ:
1303 		res = NFS4_SHARE_ACCESS_READ;
1304 		break;
1305 	case FMODE_WRITE:
1306 		res = NFS4_SHARE_ACCESS_WRITE;
1307 		break;
1308 	case FMODE_READ|FMODE_WRITE:
1309 		res = NFS4_SHARE_ACCESS_BOTH;
1310 	}
1311 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1312 		goto out;
1313 	/* Want no delegation if we're using O_DIRECT */
1314 	if (openflags & O_DIRECT)
1315 		res |= NFS4_SHARE_WANT_NO_DELEG;
1316 out:
1317 	return res;
1318 }
1319 
1320 static enum open_claim_type4
1321 nfs4_map_atomic_open_claim(struct nfs_server *server,
1322 		enum open_claim_type4 claim)
1323 {
1324 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1325 		return claim;
1326 	switch (claim) {
1327 	default:
1328 		return claim;
1329 	case NFS4_OPEN_CLAIM_FH:
1330 		return NFS4_OPEN_CLAIM_NULL;
1331 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1332 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1333 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1334 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1335 	}
1336 }
1337 
1338 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1339 {
1340 	p->o_res.f_attr = &p->f_attr;
1341 	p->o_res.seqid = p->o_arg.seqid;
1342 	p->c_res.seqid = p->c_arg.seqid;
1343 	p->o_res.server = p->o_arg.server;
1344 	p->o_res.access_request = p->o_arg.access;
1345 	nfs_fattr_init(&p->f_attr);
1346 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1347 }
1348 
1349 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1350 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1351 		const struct nfs4_open_createattrs *c,
1352 		enum open_claim_type4 claim,
1353 		gfp_t gfp_mask)
1354 {
1355 	struct dentry *parent = dget_parent(dentry);
1356 	struct inode *dir = d_inode(parent);
1357 	struct nfs_server *server = NFS_SERVER(dir);
1358 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1359 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1360 	struct nfs4_opendata *p;
1361 
1362 	p = kzalloc(sizeof(*p), gfp_mask);
1363 	if (p == NULL)
1364 		goto err;
1365 
1366 	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1367 	if (IS_ERR(p->f_attr.label))
1368 		goto err_free_p;
1369 
1370 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1371 	if (IS_ERR(p->a_label))
1372 		goto err_free_f;
1373 
1374 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1375 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1376 	if (IS_ERR(p->o_arg.seqid))
1377 		goto err_free_label;
1378 	nfs_sb_active(dentry->d_sb);
1379 	p->dentry = dget(dentry);
1380 	p->dir = parent;
1381 	p->owner = sp;
1382 	atomic_inc(&sp->so_count);
1383 	p->o_arg.open_flags = flags;
1384 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1385 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1386 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1387 			fmode, flags);
1388 	if (flags & O_CREAT) {
1389 		p->o_arg.umask = current_umask();
1390 		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1391 		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1392 			p->o_arg.u.attrs = &p->attrs;
1393 			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1394 
1395 			memcpy(p->o_arg.u.verifier.data, c->verf,
1396 					sizeof(p->o_arg.u.verifier.data));
1397 		}
1398 	}
1399 	/* ask server to check for all possible rights as results
1400 	 * are cached */
1401 	switch (p->o_arg.claim) {
1402 	default:
1403 		break;
1404 	case NFS4_OPEN_CLAIM_NULL:
1405 	case NFS4_OPEN_CLAIM_FH:
1406 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1407 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1408 				  NFS4_ACCESS_EXECUTE |
1409 				  nfs_access_xattr_mask(server);
1410 	}
1411 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1412 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1413 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1414 	p->o_arg.name = &dentry->d_name;
1415 	p->o_arg.server = server;
1416 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1417 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1418 	switch (p->o_arg.claim) {
1419 	case NFS4_OPEN_CLAIM_NULL:
1420 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1421 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1422 		p->o_arg.fh = NFS_FH(dir);
1423 		break;
1424 	case NFS4_OPEN_CLAIM_PREVIOUS:
1425 	case NFS4_OPEN_CLAIM_FH:
1426 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1427 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1428 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1429 	}
1430 	p->c_arg.fh = &p->o_res.fh;
1431 	p->c_arg.stateid = &p->o_res.stateid;
1432 	p->c_arg.seqid = p->o_arg.seqid;
1433 	nfs4_init_opendata_res(p);
1434 	kref_init(&p->kref);
1435 	return p;
1436 
1437 err_free_label:
1438 	nfs4_label_free(p->a_label);
1439 err_free_f:
1440 	nfs4_label_free(p->f_attr.label);
1441 err_free_p:
1442 	kfree(p);
1443 err:
1444 	dput(parent);
1445 	return NULL;
1446 }
1447 
1448 static void nfs4_opendata_free(struct kref *kref)
1449 {
1450 	struct nfs4_opendata *p = container_of(kref,
1451 			struct nfs4_opendata, kref);
1452 	struct super_block *sb = p->dentry->d_sb;
1453 
1454 	nfs4_lgopen_release(p->lgp);
1455 	nfs_free_seqid(p->o_arg.seqid);
1456 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1457 	if (p->state != NULL)
1458 		nfs4_put_open_state(p->state);
1459 	nfs4_put_state_owner(p->owner);
1460 
1461 	nfs4_label_free(p->a_label);
1462 	nfs4_label_free(p->f_attr.label);
1463 
1464 	dput(p->dir);
1465 	dput(p->dentry);
1466 	nfs_sb_deactive(sb);
1467 	nfs_fattr_free_names(&p->f_attr);
1468 	kfree(p->f_attr.mdsthreshold);
1469 	kfree(p);
1470 }
1471 
1472 static void nfs4_opendata_put(struct nfs4_opendata *p)
1473 {
1474 	if (p != NULL)
1475 		kref_put(&p->kref, nfs4_opendata_free);
1476 }
1477 
1478 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1479 		fmode_t fmode)
1480 {
1481 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1482 	case FMODE_READ|FMODE_WRITE:
1483 		return state->n_rdwr != 0;
1484 	case FMODE_WRITE:
1485 		return state->n_wronly != 0;
1486 	case FMODE_READ:
1487 		return state->n_rdonly != 0;
1488 	}
1489 	WARN_ON_ONCE(1);
1490 	return false;
1491 }
1492 
1493 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1494 		int open_mode, enum open_claim_type4 claim)
1495 {
1496 	int ret = 0;
1497 
1498 	if (open_mode & (O_EXCL|O_TRUNC))
1499 		goto out;
1500 	switch (claim) {
1501 	case NFS4_OPEN_CLAIM_NULL:
1502 	case NFS4_OPEN_CLAIM_FH:
1503 		goto out;
1504 	default:
1505 		break;
1506 	}
1507 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1508 		case FMODE_READ:
1509 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1510 				&& state->n_rdonly != 0;
1511 			break;
1512 		case FMODE_WRITE:
1513 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1514 				&& state->n_wronly != 0;
1515 			break;
1516 		case FMODE_READ|FMODE_WRITE:
1517 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1518 				&& state->n_rdwr != 0;
1519 	}
1520 out:
1521 	return ret;
1522 }
1523 
1524 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1525 		enum open_claim_type4 claim)
1526 {
1527 	if (delegation == NULL)
1528 		return 0;
1529 	if ((delegation->type & fmode) != fmode)
1530 		return 0;
1531 	switch (claim) {
1532 	case NFS4_OPEN_CLAIM_NULL:
1533 	case NFS4_OPEN_CLAIM_FH:
1534 		break;
1535 	case NFS4_OPEN_CLAIM_PREVIOUS:
1536 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1537 			break;
1538 		fallthrough;
1539 	default:
1540 		return 0;
1541 	}
1542 	nfs_mark_delegation_referenced(delegation);
1543 	return 1;
1544 }
1545 
1546 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1547 {
1548 	switch (fmode) {
1549 		case FMODE_WRITE:
1550 			state->n_wronly++;
1551 			break;
1552 		case FMODE_READ:
1553 			state->n_rdonly++;
1554 			break;
1555 		case FMODE_READ|FMODE_WRITE:
1556 			state->n_rdwr++;
1557 	}
1558 	nfs4_state_set_mode_locked(state, state->state | fmode);
1559 }
1560 
1561 #ifdef CONFIG_NFS_V4_1
1562 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1563 {
1564 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1565 		return true;
1566 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1567 		return true;
1568 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1569 		return true;
1570 	return false;
1571 }
1572 #endif /* CONFIG_NFS_V4_1 */
1573 
1574 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1575 {
1576 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1577 		wake_up_all(&state->waitq);
1578 }
1579 
1580 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1581 {
1582 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1583 	bool need_recover = false;
1584 
1585 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1586 		need_recover = true;
1587 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1588 		need_recover = true;
1589 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1590 		need_recover = true;
1591 	if (need_recover)
1592 		nfs4_state_mark_reclaim_nograce(clp, state);
1593 }
1594 
1595 /*
1596  * Check for whether or not the caller may update the open stateid
1597  * to the value passed in by stateid.
1598  *
1599  * Note: This function relies heavily on the server implementing
1600  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1601  * correctly.
1602  * i.e. The stateid seqids have to be initialised to 1, and
1603  * are then incremented on every state transition.
1604  */
1605 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1606 		const nfs4_stateid *stateid)
1607 {
1608 	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1609 		/* The common case - we're updating to a new sequence number */
1610 		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1611 			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1612 				return true;
1613 			return false;
1614 		}
1615 		/* The server returned a new stateid */
1616 	}
1617 	/* This is the first OPEN in this generation */
1618 	if (stateid->seqid == cpu_to_be32(1))
1619 		return true;
1620 	return false;
1621 }
1622 
1623 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1624 {
1625 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1626 		return;
1627 	if (state->n_wronly)
1628 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1629 	if (state->n_rdonly)
1630 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1631 	if (state->n_rdwr)
1632 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1633 	set_bit(NFS_OPEN_STATE, &state->flags);
1634 }
1635 
1636 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1637 		nfs4_stateid *stateid, fmode_t fmode)
1638 {
1639 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1640 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1641 	case FMODE_WRITE:
1642 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1643 		break;
1644 	case FMODE_READ:
1645 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1646 		break;
1647 	case 0:
1648 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1649 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1650 		clear_bit(NFS_OPEN_STATE, &state->flags);
1651 	}
1652 	if (stateid == NULL)
1653 		return;
1654 	/* Handle OPEN+OPEN_DOWNGRADE races */
1655 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1656 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1657 		nfs_resync_open_stateid_locked(state);
1658 		goto out;
1659 	}
1660 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1661 		nfs4_stateid_copy(&state->stateid, stateid);
1662 	nfs4_stateid_copy(&state->open_stateid, stateid);
1663 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1664 out:
1665 	nfs_state_log_update_open_stateid(state);
1666 }
1667 
1668 static void nfs_clear_open_stateid(struct nfs4_state *state,
1669 	nfs4_stateid *arg_stateid,
1670 	nfs4_stateid *stateid, fmode_t fmode)
1671 {
1672 	write_seqlock(&state->seqlock);
1673 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1674 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1675 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1676 	write_sequnlock(&state->seqlock);
1677 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1678 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1679 }
1680 
1681 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1682 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1683 	__must_hold(&state->owner->so_lock)
1684 	__must_hold(&state->seqlock)
1685 	__must_hold(RCU)
1686 
1687 {
1688 	DEFINE_WAIT(wait);
1689 	int status = 0;
1690 	for (;;) {
1691 
1692 		if (nfs_stateid_is_sequential(state, stateid))
1693 			break;
1694 
1695 		if (status)
1696 			break;
1697 		/* Rely on seqids for serialisation with NFSv4.0 */
1698 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1699 			break;
1700 
1701 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1702 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1703 		/*
1704 		 * Ensure we process the state changes in the same order
1705 		 * in which the server processed them by delaying the
1706 		 * update of the stateid until we are in sequence.
1707 		 */
1708 		write_sequnlock(&state->seqlock);
1709 		spin_unlock(&state->owner->so_lock);
1710 		rcu_read_unlock();
1711 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1712 
1713 		if (!fatal_signal_pending(current)) {
1714 			if (schedule_timeout(5*HZ) == 0)
1715 				status = -EAGAIN;
1716 			else
1717 				status = 0;
1718 		} else
1719 			status = -EINTR;
1720 		finish_wait(&state->waitq, &wait);
1721 		rcu_read_lock();
1722 		spin_lock(&state->owner->so_lock);
1723 		write_seqlock(&state->seqlock);
1724 	}
1725 
1726 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1727 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1728 		nfs4_stateid_copy(freeme, &state->open_stateid);
1729 		nfs_test_and_clear_all_open_stateid(state);
1730 	}
1731 
1732 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1733 		nfs4_stateid_copy(&state->stateid, stateid);
1734 	nfs4_stateid_copy(&state->open_stateid, stateid);
1735 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1736 	nfs_state_log_update_open_stateid(state);
1737 }
1738 
1739 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1740 		const nfs4_stateid *open_stateid,
1741 		fmode_t fmode,
1742 		nfs4_stateid *freeme)
1743 {
1744 	/*
1745 	 * Protect the call to nfs4_state_set_mode_locked and
1746 	 * serialise the stateid update
1747 	 */
1748 	write_seqlock(&state->seqlock);
1749 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1750 	switch (fmode) {
1751 	case FMODE_READ:
1752 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1753 		break;
1754 	case FMODE_WRITE:
1755 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1756 		break;
1757 	case FMODE_READ|FMODE_WRITE:
1758 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1759 	}
1760 	set_bit(NFS_OPEN_STATE, &state->flags);
1761 	write_sequnlock(&state->seqlock);
1762 }
1763 
1764 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1765 {
1766 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1767 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1768 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1769 	clear_bit(NFS_OPEN_STATE, &state->flags);
1770 }
1771 
1772 static void nfs_state_set_delegation(struct nfs4_state *state,
1773 		const nfs4_stateid *deleg_stateid,
1774 		fmode_t fmode)
1775 {
1776 	/*
1777 	 * Protect the call to nfs4_state_set_mode_locked and
1778 	 * serialise the stateid update
1779 	 */
1780 	write_seqlock(&state->seqlock);
1781 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1782 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1783 	write_sequnlock(&state->seqlock);
1784 }
1785 
1786 static void nfs_state_clear_delegation(struct nfs4_state *state)
1787 {
1788 	write_seqlock(&state->seqlock);
1789 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1790 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1791 	write_sequnlock(&state->seqlock);
1792 }
1793 
1794 int update_open_stateid(struct nfs4_state *state,
1795 		const nfs4_stateid *open_stateid,
1796 		const nfs4_stateid *delegation,
1797 		fmode_t fmode)
1798 {
1799 	struct nfs_server *server = NFS_SERVER(state->inode);
1800 	struct nfs_client *clp = server->nfs_client;
1801 	struct nfs_inode *nfsi = NFS_I(state->inode);
1802 	struct nfs_delegation *deleg_cur;
1803 	nfs4_stateid freeme = { };
1804 	int ret = 0;
1805 
1806 	fmode &= (FMODE_READ|FMODE_WRITE);
1807 
1808 	rcu_read_lock();
1809 	spin_lock(&state->owner->so_lock);
1810 	if (open_stateid != NULL) {
1811 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1812 		ret = 1;
1813 	}
1814 
1815 	deleg_cur = nfs4_get_valid_delegation(state->inode);
1816 	if (deleg_cur == NULL)
1817 		goto no_delegation;
1818 
1819 	spin_lock(&deleg_cur->lock);
1820 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1821 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1822 	    (deleg_cur->type & fmode) != fmode)
1823 		goto no_delegation_unlock;
1824 
1825 	if (delegation == NULL)
1826 		delegation = &deleg_cur->stateid;
1827 	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1828 		goto no_delegation_unlock;
1829 
1830 	nfs_mark_delegation_referenced(deleg_cur);
1831 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1832 	ret = 1;
1833 no_delegation_unlock:
1834 	spin_unlock(&deleg_cur->lock);
1835 no_delegation:
1836 	if (ret)
1837 		update_open_stateflags(state, fmode);
1838 	spin_unlock(&state->owner->so_lock);
1839 	rcu_read_unlock();
1840 
1841 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1842 		nfs4_schedule_state_manager(clp);
1843 	if (freeme.type != 0)
1844 		nfs4_test_and_free_stateid(server, &freeme,
1845 				state->owner->so_cred);
1846 
1847 	return ret;
1848 }
1849 
1850 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1851 		const nfs4_stateid *stateid)
1852 {
1853 	struct nfs4_state *state = lsp->ls_state;
1854 	bool ret = false;
1855 
1856 	spin_lock(&state->state_lock);
1857 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1858 		goto out_noupdate;
1859 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1860 		goto out_noupdate;
1861 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1862 	ret = true;
1863 out_noupdate:
1864 	spin_unlock(&state->state_lock);
1865 	return ret;
1866 }
1867 
1868 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1869 {
1870 	struct nfs_delegation *delegation;
1871 
1872 	fmode &= FMODE_READ|FMODE_WRITE;
1873 	rcu_read_lock();
1874 	delegation = nfs4_get_valid_delegation(inode);
1875 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1876 		rcu_read_unlock();
1877 		return;
1878 	}
1879 	rcu_read_unlock();
1880 	nfs4_inode_return_delegation(inode);
1881 }
1882 
1883 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1884 {
1885 	struct nfs4_state *state = opendata->state;
1886 	struct nfs_delegation *delegation;
1887 	int open_mode = opendata->o_arg.open_flags;
1888 	fmode_t fmode = opendata->o_arg.fmode;
1889 	enum open_claim_type4 claim = opendata->o_arg.claim;
1890 	nfs4_stateid stateid;
1891 	int ret = -EAGAIN;
1892 
1893 	for (;;) {
1894 		spin_lock(&state->owner->so_lock);
1895 		if (can_open_cached(state, fmode, open_mode, claim)) {
1896 			update_open_stateflags(state, fmode);
1897 			spin_unlock(&state->owner->so_lock);
1898 			goto out_return_state;
1899 		}
1900 		spin_unlock(&state->owner->so_lock);
1901 		rcu_read_lock();
1902 		delegation = nfs4_get_valid_delegation(state->inode);
1903 		if (!can_open_delegated(delegation, fmode, claim)) {
1904 			rcu_read_unlock();
1905 			break;
1906 		}
1907 		/* Save the delegation */
1908 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1909 		rcu_read_unlock();
1910 		nfs_release_seqid(opendata->o_arg.seqid);
1911 		if (!opendata->is_recover) {
1912 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1913 			if (ret != 0)
1914 				goto out;
1915 		}
1916 		ret = -EAGAIN;
1917 
1918 		/* Try to update the stateid using the delegation */
1919 		if (update_open_stateid(state, NULL, &stateid, fmode))
1920 			goto out_return_state;
1921 	}
1922 out:
1923 	return ERR_PTR(ret);
1924 out_return_state:
1925 	refcount_inc(&state->count);
1926 	return state;
1927 }
1928 
1929 static void
1930 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1931 {
1932 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1933 	struct nfs_delegation *delegation;
1934 	int delegation_flags = 0;
1935 
1936 	rcu_read_lock();
1937 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1938 	if (delegation)
1939 		delegation_flags = delegation->flags;
1940 	rcu_read_unlock();
1941 	switch (data->o_arg.claim) {
1942 	default:
1943 		break;
1944 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1945 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1946 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1947 				   "returning a delegation for "
1948 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1949 				   clp->cl_hostname);
1950 		return;
1951 	}
1952 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1953 		nfs_inode_set_delegation(state->inode,
1954 				data->owner->so_cred,
1955 				data->o_res.delegation_type,
1956 				&data->o_res.delegation,
1957 				data->o_res.pagemod_limit);
1958 	else
1959 		nfs_inode_reclaim_delegation(state->inode,
1960 				data->owner->so_cred,
1961 				data->o_res.delegation_type,
1962 				&data->o_res.delegation,
1963 				data->o_res.pagemod_limit);
1964 
1965 	if (data->o_res.do_recall)
1966 		nfs_async_inode_return_delegation(state->inode,
1967 						  &data->o_res.delegation);
1968 }
1969 
1970 /*
1971  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1972  * and update the nfs4_state.
1973  */
1974 static struct nfs4_state *
1975 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1976 {
1977 	struct inode *inode = data->state->inode;
1978 	struct nfs4_state *state = data->state;
1979 	int ret;
1980 
1981 	if (!data->rpc_done) {
1982 		if (data->rpc_status)
1983 			return ERR_PTR(data->rpc_status);
1984 		return nfs4_try_open_cached(data);
1985 	}
1986 
1987 	ret = nfs_refresh_inode(inode, &data->f_attr);
1988 	if (ret)
1989 		return ERR_PTR(ret);
1990 
1991 	if (data->o_res.delegation_type != 0)
1992 		nfs4_opendata_check_deleg(data, state);
1993 
1994 	if (!update_open_stateid(state, &data->o_res.stateid,
1995 				NULL, data->o_arg.fmode))
1996 		return ERR_PTR(-EAGAIN);
1997 	refcount_inc(&state->count);
1998 
1999 	return state;
2000 }
2001 
2002 static struct inode *
2003 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2004 {
2005 	struct inode *inode;
2006 
2007 	switch (data->o_arg.claim) {
2008 	case NFS4_OPEN_CLAIM_NULL:
2009 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2010 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2011 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2012 			return ERR_PTR(-EAGAIN);
2013 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2014 				&data->f_attr);
2015 		break;
2016 	default:
2017 		inode = d_inode(data->dentry);
2018 		ihold(inode);
2019 		nfs_refresh_inode(inode, &data->f_attr);
2020 	}
2021 	return inode;
2022 }
2023 
2024 static struct nfs4_state *
2025 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2026 {
2027 	struct nfs4_state *state;
2028 	struct inode *inode;
2029 
2030 	inode = nfs4_opendata_get_inode(data);
2031 	if (IS_ERR(inode))
2032 		return ERR_CAST(inode);
2033 	if (data->state != NULL && data->state->inode == inode) {
2034 		state = data->state;
2035 		refcount_inc(&state->count);
2036 	} else
2037 		state = nfs4_get_open_state(inode, data->owner);
2038 	iput(inode);
2039 	if (state == NULL)
2040 		state = ERR_PTR(-ENOMEM);
2041 	return state;
2042 }
2043 
2044 static struct nfs4_state *
2045 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2046 {
2047 	struct nfs4_state *state;
2048 
2049 	if (!data->rpc_done) {
2050 		state = nfs4_try_open_cached(data);
2051 		trace_nfs4_cached_open(data->state);
2052 		goto out;
2053 	}
2054 
2055 	state = nfs4_opendata_find_nfs4_state(data);
2056 	if (IS_ERR(state))
2057 		goto out;
2058 
2059 	if (data->o_res.delegation_type != 0)
2060 		nfs4_opendata_check_deleg(data, state);
2061 	if (!update_open_stateid(state, &data->o_res.stateid,
2062 				NULL, data->o_arg.fmode)) {
2063 		nfs4_put_open_state(state);
2064 		state = ERR_PTR(-EAGAIN);
2065 	}
2066 out:
2067 	nfs_release_seqid(data->o_arg.seqid);
2068 	return state;
2069 }
2070 
2071 static struct nfs4_state *
2072 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2073 {
2074 	struct nfs4_state *ret;
2075 
2076 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2077 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2078 	else
2079 		ret = _nfs4_opendata_to_nfs4_state(data);
2080 	nfs4_sequence_free_slot(&data->o_res.seq_res);
2081 	return ret;
2082 }
2083 
2084 static struct nfs_open_context *
2085 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2086 {
2087 	struct nfs_inode *nfsi = NFS_I(state->inode);
2088 	struct nfs_open_context *ctx;
2089 
2090 	rcu_read_lock();
2091 	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2092 		if (ctx->state != state)
2093 			continue;
2094 		if ((ctx->mode & mode) != mode)
2095 			continue;
2096 		if (!get_nfs_open_context(ctx))
2097 			continue;
2098 		rcu_read_unlock();
2099 		return ctx;
2100 	}
2101 	rcu_read_unlock();
2102 	return ERR_PTR(-ENOENT);
2103 }
2104 
2105 static struct nfs_open_context *
2106 nfs4_state_find_open_context(struct nfs4_state *state)
2107 {
2108 	struct nfs_open_context *ctx;
2109 
2110 	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2111 	if (!IS_ERR(ctx))
2112 		return ctx;
2113 	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2114 	if (!IS_ERR(ctx))
2115 		return ctx;
2116 	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2117 }
2118 
2119 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2120 		struct nfs4_state *state, enum open_claim_type4 claim)
2121 {
2122 	struct nfs4_opendata *opendata;
2123 
2124 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2125 			NULL, claim, GFP_NOFS);
2126 	if (opendata == NULL)
2127 		return ERR_PTR(-ENOMEM);
2128 	opendata->state = state;
2129 	refcount_inc(&state->count);
2130 	return opendata;
2131 }
2132 
2133 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2134 				    fmode_t fmode)
2135 {
2136 	struct nfs4_state *newstate;
2137 	struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2138 	int openflags = opendata->o_arg.open_flags;
2139 	int ret;
2140 
2141 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2142 		return 0;
2143 	opendata->o_arg.fmode = fmode;
2144 	opendata->o_arg.share_access =
2145 		nfs4_map_atomic_open_share(server, fmode, openflags);
2146 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2147 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2148 	nfs4_init_opendata_res(opendata);
2149 	ret = _nfs4_recover_proc_open(opendata);
2150 	if (ret != 0)
2151 		return ret;
2152 	newstate = nfs4_opendata_to_nfs4_state(opendata);
2153 	if (IS_ERR(newstate))
2154 		return PTR_ERR(newstate);
2155 	if (newstate != opendata->state)
2156 		ret = -ESTALE;
2157 	nfs4_close_state(newstate, fmode);
2158 	return ret;
2159 }
2160 
2161 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2162 {
2163 	int ret;
2164 
2165 	/* memory barrier prior to reading state->n_* */
2166 	smp_rmb();
2167 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2168 	if (ret != 0)
2169 		return ret;
2170 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2171 	if (ret != 0)
2172 		return ret;
2173 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2174 	if (ret != 0)
2175 		return ret;
2176 	/*
2177 	 * We may have performed cached opens for all three recoveries.
2178 	 * Check if we need to update the current stateid.
2179 	 */
2180 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2181 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2182 		write_seqlock(&state->seqlock);
2183 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2184 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2185 		write_sequnlock(&state->seqlock);
2186 	}
2187 	return 0;
2188 }
2189 
2190 /*
2191  * OPEN_RECLAIM:
2192  * 	reclaim state on the server after a reboot.
2193  */
2194 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2195 {
2196 	struct nfs_delegation *delegation;
2197 	struct nfs4_opendata *opendata;
2198 	fmode_t delegation_type = 0;
2199 	int status;
2200 
2201 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2202 			NFS4_OPEN_CLAIM_PREVIOUS);
2203 	if (IS_ERR(opendata))
2204 		return PTR_ERR(opendata);
2205 	rcu_read_lock();
2206 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2207 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2208 		delegation_type = delegation->type;
2209 	rcu_read_unlock();
2210 	opendata->o_arg.u.delegation_type = delegation_type;
2211 	status = nfs4_open_recover(opendata, state);
2212 	nfs4_opendata_put(opendata);
2213 	return status;
2214 }
2215 
2216 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2217 {
2218 	struct nfs_server *server = NFS_SERVER(state->inode);
2219 	struct nfs4_exception exception = { };
2220 	int err;
2221 	do {
2222 		err = _nfs4_do_open_reclaim(ctx, state);
2223 		trace_nfs4_open_reclaim(ctx, 0, err);
2224 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2225 			continue;
2226 		if (err != -NFS4ERR_DELAY)
2227 			break;
2228 		nfs4_handle_exception(server, err, &exception);
2229 	} while (exception.retry);
2230 	return err;
2231 }
2232 
2233 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2234 {
2235 	struct nfs_open_context *ctx;
2236 	int ret;
2237 
2238 	ctx = nfs4_state_find_open_context(state);
2239 	if (IS_ERR(ctx))
2240 		return -EAGAIN;
2241 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2242 	nfs_state_clear_open_state_flags(state);
2243 	ret = nfs4_do_open_reclaim(ctx, state);
2244 	put_nfs_open_context(ctx);
2245 	return ret;
2246 }
2247 
2248 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)
2249 {
2250 	switch (err) {
2251 		default:
2252 			printk(KERN_ERR "NFS: %s: unhandled error "
2253 					"%d.\n", __func__, err);
2254 			fallthrough;
2255 		case 0:
2256 		case -ENOENT:
2257 		case -EAGAIN:
2258 		case -ESTALE:
2259 		case -ETIMEDOUT:
2260 			break;
2261 		case -NFS4ERR_BADSESSION:
2262 		case -NFS4ERR_BADSLOT:
2263 		case -NFS4ERR_BAD_HIGH_SLOT:
2264 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2265 		case -NFS4ERR_DEADSESSION:
2266 			return -EAGAIN;
2267 		case -NFS4ERR_STALE_CLIENTID:
2268 		case -NFS4ERR_STALE_STATEID:
2269 			/* Don't recall a delegation if it was lost */
2270 			nfs4_schedule_lease_recovery(server->nfs_client);
2271 			return -EAGAIN;
2272 		case -NFS4ERR_MOVED:
2273 			nfs4_schedule_migration_recovery(server);
2274 			return -EAGAIN;
2275 		case -NFS4ERR_LEASE_MOVED:
2276 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2277 			return -EAGAIN;
2278 		case -NFS4ERR_DELEG_REVOKED:
2279 		case -NFS4ERR_ADMIN_REVOKED:
2280 		case -NFS4ERR_EXPIRED:
2281 		case -NFS4ERR_BAD_STATEID:
2282 		case -NFS4ERR_OPENMODE:
2283 			nfs_inode_find_state_and_recover(state->inode,
2284 					stateid);
2285 			nfs4_schedule_stateid_recovery(server, state);
2286 			return -EAGAIN;
2287 		case -NFS4ERR_DELAY:
2288 		case -NFS4ERR_GRACE:
2289 			ssleep(1);
2290 			return -EAGAIN;
2291 		case -ENOMEM:
2292 		case -NFS4ERR_DENIED:
2293 			if (fl) {
2294 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2295 				if (lsp)
2296 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2297 			}
2298 			return 0;
2299 	}
2300 	return err;
2301 }
2302 
2303 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2304 		struct nfs4_state *state, const nfs4_stateid *stateid)
2305 {
2306 	struct nfs_server *server = NFS_SERVER(state->inode);
2307 	struct nfs4_opendata *opendata;
2308 	int err = 0;
2309 
2310 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2311 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2312 	if (IS_ERR(opendata))
2313 		return PTR_ERR(opendata);
2314 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2315 	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2316 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2317 		if (err)
2318 			goto out;
2319 	}
2320 	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2321 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2322 		if (err)
2323 			goto out;
2324 	}
2325 	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2326 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2327 		if (err)
2328 			goto out;
2329 	}
2330 	nfs_state_clear_delegation(state);
2331 out:
2332 	nfs4_opendata_put(opendata);
2333 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2334 }
2335 
2336 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2337 {
2338 	struct nfs4_opendata *data = calldata;
2339 
2340 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2341 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2342 }
2343 
2344 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2345 {
2346 	struct nfs4_opendata *data = calldata;
2347 
2348 	nfs40_sequence_done(task, &data->c_res.seq_res);
2349 
2350 	data->rpc_status = task->tk_status;
2351 	if (data->rpc_status == 0) {
2352 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2353 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2354 		renew_lease(data->o_res.server, data->timestamp);
2355 		data->rpc_done = true;
2356 	}
2357 }
2358 
2359 static void nfs4_open_confirm_release(void *calldata)
2360 {
2361 	struct nfs4_opendata *data = calldata;
2362 	struct nfs4_state *state = NULL;
2363 
2364 	/* If this request hasn't been cancelled, do nothing */
2365 	if (!data->cancelled)
2366 		goto out_free;
2367 	/* In case of error, no cleanup! */
2368 	if (!data->rpc_done)
2369 		goto out_free;
2370 	state = nfs4_opendata_to_nfs4_state(data);
2371 	if (!IS_ERR(state))
2372 		nfs4_close_state(state, data->o_arg.fmode);
2373 out_free:
2374 	nfs4_opendata_put(data);
2375 }
2376 
2377 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2378 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2379 	.rpc_call_done = nfs4_open_confirm_done,
2380 	.rpc_release = nfs4_open_confirm_release,
2381 };
2382 
2383 /*
2384  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2385  */
2386 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2387 {
2388 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2389 	struct rpc_task *task;
2390 	struct  rpc_message msg = {
2391 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2392 		.rpc_argp = &data->c_arg,
2393 		.rpc_resp = &data->c_res,
2394 		.rpc_cred = data->owner->so_cred,
2395 	};
2396 	struct rpc_task_setup task_setup_data = {
2397 		.rpc_client = server->client,
2398 		.rpc_message = &msg,
2399 		.callback_ops = &nfs4_open_confirm_ops,
2400 		.callback_data = data,
2401 		.workqueue = nfsiod_workqueue,
2402 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2403 	};
2404 	int status;
2405 
2406 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2407 				data->is_recover);
2408 	kref_get(&data->kref);
2409 	data->rpc_done = false;
2410 	data->rpc_status = 0;
2411 	data->timestamp = jiffies;
2412 	task = rpc_run_task(&task_setup_data);
2413 	if (IS_ERR(task))
2414 		return PTR_ERR(task);
2415 	status = rpc_wait_for_completion_task(task);
2416 	if (status != 0) {
2417 		data->cancelled = true;
2418 		smp_wmb();
2419 	} else
2420 		status = data->rpc_status;
2421 	rpc_put_task(task);
2422 	return status;
2423 }
2424 
2425 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2426 {
2427 	struct nfs4_opendata *data = calldata;
2428 	struct nfs4_state_owner *sp = data->owner;
2429 	struct nfs_client *clp = sp->so_server->nfs_client;
2430 	enum open_claim_type4 claim = data->o_arg.claim;
2431 
2432 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2433 		goto out_wait;
2434 	/*
2435 	 * Check if we still need to send an OPEN call, or if we can use
2436 	 * a delegation instead.
