xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 4161b450)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68 
69 #include "nfs4trace.h"
70 
71 #define NFSDBG_FACILITY		NFSDBG_PROC
72 
73 #define NFS4_POLL_RETRY_MIN	(HZ/10)
74 #define NFS4_POLL_RETRY_MAX	(15*HZ)
75 
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_FILE_OPEN:
162 		return -EBUSY;
163 	default:
164 		dprintk("%s could not handle NFSv4 error %d\n",
165 				__func__, -err);
166 		break;
167 	}
168 	return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175 	FATTR4_WORD0_TYPE
176 	| FATTR4_WORD0_CHANGE
177 	| FATTR4_WORD0_SIZE
178 	| FATTR4_WORD0_FSID
179 	| FATTR4_WORD0_FILEID,
180 	FATTR4_WORD1_MODE
181 	| FATTR4_WORD1_NUMLINKS
182 	| FATTR4_WORD1_OWNER
183 	| FATTR4_WORD1_OWNER_GROUP
184 	| FATTR4_WORD1_RAWDEV
185 	| FATTR4_WORD1_SPACE_USED
186 	| FATTR4_WORD1_TIME_ACCESS
187 	| FATTR4_WORD1_TIME_METADATA
188 	| FATTR4_WORD1_TIME_MODIFY,
189 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
190 	FATTR4_WORD2_SECURITY_LABEL
191 #endif
192 };
193 
194 static const u32 nfs4_pnfs_open_bitmap[3] = {
195 	FATTR4_WORD0_TYPE
196 	| FATTR4_WORD0_CHANGE
197 	| FATTR4_WORD0_SIZE
198 	| FATTR4_WORD0_FSID
199 	| FATTR4_WORD0_FILEID,
200 	FATTR4_WORD1_MODE
201 	| FATTR4_WORD1_NUMLINKS
202 	| FATTR4_WORD1_OWNER
203 	| FATTR4_WORD1_OWNER_GROUP
204 	| FATTR4_WORD1_RAWDEV
205 	| FATTR4_WORD1_SPACE_USED
206 	| FATTR4_WORD1_TIME_ACCESS
207 	| FATTR4_WORD1_TIME_METADATA
208 	| FATTR4_WORD1_TIME_MODIFY,
209 	FATTR4_WORD2_MDSTHRESHOLD
210 };
211 
212 static const u32 nfs4_open_noattr_bitmap[3] = {
213 	FATTR4_WORD0_TYPE
214 	| FATTR4_WORD0_CHANGE
215 	| FATTR4_WORD0_FILEID,
216 };
217 
218 const u32 nfs4_statfs_bitmap[3] = {
219 	FATTR4_WORD0_FILES_AVAIL
220 	| FATTR4_WORD0_FILES_FREE
221 	| FATTR4_WORD0_FILES_TOTAL,
222 	FATTR4_WORD1_SPACE_AVAIL
223 	| FATTR4_WORD1_SPACE_FREE
224 	| FATTR4_WORD1_SPACE_TOTAL
225 };
226 
227 const u32 nfs4_pathconf_bitmap[3] = {
228 	FATTR4_WORD0_MAXLINK
229 	| FATTR4_WORD0_MAXNAME,
230 	0
231 };
232 
233 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
234 			| FATTR4_WORD0_MAXREAD
235 			| FATTR4_WORD0_MAXWRITE
236 			| FATTR4_WORD0_LEASE_TIME,
237 			FATTR4_WORD1_TIME_DELTA
238 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
239 			FATTR4_WORD2_LAYOUT_BLKSIZE
240 };
241 
242 const u32 nfs4_fs_locations_bitmap[3] = {
243 	FATTR4_WORD0_TYPE
244 	| FATTR4_WORD0_CHANGE
245 	| FATTR4_WORD0_SIZE
246 	| FATTR4_WORD0_FSID
247 	| FATTR4_WORD0_FILEID
248 	| FATTR4_WORD0_FS_LOCATIONS,
249 	FATTR4_WORD1_MODE
250 	| FATTR4_WORD1_NUMLINKS
251 	| FATTR4_WORD1_OWNER
252 	| FATTR4_WORD1_OWNER_GROUP
253 	| FATTR4_WORD1_RAWDEV
254 	| FATTR4_WORD1_SPACE_USED
255 	| FATTR4_WORD1_TIME_ACCESS
256 	| FATTR4_WORD1_TIME_METADATA
257 	| FATTR4_WORD1_TIME_MODIFY
258 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
259 };
260 
261 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
262 		struct nfs4_readdir_arg *readdir)
263 {
264 	__be32 *start, *p;
265 
266 	if (cookie > 2) {
267 		readdir->cookie = cookie;
268 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
269 		return;
270 	}
271 
272 	readdir->cookie = 0;
273 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
274 	if (cookie == 2)
275 		return;
276 
277 	/*
278 	 * NFSv4 servers do not return entries for '.' and '..'
279 	 * Therefore, we fake these entries here.  We let '.'
280 	 * have cookie 0 and '..' have cookie 1.  Note that
281 	 * when talking to the server, we always send cookie 0
282 	 * instead of 1 or 2.
283 	 */
284 	start = p = kmap_atomic(*readdir->pages);
285 
286 	if (cookie == 0) {
287 		*p++ = xdr_one;                                  /* next */
288 		*p++ = xdr_zero;                   /* cookie, first word */
289 		*p++ = xdr_one;                   /* cookie, second word */
290 		*p++ = xdr_one;                             /* entry len */
291 		memcpy(p, ".\0\0\0", 4);                        /* entry */
292 		p++;
293 		*p++ = xdr_one;                         /* bitmap length */
294 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
295 		*p++ = htonl(8);              /* attribute buffer length */
296 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
297 	}
298 
299 	*p++ = xdr_one;                                  /* next */
300 	*p++ = xdr_zero;                   /* cookie, first word */
301 	*p++ = xdr_two;                   /* cookie, second word */
302 	*p++ = xdr_two;                             /* entry len */
303 	memcpy(p, "..\0\0", 4);                         /* entry */
304 	p++;
305 	*p++ = xdr_one;                         /* bitmap length */
306 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
307 	*p++ = htonl(8);              /* attribute buffer length */
308 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
309 
310 	readdir->pgbase = (char *)p - (char *)start;
311 	readdir->count -= readdir->pgbase;
312 	kunmap_atomic(start);
313 }
314 
315 static long nfs4_update_delay(long *timeout)
316 {
317 	long ret;
318 	if (!timeout)
319 		return NFS4_POLL_RETRY_MAX;
320 	if (*timeout <= 0)
321 		*timeout = NFS4_POLL_RETRY_MIN;
322 	if (*timeout > NFS4_POLL_RETRY_MAX)
323 		*timeout = NFS4_POLL_RETRY_MAX;
324 	ret = *timeout;
325 	*timeout <<= 1;
326 	return ret;
327 }
328 
329 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
330 {
331 	int res = 0;
332 
333 	might_sleep();
334 
335 	freezable_schedule_timeout_killable_unsafe(
336 		nfs4_update_delay(timeout));
337 	if (fatal_signal_pending(current))
338 		res = -ERESTARTSYS;
339 	return res;
340 }
341 
342 /* This is the error handling routine for processes that are allowed
343  * to sleep.
344  */
345 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
346 {
347 	struct nfs_client *clp = server->nfs_client;
348 	struct nfs4_state *state = exception->state;
349 	struct inode *inode = exception->inode;
350 	int ret = errorcode;
351 
352 	exception->retry = 0;
353 	switch(errorcode) {
354 		case 0:
355 			return 0;
356 		case -NFS4ERR_OPENMODE:
357 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
358 				nfs4_inode_return_delegation(inode);
359 				exception->retry = 1;
360 				return 0;
361 			}
362 			if (state == NULL)
363 				break;
364 			ret = nfs4_schedule_stateid_recovery(server, state);
365 			if (ret < 0)
366 				break;
367 			goto wait_on_recovery;
368 		case -NFS4ERR_DELEG_REVOKED:
369 		case -NFS4ERR_ADMIN_REVOKED:
370 		case -NFS4ERR_BAD_STATEID:
371 			if (state == NULL)
372 				break;
373 			ret = nfs4_schedule_stateid_recovery(server, state);
374 			if (ret < 0)
375 				break;
376 			goto wait_on_recovery;
377 		case -NFS4ERR_EXPIRED:
378 			if (state != NULL) {
379 				ret = nfs4_schedule_stateid_recovery(server, state);
380 				if (ret < 0)
381 					break;
382 			}
383 		case -NFS4ERR_STALE_STATEID:
384 		case -NFS4ERR_STALE_CLIENTID:
385 			nfs4_schedule_lease_recovery(clp);
386 			goto wait_on_recovery;
387 		case -NFS4ERR_MOVED:
388 			ret = nfs4_schedule_migration_recovery(server);
389 			if (ret < 0)
390 				break;
391 			goto wait_on_recovery;
392 		case -NFS4ERR_LEASE_MOVED:
393 			nfs4_schedule_lease_moved_recovery(clp);
394 			goto wait_on_recovery;
395 #if defined(CONFIG_NFS_V4_1)
396 		case -NFS4ERR_BADSESSION:
397 		case -NFS4ERR_BADSLOT:
398 		case -NFS4ERR_BAD_HIGH_SLOT:
399 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
400 		case -NFS4ERR_DEADSESSION:
401 		case -NFS4ERR_SEQ_FALSE_RETRY:
402 		case -NFS4ERR_SEQ_MISORDERED:
403 			dprintk("%s ERROR: %d Reset session\n", __func__,
404 				errorcode);
405 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
406 			goto wait_on_recovery;
407 #endif /* defined(CONFIG_NFS_V4_1) */
408 		case -NFS4ERR_FILE_OPEN:
409 			if (exception->timeout > HZ) {
410 				/* We have retried a decent amount, time to
411 				 * fail
412 				 */
413 				ret = -EBUSY;
414 				break;
415 			}
416 		case -NFS4ERR_GRACE:
417 		case -NFS4ERR_DELAY:
418 			ret = nfs4_delay(server->client, &exception->timeout);
419 			if (ret != 0)
420 				break;
421 		case -NFS4ERR_RETRY_UNCACHED_REP:
422 		case -NFS4ERR_OLD_STATEID:
423 			exception->retry = 1;
424 			break;
425 		case -NFS4ERR_BADOWNER:
426 			/* The following works around a Linux server bug! */
427 		case -NFS4ERR_BADNAME:
428 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
429 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
430 				exception->retry = 1;
431 				printk(KERN_WARNING "NFS: v4 server %s "
432 						"does not accept raw "
433 						"uid/gids. "
434 						"Reenabling the idmapper.\n",
435 						server->nfs_client->cl_hostname);
436 			}
437 	}
438 	/* We failed to handle the error */
439 	return nfs4_map_errors(ret);
440 wait_on_recovery:
441 	ret = nfs4_wait_clnt_recover(clp);
442 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
443 		return -EIO;
444 	if (ret == 0)
445 		exception->retry = 1;
446 	return ret;
447 }
448 
449 /*
450  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
451  * or 'false' otherwise.
452  */
453 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
454 {
455 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
456 
457 	if (flavor == RPC_AUTH_GSS_KRB5I ||
458 	    flavor == RPC_AUTH_GSS_KRB5P)
459 		return true;
460 
461 	return false;
462 }
463 
464 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
465 {
466 	spin_lock(&clp->cl_lock);
467 	if (time_before(clp->cl_last_renewal,timestamp))
468 		clp->cl_last_renewal = timestamp;
469 	spin_unlock(&clp->cl_lock);
470 }
471 
472 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
473 {
474 	do_renew_lease(server->nfs_client, timestamp);
475 }
476 
477 struct nfs4_call_sync_data {
478 	const struct nfs_server *seq_server;
479 	struct nfs4_sequence_args *seq_args;
480 	struct nfs4_sequence_res *seq_res;
481 };
482 
483 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
484 			       struct nfs4_sequence_res *res, int cache_reply)
485 {
486 	args->sa_slot = NULL;
487 	args->sa_cache_this = cache_reply;
488 	args->sa_privileged = 0;
489 
490 	res->sr_slot = NULL;
491 }
492 
493 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
494 {
495 	args->sa_privileged = 1;
496 }
497 
498 static int nfs40_setup_sequence(const struct nfs_server *server,
499 				struct nfs4_sequence_args *args,
500 				struct nfs4_sequence_res *res,
501 				struct rpc_task *task)
502 {
503 	struct nfs4_slot_table *tbl = server->nfs_client->cl_slot_tbl;
504 	struct nfs4_slot *slot;
505 
506 	/* slot already allocated? */
507 	if (res->sr_slot != NULL)
508 		goto out_start;
509 
510 	spin_lock(&tbl->slot_tbl_lock);
511 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
512 		goto out_sleep;
513 
514 	slot = nfs4_alloc_slot(tbl);
515 	if (IS_ERR(slot)) {
516 		if (slot == ERR_PTR(-ENOMEM))
517 			task->tk_timeout = HZ >> 2;
518 		goto out_sleep;
519 	}
520 	spin_unlock(&tbl->slot_tbl_lock);
521 
522 	args->sa_slot = slot;
523 	res->sr_slot = slot;
524 
525 out_start:
526 	rpc_call_start(task);
527 	return 0;
528 
529 out_sleep:
530 	if (args->sa_privileged)
531 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
532 				NULL, RPC_PRIORITY_PRIVILEGED);
533 	else
534 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
535 	spin_unlock(&tbl->slot_tbl_lock);
536 	return -EAGAIN;
537 }
538 
539 static int nfs40_sequence_done(struct rpc_task *task,
540 			       struct nfs4_sequence_res *res)
541 {
542 	struct nfs4_slot *slot = res->sr_slot;
543 	struct nfs4_slot_table *tbl;
544 
545 	if (slot == NULL)
546 		goto out;
547 
548 	tbl = slot->table;
549 	spin_lock(&tbl->slot_tbl_lock);
550 	if (!nfs41_wake_and_assign_slot(tbl, slot))
551 		nfs4_free_slot(tbl, slot);
552 	spin_unlock(&tbl->slot_tbl_lock);
553 
554 	res->sr_slot = NULL;
555 out:
556 	return 1;
557 }
558 
559 #if defined(CONFIG_NFS_V4_1)
560 
561 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
562 {
563 	struct nfs4_session *session;
564 	struct nfs4_slot_table *tbl;
565 	struct nfs4_slot *slot = res->sr_slot;
566 	bool send_new_highest_used_slotid = false;
567 
568 	tbl = slot->table;
569 	session = tbl->session;
570 
571 	spin_lock(&tbl->slot_tbl_lock);
572 	/* Be nice to the server: try to ensure that the last transmitted
573 	 * value for highest_user_slotid <= target_highest_slotid
574 	 */
575 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
576 		send_new_highest_used_slotid = true;
577 
578 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
579 		send_new_highest_used_slotid = false;
580 		goto out_unlock;
581 	}
582 	nfs4_free_slot(tbl, slot);
583 
584 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
585 		send_new_highest_used_slotid = false;
586 out_unlock:
587 	spin_unlock(&tbl->slot_tbl_lock);
588 	res->sr_slot = NULL;
589 	if (send_new_highest_used_slotid)
590 		nfs41_server_notify_highest_slotid_update(session->clp);
591 }
592 
593 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
594 {
595 	struct nfs4_session *session;
596 	struct nfs4_slot *slot = res->sr_slot;
597 	struct nfs_client *clp;
598 	bool interrupted = false;
599 	int ret = 1;
600 
601 	if (slot == NULL)
602 		goto out_noaction;
603 	/* don't increment the sequence number if the task wasn't sent */
604 	if (!RPC_WAS_SENT(task))
605 		goto out;
606 
607 	session = slot->table->session;
608 
609 	if (slot->interrupted) {
610 		slot->interrupted = 0;
611 		interrupted = true;
612 	}
613 
614 	trace_nfs4_sequence_done(session, res);
615 	/* Check the SEQUENCE operation status */
616 	switch (res->sr_status) {
617 	case 0:
618 		/* Update the slot's sequence and clientid lease timer */
619 		++slot->seq_nr;
620 		clp = session->clp;
621 		do_renew_lease(clp, res->sr_timestamp);
622 		/* Check sequence flags */
623 		if (res->sr_status_flags != 0)
624 			nfs4_schedule_lease_recovery(clp);
625 		nfs41_update_target_slotid(slot->table, slot, res);
626 		break;
627 	case 1:
628 		/*
629 		 * sr_status remains 1 if an RPC level error occurred.
630 		 * The server may or may not have processed the sequence
631 		 * operation..
632 		 * Mark the slot as having hosted an interrupted RPC call.
633 		 */
634 		slot->interrupted = 1;
635 		goto out;
636 	case -NFS4ERR_DELAY:
637 		/* The server detected a resend of the RPC call and
638 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
639 		 * of RFC5661.
640 		 */
641 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
642 			__func__,
643 			slot->slot_nr,
644 			slot->seq_nr);
645 		goto out_retry;
646 	case -NFS4ERR_BADSLOT:
647 		/*
648 		 * The slot id we used was probably retired. Try again
649 		 * using a different slot id.
650 		 */
651 		goto retry_nowait;
652 	case -NFS4ERR_SEQ_MISORDERED:
653 		/*
654 		 * Was the last operation on this sequence interrupted?
655 		 * If so, retry after bumping the sequence number.
656 		 */
657 		if (interrupted) {
658 			++slot->seq_nr;
659 			goto retry_nowait;
660 		}
661 		/*
662 		 * Could this slot have been previously retired?
663 		 * If so, then the server may be expecting seq_nr = 1!
664 		 */
665 		if (slot->seq_nr != 1) {
666 			slot->seq_nr = 1;
667 			goto retry_nowait;
668 		}
669 		break;
670 	case -NFS4ERR_SEQ_FALSE_RETRY:
671 		++slot->seq_nr;
672 		goto retry_nowait;
673 	default:
674 		/* Just update the slot sequence no. */
675 		++slot->seq_nr;
676 	}
677 out:
678 	/* The session may be reset by one of the error handlers. */
679 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
680 	nfs41_sequence_free_slot(res);
681 out_noaction:
682 	return ret;
683 retry_nowait:
684 	if (rpc_restart_call_prepare(task)) {
685 		task->tk_status = 0;
686 		ret = 0;
687 	}
688 	goto out;
689 out_retry:
690 	if (!rpc_restart_call(task))
691 		goto out;
692 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
693 	return 0;
694 }
695 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
696 
697 static int nfs4_sequence_done(struct rpc_task *task,
698 			       struct nfs4_sequence_res *res)
699 {
700 	if (res->sr_slot == NULL)
701 		return 1;
702 	if (!res->sr_slot->table->session)
703 		return nfs40_sequence_done(task, res);
704 	return nfs41_sequence_done(task, res);
705 }
706 
707 int nfs41_setup_sequence(struct nfs4_session *session,
708 				struct nfs4_sequence_args *args,
709 				struct nfs4_sequence_res *res,
710 				struct rpc_task *task)
711 {
712 	struct nfs4_slot *slot;
713 	struct nfs4_slot_table *tbl;
714 
715 	dprintk("--> %s\n", __func__);
716 	/* slot already allocated? */
717 	if (res->sr_slot != NULL)
718 		goto out_success;
719 
720 	tbl = &session->fc_slot_table;
721 
722 	task->tk_timeout = 0;
723 
724 	spin_lock(&tbl->slot_tbl_lock);
725 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
726 	    !args->sa_privileged) {
727 		/* The state manager will wait until the slot table is empty */
728 		dprintk("%s session is draining\n", __func__);
729 		goto out_sleep;
730 	}
731 
732 	slot = nfs4_alloc_slot(tbl);
733 	if (IS_ERR(slot)) {
734 		/* If out of memory, try again in 1/4 second */
735 		if (slot == ERR_PTR(-ENOMEM))
736 			task->tk_timeout = HZ >> 2;
737 		dprintk("<-- %s: no free slots\n", __func__);
738 		goto out_sleep;
739 	}
740 	spin_unlock(&tbl->slot_tbl_lock);
741 
742 	args->sa_slot = slot;
743 
744 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
745 			slot->slot_nr, slot->seq_nr);
746 
747 	res->sr_slot = slot;
748 	res->sr_timestamp = jiffies;
749 	res->sr_status_flags = 0;
750 	/*
751 	 * sr_status is only set in decode_sequence, and so will remain
752 	 * set to 1 if an rpc level failure occurs.
753 	 */
754 	res->sr_status = 1;
755 	trace_nfs4_setup_sequence(session, args);
756 out_success:
757 	rpc_call_start(task);
758 	return 0;
759 out_sleep:
760 	/* Privileged tasks are queued with top priority */
761 	if (args->sa_privileged)
762 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
763 				NULL, RPC_PRIORITY_PRIVILEGED);
764 	else
765 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
766 	spin_unlock(&tbl->slot_tbl_lock);
767 	return -EAGAIN;
768 }
769 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
770 
771 static int nfs4_setup_sequence(const struct nfs_server *server,
772 			       struct nfs4_sequence_args *args,
773 			       struct nfs4_sequence_res *res,
774 			       struct rpc_task *task)
775 {
776 	struct nfs4_session *session = nfs4_get_session(server);
777 	int ret = 0;
778 
779 	if (!session)
780 		return nfs40_setup_sequence(server, args, res, task);
781 
782 	dprintk("--> %s clp %p session %p sr_slot %u\n",
783 		__func__, session->clp, session, res->sr_slot ?
784 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
785 
786 	ret = nfs41_setup_sequence(session, args, res, task);
787 
788 	dprintk("<-- %s status=%d\n", __func__, ret);
789 	return ret;
790 }
791 
792 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
793 {
794 	struct nfs4_call_sync_data *data = calldata;
795 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
796 
797 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
798 
799 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
800 }
801 
802 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
803 {
804 	struct nfs4_call_sync_data *data = calldata;
805 
806 	nfs41_sequence_done(task, data->seq_res);
807 }
808 
809 static const struct rpc_call_ops nfs41_call_sync_ops = {
810 	.rpc_call_prepare = nfs41_call_sync_prepare,
811 	.rpc_call_done = nfs41_call_sync_done,
812 };
813 
814 #else	/* !CONFIG_NFS_V4_1 */
815 
816 static int nfs4_setup_sequence(const struct nfs_server *server,
817 			       struct nfs4_sequence_args *args,
818 			       struct nfs4_sequence_res *res,
819 			       struct rpc_task *task)
820 {
821 	return nfs40_setup_sequence(server, args, res, task);
822 }
823 
824 static int nfs4_sequence_done(struct rpc_task *task,
825 			       struct nfs4_sequence_res *res)
826 {
827 	return nfs40_sequence_done(task, res);
828 }
829 
830 #endif	/* !CONFIG_NFS_V4_1 */
831 
832 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
833 {
834 	struct nfs4_call_sync_data *data = calldata;
835 	nfs4_setup_sequence(data->seq_server,
836 				data->seq_args, data->seq_res, task);
837 }
838 
839 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
840 {
841 	struct nfs4_call_sync_data *data = calldata;
842 	nfs4_sequence_done(task, data->seq_res);
843 }
844 
845 static const struct rpc_call_ops nfs40_call_sync_ops = {
846 	.rpc_call_prepare = nfs40_call_sync_prepare,
847 	.rpc_call_done = nfs40_call_sync_done,
848 };
849 
850 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
851 				   struct nfs_server *server,
852 				   struct rpc_message *msg,
853 				   struct nfs4_sequence_args *args,
854 				   struct nfs4_sequence_res *res)
855 {
856 	int ret;
857 	struct rpc_task *task;
858 	struct nfs_client *clp = server->nfs_client;
859 	struct nfs4_call_sync_data data = {
860 		.seq_server = server,
861 		.seq_args = args,
862 		.seq_res = res,
863 	};
864 	struct rpc_task_setup task_setup = {
865 		.rpc_client = clnt,
866 		.rpc_message = msg,
867 		.callback_ops = clp->cl_mvops->call_sync_ops,
868 		.callback_data = &data
869 	};
870 
871 	task = rpc_run_task(&task_setup);
872 	if (IS_ERR(task))
873 		ret = PTR_ERR(task);
874 	else {
875 		ret = task->tk_status;
876 		rpc_put_task(task);
877 	}
878 	return ret;
879 }
880 
881 int nfs4_call_sync(struct rpc_clnt *clnt,
882 		   struct nfs_server *server,
883 		   struct rpc_message *msg,
884 		   struct nfs4_sequence_args *args,
885 		   struct nfs4_sequence_res *res,
886 		   int cache_reply)
887 {
888 	nfs4_init_sequence(args, res, cache_reply);
889 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
890 }
891 
892 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
893 {
894 	struct nfs_inode *nfsi = NFS_I(dir);
895 
896 	spin_lock(&dir->i_lock);
897 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
898 	if (!cinfo->atomic || cinfo->before != dir->i_version)
899 		nfs_force_lookup_revalidate(dir);
900 	dir->i_version = cinfo->after;
901 	nfs_fscache_invalidate(dir);
902 	spin_unlock(&dir->i_lock);
903 }
904 
905 struct nfs4_opendata {
906 	struct kref kref;
907 	struct nfs_openargs o_arg;
908 	struct nfs_openres o_res;
909 	struct nfs_open_confirmargs c_arg;
910 	struct nfs_open_confirmres c_res;
911 	struct nfs4_string owner_name;
912 	struct nfs4_string group_name;
913 	struct nfs_fattr f_attr;
914 	struct nfs4_label *f_label;
915 	struct dentry *dir;
916 	struct dentry *dentry;
917 	struct nfs4_state_owner *owner;
918 	struct nfs4_state *state;
919 	struct iattr attrs;
920 	unsigned long timestamp;
921 	unsigned int rpc_done : 1;
922 	unsigned int file_created : 1;
923 	unsigned int is_recover : 1;
924 	int rpc_status;
925 	int cancelled;
926 };
927 
928 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
929 		int err, struct nfs4_exception *exception)
930 {
931 	if (err != -EINVAL)
932 		return false;
933 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
934 		return false;
935 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
936 	exception->retry = 1;
937 	return true;
938 }
939 
940 static enum open_claim_type4
941 nfs4_map_atomic_open_claim(struct nfs_server *server,
942 		enum open_claim_type4 claim)
943 {
944 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
945 		return claim;
946 	switch (claim) {
947 	default:
948 		return claim;
949 	case NFS4_OPEN_CLAIM_FH:
950 		return NFS4_OPEN_CLAIM_NULL;
951 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
952 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
953 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
954 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
955 	}
956 }
957 
958 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
959 {
960 	p->o_res.f_attr = &p->f_attr;
961 	p->o_res.f_label = p->f_label;
962 	p->o_res.seqid = p->o_arg.seqid;
963 	p->c_res.seqid = p->c_arg.seqid;
964 	p->o_res.server = p->o_arg.server;
965 	p->o_res.access_request = p->o_arg.access;
966 	nfs_fattr_init(&p->f_attr);
967 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
968 }
969 
970 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
971 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
972 		const struct iattr *attrs,
973 		struct nfs4_label *label,
974 		enum open_claim_type4 claim,
975 		gfp_t gfp_mask)
976 {
977 	struct dentry *parent = dget_parent(dentry);
978 	struct inode *dir = parent->d_inode;
979 	struct nfs_server *server = NFS_SERVER(dir);
980 	struct nfs4_opendata *p;
981 
982 	p = kzalloc(sizeof(*p), gfp_mask);
983 	if (p == NULL)
984 		goto err;
985 
986 	p->f_label = nfs4_label_alloc(server, gfp_mask);
987 	if (IS_ERR(p->f_label))
988 		goto err_free_p;
989 
990 	p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
991 	if (p->o_arg.seqid == NULL)
992 		goto err_free_label;
993 	nfs_sb_active(dentry->d_sb);
994 	p->dentry = dget(dentry);
995 	p->dir = parent;
996 	p->owner = sp;
997 	atomic_inc(&sp->so_count);
998 	p->o_arg.open_flags = flags;
999 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1000 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1001 	 * will return permission denied for all bits until close */
1002 	if (!(flags & O_EXCL)) {
1003 		/* ask server to check for all possible rights as results
1004 		 * are cached */
1005 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1006 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1007 	}
1008 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1009 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1010 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1011 	p->o_arg.name = &dentry->d_name;
1012 	p->o_arg.server = server;
1013 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1014 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1015 	p->o_arg.label = label;
1016 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1017 	switch (p->o_arg.claim) {
1018 	case NFS4_OPEN_CLAIM_NULL:
1019 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1020 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1021 		p->o_arg.fh = NFS_FH(dir);
1022 		break;
1023 	case NFS4_OPEN_CLAIM_PREVIOUS:
1024 	case NFS4_OPEN_CLAIM_FH:
1025 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1026 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1027 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1028 	}
1029 	if (attrs != NULL && attrs->ia_valid != 0) {
1030 		__u32 verf[2];
1031 
1032 		p->o_arg.u.attrs = &p->attrs;
1033 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1034 
1035 		verf[0] = jiffies;
1036 		verf[1] = current->pid;
1037 		memcpy(p->o_arg.u.verifier.data, verf,
1038 				sizeof(p->o_arg.u.verifier.data));
1039 	}
1040 	p->c_arg.fh = &p->o_res.fh;
1041 	p->c_arg.stateid = &p->o_res.stateid;
1042 	p->c_arg.seqid = p->o_arg.seqid;
1043 	nfs4_init_opendata_res(p);
1044 	kref_init(&p->kref);
1045 	return p;
1046 
1047 err_free_label:
1048 	nfs4_label_free(p->f_label);
1049 err_free_p:
1050 	kfree(p);
1051 err:
1052 	dput(parent);
1053 	return NULL;
1054 }
1055 
1056 static void nfs4_opendata_free(struct kref *kref)
1057 {
1058 	struct nfs4_opendata *p = container_of(kref,
1059 			struct nfs4_opendata, kref);
1060 	struct super_block *sb = p->dentry->d_sb;
1061 
1062 	nfs_free_seqid(p->o_arg.seqid);
1063 	if (p->state != NULL)
1064 		nfs4_put_open_state(p->state);
1065 	nfs4_put_state_owner(p->owner);
1066 
1067 	nfs4_label_free(p->f_label);
1068 
1069 	dput(p->dir);
1070 	dput(p->dentry);
1071 	nfs_sb_deactive(sb);
1072 	nfs_fattr_free_names(&p->f_attr);
1073 	kfree(p->f_attr.mdsthreshold);
1074 	kfree(p);
1075 }
1076 
1077 static void nfs4_opendata_put(struct nfs4_opendata *p)
1078 {
1079 	if (p != NULL)
1080 		kref_put(&p->kref, nfs4_opendata_free);
1081 }
1082 
1083 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1084 {
1085 	int ret;
1086 
1087 	ret = rpc_wait_for_completion_task(task);
1088 	return ret;
1089 }
1090 
1091 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1092 {
1093 	int ret = 0;
1094 
1095 	if (open_mode & (O_EXCL|O_TRUNC))
1096 		goto out;
1097 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1098 		case FMODE_READ:
1099 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1100 				&& state->n_rdonly != 0;
1101 			break;
1102 		case FMODE_WRITE:
1103 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1104 				&& state->n_wronly != 0;
1105 			break;
1106 		case FMODE_READ|FMODE_WRITE:
1107 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1108 				&& state->n_rdwr != 0;
1109 	}
1110 out:
1111 	return ret;
1112 }
1113 
1114 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1115 {
1116 	if (delegation == NULL)
1117 		return 0;
1118 	if ((delegation->type & fmode) != fmode)
1119 		return 0;
1120 	if (test_bit(NFS_DELEGATION_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 		spin_lock(&state->owner->so_lock);
1311 		if (can_open_cached(state, fmode, open_mode)) {
1312 			update_open_stateflags(state, fmode);
1313 			spin_unlock(&state->owner->so_lock);
1314 			goto out_return_state;
1315 		}
1316 		spin_unlock(&state->owner->so_lock);
1317 		rcu_read_lock();
1318 		delegation = rcu_dereference(nfsi->delegation);
1319 		if (!can_open_delegated(delegation, fmode)) {
1320 			rcu_read_unlock();
1321 			break;
1322 		}
1323 		/* Save the delegation */
1324 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1325 		rcu_read_unlock();
1326 		nfs_release_seqid(opendata->o_arg.seqid);
1327 		if (!opendata->is_recover) {
1328 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1329 			if (ret != 0)
1330 				goto out;
1331 		}
1332 		ret = -EAGAIN;
1333 
1334 		/* Try to update the stateid using the delegation */
1335 		if (update_open_stateid(state, NULL, &stateid, fmode))
1336 			goto out_return_state;
1337 	}
1338 out:
1339 	return ERR_PTR(ret);
1340 out_return_state:
1341 	atomic_inc(&state->count);
1342 	return state;
1343 }
1344 
1345 static void
1346 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1347 {
1348 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1349 	struct nfs_delegation *delegation;
1350 	int delegation_flags = 0;
1351 
1352 	rcu_read_lock();
1353 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1354 	if (delegation)
1355 		delegation_flags = delegation->flags;
1356 	rcu_read_unlock();
1357 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1358 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1359 				   "returning a delegation for "
1360 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1361 				   clp->cl_hostname);
1362 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1363 		nfs_inode_set_delegation(state->inode,
1364 					 data->owner->so_cred,
1365 					 &data->o_res);
1366 	else
1367 		nfs_inode_reclaim_delegation(state->inode,
1368 					     data->owner->so_cred,
1369 					     &data->o_res);
1370 }
1371 
1372 /*
1373  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1374  * and update the nfs4_state.
