xref: /linux/fs/nfs/nfs4proc.c (revision bb9707077b4ee5f77bc9939b057ff8a0d410296f)
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/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.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 "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69 
70 #include "nfs4trace.h"
71 
72 #define NFSDBG_FACILITY		NFSDBG_PROC
73 
74 #define NFS4_POLL_RETRY_MIN	(HZ/10)
75 #define NFS4_POLL_RETRY_MAX	(15*HZ)
76 
77 /* file attributes which can be mapped to nfs attributes */
78 #define NFS4_VALID_ATTRS (ATTR_MODE \
79 	| ATTR_UID \
80 	| ATTR_GID \
81 	| ATTR_SIZE \
82 	| ATTR_ATIME \
83 	| ATTR_MTIME \
84 	| ATTR_CTIME \
85 	| ATTR_ATIME_SET \
86 	| ATTR_MTIME_SET)
87 
88 struct nfs4_opendata;
89 static int _nfs4_proc_open(struct nfs4_opendata *data);
90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
96 			    struct nfs_fattr *fattr, struct iattr *sattr,
97 			    struct nfs4_state *state, struct nfs4_label *ilabel,
98 			    struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
101 		struct rpc_cred *);
102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
103 		struct rpc_cred *);
104 #endif
105 
106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
107 static inline struct nfs4_label *
108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
109 	struct iattr *sattr, struct nfs4_label *label)
110 {
111 	int err;
112 
113 	if (label == NULL)
114 		return NULL;
115 
116 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
117 		return NULL;
118 
119 	err = security_dentry_init_security(dentry, sattr->ia_mode,
120 				&dentry->d_name, (void **)&label->label, &label->len);
121 	if (err == 0)
122 		return label;
123 
124 	return NULL;
125 }
126 static inline void
127 nfs4_label_release_security(struct nfs4_label *label)
128 {
129 	if (label)
130 		security_release_secctx(label->label, label->len);
131 }
132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
133 {
134 	if (label)
135 		return server->attr_bitmask;
136 
137 	return server->attr_bitmask_nl;
138 }
139 #else
140 static inline struct nfs4_label *
141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
142 	struct iattr *sattr, struct nfs4_label *l)
143 { return NULL; }
144 static inline void
145 nfs4_label_release_security(struct nfs4_label *label)
146 { return; }
147 static inline u32 *
148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
149 { return server->attr_bitmask; }
150 #endif
151 
152 /* Prevent leaks of NFSv4 errors into userland */
153 static int nfs4_map_errors(int err)
154 {
155 	if (err >= -1000)
156 		return err;
157 	switch (err) {
158 	case -NFS4ERR_RESOURCE:
159 	case -NFS4ERR_LAYOUTTRYLATER:
160 	case -NFS4ERR_RECALLCONFLICT:
161 		return -EREMOTEIO;
162 	case -NFS4ERR_WRONGSEC:
163 	case -NFS4ERR_WRONG_CRED:
164 		return -EPERM;
165 	case -NFS4ERR_BADOWNER:
166 	case -NFS4ERR_BADNAME:
167 		return -EINVAL;
168 	case -NFS4ERR_SHARE_DENIED:
169 		return -EACCES;
170 	case -NFS4ERR_MINOR_VERS_MISMATCH:
171 		return -EPROTONOSUPPORT;
172 	case -NFS4ERR_FILE_OPEN:
173 		return -EBUSY;
174 	default:
175 		dprintk("%s could not handle NFSv4 error %d\n",
176 				__func__, -err);
177 		break;
178 	}
179 	return -EIO;
180 }
181 
182 /*
183  * This is our standard bitmap for GETATTR requests.
184  */
185 const u32 nfs4_fattr_bitmap[3] = {
186 	FATTR4_WORD0_TYPE
187 	| FATTR4_WORD0_CHANGE
188 	| FATTR4_WORD0_SIZE
189 	| FATTR4_WORD0_FSID
190 	| FATTR4_WORD0_FILEID,
191 	FATTR4_WORD1_MODE
192 	| FATTR4_WORD1_NUMLINKS
193 	| FATTR4_WORD1_OWNER
194 	| FATTR4_WORD1_OWNER_GROUP
195 	| FATTR4_WORD1_RAWDEV
196 	| FATTR4_WORD1_SPACE_USED
197 	| FATTR4_WORD1_TIME_ACCESS
198 	| FATTR4_WORD1_TIME_METADATA
199 	| FATTR4_WORD1_TIME_MODIFY
200 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
202 	FATTR4_WORD2_SECURITY_LABEL
203 #endif
204 };
205 
206 static const u32 nfs4_pnfs_open_bitmap[3] = {
207 	FATTR4_WORD0_TYPE
208 	| FATTR4_WORD0_CHANGE
209 	| FATTR4_WORD0_SIZE
210 	| FATTR4_WORD0_FSID
211 	| FATTR4_WORD0_FILEID,
212 	FATTR4_WORD1_MODE
213 	| FATTR4_WORD1_NUMLINKS
214 	| FATTR4_WORD1_OWNER
215 	| FATTR4_WORD1_OWNER_GROUP
216 	| FATTR4_WORD1_RAWDEV
217 	| FATTR4_WORD1_SPACE_USED
218 	| FATTR4_WORD1_TIME_ACCESS
219 	| FATTR4_WORD1_TIME_METADATA
220 	| FATTR4_WORD1_TIME_MODIFY,
221 	FATTR4_WORD2_MDSTHRESHOLD
222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
223 	| FATTR4_WORD2_SECURITY_LABEL
224 #endif
225 };
226 
227 static const u32 nfs4_open_noattr_bitmap[3] = {
228 	FATTR4_WORD0_TYPE
229 	| FATTR4_WORD0_CHANGE
230 	| FATTR4_WORD0_FILEID,
231 };
232 
233 const u32 nfs4_statfs_bitmap[3] = {
234 	FATTR4_WORD0_FILES_AVAIL
235 	| FATTR4_WORD0_FILES_FREE
236 	| FATTR4_WORD0_FILES_TOTAL,
237 	FATTR4_WORD1_SPACE_AVAIL
238 	| FATTR4_WORD1_SPACE_FREE
239 	| FATTR4_WORD1_SPACE_TOTAL
240 };
241 
242 const u32 nfs4_pathconf_bitmap[3] = {
243 	FATTR4_WORD0_MAXLINK
244 	| FATTR4_WORD0_MAXNAME,
245 	0
246 };
247 
248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
249 			| FATTR4_WORD0_MAXREAD
250 			| FATTR4_WORD0_MAXWRITE
251 			| FATTR4_WORD0_LEASE_TIME,
252 			FATTR4_WORD1_TIME_DELTA
253 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
254 			FATTR4_WORD2_LAYOUT_BLKSIZE
255 			| FATTR4_WORD2_CLONE_BLKSIZE
256 };
257 
258 const u32 nfs4_fs_locations_bitmap[3] = {
259 	FATTR4_WORD0_TYPE
260 	| FATTR4_WORD0_CHANGE
261 	| FATTR4_WORD0_SIZE
262 	| FATTR4_WORD0_FSID
263 	| FATTR4_WORD0_FILEID
264 	| FATTR4_WORD0_FS_LOCATIONS,
265 	FATTR4_WORD1_MODE
266 	| FATTR4_WORD1_NUMLINKS
267 	| FATTR4_WORD1_OWNER
268 	| FATTR4_WORD1_OWNER_GROUP
269 	| FATTR4_WORD1_RAWDEV
270 	| FATTR4_WORD1_SPACE_USED
271 	| FATTR4_WORD1_TIME_ACCESS
272 	| FATTR4_WORD1_TIME_METADATA
273 	| FATTR4_WORD1_TIME_MODIFY
274 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
275 };
276 
277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
278 		struct nfs4_readdir_arg *readdir)
279 {
280 	__be32 *start, *p;
281 
282 	if (cookie > 2) {
283 		readdir->cookie = cookie;
284 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
285 		return;
286 	}
287 
288 	readdir->cookie = 0;
289 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
290 	if (cookie == 2)
291 		return;
292 
293 	/*
294 	 * NFSv4 servers do not return entries for '.' and '..'
295 	 * Therefore, we fake these entries here.  We let '.'
296 	 * have cookie 0 and '..' have cookie 1.  Note that
297 	 * when talking to the server, we always send cookie 0
298 	 * instead of 1 or 2.
299 	 */
300 	start = p = kmap_atomic(*readdir->pages);
301 
302 	if (cookie == 0) {
303 		*p++ = xdr_one;                                  /* next */
304 		*p++ = xdr_zero;                   /* cookie, first word */
305 		*p++ = xdr_one;                   /* cookie, second word */
306 		*p++ = xdr_one;                             /* entry len */
307 		memcpy(p, ".\0\0\0", 4);                        /* entry */
308 		p++;
309 		*p++ = xdr_one;                         /* bitmap length */
310 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
311 		*p++ = htonl(8);              /* attribute buffer length */
312 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
313 	}
314 
315 	*p++ = xdr_one;                                  /* next */
316 	*p++ = xdr_zero;                   /* cookie, first word */
317 	*p++ = xdr_two;                   /* cookie, second word */
318 	*p++ = xdr_two;                             /* entry len */
319 	memcpy(p, "..\0\0", 4);                         /* entry */
320 	p++;
321 	*p++ = xdr_one;                         /* bitmap length */
322 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
323 	*p++ = htonl(8);              /* attribute buffer length */
324 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
325 
326 	readdir->pgbase = (char *)p - (char *)start;
327 	readdir->count -= readdir->pgbase;
328 	kunmap_atomic(start);
329 }
330 
331 static long nfs4_update_delay(long *timeout)
332 {
333 	long ret;
334 	if (!timeout)
335 		return NFS4_POLL_RETRY_MAX;
336 	if (*timeout <= 0)
337 		*timeout = NFS4_POLL_RETRY_MIN;
338 	if (*timeout > NFS4_POLL_RETRY_MAX)
339 		*timeout = NFS4_POLL_RETRY_MAX;
340 	ret = *timeout;
341 	*timeout <<= 1;
342 	return ret;
343 }
344 
345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
346 {
347 	int res = 0;
348 
349 	might_sleep();
350 
351 	freezable_schedule_timeout_killable_unsafe(
352 		nfs4_update_delay(timeout));
353 	if (fatal_signal_pending(current))
354 		res = -ERESTARTSYS;
355 	return res;
356 }
357 
358 /* This is the error handling routine for processes that are allowed
359  * to sleep.
360  */
361 static int nfs4_do_handle_exception(struct nfs_server *server,
362 		int errorcode, struct nfs4_exception *exception)
363 {
364 	struct nfs_client *clp = server->nfs_client;
365 	struct nfs4_state *state = exception->state;
366 	struct inode *inode = exception->inode;
367 	int ret = errorcode;
368 
369 	exception->delay = 0;
370 	exception->recovering = 0;
371 	exception->retry = 0;
372 	switch(errorcode) {
373 		case 0:
374 			return 0;
375 		case -NFS4ERR_OPENMODE:
376 		case -NFS4ERR_DELEG_REVOKED:
377 		case -NFS4ERR_ADMIN_REVOKED:
378 		case -NFS4ERR_BAD_STATEID:
379 			if (inode && nfs_async_inode_return_delegation(inode,
380 						NULL) == 0)
381 				goto wait_on_recovery;
382 			if (state == NULL)
383 				break;
384 			ret = nfs4_schedule_stateid_recovery(server, state);
385 			if (ret < 0)
386 				break;
387 			goto wait_on_recovery;
388 		case -NFS4ERR_EXPIRED:
389 			if (state != NULL) {
390 				ret = nfs4_schedule_stateid_recovery(server, state);
391 				if (ret < 0)
392 					break;
393 			}
394 		case -NFS4ERR_STALE_STATEID:
395 		case -NFS4ERR_STALE_CLIENTID:
396 			nfs4_schedule_lease_recovery(clp);
397 			goto wait_on_recovery;
398 		case -NFS4ERR_MOVED:
399 			ret = nfs4_schedule_migration_recovery(server);
400 			if (ret < 0)
401 				break;
402 			goto wait_on_recovery;
403 		case -NFS4ERR_LEASE_MOVED:
404 			nfs4_schedule_lease_moved_recovery(clp);
405 			goto wait_on_recovery;
406 #if defined(CONFIG_NFS_V4_1)
407 		case -NFS4ERR_BADSESSION:
408 		case -NFS4ERR_BADSLOT:
409 		case -NFS4ERR_BAD_HIGH_SLOT:
410 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
411 		case -NFS4ERR_DEADSESSION:
412 		case -NFS4ERR_SEQ_FALSE_RETRY:
413 		case -NFS4ERR_SEQ_MISORDERED:
414 			dprintk("%s ERROR: %d Reset session\n", __func__,
415 				errorcode);
416 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
417 			goto wait_on_recovery;
418 #endif /* defined(CONFIG_NFS_V4_1) */
419 		case -NFS4ERR_FILE_OPEN:
420 			if (exception->timeout > HZ) {
421 				/* We have retried a decent amount, time to
422 				 * fail
423 				 */
424 				ret = -EBUSY;
425 				break;
426 			}
427 		case -NFS4ERR_DELAY:
428 			nfs_inc_server_stats(server, NFSIOS_DELAY);
429 		case -NFS4ERR_GRACE:
430 		case -NFS4ERR_RECALLCONFLICT:
431 			exception->delay = 1;
432 			return 0;
433 
434 		case -NFS4ERR_RETRY_UNCACHED_REP:
435 		case -NFS4ERR_OLD_STATEID:
436 			exception->retry = 1;
437 			break;
438 		case -NFS4ERR_BADOWNER:
439 			/* The following works around a Linux server bug! */
440 		case -NFS4ERR_BADNAME:
441 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
442 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
443 				exception->retry = 1;
444 				printk(KERN_WARNING "NFS: v4 server %s "
445 						"does not accept raw "
446 						"uid/gids. "
447 						"Reenabling the idmapper.\n",
448 						server->nfs_client->cl_hostname);
449 			}
450 	}
451 	/* We failed to handle the error */
452 	return nfs4_map_errors(ret);
453 wait_on_recovery:
454 	exception->recovering = 1;
455 	return 0;
456 }
457 
458 /* This is the error handling routine for processes that are allowed
459  * to sleep.
460  */
461 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
462 {
463 	struct nfs_client *clp = server->nfs_client;
464 	int ret;
465 
466 	ret = nfs4_do_handle_exception(server, errorcode, exception);
467 	if (exception->delay) {
468 		ret = nfs4_delay(server->client, &exception->timeout);
469 		goto out_retry;
470 	}
471 	if (exception->recovering) {
472 		ret = nfs4_wait_clnt_recover(clp);
473 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
474 			return -EIO;
475 		goto out_retry;
476 	}
477 	return ret;
478 out_retry:
479 	if (ret == 0)
480 		exception->retry = 1;
481 	return ret;
482 }
483 
484 static int
485 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
486 		int errorcode, struct nfs4_exception *exception)
487 {
488 	struct nfs_client *clp = server->nfs_client;
489 	int ret;
490 
491 	ret = nfs4_do_handle_exception(server, errorcode, exception);
492 	if (exception->delay) {
493 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
494 		goto out_retry;
495 	}
496 	if (exception->recovering) {
497 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
498 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
499 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
500 		goto out_retry;
501 	}
502 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
503 		ret = -EIO;
504 	return ret;
505 out_retry:
506 	if (ret == 0)
507 		exception->retry = 1;
508 	return ret;
509 }
510 
511 static int
512 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
513 			struct nfs4_state *state, long *timeout)
514 {
515 	struct nfs4_exception exception = {
516 		.state = state,
517 	};
518 
519 	if (task->tk_status >= 0)
520 		return 0;
521 	if (timeout)
522 		exception.timeout = *timeout;
523 	task->tk_status = nfs4_async_handle_exception(task, server,
524 			task->tk_status,
525 			&exception);
526 	if (exception.delay && timeout)
527 		*timeout = exception.timeout;
528 	if (exception.retry)
529 		return -EAGAIN;
530 	return 0;
531 }
532 
533 /*
534  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
535  * or 'false' otherwise.
536  */
537 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
538 {
539 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
540 
541 	if (flavor == RPC_AUTH_GSS_KRB5I ||
542 	    flavor == RPC_AUTH_GSS_KRB5P)
543 		return true;
544 
545 	return false;
546 }
547 
548 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
549 {
550 	spin_lock(&clp->cl_lock);
551 	if (time_before(clp->cl_last_renewal,timestamp))
552 		clp->cl_last_renewal = timestamp;
553 	spin_unlock(&clp->cl_lock);
554 }
555 
556 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
557 {
558 	struct nfs_client *clp = server->nfs_client;
559 
560 	if (!nfs4_has_session(clp))
561 		do_renew_lease(clp, timestamp);
562 }
563 
564 struct nfs4_call_sync_data {
565 	const struct nfs_server *seq_server;
566 	struct nfs4_sequence_args *seq_args;
567 	struct nfs4_sequence_res *seq_res;
568 };
569 
570 void nfs4_init_sequence(struct nfs4_sequence_args *args,
571 			struct nfs4_sequence_res *res, int cache_reply)
572 {
573 	args->sa_slot = NULL;
574 	args->sa_cache_this = cache_reply;
575 	args->sa_privileged = 0;
576 
577 	res->sr_slot = NULL;
578 }
579 
580 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
581 {
582 	args->sa_privileged = 1;
583 }
584 
585 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
586 			 struct nfs4_sequence_args *args,
587 			 struct nfs4_sequence_res *res,
588 			 struct rpc_task *task)
589 {
590 	struct nfs4_slot *slot;
591 
592 	/* slot already allocated? */
593 	if (res->sr_slot != NULL)
594 		goto out_start;
595 
596 	spin_lock(&tbl->slot_tbl_lock);
597 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
598 		goto out_sleep;
599 
600 	slot = nfs4_alloc_slot(tbl);
601 	if (IS_ERR(slot)) {
602 		if (slot == ERR_PTR(-ENOMEM))
603 			task->tk_timeout = HZ >> 2;
604 		goto out_sleep;
605 	}
606 	spin_unlock(&tbl->slot_tbl_lock);
607 
608 	args->sa_slot = slot;
609 	res->sr_slot = slot;
610 
611 out_start:
612 	rpc_call_start(task);
613 	return 0;
614 
615 out_sleep:
616 	if (args->sa_privileged)
617 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
618 				NULL, RPC_PRIORITY_PRIVILEGED);
619 	else
620 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
621 	spin_unlock(&tbl->slot_tbl_lock);
622 	return -EAGAIN;
623 }
624 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
625 
626 static int nfs40_sequence_done(struct rpc_task *task,
627 			       struct nfs4_sequence_res *res)
628 {
629 	struct nfs4_slot *slot = res->sr_slot;
630 	struct nfs4_slot_table *tbl;
631 
632 	if (slot == NULL)
633 		goto out;
634 
635 	tbl = slot->table;
636 	spin_lock(&tbl->slot_tbl_lock);
637 	if (!nfs41_wake_and_assign_slot(tbl, slot))
638 		nfs4_free_slot(tbl, slot);
639 	spin_unlock(&tbl->slot_tbl_lock);
640 
641 	res->sr_slot = NULL;
642 out:
643 	return 1;
644 }
645 
646 #if defined(CONFIG_NFS_V4_1)
647 
648 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
649 {
650 	struct nfs4_session *session;
651 	struct nfs4_slot_table *tbl;
652 	struct nfs4_slot *slot = res->sr_slot;
653 	bool send_new_highest_used_slotid = false;
654 
655 	tbl = slot->table;
656 	session = tbl->session;
657 
658 	spin_lock(&tbl->slot_tbl_lock);
659 	/* Be nice to the server: try to ensure that the last transmitted
660 	 * value for highest_user_slotid <= target_highest_slotid
661 	 */
662 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
663 		send_new_highest_used_slotid = true;
664 
665 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
666 		send_new_highest_used_slotid = false;
667 		goto out_unlock;
668 	}
669 	nfs4_free_slot(tbl, slot);
670 
671 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
672 		send_new_highest_used_slotid = false;
673 out_unlock:
674 	spin_unlock(&tbl->slot_tbl_lock);
675 	res->sr_slot = NULL;
676 	if (send_new_highest_used_slotid)
677 		nfs41_notify_server(session->clp);
678 }
679 
680 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
681 {
682 	struct nfs4_session *session;
683 	struct nfs4_slot *slot = res->sr_slot;
684 	struct nfs_client *clp;
685 	bool interrupted = false;
686 	int ret = 1;
687 
688 	if (slot == NULL)
689 		goto out_noaction;
690 	/* don't increment the sequence number if the task wasn't sent */
691 	if (!RPC_WAS_SENT(task))
692 		goto out;
693 
694 	session = slot->table->session;
695 
696 	if (slot->interrupted) {
697 		slot->interrupted = 0;
698 		interrupted = true;
699 	}
700 
701 	trace_nfs4_sequence_done(session, res);
702 	/* Check the SEQUENCE operation status */
703 	switch (res->sr_status) {
704 	case 0:
705 		/* Update the slot's sequence and clientid lease timer */
706 		++slot->seq_nr;
707 		clp = session->clp;
708 		do_renew_lease(clp, res->sr_timestamp);
709 		/* Check sequence flags */
710 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
711 		nfs41_update_target_slotid(slot->table, slot, res);
712 		break;
713 	case 1:
714 		/*
715 		 * sr_status remains 1 if an RPC level error occurred.
716 		 * The server may or may not have processed the sequence
717 		 * operation..
718 		 * Mark the slot as having hosted an interrupted RPC call.
719 		 */
720 		slot->interrupted = 1;
721 		goto out;
722 	case -NFS4ERR_DELAY:
723 		/* The server detected a resend of the RPC call and
724 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
725 		 * of RFC5661.
726 		 */
727 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
728 			__func__,
729 			slot->slot_nr,
730 			slot->seq_nr);
731 		goto out_retry;
732 	case -NFS4ERR_BADSLOT:
733 		/*
734 		 * The slot id we used was probably retired. Try again
735 		 * using a different slot id.
736 		 */
737 		goto retry_nowait;
738 	case -NFS4ERR_SEQ_MISORDERED:
739 		/*
740 		 * Was the last operation on this sequence interrupted?
741 		 * If so, retry after bumping the sequence number.
742 		 */
743 		if (interrupted) {
744 			++slot->seq_nr;
745 			goto retry_nowait;
746 		}
747 		/*
748 		 * Could this slot have been previously retired?
749 		 * If so, then the server may be expecting seq_nr = 1!
750 		 */
751 		if (slot->seq_nr != 1) {
752 			slot->seq_nr = 1;
753 			goto retry_nowait;
754 		}
755 		break;
756 	case -NFS4ERR_SEQ_FALSE_RETRY:
757 		++slot->seq_nr;
758 		goto retry_nowait;
759 	default:
760 		/* Just update the slot sequence no. */
761 		++slot->seq_nr;
762 	}
763 out:
764 	/* The session may be reset by one of the error handlers. */
765 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
766 	nfs41_sequence_free_slot(res);
767 out_noaction:
768 	return ret;
769 retry_nowait:
770 	if (rpc_restart_call_prepare(task)) {
771 		task->tk_status = 0;
772 		ret = 0;
773 	}
774 	goto out;
775 out_retry:
776 	if (!rpc_restart_call(task))
777 		goto out;
778 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
779 	return 0;
780 }
781 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
782 
783 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
784 {
785 	if (res->sr_slot == NULL)
786 		return 1;
787 	if (!res->sr_slot->table->session)
788 		return nfs40_sequence_done(task, res);
789 	return nfs41_sequence_done(task, res);
790 }
791 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
792 
793 int nfs41_setup_sequence(struct nfs4_session *session,
794 				struct nfs4_sequence_args *args,
795 				struct nfs4_sequence_res *res,
796 				struct rpc_task *task)
797 {
798 	struct nfs4_slot *slot;
799 	struct nfs4_slot_table *tbl;
800 
801 	dprintk("--> %s\n", __func__);
802 	/* slot already allocated? */
803 	if (res->sr_slot != NULL)
804 		goto out_success;
805 
806 	tbl = &session->fc_slot_table;
807 
808 	task->tk_timeout = 0;
809 
810 	spin_lock(&tbl->slot_tbl_lock);
811 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
812 	    !args->sa_privileged) {
813 		/* The state manager will wait until the slot table is empty */
814 		dprintk("%s session is draining\n", __func__);
815 		goto out_sleep;
816 	}
817 
818 	slot = nfs4_alloc_slot(tbl);
819 	if (IS_ERR(slot)) {
820 		/* If out of memory, try again in 1/4 second */
821 		if (slot == ERR_PTR(-ENOMEM))
822 			task->tk_timeout = HZ >> 2;
823 		dprintk("<-- %s: no free slots\n", __func__);
824 		goto out_sleep;
825 	}
826 	spin_unlock(&tbl->slot_tbl_lock);
827 
828 	args->sa_slot = slot;
829 
830 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
831 			slot->slot_nr, slot->seq_nr);
832 
833 	res->sr_slot = slot;
834 	res->sr_timestamp = jiffies;
835 	res->sr_status_flags = 0;
836 	/*
837 	 * sr_status is only set in decode_sequence, and so will remain
838 	 * set to 1 if an rpc level failure occurs.
839 	 */
840 	res->sr_status = 1;
841 	trace_nfs4_setup_sequence(session, args);
842 out_success:
843 	rpc_call_start(task);
844 	return 0;
845 out_sleep:
846 	/* Privileged tasks are queued with top priority */
847 	if (args->sa_privileged)
848 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
849 				NULL, RPC_PRIORITY_PRIVILEGED);
850 	else
851 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
852 	spin_unlock(&tbl->slot_tbl_lock);
853 	return -EAGAIN;
854 }
855 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
856 
857 static int nfs4_setup_sequence(const struct nfs_server *server,
858 			       struct nfs4_sequence_args *args,
859 			       struct nfs4_sequence_res *res,
860 			       struct rpc_task *task)
861 {
862 	struct nfs4_session *session = nfs4_get_session(server);
863 	int ret = 0;
864 
865 	if (!session)
866 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
867 					    args, res, task);
868 
869 	dprintk("--> %s clp %p session %p sr_slot %u\n",
870 		__func__, session->clp, session, res->sr_slot ?
