xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 12eb4683)
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 (!RPC_WAS_SENT(task))
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 	bool send_new_highest_used_slotid = false;
563 
564 	if (!res->sr_slot) {
565 		/* just wake up the next guy waiting since
566 		 * we may have not consumed a slot after all */
567 		dprintk("%s: No slot\n", __func__);
568 		return;
569 	}
570 	tbl = res->sr_slot->table;
571 	session = tbl->session;
572 
573 	spin_lock(&tbl->slot_tbl_lock);
574 	/* Be nice to the server: try to ensure that the last transmitted
575 	 * value for highest_user_slotid <= target_highest_slotid
576 	 */
577 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
578 		send_new_highest_used_slotid = true;
579 
580 	if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
581 		send_new_highest_used_slotid = false;
582 		goto out_unlock;
583 	}
584 	nfs4_free_slot(tbl, res->sr_slot);
585 
586 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
587 		send_new_highest_used_slotid = false;
588 out_unlock:
589 	spin_unlock(&tbl->slot_tbl_lock);
590 	res->sr_slot = NULL;
591 	if (send_new_highest_used_slotid)
592 		nfs41_server_notify_highest_slotid_update(session->clp);
593 }
594 
595 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
596 {
597 	struct nfs4_session *session;
598 	struct nfs4_slot *slot;
599 	struct nfs_client *clp;
600 	bool interrupted = false;
601 	int ret = 1;
602 
603 	/* don't increment the sequence number if the task wasn't sent */
604 	if (!RPC_WAS_SENT(task))
605 		goto out;
606 
607 	slot = res->sr_slot;
608 	session = slot->table->session;
609 
610 	if (slot->interrupted) {
611 		slot->interrupted = 0;
612 		interrupted = true;
613 	}
614 
615 	trace_nfs4_sequence_done(session, res);
616 	/* Check the SEQUENCE operation status */
617 	switch (res->sr_status) {
618 	case 0:
619 		/* Update the slot's sequence and clientid lease timer */
620 		++slot->seq_nr;
621 		clp = session->clp;
622 		do_renew_lease(clp, res->sr_timestamp);
623 		/* Check sequence flags */
624 		if (res->sr_status_flags != 0)
625 			nfs4_schedule_lease_recovery(clp);
626 		nfs41_update_target_slotid(slot->table, slot, res);
627 		break;
628 	case 1:
629 		/*
630 		 * sr_status remains 1 if an RPC level error occurred.
631 		 * The server may or may not have processed the sequence
632 		 * operation..
633 		 * Mark the slot as having hosted an interrupted RPC call.
634 		 */
635 		slot->interrupted = 1;
636 		goto out;
637 	case -NFS4ERR_DELAY:
638 		/* The server detected a resend of the RPC call and
639 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
640 		 * of RFC5661.
641 		 */
642 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
643 			__func__,
644 			slot->slot_nr,
645 			slot->seq_nr);
646 		goto out_retry;
647 	case -NFS4ERR_BADSLOT:
648 		/*
649 		 * The slot id we used was probably retired. Try again
650 		 * using a different slot id.
651 		 */
652 		goto retry_nowait;
653 	case -NFS4ERR_SEQ_MISORDERED:
654 		/*
655 		 * Was the last operation on this sequence interrupted?
656 		 * If so, retry after bumping the sequence number.
657 		 */
658 		if (interrupted) {
659 			++slot->seq_nr;
660 			goto retry_nowait;
661 		}
662 		/*
663 		 * Could this slot have been previously retired?
664 		 * If so, then the server may be expecting seq_nr = 1!
665 		 */
666 		if (slot->seq_nr != 1) {
667 			slot->seq_nr = 1;
668 			goto retry_nowait;
669 		}
670 		break;
671 	case -NFS4ERR_SEQ_FALSE_RETRY:
672 		++slot->seq_nr;
673 		goto retry_nowait;
674 	default:
675 		/* Just update the slot sequence no. */
676 		++slot->seq_nr;
677 	}
678 out:
679 	/* The session may be reset by one of the error handlers. */
680 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
681 	nfs41_sequence_free_slot(res);
682 	return ret;
683 retry_nowait:
684 	if (rpc_restart_call_prepare(task)) {
685 		task->tk_status = 0;
686 		ret = 0;
687 	}
688 	goto out;
689 out_retry:
690 	if (!rpc_restart_call(task))
691 		goto out;
692 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
693 	return 0;
694 }
695 
696 static int nfs4_sequence_done(struct rpc_task *task,
697 			       struct nfs4_sequence_res *res)
698 {
699 	if (res->sr_slot == NULL)
700 		return 1;
701 	if (!res->sr_slot->table->session)
702 		return nfs40_sequence_done(task, res);
703 	return nfs41_sequence_done(task, res);
704 }
705 
706 int nfs41_setup_sequence(struct nfs4_session *session,
707 				struct nfs4_sequence_args *args,
708 				struct nfs4_sequence_res *res,
709 				struct rpc_task *task)
710 {
711 	struct nfs4_slot *slot;
712 	struct nfs4_slot_table *tbl;
713 
714 	dprintk("--> %s\n", __func__);
715 	/* slot already allocated? */
716 	if (res->sr_slot != NULL)
717 		goto out_success;
718 
719 	tbl = &session->fc_slot_table;
720 
721 	task->tk_timeout = 0;
722 
723 	spin_lock(&tbl->slot_tbl_lock);
724 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
725 	    !args->sa_privileged) {
726 		/* The state manager will wait until the slot table is empty */
727 		dprintk("%s session is draining\n", __func__);
728 		goto out_sleep;
729 	}
730 
731 	slot = nfs4_alloc_slot(tbl);
732 	if (IS_ERR(slot)) {
733 		/* If out of memory, try again in 1/4 second */
734 		if (slot == ERR_PTR(-ENOMEM))
735 			task->tk_timeout = HZ >> 2;
736 		dprintk("<-- %s: no free slots\n", __func__);
737 		goto out_sleep;
738 	}
739 	spin_unlock(&tbl->slot_tbl_lock);
740 
741 	args->sa_slot = slot;
742 
743 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
744 			slot->slot_nr, slot->seq_nr);
745 
746 	res->sr_slot = slot;
747 	res->sr_timestamp = jiffies;
748 	res->sr_status_flags = 0;
749 	/*
750 	 * sr_status is only set in decode_sequence, and so will remain
751 	 * set to 1 if an rpc level failure occurs.
752 	 */
753 	res->sr_status = 1;
754 	trace_nfs4_setup_sequence(session, args);
755 out_success:
756 	rpc_call_start(task);
757 	return 0;
758 out_sleep:
759 	/* Privileged tasks are queued with top priority */
760 	if (args->sa_privileged)
761 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
762 				NULL, RPC_PRIORITY_PRIVILEGED);
763 	else
764 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
765 	spin_unlock(&tbl->slot_tbl_lock);
766 	return -EAGAIN;
767 }
768 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
769 
770 static int nfs4_setup_sequence(const struct nfs_server *server,
771 			       struct nfs4_sequence_args *args,
772 			       struct nfs4_sequence_res *res,
773 			       struct rpc_task *task)
774 {
775 	struct nfs4_session *session = nfs4_get_session(server);
776 	int ret = 0;
777 
778 	if (!session)
779 		return nfs40_setup_sequence(server, args, res, task);
780 
781 	dprintk("--> %s clp %p session %p sr_slot %u\n",
782 		__func__, session->clp, session, res->sr_slot ?
783 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
784 
785 	ret = nfs41_setup_sequence(session, args, res, task);
786 
787 	dprintk("<-- %s status=%d\n", __func__, ret);
788 	return ret;
789 }
790 
791 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
792 {
793 	struct nfs4_call_sync_data *data = calldata;
794 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
795 
796 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
797 
798 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
799 }
800 
801 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
802 {
803 	struct nfs4_call_sync_data *data = calldata;
804 
805 	nfs41_sequence_done(task, data->seq_res);
806 }
807 
808 static const struct rpc_call_ops nfs41_call_sync_ops = {
809 	.rpc_call_prepare = nfs41_call_sync_prepare,
810 	.rpc_call_done = nfs41_call_sync_done,
811 };
812 
813 #else	/* !CONFIG_NFS_V4_1 */
814 
815 static int nfs4_setup_sequence(const struct nfs_server *server,
816 			       struct nfs4_sequence_args *args,
817 			       struct nfs4_sequence_res *res,
818 			       struct rpc_task *task)
819 {
820 	return nfs40_setup_sequence(server, args, res, task);
821 }
822 
823 static int nfs4_sequence_done(struct rpc_task *task,
824 			       struct nfs4_sequence_res *res)
825 {
826 	return nfs40_sequence_done(task, res);
827 }
828 
829 #endif	/* !CONFIG_NFS_V4_1 */
830 
831 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
832 {
833 	struct nfs4_call_sync_data *data = calldata;
834 	nfs4_setup_sequence(data->seq_server,
835 				data->seq_args, data->seq_res, task);
836 }
837 
838 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
839 {
840 	struct nfs4_call_sync_data *data = calldata;
841 	nfs4_sequence_done(task, data->seq_res);
842 }
843 
844 static const struct rpc_call_ops nfs40_call_sync_ops = {
845 	.rpc_call_prepare = nfs40_call_sync_prepare,
846 	.rpc_call_done = nfs40_call_sync_done,
847 };
848 
849 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
850 				   struct nfs_server *server,
851 				   struct rpc_message *msg,
852 				   struct nfs4_sequence_args *args,
853 				   struct nfs4_sequence_res *res)
854 {
855 	int ret;
856 	struct rpc_task *task;
857 	struct nfs_client *clp = server->nfs_client;
858 	struct nfs4_call_sync_data data = {
859 		.seq_server = server,
860 		.seq_args = args,
861 		.seq_res = res,
862 	};
863 	struct rpc_task_setup task_setup = {
864 		.rpc_client = clnt,
865 		.rpc_message = msg,
866 		.callback_ops = clp->cl_mvops->call_sync_ops,
867 		.callback_data = &data
868 	};
869 
870 	task = rpc_run_task(&task_setup);
871 	if (IS_ERR(task))
872 		ret = PTR_ERR(task);
873 	else {
874 		ret = task->tk_status;
875 		rpc_put_task(task);
876 	}
877 	return ret;
878 }
879 
880 static
881 int nfs4_call_sync(struct rpc_clnt *clnt,
882 		   struct nfs_server *server,
883 		   struct rpc_message *msg,
884 		   struct nfs4_sequence_args *args,
885 		   struct nfs4_sequence_res *res,
886 		   int cache_reply)
887 {
888 	nfs4_init_sequence(args, res, cache_reply);
889 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
890 }
891 
892 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
893 {
894 	struct nfs_inode *nfsi = NFS_I(dir);
895 
896 	spin_lock(&dir->i_lock);
897 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
898 	if (!cinfo->atomic || cinfo->before != dir->i_version)
899 		nfs_force_lookup_revalidate(dir);
900 	dir->i_version = cinfo->after;
901 	nfs_fscache_invalidate(dir);
902 	spin_unlock(&dir->i_lock);
903 }
904 
905 struct nfs4_opendata {
906 	struct kref kref;
907 	struct nfs_openargs o_arg;
908 	struct nfs_openres o_res;
909 	struct nfs_open_confirmargs c_arg;
910 	struct nfs_open_confirmres c_res;
911 	struct nfs4_string owner_name;
912 	struct nfs4_string group_name;
913 	struct nfs_fattr f_attr;
914 	struct nfs4_label *f_label;
915 	struct dentry *dir;
916 	struct dentry *dentry;
917 	struct nfs4_state_owner *owner;
918 	struct nfs4_state *state;
919 	struct iattr attrs;
920 	unsigned long timestamp;
921 	unsigned int rpc_done : 1;
922 	unsigned int file_created : 1;
923 	unsigned int is_recover : 1;
924 	int rpc_status;
925 	int cancelled;
926 };
927 
928 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
929 		int err, struct nfs4_exception *exception)
930 {
931 	if (err != -EINVAL)
932 		return false;
933 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
934 		return false;
935 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
936 	exception->retry = 1;
937 	return true;
938 }
939 
940 static enum open_claim_type4
941 nfs4_map_atomic_open_claim(struct nfs_server *server,
942 		enum open_claim_type4 claim)
943 {
944 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
945 		return claim;
946 	switch (claim) {
947 	default:
948 		return claim;
949 	case NFS4_OPEN_CLAIM_FH:
950 		return NFS4_OPEN_CLAIM_NULL;
951 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
952 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
953 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
954 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
955 	}
956 }
957 
958 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
959 {
960 	p->o_res.f_attr = &p->f_attr;
961 	p->o_res.f_label = p->f_label;
962 	p->o_res.seqid = p->o_arg.seqid;
963 	p->c_res.seqid = p->c_arg.seqid;
964 	p->o_res.server = p->o_arg.server;
965 	p->o_res.access_request = p->o_arg.access;
966 	nfs_fattr_init(&p->f_attr);
967 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
968 }
969 
970 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
971 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
972 		const struct iattr *attrs,
973 		struct nfs4_label *label,
974 		enum open_claim_type4 claim,
975 		gfp_t gfp_mask)
976 {
977 	struct dentry *parent = dget_parent(dentry);
978 	struct inode *dir = parent->d_inode;
979 	struct nfs_server *server = NFS_SERVER(dir);
980 	struct nfs4_opendata *p;
981 
982 	p = kzalloc(sizeof(*p), gfp_mask);
983 	if (p == NULL)
984 		goto err;
985 
986 	p->f_label = nfs4_label_alloc(server, gfp_mask);
987 	if (IS_ERR(p->f_label))
988 		goto err_free_p;
989 
990 	p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
991 	if (p->o_arg.seqid == NULL)
992 		goto err_free_label;
993 	nfs_sb_active(dentry->d_sb);
994 	p->dentry = dget(dentry);
995 	p->dir = parent;
996 	p->owner = sp;
997 	atomic_inc(&sp->so_count);
998 	p->o_arg.open_flags = flags;
999 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1000 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1001 	 * will return permission denied for all bits until close */
1002 	if (!(flags & O_EXCL)) {
1003 		/* ask server to check for all possible rights as results
1004 		 * are cached */
1005 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1006 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1007 	}
1008 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1009 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1010 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1011 	p->o_arg.name = &dentry->d_name;
1012 	p->o_arg.server = server;
1013 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1014 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1015 	p->o_arg.label = label;
1016 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1017 	switch (p->o_arg.claim) {
1018 	case NFS4_OPEN_CLAIM_NULL:
1019 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1020 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1021 		p->o_arg.fh = NFS_FH(dir);
1022 		break;
1023 	case NFS4_OPEN_CLAIM_PREVIOUS:
1024 	case NFS4_OPEN_CLAIM_FH:
1025 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1026 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1027 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1028 	}
1029 	if (attrs != NULL && attrs->ia_valid != 0) {
1030 		__u32 verf[2];
1031 
1032 		p->o_arg.u.attrs = &p->attrs;
1033 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1034 
1035 		verf[0] = jiffies;
1036 		verf[1] = current->pid;
1037 		memcpy(p->o_arg.u.verifier.data, verf,
1038 				sizeof(p->o_arg.u.verifier.data));
1039 	}
1040 	p->c_arg.fh = &p->o_res.fh;
1041 	p->c_arg.stateid = &p->o_res.stateid;
1042 	p->c_arg.seqid = p->o_arg.seqid;
1043 	nfs4_init_opendata_res(p);
1044 	kref_init(&p->kref);
1045 	return p;
1046 
1047 err_free_label:
1048 	nfs4_label_free(p->f_label);
1049 err_free_p:
1050 	kfree(p);
1051 err:
1052 	dput(parent);
1053 	return NULL;
1054 }
1055 
1056 static void nfs4_opendata_free(struct kref *kref)
1057 {
1058 	struct nfs4_opendata *p = container_of(kref,
1059 			struct nfs4_opendata, kref);
1060 	struct super_block *sb = p->dentry->d_sb;
1061 
1062 	nfs_free_seqid(p->o_arg.seqid);
1063 	if (p->state != NULL)
1064 		nfs4_put_open_state(p->state);
1065 	nfs4_put_state_owner(p->owner);
1066 
1067 	nfs4_label_free(p->f_label);
1068 
1069 	dput(p->dir);
1070 	dput(p->dentry);
1071 	nfs_sb_deactive(sb);
1072 	nfs_fattr_free_names(&p->f_attr);
1073 	kfree(p);
1074 }
1075 
1076 static void nfs4_opendata_put(struct nfs4_opendata *p)
1077 {
1078 	if (p != NULL)
1079 		kref_put(&p->kref, nfs4_opendata_free);
1080 }
1081 
1082 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1083 {
1084 	int ret;
1085 
1086 	ret = rpc_wait_for_completion_task(task);
1087 	return ret;
1088 }
1089 
1090 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1091 {
1092 	int ret = 0;
1093 
1094 	if (open_mode & (O_EXCL|O_TRUNC))
1095 		goto out;
1096 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1097 		case FMODE_READ:
1098 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1099 				&& state->n_rdonly != 0;
1100 			break;
1101 		case FMODE_WRITE:
1102 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1103 				&& state->n_wronly != 0;
1104 			break;
1105 		case FMODE_READ|FMODE_WRITE:
1106 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1107 				&& state->n_rdwr != 0;
1108 	}
1109 out:
1110 	return ret;
1111 }
1112 
1113 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1114 {
1115 	if (delegation == NULL)
1116 		return 0;
1117 	if ((delegation->type & fmode) != fmode)
1118 		return 0;
1119 	if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1120 		return 0;
1121 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1122 		return 0;
1123 	nfs_mark_delegation_referenced(delegation);
1124 	return 1;
1125 }
1126 
1127 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1128 {
1129 	switch (fmode) {
1130 		case FMODE_WRITE:
1131 			state->n_wronly++;
1132 			break;
1133 		case FMODE_READ:
1134 			state->n_rdonly++;
1135 			break;
1136 		case FMODE_READ|FMODE_WRITE:
1137 			state->n_rdwr++;
1138 	}
1139 	nfs4_state_set_mode_locked(state, state->state | fmode);
1140 }
1141 
1142 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1143 {
1144 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1145 		nfs4_stateid_copy(&state->stateid, stateid);
1146 	nfs4_stateid_copy(&state->open_stateid, stateid);
1147 	set_bit(NFS_OPEN_STATE, &state->flags);
1148 	switch (fmode) {
1149 		case FMODE_READ:
1150 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1151 			break;
1152 		case FMODE_WRITE:
1153 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1154 			break;
1155 		case FMODE_READ|FMODE_WRITE:
1156 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1157 	}
1158 }
1159 
1160 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1161 {
1162 	write_seqlock(&state->seqlock);
1163 	nfs_set_open_stateid_locked(state, stateid, fmode);
1164 	write_sequnlock(&state->seqlock);
1165 }
1166 
1167 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1168 {
1169 	/*
1170 	 * Protect the call to nfs4_state_set_mode_locked and
1171 	 * serialise the stateid update
1172 	 */
1173 	write_seqlock(&state->seqlock);
1174 	if (deleg_stateid != NULL) {
1175 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1176 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1177 	}
1178 	if (open_stateid != NULL)
1179 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1180 	write_sequnlock(&state->seqlock);
1181 	spin_lock(&state->owner->so_lock);
1182 	update_open_stateflags(state, fmode);
1183 	spin_unlock(&state->owner->so_lock);
1184 }
1185 
1186 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1187 {
1188 	struct nfs_inode *nfsi = NFS_I(state->inode);
1189 	struct nfs_delegation *deleg_cur;
1190 	int ret = 0;
1191 
1192 	fmode &= (FMODE_READ|FMODE_WRITE);
1193 
1194 	rcu_read_lock();
1195 	deleg_cur = rcu_dereference(nfsi->delegation);
1196 	if (deleg_cur == NULL)
1197 		goto no_delegation;
1198 
1199 	spin_lock(&deleg_cur->lock);
1200 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1201 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1202 	    (deleg_cur->type & fmode) != fmode)
1203 		goto no_delegation_unlock;
1204 
1205 	if (delegation == NULL)
1206 		delegation = &deleg_cur->stateid;
1207 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1208 		goto no_delegation_unlock;
1209 
1210 	nfs_mark_delegation_referenced(deleg_cur);
1211 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1212 	ret = 1;
1213 no_delegation_unlock:
1214 	spin_unlock(&deleg_cur->lock);
1215 no_delegation:
1216 	rcu_read_unlock();
1217 
1218 	if (!ret && open_stateid != NULL) {
1219 		__update_open_stateid(state, open_stateid, NULL, fmode);
1220 		ret = 1;
1221 	}
1222 
1223 	return ret;
1224 }
1225 
1226 
1227 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1228 {
1229 	struct nfs_delegation *delegation;
1230 
1231 	rcu_read_lock();
1232 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1233 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1234 		rcu_read_unlock();
1235 		return;
1236 	}
1237 	rcu_read_unlock();
1238 	nfs4_inode_return_delegation(inode);
1239 }
1240 
1241 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1242 {
1243 	struct nfs4_state *state = opendata->state;
1244 	struct nfs_inode *nfsi = NFS_I(state->inode);
1245 	struct nfs_delegation *delegation;
1246 	int open_mode = opendata->o_arg.open_flags;
1247 	fmode_t fmode = opendata->o_arg.fmode;
1248 	nfs4_stateid stateid;
1249 	int ret = -EAGAIN;
1250 
1251 	for (;;) {
1252 		if (can_open_cached(state, fmode, open_mode)) {
1253 			spin_lock(&state->owner->so_lock);
1254 			if (can_open_cached(state, fmode, open_mode)) {
1255 				update_open_stateflags(state, fmode);
1256 				spin_unlock(&state->owner->so_lock);
1257 				goto out_return_state;
1258 			}
1259 			spin_unlock(&state->owner->so_lock);
1260 		}
1261 		rcu_read_lock();
1262 		delegation = rcu_dereference(nfsi->delegation);
1263 		if (!can_open_delegated(delegation, fmode)) {
1264 			rcu_read_unlock();
1265 			break;
1266 		}
1267 		/* Save the delegation */
1268 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1269 		rcu_read_unlock();
1270 		nfs_release_seqid(opendata->o_arg.seqid);
1271 		if (!opendata->is_recover) {
1272 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1273 			if (ret != 0)
1274 				goto out;
1275 		}
1276 		ret = -EAGAIN;
1277 
1278 		/* Try to update the stateid using the delegation */
1279 		if (update_open_stateid(state, NULL, &stateid, fmode))
1280 			goto out_return_state;
1281 	}
1282 out:
1283 	return ERR_PTR(ret);
1284 out_return_state:
1285 	atomic_inc(&state->count);
1286 	return state;
1287 }
1288 
1289 static void
1290 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1291 {
1292 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1293 	struct nfs_delegation *delegation;
1294 	int delegation_flags = 0;
1295 
1296 	rcu_read_lock();
1297 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1298 	if (delegation)
1299 		delegation_flags = delegation->flags;
1300 	rcu_read_unlock();
1301 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1302 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1303 				   "returning a delegation for "
1304 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1305 				   clp->cl_hostname);
1306 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1307 		nfs_inode_set_delegation(state->inode,
1308 					 data->owner->so_cred,
1309 					 &data->o_res);
1310 	else
1311 		nfs_inode_reclaim_delegation(state->inode,
1312 					     data->owner->so_cred,
1313 					     &data->o_res);
1314 }
1315 
1316 /*
1317  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1318  * and update the nfs4_state.
1319  */
1320 static struct nfs4_state *
1321 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1322 {
1323 	struct inode *inode = data->state->inode;
1324 	struct nfs4_state *state = data->state;
1325 	int ret;
1326 
1327 	if (!data->rpc_done) {
1328 		if (data->rpc_status) {
1329 			ret = data->rpc_status;
1330 			goto err;
1331 		}
1332 		/* cached opens have already been processed */
1333 		goto update;
1334 	}
1335 
1336 	ret = nfs_refresh_inode(inode, &data->f_attr);
1337 	if (ret)
1338 		goto err;
1339 
1340 	if (data->o_res.delegation_type != 0)
1341 		nfs4_opendata_check_deleg(data, state);
1342 update:
1343 	update_open_stateid(state, &data->o_res.stateid, NULL,
1344 			    data->o_arg.fmode);
1345 	atomic_inc(&state->count);
1346 
1347 	return state;
1348 err:
1349 	return ERR_PTR(ret);
1350 
1351 }
1352 
1353 static struct nfs4_state *
1354 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1355 {
1356 	struct inode *inode;
1357 	struct nfs4_state *state = NULL;
1358 	int ret;
1359 
1360 	if (!data->rpc_done) {
1361 		state = nfs4_try_open_cached(data);
1362 		goto out;
1363 	}
1364 
1365 	ret = -EAGAIN;
1366 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1367 		goto err;
1368 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1369 	ret = PTR_ERR(inode);
1370 	if (IS_ERR(inode))
1371 		goto err;
1372 	ret = -ENOMEM;
1373 	state = nfs4_get_open_state(inode, data->owner);
1374 	if (state == NULL)
1375 		goto err_put_inode;
1376 	if (data->o_res.delegation_type != 0)
1377 		nfs4_opendata_check_deleg(data, state);
1378 	update_open_stateid(state, &data->o_res.stateid, NULL,
1379 			data->o_arg.fmode);
1380 	iput(inode);
1381 out:
1382 	nfs_release_seqid(data->o_arg.seqid);
1383 	return state;
1384 err_put_inode:
1385 	iput(inode);
1386 err:
1387 	return ERR_PTR(ret);
1388 }
1389 
1390 static struct nfs4_state *
1391 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1392 {
1393 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1394 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1395 	return _nfs4_opendata_to_nfs4_state(data);
1396 }
1397 
1398 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1399 {
1400 	struct nfs_inode *nfsi = NFS_I(state->inode);
1401 	struct nfs_open_context *ctx;
1402 
1403 	spin_lock(&state->inode->i_lock);
1404 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1405 		if (ctx->state != state)
1406 			continue;
1407 		get_nfs_open_context(ctx);
1408 		spin_unlock(&state->inode->i_lock);
1409 		return ctx;
1410 	}
1411 	spin_unlock(&state->inode->i_lock);
1412 	return ERR_PTR(-ENOENT);
1413 }
1414 
1415 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1416 		struct nfs4_state *state, enum open_claim_type4 claim)
1417 {
1418 	struct nfs4_opendata *opendata;
1419 
1420 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1421 			NULL, NULL, claim, GFP_NOFS);
1422 	if (opendata == NULL)
1423 		return ERR_PTR(-ENOMEM);
1424 	opendata->state = state;
1425 	atomic_inc(&state->count);
1426 	return opendata;
1427 }
1428 
1429 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1430 {
1431 	struct nfs4_state *newstate;
1432 	int ret;
1433 
1434 	opendata->o_arg.open_flags = 0;
1435 	opendata->o_arg.fmode = fmode;
1436 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1437 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1438 	nfs4_init_opendata_res(opendata);
1439 	ret = _nfs4_recover_proc_open(opendata);
1440 	if (ret != 0)
1441 		return ret;
1442 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1443 	if (IS_ERR(newstate))
1444 		return PTR_ERR(newstate);
1445 	nfs4_close_state(newstate, fmode);
1446 	*res = newstate;
1447 	return 0;
1448 }
1449 
1450 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1451 {
1452 	struct nfs4_state *newstate;
1453 	int ret;
1454 
1455 	/* memory barrier prior to reading state->n_* */
1456 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1457 	clear_bit(NFS_OPEN_STATE, &state->flags);
1458 	smp_rmb();
1459 	if (state->n_rdwr != 0) {
1460 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
1461 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1462 		if (ret != 0)
1463 			return ret;
1464 		if (newstate != state)
1465 			return -ESTALE;
1466 	}
1467 	if (state->n_wronly != 0) {
1468 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1469 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1470 		if (ret != 0)
1471 			return ret;
1472 		if (newstate != state)
1473 			return -ESTALE;
1474 	}
1475 	if (state->n_rdonly != 0) {
1476 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1477 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1478 		if (ret != 0)
1479 			return ret;
1480 		if (newstate != state)
1481 			return -ESTALE;
1482 	}
1483 	/*
1484 	 * We may have performed cached opens for all three recoveries.
1485 	 * Check if we need to update the current stateid.
1486 	 */
1487 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1488 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1489 		write_seqlock(&state->seqlock);
1490 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1491 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1492 		write_sequnlock(&state->seqlock);
1493 	}
1494 	return 0;
1495 }
1496 
1497 /*
1498  * OPEN_RECLAIM:
1499  * 	reclaim state on the server after a reboot.