1375  */
1376 static struct nfs4_state *
1377 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1378 {
1379 	struct inode *inode = data->state->inode;
1380 	struct nfs4_state *state = data->state;
1381 	int ret;
1382 
1383 	if (!data->rpc_done) {
1384 		if (data->rpc_status) {
1385 			ret = data->rpc_status;
1386 			goto err;
1387 		}
1388 		/* cached opens have already been processed */
1389 		goto update;
1390 	}
1391 
1392 	ret = nfs_refresh_inode(inode, &data->f_attr);
1393 	if (ret)
1394 		goto err;
1395 
1396 	if (data->o_res.delegation_type != 0)
1397 		nfs4_opendata_check_deleg(data, state);
1398 update:
1399 	update_open_stateid(state, &data->o_res.stateid, NULL,
1400 			    data->o_arg.fmode);
1401 	atomic_inc(&state->count);
1402 
1403 	return state;
1404 err:
1405 	return ERR_PTR(ret);
1406 
1407 }
1408 
1409 static struct nfs4_state *
1410 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1411 {
1412 	struct inode *inode;
1413 	struct nfs4_state *state = NULL;
1414 	int ret;
1415 
1416 	if (!data->rpc_done) {
1417 		state = nfs4_try_open_cached(data);
1418 		goto out;
1419 	}
1420 
1421 	ret = -EAGAIN;
1422 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1423 		goto err;
1424 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1425 	ret = PTR_ERR(inode);
1426 	if (IS_ERR(inode))
1427 		goto err;
1428 	ret = -ENOMEM;
1429 	state = nfs4_get_open_state(inode, data->owner);
1430 	if (state == NULL)
1431 		goto err_put_inode;
1432 	if (data->o_res.delegation_type != 0)
1433 		nfs4_opendata_check_deleg(data, state);
1434 	update_open_stateid(state, &data->o_res.stateid, NULL,
1435 			data->o_arg.fmode);
1436 	iput(inode);
1437 out:
1438 	nfs_release_seqid(data->o_arg.seqid);
1439 	return state;
1440 err_put_inode:
1441 	iput(inode);
1442 err:
1443 	return ERR_PTR(ret);
1444 }
1445 
1446 static struct nfs4_state *
1447 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1448 {
1449 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1450 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1451 	return _nfs4_opendata_to_nfs4_state(data);
1452 }
1453 
1454 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1455 {
1456 	struct nfs_inode *nfsi = NFS_I(state->inode);
1457 	struct nfs_open_context *ctx;
1458 
1459 	spin_lock(&state->inode->i_lock);
1460 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1461 		if (ctx->state != state)
1462 			continue;
1463 		get_nfs_open_context(ctx);
1464 		spin_unlock(&state->inode->i_lock);
1465 		return ctx;
1466 	}
1467 	spin_unlock(&state->inode->i_lock);
1468 	return ERR_PTR(-ENOENT);
1469 }
1470 
1471 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1472 		struct nfs4_state *state, enum open_claim_type4 claim)
1473 {
1474 	struct nfs4_opendata *opendata;
1475 
1476 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1477 			NULL, NULL, claim, GFP_NOFS);
1478 	if (opendata == NULL)
1479 		return ERR_PTR(-ENOMEM);
1480 	opendata->state = state;
1481 	atomic_inc(&state->count);
1482 	return opendata;
1483 }
1484 
1485 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1486 {
1487 	struct nfs4_state *newstate;
1488 	int ret;
1489 
1490 	opendata->o_arg.open_flags = 0;
1491 	opendata->o_arg.fmode = fmode;
1492 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1493 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1494 	nfs4_init_opendata_res(opendata);
1495 	ret = _nfs4_recover_proc_open(opendata);
1496 	if (ret != 0)
1497 		return ret;
1498 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1499 	if (IS_ERR(newstate))
1500 		return PTR_ERR(newstate);
1501 	nfs4_close_state(newstate, fmode);
1502 	*res = newstate;
1503 	return 0;
1504 }
1505 
1506 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1507 {
1508 	struct nfs4_state *newstate;
1509 	int ret;
1510 
1511 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1512 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1513 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1514 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1515 	/* memory barrier prior to reading state->n_* */
1516 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1517 	clear_bit(NFS_OPEN_STATE, &state->flags);
1518 	smp_rmb();
1519 	if (state->n_rdwr != 0) {
1520 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1521 		if (ret != 0)
1522 			return ret;
1523 		if (newstate != state)
1524 			return -ESTALE;
1525 	}
1526 	if (state->n_wronly != 0) {
1527 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1528 		if (ret != 0)
1529 			return ret;
1530 		if (newstate != state)
1531 			return -ESTALE;
1532 	}
1533 	if (state->n_rdonly != 0) {
1534 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1535 		if (ret != 0)
1536 			return ret;
1537 		if (newstate != state)
1538 			return -ESTALE;
1539 	}
1540 	/*
1541 	 * We may have performed cached opens for all three recoveries.
1542 	 * Check if we need to update the current stateid.
1543 	 */
1544 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1545 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1546 		write_seqlock(&state->seqlock);
1547 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1548 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1549 		write_sequnlock(&state->seqlock);
1550 	}
1551 	return 0;
1552 }
1553 
1554 /*
1555  * OPEN_RECLAIM:
1556  * 	reclaim state on the server after a reboot.
1557  */
1558 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1559 {
1560 	struct nfs_delegation *delegation;
1561 	struct nfs4_opendata *opendata;
1562 	fmode_t delegation_type = 0;
1563 	int status;
1564 
1565 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1566 			NFS4_OPEN_CLAIM_PREVIOUS);
1567 	if (IS_ERR(opendata))
1568 		return PTR_ERR(opendata);
1569 	rcu_read_lock();
1570 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1571 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1572 		delegation_type = delegation->type;
1573 	rcu_read_unlock();
1574 	opendata->o_arg.u.delegation_type = delegation_type;
1575 	status = nfs4_open_recover(opendata, state);
1576 	nfs4_opendata_put(opendata);
1577 	return status;
1578 }
1579 
1580 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1581 {
1582 	struct nfs_server *server = NFS_SERVER(state->inode);
1583 	struct nfs4_exception exception = { };
1584 	int err;
1585 	do {
1586 		err = _nfs4_do_open_reclaim(ctx, state);
1587 		trace_nfs4_open_reclaim(ctx, 0, err);
1588 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1589 			continue;
1590 		if (err != -NFS4ERR_DELAY)
1591 			break;
1592 		nfs4_handle_exception(server, err, &exception);
1593 	} while (exception.retry);
1594 	return err;
1595 }
1596 
1597 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1598 {
1599 	struct nfs_open_context *ctx;
1600 	int ret;
1601 
1602 	ctx = nfs4_state_find_open_context(state);
1603 	if (IS_ERR(ctx))
1604 		return -EAGAIN;
1605 	ret = nfs4_do_open_reclaim(ctx, state);
1606 	put_nfs_open_context(ctx);
1607 	return ret;
1608 }
1609 
1610 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1611 {
1612 	switch (err) {
1613 		default:
1614 			printk(KERN_ERR "NFS: %s: unhandled error "
1615 					"%d.\n", __func__, err);
1616 		case 0:
1617 		case -ENOENT:
1618 		case -ESTALE:
1619 			break;
1620 		case -NFS4ERR_BADSESSION:
1621 		case -NFS4ERR_BADSLOT:
1622 		case -NFS4ERR_BAD_HIGH_SLOT:
1623 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1624 		case -NFS4ERR_DEADSESSION:
1625 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1626 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1627 			return -EAGAIN;
1628 		case -NFS4ERR_STALE_CLIENTID:
1629 		case -NFS4ERR_STALE_STATEID:
1630 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1631 		case -NFS4ERR_EXPIRED:
1632 			/* Don't recall a delegation if it was lost */
1633 			nfs4_schedule_lease_recovery(server->nfs_client);
1634 			return -EAGAIN;
1635 		case -NFS4ERR_MOVED:
1636 			nfs4_schedule_migration_recovery(server);
1637 			return -EAGAIN;
1638 		case -NFS4ERR_LEASE_MOVED:
1639 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1640 			return -EAGAIN;
1641 		case -NFS4ERR_DELEG_REVOKED:
1642 		case -NFS4ERR_ADMIN_REVOKED:
1643 		case -NFS4ERR_BAD_STATEID:
1644 		case -NFS4ERR_OPENMODE:
1645 			nfs_inode_find_state_and_recover(state->inode,
1646 					stateid);
1647 			nfs4_schedule_stateid_recovery(server, state);
1648 			return -EAGAIN;
1649 		case -NFS4ERR_DELAY:
1650 		case -NFS4ERR_GRACE:
1651 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1652 			ssleep(1);
1653 			return -EAGAIN;
1654 		case -ENOMEM:
1655 		case -NFS4ERR_DENIED:
1656 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1657 			return 0;
1658 	}
1659 	return err;
1660 }
1661 
1662 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1663 {
1664 	struct nfs_server *server = NFS_SERVER(state->inode);
1665 	struct nfs4_opendata *opendata;
1666 	int err;
1667 
1668 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1669 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1670 	if (IS_ERR(opendata))
1671 		return PTR_ERR(opendata);
1672 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1673 	err = nfs4_open_recover(opendata, state);
1674 	nfs4_opendata_put(opendata);
1675 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1676 }
1677 
1678 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1679 {
1680 	struct nfs4_opendata *data = calldata;
1681 
1682 	nfs40_setup_sequence(data->o_arg.server, &data->c_arg.seq_args,
1683 				&data->c_res.seq_res, task);
1684 }
1685 
1686 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1687 {
1688 	struct nfs4_opendata *data = calldata;
1689 
1690 	nfs40_sequence_done(task, &data->c_res.seq_res);
1691 
1692 	data->rpc_status = task->tk_status;
1693 	if (data->rpc_status == 0) {
1694 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1695 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1696 		renew_lease(data->o_res.server, data->timestamp);
1697 		data->rpc_done = 1;
1698 	}
1699 }
1700 
1701 static void nfs4_open_confirm_release(void *calldata)
1702 {
1703 	struct nfs4_opendata *data = calldata;
1704 	struct nfs4_state *state = NULL;
1705 
1706 	/* If this request hasn't been cancelled, do nothing */
1707 	if (data->cancelled == 0)
1708 		goto out_free;
1709 	/* In case of error, no cleanup! */
1710 	if (!data->rpc_done)
1711 		goto out_free;
1712 	state = nfs4_opendata_to_nfs4_state(data);
1713 	if (!IS_ERR(state))
1714 		nfs4_close_state(state, data->o_arg.fmode);
1715 out_free:
1716 	nfs4_opendata_put(data);
1717 }
1718 
1719 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1720 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1721 	.rpc_call_done = nfs4_open_confirm_done,
1722 	.rpc_release = nfs4_open_confirm_release,
1723 };
1724 
1725 /*
1726  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1727  */
1728 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1729 {
1730 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1731 	struct rpc_task *task;
1732 	struct  rpc_message msg = {
1733 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1734 		.rpc_argp = &data->c_arg,
1735 		.rpc_resp = &data->c_res,
1736 		.rpc_cred = data->owner->so_cred,
1737 	};
1738 	struct rpc_task_setup task_setup_data = {
1739 		.rpc_client = server->client,
1740 		.rpc_message = &msg,
1741 		.callback_ops = &nfs4_open_confirm_ops,
1742 		.callback_data = data,
1743 		.workqueue = nfsiod_workqueue,
1744 		.flags = RPC_TASK_ASYNC,
1745 	};
1746 	int status;
1747 
1748 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1749 	kref_get(&data->kref);
1750 	data->rpc_done = 0;
1751 	data->rpc_status = 0;
1752 	data->timestamp = jiffies;
1753 	task = rpc_run_task(&task_setup_data);
1754 	if (IS_ERR(task))
1755 		return PTR_ERR(task);
1756 	status = nfs4_wait_for_completion_rpc_task(task);
1757 	if (status != 0) {
1758 		data->cancelled = 1;
1759 		smp_wmb();
1760 	} else
1761 		status = data->rpc_status;
1762 	rpc_put_task(task);
1763 	return status;
1764 }
1765 
1766 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1767 {
1768 	struct nfs4_opendata *data = calldata;
1769 	struct nfs4_state_owner *sp = data->owner;
1770 	struct nfs_client *clp = sp->so_server->nfs_client;
1771 
1772 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1773 		goto out_wait;
1774 	/*
1775 	 * Check if we still need to send an OPEN call, or if we can use
1776 	 * a delegation instead.
1777 	 */
1778 	if (data->state != NULL) {
1779 		struct nfs_delegation *delegation;
1780 
1781 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1782 			goto out_no_action;
1783 		rcu_read_lock();
1784 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1785 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1786 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1787 		    can_open_delegated(delegation, data->o_arg.fmode))
1788 			goto unlock_no_action;
1789 		rcu_read_unlock();
1790 	}
1791 	/* Update client id. */
1792 	data->o_arg.clientid = clp->cl_clientid;
1793 	switch (data->o_arg.claim) {
1794 	case NFS4_OPEN_CLAIM_PREVIOUS:
1795 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1796 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1797 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1798 	case NFS4_OPEN_CLAIM_FH:
1799 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1800 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1801 	}
1802 	data->timestamp = jiffies;
1803 	if (nfs4_setup_sequence(data->o_arg.server,
1804 				&data->o_arg.seq_args,
1805 				&data->o_res.seq_res,
1806 				task) != 0)
1807 		nfs_release_seqid(data->o_arg.seqid);
1808 
1809 	/* Set the create mode (note dependency on the session type) */
1810 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1811 	if (data->o_arg.open_flags & O_EXCL) {
1812 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1813 		if (nfs4_has_persistent_session(clp))
1814 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1815 		else if (clp->cl_mvops->minor_version > 0)
1816 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1817 	}
1818 	return;
1819 unlock_no_action:
1820 	rcu_read_unlock();
1821 out_no_action:
1822 	task->tk_action = NULL;
1823 out_wait:
1824 	nfs4_sequence_done(task, &data->o_res.seq_res);
1825 }
1826 
1827 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1828 {
1829 	struct nfs4_opendata *data = calldata;
1830 
1831 	data->rpc_status = task->tk_status;
1832 
1833 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1834 		return;
1835 
1836 	if (task->tk_status == 0) {
1837 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1838 			switch (data->o_res.f_attr->mode & S_IFMT) {
1839 			case S_IFREG:
1840 				break;
1841 			case S_IFLNK:
1842 				data->rpc_status = -ELOOP;
1843 				break;
1844 			case S_IFDIR:
1845 				data->rpc_status = -EISDIR;
1846 				break;
1847 			default:
1848 				data->rpc_status = -ENOTDIR;
1849 			}
1850 		}
1851 		renew_lease(data->o_res.server, data->timestamp);
1852 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1853 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1854 	}
1855 	data->rpc_done = 1;
1856 }
1857 
1858 static void nfs4_open_release(void *calldata)
1859 {
1860 	struct nfs4_opendata *data = calldata;
1861 	struct nfs4_state *state = NULL;
1862 
1863 	/* If this request hasn't been cancelled, do nothing */
1864 	if (data->cancelled == 0)
1865 		goto out_free;
1866 	/* In case of error, no cleanup! */
1867 	if (data->rpc_status != 0 || !data->rpc_done)
1868 		goto out_free;
1869 	/* In case we need an open_confirm, no cleanup! */
1870 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1871 		goto out_free;
1872 	state = nfs4_opendata_to_nfs4_state(data);
1873 	if (!IS_ERR(state))
1874 		nfs4_close_state(state, data->o_arg.fmode);
1875 out_free:
1876 	nfs4_opendata_put(data);
1877 }
1878 
1879 static const struct rpc_call_ops nfs4_open_ops = {
1880 	.rpc_call_prepare = nfs4_open_prepare,
1881 	.rpc_call_done = nfs4_open_done,
1882 	.rpc_release = nfs4_open_release,
1883 };
1884 
1885 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1886 {
1887 	struct inode *dir = data->dir->d_inode;
1888 	struct nfs_server *server = NFS_SERVER(dir);
1889 	struct nfs_openargs *o_arg = &data->o_arg;
1890 	struct nfs_openres *o_res = &data->o_res;
1891 	struct rpc_task *task;
1892 	struct rpc_message msg = {
1893 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1894 		.rpc_argp = o_arg,
1895 		.rpc_resp = o_res,
1896 		.rpc_cred = data->owner->so_cred,
1897 	};
1898 	struct rpc_task_setup task_setup_data = {
1899 		.rpc_client = server->client,
1900 		.rpc_message = &msg,
1901 		.callback_ops = &nfs4_open_ops,
1902 		.callback_data = data,
1903 		.workqueue = nfsiod_workqueue,
1904 		.flags = RPC_TASK_ASYNC,
1905 	};
1906 	int status;
1907 
1908 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1909 	kref_get(&data->kref);
1910 	data->rpc_done = 0;
1911 	data->rpc_status = 0;
1912 	data->cancelled = 0;
1913 	data->is_recover = 0;
1914 	if (isrecover) {
1915 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1916 		data->is_recover = 1;
1917 	}
1918 	task = rpc_run_task(&task_setup_data);
1919         if (IS_ERR(task))
1920                 return PTR_ERR(task);
1921         status = nfs4_wait_for_completion_rpc_task(task);
1922         if (status != 0) {
1923                 data->cancelled = 1;
1924                 smp_wmb();
1925         } else
1926                 status = data->rpc_status;
1927         rpc_put_task(task);
1928 
1929 	return status;
1930 }
1931 
1932 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1933 {
1934 	struct inode *dir = data->dir->d_inode;
1935 	struct nfs_openres *o_res = &data->o_res;
1936         int status;
1937 
1938 	status = nfs4_run_open_task(data, 1);
1939 	if (status != 0 || !data->rpc_done)
1940 		return status;
1941 
1942 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1943 
1944 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1945 		status = _nfs4_proc_open_confirm(data);
1946 		if (status != 0)
1947 			return status;
1948 	}
1949 
1950 	return status;
1951 }
1952 
1953 /*
1954  * Additional permission checks in order to distinguish between an
1955  * open for read, and an open for execute. This works around the
1956  * fact that NFSv4 OPEN treats read and execute permissions as being
1957  * the same.
1958  * Note that in the non-execute case, we want to turn off permission
1959  * checking if we just created a new file (POSIX open() semantics).
1960  */
1961 static int nfs4_opendata_access(struct rpc_cred *cred,
1962 				struct nfs4_opendata *opendata,
1963 				struct nfs4_state *state, fmode_t fmode,
1964 				int openflags)
1965 {
1966 	struct nfs_access_entry cache;
1967 	u32 mask;
1968 
1969 	/* access call failed or for some reason the server doesn't
1970 	 * support any access modes -- defer access call until later */
1971 	if (opendata->o_res.access_supported == 0)
1972 		return 0;
1973 
1974 	mask = 0;
1975 	/*
1976 	 * Use openflags to check for exec, because fmode won't
1977 	 * always have FMODE_EXEC set when file open for exec.
1978 	 */
1979 	if (openflags & __FMODE_EXEC) {
1980 		/* ONLY check for exec rights */
1981 		mask = MAY_EXEC;
1982 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
1983 		mask = MAY_READ;
1984 
1985 	cache.cred = cred;
1986 	cache.jiffies = jiffies;
1987 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
1988 	nfs_access_add_cache(state->inode, &cache);
1989 
1990 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1991 		return 0;
1992 
1993 	/* even though OPEN succeeded, access is denied. Close the file */
1994 	nfs4_close_state(state, fmode);
1995 	return -EACCES;
1996 }
1997 
1998 /*
1999  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2000  */
2001 static int _nfs4_proc_open(struct nfs4_opendata *data)
2002 {
2003 	struct inode *dir = data->dir->d_inode;
2004 	struct nfs_server *server = NFS_SERVER(dir);
2005 	struct nfs_openargs *o_arg = &data->o_arg;
2006 	struct nfs_openres *o_res = &data->o_res;
2007 	int status;
2008 
2009 	status = nfs4_run_open_task(data, 0);
2010 	if (!data->rpc_done)
2011 		return status;
2012 	if (status != 0) {
2013 		if (status == -NFS4ERR_BADNAME &&
2014 				!(o_arg->open_flags & O_CREAT))
2015 			return -ENOENT;
2016 		return status;
2017 	}
2018 
2019 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2020 
2021 	if (o_arg->open_flags & O_CREAT) {
2022 		update_changeattr(dir, &o_res->cinfo);
2023 		if (o_arg->open_flags & O_EXCL)
2024 			data->file_created = 1;
2025 		else if (o_res->cinfo.before != o_res->cinfo.after)
2026 			data->file_created = 1;
2027 	}
2028 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2029 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2030 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2031 		status = _nfs4_proc_open_confirm(data);
2032 		if (status != 0)
2033 			return status;
2034 	}
2035 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2036 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2037 	return 0;
2038 }
2039 
2040 static int nfs4_recover_expired_lease(struct nfs_server *server)
2041 {
2042 	return nfs4_client_recover_expired_lease(server->nfs_client);
2043 }
2044 
2045 /*
2046  * OPEN_EXPIRED:
2047  * 	reclaim state on the server after a network partition.
2048  * 	Assumes caller holds the appropriate lock
2049  */
2050 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2051 {
2052 	struct nfs4_opendata *opendata;
2053 	int ret;
2054 
2055 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2056 			NFS4_OPEN_CLAIM_FH);
2057 	if (IS_ERR(opendata))
2058 		return PTR_ERR(opendata);
2059 	ret = nfs4_open_recover(opendata, state);
2060 	if (ret == -ESTALE)
2061 		d_drop(ctx->dentry);
2062 	nfs4_opendata_put(opendata);
2063 	return ret;
2064 }
2065 
2066 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2067 {
2068 	struct nfs_server *server = NFS_SERVER(state->inode);
2069 	struct nfs4_exception exception = { };
2070 	int err;
2071 
2072 	do {
2073 		err = _nfs4_open_expired(ctx, state);
2074 		trace_nfs4_open_expired(ctx, 0, err);
2075 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2076 			continue;
2077 		switch (err) {
2078 		default:
2079 			goto out;
2080 		case -NFS4ERR_GRACE:
2081 		case -NFS4ERR_DELAY:
2082 			nfs4_handle_exception(server, err, &exception);
2083 			err = 0;
2084 		}
2085 	} while (exception.retry);
2086 out:
2087 	return err;
2088 }
2089 
2090 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2091 {
2092 	struct nfs_open_context *ctx;
2093 	int ret;
2094 
2095 	ctx = nfs4_state_find_open_context(state);
2096 	if (IS_ERR(ctx))
2097 		return -EAGAIN;
2098 	ret = nfs4_do_open_expired(ctx, state);
2099 	put_nfs_open_context(ctx);
2100 	return ret;
2101 }
2102 
2103 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2104 {
2105 	nfs_remove_bad_delegation(state->inode);
2106 	write_seqlock(&state->seqlock);
2107 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2108 	write_sequnlock(&state->seqlock);
2109 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2110 }
2111 
2112 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2113 {
2114 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2115 		nfs_finish_clear_delegation_stateid(state);
2116 }
2117 
2118 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2119 {
2120 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2121 	nfs40_clear_delegation_stateid(state);
2122 	return nfs4_open_expired(sp, state);
2123 }
2124 
2125 #if defined(CONFIG_NFS_V4_1)
2126 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2127 {
2128 	struct nfs_server *server = NFS_SERVER(state->inode);
2129 	nfs4_stateid stateid;
2130 	struct nfs_delegation *delegation;
2131 	struct rpc_cred *cred;
2132 	int status;
2133 
2134 	/* Get the delegation credential for use by test/free_stateid */
2135 	rcu_read_lock();
2136 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2137 	if (delegation == NULL) {
2138 		rcu_read_unlock();
2139 		return;
2140 	}
2141 
2142 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2143 	cred = get_rpccred(delegation->cred);
2144 	rcu_read_unlock();
2145 	status = nfs41_test_stateid(server, &stateid, cred);
2146 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2147 
2148 	if (status != NFS_OK) {
2149 		/* Free the stateid unless the server explicitly
2150 		 * informs us the stateid is unrecognized. */
2151 		if (status != -NFS4ERR_BAD_STATEID)
2152 			nfs41_free_stateid(server, &stateid, cred);
2153 		nfs_finish_clear_delegation_stateid(state);
2154 	}
2155 
2156 	put_rpccred(cred);
2157 }
2158 
2159 /**
2160  * nfs41_check_open_stateid - possibly free an open stateid
2161  *
2162  * @state: NFSv4 state for an inode
2163  *
2164  * Returns NFS_OK if recovery for this stateid is now finished.
2165  * Otherwise a negative NFS4ERR value is returned.