871 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
872 
873 	ret = nfs41_setup_sequence(session, args, res, task);
874 
875 	dprintk("<-- %s status=%d\n", __func__, ret);
876 	return ret;
877 }
878 
879 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
880 {
881 	struct nfs4_call_sync_data *data = calldata;
882 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
883 
884 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
885 
886 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
887 }
888 
889 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
890 {
891 	struct nfs4_call_sync_data *data = calldata;
892 
893 	nfs41_sequence_done(task, data->seq_res);
894 }
895 
896 static const struct rpc_call_ops nfs41_call_sync_ops = {
897 	.rpc_call_prepare = nfs41_call_sync_prepare,
898 	.rpc_call_done = nfs41_call_sync_done,
899 };
900 
901 #else	/* !CONFIG_NFS_V4_1 */
902 
903 static int nfs4_setup_sequence(const struct nfs_server *server,
904 			       struct nfs4_sequence_args *args,
905 			       struct nfs4_sequence_res *res,
906 			       struct rpc_task *task)
907 {
908 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
909 				    args, res, task);
910 }
911 
912 int nfs4_sequence_done(struct rpc_task *task,
913 		       struct nfs4_sequence_res *res)
914 {
915 	return nfs40_sequence_done(task, res);
916 }
917 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
918 
919 #endif	/* !CONFIG_NFS_V4_1 */
920 
921 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
922 {
923 	struct nfs4_call_sync_data *data = calldata;
924 	nfs4_setup_sequence(data->seq_server,
925 				data->seq_args, data->seq_res, task);
926 }
927 
928 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
929 {
930 	struct nfs4_call_sync_data *data = calldata;
931 	nfs4_sequence_done(task, data->seq_res);
932 }
933 
934 static const struct rpc_call_ops nfs40_call_sync_ops = {
935 	.rpc_call_prepare = nfs40_call_sync_prepare,
936 	.rpc_call_done = nfs40_call_sync_done,
937 };
938 
939 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
940 				   struct nfs_server *server,
941 				   struct rpc_message *msg,
942 				   struct nfs4_sequence_args *args,
943 				   struct nfs4_sequence_res *res)
944 {
945 	int ret;
946 	struct rpc_task *task;
947 	struct nfs_client *clp = server->nfs_client;
948 	struct nfs4_call_sync_data data = {
949 		.seq_server = server,
950 		.seq_args = args,
951 		.seq_res = res,
952 	};
953 	struct rpc_task_setup task_setup = {
954 		.rpc_client = clnt,
955 		.rpc_message = msg,
956 		.callback_ops = clp->cl_mvops->call_sync_ops,
957 		.callback_data = &data
958 	};
959 
960 	task = rpc_run_task(&task_setup);
961 	if (IS_ERR(task))
962 		ret = PTR_ERR(task);
963 	else {
964 		ret = task->tk_status;
965 		rpc_put_task(task);
966 	}
967 	return ret;
968 }
969 
970 int nfs4_call_sync(struct rpc_clnt *clnt,
971 		   struct nfs_server *server,
972 		   struct rpc_message *msg,
973 		   struct nfs4_sequence_args *args,
974 		   struct nfs4_sequence_res *res,
975 		   int cache_reply)
976 {
977 	nfs4_init_sequence(args, res, cache_reply);
978 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
979 }
980 
981 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
982 {
983 	struct nfs_inode *nfsi = NFS_I(dir);
984 
985 	spin_lock(&dir->i_lock);
986 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
987 	if (!cinfo->atomic || cinfo->before != dir->i_version)
988 		nfs_force_lookup_revalidate(dir);
989 	dir->i_version = cinfo->after;
990 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
991 	nfs_fscache_invalidate(dir);
992 	spin_unlock(&dir->i_lock);
993 }
994 
995 struct nfs4_opendata {
996 	struct kref kref;
997 	struct nfs_openargs o_arg;
998 	struct nfs_openres o_res;
999 	struct nfs_open_confirmargs c_arg;
1000 	struct nfs_open_confirmres c_res;
1001 	struct nfs4_string owner_name;
1002 	struct nfs4_string group_name;
1003 	struct nfs4_label *a_label;
1004 	struct nfs_fattr f_attr;
1005 	struct nfs4_label *f_label;
1006 	struct dentry *dir;
1007 	struct dentry *dentry;
1008 	struct nfs4_state_owner *owner;
1009 	struct nfs4_state *state;
1010 	struct iattr attrs;
1011 	unsigned long timestamp;
1012 	unsigned int rpc_done : 1;
1013 	unsigned int file_created : 1;
1014 	unsigned int is_recover : 1;
1015 	int rpc_status;
1016 	int cancelled;
1017 };
1018 
1019 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1020 		int err, struct nfs4_exception *exception)
1021 {
1022 	if (err != -EINVAL)
1023 		return false;
1024 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1025 		return false;
1026 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1027 	exception->retry = 1;
1028 	return true;
1029 }
1030 
1031 static u32
1032 nfs4_map_atomic_open_share(struct nfs_server *server,
1033 		fmode_t fmode, int openflags)
1034 {
1035 	u32 res = 0;
1036 
1037 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1038 	case FMODE_READ:
1039 		res = NFS4_SHARE_ACCESS_READ;
1040 		break;
1041 	case FMODE_WRITE:
1042 		res = NFS4_SHARE_ACCESS_WRITE;
1043 		break;
1044 	case FMODE_READ|FMODE_WRITE:
1045 		res = NFS4_SHARE_ACCESS_BOTH;
1046 	}
1047 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1048 		goto out;
1049 	/* Want no delegation if we're using O_DIRECT */
1050 	if (openflags & O_DIRECT)
1051 		res |= NFS4_SHARE_WANT_NO_DELEG;
1052 out:
1053 	return res;
1054 }
1055 
1056 static enum open_claim_type4
1057 nfs4_map_atomic_open_claim(struct nfs_server *server,
1058 		enum open_claim_type4 claim)
1059 {
1060 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1061 		return claim;
1062 	switch (claim) {
1063 	default:
1064 		return claim;
1065 	case NFS4_OPEN_CLAIM_FH:
1066 		return NFS4_OPEN_CLAIM_NULL;
1067 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1068 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1069 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1070 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1071 	}
1072 }
1073 
1074 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1075 {
1076 	p->o_res.f_attr = &p->f_attr;
1077 	p->o_res.f_label = p->f_label;
1078 	p->o_res.seqid = p->o_arg.seqid;
1079 	p->c_res.seqid = p->c_arg.seqid;
1080 	p->o_res.server = p->o_arg.server;
1081 	p->o_res.access_request = p->o_arg.access;
1082 	nfs_fattr_init(&p->f_attr);
1083 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1084 }
1085 
1086 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1087 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1088 		const struct iattr *attrs,
1089 		struct nfs4_label *label,
1090 		enum open_claim_type4 claim,
1091 		gfp_t gfp_mask)
1092 {
1093 	struct dentry *parent = dget_parent(dentry);
1094 	struct inode *dir = d_inode(parent);
1095 	struct nfs_server *server = NFS_SERVER(dir);
1096 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1097 	struct nfs4_opendata *p;
1098 
1099 	p = kzalloc(sizeof(*p), gfp_mask);
1100 	if (p == NULL)
1101 		goto err;
1102 
1103 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1104 	if (IS_ERR(p->f_label))
1105 		goto err_free_p;
1106 
1107 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1108 	if (IS_ERR(p->a_label))
1109 		goto err_free_f;
1110 
1111 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1112 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1113 	if (IS_ERR(p->o_arg.seqid))
1114 		goto err_free_label;
1115 	nfs_sb_active(dentry->d_sb);
1116 	p->dentry = dget(dentry);
1117 	p->dir = parent;
1118 	p->owner = sp;
1119 	atomic_inc(&sp->so_count);
1120 	p->o_arg.open_flags = flags;
1121 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1122 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1123 			fmode, flags);
1124 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1125 	 * will return permission denied for all bits until close */
1126 	if (!(flags & O_EXCL)) {
1127 		/* ask server to check for all possible rights as results
1128 		 * are cached */
1129 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1130 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1131 	}
1132 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1133 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1134 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1135 	p->o_arg.name = &dentry->d_name;
1136 	p->o_arg.server = server;
1137 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1138 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1139 	p->o_arg.label = nfs4_label_copy(p->a_label, label);
1140 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1141 	switch (p->o_arg.claim) {
1142 	case NFS4_OPEN_CLAIM_NULL:
1143 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1144 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1145 		p->o_arg.fh = NFS_FH(dir);
1146 		break;
1147 	case NFS4_OPEN_CLAIM_PREVIOUS:
1148 	case NFS4_OPEN_CLAIM_FH:
1149 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1150 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1151 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1152 	}
1153 	if (attrs != NULL && attrs->ia_valid != 0) {
1154 		__u32 verf[2];
1155 
1156 		p->o_arg.u.attrs = &p->attrs;
1157 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1158 
1159 		verf[0] = jiffies;
1160 		verf[1] = current->pid;
1161 		memcpy(p->o_arg.u.verifier.data, verf,
1162 				sizeof(p->o_arg.u.verifier.data));
1163 	}
1164 	p->c_arg.fh = &p->o_res.fh;
1165 	p->c_arg.stateid = &p->o_res.stateid;
1166 	p->c_arg.seqid = p->o_arg.seqid;
1167 	nfs4_init_opendata_res(p);
1168 	kref_init(&p->kref);
1169 	return p;
1170 
1171 err_free_label:
1172 	nfs4_label_free(p->a_label);
1173 err_free_f:
1174 	nfs4_label_free(p->f_label);
1175 err_free_p:
1176 	kfree(p);
1177 err:
1178 	dput(parent);
1179 	return NULL;
1180 }
1181 
1182 static void nfs4_opendata_free(struct kref *kref)
1183 {
1184 	struct nfs4_opendata *p = container_of(kref,
1185 			struct nfs4_opendata, kref);
1186 	struct super_block *sb = p->dentry->d_sb;
1187 
1188 	nfs_free_seqid(p->o_arg.seqid);
1189 	if (p->state != NULL)
1190 		nfs4_put_open_state(p->state);
1191 	nfs4_put_state_owner(p->owner);
1192 
1193 	nfs4_label_free(p->a_label);
1194 	nfs4_label_free(p->f_label);
1195 
1196 	dput(p->dir);
1197 	dput(p->dentry);
1198 	nfs_sb_deactive(sb);
1199 	nfs_fattr_free_names(&p->f_attr);
1200 	kfree(p->f_attr.mdsthreshold);
1201 	kfree(p);
1202 }
1203 
1204 static void nfs4_opendata_put(struct nfs4_opendata *p)
1205 {
1206 	if (p != NULL)
1207 		kref_put(&p->kref, nfs4_opendata_free);
1208 }
1209 
1210 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1211 {
1212 	int ret;
1213 
1214 	ret = rpc_wait_for_completion_task(task);
1215 	return ret;
1216 }
1217 
1218 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1219 		fmode_t fmode)
1220 {
1221 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1222 	case FMODE_READ|FMODE_WRITE:
1223 		return state->n_rdwr != 0;
1224 	case FMODE_WRITE:
1225 		return state->n_wronly != 0;
1226 	case FMODE_READ:
1227 		return state->n_rdonly != 0;
1228 	}
1229 	WARN_ON_ONCE(1);
1230 	return false;
1231 }
1232 
1233 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1234 {
1235 	int ret = 0;
1236 
1237 	if (open_mode & (O_EXCL|O_TRUNC))
1238 		goto out;
1239 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1240 		case FMODE_READ:
1241 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1242 				&& state->n_rdonly != 0;
1243 			break;
1244 		case FMODE_WRITE:
1245 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1246 				&& state->n_wronly != 0;
1247 			break;
1248 		case FMODE_READ|FMODE_WRITE:
1249 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1250 				&& state->n_rdwr != 0;
1251 	}
1252 out:
1253 	return ret;
1254 }
1255 
1256 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1257 		enum open_claim_type4 claim)
1258 {
1259 	if (delegation == NULL)
1260 		return 0;
1261 	if ((delegation->type & fmode) != fmode)
1262 		return 0;
1263 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1264 		return 0;
1265 	switch (claim) {
1266 	case NFS4_OPEN_CLAIM_NULL:
1267 	case NFS4_OPEN_CLAIM_FH:
1268 		break;
1269 	case NFS4_OPEN_CLAIM_PREVIOUS:
1270 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1271 			break;
1272 	default:
1273 		return 0;
1274 	}
1275 	nfs_mark_delegation_referenced(delegation);
1276 	return 1;
1277 }
1278 
1279 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1280 {
1281 	switch (fmode) {
1282 		case FMODE_WRITE:
1283 			state->n_wronly++;
1284 			break;
1285 		case FMODE_READ:
1286 			state->n_rdonly++;
1287 			break;
1288 		case FMODE_READ|FMODE_WRITE:
1289 			state->n_rdwr++;
1290 	}
1291 	nfs4_state_set_mode_locked(state, state->state | fmode);
1292 }
1293 
1294 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1295 {
1296 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1297 	bool need_recover = false;
1298 
1299 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1300 		need_recover = true;
1301 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1302 		need_recover = true;
1303 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1304 		need_recover = true;
1305 	if (need_recover)
1306 		nfs4_state_mark_reclaim_nograce(clp, state);
1307 }
1308 
1309 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1310 		nfs4_stateid *stateid)
1311 {
1312 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1313 		return true;
1314 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1315 		nfs_test_and_clear_all_open_stateid(state);
1316 		return true;
1317 	}
1318 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1319 		return true;
1320 	return false;
1321 }
1322 
1323 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1324 {
1325 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1326 		return;
1327 	if (state->n_wronly)
1328 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1329 	if (state->n_rdonly)
1330 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1331 	if (state->n_rdwr)
1332 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1333 	set_bit(NFS_OPEN_STATE, &state->flags);
1334 }
1335 
1336 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1337 		nfs4_stateid *arg_stateid,
1338 		nfs4_stateid *stateid, fmode_t fmode)
1339 {
1340 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1341 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1342 	case FMODE_WRITE:
1343 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1344 		break;
1345 	case FMODE_READ:
1346 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1347 		break;
1348 	case 0:
1349 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1350 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1351 		clear_bit(NFS_OPEN_STATE, &state->flags);
1352 	}
1353 	if (stateid == NULL)
1354 		return;
1355 	/* Handle races with OPEN */
1356 	if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1357 	    (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1358 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1359 		nfs_resync_open_stateid_locked(state);
1360 		return;
1361 	}
1362 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1363 		nfs4_stateid_copy(&state->stateid, stateid);
1364 	nfs4_stateid_copy(&state->open_stateid, stateid);
1365 }
1366 
1367 static void nfs_clear_open_stateid(struct nfs4_state *state,
1368 	nfs4_stateid *arg_stateid,
1369 	nfs4_stateid *stateid, fmode_t fmode)
1370 {
1371 	write_seqlock(&state->seqlock);
1372 	nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1373 	write_sequnlock(&state->seqlock);
1374 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1375 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1376 }
1377 
1378 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1379 {
1380 	switch (fmode) {
1381 		case FMODE_READ:
1382 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1383 			break;
1384 		case FMODE_WRITE:
1385 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1386 			break;
1387 		case FMODE_READ|FMODE_WRITE:
1388 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1389 	}
1390 	if (!nfs_need_update_open_stateid(state, stateid))
1391 		return;
1392 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1393 		nfs4_stateid_copy(&state->stateid, stateid);
1394 	nfs4_stateid_copy(&state->open_stateid, stateid);
1395 }
1396 
1397 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1398 {
1399 	/*
1400 	 * Protect the call to nfs4_state_set_mode_locked and
1401 	 * serialise the stateid update
1402 	 */
1403 	spin_lock(&state->owner->so_lock);
1404 	write_seqlock(&state->seqlock);
1405 	if (deleg_stateid != NULL) {
1406 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1407 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1408 	}
1409 	if (open_stateid != NULL)
1410 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1411 	write_sequnlock(&state->seqlock);
1412 	update_open_stateflags(state, fmode);
1413 	spin_unlock(&state->owner->so_lock);
1414 }
1415 
1416 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1417 {
1418 	struct nfs_inode *nfsi = NFS_I(state->inode);
1419 	struct nfs_delegation *deleg_cur;
1420 	int ret = 0;
1421 
1422 	fmode &= (FMODE_READ|FMODE_WRITE);
1423 
1424 	rcu_read_lock();
1425 	deleg_cur = rcu_dereference(nfsi->delegation);
1426 	if (deleg_cur == NULL)
1427 		goto no_delegation;
1428 
1429 	spin_lock(&deleg_cur->lock);
1430 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1431 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1432 	    (deleg_cur->type & fmode) != fmode)
1433 		goto no_delegation_unlock;
1434 
1435 	if (delegation == NULL)
1436 		delegation = &deleg_cur->stateid;
1437 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1438 		goto no_delegation_unlock;
1439 
1440 	nfs_mark_delegation_referenced(deleg_cur);
1441 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1442 	ret = 1;
1443 no_delegation_unlock:
1444 	spin_unlock(&deleg_cur->lock);
1445 no_delegation:
1446 	rcu_read_unlock();
1447 
1448 	if (!ret && open_stateid != NULL) {
1449 		__update_open_stateid(state, open_stateid, NULL, fmode);
1450 		ret = 1;
1451 	}
1452 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1453 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1454 
1455 	return ret;
1456 }
1457 
1458 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1459 		const nfs4_stateid *stateid)
1460 {
1461 	struct nfs4_state *state = lsp->ls_state;
1462 	bool ret = false;
1463 
1464 	spin_lock(&state->state_lock);
1465 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1466 		goto out_noupdate;
1467 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1468 		goto out_noupdate;
1469 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1470 	ret = true;
1471 out_noupdate:
1472 	spin_unlock(&state->state_lock);
1473 	return ret;
1474 }
1475 
1476 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1477 {
1478 	struct nfs_delegation *delegation;
1479 
1480 	rcu_read_lock();
1481 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1482 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1483 		rcu_read_unlock();
1484 		return;
1485 	}
1486 	rcu_read_unlock();
1487 	nfs4_inode_return_delegation(inode);
1488 }
1489 
1490 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1491 {
1492 	struct nfs4_state *state = opendata->state;
1493 	struct nfs_inode *nfsi = NFS_I(state->inode);
1494 	struct nfs_delegation *delegation;
1495 	int open_mode = opendata->o_arg.open_flags;
1496 	fmode_t fmode = opendata->o_arg.fmode;
1497 	enum open_claim_type4 claim = opendata->o_arg.claim;
1498 	nfs4_stateid stateid;
1499 	int ret = -EAGAIN;
1500 
1501 	for (;;) {
1502 		spin_lock(&state->owner->so_lock);
1503 		if (can_open_cached(state, fmode, open_mode)) {
1504 			update_open_stateflags(state, fmode);
1505 			spin_unlock(&state->owner->so_lock);
1506 			goto out_return_state;
1507 		}
1508 		spin_unlock(&state->owner->so_lock);
1509 		rcu_read_lock();
1510 		delegation = rcu_dereference(nfsi->delegation);
1511 		if (!can_open_delegated(delegation, fmode, claim)) {
1512 			rcu_read_unlock();
1513 			break;
1514 		}
1515 		/* Save the delegation */
1516 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1517 		rcu_read_unlock();
1518 		nfs_release_seqid(opendata->o_arg.seqid);
1519 		if (!opendata->is_recover) {
1520 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1521 			if (ret != 0)
1522 				goto out;
1523 		}
1524 		ret = -EAGAIN;
1525 
1526 		/* Try to update the stateid using the delegation */
1527 		if (update_open_stateid(state, NULL, &stateid, fmode))
1528 			goto out_return_state;
1529 	}
1530 out:
1531 	return ERR_PTR(ret);
1532 out_return_state:
1533 	atomic_inc(&state->count);
1534 	return state;
1535 }
1536 
1537 static void
1538 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1539 {
1540 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1541 	struct nfs_delegation *delegation;
1542 	int delegation_flags = 0;
1543 
1544 	rcu_read_lock();
1545 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1546 	if (delegation)
1547 		delegation_flags = delegation->flags;
1548 	rcu_read_unlock();
1549 	switch (data->o_arg.claim) {
1550 	default:
1551 		break;
1552 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1553 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1554 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1555 				   "returning a delegation for "
1556 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1557 				   clp->cl_hostname);
1558 		return;
1559 	}
1560 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1561 		nfs_inode_set_delegation(state->inode,
1562 					 data->owner->so_cred,
1563 					 &data->o_res);
1564 	else
1565 		nfs_inode_reclaim_delegation(state->inode,
1566 					     data->owner->so_cred,
1567 					     &data->o_res);
1568 }
1569 
1570 /*
1571  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1572  * and update the nfs4_state.
1573  */
1574 static struct nfs4_state *
1575 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1576 {
1577 	struct inode *inode = data->state->inode;
1578 	struct nfs4_state *state = data->state;
1579 	int ret;
1580 
1581 	if (!data->rpc_done) {
1582 		if (data->rpc_status) {
1583 			ret = data->rpc_status;
1584 			goto err;
1585 		}
1586 		/* cached opens have already been processed */
1587 		goto update;
1588 	}
1589 
1590 	ret = nfs_refresh_inode(inode, &data->f_attr);
1591 	if (ret)
1592 		goto err;
1593 
1594 	if (data->o_res.delegation_type != 0)
1595 		nfs4_opendata_check_deleg(data, state);
1596 update:
1597 	update_open_stateid(state, &data->o_res.stateid, NULL,
1598 			    data->o_arg.fmode);
1599 	atomic_inc(&state->count);
1600 
1601 	return state;
1602 err:
1603 	return ERR_PTR(ret);
1604 
1605 }
1606 
1607 static struct nfs4_state *
1608 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1609 {
1610 	struct inode *inode;
1611 	struct nfs4_state *state = NULL;
1612 	int ret;
1613 
1614 	if (!data->rpc_done) {
1615 		state = nfs4_try_open_cached(data);
1616 		trace_nfs4_cached_open(data->state);
1617 		goto out;
1618 	}
1619 
1620 	ret = -EAGAIN;
1621 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1622 		goto err;
1623 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1624 	ret = PTR_ERR(inode);
1625 	if (IS_ERR(inode))
1626 		goto err;
1627 	ret = -ENOMEM;
1628 	state = nfs4_get_open_state(inode, data->owner);
1629 	if (state == NULL)
1630 		goto err_put_inode;
1631 	if (data->o_res.delegation_type != 0)
1632 		nfs4_opendata_check_deleg(data, state);
1633 	update_open_stateid(state, &data->o_res.stateid, NULL,
1634 			data->o_arg.fmode);
1635 	iput(inode);
1636 out:
1637 	nfs_release_seqid(data->o_arg.seqid);
1638 	return state;
1639 err_put_inode:
1640 	iput(inode);
1641 err:
1642 	return ERR_PTR(ret);
1643 }
1644 
1645 static struct nfs4_state *
1646 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1647 {
1648 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1649 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1650 	return _nfs4_opendata_to_nfs4_state(data);
1651 }
1652 
1653 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1654 {
1655 	struct nfs_inode *nfsi = NFS_I(state->inode);
1656 	struct nfs_open_context *ctx;
1657 
1658 	spin_lock(&state->inode->i_lock);
1659 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1660 		if (ctx->state != state)
1661 			continue;
1662 		get_nfs_open_context(ctx);
1663 		spin_unlock(&state->inode->i_lock);
1664 		return ctx;
1665 	}
1666 	spin_unlock(&state->inode->i_lock);
1667 	return ERR_PTR(-ENOENT);
1668 }
1669 
1670 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1671 		struct nfs4_state *state, enum open_claim_type4 claim)
1672 {
1673 	struct nfs4_opendata *opendata;
1674 
1675 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1676 			NULL, NULL, claim, GFP_NOFS);
1677 	if (opendata == NULL)
1678 		return ERR_PTR(-ENOMEM);
1679 	opendata->state = state;
1680 	atomic_inc(&state->count);
1681 	return opendata;
1682 }
1683 
1684 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1685 		fmode_t fmode)
1686 {
1687 	struct nfs4_state *newstate;
1688 	int ret;
1689 
1690 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1691 		return 0;
1692 	opendata->o_arg.open_flags = 0;
1693 	opendata->o_arg.fmode = fmode;
1694 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1695 			NFS_SB(opendata->dentry->d_sb),
1696 			fmode, 0);
1697 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1698 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1699 	nfs4_init_opendata_res(opendata);
1700 	ret = _nfs4_recover_proc_open(opendata);
1701 	if (ret != 0)
1702 		return ret;
1703 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1704 	if (IS_ERR(newstate))
1705 		return PTR_ERR(newstate);
1706 	if (newstate != opendata->state)
1707 		ret = -ESTALE;
1708 	nfs4_close_state(newstate, fmode);
1709 	return ret;
1710 }
1711 
1712 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1713 {
1714 	int ret;
1715 
1716 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1717 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1718 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1719 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1720 	/* memory barrier prior to reading state->n_* */
1721 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1722 	clear_bit(NFS_OPEN_STATE, &state->flags);
1723 	smp_rmb();
1724 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1725 	if (ret != 0)
1726 		return ret;
1727 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1728 	if (ret != 0)
1729 		return ret;
1730 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1731 	if (ret != 0)
1732 		return ret;
1733 	/*
1734 	 * We may have performed cached opens for all three recoveries.
1735 	 * Check if we need to update the current stateid.
1736 	 */
1737 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1738 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1739 		write_seqlock(&state->seqlock);
1740 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1741 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1742 		write_sequnlock(&state->seqlock);
1743 	}
1744 	return 0;
1745 }
1746 
1747 /*
1748  * OPEN_RECLAIM:
1749  * 	reclaim state on the server after a reboot.
1750  */
1751 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1752 {
1753 	struct nfs_delegation *delegation;
1754 	struct nfs4_opendata *opendata;
1755 	fmode_t delegation_type = 0;
1756 	int status;
1757 
1758 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1759 			NFS4_OPEN_CLAIM_PREVIOUS);
1760 	if (IS_ERR(opendata))
1761 		return PTR_ERR(opendata);
1762 	rcu_read_lock();
1763 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1764 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1765 		delegation_type = delegation->type;
1766 	rcu_read_unlock();
1767 	opendata->o_arg.u.delegation_type = delegation_type;
1768 	status = nfs4_open_recover(opendata, state);
1769 	nfs4_opendata_put(opendata);
1770 	return status;
1771 }
1772 
1773 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1774 {
1775 	struct nfs_server *server = NFS_SERVER(state->inode);
1776 	struct nfs4_exception exception = { };
1777 	int err;
1778 	do {
1779 		err = _nfs4_do_open_reclaim(ctx, state);
1780 		trace_nfs4_open_reclaim(ctx, 0, err);
1781 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1782 			continue;
1783 		if (err != -NFS4ERR_DELAY)
1784 			break;
1785 		nfs4_handle_exception(server, err, &exception);
1786 	} while (exception.retry);
1787 	return err;
1788 }
1789 
1790 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1791 {
1792 	struct nfs_open_context *ctx;
1793 	int ret;
1794 
1795 	ctx = nfs4_state_find_open_context(state);
1796 	if (IS_ERR(ctx))
1797 		return -EAGAIN;
1798 	ret = nfs4_do_open_reclaim(ctx, state);
1799 	put_nfs_open_context(ctx);
1800 	return ret;
1801 }
1802 
1803 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1804 {
1805 	switch (err) {
1806 		default:
1807 			printk(KERN_ERR "NFS: %s: unhandled error "
1808 					"%d.\n", __func__, err);
1809 		case 0:
1810 		case -ENOENT:
1811 		case -EAGAIN:
1812 		case -ESTALE:
1813 			break;
1814 		case -NFS4ERR_BADSESSION:
1815 		case -NFS4ERR_BADSLOT:
1816 		case -NFS4ERR_BAD_HIGH_SLOT:
1817 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1818 		case -NFS4ERR_DEADSESSION:
1819 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1820 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1821 			return -EAGAIN;
1822 		case -NFS4ERR_STALE_CLIENTID:
1823 		case -NFS4ERR_STALE_STATEID:
1824 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1825 		case -NFS4ERR_EXPIRED:
1826 			/* Don't recall a delegation if it was lost */
1827 			nfs4_schedule_lease_recovery(server->nfs_client);
1828 			return -EAGAIN;
1829 		case -NFS4ERR_MOVED:
1830 			nfs4_schedule_migration_recovery(server);
1831 			return -EAGAIN;
1832 		case -NFS4ERR_LEASE_MOVED:
1833 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1834 			return -EAGAIN;
1835 		case -NFS4ERR_DELEG_REVOKED:
1836 		case -NFS4ERR_ADMIN_REVOKED:
1837 		case -NFS4ERR_BAD_STATEID:
1838 		case -NFS4ERR_OPENMODE:
1839 			nfs_inode_find_state_and_recover(state->inode,
1840 					stateid);
1841 			nfs4_schedule_stateid_recovery(server, state);
1842 			return -EAGAIN;
1843 		case -NFS4ERR_DELAY:
1844 		case -NFS4ERR_GRACE:
1845 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1846 			ssleep(1);
1847 			return -EAGAIN;
1848 		case -ENOMEM:
1849 		case -NFS4ERR_DENIED:
1850 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1851 			return 0;
1852 	}
1853 	return err;
1854 }
1855 
1856 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1857 		struct nfs4_state *state, const nfs4_stateid *stateid,
1858 		fmode_t type)
1859 {
1860 	struct nfs_server *server = NFS_SERVER(state->inode);
1861 	struct nfs4_opendata *opendata;
1862 	int err = 0;
1863 
1864 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1865 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1866 	if (IS_ERR(opendata))
1867 		return PTR_ERR(opendata);
1868 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1869 	write_seqlock(&state->seqlock);
1870 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1871 	write_sequnlock(&state->seqlock);
1872 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1873 	switch (type & (FMODE_READ|FMODE_WRITE)) {
1874 	case FMODE_READ|FMODE_WRITE:
1875 	case FMODE_WRITE:
1876 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1877 		if (err)
1878 			break;
1879 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1880 		if (err)
1881 			break;
1882 	case FMODE_READ:
1883 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
1884 	}
1885 	nfs4_opendata_put(opendata);
1886 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1887 }
1888 
1889 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1890 {
1891 	struct nfs4_opendata *data = calldata;
1892 
1893 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1894 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1895 }
1896 
1897 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1898 {
1899 	struct nfs4_opendata *data = calldata;
1900 
1901 	nfs40_sequence_done(task, &data->c_res.seq_res);
1902 
1903 	data->rpc_status = task->tk_status;
1904 	if (data->rpc_status == 0) {
1905 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1906 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1907 		renew_lease(data->o_res.server, data->timestamp);
1908 		data->rpc_done = 1;
1909 	}
1910 }
1911 
1912 static void nfs4_open_confirm_release(void *calldata)
1913 {
1914 	struct nfs4_opendata *data = calldata;
1915 	struct nfs4_state *state = NULL;
1916 
1917 	/* If this request hasn't been cancelled, do nothing */
1918 	if (data->cancelled == 0)
1919 		goto out_free;
1920 	/* In case of error, no cleanup! */
1921 	if (!data->rpc_done)
1922 		goto out_free;
1923 	state = nfs4_opendata_to_nfs4_state(data);
1924 	if (!IS_ERR(state))
1925 		nfs4_close_state(state, data->o_arg.fmode);
1926 out_free:
1927 	nfs4_opendata_put(data);
1928 }
1929 
1930 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1931 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1932 	.rpc_call_done = nfs4_open_confirm_done,
1933 	.rpc_release = nfs4_open_confirm_release,
1934 };
1935 
1936 /*
1937  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1938  */
1939 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1940 {
1941 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1942 	struct rpc_task *task;
1943 	struct  rpc_message msg = {
1944 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1945 		.rpc_argp = &data->c_arg,
1946 		.rpc_resp = &data->c_res,
1947 		.rpc_cred = data->owner->so_cred,
1948 	};
1949 	struct rpc_task_setup task_setup_data = {
1950 		.rpc_client = server->client,
1951 		.rpc_message = &msg,
1952 		.callback_ops = &nfs4_open_confirm_ops,
1953 		.callback_data = data,
1954 		.workqueue = nfsiod_workqueue,
1955 		.flags = RPC_TASK_ASYNC,
1956 	};
1957 	int status;
1958 
1959 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1960 	kref_get(&data->kref);
1961 	data->rpc_done = 0;
1962 	data->rpc_status = 0;
1963 	data->timestamp = jiffies;
1964 	if (data->is_recover)
1965 		nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1966 	task = rpc_run_task(&task_setup_data);
1967 	if (IS_ERR(task))
1968 		return PTR_ERR(task);
1969 	status = nfs4_wait_for_completion_rpc_task(task);
1970 	if (status != 0) {
1971 		data->cancelled = 1;
1972 		smp_wmb();
1973 	} else
1974 		status = data->rpc_status;
1975 	rpc_put_task(task);
1976 	return status;
1977 }
1978 
1979 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1980 {
1981 	struct nfs4_opendata *data = calldata;
1982 	struct nfs4_state_owner *sp = data->owner;
1983 	struct nfs_client *clp = sp->so_server->nfs_client;
1984 	enum open_claim_type4 claim = data->o_arg.claim;
1985 
1986 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1987 		goto out_wait;
1988 	/*
1989 	 * Check if we still need to send an OPEN call, or if we can use
1990 	 * a delegation instead.
1991 	 */
1992 	if (data->state != NULL) {
1993 		struct nfs_delegation *delegation;
1994 
1995 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1996 			goto out_no_action;
1997 		rcu_read_lock();
1998 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1999 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2000 			goto unlock_no_action;
2001 		rcu_read_unlock();
2002 	}
2003 	/* Update client id. */
2004 	data->o_arg.clientid = clp->cl_clientid;
2005 	switch (claim) {
2006 	default:
2007 		break;
2008 	case NFS4_OPEN_CLAIM_PREVIOUS:
2009 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2010 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2011 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2012 	case NFS4_OPEN_CLAIM_FH:
2013 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2014 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2015 	}
2016 	data->timestamp = jiffies;
2017 	if (nfs4_setup_sequence(data->o_arg.server,
2018 				&data->o_arg.seq_args,
2019 				&data->o_res.seq_res,
2020 				task) != 0)
2021 		nfs_release_seqid(data->o_arg.seqid);
2022 
2023 	/* Set the create mode (note dependency on the session type) */
2024 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2025 	if (data->o_arg.open_flags & O_EXCL) {
2026 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2027 		if (nfs4_has_persistent_session(clp))
2028 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2029 		else if (clp->cl_mvops->minor_version > 0)
2030 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2031 	}
2032 	return;
2033 unlock_no_action:
2034 	trace_nfs4_cached_open(data->state);
2035 	rcu_read_unlock();
2036 out_no_action:
2037 	task->tk_action = NULL;
2038 out_wait:
2039 	nfs4_sequence_done(task, &data->o_res.seq_res);
2040 }
2041 
2042 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2043 {
2044 	struct nfs4_opendata *data = calldata;
2045 
2046 	data->rpc_status = task->tk_status;
2047 
2048 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2049 		return;
2050 
2051 	if (task->tk_status == 0) {
2052 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2053 			switch (data->o_res.f_attr->mode & S_IFMT) {
2054 			case S_IFREG:
2055 				break;
2056 			case S_IFLNK:
2057 				data->rpc_status = -ELOOP;
2058 				break;
2059 			case S_IFDIR:
2060 				data->rpc_status = -EISDIR;
2061 				break;
2062 			default:
2063 				data->rpc_status = -ENOTDIR;
2064 			}
2065 		}
2066 		renew_lease(data->o_res.server, data->timestamp);
2067 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2068 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2069 	}
2070 	data->rpc_done = 1;
2071 }
2072 
2073 static void nfs4_open_release(void *calldata)
2074 {
2075 	struct nfs4_opendata *data = calldata;
2076 	struct nfs4_state *state = NULL;
2077 
2078 	/* If this request hasn't been cancelled, do nothing */
2079 	if (data->cancelled == 0)
2080 		goto out_free;
2081 	/* In case of error, no cleanup! */
2082 	if (data->rpc_status != 0 || !data->rpc_done)
2083 		goto out_free;
2084 	/* In case we need an open_confirm, no cleanup! */
2085 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2086 		goto out_free;
2087 	state = nfs4_opendata_to_nfs4_state(data);
2088 	if (!IS_ERR(state))
2089 		nfs4_close_state(state, data->o_arg.fmode);
2090 out_free:
2091 	nfs4_opendata_put(data);
2092 }
2093 
2094 static const struct rpc_call_ops nfs4_open_ops = {
2095 	.rpc_call_prepare = nfs4_open_prepare,
2096 	.rpc_call_done = nfs4_open_done,
2097 	.rpc_release = nfs4_open_release,
2098 };
2099 
2100 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2101 {
2102 	struct inode *dir = d_inode(data->dir);
2103 	struct nfs_server *server = NFS_SERVER(dir);
2104 	struct nfs_openargs *o_arg = &data->o_arg;
2105 	struct nfs_openres *o_res = &data->o_res;
2106 	struct rpc_task *task;
2107 	struct rpc_message msg = {
2108 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2109 		.rpc_argp = o_arg,
2110 		.rpc_resp = o_res,
2111 		.rpc_cred = data->owner->so_cred,
2112 	};
2113 	struct rpc_task_setup task_setup_data = {
2114 		.rpc_client = server->client,
2115 		.rpc_message = &msg,
2116 		.callback_ops = &nfs4_open_ops,
2117 		.callback_data = data,
2118 		.workqueue = nfsiod_workqueue,
2119 		.flags = RPC_TASK_ASYNC,
2120 	};
2121 	int status;
2122 
2123 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2124 	kref_get(&data->kref);
2125 	data->rpc_done = 0;
2126 	data->rpc_status = 0;
2127 	data->cancelled = 0;
2128 	data->is_recover = 0;
2129 	if (isrecover) {
2130 		nfs4_set_sequence_privileged(&o_arg->seq_args);
2131 		data->is_recover = 1;
2132 	}
2133 	task = rpc_run_task(&task_setup_data);
2134         if (IS_ERR(task))
2135                 return PTR_ERR(task);
2136         status = nfs4_wait_for_completion_rpc_task(task);
2137         if (status != 0) {
2138                 data->cancelled = 1;
2139                 smp_wmb();
2140         } else
2141                 status = data->rpc_status;
2142         rpc_put_task(task);
2143 
2144 	return status;
2145 }
2146 
2147 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2148 {
2149 	struct inode *dir = d_inode(data->dir);
2150 	struct nfs_openres *o_res = &data->o_res;
2151         int status;
2152 
2153 	status = nfs4_run_open_task(data, 1);
2154 	if (status != 0 || !data->rpc_done)
2155 		return status;
2156 
2157 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2158 
2159 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2160 		status = _nfs4_proc_open_confirm(data);
2161 		if (status != 0)
2162 			return status;
2163 	}
2164 
2165 	return status;
2166 }
2167 
2168 /*
2169  * Additional permission checks in order to distinguish between an
2170  * open for read, and an open for execute. This works around the
2171  * fact that NFSv4 OPEN treats read and execute permissions as being
2172  * the same.
2173  * Note that in the non-execute case, we want to turn off permission
2174  * checking if we just created a new file (POSIX open() semantics).
2175  */
2176 static int nfs4_opendata_access(struct rpc_cred *cred,
2177 				struct nfs4_opendata *opendata,
2178 				struct nfs4_state *state, fmode_t fmode,
2179 				int openflags)
2180 {
2181 	struct nfs_access_entry cache;
2182 	u32 mask;
2183 
2184 	/* access call failed or for some reason the server doesn't
2185 	 * support any access modes -- defer access call until later */
2186 	if (opendata->o_res.access_supported == 0)
2187 		return 0;
2188 
2189 	mask = 0;
2190 	/*
2191 	 * Use openflags to check for exec, because fmode won't
2192 	 * always have FMODE_EXEC set when file open for exec.
2193 	 */
2194 	if (openflags & __FMODE_EXEC) {
2195 		/* ONLY check for exec rights */
2196 		mask = MAY_EXEC;
2197 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2198 		mask = MAY_READ;
2199 
2200 	cache.cred = cred;
2201 	cache.jiffies = jiffies;
2202 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2203 	nfs_access_add_cache(state->inode, &cache);
2204 
2205 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2206 		return 0;
2207 
2208 	/* even though OPEN succeeded, access is denied. Close the file */
2209 	nfs4_close_state(state, fmode);
2210 	return -EACCES;
2211 }
2212 
2213 /*
2214  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2215  */
2216 static int _nfs4_proc_open(struct nfs4_opendata *data)
2217 {
2218 	struct inode *dir = d_inode(data->dir);
2219 	struct nfs_server *server = NFS_SERVER(dir);
2220 	struct nfs_openargs *o_arg = &data->o_arg;
2221 	struct nfs_openres *o_res = &data->o_res;
2222 	int status;
2223 
2224 	status = nfs4_run_open_task(data, 0);
2225 	if (!data->rpc_done)
2226 		return status;
2227 	if (status != 0) {
2228 		if (status == -NFS4ERR_BADNAME &&
2229 				!(o_arg->open_flags & O_CREAT))
2230 			return -ENOENT;
2231 		return status;
2232 	}
2233 
2234 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2235 
2236 	if (o_arg->open_flags & O_CREAT) {
2237 		update_changeattr(dir, &o_res->cinfo);
2238 		if (o_arg->open_flags & O_EXCL)
2239 			data->file_created = 1;
2240 		else if (o_res->cinfo.before != o_res->cinfo.after)
2241 			data->file_created = 1;
2242 	}
2243 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2244 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2245 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2246 		status = _nfs4_proc_open_confirm(data);
2247 		if (status != 0)
2248 			return status;
2249 	}
2250 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2251 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2252 	return 0;
2253 }
2254 
2255 static int nfs4_recover_expired_lease(struct nfs_server *server)
2256 {
2257 	return nfs4_client_recover_expired_lease(server->nfs_client);
2258 }
2259 
2260 /*
2261  * OPEN_EXPIRED:
2262  * 	reclaim state on the server after a network partition.