1500  */
1501 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1502 {
1503 	struct nfs_delegation *delegation;
1504 	struct nfs4_opendata *opendata;
1505 	fmode_t delegation_type = 0;
1506 	int status;
1507 
1508 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1509 			NFS4_OPEN_CLAIM_PREVIOUS);
1510 	if (IS_ERR(opendata))
1511 		return PTR_ERR(opendata);
1512 	rcu_read_lock();
1513 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1514 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1515 		delegation_type = delegation->type;
1516 	rcu_read_unlock();
1517 	opendata->o_arg.u.delegation_type = delegation_type;
1518 	status = nfs4_open_recover(opendata, state);
1519 	nfs4_opendata_put(opendata);
1520 	return status;
1521 }
1522 
1523 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1524 {
1525 	struct nfs_server *server = NFS_SERVER(state->inode);
1526 	struct nfs4_exception exception = { };
1527 	int err;
1528 	do {
1529 		err = _nfs4_do_open_reclaim(ctx, state);
1530 		trace_nfs4_open_reclaim(ctx, 0, err);
1531 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1532 			continue;
1533 		if (err != -NFS4ERR_DELAY)
1534 			break;
1535 		nfs4_handle_exception(server, err, &exception);
1536 	} while (exception.retry);
1537 	return err;
1538 }
1539 
1540 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1541 {
1542 	struct nfs_open_context *ctx;
1543 	int ret;
1544 
1545 	ctx = nfs4_state_find_open_context(state);
1546 	if (IS_ERR(ctx))
1547 		return -EAGAIN;
1548 	ret = nfs4_do_open_reclaim(ctx, state);
1549 	put_nfs_open_context(ctx);
1550 	return ret;
1551 }
1552 
1553 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1554 {
1555 	switch (err) {
1556 		default:
1557 			printk(KERN_ERR "NFS: %s: unhandled error "
1558 					"%d.\n", __func__, err);
1559 		case 0:
1560 		case -ENOENT:
1561 		case -ESTALE:
1562 			break;
1563 		case -NFS4ERR_BADSESSION:
1564 		case -NFS4ERR_BADSLOT:
1565 		case -NFS4ERR_BAD_HIGH_SLOT:
1566 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1567 		case -NFS4ERR_DEADSESSION:
1568 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1569 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1570 			return -EAGAIN;
1571 		case -NFS4ERR_STALE_CLIENTID:
1572 		case -NFS4ERR_STALE_STATEID:
1573 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1574 		case -NFS4ERR_EXPIRED:
1575 			/* Don't recall a delegation if it was lost */
1576 			nfs4_schedule_lease_recovery(server->nfs_client);
1577 			return -EAGAIN;
1578 		case -NFS4ERR_MOVED:
1579 			nfs4_schedule_migration_recovery(server);
1580 			return -EAGAIN;
1581 		case -NFS4ERR_LEASE_MOVED:
1582 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1583 			return -EAGAIN;
1584 		case -NFS4ERR_DELEG_REVOKED:
1585 		case -NFS4ERR_ADMIN_REVOKED:
1586 		case -NFS4ERR_BAD_STATEID:
1587 		case -NFS4ERR_OPENMODE:
1588 			nfs_inode_find_state_and_recover(state->inode,
1589 					stateid);
1590 			nfs4_schedule_stateid_recovery(server, state);
1591 			return 0;
1592 		case -NFS4ERR_DELAY:
1593 		case -NFS4ERR_GRACE:
1594 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1595 			ssleep(1);
1596 			return -EAGAIN;
1597 		case -ENOMEM:
1598 		case -NFS4ERR_DENIED:
1599 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1600 			return 0;
1601 	}
1602 	return err;
1603 }
1604 
1605 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1606 {
1607 	struct nfs_server *server = NFS_SERVER(state->inode);
1608 	struct nfs4_opendata *opendata;
1609 	int err;
1610 
1611 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1612 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1613 	if (IS_ERR(opendata))
1614 		return PTR_ERR(opendata);
1615 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1616 	err = nfs4_open_recover(opendata, state);
1617 	nfs4_opendata_put(opendata);
1618 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1619 }
1620 
1621 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1622 {
1623 	struct nfs4_opendata *data = calldata;
1624 
1625 	nfs40_setup_sequence(data->o_arg.server, &data->o_arg.seq_args,
1626 				&data->o_res.seq_res, task);
1627 }
1628 
1629 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1630 {
1631 	struct nfs4_opendata *data = calldata;
1632 
1633 	nfs40_sequence_done(task, &data->o_res.seq_res);
1634 
1635 	data->rpc_status = task->tk_status;
1636 	if (data->rpc_status == 0) {
1637 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1638 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1639 		renew_lease(data->o_res.server, data->timestamp);
1640 		data->rpc_done = 1;
1641 	}
1642 }
1643 
1644 static void nfs4_open_confirm_release(void *calldata)
1645 {
1646 	struct nfs4_opendata *data = calldata;
1647 	struct nfs4_state *state = NULL;
1648 
1649 	/* If this request hasn't been cancelled, do nothing */
1650 	if (data->cancelled == 0)
1651 		goto out_free;
1652 	/* In case of error, no cleanup! */
1653 	if (!data->rpc_done)
1654 		goto out_free;
1655 	state = nfs4_opendata_to_nfs4_state(data);
1656 	if (!IS_ERR(state))
1657 		nfs4_close_state(state, data->o_arg.fmode);
1658 out_free:
1659 	nfs4_opendata_put(data);
1660 }
1661 
1662 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1663 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1664 	.rpc_call_done = nfs4_open_confirm_done,
1665 	.rpc_release = nfs4_open_confirm_release,
1666 };
1667 
1668 /*
1669  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1670  */
1671 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1672 {
1673 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1674 	struct rpc_task *task;
1675 	struct  rpc_message msg = {
1676 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1677 		.rpc_argp = &data->c_arg,
1678 		.rpc_resp = &data->c_res,
1679 		.rpc_cred = data->owner->so_cred,
1680 	};
1681 	struct rpc_task_setup task_setup_data = {
1682 		.rpc_client = server->client,
1683 		.rpc_message = &msg,
1684 		.callback_ops = &nfs4_open_confirm_ops,
1685 		.callback_data = data,
1686 		.workqueue = nfsiod_workqueue,
1687 		.flags = RPC_TASK_ASYNC,
1688 	};
1689 	int status;
1690 
1691 	nfs4_init_sequence(&data->o_arg.seq_args, &data->o_res.seq_res, 1);
1692 	kref_get(&data->kref);
1693 	data->rpc_done = 0;
1694 	data->rpc_status = 0;
1695 	data->timestamp = jiffies;
1696 	task = rpc_run_task(&task_setup_data);
1697 	if (IS_ERR(task))
1698 		return PTR_ERR(task);
1699 	status = nfs4_wait_for_completion_rpc_task(task);
1700 	if (status != 0) {
1701 		data->cancelled = 1;
1702 		smp_wmb();
1703 	} else
1704 		status = data->rpc_status;
1705 	rpc_put_task(task);
1706 	return status;
1707 }
1708 
1709 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1710 {
1711 	struct nfs4_opendata *data = calldata;
1712 	struct nfs4_state_owner *sp = data->owner;
1713 	struct nfs_client *clp = sp->so_server->nfs_client;
1714 
1715 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1716 		goto out_wait;
1717 	/*
1718 	 * Check if we still need to send an OPEN call, or if we can use
1719 	 * a delegation instead.
1720 	 */
1721 	if (data->state != NULL) {
1722 		struct nfs_delegation *delegation;
1723 
1724 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1725 			goto out_no_action;
1726 		rcu_read_lock();
1727 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1728 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1729 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1730 		    can_open_delegated(delegation, data->o_arg.fmode))
1731 			goto unlock_no_action;
1732 		rcu_read_unlock();
1733 	}
1734 	/* Update client id. */
1735 	data->o_arg.clientid = clp->cl_clientid;
1736 	switch (data->o_arg.claim) {
1737 	case NFS4_OPEN_CLAIM_PREVIOUS:
1738 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1739 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1740 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1741 	case NFS4_OPEN_CLAIM_FH:
1742 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1743 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1744 	}
1745 	data->timestamp = jiffies;
1746 	if (nfs4_setup_sequence(data->o_arg.server,
1747 				&data->o_arg.seq_args,
1748 				&data->o_res.seq_res,
1749 				task) != 0)
1750 		nfs_release_seqid(data->o_arg.seqid);
1751 
1752 	/* Set the create mode (note dependency on the session type) */
1753 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1754 	if (data->o_arg.open_flags & O_EXCL) {
1755 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1756 		if (nfs4_has_persistent_session(clp))
1757 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1758 		else if (clp->cl_mvops->minor_version > 0)
1759 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1760 	}
1761 	return;
1762 unlock_no_action:
1763 	rcu_read_unlock();
1764 out_no_action:
1765 	task->tk_action = NULL;
1766 out_wait:
1767 	nfs4_sequence_done(task, &data->o_res.seq_res);
1768 }
1769 
1770 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1771 {
1772 	struct nfs4_opendata *data = calldata;
1773 
1774 	data->rpc_status = task->tk_status;
1775 
1776 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1777 		return;
1778 
1779 	if (task->tk_status == 0) {
1780 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1781 			switch (data->o_res.f_attr->mode & S_IFMT) {
1782 			case S_IFREG:
1783 				break;
1784 			case S_IFLNK:
1785 				data->rpc_status = -ELOOP;
1786 				break;
1787 			case S_IFDIR:
1788 				data->rpc_status = -EISDIR;
1789 				break;
1790 			default:
1791 				data->rpc_status = -ENOTDIR;
1792 			}
1793 		}
1794 		renew_lease(data->o_res.server, data->timestamp);
1795 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1796 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1797 	}
1798 	data->rpc_done = 1;
1799 }
1800 
1801 static void nfs4_open_release(void *calldata)
1802 {
1803 	struct nfs4_opendata *data = calldata;
1804 	struct nfs4_state *state = NULL;
1805 
1806 	/* If this request hasn't been cancelled, do nothing */
1807 	if (data->cancelled == 0)
1808 		goto out_free;
1809 	/* In case of error, no cleanup! */
1810 	if (data->rpc_status != 0 || !data->rpc_done)
1811 		goto out_free;
1812 	/* In case we need an open_confirm, no cleanup! */
1813 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1814 		goto out_free;
1815 	state = nfs4_opendata_to_nfs4_state(data);
1816 	if (!IS_ERR(state))
1817 		nfs4_close_state(state, data->o_arg.fmode);
1818 out_free:
1819 	nfs4_opendata_put(data);
1820 }
1821 
1822 static const struct rpc_call_ops nfs4_open_ops = {
1823 	.rpc_call_prepare = nfs4_open_prepare,
1824 	.rpc_call_done = nfs4_open_done,
1825 	.rpc_release = nfs4_open_release,
1826 };
1827 
1828 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1829 {
1830 	struct inode *dir = data->dir->d_inode;
1831 	struct nfs_server *server = NFS_SERVER(dir);
1832 	struct nfs_openargs *o_arg = &data->o_arg;
1833 	struct nfs_openres *o_res = &data->o_res;
1834 	struct rpc_task *task;
1835 	struct rpc_message msg = {
1836 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1837 		.rpc_argp = o_arg,
1838 		.rpc_resp = o_res,
1839 		.rpc_cred = data->owner->so_cred,
1840 	};
1841 	struct rpc_task_setup task_setup_data = {
1842 		.rpc_client = server->client,
1843 		.rpc_message = &msg,
1844 		.callback_ops = &nfs4_open_ops,
1845 		.callback_data = data,
1846 		.workqueue = nfsiod_workqueue,
1847 		.flags = RPC_TASK_ASYNC,
1848 	};
1849 	int status;
1850 
1851 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1852 	kref_get(&data->kref);
1853 	data->rpc_done = 0;
1854 	data->rpc_status = 0;
1855 	data->cancelled = 0;
1856 	data->is_recover = 0;
1857 	if (isrecover) {
1858 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1859 		data->is_recover = 1;
1860 	}
1861 	task = rpc_run_task(&task_setup_data);
1862         if (IS_ERR(task))
1863                 return PTR_ERR(task);
1864         status = nfs4_wait_for_completion_rpc_task(task);
1865         if (status != 0) {
1866                 data->cancelled = 1;
1867                 smp_wmb();
1868         } else
1869                 status = data->rpc_status;
1870         rpc_put_task(task);
1871 
1872 	return status;
1873 }
1874 
1875 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1876 {
1877 	struct inode *dir = data->dir->d_inode;
1878 	struct nfs_openres *o_res = &data->o_res;
1879         int status;
1880 
1881 	status = nfs4_run_open_task(data, 1);
1882 	if (status != 0 || !data->rpc_done)
1883 		return status;
1884 
1885 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1886 
1887 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1888 		status = _nfs4_proc_open_confirm(data);
1889 		if (status != 0)
1890 			return status;
1891 	}
1892 
1893 	return status;
1894 }
1895 
1896 static int nfs4_opendata_access(struct rpc_cred *cred,
1897 				struct nfs4_opendata *opendata,
1898 				struct nfs4_state *state, fmode_t fmode,
1899 				int openflags)
1900 {
1901 	struct nfs_access_entry cache;
1902 	u32 mask;
1903 
1904 	/* access call failed or for some reason the server doesn't
1905 	 * support any access modes -- defer access call until later */
1906 	if (opendata->o_res.access_supported == 0)
1907 		return 0;
1908 
1909 	mask = 0;
1910 	/* don't check MAY_WRITE - a newly created file may not have
1911 	 * write mode bits, but POSIX allows the creating process to write.
1912 	 * use openflags to check for exec, because fmode won't
1913 	 * always have FMODE_EXEC set when file open for exec. */
1914 	if (openflags & __FMODE_EXEC) {
1915 		/* ONLY check for exec rights */
1916 		mask = MAY_EXEC;
1917 	} else if (fmode & FMODE_READ)
1918 		mask = MAY_READ;
1919 
1920 	cache.cred = cred;
1921 	cache.jiffies = jiffies;
1922 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
1923 	nfs_access_add_cache(state->inode, &cache);
1924 
1925 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1926 		return 0;
1927 
1928 	/* even though OPEN succeeded, access is denied. Close the file */
1929 	nfs4_close_state(state, fmode);
1930 	return -EACCES;
1931 }
1932 
1933 /*
1934  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1935  */
1936 static int _nfs4_proc_open(struct nfs4_opendata *data)
1937 {
1938 	struct inode *dir = data->dir->d_inode;
1939 	struct nfs_server *server = NFS_SERVER(dir);
1940 	struct nfs_openargs *o_arg = &data->o_arg;
1941 	struct nfs_openres *o_res = &data->o_res;
1942 	int status;
1943 
1944 	status = nfs4_run_open_task(data, 0);
1945 	if (!data->rpc_done)
1946 		return status;
1947 	if (status != 0) {
1948 		if (status == -NFS4ERR_BADNAME &&
1949 				!(o_arg->open_flags & O_CREAT))
1950 			return -ENOENT;
1951 		return status;
1952 	}
1953 
1954 	nfs_fattr_map_and_free_names(server, &data->f_attr);
1955 
1956 	if (o_arg->open_flags & O_CREAT) {
1957 		update_changeattr(dir, &o_res->cinfo);
1958 		if (o_arg->open_flags & O_EXCL)
1959 			data->file_created = 1;
1960 		else if (o_res->cinfo.before != o_res->cinfo.after)
1961 			data->file_created = 1;
1962 	}
1963 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1964 		server->caps &= ~NFS_CAP_POSIX_LOCK;
1965 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1966 		status = _nfs4_proc_open_confirm(data);
1967 		if (status != 0)
1968 			return status;
1969 	}
1970 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1971 		_nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
1972 	return 0;
1973 }
1974 
1975 static int nfs4_recover_expired_lease(struct nfs_server *server)
1976 {
1977 	return nfs4_client_recover_expired_lease(server->nfs_client);
1978 }
1979 
1980 /*
1981  * OPEN_EXPIRED:
1982  * 	reclaim state on the server after a network partition.
1983  * 	Assumes caller holds the appropriate lock
1984  */
1985 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1986 {
1987 	struct nfs4_opendata *opendata;
1988 	int ret;
1989 
1990 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1991 			NFS4_OPEN_CLAIM_FH);
1992 	if (IS_ERR(opendata))
1993 		return PTR_ERR(opendata);
1994 	ret = nfs4_open_recover(opendata, state);
1995 	if (ret == -ESTALE)
1996 		d_drop(ctx->dentry);
1997 	nfs4_opendata_put(opendata);
1998 	return ret;
1999 }
2000 
2001 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2002 {
2003 	struct nfs_server *server = NFS_SERVER(state->inode);
2004 	struct nfs4_exception exception = { };
2005 	int err;
2006 
2007 	do {
2008 		err = _nfs4_open_expired(ctx, state);
2009 		trace_nfs4_open_expired(ctx, 0, err);
2010 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2011 			continue;
2012 		switch (err) {
2013 		default:
2014 			goto out;
2015 		case -NFS4ERR_GRACE:
2016 		case -NFS4ERR_DELAY:
2017 			nfs4_handle_exception(server, err, &exception);
2018 			err = 0;
2019 		}
2020 	} while (exception.retry);
2021 out:
2022 	return err;
2023 }
2024 
2025 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2026 {
2027 	struct nfs_open_context *ctx;
2028 	int ret;
2029 
2030 	ctx = nfs4_state_find_open_context(state);
2031 	if (IS_ERR(ctx))
2032 		return -EAGAIN;
2033 	ret = nfs4_do_open_expired(ctx, state);
2034 	put_nfs_open_context(ctx);
2035 	return ret;
2036 }
2037 
2038 #if defined(CONFIG_NFS_V4_1)
2039 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
2040 {
2041 	struct nfs_server *server = NFS_SERVER(state->inode);
2042 	nfs4_stateid *stateid = &state->stateid;
2043 	struct nfs_delegation *delegation;
2044 	struct rpc_cred *cred = NULL;
2045 	int status = -NFS4ERR_BAD_STATEID;
2046 
2047 	/* If a state reset has been done, test_stateid is unneeded */
2048 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2049 		return;
2050 
2051 	/* Get the delegation credential for use by test/free_stateid */
2052 	rcu_read_lock();
2053 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2054 	if (delegation != NULL &&
2055 	    nfs4_stateid_match(&delegation->stateid, stateid)) {
2056 		cred = get_rpccred(delegation->cred);
2057 		rcu_read_unlock();
2058 		status = nfs41_test_stateid(server, stateid, cred);
2059 		trace_nfs4_test_delegation_stateid(state, NULL, status);
2060 	} else
2061 		rcu_read_unlock();
2062 
2063 	if (status != NFS_OK) {
2064 		/* Free the stateid unless the server explicitly
2065 		 * informs us the stateid is unrecognized. */
2066 		if (status != -NFS4ERR_BAD_STATEID)
2067 			nfs41_free_stateid(server, stateid, cred);
2068 		nfs_remove_bad_delegation(state->inode);
2069 
2070 		write_seqlock(&state->seqlock);
2071 		nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2072 		write_sequnlock(&state->seqlock);
2073 		clear_bit(NFS_DELEGATED_STATE, &state->flags);
2074 	}
2075 
2076 	if (cred != NULL)
2077 		put_rpccred(cred);
2078 }
2079 
2080 /**
2081  * nfs41_check_open_stateid - possibly free an open stateid
2082  *
2083  * @state: NFSv4 state for an inode
2084  *
2085  * Returns NFS_OK if recovery for this stateid is now finished.
2086  * Otherwise a negative NFS4ERR value is returned.
2087  */
2088 static int nfs41_check_open_stateid(struct nfs4_state *state)
2089 {
2090 	struct nfs_server *server = NFS_SERVER(state->inode);
2091 	nfs4_stateid *stateid = &state->open_stateid;
2092 	struct rpc_cred *cred = state->owner->so_cred;
2093 	int status;
2094 
2095 	/* If a state reset has been done, test_stateid is unneeded */
2096 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2097 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2098 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2099 		return -NFS4ERR_BAD_STATEID;
2100 
2101 	status = nfs41_test_stateid(server, stateid, cred);
2102 	trace_nfs4_test_open_stateid(state, NULL, status);
2103 	if (status != NFS_OK) {
2104 		/* Free the stateid unless the server explicitly
2105 		 * informs us the stateid is unrecognized. */
2106 		if (status != -NFS4ERR_BAD_STATEID)
2107 			nfs41_free_stateid(server, stateid, cred);
2108 
2109 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2110 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2111 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2112 		clear_bit(NFS_OPEN_STATE, &state->flags);
2113 	}
2114 	return status;
2115 }
2116 
2117 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2118 {
2119 	int status;
2120 
2121 	nfs41_clear_delegation_stateid(state);
2122 	status = nfs41_check_open_stateid(state);
2123 	if (status != NFS_OK)
2124 		status = nfs4_open_expired(sp, state);
2125 	return status;
2126 }
2127 #endif
2128 
2129 /*
2130  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2131  * fields corresponding to attributes that were used to store the verifier.
2132  * Make sure we clobber those fields in the later setattr call
2133  */
2134 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2135 {
2136 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2137 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2138 		sattr->ia_valid |= ATTR_ATIME;
2139 
2140 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2141 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2142 		sattr->ia_valid |= ATTR_MTIME;
2143 }
2144 
2145 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2146 		fmode_t fmode,
2147 		int flags,
2148 		struct nfs_open_context *ctx)
2149 {
2150 	struct nfs4_state_owner *sp = opendata->owner;
2151 	struct nfs_server *server = sp->so_server;
2152 	struct dentry *dentry;
2153 	struct nfs4_state *state;
2154 	unsigned int seq;
2155 	int ret;
2156 
2157 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2158 
2159 	ret = _nfs4_proc_open(opendata);
2160 	if (ret != 0)
2161 		goto out;
2162 
2163 	state = nfs4_opendata_to_nfs4_state(opendata);
2164 	ret = PTR_ERR(state);
2165 	if (IS_ERR(state))
2166 		goto out;
2167 	if (server->caps & NFS_CAP_POSIX_LOCK)
2168 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2169 
2170 	dentry = opendata->dentry;
2171 	if (dentry->d_inode == NULL) {
2172 		/* FIXME: Is this d_drop() ever needed? */
2173 		d_drop(dentry);
2174 		dentry = d_add_unique(dentry, igrab(state->inode));
2175 		if (dentry == NULL) {
2176 			dentry = opendata->dentry;
2177 		} else if (dentry != ctx->dentry) {
2178 			dput(ctx->dentry);
2179 			ctx->dentry = dget(dentry);
2180 		}
2181 		nfs_set_verifier(dentry,
2182 				nfs_save_change_attribute(opendata->dir->d_inode));
2183 	}
2184 
2185 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2186 	if (ret != 0)
2187 		goto out;
2188 
2189 	ctx->state = state;
2190 	if (dentry->d_inode == state->inode) {
2191 		nfs_inode_attach_open_context(ctx);
2192 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2193 			nfs4_schedule_stateid_recovery(server, state);
2194 	}
2195 out:
2196 	return ret;
2197 }
2198 
2199 /*
2200  * Returns a referenced nfs4_state
2201  */
2202 static int _nfs4_do_open(struct inode *dir,
2203 			struct nfs_open_context *ctx,
2204 			int flags,
2205 			struct iattr *sattr,
2206 			struct nfs4_label *label,
2207 			int *opened)
2208 {
2209 	struct nfs4_state_owner  *sp;
2210 	struct nfs4_state     *state = NULL;
2211 	struct nfs_server       *server = NFS_SERVER(dir);
2212 	struct nfs4_opendata *opendata;
2213 	struct dentry *dentry = ctx->dentry;
2214 	struct rpc_cred *cred = ctx->cred;
2215 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2216 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2217 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2218 	struct nfs4_label *olabel = NULL;
2219 	int status;
2220 
2221 	/* Protect against reboot recovery conflicts */
2222 	status = -ENOMEM;
2223 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2224 	if (sp == NULL) {
2225 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2226 		goto out_err;
2227 	}
2228 	status = nfs4_recover_expired_lease(server);
2229 	if (status != 0)
2230 		goto err_put_state_owner;
2231 	if (dentry->d_inode != NULL)
2232 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2233 	status = -ENOMEM;
2234 	if (dentry->d_inode)
2235 		claim = NFS4_OPEN_CLAIM_FH;
2236 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2237 			label, claim, GFP_KERNEL);
2238 	if (opendata == NULL)
2239 		goto err_put_state_owner;
2240 
2241 	if (label) {
2242 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2243 		if (IS_ERR(olabel)) {
2244 			status = PTR_ERR(olabel);
2245 			goto err_opendata_put;
2246 		}
2247 	}
2248 
2249 	if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2250 		opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2251 		if (!opendata->f_attr.mdsthreshold)
2252 			goto err_free_label;
2253 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2254 	}
2255 	if (dentry->d_inode != NULL)
2256 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2257 
2258 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2259 	if (status != 0)
2260 		goto err_free_label;
2261 	state = ctx->state;
2262 
2263 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2264 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2265 		nfs4_exclusive_attrset(opendata, sattr);
2266 
2267 		nfs_fattr_init(opendata->o_res.f_attr);
2268 		status = nfs4_do_setattr(state->inode, cred,
2269 				opendata->o_res.f_attr, sattr,
2270 				state, label, olabel);
2271 		if (status == 0) {
2272 			nfs_setattr_update_inode(state->inode, sattr);
2273 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2274 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2275 		}
2276 	}
2277 	if (opendata->file_created)
2278 		*opened |= FILE_CREATED;
2279 
2280 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2281 		*ctx_th = opendata->f_attr.mdsthreshold;
2282 	else
2283 		kfree(opendata->f_attr.mdsthreshold);
2284 	opendata->f_attr.mdsthreshold = NULL;
2285 
2286 	nfs4_label_free(olabel);
2287 
2288 	nfs4_opendata_put(opendata);
2289 	nfs4_put_state_owner(sp);
2290 	return 0;
2291 err_free_label:
2292 	nfs4_label_free(olabel);
2293 err_opendata_put:
2294 	kfree(opendata->f_attr.mdsthreshold);
2295 	nfs4_opendata_put(opendata);
2296 err_put_state_owner:
2297 	nfs4_put_state_owner(sp);
2298 out_err:
2299 	return status;
2300 }
2301 
2302 
2303 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2304 					struct nfs_open_context *ctx,
2305 					int flags,
2306 					struct iattr *sattr,
2307 					struct nfs4_label *label,
2308 					int *opened)
2309 {
2310 	struct nfs_server *server = NFS_SERVER(dir);
2311 	struct nfs4_exception exception = { };
2312 	struct nfs4_state *res;
2313 	int status;
2314 
2315 	do {
2316 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2317 		res = ctx->state;
2318 		trace_nfs4_open_file(ctx, flags, status);
2319 		if (status == 0)
2320 			break;
2321 		/* NOTE: BAD_SEQID means the server and client disagree about the
2322 		 * book-keeping w.r.t. state-changing operations
2323 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2324 		 * It is actually a sign of a bug on the client or on the server.
2325 		 *
2326 		 * If we receive a BAD_SEQID error in the particular case of
2327 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2328 		 * have unhashed the old state_owner for us, and that we can
2329 		 * therefore safely retry using a new one. We should still warn
2330 		 * the user though...
2331 		 */
2332 		if (status == -NFS4ERR_BAD_SEQID) {
2333 			pr_warn_ratelimited("NFS: v4 server %s "
2334 					" returned a bad sequence-id error!\n",
2335 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2336 			exception.retry = 1;
2337 			continue;
2338 		}
2339 		/*
2340 		 * BAD_STATEID on OPEN means that the server cancelled our
2341 		 * state before it received the OPEN_CONFIRM.
2342 		 * Recover by retrying the request as per the discussion
2343 		 * on Page 181 of RFC3530.
