xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 588b48ca)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68 
69 #include "nfs4trace.h"
70 
71 #define NFSDBG_FACILITY		NFSDBG_PROC
72 
73 #define NFS4_POLL_RETRY_MIN	(HZ/10)
74 #define NFS4_POLL_RETRY_MAX	(15*HZ)
75 
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_ACCESS:
162 		return -EACCES;
163 	case -NFS4ERR_FILE_OPEN:
164 		return -EBUSY;
165 	default:
166 		dprintk("%s could not handle NFSv4 error %d\n",
167 				__func__, -err);
168 		break;
169 	}
170 	return -EIO;
171 }
172 
173 /*
174  * This is our standard bitmap for GETATTR requests.
175  */
176 const u32 nfs4_fattr_bitmap[3] = {
177 	FATTR4_WORD0_TYPE
178 	| FATTR4_WORD0_CHANGE
179 	| FATTR4_WORD0_SIZE
180 	| FATTR4_WORD0_FSID
181 	| FATTR4_WORD0_FILEID,
182 	FATTR4_WORD1_MODE
183 	| FATTR4_WORD1_NUMLINKS
184 	| FATTR4_WORD1_OWNER
185 	| FATTR4_WORD1_OWNER_GROUP
186 	| FATTR4_WORD1_RAWDEV
187 	| FATTR4_WORD1_SPACE_USED
188 	| FATTR4_WORD1_TIME_ACCESS
189 	| FATTR4_WORD1_TIME_METADATA
190 	| FATTR4_WORD1_TIME_MODIFY,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192 	FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195 
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197 	FATTR4_WORD0_TYPE
198 	| FATTR4_WORD0_CHANGE
199 	| FATTR4_WORD0_SIZE
200 	| FATTR4_WORD0_FSID
201 	| FATTR4_WORD0_FILEID,
202 	FATTR4_WORD1_MODE
203 	| FATTR4_WORD1_NUMLINKS
204 	| FATTR4_WORD1_OWNER
205 	| FATTR4_WORD1_OWNER_GROUP
206 	| FATTR4_WORD1_RAWDEV
207 	| FATTR4_WORD1_SPACE_USED
208 	| FATTR4_WORD1_TIME_ACCESS
209 	| FATTR4_WORD1_TIME_METADATA
210 	| FATTR4_WORD1_TIME_MODIFY,
211 	FATTR4_WORD2_MDSTHRESHOLD
212 };
213 
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215 	FATTR4_WORD0_TYPE
216 	| FATTR4_WORD0_CHANGE
217 	| FATTR4_WORD0_FILEID,
218 };
219 
220 const u32 nfs4_statfs_bitmap[3] = {
221 	FATTR4_WORD0_FILES_AVAIL
222 	| FATTR4_WORD0_FILES_FREE
223 	| FATTR4_WORD0_FILES_TOTAL,
224 	FATTR4_WORD1_SPACE_AVAIL
225 	| FATTR4_WORD1_SPACE_FREE
226 	| FATTR4_WORD1_SPACE_TOTAL
227 };
228 
229 const u32 nfs4_pathconf_bitmap[3] = {
230 	FATTR4_WORD0_MAXLINK
231 	| FATTR4_WORD0_MAXNAME,
232 	0
233 };
234 
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236 			| FATTR4_WORD0_MAXREAD
237 			| FATTR4_WORD0_MAXWRITE
238 			| FATTR4_WORD0_LEASE_TIME,
239 			FATTR4_WORD1_TIME_DELTA
240 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
241 			FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243 
244 const u32 nfs4_fs_locations_bitmap[3] = {
245 	FATTR4_WORD0_TYPE
246 	| FATTR4_WORD0_CHANGE
247 	| FATTR4_WORD0_SIZE
248 	| FATTR4_WORD0_FSID
249 	| FATTR4_WORD0_FILEID
250 	| FATTR4_WORD0_FS_LOCATIONS,
251 	FATTR4_WORD1_MODE
252 	| FATTR4_WORD1_NUMLINKS
253 	| FATTR4_WORD1_OWNER
254 	| FATTR4_WORD1_OWNER_GROUP
255 	| FATTR4_WORD1_RAWDEV
256 	| FATTR4_WORD1_SPACE_USED
257 	| FATTR4_WORD1_TIME_ACCESS
258 	| FATTR4_WORD1_TIME_METADATA
259 	| FATTR4_WORD1_TIME_MODIFY
260 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262 
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264 		struct nfs4_readdir_arg *readdir)
265 {
266 	__be32 *start, *p;
267 
268 	if (cookie > 2) {
269 		readdir->cookie = cookie;
270 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271 		return;
272 	}
273 
274 	readdir->cookie = 0;
275 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276 	if (cookie == 2)
277 		return;
278 
279 	/*
280 	 * NFSv4 servers do not return entries for '.' and '..'
281 	 * Therefore, we fake these entries here.  We let '.'
282 	 * have cookie 0 and '..' have cookie 1.  Note that
283 	 * when talking to the server, we always send cookie 0
284 	 * instead of 1 or 2.
285 	 */
286 	start = p = kmap_atomic(*readdir->pages);
287 
288 	if (cookie == 0) {
289 		*p++ = xdr_one;                                  /* next */
290 		*p++ = xdr_zero;                   /* cookie, first word */
291 		*p++ = xdr_one;                   /* cookie, second word */
292 		*p++ = xdr_one;                             /* entry len */
293 		memcpy(p, ".\0\0\0", 4);                        /* entry */
294 		p++;
295 		*p++ = xdr_one;                         /* bitmap length */
296 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297 		*p++ = htonl(8);              /* attribute buffer length */
298 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
299 	}
300 
301 	*p++ = xdr_one;                                  /* next */
302 	*p++ = xdr_zero;                   /* cookie, first word */
303 	*p++ = xdr_two;                   /* cookie, second word */
304 	*p++ = xdr_two;                             /* entry len */
305 	memcpy(p, "..\0\0", 4);                         /* entry */
306 	p++;
307 	*p++ = xdr_one;                         /* bitmap length */
308 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309 	*p++ = htonl(8);              /* attribute buffer length */
310 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
311 
312 	readdir->pgbase = (char *)p - (char *)start;
313 	readdir->count -= readdir->pgbase;
314 	kunmap_atomic(start);
315 }
316 
317 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
318 {
319 	int res = 0;
320 
321 	might_sleep();
322 
323 	if (*timeout <= 0)
324 		*timeout = NFS4_POLL_RETRY_MIN;
325 	if (*timeout > NFS4_POLL_RETRY_MAX)
326 		*timeout = NFS4_POLL_RETRY_MAX;
327 	freezable_schedule_timeout_killable_unsafe(*timeout);
328 	if (fatal_signal_pending(current))
329 		res = -ERESTARTSYS;
330 	*timeout <<= 1;
331 	return res;
332 }
333 
334 /* This is the error handling routine for processes that are allowed
335  * to sleep.
336  */
337 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
338 {
339 	struct nfs_client *clp = server->nfs_client;
340 	struct nfs4_state *state = exception->state;
341 	struct inode *inode = exception->inode;
342 	int ret = errorcode;
343 
344 	exception->retry = 0;
345 	switch(errorcode) {
346 		case 0:
347 			return 0;
348 		case -NFS4ERR_OPENMODE:
349 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
350 				nfs4_inode_return_delegation(inode);
351 				exception->retry = 1;
352 				return 0;
353 			}
354 			if (state == NULL)
355 				break;
356 			ret = nfs4_schedule_stateid_recovery(server, state);
357 			if (ret < 0)
358 				break;
359 			goto wait_on_recovery;
360 		case -NFS4ERR_DELEG_REVOKED:
361 		case -NFS4ERR_ADMIN_REVOKED:
362 		case -NFS4ERR_BAD_STATEID:
363 			if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
364 				nfs_remove_bad_delegation(inode);
365 				exception->retry = 1;
366 				break;
367 			}
368 			if (state == NULL)
369 				break;
370 			ret = nfs4_schedule_stateid_recovery(server, state);
371 			if (ret < 0)
372 				break;
373 			goto wait_on_recovery;
374 		case -NFS4ERR_EXPIRED:
375 			if (state != NULL) {
376 				ret = nfs4_schedule_stateid_recovery(server, state);
377 				if (ret < 0)
378 					break;
379 			}
380 		case -NFS4ERR_STALE_STATEID:
381 		case -NFS4ERR_STALE_CLIENTID:
382 			nfs4_schedule_lease_recovery(clp);
383 			goto wait_on_recovery;
384 		case -NFS4ERR_MOVED:
385 			ret = nfs4_schedule_migration_recovery(server);
386 			if (ret < 0)
387 				break;
388 			goto wait_on_recovery;
389 		case -NFS4ERR_LEASE_MOVED:
390 			nfs4_schedule_lease_moved_recovery(clp);
391 			goto wait_on_recovery;
392 #if defined(CONFIG_NFS_V4_1)
393 		case -NFS4ERR_BADSESSION:
394 		case -NFS4ERR_BADSLOT:
395 		case -NFS4ERR_BAD_HIGH_SLOT:
396 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
397 		case -NFS4ERR_DEADSESSION:
398 		case -NFS4ERR_SEQ_FALSE_RETRY:
399 		case -NFS4ERR_SEQ_MISORDERED:
400 			dprintk("%s ERROR: %d Reset session\n", __func__,
401 				errorcode);
402 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
403 			goto wait_on_recovery;
404 #endif /* defined(CONFIG_NFS_V4_1) */
405 		case -NFS4ERR_FILE_OPEN:
406 			if (exception->timeout > HZ) {
407 				/* We have retried a decent amount, time to
408 				 * fail
409 				 */
410 				ret = -EBUSY;
411 				break;
412 			}
413 		case -NFS4ERR_GRACE:
414 		case -NFS4ERR_DELAY:
415 			ret = nfs4_delay(server->client, &exception->timeout);
416 			if (ret != 0)
417 				break;
418 		case -NFS4ERR_RETRY_UNCACHED_REP:
419 		case -NFS4ERR_OLD_STATEID:
420 			exception->retry = 1;
421 			break;
422 		case -NFS4ERR_BADOWNER:
423 			/* The following works around a Linux server bug! */
424 		case -NFS4ERR_BADNAME:
425 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
426 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
427 				exception->retry = 1;
428 				printk(KERN_WARNING "NFS: v4 server %s "
429 						"does not accept raw "
430 						"uid/gids. "
431 						"Reenabling the idmapper.\n",
432 						server->nfs_client->cl_hostname);
433 			}
434 	}
435 	/* We failed to handle the error */
436 	return nfs4_map_errors(ret);
437 wait_on_recovery:
438 	ret = nfs4_wait_clnt_recover(clp);
439 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
440 		return -EIO;
441 	if (ret == 0)
442 		exception->retry = 1;
443 	return ret;
444 }
445 
446 /*
447  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
448  * or 'false' otherwise.
449  */
450 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
451 {
452 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
453 
454 	if (flavor == RPC_AUTH_GSS_KRB5I ||
455 	    flavor == RPC_AUTH_GSS_KRB5P)
456 		return true;
457 
458 	return false;
459 }
460 
461 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
462 {
463 	spin_lock(&clp->cl_lock);
464 	if (time_before(clp->cl_last_renewal,timestamp))
465 		clp->cl_last_renewal = timestamp;
466 	spin_unlock(&clp->cl_lock);
467 }
468 
469 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
470 {
471 	do_renew_lease(server->nfs_client, timestamp);
472 }
473 
474 struct nfs4_call_sync_data {
475 	const struct nfs_server *seq_server;
476 	struct nfs4_sequence_args *seq_args;
477 	struct nfs4_sequence_res *seq_res;
478 };
479 
480 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
481 			       struct nfs4_sequence_res *res, int cache_reply)
482 {
483 	args->sa_slot = NULL;
484 	args->sa_cache_this = cache_reply;
485 	args->sa_privileged = 0;
486 
487 	res->sr_slot = NULL;
488 }
489 
490 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
491 {
492 	args->sa_privileged = 1;
493 }
494 
495 static int nfs40_setup_sequence(const struct nfs_server *server,
496 				struct nfs4_sequence_args *args,
497 				struct nfs4_sequence_res *res,
498 				struct rpc_task *task)
499 {
500 	struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
501 	struct nfs4_slot *slot;
502 
503 	/* slot already allocated? */
504 	if (res->sr_slot != NULL)
505 		goto out_start;
506 
507 	spin_lock(&tbl->slot_tbl_lock);
508 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
509 		goto out_sleep;
510 
511 	slot = nfs4_alloc_slot(tbl);
512 	if (IS_ERR(slot)) {
513 		if (slot == ERR_PTR(-ENOMEM))
514 			task->tk_timeout = HZ >> 2;
515 		goto out_sleep;
516 	}
517 	spin_unlock(&tbl->slot_tbl_lock);
518 
519 	args->sa_slot = slot;
520 	res->sr_slot = slot;
521 
522 out_start:
523 	rpc_call_start(task);
524 	return 0;
525 
526 out_sleep:
527 	if (args->sa_privileged)
528 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
529 				NULL, RPC_PRIORITY_PRIVILEGED);
530 	else
531 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
532 	spin_unlock(&tbl->slot_tbl_lock);
533 	return -EAGAIN;
534 }
535 
536 static int nfs40_sequence_done(struct rpc_task *task,
537 			       struct nfs4_sequence_res *res)
538 {
539 	struct nfs4_slot *slot = res->sr_slot;
540 	struct nfs4_slot_table *tbl;
541 
542 	if (slot == NULL)
543 		goto out;
544 
545 	tbl = slot->table;
546 	spin_lock(&tbl->slot_tbl_lock);
547 	if (!nfs41_wake_and_assign_slot(tbl, slot))
548 		nfs4_free_slot(tbl, slot);
549 	spin_unlock(&tbl->slot_tbl_lock);
550 
551 	res->sr_slot = NULL;
552 out:
553 	return 1;
554 }
555 
556 #if defined(CONFIG_NFS_V4_1)
557 
558 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
559 {
560 	struct nfs4_session *session;
561 	struct nfs4_slot_table *tbl;
562 	struct nfs4_slot *slot = res->sr_slot;
563 	bool send_new_highest_used_slotid = false;
564 
565 	tbl = slot->table;
566 	session = tbl->session;
567 
568 	spin_lock(&tbl->slot_tbl_lock);
569 	/* Be nice to the server: try to ensure that the last transmitted
570 	 * value for highest_user_slotid <= target_highest_slotid
571 	 */
572 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
573 		send_new_highest_used_slotid = true;
574 
575 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
576 		send_new_highest_used_slotid = false;
577 		goto out_unlock;
578 	}
579 	nfs4_free_slot(tbl, slot);
580 
581 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
582 		send_new_highest_used_slotid = false;
583 out_unlock:
584 	spin_unlock(&tbl->slot_tbl_lock);
585 	res->sr_slot = NULL;
586 	if (send_new_highest_used_slotid)
587 		nfs41_server_notify_highest_slotid_update(session->clp);
588 }
589 
590 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
591 {
592 	struct nfs4_session *session;
593 	struct nfs4_slot *slot = res->sr_slot;
594 	struct nfs_client *clp;
595 	bool interrupted = false;
596 	int ret = 1;
597 
598 	if (slot == NULL)
599 		goto out_noaction;
600 	/* don't increment the sequence number if the task wasn't sent */
601 	if (!RPC_WAS_SENT(task))
602 		goto out;
603 
604 	session = slot->table->session;
605 
606 	if (slot->interrupted) {
607 		slot->interrupted = 0;
608 		interrupted = true;
609 	}
610 
611 	trace_nfs4_sequence_done(session, res);
612 	/* Check the SEQUENCE operation status */
613 	switch (res->sr_status) {
614 	case 0:
615 		/* Update the slot's sequence and clientid lease timer */
616 		++slot->seq_nr;
617 		clp = session->clp;
618 		do_renew_lease(clp, res->sr_timestamp);
619 		/* Check sequence flags */
620 		if (res->sr_status_flags != 0)
621 			nfs4_schedule_lease_recovery(clp);
622 		nfs41_update_target_slotid(slot->table, slot, res);
623 		break;
624 	case 1:
625 		/*
626 		 * sr_status remains 1 if an RPC level error occurred.
627 		 * The server may or may not have processed the sequence
628 		 * operation..
629 		 * Mark the slot as having hosted an interrupted RPC call.
630 		 */
631 		slot->interrupted = 1;
632 		goto out;
633 	case -NFS4ERR_DELAY:
634 		/* The server detected a resend of the RPC call and
635 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
636 		 * of RFC5661.
637 		 */
638 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
639 			__func__,
640 			slot->slot_nr,
641 			slot->seq_nr);
642 		goto out_retry;
643 	case -NFS4ERR_BADSLOT:
644 		/*
645 		 * The slot id we used was probably retired. Try again
646 		 * using a different slot id.
647 		 */
648 		goto retry_nowait;
649 	case -NFS4ERR_SEQ_MISORDERED:
650 		/*
651 		 * Was the last operation on this sequence interrupted?
652 		 * If so, retry after bumping the sequence number.
653 		 */
654 		if (interrupted) {
655 			++slot->seq_nr;
656 			goto retry_nowait;
657 		}
658 		/*
659 		 * Could this slot have been previously retired?
660 		 * If so, then the server may be expecting seq_nr = 1!
661 		 */
662 		if (slot->seq_nr != 1) {
663 			slot->seq_nr = 1;
664 			goto retry_nowait;
665 		}
666 		break;
667 	case -NFS4ERR_SEQ_FALSE_RETRY:
668 		++slot->seq_nr;
669 		goto retry_nowait;
670 	default:
671 		/* Just update the slot sequence no. */
672 		++slot->seq_nr;
673 	}
674 out:
675 	/* The session may be reset by one of the error handlers. */
676 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
677 	nfs41_sequence_free_slot(res);
678 out_noaction:
679 	return ret;
680 retry_nowait:
681 	if (rpc_restart_call_prepare(task)) {
682 		task->tk_status = 0;
683 		ret = 0;
684 	}
685 	goto out;
686 out_retry:
687 	if (!rpc_restart_call(task))
688 		goto out;
689 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
690 	return 0;
691 }
692 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
693 
694 static int nfs4_sequence_done(struct rpc_task *task,
695 			       struct nfs4_sequence_res *res)
696 {
697 	if (res->sr_slot == NULL)
698 		return 1;
699 	if (!res->sr_slot->table->session)
700 		return nfs40_sequence_done(task, res);
701 	return nfs41_sequence_done(task, res);
702 }
703 
704 int nfs41_setup_sequence(struct nfs4_session *session,
705 				struct nfs4_sequence_args *args,
706 				struct nfs4_sequence_res *res,
707 				struct rpc_task *task)
708 {
709 	struct nfs4_slot *slot;
710 	struct nfs4_slot_table *tbl;
711 
712 	dprintk("--> %s\n", __func__);
713 	/* slot already allocated? */
714 	if (res->sr_slot != NULL)
715 		goto out_success;
716 
717 	tbl = &session->fc_slot_table;
718 
719 	task->tk_timeout = 0;
720 
721 	spin_lock(&tbl->slot_tbl_lock);
722 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
723 	    !args->sa_privileged) {
724 		/* The state manager will wait until the slot table is empty */
725 		dprintk("%s session is draining\n", __func__);
726 		goto out_sleep;
727 	}
728 
729 	slot = nfs4_alloc_slot(tbl);
730 	if (IS_ERR(slot)) {
731 		/* If out of memory, try again in 1/4 second */
732 		if (slot == ERR_PTR(-ENOMEM))
733 			task->tk_timeout = HZ >> 2;
734 		dprintk("<-- %s: no free slots\n", __func__);
735 		goto out_sleep;
736 	}
737 	spin_unlock(&tbl->slot_tbl_lock);
738 
739 	args->sa_slot = slot;
740 
741 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
742 			slot->slot_nr, slot->seq_nr);
743 
744 	res->sr_slot = slot;
745 	res->sr_timestamp = jiffies;
746 	res->sr_status_flags = 0;
747 	/*
748 	 * sr_status is only set in decode_sequence, and so will remain
749 	 * set to 1 if an rpc level failure occurs.
750 	 */
751 	res->sr_status = 1;
752 	trace_nfs4_setup_sequence(session, args);
753 out_success:
754 	rpc_call_start(task);
755 	return 0;
756 out_sleep:
757 	/* Privileged tasks are queued with top priority */
758 	if (args->sa_privileged)
759 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
760 				NULL, RPC_PRIORITY_PRIVILEGED);
761 	else
762 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
763 	spin_unlock(&tbl->slot_tbl_lock);
764 	return -EAGAIN;
765 }
766 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
767 
768 static int nfs4_setup_sequence(const struct nfs_server *server,
769 			       struct nfs4_sequence_args *args,
770 			       struct nfs4_sequence_res *res,
771 			       struct rpc_task *task)
772 {
773 	struct nfs4_session *session = nfs4_get_session(server);
774 	int ret = 0;
775 
776 	if (!session)
777 		return nfs40_setup_sequence(server, args, res, task);
778 
779 	dprintk("--> %s clp %p session %p sr_slot %u\n",
780 		__func__, session->clp, session, res->sr_slot ?
781 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
782 
783 	ret = nfs41_setup_sequence(session, args, res, task);
784 
785 	dprintk("<-- %s status=%d\n", __func__, ret);
786 	return ret;
787 }
788 
789 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
790 {
791 	struct nfs4_call_sync_data *data = calldata;
792 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
793 
794 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
795 
796 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
797 }
798 
799 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
800 {
801 	struct nfs4_call_sync_data *data = calldata;
802 
803 	nfs41_sequence_done(task, data->seq_res);
804 }
805 
806 static const struct rpc_call_ops nfs41_call_sync_ops = {
807 	.rpc_call_prepare = nfs41_call_sync_prepare,
808 	.rpc_call_done = nfs41_call_sync_done,
809 };
810 
811 #else	/* !CONFIG_NFS_V4_1 */
812 
813 static int nfs4_setup_sequence(const struct nfs_server *server,
814 			       struct nfs4_sequence_args *args,
815 			       struct nfs4_sequence_res *res,
816 			       struct rpc_task *task)
817 {
818 	return nfs40_setup_sequence(server, args, res, task);
819 }
820 
821 static int nfs4_sequence_done(struct rpc_task *task,
822 			       struct nfs4_sequence_res *res)
823 {
824 	return nfs40_sequence_done(task, res);
825 }
826 
827 #endif	/* !CONFIG_NFS_V4_1 */
828 
829 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
830 {
831 	struct nfs4_call_sync_data *data = calldata;
832 	nfs4_setup_sequence(data->seq_server,
833 				data->seq_args, data->seq_res, task);
834 }
835 
836 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
837 {
838 	struct nfs4_call_sync_data *data = calldata;
839 	nfs4_sequence_done(task, data->seq_res);
840 }
841 
842 static const struct rpc_call_ops nfs40_call_sync_ops = {
843 	.rpc_call_prepare = nfs40_call_sync_prepare,
844 	.rpc_call_done = nfs40_call_sync_done,
845 };
846 
847 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
848 				   struct nfs_server *server,
849 				   struct rpc_message *msg,
850 				   struct nfs4_sequence_args *args,
851 				   struct nfs4_sequence_res *res)
852 {
853 	int ret;
854 	struct rpc_task *task;
855 	struct nfs_client *clp = server->nfs_client;
856 	struct nfs4_call_sync_data data = {
857 		.seq_server = server,
858 		.seq_args = args,
859 		.seq_res = res,
860 	};
861 	struct rpc_task_setup task_setup = {
862 		.rpc_client = clnt,
863 		.rpc_message = msg,
864 		.callback_ops = clp->cl_mvops->call_sync_ops,
865 		.callback_data = &data
866 	};
867 
868 	task = rpc_run_task(&task_setup);
869 	if (IS_ERR(task))
870 		ret = PTR_ERR(task);
871 	else {
872 		ret = task->tk_status;
873 		rpc_put_task(task);
874 	}
875 	return ret;
876 }
877 
878 static
879 int nfs4_call_sync(struct rpc_clnt *clnt,
880 		   struct nfs_server *server,
881 		   struct rpc_message *msg,
882 		   struct nfs4_sequence_args *args,
883 		   struct nfs4_sequence_res *res,
884 		   int cache_reply)
885 {
886 	nfs4_init_sequence(args, res, cache_reply);
887 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
888 }
889 
890 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
891 {
892 	struct nfs_inode *nfsi = NFS_I(dir);
893 
894 	spin_lock(&dir->i_lock);
895 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
896 	if (!cinfo->atomic || cinfo->before != dir->i_version)
897 		nfs_force_lookup_revalidate(dir);
898 	dir->i_version = cinfo->after;
899 	nfs_fscache_invalidate(dir);
900 	spin_unlock(&dir->i_lock);
901 }
902 
903 struct nfs4_opendata {
904 	struct kref kref;
905 	struct nfs_openargs o_arg;
906 	struct nfs_openres o_res;
907 	struct nfs_open_confirmargs c_arg;
908 	struct nfs_open_confirmres c_res;
909 	struct nfs4_string owner_name;
910 	struct nfs4_string group_name;
911 	struct nfs_fattr f_attr;
912 	struct nfs4_label *f_label;
913 	struct dentry *dir;
914 	struct dentry *dentry;
915 	struct nfs4_state_owner *owner;
916 	struct nfs4_state *state;
917 	struct iattr attrs;
918 	unsigned long timestamp;
919 	unsigned int rpc_done : 1;
920 	unsigned int file_created : 1;
921 	unsigned int is_recover : 1;
922 	int rpc_status;
923 	int cancelled;
924 };
925 
926 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
927 		int err, struct nfs4_exception *exception)
928 {
929 	if (err != -EINVAL)
930 		return false;
931 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
932 		return false;
933 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
934 	exception->retry = 1;
935 	return true;
936 }
937 
938 static enum open_claim_type4
939 nfs4_map_atomic_open_claim(struct nfs_server *server,
940 		enum open_claim_type4 claim)
941 {
942 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
943 		return claim;
944 	switch (claim) {
945 	default:
946 		return claim;
947 	case NFS4_OPEN_CLAIM_FH:
948 		return NFS4_OPEN_CLAIM_NULL;
949 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
950 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
951 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
952 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
953 	}
954 }
955 
956 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
957 {
958 	p->o_res.f_attr = &p->f_attr;
959 	p->o_res.f_label = p->f_label;
960 	p->o_res.seqid = p->o_arg.seqid;
961 	p->c_res.seqid = p->c_arg.seqid;
962 	p->o_res.server = p->o_arg.server;
963 	p->o_res.access_request = p->o_arg.access;
964 	nfs_fattr_init(&p->f_attr);
965 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
966 }
967 
968 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
969 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
970 		const struct iattr *attrs,
971 		struct nfs4_label *label,
972 		enum open_claim_type4 claim,
973 		gfp_t gfp_mask)
974 {
975 	struct dentry *parent = dget_parent(dentry);
976 	struct inode *dir = parent->d_inode;
977 	struct nfs_server *server = NFS_SERVER(dir);
978 	struct nfs4_opendata *p;
979 
980 	p = kzalloc(sizeof(*p), gfp_mask);
981 	if (p == NULL)
982 		goto err;
983 
984 	p->f_label = nfs4_label_alloc(server, gfp_mask);
985 	if (IS_ERR(p->f_label))
986 		goto err_free_p;
987 
988 	p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
989 	if (p->o_arg.seqid == NULL)
990 		goto err_free_label;
991 	nfs_sb_active(dentry->d_sb);
992 	p->dentry = dget(dentry);
993 	p->dir = parent;
994 	p->owner = sp;
995 	atomic_inc(&sp->so_count);
996 	p->o_arg.open_flags = flags;
997 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
998 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
999 	 * will return permission denied for all bits until close */
1000 	if (!(flags & O_EXCL)) {
1001 		/* ask server to check for all possible rights as results
1002 		 * are cached */
1003 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1004 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1005 	}
1006 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1007 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1008 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1009 	p->o_arg.name = &dentry->d_name;
1010 	p->o_arg.server = server;
1011 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1012 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1013 	p->o_arg.label = label;
1014 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1015 	switch (p->o_arg.claim) {
1016 	case NFS4_OPEN_CLAIM_NULL:
1017 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1018 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1019 		p->o_arg.fh = NFS_FH(dir);
1020 		break;
1021 	case NFS4_OPEN_CLAIM_PREVIOUS:
1022 	case NFS4_OPEN_CLAIM_FH:
1023 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1024 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1025 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1026 	}
1027 	if (attrs != NULL && attrs->ia_valid != 0) {
1028 		__u32 verf[2];
1029 
1030 		p->o_arg.u.attrs = &p->attrs;
1031 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1032 
1033 		verf[0] = jiffies;
1034 		verf[1] = current->pid;
1035 		memcpy(p->o_arg.u.verifier.data, verf,
1036 				sizeof(p->o_arg.u.verifier.data));
1037 	}
1038 	p->c_arg.fh = &p->o_res.fh;
1039 	p->c_arg.stateid = &p->o_res.stateid;
1040 	p->c_arg.seqid = p->o_arg.seqid;
1041 	nfs4_init_opendata_res(p);
1042 	kref_init(&p->kref);
1043 	return p;
1044 
1045 err_free_label:
1046 	nfs4_label_free(p->f_label);
1047 err_free_p:
1048 	kfree(p);
1049 err:
1050 	dput(parent);
1051 	return NULL;
1052 }
1053 
1054 static void nfs4_opendata_free(struct kref *kref)
1055 {
1056 	struct nfs4_opendata *p = container_of(kref,
1057 			struct nfs4_opendata, kref);
1058 	struct super_block *sb = p->dentry->d_sb;
1059 
1060 	nfs_free_seqid(p->o_arg.seqid);
1061 	if (p->state != NULL)
1062 		nfs4_put_open_state(p->state);
1063 	nfs4_put_state_owner(p->owner);
1064 
1065 	nfs4_label_free(p->f_label);
1066 
1067 	dput(p->dir);
1068 	dput(p->dentry);
1069 	nfs_sb_deactive(sb);
1070 	nfs_fattr_free_names(&p->f_attr);
1071 	kfree(p->f_attr.mdsthreshold);
1072 	kfree(p);
1073 }
1074 
1075 static void nfs4_opendata_put(struct nfs4_opendata *p)
1076 {
1077 	if (p != NULL)
1078 		kref_put(&p->kref, nfs4_opendata_free);
1079 }
1080 
1081 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1082 {
1083 	int ret;
1084 
1085 	ret = rpc_wait_for_completion_task(task);
1086 	return ret;
1087 }
1088 
1089 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1090 {
1091 	int ret = 0;
1092 
1093 	if (open_mode & (O_EXCL|O_TRUNC))
1094 		goto out;
1095 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1096 		case FMODE_READ:
1097 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1098 				&& state->n_rdonly != 0;
1099 			break;
1100 		case FMODE_WRITE:
1101 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1102 				&& state->n_wronly != 0;
1103 			break;
1104 		case FMODE_READ|FMODE_WRITE:
1105 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1106 				&& state->n_rdwr != 0;
1107 	}
1108 out:
1109 	return ret;
1110 }
1111 
1112 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1113 {
1114 	if (delegation == NULL)
1115 		return 0;
1116 	if ((delegation->type & fmode) != fmode)
1117 		return 0;
1118 	if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1119 		return 0;
1120 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1121 		return 0;
1122 	nfs_mark_delegation_referenced(delegation);
1123 	return 1;
1124 }
1125 
1126 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1127 {
1128 	switch (fmode) {
1129 		case FMODE_WRITE:
1130 			state->n_wronly++;
1131 			break;
1132 		case FMODE_READ:
1133 			state->n_rdonly++;
1134 			break;
1135 		case FMODE_READ|FMODE_WRITE:
1136 			state->n_rdwr++;
1137 	}
1138 	nfs4_state_set_mode_locked(state, state->state | fmode);
1139 }
1140 
1141 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1142 {
1143 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1144 	bool need_recover = false;
1145 
1146 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1147 		need_recover = true;
1148 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1149 		need_recover = true;
1150 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1151 		need_recover = true;
1152 	if (need_recover)
1153 		nfs4_state_mark_reclaim_nograce(clp, state);
1154 }
1155 
1156 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1157 		nfs4_stateid *stateid)
1158 {
1159 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1160 		return true;
1161 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1162 		nfs_test_and_clear_all_open_stateid(state);
1163 		return true;
1164 	}
1165 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1166 		return true;
1167 	return false;
1168 }
1169 
1170 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1171 		nfs4_stateid *stateid, fmode_t fmode)
1172 {
1173 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1174 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1175 	case FMODE_WRITE:
1176 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1177 		break;
1178 	case FMODE_READ:
1179 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1180 		break;
1181 	case 0:
1182 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1183 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1184 		clear_bit(NFS_OPEN_STATE, &state->flags);
1185 	}
1186 	if (stateid == NULL)
1187 		return;
1188 	if (!nfs_need_update_open_stateid(state, stateid))
1189 		return;
1190 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1191 		nfs4_stateid_copy(&state->stateid, stateid);
1192 	nfs4_stateid_copy(&state->open_stateid, stateid);
1193 }
1194 
1195 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1196 {
1197 	write_seqlock(&state->seqlock);
1198 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1199 	write_sequnlock(&state->seqlock);
1200 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1201 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1202 }
1203 
1204 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1205 {
1206 	switch (fmode) {
1207 		case FMODE_READ:
1208 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1209 			break;
1210 		case FMODE_WRITE:
1211 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1212 			break;
1213 		case FMODE_READ|FMODE_WRITE:
1214 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1215 	}
1216 	if (!nfs_need_update_open_stateid(state, stateid))
1217 		return;
1218 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1219 		nfs4_stateid_copy(&state->stateid, stateid);
1220 	nfs4_stateid_copy(&state->open_stateid, stateid);
1221 }
1222 
1223 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1224 {
1225 	/*
1226 	 * Protect the call to nfs4_state_set_mode_locked and
1227 	 * serialise the stateid update
1228 	 */
1229 	write_seqlock(&state->seqlock);
1230 	if (deleg_stateid != NULL) {
1231 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1232 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1233 	}
1234 	if (open_stateid != NULL)
1235 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1236 	write_sequnlock(&state->seqlock);
1237 	spin_lock(&state->owner->so_lock);
1238 	update_open_stateflags(state, fmode);
1239 	spin_unlock(&state->owner->so_lock);
1240 }
1241 
1242 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1243 {
1244 	struct nfs_inode *nfsi = NFS_I(state->inode);
1245 	struct nfs_delegation *deleg_cur;
1246 	int ret = 0;
1247 
1248 	fmode &= (FMODE_READ|FMODE_WRITE);
1249 
1250 	rcu_read_lock();
1251 	deleg_cur = rcu_dereference(nfsi->delegation);
1252 	if (deleg_cur == NULL)
1253 		goto no_delegation;
1254 
1255 	spin_lock(&deleg_cur->lock);
1256 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1257 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1258 	    (deleg_cur->type & fmode) != fmode)
1259 		goto no_delegation_unlock;
1260 
1261 	if (delegation == NULL)
1262 		delegation = &deleg_cur->stateid;
1263 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1264 		goto no_delegation_unlock;
1265 
1266 	nfs_mark_delegation_referenced(deleg_cur);
1267 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1268 	ret = 1;
1269 no_delegation_unlock:
1270 	spin_unlock(&deleg_cur->lock);
1271 no_delegation:
1272 	rcu_read_unlock();
1273 
1274 	if (!ret && open_stateid != NULL) {
1275 		__update_open_stateid(state, open_stateid, NULL, fmode);
1276 		ret = 1;
1277 	}
1278 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1279 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1280 
1281 	return ret;
1282 }
1283 
1284 
1285 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1286 {
1287 	struct nfs_delegation *delegation;
1288 
1289 	rcu_read_lock();
1290 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1291 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1292 		rcu_read_unlock();
1293 		return;
1294 	}
1295 	rcu_read_unlock();
1296 	nfs4_inode_return_delegation(inode);
1297 }
1298 
1299 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1300 {
1301 	struct nfs4_state *state = opendata->state;
1302 	struct nfs_inode *nfsi = NFS_I(state->inode);
1303 	struct nfs_delegation *delegation;
1304 	int open_mode = opendata->o_arg.open_flags;
1305 	fmode_t fmode = opendata->o_arg.fmode;
1306 	nfs4_stateid stateid;
1307 	int ret = -EAGAIN;
1308 
1309 	for (;;) {
1310 		if (can_open_cached(state, fmode, open_mode)) {
1311 			spin_lock(&state->owner->so_lock);
1312 			if (can_open_cached(state, fmode, open_mode)) {
1313 				update_open_stateflags(state, fmode);
1314 				spin_unlock(&state->owner->so_lock);
1315 				goto out_return_state;
1316 			}
1317 			spin_unlock(&state->owner->so_lock);
1318 		}
1319 		rcu_read_lock();
1320 		delegation = rcu_dereference(nfsi->delegation);
1321 		if (!can_open_delegated(delegation, fmode)) {
1322 			rcu_read_unlock();
1323 			break;
1324 		}
1325 		/* Save the delegation */
1326 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1327 		rcu_read_unlock();
1328 		nfs_release_seqid(opendata->o_arg.seqid);
1329 		if (!opendata->is_recover) {
1330 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1331 			if (ret != 0)
1332 				goto out;
1333 		}
1334 		ret = -EAGAIN;
1335 
1336 		/* Try to update the stateid using the delegation */
1337 		if (update_open_stateid(state, NULL, &stateid, fmode))
1338 			goto out_return_state;
1339 	}
1340 out:
1341 	return ERR_PTR(ret);
1342 out_return_state:
1343 	atomic_inc(&state->count);
1344 	return state;
1345 }
1346 
1347 static void
1348 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1349 {
1350 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1351 	struct nfs_delegation *delegation;
1352 	int delegation_flags = 0;
1353 
1354 	rcu_read_lock();
1355 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1356 	if (delegation)
1357 		delegation_flags = delegation->flags;
1358 	rcu_read_unlock();
1359 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1360 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1361 				   "returning a delegation for "
1362 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1363 				   clp->cl_hostname);
1364 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1365 		nfs_inode_set_delegation(state->inode,
1366 					 data->owner->so_cred,
1367 					 &data->o_res);
1368 	else
1369 		nfs_inode_reclaim_delegation(state->inode,
1370 					     data->owner->so_cred,
1371 					     &data->o_res);
1372 }
1373 
1374 /*
1375  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1376  * and update the nfs4_state.
