xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 867a0e05)
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 *, long *);
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_FILE_OPEN:
162 		return -EBUSY;
163 	default:
164 		dprintk("%s could not handle NFSv4 error %d\n",
165 				__func__, -err);
166 		break;
167 	}
168 	return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175 	FATTR4_WORD0_TYPE
176 	| FATTR4_WORD0_CHANGE
177 	| FATTR4_WORD0_SIZE
178 	| FATTR4_WORD0_FSID
179 	| FATTR4_WORD0_FILEID,
180 	FATTR4_WORD1_MODE
181 	| FATTR4_WORD1_NUMLINKS
182 	| FATTR4_WORD1_OWNER
183 	| FATTR4_WORD1_OWNER_GROUP
184 	| FATTR4_WORD1_RAWDEV
185 	| FATTR4_WORD1_SPACE_USED
186 	| FATTR4_WORD1_TIME_ACCESS
187 	| FATTR4_WORD1_TIME_METADATA
188 	| FATTR4_WORD1_TIME_MODIFY,
189 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
190 	FATTR4_WORD2_SECURITY_LABEL
191 #endif
192 };
193 
194 static const u32 nfs4_pnfs_open_bitmap[3] = {
195 	FATTR4_WORD0_TYPE
196 	| FATTR4_WORD0_CHANGE
197 	| FATTR4_WORD0_SIZE
198 	| FATTR4_WORD0_FSID
199 	| FATTR4_WORD0_FILEID,
200 	FATTR4_WORD1_MODE
201 	| FATTR4_WORD1_NUMLINKS
202 	| FATTR4_WORD1_OWNER
203 	| FATTR4_WORD1_OWNER_GROUP
204 	| FATTR4_WORD1_RAWDEV
205 	| FATTR4_WORD1_SPACE_USED
206 	| FATTR4_WORD1_TIME_ACCESS
207 	| FATTR4_WORD1_TIME_METADATA
208 	| FATTR4_WORD1_TIME_MODIFY,
209 	FATTR4_WORD2_MDSTHRESHOLD
210 };
211 
212 static const u32 nfs4_open_noattr_bitmap[3] = {
213 	FATTR4_WORD0_TYPE
214 	| FATTR4_WORD0_CHANGE
215 	| FATTR4_WORD0_FILEID,
216 };
217 
218 const u32 nfs4_statfs_bitmap[3] = {
219 	FATTR4_WORD0_FILES_AVAIL
220 	| FATTR4_WORD0_FILES_FREE
221 	| FATTR4_WORD0_FILES_TOTAL,
222 	FATTR4_WORD1_SPACE_AVAIL
223 	| FATTR4_WORD1_SPACE_FREE
224 	| FATTR4_WORD1_SPACE_TOTAL
225 };
226 
227 const u32 nfs4_pathconf_bitmap[3] = {
228 	FATTR4_WORD0_MAXLINK
229 	| FATTR4_WORD0_MAXNAME,
230 	0
231 };
232 
233 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
234 			| FATTR4_WORD0_MAXREAD
235 			| FATTR4_WORD0_MAXWRITE
236 			| FATTR4_WORD0_LEASE_TIME,
237 			FATTR4_WORD1_TIME_DELTA
238 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
239 			FATTR4_WORD2_LAYOUT_BLKSIZE
240 };
241 
242 const u32 nfs4_fs_locations_bitmap[3] = {
243 	FATTR4_WORD0_TYPE
244 	| FATTR4_WORD0_CHANGE
245 	| FATTR4_WORD0_SIZE
246 	| FATTR4_WORD0_FSID
247 	| FATTR4_WORD0_FILEID
248 	| FATTR4_WORD0_FS_LOCATIONS,
249 	FATTR4_WORD1_MODE
250 	| FATTR4_WORD1_NUMLINKS
251 	| FATTR4_WORD1_OWNER
252 	| FATTR4_WORD1_OWNER_GROUP
253 	| FATTR4_WORD1_RAWDEV
254 	| FATTR4_WORD1_SPACE_USED
255 	| FATTR4_WORD1_TIME_ACCESS
256 	| FATTR4_WORD1_TIME_METADATA
257 	| FATTR4_WORD1_TIME_MODIFY
258 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
259 };
260 
261 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
262 		struct nfs4_readdir_arg *readdir)
263 {
264 	__be32 *start, *p;
265 
266 	if (cookie > 2) {
267 		readdir->cookie = cookie;
268 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
269 		return;
270 	}
271 
272 	readdir->cookie = 0;
273 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
274 	if (cookie == 2)
275 		return;
276 
277 	/*
278 	 * NFSv4 servers do not return entries for '.' and '..'
279 	 * Therefore, we fake these entries here.  We let '.'
280 	 * have cookie 0 and '..' have cookie 1.  Note that
281 	 * when talking to the server, we always send cookie 0
282 	 * instead of 1 or 2.
283 	 */
284 	start = p = kmap_atomic(*readdir->pages);
285 
286 	if (cookie == 0) {
287 		*p++ = xdr_one;                                  /* next */
288 		*p++ = xdr_zero;                   /* cookie, first word */
289 		*p++ = xdr_one;                   /* cookie, second word */
290 		*p++ = xdr_one;                             /* entry len */
291 		memcpy(p, ".\0\0\0", 4);                        /* entry */
292 		p++;
293 		*p++ = xdr_one;                         /* bitmap length */
294 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
295 		*p++ = htonl(8);              /* attribute buffer length */
296 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
297 	}
298 
299 	*p++ = xdr_one;                                  /* next */
300 	*p++ = xdr_zero;                   /* cookie, first word */
301 	*p++ = xdr_two;                   /* cookie, second word */
302 	*p++ = xdr_two;                             /* entry len */
303 	memcpy(p, "..\0\0", 4);                         /* entry */
304 	p++;
305 	*p++ = xdr_one;                         /* bitmap length */
306 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
307 	*p++ = htonl(8);              /* attribute buffer length */
308 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
309 
310 	readdir->pgbase = (char *)p - (char *)start;
311 	readdir->count -= readdir->pgbase;
312 	kunmap_atomic(start);
313 }
314 
315 static long nfs4_update_delay(long *timeout)
316 {
317 	long ret;
318 	if (!timeout)
319 		return NFS4_POLL_RETRY_MAX;
320 	if (*timeout <= 0)
321 		*timeout = NFS4_POLL_RETRY_MIN;
322 	if (*timeout > NFS4_POLL_RETRY_MAX)
323 		*timeout = NFS4_POLL_RETRY_MAX;
324 	ret = *timeout;
325 	*timeout <<= 1;
326 	return ret;
327 }
328 
329 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
330 {
331 	int res = 0;
332 
333 	might_sleep();
334 
335 	freezable_schedule_timeout_killable_unsafe(
336 		nfs4_update_delay(timeout));
337 	if (fatal_signal_pending(current))
338 		res = -ERESTARTSYS;
339 	return res;
340 }
341 
342 /* This is the error handling routine for processes that are allowed
343  * to sleep.
344  */
345 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
346 {
347 	struct nfs_client *clp = server->nfs_client;
348 	struct nfs4_state *state = exception->state;
349 	struct inode *inode = exception->inode;
350 	int ret = errorcode;
351 
352 	exception->retry = 0;
353 	switch(errorcode) {
354 		case 0:
355 			return 0;
356 		case -NFS4ERR_OPENMODE:
357 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
358 				nfs4_inode_return_delegation(inode);
359 				exception->retry = 1;
360 				return 0;
361 			}
362 			if (state == NULL)
363 				break;
364 			ret = nfs4_schedule_stateid_recovery(server, state);
365 			if (ret < 0)
366 				break;
367 			goto wait_on_recovery;
368 		case -NFS4ERR_DELEG_REVOKED:
369 		case -NFS4ERR_ADMIN_REVOKED:
370 		case -NFS4ERR_BAD_STATEID:
371 			if (state == NULL)
372 				break;
373 			ret = nfs4_schedule_stateid_recovery(server, state);
374 			if (ret < 0)
375 				break;
376 			goto wait_on_recovery;
377 		case -NFS4ERR_EXPIRED:
378 			if (state != NULL) {
379 				ret = nfs4_schedule_stateid_recovery(server, state);
380 				if (ret < 0)
381 					break;
382 			}
383 		case -NFS4ERR_STALE_STATEID:
384 		case -NFS4ERR_STALE_CLIENTID:
385 			nfs4_schedule_lease_recovery(clp);
386 			goto wait_on_recovery;
387 		case -NFS4ERR_MOVED:
388 			ret = nfs4_schedule_migration_recovery(server);
389 			if (ret < 0)
390 				break;
391 			goto wait_on_recovery;
392 		case -NFS4ERR_LEASE_MOVED:
393 			nfs4_schedule_lease_moved_recovery(clp);
394 			goto wait_on_recovery;
395 #if defined(CONFIG_NFS_V4_1)
396 		case -NFS4ERR_BADSESSION:
397 		case -NFS4ERR_BADSLOT:
398 		case -NFS4ERR_BAD_HIGH_SLOT:
399 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
400 		case -NFS4ERR_DEADSESSION:
401 		case -NFS4ERR_SEQ_FALSE_RETRY:
402 		case -NFS4ERR_SEQ_MISORDERED:
403 			dprintk("%s ERROR: %d Reset session\n", __func__,
404 				errorcode);
405 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
406 			goto wait_on_recovery;
407 #endif /* defined(CONFIG_NFS_V4_1) */
408 		case -NFS4ERR_FILE_OPEN:
409 			if (exception->timeout > HZ) {
410 				/* We have retried a decent amount, time to
411 				 * fail
412 				 */
413 				ret = -EBUSY;
414 				break;
415 			}
416 		case -NFS4ERR_GRACE:
417 		case -NFS4ERR_DELAY:
418 			ret = nfs4_delay(server->client, &exception->timeout);
419 			if (ret != 0)
420 				break;
421 		case -NFS4ERR_RETRY_UNCACHED_REP:
422 		case -NFS4ERR_OLD_STATEID:
423 			exception->retry = 1;
424 			break;
425 		case -NFS4ERR_BADOWNER:
426 			/* The following works around a Linux server bug! */
427 		case -NFS4ERR_BADNAME:
428 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
429 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
430 				exception->retry = 1;
431 				printk(KERN_WARNING "NFS: v4 server %s "
432 						"does not accept raw "
433 						"uid/gids. "
434 						"Reenabling the idmapper.\n",
435 						server->nfs_client->cl_hostname);
436 			}
437 	}
438 	/* We failed to handle the error */
439 	return nfs4_map_errors(ret);
440 wait_on_recovery:
441 	ret = nfs4_wait_clnt_recover(clp);
442 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
443 		return -EIO;
444 	if (ret == 0)
445 		exception->retry = 1;
446 	return ret;
447 }
448 
449 /*
450  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
451  * or 'false' otherwise.
452  */
453 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
454 {
455 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
456 
457 	if (flavor == RPC_AUTH_GSS_KRB5I ||
458 	    flavor == RPC_AUTH_GSS_KRB5P)
459 		return true;
460 
461 	return false;
462 }
463 
464 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
465 {
466 	spin_lock(&clp->cl_lock);
467 	if (time_before(clp->cl_last_renewal,timestamp))
468 		clp->cl_last_renewal = timestamp;
469 	spin_unlock(&clp->cl_lock);
470 }
471 
472 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
473 {
474 	do_renew_lease(server->nfs_client, timestamp);
475 }
476 
477 struct nfs4_call_sync_data {
478 	const struct nfs_server *seq_server;
479 	struct nfs4_sequence_args *seq_args;
480 	struct nfs4_sequence_res *seq_res;
481 };
482 
483 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
484 			       struct nfs4_sequence_res *res, int cache_reply)
485 {
486 	args->sa_slot = NULL;
487 	args->sa_cache_this = cache_reply;
488 	args->sa_privileged = 0;
489 
490 	res->sr_slot = NULL;
491 }
492 
493 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
494 {
495 	args->sa_privileged = 1;
496 }
497 
498 static int nfs40_setup_sequence(const struct nfs_server *server,
499 				struct nfs4_sequence_args *args,
500 				struct nfs4_sequence_res *res,
501 				struct rpc_task *task)
502 {
503 	struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
504 	struct nfs4_slot *slot;
505 
506 	/* slot already allocated? */
507 	if (res->sr_slot != NULL)
508 		goto out_start;
509 
510 	spin_lock(&tbl->slot_tbl_lock);
511 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
512 		goto out_sleep;
513 
514 	slot = nfs4_alloc_slot(tbl);
515 	if (IS_ERR(slot)) {
516 		if (slot == ERR_PTR(-ENOMEM))
517 			task->tk_timeout = HZ >> 2;
518 		goto out_sleep;
519 	}
520 	spin_unlock(&tbl->slot_tbl_lock);
521 
522 	args->sa_slot = slot;
523 	res->sr_slot = slot;
524 
525 out_start:
526 	rpc_call_start(task);
527 	return 0;
528 
529 out_sleep:
530 	if (args->sa_privileged)
531 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
532 				NULL, RPC_PRIORITY_PRIVILEGED);
533 	else
534 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
535 	spin_unlock(&tbl->slot_tbl_lock);
536 	return -EAGAIN;
537 }
538 
539 static int nfs40_sequence_done(struct rpc_task *task,
540 			       struct nfs4_sequence_res *res)
541 {
542 	struct nfs4_slot *slot = res->sr_slot;
543 	struct nfs4_slot_table *tbl;
544 
545 	if (slot == NULL)
546 		goto out;
547 
548 	tbl = slot->table;
549 	spin_lock(&tbl->slot_tbl_lock);
550 	if (!nfs41_wake_and_assign_slot(tbl, slot))
551 		nfs4_free_slot(tbl, slot);
552 	spin_unlock(&tbl->slot_tbl_lock);
553 
554 	res->sr_slot = NULL;
555 out:
556 	return 1;
557 }
558 
559 #if defined(CONFIG_NFS_V4_1)
560 
561 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
562 {
563 	struct nfs4_session *session;
564 	struct nfs4_slot_table *tbl;
565 	struct nfs4_slot *slot = res->sr_slot;
566 	bool send_new_highest_used_slotid = false;
567 
568 	tbl = slot->table;
569 	session = tbl->session;
570 
571 	spin_lock(&tbl->slot_tbl_lock);
572 	/* Be nice to the server: try to ensure that the last transmitted
573 	 * value for highest_user_slotid <= target_highest_slotid
574 	 */
575 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
576 		send_new_highest_used_slotid = true;
577 
578 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
579 		send_new_highest_used_slotid = false;
580 		goto out_unlock;
581 	}
582 	nfs4_free_slot(tbl, slot);
583 
584 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
585 		send_new_highest_used_slotid = false;
586 out_unlock:
587 	spin_unlock(&tbl->slot_tbl_lock);
588 	res->sr_slot = NULL;
589 	if (send_new_highest_used_slotid)
590 		nfs41_server_notify_highest_slotid_update(session->clp);
591 }
592 
593 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
594 {
595 	struct nfs4_session *session;
596 	struct nfs4_slot *slot = res->sr_slot;
597 	struct nfs_client *clp;
598 	bool interrupted = false;
599 	int ret = 1;
600 
601 	if (slot == NULL)
602 		goto out_noaction;
603 	/* don't increment the sequence number if the task wasn't sent */
604 	if (!RPC_WAS_SENT(task))
605 		goto out;
606 
607 	session = slot->table->session;
608 
609 	if (slot->interrupted) {
610 		slot->interrupted = 0;
611 		interrupted = true;
612 	}
613 
614 	trace_nfs4_sequence_done(session, res);
615 	/* Check the SEQUENCE operation status */
616 	switch (res->sr_status) {
617 	case 0:
618 		/* Update the slot's sequence and clientid lease timer */
619 		++slot->seq_nr;
620 		clp = session->clp;
621 		do_renew_lease(clp, res->sr_timestamp);
622 		/* Check sequence flags */
623 		if (res->sr_status_flags != 0)
624 			nfs4_schedule_lease_recovery(clp);
625 		nfs41_update_target_slotid(slot->table, slot, res);
626 		break;
627 	case 1:
628 		/*
629 		 * sr_status remains 1 if an RPC level error occurred.
630 		 * The server may or may not have processed the sequence
631 		 * operation..
632 		 * Mark the slot as having hosted an interrupted RPC call.
633 		 */
634 		slot->interrupted = 1;
635 		goto out;
636 	case -NFS4ERR_DELAY:
637 		/* The server detected a resend of the RPC call and
638 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
639 		 * of RFC5661.
640 		 */
641 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
642 			__func__,
643 			slot->slot_nr,
644 			slot->seq_nr);
645 		goto out_retry;
646 	case -NFS4ERR_BADSLOT:
647 		/*
648 		 * The slot id we used was probably retired. Try again
649 		 * using a different slot id.
650 		 */
651 		goto retry_nowait;
652 	case -NFS4ERR_SEQ_MISORDERED:
653 		/*
654 		 * Was the last operation on this sequence interrupted?
655 		 * If so, retry after bumping the sequence number.
656 		 */
657 		if (interrupted) {
658 			++slot->seq_nr;
659 			goto retry_nowait;
660 		}
661 		/*
662 		 * Could this slot have been previously retired?
663 		 * If so, then the server may be expecting seq_nr = 1!
664 		 */
665 		if (slot->seq_nr != 1) {
666 			slot->seq_nr = 1;
667 			goto retry_nowait;
668 		}
669 		break;
670 	case -NFS4ERR_SEQ_FALSE_RETRY:
671 		++slot->seq_nr;
672 		goto retry_nowait;
673 	default:
674 		/* Just update the slot sequence no. */
675 		++slot->seq_nr;
676 	}
677 out:
678 	/* The session may be reset by one of the error handlers. */
679 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
680 	nfs41_sequence_free_slot(res);
681 out_noaction:
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 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
696 
697 static int nfs4_sequence_done(struct rpc_task *task,
698 			       struct nfs4_sequence_res *res)
699 {
700 	if (res->sr_slot == NULL)
701 		return 1;
702 	if (!res->sr_slot->table->session)
703 		return nfs40_sequence_done(task, res);
704 	return nfs41_sequence_done(task, res);
705 }
706 
707 int nfs41_setup_sequence(struct nfs4_session *session,
708 				struct nfs4_sequence_args *args,
709 				struct nfs4_sequence_res *res,
710 				struct rpc_task *task)
711 {
712 	struct nfs4_slot *slot;
713 	struct nfs4_slot_table *tbl;
714 
715 	dprintk("--> %s\n", __func__);
716 	/* slot already allocated? */
717 	if (res->sr_slot != NULL)
718 		goto out_success;
719 
720 	tbl = &session->fc_slot_table;
721 
722 	task->tk_timeout = 0;
723 
724 	spin_lock(&tbl->slot_tbl_lock);
725 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
726 	    !args->sa_privileged) {
727 		/* The state manager will wait until the slot table is empty */
728 		dprintk("%s session is draining\n", __func__);
729 		goto out_sleep;
730 	}
731 
732 	slot = nfs4_alloc_slot(tbl);
733 	if (IS_ERR(slot)) {
734 		/* If out of memory, try again in 1/4 second */
735 		if (slot == ERR_PTR(-ENOMEM))
736 			task->tk_timeout = HZ >> 2;
737 		dprintk("<-- %s: no free slots\n", __func__);
738 		goto out_sleep;
739 	}
740 	spin_unlock(&tbl->slot_tbl_lock);
741 
742 	args->sa_slot = slot;
743 
744 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
745 			slot->slot_nr, slot->seq_nr);
746 
747 	res->sr_slot = slot;
748 	res->sr_timestamp = jiffies;
749 	res->sr_status_flags = 0;
750 	/*
751 	 * sr_status is only set in decode_sequence, and so will remain
752 	 * set to 1 if an rpc level failure occurs.
753 	 */
754 	res->sr_status = 1;
755 	trace_nfs4_setup_sequence(session, args);
756 out_success:
757 	rpc_call_start(task);
758 	return 0;
759 out_sleep:
760 	/* Privileged tasks are queued with top priority */
761 	if (args->sa_privileged)
762 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
763 				NULL, RPC_PRIORITY_PRIVILEGED);
764 	else
765 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
766 	spin_unlock(&tbl->slot_tbl_lock);
767 	return -EAGAIN;
768 }
769 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
770 
771 static int nfs4_setup_sequence(const struct nfs_server *server,
772 			       struct nfs4_sequence_args *args,
773 			       struct nfs4_sequence_res *res,
774 			       struct rpc_task *task)
775 {
776 	struct nfs4_session *session = nfs4_get_session(server);
777 	int ret = 0;
778 
779 	if (!session)
780 		return nfs40_setup_sequence(server, args, res, task);
781 
782 	dprintk("--> %s clp %p session %p sr_slot %u\n",
783 		__func__, session->clp, session, res->sr_slot ?
784 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
785 
786 	ret = nfs41_setup_sequence(session, args, res, task);
787 
788 	dprintk("<-- %s status=%d\n", __func__, ret);
789 	return ret;
790 }
791 
792 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
793 {
794 	struct nfs4_call_sync_data *data = calldata;
795 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
796 
797 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
798 
799 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
800 }
801 
802 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
803 {
804 	struct nfs4_call_sync_data *data = calldata;
805 
806 	nfs41_sequence_done(task, data->seq_res);
807 }
808 
809 static const struct rpc_call_ops nfs41_call_sync_ops = {
810 	.rpc_call_prepare = nfs41_call_sync_prepare,
811 	.rpc_call_done = nfs41_call_sync_done,
812 };
813 
814 #else	/* !CONFIG_NFS_V4_1 */
815 
816 static int nfs4_setup_sequence(const struct nfs_server *server,
817 			       struct nfs4_sequence_args *args,
818 			       struct nfs4_sequence_res *res,
819 			       struct rpc_task *task)
820 {
821 	return nfs40_setup_sequence(server, args, res, task);
822 }
823 
824 static int nfs4_sequence_done(struct rpc_task *task,
825 			       struct nfs4_sequence_res *res)
826 {
827 	return nfs40_sequence_done(task, res);
828 }
829 
830 #endif	/* !CONFIG_NFS_V4_1 */
831 
832 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
833 {
834 	struct nfs4_call_sync_data *data = calldata;
835 	nfs4_setup_sequence(data->seq_server,
836 				data->seq_args, data->seq_res, task);
837 }
838 
839 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
840 {
841 	struct nfs4_call_sync_data *data = calldata;
842 	nfs4_sequence_done(task, data->seq_res);
843 }
844 
845 static const struct rpc_call_ops nfs40_call_sync_ops = {
846 	.rpc_call_prepare = nfs40_call_sync_prepare,
847 	.rpc_call_done = nfs40_call_sync_done,
848 };
849 
850 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
851 				   struct nfs_server *server,
852 				   struct rpc_message *msg,
853 				   struct nfs4_sequence_args *args,
854 				   struct nfs4_sequence_res *res)
855 {
856 	int ret;
857 	struct rpc_task *task;
858 	struct nfs_client *clp = server->nfs_client;
859 	struct nfs4_call_sync_data data = {
860 		.seq_server = server,
861 		.seq_args = args,
862 		.seq_res = res,
863 	};
864 	struct rpc_task_setup task_setup = {
865 		.rpc_client = clnt,
866 		.rpc_message = msg,
867 		.callback_ops = clp->cl_mvops->call_sync_ops,
868 		.callback_data = &data
869 	};
870 
871 	task = rpc_run_task(&task_setup);
872 	if (IS_ERR(task))
873 		ret = PTR_ERR(task);
874 	else {
875 		ret = task->tk_status;
876 		rpc_put_task(task);
877 	}
878 	return ret;
879 }
880 
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->f_attr.mdsthreshold);
1074 	kfree(p);
1075 }
1076 
1077 static void nfs4_opendata_put(struct nfs4_opendata *p)
1078 {
1079 	if (p != NULL)
1080 		kref_put(&p->kref, nfs4_opendata_free);
1081 }
1082 
1083 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1084 {
1085 	int ret;
1086 
1087 	ret = rpc_wait_for_completion_task(task);
1088 	return ret;
1089 }
1090 
1091 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1092 {
1093 	int ret = 0;
1094 
1095 	if (open_mode & (O_EXCL|O_TRUNC))
1096 		goto out;
1097 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1098 		case FMODE_READ:
1099 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1100 				&& state->n_rdonly != 0;
1101 			break;
1102 		case FMODE_WRITE:
1103 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1104 				&& state->n_wronly != 0;
1105 			break;
1106 		case FMODE_READ|FMODE_WRITE:
1107 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1108 				&& state->n_rdwr != 0;
1109 	}
1110 out:
1111 	return ret;
1112 }
1113 
1114 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1115 {
1116 	if (delegation == NULL)
1117 		return 0;
1118 	if ((delegation->type & fmode) != fmode)
1119 		return 0;
1120 	if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1121 		return 0;
1122 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1123 		return 0;
1124 	nfs_mark_delegation_referenced(delegation);
1125 	return 1;
1126 }
1127 
1128 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1129 {
1130 	switch (fmode) {
1131 		case FMODE_WRITE:
1132 			state->n_wronly++;
1133 			break;
1134 		case FMODE_READ:
1135 			state->n_rdonly++;
1136 			break;
1137 		case FMODE_READ|FMODE_WRITE:
1138 			state->n_rdwr++;
1139 	}
1140 	nfs4_state_set_mode_locked(state, state->state | fmode);
1141 }
1142 
1143 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1144 {
1145 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1146 	bool need_recover = false;
1147 
1148 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1149 		need_recover = true;
1150 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1151 		need_recover = true;
1152 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1153 		need_recover = true;
1154 	if (need_recover)
1155 		nfs4_state_mark_reclaim_nograce(clp, state);
1156 }
1157 
1158 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1159 		nfs4_stateid *stateid)
1160 {
1161 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1162 		return true;
1163 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1164 		nfs_test_and_clear_all_open_stateid(state);
1165 		return true;
1166 	}
1167 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1168 		return true;
1169 	return false;
1170 }
1171 
1172 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1173 		nfs4_stateid *stateid, fmode_t fmode)
1174 {
1175 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1176 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1177 	case FMODE_WRITE:
1178 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1179 		break;
1180 	case FMODE_READ:
1181 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1182 		break;
1183 	case 0:
1184 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1185 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1186 		clear_bit(NFS_OPEN_STATE, &state->flags);
1187 	}
1188 	if (stateid == NULL)
1189 		return;
1190 	if (!nfs_need_update_open_stateid(state, stateid))
1191 		return;
1192 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1193 		nfs4_stateid_copy(&state->stateid, stateid);
1194 	nfs4_stateid_copy(&state->open_stateid, stateid);
1195 }
1196 
1197 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1198 {
1199 	write_seqlock(&state->seqlock);
1200 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1201 	write_sequnlock(&state->seqlock);
1202 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1203 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1204 }
1205 
1206 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1207 {
1208 	switch (fmode) {
1209 		case FMODE_READ:
1210 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1211 			break;
1212 		case FMODE_WRITE:
1213 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1214 			break;
1215 		case FMODE_READ|FMODE_WRITE:
1216 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1217 	}
1218 	if (!nfs_need_update_open_stateid(state, stateid))
1219 		return;
1220 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1221 		nfs4_stateid_copy(&state->stateid, stateid);
1222 	nfs4_stateid_copy(&state->open_stateid, stateid);
1223 }
1224 
1225 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1226 {
1227 	/*
1228 	 * Protect the call to nfs4_state_set_mode_locked and
1229 	 * serialise the stateid update
1230 	 */
1231 	write_seqlock(&state->seqlock);
1232 	if (deleg_stateid != NULL) {
1233 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1234 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1235 	}
1236 	if (open_stateid != NULL)
1237 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1238 	write_sequnlock(&state->seqlock);
1239 	spin_lock(&state->owner->so_lock);
1240 	update_open_stateflags(state, fmode);
1241 	spin_unlock(&state->owner->so_lock);
1242 }
1243 
1244 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1245 {
1246 	struct nfs_inode *nfsi = NFS_I(state->inode);
1247 	struct nfs_delegation *deleg_cur;
1248 	int ret = 0;
1249 
1250 	fmode &= (FMODE_READ|FMODE_WRITE);
1251 
1252 	rcu_read_lock();
1253 	deleg_cur = rcu_dereference(nfsi->delegation);
1254 	if (deleg_cur == NULL)
1255 		goto no_delegation;
1256 
1257 	spin_lock(&deleg_cur->lock);
1258 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1259 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1260 	    (deleg_cur->type & fmode) != fmode)
1261 		goto no_delegation_unlock;
1262 
1263 	if (delegation == NULL)
1264 		delegation = &deleg_cur->stateid;
1265 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1266 		goto no_delegation_unlock;
1267 
1268 	nfs_mark_delegation_referenced(deleg_cur);
1269 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1270 	ret = 1;
1271 no_delegation_unlock:
1272 	spin_unlock(&deleg_cur->lock);
1273 no_delegation:
1274 	rcu_read_unlock();
1275 
1276 	if (!ret && open_stateid != NULL) {
1277 		__update_open_stateid(state, open_stateid, NULL, fmode);
1278 		ret = 1;
1279 	}
1280 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1281 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1282 
1283 	return ret;
1284 }
1285 
1286 
1287 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1288 {
1289 	struct nfs_delegation *delegation;
1290 
1291 	rcu_read_lock();
1292 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1293 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1294 		rcu_read_unlock();
1295 		return;
1296 	}
1297 	rcu_read_unlock();
1298 	nfs4_inode_return_delegation(inode);
1299 }
1300 
1301 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1302 {
1303 	struct nfs4_state *state = opendata->state;
1304 	struct nfs_inode *nfsi = NFS_I(state->inode);
1305 	struct nfs_delegation *delegation;
1306 	int open_mode = opendata->o_arg.open_flags;
1307 	fmode_t fmode = opendata->o_arg.fmode;
1308 	nfs4_stateid stateid;
1309 	int ret = -EAGAIN;
1310 
1311 	for (;;) {
1312 		spin_lock(&state->owner->so_lock);
1313 		if (can_open_cached(state, fmode, open_mode)) {
1314 			update_open_stateflags(state, fmode);
1315 			spin_unlock(&state->owner->so_lock);
1316 			goto out_return_state;
1317 		}
1318 		spin_unlock(&state->owner->so_lock);
1319 		rcu_read_lock();
1320 		delegation = rcu_dereference(nfsi->delegation);
1321 		if (!can_open_delegated(delegation, fmode)) {
1322 			rcu_read_unlock();
1323 			break;
1324 		}
1325 		/* Save the delegation */
1326 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1327 		rcu_read_unlock();
1328 		nfs_release_seqid(opendata->o_arg.seqid);
1329 		if (!opendata->is_recover) {
1330 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1331 			if (ret != 0)
1332 				goto out;
1333 		}
1334 		ret = -EAGAIN;
1335 
1336 		/* Try to update the stateid using the delegation */
1337 		if (update_open_stateid(state, NULL, &stateid, fmode))
1338 			goto out_return_state;
1339 	}
1340 out:
1341 	return ERR_PTR(ret);
1342 out_return_state:
1343 	atomic_inc(&state->count);
1344 	return state;
1345 }
1346 
1347 static void
1348 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1349 {
1350 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1351 	struct nfs_delegation *delegation;
1352 	int delegation_flags = 0;
1353 
1354 	rcu_read_lock();
1355 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1356 	if (delegation)
1357 		delegation_flags = delegation->flags;
1358 	rcu_read_unlock();
1359 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1360 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1361 				   "returning a delegation for "
1362 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1363 				   clp->cl_hostname);
1364 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1365 		nfs_inode_set_delegation(state->inode,
1366 					 data->owner->so_cred,
1367 					 &data->o_res);
1368 	else
1369 		nfs_inode_reclaim_delegation(state->inode,
1370 					     data->owner->so_cred,
1371 					     &data->o_res);
1372 }
1373 
1374 /*
1375  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1376  * and update the nfs4_state.