2344 		 */
2345 		if (status == -NFS4ERR_BAD_STATEID) {
2346 			exception.retry = 1;
2347 			continue;
2348 		}
2349 		if (status == -EAGAIN) {
2350 			/* We must have found a delegation */
2351 			exception.retry = 1;
2352 			continue;
2353 		}
2354 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2355 			continue;
2356 		res = ERR_PTR(nfs4_handle_exception(server,
2357 					status, &exception));
2358 	} while (exception.retry);
2359 	return res;
2360 }
2361 
2362 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2363 			    struct nfs_fattr *fattr, struct iattr *sattr,
2364 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2365 			    struct nfs4_label *olabel)
2366 {
2367 	struct nfs_server *server = NFS_SERVER(inode);
2368         struct nfs_setattrargs  arg = {
2369                 .fh             = NFS_FH(inode),
2370                 .iap            = sattr,
2371 		.server		= server,
2372 		.bitmask = server->attr_bitmask,
2373 		.label		= ilabel,
2374         };
2375         struct nfs_setattrres  res = {
2376 		.fattr		= fattr,
2377 		.label		= olabel,
2378 		.server		= server,
2379         };
2380         struct rpc_message msg = {
2381 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2382 		.rpc_argp	= &arg,
2383 		.rpc_resp	= &res,
2384 		.rpc_cred	= cred,
2385         };
2386 	unsigned long timestamp = jiffies;
2387 	fmode_t fmode;
2388 	bool truncate;
2389 	int status;
2390 
2391 	arg.bitmask = nfs4_bitmask(server, ilabel);
2392 	if (ilabel)
2393 		arg.bitmask = nfs4_bitmask(server, olabel);
2394 
2395 	nfs_fattr_init(fattr);
2396 
2397 	/* Servers should only apply open mode checks for file size changes */
2398 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2399 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2400 
2401 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2402 		/* Use that stateid */
2403 	} else if (truncate && state != NULL && nfs4_valid_open_stateid(state)) {
2404 		struct nfs_lockowner lockowner = {
2405 			.l_owner = current->files,
2406 			.l_pid = current->tgid,
2407 		};
2408 		nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2409 				&lockowner);
2410 	} else
2411 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2412 
2413 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2414 	if (status == 0 && state != NULL)
2415 		renew_lease(server, timestamp);
2416 	return status;
2417 }
2418 
2419 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2420 			   struct nfs_fattr *fattr, struct iattr *sattr,
2421 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2422 			   struct nfs4_label *olabel)
2423 {
2424 	struct nfs_server *server = NFS_SERVER(inode);
2425 	struct nfs4_exception exception = {
2426 		.state = state,
2427 		.inode = inode,
2428 	};
2429 	int err;
2430 	do {
2431 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2432 		trace_nfs4_setattr(inode, err);
2433 		switch (err) {
2434 		case -NFS4ERR_OPENMODE:
2435 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2436 				pr_warn_once("NFSv4: server %s is incorrectly "
2437 						"applying open mode checks to "
2438 						"a SETATTR that is not "
2439 						"changing file size.\n",
2440 						server->nfs_client->cl_hostname);
2441 			}
2442 			if (state && !(state->state & FMODE_WRITE)) {
2443 				err = -EBADF;
2444 				if (sattr->ia_valid & ATTR_OPEN)
2445 					err = -EACCES;
2446 				goto out;
2447 			}
2448 		}
2449 		err = nfs4_handle_exception(server, err, &exception);
2450 	} while (exception.retry);
2451 out:
2452 	return err;
2453 }
2454 
2455 struct nfs4_closedata {
2456 	struct inode *inode;
2457 	struct nfs4_state *state;
2458 	struct nfs_closeargs arg;
2459 	struct nfs_closeres res;
2460 	struct nfs_fattr fattr;
2461 	unsigned long timestamp;
2462 	bool roc;
2463 	u32 roc_barrier;
2464 };
2465 
2466 static void nfs4_free_closedata(void *data)
2467 {
2468 	struct nfs4_closedata *calldata = data;
2469 	struct nfs4_state_owner *sp = calldata->state->owner;
2470 	struct super_block *sb = calldata->state->inode->i_sb;
2471 
2472 	if (calldata->roc)
2473 		pnfs_roc_release(calldata->state->inode);
2474 	nfs4_put_open_state(calldata->state);
2475 	nfs_free_seqid(calldata->arg.seqid);
2476 	nfs4_put_state_owner(sp);
2477 	nfs_sb_deactive(sb);
2478 	kfree(calldata);
2479 }
2480 
2481 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2482 		fmode_t fmode)
2483 {
2484 	spin_lock(&state->owner->so_lock);
2485 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
2486 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
2487 	case FMODE_WRITE:
2488 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2489 		break;
2490 	case FMODE_READ:
2491 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2492 		break;
2493 	case 0:
2494 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2495 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2496 		clear_bit(NFS_OPEN_STATE, &state->flags);
2497 	}
2498 	spin_unlock(&state->owner->so_lock);
2499 }
2500 
2501 static void nfs4_close_done(struct rpc_task *task, void *data)
2502 {
2503 	struct nfs4_closedata *calldata = data;
2504 	struct nfs4_state *state = calldata->state;
2505 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2506 
2507 	dprintk("%s: begin!\n", __func__);
2508 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2509 		return;
2510 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2511         /* hmm. we are done with the inode, and in the process of freeing
2512 	 * the state_owner. we keep this around to process errors
2513 	 */
2514 	switch (task->tk_status) {
2515 		case 0:
2516 			if (calldata->roc)
2517 				pnfs_roc_set_barrier(state->inode,
2518 						     calldata->roc_barrier);
2519 			nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2520 			renew_lease(server, calldata->timestamp);
2521 			break;
2522 		case -NFS4ERR_ADMIN_REVOKED:
2523 		case -NFS4ERR_STALE_STATEID:
2524 		case -NFS4ERR_OLD_STATEID:
2525 		case -NFS4ERR_BAD_STATEID:
2526 		case -NFS4ERR_EXPIRED:
2527 			if (calldata->arg.fmode == 0)
2528 				break;
2529 		default:
2530 			if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
2531 				rpc_restart_call_prepare(task);
2532 				goto out_release;
2533 			}
2534 	}
2535 	nfs4_close_clear_stateid_flags(state, calldata->arg.fmode);
2536 out_release:
2537 	nfs_release_seqid(calldata->arg.seqid);
2538 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2539 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2540 }
2541 
2542 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2543 {
2544 	struct nfs4_closedata *calldata = data;
2545 	struct nfs4_state *state = calldata->state;
2546 	struct inode *inode = calldata->inode;
2547 	int call_close = 0;
2548 
2549 	dprintk("%s: begin!\n", __func__);
2550 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2551 		goto out_wait;
2552 
2553 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2554 	calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2555 	spin_lock(&state->owner->so_lock);
2556 	/* Calculate the change in open mode */
2557 	if (state->n_rdwr == 0) {
2558 		if (state->n_rdonly == 0) {
2559 			call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2560 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2561 			calldata->arg.fmode &= ~FMODE_READ;
2562 		}
2563 		if (state->n_wronly == 0) {
2564 			call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2565 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2566 			calldata->arg.fmode &= ~FMODE_WRITE;
2567 		}
2568 	}
2569 	if (!nfs4_valid_open_stateid(state))
2570 		call_close = 0;
2571 	spin_unlock(&state->owner->so_lock);
2572 
2573 	if (!call_close) {
2574 		/* Note: exit _without_ calling nfs4_close_done */
2575 		goto out_no_action;
2576 	}
2577 
2578 	if (calldata->arg.fmode == 0) {
2579 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2580 		if (calldata->roc &&
2581 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2582 			nfs_release_seqid(calldata->arg.seqid);
2583 			goto out_wait;
2584 		    }
2585 	}
2586 
2587 	nfs_fattr_init(calldata->res.fattr);
2588 	calldata->timestamp = jiffies;
2589 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2590 				&calldata->arg.seq_args,
2591 				&calldata->res.seq_res,
2592 				task) != 0)
2593 		nfs_release_seqid(calldata->arg.seqid);
2594 	dprintk("%s: done!\n", __func__);
2595 	return;
2596 out_no_action:
2597 	task->tk_action = NULL;
2598 out_wait:
2599 	nfs4_sequence_done(task, &calldata->res.seq_res);
2600 }
2601 
2602 static const struct rpc_call_ops nfs4_close_ops = {
2603 	.rpc_call_prepare = nfs4_close_prepare,
2604 	.rpc_call_done = nfs4_close_done,
2605 	.rpc_release = nfs4_free_closedata,
2606 };
2607 
2608 /*
2609  * It is possible for data to be read/written from a mem-mapped file
2610  * after the sys_close call (which hits the vfs layer as a flush).
2611  * This means that we can't safely call nfsv4 close on a file until
2612  * the inode is cleared. This in turn means that we are not good
2613  * NFSv4 citizens - we do not indicate to the server to update the file's
2614  * share state even when we are done with one of the three share
2615  * stateid's in the inode.
2616  *
2617  * NOTE: Caller must be holding the sp->so_owner semaphore!
2618  */
2619 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2620 {
2621 	struct nfs_server *server = NFS_SERVER(state->inode);
2622 	struct nfs4_closedata *calldata;
2623 	struct nfs4_state_owner *sp = state->owner;
2624 	struct rpc_task *task;
2625 	struct rpc_message msg = {
2626 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2627 		.rpc_cred = state->owner->so_cred,
2628 	};
2629 	struct rpc_task_setup task_setup_data = {
2630 		.rpc_client = server->client,
2631 		.rpc_message = &msg,
2632 		.callback_ops = &nfs4_close_ops,
2633 		.workqueue = nfsiod_workqueue,
2634 		.flags = RPC_TASK_ASYNC,
2635 	};
2636 	int status = -ENOMEM;
2637 
2638 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2639 		&task_setup_data.rpc_client, &msg);
2640 
2641 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2642 	if (calldata == NULL)
2643 		goto out;
2644 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2645 	calldata->inode = state->inode;
2646 	calldata->state = state;
2647 	calldata->arg.fh = NFS_FH(state->inode);
2648 	calldata->arg.stateid = &state->open_stateid;
2649 	/* Serialization for the sequence id */
2650 	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2651 	if (calldata->arg.seqid == NULL)
2652 		goto out_free_calldata;
2653 	calldata->arg.fmode = 0;
2654 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2655 	calldata->res.fattr = &calldata->fattr;
2656 	calldata->res.seqid = calldata->arg.seqid;
2657 	calldata->res.server = server;
2658 	calldata->roc = pnfs_roc(state->inode);
2659 	nfs_sb_active(calldata->inode->i_sb);
2660 
2661 	msg.rpc_argp = &calldata->arg;
2662 	msg.rpc_resp = &calldata->res;
2663 	task_setup_data.callback_data = calldata;
2664 	task = rpc_run_task(&task_setup_data);
2665 	if (IS_ERR(task))
2666 		return PTR_ERR(task);
2667 	status = 0;
2668 	if (wait)
2669 		status = rpc_wait_for_completion_task(task);
2670 	rpc_put_task(task);
2671 	return status;
2672 out_free_calldata:
2673 	kfree(calldata);
2674 out:
2675 	nfs4_put_open_state(state);
2676 	nfs4_put_state_owner(sp);
2677 	return status;
2678 }
2679 
2680 static struct inode *
2681 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2682 		int open_flags, struct iattr *attr, int *opened)
2683 {
2684 	struct nfs4_state *state;
2685 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2686 
2687 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2688 
2689 	/* Protect against concurrent sillydeletes */
2690 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2691 
2692 	nfs4_label_release_security(label);
2693 
2694 	if (IS_ERR(state))
2695 		return ERR_CAST(state);
2696 	return state->inode;
2697 }
2698 
2699 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2700 {
2701 	if (ctx->state == NULL)
2702 		return;
2703 	if (is_sync)
2704 		nfs4_close_sync(ctx->state, ctx->mode);
2705 	else
2706 		nfs4_close_state(ctx->state, ctx->mode);
2707 }
2708 
2709 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2710 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2711 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_CHANGE_SECURITY_LABEL - 1UL)
2712 
2713 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2714 {
2715 	struct nfs4_server_caps_arg args = {
2716 		.fhandle = fhandle,
2717 	};
2718 	struct nfs4_server_caps_res res = {};
2719 	struct rpc_message msg = {
2720 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2721 		.rpc_argp = &args,
2722 		.rpc_resp = &res,
2723 	};
2724 	int status;
2725 
2726 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2727 	if (status == 0) {
2728 		/* Sanity check the server answers */
2729 		switch (server->nfs_client->cl_minorversion) {
2730 		case 0:
2731 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2732 			res.attr_bitmask[2] = 0;
2733 			break;
2734 		case 1:
2735 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2736 			break;
2737 		case 2:
2738 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2739 		}
2740 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2741 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2742 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2743 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2744 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2745 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2746 				NFS_CAP_SECURITY_LABEL);
2747 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2748 			server->caps |= NFS_CAP_ACLS;
2749 		if (res.has_links != 0)
2750 			server->caps |= NFS_CAP_HARDLINKS;
2751 		if (res.has_symlinks != 0)
2752 			server->caps |= NFS_CAP_SYMLINKS;
2753 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2754 			server->caps |= NFS_CAP_FILEID;
2755 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2756 			server->caps |= NFS_CAP_MODE;
2757 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2758 			server->caps |= NFS_CAP_NLINK;
2759 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2760 			server->caps |= NFS_CAP_OWNER;
2761 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2762 			server->caps |= NFS_CAP_OWNER_GROUP;
2763 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2764 			server->caps |= NFS_CAP_ATIME;
2765 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2766 			server->caps |= NFS_CAP_CTIME;
2767 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2768 			server->caps |= NFS_CAP_MTIME;
2769 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2770 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2771 			server->caps |= NFS_CAP_SECURITY_LABEL;
2772 #endif
2773 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2774 				sizeof(server->attr_bitmask));
2775 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2776 
2777 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2778 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2779 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2780 		server->cache_consistency_bitmask[2] = 0;
2781 		server->acl_bitmask = res.acl_bitmask;
2782 		server->fh_expire_type = res.fh_expire_type;
2783 	}
2784 
2785 	return status;
2786 }
2787 
2788 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2789 {
2790 	struct nfs4_exception exception = { };
2791 	int err;
2792 	do {
2793 		err = nfs4_handle_exception(server,
2794 				_nfs4_server_capabilities(server, fhandle),
2795 				&exception);
2796 	} while (exception.retry);
2797 	return err;
2798 }
2799 
2800 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2801 		struct nfs_fsinfo *info)
2802 {
2803 	u32 bitmask[3];
2804 	struct nfs4_lookup_root_arg args = {
2805 		.bitmask = bitmask,
2806 	};
2807 	struct nfs4_lookup_res res = {
2808 		.server = server,
2809 		.fattr = info->fattr,
2810 		.fh = fhandle,
2811 	};
2812 	struct rpc_message msg = {
2813 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2814 		.rpc_argp = &args,
2815 		.rpc_resp = &res,
2816 	};
2817 
2818 	bitmask[0] = nfs4_fattr_bitmap[0];
2819 	bitmask[1] = nfs4_fattr_bitmap[1];
2820 	/*
2821 	 * Process the label in the upcoming getfattr
2822 	 */
2823 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2824 
2825 	nfs_fattr_init(info->fattr);
2826 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2827 }
2828 
2829 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2830 		struct nfs_fsinfo *info)
2831 {
2832 	struct nfs4_exception exception = { };
2833 	int err;
2834 	do {
2835 		err = _nfs4_lookup_root(server, fhandle, info);
2836 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2837 		switch (err) {
2838 		case 0:
2839 		case -NFS4ERR_WRONGSEC:
2840 			goto out;
2841 		default:
2842 			err = nfs4_handle_exception(server, err, &exception);
2843 		}
2844 	} while (exception.retry);
2845 out:
2846 	return err;
2847 }
2848 
2849 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2850 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2851 {
2852 	struct rpc_auth_create_args auth_args = {
2853 		.pseudoflavor = flavor,
2854 	};
2855 	struct rpc_auth *auth;
2856 	int ret;
2857 
2858 	auth = rpcauth_create(&auth_args, server->client);
2859 	if (IS_ERR(auth)) {
2860 		ret = -EACCES;
2861 		goto out;
2862 	}
2863 	ret = nfs4_lookup_root(server, fhandle, info);
2864 out:
2865 	return ret;
2866 }
2867 
2868 /*
2869  * Retry pseudoroot lookup with various security flavors.  We do this when:
2870  *
2871  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2872  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2873  *
2874  * Returns zero on success, or a negative NFS4ERR value, or a
2875  * negative errno value.
2876  */
2877 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2878 			      struct nfs_fsinfo *info)
2879 {
2880 	/* Per 3530bis 15.33.5 */
2881 	static const rpc_authflavor_t flav_array[] = {
2882 		RPC_AUTH_GSS_KRB5P,
2883 		RPC_AUTH_GSS_KRB5I,
2884 		RPC_AUTH_GSS_KRB5,
2885 		RPC_AUTH_UNIX,			/* courtesy */
2886 		RPC_AUTH_NULL,
2887 	};
2888 	int status = -EPERM;
2889 	size_t i;
2890 
2891 	if (server->auth_info.flavor_len > 0) {
2892 		/* try each flavor specified by user */
2893 		for (i = 0; i < server->auth_info.flavor_len; i++) {
2894 			status = nfs4_lookup_root_sec(server, fhandle, info,
2895 						server->auth_info.flavors[i]);
2896 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2897 				continue;
2898 			break;
2899 		}
2900 	} else {
2901 		/* no flavors specified by user, try default list */
2902 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2903 			status = nfs4_lookup_root_sec(server, fhandle, info,
2904 						      flav_array[i]);
2905 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2906 				continue;
2907 			break;
2908 		}
2909 	}
2910 
2911 	/*
2912 	 * -EACCESS could mean that the user doesn't have correct permissions
2913 	 * to access the mount.  It could also mean that we tried to mount
2914 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2915 	 * existing mount programs don't handle -EACCES very well so it should
2916 	 * be mapped to -EPERM instead.
2917 	 */
2918 	if (status == -EACCES)
2919 		status = -EPERM;
2920 	return status;
2921 }
2922 
2923 static int nfs4_do_find_root_sec(struct nfs_server *server,
2924 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2925 {
2926 	int mv = server->nfs_client->cl_minorversion;
2927 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2928 }
2929 
2930 /**
2931  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2932  * @server: initialized nfs_server handle
2933  * @fhandle: we fill in the pseudo-fs root file handle
2934  * @info: we fill in an FSINFO struct
2935  * @auth_probe: probe the auth flavours
2936  *
2937  * Returns zero on success, or a negative errno.
2938  */
2939 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2940 			 struct nfs_fsinfo *info,
2941 			 bool auth_probe)
2942 {
2943 	int status;
2944 
2945 	switch (auth_probe) {
2946 	case false:
2947 		status = nfs4_lookup_root(server, fhandle, info);
2948 		if (status != -NFS4ERR_WRONGSEC)
2949 			break;
2950 	default:
2951 		status = nfs4_do_find_root_sec(server, fhandle, info);
2952 	}
2953 
2954 	if (status == 0)
2955 		status = nfs4_server_capabilities(server, fhandle);
2956 	if (status == 0)
2957 		status = nfs4_do_fsinfo(server, fhandle, info);
2958 
2959 	return nfs4_map_errors(status);
2960 }
2961 
2962 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2963 			      struct nfs_fsinfo *info)
2964 {
2965 	int error;
2966 	struct nfs_fattr *fattr = info->fattr;
2967 	struct nfs4_label *label = NULL;
2968 
2969 	error = nfs4_server_capabilities(server, mntfh);
2970 	if (error < 0) {
2971 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2972 		return error;
2973 	}
2974 
2975 	label = nfs4_label_alloc(server, GFP_KERNEL);
2976 	if (IS_ERR(label))
2977 		return PTR_ERR(label);
2978 
2979 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
2980 	if (error < 0) {
2981 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
2982 		goto err_free_label;
2983 	}
2984 
2985 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2986 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2987 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2988 
2989 err_free_label:
2990 	nfs4_label_free(label);
2991 
2992 	return error;
2993 }
2994 
2995 /*
2996  * Get locations and (maybe) other attributes of a referral.
2997  * Note that we'll actually follow the referral later when
2998  * we detect fsid mismatch in inode revalidation
2999  */
3000 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3001 			     const struct qstr *name, struct nfs_fattr *fattr,
3002 			     struct nfs_fh *fhandle)
3003 {
3004 	int status = -ENOMEM;
3005 	struct page *page = NULL;
3006 	struct nfs4_fs_locations *locations = NULL;
3007 
3008 	page = alloc_page(GFP_KERNEL);
3009 	if (page == NULL)
3010 		goto out;
3011 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3012 	if (locations == NULL)
3013 		goto out;
3014 
3015 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3016 	if (status != 0)
3017 		goto out;
3018 
3019 	/*
3020 	 * If the fsid didn't change, this is a migration event, not a
3021 	 * referral.  Cause us to drop into the exception handler, which
3022 	 * will kick off migration recovery.
3023 	 */
3024 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3025 		dprintk("%s: server did not return a different fsid for"
3026 			" a referral at %s\n", __func__, name->name);
3027 		status = -NFS4ERR_MOVED;
3028 		goto out;
3029 	}
3030 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3031 	nfs_fixup_referral_attributes(&locations->fattr);
3032 
3033 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3034 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3035 	memset(fhandle, 0, sizeof(struct nfs_fh));
3036 out:
3037 	if (page)
3038 		__free_page(page);
3039 	kfree(locations);
3040 	return status;
3041 }
3042 
3043 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3044 				struct nfs_fattr *fattr, struct nfs4_label *label)
3045 {
3046 	struct nfs4_getattr_arg args = {
3047 		.fh = fhandle,
3048 		.bitmask = server->attr_bitmask,
3049 	};
3050 	struct nfs4_getattr_res res = {
3051 		.fattr = fattr,
3052 		.label = label,
3053 		.server = server,
3054 	};
3055 	struct rpc_message msg = {
3056 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3057 		.rpc_argp = &args,
3058 		.rpc_resp = &res,
3059 	};
3060 
3061 	args.bitmask = nfs4_bitmask(server, label);
3062 
3063 	nfs_fattr_init(fattr);
3064 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3065 }
3066 
3067 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3068 				struct nfs_fattr *fattr, struct nfs4_label *label)
3069 {
3070 	struct nfs4_exception exception = { };
3071 	int err;
3072 	do {
3073 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3074 		trace_nfs4_getattr(server, fhandle, fattr, err);
3075 		err = nfs4_handle_exception(server, err,
3076 				&exception);
3077 	} while (exception.retry);
3078 	return err;
3079 }
3080 
3081 /*
3082  * The file is not closed if it is opened due to the a request to change
3083  * the size of the file. The open call will not be needed once the
3084  * VFS layer lookup-intents are implemented.
3085  *
3086  * Close is called when the inode is destroyed.
3087  * If we haven't opened the file for O_WRONLY, we
3088  * need to in the size_change case to obtain a stateid.
3089  *
3090  * Got race?
3091  * Because OPEN is always done by name in nfsv4, it is
3092  * possible that we opened a different file by the same
3093  * name.  We can recognize this race condition, but we
3094  * can't do anything about it besides returning an error.
3095  *
3096  * This will be fixed with VFS changes (lookup-intent).
3097  */
3098 static int
3099 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3100 		  struct iattr *sattr)
3101 {
3102 	struct inode *inode = dentry->d_inode;
3103 	struct rpc_cred *cred = NULL;
3104 	struct nfs4_state *state = NULL;
3105 	struct nfs4_label *label = NULL;
3106 	int status;
3107 
3108 	if (pnfs_ld_layoutret_on_setattr(inode))
3109 		pnfs_commit_and_return_layout(inode);
3110 
3111 	nfs_fattr_init(fattr);
3112 
3113 	/* Deal with open(O_TRUNC) */
3114 	if (sattr->ia_valid & ATTR_OPEN)
3115 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3116 
3117 	/* Optimization: if the end result is no change, don't RPC */
3118 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3119 		return 0;
3120 
3121 	/* Search for an existing open(O_WRITE) file */
3122 	if (sattr->ia_valid & ATTR_FILE) {
3123 		struct nfs_open_context *ctx;
3124 
3125 		ctx = nfs_file_open_context(sattr->ia_file);
3126 		if (ctx) {
3127 			cred = ctx->cred;
3128 			state = ctx->state;
3129 		}
3130 	}
3131 
3132 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3133 	if (IS_ERR(label))
3134 		return PTR_ERR(label);
3135 
3136 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3137 	if (status == 0) {
3138 		nfs_setattr_update_inode(inode, sattr);
3139 		nfs_setsecurity(inode, fattr, label);
3140 	}
3141 	nfs4_label_free(label);
3142 	return status;
3143 }
3144 
3145 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3146 		const struct qstr *name, struct nfs_fh *fhandle,
3147 		struct nfs_fattr *fattr, struct nfs4_label *label)
3148 {
3149 	struct nfs_server *server = NFS_SERVER(dir);
3150 	int		       status;
3151 	struct nfs4_lookup_arg args = {
3152 		.bitmask = server->attr_bitmask,
3153 		.dir_fh = NFS_FH(dir),
3154 		.name = name,
3155 	};
3156 	struct nfs4_lookup_res res = {
3157 		.server = server,
3158 		.fattr = fattr,
3159 		.label = label,
3160 		.fh = fhandle,
3161 	};
3162 	struct rpc_message msg = {
3163 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3164 		.rpc_argp = &args,
3165 		.rpc_resp = &res,
3166 	};
3167 
3168 	args.bitmask = nfs4_bitmask(server, label);
3169 
3170 	nfs_fattr_init(fattr);
3171 
3172 	dprintk("NFS call  lookup %s\n", name->name);
3173 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3174 	dprintk("NFS reply lookup: %d\n", status);
3175 	return status;
3176 }
3177 
3178 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3179 {
3180 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3181 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3182 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3183 	fattr->nlink = 2;
3184 }
3185 
3186 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3187 				   struct qstr *name, struct nfs_fh *fhandle,
3188 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3189 {
3190 	struct nfs4_exception exception = { };
3191 	struct rpc_clnt *client = *clnt;
3192 	int err;
3193 	do {
3194 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3195 		trace_nfs4_lookup(dir, name, err);
3196 		switch (err) {
3197 		case -NFS4ERR_BADNAME:
3198 			err = -ENOENT;
3199 			goto out;
3200 		case -NFS4ERR_MOVED:
3201 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3202 			goto out;
3203 		case -NFS4ERR_WRONGSEC:
3204 			err = -EPERM;
3205 			if (client != *clnt)
3206 				goto out;
3207 			client = nfs4_create_sec_client(client, dir, name);
3208 			if (IS_ERR(client))
3209 				return PTR_ERR(client);
3210 
3211 			exception.retry = 1;
3212 			break;
3213 		default:
3214 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3215 		}
3216 	} while (exception.retry);
3217 
3218 out:
3219 	if (err == 0)
3220 		*clnt = client;
3221 	else if (client != *clnt)
3222 		rpc_shutdown_client(client);
3223 
3224 	return err;
3225 }
3226 
3227 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3228 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3229 			    struct nfs4_label *label)
3230 {
3231 	int status;
3232 	struct rpc_clnt *client = NFS_CLIENT(dir);
3233 
3234 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3235 	if (client != NFS_CLIENT(dir)) {
3236 		rpc_shutdown_client(client);
3237 		nfs_fixup_secinfo_attributes(fattr);
3238 	}
3239 	return status;
3240 }
3241 
3242 struct rpc_clnt *
3243 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3244 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3245 {
3246 	struct rpc_clnt *client = NFS_CLIENT(dir);
3247 	int status;
3248 
3249 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3250 	if (status < 0)
3251 		return ERR_PTR(status);
3252 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3253 }
3254 
3255 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3256 {
3257 	struct nfs_server *server = NFS_SERVER(inode);
3258 	struct nfs4_accessargs args = {
3259 		.fh = NFS_FH(inode),
3260 		.bitmask = server->cache_consistency_bitmask,
3261 	};
3262 	struct nfs4_accessres res = {
3263 		.server = server,
3264 	};
3265 	struct rpc_message msg = {
3266 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3267 		.rpc_argp = &args,
3268 		.rpc_resp = &res,
3269 		.rpc_cred = entry->cred,
3270 	};
3271 	int mode = entry->mask;
3272 	int status = 0;
3273 
3274 	/*
3275 	 * Determine which access bits we want to ask for...
3276 	 */
3277 	if (mode & MAY_READ)
3278 		args.access |= NFS4_ACCESS_READ;
3279 	if (S_ISDIR(inode->i_mode)) {
3280 		if (mode & MAY_WRITE)
3281 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3282 		if (mode & MAY_EXEC)
3283 			args.access |= NFS4_ACCESS_LOOKUP;
3284 	} else {
3285 		if (mode & MAY_WRITE)
3286 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3287 		if (mode & MAY_EXEC)
3288 			args.access |= NFS4_ACCESS_EXECUTE;
3289 	}
3290 
3291 	res.fattr = nfs_alloc_fattr();
3292 	if (res.fattr == NULL)
3293 		return -ENOMEM;
3294 
3295 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3296 	if (!status) {
3297 		nfs_access_set_mask(entry, res.access);
3298 		nfs_refresh_inode(inode, res.fattr);
3299 	}
3300 	nfs_free_fattr(res.fattr);
3301 	return status;
3302 }
3303 
3304 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3305 {
3306 	struct nfs4_exception exception = { };
3307 	int err;
3308 	do {
3309 		err = _nfs4_proc_access(inode, entry);
3310 		trace_nfs4_access(inode, err);
3311 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3312 				&exception);
3313 	} while (exception.retry);
3314 	return err;
3315 }
3316 
3317 /*
3318  * TODO: For the time being, we don't try to get any attributes
3319  * along with any of the zero-copy operations READ, READDIR,
3320  * READLINK, WRITE.
3321  *
3322  * In the case of the first three, we want to put the GETATTR
3323  * after the read-type operation -- this is because it is hard
3324  * to predict the length of a GETATTR response in v4, and thus
3325  * align the READ data correctly.  This means that the GETATTR
3326  * may end up partially falling into the page cache, and we should
3327  * shift it into the 'tail' of the xdr_buf before processing.