1377  */
1378 static struct nfs4_state *
1379 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1380 {
1381 	struct inode *inode = data->state->inode;
1382 	struct nfs4_state *state = data->state;
1383 	int ret;
1384 
1385 	if (!data->rpc_done) {
1386 		if (data->rpc_status) {
1387 			ret = data->rpc_status;
1388 			goto err;
1389 		}
1390 		/* cached opens have already been processed */
1391 		goto update;
1392 	}
1393 
1394 	ret = nfs_refresh_inode(inode, &data->f_attr);
1395 	if (ret)
1396 		goto err;
1397 
1398 	if (data->o_res.delegation_type != 0)
1399 		nfs4_opendata_check_deleg(data, state);
1400 update:
1401 	update_open_stateid(state, &data->o_res.stateid, NULL,
1402 			    data->o_arg.fmode);
1403 	atomic_inc(&state->count);
1404 
1405 	return state;
1406 err:
1407 	return ERR_PTR(ret);
1408 
1409 }
1410 
1411 static struct nfs4_state *
1412 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1413 {
1414 	struct inode *inode;
1415 	struct nfs4_state *state = NULL;
1416 	int ret;
1417 
1418 	if (!data->rpc_done) {
1419 		state = nfs4_try_open_cached(data);
1420 		goto out;
1421 	}
1422 
1423 	ret = -EAGAIN;
1424 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1425 		goto err;
1426 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1427 	ret = PTR_ERR(inode);
1428 	if (IS_ERR(inode))
1429 		goto err;
1430 	ret = -ENOMEM;
1431 	state = nfs4_get_open_state(inode, data->owner);
1432 	if (state == NULL)
1433 		goto err_put_inode;
1434 	if (data->o_res.delegation_type != 0)
1435 		nfs4_opendata_check_deleg(data, state);
1436 	update_open_stateid(state, &data->o_res.stateid, NULL,
1437 			data->o_arg.fmode);
1438 	iput(inode);
1439 out:
1440 	nfs_release_seqid(data->o_arg.seqid);
1441 	return state;
1442 err_put_inode:
1443 	iput(inode);
1444 err:
1445 	return ERR_PTR(ret);
1446 }
1447 
1448 static struct nfs4_state *
1449 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1450 {
1451 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1452 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1453 	return _nfs4_opendata_to_nfs4_state(data);
1454 }
1455 
1456 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1457 {
1458 	struct nfs_inode *nfsi = NFS_I(state->inode);
1459 	struct nfs_open_context *ctx;
1460 
1461 	spin_lock(&state->inode->i_lock);
1462 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1463 		if (ctx->state != state)
1464 			continue;
1465 		get_nfs_open_context(ctx);
1466 		spin_unlock(&state->inode->i_lock);
1467 		return ctx;
1468 	}
1469 	spin_unlock(&state->inode->i_lock);
1470 	return ERR_PTR(-ENOENT);
1471 }
1472 
1473 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1474 		struct nfs4_state *state, enum open_claim_type4 claim)
1475 {
1476 	struct nfs4_opendata *opendata;
1477 
1478 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1479 			NULL, NULL, claim, GFP_NOFS);
1480 	if (opendata == NULL)
1481 		return ERR_PTR(-ENOMEM);
1482 	opendata->state = state;
1483 	atomic_inc(&state->count);
1484 	return opendata;
1485 }
1486 
1487 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1488 {
1489 	struct nfs4_state *newstate;
1490 	int ret;
1491 
1492 	opendata->o_arg.open_flags = 0;
1493 	opendata->o_arg.fmode = fmode;
1494 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1495 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1496 	nfs4_init_opendata_res(opendata);
1497 	ret = _nfs4_recover_proc_open(opendata);
1498 	if (ret != 0)
1499 		return ret;
1500 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1501 	if (IS_ERR(newstate))
1502 		return PTR_ERR(newstate);
1503 	nfs4_close_state(newstate, fmode);
1504 	*res = newstate;
1505 	return 0;
1506 }
1507 
1508 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1509 {
1510 	struct nfs4_state *newstate;
1511 	int ret;
1512 
1513 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1514 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1515 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1516 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1517 	/* memory barrier prior to reading state->n_* */
1518 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1519 	clear_bit(NFS_OPEN_STATE, &state->flags);
1520 	smp_rmb();
1521 	if (state->n_rdwr != 0) {
1522 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1523 		if (ret != 0)
1524 			return ret;
1525 		if (newstate != state)
1526 			return -ESTALE;
1527 	}
1528 	if (state->n_wronly != 0) {
1529 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1530 		if (ret != 0)
1531 			return ret;
1532 		if (newstate != state)
1533 			return -ESTALE;
1534 	}
1535 	if (state->n_rdonly != 0) {
1536 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1537 		if (ret != 0)
1538 			return ret;
1539 		if (newstate != state)
1540 			return -ESTALE;
1541 	}
1542 	/*
1543 	 * We may have performed cached opens for all three recoveries.
1544 	 * Check if we need to update the current stateid.
1545 	 */
1546 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1547 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1548 		write_seqlock(&state->seqlock);
1549 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1550 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1551 		write_sequnlock(&state->seqlock);
1552 	}
1553 	return 0;
1554 }
1555 
1556 /*
1557  * OPEN_RECLAIM:
1558  * 	reclaim state on the server after a reboot.
1559  */
1560 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1561 {
1562 	struct nfs_delegation *delegation;
1563 	struct nfs4_opendata *opendata;
1564 	fmode_t delegation_type = 0;
1565 	int status;
1566 
1567 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1568 			NFS4_OPEN_CLAIM_PREVIOUS);
1569 	if (IS_ERR(opendata))
1570 		return PTR_ERR(opendata);
1571 	rcu_read_lock();
1572 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1573 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1574 		delegation_type = delegation->type;
1575 	rcu_read_unlock();
1576 	opendata->o_arg.u.delegation_type = delegation_type;
1577 	status = nfs4_open_recover(opendata, state);
1578 	nfs4_opendata_put(opendata);
1579 	return status;
1580 }
1581 
1582 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1583 {
1584 	struct nfs_server *server = NFS_SERVER(state->inode);
1585 	struct nfs4_exception exception = { };
1586 	int err;
1587 	do {
1588 		err = _nfs4_do_open_reclaim(ctx, state);
1589 		trace_nfs4_open_reclaim(ctx, 0, err);
1590 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1591 			continue;
1592 		if (err != -NFS4ERR_DELAY)
1593 			break;
1594 		nfs4_handle_exception(server, err, &exception);
1595 	} while (exception.retry);
1596 	return err;
1597 }
1598 
1599 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1600 {
1601 	struct nfs_open_context *ctx;
1602 	int ret;
1603 
1604 	ctx = nfs4_state_find_open_context(state);
1605 	if (IS_ERR(ctx))
1606 		return -EAGAIN;
1607 	ret = nfs4_do_open_reclaim(ctx, state);
1608 	put_nfs_open_context(ctx);
1609 	return ret;
1610 }
1611 
1612 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1613 {
1614 	switch (err) {
1615 		default:
1616 			printk(KERN_ERR "NFS: %s: unhandled error "
1617 					"%d.\n", __func__, err);
1618 		case 0:
1619 		case -ENOENT:
1620 		case -ESTALE:
1621 			break;
1622 		case -NFS4ERR_BADSESSION:
1623 		case -NFS4ERR_BADSLOT:
1624 		case -NFS4ERR_BAD_HIGH_SLOT:
1625 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1626 		case -NFS4ERR_DEADSESSION:
1627 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1628 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1629 			return -EAGAIN;
1630 		case -NFS4ERR_STALE_CLIENTID:
1631 		case -NFS4ERR_STALE_STATEID:
1632 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1633 		case -NFS4ERR_EXPIRED:
1634 			/* Don't recall a delegation if it was lost */
1635 			nfs4_schedule_lease_recovery(server->nfs_client);
1636 			return -EAGAIN;
1637 		case -NFS4ERR_MOVED:
1638 			nfs4_schedule_migration_recovery(server);
1639 			return -EAGAIN;
1640 		case -NFS4ERR_LEASE_MOVED:
1641 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1642 			return -EAGAIN;
1643 		case -NFS4ERR_DELEG_REVOKED:
1644 		case -NFS4ERR_ADMIN_REVOKED:
1645 		case -NFS4ERR_BAD_STATEID:
1646 		case -NFS4ERR_OPENMODE:
1647 			nfs_inode_find_state_and_recover(state->inode,
1648 					stateid);
1649 			nfs4_schedule_stateid_recovery(server, state);
1650 			return 0;
1651 		case -NFS4ERR_DELAY:
1652 		case -NFS4ERR_GRACE:
1653 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1654 			ssleep(1);
1655 			return -EAGAIN;
1656 		case -ENOMEM:
1657 		case -NFS4ERR_DENIED:
1658 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1659 			return 0;
1660 	}
1661 	return err;
1662 }
1663 
1664 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1665 {
1666 	struct nfs_server *server = NFS_SERVER(state->inode);
1667 	struct nfs4_opendata *opendata;
1668 	int err;
1669 
1670 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1671 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1672 	if (IS_ERR(opendata))
1673 		return PTR_ERR(opendata);
1674 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1675 	err = nfs4_open_recover(opendata, state);
1676 	nfs4_opendata_put(opendata);
1677 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1678 }
1679 
1680 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1681 {
1682 	struct nfs4_opendata *data = calldata;
1683 
1684 	nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1685 				&data->c_res.seq_res, task);
1686 }
1687 
1688 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1689 {
1690 	struct nfs4_opendata *data = calldata;
1691 
1692 	nfs40_sequence_done(task, &data->c_res.seq_res);
1693 
1694 	data->rpc_status = task->tk_status;
1695 	if (data->rpc_status == 0) {
1696 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1697 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1698 		renew_lease(data->o_res.server, data->timestamp);
1699 		data->rpc_done = 1;
1700 	}
1701 }
1702 
1703 static void nfs4_open_confirm_release(void *calldata)
1704 {
1705 	struct nfs4_opendata *data = calldata;
1706 	struct nfs4_state *state = NULL;
1707 
1708 	/* If this request hasn't been cancelled, do nothing */
1709 	if (data->cancelled == 0)
1710 		goto out_free;
1711 	/* In case of error, no cleanup! */
1712 	if (!data->rpc_done)
1713 		goto out_free;
1714 	state = nfs4_opendata_to_nfs4_state(data);
1715 	if (!IS_ERR(state))
1716 		nfs4_close_state(state, data->o_arg.fmode);
1717 out_free:
1718 	nfs4_opendata_put(data);
1719 }
1720 
1721 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1722 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1723 	.rpc_call_done = nfs4_open_confirm_done,
1724 	.rpc_release = nfs4_open_confirm_release,
1725 };
1726 
1727 /*
1728  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1729  */
1730 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1731 {
1732 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1733 	struct rpc_task *task;
1734 	struct  rpc_message msg = {
1735 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1736 		.rpc_argp = &data->c_arg,
1737 		.rpc_resp = &data->c_res,
1738 		.rpc_cred = data->owner->so_cred,
1739 	};
1740 	struct rpc_task_setup task_setup_data = {
1741 		.rpc_client = server->client,
1742 		.rpc_message = &msg,
1743 		.callback_ops = &nfs4_open_confirm_ops,
1744 		.callback_data = data,
1745 		.workqueue = nfsiod_workqueue,
1746 		.flags = RPC_TASK_ASYNC,
1747 	};
1748 	int status;
1749 
1750 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1751 	kref_get(&data->kref);
1752 	data->rpc_done = 0;
1753 	data->rpc_status = 0;
1754 	data->timestamp = jiffies;
1755 	task = rpc_run_task(&task_setup_data);
1756 	if (IS_ERR(task))
1757 		return PTR_ERR(task);
1758 	status = nfs4_wait_for_completion_rpc_task(task);
1759 	if (status != 0) {
1760 		data->cancelled = 1;
1761 		smp_wmb();
1762 	} else
1763 		status = data->rpc_status;
1764 	rpc_put_task(task);
1765 	return status;
1766 }
1767 
1768 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1769 {
1770 	struct nfs4_opendata *data = calldata;
1771 	struct nfs4_state_owner *sp = data->owner;
1772 	struct nfs_client *clp = sp->so_server->nfs_client;
1773 
1774 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1775 		goto out_wait;
1776 	/*
1777 	 * Check if we still need to send an OPEN call, or if we can use
1778 	 * a delegation instead.
1779 	 */
1780 	if (data->state != NULL) {
1781 		struct nfs_delegation *delegation;
1782 
1783 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1784 			goto out_no_action;
1785 		rcu_read_lock();
1786 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1787 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1788 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1789 		    can_open_delegated(delegation, data->o_arg.fmode))
1790 			goto unlock_no_action;
1791 		rcu_read_unlock();
1792 	}
1793 	/* Update client id. */
1794 	data->o_arg.clientid = clp->cl_clientid;
1795 	switch (data->o_arg.claim) {
1796 	case NFS4_OPEN_CLAIM_PREVIOUS:
1797 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1798 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1799 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1800 	case NFS4_OPEN_CLAIM_FH:
1801 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1802 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1803 	}
1804 	data->timestamp = jiffies;
1805 	if (nfs4_setup_sequence(data->o_arg.server,
1806 				&data->o_arg.seq_args,
1807 				&data->o_res.seq_res,
1808 				task) != 0)
1809 		nfs_release_seqid(data->o_arg.seqid);
1810 
1811 	/* Set the create mode (note dependency on the session type) */
1812 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1813 	if (data->o_arg.open_flags & O_EXCL) {
1814 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1815 		if (nfs4_has_persistent_session(clp))
1816 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1817 		else if (clp->cl_mvops->minor_version > 0)
1818 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1819 	}
1820 	return;
1821 unlock_no_action:
1822 	rcu_read_unlock();
1823 out_no_action:
1824 	task->tk_action = NULL;
1825 out_wait:
1826 	nfs4_sequence_done(task, &data->o_res.seq_res);
1827 }
1828 
1829 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1830 {
1831 	struct nfs4_opendata *data = calldata;
1832 
1833 	data->rpc_status = task->tk_status;
1834 
1835 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1836 		return;
1837 
1838 	if (task->tk_status == 0) {
1839 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1840 			switch (data->o_res.f_attr->mode & S_IFMT) {
1841 			case S_IFREG:
1842 				break;
1843 			case S_IFLNK:
1844 				data->rpc_status = -ELOOP;
1845 				break;
1846 			case S_IFDIR:
1847 				data->rpc_status = -EISDIR;
1848 				break;
1849 			default:
1850 				data->rpc_status = -ENOTDIR;
1851 			}
1852 		}
1853 		renew_lease(data->o_res.server, data->timestamp);
1854 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1855 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1856 	}
1857 	data->rpc_done = 1;
1858 }
1859 
1860 static void nfs4_open_release(void *calldata)
1861 {
1862 	struct nfs4_opendata *data = calldata;
1863 	struct nfs4_state *state = NULL;
1864 
1865 	/* If this request hasn't been cancelled, do nothing */
1866 	if (data->cancelled == 0)
1867 		goto out_free;
1868 	/* In case of error, no cleanup! */
1869 	if (data->rpc_status != 0 || !data->rpc_done)
1870 		goto out_free;
1871 	/* In case we need an open_confirm, no cleanup! */
1872 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1873 		goto out_free;
1874 	state = nfs4_opendata_to_nfs4_state(data);
1875 	if (!IS_ERR(state))
1876 		nfs4_close_state(state, data->o_arg.fmode);
1877 out_free:
1878 	nfs4_opendata_put(data);
1879 }
1880 
1881 static const struct rpc_call_ops nfs4_open_ops = {
1882 	.rpc_call_prepare = nfs4_open_prepare,
1883 	.rpc_call_done = nfs4_open_done,
1884 	.rpc_release = nfs4_open_release,
1885 };
1886 
1887 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1888 {
1889 	struct inode *dir = data->dir->d_inode;
1890 	struct nfs_server *server = NFS_SERVER(dir);
1891 	struct nfs_openargs *o_arg = &data->o_arg;
1892 	struct nfs_openres *o_res = &data->o_res;
1893 	struct rpc_task *task;
1894 	struct rpc_message msg = {
1895 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1896 		.rpc_argp = o_arg,
1897 		.rpc_resp = o_res,
1898 		.rpc_cred = data->owner->so_cred,
1899 	};
1900 	struct rpc_task_setup task_setup_data = {
1901 		.rpc_client = server->client,
1902 		.rpc_message = &msg,
1903 		.callback_ops = &nfs4_open_ops,
1904 		.callback_data = data,
1905 		.workqueue = nfsiod_workqueue,
1906 		.flags = RPC_TASK_ASYNC,
1907 	};
1908 	int status;
1909 
1910 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1911 	kref_get(&data->kref);
1912 	data->rpc_done = 0;
1913 	data->rpc_status = 0;
1914 	data->cancelled = 0;
1915 	data->is_recover = 0;
1916 	if (isrecover) {
1917 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1918 		data->is_recover = 1;
1919 	}
1920 	task = rpc_run_task(&task_setup_data);
1921         if (IS_ERR(task))
1922                 return PTR_ERR(task);
1923         status = nfs4_wait_for_completion_rpc_task(task);
1924         if (status != 0) {
1925                 data->cancelled = 1;
1926                 smp_wmb();
1927         } else
1928                 status = data->rpc_status;
1929         rpc_put_task(task);
1930 
1931 	return status;
1932 }
1933 
1934 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1935 {
1936 	struct inode *dir = data->dir->d_inode;
1937 	struct nfs_openres *o_res = &data->o_res;
1938         int status;
1939 
1940 	status = nfs4_run_open_task(data, 1);
1941 	if (status != 0 || !data->rpc_done)
1942 		return status;
1943 
1944 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1945 
1946 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1947 		status = _nfs4_proc_open_confirm(data);
1948 		if (status != 0)
1949 			return status;
1950 	}
1951 
1952 	return status;
1953 }
1954 
1955 /*
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 #if defined(CONFIG_NFS_V4_1)
2106 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
2107 {
2108 	struct nfs_server *server = NFS_SERVER(state->inode);
2109 	nfs4_stateid *stateid = &state->stateid;
2110 	struct nfs_delegation *delegation;
2111 	struct rpc_cred *cred = NULL;
2112 	int status = -NFS4ERR_BAD_STATEID;
2113 
2114 	/* If a state reset has been done, test_stateid is unneeded */
2115 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2116 		return;
2117 
2118 	/* Get the delegation credential for use by test/free_stateid */
2119 	rcu_read_lock();
2120 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2121 	if (delegation != NULL &&
2122 	    nfs4_stateid_match(&delegation->stateid, stateid)) {
2123 		cred = get_rpccred(delegation->cred);
2124 		rcu_read_unlock();
2125 		status = nfs41_test_stateid(server, stateid, cred);
2126 		trace_nfs4_test_delegation_stateid(state, NULL, status);
2127 	} else
2128 		rcu_read_unlock();
2129 
2130 	if (status != NFS_OK) {
2131 		/* Free the stateid unless the server explicitly
2132 		 * informs us the stateid is unrecognized. */
2133 		if (status != -NFS4ERR_BAD_STATEID)
2134 			nfs41_free_stateid(server, stateid, cred);
2135 		nfs_remove_bad_delegation(state->inode);
2136 
2137 		write_seqlock(&state->seqlock);
2138 		nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2139 		write_sequnlock(&state->seqlock);
2140 		clear_bit(NFS_DELEGATED_STATE, &state->flags);
2141 	}
2142 
2143 	if (cred != NULL)
2144 		put_rpccred(cred);
2145 }
2146 
2147 /**
2148  * nfs41_check_open_stateid - possibly free an open stateid
2149  *
2150  * @state: NFSv4 state for an inode
2151  *
2152  * Returns NFS_OK if recovery for this stateid is now finished.
2153  * Otherwise a negative NFS4ERR value is returned.
2154  */
2155 static int nfs41_check_open_stateid(struct nfs4_state *state)
2156 {
2157 	struct nfs_server *server = NFS_SERVER(state->inode);
2158 	nfs4_stateid *stateid = &state->open_stateid;
2159 	struct rpc_cred *cred = state->owner->so_cred;
2160 	int status;
2161 
2162 	/* If a state reset has been done, test_stateid is unneeded */
2163 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2164 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2165 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2166 		return -NFS4ERR_BAD_STATEID;
2167 
2168 	status = nfs41_test_stateid(server, stateid, cred);
2169 	trace_nfs4_test_open_stateid(state, NULL, status);
2170 	if (status != NFS_OK) {
2171 		/* Free the stateid unless the server explicitly
2172 		 * informs us the stateid is unrecognized. */
2173 		if (status != -NFS4ERR_BAD_STATEID)
2174 			nfs41_free_stateid(server, stateid, cred);
2175 
2176 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2177 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2178 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2179 		clear_bit(NFS_OPEN_STATE, &state->flags);
2180 	}
2181 	return status;
2182 }
2183 
2184 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2185 {
2186 	int status;
2187 
2188 	nfs41_clear_delegation_stateid(state);
2189 	status = nfs41_check_open_stateid(state);
2190 	if (status != NFS_OK)
2191 		status = nfs4_open_expired(sp, state);
2192 	return status;
2193 }
2194 #endif
2195 
2196 /*
2197  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2198  * fields corresponding to attributes that were used to store the verifier.
2199  * Make sure we clobber those fields in the later setattr call
2200  */
2201 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2202 {
2203 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2204 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2205 		sattr->ia_valid |= ATTR_ATIME;
2206 
2207 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2208 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2209 		sattr->ia_valid |= ATTR_MTIME;
2210 }
2211 
2212 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2213 		fmode_t fmode,
2214 		int flags,
2215 		struct nfs_open_context *ctx)
2216 {
2217 	struct nfs4_state_owner *sp = opendata->owner;
2218 	struct nfs_server *server = sp->so_server;
2219 	struct dentry *dentry;
2220 	struct nfs4_state *state;
2221 	unsigned int seq;
2222 	int ret;
2223 
2224 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2225 
2226 	ret = _nfs4_proc_open(opendata);
2227 	if (ret != 0) {
2228 		if (ret == -ENOENT) {
2229 			d_drop(opendata->dentry);
2230 			d_add(opendata->dentry, NULL);
2231 			nfs_set_verifier(opendata->dentry,
2232 					 nfs_save_change_attribute(opendata->dir->d_inode));
2233 		}
2234 		goto out;
2235 	}
2236 
2237 	state = nfs4_opendata_to_nfs4_state(opendata);
2238 	ret = PTR_ERR(state);
2239 	if (IS_ERR(state))
2240 		goto out;
2241 	if (server->caps & NFS_CAP_POSIX_LOCK)
2242 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2243 
2244 	dentry = opendata->dentry;
2245 	if (dentry->d_inode == NULL) {
2246 		/* FIXME: Is this d_drop() ever needed? */
2247 		d_drop(dentry);
2248 		dentry = d_add_unique(dentry, igrab(state->inode));
2249 		if (dentry == NULL) {
2250 			dentry = opendata->dentry;
2251 		} else if (dentry != ctx->dentry) {
2252 			dput(ctx->dentry);
2253 			ctx->dentry = dget(dentry);
2254 		}
2255 		nfs_set_verifier(dentry,
2256 				nfs_save_change_attribute(opendata->dir->d_inode));
2257 	}
2258 
2259 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2260 	if (ret != 0)
2261 		goto out;
2262 
2263 	ctx->state = state;
2264 	if (dentry->d_inode == state->inode) {
2265 		nfs_inode_attach_open_context(ctx);
2266 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2267 			nfs4_schedule_stateid_recovery(server, state);
2268 	}
2269 out:
2270 	return ret;
2271 }
2272 
2273 /*
2274  * Returns a referenced nfs4_state
2275  */
2276 static int _nfs4_do_open(struct inode *dir,
2277 			struct nfs_open_context *ctx,
2278 			int flags,
2279 			struct iattr *sattr,
2280 			struct nfs4_label *label,
2281 			int *opened)
2282 {
2283 	struct nfs4_state_owner  *sp;
2284 	struct nfs4_state     *state = NULL;
2285 	struct nfs_server       *server = NFS_SERVER(dir);
2286 	struct nfs4_opendata *opendata;
2287 	struct dentry *dentry = ctx->dentry;
2288 	struct rpc_cred *cred = ctx->cred;
2289 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2290 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2291 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2292 	struct nfs4_label *olabel = NULL;
2293 	int status;
2294 
2295 	/* Protect against reboot recovery conflicts */
2296 	status = -ENOMEM;
2297 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2298 	if (sp == NULL) {
2299 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2300 		goto out_err;
2301 	}
2302 	status = nfs4_recover_expired_lease(server);
2303 	if (status != 0)
2304 		goto err_put_state_owner;
2305 	if (dentry->d_inode != NULL)
2306 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2307 	status = -ENOMEM;
2308 	if (dentry->d_inode)
2309 		claim = NFS4_OPEN_CLAIM_FH;
2310 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2311 			label, claim, GFP_KERNEL);
2312 	if (opendata == NULL)
2313 		goto err_put_state_owner;
2314 
2315 	if (label) {
2316 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2317 		if (IS_ERR(olabel)) {
2318 			status = PTR_ERR(olabel);
2319 			goto err_opendata_put;
2320 		}
2321 	}
2322 
2323 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2324 		if (!opendata->f_attr.mdsthreshold) {
2325 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2326 			if (!opendata->f_attr.mdsthreshold)
2327 				goto err_free_label;
2328 		}
2329 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2330 	}
2331 	if (dentry->d_inode != NULL)
2332 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2333 
2334 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2335 	if (status != 0)
2336 		goto err_free_label;
2337 	state = ctx->state;
2338 
2339 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2340 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2341 		nfs4_exclusive_attrset(opendata, sattr);
2342 
2343 		nfs_fattr_init(opendata->o_res.f_attr);
2344 		status = nfs4_do_setattr(state->inode, cred,
2345 				opendata->o_res.f_attr, sattr,
2346 				state, label, olabel);
2347 		if (status == 0) {
2348 			nfs_setattr_update_inode(state->inode, sattr);
2349 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2350 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2351 		}
2352 	}
2353 	if (opendata->file_created)
2354 		*opened |= FILE_CREATED;
2355 
2356 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2357 		*ctx_th = opendata->f_attr.mdsthreshold;
2358 		opendata->f_attr.mdsthreshold = NULL;
2359 	}
2360 
2361 	nfs4_label_free(olabel);
2362 
2363 	nfs4_opendata_put(opendata);
2364 	nfs4_put_state_owner(sp);
2365 	return 0;
2366 err_free_label:
2367 	nfs4_label_free(olabel);
2368 err_opendata_put:
2369 	nfs4_opendata_put(opendata);
2370 err_put_state_owner:
2371 	nfs4_put_state_owner(sp);
2372 out_err:
2373 	return status;
2374 }
2375 
2376 
2377 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2378 					struct nfs_open_context *ctx,
2379 					int flags,
2380 					struct iattr *sattr,
2381 					struct nfs4_label *label,
2382 					int *opened)
2383 {
2384 	struct nfs_server *server = NFS_SERVER(dir);
2385 	struct nfs4_exception exception = { };
2386 	struct nfs4_state *res;
2387 	int status;
2388 
2389 	do {
2390 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2391 		res = ctx->state;
2392 		trace_nfs4_open_file(ctx, flags, status);
2393 		if (status == 0)
2394 			break;
2395 		/* NOTE: BAD_SEQID means the server and client disagree about the
2396 		 * book-keeping w.r.t. state-changing operations
2397 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2398 		 * It is actually a sign of a bug on the client or on the server.