1377  */
1378 static struct nfs4_state *
1379 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1380 {
1381 	struct inode *inode = data->state->inode;
1382 	struct nfs4_state *state = data->state;
1383 	int ret;
1384 
1385 	if (!data->rpc_done) {
1386 		if (data->rpc_status) {
1387 			ret = data->rpc_status;
1388 			goto err;
1389 		}
1390 		/* cached opens have already been processed */
1391 		goto update;
1392 	}
1393 
1394 	ret = nfs_refresh_inode(inode, &data->f_attr);
1395 	if (ret)
1396 		goto err;
1397 
1398 	if (data->o_res.delegation_type != 0)
1399 		nfs4_opendata_check_deleg(data, state);
1400 update:
1401 	update_open_stateid(state, &data->o_res.stateid, NULL,
1402 			    data->o_arg.fmode);
1403 	atomic_inc(&state->count);
1404 
1405 	return state;
1406 err:
1407 	return ERR_PTR(ret);
1408 
1409 }
1410 
1411 static struct nfs4_state *
1412 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1413 {
1414 	struct inode *inode;
1415 	struct nfs4_state *state = NULL;
1416 	int ret;
1417 
1418 	if (!data->rpc_done) {
1419 		state = nfs4_try_open_cached(data);
1420 		goto out;
1421 	}
1422 
1423 	ret = -EAGAIN;
1424 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1425 		goto err;
1426 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1427 	ret = PTR_ERR(inode);
1428 	if (IS_ERR(inode))
1429 		goto err;
1430 	ret = -ENOMEM;
1431 	state = nfs4_get_open_state(inode, data->owner);
1432 	if (state == NULL)
1433 		goto err_put_inode;
1434 	if (data->o_res.delegation_type != 0)
1435 		nfs4_opendata_check_deleg(data, state);
1436 	update_open_stateid(state, &data->o_res.stateid, NULL,
1437 			data->o_arg.fmode);
1438 	iput(inode);
1439 out:
1440 	nfs_release_seqid(data->o_arg.seqid);
1441 	return state;
1442 err_put_inode:
1443 	iput(inode);
1444 err:
1445 	return ERR_PTR(ret);
1446 }
1447 
1448 static struct nfs4_state *
1449 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1450 {
1451 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1452 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1453 	return _nfs4_opendata_to_nfs4_state(data);
1454 }
1455 
1456 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1457 {
1458 	struct nfs_inode *nfsi = NFS_I(state->inode);
1459 	struct nfs_open_context *ctx;
1460 
1461 	spin_lock(&state->inode->i_lock);
1462 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1463 		if (ctx->state != state)
1464 			continue;
1465 		get_nfs_open_context(ctx);
1466 		spin_unlock(&state->inode->i_lock);
1467 		return ctx;
1468 	}
1469 	spin_unlock(&state->inode->i_lock);
1470 	return ERR_PTR(-ENOENT);
1471 }
1472 
1473 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1474 		struct nfs4_state *state, enum open_claim_type4 claim)
1475 {
1476 	struct nfs4_opendata *opendata;
1477 
1478 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1479 			NULL, NULL, claim, GFP_NOFS);
1480 	if (opendata == NULL)
1481 		return ERR_PTR(-ENOMEM);
1482 	opendata->state = state;
1483 	atomic_inc(&state->count);
1484 	return opendata;
1485 }
1486 
1487 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1488 {
1489 	struct nfs4_state *newstate;
1490 	int ret;
1491 
1492 	opendata->o_arg.open_flags = 0;
1493 	opendata->o_arg.fmode = fmode;
1494 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1495 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1496 	nfs4_init_opendata_res(opendata);
1497 	ret = _nfs4_recover_proc_open(opendata);
1498 	if (ret != 0)
1499 		return ret;
1500 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1501 	if (IS_ERR(newstate))
1502 		return PTR_ERR(newstate);
1503 	nfs4_close_state(newstate, fmode);
1504 	*res = newstate;
1505 	return 0;
1506 }
1507 
1508 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1509 {
1510 	struct nfs4_state *newstate;
1511 	int ret;
1512 
1513 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1514 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1515 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1516 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1517 	/* memory barrier prior to reading state->n_* */
1518 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1519 	clear_bit(NFS_OPEN_STATE, &state->flags);
1520 	smp_rmb();
1521 	if (state->n_rdwr != 0) {
1522 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1523 		if (ret != 0)
1524 			return ret;
1525 		if (newstate != state)
1526 			return -ESTALE;
1527 	}
1528 	if (state->n_wronly != 0) {
1529 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1530 		if (ret != 0)
1531 			return ret;
1532 		if (newstate != state)
1533 			return -ESTALE;
1534 	}
1535 	if (state->n_rdonly != 0) {
1536 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1537 		if (ret != 0)
1538 			return ret;
1539 		if (newstate != state)
1540 			return -ESTALE;
1541 	}
1542 	/*
1543 	 * We may have performed cached opens for all three recoveries.
1544 	 * Check if we need to update the current stateid.
1545 	 */
1546 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1547 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1548 		write_seqlock(&state->seqlock);
1549 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1550 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1551 		write_sequnlock(&state->seqlock);
1552 	}
1553 	return 0;
1554 }
1555 
1556 /*
1557  * OPEN_RECLAIM:
1558  * 	reclaim state on the server after a reboot.
1559  */
1560 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1561 {
1562 	struct nfs_delegation *delegation;
1563 	struct nfs4_opendata *opendata;
1564 	fmode_t delegation_type = 0;
1565 	int status;
1566 
1567 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1568 			NFS4_OPEN_CLAIM_PREVIOUS);
1569 	if (IS_ERR(opendata))
1570 		return PTR_ERR(opendata);
1571 	rcu_read_lock();
1572 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1573 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1574 		delegation_type = delegation->type;
1575 	rcu_read_unlock();
1576 	opendata->o_arg.u.delegation_type = delegation_type;
1577 	status = nfs4_open_recover(opendata, state);
1578 	nfs4_opendata_put(opendata);
1579 	return status;
1580 }
1581 
1582 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1583 {
1584 	struct nfs_server *server = NFS_SERVER(state->inode);
1585 	struct nfs4_exception exception = { };
1586 	int err;
1587 	do {
1588 		err = _nfs4_do_open_reclaim(ctx, state);
1589 		trace_nfs4_open_reclaim(ctx, 0, err);
1590 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1591 			continue;
1592 		if (err != -NFS4ERR_DELAY)
1593 			break;
1594 		nfs4_handle_exception(server, err, &exception);
1595 	} while (exception.retry);
1596 	return err;
1597 }
1598 
1599 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1600 {
1601 	struct nfs_open_context *ctx;
1602 	int ret;
1603 
1604 	ctx = nfs4_state_find_open_context(state);
1605 	if (IS_ERR(ctx))
1606 		return -EAGAIN;
1607 	ret = nfs4_do_open_reclaim(ctx, state);
1608 	put_nfs_open_context(ctx);
1609 	return ret;
1610 }
1611 
1612 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1613 {
1614 	switch (err) {
1615 		default:
1616 			printk(KERN_ERR "NFS: %s: unhandled error "
1617 					"%d.\n", __func__, err);
1618 		case 0:
1619 		case -ENOENT:
1620 		case -ESTALE:
1621 			break;
1622 		case -NFS4ERR_BADSESSION:
1623 		case -NFS4ERR_BADSLOT:
1624 		case -NFS4ERR_BAD_HIGH_SLOT:
1625 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1626 		case -NFS4ERR_DEADSESSION:
1627 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1628 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1629 			return -EAGAIN;
1630 		case -NFS4ERR_STALE_CLIENTID:
1631 		case -NFS4ERR_STALE_STATEID:
1632 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1633 		case -NFS4ERR_EXPIRED:
1634 			/* Don't recall a delegation if it was lost */
1635 			nfs4_schedule_lease_recovery(server->nfs_client);
1636 			return -EAGAIN;
1637 		case -NFS4ERR_MOVED:
1638 			nfs4_schedule_migration_recovery(server);
1639 			return -EAGAIN;
1640 		case -NFS4ERR_LEASE_MOVED:
1641 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1642 			return -EAGAIN;
1643 		case -NFS4ERR_DELEG_REVOKED:
1644 		case -NFS4ERR_ADMIN_REVOKED:
1645 		case -NFS4ERR_BAD_STATEID:
1646 		case -NFS4ERR_OPENMODE:
1647 			nfs_inode_find_state_and_recover(state->inode,
1648 					stateid);
1649 			nfs4_schedule_stateid_recovery(server, state);
1650 			return -EAGAIN;
1651 		case -NFS4ERR_DELAY:
1652 		case -NFS4ERR_GRACE:
1653 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1654 			ssleep(1);
1655 			return -EAGAIN;
1656 		case -ENOMEM:
1657 		case -NFS4ERR_DENIED:
1658 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1659 			return 0;
1660 	}
1661 	return err;
1662 }
1663 
1664 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1665 {
1666 	struct nfs_server *server = NFS_SERVER(state->inode);
1667 	struct nfs4_opendata *opendata;
1668 	int err;
1669 
1670 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1671 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1672 	if (IS_ERR(opendata))
1673 		return PTR_ERR(opendata);
1674 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1675 	err = nfs4_open_recover(opendata, state);
1676 	nfs4_opendata_put(opendata);
1677 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1678 }
1679 
1680 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1681 {
1682 	struct nfs4_opendata *data = calldata;
1683 
1684 	nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1685 				&data->c_res.seq_res, task);
1686 }
1687 
1688 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1689 {
1690 	struct nfs4_opendata *data = calldata;
1691 
1692 	nfs40_sequence_done(task, &data->c_res.seq_res);
1693 
1694 	data->rpc_status = task->tk_status;
1695 	if (data->rpc_status == 0) {
1696 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1697 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1698 		renew_lease(data->o_res.server, data->timestamp);
1699 		data->rpc_done = 1;
1700 	}
1701 }
1702 
1703 static void nfs4_open_confirm_release(void *calldata)
1704 {
1705 	struct nfs4_opendata *data = calldata;
1706 	struct nfs4_state *state = NULL;
1707 
1708 	/* If this request hasn't been cancelled, do nothing */
1709 	if (data->cancelled == 0)
1710 		goto out_free;
1711 	/* In case of error, no cleanup! */
1712 	if (!data->rpc_done)
1713 		goto out_free;
1714 	state = nfs4_opendata_to_nfs4_state(data);
1715 	if (!IS_ERR(state))
1716 		nfs4_close_state(state, data->o_arg.fmode);
1717 out_free:
1718 	nfs4_opendata_put(data);
1719 }
1720 
1721 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1722 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1723 	.rpc_call_done = nfs4_open_confirm_done,
1724 	.rpc_release = nfs4_open_confirm_release,
1725 };
1726 
1727 /*
1728  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1729  */
1730 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1731 {
1732 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1733 	struct rpc_task *task;
1734 	struct  rpc_message msg = {
1735 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1736 		.rpc_argp = &data->c_arg,
1737 		.rpc_resp = &data->c_res,
1738 		.rpc_cred = data->owner->so_cred,
1739 	};
1740 	struct rpc_task_setup task_setup_data = {
1741 		.rpc_client = server->client,
1742 		.rpc_message = &msg,
1743 		.callback_ops = &nfs4_open_confirm_ops,
1744 		.callback_data = data,
1745 		.workqueue = nfsiod_workqueue,
1746 		.flags = RPC_TASK_ASYNC,
1747 	};
1748 	int status;
1749 
1750 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1751 	kref_get(&data->kref);
1752 	data->rpc_done = 0;
1753 	data->rpc_status = 0;
1754 	data->timestamp = jiffies;
1755 	task = rpc_run_task(&task_setup_data);
1756 	if (IS_ERR(task))
1757 		return PTR_ERR(task);
1758 	status = nfs4_wait_for_completion_rpc_task(task);
1759 	if (status != 0) {
1760 		data->cancelled = 1;
1761 		smp_wmb();
1762 	} else
1763 		status = data->rpc_status;
1764 	rpc_put_task(task);
1765 	return status;
1766 }
1767 
1768 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1769 {
1770 	struct nfs4_opendata *data = calldata;
1771 	struct nfs4_state_owner *sp = data->owner;
1772 	struct nfs_client *clp = sp->so_server->nfs_client;
1773 
1774 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1775 		goto out_wait;
1776 	/*
1777 	 * Check if we still need to send an OPEN call, or if we can use
1778 	 * a delegation instead.
1779 	 */
1780 	if (data->state != NULL) {
1781 		struct nfs_delegation *delegation;
1782 
1783 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1784 			goto out_no_action;
1785 		rcu_read_lock();
1786 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1787 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1788 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1789 		    can_open_delegated(delegation, data->o_arg.fmode))
1790 			goto unlock_no_action;
1791 		rcu_read_unlock();
1792 	}
1793 	/* Update client id. */
1794 	data->o_arg.clientid = clp->cl_clientid;
1795 	switch (data->o_arg.claim) {
1796 	case NFS4_OPEN_CLAIM_PREVIOUS:
1797 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1798 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1799 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1800 	case NFS4_OPEN_CLAIM_FH:
1801 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1802 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1803 	}
1804 	data->timestamp = jiffies;
1805 	if (nfs4_setup_sequence(data->o_arg.server,
1806 				&data->o_arg.seq_args,
1807 				&data->o_res.seq_res,
1808 				task) != 0)
1809 		nfs_release_seqid(data->o_arg.seqid);
1810 
1811 	/* Set the create mode (note dependency on the session type) */
1812 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1813 	if (data->o_arg.open_flags & O_EXCL) {
1814 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1815 		if (nfs4_has_persistent_session(clp))
1816 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1817 		else if (clp->cl_mvops->minor_version > 0)
1818 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1819 	}
1820 	return;
1821 unlock_no_action:
1822 	rcu_read_unlock();
1823 out_no_action:
1824 	task->tk_action = NULL;
1825 out_wait:
1826 	nfs4_sequence_done(task, &data->o_res.seq_res);
1827 }
1828 
1829 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1830 {
1831 	struct nfs4_opendata *data = calldata;
1832 
1833 	data->rpc_status = task->tk_status;
1834 
1835 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1836 		return;
1837 
1838 	if (task->tk_status == 0) {
1839 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1840 			switch (data->o_res.f_attr->mode & S_IFMT) {
1841 			case S_IFREG:
1842 				break;
1843 			case S_IFLNK:
1844 				data->rpc_status = -ELOOP;
1845 				break;
1846 			case S_IFDIR:
1847 				data->rpc_status = -EISDIR;
1848 				break;
1849 			default:
1850 				data->rpc_status = -ENOTDIR;
1851 			}
1852 		}
1853 		renew_lease(data->o_res.server, data->timestamp);
1854 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1855 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1856 	}
1857 	data->rpc_done = 1;
1858 }
1859 
1860 static void nfs4_open_release(void *calldata)
1861 {
1862 	struct nfs4_opendata *data = calldata;
1863 	struct nfs4_state *state = NULL;
1864 
1865 	/* If this request hasn't been cancelled, do nothing */
1866 	if (data->cancelled == 0)
1867 		goto out_free;
1868 	/* In case of error, no cleanup! */
1869 	if (data->rpc_status != 0 || !data->rpc_done)
1870 		goto out_free;
1871 	/* In case we need an open_confirm, no cleanup! */
1872 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1873 		goto out_free;
1874 	state = nfs4_opendata_to_nfs4_state(data);
1875 	if (!IS_ERR(state))
1876 		nfs4_close_state(state, data->o_arg.fmode);
1877 out_free:
1878 	nfs4_opendata_put(data);
1879 }
1880 
1881 static const struct rpc_call_ops nfs4_open_ops = {
1882 	.rpc_call_prepare = nfs4_open_prepare,
1883 	.rpc_call_done = nfs4_open_done,
1884 	.rpc_release = nfs4_open_release,
1885 };
1886 
1887 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1888 {
1889 	struct inode *dir = data->dir->d_inode;
1890 	struct nfs_server *server = NFS_SERVER(dir);
1891 	struct nfs_openargs *o_arg = &data->o_arg;
1892 	struct nfs_openres *o_res = &data->o_res;
1893 	struct rpc_task *task;
1894 	struct rpc_message msg = {
1895 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1896 		.rpc_argp = o_arg,
1897 		.rpc_resp = o_res,
1898 		.rpc_cred = data->owner->so_cred,
1899 	};
1900 	struct rpc_task_setup task_setup_data = {
1901 		.rpc_client = server->client,
1902 		.rpc_message = &msg,
1903 		.callback_ops = &nfs4_open_ops,
1904 		.callback_data = data,
1905 		.workqueue = nfsiod_workqueue,
1906 		.flags = RPC_TASK_ASYNC,
1907 	};
1908 	int status;
1909 
1910 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1911 	kref_get(&data->kref);
1912 	data->rpc_done = 0;
1913 	data->rpc_status = 0;
1914 	data->cancelled = 0;
1915 	data->is_recover = 0;
1916 	if (isrecover) {
1917 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1918 		data->is_recover = 1;
1919 	}
1920 	task = rpc_run_task(&task_setup_data);
1921         if (IS_ERR(task))
1922                 return PTR_ERR(task);
1923         status = nfs4_wait_for_completion_rpc_task(task);
1924         if (status != 0) {
1925                 data->cancelled = 1;
1926                 smp_wmb();
1927         } else
1928                 status = data->rpc_status;
1929         rpc_put_task(task);
1930 
1931 	return status;
1932 }
1933 
1934 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1935 {
1936 	struct inode *dir = data->dir->d_inode;
1937 	struct nfs_openres *o_res = &data->o_res;
1938         int status;
1939 
1940 	status = nfs4_run_open_task(data, 1);
1941 	if (status != 0 || !data->rpc_done)
1942 		return status;
1943 
1944 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1945 
1946 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1947 		status = _nfs4_proc_open_confirm(data);
1948 		if (status != 0)
1949 			return status;
1950 	}
1951 
1952 	return status;
1953 }
1954 
1955 /*
1956  * Additional permission checks in order to distinguish between an
1957  * open for read, and an open for execute. This works around the
1958  * fact that NFSv4 OPEN treats read and execute permissions as being
1959  * the same.
1960  * Note that in the non-execute case, we want to turn off permission
1961  * checking if we just created a new file (POSIX open() semantics).
1962  */
1963 static int nfs4_opendata_access(struct rpc_cred *cred,
1964 				struct nfs4_opendata *opendata,
1965 				struct nfs4_state *state, fmode_t fmode,
1966 				int openflags)
1967 {
1968 	struct nfs_access_entry cache;
1969 	u32 mask;
1970 
1971 	/* access call failed or for some reason the server doesn't
1972 	 * support any access modes -- defer access call until later */
1973 	if (opendata->o_res.access_supported == 0)
1974 		return 0;
1975 
1976 	mask = 0;
1977 	/*
1978 	 * Use openflags to check for exec, because fmode won't
1979 	 * always have FMODE_EXEC set when file open for exec.
1980 	 */
1981 	if (openflags & __FMODE_EXEC) {
1982 		/* ONLY check for exec rights */
1983 		mask = MAY_EXEC;
1984 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
1985 		mask = MAY_READ;
1986 
1987 	cache.cred = cred;
1988 	cache.jiffies = jiffies;
1989 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
1990 	nfs_access_add_cache(state->inode, &cache);
1991 
1992 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1993 		return 0;
1994 
1995 	/* even though OPEN succeeded, access is denied. Close the file */
1996 	nfs4_close_state(state, fmode);
1997 	return -EACCES;
1998 }
1999 
2000 /*
2001  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2002  */
2003 static int _nfs4_proc_open(struct nfs4_opendata *data)
2004 {
2005 	struct inode *dir = data->dir->d_inode;
2006 	struct nfs_server *server = NFS_SERVER(dir);
2007 	struct nfs_openargs *o_arg = &data->o_arg;
2008 	struct nfs_openres *o_res = &data->o_res;
2009 	int status;
2010 
2011 	status = nfs4_run_open_task(data, 0);
2012 	if (!data->rpc_done)
2013 		return status;
2014 	if (status != 0) {
2015 		if (status == -NFS4ERR_BADNAME &&
2016 				!(o_arg->open_flags & O_CREAT))
2017 			return -ENOENT;
2018 		return status;
2019 	}
2020 
2021 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2022 
2023 	if (o_arg->open_flags & O_CREAT) {
2024 		update_changeattr(dir, &o_res->cinfo);
2025 		if (o_arg->open_flags & O_EXCL)
2026 			data->file_created = 1;
2027 		else if (o_res->cinfo.before != o_res->cinfo.after)
2028 			data->file_created = 1;
2029 	}
2030 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2031 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2032 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2033 		status = _nfs4_proc_open_confirm(data);
2034 		if (status != 0)
2035 			return status;
2036 	}
2037 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2038 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2039 	return 0;
2040 }
2041 
2042 static int nfs4_recover_expired_lease(struct nfs_server *server)
2043 {
2044 	return nfs4_client_recover_expired_lease(server->nfs_client);
2045 }
2046 
2047 /*
2048  * OPEN_EXPIRED:
2049  * 	reclaim state on the server after a network partition.
2050  * 	Assumes caller holds the appropriate lock
2051  */
2052 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2053 {
2054 	struct nfs4_opendata *opendata;
2055 	int ret;
2056 
2057 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2058 			NFS4_OPEN_CLAIM_FH);
2059 	if (IS_ERR(opendata))
2060 		return PTR_ERR(opendata);
2061 	ret = nfs4_open_recover(opendata, state);
2062 	if (ret == -ESTALE)
2063 		d_drop(ctx->dentry);
2064 	nfs4_opendata_put(opendata);
2065 	return ret;
2066 }
2067 
2068 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2069 {
2070 	struct nfs_server *server = NFS_SERVER(state->inode);
2071 	struct nfs4_exception exception = { };
2072 	int err;
2073 
2074 	do {
2075 		err = _nfs4_open_expired(ctx, state);
2076 		trace_nfs4_open_expired(ctx, 0, err);
2077 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2078 			continue;
2079 		switch (err) {
2080 		default:
2081 			goto out;
2082 		case -NFS4ERR_GRACE:
2083 		case -NFS4ERR_DELAY:
2084 			nfs4_handle_exception(server, err, &exception);
2085 			err = 0;
2086 		}
2087 	} while (exception.retry);
2088 out:
2089 	return err;
2090 }
2091 
2092 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2093 {
2094 	struct nfs_open_context *ctx;
2095 	int ret;
2096 
2097 	ctx = nfs4_state_find_open_context(state);
2098 	if (IS_ERR(ctx))
2099 		return -EAGAIN;
2100 	ret = nfs4_do_open_expired(ctx, state);
2101 	put_nfs_open_context(ctx);
2102 	return ret;
2103 }
2104 
2105 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2106 {
2107 	nfs_remove_bad_delegation(state->inode);
2108 	write_seqlock(&state->seqlock);
2109 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2110 	write_sequnlock(&state->seqlock);
2111 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2112 }
2113 
2114 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2115 {
2116 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2117 		nfs_finish_clear_delegation_stateid(state);
2118 }
2119 
2120 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2121 {
2122 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2123 	nfs40_clear_delegation_stateid(state);
2124 	return nfs4_open_expired(sp, state);
2125 }
2126 
2127 #if defined(CONFIG_NFS_V4_1)
2128 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2129 {
2130 	struct nfs_server *server = NFS_SERVER(state->inode);
2131 	nfs4_stateid stateid;
2132 	struct nfs_delegation *delegation;
2133 	struct rpc_cred *cred;
2134 	int status;
2135 
2136 	/* Get the delegation credential for use by test/free_stateid */
2137 	rcu_read_lock();
2138 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2139 	if (delegation == NULL) {
2140 		rcu_read_unlock();
2141 		return;
2142 	}
2143 
2144 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2145 	cred = get_rpccred(delegation->cred);
2146 	rcu_read_unlock();
2147 	status = nfs41_test_stateid(server, &stateid, cred);
2148 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2149 
2150 	if (status != NFS_OK) {
2151 		/* Free the stateid unless the server explicitly
2152 		 * informs us the stateid is unrecognized. */
2153 		if (status != -NFS4ERR_BAD_STATEID)
2154 			nfs41_free_stateid(server, &stateid, cred);
2155 		nfs_finish_clear_delegation_stateid(state);
2156 	}
2157 
2158 	put_rpccred(cred);
2159 }
2160 
2161 /**
2162  * nfs41_check_open_stateid - possibly free an open stateid
2163  *
2164  * @state: NFSv4 state for an inode
2165  *
2166  * Returns NFS_OK if recovery for this stateid is now finished.
2167  * Otherwise a negative NFS4ERR value is returned.
2168  */
2169 static int nfs41_check_open_stateid(struct nfs4_state *state)
2170 {
2171 	struct nfs_server *server = NFS_SERVER(state->inode);
2172 	nfs4_stateid *stateid = &state->open_stateid;
2173 	struct rpc_cred *cred = state->owner->so_cred;
2174 	int status;
2175 
2176 	/* If a state reset has been done, test_stateid is unneeded */
2177 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2178 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2179 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2180 		return -NFS4ERR_BAD_STATEID;
2181 
2182 	status = nfs41_test_stateid(server, stateid, cred);
2183 	trace_nfs4_test_open_stateid(state, NULL, status);
2184 	if (status != NFS_OK) {
2185 		/* Free the stateid unless the server explicitly
2186 		 * informs us the stateid is unrecognized. */
2187 		if (status != -NFS4ERR_BAD_STATEID)
2188 			nfs41_free_stateid(server, stateid, cred);
2189 
2190 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2191 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2192 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2193 		clear_bit(NFS_OPEN_STATE, &state->flags);
2194 	}
2195 	return status;
2196 }
2197 
2198 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2199 {
2200 	int status;
2201 
2202 	nfs41_check_delegation_stateid(state);
2203 	status = nfs41_check_open_stateid(state);
2204 	if (status != NFS_OK)
2205 		status = nfs4_open_expired(sp, state);
2206 	return status;
2207 }
2208 #endif
2209 
2210 /*
2211  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2212  * fields corresponding to attributes that were used to store the verifier.
2213  * Make sure we clobber those fields in the later setattr call
2214  */
2215 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2216 {
2217 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2218 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2219 		sattr->ia_valid |= ATTR_ATIME;
2220 
2221 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2222 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2223 		sattr->ia_valid |= ATTR_MTIME;
2224 }
2225 
2226 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2227 		fmode_t fmode,
2228 		int flags,
2229 		struct nfs_open_context *ctx)
2230 {
2231 	struct nfs4_state_owner *sp = opendata->owner;
2232 	struct nfs_server *server = sp->so_server;
2233 	struct dentry *dentry;
2234 	struct nfs4_state *state;
2235 	unsigned int seq;
2236 	int ret;
2237 
2238 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2239 
2240 	ret = _nfs4_proc_open(opendata);
2241 	if (ret != 0)
2242 		goto out;
2243 
2244 	state = nfs4_opendata_to_nfs4_state(opendata);
2245 	ret = PTR_ERR(state);
2246 	if (IS_ERR(state))
2247 		goto out;
2248 	if (server->caps & NFS_CAP_POSIX_LOCK)
2249 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2250 
2251 	dentry = opendata->dentry;
2252 	if (dentry->d_inode == NULL) {
2253 		/* FIXME: Is this d_drop() ever needed? */
2254 		d_drop(dentry);
2255 		dentry = d_add_unique(dentry, igrab(state->inode));
2256 		if (dentry == NULL) {
2257 			dentry = opendata->dentry;
2258 		} else if (dentry != ctx->dentry) {
2259 			dput(ctx->dentry);
2260 			ctx->dentry = dget(dentry);
2261 		}
2262 		nfs_set_verifier(dentry,
2263 				nfs_save_change_attribute(opendata->dir->d_inode));
2264 	}
2265 
2266 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2267 	if (ret != 0)
2268 		goto out;
2269 
2270 	ctx->state = state;
2271 	if (dentry->d_inode == state->inode) {
2272 		nfs_inode_attach_open_context(ctx);
2273 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2274 			nfs4_schedule_stateid_recovery(server, state);
2275 	}
2276 out:
2277 	return ret;
2278 }
2279 
2280 /*
2281  * Returns a referenced nfs4_state
2282  */
2283 static int _nfs4_do_open(struct inode *dir,
2284 			struct nfs_open_context *ctx,
2285 			int flags,
2286 			struct iattr *sattr,
2287 			struct nfs4_label *label,
2288 			int *opened)
2289 {
2290 	struct nfs4_state_owner  *sp;
2291 	struct nfs4_state     *state = NULL;
2292 	struct nfs_server       *server = NFS_SERVER(dir);
2293 	struct nfs4_opendata *opendata;
2294 	struct dentry *dentry = ctx->dentry;
2295 	struct rpc_cred *cred = ctx->cred;
2296 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2297 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2298 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2299 	struct nfs4_label *olabel = NULL;
2300 	int status;
2301 
2302 	/* Protect against reboot recovery conflicts */
2303 	status = -ENOMEM;
2304 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2305 	if (sp == NULL) {
2306 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2307 		goto out_err;
2308 	}
2309 	status = nfs4_recover_expired_lease(server);
2310 	if (status != 0)
2311 		goto err_put_state_owner;
2312 	if (dentry->d_inode != NULL)
2313 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2314 	status = -ENOMEM;
2315 	if (dentry->d_inode)
2316 		claim = NFS4_OPEN_CLAIM_FH;
2317 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2318 			label, claim, GFP_KERNEL);
2319 	if (opendata == NULL)
2320 		goto err_put_state_owner;
2321 
2322 	if (label) {
2323 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2324 		if (IS_ERR(olabel)) {
2325 			status = PTR_ERR(olabel);
2326 			goto err_opendata_put;
2327 		}
2328 	}
2329 
2330 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2331 		if (!opendata->f_attr.mdsthreshold) {
2332 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2333 			if (!opendata->f_attr.mdsthreshold)
2334 				goto err_free_label;
2335 		}
2336 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2337 	}
2338 	if (dentry->d_inode != NULL)
2339 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2340 
2341 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2342 	if (status != 0)
2343 		goto err_free_label;
2344 	state = ctx->state;
2345 
2346 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2347 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2348 		nfs4_exclusive_attrset(opendata, sattr);
2349 
2350 		nfs_fattr_init(opendata->o_res.f_attr);
2351 		status = nfs4_do_setattr(state->inode, cred,
2352 				opendata->o_res.f_attr, sattr,
2353 				state, label, olabel);
2354 		if (status == 0) {
2355 			nfs_setattr_update_inode(state->inode, sattr);
2356 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2357 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2358 		}
2359 	}
2360 	if (opendata->file_created)
2361 		*opened |= FILE_CREATED;
2362 
2363 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2364 		*ctx_th = opendata->f_attr.mdsthreshold;
2365 		opendata->f_attr.mdsthreshold = NULL;
2366 	}
2367 
2368 	nfs4_label_free(olabel);
2369 
2370 	nfs4_opendata_put(opendata);
2371 	nfs4_put_state_owner(sp);
2372 	return 0;
2373 err_free_label:
2374 	nfs4_label_free(olabel);
2375 err_opendata_put:
2376 	nfs4_opendata_put(opendata);
2377 err_put_state_owner:
2378 	nfs4_put_state_owner(sp);
2379 out_err:
2380 	return status;
2381 }
2382 
2383 
2384 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2385 					struct nfs_open_context *ctx,
2386 					int flags,
2387 					struct iattr *sattr,
2388 					struct nfs4_label *label,
2389 					int *opened)
2390 {
2391 	struct nfs_server *server = NFS_SERVER(dir);
2392 	struct nfs4_exception exception = { };
2393 	struct nfs4_state *res;
2394 	int status;
2395 
2396 	do {
2397 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2398 		res = ctx->state;
2399 		trace_nfs4_open_file(ctx, flags, status);
2400 		if (status == 0)
2401 			break;
2402 		/* NOTE: BAD_SEQID means the server and client disagree about the
2403 		 * book-keeping w.r.t. state-changing operations
2404 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2405 		 * It is actually a sign of a bug on the client or on the server.