2437 	 */
2438 	if (data->state != NULL) {
2439 		struct nfs_delegation *delegation;
2440 
2441 		if (can_open_cached(data->state, data->o_arg.fmode,
2442 					data->o_arg.open_flags, claim))
2443 			goto out_no_action;
2444 		rcu_read_lock();
2445 		delegation = nfs4_get_valid_delegation(data->state->inode);
2446 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2447 			goto unlock_no_action;
2448 		rcu_read_unlock();
2449 	}
2450 	/* Update client id. */
2451 	data->o_arg.clientid = clp->cl_clientid;
2452 	switch (claim) {
2453 	default:
2454 		break;
2455 	case NFS4_OPEN_CLAIM_PREVIOUS:
2456 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2457 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2458 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2459 		fallthrough;
2460 	case NFS4_OPEN_CLAIM_FH:
2461 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2462 	}
2463 	data->timestamp = jiffies;
2464 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2465 				&data->o_arg.seq_args,
2466 				&data->o_res.seq_res,
2467 				task) != 0)
2468 		nfs_release_seqid(data->o_arg.seqid);
2469 
2470 	/* Set the create mode (note dependency on the session type) */
2471 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2472 	if (data->o_arg.open_flags & O_EXCL) {
2473 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2474 		if (clp->cl_mvops->minor_version == 0) {
2475 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2476 			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2477 			 * because ACCESS will return permission denied for
2478 			 * all bits until close */
2479 			data->o_res.access_request = data->o_arg.access = 0;
2480 		} else if (nfs4_has_persistent_session(clp))
2481 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2482 	}
2483 	return;
2484 unlock_no_action:
2485 	trace_nfs4_cached_open(data->state);
2486 	rcu_read_unlock();
2487 out_no_action:
2488 	task->tk_action = NULL;
2489 out_wait:
2490 	nfs4_sequence_done(task, &data->o_res.seq_res);
2491 }
2492 
2493 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2494 {
2495 	struct nfs4_opendata *data = calldata;
2496 
2497 	data->rpc_status = task->tk_status;
2498 
2499 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2500 		return;
2501 
2502 	if (task->tk_status == 0) {
2503 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2504 			switch (data->o_res.f_attr->mode & S_IFMT) {
2505 			case S_IFREG:
2506 				break;
2507 			case S_IFLNK:
2508 				data->rpc_status = -ELOOP;
2509 				break;
2510 			case S_IFDIR:
2511 				data->rpc_status = -EISDIR;
2512 				break;
2513 			default:
2514 				data->rpc_status = -ENOTDIR;
2515 			}
2516 		}
2517 		renew_lease(data->o_res.server, data->timestamp);
2518 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2519 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2520 	}
2521 	data->rpc_done = true;
2522 }
2523 
2524 static void nfs4_open_release(void *calldata)
2525 {
2526 	struct nfs4_opendata *data = calldata;
2527 	struct nfs4_state *state = NULL;
2528 
2529 	/* If this request hasn't been cancelled, do nothing */
2530 	if (!data->cancelled)
2531 		goto out_free;
2532 	/* In case of error, no cleanup! */
2533 	if (data->rpc_status != 0 || !data->rpc_done)
2534 		goto out_free;
2535 	/* In case we need an open_confirm, no cleanup! */
2536 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2537 		goto out_free;
2538 	state = nfs4_opendata_to_nfs4_state(data);
2539 	if (!IS_ERR(state))
2540 		nfs4_close_state(state, data->o_arg.fmode);
2541 out_free:
2542 	nfs4_opendata_put(data);
2543 }
2544 
2545 static const struct rpc_call_ops nfs4_open_ops = {
2546 	.rpc_call_prepare = nfs4_open_prepare,
2547 	.rpc_call_done = nfs4_open_done,
2548 	.rpc_release = nfs4_open_release,
2549 };
2550 
2551 static int nfs4_run_open_task(struct nfs4_opendata *data,
2552 			      struct nfs_open_context *ctx)
2553 {
2554 	struct inode *dir = d_inode(data->dir);
2555 	struct nfs_server *server = NFS_SERVER(dir);
2556 	struct nfs_openargs *o_arg = &data->o_arg;
2557 	struct nfs_openres *o_res = &data->o_res;
2558 	struct rpc_task *task;
2559 	struct rpc_message msg = {
2560 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2561 		.rpc_argp = o_arg,
2562 		.rpc_resp = o_res,
2563 		.rpc_cred = data->owner->so_cred,
2564 	};
2565 	struct rpc_task_setup task_setup_data = {
2566 		.rpc_client = server->client,
2567 		.rpc_message = &msg,
2568 		.callback_ops = &nfs4_open_ops,
2569 		.callback_data = data,
2570 		.workqueue = nfsiod_workqueue,
2571 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2572 	};
2573 	int status;
2574 
2575 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2576 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2577 
2578 	kref_get(&data->kref);
2579 	data->rpc_done = false;
2580 	data->rpc_status = 0;
2581 	data->cancelled = false;
2582 	data->is_recover = false;
2583 	if (!ctx) {
2584 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2585 		data->is_recover = true;
2586 		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2587 	} else {
2588 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2589 		pnfs_lgopen_prepare(data, ctx);
2590 	}
2591 	task = rpc_run_task(&task_setup_data);
2592 	if (IS_ERR(task))
2593 		return PTR_ERR(task);
2594 	status = rpc_wait_for_completion_task(task);
2595 	if (status != 0) {
2596 		data->cancelled = true;
2597 		smp_wmb();
2598 	} else
2599 		status = data->rpc_status;
2600 	rpc_put_task(task);
2601 
2602 	return status;
2603 }
2604 
2605 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2606 {
2607 	struct inode *dir = d_inode(data->dir);
2608 	struct nfs_openres *o_res = &data->o_res;
2609 	int status;
2610 
2611 	status = nfs4_run_open_task(data, NULL);
2612 	if (status != 0 || !data->rpc_done)
2613 		return status;
2614 
2615 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2616 
2617 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2618 		status = _nfs4_proc_open_confirm(data);
2619 
2620 	return status;
2621 }
2622 
2623 /*
2624  * Additional permission checks in order to distinguish between an
2625  * open for read, and an open for execute. This works around the
2626  * fact that NFSv4 OPEN treats read and execute permissions as being
2627  * the same.
2628  * Note that in the non-execute case, we want to turn off permission
2629  * checking if we just created a new file (POSIX open() semantics).
2630  */
2631 static int nfs4_opendata_access(const struct cred *cred,
2632 				struct nfs4_opendata *opendata,
2633 				struct nfs4_state *state, fmode_t fmode)
2634 {
2635 	struct nfs_access_entry cache;
2636 	u32 mask, flags;
2637 
2638 	/* access call failed or for some reason the server doesn't
2639 	 * support any access modes -- defer access call until later */
2640 	if (opendata->o_res.access_supported == 0)
2641 		return 0;
2642 
2643 	mask = 0;
2644 	if (fmode & FMODE_EXEC) {
2645 		/* ONLY check for exec rights */
2646 		if (S_ISDIR(state->inode->i_mode))
2647 			mask = NFS4_ACCESS_LOOKUP;
2648 		else
2649 			mask = NFS4_ACCESS_EXECUTE;
2650 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2651 		mask = NFS4_ACCESS_READ;
2652 
2653 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2654 	nfs_access_add_cache(state->inode, &cache, cred);
2655 
2656 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2657 	if ((mask & ~cache.mask & flags) == 0)
2658 		return 0;
2659 
2660 	return -EACCES;
2661 }
2662 
2663 /*
2664  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2665  */
2666 static int _nfs4_proc_open(struct nfs4_opendata *data,
2667 			   struct nfs_open_context *ctx)
2668 {
2669 	struct inode *dir = d_inode(data->dir);
2670 	struct nfs_server *server = NFS_SERVER(dir);
2671 	struct nfs_openargs *o_arg = &data->o_arg;
2672 	struct nfs_openres *o_res = &data->o_res;
2673 	int status;
2674 
2675 	status = nfs4_run_open_task(data, ctx);
2676 	if (!data->rpc_done)
2677 		return status;
2678 	if (status != 0) {
2679 		if (status == -NFS4ERR_BADNAME &&
2680 				!(o_arg->open_flags & O_CREAT))
2681 			return -ENOENT;
2682 		return status;
2683 	}
2684 
2685 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2686 
2687 	if (o_arg->open_flags & O_CREAT) {
2688 		if (o_arg->open_flags & O_EXCL)
2689 			data->file_created = true;
2690 		else if (o_res->cinfo.before != o_res->cinfo.after)
2691 			data->file_created = true;
2692 		if (data->file_created ||
2693 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2694 			nfs4_update_changeattr(dir, &o_res->cinfo,
2695 					o_res->f_attr->time_start,
2696 					NFS_INO_INVALID_DATA);
2697 	}
2698 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2699 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2700 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2701 		status = _nfs4_proc_open_confirm(data);
2702 		if (status != 0)
2703 			return status;
2704 	}
2705 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2706 		struct nfs_fh *fh = &o_res->fh;
2707 
2708 		nfs4_sequence_free_slot(&o_res->seq_res);
2709 		if (o_arg->claim == NFS4_OPEN_CLAIM_FH)
2710 			fh = NFS_FH(d_inode(data->dentry));
2711 		nfs4_proc_getattr(server, fh, o_res->f_attr, NULL);
2712 	}
2713 	return 0;
2714 }
2715 
2716 /*
2717  * OPEN_EXPIRED:
2718  * 	reclaim state on the server after a network partition.
2719  * 	Assumes caller holds the appropriate lock
2720  */
2721 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2722 {
2723 	struct nfs4_opendata *opendata;
2724 	int ret;
2725 
2726 	opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2727 	if (IS_ERR(opendata))
2728 		return PTR_ERR(opendata);
2729 	/*
2730 	 * We're not recovering a delegation, so ask for no delegation.
2731 	 * Otherwise the recovery thread could deadlock with an outstanding
2732 	 * delegation return.
2733 	 */
2734 	opendata->o_arg.open_flags = O_DIRECT;
2735 	ret = nfs4_open_recover(opendata, state);
2736 	if (ret == -ESTALE)
2737 		d_drop(ctx->dentry);
2738 	nfs4_opendata_put(opendata);
2739 	return ret;
2740 }
2741 
2742 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2743 {
2744 	struct nfs_server *server = NFS_SERVER(state->inode);
2745 	struct nfs4_exception exception = { };
2746 	int err;
2747 
2748 	do {
2749 		err = _nfs4_open_expired(ctx, state);
2750 		trace_nfs4_open_expired(ctx, 0, err);
2751 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2752 			continue;
2753 		switch (err) {
2754 		default:
2755 			goto out;
2756 		case -NFS4ERR_GRACE:
2757 		case -NFS4ERR_DELAY:
2758 			nfs4_handle_exception(server, err, &exception);
2759 			err = 0;
2760 		}
2761 	} while (exception.retry);
2762 out:
2763 	return err;
2764 }
2765 
2766 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2767 {
2768 	struct nfs_open_context *ctx;
2769 	int ret;
2770 
2771 	ctx = nfs4_state_find_open_context(state);
2772 	if (IS_ERR(ctx))
2773 		return -EAGAIN;
2774 	ret = nfs4_do_open_expired(ctx, state);
2775 	put_nfs_open_context(ctx);
2776 	return ret;
2777 }
2778 
2779 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2780 		const nfs4_stateid *stateid)
2781 {
2782 	nfs_remove_bad_delegation(state->inode, stateid);
2783 	nfs_state_clear_delegation(state);
2784 }
2785 
2786 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2787 {
2788 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2789 		nfs_finish_clear_delegation_stateid(state, NULL);
2790 }
2791 
2792 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2793 {
2794 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2795 	nfs40_clear_delegation_stateid(state);
2796 	nfs_state_clear_open_state_flags(state);
2797 	return nfs4_open_expired(sp, state);
2798 }
2799 
2800 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2801 		nfs4_stateid *stateid,
2802 		const struct cred *cred)
2803 {
2804 	return -NFS4ERR_BAD_STATEID;
2805 }
2806 
2807 #if defined(CONFIG_NFS_V4_1)
2808 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2809 		nfs4_stateid *stateid,
2810 		const struct cred *cred)
2811 {
2812 	int status;
2813 
2814 	switch (stateid->type) {
2815 	default:
2816 		break;
2817 	case NFS4_INVALID_STATEID_TYPE:
2818 	case NFS4_SPECIAL_STATEID_TYPE:
2819 		return -NFS4ERR_BAD_STATEID;
2820 	case NFS4_REVOKED_STATEID_TYPE:
2821 		goto out_free;
2822 	}
2823 
2824 	status = nfs41_test_stateid(server, stateid, cred);
2825 	switch (status) {
2826 	case -NFS4ERR_EXPIRED:
2827 	case -NFS4ERR_ADMIN_REVOKED:
2828 	case -NFS4ERR_DELEG_REVOKED:
2829 		break;
2830 	default:
2831 		return status;
2832 	}
2833 out_free:
2834 	/* Ack the revoked state to the server */
2835 	nfs41_free_stateid(server, stateid, cred, true);
2836 	return -NFS4ERR_EXPIRED;
2837 }
2838 
2839 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2840 {
2841 	struct nfs_server *server = NFS_SERVER(state->inode);
2842 	nfs4_stateid stateid;
2843 	struct nfs_delegation *delegation;
2844 	const struct cred *cred = NULL;
2845 	int status, ret = NFS_OK;
2846 
2847 	/* Get the delegation credential for use by test/free_stateid */
2848 	rcu_read_lock();
2849 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2850 	if (delegation == NULL) {
2851 		rcu_read_unlock();
2852 		nfs_state_clear_delegation(state);
2853 		return NFS_OK;
2854 	}
2855 
2856 	spin_lock(&delegation->lock);
2857 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2858 
2859 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2860 				&delegation->flags)) {
2861 		spin_unlock(&delegation->lock);
2862 		rcu_read_unlock();
2863 		return NFS_OK;
2864 	}
2865 
2866 	if (delegation->cred)
2867 		cred = get_cred(delegation->cred);
2868 	spin_unlock(&delegation->lock);
2869 	rcu_read_unlock();
2870 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2871 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2872 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2873 		nfs_finish_clear_delegation_stateid(state, &stateid);
2874 	else
2875 		ret = status;
2876 
2877 	put_cred(cred);
2878 	return ret;
2879 }
2880 
2881 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2882 {
2883 	nfs4_stateid tmp;
2884 
2885 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2886 	    nfs4_copy_delegation_stateid(state->inode, state->state,
2887 				&tmp, NULL) &&
2888 	    nfs4_stateid_match_other(&state->stateid, &tmp))
2889 		nfs_state_set_delegation(state, &tmp, state->state);
2890 	else
2891 		nfs_state_clear_delegation(state);
2892 }
2893 
2894 /**
2895  * nfs41_check_expired_locks - possibly free a lock stateid
2896  *
2897  * @state: NFSv4 state for an inode
2898  *
2899  * Returns NFS_OK if recovery for this stateid is now finished.
2900  * Otherwise a negative NFS4ERR value is returned.
2901  */
2902 static int nfs41_check_expired_locks(struct nfs4_state *state)
2903 {
2904 	int status, ret = NFS_OK;
2905 	struct nfs4_lock_state *lsp, *prev = NULL;
2906 	struct nfs_server *server = NFS_SERVER(state->inode);
2907 
2908 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2909 		goto out;
2910 
2911 	spin_lock(&state->state_lock);
2912 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2913 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2914 			const struct cred *cred = lsp->ls_state->owner->so_cred;
2915 
2916 			refcount_inc(&lsp->ls_count);
2917 			spin_unlock(&state->state_lock);
2918 
2919 			nfs4_put_lock_state(prev);
2920 			prev = lsp;
2921 
2922 			status = nfs41_test_and_free_expired_stateid(server,
2923 					&lsp->ls_stateid,
2924 					cred);
2925 			trace_nfs4_test_lock_stateid(state, lsp, status);
2926 			if (status == -NFS4ERR_EXPIRED ||
2927 			    status == -NFS4ERR_BAD_STATEID) {
2928 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2929 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2930 				if (!recover_lost_locks)
2931 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2932 			} else if (status != NFS_OK) {
2933 				ret = status;
2934 				nfs4_put_lock_state(prev);
2935 				goto out;
2936 			}
2937 			spin_lock(&state->state_lock);
2938 		}
2939 	}
2940 	spin_unlock(&state->state_lock);
2941 	nfs4_put_lock_state(prev);
2942 out:
2943 	return ret;
2944 }
2945 
2946 /**
2947  * nfs41_check_open_stateid - possibly free an open stateid
2948  *
2949  * @state: NFSv4 state for an inode
2950  *
2951  * Returns NFS_OK if recovery for this stateid is now finished.
2952  * Otherwise a negative NFS4ERR value is returned.
2953  */
2954 static int nfs41_check_open_stateid(struct nfs4_state *state)
2955 {
2956 	struct nfs_server *server = NFS_SERVER(state->inode);
2957 	nfs4_stateid *stateid = &state->open_stateid;
2958 	const struct cred *cred = state->owner->so_cred;
2959 	int status;
2960 
2961 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2962 		return -NFS4ERR_BAD_STATEID;
2963 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2964 	trace_nfs4_test_open_stateid(state, NULL, status);
2965 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2966 		nfs_state_clear_open_state_flags(state);
2967 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2968 		return status;
2969 	}
2970 	if (nfs_open_stateid_recover_openmode(state))
2971 		return -NFS4ERR_OPENMODE;
2972 	return NFS_OK;
2973 }
2974 
2975 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2976 {
2977 	int status;
2978 
2979 	status = nfs41_check_delegation_stateid(state);
2980 	if (status != NFS_OK)
2981 		return status;
2982 	nfs41_delegation_recover_stateid(state);
2983 
2984 	status = nfs41_check_expired_locks(state);
2985 	if (status != NFS_OK)
2986 		return status;
2987 	status = nfs41_check_open_stateid(state);
2988 	if (status != NFS_OK)
2989 		status = nfs4_open_expired(sp, state);
2990 	return status;
2991 }
2992 #endif
2993 
2994 /*
2995  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2996  * fields corresponding to attributes that were used to store the verifier.
2997  * Make sure we clobber those fields in the later setattr call
2998  */
2999 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
3000 				struct iattr *sattr, struct nfs4_label **label)
3001 {
3002 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
3003 	__u32 attrset[3];
3004 	unsigned ret;
3005 	unsigned i;
3006 
3007 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3008 		attrset[i] = opendata->o_res.attrset[i];
3009 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3010 			attrset[i] &= ~bitmask[i];
3011 	}
3012 
3013 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3014 		sattr->ia_valid : 0;
3015 
3016 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3017 		if (sattr->ia_valid & ATTR_ATIME_SET)
3018 			ret |= ATTR_ATIME_SET;
3019 		else
3020 			ret |= ATTR_ATIME;
3021 	}
3022 
3023 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3024 		if (sattr->ia_valid & ATTR_MTIME_SET)
3025 			ret |= ATTR_MTIME_SET;
3026 		else
3027 			ret |= ATTR_MTIME;
3028 	}
3029 
3030 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3031 		*label = NULL;
3032 	return ret;
3033 }
3034 
3035 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3036 		struct nfs_open_context *ctx)
3037 {
3038 	struct nfs4_state_owner *sp = opendata->owner;
3039 	struct nfs_server *server = sp->so_server;
3040 	struct dentry *dentry;
3041 	struct nfs4_state *state;
3042 	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3043 	struct inode *dir = d_inode(opendata->dir);
3044 	unsigned long dir_verifier;
3045 	unsigned int seq;
3046 	int ret;
3047 
3048 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3049 	dir_verifier = nfs_save_change_attribute(dir);
3050 
3051 	ret = _nfs4_proc_open(opendata, ctx);
3052 	if (ret != 0)
3053 		goto out;
3054 
3055 	state = _nfs4_opendata_to_nfs4_state(opendata);
3056 	ret = PTR_ERR(state);
3057 	if (IS_ERR(state))
3058 		goto out;
3059 	ctx->state = state;
3060 	if (server->caps & NFS_CAP_POSIX_LOCK)
3061 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3062 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3063 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3064 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3065 		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3066 
3067 	dentry = opendata->dentry;
3068 	if (d_really_is_negative(dentry)) {
3069 		struct dentry *alias;
3070 		d_drop(dentry);
3071 		alias = d_exact_alias(dentry, state->inode);
3072 		if (!alias)
3073 			alias = d_splice_alias(igrab(state->inode), dentry);
3074 		/* d_splice_alias() can't fail here - it's a non-directory */
3075 		if (alias) {
3076 			dput(ctx->dentry);
3077 			ctx->dentry = dentry = alias;
3078 		}
3079 	}
3080 
3081 	switch(opendata->o_arg.claim) {
3082 	default:
3083 		break;
3084 	case NFS4_OPEN_CLAIM_NULL:
3085 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3086 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3087 		if (!opendata->rpc_done)
3088 			break;
3089 		if (opendata->o_res.delegation_type != 0)
3090 			dir_verifier = nfs_save_change_attribute(dir);
3091 		nfs_set_verifier(dentry, dir_verifier);
3092 	}
3093 
3094 	/* Parse layoutget results before we check for access */
3095 	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3096 
3097 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3098 	if (ret != 0)
3099 		goto out;
3100 
3101 	if (d_inode(dentry) == state->inode) {
3102 		nfs_inode_attach_open_context(ctx);
3103 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3104 			nfs4_schedule_stateid_recovery(server, state);
3105 	}
3106 
3107 out:
3108 	if (!opendata->cancelled) {
3109 		if (opendata->lgp) {
3110 			nfs4_lgopen_release(opendata->lgp);
3111 			opendata->lgp = NULL;
3112 		}
3113 		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3114 	}
3115 	return ret;
3116 }
3117 
3118 /*
3119  * Returns a referenced nfs4_state
3120  */
3121 static int _nfs4_do_open(struct inode *dir,
3122 			struct nfs_open_context *ctx,
3123 			int flags,
3124 			const struct nfs4_open_createattrs *c,
3125 			int *opened)
3126 {
3127 	struct nfs4_state_owner  *sp;
3128 	struct nfs4_state     *state = NULL;
3129 	struct nfs_server       *server = NFS_SERVER(dir);
3130 	struct nfs4_opendata *opendata;
3131 	struct dentry *dentry = ctx->dentry;
3132 	const struct cred *cred = ctx->cred;
3133 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3134 	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3135 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3136 	struct iattr *sattr = c->sattr;
3137 	struct nfs4_label *label = c->label;
3138 	int status;
3139 
3140 	/* Protect against reboot recovery conflicts */
3141 	status = -ENOMEM;
3142 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3143 	if (sp == NULL) {
3144 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3145 		goto out_err;
3146 	}
3147 	status = nfs4_client_recover_expired_lease(server->nfs_client);
3148 	if (status != 0)
3149 		goto err_put_state_owner;
3150 	if (d_really_is_positive(dentry))
3151 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3152 	status = -ENOMEM;
3153 	if (d_really_is_positive(dentry))
3154 		claim = NFS4_OPEN_CLAIM_FH;
3155 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3156 			c, claim, GFP_KERNEL);
3157 	if (opendata == NULL)
3158 		goto err_put_state_owner;
3159 
3160 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3161 		if (!opendata->f_attr.mdsthreshold) {
3162 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3163 			if (!opendata->f_attr.mdsthreshold)
3164 				goto err_opendata_put;
3165 		}
3166 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3167 	}
3168 	if (d_really_is_positive(dentry))
3169 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3170 
3171 	status = _nfs4_open_and_get_state(opendata, ctx);
3172 	if (status != 0)
3173 		goto err_opendata_put;
3174 	state = ctx->state;
3175 
3176 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3177 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3178 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3179 		/*
3180 		 * send create attributes which was not set by open
3181 		 * with an extra setattr.
3182 		 */
3183 		if (attrs || label) {
3184 			unsigned ia_old = sattr->ia_valid;
3185 
3186 			sattr->ia_valid = attrs;
3187 			nfs_fattr_init(opendata->o_res.f_attr);
3188 			status = nfs4_do_setattr(state->inode, cred,
3189 					opendata->o_res.f_attr, sattr,
3190 					ctx, label);
3191 			if (status == 0) {
3192 				nfs_setattr_update_inode(state->inode, sattr,
3193 						opendata->o_res.f_attr);
3194 				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3195 			}
3196 			sattr->ia_valid = ia_old;
3197 		}
3198 	}
3199 	if (opened && opendata->file_created)
3200 		*opened = 1;
3201 
3202 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3203 		*ctx_th = opendata->f_attr.mdsthreshold;
3204 		opendata->f_attr.mdsthreshold = NULL;
3205 	}
3206 
3207 	nfs4_opendata_put(opendata);
3208 	nfs4_put_state_owner(sp);
3209 	return 0;
3210 err_opendata_put:
3211 	nfs4_opendata_put(opendata);
3212 err_put_state_owner:
3213 	nfs4_put_state_owner(sp);
3214 out_err:
3215 	return status;
3216 }
3217 
3218 
3219 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3220 					struct nfs_open_context *ctx,
3221 					int flags,
3222 					struct iattr *sattr,
3223 					struct nfs4_label *label,
3224 					int *opened)
3225 {
3226 	struct nfs_server *server = NFS_SERVER(dir);
3227 	struct nfs4_exception exception = {
3228 		.interruptible = true,
3229 	};
3230 	struct nfs4_state *res;
3231 	struct nfs4_open_createattrs c = {
3232 		.label = label,
3233 		.sattr = sattr,
3234 		.verf = {
3235 			[0] = (__u32)jiffies,
3236 			[1] = (__u32)current->pid,
3237 		},
3238 	};
3239 	int status;
3240 
3241 	do {
3242 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3243 		res = ctx->state;
3244 		trace_nfs4_open_file(ctx, flags, status);
3245 		if (status == 0)
3246 			break;
3247 		/* NOTE: BAD_SEQID means the server and client disagree about the
3248 		 * book-keeping w.r.t. state-changing operations
3249 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3250 		 * It is actually a sign of a bug on the client or on the server.
3251 		 *
3252 		 * If we receive a BAD_SEQID error in the particular case of
3253 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3254 		 * have unhashed the old state_owner for us, and that we can
3255 		 * therefore safely retry using a new one. We should still warn
3256 		 * the user though...
3257 		 */
3258 		if (status == -NFS4ERR_BAD_SEQID) {
3259 			pr_warn_ratelimited("NFS: v4 server %s "
3260 					" returned a bad sequence-id error!\n",
3261 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3262 			exception.retry = 1;
3263 			continue;
3264 		}
3265 		/*
3266 		 * BAD_STATEID on OPEN means that the server cancelled our
3267 		 * state before it received the OPEN_CONFIRM.
3268 		 * Recover by retrying the request as per the discussion
3269 		 * on Page 181 of RFC3530.
3270 		 */
3271 		if (status == -NFS4ERR_BAD_STATEID) {
3272 			exception.retry = 1;
3273 			continue;
3274 		}
3275 		if (status == -NFS4ERR_EXPIRED) {
3276 			nfs4_schedule_lease_recovery(server->nfs_client);
3277 			exception.retry = 1;
3278 			continue;
3279 		}
3280 		if (status == -EAGAIN) {
3281 			/* We must have found a delegation */
3282 			exception.retry = 1;
3283 			continue;
3284 		}
3285 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3286 			continue;
3287 		res = ERR_PTR(nfs4_handle_exception(server,
3288 					status, &exception));
3289 	} while (exception.retry);
3290 	return res;
3291 }
3292 
3293 static int _nfs4_do_setattr(struct inode *inode,
3294 			    struct nfs_setattrargs *arg,
3295 			    struct nfs_setattrres *res,
3296 			    const struct cred *cred,
3297 			    struct nfs_open_context *ctx)
3298 {
3299 	struct nfs_server *server = NFS_SERVER(inode);
3300 	struct rpc_message msg = {
3301 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3302 		.rpc_argp	= arg,
3303 		.rpc_resp	= res,
3304 		.rpc_cred	= cred,
3305 	};
3306 	const struct cred *delegation_cred = NULL;
3307 	unsigned long timestamp = jiffies;
3308 	bool truncate;
3309 	int status;
3310 
3311 	nfs_fattr_init(res->fattr);
3312 
3313 	/* Servers should only apply open mode checks for file size changes */
3314 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3315 	if (!truncate) {
3316 		nfs4_inode_make_writeable(inode);
3317 		goto zero_stateid;
3318 	}
3319 
3320 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3321 		/* Use that stateid */
3322 	} else if (ctx != NULL && ctx->state) {
3323 		struct nfs_lock_context *l_ctx;
3324 		if (!nfs4_valid_open_stateid(ctx->state))
3325 			return -EBADF;
3326 		l_ctx = nfs_get_lock_context(ctx);
3327 		if (IS_ERR(l_ctx))
3328 			return PTR_ERR(l_ctx);
3329 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3330 						&arg->stateid, &delegation_cred);
3331 		nfs_put_lock_context(l_ctx);
3332 		if (status == -EIO)
3333 			return -EBADF;
3334 		else if (status == -EAGAIN)
3335 			goto zero_stateid;
3336 	} else {
3337 zero_stateid:
3338 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3339 	}
3340 	if (delegation_cred)
3341 		msg.rpc_cred = delegation_cred;
3342 
3343 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3344 
3345 	put_cred(delegation_cred);
3346 	if (status == 0 && ctx != NULL)
3347 		renew_lease(server, timestamp);
3348 	trace_nfs4_setattr(inode, &arg->stateid, status);
3349 	return status;
3350 }
3351 
3352 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3353 			   struct nfs_fattr *fattr, struct iattr *sattr,
3354 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3355 {
3356 	struct nfs_server *server = NFS_SERVER(inode);
3357 	__u32 bitmask[NFS4_BITMASK_SZ];
3358 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3359 	struct nfs_setattrargs	arg = {
3360 		.fh		= NFS_FH(inode),
3361 		.iap		= sattr,
3362 		.server		= server,
3363 		.bitmask = bitmask,
3364 		.label		= ilabel,
3365 	};
3366 	struct nfs_setattrres  res = {
3367 		.fattr		= fattr,
3368 		.server		= server,
3369 	};
3370 	struct nfs4_exception exception = {
3371 		.state = state,
3372 		.inode = inode,
3373 		.stateid = &arg.stateid,
3374 	};
3375 	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3376 	int err;
3377 
3378 	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3379 		adjust_flags |= NFS_INO_INVALID_MODE;
3380 	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3381 		adjust_flags |= NFS_INO_INVALID_OTHER;
3382 
3383 	do {
3384 		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3385 					inode, adjust_flags);
3386 
3387 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3388 		switch (err) {
3389 		case -NFS4ERR_OPENMODE:
3390 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3391 				pr_warn_once("NFSv4: server %s is incorrectly "
3392 						"applying open mode checks to "
3393 						"a SETATTR that is not "
3394 						"changing file size.\n",
3395 						server->nfs_client->cl_hostname);
3396 			}
3397 			if (state && !(state->state & FMODE_WRITE)) {
3398 				err = -EBADF;
3399 				if (sattr->ia_valid & ATTR_OPEN)
3400 					err = -EACCES;
3401 				goto out;
3402 			}
3403 		}
3404 		err = nfs4_handle_exception(server, err, &exception);
3405 	} while (exception.retry);
3406 out:
3407 	return err;
3408 }
3409 
3410 static bool
3411 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3412 {
3413 	if (inode == NULL || !nfs_have_layout(inode))
3414 		return false;
3415 
3416 	return pnfs_wait_on_layoutreturn(inode, task);
3417 }
3418 
3419 /*
3420  * Update the seqid of an open stateid
3421  */
3422 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3423 		struct nfs4_state *state)
3424 {
3425 	__be32 seqid_open;
3426 	u32 dst_seqid;
3427 	int seq;
3428 
3429 	for (;;) {
3430 		if (!nfs4_valid_open_stateid(state))
3431 			break;
3432 		seq = read_seqbegin(&state->seqlock);
3433 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3434 			nfs4_stateid_copy(dst, &state->open_stateid);
3435 			if (read_seqretry(&state->seqlock, seq))
3436 				continue;
3437 			break;
3438 		}
3439 		seqid_open = state->open_stateid.seqid;
3440 		if (read_seqretry(&state->seqlock, seq))
3441 			continue;
3442 
3443 		dst_seqid = be32_to_cpu(dst->seqid);
3444 		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3445 			dst->seqid = seqid_open;
3446 		break;
3447 	}
3448 }
3449 
3450 /*
3451  * Update the seqid of an open stateid after receiving
3452  * NFS4ERR_OLD_STATEID
3453  */
3454 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3455 		struct nfs4_state *state)
3456 {
3457 	__be32 seqid_open;
3458 	u32 dst_seqid;
3459 	bool ret;
3460 	int seq, status = -EAGAIN;
3461 	DEFINE_WAIT(wait);
3462 
3463 	for (;;) {
3464 		ret = false;
3465 		if (!nfs4_valid_open_stateid(state))
3466 			break;
3467 		seq = read_seqbegin(&state->seqlock);
3468 		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3469 			if (read_seqretry(&state->seqlock, seq))
3470 				continue;
3471 			break;
3472 		}
3473 
3474 		write_seqlock(&state->seqlock);
3475 		seqid_open = state->open_stateid.seqid;
3476 
3477 		dst_seqid = be32_to_cpu(dst->seqid);
3478 
3479 		/* Did another OPEN bump the state's seqid?  try again: */
3480 		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3481 			dst->seqid = seqid_open;
3482 			write_sequnlock(&state->seqlock);
3483 			ret = true;
3484 			break;
3485 		}
3486 
3487 		/* server says we're behind but we haven't seen the update yet */
3488 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3489 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3490 		write_sequnlock(&state->seqlock);
3491 		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3492 
3493 		if (fatal_signal_pending(current))
3494 			status = -EINTR;
3495 		else
3496 			if (schedule_timeout(5*HZ) != 0)
3497 				status = 0;
3498 
3499 		finish_wait(&state->waitq, &wait);
3500 
3501 		if (!status)
3502 			continue;
3503 		if (status == -EINTR)
3504 			break;
3505 
3506 		/* we slept the whole 5 seconds, we must have lost a seqid */
3507 		dst->seqid = cpu_to_be32(dst_seqid + 1);
3508 		ret = true;
3509 		break;
3510 	}
3511 
3512 	return ret;
3513 }
3514 
3515 struct nfs4_closedata {
3516 	struct inode *inode;
3517 	struct nfs4_state *state;
3518 	struct nfs_closeargs arg;
3519 	struct nfs_closeres res;
3520 	struct {
3521 		struct nfs4_layoutreturn_args arg;
3522 		struct nfs4_layoutreturn_res res;
3523 		struct nfs4_xdr_opaque_data ld_private;
3524 		u32 roc_barrier;
3525 		bool roc;
3526 	} lr;
3527 	struct nfs_fattr fattr;
3528 	unsigned long timestamp;
3529 };
3530 
3531 static void nfs4_free_closedata(void *data)
3532 {
3533 	struct nfs4_closedata *calldata = data;
3534 	struct nfs4_state_owner *sp = calldata->state->owner;
3535 	struct super_block *sb = calldata->state->inode->i_sb;
3536 
3537 	if (calldata->lr.roc)
3538 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3539 				calldata->res.lr_ret);
3540 	nfs4_put_open_state(calldata->state);
3541 	nfs_free_seqid(calldata->arg.seqid);
3542 	nfs4_put_state_owner(sp);
3543 	nfs_sb_deactive(sb);
3544 	kfree(calldata);
3545 }
3546 
3547 static void nfs4_close_done(struct rpc_task *task, void *data)
3548 {
3549 	struct nfs4_closedata *calldata = data;
3550 	struct nfs4_state *state = calldata->state;
3551 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3552 	nfs4_stateid *res_stateid = NULL;
3553 	struct nfs4_exception exception = {
3554 		.state = state,
3555 		.inode = calldata->inode,
3556 		.stateid = &calldata->arg.stateid,
3557 	};
3558 
3559 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3560 		return;
3561 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3562 
3563 	/* Handle Layoutreturn errors */
3564 	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3565 			  &calldata->res.lr_ret) == -EAGAIN)
3566 		goto out_restart;
3567 
3568 	/* hmm. we are done with the inode, and in the process of freeing
3569 	 * the state_owner. we keep this around to process errors
3570 	 */
3571 	switch (task->tk_status) {
3572 		case 0:
3573 			res_stateid = &calldata->res.stateid;
3574 			renew_lease(server, calldata->timestamp);
3575 			break;
3576 		case -NFS4ERR_ACCESS:
3577 			if (calldata->arg.bitmask != NULL) {
3578 				calldata->arg.bitmask = NULL;
3579 				calldata->res.fattr = NULL;
3580 				goto out_restart;
3581 
3582 			}
3583 			break;
3584 		case -NFS4ERR_OLD_STATEID:
3585 			/* Did we race with OPEN? */
3586 			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3587 						state))
3588 				goto out_restart;
3589 			goto out_release;
3590 		case -NFS4ERR_ADMIN_REVOKED:
3591 		case -NFS4ERR_STALE_STATEID:
3592 		case -NFS4ERR_EXPIRED:
3593 			nfs4_free_revoked_stateid(server,
3594 					&calldata->arg.stateid,
3595 					task->tk_msg.rpc_cred);
3596 			fallthrough;
3597 		case -NFS4ERR_BAD_STATEID:
3598 			if (calldata->arg.fmode == 0)
3599 				break;
3600 			fallthrough;
3601 		default:
3602 			task->tk_status = nfs4_async_handle_exception(task,
3603 					server, task->tk_status, &exception);
3604 			if (exception.retry)
3605 				goto out_restart;
3606 	}
3607 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3608 			res_stateid, calldata->arg.fmode);
3609 out_release:
3610 	task->tk_status = 0;
3611 	nfs_release_seqid(calldata->arg.seqid);
3612 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3613 	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3614 	return;
3615 out_restart:
3616 	task->tk_status = 0;
3617 	rpc_restart_call_prepare(task);
3618 	goto out_release;
3619 }
3620 
3621 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3622 {
3623 	struct nfs4_closedata *calldata = data;
3624 	struct nfs4_state *state = calldata->state;
3625 	struct inode *inode = calldata->inode;
3626 	struct nfs_server *server = NFS_SERVER(inode);
3627 	struct pnfs_layout_hdr *lo;
3628 	bool is_rdonly, is_wronly, is_rdwr;
3629 	int call_close = 0;
3630 
3631 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3632 		goto out_wait;
3633 
3634 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3635 	spin_lock(&state->owner->so_lock);
3636 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3637 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3638 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3639 	/* Calculate the change in open mode */
3640 	calldata->arg.fmode = 0;
3641 	if (state->n_rdwr == 0) {
3642 		if (state->n_rdonly == 0)
3643 			call_close |= is_rdonly;
3644 		else if (is_rdonly)
3645 			calldata->arg.fmode |= FMODE_READ;
3646 		if (state->n_wronly == 0)
3647 			call_close |= is_wronly;
3648 		else if (is_wronly)
3649 			calldata->arg.fmode |= FMODE_WRITE;
3650 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3651 			call_close |= is_rdwr;
3652 	} else if (is_rdwr)
3653 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3654 
3655 	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3656 	if (!nfs4_valid_open_stateid(state))
3657 		call_close = 0;
3658 	spin_unlock(&state->owner->so_lock);
3659 
3660 	if (!call_close) {
3661 		/* Note: exit _without_ calling nfs4_close_done */
3662 		goto out_no_action;
3663 	}
3664 
3665 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3666 		nfs_release_seqid(calldata->arg.seqid);
3667 		goto out_wait;
3668 	}
3669 
3670 	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3671 	if (lo && !pnfs_layout_is_valid(lo)) {
3672 		calldata->arg.lr_args = NULL;
3673 		calldata->res.lr_res = NULL;
3674 	}
3675 
3676 	if (calldata->arg.fmode == 0)
3677 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3678 
3679 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3680 		/* Close-to-open cache consistency revalidation */
3681 		if (!nfs4_have_delegation(inode, FMODE_READ)) {
3682 			nfs4_bitmask_set(calldata->arg.bitmask_store,
3683 					 server->cache_consistency_bitmask,
3684 					 inode, 0);
3685 			calldata->arg.bitmask = calldata->arg.bitmask_store;
3686 		} else
3687 			calldata->arg.bitmask = NULL;
3688 	}
3689 
3690 	calldata->arg.share_access =
3691 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3692 				calldata->arg.fmode, 0);
3693 
3694 	if (calldata->res.fattr == NULL)
3695 		calldata->arg.bitmask = NULL;
3696 	else if (calldata->arg.bitmask == NULL)
3697 		calldata->res.fattr = NULL;
3698 	calldata->timestamp = jiffies;
3699 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3700 				&calldata->arg.seq_args,
3701 				&calldata->res.seq_res,
3702 				task) != 0)
3703 		nfs_release_seqid(calldata->arg.seqid);
3704 	return;
3705 out_no_action:
3706 	task->tk_action = NULL;
3707 out_wait:
3708 	nfs4_sequence_done(task, &calldata->res.seq_res);
3709 }
3710 
3711 static const struct rpc_call_ops nfs4_close_ops = {
3712 	.rpc_call_prepare = nfs4_close_prepare,
3713 	.rpc_call_done = nfs4_close_done,
3714 	.rpc_release = nfs4_free_closedata,
3715 };
3716 
3717 /*
3718  * It is possible for data to be read/written from a mem-mapped file
3719  * after the sys_close call (which hits the vfs layer as a flush).