2399 		 *
2400 		 * If we receive a BAD_SEQID error in the particular case of
2401 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2402 		 * have unhashed the old state_owner for us, and that we can
2403 		 * therefore safely retry using a new one. We should still warn
2404 		 * the user though...
2405 		 */
2406 		if (status == -NFS4ERR_BAD_SEQID) {
2407 			pr_warn_ratelimited("NFS: v4 server %s "
2408 					" returned a bad sequence-id error!\n",
2409 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2410 			exception.retry = 1;
2411 			continue;
2412 		}
2413 		/*
2414 		 * BAD_STATEID on OPEN means that the server cancelled our
2415 		 * state before it received the OPEN_CONFIRM.
2416 		 * Recover by retrying the request as per the discussion
2417 		 * on Page 181 of RFC3530.
2418 		 */
2419 		if (status == -NFS4ERR_BAD_STATEID) {
2420 			exception.retry = 1;
2421 			continue;
2422 		}
2423 		if (status == -EAGAIN) {
2424 			/* We must have found a delegation */
2425 			exception.retry = 1;
2426 			continue;
2427 		}
2428 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2429 			continue;
2430 		res = ERR_PTR(nfs4_handle_exception(server,
2431 					status, &exception));
2432 	} while (exception.retry);
2433 	return res;
2434 }
2435 
2436 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2437 			    struct nfs_fattr *fattr, struct iattr *sattr,
2438 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2439 			    struct nfs4_label *olabel)
2440 {
2441 	struct nfs_server *server = NFS_SERVER(inode);
2442         struct nfs_setattrargs  arg = {
2443                 .fh             = NFS_FH(inode),
2444                 .iap            = sattr,
2445 		.server		= server,
2446 		.bitmask = server->attr_bitmask,
2447 		.label		= ilabel,
2448         };
2449         struct nfs_setattrres  res = {
2450 		.fattr		= fattr,
2451 		.label		= olabel,
2452 		.server		= server,
2453         };
2454         struct rpc_message msg = {
2455 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2456 		.rpc_argp	= &arg,
2457 		.rpc_resp	= &res,
2458 		.rpc_cred	= cred,
2459         };
2460 	unsigned long timestamp = jiffies;
2461 	fmode_t fmode;
2462 	bool truncate;
2463 	int status;
2464 
2465 	arg.bitmask = nfs4_bitmask(server, ilabel);
2466 	if (ilabel)
2467 		arg.bitmask = nfs4_bitmask(server, olabel);
2468 
2469 	nfs_fattr_init(fattr);
2470 
2471 	/* Servers should only apply open mode checks for file size changes */
2472 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2473 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2474 
2475 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2476 		/* Use that stateid */
2477 	} else if (truncate && state != NULL) {
2478 		struct nfs_lockowner lockowner = {
2479 			.l_owner = current->files,
2480 			.l_pid = current->tgid,
2481 		};
2482 		if (!nfs4_valid_open_stateid(state))
2483 			return -EBADF;
2484 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2485 				&lockowner) == -EIO)
2486 			return -EBADF;
2487 	} else
2488 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2489 
2490 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2491 	if (status == 0 && state != NULL)
2492 		renew_lease(server, timestamp);
2493 	return status;
2494 }
2495 
2496 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2497 			   struct nfs_fattr *fattr, struct iattr *sattr,
2498 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2499 			   struct nfs4_label *olabel)
2500 {
2501 	struct nfs_server *server = NFS_SERVER(inode);
2502 	struct nfs4_exception exception = {
2503 		.state = state,
2504 		.inode = inode,
2505 	};
2506 	int err;
2507 	do {
2508 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2509 		trace_nfs4_setattr(inode, err);
2510 		switch (err) {
2511 		case -NFS4ERR_OPENMODE:
2512 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2513 				pr_warn_once("NFSv4: server %s is incorrectly "
2514 						"applying open mode checks to "
2515 						"a SETATTR that is not "
2516 						"changing file size.\n",
2517 						server->nfs_client->cl_hostname);
2518 			}
2519 			if (state && !(state->state & FMODE_WRITE)) {
2520 				err = -EBADF;
2521 				if (sattr->ia_valid & ATTR_OPEN)
2522 					err = -EACCES;
2523 				goto out;
2524 			}
2525 		}
2526 		err = nfs4_handle_exception(server, err, &exception);
2527 	} while (exception.retry);
2528 out:
2529 	return err;
2530 }
2531 
2532 struct nfs4_closedata {
2533 	struct inode *inode;
2534 	struct nfs4_state *state;
2535 	struct nfs_closeargs arg;
2536 	struct nfs_closeres res;
2537 	struct nfs_fattr fattr;
2538 	unsigned long timestamp;
2539 	bool roc;
2540 	u32 roc_barrier;
2541 };
2542 
2543 static void nfs4_free_closedata(void *data)
2544 {
2545 	struct nfs4_closedata *calldata = data;
2546 	struct nfs4_state_owner *sp = calldata->state->owner;
2547 	struct super_block *sb = calldata->state->inode->i_sb;
2548 
2549 	if (calldata->roc)
2550 		pnfs_roc_release(calldata->state->inode);
2551 	nfs4_put_open_state(calldata->state);
2552 	nfs_free_seqid(calldata->arg.seqid);
2553 	nfs4_put_state_owner(sp);
2554 	nfs_sb_deactive(sb);
2555 	kfree(calldata);
2556 }
2557 
2558 static void nfs4_close_done(struct rpc_task *task, void *data)
2559 {
2560 	struct nfs4_closedata *calldata = data;
2561 	struct nfs4_state *state = calldata->state;
2562 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2563 
2564 	dprintk("%s: begin!\n", __func__);
2565 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2566 		return;
2567 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2568         /* hmm. we are done with the inode, and in the process of freeing
2569 	 * the state_owner. we keep this around to process errors
2570 	 */
2571 	switch (task->tk_status) {
2572 		case 0:
2573 			if (calldata->roc)
2574 				pnfs_roc_set_barrier(state->inode,
2575 						     calldata->roc_barrier);
2576 			nfs_clear_open_stateid(state, &calldata->res.stateid, 0);
2577 			renew_lease(server, calldata->timestamp);
2578 			goto out_release;
2579 		case -NFS4ERR_ADMIN_REVOKED:
2580 		case -NFS4ERR_STALE_STATEID:
2581 		case -NFS4ERR_OLD_STATEID:
2582 		case -NFS4ERR_BAD_STATEID:
2583 		case -NFS4ERR_EXPIRED:
2584 			if (calldata->arg.fmode == 0)
2585 				break;
2586 		default:
2587 			if (nfs4_async_handle_error(task, server, state) == -EAGAIN) {
2588 				rpc_restart_call_prepare(task);
2589 				goto out_release;
2590 			}
2591 	}
2592 	nfs_clear_open_stateid(state, NULL, calldata->arg.fmode);
2593 out_release:
2594 	nfs_release_seqid(calldata->arg.seqid);
2595 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2596 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2597 }
2598 
2599 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2600 {
2601 	struct nfs4_closedata *calldata = data;
2602 	struct nfs4_state *state = calldata->state;
2603 	struct inode *inode = calldata->inode;
2604 	int call_close = 0;
2605 
2606 	dprintk("%s: begin!\n", __func__);
2607 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2608 		goto out_wait;
2609 
2610 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2611 	calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2612 	spin_lock(&state->owner->so_lock);
2613 	/* Calculate the change in open mode */
2614 	if (state->n_rdwr == 0) {
2615 		if (state->n_rdonly == 0) {
2616 			call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2617 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2618 			calldata->arg.fmode &= ~FMODE_READ;
2619 		}
2620 		if (state->n_wronly == 0) {
2621 			call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2622 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2623 			calldata->arg.fmode &= ~FMODE_WRITE;
2624 		}
2625 	}
2626 	if (!nfs4_valid_open_stateid(state))
2627 		call_close = 0;
2628 	spin_unlock(&state->owner->so_lock);
2629 
2630 	if (!call_close) {
2631 		/* Note: exit _without_ calling nfs4_close_done */
2632 		goto out_no_action;
2633 	}
2634 
2635 	if (calldata->arg.fmode == 0) {
2636 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2637 		if (calldata->roc &&
2638 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2639 			nfs_release_seqid(calldata->arg.seqid);
2640 			goto out_wait;
2641 		    }
2642 	}
2643 
2644 	nfs_fattr_init(calldata->res.fattr);
2645 	calldata->timestamp = jiffies;
2646 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2647 				&calldata->arg.seq_args,
2648 				&calldata->res.seq_res,
2649 				task) != 0)
2650 		nfs_release_seqid(calldata->arg.seqid);
2651 	dprintk("%s: done!\n", __func__);
2652 	return;
2653 out_no_action:
2654 	task->tk_action = NULL;
2655 out_wait:
2656 	nfs4_sequence_done(task, &calldata->res.seq_res);
2657 }
2658 
2659 static const struct rpc_call_ops nfs4_close_ops = {
2660 	.rpc_call_prepare = nfs4_close_prepare,
2661 	.rpc_call_done = nfs4_close_done,
2662 	.rpc_release = nfs4_free_closedata,
2663 };
2664 
2665 static bool nfs4_state_has_opener(struct nfs4_state *state)
2666 {
2667 	/* first check existing openers */
2668 	if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0 &&
2669 	    state->n_rdonly != 0)
2670 		return true;
2671 
2672 	if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0 &&
2673 	    state->n_wronly != 0)
2674 		return true;
2675 
2676 	if (test_bit(NFS_O_RDWR_STATE, &state->flags) != 0 &&
2677 	    state->n_rdwr != 0)
2678 		return true;
2679 
2680 	return false;
2681 }
2682 
2683 static bool nfs4_roc(struct inode *inode)
2684 {
2685 	struct nfs_inode *nfsi = NFS_I(inode);
2686 	struct nfs_open_context *ctx;
2687 	struct nfs4_state *state;
2688 
2689 	spin_lock(&inode->i_lock);
2690 	list_for_each_entry(ctx, &nfsi->open_files, list) {
2691 		state = ctx->state;
2692 		if (state == NULL)
2693 			continue;
2694 		if (nfs4_state_has_opener(state)) {
2695 			spin_unlock(&inode->i_lock);
2696 			return false;
2697 		}
2698 	}
2699 	spin_unlock(&inode->i_lock);
2700 
2701 	if (nfs4_check_delegation(inode, FMODE_READ))
2702 		return false;
2703 
2704 	return pnfs_roc(inode);
2705 }
2706 
2707 /*
2708  * It is possible for data to be read/written from a mem-mapped file
2709  * after the sys_close call (which hits the vfs layer as a flush).
2710  * This means that we can't safely call nfsv4 close on a file until
2711  * the inode is cleared. This in turn means that we are not good
2712  * NFSv4 citizens - we do not indicate to the server to update the file's
2713  * share state even when we are done with one of the three share
2714  * stateid's in the inode.
2715  *
2716  * NOTE: Caller must be holding the sp->so_owner semaphore!
2717  */
2718 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2719 {
2720 	struct nfs_server *server = NFS_SERVER(state->inode);
2721 	struct nfs4_closedata *calldata;
2722 	struct nfs4_state_owner *sp = state->owner;
2723 	struct rpc_task *task;
2724 	struct rpc_message msg = {
2725 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2726 		.rpc_cred = state->owner->so_cred,
2727 	};
2728 	struct rpc_task_setup task_setup_data = {
2729 		.rpc_client = server->client,
2730 		.rpc_message = &msg,
2731 		.callback_ops = &nfs4_close_ops,
2732 		.workqueue = nfsiod_workqueue,
2733 		.flags = RPC_TASK_ASYNC,
2734 	};
2735 	int status = -ENOMEM;
2736 
2737 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2738 		&task_setup_data.rpc_client, &msg);
2739 
2740 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2741 	if (calldata == NULL)
2742 		goto out;
2743 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2744 	calldata->inode = state->inode;
2745 	calldata->state = state;
2746 	calldata->arg.fh = NFS_FH(state->inode);
2747 	calldata->arg.stateid = &state->open_stateid;
2748 	/* Serialization for the sequence id */
2749 	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2750 	if (calldata->arg.seqid == NULL)
2751 		goto out_free_calldata;
2752 	calldata->arg.fmode = 0;
2753 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2754 	calldata->res.fattr = &calldata->fattr;
2755 	calldata->res.seqid = calldata->arg.seqid;
2756 	calldata->res.server = server;
2757 	calldata->roc = nfs4_roc(state->inode);
2758 	nfs_sb_active(calldata->inode->i_sb);
2759 
2760 	msg.rpc_argp = &calldata->arg;
2761 	msg.rpc_resp = &calldata->res;
2762 	task_setup_data.callback_data = calldata;
2763 	task = rpc_run_task(&task_setup_data);
2764 	if (IS_ERR(task))
2765 		return PTR_ERR(task);
2766 	status = 0;
2767 	if (wait)
2768 		status = rpc_wait_for_completion_task(task);
2769 	rpc_put_task(task);
2770 	return status;
2771 out_free_calldata:
2772 	kfree(calldata);
2773 out:
2774 	nfs4_put_open_state(state);
2775 	nfs4_put_state_owner(sp);
2776 	return status;
2777 }
2778 
2779 static struct inode *
2780 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2781 		int open_flags, struct iattr *attr, int *opened)
2782 {
2783 	struct nfs4_state *state;
2784 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2785 
2786 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2787 
2788 	/* Protect against concurrent sillydeletes */
2789 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2790 
2791 	nfs4_label_release_security(label);
2792 
2793 	if (IS_ERR(state))
2794 		return ERR_CAST(state);
2795 	return state->inode;
2796 }
2797 
2798 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2799 {
2800 	if (ctx->state == NULL)
2801 		return;
2802 	if (is_sync)
2803 		nfs4_close_sync(ctx->state, ctx->mode);
2804 	else
2805 		nfs4_close_state(ctx->state, ctx->mode);
2806 }
2807 
2808 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2809 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2810 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2811 
2812 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2813 {
2814 	struct nfs4_server_caps_arg args = {
2815 		.fhandle = fhandle,
2816 	};
2817 	struct nfs4_server_caps_res res = {};
2818 	struct rpc_message msg = {
2819 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2820 		.rpc_argp = &args,
2821 		.rpc_resp = &res,
2822 	};
2823 	int status;
2824 
2825 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2826 	if (status == 0) {
2827 		/* Sanity check the server answers */
2828 		switch (server->nfs_client->cl_minorversion) {
2829 		case 0:
2830 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2831 			res.attr_bitmask[2] = 0;
2832 			break;
2833 		case 1:
2834 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2835 			break;
2836 		case 2:
2837 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2838 		}
2839 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2840 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2841 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2842 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2843 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2844 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2845 				NFS_CAP_SECURITY_LABEL);
2846 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2847 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2848 			server->caps |= NFS_CAP_ACLS;
2849 		if (res.has_links != 0)
2850 			server->caps |= NFS_CAP_HARDLINKS;
2851 		if (res.has_symlinks != 0)
2852 			server->caps |= NFS_CAP_SYMLINKS;
2853 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2854 			server->caps |= NFS_CAP_FILEID;
2855 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2856 			server->caps |= NFS_CAP_MODE;
2857 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2858 			server->caps |= NFS_CAP_NLINK;
2859 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2860 			server->caps |= NFS_CAP_OWNER;
2861 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2862 			server->caps |= NFS_CAP_OWNER_GROUP;
2863 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2864 			server->caps |= NFS_CAP_ATIME;
2865 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2866 			server->caps |= NFS_CAP_CTIME;
2867 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2868 			server->caps |= NFS_CAP_MTIME;
2869 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2870 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2871 			server->caps |= NFS_CAP_SECURITY_LABEL;
2872 #endif
2873 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2874 				sizeof(server->attr_bitmask));
2875 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2876 
2877 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2878 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2879 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2880 		server->cache_consistency_bitmask[2] = 0;
2881 		server->acl_bitmask = res.acl_bitmask;
2882 		server->fh_expire_type = res.fh_expire_type;
2883 	}
2884 
2885 	return status;
2886 }
2887 
2888 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2889 {
2890 	struct nfs4_exception exception = { };
2891 	int err;
2892 	do {
2893 		err = nfs4_handle_exception(server,
2894 				_nfs4_server_capabilities(server, fhandle),
2895 				&exception);
2896 	} while (exception.retry);
2897 	return err;
2898 }
2899 
2900 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2901 		struct nfs_fsinfo *info)
2902 {
2903 	u32 bitmask[3];
2904 	struct nfs4_lookup_root_arg args = {
2905 		.bitmask = bitmask,
2906 	};
2907 	struct nfs4_lookup_res res = {
2908 		.server = server,
2909 		.fattr = info->fattr,
2910 		.fh = fhandle,
2911 	};
2912 	struct rpc_message msg = {
2913 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2914 		.rpc_argp = &args,
2915 		.rpc_resp = &res,
2916 	};
2917 
2918 	bitmask[0] = nfs4_fattr_bitmap[0];
2919 	bitmask[1] = nfs4_fattr_bitmap[1];
2920 	/*
2921 	 * Process the label in the upcoming getfattr
2922 	 */
2923 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2924 
2925 	nfs_fattr_init(info->fattr);
2926 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2927 }
2928 
2929 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2930 		struct nfs_fsinfo *info)
2931 {
2932 	struct nfs4_exception exception = { };
2933 	int err;
2934 	do {
2935 		err = _nfs4_lookup_root(server, fhandle, info);
2936 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2937 		switch (err) {
2938 		case 0:
2939 		case -NFS4ERR_WRONGSEC:
2940 			goto out;
2941 		default:
2942 			err = nfs4_handle_exception(server, err, &exception);
2943 		}
2944 	} while (exception.retry);
2945 out:
2946 	return err;
2947 }
2948 
2949 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2950 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2951 {
2952 	struct rpc_auth_create_args auth_args = {
2953 		.pseudoflavor = flavor,
2954 	};
2955 	struct rpc_auth *auth;
2956 	int ret;
2957 
2958 	auth = rpcauth_create(&auth_args, server->client);
2959 	if (IS_ERR(auth)) {
2960 		ret = -EACCES;
2961 		goto out;
2962 	}
2963 	ret = nfs4_lookup_root(server, fhandle, info);
2964 out:
2965 	return ret;
2966 }
2967 
2968 /*
2969  * Retry pseudoroot lookup with various security flavors.  We do this when:
2970  *
2971  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2972  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2973  *
2974  * Returns zero on success, or a negative NFS4ERR value, or a
2975  * negative errno value.
2976  */
2977 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2978 			      struct nfs_fsinfo *info)
2979 {
2980 	/* Per 3530bis 15.33.5 */
2981 	static const rpc_authflavor_t flav_array[] = {
2982 		RPC_AUTH_GSS_KRB5P,
2983 		RPC_AUTH_GSS_KRB5I,
2984 		RPC_AUTH_GSS_KRB5,
2985 		RPC_AUTH_UNIX,			/* courtesy */
2986 		RPC_AUTH_NULL,
2987 	};
2988 	int status = -EPERM;
2989 	size_t i;
2990 
2991 	if (server->auth_info.flavor_len > 0) {
2992 		/* try each flavor specified by user */
2993 		for (i = 0; i < server->auth_info.flavor_len; i++) {
2994 			status = nfs4_lookup_root_sec(server, fhandle, info,
2995 						server->auth_info.flavors[i]);
2996 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2997 				continue;
2998 			break;
2999 		}
3000 	} else {
3001 		/* no flavors specified by user, try default list */
3002 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3003 			status = nfs4_lookup_root_sec(server, fhandle, info,
3004 						      flav_array[i]);
3005 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3006 				continue;
3007 			break;
3008 		}
3009 	}
3010 
3011 	/*
3012 	 * -EACCESS could mean that the user doesn't have correct permissions
3013 	 * to access the mount.  It could also mean that we tried to mount
3014 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3015 	 * existing mount programs don't handle -EACCES very well so it should
3016 	 * be mapped to -EPERM instead.
3017 	 */
3018 	if (status == -EACCES)
3019 		status = -EPERM;
3020 	return status;
3021 }
3022 
3023 static int nfs4_do_find_root_sec(struct nfs_server *server,
3024 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3025 {
3026 	int mv = server->nfs_client->cl_minorversion;
3027 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3028 }
3029 
3030 /**
3031  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3032  * @server: initialized nfs_server handle
3033  * @fhandle: we fill in the pseudo-fs root file handle
3034  * @info: we fill in an FSINFO struct
3035  * @auth_probe: probe the auth flavours
3036  *
3037  * Returns zero on success, or a negative errno.
3038  */
3039 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3040 			 struct nfs_fsinfo *info,
3041 			 bool auth_probe)
3042 {
3043 	int status;
3044 
3045 	switch (auth_probe) {
3046 	case false:
3047 		status = nfs4_lookup_root(server, fhandle, info);
3048 		if (status != -NFS4ERR_WRONGSEC)
3049 			break;
3050 	default:
3051 		status = nfs4_do_find_root_sec(server, fhandle, info);
3052 	}
3053 
3054 	if (status == 0)
3055 		status = nfs4_server_capabilities(server, fhandle);
3056 	if (status == 0)
3057 		status = nfs4_do_fsinfo(server, fhandle, info);
3058 
3059 	return nfs4_map_errors(status);
3060 }
3061 
3062 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3063 			      struct nfs_fsinfo *info)
3064 {
3065 	int error;
3066 	struct nfs_fattr *fattr = info->fattr;
3067 	struct nfs4_label *label = NULL;
3068 
3069 	error = nfs4_server_capabilities(server, mntfh);
3070 	if (error < 0) {
3071 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3072 		return error;
3073 	}
3074 
3075 	label = nfs4_label_alloc(server, GFP_KERNEL);
3076 	if (IS_ERR(label))
3077 		return PTR_ERR(label);
3078 
3079 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3080 	if (error < 0) {
3081 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3082 		goto err_free_label;
3083 	}
3084 
3085 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3086 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3087 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3088 
3089 err_free_label:
3090 	nfs4_label_free(label);
3091 
3092 	return error;
3093 }
3094 
3095 /*
3096  * Get locations and (maybe) other attributes of a referral.
3097  * Note that we'll actually follow the referral later when
3098  * we detect fsid mismatch in inode revalidation
3099  */
3100 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3101 			     const struct qstr *name, struct nfs_fattr *fattr,
3102 			     struct nfs_fh *fhandle)
3103 {
3104 	int status = -ENOMEM;
3105 	struct page *page = NULL;
3106 	struct nfs4_fs_locations *locations = NULL;
3107 
3108 	page = alloc_page(GFP_KERNEL);
3109 	if (page == NULL)
3110 		goto out;
3111 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3112 	if (locations == NULL)
3113 		goto out;
3114 
3115 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3116 	if (status != 0)
3117 		goto out;
3118 
3119 	/*
3120 	 * If the fsid didn't change, this is a migration event, not a
3121 	 * referral.  Cause us to drop into the exception handler, which
3122 	 * will kick off migration recovery.
3123 	 */
3124 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3125 		dprintk("%s: server did not return a different fsid for"
3126 			" a referral at %s\n", __func__, name->name);
3127 		status = -NFS4ERR_MOVED;
3128 		goto out;
3129 	}
3130 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3131 	nfs_fixup_referral_attributes(&locations->fattr);
3132 
3133 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3134 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3135 	memset(fhandle, 0, sizeof(struct nfs_fh));
3136 out:
3137 	if (page)
3138 		__free_page(page);
3139 	kfree(locations);
3140 	return status;
3141 }
3142 
3143 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3144 				struct nfs_fattr *fattr, struct nfs4_label *label)
3145 {
3146 	struct nfs4_getattr_arg args = {
3147 		.fh = fhandle,
3148 		.bitmask = server->attr_bitmask,
3149 	};
3150 	struct nfs4_getattr_res res = {
3151 		.fattr = fattr,
3152 		.label = label,
3153 		.server = server,
3154 	};
3155 	struct rpc_message msg = {
3156 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3157 		.rpc_argp = &args,
3158 		.rpc_resp = &res,
3159 	};
3160 
3161 	args.bitmask = nfs4_bitmask(server, label);
3162 
3163 	nfs_fattr_init(fattr);
3164 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3165 }
3166 
3167 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3168 				struct nfs_fattr *fattr, struct nfs4_label *label)
3169 {
3170 	struct nfs4_exception exception = { };
3171 	int err;
3172 	do {
3173 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3174 		trace_nfs4_getattr(server, fhandle, fattr, err);
3175 		err = nfs4_handle_exception(server, err,
3176 				&exception);
3177 	} while (exception.retry);
3178 	return err;
3179 }
3180 
3181 /*
3182  * The file is not closed if it is opened due to the a request to change
3183  * the size of the file. The open call will not be needed once the
3184  * VFS layer lookup-intents are implemented.
3185  *
3186  * Close is called when the inode is destroyed.
3187  * If we haven't opened the file for O_WRONLY, we
3188  * need to in the size_change case to obtain a stateid.
3189  *
3190  * Got race?
3191  * Because OPEN is always done by name in nfsv4, it is
3192  * possible that we opened a different file by the same
3193  * name.  We can recognize this race condition, but we
3194  * can't do anything about it besides returning an error.
3195  *
3196  * This will be fixed with VFS changes (lookup-intent).
3197  */
3198 static int
3199 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3200 		  struct iattr *sattr)
3201 {
3202 	struct inode *inode = dentry->d_inode;
3203 	struct rpc_cred *cred = NULL;
3204 	struct nfs4_state *state = NULL;
3205 	struct nfs4_label *label = NULL;
3206 	int status;
3207 
3208 	if (pnfs_ld_layoutret_on_setattr(inode))
3209 		pnfs_commit_and_return_layout(inode);
3210 
3211 	nfs_fattr_init(fattr);
3212 
3213 	/* Deal with open(O_TRUNC) */
3214 	if (sattr->ia_valid & ATTR_OPEN)
3215 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3216 
3217 	/* Optimization: if the end result is no change, don't RPC */
3218 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3219 		return 0;
3220 
3221 	/* Search for an existing open(O_WRITE) file */
3222 	if (sattr->ia_valid & ATTR_FILE) {
3223 		struct nfs_open_context *ctx;
3224 
3225 		ctx = nfs_file_open_context(sattr->ia_file);
3226 		if (ctx) {
3227 			cred = ctx->cred;
3228 			state = ctx->state;
3229 		}
3230 	}
3231 
3232 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3233 	if (IS_ERR(label))
3234 		return PTR_ERR(label);
3235 
3236 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3237 	if (status == 0) {
3238 		nfs_setattr_update_inode(inode, sattr);
3239 		nfs_setsecurity(inode, fattr, label);
3240 	}
3241 	nfs4_label_free(label);
3242 	return status;
3243 }
3244 
3245 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3246 		const struct qstr *name, struct nfs_fh *fhandle,
3247 		struct nfs_fattr *fattr, struct nfs4_label *label)
3248 {
3249 	struct nfs_server *server = NFS_SERVER(dir);
3250 	int		       status;
3251 	struct nfs4_lookup_arg args = {
3252 		.bitmask = server->attr_bitmask,
3253 		.dir_fh = NFS_FH(dir),
3254 		.name = name,
3255 	};
3256 	struct nfs4_lookup_res res = {
3257 		.server = server,
3258 		.fattr = fattr,
3259 		.label = label,
3260 		.fh = fhandle,
3261 	};
3262 	struct rpc_message msg = {
3263 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3264 		.rpc_argp = &args,
3265 		.rpc_resp = &res,
3266 	};
3267 
3268 	args.bitmask = nfs4_bitmask(server, label);
3269 
3270 	nfs_fattr_init(fattr);
3271 
3272 	dprintk("NFS call  lookup %s\n", name->name);
3273 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3274 	dprintk("NFS reply lookup: %d\n", status);
3275 	return status;
3276 }
3277 
3278 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3279 {
3280 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3281 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3282 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3283 	fattr->nlink = 2;
3284 }
3285 
3286 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3287 				   struct qstr *name, struct nfs_fh *fhandle,
3288 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3289 {
3290 	struct nfs4_exception exception = { };
3291 	struct rpc_clnt *client = *clnt;
3292 	int err;
3293 	do {
3294 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3295 		trace_nfs4_lookup(dir, name, err);
3296 		switch (err) {
3297 		case -NFS4ERR_BADNAME:
3298 			err = -ENOENT;
3299 			goto out;
3300 		case -NFS4ERR_MOVED:
3301 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3302 			goto out;
3303 		case -NFS4ERR_WRONGSEC:
3304 			err = -EPERM;
3305 			if (client != *clnt)
3306 				goto out;
3307 			client = nfs4_negotiate_security(client, dir, name);
3308 			if (IS_ERR(client))
3309 				return PTR_ERR(client);
3310 
3311 			exception.retry = 1;
3312 			break;
3313 		default:
3314 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3315 		}
3316 	} while (exception.retry);
3317 
3318 out:
3319 	if (err == 0)
3320 		*clnt = client;
3321 	else if (client != *clnt)
3322 		rpc_shutdown_client(client);
3323 
3324 	return err;
3325 }
3326 
3327 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3328 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3329 			    struct nfs4_label *label)
3330 {
3331 	int status;
3332 	struct rpc_clnt *client = NFS_CLIENT(dir);
3333 
3334 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3335 	if (client != NFS_CLIENT(dir)) {
3336 		rpc_shutdown_client(client);
3337 		nfs_fixup_secinfo_attributes(fattr);
3338 	}
3339 	return status;
3340 }
3341 
3342 struct rpc_clnt *
3343 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3344 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3345 {
3346 	struct rpc_clnt *client = NFS_CLIENT(dir);
3347 	int status;
3348 
3349 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3350 	if (status < 0)
3351 		return ERR_PTR(status);
3352 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3353 }
3354 
3355 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3356 {
3357 	struct nfs_server *server = NFS_SERVER(inode);
3358 	struct nfs4_accessargs args = {
3359 		.fh = NFS_FH(inode),
3360 		.bitmask = server->cache_consistency_bitmask,
3361 	};
3362 	struct nfs4_accessres res = {
3363 		.server = server,
3364 	};
3365 	struct rpc_message msg = {
3366 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3367 		.rpc_argp = &args,
3368 		.rpc_resp = &res,
3369 		.rpc_cred = entry->cred,
3370 	};
3371 	int mode = entry->mask;
3372 	int status = 0;
3373 
3374 	/*
3375 	 * Determine which access bits we want to ask for...