3328  * To do this efficiently, we need to know the total length
3329  * of data received, which doesn't seem to be available outside
3330  * of the RPC layer.
3331  *
3332  * In the case of WRITE, we also want to put the GETATTR after
3333  * the operation -- in this case because we want to make sure
3334  * we get the post-operation mtime and size.
3335  *
3336  * Both of these changes to the XDR layer would in fact be quite
3337  * minor, but I decided to leave them for a subsequent patch.
3338  */
3339 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3340 		unsigned int pgbase, unsigned int pglen)
3341 {
3342 	struct nfs4_readlink args = {
3343 		.fh       = NFS_FH(inode),
3344 		.pgbase	  = pgbase,
3345 		.pglen    = pglen,
3346 		.pages    = &page,
3347 	};
3348 	struct nfs4_readlink_res res;
3349 	struct rpc_message msg = {
3350 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3351 		.rpc_argp = &args,
3352 		.rpc_resp = &res,
3353 	};
3354 
3355 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3356 }
3357 
3358 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3359 		unsigned int pgbase, unsigned int pglen)
3360 {
3361 	struct nfs4_exception exception = { };
3362 	int err;
3363 	do {
3364 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3365 		trace_nfs4_readlink(inode, err);
3366 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3367 				&exception);
3368 	} while (exception.retry);
3369 	return err;
3370 }
3371 
3372 /*
3373  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3374  */
3375 static int
3376 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3377 		 int flags)
3378 {
3379 	struct nfs4_label l, *ilabel = NULL;
3380 	struct nfs_open_context *ctx;
3381 	struct nfs4_state *state;
3382 	int opened = 0;
3383 	int status = 0;
3384 
3385 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3386 	if (IS_ERR(ctx))
3387 		return PTR_ERR(ctx);
3388 
3389 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3390 
3391 	sattr->ia_mode &= ~current_umask();
3392 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3393 	if (IS_ERR(state)) {
3394 		status = PTR_ERR(state);
3395 		goto out;
3396 	}
3397 out:
3398 	nfs4_label_release_security(ilabel);
3399 	put_nfs_open_context(ctx);
3400 	return status;
3401 }
3402 
3403 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3404 {
3405 	struct nfs_server *server = NFS_SERVER(dir);
3406 	struct nfs_removeargs args = {
3407 		.fh = NFS_FH(dir),
3408 		.name = *name,
3409 	};
3410 	struct nfs_removeres res = {
3411 		.server = server,
3412 	};
3413 	struct rpc_message msg = {
3414 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3415 		.rpc_argp = &args,
3416 		.rpc_resp = &res,
3417 	};
3418 	int status;
3419 
3420 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3421 	if (status == 0)
3422 		update_changeattr(dir, &res.cinfo);
3423 	return status;
3424 }
3425 
3426 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3427 {
3428 	struct nfs4_exception exception = { };
3429 	int err;
3430 	do {
3431 		err = _nfs4_proc_remove(dir, name);
3432 		trace_nfs4_remove(dir, name, err);
3433 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3434 				&exception);
3435 	} while (exception.retry);
3436 	return err;
3437 }
3438 
3439 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3440 {
3441 	struct nfs_server *server = NFS_SERVER(dir);
3442 	struct nfs_removeargs *args = msg->rpc_argp;
3443 	struct nfs_removeres *res = msg->rpc_resp;
3444 
3445 	res->server = server;
3446 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3447 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3448 
3449 	nfs_fattr_init(res->dir_attr);
3450 }
3451 
3452 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3453 {
3454 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3455 			&data->args.seq_args,
3456 			&data->res.seq_res,
3457 			task);
3458 }
3459 
3460 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3461 {
3462 	struct nfs_unlinkdata *data = task->tk_calldata;
3463 	struct nfs_removeres *res = &data->res;
3464 
3465 	if (!nfs4_sequence_done(task, &res->seq_res))
3466 		return 0;
3467 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3468 		return 0;
3469 	update_changeattr(dir, &res->cinfo);
3470 	return 1;
3471 }
3472 
3473 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3474 {
3475 	struct nfs_server *server = NFS_SERVER(dir);
3476 	struct nfs_renameargs *arg = msg->rpc_argp;
3477 	struct nfs_renameres *res = msg->rpc_resp;
3478 
3479 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3480 	res->server = server;
3481 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3482 }
3483 
3484 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3485 {
3486 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3487 			&data->args.seq_args,
3488 			&data->res.seq_res,
3489 			task);
3490 }
3491 
3492 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3493 				 struct inode *new_dir)
3494 {
3495 	struct nfs_renamedata *data = task->tk_calldata;
3496 	struct nfs_renameres *res = &data->res;
3497 
3498 	if (!nfs4_sequence_done(task, &res->seq_res))
3499 		return 0;
3500 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3501 		return 0;
3502 
3503 	update_changeattr(old_dir, &res->old_cinfo);
3504 	update_changeattr(new_dir, &res->new_cinfo);
3505 	return 1;
3506 }
3507 
3508 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3509 		struct inode *new_dir, struct qstr *new_name)
3510 {
3511 	struct nfs_server *server = NFS_SERVER(old_dir);
3512 	struct nfs_renameargs arg = {
3513 		.old_dir = NFS_FH(old_dir),
3514 		.new_dir = NFS_FH(new_dir),
3515 		.old_name = old_name,
3516 		.new_name = new_name,
3517 	};
3518 	struct nfs_renameres res = {
3519 		.server = server,
3520 	};
3521 	struct rpc_message msg = {
3522 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3523 		.rpc_argp = &arg,
3524 		.rpc_resp = &res,
3525 	};
3526 	int status = -ENOMEM;
3527 
3528 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3529 	if (!status) {
3530 		update_changeattr(old_dir, &res.old_cinfo);
3531 		update_changeattr(new_dir, &res.new_cinfo);
3532 	}
3533 	return status;
3534 }
3535 
3536 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3537 		struct inode *new_dir, struct qstr *new_name)
3538 {
3539 	struct nfs4_exception exception = { };
3540 	int err;
3541 	do {
3542 		err = _nfs4_proc_rename(old_dir, old_name,
3543 					new_dir, new_name);
3544 		trace_nfs4_rename(old_dir, old_name, new_dir, new_name, err);
3545 		err = nfs4_handle_exception(NFS_SERVER(old_dir), err,
3546 				&exception);
3547 	} while (exception.retry);
3548 	return err;
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 (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode))
4016 		return false;
4017 	return nfs4_stateid_match(stateid, &current_stateid);
4018 }
4019 
4020 static bool nfs4_error_stateid_expired(int err)
4021 {
4022 	switch (err) {
4023 	case -NFS4ERR_DELEG_REVOKED:
4024 	case -NFS4ERR_ADMIN_REVOKED:
4025 	case -NFS4ERR_BAD_STATEID:
4026 	case -NFS4ERR_STALE_STATEID:
4027 	case -NFS4ERR_OLD_STATEID:
4028 	case -NFS4ERR_OPENMODE:
4029 	case -NFS4ERR_EXPIRED:
4030 		return true;
4031 	}
4032 	return false;
4033 }
4034 
4035 void __nfs4_read_done_cb(struct nfs_read_data *data)
4036 {
4037 	nfs_invalidate_atime(data->header->inode);
4038 }
4039 
4040 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
4041 {
4042 	struct nfs_server *server = NFS_SERVER(data->header->inode);
4043 
4044 	trace_nfs4_read(data, task->tk_status);
4045 	if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
4046 		rpc_restart_call_prepare(task);
4047 		return -EAGAIN;
4048 	}
4049 
4050 	__nfs4_read_done_cb(data);
4051 	if (task->tk_status > 0)
4052 		renew_lease(server, data->timestamp);
4053 	return 0;
4054 }
4055 
4056 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4057 		struct nfs_readargs *args)
4058 {
4059 
4060 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4061 		nfs4_stateid_is_current(&args->stateid,
4062 				args->context,
4063 				args->lock_context,
4064 				FMODE_READ))
4065 		return false;
4066 	rpc_restart_call_prepare(task);
4067 	return true;
4068 }
4069 
4070 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
4071 {
4072 
4073 	dprintk("--> %s\n", __func__);
4074 
4075 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4076 		return -EAGAIN;
4077 	if (nfs4_read_stateid_changed(task, &data->args))
4078 		return -EAGAIN;
4079 	return data->read_done_cb ? data->read_done_cb(task, data) :
4080 				    nfs4_read_done_cb(task, data);
4081 }
4082 
4083 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
4084 {
4085 	data->timestamp   = jiffies;
4086 	data->read_done_cb = nfs4_read_done_cb;
4087 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4088 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
4089 }
4090 
4091 static int nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
4092 {
4093 	if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
4094 			&data->args.seq_args,
4095 			&data->res.seq_res,
4096 			task))
4097 		return 0;
4098 	if (nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
4099 				data->args.lock_context, FMODE_READ) == -EIO)
4100 		return -EIO;
4101 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &data->args.context->flags)))
4102 		return -EIO;
4103 	return 0;
4104 }
4105 
4106 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
4107 {
4108 	struct inode *inode = data->header->inode;
4109 
4110 	trace_nfs4_write(data, task->tk_status);
4111 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
4112 		rpc_restart_call_prepare(task);
4113 		return -EAGAIN;
4114 	}
4115 	if (task->tk_status >= 0) {
4116 		renew_lease(NFS_SERVER(inode), data->timestamp);
4117 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4118 	}
4119 	return 0;
4120 }
4121 
4122 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4123 		struct nfs_writeargs *args)
4124 {
4125 
4126 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4127 		nfs4_stateid_is_current(&args->stateid,
4128 				args->context,
4129 				args->lock_context,
4130 				FMODE_WRITE))
4131 		return false;
4132 	rpc_restart_call_prepare(task);
4133 	return true;
4134 }
4135 
4136 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
4137 {
4138 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4139 		return -EAGAIN;
4140 	if (nfs4_write_stateid_changed(task, &data->args))
4141 		return -EAGAIN;
4142 	return data->write_done_cb ? data->write_done_cb(task, data) :
4143 		nfs4_write_done_cb(task, data);
4144 }
4145 
4146 static
4147 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
4148 {
4149 	const struct nfs_pgio_header *hdr = data->header;
4150 
4151 	/* Don't request attributes for pNFS or O_DIRECT writes */
4152 	if (data->ds_clp != NULL || hdr->dreq != NULL)
4153 		return false;
4154 	/* Otherwise, request attributes if and only if we don't hold
4155 	 * a delegation
4156 	 */
4157 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4158 }
4159 
4160 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
4161 {
4162 	struct nfs_server *server = NFS_SERVER(data->header->inode);
4163 
4164 	if (!nfs4_write_need_cache_consistency_data(data)) {
4165 		data->args.bitmask = NULL;
4166 		data->res.fattr = NULL;
4167 	} else
4168 		data->args.bitmask = server->cache_consistency_bitmask;
4169 
4170 	if (!data->write_done_cb)
4171 		data->write_done_cb = nfs4_write_done_cb;
4172 	data->res.server = server;
4173 	data->timestamp   = jiffies;
4174 
4175 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4176 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4177 }
4178 
4179 static int nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
4180 {
4181 	if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
4182 			&data->args.seq_args,
4183 			&data->res.seq_res,
4184 			task))
4185 		return 0;
4186 	if (nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
4187 				data->args.lock_context, FMODE_WRITE) == -EIO)
4188 		return -EIO;
4189 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &data->args.context->flags)))
4190 		return -EIO;
4191 	return 0;
4192 }
4193 
4194 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4195 {
4196 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4197 			&data->args.seq_args,
4198 			&data->res.seq_res,
4199 			task);
4200 }
4201 
4202 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4203 {
4204 	struct inode *inode = data->inode;
4205 
4206 	trace_nfs4_commit(data, task->tk_status);
4207 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4208 		rpc_restart_call_prepare(task);
4209 		return -EAGAIN;
4210 	}
4211 	return 0;
4212 }
4213 
4214 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4215 {
4216 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4217 		return -EAGAIN;
4218 	return data->commit_done_cb(task, data);
4219 }
4220 
4221 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4222 {
4223 	struct nfs_server *server = NFS_SERVER(data->inode);
4224 
4225 	if (data->commit_done_cb == NULL)
4226 		data->commit_done_cb = nfs4_commit_done_cb;
4227 	data->res.server = server;
4228 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4229 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4230 }
4231 
4232 struct nfs4_renewdata {
4233 	struct nfs_client	*client;
4234 	unsigned long		timestamp;
4235 };
4236 
4237 /*
4238  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4239  * standalone procedure for queueing an asynchronous RENEW.
4240  */
4241 static void nfs4_renew_release(void *calldata)
4242 {
4243 	struct nfs4_renewdata *data = calldata;
4244 	struct nfs_client *clp = data->client;
4245 
4246 	if (atomic_read(&clp->cl_count) > 1)
4247 		nfs4_schedule_state_renewal(clp);
4248 	nfs_put_client(clp);
4249 	kfree(data);
4250 }
4251 
4252 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4253 {
4254 	struct nfs4_renewdata *data = calldata;
4255 	struct nfs_client *clp = data->client;
4256 	unsigned long timestamp = data->timestamp;
4257 
4258 	trace_nfs4_renew_async(clp, task->tk_status);
4259 	switch (task->tk_status) {
4260 	case 0:
4261 		break;
4262 	case -NFS4ERR_LEASE_MOVED:
4263 		nfs4_schedule_lease_moved_recovery(clp);
4264 		break;
4265 	default:
4266 		/* Unless we're shutting down, schedule state recovery! */
4267 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4268 			return;
4269 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4270 			nfs4_schedule_lease_recovery(clp);
4271 			return;
4272 		}
4273 		nfs4_schedule_path_down_recovery(clp);
4274 	}
4275 	do_renew_lease(clp, timestamp);
4276 }
4277 
4278 static const struct rpc_call_ops nfs4_renew_ops = {
4279 	.rpc_call_done = nfs4_renew_done,
4280 	.rpc_release = nfs4_renew_release,
4281 };
4282 
4283 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4284 {
4285 	struct rpc_message msg = {
4286 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4287 		.rpc_argp	= clp,
4288 		.rpc_cred	= cred,
4289 	};
4290 	struct nfs4_renewdata *data;
4291 
4292 	if (renew_flags == 0)
4293 		return 0;
4294 	if (!atomic_inc_not_zero(&clp->cl_count))
4295 		return -EIO;
4296 	data = kmalloc(sizeof(*data), GFP_NOFS);
4297 	if (data == NULL)
4298 		return -ENOMEM;
4299 	data->client = clp;
4300 	data->timestamp = jiffies;
4301 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4302 			&nfs4_renew_ops, data);
4303 }
4304 
4305 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4306 {
4307 	struct rpc_message msg = {
4308 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4309 		.rpc_argp	= clp,
4310 		.rpc_cred	= cred,
4311 	};
4312 	unsigned long now = jiffies;
4313 	int status;
4314 
4315 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4316 	if (status < 0)
4317 		return status;
4318 	do_renew_lease(clp, now);
4319 	return 0;
4320 }
4321 
4322 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4323 {
4324 	return (server->caps & NFS_CAP_ACLS)
4325 		&& (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
4326 		&& (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
4327 }
4328 
4329 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4330  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4331  * the stack.
4332  */
4333 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4334 
4335 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4336 		struct page **pages, unsigned int *pgbase)
4337 {
4338 	struct page *newpage, **spages;
4339 	int rc = 0;
4340 	size_t len;
4341 	spages = pages;
4342 
4343 	do {
4344 		len = min_t(size_t, PAGE_SIZE, buflen);
4345 		newpage = alloc_page(GFP_KERNEL);
4346 
4347 		if (newpage == NULL)
4348 			goto unwind;
4349 		memcpy(page_address(newpage), buf, len);
4350                 buf += len;
4351                 buflen -= len;
4352 		*pages++ = newpage;
4353 		rc++;
4354 	} while (buflen != 0);
4355 
4356 	return rc;
4357 
4358 unwind:
4359 	for(; rc > 0; rc--)
4360 		__free_page(spages[rc-1]);
4361 	return -ENOMEM;
4362 }
4363 
4364 struct nfs4_cached_acl {
4365 	int cached;
4366 	size_t len;
4367 	char data[0];
4368 };
4369 
4370 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4371 {
4372 	struct nfs_inode *nfsi = NFS_I(inode);
4373 
4374 	spin_lock(&inode->i_lock);
4375 	kfree(nfsi->nfs4_acl);
4376 	nfsi->nfs4_acl = acl;
4377 	spin_unlock(&inode->i_lock);
4378 }
4379 
4380 static void nfs4_zap_acl_attr(struct inode *inode)
4381 {
4382 	nfs4_set_cached_acl(inode, NULL);
4383 }
4384 
4385 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4386 {
4387 	struct nfs_inode *nfsi = NFS_I(inode);
4388 	struct nfs4_cached_acl *acl;
4389 	int ret = -ENOENT;
4390 
4391 	spin_lock(&inode->i_lock);
4392 	acl = nfsi->nfs4_acl;
4393 	if (acl == NULL)
4394 		goto out;
4395 	if (buf == NULL) /* user is just asking for length */
4396 		goto out_len;
4397 	if (acl->cached == 0)
4398 		goto out;
4399 	ret = -ERANGE; /* see getxattr(2) man page */
4400 	if (acl->len > buflen)
4401 		goto out;
4402 	memcpy(buf, acl->data, acl->len);
4403 out_len:
4404 	ret = acl->len;
4405 out:
4406 	spin_unlock(&inode->i_lock);
4407 	return ret;
4408 }
4409 
4410 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4411 {
4412 	struct nfs4_cached_acl *acl;
4413 	size_t buflen = sizeof(*acl) + acl_len;
4414 
4415 	if (buflen <= PAGE_SIZE) {
4416 		acl = kmalloc(buflen, GFP_KERNEL);
4417 		if (acl == NULL)
4418 			goto out;
4419 		acl->cached = 1;
4420 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4421 	} else {
4422 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4423 		if (acl == NULL)
4424 			goto out;
4425 		acl->cached = 0;
4426 	}
4427 	acl->len = acl_len;
4428 out:
4429 	nfs4_set_cached_acl(inode, acl);
4430 }
4431 
4432 /*
4433  * The getxattr API returns the required buffer length when called with a
4434  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4435  * the required buf.  On a NULL buf, we send a page of data to the server
4436  * guessing that the ACL request can be serviced by a page. If so, we cache
4437  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4438  * the cache. If not so, we throw away the page, and cache the required
4439  * length. The next getxattr call will then produce another round trip to
4440  * the server, this time with the input buf of the required size.
4441  */
4442 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4443 {
4444 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4445 	struct nfs_getaclargs args = {
4446 		.fh = NFS_FH(inode),
4447 		.acl_pages = pages,
4448 		.acl_len = buflen,
4449 	};
4450 	struct nfs_getaclres res = {
4451 		.acl_len = buflen,
4452 	};
4453 	struct rpc_message msg = {
4454 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4455 		.rpc_argp = &args,
4456 		.rpc_resp = &res,
4457 	};
4458 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4459 	int ret = -ENOMEM, i;
4460 
4461 	/* As long as we're doing a round trip to the server anyway,
4462 	 * let's be prepared for a page of acl data. */
4463 	if (npages == 0)
4464 		npages = 1;
4465 	if (npages > ARRAY_SIZE(pages))
4466 		return -ERANGE;
4467 
4468 	for (i = 0; i < npages; i++) {
4469 		pages[i] = alloc_page(GFP_KERNEL);
4470 		if (!pages[i])
4471 			goto out_free;
4472 	}
4473 
4474 	/* for decoding across pages */
4475 	res.acl_scratch = alloc_page(GFP_KERNEL);
4476 	if (!res.acl_scratch)
4477 		goto out_free;
4478 
4479 	args.acl_len = npages * PAGE_SIZE;
4480 	args.acl_pgbase = 0;
4481 
4482 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4483 		__func__, buf, buflen, npages, args.acl_len);
4484 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4485 			     &msg, &args.seq_args, &res.seq_res, 0);
4486 	if (ret)
4487 		goto out_free;
4488 
4489 	/* Handle the case where the passed-in buffer is too short */
4490 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4491 		/* Did the user only issue a request for the acl length? */
4492 		if (buf == NULL)
4493 			goto out_ok;
4494 		ret = -ERANGE;
4495 		goto out_free;
4496 	}
4497 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4498 	if (buf) {
4499 		if (res.acl_len > buflen) {
4500 			ret = -ERANGE;
4501 			goto out_free;
4502 		}
4503 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4504 	}
4505 out_ok:
4506 	ret = res.acl_len;
4507 out_free:
4508 	for (i = 0; i < npages; i++)
4509 		if (pages[i])
4510 			__free_page(pages[i]);
4511 	if (res.acl_scratch)
4512 		__free_page(res.acl_scratch);
4513 	return ret;
4514 }
4515 
4516 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4517 {
4518 	struct nfs4_exception exception = { };
4519 	ssize_t ret;
4520 	do {
4521 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4522 		trace_nfs4_get_acl(inode, ret);
4523 		if (ret >= 0)
4524 			break;
4525 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4526 	} while (exception.retry);
4527 	return ret;
4528 }
4529 
4530 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4531 {
4532 	struct nfs_server *server = NFS_SERVER(inode);
4533 	int ret;
4534 
4535 	if (!nfs4_server_supports_acls(server))
4536 		return -EOPNOTSUPP;
4537 	ret = nfs_revalidate_inode(server, inode);
4538 	if (ret < 0)
4539 		return ret;
4540 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4541 		nfs_zap_acl_cache(inode);
4542 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4543 	if (ret != -ENOENT)
4544 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4545 		 * but no cached acl data, just the acl length */
4546 		return ret;
4547 	return nfs4_get_acl_uncached(inode, buf, buflen);
4548 }
4549 
4550 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4551 {
4552 	struct nfs_server *server = NFS_SERVER(inode);
4553 	struct page *pages[NFS4ACL_MAXPAGES];
4554 	struct nfs_setaclargs arg = {
4555 		.fh		= NFS_FH(inode),
4556 		.acl_pages	= pages,
4557 		.acl_len	= buflen,
4558 	};
4559 	struct nfs_setaclres res;
4560 	struct rpc_message msg = {
4561 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4562 		.rpc_argp	= &arg,
4563 		.rpc_resp	= &res,
4564 	};
4565 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4566 	int ret, i;
4567 
4568 	if (!nfs4_server_supports_acls(server))
4569 		return -EOPNOTSUPP;
4570 	if (npages > ARRAY_SIZE(pages))
4571 		return -ERANGE;
4572 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4573 	if (i < 0)
4574 		return i;
4575 	nfs4_inode_return_delegation(inode);
4576 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4577 
4578 	/*
4579 	 * Free each page after tx, so the only ref left is
4580 	 * held by the network stack
4581 	 */
4582 	for (; i > 0; i--)
4583 		put_page(pages[i-1]);
4584 
4585 	/*
4586 	 * Acl update can result in inode attribute update.
4587 	 * so mark the attribute cache invalid.
