xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 206a81c1)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.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 "nfs4session.h"
67 #include "fscache.h"
68 
69 #include "nfs4trace.h"
70 
71 #define NFSDBG_FACILITY		NFSDBG_PROC
72 
73 #define NFS4_POLL_RETRY_MIN	(HZ/10)
74 #define NFS4_POLL_RETRY_MAX	(15*HZ)
75 
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_ACCESS:
162 		return -EACCES;
163 	case -NFS4ERR_FILE_OPEN:
164 		return -EBUSY;
165 	default:
166 		dprintk("%s could not handle NFSv4 error %d\n",
167 				__func__, -err);
168 		break;
169 	}
170 	return -EIO;
171 }
172 
173 /*
174  * This is our standard bitmap for GETATTR requests.
175  */
176 const u32 nfs4_fattr_bitmap[3] = {
177 	FATTR4_WORD0_TYPE
178 	| FATTR4_WORD0_CHANGE
179 	| FATTR4_WORD0_SIZE
180 	| FATTR4_WORD0_FSID
181 	| FATTR4_WORD0_FILEID,
182 	FATTR4_WORD1_MODE
183 	| FATTR4_WORD1_NUMLINKS
184 	| FATTR4_WORD1_OWNER
185 	| FATTR4_WORD1_OWNER_GROUP
186 	| FATTR4_WORD1_RAWDEV
187 	| FATTR4_WORD1_SPACE_USED
188 	| FATTR4_WORD1_TIME_ACCESS
189 	| FATTR4_WORD1_TIME_METADATA
190 	| FATTR4_WORD1_TIME_MODIFY,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192 	FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195 
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197 	FATTR4_WORD0_TYPE
198 	| FATTR4_WORD0_CHANGE
199 	| FATTR4_WORD0_SIZE
200 	| FATTR4_WORD0_FSID
201 	| FATTR4_WORD0_FILEID,
202 	FATTR4_WORD1_MODE
203 	| FATTR4_WORD1_NUMLINKS
204 	| FATTR4_WORD1_OWNER
205 	| FATTR4_WORD1_OWNER_GROUP
206 	| FATTR4_WORD1_RAWDEV
207 	| FATTR4_WORD1_SPACE_USED
208 	| FATTR4_WORD1_TIME_ACCESS
209 	| FATTR4_WORD1_TIME_METADATA
210 	| FATTR4_WORD1_TIME_MODIFY,
211 	FATTR4_WORD2_MDSTHRESHOLD
212 };
213 
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215 	FATTR4_WORD0_TYPE
216 	| FATTR4_WORD0_CHANGE
217 	| FATTR4_WORD0_FILEID,
218 };
219 
220 const u32 nfs4_statfs_bitmap[3] = {
221 	FATTR4_WORD0_FILES_AVAIL
222 	| FATTR4_WORD0_FILES_FREE
223 	| FATTR4_WORD0_FILES_TOTAL,
224 	FATTR4_WORD1_SPACE_AVAIL
225 	| FATTR4_WORD1_SPACE_FREE
226 	| FATTR4_WORD1_SPACE_TOTAL
227 };
228 
229 const u32 nfs4_pathconf_bitmap[3] = {
230 	FATTR4_WORD0_MAXLINK
231 	| FATTR4_WORD0_MAXNAME,
232 	0
233 };
234 
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236 			| FATTR4_WORD0_MAXREAD
237 			| FATTR4_WORD0_MAXWRITE
238 			| FATTR4_WORD0_LEASE_TIME,
239 			FATTR4_WORD1_TIME_DELTA
240 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
241 			FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243 
244 const u32 nfs4_fs_locations_bitmap[3] = {
245 	FATTR4_WORD0_TYPE
246 	| FATTR4_WORD0_CHANGE
247 	| FATTR4_WORD0_SIZE
248 	| FATTR4_WORD0_FSID
249 	| FATTR4_WORD0_FILEID
250 	| FATTR4_WORD0_FS_LOCATIONS,
251 	FATTR4_WORD1_MODE
252 	| FATTR4_WORD1_NUMLINKS
253 	| FATTR4_WORD1_OWNER
254 	| FATTR4_WORD1_OWNER_GROUP
255 	| FATTR4_WORD1_RAWDEV
256 	| FATTR4_WORD1_SPACE_USED
257 	| FATTR4_WORD1_TIME_ACCESS
258 	| FATTR4_WORD1_TIME_METADATA
259 	| FATTR4_WORD1_TIME_MODIFY
260 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262 
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264 		struct nfs4_readdir_arg *readdir)
265 {
266 	__be32 *start, *p;
267 
268 	if (cookie > 2) {
269 		readdir->cookie = cookie;
270 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271 		return;
272 	}
273 
274 	readdir->cookie = 0;
275 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276 	if (cookie == 2)
277 		return;
278 
279 	/*
280 	 * NFSv4 servers do not return entries for '.' and '..'
281 	 * Therefore, we fake these entries here.  We let '.'
282 	 * have cookie 0 and '..' have cookie 1.  Note that
283 	 * when talking to the server, we always send cookie 0
284 	 * instead of 1 or 2.
285 	 */
286 	start = p = kmap_atomic(*readdir->pages);
287 
288 	if (cookie == 0) {
289 		*p++ = xdr_one;                                  /* next */
290 		*p++ = xdr_zero;                   /* cookie, first word */
291 		*p++ = xdr_one;                   /* cookie, second word */
292 		*p++ = xdr_one;                             /* entry len */
293 		memcpy(p, ".\0\0\0", 4);                        /* entry */
294 		p++;
295 		*p++ = xdr_one;                         /* bitmap length */
296 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297 		*p++ = htonl(8);              /* attribute buffer length */
298 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
299 	}
300 
301 	*p++ = xdr_one;                                  /* next */
302 	*p++ = xdr_zero;                   /* cookie, first word */
303 	*p++ = xdr_two;                   /* cookie, second word */
304 	*p++ = xdr_two;                             /* entry len */
305 	memcpy(p, "..\0\0", 4);                         /* entry */
306 	p++;
307 	*p++ = xdr_one;                         /* bitmap length */
308 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309 	*p++ = htonl(8);              /* attribute buffer length */
310 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
311 
312 	readdir->pgbase = (char *)p - (char *)start;
313 	readdir->count -= readdir->pgbase;
314 	kunmap_atomic(start);
315 }
316 
317 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
318 {
319 	int res = 0;
320 
321 	might_sleep();
322 
323 	if (*timeout <= 0)
324 		*timeout = NFS4_POLL_RETRY_MIN;
325 	if (*timeout > NFS4_POLL_RETRY_MAX)
326 		*timeout = NFS4_POLL_RETRY_MAX;
327 	freezable_schedule_timeout_killable_unsafe(*timeout);
328 	if (fatal_signal_pending(current))
329 		res = -ERESTARTSYS;
330 	*timeout <<= 1;
331 	return res;
332 }
333 
334 /* This is the error handling routine for processes that are allowed
335  * to sleep.
336  */
337 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
338 {
339 	struct nfs_client *clp = server->nfs_client;
340 	struct nfs4_state *state = exception->state;
341 	struct inode *inode = exception->inode;
342 	int ret = errorcode;
343 
344 	exception->retry = 0;
345 	switch(errorcode) {
346 		case 0:
347 			return 0;
348 		case -NFS4ERR_OPENMODE:
349 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
350 				nfs4_inode_return_delegation(inode);
351 				exception->retry = 1;
352 				return 0;
353 			}
354 			if (state == NULL)
355 				break;
356 			ret = nfs4_schedule_stateid_recovery(server, state);
357 			if (ret < 0)
358 				break;
359 			goto wait_on_recovery;
360 		case -NFS4ERR_DELEG_REVOKED:
361 		case -NFS4ERR_ADMIN_REVOKED:
362 		case -NFS4ERR_BAD_STATEID:
363 			if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
364 				nfs_remove_bad_delegation(inode);
365 				exception->retry = 1;
366 				break;
367 			}
368 			if (state == NULL)
369 				break;
370 			ret = nfs4_schedule_stateid_recovery(server, state);
371 			if (ret < 0)
372 				break;
373 			goto wait_on_recovery;
374 		case -NFS4ERR_EXPIRED:
375 			if (state != NULL) {
376 				ret = nfs4_schedule_stateid_recovery(server, state);
377 				if (ret < 0)
378 					break;
379 			}
380 		case -NFS4ERR_STALE_STATEID:
381 		case -NFS4ERR_STALE_CLIENTID:
382 			nfs4_schedule_lease_recovery(clp);
383 			goto wait_on_recovery;
384 		case -NFS4ERR_MOVED:
385 			ret = nfs4_schedule_migration_recovery(server);
386 			if (ret < 0)
387 				break;
388 			goto wait_on_recovery;
389 		case -NFS4ERR_LEASE_MOVED:
390 			nfs4_schedule_lease_moved_recovery(clp);
391 			goto wait_on_recovery;
392 #if defined(CONFIG_NFS_V4_1)
393 		case -NFS4ERR_BADSESSION:
394 		case -NFS4ERR_BADSLOT:
395 		case -NFS4ERR_BAD_HIGH_SLOT:
396 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
397 		case -NFS4ERR_DEADSESSION:
398 		case -NFS4ERR_SEQ_FALSE_RETRY:
399 		case -NFS4ERR_SEQ_MISORDERED:
400 			dprintk("%s ERROR: %d Reset session\n", __func__,
401 				errorcode);
402 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
403 			goto wait_on_recovery;
404 #endif /* defined(CONFIG_NFS_V4_1) */
405 		case -NFS4ERR_FILE_OPEN:
406 			if (exception->timeout > HZ) {
407 				/* We have retried a decent amount, time to
408 				 * fail
409 				 */
410 				ret = -EBUSY;
411 				break;
412 			}
413 		case -NFS4ERR_GRACE:
414 		case -NFS4ERR_DELAY:
415 			ret = nfs4_delay(server->client, &exception->timeout);
416 			if (ret != 0)
417 				break;
418 		case -NFS4ERR_RETRY_UNCACHED_REP:
419 		case -NFS4ERR_OLD_STATEID:
420 			exception->retry = 1;
421 			break;
422 		case -NFS4ERR_BADOWNER:
423 			/* The following works around a Linux server bug! */
424 		case -NFS4ERR_BADNAME:
425 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
426 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
427 				exception->retry = 1;
428 				printk(KERN_WARNING "NFS: v4 server %s "
429 						"does not accept raw "
430 						"uid/gids. "
431 						"Reenabling the idmapper.\n",
432 						server->nfs_client->cl_hostname);
433 			}
434 	}
435 	/* We failed to handle the error */
436 	return nfs4_map_errors(ret);
437 wait_on_recovery:
438 	ret = nfs4_wait_clnt_recover(clp);
439 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
440 		return -EIO;
441 	if (ret == 0)
442 		exception->retry = 1;
443 	return ret;
444 }
445 
446 /*
447  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
448  * or 'false' otherwise.
449  */
450 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
451 {
452 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
453 
454 	if (flavor == RPC_AUTH_GSS_KRB5I ||
455 	    flavor == RPC_AUTH_GSS_KRB5P)
456 		return true;
457 
458 	return false;
459 }
460 
461 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
462 {
463 	spin_lock(&clp->cl_lock);
464 	if (time_before(clp->cl_last_renewal,timestamp))
465 		clp->cl_last_renewal = timestamp;
466 	spin_unlock(&clp->cl_lock);
467 }
468 
469 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
470 {
471 	do_renew_lease(server->nfs_client, timestamp);
472 }
473 
474 struct nfs4_call_sync_data {
475 	const struct nfs_server *seq_server;
476 	struct nfs4_sequence_args *seq_args;
477 	struct nfs4_sequence_res *seq_res;
478 };
479 
480 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
481 			       struct nfs4_sequence_res *res, int cache_reply)
482 {
483 	args->sa_slot = NULL;
484 	args->sa_cache_this = cache_reply;
485 	args->sa_privileged = 0;
486 
487 	res->sr_slot = NULL;
488 }
489 
490 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
491 {
492 	args->sa_privileged = 1;
493 }
494 
495 static int nfs40_setup_sequence(const struct nfs_server *server,
496 				struct nfs4_sequence_args *args,
497 				struct nfs4_sequence_res *res,
498 				struct rpc_task *task)
499 {
500 	struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
501 	struct nfs4_slot *slot;
502 
503 	/* slot already allocated? */
504 	if (res->sr_slot != NULL)
505 		goto out_start;
506 
507 	spin_lock(&tbl->slot_tbl_lock);
508 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
509 		goto out_sleep;
510 
511 	slot = nfs4_alloc_slot(tbl);
512 	if (IS_ERR(slot)) {
513 		if (slot == ERR_PTR(-ENOMEM))
514 			task->tk_timeout = HZ >> 2;
515 		goto out_sleep;
516 	}
517 	spin_unlock(&tbl->slot_tbl_lock);
518 
519 	args->sa_slot = slot;
520 	res->sr_slot = slot;
521 
522 out_start:
523 	rpc_call_start(task);
524 	return 0;
525 
526 out_sleep:
527 	if (args->sa_privileged)
528 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
529 				NULL, RPC_PRIORITY_PRIVILEGED);
530 	else
531 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
532 	spin_unlock(&tbl->slot_tbl_lock);
533 	return -EAGAIN;
534 }
535 
536 static int nfs40_sequence_done(struct rpc_task *task,
537 			       struct nfs4_sequence_res *res)
538 {
539 	struct nfs4_slot *slot = res->sr_slot;
540 	struct nfs4_slot_table *tbl;
541 
542 	if (slot == NULL)
543 		goto out;
544 
545 	tbl = slot->table;
546 	spin_lock(&tbl->slot_tbl_lock);
547 	if (!nfs41_wake_and_assign_slot(tbl, slot))
548 		nfs4_free_slot(tbl, slot);
549 	spin_unlock(&tbl->slot_tbl_lock);
550 
551 	res->sr_slot = NULL;
552 out:
553 	return 1;
554 }
555 
556 #if defined(CONFIG_NFS_V4_1)
557 
558 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
559 {
560 	struct nfs4_session *session;
561 	struct nfs4_slot_table *tbl;
562 	struct nfs4_slot *slot = res->sr_slot;
563 	bool send_new_highest_used_slotid = false;
564 
565 	tbl = slot->table;
566 	session = tbl->session;
567 
568 	spin_lock(&tbl->slot_tbl_lock);
569 	/* Be nice to the server: try to ensure that the last transmitted
570 	 * value for highest_user_slotid <= target_highest_slotid
571 	 */
572 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
573 		send_new_highest_used_slotid = true;
574 
575 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
576 		send_new_highest_used_slotid = false;
577 		goto out_unlock;
578 	}
579 	nfs4_free_slot(tbl, slot);
580 
581 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
582 		send_new_highest_used_slotid = false;
583 out_unlock:
584 	spin_unlock(&tbl->slot_tbl_lock);
585 	res->sr_slot = NULL;
586 	if (send_new_highest_used_slotid)
587 		nfs41_server_notify_highest_slotid_update(session->clp);
588 }
589 
590 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
591 {
592 	struct nfs4_session *session;
593 	struct nfs4_slot *slot = res->sr_slot;
594 	struct nfs_client *clp;
595 	bool interrupted = false;
596 	int ret = 1;
597 
598 	if (slot == NULL)
599 		goto out_noaction;
600 	/* don't increment the sequence number if the task wasn't sent */
601 	if (!RPC_WAS_SENT(task))
602 		goto out;
603 
604 	session = slot->table->session;
605 
606 	if (slot->interrupted) {
607 		slot->interrupted = 0;
608 		interrupted = true;
609 	}
610 
611 	trace_nfs4_sequence_done(session, res);
612 	/* Check the SEQUENCE operation status */
613 	switch (res->sr_status) {
614 	case 0:
615 		/* Update the slot's sequence and clientid lease timer */
616 		++slot->seq_nr;
617 		clp = session->clp;
618 		do_renew_lease(clp, res->sr_timestamp);
619 		/* Check sequence flags */
620 		if (res->sr_status_flags != 0)
621 			nfs4_schedule_lease_recovery(clp);
622 		nfs41_update_target_slotid(slot->table, slot, res);
623 		break;
624 	case 1:
625 		/*
626 		 * sr_status remains 1 if an RPC level error occurred.
627 		 * The server may or may not have processed the sequence
628 		 * operation..
629 		 * Mark the slot as having hosted an interrupted RPC call.
630 		 */
631 		slot->interrupted = 1;
632 		goto out;
633 	case -NFS4ERR_DELAY:
634 		/* The server detected a resend of the RPC call and
635 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
636 		 * of RFC5661.
637 		 */
638 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
639 			__func__,
640 			slot->slot_nr,
641 			slot->seq_nr);
642 		goto out_retry;
643 	case -NFS4ERR_BADSLOT:
644 		/*
645 		 * The slot id we used was probably retired. Try again
646 		 * using a different slot id.
647 		 */
648 		goto retry_nowait;
649 	case -NFS4ERR_SEQ_MISORDERED:
650 		/*
651 		 * Was the last operation on this sequence interrupted?
652 		 * If so, retry after bumping the sequence number.
653 		 */
654 		if (interrupted) {
655 			++slot->seq_nr;
656 			goto retry_nowait;
657 		}
658 		/*
659 		 * Could this slot have been previously retired?
660 		 * If so, then the server may be expecting seq_nr = 1!
661 		 */
662 		if (slot->seq_nr != 1) {
663 			slot->seq_nr = 1;
664 			goto retry_nowait;
665 		}
666 		break;
667 	case -NFS4ERR_SEQ_FALSE_RETRY:
668 		++slot->seq_nr;
669 		goto retry_nowait;
670 	default:
671 		/* Just update the slot sequence no. */
672 		++slot->seq_nr;
673 	}
674 out:
675 	/* The session may be reset by one of the error handlers. */
676 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
677 	nfs41_sequence_free_slot(res);
678 out_noaction:
679 	return ret;
680 retry_nowait:
681 	if (rpc_restart_call_prepare(task)) {
682 		task->tk_status = 0;
683 		ret = 0;
684 	}
685 	goto out;
686 out_retry:
687 	if (!rpc_restart_call(task))
688 		goto out;
689 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
690 	return 0;
691 }
692 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
693 
694 static int nfs4_sequence_done(struct rpc_task *task,
695 			       struct nfs4_sequence_res *res)
696 {
697 	if (res->sr_slot == NULL)
698 		return 1;
699 	if (!res->sr_slot->table->session)
700 		return nfs40_sequence_done(task, res);
701 	return nfs41_sequence_done(task, res);
702 }
703 
704 int nfs41_setup_sequence(struct nfs4_session *session,
705 				struct nfs4_sequence_args *args,
706 				struct nfs4_sequence_res *res,
707 				struct rpc_task *task)
708 {
709 	struct nfs4_slot *slot;
710 	struct nfs4_slot_table *tbl;
711 
712 	dprintk("--> %s\n", __func__);
713 	/* slot already allocated? */
714 	if (res->sr_slot != NULL)
715 		goto out_success;
716 
717 	tbl = &session->fc_slot_table;
718 
719 	task->tk_timeout = 0;
720 
721 	spin_lock(&tbl->slot_tbl_lock);
722 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
723 	    !args->sa_privileged) {
724 		/* The state manager will wait until the slot table is empty */
725 		dprintk("%s session is draining\n", __func__);
726 		goto out_sleep;
727 	}
728 
729 	slot = nfs4_alloc_slot(tbl);
730 	if (IS_ERR(slot)) {
731 		/* If out of memory, try again in 1/4 second */
732 		if (slot == ERR_PTR(-ENOMEM))
733 			task->tk_timeout = HZ >> 2;
734 		dprintk("<-- %s: no free slots\n", __func__);
735 		goto out_sleep;
736 	}
737 	spin_unlock(&tbl->slot_tbl_lock);
738 
739 	args->sa_slot = slot;
740 
741 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
742 			slot->slot_nr, slot->seq_nr);
743 
744 	res->sr_slot = slot;
745 	res->sr_timestamp = jiffies;
746 	res->sr_status_flags = 0;
747 	/*
748 	 * sr_status is only set in decode_sequence, and so will remain
749 	 * set to 1 if an rpc level failure occurs.
750 	 */
751 	res->sr_status = 1;
752 	trace_nfs4_setup_sequence(session, args);
753 out_success:
754 	rpc_call_start(task);
755 	return 0;
756 out_sleep:
757 	/* Privileged tasks are queued with top priority */
758 	if (args->sa_privileged)
759 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
760 				NULL, RPC_PRIORITY_PRIVILEGED);
761 	else
762 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
763 	spin_unlock(&tbl->slot_tbl_lock);
764 	return -EAGAIN;
765 }
766 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
767 
768 static int nfs4_setup_sequence(const struct nfs_server *server,
769 			       struct nfs4_sequence_args *args,
770 			       struct nfs4_sequence_res *res,
771 			       struct rpc_task *task)
772 {
773 	struct nfs4_session *session = nfs4_get_session(server);
774 	int ret = 0;
775 
776 	if (!session)
777 		return nfs40_setup_sequence(server, args, res, task);
778 
779 	dprintk("--> %s clp %p session %p sr_slot %u\n",
780 		__func__, session->clp, session, res->sr_slot ?
781 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
782 
783 	ret = nfs41_setup_sequence(session, args, res, task);
784 
785 	dprintk("<-- %s status=%d\n", __func__, ret);
786 	return ret;
787 }
788 
789 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
790 {
791 	struct nfs4_call_sync_data *data = calldata;
792 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
793 
794 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
795 
796 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
797 }
798 
799 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
800 {
801 	struct nfs4_call_sync_data *data = calldata;
802 
803 	nfs41_sequence_done(task, data->seq_res);
804 }
805 
806 static const struct rpc_call_ops nfs41_call_sync_ops = {
807 	.rpc_call_prepare = nfs41_call_sync_prepare,
808 	.rpc_call_done = nfs41_call_sync_done,
809 };
810 
811 #else	/* !CONFIG_NFS_V4_1 */
812 
813 static int nfs4_setup_sequence(const struct nfs_server *server,
814 			       struct nfs4_sequence_args *args,
815 			       struct nfs4_sequence_res *res,
816 			       struct rpc_task *task)
817 {
818 	return nfs40_setup_sequence(server, args, res, task);
819 }
820 
821 static int nfs4_sequence_done(struct rpc_task *task,
822 			       struct nfs4_sequence_res *res)
823 {
824 	return nfs40_sequence_done(task, res);
825 }
826 
827 #endif	/* !CONFIG_NFS_V4_1 */
828 
829 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
830 {
831 	struct nfs4_call_sync_data *data = calldata;
832 	nfs4_setup_sequence(data->seq_server,
833 				data->seq_args, data->seq_res, task);
834 }
835 
836 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
837 {
838 	struct nfs4_call_sync_data *data = calldata;
839 	nfs4_sequence_done(task, data->seq_res);
840 }
841 
842 static const struct rpc_call_ops nfs40_call_sync_ops = {
843 	.rpc_call_prepare = nfs40_call_sync_prepare,
844 	.rpc_call_done = nfs40_call_sync_done,
845 };
846 
847 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
848 				   struct nfs_server *server,
849 				   struct rpc_message *msg,
850 				   struct nfs4_sequence_args *args,
851 				   struct nfs4_sequence_res *res)
852 {
853 	int ret;
854 	struct rpc_task *task;
855 	struct nfs_client *clp = server->nfs_client;
856 	struct nfs4_call_sync_data data = {
857 		.seq_server = server,
858 		.seq_args = args,
859 		.seq_res = res,
860 	};
861 	struct rpc_task_setup task_setup = {
862 		.rpc_client = clnt,
863 		.rpc_message = msg,
864 		.callback_ops = clp->cl_mvops->call_sync_ops,
865 		.callback_data = &data
866 	};
867 
868 	task = rpc_run_task(&task_setup);
869 	if (IS_ERR(task))
870 		ret = PTR_ERR(task);
871 	else {
872 		ret = task->tk_status;
873 		rpc_put_task(task);
874 	}
875 	return ret;
876 }
877 
878 static
879 int nfs4_call_sync(struct rpc_clnt *clnt,
880 		   struct nfs_server *server,
881 		   struct rpc_message *msg,
882 		   struct nfs4_sequence_args *args,
883 		   struct nfs4_sequence_res *res,
884 		   int cache_reply)
885 {
886 	nfs4_init_sequence(args, res, cache_reply);
887 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
888 }
889 
890 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
891 {
892 	struct nfs_inode *nfsi = NFS_I(dir);
893 
894 	spin_lock(&dir->i_lock);
895 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
896 	if (!cinfo->atomic || cinfo->before != dir->i_version)
897 		nfs_force_lookup_revalidate(dir);
898 	dir->i_version = cinfo->after;
899 	nfs_fscache_invalidate(dir);
900 	spin_unlock(&dir->i_lock);
901 }
902 
903 struct nfs4_opendata {
904 	struct kref kref;
905 	struct nfs_openargs o_arg;
906 	struct nfs_openres o_res;
907 	struct nfs_open_confirmargs c_arg;
908 	struct nfs_open_confirmres c_res;
909 	struct nfs4_string owner_name;
910 	struct nfs4_string group_name;
911 	struct nfs_fattr f_attr;
912 	struct nfs4_label *f_label;
913 	struct dentry *dir;
914 	struct dentry *dentry;
915 	struct nfs4_state_owner *owner;
916 	struct nfs4_state *state;
917 	struct iattr attrs;
918 	unsigned long timestamp;
919 	unsigned int rpc_done : 1;
920 	unsigned int file_created : 1;
921 	unsigned int is_recover : 1;
922 	int rpc_status;
923 	int cancelled;
924 };
925 
926 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
927 		int err, struct nfs4_exception *exception)
928 {
929 	if (err != -EINVAL)
930 		return false;
931 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
932 		return false;
933 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
934 	exception->retry = 1;
935 	return true;
936 }
937 
938 static enum open_claim_type4
939 nfs4_map_atomic_open_claim(struct nfs_server *server,
940 		enum open_claim_type4 claim)
941 {
942 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
943 		return claim;
944 	switch (claim) {
945 	default:
946 		return claim;
947 	case NFS4_OPEN_CLAIM_FH:
948 		return NFS4_OPEN_CLAIM_NULL;
949 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
950 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
951 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
952 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
953 	}
954 }
955 
956 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
957 {
958 	p->o_res.f_attr = &p->f_attr;
959 	p->o_res.f_label = p->f_label;
960 	p->o_res.seqid = p->o_arg.seqid;
961 	p->c_res.seqid = p->c_arg.seqid;
962 	p->o_res.server = p->o_arg.server;
963 	p->o_res.access_request = p->o_arg.access;
964 	nfs_fattr_init(&p->f_attr);
965 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
966 }
967 
968 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
969 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
970 		const struct iattr *attrs,
971 		struct nfs4_label *label,
972 		enum open_claim_type4 claim,
973 		gfp_t gfp_mask)
974 {
975 	struct dentry *parent = dget_parent(dentry);
976 	struct inode *dir = parent->d_inode;
977 	struct nfs_server *server = NFS_SERVER(dir);
978 	struct nfs4_opendata *p;
979 
980 	p = kzalloc(sizeof(*p), gfp_mask);
981 	if (p == NULL)
982 		goto err;
983 
984 	p->f_label = nfs4_label_alloc(server, gfp_mask);
985 	if (IS_ERR(p->f_label))
986 		goto err_free_p;
987 
988 	p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
989 	if (p->o_arg.seqid == NULL)
990 		goto err_free_label;
991 	nfs_sb_active(dentry->d_sb);
992 	p->dentry = dget(dentry);
993 	p->dir = parent;
994 	p->owner = sp;
995 	atomic_inc(&sp->so_count);
996 	p->o_arg.open_flags = flags;
997 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
998 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
999 	 * will return permission denied for all bits until close */
1000 	if (!(flags & O_EXCL)) {
1001 		/* ask server to check for all possible rights as results
1002 		 * are cached */
1003 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1004 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1005 	}
1006 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1007 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1008 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1009 	p->o_arg.name = &dentry->d_name;
1010 	p->o_arg.server = server;
1011 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1012 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1013 	p->o_arg.label = label;
1014 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1015 	switch (p->o_arg.claim) {
1016 	case NFS4_OPEN_CLAIM_NULL:
1017 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1018 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1019 		p->o_arg.fh = NFS_FH(dir);
1020 		break;
1021 	case NFS4_OPEN_CLAIM_PREVIOUS:
1022 	case NFS4_OPEN_CLAIM_FH:
1023 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1024 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1025 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1026 	}
1027 	if (attrs != NULL && attrs->ia_valid != 0) {
1028 		__u32 verf[2];
1029 
1030 		p->o_arg.u.attrs = &p->attrs;
1031 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1032 
1033 		verf[0] = jiffies;
1034 		verf[1] = current->pid;
1035 		memcpy(p->o_arg.u.verifier.data, verf,
1036 				sizeof(p->o_arg.u.verifier.data));
1037 	}
1038 	p->c_arg.fh = &p->o_res.fh;
1039 	p->c_arg.stateid = &p->o_res.stateid;
1040 	p->c_arg.seqid = p->o_arg.seqid;
1041 	nfs4_init_opendata_res(p);
1042 	kref_init(&p->kref);
1043 	return p;
1044 
1045 err_free_label:
1046 	nfs4_label_free(p->f_label);
1047 err_free_p:
1048 	kfree(p);
1049 err:
1050 	dput(parent);
1051 	return NULL;
1052 }
1053 
1054 static void nfs4_opendata_free(struct kref *kref)
1055 {
1056 	struct nfs4_opendata *p = container_of(kref,
1057 			struct nfs4_opendata, kref);
1058 	struct super_block *sb = p->dentry->d_sb;
1059 
1060 	nfs_free_seqid(p->o_arg.seqid);
1061 	if (p->state != NULL)
1062 		nfs4_put_open_state(p->state);
1063 	nfs4_put_state_owner(p->owner);
1064 
1065 	nfs4_label_free(p->f_label);
1066 
1067 	dput(p->dir);
1068 	dput(p->dentry);
1069 	nfs_sb_deactive(sb);
1070 	nfs_fattr_free_names(&p->f_attr);
1071 	kfree(p->f_attr.mdsthreshold);
1072 	kfree(p);
1073 }
1074 
1075 static void nfs4_opendata_put(struct nfs4_opendata *p)
1076 {
1077 	if (p != NULL)
1078 		kref_put(&p->kref, nfs4_opendata_free);
1079 }
1080 
1081 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1082 {
1083 	int ret;
1084 
1085 	ret = rpc_wait_for_completion_task(task);
1086 	return ret;
1087 }
1088 
1089 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1090 {
1091 	int ret = 0;
1092 
1093 	if (open_mode & (O_EXCL|O_TRUNC))
1094 		goto out;
1095 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1096 		case FMODE_READ:
1097 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1098 				&& state->n_rdonly != 0;
1099 			break;
1100 		case FMODE_WRITE:
1101 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1102 				&& state->n_wronly != 0;
1103 			break;
1104 		case FMODE_READ|FMODE_WRITE:
1105 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1106 				&& state->n_rdwr != 0;
1107 	}
1108 out:
1109 	return ret;
1110 }
1111 
1112 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1113 {
1114 	if (delegation == NULL)
1115 		return 0;
1116 	if ((delegation->type & fmode) != fmode)
1117 		return 0;
1118 	if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1119 		return 0;
1120 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1121 		return 0;
1122 	nfs_mark_delegation_referenced(delegation);
1123 	return 1;
1124 }
1125 
1126 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1127 {
1128 	switch (fmode) {
1129 		case FMODE_WRITE:
1130 			state->n_wronly++;
1131 			break;
1132 		case FMODE_READ:
1133 			state->n_rdonly++;
1134 			break;
1135 		case FMODE_READ|FMODE_WRITE:
1136 			state->n_rdwr++;
1137 	}
1138 	nfs4_state_set_mode_locked(state, state->state | fmode);
1139 }
1140 
1141 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1142 {
1143 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1144 	bool need_recover = false;
1145 
1146 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1147 		need_recover = true;
1148 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1149 		need_recover = true;
1150 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1151 		need_recover = true;
1152 	if (need_recover)
1153 		nfs4_state_mark_reclaim_nograce(clp, state);
1154 }
1155 
1156 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1157 		nfs4_stateid *stateid)
1158 {
1159 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1160 		return true;
1161 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1162 		nfs_test_and_clear_all_open_stateid(state);
1163 		return true;
1164 	}
1165 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1166 		return true;
1167 	return false;
1168 }
1169 
1170 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1171 		nfs4_stateid *stateid, fmode_t fmode)
1172 {
1173 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1174 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1175 	case FMODE_WRITE:
1176 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1177 		break;
1178 	case FMODE_READ:
1179 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1180 		break;
1181 	case 0:
1182 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1183 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1184 		clear_bit(NFS_OPEN_STATE, &state->flags);
1185 	}
1186 	if (stateid == NULL)
1187 		return;
1188 	if (!nfs_need_update_open_stateid(state, stateid))
1189 		return;
1190 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1191 		nfs4_stateid_copy(&state->stateid, stateid);
1192 	nfs4_stateid_copy(&state->open_stateid, stateid);
1193 }
1194 
1195 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1196 {
1197 	write_seqlock(&state->seqlock);
1198 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1199 	write_sequnlock(&state->seqlock);
1200 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1201 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1202 }
1203 
1204 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1205 {
1206 	switch (fmode) {
1207 		case FMODE_READ:
1208 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1209 			break;
1210 		case FMODE_WRITE:
1211 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1212 			break;
1213 		case FMODE_READ|FMODE_WRITE:
1214 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1215 	}
1216 	if (!nfs_need_update_open_stateid(state, stateid))
1217 		return;
1218 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1219 		nfs4_stateid_copy(&state->stateid, stateid);
1220 	nfs4_stateid_copy(&state->open_stateid, stateid);
1221 }
1222 
1223 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1224 {
1225 	/*
1226 	 * Protect the call to nfs4_state_set_mode_locked and
1227 	 * serialise the stateid update
1228 	 */
1229 	write_seqlock(&state->seqlock);
1230 	if (deleg_stateid != NULL) {
1231 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1232 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1233 	}
1234 	if (open_stateid != NULL)
1235 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1236 	write_sequnlock(&state->seqlock);
1237 	spin_lock(&state->owner->so_lock);
1238 	update_open_stateflags(state, fmode);
1239 	spin_unlock(&state->owner->so_lock);
1240 }
1241 
1242 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1243 {
1244 	struct nfs_inode *nfsi = NFS_I(state->inode);
1245 	struct nfs_delegation *deleg_cur;
1246 	int ret = 0;
1247 
1248 	fmode &= (FMODE_READ|FMODE_WRITE);
1249 
1250 	rcu_read_lock();
1251 	deleg_cur = rcu_dereference(nfsi->delegation);
1252 	if (deleg_cur == NULL)
1253 		goto no_delegation;
1254 
1255 	spin_lock(&deleg_cur->lock);
1256 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1257 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1258 	    (deleg_cur->type & fmode) != fmode)
1259 		goto no_delegation_unlock;
1260 
1261 	if (delegation == NULL)
1262 		delegation = &deleg_cur->stateid;
1263 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1264 		goto no_delegation_unlock;
1265 
1266 	nfs_mark_delegation_referenced(deleg_cur);
1267 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1268 	ret = 1;
1269 no_delegation_unlock:
1270 	spin_unlock(&deleg_cur->lock);
1271 no_delegation:
1272 	rcu_read_unlock();
1273 
1274 	if (!ret && open_stateid != NULL) {
1275 		__update_open_stateid(state, open_stateid, NULL, fmode);
1276 		ret = 1;
1277 	}
1278 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1279 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1280 
1281 	return ret;
1282 }
1283 
1284 
1285 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1286 {
1287 	struct nfs_delegation *delegation;
1288 
1289 	rcu_read_lock();
1290 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1291 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1292 		rcu_read_unlock();
1293 		return;
1294 	}
1295 	rcu_read_unlock();
1296 	nfs4_inode_return_delegation(inode);
1297 }
1298 
1299 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1300 {
1301 	struct nfs4_state *state = opendata->state;
1302 	struct nfs_inode *nfsi = NFS_I(state->inode);
1303 	struct nfs_delegation *delegation;
1304 	int open_mode = opendata->o_arg.open_flags;
1305 	fmode_t fmode = opendata->o_arg.fmode;
1306 	nfs4_stateid stateid;
1307 	int ret = -EAGAIN;
1308 
1309 	for (;;) {
1310 		if (can_open_cached(state, fmode, open_mode)) {
1311 			spin_lock(&state->owner->so_lock);
1312 			if (can_open_cached(state, fmode, open_mode)) {
1313 				update_open_stateflags(state, fmode);
1314 				spin_unlock(&state->owner->so_lock);
1315 				goto out_return_state;
1316 			}
1317 			spin_unlock(&state->owner->so_lock);
1318 		}
1319 		rcu_read_lock();
1320 		delegation = rcu_dereference(nfsi->delegation);
1321 		if (!can_open_delegated(delegation, fmode)) {
1322 			rcu_read_unlock();
1323 			break;
1324 		}
1325 		/* Save the delegation */
1326 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1327 		rcu_read_unlock();
1328 		nfs_release_seqid(opendata->o_arg.seqid);
1329 		if (!opendata->is_recover) {
1330 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1331 			if (ret != 0)
1332 				goto out;
1333 		}
1334 		ret = -EAGAIN;
1335 
1336 		/* Try to update the stateid using the delegation */
1337 		if (update_open_stateid(state, NULL, &stateid, fmode))
1338 			goto out_return_state;
1339 	}
1340 out:
1341 	return ERR_PTR(ret);
1342 out_return_state:
1343 	atomic_inc(&state->count);
1344 	return state;
1345 }
1346 
1347 static void
1348 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1349 {
1350 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1351 	struct nfs_delegation *delegation;
1352 	int delegation_flags = 0;
1353 
1354 	rcu_read_lock();
1355 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1356 	if (delegation)
1357 		delegation_flags = delegation->flags;
1358 	rcu_read_unlock();
1359 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1360 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1361 				   "returning a delegation for "
1362 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1363 				   clp->cl_hostname);
1364 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1365 		nfs_inode_set_delegation(state->inode,
1366 					 data->owner->so_cred,
1367 					 &data->o_res);
1368 	else
1369 		nfs_inode_reclaim_delegation(state->inode,
1370 					     data->owner->so_cred,
1371 					     &data->o_res);
1372 }
1373 
1374 /*
1375  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1376  * and update the nfs4_state.