2166  */
2167 static int nfs41_check_open_stateid(struct nfs4_state *state)
2168 {
2169 	struct nfs_server *server = NFS_SERVER(state->inode);
2170 	nfs4_stateid *stateid = &state->open_stateid;
2171 	struct rpc_cred *cred = state->owner->so_cred;
2172 	int status;
2173 
2174 	/* If a state reset has been done, test_stateid is unneeded */
2175 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2176 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2177 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2178 		return -NFS4ERR_BAD_STATEID;
2179 
2180 	status = nfs41_test_stateid(server, stateid, cred);
2181 	trace_nfs4_test_open_stateid(state, NULL, status);
2182 	if (status != NFS_OK) {
2183 		/* Free the stateid unless the server explicitly
2184 		 * informs us the stateid is unrecognized. */
2185 		if (status != -NFS4ERR_BAD_STATEID)
2186 			nfs41_free_stateid(server, stateid, cred);
2187 
2188 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2189 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2190 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2191 		clear_bit(NFS_OPEN_STATE, &state->flags);
2192 	}
2193 	return status;
2194 }
2195 
2196 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2197 {
2198 	int status;
2199 
2200 	nfs41_check_delegation_stateid(state);
2201 	status = nfs41_check_open_stateid(state);
2202 	if (status != NFS_OK)
2203 		status = nfs4_open_expired(sp, state);
2204 	return status;
2205 }
2206 #endif
2207 
2208 /*
2209  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2210  * fields corresponding to attributes that were used to store the verifier.
2211  * Make sure we clobber those fields in the later setattr call
2212  */
2213 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2214 {
2215 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2216 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2217 		sattr->ia_valid |= ATTR_ATIME;
2218 
2219 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2220 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2221 		sattr->ia_valid |= ATTR_MTIME;
2222 }
2223 
2224 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2225 		fmode_t fmode,
2226 		int flags,
2227 		struct nfs_open_context *ctx)
2228 {
2229 	struct nfs4_state_owner *sp = opendata->owner;
2230 	struct nfs_server *server = sp->so_server;
2231 	struct dentry *dentry;
2232 	struct nfs4_state *state;
2233 	unsigned int seq;
2234 	int ret;
2235 
2236 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2237 
2238 	ret = _nfs4_proc_open(opendata);
2239 	if (ret != 0)
2240 		goto out;
2241 
2242 	state = nfs4_opendata_to_nfs4_state(opendata);
2243 	ret = PTR_ERR(state);
2244 	if (IS_ERR(state))
2245 		goto out;
2246 	if (server->caps & NFS_CAP_POSIX_LOCK)
2247 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2248 
2249 	dentry = opendata->dentry;
2250 	if (dentry->d_inode == NULL) {
2251 		/* FIXME: Is this d_drop() ever needed? */
2252 		d_drop(dentry);
2253 		dentry = d_add_unique(dentry, igrab(state->inode));
2254 		if (dentry == NULL) {
2255 			dentry = opendata->dentry;
2256 		} else if (dentry != ctx->dentry) {
2257 			dput(ctx->dentry);
2258 			ctx->dentry = dget(dentry);
2259 		}
2260 		nfs_set_verifier(dentry,
2261 				nfs_save_change_attribute(opendata->dir->d_inode));
2262 	}
2263 
2264 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2265 	if (ret != 0)
2266 		goto out;
2267 
2268 	ctx->state = state;
2269 	if (dentry->d_inode == state->inode) {
2270 		nfs_inode_attach_open_context(ctx);
2271 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2272 			nfs4_schedule_stateid_recovery(server, state);
2273 	}
2274 out:
2275 	return ret;
2276 }
2277 
2278 /*
2279  * Returns a referenced nfs4_state
2280  */
2281 static int _nfs4_do_open(struct inode *dir,
2282 			struct nfs_open_context *ctx,
2283 			int flags,
2284 			struct iattr *sattr,
2285 			struct nfs4_label *label,
2286 			int *opened)
2287 {
2288 	struct nfs4_state_owner  *sp;
2289 	struct nfs4_state     *state = NULL;
2290 	struct nfs_server       *server = NFS_SERVER(dir);
2291 	struct nfs4_opendata *opendata;
2292 	struct dentry *dentry = ctx->dentry;
2293 	struct rpc_cred *cred = ctx->cred;
2294 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2295 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2296 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2297 	struct nfs4_label *olabel = NULL;
2298 	int status;
2299 
2300 	/* Protect against reboot recovery conflicts */
2301 	status = -ENOMEM;
2302 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2303 	if (sp == NULL) {
2304 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2305 		goto out_err;
2306 	}
2307 	status = nfs4_recover_expired_lease(server);
2308 	if (status != 0)
2309 		goto err_put_state_owner;
2310 	if (dentry->d_inode != NULL)
2311 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2312 	status = -ENOMEM;
2313 	if (dentry->d_inode)
2314 		claim = NFS4_OPEN_CLAIM_FH;
2315 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2316 			label, claim, GFP_KERNEL);
2317 	if (opendata == NULL)
2318 		goto err_put_state_owner;
2319 
2320 	if (label) {
2321 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2322 		if (IS_ERR(olabel)) {
2323 			status = PTR_ERR(olabel);
2324 			goto err_opendata_put;
2325 		}
2326 	}
2327 
2328 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2329 		if (!opendata->f_attr.mdsthreshold) {
2330 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2331 			if (!opendata->f_attr.mdsthreshold)
2332 				goto err_free_label;
2333 		}
2334 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2335 	}
2336 	if (dentry->d_inode != NULL)
2337 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2338 
2339 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2340 	if (status != 0)
2341 		goto err_free_label;
2342 	state = ctx->state;
2343 
2344 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2345 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2346 		nfs4_exclusive_attrset(opendata, sattr);
2347 
2348 		nfs_fattr_init(opendata->o_res.f_attr);
2349 		status = nfs4_do_setattr(state->inode, cred,
2350 				opendata->o_res.f_attr, sattr,
2351 				state, label, olabel);
2352 		if (status == 0) {
2353 			nfs_setattr_update_inode(state->inode, sattr);
2354 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2355 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2356 		}
2357 	}
2358 	if (opendata->file_created)
2359 		*opened |= FILE_CREATED;
2360 
2361 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2362 		*ctx_th = opendata->f_attr.mdsthreshold;
2363 		opendata->f_attr.mdsthreshold = NULL;
2364 	}
2365 
2366 	nfs4_label_free(olabel);
2367 
2368 	nfs4_opendata_put(opendata);
2369 	nfs4_put_state_owner(sp);
2370 	return 0;
2371 err_free_label:
2372 	nfs4_label_free(olabel);
2373 err_opendata_put:
2374 	nfs4_opendata_put(opendata);
2375 err_put_state_owner:
2376 	nfs4_put_state_owner(sp);
2377 out_err:
2378 	return status;
2379 }
2380 
2381 
2382 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2383 					struct nfs_open_context *ctx,
2384 					int flags,
2385 					struct iattr *sattr,
2386 					struct nfs4_label *label,
2387 					int *opened)
2388 {
2389 	struct nfs_server *server = NFS_SERVER(dir);
2390 	struct nfs4_exception exception = { };
2391 	struct nfs4_state *res;
2392 	int status;
2393 
2394 	do {
2395 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2396 		res = ctx->state;
2397 		trace_nfs4_open_file(ctx, flags, status);
2398 		if (status == 0)
2399 			break;
2400 		/* NOTE: BAD_SEQID means the server and client disagree about the
2401 		 * book-keeping w.r.t. state-changing operations
2402 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2403 		 * It is actually a sign of a bug on the client or on the server.
2404 		 *
2405 		 * If we receive a BAD_SEQID error in the particular case of
2406 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2407 		 * have unhashed the old state_owner for us, and that we can
2408 		 * therefore safely retry using a new one. We should still warn
2409 		 * the user though...
2410 		 */
2411 		if (status == -NFS4ERR_BAD_SEQID) {
2412 			pr_warn_ratelimited("NFS: v4 server %s "
2413 					" returned a bad sequence-id error!\n",
2414 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2415 			exception.retry = 1;
2416 			continue;
2417 		}
2418 		/*
2419 		 * BAD_STATEID on OPEN means that the server cancelled our
2420 		 * state before it received the OPEN_CONFIRM.
2421 		 * Recover by retrying the request as per the discussion
2422 		 * on Page 181 of RFC3530.
2423 		 */
2424 		if (status == -NFS4ERR_BAD_STATEID) {
2425 			exception.retry = 1;
2426 			continue;
2427 		}
2428 		if (status == -EAGAIN) {
2429 			/* We must have found a delegation */
2430 			exception.retry = 1;
2431 			continue;
2432 		}
2433 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2434 			continue;
2435 		res = ERR_PTR(nfs4_handle_exception(server,
2436 					status, &exception));
2437 	} while (exception.retry);
2438 	return res;
2439 }
2440 
2441 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2442 			    struct nfs_fattr *fattr, struct iattr *sattr,
2443 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2444 			    struct nfs4_label *olabel)
2445 {
2446 	struct nfs_server *server = NFS_SERVER(inode);
2447         struct nfs_setattrargs  arg = {
2448                 .fh             = NFS_FH(inode),
2449                 .iap            = sattr,
2450 		.server		= server,
2451 		.bitmask = server->attr_bitmask,
2452 		.label		= ilabel,
2453         };
2454         struct nfs_setattrres  res = {
2455 		.fattr		= fattr,
2456 		.label		= olabel,
2457 		.server		= server,
2458         };
2459         struct rpc_message msg = {
2460 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2461 		.rpc_argp	= &arg,
2462 		.rpc_resp	= &res,
2463 		.rpc_cred	= cred,
2464         };
2465 	unsigned long timestamp = jiffies;
2466 	fmode_t fmode;
2467 	bool truncate;
2468 	int status;
2469 
2470 	arg.bitmask = nfs4_bitmask(server, ilabel);
2471 	if (ilabel)
2472 		arg.bitmask = nfs4_bitmask(server, olabel);
2473 
2474 	nfs_fattr_init(fattr);
2475 
2476 	/* Servers should only apply open mode checks for file size changes */
2477 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2478 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2479 
2480 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2481 		/* Use that stateid */
2482 	} else if (truncate && state != NULL) {
2483 		struct nfs_lockowner lockowner = {
2484 			.l_owner = current->files,
2485 			.l_pid = current->tgid,
2486 		};
2487 		if (!nfs4_valid_open_stateid(state))
2488 			return -EBADF;
2489 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2490 				&lockowner) == -EIO)
2491 			return -EBADF;
2492 	} else
2493 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2494 
2495 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2496 	if (status == 0 && state != NULL)
2497 		renew_lease(server, timestamp);
2498 	return status;
2499 }
2500 
2501 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2502 			   struct nfs_fattr *fattr, struct iattr *sattr,
2503 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2504 			   struct nfs4_label *olabel)
2505 {
2506 	struct nfs_server *server = NFS_SERVER(inode);
2507 	struct nfs4_exception exception = {
2508 		.state = state,
2509 		.inode = inode,
2510 	};
2511 	int err;
2512 	do {
2513 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2514 		trace_nfs4_setattr(inode, err);
2515 		switch (err) {
2516 		case -NFS4ERR_OPENMODE:
2517 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2518 				pr_warn_once("NFSv4: server %s is incorrectly "
2519 						"applying open mode checks to "
2520 						"a SETATTR that is not "
2521 						"changing file size.\n",
2522 						server->nfs_client->cl_hostname);
2523 			}
2524 			if (state && !(state->state & FMODE_WRITE)) {
2525 				err = -EBADF;
2526 				if (sattr->ia_valid & ATTR_OPEN)
2527 					err = -EACCES;
2528 				goto out;
2529 			}
2530 		}
2531 		err = nfs4_handle_exception(server, err, &exception);
2532 	} while (exception.retry);
2533 out:
2534 	return err;
2535 }
2536 
2537 struct nfs4_closedata {
2538 	struct inode *inode;
2539 	struct nfs4_state *state;
2540 	struct nfs_closeargs arg;
2541 	struct nfs_closeres res;
2542 	struct nfs_fattr fattr;
2543 	unsigned long timestamp;
2544 	bool roc;
2545 	u32 roc_barrier;
2546 };
2547 
2548 static void nfs4_free_closedata(void *data)
2549 {
2550 	struct nfs4_closedata *calldata = data;
2551 	struct nfs4_state_owner *sp = calldata->state->owner;
2552 	struct super_block *sb = calldata->state->inode->i_sb;
2553 
2554 	if (calldata->roc)
2555 		pnfs_roc_release(calldata->state->inode);
2556 	nfs4_put_open_state(calldata->state);
2557 	nfs_free_seqid(calldata->arg.seqid);
2558 	nfs4_put_state_owner(sp);
2559 	nfs_sb_deactive(sb);
2560 	kfree(calldata);
2561 }
2562 
2563 static void nfs4_close_done(struct rpc_task *task, void *data)
2564 {
2565 	struct nfs4_closedata *calldata = data;
2566 	struct nfs4_state *state = calldata->state;
2567 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2568 	nfs4_stateid *res_stateid = NULL;
2569 
2570 	dprintk("%s: begin!\n", __func__);
2571 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2572 		return;
2573 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2574         /* hmm. we are done with the inode, and in the process of freeing
2575 	 * the state_owner. we keep this around to process errors
2576 	 */
2577 	switch (task->tk_status) {
2578 		case 0:
2579 			res_stateid = &calldata->res.stateid;
2580 			if (calldata->arg.fmode == 0 && calldata->roc)
2581 				pnfs_roc_set_barrier(state->inode,
2582 						     calldata->roc_barrier);
2583 			renew_lease(server, calldata->timestamp);
2584 			break;
2585 		case -NFS4ERR_ADMIN_REVOKED:
2586 		case -NFS4ERR_STALE_STATEID:
2587 		case -NFS4ERR_OLD_STATEID:
2588 		case -NFS4ERR_BAD_STATEID:
2589 		case -NFS4ERR_EXPIRED:
2590 			if (calldata->arg.fmode == 0)
2591 				break;
2592 		default:
2593 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2594 				rpc_restart_call_prepare(task);
2595 				goto out_release;
2596 			}
2597 	}
2598 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2599 out_release:
2600 	nfs_release_seqid(calldata->arg.seqid);
2601 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2602 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2603 }
2604 
2605 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2606 {
2607 	struct nfs4_closedata *calldata = data;
2608 	struct nfs4_state *state = calldata->state;
2609 	struct inode *inode = calldata->inode;
2610 	bool is_rdonly, is_wronly, is_rdwr;
2611 	int call_close = 0;
2612 
2613 	dprintk("%s: begin!\n", __func__);
2614 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2615 		goto out_wait;
2616 
2617 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2618 	spin_lock(&state->owner->so_lock);
2619 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2620 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2621 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2622 	/* Calculate the change in open mode */
2623 	calldata->arg.fmode = 0;
2624 	if (state->n_rdwr == 0) {
2625 		if (state->n_rdonly == 0)
2626 			call_close |= is_rdonly;
2627 		else if (is_rdonly)
2628 			calldata->arg.fmode |= FMODE_READ;
2629 		if (state->n_wronly == 0)
2630 			call_close |= is_wronly;
2631 		else if (is_wronly)
2632 			calldata->arg.fmode |= FMODE_WRITE;
2633 	} else if (is_rdwr)
2634 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2635 
2636 	if (calldata->arg.fmode == 0)
2637 		call_close |= is_rdwr;
2638 
2639 	if (!nfs4_valid_open_stateid(state))
2640 		call_close = 0;
2641 	spin_unlock(&state->owner->so_lock);
2642 
2643 	if (!call_close) {
2644 		/* Note: exit _without_ calling nfs4_close_done */
2645 		goto out_no_action;
2646 	}
2647 
2648 	if (calldata->arg.fmode == 0) {
2649 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2650 		if (calldata->roc &&
2651 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2652 			nfs_release_seqid(calldata->arg.seqid);
2653 			goto out_wait;
2654 		    }
2655 	}
2656 
2657 	nfs_fattr_init(calldata->res.fattr);
2658 	calldata->timestamp = jiffies;
2659 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2660 				&calldata->arg.seq_args,
2661 				&calldata->res.seq_res,
2662 				task) != 0)
2663 		nfs_release_seqid(calldata->arg.seqid);
2664 	dprintk("%s: done!\n", __func__);
2665 	return;
2666 out_no_action:
2667 	task->tk_action = NULL;
2668 out_wait:
2669 	nfs4_sequence_done(task, &calldata->res.seq_res);
2670 }
2671 
2672 static const struct rpc_call_ops nfs4_close_ops = {
2673 	.rpc_call_prepare = nfs4_close_prepare,
2674 	.rpc_call_done = nfs4_close_done,
2675 	.rpc_release = nfs4_free_closedata,
2676 };
2677 
2678 static bool nfs4_state_has_opener(struct nfs4_state *state)
2679 {
2680 	/* first check existing openers */
2681 	if (test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0 &&
2682 	    state->n_rdonly != 0)
2683 		return true;
2684 
2685 	if (test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0 &&
2686 	    state->n_wronly != 0)
2687 		return true;
2688 
2689 	if (test_bit(NFS_O_RDWR_STATE, &state->flags) != 0 &&
2690 	    state->n_rdwr != 0)
2691 		return true;
2692 
2693 	return false;
2694 }
2695 
2696 static bool nfs4_roc(struct inode *inode)
2697 {
2698 	struct nfs_inode *nfsi = NFS_I(inode);
2699 	struct nfs_open_context *ctx;
2700 	struct nfs4_state *state;
2701 
2702 	spin_lock(&inode->i_lock);
2703 	list_for_each_entry(ctx, &nfsi->open_files, list) {
2704 		state = ctx->state;
2705 		if (state == NULL)
2706 			continue;
2707 		if (nfs4_state_has_opener(state)) {
2708 			spin_unlock(&inode->i_lock);
2709 			return false;
2710 		}
2711 	}
2712 	spin_unlock(&inode->i_lock);
2713 
2714 	if (nfs4_check_delegation(inode, FMODE_READ))
2715 		return false;
2716 
2717 	return pnfs_roc(inode);
2718 }
2719 
2720 /*
2721  * It is possible for data to be read/written from a mem-mapped file
2722  * after the sys_close call (which hits the vfs layer as a flush).
2723  * This means that we can't safely call nfsv4 close on a file until
2724  * the inode is cleared. This in turn means that we are not good
2725  * NFSv4 citizens - we do not indicate to the server to update the file's
2726  * share state even when we are done with one of the three share
2727  * stateid's in the inode.
2728  *
2729  * NOTE: Caller must be holding the sp->so_owner semaphore!
2730  */
2731 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2732 {
2733 	struct nfs_server *server = NFS_SERVER(state->inode);
2734 	struct nfs4_closedata *calldata;
2735 	struct nfs4_state_owner *sp = state->owner;
2736 	struct rpc_task *task;
2737 	struct rpc_message msg = {
2738 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2739 		.rpc_cred = state->owner->so_cred,
2740 	};
2741 	struct rpc_task_setup task_setup_data = {
2742 		.rpc_client = server->client,
2743 		.rpc_message = &msg,
2744 		.callback_ops = &nfs4_close_ops,
2745 		.workqueue = nfsiod_workqueue,
2746 		.flags = RPC_TASK_ASYNC,
2747 	};
2748 	int status = -ENOMEM;
2749 
2750 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2751 		&task_setup_data.rpc_client, &msg);
2752 
2753 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2754 	if (calldata == NULL)
2755 		goto out;
2756 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2757 	calldata->inode = state->inode;
2758 	calldata->state = state;
2759 	calldata->arg.fh = NFS_FH(state->inode);
2760 	calldata->arg.stateid = &state->open_stateid;
2761 	/* Serialization for the sequence id */
2762 	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2763 	if (calldata->arg.seqid == NULL)
2764 		goto out_free_calldata;
2765 	calldata->arg.fmode = 0;
2766 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2767 	calldata->res.fattr = &calldata->fattr;
2768 	calldata->res.seqid = calldata->arg.seqid;
2769 	calldata->res.server = server;
2770 	calldata->roc = nfs4_roc(state->inode);
2771 	nfs_sb_active(calldata->inode->i_sb);
2772 
2773 	msg.rpc_argp = &calldata->arg;
2774 	msg.rpc_resp = &calldata->res;
2775 	task_setup_data.callback_data = calldata;
2776 	task = rpc_run_task(&task_setup_data);
2777 	if (IS_ERR(task))
2778 		return PTR_ERR(task);
2779 	status = 0;
2780 	if (wait)
2781 		status = rpc_wait_for_completion_task(task);
2782 	rpc_put_task(task);
2783 	return status;
2784 out_free_calldata:
2785 	kfree(calldata);
2786 out:
2787 	nfs4_put_open_state(state);
2788 	nfs4_put_state_owner(sp);
2789 	return status;
2790 }
2791 
2792 static struct inode *
2793 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2794 		int open_flags, struct iattr *attr, int *opened)
2795 {
2796 	struct nfs4_state *state;
2797 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2798 
2799 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2800 
2801 	/* Protect against concurrent sillydeletes */
2802 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2803 
2804 	nfs4_label_release_security(label);
2805 
2806 	if (IS_ERR(state))
2807 		return ERR_CAST(state);
2808 	return state->inode;
2809 }
2810 
2811 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2812 {
2813 	if (ctx->state == NULL)
2814 		return;
2815 	if (is_sync)
2816 		nfs4_close_sync(ctx->state, ctx->mode);
2817 	else
2818 		nfs4_close_state(ctx->state, ctx->mode);
2819 }
2820 
2821 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2822 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2823 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2824 
2825 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2826 {
2827 	struct nfs4_server_caps_arg args = {
2828 		.fhandle = fhandle,
2829 	};
2830 	struct nfs4_server_caps_res res = {};
2831 	struct rpc_message msg = {
2832 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2833 		.rpc_argp = &args,
2834 		.rpc_resp = &res,
2835 	};
2836 	int status;
2837 
2838 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2839 	if (status == 0) {
2840 		/* Sanity check the server answers */
2841 		switch (server->nfs_client->cl_minorversion) {
2842 		case 0:
2843 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2844 			res.attr_bitmask[2] = 0;
2845 			break;
2846 		case 1:
2847 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2848 			break;
2849 		case 2:
2850 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2851 		}
2852 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2853 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2854 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2855 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2856 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2857 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2858 				NFS_CAP_SECURITY_LABEL);
2859 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2860 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2861 			server->caps |= NFS_CAP_ACLS;
2862 		if (res.has_links != 0)
2863 			server->caps |= NFS_CAP_HARDLINKS;
2864 		if (res.has_symlinks != 0)
2865 			server->caps |= NFS_CAP_SYMLINKS;
2866 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2867 			server->caps |= NFS_CAP_FILEID;
2868 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2869 			server->caps |= NFS_CAP_MODE;
2870 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2871 			server->caps |= NFS_CAP_NLINK;
2872 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2873 			server->caps |= NFS_CAP_OWNER;
2874 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2875 			server->caps |= NFS_CAP_OWNER_GROUP;
2876 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2877 			server->caps |= NFS_CAP_ATIME;
2878 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2879 			server->caps |= NFS_CAP_CTIME;
2880 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2881 			server->caps |= NFS_CAP_MTIME;
2882 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2883 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2884 			server->caps |= NFS_CAP_SECURITY_LABEL;
2885 #endif
2886 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2887 				sizeof(server->attr_bitmask));
2888 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2889 
2890 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2891 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2892 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2893 		server->cache_consistency_bitmask[2] = 0;
2894 		server->acl_bitmask = res.acl_bitmask;
2895 		server->fh_expire_type = res.fh_expire_type;
2896 	}
2897 
2898 	return status;
2899 }
2900 
2901 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2902 {
2903 	struct nfs4_exception exception = { };
2904 	int err;
2905 	do {
2906 		err = nfs4_handle_exception(server,
2907 				_nfs4_server_capabilities(server, fhandle),
2908 				&exception);
2909 	} while (exception.retry);
2910 	return err;
2911 }
2912 
2913 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2914 		struct nfs_fsinfo *info)
2915 {
2916 	u32 bitmask[3];
2917 	struct nfs4_lookup_root_arg args = {
2918 		.bitmask = bitmask,
2919 	};
2920 	struct nfs4_lookup_res res = {
2921 		.server = server,
2922 		.fattr = info->fattr,
2923 		.fh = fhandle,
2924 	};
2925 	struct rpc_message msg = {
2926 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2927 		.rpc_argp = &args,
2928 		.rpc_resp = &res,
2929 	};
2930 
2931 	bitmask[0] = nfs4_fattr_bitmap[0];
2932 	bitmask[1] = nfs4_fattr_bitmap[1];
2933 	/*
2934 	 * Process the label in the upcoming getfattr
2935 	 */
2936 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2937 
2938 	nfs_fattr_init(info->fattr);
2939 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2940 }
2941 
2942 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2943 		struct nfs_fsinfo *info)
2944 {
2945 	struct nfs4_exception exception = { };
2946 	int err;
2947 	do {
2948 		err = _nfs4_lookup_root(server, fhandle, info);
2949 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2950 		switch (err) {
2951 		case 0:
2952 		case -NFS4ERR_WRONGSEC:
2953 			goto out;
2954 		default:
2955 			err = nfs4_handle_exception(server, err, &exception);
2956 		}
2957 	} while (exception.retry);
2958 out:
2959 	return err;
2960 }
2961 
2962 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2963 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2964 {
2965 	struct rpc_auth_create_args auth_args = {
2966 		.pseudoflavor = flavor,
2967 	};
2968 	struct rpc_auth *auth;
2969 	int ret;
2970 
2971 	auth = rpcauth_create(&auth_args, server->client);
2972 	if (IS_ERR(auth)) {
2973 		ret = -EACCES;
2974 		goto out;
2975 	}
2976 	ret = nfs4_lookup_root(server, fhandle, info);
2977 out:
2978 	return ret;
2979 }
2980 
2981 /*
2982  * Retry pseudoroot lookup with various security flavors.  We do this when:
2983  *
2984  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2985  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2986  *
2987  * Returns zero on success, or a negative NFS4ERR value, or a
2988  * negative errno value.
2989  */
2990 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2991 			      struct nfs_fsinfo *info)
2992 {
2993 	/* Per 3530bis 15.33.5 */
2994 	static const rpc_authflavor_t flav_array[] = {
2995 		RPC_AUTH_GSS_KRB5P,
2996 		RPC_AUTH_GSS_KRB5I,
2997 		RPC_AUTH_GSS_KRB5,
2998 		RPC_AUTH_UNIX,			/* courtesy */
2999 		RPC_AUTH_NULL,
3000 	};
3001 	int status = -EPERM;
3002 	size_t i;
3003 
3004 	if (server->auth_info.flavor_len > 0) {
3005 		/* try each flavor specified by user */
3006 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3007 			status = nfs4_lookup_root_sec(server, fhandle, info,
3008 						server->auth_info.flavors[i]);
3009 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3010 				continue;
3011 			break;
3012 		}
3013 	} else {
3014 		/* no flavors specified by user, try default list */
3015 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3016 			status = nfs4_lookup_root_sec(server, fhandle, info,
3017 						      flav_array[i]);
3018 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3019 				continue;
3020 			break;
3021 		}
3022 	}
3023 
3024 	/*
3025 	 * -EACCESS could mean that the user doesn't have correct permissions
3026 	 * to access the mount.  It could also mean that we tried to mount
3027 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3028 	 * existing mount programs don't handle -EACCES very well so it should
3029 	 * be mapped to -EPERM instead.
3030 	 */
3031 	if (status == -EACCES)
3032 		status = -EPERM;
3033 	return status;
3034 }
3035 
3036 static int nfs4_do_find_root_sec(struct nfs_server *server,
3037 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3038 {
3039 	int mv = server->nfs_client->cl_minorversion;
3040 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3041 }
3042 
3043 /**
3044  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3045  * @server: initialized nfs_server handle
3046  * @fhandle: we fill in the pseudo-fs root file handle
3047  * @info: we fill in an FSINFO struct
3048  * @auth_probe: probe the auth flavours
3049  *
3050  * Returns zero on success, or a negative errno.
3051  */
3052 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3053 			 struct nfs_fsinfo *info,
3054 			 bool auth_probe)
3055 {
3056 	int status;
3057 
3058 	switch (auth_probe) {
3059 	case false:
3060 		status = nfs4_lookup_root(server, fhandle, info);
3061 		if (status != -NFS4ERR_WRONGSEC)
3062 			break;
3063 	default:
3064 		status = nfs4_do_find_root_sec(server, fhandle, info);
3065 	}
3066 
3067 	if (status == 0)
3068 		status = nfs4_server_capabilities(server, fhandle);
3069 	if (status == 0)
3070 		status = nfs4_do_fsinfo(server, fhandle, info);
3071 
3072 	return nfs4_map_errors(status);
3073 }
3074 
3075 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3076 			      struct nfs_fsinfo *info)
3077 {
3078 	int error;
3079 	struct nfs_fattr *fattr = info->fattr;
3080 	struct nfs4_label *label = NULL;
3081 
3082 	error = nfs4_server_capabilities(server, mntfh);
3083 	if (error < 0) {
3084 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3085 		return error;
3086 	}
3087 
3088 	label = nfs4_label_alloc(server, GFP_KERNEL);
3089 	if (IS_ERR(label))
3090 		return PTR_ERR(label);
3091 
3092 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3093 	if (error < 0) {
3094 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3095 		goto err_free_label;
3096 	}
3097 
3098 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3099 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3100 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3101 
3102 err_free_label:
3103 	nfs4_label_free(label);
3104 
3105 	return error;
3106 }
3107 
3108 /*
3109  * Get locations and (maybe) other attributes of a referral.