3720  * This means that we can't safely call nfsv4 close on a file until
3721  * the inode is cleared. This in turn means that we are not good
3722  * NFSv4 citizens - we do not indicate to the server to update the file's
3723  * share state even when we are done with one of the three share
3724  * stateid's in the inode.
3725  *
3726  * NOTE: Caller must be holding the sp->so_owner semaphore!
3727  */
3728 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3729 {
3730 	struct nfs_server *server = NFS_SERVER(state->inode);
3731 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3732 	struct nfs4_closedata *calldata;
3733 	struct nfs4_state_owner *sp = state->owner;
3734 	struct rpc_task *task;
3735 	struct rpc_message msg = {
3736 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3737 		.rpc_cred = state->owner->so_cred,
3738 	};
3739 	struct rpc_task_setup task_setup_data = {
3740 		.rpc_client = server->client,
3741 		.rpc_message = &msg,
3742 		.callback_ops = &nfs4_close_ops,
3743 		.workqueue = nfsiod_workqueue,
3744 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3745 	};
3746 	int status = -ENOMEM;
3747 
3748 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3749 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3750 
3751 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3752 		&task_setup_data.rpc_client, &msg);
3753 
3754 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3755 	if (calldata == NULL)
3756 		goto out;
3757 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3758 	calldata->inode = state->inode;
3759 	calldata->state = state;
3760 	calldata->arg.fh = NFS_FH(state->inode);
3761 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3762 		goto out_free_calldata;
3763 	/* Serialization for the sequence id */
3764 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3765 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3766 	if (IS_ERR(calldata->arg.seqid))
3767 		goto out_free_calldata;
3768 	nfs_fattr_init(&calldata->fattr);
3769 	calldata->arg.fmode = 0;
3770 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3771 	calldata->res.fattr = &calldata->fattr;
3772 	calldata->res.seqid = calldata->arg.seqid;
3773 	calldata->res.server = server;
3774 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3775 	calldata->lr.roc = pnfs_roc(state->inode,
3776 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3777 	if (calldata->lr.roc) {
3778 		calldata->arg.lr_args = &calldata->lr.arg;
3779 		calldata->res.lr_res = &calldata->lr.res;
3780 	}
3781 	nfs_sb_active(calldata->inode->i_sb);
3782 
3783 	msg.rpc_argp = &calldata->arg;
3784 	msg.rpc_resp = &calldata->res;
3785 	task_setup_data.callback_data = calldata;
3786 	task = rpc_run_task(&task_setup_data);
3787 	if (IS_ERR(task))
3788 		return PTR_ERR(task);
3789 	status = 0;
3790 	if (wait)
3791 		status = rpc_wait_for_completion_task(task);
3792 	rpc_put_task(task);
3793 	return status;
3794 out_free_calldata:
3795 	kfree(calldata);
3796 out:
3797 	nfs4_put_open_state(state);
3798 	nfs4_put_state_owner(sp);
3799 	return status;
3800 }
3801 
3802 static struct inode *
3803 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3804 		int open_flags, struct iattr *attr, int *opened)
3805 {
3806 	struct nfs4_state *state;
3807 	struct nfs4_label l, *label;
3808 
3809 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3810 
3811 	/* Protect against concurrent sillydeletes */
3812 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3813 
3814 	nfs4_label_release_security(label);
3815 
3816 	if (IS_ERR(state))
3817 		return ERR_CAST(state);
3818 	return state->inode;
3819 }
3820 
3821 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3822 {
3823 	if (ctx->state == NULL)
3824 		return;
3825 	if (is_sync)
3826 		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3827 	else
3828 		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3829 }
3830 
3831 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3832 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3833 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3834 
3835 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3836 {
3837 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3838 	struct nfs4_server_caps_arg args = {
3839 		.fhandle = fhandle,
3840 		.bitmask = bitmask,
3841 	};
3842 	struct nfs4_server_caps_res res = {};
3843 	struct rpc_message msg = {
3844 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3845 		.rpc_argp = &args,
3846 		.rpc_resp = &res,
3847 	};
3848 	int status;
3849 	int i;
3850 
3851 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3852 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3853 		     FATTR4_WORD0_LINK_SUPPORT |
3854 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3855 		     FATTR4_WORD0_ACLSUPPORT |
3856 		     FATTR4_WORD0_CASE_INSENSITIVE |
3857 		     FATTR4_WORD0_CASE_PRESERVING;
3858 	if (minorversion)
3859 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3860 
3861 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3862 	if (status == 0) {
3863 		/* Sanity check the server answers */
3864 		switch (minorversion) {
3865 		case 0:
3866 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3867 			res.attr_bitmask[2] = 0;
3868 			break;
3869 		case 1:
3870 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3871 			break;
3872 		case 2:
3873 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3874 		}
3875 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3876 		server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3877 				  NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3878 		server->fattr_valid = NFS_ATTR_FATTR_V4;
3879 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3880 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3881 			server->caps |= NFS_CAP_ACLS;
3882 		if (res.has_links != 0)
3883 			server->caps |= NFS_CAP_HARDLINKS;
3884 		if (res.has_symlinks != 0)
3885 			server->caps |= NFS_CAP_SYMLINKS;
3886 		if (res.case_insensitive)
3887 			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3888 		if (res.case_preserving)
3889 			server->caps |= NFS_CAP_CASE_PRESERVING;
3890 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3891 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3892 			server->caps |= NFS_CAP_SECURITY_LABEL;
3893 #endif
3894 		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3895 			server->caps |= NFS_CAP_FS_LOCATIONS;
3896 		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3897 			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3898 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3899 			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3900 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3901 			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3902 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3903 			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3904 				NFS_ATTR_FATTR_OWNER_NAME);
3905 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3906 			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3907 				NFS_ATTR_FATTR_GROUP_NAME);
3908 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3909 			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3910 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3911 			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3912 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3913 			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3914 		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3915 			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3916 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3917 				sizeof(server->attr_bitmask));
3918 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3919 
3920 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3921 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3922 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3923 		server->cache_consistency_bitmask[2] = 0;
3924 
3925 		/* Avoid a regression due to buggy server */
3926 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3927 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3928 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3929 			sizeof(server->exclcreat_bitmask));
3930 
3931 		server->acl_bitmask = res.acl_bitmask;
3932 		server->fh_expire_type = res.fh_expire_type;
3933 	}
3934 
3935 	return status;
3936 }
3937 
3938 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3939 {
3940 	struct nfs4_exception exception = {
3941 		.interruptible = true,
3942 	};
3943 	int err;
3944 
3945 	nfs4_server_set_init_caps(server);
3946 	do {
3947 		err = nfs4_handle_exception(server,
3948 				_nfs4_server_capabilities(server, fhandle),
3949 				&exception);
3950 	} while (exception.retry);
3951 	return err;
3952 }
3953 
3954 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3955 					  struct nfs_client *clp,
3956 					  struct nfs_server *server)
3957 {
3958 	int i;
3959 
3960 	for (i = 0; i < location->nservers; i++) {
3961 		struct nfs4_string *srv_loc = &location->servers[i];
3962 		struct sockaddr_storage addr;
3963 		size_t addrlen;
3964 		struct xprt_create xprt_args = {
3965 			.ident = 0,
3966 			.net = clp->cl_net,
3967 		};
3968 		struct nfs4_add_xprt_data xprtdata = {
3969 			.clp = clp,
3970 		};
3971 		struct rpc_add_xprt_test rpcdata = {
3972 			.add_xprt_test = clp->cl_mvops->session_trunk,
3973 			.data = &xprtdata,
3974 		};
3975 		char *servername = NULL;
3976 
3977 		if (!srv_loc->len)
3978 			continue;
3979 
3980 		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3981 						&addr, sizeof(addr),
3982 						clp->cl_net, server->port);
3983 		if (!addrlen)
3984 			return;
3985 		xprt_args.dstaddr = (struct sockaddr *)&addr;
3986 		xprt_args.addrlen = addrlen;
3987 		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3988 		if (!servername)
3989 			return;
3990 		memcpy(servername, srv_loc->data, srv_loc->len);
3991 		servername[srv_loc->len] = '\0';
3992 		xprt_args.servername = servername;
3993 
3994 		xprtdata.cred = nfs4_get_clid_cred(clp);
3995 		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3996 				  rpc_clnt_setup_test_and_add_xprt,
3997 				  &rpcdata);
3998 		if (xprtdata.cred)
3999 			put_cred(xprtdata.cred);
4000 		kfree(servername);
4001 	}
4002 }
4003 
4004 static int _nfs4_discover_trunking(struct nfs_server *server,
4005 				   struct nfs_fh *fhandle)
4006 {
4007 	struct nfs4_fs_locations *locations = NULL;
4008 	struct page *page;
4009 	const struct cred *cred;
4010 	struct nfs_client *clp = server->nfs_client;
4011 	const struct nfs4_state_maintenance_ops *ops =
4012 		clp->cl_mvops->state_renewal_ops;
4013 	int status = -ENOMEM, i;
4014 
4015 	cred = ops->get_state_renewal_cred(clp);
4016 	if (cred == NULL) {
4017 		cred = nfs4_get_clid_cred(clp);
4018 		if (cred == NULL)
4019 			return -ENOKEY;
4020 	}
4021 
4022 	page = alloc_page(GFP_KERNEL);
4023 	if (!page)
4024 		goto out_put_cred;
4025 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4026 	if (!locations)
4027 		goto out_free;
4028 	locations->fattr = nfs_alloc_fattr();
4029 	if (!locations->fattr)
4030 		goto out_free_2;
4031 
4032 	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4033 					 cred);
4034 	if (status)
4035 		goto out_free_3;
4036 
4037 	for (i = 0; i < locations->nlocations; i++)
4038 		test_fs_location_for_trunking(&locations->locations[i], clp,
4039 					      server);
4040 out_free_3:
4041 	kfree(locations->fattr);
4042 out_free_2:
4043 	kfree(locations);
4044 out_free:
4045 	__free_page(page);
4046 out_put_cred:
4047 	put_cred(cred);
4048 	return status;
4049 }
4050 
4051 static int nfs4_discover_trunking(struct nfs_server *server,
4052 				  struct nfs_fh *fhandle)
4053 {
4054 	struct nfs4_exception exception = {
4055 		.interruptible = true,
4056 	};
4057 	struct nfs_client *clp = server->nfs_client;
4058 	int err = 0;
4059 
4060 	if (!nfs4_has_session(clp))
4061 		goto out;
4062 	do {
4063 		err = nfs4_handle_exception(server,
4064 				_nfs4_discover_trunking(server, fhandle),
4065 				&exception);
4066 	} while (exception.retry);
4067 out:
4068 	return err;
4069 }
4070 
4071 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4072 		struct nfs_fsinfo *info)
4073 {
4074 	u32 bitmask[3];
4075 	struct nfs4_lookup_root_arg args = {
4076 		.bitmask = bitmask,
4077 	};
4078 	struct nfs4_lookup_res res = {
4079 		.server = server,
4080 		.fattr = info->fattr,
4081 		.fh = fhandle,
4082 	};
4083 	struct rpc_message msg = {
4084 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4085 		.rpc_argp = &args,
4086 		.rpc_resp = &res,
4087 	};
4088 
4089 	bitmask[0] = nfs4_fattr_bitmap[0];
4090 	bitmask[1] = nfs4_fattr_bitmap[1];
4091 	/*
4092 	 * Process the label in the upcoming getfattr
4093 	 */
4094 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4095 
4096 	nfs_fattr_init(info->fattr);
4097 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4098 }
4099 
4100 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4101 		struct nfs_fsinfo *info)
4102 {
4103 	struct nfs4_exception exception = {
4104 		.interruptible = true,
4105 	};
4106 	int err;
4107 	do {
4108 		err = _nfs4_lookup_root(server, fhandle, info);
4109 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4110 		switch (err) {
4111 		case 0:
4112 		case -NFS4ERR_WRONGSEC:
4113 			goto out;
4114 		default:
4115 			err = nfs4_handle_exception(server, err, &exception);
4116 		}
4117 	} while (exception.retry);
4118 out:
4119 	return err;
4120 }
4121 
4122 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4123 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4124 {
4125 	struct rpc_auth_create_args auth_args = {
4126 		.pseudoflavor = flavor,
4127 	};
4128 	struct rpc_auth *auth;
4129 
4130 	auth = rpcauth_create(&auth_args, server->client);
4131 	if (IS_ERR(auth))
4132 		return -EACCES;
4133 	return nfs4_lookup_root(server, fhandle, info);
4134 }
4135 
4136 /*
4137  * Retry pseudoroot lookup with various security flavors.  We do this when:
4138  *
4139  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4140  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4141  *
4142  * Returns zero on success, or a negative NFS4ERR value, or a
4143  * negative errno value.
4144  */
4145 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4146 			      struct nfs_fsinfo *info)
4147 {
4148 	/* Per 3530bis 15.33.5 */
4149 	static const rpc_authflavor_t flav_array[] = {
4150 		RPC_AUTH_GSS_KRB5P,
4151 		RPC_AUTH_GSS_KRB5I,
4152 		RPC_AUTH_GSS_KRB5,
4153 		RPC_AUTH_UNIX,			/* courtesy */
4154 		RPC_AUTH_NULL,
4155 	};
4156 	int status = -EPERM;
4157 	size_t i;
4158 
4159 	if (server->auth_info.flavor_len > 0) {
4160 		/* try each flavor specified by user */
4161 		for (i = 0; i < server->auth_info.flavor_len; i++) {
4162 			status = nfs4_lookup_root_sec(server, fhandle, info,
4163 						server->auth_info.flavors[i]);
4164 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4165 				continue;
4166 			break;
4167 		}
4168 	} else {
4169 		/* no flavors specified by user, try default list */
4170 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4171 			status = nfs4_lookup_root_sec(server, fhandle, info,
4172 						      flav_array[i]);
4173 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4174 				continue;
4175 			break;
4176 		}
4177 	}
4178 
4179 	/*
4180 	 * -EACCES could mean that the user doesn't have correct permissions
4181 	 * to access the mount.  It could also mean that we tried to mount
4182 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4183 	 * existing mount programs don't handle -EACCES very well so it should
4184 	 * be mapped to -EPERM instead.
4185 	 */
4186 	if (status == -EACCES)
4187 		status = -EPERM;
4188 	return status;
4189 }
4190 
4191 /**
4192  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4193  * @server: initialized nfs_server handle
4194  * @fhandle: we fill in the pseudo-fs root file handle
4195  * @info: we fill in an FSINFO struct
4196  * @auth_probe: probe the auth flavours
4197  *
4198  * Returns zero on success, or a negative errno.
4199  */
4200 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4201 			 struct nfs_fsinfo *info,
4202 			 bool auth_probe)
4203 {
4204 	int status = 0;
4205 
4206 	if (!auth_probe)
4207 		status = nfs4_lookup_root(server, fhandle, info);
4208 
4209 	if (auth_probe || status == NFS4ERR_WRONGSEC)
4210 		status = server->nfs_client->cl_mvops->find_root_sec(server,
4211 				fhandle, info);
4212 
4213 	if (status == 0)
4214 		status = nfs4_server_capabilities(server, fhandle);
4215 	if (status == 0)
4216 		status = nfs4_do_fsinfo(server, fhandle, info);
4217 
4218 	return nfs4_map_errors(status);
4219 }
4220 
4221 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4222 			      struct nfs_fsinfo *info)
4223 {
4224 	int error;
4225 	struct nfs_fattr *fattr = info->fattr;
4226 
4227 	error = nfs4_server_capabilities(server, mntfh);
4228 	if (error < 0) {
4229 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4230 		return error;
4231 	}
4232 
4233 	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4234 	if (error < 0) {
4235 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4236 		goto out;
4237 	}
4238 
4239 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4240 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4241 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4242 
4243 out:
4244 	return error;
4245 }
4246 
4247 /*
4248  * Get locations and (maybe) other attributes of a referral.
4249  * Note that we'll actually follow the referral later when
4250  * we detect fsid mismatch in inode revalidation
4251  */
4252 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4253 			     const struct qstr *name, struct nfs_fattr *fattr,
4254 			     struct nfs_fh *fhandle)
4255 {
4256 	int status = -ENOMEM;
4257 	struct page *page = NULL;
4258 	struct nfs4_fs_locations *locations = NULL;
4259 
4260 	page = alloc_page(GFP_KERNEL);
4261 	if (page == NULL)
4262 		goto out;
4263 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4264 	if (locations == NULL)
4265 		goto out;
4266 
4267 	locations->fattr = fattr;
4268 
4269 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4270 	if (status != 0)
4271 		goto out;
4272 
4273 	/*
4274 	 * If the fsid didn't change, this is a migration event, not a
4275 	 * referral.  Cause us to drop into the exception handler, which
4276 	 * will kick off migration recovery.
4277 	 */
4278 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4279 		dprintk("%s: server did not return a different fsid for"
4280 			" a referral at %s\n", __func__, name->name);
4281 		status = -NFS4ERR_MOVED;
4282 		goto out;
4283 	}
4284 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4285 	nfs_fixup_referral_attributes(fattr);
4286 	memset(fhandle, 0, sizeof(struct nfs_fh));
4287 out:
4288 	if (page)
4289 		__free_page(page);
4290 	kfree(locations);
4291 	return status;
4292 }
4293 
4294 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4295 				struct nfs_fattr *fattr, struct inode *inode)
4296 {
4297 	__u32 bitmask[NFS4_BITMASK_SZ];
4298 	struct nfs4_getattr_arg args = {
4299 		.fh = fhandle,
4300 		.bitmask = bitmask,
4301 	};
4302 	struct nfs4_getattr_res res = {
4303 		.fattr = fattr,
4304 		.server = server,
4305 	};
4306 	struct rpc_message msg = {
4307 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4308 		.rpc_argp = &args,
4309 		.rpc_resp = &res,
4310 	};
4311 	unsigned short task_flags = 0;
4312 
4313 	if (nfs4_has_session(server->nfs_client))
4314 		task_flags = RPC_TASK_MOVEABLE;
4315 
4316 	/* Is this is an attribute revalidation, subject to softreval? */
4317 	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4318 		task_flags |= RPC_TASK_TIMEOUT;
4319 
4320 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4321 	nfs_fattr_init(fattr);
4322 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4323 	return nfs4_do_call_sync(server->client, server, &msg,
4324 			&args.seq_args, &res.seq_res, task_flags);
4325 }
4326 
4327 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4328 				struct nfs_fattr *fattr, struct inode *inode)
4329 {
4330 	struct nfs4_exception exception = {
4331 		.interruptible = true,
4332 	};
4333 	int err;
4334 	do {
4335 		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4336 		trace_nfs4_getattr(server, fhandle, fattr, err);
4337 		err = nfs4_handle_exception(server, err,
4338 				&exception);
4339 	} while (exception.retry);
4340 	return err;
4341 }
4342 
4343 /*
4344  * The file is not closed if it is opened due to the a request to change
4345  * the size of the file. The open call will not be needed once the
4346  * VFS layer lookup-intents are implemented.
4347  *
4348  * Close is called when the inode is destroyed.
4349  * If we haven't opened the file for O_WRONLY, we
4350  * need to in the size_change case to obtain a stateid.
4351  *
4352  * Got race?
4353  * Because OPEN is always done by name in nfsv4, it is
4354  * possible that we opened a different file by the same
4355  * name.  We can recognize this race condition, but we
4356  * can't do anything about it besides returning an error.
4357  *
4358  * This will be fixed with VFS changes (lookup-intent).
4359  */
4360 static int
4361 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4362 		  struct iattr *sattr)
4363 {
4364 	struct inode *inode = d_inode(dentry);
4365 	const struct cred *cred = NULL;
4366 	struct nfs_open_context *ctx = NULL;
4367 	int status;
4368 
4369 	if (pnfs_ld_layoutret_on_setattr(inode) &&
4370 	    sattr->ia_valid & ATTR_SIZE &&
4371 	    sattr->ia_size < i_size_read(inode))
4372 		pnfs_commit_and_return_layout(inode);
4373 
4374 	nfs_fattr_init(fattr);
4375 
4376 	/* Deal with open(O_TRUNC) */
4377 	if (sattr->ia_valid & ATTR_OPEN)
4378 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4379 
4380 	/* Optimization: if the end result is no change, don't RPC */
4381 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4382 		return 0;
4383 
4384 	/* Search for an existing open(O_WRITE) file */
4385 	if (sattr->ia_valid & ATTR_FILE) {
4386 
4387 		ctx = nfs_file_open_context(sattr->ia_file);
4388 		if (ctx)
4389 			cred = ctx->cred;
4390 	}
4391 
4392 	/* Return any delegations if we're going to change ACLs */
4393 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4394 		nfs4_inode_make_writeable(inode);
4395 
4396 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4397 	if (status == 0) {
4398 		nfs_setattr_update_inode(inode, sattr, fattr);
4399 		nfs_setsecurity(inode, fattr);
4400 	}
4401 	return status;
4402 }
4403 
4404 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4405 		struct dentry *dentry, struct nfs_fh *fhandle,
4406 		struct nfs_fattr *fattr)
4407 {
4408 	struct nfs_server *server = NFS_SERVER(dir);
4409 	int		       status;
4410 	struct nfs4_lookup_arg args = {
4411 		.bitmask = server->attr_bitmask,
4412 		.dir_fh = NFS_FH(dir),
4413 		.name = &dentry->d_name,
4414 	};
4415 	struct nfs4_lookup_res res = {
4416 		.server = server,
4417 		.fattr = fattr,
4418 		.fh = fhandle,
4419 	};
4420 	struct rpc_message msg = {
4421 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4422 		.rpc_argp = &args,
4423 		.rpc_resp = &res,
4424 	};
4425 	unsigned short task_flags = 0;
4426 
4427 	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4428 		task_flags = RPC_TASK_MOVEABLE;
4429 
4430 	/* Is this is an attribute revalidation, subject to softreval? */
4431 	if (nfs_lookup_is_soft_revalidate(dentry))
4432 		task_flags |= RPC_TASK_TIMEOUT;
4433 
4434 	args.bitmask = nfs4_bitmask(server, fattr->label);
4435 
4436 	nfs_fattr_init(fattr);
4437 
4438 	dprintk("NFS call  lookup %pd2\n", dentry);
4439 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4440 	status = nfs4_do_call_sync(clnt, server, &msg,
4441 			&args.seq_args, &res.seq_res, task_flags);
4442 	dprintk("NFS reply lookup: %d\n", status);
4443 	return status;
4444 }
4445 
4446 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4447 {
4448 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4449 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4450 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4451 	fattr->nlink = 2;
4452 }
4453 
4454 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4455 				   struct dentry *dentry, struct nfs_fh *fhandle,
4456 				   struct nfs_fattr *fattr)
4457 {
4458 	struct nfs4_exception exception = {
4459 		.interruptible = true,
4460 	};
4461 	struct rpc_clnt *client = *clnt;
4462 	const struct qstr *name = &dentry->d_name;
4463 	int err;
4464 	do {
4465 		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4466 		trace_nfs4_lookup(dir, name, err);
4467 		switch (err) {
4468 		case -NFS4ERR_BADNAME:
4469 			err = -ENOENT;
4470 			goto out;
4471 		case -NFS4ERR_MOVED:
4472 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4473 			if (err == -NFS4ERR_MOVED)
4474 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4475 			goto out;
4476 		case -NFS4ERR_WRONGSEC:
4477 			err = -EPERM;
4478 			if (client != *clnt)
4479 				goto out;
4480 			client = nfs4_negotiate_security(client, dir, name);
4481 			if (IS_ERR(client))
4482 				return PTR_ERR(client);
4483 
4484 			exception.retry = 1;
4485 			break;
4486 		default:
4487 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4488 		}
4489 	} while (exception.retry);
4490 
4491 out:
4492 	if (err == 0)
4493 		*clnt = client;
4494 	else if (client != *clnt)
4495 		rpc_shutdown_client(client);
4496 
4497 	return err;
4498 }
4499 
4500 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4501 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4502 {
4503 	int status;
4504 	struct rpc_clnt *client = NFS_CLIENT(dir);
4505 
4506 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4507 	if (client != NFS_CLIENT(dir)) {
4508 		rpc_shutdown_client(client);
4509 		nfs_fixup_secinfo_attributes(fattr);
4510 	}
4511 	return status;
4512 }
4513 
4514 struct rpc_clnt *
4515 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4516 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4517 {
4518 	struct rpc_clnt *client = NFS_CLIENT(dir);
4519 	int status;
4520 
4521 	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4522 	if (status < 0)
4523 		return ERR_PTR(status);
4524 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4525 }
4526 
4527 static int _nfs4_proc_lookupp(struct inode *inode,
4528 		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4529 {
4530 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4531 	struct nfs_server *server = NFS_SERVER(inode);
4532 	int		       status;
4533 	struct nfs4_lookupp_arg args = {
4534 		.bitmask = server->attr_bitmask,
4535 		.fh = NFS_FH(inode),
4536 	};
4537 	struct nfs4_lookupp_res res = {
4538 		.server = server,
4539 		.fattr = fattr,
4540 		.fh = fhandle,
4541 	};
4542 	struct rpc_message msg = {
4543 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4544 		.rpc_argp = &args,
4545 		.rpc_resp = &res,
4546 	};
4547 	unsigned short task_flags = 0;
4548 
4549 	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4550 		task_flags |= RPC_TASK_TIMEOUT;
4551 
4552 	args.bitmask = nfs4_bitmask(server, fattr->label);
4553 
4554 	nfs_fattr_init(fattr);
4555 
4556 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4557 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4558 				&res.seq_res, task_flags);
4559 	dprintk("NFS reply lookupp: %d\n", status);
4560 	return status;
4561 }
4562 
4563 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4564 			     struct nfs_fattr *fattr)
4565 {
4566 	struct nfs4_exception exception = {
4567 		.interruptible = true,
4568 	};
4569 	int err;
4570 	do {
4571 		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4572 		trace_nfs4_lookupp(inode, err);
4573 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4574 				&exception);
4575 	} while (exception.retry);
4576 	return err;
4577 }
4578 
4579 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4580 			     const struct cred *cred)
4581 {
4582 	struct nfs_server *server = NFS_SERVER(inode);
4583 	struct nfs4_accessargs args = {
4584 		.fh = NFS_FH(inode),
4585 		.access = entry->mask,
4586 	};
4587 	struct nfs4_accessres res = {
4588 		.server = server,
4589 	};
4590 	struct rpc_message msg = {
4591 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4592 		.rpc_argp = &args,
4593 		.rpc_resp = &res,
4594 		.rpc_cred = cred,
4595 	};
4596 	int status = 0;
4597 
4598 	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4599 		res.fattr = nfs_alloc_fattr();
4600 		if (res.fattr == NULL)
4601 			return -ENOMEM;
4602 		args.bitmask = server->cache_consistency_bitmask;
4603 	}
4604 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4605 	if (!status) {
4606 		nfs_access_set_mask(entry, res.access);
4607 		if (res.fattr)
4608 			nfs_refresh_inode(inode, res.fattr);
4609 	}
4610 	nfs_free_fattr(res.fattr);
4611 	return status;
4612 }
4613 
4614 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4615 			    const struct cred *cred)
4616 {
4617 	struct nfs4_exception exception = {
4618 		.interruptible = true,
4619 	};
4620 	int err;
4621 	do {
4622 		err = _nfs4_proc_access(inode, entry, cred);
4623 		trace_nfs4_access(inode, err);
4624 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4625 				&exception);
4626 	} while (exception.retry);
4627 	return err;
4628 }
4629 
4630 /*
4631  * TODO: For the time being, we don't try to get any attributes
4632  * along with any of the zero-copy operations READ, READDIR,
4633  * READLINK, WRITE.
4634  *
4635  * In the case of the first three, we want to put the GETATTR
4636  * after the read-type operation -- this is because it is hard
4637  * to predict the length of a GETATTR response in v4, and thus
4638  * align the READ data correctly.  This means that the GETATTR
4639  * may end up partially falling into the page cache, and we should
4640  * shift it into the 'tail' of the xdr_buf before processing.
4641  * To do this efficiently, we need to know the total length
4642  * of data received, which doesn't seem to be available outside
4643  * of the RPC layer.
4644  *
4645  * In the case of WRITE, we also want to put the GETATTR after
4646  * the operation -- in this case because we want to make sure
4647  * we get the post-operation mtime and size.
4648  *
4649  * Both of these changes to the XDR layer would in fact be quite
4650  * minor, but I decided to leave them for a subsequent patch.