2263  * 	Assumes caller holds the appropriate lock
2264  */
2265 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2266 {
2267 	struct nfs4_opendata *opendata;
2268 	int ret;
2269 
2270 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2271 			NFS4_OPEN_CLAIM_FH);
2272 	if (IS_ERR(opendata))
2273 		return PTR_ERR(opendata);
2274 	ret = nfs4_open_recover(opendata, state);
2275 	if (ret == -ESTALE)
2276 		d_drop(ctx->dentry);
2277 	nfs4_opendata_put(opendata);
2278 	return ret;
2279 }
2280 
2281 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2282 {
2283 	struct nfs_server *server = NFS_SERVER(state->inode);
2284 	struct nfs4_exception exception = { };
2285 	int err;
2286 
2287 	do {
2288 		err = _nfs4_open_expired(ctx, state);
2289 		trace_nfs4_open_expired(ctx, 0, err);
2290 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2291 			continue;
2292 		switch (err) {
2293 		default:
2294 			goto out;
2295 		case -NFS4ERR_GRACE:
2296 		case -NFS4ERR_DELAY:
2297 			nfs4_handle_exception(server, err, &exception);
2298 			err = 0;
2299 		}
2300 	} while (exception.retry);
2301 out:
2302 	return err;
2303 }
2304 
2305 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2306 {
2307 	struct nfs_open_context *ctx;
2308 	int ret;
2309 
2310 	ctx = nfs4_state_find_open_context(state);
2311 	if (IS_ERR(ctx))
2312 		return -EAGAIN;
2313 	ret = nfs4_do_open_expired(ctx, state);
2314 	put_nfs_open_context(ctx);
2315 	return ret;
2316 }
2317 
2318 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2319 {
2320 	nfs_remove_bad_delegation(state->inode);
2321 	write_seqlock(&state->seqlock);
2322 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2323 	write_sequnlock(&state->seqlock);
2324 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2325 }
2326 
2327 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2328 {
2329 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2330 		nfs_finish_clear_delegation_stateid(state);
2331 }
2332 
2333 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2334 {
2335 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2336 	nfs40_clear_delegation_stateid(state);
2337 	return nfs4_open_expired(sp, state);
2338 }
2339 
2340 #if defined(CONFIG_NFS_V4_1)
2341 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2342 {
2343 	struct nfs_server *server = NFS_SERVER(state->inode);
2344 	nfs4_stateid stateid;
2345 	struct nfs_delegation *delegation;
2346 	struct rpc_cred *cred;
2347 	int status;
2348 
2349 	/* Get the delegation credential for use by test/free_stateid */
2350 	rcu_read_lock();
2351 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2352 	if (delegation == NULL) {
2353 		rcu_read_unlock();
2354 		return;
2355 	}
2356 
2357 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2358 	cred = get_rpccred(delegation->cred);
2359 	rcu_read_unlock();
2360 	status = nfs41_test_stateid(server, &stateid, cred);
2361 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2362 
2363 	if (status != NFS_OK) {
2364 		/* Free the stateid unless the server explicitly
2365 		 * informs us the stateid is unrecognized. */
2366 		if (status != -NFS4ERR_BAD_STATEID)
2367 			nfs41_free_stateid(server, &stateid, cred);
2368 		nfs_finish_clear_delegation_stateid(state);
2369 	}
2370 
2371 	put_rpccred(cred);
2372 }
2373 
2374 /**
2375  * nfs41_check_open_stateid - possibly free an open stateid
2376  *
2377  * @state: NFSv4 state for an inode
2378  *
2379  * Returns NFS_OK if recovery for this stateid is now finished.
2380  * Otherwise a negative NFS4ERR value is returned.
2381  */
2382 static int nfs41_check_open_stateid(struct nfs4_state *state)
2383 {
2384 	struct nfs_server *server = NFS_SERVER(state->inode);
2385 	nfs4_stateid *stateid = &state->open_stateid;
2386 	struct rpc_cred *cred = state->owner->so_cred;
2387 	int status;
2388 
2389 	/* If a state reset has been done, test_stateid is unneeded */
2390 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2391 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2392 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2393 		return -NFS4ERR_BAD_STATEID;
2394 
2395 	status = nfs41_test_stateid(server, stateid, cred);
2396 	trace_nfs4_test_open_stateid(state, NULL, status);
2397 	if (status != NFS_OK) {
2398 		/* Free the stateid unless the server explicitly
2399 		 * informs us the stateid is unrecognized. */
2400 		if (status != -NFS4ERR_BAD_STATEID)
2401 			nfs41_free_stateid(server, stateid, cred);
2402 
2403 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2404 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2405 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2406 		clear_bit(NFS_OPEN_STATE, &state->flags);
2407 	}
2408 	return status;
2409 }
2410 
2411 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2412 {
2413 	int status;
2414 
2415 	nfs41_check_delegation_stateid(state);
2416 	status = nfs41_check_open_stateid(state);
2417 	if (status != NFS_OK)
2418 		status = nfs4_open_expired(sp, state);
2419 	return status;
2420 }
2421 #endif
2422 
2423 /*
2424  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2425  * fields corresponding to attributes that were used to store the verifier.
2426  * Make sure we clobber those fields in the later setattr call
2427  */
2428 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2429 				struct iattr *sattr, struct nfs4_label **label)
2430 {
2431 	const u32 *attrset = opendata->o_res.attrset;
2432 
2433 	if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2434 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2435 		sattr->ia_valid |= ATTR_ATIME;
2436 
2437 	if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2438 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2439 		sattr->ia_valid |= ATTR_MTIME;
2440 
2441 	/* Except MODE, it seems harmless of setting twice. */
2442 	if ((attrset[1] & FATTR4_WORD1_MODE))
2443 		sattr->ia_valid &= ~ATTR_MODE;
2444 
2445 	if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2446 		*label = NULL;
2447 }
2448 
2449 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2450 		fmode_t fmode,
2451 		int flags,
2452 		struct nfs_open_context *ctx)
2453 {
2454 	struct nfs4_state_owner *sp = opendata->owner;
2455 	struct nfs_server *server = sp->so_server;
2456 	struct dentry *dentry;
2457 	struct nfs4_state *state;
2458 	unsigned int seq;
2459 	int ret;
2460 
2461 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2462 
2463 	ret = _nfs4_proc_open(opendata);
2464 	if (ret != 0)
2465 		goto out;
2466 
2467 	state = nfs4_opendata_to_nfs4_state(opendata);
2468 	ret = PTR_ERR(state);
2469 	if (IS_ERR(state))
2470 		goto out;
2471 	if (server->caps & NFS_CAP_POSIX_LOCK)
2472 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2473 
2474 	dentry = opendata->dentry;
2475 	if (d_really_is_negative(dentry)) {
2476 		struct dentry *alias;
2477 		d_drop(dentry);
2478 		alias = d_exact_alias(dentry, state->inode);
2479 		if (!alias)
2480 			alias = d_splice_alias(igrab(state->inode), dentry);
2481 		/* d_splice_alias() can't fail here - it's a non-directory */
2482 		if (alias) {
2483 			dput(ctx->dentry);
2484 			ctx->dentry = dentry = alias;
2485 		}
2486 		nfs_set_verifier(dentry,
2487 				nfs_save_change_attribute(d_inode(opendata->dir)));
2488 	}
2489 
2490 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2491 	if (ret != 0)
2492 		goto out;
2493 
2494 	ctx->state = state;
2495 	if (d_inode(dentry) == state->inode) {
2496 		nfs_inode_attach_open_context(ctx);
2497 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2498 			nfs4_schedule_stateid_recovery(server, state);
2499 	}
2500 out:
2501 	return ret;
2502 }
2503 
2504 /*
2505  * Returns a referenced nfs4_state
2506  */
2507 static int _nfs4_do_open(struct inode *dir,
2508 			struct nfs_open_context *ctx,
2509 			int flags,
2510 			struct iattr *sattr,
2511 			struct nfs4_label *label,
2512 			int *opened)
2513 {
2514 	struct nfs4_state_owner  *sp;
2515 	struct nfs4_state     *state = NULL;
2516 	struct nfs_server       *server = NFS_SERVER(dir);
2517 	struct nfs4_opendata *opendata;
2518 	struct dentry *dentry = ctx->dentry;
2519 	struct rpc_cred *cred = ctx->cred;
2520 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2521 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2522 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2523 	struct nfs4_label *olabel = NULL;
2524 	int status;
2525 
2526 	/* Protect against reboot recovery conflicts */
2527 	status = -ENOMEM;
2528 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2529 	if (sp == NULL) {
2530 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2531 		goto out_err;
2532 	}
2533 	status = nfs4_recover_expired_lease(server);
2534 	if (status != 0)
2535 		goto err_put_state_owner;
2536 	if (d_really_is_positive(dentry))
2537 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2538 	status = -ENOMEM;
2539 	if (d_really_is_positive(dentry))
2540 		claim = NFS4_OPEN_CLAIM_FH;
2541 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2542 			label, claim, GFP_KERNEL);
2543 	if (opendata == NULL)
2544 		goto err_put_state_owner;
2545 
2546 	if (label) {
2547 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2548 		if (IS_ERR(olabel)) {
2549 			status = PTR_ERR(olabel);
2550 			goto err_opendata_put;
2551 		}
2552 	}
2553 
2554 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2555 		if (!opendata->f_attr.mdsthreshold) {
2556 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2557 			if (!opendata->f_attr.mdsthreshold)
2558 				goto err_free_label;
2559 		}
2560 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2561 	}
2562 	if (d_really_is_positive(dentry))
2563 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2564 
2565 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2566 	if (status != 0)
2567 		goto err_free_label;
2568 	state = ctx->state;
2569 
2570 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2571 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2572 		nfs4_exclusive_attrset(opendata, sattr, &label);
2573 		/*
2574 		 * send create attributes which was not set by open
2575 		 * with an extra setattr.
2576 		 */
2577 		if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2578 			nfs_fattr_init(opendata->o_res.f_attr);
2579 			status = nfs4_do_setattr(state->inode, cred,
2580 					opendata->o_res.f_attr, sattr,
2581 					state, label, olabel);
2582 			if (status == 0) {
2583 				nfs_setattr_update_inode(state->inode, sattr,
2584 						opendata->o_res.f_attr);
2585 				nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2586 			}
2587 		}
2588 	}
2589 	if (opened && opendata->file_created)
2590 		*opened |= FILE_CREATED;
2591 
2592 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2593 		*ctx_th = opendata->f_attr.mdsthreshold;
2594 		opendata->f_attr.mdsthreshold = NULL;
2595 	}
2596 
2597 	nfs4_label_free(olabel);
2598 
2599 	nfs4_opendata_put(opendata);
2600 	nfs4_put_state_owner(sp);
2601 	return 0;
2602 err_free_label:
2603 	nfs4_label_free(olabel);
2604 err_opendata_put:
2605 	nfs4_opendata_put(opendata);
2606 err_put_state_owner:
2607 	nfs4_put_state_owner(sp);
2608 out_err:
2609 	return status;
2610 }
2611 
2612 
2613 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2614 					struct nfs_open_context *ctx,
2615 					int flags,
2616 					struct iattr *sattr,
2617 					struct nfs4_label *label,
2618 					int *opened)
2619 {
2620 	struct nfs_server *server = NFS_SERVER(dir);
2621 	struct nfs4_exception exception = { };
2622 	struct nfs4_state *res;
2623 	int status;
2624 
2625 	do {
2626 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2627 		res = ctx->state;
2628 		trace_nfs4_open_file(ctx, flags, status);
2629 		if (status == 0)
2630 			break;
2631 		/* NOTE: BAD_SEQID means the server and client disagree about the
2632 		 * book-keeping w.r.t. state-changing operations
2633 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2634 		 * It is actually a sign of a bug on the client or on the server.
2635 		 *
2636 		 * If we receive a BAD_SEQID error in the particular case of
2637 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2638 		 * have unhashed the old state_owner for us, and that we can
2639 		 * therefore safely retry using a new one. We should still warn
2640 		 * the user though...
2641 		 */
2642 		if (status == -NFS4ERR_BAD_SEQID) {
2643 			pr_warn_ratelimited("NFS: v4 server %s "
2644 					" returned a bad sequence-id error!\n",
2645 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2646 			exception.retry = 1;
2647 			continue;
2648 		}
2649 		/*
2650 		 * BAD_STATEID on OPEN means that the server cancelled our
2651 		 * state before it received the OPEN_CONFIRM.
2652 		 * Recover by retrying the request as per the discussion
2653 		 * on Page 181 of RFC3530.
2654 		 */
2655 		if (status == -NFS4ERR_BAD_STATEID) {
2656 			exception.retry = 1;
2657 			continue;
2658 		}
2659 		if (status == -EAGAIN) {
2660 			/* We must have found a delegation */
2661 			exception.retry = 1;
2662 			continue;
2663 		}
2664 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2665 			continue;
2666 		res = ERR_PTR(nfs4_handle_exception(server,
2667 					status, &exception));
2668 	} while (exception.retry);
2669 	return res;
2670 }
2671 
2672 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2673 			    struct nfs_fattr *fattr, struct iattr *sattr,
2674 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2675 			    struct nfs4_label *olabel)
2676 {
2677 	struct nfs_server *server = NFS_SERVER(inode);
2678         struct nfs_setattrargs  arg = {
2679                 .fh             = NFS_FH(inode),
2680                 .iap            = sattr,
2681 		.server		= server,
2682 		.bitmask = server->attr_bitmask,
2683 		.label		= ilabel,
2684         };
2685         struct nfs_setattrres  res = {
2686 		.fattr		= fattr,
2687 		.label		= olabel,
2688 		.server		= server,
2689         };
2690         struct rpc_message msg = {
2691 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2692 		.rpc_argp	= &arg,
2693 		.rpc_resp	= &res,
2694 		.rpc_cred	= cred,
2695         };
2696 	struct rpc_cred *delegation_cred = NULL;
2697 	unsigned long timestamp = jiffies;
2698 	fmode_t fmode;
2699 	bool truncate;
2700 	int status;
2701 
2702 	arg.bitmask = nfs4_bitmask(server, ilabel);
2703 	if (ilabel)
2704 		arg.bitmask = nfs4_bitmask(server, olabel);
2705 
2706 	nfs_fattr_init(fattr);
2707 
2708 	/* Servers should only apply open mode checks for file size changes */
2709 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2710 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2711 
2712 	if (nfs4_copy_delegation_stateid(inode, fmode, &arg.stateid, &delegation_cred)) {
2713 		/* Use that stateid */
2714 	} else if (truncate && state != NULL) {
2715 		struct nfs_lockowner lockowner = {
2716 			.l_owner = current->files,
2717 			.l_pid = current->tgid,
2718 		};
2719 		if (!nfs4_valid_open_stateid(state))
2720 			return -EBADF;
2721 		if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2722 				&arg.stateid, &delegation_cred) == -EIO)
2723 			return -EBADF;
2724 	} else
2725 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2726 	if (delegation_cred)
2727 		msg.rpc_cred = delegation_cred;
2728 
2729 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2730 
2731 	put_rpccred(delegation_cred);
2732 	if (status == 0 && state != NULL)
2733 		renew_lease(server, timestamp);
2734 	trace_nfs4_setattr(inode, &arg.stateid, status);
2735 	return status;
2736 }
2737 
2738 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2739 			   struct nfs_fattr *fattr, struct iattr *sattr,
2740 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2741 			   struct nfs4_label *olabel)
2742 {
2743 	struct nfs_server *server = NFS_SERVER(inode);
2744 	struct nfs4_exception exception = {
2745 		.state = state,
2746 		.inode = inode,
2747 	};
2748 	int err;
2749 	do {
2750 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2751 		switch (err) {
2752 		case -NFS4ERR_OPENMODE:
2753 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2754 				pr_warn_once("NFSv4: server %s is incorrectly "
2755 						"applying open mode checks to "
2756 						"a SETATTR that is not "
2757 						"changing file size.\n",
2758 						server->nfs_client->cl_hostname);
2759 			}
2760 			if (state && !(state->state & FMODE_WRITE)) {
2761 				err = -EBADF;
2762 				if (sattr->ia_valid & ATTR_OPEN)
2763 					err = -EACCES;
2764 				goto out;
2765 			}
2766 		}
2767 		err = nfs4_handle_exception(server, err, &exception);
2768 	} while (exception.retry);
2769 out:
2770 	return err;
2771 }
2772 
2773 static bool
2774 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2775 {
2776 	if (inode == NULL || !nfs_have_layout(inode))
2777 		return false;
2778 
2779 	return pnfs_wait_on_layoutreturn(inode, task);
2780 }
2781 
2782 struct nfs4_closedata {
2783 	struct inode *inode;
2784 	struct nfs4_state *state;
2785 	struct nfs_closeargs arg;
2786 	struct nfs_closeres res;
2787 	struct nfs_fattr fattr;
2788 	unsigned long timestamp;
2789 	bool roc;
2790 	u32 roc_barrier;
2791 };
2792 
2793 static void nfs4_free_closedata(void *data)
2794 {
2795 	struct nfs4_closedata *calldata = data;
2796 	struct nfs4_state_owner *sp = calldata->state->owner;
2797 	struct super_block *sb = calldata->state->inode->i_sb;
2798 
2799 	if (calldata->roc)
2800 		pnfs_roc_release(calldata->state->inode);
2801 	nfs4_put_open_state(calldata->state);
2802 	nfs_free_seqid(calldata->arg.seqid);
2803 	nfs4_put_state_owner(sp);
2804 	nfs_sb_deactive(sb);
2805 	kfree(calldata);
2806 }
2807 
2808 static void nfs4_close_done(struct rpc_task *task, void *data)
2809 {
2810 	struct nfs4_closedata *calldata = data;
2811 	struct nfs4_state *state = calldata->state;
2812 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2813 	nfs4_stateid *res_stateid = NULL;
2814 
2815 	dprintk("%s: begin!\n", __func__);
2816 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2817 		return;
2818 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2819         /* hmm. we are done with the inode, and in the process of freeing
2820 	 * the state_owner. we keep this around to process errors
2821 	 */
2822 	switch (task->tk_status) {
2823 		case 0:
2824 			res_stateid = &calldata->res.stateid;
2825 			if (calldata->roc)
2826 				pnfs_roc_set_barrier(state->inode,
2827 						     calldata->roc_barrier);
2828 			renew_lease(server, calldata->timestamp);
2829 			break;
2830 		case -NFS4ERR_ADMIN_REVOKED:
2831 		case -NFS4ERR_STALE_STATEID:
2832 		case -NFS4ERR_OLD_STATEID:
2833 		case -NFS4ERR_BAD_STATEID:
2834 		case -NFS4ERR_EXPIRED:
2835 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2836 						&state->open_stateid)) {
2837 				rpc_restart_call_prepare(task);
2838 				goto out_release;
2839 			}
2840 			if (calldata->arg.fmode == 0)
2841 				break;
2842 		default:
2843 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2844 				rpc_restart_call_prepare(task);
2845 				goto out_release;
2846 			}
2847 	}
2848 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
2849 			res_stateid, calldata->arg.fmode);
2850 out_release:
2851 	nfs_release_seqid(calldata->arg.seqid);
2852 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2853 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2854 }
2855 
2856 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2857 {
2858 	struct nfs4_closedata *calldata = data;
2859 	struct nfs4_state *state = calldata->state;
2860 	struct inode *inode = calldata->inode;
2861 	bool is_rdonly, is_wronly, is_rdwr;
2862 	int call_close = 0;
2863 
2864 	dprintk("%s: begin!\n", __func__);
2865 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2866 		goto out_wait;
2867 
2868 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2869 	spin_lock(&state->owner->so_lock);
2870 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2871 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2872 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2873 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2874 	/* Calculate the change in open mode */
2875 	calldata->arg.fmode = 0;
2876 	if (state->n_rdwr == 0) {
2877 		if (state->n_rdonly == 0)
2878 			call_close |= is_rdonly;
2879 		else if (is_rdonly)
2880 			calldata->arg.fmode |= FMODE_READ;
2881 		if (state->n_wronly == 0)
2882 			call_close |= is_wronly;
2883 		else if (is_wronly)
2884 			calldata->arg.fmode |= FMODE_WRITE;
2885 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2886 			call_close |= is_rdwr;
2887 	} else if (is_rdwr)
2888 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2889 
2890 	if (!nfs4_valid_open_stateid(state))
2891 		call_close = 0;
2892 	spin_unlock(&state->owner->so_lock);
2893 
2894 	if (!call_close) {
2895 		/* Note: exit _without_ calling nfs4_close_done */
2896 		goto out_no_action;
2897 	}
2898 
2899 	if (nfs4_wait_on_layoutreturn(inode, task)) {
2900 		nfs_release_seqid(calldata->arg.seqid);
2901 		goto out_wait;
2902 	}
2903 
2904 	if (calldata->arg.fmode == 0)
2905 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2906 	if (calldata->roc)
2907 		pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2908 
2909 	calldata->arg.share_access =
2910 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2911 				calldata->arg.fmode, 0);
2912 
2913 	nfs_fattr_init(calldata->res.fattr);
2914 	calldata->timestamp = jiffies;
2915 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2916 				&calldata->arg.seq_args,
2917 				&calldata->res.seq_res,
2918 				task) != 0)
2919 		nfs_release_seqid(calldata->arg.seqid);
2920 	dprintk("%s: done!\n", __func__);
2921 	return;
2922 out_no_action:
2923 	task->tk_action = NULL;
2924 out_wait:
2925 	nfs4_sequence_done(task, &calldata->res.seq_res);
2926 }
2927 
2928 static const struct rpc_call_ops nfs4_close_ops = {
2929 	.rpc_call_prepare = nfs4_close_prepare,
2930 	.rpc_call_done = nfs4_close_done,
2931 	.rpc_release = nfs4_free_closedata,
2932 };
2933 
2934 static bool nfs4_roc(struct inode *inode)
2935 {
2936 	if (!nfs_have_layout(inode))
2937 		return false;
2938 	return pnfs_roc(inode);
2939 }
2940 
2941 /*
2942  * It is possible for data to be read/written from a mem-mapped file
2943  * after the sys_close call (which hits the vfs layer as a flush).
2944  * This means that we can't safely call nfsv4 close on a file until
2945  * the inode is cleared. This in turn means that we are not good
2946  * NFSv4 citizens - we do not indicate to the server to update the file's
2947  * share state even when we are done with one of the three share
2948  * stateid's in the inode.
2949  *
2950  * NOTE: Caller must be holding the sp->so_owner semaphore!
2951  */
2952 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2953 {
2954 	struct nfs_server *server = NFS_SERVER(state->inode);
2955 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2956 	struct nfs4_closedata *calldata;
2957 	struct nfs4_state_owner *sp = state->owner;
2958 	struct rpc_task *task;
2959 	struct rpc_message msg = {
2960 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2961 		.rpc_cred = state->owner->so_cred,
2962 	};
2963 	struct rpc_task_setup task_setup_data = {
2964 		.rpc_client = server->client,
2965 		.rpc_message = &msg,
2966 		.callback_ops = &nfs4_close_ops,
2967 		.workqueue = nfsiod_workqueue,
2968 		.flags = RPC_TASK_ASYNC,
2969 	};
2970 	int status = -ENOMEM;
2971 
2972 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2973 		&task_setup_data.rpc_client, &msg);
2974 
2975 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2976 	if (calldata == NULL)
2977 		goto out;
2978 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2979 	calldata->inode = state->inode;
2980 	calldata->state = state;
2981 	calldata->arg.fh = NFS_FH(state->inode);
2982 	/* Serialization for the sequence id */
2983 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2984 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2985 	if (IS_ERR(calldata->arg.seqid))
2986 		goto out_free_calldata;
2987 	calldata->arg.fmode = 0;
2988 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2989 	calldata->res.fattr = &calldata->fattr;
2990 	calldata->res.seqid = calldata->arg.seqid;
2991 	calldata->res.server = server;
2992 	calldata->roc = nfs4_roc(state->inode);
2993 	nfs_sb_active(calldata->inode->i_sb);
2994 
2995 	msg.rpc_argp = &calldata->arg;
2996 	msg.rpc_resp = &calldata->res;
2997 	task_setup_data.callback_data = calldata;
2998 	task = rpc_run_task(&task_setup_data);
2999 	if (IS_ERR(task))
3000 		return PTR_ERR(task);
3001 	status = 0;
3002 	if (wait)
3003 		status = rpc_wait_for_completion_task(task);
3004 	rpc_put_task(task);
3005 	return status;
3006 out_free_calldata:
3007 	kfree(calldata);
3008 out:
3009 	nfs4_put_open_state(state);
3010 	nfs4_put_state_owner(sp);
3011 	return status;
3012 }
3013 
3014 static struct inode *
3015 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3016 		int open_flags, struct iattr *attr, int *opened)
3017 {
3018 	struct nfs4_state *state;
3019 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3020 
3021 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3022 
3023 	/* Protect against concurrent sillydeletes */
3024 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3025 
3026 	nfs4_label_release_security(label);
3027 
3028 	if (IS_ERR(state))
3029 		return ERR_CAST(state);
3030 	return state->inode;
3031 }
3032 
3033 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3034 {
3035 	if (ctx->state == NULL)
3036 		return;
3037 	if (is_sync)
3038 		nfs4_close_sync(ctx->state, ctx->mode);
3039 	else
3040 		nfs4_close_state(ctx->state, ctx->mode);
3041 }
3042 
3043 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3044 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3045 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3046 
3047 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3048 {
3049 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3050 	struct nfs4_server_caps_arg args = {
3051 		.fhandle = fhandle,
3052 		.bitmask = bitmask,
3053 	};
3054 	struct nfs4_server_caps_res res = {};
3055 	struct rpc_message msg = {
3056 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3057 		.rpc_argp = &args,
3058 		.rpc_resp = &res,
3059 	};
3060 	int status;
3061 
3062 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3063 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3064 		     FATTR4_WORD0_LINK_SUPPORT |
3065 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3066 		     FATTR4_WORD0_ACLSUPPORT;
3067 	if (minorversion)
3068 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3069 
3070 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3071 	if (status == 0) {
3072 		/* Sanity check the server answers */
3073 		switch (minorversion) {
3074 		case 0:
3075 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3076 			res.attr_bitmask[2] = 0;
3077 			break;
3078 		case 1:
3079 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3080 			break;
3081 		case 2:
3082 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3083 		}
3084 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3085 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3086 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3087 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3088 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3089 				NFS_CAP_CTIME|NFS_CAP_MTIME|
3090 				NFS_CAP_SECURITY_LABEL);
3091 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3092 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3093 			server->caps |= NFS_CAP_ACLS;
3094 		if (res.has_links != 0)
3095 			server->caps |= NFS_CAP_HARDLINKS;
3096 		if (res.has_symlinks != 0)
3097 			server->caps |= NFS_CAP_SYMLINKS;
3098 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3099 			server->caps |= NFS_CAP_FILEID;
3100 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3101 			server->caps |= NFS_CAP_MODE;
3102 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3103 			server->caps |= NFS_CAP_NLINK;
3104 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3105 			server->caps |= NFS_CAP_OWNER;
3106 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3107 			server->caps |= NFS_CAP_OWNER_GROUP;
3108 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3109 			server->caps |= NFS_CAP_ATIME;
3110 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3111 			server->caps |= NFS_CAP_CTIME;
3112 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3113 			server->caps |= NFS_CAP_MTIME;
3114 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3115 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3116 			server->caps |= NFS_CAP_SECURITY_LABEL;
3117 #endif
3118 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3119 				sizeof(server->attr_bitmask));
3120 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3121 
3122 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3123 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3124 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3125 		server->cache_consistency_bitmask[2] = 0;
3126 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3127 			sizeof(server->exclcreat_bitmask));
3128 		server->acl_bitmask = res.acl_bitmask;
3129 		server->fh_expire_type = res.fh_expire_type;
3130 	}
3131 
3132 	return status;
3133 }
3134 
3135 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3136 {
3137 	struct nfs4_exception exception = { };
3138 	int err;
3139 	do {
3140 		err = nfs4_handle_exception(server,
3141 				_nfs4_server_capabilities(server, fhandle),
3142 				&exception);
3143 	} while (exception.retry);
3144 	return err;
3145 }
3146 
3147 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3148 		struct nfs_fsinfo *info)
3149 {
3150 	u32 bitmask[3];
3151 	struct nfs4_lookup_root_arg args = {
3152 		.bitmask = bitmask,
3153 	};
3154 	struct nfs4_lookup_res res = {
3155 		.server = server,
3156 		.fattr = info->fattr,
3157 		.fh = fhandle,
3158 	};
3159 	struct rpc_message msg = {
3160 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3161 		.rpc_argp = &args,
3162 		.rpc_resp = &res,
3163 	};
3164 
3165 	bitmask[0] = nfs4_fattr_bitmap[0];
3166 	bitmask[1] = nfs4_fattr_bitmap[1];
3167 	/*
3168 	 * Process the label in the upcoming getfattr
3169 	 */
3170 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3171 
3172 	nfs_fattr_init(info->fattr);
3173 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3174 }
3175 
3176 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3177 		struct nfs_fsinfo *info)
3178 {
3179 	struct nfs4_exception exception = { };
3180 	int err;
3181 	do {
3182 		err = _nfs4_lookup_root(server, fhandle, info);
3183 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3184 		switch (err) {
3185 		case 0:
3186 		case -NFS4ERR_WRONGSEC:
3187 			goto out;
3188 		default:
3189 			err = nfs4_handle_exception(server, err, &exception);
3190 		}
3191 	} while (exception.retry);
3192 out:
3193 	return err;
3194 }
3195 
3196 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3197 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3198 {
3199 	struct rpc_auth_create_args auth_args = {
3200 		.pseudoflavor = flavor,
3201 	};
3202 	struct rpc_auth *auth;
3203 	int ret;
3204 
3205 	auth = rpcauth_create(&auth_args, server->client);
3206 	if (IS_ERR(auth)) {
3207 		ret = -EACCES;
3208 		goto out;
3209 	}
3210 	ret = nfs4_lookup_root(server, fhandle, info);
3211 out:
3212 	return ret;
3213 }
3214 
3215 /*
3216  * Retry pseudoroot lookup with various security flavors.  We do this when:
3217  *
3218  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3219  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3220  *
3221  * Returns zero on success, or a negative NFS4ERR value, or a
3222  * negative errno value.
3223  */
3224 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3225 			      struct nfs_fsinfo *info)
3226 {
3227 	/* Per 3530bis 15.33.5 */
3228 	static const rpc_authflavor_t flav_array[] = {
3229 		RPC_AUTH_GSS_KRB5P,
3230 		RPC_AUTH_GSS_KRB5I,
3231 		RPC_AUTH_GSS_KRB5,
3232 		RPC_AUTH_UNIX,			/* courtesy */
3233 		RPC_AUTH_NULL,
3234 	};
3235 	int status = -EPERM;
3236 	size_t i;
3237 
3238 	if (server->auth_info.flavor_len > 0) {
3239 		/* try each flavor specified by user */
3240 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3241 			status = nfs4_lookup_root_sec(server, fhandle, info,
3242 						server->auth_info.flavors[i]);
3243 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3244 				continue;
3245 			break;
3246 		}
3247 	} else {
3248 		/* no flavors specified by user, try default list */
3249 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3250 			status = nfs4_lookup_root_sec(server, fhandle, info,
3251 						      flav_array[i]);
3252 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3253 				continue;
3254 			break;
3255 		}
3256 	}
3257 
3258 	/*
3259 	 * -EACCESS could mean that the user doesn't have correct permissions
3260 	 * to access the mount.  It could also mean that we tried to mount
3261 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3262 	 * existing mount programs don't handle -EACCES very well so it should
3263 	 * be mapped to -EPERM instead.
3264 	 */
3265 	if (status == -EACCES)
3266 		status = -EPERM;
3267 	return status;
3268 }
3269 
3270 static int nfs4_do_find_root_sec(struct nfs_server *server,
3271 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3272 {
3273 	int mv = server->nfs_client->cl_minorversion;
3274 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3275 }
3276 
3277 /**
3278  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3279  * @server: initialized nfs_server handle
3280  * @fhandle: we fill in the pseudo-fs root file handle
3281  * @info: we fill in an FSINFO struct
3282  * @auth_probe: probe the auth flavours
3283  *
3284  * Returns zero on success, or a negative errno.
3285  */
3286 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3287 			 struct nfs_fsinfo *info,
3288 			 bool auth_probe)
3289 {
3290 	int status = 0;
3291 
3292 	if (!auth_probe)
3293 		status = nfs4_lookup_root(server, fhandle, info);
3294 
3295 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3296 		status = nfs4_do_find_root_sec(server, fhandle, info);
3297 
3298 	if (status == 0)
3299 		status = nfs4_server_capabilities(server, fhandle);
3300 	if (status == 0)
3301 		status = nfs4_do_fsinfo(server, fhandle, info);
3302 
3303 	return nfs4_map_errors(status);
3304 }
3305 
3306 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3307 			      struct nfs_fsinfo *info)
3308 {
3309 	int error;
3310 	struct nfs_fattr *fattr = info->fattr;
3311 	struct nfs4_label *label = NULL;
3312 
3313 	error = nfs4_server_capabilities(server, mntfh);
3314 	if (error < 0) {
3315 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3316 		return error;
3317 	}
3318 
3319 	label = nfs4_label_alloc(server, GFP_KERNEL);
3320 	if (IS_ERR(label))
3321 		return PTR_ERR(label);
3322 
3323 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3324 	if (error < 0) {
3325 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3326 		goto err_free_label;
3327 	}
3328 
3329 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3330 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3331 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3332 
3333 err_free_label:
3334 	nfs4_label_free(label);
3335 
3336 	return error;
3337 }
3338 
3339 /*
3340  * Get locations and (maybe) other attributes of a referral.
3341  * Note that we'll actually follow the referral later when
3342  * we detect fsid mismatch in inode revalidation
3343  */
3344 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3345 			     const struct qstr *name, struct nfs_fattr *fattr,
3346 			     struct nfs_fh *fhandle)
3347 {
3348 	int status = -ENOMEM;
3349 	struct page *page = NULL;
3350 	struct nfs4_fs_locations *locations = NULL;
3351 
3352 	page = alloc_page(GFP_KERNEL);
3353 	if (page == NULL)
3354 		goto out;
3355 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3356 	if (locations == NULL)
3357 		goto out;
3358 
3359 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3360 	if (status != 0)
3361 		goto out;
3362 
3363 	/*
3364 	 * If the fsid didn't change, this is a migration event, not a
3365 	 * referral.  Cause us to drop into the exception handler, which
3366 	 * will kick off migration recovery.