1377  */
1378 static struct nfs4_state *
1379 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1380 {
1381 	struct inode *inode = data->state->inode;
1382 	struct nfs4_state *state = data->state;
1383 	int ret;
1384 
1385 	if (!data->rpc_done) {
1386 		if (data->rpc_status) {
1387 			ret = data->rpc_status;
1388 			goto err;
1389 		}
1390 		/* cached opens have already been processed */
1391 		goto update;
1392 	}
1393 
1394 	ret = nfs_refresh_inode(inode, &data->f_attr);
1395 	if (ret)
1396 		goto err;
1397 
1398 	if (data->o_res.delegation_type != 0)
1399 		nfs4_opendata_check_deleg(data, state);
1400 update:
1401 	update_open_stateid(state, &data->o_res.stateid, NULL,
1402 			    data->o_arg.fmode);
1403 	atomic_inc(&state->count);
1404 
1405 	return state;
1406 err:
1407 	return ERR_PTR(ret);
1408 
1409 }
1410 
1411 static struct nfs4_state *
1412 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1413 {
1414 	struct inode *inode;
1415 	struct nfs4_state *state = NULL;
1416 	int ret;
1417 
1418 	if (!data->rpc_done) {
1419 		state = nfs4_try_open_cached(data);
1420 		goto out;
1421 	}
1422 
1423 	ret = -EAGAIN;
1424 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1425 		goto err;
1426 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1427 	ret = PTR_ERR(inode);
1428 	if (IS_ERR(inode))
1429 		goto err;
1430 	ret = -ENOMEM;
1431 	state = nfs4_get_open_state(inode, data->owner);
1432 	if (state == NULL)
1433 		goto err_put_inode;
1434 	if (data->o_res.delegation_type != 0)
1435 		nfs4_opendata_check_deleg(data, state);
1436 	update_open_stateid(state, &data->o_res.stateid, NULL,
1437 			data->o_arg.fmode);
1438 	iput(inode);
1439 out:
1440 	nfs_release_seqid(data->o_arg.seqid);
1441 	return state;
1442 err_put_inode:
1443 	iput(inode);
1444 err:
1445 	return ERR_PTR(ret);
1446 }
1447 
1448 static struct nfs4_state *
1449 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1450 {
1451 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1452 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1453 	return _nfs4_opendata_to_nfs4_state(data);
1454 }
1455 
1456 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1457 {
1458 	struct nfs_inode *nfsi = NFS_I(state->inode);
1459 	struct nfs_open_context *ctx;
1460 
1461 	spin_lock(&state->inode->i_lock);
1462 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1463 		if (ctx->state != state)
1464 			continue;
1465 		get_nfs_open_context(ctx);
1466 		spin_unlock(&state->inode->i_lock);
1467 		return ctx;
1468 	}
1469 	spin_unlock(&state->inode->i_lock);
1470 	return ERR_PTR(-ENOENT);
1471 }
1472 
1473 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1474 		struct nfs4_state *state, enum open_claim_type4 claim)
1475 {
1476 	struct nfs4_opendata *opendata;
1477 
1478 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1479 			NULL, NULL, claim, GFP_NOFS);
1480 	if (opendata == NULL)
1481 		return ERR_PTR(-ENOMEM);
1482 	opendata->state = state;
1483 	atomic_inc(&state->count);
1484 	return opendata;
1485 }
1486 
1487 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1488 {
1489 	struct nfs4_state *newstate;
1490 	int ret;
1491 
1492 	opendata->o_arg.open_flags = 0;
1493 	opendata->o_arg.fmode = fmode;
1494 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1495 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1496 	nfs4_init_opendata_res(opendata);
1497 	ret = _nfs4_recover_proc_open(opendata);
1498 	if (ret != 0)
1499 		return ret;
1500 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1501 	if (IS_ERR(newstate))
1502 		return PTR_ERR(newstate);
1503 	nfs4_close_state(newstate, fmode);
1504 	*res = newstate;
1505 	return 0;
1506 }
1507 
1508 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1509 {
1510 	struct nfs4_state *newstate;
1511 	int ret;
1512 
1513 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1514 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1515 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1516 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1517 	/* memory barrier prior to reading state->n_* */
1518 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1519 	clear_bit(NFS_OPEN_STATE, &state->flags);
1520 	smp_rmb();
1521 	if (state->n_rdwr != 0) {
1522 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1523 		if (ret != 0)
1524 			return ret;
1525 		if (newstate != state)
1526 			return -ESTALE;
1527 	}
1528 	if (state->n_wronly != 0) {
1529 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1530 		if (ret != 0)
1531 			return ret;
1532 		if (newstate != state)
1533 			return -ESTALE;
1534 	}
1535 	if (state->n_rdonly != 0) {
1536 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1537 		if (ret != 0)
1538 			return ret;
1539 		if (newstate != state)
1540 			return -ESTALE;
1541 	}
1542 	/*
1543 	 * We may have performed cached opens for all three recoveries.
1544 	 * Check if we need to update the current stateid.
1545 	 */
1546 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1547 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1548 		write_seqlock(&state->seqlock);
1549 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1550 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1551 		write_sequnlock(&state->seqlock);
1552 	}
1553 	return 0;
1554 }
1555 
1556 /*
1557  * OPEN_RECLAIM:
1558  * 	reclaim state on the server after a reboot.
1559  */
1560 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1561 {
1562 	struct nfs_delegation *delegation;
1563 	struct nfs4_opendata *opendata;
1564 	fmode_t delegation_type = 0;
1565 	int status;
1566 
1567 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1568 			NFS4_OPEN_CLAIM_PREVIOUS);
1569 	if (IS_ERR(opendata))
1570 		return PTR_ERR(opendata);
1571 	rcu_read_lock();
1572 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1573 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1574 		delegation_type = delegation->type;
1575 	rcu_read_unlock();
1576 	opendata->o_arg.u.delegation_type = delegation_type;
1577 	status = nfs4_open_recover(opendata, state);
1578 	nfs4_opendata_put(opendata);
1579 	return status;
1580 }
1581 
1582 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1583 {
1584 	struct nfs_server *server = NFS_SERVER(state->inode);
1585 	struct nfs4_exception exception = { };
1586 	int err;
1587 	do {
1588 		err = _nfs4_do_open_reclaim(ctx, state);
1589 		trace_nfs4_open_reclaim(ctx, 0, err);
1590 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1591 			continue;
1592 		if (err != -NFS4ERR_DELAY)
1593 			break;
1594 		nfs4_handle_exception(server, err, &exception);
1595 	} while (exception.retry);
1596 	return err;
1597 }
1598 
1599 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1600 {
1601 	struct nfs_open_context *ctx;
1602 	int ret;
1603 
1604 	ctx = nfs4_state_find_open_context(state);
1605 	if (IS_ERR(ctx))
1606 		return -EAGAIN;
1607 	ret = nfs4_do_open_reclaim(ctx, state);
1608 	put_nfs_open_context(ctx);
1609 	return ret;
1610 }
1611 
1612 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1613 {
1614 	switch (err) {
1615 		default:
1616 			printk(KERN_ERR "NFS: %s: unhandled error "
1617 					"%d.\n", __func__, err);
1618 		case 0:
1619 		case -ENOENT:
1620 		case -ESTALE:
1621 			break;
1622 		case -NFS4ERR_BADSESSION:
1623 		case -NFS4ERR_BADSLOT:
1624 		case -NFS4ERR_BAD_HIGH_SLOT:
1625 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1626 		case -NFS4ERR_DEADSESSION:
1627 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1628 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1629 			return -EAGAIN;
1630 		case -NFS4ERR_STALE_CLIENTID:
1631 		case -NFS4ERR_STALE_STATEID:
1632 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1633 		case -NFS4ERR_EXPIRED:
1634 			/* Don't recall a delegation if it was lost */
1635 			nfs4_schedule_lease_recovery(server->nfs_client);
1636 			return -EAGAIN;
1637 		case -NFS4ERR_MOVED:
1638 			nfs4_schedule_migration_recovery(server);
1639 			return -EAGAIN;
1640 		case -NFS4ERR_LEASE_MOVED:
1641 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1642 			return -EAGAIN;
1643 		case -NFS4ERR_DELEG_REVOKED:
1644 		case -NFS4ERR_ADMIN_REVOKED:
1645 		case -NFS4ERR_BAD_STATEID:
1646 		case -NFS4ERR_OPENMODE:
1647 			nfs_inode_find_state_and_recover(state->inode,
1648 					stateid);
1649 			nfs4_schedule_stateid_recovery(server, state);
1650 			return 0;
1651 		case -NFS4ERR_DELAY:
1652 		case -NFS4ERR_GRACE:
1653 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1654 			ssleep(1);
1655 			return -EAGAIN;
1656 		case -ENOMEM:
1657 		case -NFS4ERR_DENIED:
1658 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1659 			return 0;
1660 	}
1661 	return err;
1662 }
1663 
1664 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1665 {
1666 	struct nfs_server *server = NFS_SERVER(state->inode);
1667 	struct nfs4_opendata *opendata;
1668 	int err;
1669 
1670 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1671 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1672 	if (IS_ERR(opendata))
1673 		return PTR_ERR(opendata);
1674 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1675 	err = nfs4_open_recover(opendata, state);
1676 	nfs4_opendata_put(opendata);
1677 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1678 }
1679 
1680 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1681 {
1682 	struct nfs4_opendata *data = calldata;
1683 
1684 	nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1685 				&data->c_res.seq_res, task);
1686 }
1687 
1688 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1689 {
1690 	struct nfs4_opendata *data = calldata;
1691 
1692 	nfs40_sequence_done(task, &data->c_res.seq_res);
1693 
1694 	data->rpc_status = task->tk_status;
1695 	if (data->rpc_status == 0) {
1696 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1697 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1698 		renew_lease(data->o_res.server, data->timestamp);
1699 		data->rpc_done = 1;
1700 	}
1701 }
1702 
1703 static void nfs4_open_confirm_release(void *calldata)
1704 {
1705 	struct nfs4_opendata *data = calldata;
1706 	struct nfs4_state *state = NULL;
1707 
1708 	/* If this request hasn't been cancelled, do nothing */
1709 	if (data->cancelled == 0)
1710 		goto out_free;
1711 	/* In case of error, no cleanup! */
1712 	if (!data->rpc_done)
1713 		goto out_free;
1714 	state = nfs4_opendata_to_nfs4_state(data);
1715 	if (!IS_ERR(state))
1716 		nfs4_close_state(state, data->o_arg.fmode);
1717 out_free:
1718 	nfs4_opendata_put(data);
1719 }
1720 
1721 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1722 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1723 	.rpc_call_done = nfs4_open_confirm_done,
1724 	.rpc_release = nfs4_open_confirm_release,
1725 };
1726 
1727 /*
1728  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1729  */
1730 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1731 {
1732 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1733 	struct rpc_task *task;
1734 	struct  rpc_message msg = {
1735 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1736 		.rpc_argp = &data->c_arg,
1737 		.rpc_resp = &data->c_res,
1738 		.rpc_cred = data->owner->so_cred,
1739 	};
1740 	struct rpc_task_setup task_setup_data = {
1741 		.rpc_client = server->client,
1742 		.rpc_message = &msg,
1743 		.callback_ops = &nfs4_open_confirm_ops,
1744 		.callback_data = data,
1745 		.workqueue = nfsiod_workqueue,
1746 		.flags = RPC_TASK_ASYNC,
1747 	};
1748 	int status;
1749 
1750 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1751 	kref_get(&data->kref);
1752 	data->rpc_done = 0;
1753 	data->rpc_status = 0;
1754 	data->timestamp = jiffies;
1755 	task = rpc_run_task(&task_setup_data);
1756 	if (IS_ERR(task))
1757 		return PTR_ERR(task);
1758 	status = nfs4_wait_for_completion_rpc_task(task);
1759 	if (status != 0) {
1760 		data->cancelled = 1;
1761 		smp_wmb();
1762 	} else
1763 		status = data->rpc_status;
1764 	rpc_put_task(task);
1765 	return status;
1766 }
1767 
1768 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1769 {
1770 	struct nfs4_opendata *data = calldata;
1771 	struct nfs4_state_owner *sp = data->owner;
1772 	struct nfs_client *clp = sp->so_server->nfs_client;
1773 
1774 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1775 		goto out_wait;
1776 	/*
1777 	 * Check if we still need to send an OPEN call, or if we can use
1778 	 * a delegation instead.
1779 	 */
1780 	if (data->state != NULL) {
1781 		struct nfs_delegation *delegation;
1782 
1783 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1784 			goto out_no_action;
1785 		rcu_read_lock();
1786 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1787 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1788 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1789 		    can_open_delegated(delegation, data->o_arg.fmode))
1790 			goto unlock_no_action;
1791 		rcu_read_unlock();
1792 	}
1793 	/* Update client id. */
1794 	data->o_arg.clientid = clp->cl_clientid;
1795 	switch (data->o_arg.claim) {
1796 	case NFS4_OPEN_CLAIM_PREVIOUS:
1797 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1798 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1799 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1800 	case NFS4_OPEN_CLAIM_FH:
1801 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1802 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1803 	}
1804 	data->timestamp = jiffies;
1805 	if (nfs4_setup_sequence(data->o_arg.server,
1806 				&data->o_arg.seq_args,
1807 				&data->o_res.seq_res,
1808 				task) != 0)
1809 		nfs_release_seqid(data->o_arg.seqid);
1810 
1811 	/* Set the create mode (note dependency on the session type) */
1812 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1813 	if (data->o_arg.open_flags & O_EXCL) {
1814 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1815 		if (nfs4_has_persistent_session(clp))
1816 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1817 		else if (clp->cl_mvops->minor_version > 0)
1818 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1819 	}
1820 	return;
1821 unlock_no_action:
1822 	rcu_read_unlock();
1823 out_no_action:
1824 	task->tk_action = NULL;
1825 out_wait:
1826 	nfs4_sequence_done(task, &data->o_res.seq_res);
1827 }
1828 
1829 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1830 {
1831 	struct nfs4_opendata *data = calldata;
1832 
1833 	data->rpc_status = task->tk_status;
1834 
1835 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1836 		return;
1837 
1838 	if (task->tk_status == 0) {
1839 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1840 			switch (data->o_res.f_attr->mode & S_IFMT) {
1841 			case S_IFREG:
1842 				break;
1843 			case S_IFLNK:
1844 				data->rpc_status = -ELOOP;
1845 				break;
1846 			case S_IFDIR:
1847 				data->rpc_status = -EISDIR;
1848 				break;
1849 			default:
1850 				data->rpc_status = -ENOTDIR;
1851 			}
1852 		}
1853 		renew_lease(data->o_res.server, data->timestamp);
1854 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1855 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1856 	}
1857 	data->rpc_done = 1;
1858 }
1859 
1860 static void nfs4_open_release(void *calldata)
1861 {
1862 	struct nfs4_opendata *data = calldata;
1863 	struct nfs4_state *state = NULL;
1864 
1865 	/* If this request hasn't been cancelled, do nothing */
1866 	if (data->cancelled == 0)
1867 		goto out_free;
1868 	/* In case of error, no cleanup! */
1869 	if (data->rpc_status != 0 || !data->rpc_done)
1870 		goto out_free;
1871 	/* In case we need an open_confirm, no cleanup! */
1872 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1873 		goto out_free;
1874 	state = nfs4_opendata_to_nfs4_state(data);
1875 	if (!IS_ERR(state))
1876 		nfs4_close_state(state, data->o_arg.fmode);
1877 out_free:
1878 	nfs4_opendata_put(data);
1879 }
1880 
1881 static const struct rpc_call_ops nfs4_open_ops = {
1882 	.rpc_call_prepare = nfs4_open_prepare,
1883 	.rpc_call_done = nfs4_open_done,
1884 	.rpc_release = nfs4_open_release,
1885 };
1886 
1887 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1888 {
1889 	struct inode *dir = data->dir->d_inode;
1890 	struct nfs_server *server = NFS_SERVER(dir);
1891 	struct nfs_openargs *o_arg = &data->o_arg;
1892 	struct nfs_openres *o_res = &data->o_res;
1893 	struct rpc_task *task;
1894 	struct rpc_message msg = {
1895 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1896 		.rpc_argp = o_arg,
1897 		.rpc_resp = o_res,
1898 		.rpc_cred = data->owner->so_cred,
1899 	};
1900 	struct rpc_task_setup task_setup_data = {
1901 		.rpc_client = server->client,
1902 		.rpc_message = &msg,
1903 		.callback_ops = &nfs4_open_ops,
1904 		.callback_data = data,
1905 		.workqueue = nfsiod_workqueue,
1906 		.flags = RPC_TASK_ASYNC,
1907 	};
1908 	int status;
1909 
1910 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1911 	kref_get(&data->kref);
1912 	data->rpc_done = 0;
1913 	data->rpc_status = 0;
1914 	data->cancelled = 0;
1915 	data->is_recover = 0;
1916 	if (isrecover) {
1917 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1918 		data->is_recover = 1;
1919 	}
1920 	task = rpc_run_task(&task_setup_data);
1921         if (IS_ERR(task))
1922                 return PTR_ERR(task);
1923         status = nfs4_wait_for_completion_rpc_task(task);
1924         if (status != 0) {
1925                 data->cancelled = 1;
1926                 smp_wmb();
1927         } else
1928                 status = data->rpc_status;
1929         rpc_put_task(task);
1930 
1931 	return status;
1932 }
1933 
1934 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1935 {
1936 	struct inode *dir = data->dir->d_inode;
1937 	struct nfs_openres *o_res = &data->o_res;
1938         int status;
1939 
1940 	status = nfs4_run_open_task(data, 1);
1941 	if (status != 0 || !data->rpc_done)
1942 		return status;
1943 
1944 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1945 
1946 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1947 		status = _nfs4_proc_open_confirm(data);
1948 		if (status != 0)
1949 			return status;
1950 	}
1951 
1952 	return status;
1953 }
1954 
1955 static int nfs4_opendata_access(struct rpc_cred *cred,
1956 				struct nfs4_opendata *opendata,
1957 				struct nfs4_state *state, fmode_t fmode,
1958 				int openflags)
1959 {
1960 	struct nfs_access_entry cache;
1961 	u32 mask;
1962 
1963 	/* access call failed or for some reason the server doesn't
1964 	 * support any access modes -- defer access call until later */
1965 	if (opendata->o_res.access_supported == 0)
1966 		return 0;
1967 
1968 	mask = 0;
1969 	/* don't check MAY_WRITE - a newly created file may not have
1970 	 * write mode bits, but POSIX allows the creating process to write.
1971 	 * use openflags to check for exec, because fmode won't
1972 	 * always have FMODE_EXEC set when file open for exec. */
1973 	if (openflags & __FMODE_EXEC) {
1974 		/* ONLY check for exec rights */
1975 		mask = MAY_EXEC;
1976 	} else if (fmode & FMODE_READ)
1977 		mask = MAY_READ;
1978 
1979 	cache.cred = cred;
1980 	cache.jiffies = jiffies;
1981 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
1982 	nfs_access_add_cache(state->inode, &cache);
1983 
1984 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1985 		return 0;
1986 
1987 	/* even though OPEN succeeded, access is denied. Close the file */
1988 	nfs4_close_state(state, fmode);
1989 	return -EACCES;
1990 }
1991 
1992 /*
1993  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1994  */
1995 static int _nfs4_proc_open(struct nfs4_opendata *data)
1996 {
1997 	struct inode *dir = data->dir->d_inode;
1998 	struct nfs_server *server = NFS_SERVER(dir);
1999 	struct nfs_openargs *o_arg = &data->o_arg;
2000 	struct nfs_openres *o_res = &data->o_res;
2001 	int status;
2002 
2003 	status = nfs4_run_open_task(data, 0);
2004 	if (!data->rpc_done)
2005 		return status;
2006 	if (status != 0) {
2007 		if (status == -NFS4ERR_BADNAME &&
2008 				!(o_arg->open_flags & O_CREAT))
2009 			return -ENOENT;
2010 		return status;
2011 	}
2012 
2013 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2014 
2015 	if (o_arg->open_flags & O_CREAT) {
2016 		update_changeattr(dir, &o_res->cinfo);
2017 		if (o_arg->open_flags & O_EXCL)
2018 			data->file_created = 1;
2019 		else if (o_res->cinfo.before != o_res->cinfo.after)
2020 			data->file_created = 1;
2021 	}
2022 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2023 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2024 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2025 		status = _nfs4_proc_open_confirm(data);
2026 		if (status != 0)
2027 			return status;
2028 	}
2029 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2030 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2031 	return 0;
2032 }
2033 
2034 static int nfs4_recover_expired_lease(struct nfs_server *server)
2035 {
2036 	return nfs4_client_recover_expired_lease(server->nfs_client);
2037 }
2038 
2039 /*
2040  * OPEN_EXPIRED:
2041  * 	reclaim state on the server after a network partition.
2042  * 	Assumes caller holds the appropriate lock
2043  */
2044 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2045 {
2046 	struct nfs4_opendata *opendata;
2047 	int ret;
2048 
2049 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2050 			NFS4_OPEN_CLAIM_FH);
2051 	if (IS_ERR(opendata))
2052 		return PTR_ERR(opendata);
2053 	ret = nfs4_open_recover(opendata, state);
2054 	if (ret == -ESTALE)
2055 		d_drop(ctx->dentry);
2056 	nfs4_opendata_put(opendata);
2057 	return ret;
2058 }
2059 
2060 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2061 {
2062 	struct nfs_server *server = NFS_SERVER(state->inode);
2063 	struct nfs4_exception exception = { };
2064 	int err;
2065 
2066 	do {
2067 		err = _nfs4_open_expired(ctx, state);
2068 		trace_nfs4_open_expired(ctx, 0, err);
2069 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2070 			continue;
2071 		switch (err) {
2072 		default:
2073 			goto out;
2074 		case -NFS4ERR_GRACE:
2075 		case -NFS4ERR_DELAY:
2076 			nfs4_handle_exception(server, err, &exception);
2077 			err = 0;
2078 		}
2079 	} while (exception.retry);
2080 out:
2081 	return err;
2082 }
2083 
2084 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2085 {
2086 	struct nfs_open_context *ctx;
2087 	int ret;
2088 
2089 	ctx = nfs4_state_find_open_context(state);
2090 	if (IS_ERR(ctx))
2091 		return -EAGAIN;
2092 	ret = nfs4_do_open_expired(ctx, state);
2093 	put_nfs_open_context(ctx);
2094 	return ret;
2095 }
2096 
2097 #if defined(CONFIG_NFS_V4_1)
2098 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
2099 {
2100 	struct nfs_server *server = NFS_SERVER(state->inode);
2101 	nfs4_stateid *stateid = &state->stateid;
2102 	struct nfs_delegation *delegation;
2103 	struct rpc_cred *cred = NULL;
2104 	int status = -NFS4ERR_BAD_STATEID;
2105 
2106 	/* If a state reset has been done, test_stateid is unneeded */
2107 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2108 		return;
2109 
2110 	/* Get the delegation credential for use by test/free_stateid */
2111 	rcu_read_lock();
2112 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2113 	if (delegation != NULL &&
2114 	    nfs4_stateid_match(&delegation->stateid, stateid)) {
2115 		cred = get_rpccred(delegation->cred);
2116 		rcu_read_unlock();
2117 		status = nfs41_test_stateid(server, stateid, cred);
2118 		trace_nfs4_test_delegation_stateid(state, NULL, status);
2119 	} else
2120 		rcu_read_unlock();
2121 
2122 	if (status != NFS_OK) {
2123 		/* Free the stateid unless the server explicitly
2124 		 * informs us the stateid is unrecognized. */
2125 		if (status != -NFS4ERR_BAD_STATEID)
2126 			nfs41_free_stateid(server, stateid, cred);
2127 		nfs_remove_bad_delegation(state->inode);
2128 
2129 		write_seqlock(&state->seqlock);
2130 		nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2131 		write_sequnlock(&state->seqlock);
2132 		clear_bit(NFS_DELEGATED_STATE, &state->flags);
2133 	}
2134 
2135 	if (cred != NULL)
2136 		put_rpccred(cred);
2137 }
2138 
2139 /**
2140  * nfs41_check_open_stateid - possibly free an open stateid
2141  *
2142  * @state: NFSv4 state for an inode
2143  *
2144  * Returns NFS_OK if recovery for this stateid is now finished.
2145  * Otherwise a negative NFS4ERR value is returned.
2146  */
2147 static int nfs41_check_open_stateid(struct nfs4_state *state)
2148 {
2149 	struct nfs_server *server = NFS_SERVER(state->inode);
2150 	nfs4_stateid *stateid = &state->open_stateid;
2151 	struct rpc_cred *cred = state->owner->so_cred;
2152 	int status;
2153 
2154 	/* If a state reset has been done, test_stateid is unneeded */
2155 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2156 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2157 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2158 		return -NFS4ERR_BAD_STATEID;
2159 
2160 	status = nfs41_test_stateid(server, stateid, cred);
2161 	trace_nfs4_test_open_stateid(state, NULL, status);
2162 	if (status != NFS_OK) {
2163 		/* Free the stateid unless the server explicitly
2164 		 * informs us the stateid is unrecognized. */
2165 		if (status != -NFS4ERR_BAD_STATEID)
2166 			nfs41_free_stateid(server, stateid, cred);
2167 
2168 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2169 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2170 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2171 		clear_bit(NFS_OPEN_STATE, &state->flags);
2172 	}
2173 	return status;
2174 }
2175 
2176 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2177 {
2178 	int status;
2179 
2180 	nfs41_clear_delegation_stateid(state);
2181 	status = nfs41_check_open_stateid(state);
2182 	if (status != NFS_OK)
2183 		status = nfs4_open_expired(sp, state);
2184 	return status;
2185 }
2186 #endif
2187 
2188 /*
2189  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2190  * fields corresponding to attributes that were used to store the verifier.