4651  */
4652 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4653 		unsigned int pgbase, unsigned int pglen)
4654 {
4655 	struct nfs4_readlink args = {
4656 		.fh       = NFS_FH(inode),
4657 		.pgbase	  = pgbase,
4658 		.pglen    = pglen,
4659 		.pages    = &page,
4660 	};
4661 	struct nfs4_readlink_res res;
4662 	struct rpc_message msg = {
4663 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4664 		.rpc_argp = &args,
4665 		.rpc_resp = &res,
4666 	};
4667 
4668 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4669 }
4670 
4671 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4672 		unsigned int pgbase, unsigned int pglen)
4673 {
4674 	struct nfs4_exception exception = {
4675 		.interruptible = true,
4676 	};
4677 	int err;
4678 	do {
4679 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4680 		trace_nfs4_readlink(inode, err);
4681 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4682 				&exception);
4683 	} while (exception.retry);
4684 	return err;
4685 }
4686 
4687 /*
4688  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4689  */
4690 static int
4691 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4692 		 int flags)
4693 {
4694 	struct nfs_server *server = NFS_SERVER(dir);
4695 	struct nfs4_label l, *ilabel;
4696 	struct nfs_open_context *ctx;
4697 	struct nfs4_state *state;
4698 	int status = 0;
4699 
4700 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4701 	if (IS_ERR(ctx))
4702 		return PTR_ERR(ctx);
4703 
4704 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4705 
4706 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4707 		sattr->ia_mode &= ~current_umask();
4708 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4709 	if (IS_ERR(state)) {
4710 		status = PTR_ERR(state);
4711 		goto out;
4712 	}
4713 out:
4714 	nfs4_label_release_security(ilabel);
4715 	put_nfs_open_context(ctx);
4716 	return status;
4717 }
4718 
4719 static int
4720 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4721 {
4722 	struct nfs_server *server = NFS_SERVER(dir);
4723 	struct nfs_removeargs args = {
4724 		.fh = NFS_FH(dir),
4725 		.name = *name,
4726 	};
4727 	struct nfs_removeres res = {
4728 		.server = server,
4729 	};
4730 	struct rpc_message msg = {
4731 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4732 		.rpc_argp = &args,
4733 		.rpc_resp = &res,
4734 	};
4735 	unsigned long timestamp = jiffies;
4736 	int status;
4737 
4738 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4739 	if (status == 0) {
4740 		spin_lock(&dir->i_lock);
4741 		/* Removing a directory decrements nlink in the parent */
4742 		if (ftype == NF4DIR && dir->i_nlink > 2)
4743 			nfs4_dec_nlink_locked(dir);
4744 		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4745 					      NFS_INO_INVALID_DATA);
4746 		spin_unlock(&dir->i_lock);
4747 	}
4748 	return status;
4749 }
4750 
4751 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4752 {
4753 	struct nfs4_exception exception = {
4754 		.interruptible = true,
4755 	};
4756 	struct inode *inode = d_inode(dentry);
4757 	int err;
4758 
4759 	if (inode) {
4760 		if (inode->i_nlink == 1)
4761 			nfs4_inode_return_delegation(inode);
4762 		else
4763 			nfs4_inode_make_writeable(inode);
4764 	}
4765 	do {
4766 		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4767 		trace_nfs4_remove(dir, &dentry->d_name, err);
4768 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4769 				&exception);
4770 	} while (exception.retry);
4771 	return err;
4772 }
4773 
4774 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4775 {
4776 	struct nfs4_exception exception = {
4777 		.interruptible = true,
4778 	};
4779 	int err;
4780 
4781 	do {
4782 		err = _nfs4_proc_remove(dir, name, NF4DIR);
4783 		trace_nfs4_remove(dir, name, err);
4784 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4785 				&exception);
4786 	} while (exception.retry);
4787 	return err;
4788 }
4789 
4790 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4791 		struct dentry *dentry,
4792 		struct inode *inode)
4793 {
4794 	struct nfs_removeargs *args = msg->rpc_argp;
4795 	struct nfs_removeres *res = msg->rpc_resp;
4796 
4797 	res->server = NFS_SB(dentry->d_sb);
4798 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4799 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4800 
4801 	nfs_fattr_init(res->dir_attr);
4802 
4803 	if (inode) {
4804 		nfs4_inode_return_delegation(inode);
4805 		nfs_d_prune_case_insensitive_aliases(inode);
4806 	}
4807 }
4808 
4809 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4810 {
4811 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4812 			&data->args.seq_args,
4813 			&data->res.seq_res,
4814 			task);
4815 }
4816 
4817 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4818 {
4819 	struct nfs_unlinkdata *data = task->tk_calldata;
4820 	struct nfs_removeres *res = &data->res;
4821 
4822 	if (!nfs4_sequence_done(task, &res->seq_res))
4823 		return 0;
4824 	if (nfs4_async_handle_error(task, res->server, NULL,
4825 				    &data->timeout) == -EAGAIN)
4826 		return 0;
4827 	if (task->tk_status == 0)
4828 		nfs4_update_changeattr(dir, &res->cinfo,
4829 				res->dir_attr->time_start,
4830 				NFS_INO_INVALID_DATA);
4831 	return 1;
4832 }
4833 
4834 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4835 		struct dentry *old_dentry,
4836 		struct dentry *new_dentry)
4837 {
4838 	struct nfs_renameargs *arg = msg->rpc_argp;
4839 	struct nfs_renameres *res = msg->rpc_resp;
4840 	struct inode *old_inode = d_inode(old_dentry);
4841 	struct inode *new_inode = d_inode(new_dentry);
4842 
4843 	if (old_inode)
4844 		nfs4_inode_make_writeable(old_inode);
4845 	if (new_inode)
4846 		nfs4_inode_return_delegation(new_inode);
4847 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4848 	res->server = NFS_SB(old_dentry->d_sb);
4849 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4850 }
4851 
4852 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4853 {
4854 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4855 			&data->args.seq_args,
4856 			&data->res.seq_res,
4857 			task);
4858 }
4859 
4860 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4861 				 struct inode *new_dir)
4862 {
4863 	struct nfs_renamedata *data = task->tk_calldata;
4864 	struct nfs_renameres *res = &data->res;
4865 
4866 	if (!nfs4_sequence_done(task, &res->seq_res))
4867 		return 0;
4868 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4869 		return 0;
4870 
4871 	if (task->tk_status == 0) {
4872 		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4873 		if (new_dir != old_dir) {
4874 			/* Note: If we moved a directory, nlink will change */
4875 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4876 					res->old_fattr->time_start,
4877 					NFS_INO_INVALID_NLINK |
4878 					    NFS_INO_INVALID_DATA);
4879 			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4880 					res->new_fattr->time_start,
4881 					NFS_INO_INVALID_NLINK |
4882 					    NFS_INO_INVALID_DATA);
4883 		} else
4884 			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4885 					res->old_fattr->time_start,
4886 					NFS_INO_INVALID_DATA);
4887 	}
4888 	return 1;
4889 }
4890 
4891 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4892 {
4893 	struct nfs_server *server = NFS_SERVER(inode);
4894 	__u32 bitmask[NFS4_BITMASK_SZ];
4895 	struct nfs4_link_arg arg = {
4896 		.fh     = NFS_FH(inode),
4897 		.dir_fh = NFS_FH(dir),
4898 		.name   = name,
4899 		.bitmask = bitmask,
4900 	};
4901 	struct nfs4_link_res res = {
4902 		.server = server,
4903 	};
4904 	struct rpc_message msg = {
4905 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4906 		.rpc_argp = &arg,
4907 		.rpc_resp = &res,
4908 	};
4909 	int status = -ENOMEM;
4910 
4911 	res.fattr = nfs_alloc_fattr_with_label(server);
4912 	if (res.fattr == NULL)
4913 		goto out;
4914 
4915 	nfs4_inode_make_writeable(inode);
4916 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4917 				NFS_INO_INVALID_CHANGE);
4918 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4919 	if (!status) {
4920 		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4921 				       NFS_INO_INVALID_DATA);
4922 		nfs4_inc_nlink(inode);
4923 		status = nfs_post_op_update_inode(inode, res.fattr);
4924 		if (!status)
4925 			nfs_setsecurity(inode, res.fattr);
4926 	}
4927 
4928 out:
4929 	nfs_free_fattr(res.fattr);
4930 	return status;
4931 }
4932 
4933 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4934 {
4935 	struct nfs4_exception exception = {
4936 		.interruptible = true,
4937 	};
4938 	int err;
4939 	do {
4940 		err = nfs4_handle_exception(NFS_SERVER(inode),
4941 				_nfs4_proc_link(inode, dir, name),
4942 				&exception);
4943 	} while (exception.retry);
4944 	return err;
4945 }
4946 
4947 struct nfs4_createdata {
4948 	struct rpc_message msg;
4949 	struct nfs4_create_arg arg;
4950 	struct nfs4_create_res res;
4951 	struct nfs_fh fh;
4952 	struct nfs_fattr fattr;
4953 };
4954 
4955 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4956 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4957 {
4958 	struct nfs4_createdata *data;
4959 
4960 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4961 	if (data != NULL) {
4962 		struct nfs_server *server = NFS_SERVER(dir);
4963 
4964 		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4965 		if (IS_ERR(data->fattr.label))
4966 			goto out_free;
4967 
4968 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4969 		data->msg.rpc_argp = &data->arg;
4970 		data->msg.rpc_resp = &data->res;
4971 		data->arg.dir_fh = NFS_FH(dir);
4972 		data->arg.server = server;
4973 		data->arg.name = name;
4974 		data->arg.attrs = sattr;
4975 		data->arg.ftype = ftype;
4976 		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4977 		data->arg.umask = current_umask();
4978 		data->res.server = server;
4979 		data->res.fh = &data->fh;
4980 		data->res.fattr = &data->fattr;
4981 		nfs_fattr_init(data->res.fattr);
4982 	}
4983 	return data;
4984 out_free:
4985 	kfree(data);
4986 	return NULL;
4987 }
4988 
4989 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4990 {
4991 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4992 				    &data->arg.seq_args, &data->res.seq_res, 1);
4993 	if (status == 0) {
4994 		spin_lock(&dir->i_lock);
4995 		/* Creating a directory bumps nlink in the parent */
4996 		if (data->arg.ftype == NF4DIR)
4997 			nfs4_inc_nlink_locked(dir);
4998 		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4999 					      data->res.fattr->time_start,
5000 					      NFS_INO_INVALID_DATA);
5001 		spin_unlock(&dir->i_lock);
5002 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
5003 	}
5004 	return status;
5005 }
5006 
5007 static void nfs4_free_createdata(struct nfs4_createdata *data)
5008 {
5009 	nfs4_label_free(data->fattr.label);
5010 	kfree(data);
5011 }
5012 
5013 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5014 		struct page *page, unsigned int len, struct iattr *sattr,
5015 		struct nfs4_label *label)
5016 {
5017 	struct nfs4_createdata *data;
5018 	int status = -ENAMETOOLONG;
5019 
5020 	if (len > NFS4_MAXPATHLEN)
5021 		goto out;
5022 
5023 	status = -ENOMEM;
5024 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5025 	if (data == NULL)
5026 		goto out;
5027 
5028 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5029 	data->arg.u.symlink.pages = &page;
5030 	data->arg.u.symlink.len = len;
5031 	data->arg.label = label;
5032 
5033 	status = nfs4_do_create(dir, dentry, data);
5034 
5035 	nfs4_free_createdata(data);
5036 out:
5037 	return status;
5038 }
5039 
5040 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5041 		struct page *page, unsigned int len, struct iattr *sattr)
5042 {
5043 	struct nfs4_exception exception = {
5044 		.interruptible = true,
5045 	};
5046 	struct nfs4_label l, *label;
5047 	int err;
5048 
5049 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5050 
5051 	do {
5052 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5053 		trace_nfs4_symlink(dir, &dentry->d_name, err);
5054 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5055 				&exception);
5056 	} while (exception.retry);
5057 
5058 	nfs4_label_release_security(label);
5059 	return err;
5060 }
5061 
5062 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5063 		struct iattr *sattr, struct nfs4_label *label)
5064 {
5065 	struct nfs4_createdata *data;
5066 	int status = -ENOMEM;
5067 
5068 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5069 	if (data == NULL)
5070 		goto out;
5071 
5072 	data->arg.label = label;
5073 	status = nfs4_do_create(dir, dentry, data);
5074 
5075 	nfs4_free_createdata(data);
5076 out:
5077 	return status;
5078 }
5079 
5080 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5081 		struct iattr *sattr)
5082 {
5083 	struct nfs_server *server = NFS_SERVER(dir);
5084 	struct nfs4_exception exception = {
5085 		.interruptible = true,
5086 	};
5087 	struct nfs4_label l, *label;
5088 	int err;
5089 
5090 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5091 
5092 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5093 		sattr->ia_mode &= ~current_umask();
5094 	do {
5095 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5096 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5097 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5098 				&exception);
5099 	} while (exception.retry);
5100 	nfs4_label_release_security(label);
5101 
5102 	return err;
5103 }
5104 
5105 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5106 			      struct nfs_readdir_res *nr_res)
5107 {
5108 	struct inode		*dir = d_inode(nr_arg->dentry);
5109 	struct nfs_server	*server = NFS_SERVER(dir);
5110 	struct nfs4_readdir_arg args = {
5111 		.fh = NFS_FH(dir),
5112 		.pages = nr_arg->pages,
5113 		.pgbase = 0,
5114 		.count = nr_arg->page_len,
5115 		.plus = nr_arg->plus,
5116 	};
5117 	struct nfs4_readdir_res res;
5118 	struct rpc_message msg = {
5119 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5120 		.rpc_argp = &args,
5121 		.rpc_resp = &res,
5122 		.rpc_cred = nr_arg->cred,
5123 	};
5124 	int			status;
5125 
5126 	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5127 		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5128 	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5129 		args.bitmask = server->attr_bitmask_nl;
5130 	else
5131 		args.bitmask = server->attr_bitmask;
5132 
5133 	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5134 	res.pgbase = args.pgbase;
5135 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5136 			&res.seq_res, 0);
5137 	if (status >= 0) {
5138 		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5139 		status += args.pgbase;
5140 	}
5141 
5142 	nfs_invalidate_atime(dir);
5143 
5144 	dprintk("%s: returns %d\n", __func__, status);
5145 	return status;
5146 }
5147 
5148 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5149 			     struct nfs_readdir_res *res)
5150 {
5151 	struct nfs4_exception exception = {
5152 		.interruptible = true,
5153 	};
5154 	int err;
5155 	do {
5156 		err = _nfs4_proc_readdir(arg, res);
5157 		trace_nfs4_readdir(d_inode(arg->dentry), err);
5158 		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5159 					    err, &exception);
5160 	} while (exception.retry);
5161 	return err;
5162 }
5163 
5164 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5165 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5166 {
5167 	struct nfs4_createdata *data;
5168 	int mode = sattr->ia_mode;
5169 	int status = -ENOMEM;
5170 
5171 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5172 	if (data == NULL)
5173 		goto out;
5174 
5175 	if (S_ISFIFO(mode))
5176 		data->arg.ftype = NF4FIFO;
5177 	else if (S_ISBLK(mode)) {
5178 		data->arg.ftype = NF4BLK;
5179 		data->arg.u.device.specdata1 = MAJOR(rdev);
5180 		data->arg.u.device.specdata2 = MINOR(rdev);
5181 	}
5182 	else if (S_ISCHR(mode)) {
5183 		data->arg.ftype = NF4CHR;
5184 		data->arg.u.device.specdata1 = MAJOR(rdev);
5185 		data->arg.u.device.specdata2 = MINOR(rdev);
5186 	} else if (!S_ISSOCK(mode)) {
5187 		status = -EINVAL;
5188 		goto out_free;
5189 	}
5190 
5191 	data->arg.label = label;
5192 	status = nfs4_do_create(dir, dentry, data);
5193 out_free:
5194 	nfs4_free_createdata(data);
5195 out:
5196 	return status;
5197 }
5198 
5199 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5200 		struct iattr *sattr, dev_t rdev)
5201 {
5202 	struct nfs_server *server = NFS_SERVER(dir);
5203 	struct nfs4_exception exception = {
5204 		.interruptible = true,
5205 	};
5206 	struct nfs4_label l, *label;
5207 	int err;
5208 
5209 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5210 
5211 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5212 		sattr->ia_mode &= ~current_umask();
5213 	do {
5214 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5215 		trace_nfs4_mknod(dir, &dentry->d_name, err);
5216 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5217 				&exception);
5218 	} while (exception.retry);
5219 
5220 	nfs4_label_release_security(label);
5221 
5222 	return err;
5223 }
5224 
5225 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5226 		 struct nfs_fsstat *fsstat)
5227 {
5228 	struct nfs4_statfs_arg args = {
5229 		.fh = fhandle,
5230 		.bitmask = server->attr_bitmask,
5231 	};
5232 	struct nfs4_statfs_res res = {
5233 		.fsstat = fsstat,
5234 	};
5235 	struct rpc_message msg = {
5236 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5237 		.rpc_argp = &args,
5238 		.rpc_resp = &res,
5239 	};
5240 
5241 	nfs_fattr_init(fsstat->fattr);
5242 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5243 }
5244 
5245 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5246 {
5247 	struct nfs4_exception exception = {
5248 		.interruptible = true,
5249 	};
5250 	int err;
5251 	do {
5252 		err = nfs4_handle_exception(server,
5253 				_nfs4_proc_statfs(server, fhandle, fsstat),
5254 				&exception);
5255 	} while (exception.retry);
5256 	return err;
5257 }
5258 
5259 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5260 		struct nfs_fsinfo *fsinfo)
5261 {
5262 	struct nfs4_fsinfo_arg args = {
5263 		.fh = fhandle,
5264 		.bitmask = server->attr_bitmask,
5265 	};
5266 	struct nfs4_fsinfo_res res = {
5267 		.fsinfo = fsinfo,
5268 	};
5269 	struct rpc_message msg = {
5270 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5271 		.rpc_argp = &args,
5272 		.rpc_resp = &res,
5273 	};
5274 
5275 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5276 }
5277 
5278 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5279 {
5280 	struct nfs4_exception exception = {
5281 		.interruptible = true,
5282 	};
5283 	int err;
5284 
5285 	do {
5286 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5287 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5288 		if (err == 0) {
5289 			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5290 			break;
5291 		}
5292 		err = nfs4_handle_exception(server, err, &exception);
5293 	} while (exception.retry);
5294 	return err;
5295 }
5296 
5297 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5298 {
5299 	int error;
5300 
5301 	nfs_fattr_init(fsinfo->fattr);
5302 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5303 	if (error == 0) {
5304 		/* block layout checks this! */
5305 		server->pnfs_blksize = fsinfo->blksize;
5306 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5307 	}
5308 
5309 	return error;
5310 }
5311 
5312 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5313 		struct nfs_pathconf *pathconf)
5314 {
5315 	struct nfs4_pathconf_arg args = {
5316 		.fh = fhandle,
5317 		.bitmask = server->attr_bitmask,
5318 	};
5319 	struct nfs4_pathconf_res res = {
5320 		.pathconf = pathconf,
5321 	};
5322 	struct rpc_message msg = {
5323 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5324 		.rpc_argp = &args,
5325 		.rpc_resp = &res,
5326 	};
5327 
5328 	/* None of the pathconf attributes are mandatory to implement */
5329 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5330 		memset(pathconf, 0, sizeof(*pathconf));
5331 		return 0;
5332 	}
5333 
5334 	nfs_fattr_init(pathconf->fattr);
5335 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5336 }
5337 
5338 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5339 		struct nfs_pathconf *pathconf)
5340 {
5341 	struct nfs4_exception exception = {
5342 		.interruptible = true,
5343 	};
5344 	int err;
5345 
5346 	do {
5347 		err = nfs4_handle_exception(server,
5348 				_nfs4_proc_pathconf(server, fhandle, pathconf),
5349 				&exception);
5350 	} while (exception.retry);
5351 	return err;
5352 }
5353 
5354 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5355 		const struct nfs_open_context *ctx,
5356 		const struct nfs_lock_context *l_ctx,
5357 		fmode_t fmode)
5358 {
5359 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5360 }
5361 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5362 
5363 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5364 		const struct nfs_open_context *ctx,
5365 		const struct nfs_lock_context *l_ctx,
5366 		fmode_t fmode)
5367 {
5368 	nfs4_stateid _current_stateid;
5369 
5370 	/* If the current stateid represents a lost lock, then exit */
5371 	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5372 		return true;
5373 	return nfs4_stateid_match(stateid, &_current_stateid);
5374 }
5375 
5376 static bool nfs4_error_stateid_expired(int err)
5377 {
5378 	switch (err) {
5379 	case -NFS4ERR_DELEG_REVOKED:
5380 	case -NFS4ERR_ADMIN_REVOKED:
5381 	case -NFS4ERR_BAD_STATEID:
5382 	case -NFS4ERR_STALE_STATEID:
5383 	case -NFS4ERR_OLD_STATEID:
5384 	case -NFS4ERR_OPENMODE:
5385 	case -NFS4ERR_EXPIRED:
5386 		return true;
5387 	}
5388 	return false;
5389 }
5390 
5391 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5392 {
5393 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5394 
5395 	trace_nfs4_read(hdr, task->tk_status);
5396 	if (task->tk_status < 0) {
5397 		struct nfs4_exception exception = {
5398 			.inode = hdr->inode,
5399 			.state = hdr->args.context->state,
5400 			.stateid = &hdr->args.stateid,
5401 		};
5402 		task->tk_status = nfs4_async_handle_exception(task,
5403 				server, task->tk_status, &exception);
5404 		if (exception.retry) {
5405 			rpc_restart_call_prepare(task);
5406 			return -EAGAIN;
5407 		}
5408 	}
5409 
5410 	if (task->tk_status > 0)
5411 		renew_lease(server, hdr->timestamp);
5412 	return 0;
5413 }
5414 
5415 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5416 		struct nfs_pgio_args *args)
5417 {
5418 
5419 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5420 		nfs4_stateid_is_current(&args->stateid,
5421 				args->context,
5422 				args->lock_context,
5423 				FMODE_READ))
5424 		return false;
5425 	rpc_restart_call_prepare(task);
5426 	return true;
5427 }
5428 
5429 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5430 					 struct nfs_pgio_header *hdr)
5431 {
5432 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5433 	struct rpc_message *msg = &task->tk_msg;
5434 
5435 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5436 	    server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5437 		server->caps &= ~NFS_CAP_READ_PLUS;
5438 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5439 		rpc_restart_call_prepare(task);
5440 		return true;
5441 	}
5442 	return false;
5443 }
5444 
5445 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5446 {
5447 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5448 		return -EAGAIN;
5449 	if (nfs4_read_stateid_changed(task, &hdr->args))
5450 		return -EAGAIN;
5451 	if (nfs4_read_plus_not_supported(task, hdr))
5452 		return -EAGAIN;
5453 	if (task->tk_status > 0)
5454 		nfs_invalidate_atime(hdr->inode);
5455 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5456 				    nfs4_read_done_cb(task, hdr);
5457 }
5458 
5459 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5460 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5461 				    struct rpc_message *msg)
5462 {
5463 	/* Note: We don't use READ_PLUS with pNFS yet */
5464 	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5465 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5466 		return nfs_read_alloc_scratch(hdr, READ_PLUS_SCRATCH_SIZE);
5467 	}
5468 	return false;
5469 }
5470 #else
5471 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5472 				    struct rpc_message *msg)
5473 {
5474 	return false;
5475 }
5476 #endif /* CONFIG_NFS_V4_2 */
5477 
5478 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5479 				 struct rpc_message *msg)
5480 {
5481 	hdr->timestamp   = jiffies;
5482 	if (!hdr->pgio_done_cb)
5483 		hdr->pgio_done_cb = nfs4_read_done_cb;
5484 	if (!nfs42_read_plus_support(hdr, msg))
5485 		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5486 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5487 }
5488 
5489 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5490 				      struct nfs_pgio_header *hdr)
5491 {
5492 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5493 			&hdr->args.seq_args,
5494 			&hdr->res.seq_res,
5495 			task))
5496 		return 0;
5497 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5498 				hdr->args.lock_context,
5499 				hdr->rw_mode) == -EIO)
5500 		return -EIO;
5501 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5502 		return -EIO;
5503 	return 0;
5504 }
5505 
5506 static int nfs4_write_done_cb(struct rpc_task *task,
5507 			      struct nfs_pgio_header *hdr)
5508 {
5509 	struct inode *inode = hdr->inode;
5510 
5511 	trace_nfs4_write(hdr, task->tk_status);
5512 	if (task->tk_status < 0) {
5513 		struct nfs4_exception exception = {
5514 			.inode = hdr->inode,
5515 			.state = hdr->args.context->state,
5516 			.stateid = &hdr->args.stateid,
5517 		};
5518 		task->tk_status = nfs4_async_handle_exception(task,
5519 				NFS_SERVER(inode), task->tk_status,
5520 				&exception);
5521 		if (exception.retry) {
5522 			rpc_restart_call_prepare(task);
5523 			return -EAGAIN;
5524 		}
5525 	}
5526 	if (task->tk_status >= 0) {
5527 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5528 		nfs_writeback_update_inode(hdr);
5529 	}
5530 	return 0;
5531 }
5532 
5533 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5534 		struct nfs_pgio_args *args)
5535 {
5536 
5537 	if (!nfs4_error_stateid_expired(task->tk_status) ||
5538 		nfs4_stateid_is_current(&args->stateid,
5539 				args->context,
5540 				args->lock_context,
5541 				FMODE_WRITE))
5542 		return false;
5543 	rpc_restart_call_prepare(task);
5544 	return true;
5545 }
5546 
5547 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5548 {
5549 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5550 		return -EAGAIN;
5551 	if (nfs4_write_stateid_changed(task, &hdr->args))
5552 		return -EAGAIN;
5553 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5554 		nfs4_write_done_cb(task, hdr);
5555 }
5556 
5557 static
5558 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5559 {
5560 	/* Don't request attributes for pNFS or O_DIRECT writes */
5561 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5562 		return false;
5563 	/* Otherwise, request attributes if and only if we don't hold
5564 	 * a delegation
5565 	 */
5566 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5567 }
5568 
5569 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5570 		      struct inode *inode, unsigned long cache_validity)
5571 {
5572 	struct nfs_server *server = NFS_SERVER(inode);
5573 	unsigned int i;
5574 
5575 	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5576 	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5577 
5578 	if (cache_validity & NFS_INO_INVALID_CHANGE)
5579 		bitmask[0] |= FATTR4_WORD0_CHANGE;
5580 	if (cache_validity & NFS_INO_INVALID_ATIME)
5581 		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5582 	if (cache_validity & NFS_INO_INVALID_MODE)
5583 		bitmask[1] |= FATTR4_WORD1_MODE;
5584 	if (cache_validity & NFS_INO_INVALID_OTHER)
5585 		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5586 	if (cache_validity & NFS_INO_INVALID_NLINK)
5587 		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5588 	if (cache_validity & NFS_INO_INVALID_CTIME)
5589 		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5590 	if (cache_validity & NFS_INO_INVALID_MTIME)
5591 		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5592 	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5593 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5594 
5595 	if (cache_validity & NFS_INO_INVALID_SIZE)
5596 		bitmask[0] |= FATTR4_WORD0_SIZE;
5597 
5598 	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5599 		bitmask[i] &= server->attr_bitmask[i];
5600 }
5601 
5602 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5603 				  struct rpc_message *msg,
5604 				  struct rpc_clnt **clnt)
5605 {
5606 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5607 
5608 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5609 		hdr->args.bitmask = NULL;
5610 		hdr->res.fattr = NULL;
5611 	} else {
5612 		nfs4_bitmask_set(hdr->args.bitmask_store,
5613 				 server->cache_consistency_bitmask,
5614 				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5615 		hdr->args.bitmask = hdr->args.bitmask_store;
5616 	}
5617 
5618 	if (!hdr->pgio_done_cb)
5619 		hdr->pgio_done_cb = nfs4_write_done_cb;
5620 	hdr->res.server = server;
5621 	hdr->timestamp   = jiffies;
5622 
5623 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5624 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5625 	nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5626 }
5627 
5628 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5629 {
5630 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5631 			&data->args.seq_args,
5632 			&data->res.seq_res,
5633 			task);
5634 }
5635 
5636 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5637 {
5638 	struct inode *inode = data->inode;
5639 
5640 	trace_nfs4_commit(data, task->tk_status);
5641 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5642 				    NULL, NULL) == -EAGAIN) {
5643 		rpc_restart_call_prepare(task);
5644 		return -EAGAIN;
5645 	}
5646 	return 0;
5647 }
5648 
5649 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5650 {
5651 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5652 		return -EAGAIN;
5653 	return data->commit_done_cb(task, data);
5654 }
5655 
5656 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5657 				   struct rpc_clnt **clnt)
5658 {
5659 	struct nfs_server *server = NFS_SERVER(data->inode);
5660 
5661 	if (data->commit_done_cb == NULL)
5662 		data->commit_done_cb = nfs4_commit_done_cb;
5663 	data->res.server = server;
5664 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5665 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5666 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5667 }
5668 
5669 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5670 				struct nfs_commitres *res)
5671 {
5672 	struct inode *dst_inode = file_inode(dst);
5673 	struct nfs_server *server = NFS_SERVER(dst_inode);
5674 	struct rpc_message msg = {
5675 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5676 		.rpc_argp = args,
5677 		.rpc_resp = res,
5678 	};
5679 
5680 	args->fh = NFS_FH(dst_inode);
5681 	return nfs4_call_sync(server->client, server, &msg,
5682 			&args->seq_args, &res->seq_res, 1);
5683 }
5684 
5685 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5686 {
5687 	struct nfs_commitargs args = {
5688 		.offset = offset,
5689 		.count = count,
5690 	};
5691 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5692 	struct nfs4_exception exception = { };
5693 	int status;
5694 
5695 	do {
5696 		status = _nfs4_proc_commit(dst, &args, res);
5697 		status = nfs4_handle_exception(dst_server, status, &exception);
5698 	} while (exception.retry);
5699 
5700 	return status;
5701 }
5702 
5703 struct nfs4_renewdata {
5704 	struct nfs_client	*client;
5705 	unsigned long		timestamp;
5706 };
5707 
5708 /*
5709  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5710  * standalone procedure for queueing an asynchronous RENEW.
5711  */
5712 static void nfs4_renew_release(void *calldata)
5713 {
5714 	struct nfs4_renewdata *data = calldata;
5715 	struct nfs_client *clp = data->client;
5716 
5717 	if (refcount_read(&clp->cl_count) > 1)
5718 		nfs4_schedule_state_renewal(clp);
5719 	nfs_put_client(clp);
5720 	kfree(data);
5721 }
5722 
5723 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5724 {
5725 	struct nfs4_renewdata *data = calldata;
5726 	struct nfs_client *clp = data->client;
5727 	unsigned long timestamp = data->timestamp;
5728 
5729 	trace_nfs4_renew_async(clp, task->tk_status);
5730 	switch (task->tk_status) {
5731 	case 0:
5732 		break;
5733 	case -NFS4ERR_LEASE_MOVED:
5734 		nfs4_schedule_lease_moved_recovery(clp);
5735 		break;
5736 	default:
5737 		/* Unless we're shutting down, schedule state recovery! */
5738 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5739 			return;
5740 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5741 			nfs4_schedule_lease_recovery(clp);
5742 			return;
5743 		}
5744 		nfs4_schedule_path_down_recovery(clp);
5745 	}
5746 	do_renew_lease(clp, timestamp);
5747 }
5748 
5749 static const struct rpc_call_ops nfs4_renew_ops = {
5750 	.rpc_call_done = nfs4_renew_done,
5751 	.rpc_release = nfs4_renew_release,
5752 };
5753 
5754 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5755 {
5756 	struct rpc_message msg = {
5757 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5758 		.rpc_argp	= clp,
5759 		.rpc_cred	= cred,
5760 	};
5761 	struct nfs4_renewdata *data;
5762 
5763 	if (renew_flags == 0)
5764 		return 0;
5765 	if (!refcount_inc_not_zero(&clp->cl_count))
5766 		return -EIO;
5767 	data = kmalloc(sizeof(*data), GFP_NOFS);
5768 	if (data == NULL) {
5769 		nfs_put_client(clp);
5770 		return -ENOMEM;
5771 	}
5772 	data->client = clp;
5773 	data->timestamp = jiffies;
5774 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5775 			&nfs4_renew_ops, data);
5776 }
5777 
5778 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5779 {
5780 	struct rpc_message msg = {
5781 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5782 		.rpc_argp	= clp,
5783 		.rpc_cred	= cred,
5784 	};
5785 	unsigned long now = jiffies;
5786 	int status;
5787 
5788 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5789 	if (status < 0)
5790 		return status;
5791 	do_renew_lease(clp, now);
5792 	return 0;
5793 }
5794 
5795 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5796 				      enum nfs4_acl_type type)
5797 {
5798 	switch (type) {
5799 	default:
5800 		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5801 	case NFS4ACL_DACL:
5802 		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5803 	case NFS4ACL_SACL:
5804 		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5805 	}
5806 }
5807 
5808 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5809  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5810  * the stack.
5811  */
5812 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5813 
5814 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5815 		struct page **pages)
5816 {
5817 	struct page *newpage, **spages;
5818 	int rc = 0;
5819 	size_t len;
5820 	spages = pages;
5821 
5822 	do {
5823 		len = min_t(size_t, PAGE_SIZE, buflen);
5824 		newpage = alloc_page(GFP_KERNEL);
5825 
5826 		if (newpage == NULL)
5827 			goto unwind;
5828 		memcpy(page_address(newpage), buf, len);
5829 		buf += len;
5830 		buflen -= len;
5831 		*pages++ = newpage;
5832 		rc++;
5833 	} while (buflen != 0);
5834 
5835 	return rc;
5836 
5837 unwind:
5838 	for(; rc > 0; rc--)
5839 		__free_page(spages[rc-1]);
5840 	return -ENOMEM;
5841 }
5842 
5843 struct nfs4_cached_acl {
5844 	enum nfs4_acl_type type;
5845 	int cached;
5846 	size_t len;
5847 	char data[];
5848 };
5849 
5850 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5851 {
5852 	struct nfs_inode *nfsi = NFS_I(inode);
5853 
5854 	spin_lock(&inode->i_lock);
5855 	kfree(nfsi->nfs4_acl);
5856 	nfsi->nfs4_acl = acl;
5857 	spin_unlock(&inode->i_lock);
5858 }
5859 
5860 static void nfs4_zap_acl_attr(struct inode *inode)
5861 {
5862 	nfs4_set_cached_acl(inode, NULL);
5863 }
5864 
5865 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5866 				    size_t buflen, enum nfs4_acl_type type)
5867 {
5868 	struct nfs_inode *nfsi = NFS_I(inode);
5869 	struct nfs4_cached_acl *acl;
5870 	int ret = -ENOENT;
5871 
5872 	spin_lock(&inode->i_lock);
5873 	acl = nfsi->nfs4_acl;
5874 	if (acl == NULL)
5875 		goto out;
5876 	if (acl->type != type)
5877 		goto out;
5878 	if (buf == NULL) /* user is just asking for length */
5879 		goto out_len;
5880 	if (acl->cached == 0)
5881 		goto out;
5882 	ret = -ERANGE; /* see getxattr(2) man page */
5883 	if (acl->len > buflen)
5884 		goto out;
5885 	memcpy(buf, acl->data, acl->len);
5886 out_len:
5887 	ret = acl->len;
5888 out:
5889 	spin_unlock(&inode->i_lock);
5890 	return ret;
5891 }
5892 
5893 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5894 				  size_t pgbase, size_t acl_len,
5895 				  enum nfs4_acl_type type)
5896 {
5897 	struct nfs4_cached_acl *acl;
5898 	size_t buflen = sizeof(*acl) + acl_len;
5899 
5900 	if (buflen <= PAGE_SIZE) {
5901 		acl = kmalloc(buflen, GFP_KERNEL);
5902 		if (acl == NULL)
5903 			goto out;
5904 		acl->cached = 1;
5905 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5906 	} else {
5907 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5908 		if (acl == NULL)
5909 			goto out;
5910 		acl->cached = 0;
5911 	}
5912 	acl->type = type;
5913 	acl->len = acl_len;
5914 out:
5915 	nfs4_set_cached_acl(inode, acl);
5916 }
5917 
5918 /*
5919  * The getxattr API returns the required buffer length when called with a
5920  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5921  * the required buf.  On a NULL buf, we send a page of data to the server
5922  * guessing that the ACL request can be serviced by a page. If so, we cache
5923  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5924  * the cache. If not so, we throw away the page, and cache the required
5925  * length. The next getxattr call will then produce another round trip to
5926  * the server, this time with the input buf of the required size.
5927  */
5928 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5929 				       size_t buflen, enum nfs4_acl_type type)
5930 {
5931 	struct page **pages;
5932 	struct nfs_getaclargs args = {
5933 		.fh = NFS_FH(inode),
5934 		.acl_type = type,
5935 		.acl_len = buflen,
5936 	};
5937 	struct nfs_getaclres res = {
5938 		.acl_type = type,
5939 		.acl_len = buflen,
5940 	};
5941 	struct rpc_message msg = {
5942 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5943 		.rpc_argp = &args,
5944 		.rpc_resp = &res,
5945 	};
5946 	unsigned int npages;
5947 	int ret = -ENOMEM, i;
5948 	struct nfs_server *server = NFS_SERVER(inode);
5949 
5950 	if (buflen == 0)
5951 		buflen = server->rsize;
5952 
5953 	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5954 	pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5955 	if (!pages)
5956 		return -ENOMEM;
5957 
5958 	args.acl_pages = pages;
5959 
5960 	for (i = 0; i < npages; i++) {
5961 		pages[i] = alloc_page(GFP_KERNEL);
5962 		if (!pages[i])
5963 			goto out_free;
5964 	}
5965 
5966 	/* for decoding across pages */
5967 	res.acl_scratch = alloc_page(GFP_KERNEL);
5968 	if (!res.acl_scratch)
5969 		goto out_free;
5970 
5971 	args.acl_len = npages * PAGE_SIZE;
5972 
5973 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5974 		__func__, buf, buflen, npages, args.acl_len);
5975 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5976 			     &msg, &args.seq_args, &res.seq_res, 0);
5977 	if (ret)
5978 		goto out_free;
5979 
5980 	/* Handle the case where the passed-in buffer is too short */
5981 	if (res.acl_flags & NFS4_ACL_TRUNC) {
5982 		/* Did the user only issue a request for the acl length? */
5983 		if (buf == NULL)
5984 			goto out_ok;
5985 		ret = -ERANGE;
5986 		goto out_free;
5987 	}
5988 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
5989 			      type);
5990 	if (buf) {
5991 		if (res.acl_len > buflen) {
5992 			ret = -ERANGE;
5993 			goto out_free;
5994 		}
5995 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5996 	}
5997 out_ok:
5998 	ret = res.acl_len;
5999 out_free:
6000 	while (--i >= 0)
6001 		__free_page(pages[i]);
6002 	if (res.acl_scratch)
6003 		__free_page(res.acl_scratch);
6004 	kfree(pages);
6005 	return ret;
6006 }
6007 
6008 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6009 				     size_t buflen, enum nfs4_acl_type type)
6010 {
6011 	struct nfs4_exception exception = {
6012 		.interruptible = true,
6013 	};
6014 	ssize_t ret;
6015 	do {
6016 		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6017 		trace_nfs4_get_acl(inode, ret);
6018 		if (ret >= 0)
6019 			break;
6020 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6021 	} while (exception.retry);
6022 	return ret;
6023 }
6024 
6025 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6026 				 enum nfs4_acl_type type)
6027 {
6028 	struct nfs_server *server = NFS_SERVER(inode);
6029 	int ret;
6030 
6031 	if (!nfs4_server_supports_acls(server, type))
6032 		return -EOPNOTSUPP;
6033 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6034 	if (ret < 0)
6035 		return ret;
6036 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6037 		nfs_zap_acl_cache(inode);
6038 	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6039 	if (ret != -ENOENT)
6040 		/* -ENOENT is returned if there is no ACL or if there is an ACL
6041 		 * but no cached acl data, just the acl length */
6042 		return ret;
6043 	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6044 }
6045 
6046 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6047 			       size_t buflen, enum nfs4_acl_type type)
6048 {
6049 	struct nfs_server *server = NFS_SERVER(inode);
6050 	struct page *pages[NFS4ACL_MAXPAGES];
6051 	struct nfs_setaclargs arg = {
6052 		.fh = NFS_FH(inode),
6053 		.acl_type = type,
6054 		.acl_len = buflen,
6055 		.acl_pages = pages,
6056 	};
6057 	struct nfs_setaclres res;
6058 	struct rpc_message msg = {
6059 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6060 		.rpc_argp	= &arg,
6061 		.rpc_resp	= &res,
6062 	};
6063 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6064 	int ret, i;
6065 
6066 	/* You can't remove system.nfs4_acl: */
6067 	if (buflen == 0)
6068 		return -EINVAL;
6069 	if (!nfs4_server_supports_acls(server, type))
6070 		return -EOPNOTSUPP;
6071 	if (npages > ARRAY_SIZE(pages))
6072 		return -ERANGE;
6073 	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6074 	if (i < 0)
6075 		return i;
6076 	nfs4_inode_make_writeable(inode);
6077 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6078 
6079 	/*
6080 	 * Free each page after tx, so the only ref left is
6081 	 * held by the network stack
6082 	 */
6083 	for (; i > 0; i--)
6084 		put_page(pages[i-1]);
6085 
6086 	/*
6087 	 * Acl update can result in inode attribute update.