2406 		 *
2407 		 * If we receive a BAD_SEQID error in the particular case of
2408 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2409 		 * have unhashed the old state_owner for us, and that we can
2410 		 * therefore safely retry using a new one. We should still warn
2411 		 * the user though...
2412 		 */
2413 		if (status == -NFS4ERR_BAD_SEQID) {
2414 			pr_warn_ratelimited("NFS: v4 server %s "
2415 					" returned a bad sequence-id error!\n",
2416 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2417 			exception.retry = 1;
2418 			continue;
2419 		}
2420 		/*
2421 		 * BAD_STATEID on OPEN means that the server cancelled our
2422 		 * state before it received the OPEN_CONFIRM.
2423 		 * Recover by retrying the request as per the discussion
2424 		 * on Page 181 of RFC3530.
2425 		 */
2426 		if (status == -NFS4ERR_BAD_STATEID) {
2427 			exception.retry = 1;
2428 			continue;
2429 		}
2430 		if (status == -EAGAIN) {
2431 			/* We must have found a delegation */
2432 			exception.retry = 1;
2433 			continue;
2434 		}
2435 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2436 			continue;
2437 		res = ERR_PTR(nfs4_handle_exception(server,
2438 					status, &exception));
2439 	} while (exception.retry);
2440 	return res;
2441 }
2442 
2443 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2444 			    struct nfs_fattr *fattr, struct iattr *sattr,
2445 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2446 			    struct nfs4_label *olabel)
2447 {
2448 	struct nfs_server *server = NFS_SERVER(inode);
2449         struct nfs_setattrargs  arg = {
2450                 .fh             = NFS_FH(inode),
2451                 .iap            = sattr,
2452 		.server		= server,
2453 		.bitmask = server->attr_bitmask,
2454 		.label		= ilabel,
2455         };
2456         struct nfs_setattrres  res = {
2457 		.fattr		= fattr,
2458 		.label		= olabel,
2459 		.server		= server,
2460         };
2461         struct rpc_message msg = {
2462 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2463 		.rpc_argp	= &arg,
2464 		.rpc_resp	= &res,
2465 		.rpc_cred	= cred,
2466         };
2467 	unsigned long timestamp = jiffies;
2468 	fmode_t fmode;
2469 	bool truncate;
2470 	int status;
2471 
2472 	arg.bitmask = nfs4_bitmask(server, ilabel);
2473 	if (ilabel)
2474 		arg.bitmask = nfs4_bitmask(server, olabel);
2475 
2476 	nfs_fattr_init(fattr);
2477 
2478 	/* Servers should only apply open mode checks for file size changes */
2479 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2480 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2481 
2482 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2483 		/* Use that stateid */
2484 	} else if (truncate && state != NULL) {
2485 		struct nfs_lockowner lockowner = {
2486 			.l_owner = current->files,
2487 			.l_pid = current->tgid,
2488 		};
2489 		if (!nfs4_valid_open_stateid(state))
2490 			return -EBADF;
2491 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2492 				&lockowner) == -EIO)
2493 			return -EBADF;
2494 	} else
2495 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2496 
2497 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2498 	if (status == 0 && state != NULL)
2499 		renew_lease(server, timestamp);
2500 	return status;
2501 }
2502 
2503 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2504 			   struct nfs_fattr *fattr, struct iattr *sattr,
2505 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2506 			   struct nfs4_label *olabel)
2507 {
2508 	struct nfs_server *server = NFS_SERVER(inode);
2509 	struct nfs4_exception exception = {
2510 		.state = state,
2511 		.inode = inode,
2512 	};
2513 	int err;
2514 	do {
2515 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2516 		trace_nfs4_setattr(inode, err);
2517 		switch (err) {
2518 		case -NFS4ERR_OPENMODE:
2519 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2520 				pr_warn_once("NFSv4: server %s is incorrectly "
2521 						"applying open mode checks to "
2522 						"a SETATTR that is not "
2523 						"changing file size.\n",
2524 						server->nfs_client->cl_hostname);
2525 			}
2526 			if (state && !(state->state & FMODE_WRITE)) {
2527 				err = -EBADF;
2528 				if (sattr->ia_valid & ATTR_OPEN)
2529 					err = -EACCES;
2530 				goto out;
2531 			}
2532 		}
2533 		err = nfs4_handle_exception(server, err, &exception);
2534 	} while (exception.retry);
2535 out:
2536 	return err;
2537 }
2538 
2539 struct nfs4_closedata {
2540 	struct inode *inode;
2541 	struct nfs4_state *state;
2542 	struct nfs_closeargs arg;
2543 	struct nfs_closeres res;
2544 	struct nfs_fattr fattr;
2545 	unsigned long timestamp;
2546 	bool roc;
2547 	u32 roc_barrier;
2548 };
2549 
2550 static void nfs4_free_closedata(void *data)
2551 {
2552 	struct nfs4_closedata *calldata = data;
2553 	struct nfs4_state_owner *sp = calldata->state->owner;
2554 	struct super_block *sb = calldata->state->inode->i_sb;
2555 
2556 	if (calldata->roc)
2557 		pnfs_roc_release(calldata->state->inode);
2558 	nfs4_put_open_state(calldata->state);
2559 	nfs_free_seqid(calldata->arg.seqid);
2560 	nfs4_put_state_owner(sp);
2561 	nfs_sb_deactive(sb);
2562 	kfree(calldata);
2563 }
2564 
2565 static void nfs4_close_done(struct rpc_task *task, void *data)
2566 {
2567 	struct nfs4_closedata *calldata = data;
2568 	struct nfs4_state *state = calldata->state;
2569 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2570 	nfs4_stateid *res_stateid = NULL;
2571 
2572 	dprintk("%s: begin!\n", __func__);
2573 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2574 		return;
2575 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2576         /* hmm. we are done with the inode, and in the process of freeing
2577 	 * the state_owner. we keep this around to process errors
2578 	 */
2579 	switch (task->tk_status) {
2580 		case 0:
2581 			res_stateid = &calldata->res.stateid;
2582 			if (calldata->arg.fmode == 0 && calldata->roc)
2583 				pnfs_roc_set_barrier(state->inode,
2584 						     calldata->roc_barrier);
2585 			renew_lease(server, calldata->timestamp);
2586 			break;
2587 		case -NFS4ERR_ADMIN_REVOKED:
2588 		case -NFS4ERR_STALE_STATEID:
2589 		case -NFS4ERR_OLD_STATEID:
2590 		case -NFS4ERR_BAD_STATEID:
2591 		case -NFS4ERR_EXPIRED:
2592 			if (calldata->arg.fmode == 0)
2593 				break;
2594 		default:
2595 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2596 				rpc_restart_call_prepare(task);
2597 				goto out_release;
2598 			}
2599 	}
2600 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2601 out_release:
2602 	nfs_release_seqid(calldata->arg.seqid);
2603 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2604 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2605 }
2606 
2607 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2608 {
2609 	struct nfs4_closedata *calldata = data;
2610 	struct nfs4_state *state = calldata->state;
2611 	struct inode *inode = calldata->inode;
2612 	bool is_rdonly, is_wronly, is_rdwr;
2613 	int call_close = 0;
2614 
2615 	dprintk("%s: begin!\n", __func__);
2616 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2617 		goto out_wait;
2618 
2619 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2620 	spin_lock(&state->owner->so_lock);
2621 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2622 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2623 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2624 	/* Calculate the change in open mode */
2625 	calldata->arg.fmode = 0;
2626 	if (state->n_rdwr == 0) {
2627 		if (state->n_rdonly == 0)
2628 			call_close |= is_rdonly;
2629 		else if (is_rdonly)
2630 			calldata->arg.fmode |= FMODE_READ;
2631 		if (state->n_wronly == 0)
2632 			call_close |= is_wronly;
2633 		else if (is_wronly)
2634 			calldata->arg.fmode |= FMODE_WRITE;
2635 	} else if (is_rdwr)
2636 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2637 
2638 	if (calldata->arg.fmode == 0)
2639 		call_close |= is_rdwr;
2640 
2641 	if (!nfs4_valid_open_stateid(state))
2642 		call_close = 0;
2643 	spin_unlock(&state->owner->so_lock);
2644 
2645 	if (!call_close) {
2646 		/* Note: exit _without_ calling nfs4_close_done */
2647 		goto out_no_action;
2648 	}
2649 
2650 	if (calldata->arg.fmode == 0) {
2651 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2652 		if (calldata->roc &&
2653 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2654 			nfs_release_seqid(calldata->arg.seqid);
2655 			goto out_wait;
2656 		    }
2657 	}
2658 
2659 	nfs_fattr_init(calldata->res.fattr);
2660 	calldata->timestamp = jiffies;
2661 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2662 				&calldata->arg.seq_args,
2663 				&calldata->res.seq_res,
2664 				task) != 0)
2665 		nfs_release_seqid(calldata->arg.seqid);
2666 	dprintk("%s: done!\n", __func__);
2667 	return;
2668 out_no_action:
2669 	task->tk_action = NULL;
2670 out_wait:
2671 	nfs4_sequence_done(task, &calldata->res.seq_res);
2672 }
2673 
2674 static const struct rpc_call_ops nfs4_close_ops = {
2675 	.rpc_call_prepare = nfs4_close_prepare,
2676 	.rpc_call_done = nfs4_close_done,
2677 	.rpc_release = nfs4_free_closedata,
2678 };
2679 
2680 static bool nfs4_state_has_opener(struct nfs4_state *state)
2681 {
2682 	/* first check existing openers */
2683 	if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0 &&
2684 	    state->n_rdonly != 0)
2685 		return true;
2686 
2687 	if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0 &&
2688 	    state->n_wronly != 0)
2689 		return true;
2690 
2691 	if (test_bit(NFS_O_RDWR_STATE, &state->flags) != 0 &&
2692 	    state->n_rdwr != 0)
2693 		return true;
2694 
2695 	return false;
2696 }
2697 
2698 static bool nfs4_roc(struct inode *inode)
2699 {
2700 	struct nfs_inode *nfsi = NFS_I(inode);
2701 	struct nfs_open_context *ctx;
2702 	struct nfs4_state *state;
2703 
2704 	spin_lock(&inode->i_lock);
2705 	list_for_each_entry(ctx, &nfsi->open_files, list) {
2706 		state = ctx->state;
2707 		if (state == NULL)
2708 			continue;
2709 		if (nfs4_state_has_opener(state)) {
2710 			spin_unlock(&inode->i_lock);
2711 			return false;
2712 		}
2713 	}
2714 	spin_unlock(&inode->i_lock);
2715 
2716 	if (nfs4_check_delegation(inode, FMODE_READ))
2717 		return false;
2718 
2719 	return pnfs_roc(inode);
2720 }
2721 
2722 /*
2723  * It is possible for data to be read/written from a mem-mapped file
2724  * after the sys_close call (which hits the vfs layer as a flush).
2725  * This means that we can't safely call nfsv4 close on a file until
2726  * the inode is cleared. This in turn means that we are not good
2727  * NFSv4 citizens - we do not indicate to the server to update the file's
2728  * share state even when we are done with one of the three share
2729  * stateid's in the inode.
2730  *
2731  * NOTE: Caller must be holding the sp->so_owner semaphore!
2732  */
2733 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2734 {
2735 	struct nfs_server *server = NFS_SERVER(state->inode);
2736 	struct nfs4_closedata *calldata;
2737 	struct nfs4_state_owner *sp = state->owner;
2738 	struct rpc_task *task;
2739 	struct rpc_message msg = {
2740 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2741 		.rpc_cred = state->owner->so_cred,
2742 	};
2743 	struct rpc_task_setup task_setup_data = {
2744 		.rpc_client = server->client,
2745 		.rpc_message = &msg,
2746 		.callback_ops = &nfs4_close_ops,
2747 		.workqueue = nfsiod_workqueue,
2748 		.flags = RPC_TASK_ASYNC,
2749 	};
2750 	int status = -ENOMEM;
2751 
2752 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2753 		&task_setup_data.rpc_client, &msg);
2754 
2755 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2756 	if (calldata == NULL)
2757 		goto out;
2758 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2759 	calldata->inode = state->inode;
2760 	calldata->state = state;
2761 	calldata->arg.fh = NFS_FH(state->inode);
2762 	calldata->arg.stateid = &state->open_stateid;
2763 	/* Serialization for the sequence id */
2764 	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2765 	if (calldata->arg.seqid == NULL)
2766 		goto out_free_calldata;
2767 	calldata->arg.fmode = 0;
2768 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2769 	calldata->res.fattr = &calldata->fattr;
2770 	calldata->res.seqid = calldata->arg.seqid;
2771 	calldata->res.server = server;
2772 	calldata->roc = nfs4_roc(state->inode);
2773 	nfs_sb_active(calldata->inode->i_sb);
2774 
2775 	msg.rpc_argp = &calldata->arg;
2776 	msg.rpc_resp = &calldata->res;
2777 	task_setup_data.callback_data = calldata;
2778 	task = rpc_run_task(&task_setup_data);
2779 	if (IS_ERR(task))
2780 		return PTR_ERR(task);
2781 	status = 0;
2782 	if (wait)
2783 		status = rpc_wait_for_completion_task(task);
2784 	rpc_put_task(task);
2785 	return status;
2786 out_free_calldata:
2787 	kfree(calldata);
2788 out:
2789 	nfs4_put_open_state(state);
2790 	nfs4_put_state_owner(sp);
2791 	return status;
2792 }
2793 
2794 static struct inode *
2795 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2796 		int open_flags, struct iattr *attr, int *opened)
2797 {
2798 	struct nfs4_state *state;
2799 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2800 
2801 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2802 
2803 	/* Protect against concurrent sillydeletes */
2804 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2805 
2806 	nfs4_label_release_security(label);
2807 
2808 	if (IS_ERR(state))
2809 		return ERR_CAST(state);
2810 	return state->inode;
2811 }
2812 
2813 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2814 {
2815 	if (ctx->state == NULL)
2816 		return;
2817 	if (is_sync)
2818 		nfs4_close_sync(ctx->state, ctx->mode);
2819 	else
2820 		nfs4_close_state(ctx->state, ctx->mode);
2821 }
2822 
2823 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2824 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2825 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2826 
2827 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2828 {
2829 	struct nfs4_server_caps_arg args = {
2830 		.fhandle = fhandle,
2831 	};
2832 	struct nfs4_server_caps_res res = {};
2833 	struct rpc_message msg = {
2834 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2835 		.rpc_argp = &args,
2836 		.rpc_resp = &res,
2837 	};
2838 	int status;
2839 
2840 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2841 	if (status == 0) {
2842 		/* Sanity check the server answers */
2843 		switch (server->nfs_client->cl_minorversion) {
2844 		case 0:
2845 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2846 			res.attr_bitmask[2] = 0;
2847 			break;
2848 		case 1:
2849 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2850 			break;
2851 		case 2:
2852 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2853 		}
2854 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2855 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2856 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2857 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2858 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2859 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2860 				NFS_CAP_SECURITY_LABEL);
2861 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2862 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2863 			server->caps |= NFS_CAP_ACLS;
2864 		if (res.has_links != 0)
2865 			server->caps |= NFS_CAP_HARDLINKS;
2866 		if (res.has_symlinks != 0)
2867 			server->caps |= NFS_CAP_SYMLINKS;
2868 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2869 			server->caps |= NFS_CAP_FILEID;
2870 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2871 			server->caps |= NFS_CAP_MODE;
2872 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2873 			server->caps |= NFS_CAP_NLINK;
2874 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2875 			server->caps |= NFS_CAP_OWNER;
2876 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2877 			server->caps |= NFS_CAP_OWNER_GROUP;
2878 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2879 			server->caps |= NFS_CAP_ATIME;
2880 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2881 			server->caps |= NFS_CAP_CTIME;
2882 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2883 			server->caps |= NFS_CAP_MTIME;
2884 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2885 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2886 			server->caps |= NFS_CAP_SECURITY_LABEL;
2887 #endif
2888 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2889 				sizeof(server->attr_bitmask));
2890 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2891 
2892 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2893 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2894 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2895 		server->cache_consistency_bitmask[2] = 0;
2896 		server->acl_bitmask = res.acl_bitmask;
2897 		server->fh_expire_type = res.fh_expire_type;
2898 	}
2899 
2900 	return status;
2901 }
2902 
2903 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2904 {
2905 	struct nfs4_exception exception = { };
2906 	int err;
2907 	do {
2908 		err = nfs4_handle_exception(server,
2909 				_nfs4_server_capabilities(server, fhandle),
2910 				&exception);
2911 	} while (exception.retry);
2912 	return err;
2913 }
2914 
2915 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2916 		struct nfs_fsinfo *info)
2917 {
2918 	u32 bitmask[3];
2919 	struct nfs4_lookup_root_arg args = {
2920 		.bitmask = bitmask,
2921 	};
2922 	struct nfs4_lookup_res res = {
2923 		.server = server,
2924 		.fattr = info->fattr,
2925 		.fh = fhandle,
2926 	};
2927 	struct rpc_message msg = {
2928 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2929 		.rpc_argp = &args,
2930 		.rpc_resp = &res,
2931 	};
2932 
2933 	bitmask[0] = nfs4_fattr_bitmap[0];
2934 	bitmask[1] = nfs4_fattr_bitmap[1];
2935 	/*
2936 	 * Process the label in the upcoming getfattr
2937 	 */
2938 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2939 
2940 	nfs_fattr_init(info->fattr);
2941 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2942 }
2943 
2944 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2945 		struct nfs_fsinfo *info)
2946 {
2947 	struct nfs4_exception exception = { };
2948 	int err;
2949 	do {
2950 		err = _nfs4_lookup_root(server, fhandle, info);
2951 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2952 		switch (err) {
2953 		case 0:
2954 		case -NFS4ERR_WRONGSEC:
2955 			goto out;
2956 		default:
2957 			err = nfs4_handle_exception(server, err, &exception);
2958 		}
2959 	} while (exception.retry);
2960 out:
2961 	return err;
2962 }
2963 
2964 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2965 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2966 {
2967 	struct rpc_auth_create_args auth_args = {
2968 		.pseudoflavor = flavor,
2969 	};
2970 	struct rpc_auth *auth;
2971 	int ret;
2972 
2973 	auth = rpcauth_create(&auth_args, server->client);
2974 	if (IS_ERR(auth)) {
2975 		ret = -EACCES;
2976 		goto out;
2977 	}
2978 	ret = nfs4_lookup_root(server, fhandle, info);
2979 out:
2980 	return ret;
2981 }
2982 
2983 /*
2984  * Retry pseudoroot lookup with various security flavors.  We do this when:
2985  *
2986  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2987  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2988  *
2989  * Returns zero on success, or a negative NFS4ERR value, or a
2990  * negative errno value.
2991  */
2992 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2993 			      struct nfs_fsinfo *info)
2994 {
2995 	/* Per 3530bis 15.33.5 */
2996 	static const rpc_authflavor_t flav_array[] = {
2997 		RPC_AUTH_GSS_KRB5P,
2998 		RPC_AUTH_GSS_KRB5I,
2999 		RPC_AUTH_GSS_KRB5,
3000 		RPC_AUTH_UNIX,			/* courtesy */
3001 		RPC_AUTH_NULL,
3002 	};
3003 	int status = -EPERM;
3004 	size_t i;
3005 
3006 	if (server->auth_info.flavor_len > 0) {
3007 		/* try each flavor specified by user */
3008 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3009 			status = nfs4_lookup_root_sec(server, fhandle, info,
3010 						server->auth_info.flavors[i]);
3011 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3012 				continue;
3013 			break;
3014 		}
3015 	} else {
3016 		/* no flavors specified by user, try default list */
3017 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3018 			status = nfs4_lookup_root_sec(server, fhandle, info,
3019 						      flav_array[i]);
3020 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3021 				continue;
3022 			break;
3023 		}
3024 	}
3025 
3026 	/*
3027 	 * -EACCESS could mean that the user doesn't have correct permissions
3028 	 * to access the mount.  It could also mean that we tried to mount
3029 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3030 	 * existing mount programs don't handle -EACCES very well so it should
3031 	 * be mapped to -EPERM instead.
3032 	 */
3033 	if (status == -EACCES)
3034 		status = -EPERM;
3035 	return status;
3036 }
3037 
3038 static int nfs4_do_find_root_sec(struct nfs_server *server,
3039 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3040 {
3041 	int mv = server->nfs_client->cl_minorversion;
3042 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3043 }
3044 
3045 /**
3046  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3047  * @server: initialized nfs_server handle
3048  * @fhandle: we fill in the pseudo-fs root file handle
3049  * @info: we fill in an FSINFO struct
3050  * @auth_probe: probe the auth flavours
3051  *
3052  * Returns zero on success, or a negative errno.
3053  */
3054 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3055 			 struct nfs_fsinfo *info,
3056 			 bool auth_probe)
3057 {
3058 	int status;
3059 
3060 	switch (auth_probe) {
3061 	case false:
3062 		status = nfs4_lookup_root(server, fhandle, info);
3063 		if (status != -NFS4ERR_WRONGSEC)
3064 			break;
3065 	default:
3066 		status = nfs4_do_find_root_sec(server, fhandle, info);
3067 	}
3068 
3069 	if (status == 0)
3070 		status = nfs4_server_capabilities(server, fhandle);
3071 	if (status == 0)
3072 		status = nfs4_do_fsinfo(server, fhandle, info);
3073 
3074 	return nfs4_map_errors(status);
3075 }
3076 
3077 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3078 			      struct nfs_fsinfo *info)
3079 {
3080 	int error;
3081 	struct nfs_fattr *fattr = info->fattr;
3082 	struct nfs4_label *label = NULL;
3083 
3084 	error = nfs4_server_capabilities(server, mntfh);
3085 	if (error < 0) {
3086 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3087 		return error;
3088 	}
3089 
3090 	label = nfs4_label_alloc(server, GFP_KERNEL);
3091 	if (IS_ERR(label))
3092 		return PTR_ERR(label);
3093 
3094 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3095 	if (error < 0) {
3096 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3097 		goto err_free_label;
3098 	}
3099 
3100 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3101 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3102 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3103 
3104 err_free_label:
3105 	nfs4_label_free(label);
3106 
3107 	return error;
3108 }
3109 
3110 /*
3111  * Get locations and (maybe) other attributes of a referral.
3112  * Note that we'll actually follow the referral later when
3113  * we detect fsid mismatch in inode revalidation
3114  */
3115 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3116 			     const struct qstr *name, struct nfs_fattr *fattr,
3117 			     struct nfs_fh *fhandle)
3118 {
3119 	int status = -ENOMEM;
3120 	struct page *page = NULL;
3121 	struct nfs4_fs_locations *locations = NULL;
3122 
3123 	page = alloc_page(GFP_KERNEL);
3124 	if (page == NULL)
3125 		goto out;
3126 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3127 	if (locations == NULL)
3128 		goto out;
3129 
3130 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3131 	if (status != 0)
3132 		goto out;
3133 
3134 	/*
3135 	 * If the fsid didn't change, this is a migration event, not a
3136 	 * referral.  Cause us to drop into the exception handler, which
3137 	 * will kick off migration recovery.
3138 	 */
3139 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3140 		dprintk("%s: server did not return a different fsid for"
3141 			" a referral at %s\n", __func__, name->name);
3142 		status = -NFS4ERR_MOVED;
3143 		goto out;
3144 	}
3145 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3146 	nfs_fixup_referral_attributes(&locations->fattr);
3147 
3148 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3149 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3150 	memset(fhandle, 0, sizeof(struct nfs_fh));
3151 out:
3152 	if (page)
3153 		__free_page(page);
3154 	kfree(locations);
3155 	return status;
3156 }
3157 
3158 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3159 				struct nfs_fattr *fattr, struct nfs4_label *label)
3160 {
3161 	struct nfs4_getattr_arg args = {
3162 		.fh = fhandle,
3163 		.bitmask = server->attr_bitmask,
3164 	};
3165 	struct nfs4_getattr_res res = {
3166 		.fattr = fattr,
3167 		.label = label,
3168 		.server = server,
3169 	};
3170 	struct rpc_message msg = {
3171 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3172 		.rpc_argp = &args,
3173 		.rpc_resp = &res,
3174 	};
3175 
3176 	args.bitmask = nfs4_bitmask(server, label);
3177 
3178 	nfs_fattr_init(fattr);
3179 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3180 }
3181 
3182 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3183 				struct nfs_fattr *fattr, struct nfs4_label *label)
3184 {
3185 	struct nfs4_exception exception = { };
3186 	int err;
3187 	do {
3188 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3189 		trace_nfs4_getattr(server, fhandle, fattr, err);
3190 		err = nfs4_handle_exception(server, err,
3191 				&exception);
3192 	} while (exception.retry);
3193 	return err;
3194 }
3195 
3196 /*
3197  * The file is not closed if it is opened due to the a request to change
3198  * the size of the file. The open call will not be needed once the
3199  * VFS layer lookup-intents are implemented.
3200  *
3201  * Close is called when the inode is destroyed.
3202  * If we haven't opened the file for O_WRONLY, we
3203  * need to in the size_change case to obtain a stateid.
3204  *
3205  * Got race?
3206  * Because OPEN is always done by name in nfsv4, it is
3207  * possible that we opened a different file by the same
3208  * name.  We can recognize this race condition, but we
3209  * can't do anything about it besides returning an error.
3210  *
3211  * This will be fixed with VFS changes (lookup-intent).
3212  */
3213 static int
3214 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3215 		  struct iattr *sattr)
3216 {
3217 	struct inode *inode = dentry->d_inode;
3218 	struct rpc_cred *cred = NULL;
3219 	struct nfs4_state *state = NULL;
3220 	struct nfs4_label *label = NULL;
3221 	int status;
3222 
3223 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3224 	    sattr->ia_valid & ATTR_SIZE &&
3225 	    sattr->ia_size < i_size_read(inode))
3226 		pnfs_commit_and_return_layout(inode);
3227 
3228 	nfs_fattr_init(fattr);
3229 
3230 	/* Deal with open(O_TRUNC) */
3231 	if (sattr->ia_valid & ATTR_OPEN)
3232 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3233 
3234 	/* Optimization: if the end result is no change, don't RPC */
3235 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3236 		return 0;
3237 
3238 	/* Search for an existing open(O_WRITE) file */
3239 	if (sattr->ia_valid & ATTR_FILE) {
3240 		struct nfs_open_context *ctx;
3241 
3242 		ctx = nfs_file_open_context(sattr->ia_file);
3243 		if (ctx) {
3244 			cred = ctx->cred;
3245 			state = ctx->state;
3246 		}
3247 	}
3248 
3249 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3250 	if (IS_ERR(label))
3251 		return PTR_ERR(label);
3252 
3253 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3254 	if (status == 0) {
3255 		nfs_setattr_update_inode(inode, sattr);
3256 		nfs_setsecurity(inode, fattr, label);
3257 	}
3258 	nfs4_label_free(label);
3259 	return status;
3260 }
3261 
3262 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3263 		const struct qstr *name, struct nfs_fh *fhandle,
3264 		struct nfs_fattr *fattr, struct nfs4_label *label)
3265 {
3266 	struct nfs_server *server = NFS_SERVER(dir);
3267 	int		       status;
3268 	struct nfs4_lookup_arg args = {
3269 		.bitmask = server->attr_bitmask,
3270 		.dir_fh = NFS_FH(dir),
3271 		.name = name,
3272 	};
3273 	struct nfs4_lookup_res res = {
3274 		.server = server,
3275 		.fattr = fattr,
3276 		.label = label,
3277 		.fh = fhandle,
3278 	};
3279 	struct rpc_message msg = {
3280 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3281 		.rpc_argp = &args,
3282 		.rpc_resp = &res,
3283 	};
3284 
3285 	args.bitmask = nfs4_bitmask(server, label);
3286 
3287 	nfs_fattr_init(fattr);
3288 
3289 	dprintk("NFS call  lookup %s\n", name->name);
3290 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3291 	dprintk("NFS reply lookup: %d\n", status);
3292 	return status;
3293 }
3294 
3295 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3296 {
3297 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3298 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3299 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3300 	fattr->nlink = 2;
3301 }
3302 
3303 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3304 				   struct qstr *name, struct nfs_fh *fhandle,
3305 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3306 {
3307 	struct nfs4_exception exception = { };
3308 	struct rpc_clnt *client = *clnt;
3309 	int err;
3310 	do {
3311 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3312 		trace_nfs4_lookup(dir, name, err);
3313 		switch (err) {
3314 		case -NFS4ERR_BADNAME:
3315 			err = -ENOENT;
3316 			goto out;
3317 		case -NFS4ERR_MOVED:
3318 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3319 			goto out;
3320 		case -NFS4ERR_WRONGSEC:
3321 			err = -EPERM;
3322 			if (client != *clnt)
3323 				goto out;
3324 			client = nfs4_negotiate_security(client, dir, name);
3325 			if (IS_ERR(client))
3326 				return PTR_ERR(client);
3327 
3328 			exception.retry = 1;
3329 			break;
3330 		default:
3331 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3332 		}
3333 	} while (exception.retry);
3334 
3335 out:
3336 	if (err == 0)
3337 		*clnt = client;
3338 	else if (client != *clnt)
3339 		rpc_shutdown_client(client);
3340 
3341 	return err;
3342 }
3343 
3344 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3345 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3346 			    struct nfs4_label *label)
3347 {
3348 	int status;
3349 	struct rpc_clnt *client = NFS_CLIENT(dir);
3350 
3351 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3352 	if (client != NFS_CLIENT(dir)) {
3353 		rpc_shutdown_client(client);
3354 		nfs_fixup_secinfo_attributes(fattr);
3355 	}
3356 	return status;
3357 }
3358 
3359 struct rpc_clnt *
3360 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3361 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3362 {
3363 	struct rpc_clnt *client = NFS_CLIENT(dir);
3364 	int status;
3365 
3366 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3367 	if (status < 0)
3368 		return ERR_PTR(status);
3369 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3370 }
3371 
3372 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3373 {
3374 	struct nfs_server *server = NFS_SERVER(inode);
3375 	struct nfs4_accessargs args = {
3376 		.fh = NFS_FH(inode),
3377 		.bitmask = server->cache_consistency_bitmask,
3378 	};
3379 	struct nfs4_accessres res = {
3380 		.server = server,
3381 	};
3382 	struct rpc_message msg = {
3383 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3384 		.rpc_argp = &args,
3385 		.rpc_resp = &res,
3386 		.rpc_cred = entry->cred,
3387 	};
3388 	int mode = entry->mask;
3389 	int status = 0;
3390 
3391 	/*
3392 	 * Determine which access bits we want to ask for...