3376 	 */
3377 	if (mode & MAY_READ)
3378 		args.access |= NFS4_ACCESS_READ;
3379 	if (S_ISDIR(inode->i_mode)) {
3380 		if (mode & MAY_WRITE)
3381 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3382 		if (mode & MAY_EXEC)
3383 			args.access |= NFS4_ACCESS_LOOKUP;
3384 	} else {
3385 		if (mode & MAY_WRITE)
3386 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3387 		if (mode & MAY_EXEC)
3388 			args.access |= NFS4_ACCESS_EXECUTE;
3389 	}
3390 
3391 	res.fattr = nfs_alloc_fattr();
3392 	if (res.fattr == NULL)
3393 		return -ENOMEM;
3394 
3395 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3396 	if (!status) {
3397 		nfs_access_set_mask(entry, res.access);
3398 		nfs_refresh_inode(inode, res.fattr);
3399 	}
3400 	nfs_free_fattr(res.fattr);
3401 	return status;
3402 }
3403 
3404 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3405 {
3406 	struct nfs4_exception exception = { };
3407 	int err;
3408 	do {
3409 		err = _nfs4_proc_access(inode, entry);
3410 		trace_nfs4_access(inode, err);
3411 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3412 				&exception);
3413 	} while (exception.retry);
3414 	return err;
3415 }
3416 
3417 /*
3418  * TODO: For the time being, we don't try to get any attributes
3419  * along with any of the zero-copy operations READ, READDIR,
3420  * READLINK, WRITE.
3421  *
3422  * In the case of the first three, we want to put the GETATTR
3423  * after the read-type operation -- this is because it is hard
3424  * to predict the length of a GETATTR response in v4, and thus
3425  * align the READ data correctly.  This means that the GETATTR
3426  * may end up partially falling into the page cache, and we should
3427  * shift it into the 'tail' of the xdr_buf before processing.
3428  * To do this efficiently, we need to know the total length
3429  * of data received, which doesn't seem to be available outside
3430  * of the RPC layer.
3431  *
3432  * In the case of WRITE, we also want to put the GETATTR after
3433  * the operation -- in this case because we want to make sure
3434  * we get the post-operation mtime and size.
3435  *
3436  * Both of these changes to the XDR layer would in fact be quite
3437  * minor, but I decided to leave them for a subsequent patch.
3438  */
3439 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3440 		unsigned int pgbase, unsigned int pglen)
3441 {
3442 	struct nfs4_readlink args = {
3443 		.fh       = NFS_FH(inode),
3444 		.pgbase	  = pgbase,
3445 		.pglen    = pglen,
3446 		.pages    = &page,
3447 	};
3448 	struct nfs4_readlink_res res;
3449 	struct rpc_message msg = {
3450 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3451 		.rpc_argp = &args,
3452 		.rpc_resp = &res,
3453 	};
3454 
3455 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3456 }
3457 
3458 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3459 		unsigned int pgbase, unsigned int pglen)
3460 {
3461 	struct nfs4_exception exception = { };
3462 	int err;
3463 	do {
3464 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3465 		trace_nfs4_readlink(inode, err);
3466 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3467 				&exception);
3468 	} while (exception.retry);
3469 	return err;
3470 }
3471 
3472 /*
3473  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3474  */
3475 static int
3476 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3477 		 int flags)
3478 {
3479 	struct nfs4_label l, *ilabel = NULL;
3480 	struct nfs_open_context *ctx;
3481 	struct nfs4_state *state;
3482 	int opened = 0;
3483 	int status = 0;
3484 
3485 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3486 	if (IS_ERR(ctx))
3487 		return PTR_ERR(ctx);
3488 
3489 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3490 
3491 	sattr->ia_mode &= ~current_umask();
3492 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3493 	if (IS_ERR(state)) {
3494 		status = PTR_ERR(state);
3495 		goto out;
3496 	}
3497 out:
3498 	nfs4_label_release_security(ilabel);
3499 	put_nfs_open_context(ctx);
3500 	return status;
3501 }
3502 
3503 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3504 {
3505 	struct nfs_server *server = NFS_SERVER(dir);
3506 	struct nfs_removeargs args = {
3507 		.fh = NFS_FH(dir),
3508 		.name = *name,
3509 	};
3510 	struct nfs_removeres res = {
3511 		.server = server,
3512 	};
3513 	struct rpc_message msg = {
3514 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3515 		.rpc_argp = &args,
3516 		.rpc_resp = &res,
3517 	};
3518 	int status;
3519 
3520 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3521 	if (status == 0)
3522 		update_changeattr(dir, &res.cinfo);
3523 	return status;
3524 }
3525 
3526 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3527 {
3528 	struct nfs4_exception exception = { };
3529 	int err;
3530 	do {
3531 		err = _nfs4_proc_remove(dir, name);
3532 		trace_nfs4_remove(dir, name, err);
3533 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3534 				&exception);
3535 	} while (exception.retry);
3536 	return err;
3537 }
3538 
3539 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3540 {
3541 	struct nfs_server *server = NFS_SERVER(dir);
3542 	struct nfs_removeargs *args = msg->rpc_argp;
3543 	struct nfs_removeres *res = msg->rpc_resp;
3544 
3545 	res->server = server;
3546 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3547 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3548 
3549 	nfs_fattr_init(res->dir_attr);
3550 }
3551 
3552 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3553 {
3554 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3555 			&data->args.seq_args,
3556 			&data->res.seq_res,
3557 			task);
3558 }
3559 
3560 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3561 {
3562 	struct nfs_unlinkdata *data = task->tk_calldata;
3563 	struct nfs_removeres *res = &data->res;
3564 
3565 	if (!nfs4_sequence_done(task, &res->seq_res))
3566 		return 0;
3567 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3568 		return 0;
3569 	update_changeattr(dir, &res->cinfo);
3570 	return 1;
3571 }
3572 
3573 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3574 {
3575 	struct nfs_server *server = NFS_SERVER(dir);
3576 	struct nfs_renameargs *arg = msg->rpc_argp;
3577 	struct nfs_renameres *res = msg->rpc_resp;
3578 
3579 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3580 	res->server = server;
3581 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3582 }
3583 
3584 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3585 {
3586 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3587 			&data->args.seq_args,
3588 			&data->res.seq_res,
3589 			task);
3590 }
3591 
3592 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3593 				 struct inode *new_dir)
3594 {
3595 	struct nfs_renamedata *data = task->tk_calldata;
3596 	struct nfs_renameres *res = &data->res;
3597 
3598 	if (!nfs4_sequence_done(task, &res->seq_res))
3599 		return 0;
3600 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3601 		return 0;
3602 
3603 	update_changeattr(old_dir, &res->old_cinfo);
3604 	update_changeattr(new_dir, &res->new_cinfo);
3605 	return 1;
3606 }
3607 
3608 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3609 {
3610 	struct nfs_server *server = NFS_SERVER(inode);
3611 	struct nfs4_link_arg arg = {
3612 		.fh     = NFS_FH(inode),
3613 		.dir_fh = NFS_FH(dir),
3614 		.name   = name,
3615 		.bitmask = server->attr_bitmask,
3616 	};
3617 	struct nfs4_link_res res = {
3618 		.server = server,
3619 		.label = NULL,
3620 	};
3621 	struct rpc_message msg = {
3622 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3623 		.rpc_argp = &arg,
3624 		.rpc_resp = &res,
3625 	};
3626 	int status = -ENOMEM;
3627 
3628 	res.fattr = nfs_alloc_fattr();
3629 	if (res.fattr == NULL)
3630 		goto out;
3631 
3632 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3633 	if (IS_ERR(res.label)) {
3634 		status = PTR_ERR(res.label);
3635 		goto out;
3636 	}
3637 	arg.bitmask = nfs4_bitmask(server, res.label);
3638 
3639 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3640 	if (!status) {
3641 		update_changeattr(dir, &res.cinfo);
3642 		status = nfs_post_op_update_inode(inode, res.fattr);
3643 		if (!status)
3644 			nfs_setsecurity(inode, res.fattr, res.label);
3645 	}
3646 
3647 
3648 	nfs4_label_free(res.label);
3649 
3650 out:
3651 	nfs_free_fattr(res.fattr);
3652 	return status;
3653 }
3654 
3655 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3656 {
3657 	struct nfs4_exception exception = { };
3658 	int err;
3659 	do {
3660 		err = nfs4_handle_exception(NFS_SERVER(inode),
3661 				_nfs4_proc_link(inode, dir, name),
3662 				&exception);
3663 	} while (exception.retry);
3664 	return err;
3665 }
3666 
3667 struct nfs4_createdata {
3668 	struct rpc_message msg;
3669 	struct nfs4_create_arg arg;
3670 	struct nfs4_create_res res;
3671 	struct nfs_fh fh;
3672 	struct nfs_fattr fattr;
3673 	struct nfs4_label *label;
3674 };
3675 
3676 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3677 		struct qstr *name, struct iattr *sattr, u32 ftype)
3678 {
3679 	struct nfs4_createdata *data;
3680 
3681 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3682 	if (data != NULL) {
3683 		struct nfs_server *server = NFS_SERVER(dir);
3684 
3685 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3686 		if (IS_ERR(data->label))
3687 			goto out_free;
3688 
3689 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3690 		data->msg.rpc_argp = &data->arg;
3691 		data->msg.rpc_resp = &data->res;
3692 		data->arg.dir_fh = NFS_FH(dir);
3693 		data->arg.server = server;
3694 		data->arg.name = name;
3695 		data->arg.attrs = sattr;
3696 		data->arg.ftype = ftype;
3697 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3698 		data->res.server = server;
3699 		data->res.fh = &data->fh;
3700 		data->res.fattr = &data->fattr;
3701 		data->res.label = data->label;
3702 		nfs_fattr_init(data->res.fattr);
3703 	}
3704 	return data;
3705 out_free:
3706 	kfree(data);
3707 	return NULL;
3708 }
3709 
3710 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3711 {
3712 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3713 				    &data->arg.seq_args, &data->res.seq_res, 1);
3714 	if (status == 0) {
3715 		update_changeattr(dir, &data->res.dir_cinfo);
3716 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3717 	}
3718 	return status;
3719 }
3720 
3721 static void nfs4_free_createdata(struct nfs4_createdata *data)
3722 {
3723 	nfs4_label_free(data->label);
3724 	kfree(data);
3725 }
3726 
3727 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3728 		struct page *page, unsigned int len, struct iattr *sattr,
3729 		struct nfs4_label *label)
3730 {
3731 	struct nfs4_createdata *data;
3732 	int status = -ENAMETOOLONG;
3733 
3734 	if (len > NFS4_MAXPATHLEN)
3735 		goto out;
3736 
3737 	status = -ENOMEM;
3738 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3739 	if (data == NULL)
3740 		goto out;
3741 
3742 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3743 	data->arg.u.symlink.pages = &page;
3744 	data->arg.u.symlink.len = len;
3745 	data->arg.label = label;
3746 
3747 	status = nfs4_do_create(dir, dentry, data);
3748 
3749 	nfs4_free_createdata(data);
3750 out:
3751 	return status;
3752 }
3753 
3754 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3755 		struct page *page, unsigned int len, struct iattr *sattr)
3756 {
3757 	struct nfs4_exception exception = { };
3758 	struct nfs4_label l, *label = NULL;
3759 	int err;
3760 
3761 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3762 
3763 	do {
3764 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3765 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3766 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3767 				&exception);
3768 	} while (exception.retry);
3769 
3770 	nfs4_label_release_security(label);
3771 	return err;
3772 }
3773 
3774 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3775 		struct iattr *sattr, struct nfs4_label *label)
3776 {
3777 	struct nfs4_createdata *data;
3778 	int status = -ENOMEM;
3779 
3780 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3781 	if (data == NULL)
3782 		goto out;
3783 
3784 	data->arg.label = label;
3785 	status = nfs4_do_create(dir, dentry, data);
3786 
3787 	nfs4_free_createdata(data);
3788 out:
3789 	return status;
3790 }
3791 
3792 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3793 		struct iattr *sattr)
3794 {
3795 	struct nfs4_exception exception = { };
3796 	struct nfs4_label l, *label = NULL;
3797 	int err;
3798 
3799 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3800 
3801 	sattr->ia_mode &= ~current_umask();
3802 	do {
3803 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3804 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3805 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3806 				&exception);
3807 	} while (exception.retry);
3808 	nfs4_label_release_security(label);
3809 
3810 	return err;
3811 }
3812 
3813 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3814 		u64 cookie, struct page **pages, unsigned int count, int plus)
3815 {
3816 	struct inode		*dir = dentry->d_inode;
3817 	struct nfs4_readdir_arg args = {
3818 		.fh = NFS_FH(dir),
3819 		.pages = pages,
3820 		.pgbase = 0,
3821 		.count = count,
3822 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3823 		.plus = plus,
3824 	};
3825 	struct nfs4_readdir_res res;
3826 	struct rpc_message msg = {
3827 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3828 		.rpc_argp = &args,
3829 		.rpc_resp = &res,
3830 		.rpc_cred = cred,
3831 	};
3832 	int			status;
3833 
3834 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3835 			dentry,
3836 			(unsigned long long)cookie);
3837 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3838 	res.pgbase = args.pgbase;
3839 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3840 	if (status >= 0) {
3841 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3842 		status += args.pgbase;
3843 	}
3844 
3845 	nfs_invalidate_atime(dir);
3846 
3847 	dprintk("%s: returns %d\n", __func__, status);
3848 	return status;
3849 }
3850 
3851 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3852 		u64 cookie, struct page **pages, unsigned int count, int plus)
3853 {
3854 	struct nfs4_exception exception = { };
3855 	int err;
3856 	do {
3857 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3858 				pages, count, plus);
3859 		trace_nfs4_readdir(dentry->d_inode, err);
3860 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3861 				&exception);
3862 	} while (exception.retry);
3863 	return err;
3864 }
3865 
3866 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3867 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3868 {
3869 	struct nfs4_createdata *data;
3870 	int mode = sattr->ia_mode;
3871 	int status = -ENOMEM;
3872 
3873 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3874 	if (data == NULL)
3875 		goto out;
3876 
3877 	if (S_ISFIFO(mode))
3878 		data->arg.ftype = NF4FIFO;
3879 	else if (S_ISBLK(mode)) {
3880 		data->arg.ftype = NF4BLK;
3881 		data->arg.u.device.specdata1 = MAJOR(rdev);
3882 		data->arg.u.device.specdata2 = MINOR(rdev);
3883 	}
3884 	else if (S_ISCHR(mode)) {
3885 		data->arg.ftype = NF4CHR;
3886 		data->arg.u.device.specdata1 = MAJOR(rdev);
3887 		data->arg.u.device.specdata2 = MINOR(rdev);
3888 	} else if (!S_ISSOCK(mode)) {
3889 		status = -EINVAL;
3890 		goto out_free;
3891 	}
3892 
3893 	data->arg.label = label;
3894 	status = nfs4_do_create(dir, dentry, data);
3895 out_free:
3896 	nfs4_free_createdata(data);
3897 out:
3898 	return status;
3899 }
3900 
3901 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3902 		struct iattr *sattr, dev_t rdev)
3903 {
3904 	struct nfs4_exception exception = { };
3905 	struct nfs4_label l, *label = NULL;
3906 	int err;
3907 
3908 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3909 
3910 	sattr->ia_mode &= ~current_umask();
3911 	do {
3912 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3913 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3914 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3915 				&exception);
3916 	} while (exception.retry);
3917 
3918 	nfs4_label_release_security(label);
3919 
3920 	return err;
3921 }
3922 
3923 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3924 		 struct nfs_fsstat *fsstat)
3925 {
3926 	struct nfs4_statfs_arg args = {
3927 		.fh = fhandle,
3928 		.bitmask = server->attr_bitmask,
3929 	};
3930 	struct nfs4_statfs_res res = {
3931 		.fsstat = fsstat,
3932 	};
3933 	struct rpc_message msg = {
3934 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3935 		.rpc_argp = &args,
3936 		.rpc_resp = &res,
3937 	};
3938 
3939 	nfs_fattr_init(fsstat->fattr);
3940 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3941 }
3942 
3943 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3944 {
3945 	struct nfs4_exception exception = { };
3946 	int err;
3947 	do {
3948 		err = nfs4_handle_exception(server,
3949 				_nfs4_proc_statfs(server, fhandle, fsstat),
3950 				&exception);
3951 	} while (exception.retry);
3952 	return err;
3953 }
3954 
3955 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3956 		struct nfs_fsinfo *fsinfo)
3957 {
3958 	struct nfs4_fsinfo_arg args = {
3959 		.fh = fhandle,
3960 		.bitmask = server->attr_bitmask,
3961 	};
3962 	struct nfs4_fsinfo_res res = {
3963 		.fsinfo = fsinfo,
3964 	};
3965 	struct rpc_message msg = {
3966 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3967 		.rpc_argp = &args,
3968 		.rpc_resp = &res,
3969 	};
3970 
3971 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3972 }
3973 
3974 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3975 {
3976 	struct nfs4_exception exception = { };
3977 	unsigned long now = jiffies;
3978 	int err;
3979 
3980 	do {
3981 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3982 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
3983 		if (err == 0) {
3984 			struct nfs_client *clp = server->nfs_client;
3985 
3986 			spin_lock(&clp->cl_lock);
3987 			clp->cl_lease_time = fsinfo->lease_time * HZ;
3988 			clp->cl_last_renewal = now;
3989 			spin_unlock(&clp->cl_lock);
3990 			break;
3991 		}
3992 		err = nfs4_handle_exception(server, err, &exception);
3993 	} while (exception.retry);
3994 	return err;
3995 }
3996 
3997 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3998 {
3999 	int error;
4000 
4001 	nfs_fattr_init(fsinfo->fattr);
4002 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4003 	if (error == 0) {
4004 		/* block layout checks this! */
4005 		server->pnfs_blksize = fsinfo->blksize;
4006 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4007 	}
4008 
4009 	return error;
4010 }
4011 
4012 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4013 		struct nfs_pathconf *pathconf)
4014 {
4015 	struct nfs4_pathconf_arg args = {
4016 		.fh = fhandle,
4017 		.bitmask = server->attr_bitmask,
4018 	};
4019 	struct nfs4_pathconf_res res = {
4020 		.pathconf = pathconf,
4021 	};
4022 	struct rpc_message msg = {
4023 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4024 		.rpc_argp = &args,
4025 		.rpc_resp = &res,
4026 	};
4027 
4028 	/* None of the pathconf attributes are mandatory to implement */
4029 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4030 		memset(pathconf, 0, sizeof(*pathconf));
4031 		return 0;
4032 	}
4033 
4034 	nfs_fattr_init(pathconf->fattr);
4035 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4036 }
4037 
4038 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4039 		struct nfs_pathconf *pathconf)
4040 {
4041 	struct nfs4_exception exception = { };
4042 	int err;
4043 
4044 	do {
4045 		err = nfs4_handle_exception(server,
4046 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4047 				&exception);
4048 	} while (exception.retry);
4049 	return err;
4050 }
4051 
4052 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4053 		const struct nfs_open_context *ctx,
4054 		const struct nfs_lock_context *l_ctx,
4055 		fmode_t fmode)
4056 {
4057 	const struct nfs_lockowner *lockowner = NULL;
4058 
4059 	if (l_ctx != NULL)
4060 		lockowner = &l_ctx->lockowner;
4061 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4062 }
4063 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4064 
4065 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4066 		const struct nfs_open_context *ctx,
4067 		const struct nfs_lock_context *l_ctx,
4068 		fmode_t fmode)
4069 {
4070 	nfs4_stateid current_stateid;
4071 
4072 	/* If the current stateid represents a lost lock, then exit */
4073 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4074 		return true;
4075 	return nfs4_stateid_match(stateid, &current_stateid);
4076 }
4077 
4078 static bool nfs4_error_stateid_expired(int err)
4079 {
4080 	switch (err) {
4081 	case -NFS4ERR_DELEG_REVOKED:
4082 	case -NFS4ERR_ADMIN_REVOKED:
4083 	case -NFS4ERR_BAD_STATEID:
4084 	case -NFS4ERR_STALE_STATEID:
4085 	case -NFS4ERR_OLD_STATEID:
4086 	case -NFS4ERR_OPENMODE:
4087 	case -NFS4ERR_EXPIRED:
4088 		return true;
4089 	}
4090 	return false;
4091 }
4092 
4093 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4094 {
4095 	nfs_invalidate_atime(hdr->inode);
4096 }
4097 
4098 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4099 {
4100 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4101 
4102 	trace_nfs4_read(hdr, task->tk_status);
4103 	if (nfs4_async_handle_error(task, server,
4104 				    hdr->args.context->state) == -EAGAIN) {
4105 		rpc_restart_call_prepare(task);
4106 		return -EAGAIN;
4107 	}
4108 
4109 	__nfs4_read_done_cb(hdr);
4110 	if (task->tk_status > 0)
4111 		renew_lease(server, hdr->timestamp);
4112 	return 0;
4113 }
4114 
4115 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4116 		struct nfs_pgio_args *args)
4117 {
4118 
4119 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4120 		nfs4_stateid_is_current(&args->stateid,
4121 				args->context,
4122 				args->lock_context,
4123 				FMODE_READ))
4124 		return false;
4125 	rpc_restart_call_prepare(task);
4126 	return true;
4127 }
4128 
4129 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4130 {
4131 
4132 	dprintk("--> %s\n", __func__);
4133 
4134 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4135 		return -EAGAIN;
4136 	if (nfs4_read_stateid_changed(task, &hdr->args))
4137 		return -EAGAIN;
4138 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4139 				    nfs4_read_done_cb(task, hdr);
4140 }
4141 
4142 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4143 				 struct rpc_message *msg)
4144 {
4145 	hdr->timestamp   = jiffies;
4146 	hdr->pgio_done_cb = nfs4_read_done_cb;
4147 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4148 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4149 }
4150 
4151 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4152 				      struct nfs_pgio_header *hdr)
4153 {
4154 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4155 			&hdr->args.seq_args,
4156 			&hdr->res.seq_res,
4157 			task))
4158 		return 0;
4159 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4160 				hdr->args.lock_context,
4161 				hdr->rw_ops->rw_mode) == -EIO)
4162 		return -EIO;
4163 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4164 		return -EIO;
4165 	return 0;
4166 }
4167 
4168 static int nfs4_write_done_cb(struct rpc_task *task,
4169 			      struct nfs_pgio_header *hdr)
4170 {
4171 	struct inode *inode = hdr->inode;
4172 
4173 	trace_nfs4_write(hdr, task->tk_status);
4174 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4175 				    hdr->args.context->state) == -EAGAIN) {
4176 		rpc_restart_call_prepare(task);
4177 		return -EAGAIN;
4178 	}
4179 	if (task->tk_status >= 0) {
4180 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4181 		nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4182 	}
4183 	return 0;
4184 }
4185 
4186 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4187 		struct nfs_pgio_args *args)
4188 {
4189 
4190 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4191 		nfs4_stateid_is_current(&args->stateid,
4192 				args->context,
4193 				args->lock_context,
4194 				FMODE_WRITE))
4195 		return false;
4196 	rpc_restart_call_prepare(task);
4197 	return true;
4198 }
4199 
4200 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4201 {
4202 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4203 		return -EAGAIN;
4204 	if (nfs4_write_stateid_changed(task, &hdr->args))
4205 		return -EAGAIN;
4206 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4207 		nfs4_write_done_cb(task, hdr);
4208 }
4209 
4210 static
4211 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4212 {
4213 	/* Don't request attributes for pNFS or O_DIRECT writes */
4214 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4215 		return false;
4216 	/* Otherwise, request attributes if and only if we don't hold
4217 	 * a delegation
4218 	 */
4219 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4220 }
4221 
4222 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4223 				  struct rpc_message *msg)
4224 {
4225 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4226 
4227 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4228 		hdr->args.bitmask = NULL;
4229 		hdr->res.fattr = NULL;
4230 	} else
4231 		hdr->args.bitmask = server->cache_consistency_bitmask;
4232 
4233 	if (!hdr->pgio_done_cb)
4234 		hdr->pgio_done_cb = nfs4_write_done_cb;
4235 	hdr->res.server = server;
4236 	hdr->timestamp   = jiffies;
4237 
4238 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4239 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4240 }
4241 
4242 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4243 {
4244 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4245 			&data->args.seq_args,
4246 			&data->res.seq_res,
4247 			task);
4248 }
4249 
4250 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4251 {
4252 	struct inode *inode = data->inode;
4253 
4254 	trace_nfs4_commit(data, task->tk_status);
4255 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4256 		rpc_restart_call_prepare(task);
4257 		return -EAGAIN;
4258 	}
4259 	return 0;
4260 }
4261 
4262 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4263 {
4264 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4265 		return -EAGAIN;
4266 	return data->commit_done_cb(task, data);
4267 }
4268 
4269 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4270 {
4271 	struct nfs_server *server = NFS_SERVER(data->inode);
4272 
4273 	if (data->commit_done_cb == NULL)
4274 		data->commit_done_cb = nfs4_commit_done_cb;
4275 	data->res.server = server;
4276 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4277 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4278 }
4279 
4280 struct nfs4_renewdata {
4281 	struct nfs_client	*client;
4282 	unsigned long		timestamp;
4283 };
4284 
4285 /*
4286  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4287  * standalone procedure for queueing an asynchronous RENEW.
4288  */
4289 static void nfs4_renew_release(void *calldata)
4290 {
4291 	struct nfs4_renewdata *data = calldata;
4292 	struct nfs_client *clp = data->client;
4293 
4294 	if (atomic_read(&clp->cl_count) > 1)
4295 		nfs4_schedule_state_renewal(clp);
4296 	nfs_put_client(clp);
4297 	kfree(data);
4298 }
4299 
4300 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4301 {
4302 	struct nfs4_renewdata *data = calldata;
4303 	struct nfs_client *clp = data->client;
4304 	unsigned long timestamp = data->timestamp;
4305 
4306 	trace_nfs4_renew_async(clp, task->tk_status);
4307 	switch (task->tk_status) {
4308 	case 0:
4309 		break;
4310 	case -NFS4ERR_LEASE_MOVED:
4311 		nfs4_schedule_lease_moved_recovery(clp);
4312 		break;
4313 	default:
4314 		/* Unless we're shutting down, schedule state recovery! */
4315 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4316 			return;
4317 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4318 			nfs4_schedule_lease_recovery(clp);
4319 			return;
4320 		}
4321 		nfs4_schedule_path_down_recovery(clp);
4322 	}
4323 	do_renew_lease(clp, timestamp);
4324 }
4325 
4326 static const struct rpc_call_ops nfs4_renew_ops = {
4327 	.rpc_call_done = nfs4_renew_done,
4328 	.rpc_release = nfs4_renew_release,
4329 };
4330 
4331 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4332 {
4333 	struct rpc_message msg = {
4334 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4335 		.rpc_argp	= clp,
4336 		.rpc_cred	= cred,
4337 	};
4338 	struct nfs4_renewdata *data;
4339 
4340 	if (renew_flags == 0)
4341 		return 0;
4342 	if (!atomic_inc_not_zero(&clp->cl_count))
4343 		return -EIO;
4344 	data = kmalloc(sizeof(*data), GFP_NOFS);
4345 	if (data == NULL)
4346 		return -ENOMEM;
4347 	data->client = clp;
4348 	data->timestamp = jiffies;
4349 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4350 			&nfs4_renew_ops, data);
4351 }
4352 
4353 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4354 {
4355 	struct rpc_message msg = {
4356 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4357 		.rpc_argp	= clp,
4358 		.rpc_cred	= cred,
4359 	};
4360 	unsigned long now = jiffies;
4361 	int status;
4362 
4363 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4364 	if (status < 0)
4365 		return status;
4366 	do_renew_lease(clp, now);
4367 	return 0;
4368 }
4369 
4370 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4371 {
4372 	return server->caps & NFS_CAP_ACLS;
4373 }
4374 
4375 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4376  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4377  * the stack.
4378  */
4379 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4380 
4381 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4382 		struct page **pages, unsigned int *pgbase)
4383 {
4384 	struct page *newpage, **spages;
4385 	int rc = 0;
4386 	size_t len;
4387 	spages = pages;
4388 
4389 	do {
4390 		len = min_t(size_t, PAGE_SIZE, buflen);
4391 		newpage = alloc_page(GFP_KERNEL);
4392 
4393 		if (newpage == NULL)
4394 			goto unwind;
4395 		memcpy(page_address(newpage), buf, len);
4396                 buf += len;
4397                 buflen -= len;
4398 		*pages++ = newpage;
4399 		rc++;
4400 	} while (buflen != 0);
4401 
4402 	return rc;
4403 
4404 unwind:
4405 	for(; rc > 0; rc--)
4406 		__free_page(spages[rc-1]);
4407 	return -ENOMEM;
4408 }
4409 
4410 struct nfs4_cached_acl {
4411 	int cached;
4412 	size_t len;
4413 	char data[0];
4414 };
4415 
4416 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4417 {
4418 	struct nfs_inode *nfsi = NFS_I(inode);
4419 
4420 	spin_lock(&inode->i_lock);
4421 	kfree(nfsi->nfs4_acl);
4422 	nfsi->nfs4_acl = acl;
4423 	spin_unlock(&inode->i_lock);
4424 }
4425 
4426 static void nfs4_zap_acl_attr(struct inode *inode)
4427 {
4428 	nfs4_set_cached_acl(inode, NULL);
4429 }
4430 
4431 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4432 {
4433 	struct nfs_inode *nfsi = NFS_I(inode);
4434 	struct nfs4_cached_acl *acl;
4435 	int ret = -ENOENT;
4436 
4437 	spin_lock(&inode->i_lock);
4438 	acl = nfsi->nfs4_acl;
4439 	if (acl == NULL)
4440 		goto out;
4441 	if (buf == NULL) /* user is just asking for length */
4442 		goto out_len;
4443 	if (acl->cached == 0)
4444 		goto out;
4445 	ret = -ERANGE; /* see getxattr(2) man page */
4446 	if (acl->len > buflen)
4447 		goto out;
4448 	memcpy(buf, acl->data, acl->len);
4449 out_len:
4450 	ret = acl->len;
4451 out:
4452 	spin_unlock(&inode->i_lock);
4453 	return ret;
4454 }
4455 
4456 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4457 {
4458 	struct nfs4_cached_acl *acl;
4459 	size_t buflen = sizeof(*acl) + acl_len;
4460 
4461 	if (buflen <= PAGE_SIZE) {
4462 		acl = kmalloc(buflen, GFP_KERNEL);
4463 		if (acl == NULL)
4464 			goto out;
4465 		acl->cached = 1;
4466 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4467 	} else {
4468 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4469 		if (acl == NULL)
4470 			goto out;
4471 		acl->cached = 0;
4472 	}
4473 	acl->len = acl_len;
4474 out:
4475 	nfs4_set_cached_acl(inode, acl);
4476 }
4477 
4478 /*
4479  * The getxattr API returns the required buffer length when called with a
4480  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4481  * the required buf.  On a NULL buf, we send a page of data to the server
4482  * guessing that the ACL request can be serviced by a page. If so, we cache
4483  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4484  * the cache. If not so, we throw away the page, and cache the required
4485  * length. The next getxattr call will then produce another round trip to
4486  * the server, this time with the input buf of the required size.