6088 	 * so mark the attribute cache invalid.
6089 	 */
6090 	spin_lock(&inode->i_lock);
6091 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6092 					     NFS_INO_INVALID_CTIME |
6093 					     NFS_INO_REVAL_FORCED);
6094 	spin_unlock(&inode->i_lock);
6095 	nfs_access_zap_cache(inode);
6096 	nfs_zap_acl_cache(inode);
6097 	return ret;
6098 }
6099 
6100 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6101 			     size_t buflen, enum nfs4_acl_type type)
6102 {
6103 	struct nfs4_exception exception = { };
6104 	int err;
6105 	do {
6106 		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6107 		trace_nfs4_set_acl(inode, err);
6108 		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6109 			/*
6110 			 * no need to retry since the kernel
6111 			 * isn't involved in encoding the ACEs.
6112 			 */
6113 			err = -EINVAL;
6114 			break;
6115 		}
6116 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6117 				&exception);
6118 	} while (exception.retry);
6119 	return err;
6120 }
6121 
6122 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6123 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6124 					size_t buflen)
6125 {
6126 	struct nfs_server *server = NFS_SERVER(inode);
6127 	struct nfs4_label label = {0, 0, buflen, buf};
6128 
6129 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6130 	struct nfs_fattr fattr = {
6131 		.label = &label,
6132 	};
6133 	struct nfs4_getattr_arg arg = {
6134 		.fh		= NFS_FH(inode),
6135 		.bitmask	= bitmask,
6136 	};
6137 	struct nfs4_getattr_res res = {
6138 		.fattr		= &fattr,
6139 		.server		= server,
6140 	};
6141 	struct rpc_message msg = {
6142 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6143 		.rpc_argp	= &arg,
6144 		.rpc_resp	= &res,
6145 	};
6146 	int ret;
6147 
6148 	nfs_fattr_init(&fattr);
6149 
6150 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6151 	if (ret)
6152 		return ret;
6153 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6154 		return -ENOENT;
6155 	return label.len;
6156 }
6157 
6158 static int nfs4_get_security_label(struct inode *inode, void *buf,
6159 					size_t buflen)
6160 {
6161 	struct nfs4_exception exception = {
6162 		.interruptible = true,
6163 	};
6164 	int err;
6165 
6166 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6167 		return -EOPNOTSUPP;
6168 
6169 	do {
6170 		err = _nfs4_get_security_label(inode, buf, buflen);
6171 		trace_nfs4_get_security_label(inode, err);
6172 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6173 				&exception);
6174 	} while (exception.retry);
6175 	return err;
6176 }
6177 
6178 static int _nfs4_do_set_security_label(struct inode *inode,
6179 		struct nfs4_label *ilabel,
6180 		struct nfs_fattr *fattr)
6181 {
6182 
6183 	struct iattr sattr = {0};
6184 	struct nfs_server *server = NFS_SERVER(inode);
6185 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6186 	struct nfs_setattrargs arg = {
6187 		.fh		= NFS_FH(inode),
6188 		.iap		= &sattr,
6189 		.server		= server,
6190 		.bitmask	= bitmask,
6191 		.label		= ilabel,
6192 	};
6193 	struct nfs_setattrres res = {
6194 		.fattr		= fattr,
6195 		.server		= server,
6196 	};
6197 	struct rpc_message msg = {
6198 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6199 		.rpc_argp	= &arg,
6200 		.rpc_resp	= &res,
6201 	};
6202 	int status;
6203 
6204 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6205 
6206 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6207 	if (status)
6208 		dprintk("%s failed: %d\n", __func__, status);
6209 
6210 	return status;
6211 }
6212 
6213 static int nfs4_do_set_security_label(struct inode *inode,
6214 		struct nfs4_label *ilabel,
6215 		struct nfs_fattr *fattr)
6216 {
6217 	struct nfs4_exception exception = { };
6218 	int err;
6219 
6220 	do {
6221 		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6222 		trace_nfs4_set_security_label(inode, err);
6223 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6224 				&exception);
6225 	} while (exception.retry);
6226 	return err;
6227 }
6228 
6229 static int
6230 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6231 {
6232 	struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6233 	struct nfs_fattr *fattr;
6234 	int status;
6235 
6236 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6237 		return -EOPNOTSUPP;
6238 
6239 	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6240 	if (fattr == NULL)
6241 		return -ENOMEM;
6242 
6243 	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6244 	if (status == 0)
6245 		nfs_setsecurity(inode, fattr);
6246 
6247 	return status;
6248 }
6249 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6250 
6251 
6252 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6253 				    nfs4_verifier *bootverf)
6254 {
6255 	__be32 verf[2];
6256 
6257 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6258 		/* An impossible timestamp guarantees this value
6259 		 * will never match a generated boot time. */
6260 		verf[0] = cpu_to_be32(U32_MAX);
6261 		verf[1] = cpu_to_be32(U32_MAX);
6262 	} else {
6263 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6264 		u64 ns = ktime_to_ns(nn->boot_time);
6265 
6266 		verf[0] = cpu_to_be32(ns >> 32);
6267 		verf[1] = cpu_to_be32(ns);
6268 	}
6269 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6270 }
6271 
6272 static size_t
6273 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6274 {
6275 	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6276 	struct nfs_netns_client *nn_clp = nn->nfs_client;
6277 	const char *id;
6278 
6279 	buf[0] = '\0';
6280 
6281 	if (nn_clp) {
6282 		rcu_read_lock();
6283 		id = rcu_dereference(nn_clp->identifier);
6284 		if (id)
6285 			strscpy(buf, id, buflen);
6286 		rcu_read_unlock();
6287 	}
6288 
6289 	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6290 		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6291 
6292 	return strlen(buf);
6293 }
6294 
6295 static int
6296 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6297 {
6298 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6299 	size_t buflen;
6300 	size_t len;
6301 	char *str;
6302 
6303 	if (clp->cl_owner_id != NULL)
6304 		return 0;
6305 
6306 	rcu_read_lock();
6307 	len = 14 +
6308 		strlen(clp->cl_rpcclient->cl_nodename) +
6309 		1 +
6310 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6311 		1;
6312 	rcu_read_unlock();
6313 
6314 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6315 	if (buflen)
6316 		len += buflen + 1;
6317 
6318 	if (len > NFS4_OPAQUE_LIMIT + 1)
6319 		return -EINVAL;
6320 
6321 	/*
6322 	 * Since this string is allocated at mount time, and held until the
6323 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6324 	 * about a memory-reclaim deadlock.
6325 	 */
6326 	str = kmalloc(len, GFP_KERNEL);
6327 	if (!str)
6328 		return -ENOMEM;
6329 
6330 	rcu_read_lock();
6331 	if (buflen)
6332 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6333 			  clp->cl_rpcclient->cl_nodename, buf,
6334 			  rpc_peeraddr2str(clp->cl_rpcclient,
6335 					   RPC_DISPLAY_ADDR));
6336 	else
6337 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6338 			  clp->cl_rpcclient->cl_nodename,
6339 			  rpc_peeraddr2str(clp->cl_rpcclient,
6340 					   RPC_DISPLAY_ADDR));
6341 	rcu_read_unlock();
6342 
6343 	clp->cl_owner_id = str;
6344 	return 0;
6345 }
6346 
6347 static int
6348 nfs4_init_uniform_client_string(struct nfs_client *clp)
6349 {
6350 	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6351 	size_t buflen;
6352 	size_t len;
6353 	char *str;
6354 
6355 	if (clp->cl_owner_id != NULL)
6356 		return 0;
6357 
6358 	len = 10 + 10 + 1 + 10 + 1 +
6359 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6360 
6361 	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6362 	if (buflen)
6363 		len += buflen + 1;
6364 
6365 	if (len > NFS4_OPAQUE_LIMIT + 1)
6366 		return -EINVAL;
6367 
6368 	/*
6369 	 * Since this string is allocated at mount time, and held until the
6370 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6371 	 * about a memory-reclaim deadlock.
6372 	 */
6373 	str = kmalloc(len, GFP_KERNEL);
6374 	if (!str)
6375 		return -ENOMEM;
6376 
6377 	if (buflen)
6378 		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6379 			  clp->rpc_ops->version, clp->cl_minorversion,
6380 			  buf, clp->cl_rpcclient->cl_nodename);
6381 	else
6382 		scnprintf(str, len, "Linux NFSv%u.%u %s",
6383 			  clp->rpc_ops->version, clp->cl_minorversion,
6384 			  clp->cl_rpcclient->cl_nodename);
6385 	clp->cl_owner_id = str;
6386 	return 0;
6387 }
6388 
6389 /*
6390  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6391  * services.  Advertise one based on the address family of the
6392  * clientaddr.
6393  */
6394 static unsigned int
6395 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6396 {
6397 	if (strchr(clp->cl_ipaddr, ':') != NULL)
6398 		return scnprintf(buf, len, "tcp6");
6399 	else
6400 		return scnprintf(buf, len, "tcp");
6401 }
6402 
6403 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6404 {
6405 	struct nfs4_setclientid *sc = calldata;
6406 
6407 	if (task->tk_status == 0)
6408 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6409 }
6410 
6411 static const struct rpc_call_ops nfs4_setclientid_ops = {
6412 	.rpc_call_done = nfs4_setclientid_done,
6413 };
6414 
6415 /**
6416  * nfs4_proc_setclientid - Negotiate client ID
6417  * @clp: state data structure
6418  * @program: RPC program for NFSv4 callback service
6419  * @port: IP port number for NFS4 callback service
6420  * @cred: credential to use for this call
6421  * @res: where to place the result
6422  *
6423  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6424  */
6425 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6426 		unsigned short port, const struct cred *cred,
6427 		struct nfs4_setclientid_res *res)
6428 {
6429 	nfs4_verifier sc_verifier;
6430 	struct nfs4_setclientid setclientid = {
6431 		.sc_verifier = &sc_verifier,
6432 		.sc_prog = program,
6433 		.sc_clnt = clp,
6434 	};
6435 	struct rpc_message msg = {
6436 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6437 		.rpc_argp = &setclientid,
6438 		.rpc_resp = res,
6439 		.rpc_cred = cred,
6440 	};
6441 	struct rpc_task_setup task_setup_data = {
6442 		.rpc_client = clp->cl_rpcclient,
6443 		.rpc_message = &msg,
6444 		.callback_ops = &nfs4_setclientid_ops,
6445 		.callback_data = &setclientid,
6446 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6447 	};
6448 	unsigned long now = jiffies;
6449 	int status;
6450 
6451 	/* nfs_client_id4 */
6452 	nfs4_init_boot_verifier(clp, &sc_verifier);
6453 
6454 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6455 		status = nfs4_init_uniform_client_string(clp);
6456 	else
6457 		status = nfs4_init_nonuniform_client_string(clp);
6458 
6459 	if (status)
6460 		goto out;
6461 
6462 	/* cb_client4 */
6463 	setclientid.sc_netid_len =
6464 				nfs4_init_callback_netid(clp,
6465 						setclientid.sc_netid,
6466 						sizeof(setclientid.sc_netid));
6467 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6468 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6469 				clp->cl_ipaddr, port >> 8, port & 255);
6470 
6471 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6472 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6473 		clp->cl_owner_id);
6474 
6475 	status = nfs4_call_sync_custom(&task_setup_data);
6476 	if (setclientid.sc_cred) {
6477 		kfree(clp->cl_acceptor);
6478 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6479 		put_rpccred(setclientid.sc_cred);
6480 	}
6481 
6482 	if (status == 0)
6483 		do_renew_lease(clp, now);
6484 out:
6485 	trace_nfs4_setclientid(clp, status);
6486 	dprintk("NFS reply setclientid: %d\n", status);
6487 	return status;
6488 }
6489 
6490 /**
6491  * nfs4_proc_setclientid_confirm - Confirm client ID
6492  * @clp: state data structure
6493  * @arg: result of a previous SETCLIENTID
6494  * @cred: credential to use for this call
6495  *
6496  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6497  */
6498 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6499 		struct nfs4_setclientid_res *arg,
6500 		const struct cred *cred)
6501 {
6502 	struct rpc_message msg = {
6503 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6504 		.rpc_argp = arg,
6505 		.rpc_cred = cred,
6506 	};
6507 	int status;
6508 
6509 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6510 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6511 		clp->cl_clientid);
6512 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6513 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6514 	trace_nfs4_setclientid_confirm(clp, status);
6515 	dprintk("NFS reply setclientid_confirm: %d\n", status);
6516 	return status;
6517 }
6518 
6519 struct nfs4_delegreturndata {
6520 	struct nfs4_delegreturnargs args;
6521 	struct nfs4_delegreturnres res;
6522 	struct nfs_fh fh;
6523 	nfs4_stateid stateid;
6524 	unsigned long timestamp;
6525 	struct {
6526 		struct nfs4_layoutreturn_args arg;
6527 		struct nfs4_layoutreturn_res res;
6528 		struct nfs4_xdr_opaque_data ld_private;
6529 		u32 roc_barrier;
6530 		bool roc;
6531 	} lr;
6532 	struct nfs_fattr fattr;
6533 	int rpc_status;
6534 	struct inode *inode;
6535 };
6536 
6537 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6538 {
6539 	struct nfs4_delegreturndata *data = calldata;
6540 	struct nfs4_exception exception = {
6541 		.inode = data->inode,
6542 		.stateid = &data->stateid,
6543 		.task_is_privileged = data->args.seq_args.sa_privileged,
6544 	};
6545 
6546 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6547 		return;
6548 
6549 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6550 
6551 	/* Handle Layoutreturn errors */
6552 	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6553 			  &data->res.lr_ret) == -EAGAIN)
6554 		goto out_restart;
6555 
6556 	switch (task->tk_status) {
6557 	case 0:
6558 		renew_lease(data->res.server, data->timestamp);
6559 		break;
6560 	case -NFS4ERR_ADMIN_REVOKED:
6561 	case -NFS4ERR_DELEG_REVOKED:
6562 	case -NFS4ERR_EXPIRED:
6563 		nfs4_free_revoked_stateid(data->res.server,
6564 				data->args.stateid,
6565 				task->tk_msg.rpc_cred);
6566 		fallthrough;
6567 	case -NFS4ERR_BAD_STATEID:
6568 	case -NFS4ERR_STALE_STATEID:
6569 	case -ETIMEDOUT:
6570 		task->tk_status = 0;
6571 		break;
6572 	case -NFS4ERR_OLD_STATEID:
6573 		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6574 			nfs4_stateid_seqid_inc(&data->stateid);
6575 		if (data->args.bitmask) {
6576 			data->args.bitmask = NULL;
6577 			data->res.fattr = NULL;
6578 		}
6579 		goto out_restart;
6580 	case -NFS4ERR_ACCESS:
6581 		if (data->args.bitmask) {
6582 			data->args.bitmask = NULL;
6583 			data->res.fattr = NULL;
6584 			goto out_restart;
6585 		}
6586 		fallthrough;
6587 	default:
6588 		task->tk_status = nfs4_async_handle_exception(task,
6589 				data->res.server, task->tk_status,
6590 				&exception);
6591 		if (exception.retry)
6592 			goto out_restart;
6593 	}
6594 	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6595 	data->rpc_status = task->tk_status;
6596 	return;
6597 out_restart:
6598 	task->tk_status = 0;
6599 	rpc_restart_call_prepare(task);
6600 }
6601 
6602 static void nfs4_delegreturn_release(void *calldata)
6603 {
6604 	struct nfs4_delegreturndata *data = calldata;
6605 	struct inode *inode = data->inode;
6606 
6607 	if (data->lr.roc)
6608 		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6609 				 data->res.lr_ret);
6610 	if (inode) {
6611 		nfs4_fattr_set_prechange(&data->fattr,
6612 					 inode_peek_iversion_raw(inode));
6613 		nfs_refresh_inode(inode, &data->fattr);
6614 		nfs_iput_and_deactive(inode);
6615 	}
6616 	kfree(calldata);
6617 }
6618 
6619 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6620 {
6621 	struct nfs4_delegreturndata *d_data;
6622 	struct pnfs_layout_hdr *lo;
6623 
6624 	d_data = data;
6625 
6626 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6627 		nfs4_sequence_done(task, &d_data->res.seq_res);
6628 		return;
6629 	}
6630 
6631 	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6632 	if (lo && !pnfs_layout_is_valid(lo)) {
6633 		d_data->args.lr_args = NULL;
6634 		d_data->res.lr_res = NULL;
6635 	}
6636 
6637 	nfs4_setup_sequence(d_data->res.server->nfs_client,
6638 			&d_data->args.seq_args,
6639 			&d_data->res.seq_res,
6640 			task);
6641 }
6642 
6643 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6644 	.rpc_call_prepare = nfs4_delegreturn_prepare,
6645 	.rpc_call_done = nfs4_delegreturn_done,
6646 	.rpc_release = nfs4_delegreturn_release,
6647 };
6648 
6649 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6650 {
6651 	struct nfs4_delegreturndata *data;
6652 	struct nfs_server *server = NFS_SERVER(inode);
6653 	struct rpc_task *task;
6654 	struct rpc_message msg = {
6655 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6656 		.rpc_cred = cred,
6657 	};
6658 	struct rpc_task_setup task_setup_data = {
6659 		.rpc_client = server->client,
6660 		.rpc_message = &msg,
6661 		.callback_ops = &nfs4_delegreturn_ops,
6662 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6663 	};
6664 	int status = 0;
6665 
6666 	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6667 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6668 
6669 	data = kzalloc(sizeof(*data), GFP_KERNEL);
6670 	if (data == NULL)
6671 		return -ENOMEM;
6672 
6673 	nfs4_state_protect(server->nfs_client,
6674 			NFS_SP4_MACH_CRED_CLEANUP,
6675 			&task_setup_data.rpc_client, &msg);
6676 
6677 	data->args.fhandle = &data->fh;
6678 	data->args.stateid = &data->stateid;
6679 	nfs4_bitmask_set(data->args.bitmask_store,
6680 			 server->cache_consistency_bitmask, inode, 0);
6681 	data->args.bitmask = data->args.bitmask_store;
6682 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6683 	nfs4_stateid_copy(&data->stateid, stateid);
6684 	data->res.fattr = &data->fattr;
6685 	data->res.server = server;
6686 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6687 	data->lr.arg.ld_private = &data->lr.ld_private;
6688 	nfs_fattr_init(data->res.fattr);
6689 	data->timestamp = jiffies;
6690 	data->rpc_status = 0;
6691 	data->inode = nfs_igrab_and_active(inode);
6692 	if (data->inode || issync) {
6693 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6694 					cred);
6695 		if (data->lr.roc) {
6696 			data->args.lr_args = &data->lr.arg;
6697 			data->res.lr_res = &data->lr.res;
6698 		}
6699 	}
6700 
6701 	if (!data->inode)
6702 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6703 				   1);
6704 	else
6705 		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6706 				   0);
6707 	task_setup_data.callback_data = data;
6708 	msg.rpc_argp = &data->args;
6709 	msg.rpc_resp = &data->res;
6710 	task = rpc_run_task(&task_setup_data);
6711 	if (IS_ERR(task))
6712 		return PTR_ERR(task);
6713 	if (!issync)
6714 		goto out;
6715 	status = rpc_wait_for_completion_task(task);
6716 	if (status != 0)
6717 		goto out;
6718 	status = data->rpc_status;
6719 out:
6720 	rpc_put_task(task);
6721 	return status;
6722 }
6723 
6724 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6725 {
6726 	struct nfs_server *server = NFS_SERVER(inode);
6727 	struct nfs4_exception exception = { };
6728 	int err;
6729 	do {
6730 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6731 		trace_nfs4_delegreturn(inode, stateid, err);
6732 		switch (err) {
6733 			case -NFS4ERR_STALE_STATEID:
6734 			case -NFS4ERR_EXPIRED:
6735 			case 0:
6736 				return 0;
6737 		}
6738 		err = nfs4_handle_exception(server, err, &exception);
6739 	} while (exception.retry);
6740 	return err;
6741 }
6742 
6743 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6744 {
6745 	struct inode *inode = state->inode;
6746 	struct nfs_server *server = NFS_SERVER(inode);
6747 	struct nfs_client *clp = server->nfs_client;
6748 	struct nfs_lockt_args arg = {
6749 		.fh = NFS_FH(inode),
6750 		.fl = request,
6751 	};
6752 	struct nfs_lockt_res res = {
6753 		.denied = request,
6754 	};
6755 	struct rpc_message msg = {
6756 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6757 		.rpc_argp	= &arg,
6758 		.rpc_resp	= &res,
6759 		.rpc_cred	= state->owner->so_cred,
6760 	};
6761 	struct nfs4_lock_state *lsp;
6762 	int status;
6763 
6764 	arg.lock_owner.clientid = clp->cl_clientid;
6765 	status = nfs4_set_lock_state(state, request);
6766 	if (status != 0)
6767 		goto out;
6768 	lsp = request->fl_u.nfs4_fl.owner;
6769 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6770 	arg.lock_owner.s_dev = server->s_dev;
6771 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6772 	switch (status) {
6773 		case 0:
6774 			request->fl_type = F_UNLCK;
6775 			break;
6776 		case -NFS4ERR_DENIED:
6777 			status = 0;
6778 	}
6779 	request->fl_ops->fl_release_private(request);
6780 	request->fl_ops = NULL;
6781 out:
6782 	return status;
6783 }
6784 
6785 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6786 {
6787 	struct nfs4_exception exception = {
6788 		.interruptible = true,
6789 	};
6790 	int err;
6791 
6792 	do {
6793 		err = _nfs4_proc_getlk(state, cmd, request);
6794 		trace_nfs4_get_lock(request, state, cmd, err);
6795 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6796 				&exception);
6797 	} while (exception.retry);
6798 	return err;
6799 }
6800 
6801 /*
6802  * Update the seqid of a lock stateid after receiving
6803  * NFS4ERR_OLD_STATEID
6804  */
6805 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6806 		struct nfs4_lock_state *lsp)
6807 {
6808 	struct nfs4_state *state = lsp->ls_state;
6809 	bool ret = false;
6810 
6811 	spin_lock(&state->state_lock);
6812 	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6813 		goto out;
6814 	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6815 		nfs4_stateid_seqid_inc(dst);
6816 	else
6817 		dst->seqid = lsp->ls_stateid.seqid;
6818 	ret = true;
6819 out:
6820 	spin_unlock(&state->state_lock);
6821 	return ret;
6822 }
6823 
6824 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6825 		struct nfs4_lock_state *lsp)
6826 {
6827 	struct nfs4_state *state = lsp->ls_state;
6828 	bool ret;
6829 
6830 	spin_lock(&state->state_lock);
6831 	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6832 	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6833 	spin_unlock(&state->state_lock);
6834 	return ret;
6835 }
6836 
6837 struct nfs4_unlockdata {
6838 	struct nfs_locku_args arg;
6839 	struct nfs_locku_res res;
6840 	struct nfs4_lock_state *lsp;
6841 	struct nfs_open_context *ctx;
6842 	struct nfs_lock_context *l_ctx;
6843 	struct file_lock fl;
6844 	struct nfs_server *server;
6845 	unsigned long timestamp;
6846 };
6847 
6848 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6849 		struct nfs_open_context *ctx,
6850 		struct nfs4_lock_state *lsp,
6851 		struct nfs_seqid *seqid)
6852 {
6853 	struct nfs4_unlockdata *p;
6854 	struct nfs4_state *state = lsp->ls_state;
6855 	struct inode *inode = state->inode;
6856 
6857 	p = kzalloc(sizeof(*p), GFP_KERNEL);
6858 	if (p == NULL)
6859 		return NULL;
6860 	p->arg.fh = NFS_FH(inode);
6861 	p->arg.fl = &p->fl;
6862 	p->arg.seqid = seqid;
6863 	p->res.seqid = seqid;
6864 	p->lsp = lsp;
6865 	/* Ensure we don't close file until we're done freeing locks! */
6866 	p->ctx = get_nfs_open_context(ctx);
6867 	p->l_ctx = nfs_get_lock_context(ctx);
6868 	locks_init_lock(&p->fl);
6869 	locks_copy_lock(&p->fl, fl);
6870 	p->server = NFS_SERVER(inode);
6871 	spin_lock(&state->state_lock);
6872 	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6873 	spin_unlock(&state->state_lock);
6874 	return p;
6875 }
6876 
6877 static void nfs4_locku_release_calldata(void *data)
6878 {
6879 	struct nfs4_unlockdata *calldata = data;
6880 	nfs_free_seqid(calldata->arg.seqid);
6881 	nfs4_put_lock_state(calldata->lsp);
6882 	nfs_put_lock_context(calldata->l_ctx);
6883 	put_nfs_open_context(calldata->ctx);
6884 	kfree(calldata);
6885 }
6886 
6887 static void nfs4_locku_done(struct rpc_task *task, void *data)
6888 {
6889 	struct nfs4_unlockdata *calldata = data;
6890 	struct nfs4_exception exception = {
6891 		.inode = calldata->lsp->ls_state->inode,
6892 		.stateid = &calldata->arg.stateid,
6893 	};
6894 
6895 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6896 		return;
6897 	switch (task->tk_status) {
6898 		case 0:
6899 			renew_lease(calldata->server, calldata->timestamp);
6900 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6901 			if (nfs4_update_lock_stateid(calldata->lsp,
6902 					&calldata->res.stateid))
6903 				break;
6904 			fallthrough;
6905 		case -NFS4ERR_ADMIN_REVOKED:
6906 		case -NFS4ERR_EXPIRED:
6907 			nfs4_free_revoked_stateid(calldata->server,
6908 					&calldata->arg.stateid,
6909 					task->tk_msg.rpc_cred);
6910 			fallthrough;
6911 		case -NFS4ERR_BAD_STATEID:
6912 		case -NFS4ERR_STALE_STATEID:
6913 			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6914 						calldata->lsp))
6915 				rpc_restart_call_prepare(task);
6916 			break;
6917 		case -NFS4ERR_OLD_STATEID:
6918 			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6919 						calldata->lsp))
6920 				rpc_restart_call_prepare(task);
6921 			break;
6922 		default:
6923 			task->tk_status = nfs4_async_handle_exception(task,
6924 					calldata->server, task->tk_status,
6925 					&exception);
6926 			if (exception.retry)
6927 				rpc_restart_call_prepare(task);
6928 	}
6929 	nfs_release_seqid(calldata->arg.seqid);
6930 }
6931 
6932 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6933 {
6934 	struct nfs4_unlockdata *calldata = data;
6935 
6936 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6937 		nfs_async_iocounter_wait(task, calldata->l_ctx))
6938 		return;
6939 
6940 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6941 		goto out_wait;
6942 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6943 		/* Note: exit _without_ running nfs4_locku_done */
6944 		goto out_no_action;
6945 	}
6946 	calldata->timestamp = jiffies;
6947 	if (nfs4_setup_sequence(calldata->server->nfs_client,
6948 				&calldata->arg.seq_args,
6949 				&calldata->res.seq_res,
6950 				task) != 0)
6951 		nfs_release_seqid(calldata->arg.seqid);
6952 	return;
6953 out_no_action:
6954 	task->tk_action = NULL;
6955 out_wait:
6956 	nfs4_sequence_done(task, &calldata->res.seq_res);
6957 }
6958 
6959 static const struct rpc_call_ops nfs4_locku_ops = {
6960 	.rpc_call_prepare = nfs4_locku_prepare,
6961 	.rpc_call_done = nfs4_locku_done,
6962 	.rpc_release = nfs4_locku_release_calldata,
6963 };
6964 
6965 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6966 		struct nfs_open_context *ctx,
6967 		struct nfs4_lock_state *lsp,
6968 		struct nfs_seqid *seqid)
6969 {
6970 	struct nfs4_unlockdata *data;
6971 	struct rpc_message msg = {
6972 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6973 		.rpc_cred = ctx->cred,
6974 	};
6975 	struct rpc_task_setup task_setup_data = {
6976 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6977 		.rpc_message = &msg,
6978 		.callback_ops = &nfs4_locku_ops,
6979 		.workqueue = nfsiod_workqueue,
6980 		.flags = RPC_TASK_ASYNC,
6981 	};
6982 
6983 	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
6984 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6985 
6986 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6987 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6988 
6989 	/* Ensure this is an unlock - when canceling a lock, the
6990 	 * canceled lock is passed in, and it won't be an unlock.