3110  * Note that we'll actually follow the referral later when
3111  * we detect fsid mismatch in inode revalidation
3112  */
3113 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3114 			     const struct qstr *name, struct nfs_fattr *fattr,
3115 			     struct nfs_fh *fhandle)
3116 {
3117 	int status = -ENOMEM;
3118 	struct page *page = NULL;
3119 	struct nfs4_fs_locations *locations = NULL;
3120 
3121 	page = alloc_page(GFP_KERNEL);
3122 	if (page == NULL)
3123 		goto out;
3124 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3125 	if (locations == NULL)
3126 		goto out;
3127 
3128 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3129 	if (status != 0)
3130 		goto out;
3131 
3132 	/*
3133 	 * If the fsid didn't change, this is a migration event, not a
3134 	 * referral.  Cause us to drop into the exception handler, which
3135 	 * will kick off migration recovery.
3136 	 */
3137 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3138 		dprintk("%s: server did not return a different fsid for"
3139 			" a referral at %s\n", __func__, name->name);
3140 		status = -NFS4ERR_MOVED;
3141 		goto out;
3142 	}
3143 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3144 	nfs_fixup_referral_attributes(&locations->fattr);
3145 
3146 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3147 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3148 	memset(fhandle, 0, sizeof(struct nfs_fh));
3149 out:
3150 	if (page)
3151 		__free_page(page);
3152 	kfree(locations);
3153 	return status;
3154 }
3155 
3156 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3157 				struct nfs_fattr *fattr, struct nfs4_label *label)
3158 {
3159 	struct nfs4_getattr_arg args = {
3160 		.fh = fhandle,
3161 		.bitmask = server->attr_bitmask,
3162 	};
3163 	struct nfs4_getattr_res res = {
3164 		.fattr = fattr,
3165 		.label = label,
3166 		.server = server,
3167 	};
3168 	struct rpc_message msg = {
3169 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3170 		.rpc_argp = &args,
3171 		.rpc_resp = &res,
3172 	};
3173 
3174 	args.bitmask = nfs4_bitmask(server, label);
3175 
3176 	nfs_fattr_init(fattr);
3177 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3178 }
3179 
3180 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3181 				struct nfs_fattr *fattr, struct nfs4_label *label)
3182 {
3183 	struct nfs4_exception exception = { };
3184 	int err;
3185 	do {
3186 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3187 		trace_nfs4_getattr(server, fhandle, fattr, err);
3188 		err = nfs4_handle_exception(server, err,
3189 				&exception);
3190 	} while (exception.retry);
3191 	return err;
3192 }
3193 
3194 /*
3195  * The file is not closed if it is opened due to the a request to change
3196  * the size of the file. The open call will not be needed once the
3197  * VFS layer lookup-intents are implemented.
3198  *
3199  * Close is called when the inode is destroyed.
3200  * If we haven't opened the file for O_WRONLY, we
3201  * need to in the size_change case to obtain a stateid.
3202  *
3203  * Got race?
3204  * Because OPEN is always done by name in nfsv4, it is
3205  * possible that we opened a different file by the same
3206  * name.  We can recognize this race condition, but we
3207  * can't do anything about it besides returning an error.
3208  *
3209  * This will be fixed with VFS changes (lookup-intent).
3210  */
3211 static int
3212 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3213 		  struct iattr *sattr)
3214 {
3215 	struct inode *inode = dentry->d_inode;
3216 	struct rpc_cred *cred = NULL;
3217 	struct nfs4_state *state = NULL;
3218 	struct nfs4_label *label = NULL;
3219 	int status;
3220 
3221 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3222 	    sattr->ia_valid & ATTR_SIZE &&
3223 	    sattr->ia_size < i_size_read(inode))
3224 		pnfs_commit_and_return_layout(inode);
3225 
3226 	nfs_fattr_init(fattr);
3227 
3228 	/* Deal with open(O_TRUNC) */
3229 	if (sattr->ia_valid & ATTR_OPEN)
3230 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3231 
3232 	/* Optimization: if the end result is no change, don't RPC */
3233 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3234 		return 0;
3235 
3236 	/* Search for an existing open(O_WRITE) file */
3237 	if (sattr->ia_valid & ATTR_FILE) {
3238 		struct nfs_open_context *ctx;
3239 
3240 		ctx = nfs_file_open_context(sattr->ia_file);
3241 		if (ctx) {
3242 			cred = ctx->cred;
3243 			state = ctx->state;
3244 		}
3245 	}
3246 
3247 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3248 	if (IS_ERR(label))
3249 		return PTR_ERR(label);
3250 
3251 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3252 	if (status == 0) {
3253 		nfs_setattr_update_inode(inode, sattr);
3254 		nfs_setsecurity(inode, fattr, label);
3255 	}
3256 	nfs4_label_free(label);
3257 	return status;
3258 }
3259 
3260 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3261 		const struct qstr *name, struct nfs_fh *fhandle,
3262 		struct nfs_fattr *fattr, struct nfs4_label *label)
3263 {
3264 	struct nfs_server *server = NFS_SERVER(dir);
3265 	int		       status;
3266 	struct nfs4_lookup_arg args = {
3267 		.bitmask = server->attr_bitmask,
3268 		.dir_fh = NFS_FH(dir),
3269 		.name = name,
3270 	};
3271 	struct nfs4_lookup_res res = {
3272 		.server = server,
3273 		.fattr = fattr,
3274 		.label = label,
3275 		.fh = fhandle,
3276 	};
3277 	struct rpc_message msg = {
3278 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3279 		.rpc_argp = &args,
3280 		.rpc_resp = &res,
3281 	};
3282 
3283 	args.bitmask = nfs4_bitmask(server, label);
3284 
3285 	nfs_fattr_init(fattr);
3286 
3287 	dprintk("NFS call  lookup %s\n", name->name);
3288 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3289 	dprintk("NFS reply lookup: %d\n", status);
3290 	return status;
3291 }
3292 
3293 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3294 {
3295 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3296 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3297 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3298 	fattr->nlink = 2;
3299 }
3300 
3301 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3302 				   struct qstr *name, struct nfs_fh *fhandle,
3303 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3304 {
3305 	struct nfs4_exception exception = { };
3306 	struct rpc_clnt *client = *clnt;
3307 	int err;
3308 	do {
3309 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3310 		trace_nfs4_lookup(dir, name, err);
3311 		switch (err) {
3312 		case -NFS4ERR_BADNAME:
3313 			err = -ENOENT;
3314 			goto out;
3315 		case -NFS4ERR_MOVED:
3316 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3317 			goto out;
3318 		case -NFS4ERR_WRONGSEC:
3319 			err = -EPERM;
3320 			if (client != *clnt)
3321 				goto out;
3322 			client = nfs4_negotiate_security(client, dir, name);
3323 			if (IS_ERR(client))
3324 				return PTR_ERR(client);
3325 
3326 			exception.retry = 1;
3327 			break;
3328 		default:
3329 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3330 		}
3331 	} while (exception.retry);
3332 
3333 out:
3334 	if (err == 0)
3335 		*clnt = client;
3336 	else if (client != *clnt)
3337 		rpc_shutdown_client(client);
3338 
3339 	return err;
3340 }
3341 
3342 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3343 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3344 			    struct nfs4_label *label)
3345 {
3346 	int status;
3347 	struct rpc_clnt *client = NFS_CLIENT(dir);
3348 
3349 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3350 	if (client != NFS_CLIENT(dir)) {
3351 		rpc_shutdown_client(client);
3352 		nfs_fixup_secinfo_attributes(fattr);
3353 	}
3354 	return status;
3355 }
3356 
3357 struct rpc_clnt *
3358 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3359 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3360 {
3361 	struct rpc_clnt *client = NFS_CLIENT(dir);
3362 	int status;
3363 
3364 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3365 	if (status < 0)
3366 		return ERR_PTR(status);
3367 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3368 }
3369 
3370 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3371 {
3372 	struct nfs_server *server = NFS_SERVER(inode);
3373 	struct nfs4_accessargs args = {
3374 		.fh = NFS_FH(inode),
3375 		.bitmask = server->cache_consistency_bitmask,
3376 	};
3377 	struct nfs4_accessres res = {
3378 		.server = server,
3379 	};
3380 	struct rpc_message msg = {
3381 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3382 		.rpc_argp = &args,
3383 		.rpc_resp = &res,
3384 		.rpc_cred = entry->cred,
3385 	};
3386 	int mode = entry->mask;
3387 	int status = 0;
3388 
3389 	/*
3390 	 * Determine which access bits we want to ask for...
3391 	 */
3392 	if (mode & MAY_READ)
3393 		args.access |= NFS4_ACCESS_READ;
3394 	if (S_ISDIR(inode->i_mode)) {
3395 		if (mode & MAY_WRITE)
3396 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3397 		if (mode & MAY_EXEC)
3398 			args.access |= NFS4_ACCESS_LOOKUP;
3399 	} else {
3400 		if (mode & MAY_WRITE)
3401 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3402 		if (mode & MAY_EXEC)
3403 			args.access |= NFS4_ACCESS_EXECUTE;
3404 	}
3405 
3406 	res.fattr = nfs_alloc_fattr();
3407 	if (res.fattr == NULL)
3408 		return -ENOMEM;
3409 
3410 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3411 	if (!status) {
3412 		nfs_access_set_mask(entry, res.access);
3413 		nfs_refresh_inode(inode, res.fattr);
3414 	}
3415 	nfs_free_fattr(res.fattr);
3416 	return status;
3417 }
3418 
3419 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3420 {
3421 	struct nfs4_exception exception = { };
3422 	int err;
3423 	do {
3424 		err = _nfs4_proc_access(inode, entry);
3425 		trace_nfs4_access(inode, err);
3426 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3427 				&exception);
3428 	} while (exception.retry);
3429 	return err;
3430 }
3431 
3432 /*
3433  * TODO: For the time being, we don't try to get any attributes
3434  * along with any of the zero-copy operations READ, READDIR,
3435  * READLINK, WRITE.
3436  *
3437  * In the case of the first three, we want to put the GETATTR
3438  * after the read-type operation -- this is because it is hard
3439  * to predict the length of a GETATTR response in v4, and thus
3440  * align the READ data correctly.  This means that the GETATTR
3441  * may end up partially falling into the page cache, and we should
3442  * shift it into the 'tail' of the xdr_buf before processing.
3443  * To do this efficiently, we need to know the total length
3444  * of data received, which doesn't seem to be available outside
3445  * of the RPC layer.
3446  *
3447  * In the case of WRITE, we also want to put the GETATTR after
3448  * the operation -- in this case because we want to make sure
3449  * we get the post-operation mtime and size.
3450  *
3451  * Both of these changes to the XDR layer would in fact be quite
3452  * minor, but I decided to leave them for a subsequent patch.
3453  */
3454 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3455 		unsigned int pgbase, unsigned int pglen)
3456 {
3457 	struct nfs4_readlink args = {
3458 		.fh       = NFS_FH(inode),
3459 		.pgbase	  = pgbase,
3460 		.pglen    = pglen,
3461 		.pages    = &page,
3462 	};
3463 	struct nfs4_readlink_res res;
3464 	struct rpc_message msg = {
3465 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3466 		.rpc_argp = &args,
3467 		.rpc_resp = &res,
3468 	};
3469 
3470 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3471 }
3472 
3473 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3474 		unsigned int pgbase, unsigned int pglen)
3475 {
3476 	struct nfs4_exception exception = { };
3477 	int err;
3478 	do {
3479 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3480 		trace_nfs4_readlink(inode, err);
3481 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3482 				&exception);
3483 	} while (exception.retry);
3484 	return err;
3485 }
3486 
3487 /*
3488  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3489  */
3490 static int
3491 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3492 		 int flags)
3493 {
3494 	struct nfs4_label l, *ilabel = NULL;
3495 	struct nfs_open_context *ctx;
3496 	struct nfs4_state *state;
3497 	int opened = 0;
3498 	int status = 0;
3499 
3500 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3501 	if (IS_ERR(ctx))
3502 		return PTR_ERR(ctx);
3503 
3504 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3505 
3506 	sattr->ia_mode &= ~current_umask();
3507 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3508 	if (IS_ERR(state)) {
3509 		status = PTR_ERR(state);
3510 		goto out;
3511 	}
3512 out:
3513 	nfs4_label_release_security(ilabel);
3514 	put_nfs_open_context(ctx);
3515 	return status;
3516 }
3517 
3518 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3519 {
3520 	struct nfs_server *server = NFS_SERVER(dir);
3521 	struct nfs_removeargs args = {
3522 		.fh = NFS_FH(dir),
3523 		.name = *name,
3524 	};
3525 	struct nfs_removeres res = {
3526 		.server = server,
3527 	};
3528 	struct rpc_message msg = {
3529 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3530 		.rpc_argp = &args,
3531 		.rpc_resp = &res,
3532 	};
3533 	int status;
3534 
3535 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3536 	if (status == 0)
3537 		update_changeattr(dir, &res.cinfo);
3538 	return status;
3539 }
3540 
3541 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3542 {
3543 	struct nfs4_exception exception = { };
3544 	int err;
3545 	do {
3546 		err = _nfs4_proc_remove(dir, name);
3547 		trace_nfs4_remove(dir, name, err);
3548 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3549 				&exception);
3550 	} while (exception.retry);
3551 	return err;
3552 }
3553 
3554 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3555 {
3556 	struct nfs_server *server = NFS_SERVER(dir);
3557 	struct nfs_removeargs *args = msg->rpc_argp;
3558 	struct nfs_removeres *res = msg->rpc_resp;
3559 
3560 	res->server = server;
3561 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3562 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3563 
3564 	nfs_fattr_init(res->dir_attr);
3565 }
3566 
3567 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3568 {
3569 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3570 			&data->args.seq_args,
3571 			&data->res.seq_res,
3572 			task);
3573 }
3574 
3575 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3576 {
3577 	struct nfs_unlinkdata *data = task->tk_calldata;
3578 	struct nfs_removeres *res = &data->res;
3579 
3580 	if (!nfs4_sequence_done(task, &res->seq_res))
3581 		return 0;
3582 	if (nfs4_async_handle_error(task, res->server, NULL,
3583 				    &data->timeout) == -EAGAIN)
3584 		return 0;
3585 	update_changeattr(dir, &res->cinfo);
3586 	return 1;
3587 }
3588 
3589 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3590 {
3591 	struct nfs_server *server = NFS_SERVER(dir);
3592 	struct nfs_renameargs *arg = msg->rpc_argp;
3593 	struct nfs_renameres *res = msg->rpc_resp;
3594 
3595 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3596 	res->server = server;
3597 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3598 }
3599 
3600 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3601 {
3602 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3603 			&data->args.seq_args,
3604 			&data->res.seq_res,
3605 			task);
3606 }
3607 
3608 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3609 				 struct inode *new_dir)
3610 {
3611 	struct nfs_renamedata *data = task->tk_calldata;
3612 	struct nfs_renameres *res = &data->res;
3613 
3614 	if (!nfs4_sequence_done(task, &res->seq_res))
3615 		return 0;
3616 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3617 		return 0;
3618 
3619 	update_changeattr(old_dir, &res->old_cinfo);
3620 	update_changeattr(new_dir, &res->new_cinfo);
3621 	return 1;
3622 }
3623 
3624 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3625 {
3626 	struct nfs_server *server = NFS_SERVER(inode);
3627 	struct nfs4_link_arg arg = {
3628 		.fh     = NFS_FH(inode),
3629 		.dir_fh = NFS_FH(dir),
3630 		.name   = name,
3631 		.bitmask = server->attr_bitmask,
3632 	};
3633 	struct nfs4_link_res res = {
3634 		.server = server,
3635 		.label = NULL,
3636 	};
3637 	struct rpc_message msg = {
3638 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3639 		.rpc_argp = &arg,
3640 		.rpc_resp = &res,
3641 	};
3642 	int status = -ENOMEM;
3643 
3644 	res.fattr = nfs_alloc_fattr();
3645 	if (res.fattr == NULL)
3646 		goto out;
3647 
3648 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3649 	if (IS_ERR(res.label)) {
3650 		status = PTR_ERR(res.label);
3651 		goto out;
3652 	}
3653 	arg.bitmask = nfs4_bitmask(server, res.label);
3654 
3655 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3656 	if (!status) {
3657 		update_changeattr(dir, &res.cinfo);
3658 		status = nfs_post_op_update_inode(inode, res.fattr);
3659 		if (!status)
3660 			nfs_setsecurity(inode, res.fattr, res.label);
3661 	}
3662 
3663 
3664 	nfs4_label_free(res.label);
3665 
3666 out:
3667 	nfs_free_fattr(res.fattr);
3668 	return status;
3669 }
3670 
3671 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3672 {
3673 	struct nfs4_exception exception = { };
3674 	int err;
3675 	do {
3676 		err = nfs4_handle_exception(NFS_SERVER(inode),
3677 				_nfs4_proc_link(inode, dir, name),
3678 				&exception);
3679 	} while (exception.retry);
3680 	return err;
3681 }
3682 
3683 struct nfs4_createdata {
3684 	struct rpc_message msg;
3685 	struct nfs4_create_arg arg;
3686 	struct nfs4_create_res res;
3687 	struct nfs_fh fh;
3688 	struct nfs_fattr fattr;
3689 	struct nfs4_label *label;
3690 };
3691 
3692 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3693 		struct qstr *name, struct iattr *sattr, u32 ftype)
3694 {
3695 	struct nfs4_createdata *data;
3696 
3697 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3698 	if (data != NULL) {
3699 		struct nfs_server *server = NFS_SERVER(dir);
3700 
3701 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3702 		if (IS_ERR(data->label))
3703 			goto out_free;
3704 
3705 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3706 		data->msg.rpc_argp = &data->arg;
3707 		data->msg.rpc_resp = &data->res;
3708 		data->arg.dir_fh = NFS_FH(dir);
3709 		data->arg.server = server;
3710 		data->arg.name = name;
3711 		data->arg.attrs = sattr;
3712 		data->arg.ftype = ftype;
3713 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3714 		data->res.server = server;
3715 		data->res.fh = &data->fh;
3716 		data->res.fattr = &data->fattr;
3717 		data->res.label = data->label;
3718 		nfs_fattr_init(data->res.fattr);
3719 	}
3720 	return data;
3721 out_free:
3722 	kfree(data);
3723 	return NULL;
3724 }
3725 
3726 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3727 {
3728 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3729 				    &data->arg.seq_args, &data->res.seq_res, 1);
3730 	if (status == 0) {
3731 		update_changeattr(dir, &data->res.dir_cinfo);
3732 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3733 	}
3734 	return status;
3735 }
3736 
3737 static void nfs4_free_createdata(struct nfs4_createdata *data)
3738 {
3739 	nfs4_label_free(data->label);
3740 	kfree(data);
3741 }
3742 
3743 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3744 		struct page *page, unsigned int len, struct iattr *sattr,
3745 		struct nfs4_label *label)
3746 {
3747 	struct nfs4_createdata *data;
3748 	int status = -ENAMETOOLONG;
3749 
3750 	if (len > NFS4_MAXPATHLEN)
3751 		goto out;
3752 
3753 	status = -ENOMEM;
3754 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3755 	if (data == NULL)
3756 		goto out;
3757 
3758 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3759 	data->arg.u.symlink.pages = &page;
3760 	data->arg.u.symlink.len = len;
3761 	data->arg.label = label;
3762 
3763 	status = nfs4_do_create(dir, dentry, data);
3764 
3765 	nfs4_free_createdata(data);
3766 out:
3767 	return status;
3768 }
3769 
3770 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3771 		struct page *page, unsigned int len, struct iattr *sattr)
3772 {
3773 	struct nfs4_exception exception = { };
3774 	struct nfs4_label l, *label = NULL;
3775 	int err;
3776 
3777 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3778 
3779 	do {
3780 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3781 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3782 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3783 				&exception);
3784 	} while (exception.retry);
3785 
3786 	nfs4_label_release_security(label);
3787 	return err;
3788 }
3789 
3790 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3791 		struct iattr *sattr, struct nfs4_label *label)
3792 {
3793 	struct nfs4_createdata *data;
3794 	int status = -ENOMEM;
3795 
3796 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3797 	if (data == NULL)
3798 		goto out;
3799 
3800 	data->arg.label = label;
3801 	status = nfs4_do_create(dir, dentry, data);
3802 
3803 	nfs4_free_createdata(data);
3804 out:
3805 	return status;
3806 }
3807 
3808 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3809 		struct iattr *sattr)
3810 {
3811 	struct nfs4_exception exception = { };
3812 	struct nfs4_label l, *label = NULL;
3813 	int err;
3814 
3815 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3816 
3817 	sattr->ia_mode &= ~current_umask();
3818 	do {
3819 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3820 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3821 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3822 				&exception);
3823 	} while (exception.retry);
3824 	nfs4_label_release_security(label);
3825 
3826 	return err;
3827 }
3828 
3829 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3830 		u64 cookie, struct page **pages, unsigned int count, int plus)
3831 {
3832 	struct inode		*dir = dentry->d_inode;
3833 	struct nfs4_readdir_arg args = {
3834 		.fh = NFS_FH(dir),
3835 		.pages = pages,
3836 		.pgbase = 0,
3837 		.count = count,
3838 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3839 		.plus = plus,
3840 	};
3841 	struct nfs4_readdir_res res;
3842 	struct rpc_message msg = {
3843 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3844 		.rpc_argp = &args,
3845 		.rpc_resp = &res,
3846 		.rpc_cred = cred,
3847 	};
3848 	int			status;
3849 
3850 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3851 			dentry,
3852 			(unsigned long long)cookie);
3853 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3854 	res.pgbase = args.pgbase;
3855 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3856 	if (status >= 0) {
3857 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3858 		status += args.pgbase;
3859 	}
3860 
3861 	nfs_invalidate_atime(dir);
3862 
3863 	dprintk("%s: returns %d\n", __func__, status);
3864 	return status;
3865 }
3866 
3867 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3868 		u64 cookie, struct page **pages, unsigned int count, int plus)
3869 {
3870 	struct nfs4_exception exception = { };
3871 	int err;
3872 	do {
3873 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3874 				pages, count, plus);
3875 		trace_nfs4_readdir(dentry->d_inode, err);
3876 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3877 				&exception);
3878 	} while (exception.retry);
3879 	return err;
3880 }
3881 
3882 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3883 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3884 {
3885 	struct nfs4_createdata *data;
3886 	int mode = sattr->ia_mode;
3887 	int status = -ENOMEM;
3888 
3889 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3890 	if (data == NULL)
3891 		goto out;
3892 
3893 	if (S_ISFIFO(mode))
3894 		data->arg.ftype = NF4FIFO;
3895 	else if (S_ISBLK(mode)) {
3896 		data->arg.ftype = NF4BLK;
3897 		data->arg.u.device.specdata1 = MAJOR(rdev);
3898 		data->arg.u.device.specdata2 = MINOR(rdev);
3899 	}
3900 	else if (S_ISCHR(mode)) {
3901 		data->arg.ftype = NF4CHR;
3902 		data->arg.u.device.specdata1 = MAJOR(rdev);
3903 		data->arg.u.device.specdata2 = MINOR(rdev);
3904 	} else if (!S_ISSOCK(mode)) {
3905 		status = -EINVAL;
3906 		goto out_free;
3907 	}
3908 
3909 	data->arg.label = label;
3910 	status = nfs4_do_create(dir, dentry, data);
3911 out_free:
3912 	nfs4_free_createdata(data);
3913 out:
3914 	return status;
3915 }
3916 
3917 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3918 		struct iattr *sattr, dev_t rdev)
3919 {
3920 	struct nfs4_exception exception = { };
3921 	struct nfs4_label l, *label = NULL;
3922 	int err;
3923 
3924 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3925 
3926 	sattr->ia_mode &= ~current_umask();
3927 	do {
3928 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3929 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3930 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3931 				&exception);
3932 	} while (exception.retry);
3933 
3934 	nfs4_label_release_security(label);
3935 
3936 	return err;
3937 }
3938 
3939 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3940 		 struct nfs_fsstat *fsstat)
3941 {
3942 	struct nfs4_statfs_arg args = {
3943 		.fh = fhandle,
3944 		.bitmask = server->attr_bitmask,
3945 	};
3946 	struct nfs4_statfs_res res = {
3947 		.fsstat = fsstat,
3948 	};
3949 	struct rpc_message msg = {
3950 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3951 		.rpc_argp = &args,
3952 		.rpc_resp = &res,
3953 	};
3954 
3955 	nfs_fattr_init(fsstat->fattr);
3956 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3957 }
3958 
3959 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3960 {
3961 	struct nfs4_exception exception = { };
3962 	int err;
3963 	do {
3964 		err = nfs4_handle_exception(server,
3965 				_nfs4_proc_statfs(server, fhandle, fsstat),
3966 				&exception);
3967 	} while (exception.retry);
3968 	return err;
3969 }
3970 
3971 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3972 		struct nfs_fsinfo *fsinfo)
3973 {
3974 	struct nfs4_fsinfo_arg args = {
3975 		.fh = fhandle,
3976 		.bitmask = server->attr_bitmask,
3977 	};
3978 	struct nfs4_fsinfo_res res = {
3979 		.fsinfo = fsinfo,
3980 	};
3981 	struct rpc_message msg = {
3982 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3983 		.rpc_argp = &args,
3984 		.rpc_resp = &res,
3985 	};
3986 
3987 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3988 }
3989 
3990 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3991 {
3992 	struct nfs4_exception exception = { };
3993 	unsigned long now = jiffies;
3994 	int err;
3995 
3996 	do {
3997 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3998 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
3999 		if (err == 0) {
4000 			struct nfs_client *clp = server->nfs_client;
4001 
4002 			spin_lock(&clp->cl_lock);
4003 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4004 			clp->cl_last_renewal = now;
4005 			spin_unlock(&clp->cl_lock);
4006 			break;
4007 		}
4008 		err = nfs4_handle_exception(server, err, &exception);
4009 	} while (exception.retry);
4010 	return err;
4011 }
4012 
4013 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4014 {
4015 	int error;
4016 
4017 	nfs_fattr_init(fsinfo->fattr);
4018 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4019 	if (error == 0) {
4020 		/* block layout checks this! */
4021 		server->pnfs_blksize = fsinfo->blksize;
4022 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4023 	}
4024 
4025 	return error;
4026 }
4027 
4028 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4029 		struct nfs_pathconf *pathconf)
4030 {
4031 	struct nfs4_pathconf_arg args = {
4032 		.fh = fhandle,
4033 		.bitmask = server->attr_bitmask,
4034 	};
4035 	struct nfs4_pathconf_res res = {
4036 		.pathconf = pathconf,
4037 	};
4038 	struct rpc_message msg = {
4039 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4040 		.rpc_argp = &args,
4041 		.rpc_resp = &res,
4042 	};
4043 
4044 	/* None of the pathconf attributes are mandatory to implement */
4045 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4046 		memset(pathconf, 0, sizeof(*pathconf));
4047 		return 0;
4048 	}
4049 
4050 	nfs_fattr_init(pathconf->fattr);
4051 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4052 }
4053 
4054 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4055 		struct nfs_pathconf *pathconf)
4056 {
4057 	struct nfs4_exception exception = { };
4058 	int err;
4059 
4060 	do {
4061 		err = nfs4_handle_exception(server,
4062 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4063 				&exception);
4064 	} while (exception.retry);
4065 	return err;
4066 }
4067 
4068 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4069 		const struct nfs_open_context *ctx,
4070 		const struct nfs_lock_context *l_ctx,
4071 		fmode_t fmode)
4072 {
4073 	const struct nfs_lockowner *lockowner = NULL;
4074 
4075 	if (l_ctx != NULL)
4076 		lockowner = &l_ctx->lockowner;
4077 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4078 }
4079 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4080 
4081 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4082 		const struct nfs_open_context *ctx,
4083 		const struct nfs_lock_context *l_ctx,
4084 		fmode_t fmode)
4085 {
4086 	nfs4_stateid current_stateid;
4087 
4088 	/* If the current stateid represents a lost lock, then exit */
4089 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4090 		return true;
4091 	return nfs4_stateid_match(stateid, &current_stateid);
4092 }
4093 
4094 static bool nfs4_error_stateid_expired(int err)
4095 {
4096 	switch (err) {
4097 	case -NFS4ERR_DELEG_REVOKED:
4098 	case -NFS4ERR_ADMIN_REVOKED:
4099 	case -NFS4ERR_BAD_STATEID:
4100 	case -NFS4ERR_STALE_STATEID:
4101 	case -NFS4ERR_OLD_STATEID:
4102 	case -NFS4ERR_OPENMODE:
4103 	case -NFS4ERR_EXPIRED:
4104 		return true;
4105 	}
4106 	return false;
4107 }
4108 
4109 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4110 {
4111 	nfs_invalidate_atime(hdr->inode);
4112 }
4113 
4114 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4115 {
4116 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4117 
4118 	trace_nfs4_read(hdr, task->tk_status);
4119 	if (nfs4_async_handle_error(task, server,
4120 				    hdr->args.context->state,
4121 				    NULL) == -EAGAIN) {
4122 		rpc_restart_call_prepare(task);
4123 		return -EAGAIN;
4124 	}
4125 
4126 	__nfs4_read_done_cb(hdr);
4127 	if (task->tk_status > 0)
4128 		renew_lease(server, hdr->timestamp);
4129 	return 0;
4130 }
4131 
4132 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4133 		struct nfs_pgio_args *args)
4134 {
4135 
4136 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4137 		nfs4_stateid_is_current(&args->stateid,
4138 				args->context,
4139 				args->lock_context,
4140 				FMODE_READ))
4141 		return false;
4142 	rpc_restart_call_prepare(task);
4143 	return true;
4144 }
4145 
4146 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4147 {
4148 
4149 	dprintk("--> %s\n", __func__);
4150 
4151 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4152 		return -EAGAIN;
4153 	if (nfs4_read_stateid_changed(task, &hdr->args))
4154 		return -EAGAIN;
4155 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4156 				    nfs4_read_done_cb(task, hdr);
4157 }
4158 
4159 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4160 				 struct rpc_message *msg)
4161 {
4162 	hdr->timestamp   = jiffies;
4163 	hdr->pgio_done_cb = nfs4_read_done_cb;
4164 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4165 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4166 }
4167 
4168 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4169 				      struct nfs_pgio_header *hdr)
4170 {
4171 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4172 			&hdr->args.seq_args,
4173 			&hdr->res.seq_res,
4174 			task))
4175 		return 0;
4176 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4177 				hdr->args.lock_context,
4178 				hdr->rw_ops->rw_mode) == -EIO)
4179 		return -EIO;
4180 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4181 		return -EIO;
4182 	return 0;
4183 }
4184 
4185 static int nfs4_write_done_cb(struct rpc_task *task,
4186 			      struct nfs_pgio_header *hdr)
4187 {
4188 	struct inode *inode = hdr->inode;
4189 
4190 	trace_nfs4_write(hdr, task->tk_status);
4191 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4192 				    hdr->args.context->state,
4193 				    NULL) == -EAGAIN) {
4194 		rpc_restart_call_prepare(task);
4195 		return -EAGAIN;
4196 	}
4197 	if (task->tk_status >= 0) {
4198 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4199 		nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4200 	}
4201 	return 0;
4202 }
4203 
4204 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4205 		struct nfs_pgio_args *args)
4206 {
4207 
4208 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4209 		nfs4_stateid_is_current(&args->stateid,
4210 				args->context,
4211 				args->lock_context,
4212 				FMODE_WRITE))
4213 		return false;
4214 	rpc_restart_call_prepare(task);
4215 	return true;
4216 }
4217 
4218 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4219 {
4220 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4221 		return -EAGAIN;
4222 	if (nfs4_write_stateid_changed(task, &hdr->args))
4223 		return -EAGAIN;
4224 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4225 		nfs4_write_done_cb(task, hdr);
4226 }
4227 
4228 static
4229 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4230 {
4231 	/* Don't request attributes for pNFS or O_DIRECT writes */
4232 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4233 		return false;
4234 	/* Otherwise, request attributes if and only if we don't hold
4235 	 * a delegation
4236 	 */
4237 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4238 }
4239 
4240 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4241 				  struct rpc_message *msg)
4242 {
4243 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4244 
4245 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4246 		hdr->args.bitmask = NULL;
4247 		hdr->res.fattr = NULL;
4248 	} else
4249 		hdr->args.bitmask = server->cache_consistency_bitmask;
4250 
4251 	if (!hdr->pgio_done_cb)
4252 		hdr->pgio_done_cb = nfs4_write_done_cb;
4253 	hdr->res.server = server;
4254 	hdr->timestamp   = jiffies;
4255 
4256 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4257 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4258 }
4259 
4260 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4261 {
4262 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4263 			&data->args.seq_args,
4264 			&data->res.seq_res,
4265 			task);
4266 }
4267 
4268 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4269 {
4270 	struct inode *inode = data->inode;
4271 
4272 	trace_nfs4_commit(data, task->tk_status);
4273 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4274 				    NULL, NULL) == -EAGAIN) {
4275 		rpc_restart_call_prepare(task);
4276 		return -EAGAIN;
4277 	}
4278 	return 0;
4279 }
4280 
4281 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4282 {
4283 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4284 		return -EAGAIN;
4285 	return data->commit_done_cb(task, data);
4286 }
4287 
4288 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4289 {
4290 	struct nfs_server *server = NFS_SERVER(data->inode);
4291 
4292 	if (data->commit_done_cb == NULL)
4293 		data->commit_done_cb = nfs4_commit_done_cb;
4294 	data->res.server = server;
4295 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4296 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4297 }
4298 
4299 struct nfs4_renewdata {
4300 	struct nfs_client	*client;
4301 	unsigned long		timestamp;
4302 };
4303 
4304 /*
4305  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4306  * standalone procedure for queueing an asynchronous RENEW.