4487  */
4488 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4489 {
4490 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4491 	struct nfs_getaclargs args = {
4492 		.fh = NFS_FH(inode),
4493 		.acl_pages = pages,
4494 		.acl_len = buflen,
4495 	};
4496 	struct nfs_getaclres res = {
4497 		.acl_len = buflen,
4498 	};
4499 	struct rpc_message msg = {
4500 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4501 		.rpc_argp = &args,
4502 		.rpc_resp = &res,
4503 	};
4504 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4505 	int ret = -ENOMEM, i;
4506 
4507 	/* As long as we're doing a round trip to the server anyway,
4508 	 * let's be prepared for a page of acl data. */
4509 	if (npages == 0)
4510 		npages = 1;
4511 	if (npages > ARRAY_SIZE(pages))
4512 		return -ERANGE;
4513 
4514 	for (i = 0; i < npages; i++) {
4515 		pages[i] = alloc_page(GFP_KERNEL);
4516 		if (!pages[i])
4517 			goto out_free;
4518 	}
4519 
4520 	/* for decoding across pages */
4521 	res.acl_scratch = alloc_page(GFP_KERNEL);
4522 	if (!res.acl_scratch)
4523 		goto out_free;
4524 
4525 	args.acl_len = npages * PAGE_SIZE;
4526 	args.acl_pgbase = 0;
4527 
4528 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4529 		__func__, buf, buflen, npages, args.acl_len);
4530 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4531 			     &msg, &args.seq_args, &res.seq_res, 0);
4532 	if (ret)
4533 		goto out_free;
4534 
4535 	/* Handle the case where the passed-in buffer is too short */
4536 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4537 		/* Did the user only issue a request for the acl length? */
4538 		if (buf == NULL)
4539 			goto out_ok;
4540 		ret = -ERANGE;
4541 		goto out_free;
4542 	}
4543 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4544 	if (buf) {
4545 		if (res.acl_len > buflen) {
4546 			ret = -ERANGE;
4547 			goto out_free;
4548 		}
4549 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4550 	}
4551 out_ok:
4552 	ret = res.acl_len;
4553 out_free:
4554 	for (i = 0; i < npages; i++)
4555 		if (pages[i])
4556 			__free_page(pages[i]);
4557 	if (res.acl_scratch)
4558 		__free_page(res.acl_scratch);
4559 	return ret;
4560 }
4561 
4562 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4563 {
4564 	struct nfs4_exception exception = { };
4565 	ssize_t ret;
4566 	do {
4567 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4568 		trace_nfs4_get_acl(inode, ret);
4569 		if (ret >= 0)
4570 			break;
4571 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4572 	} while (exception.retry);
4573 	return ret;
4574 }
4575 
4576 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4577 {
4578 	struct nfs_server *server = NFS_SERVER(inode);
4579 	int ret;
4580 
4581 	if (!nfs4_server_supports_acls(server))
4582 		return -EOPNOTSUPP;
4583 	ret = nfs_revalidate_inode(server, inode);
4584 	if (ret < 0)
4585 		return ret;
4586 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4587 		nfs_zap_acl_cache(inode);
4588 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4589 	if (ret != -ENOENT)
4590 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4591 		 * but no cached acl data, just the acl length */
4592 		return ret;
4593 	return nfs4_get_acl_uncached(inode, buf, buflen);
4594 }
4595 
4596 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4597 {
4598 	struct nfs_server *server = NFS_SERVER(inode);
4599 	struct page *pages[NFS4ACL_MAXPAGES];
4600 	struct nfs_setaclargs arg = {
4601 		.fh		= NFS_FH(inode),
4602 		.acl_pages	= pages,
4603 		.acl_len	= buflen,
4604 	};
4605 	struct nfs_setaclres res;
4606 	struct rpc_message msg = {
4607 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4608 		.rpc_argp	= &arg,
4609 		.rpc_resp	= &res,
4610 	};
4611 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4612 	int ret, i;
4613 
4614 	if (!nfs4_server_supports_acls(server))
4615 		return -EOPNOTSUPP;
4616 	if (npages > ARRAY_SIZE(pages))
4617 		return -ERANGE;
4618 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4619 	if (i < 0)
4620 		return i;
4621 	nfs4_inode_return_delegation(inode);
4622 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4623 
4624 	/*
4625 	 * Free each page after tx, so the only ref left is
4626 	 * held by the network stack
4627 	 */
4628 	for (; i > 0; i--)
4629 		put_page(pages[i-1]);
4630 
4631 	/*
4632 	 * Acl update can result in inode attribute update.
4633 	 * so mark the attribute cache invalid.
4634 	 */
4635 	spin_lock(&inode->i_lock);
4636 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4637 	spin_unlock(&inode->i_lock);
4638 	nfs_access_zap_cache(inode);
4639 	nfs_zap_acl_cache(inode);
4640 	return ret;
4641 }
4642 
4643 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4644 {
4645 	struct nfs4_exception exception = { };
4646 	int err;
4647 	do {
4648 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4649 		trace_nfs4_set_acl(inode, err);
4650 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4651 				&exception);
4652 	} while (exception.retry);
4653 	return err;
4654 }
4655 
4656 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4657 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4658 					size_t buflen)
4659 {
4660 	struct nfs_server *server = NFS_SERVER(inode);
4661 	struct nfs_fattr fattr;
4662 	struct nfs4_label label = {0, 0, buflen, buf};
4663 
4664 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4665 	struct nfs4_getattr_arg arg = {
4666 		.fh		= NFS_FH(inode),
4667 		.bitmask	= bitmask,
4668 	};
4669 	struct nfs4_getattr_res res = {
4670 		.fattr		= &fattr,
4671 		.label		= &label,
4672 		.server		= server,
4673 	};
4674 	struct rpc_message msg = {
4675 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4676 		.rpc_argp	= &arg,
4677 		.rpc_resp	= &res,
4678 	};
4679 	int ret;
4680 
4681 	nfs_fattr_init(&fattr);
4682 
4683 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4684 	if (ret)
4685 		return ret;
4686 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4687 		return -ENOENT;
4688 	if (buflen < label.len)
4689 		return -ERANGE;
4690 	return 0;
4691 }
4692 
4693 static int nfs4_get_security_label(struct inode *inode, void *buf,
4694 					size_t buflen)
4695 {
4696 	struct nfs4_exception exception = { };
4697 	int err;
4698 
4699 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4700 		return -EOPNOTSUPP;
4701 
4702 	do {
4703 		err = _nfs4_get_security_label(inode, buf, buflen);
4704 		trace_nfs4_get_security_label(inode, err);
4705 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4706 				&exception);
4707 	} while (exception.retry);
4708 	return err;
4709 }
4710 
4711 static int _nfs4_do_set_security_label(struct inode *inode,
4712 		struct nfs4_label *ilabel,
4713 		struct nfs_fattr *fattr,
4714 		struct nfs4_label *olabel)
4715 {
4716 
4717 	struct iattr sattr = {0};
4718 	struct nfs_server *server = NFS_SERVER(inode);
4719 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4720 	struct nfs_setattrargs arg = {
4721 		.fh             = NFS_FH(inode),
4722 		.iap            = &sattr,
4723 		.server		= server,
4724 		.bitmask	= bitmask,
4725 		.label		= ilabel,
4726 	};
4727 	struct nfs_setattrres res = {
4728 		.fattr		= fattr,
4729 		.label		= olabel,
4730 		.server		= server,
4731 	};
4732 	struct rpc_message msg = {
4733 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4734 		.rpc_argp       = &arg,
4735 		.rpc_resp       = &res,
4736 	};
4737 	int status;
4738 
4739 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4740 
4741 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4742 	if (status)
4743 		dprintk("%s failed: %d\n", __func__, status);
4744 
4745 	return status;
4746 }
4747 
4748 static int nfs4_do_set_security_label(struct inode *inode,
4749 		struct nfs4_label *ilabel,
4750 		struct nfs_fattr *fattr,
4751 		struct nfs4_label *olabel)
4752 {
4753 	struct nfs4_exception exception = { };
4754 	int err;
4755 
4756 	do {
4757 		err = _nfs4_do_set_security_label(inode, ilabel,
4758 				fattr, olabel);
4759 		trace_nfs4_set_security_label(inode, err);
4760 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4761 				&exception);
4762 	} while (exception.retry);
4763 	return err;
4764 }
4765 
4766 static int
4767 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4768 {
4769 	struct nfs4_label ilabel, *olabel = NULL;
4770 	struct nfs_fattr fattr;
4771 	struct rpc_cred *cred;
4772 	struct inode *inode = dentry->d_inode;
4773 	int status;
4774 
4775 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4776 		return -EOPNOTSUPP;
4777 
4778 	nfs_fattr_init(&fattr);
4779 
4780 	ilabel.pi = 0;
4781 	ilabel.lfs = 0;
4782 	ilabel.label = (char *)buf;
4783 	ilabel.len = buflen;
4784 
4785 	cred = rpc_lookup_cred();
4786 	if (IS_ERR(cred))
4787 		return PTR_ERR(cred);
4788 
4789 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4790 	if (IS_ERR(olabel)) {
4791 		status = -PTR_ERR(olabel);
4792 		goto out;
4793 	}
4794 
4795 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4796 	if (status == 0)
4797 		nfs_setsecurity(inode, &fattr, olabel);
4798 
4799 	nfs4_label_free(olabel);
4800 out:
4801 	put_rpccred(cred);
4802 	return status;
4803 }
4804 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4805 
4806 
4807 static int
4808 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4809 {
4810 	struct nfs_client *clp = server->nfs_client;
4811 
4812 	if (task->tk_status >= 0)
4813 		return 0;
4814 	switch(task->tk_status) {
4815 		case -NFS4ERR_DELEG_REVOKED:
4816 		case -NFS4ERR_ADMIN_REVOKED:
4817 		case -NFS4ERR_BAD_STATEID:
4818 			if (state == NULL)
4819 				break;
4820 			nfs_remove_bad_delegation(state->inode);
4821 		case -NFS4ERR_OPENMODE:
4822 			if (state == NULL)
4823 				break;
4824 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4825 				goto recovery_failed;
4826 			goto wait_on_recovery;
4827 		case -NFS4ERR_EXPIRED:
4828 			if (state != NULL) {
4829 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4830 					goto recovery_failed;
4831 			}
4832 		case -NFS4ERR_STALE_STATEID:
4833 		case -NFS4ERR_STALE_CLIENTID:
4834 			nfs4_schedule_lease_recovery(clp);
4835 			goto wait_on_recovery;
4836 		case -NFS4ERR_MOVED:
4837 			if (nfs4_schedule_migration_recovery(server) < 0)
4838 				goto recovery_failed;
4839 			goto wait_on_recovery;
4840 		case -NFS4ERR_LEASE_MOVED:
4841 			nfs4_schedule_lease_moved_recovery(clp);
4842 			goto wait_on_recovery;
4843 #if defined(CONFIG_NFS_V4_1)
4844 		case -NFS4ERR_BADSESSION:
4845 		case -NFS4ERR_BADSLOT:
4846 		case -NFS4ERR_BAD_HIGH_SLOT:
4847 		case -NFS4ERR_DEADSESSION:
4848 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4849 		case -NFS4ERR_SEQ_FALSE_RETRY:
4850 		case -NFS4ERR_SEQ_MISORDERED:
4851 			dprintk("%s ERROR %d, Reset session\n", __func__,
4852 				task->tk_status);
4853 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4854 			goto wait_on_recovery;
4855 #endif /* CONFIG_NFS_V4_1 */
4856 		case -NFS4ERR_DELAY:
4857 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4858 		case -NFS4ERR_GRACE:
4859 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4860 		case -NFS4ERR_RETRY_UNCACHED_REP:
4861 		case -NFS4ERR_OLD_STATEID:
4862 			goto restart_call;
4863 	}
4864 	task->tk_status = nfs4_map_errors(task->tk_status);
4865 	return 0;
4866 recovery_failed:
4867 	task->tk_status = -EIO;
4868 	return 0;
4869 wait_on_recovery:
4870 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4871 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4872 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4873 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4874 		goto recovery_failed;
4875 restart_call:
4876 	task->tk_status = 0;
4877 	return -EAGAIN;
4878 }
4879 
4880 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4881 				    nfs4_verifier *bootverf)
4882 {
4883 	__be32 verf[2];
4884 
4885 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4886 		/* An impossible timestamp guarantees this value
4887 		 * will never match a generated boot time. */
4888 		verf[0] = 0;
4889 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4890 	} else {
4891 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4892 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4893 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4894 	}
4895 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4896 }
4897 
4898 static unsigned int
4899 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4900 				   char *buf, size_t len)
4901 {
4902 	unsigned int result;
4903 
4904 	rcu_read_lock();
4905 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4906 				clp->cl_ipaddr,
4907 				rpc_peeraddr2str(clp->cl_rpcclient,
4908 							RPC_DISPLAY_ADDR),
4909 				rpc_peeraddr2str(clp->cl_rpcclient,
4910 							RPC_DISPLAY_PROTO));
4911 	rcu_read_unlock();
4912 	return result;
4913 }
4914 
4915 static unsigned int
4916 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4917 				char *buf, size_t len)
4918 {
4919 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4920 
4921 	if (nfs4_client_id_uniquifier[0] != '\0')
4922 		return scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4923 				clp->rpc_ops->version,
4924 				clp->cl_minorversion,
4925 				nfs4_client_id_uniquifier,
4926 				nodename);
4927 	return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4928 				clp->rpc_ops->version, clp->cl_minorversion,
4929 				nodename);
4930 }
4931 
4932 /*
4933  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4934  * services.  Advertise one based on the address family of the
4935  * clientaddr.
4936  */
4937 static unsigned int
4938 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4939 {
4940 	if (strchr(clp->cl_ipaddr, ':') != NULL)
4941 		return scnprintf(buf, len, "tcp6");
4942 	else
4943 		return scnprintf(buf, len, "tcp");
4944 }
4945 
4946 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
4947 {
4948 	struct nfs4_setclientid *sc = calldata;
4949 
4950 	if (task->tk_status == 0)
4951 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
4952 }
4953 
4954 static const struct rpc_call_ops nfs4_setclientid_ops = {
4955 	.rpc_call_done = nfs4_setclientid_done,
4956 };
4957 
4958 /**
4959  * nfs4_proc_setclientid - Negotiate client ID
4960  * @clp: state data structure
4961  * @program: RPC program for NFSv4 callback service
4962  * @port: IP port number for NFS4 callback service
4963  * @cred: RPC credential to use for this call
4964  * @res: where to place the result
4965  *
4966  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4967  */
4968 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4969 		unsigned short port, struct rpc_cred *cred,
4970 		struct nfs4_setclientid_res *res)
4971 {
4972 	nfs4_verifier sc_verifier;
4973 	struct nfs4_setclientid setclientid = {
4974 		.sc_verifier = &sc_verifier,
4975 		.sc_prog = program,
4976 		.sc_cb_ident = clp->cl_cb_ident,
4977 	};
4978 	struct rpc_message msg = {
4979 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4980 		.rpc_argp = &setclientid,
4981 		.rpc_resp = res,
4982 		.rpc_cred = cred,
4983 	};
4984 	struct rpc_task *task;
4985 	struct rpc_task_setup task_setup_data = {
4986 		.rpc_client = clp->cl_rpcclient,
4987 		.rpc_message = &msg,
4988 		.callback_ops = &nfs4_setclientid_ops,
4989 		.callback_data = &setclientid,
4990 		.flags = RPC_TASK_TIMEOUT,
4991 	};
4992 	int status;
4993 
4994 	/* nfs_client_id4 */
4995 	nfs4_init_boot_verifier(clp, &sc_verifier);
4996 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4997 		setclientid.sc_name_len =
4998 				nfs4_init_uniform_client_string(clp,
4999 						setclientid.sc_name,
5000 						sizeof(setclientid.sc_name));
5001 	else
5002 		setclientid.sc_name_len =
5003 				nfs4_init_nonuniform_client_string(clp,
5004 						setclientid.sc_name,
5005 						sizeof(setclientid.sc_name));
5006 	/* cb_client4 */
5007 	setclientid.sc_netid_len =
5008 				nfs4_init_callback_netid(clp,
5009 						setclientid.sc_netid,
5010 						sizeof(setclientid.sc_netid));
5011 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5012 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5013 				clp->cl_ipaddr, port >> 8, port & 255);
5014 
5015 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5016 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5017 		setclientid.sc_name_len, setclientid.sc_name);
5018 	task = rpc_run_task(&task_setup_data);
5019 	if (IS_ERR(task)) {
5020 		status = PTR_ERR(task);
5021 		goto out;
5022 	}
5023 	status = task->tk_status;
5024 	if (setclientid.sc_cred) {
5025 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5026 		put_rpccred(setclientid.sc_cred);
5027 	}
5028 	rpc_put_task(task);
5029 out:
5030 	trace_nfs4_setclientid(clp, status);
5031 	dprintk("NFS reply setclientid: %d\n", status);
5032 	return status;
5033 }
5034 
5035 /**
5036  * nfs4_proc_setclientid_confirm - Confirm client ID
5037  * @clp: state data structure
5038  * @res: result of a previous SETCLIENTID
5039  * @cred: RPC credential to use for this call
5040  *
5041  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5042  */
5043 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5044 		struct nfs4_setclientid_res *arg,
5045 		struct rpc_cred *cred)
5046 {
5047 	struct rpc_message msg = {
5048 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5049 		.rpc_argp = arg,
5050 		.rpc_cred = cred,
5051 	};
5052 	int status;
5053 
5054 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5055 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5056 		clp->cl_clientid);
5057 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5058 	trace_nfs4_setclientid_confirm(clp, status);
5059 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5060 	return status;
5061 }
5062 
5063 struct nfs4_delegreturndata {
5064 	struct nfs4_delegreturnargs args;
5065 	struct nfs4_delegreturnres res;
5066 	struct nfs_fh fh;
5067 	nfs4_stateid stateid;
5068 	unsigned long timestamp;
5069 	struct nfs_fattr fattr;
5070 	int rpc_status;
5071 	struct inode *inode;
5072 	bool roc;
5073 	u32 roc_barrier;
5074 };
5075 
5076 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5077 {
5078 	struct nfs4_delegreturndata *data = calldata;
5079 
5080 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5081 		return;
5082 
5083 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5084 	switch (task->tk_status) {
5085 	case 0:
5086 		renew_lease(data->res.server, data->timestamp);
5087 	case -NFS4ERR_ADMIN_REVOKED:
5088 	case -NFS4ERR_DELEG_REVOKED:
5089 	case -NFS4ERR_BAD_STATEID:
5090 	case -NFS4ERR_OLD_STATEID:
5091 	case -NFS4ERR_STALE_STATEID:
5092 	case -NFS4ERR_EXPIRED:
5093 		task->tk_status = 0;
5094 		if (data->roc)
5095 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5096 		break;
5097 	default:
5098 		if (nfs4_async_handle_error(task, data->res.server, NULL) ==
5099 				-EAGAIN) {
5100 			rpc_restart_call_prepare(task);
5101 			return;
5102 		}
5103 	}
5104 	data->rpc_status = task->tk_status;
5105 }
5106 
5107 static void nfs4_delegreturn_release(void *calldata)
5108 {
5109 	struct nfs4_delegreturndata *data = calldata;
5110 
5111 	if (data->roc)
5112 		pnfs_roc_release(data->inode);
5113 	kfree(calldata);
5114 }
5115 
5116 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5117 {
5118 	struct nfs4_delegreturndata *d_data;
5119 
5120 	d_data = (struct nfs4_delegreturndata *)data;
5121 
5122 	if (d_data->roc &&
5123 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5124 		return;
5125 
5126 	nfs4_setup_sequence(d_data->res.server,
5127 			&d_data->args.seq_args,
5128 			&d_data->res.seq_res,
5129 			task);
5130 }
5131 
5132 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5133 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5134 	.rpc_call_done = nfs4_delegreturn_done,
5135 	.rpc_release = nfs4_delegreturn_release,
5136 };
5137 
5138 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5139 {
5140 	struct nfs4_delegreturndata *data;
5141 	struct nfs_server *server = NFS_SERVER(inode);
5142 	struct rpc_task *task;
5143 	struct rpc_message msg = {
5144 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5145 		.rpc_cred = cred,
5146 	};
5147 	struct rpc_task_setup task_setup_data = {
5148 		.rpc_client = server->client,
5149 		.rpc_message = &msg,
5150 		.callback_ops = &nfs4_delegreturn_ops,
5151 		.flags = RPC_TASK_ASYNC,
5152 	};
5153 	int status = 0;
5154 
5155 	data = kzalloc(sizeof(*data), GFP_NOFS);
5156 	if (data == NULL)
5157 		return -ENOMEM;
5158 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5159 	data->args.fhandle = &data->fh;
5160 	data->args.stateid = &data->stateid;
5161 	data->args.bitmask = server->cache_consistency_bitmask;
5162 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5163 	nfs4_stateid_copy(&data->stateid, stateid);
5164 	data->res.fattr = &data->fattr;
5165 	data->res.server = server;
5166 	nfs_fattr_init(data->res.fattr);
5167 	data->timestamp = jiffies;
5168 	data->rpc_status = 0;
5169 	data->inode = inode;
5170 	data->roc = list_empty(&NFS_I(inode)->open_files) ?
5171 		    pnfs_roc(inode) : false;
5172 
5173 	task_setup_data.callback_data = data;
5174 	msg.rpc_argp = &data->args;
5175 	msg.rpc_resp = &data->res;
5176 	task = rpc_run_task(&task_setup_data);
5177 	if (IS_ERR(task))
5178 		return PTR_ERR(task);
5179 	if (!issync)
5180 		goto out;
5181 	status = nfs4_wait_for_completion_rpc_task(task);
5182 	if (status != 0)
5183 		goto out;
5184 	status = data->rpc_status;
5185 	if (status == 0)
5186 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5187 	else
5188 		nfs_refresh_inode(inode, &data->fattr);
5189 out:
5190 	rpc_put_task(task);
5191 	return status;
5192 }
5193 
5194 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5195 {
5196 	struct nfs_server *server = NFS_SERVER(inode);
5197 	struct nfs4_exception exception = { };
5198 	int err;
5199 	do {
5200 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5201 		trace_nfs4_delegreturn(inode, err);
5202 		switch (err) {
5203 			case -NFS4ERR_STALE_STATEID:
5204 			case -NFS4ERR_EXPIRED:
5205 			case 0:
5206 				return 0;
5207 		}
5208 		err = nfs4_handle_exception(server, err, &exception);
5209 	} while (exception.retry);
5210 	return err;
5211 }
5212 
5213 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5214 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5215 
5216 /*
5217  * sleep, with exponential backoff, and retry the LOCK operation.
5218  */
5219 static unsigned long
5220 nfs4_set_lock_task_retry(unsigned long timeout)
5221 {
5222 	freezable_schedule_timeout_killable_unsafe(timeout);
5223 	timeout <<= 1;
5224 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5225 		return NFS4_LOCK_MAXTIMEOUT;
5226 	return timeout;
5227 }
5228 
5229 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5230 {
5231 	struct inode *inode = state->inode;
5232 	struct nfs_server *server = NFS_SERVER(inode);
5233 	struct nfs_client *clp = server->nfs_client;
5234 	struct nfs_lockt_args arg = {
5235 		.fh = NFS_FH(inode),
5236 		.fl = request,
5237 	};
5238 	struct nfs_lockt_res res = {
5239 		.denied = request,
5240 	};
5241 	struct rpc_message msg = {
5242 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5243 		.rpc_argp       = &arg,
5244 		.rpc_resp       = &res,
5245 		.rpc_cred	= state->owner->so_cred,
5246 	};
5247 	struct nfs4_lock_state *lsp;
5248 	int status;
5249 
5250 	arg.lock_owner.clientid = clp->cl_clientid;
5251 	status = nfs4_set_lock_state(state, request);
5252 	if (status != 0)
5253 		goto out;
5254 	lsp = request->fl_u.nfs4_fl.owner;
5255 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5256 	arg.lock_owner.s_dev = server->s_dev;
5257 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5258 	switch (status) {
5259 		case 0:
5260 			request->fl_type = F_UNLCK;
5261 			break;
5262 		case -NFS4ERR_DENIED:
5263 			status = 0;
5264 	}
5265 	request->fl_ops->fl_release_private(request);
5266 	request->fl_ops = NULL;
5267 out:
5268 	return status;
5269 }
5270 
5271 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5272 {
5273 	struct nfs4_exception exception = { };
5274 	int err;
5275 
5276 	do {
5277 		err = _nfs4_proc_getlk(state, cmd, request);
5278 		trace_nfs4_get_lock(request, state, cmd, err);
5279 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5280 				&exception);
5281 	} while (exception.retry);
5282 	return err;
5283 }
5284 
5285 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5286 {
5287 	int res = 0;
5288 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5289 		case FL_POSIX:
5290 			res = posix_lock_file_wait(file, fl);
5291 			break;
5292 		case FL_FLOCK:
5293 			res = flock_lock_file_wait(file, fl);
5294 			break;
5295 		default:
5296 			BUG();
5297 	}
5298 	return res;
5299 }
5300 
5301 struct nfs4_unlockdata {
5302 	struct nfs_locku_args arg;
5303 	struct nfs_locku_res res;
5304 	struct nfs4_lock_state *lsp;
5305 	struct nfs_open_context *ctx;
5306 	struct file_lock fl;
5307 	const struct nfs_server *server;
5308 	unsigned long timestamp;
5309 };
5310 
5311 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5312 		struct nfs_open_context *ctx,
5313 		struct nfs4_lock_state *lsp,
5314 		struct nfs_seqid *seqid)
5315 {
5316 	struct nfs4_unlockdata *p;
5317 	struct inode *inode = lsp->ls_state->inode;
5318 
5319 	p = kzalloc(sizeof(*p), GFP_NOFS);
5320 	if (p == NULL)
5321 		return NULL;
5322 	p->arg.fh = NFS_FH(inode);
5323 	p->arg.fl = &p->fl;
5324 	p->arg.seqid = seqid;
5325 	p->res.seqid = seqid;
5326 	p->arg.stateid = &lsp->ls_stateid;
5327 	p->lsp = lsp;
5328 	atomic_inc(&lsp->ls_count);
5329 	/* Ensure we don't close file until we're done freeing locks! */
5330 	p->ctx = get_nfs_open_context(ctx);
5331 	memcpy(&p->fl, fl, sizeof(p->fl));
5332 	p->server = NFS_SERVER(inode);
5333 	return p;
5334 }
5335 
5336 static void nfs4_locku_release_calldata(void *data)
5337 {
5338 	struct nfs4_unlockdata *calldata = data;
5339 	nfs_free_seqid(calldata->arg.seqid);
5340 	nfs4_put_lock_state(calldata->lsp);
5341 	put_nfs_open_context(calldata->ctx);
5342 	kfree(calldata);
5343 }
5344 
5345 static void nfs4_locku_done(struct rpc_task *task, void *data)
5346 {
5347 	struct nfs4_unlockdata *calldata = data;
5348 
5349 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5350 		return;
5351 	switch (task->tk_status) {
5352 		case 0:
5353 			nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5354 					&calldata->res.stateid);
5355 			renew_lease(calldata->server, calldata->timestamp);
5356 			break;
5357 		case -NFS4ERR_BAD_STATEID:
5358 		case -NFS4ERR_OLD_STATEID:
5359 		case -NFS4ERR_STALE_STATEID:
5360 		case -NFS4ERR_EXPIRED:
5361 			break;
5362 		default:
5363 			if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5364 				rpc_restart_call_prepare(task);
5365 	}
5366 	nfs_release_seqid(calldata->arg.seqid);
5367 }
5368 
5369 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5370 {
5371 	struct nfs4_unlockdata *calldata = data;
5372 
5373 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5374 		goto out_wait;
5375 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5376 		/* Note: exit _without_ running nfs4_locku_done */
5377 		goto out_no_action;
5378 	}
5379 	calldata->timestamp = jiffies;
5380 	if (nfs4_setup_sequence(calldata->server,
5381 				&calldata->arg.seq_args,
5382 				&calldata->res.seq_res,
5383 				task) != 0)
5384 		nfs_release_seqid(calldata->arg.seqid);
5385 	return;
5386 out_no_action:
5387 	task->tk_action = NULL;
5388 out_wait:
5389 	nfs4_sequence_done(task, &calldata->res.seq_res);
5390 }
5391 
5392 static const struct rpc_call_ops nfs4_locku_ops = {
5393 	.rpc_call_prepare = nfs4_locku_prepare,
5394 	.rpc_call_done = nfs4_locku_done,
5395 	.rpc_release = nfs4_locku_release_calldata,
5396 };
5397 
5398 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5399 		struct nfs_open_context *ctx,
5400 		struct nfs4_lock_state *lsp,
5401 		struct nfs_seqid *seqid)
5402 {
5403 	struct nfs4_unlockdata *data;
5404 	struct rpc_message msg = {
5405 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5406 		.rpc_cred = ctx->cred,
5407 	};
5408 	struct rpc_task_setup task_setup_data = {
5409 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5410 		.rpc_message = &msg,
5411 		.callback_ops = &nfs4_locku_ops,
5412 		.workqueue = nfsiod_workqueue,
5413 		.flags = RPC_TASK_ASYNC,
5414 	};
5415 
5416 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5417 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5418 
5419 	/* Ensure this is an unlock - when canceling a lock, the
5420 	 * canceled lock is passed in, and it won't be an unlock.