6991 	 */
6992 	fl->fl_type = F_UNLCK;
6993 	if (fl->fl_flags & FL_CLOSE)
6994 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6995 
6996 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6997 	if (data == NULL) {
6998 		nfs_free_seqid(seqid);
6999 		return ERR_PTR(-ENOMEM);
7000 	}
7001 
7002 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
7003 	msg.rpc_argp = &data->arg;
7004 	msg.rpc_resp = &data->res;
7005 	task_setup_data.callback_data = data;
7006 	return rpc_run_task(&task_setup_data);
7007 }
7008 
7009 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7010 {
7011 	struct inode *inode = state->inode;
7012 	struct nfs4_state_owner *sp = state->owner;
7013 	struct nfs_inode *nfsi = NFS_I(inode);
7014 	struct nfs_seqid *seqid;
7015 	struct nfs4_lock_state *lsp;
7016 	struct rpc_task *task;
7017 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7018 	int status = 0;
7019 	unsigned char fl_flags = request->fl_flags;
7020 
7021 	status = nfs4_set_lock_state(state, request);
7022 	/* Unlock _before_ we do the RPC call */
7023 	request->fl_flags |= FL_EXISTS;
7024 	/* Exclude nfs_delegation_claim_locks() */
7025 	mutex_lock(&sp->so_delegreturn_mutex);
7026 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7027 	down_read(&nfsi->rwsem);
7028 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7029 		up_read(&nfsi->rwsem);
7030 		mutex_unlock(&sp->so_delegreturn_mutex);
7031 		goto out;
7032 	}
7033 	lsp = request->fl_u.nfs4_fl.owner;
7034 	set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7035 	up_read(&nfsi->rwsem);
7036 	mutex_unlock(&sp->so_delegreturn_mutex);
7037 	if (status != 0)
7038 		goto out;
7039 	/* Is this a delegated lock? */
7040 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7041 		goto out;
7042 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7043 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7044 	status = -ENOMEM;
7045 	if (IS_ERR(seqid))
7046 		goto out;
7047 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
7048 	status = PTR_ERR(task);
7049 	if (IS_ERR(task))
7050 		goto out;
7051 	status = rpc_wait_for_completion_task(task);
7052 	rpc_put_task(task);
7053 out:
7054 	request->fl_flags = fl_flags;
7055 	trace_nfs4_unlock(request, state, F_SETLK, status);
7056 	return status;
7057 }
7058 
7059 struct nfs4_lockdata {
7060 	struct nfs_lock_args arg;
7061 	struct nfs_lock_res res;
7062 	struct nfs4_lock_state *lsp;
7063 	struct nfs_open_context *ctx;
7064 	struct file_lock fl;
7065 	unsigned long timestamp;
7066 	int rpc_status;
7067 	int cancelled;
7068 	struct nfs_server *server;
7069 };
7070 
7071 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7072 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7073 		gfp_t gfp_mask)
7074 {
7075 	struct nfs4_lockdata *p;
7076 	struct inode *inode = lsp->ls_state->inode;
7077 	struct nfs_server *server = NFS_SERVER(inode);
7078 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7079 
7080 	p = kzalloc(sizeof(*p), gfp_mask);
7081 	if (p == NULL)
7082 		return NULL;
7083 
7084 	p->arg.fh = NFS_FH(inode);
7085 	p->arg.fl = &p->fl;
7086 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7087 	if (IS_ERR(p->arg.open_seqid))
7088 		goto out_free;
7089 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7090 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7091 	if (IS_ERR(p->arg.lock_seqid))
7092 		goto out_free_seqid;
7093 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7094 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7095 	p->arg.lock_owner.s_dev = server->s_dev;
7096 	p->res.lock_seqid = p->arg.lock_seqid;
7097 	p->lsp = lsp;
7098 	p->server = server;
7099 	p->ctx = get_nfs_open_context(ctx);
7100 	locks_init_lock(&p->fl);
7101 	locks_copy_lock(&p->fl, fl);
7102 	return p;
7103 out_free_seqid:
7104 	nfs_free_seqid(p->arg.open_seqid);
7105 out_free:
7106 	kfree(p);
7107 	return NULL;
7108 }
7109 
7110 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7111 {
7112 	struct nfs4_lockdata *data = calldata;
7113 	struct nfs4_state *state = data->lsp->ls_state;
7114 
7115 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7116 		goto out_wait;
7117 	/* Do we need to do an open_to_lock_owner? */
7118 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7119 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7120 			goto out_release_lock_seqid;
7121 		}
7122 		nfs4_stateid_copy(&data->arg.open_stateid,
7123 				&state->open_stateid);
7124 		data->arg.new_lock_owner = 1;
7125 		data->res.open_seqid = data->arg.open_seqid;
7126 	} else {
7127 		data->arg.new_lock_owner = 0;
7128 		nfs4_stateid_copy(&data->arg.lock_stateid,
7129 				&data->lsp->ls_stateid);
7130 	}
7131 	if (!nfs4_valid_open_stateid(state)) {
7132 		data->rpc_status = -EBADF;
7133 		task->tk_action = NULL;
7134 		goto out_release_open_seqid;
7135 	}
7136 	data->timestamp = jiffies;
7137 	if (nfs4_setup_sequence(data->server->nfs_client,
7138 				&data->arg.seq_args,
7139 				&data->res.seq_res,
7140 				task) == 0)
7141 		return;
7142 out_release_open_seqid:
7143 	nfs_release_seqid(data->arg.open_seqid);
7144 out_release_lock_seqid:
7145 	nfs_release_seqid(data->arg.lock_seqid);
7146 out_wait:
7147 	nfs4_sequence_done(task, &data->res.seq_res);
7148 	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7149 }
7150 
7151 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7152 {
7153 	struct nfs4_lockdata *data = calldata;
7154 	struct nfs4_lock_state *lsp = data->lsp;
7155 
7156 	if (!nfs4_sequence_done(task, &data->res.seq_res))
7157 		return;
7158 
7159 	data->rpc_status = task->tk_status;
7160 	switch (task->tk_status) {
7161 	case 0:
7162 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7163 				data->timestamp);
7164 		if (data->arg.new_lock && !data->cancelled) {
7165 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7166 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7167 				goto out_restart;
7168 		}
7169 		if (data->arg.new_lock_owner != 0) {
7170 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7171 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7172 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7173 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7174 			goto out_restart;
7175 		break;
7176 	case -NFS4ERR_OLD_STATEID:
7177 		if (data->arg.new_lock_owner != 0 &&
7178 			nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7179 					lsp->ls_state))
7180 			goto out_restart;
7181 		if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7182 			goto out_restart;
7183 		fallthrough;
7184 	case -NFS4ERR_BAD_STATEID:
7185 	case -NFS4ERR_STALE_STATEID:
7186 	case -NFS4ERR_EXPIRED:
7187 		if (data->arg.new_lock_owner != 0) {
7188 			if (!nfs4_stateid_match(&data->arg.open_stateid,
7189 						&lsp->ls_state->open_stateid))
7190 				goto out_restart;
7191 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7192 						&lsp->ls_stateid))
7193 				goto out_restart;
7194 	}
7195 out_done:
7196 	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7197 	return;
7198 out_restart:
7199 	if (!data->cancelled)
7200 		rpc_restart_call_prepare(task);
7201 	goto out_done;
7202 }
7203 
7204 static void nfs4_lock_release(void *calldata)
7205 {
7206 	struct nfs4_lockdata *data = calldata;
7207 
7208 	nfs_free_seqid(data->arg.open_seqid);
7209 	if (data->cancelled && data->rpc_status == 0) {
7210 		struct rpc_task *task;
7211 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7212 				data->arg.lock_seqid);
7213 		if (!IS_ERR(task))
7214 			rpc_put_task_async(task);
7215 		dprintk("%s: cancelling lock!\n", __func__);
7216 	} else
7217 		nfs_free_seqid(data->arg.lock_seqid);
7218 	nfs4_put_lock_state(data->lsp);
7219 	put_nfs_open_context(data->ctx);
7220 	kfree(data);
7221 }
7222 
7223 static const struct rpc_call_ops nfs4_lock_ops = {
7224 	.rpc_call_prepare = nfs4_lock_prepare,
7225 	.rpc_call_done = nfs4_lock_done,
7226 	.rpc_release = nfs4_lock_release,
7227 };
7228 
7229 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7230 {
7231 	switch (error) {
7232 	case -NFS4ERR_ADMIN_REVOKED:
7233 	case -NFS4ERR_EXPIRED:
7234 	case -NFS4ERR_BAD_STATEID:
7235 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7236 		if (new_lock_owner != 0 ||
7237 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7238 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7239 		break;
7240 	case -NFS4ERR_STALE_STATEID:
7241 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7242 		nfs4_schedule_lease_recovery(server->nfs_client);
7243 	}
7244 }
7245 
7246 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7247 {
7248 	struct nfs4_lockdata *data;
7249 	struct rpc_task *task;
7250 	struct rpc_message msg = {
7251 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7252 		.rpc_cred = state->owner->so_cred,
7253 	};
7254 	struct rpc_task_setup task_setup_data = {
7255 		.rpc_client = NFS_CLIENT(state->inode),
7256 		.rpc_message = &msg,
7257 		.callback_ops = &nfs4_lock_ops,
7258 		.workqueue = nfsiod_workqueue,
7259 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7260 	};
7261 	int ret;
7262 
7263 	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7264 		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7265 
7266 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7267 				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7268 	if (data == NULL)
7269 		return -ENOMEM;
7270 	if (IS_SETLKW(cmd))
7271 		data->arg.block = 1;
7272 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7273 				recovery_type > NFS_LOCK_NEW);
7274 	msg.rpc_argp = &data->arg;
7275 	msg.rpc_resp = &data->res;
7276 	task_setup_data.callback_data = data;
7277 	if (recovery_type > NFS_LOCK_NEW) {
7278 		if (recovery_type == NFS_LOCK_RECLAIM)
7279 			data->arg.reclaim = NFS_LOCK_RECLAIM;
7280 	} else
7281 		data->arg.new_lock = 1;
7282 	task = rpc_run_task(&task_setup_data);
7283 	if (IS_ERR(task))
7284 		return PTR_ERR(task);
7285 	ret = rpc_wait_for_completion_task(task);
7286 	if (ret == 0) {
7287 		ret = data->rpc_status;
7288 		if (ret)
7289 			nfs4_handle_setlk_error(data->server, data->lsp,
7290 					data->arg.new_lock_owner, ret);
7291 	} else
7292 		data->cancelled = true;
7293 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7294 	rpc_put_task(task);
7295 	dprintk("%s: ret = %d\n", __func__, ret);
7296 	return ret;
7297 }
7298 
7299 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7300 {
7301 	struct nfs_server *server = NFS_SERVER(state->inode);
7302 	struct nfs4_exception exception = {
7303 		.inode = state->inode,
7304 	};
7305 	int err;
7306 
7307 	do {
7308 		/* Cache the lock if possible... */
7309 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7310 			return 0;
7311 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7312 		if (err != -NFS4ERR_DELAY)
7313 			break;
7314 		nfs4_handle_exception(server, err, &exception);
7315 	} while (exception.retry);
7316 	return err;
7317 }
7318 
7319 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7320 {
7321 	struct nfs_server *server = NFS_SERVER(state->inode);
7322 	struct nfs4_exception exception = {
7323 		.inode = state->inode,
7324 	};
7325 	int err;
7326 
7327 	err = nfs4_set_lock_state(state, request);
7328 	if (err != 0)
7329 		return err;
7330 	if (!recover_lost_locks) {
7331 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7332 		return 0;
7333 	}
7334 	do {
7335 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7336 			return 0;
7337 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7338 		switch (err) {
7339 		default:
7340 			goto out;
7341 		case -NFS4ERR_GRACE:
7342 		case -NFS4ERR_DELAY:
7343 			nfs4_handle_exception(server, err, &exception);
7344 			err = 0;
7345 		}
7346 	} while (exception.retry);
7347 out:
7348 	return err;
7349 }
7350 
7351 #if defined(CONFIG_NFS_V4_1)
7352 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7353 {
7354 	struct nfs4_lock_state *lsp;
7355 	int status;
7356 
7357 	status = nfs4_set_lock_state(state, request);
7358 	if (status != 0)
7359 		return status;
7360 	lsp = request->fl_u.nfs4_fl.owner;
7361 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7362 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7363 		return 0;
7364 	return nfs4_lock_expired(state, request);
7365 }
7366 #endif
7367 
7368 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7369 {
7370 	struct nfs_inode *nfsi = NFS_I(state->inode);
7371 	struct nfs4_state_owner *sp = state->owner;
7372 	unsigned char fl_flags = request->fl_flags;
7373 	int status;
7374 
7375 	request->fl_flags |= FL_ACCESS;
7376 	status = locks_lock_inode_wait(state->inode, request);
7377 	if (status < 0)
7378 		goto out;
7379 	mutex_lock(&sp->so_delegreturn_mutex);
7380 	down_read(&nfsi->rwsem);
7381 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7382 		/* Yes: cache locks! */
7383 		/* ...but avoid races with delegation recall... */
7384 		request->fl_flags = fl_flags & ~FL_SLEEP;
7385 		status = locks_lock_inode_wait(state->inode, request);
7386 		up_read(&nfsi->rwsem);
7387 		mutex_unlock(&sp->so_delegreturn_mutex);
7388 		goto out;
7389 	}
7390 	up_read(&nfsi->rwsem);
7391 	mutex_unlock(&sp->so_delegreturn_mutex);
7392 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7393 out:
7394 	request->fl_flags = fl_flags;
7395 	return status;
7396 }
7397 
7398 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7399 {
7400 	struct nfs4_exception exception = {
7401 		.state = state,
7402 		.inode = state->inode,
7403 		.interruptible = true,
7404 	};
7405 	int err;
7406 
7407 	do {
7408 		err = _nfs4_proc_setlk(state, cmd, request);
7409 		if (err == -NFS4ERR_DENIED)
7410 			err = -EAGAIN;
7411 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7412 				err, &exception);
7413 	} while (exception.retry);
7414 	return err;
7415 }
7416 
7417 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7418 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7419 
7420 static int
7421 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7422 			struct file_lock *request)
7423 {
7424 	int		status = -ERESTARTSYS;
7425 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7426 
7427 	while(!signalled()) {
7428 		status = nfs4_proc_setlk(state, cmd, request);
7429 		if ((status != -EAGAIN) || IS_SETLK(cmd))
7430 			break;
7431 		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7432 		schedule_timeout(timeout);
7433 		timeout *= 2;
7434 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7435 		status = -ERESTARTSYS;
7436 	}
7437 	return status;
7438 }
7439 
7440 #ifdef CONFIG_NFS_V4_1
7441 struct nfs4_lock_waiter {
7442 	struct inode		*inode;
7443 	struct nfs_lowner	owner;
7444 	wait_queue_entry_t	wait;
7445 };
7446 
7447 static int
7448 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7449 {
7450 	struct nfs4_lock_waiter	*waiter	=
7451 		container_of(wait, struct nfs4_lock_waiter, wait);
7452 
7453 	/* NULL key means to wake up everyone */
7454 	if (key) {
7455 		struct cb_notify_lock_args	*cbnl = key;
7456 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7457 						*wowner = &waiter->owner;
7458 
7459 		/* Only wake if the callback was for the same owner. */
7460 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7461 			return 0;
7462 
7463 		/* Make sure it's for the right inode */
7464 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7465 			return 0;
7466 	}
7467 
7468 	return woken_wake_function(wait, mode, flags, key);
7469 }
7470 
7471 static int
7472 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7473 {
7474 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7475 	struct nfs_server *server = NFS_SERVER(state->inode);
7476 	struct nfs_client *clp = server->nfs_client;
7477 	wait_queue_head_t *q = &clp->cl_lock_waitq;
7478 	struct nfs4_lock_waiter waiter = {
7479 		.inode = state->inode,
7480 		.owner = { .clientid = clp->cl_clientid,
7481 			   .id = lsp->ls_seqid.owner_id,
7482 			   .s_dev = server->s_dev },
7483 	};
7484 	int status;
7485 
7486 	/* Don't bother with waitqueue if we don't expect a callback */
7487 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7488 		return nfs4_retry_setlk_simple(state, cmd, request);
7489 
7490 	init_wait(&waiter.wait);
7491 	waiter.wait.func = nfs4_wake_lock_waiter;
7492 	add_wait_queue(q, &waiter.wait);
7493 
7494 	do {
7495 		status = nfs4_proc_setlk(state, cmd, request);
7496 		if (status != -EAGAIN || IS_SETLK(cmd))
7497 			break;
7498 
7499 		status = -ERESTARTSYS;
7500 		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7501 			   NFS4_LOCK_MAXTIMEOUT);
7502 	} while (!signalled());
7503 
7504 	remove_wait_queue(q, &waiter.wait);
7505 
7506 	return status;
7507 }
7508 #else /* !CONFIG_NFS_V4_1 */
7509 static inline int
7510 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7511 {
7512 	return nfs4_retry_setlk_simple(state, cmd, request);
7513 }
7514 #endif
7515 
7516 static int
7517 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7518 {
7519 	struct nfs_open_context *ctx;
7520 	struct nfs4_state *state;
7521 	int status;
7522 
7523 	/* verify open state */
7524 	ctx = nfs_file_open_context(filp);
7525 	state = ctx->state;
7526 
7527 	if (IS_GETLK(cmd)) {
7528 		if (state != NULL)
7529 			return nfs4_proc_getlk(state, F_GETLK, request);
7530 		return 0;
7531 	}
7532 
7533 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7534 		return -EINVAL;
7535 
7536 	if (request->fl_type == F_UNLCK) {
7537 		if (state != NULL)
7538 			return nfs4_proc_unlck(state, cmd, request);
7539 		return 0;
7540 	}
7541 
7542 	if (state == NULL)
7543 		return -ENOLCK;
7544 
7545 	if ((request->fl_flags & FL_POSIX) &&
7546 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7547 		return -ENOLCK;
7548 
7549 	/*
7550 	 * Don't rely on the VFS having checked the file open mode,
7551 	 * since it won't do this for flock() locks.
7552 	 */
7553 	switch (request->fl_type) {
7554 	case F_RDLCK:
7555 		if (!(filp->f_mode & FMODE_READ))
7556 			return -EBADF;
7557 		break;
7558 	case F_WRLCK:
7559 		if (!(filp->f_mode & FMODE_WRITE))
7560 			return -EBADF;
7561 	}
7562 
7563 	status = nfs4_set_lock_state(state, request);
7564 	if (status != 0)
7565 		return status;
7566 
7567 	return nfs4_retry_setlk(state, cmd, request);
7568 }
7569 
7570 static int nfs4_delete_lease(struct file *file, void **priv)
7571 {
7572 	return generic_setlease(file, F_UNLCK, NULL, priv);
7573 }
7574 
7575 static int nfs4_add_lease(struct file *file, int arg, struct file_lock **lease,
7576 			  void **priv)
7577 {
7578 	struct inode *inode = file_inode(file);
7579 	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7580 	int ret;
7581 
7582 	/* No delegation, no lease */
7583 	if (!nfs4_have_delegation(inode, type))
7584 		return -EAGAIN;
7585 	ret = generic_setlease(file, arg, lease, priv);
7586 	if (ret || nfs4_have_delegation(inode, type))
7587 		return ret;
7588 	/* We raced with a delegation return */
7589 	nfs4_delete_lease(file, priv);
7590 	return -EAGAIN;
7591 }
7592 
7593 int nfs4_proc_setlease(struct file *file, int arg, struct file_lock **lease,
7594 		       void **priv)
7595 {
7596 	switch (arg) {
7597 	case F_RDLCK:
7598 	case F_WRLCK:
7599 		return nfs4_add_lease(file, arg, lease, priv);
7600 	case F_UNLCK:
7601 		return nfs4_delete_lease(file, priv);
7602 	default:
7603 		return -EINVAL;
7604 	}
7605 }
7606 
7607 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7608 {
7609 	struct nfs_server *server = NFS_SERVER(state->inode);
7610 	int err;
7611 
7612 	err = nfs4_set_lock_state(state, fl);
7613 	if (err != 0)
7614 		return err;
7615 	do {
7616 		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7617 		if (err != -NFS4ERR_DELAY)
7618 			break;
7619 		ssleep(1);
7620 	} while (err == -NFS4ERR_DELAY);
7621 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7622 }
7623 
7624 struct nfs_release_lockowner_data {
7625 	struct nfs4_lock_state *lsp;
7626 	struct nfs_server *server;
7627 	struct nfs_release_lockowner_args args;
7628 	struct nfs_release_lockowner_res res;
7629 	unsigned long timestamp;
7630 };
7631 
7632 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7633 {
7634 	struct nfs_release_lockowner_data *data = calldata;
7635 	struct nfs_server *server = data->server;
7636 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7637 			   &data->res.seq_res, task);
7638 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7639 	data->timestamp = jiffies;
7640 }
7641 
7642 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7643 {
7644 	struct nfs_release_lockowner_data *data = calldata;
7645 	struct nfs_server *server = data->server;
7646 
7647 	nfs40_sequence_done(task, &data->res.seq_res);
7648 
7649 	switch (task->tk_status) {
7650 	case 0:
7651 		renew_lease(server, data->timestamp);
7652 		break;
7653 	case -NFS4ERR_STALE_CLIENTID:
7654 	case -NFS4ERR_EXPIRED:
7655 		nfs4_schedule_lease_recovery(server->nfs_client);
7656 		break;
7657 	case -NFS4ERR_LEASE_MOVED:
7658 	case -NFS4ERR_DELAY:
7659 		if (nfs4_async_handle_error(task, server,
7660 					    NULL, NULL) == -EAGAIN)
7661 			rpc_restart_call_prepare(task);
7662 	}
7663 }
7664 
7665 static void nfs4_release_lockowner_release(void *calldata)
7666 {
7667 	struct nfs_release_lockowner_data *data = calldata;
7668 	nfs4_free_lock_state(data->server, data->lsp);
7669 	kfree(calldata);
7670 }
7671 
7672 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7673 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7674 	.rpc_call_done = nfs4_release_lockowner_done,
7675 	.rpc_release = nfs4_release_lockowner_release,
7676 };
7677 
7678 static void
7679 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7680 {
7681 	struct nfs_release_lockowner_data *data;
7682 	struct rpc_message msg = {
7683 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7684 	};
7685 
7686 	if (server->nfs_client->cl_mvops->minor_version != 0)
7687 		return;
7688 
7689 	data = kmalloc(sizeof(*data), GFP_KERNEL);
7690 	if (!data)
7691 		return;
7692 	data->lsp = lsp;
7693 	data->server = server;
7694 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7695 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7696 	data->args.lock_owner.s_dev = server->s_dev;
7697 
7698 	msg.rpc_argp = &data->args;
7699 	msg.rpc_resp = &data->res;
7700 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7701 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7702 }
7703 
7704 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7705 
7706 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7707 				   struct mnt_idmap *idmap,
7708 				   struct dentry *unused, struct inode *inode,
7709 				   const char *key, const void *buf,
7710 				   size_t buflen, int flags)
7711 {
7712 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7713 }
7714 
7715 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7716 				   struct dentry *unused, struct inode *inode,
7717 				   const char *key, void *buf, size_t buflen)
7718 {
7719 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7720 }
7721 
7722 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7723 {
7724 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7725 }
7726 
7727 #if defined(CONFIG_NFS_V4_1)
7728 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7729 
7730 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7731 				    struct mnt_idmap *idmap,
7732 				    struct dentry *unused, struct inode *inode,
7733 				    const char *key, const void *buf,
7734 				    size_t buflen, int flags)
7735 {
7736 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7737 }
7738 
7739 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7740 				    struct dentry *unused, struct inode *inode,
7741 				    const char *key, void *buf, size_t buflen)
7742 {
7743 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7744 }
7745 
7746 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7747 {
7748 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7749 }
7750 
7751 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7752 
7753 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7754 				    struct mnt_idmap *idmap,
7755 				    struct dentry *unused, struct inode *inode,
7756 				    const char *key, const void *buf,
7757 				    size_t buflen, int flags)
7758 {
7759 	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7760 }
7761 
7762 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7763 				    struct dentry *unused, struct inode *inode,
7764 				    const char *key, void *buf, size_t buflen)
7765 {
7766 	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7767 }
7768 
7769 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7770 {
7771 	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7772 }
7773 
7774 #endif
7775 
7776 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7777 
7778 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7779 				     struct mnt_idmap *idmap,
7780 				     struct dentry *unused, struct inode *inode,
7781 				     const char *key, const void *buf,
7782 				     size_t buflen, int flags)
7783 {
7784 	if (security_ismaclabel(key))
7785 		return nfs4_set_security_label(inode, buf, buflen);
7786 
7787 	return -EOPNOTSUPP;
7788 }
7789 
7790 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7791 				     struct dentry *unused, struct inode *inode,
7792 				     const char *key, void *buf, size_t buflen)
7793 {
7794 	if (security_ismaclabel(key))
7795 		return nfs4_get_security_label(inode, buf, buflen);
7796 	return -EOPNOTSUPP;
7797 }
7798 
7799 static ssize_t
7800 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7801 {
7802 	int len = 0;
7803 
7804 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7805 		len = security_inode_listsecurity(inode, list, list_len);
7806 		if (len >= 0 && list_len && len > list_len)
7807 			return -ERANGE;
7808 	}
7809 	return len;
7810 }
7811 
7812 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7813 	.prefix = XATTR_SECURITY_PREFIX,
7814 	.get	= nfs4_xattr_get_nfs4_label,
7815 	.set	= nfs4_xattr_set_nfs4_label,
7816 };
7817 
7818 #else
7819 
7820 static ssize_t
7821 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7822 {
7823 	return 0;
7824 }
7825 
7826 #endif
7827 
7828 #ifdef CONFIG_NFS_V4_2
7829 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7830 				    struct mnt_idmap *idmap,
7831 				    struct dentry *unused, struct inode *inode,
7832 				    const char *key, const void *buf,
7833 				    size_t buflen, int flags)
7834 {
7835 	u32 mask;
7836 	int ret;
7837 
7838 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7839 		return -EOPNOTSUPP;
7840 
7841 	/*
7842 	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7843 	 * flags right now. Handling of xattr operations use the normal
7844 	 * file read/write permissions.
7845 	 *
7846 	 * Just in case the server has other ideas (which RFC 8276 allows),
7847 	 * do a cached access check for the XA* flags to possibly avoid
7848 	 * doing an RPC and getting EACCES back.
7849 	 */
7850 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7851 		if (!(mask & NFS_ACCESS_XAWRITE))
7852 			return -EACCES;
7853 	}
7854 
7855 	if (buf == NULL) {
7856 		ret = nfs42_proc_removexattr(inode, key);
7857 		if (!ret)
7858 			nfs4_xattr_cache_remove(inode, key);
7859 	} else {
7860 		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7861 		if (!ret)
7862 			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7863 	}
7864 
7865 	return ret;
7866 }
7867 
7868 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7869 				    struct dentry *unused, struct inode *inode,
7870 				    const char *key, void *buf, size_t buflen)
7871 {
7872 	u32 mask;
7873 	ssize_t ret;
7874 
7875 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7876 		return -EOPNOTSUPP;
7877 
7878 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7879 		if (!(mask & NFS_ACCESS_XAREAD))
7880 			return -EACCES;
7881 	}
7882 
7883 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7884 	if (ret)
7885 		return ret;
7886 
7887 	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7888 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7889 		return ret;
7890 
7891 	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7892 
7893 	return ret;
7894 }
7895 
7896 static ssize_t
7897 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7898 {
7899 	u64 cookie;
7900 	bool eof;
7901 	ssize_t ret, size;
7902 	char *buf;
7903 	size_t buflen;
7904 	u32 mask;
7905 
7906 	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7907 		return 0;
7908 
7909 	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7910 		if (!(mask & NFS_ACCESS_XALIST))
7911 			return 0;
7912 	}
7913 
7914 	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7915 	if (ret)
7916 		return ret;
7917 
7918 	ret = nfs4_xattr_cache_list(inode, list, list_len);
7919 	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7920 		return ret;
7921 
7922 	cookie = 0;
7923 	eof = false;
7924 	buflen = list_len ? list_len : XATTR_LIST_MAX;
7925 	buf = list_len ? list : NULL;
7926 	size = 0;
7927 
7928 	while (!eof) {
7929 		ret = nfs42_proc_listxattrs(inode, buf, buflen,
7930 		    &cookie, &eof);
7931 		if (ret < 0)
7932 			return ret;
7933 
7934 		if (list_len) {
7935 			buf += ret;
7936 			buflen -= ret;
7937 		}
7938 		size += ret;
7939 	}
7940 
7941 	if (list_len)
7942 		nfs4_xattr_cache_set_list(inode, list, size);
7943 
7944 	return size;
7945 }
7946 
7947 #else
7948 
7949 static ssize_t
7950 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7951 {
7952 	return 0;
7953 }
7954 #endif /* CONFIG_NFS_V4_2 */
7955 
7956 /*
7957  * nfs_fhget will use either the mounted_on_fileid or the fileid
7958  */
7959 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7960 {
7961 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7962 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7963 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7964 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7965 		return;
7966 
7967 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7968 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7969 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7970 	fattr->nlink = 2;
7971 }
7972 
7973 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7974 				   const struct qstr *name,
7975 				   struct nfs4_fs_locations *fs_locations,
7976 				   struct page *page)
7977 {
7978 	struct nfs_server *server = NFS_SERVER(dir);
7979 	u32 bitmask[3];
7980 	struct nfs4_fs_locations_arg args = {
7981 		.dir_fh = NFS_FH(dir),
7982 		.name = name,
7983 		.page = page,
7984 		.bitmask = bitmask,
7985 	};
7986 	struct nfs4_fs_locations_res res = {
7987 		.fs_locations = fs_locations,
7988 	};
7989 	struct rpc_message msg = {
7990 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7991 		.rpc_argp = &args,
7992 		.rpc_resp = &res,
7993 	};
7994 	int status;
7995 
7996 	dprintk("%s: start\n", __func__);
7997 
7998 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7999 	bitmask[1] = nfs4_fattr_bitmap[1];
8000 
8001 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
8002 	 * is not supported */
8003 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
8004 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
8005 	else
8006 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8007 
8008 	nfs_fattr_init(fs_locations->fattr);
8009 	fs_locations->server = server;
8010 	fs_locations->nlocations = 0;
8011 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8012 	dprintk("%s: returned status = %d\n", __func__, status);
8013 	return status;
8014 }
8015 
8016 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8017 			   const struct qstr *name,
8018 			   struct nfs4_fs_locations *fs_locations,
8019 			   struct page *page)
8020 {
8021 	struct nfs4_exception exception = {
8022 		.interruptible = true,
8023 	};
8024 	int err;
8025 	do {
8026 		err = _nfs4_proc_fs_locations(client, dir, name,
8027 				fs_locations, page);
8028 		trace_nfs4_get_fs_locations(dir, name, err);
8029 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8030 				&exception);
8031 	} while (exception.retry);
8032 	return err;
8033 }
8034 
8035 /*
8036  * This operation also signals the server that this client is
8037  * performing migration recovery.  The server can stop returning
8038  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8039  * appended to this compound to identify the client ID which is
8040  * performing recovery.
8041  */
8042 static int _nfs40_proc_get_locations(struct nfs_server *server,
8043 				     struct nfs_fh *fhandle,
8044 				     struct nfs4_fs_locations *locations,
8045 				     struct page *page, const struct cred *cred)
8046 {
8047 	struct rpc_clnt *clnt = server->client;
8048 	u32 bitmask[2] = {
8049 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8050 	};
8051 	struct nfs4_fs_locations_arg args = {
8052 		.clientid	= server->nfs_client->cl_clientid,
8053 		.fh		= fhandle,
8054 		.page		= page,
8055 		.bitmask	= bitmask,
8056 		.migration	= 1,		/* skip LOOKUP */
8057 		.renew		= 1,		/* append RENEW */
8058 	};
8059 	struct nfs4_fs_locations_res res = {
8060 		.fs_locations	= locations,
8061 		.migration	= 1,
8062 		.renew		= 1,
8063 	};
8064 	struct rpc_message msg = {
8065 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8066 		.rpc_argp	= &args,
8067 		.rpc_resp	= &res,
8068 		.rpc_cred	= cred,
8069 	};
8070 	unsigned long now = jiffies;
8071 	int status;
8072 
8073 	nfs_fattr_init(locations->fattr);
8074 	locations->server = server;
8075 	locations->nlocations = 0;
8076 
8077 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8078 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8079 					&args.seq_args, &res.seq_res);
8080 	if (status)
8081 		return status;
8082 
8083 	renew_lease(server, now);
8084 	return 0;
8085 }
8086 
8087 #ifdef CONFIG_NFS_V4_1
8088 
8089 /*
8090  * This operation also signals the server that this client is
8091  * performing migration recovery.  The server can stop asserting
8092  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8093  * performing this operation is identified in the SEQUENCE
8094  * operation in this compound.
8095  *
8096  * When the client supports GETATTR(fs_locations_info), it can
8097  * be plumbed in here.
8098  */
8099 static int _nfs41_proc_get_locations(struct nfs_server *server,
8100 				     struct nfs_fh *fhandle,
8101 				     struct nfs4_fs_locations *locations,
8102 				     struct page *page, const struct cred *cred)
8103 {
8104 	struct rpc_clnt *clnt = server->client;
8105 	u32 bitmask[2] = {
8106 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8107 	};
8108 	struct nfs4_fs_locations_arg args = {
8109 		.fh		= fhandle,
8110 		.page		= page,
8111 		.bitmask	= bitmask,
8112 		.migration	= 1,		/* skip LOOKUP */
8113 	};
8114 	struct nfs4_fs_locations_res res = {
8115 		.fs_locations	= locations,
8116 		.migration	= 1,
8117 	};
8118 	struct rpc_message msg = {
8119 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8120 		.rpc_argp	= &args,
8121 		.rpc_resp	= &res,
8122 		.rpc_cred	= cred,
8123 	};
8124 	struct nfs4_call_sync_data data = {
8125 		.seq_server = server,
8126 		.seq_args = &args.seq_args,
8127 		.seq_res = &res.seq_res,
8128 	};
8129 	struct rpc_task_setup task_setup_data = {
8130 		.rpc_client = clnt,
8131 		.rpc_message = &msg,
8132 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8133 		.callback_data = &data,
8134 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8135 	};
8136 	int status;
8137 
8138 	nfs_fattr_init(locations->fattr);
8139 	locations->server = server;
8140 	locations->nlocations = 0;
8141 
8142 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8143 	status = nfs4_call_sync_custom(&task_setup_data);
8144 	if (status == NFS4_OK &&
8145 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8146 		status = -NFS4ERR_LEASE_MOVED;
8147 	return status;
8148 }
8149 
8150 #endif	/* CONFIG_NFS_V4_1 */
8151 
8152 /**
8153  * nfs4_proc_get_locations - discover locations for a migrated FSID
8154  * @server: pointer to nfs_server to process
8155  * @fhandle: pointer to the kernel NFS client file handle
8156  * @locations: result of query
8157  * @page: buffer
8158  * @cred: credential to use for this operation
8159  *
8160  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8161  * operation failed, or a negative errno if a local error occurred.
8162  *
8163  * On success, "locations" is filled in, but if the server has
8164  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8165  * asserted.
8166  *
8167  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8168  * from this client that require migration recovery.
8169  */
8170 int nfs4_proc_get_locations(struct nfs_server *server,
8171 			    struct nfs_fh *fhandle,
8172 			    struct nfs4_fs_locations *locations,
8173 			    struct page *page, const struct cred *cred)
8174 {
8175 	struct nfs_client *clp = server->nfs_client;
8176 	const struct nfs4_mig_recovery_ops *ops =
8177 					clp->cl_mvops->mig_recovery_ops;
8178 	struct nfs4_exception exception = {
8179 		.interruptible = true,
8180 	};
8181 	int status;
8182 
8183 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8184 		(unsigned long long)server->fsid.major,
8185 		(unsigned long long)server->fsid.minor,
8186 		clp->cl_hostname);
8187 	nfs_display_fhandle(fhandle, __func__);
8188 
8189 	do {
8190 		status = ops->get_locations(server, fhandle, locations, page,
8191 					    cred);
8192 		if (status != -NFS4ERR_DELAY)
8193 			break;
8194 		nfs4_handle_exception(server, status, &exception);
8195 	} while (exception.retry);
8196 	return status;
8197 }
8198 
8199 /*
8200  * This operation also signals the server that this client is
8201  * performing "lease moved" recovery.  The server can stop
8202  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8203  * is appended to this compound to identify the client ID which is
8204  * performing recovery.
8205  */
8206 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8207 {
8208 	struct nfs_server *server = NFS_SERVER(inode);
8209 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8210 	struct rpc_clnt *clnt = server->client;
8211 	struct nfs4_fsid_present_arg args = {
8212 		.fh		= NFS_FH(inode),
8213 		.clientid	= clp->cl_clientid,
8214 		.renew		= 1,		/* append RENEW */
8215 	};
8216 	struct nfs4_fsid_present_res res = {
8217 		.renew		= 1,
8218 	};
8219 	struct rpc_message msg = {
8220 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8221 		.rpc_argp	= &args,
8222 		.rpc_resp	= &res,
8223 		.rpc_cred	= cred,
8224 	};
8225 	unsigned long now = jiffies;
8226 	int status;
8227 
8228 	res.fh = nfs_alloc_fhandle();
8229 	if (res.fh == NULL)
8230 		return -ENOMEM;
8231 
8232 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8233 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8234 						&args.seq_args, &res.seq_res);
8235 	nfs_free_fhandle(res.fh);
8236 	if (status)
8237 		return status;
8238 
8239 	do_renew_lease(clp, now);
8240 	return 0;
8241 }
8242 
8243 #ifdef CONFIG_NFS_V4_1
8244 
8245 /*
8246  * This operation also signals the server that this client is
8247  * performing "lease moved" recovery.  The server can stop asserting
8248  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8249  * this operation is identified in the SEQUENCE operation in this
8250  * compound.
8251  */
8252 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8253 {
8254 	struct nfs_server *server = NFS_SERVER(inode);
8255 	struct rpc_clnt *clnt = server->client;
8256 	struct nfs4_fsid_present_arg args = {
8257 		.fh		= NFS_FH(inode),
8258 	};
8259 	struct nfs4_fsid_present_res res = {
8260 	};
8261 	struct rpc_message msg = {
8262 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8263 		.rpc_argp	= &args,
8264 		.rpc_resp	= &res,
8265 		.rpc_cred	= cred,
8266 	};
8267 	int status;
8268 
8269 	res.fh = nfs_alloc_fhandle();
8270 	if (res.fh == NULL)
8271 		return -ENOMEM;
8272 
8273 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8274 	status = nfs4_call_sync_sequence(clnt, server, &msg,
8275 						&args.seq_args, &res.seq_res);
8276 	nfs_free_fhandle(res.fh);
8277 	if (status == NFS4_OK &&
8278 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8279 		status = -NFS4ERR_LEASE_MOVED;
8280 	return status;
8281 }
8282 
8283 #endif	/* CONFIG_NFS_V4_1 */
8284 
8285 /**
8286  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8287  * @inode: inode on FSID to check
8288  * @cred: credential to use for this operation
8289  *
8290  * Server indicates whether the FSID is present, moved, or not
8291  * recognized.  This operation is necessary to clear a LEASE_MOVED
8292  * condition for this client ID.
8293  *
8294  * Returns NFS4_OK if the FSID is present on this server,
8295  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8296  *  NFS4ERR code if some error occurred on the server, or a
8297  *  negative errno if a local failure occurred.
8298  */
8299 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8300 {
8301 	struct nfs_server *server = NFS_SERVER(inode);
8302 	struct nfs_client *clp = server->nfs_client;
8303 	const struct nfs4_mig_recovery_ops *ops =
8304 					clp->cl_mvops->mig_recovery_ops;
8305 	struct nfs4_exception exception = {
8306 		.interruptible = true,
8307 	};
8308 	int status;
8309 
8310 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8311 		(unsigned long long)server->fsid.major,
8312 		(unsigned long long)server->fsid.minor,
8313 		clp->cl_hostname);
8314 	nfs_display_fhandle(NFS_FH(inode), __func__);
8315 
8316 	do {
8317 		status = ops->fsid_present(inode, cred);
8318 		if (status != -NFS4ERR_DELAY)
8319 			break;
8320 		nfs4_handle_exception(server, status, &exception);
8321 	} while (exception.retry);
8322 	return status;
8323 }
8324 
8325 /*
8326  * If 'use_integrity' is true and the state managment nfs_client
8327  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8328  * and the machine credential as per RFC3530bis and RFC5661 Security
8329  * Considerations sections. Otherwise, just use the user cred with the
8330  * filesystem's rpc_client.
8331  */
8332 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8333 {
8334 	int status;
8335 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8336 	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8337 	struct nfs4_secinfo_arg args = {
8338 		.dir_fh = NFS_FH(dir),
8339 		.name   = name,
8340 	};
8341 	struct nfs4_secinfo_res res = {
8342 		.flavors     = flavors,
8343 	};
8344 	struct rpc_message msg = {
8345 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8346 		.rpc_argp = &args,
8347 		.rpc_resp = &res,
8348 	};
8349 	struct nfs4_call_sync_data data = {
8350 		.seq_server = NFS_SERVER(dir),
8351 		.seq_args = &args.seq_args,
8352 		.seq_res = &res.seq_res,
8353 	};
8354 	struct rpc_task_setup task_setup = {
8355 		.rpc_client = clnt,
8356 		.rpc_message = &msg,
8357 		.callback_ops = clp->cl_mvops->call_sync_ops,
8358 		.callback_data = &data,
8359 		.flags = RPC_TASK_NO_ROUND_ROBIN,
8360 	};
8361 	const struct cred *cred = NULL;
8362 
8363 	if (use_integrity) {
8364 		clnt = clp->cl_rpcclient;
8365 		task_setup.rpc_client = clnt;
8366 
8367 		cred = nfs4_get_clid_cred(clp);
8368 		msg.rpc_cred = cred;
8369 	}
8370 
8371 	dprintk("NFS call  secinfo %s\n", name->name);
8372 
8373 	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8374 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8375 	status = nfs4_call_sync_custom(&task_setup);
8376 
8377 	dprintk("NFS reply  secinfo: %d\n", status);
8378 
8379 	put_cred(cred);
8380 	return status;
8381 }
8382 
8383 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8384 		      struct nfs4_secinfo_flavors *flavors)
8385 {
8386 	struct nfs4_exception exception = {
8387 		.interruptible = true,
8388 	};
8389 	int err;
8390 	do {
8391 		err = -NFS4ERR_WRONGSEC;
8392 
8393 		/* try to use integrity protection with machine cred */
8394 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8395 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8396 
8397 		/*
8398 		 * if unable to use integrity protection, or SECINFO with
8399 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8400 		 * disallowed by spec, but exists in deployed servers) use
8401 		 * the current filesystem's rpc_client and the user cred.
8402 		 */
8403 		if (err == -NFS4ERR_WRONGSEC)
8404 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8405 
8406 		trace_nfs4_secinfo(dir, name, err);
8407 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8408 				&exception);
8409 	} while (exception.retry);
8410 	return err;
8411 }
8412 
8413 #ifdef CONFIG_NFS_V4_1
8414 /*
8415  * Check the exchange flags returned by the server for invalid flags, having
8416  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8417  * DS flags set.
8418  */
8419 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8420 {
8421 	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8422 		goto out_inval;
8423 	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8424 		goto out_inval;
8425 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8426 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8427 		goto out_inval;
8428 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8429 		goto out_inval;
8430 	return NFS_OK;
8431 out_inval:
8432 	return -NFS4ERR_INVAL;
8433 }
8434 
8435 static bool
8436 nfs41_same_server_scope(struct nfs41_server_scope *a,
8437 			struct nfs41_server_scope *b)
8438 {
8439 	if (a->server_scope_sz != b->server_scope_sz)
8440 		return false;
8441 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8442 }
8443 
8444 static void
8445 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8446 {
8447 	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8448 	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8449 	struct nfs_client *clp = args->client;
8450 
8451 	switch (task->tk_status) {
8452 	case -NFS4ERR_BADSESSION:
8453 	case -NFS4ERR_DEADSESSION:
8454 		nfs4_schedule_session_recovery(clp->cl_session,
8455 				task->tk_status);
8456 		return;
8457 	}
8458 	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8459 			res->dir != NFS4_CDFS4_BOTH) {
8460 		rpc_task_close_connection(task);
8461 		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8462 			rpc_restart_call(task);
8463 	}
8464 }
8465 
8466 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8467 	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8468 };
8469 
8470 /*
8471  * nfs4_proc_bind_one_conn_to_session()
8472  *
8473  * The 4.1 client currently uses the same TCP connection for the
8474  * fore and backchannel.