4588 	 */
4589 	spin_lock(&inode->i_lock);
4590 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4591 	spin_unlock(&inode->i_lock);
4592 	nfs_access_zap_cache(inode);
4593 	nfs_zap_acl_cache(inode);
4594 	return ret;
4595 }
4596 
4597 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4598 {
4599 	struct nfs4_exception exception = { };
4600 	int err;
4601 	do {
4602 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4603 		trace_nfs4_set_acl(inode, err);
4604 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4605 				&exception);
4606 	} while (exception.retry);
4607 	return err;
4608 }
4609 
4610 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4611 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4612 					size_t buflen)
4613 {
4614 	struct nfs_server *server = NFS_SERVER(inode);
4615 	struct nfs_fattr fattr;
4616 	struct nfs4_label label = {0, 0, buflen, buf};
4617 
4618 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4619 	struct nfs4_getattr_arg arg = {
4620 		.fh		= NFS_FH(inode),
4621 		.bitmask	= bitmask,
4622 	};
4623 	struct nfs4_getattr_res res = {
4624 		.fattr		= &fattr,
4625 		.label		= &label,
4626 		.server		= server,
4627 	};
4628 	struct rpc_message msg = {
4629 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4630 		.rpc_argp	= &arg,
4631 		.rpc_resp	= &res,
4632 	};
4633 	int ret;
4634 
4635 	nfs_fattr_init(&fattr);
4636 
4637 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4638 	if (ret)
4639 		return ret;
4640 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4641 		return -ENOENT;
4642 	if (buflen < label.len)
4643 		return -ERANGE;
4644 	return 0;
4645 }
4646 
4647 static int nfs4_get_security_label(struct inode *inode, void *buf,
4648 					size_t buflen)
4649 {
4650 	struct nfs4_exception exception = { };
4651 	int err;
4652 
4653 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4654 		return -EOPNOTSUPP;
4655 
4656 	do {
4657 		err = _nfs4_get_security_label(inode, buf, buflen);
4658 		trace_nfs4_get_security_label(inode, err);
4659 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4660 				&exception);
4661 	} while (exception.retry);
4662 	return err;
4663 }
4664 
4665 static int _nfs4_do_set_security_label(struct inode *inode,
4666 		struct nfs4_label *ilabel,
4667 		struct nfs_fattr *fattr,
4668 		struct nfs4_label *olabel)
4669 {
4670 
4671 	struct iattr sattr = {0};
4672 	struct nfs_server *server = NFS_SERVER(inode);
4673 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4674 	struct nfs_setattrargs arg = {
4675 		.fh             = NFS_FH(inode),
4676 		.iap            = &sattr,
4677 		.server		= server,
4678 		.bitmask	= bitmask,
4679 		.label		= ilabel,
4680 	};
4681 	struct nfs_setattrres res = {
4682 		.fattr		= fattr,
4683 		.label		= olabel,
4684 		.server		= server,
4685 	};
4686 	struct rpc_message msg = {
4687 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4688 		.rpc_argp       = &arg,
4689 		.rpc_resp       = &res,
4690 	};
4691 	int status;
4692 
4693 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4694 
4695 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4696 	if (status)
4697 		dprintk("%s failed: %d\n", __func__, status);
4698 
4699 	return status;
4700 }
4701 
4702 static int nfs4_do_set_security_label(struct inode *inode,
4703 		struct nfs4_label *ilabel,
4704 		struct nfs_fattr *fattr,
4705 		struct nfs4_label *olabel)
4706 {
4707 	struct nfs4_exception exception = { };
4708 	int err;
4709 
4710 	do {
4711 		err = _nfs4_do_set_security_label(inode, ilabel,
4712 				fattr, olabel);
4713 		trace_nfs4_set_security_label(inode, err);
4714 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4715 				&exception);
4716 	} while (exception.retry);
4717 	return err;
4718 }
4719 
4720 static int
4721 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4722 {
4723 	struct nfs4_label ilabel, *olabel = NULL;
4724 	struct nfs_fattr fattr;
4725 	struct rpc_cred *cred;
4726 	struct inode *inode = dentry->d_inode;
4727 	int status;
4728 
4729 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4730 		return -EOPNOTSUPP;
4731 
4732 	nfs_fattr_init(&fattr);
4733 
4734 	ilabel.pi = 0;
4735 	ilabel.lfs = 0;
4736 	ilabel.label = (char *)buf;
4737 	ilabel.len = buflen;
4738 
4739 	cred = rpc_lookup_cred();
4740 	if (IS_ERR(cred))
4741 		return PTR_ERR(cred);
4742 
4743 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4744 	if (IS_ERR(olabel)) {
4745 		status = -PTR_ERR(olabel);
4746 		goto out;
4747 	}
4748 
4749 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4750 	if (status == 0)
4751 		nfs_setsecurity(inode, &fattr, olabel);
4752 
4753 	nfs4_label_free(olabel);
4754 out:
4755 	put_rpccred(cred);
4756 	return status;
4757 }
4758 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4759 
4760 
4761 static int
4762 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4763 {
4764 	struct nfs_client *clp = server->nfs_client;
4765 
4766 	if (task->tk_status >= 0)
4767 		return 0;
4768 	switch(task->tk_status) {
4769 		case -NFS4ERR_DELEG_REVOKED:
4770 		case -NFS4ERR_ADMIN_REVOKED:
4771 		case -NFS4ERR_BAD_STATEID:
4772 			if (state == NULL)
4773 				break;
4774 			nfs_remove_bad_delegation(state->inode);
4775 		case -NFS4ERR_OPENMODE:
4776 			if (state == NULL)
4777 				break;
4778 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4779 				goto recovery_failed;
4780 			goto wait_on_recovery;
4781 		case -NFS4ERR_EXPIRED:
4782 			if (state != NULL) {
4783 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4784 					goto recovery_failed;
4785 			}
4786 		case -NFS4ERR_STALE_STATEID:
4787 		case -NFS4ERR_STALE_CLIENTID:
4788 			nfs4_schedule_lease_recovery(clp);
4789 			goto wait_on_recovery;
4790 		case -NFS4ERR_MOVED:
4791 			if (nfs4_schedule_migration_recovery(server) < 0)
4792 				goto recovery_failed;
4793 			goto wait_on_recovery;
4794 		case -NFS4ERR_LEASE_MOVED:
4795 			nfs4_schedule_lease_moved_recovery(clp);
4796 			goto wait_on_recovery;
4797 #if defined(CONFIG_NFS_V4_1)
4798 		case -NFS4ERR_BADSESSION:
4799 		case -NFS4ERR_BADSLOT:
4800 		case -NFS4ERR_BAD_HIGH_SLOT:
4801 		case -NFS4ERR_DEADSESSION:
4802 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4803 		case -NFS4ERR_SEQ_FALSE_RETRY:
4804 		case -NFS4ERR_SEQ_MISORDERED:
4805 			dprintk("%s ERROR %d, Reset session\n", __func__,
4806 				task->tk_status);
4807 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4808 			goto wait_on_recovery;
4809 #endif /* CONFIG_NFS_V4_1 */
4810 		case -NFS4ERR_DELAY:
4811 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4812 		case -NFS4ERR_GRACE:
4813 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4814 		case -NFS4ERR_RETRY_UNCACHED_REP:
4815 		case -NFS4ERR_OLD_STATEID:
4816 			goto restart_call;
4817 	}
4818 	task->tk_status = nfs4_map_errors(task->tk_status);
4819 	return 0;
4820 recovery_failed:
4821 	task->tk_status = -EIO;
4822 	return 0;
4823 wait_on_recovery:
4824 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4825 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4826 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4827 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4828 		goto recovery_failed;
4829 restart_call:
4830 	task->tk_status = 0;
4831 	return -EAGAIN;
4832 }
4833 
4834 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4835 				    nfs4_verifier *bootverf)
4836 {
4837 	__be32 verf[2];
4838 
4839 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4840 		/* An impossible timestamp guarantees this value
4841 		 * will never match a generated boot time. */
4842 		verf[0] = 0;
4843 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4844 	} else {
4845 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4846 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4847 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4848 	}
4849 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4850 }
4851 
4852 static unsigned int
4853 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4854 				   char *buf, size_t len)
4855 {
4856 	unsigned int result;
4857 
4858 	rcu_read_lock();
4859 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4860 				clp->cl_ipaddr,
4861 				rpc_peeraddr2str(clp->cl_rpcclient,
4862 							RPC_DISPLAY_ADDR),
4863 				rpc_peeraddr2str(clp->cl_rpcclient,
4864 							RPC_DISPLAY_PROTO));
4865 	rcu_read_unlock();
4866 	return result;
4867 }
4868 
4869 static unsigned int
4870 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4871 				char *buf, size_t len)
4872 {
4873 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4874 
4875 	if (nfs4_client_id_uniquifier[0] != '\0')
4876 		return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4877 				clp->rpc_ops->version,
4878 				clp->cl_minorversion,
4879 				nfs4_client_id_uniquifier,
4880 				nodename);
4881 	return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4882 				clp->rpc_ops->version, clp->cl_minorversion,
4883 				nodename);
4884 }
4885 
4886 /**
4887  * nfs4_proc_setclientid - Negotiate client ID
4888  * @clp: state data structure
4889  * @program: RPC program for NFSv4 callback service
4890  * @port: IP port number for NFS4 callback service
4891  * @cred: RPC credential to use for this call
4892  * @res: where to place the result
4893  *
4894  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4895  */
4896 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4897 		unsigned short port, struct rpc_cred *cred,
4898 		struct nfs4_setclientid_res *res)
4899 {
4900 	nfs4_verifier sc_verifier;
4901 	struct nfs4_setclientid setclientid = {
4902 		.sc_verifier = &sc_verifier,
4903 		.sc_prog = program,
4904 		.sc_cb_ident = clp->cl_cb_ident,
4905 	};
4906 	struct rpc_message msg = {
4907 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4908 		.rpc_argp = &setclientid,
4909 		.rpc_resp = res,
4910 		.rpc_cred = cred,
4911 	};
4912 	int status;
4913 
4914 	/* nfs_client_id4 */
4915 	nfs4_init_boot_verifier(clp, &sc_verifier);
4916 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4917 		setclientid.sc_name_len =
4918 				nfs4_init_uniform_client_string(clp,
4919 						setclientid.sc_name,
4920 						sizeof(setclientid.sc_name));
4921 	else
4922 		setclientid.sc_name_len =
4923 				nfs4_init_nonuniform_client_string(clp,
4924 						setclientid.sc_name,
4925 						sizeof(setclientid.sc_name));
4926 	/* cb_client4 */
4927 	rcu_read_lock();
4928 	setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4929 				sizeof(setclientid.sc_netid), "%s",
4930 				rpc_peeraddr2str(clp->cl_rpcclient,
4931 							RPC_DISPLAY_NETID));
4932 	rcu_read_unlock();
4933 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4934 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4935 				clp->cl_ipaddr, port >> 8, port & 255);
4936 
4937 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4938 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
4939 		setclientid.sc_name_len, setclientid.sc_name);
4940 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4941 	trace_nfs4_setclientid(clp, status);
4942 	dprintk("NFS reply setclientid: %d\n", status);
4943 	return status;
4944 }
4945 
4946 /**
4947  * nfs4_proc_setclientid_confirm - Confirm client ID
4948  * @clp: state data structure
4949  * @res: result of a previous SETCLIENTID
4950  * @cred: RPC credential to use for this call
4951  *
4952  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4953  */
4954 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4955 		struct nfs4_setclientid_res *arg,
4956 		struct rpc_cred *cred)
4957 {
4958 	struct rpc_message msg = {
4959 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4960 		.rpc_argp = arg,
4961 		.rpc_cred = cred,
4962 	};
4963 	int status;
4964 
4965 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4966 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
4967 		clp->cl_clientid);
4968 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4969 	trace_nfs4_setclientid_confirm(clp, status);
4970 	dprintk("NFS reply setclientid_confirm: %d\n", status);
4971 	return status;
4972 }
4973 
4974 struct nfs4_delegreturndata {
4975 	struct nfs4_delegreturnargs args;
4976 	struct nfs4_delegreturnres res;
4977 	struct nfs_fh fh;
4978 	nfs4_stateid stateid;
4979 	unsigned long timestamp;
4980 	struct nfs_fattr fattr;
4981 	int rpc_status;
4982 };
4983 
4984 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4985 {
4986 	struct nfs4_delegreturndata *data = calldata;
4987 
4988 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4989 		return;
4990 
4991 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
4992 	switch (task->tk_status) {
4993 	case 0:
4994 		renew_lease(data->res.server, data->timestamp);
4995 		break;
4996 	case -NFS4ERR_ADMIN_REVOKED:
4997 	case -NFS4ERR_DELEG_REVOKED:
4998 	case -NFS4ERR_BAD_STATEID:
4999 	case -NFS4ERR_OLD_STATEID:
5000 	case -NFS4ERR_STALE_STATEID:
5001 	case -NFS4ERR_EXPIRED:
5002 		task->tk_status = 0;
5003 		break;
5004 	default:
5005 		if (nfs4_async_handle_error(task, data->res.server, NULL) ==
5006 				-EAGAIN) {
5007 			rpc_restart_call_prepare(task);
5008 			return;
5009 		}
5010 	}
5011 	data->rpc_status = task->tk_status;
5012 }
5013 
5014 static void nfs4_delegreturn_release(void *calldata)
5015 {
5016 	kfree(calldata);
5017 }
5018 
5019 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5020 {
5021 	struct nfs4_delegreturndata *d_data;
5022 
5023 	d_data = (struct nfs4_delegreturndata *)data;
5024 
5025 	nfs4_setup_sequence(d_data->res.server,
5026 			&d_data->args.seq_args,
5027 			&d_data->res.seq_res,
5028 			task);
5029 }
5030 
5031 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5032 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5033 	.rpc_call_done = nfs4_delegreturn_done,
5034 	.rpc_release = nfs4_delegreturn_release,
5035 };
5036 
5037 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5038 {
5039 	struct nfs4_delegreturndata *data;
5040 	struct nfs_server *server = NFS_SERVER(inode);
5041 	struct rpc_task *task;
5042 	struct rpc_message msg = {
5043 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5044 		.rpc_cred = cred,
5045 	};
5046 	struct rpc_task_setup task_setup_data = {
5047 		.rpc_client = server->client,
5048 		.rpc_message = &msg,
5049 		.callback_ops = &nfs4_delegreturn_ops,
5050 		.flags = RPC_TASK_ASYNC,
5051 	};
5052 	int status = 0;
5053 
5054 	data = kzalloc(sizeof(*data), GFP_NOFS);
5055 	if (data == NULL)
5056 		return -ENOMEM;
5057 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5058 	data->args.fhandle = &data->fh;
5059 	data->args.stateid = &data->stateid;
5060 	data->args.bitmask = server->cache_consistency_bitmask;
5061 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5062 	nfs4_stateid_copy(&data->stateid, stateid);
5063 	data->res.fattr = &data->fattr;
5064 	data->res.server = server;
5065 	nfs_fattr_init(data->res.fattr);
5066 	data->timestamp = jiffies;
5067 	data->rpc_status = 0;
5068 
5069 	task_setup_data.callback_data = data;
5070 	msg.rpc_argp = &data->args;
5071 	msg.rpc_resp = &data->res;
5072 	task = rpc_run_task(&task_setup_data);
5073 	if (IS_ERR(task))
5074 		return PTR_ERR(task);
5075 	if (!issync)
5076 		goto out;
5077 	status = nfs4_wait_for_completion_rpc_task(task);
5078 	if (status != 0)
5079 		goto out;
5080 	status = data->rpc_status;
5081 	if (status == 0)
5082 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5083 	else
5084 		nfs_refresh_inode(inode, &data->fattr);
5085 out:
5086 	rpc_put_task(task);
5087 	return status;
5088 }
5089 
5090 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5091 {
5092 	struct nfs_server *server = NFS_SERVER(inode);
5093 	struct nfs4_exception exception = { };
5094 	int err;
5095 	do {
5096 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5097 		trace_nfs4_delegreturn(inode, err);
5098 		switch (err) {
5099 			case -NFS4ERR_STALE_STATEID:
5100 			case -NFS4ERR_EXPIRED:
5101 			case 0:
5102 				return 0;
5103 		}
5104 		err = nfs4_handle_exception(server, err, &exception);
5105 	} while (exception.retry);
5106 	return err;
5107 }
5108 
5109 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5110 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5111 
5112 /*
5113  * sleep, with exponential backoff, and retry the LOCK operation.
5114  */
5115 static unsigned long
5116 nfs4_set_lock_task_retry(unsigned long timeout)
5117 {
5118 	freezable_schedule_timeout_killable_unsafe(timeout);
5119 	timeout <<= 1;
5120 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5121 		return NFS4_LOCK_MAXTIMEOUT;
5122 	return timeout;
5123 }
5124 
5125 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5126 {
5127 	struct inode *inode = state->inode;
5128 	struct nfs_server *server = NFS_SERVER(inode);
5129 	struct nfs_client *clp = server->nfs_client;
5130 	struct nfs_lockt_args arg = {
5131 		.fh = NFS_FH(inode),
5132 		.fl = request,
5133 	};
5134 	struct nfs_lockt_res res = {
5135 		.denied = request,
5136 	};
5137 	struct rpc_message msg = {
5138 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5139 		.rpc_argp       = &arg,
5140 		.rpc_resp       = &res,
5141 		.rpc_cred	= state->owner->so_cred,
5142 	};
5143 	struct nfs4_lock_state *lsp;
5144 	int status;
5145 
5146 	arg.lock_owner.clientid = clp->cl_clientid;
5147 	status = nfs4_set_lock_state(state, request);
5148 	if (status != 0)
5149 		goto out;
5150 	lsp = request->fl_u.nfs4_fl.owner;
5151 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5152 	arg.lock_owner.s_dev = server->s_dev;
5153 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5154 	switch (status) {
5155 		case 0:
5156 			request->fl_type = F_UNLCK;
5157 			break;
5158 		case -NFS4ERR_DENIED:
5159 			status = 0;
5160 	}
5161 	request->fl_ops->fl_release_private(request);
5162 	request->fl_ops = NULL;
5163 out:
5164 	return status;
5165 }
5166 
5167 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5168 {
5169 	struct nfs4_exception exception = { };
5170 	int err;
5171 
5172 	do {
5173 		err = _nfs4_proc_getlk(state, cmd, request);
5174 		trace_nfs4_get_lock(request, state, cmd, err);
5175 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5176 				&exception);
5177 	} while (exception.retry);
5178 	return err;
5179 }
5180 
5181 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5182 {
5183 	int res = 0;
5184 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5185 		case FL_POSIX:
5186 			res = posix_lock_file_wait(file, fl);
5187 			break;
5188 		case FL_FLOCK:
5189 			res = flock_lock_file_wait(file, fl);
5190 			break;
5191 		default:
5192 			BUG();
5193 	}
5194 	return res;
5195 }
5196 
5197 struct nfs4_unlockdata {
5198 	struct nfs_locku_args arg;
5199 	struct nfs_locku_res res;
5200 	struct nfs4_lock_state *lsp;
5201 	struct nfs_open_context *ctx;
5202 	struct file_lock fl;
5203 	const struct nfs_server *server;
5204 	unsigned long timestamp;
5205 };
5206 
5207 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5208 		struct nfs_open_context *ctx,
5209 		struct nfs4_lock_state *lsp,
5210 		struct nfs_seqid *seqid)
5211 {
5212 	struct nfs4_unlockdata *p;
5213 	struct inode *inode = lsp->ls_state->inode;
5214 
5215 	p = kzalloc(sizeof(*p), GFP_NOFS);
5216 	if (p == NULL)
5217 		return NULL;
5218 	p->arg.fh = NFS_FH(inode);
5219 	p->arg.fl = &p->fl;
5220 	p->arg.seqid = seqid;
5221 	p->res.seqid = seqid;
5222 	p->arg.stateid = &lsp->ls_stateid;
5223 	p->lsp = lsp;
5224 	atomic_inc(&lsp->ls_count);
5225 	/* Ensure we don't close file until we're done freeing locks! */
5226 	p->ctx = get_nfs_open_context(ctx);
5227 	memcpy(&p->fl, fl, sizeof(p->fl));
5228 	p->server = NFS_SERVER(inode);
5229 	return p;
5230 }
5231 
5232 static void nfs4_locku_release_calldata(void *data)
5233 {
5234 	struct nfs4_unlockdata *calldata = data;
5235 	nfs_free_seqid(calldata->arg.seqid);
5236 	nfs4_put_lock_state(calldata->lsp);
5237 	put_nfs_open_context(calldata->ctx);
5238 	kfree(calldata);
5239 }
5240 
5241 static void nfs4_locku_done(struct rpc_task *task, void *data)
5242 {
5243 	struct nfs4_unlockdata *calldata = data;
5244 
5245 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5246 		return;
5247 	switch (task->tk_status) {
5248 		case 0:
5249 			nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5250 					&calldata->res.stateid);
5251 			renew_lease(calldata->server, calldata->timestamp);
5252 			break;
5253 		case -NFS4ERR_BAD_STATEID:
5254 		case -NFS4ERR_OLD_STATEID:
5255 		case -NFS4ERR_STALE_STATEID:
5256 		case -NFS4ERR_EXPIRED:
5257 			break;
5258 		default:
5259 			if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5260 				rpc_restart_call_prepare(task);
5261 	}
5262 	nfs_release_seqid(calldata->arg.seqid);
5263 }
5264 
5265 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5266 {
5267 	struct nfs4_unlockdata *calldata = data;
5268 
5269 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5270 		goto out_wait;
5271 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5272 		/* Note: exit _without_ running nfs4_locku_done */
5273 		goto out_no_action;
5274 	}
5275 	calldata->timestamp = jiffies;
5276 	if (nfs4_setup_sequence(calldata->server,
5277 				&calldata->arg.seq_args,
5278 				&calldata->res.seq_res,
5279 				task) != 0)
5280 		nfs_release_seqid(calldata->arg.seqid);
5281 	return;
5282 out_no_action:
5283 	task->tk_action = NULL;
5284 out_wait:
5285 	nfs4_sequence_done(task, &calldata->res.seq_res);
5286 }
5287 
5288 static const struct rpc_call_ops nfs4_locku_ops = {
5289 	.rpc_call_prepare = nfs4_locku_prepare,
5290 	.rpc_call_done = nfs4_locku_done,
5291 	.rpc_release = nfs4_locku_release_calldata,
5292 };
5293 
5294 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5295 		struct nfs_open_context *ctx,
5296 		struct nfs4_lock_state *lsp,
5297 		struct nfs_seqid *seqid)
5298 {
5299 	struct nfs4_unlockdata *data;
5300 	struct rpc_message msg = {
5301 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5302 		.rpc_cred = ctx->cred,
5303 	};
5304 	struct rpc_task_setup task_setup_data = {
5305 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5306 		.rpc_message = &msg,
5307 		.callback_ops = &nfs4_locku_ops,
5308 		.workqueue = nfsiod_workqueue,
5309 		.flags = RPC_TASK_ASYNC,
5310 	};
5311 
5312 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5313 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5314 
5315 	/* Ensure this is an unlock - when canceling a lock, the
5316 	 * canceled lock is passed in, and it won't be an unlock.
5317 	 */
5318 	fl->fl_type = F_UNLCK;
5319 
5320 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5321 	if (data == NULL) {
5322 		nfs_free_seqid(seqid);
5323 		return ERR_PTR(-ENOMEM);
5324 	}
5325 
5326 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5327 	msg.rpc_argp = &data->arg;
5328 	msg.rpc_resp = &data->res;
5329 	task_setup_data.callback_data = data;
5330 	return rpc_run_task(&task_setup_data);
5331 }
5332 
5333 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5334 {
5335 	struct inode *inode = state->inode;
5336 	struct nfs4_state_owner *sp = state->owner;
5337 	struct nfs_inode *nfsi = NFS_I(inode);
5338 	struct nfs_seqid *seqid;
5339 	struct nfs4_lock_state *lsp;
5340 	struct rpc_task *task;
5341 	int status = 0;
5342 	unsigned char fl_flags = request->fl_flags;
5343 
5344 	status = nfs4_set_lock_state(state, request);
5345 	/* Unlock _before_ we do the RPC call */
5346 	request->fl_flags |= FL_EXISTS;
5347 	/* Exclude nfs_delegation_claim_locks() */
5348 	mutex_lock(&sp->so_delegreturn_mutex);
5349 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5350 	down_read(&nfsi->rwsem);
5351 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5352 		up_read(&nfsi->rwsem);
5353 		mutex_unlock(&sp->so_delegreturn_mutex);
5354 		goto out;
5355 	}
5356 	up_read(&nfsi->rwsem);
5357 	mutex_unlock(&sp->so_delegreturn_mutex);
5358 	if (status != 0)
5359 		goto out;
5360 	/* Is this a delegated lock? */
5361 	lsp = request->fl_u.nfs4_fl.owner;
5362 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5363 		goto out;
5364 	seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5365 	status = -ENOMEM;
5366 	if (seqid == NULL)
5367 		goto out;
5368 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5369 	status = PTR_ERR(task);
5370 	if (IS_ERR(task))
5371 		goto out;
5372 	status = nfs4_wait_for_completion_rpc_task(task);
5373 	rpc_put_task(task);
5374 out:
5375 	request->fl_flags = fl_flags;
5376 	trace_nfs4_unlock(request, state, F_SETLK, status);
5377 	return status;
5378 }
5379 
5380 struct nfs4_lockdata {
5381 	struct nfs_lock_args arg;
5382 	struct nfs_lock_res res;
5383 	struct nfs4_lock_state *lsp;
5384 	struct nfs_open_context *ctx;
5385 	struct file_lock fl;
5386 	unsigned long timestamp;
5387 	int rpc_status;
5388 	int cancelled;
5389 	struct nfs_server *server;
5390 };
5391 
5392 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5393 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5394 		gfp_t gfp_mask)
5395 {
5396 	struct nfs4_lockdata *p;
5397 	struct inode *inode = lsp->ls_state->inode;
5398 	struct nfs_server *server = NFS_SERVER(inode);
5399 
5400 	p = kzalloc(sizeof(*p), gfp_mask);
5401 	if (p == NULL)
5402 		return NULL;
5403 
5404 	p->arg.fh = NFS_FH(inode);
5405 	p->arg.fl = &p->fl;
5406 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5407 	if (p->arg.open_seqid == NULL)
5408 		goto out_free;
5409 	p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5410 	if (p->arg.lock_seqid == NULL)
5411 		goto out_free_seqid;
5412 	p->arg.lock_stateid = &lsp->ls_stateid;
5413 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5414 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5415 	p->arg.lock_owner.s_dev = server->s_dev;
5416 	p->res.lock_seqid = p->arg.lock_seqid;
5417 	p->lsp = lsp;
5418 	p->server = server;
5419 	atomic_inc(&lsp->ls_count);
5420 	p->ctx = get_nfs_open_context(ctx);
5421 	memcpy(&p->fl, fl, sizeof(p->fl));
5422 	return p;
5423 out_free_seqid:
5424 	nfs_free_seqid(p->arg.open_seqid);
5425 out_free:
5426 	kfree(p);
5427 	return NULL;
5428 }
5429 
5430 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5431 {
5432 	struct nfs4_lockdata *data = calldata;
5433 	struct nfs4_state *state = data->lsp->ls_state;
5434 
5435 	dprintk("%s: begin!\n", __func__);
5436 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5437 		goto out_wait;
5438 	/* Do we need to do an open_to_lock_owner? */
5439 	if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5440 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5441 			goto out_release_lock_seqid;
5442 		}
5443 		data->arg.open_stateid = &state->open_stateid;
5444 		data->arg.new_lock_owner = 1;
5445 		data->res.open_seqid = data->arg.open_seqid;
5446 	} else
5447 		data->arg.new_lock_owner = 0;
5448 	if (!nfs4_valid_open_stateid(state)) {
5449 		data->rpc_status = -EBADF;
5450 		task->tk_action = NULL;
5451 		goto out_release_open_seqid;
5452 	}
5453 	data->timestamp = jiffies;
5454 	if (nfs4_setup_sequence(data->server,
5455 				&data->arg.seq_args,
5456 				&data->res.seq_res,
5457 				task) == 0)
5458 		return;
5459 out_release_open_seqid:
5460 	nfs_release_seqid(data->arg.open_seqid);
5461 out_release_lock_seqid:
5462 	nfs_release_seqid(data->arg.lock_seqid);
5463 out_wait:
5464 	nfs4_sequence_done(task, &data->res.seq_res);
5465 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5466 }
5467 
5468 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5469 {
5470 	struct nfs4_lockdata *data = calldata;
5471 
5472 	dprintk("%s: begin!\n", __func__);
5473 
5474 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5475 		return;
5476 
5477 	data->rpc_status = task->tk_status;
5478 	if (data->arg.new_lock_owner != 0) {
5479 		if (data->rpc_status == 0)
5480 			nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5481 		else
5482 			goto out;
5483 	}
5484 	if (data->rpc_status == 0) {
5485 		nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5486 		set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5487 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5488 	}
5489 out:
5490 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5491 }
5492 
5493 static void nfs4_lock_release(void *calldata)
5494 {
5495 	struct nfs4_lockdata *data = calldata;
5496 
5497 	dprintk("%s: begin!\n", __func__);
5498 	nfs_free_seqid(data->arg.open_seqid);
5499 	if (data->cancelled != 0) {
5500 		struct rpc_task *task;
5501 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5502 				data->arg.lock_seqid);
5503 		if (!IS_ERR(task))
5504 			rpc_put_task_async(task);
5505 		dprintk("%s: cancelling lock!\n", __func__);
5506 	} else
5507 		nfs_free_seqid(data->arg.lock_seqid);
5508 	nfs4_put_lock_state(data->lsp);
5509 	put_nfs_open_context(data->ctx);
5510 	kfree(data);
5511 	dprintk("%s: done!\n", __func__);
5512 }
5513 
5514 static const struct rpc_call_ops nfs4_lock_ops = {
5515 	.rpc_call_prepare = nfs4_lock_prepare,
5516 	.rpc_call_done = nfs4_lock_done,
5517 	.rpc_release = nfs4_lock_release,
5518 };
5519 
5520 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5521 {
5522 	switch (error) {
5523 	case -NFS4ERR_ADMIN_REVOKED:
5524 	case -NFS4ERR_BAD_STATEID:
5525 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5526 		if (new_lock_owner != 0 ||
5527 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5528 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5529 		break;
5530 	case -NFS4ERR_STALE_STATEID:
5531 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5532 	case -NFS4ERR_EXPIRED:
5533 		nfs4_schedule_lease_recovery(server->nfs_client);
5534 	};
5535 }
5536 
5537 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5538 {
5539 	struct nfs4_lockdata *data;
5540 	struct rpc_task *task;
5541 	struct rpc_message msg = {
5542 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5543 		.rpc_cred = state->owner->so_cred,
5544 	};
5545 	struct rpc_task_setup task_setup_data = {
5546 		.rpc_client = NFS_CLIENT(state->inode),
5547 		.rpc_message = &msg,
5548 		.callback_ops = &nfs4_lock_ops,
5549 		.workqueue = nfsiod_workqueue,
5550 		.flags = RPC_TASK_ASYNC,
5551 	};
5552 	int ret;
5553 
5554 	dprintk("%s: begin!\n", __func__);
5555 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5556 			fl->fl_u.nfs4_fl.owner,
5557 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5558 	if (data == NULL)
5559 		return -ENOMEM;
5560 	if (IS_SETLKW(cmd))
5561 		data->arg.block = 1;
5562 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5563 	msg.rpc_argp = &data->arg;
5564 	msg.rpc_resp = &data->res;
5565 	task_setup_data.callback_data = data;
5566 	if (recovery_type > NFS_LOCK_NEW) {
5567 		if (recovery_type == NFS_LOCK_RECLAIM)
5568 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5569 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5570 	}
5571 	task = rpc_run_task(&task_setup_data);
5572 	if (IS_ERR(task))
5573 		return PTR_ERR(task);
5574 	ret = nfs4_wait_for_completion_rpc_task(task);
5575 	if (ret == 0) {
5576 		ret = data->rpc_status;
5577 		if (ret)
5578 			nfs4_handle_setlk_error(data->server, data->lsp,
5579 					data->arg.new_lock_owner, ret);
5580 	} else
5581 		data->cancelled = 1;
5582 	rpc_put_task(task);
5583 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5584 	return ret;
5585 }
5586 
5587 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5588 {
5589 	struct nfs_server *server = NFS_SERVER(state->inode);
5590 	struct nfs4_exception exception = {
5591 		.inode = state->inode,
5592 	};
5593 	int err;
5594 
5595 	do {
5596 		/* Cache the lock if possible... */
5597 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5598 			return 0;
5599 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5600 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5601 		if (err != -NFS4ERR_DELAY)
5602 			break;
5603 		nfs4_handle_exception(server, err, &exception);
5604 	} while (exception.retry);
5605 	return err;
5606 }
5607 
5608 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5609 {
5610 	struct nfs_server *server = NFS_SERVER(state->inode);
5611 	struct nfs4_exception exception = {
5612 		.inode = state->inode,
5613 	};
5614 	int err;
5615 
5616 	err = nfs4_set_lock_state(state, request);
5617 	if (err != 0)
5618 		return err;
5619 	if (!recover_lost_locks) {
5620 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5621 		return 0;
5622 	}
5623 	do {
5624 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5625 			return 0;
5626 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5627 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5628 		switch (err) {
5629 		default:
5630 			goto out;
5631 		case -NFS4ERR_GRACE:
5632 		case -NFS4ERR_DELAY:
5633 			nfs4_handle_exception(server, err, &exception);
5634 			err = 0;
5635 		}
5636 	} while (exception.retry);
5637 out:
5638 	return err;
5639 }
5640 
5641 #if defined(CONFIG_NFS_V4_1)
5642 /**
5643  * nfs41_check_expired_locks - possibly free a lock stateid
5644  *
5645  * @state: NFSv4 state for an inode
5646  *
5647  * Returns NFS_OK if recovery for this stateid is now finished.
5648  * Otherwise a negative NFS4ERR value is returned.