2191  * Make sure we clobber those fields in the later setattr call
2192  */
2193 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2194 {
2195 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2196 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2197 		sattr->ia_valid |= ATTR_ATIME;
2198 
2199 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2200 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2201 		sattr->ia_valid |= ATTR_MTIME;
2202 }
2203 
2204 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2205 		fmode_t fmode,
2206 		int flags,
2207 		struct nfs_open_context *ctx)
2208 {
2209 	struct nfs4_state_owner *sp = opendata->owner;
2210 	struct nfs_server *server = sp->so_server;
2211 	struct dentry *dentry;
2212 	struct nfs4_state *state;
2213 	unsigned int seq;
2214 	int ret;
2215 
2216 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2217 
2218 	ret = _nfs4_proc_open(opendata);
2219 	if (ret != 0)
2220 		goto out;
2221 
2222 	state = nfs4_opendata_to_nfs4_state(opendata);
2223 	ret = PTR_ERR(state);
2224 	if (IS_ERR(state))
2225 		goto out;
2226 	if (server->caps & NFS_CAP_POSIX_LOCK)
2227 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2228 
2229 	dentry = opendata->dentry;
2230 	if (dentry->d_inode == NULL) {
2231 		/* FIXME: Is this d_drop() ever needed? */
2232 		d_drop(dentry);
2233 		dentry = d_add_unique(dentry, igrab(state->inode));
2234 		if (dentry == NULL) {
2235 			dentry = opendata->dentry;
2236 		} else if (dentry != ctx->dentry) {
2237 			dput(ctx->dentry);
2238 			ctx->dentry = dget(dentry);
2239 		}
2240 		nfs_set_verifier(dentry,
2241 				nfs_save_change_attribute(opendata->dir->d_inode));
2242 	}
2243 
2244 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2245 	if (ret != 0)
2246 		goto out;
2247 
2248 	ctx->state = state;
2249 	if (dentry->d_inode == state->inode) {
2250 		nfs_inode_attach_open_context(ctx);
2251 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2252 			nfs4_schedule_stateid_recovery(server, state);
2253 	}
2254 out:
2255 	return ret;
2256 }
2257 
2258 /*
2259  * Returns a referenced nfs4_state
2260  */
2261 static int _nfs4_do_open(struct inode *dir,
2262 			struct nfs_open_context *ctx,
2263 			int flags,
2264 			struct iattr *sattr,
2265 			struct nfs4_label *label,
2266 			int *opened)
2267 {
2268 	struct nfs4_state_owner  *sp;
2269 	struct nfs4_state     *state = NULL;
2270 	struct nfs_server       *server = NFS_SERVER(dir);
2271 	struct nfs4_opendata *opendata;
2272 	struct dentry *dentry = ctx->dentry;
2273 	struct rpc_cred *cred = ctx->cred;
2274 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2275 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2276 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2277 	struct nfs4_label *olabel = NULL;
2278 	int status;
2279 
2280 	/* Protect against reboot recovery conflicts */
2281 	status = -ENOMEM;
2282 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2283 	if (sp == NULL) {
2284 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2285 		goto out_err;
2286 	}
2287 	status = nfs4_recover_expired_lease(server);
2288 	if (status != 0)
2289 		goto err_put_state_owner;
2290 	if (dentry->d_inode != NULL)
2291 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2292 	status = -ENOMEM;
2293 	if (dentry->d_inode)
2294 		claim = NFS4_OPEN_CLAIM_FH;
2295 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2296 			label, claim, GFP_KERNEL);
2297 	if (opendata == NULL)
2298 		goto err_put_state_owner;
2299 
2300 	if (label) {
2301 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2302 		if (IS_ERR(olabel)) {
2303 			status = PTR_ERR(olabel);
2304 			goto err_opendata_put;
2305 		}
2306 	}
2307 
2308 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2309 		if (!opendata->f_attr.mdsthreshold) {
2310 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2311 			if (!opendata->f_attr.mdsthreshold)
2312 				goto err_free_label;
2313 		}
2314 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2315 	}
2316 	if (dentry->d_inode != NULL)
2317 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2318 
2319 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2320 	if (status != 0)
2321 		goto err_free_label;
2322 	state = ctx->state;
2323 
2324 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2325 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2326 		nfs4_exclusive_attrset(opendata, sattr);
2327 
2328 		nfs_fattr_init(opendata->o_res.f_attr);
2329 		status = nfs4_do_setattr(state->inode, cred,
2330 				opendata->o_res.f_attr, sattr,
2331 				state, label, olabel);
2332 		if (status == 0) {
2333 			nfs_setattr_update_inode(state->inode, sattr);
2334 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2335 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2336 		}
2337 	}
2338 	if (opendata->file_created)
2339 		*opened |= FILE_CREATED;
2340 
2341 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2342 		*ctx_th = opendata->f_attr.mdsthreshold;
2343 		opendata->f_attr.mdsthreshold = NULL;
2344 	}
2345 
2346 	nfs4_label_free(olabel);
2347 
2348 	nfs4_opendata_put(opendata);
2349 	nfs4_put_state_owner(sp);
2350 	return 0;
2351 err_free_label:
2352 	nfs4_label_free(olabel);
2353 err_opendata_put:
2354 	nfs4_opendata_put(opendata);
2355 err_put_state_owner:
2356 	nfs4_put_state_owner(sp);
2357 out_err:
2358 	return status;
2359 }
2360 
2361 
2362 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2363 					struct nfs_open_context *ctx,
2364 					int flags,
2365 					struct iattr *sattr,
2366 					struct nfs4_label *label,
2367 					int *opened)
2368 {
2369 	struct nfs_server *server = NFS_SERVER(dir);
2370 	struct nfs4_exception exception = { };
2371 	struct nfs4_state *res;
2372 	int status;
2373 
2374 	do {
2375 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2376 		res = ctx->state;
2377 		trace_nfs4_open_file(ctx, flags, status);
2378 		if (status == 0)
2379 			break;
2380 		/* NOTE: BAD_SEQID means the server and client disagree about the
2381 		 * book-keeping w.r.t. state-changing operations
2382 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2383 		 * It is actually a sign of a bug on the client or on the server.
2384 		 *
2385 		 * If we receive a BAD_SEQID error in the particular case of
2386 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2387 		 * have unhashed the old state_owner for us, and that we can
2388 		 * therefore safely retry using a new one. We should still warn
2389 		 * the user though...
2390 		 */
2391 		if (status == -NFS4ERR_BAD_SEQID) {
2392 			pr_warn_ratelimited("NFS: v4 server %s "
2393 					" returned a bad sequence-id error!\n",
2394 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2395 			exception.retry = 1;
2396 			continue;
2397 		}
2398 		/*
2399 		 * BAD_STATEID on OPEN means that the server cancelled our
2400 		 * state before it received the OPEN_CONFIRM.
2401 		 * Recover by retrying the request as per the discussion
2402 		 * on Page 181 of RFC3530.
2403 		 */
2404 		if (status == -NFS4ERR_BAD_STATEID) {
2405 			exception.retry = 1;
2406 			continue;
2407 		}
2408 		if (status == -EAGAIN) {
2409 			/* We must have found a delegation */
2410 			exception.retry = 1;
2411 			continue;
2412 		}
2413 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2414 			continue;
2415 		res = ERR_PTR(nfs4_handle_exception(server,
2416 					status, &exception));
2417 	} while (exception.retry);
2418 	return res;
2419 }
2420 
2421 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2422 			    struct nfs_fattr *fattr, struct iattr *sattr,
2423 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2424 			    struct nfs4_label *olabel)
2425 {
2426 	struct nfs_server *server = NFS_SERVER(inode);
2427         struct nfs_setattrargs  arg = {
2428                 .fh             = NFS_FH(inode),
2429                 .iap            = sattr,
2430 		.server		= server,
2431 		.bitmask = server->attr_bitmask,
2432 		.label		= ilabel,
2433         };
2434         struct nfs_setattrres  res = {
2435 		.fattr		= fattr,
2436 		.label		= olabel,
2437 		.server		= server,
2438         };
2439         struct rpc_message msg = {
2440 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2441 		.rpc_argp	= &arg,
2442 		.rpc_resp	= &res,
2443 		.rpc_cred	= cred,
2444         };
2445 	unsigned long timestamp = jiffies;
2446 	fmode_t fmode;
2447 	bool truncate;
2448 	int status;
2449 
2450 	arg.bitmask = nfs4_bitmask(server, ilabel);
2451 	if (ilabel)
2452 		arg.bitmask = nfs4_bitmask(server, olabel);
2453 
2454 	nfs_fattr_init(fattr);
2455 
2456 	/* Servers should only apply open mode checks for file size changes */
2457 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2458 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2459 
2460 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2461 		/* Use that stateid */
2462 	} else if (truncate && state != NULL) {
2463 		struct nfs_lockowner lockowner = {
2464 			.l_owner = current->files,
2465 			.l_pid = current->tgid,
2466 		};
2467 		if (!nfs4_valid_open_stateid(state))
2468 			return -EBADF;
2469 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2470 				&lockowner) == -EIO)
2471 			return -EBADF;
2472 	} else
2473 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2474 
2475 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2476 	if (status == 0 && state != NULL)
2477 		renew_lease(server, timestamp);
2478 	return status;
2479 }
2480 
2481 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2482 			   struct nfs_fattr *fattr, struct iattr *sattr,
2483 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2484 			   struct nfs4_label *olabel)
2485 {
2486 	struct nfs_server *server = NFS_SERVER(inode);
2487 	struct nfs4_exception exception = {
2488 		.state = state,
2489 		.inode = inode,
2490 	};
2491 	int err;
2492 	do {
2493 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2494 		trace_nfs4_setattr(inode, err);
2495 		switch (err) {
2496 		case -NFS4ERR_OPENMODE:
2497 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2498 				pr_warn_once("NFSv4: server %s is incorrectly "
2499 						"applying open mode checks to "
2500 						"a SETATTR that is not "
2501 						"changing file size.\n",
2502 						server->nfs_client->cl_hostname);
2503 			}
2504 			if (state && !(state->state & FMODE_WRITE)) {
2505 				err = -EBADF;
2506 				if (sattr->ia_valid & ATTR_OPEN)
2507 					err = -EACCES;
2508 				goto out;
2509 			}
2510 		}
2511 		err = nfs4_handle_exception(server, err, &exception);
2512 	} while (exception.retry);
2513 out:
2514 	return err;
2515 }
2516 
2517 struct nfs4_closedata {
2518 	struct inode *inode;
2519 	struct nfs4_state *state;
2520 	struct nfs_closeargs arg;
2521 	struct nfs_closeres res;
2522 	struct nfs_fattr fattr;
2523 	unsigned long timestamp;
2524 	bool roc;
2525 	u32 roc_barrier;
2526 };
2527 
2528 static void nfs4_free_closedata(void *data)
2529 {
2530 	struct nfs4_closedata *calldata = data;
2531 	struct nfs4_state_owner *sp = calldata->state->owner;
2532 	struct super_block *sb = calldata->state->inode->i_sb;
2533 
2534 	if (calldata->roc)
2535 		pnfs_roc_release(calldata->state->inode);
2536 	nfs4_put_open_state(calldata->state);
2537 	nfs_free_seqid(calldata->arg.seqid);
2538 	nfs4_put_state_owner(sp);
2539 	nfs_sb_deactive(sb);
2540 	kfree(calldata);
2541 }
2542 
2543 static void nfs4_close_done(struct rpc_task *task, void *data)
2544 {
2545 	struct nfs4_closedata *calldata = data;
2546 	struct nfs4_state *state = calldata->state;
2547 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2548 
2549 	dprintk("%s: begin!\n", __func__);
2550 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2551 		return;
2552 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2553         /* hmm. we are done with the inode, and in the process of freeing
2554 	 * the state_owner. we keep this around to process errors
2555 	 */
2556 	switch (task->tk_status) {
2557 		case 0:
2558 			if (calldata->roc)
2559 				pnfs_roc_set_barrier(state->inode,
2560 						     calldata->roc_barrier);
2561 			nfs_clear_open_stateid(state, &calldata->res.stateid, 0);
2562 			renew_lease(server, calldata->timestamp);
2563 			goto out_release;
2564 		case -NFS4ERR_ADMIN_REVOKED:
2565 		case -NFS4ERR_STALE_STATEID:
2566 		case -NFS4ERR_OLD_STATEID:
2567 		case -NFS4ERR_BAD_STATEID:
2568 		case -NFS4ERR_EXPIRED:
2569 			if (calldata->arg.fmode == 0)
2570 				break;
2571 		default:
2572 			if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
2573 				rpc_restart_call_prepare(task);
2574 				goto out_release;
2575 			}
2576 	}
2577 	nfs_clear_open_stateid(state, NULL, calldata->arg.fmode);
2578 out_release:
2579 	nfs_release_seqid(calldata->arg.seqid);
2580 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2581 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2582 }
2583 
2584 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2585 {
2586 	struct nfs4_closedata *calldata = data;
2587 	struct nfs4_state *state = calldata->state;
2588 	struct inode *inode = calldata->inode;
2589 	int call_close = 0;
2590 
2591 	dprintk("%s: begin!\n", __func__);
2592 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2593 		goto out_wait;
2594 
2595 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2596 	calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2597 	spin_lock(&state->owner->so_lock);
2598 	/* Calculate the change in open mode */
2599 	if (state->n_rdwr == 0) {
2600 		if (state->n_rdonly == 0) {
2601 			call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2602 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2603 			calldata->arg.fmode &= ~FMODE_READ;
2604 		}
2605 		if (state->n_wronly == 0) {
2606 			call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2607 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2608 			calldata->arg.fmode &= ~FMODE_WRITE;
2609 		}
2610 	}
2611 	if (!nfs4_valid_open_stateid(state))
2612 		call_close = 0;
2613 	spin_unlock(&state->owner->so_lock);
2614 
2615 	if (!call_close) {
2616 		/* Note: exit _without_ calling nfs4_close_done */
2617 		goto out_no_action;
2618 	}
2619 
2620 	if (calldata->arg.fmode == 0) {
2621 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2622 		if (calldata->roc &&
2623 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2624 			nfs_release_seqid(calldata->arg.seqid);
2625 			goto out_wait;
2626 		    }
2627 	}
2628 
2629 	nfs_fattr_init(calldata->res.fattr);
2630 	calldata->timestamp = jiffies;
2631 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2632 				&calldata->arg.seq_args,
2633 				&calldata->res.seq_res,
2634 				task) != 0)
2635 		nfs_release_seqid(calldata->arg.seqid);
2636 	dprintk("%s: done!\n", __func__);
2637 	return;
2638 out_no_action:
2639 	task->tk_action = NULL;
2640 out_wait:
2641 	nfs4_sequence_done(task, &calldata->res.seq_res);
2642 }
2643 
2644 static const struct rpc_call_ops nfs4_close_ops = {
2645 	.rpc_call_prepare = nfs4_close_prepare,
2646 	.rpc_call_done = nfs4_close_done,
2647 	.rpc_release = nfs4_free_closedata,
2648 };
2649 
2650 /*
2651  * It is possible for data to be read/written from a mem-mapped file
2652  * after the sys_close call (which hits the vfs layer as a flush).
2653  * This means that we can't safely call nfsv4 close on a file until
2654  * the inode is cleared. This in turn means that we are not good
2655  * NFSv4 citizens - we do not indicate to the server to update the file's
2656  * share state even when we are done with one of the three share
2657  * stateid's in the inode.
2658  *
2659  * NOTE: Caller must be holding the sp->so_owner semaphore!
2660  */
2661 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2662 {
2663 	struct nfs_server *server = NFS_SERVER(state->inode);
2664 	struct nfs4_closedata *calldata;
2665 	struct nfs4_state_owner *sp = state->owner;
2666 	struct rpc_task *task;
2667 	struct rpc_message msg = {
2668 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2669 		.rpc_cred = state->owner->so_cred,
2670 	};
2671 	struct rpc_task_setup task_setup_data = {
2672 		.rpc_client = server->client,
2673 		.rpc_message = &msg,
2674 		.callback_ops = &nfs4_close_ops,
2675 		.workqueue = nfsiod_workqueue,
2676 		.flags = RPC_TASK_ASYNC,
2677 	};
2678 	int status = -ENOMEM;
2679 
2680 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2681 		&task_setup_data.rpc_client, &msg);
2682 
2683 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2684 	if (calldata == NULL)
2685 		goto out;
2686 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2687 	calldata->inode = state->inode;
2688 	calldata->state = state;
2689 	calldata->arg.fh = NFS_FH(state->inode);
2690 	calldata->arg.stateid = &state->open_stateid;
2691 	/* Serialization for the sequence id */
2692 	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2693 	if (calldata->arg.seqid == NULL)
2694 		goto out_free_calldata;
2695 	calldata->arg.fmode = 0;
2696 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2697 	calldata->res.fattr = &calldata->fattr;
2698 	calldata->res.seqid = calldata->arg.seqid;
2699 	calldata->res.server = server;
2700 	calldata->roc = pnfs_roc(state->inode);
2701 	nfs_sb_active(calldata->inode->i_sb);
2702 
2703 	msg.rpc_argp = &calldata->arg;
2704 	msg.rpc_resp = &calldata->res;
2705 	task_setup_data.callback_data = calldata;
2706 	task = rpc_run_task(&task_setup_data);
2707 	if (IS_ERR(task))
2708 		return PTR_ERR(task);
2709 	status = 0;
2710 	if (wait)
2711 		status = rpc_wait_for_completion_task(task);
2712 	rpc_put_task(task);
2713 	return status;
2714 out_free_calldata:
2715 	kfree(calldata);
2716 out:
2717 	nfs4_put_open_state(state);
2718 	nfs4_put_state_owner(sp);
2719 	return status;
2720 }
2721 
2722 static struct inode *
2723 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2724 		int open_flags, struct iattr *attr, int *opened)
2725 {
2726 	struct nfs4_state *state;
2727 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2728 
2729 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2730 
2731 	/* Protect against concurrent sillydeletes */
2732 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2733 
2734 	nfs4_label_release_security(label);
2735 
2736 	if (IS_ERR(state))
2737 		return ERR_CAST(state);
2738 	return state->inode;
2739 }
2740 
2741 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2742 {
2743 	if (ctx->state == NULL)
2744 		return;
2745 	if (is_sync)
2746 		nfs4_close_sync(ctx->state, ctx->mode);
2747 	else
2748 		nfs4_close_state(ctx->state, ctx->mode);
2749 }
2750 
2751 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2752 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2753 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2754 
2755 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2756 {
2757 	struct nfs4_server_caps_arg args = {
2758 		.fhandle = fhandle,
2759 	};
2760 	struct nfs4_server_caps_res res = {};
2761 	struct rpc_message msg = {
2762 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2763 		.rpc_argp = &args,
2764 		.rpc_resp = &res,
2765 	};
2766 	int status;
2767 
2768 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2769 	if (status == 0) {
2770 		/* Sanity check the server answers */
2771 		switch (server->nfs_client->cl_minorversion) {
2772 		case 0:
2773 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2774 			res.attr_bitmask[2] = 0;
2775 			break;
2776 		case 1:
2777 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2778 			break;
2779 		case 2:
2780 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2781 		}
2782 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2783 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2784 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2785 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2786 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2787 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2788 				NFS_CAP_SECURITY_LABEL);
2789 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2790 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2791 			server->caps |= NFS_CAP_ACLS;
2792 		if (res.has_links != 0)
2793 			server->caps |= NFS_CAP_HARDLINKS;
2794 		if (res.has_symlinks != 0)
2795 			server->caps |= NFS_CAP_SYMLINKS;
2796 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2797 			server->caps |= NFS_CAP_FILEID;
2798 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2799 			server->caps |= NFS_CAP_MODE;
2800 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2801 			server->caps |= NFS_CAP_NLINK;
2802 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2803 			server->caps |= NFS_CAP_OWNER;
2804 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2805 			server->caps |= NFS_CAP_OWNER_GROUP;
2806 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2807 			server->caps |= NFS_CAP_ATIME;
2808 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2809 			server->caps |= NFS_CAP_CTIME;
2810 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2811 			server->caps |= NFS_CAP_MTIME;
2812 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2813 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2814 			server->caps |= NFS_CAP_SECURITY_LABEL;
2815 #endif
2816 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2817 				sizeof(server->attr_bitmask));
2818 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2819 
2820 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2821 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2822 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2823 		server->cache_consistency_bitmask[2] = 0;
2824 		server->acl_bitmask = res.acl_bitmask;
2825 		server->fh_expire_type = res.fh_expire_type;
2826 	}
2827 
2828 	return status;
2829 }
2830 
2831 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2832 {
2833 	struct nfs4_exception exception = { };
2834 	int err;
2835 	do {
2836 		err = nfs4_handle_exception(server,
2837 				_nfs4_server_capabilities(server, fhandle),
2838 				&exception);
2839 	} while (exception.retry);
2840 	return err;
2841 }
2842 
2843 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2844 		struct nfs_fsinfo *info)
2845 {
2846 	u32 bitmask[3];
2847 	struct nfs4_lookup_root_arg args = {
2848 		.bitmask = bitmask,
2849 	};
2850 	struct nfs4_lookup_res res = {
2851 		.server = server,
2852 		.fattr = info->fattr,
2853 		.fh = fhandle,
2854 	};
2855 	struct rpc_message msg = {
2856 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2857 		.rpc_argp = &args,
2858 		.rpc_resp = &res,
2859 	};
2860 
2861 	bitmask[0] = nfs4_fattr_bitmap[0];
2862 	bitmask[1] = nfs4_fattr_bitmap[1];
2863 	/*
2864 	 * Process the label in the upcoming getfattr
2865 	 */
2866 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2867 
2868 	nfs_fattr_init(info->fattr);
2869 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2870 }
2871 
2872 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2873 		struct nfs_fsinfo *info)
2874 {
2875 	struct nfs4_exception exception = { };
2876 	int err;
2877 	do {
2878 		err = _nfs4_lookup_root(server, fhandle, info);
2879 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2880 		switch (err) {
2881 		case 0:
2882 		case -NFS4ERR_WRONGSEC:
2883 			goto out;
2884 		default:
2885 			err = nfs4_handle_exception(server, err, &exception);
2886 		}
2887 	} while (exception.retry);
2888 out:
2889 	return err;
2890 }
2891 
2892 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2893 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2894 {
2895 	struct rpc_auth_create_args auth_args = {
2896 		.pseudoflavor = flavor,
2897 	};
2898 	struct rpc_auth *auth;
2899 	int ret;
2900 
2901 	auth = rpcauth_create(&auth_args, server->client);
2902 	if (IS_ERR(auth)) {
2903 		ret = -EACCES;
2904 		goto out;
2905 	}
2906 	ret = nfs4_lookup_root(server, fhandle, info);
2907 out:
2908 	return ret;
2909 }
2910 
2911 /*
2912  * Retry pseudoroot lookup with various security flavors.  We do this when:
2913  *
2914  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2915  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2916  *
2917  * Returns zero on success, or a negative NFS4ERR value, or a
2918  * negative errno value.
2919  */
2920 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2921 			      struct nfs_fsinfo *info)
2922 {
2923 	/* Per 3530bis 15.33.5 */
2924 	static const rpc_authflavor_t flav_array[] = {
2925 		RPC_AUTH_GSS_KRB5P,
2926 		RPC_AUTH_GSS_KRB5I,
2927 		RPC_AUTH_GSS_KRB5,
2928 		RPC_AUTH_UNIX,			/* courtesy */
2929 		RPC_AUTH_NULL,
2930 	};
2931 	int status = -EPERM;
2932 	size_t i;
2933 
2934 	if (server->auth_info.flavor_len > 0) {
2935 		/* try each flavor specified by user */
2936 		for (i = 0; i < server->auth_info.flavor_len; i++) {
2937 			status = nfs4_lookup_root_sec(server, fhandle, info,
2938 						server->auth_info.flavors[i]);
2939 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2940 				continue;
2941 			break;
2942 		}
2943 	} else {
2944 		/* no flavors specified by user, try default list */
2945 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2946 			status = nfs4_lookup_root_sec(server, fhandle, info,
2947 						      flav_array[i]);
2948 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2949 				continue;
2950 			break;
2951 		}
2952 	}
2953 
2954 	/*
2955 	 * -EACCESS could mean that the user doesn't have correct permissions
2956 	 * to access the mount.  It could also mean that we tried to mount
2957 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2958 	 * existing mount programs don't handle -EACCES very well so it should
2959 	 * be mapped to -EPERM instead.
2960 	 */
2961 	if (status == -EACCES)
2962 		status = -EPERM;
2963 	return status;
2964 }
2965 
2966 static int nfs4_do_find_root_sec(struct nfs_server *server,
2967 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2968 {
2969 	int mv = server->nfs_client->cl_minorversion;
2970 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2971 }
2972 
2973 /**
2974  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2975  * @server: initialized nfs_server handle
2976  * @fhandle: we fill in the pseudo-fs root file handle
2977  * @info: we fill in an FSINFO struct
2978  * @auth_probe: probe the auth flavours
2979  *
2980  * Returns zero on success, or a negative errno.
2981  */
2982 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2983 			 struct nfs_fsinfo *info,
2984 			 bool auth_probe)
2985 {
2986 	int status;
2987 
2988 	switch (auth_probe) {
2989 	case false:
2990 		status = nfs4_lookup_root(server, fhandle, info);
2991 		if (status != -NFS4ERR_WRONGSEC)
2992 			break;
2993 	default:
2994 		status = nfs4_do_find_root_sec(server, fhandle, info);
2995 	}
2996 
2997 	if (status == 0)
2998 		status = nfs4_server_capabilities(server, fhandle);
2999 	if (status == 0)
3000 		status = nfs4_do_fsinfo(server, fhandle, info);
3001 
3002 	return nfs4_map_errors(status);
3003 }
3004 
3005 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3006 			      struct nfs_fsinfo *info)
3007 {
3008 	int error;
3009 	struct nfs_fattr *fattr = info->fattr;
3010 	struct nfs4_label *label = NULL;
3011 
3012 	error = nfs4_server_capabilities(server, mntfh);
3013 	if (error < 0) {
3014 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3015 		return error;
3016 	}
3017 
3018 	label = nfs4_label_alloc(server, GFP_KERNEL);
3019 	if (IS_ERR(label))
3020 		return PTR_ERR(label);
3021 
3022 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3023 	if (error < 0) {
3024 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3025 		goto err_free_label;
3026 	}
3027 
3028 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3029 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3030 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3031 
3032 err_free_label:
3033 	nfs4_label_free(label);
3034 
3035 	return error;
3036 }
3037 
3038 /*
3039  * Get locations and (maybe) other attributes of a referral.
3040  * Note that we'll actually follow the referral later when
3041  * we detect fsid mismatch in inode revalidation
3042  */
3043 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3044 			     const struct qstr *name, struct nfs_fattr *fattr,
3045 			     struct nfs_fh *fhandle)
3046 {
3047 	int status = -ENOMEM;
3048 	struct page *page = NULL;
3049 	struct nfs4_fs_locations *locations = NULL;
3050 
3051 	page = alloc_page(GFP_KERNEL);
3052 	if (page == NULL)
3053 		goto out;
3054 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3055 	if (locations == NULL)
3056 		goto out;
3057 
3058 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3059 	if (status != 0)
3060 		goto out;
3061 
3062 	/*
3063 	 * If the fsid didn't change, this is a migration event, not a
3064 	 * referral.  Cause us to drop into the exception handler, which
3065 	 * will kick off migration recovery.
3066 	 */
3067 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3068 		dprintk("%s: server did not return a different fsid for"
3069 			" a referral at %s\n", __func__, name->name);
3070 		status = -NFS4ERR_MOVED;
3071 		goto out;
3072 	}
3073 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3074 	nfs_fixup_referral_attributes(&locations->fattr);
3075 
3076 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3077 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3078 	memset(fhandle, 0, sizeof(struct nfs_fh));
3079 out:
3080 	if (page)
3081 		__free_page(page);
3082 	kfree(locations);
3083 	return status;
3084 }
3085 
3086 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3087 				struct nfs_fattr *fattr, struct nfs4_label *label)
3088 {
3089 	struct nfs4_getattr_arg args = {
3090 		.fh = fhandle,
3091 		.bitmask = server->attr_bitmask,
3092 	};
3093 	struct nfs4_getattr_res res = {
3094 		.fattr = fattr,
3095 		.label = label,
3096 		.server = server,
3097 	};
3098 	struct rpc_message msg = {
3099 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3100 		.rpc_argp = &args,
3101 		.rpc_resp = &res,
3102 	};
3103 
3104 	args.bitmask = nfs4_bitmask(server, label);
3105 
3106 	nfs_fattr_init(fattr);
3107 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3108 }
3109 
3110 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3111 				struct nfs_fattr *fattr, struct nfs4_label *label)
3112 {
3113 	struct nfs4_exception exception = { };
3114 	int err;
3115 	do {
3116 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3117 		trace_nfs4_getattr(server, fhandle, fattr, err);
3118 		err = nfs4_handle_exception(server, err,
3119 				&exception);
3120 	} while (exception.retry);
3121 	return err;
3122 }
3123 
3124 /*
3125  * The file is not closed if it is opened due to the a request to change
3126  * the size of the file. The open call will not be needed once the
3127  * VFS layer lookup-intents are implemented.
3128  *
3129  * Close is called when the inode is destroyed.
3130  * If we haven't opened the file for O_WRONLY, we
3131  * need to in the size_change case to obtain a stateid.
3132  *
3133  * Got race?
3134  * Because OPEN is always done by name in nfsv4, it is
3135  * possible that we opened a different file by the same
3136  * name.  We can recognize this race condition, but we
3137  * can't do anything about it besides returning an error.
3138  *
3139  * This will be fixed with VFS changes (lookup-intent).