3367 	 */
3368 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3369 		dprintk("%s: server did not return a different fsid for"
3370 			" a referral at %s\n", __func__, name->name);
3371 		status = -NFS4ERR_MOVED;
3372 		goto out;
3373 	}
3374 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3375 	nfs_fixup_referral_attributes(&locations->fattr);
3376 
3377 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3378 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3379 	memset(fhandle, 0, sizeof(struct nfs_fh));
3380 out:
3381 	if (page)
3382 		__free_page(page);
3383 	kfree(locations);
3384 	return status;
3385 }
3386 
3387 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3388 				struct nfs_fattr *fattr, struct nfs4_label *label)
3389 {
3390 	struct nfs4_getattr_arg args = {
3391 		.fh = fhandle,
3392 		.bitmask = server->attr_bitmask,
3393 	};
3394 	struct nfs4_getattr_res res = {
3395 		.fattr = fattr,
3396 		.label = label,
3397 		.server = server,
3398 	};
3399 	struct rpc_message msg = {
3400 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3401 		.rpc_argp = &args,
3402 		.rpc_resp = &res,
3403 	};
3404 
3405 	args.bitmask = nfs4_bitmask(server, label);
3406 
3407 	nfs_fattr_init(fattr);
3408 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3409 }
3410 
3411 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3412 				struct nfs_fattr *fattr, struct nfs4_label *label)
3413 {
3414 	struct nfs4_exception exception = { };
3415 	int err;
3416 	do {
3417 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3418 		trace_nfs4_getattr(server, fhandle, fattr, err);
3419 		err = nfs4_handle_exception(server, err,
3420 				&exception);
3421 	} while (exception.retry);
3422 	return err;
3423 }
3424 
3425 /*
3426  * The file is not closed if it is opened due to the a request to change
3427  * the size of the file. The open call will not be needed once the
3428  * VFS layer lookup-intents are implemented.
3429  *
3430  * Close is called when the inode is destroyed.
3431  * If we haven't opened the file for O_WRONLY, we
3432  * need to in the size_change case to obtain a stateid.
3433  *
3434  * Got race?
3435  * Because OPEN is always done by name in nfsv4, it is
3436  * possible that we opened a different file by the same
3437  * name.  We can recognize this race condition, but we
3438  * can't do anything about it besides returning an error.
3439  *
3440  * This will be fixed with VFS changes (lookup-intent).
3441  */
3442 static int
3443 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3444 		  struct iattr *sattr)
3445 {
3446 	struct inode *inode = d_inode(dentry);
3447 	struct rpc_cred *cred = NULL;
3448 	struct nfs4_state *state = NULL;
3449 	struct nfs4_label *label = NULL;
3450 	int status;
3451 
3452 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3453 	    sattr->ia_valid & ATTR_SIZE &&
3454 	    sattr->ia_size < i_size_read(inode))
3455 		pnfs_commit_and_return_layout(inode);
3456 
3457 	nfs_fattr_init(fattr);
3458 
3459 	/* Deal with open(O_TRUNC) */
3460 	if (sattr->ia_valid & ATTR_OPEN)
3461 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3462 
3463 	/* Optimization: if the end result is no change, don't RPC */
3464 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3465 		return 0;
3466 
3467 	/* Search for an existing open(O_WRITE) file */
3468 	if (sattr->ia_valid & ATTR_FILE) {
3469 		struct nfs_open_context *ctx;
3470 
3471 		ctx = nfs_file_open_context(sattr->ia_file);
3472 		if (ctx) {
3473 			cred = ctx->cred;
3474 			state = ctx->state;
3475 		}
3476 	}
3477 
3478 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3479 	if (IS_ERR(label))
3480 		return PTR_ERR(label);
3481 
3482 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3483 	if (status == 0) {
3484 		nfs_setattr_update_inode(inode, sattr, fattr);
3485 		nfs_setsecurity(inode, fattr, label);
3486 	}
3487 	nfs4_label_free(label);
3488 	return status;
3489 }
3490 
3491 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3492 		const struct qstr *name, struct nfs_fh *fhandle,
3493 		struct nfs_fattr *fattr, struct nfs4_label *label)
3494 {
3495 	struct nfs_server *server = NFS_SERVER(dir);
3496 	int		       status;
3497 	struct nfs4_lookup_arg args = {
3498 		.bitmask = server->attr_bitmask,
3499 		.dir_fh = NFS_FH(dir),
3500 		.name = name,
3501 	};
3502 	struct nfs4_lookup_res res = {
3503 		.server = server,
3504 		.fattr = fattr,
3505 		.label = label,
3506 		.fh = fhandle,
3507 	};
3508 	struct rpc_message msg = {
3509 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3510 		.rpc_argp = &args,
3511 		.rpc_resp = &res,
3512 	};
3513 
3514 	args.bitmask = nfs4_bitmask(server, label);
3515 
3516 	nfs_fattr_init(fattr);
3517 
3518 	dprintk("NFS call  lookup %s\n", name->name);
3519 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3520 	dprintk("NFS reply lookup: %d\n", status);
3521 	return status;
3522 }
3523 
3524 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3525 {
3526 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3527 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3528 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3529 	fattr->nlink = 2;
3530 }
3531 
3532 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3533 				   struct qstr *name, struct nfs_fh *fhandle,
3534 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3535 {
3536 	struct nfs4_exception exception = { };
3537 	struct rpc_clnt *client = *clnt;
3538 	int err;
3539 	do {
3540 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3541 		trace_nfs4_lookup(dir, name, err);
3542 		switch (err) {
3543 		case -NFS4ERR_BADNAME:
3544 			err = -ENOENT;
3545 			goto out;
3546 		case -NFS4ERR_MOVED:
3547 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3548 			if (err == -NFS4ERR_MOVED)
3549 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3550 			goto out;
3551 		case -NFS4ERR_WRONGSEC:
3552 			err = -EPERM;
3553 			if (client != *clnt)
3554 				goto out;
3555 			client = nfs4_negotiate_security(client, dir, name);
3556 			if (IS_ERR(client))
3557 				return PTR_ERR(client);
3558 
3559 			exception.retry = 1;
3560 			break;
3561 		default:
3562 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3563 		}
3564 	} while (exception.retry);
3565 
3566 out:
3567 	if (err == 0)
3568 		*clnt = client;
3569 	else if (client != *clnt)
3570 		rpc_shutdown_client(client);
3571 
3572 	return err;
3573 }
3574 
3575 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3576 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3577 			    struct nfs4_label *label)
3578 {
3579 	int status;
3580 	struct rpc_clnt *client = NFS_CLIENT(dir);
3581 
3582 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3583 	if (client != NFS_CLIENT(dir)) {
3584 		rpc_shutdown_client(client);
3585 		nfs_fixup_secinfo_attributes(fattr);
3586 	}
3587 	return status;
3588 }
3589 
3590 struct rpc_clnt *
3591 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3592 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3593 {
3594 	struct rpc_clnt *client = NFS_CLIENT(dir);
3595 	int status;
3596 
3597 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3598 	if (status < 0)
3599 		return ERR_PTR(status);
3600 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3601 }
3602 
3603 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3604 {
3605 	struct nfs_server *server = NFS_SERVER(inode);
3606 	struct nfs4_accessargs args = {
3607 		.fh = NFS_FH(inode),
3608 		.bitmask = server->cache_consistency_bitmask,
3609 	};
3610 	struct nfs4_accessres res = {
3611 		.server = server,
3612 	};
3613 	struct rpc_message msg = {
3614 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3615 		.rpc_argp = &args,
3616 		.rpc_resp = &res,
3617 		.rpc_cred = entry->cred,
3618 	};
3619 	int mode = entry->mask;
3620 	int status = 0;
3621 
3622 	/*
3623 	 * Determine which access bits we want to ask for...
3624 	 */
3625 	if (mode & MAY_READ)
3626 		args.access |= NFS4_ACCESS_READ;
3627 	if (S_ISDIR(inode->i_mode)) {
3628 		if (mode & MAY_WRITE)
3629 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3630 		if (mode & MAY_EXEC)
3631 			args.access |= NFS4_ACCESS_LOOKUP;
3632 	} else {
3633 		if (mode & MAY_WRITE)
3634 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3635 		if (mode & MAY_EXEC)
3636 			args.access |= NFS4_ACCESS_EXECUTE;
3637 	}
3638 
3639 	res.fattr = nfs_alloc_fattr();
3640 	if (res.fattr == NULL)
3641 		return -ENOMEM;
3642 
3643 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3644 	if (!status) {
3645 		nfs_access_set_mask(entry, res.access);
3646 		nfs_refresh_inode(inode, res.fattr);
3647 	}
3648 	nfs_free_fattr(res.fattr);
3649 	return status;
3650 }
3651 
3652 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3653 {
3654 	struct nfs4_exception exception = { };
3655 	int err;
3656 	do {
3657 		err = _nfs4_proc_access(inode, entry);
3658 		trace_nfs4_access(inode, err);
3659 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3660 				&exception);
3661 	} while (exception.retry);
3662 	return err;
3663 }
3664 
3665 /*
3666  * TODO: For the time being, we don't try to get any attributes
3667  * along with any of the zero-copy operations READ, READDIR,
3668  * READLINK, WRITE.
3669  *
3670  * In the case of the first three, we want to put the GETATTR
3671  * after the read-type operation -- this is because it is hard
3672  * to predict the length of a GETATTR response in v4, and thus
3673  * align the READ data correctly.  This means that the GETATTR
3674  * may end up partially falling into the page cache, and we should
3675  * shift it into the 'tail' of the xdr_buf before processing.
3676  * To do this efficiently, we need to know the total length
3677  * of data received, which doesn't seem to be available outside
3678  * of the RPC layer.
3679  *
3680  * In the case of WRITE, we also want to put the GETATTR after
3681  * the operation -- in this case because we want to make sure
3682  * we get the post-operation mtime and size.
3683  *
3684  * Both of these changes to the XDR layer would in fact be quite
3685  * minor, but I decided to leave them for a subsequent patch.
3686  */
3687 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3688 		unsigned int pgbase, unsigned int pglen)
3689 {
3690 	struct nfs4_readlink args = {
3691 		.fh       = NFS_FH(inode),
3692 		.pgbase	  = pgbase,
3693 		.pglen    = pglen,
3694 		.pages    = &page,
3695 	};
3696 	struct nfs4_readlink_res res;
3697 	struct rpc_message msg = {
3698 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3699 		.rpc_argp = &args,
3700 		.rpc_resp = &res,
3701 	};
3702 
3703 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3704 }
3705 
3706 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3707 		unsigned int pgbase, unsigned int pglen)
3708 {
3709 	struct nfs4_exception exception = { };
3710 	int err;
3711 	do {
3712 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3713 		trace_nfs4_readlink(inode, err);
3714 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3715 				&exception);
3716 	} while (exception.retry);
3717 	return err;
3718 }
3719 
3720 /*
3721  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3722  */
3723 static int
3724 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3725 		 int flags)
3726 {
3727 	struct nfs4_label l, *ilabel = NULL;
3728 	struct nfs_open_context *ctx;
3729 	struct nfs4_state *state;
3730 	int status = 0;
3731 
3732 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3733 	if (IS_ERR(ctx))
3734 		return PTR_ERR(ctx);
3735 
3736 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3737 
3738 	sattr->ia_mode &= ~current_umask();
3739 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3740 	if (IS_ERR(state)) {
3741 		status = PTR_ERR(state);
3742 		goto out;
3743 	}
3744 out:
3745 	nfs4_label_release_security(ilabel);
3746 	put_nfs_open_context(ctx);
3747 	return status;
3748 }
3749 
3750 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3751 {
3752 	struct nfs_server *server = NFS_SERVER(dir);
3753 	struct nfs_removeargs args = {
3754 		.fh = NFS_FH(dir),
3755 		.name = *name,
3756 	};
3757 	struct nfs_removeres res = {
3758 		.server = server,
3759 	};
3760 	struct rpc_message msg = {
3761 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3762 		.rpc_argp = &args,
3763 		.rpc_resp = &res,
3764 	};
3765 	int status;
3766 
3767 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3768 	if (status == 0)
3769 		update_changeattr(dir, &res.cinfo);
3770 	return status;
3771 }
3772 
3773 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3774 {
3775 	struct nfs4_exception exception = { };
3776 	int err;
3777 	do {
3778 		err = _nfs4_proc_remove(dir, name);
3779 		trace_nfs4_remove(dir, name, err);
3780 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3781 				&exception);
3782 	} while (exception.retry);
3783 	return err;
3784 }
3785 
3786 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3787 {
3788 	struct nfs_server *server = NFS_SERVER(dir);
3789 	struct nfs_removeargs *args = msg->rpc_argp;
3790 	struct nfs_removeres *res = msg->rpc_resp;
3791 
3792 	res->server = server;
3793 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3794 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3795 
3796 	nfs_fattr_init(res->dir_attr);
3797 }
3798 
3799 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3800 {
3801 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3802 			&data->args.seq_args,
3803 			&data->res.seq_res,
3804 			task);
3805 }
3806 
3807 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3808 {
3809 	struct nfs_unlinkdata *data = task->tk_calldata;
3810 	struct nfs_removeres *res = &data->res;
3811 
3812 	if (!nfs4_sequence_done(task, &res->seq_res))
3813 		return 0;
3814 	if (nfs4_async_handle_error(task, res->server, NULL,
3815 				    &data->timeout) == -EAGAIN)
3816 		return 0;
3817 	update_changeattr(dir, &res->cinfo);
3818 	return 1;
3819 }
3820 
3821 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3822 {
3823 	struct nfs_server *server = NFS_SERVER(dir);
3824 	struct nfs_renameargs *arg = msg->rpc_argp;
3825 	struct nfs_renameres *res = msg->rpc_resp;
3826 
3827 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3828 	res->server = server;
3829 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3830 }
3831 
3832 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3833 {
3834 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3835 			&data->args.seq_args,
3836 			&data->res.seq_res,
3837 			task);
3838 }
3839 
3840 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3841 				 struct inode *new_dir)
3842 {
3843 	struct nfs_renamedata *data = task->tk_calldata;
3844 	struct nfs_renameres *res = &data->res;
3845 
3846 	if (!nfs4_sequence_done(task, &res->seq_res))
3847 		return 0;
3848 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3849 		return 0;
3850 
3851 	update_changeattr(old_dir, &res->old_cinfo);
3852 	update_changeattr(new_dir, &res->new_cinfo);
3853 	return 1;
3854 }
3855 
3856 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3857 {
3858 	struct nfs_server *server = NFS_SERVER(inode);
3859 	struct nfs4_link_arg arg = {
3860 		.fh     = NFS_FH(inode),
3861 		.dir_fh = NFS_FH(dir),
3862 		.name   = name,
3863 		.bitmask = server->attr_bitmask,
3864 	};
3865 	struct nfs4_link_res res = {
3866 		.server = server,
3867 		.label = NULL,
3868 	};
3869 	struct rpc_message msg = {
3870 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3871 		.rpc_argp = &arg,
3872 		.rpc_resp = &res,
3873 	};
3874 	int status = -ENOMEM;
3875 
3876 	res.fattr = nfs_alloc_fattr();
3877 	if (res.fattr == NULL)
3878 		goto out;
3879 
3880 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3881 	if (IS_ERR(res.label)) {
3882 		status = PTR_ERR(res.label);
3883 		goto out;
3884 	}
3885 	arg.bitmask = nfs4_bitmask(server, res.label);
3886 
3887 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3888 	if (!status) {
3889 		update_changeattr(dir, &res.cinfo);
3890 		status = nfs_post_op_update_inode(inode, res.fattr);
3891 		if (!status)
3892 			nfs_setsecurity(inode, res.fattr, res.label);
3893 	}
3894 
3895 
3896 	nfs4_label_free(res.label);
3897 
3898 out:
3899 	nfs_free_fattr(res.fattr);
3900 	return status;
3901 }
3902 
3903 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3904 {
3905 	struct nfs4_exception exception = { };
3906 	int err;
3907 	do {
3908 		err = nfs4_handle_exception(NFS_SERVER(inode),
3909 				_nfs4_proc_link(inode, dir, name),
3910 				&exception);
3911 	} while (exception.retry);
3912 	return err;
3913 }
3914 
3915 struct nfs4_createdata {
3916 	struct rpc_message msg;
3917 	struct nfs4_create_arg arg;
3918 	struct nfs4_create_res res;
3919 	struct nfs_fh fh;
3920 	struct nfs_fattr fattr;
3921 	struct nfs4_label *label;
3922 };
3923 
3924 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3925 		struct qstr *name, struct iattr *sattr, u32 ftype)
3926 {
3927 	struct nfs4_createdata *data;
3928 
3929 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3930 	if (data != NULL) {
3931 		struct nfs_server *server = NFS_SERVER(dir);
3932 
3933 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3934 		if (IS_ERR(data->label))
3935 			goto out_free;
3936 
3937 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3938 		data->msg.rpc_argp = &data->arg;
3939 		data->msg.rpc_resp = &data->res;
3940 		data->arg.dir_fh = NFS_FH(dir);
3941 		data->arg.server = server;
3942 		data->arg.name = name;
3943 		data->arg.attrs = sattr;
3944 		data->arg.ftype = ftype;
3945 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3946 		data->res.server = server;
3947 		data->res.fh = &data->fh;
3948 		data->res.fattr = &data->fattr;
3949 		data->res.label = data->label;
3950 		nfs_fattr_init(data->res.fattr);
3951 	}
3952 	return data;
3953 out_free:
3954 	kfree(data);
3955 	return NULL;
3956 }
3957 
3958 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3959 {
3960 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3961 				    &data->arg.seq_args, &data->res.seq_res, 1);
3962 	if (status == 0) {
3963 		update_changeattr(dir, &data->res.dir_cinfo);
3964 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3965 	}
3966 	return status;
3967 }
3968 
3969 static void nfs4_free_createdata(struct nfs4_createdata *data)
3970 {
3971 	nfs4_label_free(data->label);
3972 	kfree(data);
3973 }
3974 
3975 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3976 		struct page *page, unsigned int len, struct iattr *sattr,
3977 		struct nfs4_label *label)
3978 {
3979 	struct nfs4_createdata *data;
3980 	int status = -ENAMETOOLONG;
3981 
3982 	if (len > NFS4_MAXPATHLEN)
3983 		goto out;
3984 
3985 	status = -ENOMEM;
3986 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3987 	if (data == NULL)
3988 		goto out;
3989 
3990 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3991 	data->arg.u.symlink.pages = &page;
3992 	data->arg.u.symlink.len = len;
3993 	data->arg.label = label;
3994 
3995 	status = nfs4_do_create(dir, dentry, data);
3996 
3997 	nfs4_free_createdata(data);
3998 out:
3999 	return status;
4000 }
4001 
4002 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4003 		struct page *page, unsigned int len, struct iattr *sattr)
4004 {
4005 	struct nfs4_exception exception = { };
4006 	struct nfs4_label l, *label = NULL;
4007 	int err;
4008 
4009 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4010 
4011 	do {
4012 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4013 		trace_nfs4_symlink(dir, &dentry->d_name, err);
4014 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4015 				&exception);
4016 	} while (exception.retry);
4017 
4018 	nfs4_label_release_security(label);
4019 	return err;
4020 }
4021 
4022 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4023 		struct iattr *sattr, struct nfs4_label *label)
4024 {
4025 	struct nfs4_createdata *data;
4026 	int status = -ENOMEM;
4027 
4028 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4029 	if (data == NULL)
4030 		goto out;
4031 
4032 	data->arg.label = label;
4033 	status = nfs4_do_create(dir, dentry, data);
4034 
4035 	nfs4_free_createdata(data);
4036 out:
4037 	return status;
4038 }
4039 
4040 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4041 		struct iattr *sattr)
4042 {
4043 	struct nfs4_exception exception = { };
4044 	struct nfs4_label l, *label = NULL;
4045 	int err;
4046 
4047 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4048 
4049 	sattr->ia_mode &= ~current_umask();
4050 	do {
4051 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4052 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
4053 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4054 				&exception);
4055 	} while (exception.retry);
4056 	nfs4_label_release_security(label);
4057 
4058 	return err;
4059 }
4060 
4061 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4062 		u64 cookie, struct page **pages, unsigned int count, int plus)
4063 {
4064 	struct inode		*dir = d_inode(dentry);
4065 	struct nfs4_readdir_arg args = {
4066 		.fh = NFS_FH(dir),
4067 		.pages = pages,
4068 		.pgbase = 0,
4069 		.count = count,
4070 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4071 		.plus = plus,
4072 	};
4073 	struct nfs4_readdir_res res;
4074 	struct rpc_message msg = {
4075 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4076 		.rpc_argp = &args,
4077 		.rpc_resp = &res,
4078 		.rpc_cred = cred,
4079 	};
4080 	int			status;
4081 
4082 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4083 			dentry,
4084 			(unsigned long long)cookie);
4085 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4086 	res.pgbase = args.pgbase;
4087 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4088 	if (status >= 0) {
4089 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4090 		status += args.pgbase;
4091 	}
4092 
4093 	nfs_invalidate_atime(dir);
4094 
4095 	dprintk("%s: returns %d\n", __func__, status);
4096 	return status;
4097 }
4098 
4099 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4100 		u64 cookie, struct page **pages, unsigned int count, int plus)
4101 {
4102 	struct nfs4_exception exception = { };
4103 	int err;
4104 	do {
4105 		err = _nfs4_proc_readdir(dentry, cred, cookie,
4106 				pages, count, plus);
4107 		trace_nfs4_readdir(d_inode(dentry), err);
4108 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4109 				&exception);
4110 	} while (exception.retry);
4111 	return err;
4112 }
4113 
4114 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4115 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4116 {
4117 	struct nfs4_createdata *data;
4118 	int mode = sattr->ia_mode;
4119 	int status = -ENOMEM;
4120 
4121 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4122 	if (data == NULL)
4123 		goto out;
4124 
4125 	if (S_ISFIFO(mode))
4126 		data->arg.ftype = NF4FIFO;
4127 	else if (S_ISBLK(mode)) {
4128 		data->arg.ftype = NF4BLK;
4129 		data->arg.u.device.specdata1 = MAJOR(rdev);
4130 		data->arg.u.device.specdata2 = MINOR(rdev);
4131 	}
4132 	else if (S_ISCHR(mode)) {
4133 		data->arg.ftype = NF4CHR;
4134 		data->arg.u.device.specdata1 = MAJOR(rdev);
4135 		data->arg.u.device.specdata2 = MINOR(rdev);
4136 	} else if (!S_ISSOCK(mode)) {
4137 		status = -EINVAL;
4138 		goto out_free;
4139 	}
4140 
4141 	data->arg.label = label;
4142 	status = nfs4_do_create(dir, dentry, data);
4143 out_free:
4144 	nfs4_free_createdata(data);
4145 out:
4146 	return status;
4147 }
4148 
4149 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4150 		struct iattr *sattr, dev_t rdev)
4151 {
4152 	struct nfs4_exception exception = { };
4153 	struct nfs4_label l, *label = NULL;
4154 	int err;
4155 
4156 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4157 
4158 	sattr->ia_mode &= ~current_umask();
4159 	do {
4160 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4161 		trace_nfs4_mknod(dir, &dentry->d_name, err);
4162 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4163 				&exception);
4164 	} while (exception.retry);
4165 
4166 	nfs4_label_release_security(label);
4167 
4168 	return err;
4169 }
4170 
4171 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4172 		 struct nfs_fsstat *fsstat)
4173 {
4174 	struct nfs4_statfs_arg args = {
4175 		.fh = fhandle,
4176 		.bitmask = server->attr_bitmask,
4177 	};
4178 	struct nfs4_statfs_res res = {
4179 		.fsstat = fsstat,
4180 	};
4181 	struct rpc_message msg = {
4182 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4183 		.rpc_argp = &args,
4184 		.rpc_resp = &res,
4185 	};
4186 
4187 	nfs_fattr_init(fsstat->fattr);
4188 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4189 }
4190 
4191 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4192 {
4193 	struct nfs4_exception exception = { };
4194 	int err;
4195 	do {
4196 		err = nfs4_handle_exception(server,
4197 				_nfs4_proc_statfs(server, fhandle, fsstat),
4198 				&exception);
4199 	} while (exception.retry);
4200 	return err;
4201 }
4202 
4203 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4204 		struct nfs_fsinfo *fsinfo)
4205 {
4206 	struct nfs4_fsinfo_arg args = {
4207 		.fh = fhandle,
4208 		.bitmask = server->attr_bitmask,
4209 	};
4210 	struct nfs4_fsinfo_res res = {
4211 		.fsinfo = fsinfo,
4212 	};
4213 	struct rpc_message msg = {
4214 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4215 		.rpc_argp = &args,
4216 		.rpc_resp = &res,
4217 	};
4218 
4219 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4220 }
4221 
4222 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4223 {
4224 	struct nfs4_exception exception = { };
4225 	unsigned long now = jiffies;
4226 	int err;
4227 
4228 	do {
4229 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4230 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4231 		if (err == 0) {
4232 			struct nfs_client *clp = server->nfs_client;
4233 
4234 			spin_lock(&clp->cl_lock);
4235 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4236 			clp->cl_last_renewal = now;
4237 			spin_unlock(&clp->cl_lock);
4238 			break;
4239 		}
4240 		err = nfs4_handle_exception(server, err, &exception);
4241 	} while (exception.retry);
4242 	return err;
4243 }
4244 
4245 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4246 {
4247 	int error;
4248 
4249 	nfs_fattr_init(fsinfo->fattr);
4250 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4251 	if (error == 0) {
4252 		/* block layout checks this! */
4253 		server->pnfs_blksize = fsinfo->blksize;
4254 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4255 	}
4256 
4257 	return error;
4258 }
4259 
4260 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4261 		struct nfs_pathconf *pathconf)
4262 {
4263 	struct nfs4_pathconf_arg args = {
4264 		.fh = fhandle,
4265 		.bitmask = server->attr_bitmask,
4266 	};
4267 	struct nfs4_pathconf_res res = {
4268 		.pathconf = pathconf,
4269 	};
4270 	struct rpc_message msg = {
4271 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4272 		.rpc_argp = &args,
4273 		.rpc_resp = &res,
4274 	};
4275 
4276 	/* None of the pathconf attributes are mandatory to implement */
4277 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4278 		memset(pathconf, 0, sizeof(*pathconf));
4279 		return 0;
4280 	}
4281 
4282 	nfs_fattr_init(pathconf->fattr);
4283 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4284 }
4285 
4286 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4287 		struct nfs_pathconf *pathconf)
4288 {
4289 	struct nfs4_exception exception = { };
4290 	int err;
4291 
4292 	do {
4293 		err = nfs4_handle_exception(server,
4294 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4295 				&exception);
4296 	} while (exception.retry);
4297 	return err;
4298 }
4299 
4300 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4301 		const struct nfs_open_context *ctx,
4302 		const struct nfs_lock_context *l_ctx,
4303 		fmode_t fmode)
4304 {
4305 	const struct nfs_lockowner *lockowner = NULL;
4306 
4307 	if (l_ctx != NULL)
4308 		lockowner = &l_ctx->lockowner;
4309 	return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4310 }
4311 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4312 
4313 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4314 		const struct nfs_open_context *ctx,
4315 		const struct nfs_lock_context *l_ctx,
4316 		fmode_t fmode)
4317 {
4318 	nfs4_stateid current_stateid;
4319 
4320 	/* If the current stateid represents a lost lock, then exit */
4321 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4322 		return true;
4323 	return nfs4_stateid_match(stateid, &current_stateid);
4324 }
4325 
4326 static bool nfs4_error_stateid_expired(int err)
4327 {
4328 	switch (err) {
4329 	case -NFS4ERR_DELEG_REVOKED:
4330 	case -NFS4ERR_ADMIN_REVOKED:
4331 	case -NFS4ERR_BAD_STATEID:
4332 	case -NFS4ERR_STALE_STATEID:
4333 	case -NFS4ERR_OLD_STATEID:
4334 	case -NFS4ERR_OPENMODE:
4335 	case -NFS4ERR_EXPIRED:
4336 		return true;
4337 	}
4338 	return false;
4339 }
4340 
4341 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4342 {
4343 	nfs_invalidate_atime(hdr->inode);
4344 }
4345 
4346 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4347 {
4348 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4349 
4350 	trace_nfs4_read(hdr, task->tk_status);
4351 	if (nfs4_async_handle_error(task, server,
4352 				    hdr->args.context->state,
4353 				    NULL) == -EAGAIN) {
4354 		rpc_restart_call_prepare(task);
4355 		return -EAGAIN;
4356 	}
4357 
4358 	__nfs4_read_done_cb(hdr);
4359 	if (task->tk_status > 0)
4360 		renew_lease(server, hdr->timestamp);
4361 	return 0;
4362 }
4363 
4364 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4365 		struct nfs_pgio_args *args)
4366 {
4367 
4368 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4369 		nfs4_stateid_is_current(&args->stateid,
4370 				args->context,
4371 				args->lock_context,
4372 				FMODE_READ))
4373 		return false;
4374 	rpc_restart_call_prepare(task);
4375 	return true;
4376 }
4377 
4378 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4379 {
4380 
4381 	dprintk("--> %s\n", __func__);
4382 
4383 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4384 		return -EAGAIN;
4385 	if (nfs4_read_stateid_changed(task, &hdr->args))
4386 		return -EAGAIN;
4387 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4388 				    nfs4_read_done_cb(task, hdr);
4389 }
4390 
4391 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4392 				 struct rpc_message *msg)
4393 {
4394 	hdr->timestamp   = jiffies;
4395 	hdr->pgio_done_cb = nfs4_read_done_cb;
4396 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4397 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4398 }
4399 
4400 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4401 				      struct nfs_pgio_header *hdr)
4402 {
4403 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4404 			&hdr->args.seq_args,
4405 			&hdr->res.seq_res,
4406 			task))
4407 		return 0;
4408 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4409 				hdr->args.lock_context,
4410 				hdr->rw_ops->rw_mode) == -EIO)
4411 		return -EIO;
4412 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4413 		return -EIO;
4414 	return 0;
4415 }
4416 
4417 static int nfs4_write_done_cb(struct rpc_task *task,
4418 			      struct nfs_pgio_header *hdr)
4419 {
4420 	struct inode *inode = hdr->inode;
4421 
4422 	trace_nfs4_write(hdr, task->tk_status);
4423 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4424 				    hdr->args.context->state,
4425 				    NULL) == -EAGAIN) {
4426 		rpc_restart_call_prepare(task);
4427 		return -EAGAIN;
4428 	}
4429 	if (task->tk_status >= 0) {
4430 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4431 		nfs_writeback_update_inode(hdr);
4432 	}
4433 	return 0;
4434 }
4435 
4436 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4437 		struct nfs_pgio_args *args)
4438 {
4439 
4440 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4441 		nfs4_stateid_is_current(&args->stateid,
4442 				args->context,
4443 				args->lock_context,
4444 				FMODE_WRITE))
4445 		return false;
4446 	rpc_restart_call_prepare(task);
4447 	return true;
4448 }
4449 
4450 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4451 {
4452 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4453 		return -EAGAIN;
4454 	if (nfs4_write_stateid_changed(task, &hdr->args))
4455 		return -EAGAIN;
4456 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4457 		nfs4_write_done_cb(task, hdr);
4458 }
4459 
4460 static
4461 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4462 {
4463 	/* Don't request attributes for pNFS or O_DIRECT writes */
4464 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4465 		return false;
4466 	/* Otherwise, request attributes if and only if we don't hold
4467 	 * a delegation
4468 	 */
4469 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4470 }
4471 
4472 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4473 				  struct rpc_message *msg)
4474 {
4475 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4476 
4477 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4478 		hdr->args.bitmask = NULL;
4479 		hdr->res.fattr = NULL;
4480 	} else
4481 		hdr->args.bitmask = server->cache_consistency_bitmask;
4482 
4483 	if (!hdr->pgio_done_cb)
4484 		hdr->pgio_done_cb = nfs4_write_done_cb;
4485 	hdr->res.server = server;
4486 	hdr->timestamp   = jiffies;
4487 
4488 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4489 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4490 }
4491 
4492 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4493 {
4494 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4495 			&data->args.seq_args,
4496 			&data->res.seq_res,
4497 			task);
4498 }
4499 
4500 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4501 {
4502 	struct inode *inode = data->inode;
4503 
4504 	trace_nfs4_commit(data, task->tk_status);
4505 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4506 				    NULL, NULL) == -EAGAIN) {
4507 		rpc_restart_call_prepare(task);
4508 		return -EAGAIN;
4509 	}
4510 	return 0;
4511 }
4512 
4513 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4514 {
4515 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4516 		return -EAGAIN;
4517 	return data->commit_done_cb(task, data);
4518 }
4519 
4520 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4521 {
4522 	struct nfs_server *server = NFS_SERVER(data->inode);
4523 
4524 	if (data->commit_done_cb == NULL)
4525 		data->commit_done_cb = nfs4_commit_done_cb;
4526 	data->res.server = server;
4527 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4528 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4529 }
4530 
4531 struct nfs4_renewdata {
4532 	struct nfs_client	*client;
4533 	unsigned long		timestamp;
4534 };
4535 
4536 /*
4537  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4538  * standalone procedure for queueing an asynchronous RENEW.