3393 	 */
3394 	if (mode & MAY_READ)
3395 		args.access |= NFS4_ACCESS_READ;
3396 	if (S_ISDIR(inode->i_mode)) {
3397 		if (mode & MAY_WRITE)
3398 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3399 		if (mode & MAY_EXEC)
3400 			args.access |= NFS4_ACCESS_LOOKUP;
3401 	} else {
3402 		if (mode & MAY_WRITE)
3403 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3404 		if (mode & MAY_EXEC)
3405 			args.access |= NFS4_ACCESS_EXECUTE;
3406 	}
3407 
3408 	res.fattr = nfs_alloc_fattr();
3409 	if (res.fattr == NULL)
3410 		return -ENOMEM;
3411 
3412 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3413 	if (!status) {
3414 		nfs_access_set_mask(entry, res.access);
3415 		nfs_refresh_inode(inode, res.fattr);
3416 	}
3417 	nfs_free_fattr(res.fattr);
3418 	return status;
3419 }
3420 
3421 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3422 {
3423 	struct nfs4_exception exception = { };
3424 	int err;
3425 	do {
3426 		err = _nfs4_proc_access(inode, entry);
3427 		trace_nfs4_access(inode, err);
3428 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3429 				&exception);
3430 	} while (exception.retry);
3431 	return err;
3432 }
3433 
3434 /*
3435  * TODO: For the time being, we don't try to get any attributes
3436  * along with any of the zero-copy operations READ, READDIR,
3437  * READLINK, WRITE.
3438  *
3439  * In the case of the first three, we want to put the GETATTR
3440  * after the read-type operation -- this is because it is hard
3441  * to predict the length of a GETATTR response in v4, and thus
3442  * align the READ data correctly.  This means that the GETATTR
3443  * may end up partially falling into the page cache, and we should
3444  * shift it into the 'tail' of the xdr_buf before processing.
3445  * To do this efficiently, we need to know the total length
3446  * of data received, which doesn't seem to be available outside
3447  * of the RPC layer.
3448  *
3449  * In the case of WRITE, we also want to put the GETATTR after
3450  * the operation -- in this case because we want to make sure
3451  * we get the post-operation mtime and size.
3452  *
3453  * Both of these changes to the XDR layer would in fact be quite
3454  * minor, but I decided to leave them for a subsequent patch.
3455  */
3456 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3457 		unsigned int pgbase, unsigned int pglen)
3458 {
3459 	struct nfs4_readlink args = {
3460 		.fh       = NFS_FH(inode),
3461 		.pgbase	  = pgbase,
3462 		.pglen    = pglen,
3463 		.pages    = &page,
3464 	};
3465 	struct nfs4_readlink_res res;
3466 	struct rpc_message msg = {
3467 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3468 		.rpc_argp = &args,
3469 		.rpc_resp = &res,
3470 	};
3471 
3472 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3473 }
3474 
3475 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3476 		unsigned int pgbase, unsigned int pglen)
3477 {
3478 	struct nfs4_exception exception = { };
3479 	int err;
3480 	do {
3481 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3482 		trace_nfs4_readlink(inode, err);
3483 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3484 				&exception);
3485 	} while (exception.retry);
3486 	return err;
3487 }
3488 
3489 /*
3490  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3491  */
3492 static int
3493 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3494 		 int flags)
3495 {
3496 	struct nfs4_label l, *ilabel = NULL;
3497 	struct nfs_open_context *ctx;
3498 	struct nfs4_state *state;
3499 	int opened = 0;
3500 	int status = 0;
3501 
3502 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3503 	if (IS_ERR(ctx))
3504 		return PTR_ERR(ctx);
3505 
3506 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3507 
3508 	sattr->ia_mode &= ~current_umask();
3509 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3510 	if (IS_ERR(state)) {
3511 		status = PTR_ERR(state);
3512 		goto out;
3513 	}
3514 out:
3515 	nfs4_label_release_security(ilabel);
3516 	put_nfs_open_context(ctx);
3517 	return status;
3518 }
3519 
3520 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3521 {
3522 	struct nfs_server *server = NFS_SERVER(dir);
3523 	struct nfs_removeargs args = {
3524 		.fh = NFS_FH(dir),
3525 		.name = *name,
3526 	};
3527 	struct nfs_removeres res = {
3528 		.server = server,
3529 	};
3530 	struct rpc_message msg = {
3531 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3532 		.rpc_argp = &args,
3533 		.rpc_resp = &res,
3534 	};
3535 	int status;
3536 
3537 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3538 	if (status == 0)
3539 		update_changeattr(dir, &res.cinfo);
3540 	return status;
3541 }
3542 
3543 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3544 {
3545 	struct nfs4_exception exception = { };
3546 	int err;
3547 	do {
3548 		err = _nfs4_proc_remove(dir, name);
3549 		trace_nfs4_remove(dir, name, err);
3550 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3551 				&exception);
3552 	} while (exception.retry);
3553 	return err;
3554 }
3555 
3556 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3557 {
3558 	struct nfs_server *server = NFS_SERVER(dir);
3559 	struct nfs_removeargs *args = msg->rpc_argp;
3560 	struct nfs_removeres *res = msg->rpc_resp;
3561 
3562 	res->server = server;
3563 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3564 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3565 
3566 	nfs_fattr_init(res->dir_attr);
3567 }
3568 
3569 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3570 {
3571 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3572 			&data->args.seq_args,
3573 			&data->res.seq_res,
3574 			task);
3575 }
3576 
3577 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3578 {
3579 	struct nfs_unlinkdata *data = task->tk_calldata;
3580 	struct nfs_removeres *res = &data->res;
3581 
3582 	if (!nfs4_sequence_done(task, &res->seq_res))
3583 		return 0;
3584 	if (nfs4_async_handle_error(task, res->server, NULL,
3585 				    &data->timeout) == -EAGAIN)
3586 		return 0;
3587 	update_changeattr(dir, &res->cinfo);
3588 	return 1;
3589 }
3590 
3591 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3592 {
3593 	struct nfs_server *server = NFS_SERVER(dir);
3594 	struct nfs_renameargs *arg = msg->rpc_argp;
3595 	struct nfs_renameres *res = msg->rpc_resp;
3596 
3597 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3598 	res->server = server;
3599 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3600 }
3601 
3602 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3603 {
3604 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3605 			&data->args.seq_args,
3606 			&data->res.seq_res,
3607 			task);
3608 }
3609 
3610 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3611 				 struct inode *new_dir)
3612 {
3613 	struct nfs_renamedata *data = task->tk_calldata;
3614 	struct nfs_renameres *res = &data->res;
3615 
3616 	if (!nfs4_sequence_done(task, &res->seq_res))
3617 		return 0;
3618 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3619 		return 0;
3620 
3621 	update_changeattr(old_dir, &res->old_cinfo);
3622 	update_changeattr(new_dir, &res->new_cinfo);
3623 	return 1;
3624 }
3625 
3626 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3627 {
3628 	struct nfs_server *server = NFS_SERVER(inode);
3629 	struct nfs4_link_arg arg = {
3630 		.fh     = NFS_FH(inode),
3631 		.dir_fh = NFS_FH(dir),
3632 		.name   = name,
3633 		.bitmask = server->attr_bitmask,
3634 	};
3635 	struct nfs4_link_res res = {
3636 		.server = server,
3637 		.label = NULL,
3638 	};
3639 	struct rpc_message msg = {
3640 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3641 		.rpc_argp = &arg,
3642 		.rpc_resp = &res,
3643 	};
3644 	int status = -ENOMEM;
3645 
3646 	res.fattr = nfs_alloc_fattr();
3647 	if (res.fattr == NULL)
3648 		goto out;
3649 
3650 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3651 	if (IS_ERR(res.label)) {
3652 		status = PTR_ERR(res.label);
3653 		goto out;
3654 	}
3655 	arg.bitmask = nfs4_bitmask(server, res.label);
3656 
3657 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3658 	if (!status) {
3659 		update_changeattr(dir, &res.cinfo);
3660 		status = nfs_post_op_update_inode(inode, res.fattr);
3661 		if (!status)
3662 			nfs_setsecurity(inode, res.fattr, res.label);
3663 	}
3664 
3665 
3666 	nfs4_label_free(res.label);
3667 
3668 out:
3669 	nfs_free_fattr(res.fattr);
3670 	return status;
3671 }
3672 
3673 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3674 {
3675 	struct nfs4_exception exception = { };
3676 	int err;
3677 	do {
3678 		err = nfs4_handle_exception(NFS_SERVER(inode),
3679 				_nfs4_proc_link(inode, dir, name),
3680 				&exception);
3681 	} while (exception.retry);
3682 	return err;
3683 }
3684 
3685 struct nfs4_createdata {
3686 	struct rpc_message msg;
3687 	struct nfs4_create_arg arg;
3688 	struct nfs4_create_res res;
3689 	struct nfs_fh fh;
3690 	struct nfs_fattr fattr;
3691 	struct nfs4_label *label;
3692 };
3693 
3694 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3695 		struct qstr *name, struct iattr *sattr, u32 ftype)
3696 {
3697 	struct nfs4_createdata *data;
3698 
3699 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3700 	if (data != NULL) {
3701 		struct nfs_server *server = NFS_SERVER(dir);
3702 
3703 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3704 		if (IS_ERR(data->label))
3705 			goto out_free;
3706 
3707 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3708 		data->msg.rpc_argp = &data->arg;
3709 		data->msg.rpc_resp = &data->res;
3710 		data->arg.dir_fh = NFS_FH(dir);
3711 		data->arg.server = server;
3712 		data->arg.name = name;
3713 		data->arg.attrs = sattr;
3714 		data->arg.ftype = ftype;
3715 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3716 		data->res.server = server;
3717 		data->res.fh = &data->fh;
3718 		data->res.fattr = &data->fattr;
3719 		data->res.label = data->label;
3720 		nfs_fattr_init(data->res.fattr);
3721 	}
3722 	return data;
3723 out_free:
3724 	kfree(data);
3725 	return NULL;
3726 }
3727 
3728 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3729 {
3730 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3731 				    &data->arg.seq_args, &data->res.seq_res, 1);
3732 	if (status == 0) {
3733 		update_changeattr(dir, &data->res.dir_cinfo);
3734 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3735 	}
3736 	return status;
3737 }
3738 
3739 static void nfs4_free_createdata(struct nfs4_createdata *data)
3740 {
3741 	nfs4_label_free(data->label);
3742 	kfree(data);
3743 }
3744 
3745 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3746 		struct page *page, unsigned int len, struct iattr *sattr,
3747 		struct nfs4_label *label)
3748 {
3749 	struct nfs4_createdata *data;
3750 	int status = -ENAMETOOLONG;
3751 
3752 	if (len > NFS4_MAXPATHLEN)
3753 		goto out;
3754 
3755 	status = -ENOMEM;
3756 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3757 	if (data == NULL)
3758 		goto out;
3759 
3760 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3761 	data->arg.u.symlink.pages = &page;
3762 	data->arg.u.symlink.len = len;
3763 	data->arg.label = label;
3764 
3765 	status = nfs4_do_create(dir, dentry, data);
3766 
3767 	nfs4_free_createdata(data);
3768 out:
3769 	return status;
3770 }
3771 
3772 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3773 		struct page *page, unsigned int len, struct iattr *sattr)
3774 {
3775 	struct nfs4_exception exception = { };
3776 	struct nfs4_label l, *label = NULL;
3777 	int err;
3778 
3779 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3780 
3781 	do {
3782 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3783 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3784 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3785 				&exception);
3786 	} while (exception.retry);
3787 
3788 	nfs4_label_release_security(label);
3789 	return err;
3790 }
3791 
3792 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3793 		struct iattr *sattr, struct nfs4_label *label)
3794 {
3795 	struct nfs4_createdata *data;
3796 	int status = -ENOMEM;
3797 
3798 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3799 	if (data == NULL)
3800 		goto out;
3801 
3802 	data->arg.label = label;
3803 	status = nfs4_do_create(dir, dentry, data);
3804 
3805 	nfs4_free_createdata(data);
3806 out:
3807 	return status;
3808 }
3809 
3810 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3811 		struct iattr *sattr)
3812 {
3813 	struct nfs4_exception exception = { };
3814 	struct nfs4_label l, *label = NULL;
3815 	int err;
3816 
3817 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3818 
3819 	sattr->ia_mode &= ~current_umask();
3820 	do {
3821 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3822 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3823 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3824 				&exception);
3825 	} while (exception.retry);
3826 	nfs4_label_release_security(label);
3827 
3828 	return err;
3829 }
3830 
3831 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3832 		u64 cookie, struct page **pages, unsigned int count, int plus)
3833 {
3834 	struct inode		*dir = dentry->d_inode;
3835 	struct nfs4_readdir_arg args = {
3836 		.fh = NFS_FH(dir),
3837 		.pages = pages,
3838 		.pgbase = 0,
3839 		.count = count,
3840 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3841 		.plus = plus,
3842 	};
3843 	struct nfs4_readdir_res res;
3844 	struct rpc_message msg = {
3845 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3846 		.rpc_argp = &args,
3847 		.rpc_resp = &res,
3848 		.rpc_cred = cred,
3849 	};
3850 	int			status;
3851 
3852 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3853 			dentry,
3854 			(unsigned long long)cookie);
3855 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3856 	res.pgbase = args.pgbase;
3857 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3858 	if (status >= 0) {
3859 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3860 		status += args.pgbase;
3861 	}
3862 
3863 	nfs_invalidate_atime(dir);
3864 
3865 	dprintk("%s: returns %d\n", __func__, status);
3866 	return status;
3867 }
3868 
3869 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3870 		u64 cookie, struct page **pages, unsigned int count, int plus)
3871 {
3872 	struct nfs4_exception exception = { };
3873 	int err;
3874 	do {
3875 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3876 				pages, count, plus);
3877 		trace_nfs4_readdir(dentry->d_inode, err);
3878 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3879 				&exception);
3880 	} while (exception.retry);
3881 	return err;
3882 }
3883 
3884 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3885 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3886 {
3887 	struct nfs4_createdata *data;
3888 	int mode = sattr->ia_mode;
3889 	int status = -ENOMEM;
3890 
3891 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3892 	if (data == NULL)
3893 		goto out;
3894 
3895 	if (S_ISFIFO(mode))
3896 		data->arg.ftype = NF4FIFO;
3897 	else if (S_ISBLK(mode)) {
3898 		data->arg.ftype = NF4BLK;
3899 		data->arg.u.device.specdata1 = MAJOR(rdev);
3900 		data->arg.u.device.specdata2 = MINOR(rdev);
3901 	}
3902 	else if (S_ISCHR(mode)) {
3903 		data->arg.ftype = NF4CHR;
3904 		data->arg.u.device.specdata1 = MAJOR(rdev);
3905 		data->arg.u.device.specdata2 = MINOR(rdev);
3906 	} else if (!S_ISSOCK(mode)) {
3907 		status = -EINVAL;
3908 		goto out_free;
3909 	}
3910 
3911 	data->arg.label = label;
3912 	status = nfs4_do_create(dir, dentry, data);
3913 out_free:
3914 	nfs4_free_createdata(data);
3915 out:
3916 	return status;
3917 }
3918 
3919 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3920 		struct iattr *sattr, dev_t rdev)
3921 {
3922 	struct nfs4_exception exception = { };
3923 	struct nfs4_label l, *label = NULL;
3924 	int err;
3925 
3926 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3927 
3928 	sattr->ia_mode &= ~current_umask();
3929 	do {
3930 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3931 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3932 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3933 				&exception);
3934 	} while (exception.retry);
3935 
3936 	nfs4_label_release_security(label);
3937 
3938 	return err;
3939 }
3940 
3941 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3942 		 struct nfs_fsstat *fsstat)
3943 {
3944 	struct nfs4_statfs_arg args = {
3945 		.fh = fhandle,
3946 		.bitmask = server->attr_bitmask,
3947 	};
3948 	struct nfs4_statfs_res res = {
3949 		.fsstat = fsstat,
3950 	};
3951 	struct rpc_message msg = {
3952 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3953 		.rpc_argp = &args,
3954 		.rpc_resp = &res,
3955 	};
3956 
3957 	nfs_fattr_init(fsstat->fattr);
3958 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3959 }
3960 
3961 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3962 {
3963 	struct nfs4_exception exception = { };
3964 	int err;
3965 	do {
3966 		err = nfs4_handle_exception(server,
3967 				_nfs4_proc_statfs(server, fhandle, fsstat),
3968 				&exception);
3969 	} while (exception.retry);
3970 	return err;
3971 }
3972 
3973 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3974 		struct nfs_fsinfo *fsinfo)
3975 {
3976 	struct nfs4_fsinfo_arg args = {
3977 		.fh = fhandle,
3978 		.bitmask = server->attr_bitmask,
3979 	};
3980 	struct nfs4_fsinfo_res res = {
3981 		.fsinfo = fsinfo,
3982 	};
3983 	struct rpc_message msg = {
3984 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3985 		.rpc_argp = &args,
3986 		.rpc_resp = &res,
3987 	};
3988 
3989 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3990 }
3991 
3992 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3993 {
3994 	struct nfs4_exception exception = { };
3995 	unsigned long now = jiffies;
3996 	int err;
3997 
3998 	do {
3999 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4000 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4001 		if (err == 0) {
4002 			struct nfs_client *clp = server->nfs_client;
4003 
4004 			spin_lock(&clp->cl_lock);
4005 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4006 			clp->cl_last_renewal = now;
4007 			spin_unlock(&clp->cl_lock);
4008 			break;
4009 		}
4010 		err = nfs4_handle_exception(server, err, &exception);
4011 	} while (exception.retry);
4012 	return err;
4013 }
4014 
4015 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4016 {
4017 	int error;
4018 
4019 	nfs_fattr_init(fsinfo->fattr);
4020 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4021 	if (error == 0) {
4022 		/* block layout checks this! */
4023 		server->pnfs_blksize = fsinfo->blksize;
4024 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4025 	}
4026 
4027 	return error;
4028 }
4029 
4030 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4031 		struct nfs_pathconf *pathconf)
4032 {
4033 	struct nfs4_pathconf_arg args = {
4034 		.fh = fhandle,
4035 		.bitmask = server->attr_bitmask,
4036 	};
4037 	struct nfs4_pathconf_res res = {
4038 		.pathconf = pathconf,
4039 	};
4040 	struct rpc_message msg = {
4041 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4042 		.rpc_argp = &args,
4043 		.rpc_resp = &res,
4044 	};
4045 
4046 	/* None of the pathconf attributes are mandatory to implement */
4047 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4048 		memset(pathconf, 0, sizeof(*pathconf));
4049 		return 0;
4050 	}
4051 
4052 	nfs_fattr_init(pathconf->fattr);
4053 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4054 }
4055 
4056 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4057 		struct nfs_pathconf *pathconf)
4058 {
4059 	struct nfs4_exception exception = { };
4060 	int err;
4061 
4062 	do {
4063 		err = nfs4_handle_exception(server,
4064 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4065 				&exception);
4066 	} while (exception.retry);
4067 	return err;
4068 }
4069 
4070 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4071 		const struct nfs_open_context *ctx,
4072 		const struct nfs_lock_context *l_ctx,
4073 		fmode_t fmode)
4074 {
4075 	const struct nfs_lockowner *lockowner = NULL;
4076 
4077 	if (l_ctx != NULL)
4078 		lockowner = &l_ctx->lockowner;
4079 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4080 }
4081 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4082 
4083 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4084 		const struct nfs_open_context *ctx,
4085 		const struct nfs_lock_context *l_ctx,
4086 		fmode_t fmode)
4087 {
4088 	nfs4_stateid current_stateid;
4089 
4090 	/* If the current stateid represents a lost lock, then exit */
4091 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4092 		return true;
4093 	return nfs4_stateid_match(stateid, &current_stateid);
4094 }
4095 
4096 static bool nfs4_error_stateid_expired(int err)
4097 {
4098 	switch (err) {
4099 	case -NFS4ERR_DELEG_REVOKED:
4100 	case -NFS4ERR_ADMIN_REVOKED:
4101 	case -NFS4ERR_BAD_STATEID:
4102 	case -NFS4ERR_STALE_STATEID:
4103 	case -NFS4ERR_OLD_STATEID:
4104 	case -NFS4ERR_OPENMODE:
4105 	case -NFS4ERR_EXPIRED:
4106 		return true;
4107 	}
4108 	return false;
4109 }
4110 
4111 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4112 {
4113 	nfs_invalidate_atime(hdr->inode);
4114 }
4115 
4116 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4117 {
4118 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4119 
4120 	trace_nfs4_read(hdr, task->tk_status);
4121 	if (nfs4_async_handle_error(task, server,
4122 				    hdr->args.context->state,
4123 				    NULL) == -EAGAIN) {
4124 		rpc_restart_call_prepare(task);
4125 		return -EAGAIN;
4126 	}
4127 
4128 	__nfs4_read_done_cb(hdr);
4129 	if (task->tk_status > 0)
4130 		renew_lease(server, hdr->timestamp);
4131 	return 0;
4132 }
4133 
4134 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4135 		struct nfs_pgio_args *args)
4136 {
4137 
4138 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4139 		nfs4_stateid_is_current(&args->stateid,
4140 				args->context,
4141 				args->lock_context,
4142 				FMODE_READ))
4143 		return false;
4144 	rpc_restart_call_prepare(task);
4145 	return true;
4146 }
4147 
4148 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4149 {
4150 
4151 	dprintk("--> %s\n", __func__);
4152 
4153 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4154 		return -EAGAIN;
4155 	if (nfs4_read_stateid_changed(task, &hdr->args))
4156 		return -EAGAIN;
4157 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4158 				    nfs4_read_done_cb(task, hdr);
4159 }
4160 
4161 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4162 				 struct rpc_message *msg)
4163 {
4164 	hdr->timestamp   = jiffies;
4165 	hdr->pgio_done_cb = nfs4_read_done_cb;
4166 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4167 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4168 }
4169 
4170 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4171 				      struct nfs_pgio_header *hdr)
4172 {
4173 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4174 			&hdr->args.seq_args,
4175 			&hdr->res.seq_res,
4176 			task))
4177 		return 0;
4178 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4179 				hdr->args.lock_context,
4180 				hdr->rw_ops->rw_mode) == -EIO)
4181 		return -EIO;
4182 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4183 		return -EIO;
4184 	return 0;
4185 }
4186 
4187 static int nfs4_write_done_cb(struct rpc_task *task,
4188 			      struct nfs_pgio_header *hdr)
4189 {
4190 	struct inode *inode = hdr->inode;
4191 
4192 	trace_nfs4_write(hdr, task->tk_status);
4193 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4194 				    hdr->args.context->state,
4195 				    NULL) == -EAGAIN) {
4196 		rpc_restart_call_prepare(task);
4197 		return -EAGAIN;
4198 	}
4199 	if (task->tk_status >= 0) {
4200 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4201 		nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4202 	}
4203 	return 0;
4204 }
4205 
4206 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4207 		struct nfs_pgio_args *args)
4208 {
4209 
4210 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4211 		nfs4_stateid_is_current(&args->stateid,
4212 				args->context,
4213 				args->lock_context,
4214 				FMODE_WRITE))
4215 		return false;
4216 	rpc_restart_call_prepare(task);
4217 	return true;
4218 }
4219 
4220 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4221 {
4222 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4223 		return -EAGAIN;
4224 	if (nfs4_write_stateid_changed(task, &hdr->args))
4225 		return -EAGAIN;
4226 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4227 		nfs4_write_done_cb(task, hdr);
4228 }
4229 
4230 static
4231 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4232 {
4233 	/* Don't request attributes for pNFS or O_DIRECT writes */
4234 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4235 		return false;
4236 	/* Otherwise, request attributes if and only if we don't hold
4237 	 * a delegation
4238 	 */
4239 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4240 }
4241 
4242 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4243 				  struct rpc_message *msg)
4244 {
4245 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4246 
4247 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4248 		hdr->args.bitmask = NULL;
4249 		hdr->res.fattr = NULL;
4250 	} else
4251 		hdr->args.bitmask = server->cache_consistency_bitmask;
4252 
4253 	if (!hdr->pgio_done_cb)
4254 		hdr->pgio_done_cb = nfs4_write_done_cb;
4255 	hdr->res.server = server;
4256 	hdr->timestamp   = jiffies;
4257 
4258 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4259 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4260 }
4261 
4262 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4263 {
4264 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4265 			&data->args.seq_args,
4266 			&data->res.seq_res,
4267 			task);
4268 }
4269 
4270 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4271 {
4272 	struct inode *inode = data->inode;
4273 
4274 	trace_nfs4_commit(data, task->tk_status);
4275 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4276 				    NULL, NULL) == -EAGAIN) {
4277 		rpc_restart_call_prepare(task);
4278 		return -EAGAIN;
4279 	}
4280 	return 0;
4281 }
4282 
4283 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4284 {
4285 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4286 		return -EAGAIN;
4287 	return data->commit_done_cb(task, data);
4288 }
4289 
4290 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4291 {
4292 	struct nfs_server *server = NFS_SERVER(data->inode);
4293 
4294 	if (data->commit_done_cb == NULL)
4295 		data->commit_done_cb = nfs4_commit_done_cb;
4296 	data->res.server = server;
4297 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4298 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4299 }
4300 
4301 struct nfs4_renewdata {
4302 	struct nfs_client	*client;
4303 	unsigned long		timestamp;
4304 };
4305 
4306 /*
4307  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4308  * standalone procedure for queueing an asynchronous RENEW.
4309  */
4310 static void nfs4_renew_release(void *calldata)
4311 {
4312 	struct nfs4_renewdata *data = calldata;
4313 	struct nfs_client *clp = data->client;
4314 
4315 	if (atomic_read(&clp->cl_count) > 1)
4316 		nfs4_schedule_state_renewal(clp);
4317 	nfs_put_client(clp);
4318 	kfree(data);
4319 }
4320 
4321 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4322 {
4323 	struct nfs4_renewdata *data = calldata;
4324 	struct nfs_client *clp = data->client;
4325 	unsigned long timestamp = data->timestamp;
4326 
4327 	trace_nfs4_renew_async(clp, task->tk_status);
4328 	switch (task->tk_status) {
4329 	case 0:
4330 		break;
4331 	case -NFS4ERR_LEASE_MOVED:
4332 		nfs4_schedule_lease_moved_recovery(clp);
4333 		break;
4334 	default:
4335 		/* Unless we're shutting down, schedule state recovery! */
4336 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4337 			return;
4338 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4339 			nfs4_schedule_lease_recovery(clp);
4340 			return;
4341 		}
4342 		nfs4_schedule_path_down_recovery(clp);
4343 	}
4344 	do_renew_lease(clp, timestamp);
4345 }
4346 
4347 static const struct rpc_call_ops nfs4_renew_ops = {
4348 	.rpc_call_done = nfs4_renew_done,
4349 	.rpc_release = nfs4_renew_release,
4350 };
4351 
4352 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4353 {
4354 	struct rpc_message msg = {
4355 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4356 		.rpc_argp	= clp,
4357 		.rpc_cred	= cred,
4358 	};
4359 	struct nfs4_renewdata *data;
4360 
4361 	if (renew_flags == 0)
4362 		return 0;
4363 	if (!atomic_inc_not_zero(&clp->cl_count))
4364 		return -EIO;
4365 	data = kmalloc(sizeof(*data), GFP_NOFS);
4366 	if (data == NULL)
4367 		return -ENOMEM;
4368 	data->client = clp;
4369 	data->timestamp = jiffies;
4370 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4371 			&nfs4_renew_ops, data);
4372 }
4373 
4374 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4375 {
4376 	struct rpc_message msg = {
4377 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4378 		.rpc_argp	= clp,
4379 		.rpc_cred	= cred,
4380 	};
4381 	unsigned long now = jiffies;
4382 	int status;
4383 
4384 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4385 	if (status < 0)
4386 		return status;
4387 	do_renew_lease(clp, now);
4388 	return 0;
4389 }
4390 
4391 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4392 {
4393 	return server->caps & NFS_CAP_ACLS;
4394 }
4395 
4396 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4397  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4398  * the stack.
4399  */
4400 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4401 
4402 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4403 		struct page **pages, unsigned int *pgbase)
4404 {
4405 	struct page *newpage, **spages;
4406 	int rc = 0;
4407 	size_t len;
4408 	spages = pages;
4409 
4410 	do {
4411 		len = min_t(size_t, PAGE_SIZE, buflen);
4412 		newpage = alloc_page(GFP_KERNEL);
4413 
4414 		if (newpage == NULL)
4415 			goto unwind;
4416 		memcpy(page_address(newpage), buf, len);
4417                 buf += len;
4418                 buflen -= len;
4419 		*pages++ = newpage;
4420 		rc++;
4421 	} while (buflen != 0);
4422 
4423 	return rc;
4424 
4425 unwind:
4426 	for(; rc > 0; rc--)
4427 		__free_page(spages[rc-1]);
4428 	return -ENOMEM;
4429 }
4430 
4431 struct nfs4_cached_acl {
4432 	int cached;
4433 	size_t len;
4434 	char data[0];
4435 };
4436 
4437 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4438 {
4439 	struct nfs_inode *nfsi = NFS_I(inode);
4440 
4441 	spin_lock(&inode->i_lock);
4442 	kfree(nfsi->nfs4_acl);
4443 	nfsi->nfs4_acl = acl;
4444 	spin_unlock(&inode->i_lock);
4445 }
4446 
4447 static void nfs4_zap_acl_attr(struct inode *inode)
4448 {
4449 	nfs4_set_cached_acl(inode, NULL);
4450 }
4451 
4452 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4453 {
4454 	struct nfs_inode *nfsi = NFS_I(inode);
4455 	struct nfs4_cached_acl *acl;
4456 	int ret = -ENOENT;
4457 
4458 	spin_lock(&inode->i_lock);
4459 	acl = nfsi->nfs4_acl;
4460 	if (acl == NULL)
4461 		goto out;
4462 	if (buf == NULL) /* user is just asking for length */
4463 		goto out_len;
4464 	if (acl->cached == 0)
4465 		goto out;
4466 	ret = -ERANGE; /* see getxattr(2) man page */
4467 	if (acl->len > buflen)
4468 		goto out;
4469 	memcpy(buf, acl->data, acl->len);
4470 out_len:
4471 	ret = acl->len;
4472 out:
4473 	spin_unlock(&inode->i_lock);
4474 	return ret;
4475 }
4476 
4477 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4478 {
4479 	struct nfs4_cached_acl *acl;
4480 	size_t buflen = sizeof(*acl) + acl_len;
4481 
4482 	if (buflen <= PAGE_SIZE) {
4483 		acl = kmalloc(buflen, GFP_KERNEL);
4484 		if (acl == NULL)
4485 			goto out;
4486 		acl->cached = 1;
4487 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4488 	} else {
4489 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4490 		if (acl == NULL)
4491 			goto out;
4492 		acl->cached = 0;
4493 	}
4494 	acl->len = acl_len;
4495 out:
4496 	nfs4_set_cached_acl(inode, acl);
4497 }
4498 
4499 /*
4500  * The getxattr API returns the required buffer length when called with a
4501  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4502  * the required buf.  On a NULL buf, we send a page of data to the server
4503  * guessing that the ACL request can be serviced by a page. If so, we cache
4504  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4505  * the cache. If not so, we throw away the page, and cache the required
4506  * length. The next getxattr call will then produce another round trip to
4507  * the server, this time with the input buf of the required size.