5421 	 */
5422 	fl->fl_type = F_UNLCK;
5423 
5424 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5425 	if (data == NULL) {
5426 		nfs_free_seqid(seqid);
5427 		return ERR_PTR(-ENOMEM);
5428 	}
5429 
5430 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5431 	msg.rpc_argp = &data->arg;
5432 	msg.rpc_resp = &data->res;
5433 	task_setup_data.callback_data = data;
5434 	return rpc_run_task(&task_setup_data);
5435 }
5436 
5437 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5438 {
5439 	struct inode *inode = state->inode;
5440 	struct nfs4_state_owner *sp = state->owner;
5441 	struct nfs_inode *nfsi = NFS_I(inode);
5442 	struct nfs_seqid *seqid;
5443 	struct nfs4_lock_state *lsp;
5444 	struct rpc_task *task;
5445 	int status = 0;
5446 	unsigned char fl_flags = request->fl_flags;
5447 
5448 	status = nfs4_set_lock_state(state, request);
5449 	/* Unlock _before_ we do the RPC call */
5450 	request->fl_flags |= FL_EXISTS;
5451 	/* Exclude nfs_delegation_claim_locks() */
5452 	mutex_lock(&sp->so_delegreturn_mutex);
5453 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5454 	down_read(&nfsi->rwsem);
5455 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5456 		up_read(&nfsi->rwsem);
5457 		mutex_unlock(&sp->so_delegreturn_mutex);
5458 		goto out;
5459 	}
5460 	up_read(&nfsi->rwsem);
5461 	mutex_unlock(&sp->so_delegreturn_mutex);
5462 	if (status != 0)
5463 		goto out;
5464 	/* Is this a delegated lock? */
5465 	lsp = request->fl_u.nfs4_fl.owner;
5466 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5467 		goto out;
5468 	seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5469 	status = -ENOMEM;
5470 	if (seqid == NULL)
5471 		goto out;
5472 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5473 	status = PTR_ERR(task);
5474 	if (IS_ERR(task))
5475 		goto out;
5476 	status = nfs4_wait_for_completion_rpc_task(task);
5477 	rpc_put_task(task);
5478 out:
5479 	request->fl_flags = fl_flags;
5480 	trace_nfs4_unlock(request, state, F_SETLK, status);
5481 	return status;
5482 }
5483 
5484 struct nfs4_lockdata {
5485 	struct nfs_lock_args arg;
5486 	struct nfs_lock_res res;
5487 	struct nfs4_lock_state *lsp;
5488 	struct nfs_open_context *ctx;
5489 	struct file_lock fl;
5490 	unsigned long timestamp;
5491 	int rpc_status;
5492 	int cancelled;
5493 	struct nfs_server *server;
5494 };
5495 
5496 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5497 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5498 		gfp_t gfp_mask)
5499 {
5500 	struct nfs4_lockdata *p;
5501 	struct inode *inode = lsp->ls_state->inode;
5502 	struct nfs_server *server = NFS_SERVER(inode);
5503 
5504 	p = kzalloc(sizeof(*p), gfp_mask);
5505 	if (p == NULL)
5506 		return NULL;
5507 
5508 	p->arg.fh = NFS_FH(inode);
5509 	p->arg.fl = &p->fl;
5510 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5511 	if (p->arg.open_seqid == NULL)
5512 		goto out_free;
5513 	p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5514 	if (p->arg.lock_seqid == NULL)
5515 		goto out_free_seqid;
5516 	p->arg.lock_stateid = &lsp->ls_stateid;
5517 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5518 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5519 	p->arg.lock_owner.s_dev = server->s_dev;
5520 	p->res.lock_seqid = p->arg.lock_seqid;
5521 	p->lsp = lsp;
5522 	p->server = server;
5523 	atomic_inc(&lsp->ls_count);
5524 	p->ctx = get_nfs_open_context(ctx);
5525 	memcpy(&p->fl, fl, sizeof(p->fl));
5526 	return p;
5527 out_free_seqid:
5528 	nfs_free_seqid(p->arg.open_seqid);
5529 out_free:
5530 	kfree(p);
5531 	return NULL;
5532 }
5533 
5534 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5535 {
5536 	struct nfs4_lockdata *data = calldata;
5537 	struct nfs4_state *state = data->lsp->ls_state;
5538 
5539 	dprintk("%s: begin!\n", __func__);
5540 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5541 		goto out_wait;
5542 	/* Do we need to do an open_to_lock_owner? */
5543 	if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5544 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5545 			goto out_release_lock_seqid;
5546 		}
5547 		data->arg.open_stateid = &state->open_stateid;
5548 		data->arg.new_lock_owner = 1;
5549 		data->res.open_seqid = data->arg.open_seqid;
5550 	} else
5551 		data->arg.new_lock_owner = 0;
5552 	if (!nfs4_valid_open_stateid(state)) {
5553 		data->rpc_status = -EBADF;
5554 		task->tk_action = NULL;
5555 		goto out_release_open_seqid;
5556 	}
5557 	data->timestamp = jiffies;
5558 	if (nfs4_setup_sequence(data->server,
5559 				&data->arg.seq_args,
5560 				&data->res.seq_res,
5561 				task) == 0)
5562 		return;
5563 out_release_open_seqid:
5564 	nfs_release_seqid(data->arg.open_seqid);
5565 out_release_lock_seqid:
5566 	nfs_release_seqid(data->arg.lock_seqid);
5567 out_wait:
5568 	nfs4_sequence_done(task, &data->res.seq_res);
5569 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5570 }
5571 
5572 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5573 {
5574 	struct nfs4_lockdata *data = calldata;
5575 
5576 	dprintk("%s: begin!\n", __func__);
5577 
5578 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5579 		return;
5580 
5581 	data->rpc_status = task->tk_status;
5582 	if (data->arg.new_lock_owner != 0) {
5583 		if (data->rpc_status == 0)
5584 			nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5585 		else
5586 			goto out;
5587 	}
5588 	if (data->rpc_status == 0) {
5589 		nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5590 		set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5591 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5592 	}
5593 out:
5594 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5595 }
5596 
5597 static void nfs4_lock_release(void *calldata)
5598 {
5599 	struct nfs4_lockdata *data = calldata;
5600 
5601 	dprintk("%s: begin!\n", __func__);
5602 	nfs_free_seqid(data->arg.open_seqid);
5603 	if (data->cancelled != 0) {
5604 		struct rpc_task *task;
5605 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5606 				data->arg.lock_seqid);
5607 		if (!IS_ERR(task))
5608 			rpc_put_task_async(task);
5609 		dprintk("%s: cancelling lock!\n", __func__);
5610 	} else
5611 		nfs_free_seqid(data->arg.lock_seqid);
5612 	nfs4_put_lock_state(data->lsp);
5613 	put_nfs_open_context(data->ctx);
5614 	kfree(data);
5615 	dprintk("%s: done!\n", __func__);
5616 }
5617 
5618 static const struct rpc_call_ops nfs4_lock_ops = {
5619 	.rpc_call_prepare = nfs4_lock_prepare,
5620 	.rpc_call_done = nfs4_lock_done,
5621 	.rpc_release = nfs4_lock_release,
5622 };
5623 
5624 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5625 {
5626 	switch (error) {
5627 	case -NFS4ERR_ADMIN_REVOKED:
5628 	case -NFS4ERR_BAD_STATEID:
5629 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5630 		if (new_lock_owner != 0 ||
5631 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5632 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5633 		break;
5634 	case -NFS4ERR_STALE_STATEID:
5635 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5636 	case -NFS4ERR_EXPIRED:
5637 		nfs4_schedule_lease_recovery(server->nfs_client);
5638 	};
5639 }
5640 
5641 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5642 {
5643 	struct nfs4_lockdata *data;
5644 	struct rpc_task *task;
5645 	struct rpc_message msg = {
5646 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5647 		.rpc_cred = state->owner->so_cred,
5648 	};
5649 	struct rpc_task_setup task_setup_data = {
5650 		.rpc_client = NFS_CLIENT(state->inode),
5651 		.rpc_message = &msg,
5652 		.callback_ops = &nfs4_lock_ops,
5653 		.workqueue = nfsiod_workqueue,
5654 		.flags = RPC_TASK_ASYNC,
5655 	};
5656 	int ret;
5657 
5658 	dprintk("%s: begin!\n", __func__);
5659 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5660 			fl->fl_u.nfs4_fl.owner,
5661 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5662 	if (data == NULL)
5663 		return -ENOMEM;
5664 	if (IS_SETLKW(cmd))
5665 		data->arg.block = 1;
5666 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5667 	msg.rpc_argp = &data->arg;
5668 	msg.rpc_resp = &data->res;
5669 	task_setup_data.callback_data = data;
5670 	if (recovery_type > NFS_LOCK_NEW) {
5671 		if (recovery_type == NFS_LOCK_RECLAIM)
5672 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5673 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5674 	}
5675 	task = rpc_run_task(&task_setup_data);
5676 	if (IS_ERR(task))
5677 		return PTR_ERR(task);
5678 	ret = nfs4_wait_for_completion_rpc_task(task);
5679 	if (ret == 0) {
5680 		ret = data->rpc_status;
5681 		if (ret)
5682 			nfs4_handle_setlk_error(data->server, data->lsp,
5683 					data->arg.new_lock_owner, ret);
5684 	} else
5685 		data->cancelled = 1;
5686 	rpc_put_task(task);
5687 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5688 	return ret;
5689 }
5690 
5691 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5692 {
5693 	struct nfs_server *server = NFS_SERVER(state->inode);
5694 	struct nfs4_exception exception = {
5695 		.inode = state->inode,
5696 	};
5697 	int err;
5698 
5699 	do {
5700 		/* Cache the lock if possible... */
5701 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5702 			return 0;
5703 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5704 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5705 		if (err != -NFS4ERR_DELAY)
5706 			break;
5707 		nfs4_handle_exception(server, err, &exception);
5708 	} while (exception.retry);
5709 	return err;
5710 }
5711 
5712 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5713 {
5714 	struct nfs_server *server = NFS_SERVER(state->inode);
5715 	struct nfs4_exception exception = {
5716 		.inode = state->inode,
5717 	};
5718 	int err;
5719 
5720 	err = nfs4_set_lock_state(state, request);
5721 	if (err != 0)
5722 		return err;
5723 	if (!recover_lost_locks) {
5724 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5725 		return 0;
5726 	}
5727 	do {
5728 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5729 			return 0;
5730 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5731 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5732 		switch (err) {
5733 		default:
5734 			goto out;
5735 		case -NFS4ERR_GRACE:
5736 		case -NFS4ERR_DELAY:
5737 			nfs4_handle_exception(server, err, &exception);
5738 			err = 0;
5739 		}
5740 	} while (exception.retry);
5741 out:
5742 	return err;
5743 }
5744 
5745 #if defined(CONFIG_NFS_V4_1)
5746 /**
5747  * nfs41_check_expired_locks - possibly free a lock stateid
5748  *
5749  * @state: NFSv4 state for an inode
5750  *
5751  * Returns NFS_OK if recovery for this stateid is now finished.
5752  * Otherwise a negative NFS4ERR value is returned.
5753  */
5754 static int nfs41_check_expired_locks(struct nfs4_state *state)
5755 {
5756 	int status, ret = -NFS4ERR_BAD_STATEID;
5757 	struct nfs4_lock_state *lsp;
5758 	struct nfs_server *server = NFS_SERVER(state->inode);
5759 
5760 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5761 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5762 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5763 
5764 			status = nfs41_test_stateid(server,
5765 					&lsp->ls_stateid,
5766 					cred);
5767 			trace_nfs4_test_lock_stateid(state, lsp, status);
5768 			if (status != NFS_OK) {
5769 				/* Free the stateid unless the server
5770 				 * informs us the stateid is unrecognized. */
5771 				if (status != -NFS4ERR_BAD_STATEID)
5772 					nfs41_free_stateid(server,
5773 							&lsp->ls_stateid,
5774 							cred);
5775 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5776 				ret = status;
5777 			}
5778 		}
5779 	};
5780 
5781 	return ret;
5782 }
5783 
5784 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5785 {
5786 	int status = NFS_OK;
5787 
5788 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5789 		status = nfs41_check_expired_locks(state);
5790 	if (status != NFS_OK)
5791 		status = nfs4_lock_expired(state, request);
5792 	return status;
5793 }
5794 #endif
5795 
5796 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5797 {
5798 	struct nfs4_state_owner *sp = state->owner;
5799 	struct nfs_inode *nfsi = NFS_I(state->inode);
5800 	unsigned char fl_flags = request->fl_flags;
5801 	unsigned int seq;
5802 	int status = -ENOLCK;
5803 
5804 	if ((fl_flags & FL_POSIX) &&
5805 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5806 		goto out;
5807 	/* Is this a delegated open? */
5808 	status = nfs4_set_lock_state(state, request);
5809 	if (status != 0)
5810 		goto out;
5811 	request->fl_flags |= FL_ACCESS;
5812 	status = do_vfs_lock(request->fl_file, request);
5813 	if (status < 0)
5814 		goto out;
5815 	down_read(&nfsi->rwsem);
5816 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5817 		/* Yes: cache locks! */
5818 		/* ...but avoid races with delegation recall... */
5819 		request->fl_flags = fl_flags & ~FL_SLEEP;
5820 		status = do_vfs_lock(request->fl_file, request);
5821 		goto out_unlock;
5822 	}
5823 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5824 	up_read(&nfsi->rwsem);
5825 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5826 	if (status != 0)
5827 		goto out;
5828 	down_read(&nfsi->rwsem);
5829 	if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5830 		status = -NFS4ERR_DELAY;
5831 		goto out_unlock;
5832 	}
5833 	/* Note: we always want to sleep here! */
5834 	request->fl_flags = fl_flags | FL_SLEEP;
5835 	if (do_vfs_lock(request->fl_file, request) < 0)
5836 		printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5837 			"manager!\n", __func__);
5838 out_unlock:
5839 	up_read(&nfsi->rwsem);
5840 out:
5841 	request->fl_flags = fl_flags;
5842 	return status;
5843 }
5844 
5845 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5846 {
5847 	struct nfs4_exception exception = {
5848 		.state = state,
5849 		.inode = state->inode,
5850 	};
5851 	int err;
5852 
5853 	do {
5854 		err = _nfs4_proc_setlk(state, cmd, request);
5855 		trace_nfs4_set_lock(request, state, cmd, err);
5856 		if (err == -NFS4ERR_DENIED)
5857 			err = -EAGAIN;
5858 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5859 				err, &exception);
5860 	} while (exception.retry);
5861 	return err;
5862 }
5863 
5864 static int
5865 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5866 {
5867 	struct nfs_open_context *ctx;
5868 	struct nfs4_state *state;
5869 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5870 	int status;
5871 
5872 	/* verify open state */
5873 	ctx = nfs_file_open_context(filp);
5874 	state = ctx->state;
5875 
5876 	if (request->fl_start < 0 || request->fl_end < 0)
5877 		return -EINVAL;
5878 
5879 	if (IS_GETLK(cmd)) {
5880 		if (state != NULL)
5881 			return nfs4_proc_getlk(state, F_GETLK, request);
5882 		return 0;
5883 	}
5884 
5885 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5886 		return -EINVAL;
5887 
5888 	if (request->fl_type == F_UNLCK) {
5889 		if (state != NULL)
5890 			return nfs4_proc_unlck(state, cmd, request);
5891 		return 0;
5892 	}
5893 
5894 	if (state == NULL)
5895 		return -ENOLCK;
5896 	/*
5897 	 * Don't rely on the VFS having checked the file open mode,
5898 	 * since it won't do this for flock() locks.
5899 	 */
5900 	switch (request->fl_type) {
5901 	case F_RDLCK:
5902 		if (!(filp->f_mode & FMODE_READ))
5903 			return -EBADF;
5904 		break;
5905 	case F_WRLCK:
5906 		if (!(filp->f_mode & FMODE_WRITE))
5907 			return -EBADF;
5908 	}
5909 
5910 	do {
5911 		status = nfs4_proc_setlk(state, cmd, request);
5912 		if ((status != -EAGAIN) || IS_SETLK(cmd))
5913 			break;
5914 		timeout = nfs4_set_lock_task_retry(timeout);
5915 		status = -ERESTARTSYS;
5916 		if (signalled())
5917 			break;
5918 	} while(status < 0);
5919 	return status;
5920 }
5921 
5922 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5923 {
5924 	struct nfs_server *server = NFS_SERVER(state->inode);
5925 	int err;
5926 
5927 	err = nfs4_set_lock_state(state, fl);
5928 	if (err != 0)
5929 		return err;
5930 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5931 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5932 }
5933 
5934 struct nfs_release_lockowner_data {
5935 	struct nfs4_lock_state *lsp;
5936 	struct nfs_server *server;
5937 	struct nfs_release_lockowner_args args;
5938 	struct nfs_release_lockowner_res res;
5939 	unsigned long timestamp;
5940 };
5941 
5942 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5943 {
5944 	struct nfs_release_lockowner_data *data = calldata;
5945 	struct nfs_server *server = data->server;
5946 	nfs40_setup_sequence(server, &data->args.seq_args,
5947 				&data->res.seq_res, task);
5948 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5949 	data->timestamp = jiffies;
5950 }
5951 
5952 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5953 {
5954 	struct nfs_release_lockowner_data *data = calldata;
5955 	struct nfs_server *server = data->server;
5956 
5957 	nfs40_sequence_done(task, &data->res.seq_res);
5958 
5959 	switch (task->tk_status) {
5960 	case 0:
5961 		renew_lease(server, data->timestamp);
5962 		break;
5963 	case -NFS4ERR_STALE_CLIENTID:
5964 	case -NFS4ERR_EXPIRED:
5965 		nfs4_schedule_lease_recovery(server->nfs_client);
5966 		break;
5967 	case -NFS4ERR_LEASE_MOVED:
5968 	case -NFS4ERR_DELAY:
5969 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN)
5970 			rpc_restart_call_prepare(task);
5971 	}
5972 }
5973 
5974 static void nfs4_release_lockowner_release(void *calldata)
5975 {
5976 	struct nfs_release_lockowner_data *data = calldata;
5977 	nfs4_free_lock_state(data->server, data->lsp);
5978 	kfree(calldata);
5979 }
5980 
5981 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5982 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
5983 	.rpc_call_done = nfs4_release_lockowner_done,
5984 	.rpc_release = nfs4_release_lockowner_release,
5985 };
5986 
5987 static void
5988 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5989 {
5990 	struct nfs_release_lockowner_data *data;
5991 	struct rpc_message msg = {
5992 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5993 	};
5994 
5995 	if (server->nfs_client->cl_mvops->minor_version != 0)
5996 		return;
5997 
5998 	data = kmalloc(sizeof(*data), GFP_NOFS);
5999 	if (!data)
6000 		return;
6001 	data->lsp = lsp;
6002 	data->server = server;
6003 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6004 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6005 	data->args.lock_owner.s_dev = server->s_dev;
6006 
6007 	msg.rpc_argp = &data->args;
6008 	msg.rpc_resp = &data->res;
6009 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6010 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6011 }
6012 
6013 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6014 
6015 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6016 				   const void *buf, size_t buflen,
6017 				   int flags, int type)
6018 {
6019 	if (strcmp(key, "") != 0)
6020 		return -EINVAL;
6021 
6022 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6023 }
6024 
6025 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6026 				   void *buf, size_t buflen, int type)
6027 {
6028 	if (strcmp(key, "") != 0)
6029 		return -EINVAL;
6030 
6031 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6032 }
6033 
6034 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6035 				       size_t list_len, const char *name,
6036 				       size_t name_len, int type)
6037 {
6038 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6039 
6040 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6041 		return 0;
6042 
6043 	if (list && len <= list_len)
6044 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6045 	return len;
6046 }
6047 
6048 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6049 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6050 {
6051 	return server->caps & NFS_CAP_SECURITY_LABEL;
6052 }
6053 
6054 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6055 				   const void *buf, size_t buflen,
6056 				   int flags, int type)
6057 {
6058 	if (security_ismaclabel(key))
6059 		return nfs4_set_security_label(dentry, buf, buflen);
6060 
6061 	return -EOPNOTSUPP;
6062 }
6063 
6064 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6065 				   void *buf, size_t buflen, int type)
6066 {
6067 	if (security_ismaclabel(key))
6068 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6069 	return -EOPNOTSUPP;
6070 }
6071 
6072 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6073 				       size_t list_len, const char *name,
6074 				       size_t name_len, int type)
6075 {
6076 	size_t len = 0;
6077 
6078 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6079 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6080 		if (list && len <= list_len)
6081 			security_inode_listsecurity(dentry->d_inode, list, len);
6082 	}
6083 	return len;
6084 }
6085 
6086 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6087 	.prefix = XATTR_SECURITY_PREFIX,
6088 	.list	= nfs4_xattr_list_nfs4_label,
6089 	.get	= nfs4_xattr_get_nfs4_label,
6090 	.set	= nfs4_xattr_set_nfs4_label,
6091 };
6092 #endif
6093 
6094 
6095 /*
6096  * nfs_fhget will use either the mounted_on_fileid or the fileid
6097  */
6098 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6099 {
6100 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6101 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6102 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6103 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6104 		return;
6105 
6106 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6107 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6108 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6109 	fattr->nlink = 2;
6110 }
6111 
6112 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6113 				   const struct qstr *name,
6114 				   struct nfs4_fs_locations *fs_locations,
6115 				   struct page *page)
6116 {
6117 	struct nfs_server *server = NFS_SERVER(dir);
6118 	u32 bitmask[3] = {
6119 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6120 	};
6121 	struct nfs4_fs_locations_arg args = {
6122 		.dir_fh = NFS_FH(dir),
6123 		.name = name,
6124 		.page = page,
6125 		.bitmask = bitmask,
6126 	};
6127 	struct nfs4_fs_locations_res res = {
6128 		.fs_locations = fs_locations,
6129 	};
6130 	struct rpc_message msg = {
6131 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6132 		.rpc_argp = &args,
6133 		.rpc_resp = &res,
6134 	};
6135 	int status;
6136 
6137 	dprintk("%s: start\n", __func__);
6138 
6139 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6140 	 * is not supported */
6141 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6142 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6143 	else
6144 		bitmask[0] |= FATTR4_WORD0_FILEID;
6145 
6146 	nfs_fattr_init(&fs_locations->fattr);
6147 	fs_locations->server = server;
6148 	fs_locations->nlocations = 0;
6149 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6150 	dprintk("%s: returned status = %d\n", __func__, status);
6151 	return status;
6152 }
6153 
6154 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6155 			   const struct qstr *name,
6156 			   struct nfs4_fs_locations *fs_locations,
6157 			   struct page *page)
6158 {
6159 	struct nfs4_exception exception = { };
6160 	int err;
6161 	do {
6162 		err = _nfs4_proc_fs_locations(client, dir, name,
6163 				fs_locations, page);
6164 		trace_nfs4_get_fs_locations(dir, name, err);
6165 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6166 				&exception);
6167 	} while (exception.retry);
6168 	return err;
6169 }
6170 
6171 /*
6172  * This operation also signals the server that this client is
6173  * performing migration recovery.  The server can stop returning
6174  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6175  * appended to this compound to identify the client ID which is
6176  * performing recovery.
6177  */
6178 static int _nfs40_proc_get_locations(struct inode *inode,
6179 				     struct nfs4_fs_locations *locations,
6180 				     struct page *page, struct rpc_cred *cred)
6181 {
6182 	struct nfs_server *server = NFS_SERVER(inode);
6183 	struct rpc_clnt *clnt = server->client;
6184 	u32 bitmask[2] = {
6185 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6186 	};
6187 	struct nfs4_fs_locations_arg args = {
6188 		.clientid	= server->nfs_client->cl_clientid,
6189 		.fh		= NFS_FH(inode),
6190 		.page		= page,
6191 		.bitmask	= bitmask,
6192 		.migration	= 1,		/* skip LOOKUP */
6193 		.renew		= 1,		/* append RENEW */
6194 	};
6195 	struct nfs4_fs_locations_res res = {
6196 		.fs_locations	= locations,
6197 		.migration	= 1,
6198 		.renew		= 1,
6199 	};
6200 	struct rpc_message msg = {
6201 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6202 		.rpc_argp	= &args,
6203 		.rpc_resp	= &res,
6204 		.rpc_cred	= cred,
6205 	};
6206 	unsigned long now = jiffies;
6207 	int status;
6208 
6209 	nfs_fattr_init(&locations->fattr);
6210 	locations->server = server;
6211 	locations->nlocations = 0;
6212 
6213 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6214 	nfs4_set_sequence_privileged(&args.seq_args);
6215 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6216 					&args.seq_args, &res.seq_res);
6217 	if (status)
6218 		return status;
6219 
6220 	renew_lease(server, now);
6221 	return 0;
6222 }
6223 
6224 #ifdef CONFIG_NFS_V4_1
6225 
6226 /*
6227  * This operation also signals the server that this client is
6228  * performing migration recovery.  The server can stop asserting
6229  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6230  * performing this operation is identified in the SEQUENCE
6231  * operation in this compound.
6232  *
6233  * When the client supports GETATTR(fs_locations_info), it can
6234  * be plumbed in here.
6235  */
6236 static int _nfs41_proc_get_locations(struct inode *inode,
6237 				     struct nfs4_fs_locations *locations,
6238 				     struct page *page, struct rpc_cred *cred)
6239 {
6240 	struct nfs_server *server = NFS_SERVER(inode);
6241 	struct rpc_clnt *clnt = server->client;
6242 	u32 bitmask[2] = {
6243 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6244 	};
6245 	struct nfs4_fs_locations_arg args = {
6246 		.fh		= NFS_FH(inode),
6247 		.page		= page,
6248 		.bitmask	= bitmask,
6249 		.migration	= 1,		/* skip LOOKUP */
6250 	};
6251 	struct nfs4_fs_locations_res res = {
6252 		.fs_locations	= locations,
6253 		.migration	= 1,
6254 	};
6255 	struct rpc_message msg = {
6256 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6257 		.rpc_argp	= &args,
6258 		.rpc_resp	= &res,
6259 		.rpc_cred	= cred,
6260 	};
6261 	int status;
6262 
6263 	nfs_fattr_init(&locations->fattr);
6264 	locations->server = server;
6265 	locations->nlocations = 0;
6266 
6267 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6268 	nfs4_set_sequence_privileged(&args.seq_args);
6269 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6270 					&args.seq_args, &res.seq_res);
6271 	if (status == NFS4_OK &&
6272 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6273 		status = -NFS4ERR_LEASE_MOVED;
6274 	return status;
6275 }
6276 
6277 #endif	/* CONFIG_NFS_V4_1 */
6278 
6279 /**
6280  * nfs4_proc_get_locations - discover locations for a migrated FSID
6281  * @inode: inode on FSID that is migrating
6282  * @locations: result of query
6283  * @page: buffer
6284  * @cred: credential to use for this operation
6285  *
6286  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6287  * operation failed, or a negative errno if a local error occurred.
6288  *
6289  * On success, "locations" is filled in, but if the server has
6290  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6291  * asserted.
6292  *
6293  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6294  * from this client that require migration recovery.
6295  */
6296 int nfs4_proc_get_locations(struct inode *inode,
6297 			    struct nfs4_fs_locations *locations,
6298 			    struct page *page, struct rpc_cred *cred)
6299 {
6300 	struct nfs_server *server = NFS_SERVER(inode);
6301 	struct nfs_client *clp = server->nfs_client;
6302 	const struct nfs4_mig_recovery_ops *ops =
6303 					clp->cl_mvops->mig_recovery_ops;
6304 	struct nfs4_exception exception = { };
6305 	int status;
6306 
6307 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6308 		(unsigned long long)server->fsid.major,
6309 		(unsigned long long)server->fsid.minor,
6310 		clp->cl_hostname);
6311 	nfs_display_fhandle(NFS_FH(inode), __func__);
6312 
6313 	do {
6314 		status = ops->get_locations(inode, locations, page, cred);
6315 		if (status != -NFS4ERR_DELAY)
6316 			break;
6317 		nfs4_handle_exception(server, status, &exception);
6318 	} while (exception.retry);
6319 	return status;
6320 }
6321 
6322 /*
6323  * This operation also signals the server that this client is
6324  * performing "lease moved" recovery.  The server can stop
6325  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6326  * is appended to this compound to identify the client ID which is
6327  * performing recovery.
6328  */
6329 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6330 {
6331 	struct nfs_server *server = NFS_SERVER(inode);
6332 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6333 	struct rpc_clnt *clnt = server->client;
6334 	struct nfs4_fsid_present_arg args = {
6335 		.fh		= NFS_FH(inode),
6336 		.clientid	= clp->cl_clientid,
6337 		.renew		= 1,		/* append RENEW */
6338 	};
6339 	struct nfs4_fsid_present_res res = {
6340 		.renew		= 1,
6341 	};
6342 	struct rpc_message msg = {
6343 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6344 		.rpc_argp	= &args,
6345 		.rpc_resp	= &res,
6346 		.rpc_cred	= cred,
6347 	};
6348 	unsigned long now = jiffies;
6349 	int status;
6350 
6351 	res.fh = nfs_alloc_fhandle();
6352 	if (res.fh == NULL)
6353 		return -ENOMEM;
6354 
6355 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6356 	nfs4_set_sequence_privileged(&args.seq_args);
6357 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6358 						&args.seq_args, &res.seq_res);
6359 	nfs_free_fhandle(res.fh);
6360 	if (status)
6361 		return status;
6362 
6363 	do_renew_lease(clp, now);
6364 	return 0;
6365 }
6366 
6367 #ifdef CONFIG_NFS_V4_1
6368 
6369 /*
6370  * This operation also signals the server that this client is
6371  * performing "lease moved" recovery.  The server can stop asserting
6372  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6373  * this operation is identified in the SEQUENCE operation in this
6374  * compound.
6375  */
6376 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6377 {
6378 	struct nfs_server *server = NFS_SERVER(inode);
6379 	struct rpc_clnt *clnt = server->client;
6380 	struct nfs4_fsid_present_arg args = {
6381 		.fh		= NFS_FH(inode),
6382 	};
6383 	struct nfs4_fsid_present_res res = {
6384 	};
6385 	struct rpc_message msg = {
6386 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6387 		.rpc_argp	= &args,
6388 		.rpc_resp	= &res,
6389 		.rpc_cred	= cred,
6390 	};
6391 	int status;
6392 
6393 	res.fh = nfs_alloc_fhandle();
6394 	if (res.fh == NULL)
6395 		return -ENOMEM;
6396 
6397 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6398 	nfs4_set_sequence_privileged(&args.seq_args);
6399 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6400 						&args.seq_args, &res.seq_res);
6401 	nfs_free_fhandle(res.fh);
6402 	if (status == NFS4_OK &&
6403 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6404 		status = -NFS4ERR_LEASE_MOVED;
6405 	return status;
6406 }
6407 
6408 #endif	/* CONFIG_NFS_V4_1 */
6409 
6410 /**
6411  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6412  * @inode: inode on FSID to check
6413  * @cred: credential to use for this operation
6414  *
6415  * Server indicates whether the FSID is present, moved, or not
6416  * recognized.  This operation is necessary to clear a LEASE_MOVED
6417  * condition for this client ID.
6418  *
6419  * Returns NFS4_OK if the FSID is present on this server,
6420  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6421  *  NFS4ERR code if some error occurred on the server, or a
6422  *  negative errno if a local failure occurred.
6423  */
6424 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6425 {
6426 	struct nfs_server *server = NFS_SERVER(inode);
6427 	struct nfs_client *clp = server->nfs_client;
6428 	const struct nfs4_mig_recovery_ops *ops =
6429 					clp->cl_mvops->mig_recovery_ops;
6430 	struct nfs4_exception exception = { };
6431 	int status;
6432 
6433 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6434 		(unsigned long long)server->fsid.major,
6435 		(unsigned long long)server->fsid.minor,
6436 		clp->cl_hostname);
6437 	nfs_display_fhandle(NFS_FH(inode), __func__);
6438 
6439 	do {
6440 		status = ops->fsid_present(inode, cred);
6441 		if (status != -NFS4ERR_DELAY)
6442 			break;
6443 		nfs4_handle_exception(server, status, &exception);
6444 	} while (exception.retry);
6445 	return status;
6446 }
6447 
6448 /**
6449  * If 'use_integrity' is true and the state managment nfs_client
6450  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6451  * and the machine credential as per RFC3530bis and RFC5661 Security
6452  * Considerations sections. Otherwise, just use the user cred with the
6453  * filesystem's rpc_client.
6454  */
6455 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6456 {
6457 	int status;
6458 	struct nfs4_secinfo_arg args = {
6459 		.dir_fh = NFS_FH(dir),
6460 		.name   = name,
6461 	};
6462 	struct nfs4_secinfo_res res = {
6463 		.flavors     = flavors,
6464 	};
6465 	struct rpc_message msg = {
6466 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6467 		.rpc_argp = &args,
6468 		.rpc_resp = &res,
6469 	};
6470 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6471 	struct rpc_cred *cred = NULL;
6472 
6473 	if (use_integrity) {
6474 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6475 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6476 		msg.rpc_cred = cred;
6477 	}
6478 
6479 	dprintk("NFS call  secinfo %s\n", name->name);
6480 
6481 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6482 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6483 
6484 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6485 				&res.seq_res, 0);
6486 	dprintk("NFS reply  secinfo: %d\n", status);
6487 
6488 	if (cred)
6489 		put_rpccred(cred);
6490 
6491 	return status;
6492 }
6493 
6494 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6495 		      struct nfs4_secinfo_flavors *flavors)
6496 {
6497 	struct nfs4_exception exception = { };
6498 	int err;
6499 	do {
6500 		err = -NFS4ERR_WRONGSEC;
6501 
6502 		/* try to use integrity protection with machine cred */
6503 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6504 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6505 
6506 		/*
6507 		 * if unable to use integrity protection, or SECINFO with
6508 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6509 		 * disallowed by spec, but exists in deployed servers) use
6510 		 * the current filesystem's rpc_client and the user cred.