3140  */
3141 static int
3142 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3143 		  struct iattr *sattr)
3144 {
3145 	struct inode *inode = dentry->d_inode;
3146 	struct rpc_cred *cred = NULL;
3147 	struct nfs4_state *state = NULL;
3148 	struct nfs4_label *label = NULL;
3149 	int status;
3150 
3151 	if (pnfs_ld_layoutret_on_setattr(inode))
3152 		pnfs_commit_and_return_layout(inode);
3153 
3154 	nfs_fattr_init(fattr);
3155 
3156 	/* Deal with open(O_TRUNC) */
3157 	if (sattr->ia_valid & ATTR_OPEN)
3158 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3159 
3160 	/* Optimization: if the end result is no change, don't RPC */
3161 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3162 		return 0;
3163 
3164 	/* Search for an existing open(O_WRITE) file */
3165 	if (sattr->ia_valid & ATTR_FILE) {
3166 		struct nfs_open_context *ctx;
3167 
3168 		ctx = nfs_file_open_context(sattr->ia_file);
3169 		if (ctx) {
3170 			cred = ctx->cred;
3171 			state = ctx->state;
3172 		}
3173 	}
3174 
3175 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3176 	if (IS_ERR(label))
3177 		return PTR_ERR(label);
3178 
3179 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3180 	if (status == 0) {
3181 		nfs_setattr_update_inode(inode, sattr);
3182 		nfs_setsecurity(inode, fattr, label);
3183 	}
3184 	nfs4_label_free(label);
3185 	return status;
3186 }
3187 
3188 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3189 		const struct qstr *name, struct nfs_fh *fhandle,
3190 		struct nfs_fattr *fattr, struct nfs4_label *label)
3191 {
3192 	struct nfs_server *server = NFS_SERVER(dir);
3193 	int		       status;
3194 	struct nfs4_lookup_arg args = {
3195 		.bitmask = server->attr_bitmask,
3196 		.dir_fh = NFS_FH(dir),
3197 		.name = name,
3198 	};
3199 	struct nfs4_lookup_res res = {
3200 		.server = server,
3201 		.fattr = fattr,
3202 		.label = label,
3203 		.fh = fhandle,
3204 	};
3205 	struct rpc_message msg = {
3206 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3207 		.rpc_argp = &args,
3208 		.rpc_resp = &res,
3209 	};
3210 
3211 	args.bitmask = nfs4_bitmask(server, label);
3212 
3213 	nfs_fattr_init(fattr);
3214 
3215 	dprintk("NFS call  lookup %s\n", name->name);
3216 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3217 	dprintk("NFS reply lookup: %d\n", status);
3218 	return status;
3219 }
3220 
3221 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3222 {
3223 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3224 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3225 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3226 	fattr->nlink = 2;
3227 }
3228 
3229 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3230 				   struct qstr *name, struct nfs_fh *fhandle,
3231 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3232 {
3233 	struct nfs4_exception exception = { };
3234 	struct rpc_clnt *client = *clnt;
3235 	int err;
3236 	do {
3237 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3238 		trace_nfs4_lookup(dir, name, err);
3239 		switch (err) {
3240 		case -NFS4ERR_BADNAME:
3241 			err = -ENOENT;
3242 			goto out;
3243 		case -NFS4ERR_MOVED:
3244 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3245 			goto out;
3246 		case -NFS4ERR_WRONGSEC:
3247 			err = -EPERM;
3248 			if (client != *clnt)
3249 				goto out;
3250 			client = nfs4_create_sec_client(client, dir, name);
3251 			if (IS_ERR(client))
3252 				return PTR_ERR(client);
3253 
3254 			exception.retry = 1;
3255 			break;
3256 		default:
3257 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3258 		}
3259 	} while (exception.retry);
3260 
3261 out:
3262 	if (err == 0)
3263 		*clnt = client;
3264 	else if (client != *clnt)
3265 		rpc_shutdown_client(client);
3266 
3267 	return err;
3268 }
3269 
3270 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3271 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3272 			    struct nfs4_label *label)
3273 {
3274 	int status;
3275 	struct rpc_clnt *client = NFS_CLIENT(dir);
3276 
3277 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3278 	if (client != NFS_CLIENT(dir)) {
3279 		rpc_shutdown_client(client);
3280 		nfs_fixup_secinfo_attributes(fattr);
3281 	}
3282 	return status;
3283 }
3284 
3285 struct rpc_clnt *
3286 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3287 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3288 {
3289 	struct rpc_clnt *client = NFS_CLIENT(dir);
3290 	int status;
3291 
3292 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3293 	if (status < 0)
3294 		return ERR_PTR(status);
3295 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3296 }
3297 
3298 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3299 {
3300 	struct nfs_server *server = NFS_SERVER(inode);
3301 	struct nfs4_accessargs args = {
3302 		.fh = NFS_FH(inode),
3303 		.bitmask = server->cache_consistency_bitmask,
3304 	};
3305 	struct nfs4_accessres res = {
3306 		.server = server,
3307 	};
3308 	struct rpc_message msg = {
3309 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3310 		.rpc_argp = &args,
3311 		.rpc_resp = &res,
3312 		.rpc_cred = entry->cred,
3313 	};
3314 	int mode = entry->mask;
3315 	int status = 0;
3316 
3317 	/*
3318 	 * Determine which access bits we want to ask for...
3319 	 */
3320 	if (mode & MAY_READ)
3321 		args.access |= NFS4_ACCESS_READ;
3322 	if (S_ISDIR(inode->i_mode)) {
3323 		if (mode & MAY_WRITE)
3324 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3325 		if (mode & MAY_EXEC)
3326 			args.access |= NFS4_ACCESS_LOOKUP;
3327 	} else {
3328 		if (mode & MAY_WRITE)
3329 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3330 		if (mode & MAY_EXEC)
3331 			args.access |= NFS4_ACCESS_EXECUTE;
3332 	}
3333 
3334 	res.fattr = nfs_alloc_fattr();
3335 	if (res.fattr == NULL)
3336 		return -ENOMEM;
3337 
3338 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3339 	if (!status) {
3340 		nfs_access_set_mask(entry, res.access);
3341 		nfs_refresh_inode(inode, res.fattr);
3342 	}
3343 	nfs_free_fattr(res.fattr);
3344 	return status;
3345 }
3346 
3347 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3348 {
3349 	struct nfs4_exception exception = { };
3350 	int err;
3351 	do {
3352 		err = _nfs4_proc_access(inode, entry);
3353 		trace_nfs4_access(inode, err);
3354 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3355 				&exception);
3356 	} while (exception.retry);
3357 	return err;
3358 }
3359 
3360 /*
3361  * TODO: For the time being, we don't try to get any attributes
3362  * along with any of the zero-copy operations READ, READDIR,
3363  * READLINK, WRITE.
3364  *
3365  * In the case of the first three, we want to put the GETATTR
3366  * after the read-type operation -- this is because it is hard
3367  * to predict the length of a GETATTR response in v4, and thus
3368  * align the READ data correctly.  This means that the GETATTR
3369  * may end up partially falling into the page cache, and we should
3370  * shift it into the 'tail' of the xdr_buf before processing.
3371  * To do this efficiently, we need to know the total length
3372  * of data received, which doesn't seem to be available outside
3373  * of the RPC layer.
3374  *
3375  * In the case of WRITE, we also want to put the GETATTR after
3376  * the operation -- in this case because we want to make sure
3377  * we get the post-operation mtime and size.
3378  *
3379  * Both of these changes to the XDR layer would in fact be quite
3380  * minor, but I decided to leave them for a subsequent patch.
3381  */
3382 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3383 		unsigned int pgbase, unsigned int pglen)
3384 {
3385 	struct nfs4_readlink args = {
3386 		.fh       = NFS_FH(inode),
3387 		.pgbase	  = pgbase,
3388 		.pglen    = pglen,
3389 		.pages    = &page,
3390 	};
3391 	struct nfs4_readlink_res res;
3392 	struct rpc_message msg = {
3393 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3394 		.rpc_argp = &args,
3395 		.rpc_resp = &res,
3396 	};
3397 
3398 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3399 }
3400 
3401 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3402 		unsigned int pgbase, unsigned int pglen)
3403 {
3404 	struct nfs4_exception exception = { };
3405 	int err;
3406 	do {
3407 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3408 		trace_nfs4_readlink(inode, err);
3409 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3410 				&exception);
3411 	} while (exception.retry);
3412 	return err;
3413 }
3414 
3415 /*
3416  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3417  */
3418 static int
3419 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3420 		 int flags)
3421 {
3422 	struct nfs4_label l, *ilabel = NULL;
3423 	struct nfs_open_context *ctx;
3424 	struct nfs4_state *state;
3425 	int opened = 0;
3426 	int status = 0;
3427 
3428 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3429 	if (IS_ERR(ctx))
3430 		return PTR_ERR(ctx);
3431 
3432 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3433 
3434 	sattr->ia_mode &= ~current_umask();
3435 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3436 	if (IS_ERR(state)) {
3437 		status = PTR_ERR(state);
3438 		goto out;
3439 	}
3440 out:
3441 	nfs4_label_release_security(ilabel);
3442 	put_nfs_open_context(ctx);
3443 	return status;
3444 }
3445 
3446 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3447 {
3448 	struct nfs_server *server = NFS_SERVER(dir);
3449 	struct nfs_removeargs args = {
3450 		.fh = NFS_FH(dir),
3451 		.name = *name,
3452 	};
3453 	struct nfs_removeres res = {
3454 		.server = server,
3455 	};
3456 	struct rpc_message msg = {
3457 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3458 		.rpc_argp = &args,
3459 		.rpc_resp = &res,
3460 	};
3461 	int status;
3462 
3463 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3464 	if (status == 0)
3465 		update_changeattr(dir, &res.cinfo);
3466 	return status;
3467 }
3468 
3469 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3470 {
3471 	struct nfs4_exception exception = { };
3472 	int err;
3473 	do {
3474 		err = _nfs4_proc_remove(dir, name);
3475 		trace_nfs4_remove(dir, name, err);
3476 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3477 				&exception);
3478 	} while (exception.retry);
3479 	return err;
3480 }
3481 
3482 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3483 {
3484 	struct nfs_server *server = NFS_SERVER(dir);
3485 	struct nfs_removeargs *args = msg->rpc_argp;
3486 	struct nfs_removeres *res = msg->rpc_resp;
3487 
3488 	res->server = server;
3489 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3490 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3491 
3492 	nfs_fattr_init(res->dir_attr);
3493 }
3494 
3495 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3496 {
3497 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3498 			&data->args.seq_args,
3499 			&data->res.seq_res,
3500 			task);
3501 }
3502 
3503 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3504 {
3505 	struct nfs_unlinkdata *data = task->tk_calldata;
3506 	struct nfs_removeres *res = &data->res;
3507 
3508 	if (!nfs4_sequence_done(task, &res->seq_res))
3509 		return 0;
3510 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3511 		return 0;
3512 	update_changeattr(dir, &res->cinfo);
3513 	return 1;
3514 }
3515 
3516 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3517 {
3518 	struct nfs_server *server = NFS_SERVER(dir);
3519 	struct nfs_renameargs *arg = msg->rpc_argp;
3520 	struct nfs_renameres *res = msg->rpc_resp;
3521 
3522 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3523 	res->server = server;
3524 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3525 }
3526 
3527 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3528 {
3529 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3530 			&data->args.seq_args,
3531 			&data->res.seq_res,
3532 			task);
3533 }
3534 
3535 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3536 				 struct inode *new_dir)
3537 {
3538 	struct nfs_renamedata *data = task->tk_calldata;
3539 	struct nfs_renameres *res = &data->res;
3540 
3541 	if (!nfs4_sequence_done(task, &res->seq_res))
3542 		return 0;
3543 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3544 		return 0;
3545 
3546 	update_changeattr(old_dir, &res->old_cinfo);
3547 	update_changeattr(new_dir, &res->new_cinfo);
3548 	return 1;
3549 }
3550 
3551 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3552 {
3553 	struct nfs_server *server = NFS_SERVER(inode);
3554 	struct nfs4_link_arg arg = {
3555 		.fh     = NFS_FH(inode),
3556 		.dir_fh = NFS_FH(dir),
3557 		.name   = name,
3558 		.bitmask = server->attr_bitmask,
3559 	};
3560 	struct nfs4_link_res res = {
3561 		.server = server,
3562 		.label = NULL,
3563 	};
3564 	struct rpc_message msg = {
3565 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3566 		.rpc_argp = &arg,
3567 		.rpc_resp = &res,
3568 	};
3569 	int status = -ENOMEM;
3570 
3571 	res.fattr = nfs_alloc_fattr();
3572 	if (res.fattr == NULL)
3573 		goto out;
3574 
3575 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3576 	if (IS_ERR(res.label)) {
3577 		status = PTR_ERR(res.label);
3578 		goto out;
3579 	}
3580 	arg.bitmask = nfs4_bitmask(server, res.label);
3581 
3582 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3583 	if (!status) {
3584 		update_changeattr(dir, &res.cinfo);
3585 		status = nfs_post_op_update_inode(inode, res.fattr);
3586 		if (!status)
3587 			nfs_setsecurity(inode, res.fattr, res.label);
3588 	}
3589 
3590 
3591 	nfs4_label_free(res.label);
3592 
3593 out:
3594 	nfs_free_fattr(res.fattr);
3595 	return status;
3596 }
3597 
3598 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3599 {
3600 	struct nfs4_exception exception = { };
3601 	int err;
3602 	do {
3603 		err = nfs4_handle_exception(NFS_SERVER(inode),
3604 				_nfs4_proc_link(inode, dir, name),
3605 				&exception);
3606 	} while (exception.retry);
3607 	return err;
3608 }
3609 
3610 struct nfs4_createdata {
3611 	struct rpc_message msg;
3612 	struct nfs4_create_arg arg;
3613 	struct nfs4_create_res res;
3614 	struct nfs_fh fh;
3615 	struct nfs_fattr fattr;
3616 	struct nfs4_label *label;
3617 };
3618 
3619 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3620 		struct qstr *name, struct iattr *sattr, u32 ftype)
3621 {
3622 	struct nfs4_createdata *data;
3623 
3624 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3625 	if (data != NULL) {
3626 		struct nfs_server *server = NFS_SERVER(dir);
3627 
3628 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3629 		if (IS_ERR(data->label))
3630 			goto out_free;
3631 
3632 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3633 		data->msg.rpc_argp = &data->arg;
3634 		data->msg.rpc_resp = &data->res;
3635 		data->arg.dir_fh = NFS_FH(dir);
3636 		data->arg.server = server;
3637 		data->arg.name = name;
3638 		data->arg.attrs = sattr;
3639 		data->arg.ftype = ftype;
3640 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3641 		data->res.server = server;
3642 		data->res.fh = &data->fh;
3643 		data->res.fattr = &data->fattr;
3644 		data->res.label = data->label;
3645 		nfs_fattr_init(data->res.fattr);
3646 	}
3647 	return data;
3648 out_free:
3649 	kfree(data);
3650 	return NULL;
3651 }
3652 
3653 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3654 {
3655 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3656 				    &data->arg.seq_args, &data->res.seq_res, 1);
3657 	if (status == 0) {
3658 		update_changeattr(dir, &data->res.dir_cinfo);
3659 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3660 	}
3661 	return status;
3662 }
3663 
3664 static void nfs4_free_createdata(struct nfs4_createdata *data)
3665 {
3666 	nfs4_label_free(data->label);
3667 	kfree(data);
3668 }
3669 
3670 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3671 		struct page *page, unsigned int len, struct iattr *sattr,
3672 		struct nfs4_label *label)
3673 {
3674 	struct nfs4_createdata *data;
3675 	int status = -ENAMETOOLONG;
3676 
3677 	if (len > NFS4_MAXPATHLEN)
3678 		goto out;
3679 
3680 	status = -ENOMEM;
3681 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3682 	if (data == NULL)
3683 		goto out;
3684 
3685 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3686 	data->arg.u.symlink.pages = &page;
3687 	data->arg.u.symlink.len = len;
3688 	data->arg.label = label;
3689 
3690 	status = nfs4_do_create(dir, dentry, data);
3691 
3692 	nfs4_free_createdata(data);
3693 out:
3694 	return status;
3695 }
3696 
3697 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3698 		struct page *page, unsigned int len, struct iattr *sattr)
3699 {
3700 	struct nfs4_exception exception = { };
3701 	struct nfs4_label l, *label = NULL;
3702 	int err;
3703 
3704 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3705 
3706 	do {
3707 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3708 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3709 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3710 				&exception);
3711 	} while (exception.retry);
3712 
3713 	nfs4_label_release_security(label);
3714 	return err;
3715 }
3716 
3717 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3718 		struct iattr *sattr, struct nfs4_label *label)
3719 {
3720 	struct nfs4_createdata *data;
3721 	int status = -ENOMEM;
3722 
3723 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3724 	if (data == NULL)
3725 		goto out;
3726 
3727 	data->arg.label = label;
3728 	status = nfs4_do_create(dir, dentry, data);
3729 
3730 	nfs4_free_createdata(data);
3731 out:
3732 	return status;
3733 }
3734 
3735 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3736 		struct iattr *sattr)
3737 {
3738 	struct nfs4_exception exception = { };
3739 	struct nfs4_label l, *label = NULL;
3740 	int err;
3741 
3742 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3743 
3744 	sattr->ia_mode &= ~current_umask();
3745 	do {
3746 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3747 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3748 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3749 				&exception);
3750 	} while (exception.retry);
3751 	nfs4_label_release_security(label);
3752 
3753 	return err;
3754 }
3755 
3756 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3757 		u64 cookie, struct page **pages, unsigned int count, int plus)
3758 {
3759 	struct inode		*dir = dentry->d_inode;
3760 	struct nfs4_readdir_arg args = {
3761 		.fh = NFS_FH(dir),
3762 		.pages = pages,
3763 		.pgbase = 0,
3764 		.count = count,
3765 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3766 		.plus = plus,
3767 	};
3768 	struct nfs4_readdir_res res;
3769 	struct rpc_message msg = {
3770 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3771 		.rpc_argp = &args,
3772 		.rpc_resp = &res,
3773 		.rpc_cred = cred,
3774 	};
3775 	int			status;
3776 
3777 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3778 			dentry,
3779 			(unsigned long long)cookie);
3780 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3781 	res.pgbase = args.pgbase;
3782 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3783 	if (status >= 0) {
3784 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3785 		status += args.pgbase;
3786 	}
3787 
3788 	nfs_invalidate_atime(dir);
3789 
3790 	dprintk("%s: returns %d\n", __func__, status);
3791 	return status;
3792 }
3793 
3794 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3795 		u64 cookie, struct page **pages, unsigned int count, int plus)
3796 {
3797 	struct nfs4_exception exception = { };
3798 	int err;
3799 	do {
3800 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3801 				pages, count, plus);
3802 		trace_nfs4_readdir(dentry->d_inode, err);
3803 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3804 				&exception);
3805 	} while (exception.retry);
3806 	return err;
3807 }
3808 
3809 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3810 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3811 {
3812 	struct nfs4_createdata *data;
3813 	int mode = sattr->ia_mode;
3814 	int status = -ENOMEM;
3815 
3816 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3817 	if (data == NULL)
3818 		goto out;
3819 
3820 	if (S_ISFIFO(mode))
3821 		data->arg.ftype = NF4FIFO;
3822 	else if (S_ISBLK(mode)) {
3823 		data->arg.ftype = NF4BLK;
3824 		data->arg.u.device.specdata1 = MAJOR(rdev);
3825 		data->arg.u.device.specdata2 = MINOR(rdev);
3826 	}
3827 	else if (S_ISCHR(mode)) {
3828 		data->arg.ftype = NF4CHR;
3829 		data->arg.u.device.specdata1 = MAJOR(rdev);
3830 		data->arg.u.device.specdata2 = MINOR(rdev);
3831 	} else if (!S_ISSOCK(mode)) {
3832 		status = -EINVAL;
3833 		goto out_free;
3834 	}
3835 
3836 	data->arg.label = label;
3837 	status = nfs4_do_create(dir, dentry, data);
3838 out_free:
3839 	nfs4_free_createdata(data);
3840 out:
3841 	return status;
3842 }
3843 
3844 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3845 		struct iattr *sattr, dev_t rdev)
3846 {
3847 	struct nfs4_exception exception = { };
3848 	struct nfs4_label l, *label = NULL;
3849 	int err;
3850 
3851 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3852 
3853 	sattr->ia_mode &= ~current_umask();
3854 	do {
3855 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3856 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3857 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3858 				&exception);
3859 	} while (exception.retry);
3860 
3861 	nfs4_label_release_security(label);
3862 
3863 	return err;
3864 }
3865 
3866 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3867 		 struct nfs_fsstat *fsstat)
3868 {
3869 	struct nfs4_statfs_arg args = {
3870 		.fh = fhandle,
3871 		.bitmask = server->attr_bitmask,
3872 	};
3873 	struct nfs4_statfs_res res = {
3874 		.fsstat = fsstat,
3875 	};
3876 	struct rpc_message msg = {
3877 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3878 		.rpc_argp = &args,
3879 		.rpc_resp = &res,
3880 	};
3881 
3882 	nfs_fattr_init(fsstat->fattr);
3883 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3884 }
3885 
3886 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3887 {
3888 	struct nfs4_exception exception = { };
3889 	int err;
3890 	do {
3891 		err = nfs4_handle_exception(server,
3892 				_nfs4_proc_statfs(server, fhandle, fsstat),
3893 				&exception);
3894 	} while (exception.retry);
3895 	return err;
3896 }
3897 
3898 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3899 		struct nfs_fsinfo *fsinfo)
3900 {
3901 	struct nfs4_fsinfo_arg args = {
3902 		.fh = fhandle,
3903 		.bitmask = server->attr_bitmask,
3904 	};
3905 	struct nfs4_fsinfo_res res = {
3906 		.fsinfo = fsinfo,
3907 	};
3908 	struct rpc_message msg = {
3909 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3910 		.rpc_argp = &args,
3911 		.rpc_resp = &res,
3912 	};
3913 
3914 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3915 }
3916 
3917 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3918 {
3919 	struct nfs4_exception exception = { };
3920 	unsigned long now = jiffies;
3921 	int err;
3922 
3923 	do {
3924 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3925 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
3926 		if (err == 0) {
3927 			struct nfs_client *clp = server->nfs_client;
3928 
3929 			spin_lock(&clp->cl_lock);
3930 			clp->cl_lease_time = fsinfo->lease_time * HZ;
3931 			clp->cl_last_renewal = now;
3932 			spin_unlock(&clp->cl_lock);
3933 			break;
3934 		}
3935 		err = nfs4_handle_exception(server, err, &exception);
3936 	} while (exception.retry);
3937 	return err;
3938 }
3939 
3940 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3941 {
3942 	int error;
3943 
3944 	nfs_fattr_init(fsinfo->fattr);
3945 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3946 	if (error == 0) {
3947 		/* block layout checks this! */
3948 		server->pnfs_blksize = fsinfo->blksize;
3949 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3950 	}
3951 
3952 	return error;
3953 }
3954 
3955 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3956 		struct nfs_pathconf *pathconf)
3957 {
3958 	struct nfs4_pathconf_arg args = {
3959 		.fh = fhandle,
3960 		.bitmask = server->attr_bitmask,
3961 	};
3962 	struct nfs4_pathconf_res res = {
3963 		.pathconf = pathconf,
3964 	};
3965 	struct rpc_message msg = {
3966 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3967 		.rpc_argp = &args,
3968 		.rpc_resp = &res,
3969 	};
3970 
3971 	/* None of the pathconf attributes are mandatory to implement */
3972 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3973 		memset(pathconf, 0, sizeof(*pathconf));
3974 		return 0;
3975 	}
3976 
3977 	nfs_fattr_init(pathconf->fattr);
3978 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3979 }
3980 
3981 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3982 		struct nfs_pathconf *pathconf)
3983 {
3984 	struct nfs4_exception exception = { };
3985 	int err;
3986 
3987 	do {
3988 		err = nfs4_handle_exception(server,
3989 				_nfs4_proc_pathconf(server, fhandle, pathconf),
3990 				&exception);
3991 	} while (exception.retry);
3992 	return err;
3993 }
3994 
3995 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3996 		const struct nfs_open_context *ctx,
3997 		const struct nfs_lock_context *l_ctx,
3998 		fmode_t fmode)
3999 {
4000 	const struct nfs_lockowner *lockowner = NULL;
4001 
4002 	if (l_ctx != NULL)
4003 		lockowner = &l_ctx->lockowner;
4004 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4005 }
4006 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4007 
4008 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4009 		const struct nfs_open_context *ctx,
4010 		const struct nfs_lock_context *l_ctx,
4011 		fmode_t fmode)
4012 {
4013 	nfs4_stateid current_stateid;
4014 
4015 	/* If the current stateid represents a lost lock, then exit */
4016 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4017 		return true;
4018 	return nfs4_stateid_match(stateid, &current_stateid);
4019 }
4020 
4021 static bool nfs4_error_stateid_expired(int err)
4022 {
4023 	switch (err) {
4024 	case -NFS4ERR_DELEG_REVOKED:
4025 	case -NFS4ERR_ADMIN_REVOKED:
4026 	case -NFS4ERR_BAD_STATEID:
4027 	case -NFS4ERR_STALE_STATEID:
4028 	case -NFS4ERR_OLD_STATEID:
4029 	case -NFS4ERR_OPENMODE:
4030 	case -NFS4ERR_EXPIRED:
4031 		return true;
4032 	}
4033 	return false;
4034 }
4035 
4036 void __nfs4_read_done_cb(struct nfs_pgio_data *data)
4037 {
4038 	nfs_invalidate_atime(data->header->inode);
4039 }
4040 
4041 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_data *data)
4042 {
4043 	struct nfs_server *server = NFS_SERVER(data->header->inode);
4044 
4045 	trace_nfs4_read(data, task->tk_status);
4046 	if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
4047 		rpc_restart_call_prepare(task);
4048 		return -EAGAIN;
4049 	}
4050 
4051 	__nfs4_read_done_cb(data);
4052 	if (task->tk_status > 0)
4053 		renew_lease(server, data->timestamp);
4054 	return 0;
4055 }
4056 
4057 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4058 		struct nfs_pgio_args *args)
4059 {
4060 
4061 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4062 		nfs4_stateid_is_current(&args->stateid,
4063 				args->context,
4064 				args->lock_context,
4065 				FMODE_READ))
4066 		return false;
4067 	rpc_restart_call_prepare(task);
4068 	return true;
4069 }
4070 
4071 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_data *data)
4072 {
4073 
4074 	dprintk("--> %s\n", __func__);
4075 
4076 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4077 		return -EAGAIN;
4078 	if (nfs4_read_stateid_changed(task, &data->args))
4079 		return -EAGAIN;
4080 	return data->pgio_done_cb ? data->pgio_done_cb(task, data) :
4081 				    nfs4_read_done_cb(task, data);
4082 }
4083 
4084 static void nfs4_proc_read_setup(struct nfs_pgio_data *data, struct rpc_message *msg)
4085 {
4086 	data->timestamp   = jiffies;
4087 	data->pgio_done_cb = nfs4_read_done_cb;
4088 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4089 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
4090 }
4091 
4092 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task, struct nfs_pgio_data *data)
4093 {
4094 	if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
4095 			&data->args.seq_args,
4096 			&data->res.seq_res,
4097 			task))
4098 		return 0;
4099 	if (nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
4100 				data->args.lock_context, data->header->rw_ops->rw_mode) == -EIO)
4101 		return -EIO;
4102 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &data->args.context->flags)))
4103 		return -EIO;
4104 	return 0;
4105 }
4106 
4107 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_pgio_data *data)
4108 {
4109 	struct inode *inode = data->header->inode;
4110 
4111 	trace_nfs4_write(data, task->tk_status);
4112 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
4113 		rpc_restart_call_prepare(task);
4114 		return -EAGAIN;
4115 	}
4116 	if (task->tk_status >= 0) {
4117 		renew_lease(NFS_SERVER(inode), data->timestamp);
4118 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4119 	}
4120 	return 0;
4121 }
4122 
4123 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4124 		struct nfs_pgio_args *args)
4125 {
4126 
4127 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4128 		nfs4_stateid_is_current(&args->stateid,
4129 				args->context,
4130 				args->lock_context,
4131 				FMODE_WRITE))
4132 		return false;
4133 	rpc_restart_call_prepare(task);
4134 	return true;
4135 }
4136 
4137 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_data *data)
4138 {
4139 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4140 		return -EAGAIN;
4141 	if (nfs4_write_stateid_changed(task, &data->args))
4142 		return -EAGAIN;
4143 	return data->pgio_done_cb ? data->pgio_done_cb(task, data) :
4144 		nfs4_write_done_cb(task, data);
4145 }
4146 
4147 static
4148 bool nfs4_write_need_cache_consistency_data(const struct nfs_pgio_data *data)
4149 {
4150 	const struct nfs_pgio_header *hdr = data->header;
4151 
4152 	/* Don't request attributes for pNFS or O_DIRECT writes */
4153 	if (data->ds_clp != NULL || hdr->dreq != NULL)
4154 		return false;
4155 	/* Otherwise, request attributes if and only if we don't hold
4156 	 * a delegation
4157 	 */
4158 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4159 }
4160 
4161 static void nfs4_proc_write_setup(struct nfs_pgio_data *data, struct rpc_message *msg)
4162 {
4163 	struct nfs_server *server = NFS_SERVER(data->header->inode);
4164 
4165 	if (!nfs4_write_need_cache_consistency_data(data)) {
4166 		data->args.bitmask = NULL;
4167 		data->res.fattr = NULL;
4168 	} else
4169 		data->args.bitmask = server->cache_consistency_bitmask;
4170 
4171 	if (!data->pgio_done_cb)
4172 		data->pgio_done_cb = nfs4_write_done_cb;
4173 	data->res.server = server;
4174 	data->timestamp   = jiffies;
4175 
4176 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4177 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4178 }
4179 
4180 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4181 {
4182 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4183 			&data->args.seq_args,
4184 			&data->res.seq_res,
4185 			task);
4186 }
4187 
4188 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4189 {
4190 	struct inode *inode = data->inode;
4191 
4192 	trace_nfs4_commit(data, task->tk_status);
4193 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4194 		rpc_restart_call_prepare(task);
4195 		return -EAGAIN;
4196 	}
4197 	return 0;
4198 }
4199 
4200 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4201 {
4202 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4203 		return -EAGAIN;
4204 	return data->commit_done_cb(task, data);
4205 }
4206 
4207 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4208 {
4209 	struct nfs_server *server = NFS_SERVER(data->inode);
4210 
4211 	if (data->commit_done_cb == NULL)
4212 		data->commit_done_cb = nfs4_commit_done_cb;
4213 	data->res.server = server;
4214 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4215 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4216 }
4217 
4218 struct nfs4_renewdata {
4219 	struct nfs_client	*client;
4220 	unsigned long		timestamp;
4221 };
4222 
4223 /*
4224  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4225  * standalone procedure for queueing an asynchronous RENEW.
4226  */
4227 static void nfs4_renew_release(void *calldata)
4228 {
4229 	struct nfs4_renewdata *data = calldata;
4230 	struct nfs_client *clp = data->client;
4231 
4232 	if (atomic_read(&clp->cl_count) > 1)
4233 		nfs4_schedule_state_renewal(clp);
4234 	nfs_put_client(clp);
4235 	kfree(data);
4236 }
4237 
4238 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4239 {
4240 	struct nfs4_renewdata *data = calldata;
4241 	struct nfs_client *clp = data->client;
4242 	unsigned long timestamp = data->timestamp;
4243 
4244 	trace_nfs4_renew_async(clp, task->tk_status);
4245 	switch (task->tk_status) {
4246 	case 0:
4247 		break;
4248 	case -NFS4ERR_LEASE_MOVED:
4249 		nfs4_schedule_lease_moved_recovery(clp);
4250 		break;
4251 	default:
4252 		/* Unless we're shutting down, schedule state recovery! */
4253 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4254 			return;
4255 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4256 			nfs4_schedule_lease_recovery(clp);
4257 			return;
4258 		}
4259 		nfs4_schedule_path_down_recovery(clp);
4260 	}
4261 	do_renew_lease(clp, timestamp);
4262 }
4263 
4264 static const struct rpc_call_ops nfs4_renew_ops = {
4265 	.rpc_call_done = nfs4_renew_done,
4266 	.rpc_release = nfs4_renew_release,
4267 };
4268 
4269 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4270 {
4271 	struct rpc_message msg = {
4272 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4273 		.rpc_argp	= clp,
4274 		.rpc_cred	= cred,
4275 	};
4276 	struct nfs4_renewdata *data;
4277 
4278 	if (renew_flags == 0)
4279 		return 0;
4280 	if (!atomic_inc_not_zero(&clp->cl_count))
4281 		return -EIO;
4282 	data = kmalloc(sizeof(*data), GFP_NOFS);
4283 	if (data == NULL)
4284 		return -ENOMEM;
4285 	data->client = clp;
4286 	data->timestamp = jiffies;
4287 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4288 			&nfs4_renew_ops, data);
4289 }
4290 
4291 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4292 {
4293 	struct rpc_message msg = {
4294 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4295 		.rpc_argp	= clp,
4296 		.rpc_cred	= cred,
4297 	};
4298 	unsigned long now = jiffies;
4299 	int status;
4300 
4301 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4302 	if (status < 0)
4303 		return status;
4304 	do_renew_lease(clp, now);
4305 	return 0;
4306 }
4307 
4308 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4309 {
4310 	return server->caps & NFS_CAP_ACLS;
4311 }
4312 
4313 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4314  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4315  * the stack.