4307  */
4308 static void nfs4_renew_release(void *calldata)
4309 {
4310 	struct nfs4_renewdata *data = calldata;
4311 	struct nfs_client *clp = data->client;
4312 
4313 	if (atomic_read(&clp->cl_count) > 1)
4314 		nfs4_schedule_state_renewal(clp);
4315 	nfs_put_client(clp);
4316 	kfree(data);
4317 }
4318 
4319 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4320 {
4321 	struct nfs4_renewdata *data = calldata;
4322 	struct nfs_client *clp = data->client;
4323 	unsigned long timestamp = data->timestamp;
4324 
4325 	trace_nfs4_renew_async(clp, task->tk_status);
4326 	switch (task->tk_status) {
4327 	case 0:
4328 		break;
4329 	case -NFS4ERR_LEASE_MOVED:
4330 		nfs4_schedule_lease_moved_recovery(clp);
4331 		break;
4332 	default:
4333 		/* Unless we're shutting down, schedule state recovery! */
4334 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4335 			return;
4336 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4337 			nfs4_schedule_lease_recovery(clp);
4338 			return;
4339 		}
4340 		nfs4_schedule_path_down_recovery(clp);
4341 	}
4342 	do_renew_lease(clp, timestamp);
4343 }
4344 
4345 static const struct rpc_call_ops nfs4_renew_ops = {
4346 	.rpc_call_done = nfs4_renew_done,
4347 	.rpc_release = nfs4_renew_release,
4348 };
4349 
4350 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4351 {
4352 	struct rpc_message msg = {
4353 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4354 		.rpc_argp	= clp,
4355 		.rpc_cred	= cred,
4356 	};
4357 	struct nfs4_renewdata *data;
4358 
4359 	if (renew_flags == 0)
4360 		return 0;
4361 	if (!atomic_inc_not_zero(&clp->cl_count))
4362 		return -EIO;
4363 	data = kmalloc(sizeof(*data), GFP_NOFS);
4364 	if (data == NULL)
4365 		return -ENOMEM;
4366 	data->client = clp;
4367 	data->timestamp = jiffies;
4368 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4369 			&nfs4_renew_ops, data);
4370 }
4371 
4372 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4373 {
4374 	struct rpc_message msg = {
4375 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4376 		.rpc_argp	= clp,
4377 		.rpc_cred	= cred,
4378 	};
4379 	unsigned long now = jiffies;
4380 	int status;
4381 
4382 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4383 	if (status < 0)
4384 		return status;
4385 	do_renew_lease(clp, now);
4386 	return 0;
4387 }
4388 
4389 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4390 {
4391 	return server->caps & NFS_CAP_ACLS;
4392 }
4393 
4394 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4395  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4396  * the stack.
4397  */
4398 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4399 
4400 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4401 		struct page **pages, unsigned int *pgbase)
4402 {
4403 	struct page *newpage, **spages;
4404 	int rc = 0;
4405 	size_t len;
4406 	spages = pages;
4407 
4408 	do {
4409 		len = min_t(size_t, PAGE_SIZE, buflen);
4410 		newpage = alloc_page(GFP_KERNEL);
4411 
4412 		if (newpage == NULL)
4413 			goto unwind;
4414 		memcpy(page_address(newpage), buf, len);
4415                 buf += len;
4416                 buflen -= len;
4417 		*pages++ = newpage;
4418 		rc++;
4419 	} while (buflen != 0);
4420 
4421 	return rc;
4422 
4423 unwind:
4424 	for(; rc > 0; rc--)
4425 		__free_page(spages[rc-1]);
4426 	return -ENOMEM;
4427 }
4428 
4429 struct nfs4_cached_acl {
4430 	int cached;
4431 	size_t len;
4432 	char data[0];
4433 };
4434 
4435 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4436 {
4437 	struct nfs_inode *nfsi = NFS_I(inode);
4438 
4439 	spin_lock(&inode->i_lock);
4440 	kfree(nfsi->nfs4_acl);
4441 	nfsi->nfs4_acl = acl;
4442 	spin_unlock(&inode->i_lock);
4443 }
4444 
4445 static void nfs4_zap_acl_attr(struct inode *inode)
4446 {
4447 	nfs4_set_cached_acl(inode, NULL);
4448 }
4449 
4450 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4451 {
4452 	struct nfs_inode *nfsi = NFS_I(inode);
4453 	struct nfs4_cached_acl *acl;
4454 	int ret = -ENOENT;
4455 
4456 	spin_lock(&inode->i_lock);
4457 	acl = nfsi->nfs4_acl;
4458 	if (acl == NULL)
4459 		goto out;
4460 	if (buf == NULL) /* user is just asking for length */
4461 		goto out_len;
4462 	if (acl->cached == 0)
4463 		goto out;
4464 	ret = -ERANGE; /* see getxattr(2) man page */
4465 	if (acl->len > buflen)
4466 		goto out;
4467 	memcpy(buf, acl->data, acl->len);
4468 out_len:
4469 	ret = acl->len;
4470 out:
4471 	spin_unlock(&inode->i_lock);
4472 	return ret;
4473 }
4474 
4475 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4476 {
4477 	struct nfs4_cached_acl *acl;
4478 	size_t buflen = sizeof(*acl) + acl_len;
4479 
4480 	if (buflen <= PAGE_SIZE) {
4481 		acl = kmalloc(buflen, GFP_KERNEL);
4482 		if (acl == NULL)
4483 			goto out;
4484 		acl->cached = 1;
4485 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4486 	} else {
4487 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4488 		if (acl == NULL)
4489 			goto out;
4490 		acl->cached = 0;
4491 	}
4492 	acl->len = acl_len;
4493 out:
4494 	nfs4_set_cached_acl(inode, acl);
4495 }
4496 
4497 /*
4498  * The getxattr API returns the required buffer length when called with a
4499  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4500  * the required buf.  On a NULL buf, we send a page of data to the server
4501  * guessing that the ACL request can be serviced by a page. If so, we cache
4502  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4503  * the cache. If not so, we throw away the page, and cache the required
4504  * length. The next getxattr call will then produce another round trip to
4505  * the server, this time with the input buf of the required size.
4506  */
4507 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4508 {
4509 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4510 	struct nfs_getaclargs args = {
4511 		.fh = NFS_FH(inode),
4512 		.acl_pages = pages,
4513 		.acl_len = buflen,
4514 	};
4515 	struct nfs_getaclres res = {
4516 		.acl_len = buflen,
4517 	};
4518 	struct rpc_message msg = {
4519 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4520 		.rpc_argp = &args,
4521 		.rpc_resp = &res,
4522 	};
4523 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4524 	int ret = -ENOMEM, i;
4525 
4526 	/* As long as we're doing a round trip to the server anyway,
4527 	 * let's be prepared for a page of acl data. */
4528 	if (npages == 0)
4529 		npages = 1;
4530 	if (npages > ARRAY_SIZE(pages))
4531 		return -ERANGE;
4532 
4533 	for (i = 0; i < npages; i++) {
4534 		pages[i] = alloc_page(GFP_KERNEL);
4535 		if (!pages[i])
4536 			goto out_free;
4537 	}
4538 
4539 	/* for decoding across pages */
4540 	res.acl_scratch = alloc_page(GFP_KERNEL);
4541 	if (!res.acl_scratch)
4542 		goto out_free;
4543 
4544 	args.acl_len = npages * PAGE_SIZE;
4545 	args.acl_pgbase = 0;
4546 
4547 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4548 		__func__, buf, buflen, npages, args.acl_len);
4549 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4550 			     &msg, &args.seq_args, &res.seq_res, 0);
4551 	if (ret)
4552 		goto out_free;
4553 
4554 	/* Handle the case where the passed-in buffer is too short */
4555 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4556 		/* Did the user only issue a request for the acl length? */
4557 		if (buf == NULL)
4558 			goto out_ok;
4559 		ret = -ERANGE;
4560 		goto out_free;
4561 	}
4562 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4563 	if (buf) {
4564 		if (res.acl_len > buflen) {
4565 			ret = -ERANGE;
4566 			goto out_free;
4567 		}
4568 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4569 	}
4570 out_ok:
4571 	ret = res.acl_len;
4572 out_free:
4573 	for (i = 0; i < npages; i++)
4574 		if (pages[i])
4575 			__free_page(pages[i]);
4576 	if (res.acl_scratch)
4577 		__free_page(res.acl_scratch);
4578 	return ret;
4579 }
4580 
4581 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4582 {
4583 	struct nfs4_exception exception = { };
4584 	ssize_t ret;
4585 	do {
4586 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4587 		trace_nfs4_get_acl(inode, ret);
4588 		if (ret >= 0)
4589 			break;
4590 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4591 	} while (exception.retry);
4592 	return ret;
4593 }
4594 
4595 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4596 {
4597 	struct nfs_server *server = NFS_SERVER(inode);
4598 	int ret;
4599 
4600 	if (!nfs4_server_supports_acls(server))
4601 		return -EOPNOTSUPP;
4602 	ret = nfs_revalidate_inode(server, inode);
4603 	if (ret < 0)
4604 		return ret;
4605 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4606 		nfs_zap_acl_cache(inode);
4607 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4608 	if (ret != -ENOENT)
4609 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4610 		 * but no cached acl data, just the acl length */
4611 		return ret;
4612 	return nfs4_get_acl_uncached(inode, buf, buflen);
4613 }
4614 
4615 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4616 {
4617 	struct nfs_server *server = NFS_SERVER(inode);
4618 	struct page *pages[NFS4ACL_MAXPAGES];
4619 	struct nfs_setaclargs arg = {
4620 		.fh		= NFS_FH(inode),
4621 		.acl_pages	= pages,
4622 		.acl_len	= buflen,
4623 	};
4624 	struct nfs_setaclres res;
4625 	struct rpc_message msg = {
4626 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4627 		.rpc_argp	= &arg,
4628 		.rpc_resp	= &res,
4629 	};
4630 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4631 	int ret, i;
4632 
4633 	if (!nfs4_server_supports_acls(server))
4634 		return -EOPNOTSUPP;
4635 	if (npages > ARRAY_SIZE(pages))
4636 		return -ERANGE;
4637 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4638 	if (i < 0)
4639 		return i;
4640 	nfs4_inode_return_delegation(inode);
4641 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4642 
4643 	/*
4644 	 * Free each page after tx, so the only ref left is
4645 	 * held by the network stack
4646 	 */
4647 	for (; i > 0; i--)
4648 		put_page(pages[i-1]);
4649 
4650 	/*
4651 	 * Acl update can result in inode attribute update.
4652 	 * so mark the attribute cache invalid.
4653 	 */
4654 	spin_lock(&inode->i_lock);
4655 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4656 	spin_unlock(&inode->i_lock);
4657 	nfs_access_zap_cache(inode);
4658 	nfs_zap_acl_cache(inode);
4659 	return ret;
4660 }
4661 
4662 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4663 {
4664 	struct nfs4_exception exception = { };
4665 	int err;
4666 	do {
4667 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4668 		trace_nfs4_set_acl(inode, err);
4669 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4670 				&exception);
4671 	} while (exception.retry);
4672 	return err;
4673 }
4674 
4675 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4676 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4677 					size_t buflen)
4678 {
4679 	struct nfs_server *server = NFS_SERVER(inode);
4680 	struct nfs_fattr fattr;
4681 	struct nfs4_label label = {0, 0, buflen, buf};
4682 
4683 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4684 	struct nfs4_getattr_arg arg = {
4685 		.fh		= NFS_FH(inode),
4686 		.bitmask	= bitmask,
4687 	};
4688 	struct nfs4_getattr_res res = {
4689 		.fattr		= &fattr,
4690 		.label		= &label,
4691 		.server		= server,
4692 	};
4693 	struct rpc_message msg = {
4694 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4695 		.rpc_argp	= &arg,
4696 		.rpc_resp	= &res,
4697 	};
4698 	int ret;
4699 
4700 	nfs_fattr_init(&fattr);
4701 
4702 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4703 	if (ret)
4704 		return ret;
4705 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4706 		return -ENOENT;
4707 	if (buflen < label.len)
4708 		return -ERANGE;
4709 	return 0;
4710 }
4711 
4712 static int nfs4_get_security_label(struct inode *inode, void *buf,
4713 					size_t buflen)
4714 {
4715 	struct nfs4_exception exception = { };
4716 	int err;
4717 
4718 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4719 		return -EOPNOTSUPP;
4720 
4721 	do {
4722 		err = _nfs4_get_security_label(inode, buf, buflen);
4723 		trace_nfs4_get_security_label(inode, err);
4724 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4725 				&exception);
4726 	} while (exception.retry);
4727 	return err;
4728 }
4729 
4730 static int _nfs4_do_set_security_label(struct inode *inode,
4731 		struct nfs4_label *ilabel,
4732 		struct nfs_fattr *fattr,
4733 		struct nfs4_label *olabel)
4734 {
4735 
4736 	struct iattr sattr = {0};
4737 	struct nfs_server *server = NFS_SERVER(inode);
4738 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4739 	struct nfs_setattrargs arg = {
4740 		.fh             = NFS_FH(inode),
4741 		.iap            = &sattr,
4742 		.server		= server,
4743 		.bitmask	= bitmask,
4744 		.label		= ilabel,
4745 	};
4746 	struct nfs_setattrres res = {
4747 		.fattr		= fattr,
4748 		.label		= olabel,
4749 		.server		= server,
4750 	};
4751 	struct rpc_message msg = {
4752 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4753 		.rpc_argp       = &arg,
4754 		.rpc_resp       = &res,
4755 	};
4756 	int status;
4757 
4758 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4759 
4760 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4761 	if (status)
4762 		dprintk("%s failed: %d\n", __func__, status);
4763 
4764 	return status;
4765 }
4766 
4767 static int nfs4_do_set_security_label(struct inode *inode,
4768 		struct nfs4_label *ilabel,
4769 		struct nfs_fattr *fattr,
4770 		struct nfs4_label *olabel)
4771 {
4772 	struct nfs4_exception exception = { };
4773 	int err;
4774 
4775 	do {
4776 		err = _nfs4_do_set_security_label(inode, ilabel,
4777 				fattr, olabel);
4778 		trace_nfs4_set_security_label(inode, err);
4779 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4780 				&exception);
4781 	} while (exception.retry);
4782 	return err;
4783 }
4784 
4785 static int
4786 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4787 {
4788 	struct nfs4_label ilabel, *olabel = NULL;
4789 	struct nfs_fattr fattr;
4790 	struct rpc_cred *cred;
4791 	struct inode *inode = dentry->d_inode;
4792 	int status;
4793 
4794 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4795 		return -EOPNOTSUPP;
4796 
4797 	nfs_fattr_init(&fattr);
4798 
4799 	ilabel.pi = 0;
4800 	ilabel.lfs = 0;
4801 	ilabel.label = (char *)buf;
4802 	ilabel.len = buflen;
4803 
4804 	cred = rpc_lookup_cred();
4805 	if (IS_ERR(cred))
4806 		return PTR_ERR(cred);
4807 
4808 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4809 	if (IS_ERR(olabel)) {
4810 		status = -PTR_ERR(olabel);
4811 		goto out;
4812 	}
4813 
4814 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4815 	if (status == 0)
4816 		nfs_setsecurity(inode, &fattr, olabel);
4817 
4818 	nfs4_label_free(olabel);
4819 out:
4820 	put_rpccred(cred);
4821 	return status;
4822 }
4823 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4824 
4825 
4826 static int
4827 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4828 			struct nfs4_state *state, long *timeout)
4829 {
4830 	struct nfs_client *clp = server->nfs_client;
4831 
4832 	if (task->tk_status >= 0)
4833 		return 0;
4834 	switch(task->tk_status) {
4835 		case -NFS4ERR_DELEG_REVOKED:
4836 		case -NFS4ERR_ADMIN_REVOKED:
4837 		case -NFS4ERR_BAD_STATEID:
4838 		case -NFS4ERR_OPENMODE:
4839 			if (state == NULL)
4840 				break;
4841 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4842 				goto recovery_failed;
4843 			goto wait_on_recovery;
4844 		case -NFS4ERR_EXPIRED:
4845 			if (state != NULL) {
4846 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4847 					goto recovery_failed;
4848 			}
4849 		case -NFS4ERR_STALE_STATEID:
4850 		case -NFS4ERR_STALE_CLIENTID:
4851 			nfs4_schedule_lease_recovery(clp);
4852 			goto wait_on_recovery;
4853 		case -NFS4ERR_MOVED:
4854 			if (nfs4_schedule_migration_recovery(server) < 0)
4855 				goto recovery_failed;
4856 			goto wait_on_recovery;
4857 		case -NFS4ERR_LEASE_MOVED:
4858 			nfs4_schedule_lease_moved_recovery(clp);
4859 			goto wait_on_recovery;
4860 #if defined(CONFIG_NFS_V4_1)
4861 		case -NFS4ERR_BADSESSION:
4862 		case -NFS4ERR_BADSLOT:
4863 		case -NFS4ERR_BAD_HIGH_SLOT:
4864 		case -NFS4ERR_DEADSESSION:
4865 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4866 		case -NFS4ERR_SEQ_FALSE_RETRY:
4867 		case -NFS4ERR_SEQ_MISORDERED:
4868 			dprintk("%s ERROR %d, Reset session\n", __func__,
4869 				task->tk_status);
4870 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4871 			goto wait_on_recovery;
4872 #endif /* CONFIG_NFS_V4_1 */
4873 		case -NFS4ERR_DELAY:
4874 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4875 			rpc_delay(task, nfs4_update_delay(timeout));
4876 			goto restart_call;
4877 		case -NFS4ERR_GRACE:
4878 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4879 		case -NFS4ERR_RETRY_UNCACHED_REP:
4880 		case -NFS4ERR_OLD_STATEID:
4881 			goto restart_call;
4882 	}
4883 	task->tk_status = nfs4_map_errors(task->tk_status);
4884 	return 0;
4885 recovery_failed:
4886 	task->tk_status = -EIO;
4887 	return 0;
4888 wait_on_recovery:
4889 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4890 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4891 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4892 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4893 		goto recovery_failed;
4894 restart_call:
4895 	task->tk_status = 0;
4896 	return -EAGAIN;
4897 }
4898 
4899 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4900 				    nfs4_verifier *bootverf)
4901 {
4902 	__be32 verf[2];
4903 
4904 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4905 		/* An impossible timestamp guarantees this value
4906 		 * will never match a generated boot time. */
4907 		verf[0] = 0;
4908 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4909 	} else {
4910 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4911 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4912 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4913 	}
4914 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4915 }
4916 
4917 static unsigned int
4918 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4919 				   char *buf, size_t len)
4920 {
4921 	unsigned int result;
4922 
4923 	if (clp->cl_owner_id != NULL)
4924 		return strlcpy(buf, clp->cl_owner_id, len);
4925 
4926 	rcu_read_lock();
4927 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4928 				clp->cl_ipaddr,
4929 				rpc_peeraddr2str(clp->cl_rpcclient,
4930 							RPC_DISPLAY_ADDR),
4931 				rpc_peeraddr2str(clp->cl_rpcclient,
4932 							RPC_DISPLAY_PROTO));
4933 	rcu_read_unlock();
4934 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4935 	return result;
4936 }
4937 
4938 static unsigned int
4939 nfs4_init_uniform_client_string(struct nfs_client *clp,
4940 				char *buf, size_t len)
4941 {
4942 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4943 	unsigned int result;
4944 
4945 	if (clp->cl_owner_id != NULL)
4946 		return strlcpy(buf, clp->cl_owner_id, len);
4947 
4948 	if (nfs4_client_id_uniquifier[0] != '\0')
4949 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4950 				clp->rpc_ops->version,
4951 				clp->cl_minorversion,
4952 				nfs4_client_id_uniquifier,
4953 				nodename);
4954 	else
4955 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4956 				clp->rpc_ops->version, clp->cl_minorversion,
4957 				nodename);
4958 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4959 	return result;
4960 }
4961 
4962 /*
4963  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
4964  * services.  Advertise one based on the address family of the
4965  * clientaddr.
4966  */
4967 static unsigned int
4968 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
4969 {
4970 	if (strchr(clp->cl_ipaddr, ':') != NULL)
4971 		return scnprintf(buf, len, "tcp6");
4972 	else
4973 		return scnprintf(buf, len, "tcp");
4974 }
4975 
4976 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
4977 {
4978 	struct nfs4_setclientid *sc = calldata;
4979 
4980 	if (task->tk_status == 0)
4981 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
4982 }
4983 
4984 static const struct rpc_call_ops nfs4_setclientid_ops = {
4985 	.rpc_call_done = nfs4_setclientid_done,
4986 };
4987 
4988 /**
4989  * nfs4_proc_setclientid - Negotiate client ID
4990  * @clp: state data structure
4991  * @program: RPC program for NFSv4 callback service
4992  * @port: IP port number for NFS4 callback service
4993  * @cred: RPC credential to use for this call
4994  * @res: where to place the result
4995  *
4996  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4997  */
4998 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4999 		unsigned short port, struct rpc_cred *cred,
5000 		struct nfs4_setclientid_res *res)
5001 {
5002 	nfs4_verifier sc_verifier;
5003 	struct nfs4_setclientid setclientid = {
5004 		.sc_verifier = &sc_verifier,
5005 		.sc_prog = program,
5006 		.sc_cb_ident = clp->cl_cb_ident,
5007 	};
5008 	struct rpc_message msg = {
5009 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5010 		.rpc_argp = &setclientid,
5011 		.rpc_resp = res,
5012 		.rpc_cred = cred,
5013 	};
5014 	struct rpc_task *task;
5015 	struct rpc_task_setup task_setup_data = {
5016 		.rpc_client = clp->cl_rpcclient,
5017 		.rpc_message = &msg,
5018 		.callback_ops = &nfs4_setclientid_ops,
5019 		.callback_data = &setclientid,
5020 		.flags = RPC_TASK_TIMEOUT,
5021 	};
5022 	int status;
5023 
5024 	/* nfs_client_id4 */
5025 	nfs4_init_boot_verifier(clp, &sc_verifier);
5026 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5027 		setclientid.sc_name_len =
5028 				nfs4_init_uniform_client_string(clp,
5029 						setclientid.sc_name,
5030 						sizeof(setclientid.sc_name));
5031 	else
5032 		setclientid.sc_name_len =
5033 				nfs4_init_nonuniform_client_string(clp,
5034 						setclientid.sc_name,
5035 						sizeof(setclientid.sc_name));
5036 	/* cb_client4 */
5037 	setclientid.sc_netid_len =
5038 				nfs4_init_callback_netid(clp,
5039 						setclientid.sc_netid,
5040 						sizeof(setclientid.sc_netid));
5041 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5042 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5043 				clp->cl_ipaddr, port >> 8, port & 255);
5044 
5045 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5046 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5047 		setclientid.sc_name_len, setclientid.sc_name);
5048 	task = rpc_run_task(&task_setup_data);
5049 	if (IS_ERR(task)) {
5050 		status = PTR_ERR(task);
5051 		goto out;
5052 	}
5053 	status = task->tk_status;
5054 	if (setclientid.sc_cred) {
5055 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5056 		put_rpccred(setclientid.sc_cred);
5057 	}
5058 	rpc_put_task(task);
5059 out:
5060 	trace_nfs4_setclientid(clp, status);
5061 	dprintk("NFS reply setclientid: %d\n", status);
5062 	return status;
5063 }
5064 
5065 /**
5066  * nfs4_proc_setclientid_confirm - Confirm client ID
5067  * @clp: state data structure
5068  * @res: result of a previous SETCLIENTID
5069  * @cred: RPC credential to use for this call
5070  *
5071  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5072  */
5073 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5074 		struct nfs4_setclientid_res *arg,
5075 		struct rpc_cred *cred)
5076 {
5077 	struct rpc_message msg = {
5078 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5079 		.rpc_argp = arg,
5080 		.rpc_cred = cred,
5081 	};
5082 	int status;
5083 
5084 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5085 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5086 		clp->cl_clientid);
5087 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5088 	trace_nfs4_setclientid_confirm(clp, status);
5089 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5090 	return status;
5091 }
5092 
5093 struct nfs4_delegreturndata {
5094 	struct nfs4_delegreturnargs args;
5095 	struct nfs4_delegreturnres res;
5096 	struct nfs_fh fh;
5097 	nfs4_stateid stateid;
5098 	unsigned long timestamp;
5099 	struct nfs_fattr fattr;
5100 	int rpc_status;
5101 	struct inode *inode;
5102 	bool roc;
5103 	u32 roc_barrier;
5104 };
5105 
5106 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5107 {
5108 	struct nfs4_delegreturndata *data = calldata;
5109 
5110 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5111 		return;
5112 
5113 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5114 	switch (task->tk_status) {
5115 	case 0:
5116 		renew_lease(data->res.server, data->timestamp);
5117 	case -NFS4ERR_ADMIN_REVOKED:
5118 	case -NFS4ERR_DELEG_REVOKED:
5119 	case -NFS4ERR_BAD_STATEID:
5120 	case -NFS4ERR_OLD_STATEID:
5121 	case -NFS4ERR_STALE_STATEID:
5122 	case -NFS4ERR_EXPIRED:
5123 		task->tk_status = 0;
5124 		if (data->roc)
5125 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5126 		break;
5127 	default:
5128 		if (nfs4_async_handle_error(task, data->res.server,
5129 					    NULL, NULL) == -EAGAIN) {
5130 			rpc_restart_call_prepare(task);
5131 			return;
5132 		}
5133 	}
5134 	data->rpc_status = task->tk_status;
5135 }
5136 
5137 static void nfs4_delegreturn_release(void *calldata)
5138 {
5139 	struct nfs4_delegreturndata *data = calldata;
5140 
5141 	if (data->roc)
5142 		pnfs_roc_release(data->inode);
5143 	kfree(calldata);
5144 }
5145 
5146 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5147 {
5148 	struct nfs4_delegreturndata *d_data;
5149 
5150 	d_data = (struct nfs4_delegreturndata *)data;
5151 
5152 	if (d_data->roc &&
5153 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5154 		return;
5155 
5156 	nfs4_setup_sequence(d_data->res.server,
5157 			&d_data->args.seq_args,
5158 			&d_data->res.seq_res,
5159 			task);
5160 }
5161 
5162 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5163 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5164 	.rpc_call_done = nfs4_delegreturn_done,
5165 	.rpc_release = nfs4_delegreturn_release,
5166 };
5167 
5168 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5169 {
5170 	struct nfs4_delegreturndata *data;
5171 	struct nfs_server *server = NFS_SERVER(inode);
5172 	struct rpc_task *task;
5173 	struct rpc_message msg = {
5174 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5175 		.rpc_cred = cred,
5176 	};
5177 	struct rpc_task_setup task_setup_data = {
5178 		.rpc_client = server->client,
5179 		.rpc_message = &msg,
5180 		.callback_ops = &nfs4_delegreturn_ops,
5181 		.flags = RPC_TASK_ASYNC,
5182 	};
5183 	int status = 0;
5184 
5185 	data = kzalloc(sizeof(*data), GFP_NOFS);
5186 	if (data == NULL)
5187 		return -ENOMEM;
5188 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5189 	data->args.fhandle = &data->fh;
5190 	data->args.stateid = &data->stateid;
5191 	data->args.bitmask = server->cache_consistency_bitmask;
5192 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5193 	nfs4_stateid_copy(&data->stateid, stateid);
5194 	data->res.fattr = &data->fattr;
5195 	data->res.server = server;
5196 	nfs_fattr_init(data->res.fattr);
5197 	data->timestamp = jiffies;
5198 	data->rpc_status = 0;
5199 	data->inode = inode;
5200 	data->roc = list_empty(&NFS_I(inode)->open_files) ?