6511 		 */
6512 		if (err == -NFS4ERR_WRONGSEC)
6513 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6514 
6515 		trace_nfs4_secinfo(dir, name, err);
6516 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6517 				&exception);
6518 	} while (exception.retry);
6519 	return err;
6520 }
6521 
6522 #ifdef CONFIG_NFS_V4_1
6523 /*
6524  * Check the exchange flags returned by the server for invalid flags, having
6525  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6526  * DS flags set.
6527  */
6528 static int nfs4_check_cl_exchange_flags(u32 flags)
6529 {
6530 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6531 		goto out_inval;
6532 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6533 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6534 		goto out_inval;
6535 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6536 		goto out_inval;
6537 	return NFS_OK;
6538 out_inval:
6539 	return -NFS4ERR_INVAL;
6540 }
6541 
6542 static bool
6543 nfs41_same_server_scope(struct nfs41_server_scope *a,
6544 			struct nfs41_server_scope *b)
6545 {
6546 	if (a->server_scope_sz == b->server_scope_sz &&
6547 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6548 		return true;
6549 
6550 	return false;
6551 }
6552 
6553 /*
6554  * nfs4_proc_bind_conn_to_session()
6555  *
6556  * The 4.1 client currently uses the same TCP connection for the
6557  * fore and backchannel.
6558  */
6559 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6560 {
6561 	int status;
6562 	struct nfs41_bind_conn_to_session_res res;
6563 	struct rpc_message msg = {
6564 		.rpc_proc =
6565 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6566 		.rpc_argp = clp,
6567 		.rpc_resp = &res,
6568 		.rpc_cred = cred,
6569 	};
6570 
6571 	dprintk("--> %s\n", __func__);
6572 
6573 	res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6574 	if (unlikely(res.session == NULL)) {
6575 		status = -ENOMEM;
6576 		goto out;
6577 	}
6578 
6579 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6580 	trace_nfs4_bind_conn_to_session(clp, status);
6581 	if (status == 0) {
6582 		if (memcmp(res.session->sess_id.data,
6583 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6584 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6585 			status = -EIO;
6586 			goto out_session;
6587 		}
6588 		if (res.dir != NFS4_CDFS4_BOTH) {
6589 			dprintk("NFS: %s: Unexpected direction from server\n",
6590 				__func__);
6591 			status = -EIO;
6592 			goto out_session;
6593 		}
6594 		if (res.use_conn_in_rdma_mode) {
6595 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6596 				__func__);
6597 			status = -EIO;
6598 			goto out_session;
6599 		}
6600 	}
6601 out_session:
6602 	kfree(res.session);
6603 out:
6604 	dprintk("<-- %s status= %d\n", __func__, status);
6605 	return status;
6606 }
6607 
6608 /*
6609  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6610  * and operations we'd like to see to enable certain features in the allow map
6611  */
6612 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6613 	.how = SP4_MACH_CRED,
6614 	.enforce.u.words = {
6615 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6616 		      1 << (OP_EXCHANGE_ID - 32) |
6617 		      1 << (OP_CREATE_SESSION - 32) |
6618 		      1 << (OP_DESTROY_SESSION - 32) |
6619 		      1 << (OP_DESTROY_CLIENTID - 32)
6620 	},
6621 	.allow.u.words = {
6622 		[0] = 1 << (OP_CLOSE) |
6623 		      1 << (OP_LOCKU) |
6624 		      1 << (OP_COMMIT),
6625 		[1] = 1 << (OP_SECINFO - 32) |
6626 		      1 << (OP_SECINFO_NO_NAME - 32) |
6627 		      1 << (OP_TEST_STATEID - 32) |
6628 		      1 << (OP_FREE_STATEID - 32) |
6629 		      1 << (OP_WRITE - 32)
6630 	}
6631 };
6632 
6633 /*
6634  * Select the state protection mode for client `clp' given the server results
6635  * from exchange_id in `sp'.
6636  *
6637  * Returns 0 on success, negative errno otherwise.
6638  */
6639 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6640 				 struct nfs41_state_protection *sp)
6641 {
6642 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6643 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6644 		      1 << (OP_EXCHANGE_ID - 32) |
6645 		      1 << (OP_CREATE_SESSION - 32) |
6646 		      1 << (OP_DESTROY_SESSION - 32) |
6647 		      1 << (OP_DESTROY_CLIENTID - 32)
6648 	};
6649 	unsigned int i;
6650 
6651 	if (sp->how == SP4_MACH_CRED) {
6652 		/* Print state protect result */
6653 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6654 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6655 			if (test_bit(i, sp->enforce.u.longs))
6656 				dfprintk(MOUNT, "  enforce op %d\n", i);
6657 			if (test_bit(i, sp->allow.u.longs))
6658 				dfprintk(MOUNT, "  allow op %d\n", i);
6659 		}
6660 
6661 		/* make sure nothing is on enforce list that isn't supported */
6662 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6663 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6664 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6665 				return -EINVAL;
6666 			}
6667 		}
6668 
6669 		/*
6670 		 * Minimal mode - state operations are allowed to use machine
6671 		 * credential.  Note this already happens by default, so the
6672 		 * client doesn't have to do anything more than the negotiation.
6673 		 *
6674 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6675 		 *       we're already using the machine cred for exchange_id
6676 		 *       and will never use a different cred.
6677 		 */
6678 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6679 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6680 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6681 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6682 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6683 			dfprintk(MOUNT, "  minimal mode enabled\n");
6684 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6685 		} else {
6686 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6687 			return -EINVAL;
6688 		}
6689 
6690 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6691 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6692 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6693 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6694 		}
6695 
6696 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6697 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6698 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6699 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6700 		}
6701 
6702 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6703 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6704 			dfprintk(MOUNT, "  stateid mode enabled\n");
6705 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6706 		}
6707 
6708 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6709 			dfprintk(MOUNT, "  write mode enabled\n");
6710 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6711 		}
6712 
6713 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6714 			dfprintk(MOUNT, "  commit mode enabled\n");
6715 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6716 		}
6717 	}
6718 
6719 	return 0;
6720 }
6721 
6722 /*
6723  * _nfs4_proc_exchange_id()
6724  *
6725  * Wrapper for EXCHANGE_ID operation.
6726  */
6727 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6728 	u32 sp4_how)
6729 {
6730 	nfs4_verifier verifier;
6731 	struct nfs41_exchange_id_args args = {
6732 		.verifier = &verifier,
6733 		.client = clp,
6734 #ifdef CONFIG_NFS_V4_1_MIGRATION
6735 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6736 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6737 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6738 #else
6739 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6740 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6741 #endif
6742 	};
6743 	struct nfs41_exchange_id_res res = {
6744 		0
6745 	};
6746 	int status;
6747 	struct rpc_message msg = {
6748 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6749 		.rpc_argp = &args,
6750 		.rpc_resp = &res,
6751 		.rpc_cred = cred,
6752 	};
6753 
6754 	nfs4_init_boot_verifier(clp, &verifier);
6755 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6756 							sizeof(args.id));
6757 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6758 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6759 		args.id_len, args.id);
6760 
6761 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6762 					GFP_NOFS);
6763 	if (unlikely(res.server_owner == NULL)) {
6764 		status = -ENOMEM;
6765 		goto out;
6766 	}
6767 
6768 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6769 					GFP_NOFS);
6770 	if (unlikely(res.server_scope == NULL)) {
6771 		status = -ENOMEM;
6772 		goto out_server_owner;
6773 	}
6774 
6775 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6776 	if (unlikely(res.impl_id == NULL)) {
6777 		status = -ENOMEM;
6778 		goto out_server_scope;
6779 	}
6780 
6781 	switch (sp4_how) {
6782 	case SP4_NONE:
6783 		args.state_protect.how = SP4_NONE;
6784 		break;
6785 
6786 	case SP4_MACH_CRED:
6787 		args.state_protect = nfs4_sp4_mach_cred_request;
6788 		break;
6789 
6790 	default:
6791 		/* unsupported! */
6792 		WARN_ON_ONCE(1);
6793 		status = -EINVAL;
6794 		goto out_server_scope;
6795 	}
6796 
6797 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6798 	trace_nfs4_exchange_id(clp, status);
6799 	if (status == 0)
6800 		status = nfs4_check_cl_exchange_flags(res.flags);
6801 
6802 	if (status == 0)
6803 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6804 
6805 	if (status == 0) {
6806 		clp->cl_clientid = res.clientid;
6807 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6808 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6809 			clp->cl_seqid = res.seqid;
6810 
6811 		kfree(clp->cl_serverowner);
6812 		clp->cl_serverowner = res.server_owner;
6813 		res.server_owner = NULL;
6814 
6815 		/* use the most recent implementation id */
6816 		kfree(clp->cl_implid);
6817 		clp->cl_implid = res.impl_id;
6818 
6819 		if (clp->cl_serverscope != NULL &&
6820 		    !nfs41_same_server_scope(clp->cl_serverscope,
6821 					     res.server_scope)) {
6822 			dprintk("%s: server_scope mismatch detected\n",
6823 				__func__);
6824 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6825 			kfree(clp->cl_serverscope);
6826 			clp->cl_serverscope = NULL;
6827 		}
6828 
6829 		if (clp->cl_serverscope == NULL) {
6830 			clp->cl_serverscope = res.server_scope;
6831 			goto out;
6832 		}
6833 	} else
6834 		kfree(res.impl_id);
6835 
6836 out_server_owner:
6837 	kfree(res.server_owner);
6838 out_server_scope:
6839 	kfree(res.server_scope);
6840 out:
6841 	if (clp->cl_implid != NULL)
6842 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6843 			"domain: %s, name: %s, date: %llu,%u\n",
6844 			clp->cl_implid->domain, clp->cl_implid->name,
6845 			clp->cl_implid->date.seconds,
6846 			clp->cl_implid->date.nseconds);
6847 	dprintk("NFS reply exchange_id: %d\n", status);
6848 	return status;
6849 }
6850 
6851 /*
6852  * nfs4_proc_exchange_id()
6853  *
6854  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6855  *
6856  * Since the clientid has expired, all compounds using sessions
6857  * associated with the stale clientid will be returning
6858  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6859  * be in some phase of session reset.
6860  *
6861  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6862  */
6863 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6864 {
6865 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6866 	int status;
6867 
6868 	/* try SP4_MACH_CRED if krb5i/p	*/
6869 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6870 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6871 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6872 		if (!status)
6873 			return 0;
6874 	}
6875 
6876 	/* try SP4_NONE */
6877 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6878 }
6879 
6880 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6881 		struct rpc_cred *cred)
6882 {
6883 	struct rpc_message msg = {
6884 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6885 		.rpc_argp = clp,
6886 		.rpc_cred = cred,
6887 	};
6888 	int status;
6889 
6890 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6891 	trace_nfs4_destroy_clientid(clp, status);
6892 	if (status)
6893 		dprintk("NFS: Got error %d from the server %s on "
6894 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6895 	return status;
6896 }
6897 
6898 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6899 		struct rpc_cred *cred)
6900 {
6901 	unsigned int loop;
6902 	int ret;
6903 
6904 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6905 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6906 		switch (ret) {
6907 		case -NFS4ERR_DELAY:
6908 		case -NFS4ERR_CLIENTID_BUSY:
6909 			ssleep(1);
6910 			break;
6911 		default:
6912 			return ret;
6913 		}
6914 	}
6915 	return 0;
6916 }
6917 
6918 int nfs4_destroy_clientid(struct nfs_client *clp)
6919 {
6920 	struct rpc_cred *cred;
6921 	int ret = 0;
6922 
6923 	if (clp->cl_mvops->minor_version < 1)
6924 		goto out;
6925 	if (clp->cl_exchange_flags == 0)
6926 		goto out;
6927 	if (clp->cl_preserve_clid)
6928 		goto out;
6929 	cred = nfs4_get_clid_cred(clp);
6930 	ret = nfs4_proc_destroy_clientid(clp, cred);
6931 	if (cred)
6932 		put_rpccred(cred);
6933 	switch (ret) {
6934 	case 0:
6935 	case -NFS4ERR_STALE_CLIENTID:
6936 		clp->cl_exchange_flags = 0;
6937 	}
6938 out:
6939 	return ret;
6940 }
6941 
6942 struct nfs4_get_lease_time_data {
6943 	struct nfs4_get_lease_time_args *args;
6944 	struct nfs4_get_lease_time_res *res;
6945 	struct nfs_client *clp;
6946 };
6947 
6948 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6949 					void *calldata)
6950 {
6951 	struct nfs4_get_lease_time_data *data =
6952 			(struct nfs4_get_lease_time_data *)calldata;
6953 
6954 	dprintk("--> %s\n", __func__);
6955 	/* just setup sequence, do not trigger session recovery
6956 	   since we're invoked within one */
6957 	nfs41_setup_sequence(data->clp->cl_session,
6958 			&data->args->la_seq_args,
6959 			&data->res->lr_seq_res,
6960 			task);
6961 	dprintk("<-- %s\n", __func__);
6962 }
6963 
6964 /*
6965  * Called from nfs4_state_manager thread for session setup, so don't recover
6966  * from sequence operation or clientid errors.
6967  */
6968 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6969 {
6970 	struct nfs4_get_lease_time_data *data =
6971 			(struct nfs4_get_lease_time_data *)calldata;
6972 
6973 	dprintk("--> %s\n", __func__);
6974 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6975 		return;
6976 	switch (task->tk_status) {
6977 	case -NFS4ERR_DELAY:
6978 	case -NFS4ERR_GRACE:
6979 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6980 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
6981 		task->tk_status = 0;
6982 		/* fall through */
6983 	case -NFS4ERR_RETRY_UNCACHED_REP:
6984 		rpc_restart_call_prepare(task);
6985 		return;
6986 	}
6987 	dprintk("<-- %s\n", __func__);
6988 }
6989 
6990 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
6991 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
6992 	.rpc_call_done = nfs4_get_lease_time_done,
6993 };
6994 
6995 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
6996 {
6997 	struct rpc_task *task;
6998 	struct nfs4_get_lease_time_args args;
6999 	struct nfs4_get_lease_time_res res = {
7000 		.lr_fsinfo = fsinfo,
7001 	};
7002 	struct nfs4_get_lease_time_data data = {
7003 		.args = &args,
7004 		.res = &res,
7005 		.clp = clp,
7006 	};
7007 	struct rpc_message msg = {
7008 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7009 		.rpc_argp = &args,
7010 		.rpc_resp = &res,
7011 	};
7012 	struct rpc_task_setup task_setup = {
7013 		.rpc_client = clp->cl_rpcclient,
7014 		.rpc_message = &msg,
7015 		.callback_ops = &nfs4_get_lease_time_ops,
7016 		.callback_data = &data,
7017 		.flags = RPC_TASK_TIMEOUT,
7018 	};
7019 	int status;
7020 
7021 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7022 	nfs4_set_sequence_privileged(&args.la_seq_args);
7023 	dprintk("--> %s\n", __func__);
7024 	task = rpc_run_task(&task_setup);
7025 
7026 	if (IS_ERR(task))
7027 		status = PTR_ERR(task);
7028 	else {
7029 		status = task->tk_status;
7030 		rpc_put_task(task);
7031 	}
7032 	dprintk("<-- %s return %d\n", __func__, status);
7033 
7034 	return status;
7035 }
7036 
7037 /*
7038  * Initialize the values to be used by the client in CREATE_SESSION
7039  * If nfs4_init_session set the fore channel request and response sizes,
7040  * use them.
7041  *
7042  * Set the back channel max_resp_sz_cached to zero to force the client to
7043  * always set csa_cachethis to FALSE because the current implementation
7044  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7045  */
7046 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7047 {
7048 	unsigned int max_rqst_sz, max_resp_sz;
7049 
7050 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7051 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7052 
7053 	/* Fore channel attributes */
7054 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7055 	args->fc_attrs.max_resp_sz = max_resp_sz;
7056 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7057 	args->fc_attrs.max_reqs = max_session_slots;
7058 
7059 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7060 		"max_ops=%u max_reqs=%u\n",
7061 		__func__,
7062 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7063 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7064 
7065 	/* Back channel attributes */
7066 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7067 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7068 	args->bc_attrs.max_resp_sz_cached = 0;
7069 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7070 	args->bc_attrs.max_reqs = 1;
7071 
7072 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7073 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7074 		__func__,
7075 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7076 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7077 		args->bc_attrs.max_reqs);
7078 }
7079 
7080 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7081 {
7082 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7083 	struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7084 
7085 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7086 		return -EINVAL;
7087 	/*
7088 	 * Our requested max_ops is the minimum we need; we're not
7089 	 * prepared to break up compounds into smaller pieces than that.
7090 	 * So, no point even trying to continue if the server won't
7091 	 * cooperate:
7092 	 */
7093 	if (rcvd->max_ops < sent->max_ops)
7094 		return -EINVAL;
7095 	if (rcvd->max_reqs == 0)
7096 		return -EINVAL;
7097 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7098 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7099 	return 0;
7100 }
7101 
7102 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7103 {
7104 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7105 	struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7106 
7107 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7108 		return -EINVAL;
7109 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7110 		return -EINVAL;
7111 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7112 		return -EINVAL;
7113 	/* These would render the backchannel useless: */
7114 	if (rcvd->max_ops != sent->max_ops)
7115 		return -EINVAL;
7116 	if (rcvd->max_reqs != sent->max_reqs)
7117 		return -EINVAL;
7118 	return 0;
7119 }
7120 
7121 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7122 				     struct nfs4_session *session)
7123 {
7124 	int ret;
7125 
7126 	ret = nfs4_verify_fore_channel_attrs(args, session);
7127 	if (ret)
7128 		return ret;
7129 	return nfs4_verify_back_channel_attrs(args, session);
7130 }
7131 
7132 static int _nfs4_proc_create_session(struct nfs_client *clp,
7133 		struct rpc_cred *cred)
7134 {
7135 	struct nfs4_session *session = clp->cl_session;
7136 	struct nfs41_create_session_args args = {
7137 		.client = clp,
7138 		.cb_program = NFS4_CALLBACK,
7139 	};
7140 	struct nfs41_create_session_res res = {
7141 		.client = clp,
7142 	};
7143 	struct rpc_message msg = {
7144 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7145 		.rpc_argp = &args,
7146 		.rpc_resp = &res,
7147 		.rpc_cred = cred,
7148 	};
7149 	int status;
7150 
7151 	nfs4_init_channel_attrs(&args);
7152 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7153 
7154 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7155 	trace_nfs4_create_session(clp, status);
7156 
7157 	if (!status) {
7158 		/* Verify the session's negotiated channel_attrs values */
7159 		status = nfs4_verify_channel_attrs(&args, session);
7160 		/* Increment the clientid slot sequence id */
7161 		clp->cl_seqid++;
7162 	}
7163 
7164 	return status;
7165 }
7166 
7167 /*
7168  * Issues a CREATE_SESSION operation to the server.
7169  * It is the responsibility of the caller to verify the session is
7170  * expired before calling this routine.
7171  */
7172 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7173 {
7174 	int status;
7175 	unsigned *ptr;
7176 	struct nfs4_session *session = clp->cl_session;
7177 
7178 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7179 
7180 	status = _nfs4_proc_create_session(clp, cred);
7181 	if (status)
7182 		goto out;
7183 
7184 	/* Init or reset the session slot tables */
7185 	status = nfs4_setup_session_slot_tables(session);
7186 	dprintk("slot table setup returned %d\n", status);
7187 	if (status)
7188 		goto out;
7189 
7190 	ptr = (unsigned *)&session->sess_id.data[0];
7191 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7192 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7193 out:
7194 	dprintk("<-- %s\n", __func__);
7195 	return status;
7196 }
7197 
7198 /*
7199  * Issue the over-the-wire RPC DESTROY_SESSION.
7200  * The caller must serialize access to this routine.
7201  */
7202 int nfs4_proc_destroy_session(struct nfs4_session *session,
7203 		struct rpc_cred *cred)
7204 {
7205 	struct rpc_message msg = {
7206 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7207 		.rpc_argp = session,
7208 		.rpc_cred = cred,
7209 	};
7210 	int status = 0;
7211 
7212 	dprintk("--> nfs4_proc_destroy_session\n");
7213 
7214 	/* session is still being setup */
7215 	if (session->clp->cl_cons_state != NFS_CS_READY)
7216 		return status;
7217 
7218 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7219 	trace_nfs4_destroy_session(session->clp, status);
7220 
7221 	if (status)
7222 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7223 			"Session has been destroyed regardless...\n", status);
7224 
7225 	dprintk("<-- nfs4_proc_destroy_session\n");
7226 	return status;
7227 }
7228 
7229 /*
7230  * Renew the cl_session lease.
7231  */
7232 struct nfs4_sequence_data {
7233 	struct nfs_client *clp;
7234 	struct nfs4_sequence_args args;
7235 	struct nfs4_sequence_res res;
7236 };
7237 
7238 static void nfs41_sequence_release(void *data)
7239 {
7240 	struct nfs4_sequence_data *calldata = data;
7241 	struct nfs_client *clp = calldata->clp;
7242 
7243 	if (atomic_read(&clp->cl_count) > 1)
7244 		nfs4_schedule_state_renewal(clp);
7245 	nfs_put_client(clp);
7246 	kfree(calldata);
7247 }
7248 
7249 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7250 {
7251 	switch(task->tk_status) {
7252 	case -NFS4ERR_DELAY:
7253 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7254 		return -EAGAIN;
7255 	default:
7256 		nfs4_schedule_lease_recovery(clp);
7257 	}
7258 	return 0;
7259 }
7260 
7261 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7262 {
7263 	struct nfs4_sequence_data *calldata = data;
7264 	struct nfs_client *clp = calldata->clp;
7265 
7266 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7267 		return;
7268 
7269 	trace_nfs4_sequence(clp, task->tk_status);
7270 	if (task->tk_status < 0) {
7271 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7272 		if (atomic_read(&clp->cl_count) == 1)
7273 			goto out;
7274 
7275 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7276 			rpc_restart_call_prepare(task);
7277 			return;
7278 		}
7279 	}
7280 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7281 out:
7282 	dprintk("<-- %s\n", __func__);
7283 }
7284 
7285 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7286 {
7287 	struct nfs4_sequence_data *calldata = data;
7288 	struct nfs_client *clp = calldata->clp;
7289 	struct nfs4_sequence_args *args;
7290 	struct nfs4_sequence_res *res;
7291 
7292 	args = task->tk_msg.rpc_argp;
7293 	res = task->tk_msg.rpc_resp;
7294 
7295 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7296 }
7297 
7298 static const struct rpc_call_ops nfs41_sequence_ops = {
7299 	.rpc_call_done = nfs41_sequence_call_done,
7300 	.rpc_call_prepare = nfs41_sequence_prepare,
7301 	.rpc_release = nfs41_sequence_release,
7302 };
7303 
7304 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7305 		struct rpc_cred *cred,
7306 		bool is_privileged)
7307 {
7308 	struct nfs4_sequence_data *calldata;
7309 	struct rpc_message msg = {
7310 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7311 		.rpc_cred = cred,
7312 	};
7313 	struct rpc_task_setup task_setup_data = {
7314 		.rpc_client = clp->cl_rpcclient,
7315 		.rpc_message = &msg,
7316 		.callback_ops = &nfs41_sequence_ops,
7317 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7318 	};
7319 
7320 	if (!atomic_inc_not_zero(&clp->cl_count))
7321 		return ERR_PTR(-EIO);
7322 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7323 	if (calldata == NULL) {
7324 		nfs_put_client(clp);
7325 		return ERR_PTR(-ENOMEM);
7326 	}
7327 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7328 	if (is_privileged)
7329 		nfs4_set_sequence_privileged(&calldata->args);
7330 	msg.rpc_argp = &calldata->args;
7331 	msg.rpc_resp = &calldata->res;
7332 	calldata->clp = clp;
7333 	task_setup_data.callback_data = calldata;
7334 
7335 	return rpc_run_task(&task_setup_data);
7336 }
7337 
7338 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7339 {
7340 	struct rpc_task *task;
7341 	int ret = 0;
7342 
7343 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7344 		return 0;
7345 	task = _nfs41_proc_sequence(clp, cred, false);
7346 	if (IS_ERR(task))
7347 		ret = PTR_ERR(task);
7348 	else
7349 		rpc_put_task_async(task);
7350 	dprintk("<-- %s status=%d\n", __func__, ret);
7351 	return ret;
7352 }
7353 
7354 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7355 {
7356 	struct rpc_task *task;
7357 	int ret;
7358 
7359 	task = _nfs41_proc_sequence(clp, cred, true);
7360 	if (IS_ERR(task)) {
7361 		ret = PTR_ERR(task);
7362 		goto out;
7363 	}
7364 	ret = rpc_wait_for_completion_task(task);
7365 	if (!ret) {
7366 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7367 
7368 		if (task->tk_status == 0)
7369 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7370 		ret = task->tk_status;
7371 	}
7372 	rpc_put_task(task);
7373 out:
7374 	dprintk("<-- %s status=%d\n", __func__, ret);
7375 	return ret;
7376 }
7377 
7378 struct nfs4_reclaim_complete_data {
7379 	struct nfs_client *clp;
7380 	struct nfs41_reclaim_complete_args arg;
7381 	struct nfs41_reclaim_complete_res res;
7382 };
7383 
7384 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7385 {
7386 	struct nfs4_reclaim_complete_data *calldata = data;
7387 
7388 	nfs41_setup_sequence(calldata->clp->cl_session,
7389 			&calldata->arg.seq_args,
7390 			&calldata->res.seq_res,
7391 			task);
7392 }
7393 
7394 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7395 {
7396 	switch(task->tk_status) {
7397 	case 0:
7398 	case -NFS4ERR_COMPLETE_ALREADY:
7399 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7400 		break;
7401 	case -NFS4ERR_DELAY:
7402 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7403 		/* fall through */
7404 	case -NFS4ERR_RETRY_UNCACHED_REP:
7405 		return -EAGAIN;
7406 	default:
7407 		nfs4_schedule_lease_recovery(clp);
7408 	}
7409 	return 0;
7410 }
7411 
7412 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7413 {
7414 	struct nfs4_reclaim_complete_data *calldata = data;
7415 	struct nfs_client *clp = calldata->clp;
7416 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7417 
7418 	dprintk("--> %s\n", __func__);
7419 	if (!nfs41_sequence_done(task, res))
7420 		return;
7421 
7422 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7423 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7424 		rpc_restart_call_prepare(task);
7425 		return;
7426 	}
7427 	dprintk("<-- %s\n", __func__);
7428 }
7429 
7430 static void nfs4_free_reclaim_complete_data(void *data)
7431 {
7432 	struct nfs4_reclaim_complete_data *calldata = data;
7433 
7434 	kfree(calldata);
7435 }
7436 
7437 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7438 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7439 	.rpc_call_done = nfs4_reclaim_complete_done,
7440 	.rpc_release = nfs4_free_reclaim_complete_data,
7441 };
7442 
7443 /*
7444  * Issue a global reclaim complete.
7445  */
7446 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7447 		struct rpc_cred *cred)
7448 {
7449 	struct nfs4_reclaim_complete_data *calldata;
7450 	struct rpc_task *task;
7451 	struct rpc_message msg = {
7452 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7453 		.rpc_cred = cred,
7454 	};
7455 	struct rpc_task_setup task_setup_data = {
7456 		.rpc_client = clp->cl_rpcclient,
7457 		.rpc_message = &msg,
7458 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7459 		.flags = RPC_TASK_ASYNC,
7460 	};
7461 	int status = -ENOMEM;
7462 
7463 	dprintk("--> %s\n", __func__);
7464 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7465 	if (calldata == NULL)
7466 		goto out;
7467 	calldata->clp = clp;
7468 	calldata->arg.one_fs = 0;
7469 
7470 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7471 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7472 	msg.rpc_argp = &calldata->arg;
7473 	msg.rpc_resp = &calldata->res;
7474 	task_setup_data.callback_data = calldata;
7475 	task = rpc_run_task(&task_setup_data);
7476 	if (IS_ERR(task)) {
7477 		status = PTR_ERR(task);
7478 		goto out;
7479 	}
7480 	status = nfs4_wait_for_completion_rpc_task(task);
7481 	if (status == 0)
7482 		status = task->tk_status;
7483 	rpc_put_task(task);
7484 	return 0;
7485 out:
7486 	dprintk("<-- %s status=%d\n", __func__, status);
7487 	return status;
7488 }
7489 
7490 static void
7491 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7492 {
7493 	struct nfs4_layoutget *lgp = calldata;
7494 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7495 	struct nfs4_session *session = nfs4_get_session(server);
7496 
7497 	dprintk("--> %s\n", __func__);
7498 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7499 	 * right now covering the LAYOUTGET we are about to send.
7500 	 * However, that is not so catastrophic, and there seems
7501 	 * to be no way to prevent it completely.
7502 	 */
7503 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7504 				&lgp->res.seq_res, task))
7505 		return;
7506 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7507 					  NFS_I(lgp->args.inode)->layout,
7508 					  lgp->args.ctx->state)) {
7509 		rpc_exit(task, NFS4_OK);
7510 	}
7511 }
7512 
7513 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7514 {
7515 	struct nfs4_layoutget *lgp = calldata;
7516 	struct inode *inode = lgp->args.inode;
7517 	struct nfs_server *server = NFS_SERVER(inode);
7518 	struct pnfs_layout_hdr *lo;
7519 	struct nfs4_state *state = NULL;
7520 	unsigned long timeo, now, giveup;
7521 
7522 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7523 
7524 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7525 		goto out;
7526 
7527 	switch (task->tk_status) {
7528 	case 0:
7529 		goto out;
7530 	/*
7531 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7532 	 * (or clients) writing to the same RAID stripe
7533 	 */
7534 	case -NFS4ERR_LAYOUTTRYLATER:
7535 	/*
7536 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7537 	 * existing layout before getting a new one).
7538 	 */
7539 	case -NFS4ERR_RECALLCONFLICT:
7540 		timeo = rpc_get_timeout(task->tk_client);
7541 		giveup = lgp->args.timestamp + timeo;
7542 		now = jiffies;
7543 		if (time_after(giveup, now)) {
7544 			unsigned long delay;
7545 
7546 			/* Delay for:
7547 			 * - Not less then NFS4_POLL_RETRY_MIN.