4539  */
4540 static void nfs4_renew_release(void *calldata)
4541 {
4542 	struct nfs4_renewdata *data = calldata;
4543 	struct nfs_client *clp = data->client;
4544 
4545 	if (atomic_read(&clp->cl_count) > 1)
4546 		nfs4_schedule_state_renewal(clp);
4547 	nfs_put_client(clp);
4548 	kfree(data);
4549 }
4550 
4551 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4552 {
4553 	struct nfs4_renewdata *data = calldata;
4554 	struct nfs_client *clp = data->client;
4555 	unsigned long timestamp = data->timestamp;
4556 
4557 	trace_nfs4_renew_async(clp, task->tk_status);
4558 	switch (task->tk_status) {
4559 	case 0:
4560 		break;
4561 	case -NFS4ERR_LEASE_MOVED:
4562 		nfs4_schedule_lease_moved_recovery(clp);
4563 		break;
4564 	default:
4565 		/* Unless we're shutting down, schedule state recovery! */
4566 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4567 			return;
4568 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4569 			nfs4_schedule_lease_recovery(clp);
4570 			return;
4571 		}
4572 		nfs4_schedule_path_down_recovery(clp);
4573 	}
4574 	do_renew_lease(clp, timestamp);
4575 }
4576 
4577 static const struct rpc_call_ops nfs4_renew_ops = {
4578 	.rpc_call_done = nfs4_renew_done,
4579 	.rpc_release = nfs4_renew_release,
4580 };
4581 
4582 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4583 {
4584 	struct rpc_message msg = {
4585 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4586 		.rpc_argp	= clp,
4587 		.rpc_cred	= cred,
4588 	};
4589 	struct nfs4_renewdata *data;
4590 
4591 	if (renew_flags == 0)
4592 		return 0;
4593 	if (!atomic_inc_not_zero(&clp->cl_count))
4594 		return -EIO;
4595 	data = kmalloc(sizeof(*data), GFP_NOFS);
4596 	if (data == NULL)
4597 		return -ENOMEM;
4598 	data->client = clp;
4599 	data->timestamp = jiffies;
4600 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4601 			&nfs4_renew_ops, data);
4602 }
4603 
4604 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4605 {
4606 	struct rpc_message msg = {
4607 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4608 		.rpc_argp	= clp,
4609 		.rpc_cred	= cred,
4610 	};
4611 	unsigned long now = jiffies;
4612 	int status;
4613 
4614 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4615 	if (status < 0)
4616 		return status;
4617 	do_renew_lease(clp, now);
4618 	return 0;
4619 }
4620 
4621 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4622 {
4623 	return server->caps & NFS_CAP_ACLS;
4624 }
4625 
4626 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4627  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4628  * the stack.
4629  */
4630 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4631 
4632 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4633 		struct page **pages)
4634 {
4635 	struct page *newpage, **spages;
4636 	int rc = 0;
4637 	size_t len;
4638 	spages = pages;
4639 
4640 	do {
4641 		len = min_t(size_t, PAGE_SIZE, buflen);
4642 		newpage = alloc_page(GFP_KERNEL);
4643 
4644 		if (newpage == NULL)
4645 			goto unwind;
4646 		memcpy(page_address(newpage), buf, len);
4647                 buf += len;
4648                 buflen -= len;
4649 		*pages++ = newpage;
4650 		rc++;
4651 	} while (buflen != 0);
4652 
4653 	return rc;
4654 
4655 unwind:
4656 	for(; rc > 0; rc--)
4657 		__free_page(spages[rc-1]);
4658 	return -ENOMEM;
4659 }
4660 
4661 struct nfs4_cached_acl {
4662 	int cached;
4663 	size_t len;
4664 	char data[0];
4665 };
4666 
4667 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4668 {
4669 	struct nfs_inode *nfsi = NFS_I(inode);
4670 
4671 	spin_lock(&inode->i_lock);
4672 	kfree(nfsi->nfs4_acl);
4673 	nfsi->nfs4_acl = acl;
4674 	spin_unlock(&inode->i_lock);
4675 }
4676 
4677 static void nfs4_zap_acl_attr(struct inode *inode)
4678 {
4679 	nfs4_set_cached_acl(inode, NULL);
4680 }
4681 
4682 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4683 {
4684 	struct nfs_inode *nfsi = NFS_I(inode);
4685 	struct nfs4_cached_acl *acl;
4686 	int ret = -ENOENT;
4687 
4688 	spin_lock(&inode->i_lock);
4689 	acl = nfsi->nfs4_acl;
4690 	if (acl == NULL)
4691 		goto out;
4692 	if (buf == NULL) /* user is just asking for length */
4693 		goto out_len;
4694 	if (acl->cached == 0)
4695 		goto out;
4696 	ret = -ERANGE; /* see getxattr(2) man page */
4697 	if (acl->len > buflen)
4698 		goto out;
4699 	memcpy(buf, acl->data, acl->len);
4700 out_len:
4701 	ret = acl->len;
4702 out:
4703 	spin_unlock(&inode->i_lock);
4704 	return ret;
4705 }
4706 
4707 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4708 {
4709 	struct nfs4_cached_acl *acl;
4710 	size_t buflen = sizeof(*acl) + acl_len;
4711 
4712 	if (buflen <= PAGE_SIZE) {
4713 		acl = kmalloc(buflen, GFP_KERNEL);
4714 		if (acl == NULL)
4715 			goto out;
4716 		acl->cached = 1;
4717 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4718 	} else {
4719 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4720 		if (acl == NULL)
4721 			goto out;
4722 		acl->cached = 0;
4723 	}
4724 	acl->len = acl_len;
4725 out:
4726 	nfs4_set_cached_acl(inode, acl);
4727 }
4728 
4729 /*
4730  * The getxattr API returns the required buffer length when called with a
4731  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4732  * the required buf.  On a NULL buf, we send a page of data to the server
4733  * guessing that the ACL request can be serviced by a page. If so, we cache
4734  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4735  * the cache. If not so, we throw away the page, and cache the required
4736  * length. The next getxattr call will then produce another round trip to
4737  * the server, this time with the input buf of the required size.
4738  */
4739 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4740 {
4741 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4742 	struct nfs_getaclargs args = {
4743 		.fh = NFS_FH(inode),
4744 		.acl_pages = pages,
4745 		.acl_len = buflen,
4746 	};
4747 	struct nfs_getaclres res = {
4748 		.acl_len = buflen,
4749 	};
4750 	struct rpc_message msg = {
4751 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4752 		.rpc_argp = &args,
4753 		.rpc_resp = &res,
4754 	};
4755 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4756 	int ret = -ENOMEM, i;
4757 
4758 	/* As long as we're doing a round trip to the server anyway,
4759 	 * let's be prepared for a page of acl data. */
4760 	if (npages == 0)
4761 		npages = 1;
4762 	if (npages > ARRAY_SIZE(pages))
4763 		return -ERANGE;
4764 
4765 	for (i = 0; i < npages; i++) {
4766 		pages[i] = alloc_page(GFP_KERNEL);
4767 		if (!pages[i])
4768 			goto out_free;
4769 	}
4770 
4771 	/* for decoding across pages */
4772 	res.acl_scratch = alloc_page(GFP_KERNEL);
4773 	if (!res.acl_scratch)
4774 		goto out_free;
4775 
4776 	args.acl_len = npages * PAGE_SIZE;
4777 
4778 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4779 		__func__, buf, buflen, npages, args.acl_len);
4780 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4781 			     &msg, &args.seq_args, &res.seq_res, 0);
4782 	if (ret)
4783 		goto out_free;
4784 
4785 	/* Handle the case where the passed-in buffer is too short */
4786 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4787 		/* Did the user only issue a request for the acl length? */
4788 		if (buf == NULL)
4789 			goto out_ok;
4790 		ret = -ERANGE;
4791 		goto out_free;
4792 	}
4793 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4794 	if (buf) {
4795 		if (res.acl_len > buflen) {
4796 			ret = -ERANGE;
4797 			goto out_free;
4798 		}
4799 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4800 	}
4801 out_ok:
4802 	ret = res.acl_len;
4803 out_free:
4804 	for (i = 0; i < npages; i++)
4805 		if (pages[i])
4806 			__free_page(pages[i]);
4807 	if (res.acl_scratch)
4808 		__free_page(res.acl_scratch);
4809 	return ret;
4810 }
4811 
4812 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4813 {
4814 	struct nfs4_exception exception = { };
4815 	ssize_t ret;
4816 	do {
4817 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4818 		trace_nfs4_get_acl(inode, ret);
4819 		if (ret >= 0)
4820 			break;
4821 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4822 	} while (exception.retry);
4823 	return ret;
4824 }
4825 
4826 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4827 {
4828 	struct nfs_server *server = NFS_SERVER(inode);
4829 	int ret;
4830 
4831 	if (!nfs4_server_supports_acls(server))
4832 		return -EOPNOTSUPP;
4833 	ret = nfs_revalidate_inode(server, inode);
4834 	if (ret < 0)
4835 		return ret;
4836 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4837 		nfs_zap_acl_cache(inode);
4838 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4839 	if (ret != -ENOENT)
4840 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4841 		 * but no cached acl data, just the acl length */
4842 		return ret;
4843 	return nfs4_get_acl_uncached(inode, buf, buflen);
4844 }
4845 
4846 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4847 {
4848 	struct nfs_server *server = NFS_SERVER(inode);
4849 	struct page *pages[NFS4ACL_MAXPAGES];
4850 	struct nfs_setaclargs arg = {
4851 		.fh		= NFS_FH(inode),
4852 		.acl_pages	= pages,
4853 		.acl_len	= buflen,
4854 	};
4855 	struct nfs_setaclres res;
4856 	struct rpc_message msg = {
4857 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4858 		.rpc_argp	= &arg,
4859 		.rpc_resp	= &res,
4860 	};
4861 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4862 	int ret, i;
4863 
4864 	if (!nfs4_server_supports_acls(server))
4865 		return -EOPNOTSUPP;
4866 	if (npages > ARRAY_SIZE(pages))
4867 		return -ERANGE;
4868 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4869 	if (i < 0)
4870 		return i;
4871 	nfs4_inode_return_delegation(inode);
4872 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4873 
4874 	/*
4875 	 * Free each page after tx, so the only ref left is
4876 	 * held by the network stack
4877 	 */
4878 	for (; i > 0; i--)
4879 		put_page(pages[i-1]);
4880 
4881 	/*
4882 	 * Acl update can result in inode attribute update.
4883 	 * so mark the attribute cache invalid.
4884 	 */
4885 	spin_lock(&inode->i_lock);
4886 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4887 	spin_unlock(&inode->i_lock);
4888 	nfs_access_zap_cache(inode);
4889 	nfs_zap_acl_cache(inode);
4890 	return ret;
4891 }
4892 
4893 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4894 {
4895 	struct nfs4_exception exception = { };
4896 	int err;
4897 	do {
4898 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4899 		trace_nfs4_set_acl(inode, err);
4900 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4901 				&exception);
4902 	} while (exception.retry);
4903 	return err;
4904 }
4905 
4906 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4907 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4908 					size_t buflen)
4909 {
4910 	struct nfs_server *server = NFS_SERVER(inode);
4911 	struct nfs_fattr fattr;
4912 	struct nfs4_label label = {0, 0, buflen, buf};
4913 
4914 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4915 	struct nfs4_getattr_arg arg = {
4916 		.fh		= NFS_FH(inode),
4917 		.bitmask	= bitmask,
4918 	};
4919 	struct nfs4_getattr_res res = {
4920 		.fattr		= &fattr,
4921 		.label		= &label,
4922 		.server		= server,
4923 	};
4924 	struct rpc_message msg = {
4925 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4926 		.rpc_argp	= &arg,
4927 		.rpc_resp	= &res,
4928 	};
4929 	int ret;
4930 
4931 	nfs_fattr_init(&fattr);
4932 
4933 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4934 	if (ret)
4935 		return ret;
4936 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4937 		return -ENOENT;
4938 	if (buflen < label.len)
4939 		return -ERANGE;
4940 	return 0;
4941 }
4942 
4943 static int nfs4_get_security_label(struct inode *inode, void *buf,
4944 					size_t buflen)
4945 {
4946 	struct nfs4_exception exception = { };
4947 	int err;
4948 
4949 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4950 		return -EOPNOTSUPP;
4951 
4952 	do {
4953 		err = _nfs4_get_security_label(inode, buf, buflen);
4954 		trace_nfs4_get_security_label(inode, err);
4955 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4956 				&exception);
4957 	} while (exception.retry);
4958 	return err;
4959 }
4960 
4961 static int _nfs4_do_set_security_label(struct inode *inode,
4962 		struct nfs4_label *ilabel,
4963 		struct nfs_fattr *fattr,
4964 		struct nfs4_label *olabel)
4965 {
4966 
4967 	struct iattr sattr = {0};
4968 	struct nfs_server *server = NFS_SERVER(inode);
4969 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4970 	struct nfs_setattrargs arg = {
4971 		.fh             = NFS_FH(inode),
4972 		.iap            = &sattr,
4973 		.server		= server,
4974 		.bitmask	= bitmask,
4975 		.label		= ilabel,
4976 	};
4977 	struct nfs_setattrres res = {
4978 		.fattr		= fattr,
4979 		.label		= olabel,
4980 		.server		= server,
4981 	};
4982 	struct rpc_message msg = {
4983 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4984 		.rpc_argp       = &arg,
4985 		.rpc_resp       = &res,
4986 	};
4987 	int status;
4988 
4989 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4990 
4991 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4992 	if (status)
4993 		dprintk("%s failed: %d\n", __func__, status);
4994 
4995 	return status;
4996 }
4997 
4998 static int nfs4_do_set_security_label(struct inode *inode,
4999 		struct nfs4_label *ilabel,
5000 		struct nfs_fattr *fattr,
5001 		struct nfs4_label *olabel)
5002 {
5003 	struct nfs4_exception exception = { };
5004 	int err;
5005 
5006 	do {
5007 		err = _nfs4_do_set_security_label(inode, ilabel,
5008 				fattr, olabel);
5009 		trace_nfs4_set_security_label(inode, err);
5010 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5011 				&exception);
5012 	} while (exception.retry);
5013 	return err;
5014 }
5015 
5016 static int
5017 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5018 {
5019 	struct nfs4_label ilabel, *olabel = NULL;
5020 	struct nfs_fattr fattr;
5021 	struct rpc_cred *cred;
5022 	int status;
5023 
5024 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5025 		return -EOPNOTSUPP;
5026 
5027 	nfs_fattr_init(&fattr);
5028 
5029 	ilabel.pi = 0;
5030 	ilabel.lfs = 0;
5031 	ilabel.label = (char *)buf;
5032 	ilabel.len = buflen;
5033 
5034 	cred = rpc_lookup_cred();
5035 	if (IS_ERR(cred))
5036 		return PTR_ERR(cred);
5037 
5038 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5039 	if (IS_ERR(olabel)) {
5040 		status = -PTR_ERR(olabel);
5041 		goto out;
5042 	}
5043 
5044 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5045 	if (status == 0)
5046 		nfs_setsecurity(inode, &fattr, olabel);
5047 
5048 	nfs4_label_free(olabel);
5049 out:
5050 	put_rpccred(cred);
5051 	return status;
5052 }
5053 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
5054 
5055 
5056 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5057 				    nfs4_verifier *bootverf)
5058 {
5059 	__be32 verf[2];
5060 
5061 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5062 		/* An impossible timestamp guarantees this value
5063 		 * will never match a generated boot time. */
5064 		verf[0] = 0;
5065 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5066 	} else {
5067 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5068 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5069 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5070 	}
5071 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
5072 }
5073 
5074 static int
5075 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5076 {
5077 	size_t len;
5078 	char *str;
5079 
5080 	if (clp->cl_owner_id != NULL)
5081 		return 0;
5082 
5083 	rcu_read_lock();
5084 	len = 14 + strlen(clp->cl_ipaddr) + 1 +
5085 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5086 		1 +
5087 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5088 		1;
5089 	rcu_read_unlock();
5090 
5091 	if (len > NFS4_OPAQUE_LIMIT + 1)
5092 		return -EINVAL;
5093 
5094 	/*
5095 	 * Since this string is allocated at mount time, and held until the
5096 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5097 	 * about a memory-reclaim deadlock.
5098 	 */
5099 	str = kmalloc(len, GFP_KERNEL);
5100 	if (!str)
5101 		return -ENOMEM;
5102 
5103 	rcu_read_lock();
5104 	scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5105 			clp->cl_ipaddr,
5106 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5107 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5108 	rcu_read_unlock();
5109 
5110 	clp->cl_owner_id = str;
5111 	return 0;
5112 }
5113 
5114 static int
5115 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5116 {
5117 	size_t len;
5118 	char *str;
5119 
5120 	len = 10 + 10 + 1 + 10 + 1 +
5121 		strlen(nfs4_client_id_uniquifier) + 1 +
5122 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5123 
5124 	if (len > NFS4_OPAQUE_LIMIT + 1)
5125 		return -EINVAL;
5126 
5127 	/*
5128 	 * Since this string is allocated at mount time, and held until the
5129 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5130 	 * about a memory-reclaim deadlock.
5131 	 */
5132 	str = kmalloc(len, GFP_KERNEL);
5133 	if (!str)
5134 		return -ENOMEM;
5135 
5136 	scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5137 			clp->rpc_ops->version, clp->cl_minorversion,
5138 			nfs4_client_id_uniquifier,
5139 			clp->cl_rpcclient->cl_nodename);
5140 	clp->cl_owner_id = str;
5141 	return 0;
5142 }
5143 
5144 static int
5145 nfs4_init_uniform_client_string(struct nfs_client *clp)
5146 {
5147 	size_t len;
5148 	char *str;
5149 
5150 	if (clp->cl_owner_id != NULL)
5151 		return 0;
5152 
5153 	if (nfs4_client_id_uniquifier[0] != '\0')
5154 		return nfs4_init_uniquifier_client_string(clp);
5155 
5156 	len = 10 + 10 + 1 + 10 + 1 +
5157 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5158 
5159 	if (len > NFS4_OPAQUE_LIMIT + 1)
5160 		return -EINVAL;
5161 
5162 	/*
5163 	 * Since this string is allocated at mount time, and held until the
5164 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5165 	 * about a memory-reclaim deadlock.
5166 	 */
5167 	str = kmalloc(len, GFP_KERNEL);
5168 	if (!str)
5169 		return -ENOMEM;
5170 
5171 	scnprintf(str, len, "Linux NFSv%u.%u %s",
5172 			clp->rpc_ops->version, clp->cl_minorversion,
5173 			clp->cl_rpcclient->cl_nodename);
5174 	clp->cl_owner_id = str;
5175 	return 0;
5176 }
5177 
5178 /*
5179  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5180  * services.  Advertise one based on the address family of the
5181  * clientaddr.
5182  */
5183 static unsigned int
5184 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5185 {
5186 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5187 		return scnprintf(buf, len, "tcp6");
5188 	else
5189 		return scnprintf(buf, len, "tcp");
5190 }
5191 
5192 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5193 {
5194 	struct nfs4_setclientid *sc = calldata;
5195 
5196 	if (task->tk_status == 0)
5197 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5198 }
5199 
5200 static const struct rpc_call_ops nfs4_setclientid_ops = {
5201 	.rpc_call_done = nfs4_setclientid_done,
5202 };
5203 
5204 /**
5205  * nfs4_proc_setclientid - Negotiate client ID
5206  * @clp: state data structure
5207  * @program: RPC program for NFSv4 callback service
5208  * @port: IP port number for NFS4 callback service
5209  * @cred: RPC credential to use for this call
5210  * @res: where to place the result
5211  *
5212  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5213  */
5214 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5215 		unsigned short port, struct rpc_cred *cred,
5216 		struct nfs4_setclientid_res *res)
5217 {
5218 	nfs4_verifier sc_verifier;
5219 	struct nfs4_setclientid setclientid = {
5220 		.sc_verifier = &sc_verifier,
5221 		.sc_prog = program,
5222 		.sc_clnt = clp,
5223 	};
5224 	struct rpc_message msg = {
5225 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5226 		.rpc_argp = &setclientid,
5227 		.rpc_resp = res,
5228 		.rpc_cred = cred,
5229 	};
5230 	struct rpc_task *task;
5231 	struct rpc_task_setup task_setup_data = {
5232 		.rpc_client = clp->cl_rpcclient,
5233 		.rpc_message = &msg,
5234 		.callback_ops = &nfs4_setclientid_ops,
5235 		.callback_data = &setclientid,
5236 		.flags = RPC_TASK_TIMEOUT,
5237 	};
5238 	int status;
5239 
5240 	/* nfs_client_id4 */
5241 	nfs4_init_boot_verifier(clp, &sc_verifier);
5242 
5243 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5244 		status = nfs4_init_uniform_client_string(clp);
5245 	else
5246 		status = nfs4_init_nonuniform_client_string(clp);
5247 
5248 	if (status)
5249 		goto out;
5250 
5251 	/* cb_client4 */
5252 	setclientid.sc_netid_len =
5253 				nfs4_init_callback_netid(clp,
5254 						setclientid.sc_netid,
5255 						sizeof(setclientid.sc_netid));
5256 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5257 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5258 				clp->cl_ipaddr, port >> 8, port & 255);
5259 
5260 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
5261 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5262 		clp->cl_owner_id);
5263 	task = rpc_run_task(&task_setup_data);
5264 	if (IS_ERR(task)) {
5265 		status = PTR_ERR(task);
5266 		goto out;
5267 	}
5268 	status = task->tk_status;
5269 	if (setclientid.sc_cred) {
5270 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5271 		put_rpccred(setclientid.sc_cred);
5272 	}
5273 	rpc_put_task(task);
5274 out:
5275 	trace_nfs4_setclientid(clp, status);
5276 	dprintk("NFS reply setclientid: %d\n", status);
5277 	return status;
5278 }
5279 
5280 /**
5281  * nfs4_proc_setclientid_confirm - Confirm client ID
5282  * @clp: state data structure
5283  * @res: result of a previous SETCLIENTID
5284  * @cred: RPC credential to use for this call
5285  *
5286  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5287  */
5288 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5289 		struct nfs4_setclientid_res *arg,
5290 		struct rpc_cred *cred)
5291 {
5292 	struct rpc_message msg = {
5293 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5294 		.rpc_argp = arg,
5295 		.rpc_cred = cred,
5296 	};
5297 	int status;
5298 
5299 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5300 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5301 		clp->cl_clientid);
5302 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5303 	trace_nfs4_setclientid_confirm(clp, status);
5304 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5305 	return status;
5306 }
5307 
5308 struct nfs4_delegreturndata {
5309 	struct nfs4_delegreturnargs args;
5310 	struct nfs4_delegreturnres res;
5311 	struct nfs_fh fh;
5312 	nfs4_stateid stateid;
5313 	unsigned long timestamp;
5314 	struct nfs_fattr fattr;
5315 	int rpc_status;
5316 	struct inode *inode;
5317 	bool roc;
5318 	u32 roc_barrier;
5319 };
5320 
5321 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5322 {
5323 	struct nfs4_delegreturndata *data = calldata;
5324 
5325 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5326 		return;
5327 
5328 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5329 	switch (task->tk_status) {
5330 	case 0:
5331 		renew_lease(data->res.server, data->timestamp);
5332 	case -NFS4ERR_ADMIN_REVOKED:
5333 	case -NFS4ERR_DELEG_REVOKED:
5334 	case -NFS4ERR_BAD_STATEID:
5335 	case -NFS4ERR_OLD_STATEID:
5336 	case -NFS4ERR_STALE_STATEID:
5337 	case -NFS4ERR_EXPIRED:
5338 		task->tk_status = 0;
5339 		if (data->roc)
5340 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5341 		break;
5342 	default:
5343 		if (nfs4_async_handle_error(task, data->res.server,
5344 					    NULL, NULL) == -EAGAIN) {
5345 			rpc_restart_call_prepare(task);
5346 			return;
5347 		}
5348 	}
5349 	data->rpc_status = task->tk_status;
5350 }
5351 
5352 static void nfs4_delegreturn_release(void *calldata)
5353 {
5354 	struct nfs4_delegreturndata *data = calldata;
5355 	struct inode *inode = data->inode;
5356 
5357 	if (inode) {
5358 		if (data->roc)
5359 			pnfs_roc_release(inode);
5360 		nfs_iput_and_deactive(inode);
5361 	}
5362 	kfree(calldata);
5363 }
5364 
5365 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5366 {
5367 	struct nfs4_delegreturndata *d_data;
5368 
5369 	d_data = (struct nfs4_delegreturndata *)data;
5370 
5371 	if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5372 		return;
5373 
5374 	if (d_data->roc)
5375 		pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5376 
5377 	nfs4_setup_sequence(d_data->res.server,
5378 			&d_data->args.seq_args,
5379 			&d_data->res.seq_res,
5380 			task);
5381 }
5382 
5383 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5384 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5385 	.rpc_call_done = nfs4_delegreturn_done,
5386 	.rpc_release = nfs4_delegreturn_release,
5387 };
5388 
5389 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5390 {
5391 	struct nfs4_delegreturndata *data;
5392 	struct nfs_server *server = NFS_SERVER(inode);
5393 	struct rpc_task *task;
5394 	struct rpc_message msg = {
5395 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5396 		.rpc_cred = cred,
5397 	};
5398 	struct rpc_task_setup task_setup_data = {
5399 		.rpc_client = server->client,
5400 		.rpc_message = &msg,
5401 		.callback_ops = &nfs4_delegreturn_ops,
5402 		.flags = RPC_TASK_ASYNC,
5403 	};
5404 	int status = 0;
5405 
5406 	data = kzalloc(sizeof(*data), GFP_NOFS);
5407 	if (data == NULL)
5408 		return -ENOMEM;
5409 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5410 
5411 	nfs4_state_protect(server->nfs_client,
5412 			NFS_SP4_MACH_CRED_CLEANUP,
5413 			&task_setup_data.rpc_client, &msg);
5414 
5415 	data->args.fhandle = &data->fh;
5416 	data->args.stateid = &data->stateid;
5417 	data->args.bitmask = server->cache_consistency_bitmask;
5418 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5419 	nfs4_stateid_copy(&data->stateid, stateid);
5420 	data->res.fattr = &data->fattr;
5421 	data->res.server = server;
5422 	nfs_fattr_init(data->res.fattr);
5423 	data->timestamp = jiffies;
5424 	data->rpc_status = 0;
5425 	data->inode = nfs_igrab_and_active(inode);
5426 	if (data->inode)
5427 		data->roc = nfs4_roc(inode);
5428 
5429 	task_setup_data.callback_data = data;
5430 	msg.rpc_argp = &data->args;
5431 	msg.rpc_resp = &data->res;
5432 	task = rpc_run_task(&task_setup_data);
5433 	if (IS_ERR(task))
5434 		return PTR_ERR(task);
5435 	if (!issync)
5436 		goto out;
5437 	status = nfs4_wait_for_completion_rpc_task(task);
5438 	if (status != 0)
5439 		goto out;
5440 	status = data->rpc_status;
5441 	if (status == 0)
5442 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5443 	else
5444 		nfs_refresh_inode(inode, &data->fattr);
5445 out:
5446 	rpc_put_task(task);
5447 	return status;
5448 }
5449 
5450 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5451 {
5452 	struct nfs_server *server = NFS_SERVER(inode);
5453 	struct nfs4_exception exception = { };
5454 	int err;
5455 	do {
5456 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5457 		trace_nfs4_delegreturn(inode, stateid, err);
5458 		switch (err) {
5459 			case -NFS4ERR_STALE_STATEID:
5460 			case -NFS4ERR_EXPIRED:
5461 			case 0:
5462 				return 0;
5463 		}
5464 		err = nfs4_handle_exception(server, err, &exception);
5465 	} while (exception.retry);
5466 	return err;
5467 }
5468 
5469 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5470 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5471 
5472 /*
5473  * sleep, with exponential backoff, and retry the LOCK operation.
5474  */
5475 static unsigned long
5476 nfs4_set_lock_task_retry(unsigned long timeout)
5477 {
5478 	freezable_schedule_timeout_killable_unsafe(timeout);
5479 	timeout <<= 1;
5480 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5481 		return NFS4_LOCK_MAXTIMEOUT;
5482 	return timeout;
5483 }
5484 
5485 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5486 {
5487 	struct inode *inode = state->inode;
5488 	struct nfs_server *server = NFS_SERVER(inode);
5489 	struct nfs_client *clp = server->nfs_client;
5490 	struct nfs_lockt_args arg = {
5491 		.fh = NFS_FH(inode),
5492 		.fl = request,
5493 	};
5494 	struct nfs_lockt_res res = {
5495 		.denied = request,
5496 	};
5497 	struct rpc_message msg = {
5498 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5499 		.rpc_argp       = &arg,
5500 		.rpc_resp       = &res,
5501 		.rpc_cred	= state->owner->so_cred,
5502 	};
5503 	struct nfs4_lock_state *lsp;
5504 	int status;
5505 
5506 	arg.lock_owner.clientid = clp->cl_clientid;
5507 	status = nfs4_set_lock_state(state, request);
5508 	if (status != 0)
5509 		goto out;
5510 	lsp = request->fl_u.nfs4_fl.owner;
5511 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5512 	arg.lock_owner.s_dev = server->s_dev;
5513 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5514 	switch (status) {
5515 		case 0:
5516 			request->fl_type = F_UNLCK;
5517 			break;
5518 		case -NFS4ERR_DENIED:
5519 			status = 0;
5520 	}
5521 	request->fl_ops->fl_release_private(request);
5522 	request->fl_ops = NULL;
5523 out:
5524 	return status;
5525 }
5526 
5527 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5528 {
5529 	struct nfs4_exception exception = { };
5530 	int err;
5531 
5532 	do {
5533 		err = _nfs4_proc_getlk(state, cmd, request);
5534 		trace_nfs4_get_lock(request, state, cmd, err);
5535 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5536 				&exception);
5537 	} while (exception.retry);
5538 	return err;
5539 }
5540 
5541 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5542 {
5543 	return locks_lock_inode_wait(inode, fl);
5544 }
5545 
5546 struct nfs4_unlockdata {
5547 	struct nfs_locku_args arg;
5548 	struct nfs_locku_res res;
5549 	struct nfs4_lock_state *lsp;
5550 	struct nfs_open_context *ctx;
5551 	struct file_lock fl;
5552 	struct nfs_server *server;
5553 	unsigned long timestamp;
5554 };
5555 
5556 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5557 		struct nfs_open_context *ctx,
5558 		struct nfs4_lock_state *lsp,
5559 		struct nfs_seqid *seqid)
5560 {
5561 	struct nfs4_unlockdata *p;
5562 	struct inode *inode = lsp->ls_state->inode;
5563 
5564 	p = kzalloc(sizeof(*p), GFP_NOFS);
5565 	if (p == NULL)
5566 		return NULL;
5567 	p->arg.fh = NFS_FH(inode);
5568 	p->arg.fl = &p->fl;
5569 	p->arg.seqid = seqid;
5570 	p->res.seqid = seqid;
5571 	p->lsp = lsp;
5572 	atomic_inc(&lsp->ls_count);
5573 	/* Ensure we don't close file until we're done freeing locks! */
5574 	p->ctx = get_nfs_open_context(ctx);
5575 	memcpy(&p->fl, fl, sizeof(p->fl));
5576 	p->server = NFS_SERVER(inode);
5577 	return p;
5578 }
5579 
5580 static void nfs4_locku_release_calldata(void *data)
5581 {
5582 	struct nfs4_unlockdata *calldata = data;
5583 	nfs_free_seqid(calldata->arg.seqid);
5584 	nfs4_put_lock_state(calldata->lsp);
5585 	put_nfs_open_context(calldata->ctx);
5586 	kfree(calldata);
5587 }
5588 
5589 static void nfs4_locku_done(struct rpc_task *task, void *data)
5590 {
5591 	struct nfs4_unlockdata *calldata = data;
5592 
5593 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5594 		return;
5595 	switch (task->tk_status) {
5596 		case 0:
5597 			renew_lease(calldata->server, calldata->timestamp);
5598 			do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5599 			if (nfs4_update_lock_stateid(calldata->lsp,
5600 					&calldata->res.stateid))
5601 				break;
5602 		case -NFS4ERR_BAD_STATEID:
5603 		case -NFS4ERR_OLD_STATEID:
5604 		case -NFS4ERR_STALE_STATEID:
5605 		case -NFS4ERR_EXPIRED:
5606 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5607 						&calldata->lsp->ls_stateid))
5608 				rpc_restart_call_prepare(task);
5609 			break;
5610 		default:
5611 			if (nfs4_async_handle_error(task, calldata->server,
5612 						    NULL, NULL) == -EAGAIN)
5613 				rpc_restart_call_prepare(task);
5614 	}
5615 	nfs_release_seqid(calldata->arg.seqid);
5616 }
5617 
5618 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5619 {
5620 	struct nfs4_unlockdata *calldata = data;
5621 
5622 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5623 		goto out_wait;
5624 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5625 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5626 		/* Note: exit _without_ running nfs4_locku_done */
5627 		goto out_no_action;
5628 	}
5629 	calldata->timestamp = jiffies;
5630 	if (nfs4_setup_sequence(calldata->server,
5631 				&calldata->arg.seq_args,
5632 				&calldata->res.seq_res,
5633 				task) != 0)
5634 		nfs_release_seqid(calldata->arg.seqid);
5635 	return;
5636 out_no_action:
5637 	task->tk_action = NULL;
5638 out_wait:
5639 	nfs4_sequence_done(task, &calldata->res.seq_res);
5640 }
5641 
5642 static const struct rpc_call_ops nfs4_locku_ops = {
5643 	.rpc_call_prepare = nfs4_locku_prepare,
5644 	.rpc_call_done = nfs4_locku_done,
5645 	.rpc_release = nfs4_locku_release_calldata,
5646 };
5647 
5648 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5649 		struct nfs_open_context *ctx,
5650 		struct nfs4_lock_state *lsp,
5651 		struct nfs_seqid *seqid)
5652 {
5653 	struct nfs4_unlockdata *data;
5654 	struct rpc_message msg = {
5655 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5656 		.rpc_cred = ctx->cred,
5657 	};
5658 	struct rpc_task_setup task_setup_data = {
5659 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5660 		.rpc_message = &msg,
5661 		.callback_ops = &nfs4_locku_ops,
5662 		.workqueue = nfsiod_workqueue,
5663 		.flags = RPC_TASK_ASYNC,
5664 	};
5665 
5666 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5667 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5668 
5669 	/* Ensure this is an unlock - when canceling a lock, the
5670 	 * canceled lock is passed in, and it won't be an unlock.