4316  */
4317 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4318 
4319 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4320 		struct page **pages, unsigned int *pgbase)
4321 {
4322 	struct page *newpage, **spages;
4323 	int rc = 0;
4324 	size_t len;
4325 	spages = pages;
4326 
4327 	do {
4328 		len = min_t(size_t, PAGE_SIZE, buflen);
4329 		newpage = alloc_page(GFP_KERNEL);
4330 
4331 		if (newpage == NULL)
4332 			goto unwind;
4333 		memcpy(page_address(newpage), buf, len);
4334                 buf += len;
4335                 buflen -= len;
4336 		*pages++ = newpage;
4337 		rc++;
4338 	} while (buflen != 0);
4339 
4340 	return rc;
4341 
4342 unwind:
4343 	for(; rc > 0; rc--)
4344 		__free_page(spages[rc-1]);
4345 	return -ENOMEM;
4346 }
4347 
4348 struct nfs4_cached_acl {
4349 	int cached;
4350 	size_t len;
4351 	char data[0];
4352 };
4353 
4354 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4355 {
4356 	struct nfs_inode *nfsi = NFS_I(inode);
4357 
4358 	spin_lock(&inode->i_lock);
4359 	kfree(nfsi->nfs4_acl);
4360 	nfsi->nfs4_acl = acl;
4361 	spin_unlock(&inode->i_lock);
4362 }
4363 
4364 static void nfs4_zap_acl_attr(struct inode *inode)
4365 {
4366 	nfs4_set_cached_acl(inode, NULL);
4367 }
4368 
4369 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4370 {
4371 	struct nfs_inode *nfsi = NFS_I(inode);
4372 	struct nfs4_cached_acl *acl;
4373 	int ret = -ENOENT;
4374 
4375 	spin_lock(&inode->i_lock);
4376 	acl = nfsi->nfs4_acl;
4377 	if (acl == NULL)
4378 		goto out;
4379 	if (buf == NULL) /* user is just asking for length */
4380 		goto out_len;
4381 	if (acl->cached == 0)
4382 		goto out;
4383 	ret = -ERANGE; /* see getxattr(2) man page */
4384 	if (acl->len > buflen)
4385 		goto out;
4386 	memcpy(buf, acl->data, acl->len);
4387 out_len:
4388 	ret = acl->len;
4389 out:
4390 	spin_unlock(&inode->i_lock);
4391 	return ret;
4392 }
4393 
4394 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4395 {
4396 	struct nfs4_cached_acl *acl;
4397 	size_t buflen = sizeof(*acl) + acl_len;
4398 
4399 	if (buflen <= PAGE_SIZE) {
4400 		acl = kmalloc(buflen, GFP_KERNEL);
4401 		if (acl == NULL)
4402 			goto out;
4403 		acl->cached = 1;
4404 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4405 	} else {
4406 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4407 		if (acl == NULL)
4408 			goto out;
4409 		acl->cached = 0;
4410 	}
4411 	acl->len = acl_len;
4412 out:
4413 	nfs4_set_cached_acl(inode, acl);
4414 }
4415 
4416 /*
4417  * The getxattr API returns the required buffer length when called with a
4418  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4419  * the required buf.  On a NULL buf, we send a page of data to the server
4420  * guessing that the ACL request can be serviced by a page. If so, we cache
4421  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4422  * the cache. If not so, we throw away the page, and cache the required
4423  * length. The next getxattr call will then produce another round trip to
4424  * the server, this time with the input buf of the required size.
4425  */
4426 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4427 {
4428 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4429 	struct nfs_getaclargs args = {
4430 		.fh = NFS_FH(inode),
4431 		.acl_pages = pages,
4432 		.acl_len = buflen,
4433 	};
4434 	struct nfs_getaclres res = {
4435 		.acl_len = buflen,
4436 	};
4437 	struct rpc_message msg = {
4438 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4439 		.rpc_argp = &args,
4440 		.rpc_resp = &res,
4441 	};
4442 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4443 	int ret = -ENOMEM, i;
4444 
4445 	/* As long as we're doing a round trip to the server anyway,
4446 	 * let's be prepared for a page of acl data. */
4447 	if (npages == 0)
4448 		npages = 1;
4449 	if (npages > ARRAY_SIZE(pages))
4450 		return -ERANGE;
4451 
4452 	for (i = 0; i < npages; i++) {
4453 		pages[i] = alloc_page(GFP_KERNEL);
4454 		if (!pages[i])
4455 			goto out_free;
4456 	}
4457 
4458 	/* for decoding across pages */
4459 	res.acl_scratch = alloc_page(GFP_KERNEL);
4460 	if (!res.acl_scratch)
4461 		goto out_free;
4462 
4463 	args.acl_len = npages * PAGE_SIZE;
4464 	args.acl_pgbase = 0;
4465 
4466 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4467 		__func__, buf, buflen, npages, args.acl_len);
4468 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4469 			     &msg, &args.seq_args, &res.seq_res, 0);
4470 	if (ret)
4471 		goto out_free;
4472 
4473 	/* Handle the case where the passed-in buffer is too short */
4474 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4475 		/* Did the user only issue a request for the acl length? */
4476 		if (buf == NULL)
4477 			goto out_ok;
4478 		ret = -ERANGE;
4479 		goto out_free;
4480 	}
4481 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4482 	if (buf) {
4483 		if (res.acl_len > buflen) {
4484 			ret = -ERANGE;
4485 			goto out_free;
4486 		}
4487 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4488 	}
4489 out_ok:
4490 	ret = res.acl_len;
4491 out_free:
4492 	for (i = 0; i < npages; i++)
4493 		if (pages[i])
4494 			__free_page(pages[i]);
4495 	if (res.acl_scratch)
4496 		__free_page(res.acl_scratch);
4497 	return ret;
4498 }
4499 
4500 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4501 {
4502 	struct nfs4_exception exception = { };
4503 	ssize_t ret;
4504 	do {
4505 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4506 		trace_nfs4_get_acl(inode, ret);
4507 		if (ret >= 0)
4508 			break;
4509 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4510 	} while (exception.retry);
4511 	return ret;
4512 }
4513 
4514 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4515 {
4516 	struct nfs_server *server = NFS_SERVER(inode);
4517 	int ret;
4518 
4519 	if (!nfs4_server_supports_acls(server))
4520 		return -EOPNOTSUPP;
4521 	ret = nfs_revalidate_inode(server, inode);
4522 	if (ret < 0)
4523 		return ret;
4524 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4525 		nfs_zap_acl_cache(inode);
4526 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4527 	if (ret != -ENOENT)
4528 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4529 		 * but no cached acl data, just the acl length */
4530 		return ret;
4531 	return nfs4_get_acl_uncached(inode, buf, buflen);
4532 }
4533 
4534 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4535 {
4536 	struct nfs_server *server = NFS_SERVER(inode);
4537 	struct page *pages[NFS4ACL_MAXPAGES];
4538 	struct nfs_setaclargs arg = {
4539 		.fh		= NFS_FH(inode),
4540 		.acl_pages	= pages,
4541 		.acl_len	= buflen,
4542 	};
4543 	struct nfs_setaclres res;
4544 	struct rpc_message msg = {
4545 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4546 		.rpc_argp	= &arg,
4547 		.rpc_resp	= &res,
4548 	};
4549 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4550 	int ret, i;
4551 
4552 	if (!nfs4_server_supports_acls(server))
4553 		return -EOPNOTSUPP;
4554 	if (npages > ARRAY_SIZE(pages))
4555 		return -ERANGE;
4556 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4557 	if (i < 0)
4558 		return i;
4559 	nfs4_inode_return_delegation(inode);
4560 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4561 
4562 	/*
4563 	 * Free each page after tx, so the only ref left is
4564 	 * held by the network stack
4565 	 */
4566 	for (; i > 0; i--)
4567 		put_page(pages[i-1]);
4568 
4569 	/*
4570 	 * Acl update can result in inode attribute update.
4571 	 * so mark the attribute cache invalid.
4572 	 */
4573 	spin_lock(&inode->i_lock);
4574 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4575 	spin_unlock(&inode->i_lock);
4576 	nfs_access_zap_cache(inode);
4577 	nfs_zap_acl_cache(inode);
4578 	return ret;
4579 }
4580 
4581 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4582 {
4583 	struct nfs4_exception exception = { };
4584 	int err;
4585 	do {
4586 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4587 		trace_nfs4_set_acl(inode, err);
4588 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4589 				&exception);
4590 	} while (exception.retry);
4591 	return err;
4592 }
4593 
4594 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4595 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4596 					size_t buflen)
4597 {
4598 	struct nfs_server *server = NFS_SERVER(inode);
4599 	struct nfs_fattr fattr;
4600 	struct nfs4_label label = {0, 0, buflen, buf};
4601 
4602 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4603 	struct nfs4_getattr_arg arg = {
4604 		.fh		= NFS_FH(inode),
4605 		.bitmask	= bitmask,
4606 	};
4607 	struct nfs4_getattr_res res = {
4608 		.fattr		= &fattr,
4609 		.label		= &label,
4610 		.server		= server,
4611 	};
4612 	struct rpc_message msg = {
4613 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4614 		.rpc_argp	= &arg,
4615 		.rpc_resp	= &res,
4616 	};
4617 	int ret;
4618 
4619 	nfs_fattr_init(&fattr);
4620 
4621 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4622 	if (ret)
4623 		return ret;
4624 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4625 		return -ENOENT;
4626 	if (buflen < label.len)
4627 		return -ERANGE;
4628 	return 0;
4629 }
4630 
4631 static int nfs4_get_security_label(struct inode *inode, void *buf,
4632 					size_t buflen)
4633 {
4634 	struct nfs4_exception exception = { };
4635 	int err;
4636 
4637 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4638 		return -EOPNOTSUPP;
4639 
4640 	do {
4641 		err = _nfs4_get_security_label(inode, buf, buflen);
4642 		trace_nfs4_get_security_label(inode, err);
4643 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4644 				&exception);
4645 	} while (exception.retry);
4646 	return err;
4647 }
4648 
4649 static int _nfs4_do_set_security_label(struct inode *inode,
4650 		struct nfs4_label *ilabel,
4651 		struct nfs_fattr *fattr,
4652 		struct nfs4_label *olabel)
4653 {
4654 
4655 	struct iattr sattr = {0};
4656 	struct nfs_server *server = NFS_SERVER(inode);
4657 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4658 	struct nfs_setattrargs arg = {
4659 		.fh             = NFS_FH(inode),
4660 		.iap            = &sattr,
4661 		.server		= server,
4662 		.bitmask	= bitmask,
4663 		.label		= ilabel,
4664 	};
4665 	struct nfs_setattrres res = {
4666 		.fattr		= fattr,
4667 		.label		= olabel,
4668 		.server		= server,
4669 	};
4670 	struct rpc_message msg = {
4671 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4672 		.rpc_argp       = &arg,
4673 		.rpc_resp       = &res,
4674 	};
4675 	int status;
4676 
4677 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4678 
4679 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4680 	if (status)
4681 		dprintk("%s failed: %d\n", __func__, status);
4682 
4683 	return status;
4684 }
4685 
4686 static int nfs4_do_set_security_label(struct inode *inode,
4687 		struct nfs4_label *ilabel,
4688 		struct nfs_fattr *fattr,
4689 		struct nfs4_label *olabel)
4690 {
4691 	struct nfs4_exception exception = { };
4692 	int err;
4693 
4694 	do {
4695 		err = _nfs4_do_set_security_label(inode, ilabel,
4696 				fattr, olabel);
4697 		trace_nfs4_set_security_label(inode, err);
4698 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4699 				&exception);
4700 	} while (exception.retry);
4701 	return err;
4702 }
4703 
4704 static int
4705 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4706 {
4707 	struct nfs4_label ilabel, *olabel = NULL;
4708 	struct nfs_fattr fattr;
4709 	struct rpc_cred *cred;
4710 	struct inode *inode = dentry->d_inode;
4711 	int status;
4712 
4713 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4714 		return -EOPNOTSUPP;
4715 
4716 	nfs_fattr_init(&fattr);
4717 
4718 	ilabel.pi = 0;
4719 	ilabel.lfs = 0;
4720 	ilabel.label = (char *)buf;
4721 	ilabel.len = buflen;
4722 
4723 	cred = rpc_lookup_cred();
4724 	if (IS_ERR(cred))
4725 		return PTR_ERR(cred);
4726 
4727 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4728 	if (IS_ERR(olabel)) {
4729 		status = -PTR_ERR(olabel);
4730 		goto out;
4731 	}
4732 
4733 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4734 	if (status == 0)
4735 		nfs_setsecurity(inode, &fattr, olabel);
4736 
4737 	nfs4_label_free(olabel);
4738 out:
4739 	put_rpccred(cred);
4740 	return status;
4741 }
4742 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4743 
4744 
4745 static int
4746 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4747 {
4748 	struct nfs_client *clp = server->nfs_client;
4749 
4750 	if (task->tk_status >= 0)
4751 		return 0;
4752 	switch(task->tk_status) {
4753 		case -NFS4ERR_DELEG_REVOKED:
4754 		case -NFS4ERR_ADMIN_REVOKED:
4755 		case -NFS4ERR_BAD_STATEID:
4756 			if (state == NULL)
4757 				break;
4758 			nfs_remove_bad_delegation(state->inode);
4759 		case -NFS4ERR_OPENMODE:
4760 			if (state == NULL)
4761 				break;
4762 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4763 				goto recovery_failed;
4764 			goto wait_on_recovery;
4765 		case -NFS4ERR_EXPIRED:
4766 			if (state != NULL) {
4767 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4768 					goto recovery_failed;
4769 			}
4770 		case -NFS4ERR_STALE_STATEID:
4771 		case -NFS4ERR_STALE_CLIENTID:
4772 			nfs4_schedule_lease_recovery(clp);
4773 			goto wait_on_recovery;
4774 		case -NFS4ERR_MOVED:
4775 			if (nfs4_schedule_migration_recovery(server) < 0)
4776 				goto recovery_failed;
4777 			goto wait_on_recovery;
4778 		case -NFS4ERR_LEASE_MOVED:
4779 			nfs4_schedule_lease_moved_recovery(clp);
4780 			goto wait_on_recovery;
4781 #if defined(CONFIG_NFS_V4_1)
4782 		case -NFS4ERR_BADSESSION:
4783 		case -NFS4ERR_BADSLOT:
4784 		case -NFS4ERR_BAD_HIGH_SLOT:
4785 		case -NFS4ERR_DEADSESSION:
4786 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4787 		case -NFS4ERR_SEQ_FALSE_RETRY:
4788 		case -NFS4ERR_SEQ_MISORDERED:
4789 			dprintk("%s ERROR %d, Reset session\n", __func__,
4790 				task->tk_status);
4791 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4792 			goto wait_on_recovery;
4793 #endif /* CONFIG_NFS_V4_1 */
4794 		case -NFS4ERR_DELAY:
4795 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4796 		case -NFS4ERR_GRACE:
4797 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4798 		case -NFS4ERR_RETRY_UNCACHED_REP:
4799 		case -NFS4ERR_OLD_STATEID:
4800 			goto restart_call;
4801 	}
4802 	task->tk_status = nfs4_map_errors(task->tk_status);
4803 	return 0;
4804 recovery_failed:
4805 	task->tk_status = -EIO;
4806 	return 0;
4807 wait_on_recovery:
4808 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4809 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4810 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4811 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4812 		goto recovery_failed;
4813 restart_call:
4814 	task->tk_status = 0;
4815 	return -EAGAIN;
4816 }
4817 
4818 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4819 				    nfs4_verifier *bootverf)
4820 {
4821 	__be32 verf[2];
4822 
4823 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4824 		/* An impossible timestamp guarantees this value
4825 		 * will never match a generated boot time. */
4826 		verf[0] = 0;
4827 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4828 	} else {
4829 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4830 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4831 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4832 	}
4833 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4834 }
4835 
4836 static unsigned int
4837 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4838 				   char *buf, size_t len)
4839 {
4840 	unsigned int result;
4841 
4842 	rcu_read_lock();
4843 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4844 				clp->cl_ipaddr,
4845 				rpc_peeraddr2str(clp->cl_rpcclient,
4846 							RPC_DISPLAY_ADDR),
4847 				rpc_peeraddr2str(clp->cl_rpcclient,
4848 							RPC_DISPLAY_PROTO));
4849 	rcu_read_unlock();
4850 	return result;
4851 }
4852 
4853 static unsigned int
4854 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4855 				char *buf, size_t len)
4856 {
4857 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4858 
4859 	if (nfs4_client_id_uniquifier[0] != '\0')
4860 		return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4861 				clp->rpc_ops->version,
4862 				clp->cl_minorversion,
4863 				nfs4_client_id_uniquifier,
4864 				nodename);
4865 	return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4866 				clp->rpc_ops->version, clp->cl_minorversion,
4867 				nodename);
4868 }
4869 
4870 /*
4871  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4872  * services.  Advertise one based on the address family of the
4873  * clientaddr.
4874  */
4875 static unsigned int
4876 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4877 {
4878 	if (strchr(clp->cl_ipaddr, ':') != NULL)
4879 		return scnprintf(buf, len, "tcp6");
4880 	else
4881 		return scnprintf(buf, len, "tcp");
4882 }
4883 
4884 /**
4885  * nfs4_proc_setclientid - Negotiate client ID
4886  * @clp: state data structure
4887  * @program: RPC program for NFSv4 callback service
4888  * @port: IP port number for NFS4 callback service
4889  * @cred: RPC credential to use for this call
4890  * @res: where to place the result
4891  *
4892  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4893  */
4894 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4895 		unsigned short port, struct rpc_cred *cred,
4896 		struct nfs4_setclientid_res *res)
4897 {
4898 	nfs4_verifier sc_verifier;
4899 	struct nfs4_setclientid setclientid = {
4900 		.sc_verifier = &sc_verifier,
4901 		.sc_prog = program,
4902 		.sc_cb_ident = clp->cl_cb_ident,
4903 	};
4904 	struct rpc_message msg = {
4905 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4906 		.rpc_argp = &setclientid,
4907 		.rpc_resp = res,
4908 		.rpc_cred = cred,
4909 	};
4910 	int status;
4911 
4912 	/* nfs_client_id4 */
4913 	nfs4_init_boot_verifier(clp, &sc_verifier);
4914 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4915 		setclientid.sc_name_len =
4916 				nfs4_init_uniform_client_string(clp,
4917 						setclientid.sc_name,
4918 						sizeof(setclientid.sc_name));
4919 	else
4920 		setclientid.sc_name_len =
4921 				nfs4_init_nonuniform_client_string(clp,
4922 						setclientid.sc_name,
4923 						sizeof(setclientid.sc_name));
4924 	/* cb_client4 */
4925 	setclientid.sc_netid_len =
4926 				nfs4_init_callback_netid(clp,
4927 						setclientid.sc_netid,
4928 						sizeof(setclientid.sc_netid));
4929 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4930 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4931 				clp->cl_ipaddr, port >> 8, port & 255);
4932 
4933 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4934 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
4935 		setclientid.sc_name_len, setclientid.sc_name);
4936 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4937 	trace_nfs4_setclientid(clp, status);
4938 	dprintk("NFS reply setclientid: %d\n", status);
4939 	return status;
4940 }
4941 
4942 /**
4943  * nfs4_proc_setclientid_confirm - Confirm client ID
4944  * @clp: state data structure
4945  * @res: result of a previous SETCLIENTID
4946  * @cred: RPC credential to use for this call
4947  *
4948  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4949  */
4950 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4951 		struct nfs4_setclientid_res *arg,
4952 		struct rpc_cred *cred)
4953 {
4954 	struct rpc_message msg = {
4955 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4956 		.rpc_argp = arg,
4957 		.rpc_cred = cred,
4958 	};
4959 	int status;
4960 
4961 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4962 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
4963 		clp->cl_clientid);
4964 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4965 	trace_nfs4_setclientid_confirm(clp, status);
4966 	dprintk("NFS reply setclientid_confirm: %d\n", status);
4967 	return status;
4968 }
4969 
4970 struct nfs4_delegreturndata {
4971 	struct nfs4_delegreturnargs args;
4972 	struct nfs4_delegreturnres res;
4973 	struct nfs_fh fh;
4974 	nfs4_stateid stateid;
4975 	unsigned long timestamp;
4976 	struct nfs_fattr fattr;
4977 	int rpc_status;
4978 };
4979 
4980 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4981 {
4982 	struct nfs4_delegreturndata *data = calldata;
4983 
4984 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4985 		return;
4986 
4987 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
4988 	switch (task->tk_status) {
4989 	case 0:
4990 		renew_lease(data->res.server, data->timestamp);
4991 		break;
4992 	case -NFS4ERR_ADMIN_REVOKED:
4993 	case -NFS4ERR_DELEG_REVOKED:
4994 	case -NFS4ERR_BAD_STATEID:
4995 	case -NFS4ERR_OLD_STATEID:
4996 	case -NFS4ERR_STALE_STATEID:
4997 	case -NFS4ERR_EXPIRED:
4998 		task->tk_status = 0;
4999 		break;
5000 	default:
5001 		if (nfs4_async_handle_error(task, data->res.server, NULL) ==
5002 				-EAGAIN) {
5003 			rpc_restart_call_prepare(task);
5004 			return;
5005 		}
5006 	}
5007 	data->rpc_status = task->tk_status;
5008 }
5009 
5010 static void nfs4_delegreturn_release(void *calldata)
5011 {
5012 	kfree(calldata);
5013 }
5014 
5015 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5016 {
5017 	struct nfs4_delegreturndata *d_data;
5018 
5019 	d_data = (struct nfs4_delegreturndata *)data;
5020 
5021 	nfs4_setup_sequence(d_data->res.server,
5022 			&d_data->args.seq_args,
5023 			&d_data->res.seq_res,
5024 			task);
5025 }
5026 
5027 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5028 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5029 	.rpc_call_done = nfs4_delegreturn_done,
5030 	.rpc_release = nfs4_delegreturn_release,
5031 };
5032 
5033 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5034 {
5035 	struct nfs4_delegreturndata *data;
5036 	struct nfs_server *server = NFS_SERVER(inode);
5037 	struct rpc_task *task;
5038 	struct rpc_message msg = {
5039 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5040 		.rpc_cred = cred,
5041 	};
5042 	struct rpc_task_setup task_setup_data = {
5043 		.rpc_client = server->client,
5044 		.rpc_message = &msg,
5045 		.callback_ops = &nfs4_delegreturn_ops,
5046 		.flags = RPC_TASK_ASYNC,
5047 	};
5048 	int status = 0;
5049 
5050 	data = kzalloc(sizeof(*data), GFP_NOFS);
5051 	if (data == NULL)
5052 		return -ENOMEM;
5053 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5054 	data->args.fhandle = &data->fh;
5055 	data->args.stateid = &data->stateid;
5056 	data->args.bitmask = server->cache_consistency_bitmask;
5057 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5058 	nfs4_stateid_copy(&data->stateid, stateid);
5059 	data->res.fattr = &data->fattr;
5060 	data->res.server = server;
5061 	nfs_fattr_init(data->res.fattr);
5062 	data->timestamp = jiffies;
5063 	data->rpc_status = 0;
5064 
5065 	task_setup_data.callback_data = data;
5066 	msg.rpc_argp = &data->args;
5067 	msg.rpc_resp = &data->res;
5068 	task = rpc_run_task(&task_setup_data);
5069 	if (IS_ERR(task))
5070 		return PTR_ERR(task);
5071 	if (!issync)
5072 		goto out;
5073 	status = nfs4_wait_for_completion_rpc_task(task);
5074 	if (status != 0)
5075 		goto out;
5076 	status = data->rpc_status;
5077 	if (status == 0)
5078 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5079 	else
5080 		nfs_refresh_inode(inode, &data->fattr);
5081 out:
5082 	rpc_put_task(task);
5083 	return status;
5084 }
5085 
5086 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5087 {
5088 	struct nfs_server *server = NFS_SERVER(inode);
5089 	struct nfs4_exception exception = { };
5090 	int err;
5091 	do {
5092 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5093 		trace_nfs4_delegreturn(inode, err);
5094 		switch (err) {
5095 			case -NFS4ERR_STALE_STATEID:
5096 			case -NFS4ERR_EXPIRED:
5097 			case 0:
5098 				return 0;
5099 		}
5100 		err = nfs4_handle_exception(server, err, &exception);
5101 	} while (exception.retry);
5102 	return err;
5103 }
5104 
5105 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5106 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5107 
5108 /*
5109  * sleep, with exponential backoff, and retry the LOCK operation.
5110  */
5111 static unsigned long
5112 nfs4_set_lock_task_retry(unsigned long timeout)
5113 {
5114 	freezable_schedule_timeout_killable_unsafe(timeout);
5115 	timeout <<= 1;
5116 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5117 		return NFS4_LOCK_MAXTIMEOUT;
5118 	return timeout;
5119 }
5120 
5121 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5122 {
5123 	struct inode *inode = state->inode;
5124 	struct nfs_server *server = NFS_SERVER(inode);
5125 	struct nfs_client *clp = server->nfs_client;
5126 	struct nfs_lockt_args arg = {
5127 		.fh = NFS_FH(inode),
5128 		.fl = request,
5129 	};
5130 	struct nfs_lockt_res res = {
5131 		.denied = request,
5132 	};
5133 	struct rpc_message msg = {
5134 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5135 		.rpc_argp       = &arg,
5136 		.rpc_resp       = &res,
5137 		.rpc_cred	= state->owner->so_cred,
5138 	};
5139 	struct nfs4_lock_state *lsp;
5140 	int status;
5141 
5142 	arg.lock_owner.clientid = clp->cl_clientid;
5143 	status = nfs4_set_lock_state(state, request);
5144 	if (status != 0)
5145 		goto out;
5146 	lsp = request->fl_u.nfs4_fl.owner;
5147 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5148 	arg.lock_owner.s_dev = server->s_dev;
5149 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5150 	switch (status) {
5151 		case 0:
5152 			request->fl_type = F_UNLCK;
5153 			break;
5154 		case -NFS4ERR_DENIED:
5155 			status = 0;
5156 	}
5157 	request->fl_ops->fl_release_private(request);
5158 	request->fl_ops = NULL;
5159 out:
5160 	return status;
5161 }
5162 
5163 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5164 {
5165 	struct nfs4_exception exception = { };
5166 	int err;
5167 
5168 	do {
5169 		err = _nfs4_proc_getlk(state, cmd, request);
5170 		trace_nfs4_get_lock(request, state, cmd, err);
5171 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5172 				&exception);
5173 	} while (exception.retry);
5174 	return err;
5175 }
5176 
5177 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5178 {
5179 	int res = 0;
5180 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5181 		case FL_POSIX:
5182 			res = posix_lock_file_wait(file, fl);
5183 			break;
5184 		case FL_FLOCK:
5185 			res = flock_lock_file_wait(file, fl);
5186 			break;
5187 		default:
5188 			BUG();
5189 	}
5190 	return res;
5191 }
5192 
5193 struct nfs4_unlockdata {
5194 	struct nfs_locku_args arg;
5195 	struct nfs_locku_res res;
5196 	struct nfs4_lock_state *lsp;
5197 	struct nfs_open_context *ctx;
5198 	struct file_lock fl;
5199 	const struct nfs_server *server;
5200 	unsigned long timestamp;
5201 };
5202 
5203 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5204 		struct nfs_open_context *ctx,
5205 		struct nfs4_lock_state *lsp,
5206 		struct nfs_seqid *seqid)
5207 {
5208 	struct nfs4_unlockdata *p;
5209 	struct inode *inode = lsp->ls_state->inode;
5210 
5211 	p = kzalloc(sizeof(*p), GFP_NOFS);
5212 	if (p == NULL)
5213 		return NULL;
5214 	p->arg.fh = NFS_FH(inode);
5215 	p->arg.fl = &p->fl;
5216 	p->arg.seqid = seqid;
5217 	p->res.seqid = seqid;
5218 	p->arg.stateid = &lsp->ls_stateid;
5219 	p->lsp = lsp;
5220 	atomic_inc(&lsp->ls_count);
5221 	/* Ensure we don't close file until we're done freeing locks! */
5222 	p->ctx = get_nfs_open_context(ctx);
5223 	memcpy(&p->fl, fl, sizeof(p->fl));
5224 	p->server = NFS_SERVER(inode);
5225 	return p;
5226 }
5227 
5228 static void nfs4_locku_release_calldata(void *data)
5229 {
5230 	struct nfs4_unlockdata *calldata = data;
5231 	nfs_free_seqid(calldata->arg.seqid);
5232 	nfs4_put_lock_state(calldata->lsp);
5233 	put_nfs_open_context(calldata->ctx);
5234 	kfree(calldata);
5235 }
5236 
5237 static void nfs4_locku_done(struct rpc_task *task, void *data)
5238 {
5239 	struct nfs4_unlockdata *calldata = data;
5240 
5241 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5242 		return;
5243 	switch (task->tk_status) {
5244 		case 0:
5245 			nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5246 					&calldata->res.stateid);
5247 			renew_lease(calldata->server, calldata->timestamp);
5248 			break;
5249 		case -NFS4ERR_BAD_STATEID:
5250 		case -NFS4ERR_OLD_STATEID:
5251 		case -NFS4ERR_STALE_STATEID:
5252 		case -NFS4ERR_EXPIRED:
5253 			break;
5254 		default:
5255 			if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5256 				rpc_restart_call_prepare(task);
5257 	}
5258 	nfs_release_seqid(calldata->arg.seqid);
5259 }
5260 
5261 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5262 {
5263 	struct nfs4_unlockdata *calldata = data;
5264 
5265 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5266 		goto out_wait;
5267 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5268 		/* Note: exit _without_ running nfs4_locku_done */
5269 		goto out_no_action;
5270 	}
5271 	calldata->timestamp = jiffies;
5272 	if (nfs4_setup_sequence(calldata->server,
5273 				&calldata->arg.seq_args,
5274 				&calldata->res.seq_res,
5275 				task) != 0)
5276 		nfs_release_seqid(calldata->arg.seqid);
5277 	return;
5278 out_no_action:
5279 	task->tk_action = NULL;
5280 out_wait:
5281 	nfs4_sequence_done(task, &calldata->res.seq_res);
5282 }
5283 
5284 static const struct rpc_call_ops nfs4_locku_ops = {
5285 	.rpc_call_prepare = nfs4_locku_prepare,
5286 	.rpc_call_done = nfs4_locku_done,
5287 	.rpc_release = nfs4_locku_release_calldata,
5288 };
5289 
5290 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5291 		struct nfs_open_context *ctx,
5292 		struct nfs4_lock_state *lsp,
5293 		struct nfs_seqid *seqid)
5294 {
5295 	struct nfs4_unlockdata *data;
5296 	struct rpc_message msg = {
5297 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5298 		.rpc_cred = ctx->cred,
5299 	};
5300 	struct rpc_task_setup task_setup_data = {
5301 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5302 		.rpc_message = &msg,
5303 		.callback_ops = &nfs4_locku_ops,
5304 		.workqueue = nfsiod_workqueue,
5305 		.flags = RPC_TASK_ASYNC,
5306 	};
5307 
5308 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5309 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5310 
5311 	/* Ensure this is an unlock - when canceling a lock, the
5312 	 * canceled lock is passed in, and it won't be an unlock.