7548 			 * - One last time a jiffie before we give up
7549 			 * - exponential backoff (time_now minus start_attempt)
7550 			 */
7551 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7552 				    min((giveup - now - 1),
7553 					now - lgp->args.timestamp));
7554 
7555 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7556 				__func__, delay);
7557 			rpc_delay(task, delay);
7558 			task->tk_status = 0;
7559 			rpc_restart_call_prepare(task);
7560 			goto out; /* Do not call nfs4_async_handle_error() */
7561 		}
7562 		break;
7563 	case -NFS4ERR_EXPIRED:
7564 	case -NFS4ERR_BAD_STATEID:
7565 		spin_lock(&inode->i_lock);
7566 		lo = NFS_I(inode)->layout;
7567 		if (!lo || list_empty(&lo->plh_segs)) {
7568 			spin_unlock(&inode->i_lock);
7569 			/* If the open stateid was bad, then recover it. */
7570 			state = lgp->args.ctx->state;
7571 		} else {
7572 			LIST_HEAD(head);
7573 
7574 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7575 			spin_unlock(&inode->i_lock);
7576 			/* Mark the bad layout state as invalid, then
7577 			 * retry using the open stateid. */
7578 			pnfs_free_lseg_list(&head);
7579 		}
7580 	}
7581 	if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
7582 		rpc_restart_call_prepare(task);
7583 out:
7584 	dprintk("<-- %s\n", __func__);
7585 }
7586 
7587 static size_t max_response_pages(struct nfs_server *server)
7588 {
7589 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7590 	return nfs_page_array_len(0, max_resp_sz);
7591 }
7592 
7593 static void nfs4_free_pages(struct page **pages, size_t size)
7594 {
7595 	int i;
7596 
7597 	if (!pages)
7598 		return;
7599 
7600 	for (i = 0; i < size; i++) {
7601 		if (!pages[i])
7602 			break;
7603 		__free_page(pages[i]);
7604 	}
7605 	kfree(pages);
7606 }
7607 
7608 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7609 {
7610 	struct page **pages;
7611 	int i;
7612 
7613 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7614 	if (!pages) {
7615 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7616 		return NULL;
7617 	}
7618 
7619 	for (i = 0; i < size; i++) {
7620 		pages[i] = alloc_page(gfp_flags);
7621 		if (!pages[i]) {
7622 			dprintk("%s: failed to allocate page\n", __func__);
7623 			nfs4_free_pages(pages, size);
7624 			return NULL;
7625 		}
7626 	}
7627 
7628 	return pages;
7629 }
7630 
7631 static void nfs4_layoutget_release(void *calldata)
7632 {
7633 	struct nfs4_layoutget *lgp = calldata;
7634 	struct inode *inode = lgp->args.inode;
7635 	struct nfs_server *server = NFS_SERVER(inode);
7636 	size_t max_pages = max_response_pages(server);
7637 
7638 	dprintk("--> %s\n", __func__);
7639 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7640 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7641 	put_nfs_open_context(lgp->args.ctx);
7642 	kfree(calldata);
7643 	dprintk("<-- %s\n", __func__);
7644 }
7645 
7646 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7647 	.rpc_call_prepare = nfs4_layoutget_prepare,
7648 	.rpc_call_done = nfs4_layoutget_done,
7649 	.rpc_release = nfs4_layoutget_release,
7650 };
7651 
7652 struct pnfs_layout_segment *
7653 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7654 {
7655 	struct inode *inode = lgp->args.inode;
7656 	struct nfs_server *server = NFS_SERVER(inode);
7657 	size_t max_pages = max_response_pages(server);
7658 	struct rpc_task *task;
7659 	struct rpc_message msg = {
7660 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7661 		.rpc_argp = &lgp->args,
7662 		.rpc_resp = &lgp->res,
7663 		.rpc_cred = lgp->cred,
7664 	};
7665 	struct rpc_task_setup task_setup_data = {
7666 		.rpc_client = server->client,
7667 		.rpc_message = &msg,
7668 		.callback_ops = &nfs4_layoutget_call_ops,
7669 		.callback_data = lgp,
7670 		.flags = RPC_TASK_ASYNC,
7671 	};
7672 	struct pnfs_layout_segment *lseg = NULL;
7673 	int status = 0;
7674 
7675 	dprintk("--> %s\n", __func__);
7676 
7677 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7678 	if (!lgp->args.layout.pages) {
7679 		nfs4_layoutget_release(lgp);
7680 		return ERR_PTR(-ENOMEM);
7681 	}
7682 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7683 	lgp->args.timestamp = jiffies;
7684 
7685 	lgp->res.layoutp = &lgp->args.layout;
7686 	lgp->res.seq_res.sr_slot = NULL;
7687 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7688 
7689 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7690 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7691 
7692 	task = rpc_run_task(&task_setup_data);
7693 	if (IS_ERR(task))
7694 		return ERR_CAST(task);
7695 	status = nfs4_wait_for_completion_rpc_task(task);
7696 	if (status == 0)
7697 		status = task->tk_status;
7698 	trace_nfs4_layoutget(lgp->args.ctx,
7699 			&lgp->args.range,
7700 			&lgp->res.range,
7701 			status);
7702 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7703 	if (status == 0 && lgp->res.layoutp->len)
7704 		lseg = pnfs_layout_process(lgp);
7705 	rpc_put_task(task);
7706 	dprintk("<-- %s status=%d\n", __func__, status);
7707 	if (status)
7708 		return ERR_PTR(status);
7709 	return lseg;
7710 }
7711 
7712 static void
7713 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7714 {
7715 	struct nfs4_layoutreturn *lrp = calldata;
7716 
7717 	dprintk("--> %s\n", __func__);
7718 	nfs41_setup_sequence(lrp->clp->cl_session,
7719 			&lrp->args.seq_args,
7720 			&lrp->res.seq_res,
7721 			task);
7722 }
7723 
7724 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7725 {
7726 	struct nfs4_layoutreturn *lrp = calldata;
7727 	struct nfs_server *server;
7728 
7729 	dprintk("--> %s\n", __func__);
7730 
7731 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7732 		return;
7733 
7734 	server = NFS_SERVER(lrp->args.inode);
7735 	switch (task->tk_status) {
7736 	default:
7737 		task->tk_status = 0;
7738 	case 0:
7739 		break;
7740 	case -NFS4ERR_DELAY:
7741 		if (nfs4_async_handle_error(task, server, NULL) != -EAGAIN)
7742 			break;
7743 		rpc_restart_call_prepare(task);
7744 		return;
7745 	}
7746 	dprintk("<-- %s\n", __func__);
7747 }
7748 
7749 static void nfs4_layoutreturn_release(void *calldata)
7750 {
7751 	struct nfs4_layoutreturn *lrp = calldata;
7752 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7753 
7754 	dprintk("--> %s\n", __func__);
7755 	spin_lock(&lo->plh_inode->i_lock);
7756 	if (lrp->res.lrs_present)
7757 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7758 	lo->plh_block_lgets--;
7759 	spin_unlock(&lo->plh_inode->i_lock);
7760 	pnfs_put_layout_hdr(lrp->args.layout);
7761 	kfree(calldata);
7762 	dprintk("<-- %s\n", __func__);
7763 }
7764 
7765 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7766 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7767 	.rpc_call_done = nfs4_layoutreturn_done,
7768 	.rpc_release = nfs4_layoutreturn_release,
7769 };
7770 
7771 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7772 {
7773 	struct rpc_task *task;
7774 	struct rpc_message msg = {
7775 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7776 		.rpc_argp = &lrp->args,
7777 		.rpc_resp = &lrp->res,
7778 		.rpc_cred = lrp->cred,
7779 	};
7780 	struct rpc_task_setup task_setup_data = {
7781 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7782 		.rpc_message = &msg,
7783 		.callback_ops = &nfs4_layoutreturn_call_ops,
7784 		.callback_data = lrp,
7785 	};
7786 	int status;
7787 
7788 	dprintk("--> %s\n", __func__);
7789 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7790 	task = rpc_run_task(&task_setup_data);
7791 	if (IS_ERR(task))
7792 		return PTR_ERR(task);
7793 	status = task->tk_status;
7794 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7795 	dprintk("<-- %s status=%d\n", __func__, status);
7796 	rpc_put_task(task);
7797 	return status;
7798 }
7799 
7800 /*
7801  * Retrieve the list of Data Server devices from the MDS.
7802  */
7803 static int _nfs4_getdevicelist(struct nfs_server *server,
7804 				    const struct nfs_fh *fh,
7805 				    struct pnfs_devicelist *devlist)
7806 {
7807 	struct nfs4_getdevicelist_args args = {
7808 		.fh = fh,
7809 		.layoutclass = server->pnfs_curr_ld->id,
7810 	};
7811 	struct nfs4_getdevicelist_res res = {
7812 		.devlist = devlist,
7813 	};
7814 	struct rpc_message msg = {
7815 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
7816 		.rpc_argp = &args,
7817 		.rpc_resp = &res,
7818 	};
7819 	int status;
7820 
7821 	dprintk("--> %s\n", __func__);
7822 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
7823 				&res.seq_res, 0);
7824 	dprintk("<-- %s status=%d\n", __func__, status);
7825 	return status;
7826 }
7827 
7828 int nfs4_proc_getdevicelist(struct nfs_server *server,
7829 			    const struct nfs_fh *fh,
7830 			    struct pnfs_devicelist *devlist)
7831 {
7832 	struct nfs4_exception exception = { };
7833 	int err;
7834 
7835 	do {
7836 		err = nfs4_handle_exception(server,
7837 				_nfs4_getdevicelist(server, fh, devlist),
7838 				&exception);
7839 	} while (exception.retry);
7840 
7841 	dprintk("%s: err=%d, num_devs=%u\n", __func__,
7842 		err, devlist->num_devs);
7843 
7844 	return err;
7845 }
7846 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
7847 
7848 static int
7849 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7850 		struct pnfs_device *pdev,
7851 		struct rpc_cred *cred)
7852 {
7853 	struct nfs4_getdeviceinfo_args args = {
7854 		.pdev = pdev,
7855 	};
7856 	struct nfs4_getdeviceinfo_res res = {
7857 		.pdev = pdev,
7858 	};
7859 	struct rpc_message msg = {
7860 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7861 		.rpc_argp = &args,
7862 		.rpc_resp = &res,
7863 		.rpc_cred = cred,
7864 	};
7865 	int status;
7866 
7867 	dprintk("--> %s\n", __func__);
7868 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7869 	dprintk("<-- %s status=%d\n", __func__, status);
7870 
7871 	return status;
7872 }
7873 
7874 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7875 		struct pnfs_device *pdev,
7876 		struct rpc_cred *cred)
7877 {
7878 	struct nfs4_exception exception = { };
7879 	int err;
7880 
7881 	do {
7882 		err = nfs4_handle_exception(server,
7883 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7884 					&exception);
7885 	} while (exception.retry);
7886 	return err;
7887 }
7888 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7889 
7890 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7891 {
7892 	struct nfs4_layoutcommit_data *data = calldata;
7893 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7894 	struct nfs4_session *session = nfs4_get_session(server);
7895 
7896 	nfs41_setup_sequence(session,
7897 			&data->args.seq_args,
7898 			&data->res.seq_res,
7899 			task);
7900 }
7901 
7902 static void
7903 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7904 {
7905 	struct nfs4_layoutcommit_data *data = calldata;
7906 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7907 
7908 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7909 		return;
7910 
7911 	switch (task->tk_status) { /* Just ignore these failures */
7912 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7913 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7914 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7915 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7916 		task->tk_status = 0;
7917 	case 0:
7918 		break;
7919 	default:
7920 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7921 			rpc_restart_call_prepare(task);
7922 			return;
7923 		}
7924 	}
7925 }
7926 
7927 static void nfs4_layoutcommit_release(void *calldata)
7928 {
7929 	struct nfs4_layoutcommit_data *data = calldata;
7930 
7931 	pnfs_cleanup_layoutcommit(data);
7932 	nfs_post_op_update_inode_force_wcc(data->args.inode,
7933 					   data->res.fattr);
7934 	put_rpccred(data->cred);
7935 	kfree(data);
7936 }
7937 
7938 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7939 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
7940 	.rpc_call_done = nfs4_layoutcommit_done,
7941 	.rpc_release = nfs4_layoutcommit_release,
7942 };
7943 
7944 int
7945 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7946 {
7947 	struct rpc_message msg = {
7948 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7949 		.rpc_argp = &data->args,
7950 		.rpc_resp = &data->res,
7951 		.rpc_cred = data->cred,
7952 	};
7953 	struct rpc_task_setup task_setup_data = {
7954 		.task = &data->task,
7955 		.rpc_client = NFS_CLIENT(data->args.inode),
7956 		.rpc_message = &msg,
7957 		.callback_ops = &nfs4_layoutcommit_ops,
7958 		.callback_data = data,
7959 		.flags = RPC_TASK_ASYNC,
7960 	};
7961 	struct rpc_task *task;
7962 	int status = 0;
7963 
7964 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7965 		"lbw: %llu inode %lu\n",
7966 		data->task.tk_pid, sync,
7967 		data->args.lastbytewritten,
7968 		data->args.inode->i_ino);
7969 
7970 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7971 	task = rpc_run_task(&task_setup_data);
7972 	if (IS_ERR(task))
7973 		return PTR_ERR(task);
7974 	if (sync == false)
7975 		goto out;
7976 	status = nfs4_wait_for_completion_rpc_task(task);
7977 	if (status != 0)
7978 		goto out;
7979 	status = task->tk_status;
7980 	trace_nfs4_layoutcommit(data->args.inode, status);
7981 out:
7982 	dprintk("%s: status %d\n", __func__, status);
7983 	rpc_put_task(task);
7984 	return status;
7985 }
7986 
7987 /**
7988  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7989  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7990  */
7991 static int
7992 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7993 		    struct nfs_fsinfo *info,
7994 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7995 {
7996 	struct nfs41_secinfo_no_name_args args = {
7997 		.style = SECINFO_STYLE_CURRENT_FH,
7998 	};
7999 	struct nfs4_secinfo_res res = {
8000 		.flavors = flavors,
8001 	};
8002 	struct rpc_message msg = {
8003 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8004 		.rpc_argp = &args,
8005 		.rpc_resp = &res,
8006 	};
8007 	struct rpc_clnt *clnt = server->client;
8008 	struct rpc_cred *cred = NULL;
8009 	int status;
8010 
8011 	if (use_integrity) {
8012 		clnt = server->nfs_client->cl_rpcclient;
8013 		cred = nfs4_get_clid_cred(server->nfs_client);
8014 		msg.rpc_cred = cred;
8015 	}
8016 
8017 	dprintk("--> %s\n", __func__);
8018 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8019 				&res.seq_res, 0);
8020 	dprintk("<-- %s status=%d\n", __func__, status);
8021 
8022 	if (cred)
8023 		put_rpccred(cred);
8024 
8025 	return status;
8026 }
8027 
8028 static int
8029 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8030 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8031 {
8032 	struct nfs4_exception exception = { };
8033 	int err;
8034 	do {
8035 		/* first try using integrity protection */
8036 		err = -NFS4ERR_WRONGSEC;
8037 
8038 		/* try to use integrity protection with machine cred */
8039 		if (_nfs4_is_integrity_protected(server->nfs_client))
8040 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8041 							  flavors, true);
8042 
8043 		/*
8044 		 * if unable to use integrity protection, or SECINFO with
8045 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8046 		 * disallowed by spec, but exists in deployed servers) use
8047 		 * the current filesystem's rpc_client and the user cred.
8048 		 */
8049 		if (err == -NFS4ERR_WRONGSEC)
8050 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8051 							  flavors, false);
8052 
8053 		switch (err) {
8054 		case 0:
8055 		case -NFS4ERR_WRONGSEC:
8056 		case -ENOTSUPP:
8057 			goto out;
8058 		default:
8059 			err = nfs4_handle_exception(server, err, &exception);
8060 		}
8061 	} while (exception.retry);
8062 out:
8063 	return err;
8064 }
8065 
8066 static int
8067 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8068 		    struct nfs_fsinfo *info)
8069 {
8070 	int err;
8071 	struct page *page;
8072 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8073 	struct nfs4_secinfo_flavors *flavors;
8074 	struct nfs4_secinfo4 *secinfo;
8075 	int i;
8076 
8077 	page = alloc_page(GFP_KERNEL);
8078 	if (!page) {
8079 		err = -ENOMEM;
8080 		goto out;
8081 	}
8082 
8083 	flavors = page_address(page);
8084 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8085 
8086 	/*
8087 	 * Fall back on "guess and check" method if
8088 	 * the server doesn't support SECINFO_NO_NAME
8089 	 */
8090 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8091 		err = nfs4_find_root_sec(server, fhandle, info);
8092 		goto out_freepage;
8093 	}
8094 	if (err)
8095 		goto out_freepage;
8096 
8097 	for (i = 0; i < flavors->num_flavors; i++) {
8098 		secinfo = &flavors->flavors[i];
8099 
8100 		switch (secinfo->flavor) {
8101 		case RPC_AUTH_NULL:
8102 		case RPC_AUTH_UNIX:
8103 		case RPC_AUTH_GSS:
8104 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8105 					&secinfo->flavor_info);
8106 			break;
8107 		default:
8108 			flavor = RPC_AUTH_MAXFLAVOR;
8109 			break;
8110 		}
8111 
8112 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8113 			flavor = RPC_AUTH_MAXFLAVOR;
8114 
8115 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8116 			err = nfs4_lookup_root_sec(server, fhandle,
8117 						   info, flavor);
8118 			if (!err)
8119 				break;
8120 		}
8121 	}
8122 
8123 	if (flavor == RPC_AUTH_MAXFLAVOR)
8124 		err = -EPERM;
8125 
8126 out_freepage:
8127 	put_page(page);
8128 	if (err == -EACCES)
8129 		return -EPERM;
8130 out:
8131 	return err;
8132 }
8133 
8134 static int _nfs41_test_stateid(struct nfs_server *server,
8135 		nfs4_stateid *stateid,
8136 		struct rpc_cred *cred)
8137 {
8138 	int status;
8139 	struct nfs41_test_stateid_args args = {
8140 		.stateid = stateid,
8141 	};
8142 	struct nfs41_test_stateid_res res;
8143 	struct rpc_message msg = {
8144 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8145 		.rpc_argp = &args,
8146 		.rpc_resp = &res,
8147 		.rpc_cred = cred,
8148 	};
8149 	struct rpc_clnt *rpc_client = server->client;
8150 
8151 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8152 		&rpc_client, &msg);
8153 
8154 	dprintk("NFS call  test_stateid %p\n", stateid);
8155 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8156 	nfs4_set_sequence_privileged(&args.seq_args);
8157 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8158 			&args.seq_args, &res.seq_res);
8159 	if (status != NFS_OK) {
8160 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8161 		return status;
8162 	}
8163 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8164 	return -res.status;
8165 }
8166 
8167 /**
8168  * nfs41_test_stateid - perform a TEST_STATEID operation
8169  *
8170  * @server: server / transport on which to perform the operation
8171  * @stateid: state ID to test
8172  * @cred: credential
8173  *
8174  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8175  * Otherwise a negative NFS4ERR value is returned if the operation
8176  * failed or the state ID is not currently valid.
8177  */
8178 static int nfs41_test_stateid(struct nfs_server *server,
8179 		nfs4_stateid *stateid,
8180 		struct rpc_cred *cred)
8181 {
8182 	struct nfs4_exception exception = { };
8183 	int err;
8184 	do {
8185 		err = _nfs41_test_stateid(server, stateid, cred);
8186 		if (err != -NFS4ERR_DELAY)
8187 			break;
8188 		nfs4_handle_exception(server, err, &exception);
8189 	} while (exception.retry);
8190 	return err;
8191 }
8192 
8193 struct nfs_free_stateid_data {
8194 	struct nfs_server *server;
8195 	struct nfs41_free_stateid_args args;
8196 	struct nfs41_free_stateid_res res;
8197 };
8198 
8199 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8200 {
8201 	struct nfs_free_stateid_data *data = calldata;
8202 	nfs41_setup_sequence(nfs4_get_session(data->server),
8203 			&data->args.seq_args,
8204 			&data->res.seq_res,
8205 			task);
8206 }
8207 
8208 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8209 {
8210 	struct nfs_free_stateid_data *data = calldata;
8211 
8212 	nfs41_sequence_done(task, &data->res.seq_res);
8213 
8214 	switch (task->tk_status) {
8215 	case -NFS4ERR_DELAY:
8216 		if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
8217 			rpc_restart_call_prepare(task);
8218 	}
8219 }
8220 
8221 static void nfs41_free_stateid_release(void *calldata)
8222 {
8223 	kfree(calldata);
8224 }
8225 
8226 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8227 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8228 	.rpc_call_done = nfs41_free_stateid_done,
8229 	.rpc_release = nfs41_free_stateid_release,
8230 };
8231 
8232 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8233 		nfs4_stateid *stateid,
8234 		struct rpc_cred *cred,
8235 		bool privileged)
8236 {
8237 	struct rpc_message msg = {
8238 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8239 		.rpc_cred = cred,
8240 	};
8241 	struct rpc_task_setup task_setup = {
8242 		.rpc_client = server->client,
8243 		.rpc_message = &msg,
8244 		.callback_ops = &nfs41_free_stateid_ops,
8245 		.flags = RPC_TASK_ASYNC,
8246 	};
8247 	struct nfs_free_stateid_data *data;
8248 
8249 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8250 		&task_setup.rpc_client, &msg);
8251 
8252 	dprintk("NFS call  free_stateid %p\n", stateid);
8253 	data = kmalloc(sizeof(*data), GFP_NOFS);
8254 	if (!data)
8255 		return ERR_PTR(-ENOMEM);
8256 	data->server = server;
8257 	nfs4_stateid_copy(&data->args.stateid, stateid);
8258 
8259 	task_setup.callback_data = data;
8260 
8261 	msg.rpc_argp = &data->args;
8262 	msg.rpc_resp = &data->res;
8263 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8264 	if (privileged)
8265 		nfs4_set_sequence_privileged(&data->args.seq_args);
8266 
8267 	return rpc_run_task(&task_setup);
8268 }
8269 
8270 /**
8271  * nfs41_free_stateid - perform a FREE_STATEID operation
8272  *
8273  * @server: server / transport on which to perform the operation
8274  * @stateid: state ID to release
8275  * @cred: credential
8276  *
8277  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8278  * negative NFS4ERR value is returned.
8279  */
8280 static int nfs41_free_stateid(struct nfs_server *server,
8281 		nfs4_stateid *stateid,
8282 		struct rpc_cred *cred)
8283 {
8284 	struct rpc_task *task;
8285 	int ret;
8286 
8287 	task = _nfs41_free_stateid(server, stateid, cred, true);
8288 	if (IS_ERR(task))
8289 		return PTR_ERR(task);
8290 	ret = rpc_wait_for_completion_task(task);
8291 	if (!ret)
8292 		ret = task->tk_status;
8293 	rpc_put_task(task);
8294 	return ret;
8295 }
8296 
8297 static void
8298 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8299 {
8300 	struct rpc_task *task;
8301 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8302 
8303 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8304 	nfs4_free_lock_state(server, lsp);
8305 	if (IS_ERR(task))
8306 		return;
8307 	rpc_put_task(task);
8308 }
8309 
8310 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8311 		const nfs4_stateid *s2)
8312 {
8313 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8314 		return false;
8315 
8316 	if (s1->seqid == s2->seqid)
8317 		return true;
8318 	if (s1->seqid == 0 || s2->seqid == 0)
8319 		return true;
8320 
8321 	return false;
8322 }
8323 
8324 #endif /* CONFIG_NFS_V4_1 */
8325 
8326 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8327 		const nfs4_stateid *s2)
8328 {
8329 	return nfs4_stateid_match(s1, s2);
8330 }
8331 
8332 
8333 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8334 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8335 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8336 	.recover_open	= nfs4_open_reclaim,
8337 	.recover_lock	= nfs4_lock_reclaim,
8338 	.establish_clid = nfs4_init_clientid,
8339 	.detect_trunking = nfs40_discover_server_trunking,
8340 };
8341 
8342 #if defined(CONFIG_NFS_V4_1)
8343 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8344 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8345 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8346 	.recover_open	= nfs4_open_reclaim,
8347 	.recover_lock	= nfs4_lock_reclaim,
8348 	.establish_clid = nfs41_init_clientid,
8349 	.reclaim_complete = nfs41_proc_reclaim_complete,
8350 	.detect_trunking = nfs41_discover_server_trunking,
8351 };
8352 #endif /* CONFIG_NFS_V4_1 */
8353 
8354 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8355 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8356 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8357 	.recover_open	= nfs4_open_expired,
8358 	.recover_lock	= nfs4_lock_expired,
8359 	.establish_clid = nfs4_init_clientid,
8360 };
8361 
8362 #if defined(CONFIG_NFS_V4_1)
8363 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8364 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8365 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8366 	.recover_open	= nfs41_open_expired,
8367 	.recover_lock	= nfs41_lock_expired,
8368 	.establish_clid = nfs41_init_clientid,
8369 };
8370 #endif /* CONFIG_NFS_V4_1 */
8371 
8372 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8373 	.sched_state_renewal = nfs4_proc_async_renew,
8374 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8375 	.renew_lease = nfs4_proc_renew,
8376 };
8377 
8378 #if defined(CONFIG_NFS_V4_1)
8379 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8380 	.sched_state_renewal = nfs41_proc_async_sequence,
8381 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8382 	.renew_lease = nfs4_proc_sequence,
8383 };
8384 #endif
8385 
8386 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8387 	.get_locations = _nfs40_proc_get_locations,
8388 	.fsid_present = _nfs40_proc_fsid_present,
8389 };
8390 
8391 #if defined(CONFIG_NFS_V4_1)
8392 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8393 	.get_locations = _nfs41_proc_get_locations,
8394 	.fsid_present = _nfs41_proc_fsid_present,
8395 };
8396 #endif	/* CONFIG_NFS_V4_1 */
8397 
8398 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8399 	.minor_version = 0,
8400 	.init_caps = NFS_CAP_READDIRPLUS
8401 		| NFS_CAP_ATOMIC_OPEN
8402 		| NFS_CAP_CHANGE_ATTR
8403 		| NFS_CAP_POSIX_LOCK,
8404 	.init_client = nfs40_init_client,
8405 	.shutdown_client = nfs40_shutdown_client,
8406 	.match_stateid = nfs4_match_stateid,
8407 	.find_root_sec = nfs4_find_root_sec,
8408 	.free_lock_state = nfs4_release_lockowner,
8409 	.call_sync_ops = &nfs40_call_sync_ops,
8410 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8411 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8412 	.state_renewal_ops = &nfs40_state_renewal_ops,
8413 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8414 };
8415 
8416 #if defined(CONFIG_NFS_V4_1)
8417 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8418 	.minor_version = 1,
8419 	.init_caps = NFS_CAP_READDIRPLUS
8420 		| NFS_CAP_ATOMIC_OPEN
8421 		| NFS_CAP_CHANGE_ATTR
8422 		| NFS_CAP_POSIX_LOCK
8423 		| NFS_CAP_STATEID_NFSV41
8424 		| NFS_CAP_ATOMIC_OPEN_V1,
8425 	.init_client = nfs41_init_client,
8426 	.shutdown_client = nfs41_shutdown_client,
8427 	.match_stateid = nfs41_match_stateid,
8428 	.find_root_sec = nfs41_find_root_sec,
8429 	.free_lock_state = nfs41_free_lock_state,
8430 	.call_sync_ops = &nfs41_call_sync_ops,
8431 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8432 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8433 	.state_renewal_ops = &nfs41_state_renewal_ops,
8434 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8435 };
8436 #endif
8437 
8438 #if defined(CONFIG_NFS_V4_2)
8439 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8440 	.minor_version = 2,
8441 	.init_caps = NFS_CAP_READDIRPLUS
8442 		| NFS_CAP_ATOMIC_OPEN
8443 		| NFS_CAP_CHANGE_ATTR
8444 		| NFS_CAP_POSIX_LOCK
8445 		| NFS_CAP_STATEID_NFSV41
8446 		| NFS_CAP_ATOMIC_OPEN_V1,
8447 	.init_client = nfs41_init_client,
8448 	.shutdown_client = nfs41_shutdown_client,
8449 	.match_stateid = nfs41_match_stateid,
8450 	.find_root_sec = nfs41_find_root_sec,
8451 	.free_lock_state = nfs41_free_lock_state,
8452 	.call_sync_ops = &nfs41_call_sync_ops,
8453 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8454 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8455 	.state_renewal_ops = &nfs41_state_renewal_ops,
8456 };
8457 #endif
8458 
8459 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8460 	[0] = &nfs_v4_0_minor_ops,
8461 #if defined(CONFIG_NFS_V4_1)
8462 	[1] = &nfs_v4_1_minor_ops,
8463 #endif
8464 #if defined(CONFIG_NFS_V4_2)
8465 	[2] = &nfs_v4_2_minor_ops,
8466 #endif
8467 };
8468 
8469 static const struct inode_operations nfs4_dir_inode_operations = {
8470 	.create		= nfs_create,
8471 	.lookup		= nfs_lookup,
8472 	.atomic_open	= nfs_atomic_open,
8473 	.link		= nfs_link,
8474 	.unlink		= nfs_unlink,
8475 	.symlink	= nfs_symlink,
8476 	.mkdir		= nfs_mkdir,
8477 	.rmdir		= nfs_rmdir,
8478 	.mknod		= nfs_mknod,
8479 	.rename		= nfs_rename,
8480 	.permission	= nfs_permission,
8481 	.getattr	= nfs_getattr,
8482 	.setattr	= nfs_setattr,
8483 	.getxattr	= generic_getxattr,
8484 	.setxattr	= generic_setxattr,
8485 	.listxattr	= generic_listxattr,
8486 	.removexattr	= generic_removexattr,
8487 };
8488 
8489 static const struct inode_operations nfs4_file_inode_operations = {
8490 	.permission	= nfs_permission,
8491 	.getattr	= nfs_getattr,
8492 	.setattr	= nfs_setattr,
8493 	.getxattr	= generic_getxattr,
8494 	.setxattr	= generic_setxattr,
8495 	.listxattr	= generic_listxattr,
8496 	.removexattr	= generic_removexattr,
8497 };
8498 
8499 const struct nfs_rpc_ops nfs_v4_clientops = {
8500 	.version	= 4,			/* protocol version */
8501 	.dentry_ops	= &nfs4_dentry_operations,
8502 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8503 	.file_inode_ops	= &nfs4_file_inode_operations,
8504 	.file_ops	= &nfs4_file_operations,
8505 	.getroot	= nfs4_proc_get_root,
8506 	.submount	= nfs4_submount,
8507 	.try_mount	= nfs4_try_mount,
8508 	.getattr	= nfs4_proc_getattr,
8509 	.setattr	= nfs4_proc_setattr,
8510 	.lookup		= nfs4_proc_lookup,
8511 	.access		= nfs4_proc_access,
8512 	.readlink	= nfs4_proc_readlink,
8513 	.create		= nfs4_proc_create,
8514 	.remove		= nfs4_proc_remove,
8515 	.unlink_setup	= nfs4_proc_unlink_setup,
8516 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8517 	.unlink_done	= nfs4_proc_unlink_done,
8518 	.rename_setup	= nfs4_proc_rename_setup,
8519 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8520 	.rename_done	= nfs4_proc_rename_done,
8521 	.link		= nfs4_proc_link,
8522 	.symlink	= nfs4_proc_symlink,
8523 	.mkdir		= nfs4_proc_mkdir,
8524 	.rmdir		= nfs4_proc_remove,
8525 	.readdir	= nfs4_proc_readdir,
8526 	.mknod		= nfs4_proc_mknod,
8527 	.statfs		= nfs4_proc_statfs,
8528 	.fsinfo		= nfs4_proc_fsinfo,
8529 	.pathconf	= nfs4_proc_pathconf,
8530 	.set_capabilities = nfs4_server_capabilities,
8531 	.decode_dirent	= nfs4_decode_dirent,
8532 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8533 	.read_setup	= nfs4_proc_read_setup,
8534 	.read_done	= nfs4_read_done,
8535 	.write_setup	= nfs4_proc_write_setup,
8536 	.write_done	= nfs4_write_done,
8537 	.commit_setup	= nfs4_proc_commit_setup,
8538 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8539 	.commit_done	= nfs4_commit_done,
8540 	.lock		= nfs4_proc_lock,
8541 	.clear_acl_cache = nfs4_zap_acl_attr,
8542 	.close_context  = nfs4_close_context,
8543 	.open_context	= nfs4_atomic_open,
8544 	.have_delegation = nfs4_have_delegation,
8545 	.return_delegation = nfs4_inode_return_delegation,
8546 	.alloc_client	= nfs4_alloc_client,
8547 	.init_client	= nfs4_init_client,
8548 	.free_client	= nfs4_free_client,
8549 	.create_server	= nfs4_create_server,
8550 	.clone_server	= nfs_clone_server,
8551 };
8552 
8553 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8554 	.prefix	= XATTR_NAME_NFSV4_ACL,
8555 	.list	= nfs4_xattr_list_nfs4_acl,
8556 	.get	= nfs4_xattr_get_nfs4_acl,
8557 	.set	= nfs4_xattr_set_nfs4_acl,
8558 };
8559 
8560 const struct xattr_handler *nfs4_xattr_handlers[] = {
8561 	&nfs4_xattr_nfs4_acl_handler,
8562 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8563 	&nfs4_xattr_nfs4_label_handler,
8564 #endif
8565 	NULL
8566 };
8567 
8568 /*
8569  * Local variables:
8570  *  c-basic-offset: 8
8571  * End:
8572  */
8573