5671 	 */
5672 	fl->fl_type = F_UNLCK;
5673 
5674 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5675 	if (data == NULL) {
5676 		nfs_free_seqid(seqid);
5677 		return ERR_PTR(-ENOMEM);
5678 	}
5679 
5680 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5681 	msg.rpc_argp = &data->arg;
5682 	msg.rpc_resp = &data->res;
5683 	task_setup_data.callback_data = data;
5684 	return rpc_run_task(&task_setup_data);
5685 }
5686 
5687 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5688 {
5689 	struct inode *inode = state->inode;
5690 	struct nfs4_state_owner *sp = state->owner;
5691 	struct nfs_inode *nfsi = NFS_I(inode);
5692 	struct nfs_seqid *seqid;
5693 	struct nfs4_lock_state *lsp;
5694 	struct rpc_task *task;
5695 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5696 	int status = 0;
5697 	unsigned char fl_flags = request->fl_flags;
5698 
5699 	status = nfs4_set_lock_state(state, request);
5700 	/* Unlock _before_ we do the RPC call */
5701 	request->fl_flags |= FL_EXISTS;
5702 	/* Exclude nfs_delegation_claim_locks() */
5703 	mutex_lock(&sp->so_delegreturn_mutex);
5704 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5705 	down_read(&nfsi->rwsem);
5706 	if (do_vfs_lock(inode, request) == -ENOENT) {
5707 		up_read(&nfsi->rwsem);
5708 		mutex_unlock(&sp->so_delegreturn_mutex);
5709 		goto out;
5710 	}
5711 	up_read(&nfsi->rwsem);
5712 	mutex_unlock(&sp->so_delegreturn_mutex);
5713 	if (status != 0)
5714 		goto out;
5715 	/* Is this a delegated lock? */
5716 	lsp = request->fl_u.nfs4_fl.owner;
5717 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5718 		goto out;
5719 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5720 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5721 	status = -ENOMEM;
5722 	if (IS_ERR(seqid))
5723 		goto out;
5724 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5725 	status = PTR_ERR(task);
5726 	if (IS_ERR(task))
5727 		goto out;
5728 	status = nfs4_wait_for_completion_rpc_task(task);
5729 	rpc_put_task(task);
5730 out:
5731 	request->fl_flags = fl_flags;
5732 	trace_nfs4_unlock(request, state, F_SETLK, status);
5733 	return status;
5734 }
5735 
5736 struct nfs4_lockdata {
5737 	struct nfs_lock_args arg;
5738 	struct nfs_lock_res res;
5739 	struct nfs4_lock_state *lsp;
5740 	struct nfs_open_context *ctx;
5741 	struct file_lock fl;
5742 	unsigned long timestamp;
5743 	int rpc_status;
5744 	int cancelled;
5745 	struct nfs_server *server;
5746 };
5747 
5748 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5749 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5750 		gfp_t gfp_mask)
5751 {
5752 	struct nfs4_lockdata *p;
5753 	struct inode *inode = lsp->ls_state->inode;
5754 	struct nfs_server *server = NFS_SERVER(inode);
5755 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5756 
5757 	p = kzalloc(sizeof(*p), gfp_mask);
5758 	if (p == NULL)
5759 		return NULL;
5760 
5761 	p->arg.fh = NFS_FH(inode);
5762 	p->arg.fl = &p->fl;
5763 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5764 	if (IS_ERR(p->arg.open_seqid))
5765 		goto out_free;
5766 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5767 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5768 	if (IS_ERR(p->arg.lock_seqid))
5769 		goto out_free_seqid;
5770 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5771 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5772 	p->arg.lock_owner.s_dev = server->s_dev;
5773 	p->res.lock_seqid = p->arg.lock_seqid;
5774 	p->lsp = lsp;
5775 	p->server = server;
5776 	atomic_inc(&lsp->ls_count);
5777 	p->ctx = get_nfs_open_context(ctx);
5778 	get_file(fl->fl_file);
5779 	memcpy(&p->fl, fl, sizeof(p->fl));
5780 	return p;
5781 out_free_seqid:
5782 	nfs_free_seqid(p->arg.open_seqid);
5783 out_free:
5784 	kfree(p);
5785 	return NULL;
5786 }
5787 
5788 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5789 {
5790 	struct nfs4_lockdata *data = calldata;
5791 	struct nfs4_state *state = data->lsp->ls_state;
5792 
5793 	dprintk("%s: begin!\n", __func__);
5794 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5795 		goto out_wait;
5796 	/* Do we need to do an open_to_lock_owner? */
5797 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5798 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5799 			goto out_release_lock_seqid;
5800 		}
5801 		nfs4_stateid_copy(&data->arg.open_stateid,
5802 				&state->open_stateid);
5803 		data->arg.new_lock_owner = 1;
5804 		data->res.open_seqid = data->arg.open_seqid;
5805 	} else {
5806 		data->arg.new_lock_owner = 0;
5807 		nfs4_stateid_copy(&data->arg.lock_stateid,
5808 				&data->lsp->ls_stateid);
5809 	}
5810 	if (!nfs4_valid_open_stateid(state)) {
5811 		data->rpc_status = -EBADF;
5812 		task->tk_action = NULL;
5813 		goto out_release_open_seqid;
5814 	}
5815 	data->timestamp = jiffies;
5816 	if (nfs4_setup_sequence(data->server,
5817 				&data->arg.seq_args,
5818 				&data->res.seq_res,
5819 				task) == 0)
5820 		return;
5821 out_release_open_seqid:
5822 	nfs_release_seqid(data->arg.open_seqid);
5823 out_release_lock_seqid:
5824 	nfs_release_seqid(data->arg.lock_seqid);
5825 out_wait:
5826 	nfs4_sequence_done(task, &data->res.seq_res);
5827 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5828 }
5829 
5830 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5831 {
5832 	struct nfs4_lockdata *data = calldata;
5833 	struct nfs4_lock_state *lsp = data->lsp;
5834 
5835 	dprintk("%s: begin!\n", __func__);
5836 
5837 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5838 		return;
5839 
5840 	data->rpc_status = task->tk_status;
5841 	switch (task->tk_status) {
5842 	case 0:
5843 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5844 				data->timestamp);
5845 		if (data->arg.new_lock) {
5846 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5847 			if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5848 				rpc_restart_call_prepare(task);
5849 				break;
5850 			}
5851 		}
5852 		if (data->arg.new_lock_owner != 0) {
5853 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5854 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5855 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5856 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5857 			rpc_restart_call_prepare(task);
5858 		break;
5859 	case -NFS4ERR_BAD_STATEID:
5860 	case -NFS4ERR_OLD_STATEID:
5861 	case -NFS4ERR_STALE_STATEID:
5862 	case -NFS4ERR_EXPIRED:
5863 		if (data->arg.new_lock_owner != 0) {
5864 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5865 						&lsp->ls_state->open_stateid))
5866 				rpc_restart_call_prepare(task);
5867 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5868 						&lsp->ls_stateid))
5869 				rpc_restart_call_prepare(task);
5870 	}
5871 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5872 }
5873 
5874 static void nfs4_lock_release(void *calldata)
5875 {
5876 	struct nfs4_lockdata *data = calldata;
5877 
5878 	dprintk("%s: begin!\n", __func__);
5879 	nfs_free_seqid(data->arg.open_seqid);
5880 	if (data->cancelled != 0) {
5881 		struct rpc_task *task;
5882 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5883 				data->arg.lock_seqid);
5884 		if (!IS_ERR(task))
5885 			rpc_put_task_async(task);
5886 		dprintk("%s: cancelling lock!\n", __func__);
5887 	} else
5888 		nfs_free_seqid(data->arg.lock_seqid);
5889 	nfs4_put_lock_state(data->lsp);
5890 	put_nfs_open_context(data->ctx);
5891 	fput(data->fl.fl_file);
5892 	kfree(data);
5893 	dprintk("%s: done!\n", __func__);
5894 }
5895 
5896 static const struct rpc_call_ops nfs4_lock_ops = {
5897 	.rpc_call_prepare = nfs4_lock_prepare,
5898 	.rpc_call_done = nfs4_lock_done,
5899 	.rpc_release = nfs4_lock_release,
5900 };
5901 
5902 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5903 {
5904 	switch (error) {
5905 	case -NFS4ERR_ADMIN_REVOKED:
5906 	case -NFS4ERR_BAD_STATEID:
5907 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5908 		if (new_lock_owner != 0 ||
5909 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5910 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5911 		break;
5912 	case -NFS4ERR_STALE_STATEID:
5913 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5914 	case -NFS4ERR_EXPIRED:
5915 		nfs4_schedule_lease_recovery(server->nfs_client);
5916 	};
5917 }
5918 
5919 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5920 {
5921 	struct nfs4_lockdata *data;
5922 	struct rpc_task *task;
5923 	struct rpc_message msg = {
5924 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5925 		.rpc_cred = state->owner->so_cred,
5926 	};
5927 	struct rpc_task_setup task_setup_data = {
5928 		.rpc_client = NFS_CLIENT(state->inode),
5929 		.rpc_message = &msg,
5930 		.callback_ops = &nfs4_lock_ops,
5931 		.workqueue = nfsiod_workqueue,
5932 		.flags = RPC_TASK_ASYNC,
5933 	};
5934 	int ret;
5935 
5936 	dprintk("%s: begin!\n", __func__);
5937 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5938 			fl->fl_u.nfs4_fl.owner,
5939 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5940 	if (data == NULL)
5941 		return -ENOMEM;
5942 	if (IS_SETLKW(cmd))
5943 		data->arg.block = 1;
5944 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5945 	msg.rpc_argp = &data->arg;
5946 	msg.rpc_resp = &data->res;
5947 	task_setup_data.callback_data = data;
5948 	if (recovery_type > NFS_LOCK_NEW) {
5949 		if (recovery_type == NFS_LOCK_RECLAIM)
5950 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5951 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5952 	} else
5953 		data->arg.new_lock = 1;
5954 	task = rpc_run_task(&task_setup_data);
5955 	if (IS_ERR(task))
5956 		return PTR_ERR(task);
5957 	ret = nfs4_wait_for_completion_rpc_task(task);
5958 	if (ret == 0) {
5959 		ret = data->rpc_status;
5960 		if (ret)
5961 			nfs4_handle_setlk_error(data->server, data->lsp,
5962 					data->arg.new_lock_owner, ret);
5963 	} else
5964 		data->cancelled = 1;
5965 	rpc_put_task(task);
5966 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5967 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5968 	return ret;
5969 }
5970 
5971 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5972 {
5973 	struct nfs_server *server = NFS_SERVER(state->inode);
5974 	struct nfs4_exception exception = {
5975 		.inode = state->inode,
5976 	};
5977 	int err;
5978 
5979 	do {
5980 		/* Cache the lock if possible... */
5981 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5982 			return 0;
5983 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5984 		if (err != -NFS4ERR_DELAY)
5985 			break;
5986 		nfs4_handle_exception(server, err, &exception);
5987 	} while (exception.retry);
5988 	return err;
5989 }
5990 
5991 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5992 {
5993 	struct nfs_server *server = NFS_SERVER(state->inode);
5994 	struct nfs4_exception exception = {
5995 		.inode = state->inode,
5996 	};
5997 	int err;
5998 
5999 	err = nfs4_set_lock_state(state, request);
6000 	if (err != 0)
6001 		return err;
6002 	if (!recover_lost_locks) {
6003 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6004 		return 0;
6005 	}
6006 	do {
6007 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6008 			return 0;
6009 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6010 		switch (err) {
6011 		default:
6012 			goto out;
6013 		case -NFS4ERR_GRACE:
6014 		case -NFS4ERR_DELAY:
6015 			nfs4_handle_exception(server, err, &exception);
6016 			err = 0;
6017 		}
6018 	} while (exception.retry);
6019 out:
6020 	return err;
6021 }
6022 
6023 #if defined(CONFIG_NFS_V4_1)
6024 /**
6025  * nfs41_check_expired_locks - possibly free a lock stateid
6026  *
6027  * @state: NFSv4 state for an inode
6028  *
6029  * Returns NFS_OK if recovery for this stateid is now finished.
6030  * Otherwise a negative NFS4ERR value is returned.
6031  */
6032 static int nfs41_check_expired_locks(struct nfs4_state *state)
6033 {
6034 	int status, ret = -NFS4ERR_BAD_STATEID;
6035 	struct nfs4_lock_state *lsp;
6036 	struct nfs_server *server = NFS_SERVER(state->inode);
6037 
6038 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6039 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6040 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6041 
6042 			status = nfs41_test_stateid(server,
6043 					&lsp->ls_stateid,
6044 					cred);
6045 			trace_nfs4_test_lock_stateid(state, lsp, status);
6046 			if (status != NFS_OK) {
6047 				/* Free the stateid unless the server
6048 				 * informs us the stateid is unrecognized. */
6049 				if (status != -NFS4ERR_BAD_STATEID)
6050 					nfs41_free_stateid(server,
6051 							&lsp->ls_stateid,
6052 							cred);
6053 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6054 				ret = status;
6055 			}
6056 		}
6057 	};
6058 
6059 	return ret;
6060 }
6061 
6062 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6063 {
6064 	int status = NFS_OK;
6065 
6066 	if (test_bit(LK_STATE_IN_USE, &state->flags))
6067 		status = nfs41_check_expired_locks(state);
6068 	if (status != NFS_OK)
6069 		status = nfs4_lock_expired(state, request);
6070 	return status;
6071 }
6072 #endif
6073 
6074 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6075 {
6076 	struct nfs_inode *nfsi = NFS_I(state->inode);
6077 	struct nfs4_state_owner *sp = state->owner;
6078 	unsigned char fl_flags = request->fl_flags;
6079 	int status = -ENOLCK;
6080 
6081 	if ((fl_flags & FL_POSIX) &&
6082 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6083 		goto out;
6084 	/* Is this a delegated open? */
6085 	status = nfs4_set_lock_state(state, request);
6086 	if (status != 0)
6087 		goto out;
6088 	request->fl_flags |= FL_ACCESS;
6089 	status = do_vfs_lock(state->inode, request);
6090 	if (status < 0)
6091 		goto out;
6092 	mutex_lock(&sp->so_delegreturn_mutex);
6093 	down_read(&nfsi->rwsem);
6094 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6095 		/* Yes: cache locks! */
6096 		/* ...but avoid races with delegation recall... */
6097 		request->fl_flags = fl_flags & ~FL_SLEEP;
6098 		status = do_vfs_lock(state->inode, request);
6099 		up_read(&nfsi->rwsem);
6100 		mutex_unlock(&sp->so_delegreturn_mutex);
6101 		goto out;
6102 	}
6103 	up_read(&nfsi->rwsem);
6104 	mutex_unlock(&sp->so_delegreturn_mutex);
6105 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6106 out:
6107 	request->fl_flags = fl_flags;
6108 	return status;
6109 }
6110 
6111 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6112 {
6113 	struct nfs4_exception exception = {
6114 		.state = state,
6115 		.inode = state->inode,
6116 	};
6117 	int err;
6118 
6119 	do {
6120 		err = _nfs4_proc_setlk(state, cmd, request);
6121 		if (err == -NFS4ERR_DENIED)
6122 			err = -EAGAIN;
6123 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
6124 				err, &exception);
6125 	} while (exception.retry);
6126 	return err;
6127 }
6128 
6129 static int
6130 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6131 {
6132 	struct nfs_open_context *ctx;
6133 	struct nfs4_state *state;
6134 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6135 	int status;
6136 
6137 	/* verify open state */
6138 	ctx = nfs_file_open_context(filp);
6139 	state = ctx->state;
6140 
6141 	if (request->fl_start < 0 || request->fl_end < 0)
6142 		return -EINVAL;
6143 
6144 	if (IS_GETLK(cmd)) {
6145 		if (state != NULL)
6146 			return nfs4_proc_getlk(state, F_GETLK, request);
6147 		return 0;
6148 	}
6149 
6150 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6151 		return -EINVAL;
6152 
6153 	if (request->fl_type == F_UNLCK) {
6154 		if (state != NULL)
6155 			return nfs4_proc_unlck(state, cmd, request);
6156 		return 0;
6157 	}
6158 
6159 	if (state == NULL)
6160 		return -ENOLCK;
6161 	/*
6162 	 * Don't rely on the VFS having checked the file open mode,
6163 	 * since it won't do this for flock() locks.
6164 	 */
6165 	switch (request->fl_type) {
6166 	case F_RDLCK:
6167 		if (!(filp->f_mode & FMODE_READ))
6168 			return -EBADF;
6169 		break;
6170 	case F_WRLCK:
6171 		if (!(filp->f_mode & FMODE_WRITE))
6172 			return -EBADF;
6173 	}
6174 
6175 	do {
6176 		status = nfs4_proc_setlk(state, cmd, request);
6177 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6178 			break;
6179 		timeout = nfs4_set_lock_task_retry(timeout);
6180 		status = -ERESTARTSYS;
6181 		if (signalled())
6182 			break;
6183 	} while(status < 0);
6184 	return status;
6185 }
6186 
6187 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6188 {
6189 	struct nfs_server *server = NFS_SERVER(state->inode);
6190 	int err;
6191 
6192 	err = nfs4_set_lock_state(state, fl);
6193 	if (err != 0)
6194 		return err;
6195 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6196 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6197 }
6198 
6199 struct nfs_release_lockowner_data {
6200 	struct nfs4_lock_state *lsp;
6201 	struct nfs_server *server;
6202 	struct nfs_release_lockowner_args args;
6203 	struct nfs_release_lockowner_res res;
6204 	unsigned long timestamp;
6205 };
6206 
6207 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6208 {
6209 	struct nfs_release_lockowner_data *data = calldata;
6210 	struct nfs_server *server = data->server;
6211 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6212 			     &data->args.seq_args, &data->res.seq_res, task);
6213 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6214 	data->timestamp = jiffies;
6215 }
6216 
6217 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6218 {
6219 	struct nfs_release_lockowner_data *data = calldata;
6220 	struct nfs_server *server = data->server;
6221 
6222 	nfs40_sequence_done(task, &data->res.seq_res);
6223 
6224 	switch (task->tk_status) {
6225 	case 0:
6226 		renew_lease(server, data->timestamp);
6227 		break;
6228 	case -NFS4ERR_STALE_CLIENTID:
6229 	case -NFS4ERR_EXPIRED:
6230 		nfs4_schedule_lease_recovery(server->nfs_client);
6231 		break;
6232 	case -NFS4ERR_LEASE_MOVED:
6233 	case -NFS4ERR_DELAY:
6234 		if (nfs4_async_handle_error(task, server,
6235 					    NULL, NULL) == -EAGAIN)
6236 			rpc_restart_call_prepare(task);
6237 	}
6238 }
6239 
6240 static void nfs4_release_lockowner_release(void *calldata)
6241 {
6242 	struct nfs_release_lockowner_data *data = calldata;
6243 	nfs4_free_lock_state(data->server, data->lsp);
6244 	kfree(calldata);
6245 }
6246 
6247 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6248 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6249 	.rpc_call_done = nfs4_release_lockowner_done,
6250 	.rpc_release = nfs4_release_lockowner_release,
6251 };
6252 
6253 static void
6254 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6255 {
6256 	struct nfs_release_lockowner_data *data;
6257 	struct rpc_message msg = {
6258 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6259 	};
6260 
6261 	if (server->nfs_client->cl_mvops->minor_version != 0)
6262 		return;
6263 
6264 	data = kmalloc(sizeof(*data), GFP_NOFS);
6265 	if (!data)
6266 		return;
6267 	data->lsp = lsp;
6268 	data->server = server;
6269 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6270 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6271 	data->args.lock_owner.s_dev = server->s_dev;
6272 
6273 	msg.rpc_argp = &data->args;
6274 	msg.rpc_resp = &data->res;
6275 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6276 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6277 }
6278 
6279 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6280 
6281 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6282 				   struct dentry *unused, struct inode *inode,
6283 				   const char *key, const void *buf,
6284 				   size_t buflen, int flags)
6285 {
6286 	return nfs4_proc_set_acl(inode, buf, buflen);
6287 }
6288 
6289 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6290 				   struct dentry *unused, struct inode *inode,
6291 				   const char *key, void *buf, size_t buflen)
6292 {
6293 	return nfs4_proc_get_acl(inode, buf, buflen);
6294 }
6295 
6296 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6297 {
6298 	return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6299 }
6300 
6301 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6302 
6303 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6304 				     struct dentry *unused, struct inode *inode,
6305 				     const char *key, const void *buf,
6306 				     size_t buflen, int flags)
6307 {
6308 	if (security_ismaclabel(key))
6309 		return nfs4_set_security_label(inode, buf, buflen);
6310 
6311 	return -EOPNOTSUPP;
6312 }
6313 
6314 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6315 				     struct dentry *unused, struct inode *inode,
6316 				     const char *key, void *buf, size_t buflen)
6317 {
6318 	if (security_ismaclabel(key))
6319 		return nfs4_get_security_label(inode, buf, buflen);
6320 	return -EOPNOTSUPP;
6321 }
6322 
6323 static ssize_t
6324 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6325 {
6326 	int len = 0;
6327 
6328 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6329 		len = security_inode_listsecurity(inode, list, list_len);
6330 		if (list_len && len > list_len)
6331 			return -ERANGE;
6332 	}
6333 	return len;
6334 }
6335 
6336 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6337 	.prefix = XATTR_SECURITY_PREFIX,
6338 	.get	= nfs4_xattr_get_nfs4_label,
6339 	.set	= nfs4_xattr_set_nfs4_label,
6340 };
6341 
6342 #else
6343 
6344 static ssize_t
6345 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6346 {
6347 	return 0;
6348 }
6349 
6350 #endif
6351 
6352 /*
6353  * nfs_fhget will use either the mounted_on_fileid or the fileid
6354  */
6355 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6356 {
6357 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6358 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6359 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6360 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6361 		return;
6362 
6363 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6364 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6365 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6366 	fattr->nlink = 2;
6367 }
6368 
6369 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6370 				   const struct qstr *name,
6371 				   struct nfs4_fs_locations *fs_locations,
6372 				   struct page *page)
6373 {
6374 	struct nfs_server *server = NFS_SERVER(dir);
6375 	u32 bitmask[3] = {
6376 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6377 	};
6378 	struct nfs4_fs_locations_arg args = {
6379 		.dir_fh = NFS_FH(dir),
6380 		.name = name,
6381 		.page = page,
6382 		.bitmask = bitmask,
6383 	};
6384 	struct nfs4_fs_locations_res res = {
6385 		.fs_locations = fs_locations,
6386 	};
6387 	struct rpc_message msg = {
6388 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6389 		.rpc_argp = &args,
6390 		.rpc_resp = &res,
6391 	};
6392 	int status;
6393 
6394 	dprintk("%s: start\n", __func__);
6395 
6396 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6397 	 * is not supported */
6398 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6399 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6400 	else
6401 		bitmask[0] |= FATTR4_WORD0_FILEID;
6402 
6403 	nfs_fattr_init(&fs_locations->fattr);
6404 	fs_locations->server = server;
6405 	fs_locations->nlocations = 0;
6406 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6407 	dprintk("%s: returned status = %d\n", __func__, status);
6408 	return status;
6409 }
6410 
6411 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6412 			   const struct qstr *name,
6413 			   struct nfs4_fs_locations *fs_locations,
6414 			   struct page *page)
6415 {
6416 	struct nfs4_exception exception = { };
6417 	int err;
6418 	do {
6419 		err = _nfs4_proc_fs_locations(client, dir, name,
6420 				fs_locations, page);
6421 		trace_nfs4_get_fs_locations(dir, name, err);
6422 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6423 				&exception);
6424 	} while (exception.retry);
6425 	return err;
6426 }
6427 
6428 /*
6429  * This operation also signals the server that this client is
6430  * performing migration recovery.  The server can stop returning
6431  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6432  * appended to this compound to identify the client ID which is
6433  * performing recovery.
6434  */
6435 static int _nfs40_proc_get_locations(struct inode *inode,
6436 				     struct nfs4_fs_locations *locations,
6437 				     struct page *page, struct rpc_cred *cred)
6438 {
6439 	struct nfs_server *server = NFS_SERVER(inode);
6440 	struct rpc_clnt *clnt = server->client;
6441 	u32 bitmask[2] = {
6442 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6443 	};
6444 	struct nfs4_fs_locations_arg args = {
6445 		.clientid	= server->nfs_client->cl_clientid,
6446 		.fh		= NFS_FH(inode),
6447 		.page		= page,
6448 		.bitmask	= bitmask,
6449 		.migration	= 1,		/* skip LOOKUP */
6450 		.renew		= 1,		/* append RENEW */
6451 	};
6452 	struct nfs4_fs_locations_res res = {
6453 		.fs_locations	= locations,
6454 		.migration	= 1,
6455 		.renew		= 1,
6456 	};
6457 	struct rpc_message msg = {
6458 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6459 		.rpc_argp	= &args,
6460 		.rpc_resp	= &res,
6461 		.rpc_cred	= cred,
6462 	};
6463 	unsigned long now = jiffies;
6464 	int status;
6465 
6466 	nfs_fattr_init(&locations->fattr);
6467 	locations->server = server;
6468 	locations->nlocations = 0;
6469 
6470 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6471 	nfs4_set_sequence_privileged(&args.seq_args);
6472 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6473 					&args.seq_args, &res.seq_res);
6474 	if (status)
6475 		return status;
6476 
6477 	renew_lease(server, now);
6478 	return 0;
6479 }
6480 
6481 #ifdef CONFIG_NFS_V4_1
6482 
6483 /*
6484  * This operation also signals the server that this client is
6485  * performing migration recovery.  The server can stop asserting
6486  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6487  * performing this operation is identified in the SEQUENCE
6488  * operation in this compound.
6489  *
6490  * When the client supports GETATTR(fs_locations_info), it can
6491  * be plumbed in here.
6492  */
6493 static int _nfs41_proc_get_locations(struct inode *inode,
6494 				     struct nfs4_fs_locations *locations,
6495 				     struct page *page, struct rpc_cred *cred)
6496 {
6497 	struct nfs_server *server = NFS_SERVER(inode);
6498 	struct rpc_clnt *clnt = server->client;
6499 	u32 bitmask[2] = {
6500 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6501 	};
6502 	struct nfs4_fs_locations_arg args = {
6503 		.fh		= NFS_FH(inode),
6504 		.page		= page,
6505 		.bitmask	= bitmask,
6506 		.migration	= 1,		/* skip LOOKUP */
6507 	};
6508 	struct nfs4_fs_locations_res res = {
6509 		.fs_locations	= locations,
6510 		.migration	= 1,
6511 	};
6512 	struct rpc_message msg = {
6513 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6514 		.rpc_argp	= &args,
6515 		.rpc_resp	= &res,
6516 		.rpc_cred	= cred,
6517 	};
6518 	int status;
6519 
6520 	nfs_fattr_init(&locations->fattr);
6521 	locations->server = server;
6522 	locations->nlocations = 0;
6523 
6524 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6525 	nfs4_set_sequence_privileged(&args.seq_args);
6526 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6527 					&args.seq_args, &res.seq_res);
6528 	if (status == NFS4_OK &&
6529 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6530 		status = -NFS4ERR_LEASE_MOVED;
6531 	return status;
6532 }
6533 
6534 #endif	/* CONFIG_NFS_V4_1 */
6535 
6536 /**
6537  * nfs4_proc_get_locations - discover locations for a migrated FSID
6538  * @inode: inode on FSID that is migrating
6539  * @locations: result of query
6540  * @page: buffer
6541  * @cred: credential to use for this operation
6542  *
6543  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6544  * operation failed, or a negative errno if a local error occurred.
6545  *
6546  * On success, "locations" is filled in, but if the server has
6547  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6548  * asserted.
6549  *
6550  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6551  * from this client that require migration recovery.
6552  */
6553 int nfs4_proc_get_locations(struct inode *inode,
6554 			    struct nfs4_fs_locations *locations,
6555 			    struct page *page, struct rpc_cred *cred)
6556 {
6557 	struct nfs_server *server = NFS_SERVER(inode);
6558 	struct nfs_client *clp = server->nfs_client;
6559 	const struct nfs4_mig_recovery_ops *ops =
6560 					clp->cl_mvops->mig_recovery_ops;
6561 	struct nfs4_exception exception = { };
6562 	int status;
6563 
6564 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6565 		(unsigned long long)server->fsid.major,
6566 		(unsigned long long)server->fsid.minor,
6567 		clp->cl_hostname);
6568 	nfs_display_fhandle(NFS_FH(inode), __func__);
6569 
6570 	do {
6571 		status = ops->get_locations(inode, locations, page, cred);
6572 		if (status != -NFS4ERR_DELAY)
6573 			break;
6574 		nfs4_handle_exception(server, status, &exception);
6575 	} while (exception.retry);
6576 	return status;
6577 }
6578 
6579 /*
6580  * This operation also signals the server that this client is
6581  * performing "lease moved" recovery.  The server can stop
6582  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6583  * is appended to this compound to identify the client ID which is
6584  * performing recovery.
6585  */
6586 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6587 {
6588 	struct nfs_server *server = NFS_SERVER(inode);
6589 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6590 	struct rpc_clnt *clnt = server->client;
6591 	struct nfs4_fsid_present_arg args = {
6592 		.fh		= NFS_FH(inode),
6593 		.clientid	= clp->cl_clientid,
6594 		.renew		= 1,		/* append RENEW */
6595 	};
6596 	struct nfs4_fsid_present_res res = {
6597 		.renew		= 1,
6598 	};
6599 	struct rpc_message msg = {
6600 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6601 		.rpc_argp	= &args,
6602 		.rpc_resp	= &res,
6603 		.rpc_cred	= cred,
6604 	};
6605 	unsigned long now = jiffies;
6606 	int status;
6607 
6608 	res.fh = nfs_alloc_fhandle();
6609 	if (res.fh == NULL)
6610 		return -ENOMEM;
6611 
6612 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6613 	nfs4_set_sequence_privileged(&args.seq_args);
6614 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6615 						&args.seq_args, &res.seq_res);
6616 	nfs_free_fhandle(res.fh);
6617 	if (status)
6618 		return status;
6619 
6620 	do_renew_lease(clp, now);
6621 	return 0;
6622 }
6623 
6624 #ifdef CONFIG_NFS_V4_1
6625 
6626 /*
6627  * This operation also signals the server that this client is
6628  * performing "lease moved" recovery.  The server can stop asserting
6629  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6630  * this operation is identified in the SEQUENCE operation in this
6631  * compound.
6632  */
6633 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6634 {
6635 	struct nfs_server *server = NFS_SERVER(inode);
6636 	struct rpc_clnt *clnt = server->client;
6637 	struct nfs4_fsid_present_arg args = {
6638 		.fh		= NFS_FH(inode),
6639 	};
6640 	struct nfs4_fsid_present_res res = {
6641 	};
6642 	struct rpc_message msg = {
6643 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6644 		.rpc_argp	= &args,
6645 		.rpc_resp	= &res,
6646 		.rpc_cred	= cred,
6647 	};
6648 	int status;
6649 
6650 	res.fh = nfs_alloc_fhandle();
6651 	if (res.fh == NULL)
6652 		return -ENOMEM;
6653 
6654 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6655 	nfs4_set_sequence_privileged(&args.seq_args);
6656 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6657 						&args.seq_args, &res.seq_res);
6658 	nfs_free_fhandle(res.fh);
6659 	if (status == NFS4_OK &&
6660 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6661 		status = -NFS4ERR_LEASE_MOVED;
6662 	return status;
6663 }
6664 
6665 #endif	/* CONFIG_NFS_V4_1 */
6666 
6667 /**
6668  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6669  * @inode: inode on FSID to check
6670  * @cred: credential to use for this operation
6671  *
6672  * Server indicates whether the FSID is present, moved, or not
6673  * recognized.  This operation is necessary to clear a LEASE_MOVED
6674  * condition for this client ID.
6675  *
6676  * Returns NFS4_OK if the FSID is present on this server,
6677  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6678  *  NFS4ERR code if some error occurred on the server, or a
6679  *  negative errno if a local failure occurred.
6680  */
6681 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6682 {
6683 	struct nfs_server *server = NFS_SERVER(inode);
6684 	struct nfs_client *clp = server->nfs_client;
6685 	const struct nfs4_mig_recovery_ops *ops =
6686 					clp->cl_mvops->mig_recovery_ops;
6687 	struct nfs4_exception exception = { };
6688 	int status;
6689 
6690 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6691 		(unsigned long long)server->fsid.major,
6692 		(unsigned long long)server->fsid.minor,
6693 		clp->cl_hostname);
6694 	nfs_display_fhandle(NFS_FH(inode), __func__);
6695 
6696 	do {
6697 		status = ops->fsid_present(inode, cred);
6698 		if (status != -NFS4ERR_DELAY)
6699 			break;
6700 		nfs4_handle_exception(server, status, &exception);
6701 	} while (exception.retry);
6702 	return status;
6703 }
6704 
6705 /**
6706  * If 'use_integrity' is true and the state managment nfs_client
6707  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6708  * and the machine credential as per RFC3530bis and RFC5661 Security
6709  * Considerations sections. Otherwise, just use the user cred with the
6710  * filesystem's rpc_client.
6711  */
6712 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6713 {
6714 	int status;
6715 	struct nfs4_secinfo_arg args = {
6716 		.dir_fh = NFS_FH(dir),
6717 		.name   = name,
6718 	};
6719 	struct nfs4_secinfo_res res = {
6720 		.flavors     = flavors,
6721 	};
6722 	struct rpc_message msg = {
6723 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6724 		.rpc_argp = &args,
6725 		.rpc_resp = &res,
6726 	};
6727 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6728 	struct rpc_cred *cred = NULL;
6729 
6730 	if (use_integrity) {
6731 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6732 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6733 		msg.rpc_cred = cred;
6734 	}
6735 
6736 	dprintk("NFS call  secinfo %s\n", name->name);
6737 
6738 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6739 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6740 
6741 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6742 				&res.seq_res, 0);
6743 	dprintk("NFS reply  secinfo: %d\n", status);
6744 
6745 	if (cred)
6746 		put_rpccred(cred);
6747 
6748 	return status;
6749 }
6750 
6751 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6752 		      struct nfs4_secinfo_flavors *flavors)
6753 {
6754 	struct nfs4_exception exception = { };
6755 	int err;
6756 	do {
6757 		err = -NFS4ERR_WRONGSEC;
6758 
6759 		/* try to use integrity protection with machine cred */
6760 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6761 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6762 
6763 		/*
6764 		 * if unable to use integrity protection, or SECINFO with
6765 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6766 		 * disallowed by spec, but exists in deployed servers) use
6767 		 * the current filesystem's rpc_client and the user cred.