5201 		    pnfs_roc(inode) : false;
5202 
5203 	task_setup_data.callback_data = data;
5204 	msg.rpc_argp = &data->args;
5205 	msg.rpc_resp = &data->res;
5206 	task = rpc_run_task(&task_setup_data);
5207 	if (IS_ERR(task))
5208 		return PTR_ERR(task);
5209 	if (!issync)
5210 		goto out;
5211 	status = nfs4_wait_for_completion_rpc_task(task);
5212 	if (status != 0)
5213 		goto out;
5214 	status = data->rpc_status;
5215 	if (status == 0)
5216 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5217 	else
5218 		nfs_refresh_inode(inode, &data->fattr);
5219 out:
5220 	rpc_put_task(task);
5221 	return status;
5222 }
5223 
5224 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5225 {
5226 	struct nfs_server *server = NFS_SERVER(inode);
5227 	struct nfs4_exception exception = { };
5228 	int err;
5229 	do {
5230 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5231 		trace_nfs4_delegreturn(inode, err);
5232 		switch (err) {
5233 			case -NFS4ERR_STALE_STATEID:
5234 			case -NFS4ERR_EXPIRED:
5235 			case 0:
5236 				return 0;
5237 		}
5238 		err = nfs4_handle_exception(server, err, &exception);
5239 	} while (exception.retry);
5240 	return err;
5241 }
5242 
5243 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5244 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5245 
5246 /*
5247  * sleep, with exponential backoff, and retry the LOCK operation.
5248  */
5249 static unsigned long
5250 nfs4_set_lock_task_retry(unsigned long timeout)
5251 {
5252 	freezable_schedule_timeout_killable_unsafe(timeout);
5253 	timeout <<= 1;
5254 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5255 		return NFS4_LOCK_MAXTIMEOUT;
5256 	return timeout;
5257 }
5258 
5259 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5260 {
5261 	struct inode *inode = state->inode;
5262 	struct nfs_server *server = NFS_SERVER(inode);
5263 	struct nfs_client *clp = server->nfs_client;
5264 	struct nfs_lockt_args arg = {
5265 		.fh = NFS_FH(inode),
5266 		.fl = request,
5267 	};
5268 	struct nfs_lockt_res res = {
5269 		.denied = request,
5270 	};
5271 	struct rpc_message msg = {
5272 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5273 		.rpc_argp       = &arg,
5274 		.rpc_resp       = &res,
5275 		.rpc_cred	= state->owner->so_cred,
5276 	};
5277 	struct nfs4_lock_state *lsp;
5278 	int status;
5279 
5280 	arg.lock_owner.clientid = clp->cl_clientid;
5281 	status = nfs4_set_lock_state(state, request);
5282 	if (status != 0)
5283 		goto out;
5284 	lsp = request->fl_u.nfs4_fl.owner;
5285 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5286 	arg.lock_owner.s_dev = server->s_dev;
5287 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5288 	switch (status) {
5289 		case 0:
5290 			request->fl_type = F_UNLCK;
5291 			break;
5292 		case -NFS4ERR_DENIED:
5293 			status = 0;
5294 	}
5295 	request->fl_ops->fl_release_private(request);
5296 	request->fl_ops = NULL;
5297 out:
5298 	return status;
5299 }
5300 
5301 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5302 {
5303 	struct nfs4_exception exception = { };
5304 	int err;
5305 
5306 	do {
5307 		err = _nfs4_proc_getlk(state, cmd, request);
5308 		trace_nfs4_get_lock(request, state, cmd, err);
5309 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5310 				&exception);
5311 	} while (exception.retry);
5312 	return err;
5313 }
5314 
5315 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5316 {
5317 	int res = 0;
5318 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5319 		case FL_POSIX:
5320 			res = posix_lock_file_wait(file, fl);
5321 			break;
5322 		case FL_FLOCK:
5323 			res = flock_lock_file_wait(file, fl);
5324 			break;
5325 		default:
5326 			BUG();
5327 	}
5328 	return res;
5329 }
5330 
5331 struct nfs4_unlockdata {
5332 	struct nfs_locku_args arg;
5333 	struct nfs_locku_res res;
5334 	struct nfs4_lock_state *lsp;
5335 	struct nfs_open_context *ctx;
5336 	struct file_lock fl;
5337 	const struct nfs_server *server;
5338 	unsigned long timestamp;
5339 };
5340 
5341 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5342 		struct nfs_open_context *ctx,
5343 		struct nfs4_lock_state *lsp,
5344 		struct nfs_seqid *seqid)
5345 {
5346 	struct nfs4_unlockdata *p;
5347 	struct inode *inode = lsp->ls_state->inode;
5348 
5349 	p = kzalloc(sizeof(*p), GFP_NOFS);
5350 	if (p == NULL)
5351 		return NULL;
5352 	p->arg.fh = NFS_FH(inode);
5353 	p->arg.fl = &p->fl;
5354 	p->arg.seqid = seqid;
5355 	p->res.seqid = seqid;
5356 	p->arg.stateid = &lsp->ls_stateid;
5357 	p->lsp = lsp;
5358 	atomic_inc(&lsp->ls_count);
5359 	/* Ensure we don't close file until we're done freeing locks! */
5360 	p->ctx = get_nfs_open_context(ctx);
5361 	memcpy(&p->fl, fl, sizeof(p->fl));
5362 	p->server = NFS_SERVER(inode);
5363 	return p;
5364 }
5365 
5366 static void nfs4_locku_release_calldata(void *data)
5367 {
5368 	struct nfs4_unlockdata *calldata = data;
5369 	nfs_free_seqid(calldata->arg.seqid);
5370 	nfs4_put_lock_state(calldata->lsp);
5371 	put_nfs_open_context(calldata->ctx);
5372 	kfree(calldata);
5373 }
5374 
5375 static void nfs4_locku_done(struct rpc_task *task, void *data)
5376 {
5377 	struct nfs4_unlockdata *calldata = data;
5378 
5379 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5380 		return;
5381 	switch (task->tk_status) {
5382 		case 0:
5383 			nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5384 					&calldata->res.stateid);
5385 			renew_lease(calldata->server, calldata->timestamp);
5386 			break;
5387 		case -NFS4ERR_BAD_STATEID:
5388 		case -NFS4ERR_OLD_STATEID:
5389 		case -NFS4ERR_STALE_STATEID:
5390 		case -NFS4ERR_EXPIRED:
5391 			break;
5392 		default:
5393 			if (nfs4_async_handle_error(task, calldata->server,
5394 						    NULL, NULL) == -EAGAIN)
5395 				rpc_restart_call_prepare(task);
5396 	}
5397 	nfs_release_seqid(calldata->arg.seqid);
5398 }
5399 
5400 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5401 {
5402 	struct nfs4_unlockdata *calldata = data;
5403 
5404 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5405 		goto out_wait;
5406 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5407 		/* Note: exit _without_ running nfs4_locku_done */
5408 		goto out_no_action;
5409 	}
5410 	calldata->timestamp = jiffies;
5411 	if (nfs4_setup_sequence(calldata->server,
5412 				&calldata->arg.seq_args,
5413 				&calldata->res.seq_res,
5414 				task) != 0)
5415 		nfs_release_seqid(calldata->arg.seqid);
5416 	return;
5417 out_no_action:
5418 	task->tk_action = NULL;
5419 out_wait:
5420 	nfs4_sequence_done(task, &calldata->res.seq_res);
5421 }
5422 
5423 static const struct rpc_call_ops nfs4_locku_ops = {
5424 	.rpc_call_prepare = nfs4_locku_prepare,
5425 	.rpc_call_done = nfs4_locku_done,
5426 	.rpc_release = nfs4_locku_release_calldata,
5427 };
5428 
5429 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5430 		struct nfs_open_context *ctx,
5431 		struct nfs4_lock_state *lsp,
5432 		struct nfs_seqid *seqid)
5433 {
5434 	struct nfs4_unlockdata *data;
5435 	struct rpc_message msg = {
5436 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5437 		.rpc_cred = ctx->cred,
5438 	};
5439 	struct rpc_task_setup task_setup_data = {
5440 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5441 		.rpc_message = &msg,
5442 		.callback_ops = &nfs4_locku_ops,
5443 		.workqueue = nfsiod_workqueue,
5444 		.flags = RPC_TASK_ASYNC,
5445 	};
5446 
5447 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5448 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5449 
5450 	/* Ensure this is an unlock - when canceling a lock, the
5451 	 * canceled lock is passed in, and it won't be an unlock.
5452 	 */
5453 	fl->fl_type = F_UNLCK;
5454 
5455 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5456 	if (data == NULL) {
5457 		nfs_free_seqid(seqid);
5458 		return ERR_PTR(-ENOMEM);
5459 	}
5460 
5461 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5462 	msg.rpc_argp = &data->arg;
5463 	msg.rpc_resp = &data->res;
5464 	task_setup_data.callback_data = data;
5465 	return rpc_run_task(&task_setup_data);
5466 }
5467 
5468 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5469 {
5470 	struct inode *inode = state->inode;
5471 	struct nfs4_state_owner *sp = state->owner;
5472 	struct nfs_inode *nfsi = NFS_I(inode);
5473 	struct nfs_seqid *seqid;
5474 	struct nfs4_lock_state *lsp;
5475 	struct rpc_task *task;
5476 	int status = 0;
5477 	unsigned char fl_flags = request->fl_flags;
5478 
5479 	status = nfs4_set_lock_state(state, request);
5480 	/* Unlock _before_ we do the RPC call */
5481 	request->fl_flags |= FL_EXISTS;
5482 	/* Exclude nfs_delegation_claim_locks() */
5483 	mutex_lock(&sp->so_delegreturn_mutex);
5484 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5485 	down_read(&nfsi->rwsem);
5486 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5487 		up_read(&nfsi->rwsem);
5488 		mutex_unlock(&sp->so_delegreturn_mutex);
5489 		goto out;
5490 	}
5491 	up_read(&nfsi->rwsem);
5492 	mutex_unlock(&sp->so_delegreturn_mutex);
5493 	if (status != 0)
5494 		goto out;
5495 	/* Is this a delegated lock? */
5496 	lsp = request->fl_u.nfs4_fl.owner;
5497 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5498 		goto out;
5499 	seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5500 	status = -ENOMEM;
5501 	if (seqid == NULL)
5502 		goto out;
5503 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5504 	status = PTR_ERR(task);
5505 	if (IS_ERR(task))
5506 		goto out;
5507 	status = nfs4_wait_for_completion_rpc_task(task);
5508 	rpc_put_task(task);
5509 out:
5510 	request->fl_flags = fl_flags;
5511 	trace_nfs4_unlock(request, state, F_SETLK, status);
5512 	return status;
5513 }
5514 
5515 struct nfs4_lockdata {
5516 	struct nfs_lock_args arg;
5517 	struct nfs_lock_res res;
5518 	struct nfs4_lock_state *lsp;
5519 	struct nfs_open_context *ctx;
5520 	struct file_lock fl;
5521 	unsigned long timestamp;
5522 	int rpc_status;
5523 	int cancelled;
5524 	struct nfs_server *server;
5525 };
5526 
5527 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5528 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5529 		gfp_t gfp_mask)
5530 {
5531 	struct nfs4_lockdata *p;
5532 	struct inode *inode = lsp->ls_state->inode;
5533 	struct nfs_server *server = NFS_SERVER(inode);
5534 
5535 	p = kzalloc(sizeof(*p), gfp_mask);
5536 	if (p == NULL)
5537 		return NULL;
5538 
5539 	p->arg.fh = NFS_FH(inode);
5540 	p->arg.fl = &p->fl;
5541 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5542 	if (p->arg.open_seqid == NULL)
5543 		goto out_free;
5544 	p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5545 	if (p->arg.lock_seqid == NULL)
5546 		goto out_free_seqid;
5547 	p->arg.lock_stateid = &lsp->ls_stateid;
5548 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5549 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5550 	p->arg.lock_owner.s_dev = server->s_dev;
5551 	p->res.lock_seqid = p->arg.lock_seqid;
5552 	p->lsp = lsp;
5553 	p->server = server;
5554 	atomic_inc(&lsp->ls_count);
5555 	p->ctx = get_nfs_open_context(ctx);
5556 	memcpy(&p->fl, fl, sizeof(p->fl));
5557 	return p;
5558 out_free_seqid:
5559 	nfs_free_seqid(p->arg.open_seqid);
5560 out_free:
5561 	kfree(p);
5562 	return NULL;
5563 }
5564 
5565 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5566 {
5567 	struct nfs4_lockdata *data = calldata;
5568 	struct nfs4_state *state = data->lsp->ls_state;
5569 
5570 	dprintk("%s: begin!\n", __func__);
5571 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5572 		goto out_wait;
5573 	/* Do we need to do an open_to_lock_owner? */
5574 	if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5575 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5576 			goto out_release_lock_seqid;
5577 		}
5578 		data->arg.open_stateid = &state->open_stateid;
5579 		data->arg.new_lock_owner = 1;
5580 		data->res.open_seqid = data->arg.open_seqid;
5581 	} else
5582 		data->arg.new_lock_owner = 0;
5583 	if (!nfs4_valid_open_stateid(state)) {
5584 		data->rpc_status = -EBADF;
5585 		task->tk_action = NULL;
5586 		goto out_release_open_seqid;
5587 	}
5588 	data->timestamp = jiffies;
5589 	if (nfs4_setup_sequence(data->server,
5590 				&data->arg.seq_args,
5591 				&data->res.seq_res,
5592 				task) == 0)
5593 		return;
5594 out_release_open_seqid:
5595 	nfs_release_seqid(data->arg.open_seqid);
5596 out_release_lock_seqid:
5597 	nfs_release_seqid(data->arg.lock_seqid);
5598 out_wait:
5599 	nfs4_sequence_done(task, &data->res.seq_res);
5600 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5601 }
5602 
5603 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5604 {
5605 	struct nfs4_lockdata *data = calldata;
5606 
5607 	dprintk("%s: begin!\n", __func__);
5608 
5609 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5610 		return;
5611 
5612 	data->rpc_status = task->tk_status;
5613 	if (data->arg.new_lock_owner != 0) {
5614 		if (data->rpc_status == 0)
5615 			nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5616 		else
5617 			goto out;
5618 	}
5619 	if (data->rpc_status == 0) {
5620 		nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5621 		set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5622 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5623 	}
5624 out:
5625 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5626 }
5627 
5628 static void nfs4_lock_release(void *calldata)
5629 {
5630 	struct nfs4_lockdata *data = calldata;
5631 
5632 	dprintk("%s: begin!\n", __func__);
5633 	nfs_free_seqid(data->arg.open_seqid);
5634 	if (data->cancelled != 0) {
5635 		struct rpc_task *task;
5636 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5637 				data->arg.lock_seqid);
5638 		if (!IS_ERR(task))
5639 			rpc_put_task_async(task);
5640 		dprintk("%s: cancelling lock!\n", __func__);
5641 	} else
5642 		nfs_free_seqid(data->arg.lock_seqid);
5643 	nfs4_put_lock_state(data->lsp);
5644 	put_nfs_open_context(data->ctx);
5645 	kfree(data);
5646 	dprintk("%s: done!\n", __func__);
5647 }
5648 
5649 static const struct rpc_call_ops nfs4_lock_ops = {
5650 	.rpc_call_prepare = nfs4_lock_prepare,
5651 	.rpc_call_done = nfs4_lock_done,
5652 	.rpc_release = nfs4_lock_release,
5653 };
5654 
5655 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5656 {
5657 	switch (error) {
5658 	case -NFS4ERR_ADMIN_REVOKED:
5659 	case -NFS4ERR_BAD_STATEID:
5660 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5661 		if (new_lock_owner != 0 ||
5662 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5663 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5664 		break;
5665 	case -NFS4ERR_STALE_STATEID:
5666 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5667 	case -NFS4ERR_EXPIRED:
5668 		nfs4_schedule_lease_recovery(server->nfs_client);
5669 	};
5670 }
5671 
5672 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5673 {
5674 	struct nfs4_lockdata *data;
5675 	struct rpc_task *task;
5676 	struct rpc_message msg = {
5677 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5678 		.rpc_cred = state->owner->so_cred,
5679 	};
5680 	struct rpc_task_setup task_setup_data = {
5681 		.rpc_client = NFS_CLIENT(state->inode),
5682 		.rpc_message = &msg,
5683 		.callback_ops = &nfs4_lock_ops,
5684 		.workqueue = nfsiod_workqueue,
5685 		.flags = RPC_TASK_ASYNC,
5686 	};
5687 	int ret;
5688 
5689 	dprintk("%s: begin!\n", __func__);
5690 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5691 			fl->fl_u.nfs4_fl.owner,
5692 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5693 	if (data == NULL)
5694 		return -ENOMEM;
5695 	if (IS_SETLKW(cmd))
5696 		data->arg.block = 1;
5697 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5698 	msg.rpc_argp = &data->arg;
5699 	msg.rpc_resp = &data->res;
5700 	task_setup_data.callback_data = data;
5701 	if (recovery_type > NFS_LOCK_NEW) {
5702 		if (recovery_type == NFS_LOCK_RECLAIM)
5703 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5704 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5705 	}
5706 	task = rpc_run_task(&task_setup_data);
5707 	if (IS_ERR(task))
5708 		return PTR_ERR(task);
5709 	ret = nfs4_wait_for_completion_rpc_task(task);
5710 	if (ret == 0) {
5711 		ret = data->rpc_status;
5712 		if (ret)
5713 			nfs4_handle_setlk_error(data->server, data->lsp,
5714 					data->arg.new_lock_owner, ret);
5715 	} else
5716 		data->cancelled = 1;
5717 	rpc_put_task(task);
5718 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5719 	return ret;
5720 }
5721 
5722 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5723 {
5724 	struct nfs_server *server = NFS_SERVER(state->inode);
5725 	struct nfs4_exception exception = {
5726 		.inode = state->inode,
5727 	};
5728 	int err;
5729 
5730 	do {
5731 		/* Cache the lock if possible... */
5732 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5733 			return 0;
5734 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5735 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5736 		if (err != -NFS4ERR_DELAY)
5737 			break;
5738 		nfs4_handle_exception(server, err, &exception);
5739 	} while (exception.retry);
5740 	return err;
5741 }
5742 
5743 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5744 {
5745 	struct nfs_server *server = NFS_SERVER(state->inode);
5746 	struct nfs4_exception exception = {
5747 		.inode = state->inode,
5748 	};
5749 	int err;
5750 
5751 	err = nfs4_set_lock_state(state, request);
5752 	if (err != 0)
5753 		return err;
5754 	if (!recover_lost_locks) {
5755 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5756 		return 0;
5757 	}
5758 	do {
5759 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5760 			return 0;
5761 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5762 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5763 		switch (err) {
5764 		default:
5765 			goto out;
5766 		case -NFS4ERR_GRACE:
5767 		case -NFS4ERR_DELAY:
5768 			nfs4_handle_exception(server, err, &exception);
5769 			err = 0;
5770 		}
5771 	} while (exception.retry);
5772 out:
5773 	return err;
5774 }
5775 
5776 #if defined(CONFIG_NFS_V4_1)
5777 /**
5778  * nfs41_check_expired_locks - possibly free a lock stateid
5779  *
5780  * @state: NFSv4 state for an inode
5781  *
5782  * Returns NFS_OK if recovery for this stateid is now finished.
5783  * Otherwise a negative NFS4ERR value is returned.
5784  */
5785 static int nfs41_check_expired_locks(struct nfs4_state *state)
5786 {
5787 	int status, ret = -NFS4ERR_BAD_STATEID;
5788 	struct nfs4_lock_state *lsp;
5789 	struct nfs_server *server = NFS_SERVER(state->inode);
5790 
5791 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5792 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5793 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5794 
5795 			status = nfs41_test_stateid(server,
5796 					&lsp->ls_stateid,
5797 					cred);
5798 			trace_nfs4_test_lock_stateid(state, lsp, status);
5799 			if (status != NFS_OK) {
5800 				/* Free the stateid unless the server
5801 				 * informs us the stateid is unrecognized. */
5802 				if (status != -NFS4ERR_BAD_STATEID)
5803 					nfs41_free_stateid(server,
5804 							&lsp->ls_stateid,
5805 							cred);
5806 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5807 				ret = status;
5808 			}
5809 		}
5810 	};
5811 
5812 	return ret;
5813 }
5814 
5815 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5816 {
5817 	int status = NFS_OK;
5818 
5819 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5820 		status = nfs41_check_expired_locks(state);
5821 	if (status != NFS_OK)
5822 		status = nfs4_lock_expired(state, request);
5823 	return status;
5824 }
5825 #endif
5826 
5827 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5828 {
5829 	struct nfs4_state_owner *sp = state->owner;
5830 	struct nfs_inode *nfsi = NFS_I(state->inode);
5831 	unsigned char fl_flags = request->fl_flags;
5832 	unsigned int seq;
5833 	int status = -ENOLCK;
5834 
5835 	if ((fl_flags & FL_POSIX) &&
5836 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5837 		goto out;
5838 	/* Is this a delegated open? */
5839 	status = nfs4_set_lock_state(state, request);
5840 	if (status != 0)
5841 		goto out;
5842 	request->fl_flags |= FL_ACCESS;
5843 	status = do_vfs_lock(request->fl_file, request);
5844 	if (status < 0)
5845 		goto out;
5846 	down_read(&nfsi->rwsem);
5847 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5848 		/* Yes: cache locks! */
5849 		/* ...but avoid races with delegation recall... */
5850 		request->fl_flags = fl_flags & ~FL_SLEEP;
5851 		status = do_vfs_lock(request->fl_file, request);
5852 		goto out_unlock;
5853 	}
5854 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5855 	up_read(&nfsi->rwsem);
5856 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5857 	if (status != 0)
5858 		goto out;
5859 	down_read(&nfsi->rwsem);
5860 	if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5861 		status = -NFS4ERR_DELAY;
5862 		goto out_unlock;
5863 	}
5864 	/* Note: we always want to sleep here! */
5865 	request->fl_flags = fl_flags | FL_SLEEP;
5866 	if (do_vfs_lock(request->fl_file, request) < 0)
5867 		printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5868 			"manager!\n", __func__);
5869 out_unlock:
5870 	up_read(&nfsi->rwsem);
5871 out:
5872 	request->fl_flags = fl_flags;
5873 	return status;
5874 }
5875 
5876 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5877 {
5878 	struct nfs4_exception exception = {
5879 		.state = state,
5880 		.inode = state->inode,
5881 	};
5882 	int err;
5883 
5884 	do {
5885 		err = _nfs4_proc_setlk(state, cmd, request);
5886 		trace_nfs4_set_lock(request, state, cmd, err);
5887 		if (err == -NFS4ERR_DENIED)
5888 			err = -EAGAIN;
5889 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5890 				err, &exception);
5891 	} while (exception.retry);
5892 	return err;
5893 }
5894 
5895 static int
5896 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5897 {
5898 	struct nfs_open_context *ctx;
5899 	struct nfs4_state *state;
5900 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5901 	int status;
5902 
5903 	/* verify open state */
5904 	ctx = nfs_file_open_context(filp);
5905 	state = ctx->state;
5906 
5907 	if (request->fl_start < 0 || request->fl_end < 0)
5908 		return -EINVAL;
5909 
5910 	if (IS_GETLK(cmd)) {
5911 		if (state != NULL)
5912 			return nfs4_proc_getlk(state, F_GETLK, request);
5913 		return 0;
5914 	}
5915 
5916 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5917 		return -EINVAL;
5918 
5919 	if (request->fl_type == F_UNLCK) {
5920 		if (state != NULL)
5921 			return nfs4_proc_unlck(state, cmd, request);
5922 		return 0;
5923 	}
5924 
5925 	if (state == NULL)
5926 		return -ENOLCK;
5927 	/*
5928 	 * Don't rely on the VFS having checked the file open mode,
5929 	 * since it won't do this for flock() locks.