5313 	 */
5314 	fl->fl_type = F_UNLCK;
5315 
5316 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5317 	if (data == NULL) {
5318 		nfs_free_seqid(seqid);
5319 		return ERR_PTR(-ENOMEM);
5320 	}
5321 
5322 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5323 	msg.rpc_argp = &data->arg;
5324 	msg.rpc_resp = &data->res;
5325 	task_setup_data.callback_data = data;
5326 	return rpc_run_task(&task_setup_data);
5327 }
5328 
5329 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5330 {
5331 	struct inode *inode = state->inode;
5332 	struct nfs4_state_owner *sp = state->owner;
5333 	struct nfs_inode *nfsi = NFS_I(inode);
5334 	struct nfs_seqid *seqid;
5335 	struct nfs4_lock_state *lsp;
5336 	struct rpc_task *task;
5337 	int status = 0;
5338 	unsigned char fl_flags = request->fl_flags;
5339 
5340 	status = nfs4_set_lock_state(state, request);
5341 	/* Unlock _before_ we do the RPC call */
5342 	request->fl_flags |= FL_EXISTS;
5343 	/* Exclude nfs_delegation_claim_locks() */
5344 	mutex_lock(&sp->so_delegreturn_mutex);
5345 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5346 	down_read(&nfsi->rwsem);
5347 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5348 		up_read(&nfsi->rwsem);
5349 		mutex_unlock(&sp->so_delegreturn_mutex);
5350 		goto out;
5351 	}
5352 	up_read(&nfsi->rwsem);
5353 	mutex_unlock(&sp->so_delegreturn_mutex);
5354 	if (status != 0)
5355 		goto out;
5356 	/* Is this a delegated lock? */
5357 	lsp = request->fl_u.nfs4_fl.owner;
5358 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5359 		goto out;
5360 	seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5361 	status = -ENOMEM;
5362 	if (seqid == NULL)
5363 		goto out;
5364 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5365 	status = PTR_ERR(task);
5366 	if (IS_ERR(task))
5367 		goto out;
5368 	status = nfs4_wait_for_completion_rpc_task(task);
5369 	rpc_put_task(task);
5370 out:
5371 	request->fl_flags = fl_flags;
5372 	trace_nfs4_unlock(request, state, F_SETLK, status);
5373 	return status;
5374 }
5375 
5376 struct nfs4_lockdata {
5377 	struct nfs_lock_args arg;
5378 	struct nfs_lock_res res;
5379 	struct nfs4_lock_state *lsp;
5380 	struct nfs_open_context *ctx;
5381 	struct file_lock fl;
5382 	unsigned long timestamp;
5383 	int rpc_status;
5384 	int cancelled;
5385 	struct nfs_server *server;
5386 };
5387 
5388 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5389 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5390 		gfp_t gfp_mask)
5391 {
5392 	struct nfs4_lockdata *p;
5393 	struct inode *inode = lsp->ls_state->inode;
5394 	struct nfs_server *server = NFS_SERVER(inode);
5395 
5396 	p = kzalloc(sizeof(*p), gfp_mask);
5397 	if (p == NULL)
5398 		return NULL;
5399 
5400 	p->arg.fh = NFS_FH(inode);
5401 	p->arg.fl = &p->fl;
5402 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5403 	if (p->arg.open_seqid == NULL)
5404 		goto out_free;
5405 	p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5406 	if (p->arg.lock_seqid == NULL)
5407 		goto out_free_seqid;
5408 	p->arg.lock_stateid = &lsp->ls_stateid;
5409 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5410 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5411 	p->arg.lock_owner.s_dev = server->s_dev;
5412 	p->res.lock_seqid = p->arg.lock_seqid;
5413 	p->lsp = lsp;
5414 	p->server = server;
5415 	atomic_inc(&lsp->ls_count);
5416 	p->ctx = get_nfs_open_context(ctx);
5417 	memcpy(&p->fl, fl, sizeof(p->fl));
5418 	return p;
5419 out_free_seqid:
5420 	nfs_free_seqid(p->arg.open_seqid);
5421 out_free:
5422 	kfree(p);
5423 	return NULL;
5424 }
5425 
5426 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5427 {
5428 	struct nfs4_lockdata *data = calldata;
5429 	struct nfs4_state *state = data->lsp->ls_state;
5430 
5431 	dprintk("%s: begin!\n", __func__);
5432 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5433 		goto out_wait;
5434 	/* Do we need to do an open_to_lock_owner? */
5435 	if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5436 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5437 			goto out_release_lock_seqid;
5438 		}
5439 		data->arg.open_stateid = &state->open_stateid;
5440 		data->arg.new_lock_owner = 1;
5441 		data->res.open_seqid = data->arg.open_seqid;
5442 	} else
5443 		data->arg.new_lock_owner = 0;
5444 	if (!nfs4_valid_open_stateid(state)) {
5445 		data->rpc_status = -EBADF;
5446 		task->tk_action = NULL;
5447 		goto out_release_open_seqid;
5448 	}
5449 	data->timestamp = jiffies;
5450 	if (nfs4_setup_sequence(data->server,
5451 				&data->arg.seq_args,
5452 				&data->res.seq_res,
5453 				task) == 0)
5454 		return;
5455 out_release_open_seqid:
5456 	nfs_release_seqid(data->arg.open_seqid);
5457 out_release_lock_seqid:
5458 	nfs_release_seqid(data->arg.lock_seqid);
5459 out_wait:
5460 	nfs4_sequence_done(task, &data->res.seq_res);
5461 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5462 }
5463 
5464 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5465 {
5466 	struct nfs4_lockdata *data = calldata;
5467 
5468 	dprintk("%s: begin!\n", __func__);
5469 
5470 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5471 		return;
5472 
5473 	data->rpc_status = task->tk_status;
5474 	if (data->arg.new_lock_owner != 0) {
5475 		if (data->rpc_status == 0)
5476 			nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5477 		else
5478 			goto out;
5479 	}
5480 	if (data->rpc_status == 0) {
5481 		nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5482 		set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5483 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5484 	}
5485 out:
5486 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5487 }
5488 
5489 static void nfs4_lock_release(void *calldata)
5490 {
5491 	struct nfs4_lockdata *data = calldata;
5492 
5493 	dprintk("%s: begin!\n", __func__);
5494 	nfs_free_seqid(data->arg.open_seqid);
5495 	if (data->cancelled != 0) {
5496 		struct rpc_task *task;
5497 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5498 				data->arg.lock_seqid);
5499 		if (!IS_ERR(task))
5500 			rpc_put_task_async(task);
5501 		dprintk("%s: cancelling lock!\n", __func__);
5502 	} else
5503 		nfs_free_seqid(data->arg.lock_seqid);
5504 	nfs4_put_lock_state(data->lsp);
5505 	put_nfs_open_context(data->ctx);
5506 	kfree(data);
5507 	dprintk("%s: done!\n", __func__);
5508 }
5509 
5510 static const struct rpc_call_ops nfs4_lock_ops = {
5511 	.rpc_call_prepare = nfs4_lock_prepare,
5512 	.rpc_call_done = nfs4_lock_done,
5513 	.rpc_release = nfs4_lock_release,
5514 };
5515 
5516 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5517 {
5518 	switch (error) {
5519 	case -NFS4ERR_ADMIN_REVOKED:
5520 	case -NFS4ERR_BAD_STATEID:
5521 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5522 		if (new_lock_owner != 0 ||
5523 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5524 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5525 		break;
5526 	case -NFS4ERR_STALE_STATEID:
5527 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5528 	case -NFS4ERR_EXPIRED:
5529 		nfs4_schedule_lease_recovery(server->nfs_client);
5530 	};
5531 }
5532 
5533 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5534 {
5535 	struct nfs4_lockdata *data;
5536 	struct rpc_task *task;
5537 	struct rpc_message msg = {
5538 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5539 		.rpc_cred = state->owner->so_cred,
5540 	};
5541 	struct rpc_task_setup task_setup_data = {
5542 		.rpc_client = NFS_CLIENT(state->inode),
5543 		.rpc_message = &msg,
5544 		.callback_ops = &nfs4_lock_ops,
5545 		.workqueue = nfsiod_workqueue,
5546 		.flags = RPC_TASK_ASYNC,
5547 	};
5548 	int ret;
5549 
5550 	dprintk("%s: begin!\n", __func__);
5551 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5552 			fl->fl_u.nfs4_fl.owner,
5553 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5554 	if (data == NULL)
5555 		return -ENOMEM;
5556 	if (IS_SETLKW(cmd))
5557 		data->arg.block = 1;
5558 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5559 	msg.rpc_argp = &data->arg;
5560 	msg.rpc_resp = &data->res;
5561 	task_setup_data.callback_data = data;
5562 	if (recovery_type > NFS_LOCK_NEW) {
5563 		if (recovery_type == NFS_LOCK_RECLAIM)
5564 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5565 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5566 	}
5567 	task = rpc_run_task(&task_setup_data);
5568 	if (IS_ERR(task))
5569 		return PTR_ERR(task);
5570 	ret = nfs4_wait_for_completion_rpc_task(task);
5571 	if (ret == 0) {
5572 		ret = data->rpc_status;
5573 		if (ret)
5574 			nfs4_handle_setlk_error(data->server, data->lsp,
5575 					data->arg.new_lock_owner, ret);
5576 	} else
5577 		data->cancelled = 1;
5578 	rpc_put_task(task);
5579 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5580 	return ret;
5581 }
5582 
5583 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5584 {
5585 	struct nfs_server *server = NFS_SERVER(state->inode);
5586 	struct nfs4_exception exception = {
5587 		.inode = state->inode,
5588 	};
5589 	int err;
5590 
5591 	do {
5592 		/* Cache the lock if possible... */
5593 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5594 			return 0;
5595 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5596 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5597 		if (err != -NFS4ERR_DELAY)
5598 			break;
5599 		nfs4_handle_exception(server, err, &exception);
5600 	} while (exception.retry);
5601 	return err;
5602 }
5603 
5604 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5605 {
5606 	struct nfs_server *server = NFS_SERVER(state->inode);
5607 	struct nfs4_exception exception = {
5608 		.inode = state->inode,
5609 	};
5610 	int err;
5611 
5612 	err = nfs4_set_lock_state(state, request);
5613 	if (err != 0)
5614 		return err;
5615 	if (!recover_lost_locks) {
5616 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5617 		return 0;
5618 	}
5619 	do {
5620 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5621 			return 0;
5622 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5623 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5624 		switch (err) {
5625 		default:
5626 			goto out;
5627 		case -NFS4ERR_GRACE:
5628 		case -NFS4ERR_DELAY:
5629 			nfs4_handle_exception(server, err, &exception);
5630 			err = 0;
5631 		}
5632 	} while (exception.retry);
5633 out:
5634 	return err;
5635 }
5636 
5637 #if defined(CONFIG_NFS_V4_1)
5638 /**
5639  * nfs41_check_expired_locks - possibly free a lock stateid
5640  *
5641  * @state: NFSv4 state for an inode
5642  *
5643  * Returns NFS_OK if recovery for this stateid is now finished.
5644  * Otherwise a negative NFS4ERR value is returned.
5645  */
5646 static int nfs41_check_expired_locks(struct nfs4_state *state)
5647 {
5648 	int status, ret = -NFS4ERR_BAD_STATEID;
5649 	struct nfs4_lock_state *lsp;
5650 	struct nfs_server *server = NFS_SERVER(state->inode);
5651 
5652 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5653 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5654 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5655 
5656 			status = nfs41_test_stateid(server,
5657 					&lsp->ls_stateid,
5658 					cred);
5659 			trace_nfs4_test_lock_stateid(state, lsp, status);
5660 			if (status != NFS_OK) {
5661 				/* Free the stateid unless the server
5662 				 * informs us the stateid is unrecognized. */
5663 				if (status != -NFS4ERR_BAD_STATEID)
5664 					nfs41_free_stateid(server,
5665 							&lsp->ls_stateid,
5666 							cred);
5667 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5668 				ret = status;
5669 			}
5670 		}
5671 	};
5672 
5673 	return ret;
5674 }
5675 
5676 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5677 {
5678 	int status = NFS_OK;
5679 
5680 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5681 		status = nfs41_check_expired_locks(state);
5682 	if (status != NFS_OK)
5683 		status = nfs4_lock_expired(state, request);
5684 	return status;
5685 }
5686 #endif
5687 
5688 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5689 {
5690 	struct nfs4_state_owner *sp = state->owner;
5691 	struct nfs_inode *nfsi = NFS_I(state->inode);
5692 	unsigned char fl_flags = request->fl_flags;
5693 	unsigned int seq;
5694 	int status = -ENOLCK;
5695 
5696 	if ((fl_flags & FL_POSIX) &&
5697 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5698 		goto out;
5699 	/* Is this a delegated open? */
5700 	status = nfs4_set_lock_state(state, request);
5701 	if (status != 0)
5702 		goto out;
5703 	request->fl_flags |= FL_ACCESS;
5704 	status = do_vfs_lock(request->fl_file, request);
5705 	if (status < 0)
5706 		goto out;
5707 	down_read(&nfsi->rwsem);
5708 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5709 		/* Yes: cache locks! */
5710 		/* ...but avoid races with delegation recall... */
5711 		request->fl_flags = fl_flags & ~FL_SLEEP;
5712 		status = do_vfs_lock(request->fl_file, request);
5713 		goto out_unlock;
5714 	}
5715 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5716 	up_read(&nfsi->rwsem);
5717 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5718 	if (status != 0)
5719 		goto out;
5720 	down_read(&nfsi->rwsem);
5721 	if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5722 		status = -NFS4ERR_DELAY;
5723 		goto out_unlock;
5724 	}
5725 	/* Note: we always want to sleep here! */
5726 	request->fl_flags = fl_flags | FL_SLEEP;
5727 	if (do_vfs_lock(request->fl_file, request) < 0)
5728 		printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5729 			"manager!\n", __func__);
5730 out_unlock:
5731 	up_read(&nfsi->rwsem);
5732 out:
5733 	request->fl_flags = fl_flags;
5734 	return status;
5735 }
5736 
5737 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5738 {
5739 	struct nfs4_exception exception = {
5740 		.state = state,
5741 		.inode = state->inode,
5742 	};
5743 	int err;
5744 
5745 	do {
5746 		err = _nfs4_proc_setlk(state, cmd, request);
5747 		trace_nfs4_set_lock(request, state, cmd, err);
5748 		if (err == -NFS4ERR_DENIED)
5749 			err = -EAGAIN;
5750 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5751 				err, &exception);
5752 	} while (exception.retry);
5753 	return err;
5754 }
5755 
5756 static int
5757 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5758 {
5759 	struct nfs_open_context *ctx;
5760 	struct nfs4_state *state;
5761 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5762 	int status;
5763 
5764 	/* verify open state */
5765 	ctx = nfs_file_open_context(filp);
5766 	state = ctx->state;
5767 
5768 	if (request->fl_start < 0 || request->fl_end < 0)
5769 		return -EINVAL;
5770 
5771 	if (IS_GETLK(cmd)) {
5772 		if (state != NULL)
5773 			return nfs4_proc_getlk(state, F_GETLK, request);
5774 		return 0;
5775 	}
5776 
5777 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5778 		return -EINVAL;
5779 
5780 	if (request->fl_type == F_UNLCK) {
5781 		if (state != NULL)
5782 			return nfs4_proc_unlck(state, cmd, request);
5783 		return 0;
5784 	}
5785 
5786 	if (state == NULL)
5787 		return -ENOLCK;
5788 	/*
5789 	 * Don't rely on the VFS having checked the file open mode,
5790 	 * since it won't do this for flock() locks.
5791 	 */
5792 	switch (request->fl_type) {
5793 	case F_RDLCK:
5794 		if (!(filp->f_mode & FMODE_READ))
5795 			return -EBADF;
5796 		break;
5797 	case F_WRLCK:
5798 		if (!(filp->f_mode & FMODE_WRITE))
5799 			return -EBADF;
5800 	}
5801 
5802 	do {
5803 		status = nfs4_proc_setlk(state, cmd, request);
5804 		if ((status != -EAGAIN) || IS_SETLK(cmd))
5805 			break;
5806 		timeout = nfs4_set_lock_task_retry(timeout);
5807 		status = -ERESTARTSYS;
5808 		if (signalled())
5809 			break;
5810 	} while(status < 0);
5811 	return status;
5812 }
5813 
5814 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5815 {
5816 	struct nfs_server *server = NFS_SERVER(state->inode);
5817 	int err;
5818 
5819 	err = nfs4_set_lock_state(state, fl);
5820 	if (err != 0)
5821 		return err;
5822 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5823 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5824 }
5825 
5826 struct nfs_release_lockowner_data {
5827 	struct nfs4_lock_state *lsp;
5828 	struct nfs_server *server;
5829 	struct nfs_release_lockowner_args args;
5830 	struct nfs_release_lockowner_res res;
5831 	unsigned long timestamp;
5832 };
5833 
5834 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5835 {
5836 	struct nfs_release_lockowner_data *data = calldata;
5837 	nfs40_setup_sequence(data->server,
5838 				&data->args.seq_args, &data->res.seq_res, task);
5839 	data->timestamp = jiffies;
5840 }
5841 
5842 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5843 {
5844 	struct nfs_release_lockowner_data *data = calldata;
5845 	struct nfs_server *server = data->server;
5846 
5847 	nfs40_sequence_done(task, &data->res.seq_res);
5848 
5849 	switch (task->tk_status) {
5850 	case 0:
5851 		renew_lease(server, data->timestamp);
5852 		break;
5853 	case -NFS4ERR_STALE_CLIENTID:
5854 	case -NFS4ERR_EXPIRED:
5855 	case -NFS4ERR_LEASE_MOVED:
5856 	case -NFS4ERR_DELAY:
5857 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN)
5858 			rpc_restart_call_prepare(task);
5859 	}
5860 }
5861 
5862 static void nfs4_release_lockowner_release(void *calldata)
5863 {
5864 	struct nfs_release_lockowner_data *data = calldata;
5865 	nfs4_free_lock_state(data->server, data->lsp);
5866 	kfree(calldata);
5867 }
5868 
5869 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5870 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
5871 	.rpc_call_done = nfs4_release_lockowner_done,
5872 	.rpc_release = nfs4_release_lockowner_release,
5873 };
5874 
5875 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5876 {
5877 	struct nfs_release_lockowner_data *data;
5878 	struct rpc_message msg = {
5879 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5880 	};
5881 
5882 	if (server->nfs_client->cl_mvops->minor_version != 0)
5883 		return -EINVAL;
5884 
5885 	data = kmalloc(sizeof(*data), GFP_NOFS);
5886 	if (!data)
5887 		return -ENOMEM;
5888 	data->lsp = lsp;
5889 	data->server = server;
5890 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5891 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5892 	data->args.lock_owner.s_dev = server->s_dev;
5893 
5894 	msg.rpc_argp = &data->args;
5895 	msg.rpc_resp = &data->res;
5896 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
5897 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5898 	return 0;
5899 }
5900 
5901 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5902 
5903 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5904 				   const void *buf, size_t buflen,
5905 				   int flags, int type)
5906 {
5907 	if (strcmp(key, "") != 0)
5908 		return -EINVAL;
5909 
5910 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5911 }
5912 
5913 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5914 				   void *buf, size_t buflen, int type)
5915 {
5916 	if (strcmp(key, "") != 0)
5917 		return -EINVAL;
5918 
5919 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5920 }
5921 
5922 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5923 				       size_t list_len, const char *name,
5924 				       size_t name_len, int type)
5925 {
5926 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5927 
5928 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5929 		return 0;
5930 
5931 	if (list && len <= list_len)
5932 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5933 	return len;
5934 }
5935 
5936 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5937 static inline int nfs4_server_supports_labels(struct nfs_server *server)
5938 {
5939 	return server->caps & NFS_CAP_SECURITY_LABEL;
5940 }
5941 
5942 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
5943 				   const void *buf, size_t buflen,
5944 				   int flags, int type)
5945 {
5946 	if (security_ismaclabel(key))
5947 		return nfs4_set_security_label(dentry, buf, buflen);
5948 
5949 	return -EOPNOTSUPP;
5950 }
5951 
5952 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
5953 				   void *buf, size_t buflen, int type)
5954 {
5955 	if (security_ismaclabel(key))
5956 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
5957 	return -EOPNOTSUPP;
5958 }
5959 
5960 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
5961 				       size_t list_len, const char *name,
5962 				       size_t name_len, int type)
5963 {
5964 	size_t len = 0;
5965 
5966 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
5967 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
5968 		if (list && len <= list_len)
5969 			security_inode_listsecurity(dentry->d_inode, list, len);
5970 	}
5971 	return len;
5972 }
5973 
5974 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
5975 	.prefix = XATTR_SECURITY_PREFIX,
5976 	.list	= nfs4_xattr_list_nfs4_label,
5977 	.get	= nfs4_xattr_get_nfs4_label,
5978 	.set	= nfs4_xattr_set_nfs4_label,
5979 };
5980 #endif
5981 
5982 
5983 /*
5984  * nfs_fhget will use either the mounted_on_fileid or the fileid
5985  */
5986 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5987 {
5988 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5989 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5990 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5991 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5992 		return;
5993 
5994 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5995 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5996 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5997 	fattr->nlink = 2;
5998 }
5999 
6000 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6001 				   const struct qstr *name,
6002 				   struct nfs4_fs_locations *fs_locations,
6003 				   struct page *page)
6004 {
6005 	struct nfs_server *server = NFS_SERVER(dir);
6006 	u32 bitmask[3] = {
6007 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6008 	};
6009 	struct nfs4_fs_locations_arg args = {
6010 		.dir_fh = NFS_FH(dir),
6011 		.name = name,
6012 		.page = page,
6013 		.bitmask = bitmask,
6014 	};
6015 	struct nfs4_fs_locations_res res = {
6016 		.fs_locations = fs_locations,
6017 	};
6018 	struct rpc_message msg = {
6019 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6020 		.rpc_argp = &args,
6021 		.rpc_resp = &res,
6022 	};
6023 	int status;
6024 
6025 	dprintk("%s: start\n", __func__);
6026 
6027 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6028 	 * is not supported */
6029 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6030 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6031 	else
6032 		bitmask[0] |= FATTR4_WORD0_FILEID;
6033 
6034 	nfs_fattr_init(&fs_locations->fattr);
6035 	fs_locations->server = server;
6036 	fs_locations->nlocations = 0;
6037 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6038 	dprintk("%s: returned status = %d\n", __func__, status);
6039 	return status;
6040 }
6041 
6042 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6043 			   const struct qstr *name,
6044 			   struct nfs4_fs_locations *fs_locations,
6045 			   struct page *page)
6046 {
6047 	struct nfs4_exception exception = { };
6048 	int err;
6049 	do {
6050 		err = _nfs4_proc_fs_locations(client, dir, name,
6051 				fs_locations, page);
6052 		trace_nfs4_get_fs_locations(dir, name, err);
6053 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6054 				&exception);
6055 	} while (exception.retry);
6056 	return err;
6057 }
6058 
6059 /*
6060  * This operation also signals the server that this client is
6061  * performing migration recovery.  The server can stop returning
6062  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6063  * appended to this compound to identify the client ID which is
6064  * performing recovery.
6065  */
6066 static int _nfs40_proc_get_locations(struct inode *inode,
6067 				     struct nfs4_fs_locations *locations,
6068 				     struct page *page, struct rpc_cred *cred)
6069 {
6070 	struct nfs_server *server = NFS_SERVER(inode);
6071 	struct rpc_clnt *clnt = server->client;
6072 	u32 bitmask[2] = {
6073 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6074 	};
6075 	struct nfs4_fs_locations_arg args = {
6076 		.clientid	= server->nfs_client->cl_clientid,
6077 		.fh		= NFS_FH(inode),
6078 		.page		= page,
6079 		.bitmask	= bitmask,
6080 		.migration	= 1,		/* skip LOOKUP */
6081 		.renew		= 1,		/* append RENEW */
6082 	};
6083 	struct nfs4_fs_locations_res res = {
6084 		.fs_locations	= locations,
6085 		.migration	= 1,
6086 		.renew		= 1,
6087 	};
6088 	struct rpc_message msg = {
6089 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6090 		.rpc_argp	= &args,
6091 		.rpc_resp	= &res,
6092 		.rpc_cred	= cred,
6093 	};
6094 	unsigned long now = jiffies;
6095 	int status;
6096 
6097 	nfs_fattr_init(&locations->fattr);
6098 	locations->server = server;
6099 	locations->nlocations = 0;
6100 
6101 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6102 	nfs4_set_sequence_privileged(&args.seq_args);
6103 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6104 					&args.seq_args, &res.seq_res);
6105 	if (status)
6106 		return status;
6107 
6108 	renew_lease(server, now);
6109 	return 0;
6110 }
6111 
6112 #ifdef CONFIG_NFS_V4_1
6113 
6114 /*
6115  * This operation also signals the server that this client is
6116  * performing migration recovery.  The server can stop asserting
6117  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6118  * performing this operation is identified in the SEQUENCE
6119  * operation in this compound.
6120  *
6121  * When the client supports GETATTR(fs_locations_info), it can
6122  * be plumbed in here.
6123  */
6124 static int _nfs41_proc_get_locations(struct inode *inode,
6125 				     struct nfs4_fs_locations *locations,
6126 				     struct page *page, struct rpc_cred *cred)
6127 {
6128 	struct nfs_server *server = NFS_SERVER(inode);
6129 	struct rpc_clnt *clnt = server->client;
6130 	u32 bitmask[2] = {
6131 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6132 	};
6133 	struct nfs4_fs_locations_arg args = {
6134 		.fh		= NFS_FH(inode),
6135 		.page		= page,
6136 		.bitmask	= bitmask,
6137 		.migration	= 1,		/* skip LOOKUP */
6138 	};
6139 	struct nfs4_fs_locations_res res = {
6140 		.fs_locations	= locations,
6141 		.migration	= 1,
6142 	};
6143 	struct rpc_message msg = {
6144 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6145 		.rpc_argp	= &args,
6146 		.rpc_resp	= &res,
6147 		.rpc_cred	= cred,
6148 	};
6149 	int status;
6150 
6151 	nfs_fattr_init(&locations->fattr);
6152 	locations->server = server;
6153 	locations->nlocations = 0;
6154 
6155 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6156 	nfs4_set_sequence_privileged(&args.seq_args);
6157 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6158 					&args.seq_args, &res.seq_res);
6159 	if (status == NFS4_OK &&
6160 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6161 		status = -NFS4ERR_LEASE_MOVED;
6162 	return status;
6163 }
6164 
6165 #endif	/* CONFIG_NFS_V4_1 */
6166 
6167 /**
6168  * nfs4_proc_get_locations - discover locations for a migrated FSID
6169  * @inode: inode on FSID that is migrating
6170  * @locations: result of query
6171  * @page: buffer
6172  * @cred: credential to use for this operation
6173  *
6174  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6175  * operation failed, or a negative errno if a local error occurred.
6176  *
6177  * On success, "locations" is filled in, but if the server has
6178  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6179  * asserted.
6180  *
6181  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6182  * from this client that require migration recovery.
6183  */
6184 int nfs4_proc_get_locations(struct inode *inode,
6185 			    struct nfs4_fs_locations *locations,
6186 			    struct page *page, struct rpc_cred *cred)
6187 {
6188 	struct nfs_server *server = NFS_SERVER(inode);
6189 	struct nfs_client *clp = server->nfs_client;
6190 	const struct nfs4_mig_recovery_ops *ops =
6191 					clp->cl_mvops->mig_recovery_ops;
6192 	struct nfs4_exception exception = { };
6193 	int status;
6194 
6195 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6196 		(unsigned long long)server->fsid.major,
6197 		(unsigned long long)server->fsid.minor,
6198 		clp->cl_hostname);
6199 	nfs_display_fhandle(NFS_FH(inode), __func__);
6200 
6201 	do {
6202 		status = ops->get_locations(inode, locations, page, cred);
6203 		if (status != -NFS4ERR_DELAY)
6204 			break;
6205 		nfs4_handle_exception(server, status, &exception);
6206 	} while (exception.retry);
6207 	return status;
6208 }
6209 
6210 /*
6211  * This operation also signals the server that this client is
6212  * performing "lease moved" recovery.  The server can stop
6213  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6214  * is appended to this compound to identify the client ID which is
6215  * performing recovery.
6216  */
6217 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6218 {
6219 	struct nfs_server *server = NFS_SERVER(inode);
6220 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6221 	struct rpc_clnt *clnt = server->client;
6222 	struct nfs4_fsid_present_arg args = {
6223 		.fh		= NFS_FH(inode),
6224 		.clientid	= clp->cl_clientid,
6225 		.renew		= 1,		/* append RENEW */
6226 	};
6227 	struct nfs4_fsid_present_res res = {
6228 		.renew		= 1,
6229 	};
6230 	struct rpc_message msg = {
6231 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6232 		.rpc_argp	= &args,
6233 		.rpc_resp	= &res,
6234 		.rpc_cred	= cred,
6235 	};
6236 	unsigned long now = jiffies;
6237 	int status;
6238 
6239 	res.fh = nfs_alloc_fhandle();
6240 	if (res.fh == NULL)
6241 		return -ENOMEM;
6242 
6243 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6244 	nfs4_set_sequence_privileged(&args.seq_args);
6245 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6246 						&args.seq_args, &res.seq_res);
6247 	nfs_free_fhandle(res.fh);
6248 	if (status)
6249 		return status;
6250 
6251 	do_renew_lease(clp, now);
6252 	return 0;
6253 }
6254 
6255 #ifdef CONFIG_NFS_V4_1
6256 
6257 /*
6258  * This operation also signals the server that this client is
6259  * performing "lease moved" recovery.  The server can stop asserting
6260  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6261  * this operation is identified in the SEQUENCE operation in this
6262  * compound.
6263  */
6264 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6265 {
6266 	struct nfs_server *server = NFS_SERVER(inode);
6267 	struct rpc_clnt *clnt = server->client;
6268 	struct nfs4_fsid_present_arg args = {
6269 		.fh		= NFS_FH(inode),
6270 	};
6271 	struct nfs4_fsid_present_res res = {
6272 	};
6273 	struct rpc_message msg = {
6274 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6275 		.rpc_argp	= &args,
6276 		.rpc_resp	= &res,
6277 		.rpc_cred	= cred,
6278 	};
6279 	int status;
6280 
6281 	res.fh = nfs_alloc_fhandle();
6282 	if (res.fh == NULL)
6283 		return -ENOMEM;
6284 
6285 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6286 	nfs4_set_sequence_privileged(&args.seq_args);
6287 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6288 						&args.seq_args, &res.seq_res);
6289 	nfs_free_fhandle(res.fh);
6290 	if (status == NFS4_OK &&
6291 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6292 		status = -NFS4ERR_LEASE_MOVED;
6293 	return status;
6294 }
6295 
6296 #endif	/* CONFIG_NFS_V4_1 */
6297 
6298 /**
6299  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6300  * @inode: inode on FSID to check
6301  * @cred: credential to use for this operation
6302  *
6303  * Server indicates whether the FSID is present, moved, or not
6304  * recognized.  This operation is necessary to clear a LEASE_MOVED
6305  * condition for this client ID.
6306  *
6307  * Returns NFS4_OK if the FSID is present on this server,
6308  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6309  *  NFS4ERR code if some error occurred on the server, or a
6310  *  negative errno if a local failure occurred.
6311  */
6312 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6313 {
6314 	struct nfs_server *server = NFS_SERVER(inode);
6315 	struct nfs_client *clp = server->nfs_client;
6316 	const struct nfs4_mig_recovery_ops *ops =
6317 					clp->cl_mvops->mig_recovery_ops;
6318 	struct nfs4_exception exception = { };
6319 	int status;
6320 
6321 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6322 		(unsigned long long)server->fsid.major,
6323 		(unsigned long long)server->fsid.minor,
6324 		clp->cl_hostname);
6325 	nfs_display_fhandle(NFS_FH(inode), __func__);
6326 
6327 	do {
6328 		status = ops->fsid_present(inode, cred);
6329 		if (status != -NFS4ERR_DELAY)
6330 			break;
6331 		nfs4_handle_exception(server, status, &exception);
6332 	} while (exception.retry);
6333 	return status;
6334 }
6335 
6336 /**
6337  * If 'use_integrity' is true and the state managment nfs_client
6338  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6339  * and the machine credential as per RFC3530bis and RFC5661 Security
6340  * Considerations sections. Otherwise, just use the user cred with the
6341  * filesystem's rpc_client.
6342  */
6343 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6344 {
6345 	int status;
6346 	struct nfs4_secinfo_arg args = {
6347 		.dir_fh = NFS_FH(dir),
6348 		.name   = name,
6349 	};
6350 	struct nfs4_secinfo_res res = {
6351 		.flavors     = flavors,
6352 	};
6353 	struct rpc_message msg = {
6354 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6355 		.rpc_argp = &args,
6356 		.rpc_resp = &res,
6357 	};
6358 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6359 	struct rpc_cred *cred = NULL;
6360 
6361 	if (use_integrity) {
6362 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6363 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6364 		msg.rpc_cred = cred;
6365 	}
6366 
6367 	dprintk("NFS call  secinfo %s\n", name->name);
6368 
6369 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6370 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6371 
6372 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6373 				&res.seq_res, 0);
6374 	dprintk("NFS reply  secinfo: %d\n", status);
6375 
6376 	if (cred)
6377 		put_rpccred(cred);
6378 
6379 	return status;
6380 }
6381 
6382 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6383 		      struct nfs4_secinfo_flavors *flavors)
6384 {
6385 	struct nfs4_exception exception = { };
6386 	int err;
6387 	do {
6388 		err = -NFS4ERR_WRONGSEC;
6389 
6390 		/* try to use integrity protection with machine cred */
6391 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6392 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6393 
6394 		/*
6395 		 * if unable to use integrity protection, or SECINFO with
6396 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6397 		 * disallowed by spec, but exists in deployed servers) use
6398 		 * the current filesystem's rpc_client and the user cred.