6768 		 */
6769 		if (err == -NFS4ERR_WRONGSEC)
6770 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6771 
6772 		trace_nfs4_secinfo(dir, name, err);
6773 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6774 				&exception);
6775 	} while (exception.retry);
6776 	return err;
6777 }
6778 
6779 #ifdef CONFIG_NFS_V4_1
6780 /*
6781  * Check the exchange flags returned by the server for invalid flags, having
6782  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6783  * DS flags set.
6784  */
6785 static int nfs4_check_cl_exchange_flags(u32 flags)
6786 {
6787 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6788 		goto out_inval;
6789 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6790 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6791 		goto out_inval;
6792 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6793 		goto out_inval;
6794 	return NFS_OK;
6795 out_inval:
6796 	return -NFS4ERR_INVAL;
6797 }
6798 
6799 static bool
6800 nfs41_same_server_scope(struct nfs41_server_scope *a,
6801 			struct nfs41_server_scope *b)
6802 {
6803 	if (a->server_scope_sz == b->server_scope_sz &&
6804 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6805 		return true;
6806 
6807 	return false;
6808 }
6809 
6810 static void
6811 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6812 {
6813 }
6814 
6815 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6816 	.rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6817 };
6818 
6819 /*
6820  * nfs4_proc_bind_one_conn_to_session()
6821  *
6822  * The 4.1 client currently uses the same TCP connection for the
6823  * fore and backchannel.
6824  */
6825 static
6826 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6827 		struct rpc_xprt *xprt,
6828 		struct nfs_client *clp,
6829 		struct rpc_cred *cred)
6830 {
6831 	int status;
6832 	struct nfs41_bind_conn_to_session_args args = {
6833 		.client = clp,
6834 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6835 	};
6836 	struct nfs41_bind_conn_to_session_res res;
6837 	struct rpc_message msg = {
6838 		.rpc_proc =
6839 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6840 		.rpc_argp = &args,
6841 		.rpc_resp = &res,
6842 		.rpc_cred = cred,
6843 	};
6844 	struct rpc_task_setup task_setup_data = {
6845 		.rpc_client = clnt,
6846 		.rpc_xprt = xprt,
6847 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
6848 		.rpc_message = &msg,
6849 		.flags = RPC_TASK_TIMEOUT,
6850 	};
6851 	struct rpc_task *task;
6852 
6853 	dprintk("--> %s\n", __func__);
6854 
6855 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6856 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6857 		args.dir = NFS4_CDFC4_FORE;
6858 
6859 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
6860 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
6861 		args.dir = NFS4_CDFC4_FORE;
6862 
6863 	task = rpc_run_task(&task_setup_data);
6864 	if (!IS_ERR(task)) {
6865 		status = task->tk_status;
6866 		rpc_put_task(task);
6867 	} else
6868 		status = PTR_ERR(task);
6869 	trace_nfs4_bind_conn_to_session(clp, status);
6870 	if (status == 0) {
6871 		if (memcmp(res.sessionid.data,
6872 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6873 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6874 			status = -EIO;
6875 			goto out;
6876 		}
6877 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6878 			dprintk("NFS: %s: Unexpected direction from server\n",
6879 				__func__);
6880 			status = -EIO;
6881 			goto out;
6882 		}
6883 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6884 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6885 				__func__);
6886 			status = -EIO;
6887 			goto out;
6888 		}
6889 	}
6890 out:
6891 	dprintk("<-- %s status= %d\n", __func__, status);
6892 	return status;
6893 }
6894 
6895 struct rpc_bind_conn_calldata {
6896 	struct nfs_client *clp;
6897 	struct rpc_cred *cred;
6898 };
6899 
6900 static int
6901 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6902 		struct rpc_xprt *xprt,
6903 		void *calldata)
6904 {
6905 	struct rpc_bind_conn_calldata *p = calldata;
6906 
6907 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6908 }
6909 
6910 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6911 {
6912 	struct rpc_bind_conn_calldata data = {
6913 		.clp = clp,
6914 		.cred = cred,
6915 	};
6916 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6917 			nfs4_proc_bind_conn_to_session_callback, &data);
6918 }
6919 
6920 /*
6921  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6922  * and operations we'd like to see to enable certain features in the allow map
6923  */
6924 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6925 	.how = SP4_MACH_CRED,
6926 	.enforce.u.words = {
6927 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6928 		      1 << (OP_EXCHANGE_ID - 32) |
6929 		      1 << (OP_CREATE_SESSION - 32) |
6930 		      1 << (OP_DESTROY_SESSION - 32) |
6931 		      1 << (OP_DESTROY_CLIENTID - 32)
6932 	},
6933 	.allow.u.words = {
6934 		[0] = 1 << (OP_CLOSE) |
6935 		      1 << (OP_OPEN_DOWNGRADE) |
6936 		      1 << (OP_LOCKU) |
6937 		      1 << (OP_DELEGRETURN) |
6938 		      1 << (OP_COMMIT),
6939 		[1] = 1 << (OP_SECINFO - 32) |
6940 		      1 << (OP_SECINFO_NO_NAME - 32) |
6941 		      1 << (OP_LAYOUTRETURN - 32) |
6942 		      1 << (OP_TEST_STATEID - 32) |
6943 		      1 << (OP_FREE_STATEID - 32) |
6944 		      1 << (OP_WRITE - 32)
6945 	}
6946 };
6947 
6948 /*
6949  * Select the state protection mode for client `clp' given the server results
6950  * from exchange_id in `sp'.
6951  *
6952  * Returns 0 on success, negative errno otherwise.
6953  */
6954 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6955 				 struct nfs41_state_protection *sp)
6956 {
6957 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6958 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6959 		      1 << (OP_EXCHANGE_ID - 32) |
6960 		      1 << (OP_CREATE_SESSION - 32) |
6961 		      1 << (OP_DESTROY_SESSION - 32) |
6962 		      1 << (OP_DESTROY_CLIENTID - 32)
6963 	};
6964 	unsigned int i;
6965 
6966 	if (sp->how == SP4_MACH_CRED) {
6967 		/* Print state protect result */
6968 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6969 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6970 			if (test_bit(i, sp->enforce.u.longs))
6971 				dfprintk(MOUNT, "  enforce op %d\n", i);
6972 			if (test_bit(i, sp->allow.u.longs))
6973 				dfprintk(MOUNT, "  allow op %d\n", i);
6974 		}
6975 
6976 		/* make sure nothing is on enforce list that isn't supported */
6977 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6978 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6979 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6980 				return -EINVAL;
6981 			}
6982 		}
6983 
6984 		/*
6985 		 * Minimal mode - state operations are allowed to use machine
6986 		 * credential.  Note this already happens by default, so the
6987 		 * client doesn't have to do anything more than the negotiation.
6988 		 *
6989 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6990 		 *       we're already using the machine cred for exchange_id
6991 		 *       and will never use a different cred.
6992 		 */
6993 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6994 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6995 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6996 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6997 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6998 			dfprintk(MOUNT, "  minimal mode enabled\n");
6999 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7000 		} else {
7001 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7002 			return -EINVAL;
7003 		}
7004 
7005 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7006 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7007 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7008 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
7009 			dfprintk(MOUNT, "  cleanup mode enabled\n");
7010 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7011 		}
7012 
7013 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7014 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7015 			set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7016 				&clp->cl_sp4_flags);
7017 		}
7018 
7019 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7020 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7021 			dfprintk(MOUNT, "  secinfo mode enabled\n");
7022 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7023 		}
7024 
7025 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7026 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7027 			dfprintk(MOUNT, "  stateid mode enabled\n");
7028 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7029 		}
7030 
7031 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7032 			dfprintk(MOUNT, "  write mode enabled\n");
7033 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7034 		}
7035 
7036 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7037 			dfprintk(MOUNT, "  commit mode enabled\n");
7038 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7039 		}
7040 	}
7041 
7042 	return 0;
7043 }
7044 
7045 /*
7046  * _nfs4_proc_exchange_id()
7047  *
7048  * Wrapper for EXCHANGE_ID operation.
7049  */
7050 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7051 	u32 sp4_how)
7052 {
7053 	nfs4_verifier verifier;
7054 	struct nfs41_exchange_id_args args = {
7055 		.verifier = &verifier,
7056 		.client = clp,
7057 #ifdef CONFIG_NFS_V4_1_MIGRATION
7058 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7059 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
7060 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7061 #else
7062 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7063 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
7064 #endif
7065 	};
7066 	struct nfs41_exchange_id_res res = {
7067 		0
7068 	};
7069 	int status;
7070 	struct rpc_message msg = {
7071 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7072 		.rpc_argp = &args,
7073 		.rpc_resp = &res,
7074 		.rpc_cred = cred,
7075 	};
7076 
7077 	nfs4_init_boot_verifier(clp, &verifier);
7078 
7079 	status = nfs4_init_uniform_client_string(clp);
7080 	if (status)
7081 		goto out;
7082 
7083 	dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7084 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
7085 		clp->cl_owner_id);
7086 
7087 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7088 					GFP_NOFS);
7089 	if (unlikely(res.server_owner == NULL)) {
7090 		status = -ENOMEM;
7091 		goto out;
7092 	}
7093 
7094 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7095 					GFP_NOFS);
7096 	if (unlikely(res.server_scope == NULL)) {
7097 		status = -ENOMEM;
7098 		goto out_server_owner;
7099 	}
7100 
7101 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7102 	if (unlikely(res.impl_id == NULL)) {
7103 		status = -ENOMEM;
7104 		goto out_server_scope;
7105 	}
7106 
7107 	switch (sp4_how) {
7108 	case SP4_NONE:
7109 		args.state_protect.how = SP4_NONE;
7110 		break;
7111 
7112 	case SP4_MACH_CRED:
7113 		args.state_protect = nfs4_sp4_mach_cred_request;
7114 		break;
7115 
7116 	default:
7117 		/* unsupported! */
7118 		WARN_ON_ONCE(1);
7119 		status = -EINVAL;
7120 		goto out_impl_id;
7121 	}
7122 
7123 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7124 	trace_nfs4_exchange_id(clp, status);
7125 	if (status == 0)
7126 		status = nfs4_check_cl_exchange_flags(res.flags);
7127 
7128 	if (status == 0)
7129 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
7130 
7131 	if (status == 0) {
7132 		clp->cl_clientid = res.clientid;
7133 		clp->cl_exchange_flags = res.flags;
7134 		/* Client ID is not confirmed */
7135 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7136 			clear_bit(NFS4_SESSION_ESTABLISHED,
7137 					&clp->cl_session->session_state);
7138 			clp->cl_seqid = res.seqid;
7139 		}
7140 
7141 		kfree(clp->cl_serverowner);
7142 		clp->cl_serverowner = res.server_owner;
7143 		res.server_owner = NULL;
7144 
7145 		/* use the most recent implementation id */
7146 		kfree(clp->cl_implid);
7147 		clp->cl_implid = res.impl_id;
7148 		res.impl_id = NULL;
7149 
7150 		if (clp->cl_serverscope != NULL &&
7151 		    !nfs41_same_server_scope(clp->cl_serverscope,
7152 					     res.server_scope)) {
7153 			dprintk("%s: server_scope mismatch detected\n",
7154 				__func__);
7155 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7156 			kfree(clp->cl_serverscope);
7157 			clp->cl_serverscope = NULL;
7158 		}
7159 
7160 		if (clp->cl_serverscope == NULL) {
7161 			clp->cl_serverscope = res.server_scope;
7162 			res.server_scope = NULL;
7163 		}
7164 	}
7165 
7166 out_impl_id:
7167 	kfree(res.impl_id);
7168 out_server_scope:
7169 	kfree(res.server_scope);
7170 out_server_owner:
7171 	kfree(res.server_owner);
7172 out:
7173 	if (clp->cl_implid != NULL)
7174 		dprintk("NFS reply exchange_id: Server Implementation ID: "
7175 			"domain: %s, name: %s, date: %llu,%u\n",
7176 			clp->cl_implid->domain, clp->cl_implid->name,
7177 			clp->cl_implid->date.seconds,
7178 			clp->cl_implid->date.nseconds);
7179 	dprintk("NFS reply exchange_id: %d\n", status);
7180 	return status;
7181 }
7182 
7183 /*
7184  * nfs4_proc_exchange_id()
7185  *
7186  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7187  *
7188  * Since the clientid has expired, all compounds using sessions
7189  * associated with the stale clientid will be returning
7190  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7191  * be in some phase of session reset.
7192  *
7193  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7194  */
7195 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7196 {
7197 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7198 	int status;
7199 
7200 	/* try SP4_MACH_CRED if krb5i/p	*/
7201 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
7202 	    authflavor == RPC_AUTH_GSS_KRB5P) {
7203 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7204 		if (!status)
7205 			return 0;
7206 	}
7207 
7208 	/* try SP4_NONE */
7209 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7210 }
7211 
7212 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7213 		struct rpc_cred *cred)
7214 {
7215 	struct rpc_message msg = {
7216 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7217 		.rpc_argp = clp,
7218 		.rpc_cred = cred,
7219 	};
7220 	int status;
7221 
7222 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7223 	trace_nfs4_destroy_clientid(clp, status);
7224 	if (status)
7225 		dprintk("NFS: Got error %d from the server %s on "
7226 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
7227 	return status;
7228 }
7229 
7230 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7231 		struct rpc_cred *cred)
7232 {
7233 	unsigned int loop;
7234 	int ret;
7235 
7236 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7237 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7238 		switch (ret) {
7239 		case -NFS4ERR_DELAY:
7240 		case -NFS4ERR_CLIENTID_BUSY:
7241 			ssleep(1);
7242 			break;
7243 		default:
7244 			return ret;
7245 		}
7246 	}
7247 	return 0;
7248 }
7249 
7250 int nfs4_destroy_clientid(struct nfs_client *clp)
7251 {
7252 	struct rpc_cred *cred;
7253 	int ret = 0;
7254 
7255 	if (clp->cl_mvops->minor_version < 1)
7256 		goto out;
7257 	if (clp->cl_exchange_flags == 0)
7258 		goto out;
7259 	if (clp->cl_preserve_clid)
7260 		goto out;
7261 	cred = nfs4_get_clid_cred(clp);
7262 	ret = nfs4_proc_destroy_clientid(clp, cred);
7263 	if (cred)
7264 		put_rpccred(cred);
7265 	switch (ret) {
7266 	case 0:
7267 	case -NFS4ERR_STALE_CLIENTID:
7268 		clp->cl_exchange_flags = 0;
7269 	}
7270 out:
7271 	return ret;
7272 }
7273 
7274 struct nfs4_get_lease_time_data {
7275 	struct nfs4_get_lease_time_args *args;
7276 	struct nfs4_get_lease_time_res *res;
7277 	struct nfs_client *clp;
7278 };
7279 
7280 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7281 					void *calldata)
7282 {
7283 	struct nfs4_get_lease_time_data *data =
7284 			(struct nfs4_get_lease_time_data *)calldata;
7285 
7286 	dprintk("--> %s\n", __func__);
7287 	/* just setup sequence, do not trigger session recovery
7288 	   since we're invoked within one */
7289 	nfs41_setup_sequence(data->clp->cl_session,
7290 			&data->args->la_seq_args,
7291 			&data->res->lr_seq_res,
7292 			task);
7293 	dprintk("<-- %s\n", __func__);
7294 }
7295 
7296 /*
7297  * Called from nfs4_state_manager thread for session setup, so don't recover
7298  * from sequence operation or clientid errors.
7299  */
7300 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7301 {
7302 	struct nfs4_get_lease_time_data *data =
7303 			(struct nfs4_get_lease_time_data *)calldata;
7304 
7305 	dprintk("--> %s\n", __func__);
7306 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7307 		return;
7308 	switch (task->tk_status) {
7309 	case -NFS4ERR_DELAY:
7310 	case -NFS4ERR_GRACE:
7311 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7312 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7313 		task->tk_status = 0;
7314 		/* fall through */
7315 	case -NFS4ERR_RETRY_UNCACHED_REP:
7316 		rpc_restart_call_prepare(task);
7317 		return;
7318 	}
7319 	dprintk("<-- %s\n", __func__);
7320 }
7321 
7322 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7323 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7324 	.rpc_call_done = nfs4_get_lease_time_done,
7325 };
7326 
7327 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7328 {
7329 	struct rpc_task *task;
7330 	struct nfs4_get_lease_time_args args;
7331 	struct nfs4_get_lease_time_res res = {
7332 		.lr_fsinfo = fsinfo,
7333 	};
7334 	struct nfs4_get_lease_time_data data = {
7335 		.args = &args,
7336 		.res = &res,
7337 		.clp = clp,
7338 	};
7339 	struct rpc_message msg = {
7340 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7341 		.rpc_argp = &args,
7342 		.rpc_resp = &res,
7343 	};
7344 	struct rpc_task_setup task_setup = {
7345 		.rpc_client = clp->cl_rpcclient,
7346 		.rpc_message = &msg,
7347 		.callback_ops = &nfs4_get_lease_time_ops,
7348 		.callback_data = &data,
7349 		.flags = RPC_TASK_TIMEOUT,
7350 	};
7351 	int status;
7352 
7353 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7354 	nfs4_set_sequence_privileged(&args.la_seq_args);
7355 	dprintk("--> %s\n", __func__);
7356 	task = rpc_run_task(&task_setup);
7357 
7358 	if (IS_ERR(task))
7359 		status = PTR_ERR(task);
7360 	else {
7361 		status = task->tk_status;
7362 		rpc_put_task(task);
7363 	}
7364 	dprintk("<-- %s return %d\n", __func__, status);
7365 
7366 	return status;
7367 }
7368 
7369 /*
7370  * Initialize the values to be used by the client in CREATE_SESSION
7371  * If nfs4_init_session set the fore channel request and response sizes,
7372  * use them.
7373  *
7374  * Set the back channel max_resp_sz_cached to zero to force the client to
7375  * always set csa_cachethis to FALSE because the current implementation
7376  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7377  */
7378 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7379 				    struct rpc_clnt *clnt)
7380 {
7381 	unsigned int max_rqst_sz, max_resp_sz;
7382 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7383 
7384 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7385 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7386 
7387 	/* Fore channel attributes */
7388 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7389 	args->fc_attrs.max_resp_sz = max_resp_sz;
7390 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7391 	args->fc_attrs.max_reqs = max_session_slots;
7392 
7393 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7394 		"max_ops=%u max_reqs=%u\n",
7395 		__func__,
7396 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7397 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7398 
7399 	/* Back channel attributes */
7400 	args->bc_attrs.max_rqst_sz = max_bc_payload;
7401 	args->bc_attrs.max_resp_sz = max_bc_payload;
7402 	args->bc_attrs.max_resp_sz_cached = 0;
7403 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7404 	args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7405 
7406 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7407 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7408 		__func__,
7409 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7410 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7411 		args->bc_attrs.max_reqs);
7412 }
7413 
7414 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7415 		struct nfs41_create_session_res *res)
7416 {
7417 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7418 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7419 
7420 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7421 		return -EINVAL;
7422 	/*
7423 	 * Our requested max_ops is the minimum we need; we're not
7424 	 * prepared to break up compounds into smaller pieces than that.
7425 	 * So, no point even trying to continue if the server won't
7426 	 * cooperate:
7427 	 */
7428 	if (rcvd->max_ops < sent->max_ops)
7429 		return -EINVAL;
7430 	if (rcvd->max_reqs == 0)
7431 		return -EINVAL;
7432 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7433 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7434 	return 0;
7435 }
7436 
7437 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7438 		struct nfs41_create_session_res *res)
7439 {
7440 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7441 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7442 
7443 	if (!(res->flags & SESSION4_BACK_CHAN))
7444 		goto out;
7445 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7446 		return -EINVAL;
7447 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7448 		return -EINVAL;
7449 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7450 		return -EINVAL;
7451 	/* These would render the backchannel useless: */
7452 	if (rcvd->max_ops != sent->max_ops)
7453 		return -EINVAL;
7454 	if (rcvd->max_reqs != sent->max_reqs)
7455 		return -EINVAL;
7456 out:
7457 	return 0;
7458 }
7459 
7460 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7461 				     struct nfs41_create_session_res *res)
7462 {
7463 	int ret;
7464 
7465 	ret = nfs4_verify_fore_channel_attrs(args, res);
7466 	if (ret)
7467 		return ret;
7468 	return nfs4_verify_back_channel_attrs(args, res);
7469 }
7470 
7471 static void nfs4_update_session(struct nfs4_session *session,
7472 		struct nfs41_create_session_res *res)
7473 {
7474 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7475 	/* Mark client id and session as being confirmed */
7476 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7477 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7478 	session->flags = res->flags;
7479 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7480 	if (res->flags & SESSION4_BACK_CHAN)
7481 		memcpy(&session->bc_attrs, &res->bc_attrs,
7482 				sizeof(session->bc_attrs));
7483 }
7484 
7485 static int _nfs4_proc_create_session(struct nfs_client *clp,
7486 		struct rpc_cred *cred)
7487 {
7488 	struct nfs4_session *session = clp->cl_session;
7489 	struct nfs41_create_session_args args = {
7490 		.client = clp,
7491 		.clientid = clp->cl_clientid,
7492 		.seqid = clp->cl_seqid,
7493 		.cb_program = NFS4_CALLBACK,
7494 	};
7495 	struct nfs41_create_session_res res;
7496 
7497 	struct rpc_message msg = {
7498 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7499 		.rpc_argp = &args,
7500 		.rpc_resp = &res,
7501 		.rpc_cred = cred,
7502 	};
7503 	int status;
7504 
7505 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7506 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7507 
7508 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7509 	trace_nfs4_create_session(clp, status);
7510 
7511 	if (!status) {
7512 		/* Verify the session's negotiated channel_attrs values */
7513 		status = nfs4_verify_channel_attrs(&args, &res);
7514 		/* Increment the clientid slot sequence id */
7515 		if (clp->cl_seqid == res.seqid)
7516 			clp->cl_seqid++;
7517 		if (status)
7518 			goto out;
7519 		nfs4_update_session(session, &res);
7520 	}
7521 out:
7522 	return status;
7523 }
7524 
7525 /*
7526  * Issues a CREATE_SESSION operation to the server.
7527  * It is the responsibility of the caller to verify the session is
7528  * expired before calling this routine.
7529  */
7530 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7531 {
7532 	int status;
7533 	unsigned *ptr;
7534 	struct nfs4_session *session = clp->cl_session;
7535 
7536 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7537 
7538 	status = _nfs4_proc_create_session(clp, cred);
7539 	if (status)
7540 		goto out;
7541 
7542 	/* Init or reset the session slot tables */
7543 	status = nfs4_setup_session_slot_tables(session);
7544 	dprintk("slot table setup returned %d\n", status);
7545 	if (status)
7546 		goto out;
7547 
7548 	ptr = (unsigned *)&session->sess_id.data[0];
7549 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7550 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7551 out:
7552 	dprintk("<-- %s\n", __func__);
7553 	return status;
7554 }
7555 
7556 /*
7557  * Issue the over-the-wire RPC DESTROY_SESSION.
7558  * The caller must serialize access to this routine.
7559  */
7560 int nfs4_proc_destroy_session(struct nfs4_session *session,
7561 		struct rpc_cred *cred)
7562 {
7563 	struct rpc_message msg = {
7564 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7565 		.rpc_argp = session,
7566 		.rpc_cred = cred,
7567 	};
7568 	int status = 0;
7569 
7570 	dprintk("--> nfs4_proc_destroy_session\n");
7571 
7572 	/* session is still being setup */
7573 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7574 		return 0;
7575 
7576 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7577 	trace_nfs4_destroy_session(session->clp, status);
7578 
7579 	if (status)
7580 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7581 			"Session has been destroyed regardless...\n", status);
7582 
7583 	dprintk("<-- nfs4_proc_destroy_session\n");
7584 	return status;
7585 }
7586 
7587 /*
7588  * Renew the cl_session lease.
7589  */
7590 struct nfs4_sequence_data {
7591 	struct nfs_client *clp;
7592 	struct nfs4_sequence_args args;
7593 	struct nfs4_sequence_res res;
7594 };
7595 
7596 static void nfs41_sequence_release(void *data)
7597 {
7598 	struct nfs4_sequence_data *calldata = data;
7599 	struct nfs_client *clp = calldata->clp;
7600 
7601 	if (atomic_read(&clp->cl_count) > 1)
7602 		nfs4_schedule_state_renewal(clp);
7603 	nfs_put_client(clp);
7604 	kfree(calldata);
7605 }
7606 
7607 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7608 {
7609 	switch(task->tk_status) {
7610 	case -NFS4ERR_DELAY:
7611 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7612 		return -EAGAIN;
7613 	default:
7614 		nfs4_schedule_lease_recovery(clp);
7615 	}
7616 	return 0;
7617 }
7618 
7619 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7620 {
7621 	struct nfs4_sequence_data *calldata = data;
7622 	struct nfs_client *clp = calldata->clp;
7623 
7624 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7625 		return;
7626 
7627 	trace_nfs4_sequence(clp, task->tk_status);
7628 	if (task->tk_status < 0) {
7629 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7630 		if (atomic_read(&clp->cl_count) == 1)
7631 			goto out;
7632 
7633 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7634 			rpc_restart_call_prepare(task);
7635 			return;
7636 		}
7637 	}
7638 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7639 out:
7640 	dprintk("<-- %s\n", __func__);
7641 }
7642 
7643 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7644 {
7645 	struct nfs4_sequence_data *calldata = data;
7646 	struct nfs_client *clp = calldata->clp;
7647 	struct nfs4_sequence_args *args;
7648 	struct nfs4_sequence_res *res;
7649 
7650 	args = task->tk_msg.rpc_argp;
7651 	res = task->tk_msg.rpc_resp;
7652 
7653 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7654 }
7655 
7656 static const struct rpc_call_ops nfs41_sequence_ops = {
7657 	.rpc_call_done = nfs41_sequence_call_done,
7658 	.rpc_call_prepare = nfs41_sequence_prepare,
7659 	.rpc_release = nfs41_sequence_release,
7660 };
7661 
7662 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7663 		struct rpc_cred *cred,
7664 		bool is_privileged)
7665 {
7666 	struct nfs4_sequence_data *calldata;
7667 	struct rpc_message msg = {
7668 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7669 		.rpc_cred = cred,
7670 	};
7671 	struct rpc_task_setup task_setup_data = {
7672 		.rpc_client = clp->cl_rpcclient,
7673 		.rpc_message = &msg,
7674 		.callback_ops = &nfs41_sequence_ops,
7675 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7676 	};
7677 
7678 	if (!atomic_inc_not_zero(&clp->cl_count))
7679 		return ERR_PTR(-EIO);
7680 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7681 	if (calldata == NULL) {
7682 		nfs_put_client(clp);
7683 		return ERR_PTR(-ENOMEM);
7684 	}
7685 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7686 	if (is_privileged)
7687 		nfs4_set_sequence_privileged(&calldata->args);
7688 	msg.rpc_argp = &calldata->args;
7689 	msg.rpc_resp = &calldata->res;
7690 	calldata->clp = clp;
7691 	task_setup_data.callback_data = calldata;
7692 
7693 	return rpc_run_task(&task_setup_data);
7694 }
7695 
7696 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7697 {
7698 	struct rpc_task *task;
7699 	int ret = 0;
7700 
7701 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7702 		return -EAGAIN;
7703 	task = _nfs41_proc_sequence(clp, cred, false);
7704 	if (IS_ERR(task))
7705 		ret = PTR_ERR(task);
7706 	else
7707 		rpc_put_task_async(task);
7708 	dprintk("<-- %s status=%d\n", __func__, ret);
7709 	return ret;
7710 }
7711 
7712 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7713 {
7714 	struct rpc_task *task;
7715 	int ret;
7716 
7717 	task = _nfs41_proc_sequence(clp, cred, true);
7718 	if (IS_ERR(task)) {
7719 		ret = PTR_ERR(task);
7720 		goto out;
7721 	}
7722 	ret = rpc_wait_for_completion_task(task);
7723 	if (!ret)
7724 		ret = task->tk_status;
7725 	rpc_put_task(task);
7726 out:
7727 	dprintk("<-- %s status=%d\n", __func__, ret);
7728 	return ret;
7729 }
7730 
7731 struct nfs4_reclaim_complete_data {
7732 	struct nfs_client *clp;
7733 	struct nfs41_reclaim_complete_args arg;
7734 	struct nfs41_reclaim_complete_res res;
7735 };
7736 
7737 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7738 {
7739 	struct nfs4_reclaim_complete_data *calldata = data;
7740 
7741 	nfs41_setup_sequence(calldata->clp->cl_session,
7742 			&calldata->arg.seq_args,
7743 			&calldata->res.seq_res,
7744 			task);
7745 }
7746 
7747 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7748 {
7749 	switch(task->tk_status) {
7750 	case 0:
7751 	case -NFS4ERR_COMPLETE_ALREADY:
7752 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7753 		break;
7754 	case -NFS4ERR_DELAY:
7755 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7756 		/* fall through */
7757 	case -NFS4ERR_RETRY_UNCACHED_REP:
7758 		return -EAGAIN;
7759 	default:
7760 		nfs4_schedule_lease_recovery(clp);
7761 	}
7762 	return 0;
7763 }
7764 
7765 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7766 {
7767 	struct nfs4_reclaim_complete_data *calldata = data;
7768 	struct nfs_client *clp = calldata->clp;
7769 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7770 
7771 	dprintk("--> %s\n", __func__);
7772 	if (!nfs41_sequence_done(task, res))
7773 		return;
7774 
7775 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7776 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7777 		rpc_restart_call_prepare(task);
7778 		return;
7779 	}
7780 	dprintk("<-- %s\n", __func__);
7781 }
7782 
7783 static void nfs4_free_reclaim_complete_data(void *data)
7784 {
7785 	struct nfs4_reclaim_complete_data *calldata = data;
7786 
7787 	kfree(calldata);
7788 }
7789 
7790 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7791 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7792 	.rpc_call_done = nfs4_reclaim_complete_done,
7793 	.rpc_release = nfs4_free_reclaim_complete_data,
7794 };
7795 
7796 /*
7797  * Issue a global reclaim complete.
7798  */
7799 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7800 		struct rpc_cred *cred)
7801 {
7802 	struct nfs4_reclaim_complete_data *calldata;
7803 	struct rpc_task *task;
7804 	struct rpc_message msg = {
7805 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7806 		.rpc_cred = cred,
7807 	};
7808 	struct rpc_task_setup task_setup_data = {
7809 		.rpc_client = clp->cl_rpcclient,
7810 		.rpc_message = &msg,
7811 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7812 		.flags = RPC_TASK_ASYNC,
7813 	};
7814 	int status = -ENOMEM;
7815 
7816 	dprintk("--> %s\n", __func__);
7817 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7818 	if (calldata == NULL)
7819 		goto out;
7820 	calldata->clp = clp;
7821 	calldata->arg.one_fs = 0;
7822 
7823 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7824 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7825 	msg.rpc_argp = &calldata->arg;
7826 	msg.rpc_resp = &calldata->res;
7827 	task_setup_data.callback_data = calldata;
7828 	task = rpc_run_task(&task_setup_data);
7829 	if (IS_ERR(task)) {
7830 		status = PTR_ERR(task);
7831 		goto out;
7832 	}
7833 	status = nfs4_wait_for_completion_rpc_task(task);
7834 	if (status == 0)
7835 		status = task->tk_status;
7836 	rpc_put_task(task);
7837 	return 0;
7838 out:
7839 	dprintk("<-- %s status=%d\n", __func__, status);
7840 	return status;
7841 }
7842 
7843 static void
7844 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7845 {
7846 	struct nfs4_layoutget *lgp = calldata;
7847 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7848 	struct nfs4_session *session = nfs4_get_session(server);
7849 
7850 	dprintk("--> %s\n", __func__);
7851 	nfs41_setup_sequence(session, &lgp->args.seq_args,
7852 				&lgp->res.seq_res, task);
7853 	dprintk("<-- %s\n", __func__);
7854 }
7855 
7856 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7857 {
7858 	struct nfs4_layoutget *lgp = calldata;
7859 
7860 	dprintk("--> %s\n", __func__);
7861 	nfs41_sequence_done(task, &lgp->res.seq_res);
7862 	dprintk("<-- %s\n", __func__);
7863 }
7864 
7865 static int
7866 nfs4_layoutget_handle_exception(struct rpc_task *task,
7867 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
7868 {
7869 	struct inode *inode = lgp->args.inode;
7870 	struct nfs_server *server = NFS_SERVER(inode);
7871 	struct pnfs_layout_hdr *lo;
7872 	int status = task->tk_status;
7873 
7874 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7875 
7876 	switch (status) {
7877 	case 0:
7878 		goto out;
7879 
7880 	/*
7881 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7882 	 * on the file. set tk_status to -ENODATA to tell upper layer to
7883 	 * retry go inband.
7884 	 */
7885 	case -NFS4ERR_LAYOUTUNAVAILABLE:
7886 		status = -ENODATA;
7887 		goto out;
7888 	/*
7889 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7890 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7891 	 */
7892 	case -NFS4ERR_BADLAYOUT:
7893 		status = -EOVERFLOW;
7894 		goto out;
7895 	/*
7896 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7897 	 * (or clients) writing to the same RAID stripe except when
7898 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
7899 	 *
7900 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
7901 	 * while, and then eventually give up.