4508  */
4509 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4510 {
4511 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4512 	struct nfs_getaclargs args = {
4513 		.fh = NFS_FH(inode),
4514 		.acl_pages = pages,
4515 		.acl_len = buflen,
4516 	};
4517 	struct nfs_getaclres res = {
4518 		.acl_len = buflen,
4519 	};
4520 	struct rpc_message msg = {
4521 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4522 		.rpc_argp = &args,
4523 		.rpc_resp = &res,
4524 	};
4525 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4526 	int ret = -ENOMEM, i;
4527 
4528 	/* As long as we're doing a round trip to the server anyway,
4529 	 * let's be prepared for a page of acl data. */
4530 	if (npages == 0)
4531 		npages = 1;
4532 	if (npages > ARRAY_SIZE(pages))
4533 		return -ERANGE;
4534 
4535 	for (i = 0; i < npages; i++) {
4536 		pages[i] = alloc_page(GFP_KERNEL);
4537 		if (!pages[i])
4538 			goto out_free;
4539 	}
4540 
4541 	/* for decoding across pages */
4542 	res.acl_scratch = alloc_page(GFP_KERNEL);
4543 	if (!res.acl_scratch)
4544 		goto out_free;
4545 
4546 	args.acl_len = npages * PAGE_SIZE;
4547 	args.acl_pgbase = 0;
4548 
4549 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4550 		__func__, buf, buflen, npages, args.acl_len);
4551 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4552 			     &msg, &args.seq_args, &res.seq_res, 0);
4553 	if (ret)
4554 		goto out_free;
4555 
4556 	/* Handle the case where the passed-in buffer is too short */
4557 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4558 		/* Did the user only issue a request for the acl length? */
4559 		if (buf == NULL)
4560 			goto out_ok;
4561 		ret = -ERANGE;
4562 		goto out_free;
4563 	}
4564 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4565 	if (buf) {
4566 		if (res.acl_len > buflen) {
4567 			ret = -ERANGE;
4568 			goto out_free;
4569 		}
4570 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4571 	}
4572 out_ok:
4573 	ret = res.acl_len;
4574 out_free:
4575 	for (i = 0; i < npages; i++)
4576 		if (pages[i])
4577 			__free_page(pages[i]);
4578 	if (res.acl_scratch)
4579 		__free_page(res.acl_scratch);
4580 	return ret;
4581 }
4582 
4583 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4584 {
4585 	struct nfs4_exception exception = { };
4586 	ssize_t ret;
4587 	do {
4588 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4589 		trace_nfs4_get_acl(inode, ret);
4590 		if (ret >= 0)
4591 			break;
4592 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4593 	} while (exception.retry);
4594 	return ret;
4595 }
4596 
4597 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4598 {
4599 	struct nfs_server *server = NFS_SERVER(inode);
4600 	int ret;
4601 
4602 	if (!nfs4_server_supports_acls(server))
4603 		return -EOPNOTSUPP;
4604 	ret = nfs_revalidate_inode(server, inode);
4605 	if (ret < 0)
4606 		return ret;
4607 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4608 		nfs_zap_acl_cache(inode);
4609 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4610 	if (ret != -ENOENT)
4611 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4612 		 * but no cached acl data, just the acl length */
4613 		return ret;
4614 	return nfs4_get_acl_uncached(inode, buf, buflen);
4615 }
4616 
4617 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4618 {
4619 	struct nfs_server *server = NFS_SERVER(inode);
4620 	struct page *pages[NFS4ACL_MAXPAGES];
4621 	struct nfs_setaclargs arg = {
4622 		.fh		= NFS_FH(inode),
4623 		.acl_pages	= pages,
4624 		.acl_len	= buflen,
4625 	};
4626 	struct nfs_setaclres res;
4627 	struct rpc_message msg = {
4628 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4629 		.rpc_argp	= &arg,
4630 		.rpc_resp	= &res,
4631 	};
4632 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4633 	int ret, i;
4634 
4635 	if (!nfs4_server_supports_acls(server))
4636 		return -EOPNOTSUPP;
4637 	if (npages > ARRAY_SIZE(pages))
4638 		return -ERANGE;
4639 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4640 	if (i < 0)
4641 		return i;
4642 	nfs4_inode_return_delegation(inode);
4643 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4644 
4645 	/*
4646 	 * Free each page after tx, so the only ref left is
4647 	 * held by the network stack
4648 	 */
4649 	for (; i > 0; i--)
4650 		put_page(pages[i-1]);
4651 
4652 	/*
4653 	 * Acl update can result in inode attribute update.
4654 	 * so mark the attribute cache invalid.
4655 	 */
4656 	spin_lock(&inode->i_lock);
4657 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4658 	spin_unlock(&inode->i_lock);
4659 	nfs_access_zap_cache(inode);
4660 	nfs_zap_acl_cache(inode);
4661 	return ret;
4662 }
4663 
4664 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4665 {
4666 	struct nfs4_exception exception = { };
4667 	int err;
4668 	do {
4669 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4670 		trace_nfs4_set_acl(inode, err);
4671 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4672 				&exception);
4673 	} while (exception.retry);
4674 	return err;
4675 }
4676 
4677 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4678 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4679 					size_t buflen)
4680 {
4681 	struct nfs_server *server = NFS_SERVER(inode);
4682 	struct nfs_fattr fattr;
4683 	struct nfs4_label label = {0, 0, buflen, buf};
4684 
4685 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4686 	struct nfs4_getattr_arg arg = {
4687 		.fh		= NFS_FH(inode),
4688 		.bitmask	= bitmask,
4689 	};
4690 	struct nfs4_getattr_res res = {
4691 		.fattr		= &fattr,
4692 		.label		= &label,
4693 		.server		= server,
4694 	};
4695 	struct rpc_message msg = {
4696 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4697 		.rpc_argp	= &arg,
4698 		.rpc_resp	= &res,
4699 	};
4700 	int ret;
4701 
4702 	nfs_fattr_init(&fattr);
4703 
4704 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4705 	if (ret)
4706 		return ret;
4707 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4708 		return -ENOENT;
4709 	if (buflen < label.len)
4710 		return -ERANGE;
4711 	return 0;
4712 }
4713 
4714 static int nfs4_get_security_label(struct inode *inode, void *buf,
4715 					size_t buflen)
4716 {
4717 	struct nfs4_exception exception = { };
4718 	int err;
4719 
4720 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4721 		return -EOPNOTSUPP;
4722 
4723 	do {
4724 		err = _nfs4_get_security_label(inode, buf, buflen);
4725 		trace_nfs4_get_security_label(inode, err);
4726 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4727 				&exception);
4728 	} while (exception.retry);
4729 	return err;
4730 }
4731 
4732 static int _nfs4_do_set_security_label(struct inode *inode,
4733 		struct nfs4_label *ilabel,
4734 		struct nfs_fattr *fattr,
4735 		struct nfs4_label *olabel)
4736 {
4737 
4738 	struct iattr sattr = {0};
4739 	struct nfs_server *server = NFS_SERVER(inode);
4740 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4741 	struct nfs_setattrargs arg = {
4742 		.fh             = NFS_FH(inode),
4743 		.iap            = &sattr,
4744 		.server		= server,
4745 		.bitmask	= bitmask,
4746 		.label		= ilabel,
4747 	};
4748 	struct nfs_setattrres res = {
4749 		.fattr		= fattr,
4750 		.label		= olabel,
4751 		.server		= server,
4752 	};
4753 	struct rpc_message msg = {
4754 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4755 		.rpc_argp       = &arg,
4756 		.rpc_resp       = &res,
4757 	};
4758 	int status;
4759 
4760 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4761 
4762 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4763 	if (status)
4764 		dprintk("%s failed: %d\n", __func__, status);
4765 
4766 	return status;
4767 }
4768 
4769 static int nfs4_do_set_security_label(struct inode *inode,
4770 		struct nfs4_label *ilabel,
4771 		struct nfs_fattr *fattr,
4772 		struct nfs4_label *olabel)
4773 {
4774 	struct nfs4_exception exception = { };
4775 	int err;
4776 
4777 	do {
4778 		err = _nfs4_do_set_security_label(inode, ilabel,
4779 				fattr, olabel);
4780 		trace_nfs4_set_security_label(inode, err);
4781 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4782 				&exception);
4783 	} while (exception.retry);
4784 	return err;
4785 }
4786 
4787 static int
4788 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4789 {
4790 	struct nfs4_label ilabel, *olabel = NULL;
4791 	struct nfs_fattr fattr;
4792 	struct rpc_cred *cred;
4793 	struct inode *inode = dentry->d_inode;
4794 	int status;
4795 
4796 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4797 		return -EOPNOTSUPP;
4798 
4799 	nfs_fattr_init(&fattr);
4800 
4801 	ilabel.pi = 0;
4802 	ilabel.lfs = 0;
4803 	ilabel.label = (char *)buf;
4804 	ilabel.len = buflen;
4805 
4806 	cred = rpc_lookup_cred();
4807 	if (IS_ERR(cred))
4808 		return PTR_ERR(cred);
4809 
4810 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4811 	if (IS_ERR(olabel)) {
4812 		status = -PTR_ERR(olabel);
4813 		goto out;
4814 	}
4815 
4816 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4817 	if (status == 0)
4818 		nfs_setsecurity(inode, &fattr, olabel);
4819 
4820 	nfs4_label_free(olabel);
4821 out:
4822 	put_rpccred(cred);
4823 	return status;
4824 }
4825 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4826 
4827 
4828 static int
4829 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4830 			struct nfs4_state *state, long *timeout)
4831 {
4832 	struct nfs_client *clp = server->nfs_client;
4833 
4834 	if (task->tk_status >= 0)
4835 		return 0;
4836 	switch(task->tk_status) {
4837 		case -NFS4ERR_DELEG_REVOKED:
4838 		case -NFS4ERR_ADMIN_REVOKED:
4839 		case -NFS4ERR_BAD_STATEID:
4840 		case -NFS4ERR_OPENMODE:
4841 			if (state == NULL)
4842 				break;
4843 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4844 				goto recovery_failed;
4845 			goto wait_on_recovery;
4846 		case -NFS4ERR_EXPIRED:
4847 			if (state != NULL) {
4848 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4849 					goto recovery_failed;
4850 			}
4851 		case -NFS4ERR_STALE_STATEID:
4852 		case -NFS4ERR_STALE_CLIENTID:
4853 			nfs4_schedule_lease_recovery(clp);
4854 			goto wait_on_recovery;
4855 		case -NFS4ERR_MOVED:
4856 			if (nfs4_schedule_migration_recovery(server) < 0)
4857 				goto recovery_failed;
4858 			goto wait_on_recovery;
4859 		case -NFS4ERR_LEASE_MOVED:
4860 			nfs4_schedule_lease_moved_recovery(clp);
4861 			goto wait_on_recovery;
4862 #if defined(CONFIG_NFS_V4_1)
4863 		case -NFS4ERR_BADSESSION:
4864 		case -NFS4ERR_BADSLOT:
4865 		case -NFS4ERR_BAD_HIGH_SLOT:
4866 		case -NFS4ERR_DEADSESSION:
4867 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4868 		case -NFS4ERR_SEQ_FALSE_RETRY:
4869 		case -NFS4ERR_SEQ_MISORDERED:
4870 			dprintk("%s ERROR %d, Reset session\n", __func__,
4871 				task->tk_status);
4872 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4873 			goto wait_on_recovery;
4874 #endif /* CONFIG_NFS_V4_1 */
4875 		case -NFS4ERR_DELAY:
4876 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4877 			rpc_delay(task, nfs4_update_delay(timeout));
4878 			goto restart_call;
4879 		case -NFS4ERR_GRACE:
4880 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4881 		case -NFS4ERR_RETRY_UNCACHED_REP:
4882 		case -NFS4ERR_OLD_STATEID:
4883 			goto restart_call;
4884 	}
4885 	task->tk_status = nfs4_map_errors(task->tk_status);
4886 	return 0;
4887 recovery_failed:
4888 	task->tk_status = -EIO;
4889 	return 0;
4890 wait_on_recovery:
4891 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4892 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4893 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4894 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4895 		goto recovery_failed;
4896 restart_call:
4897 	task->tk_status = 0;
4898 	return -EAGAIN;
4899 }
4900 
4901 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4902 				    nfs4_verifier *bootverf)
4903 {
4904 	__be32 verf[2];
4905 
4906 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4907 		/* An impossible timestamp guarantees this value
4908 		 * will never match a generated boot time. */
4909 		verf[0] = 0;
4910 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4911 	} else {
4912 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4913 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4914 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4915 	}
4916 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4917 }
4918 
4919 static unsigned int
4920 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4921 				   char *buf, size_t len)
4922 {
4923 	unsigned int result;
4924 
4925 	rcu_read_lock();
4926 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4927 				clp->cl_ipaddr,
4928 				rpc_peeraddr2str(clp->cl_rpcclient,
4929 							RPC_DISPLAY_ADDR),
4930 				rpc_peeraddr2str(clp->cl_rpcclient,
4931 							RPC_DISPLAY_PROTO));
4932 	rcu_read_unlock();
4933 	return result;
4934 }
4935 
4936 static unsigned int
4937 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4938 				char *buf, size_t len)
4939 {
4940 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4941 
4942 	if (nfs4_client_id_uniquifier[0] != '\0')
4943 		return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4944 				clp->rpc_ops->version,
4945 				clp->cl_minorversion,
4946 				nfs4_client_id_uniquifier,
4947 				nodename);
4948 	return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4949 				clp->rpc_ops->version, clp->cl_minorversion,
4950 				nodename);
4951 }
4952 
4953 /*
4954  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4955  * services.  Advertise one based on the address family of the
4956  * clientaddr.
4957  */
4958 static unsigned int
4959 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4960 {
4961 	if (strchr(clp->cl_ipaddr, ':') != NULL)
4962 		return scnprintf(buf, len, "tcp6");
4963 	else
4964 		return scnprintf(buf, len, "tcp");
4965 }
4966 
4967 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
4968 {
4969 	struct nfs4_setclientid *sc = calldata;
4970 
4971 	if (task->tk_status == 0)
4972 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
4973 }
4974 
4975 static const struct rpc_call_ops nfs4_setclientid_ops = {
4976 	.rpc_call_done = nfs4_setclientid_done,
4977 };
4978 
4979 /**
4980  * nfs4_proc_setclientid - Negotiate client ID
4981  * @clp: state data structure
4982  * @program: RPC program for NFSv4 callback service
4983  * @port: IP port number for NFS4 callback service
4984  * @cred: RPC credential to use for this call
4985  * @res: where to place the result
4986  *
4987  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4988  */
4989 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4990 		unsigned short port, struct rpc_cred *cred,
4991 		struct nfs4_setclientid_res *res)
4992 {
4993 	nfs4_verifier sc_verifier;
4994 	struct nfs4_setclientid setclientid = {
4995 		.sc_verifier = &sc_verifier,
4996 		.sc_prog = program,
4997 		.sc_cb_ident = clp->cl_cb_ident,
4998 	};
4999 	struct rpc_message msg = {
5000 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5001 		.rpc_argp = &setclientid,
5002 		.rpc_resp = res,
5003 		.rpc_cred = cred,
5004 	};
5005 	struct rpc_task *task;
5006 	struct rpc_task_setup task_setup_data = {
5007 		.rpc_client = clp->cl_rpcclient,
5008 		.rpc_message = &msg,
5009 		.callback_ops = &nfs4_setclientid_ops,
5010 		.callback_data = &setclientid,
5011 		.flags = RPC_TASK_TIMEOUT,
5012 	};
5013 	int status;
5014 
5015 	/* nfs_client_id4 */
5016 	nfs4_init_boot_verifier(clp, &sc_verifier);
5017 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5018 		setclientid.sc_name_len =
5019 				nfs4_init_uniform_client_string(clp,
5020 						setclientid.sc_name,
5021 						sizeof(setclientid.sc_name));
5022 	else
5023 		setclientid.sc_name_len =
5024 				nfs4_init_nonuniform_client_string(clp,
5025 						setclientid.sc_name,
5026 						sizeof(setclientid.sc_name));
5027 	/* cb_client4 */
5028 	setclientid.sc_netid_len =
5029 				nfs4_init_callback_netid(clp,
5030 						setclientid.sc_netid,
5031 						sizeof(setclientid.sc_netid));
5032 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5033 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5034 				clp->cl_ipaddr, port >> 8, port & 255);
5035 
5036 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5037 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5038 		setclientid.sc_name_len, setclientid.sc_name);
5039 	task = rpc_run_task(&task_setup_data);
5040 	if (IS_ERR(task)) {
5041 		status = PTR_ERR(task);
5042 		goto out;
5043 	}
5044 	status = task->tk_status;
5045 	if (setclientid.sc_cred) {
5046 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5047 		put_rpccred(setclientid.sc_cred);
5048 	}
5049 	rpc_put_task(task);
5050 out:
5051 	trace_nfs4_setclientid(clp, status);
5052 	dprintk("NFS reply setclientid: %d\n", status);
5053 	return status;
5054 }
5055 
5056 /**
5057  * nfs4_proc_setclientid_confirm - Confirm client ID
5058  * @clp: state data structure
5059  * @res: result of a previous SETCLIENTID
5060  * @cred: RPC credential to use for this call
5061  *
5062  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5063  */
5064 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5065 		struct nfs4_setclientid_res *arg,
5066 		struct rpc_cred *cred)
5067 {
5068 	struct rpc_message msg = {
5069 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5070 		.rpc_argp = arg,
5071 		.rpc_cred = cred,
5072 	};
5073 	int status;
5074 
5075 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5076 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5077 		clp->cl_clientid);
5078 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5079 	trace_nfs4_setclientid_confirm(clp, status);
5080 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5081 	return status;
5082 }
5083 
5084 struct nfs4_delegreturndata {
5085 	struct nfs4_delegreturnargs args;
5086 	struct nfs4_delegreturnres res;
5087 	struct nfs_fh fh;
5088 	nfs4_stateid stateid;
5089 	unsigned long timestamp;
5090 	struct nfs_fattr fattr;
5091 	int rpc_status;
5092 	struct inode *inode;
5093 	bool roc;
5094 	u32 roc_barrier;
5095 };
5096 
5097 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5098 {
5099 	struct nfs4_delegreturndata *data = calldata;
5100 
5101 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5102 		return;
5103 
5104 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5105 	switch (task->tk_status) {
5106 	case 0:
5107 		renew_lease(data->res.server, data->timestamp);
5108 	case -NFS4ERR_ADMIN_REVOKED:
5109 	case -NFS4ERR_DELEG_REVOKED:
5110 	case -NFS4ERR_BAD_STATEID:
5111 	case -NFS4ERR_OLD_STATEID:
5112 	case -NFS4ERR_STALE_STATEID:
5113 	case -NFS4ERR_EXPIRED:
5114 		task->tk_status = 0;
5115 		if (data->roc)
5116 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5117 		break;
5118 	default:
5119 		if (nfs4_async_handle_error(task, data->res.server,
5120 					    NULL, NULL) == -EAGAIN) {
5121 			rpc_restart_call_prepare(task);
5122 			return;
5123 		}
5124 	}
5125 	data->rpc_status = task->tk_status;
5126 }
5127 
5128 static void nfs4_delegreturn_release(void *calldata)
5129 {
5130 	struct nfs4_delegreturndata *data = calldata;
5131 
5132 	if (data->roc)
5133 		pnfs_roc_release(data->inode);
5134 	kfree(calldata);
5135 }
5136 
5137 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5138 {
5139 	struct nfs4_delegreturndata *d_data;
5140 
5141 	d_data = (struct nfs4_delegreturndata *)data;
5142 
5143 	if (d_data->roc &&
5144 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5145 		return;
5146 
5147 	nfs4_setup_sequence(d_data->res.server,
5148 			&d_data->args.seq_args,
5149 			&d_data->res.seq_res,
5150 			task);
5151 }
5152 
5153 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5154 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5155 	.rpc_call_done = nfs4_delegreturn_done,
5156 	.rpc_release = nfs4_delegreturn_release,
5157 };
5158 
5159 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5160 {
5161 	struct nfs4_delegreturndata *data;
5162 	struct nfs_server *server = NFS_SERVER(inode);
5163 	struct rpc_task *task;
5164 	struct rpc_message msg = {
5165 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5166 		.rpc_cred = cred,
5167 	};
5168 	struct rpc_task_setup task_setup_data = {
5169 		.rpc_client = server->client,
5170 		.rpc_message = &msg,
5171 		.callback_ops = &nfs4_delegreturn_ops,
5172 		.flags = RPC_TASK_ASYNC,
5173 	};
5174 	int status = 0;
5175 
5176 	data = kzalloc(sizeof(*data), GFP_NOFS);
5177 	if (data == NULL)
5178 		return -ENOMEM;
5179 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5180 	data->args.fhandle = &data->fh;
5181 	data->args.stateid = &data->stateid;
5182 	data->args.bitmask = server->cache_consistency_bitmask;
5183 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5184 	nfs4_stateid_copy(&data->stateid, stateid);
5185 	data->res.fattr = &data->fattr;
5186 	data->res.server = server;
5187 	nfs_fattr_init(data->res.fattr);
5188 	data->timestamp = jiffies;
5189 	data->rpc_status = 0;
5190 	data->inode = inode;
5191 	data->roc = list_empty(&NFS_I(inode)->open_files) ?
5192 		    pnfs_roc(inode) : false;
5193 
5194 	task_setup_data.callback_data = data;
5195 	msg.rpc_argp = &data->args;
5196 	msg.rpc_resp = &data->res;
5197 	task = rpc_run_task(&task_setup_data);
5198 	if (IS_ERR(task))
5199 		return PTR_ERR(task);
5200 	if (!issync)
5201 		goto out;
5202 	status = nfs4_wait_for_completion_rpc_task(task);
5203 	if (status != 0)
5204 		goto out;
5205 	status = data->rpc_status;
5206 	if (status == 0)
5207 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5208 	else
5209 		nfs_refresh_inode(inode, &data->fattr);
5210 out:
5211 	rpc_put_task(task);
5212 	return status;
5213 }
5214 
5215 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5216 {
5217 	struct nfs_server *server = NFS_SERVER(inode);
5218 	struct nfs4_exception exception = { };
5219 	int err;
5220 	do {
5221 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5222 		trace_nfs4_delegreturn(inode, err);
5223 		switch (err) {
5224 			case -NFS4ERR_STALE_STATEID:
5225 			case -NFS4ERR_EXPIRED:
5226 			case 0:
5227 				return 0;
5228 		}
5229 		err = nfs4_handle_exception(server, err, &exception);
5230 	} while (exception.retry);
5231 	return err;
5232 }
5233 
5234 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5235 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5236 
5237 /*
5238  * sleep, with exponential backoff, and retry the LOCK operation.
5239  */
5240 static unsigned long
5241 nfs4_set_lock_task_retry(unsigned long timeout)
5242 {
5243 	freezable_schedule_timeout_killable_unsafe(timeout);
5244 	timeout <<= 1;
5245 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5246 		return NFS4_LOCK_MAXTIMEOUT;
5247 	return timeout;
5248 }
5249 
5250 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5251 {
5252 	struct inode *inode = state->inode;
5253 	struct nfs_server *server = NFS_SERVER(inode);
5254 	struct nfs_client *clp = server->nfs_client;
5255 	struct nfs_lockt_args arg = {
5256 		.fh = NFS_FH(inode),
5257 		.fl = request,
5258 	};
5259 	struct nfs_lockt_res res = {
5260 		.denied = request,
5261 	};
5262 	struct rpc_message msg = {
5263 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5264 		.rpc_argp       = &arg,
5265 		.rpc_resp       = &res,
5266 		.rpc_cred	= state->owner->so_cred,
5267 	};
5268 	struct nfs4_lock_state *lsp;
5269 	int status;
5270 
5271 	arg.lock_owner.clientid = clp->cl_clientid;
5272 	status = nfs4_set_lock_state(state, request);
5273 	if (status != 0)
5274 		goto out;
5275 	lsp = request->fl_u.nfs4_fl.owner;
5276 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5277 	arg.lock_owner.s_dev = server->s_dev;
5278 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5279 	switch (status) {
5280 		case 0:
5281 			request->fl_type = F_UNLCK;
5282 			break;
5283 		case -NFS4ERR_DENIED:
5284 			status = 0;
5285 	}
5286 	request->fl_ops->fl_release_private(request);
5287 	request->fl_ops = NULL;
5288 out:
5289 	return status;
5290 }
5291 
5292 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5293 {
5294 	struct nfs4_exception exception = { };
5295 	int err;
5296 
5297 	do {
5298 		err = _nfs4_proc_getlk(state, cmd, request);
5299 		trace_nfs4_get_lock(request, state, cmd, err);
5300 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5301 				&exception);
5302 	} while (exception.retry);
5303 	return err;
5304 }
5305 
5306 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5307 {
5308 	int res = 0;
5309 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5310 		case FL_POSIX:
5311 			res = posix_lock_file_wait(file, fl);
5312 			break;
5313 		case FL_FLOCK:
5314 			res = flock_lock_file_wait(file, fl);
5315 			break;
5316 		default:
5317 			BUG();
5318 	}
5319 	return res;
5320 }
5321 
5322 struct nfs4_unlockdata {
5323 	struct nfs_locku_args arg;
5324 	struct nfs_locku_res res;
5325 	struct nfs4_lock_state *lsp;
5326 	struct nfs_open_context *ctx;
5327 	struct file_lock fl;
5328 	const struct nfs_server *server;
5329 	unsigned long timestamp;
5330 };
5331 
5332 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5333 		struct nfs_open_context *ctx,
5334 		struct nfs4_lock_state *lsp,
5335 		struct nfs_seqid *seqid)
5336 {
5337 	struct nfs4_unlockdata *p;
5338 	struct inode *inode = lsp->ls_state->inode;
5339 
5340 	p = kzalloc(sizeof(*p), GFP_NOFS);
5341 	if (p == NULL)
5342 		return NULL;
5343 	p->arg.fh = NFS_FH(inode);
5344 	p->arg.fl = &p->fl;
5345 	p->arg.seqid = seqid;
5346 	p->res.seqid = seqid;
5347 	p->arg.stateid = &lsp->ls_stateid;
5348 	p->lsp = lsp;
5349 	atomic_inc(&lsp->ls_count);
5350 	/* Ensure we don't close file until we're done freeing locks! */
5351 	p->ctx = get_nfs_open_context(ctx);
5352 	memcpy(&p->fl, fl, sizeof(p->fl));
5353 	p->server = NFS_SERVER(inode);
5354 	return p;
5355 }
5356 
5357 static void nfs4_locku_release_calldata(void *data)
5358 {
5359 	struct nfs4_unlockdata *calldata = data;
5360 	nfs_free_seqid(calldata->arg.seqid);
5361 	nfs4_put_lock_state(calldata->lsp);
5362 	put_nfs_open_context(calldata->ctx);
5363 	kfree(calldata);
5364 }
5365 
5366 static void nfs4_locku_done(struct rpc_task *task, void *data)
5367 {
5368 	struct nfs4_unlockdata *calldata = data;
5369 
5370 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5371 		return;
5372 	switch (task->tk_status) {
5373 		case 0:
5374 			nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5375 					&calldata->res.stateid);
5376 			renew_lease(calldata->server, calldata->timestamp);
5377 			break;
5378 		case -NFS4ERR_BAD_STATEID:
5379 		case -NFS4ERR_OLD_STATEID:
5380 		case -NFS4ERR_STALE_STATEID:
5381 		case -NFS4ERR_EXPIRED:
5382 			break;
5383 		default:
5384 			if (nfs4_async_handle_error(task, calldata->server,
5385 						    NULL, NULL) == -EAGAIN)
5386 				rpc_restart_call_prepare(task);
5387 	}
5388 	nfs_release_seqid(calldata->arg.seqid);
5389 }
5390 
5391 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5392 {
5393 	struct nfs4_unlockdata *calldata = data;
5394 
5395 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5396 		goto out_wait;
5397 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5398 		/* Note: exit _without_ running nfs4_locku_done */
5399 		goto out_no_action;
5400 	}
5401 	calldata->timestamp = jiffies;
5402 	if (nfs4_setup_sequence(calldata->server,
5403 				&calldata->arg.seq_args,
5404 				&calldata->res.seq_res,
5405 				task) != 0)
5406 		nfs_release_seqid(calldata->arg.seqid);
5407 	return;
5408 out_no_action:
5409 	task->tk_action = NULL;
5410 out_wait:
5411 	nfs4_sequence_done(task, &calldata->res.seq_res);
5412 }
5413 
5414 static const struct rpc_call_ops nfs4_locku_ops = {
5415 	.rpc_call_prepare = nfs4_locku_prepare,
5416 	.rpc_call_done = nfs4_locku_done,
5417 	.rpc_release = nfs4_locku_release_calldata,
5418 };
5419 
5420 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5421 		struct nfs_open_context *ctx,
5422 		struct nfs4_lock_state *lsp,
5423 		struct nfs_seqid *seqid)
5424 {
5425 	struct nfs4_unlockdata *data;
5426 	struct rpc_message msg = {
5427 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5428 		.rpc_cred = ctx->cred,
5429 	};
5430 	struct rpc_task_setup task_setup_data = {
5431 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5432 		.rpc_message = &msg,
5433 		.callback_ops = &nfs4_locku_ops,
5434 		.workqueue = nfsiod_workqueue,
5435 		.flags = RPC_TASK_ASYNC,
5436 	};
5437 
5438 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5439 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5440 
5441 	/* Ensure this is an unlock - when canceling a lock, the
5442 	 * canceled lock is passed in, and it won't be an unlock.