6399 		 */
6400 		if (err == -NFS4ERR_WRONGSEC)
6401 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6402 
6403 		trace_nfs4_secinfo(dir, name, err);
6404 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6405 				&exception);
6406 	} while (exception.retry);
6407 	return err;
6408 }
6409 
6410 #ifdef CONFIG_NFS_V4_1
6411 /*
6412  * Check the exchange flags returned by the server for invalid flags, having
6413  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6414  * DS flags set.
6415  */
6416 static int nfs4_check_cl_exchange_flags(u32 flags)
6417 {
6418 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6419 		goto out_inval;
6420 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6421 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6422 		goto out_inval;
6423 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6424 		goto out_inval;
6425 	return NFS_OK;
6426 out_inval:
6427 	return -NFS4ERR_INVAL;
6428 }
6429 
6430 static bool
6431 nfs41_same_server_scope(struct nfs41_server_scope *a,
6432 			struct nfs41_server_scope *b)
6433 {
6434 	if (a->server_scope_sz == b->server_scope_sz &&
6435 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6436 		return true;
6437 
6438 	return false;
6439 }
6440 
6441 /*
6442  * nfs4_proc_bind_conn_to_session()
6443  *
6444  * The 4.1 client currently uses the same TCP connection for the
6445  * fore and backchannel.
6446  */
6447 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6448 {
6449 	int status;
6450 	struct nfs41_bind_conn_to_session_res res;
6451 	struct rpc_message msg = {
6452 		.rpc_proc =
6453 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6454 		.rpc_argp = clp,
6455 		.rpc_resp = &res,
6456 		.rpc_cred = cred,
6457 	};
6458 
6459 	dprintk("--> %s\n", __func__);
6460 
6461 	res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6462 	if (unlikely(res.session == NULL)) {
6463 		status = -ENOMEM;
6464 		goto out;
6465 	}
6466 
6467 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6468 	trace_nfs4_bind_conn_to_session(clp, status);
6469 	if (status == 0) {
6470 		if (memcmp(res.session->sess_id.data,
6471 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6472 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6473 			status = -EIO;
6474 			goto out_session;
6475 		}
6476 		if (res.dir != NFS4_CDFS4_BOTH) {
6477 			dprintk("NFS: %s: Unexpected direction from server\n",
6478 				__func__);
6479 			status = -EIO;
6480 			goto out_session;
6481 		}
6482 		if (res.use_conn_in_rdma_mode) {
6483 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6484 				__func__);
6485 			status = -EIO;
6486 			goto out_session;
6487 		}
6488 	}
6489 out_session:
6490 	kfree(res.session);
6491 out:
6492 	dprintk("<-- %s status= %d\n", __func__, status);
6493 	return status;
6494 }
6495 
6496 /*
6497  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6498  * and operations we'd like to see to enable certain features in the allow map
6499  */
6500 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6501 	.how = SP4_MACH_CRED,
6502 	.enforce.u.words = {
6503 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6504 		      1 << (OP_EXCHANGE_ID - 32) |
6505 		      1 << (OP_CREATE_SESSION - 32) |
6506 		      1 << (OP_DESTROY_SESSION - 32) |
6507 		      1 << (OP_DESTROY_CLIENTID - 32)
6508 	},
6509 	.allow.u.words = {
6510 		[0] = 1 << (OP_CLOSE) |
6511 		      1 << (OP_LOCKU) |
6512 		      1 << (OP_COMMIT),
6513 		[1] = 1 << (OP_SECINFO - 32) |
6514 		      1 << (OP_SECINFO_NO_NAME - 32) |
6515 		      1 << (OP_TEST_STATEID - 32) |
6516 		      1 << (OP_FREE_STATEID - 32) |
6517 		      1 << (OP_WRITE - 32)
6518 	}
6519 };
6520 
6521 /*
6522  * Select the state protection mode for client `clp' given the server results
6523  * from exchange_id in `sp'.
6524  *
6525  * Returns 0 on success, negative errno otherwise.
6526  */
6527 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6528 				 struct nfs41_state_protection *sp)
6529 {
6530 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6531 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6532 		      1 << (OP_EXCHANGE_ID - 32) |
6533 		      1 << (OP_CREATE_SESSION - 32) |
6534 		      1 << (OP_DESTROY_SESSION - 32) |
6535 		      1 << (OP_DESTROY_CLIENTID - 32)
6536 	};
6537 	unsigned int i;
6538 
6539 	if (sp->how == SP4_MACH_CRED) {
6540 		/* Print state protect result */
6541 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6542 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6543 			if (test_bit(i, sp->enforce.u.longs))
6544 				dfprintk(MOUNT, "  enforce op %d\n", i);
6545 			if (test_bit(i, sp->allow.u.longs))
6546 				dfprintk(MOUNT, "  allow op %d\n", i);
6547 		}
6548 
6549 		/* make sure nothing is on enforce list that isn't supported */
6550 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6551 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6552 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6553 				return -EINVAL;
6554 			}
6555 		}
6556 
6557 		/*
6558 		 * Minimal mode - state operations are allowed to use machine
6559 		 * credential.  Note this already happens by default, so the
6560 		 * client doesn't have to do anything more than the negotiation.
6561 		 *
6562 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6563 		 *       we're already using the machine cred for exchange_id
6564 		 *       and will never use a different cred.
6565 		 */
6566 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6567 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6568 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6569 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6570 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6571 			dfprintk(MOUNT, "  minimal mode enabled\n");
6572 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6573 		} else {
6574 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6575 			return -EINVAL;
6576 		}
6577 
6578 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6579 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6580 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6581 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6582 		}
6583 
6584 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6585 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6586 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6587 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6588 		}
6589 
6590 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6591 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6592 			dfprintk(MOUNT, "  stateid mode enabled\n");
6593 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6594 		}
6595 
6596 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6597 			dfprintk(MOUNT, "  write mode enabled\n");
6598 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6599 		}
6600 
6601 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6602 			dfprintk(MOUNT, "  commit mode enabled\n");
6603 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6604 		}
6605 	}
6606 
6607 	return 0;
6608 }
6609 
6610 /*
6611  * _nfs4_proc_exchange_id()
6612  *
6613  * Wrapper for EXCHANGE_ID operation.
6614  */
6615 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6616 	u32 sp4_how)
6617 {
6618 	nfs4_verifier verifier;
6619 	struct nfs41_exchange_id_args args = {
6620 		.verifier = &verifier,
6621 		.client = clp,
6622 #ifdef CONFIG_NFS_V4_1_MIGRATION
6623 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6624 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6625 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6626 #else
6627 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6628 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6629 #endif
6630 	};
6631 	struct nfs41_exchange_id_res res = {
6632 		0
6633 	};
6634 	int status;
6635 	struct rpc_message msg = {
6636 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6637 		.rpc_argp = &args,
6638 		.rpc_resp = &res,
6639 		.rpc_cred = cred,
6640 	};
6641 
6642 	nfs4_init_boot_verifier(clp, &verifier);
6643 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6644 							sizeof(args.id));
6645 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6646 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6647 		args.id_len, args.id);
6648 
6649 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6650 					GFP_NOFS);
6651 	if (unlikely(res.server_owner == NULL)) {
6652 		status = -ENOMEM;
6653 		goto out;
6654 	}
6655 
6656 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6657 					GFP_NOFS);
6658 	if (unlikely(res.server_scope == NULL)) {
6659 		status = -ENOMEM;
6660 		goto out_server_owner;
6661 	}
6662 
6663 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6664 	if (unlikely(res.impl_id == NULL)) {
6665 		status = -ENOMEM;
6666 		goto out_server_scope;
6667 	}
6668 
6669 	switch (sp4_how) {
6670 	case SP4_NONE:
6671 		args.state_protect.how = SP4_NONE;
6672 		break;
6673 
6674 	case SP4_MACH_CRED:
6675 		args.state_protect = nfs4_sp4_mach_cred_request;
6676 		break;
6677 
6678 	default:
6679 		/* unsupported! */
6680 		WARN_ON_ONCE(1);
6681 		status = -EINVAL;
6682 		goto out_server_scope;
6683 	}
6684 
6685 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6686 	trace_nfs4_exchange_id(clp, status);
6687 	if (status == 0)
6688 		status = nfs4_check_cl_exchange_flags(res.flags);
6689 
6690 	if (status == 0)
6691 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6692 
6693 	if (status == 0) {
6694 		clp->cl_clientid = res.clientid;
6695 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6696 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6697 			clp->cl_seqid = res.seqid;
6698 
6699 		kfree(clp->cl_serverowner);
6700 		clp->cl_serverowner = res.server_owner;
6701 		res.server_owner = NULL;
6702 
6703 		/* use the most recent implementation id */
6704 		kfree(clp->cl_implid);
6705 		clp->cl_implid = res.impl_id;
6706 
6707 		if (clp->cl_serverscope != NULL &&
6708 		    !nfs41_same_server_scope(clp->cl_serverscope,
6709 					     res.server_scope)) {
6710 			dprintk("%s: server_scope mismatch detected\n",
6711 				__func__);
6712 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6713 			kfree(clp->cl_serverscope);
6714 			clp->cl_serverscope = NULL;
6715 		}
6716 
6717 		if (clp->cl_serverscope == NULL) {
6718 			clp->cl_serverscope = res.server_scope;
6719 			goto out;
6720 		}
6721 	} else
6722 		kfree(res.impl_id);
6723 
6724 out_server_owner:
6725 	kfree(res.server_owner);
6726 out_server_scope:
6727 	kfree(res.server_scope);
6728 out:
6729 	if (clp->cl_implid != NULL)
6730 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6731 			"domain: %s, name: %s, date: %llu,%u\n",
6732 			clp->cl_implid->domain, clp->cl_implid->name,
6733 			clp->cl_implid->date.seconds,
6734 			clp->cl_implid->date.nseconds);
6735 	dprintk("NFS reply exchange_id: %d\n", status);
6736 	return status;
6737 }
6738 
6739 /*
6740  * nfs4_proc_exchange_id()
6741  *
6742  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6743  *
6744  * Since the clientid has expired, all compounds using sessions
6745  * associated with the stale clientid will be returning
6746  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6747  * be in some phase of session reset.
6748  *
6749  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6750  */
6751 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6752 {
6753 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6754 	int status;
6755 
6756 	/* try SP4_MACH_CRED if krb5i/p	*/
6757 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6758 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6759 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6760 		if (!status)
6761 			return 0;
6762 	}
6763 
6764 	/* try SP4_NONE */
6765 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6766 }
6767 
6768 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6769 		struct rpc_cred *cred)
6770 {
6771 	struct rpc_message msg = {
6772 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6773 		.rpc_argp = clp,
6774 		.rpc_cred = cred,
6775 	};
6776 	int status;
6777 
6778 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6779 	trace_nfs4_destroy_clientid(clp, status);
6780 	if (status)
6781 		dprintk("NFS: Got error %d from the server %s on "
6782 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6783 	return status;
6784 }
6785 
6786 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6787 		struct rpc_cred *cred)
6788 {
6789 	unsigned int loop;
6790 	int ret;
6791 
6792 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6793 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6794 		switch (ret) {
6795 		case -NFS4ERR_DELAY:
6796 		case -NFS4ERR_CLIENTID_BUSY:
6797 			ssleep(1);
6798 			break;
6799 		default:
6800 			return ret;
6801 		}
6802 	}
6803 	return 0;
6804 }
6805 
6806 int nfs4_destroy_clientid(struct nfs_client *clp)
6807 {
6808 	struct rpc_cred *cred;
6809 	int ret = 0;
6810 
6811 	if (clp->cl_mvops->minor_version < 1)
6812 		goto out;
6813 	if (clp->cl_exchange_flags == 0)
6814 		goto out;
6815 	if (clp->cl_preserve_clid)
6816 		goto out;
6817 	cred = nfs4_get_clid_cred(clp);
6818 	ret = nfs4_proc_destroy_clientid(clp, cred);
6819 	if (cred)
6820 		put_rpccred(cred);
6821 	switch (ret) {
6822 	case 0:
6823 	case -NFS4ERR_STALE_CLIENTID:
6824 		clp->cl_exchange_flags = 0;
6825 	}
6826 out:
6827 	return ret;
6828 }
6829 
6830 struct nfs4_get_lease_time_data {
6831 	struct nfs4_get_lease_time_args *args;
6832 	struct nfs4_get_lease_time_res *res;
6833 	struct nfs_client *clp;
6834 };
6835 
6836 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6837 					void *calldata)
6838 {
6839 	struct nfs4_get_lease_time_data *data =
6840 			(struct nfs4_get_lease_time_data *)calldata;
6841 
6842 	dprintk("--> %s\n", __func__);
6843 	/* just setup sequence, do not trigger session recovery
6844 	   since we're invoked within one */
6845 	nfs41_setup_sequence(data->clp->cl_session,
6846 			&data->args->la_seq_args,
6847 			&data->res->lr_seq_res,
6848 			task);
6849 	dprintk("<-- %s\n", __func__);
6850 }
6851 
6852 /*
6853  * Called from nfs4_state_manager thread for session setup, so don't recover
6854  * from sequence operation or clientid errors.
6855  */
6856 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6857 {
6858 	struct nfs4_get_lease_time_data *data =
6859 			(struct nfs4_get_lease_time_data *)calldata;
6860 
6861 	dprintk("--> %s\n", __func__);
6862 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6863 		return;
6864 	switch (task->tk_status) {
6865 	case -NFS4ERR_DELAY:
6866 	case -NFS4ERR_GRACE:
6867 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6868 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
6869 		task->tk_status = 0;
6870 		/* fall through */
6871 	case -NFS4ERR_RETRY_UNCACHED_REP:
6872 		rpc_restart_call_prepare(task);
6873 		return;
6874 	}
6875 	dprintk("<-- %s\n", __func__);
6876 }
6877 
6878 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
6879 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
6880 	.rpc_call_done = nfs4_get_lease_time_done,
6881 };
6882 
6883 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
6884 {
6885 	struct rpc_task *task;
6886 	struct nfs4_get_lease_time_args args;
6887 	struct nfs4_get_lease_time_res res = {
6888 		.lr_fsinfo = fsinfo,
6889 	};
6890 	struct nfs4_get_lease_time_data data = {
6891 		.args = &args,
6892 		.res = &res,
6893 		.clp = clp,
6894 	};
6895 	struct rpc_message msg = {
6896 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
6897 		.rpc_argp = &args,
6898 		.rpc_resp = &res,
6899 	};
6900 	struct rpc_task_setup task_setup = {
6901 		.rpc_client = clp->cl_rpcclient,
6902 		.rpc_message = &msg,
6903 		.callback_ops = &nfs4_get_lease_time_ops,
6904 		.callback_data = &data,
6905 		.flags = RPC_TASK_TIMEOUT,
6906 	};
6907 	int status;
6908 
6909 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
6910 	nfs4_set_sequence_privileged(&args.la_seq_args);
6911 	dprintk("--> %s\n", __func__);
6912 	task = rpc_run_task(&task_setup);
6913 
6914 	if (IS_ERR(task))
6915 		status = PTR_ERR(task);
6916 	else {
6917 		status = task->tk_status;
6918 		rpc_put_task(task);
6919 	}
6920 	dprintk("<-- %s return %d\n", __func__, status);
6921 
6922 	return status;
6923 }
6924 
6925 /*
6926  * Initialize the values to be used by the client in CREATE_SESSION
6927  * If nfs4_init_session set the fore channel request and response sizes,
6928  * use them.
6929  *
6930  * Set the back channel max_resp_sz_cached to zero to force the client to
6931  * always set csa_cachethis to FALSE because the current implementation
6932  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
6933  */
6934 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
6935 {
6936 	unsigned int max_rqst_sz, max_resp_sz;
6937 
6938 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
6939 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
6940 
6941 	/* Fore channel attributes */
6942 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
6943 	args->fc_attrs.max_resp_sz = max_resp_sz;
6944 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
6945 	args->fc_attrs.max_reqs = max_session_slots;
6946 
6947 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
6948 		"max_ops=%u max_reqs=%u\n",
6949 		__func__,
6950 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
6951 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
6952 
6953 	/* Back channel attributes */
6954 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
6955 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
6956 	args->bc_attrs.max_resp_sz_cached = 0;
6957 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
6958 	args->bc_attrs.max_reqs = 1;
6959 
6960 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
6961 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
6962 		__func__,
6963 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
6964 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
6965 		args->bc_attrs.max_reqs);
6966 }
6967 
6968 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6969 {
6970 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
6971 	struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
6972 
6973 	if (rcvd->max_resp_sz > sent->max_resp_sz)
6974 		return -EINVAL;
6975 	/*
6976 	 * Our requested max_ops is the minimum we need; we're not
6977 	 * prepared to break up compounds into smaller pieces than that.
6978 	 * So, no point even trying to continue if the server won't
6979 	 * cooperate:
6980 	 */
6981 	if (rcvd->max_ops < sent->max_ops)
6982 		return -EINVAL;
6983 	if (rcvd->max_reqs == 0)
6984 		return -EINVAL;
6985 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
6986 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
6987 	return 0;
6988 }
6989 
6990 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6991 {
6992 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
6993 	struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
6994 
6995 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
6996 		return -EINVAL;
6997 	if (rcvd->max_resp_sz < sent->max_resp_sz)
6998 		return -EINVAL;
6999 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7000 		return -EINVAL;
7001 	/* These would render the backchannel useless: */
7002 	if (rcvd->max_ops != sent->max_ops)
7003 		return -EINVAL;
7004 	if (rcvd->max_reqs != sent->max_reqs)
7005 		return -EINVAL;
7006 	return 0;
7007 }
7008 
7009 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7010 				     struct nfs4_session *session)
7011 {
7012 	int ret;
7013 
7014 	ret = nfs4_verify_fore_channel_attrs(args, session);
7015 	if (ret)
7016 		return ret;
7017 	return nfs4_verify_back_channel_attrs(args, session);
7018 }
7019 
7020 static int _nfs4_proc_create_session(struct nfs_client *clp,
7021 		struct rpc_cred *cred)
7022 {
7023 	struct nfs4_session *session = clp->cl_session;
7024 	struct nfs41_create_session_args args = {
7025 		.client = clp,
7026 		.cb_program = NFS4_CALLBACK,
7027 	};
7028 	struct nfs41_create_session_res res = {
7029 		.client = clp,
7030 	};
7031 	struct rpc_message msg = {
7032 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7033 		.rpc_argp = &args,
7034 		.rpc_resp = &res,
7035 		.rpc_cred = cred,
7036 	};
7037 	int status;
7038 
7039 	nfs4_init_channel_attrs(&args);
7040 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7041 
7042 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7043 	trace_nfs4_create_session(clp, status);
7044 
7045 	if (!status) {
7046 		/* Verify the session's negotiated channel_attrs values */
7047 		status = nfs4_verify_channel_attrs(&args, session);
7048 		/* Increment the clientid slot sequence id */
7049 		clp->cl_seqid++;
7050 	}
7051 
7052 	return status;
7053 }
7054 
7055 /*
7056  * Issues a CREATE_SESSION operation to the server.
7057  * It is the responsibility of the caller to verify the session is
7058  * expired before calling this routine.
7059  */
7060 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7061 {
7062 	int status;
7063 	unsigned *ptr;
7064 	struct nfs4_session *session = clp->cl_session;
7065 
7066 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7067 
7068 	status = _nfs4_proc_create_session(clp, cred);
7069 	if (status)
7070 		goto out;
7071 
7072 	/* Init or reset the session slot tables */
7073 	status = nfs4_setup_session_slot_tables(session);
7074 	dprintk("slot table setup returned %d\n", status);
7075 	if (status)
7076 		goto out;
7077 
7078 	ptr = (unsigned *)&session->sess_id.data[0];
7079 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7080 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7081 out:
7082 	dprintk("<-- %s\n", __func__);
7083 	return status;
7084 }
7085 
7086 /*
7087  * Issue the over-the-wire RPC DESTROY_SESSION.
7088  * The caller must serialize access to this routine.
7089  */
7090 int nfs4_proc_destroy_session(struct nfs4_session *session,
7091 		struct rpc_cred *cred)
7092 {
7093 	struct rpc_message msg = {
7094 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7095 		.rpc_argp = session,
7096 		.rpc_cred = cred,
7097 	};
7098 	int status = 0;
7099 
7100 	dprintk("--> nfs4_proc_destroy_session\n");
7101 
7102 	/* session is still being setup */
7103 	if (session->clp->cl_cons_state != NFS_CS_READY)
7104 		return status;
7105 
7106 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7107 	trace_nfs4_destroy_session(session->clp, status);
7108 
7109 	if (status)
7110 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7111 			"Session has been destroyed regardless...\n", status);
7112 
7113 	dprintk("<-- nfs4_proc_destroy_session\n");
7114 	return status;
7115 }
7116 
7117 /*
7118  * Renew the cl_session lease.
7119  */
7120 struct nfs4_sequence_data {
7121 	struct nfs_client *clp;
7122 	struct nfs4_sequence_args args;
7123 	struct nfs4_sequence_res res;
7124 };
7125 
7126 static void nfs41_sequence_release(void *data)
7127 {
7128 	struct nfs4_sequence_data *calldata = data;
7129 	struct nfs_client *clp = calldata->clp;
7130 
7131 	if (atomic_read(&clp->cl_count) > 1)
7132 		nfs4_schedule_state_renewal(clp);
7133 	nfs_put_client(clp);
7134 	kfree(calldata);
7135 }
7136 
7137 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7138 {
7139 	switch(task->tk_status) {
7140 	case -NFS4ERR_DELAY:
7141 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7142 		return -EAGAIN;
7143 	default:
7144 		nfs4_schedule_lease_recovery(clp);
7145 	}
7146 	return 0;
7147 }
7148 
7149 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7150 {
7151 	struct nfs4_sequence_data *calldata = data;
7152 	struct nfs_client *clp = calldata->clp;
7153 
7154 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7155 		return;
7156 
7157 	trace_nfs4_sequence(clp, task->tk_status);
7158 	if (task->tk_status < 0) {
7159 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7160 		if (atomic_read(&clp->cl_count) == 1)
7161 			goto out;
7162 
7163 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7164 			rpc_restart_call_prepare(task);
7165 			return;
7166 		}
7167 	}
7168 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7169 out:
7170 	dprintk("<-- %s\n", __func__);
7171 }
7172 
7173 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7174 {
7175 	struct nfs4_sequence_data *calldata = data;
7176 	struct nfs_client *clp = calldata->clp;
7177 	struct nfs4_sequence_args *args;
7178 	struct nfs4_sequence_res *res;
7179 
7180 	args = task->tk_msg.rpc_argp;
7181 	res = task->tk_msg.rpc_resp;
7182 
7183 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7184 }
7185 
7186 static const struct rpc_call_ops nfs41_sequence_ops = {
7187 	.rpc_call_done = nfs41_sequence_call_done,
7188 	.rpc_call_prepare = nfs41_sequence_prepare,
7189 	.rpc_release = nfs41_sequence_release,
7190 };
7191 
7192 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7193 		struct rpc_cred *cred,
7194 		bool is_privileged)
7195 {
7196 	struct nfs4_sequence_data *calldata;
7197 	struct rpc_message msg = {
7198 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7199 		.rpc_cred = cred,
7200 	};
7201 	struct rpc_task_setup task_setup_data = {
7202 		.rpc_client = clp->cl_rpcclient,
7203 		.rpc_message = &msg,
7204 		.callback_ops = &nfs41_sequence_ops,
7205 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7206 	};
7207 
7208 	if (!atomic_inc_not_zero(&clp->cl_count))
7209 		return ERR_PTR(-EIO);
7210 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7211 	if (calldata == NULL) {
7212 		nfs_put_client(clp);
7213 		return ERR_PTR(-ENOMEM);
7214 	}
7215 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7216 	if (is_privileged)
7217 		nfs4_set_sequence_privileged(&calldata->args);
7218 	msg.rpc_argp = &calldata->args;
7219 	msg.rpc_resp = &calldata->res;
7220 	calldata->clp = clp;
7221 	task_setup_data.callback_data = calldata;
7222 
7223 	return rpc_run_task(&task_setup_data);
7224 }
7225 
7226 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7227 {
7228 	struct rpc_task *task;
7229 	int ret = 0;
7230 
7231 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7232 		return 0;
7233 	task = _nfs41_proc_sequence(clp, cred, false);
7234 	if (IS_ERR(task))
7235 		ret = PTR_ERR(task);
7236 	else
7237 		rpc_put_task_async(task);
7238 	dprintk("<-- %s status=%d\n", __func__, ret);
7239 	return ret;
7240 }
7241 
7242 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7243 {
7244 	struct rpc_task *task;
7245 	int ret;
7246 
7247 	task = _nfs41_proc_sequence(clp, cred, true);
7248 	if (IS_ERR(task)) {
7249 		ret = PTR_ERR(task);
7250 		goto out;
7251 	}
7252 	ret = rpc_wait_for_completion_task(task);
7253 	if (!ret) {
7254 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7255 
7256 		if (task->tk_status == 0)
7257 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7258 		ret = task->tk_status;
7259 	}
7260 	rpc_put_task(task);
7261 out:
7262 	dprintk("<-- %s status=%d\n", __func__, ret);
7263 	return ret;
7264 }
7265 
7266 struct nfs4_reclaim_complete_data {
7267 	struct nfs_client *clp;
7268 	struct nfs41_reclaim_complete_args arg;
7269 	struct nfs41_reclaim_complete_res res;
7270 };
7271 
7272 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7273 {
7274 	struct nfs4_reclaim_complete_data *calldata = data;
7275 
7276 	nfs41_setup_sequence(calldata->clp->cl_session,
7277 			&calldata->arg.seq_args,
7278 			&calldata->res.seq_res,
7279 			task);
7280 }
7281 
7282 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7283 {
7284 	switch(task->tk_status) {
7285 	case 0:
7286 	case -NFS4ERR_COMPLETE_ALREADY:
7287 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7288 		break;
7289 	case -NFS4ERR_DELAY:
7290 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7291 		/* fall through */
7292 	case -NFS4ERR_RETRY_UNCACHED_REP:
7293 		return -EAGAIN;
7294 	default:
7295 		nfs4_schedule_lease_recovery(clp);
7296 	}
7297 	return 0;
7298 }
7299 
7300 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7301 {
7302 	struct nfs4_reclaim_complete_data *calldata = data;
7303 	struct nfs_client *clp = calldata->clp;
7304 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7305 
7306 	dprintk("--> %s\n", __func__);
7307 	if (!nfs41_sequence_done(task, res))
7308 		return;
7309 
7310 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7311 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7312 		rpc_restart_call_prepare(task);
7313 		return;
7314 	}
7315 	dprintk("<-- %s\n", __func__);
7316 }
7317 
7318 static void nfs4_free_reclaim_complete_data(void *data)
7319 {
7320 	struct nfs4_reclaim_complete_data *calldata = data;
7321 
7322 	kfree(calldata);
7323 }
7324 
7325 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7326 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7327 	.rpc_call_done = nfs4_reclaim_complete_done,
7328 	.rpc_release = nfs4_free_reclaim_complete_data,
7329 };
7330 
7331 /*
7332  * Issue a global reclaim complete.
7333  */
7334 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7335 		struct rpc_cred *cred)
7336 {
7337 	struct nfs4_reclaim_complete_data *calldata;
7338 	struct rpc_task *task;
7339 	struct rpc_message msg = {
7340 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7341 		.rpc_cred = cred,
7342 	};
7343 	struct rpc_task_setup task_setup_data = {
7344 		.rpc_client = clp->cl_rpcclient,
7345 		.rpc_message = &msg,
7346 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7347 		.flags = RPC_TASK_ASYNC,
7348 	};
7349 	int status = -ENOMEM;
7350 
7351 	dprintk("--> %s\n", __func__);
7352 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7353 	if (calldata == NULL)
7354 		goto out;
7355 	calldata->clp = clp;
7356 	calldata->arg.one_fs = 0;
7357 
7358 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7359 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7360 	msg.rpc_argp = &calldata->arg;
7361 	msg.rpc_resp = &calldata->res;
7362 	task_setup_data.callback_data = calldata;
7363 	task = rpc_run_task(&task_setup_data);
7364 	if (IS_ERR(task)) {
7365 		status = PTR_ERR(task);
7366 		goto out;
7367 	}
7368 	status = nfs4_wait_for_completion_rpc_task(task);
7369 	if (status == 0)
7370 		status = task->tk_status;
7371 	rpc_put_task(task);
7372 	return 0;
7373 out:
7374 	dprintk("<-- %s status=%d\n", __func__, status);
7375 	return status;
7376 }
7377 
7378 static void
7379 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7380 {
7381 	struct nfs4_layoutget *lgp = calldata;
7382 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7383 	struct nfs4_session *session = nfs4_get_session(server);
7384 
7385 	dprintk("--> %s\n", __func__);
7386 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7387 	 * right now covering the LAYOUTGET we are about to send.
7388 	 * However, that is not so catastrophic, and there seems
7389 	 * to be no way to prevent it completely.
7390 	 */
7391 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7392 				&lgp->res.seq_res, task))
7393 		return;
7394 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7395 					  NFS_I(lgp->args.inode)->layout,
7396 					  lgp->args.ctx->state)) {
7397 		rpc_exit(task, NFS4_OK);
7398 	}
7399 }
7400 
7401 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7402 {
7403 	struct nfs4_layoutget *lgp = calldata;
7404 	struct inode *inode = lgp->args.inode;
7405 	struct nfs_server *server = NFS_SERVER(inode);
7406 	struct pnfs_layout_hdr *lo;
7407 	struct nfs4_state *state = NULL;
7408 	unsigned long timeo, now, giveup;
7409 
7410 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7411 
7412 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7413 		goto out;
7414 
7415 	switch (task->tk_status) {
7416 	case 0:
7417 		goto out;
7418 	/*
7419 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7420 	 * (or clients) writing to the same RAID stripe
7421 	 */
7422 	case -NFS4ERR_LAYOUTTRYLATER:
7423 	/*
7424 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7425 	 * existing layout before getting a new one).
7426 	 */
7427 	case -NFS4ERR_RECALLCONFLICT:
7428 		timeo = rpc_get_timeout(task->tk_client);
7429 		giveup = lgp->args.timestamp + timeo;
7430 		now = jiffies;
7431 		if (time_after(giveup, now)) {
7432 			unsigned long delay;
7433 
7434 			/* Delay for:
7435 			 * - Not less then NFS4_POLL_RETRY_MIN.