7902 	 */
7903 	case -NFS4ERR_LAYOUTTRYLATER:
7904 		if (lgp->args.minlength == 0) {
7905 			status = -EOVERFLOW;
7906 			goto out;
7907 		}
7908 		/* Fallthrough */
7909 	case -NFS4ERR_RECALLCONFLICT:
7910 		nfs4_handle_exception(server, -NFS4ERR_RECALLCONFLICT,
7911 					exception);
7912 		status = -ERECALLCONFLICT;
7913 		goto out;
7914 	case -NFS4ERR_EXPIRED:
7915 	case -NFS4ERR_BAD_STATEID:
7916 		exception->timeout = 0;
7917 		spin_lock(&inode->i_lock);
7918 		if (nfs4_stateid_match(&lgp->args.stateid,
7919 					&lgp->args.ctx->state->stateid)) {
7920 			spin_unlock(&inode->i_lock);
7921 			/* If the open stateid was bad, then recover it. */
7922 			exception->state = lgp->args.ctx->state;
7923 			break;
7924 		}
7925 		lo = NFS_I(inode)->layout;
7926 		if (lo && !test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) &&
7927 		    nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
7928 			LIST_HEAD(head);
7929 
7930 			/*
7931 			 * Mark the bad layout state as invalid, then retry
7932 			 * with the current stateid.
7933 			 */
7934 			set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7935 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0);
7936 			spin_unlock(&inode->i_lock);
7937 			pnfs_free_lseg_list(&head);
7938 			status = -EAGAIN;
7939 			goto out;
7940 		} else
7941 			spin_unlock(&inode->i_lock);
7942 	}
7943 
7944 	status = nfs4_handle_exception(server, status, exception);
7945 	if (exception->retry)
7946 		status = -EAGAIN;
7947 out:
7948 	dprintk("<-- %s\n", __func__);
7949 	return status;
7950 }
7951 
7952 static size_t max_response_pages(struct nfs_server *server)
7953 {
7954 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7955 	return nfs_page_array_len(0, max_resp_sz);
7956 }
7957 
7958 static void nfs4_free_pages(struct page **pages, size_t size)
7959 {
7960 	int i;
7961 
7962 	if (!pages)
7963 		return;
7964 
7965 	for (i = 0; i < size; i++) {
7966 		if (!pages[i])
7967 			break;
7968 		__free_page(pages[i]);
7969 	}
7970 	kfree(pages);
7971 }
7972 
7973 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7974 {
7975 	struct page **pages;
7976 	int i;
7977 
7978 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7979 	if (!pages) {
7980 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7981 		return NULL;
7982 	}
7983 
7984 	for (i = 0; i < size; i++) {
7985 		pages[i] = alloc_page(gfp_flags);
7986 		if (!pages[i]) {
7987 			dprintk("%s: failed to allocate page\n", __func__);
7988 			nfs4_free_pages(pages, size);
7989 			return NULL;
7990 		}
7991 	}
7992 
7993 	return pages;
7994 }
7995 
7996 static void nfs4_layoutget_release(void *calldata)
7997 {
7998 	struct nfs4_layoutget *lgp = calldata;
7999 	struct inode *inode = lgp->args.inode;
8000 	struct nfs_server *server = NFS_SERVER(inode);
8001 	size_t max_pages = max_response_pages(server);
8002 
8003 	dprintk("--> %s\n", __func__);
8004 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
8005 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
8006 	put_nfs_open_context(lgp->args.ctx);
8007 	kfree(calldata);
8008 	dprintk("<-- %s\n", __func__);
8009 }
8010 
8011 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8012 	.rpc_call_prepare = nfs4_layoutget_prepare,
8013 	.rpc_call_done = nfs4_layoutget_done,
8014 	.rpc_release = nfs4_layoutget_release,
8015 };
8016 
8017 struct pnfs_layout_segment *
8018 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8019 {
8020 	struct inode *inode = lgp->args.inode;
8021 	struct nfs_server *server = NFS_SERVER(inode);
8022 	size_t max_pages = max_response_pages(server);
8023 	struct rpc_task *task;
8024 	struct rpc_message msg = {
8025 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8026 		.rpc_argp = &lgp->args,
8027 		.rpc_resp = &lgp->res,
8028 		.rpc_cred = lgp->cred,
8029 	};
8030 	struct rpc_task_setup task_setup_data = {
8031 		.rpc_client = server->client,
8032 		.rpc_message = &msg,
8033 		.callback_ops = &nfs4_layoutget_call_ops,
8034 		.callback_data = lgp,
8035 		.flags = RPC_TASK_ASYNC,
8036 	};
8037 	struct pnfs_layout_segment *lseg = NULL;
8038 	struct nfs4_exception exception = {
8039 		.inode = inode,
8040 		.timeout = *timeout,
8041 	};
8042 	int status = 0;
8043 
8044 	dprintk("--> %s\n", __func__);
8045 
8046 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8047 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
8048 
8049 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8050 	if (!lgp->args.layout.pages) {
8051 		nfs4_layoutget_release(lgp);
8052 		return ERR_PTR(-ENOMEM);
8053 	}
8054 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8055 
8056 	lgp->res.layoutp = &lgp->args.layout;
8057 	lgp->res.seq_res.sr_slot = NULL;
8058 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8059 
8060 	task = rpc_run_task(&task_setup_data);
8061 	if (IS_ERR(task))
8062 		return ERR_CAST(task);
8063 	status = nfs4_wait_for_completion_rpc_task(task);
8064 	if (status == 0) {
8065 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8066 		*timeout = exception.timeout;
8067 	}
8068 
8069 	trace_nfs4_layoutget(lgp->args.ctx,
8070 			&lgp->args.range,
8071 			&lgp->res.range,
8072 			&lgp->res.stateid,
8073 			status);
8074 
8075 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8076 	if (status == 0 && lgp->res.layoutp->len)
8077 		lseg = pnfs_layout_process(lgp);
8078 	rpc_put_task(task);
8079 	dprintk("<-- %s status=%d\n", __func__, status);
8080 	if (status)
8081 		return ERR_PTR(status);
8082 	return lseg;
8083 }
8084 
8085 static void
8086 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8087 {
8088 	struct nfs4_layoutreturn *lrp = calldata;
8089 
8090 	dprintk("--> %s\n", __func__);
8091 	nfs41_setup_sequence(lrp->clp->cl_session,
8092 			&lrp->args.seq_args,
8093 			&lrp->res.seq_res,
8094 			task);
8095 }
8096 
8097 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8098 {
8099 	struct nfs4_layoutreturn *lrp = calldata;
8100 	struct nfs_server *server;
8101 
8102 	dprintk("--> %s\n", __func__);
8103 
8104 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8105 		return;
8106 
8107 	server = NFS_SERVER(lrp->args.inode);
8108 	switch (task->tk_status) {
8109 	default:
8110 		task->tk_status = 0;
8111 	case 0:
8112 		break;
8113 	case -NFS4ERR_DELAY:
8114 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8115 			break;
8116 		rpc_restart_call_prepare(task);
8117 		return;
8118 	}
8119 	dprintk("<-- %s\n", __func__);
8120 }
8121 
8122 static void nfs4_layoutreturn_release(void *calldata)
8123 {
8124 	struct nfs4_layoutreturn *lrp = calldata;
8125 	struct pnfs_layout_hdr *lo = lrp->args.layout;
8126 	LIST_HEAD(freeme);
8127 
8128 	dprintk("--> %s\n", __func__);
8129 	spin_lock(&lo->plh_inode->i_lock);
8130 	pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range,
8131 			be32_to_cpu(lrp->args.stateid.seqid));
8132 	pnfs_mark_layout_returned_if_empty(lo);
8133 	if (lrp->res.lrs_present)
8134 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8135 	pnfs_clear_layoutreturn_waitbit(lo);
8136 	spin_unlock(&lo->plh_inode->i_lock);
8137 	pnfs_free_lseg_list(&freeme);
8138 	pnfs_put_layout_hdr(lrp->args.layout);
8139 	nfs_iput_and_deactive(lrp->inode);
8140 	kfree(calldata);
8141 	dprintk("<-- %s\n", __func__);
8142 }
8143 
8144 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8145 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
8146 	.rpc_call_done = nfs4_layoutreturn_done,
8147 	.rpc_release = nfs4_layoutreturn_release,
8148 };
8149 
8150 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8151 {
8152 	struct rpc_task *task;
8153 	struct rpc_message msg = {
8154 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8155 		.rpc_argp = &lrp->args,
8156 		.rpc_resp = &lrp->res,
8157 		.rpc_cred = lrp->cred,
8158 	};
8159 	struct rpc_task_setup task_setup_data = {
8160 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
8161 		.rpc_message = &msg,
8162 		.callback_ops = &nfs4_layoutreturn_call_ops,
8163 		.callback_data = lrp,
8164 	};
8165 	int status = 0;
8166 
8167 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8168 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8169 			&task_setup_data.rpc_client, &msg);
8170 
8171 	dprintk("--> %s\n", __func__);
8172 	if (!sync) {
8173 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8174 		if (!lrp->inode) {
8175 			nfs4_layoutreturn_release(lrp);
8176 			return -EAGAIN;
8177 		}
8178 		task_setup_data.flags |= RPC_TASK_ASYNC;
8179 	}
8180 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8181 	task = rpc_run_task(&task_setup_data);
8182 	if (IS_ERR(task))
8183 		return PTR_ERR(task);
8184 	if (sync)
8185 		status = task->tk_status;
8186 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8187 	dprintk("<-- %s status=%d\n", __func__, status);
8188 	rpc_put_task(task);
8189 	return status;
8190 }
8191 
8192 static int
8193 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8194 		struct pnfs_device *pdev,
8195 		struct rpc_cred *cred)
8196 {
8197 	struct nfs4_getdeviceinfo_args args = {
8198 		.pdev = pdev,
8199 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
8200 			NOTIFY_DEVICEID4_DELETE,
8201 	};
8202 	struct nfs4_getdeviceinfo_res res = {
8203 		.pdev = pdev,
8204 	};
8205 	struct rpc_message msg = {
8206 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8207 		.rpc_argp = &args,
8208 		.rpc_resp = &res,
8209 		.rpc_cred = cred,
8210 	};
8211 	int status;
8212 
8213 	dprintk("--> %s\n", __func__);
8214 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8215 	if (res.notification & ~args.notify_types)
8216 		dprintk("%s: unsupported notification\n", __func__);
8217 	if (res.notification != args.notify_types)
8218 		pdev->nocache = 1;
8219 
8220 	dprintk("<-- %s status=%d\n", __func__, status);
8221 
8222 	return status;
8223 }
8224 
8225 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8226 		struct pnfs_device *pdev,
8227 		struct rpc_cred *cred)
8228 {
8229 	struct nfs4_exception exception = { };
8230 	int err;
8231 
8232 	do {
8233 		err = nfs4_handle_exception(server,
8234 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
8235 					&exception);
8236 	} while (exception.retry);
8237 	return err;
8238 }
8239 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8240 
8241 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8242 {
8243 	struct nfs4_layoutcommit_data *data = calldata;
8244 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8245 	struct nfs4_session *session = nfs4_get_session(server);
8246 
8247 	nfs41_setup_sequence(session,
8248 			&data->args.seq_args,
8249 			&data->res.seq_res,
8250 			task);
8251 }
8252 
8253 static void
8254 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8255 {
8256 	struct nfs4_layoutcommit_data *data = calldata;
8257 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8258 
8259 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8260 		return;
8261 
8262 	switch (task->tk_status) { /* Just ignore these failures */
8263 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8264 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8265 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8266 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8267 		task->tk_status = 0;
8268 	case 0:
8269 		break;
8270 	default:
8271 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8272 			rpc_restart_call_prepare(task);
8273 			return;
8274 		}
8275 	}
8276 }
8277 
8278 static void nfs4_layoutcommit_release(void *calldata)
8279 {
8280 	struct nfs4_layoutcommit_data *data = calldata;
8281 
8282 	pnfs_cleanup_layoutcommit(data);
8283 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8284 					   data->res.fattr);
8285 	put_rpccred(data->cred);
8286 	nfs_iput_and_deactive(data->inode);
8287 	kfree(data);
8288 }
8289 
8290 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8291 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8292 	.rpc_call_done = nfs4_layoutcommit_done,
8293 	.rpc_release = nfs4_layoutcommit_release,
8294 };
8295 
8296 int
8297 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8298 {
8299 	struct rpc_message msg = {
8300 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8301 		.rpc_argp = &data->args,
8302 		.rpc_resp = &data->res,
8303 		.rpc_cred = data->cred,
8304 	};
8305 	struct rpc_task_setup task_setup_data = {
8306 		.task = &data->task,
8307 		.rpc_client = NFS_CLIENT(data->args.inode),
8308 		.rpc_message = &msg,
8309 		.callback_ops = &nfs4_layoutcommit_ops,
8310 		.callback_data = data,
8311 	};
8312 	struct rpc_task *task;
8313 	int status = 0;
8314 
8315 	dprintk("NFS: initiating layoutcommit call. sync %d "
8316 		"lbw: %llu inode %lu\n", sync,
8317 		data->args.lastbytewritten,
8318 		data->args.inode->i_ino);
8319 
8320 	if (!sync) {
8321 		data->inode = nfs_igrab_and_active(data->args.inode);
8322 		if (data->inode == NULL) {
8323 			nfs4_layoutcommit_release(data);
8324 			return -EAGAIN;
8325 		}
8326 		task_setup_data.flags = RPC_TASK_ASYNC;
8327 	}
8328 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8329 	task = rpc_run_task(&task_setup_data);
8330 	if (IS_ERR(task))
8331 		return PTR_ERR(task);
8332 	if (sync)
8333 		status = task->tk_status;
8334 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8335 	dprintk("%s: status %d\n", __func__, status);
8336 	rpc_put_task(task);
8337 	return status;
8338 }
8339 
8340 /**
8341  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8342  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8343  */
8344 static int
8345 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8346 		    struct nfs_fsinfo *info,
8347 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8348 {
8349 	struct nfs41_secinfo_no_name_args args = {
8350 		.style = SECINFO_STYLE_CURRENT_FH,
8351 	};
8352 	struct nfs4_secinfo_res res = {
8353 		.flavors = flavors,
8354 	};
8355 	struct rpc_message msg = {
8356 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8357 		.rpc_argp = &args,
8358 		.rpc_resp = &res,
8359 	};
8360 	struct rpc_clnt *clnt = server->client;
8361 	struct rpc_cred *cred = NULL;
8362 	int status;
8363 
8364 	if (use_integrity) {
8365 		clnt = server->nfs_client->cl_rpcclient;
8366 		cred = nfs4_get_clid_cred(server->nfs_client);
8367 		msg.rpc_cred = cred;
8368 	}
8369 
8370 	dprintk("--> %s\n", __func__);
8371 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8372 				&res.seq_res, 0);
8373 	dprintk("<-- %s status=%d\n", __func__, status);
8374 
8375 	if (cred)
8376 		put_rpccred(cred);
8377 
8378 	return status;
8379 }
8380 
8381 static int
8382 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8383 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8384 {
8385 	struct nfs4_exception exception = { };
8386 	int err;
8387 	do {
8388 		/* first try using integrity protection */
8389 		err = -NFS4ERR_WRONGSEC;
8390 
8391 		/* try to use integrity protection with machine cred */
8392 		if (_nfs4_is_integrity_protected(server->nfs_client))
8393 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8394 							  flavors, true);
8395 
8396 		/*
8397 		 * if unable to use integrity protection, or SECINFO with
8398 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8399 		 * disallowed by spec, but exists in deployed servers) use
8400 		 * the current filesystem's rpc_client and the user cred.
8401 		 */
8402 		if (err == -NFS4ERR_WRONGSEC)
8403 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8404 							  flavors, false);
8405 
8406 		switch (err) {
8407 		case 0:
8408 		case -NFS4ERR_WRONGSEC:
8409 		case -ENOTSUPP:
8410 			goto out;
8411 		default:
8412 			err = nfs4_handle_exception(server, err, &exception);
8413 		}
8414 	} while (exception.retry);
8415 out:
8416 	return err;
8417 }
8418 
8419 static int
8420 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8421 		    struct nfs_fsinfo *info)
8422 {
8423 	int err;
8424 	struct page *page;
8425 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8426 	struct nfs4_secinfo_flavors *flavors;
8427 	struct nfs4_secinfo4 *secinfo;
8428 	int i;
8429 
8430 	page = alloc_page(GFP_KERNEL);
8431 	if (!page) {
8432 		err = -ENOMEM;
8433 		goto out;
8434 	}
8435 
8436 	flavors = page_address(page);
8437 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8438 
8439 	/*
8440 	 * Fall back on "guess and check" method if
8441 	 * the server doesn't support SECINFO_NO_NAME
8442 	 */
8443 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8444 		err = nfs4_find_root_sec(server, fhandle, info);
8445 		goto out_freepage;
8446 	}
8447 	if (err)
8448 		goto out_freepage;
8449 
8450 	for (i = 0; i < flavors->num_flavors; i++) {
8451 		secinfo = &flavors->flavors[i];
8452 
8453 		switch (secinfo->flavor) {
8454 		case RPC_AUTH_NULL:
8455 		case RPC_AUTH_UNIX:
8456 		case RPC_AUTH_GSS:
8457 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8458 					&secinfo->flavor_info);
8459 			break;
8460 		default:
8461 			flavor = RPC_AUTH_MAXFLAVOR;
8462 			break;
8463 		}
8464 
8465 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8466 			flavor = RPC_AUTH_MAXFLAVOR;
8467 
8468 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8469 			err = nfs4_lookup_root_sec(server, fhandle,
8470 						   info, flavor);
8471 			if (!err)
8472 				break;
8473 		}
8474 	}
8475 
8476 	if (flavor == RPC_AUTH_MAXFLAVOR)
8477 		err = -EPERM;
8478 
8479 out_freepage:
8480 	put_page(page);
8481 	if (err == -EACCES)
8482 		return -EPERM;
8483 out:
8484 	return err;
8485 }
8486 
8487 static int _nfs41_test_stateid(struct nfs_server *server,
8488 		nfs4_stateid *stateid,
8489 		struct rpc_cred *cred)
8490 {
8491 	int status;
8492 	struct nfs41_test_stateid_args args = {
8493 		.stateid = stateid,
8494 	};
8495 	struct nfs41_test_stateid_res res;
8496 	struct rpc_message msg = {
8497 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8498 		.rpc_argp = &args,
8499 		.rpc_resp = &res,
8500 		.rpc_cred = cred,
8501 	};
8502 	struct rpc_clnt *rpc_client = server->client;
8503 
8504 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8505 		&rpc_client, &msg);
8506 
8507 	dprintk("NFS call  test_stateid %p\n", stateid);
8508 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8509 	nfs4_set_sequence_privileged(&args.seq_args);
8510 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8511 			&args.seq_args, &res.seq_res);
8512 	if (status != NFS_OK) {
8513 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8514 		return status;
8515 	}
8516 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8517 	return -res.status;
8518 }
8519 
8520 /**
8521  * nfs41_test_stateid - perform a TEST_STATEID operation
8522  *
8523  * @server: server / transport on which to perform the operation
8524  * @stateid: state ID to test
8525  * @cred: credential
8526  *
8527  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8528  * Otherwise a negative NFS4ERR value is returned if the operation
8529  * failed or the state ID is not currently valid.
8530  */
8531 static int nfs41_test_stateid(struct nfs_server *server,
8532 		nfs4_stateid *stateid,
8533 		struct rpc_cred *cred)
8534 {
8535 	struct nfs4_exception exception = { };
8536 	int err;
8537 	do {
8538 		err = _nfs41_test_stateid(server, stateid, cred);
8539 		if (err != -NFS4ERR_DELAY)
8540 			break;
8541 		nfs4_handle_exception(server, err, &exception);
8542 	} while (exception.retry);
8543 	return err;
8544 }
8545 
8546 struct nfs_free_stateid_data {
8547 	struct nfs_server *server;
8548 	struct nfs41_free_stateid_args args;
8549 	struct nfs41_free_stateid_res res;
8550 };
8551 
8552 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8553 {
8554 	struct nfs_free_stateid_data *data = calldata;
8555 	nfs41_setup_sequence(nfs4_get_session(data->server),
8556 			&data->args.seq_args,
8557 			&data->res.seq_res,
8558 			task);
8559 }
8560 
8561 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8562 {
8563 	struct nfs_free_stateid_data *data = calldata;
8564 
8565 	nfs41_sequence_done(task, &data->res.seq_res);
8566 
8567 	switch (task->tk_status) {
8568 	case -NFS4ERR_DELAY:
8569 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8570 			rpc_restart_call_prepare(task);
8571 	}
8572 }
8573 
8574 static void nfs41_free_stateid_release(void *calldata)
8575 {
8576 	kfree(calldata);
8577 }
8578 
8579 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8580 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8581 	.rpc_call_done = nfs41_free_stateid_done,
8582 	.rpc_release = nfs41_free_stateid_release,
8583 };
8584 
8585 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8586 		nfs4_stateid *stateid,
8587 		struct rpc_cred *cred,
8588 		bool privileged)
8589 {
8590 	struct rpc_message msg = {
8591 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8592 		.rpc_cred = cred,
8593 	};
8594 	struct rpc_task_setup task_setup = {
8595 		.rpc_client = server->client,
8596 		.rpc_message = &msg,
8597 		.callback_ops = &nfs41_free_stateid_ops,
8598 		.flags = RPC_TASK_ASYNC,
8599 	};
8600 	struct nfs_free_stateid_data *data;
8601 
8602 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8603 		&task_setup.rpc_client, &msg);
8604 
8605 	dprintk("NFS call  free_stateid %p\n", stateid);
8606 	data = kmalloc(sizeof(*data), GFP_NOFS);
8607 	if (!data)
8608 		return ERR_PTR(-ENOMEM);
8609 	data->server = server;
8610 	nfs4_stateid_copy(&data->args.stateid, stateid);
8611 
8612 	task_setup.callback_data = data;
8613 
8614 	msg.rpc_argp = &data->args;
8615 	msg.rpc_resp = &data->res;
8616 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8617 	if (privileged)
8618 		nfs4_set_sequence_privileged(&data->args.seq_args);
8619 
8620 	return rpc_run_task(&task_setup);
8621 }
8622 
8623 /**
8624  * nfs41_free_stateid - perform a FREE_STATEID operation
8625  *
8626  * @server: server / transport on which to perform the operation
8627  * @stateid: state ID to release
8628  * @cred: credential
8629  *
8630  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8631  * negative NFS4ERR value is returned.
8632  */
8633 static int nfs41_free_stateid(struct nfs_server *server,
8634 		nfs4_stateid *stateid,
8635 		struct rpc_cred *cred)
8636 {
8637 	struct rpc_task *task;
8638 	int ret;
8639 
8640 	task = _nfs41_free_stateid(server, stateid, cred, true);
8641 	if (IS_ERR(task))
8642 		return PTR_ERR(task);
8643 	ret = rpc_wait_for_completion_task(task);
8644 	if (!ret)
8645 		ret = task->tk_status;
8646 	rpc_put_task(task);
8647 	return ret;
8648 }
8649 
8650 static void
8651 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8652 {
8653 	struct rpc_task *task;
8654 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8655 
8656 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8657 	nfs4_free_lock_state(server, lsp);
8658 	if (IS_ERR(task))
8659 		return;
8660 	rpc_put_task(task);
8661 }
8662 
8663 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8664 		const nfs4_stateid *s2)
8665 {
8666 	if (s1->type != s2->type)
8667 		return false;
8668 
8669 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8670 		return false;
8671 
8672 	if (s1->seqid == s2->seqid)
8673 		return true;
8674 	if (s1->seqid == 0 || s2->seqid == 0)
8675 		return true;
8676 
8677 	return false;
8678 }
8679 
8680 #endif /* CONFIG_NFS_V4_1 */
8681 
8682 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8683 		const nfs4_stateid *s2)
8684 {
8685 	return nfs4_stateid_match(s1, s2);
8686 }
8687 
8688 
8689 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8690 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8691 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8692 	.recover_open	= nfs4_open_reclaim,
8693 	.recover_lock	= nfs4_lock_reclaim,
8694 	.establish_clid = nfs4_init_clientid,
8695 	.detect_trunking = nfs40_discover_server_trunking,
8696 };
8697 
8698 #if defined(CONFIG_NFS_V4_1)
8699 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8700 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8701 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8702 	.recover_open	= nfs4_open_reclaim,
8703 	.recover_lock	= nfs4_lock_reclaim,
8704 	.establish_clid = nfs41_init_clientid,
8705 	.reclaim_complete = nfs41_proc_reclaim_complete,
8706 	.detect_trunking = nfs41_discover_server_trunking,
8707 };
8708 #endif /* CONFIG_NFS_V4_1 */
8709 
8710 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8711 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8712 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8713 	.recover_open	= nfs40_open_expired,
8714 	.recover_lock	= nfs4_lock_expired,
8715 	.establish_clid = nfs4_init_clientid,
8716 };
8717 
8718 #if defined(CONFIG_NFS_V4_1)
8719 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8720 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8721 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8722 	.recover_open	= nfs41_open_expired,
8723 	.recover_lock	= nfs41_lock_expired,
8724 	.establish_clid = nfs41_init_clientid,
8725 };
8726 #endif /* CONFIG_NFS_V4_1 */
8727 
8728 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8729 	.sched_state_renewal = nfs4_proc_async_renew,
8730 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8731 	.renew_lease = nfs4_proc_renew,
8732 };
8733 
8734 #if defined(CONFIG_NFS_V4_1)
8735 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8736 	.sched_state_renewal = nfs41_proc_async_sequence,
8737 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8738 	.renew_lease = nfs4_proc_sequence,
8739 };
8740 #endif
8741 
8742 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8743 	.get_locations = _nfs40_proc_get_locations,
8744 	.fsid_present = _nfs40_proc_fsid_present,
8745 };
8746 
8747 #if defined(CONFIG_NFS_V4_1)
8748 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8749 	.get_locations = _nfs41_proc_get_locations,
8750 	.fsid_present = _nfs41_proc_fsid_present,
8751 };
8752 #endif	/* CONFIG_NFS_V4_1 */
8753 
8754 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8755 	.minor_version = 0,
8756 	.init_caps = NFS_CAP_READDIRPLUS
8757 		| NFS_CAP_ATOMIC_OPEN
8758 		| NFS_CAP_POSIX_LOCK,
8759 	.init_client = nfs40_init_client,
8760 	.shutdown_client = nfs40_shutdown_client,
8761 	.match_stateid = nfs4_match_stateid,
8762 	.find_root_sec = nfs4_find_root_sec,
8763 	.free_lock_state = nfs4_release_lockowner,
8764 	.alloc_seqid = nfs_alloc_seqid,
8765 	.call_sync_ops = &nfs40_call_sync_ops,
8766 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8767 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8768 	.state_renewal_ops = &nfs40_state_renewal_ops,
8769 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8770 };
8771 
8772 #if defined(CONFIG_NFS_V4_1)
8773 static struct nfs_seqid *
8774 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8775 {
8776 	return NULL;
8777 }
8778 
8779 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8780 	.minor_version = 1,
8781 	.init_caps = NFS_CAP_READDIRPLUS
8782 		| NFS_CAP_ATOMIC_OPEN
8783 		| NFS_CAP_POSIX_LOCK
8784 		| NFS_CAP_STATEID_NFSV41
8785 		| NFS_CAP_ATOMIC_OPEN_V1,
8786 	.init_client = nfs41_init_client,
8787 	.shutdown_client = nfs41_shutdown_client,
8788 	.match_stateid = nfs41_match_stateid,
8789 	.find_root_sec = nfs41_find_root_sec,
8790 	.free_lock_state = nfs41_free_lock_state,
8791 	.alloc_seqid = nfs_alloc_no_seqid,
8792 	.call_sync_ops = &nfs41_call_sync_ops,
8793 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8794 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8795 	.state_renewal_ops = &nfs41_state_renewal_ops,
8796 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8797 };
8798 #endif
8799 
8800 #if defined(CONFIG_NFS_V4_2)
8801 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8802 	.minor_version = 2,
8803 	.init_caps = NFS_CAP_READDIRPLUS
8804 		| NFS_CAP_ATOMIC_OPEN
8805 		| NFS_CAP_POSIX_LOCK
8806 		| NFS_CAP_STATEID_NFSV41
8807 		| NFS_CAP_ATOMIC_OPEN_V1
8808 		| NFS_CAP_ALLOCATE
8809 		| NFS_CAP_COPY
8810 		| NFS_CAP_DEALLOCATE
8811 		| NFS_CAP_SEEK
8812 		| NFS_CAP_LAYOUTSTATS
8813 		| NFS_CAP_CLONE,
8814 	.init_client = nfs41_init_client,
8815 	.shutdown_client = nfs41_shutdown_client,
8816 	.match_stateid = nfs41_match_stateid,
8817 	.find_root_sec = nfs41_find_root_sec,
8818 	.free_lock_state = nfs41_free_lock_state,
8819 	.call_sync_ops = &nfs41_call_sync_ops,
8820 	.alloc_seqid = nfs_alloc_no_seqid,
8821 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8822 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8823 	.state_renewal_ops = &nfs41_state_renewal_ops,
8824 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8825 };
8826 #endif
8827 
8828 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8829 	[0] = &nfs_v4_0_minor_ops,
8830 #if defined(CONFIG_NFS_V4_1)
8831 	[1] = &nfs_v4_1_minor_ops,
8832 #endif
8833 #if defined(CONFIG_NFS_V4_2)
8834 	[2] = &nfs_v4_2_minor_ops,
8835 #endif
8836 };
8837 
8838 ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8839 {
8840 	ssize_t error, error2;
8841 
8842 	error = generic_listxattr(dentry, list, size);
8843 	if (error < 0)
8844 		return error;
8845 	if (list) {
8846 		list += error;
8847 		size -= error;
8848 	}
8849 
8850 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8851 	if (error2 < 0)
8852 		return error2;
8853 	return error + error2;
8854 }
8855 
8856 static const struct inode_operations nfs4_dir_inode_operations = {
8857 	.create		= nfs_create,
8858 	.lookup		= nfs_lookup,
8859 	.atomic_open	= nfs_atomic_open,
8860 	.link		= nfs_link,
8861 	.unlink		= nfs_unlink,
8862 	.symlink	= nfs_symlink,
8863 	.mkdir		= nfs_mkdir,
8864 	.rmdir		= nfs_rmdir,
8865 	.mknod		= nfs_mknod,
8866 	.rename		= nfs_rename,
8867 	.permission	= nfs_permission,
8868 	.getattr	= nfs_getattr,
8869 	.setattr	= nfs_setattr,
8870 	.getxattr	= generic_getxattr,
8871 	.setxattr	= generic_setxattr,
8872 	.listxattr	= nfs4_listxattr,
8873 	.removexattr	= generic_removexattr,
8874 };
8875 
8876 static const struct inode_operations nfs4_file_inode_operations = {
8877 	.permission	= nfs_permission,
8878 	.getattr	= nfs_getattr,
8879 	.setattr	= nfs_setattr,
8880 	.getxattr	= generic_getxattr,
8881 	.setxattr	= generic_setxattr,
8882 	.listxattr	= nfs4_listxattr,
8883 	.removexattr	= generic_removexattr,
8884 };
8885 
8886 const struct nfs_rpc_ops nfs_v4_clientops = {
8887 	.version	= 4,			/* protocol version */
8888 	.dentry_ops	= &nfs4_dentry_operations,
8889 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8890 	.file_inode_ops	= &nfs4_file_inode_operations,
8891 	.file_ops	= &nfs4_file_operations,
8892 	.getroot	= nfs4_proc_get_root,
8893 	.submount	= nfs4_submount,
8894 	.try_mount	= nfs4_try_mount,
8895 	.getattr	= nfs4_proc_getattr,
8896 	.setattr	= nfs4_proc_setattr,
8897 	.lookup		= nfs4_proc_lookup,
8898 	.access		= nfs4_proc_access,
8899 	.readlink	= nfs4_proc_readlink,
8900 	.create		= nfs4_proc_create,
8901 	.remove		= nfs4_proc_remove,
8902 	.unlink_setup	= nfs4_proc_unlink_setup,
8903 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8904 	.unlink_done	= nfs4_proc_unlink_done,
8905 	.rename_setup	= nfs4_proc_rename_setup,
8906 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8907 	.rename_done	= nfs4_proc_rename_done,
8908 	.link		= nfs4_proc_link,
8909 	.symlink	= nfs4_proc_symlink,
8910 	.mkdir		= nfs4_proc_mkdir,
8911 	.rmdir		= nfs4_proc_remove,
8912 	.readdir	= nfs4_proc_readdir,
8913 	.mknod		= nfs4_proc_mknod,
8914 	.statfs		= nfs4_proc_statfs,
8915 	.fsinfo		= nfs4_proc_fsinfo,
8916 	.pathconf	= nfs4_proc_pathconf,
8917 	.set_capabilities = nfs4_server_capabilities,
8918 	.decode_dirent	= nfs4_decode_dirent,
8919 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8920 	.read_setup	= nfs4_proc_read_setup,
8921 	.read_done	= nfs4_read_done,
8922 	.write_setup	= nfs4_proc_write_setup,
8923 	.write_done	= nfs4_write_done,
8924 	.commit_setup	= nfs4_proc_commit_setup,
8925 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8926 	.commit_done	= nfs4_commit_done,
8927 	.lock		= nfs4_proc_lock,
8928 	.clear_acl_cache = nfs4_zap_acl_attr,
8929 	.close_context  = nfs4_close_context,
8930 	.open_context	= nfs4_atomic_open,
8931 	.have_delegation = nfs4_have_delegation,
8932 	.return_delegation = nfs4_inode_return_delegation,
8933 	.alloc_client	= nfs4_alloc_client,
8934 	.init_client	= nfs4_init_client,
8935 	.free_client	= nfs4_free_client,
8936 	.create_server	= nfs4_create_server,
8937 	.clone_server	= nfs_clone_server,
8938 };
8939 
8940 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8941 	.name	= XATTR_NAME_NFSV4_ACL,
8942 	.list	= nfs4_xattr_list_nfs4_acl,
8943 	.get	= nfs4_xattr_get_nfs4_acl,
8944 	.set	= nfs4_xattr_set_nfs4_acl,
8945 };
8946 
8947 const struct xattr_handler *nfs4_xattr_handlers[] = {
8948 	&nfs4_xattr_nfs4_acl_handler,
8949 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8950 	&nfs4_xattr_nfs4_label_handler,
8951 #endif
8952 	NULL
8953 };
8954 
8955 /*
8956  * Local variables:
8957  *  c-basic-offset: 8
8958  * End:
8959  */
8960