5930 	 */
5931 	switch (request->fl_type) {
5932 	case F_RDLCK:
5933 		if (!(filp->f_mode & FMODE_READ))
5934 			return -EBADF;
5935 		break;
5936 	case F_WRLCK:
5937 		if (!(filp->f_mode & FMODE_WRITE))
5938 			return -EBADF;
5939 	}
5940 
5941 	do {
5942 		status = nfs4_proc_setlk(state, cmd, request);
5943 		if ((status != -EAGAIN) || IS_SETLK(cmd))
5944 			break;
5945 		timeout = nfs4_set_lock_task_retry(timeout);
5946 		status = -ERESTARTSYS;
5947 		if (signalled())
5948 			break;
5949 	} while(status < 0);
5950 	return status;
5951 }
5952 
5953 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5954 {
5955 	struct nfs_server *server = NFS_SERVER(state->inode);
5956 	int err;
5957 
5958 	err = nfs4_set_lock_state(state, fl);
5959 	if (err != 0)
5960 		return err;
5961 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5962 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5963 }
5964 
5965 struct nfs_release_lockowner_data {
5966 	struct nfs4_lock_state *lsp;
5967 	struct nfs_server *server;
5968 	struct nfs_release_lockowner_args args;
5969 	struct nfs_release_lockowner_res res;
5970 	unsigned long timestamp;
5971 };
5972 
5973 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
5974 {
5975 	struct nfs_release_lockowner_data *data = calldata;
5976 	struct nfs_server *server = data->server;
5977 	nfs40_setup_sequence(server, &data->args.seq_args,
5978 				&data->res.seq_res, task);
5979 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5980 	data->timestamp = jiffies;
5981 }
5982 
5983 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
5984 {
5985 	struct nfs_release_lockowner_data *data = calldata;
5986 	struct nfs_server *server = data->server;
5987 
5988 	nfs40_sequence_done(task, &data->res.seq_res);
5989 
5990 	switch (task->tk_status) {
5991 	case 0:
5992 		renew_lease(server, data->timestamp);
5993 		break;
5994 	case -NFS4ERR_STALE_CLIENTID:
5995 	case -NFS4ERR_EXPIRED:
5996 		nfs4_schedule_lease_recovery(server->nfs_client);
5997 		break;
5998 	case -NFS4ERR_LEASE_MOVED:
5999 	case -NFS4ERR_DELAY:
6000 		if (nfs4_async_handle_error(task, server,
6001 					    NULL, NULL) == -EAGAIN)
6002 			rpc_restart_call_prepare(task);
6003 	}
6004 }
6005 
6006 static void nfs4_release_lockowner_release(void *calldata)
6007 {
6008 	struct nfs_release_lockowner_data *data = calldata;
6009 	nfs4_free_lock_state(data->server, data->lsp);
6010 	kfree(calldata);
6011 }
6012 
6013 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6014 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6015 	.rpc_call_done = nfs4_release_lockowner_done,
6016 	.rpc_release = nfs4_release_lockowner_release,
6017 };
6018 
6019 static void
6020 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6021 {
6022 	struct nfs_release_lockowner_data *data;
6023 	struct rpc_message msg = {
6024 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6025 	};
6026 
6027 	if (server->nfs_client->cl_mvops->minor_version != 0)
6028 		return;
6029 
6030 	data = kmalloc(sizeof(*data), GFP_NOFS);
6031 	if (!data)
6032 		return;
6033 	data->lsp = lsp;
6034 	data->server = server;
6035 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6036 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6037 	data->args.lock_owner.s_dev = server->s_dev;
6038 
6039 	msg.rpc_argp = &data->args;
6040 	msg.rpc_resp = &data->res;
6041 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6042 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6043 }
6044 
6045 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6046 
6047 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6048 				   const void *buf, size_t buflen,
6049 				   int flags, int type)
6050 {
6051 	if (strcmp(key, "") != 0)
6052 		return -EINVAL;
6053 
6054 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6055 }
6056 
6057 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6058 				   void *buf, size_t buflen, int type)
6059 {
6060 	if (strcmp(key, "") != 0)
6061 		return -EINVAL;
6062 
6063 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6064 }
6065 
6066 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6067 				       size_t list_len, const char *name,
6068 				       size_t name_len, int type)
6069 {
6070 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6071 
6072 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6073 		return 0;
6074 
6075 	if (list && len <= list_len)
6076 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6077 	return len;
6078 }
6079 
6080 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6081 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6082 {
6083 	return server->caps & NFS_CAP_SECURITY_LABEL;
6084 }
6085 
6086 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6087 				   const void *buf, size_t buflen,
6088 				   int flags, int type)
6089 {
6090 	if (security_ismaclabel(key))
6091 		return nfs4_set_security_label(dentry, buf, buflen);
6092 
6093 	return -EOPNOTSUPP;
6094 }
6095 
6096 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6097 				   void *buf, size_t buflen, int type)
6098 {
6099 	if (security_ismaclabel(key))
6100 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6101 	return -EOPNOTSUPP;
6102 }
6103 
6104 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6105 				       size_t list_len, const char *name,
6106 				       size_t name_len, int type)
6107 {
6108 	size_t len = 0;
6109 
6110 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6111 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6112 		if (list && len <= list_len)
6113 			security_inode_listsecurity(dentry->d_inode, list, len);
6114 	}
6115 	return len;
6116 }
6117 
6118 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6119 	.prefix = XATTR_SECURITY_PREFIX,
6120 	.list	= nfs4_xattr_list_nfs4_label,
6121 	.get	= nfs4_xattr_get_nfs4_label,
6122 	.set	= nfs4_xattr_set_nfs4_label,
6123 };
6124 #endif
6125 
6126 
6127 /*
6128  * nfs_fhget will use either the mounted_on_fileid or the fileid
6129  */
6130 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6131 {
6132 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6133 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6134 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6135 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6136 		return;
6137 
6138 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6139 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6140 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6141 	fattr->nlink = 2;
6142 }
6143 
6144 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6145 				   const struct qstr *name,
6146 				   struct nfs4_fs_locations *fs_locations,
6147 				   struct page *page)
6148 {
6149 	struct nfs_server *server = NFS_SERVER(dir);
6150 	u32 bitmask[3] = {
6151 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6152 	};
6153 	struct nfs4_fs_locations_arg args = {
6154 		.dir_fh = NFS_FH(dir),
6155 		.name = name,
6156 		.page = page,
6157 		.bitmask = bitmask,
6158 	};
6159 	struct nfs4_fs_locations_res res = {
6160 		.fs_locations = fs_locations,
6161 	};
6162 	struct rpc_message msg = {
6163 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6164 		.rpc_argp = &args,
6165 		.rpc_resp = &res,
6166 	};
6167 	int status;
6168 
6169 	dprintk("%s: start\n", __func__);
6170 
6171 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6172 	 * is not supported */
6173 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6174 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6175 	else
6176 		bitmask[0] |= FATTR4_WORD0_FILEID;
6177 
6178 	nfs_fattr_init(&fs_locations->fattr);
6179 	fs_locations->server = server;
6180 	fs_locations->nlocations = 0;
6181 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6182 	dprintk("%s: returned status = %d\n", __func__, status);
6183 	return status;
6184 }
6185 
6186 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6187 			   const struct qstr *name,
6188 			   struct nfs4_fs_locations *fs_locations,
6189 			   struct page *page)
6190 {
6191 	struct nfs4_exception exception = { };
6192 	int err;
6193 	do {
6194 		err = _nfs4_proc_fs_locations(client, dir, name,
6195 				fs_locations, page);
6196 		trace_nfs4_get_fs_locations(dir, name, err);
6197 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6198 				&exception);
6199 	} while (exception.retry);
6200 	return err;
6201 }
6202 
6203 /*
6204  * This operation also signals the server that this client is
6205  * performing migration recovery.  The server can stop returning
6206  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6207  * appended to this compound to identify the client ID which is
6208  * performing recovery.
6209  */
6210 static int _nfs40_proc_get_locations(struct inode *inode,
6211 				     struct nfs4_fs_locations *locations,
6212 				     struct page *page, struct rpc_cred *cred)
6213 {
6214 	struct nfs_server *server = NFS_SERVER(inode);
6215 	struct rpc_clnt *clnt = server->client;
6216 	u32 bitmask[2] = {
6217 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6218 	};
6219 	struct nfs4_fs_locations_arg args = {
6220 		.clientid	= server->nfs_client->cl_clientid,
6221 		.fh		= NFS_FH(inode),
6222 		.page		= page,
6223 		.bitmask	= bitmask,
6224 		.migration	= 1,		/* skip LOOKUP */
6225 		.renew		= 1,		/* append RENEW */
6226 	};
6227 	struct nfs4_fs_locations_res res = {
6228 		.fs_locations	= locations,
6229 		.migration	= 1,
6230 		.renew		= 1,
6231 	};
6232 	struct rpc_message msg = {
6233 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6234 		.rpc_argp	= &args,
6235 		.rpc_resp	= &res,
6236 		.rpc_cred	= cred,
6237 	};
6238 	unsigned long now = jiffies;
6239 	int status;
6240 
6241 	nfs_fattr_init(&locations->fattr);
6242 	locations->server = server;
6243 	locations->nlocations = 0;
6244 
6245 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6246 	nfs4_set_sequence_privileged(&args.seq_args);
6247 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6248 					&args.seq_args, &res.seq_res);
6249 	if (status)
6250 		return status;
6251 
6252 	renew_lease(server, now);
6253 	return 0;
6254 }
6255 
6256 #ifdef CONFIG_NFS_V4_1
6257 
6258 /*
6259  * This operation also signals the server that this client is
6260  * performing migration recovery.  The server can stop asserting
6261  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6262  * performing this operation is identified in the SEQUENCE
6263  * operation in this compound.
6264  *
6265  * When the client supports GETATTR(fs_locations_info), it can
6266  * be plumbed in here.
6267  */
6268 static int _nfs41_proc_get_locations(struct inode *inode,
6269 				     struct nfs4_fs_locations *locations,
6270 				     struct page *page, struct rpc_cred *cred)
6271 {
6272 	struct nfs_server *server = NFS_SERVER(inode);
6273 	struct rpc_clnt *clnt = server->client;
6274 	u32 bitmask[2] = {
6275 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6276 	};
6277 	struct nfs4_fs_locations_arg args = {
6278 		.fh		= NFS_FH(inode),
6279 		.page		= page,
6280 		.bitmask	= bitmask,
6281 		.migration	= 1,		/* skip LOOKUP */
6282 	};
6283 	struct nfs4_fs_locations_res res = {
6284 		.fs_locations	= locations,
6285 		.migration	= 1,
6286 	};
6287 	struct rpc_message msg = {
6288 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6289 		.rpc_argp	= &args,
6290 		.rpc_resp	= &res,
6291 		.rpc_cred	= cred,
6292 	};
6293 	int status;
6294 
6295 	nfs_fattr_init(&locations->fattr);
6296 	locations->server = server;
6297 	locations->nlocations = 0;
6298 
6299 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6300 	nfs4_set_sequence_privileged(&args.seq_args);
6301 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6302 					&args.seq_args, &res.seq_res);
6303 	if (status == NFS4_OK &&
6304 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6305 		status = -NFS4ERR_LEASE_MOVED;
6306 	return status;
6307 }
6308 
6309 #endif	/* CONFIG_NFS_V4_1 */
6310 
6311 /**
6312  * nfs4_proc_get_locations - discover locations for a migrated FSID
6313  * @inode: inode on FSID that is migrating
6314  * @locations: result of query
6315  * @page: buffer
6316  * @cred: credential to use for this operation
6317  *
6318  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6319  * operation failed, or a negative errno if a local error occurred.
6320  *
6321  * On success, "locations" is filled in, but if the server has
6322  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6323  * asserted.
6324  *
6325  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6326  * from this client that require migration recovery.
6327  */
6328 int nfs4_proc_get_locations(struct inode *inode,
6329 			    struct nfs4_fs_locations *locations,
6330 			    struct page *page, struct rpc_cred *cred)
6331 {
6332 	struct nfs_server *server = NFS_SERVER(inode);
6333 	struct nfs_client *clp = server->nfs_client;
6334 	const struct nfs4_mig_recovery_ops *ops =
6335 					clp->cl_mvops->mig_recovery_ops;
6336 	struct nfs4_exception exception = { };
6337 	int status;
6338 
6339 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6340 		(unsigned long long)server->fsid.major,
6341 		(unsigned long long)server->fsid.minor,
6342 		clp->cl_hostname);
6343 	nfs_display_fhandle(NFS_FH(inode), __func__);
6344 
6345 	do {
6346 		status = ops->get_locations(inode, locations, page, cred);
6347 		if (status != -NFS4ERR_DELAY)
6348 			break;
6349 		nfs4_handle_exception(server, status, &exception);
6350 	} while (exception.retry);
6351 	return status;
6352 }
6353 
6354 /*
6355  * This operation also signals the server that this client is
6356  * performing "lease moved" recovery.  The server can stop
6357  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6358  * is appended to this compound to identify the client ID which is
6359  * performing recovery.
6360  */
6361 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6362 {
6363 	struct nfs_server *server = NFS_SERVER(inode);
6364 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6365 	struct rpc_clnt *clnt = server->client;
6366 	struct nfs4_fsid_present_arg args = {
6367 		.fh		= NFS_FH(inode),
6368 		.clientid	= clp->cl_clientid,
6369 		.renew		= 1,		/* append RENEW */
6370 	};
6371 	struct nfs4_fsid_present_res res = {
6372 		.renew		= 1,
6373 	};
6374 	struct rpc_message msg = {
6375 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6376 		.rpc_argp	= &args,
6377 		.rpc_resp	= &res,
6378 		.rpc_cred	= cred,
6379 	};
6380 	unsigned long now = jiffies;
6381 	int status;
6382 
6383 	res.fh = nfs_alloc_fhandle();
6384 	if (res.fh == NULL)
6385 		return -ENOMEM;
6386 
6387 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6388 	nfs4_set_sequence_privileged(&args.seq_args);
6389 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6390 						&args.seq_args, &res.seq_res);
6391 	nfs_free_fhandle(res.fh);
6392 	if (status)
6393 		return status;
6394 
6395 	do_renew_lease(clp, now);
6396 	return 0;
6397 }
6398 
6399 #ifdef CONFIG_NFS_V4_1
6400 
6401 /*
6402  * This operation also signals the server that this client is
6403  * performing "lease moved" recovery.  The server can stop asserting
6404  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6405  * this operation is identified in the SEQUENCE operation in this
6406  * compound.
6407  */
6408 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6409 {
6410 	struct nfs_server *server = NFS_SERVER(inode);
6411 	struct rpc_clnt *clnt = server->client;
6412 	struct nfs4_fsid_present_arg args = {
6413 		.fh		= NFS_FH(inode),
6414 	};
6415 	struct nfs4_fsid_present_res res = {
6416 	};
6417 	struct rpc_message msg = {
6418 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6419 		.rpc_argp	= &args,
6420 		.rpc_resp	= &res,
6421 		.rpc_cred	= cred,
6422 	};
6423 	int status;
6424 
6425 	res.fh = nfs_alloc_fhandle();
6426 	if (res.fh == NULL)
6427 		return -ENOMEM;
6428 
6429 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6430 	nfs4_set_sequence_privileged(&args.seq_args);
6431 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6432 						&args.seq_args, &res.seq_res);
6433 	nfs_free_fhandle(res.fh);
6434 	if (status == NFS4_OK &&
6435 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6436 		status = -NFS4ERR_LEASE_MOVED;
6437 	return status;
6438 }
6439 
6440 #endif	/* CONFIG_NFS_V4_1 */
6441 
6442 /**
6443  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6444  * @inode: inode on FSID to check
6445  * @cred: credential to use for this operation
6446  *
6447  * Server indicates whether the FSID is present, moved, or not
6448  * recognized.  This operation is necessary to clear a LEASE_MOVED
6449  * condition for this client ID.
6450  *
6451  * Returns NFS4_OK if the FSID is present on this server,
6452  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6453  *  NFS4ERR code if some error occurred on the server, or a
6454  *  negative errno if a local failure occurred.
6455  */
6456 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6457 {
6458 	struct nfs_server *server = NFS_SERVER(inode);
6459 	struct nfs_client *clp = server->nfs_client;
6460 	const struct nfs4_mig_recovery_ops *ops =
6461 					clp->cl_mvops->mig_recovery_ops;
6462 	struct nfs4_exception exception = { };
6463 	int status;
6464 
6465 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6466 		(unsigned long long)server->fsid.major,
6467 		(unsigned long long)server->fsid.minor,
6468 		clp->cl_hostname);
6469 	nfs_display_fhandle(NFS_FH(inode), __func__);
6470 
6471 	do {
6472 		status = ops->fsid_present(inode, cred);
6473 		if (status != -NFS4ERR_DELAY)
6474 			break;
6475 		nfs4_handle_exception(server, status, &exception);
6476 	} while (exception.retry);
6477 	return status;
6478 }
6479 
6480 /**
6481  * If 'use_integrity' is true and the state managment nfs_client
6482  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6483  * and the machine credential as per RFC3530bis and RFC5661 Security
6484  * Considerations sections. Otherwise, just use the user cred with the
6485  * filesystem's rpc_client.
6486  */
6487 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6488 {
6489 	int status;
6490 	struct nfs4_secinfo_arg args = {
6491 		.dir_fh = NFS_FH(dir),
6492 		.name   = name,
6493 	};
6494 	struct nfs4_secinfo_res res = {
6495 		.flavors     = flavors,
6496 	};
6497 	struct rpc_message msg = {
6498 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6499 		.rpc_argp = &args,
6500 		.rpc_resp = &res,
6501 	};
6502 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6503 	struct rpc_cred *cred = NULL;
6504 
6505 	if (use_integrity) {
6506 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6507 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6508 		msg.rpc_cred = cred;
6509 	}
6510 
6511 	dprintk("NFS call  secinfo %s\n", name->name);
6512 
6513 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6514 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6515 
6516 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6517 				&res.seq_res, 0);
6518 	dprintk("NFS reply  secinfo: %d\n", status);
6519 
6520 	if (cred)
6521 		put_rpccred(cred);
6522 
6523 	return status;
6524 }
6525 
6526 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6527 		      struct nfs4_secinfo_flavors *flavors)
6528 {
6529 	struct nfs4_exception exception = { };
6530 	int err;
6531 	do {
6532 		err = -NFS4ERR_WRONGSEC;
6533 
6534 		/* try to use integrity protection with machine cred */
6535 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6536 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6537 
6538 		/*
6539 		 * if unable to use integrity protection, or SECINFO with
6540 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6541 		 * disallowed by spec, but exists in deployed servers) use
6542 		 * the current filesystem's rpc_client and the user cred.
6543 		 */
6544 		if (err == -NFS4ERR_WRONGSEC)
6545 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6546 
6547 		trace_nfs4_secinfo(dir, name, err);
6548 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6549 				&exception);
6550 	} while (exception.retry);
6551 	return err;
6552 }
6553 
6554 #ifdef CONFIG_NFS_V4_1
6555 /*
6556  * Check the exchange flags returned by the server for invalid flags, having
6557  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6558  * DS flags set.
6559  */
6560 static int nfs4_check_cl_exchange_flags(u32 flags)
6561 {
6562 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6563 		goto out_inval;
6564 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6565 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6566 		goto out_inval;
6567 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6568 		goto out_inval;
6569 	return NFS_OK;
6570 out_inval:
6571 	return -NFS4ERR_INVAL;
6572 }
6573 
6574 static bool
6575 nfs41_same_server_scope(struct nfs41_server_scope *a,
6576 			struct nfs41_server_scope *b)
6577 {
6578 	if (a->server_scope_sz == b->server_scope_sz &&
6579 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6580 		return true;
6581 
6582 	return false;
6583 }
6584 
6585 /*
6586  * nfs4_proc_bind_conn_to_session()
6587  *
6588  * The 4.1 client currently uses the same TCP connection for the
6589  * fore and backchannel.
6590  */
6591 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6592 {
6593 	int status;
6594 	struct nfs41_bind_conn_to_session_res res;
6595 	struct rpc_message msg = {
6596 		.rpc_proc =
6597 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6598 		.rpc_argp = clp,
6599 		.rpc_resp = &res,
6600 		.rpc_cred = cred,
6601 	};
6602 
6603 	dprintk("--> %s\n", __func__);
6604 
6605 	res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
6606 	if (unlikely(res.session == NULL)) {
6607 		status = -ENOMEM;
6608 		goto out;
6609 	}
6610 
6611 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6612 	trace_nfs4_bind_conn_to_session(clp, status);
6613 	if (status == 0) {
6614 		if (memcmp(res.session->sess_id.data,
6615 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6616 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6617 			status = -EIO;
6618 			goto out_session;
6619 		}
6620 		if (res.dir != NFS4_CDFS4_BOTH) {
6621 			dprintk("NFS: %s: Unexpected direction from server\n",
6622 				__func__);
6623 			status = -EIO;
6624 			goto out_session;
6625 		}
6626 		if (res.use_conn_in_rdma_mode) {
6627 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6628 				__func__);
6629 			status = -EIO;
6630 			goto out_session;
6631 		}
6632 	}
6633 out_session:
6634 	kfree(res.session);
6635 out:
6636 	dprintk("<-- %s status= %d\n", __func__, status);
6637 	return status;
6638 }
6639 
6640 /*
6641  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6642  * and operations we'd like to see to enable certain features in the allow map
6643  */
6644 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6645 	.how = SP4_MACH_CRED,
6646 	.enforce.u.words = {
6647 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6648 		      1 << (OP_EXCHANGE_ID - 32) |
6649 		      1 << (OP_CREATE_SESSION - 32) |
6650 		      1 << (OP_DESTROY_SESSION - 32) |
6651 		      1 << (OP_DESTROY_CLIENTID - 32)
6652 	},
6653 	.allow.u.words = {
6654 		[0] = 1 << (OP_CLOSE) |
6655 		      1 << (OP_LOCKU) |
6656 		      1 << (OP_COMMIT),
6657 		[1] = 1 << (OP_SECINFO - 32) |
6658 		      1 << (OP_SECINFO_NO_NAME - 32) |
6659 		      1 << (OP_TEST_STATEID - 32) |
6660 		      1 << (OP_FREE_STATEID - 32) |
6661 		      1 << (OP_WRITE - 32)
6662 	}
6663 };
6664 
6665 /*
6666  * Select the state protection mode for client `clp' given the server results
6667  * from exchange_id in `sp'.
6668  *
6669  * Returns 0 on success, negative errno otherwise.
6670  */
6671 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6672 				 struct nfs41_state_protection *sp)
6673 {
6674 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6675 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6676 		      1 << (OP_EXCHANGE_ID - 32) |
6677 		      1 << (OP_CREATE_SESSION - 32) |
6678 		      1 << (OP_DESTROY_SESSION - 32) |
6679 		      1 << (OP_DESTROY_CLIENTID - 32)
6680 	};
6681 	unsigned int i;
6682 
6683 	if (sp->how == SP4_MACH_CRED) {
6684 		/* Print state protect result */
6685 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6686 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6687 			if (test_bit(i, sp->enforce.u.longs))
6688 				dfprintk(MOUNT, "  enforce op %d\n", i);
6689 			if (test_bit(i, sp->allow.u.longs))
6690 				dfprintk(MOUNT, "  allow op %d\n", i);
6691 		}
6692 
6693 		/* make sure nothing is on enforce list that isn't supported */
6694 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6695 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6696 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6697 				return -EINVAL;
6698 			}
6699 		}
6700 
6701 		/*
6702 		 * Minimal mode - state operations are allowed to use machine
6703 		 * credential.  Note this already happens by default, so the
6704 		 * client doesn't have to do anything more than the negotiation.
6705 		 *
6706 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6707 		 *       we're already using the machine cred for exchange_id
6708 		 *       and will never use a different cred.
6709 		 */
6710 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6711 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6712 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6713 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6714 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6715 			dfprintk(MOUNT, "  minimal mode enabled\n");
6716 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6717 		} else {
6718 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6719 			return -EINVAL;
6720 		}
6721 
6722 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6723 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6724 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6725 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6726 		}
6727 
6728 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6729 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6730 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6731 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6732 		}
6733 
6734 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6735 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6736 			dfprintk(MOUNT, "  stateid mode enabled\n");
6737 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6738 		}
6739 
6740 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6741 			dfprintk(MOUNT, "  write mode enabled\n");
6742 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6743 		}
6744 
6745 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6746 			dfprintk(MOUNT, "  commit mode enabled\n");
6747 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6748 		}
6749 	}
6750 
6751 	return 0;
6752 }
6753 
6754 /*
6755  * _nfs4_proc_exchange_id()
6756  *
6757  * Wrapper for EXCHANGE_ID operation.
6758  */
6759 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6760 	u32 sp4_how)
6761 {
6762 	nfs4_verifier verifier;
6763 	struct nfs41_exchange_id_args args = {
6764 		.verifier = &verifier,
6765 		.client = clp,
6766 #ifdef CONFIG_NFS_V4_1_MIGRATION
6767 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6768 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6769 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6770 #else
6771 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6772 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6773 #endif
6774 	};
6775 	struct nfs41_exchange_id_res res = {
6776 		0
6777 	};
6778 	int status;
6779 	struct rpc_message msg = {
6780 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6781 		.rpc_argp = &args,
6782 		.rpc_resp = &res,
6783 		.rpc_cred = cred,
6784 	};
6785 
6786 	nfs4_init_boot_verifier(clp, &verifier);
6787 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6788 							sizeof(args.id));
6789 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6790 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6791 		args.id_len, args.id);
6792 
6793 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6794 					GFP_NOFS);
6795 	if (unlikely(res.server_owner == NULL)) {
6796 		status = -ENOMEM;
6797 		goto out;
6798 	}
6799 
6800 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6801 					GFP_NOFS);
6802 	if (unlikely(res.server_scope == NULL)) {
6803 		status = -ENOMEM;
6804 		goto out_server_owner;
6805 	}
6806 
6807 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6808 	if (unlikely(res.impl_id == NULL)) {
6809 		status = -ENOMEM;
6810 		goto out_server_scope;
6811 	}
6812 
6813 	switch (sp4_how) {
6814 	case SP4_NONE:
6815 		args.state_protect.how = SP4_NONE;
6816 		break;
6817 
6818 	case SP4_MACH_CRED:
6819 		args.state_protect = nfs4_sp4_mach_cred_request;
6820 		break;
6821 
6822 	default:
6823 		/* unsupported! */
6824 		WARN_ON_ONCE(1);
6825 		status = -EINVAL;
6826 		goto out_server_scope;
6827 	}
6828 
6829 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6830 	trace_nfs4_exchange_id(clp, status);
6831 	if (status == 0)
6832 		status = nfs4_check_cl_exchange_flags(res.flags);
6833 
6834 	if (status == 0)
6835 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6836 
6837 	if (status == 0) {
6838 		clp->cl_clientid = res.clientid;
6839 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6840 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6841 			clp->cl_seqid = res.seqid;
6842 
6843 		kfree(clp->cl_serverowner);
6844 		clp->cl_serverowner = res.server_owner;
6845 		res.server_owner = NULL;
6846 
6847 		/* use the most recent implementation id */
6848 		kfree(clp->cl_implid);
6849 		clp->cl_implid = res.impl_id;
6850 
6851 		if (clp->cl_serverscope != NULL &&
6852 		    !nfs41_same_server_scope(clp->cl_serverscope,
6853 					     res.server_scope)) {
6854 			dprintk("%s: server_scope mismatch detected\n",
6855 				__func__);
6856 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6857 			kfree(clp->cl_serverscope);
6858 			clp->cl_serverscope = NULL;
6859 		}
6860 
6861 		if (clp->cl_serverscope == NULL) {
6862 			clp->cl_serverscope = res.server_scope;
6863 			goto out;
6864 		}
6865 	} else
6866 		kfree(res.impl_id);
6867 
6868 out_server_owner:
6869 	kfree(res.server_owner);
6870 out_server_scope:
6871 	kfree(res.server_scope);
6872 out:
6873 	if (clp->cl_implid != NULL)
6874 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6875 			"domain: %s, name: %s, date: %llu,%u\n",
6876 			clp->cl_implid->domain, clp->cl_implid->name,
6877 			clp->cl_implid->date.seconds,
6878 			clp->cl_implid->date.nseconds);
6879 	dprintk("NFS reply exchange_id: %d\n", status);
6880 	return status;
6881 }
6882 
6883 /*
6884  * nfs4_proc_exchange_id()
6885  *
6886  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6887  *
6888  * Since the clientid has expired, all compounds using sessions
6889  * associated with the stale clientid will be returning
6890  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6891  * be in some phase of session reset.
6892  *
6893  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6894  */
6895 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6896 {
6897 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6898 	int status;
6899 
6900 	/* try SP4_MACH_CRED if krb5i/p	*/
6901 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6902 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6903 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6904 		if (!status)
6905 			return 0;
6906 	}
6907 
6908 	/* try SP4_NONE */
6909 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6910 }
6911 
6912 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6913 		struct rpc_cred *cred)
6914 {
6915 	struct rpc_message msg = {
6916 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6917 		.rpc_argp = clp,
6918 		.rpc_cred = cred,
6919 	};
6920 	int status;
6921 
6922 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6923 	trace_nfs4_destroy_clientid(clp, status);
6924 	if (status)
6925 		dprintk("NFS: Got error %d from the server %s on "
6926 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6927 	return status;
6928 }
6929 
6930 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6931 		struct rpc_cred *cred)
6932 {
6933 	unsigned int loop;
6934 	int ret;
6935 
6936 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6937 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6938 		switch (ret) {
6939 		case -NFS4ERR_DELAY:
6940 		case -NFS4ERR_CLIENTID_BUSY:
6941 			ssleep(1);
6942 			break;
6943 		default:
6944 			return ret;
6945 		}
6946 	}
6947 	return 0;
6948 }
6949 
6950 int nfs4_destroy_clientid(struct nfs_client *clp)
6951 {
6952 	struct rpc_cred *cred;
6953 	int ret = 0;
6954 
6955 	if (clp->cl_mvops->minor_version < 1)
6956 		goto out;
6957 	if (clp->cl_exchange_flags == 0)
6958 		goto out;
6959 	if (clp->cl_preserve_clid)
6960 		goto out;
6961 	cred = nfs4_get_clid_cred(clp);
6962 	ret = nfs4_proc_destroy_clientid(clp, cred);
6963 	if (cred)
6964 		put_rpccred(cred);
6965 	switch (ret) {
6966 	case 0:
6967 	case -NFS4ERR_STALE_CLIENTID:
6968 		clp->cl_exchange_flags = 0;
6969 	}
6970 out:
6971 	return ret;
6972 }
6973 
6974 struct nfs4_get_lease_time_data {
6975 	struct nfs4_get_lease_time_args *args;
6976 	struct nfs4_get_lease_time_res *res;
6977 	struct nfs_client *clp;
6978 };
6979 
6980 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6981 					void *calldata)
6982 {
6983 	struct nfs4_get_lease_time_data *data =
6984 			(struct nfs4_get_lease_time_data *)calldata;
6985 
6986 	dprintk("--> %s\n", __func__);
6987 	/* just setup sequence, do not trigger session recovery
6988 	   since we're invoked within one */
6989 	nfs41_setup_sequence(data->clp->cl_session,
6990 			&data->args->la_seq_args,
6991 			&data->res->lr_seq_res,
6992 			task);
6993 	dprintk("<-- %s\n", __func__);
6994 }
6995 
6996 /*
6997  * Called from nfs4_state_manager thread for session setup, so don't recover
6998  * from sequence operation or clientid errors.