5443 	 */
5444 	fl->fl_type = F_UNLCK;
5445 
5446 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5447 	if (data == NULL) {
5448 		nfs_free_seqid(seqid);
5449 		return ERR_PTR(-ENOMEM);
5450 	}
5451 
5452 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5453 	msg.rpc_argp = &data->arg;
5454 	msg.rpc_resp = &data->res;
5455 	task_setup_data.callback_data = data;
5456 	return rpc_run_task(&task_setup_data);
5457 }
5458 
5459 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5460 {
5461 	struct inode *inode = state->inode;
5462 	struct nfs4_state_owner *sp = state->owner;
5463 	struct nfs_inode *nfsi = NFS_I(inode);
5464 	struct nfs_seqid *seqid;
5465 	struct nfs4_lock_state *lsp;
5466 	struct rpc_task *task;
5467 	int status = 0;
5468 	unsigned char fl_flags = request->fl_flags;
5469 
5470 	status = nfs4_set_lock_state(state, request);
5471 	/* Unlock _before_ we do the RPC call */
5472 	request->fl_flags |= FL_EXISTS;
5473 	/* Exclude nfs_delegation_claim_locks() */
5474 	mutex_lock(&sp->so_delegreturn_mutex);
5475 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5476 	down_read(&nfsi->rwsem);
5477 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5478 		up_read(&nfsi->rwsem);
5479 		mutex_unlock(&sp->so_delegreturn_mutex);
5480 		goto out;
5481 	}
5482 	up_read(&nfsi->rwsem);
5483 	mutex_unlock(&sp->so_delegreturn_mutex);
5484 	if (status != 0)
5485 		goto out;
5486 	/* Is this a delegated lock? */
5487 	lsp = request->fl_u.nfs4_fl.owner;
5488 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5489 		goto out;
5490 	seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5491 	status = -ENOMEM;
5492 	if (seqid == NULL)
5493 		goto out;
5494 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5495 	status = PTR_ERR(task);
5496 	if (IS_ERR(task))
5497 		goto out;
5498 	status = nfs4_wait_for_completion_rpc_task(task);
5499 	rpc_put_task(task);
5500 out:
5501 	request->fl_flags = fl_flags;
5502 	trace_nfs4_unlock(request, state, F_SETLK, status);
5503 	return status;
5504 }
5505 
5506 struct nfs4_lockdata {
5507 	struct nfs_lock_args arg;
5508 	struct nfs_lock_res res;
5509 	struct nfs4_lock_state *lsp;
5510 	struct nfs_open_context *ctx;
5511 	struct file_lock fl;
5512 	unsigned long timestamp;
5513 	int rpc_status;
5514 	int cancelled;
5515 	struct nfs_server *server;
5516 };
5517 
5518 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5519 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5520 		gfp_t gfp_mask)
5521 {
5522 	struct nfs4_lockdata *p;
5523 	struct inode *inode = lsp->ls_state->inode;
5524 	struct nfs_server *server = NFS_SERVER(inode);
5525 
5526 	p = kzalloc(sizeof(*p), gfp_mask);
5527 	if (p == NULL)
5528 		return NULL;
5529 
5530 	p->arg.fh = NFS_FH(inode);
5531 	p->arg.fl = &p->fl;
5532 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5533 	if (p->arg.open_seqid == NULL)
5534 		goto out_free;
5535 	p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5536 	if (p->arg.lock_seqid == NULL)
5537 		goto out_free_seqid;
5538 	p->arg.lock_stateid = &lsp->ls_stateid;
5539 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5540 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5541 	p->arg.lock_owner.s_dev = server->s_dev;
5542 	p->res.lock_seqid = p->arg.lock_seqid;
5543 	p->lsp = lsp;
5544 	p->server = server;
5545 	atomic_inc(&lsp->ls_count);
5546 	p->ctx = get_nfs_open_context(ctx);
5547 	memcpy(&p->fl, fl, sizeof(p->fl));
5548 	return p;
5549 out_free_seqid:
5550 	nfs_free_seqid(p->arg.open_seqid);
5551 out_free:
5552 	kfree(p);
5553 	return NULL;
5554 }
5555 
5556 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5557 {
5558 	struct nfs4_lockdata *data = calldata;
5559 	struct nfs4_state *state = data->lsp->ls_state;
5560 
5561 	dprintk("%s: begin!\n", __func__);
5562 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5563 		goto out_wait;
5564 	/* Do we need to do an open_to_lock_owner? */
5565 	if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5566 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5567 			goto out_release_lock_seqid;
5568 		}
5569 		data->arg.open_stateid = &state->open_stateid;
5570 		data->arg.new_lock_owner = 1;
5571 		data->res.open_seqid = data->arg.open_seqid;
5572 	} else
5573 		data->arg.new_lock_owner = 0;
5574 	if (!nfs4_valid_open_stateid(state)) {
5575 		data->rpc_status = -EBADF;
5576 		task->tk_action = NULL;
5577 		goto out_release_open_seqid;
5578 	}
5579 	data->timestamp = jiffies;
5580 	if (nfs4_setup_sequence(data->server,
5581 				&data->arg.seq_args,
5582 				&data->res.seq_res,
5583 				task) == 0)
5584 		return;
5585 out_release_open_seqid:
5586 	nfs_release_seqid(data->arg.open_seqid);
5587 out_release_lock_seqid:
5588 	nfs_release_seqid(data->arg.lock_seqid);
5589 out_wait:
5590 	nfs4_sequence_done(task, &data->res.seq_res);
5591 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5592 }
5593 
5594 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5595 {
5596 	struct nfs4_lockdata *data = calldata;
5597 
5598 	dprintk("%s: begin!\n", __func__);
5599 
5600 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5601 		return;
5602 
5603 	data->rpc_status = task->tk_status;
5604 	if (data->arg.new_lock_owner != 0) {
5605 		if (data->rpc_status == 0)
5606 			nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5607 		else
5608 			goto out;
5609 	}
5610 	if (data->rpc_status == 0) {
5611 		nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5612 		set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5613 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5614 	}
5615 out:
5616 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5617 }
5618 
5619 static void nfs4_lock_release(void *calldata)
5620 {
5621 	struct nfs4_lockdata *data = calldata;
5622 
5623 	dprintk("%s: begin!\n", __func__);
5624 	nfs_free_seqid(data->arg.open_seqid);
5625 	if (data->cancelled != 0) {
5626 		struct rpc_task *task;
5627 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5628 				data->arg.lock_seqid);
5629 		if (!IS_ERR(task))
5630 			rpc_put_task_async(task);
5631 		dprintk("%s: cancelling lock!\n", __func__);
5632 	} else
5633 		nfs_free_seqid(data->arg.lock_seqid);
5634 	nfs4_put_lock_state(data->lsp);
5635 	put_nfs_open_context(data->ctx);
5636 	kfree(data);
5637 	dprintk("%s: done!\n", __func__);
5638 }
5639 
5640 static const struct rpc_call_ops nfs4_lock_ops = {
5641 	.rpc_call_prepare = nfs4_lock_prepare,
5642 	.rpc_call_done = nfs4_lock_done,
5643 	.rpc_release = nfs4_lock_release,
5644 };
5645 
5646 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5647 {
5648 	switch (error) {
5649 	case -NFS4ERR_ADMIN_REVOKED:
5650 	case -NFS4ERR_BAD_STATEID:
5651 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5652 		if (new_lock_owner != 0 ||
5653 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5654 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5655 		break;
5656 	case -NFS4ERR_STALE_STATEID:
5657 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5658 	case -NFS4ERR_EXPIRED:
5659 		nfs4_schedule_lease_recovery(server->nfs_client);
5660 	};
5661 }
5662 
5663 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5664 {
5665 	struct nfs4_lockdata *data;
5666 	struct rpc_task *task;
5667 	struct rpc_message msg = {
5668 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5669 		.rpc_cred = state->owner->so_cred,
5670 	};
5671 	struct rpc_task_setup task_setup_data = {
5672 		.rpc_client = NFS_CLIENT(state->inode),
5673 		.rpc_message = &msg,
5674 		.callback_ops = &nfs4_lock_ops,
5675 		.workqueue = nfsiod_workqueue,
5676 		.flags = RPC_TASK_ASYNC,
5677 	};
5678 	int ret;
5679 
5680 	dprintk("%s: begin!\n", __func__);
5681 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5682 			fl->fl_u.nfs4_fl.owner,
5683 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5684 	if (data == NULL)
5685 		return -ENOMEM;
5686 	if (IS_SETLKW(cmd))
5687 		data->arg.block = 1;
5688 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5689 	msg.rpc_argp = &data->arg;
5690 	msg.rpc_resp = &data->res;
5691 	task_setup_data.callback_data = data;
5692 	if (recovery_type > NFS_LOCK_NEW) {
5693 		if (recovery_type == NFS_LOCK_RECLAIM)
5694 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5695 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5696 	}
5697 	task = rpc_run_task(&task_setup_data);
5698 	if (IS_ERR(task))
5699 		return PTR_ERR(task);
5700 	ret = nfs4_wait_for_completion_rpc_task(task);
5701 	if (ret == 0) {
5702 		ret = data->rpc_status;
5703 		if (ret)
5704 			nfs4_handle_setlk_error(data->server, data->lsp,
5705 					data->arg.new_lock_owner, ret);
5706 	} else
5707 		data->cancelled = 1;
5708 	rpc_put_task(task);
5709 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5710 	return ret;
5711 }
5712 
5713 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5714 {
5715 	struct nfs_server *server = NFS_SERVER(state->inode);
5716 	struct nfs4_exception exception = {
5717 		.inode = state->inode,
5718 	};
5719 	int err;
5720 
5721 	do {
5722 		/* Cache the lock if possible... */
5723 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5724 			return 0;
5725 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5726 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5727 		if (err != -NFS4ERR_DELAY)
5728 			break;
5729 		nfs4_handle_exception(server, err, &exception);
5730 	} while (exception.retry);
5731 	return err;
5732 }
5733 
5734 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5735 {
5736 	struct nfs_server *server = NFS_SERVER(state->inode);
5737 	struct nfs4_exception exception = {
5738 		.inode = state->inode,
5739 	};
5740 	int err;
5741 
5742 	err = nfs4_set_lock_state(state, request);
5743 	if (err != 0)
5744 		return err;
5745 	if (!recover_lost_locks) {
5746 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5747 		return 0;
5748 	}
5749 	do {
5750 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5751 			return 0;
5752 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5753 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5754 		switch (err) {
5755 		default:
5756 			goto out;
5757 		case -NFS4ERR_GRACE:
5758 		case -NFS4ERR_DELAY:
5759 			nfs4_handle_exception(server, err, &exception);
5760 			err = 0;
5761 		}
5762 	} while (exception.retry);
5763 out:
5764 	return err;
5765 }
5766 
5767 #if defined(CONFIG_NFS_V4_1)
5768 /**
5769  * nfs41_check_expired_locks - possibly free a lock stateid
5770  *
5771  * @state: NFSv4 state for an inode
5772  *
5773  * Returns NFS_OK if recovery for this stateid is now finished.
5774  * Otherwise a negative NFS4ERR value is returned.
5775  */
5776 static int nfs41_check_expired_locks(struct nfs4_state *state)
5777 {
5778 	int status, ret = -NFS4ERR_BAD_STATEID;
5779 	struct nfs4_lock_state *lsp;
5780 	struct nfs_server *server = NFS_SERVER(state->inode);
5781 
5782 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5783 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5784 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5785 
5786 			status = nfs41_test_stateid(server,
5787 					&lsp->ls_stateid,
5788 					cred);
5789 			trace_nfs4_test_lock_stateid(state, lsp, status);
5790 			if (status != NFS_OK) {
5791 				/* Free the stateid unless the server
5792 				 * informs us the stateid is unrecognized. */
5793 				if (status != -NFS4ERR_BAD_STATEID)
5794 					nfs41_free_stateid(server,
5795 							&lsp->ls_stateid,
5796 							cred);
5797 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5798 				ret = status;
5799 			}
5800 		}
5801 	};
5802 
5803 	return ret;
5804 }
5805 
5806 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5807 {
5808 	int status = NFS_OK;
5809 
5810 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5811 		status = nfs41_check_expired_locks(state);
5812 	if (status != NFS_OK)
5813 		status = nfs4_lock_expired(state, request);
5814 	return status;
5815 }
5816 #endif
5817 
5818 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5819 {
5820 	struct nfs4_state_owner *sp = state->owner;
5821 	struct nfs_inode *nfsi = NFS_I(state->inode);
5822 	unsigned char fl_flags = request->fl_flags;
5823 	unsigned int seq;
5824 	int status = -ENOLCK;
5825 
5826 	if ((fl_flags & FL_POSIX) &&
5827 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5828 		goto out;
5829 	/* Is this a delegated open? */
5830 	status = nfs4_set_lock_state(state, request);
5831 	if (status != 0)
5832 		goto out;
5833 	request->fl_flags |= FL_ACCESS;
5834 	status = do_vfs_lock(request->fl_file, request);
5835 	if (status < 0)
5836 		goto out;
5837 	down_read(&nfsi->rwsem);
5838 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5839 		/* Yes: cache locks! */
5840 		/* ...but avoid races with delegation recall... */
5841 		request->fl_flags = fl_flags & ~FL_SLEEP;
5842 		status = do_vfs_lock(request->fl_file, request);
5843 		goto out_unlock;
5844 	}
5845 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5846 	up_read(&nfsi->rwsem);
5847 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5848 	if (status != 0)
5849 		goto out;
5850 	down_read(&nfsi->rwsem);
5851 	if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5852 		status = -NFS4ERR_DELAY;
5853 		goto out_unlock;
5854 	}
5855 	/* Note: we always want to sleep here! */
5856 	request->fl_flags = fl_flags | FL_SLEEP;
5857 	if (do_vfs_lock(request->fl_file, request) < 0)
5858 		printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5859 			"manager!\n", __func__);
5860 out_unlock:
5861 	up_read(&nfsi->rwsem);
5862 out:
5863 	request->fl_flags = fl_flags;
5864 	return status;
5865 }
5866 
5867 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5868 {
5869 	struct nfs4_exception exception = {
5870 		.state = state,
5871 		.inode = state->inode,
5872 	};
5873 	int err;
5874 
5875 	do {
5876 		err = _nfs4_proc_setlk(state, cmd, request);
5877 		trace_nfs4_set_lock(request, state, cmd, err);
5878 		if (err == -NFS4ERR_DENIED)
5879 			err = -EAGAIN;
5880 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5881 				err, &exception);
5882 	} while (exception.retry);
5883 	return err;
5884 }
5885 
5886 static int
5887 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5888 {
5889 	struct nfs_open_context *ctx;
5890 	struct nfs4_state *state;
5891 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5892 	int status;
5893 
5894 	/* verify open state */
5895 	ctx = nfs_file_open_context(filp);
5896 	state = ctx->state;
5897 
5898 	if (request->fl_start < 0 || request->fl_end < 0)
5899 		return -EINVAL;
5900 
5901 	if (IS_GETLK(cmd)) {
5902 		if (state != NULL)
5903 			return nfs4_proc_getlk(state, F_GETLK, request);
5904 		return 0;
5905 	}
5906 
5907 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5908 		return -EINVAL;
5909 
5910 	if (request->fl_type == F_UNLCK) {
5911 		if (state != NULL)
5912 			return nfs4_proc_unlck(state, cmd, request);
5913 		return 0;
5914 	}
5915 
5916 	if (state == NULL)
5917 		return -ENOLCK;
5918 	/*
5919 	 * Don't rely on the VFS having checked the file open mode,
5920 	 * since it won't do this for flock() locks.
5921 	 */
5922 	switch (request->fl_type) {
5923 	case F_RDLCK:
5924 		if (!(filp->f_mode & FMODE_READ))
5925 			return -EBADF;
5926 		break;
5927 	case F_WRLCK:
5928 		if (!(filp->f_mode & FMODE_WRITE))
5929 			return -EBADF;
5930 	}
5931 
5932 	do {
5933 		status = nfs4_proc_setlk(state, cmd, request);
5934 		if ((status != -EAGAIN) || IS_SETLK(cmd))
5935 			break;
5936 		timeout = nfs4_set_lock_task_retry(timeout);
5937 		status = -ERESTARTSYS;
5938 		if (signalled())
5939 			break;
5940 	} while(status < 0);
5941 	return status;
5942 }
5943 
5944 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5945 {
5946 	struct nfs_server *server = NFS_SERVER(state->inode);
5947 	int err;
5948 
5949 	err = nfs4_set_lock_state(state, fl);
5950 	if (err != 0)
5951 		return err;
5952 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5953 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5954 }
5955 
5956 struct nfs_release_lockowner_data {
5957 	struct nfs4_lock_state *lsp;
5958 	struct nfs_server *server;
5959 	struct nfs_release_lockowner_args args;
5960 	struct nfs_release_lockowner_res res;
5961 	unsigned long timestamp;
5962 };
5963 
5964 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5965 {
5966 	struct nfs_release_lockowner_data *data = calldata;
5967 	struct nfs_server *server = data->server;
5968 	nfs40_setup_sequence(server, &data->args.seq_args,
5969 				&data->res.seq_res, task);
5970 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5971 	data->timestamp = jiffies;
5972 }
5973 
5974 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5975 {
5976 	struct nfs_release_lockowner_data *data = calldata;
5977 	struct nfs_server *server = data->server;
5978 
5979 	nfs40_sequence_done(task, &data->res.seq_res);
5980 
5981 	switch (task->tk_status) {
5982 	case 0:
5983 		renew_lease(server, data->timestamp);
5984 		break;
5985 	case -NFS4ERR_STALE_CLIENTID:
5986 	case -NFS4ERR_EXPIRED:
5987 		nfs4_schedule_lease_recovery(server->nfs_client);
5988 		break;
5989 	case -NFS4ERR_LEASE_MOVED:
5990 	case -NFS4ERR_DELAY:
5991 		if (nfs4_async_handle_error(task, server,
5992 					    NULL, NULL) == -EAGAIN)
5993 			rpc_restart_call_prepare(task);
5994 	}
5995 }
5996 
5997 static void nfs4_release_lockowner_release(void *calldata)
5998 {
5999 	struct nfs_release_lockowner_data *data = calldata;
6000 	nfs4_free_lock_state(data->server, data->lsp);
6001 	kfree(calldata);
6002 }
6003 
6004 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6005 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6006 	.rpc_call_done = nfs4_release_lockowner_done,
6007 	.rpc_release = nfs4_release_lockowner_release,
6008 };
6009 
6010 static void
6011 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6012 {
6013 	struct nfs_release_lockowner_data *data;
6014 	struct rpc_message msg = {
6015 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6016 	};
6017 
6018 	if (server->nfs_client->cl_mvops->minor_version != 0)
6019 		return;
6020 
6021 	data = kmalloc(sizeof(*data), GFP_NOFS);
6022 	if (!data)
6023 		return;
6024 	data->lsp = lsp;
6025 	data->server = server;
6026 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6027 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6028 	data->args.lock_owner.s_dev = server->s_dev;
6029 
6030 	msg.rpc_argp = &data->args;
6031 	msg.rpc_resp = &data->res;
6032 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6033 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6034 }
6035 
6036 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6037 
6038 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6039 				   const void *buf, size_t buflen,
6040 				   int flags, int type)
6041 {
6042 	if (strcmp(key, "") != 0)
6043 		return -EINVAL;
6044 
6045 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6046 }
6047 
6048 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6049 				   void *buf, size_t buflen, int type)
6050 {
6051 	if (strcmp(key, "") != 0)
6052 		return -EINVAL;
6053 
6054 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6055 }
6056 
6057 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6058 				       size_t list_len, const char *name,
6059 				       size_t name_len, int type)
6060 {
6061 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6062 
6063 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6064 		return 0;
6065 
6066 	if (list && len <= list_len)
6067 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6068 	return len;
6069 }
6070 
6071 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6072 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6073 {
6074 	return server->caps & NFS_CAP_SECURITY_LABEL;
6075 }
6076 
6077 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6078 				   const void *buf, size_t buflen,
6079 				   int flags, int type)
6080 {
6081 	if (security_ismaclabel(key))
6082 		return nfs4_set_security_label(dentry, buf, buflen);
6083 
6084 	return -EOPNOTSUPP;
6085 }
6086 
6087 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6088 				   void *buf, size_t buflen, int type)
6089 {
6090 	if (security_ismaclabel(key))
6091 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6092 	return -EOPNOTSUPP;
6093 }
6094 
6095 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6096 				       size_t list_len, const char *name,
6097 				       size_t name_len, int type)
6098 {
6099 	size_t len = 0;
6100 
6101 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6102 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6103 		if (list && len <= list_len)
6104 			security_inode_listsecurity(dentry->d_inode, list, len);
6105 	}
6106 	return len;
6107 }
6108 
6109 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6110 	.prefix = XATTR_SECURITY_PREFIX,
6111 	.list	= nfs4_xattr_list_nfs4_label,
6112 	.get	= nfs4_xattr_get_nfs4_label,
6113 	.set	= nfs4_xattr_set_nfs4_label,
6114 };
6115 #endif
6116 
6117 
6118 /*
6119  * nfs_fhget will use either the mounted_on_fileid or the fileid
6120  */
6121 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6122 {
6123 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6124 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6125 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6126 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6127 		return;
6128 
6129 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6130 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6131 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6132 	fattr->nlink = 2;
6133 }
6134 
6135 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6136 				   const struct qstr *name,
6137 				   struct nfs4_fs_locations *fs_locations,
6138 				   struct page *page)
6139 {
6140 	struct nfs_server *server = NFS_SERVER(dir);
6141 	u32 bitmask[3] = {
6142 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6143 	};
6144 	struct nfs4_fs_locations_arg args = {
6145 		.dir_fh = NFS_FH(dir),
6146 		.name = name,
6147 		.page = page,
6148 		.bitmask = bitmask,
6149 	};
6150 	struct nfs4_fs_locations_res res = {
6151 		.fs_locations = fs_locations,
6152 	};
6153 	struct rpc_message msg = {
6154 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6155 		.rpc_argp = &args,
6156 		.rpc_resp = &res,
6157 	};
6158 	int status;
6159 
6160 	dprintk("%s: start\n", __func__);
6161 
6162 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6163 	 * is not supported */
6164 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6165 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6166 	else
6167 		bitmask[0] |= FATTR4_WORD0_FILEID;
6168 
6169 	nfs_fattr_init(&fs_locations->fattr);
6170 	fs_locations->server = server;
6171 	fs_locations->nlocations = 0;
6172 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6173 	dprintk("%s: returned status = %d\n", __func__, status);
6174 	return status;
6175 }
6176 
6177 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6178 			   const struct qstr *name,
6179 			   struct nfs4_fs_locations *fs_locations,
6180 			   struct page *page)
6181 {
6182 	struct nfs4_exception exception = { };
6183 	int err;
6184 	do {
6185 		err = _nfs4_proc_fs_locations(client, dir, name,
6186 				fs_locations, page);
6187 		trace_nfs4_get_fs_locations(dir, name, err);
6188 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6189 				&exception);
6190 	} while (exception.retry);
6191 	return err;
6192 }
6193 
6194 /*
6195  * This operation also signals the server that this client is
6196  * performing migration recovery.  The server can stop returning
6197  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6198  * appended to this compound to identify the client ID which is
6199  * performing recovery.
6200  */
6201 static int _nfs40_proc_get_locations(struct inode *inode,
6202 				     struct nfs4_fs_locations *locations,
6203 				     struct page *page, struct rpc_cred *cred)
6204 {
6205 	struct nfs_server *server = NFS_SERVER(inode);
6206 	struct rpc_clnt *clnt = server->client;
6207 	u32 bitmask[2] = {
6208 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6209 	};
6210 	struct nfs4_fs_locations_arg args = {
6211 		.clientid	= server->nfs_client->cl_clientid,
6212 		.fh		= NFS_FH(inode),
6213 		.page		= page,
6214 		.bitmask	= bitmask,
6215 		.migration	= 1,		/* skip LOOKUP */
6216 		.renew		= 1,		/* append RENEW */
6217 	};
6218 	struct nfs4_fs_locations_res res = {
6219 		.fs_locations	= locations,
6220 		.migration	= 1,
6221 		.renew		= 1,
6222 	};
6223 	struct rpc_message msg = {
6224 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6225 		.rpc_argp	= &args,
6226 		.rpc_resp	= &res,
6227 		.rpc_cred	= cred,
6228 	};
6229 	unsigned long now = jiffies;
6230 	int status;
6231 
6232 	nfs_fattr_init(&locations->fattr);
6233 	locations->server = server;
6234 	locations->nlocations = 0;
6235 
6236 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6237 	nfs4_set_sequence_privileged(&args.seq_args);
6238 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6239 					&args.seq_args, &res.seq_res);
6240 	if (status)
6241 		return status;
6242 
6243 	renew_lease(server, now);
6244 	return 0;
6245 }
6246 
6247 #ifdef CONFIG_NFS_V4_1
6248 
6249 /*
6250  * This operation also signals the server that this client is
6251  * performing migration recovery.  The server can stop asserting
6252  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6253  * performing this operation is identified in the SEQUENCE
6254  * operation in this compound.
6255  *
6256  * When the client supports GETATTR(fs_locations_info), it can
6257  * be plumbed in here.
6258  */
6259 static int _nfs41_proc_get_locations(struct inode *inode,
6260 				     struct nfs4_fs_locations *locations,
6261 				     struct page *page, struct rpc_cred *cred)
6262 {
6263 	struct nfs_server *server = NFS_SERVER(inode);
6264 	struct rpc_clnt *clnt = server->client;
6265 	u32 bitmask[2] = {
6266 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6267 	};
6268 	struct nfs4_fs_locations_arg args = {
6269 		.fh		= NFS_FH(inode),
6270 		.page		= page,
6271 		.bitmask	= bitmask,
6272 		.migration	= 1,		/* skip LOOKUP */
6273 	};
6274 	struct nfs4_fs_locations_res res = {
6275 		.fs_locations	= locations,
6276 		.migration	= 1,
6277 	};
6278 	struct rpc_message msg = {
6279 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6280 		.rpc_argp	= &args,
6281 		.rpc_resp	= &res,
6282 		.rpc_cred	= cred,
6283 	};
6284 	int status;
6285 
6286 	nfs_fattr_init(&locations->fattr);
6287 	locations->server = server;
6288 	locations->nlocations = 0;
6289 
6290 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6291 	nfs4_set_sequence_privileged(&args.seq_args);
6292 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6293 					&args.seq_args, &res.seq_res);
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_get_locations - discover locations for a migrated FSID
6304  * @inode: inode on FSID that is migrating
6305  * @locations: result of query
6306  * @page: buffer
6307  * @cred: credential to use for this operation
6308  *
6309  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6310  * operation failed, or a negative errno if a local error occurred.
6311  *
6312  * On success, "locations" is filled in, but if the server has
6313  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6314  * asserted.
6315  *
6316  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6317  * from this client that require migration recovery.
6318  */
6319 int nfs4_proc_get_locations(struct inode *inode,
6320 			    struct nfs4_fs_locations *locations,
6321 			    struct page *page, struct rpc_cred *cred)
6322 {
6323 	struct nfs_server *server = NFS_SERVER(inode);
6324 	struct nfs_client *clp = server->nfs_client;
6325 	const struct nfs4_mig_recovery_ops *ops =
6326 					clp->cl_mvops->mig_recovery_ops;
6327 	struct nfs4_exception exception = { };
6328 	int status;
6329 
6330 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6331 		(unsigned long long)server->fsid.major,
6332 		(unsigned long long)server->fsid.minor,
6333 		clp->cl_hostname);
6334 	nfs_display_fhandle(NFS_FH(inode), __func__);
6335 
6336 	do {
6337 		status = ops->get_locations(inode, locations, page, cred);
6338 		if (status != -NFS4ERR_DELAY)
6339 			break;
6340 		nfs4_handle_exception(server, status, &exception);
6341 	} while (exception.retry);
6342 	return status;
6343 }
6344 
6345 /*
6346  * This operation also signals the server that this client is
6347  * performing "lease moved" recovery.  The server can stop
6348  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6349  * is appended to this compound to identify the client ID which is
6350  * performing recovery.
6351  */
6352 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6353 {
6354 	struct nfs_server *server = NFS_SERVER(inode);
6355 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6356 	struct rpc_clnt *clnt = server->client;
6357 	struct nfs4_fsid_present_arg args = {
6358 		.fh		= NFS_FH(inode),
6359 		.clientid	= clp->cl_clientid,
6360 		.renew		= 1,		/* append RENEW */
6361 	};
6362 	struct nfs4_fsid_present_res res = {
6363 		.renew		= 1,
6364 	};
6365 	struct rpc_message msg = {
6366 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6367 		.rpc_argp	= &args,
6368 		.rpc_resp	= &res,
6369 		.rpc_cred	= cred,
6370 	};
6371 	unsigned long now = jiffies;
6372 	int status;
6373 
6374 	res.fh = nfs_alloc_fhandle();
6375 	if (res.fh == NULL)
6376 		return -ENOMEM;
6377 
6378 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6379 	nfs4_set_sequence_privileged(&args.seq_args);
6380 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6381 						&args.seq_args, &res.seq_res);
6382 	nfs_free_fhandle(res.fh);
6383 	if (status)
6384 		return status;
6385 
6386 	do_renew_lease(clp, now);
6387 	return 0;
6388 }
6389 
6390 #ifdef CONFIG_NFS_V4_1
6391 
6392 /*
6393  * This operation also signals the server that this client is
6394  * performing "lease moved" recovery.  The server can stop asserting
6395  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6396  * this operation is identified in the SEQUENCE operation in this
6397  * compound.
6398  */
6399 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6400 {
6401 	struct nfs_server *server = NFS_SERVER(inode);
6402 	struct rpc_clnt *clnt = server->client;
6403 	struct nfs4_fsid_present_arg args = {
6404 		.fh		= NFS_FH(inode),
6405 	};
6406 	struct nfs4_fsid_present_res res = {
6407 	};
6408 	struct rpc_message msg = {
6409 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6410 		.rpc_argp	= &args,
6411 		.rpc_resp	= &res,
6412 		.rpc_cred	= cred,
6413 	};
6414 	int status;
6415 
6416 	res.fh = nfs_alloc_fhandle();
6417 	if (res.fh == NULL)
6418 		return -ENOMEM;
6419 
6420 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6421 	nfs4_set_sequence_privileged(&args.seq_args);
6422 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6423 						&args.seq_args, &res.seq_res);
6424 	nfs_free_fhandle(res.fh);
6425 	if (status == NFS4_OK &&
6426 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6427 		status = -NFS4ERR_LEASE_MOVED;
6428 	return status;
6429 }
6430 
6431 #endif	/* CONFIG_NFS_V4_1 */
6432 
6433 /**
6434  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6435  * @inode: inode on FSID to check
6436  * @cred: credential to use for this operation
6437  *
6438  * Server indicates whether the FSID is present, moved, or not
6439  * recognized.  This operation is necessary to clear a LEASE_MOVED
6440  * condition for this client ID.
6441  *
6442  * Returns NFS4_OK if the FSID is present on this server,
6443  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6444  *  NFS4ERR code if some error occurred on the server, or a
6445  *  negative errno if a local failure occurred.
6446  */
6447 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6448 {
6449 	struct nfs_server *server = NFS_SERVER(inode);
6450 	struct nfs_client *clp = server->nfs_client;
6451 	const struct nfs4_mig_recovery_ops *ops =
6452 					clp->cl_mvops->mig_recovery_ops;
6453 	struct nfs4_exception exception = { };
6454 	int status;
6455 
6456 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6457 		(unsigned long long)server->fsid.major,
6458 		(unsigned long long)server->fsid.minor,
6459 		clp->cl_hostname);
6460 	nfs_display_fhandle(NFS_FH(inode), __func__);
6461 
6462 	do {
6463 		status = ops->fsid_present(inode, cred);
6464 		if (status != -NFS4ERR_DELAY)
6465 			break;
6466 		nfs4_handle_exception(server, status, &exception);
6467 	} while (exception.retry);
6468 	return status;
6469 }
6470 
6471 /**
6472  * If 'use_integrity' is true and the state managment nfs_client
6473  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6474  * and the machine credential as per RFC3530bis and RFC5661 Security
6475  * Considerations sections. Otherwise, just use the user cred with the
6476  * filesystem's rpc_client.
6477  */
6478 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6479 {
6480 	int status;
6481 	struct nfs4_secinfo_arg args = {
6482 		.dir_fh = NFS_FH(dir),
6483 		.name   = name,
6484 	};
6485 	struct nfs4_secinfo_res res = {
6486 		.flavors     = flavors,
6487 	};
6488 	struct rpc_message msg = {
6489 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6490 		.rpc_argp = &args,
6491 		.rpc_resp = &res,
6492 	};
6493 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6494 	struct rpc_cred *cred = NULL;
6495 
6496 	if (use_integrity) {
6497 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6498 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6499 		msg.rpc_cred = cred;
6500 	}
6501 
6502 	dprintk("NFS call  secinfo %s\n", name->name);
6503 
6504 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6505 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6506 
6507 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6508 				&res.seq_res, 0);
6509 	dprintk("NFS reply  secinfo: %d\n", status);
6510 
6511 	if (cred)
6512 		put_rpccred(cred);
6513 
6514 	return status;
6515 }
6516 
6517 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6518 		      struct nfs4_secinfo_flavors *flavors)
6519 {
6520 	struct nfs4_exception exception = { };
6521 	int err;
6522 	do {
6523 		err = -NFS4ERR_WRONGSEC;
6524 
6525 		/* try to use integrity protection with machine cred */
6526 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6527 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6528 
6529 		/*
6530 		 * if unable to use integrity protection, or SECINFO with
6531 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6532 		 * disallowed by spec, but exists in deployed servers) use
6533 		 * the current filesystem's rpc_client and the user cred.