5649  */
5650 static int nfs41_check_expired_locks(struct nfs4_state *state)
5651 {
5652 	int status, ret = -NFS4ERR_BAD_STATEID;
5653 	struct nfs4_lock_state *lsp;
5654 	struct nfs_server *server = NFS_SERVER(state->inode);
5655 
5656 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5657 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5658 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5659 
5660 			status = nfs41_test_stateid(server,
5661 					&lsp->ls_stateid,
5662 					cred);
5663 			trace_nfs4_test_lock_stateid(state, lsp, status);
5664 			if (status != NFS_OK) {
5665 				/* Free the stateid unless the server
5666 				 * informs us the stateid is unrecognized. */
5667 				if (status != -NFS4ERR_BAD_STATEID)
5668 					nfs41_free_stateid(server,
5669 							&lsp->ls_stateid,
5670 							cred);
5671 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5672 				ret = status;
5673 			}
5674 		}
5675 	};
5676 
5677 	return ret;
5678 }
5679 
5680 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5681 {
5682 	int status = NFS_OK;
5683 
5684 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5685 		status = nfs41_check_expired_locks(state);
5686 	if (status != NFS_OK)
5687 		status = nfs4_lock_expired(state, request);
5688 	return status;
5689 }
5690 #endif
5691 
5692 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5693 {
5694 	struct nfs4_state_owner *sp = state->owner;
5695 	struct nfs_inode *nfsi = NFS_I(state->inode);
5696 	unsigned char fl_flags = request->fl_flags;
5697 	unsigned int seq;
5698 	int status = -ENOLCK;
5699 
5700 	if ((fl_flags & FL_POSIX) &&
5701 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5702 		goto out;
5703 	/* Is this a delegated open? */
5704 	status = nfs4_set_lock_state(state, request);
5705 	if (status != 0)
5706 		goto out;
5707 	request->fl_flags |= FL_ACCESS;
5708 	status = do_vfs_lock(request->fl_file, request);
5709 	if (status < 0)
5710 		goto out;
5711 	down_read(&nfsi->rwsem);
5712 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5713 		/* Yes: cache locks! */
5714 		/* ...but avoid races with delegation recall... */
5715 		request->fl_flags = fl_flags & ~FL_SLEEP;
5716 		status = do_vfs_lock(request->fl_file, request);
5717 		goto out_unlock;
5718 	}
5719 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5720 	up_read(&nfsi->rwsem);
5721 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5722 	if (status != 0)
5723 		goto out;
5724 	down_read(&nfsi->rwsem);
5725 	if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5726 		status = -NFS4ERR_DELAY;
5727 		goto out_unlock;
5728 	}
5729 	/* Note: we always want to sleep here! */
5730 	request->fl_flags = fl_flags | FL_SLEEP;
5731 	if (do_vfs_lock(request->fl_file, request) < 0)
5732 		printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5733 			"manager!\n", __func__);
5734 out_unlock:
5735 	up_read(&nfsi->rwsem);
5736 out:
5737 	request->fl_flags = fl_flags;
5738 	return status;
5739 }
5740 
5741 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5742 {
5743 	struct nfs4_exception exception = {
5744 		.state = state,
5745 		.inode = state->inode,
5746 	};
5747 	int err;
5748 
5749 	do {
5750 		err = _nfs4_proc_setlk(state, cmd, request);
5751 		trace_nfs4_set_lock(request, state, cmd, err);
5752 		if (err == -NFS4ERR_DENIED)
5753 			err = -EAGAIN;
5754 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5755 				err, &exception);
5756 	} while (exception.retry);
5757 	return err;
5758 }
5759 
5760 static int
5761 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5762 {
5763 	struct nfs_open_context *ctx;
5764 	struct nfs4_state *state;
5765 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5766 	int status;
5767 
5768 	/* verify open state */
5769 	ctx = nfs_file_open_context(filp);
5770 	state = ctx->state;
5771 
5772 	if (request->fl_start < 0 || request->fl_end < 0)
5773 		return -EINVAL;
5774 
5775 	if (IS_GETLK(cmd)) {
5776 		if (state != NULL)
5777 			return nfs4_proc_getlk(state, F_GETLK, request);
5778 		return 0;
5779 	}
5780 
5781 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5782 		return -EINVAL;
5783 
5784 	if (request->fl_type == F_UNLCK) {
5785 		if (state != NULL)
5786 			return nfs4_proc_unlck(state, cmd, request);
5787 		return 0;
5788 	}
5789 
5790 	if (state == NULL)
5791 		return -ENOLCK;
5792 	/*
5793 	 * Don't rely on the VFS having checked the file open mode,
5794 	 * since it won't do this for flock() locks.
5795 	 */
5796 	switch (request->fl_type) {
5797 	case F_RDLCK:
5798 		if (!(filp->f_mode & FMODE_READ))
5799 			return -EBADF;
5800 		break;
5801 	case F_WRLCK:
5802 		if (!(filp->f_mode & FMODE_WRITE))
5803 			return -EBADF;
5804 	}
5805 
5806 	do {
5807 		status = nfs4_proc_setlk(state, cmd, request);
5808 		if ((status != -EAGAIN) || IS_SETLK(cmd))
5809 			break;
5810 		timeout = nfs4_set_lock_task_retry(timeout);
5811 		status = -ERESTARTSYS;
5812 		if (signalled())
5813 			break;
5814 	} while(status < 0);
5815 	return status;
5816 }
5817 
5818 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5819 {
5820 	struct nfs_server *server = NFS_SERVER(state->inode);
5821 	int err;
5822 
5823 	err = nfs4_set_lock_state(state, fl);
5824 	if (err != 0)
5825 		return err;
5826 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5827 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5828 }
5829 
5830 struct nfs_release_lockowner_data {
5831 	struct nfs4_lock_state *lsp;
5832 	struct nfs_server *server;
5833 	struct nfs_release_lockowner_args args;
5834 	struct nfs4_sequence_args seq_args;
5835 	struct nfs4_sequence_res seq_res;
5836 	unsigned long timestamp;
5837 };
5838 
5839 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5840 {
5841 	struct nfs_release_lockowner_data *data = calldata;
5842 	nfs40_setup_sequence(data->server,
5843 				&data->seq_args, &data->seq_res, task);
5844 	data->timestamp = jiffies;
5845 }
5846 
5847 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5848 {
5849 	struct nfs_release_lockowner_data *data = calldata;
5850 	struct nfs_server *server = data->server;
5851 
5852 	nfs40_sequence_done(task, &data->seq_res);
5853 
5854 	switch (task->tk_status) {
5855 	case 0:
5856 		renew_lease(server, data->timestamp);
5857 		break;
5858 	case -NFS4ERR_STALE_CLIENTID:
5859 	case -NFS4ERR_EXPIRED:
5860 	case -NFS4ERR_LEASE_MOVED:
5861 	case -NFS4ERR_DELAY:
5862 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN)
5863 			rpc_restart_call_prepare(task);
5864 	}
5865 }
5866 
5867 static void nfs4_release_lockowner_release(void *calldata)
5868 {
5869 	struct nfs_release_lockowner_data *data = calldata;
5870 	nfs4_free_lock_state(data->server, data->lsp);
5871 	kfree(calldata);
5872 }
5873 
5874 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5875 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
5876 	.rpc_call_done = nfs4_release_lockowner_done,
5877 	.rpc_release = nfs4_release_lockowner_release,
5878 };
5879 
5880 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5881 {
5882 	struct nfs_release_lockowner_data *data;
5883 	struct rpc_message msg = {
5884 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5885 	};
5886 
5887 	if (server->nfs_client->cl_mvops->minor_version != 0)
5888 		return -EINVAL;
5889 
5890 	data = kmalloc(sizeof(*data), GFP_NOFS);
5891 	if (!data)
5892 		return -ENOMEM;
5893 	nfs4_init_sequence(&data->seq_args, &data->seq_res, 0);
5894 	data->lsp = lsp;
5895 	data->server = server;
5896 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5897 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5898 	data->args.lock_owner.s_dev = server->s_dev;
5899 
5900 	msg.rpc_argp = &data->args;
5901 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5902 	return 0;
5903 }
5904 
5905 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5906 
5907 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5908 				   const void *buf, size_t buflen,
5909 				   int flags, int type)
5910 {
5911 	if (strcmp(key, "") != 0)
5912 		return -EINVAL;
5913 
5914 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5915 }
5916 
5917 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5918 				   void *buf, size_t buflen, int type)
5919 {
5920 	if (strcmp(key, "") != 0)
5921 		return -EINVAL;
5922 
5923 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5924 }
5925 
5926 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5927 				       size_t list_len, const char *name,
5928 				       size_t name_len, int type)
5929 {
5930 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5931 
5932 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5933 		return 0;
5934 
5935 	if (list && len <= list_len)
5936 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5937 	return len;
5938 }
5939 
5940 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5941 static inline int nfs4_server_supports_labels(struct nfs_server *server)
5942 {
5943 	return server->caps & NFS_CAP_SECURITY_LABEL;
5944 }
5945 
5946 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
5947 				   const void *buf, size_t buflen,
5948 				   int flags, int type)
5949 {
5950 	if (security_ismaclabel(key))
5951 		return nfs4_set_security_label(dentry, buf, buflen);
5952 
5953 	return -EOPNOTSUPP;
5954 }
5955 
5956 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
5957 				   void *buf, size_t buflen, int type)
5958 {
5959 	if (security_ismaclabel(key))
5960 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
5961 	return -EOPNOTSUPP;
5962 }
5963 
5964 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
5965 				       size_t list_len, const char *name,
5966 				       size_t name_len, int type)
5967 {
5968 	size_t len = 0;
5969 
5970 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
5971 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
5972 		if (list && len <= list_len)
5973 			security_inode_listsecurity(dentry->d_inode, list, len);
5974 	}
5975 	return len;
5976 }
5977 
5978 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
5979 	.prefix = XATTR_SECURITY_PREFIX,
5980 	.list	= nfs4_xattr_list_nfs4_label,
5981 	.get	= nfs4_xattr_get_nfs4_label,
5982 	.set	= nfs4_xattr_set_nfs4_label,
5983 };
5984 #endif
5985 
5986 
5987 /*
5988  * nfs_fhget will use either the mounted_on_fileid or the fileid
5989  */
5990 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5991 {
5992 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5993 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5994 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5995 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5996 		return;
5997 
5998 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5999 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6000 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6001 	fattr->nlink = 2;
6002 }
6003 
6004 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6005 				   const struct qstr *name,
6006 				   struct nfs4_fs_locations *fs_locations,
6007 				   struct page *page)
6008 {
6009 	struct nfs_server *server = NFS_SERVER(dir);
6010 	u32 bitmask[3] = {
6011 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6012 	};
6013 	struct nfs4_fs_locations_arg args = {
6014 		.dir_fh = NFS_FH(dir),
6015 		.name = name,
6016 		.page = page,
6017 		.bitmask = bitmask,
6018 	};
6019 	struct nfs4_fs_locations_res res = {
6020 		.fs_locations = fs_locations,
6021 	};
6022 	struct rpc_message msg = {
6023 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6024 		.rpc_argp = &args,
6025 		.rpc_resp = &res,
6026 	};
6027 	int status;
6028 
6029 	dprintk("%s: start\n", __func__);
6030 
6031 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6032 	 * is not supported */
6033 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6034 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6035 	else
6036 		bitmask[0] |= FATTR4_WORD0_FILEID;
6037 
6038 	nfs_fattr_init(&fs_locations->fattr);
6039 	fs_locations->server = server;
6040 	fs_locations->nlocations = 0;
6041 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6042 	dprintk("%s: returned status = %d\n", __func__, status);
6043 	return status;
6044 }
6045 
6046 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6047 			   const struct qstr *name,
6048 			   struct nfs4_fs_locations *fs_locations,
6049 			   struct page *page)
6050 {
6051 	struct nfs4_exception exception = { };
6052 	int err;
6053 	do {
6054 		err = _nfs4_proc_fs_locations(client, dir, name,
6055 				fs_locations, page);
6056 		trace_nfs4_get_fs_locations(dir, name, err);
6057 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6058 				&exception);
6059 	} while (exception.retry);
6060 	return err;
6061 }
6062 
6063 /*
6064  * This operation also signals the server that this client is
6065  * performing migration recovery.  The server can stop returning
6066  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6067  * appended to this compound to identify the client ID which is
6068  * performing recovery.
6069  */
6070 static int _nfs40_proc_get_locations(struct inode *inode,
6071 				     struct nfs4_fs_locations *locations,
6072 				     struct page *page, struct rpc_cred *cred)
6073 {
6074 	struct nfs_server *server = NFS_SERVER(inode);
6075 	struct rpc_clnt *clnt = server->client;
6076 	u32 bitmask[2] = {
6077 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6078 	};
6079 	struct nfs4_fs_locations_arg args = {
6080 		.clientid	= server->nfs_client->cl_clientid,
6081 		.fh		= NFS_FH(inode),
6082 		.page		= page,
6083 		.bitmask	= bitmask,
6084 		.migration	= 1,		/* skip LOOKUP */
6085 		.renew		= 1,		/* append RENEW */
6086 	};
6087 	struct nfs4_fs_locations_res res = {
6088 		.fs_locations	= locations,
6089 		.migration	= 1,
6090 		.renew		= 1,
6091 	};
6092 	struct rpc_message msg = {
6093 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6094 		.rpc_argp	= &args,
6095 		.rpc_resp	= &res,
6096 		.rpc_cred	= cred,
6097 	};
6098 	unsigned long now = jiffies;
6099 	int status;
6100 
6101 	nfs_fattr_init(&locations->fattr);
6102 	locations->server = server;
6103 	locations->nlocations = 0;
6104 
6105 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6106 	nfs4_set_sequence_privileged(&args.seq_args);
6107 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6108 					&args.seq_args, &res.seq_res);
6109 	if (status)
6110 		return status;
6111 
6112 	renew_lease(server, now);
6113 	return 0;
6114 }
6115 
6116 #ifdef CONFIG_NFS_V4_1
6117 
6118 /*
6119  * This operation also signals the server that this client is
6120  * performing migration recovery.  The server can stop asserting
6121  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6122  * performing this operation is identified in the SEQUENCE
6123  * operation in this compound.
6124  *
6125  * When the client supports GETATTR(fs_locations_info), it can
6126  * be plumbed in here.
6127  */
6128 static int _nfs41_proc_get_locations(struct inode *inode,
6129 				     struct nfs4_fs_locations *locations,
6130 				     struct page *page, struct rpc_cred *cred)
6131 {
6132 	struct nfs_server *server = NFS_SERVER(inode);
6133 	struct rpc_clnt *clnt = server->client;
6134 	u32 bitmask[2] = {
6135 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6136 	};
6137 	struct nfs4_fs_locations_arg args = {
6138 		.fh		= NFS_FH(inode),
6139 		.page		= page,
6140 		.bitmask	= bitmask,
6141 		.migration	= 1,		/* skip LOOKUP */
6142 	};
6143 	struct nfs4_fs_locations_res res = {
6144 		.fs_locations	= locations,
6145 		.migration	= 1,
6146 	};
6147 	struct rpc_message msg = {
6148 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6149 		.rpc_argp	= &args,
6150 		.rpc_resp	= &res,
6151 		.rpc_cred	= cred,
6152 	};
6153 	int status;
6154 
6155 	nfs_fattr_init(&locations->fattr);
6156 	locations->server = server;
6157 	locations->nlocations = 0;
6158 
6159 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6160 	nfs4_set_sequence_privileged(&args.seq_args);
6161 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6162 					&args.seq_args, &res.seq_res);
6163 	if (status == NFS4_OK &&
6164 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6165 		status = -NFS4ERR_LEASE_MOVED;
6166 	return status;
6167 }
6168 
6169 #endif	/* CONFIG_NFS_V4_1 */
6170 
6171 /**
6172  * nfs4_proc_get_locations - discover locations for a migrated FSID
6173  * @inode: inode on FSID that is migrating
6174  * @locations: result of query
6175  * @page: buffer
6176  * @cred: credential to use for this operation
6177  *
6178  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6179  * operation failed, or a negative errno if a local error occurred.
6180  *
6181  * On success, "locations" is filled in, but if the server has
6182  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6183  * asserted.
6184  *
6185  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6186  * from this client that require migration recovery.
6187  */
6188 int nfs4_proc_get_locations(struct inode *inode,
6189 			    struct nfs4_fs_locations *locations,
6190 			    struct page *page, struct rpc_cred *cred)
6191 {
6192 	struct nfs_server *server = NFS_SERVER(inode);
6193 	struct nfs_client *clp = server->nfs_client;
6194 	const struct nfs4_mig_recovery_ops *ops =
6195 					clp->cl_mvops->mig_recovery_ops;
6196 	struct nfs4_exception exception = { };
6197 	int status;
6198 
6199 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6200 		(unsigned long long)server->fsid.major,
6201 		(unsigned long long)server->fsid.minor,
6202 		clp->cl_hostname);
6203 	nfs_display_fhandle(NFS_FH(inode), __func__);
6204 
6205 	do {
6206 		status = ops->get_locations(inode, locations, page, cred);
6207 		if (status != -NFS4ERR_DELAY)
6208 			break;
6209 		nfs4_handle_exception(server, status, &exception);
6210 	} while (exception.retry);
6211 	return status;
6212 }
6213 
6214 /*
6215  * This operation also signals the server that this client is
6216  * performing "lease moved" recovery.  The server can stop
6217  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6218  * is appended to this compound to identify the client ID which is
6219  * performing recovery.
6220  */
6221 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6222 {
6223 	struct nfs_server *server = NFS_SERVER(inode);
6224 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6225 	struct rpc_clnt *clnt = server->client;
6226 	struct nfs4_fsid_present_arg args = {
6227 		.fh		= NFS_FH(inode),
6228 		.clientid	= clp->cl_clientid,
6229 		.renew		= 1,		/* append RENEW */
6230 	};
6231 	struct nfs4_fsid_present_res res = {
6232 		.renew		= 1,
6233 	};
6234 	struct rpc_message msg = {
6235 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6236 		.rpc_argp	= &args,
6237 		.rpc_resp	= &res,
6238 		.rpc_cred	= cred,
6239 	};
6240 	unsigned long now = jiffies;
6241 	int status;
6242 
6243 	res.fh = nfs_alloc_fhandle();
6244 	if (res.fh == NULL)
6245 		return -ENOMEM;
6246 
6247 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6248 	nfs4_set_sequence_privileged(&args.seq_args);
6249 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6250 						&args.seq_args, &res.seq_res);
6251 	nfs_free_fhandle(res.fh);
6252 	if (status)
6253 		return status;
6254 
6255 	do_renew_lease(clp, now);
6256 	return 0;
6257 }
6258 
6259 #ifdef CONFIG_NFS_V4_1
6260 
6261 /*
6262  * This operation also signals the server that this client is
6263  * performing "lease moved" recovery.  The server can stop asserting
6264  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6265  * this operation is identified in the SEQUENCE operation in this
6266  * compound.
6267  */
6268 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6269 {
6270 	struct nfs_server *server = NFS_SERVER(inode);
6271 	struct rpc_clnt *clnt = server->client;
6272 	struct nfs4_fsid_present_arg args = {
6273 		.fh		= NFS_FH(inode),
6274 	};
6275 	struct nfs4_fsid_present_res res = {
6276 	};
6277 	struct rpc_message msg = {
6278 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6279 		.rpc_argp	= &args,
6280 		.rpc_resp	= &res,
6281 		.rpc_cred	= cred,
6282 	};
6283 	int status;
6284 
6285 	res.fh = nfs_alloc_fhandle();
6286 	if (res.fh == NULL)
6287 		return -ENOMEM;
6288 
6289 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6290 	nfs4_set_sequence_privileged(&args.seq_args);
6291 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6292 						&args.seq_args, &res.seq_res);
6293 	nfs_free_fhandle(res.fh);
6294 	if (status == NFS4_OK &&
6295 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6296 		status = -NFS4ERR_LEASE_MOVED;
6297 	return status;
6298 }
6299 
6300 #endif	/* CONFIG_NFS_V4_1 */
6301 
6302 /**
6303  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6304  * @inode: inode on FSID to check
6305  * @cred: credential to use for this operation
6306  *
6307  * Server indicates whether the FSID is present, moved, or not
6308  * recognized.  This operation is necessary to clear a LEASE_MOVED
6309  * condition for this client ID.
6310  *
6311  * Returns NFS4_OK if the FSID is present on this server,
6312  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6313  *  NFS4ERR code if some error occurred on the server, or a
6314  *  negative errno if a local failure occurred.
6315  */
6316 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6317 {
6318 	struct nfs_server *server = NFS_SERVER(inode);
6319 	struct nfs_client *clp = server->nfs_client;
6320 	const struct nfs4_mig_recovery_ops *ops =
6321 					clp->cl_mvops->mig_recovery_ops;
6322 	struct nfs4_exception exception = { };
6323 	int status;
6324 
6325 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6326 		(unsigned long long)server->fsid.major,
6327 		(unsigned long long)server->fsid.minor,
6328 		clp->cl_hostname);
6329 	nfs_display_fhandle(NFS_FH(inode), __func__);
6330 
6331 	do {
6332 		status = ops->fsid_present(inode, cred);
6333 		if (status != -NFS4ERR_DELAY)
6334 			break;
6335 		nfs4_handle_exception(server, status, &exception);
6336 	} while (exception.retry);
6337 	return status;
6338 }
6339 
6340 /**
6341  * If 'use_integrity' is true and the state managment nfs_client
6342  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6343  * and the machine credential as per RFC3530bis and RFC5661 Security
6344  * Considerations sections. Otherwise, just use the user cred with the
6345  * filesystem's rpc_client.
6346  */
6347 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6348 {
6349 	int status;
6350 	struct nfs4_secinfo_arg args = {
6351 		.dir_fh = NFS_FH(dir),
6352 		.name   = name,
6353 	};
6354 	struct nfs4_secinfo_res res = {
6355 		.flavors     = flavors,
6356 	};
6357 	struct rpc_message msg = {
6358 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6359 		.rpc_argp = &args,
6360 		.rpc_resp = &res,
6361 	};
6362 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6363 	struct rpc_cred *cred = NULL;
6364 
6365 	if (use_integrity) {
6366 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6367 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6368 		msg.rpc_cred = cred;
6369 	}
6370 
6371 	dprintk("NFS call  secinfo %s\n", name->name);
6372 
6373 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6374 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6375 
6376 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6377 				&res.seq_res, 0);
6378 	dprintk("NFS reply  secinfo: %d\n", status);
6379 
6380 	if (cred)
6381 		put_rpccred(cred);
6382 
6383 	return status;
6384 }
6385 
6386 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6387 		      struct nfs4_secinfo_flavors *flavors)
6388 {
6389 	struct nfs4_exception exception = { };
6390 	int err;
6391 	do {
6392 		err = -NFS4ERR_WRONGSEC;
6393 
6394 		/* try to use integrity protection with machine cred */
6395 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6396 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6397 
6398 		/*
6399 		 * if unable to use integrity protection, or SECINFO with
6400 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6401 		 * disallowed by spec, but exists in deployed servers) use
6402 		 * the current filesystem's rpc_client and the user cred.
6403 		 */
6404 		if (err == -NFS4ERR_WRONGSEC)
6405 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6406 
6407 		trace_nfs4_secinfo(dir, name, err);
6408 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6409 				&exception);
6410 	} while (exception.retry);
6411 	return err;
6412 }
6413 
6414 #ifdef CONFIG_NFS_V4_1
6415 /*
6416  * Check the exchange flags returned by the server for invalid flags, having
6417  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6418  * DS flags set.
6419  */
6420 static int nfs4_check_cl_exchange_flags(u32 flags)
6421 {
6422 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6423 		goto out_inval;
6424 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6425 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6426 		goto out_inval;
6427 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6428 		goto out_inval;
6429 	return NFS_OK;
6430 out_inval:
6431 	return -NFS4ERR_INVAL;
6432 }
6433 
6434 static bool
6435 nfs41_same_server_scope(struct nfs41_server_scope *a,
6436 			struct nfs41_server_scope *b)
6437 {
6438 	if (a->server_scope_sz == b->server_scope_sz &&
6439 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6440 		return true;
6441 
6442 	return false;
6443 }
6444 
6445 /*
6446  * nfs4_proc_bind_conn_to_session()
6447  *
6448  * The 4.1 client currently uses the same TCP connection for the
6449  * fore and backchannel.
6450  */
6451 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6452 {
6453 	int status;
6454 	struct nfs41_bind_conn_to_session_res res;
6455 	struct rpc_message msg = {
6456 		.rpc_proc =
6457 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6458 		.rpc_argp = clp,
6459 		.rpc_resp = &res,
6460 		.rpc_cred = cred,
6461 	};
6462 
6463 	dprintk("--> %s\n", __func__);
6464 
6465 	res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6466 	if (unlikely(res.session == NULL)) {
6467 		status = -ENOMEM;
6468 		goto out;
6469 	}
6470 
6471 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6472 	trace_nfs4_bind_conn_to_session(clp, status);
6473 	if (status == 0) {
6474 		if (memcmp(res.session->sess_id.data,
6475 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6476 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6477 			status = -EIO;
6478 			goto out_session;
6479 		}
6480 		if (res.dir != NFS4_CDFS4_BOTH) {
6481 			dprintk("NFS: %s: Unexpected direction from server\n",
6482 				__func__);
6483 			status = -EIO;
6484 			goto out_session;
6485 		}
6486 		if (res.use_conn_in_rdma_mode) {
6487 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6488 				__func__);
6489 			status = -EIO;
6490 			goto out_session;
6491 		}
6492 	}
6493 out_session:
6494 	kfree(res.session);
6495 out:
6496 	dprintk("<-- %s status= %d\n", __func__, status);
6497 	return status;
6498 }
6499 
6500 /*
6501  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6502  * and operations we'd like to see to enable certain features in the allow map
6503  */
6504 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6505 	.how = SP4_MACH_CRED,
6506 	.enforce.u.words = {
6507 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6508 		      1 << (OP_EXCHANGE_ID - 32) |
6509 		      1 << (OP_CREATE_SESSION - 32) |
6510 		      1 << (OP_DESTROY_SESSION - 32) |
6511 		      1 << (OP_DESTROY_CLIENTID - 32)
6512 	},
6513 	.allow.u.words = {
6514 		[0] = 1 << (OP_CLOSE) |
6515 		      1 << (OP_LOCKU) |
6516 		      1 << (OP_COMMIT),
6517 		[1] = 1 << (OP_SECINFO - 32) |
6518 		      1 << (OP_SECINFO_NO_NAME - 32) |
6519 		      1 << (OP_TEST_STATEID - 32) |
6520 		      1 << (OP_FREE_STATEID - 32) |
6521 		      1 << (OP_WRITE - 32)
6522 	}
6523 };
6524 
6525 /*
6526  * Select the state protection mode for client `clp' given the server results
6527  * from exchange_id in `sp'.
6528  *
6529  * Returns 0 on success, negative errno otherwise.
6530  */
6531 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6532 				 struct nfs41_state_protection *sp)
6533 {
6534 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6535 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6536 		      1 << (OP_EXCHANGE_ID - 32) |
6537 		      1 << (OP_CREATE_SESSION - 32) |
6538 		      1 << (OP_DESTROY_SESSION - 32) |
6539 		      1 << (OP_DESTROY_CLIENTID - 32)
6540 	};
6541 	unsigned int i;
6542 
6543 	if (sp->how == SP4_MACH_CRED) {
6544 		/* Print state protect result */
6545 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6546 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6547 			if (test_bit(i, sp->enforce.u.longs))
6548 				dfprintk(MOUNT, "  enforce op %d\n", i);
6549 			if (test_bit(i, sp->allow.u.longs))
6550 				dfprintk(MOUNT, "  allow op %d\n", i);
6551 		}
6552 
6553 		/* make sure nothing is on enforce list that isn't supported */
6554 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6555 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6556 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6557 				return -EINVAL;
6558 			}
6559 		}
6560 
6561 		/*
6562 		 * Minimal mode - state operations are allowed to use machine
6563 		 * credential.  Note this already happens by default, so the
6564 		 * client doesn't have to do anything more than the negotiation.
6565 		 *
6566 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6567 		 *       we're already using the machine cred for exchange_id
6568 		 *       and will never use a different cred.
6569 		 */
6570 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6571 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6572 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6573 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6574 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6575 			dfprintk(MOUNT, "  minimal mode enabled\n");
6576 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6577 		} else {
6578 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6579 			return -EINVAL;
6580 		}
6581 
6582 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6583 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6584 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6585 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6586 		}
6587 
6588 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6589 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6590 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6591 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6592 		}
6593 
6594 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6595 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6596 			dfprintk(MOUNT, "  stateid mode enabled\n");
6597 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6598 		}
6599 
6600 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6601 			dfprintk(MOUNT, "  write mode enabled\n");
6602 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6603 		}
6604 
6605 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6606 			dfprintk(MOUNT, "  commit mode enabled\n");
6607 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6608 		}
6609 	}
6610 
6611 	return 0;
6612 }
6613 
6614 /*
6615  * _nfs4_proc_exchange_id()
6616  *
6617  * Wrapper for EXCHANGE_ID operation.