8475  */
8476 static
8477 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8478 		struct rpc_xprt *xprt,
8479 		struct nfs_client *clp,
8480 		const struct cred *cred)
8481 {
8482 	int status;
8483 	struct nfs41_bind_conn_to_session_args args = {
8484 		.client = clp,
8485 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8486 		.retries = 0,
8487 	};
8488 	struct nfs41_bind_conn_to_session_res res;
8489 	struct rpc_message msg = {
8490 		.rpc_proc =
8491 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8492 		.rpc_argp = &args,
8493 		.rpc_resp = &res,
8494 		.rpc_cred = cred,
8495 	};
8496 	struct rpc_task_setup task_setup_data = {
8497 		.rpc_client = clnt,
8498 		.rpc_xprt = xprt,
8499 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8500 		.rpc_message = &msg,
8501 		.flags = RPC_TASK_TIMEOUT,
8502 	};
8503 	struct rpc_task *task;
8504 
8505 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8506 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8507 		args.dir = NFS4_CDFC4_FORE;
8508 
8509 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8510 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8511 		args.dir = NFS4_CDFC4_FORE;
8512 
8513 	task = rpc_run_task(&task_setup_data);
8514 	if (!IS_ERR(task)) {
8515 		status = task->tk_status;
8516 		rpc_put_task(task);
8517 	} else
8518 		status = PTR_ERR(task);
8519 	trace_nfs4_bind_conn_to_session(clp, status);
8520 	if (status == 0) {
8521 		if (memcmp(res.sessionid.data,
8522 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8523 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8524 			return -EIO;
8525 		}
8526 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8527 			dprintk("NFS: %s: Unexpected direction from server\n",
8528 				__func__);
8529 			return -EIO;
8530 		}
8531 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8532 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8533 				__func__);
8534 			return -EIO;
8535 		}
8536 	}
8537 
8538 	return status;
8539 }
8540 
8541 struct rpc_bind_conn_calldata {
8542 	struct nfs_client *clp;
8543 	const struct cred *cred;
8544 };
8545 
8546 static int
8547 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8548 		struct rpc_xprt *xprt,
8549 		void *calldata)
8550 {
8551 	struct rpc_bind_conn_calldata *p = calldata;
8552 
8553 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8554 }
8555 
8556 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8557 {
8558 	struct rpc_bind_conn_calldata data = {
8559 		.clp = clp,
8560 		.cred = cred,
8561 	};
8562 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8563 			nfs4_proc_bind_conn_to_session_callback, &data);
8564 }
8565 
8566 /*
8567  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8568  * and operations we'd like to see to enable certain features in the allow map
8569  */
8570 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8571 	.how = SP4_MACH_CRED,
8572 	.enforce.u.words = {
8573 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8574 		      1 << (OP_EXCHANGE_ID - 32) |
8575 		      1 << (OP_CREATE_SESSION - 32) |
8576 		      1 << (OP_DESTROY_SESSION - 32) |
8577 		      1 << (OP_DESTROY_CLIENTID - 32)
8578 	},
8579 	.allow.u.words = {
8580 		[0] = 1 << (OP_CLOSE) |
8581 		      1 << (OP_OPEN_DOWNGRADE) |
8582 		      1 << (OP_LOCKU) |
8583 		      1 << (OP_DELEGRETURN) |
8584 		      1 << (OP_COMMIT),
8585 		[1] = 1 << (OP_SECINFO - 32) |
8586 		      1 << (OP_SECINFO_NO_NAME - 32) |
8587 		      1 << (OP_LAYOUTRETURN - 32) |
8588 		      1 << (OP_TEST_STATEID - 32) |
8589 		      1 << (OP_FREE_STATEID - 32) |
8590 		      1 << (OP_WRITE - 32)
8591 	}
8592 };
8593 
8594 /*
8595  * Select the state protection mode for client `clp' given the server results
8596  * from exchange_id in `sp'.
8597  *
8598  * Returns 0 on success, negative errno otherwise.
8599  */
8600 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8601 				 struct nfs41_state_protection *sp)
8602 {
8603 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8604 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8605 		      1 << (OP_EXCHANGE_ID - 32) |
8606 		      1 << (OP_CREATE_SESSION - 32) |
8607 		      1 << (OP_DESTROY_SESSION - 32) |
8608 		      1 << (OP_DESTROY_CLIENTID - 32)
8609 	};
8610 	unsigned long flags = 0;
8611 	unsigned int i;
8612 	int ret = 0;
8613 
8614 	if (sp->how == SP4_MACH_CRED) {
8615 		/* Print state protect result */
8616 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8617 		for (i = 0; i <= LAST_NFS4_OP; i++) {
8618 			if (test_bit(i, sp->enforce.u.longs))
8619 				dfprintk(MOUNT, "  enforce op %d\n", i);
8620 			if (test_bit(i, sp->allow.u.longs))
8621 				dfprintk(MOUNT, "  allow op %d\n", i);
8622 		}
8623 
8624 		/* make sure nothing is on enforce list that isn't supported */
8625 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8626 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8627 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8628 				ret = -EINVAL;
8629 				goto out;
8630 			}
8631 		}
8632 
8633 		/*
8634 		 * Minimal mode - state operations are allowed to use machine
8635 		 * credential.  Note this already happens by default, so the
8636 		 * client doesn't have to do anything more than the negotiation.
8637 		 *
8638 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8639 		 *       we're already using the machine cred for exchange_id
8640 		 *       and will never use a different cred.
8641 		 */
8642 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8643 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8644 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8645 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8646 			dfprintk(MOUNT, "sp4_mach_cred:\n");
8647 			dfprintk(MOUNT, "  minimal mode enabled\n");
8648 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8649 		} else {
8650 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8651 			ret = -EINVAL;
8652 			goto out;
8653 		}
8654 
8655 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8656 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8657 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8658 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8659 			dfprintk(MOUNT, "  cleanup mode enabled\n");
8660 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8661 		}
8662 
8663 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8664 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8665 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8666 		}
8667 
8668 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8669 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8670 			dfprintk(MOUNT, "  secinfo mode enabled\n");
8671 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8672 		}
8673 
8674 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8675 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8676 			dfprintk(MOUNT, "  stateid mode enabled\n");
8677 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8678 		}
8679 
8680 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8681 			dfprintk(MOUNT, "  write mode enabled\n");
8682 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8683 		}
8684 
8685 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8686 			dfprintk(MOUNT, "  commit mode enabled\n");
8687 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8688 		}
8689 	}
8690 out:
8691 	clp->cl_sp4_flags = flags;
8692 	return ret;
8693 }
8694 
8695 struct nfs41_exchange_id_data {
8696 	struct nfs41_exchange_id_res res;
8697 	struct nfs41_exchange_id_args args;
8698 };
8699 
8700 static void nfs4_exchange_id_release(void *data)
8701 {
8702 	struct nfs41_exchange_id_data *cdata =
8703 					(struct nfs41_exchange_id_data *)data;
8704 
8705 	nfs_put_client(cdata->args.client);
8706 	kfree(cdata->res.impl_id);
8707 	kfree(cdata->res.server_scope);
8708 	kfree(cdata->res.server_owner);
8709 	kfree(cdata);
8710 }
8711 
8712 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8713 	.rpc_release = nfs4_exchange_id_release,
8714 };
8715 
8716 /*
8717  * _nfs4_proc_exchange_id()
8718  *
8719  * Wrapper for EXCHANGE_ID operation.
8720  */
8721 static struct rpc_task *
8722 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8723 			u32 sp4_how, struct rpc_xprt *xprt)
8724 {
8725 	struct rpc_message msg = {
8726 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8727 		.rpc_cred = cred,
8728 	};
8729 	struct rpc_task_setup task_setup_data = {
8730 		.rpc_client = clp->cl_rpcclient,
8731 		.callback_ops = &nfs4_exchange_id_call_ops,
8732 		.rpc_message = &msg,
8733 		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8734 	};
8735 	struct nfs41_exchange_id_data *calldata;
8736 	int status;
8737 
8738 	if (!refcount_inc_not_zero(&clp->cl_count))
8739 		return ERR_PTR(-EIO);
8740 
8741 	status = -ENOMEM;
8742 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8743 	if (!calldata)
8744 		goto out;
8745 
8746 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8747 
8748 	status = nfs4_init_uniform_client_string(clp);
8749 	if (status)
8750 		goto out_calldata;
8751 
8752 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8753 						GFP_NOFS);
8754 	status = -ENOMEM;
8755 	if (unlikely(calldata->res.server_owner == NULL))
8756 		goto out_calldata;
8757 
8758 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8759 					GFP_NOFS);
8760 	if (unlikely(calldata->res.server_scope == NULL))
8761 		goto out_server_owner;
8762 
8763 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8764 	if (unlikely(calldata->res.impl_id == NULL))
8765 		goto out_server_scope;
8766 
8767 	switch (sp4_how) {
8768 	case SP4_NONE:
8769 		calldata->args.state_protect.how = SP4_NONE;
8770 		break;
8771 
8772 	case SP4_MACH_CRED:
8773 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8774 		break;
8775 
8776 	default:
8777 		/* unsupported! */
8778 		WARN_ON_ONCE(1);
8779 		status = -EINVAL;
8780 		goto out_impl_id;
8781 	}
8782 	if (xprt) {
8783 		task_setup_data.rpc_xprt = xprt;
8784 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8785 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8786 				sizeof(calldata->args.verifier.data));
8787 	}
8788 	calldata->args.client = clp;
8789 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8790 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8791 #ifdef CONFIG_NFS_V4_1_MIGRATION
8792 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8793 #endif
8794 	if (test_bit(NFS_CS_DS, &clp->cl_flags))
8795 		calldata->args.flags |= EXCHGID4_FLAG_USE_PNFS_DS;
8796 	msg.rpc_argp = &calldata->args;
8797 	msg.rpc_resp = &calldata->res;
8798 	task_setup_data.callback_data = calldata;
8799 
8800 	return rpc_run_task(&task_setup_data);
8801 
8802 out_impl_id:
8803 	kfree(calldata->res.impl_id);
8804 out_server_scope:
8805 	kfree(calldata->res.server_scope);
8806 out_server_owner:
8807 	kfree(calldata->res.server_owner);
8808 out_calldata:
8809 	kfree(calldata);
8810 out:
8811 	nfs_put_client(clp);
8812 	return ERR_PTR(status);
8813 }
8814 
8815 /*
8816  * _nfs4_proc_exchange_id()
8817  *
8818  * Wrapper for EXCHANGE_ID operation.
8819  */
8820 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8821 			u32 sp4_how)
8822 {
8823 	struct rpc_task *task;
8824 	struct nfs41_exchange_id_args *argp;
8825 	struct nfs41_exchange_id_res *resp;
8826 	unsigned long now = jiffies;
8827 	int status;
8828 
8829 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8830 	if (IS_ERR(task))
8831 		return PTR_ERR(task);
8832 
8833 	argp = task->tk_msg.rpc_argp;
8834 	resp = task->tk_msg.rpc_resp;
8835 	status = task->tk_status;
8836 	if (status  != 0)
8837 		goto out;
8838 
8839 	status = nfs4_check_cl_exchange_flags(resp->flags,
8840 			clp->cl_mvops->minor_version);
8841 	if (status  != 0)
8842 		goto out;
8843 
8844 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8845 	if (status != 0)
8846 		goto out;
8847 
8848 	do_renew_lease(clp, now);
8849 
8850 	clp->cl_clientid = resp->clientid;
8851 	clp->cl_exchange_flags = resp->flags;
8852 	clp->cl_seqid = resp->seqid;
8853 	/* Client ID is not confirmed */
8854 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8855 		clear_bit(NFS4_SESSION_ESTABLISHED,
8856 			  &clp->cl_session->session_state);
8857 
8858 	if (clp->cl_serverscope != NULL &&
8859 	    !nfs41_same_server_scope(clp->cl_serverscope,
8860 				resp->server_scope)) {
8861 		dprintk("%s: server_scope mismatch detected\n",
8862 			__func__);
8863 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8864 	}
8865 
8866 	swap(clp->cl_serverowner, resp->server_owner);
8867 	swap(clp->cl_serverscope, resp->server_scope);
8868 	swap(clp->cl_implid, resp->impl_id);
8869 
8870 	/* Save the EXCHANGE_ID verifier session trunk tests */
8871 	memcpy(clp->cl_confirm.data, argp->verifier.data,
8872 	       sizeof(clp->cl_confirm.data));
8873 out:
8874 	trace_nfs4_exchange_id(clp, status);
8875 	rpc_put_task(task);
8876 	return status;
8877 }
8878 
8879 /*
8880  * nfs4_proc_exchange_id()
8881  *
8882  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8883  *
8884  * Since the clientid has expired, all compounds using sessions
8885  * associated with the stale clientid will be returning
8886  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8887  * be in some phase of session reset.
8888  *
8889  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8890  */
8891 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8892 {
8893 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8894 	int status;
8895 
8896 	/* try SP4_MACH_CRED if krb5i/p	*/
8897 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8898 	    authflavor == RPC_AUTH_GSS_KRB5P) {
8899 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8900 		if (!status)
8901 			return 0;
8902 	}
8903 
8904 	/* try SP4_NONE */
8905 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8906 }
8907 
8908 /**
8909  * nfs4_test_session_trunk
8910  *
8911  * This is an add_xprt_test() test function called from
8912  * rpc_clnt_setup_test_and_add_xprt.
8913  *
8914  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8915  * and is dereferrenced in nfs4_exchange_id_release
8916  *
8917  * Upon success, add the new transport to the rpc_clnt
8918  *
8919  * @clnt: struct rpc_clnt to get new transport
8920  * @xprt: the rpc_xprt to test
8921  * @data: call data for _nfs4_proc_exchange_id.
8922  */
8923 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8924 			    void *data)
8925 {
8926 	struct nfs4_add_xprt_data *adata = data;
8927 	struct rpc_task *task;
8928 	int status;
8929 
8930 	u32 sp4_how;
8931 
8932 	dprintk("--> %s try %s\n", __func__,
8933 		xprt->address_strings[RPC_DISPLAY_ADDR]);
8934 
8935 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8936 
8937 	/* Test connection for session trunking. Async exchange_id call */
8938 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8939 	if (IS_ERR(task))
8940 		return;
8941 
8942 	status = task->tk_status;
8943 	if (status == 0)
8944 		status = nfs4_detect_session_trunking(adata->clp,
8945 				task->tk_msg.rpc_resp, xprt);
8946 
8947 	if (status == 0)
8948 		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8949 	else if (rpc_clnt_xprt_switch_has_addr(clnt,
8950 				(struct sockaddr *)&xprt->addr))
8951 		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8952 
8953 	rpc_put_task(task);
8954 }
8955 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8956 
8957 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8958 		const struct cred *cred)
8959 {
8960 	struct rpc_message msg = {
8961 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8962 		.rpc_argp = clp,
8963 		.rpc_cred = cred,
8964 	};
8965 	int status;
8966 
8967 	status = rpc_call_sync(clp->cl_rpcclient, &msg,
8968 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8969 	trace_nfs4_destroy_clientid(clp, status);
8970 	if (status)
8971 		dprintk("NFS: Got error %d from the server %s on "
8972 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
8973 	return status;
8974 }
8975 
8976 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8977 		const struct cred *cred)
8978 {
8979 	unsigned int loop;
8980 	int ret;
8981 
8982 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8983 		ret = _nfs4_proc_destroy_clientid(clp, cred);
8984 		switch (ret) {
8985 		case -NFS4ERR_DELAY:
8986 		case -NFS4ERR_CLIENTID_BUSY:
8987 			ssleep(1);
8988 			break;
8989 		default:
8990 			return ret;
8991 		}
8992 	}
8993 	return 0;
8994 }
8995 
8996 int nfs4_destroy_clientid(struct nfs_client *clp)
8997 {
8998 	const struct cred *cred;
8999 	int ret = 0;
9000 
9001 	if (clp->cl_mvops->minor_version < 1)
9002 		goto out;
9003 	if (clp->cl_exchange_flags == 0)
9004 		goto out;
9005 	if (clp->cl_preserve_clid)
9006 		goto out;
9007 	cred = nfs4_get_clid_cred(clp);
9008 	ret = nfs4_proc_destroy_clientid(clp, cred);
9009 	put_cred(cred);
9010 	switch (ret) {
9011 	case 0:
9012 	case -NFS4ERR_STALE_CLIENTID:
9013 		clp->cl_exchange_flags = 0;
9014 	}
9015 out:
9016 	return ret;
9017 }
9018 
9019 #endif /* CONFIG_NFS_V4_1 */
9020 
9021 struct nfs4_get_lease_time_data {
9022 	struct nfs4_get_lease_time_args *args;
9023 	struct nfs4_get_lease_time_res *res;
9024 	struct nfs_client *clp;
9025 };
9026 
9027 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9028 					void *calldata)
9029 {
9030 	struct nfs4_get_lease_time_data *data =
9031 			(struct nfs4_get_lease_time_data *)calldata;
9032 
9033 	/* just setup sequence, do not trigger session recovery
9034 	   since we're invoked within one */
9035 	nfs4_setup_sequence(data->clp,
9036 			&data->args->la_seq_args,
9037 			&data->res->lr_seq_res,
9038 			task);
9039 }
9040 
9041 /*
9042  * Called from nfs4_state_manager thread for session setup, so don't recover
9043  * from sequence operation or clientid errors.
9044  */
9045 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9046 {
9047 	struct nfs4_get_lease_time_data *data =
9048 			(struct nfs4_get_lease_time_data *)calldata;
9049 
9050 	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9051 		return;
9052 	switch (task->tk_status) {
9053 	case -NFS4ERR_DELAY:
9054 	case -NFS4ERR_GRACE:
9055 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9056 		task->tk_status = 0;
9057 		fallthrough;
9058 	case -NFS4ERR_RETRY_UNCACHED_REP:
9059 		rpc_restart_call_prepare(task);
9060 		return;
9061 	}
9062 }
9063 
9064 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9065 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9066 	.rpc_call_done = nfs4_get_lease_time_done,
9067 };
9068 
9069 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9070 {
9071 	struct nfs4_get_lease_time_args args;
9072 	struct nfs4_get_lease_time_res res = {
9073 		.lr_fsinfo = fsinfo,
9074 	};
9075 	struct nfs4_get_lease_time_data data = {
9076 		.args = &args,
9077 		.res = &res,
9078 		.clp = clp,
9079 	};
9080 	struct rpc_message msg = {
9081 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9082 		.rpc_argp = &args,
9083 		.rpc_resp = &res,
9084 	};
9085 	struct rpc_task_setup task_setup = {
9086 		.rpc_client = clp->cl_rpcclient,
9087 		.rpc_message = &msg,
9088 		.callback_ops = &nfs4_get_lease_time_ops,
9089 		.callback_data = &data,
9090 		.flags = RPC_TASK_TIMEOUT,
9091 	};
9092 
9093 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9094 	return nfs4_call_sync_custom(&task_setup);
9095 }
9096 
9097 #ifdef CONFIG_NFS_V4_1
9098 
9099 /*
9100  * Initialize the values to be used by the client in CREATE_SESSION
9101  * If nfs4_init_session set the fore channel request and response sizes,
9102  * use them.
9103  *
9104  * Set the back channel max_resp_sz_cached to zero to force the client to
9105  * always set csa_cachethis to FALSE because the current implementation
9106  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9107  */
9108 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9109 				    struct rpc_clnt *clnt)
9110 {
9111 	unsigned int max_rqst_sz, max_resp_sz;
9112 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9113 	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9114 
9115 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9116 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9117 
9118 	/* Fore channel attributes */
9119 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9120 	args->fc_attrs.max_resp_sz = max_resp_sz;
9121 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9122 	args->fc_attrs.max_reqs = max_session_slots;
9123 
9124 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9125 		"max_ops=%u max_reqs=%u\n",
9126 		__func__,
9127 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9128 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9129 
9130 	/* Back channel attributes */
9131 	args->bc_attrs.max_rqst_sz = max_bc_payload;
9132 	args->bc_attrs.max_resp_sz = max_bc_payload;
9133 	args->bc_attrs.max_resp_sz_cached = 0;
9134 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9135 	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9136 	if (args->bc_attrs.max_reqs > max_bc_slots)
9137 		args->bc_attrs.max_reqs = max_bc_slots;
9138 
9139 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9140 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9141 		__func__,
9142 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9143 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9144 		args->bc_attrs.max_reqs);
9145 }
9146 
9147 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9148 		struct nfs41_create_session_res *res)
9149 {
9150 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9151 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9152 
9153 	if (rcvd->max_resp_sz > sent->max_resp_sz)
9154 		return -EINVAL;
9155 	/*
9156 	 * Our requested max_ops is the minimum we need; we're not
9157 	 * prepared to break up compounds into smaller pieces than that.
9158 	 * So, no point even trying to continue if the server won't
9159 	 * cooperate:
9160 	 */
9161 	if (rcvd->max_ops < sent->max_ops)
9162 		return -EINVAL;
9163 	if (rcvd->max_reqs == 0)
9164 		return -EINVAL;
9165 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9166 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9167 	return 0;
9168 }
9169 
9170 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9171 		struct nfs41_create_session_res *res)
9172 {
9173 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9174 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9175 
9176 	if (!(res->flags & SESSION4_BACK_CHAN))
9177 		goto out;
9178 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9179 		return -EINVAL;
9180 	if (rcvd->max_resp_sz < sent->max_resp_sz)
9181 		return -EINVAL;
9182 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9183 		return -EINVAL;
9184 	if (rcvd->max_ops > sent->max_ops)
9185 		return -EINVAL;
9186 	if (rcvd->max_reqs > sent->max_reqs)
9187 		return -EINVAL;
9188 out:
9189 	return 0;
9190 }
9191 
9192 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9193 				     struct nfs41_create_session_res *res)
9194 {
9195 	int ret;
9196 
9197 	ret = nfs4_verify_fore_channel_attrs(args, res);
9198 	if (ret)
9199 		return ret;
9200 	return nfs4_verify_back_channel_attrs(args, res);
9201 }
9202 
9203 static void nfs4_update_session(struct nfs4_session *session,
9204 		struct nfs41_create_session_res *res)
9205 {
9206 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9207 	/* Mark client id and session as being confirmed */
9208 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9209 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9210 	session->flags = res->flags;
9211 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9212 	if (res->flags & SESSION4_BACK_CHAN)
9213 		memcpy(&session->bc_attrs, &res->bc_attrs,
9214 				sizeof(session->bc_attrs));
9215 }
9216 
9217 static int _nfs4_proc_create_session(struct nfs_client *clp,
9218 		const struct cred *cred)
9219 {
9220 	struct nfs4_session *session = clp->cl_session;
9221 	struct nfs41_create_session_args args = {
9222 		.client = clp,
9223 		.clientid = clp->cl_clientid,
9224 		.seqid = clp->cl_seqid,
9225 		.cb_program = NFS4_CALLBACK,
9226 	};
9227 	struct nfs41_create_session_res res;
9228 
9229 	struct rpc_message msg = {
9230 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9231 		.rpc_argp = &args,
9232 		.rpc_resp = &res,
9233 		.rpc_cred = cred,
9234 	};
9235 	int status;
9236 
9237 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9238 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9239 
9240 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9241 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9242 	trace_nfs4_create_session(clp, status);
9243 
9244 	switch (status) {
9245 	case -NFS4ERR_STALE_CLIENTID:
9246 	case -NFS4ERR_DELAY:
9247 	case -ETIMEDOUT:
9248 	case -EACCES:
9249 	case -EAGAIN:
9250 		goto out;
9251 	}
9252 
9253 	clp->cl_seqid++;
9254 	if (!status) {
9255 		/* Verify the session's negotiated channel_attrs values */
9256 		status = nfs4_verify_channel_attrs(&args, &res);
9257 		/* Increment the clientid slot sequence id */
9258 		if (status)
9259 			goto out;
9260 		nfs4_update_session(session, &res);
9261 	}
9262 out:
9263 	return status;
9264 }
9265 
9266 /*
9267  * Issues a CREATE_SESSION operation to the server.
9268  * It is the responsibility of the caller to verify the session is
9269  * expired before calling this routine.
9270  */
9271 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9272 {
9273 	int status;
9274 	unsigned *ptr;
9275 	struct nfs4_session *session = clp->cl_session;
9276 	struct nfs4_add_xprt_data xprtdata = {
9277 		.clp = clp,
9278 	};
9279 	struct rpc_add_xprt_test rpcdata = {
9280 		.add_xprt_test = clp->cl_mvops->session_trunk,
9281 		.data = &xprtdata,
9282 	};
9283 
9284 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9285 
9286 	status = _nfs4_proc_create_session(clp, cred);
9287 	if (status)
9288 		goto out;
9289 
9290 	/* Init or reset the session slot tables */
9291 	status = nfs4_setup_session_slot_tables(session);
9292 	dprintk("slot table setup returned %d\n", status);
9293 	if (status)
9294 		goto out;
9295 
9296 	ptr = (unsigned *)&session->sess_id.data[0];
9297 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9298 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9299 	rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9300 out:
9301 	return status;
9302 }
9303 
9304 /*
9305  * Issue the over-the-wire RPC DESTROY_SESSION.
9306  * The caller must serialize access to this routine.
9307  */
9308 int nfs4_proc_destroy_session(struct nfs4_session *session,
9309 		const struct cred *cred)
9310 {
9311 	struct rpc_message msg = {
9312 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9313 		.rpc_argp = session,
9314 		.rpc_cred = cred,
9315 	};
9316 	int status = 0;
9317 
9318 	/* session is still being setup */
9319 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9320 		return 0;
9321 
9322 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9323 			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9324 	trace_nfs4_destroy_session(session->clp, status);
9325 
9326 	if (status)
9327 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9328 			"Session has been destroyed regardless...\n", status);
9329 	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9330 	return status;
9331 }
9332 
9333 /*
9334  * Renew the cl_session lease.
9335  */
9336 struct nfs4_sequence_data {
9337 	struct nfs_client *clp;
9338 	struct nfs4_sequence_args args;
9339 	struct nfs4_sequence_res res;
9340 };
9341 
9342 static void nfs41_sequence_release(void *data)
9343 {
9344 	struct nfs4_sequence_data *calldata = data;
9345 	struct nfs_client *clp = calldata->clp;
9346 
9347 	if (refcount_read(&clp->cl_count) > 1)
9348 		nfs4_schedule_state_renewal(clp);
9349 	nfs_put_client(clp);
9350 	kfree(calldata);
9351 }
9352 
9353 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9354 {
9355 	switch(task->tk_status) {
9356 	case -NFS4ERR_DELAY:
9357 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9358 		return -EAGAIN;
9359 	default:
9360 		nfs4_schedule_lease_recovery(clp);
9361 	}
9362 	return 0;
9363 }
9364 
9365 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9366 {
9367 	struct nfs4_sequence_data *calldata = data;
9368 	struct nfs_client *clp = calldata->clp;
9369 
9370 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9371 		return;
9372 
9373 	trace_nfs4_sequence(clp, task->tk_status);
9374 	if (task->tk_status < 0 && !task->tk_client->cl_shutdown) {
9375 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9376 		if (refcount_read(&clp->cl_count) == 1)
9377 			return;
9378 
9379 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9380 			rpc_restart_call_prepare(task);
9381 			return;
9382 		}
9383 	}
9384 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9385 }
9386 
9387 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9388 {
9389 	struct nfs4_sequence_data *calldata = data;
9390 	struct nfs_client *clp = calldata->clp;
9391 	struct nfs4_sequence_args *args;
9392 	struct nfs4_sequence_res *res;
9393 
9394 	args = task->tk_msg.rpc_argp;
9395 	res = task->tk_msg.rpc_resp;
9396 
9397 	nfs4_setup_sequence(clp, args, res, task);
9398 }
9399 
9400 static const struct rpc_call_ops nfs41_sequence_ops = {
9401 	.rpc_call_done = nfs41_sequence_call_done,
9402 	.rpc_call_prepare = nfs41_sequence_prepare,
9403 	.rpc_release = nfs41_sequence_release,
9404 };
9405 
9406 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9407 		const struct cred *cred,
9408 		struct nfs4_slot *slot,
9409 		bool is_privileged)
9410 {
9411 	struct nfs4_sequence_data *calldata;
9412 	struct rpc_message msg = {
9413 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9414 		.rpc_cred = cred,
9415 	};
9416 	struct rpc_task_setup task_setup_data = {
9417 		.rpc_client = clp->cl_rpcclient,
9418 		.rpc_message = &msg,
9419 		.callback_ops = &nfs41_sequence_ops,
9420 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9421 	};
9422 	struct rpc_task *ret;
9423 
9424 	ret = ERR_PTR(-EIO);
9425 	if (!refcount_inc_not_zero(&clp->cl_count))
9426 		goto out_err;
9427 
9428 	ret = ERR_PTR(-ENOMEM);
9429 	calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9430 	if (calldata == NULL)
9431 		goto out_put_clp;
9432 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9433 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9434 	msg.rpc_argp = &calldata->args;
9435 	msg.rpc_resp = &calldata->res;
9436 	calldata->clp = clp;
9437 	task_setup_data.callback_data = calldata;
9438 
9439 	ret = rpc_run_task(&task_setup_data);
9440 	if (IS_ERR(ret))
9441 		goto out_err;
9442 	return ret;
9443 out_put_clp:
9444 	nfs_put_client(clp);
9445 out_err:
9446 	nfs41_release_slot(slot);
9447 	return ret;
9448 }
9449 
9450 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9451 {
9452 	struct rpc_task *task;
9453 	int ret = 0;
9454 
9455 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9456 		return -EAGAIN;
9457 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9458 	if (IS_ERR(task))
9459 		ret = PTR_ERR(task);
9460 	else
9461 		rpc_put_task_async(task);
9462 	dprintk("<-- %s status=%d\n", __func__, ret);
9463 	return ret;
9464 }
9465 
9466 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9467 {
9468 	struct rpc_task *task;
9469 	int ret;
9470 
9471 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9472 	if (IS_ERR(task)) {
9473 		ret = PTR_ERR(task);
9474 		goto out;
9475 	}
9476 	ret = rpc_wait_for_completion_task(task);
9477 	if (!ret)
9478 		ret = task->tk_status;
9479 	rpc_put_task(task);
9480 out:
9481 	dprintk("<-- %s status=%d\n", __func__, ret);
9482 	return ret;
9483 }
9484 
9485 struct nfs4_reclaim_complete_data {
9486 	struct nfs_client *clp;
9487 	struct nfs41_reclaim_complete_args arg;
9488 	struct nfs41_reclaim_complete_res res;
9489 };
9490 
9491 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9492 {
9493 	struct nfs4_reclaim_complete_data *calldata = data;
9494 
9495 	nfs4_setup_sequence(calldata->clp,
9496 			&calldata->arg.seq_args,
9497 			&calldata->res.seq_res,
9498 			task);
9499 }
9500 
9501 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9502 {
9503 	switch(task->tk_status) {
9504 	case 0:
9505 		wake_up_all(&clp->cl_lock_waitq);
9506 		fallthrough;
9507 	case -NFS4ERR_COMPLETE_ALREADY:
9508 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9509 		break;
9510 	case -NFS4ERR_DELAY:
9511 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9512 		fallthrough;
9513 	case -NFS4ERR_RETRY_UNCACHED_REP:
9514 	case -EACCES:
9515 		dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9516 			__func__, task->tk_status, clp->cl_hostname);
9517 		return -EAGAIN;
9518 	case -NFS4ERR_BADSESSION:
9519 	case -NFS4ERR_DEADSESSION:
9520 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9521 		break;
9522 	default:
9523 		nfs4_schedule_lease_recovery(clp);
9524 	}
9525 	return 0;
9526 }
9527 
9528 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9529 {
9530 	struct nfs4_reclaim_complete_data *calldata = data;
9531 	struct nfs_client *clp = calldata->clp;
9532 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9533 
9534 	if (!nfs41_sequence_done(task, res))
9535 		return;
9536 
9537 	trace_nfs4_reclaim_complete(clp, task->tk_status);
9538 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9539 		rpc_restart_call_prepare(task);
9540 		return;
9541 	}
9542 }
9543 
9544 static void nfs4_free_reclaim_complete_data(void *data)
9545 {
9546 	struct nfs4_reclaim_complete_data *calldata = data;
9547 
9548 	kfree(calldata);
9549 }
9550 
9551 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9552 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9553 	.rpc_call_done = nfs4_reclaim_complete_done,
9554 	.rpc_release = nfs4_free_reclaim_complete_data,
9555 };
9556 
9557 /*
9558  * Issue a global reclaim complete.
9559  */
9560 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9561 		const struct cred *cred)
9562 {
9563 	struct nfs4_reclaim_complete_data *calldata;
9564 	struct rpc_message msg = {
9565 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9566 		.rpc_cred = cred,
9567 	};
9568 	struct rpc_task_setup task_setup_data = {
9569 		.rpc_client = clp->cl_rpcclient,
9570 		.rpc_message = &msg,
9571 		.callback_ops = &nfs4_reclaim_complete_call_ops,
9572 		.flags = RPC_TASK_NO_ROUND_ROBIN,
9573 	};
9574 	int status = -ENOMEM;
9575 
9576 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9577 	if (calldata == NULL)
9578 		goto out;
9579 	calldata->clp = clp;
9580 	calldata->arg.one_fs = 0;
9581 
9582 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9583 	msg.rpc_argp = &calldata->arg;
9584 	msg.rpc_resp = &calldata->res;
9585 	task_setup_data.callback_data = calldata;
9586 	status = nfs4_call_sync_custom(&task_setup_data);
9587 out:
9588 	dprintk("<-- %s status=%d\n", __func__, status);
9589 	return status;
9590 }
9591 
9592 static void
9593 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9594 {
9595 	struct nfs4_layoutget *lgp = calldata;
9596 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9597 
9598 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9599 				&lgp->res.seq_res, task);
9600 }
9601 
9602 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9603 {
9604 	struct nfs4_layoutget *lgp = calldata;
9605 
9606 	nfs41_sequence_process(task, &lgp->res.seq_res);
9607 }
9608 
9609 static int
9610 nfs4_layoutget_handle_exception(struct rpc_task *task,
9611 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9612 {
9613 	struct inode *inode = lgp->args.inode;
9614 	struct nfs_server *server = NFS_SERVER(inode);
9615 	struct pnfs_layout_hdr *lo = lgp->lo;
9616 	int nfs4err = task->tk_status;
9617 	int err, status = 0;
9618 	LIST_HEAD(head);
9619 
9620 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9621 
9622 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9623 
9624 	switch (nfs4err) {
9625 	case 0:
9626 		goto out;
9627 
9628 	/*
9629 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9630 	 * on the file. set tk_status to -ENODATA to tell upper layer to
9631 	 * retry go inband.
9632 	 */
9633 	case -NFS4ERR_LAYOUTUNAVAILABLE:
9634 		status = -ENODATA;
9635 		goto out;
9636 	/*
9637 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9638 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9639 	 */
9640 	case -NFS4ERR_BADLAYOUT:
9641 		status = -EOVERFLOW;
9642 		goto out;
9643 	/*
9644 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9645 	 * (or clients) writing to the same RAID stripe except when
9646 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9647 	 *
9648 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9649 	 * while, and then eventually give up.