6534 		 */
6535 		if (err == -NFS4ERR_WRONGSEC)
6536 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6537 
6538 		trace_nfs4_secinfo(dir, name, err);
6539 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6540 				&exception);
6541 	} while (exception.retry);
6542 	return err;
6543 }
6544 
6545 #ifdef CONFIG_NFS_V4_1
6546 /*
6547  * Check the exchange flags returned by the server for invalid flags, having
6548  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6549  * DS flags set.
6550  */
6551 static int nfs4_check_cl_exchange_flags(u32 flags)
6552 {
6553 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6554 		goto out_inval;
6555 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6556 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6557 		goto out_inval;
6558 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6559 		goto out_inval;
6560 	return NFS_OK;
6561 out_inval:
6562 	return -NFS4ERR_INVAL;
6563 }
6564 
6565 static bool
6566 nfs41_same_server_scope(struct nfs41_server_scope *a,
6567 			struct nfs41_server_scope *b)
6568 {
6569 	if (a->server_scope_sz == b->server_scope_sz &&
6570 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6571 		return true;
6572 
6573 	return false;
6574 }
6575 
6576 /*
6577  * nfs4_proc_bind_conn_to_session()
6578  *
6579  * The 4.1 client currently uses the same TCP connection for the
6580  * fore and backchannel.
6581  */
6582 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6583 {
6584 	int status;
6585 	struct nfs41_bind_conn_to_session_res res;
6586 	struct rpc_message msg = {
6587 		.rpc_proc =
6588 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6589 		.rpc_argp = clp,
6590 		.rpc_resp = &res,
6591 		.rpc_cred = cred,
6592 	};
6593 
6594 	dprintk("--> %s\n", __func__);
6595 
6596 	res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6597 	if (unlikely(res.session == NULL)) {
6598 		status = -ENOMEM;
6599 		goto out;
6600 	}
6601 
6602 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6603 	trace_nfs4_bind_conn_to_session(clp, status);
6604 	if (status == 0) {
6605 		if (memcmp(res.session->sess_id.data,
6606 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6607 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6608 			status = -EIO;
6609 			goto out_session;
6610 		}
6611 		if (res.dir != NFS4_CDFS4_BOTH) {
6612 			dprintk("NFS: %s: Unexpected direction from server\n",
6613 				__func__);
6614 			status = -EIO;
6615 			goto out_session;
6616 		}
6617 		if (res.use_conn_in_rdma_mode) {
6618 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6619 				__func__);
6620 			status = -EIO;
6621 			goto out_session;
6622 		}
6623 	}
6624 out_session:
6625 	kfree(res.session);
6626 out:
6627 	dprintk("<-- %s status= %d\n", __func__, status);
6628 	return status;
6629 }
6630 
6631 /*
6632  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6633  * and operations we'd like to see to enable certain features in the allow map
6634  */
6635 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6636 	.how = SP4_MACH_CRED,
6637 	.enforce.u.words = {
6638 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6639 		      1 << (OP_EXCHANGE_ID - 32) |
6640 		      1 << (OP_CREATE_SESSION - 32) |
6641 		      1 << (OP_DESTROY_SESSION - 32) |
6642 		      1 << (OP_DESTROY_CLIENTID - 32)
6643 	},
6644 	.allow.u.words = {
6645 		[0] = 1 << (OP_CLOSE) |
6646 		      1 << (OP_LOCKU) |
6647 		      1 << (OP_COMMIT),
6648 		[1] = 1 << (OP_SECINFO - 32) |
6649 		      1 << (OP_SECINFO_NO_NAME - 32) |
6650 		      1 << (OP_TEST_STATEID - 32) |
6651 		      1 << (OP_FREE_STATEID - 32) |
6652 		      1 << (OP_WRITE - 32)
6653 	}
6654 };
6655 
6656 /*
6657  * Select the state protection mode for client `clp' given the server results
6658  * from exchange_id in `sp'.
6659  *
6660  * Returns 0 on success, negative errno otherwise.
6661  */
6662 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6663 				 struct nfs41_state_protection *sp)
6664 {
6665 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6666 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6667 		      1 << (OP_EXCHANGE_ID - 32) |
6668 		      1 << (OP_CREATE_SESSION - 32) |
6669 		      1 << (OP_DESTROY_SESSION - 32) |
6670 		      1 << (OP_DESTROY_CLIENTID - 32)
6671 	};
6672 	unsigned int i;
6673 
6674 	if (sp->how == SP4_MACH_CRED) {
6675 		/* Print state protect result */
6676 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6677 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6678 			if (test_bit(i, sp->enforce.u.longs))
6679 				dfprintk(MOUNT, "  enforce op %d\n", i);
6680 			if (test_bit(i, sp->allow.u.longs))
6681 				dfprintk(MOUNT, "  allow op %d\n", i);
6682 		}
6683 
6684 		/* make sure nothing is on enforce list that isn't supported */
6685 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6686 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6687 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6688 				return -EINVAL;
6689 			}
6690 		}
6691 
6692 		/*
6693 		 * Minimal mode - state operations are allowed to use machine
6694 		 * credential.  Note this already happens by default, so the
6695 		 * client doesn't have to do anything more than the negotiation.
6696 		 *
6697 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6698 		 *       we're already using the machine cred for exchange_id
6699 		 *       and will never use a different cred.
6700 		 */
6701 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6702 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6703 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6704 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6705 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6706 			dfprintk(MOUNT, "  minimal mode enabled\n");
6707 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6708 		} else {
6709 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6710 			return -EINVAL;
6711 		}
6712 
6713 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6714 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6715 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6716 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6717 		}
6718 
6719 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6720 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6721 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6722 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6723 		}
6724 
6725 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6726 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6727 			dfprintk(MOUNT, "  stateid mode enabled\n");
6728 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6729 		}
6730 
6731 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6732 			dfprintk(MOUNT, "  write mode enabled\n");
6733 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6734 		}
6735 
6736 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6737 			dfprintk(MOUNT, "  commit mode enabled\n");
6738 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6739 		}
6740 	}
6741 
6742 	return 0;
6743 }
6744 
6745 /*
6746  * _nfs4_proc_exchange_id()
6747  *
6748  * Wrapper for EXCHANGE_ID operation.
6749  */
6750 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6751 	u32 sp4_how)
6752 {
6753 	nfs4_verifier verifier;
6754 	struct nfs41_exchange_id_args args = {
6755 		.verifier = &verifier,
6756 		.client = clp,
6757 #ifdef CONFIG_NFS_V4_1_MIGRATION
6758 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6759 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6760 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6761 #else
6762 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6763 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6764 #endif
6765 	};
6766 	struct nfs41_exchange_id_res res = {
6767 		0
6768 	};
6769 	int status;
6770 	struct rpc_message msg = {
6771 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6772 		.rpc_argp = &args,
6773 		.rpc_resp = &res,
6774 		.rpc_cred = cred,
6775 	};
6776 
6777 	nfs4_init_boot_verifier(clp, &verifier);
6778 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6779 							sizeof(args.id));
6780 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6781 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6782 		args.id_len, args.id);
6783 
6784 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6785 					GFP_NOFS);
6786 	if (unlikely(res.server_owner == NULL)) {
6787 		status = -ENOMEM;
6788 		goto out;
6789 	}
6790 
6791 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6792 					GFP_NOFS);
6793 	if (unlikely(res.server_scope == NULL)) {
6794 		status = -ENOMEM;
6795 		goto out_server_owner;
6796 	}
6797 
6798 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6799 	if (unlikely(res.impl_id == NULL)) {
6800 		status = -ENOMEM;
6801 		goto out_server_scope;
6802 	}
6803 
6804 	switch (sp4_how) {
6805 	case SP4_NONE:
6806 		args.state_protect.how = SP4_NONE;
6807 		break;
6808 
6809 	case SP4_MACH_CRED:
6810 		args.state_protect = nfs4_sp4_mach_cred_request;
6811 		break;
6812 
6813 	default:
6814 		/* unsupported! */
6815 		WARN_ON_ONCE(1);
6816 		status = -EINVAL;
6817 		goto out_server_scope;
6818 	}
6819 
6820 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6821 	trace_nfs4_exchange_id(clp, status);
6822 	if (status == 0)
6823 		status = nfs4_check_cl_exchange_flags(res.flags);
6824 
6825 	if (status == 0)
6826 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6827 
6828 	if (status == 0) {
6829 		clp->cl_clientid = res.clientid;
6830 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6831 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6832 			clp->cl_seqid = res.seqid;
6833 
6834 		kfree(clp->cl_serverowner);
6835 		clp->cl_serverowner = res.server_owner;
6836 		res.server_owner = NULL;
6837 
6838 		/* use the most recent implementation id */
6839 		kfree(clp->cl_implid);
6840 		clp->cl_implid = res.impl_id;
6841 
6842 		if (clp->cl_serverscope != NULL &&
6843 		    !nfs41_same_server_scope(clp->cl_serverscope,
6844 					     res.server_scope)) {
6845 			dprintk("%s: server_scope mismatch detected\n",
6846 				__func__);
6847 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6848 			kfree(clp->cl_serverscope);
6849 			clp->cl_serverscope = NULL;
6850 		}
6851 
6852 		if (clp->cl_serverscope == NULL) {
6853 			clp->cl_serverscope = res.server_scope;
6854 			goto out;
6855 		}
6856 	} else
6857 		kfree(res.impl_id);
6858 
6859 out_server_owner:
6860 	kfree(res.server_owner);
6861 out_server_scope:
6862 	kfree(res.server_scope);
6863 out:
6864 	if (clp->cl_implid != NULL)
6865 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6866 			"domain: %s, name: %s, date: %llu,%u\n",
6867 			clp->cl_implid->domain, clp->cl_implid->name,
6868 			clp->cl_implid->date.seconds,
6869 			clp->cl_implid->date.nseconds);
6870 	dprintk("NFS reply exchange_id: %d\n", status);
6871 	return status;
6872 }
6873 
6874 /*
6875  * nfs4_proc_exchange_id()
6876  *
6877  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6878  *
6879  * Since the clientid has expired, all compounds using sessions
6880  * associated with the stale clientid will be returning
6881  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6882  * be in some phase of session reset.
6883  *
6884  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6885  */
6886 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6887 {
6888 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6889 	int status;
6890 
6891 	/* try SP4_MACH_CRED if krb5i/p	*/
6892 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6893 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6894 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6895 		if (!status)
6896 			return 0;
6897 	}
6898 
6899 	/* try SP4_NONE */
6900 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6901 }
6902 
6903 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6904 		struct rpc_cred *cred)
6905 {
6906 	struct rpc_message msg = {
6907 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6908 		.rpc_argp = clp,
6909 		.rpc_cred = cred,
6910 	};
6911 	int status;
6912 
6913 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6914 	trace_nfs4_destroy_clientid(clp, status);
6915 	if (status)
6916 		dprintk("NFS: Got error %d from the server %s on "
6917 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6918 	return status;
6919 }
6920 
6921 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6922 		struct rpc_cred *cred)
6923 {
6924 	unsigned int loop;
6925 	int ret;
6926 
6927 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6928 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6929 		switch (ret) {
6930 		case -NFS4ERR_DELAY:
6931 		case -NFS4ERR_CLIENTID_BUSY:
6932 			ssleep(1);
6933 			break;
6934 		default:
6935 			return ret;
6936 		}
6937 	}
6938 	return 0;
6939 }
6940 
6941 int nfs4_destroy_clientid(struct nfs_client *clp)
6942 {
6943 	struct rpc_cred *cred;
6944 	int ret = 0;
6945 
6946 	if (clp->cl_mvops->minor_version < 1)
6947 		goto out;
6948 	if (clp->cl_exchange_flags == 0)
6949 		goto out;
6950 	if (clp->cl_preserve_clid)
6951 		goto out;
6952 	cred = nfs4_get_clid_cred(clp);
6953 	ret = nfs4_proc_destroy_clientid(clp, cred);
6954 	if (cred)
6955 		put_rpccred(cred);
6956 	switch (ret) {
6957 	case 0:
6958 	case -NFS4ERR_STALE_CLIENTID:
6959 		clp->cl_exchange_flags = 0;
6960 	}
6961 out:
6962 	return ret;
6963 }
6964 
6965 struct nfs4_get_lease_time_data {
6966 	struct nfs4_get_lease_time_args *args;
6967 	struct nfs4_get_lease_time_res *res;
6968 	struct nfs_client *clp;
6969 };
6970 
6971 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6972 					void *calldata)
6973 {
6974 	struct nfs4_get_lease_time_data *data =
6975 			(struct nfs4_get_lease_time_data *)calldata;
6976 
6977 	dprintk("--> %s\n", __func__);
6978 	/* just setup sequence, do not trigger session recovery
6979 	   since we're invoked within one */
6980 	nfs41_setup_sequence(data->clp->cl_session,
6981 			&data->args->la_seq_args,
6982 			&data->res->lr_seq_res,
6983 			task);
6984 	dprintk("<-- %s\n", __func__);
6985 }
6986 
6987 /*
6988  * Called from nfs4_state_manager thread for session setup, so don't recover
6989  * from sequence operation or clientid errors.
6990  */
6991 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6992 {
6993 	struct nfs4_get_lease_time_data *data =
6994 			(struct nfs4_get_lease_time_data *)calldata;
6995 
6996 	dprintk("--> %s\n", __func__);
6997 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6998 		return;
6999 	switch (task->tk_status) {
7000 	case -NFS4ERR_DELAY:
7001 	case -NFS4ERR_GRACE:
7002 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7003 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7004 		task->tk_status = 0;
7005 		/* fall through */
7006 	case -NFS4ERR_RETRY_UNCACHED_REP:
7007 		rpc_restart_call_prepare(task);
7008 		return;
7009 	}
7010 	dprintk("<-- %s\n", __func__);
7011 }
7012 
7013 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7014 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7015 	.rpc_call_done = nfs4_get_lease_time_done,
7016 };
7017 
7018 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7019 {
7020 	struct rpc_task *task;
7021 	struct nfs4_get_lease_time_args args;
7022 	struct nfs4_get_lease_time_res res = {
7023 		.lr_fsinfo = fsinfo,
7024 	};
7025 	struct nfs4_get_lease_time_data data = {
7026 		.args = &args,
7027 		.res = &res,
7028 		.clp = clp,
7029 	};
7030 	struct rpc_message msg = {
7031 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7032 		.rpc_argp = &args,
7033 		.rpc_resp = &res,
7034 	};
7035 	struct rpc_task_setup task_setup = {
7036 		.rpc_client = clp->cl_rpcclient,
7037 		.rpc_message = &msg,
7038 		.callback_ops = &nfs4_get_lease_time_ops,
7039 		.callback_data = &data,
7040 		.flags = RPC_TASK_TIMEOUT,
7041 	};
7042 	int status;
7043 
7044 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7045 	nfs4_set_sequence_privileged(&args.la_seq_args);
7046 	dprintk("--> %s\n", __func__);
7047 	task = rpc_run_task(&task_setup);
7048 
7049 	if (IS_ERR(task))
7050 		status = PTR_ERR(task);
7051 	else {
7052 		status = task->tk_status;
7053 		rpc_put_task(task);
7054 	}
7055 	dprintk("<-- %s return %d\n", __func__, status);
7056 
7057 	return status;
7058 }
7059 
7060 /*
7061  * Initialize the values to be used by the client in CREATE_SESSION
7062  * If nfs4_init_session set the fore channel request and response sizes,
7063  * use them.
7064  *
7065  * Set the back channel max_resp_sz_cached to zero to force the client to
7066  * always set csa_cachethis to FALSE because the current implementation
7067  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7068  */
7069 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7070 {
7071 	unsigned int max_rqst_sz, max_resp_sz;
7072 
7073 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7074 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7075 
7076 	/* Fore channel attributes */
7077 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7078 	args->fc_attrs.max_resp_sz = max_resp_sz;
7079 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7080 	args->fc_attrs.max_reqs = max_session_slots;
7081 
7082 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7083 		"max_ops=%u max_reqs=%u\n",
7084 		__func__,
7085 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7086 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7087 
7088 	/* Back channel attributes */
7089 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7090 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7091 	args->bc_attrs.max_resp_sz_cached = 0;
7092 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7093 	args->bc_attrs.max_reqs = 1;
7094 
7095 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7096 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7097 		__func__,
7098 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7099 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7100 		args->bc_attrs.max_reqs);
7101 }
7102 
7103 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7104 {
7105 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7106 	struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7107 
7108 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7109 		return -EINVAL;
7110 	/*
7111 	 * Our requested max_ops is the minimum we need; we're not
7112 	 * prepared to break up compounds into smaller pieces than that.
7113 	 * So, no point even trying to continue if the server won't
7114 	 * cooperate:
7115 	 */
7116 	if (rcvd->max_ops < sent->max_ops)
7117 		return -EINVAL;
7118 	if (rcvd->max_reqs == 0)
7119 		return -EINVAL;
7120 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7121 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7122 	return 0;
7123 }
7124 
7125 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7126 {
7127 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7128 	struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7129 
7130 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7131 		return -EINVAL;
7132 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7133 		return -EINVAL;
7134 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7135 		return -EINVAL;
7136 	/* These would render the backchannel useless: */
7137 	if (rcvd->max_ops != sent->max_ops)
7138 		return -EINVAL;
7139 	if (rcvd->max_reqs != sent->max_reqs)
7140 		return -EINVAL;
7141 	return 0;
7142 }
7143 
7144 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7145 				     struct nfs4_session *session)
7146 {
7147 	int ret;
7148 
7149 	ret = nfs4_verify_fore_channel_attrs(args, session);
7150 	if (ret)
7151 		return ret;
7152 	return nfs4_verify_back_channel_attrs(args, session);
7153 }
7154 
7155 static int _nfs4_proc_create_session(struct nfs_client *clp,
7156 		struct rpc_cred *cred)
7157 {
7158 	struct nfs4_session *session = clp->cl_session;
7159 	struct nfs41_create_session_args args = {
7160 		.client = clp,
7161 		.cb_program = NFS4_CALLBACK,
7162 	};
7163 	struct nfs41_create_session_res res = {
7164 		.client = clp,
7165 	};
7166 	struct rpc_message msg = {
7167 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7168 		.rpc_argp = &args,
7169 		.rpc_resp = &res,
7170 		.rpc_cred = cred,
7171 	};
7172 	int status;
7173 
7174 	nfs4_init_channel_attrs(&args);
7175 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7176 
7177 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7178 	trace_nfs4_create_session(clp, status);
7179 
7180 	if (!status) {
7181 		/* Verify the session's negotiated channel_attrs values */
7182 		status = nfs4_verify_channel_attrs(&args, session);
7183 		/* Increment the clientid slot sequence id */
7184 		clp->cl_seqid++;
7185 	}
7186 
7187 	return status;
7188 }
7189 
7190 /*
7191  * Issues a CREATE_SESSION operation to the server.
7192  * It is the responsibility of the caller to verify the session is
7193  * expired before calling this routine.
7194  */
7195 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7196 {
7197 	int status;
7198 	unsigned *ptr;
7199 	struct nfs4_session *session = clp->cl_session;
7200 
7201 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7202 
7203 	status = _nfs4_proc_create_session(clp, cred);
7204 	if (status)
7205 		goto out;
7206 
7207 	/* Init or reset the session slot tables */
7208 	status = nfs4_setup_session_slot_tables(session);
7209 	dprintk("slot table setup returned %d\n", status);
7210 	if (status)
7211 		goto out;
7212 
7213 	ptr = (unsigned *)&session->sess_id.data[0];
7214 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7215 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7216 out:
7217 	dprintk("<-- %s\n", __func__);
7218 	return status;
7219 }
7220 
7221 /*
7222  * Issue the over-the-wire RPC DESTROY_SESSION.
7223  * The caller must serialize access to this routine.
7224  */
7225 int nfs4_proc_destroy_session(struct nfs4_session *session,
7226 		struct rpc_cred *cred)
7227 {
7228 	struct rpc_message msg = {
7229 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7230 		.rpc_argp = session,
7231 		.rpc_cred = cred,
7232 	};
7233 	int status = 0;
7234 
7235 	dprintk("--> nfs4_proc_destroy_session\n");
7236 
7237 	/* session is still being setup */
7238 	if (session->clp->cl_cons_state != NFS_CS_READY)
7239 		return status;
7240 
7241 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7242 	trace_nfs4_destroy_session(session->clp, status);
7243 
7244 	if (status)
7245 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7246 			"Session has been destroyed regardless...\n", status);
7247 
7248 	dprintk("<-- nfs4_proc_destroy_session\n");
7249 	return status;
7250 }
7251 
7252 /*
7253  * Renew the cl_session lease.
7254  */
7255 struct nfs4_sequence_data {
7256 	struct nfs_client *clp;
7257 	struct nfs4_sequence_args args;
7258 	struct nfs4_sequence_res res;
7259 };
7260 
7261 static void nfs41_sequence_release(void *data)
7262 {
7263 	struct nfs4_sequence_data *calldata = data;
7264 	struct nfs_client *clp = calldata->clp;
7265 
7266 	if (atomic_read(&clp->cl_count) > 1)
7267 		nfs4_schedule_state_renewal(clp);
7268 	nfs_put_client(clp);
7269 	kfree(calldata);
7270 }
7271 
7272 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7273 {
7274 	switch(task->tk_status) {
7275 	case -NFS4ERR_DELAY:
7276 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7277 		return -EAGAIN;
7278 	default:
7279 		nfs4_schedule_lease_recovery(clp);
7280 	}
7281 	return 0;
7282 }
7283 
7284 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7285 {
7286 	struct nfs4_sequence_data *calldata = data;
7287 	struct nfs_client *clp = calldata->clp;
7288 
7289 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7290 		return;
7291 
7292 	trace_nfs4_sequence(clp, task->tk_status);
7293 	if (task->tk_status < 0) {
7294 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7295 		if (atomic_read(&clp->cl_count) == 1)
7296 			goto out;
7297 
7298 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7299 			rpc_restart_call_prepare(task);
7300 			return;
7301 		}
7302 	}
7303 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7304 out:
7305 	dprintk("<-- %s\n", __func__);
7306 }
7307 
7308 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7309 {
7310 	struct nfs4_sequence_data *calldata = data;
7311 	struct nfs_client *clp = calldata->clp;
7312 	struct nfs4_sequence_args *args;
7313 	struct nfs4_sequence_res *res;
7314 
7315 	args = task->tk_msg.rpc_argp;
7316 	res = task->tk_msg.rpc_resp;
7317 
7318 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7319 }
7320 
7321 static const struct rpc_call_ops nfs41_sequence_ops = {
7322 	.rpc_call_done = nfs41_sequence_call_done,
7323 	.rpc_call_prepare = nfs41_sequence_prepare,
7324 	.rpc_release = nfs41_sequence_release,
7325 };
7326 
7327 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7328 		struct rpc_cred *cred,
7329 		bool is_privileged)
7330 {
7331 	struct nfs4_sequence_data *calldata;
7332 	struct rpc_message msg = {
7333 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7334 		.rpc_cred = cred,
7335 	};
7336 	struct rpc_task_setup task_setup_data = {
7337 		.rpc_client = clp->cl_rpcclient,
7338 		.rpc_message = &msg,
7339 		.callback_ops = &nfs41_sequence_ops,
7340 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7341 	};
7342 
7343 	if (!atomic_inc_not_zero(&clp->cl_count))
7344 		return ERR_PTR(-EIO);
7345 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7346 	if (calldata == NULL) {
7347 		nfs_put_client(clp);
7348 		return ERR_PTR(-ENOMEM);
7349 	}
7350 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7351 	if (is_privileged)
7352 		nfs4_set_sequence_privileged(&calldata->args);
7353 	msg.rpc_argp = &calldata->args;
7354 	msg.rpc_resp = &calldata->res;
7355 	calldata->clp = clp;
7356 	task_setup_data.callback_data = calldata;
7357 
7358 	return rpc_run_task(&task_setup_data);
7359 }
7360 
7361 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7362 {
7363 	struct rpc_task *task;
7364 	int ret = 0;
7365 
7366 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7367 		return -EAGAIN;
7368 	task = _nfs41_proc_sequence(clp, cred, false);
7369 	if (IS_ERR(task))
7370 		ret = PTR_ERR(task);
7371 	else
7372 		rpc_put_task_async(task);
7373 	dprintk("<-- %s status=%d\n", __func__, ret);
7374 	return ret;
7375 }
7376 
7377 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7378 {
7379 	struct rpc_task *task;
7380 	int ret;
7381 
7382 	task = _nfs41_proc_sequence(clp, cred, true);
7383 	if (IS_ERR(task)) {
7384 		ret = PTR_ERR(task);
7385 		goto out;
7386 	}
7387 	ret = rpc_wait_for_completion_task(task);
7388 	if (!ret) {
7389 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7390 
7391 		if (task->tk_status == 0)
7392 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7393 		ret = task->tk_status;
7394 	}
7395 	rpc_put_task(task);
7396 out:
7397 	dprintk("<-- %s status=%d\n", __func__, ret);
7398 	return ret;
7399 }
7400 
7401 struct nfs4_reclaim_complete_data {
7402 	struct nfs_client *clp;
7403 	struct nfs41_reclaim_complete_args arg;
7404 	struct nfs41_reclaim_complete_res res;
7405 };
7406 
7407 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7408 {
7409 	struct nfs4_reclaim_complete_data *calldata = data;
7410 
7411 	nfs41_setup_sequence(calldata->clp->cl_session,
7412 			&calldata->arg.seq_args,
7413 			&calldata->res.seq_res,
7414 			task);
7415 }
7416 
7417 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7418 {
7419 	switch(task->tk_status) {
7420 	case 0:
7421 	case -NFS4ERR_COMPLETE_ALREADY:
7422 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7423 		break;
7424 	case -NFS4ERR_DELAY:
7425 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7426 		/* fall through */
7427 	case -NFS4ERR_RETRY_UNCACHED_REP:
7428 		return -EAGAIN;
7429 	default:
7430 		nfs4_schedule_lease_recovery(clp);
7431 	}
7432 	return 0;
7433 }
7434 
7435 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7436 {
7437 	struct nfs4_reclaim_complete_data *calldata = data;
7438 	struct nfs_client *clp = calldata->clp;
7439 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7440 
7441 	dprintk("--> %s\n", __func__);
7442 	if (!nfs41_sequence_done(task, res))
7443 		return;
7444 
7445 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7446 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7447 		rpc_restart_call_prepare(task);
7448 		return;
7449 	}
7450 	dprintk("<-- %s\n", __func__);
7451 }
7452 
7453 static void nfs4_free_reclaim_complete_data(void *data)
7454 {
7455 	struct nfs4_reclaim_complete_data *calldata = data;
7456 
7457 	kfree(calldata);
7458 }
7459 
7460 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7461 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7462 	.rpc_call_done = nfs4_reclaim_complete_done,
7463 	.rpc_release = nfs4_free_reclaim_complete_data,
7464 };
7465 
7466 /*
7467  * Issue a global reclaim complete.
7468  */
7469 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7470 		struct rpc_cred *cred)
7471 {
7472 	struct nfs4_reclaim_complete_data *calldata;
7473 	struct rpc_task *task;
7474 	struct rpc_message msg = {
7475 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7476 		.rpc_cred = cred,
7477 	};
7478 	struct rpc_task_setup task_setup_data = {
7479 		.rpc_client = clp->cl_rpcclient,
7480 		.rpc_message = &msg,
7481 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7482 		.flags = RPC_TASK_ASYNC,
7483 	};
7484 	int status = -ENOMEM;
7485 
7486 	dprintk("--> %s\n", __func__);
7487 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7488 	if (calldata == NULL)
7489 		goto out;
7490 	calldata->clp = clp;
7491 	calldata->arg.one_fs = 0;
7492 
7493 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7494 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7495 	msg.rpc_argp = &calldata->arg;
7496 	msg.rpc_resp = &calldata->res;
7497 	task_setup_data.callback_data = calldata;
7498 	task = rpc_run_task(&task_setup_data);
7499 	if (IS_ERR(task)) {
7500 		status = PTR_ERR(task);
7501 		goto out;
7502 	}
7503 	status = nfs4_wait_for_completion_rpc_task(task);
7504 	if (status == 0)
7505 		status = task->tk_status;
7506 	rpc_put_task(task);
7507 	return 0;
7508 out:
7509 	dprintk("<-- %s status=%d\n", __func__, status);
7510 	return status;
7511 }
7512 
7513 static void
7514 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7515 {
7516 	struct nfs4_layoutget *lgp = calldata;
7517 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7518 	struct nfs4_session *session = nfs4_get_session(server);
7519 
7520 	dprintk("--> %s\n", __func__);
7521 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7522 	 * right now covering the LAYOUTGET we are about to send.
7523 	 * However, that is not so catastrophic, and there seems
7524 	 * to be no way to prevent it completely.
7525 	 */
7526 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7527 				&lgp->res.seq_res, task))
7528 		return;
7529 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7530 					  NFS_I(lgp->args.inode)->layout,
7531 					  lgp->args.ctx->state)) {
7532 		rpc_exit(task, NFS4_OK);
7533 	}
7534 }
7535 
7536 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7537 {
7538 	struct nfs4_layoutget *lgp = calldata;
7539 	struct inode *inode = lgp->args.inode;
7540 	struct nfs_server *server = NFS_SERVER(inode);
7541 	struct pnfs_layout_hdr *lo;
7542 	struct nfs4_state *state = NULL;
7543 	unsigned long timeo, now, giveup;
7544 
7545 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7546 
7547 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7548 		goto out;
7549 
7550 	switch (task->tk_status) {
7551 	case 0:
7552 		goto out;
7553 	/*
7554 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7555 	 * (or clients) writing to the same RAID stripe
7556 	 */
7557 	case -NFS4ERR_LAYOUTTRYLATER:
7558 	/*
7559 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7560 	 * existing layout before getting a new one).
7561 	 */
7562 	case -NFS4ERR_RECALLCONFLICT:
7563 		timeo = rpc_get_timeout(task->tk_client);
7564 		giveup = lgp->args.timestamp + timeo;
7565 		now = jiffies;
7566 		if (time_after(giveup, now)) {
7567 			unsigned long delay;
7568 
7569 			/* Delay for:
7570 			 * - Not less then NFS4_POLL_RETRY_MIN.
7571 			 * - One last time a jiffie before we give up
7572 			 * - exponential backoff (time_now minus start_attempt)
7573 			 */
7574 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7575 				    min((giveup - now - 1),
7576 					now - lgp->args.timestamp));
7577 
7578 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7579 				__func__, delay);
7580 			rpc_delay(task, delay);
7581 			task->tk_status = 0;
7582 			rpc_restart_call_prepare(task);
7583 			goto out; /* Do not call nfs4_async_handle_error() */
7584 		}
7585 		break;
7586 	case -NFS4ERR_EXPIRED:
7587 	case -NFS4ERR_BAD_STATEID:
7588 		spin_lock(&inode->i_lock);
7589 		lo = NFS_I(inode)->layout;
7590 		if (!lo || list_empty(&lo->plh_segs)) {
7591 			spin_unlock(&inode->i_lock);
7592 			/* If the open stateid was bad, then recover it. */
7593 			state = lgp->args.ctx->state;
7594 		} else {
7595 			LIST_HEAD(head);
7596 
7597 			/*
7598 			 * Mark the bad layout state as invalid, then retry
7599 			 * with the current stateid.