6618  */
6619 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6620 	u32 sp4_how)
6621 {
6622 	nfs4_verifier verifier;
6623 	struct nfs41_exchange_id_args args = {
6624 		.verifier = &verifier,
6625 		.client = clp,
6626 #ifdef CONFIG_NFS_V4_1_MIGRATION
6627 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6628 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6629 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6630 #else
6631 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6632 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6633 #endif
6634 	};
6635 	struct nfs41_exchange_id_res res = {
6636 		0
6637 	};
6638 	int status;
6639 	struct rpc_message msg = {
6640 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6641 		.rpc_argp = &args,
6642 		.rpc_resp = &res,
6643 		.rpc_cred = cred,
6644 	};
6645 
6646 	nfs4_init_boot_verifier(clp, &verifier);
6647 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6648 							sizeof(args.id));
6649 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6650 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6651 		args.id_len, args.id);
6652 
6653 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6654 					GFP_NOFS);
6655 	if (unlikely(res.server_owner == NULL)) {
6656 		status = -ENOMEM;
6657 		goto out;
6658 	}
6659 
6660 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6661 					GFP_NOFS);
6662 	if (unlikely(res.server_scope == NULL)) {
6663 		status = -ENOMEM;
6664 		goto out_server_owner;
6665 	}
6666 
6667 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6668 	if (unlikely(res.impl_id == NULL)) {
6669 		status = -ENOMEM;
6670 		goto out_server_scope;
6671 	}
6672 
6673 	switch (sp4_how) {
6674 	case SP4_NONE:
6675 		args.state_protect.how = SP4_NONE;
6676 		break;
6677 
6678 	case SP4_MACH_CRED:
6679 		args.state_protect = nfs4_sp4_mach_cred_request;
6680 		break;
6681 
6682 	default:
6683 		/* unsupported! */
6684 		WARN_ON_ONCE(1);
6685 		status = -EINVAL;
6686 		goto out_server_scope;
6687 	}
6688 
6689 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6690 	trace_nfs4_exchange_id(clp, status);
6691 	if (status == 0)
6692 		status = nfs4_check_cl_exchange_flags(res.flags);
6693 
6694 	if (status == 0)
6695 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6696 
6697 	if (status == 0) {
6698 		clp->cl_clientid = res.clientid;
6699 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6700 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6701 			clp->cl_seqid = res.seqid;
6702 
6703 		kfree(clp->cl_serverowner);
6704 		clp->cl_serverowner = res.server_owner;
6705 		res.server_owner = NULL;
6706 
6707 		/* use the most recent implementation id */
6708 		kfree(clp->cl_implid);
6709 		clp->cl_implid = res.impl_id;
6710 
6711 		if (clp->cl_serverscope != NULL &&
6712 		    !nfs41_same_server_scope(clp->cl_serverscope,
6713 					     res.server_scope)) {
6714 			dprintk("%s: server_scope mismatch detected\n",
6715 				__func__);
6716 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6717 			kfree(clp->cl_serverscope);
6718 			clp->cl_serverscope = NULL;
6719 		}
6720 
6721 		if (clp->cl_serverscope == NULL) {
6722 			clp->cl_serverscope = res.server_scope;
6723 			goto out;
6724 		}
6725 	} else
6726 		kfree(res.impl_id);
6727 
6728 out_server_owner:
6729 	kfree(res.server_owner);
6730 out_server_scope:
6731 	kfree(res.server_scope);
6732 out:
6733 	if (clp->cl_implid != NULL)
6734 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6735 			"domain: %s, name: %s, date: %llu,%u\n",
6736 			clp->cl_implid->domain, clp->cl_implid->name,
6737 			clp->cl_implid->date.seconds,
6738 			clp->cl_implid->date.nseconds);
6739 	dprintk("NFS reply exchange_id: %d\n", status);
6740 	return status;
6741 }
6742 
6743 /*
6744  * nfs4_proc_exchange_id()
6745  *
6746  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6747  *
6748  * Since the clientid has expired, all compounds using sessions
6749  * associated with the stale clientid will be returning
6750  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6751  * be in some phase of session reset.
6752  *
6753  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6754  */
6755 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6756 {
6757 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6758 	int status;
6759 
6760 	/* try SP4_MACH_CRED if krb5i/p	*/
6761 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6762 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6763 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6764 		if (!status)
6765 			return 0;
6766 	}
6767 
6768 	/* try SP4_NONE */
6769 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6770 }
6771 
6772 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6773 		struct rpc_cred *cred)
6774 {
6775 	struct rpc_message msg = {
6776 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6777 		.rpc_argp = clp,
6778 		.rpc_cred = cred,
6779 	};
6780 	int status;
6781 
6782 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6783 	trace_nfs4_destroy_clientid(clp, status);
6784 	if (status)
6785 		dprintk("NFS: Got error %d from the server %s on "
6786 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6787 	return status;
6788 }
6789 
6790 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6791 		struct rpc_cred *cred)
6792 {
6793 	unsigned int loop;
6794 	int ret;
6795 
6796 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6797 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6798 		switch (ret) {
6799 		case -NFS4ERR_DELAY:
6800 		case -NFS4ERR_CLIENTID_BUSY:
6801 			ssleep(1);
6802 			break;
6803 		default:
6804 			return ret;
6805 		}
6806 	}
6807 	return 0;
6808 }
6809 
6810 int nfs4_destroy_clientid(struct nfs_client *clp)
6811 {
6812 	struct rpc_cred *cred;
6813 	int ret = 0;
6814 
6815 	if (clp->cl_mvops->minor_version < 1)
6816 		goto out;
6817 	if (clp->cl_exchange_flags == 0)
6818 		goto out;
6819 	if (clp->cl_preserve_clid)
6820 		goto out;
6821 	cred = nfs4_get_clid_cred(clp);
6822 	ret = nfs4_proc_destroy_clientid(clp, cred);
6823 	if (cred)
6824 		put_rpccred(cred);
6825 	switch (ret) {
6826 	case 0:
6827 	case -NFS4ERR_STALE_CLIENTID:
6828 		clp->cl_exchange_flags = 0;
6829 	}
6830 out:
6831 	return ret;
6832 }
6833 
6834 struct nfs4_get_lease_time_data {
6835 	struct nfs4_get_lease_time_args *args;
6836 	struct nfs4_get_lease_time_res *res;
6837 	struct nfs_client *clp;
6838 };
6839 
6840 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6841 					void *calldata)
6842 {
6843 	struct nfs4_get_lease_time_data *data =
6844 			(struct nfs4_get_lease_time_data *)calldata;
6845 
6846 	dprintk("--> %s\n", __func__);
6847 	/* just setup sequence, do not trigger session recovery
6848 	   since we're invoked within one */
6849 	nfs41_setup_sequence(data->clp->cl_session,
6850 			&data->args->la_seq_args,
6851 			&data->res->lr_seq_res,
6852 			task);
6853 	dprintk("<-- %s\n", __func__);
6854 }
6855 
6856 /*
6857  * Called from nfs4_state_manager thread for session setup, so don't recover
6858  * from sequence operation or clientid errors.
6859  */
6860 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6861 {
6862 	struct nfs4_get_lease_time_data *data =
6863 			(struct nfs4_get_lease_time_data *)calldata;
6864 
6865 	dprintk("--> %s\n", __func__);
6866 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6867 		return;
6868 	switch (task->tk_status) {
6869 	case -NFS4ERR_DELAY:
6870 	case -NFS4ERR_GRACE:
6871 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6872 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
6873 		task->tk_status = 0;
6874 		/* fall through */
6875 	case -NFS4ERR_RETRY_UNCACHED_REP:
6876 		rpc_restart_call_prepare(task);
6877 		return;
6878 	}
6879 	dprintk("<-- %s\n", __func__);
6880 }
6881 
6882 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
6883 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
6884 	.rpc_call_done = nfs4_get_lease_time_done,
6885 };
6886 
6887 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
6888 {
6889 	struct rpc_task *task;
6890 	struct nfs4_get_lease_time_args args;
6891 	struct nfs4_get_lease_time_res res = {
6892 		.lr_fsinfo = fsinfo,
6893 	};
6894 	struct nfs4_get_lease_time_data data = {
6895 		.args = &args,
6896 		.res = &res,
6897 		.clp = clp,
6898 	};
6899 	struct rpc_message msg = {
6900 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
6901 		.rpc_argp = &args,
6902 		.rpc_resp = &res,
6903 	};
6904 	struct rpc_task_setup task_setup = {
6905 		.rpc_client = clp->cl_rpcclient,
6906 		.rpc_message = &msg,
6907 		.callback_ops = &nfs4_get_lease_time_ops,
6908 		.callback_data = &data,
6909 		.flags = RPC_TASK_TIMEOUT,
6910 	};
6911 	int status;
6912 
6913 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
6914 	nfs4_set_sequence_privileged(&args.la_seq_args);
6915 	dprintk("--> %s\n", __func__);
6916 	task = rpc_run_task(&task_setup);
6917 
6918 	if (IS_ERR(task))
6919 		status = PTR_ERR(task);
6920 	else {
6921 		status = task->tk_status;
6922 		rpc_put_task(task);
6923 	}
6924 	dprintk("<-- %s return %d\n", __func__, status);
6925 
6926 	return status;
6927 }
6928 
6929 /*
6930  * Initialize the values to be used by the client in CREATE_SESSION
6931  * If nfs4_init_session set the fore channel request and response sizes,
6932  * use them.
6933  *
6934  * Set the back channel max_resp_sz_cached to zero to force the client to
6935  * always set csa_cachethis to FALSE because the current implementation
6936  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
6937  */
6938 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
6939 {
6940 	unsigned int max_rqst_sz, max_resp_sz;
6941 
6942 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
6943 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
6944 
6945 	/* Fore channel attributes */
6946 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
6947 	args->fc_attrs.max_resp_sz = max_resp_sz;
6948 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
6949 	args->fc_attrs.max_reqs = max_session_slots;
6950 
6951 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
6952 		"max_ops=%u max_reqs=%u\n",
6953 		__func__,
6954 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
6955 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
6956 
6957 	/* Back channel attributes */
6958 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
6959 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
6960 	args->bc_attrs.max_resp_sz_cached = 0;
6961 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
6962 	args->bc_attrs.max_reqs = 1;
6963 
6964 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
6965 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
6966 		__func__,
6967 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
6968 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
6969 		args->bc_attrs.max_reqs);
6970 }
6971 
6972 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6973 {
6974 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
6975 	struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
6976 
6977 	if (rcvd->max_resp_sz > sent->max_resp_sz)
6978 		return -EINVAL;
6979 	/*
6980 	 * Our requested max_ops is the minimum we need; we're not
6981 	 * prepared to break up compounds into smaller pieces than that.
6982 	 * So, no point even trying to continue if the server won't
6983 	 * cooperate:
6984 	 */
6985 	if (rcvd->max_ops < sent->max_ops)
6986 		return -EINVAL;
6987 	if (rcvd->max_reqs == 0)
6988 		return -EINVAL;
6989 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
6990 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
6991 	return 0;
6992 }
6993 
6994 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6995 {
6996 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
6997 	struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
6998 
6999 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7000 		return -EINVAL;
7001 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7002 		return -EINVAL;
7003 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7004 		return -EINVAL;
7005 	/* These would render the backchannel useless: */
7006 	if (rcvd->max_ops != sent->max_ops)
7007 		return -EINVAL;
7008 	if (rcvd->max_reqs != sent->max_reqs)
7009 		return -EINVAL;
7010 	return 0;
7011 }
7012 
7013 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7014 				     struct nfs4_session *session)
7015 {
7016 	int ret;
7017 
7018 	ret = nfs4_verify_fore_channel_attrs(args, session);
7019 	if (ret)
7020 		return ret;
7021 	return nfs4_verify_back_channel_attrs(args, session);
7022 }
7023 
7024 static int _nfs4_proc_create_session(struct nfs_client *clp,
7025 		struct rpc_cred *cred)
7026 {
7027 	struct nfs4_session *session = clp->cl_session;
7028 	struct nfs41_create_session_args args = {
7029 		.client = clp,
7030 		.cb_program = NFS4_CALLBACK,
7031 	};
7032 	struct nfs41_create_session_res res = {
7033 		.client = clp,
7034 	};
7035 	struct rpc_message msg = {
7036 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7037 		.rpc_argp = &args,
7038 		.rpc_resp = &res,
7039 		.rpc_cred = cred,
7040 	};
7041 	int status;
7042 
7043 	nfs4_init_channel_attrs(&args);
7044 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7045 
7046 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7047 	trace_nfs4_create_session(clp, status);
7048 
7049 	if (!status) {
7050 		/* Verify the session's negotiated channel_attrs values */
7051 		status = nfs4_verify_channel_attrs(&args, session);
7052 		/* Increment the clientid slot sequence id */
7053 		clp->cl_seqid++;
7054 	}
7055 
7056 	return status;
7057 }
7058 
7059 /*
7060  * Issues a CREATE_SESSION operation to the server.
7061  * It is the responsibility of the caller to verify the session is
7062  * expired before calling this routine.
7063  */
7064 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7065 {
7066 	int status;
7067 	unsigned *ptr;
7068 	struct nfs4_session *session = clp->cl_session;
7069 
7070 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7071 
7072 	status = _nfs4_proc_create_session(clp, cred);
7073 	if (status)
7074 		goto out;
7075 
7076 	/* Init or reset the session slot tables */
7077 	status = nfs4_setup_session_slot_tables(session);
7078 	dprintk("slot table setup returned %d\n", status);
7079 	if (status)
7080 		goto out;
7081 
7082 	ptr = (unsigned *)&session->sess_id.data[0];
7083 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7084 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7085 out:
7086 	dprintk("<-- %s\n", __func__);
7087 	return status;
7088 }
7089 
7090 /*
7091  * Issue the over-the-wire RPC DESTROY_SESSION.
7092  * The caller must serialize access to this routine.
7093  */
7094 int nfs4_proc_destroy_session(struct nfs4_session *session,
7095 		struct rpc_cred *cred)
7096 {
7097 	struct rpc_message msg = {
7098 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7099 		.rpc_argp = session,
7100 		.rpc_cred = cred,
7101 	};
7102 	int status = 0;
7103 
7104 	dprintk("--> nfs4_proc_destroy_session\n");
7105 
7106 	/* session is still being setup */
7107 	if (session->clp->cl_cons_state != NFS_CS_READY)
7108 		return status;
7109 
7110 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7111 	trace_nfs4_destroy_session(session->clp, status);
7112 
7113 	if (status)
7114 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7115 			"Session has been destroyed regardless...\n", status);
7116 
7117 	dprintk("<-- nfs4_proc_destroy_session\n");
7118 	return status;
7119 }
7120 
7121 /*
7122  * Renew the cl_session lease.
7123  */
7124 struct nfs4_sequence_data {
7125 	struct nfs_client *clp;
7126 	struct nfs4_sequence_args args;
7127 	struct nfs4_sequence_res res;
7128 };
7129 
7130 static void nfs41_sequence_release(void *data)
7131 {
7132 	struct nfs4_sequence_data *calldata = data;
7133 	struct nfs_client *clp = calldata->clp;
7134 
7135 	if (atomic_read(&clp->cl_count) > 1)
7136 		nfs4_schedule_state_renewal(clp);
7137 	nfs_put_client(clp);
7138 	kfree(calldata);
7139 }
7140 
7141 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7142 {
7143 	switch(task->tk_status) {
7144 	case -NFS4ERR_DELAY:
7145 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7146 		return -EAGAIN;
7147 	default:
7148 		nfs4_schedule_lease_recovery(clp);
7149 	}
7150 	return 0;
7151 }
7152 
7153 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7154 {
7155 	struct nfs4_sequence_data *calldata = data;
7156 	struct nfs_client *clp = calldata->clp;
7157 
7158 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7159 		return;
7160 
7161 	trace_nfs4_sequence(clp, task->tk_status);
7162 	if (task->tk_status < 0) {
7163 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7164 		if (atomic_read(&clp->cl_count) == 1)
7165 			goto out;
7166 
7167 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7168 			rpc_restart_call_prepare(task);
7169 			return;
7170 		}
7171 	}
7172 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7173 out:
7174 	dprintk("<-- %s\n", __func__);
7175 }
7176 
7177 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7178 {
7179 	struct nfs4_sequence_data *calldata = data;
7180 	struct nfs_client *clp = calldata->clp;
7181 	struct nfs4_sequence_args *args;
7182 	struct nfs4_sequence_res *res;
7183 
7184 	args = task->tk_msg.rpc_argp;
7185 	res = task->tk_msg.rpc_resp;
7186 
7187 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7188 }
7189 
7190 static const struct rpc_call_ops nfs41_sequence_ops = {
7191 	.rpc_call_done = nfs41_sequence_call_done,
7192 	.rpc_call_prepare = nfs41_sequence_prepare,
7193 	.rpc_release = nfs41_sequence_release,
7194 };
7195 
7196 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7197 		struct rpc_cred *cred,
7198 		bool is_privileged)
7199 {
7200 	struct nfs4_sequence_data *calldata;
7201 	struct rpc_message msg = {
7202 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7203 		.rpc_cred = cred,
7204 	};
7205 	struct rpc_task_setup task_setup_data = {
7206 		.rpc_client = clp->cl_rpcclient,
7207 		.rpc_message = &msg,
7208 		.callback_ops = &nfs41_sequence_ops,
7209 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7210 	};
7211 
7212 	if (!atomic_inc_not_zero(&clp->cl_count))
7213 		return ERR_PTR(-EIO);
7214 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7215 	if (calldata == NULL) {
7216 		nfs_put_client(clp);
7217 		return ERR_PTR(-ENOMEM);
7218 	}
7219 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7220 	if (is_privileged)
7221 		nfs4_set_sequence_privileged(&calldata->args);
7222 	msg.rpc_argp = &calldata->args;
7223 	msg.rpc_resp = &calldata->res;
7224 	calldata->clp = clp;
7225 	task_setup_data.callback_data = calldata;
7226 
7227 	return rpc_run_task(&task_setup_data);
7228 }
7229 
7230 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7231 {
7232 	struct rpc_task *task;
7233 	int ret = 0;
7234 
7235 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7236 		return 0;
7237 	task = _nfs41_proc_sequence(clp, cred, false);
7238 	if (IS_ERR(task))
7239 		ret = PTR_ERR(task);
7240 	else
7241 		rpc_put_task_async(task);
7242 	dprintk("<-- %s status=%d\n", __func__, ret);
7243 	return ret;
7244 }
7245 
7246 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7247 {
7248 	struct rpc_task *task;
7249 	int ret;
7250 
7251 	task = _nfs41_proc_sequence(clp, cred, true);
7252 	if (IS_ERR(task)) {
7253 		ret = PTR_ERR(task);
7254 		goto out;
7255 	}
7256 	ret = rpc_wait_for_completion_task(task);
7257 	if (!ret) {
7258 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7259 
7260 		if (task->tk_status == 0)
7261 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7262 		ret = task->tk_status;
7263 	}
7264 	rpc_put_task(task);
7265 out:
7266 	dprintk("<-- %s status=%d\n", __func__, ret);
7267 	return ret;
7268 }
7269 
7270 struct nfs4_reclaim_complete_data {
7271 	struct nfs_client *clp;
7272 	struct nfs41_reclaim_complete_args arg;
7273 	struct nfs41_reclaim_complete_res res;
7274 };
7275 
7276 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7277 {
7278 	struct nfs4_reclaim_complete_data *calldata = data;
7279 
7280 	nfs41_setup_sequence(calldata->clp->cl_session,
7281 			&calldata->arg.seq_args,
7282 			&calldata->res.seq_res,
7283 			task);
7284 }
7285 
7286 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7287 {
7288 	switch(task->tk_status) {
7289 	case 0:
7290 	case -NFS4ERR_COMPLETE_ALREADY:
7291 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7292 		break;
7293 	case -NFS4ERR_DELAY:
7294 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7295 		/* fall through */
7296 	case -NFS4ERR_RETRY_UNCACHED_REP:
7297 		return -EAGAIN;
7298 	default:
7299 		nfs4_schedule_lease_recovery(clp);
7300 	}
7301 	return 0;
7302 }
7303 
7304 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7305 {
7306 	struct nfs4_reclaim_complete_data *calldata = data;
7307 	struct nfs_client *clp = calldata->clp;
7308 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7309 
7310 	dprintk("--> %s\n", __func__);
7311 	if (!nfs41_sequence_done(task, res))
7312 		return;
7313 
7314 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7315 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7316 		rpc_restart_call_prepare(task);
7317 		return;
7318 	}
7319 	dprintk("<-- %s\n", __func__);
7320 }
7321 
7322 static void nfs4_free_reclaim_complete_data(void *data)
7323 {
7324 	struct nfs4_reclaim_complete_data *calldata = data;
7325 
7326 	kfree(calldata);
7327 }
7328 
7329 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7330 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7331 	.rpc_call_done = nfs4_reclaim_complete_done,
7332 	.rpc_release = nfs4_free_reclaim_complete_data,
7333 };
7334 
7335 /*
7336  * Issue a global reclaim complete.
7337  */
7338 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7339 		struct rpc_cred *cred)
7340 {
7341 	struct nfs4_reclaim_complete_data *calldata;
7342 	struct rpc_task *task;
7343 	struct rpc_message msg = {
7344 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7345 		.rpc_cred = cred,
7346 	};
7347 	struct rpc_task_setup task_setup_data = {
7348 		.rpc_client = clp->cl_rpcclient,
7349 		.rpc_message = &msg,
7350 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7351 		.flags = RPC_TASK_ASYNC,
7352 	};
7353 	int status = -ENOMEM;
7354 
7355 	dprintk("--> %s\n", __func__);
7356 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7357 	if (calldata == NULL)
7358 		goto out;
7359 	calldata->clp = clp;
7360 	calldata->arg.one_fs = 0;
7361 
7362 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7363 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7364 	msg.rpc_argp = &calldata->arg;
7365 	msg.rpc_resp = &calldata->res;
7366 	task_setup_data.callback_data = calldata;
7367 	task = rpc_run_task(&task_setup_data);
7368 	if (IS_ERR(task)) {
7369 		status = PTR_ERR(task);
7370 		goto out;
7371 	}
7372 	status = nfs4_wait_for_completion_rpc_task(task);
7373 	if (status == 0)
7374 		status = task->tk_status;
7375 	rpc_put_task(task);
7376 	return 0;
7377 out:
7378 	dprintk("<-- %s status=%d\n", __func__, status);
7379 	return status;
7380 }
7381 
7382 static void
7383 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7384 {
7385 	struct nfs4_layoutget *lgp = calldata;
7386 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7387 	struct nfs4_session *session = nfs4_get_session(server);
7388 
7389 	dprintk("--> %s\n", __func__);
7390 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7391 	 * right now covering the LAYOUTGET we are about to send.
7392 	 * However, that is not so catastrophic, and there seems
7393 	 * to be no way to prevent it completely.
7394 	 */
7395 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7396 				&lgp->res.seq_res, task))
7397 		return;
7398 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7399 					  NFS_I(lgp->args.inode)->layout,
7400 					  lgp->args.ctx->state)) {
7401 		rpc_exit(task, NFS4_OK);
7402 	}
7403 }
7404 
7405 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7406 {
7407 	struct nfs4_layoutget *lgp = calldata;
7408 	struct inode *inode = lgp->args.inode;
7409 	struct nfs_server *server = NFS_SERVER(inode);
7410 	struct pnfs_layout_hdr *lo;
7411 	struct nfs4_state *state = NULL;
7412 	unsigned long timeo, giveup;
7413 
7414 	dprintk("--> %s\n", __func__);
7415 
7416 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7417 		goto out;
7418 
7419 	switch (task->tk_status) {
7420 	case 0:
7421 		goto out;
7422 	case -NFS4ERR_LAYOUTTRYLATER:
7423 	case -NFS4ERR_RECALLCONFLICT:
7424 		timeo = rpc_get_timeout(task->tk_client);
7425 		giveup = lgp->args.timestamp + timeo;
7426 		if (time_after(giveup, jiffies))
7427 			task->tk_status = -NFS4ERR_DELAY;
7428 		break;
7429 	case -NFS4ERR_EXPIRED:
7430 	case -NFS4ERR_BAD_STATEID:
7431 		spin_lock(&inode->i_lock);
7432 		lo = NFS_I(inode)->layout;
7433 		if (!lo || list_empty(&lo->plh_segs)) {
7434 			spin_unlock(&inode->i_lock);
7435 			/* If the open stateid was bad, then recover it. */
7436 			state = lgp->args.ctx->state;
7437 		} else {
7438 			LIST_HEAD(head);
7439 
7440 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7441 			spin_unlock(&inode->i_lock);
7442 			/* Mark the bad layout state as invalid, then
7443 			 * retry using the open stateid. */
7444 			pnfs_free_lseg_list(&head);
7445 		}
7446 	}
7447 	if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
7448 		rpc_restart_call_prepare(task);
7449 out:
7450 	dprintk("<-- %s\n", __func__);
7451 }
7452 
7453 static size_t max_response_pages(struct nfs_server *server)
7454 {
7455 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7456 	return nfs_page_array_len(0, max_resp_sz);
7457 }
7458 
7459 static void nfs4_free_pages(struct page **pages, size_t size)
7460 {
7461 	int i;
7462 
7463 	if (!pages)
7464 		return;
7465 
7466 	for (i = 0; i < size; i++) {
7467 		if (!pages[i])
7468 			break;
7469 		__free_page(pages[i]);
7470 	}
7471 	kfree(pages);
7472 }
7473 
7474 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7475 {
7476 	struct page **pages;
7477 	int i;
7478 
7479 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7480 	if (!pages) {
7481 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7482 		return NULL;
7483 	}
7484 
7485 	for (i = 0; i < size; i++) {
7486 		pages[i] = alloc_page(gfp_flags);
7487 		if (!pages[i]) {
7488 			dprintk("%s: failed to allocate page\n", __func__);
7489 			nfs4_free_pages(pages, size);
7490 			return NULL;
7491 		}
7492 	}
7493 
7494 	return pages;
7495 }
7496 
7497 static void nfs4_layoutget_release(void *calldata)
7498 {
7499 	struct nfs4_layoutget *lgp = calldata;
7500 	struct inode *inode = lgp->args.inode;
7501 	struct nfs_server *server = NFS_SERVER(inode);
7502 	size_t max_pages = max_response_pages(server);
7503 
7504 	dprintk("--> %s\n", __func__);
7505 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7506 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7507 	put_nfs_open_context(lgp->args.ctx);
7508 	kfree(calldata);
7509 	dprintk("<-- %s\n", __func__);
7510 }
7511 
7512 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7513 	.rpc_call_prepare = nfs4_layoutget_prepare,
7514 	.rpc_call_done = nfs4_layoutget_done,
7515 	.rpc_release = nfs4_layoutget_release,
7516 };
7517 
7518 struct pnfs_layout_segment *
7519 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7520 {
7521 	struct inode *inode = lgp->args.inode;
7522 	struct nfs_server *server = NFS_SERVER(inode);
7523 	size_t max_pages = max_response_pages(server);
7524 	struct rpc_task *task;
7525 	struct rpc_message msg = {
7526 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7527 		.rpc_argp = &lgp->args,
7528 		.rpc_resp = &lgp->res,
7529 		.rpc_cred = lgp->cred,
7530 	};
7531 	struct rpc_task_setup task_setup_data = {
7532 		.rpc_client = server->client,
7533 		.rpc_message = &msg,
7534 		.callback_ops = &nfs4_layoutget_call_ops,
7535 		.callback_data = lgp,
7536 		.flags = RPC_TASK_ASYNC,
7537 	};
7538 	struct pnfs_layout_segment *lseg = NULL;
7539 	int status = 0;
7540 
7541 	dprintk("--> %s\n", __func__);
7542 
7543 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7544 	if (!lgp->args.layout.pages) {
7545 		nfs4_layoutget_release(lgp);
7546 		return ERR_PTR(-ENOMEM);
7547 	}
7548 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7549 	lgp->args.timestamp = jiffies;
7550 
7551 	lgp->res.layoutp = &lgp->args.layout;
7552 	lgp->res.seq_res.sr_slot = NULL;
7553 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7554 
7555 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7556 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7557 
7558 	task = rpc_run_task(&task_setup_data);
7559 	if (IS_ERR(task))
7560 		return ERR_CAST(task);
7561 	status = nfs4_wait_for_completion_rpc_task(task);
7562 	if (status == 0)
7563 		status = task->tk_status;
7564 	trace_nfs4_layoutget(lgp->args.ctx,
7565 			&lgp->args.range,
7566 			&lgp->res.range,
7567 			status);
7568 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7569 	if (status == 0 && lgp->res.layoutp->len)
7570 		lseg = pnfs_layout_process(lgp);
7571 	rpc_put_task(task);
7572 	dprintk("<-- %s status=%d\n", __func__, status);
7573 	if (status)
7574 		return ERR_PTR(status);
7575 	return lseg;
7576 }
7577 
7578 static void
7579 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7580 {
7581 	struct nfs4_layoutreturn *lrp = calldata;
7582 
7583 	dprintk("--> %s\n", __func__);
7584 	nfs41_setup_sequence(lrp->clp->cl_session,
7585 			&lrp->args.seq_args,
7586 			&lrp->res.seq_res,
7587 			task);
7588 }
7589 
7590 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7591 {
7592 	struct nfs4_layoutreturn *lrp = calldata;
7593 	struct nfs_server *server;
7594 
7595 	dprintk("--> %s\n", __func__);
7596 
7597 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7598 		return;
7599 
7600 	server = NFS_SERVER(lrp->args.inode);
7601 	switch (task->tk_status) {
7602 	default:
7603 		task->tk_status = 0;
7604 	case 0:
7605 		break;
7606 	case -NFS4ERR_DELAY:
7607 		if (nfs4_async_handle_error(task, server, NULL) != -EAGAIN)
7608 			break;
7609 		rpc_restart_call_prepare(task);
7610 		return;
7611 	}
7612 	dprintk("<-- %s\n", __func__);
7613 }
7614 
7615 static void nfs4_layoutreturn_release(void *calldata)
7616 {
7617 	struct nfs4_layoutreturn *lrp = calldata;
7618 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7619 
7620 	dprintk("--> %s\n", __func__);
7621 	spin_lock(&lo->plh_inode->i_lock);
7622 	if (lrp->res.lrs_present)
7623 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7624 	lo->plh_block_lgets--;
7625 	spin_unlock(&lo->plh_inode->i_lock);
7626 	pnfs_put_layout_hdr(lrp->args.layout);
7627 	kfree(calldata);
7628 	dprintk("<-- %s\n", __func__);
7629 }
7630 
7631 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7632 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7633 	.rpc_call_done = nfs4_layoutreturn_done,
7634 	.rpc_release = nfs4_layoutreturn_release,
7635 };
7636 
7637 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7638 {
7639 	struct rpc_task *task;
7640 	struct rpc_message msg = {
7641 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7642 		.rpc_argp = &lrp->args,
7643 		.rpc_resp = &lrp->res,
7644 		.rpc_cred = lrp->cred,
7645 	};
7646 	struct rpc_task_setup task_setup_data = {
7647 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7648 		.rpc_message = &msg,
7649 		.callback_ops = &nfs4_layoutreturn_call_ops,
7650 		.callback_data = lrp,
7651 	};
7652 	int status;
7653 
7654 	dprintk("--> %s\n", __func__);
7655 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7656 	task = rpc_run_task(&task_setup_data);
7657 	if (IS_ERR(task))
7658 		return PTR_ERR(task);
7659 	status = task->tk_status;
7660 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7661 	dprintk("<-- %s status=%d\n", __func__, status);
7662 	rpc_put_task(task);
7663 	return status;
7664 }
7665 
7666 /*
7667  * Retrieve the list of Data Server devices from the MDS.