7436 			 * - One last time a jiffie before we give up
7437 			 * - exponential backoff (time_now minus start_attempt)
7438 			 */
7439 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7440 				    min((giveup - now - 1),
7441 					now - lgp->args.timestamp));
7442 
7443 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7444 				__func__, delay);
7445 			rpc_delay(task, delay);
7446 			task->tk_status = 0;
7447 			rpc_restart_call_prepare(task);
7448 			goto out; /* Do not call nfs4_async_handle_error() */
7449 		}
7450 		break;
7451 	case -NFS4ERR_EXPIRED:
7452 	case -NFS4ERR_BAD_STATEID:
7453 		spin_lock(&inode->i_lock);
7454 		lo = NFS_I(inode)->layout;
7455 		if (!lo || list_empty(&lo->plh_segs)) {
7456 			spin_unlock(&inode->i_lock);
7457 			/* If the open stateid was bad, then recover it. */
7458 			state = lgp->args.ctx->state;
7459 		} else {
7460 			LIST_HEAD(head);
7461 
7462 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7463 			spin_unlock(&inode->i_lock);
7464 			/* Mark the bad layout state as invalid, then
7465 			 * retry using the open stateid. */
7466 			pnfs_free_lseg_list(&head);
7467 		}
7468 	}
7469 	if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
7470 		rpc_restart_call_prepare(task);
7471 out:
7472 	dprintk("<-- %s\n", __func__);
7473 }
7474 
7475 static size_t max_response_pages(struct nfs_server *server)
7476 {
7477 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7478 	return nfs_page_array_len(0, max_resp_sz);
7479 }
7480 
7481 static void nfs4_free_pages(struct page **pages, size_t size)
7482 {
7483 	int i;
7484 
7485 	if (!pages)
7486 		return;
7487 
7488 	for (i = 0; i < size; i++) {
7489 		if (!pages[i])
7490 			break;
7491 		__free_page(pages[i]);
7492 	}
7493 	kfree(pages);
7494 }
7495 
7496 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7497 {
7498 	struct page **pages;
7499 	int i;
7500 
7501 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7502 	if (!pages) {
7503 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7504 		return NULL;
7505 	}
7506 
7507 	for (i = 0; i < size; i++) {
7508 		pages[i] = alloc_page(gfp_flags);
7509 		if (!pages[i]) {
7510 			dprintk("%s: failed to allocate page\n", __func__);
7511 			nfs4_free_pages(pages, size);
7512 			return NULL;
7513 		}
7514 	}
7515 
7516 	return pages;
7517 }
7518 
7519 static void nfs4_layoutget_release(void *calldata)
7520 {
7521 	struct nfs4_layoutget *lgp = calldata;
7522 	struct inode *inode = lgp->args.inode;
7523 	struct nfs_server *server = NFS_SERVER(inode);
7524 	size_t max_pages = max_response_pages(server);
7525 
7526 	dprintk("--> %s\n", __func__);
7527 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7528 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7529 	put_nfs_open_context(lgp->args.ctx);
7530 	kfree(calldata);
7531 	dprintk("<-- %s\n", __func__);
7532 }
7533 
7534 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7535 	.rpc_call_prepare = nfs4_layoutget_prepare,
7536 	.rpc_call_done = nfs4_layoutget_done,
7537 	.rpc_release = nfs4_layoutget_release,
7538 };
7539 
7540 struct pnfs_layout_segment *
7541 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7542 {
7543 	struct inode *inode = lgp->args.inode;
7544 	struct nfs_server *server = NFS_SERVER(inode);
7545 	size_t max_pages = max_response_pages(server);
7546 	struct rpc_task *task;
7547 	struct rpc_message msg = {
7548 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7549 		.rpc_argp = &lgp->args,
7550 		.rpc_resp = &lgp->res,
7551 		.rpc_cred = lgp->cred,
7552 	};
7553 	struct rpc_task_setup task_setup_data = {
7554 		.rpc_client = server->client,
7555 		.rpc_message = &msg,
7556 		.callback_ops = &nfs4_layoutget_call_ops,
7557 		.callback_data = lgp,
7558 		.flags = RPC_TASK_ASYNC,
7559 	};
7560 	struct pnfs_layout_segment *lseg = NULL;
7561 	int status = 0;
7562 
7563 	dprintk("--> %s\n", __func__);
7564 
7565 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7566 	if (!lgp->args.layout.pages) {
7567 		nfs4_layoutget_release(lgp);
7568 		return ERR_PTR(-ENOMEM);
7569 	}
7570 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7571 	lgp->args.timestamp = jiffies;
7572 
7573 	lgp->res.layoutp = &lgp->args.layout;
7574 	lgp->res.seq_res.sr_slot = NULL;
7575 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7576 
7577 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7578 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7579 
7580 	task = rpc_run_task(&task_setup_data);
7581 	if (IS_ERR(task))
7582 		return ERR_CAST(task);
7583 	status = nfs4_wait_for_completion_rpc_task(task);
7584 	if (status == 0)
7585 		status = task->tk_status;
7586 	trace_nfs4_layoutget(lgp->args.ctx,
7587 			&lgp->args.range,
7588 			&lgp->res.range,
7589 			status);
7590 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7591 	if (status == 0 && lgp->res.layoutp->len)
7592 		lseg = pnfs_layout_process(lgp);
7593 	rpc_put_task(task);
7594 	dprintk("<-- %s status=%d\n", __func__, status);
7595 	if (status)
7596 		return ERR_PTR(status);
7597 	return lseg;
7598 }
7599 
7600 static void
7601 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7602 {
7603 	struct nfs4_layoutreturn *lrp = calldata;
7604 
7605 	dprintk("--> %s\n", __func__);
7606 	nfs41_setup_sequence(lrp->clp->cl_session,
7607 			&lrp->args.seq_args,
7608 			&lrp->res.seq_res,
7609 			task);
7610 }
7611 
7612 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7613 {
7614 	struct nfs4_layoutreturn *lrp = calldata;
7615 	struct nfs_server *server;
7616 
7617 	dprintk("--> %s\n", __func__);
7618 
7619 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7620 		return;
7621 
7622 	server = NFS_SERVER(lrp->args.inode);
7623 	switch (task->tk_status) {
7624 	default:
7625 		task->tk_status = 0;
7626 	case 0:
7627 		break;
7628 	case -NFS4ERR_DELAY:
7629 		if (nfs4_async_handle_error(task, server, NULL) != -EAGAIN)
7630 			break;
7631 		rpc_restart_call_prepare(task);
7632 		return;
7633 	}
7634 	dprintk("<-- %s\n", __func__);
7635 }
7636 
7637 static void nfs4_layoutreturn_release(void *calldata)
7638 {
7639 	struct nfs4_layoutreturn *lrp = calldata;
7640 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7641 
7642 	dprintk("--> %s\n", __func__);
7643 	spin_lock(&lo->plh_inode->i_lock);
7644 	if (lrp->res.lrs_present)
7645 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7646 	lo->plh_block_lgets--;
7647 	spin_unlock(&lo->plh_inode->i_lock);
7648 	pnfs_put_layout_hdr(lrp->args.layout);
7649 	kfree(calldata);
7650 	dprintk("<-- %s\n", __func__);
7651 }
7652 
7653 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7654 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7655 	.rpc_call_done = nfs4_layoutreturn_done,
7656 	.rpc_release = nfs4_layoutreturn_release,
7657 };
7658 
7659 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7660 {
7661 	struct rpc_task *task;
7662 	struct rpc_message msg = {
7663 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7664 		.rpc_argp = &lrp->args,
7665 		.rpc_resp = &lrp->res,
7666 		.rpc_cred = lrp->cred,
7667 	};
7668 	struct rpc_task_setup task_setup_data = {
7669 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7670 		.rpc_message = &msg,
7671 		.callback_ops = &nfs4_layoutreturn_call_ops,
7672 		.callback_data = lrp,
7673 	};
7674 	int status;
7675 
7676 	dprintk("--> %s\n", __func__);
7677 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7678 	task = rpc_run_task(&task_setup_data);
7679 	if (IS_ERR(task))
7680 		return PTR_ERR(task);
7681 	status = task->tk_status;
7682 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7683 	dprintk("<-- %s status=%d\n", __func__, status);
7684 	rpc_put_task(task);
7685 	return status;
7686 }
7687 
7688 /*
7689  * Retrieve the list of Data Server devices from the MDS.
7690  */
7691 static int _nfs4_getdevicelist(struct nfs_server *server,
7692 				    const struct nfs_fh *fh,
7693 				    struct pnfs_devicelist *devlist)
7694 {
7695 	struct nfs4_getdevicelist_args args = {
7696 		.fh = fh,
7697 		.layoutclass = server->pnfs_curr_ld->id,
7698 	};
7699 	struct nfs4_getdevicelist_res res = {
7700 		.devlist = devlist,
7701 	};
7702 	struct rpc_message msg = {
7703 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
7704 		.rpc_argp = &args,
7705 		.rpc_resp = &res,
7706 	};
7707 	int status;
7708 
7709 	dprintk("--> %s\n", __func__);
7710 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
7711 				&res.seq_res, 0);
7712 	dprintk("<-- %s status=%d\n", __func__, status);
7713 	return status;
7714 }
7715 
7716 int nfs4_proc_getdevicelist(struct nfs_server *server,
7717 			    const struct nfs_fh *fh,
7718 			    struct pnfs_devicelist *devlist)
7719 {
7720 	struct nfs4_exception exception = { };
7721 	int err;
7722 
7723 	do {
7724 		err = nfs4_handle_exception(server,
7725 				_nfs4_getdevicelist(server, fh, devlist),
7726 				&exception);
7727 	} while (exception.retry);
7728 
7729 	dprintk("%s: err=%d, num_devs=%u\n", __func__,
7730 		err, devlist->num_devs);
7731 
7732 	return err;
7733 }
7734 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
7735 
7736 static int
7737 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7738 		struct pnfs_device *pdev,
7739 		struct rpc_cred *cred)
7740 {
7741 	struct nfs4_getdeviceinfo_args args = {
7742 		.pdev = pdev,
7743 	};
7744 	struct nfs4_getdeviceinfo_res res = {
7745 		.pdev = pdev,
7746 	};
7747 	struct rpc_message msg = {
7748 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7749 		.rpc_argp = &args,
7750 		.rpc_resp = &res,
7751 		.rpc_cred = cred,
7752 	};
7753 	int status;
7754 
7755 	dprintk("--> %s\n", __func__);
7756 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7757 	dprintk("<-- %s status=%d\n", __func__, status);
7758 
7759 	return status;
7760 }
7761 
7762 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7763 		struct pnfs_device *pdev,
7764 		struct rpc_cred *cred)
7765 {
7766 	struct nfs4_exception exception = { };
7767 	int err;
7768 
7769 	do {
7770 		err = nfs4_handle_exception(server,
7771 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7772 					&exception);
7773 	} while (exception.retry);
7774 	return err;
7775 }
7776 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7777 
7778 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7779 {
7780 	struct nfs4_layoutcommit_data *data = calldata;
7781 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7782 	struct nfs4_session *session = nfs4_get_session(server);
7783 
7784 	nfs41_setup_sequence(session,
7785 			&data->args.seq_args,
7786 			&data->res.seq_res,
7787 			task);
7788 }
7789 
7790 static void
7791 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7792 {
7793 	struct nfs4_layoutcommit_data *data = calldata;
7794 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7795 
7796 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7797 		return;
7798 
7799 	switch (task->tk_status) { /* Just ignore these failures */
7800 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7801 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7802 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7803 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7804 		task->tk_status = 0;
7805 	case 0:
7806 		break;
7807 	default:
7808 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7809 			rpc_restart_call_prepare(task);
7810 			return;
7811 		}
7812 	}
7813 }
7814 
7815 static void nfs4_layoutcommit_release(void *calldata)
7816 {
7817 	struct nfs4_layoutcommit_data *data = calldata;
7818 
7819 	pnfs_cleanup_layoutcommit(data);
7820 	nfs_post_op_update_inode_force_wcc(data->args.inode,
7821 					   data->res.fattr);
7822 	put_rpccred(data->cred);
7823 	kfree(data);
7824 }
7825 
7826 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7827 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
7828 	.rpc_call_done = nfs4_layoutcommit_done,
7829 	.rpc_release = nfs4_layoutcommit_release,
7830 };
7831 
7832 int
7833 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7834 {
7835 	struct rpc_message msg = {
7836 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7837 		.rpc_argp = &data->args,
7838 		.rpc_resp = &data->res,
7839 		.rpc_cred = data->cred,
7840 	};
7841 	struct rpc_task_setup task_setup_data = {
7842 		.task = &data->task,
7843 		.rpc_client = NFS_CLIENT(data->args.inode),
7844 		.rpc_message = &msg,
7845 		.callback_ops = &nfs4_layoutcommit_ops,
7846 		.callback_data = data,
7847 		.flags = RPC_TASK_ASYNC,
7848 	};
7849 	struct rpc_task *task;
7850 	int status = 0;
7851 
7852 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7853 		"lbw: %llu inode %lu\n",
7854 		data->task.tk_pid, sync,
7855 		data->args.lastbytewritten,
7856 		data->args.inode->i_ino);
7857 
7858 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7859 	task = rpc_run_task(&task_setup_data);
7860 	if (IS_ERR(task))
7861 		return PTR_ERR(task);
7862 	if (sync == false)
7863 		goto out;
7864 	status = nfs4_wait_for_completion_rpc_task(task);
7865 	if (status != 0)
7866 		goto out;
7867 	status = task->tk_status;
7868 	trace_nfs4_layoutcommit(data->args.inode, status);
7869 out:
7870 	dprintk("%s: status %d\n", __func__, status);
7871 	rpc_put_task(task);
7872 	return status;
7873 }
7874 
7875 /**
7876  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7877  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7878  */
7879 static int
7880 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7881 		    struct nfs_fsinfo *info,
7882 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7883 {
7884 	struct nfs41_secinfo_no_name_args args = {
7885 		.style = SECINFO_STYLE_CURRENT_FH,
7886 	};
7887 	struct nfs4_secinfo_res res = {
7888 		.flavors = flavors,
7889 	};
7890 	struct rpc_message msg = {
7891 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
7892 		.rpc_argp = &args,
7893 		.rpc_resp = &res,
7894 	};
7895 	struct rpc_clnt *clnt = server->client;
7896 	struct rpc_cred *cred = NULL;
7897 	int status;
7898 
7899 	if (use_integrity) {
7900 		clnt = server->nfs_client->cl_rpcclient;
7901 		cred = nfs4_get_clid_cred(server->nfs_client);
7902 		msg.rpc_cred = cred;
7903 	}
7904 
7905 	dprintk("--> %s\n", __func__);
7906 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
7907 				&res.seq_res, 0);
7908 	dprintk("<-- %s status=%d\n", __func__, status);
7909 
7910 	if (cred)
7911 		put_rpccred(cred);
7912 
7913 	return status;
7914 }
7915 
7916 static int
7917 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7918 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
7919 {
7920 	struct nfs4_exception exception = { };
7921 	int err;
7922 	do {
7923 		/* first try using integrity protection */
7924 		err = -NFS4ERR_WRONGSEC;
7925 
7926 		/* try to use integrity protection with machine cred */
7927 		if (_nfs4_is_integrity_protected(server->nfs_client))
7928 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
7929 							  flavors, true);
7930 
7931 		/*
7932 		 * if unable to use integrity protection, or SECINFO with
7933 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
7934 		 * disallowed by spec, but exists in deployed servers) use
7935 		 * the current filesystem's rpc_client and the user cred.
7936 		 */
7937 		if (err == -NFS4ERR_WRONGSEC)
7938 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
7939 							  flavors, false);
7940 
7941 		switch (err) {
7942 		case 0:
7943 		case -NFS4ERR_WRONGSEC:
7944 		case -ENOTSUPP:
7945 			goto out;
7946 		default:
7947 			err = nfs4_handle_exception(server, err, &exception);
7948 		}
7949 	} while (exception.retry);
7950 out:
7951 	return err;
7952 }
7953 
7954 static int
7955 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
7956 		    struct nfs_fsinfo *info)
7957 {
7958 	int err;
7959 	struct page *page;
7960 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
7961 	struct nfs4_secinfo_flavors *flavors;
7962 	struct nfs4_secinfo4 *secinfo;
7963 	int i;
7964 
7965 	page = alloc_page(GFP_KERNEL);
7966 	if (!page) {
7967 		err = -ENOMEM;
7968 		goto out;
7969 	}
7970 
7971 	flavors = page_address(page);
7972 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
7973 
7974 	/*
7975 	 * Fall back on "guess and check" method if
7976 	 * the server doesn't support SECINFO_NO_NAME
7977 	 */
7978 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
7979 		err = nfs4_find_root_sec(server, fhandle, info);
7980 		goto out_freepage;
7981 	}
7982 	if (err)
7983 		goto out_freepage;
7984 
7985 	for (i = 0; i < flavors->num_flavors; i++) {
7986 		secinfo = &flavors->flavors[i];
7987 
7988 		switch (secinfo->flavor) {
7989 		case RPC_AUTH_NULL:
7990 		case RPC_AUTH_UNIX:
7991 		case RPC_AUTH_GSS:
7992 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
7993 					&secinfo->flavor_info);
7994 			break;
7995 		default:
7996 			flavor = RPC_AUTH_MAXFLAVOR;
7997 			break;
7998 		}
7999 
8000 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8001 			flavor = RPC_AUTH_MAXFLAVOR;
8002 
8003 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8004 			err = nfs4_lookup_root_sec(server, fhandle,
8005 						   info, flavor);
8006 			if (!err)
8007 				break;
8008 		}
8009 	}
8010 
8011 	if (flavor == RPC_AUTH_MAXFLAVOR)
8012 		err = -EPERM;
8013 
8014 out_freepage:
8015 	put_page(page);
8016 	if (err == -EACCES)
8017 		return -EPERM;
8018 out:
8019 	return err;
8020 }
8021 
8022 static int _nfs41_test_stateid(struct nfs_server *server,
8023 		nfs4_stateid *stateid,
8024 		struct rpc_cred *cred)
8025 {
8026 	int status;
8027 	struct nfs41_test_stateid_args args = {
8028 		.stateid = stateid,
8029 	};
8030 	struct nfs41_test_stateid_res res;
8031 	struct rpc_message msg = {
8032 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8033 		.rpc_argp = &args,
8034 		.rpc_resp = &res,
8035 		.rpc_cred = cred,
8036 	};
8037 	struct rpc_clnt *rpc_client = server->client;
8038 
8039 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8040 		&rpc_client, &msg);
8041 
8042 	dprintk("NFS call  test_stateid %p\n", stateid);
8043 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8044 	nfs4_set_sequence_privileged(&args.seq_args);
8045 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8046 			&args.seq_args, &res.seq_res);
8047 	if (status != NFS_OK) {
8048 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8049 		return status;
8050 	}
8051 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8052 	return -res.status;
8053 }
8054 
8055 /**
8056  * nfs41_test_stateid - perform a TEST_STATEID operation
8057  *
8058  * @server: server / transport on which to perform the operation
8059  * @stateid: state ID to test
8060  * @cred: credential
8061  *
8062  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8063  * Otherwise a negative NFS4ERR value is returned if the operation
8064  * failed or the state ID is not currently valid.
8065  */
8066 static int nfs41_test_stateid(struct nfs_server *server,
8067 		nfs4_stateid *stateid,
8068 		struct rpc_cred *cred)
8069 {
8070 	struct nfs4_exception exception = { };
8071 	int err;
8072 	do {
8073 		err = _nfs41_test_stateid(server, stateid, cred);
8074 		if (err != -NFS4ERR_DELAY)
8075 			break;
8076 		nfs4_handle_exception(server, err, &exception);
8077 	} while (exception.retry);
8078 	return err;
8079 }
8080 
8081 struct nfs_free_stateid_data {
8082 	struct nfs_server *server;
8083 	struct nfs41_free_stateid_args args;
8084 	struct nfs41_free_stateid_res res;
8085 };
8086 
8087 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8088 {
8089 	struct nfs_free_stateid_data *data = calldata;
8090 	nfs41_setup_sequence(nfs4_get_session(data->server),
8091 			&data->args.seq_args,
8092 			&data->res.seq_res,
8093 			task);
8094 }
8095 
8096 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8097 {
8098 	struct nfs_free_stateid_data *data = calldata;
8099 
8100 	nfs41_sequence_done(task, &data->res.seq_res);
8101 
8102 	switch (task->tk_status) {
8103 	case -NFS4ERR_DELAY:
8104 		if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
8105 			rpc_restart_call_prepare(task);
8106 	}
8107 }
8108 
8109 static void nfs41_free_stateid_release(void *calldata)
8110 {
8111 	kfree(calldata);
8112 }
8113 
8114 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8115 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8116 	.rpc_call_done = nfs41_free_stateid_done,
8117 	.rpc_release = nfs41_free_stateid_release,
8118 };
8119 
8120 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8121 		nfs4_stateid *stateid,
8122 		struct rpc_cred *cred,
8123 		bool privileged)
8124 {
8125 	struct rpc_message msg = {
8126 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8127 		.rpc_cred = cred,
8128 	};
8129 	struct rpc_task_setup task_setup = {
8130 		.rpc_client = server->client,
8131 		.rpc_message = &msg,
8132 		.callback_ops = &nfs41_free_stateid_ops,
8133 		.flags = RPC_TASK_ASYNC,
8134 	};
8135 	struct nfs_free_stateid_data *data;
8136 
8137 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8138 		&task_setup.rpc_client, &msg);
8139 
8140 	dprintk("NFS call  free_stateid %p\n", stateid);
8141 	data = kmalloc(sizeof(*data), GFP_NOFS);
8142 	if (!data)
8143 		return ERR_PTR(-ENOMEM);
8144 	data->server = server;
8145 	nfs4_stateid_copy(&data->args.stateid, stateid);
8146 
8147 	task_setup.callback_data = data;
8148 
8149 	msg.rpc_argp = &data->args;
8150 	msg.rpc_resp = &data->res;
8151 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8152 	if (privileged)
8153 		nfs4_set_sequence_privileged(&data->args.seq_args);
8154 
8155 	return rpc_run_task(&task_setup);
8156 }
8157 
8158 /**
8159  * nfs41_free_stateid - perform a FREE_STATEID operation
8160  *
8161  * @server: server / transport on which to perform the operation
8162  * @stateid: state ID to release
8163  * @cred: credential
8164  *
8165  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8166  * negative NFS4ERR value is returned.
8167  */
8168 static int nfs41_free_stateid(struct nfs_server *server,
8169 		nfs4_stateid *stateid,
8170 		struct rpc_cred *cred)
8171 {
8172 	struct rpc_task *task;
8173 	int ret;
8174 
8175 	task = _nfs41_free_stateid(server, stateid, cred, true);
8176 	if (IS_ERR(task))
8177 		return PTR_ERR(task);
8178 	ret = rpc_wait_for_completion_task(task);
8179 	if (!ret)
8180 		ret = task->tk_status;
8181 	rpc_put_task(task);
8182 	return ret;
8183 }
8184 
8185 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8186 {
8187 	struct rpc_task *task;
8188 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8189 
8190 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8191 	nfs4_free_lock_state(server, lsp);
8192 	if (IS_ERR(task))
8193 		return PTR_ERR(task);
8194 	rpc_put_task(task);
8195 	return 0;
8196 }
8197 
8198 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8199 		const nfs4_stateid *s2)
8200 {
8201 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8202 		return false;
8203 
8204 	if (s1->seqid == s2->seqid)
8205 		return true;
8206 	if (s1->seqid == 0 || s2->seqid == 0)
8207 		return true;
8208 
8209 	return false;
8210 }
8211 
8212 #endif /* CONFIG_NFS_V4_1 */
8213 
8214 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8215 		const nfs4_stateid *s2)
8216 {
8217 	return nfs4_stateid_match(s1, s2);
8218 }
8219 
8220 
8221 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8222 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8223 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8224 	.recover_open	= nfs4_open_reclaim,
8225 	.recover_lock	= nfs4_lock_reclaim,
8226 	.establish_clid = nfs4_init_clientid,
8227 	.detect_trunking = nfs40_discover_server_trunking,
8228 };
8229 
8230 #if defined(CONFIG_NFS_V4_1)
8231 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8232 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8233 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8234 	.recover_open	= nfs4_open_reclaim,
8235 	.recover_lock	= nfs4_lock_reclaim,
8236 	.establish_clid = nfs41_init_clientid,
8237 	.reclaim_complete = nfs41_proc_reclaim_complete,
8238 	.detect_trunking = nfs41_discover_server_trunking,
8239 };
8240 #endif /* CONFIG_NFS_V4_1 */
8241 
8242 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8243 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8244 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8245 	.recover_open	= nfs4_open_expired,
8246 	.recover_lock	= nfs4_lock_expired,
8247 	.establish_clid = nfs4_init_clientid,
8248 };
8249 
8250 #if defined(CONFIG_NFS_V4_1)
8251 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8252 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8253 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8254 	.recover_open	= nfs41_open_expired,
8255 	.recover_lock	= nfs41_lock_expired,
8256 	.establish_clid = nfs41_init_clientid,
8257 };
8258 #endif /* CONFIG_NFS_V4_1 */
8259 
8260 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8261 	.sched_state_renewal = nfs4_proc_async_renew,
8262 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8263 	.renew_lease = nfs4_proc_renew,
8264 };
8265 
8266 #if defined(CONFIG_NFS_V4_1)
8267 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8268 	.sched_state_renewal = nfs41_proc_async_sequence,
8269 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8270 	.renew_lease = nfs4_proc_sequence,
8271 };
8272 #endif
8273 
8274 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8275 	.get_locations = _nfs40_proc_get_locations,
8276 	.fsid_present = _nfs40_proc_fsid_present,
8277 };
8278 
8279 #if defined(CONFIG_NFS_V4_1)
8280 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8281 	.get_locations = _nfs41_proc_get_locations,
8282 	.fsid_present = _nfs41_proc_fsid_present,
8283 };
8284 #endif	/* CONFIG_NFS_V4_1 */
8285 
8286 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8287 	.minor_version = 0,
8288 	.init_caps = NFS_CAP_READDIRPLUS
8289 		| NFS_CAP_ATOMIC_OPEN
8290 		| NFS_CAP_CHANGE_ATTR
8291 		| NFS_CAP_POSIX_LOCK,
8292 	.init_client = nfs40_init_client,
8293 	.shutdown_client = nfs40_shutdown_client,
8294 	.match_stateid = nfs4_match_stateid,
8295 	.find_root_sec = nfs4_find_root_sec,
8296 	.free_lock_state = nfs4_release_lockowner,
8297 	.call_sync_ops = &nfs40_call_sync_ops,
8298 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8299 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8300 	.state_renewal_ops = &nfs40_state_renewal_ops,
8301 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8302 };
8303 
8304 #if defined(CONFIG_NFS_V4_1)
8305 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8306 	.minor_version = 1,
8307 	.init_caps = NFS_CAP_READDIRPLUS
8308 		| NFS_CAP_ATOMIC_OPEN
8309 		| NFS_CAP_CHANGE_ATTR
8310 		| NFS_CAP_POSIX_LOCK
8311 		| NFS_CAP_STATEID_NFSV41
8312 		| NFS_CAP_ATOMIC_OPEN_V1,
8313 	.init_client = nfs41_init_client,
8314 	.shutdown_client = nfs41_shutdown_client,
8315 	.match_stateid = nfs41_match_stateid,
8316 	.find_root_sec = nfs41_find_root_sec,
8317 	.free_lock_state = nfs41_free_lock_state,
8318 	.call_sync_ops = &nfs41_call_sync_ops,
8319 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8320 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8321 	.state_renewal_ops = &nfs41_state_renewal_ops,
8322 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8323 };
8324 #endif
8325 
8326 #if defined(CONFIG_NFS_V4_2)
8327 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8328 	.minor_version = 2,
8329 	.init_caps = NFS_CAP_READDIRPLUS
8330 		| NFS_CAP_ATOMIC_OPEN
8331 		| NFS_CAP_CHANGE_ATTR
8332 		| NFS_CAP_POSIX_LOCK
8333 		| NFS_CAP_STATEID_NFSV41
8334 		| NFS_CAP_ATOMIC_OPEN_V1,
8335 	.init_client = nfs41_init_client,
8336 	.shutdown_client = nfs41_shutdown_client,
8337 	.match_stateid = nfs41_match_stateid,
8338 	.find_root_sec = nfs41_find_root_sec,
8339 	.free_lock_state = nfs41_free_lock_state,
8340 	.call_sync_ops = &nfs41_call_sync_ops,
8341 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8342 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8343 	.state_renewal_ops = &nfs41_state_renewal_ops,
8344 };
8345 #endif
8346 
8347 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8348 	[0] = &nfs_v4_0_minor_ops,
8349 #if defined(CONFIG_NFS_V4_1)
8350 	[1] = &nfs_v4_1_minor_ops,
8351 #endif
8352 #if defined(CONFIG_NFS_V4_2)
8353 	[2] = &nfs_v4_2_minor_ops,
8354 #endif
8355 };
8356 
8357 static const struct inode_operations nfs4_dir_inode_operations = {
8358 	.create		= nfs_create,
8359 	.lookup		= nfs_lookup,
8360 	.atomic_open	= nfs_atomic_open,
8361 	.link		= nfs_link,
8362 	.unlink		= nfs_unlink,
8363 	.symlink	= nfs_symlink,
8364 	.mkdir		= nfs_mkdir,
8365 	.rmdir		= nfs_rmdir,
8366 	.mknod		= nfs_mknod,
8367 	.rename		= nfs_rename,
8368 	.permission	= nfs_permission,
8369 	.getattr	= nfs_getattr,
8370 	.setattr	= nfs_setattr,
8371 	.getxattr	= generic_getxattr,
8372 	.setxattr	= generic_setxattr,
8373 	.listxattr	= generic_listxattr,
8374 	.removexattr	= generic_removexattr,
8375 };
8376 
8377 static const struct inode_operations nfs4_file_inode_operations = {
8378 	.permission	= nfs_permission,
8379 	.getattr	= nfs_getattr,
8380 	.setattr	= nfs_setattr,
8381 	.getxattr	= generic_getxattr,
8382 	.setxattr	= generic_setxattr,
8383 	.listxattr	= generic_listxattr,
8384 	.removexattr	= generic_removexattr,
8385 };
8386 
8387 const struct nfs_rpc_ops nfs_v4_clientops = {
8388 	.version	= 4,			/* protocol version */
8389 	.dentry_ops	= &nfs4_dentry_operations,
8390 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8391 	.file_inode_ops	= &nfs4_file_inode_operations,
8392 	.file_ops	= &nfs4_file_operations,
8393 	.getroot	= nfs4_proc_get_root,
8394 	.submount	= nfs4_submount,
8395 	.try_mount	= nfs4_try_mount,
8396 	.getattr	= nfs4_proc_getattr,
8397 	.setattr	= nfs4_proc_setattr,
8398 	.lookup		= nfs4_proc_lookup,
8399 	.access		= nfs4_proc_access,
8400 	.readlink	= nfs4_proc_readlink,
8401 	.create		= nfs4_proc_create,
8402 	.remove		= nfs4_proc_remove,
8403 	.unlink_setup	= nfs4_proc_unlink_setup,
8404 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8405 	.unlink_done	= nfs4_proc_unlink_done,
8406 	.rename_setup	= nfs4_proc_rename_setup,
8407 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8408 	.rename_done	= nfs4_proc_rename_done,
8409 	.link		= nfs4_proc_link,
8410 	.symlink	= nfs4_proc_symlink,
8411 	.mkdir		= nfs4_proc_mkdir,
8412 	.rmdir		= nfs4_proc_remove,
8413 	.readdir	= nfs4_proc_readdir,
8414 	.mknod		= nfs4_proc_mknod,
8415 	.statfs		= nfs4_proc_statfs,
8416 	.fsinfo		= nfs4_proc_fsinfo,
8417 	.pathconf	= nfs4_proc_pathconf,
8418 	.set_capabilities = nfs4_server_capabilities,
8419 	.decode_dirent	= nfs4_decode_dirent,
8420 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8421 	.read_setup	= nfs4_proc_read_setup,
8422 	.read_done	= nfs4_read_done,
8423 	.write_setup	= nfs4_proc_write_setup,
8424 	.write_done	= nfs4_write_done,
8425 	.commit_setup	= nfs4_proc_commit_setup,
8426 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8427 	.commit_done	= nfs4_commit_done,
8428 	.lock		= nfs4_proc_lock,
8429 	.clear_acl_cache = nfs4_zap_acl_attr,
8430 	.close_context  = nfs4_close_context,
8431 	.open_context	= nfs4_atomic_open,
8432 	.have_delegation = nfs4_have_delegation,
8433 	.return_delegation = nfs4_inode_return_delegation,
8434 	.alloc_client	= nfs4_alloc_client,
8435 	.init_client	= nfs4_init_client,
8436 	.free_client	= nfs4_free_client,
8437 	.create_server	= nfs4_create_server,
8438 	.clone_server	= nfs_clone_server,
8439 };
8440 
8441 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8442 	.prefix	= XATTR_NAME_NFSV4_ACL,
8443 	.list	= nfs4_xattr_list_nfs4_acl,
8444 	.get	= nfs4_xattr_get_nfs4_acl,
8445 	.set	= nfs4_xattr_set_nfs4_acl,
8446 };
8447 
8448 const struct xattr_handler *nfs4_xattr_handlers[] = {
8449 	&nfs4_xattr_nfs4_acl_handler,
8450 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8451 	&nfs4_xattr_nfs4_label_handler,
8452 #endif
8453 	NULL
8454 };
8455 
8456 /*
8457  * Local variables:
8458  *  c-basic-offset: 8
8459  * End:
8460  */
8461