7600 			 */
7601 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7602 			spin_unlock(&inode->i_lock);
7603 			pnfs_free_lseg_list(&head);
7604 
7605 			task->tk_status = 0;
7606 			rpc_restart_call_prepare(task);
7607 		}
7608 	}
7609 	if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7610 		rpc_restart_call_prepare(task);
7611 out:
7612 	dprintk("<-- %s\n", __func__);
7613 }
7614 
7615 static size_t max_response_pages(struct nfs_server *server)
7616 {
7617 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7618 	return nfs_page_array_len(0, max_resp_sz);
7619 }
7620 
7621 static void nfs4_free_pages(struct page **pages, size_t size)
7622 {
7623 	int i;
7624 
7625 	if (!pages)
7626 		return;
7627 
7628 	for (i = 0; i < size; i++) {
7629 		if (!pages[i])
7630 			break;
7631 		__free_page(pages[i]);
7632 	}
7633 	kfree(pages);
7634 }
7635 
7636 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7637 {
7638 	struct page **pages;
7639 	int i;
7640 
7641 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7642 	if (!pages) {
7643 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7644 		return NULL;
7645 	}
7646 
7647 	for (i = 0; i < size; i++) {
7648 		pages[i] = alloc_page(gfp_flags);
7649 		if (!pages[i]) {
7650 			dprintk("%s: failed to allocate page\n", __func__);
7651 			nfs4_free_pages(pages, size);
7652 			return NULL;
7653 		}
7654 	}
7655 
7656 	return pages;
7657 }
7658 
7659 static void nfs4_layoutget_release(void *calldata)
7660 {
7661 	struct nfs4_layoutget *lgp = calldata;
7662 	struct inode *inode = lgp->args.inode;
7663 	struct nfs_server *server = NFS_SERVER(inode);
7664 	size_t max_pages = max_response_pages(server);
7665 
7666 	dprintk("--> %s\n", __func__);
7667 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7668 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7669 	put_nfs_open_context(lgp->args.ctx);
7670 	kfree(calldata);
7671 	dprintk("<-- %s\n", __func__);
7672 }
7673 
7674 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7675 	.rpc_call_prepare = nfs4_layoutget_prepare,
7676 	.rpc_call_done = nfs4_layoutget_done,
7677 	.rpc_release = nfs4_layoutget_release,
7678 };
7679 
7680 struct pnfs_layout_segment *
7681 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7682 {
7683 	struct inode *inode = lgp->args.inode;
7684 	struct nfs_server *server = NFS_SERVER(inode);
7685 	size_t max_pages = max_response_pages(server);
7686 	struct rpc_task *task;
7687 	struct rpc_message msg = {
7688 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7689 		.rpc_argp = &lgp->args,
7690 		.rpc_resp = &lgp->res,
7691 		.rpc_cred = lgp->cred,
7692 	};
7693 	struct rpc_task_setup task_setup_data = {
7694 		.rpc_client = server->client,
7695 		.rpc_message = &msg,
7696 		.callback_ops = &nfs4_layoutget_call_ops,
7697 		.callback_data = lgp,
7698 		.flags = RPC_TASK_ASYNC,
7699 	};
7700 	struct pnfs_layout_segment *lseg = NULL;
7701 	int status = 0;
7702 
7703 	dprintk("--> %s\n", __func__);
7704 
7705 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7706 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7707 
7708 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7709 	if (!lgp->args.layout.pages) {
7710 		nfs4_layoutget_release(lgp);
7711 		return ERR_PTR(-ENOMEM);
7712 	}
7713 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7714 	lgp->args.timestamp = jiffies;
7715 
7716 	lgp->res.layoutp = &lgp->args.layout;
7717 	lgp->res.seq_res.sr_slot = NULL;
7718 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7719 
7720 	task = rpc_run_task(&task_setup_data);
7721 	if (IS_ERR(task))
7722 		return ERR_CAST(task);
7723 	status = nfs4_wait_for_completion_rpc_task(task);
7724 	if (status == 0)
7725 		status = task->tk_status;
7726 	trace_nfs4_layoutget(lgp->args.ctx,
7727 			&lgp->args.range,
7728 			&lgp->res.range,
7729 			status);
7730 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7731 	if (status == 0 && lgp->res.layoutp->len)
7732 		lseg = pnfs_layout_process(lgp);
7733 	rpc_put_task(task);
7734 	dprintk("<-- %s status=%d\n", __func__, status);
7735 	if (status)
7736 		return ERR_PTR(status);
7737 	return lseg;
7738 }
7739 
7740 static void
7741 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7742 {
7743 	struct nfs4_layoutreturn *lrp = calldata;
7744 
7745 	dprintk("--> %s\n", __func__);
7746 	nfs41_setup_sequence(lrp->clp->cl_session,
7747 			&lrp->args.seq_args,
7748 			&lrp->res.seq_res,
7749 			task);
7750 }
7751 
7752 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7753 {
7754 	struct nfs4_layoutreturn *lrp = calldata;
7755 	struct nfs_server *server;
7756 
7757 	dprintk("--> %s\n", __func__);
7758 
7759 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7760 		return;
7761 
7762 	server = NFS_SERVER(lrp->args.inode);
7763 	switch (task->tk_status) {
7764 	default:
7765 		task->tk_status = 0;
7766 	case 0:
7767 		break;
7768 	case -NFS4ERR_DELAY:
7769 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7770 			break;
7771 		rpc_restart_call_prepare(task);
7772 		return;
7773 	}
7774 	dprintk("<-- %s\n", __func__);
7775 }
7776 
7777 static void nfs4_layoutreturn_release(void *calldata)
7778 {
7779 	struct nfs4_layoutreturn *lrp = calldata;
7780 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7781 
7782 	dprintk("--> %s\n", __func__);
7783 	spin_lock(&lo->plh_inode->i_lock);
7784 	if (lrp->res.lrs_present)
7785 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7786 	lo->plh_block_lgets--;
7787 	spin_unlock(&lo->plh_inode->i_lock);
7788 	pnfs_put_layout_hdr(lrp->args.layout);
7789 	kfree(calldata);
7790 	dprintk("<-- %s\n", __func__);
7791 }
7792 
7793 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7794 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7795 	.rpc_call_done = nfs4_layoutreturn_done,
7796 	.rpc_release = nfs4_layoutreturn_release,
7797 };
7798 
7799 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7800 {
7801 	struct rpc_task *task;
7802 	struct rpc_message msg = {
7803 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7804 		.rpc_argp = &lrp->args,
7805 		.rpc_resp = &lrp->res,
7806 		.rpc_cred = lrp->cred,
7807 	};
7808 	struct rpc_task_setup task_setup_data = {
7809 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7810 		.rpc_message = &msg,
7811 		.callback_ops = &nfs4_layoutreturn_call_ops,
7812 		.callback_data = lrp,
7813 	};
7814 	int status;
7815 
7816 	dprintk("--> %s\n", __func__);
7817 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7818 	task = rpc_run_task(&task_setup_data);
7819 	if (IS_ERR(task))
7820 		return PTR_ERR(task);
7821 	status = task->tk_status;
7822 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7823 	dprintk("<-- %s status=%d\n", __func__, status);
7824 	rpc_put_task(task);
7825 	return status;
7826 }
7827 
7828 static int
7829 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7830 		struct pnfs_device *pdev,
7831 		struct rpc_cred *cred)
7832 {
7833 	struct nfs4_getdeviceinfo_args args = {
7834 		.pdev = pdev,
7835 	};
7836 	struct nfs4_getdeviceinfo_res res = {
7837 		.pdev = pdev,
7838 	};
7839 	struct rpc_message msg = {
7840 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7841 		.rpc_argp = &args,
7842 		.rpc_resp = &res,
7843 		.rpc_cred = cred,
7844 	};
7845 	int status;
7846 
7847 	dprintk("--> %s\n", __func__);
7848 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7849 	dprintk("<-- %s status=%d\n", __func__, status);
7850 
7851 	return status;
7852 }
7853 
7854 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7855 		struct pnfs_device *pdev,
7856 		struct rpc_cred *cred)
7857 {
7858 	struct nfs4_exception exception = { };
7859 	int err;
7860 
7861 	do {
7862 		err = nfs4_handle_exception(server,
7863 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7864 					&exception);
7865 	} while (exception.retry);
7866 	return err;
7867 }
7868 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7869 
7870 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7871 {
7872 	struct nfs4_layoutcommit_data *data = calldata;
7873 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7874 	struct nfs4_session *session = nfs4_get_session(server);
7875 
7876 	nfs41_setup_sequence(session,
7877 			&data->args.seq_args,
7878 			&data->res.seq_res,
7879 			task);
7880 }
7881 
7882 static void
7883 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7884 {
7885 	struct nfs4_layoutcommit_data *data = calldata;
7886 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7887 
7888 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7889 		return;
7890 
7891 	switch (task->tk_status) { /* Just ignore these failures */
7892 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7893 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7894 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7895 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7896 		task->tk_status = 0;
7897 	case 0:
7898 		break;
7899 	default:
7900 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
7901 			rpc_restart_call_prepare(task);
7902 			return;
7903 		}
7904 	}
7905 }
7906 
7907 static void nfs4_layoutcommit_release(void *calldata)
7908 {
7909 	struct nfs4_layoutcommit_data *data = calldata;
7910 
7911 	pnfs_cleanup_layoutcommit(data);
7912 	nfs_post_op_update_inode_force_wcc(data->args.inode,
7913 					   data->res.fattr);
7914 	put_rpccred(data->cred);
7915 	kfree(data);
7916 }
7917 
7918 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7919 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
7920 	.rpc_call_done = nfs4_layoutcommit_done,
7921 	.rpc_release = nfs4_layoutcommit_release,
7922 };
7923 
7924 int
7925 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7926 {
7927 	struct rpc_message msg = {
7928 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7929 		.rpc_argp = &data->args,
7930 		.rpc_resp = &data->res,
7931 		.rpc_cred = data->cred,
7932 	};
7933 	struct rpc_task_setup task_setup_data = {
7934 		.task = &data->task,
7935 		.rpc_client = NFS_CLIENT(data->args.inode),
7936 		.rpc_message = &msg,
7937 		.callback_ops = &nfs4_layoutcommit_ops,
7938 		.callback_data = data,
7939 		.flags = RPC_TASK_ASYNC,
7940 	};
7941 	struct rpc_task *task;
7942 	int status = 0;
7943 
7944 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7945 		"lbw: %llu inode %lu\n",
7946 		data->task.tk_pid, sync,
7947 		data->args.lastbytewritten,
7948 		data->args.inode->i_ino);
7949 
7950 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7951 	task = rpc_run_task(&task_setup_data);
7952 	if (IS_ERR(task))
7953 		return PTR_ERR(task);
7954 	if (sync == false)
7955 		goto out;
7956 	status = nfs4_wait_for_completion_rpc_task(task);
7957 	if (status != 0)
7958 		goto out;
7959 	status = task->tk_status;
7960 	trace_nfs4_layoutcommit(data->args.inode, status);
7961 out:
7962 	dprintk("%s: status %d\n", __func__, status);
7963 	rpc_put_task(task);
7964 	return status;
7965 }
7966 
7967 /**
7968  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7969  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7970  */
7971 static int
7972 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7973 		    struct nfs_fsinfo *info,
7974 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7975 {
7976 	struct nfs41_secinfo_no_name_args args = {
7977 		.style = SECINFO_STYLE_CURRENT_FH,
7978 	};
7979 	struct nfs4_secinfo_res res = {
7980 		.flavors = flavors,
7981 	};
7982 	struct rpc_message msg = {
7983 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
7984 		.rpc_argp = &args,
7985 		.rpc_resp = &res,
7986 	};
7987 	struct rpc_clnt *clnt = server->client;
7988 	struct rpc_cred *cred = NULL;
7989 	int status;
7990 
7991 	if (use_integrity) {
7992 		clnt = server->nfs_client->cl_rpcclient;
7993 		cred = nfs4_get_clid_cred(server->nfs_client);
7994 		msg.rpc_cred = cred;
7995 	}
7996 
7997 	dprintk("--> %s\n", __func__);
7998 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
7999 				&res.seq_res, 0);
8000 	dprintk("<-- %s status=%d\n", __func__, status);
8001 
8002 	if (cred)
8003 		put_rpccred(cred);
8004 
8005 	return status;
8006 }
8007 
8008 static int
8009 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8010 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8011 {
8012 	struct nfs4_exception exception = { };
8013 	int err;
8014 	do {
8015 		/* first try using integrity protection */
8016 		err = -NFS4ERR_WRONGSEC;
8017 
8018 		/* try to use integrity protection with machine cred */
8019 		if (_nfs4_is_integrity_protected(server->nfs_client))
8020 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8021 							  flavors, true);
8022 
8023 		/*
8024 		 * if unable to use integrity protection, or SECINFO with
8025 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8026 		 * disallowed by spec, but exists in deployed servers) use
8027 		 * the current filesystem's rpc_client and the user cred.
8028 		 */
8029 		if (err == -NFS4ERR_WRONGSEC)
8030 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8031 							  flavors, false);
8032 
8033 		switch (err) {
8034 		case 0:
8035 		case -NFS4ERR_WRONGSEC:
8036 		case -ENOTSUPP:
8037 			goto out;
8038 		default:
8039 			err = nfs4_handle_exception(server, err, &exception);
8040 		}
8041 	} while (exception.retry);
8042 out:
8043 	return err;
8044 }
8045 
8046 static int
8047 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8048 		    struct nfs_fsinfo *info)
8049 {
8050 	int err;
8051 	struct page *page;
8052 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8053 	struct nfs4_secinfo_flavors *flavors;
8054 	struct nfs4_secinfo4 *secinfo;
8055 	int i;
8056 
8057 	page = alloc_page(GFP_KERNEL);
8058 	if (!page) {
8059 		err = -ENOMEM;
8060 		goto out;
8061 	}
8062 
8063 	flavors = page_address(page);
8064 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8065 
8066 	/*
8067 	 * Fall back on "guess and check" method if
8068 	 * the server doesn't support SECINFO_NO_NAME
8069 	 */
8070 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8071 		err = nfs4_find_root_sec(server, fhandle, info);
8072 		goto out_freepage;
8073 	}
8074 	if (err)
8075 		goto out_freepage;
8076 
8077 	for (i = 0; i < flavors->num_flavors; i++) {
8078 		secinfo = &flavors->flavors[i];
8079 
8080 		switch (secinfo->flavor) {
8081 		case RPC_AUTH_NULL:
8082 		case RPC_AUTH_UNIX:
8083 		case RPC_AUTH_GSS:
8084 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8085 					&secinfo->flavor_info);
8086 			break;
8087 		default:
8088 			flavor = RPC_AUTH_MAXFLAVOR;
8089 			break;
8090 		}
8091 
8092 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8093 			flavor = RPC_AUTH_MAXFLAVOR;
8094 
8095 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8096 			err = nfs4_lookup_root_sec(server, fhandle,
8097 						   info, flavor);
8098 			if (!err)
8099 				break;
8100 		}
8101 	}
8102 
8103 	if (flavor == RPC_AUTH_MAXFLAVOR)
8104 		err = -EPERM;
8105 
8106 out_freepage:
8107 	put_page(page);
8108 	if (err == -EACCES)
8109 		return -EPERM;
8110 out:
8111 	return err;
8112 }
8113 
8114 static int _nfs41_test_stateid(struct nfs_server *server,
8115 		nfs4_stateid *stateid,
8116 		struct rpc_cred *cred)
8117 {
8118 	int status;
8119 	struct nfs41_test_stateid_args args = {
8120 		.stateid = stateid,
8121 	};
8122 	struct nfs41_test_stateid_res res;
8123 	struct rpc_message msg = {
8124 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8125 		.rpc_argp = &args,
8126 		.rpc_resp = &res,
8127 		.rpc_cred = cred,
8128 	};
8129 	struct rpc_clnt *rpc_client = server->client;
8130 
8131 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8132 		&rpc_client, &msg);
8133 
8134 	dprintk("NFS call  test_stateid %p\n", stateid);
8135 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8136 	nfs4_set_sequence_privileged(&args.seq_args);
8137 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8138 			&args.seq_args, &res.seq_res);
8139 	if (status != NFS_OK) {
8140 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8141 		return status;
8142 	}
8143 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8144 	return -res.status;
8145 }
8146 
8147 /**
8148  * nfs41_test_stateid - perform a TEST_STATEID operation
8149  *
8150  * @server: server / transport on which to perform the operation
8151  * @stateid: state ID to test
8152  * @cred: credential
8153  *
8154  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8155  * Otherwise a negative NFS4ERR value is returned if the operation
8156  * failed or the state ID is not currently valid.
8157  */
8158 static int nfs41_test_stateid(struct nfs_server *server,
8159 		nfs4_stateid *stateid,
8160 		struct rpc_cred *cred)
8161 {
8162 	struct nfs4_exception exception = { };
8163 	int err;
8164 	do {
8165 		err = _nfs41_test_stateid(server, stateid, cred);
8166 		if (err != -NFS4ERR_DELAY)
8167 			break;
8168 		nfs4_handle_exception(server, err, &exception);
8169 	} while (exception.retry);
8170 	return err;
8171 }
8172 
8173 struct nfs_free_stateid_data {
8174 	struct nfs_server *server;
8175 	struct nfs41_free_stateid_args args;
8176 	struct nfs41_free_stateid_res res;
8177 };
8178 
8179 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8180 {
8181 	struct nfs_free_stateid_data *data = calldata;
8182 	nfs41_setup_sequence(nfs4_get_session(data->server),
8183 			&data->args.seq_args,
8184 			&data->res.seq_res,
8185 			task);
8186 }
8187 
8188 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8189 {
8190 	struct nfs_free_stateid_data *data = calldata;
8191 
8192 	nfs41_sequence_done(task, &data->res.seq_res);
8193 
8194 	switch (task->tk_status) {
8195 	case -NFS4ERR_DELAY:
8196 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8197 			rpc_restart_call_prepare(task);
8198 	}
8199 }
8200 
8201 static void nfs41_free_stateid_release(void *calldata)
8202 {
8203 	kfree(calldata);
8204 }
8205 
8206 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8207 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8208 	.rpc_call_done = nfs41_free_stateid_done,
8209 	.rpc_release = nfs41_free_stateid_release,
8210 };
8211 
8212 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8213 		nfs4_stateid *stateid,
8214 		struct rpc_cred *cred,
8215 		bool privileged)
8216 {
8217 	struct rpc_message msg = {
8218 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8219 		.rpc_cred = cred,
8220 	};
8221 	struct rpc_task_setup task_setup = {
8222 		.rpc_client = server->client,
8223 		.rpc_message = &msg,
8224 		.callback_ops = &nfs41_free_stateid_ops,
8225 		.flags = RPC_TASK_ASYNC,
8226 	};
8227 	struct nfs_free_stateid_data *data;
8228 
8229 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8230 		&task_setup.rpc_client, &msg);
8231 
8232 	dprintk("NFS call  free_stateid %p\n", stateid);
8233 	data = kmalloc(sizeof(*data), GFP_NOFS);
8234 	if (!data)
8235 		return ERR_PTR(-ENOMEM);
8236 	data->server = server;
8237 	nfs4_stateid_copy(&data->args.stateid, stateid);
8238 
8239 	task_setup.callback_data = data;
8240 
8241 	msg.rpc_argp = &data->args;
8242 	msg.rpc_resp = &data->res;
8243 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8244 	if (privileged)
8245 		nfs4_set_sequence_privileged(&data->args.seq_args);
8246 
8247 	return rpc_run_task(&task_setup);
8248 }
8249 
8250 /**
8251  * nfs41_free_stateid - perform a FREE_STATEID operation
8252  *
8253  * @server: server / transport on which to perform the operation
8254  * @stateid: state ID to release
8255  * @cred: credential
8256  *
8257  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8258  * negative NFS4ERR value is returned.
8259  */
8260 static int nfs41_free_stateid(struct nfs_server *server,
8261 		nfs4_stateid *stateid,
8262 		struct rpc_cred *cred)
8263 {
8264 	struct rpc_task *task;
8265 	int ret;
8266 
8267 	task = _nfs41_free_stateid(server, stateid, cred, true);
8268 	if (IS_ERR(task))
8269 		return PTR_ERR(task);
8270 	ret = rpc_wait_for_completion_task(task);
8271 	if (!ret)
8272 		ret = task->tk_status;
8273 	rpc_put_task(task);
8274 	return ret;
8275 }
8276 
8277 static void
8278 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8279 {
8280 	struct rpc_task *task;
8281 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8282 
8283 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8284 	nfs4_free_lock_state(server, lsp);
8285 	if (IS_ERR(task))
8286 		return;
8287 	rpc_put_task(task);
8288 }
8289 
8290 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8291 		const nfs4_stateid *s2)
8292 {
8293 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8294 		return false;
8295 
8296 	if (s1->seqid == s2->seqid)
8297 		return true;
8298 	if (s1->seqid == 0 || s2->seqid == 0)
8299 		return true;
8300 
8301 	return false;
8302 }
8303 
8304 #endif /* CONFIG_NFS_V4_1 */
8305 
8306 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8307 		const nfs4_stateid *s2)
8308 {
8309 	return nfs4_stateid_match(s1, s2);
8310 }
8311 
8312 
8313 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8314 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8315 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8316 	.recover_open	= nfs4_open_reclaim,
8317 	.recover_lock	= nfs4_lock_reclaim,
8318 	.establish_clid = nfs4_init_clientid,
8319 	.detect_trunking = nfs40_discover_server_trunking,
8320 };
8321 
8322 #if defined(CONFIG_NFS_V4_1)
8323 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8324 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8325 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8326 	.recover_open	= nfs4_open_reclaim,
8327 	.recover_lock	= nfs4_lock_reclaim,
8328 	.establish_clid = nfs41_init_clientid,
8329 	.reclaim_complete = nfs41_proc_reclaim_complete,
8330 	.detect_trunking = nfs41_discover_server_trunking,
8331 };
8332 #endif /* CONFIG_NFS_V4_1 */
8333 
8334 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8335 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8336 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8337 	.recover_open	= nfs40_open_expired,
8338 	.recover_lock	= nfs4_lock_expired,
8339 	.establish_clid = nfs4_init_clientid,
8340 };
8341 
8342 #if defined(CONFIG_NFS_V4_1)
8343 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8344 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8345 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8346 	.recover_open	= nfs41_open_expired,
8347 	.recover_lock	= nfs41_lock_expired,
8348 	.establish_clid = nfs41_init_clientid,
8349 };
8350 #endif /* CONFIG_NFS_V4_1 */
8351 
8352 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8353 	.sched_state_renewal = nfs4_proc_async_renew,
8354 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8355 	.renew_lease = nfs4_proc_renew,
8356 };
8357 
8358 #if defined(CONFIG_NFS_V4_1)
8359 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8360 	.sched_state_renewal = nfs41_proc_async_sequence,
8361 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8362 	.renew_lease = nfs4_proc_sequence,
8363 };
8364 #endif
8365 
8366 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8367 	.get_locations = _nfs40_proc_get_locations,
8368 	.fsid_present = _nfs40_proc_fsid_present,
8369 };
8370 
8371 #if defined(CONFIG_NFS_V4_1)
8372 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8373 	.get_locations = _nfs41_proc_get_locations,
8374 	.fsid_present = _nfs41_proc_fsid_present,
8375 };
8376 #endif	/* CONFIG_NFS_V4_1 */
8377 
8378 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8379 	.minor_version = 0,
8380 	.init_caps = NFS_CAP_READDIRPLUS
8381 		| NFS_CAP_ATOMIC_OPEN
8382 		| NFS_CAP_CHANGE_ATTR
8383 		| NFS_CAP_POSIX_LOCK,
8384 	.init_client = nfs40_init_client,
8385 	.shutdown_client = nfs40_shutdown_client,
8386 	.match_stateid = nfs4_match_stateid,
8387 	.find_root_sec = nfs4_find_root_sec,
8388 	.free_lock_state = nfs4_release_lockowner,
8389 	.call_sync_ops = &nfs40_call_sync_ops,
8390 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8391 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8392 	.state_renewal_ops = &nfs40_state_renewal_ops,
8393 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8394 };
8395 
8396 #if defined(CONFIG_NFS_V4_1)
8397 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8398 	.minor_version = 1,
8399 	.init_caps = NFS_CAP_READDIRPLUS
8400 		| NFS_CAP_ATOMIC_OPEN
8401 		| NFS_CAP_CHANGE_ATTR
8402 		| NFS_CAP_POSIX_LOCK
8403 		| NFS_CAP_STATEID_NFSV41
8404 		| NFS_CAP_ATOMIC_OPEN_V1,
8405 	.init_client = nfs41_init_client,
8406 	.shutdown_client = nfs41_shutdown_client,
8407 	.match_stateid = nfs41_match_stateid,
8408 	.find_root_sec = nfs41_find_root_sec,
8409 	.free_lock_state = nfs41_free_lock_state,
8410 	.call_sync_ops = &nfs41_call_sync_ops,
8411 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8412 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8413 	.state_renewal_ops = &nfs41_state_renewal_ops,
8414 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8415 };
8416 #endif
8417 
8418 #if defined(CONFIG_NFS_V4_2)
8419 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8420 	.minor_version = 2,
8421 	.init_caps = NFS_CAP_READDIRPLUS
8422 		| NFS_CAP_ATOMIC_OPEN
8423 		| NFS_CAP_CHANGE_ATTR
8424 		| NFS_CAP_POSIX_LOCK
8425 		| NFS_CAP_STATEID_NFSV41
8426 		| NFS_CAP_ATOMIC_OPEN_V1
8427 		| NFS_CAP_ALLOCATE
8428 		| NFS_CAP_DEALLOCATE
8429 		| NFS_CAP_SEEK,
8430 	.init_client = nfs41_init_client,
8431 	.shutdown_client = nfs41_shutdown_client,
8432 	.match_stateid = nfs41_match_stateid,
8433 	.find_root_sec = nfs41_find_root_sec,
8434 	.free_lock_state = nfs41_free_lock_state,
8435 	.call_sync_ops = &nfs41_call_sync_ops,
8436 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8437 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8438 	.state_renewal_ops = &nfs41_state_renewal_ops,
8439 };
8440 #endif
8441 
8442 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8443 	[0] = &nfs_v4_0_minor_ops,
8444 #if defined(CONFIG_NFS_V4_1)
8445 	[1] = &nfs_v4_1_minor_ops,
8446 #endif
8447 #if defined(CONFIG_NFS_V4_2)
8448 	[2] = &nfs_v4_2_minor_ops,
8449 #endif
8450 };
8451 
8452 static const struct inode_operations nfs4_dir_inode_operations = {
8453 	.create		= nfs_create,
8454 	.lookup		= nfs_lookup,
8455 	.atomic_open	= nfs_atomic_open,
8456 	.link		= nfs_link,
8457 	.unlink		= nfs_unlink,
8458 	.symlink	= nfs_symlink,
8459 	.mkdir		= nfs_mkdir,
8460 	.rmdir		= nfs_rmdir,
8461 	.mknod		= nfs_mknod,
8462 	.rename		= nfs_rename,
8463 	.permission	= nfs_permission,
8464 	.getattr	= nfs_getattr,
8465 	.setattr	= nfs_setattr,
8466 	.getxattr	= generic_getxattr,
8467 	.setxattr	= generic_setxattr,
8468 	.listxattr	= generic_listxattr,
8469 	.removexattr	= generic_removexattr,
8470 };
8471 
8472 static const struct inode_operations nfs4_file_inode_operations = {
8473 	.permission	= nfs_permission,
8474 	.getattr	= nfs_getattr,
8475 	.setattr	= nfs_setattr,
8476 	.getxattr	= generic_getxattr,
8477 	.setxattr	= generic_setxattr,
8478 	.listxattr	= generic_listxattr,
8479 	.removexattr	= generic_removexattr,
8480 };
8481 
8482 const struct nfs_rpc_ops nfs_v4_clientops = {
8483 	.version	= 4,			/* protocol version */
8484 	.dentry_ops	= &nfs4_dentry_operations,
8485 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8486 	.file_inode_ops	= &nfs4_file_inode_operations,
8487 	.file_ops	= &nfs4_file_operations,
8488 	.getroot	= nfs4_proc_get_root,
8489 	.submount	= nfs4_submount,
8490 	.try_mount	= nfs4_try_mount,
8491 	.getattr	= nfs4_proc_getattr,
8492 	.setattr	= nfs4_proc_setattr,
8493 	.lookup		= nfs4_proc_lookup,
8494 	.access		= nfs4_proc_access,
8495 	.readlink	= nfs4_proc_readlink,
8496 	.create		= nfs4_proc_create,
8497 	.remove		= nfs4_proc_remove,
8498 	.unlink_setup	= nfs4_proc_unlink_setup,
8499 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8500 	.unlink_done	= nfs4_proc_unlink_done,
8501 	.rename_setup	= nfs4_proc_rename_setup,
8502 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8503 	.rename_done	= nfs4_proc_rename_done,
8504 	.link		= nfs4_proc_link,
8505 	.symlink	= nfs4_proc_symlink,
8506 	.mkdir		= nfs4_proc_mkdir,
8507 	.rmdir		= nfs4_proc_remove,
8508 	.readdir	= nfs4_proc_readdir,
8509 	.mknod		= nfs4_proc_mknod,
8510 	.statfs		= nfs4_proc_statfs,
8511 	.fsinfo		= nfs4_proc_fsinfo,
8512 	.pathconf	= nfs4_proc_pathconf,
8513 	.set_capabilities = nfs4_server_capabilities,
8514 	.decode_dirent	= nfs4_decode_dirent,
8515 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8516 	.read_setup	= nfs4_proc_read_setup,
8517 	.read_done	= nfs4_read_done,
8518 	.write_setup	= nfs4_proc_write_setup,
8519 	.write_done	= nfs4_write_done,
8520 	.commit_setup	= nfs4_proc_commit_setup,
8521 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8522 	.commit_done	= nfs4_commit_done,
8523 	.lock		= nfs4_proc_lock,
8524 	.clear_acl_cache = nfs4_zap_acl_attr,
8525 	.close_context  = nfs4_close_context,
8526 	.open_context	= nfs4_atomic_open,
8527 	.have_delegation = nfs4_have_delegation,
8528 	.return_delegation = nfs4_inode_return_delegation,
8529 	.alloc_client	= nfs4_alloc_client,
8530 	.init_client	= nfs4_init_client,
8531 	.free_client	= nfs4_free_client,
8532 	.create_server	= nfs4_create_server,
8533 	.clone_server	= nfs_clone_server,
8534 };
8535 
8536 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8537 	.prefix	= XATTR_NAME_NFSV4_ACL,
8538 	.list	= nfs4_xattr_list_nfs4_acl,
8539 	.get	= nfs4_xattr_get_nfs4_acl,
8540 	.set	= nfs4_xattr_set_nfs4_acl,
8541 };
8542 
8543 const struct xattr_handler *nfs4_xattr_handlers[] = {
8544 	&nfs4_xattr_nfs4_acl_handler,
8545 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8546 	&nfs4_xattr_nfs4_label_handler,
8547 #endif
8548 	NULL
8549 };
8550 
8551 /*
8552  * Local variables:
8553  *  c-basic-offset: 8
8554  * End:
8555  */
8556