7668  */
7669 static int _nfs4_getdevicelist(struct nfs_server *server,
7670 				    const struct nfs_fh *fh,
7671 				    struct pnfs_devicelist *devlist)
7672 {
7673 	struct nfs4_getdevicelist_args args = {
7674 		.fh = fh,
7675 		.layoutclass = server->pnfs_curr_ld->id,
7676 	};
7677 	struct nfs4_getdevicelist_res res = {
7678 		.devlist = devlist,
7679 	};
7680 	struct rpc_message msg = {
7681 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
7682 		.rpc_argp = &args,
7683 		.rpc_resp = &res,
7684 	};
7685 	int status;
7686 
7687 	dprintk("--> %s\n", __func__);
7688 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
7689 				&res.seq_res, 0);
7690 	dprintk("<-- %s status=%d\n", __func__, status);
7691 	return status;
7692 }
7693 
7694 int nfs4_proc_getdevicelist(struct nfs_server *server,
7695 			    const struct nfs_fh *fh,
7696 			    struct pnfs_devicelist *devlist)
7697 {
7698 	struct nfs4_exception exception = { };
7699 	int err;
7700 
7701 	do {
7702 		err = nfs4_handle_exception(server,
7703 				_nfs4_getdevicelist(server, fh, devlist),
7704 				&exception);
7705 	} while (exception.retry);
7706 
7707 	dprintk("%s: err=%d, num_devs=%u\n", __func__,
7708 		err, devlist->num_devs);
7709 
7710 	return err;
7711 }
7712 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
7713 
7714 static int
7715 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7716 		struct pnfs_device *pdev,
7717 		struct rpc_cred *cred)
7718 {
7719 	struct nfs4_getdeviceinfo_args args = {
7720 		.pdev = pdev,
7721 	};
7722 	struct nfs4_getdeviceinfo_res res = {
7723 		.pdev = pdev,
7724 	};
7725 	struct rpc_message msg = {
7726 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7727 		.rpc_argp = &args,
7728 		.rpc_resp = &res,
7729 		.rpc_cred = cred,
7730 	};
7731 	int status;
7732 
7733 	dprintk("--> %s\n", __func__);
7734 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7735 	dprintk("<-- %s status=%d\n", __func__, status);
7736 
7737 	return status;
7738 }
7739 
7740 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7741 		struct pnfs_device *pdev,
7742 		struct rpc_cred *cred)
7743 {
7744 	struct nfs4_exception exception = { };
7745 	int err;
7746 
7747 	do {
7748 		err = nfs4_handle_exception(server,
7749 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7750 					&exception);
7751 	} while (exception.retry);
7752 	return err;
7753 }
7754 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7755 
7756 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7757 {
7758 	struct nfs4_layoutcommit_data *data = calldata;
7759 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7760 	struct nfs4_session *session = nfs4_get_session(server);
7761 
7762 	nfs41_setup_sequence(session,
7763 			&data->args.seq_args,
7764 			&data->res.seq_res,
7765 			task);
7766 }
7767 
7768 static void
7769 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7770 {
7771 	struct nfs4_layoutcommit_data *data = calldata;
7772 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7773 
7774 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7775 		return;
7776 
7777 	switch (task->tk_status) { /* Just ignore these failures */
7778 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7779 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7780 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7781 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7782 		task->tk_status = 0;
7783 		break;
7784 	case 0:
7785 		nfs_post_op_update_inode_force_wcc(data->args.inode,
7786 						   data->res.fattr);
7787 		break;
7788 	default:
7789 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7790 			rpc_restart_call_prepare(task);
7791 			return;
7792 		}
7793 	}
7794 }
7795 
7796 static void nfs4_layoutcommit_release(void *calldata)
7797 {
7798 	struct nfs4_layoutcommit_data *data = calldata;
7799 
7800 	pnfs_cleanup_layoutcommit(data);
7801 	put_rpccred(data->cred);
7802 	kfree(data);
7803 }
7804 
7805 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7806 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
7807 	.rpc_call_done = nfs4_layoutcommit_done,
7808 	.rpc_release = nfs4_layoutcommit_release,
7809 };
7810 
7811 int
7812 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7813 {
7814 	struct rpc_message msg = {
7815 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7816 		.rpc_argp = &data->args,
7817 		.rpc_resp = &data->res,
7818 		.rpc_cred = data->cred,
7819 	};
7820 	struct rpc_task_setup task_setup_data = {
7821 		.task = &data->task,
7822 		.rpc_client = NFS_CLIENT(data->args.inode),
7823 		.rpc_message = &msg,
7824 		.callback_ops = &nfs4_layoutcommit_ops,
7825 		.callback_data = data,
7826 		.flags = RPC_TASK_ASYNC,
7827 	};
7828 	struct rpc_task *task;
7829 	int status = 0;
7830 
7831 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7832 		"lbw: %llu inode %lu\n",
7833 		data->task.tk_pid, sync,
7834 		data->args.lastbytewritten,
7835 		data->args.inode->i_ino);
7836 
7837 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7838 	task = rpc_run_task(&task_setup_data);
7839 	if (IS_ERR(task))
7840 		return PTR_ERR(task);
7841 	if (sync == false)
7842 		goto out;
7843 	status = nfs4_wait_for_completion_rpc_task(task);
7844 	if (status != 0)
7845 		goto out;
7846 	status = task->tk_status;
7847 	trace_nfs4_layoutcommit(data->args.inode, status);
7848 out:
7849 	dprintk("%s: status %d\n", __func__, status);
7850 	rpc_put_task(task);
7851 	return status;
7852 }
7853 
7854 /**
7855  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7856  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7857  */
7858 static int
7859 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7860 		    struct nfs_fsinfo *info,
7861 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7862 {
7863 	struct nfs41_secinfo_no_name_args args = {
7864 		.style = SECINFO_STYLE_CURRENT_FH,
7865 	};
7866 	struct nfs4_secinfo_res res = {
7867 		.flavors = flavors,
7868 	};
7869 	struct rpc_message msg = {
7870 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
7871 		.rpc_argp = &args,
7872 		.rpc_resp = &res,
7873 	};
7874 	struct rpc_clnt *clnt = server->client;
7875 	struct rpc_cred *cred = NULL;
7876 	int status;
7877 
7878 	if (use_integrity) {
7879 		clnt = server->nfs_client->cl_rpcclient;
7880 		cred = nfs4_get_clid_cred(server->nfs_client);
7881 		msg.rpc_cred = cred;
7882 	}
7883 
7884 	dprintk("--> %s\n", __func__);
7885 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
7886 				&res.seq_res, 0);
7887 	dprintk("<-- %s status=%d\n", __func__, status);
7888 
7889 	if (cred)
7890 		put_rpccred(cred);
7891 
7892 	return status;
7893 }
7894 
7895 static int
7896 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7897 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
7898 {
7899 	struct nfs4_exception exception = { };
7900 	int err;
7901 	do {
7902 		/* first try using integrity protection */
7903 		err = -NFS4ERR_WRONGSEC;
7904 
7905 		/* try to use integrity protection with machine cred */
7906 		if (_nfs4_is_integrity_protected(server->nfs_client))
7907 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
7908 							  flavors, true);
7909 
7910 		/*
7911 		 * if unable to use integrity protection, or SECINFO with
7912 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
7913 		 * disallowed by spec, but exists in deployed servers) use
7914 		 * the current filesystem's rpc_client and the user cred.
7915 		 */
7916 		if (err == -NFS4ERR_WRONGSEC)
7917 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
7918 							  flavors, false);
7919 
7920 		switch (err) {
7921 		case 0:
7922 		case -NFS4ERR_WRONGSEC:
7923 		case -NFS4ERR_NOTSUPP:
7924 			goto out;
7925 		default:
7926 			err = nfs4_handle_exception(server, err, &exception);
7927 		}
7928 	} while (exception.retry);
7929 out:
7930 	return err;
7931 }
7932 
7933 static int
7934 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
7935 		    struct nfs_fsinfo *info)
7936 {
7937 	int err;
7938 	struct page *page;
7939 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
7940 	struct nfs4_secinfo_flavors *flavors;
7941 	struct nfs4_secinfo4 *secinfo;
7942 	int i;
7943 
7944 	page = alloc_page(GFP_KERNEL);
7945 	if (!page) {
7946 		err = -ENOMEM;
7947 		goto out;
7948 	}
7949 
7950 	flavors = page_address(page);
7951 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
7952 
7953 	/*
7954 	 * Fall back on "guess and check" method if
7955 	 * the server doesn't support SECINFO_NO_NAME
7956 	 */
7957 	if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
7958 		err = nfs4_find_root_sec(server, fhandle, info);
7959 		goto out_freepage;
7960 	}
7961 	if (err)
7962 		goto out_freepage;
7963 
7964 	for (i = 0; i < flavors->num_flavors; i++) {
7965 		secinfo = &flavors->flavors[i];
7966 
7967 		switch (secinfo->flavor) {
7968 		case RPC_AUTH_NULL:
7969 		case RPC_AUTH_UNIX:
7970 		case RPC_AUTH_GSS:
7971 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
7972 					&secinfo->flavor_info);
7973 			break;
7974 		default:
7975 			flavor = RPC_AUTH_MAXFLAVOR;
7976 			break;
7977 		}
7978 
7979 		if (!nfs_auth_info_match(&server->auth_info, flavor))
7980 			flavor = RPC_AUTH_MAXFLAVOR;
7981 
7982 		if (flavor != RPC_AUTH_MAXFLAVOR) {
7983 			err = nfs4_lookup_root_sec(server, fhandle,
7984 						   info, flavor);
7985 			if (!err)
7986 				break;
7987 		}
7988 	}
7989 
7990 	if (flavor == RPC_AUTH_MAXFLAVOR)
7991 		err = -EPERM;
7992 
7993 out_freepage:
7994 	put_page(page);
7995 	if (err == -EACCES)
7996 		return -EPERM;
7997 out:
7998 	return err;
7999 }
8000 
8001 static int _nfs41_test_stateid(struct nfs_server *server,
8002 		nfs4_stateid *stateid,
8003 		struct rpc_cred *cred)
8004 {
8005 	int status;
8006 	struct nfs41_test_stateid_args args = {
8007 		.stateid = stateid,
8008 	};
8009 	struct nfs41_test_stateid_res res;
8010 	struct rpc_message msg = {
8011 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8012 		.rpc_argp = &args,
8013 		.rpc_resp = &res,
8014 		.rpc_cred = cred,
8015 	};
8016 	struct rpc_clnt *rpc_client = server->client;
8017 
8018 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8019 		&rpc_client, &msg);
8020 
8021 	dprintk("NFS call  test_stateid %p\n", stateid);
8022 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8023 	nfs4_set_sequence_privileged(&args.seq_args);
8024 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8025 			&args.seq_args, &res.seq_res);
8026 	if (status != NFS_OK) {
8027 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8028 		return status;
8029 	}
8030 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8031 	return -res.status;
8032 }
8033 
8034 /**
8035  * nfs41_test_stateid - perform a TEST_STATEID operation
8036  *
8037  * @server: server / transport on which to perform the operation
8038  * @stateid: state ID to test
8039  * @cred: credential
8040  *
8041  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8042  * Otherwise a negative NFS4ERR value is returned if the operation
8043  * failed or the state ID is not currently valid.
8044  */
8045 static int nfs41_test_stateid(struct nfs_server *server,
8046 		nfs4_stateid *stateid,
8047 		struct rpc_cred *cred)
8048 {
8049 	struct nfs4_exception exception = { };
8050 	int err;
8051 	do {
8052 		err = _nfs41_test_stateid(server, stateid, cred);
8053 		if (err != -NFS4ERR_DELAY)
8054 			break;
8055 		nfs4_handle_exception(server, err, &exception);
8056 	} while (exception.retry);
8057 	return err;
8058 }
8059 
8060 struct nfs_free_stateid_data {
8061 	struct nfs_server *server;
8062 	struct nfs41_free_stateid_args args;
8063 	struct nfs41_free_stateid_res res;
8064 };
8065 
8066 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8067 {
8068 	struct nfs_free_stateid_data *data = calldata;
8069 	nfs41_setup_sequence(nfs4_get_session(data->server),
8070 			&data->args.seq_args,
8071 			&data->res.seq_res,
8072 			task);
8073 }
8074 
8075 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8076 {
8077 	struct nfs_free_stateid_data *data = calldata;
8078 
8079 	nfs41_sequence_done(task, &data->res.seq_res);
8080 
8081 	switch (task->tk_status) {
8082 	case -NFS4ERR_DELAY:
8083 		if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
8084 			rpc_restart_call_prepare(task);
8085 	}
8086 }
8087 
8088 static void nfs41_free_stateid_release(void *calldata)
8089 {
8090 	kfree(calldata);
8091 }
8092 
8093 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8094 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8095 	.rpc_call_done = nfs41_free_stateid_done,
8096 	.rpc_release = nfs41_free_stateid_release,
8097 };
8098 
8099 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8100 		nfs4_stateid *stateid,
8101 		struct rpc_cred *cred,
8102 		bool privileged)
8103 {
8104 	struct rpc_message msg = {
8105 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8106 		.rpc_cred = cred,
8107 	};
8108 	struct rpc_task_setup task_setup = {
8109 		.rpc_client = server->client,
8110 		.rpc_message = &msg,
8111 		.callback_ops = &nfs41_free_stateid_ops,
8112 		.flags = RPC_TASK_ASYNC,
8113 	};
8114 	struct nfs_free_stateid_data *data;
8115 
8116 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8117 		&task_setup.rpc_client, &msg);
8118 
8119 	dprintk("NFS call  free_stateid %p\n", stateid);
8120 	data = kmalloc(sizeof(*data), GFP_NOFS);
8121 	if (!data)
8122 		return ERR_PTR(-ENOMEM);
8123 	data->server = server;
8124 	nfs4_stateid_copy(&data->args.stateid, stateid);
8125 
8126 	task_setup.callback_data = data;
8127 
8128 	msg.rpc_argp = &data->args;
8129 	msg.rpc_resp = &data->res;
8130 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8131 	if (privileged)
8132 		nfs4_set_sequence_privileged(&data->args.seq_args);
8133 
8134 	return rpc_run_task(&task_setup);
8135 }
8136 
8137 /**
8138  * nfs41_free_stateid - perform a FREE_STATEID operation
8139  *
8140  * @server: server / transport on which to perform the operation
8141  * @stateid: state ID to release
8142  * @cred: credential
8143  *
8144  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8145  * negative NFS4ERR value is returned.
8146  */
8147 static int nfs41_free_stateid(struct nfs_server *server,
8148 		nfs4_stateid *stateid,
8149 		struct rpc_cred *cred)
8150 {
8151 	struct rpc_task *task;
8152 	int ret;
8153 
8154 	task = _nfs41_free_stateid(server, stateid, cred, true);
8155 	if (IS_ERR(task))
8156 		return PTR_ERR(task);
8157 	ret = rpc_wait_for_completion_task(task);
8158 	if (!ret)
8159 		ret = task->tk_status;
8160 	rpc_put_task(task);
8161 	return ret;
8162 }
8163 
8164 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8165 {
8166 	struct rpc_task *task;
8167 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8168 
8169 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8170 	nfs4_free_lock_state(server, lsp);
8171 	if (IS_ERR(task))
8172 		return PTR_ERR(task);
8173 	rpc_put_task(task);
8174 	return 0;
8175 }
8176 
8177 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8178 		const nfs4_stateid *s2)
8179 {
8180 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8181 		return false;
8182 
8183 	if (s1->seqid == s2->seqid)
8184 		return true;
8185 	if (s1->seqid == 0 || s2->seqid == 0)
8186 		return true;
8187 
8188 	return false;
8189 }
8190 
8191 #endif /* CONFIG_NFS_V4_1 */
8192 
8193 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8194 		const nfs4_stateid *s2)
8195 {
8196 	return nfs4_stateid_match(s1, s2);
8197 }
8198 
8199 
8200 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8201 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8202 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8203 	.recover_open	= nfs4_open_reclaim,
8204 	.recover_lock	= nfs4_lock_reclaim,
8205 	.establish_clid = nfs4_init_clientid,
8206 	.detect_trunking = nfs40_discover_server_trunking,
8207 };
8208 
8209 #if defined(CONFIG_NFS_V4_1)
8210 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8211 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8212 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8213 	.recover_open	= nfs4_open_reclaim,
8214 	.recover_lock	= nfs4_lock_reclaim,
8215 	.establish_clid = nfs41_init_clientid,
8216 	.reclaim_complete = nfs41_proc_reclaim_complete,
8217 	.detect_trunking = nfs41_discover_server_trunking,
8218 };
8219 #endif /* CONFIG_NFS_V4_1 */
8220 
8221 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8222 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8223 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8224 	.recover_open	= nfs4_open_expired,
8225 	.recover_lock	= nfs4_lock_expired,
8226 	.establish_clid = nfs4_init_clientid,
8227 };
8228 
8229 #if defined(CONFIG_NFS_V4_1)
8230 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8231 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8232 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8233 	.recover_open	= nfs41_open_expired,
8234 	.recover_lock	= nfs41_lock_expired,
8235 	.establish_clid = nfs41_init_clientid,
8236 };
8237 #endif /* CONFIG_NFS_V4_1 */
8238 
8239 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8240 	.sched_state_renewal = nfs4_proc_async_renew,
8241 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8242 	.renew_lease = nfs4_proc_renew,
8243 };
8244 
8245 #if defined(CONFIG_NFS_V4_1)
8246 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8247 	.sched_state_renewal = nfs41_proc_async_sequence,
8248 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8249 	.renew_lease = nfs4_proc_sequence,
8250 };
8251 #endif
8252 
8253 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8254 	.get_locations = _nfs40_proc_get_locations,
8255 	.fsid_present = _nfs40_proc_fsid_present,
8256 };
8257 
8258 #if defined(CONFIG_NFS_V4_1)
8259 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8260 	.get_locations = _nfs41_proc_get_locations,
8261 	.fsid_present = _nfs41_proc_fsid_present,
8262 };
8263 #endif	/* CONFIG_NFS_V4_1 */
8264 
8265 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8266 	.minor_version = 0,
8267 	.init_caps = NFS_CAP_READDIRPLUS
8268 		| NFS_CAP_ATOMIC_OPEN
8269 		| NFS_CAP_CHANGE_ATTR
8270 		| NFS_CAP_POSIX_LOCK,
8271 	.init_client = nfs40_init_client,
8272 	.shutdown_client = nfs40_shutdown_client,
8273 	.match_stateid = nfs4_match_stateid,
8274 	.find_root_sec = nfs4_find_root_sec,
8275 	.free_lock_state = nfs4_release_lockowner,
8276 	.call_sync_ops = &nfs40_call_sync_ops,
8277 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8278 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8279 	.state_renewal_ops = &nfs40_state_renewal_ops,
8280 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8281 };
8282 
8283 #if defined(CONFIG_NFS_V4_1)
8284 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8285 	.minor_version = 1,
8286 	.init_caps = NFS_CAP_READDIRPLUS
8287 		| NFS_CAP_ATOMIC_OPEN
8288 		| NFS_CAP_CHANGE_ATTR
8289 		| NFS_CAP_POSIX_LOCK
8290 		| NFS_CAP_STATEID_NFSV41
8291 		| NFS_CAP_ATOMIC_OPEN_V1,
8292 	.init_client = nfs41_init_client,
8293 	.shutdown_client = nfs41_shutdown_client,
8294 	.match_stateid = nfs41_match_stateid,
8295 	.find_root_sec = nfs41_find_root_sec,
8296 	.free_lock_state = nfs41_free_lock_state,
8297 	.call_sync_ops = &nfs41_call_sync_ops,
8298 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8299 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8300 	.state_renewal_ops = &nfs41_state_renewal_ops,
8301 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8302 };
8303 #endif
8304 
8305 #if defined(CONFIG_NFS_V4_2)
8306 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8307 	.minor_version = 2,
8308 	.init_caps = NFS_CAP_READDIRPLUS
8309 		| NFS_CAP_ATOMIC_OPEN
8310 		| NFS_CAP_CHANGE_ATTR
8311 		| NFS_CAP_POSIX_LOCK
8312 		| NFS_CAP_STATEID_NFSV41
8313 		| NFS_CAP_ATOMIC_OPEN_V1,
8314 	.init_client = nfs41_init_client,
8315 	.shutdown_client = nfs41_shutdown_client,
8316 	.match_stateid = nfs41_match_stateid,
8317 	.find_root_sec = nfs41_find_root_sec,
8318 	.free_lock_state = nfs41_free_lock_state,
8319 	.call_sync_ops = &nfs41_call_sync_ops,
8320 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8321 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8322 	.state_renewal_ops = &nfs41_state_renewal_ops,
8323 };
8324 #endif
8325 
8326 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8327 	[0] = &nfs_v4_0_minor_ops,
8328 #if defined(CONFIG_NFS_V4_1)
8329 	[1] = &nfs_v4_1_minor_ops,
8330 #endif
8331 #if defined(CONFIG_NFS_V4_2)
8332 	[2] = &nfs_v4_2_minor_ops,
8333 #endif
8334 };
8335 
8336 static const struct inode_operations nfs4_dir_inode_operations = {
8337 	.create		= nfs_create,
8338 	.lookup		= nfs_lookup,
8339 	.atomic_open	= nfs_atomic_open,
8340 	.link		= nfs_link,
8341 	.unlink		= nfs_unlink,
8342 	.symlink	= nfs_symlink,
8343 	.mkdir		= nfs_mkdir,
8344 	.rmdir		= nfs_rmdir,
8345 	.mknod		= nfs_mknod,
8346 	.rename		= nfs_rename,
8347 	.permission	= nfs_permission,
8348 	.getattr	= nfs_getattr,
8349 	.setattr	= nfs_setattr,
8350 	.getxattr	= generic_getxattr,
8351 	.setxattr	= generic_setxattr,
8352 	.listxattr	= generic_listxattr,
8353 	.removexattr	= generic_removexattr,
8354 };
8355 
8356 static const struct inode_operations nfs4_file_inode_operations = {
8357 	.permission	= nfs_permission,
8358 	.getattr	= nfs_getattr,
8359 	.setattr	= nfs_setattr,
8360 	.getxattr	= generic_getxattr,
8361 	.setxattr	= generic_setxattr,
8362 	.listxattr	= generic_listxattr,
8363 	.removexattr	= generic_removexattr,
8364 };
8365 
8366 const struct nfs_rpc_ops nfs_v4_clientops = {
8367 	.version	= 4,			/* protocol version */
8368 	.dentry_ops	= &nfs4_dentry_operations,
8369 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8370 	.file_inode_ops	= &nfs4_file_inode_operations,
8371 	.file_ops	= &nfs4_file_operations,
8372 	.getroot	= nfs4_proc_get_root,
8373 	.submount	= nfs4_submount,
8374 	.try_mount	= nfs4_try_mount,
8375 	.getattr	= nfs4_proc_getattr,
8376 	.setattr	= nfs4_proc_setattr,
8377 	.lookup		= nfs4_proc_lookup,
8378 	.access		= nfs4_proc_access,
8379 	.readlink	= nfs4_proc_readlink,
8380 	.create		= nfs4_proc_create,
8381 	.remove		= nfs4_proc_remove,
8382 	.unlink_setup	= nfs4_proc_unlink_setup,
8383 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8384 	.unlink_done	= nfs4_proc_unlink_done,
8385 	.rename		= nfs4_proc_rename,
8386 	.rename_setup	= nfs4_proc_rename_setup,
8387 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8388 	.rename_done	= nfs4_proc_rename_done,
8389 	.link		= nfs4_proc_link,
8390 	.symlink	= nfs4_proc_symlink,
8391 	.mkdir		= nfs4_proc_mkdir,
8392 	.rmdir		= nfs4_proc_remove,
8393 	.readdir	= nfs4_proc_readdir,
8394 	.mknod		= nfs4_proc_mknod,
8395 	.statfs		= nfs4_proc_statfs,
8396 	.fsinfo		= nfs4_proc_fsinfo,
8397 	.pathconf	= nfs4_proc_pathconf,
8398 	.set_capabilities = nfs4_server_capabilities,
8399 	.decode_dirent	= nfs4_decode_dirent,
8400 	.read_setup	= nfs4_proc_read_setup,
8401 	.read_pageio_init = pnfs_pageio_init_read,
8402 	.read_rpc_prepare = nfs4_proc_read_rpc_prepare,
8403 	.read_done	= nfs4_read_done,
8404 	.write_setup	= nfs4_proc_write_setup,
8405 	.write_pageio_init = pnfs_pageio_init_write,
8406 	.write_rpc_prepare = nfs4_proc_write_rpc_prepare,
8407 	.write_done	= nfs4_write_done,
8408 	.commit_setup	= nfs4_proc_commit_setup,
8409 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8410 	.commit_done	= nfs4_commit_done,
8411 	.lock		= nfs4_proc_lock,
8412 	.clear_acl_cache = nfs4_zap_acl_attr,
8413 	.close_context  = nfs4_close_context,
8414 	.open_context	= nfs4_atomic_open,
8415 	.have_delegation = nfs4_have_delegation,
8416 	.return_delegation = nfs4_inode_return_delegation,
8417 	.alloc_client	= nfs4_alloc_client,
8418 	.init_client	= nfs4_init_client,
8419 	.free_client	= nfs4_free_client,
8420 	.create_server	= nfs4_create_server,
8421 	.clone_server	= nfs_clone_server,
8422 };
8423 
8424 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8425 	.prefix	= XATTR_NAME_NFSV4_ACL,
8426 	.list	= nfs4_xattr_list_nfs4_acl,
8427 	.get	= nfs4_xattr_get_nfs4_acl,
8428 	.set	= nfs4_xattr_set_nfs4_acl,
8429 };
8430 
8431 const struct xattr_handler *nfs4_xattr_handlers[] = {
8432 	&nfs4_xattr_nfs4_acl_handler,
8433 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8434 	&nfs4_xattr_nfs4_label_handler,
8435 #endif
8436 	NULL
8437 };
8438 
8439 /*
8440  * Local variables:
8441  *  c-basic-offset: 8
8442  * End:
8443  */
8444