9650 	 */
9651 	case -NFS4ERR_LAYOUTTRYLATER:
9652 		if (lgp->args.minlength == 0) {
9653 			status = -EOVERFLOW;
9654 			goto out;
9655 		}
9656 		status = -EBUSY;
9657 		break;
9658 	case -NFS4ERR_RECALLCONFLICT:
9659 		status = -ERECALLCONFLICT;
9660 		break;
9661 	case -NFS4ERR_DELEG_REVOKED:
9662 	case -NFS4ERR_ADMIN_REVOKED:
9663 	case -NFS4ERR_EXPIRED:
9664 	case -NFS4ERR_BAD_STATEID:
9665 		exception->timeout = 0;
9666 		spin_lock(&inode->i_lock);
9667 		/* If the open stateid was bad, then recover it. */
9668 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9669 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9670 			spin_unlock(&inode->i_lock);
9671 			exception->state = lgp->args.ctx->state;
9672 			exception->stateid = &lgp->args.stateid;
9673 			break;
9674 		}
9675 
9676 		/*
9677 		 * Mark the bad layout state as invalid, then retry
9678 		 */
9679 		pnfs_mark_layout_stateid_invalid(lo, &head);
9680 		spin_unlock(&inode->i_lock);
9681 		nfs_commit_inode(inode, 0);
9682 		pnfs_free_lseg_list(&head);
9683 		status = -EAGAIN;
9684 		goto out;
9685 	}
9686 
9687 	err = nfs4_handle_exception(server, nfs4err, exception);
9688 	if (!status) {
9689 		if (exception->retry)
9690 			status = -EAGAIN;
9691 		else
9692 			status = err;
9693 	}
9694 out:
9695 	return status;
9696 }
9697 
9698 size_t max_response_pages(struct nfs_server *server)
9699 {
9700 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9701 	return nfs_page_array_len(0, max_resp_sz);
9702 }
9703 
9704 static void nfs4_layoutget_release(void *calldata)
9705 {
9706 	struct nfs4_layoutget *lgp = calldata;
9707 
9708 	nfs4_sequence_free_slot(&lgp->res.seq_res);
9709 	pnfs_layoutget_free(lgp);
9710 }
9711 
9712 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9713 	.rpc_call_prepare = nfs4_layoutget_prepare,
9714 	.rpc_call_done = nfs4_layoutget_done,
9715 	.rpc_release = nfs4_layoutget_release,
9716 };
9717 
9718 struct pnfs_layout_segment *
9719 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9720 {
9721 	struct inode *inode = lgp->args.inode;
9722 	struct nfs_server *server = NFS_SERVER(inode);
9723 	struct rpc_task *task;
9724 	struct rpc_message msg = {
9725 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9726 		.rpc_argp = &lgp->args,
9727 		.rpc_resp = &lgp->res,
9728 		.rpc_cred = lgp->cred,
9729 	};
9730 	struct rpc_task_setup task_setup_data = {
9731 		.rpc_client = server->client,
9732 		.rpc_message = &msg,
9733 		.callback_ops = &nfs4_layoutget_call_ops,
9734 		.callback_data = lgp,
9735 		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9736 			 RPC_TASK_MOVEABLE,
9737 	};
9738 	struct pnfs_layout_segment *lseg = NULL;
9739 	struct nfs4_exception exception = {
9740 		.inode = inode,
9741 		.timeout = *timeout,
9742 	};
9743 	int status = 0;
9744 
9745 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9746 
9747 	task = rpc_run_task(&task_setup_data);
9748 	if (IS_ERR(task))
9749 		return ERR_CAST(task);
9750 
9751 	status = rpc_wait_for_completion_task(task);
9752 	if (status != 0)
9753 		goto out;
9754 
9755 	if (task->tk_status < 0) {
9756 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9757 		*timeout = exception.timeout;
9758 	} else if (lgp->res.layoutp->len == 0) {
9759 		status = -EAGAIN;
9760 		*timeout = nfs4_update_delay(&exception.timeout);
9761 	} else
9762 		lseg = pnfs_layout_process(lgp);
9763 out:
9764 	trace_nfs4_layoutget(lgp->args.ctx,
9765 			&lgp->args.range,
9766 			&lgp->res.range,
9767 			&lgp->res.stateid,
9768 			status);
9769 
9770 	rpc_put_task(task);
9771 	dprintk("<-- %s status=%d\n", __func__, status);
9772 	if (status)
9773 		return ERR_PTR(status);
9774 	return lseg;
9775 }
9776 
9777 static void
9778 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9779 {
9780 	struct nfs4_layoutreturn *lrp = calldata;
9781 
9782 	nfs4_setup_sequence(lrp->clp,
9783 			&lrp->args.seq_args,
9784 			&lrp->res.seq_res,
9785 			task);
9786 	if (!pnfs_layout_is_valid(lrp->args.layout))
9787 		rpc_exit(task, 0);
9788 }
9789 
9790 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9791 {
9792 	struct nfs4_layoutreturn *lrp = calldata;
9793 	struct nfs_server *server;
9794 
9795 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9796 		return;
9797 
9798 	/*
9799 	 * Was there an RPC level error? Assume the call succeeded,
9800 	 * and that we need to release the layout
9801 	 */
9802 	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9803 		lrp->res.lrs_present = 0;
9804 		return;
9805 	}
9806 
9807 	server = NFS_SERVER(lrp->args.inode);
9808 	switch (task->tk_status) {
9809 	case -NFS4ERR_OLD_STATEID:
9810 		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9811 					&lrp->args.range,
9812 					lrp->args.inode))
9813 			goto out_restart;
9814 		fallthrough;
9815 	default:
9816 		task->tk_status = 0;
9817 		fallthrough;
9818 	case 0:
9819 		break;
9820 	case -NFS4ERR_DELAY:
9821 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9822 			break;
9823 		goto out_restart;
9824 	}
9825 	return;
9826 out_restart:
9827 	task->tk_status = 0;
9828 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9829 	rpc_restart_call_prepare(task);
9830 }
9831 
9832 static void nfs4_layoutreturn_release(void *calldata)
9833 {
9834 	struct nfs4_layoutreturn *lrp = calldata;
9835 	struct pnfs_layout_hdr *lo = lrp->args.layout;
9836 
9837 	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9838 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9839 	nfs4_sequence_free_slot(&lrp->res.seq_res);
9840 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9841 		lrp->ld_private.ops->free(&lrp->ld_private);
9842 	pnfs_put_layout_hdr(lrp->args.layout);
9843 	nfs_iput_and_deactive(lrp->inode);
9844 	put_cred(lrp->cred);
9845 	kfree(calldata);
9846 }
9847 
9848 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9849 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9850 	.rpc_call_done = nfs4_layoutreturn_done,
9851 	.rpc_release = nfs4_layoutreturn_release,
9852 };
9853 
9854 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9855 {
9856 	struct rpc_task *task;
9857 	struct rpc_message msg = {
9858 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9859 		.rpc_argp = &lrp->args,
9860 		.rpc_resp = &lrp->res,
9861 		.rpc_cred = lrp->cred,
9862 	};
9863 	struct rpc_task_setup task_setup_data = {
9864 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9865 		.rpc_message = &msg,
9866 		.callback_ops = &nfs4_layoutreturn_call_ops,
9867 		.callback_data = lrp,
9868 		.flags = RPC_TASK_MOVEABLE,
9869 	};
9870 	int status = 0;
9871 
9872 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9873 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9874 			&task_setup_data.rpc_client, &msg);
9875 
9876 	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9877 	if (!sync) {
9878 		if (!lrp->inode) {
9879 			nfs4_layoutreturn_release(lrp);
9880 			return -EAGAIN;
9881 		}
9882 		task_setup_data.flags |= RPC_TASK_ASYNC;
9883 	}
9884 	if (!lrp->inode)
9885 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9886 				   1);
9887 	else
9888 		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9889 				   0);
9890 	task = rpc_run_task(&task_setup_data);
9891 	if (IS_ERR(task))
9892 		return PTR_ERR(task);
9893 	if (sync)
9894 		status = task->tk_status;
9895 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9896 	dprintk("<-- %s status=%d\n", __func__, status);
9897 	rpc_put_task(task);
9898 	return status;
9899 }
9900 
9901 static int
9902 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9903 		struct pnfs_device *pdev,
9904 		const struct cred *cred)
9905 {
9906 	struct nfs4_getdeviceinfo_args args = {
9907 		.pdev = pdev,
9908 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9909 			NOTIFY_DEVICEID4_DELETE,
9910 	};
9911 	struct nfs4_getdeviceinfo_res res = {
9912 		.pdev = pdev,
9913 	};
9914 	struct rpc_message msg = {
9915 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9916 		.rpc_argp = &args,
9917 		.rpc_resp = &res,
9918 		.rpc_cred = cred,
9919 	};
9920 	int status;
9921 
9922 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9923 	if (res.notification & ~args.notify_types)
9924 		dprintk("%s: unsupported notification\n", __func__);
9925 	if (res.notification != args.notify_types)
9926 		pdev->nocache = 1;
9927 
9928 	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9929 
9930 	dprintk("<-- %s status=%d\n", __func__, status);
9931 
9932 	return status;
9933 }
9934 
9935 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9936 		struct pnfs_device *pdev,
9937 		const struct cred *cred)
9938 {
9939 	struct nfs4_exception exception = { };
9940 	int err;
9941 
9942 	do {
9943 		err = nfs4_handle_exception(server,
9944 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9945 					&exception);
9946 	} while (exception.retry);
9947 	return err;
9948 }
9949 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9950 
9951 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9952 {
9953 	struct nfs4_layoutcommit_data *data = calldata;
9954 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9955 
9956 	nfs4_setup_sequence(server->nfs_client,
9957 			&data->args.seq_args,
9958 			&data->res.seq_res,
9959 			task);
9960 }
9961 
9962 static void
9963 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9964 {
9965 	struct nfs4_layoutcommit_data *data = calldata;
9966 	struct nfs_server *server = NFS_SERVER(data->args.inode);
9967 
9968 	if (!nfs41_sequence_done(task, &data->res.seq_res))
9969 		return;
9970 
9971 	switch (task->tk_status) { /* Just ignore these failures */
9972 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9973 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9974 	case -NFS4ERR_BADLAYOUT:     /* no layout */
9975 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
9976 		task->tk_status = 0;
9977 		break;
9978 	case 0:
9979 		break;
9980 	default:
9981 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9982 			rpc_restart_call_prepare(task);
9983 			return;
9984 		}
9985 	}
9986 }
9987 
9988 static void nfs4_layoutcommit_release(void *calldata)
9989 {
9990 	struct nfs4_layoutcommit_data *data = calldata;
9991 
9992 	pnfs_cleanup_layoutcommit(data);
9993 	nfs_post_op_update_inode_force_wcc(data->args.inode,
9994 					   data->res.fattr);
9995 	put_cred(data->cred);
9996 	nfs_iput_and_deactive(data->inode);
9997 	kfree(data);
9998 }
9999 
10000 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
10001 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
10002 	.rpc_call_done = nfs4_layoutcommit_done,
10003 	.rpc_release = nfs4_layoutcommit_release,
10004 };
10005 
10006 int
10007 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10008 {
10009 	struct rpc_message msg = {
10010 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10011 		.rpc_argp = &data->args,
10012 		.rpc_resp = &data->res,
10013 		.rpc_cred = data->cred,
10014 	};
10015 	struct rpc_task_setup task_setup_data = {
10016 		.task = &data->task,
10017 		.rpc_client = NFS_CLIENT(data->args.inode),
10018 		.rpc_message = &msg,
10019 		.callback_ops = &nfs4_layoutcommit_ops,
10020 		.callback_data = data,
10021 		.flags = RPC_TASK_MOVEABLE,
10022 	};
10023 	struct rpc_task *task;
10024 	int status = 0;
10025 
10026 	dprintk("NFS: initiating layoutcommit call. sync %d "
10027 		"lbw: %llu inode %lu\n", sync,
10028 		data->args.lastbytewritten,
10029 		data->args.inode->i_ino);
10030 
10031 	if (!sync) {
10032 		data->inode = nfs_igrab_and_active(data->args.inode);
10033 		if (data->inode == NULL) {
10034 			nfs4_layoutcommit_release(data);
10035 			return -EAGAIN;
10036 		}
10037 		task_setup_data.flags = RPC_TASK_ASYNC;
10038 	}
10039 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10040 	task = rpc_run_task(&task_setup_data);
10041 	if (IS_ERR(task))
10042 		return PTR_ERR(task);
10043 	if (sync)
10044 		status = task->tk_status;
10045 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10046 	dprintk("%s: status %d\n", __func__, status);
10047 	rpc_put_task(task);
10048 	return status;
10049 }
10050 
10051 /*
10052  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10053  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10054  */
10055 static int
10056 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10057 		    struct nfs_fsinfo *info,
10058 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10059 {
10060 	struct nfs41_secinfo_no_name_args args = {
10061 		.style = SECINFO_STYLE_CURRENT_FH,
10062 	};
10063 	struct nfs4_secinfo_res res = {
10064 		.flavors = flavors,
10065 	};
10066 	struct rpc_message msg = {
10067 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10068 		.rpc_argp = &args,
10069 		.rpc_resp = &res,
10070 	};
10071 	struct nfs4_call_sync_data data = {
10072 		.seq_server = server,
10073 		.seq_args = &args.seq_args,
10074 		.seq_res = &res.seq_res,
10075 	};
10076 	struct rpc_task_setup task_setup = {
10077 		.rpc_client = server->client,
10078 		.rpc_message = &msg,
10079 		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10080 		.callback_data = &data,
10081 		.flags = RPC_TASK_NO_ROUND_ROBIN,
10082 	};
10083 	const struct cred *cred = NULL;
10084 	int status;
10085 
10086 	if (use_integrity) {
10087 		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10088 
10089 		cred = nfs4_get_clid_cred(server->nfs_client);
10090 		msg.rpc_cred = cred;
10091 	}
10092 
10093 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10094 	status = nfs4_call_sync_custom(&task_setup);
10095 	dprintk("<-- %s status=%d\n", __func__, status);
10096 
10097 	put_cred(cred);
10098 
10099 	return status;
10100 }
10101 
10102 static int
10103 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10104 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10105 {
10106 	struct nfs4_exception exception = {
10107 		.interruptible = true,
10108 	};
10109 	int err;
10110 	do {
10111 		/* first try using integrity protection */
10112 		err = -NFS4ERR_WRONGSEC;
10113 
10114 		/* try to use integrity protection with machine cred */
10115 		if (_nfs4_is_integrity_protected(server->nfs_client))
10116 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10117 							  flavors, true);
10118 
10119 		/*
10120 		 * if unable to use integrity protection, or SECINFO with
10121 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10122 		 * disallowed by spec, but exists in deployed servers) use
10123 		 * the current filesystem's rpc_client and the user cred.
10124 		 */
10125 		if (err == -NFS4ERR_WRONGSEC)
10126 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10127 							  flavors, false);
10128 
10129 		switch (err) {
10130 		case 0:
10131 		case -NFS4ERR_WRONGSEC:
10132 		case -ENOTSUPP:
10133 			goto out;
10134 		default:
10135 			err = nfs4_handle_exception(server, err, &exception);
10136 		}
10137 	} while (exception.retry);
10138 out:
10139 	return err;
10140 }
10141 
10142 static int
10143 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10144 		    struct nfs_fsinfo *info)
10145 {
10146 	int err;
10147 	struct page *page;
10148 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10149 	struct nfs4_secinfo_flavors *flavors;
10150 	struct nfs4_secinfo4 *secinfo;
10151 	int i;
10152 
10153 	page = alloc_page(GFP_KERNEL);
10154 	if (!page) {
10155 		err = -ENOMEM;
10156 		goto out;
10157 	}
10158 
10159 	flavors = page_address(page);
10160 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10161 
10162 	/*
10163 	 * Fall back on "guess and check" method if
10164 	 * the server doesn't support SECINFO_NO_NAME
10165 	 */
10166 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10167 		err = nfs4_find_root_sec(server, fhandle, info);
10168 		goto out_freepage;
10169 	}
10170 	if (err)
10171 		goto out_freepage;
10172 
10173 	for (i = 0; i < flavors->num_flavors; i++) {
10174 		secinfo = &flavors->flavors[i];
10175 
10176 		switch (secinfo->flavor) {
10177 		case RPC_AUTH_NULL:
10178 		case RPC_AUTH_UNIX:
10179 		case RPC_AUTH_GSS:
10180 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10181 					&secinfo->flavor_info);
10182 			break;
10183 		default:
10184 			flavor = RPC_AUTH_MAXFLAVOR;
10185 			break;
10186 		}
10187 
10188 		if (!nfs_auth_info_match(&server->auth_info, flavor))
10189 			flavor = RPC_AUTH_MAXFLAVOR;
10190 
10191 		if (flavor != RPC_AUTH_MAXFLAVOR) {
10192 			err = nfs4_lookup_root_sec(server, fhandle,
10193 						   info, flavor);
10194 			if (!err)
10195 				break;
10196 		}
10197 	}
10198 
10199 	if (flavor == RPC_AUTH_MAXFLAVOR)
10200 		err = -EPERM;
10201 
10202 out_freepage:
10203 	put_page(page);
10204 	if (err == -EACCES)
10205 		return -EPERM;
10206 out:
10207 	return err;
10208 }
10209 
10210 static int _nfs41_test_stateid(struct nfs_server *server,
10211 		nfs4_stateid *stateid,
10212 		const struct cred *cred)
10213 {
10214 	int status;
10215 	struct nfs41_test_stateid_args args = {
10216 		.stateid = stateid,
10217 	};
10218 	struct nfs41_test_stateid_res res;
10219 	struct rpc_message msg = {
10220 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10221 		.rpc_argp = &args,
10222 		.rpc_resp = &res,
10223 		.rpc_cred = cred,
10224 	};
10225 	struct rpc_clnt *rpc_client = server->client;
10226 
10227 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10228 		&rpc_client, &msg);
10229 
10230 	dprintk("NFS call  test_stateid %p\n", stateid);
10231 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10232 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10233 			&args.seq_args, &res.seq_res);
10234 	if (status != NFS_OK) {
10235 		dprintk("NFS reply test_stateid: failed, %d\n", status);
10236 		return status;
10237 	}
10238 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10239 	return -res.status;
10240 }
10241 
10242 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10243 		int err, struct nfs4_exception *exception)
10244 {
10245 	exception->retry = 0;
10246 	switch(err) {
10247 	case -NFS4ERR_DELAY:
10248 	case -NFS4ERR_RETRY_UNCACHED_REP:
10249 		nfs4_handle_exception(server, err, exception);
10250 		break;
10251 	case -NFS4ERR_BADSESSION:
10252 	case -NFS4ERR_BADSLOT:
10253 	case -NFS4ERR_BAD_HIGH_SLOT:
10254 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10255 	case -NFS4ERR_DEADSESSION:
10256 		nfs4_do_handle_exception(server, err, exception);
10257 	}
10258 }
10259 
10260 /**
10261  * nfs41_test_stateid - perform a TEST_STATEID operation
10262  *
10263  * @server: server / transport on which to perform the operation
10264  * @stateid: state ID to test
10265  * @cred: credential
10266  *
10267  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10268  * Otherwise a negative NFS4ERR value is returned if the operation
10269  * failed or the state ID is not currently valid.
10270  */
10271 static int nfs41_test_stateid(struct nfs_server *server,
10272 		nfs4_stateid *stateid,
10273 		const struct cred *cred)
10274 {
10275 	struct nfs4_exception exception = {
10276 		.interruptible = true,
10277 	};
10278 	int err;
10279 	do {
10280 		err = _nfs41_test_stateid(server, stateid, cred);
10281 		nfs4_handle_delay_or_session_error(server, err, &exception);
10282 	} while (exception.retry);
10283 	return err;
10284 }
10285 
10286 struct nfs_free_stateid_data {
10287 	struct nfs_server *server;
10288 	struct nfs41_free_stateid_args args;
10289 	struct nfs41_free_stateid_res res;
10290 };
10291 
10292 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10293 {
10294 	struct nfs_free_stateid_data *data = calldata;
10295 	nfs4_setup_sequence(data->server->nfs_client,
10296 			&data->args.seq_args,
10297 			&data->res.seq_res,
10298 			task);
10299 }
10300 
10301 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10302 {
10303 	struct nfs_free_stateid_data *data = calldata;
10304 
10305 	nfs41_sequence_done(task, &data->res.seq_res);
10306 
10307 	switch (task->tk_status) {
10308 	case -NFS4ERR_DELAY:
10309 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10310 			rpc_restart_call_prepare(task);
10311 	}
10312 }
10313 
10314 static void nfs41_free_stateid_release(void *calldata)
10315 {
10316 	struct nfs_free_stateid_data *data = calldata;
10317 	struct nfs_client *clp = data->server->nfs_client;
10318 
10319 	nfs_put_client(clp);
10320 	kfree(calldata);
10321 }
10322 
10323 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10324 	.rpc_call_prepare = nfs41_free_stateid_prepare,
10325 	.rpc_call_done = nfs41_free_stateid_done,
10326 	.rpc_release = nfs41_free_stateid_release,
10327 };
10328 
10329 /**
10330  * nfs41_free_stateid - perform a FREE_STATEID operation
10331  *
10332  * @server: server / transport on which to perform the operation
10333  * @stateid: state ID to release
10334  * @cred: credential
10335  * @privileged: set to true if this call needs to be privileged
10336  *
10337  * Note: this function is always asynchronous.
10338  */
10339 static int nfs41_free_stateid(struct nfs_server *server,
10340 		const nfs4_stateid *stateid,
10341 		const struct cred *cred,
10342 		bool privileged)
10343 {
10344 	struct rpc_message msg = {
10345 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10346 		.rpc_cred = cred,
10347 	};
10348 	struct rpc_task_setup task_setup = {
10349 		.rpc_client = server->client,
10350 		.rpc_message = &msg,
10351 		.callback_ops = &nfs41_free_stateid_ops,
10352 		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10353 	};
10354 	struct nfs_free_stateid_data *data;
10355 	struct rpc_task *task;
10356 	struct nfs_client *clp = server->nfs_client;
10357 
10358 	if (!refcount_inc_not_zero(&clp->cl_count))
10359 		return -EIO;
10360 
10361 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10362 		&task_setup.rpc_client, &msg);
10363 
10364 	dprintk("NFS call  free_stateid %p\n", stateid);
10365 	data = kmalloc(sizeof(*data), GFP_KERNEL);
10366 	if (!data)
10367 		return -ENOMEM;
10368 	data->server = server;
10369 	nfs4_stateid_copy(&data->args.stateid, stateid);
10370 
10371 	task_setup.callback_data = data;
10372 
10373 	msg.rpc_argp = &data->args;
10374 	msg.rpc_resp = &data->res;
10375 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10376 	task = rpc_run_task(&task_setup);
10377 	if (IS_ERR(task))
10378 		return PTR_ERR(task);
10379 	rpc_put_task(task);
10380 	return 0;
10381 }
10382 
10383 static void
10384 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10385 {
10386 	const struct cred *cred = lsp->ls_state->owner->so_cred;
10387 
10388 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10389 	nfs4_free_lock_state(server, lsp);
10390 }
10391 
10392 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10393 		const nfs4_stateid *s2)
10394 {
10395 	if (s1->type != s2->type)
10396 		return false;
10397 
10398 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10399 		return false;
10400 
10401 	if (s1->seqid == s2->seqid)
10402 		return true;
10403 
10404 	return s1->seqid == 0 || s2->seqid == 0;
10405 }
10406 
10407 #endif /* CONFIG_NFS_V4_1 */
10408 
10409 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10410 		const nfs4_stateid *s2)
10411 {
10412 	return nfs4_stateid_match(s1, s2);
10413 }
10414 
10415 
10416 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10417 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10418 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10419 	.recover_open	= nfs4_open_reclaim,
10420 	.recover_lock	= nfs4_lock_reclaim,
10421 	.establish_clid = nfs4_init_clientid,
10422 	.detect_trunking = nfs40_discover_server_trunking,
10423 };
10424 
10425 #if defined(CONFIG_NFS_V4_1)
10426 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10427 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10428 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10429 	.recover_open	= nfs4_open_reclaim,
10430 	.recover_lock	= nfs4_lock_reclaim,
10431 	.establish_clid = nfs41_init_clientid,
10432 	.reclaim_complete = nfs41_proc_reclaim_complete,
10433 	.detect_trunking = nfs41_discover_server_trunking,
10434 };
10435 #endif /* CONFIG_NFS_V4_1 */
10436 
10437 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10438 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10439 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10440 	.recover_open	= nfs40_open_expired,
10441 	.recover_lock	= nfs4_lock_expired,
10442 	.establish_clid = nfs4_init_clientid,
10443 };
10444 
10445 #if defined(CONFIG_NFS_V4_1)
10446 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10447 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10448 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10449 	.recover_open	= nfs41_open_expired,
10450 	.recover_lock	= nfs41_lock_expired,
10451 	.establish_clid = nfs41_init_clientid,
10452 };
10453 #endif /* CONFIG_NFS_V4_1 */
10454 
10455 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10456 	.sched_state_renewal = nfs4_proc_async_renew,
10457 	.get_state_renewal_cred = nfs4_get_renew_cred,
10458 	.renew_lease = nfs4_proc_renew,
10459 };
10460 
10461 #if defined(CONFIG_NFS_V4_1)
10462 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10463 	.sched_state_renewal = nfs41_proc_async_sequence,
10464 	.get_state_renewal_cred = nfs4_get_machine_cred,
10465 	.renew_lease = nfs4_proc_sequence,
10466 };
10467 #endif
10468 
10469 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10470 	.get_locations = _nfs40_proc_get_locations,
10471 	.fsid_present = _nfs40_proc_fsid_present,
10472 };
10473 
10474 #if defined(CONFIG_NFS_V4_1)
10475 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10476 	.get_locations = _nfs41_proc_get_locations,
10477 	.fsid_present = _nfs41_proc_fsid_present,
10478 };
10479 #endif	/* CONFIG_NFS_V4_1 */
10480 
10481 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10482 	.minor_version = 0,
10483 	.init_caps = NFS_CAP_READDIRPLUS
10484 		| NFS_CAP_ATOMIC_OPEN
10485 		| NFS_CAP_POSIX_LOCK,
10486 	.init_client = nfs40_init_client,
10487 	.shutdown_client = nfs40_shutdown_client,
10488 	.match_stateid = nfs4_match_stateid,
10489 	.find_root_sec = nfs4_find_root_sec,
10490 	.free_lock_state = nfs4_release_lockowner,
10491 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10492 	.alloc_seqid = nfs_alloc_seqid,
10493 	.call_sync_ops = &nfs40_call_sync_ops,
10494 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10495 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10496 	.state_renewal_ops = &nfs40_state_renewal_ops,
10497 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10498 };
10499 
10500 #if defined(CONFIG_NFS_V4_1)
10501 static struct nfs_seqid *
10502 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10503 {
10504 	return NULL;
10505 }
10506 
10507 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10508 	.minor_version = 1,
10509 	.init_caps = NFS_CAP_READDIRPLUS
10510 		| NFS_CAP_ATOMIC_OPEN
10511 		| NFS_CAP_POSIX_LOCK
10512 		| NFS_CAP_STATEID_NFSV41
10513 		| NFS_CAP_ATOMIC_OPEN_V1
10514 		| NFS_CAP_LGOPEN
10515 		| NFS_CAP_MOVEABLE,
10516 	.init_client = nfs41_init_client,
10517 	.shutdown_client = nfs41_shutdown_client,
10518 	.match_stateid = nfs41_match_stateid,
10519 	.find_root_sec = nfs41_find_root_sec,
10520 	.free_lock_state = nfs41_free_lock_state,
10521 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10522 	.alloc_seqid = nfs_alloc_no_seqid,
10523 	.session_trunk = nfs4_test_session_trunk,
10524 	.call_sync_ops = &nfs41_call_sync_ops,
10525 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10526 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10527 	.state_renewal_ops = &nfs41_state_renewal_ops,
10528 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10529 };
10530 #endif
10531 
10532 #if defined(CONFIG_NFS_V4_2)
10533 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10534 	.minor_version = 2,
10535 	.init_caps = NFS_CAP_READDIRPLUS
10536 		| NFS_CAP_ATOMIC_OPEN
10537 		| NFS_CAP_POSIX_LOCK
10538 		| NFS_CAP_STATEID_NFSV41
10539 		| NFS_CAP_ATOMIC_OPEN_V1
10540 		| NFS_CAP_LGOPEN
10541 		| NFS_CAP_ALLOCATE
10542 		| NFS_CAP_COPY
10543 		| NFS_CAP_OFFLOAD_CANCEL
10544 		| NFS_CAP_COPY_NOTIFY
10545 		| NFS_CAP_DEALLOCATE
10546 		| NFS_CAP_SEEK
10547 		| NFS_CAP_LAYOUTSTATS
10548 		| NFS_CAP_CLONE
10549 		| NFS_CAP_LAYOUTERROR
10550 		| NFS_CAP_READ_PLUS
10551 		| NFS_CAP_MOVEABLE,
10552 	.init_client = nfs41_init_client,
10553 	.shutdown_client = nfs41_shutdown_client,
10554 	.match_stateid = nfs41_match_stateid,
10555 	.find_root_sec = nfs41_find_root_sec,
10556 	.free_lock_state = nfs41_free_lock_state,
10557 	.call_sync_ops = &nfs41_call_sync_ops,
10558 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10559 	.alloc_seqid = nfs_alloc_no_seqid,
10560 	.session_trunk = nfs4_test_session_trunk,
10561 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10562 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10563 	.state_renewal_ops = &nfs41_state_renewal_ops,
10564 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10565 };
10566 #endif
10567 
10568 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10569 	[0] = &nfs_v4_0_minor_ops,
10570 #if defined(CONFIG_NFS_V4_1)
10571 	[1] = &nfs_v4_1_minor_ops,
10572 #endif
10573 #if defined(CONFIG_NFS_V4_2)
10574 	[2] = &nfs_v4_2_minor_ops,
10575 #endif
10576 };
10577 
10578 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10579 {
10580 	ssize_t error, error2, error3;
10581 
10582 	error = generic_listxattr(dentry, list, size);
10583 	if (error < 0)
10584 		return error;
10585 	if (list) {
10586 		list += error;
10587 		size -= error;
10588 	}
10589 
10590 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10591 	if (error2 < 0)
10592 		return error2;
10593 
10594 	if (list) {
10595 		list += error2;
10596 		size -= error2;
10597 	}
10598 
10599 	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10600 	if (error3 < 0)
10601 		return error3;
10602 
10603 	return error + error2 + error3;
10604 }
10605 
10606 static void nfs4_enable_swap(struct inode *inode)
10607 {
10608 	/* The state manager thread must always be running.
10609 	 * It will notice the client is a swapper, and stay put.
10610 	 */
10611 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10612 
10613 	nfs4_schedule_state_manager(clp);
10614 }
10615 
10616 static void nfs4_disable_swap(struct inode *inode)
10617 {
10618 	/* The state manager thread will now exit once it is
10619 	 * woken.
10620 	 */
10621 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10622 
10623 	set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10624 	clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10625 	wake_up_var(&clp->cl_state);
10626 }
10627 
10628 static const struct inode_operations nfs4_dir_inode_operations = {
10629 	.create		= nfs_create,
10630 	.lookup		= nfs_lookup,
10631 	.atomic_open	= nfs_atomic_open,
10632 	.link		= nfs_link,
10633 	.unlink		= nfs_unlink,
10634 	.symlink	= nfs_symlink,
10635 	.mkdir		= nfs_mkdir,
10636 	.rmdir		= nfs_rmdir,
10637 	.mknod		= nfs_mknod,
10638 	.rename		= nfs_rename,
10639 	.permission	= nfs_permission,
10640 	.getattr	= nfs_getattr,
10641 	.setattr	= nfs_setattr,
10642 	.listxattr	= nfs4_listxattr,
10643 };
10644 
10645 static const struct inode_operations nfs4_file_inode_operations = {
10646 	.permission	= nfs_permission,
10647 	.getattr	= nfs_getattr,
10648 	.setattr	= nfs_setattr,
10649 	.listxattr	= nfs4_listxattr,
10650 };
10651 
10652 const struct nfs_rpc_ops nfs_v4_clientops = {
10653 	.version	= 4,			/* protocol version */
10654 	.dentry_ops	= &nfs4_dentry_operations,
10655 	.dir_inode_ops	= &nfs4_dir_inode_operations,
10656 	.file_inode_ops	= &nfs4_file_inode_operations,
10657 	.file_ops	= &nfs4_file_operations,
10658 	.getroot	= nfs4_proc_get_root,
10659 	.submount	= nfs4_submount,
10660 	.try_get_tree	= nfs4_try_get_tree,
10661 	.getattr	= nfs4_proc_getattr,
10662 	.setattr	= nfs4_proc_setattr,
10663 	.lookup		= nfs4_proc_lookup,
10664 	.lookupp	= nfs4_proc_lookupp,
10665 	.access		= nfs4_proc_access,
10666 	.readlink	= nfs4_proc_readlink,
10667 	.create		= nfs4_proc_create,
10668 	.remove		= nfs4_proc_remove,
10669 	.unlink_setup	= nfs4_proc_unlink_setup,
10670 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10671 	.unlink_done	= nfs4_proc_unlink_done,
10672 	.rename_setup	= nfs4_proc_rename_setup,
10673 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10674 	.rename_done	= nfs4_proc_rename_done,
10675 	.link		= nfs4_proc_link,
10676 	.symlink	= nfs4_proc_symlink,
10677 	.mkdir		= nfs4_proc_mkdir,
10678 	.rmdir		= nfs4_proc_rmdir,
10679 	.readdir	= nfs4_proc_readdir,
10680 	.mknod		= nfs4_proc_mknod,
10681 	.statfs		= nfs4_proc_statfs,
10682 	.fsinfo		= nfs4_proc_fsinfo,
10683 	.pathconf	= nfs4_proc_pathconf,
10684 	.set_capabilities = nfs4_server_capabilities,
10685 	.decode_dirent	= nfs4_decode_dirent,
10686 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10687 	.read_setup	= nfs4_proc_read_setup,
10688 	.read_done	= nfs4_read_done,
10689 	.write_setup	= nfs4_proc_write_setup,
10690 	.write_done	= nfs4_write_done,
10691 	.commit_setup	= nfs4_proc_commit_setup,
10692 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10693 	.commit_done	= nfs4_commit_done,
10694 	.lock		= nfs4_proc_lock,
10695 	.clear_acl_cache = nfs4_zap_acl_attr,
10696 	.close_context  = nfs4_close_context,
10697 	.open_context	= nfs4_atomic_open,
10698 	.have_delegation = nfs4_have_delegation,
10699 	.alloc_client	= nfs4_alloc_client,
10700 	.init_client	= nfs4_init_client,
10701 	.free_client	= nfs4_free_client,
10702 	.create_server	= nfs4_create_server,
10703 	.clone_server	= nfs_clone_server,
10704 	.discover_trunking = nfs4_discover_trunking,
10705 	.enable_swap	= nfs4_enable_swap,
10706 	.disable_swap	= nfs4_disable_swap,
10707 };
10708 
10709 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10710 	.name	= XATTR_NAME_NFSV4_ACL,
10711 	.list	= nfs4_xattr_list_nfs4_acl,
10712 	.get	= nfs4_xattr_get_nfs4_acl,
10713 	.set	= nfs4_xattr_set_nfs4_acl,
10714 };
10715 
10716 #if defined(CONFIG_NFS_V4_1)
10717 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10718 	.name	= XATTR_NAME_NFSV4_DACL,
10719 	.list	= nfs4_xattr_list_nfs4_dacl,
10720 	.get	= nfs4_xattr_get_nfs4_dacl,
10721 	.set	= nfs4_xattr_set_nfs4_dacl,
10722 };
10723 
10724 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10725 	.name	= XATTR_NAME_NFSV4_SACL,
10726 	.list	= nfs4_xattr_list_nfs4_sacl,
10727 	.get	= nfs4_xattr_get_nfs4_sacl,
10728 	.set	= nfs4_xattr_set_nfs4_sacl,
10729 };
10730 #endif
10731 
10732 #ifdef CONFIG_NFS_V4_2
10733 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10734 	.prefix	= XATTR_USER_PREFIX,
10735 	.get	= nfs4_xattr_get_nfs4_user,
10736 	.set	= nfs4_xattr_set_nfs4_user,
10737 };
10738 #endif
10739 
10740 const struct xattr_handler *nfs4_xattr_handlers[] = {
10741 	&nfs4_xattr_nfs4_acl_handler,
10742 #if defined(CONFIG_NFS_V4_1)
10743 	&nfs4_xattr_nfs4_dacl_handler,
10744 	&nfs4_xattr_nfs4_sacl_handler,
10745 #endif
10746 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10747 	&nfs4_xattr_nfs4_label_handler,
10748 #endif
10749 #ifdef CONFIG_NFS_V4_2
10750 	&nfs4_xattr_nfs4_user_handler,
10751 #endif
10752 	NULL
10753 };
10754