6999  */
7000 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7001 {
7002 	struct nfs4_get_lease_time_data *data =
7003 			(struct nfs4_get_lease_time_data *)calldata;
7004 
7005 	dprintk("--> %s\n", __func__);
7006 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7007 		return;
7008 	switch (task->tk_status) {
7009 	case -NFS4ERR_DELAY:
7010 	case -NFS4ERR_GRACE:
7011 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7012 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7013 		task->tk_status = 0;
7014 		/* fall through */
7015 	case -NFS4ERR_RETRY_UNCACHED_REP:
7016 		rpc_restart_call_prepare(task);
7017 		return;
7018 	}
7019 	dprintk("<-- %s\n", __func__);
7020 }
7021 
7022 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7023 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7024 	.rpc_call_done = nfs4_get_lease_time_done,
7025 };
7026 
7027 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7028 {
7029 	struct rpc_task *task;
7030 	struct nfs4_get_lease_time_args args;
7031 	struct nfs4_get_lease_time_res res = {
7032 		.lr_fsinfo = fsinfo,
7033 	};
7034 	struct nfs4_get_lease_time_data data = {
7035 		.args = &args,
7036 		.res = &res,
7037 		.clp = clp,
7038 	};
7039 	struct rpc_message msg = {
7040 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7041 		.rpc_argp = &args,
7042 		.rpc_resp = &res,
7043 	};
7044 	struct rpc_task_setup task_setup = {
7045 		.rpc_client = clp->cl_rpcclient,
7046 		.rpc_message = &msg,
7047 		.callback_ops = &nfs4_get_lease_time_ops,
7048 		.callback_data = &data,
7049 		.flags = RPC_TASK_TIMEOUT,
7050 	};
7051 	int status;
7052 
7053 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7054 	nfs4_set_sequence_privileged(&args.la_seq_args);
7055 	dprintk("--> %s\n", __func__);
7056 	task = rpc_run_task(&task_setup);
7057 
7058 	if (IS_ERR(task))
7059 		status = PTR_ERR(task);
7060 	else {
7061 		status = task->tk_status;
7062 		rpc_put_task(task);
7063 	}
7064 	dprintk("<-- %s return %d\n", __func__, status);
7065 
7066 	return status;
7067 }
7068 
7069 /*
7070  * Initialize the values to be used by the client in CREATE_SESSION
7071  * If nfs4_init_session set the fore channel request and response sizes,
7072  * use them.
7073  *
7074  * Set the back channel max_resp_sz_cached to zero to force the client to
7075  * always set csa_cachethis to FALSE because the current implementation
7076  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7077  */
7078 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7079 {
7080 	unsigned int max_rqst_sz, max_resp_sz;
7081 
7082 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7083 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7084 
7085 	/* Fore channel attributes */
7086 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7087 	args->fc_attrs.max_resp_sz = max_resp_sz;
7088 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7089 	args->fc_attrs.max_reqs = max_session_slots;
7090 
7091 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7092 		"max_ops=%u max_reqs=%u\n",
7093 		__func__,
7094 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7095 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7096 
7097 	/* Back channel attributes */
7098 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7099 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7100 	args->bc_attrs.max_resp_sz_cached = 0;
7101 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7102 	args->bc_attrs.max_reqs = 1;
7103 
7104 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7105 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7106 		__func__,
7107 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7108 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7109 		args->bc_attrs.max_reqs);
7110 }
7111 
7112 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7113 {
7114 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7115 	struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
7116 
7117 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7118 		return -EINVAL;
7119 	/*
7120 	 * Our requested max_ops is the minimum we need; we're not
7121 	 * prepared to break up compounds into smaller pieces than that.
7122 	 * So, no point even trying to continue if the server won't
7123 	 * cooperate:
7124 	 */
7125 	if (rcvd->max_ops < sent->max_ops)
7126 		return -EINVAL;
7127 	if (rcvd->max_reqs == 0)
7128 		return -EINVAL;
7129 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7130 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7131 	return 0;
7132 }
7133 
7134 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
7135 {
7136 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7137 	struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
7138 
7139 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7140 		return -EINVAL;
7141 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7142 		return -EINVAL;
7143 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7144 		return -EINVAL;
7145 	/* These would render the backchannel useless: */
7146 	if (rcvd->max_ops != sent->max_ops)
7147 		return -EINVAL;
7148 	if (rcvd->max_reqs != sent->max_reqs)
7149 		return -EINVAL;
7150 	return 0;
7151 }
7152 
7153 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7154 				     struct nfs4_session *session)
7155 {
7156 	int ret;
7157 
7158 	ret = nfs4_verify_fore_channel_attrs(args, session);
7159 	if (ret)
7160 		return ret;
7161 	return nfs4_verify_back_channel_attrs(args, session);
7162 }
7163 
7164 static int _nfs4_proc_create_session(struct nfs_client *clp,
7165 		struct rpc_cred *cred)
7166 {
7167 	struct nfs4_session *session = clp->cl_session;
7168 	struct nfs41_create_session_args args = {
7169 		.client = clp,
7170 		.cb_program = NFS4_CALLBACK,
7171 	};
7172 	struct nfs41_create_session_res res = {
7173 		.client = clp,
7174 	};
7175 	struct rpc_message msg = {
7176 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7177 		.rpc_argp = &args,
7178 		.rpc_resp = &res,
7179 		.rpc_cred = cred,
7180 	};
7181 	int status;
7182 
7183 	nfs4_init_channel_attrs(&args);
7184 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7185 
7186 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7187 	trace_nfs4_create_session(clp, status);
7188 
7189 	if (!status) {
7190 		/* Verify the session's negotiated channel_attrs values */
7191 		status = nfs4_verify_channel_attrs(&args, session);
7192 		/* Increment the clientid slot sequence id */
7193 		clp->cl_seqid++;
7194 	}
7195 
7196 	return status;
7197 }
7198 
7199 /*
7200  * Issues a CREATE_SESSION operation to the server.
7201  * It is the responsibility of the caller to verify the session is
7202  * expired before calling this routine.
7203  */
7204 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7205 {
7206 	int status;
7207 	unsigned *ptr;
7208 	struct nfs4_session *session = clp->cl_session;
7209 
7210 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7211 
7212 	status = _nfs4_proc_create_session(clp, cred);
7213 	if (status)
7214 		goto out;
7215 
7216 	/* Init or reset the session slot tables */
7217 	status = nfs4_setup_session_slot_tables(session);
7218 	dprintk("slot table setup returned %d\n", status);
7219 	if (status)
7220 		goto out;
7221 
7222 	ptr = (unsigned *)&session->sess_id.data[0];
7223 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7224 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7225 out:
7226 	dprintk("<-- %s\n", __func__);
7227 	return status;
7228 }
7229 
7230 /*
7231  * Issue the over-the-wire RPC DESTROY_SESSION.
7232  * The caller must serialize access to this routine.
7233  */
7234 int nfs4_proc_destroy_session(struct nfs4_session *session,
7235 		struct rpc_cred *cred)
7236 {
7237 	struct rpc_message msg = {
7238 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7239 		.rpc_argp = session,
7240 		.rpc_cred = cred,
7241 	};
7242 	int status = 0;
7243 
7244 	dprintk("--> nfs4_proc_destroy_session\n");
7245 
7246 	/* session is still being setup */
7247 	if (session->clp->cl_cons_state != NFS_CS_READY)
7248 		return status;
7249 
7250 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7251 	trace_nfs4_destroy_session(session->clp, status);
7252 
7253 	if (status)
7254 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7255 			"Session has been destroyed regardless...\n", status);
7256 
7257 	dprintk("<-- nfs4_proc_destroy_session\n");
7258 	return status;
7259 }
7260 
7261 /*
7262  * Renew the cl_session lease.
7263  */
7264 struct nfs4_sequence_data {
7265 	struct nfs_client *clp;
7266 	struct nfs4_sequence_args args;
7267 	struct nfs4_sequence_res res;
7268 };
7269 
7270 static void nfs41_sequence_release(void *data)
7271 {
7272 	struct nfs4_sequence_data *calldata = data;
7273 	struct nfs_client *clp = calldata->clp;
7274 
7275 	if (atomic_read(&clp->cl_count) > 1)
7276 		nfs4_schedule_state_renewal(clp);
7277 	nfs_put_client(clp);
7278 	kfree(calldata);
7279 }
7280 
7281 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7282 {
7283 	switch(task->tk_status) {
7284 	case -NFS4ERR_DELAY:
7285 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7286 		return -EAGAIN;
7287 	default:
7288 		nfs4_schedule_lease_recovery(clp);
7289 	}
7290 	return 0;
7291 }
7292 
7293 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7294 {
7295 	struct nfs4_sequence_data *calldata = data;
7296 	struct nfs_client *clp = calldata->clp;
7297 
7298 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7299 		return;
7300 
7301 	trace_nfs4_sequence(clp, task->tk_status);
7302 	if (task->tk_status < 0) {
7303 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7304 		if (atomic_read(&clp->cl_count) == 1)
7305 			goto out;
7306 
7307 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7308 			rpc_restart_call_prepare(task);
7309 			return;
7310 		}
7311 	}
7312 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7313 out:
7314 	dprintk("<-- %s\n", __func__);
7315 }
7316 
7317 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7318 {
7319 	struct nfs4_sequence_data *calldata = data;
7320 	struct nfs_client *clp = calldata->clp;
7321 	struct nfs4_sequence_args *args;
7322 	struct nfs4_sequence_res *res;
7323 
7324 	args = task->tk_msg.rpc_argp;
7325 	res = task->tk_msg.rpc_resp;
7326 
7327 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7328 }
7329 
7330 static const struct rpc_call_ops nfs41_sequence_ops = {
7331 	.rpc_call_done = nfs41_sequence_call_done,
7332 	.rpc_call_prepare = nfs41_sequence_prepare,
7333 	.rpc_release = nfs41_sequence_release,
7334 };
7335 
7336 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7337 		struct rpc_cred *cred,
7338 		bool is_privileged)
7339 {
7340 	struct nfs4_sequence_data *calldata;
7341 	struct rpc_message msg = {
7342 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7343 		.rpc_cred = cred,
7344 	};
7345 	struct rpc_task_setup task_setup_data = {
7346 		.rpc_client = clp->cl_rpcclient,
7347 		.rpc_message = &msg,
7348 		.callback_ops = &nfs41_sequence_ops,
7349 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7350 	};
7351 
7352 	if (!atomic_inc_not_zero(&clp->cl_count))
7353 		return ERR_PTR(-EIO);
7354 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7355 	if (calldata == NULL) {
7356 		nfs_put_client(clp);
7357 		return ERR_PTR(-ENOMEM);
7358 	}
7359 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7360 	if (is_privileged)
7361 		nfs4_set_sequence_privileged(&calldata->args);
7362 	msg.rpc_argp = &calldata->args;
7363 	msg.rpc_resp = &calldata->res;
7364 	calldata->clp = clp;
7365 	task_setup_data.callback_data = calldata;
7366 
7367 	return rpc_run_task(&task_setup_data);
7368 }
7369 
7370 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7371 {
7372 	struct rpc_task *task;
7373 	int ret = 0;
7374 
7375 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7376 		return -EAGAIN;
7377 	task = _nfs41_proc_sequence(clp, cred, false);
7378 	if (IS_ERR(task))
7379 		ret = PTR_ERR(task);
7380 	else
7381 		rpc_put_task_async(task);
7382 	dprintk("<-- %s status=%d\n", __func__, ret);
7383 	return ret;
7384 }
7385 
7386 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7387 {
7388 	struct rpc_task *task;
7389 	int ret;
7390 
7391 	task = _nfs41_proc_sequence(clp, cred, true);
7392 	if (IS_ERR(task)) {
7393 		ret = PTR_ERR(task);
7394 		goto out;
7395 	}
7396 	ret = rpc_wait_for_completion_task(task);
7397 	if (!ret) {
7398 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7399 
7400 		if (task->tk_status == 0)
7401 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7402 		ret = task->tk_status;
7403 	}
7404 	rpc_put_task(task);
7405 out:
7406 	dprintk("<-- %s status=%d\n", __func__, ret);
7407 	return ret;
7408 }
7409 
7410 struct nfs4_reclaim_complete_data {
7411 	struct nfs_client *clp;
7412 	struct nfs41_reclaim_complete_args arg;
7413 	struct nfs41_reclaim_complete_res res;
7414 };
7415 
7416 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7417 {
7418 	struct nfs4_reclaim_complete_data *calldata = data;
7419 
7420 	nfs41_setup_sequence(calldata->clp->cl_session,
7421 			&calldata->arg.seq_args,
7422 			&calldata->res.seq_res,
7423 			task);
7424 }
7425 
7426 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7427 {
7428 	switch(task->tk_status) {
7429 	case 0:
7430 	case -NFS4ERR_COMPLETE_ALREADY:
7431 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7432 		break;
7433 	case -NFS4ERR_DELAY:
7434 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7435 		/* fall through */
7436 	case -NFS4ERR_RETRY_UNCACHED_REP:
7437 		return -EAGAIN;
7438 	default:
7439 		nfs4_schedule_lease_recovery(clp);
7440 	}
7441 	return 0;
7442 }
7443 
7444 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7445 {
7446 	struct nfs4_reclaim_complete_data *calldata = data;
7447 	struct nfs_client *clp = calldata->clp;
7448 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7449 
7450 	dprintk("--> %s\n", __func__);
7451 	if (!nfs41_sequence_done(task, res))
7452 		return;
7453 
7454 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7455 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7456 		rpc_restart_call_prepare(task);
7457 		return;
7458 	}
7459 	dprintk("<-- %s\n", __func__);
7460 }
7461 
7462 static void nfs4_free_reclaim_complete_data(void *data)
7463 {
7464 	struct nfs4_reclaim_complete_data *calldata = data;
7465 
7466 	kfree(calldata);
7467 }
7468 
7469 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7470 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7471 	.rpc_call_done = nfs4_reclaim_complete_done,
7472 	.rpc_release = nfs4_free_reclaim_complete_data,
7473 };
7474 
7475 /*
7476  * Issue a global reclaim complete.
7477  */
7478 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7479 		struct rpc_cred *cred)
7480 {
7481 	struct nfs4_reclaim_complete_data *calldata;
7482 	struct rpc_task *task;
7483 	struct rpc_message msg = {
7484 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7485 		.rpc_cred = cred,
7486 	};
7487 	struct rpc_task_setup task_setup_data = {
7488 		.rpc_client = clp->cl_rpcclient,
7489 		.rpc_message = &msg,
7490 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7491 		.flags = RPC_TASK_ASYNC,
7492 	};
7493 	int status = -ENOMEM;
7494 
7495 	dprintk("--> %s\n", __func__);
7496 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7497 	if (calldata == NULL)
7498 		goto out;
7499 	calldata->clp = clp;
7500 	calldata->arg.one_fs = 0;
7501 
7502 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7503 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7504 	msg.rpc_argp = &calldata->arg;
7505 	msg.rpc_resp = &calldata->res;
7506 	task_setup_data.callback_data = calldata;
7507 	task = rpc_run_task(&task_setup_data);
7508 	if (IS_ERR(task)) {
7509 		status = PTR_ERR(task);
7510 		goto out;
7511 	}
7512 	status = nfs4_wait_for_completion_rpc_task(task);
7513 	if (status == 0)
7514 		status = task->tk_status;
7515 	rpc_put_task(task);
7516 	return 0;
7517 out:
7518 	dprintk("<-- %s status=%d\n", __func__, status);
7519 	return status;
7520 }
7521 
7522 static void
7523 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7524 {
7525 	struct nfs4_layoutget *lgp = calldata;
7526 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7527 	struct nfs4_session *session = nfs4_get_session(server);
7528 
7529 	dprintk("--> %s\n", __func__);
7530 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7531 	 * right now covering the LAYOUTGET we are about to send.
7532 	 * However, that is not so catastrophic, and there seems
7533 	 * to be no way to prevent it completely.
7534 	 */
7535 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7536 				&lgp->res.seq_res, task))
7537 		return;
7538 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7539 					  NFS_I(lgp->args.inode)->layout,
7540 					  lgp->args.ctx->state)) {
7541 		rpc_exit(task, NFS4_OK);
7542 	}
7543 }
7544 
7545 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7546 {
7547 	struct nfs4_layoutget *lgp = calldata;
7548 	struct inode *inode = lgp->args.inode;
7549 	struct nfs_server *server = NFS_SERVER(inode);
7550 	struct pnfs_layout_hdr *lo;
7551 	struct nfs4_state *state = NULL;
7552 	unsigned long timeo, now, giveup;
7553 
7554 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7555 
7556 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7557 		goto out;
7558 
7559 	switch (task->tk_status) {
7560 	case 0:
7561 		goto out;
7562 	/*
7563 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7564 	 * (or clients) writing to the same RAID stripe
7565 	 */
7566 	case -NFS4ERR_LAYOUTTRYLATER:
7567 	/*
7568 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7569 	 * existing layout before getting a new one).
7570 	 */
7571 	case -NFS4ERR_RECALLCONFLICT:
7572 		timeo = rpc_get_timeout(task->tk_client);
7573 		giveup = lgp->args.timestamp + timeo;
7574 		now = jiffies;
7575 		if (time_after(giveup, now)) {
7576 			unsigned long delay;
7577 
7578 			/* Delay for:
7579 			 * - Not less then NFS4_POLL_RETRY_MIN.
7580 			 * - One last time a jiffie before we give up
7581 			 * - exponential backoff (time_now minus start_attempt)
7582 			 */
7583 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7584 				    min((giveup - now - 1),
7585 					now - lgp->args.timestamp));
7586 
7587 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7588 				__func__, delay);
7589 			rpc_delay(task, delay);
7590 			task->tk_status = 0;
7591 			rpc_restart_call_prepare(task);
7592 			goto out; /* Do not call nfs4_async_handle_error() */
7593 		}
7594 		break;
7595 	case -NFS4ERR_EXPIRED:
7596 	case -NFS4ERR_BAD_STATEID:
7597 		spin_lock(&inode->i_lock);
7598 		lo = NFS_I(inode)->layout;
7599 		if (!lo || list_empty(&lo->plh_segs)) {
7600 			spin_unlock(&inode->i_lock);
7601 			/* If the open stateid was bad, then recover it. */
7602 			state = lgp->args.ctx->state;
7603 		} else {
7604 			LIST_HEAD(head);
7605 
7606 			/*
7607 			 * Mark the bad layout state as invalid, then retry
7608 			 * with the current stateid.
7609 			 */
7610 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7611 			spin_unlock(&inode->i_lock);
7612 			pnfs_free_lseg_list(&head);
7613 
7614 			task->tk_status = 0;
7615 			rpc_restart_call_prepare(task);
7616 		}
7617 	}
7618 	if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7619 		rpc_restart_call_prepare(task);
7620 out:
7621 	dprintk("<-- %s\n", __func__);
7622 }
7623 
7624 static size_t max_response_pages(struct nfs_server *server)
7625 {
7626 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7627 	return nfs_page_array_len(0, max_resp_sz);
7628 }
7629 
7630 static void nfs4_free_pages(struct page **pages, size_t size)
7631 {
7632 	int i;
7633 
7634 	if (!pages)
7635 		return;
7636 
7637 	for (i = 0; i < size; i++) {
7638 		if (!pages[i])
7639 			break;
7640 		__free_page(pages[i]);
7641 	}
7642 	kfree(pages);
7643 }
7644 
7645 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7646 {
7647 	struct page **pages;
7648 	int i;
7649 
7650 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7651 	if (!pages) {
7652 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7653 		return NULL;
7654 	}
7655 
7656 	for (i = 0; i < size; i++) {
7657 		pages[i] = alloc_page(gfp_flags);
7658 		if (!pages[i]) {
7659 			dprintk("%s: failed to allocate page\n", __func__);
7660 			nfs4_free_pages(pages, size);
7661 			return NULL;
7662 		}
7663 	}
7664 
7665 	return pages;
7666 }
7667 
7668 static void nfs4_layoutget_release(void *calldata)
7669 {
7670 	struct nfs4_layoutget *lgp = calldata;
7671 	struct inode *inode = lgp->args.inode;
7672 	struct nfs_server *server = NFS_SERVER(inode);
7673 	size_t max_pages = max_response_pages(server);
7674 
7675 	dprintk("--> %s\n", __func__);
7676 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7677 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7678 	put_nfs_open_context(lgp->args.ctx);
7679 	kfree(calldata);
7680 	dprintk("<-- %s\n", __func__);
7681 }
7682 
7683 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7684 	.rpc_call_prepare = nfs4_layoutget_prepare,
7685 	.rpc_call_done = nfs4_layoutget_done,
7686 	.rpc_release = nfs4_layoutget_release,
7687 };
7688 
7689 struct pnfs_layout_segment *
7690 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7691 {
7692 	struct inode *inode = lgp->args.inode;
7693 	struct nfs_server *server = NFS_SERVER(inode);
7694 	size_t max_pages = max_response_pages(server);
7695 	struct rpc_task *task;
7696 	struct rpc_message msg = {
7697 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7698 		.rpc_argp = &lgp->args,
7699 		.rpc_resp = &lgp->res,
7700 		.rpc_cred = lgp->cred,
7701 	};
7702 	struct rpc_task_setup task_setup_data = {
7703 		.rpc_client = server->client,
7704 		.rpc_message = &msg,
7705 		.callback_ops = &nfs4_layoutget_call_ops,
7706 		.callback_data = lgp,
7707 		.flags = RPC_TASK_ASYNC,
7708 	};
7709 	struct pnfs_layout_segment *lseg = NULL;
7710 	int status = 0;
7711 
7712 	dprintk("--> %s\n", __func__);
7713 
7714 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7715 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7716 
7717 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7718 	if (!lgp->args.layout.pages) {
7719 		nfs4_layoutget_release(lgp);
7720 		return ERR_PTR(-ENOMEM);
7721 	}
7722 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7723 	lgp->args.timestamp = jiffies;
7724 
7725 	lgp->res.layoutp = &lgp->args.layout;
7726 	lgp->res.seq_res.sr_slot = NULL;
7727 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7728 
7729 	task = rpc_run_task(&task_setup_data);
7730 	if (IS_ERR(task))
7731 		return ERR_CAST(task);
7732 	status = nfs4_wait_for_completion_rpc_task(task);
7733 	if (status == 0)
7734 		status = task->tk_status;
7735 	trace_nfs4_layoutget(lgp->args.ctx,
7736 			&lgp->args.range,
7737 			&lgp->res.range,
7738 			status);
7739 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7740 	if (status == 0 && lgp->res.layoutp->len)
7741 		lseg = pnfs_layout_process(lgp);
7742 	rpc_put_task(task);
7743 	dprintk("<-- %s status=%d\n", __func__, status);
7744 	if (status)
7745 		return ERR_PTR(status);
7746 	return lseg;
7747 }
7748 
7749 static void
7750 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7751 {
7752 	struct nfs4_layoutreturn *lrp = calldata;
7753 
7754 	dprintk("--> %s\n", __func__);
7755 	nfs41_setup_sequence(lrp->clp->cl_session,
7756 			&lrp->args.seq_args,
7757 			&lrp->res.seq_res,
7758 			task);
7759 }
7760 
7761 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7762 {
7763 	struct nfs4_layoutreturn *lrp = calldata;
7764 	struct nfs_server *server;
7765 
7766 	dprintk("--> %s\n", __func__);
7767 
7768 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7769 		return;
7770 
7771 	server = NFS_SERVER(lrp->args.inode);
7772 	switch (task->tk_status) {
7773 	default:
7774 		task->tk_status = 0;
7775 	case 0:
7776 		break;
7777 	case -NFS4ERR_DELAY:
7778 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7779 			break;
7780 		rpc_restart_call_prepare(task);
7781 		return;
7782 	}
7783 	dprintk("<-- %s\n", __func__);
7784 }
7785 
7786 static void nfs4_layoutreturn_release(void *calldata)
7787 {
7788 	struct nfs4_layoutreturn *lrp = calldata;
7789 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7790 
7791 	dprintk("--> %s\n", __func__);
7792 	spin_lock(&lo->plh_inode->i_lock);
7793 	if (lrp->res.lrs_present)
7794 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7795 	lo->plh_block_lgets--;
7796 	spin_unlock(&lo->plh_inode->i_lock);
7797 	pnfs_put_layout_hdr(lrp->args.layout);
7798 	kfree(calldata);
7799 	dprintk("<-- %s\n", __func__);
7800 }
7801 
7802 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7803 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7804 	.rpc_call_done = nfs4_layoutreturn_done,
7805 	.rpc_release = nfs4_layoutreturn_release,
7806 };
7807 
7808 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
7809 {
7810 	struct rpc_task *task;
7811 	struct rpc_message msg = {
7812 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7813 		.rpc_argp = &lrp->args,
7814 		.rpc_resp = &lrp->res,
7815 		.rpc_cred = lrp->cred,
7816 	};
7817 	struct rpc_task_setup task_setup_data = {
7818 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7819 		.rpc_message = &msg,
7820 		.callback_ops = &nfs4_layoutreturn_call_ops,
7821 		.callback_data = lrp,
7822 	};
7823 	int status;
7824 
7825 	dprintk("--> %s\n", __func__);
7826 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7827 	task = rpc_run_task(&task_setup_data);
7828 	if (IS_ERR(task))
7829 		return PTR_ERR(task);
7830 	status = task->tk_status;
7831 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7832 	dprintk("<-- %s status=%d\n", __func__, status);
7833 	rpc_put_task(task);
7834 	return status;
7835 }
7836 
7837 static int
7838 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7839 		struct pnfs_device *pdev,
7840 		struct rpc_cred *cred)
7841 {
7842 	struct nfs4_getdeviceinfo_args args = {
7843 		.pdev = pdev,
7844 	};
7845 	struct nfs4_getdeviceinfo_res res = {
7846 		.pdev = pdev,
7847 	};
7848 	struct rpc_message msg = {
7849 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7850 		.rpc_argp = &args,
7851 		.rpc_resp = &res,
7852 		.rpc_cred = cred,
7853 	};
7854 	int status;
7855 
7856 	dprintk("--> %s\n", __func__);
7857 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7858 	dprintk("<-- %s status=%d\n", __func__, status);
7859 
7860 	return status;
7861 }
7862 
7863 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7864 		struct pnfs_device *pdev,
7865 		struct rpc_cred *cred)
7866 {
7867 	struct nfs4_exception exception = { };
7868 	int err;
7869 
7870 	do {
7871 		err = nfs4_handle_exception(server,
7872 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7873 					&exception);
7874 	} while (exception.retry);
7875 	return err;
7876 }
7877 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7878 
7879 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7880 {
7881 	struct nfs4_layoutcommit_data *data = calldata;
7882 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7883 	struct nfs4_session *session = nfs4_get_session(server);
7884 
7885 	nfs41_setup_sequence(session,
7886 			&data->args.seq_args,
7887 			&data->res.seq_res,
7888 			task);
7889 }
7890 
7891 static void
7892 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7893 {
7894 	struct nfs4_layoutcommit_data *data = calldata;
7895 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7896 
7897 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7898 		return;
7899 
7900 	switch (task->tk_status) { /* Just ignore these failures */
7901 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7902 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7903 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7904 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7905 		task->tk_status = 0;
7906 	case 0:
7907 		break;
7908 	default:
7909 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
7910 			rpc_restart_call_prepare(task);
7911 			return;
7912 		}
7913 	}
7914 }
7915 
7916 static void nfs4_layoutcommit_release(void *calldata)
7917 {
7918 	struct nfs4_layoutcommit_data *data = calldata;
7919 
7920 	pnfs_cleanup_layoutcommit(data);
7921 	nfs_post_op_update_inode_force_wcc(data->args.inode,
7922 					   data->res.fattr);
7923 	put_rpccred(data->cred);
7924 	kfree(data);
7925 }
7926 
7927 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7928 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
7929 	.rpc_call_done = nfs4_layoutcommit_done,
7930 	.rpc_release = nfs4_layoutcommit_release,
7931 };
7932 
7933 int
7934 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7935 {
7936 	struct rpc_message msg = {
7937 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7938 		.rpc_argp = &data->args,
7939 		.rpc_resp = &data->res,
7940 		.rpc_cred = data->cred,
7941 	};
7942 	struct rpc_task_setup task_setup_data = {
7943 		.task = &data->task,
7944 		.rpc_client = NFS_CLIENT(data->args.inode),
7945 		.rpc_message = &msg,
7946 		.callback_ops = &nfs4_layoutcommit_ops,
7947 		.callback_data = data,
7948 		.flags = RPC_TASK_ASYNC,
7949 	};
7950 	struct rpc_task *task;
7951 	int status = 0;
7952 
7953 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7954 		"lbw: %llu inode %lu\n",
7955 		data->task.tk_pid, sync,
7956 		data->args.lastbytewritten,
7957 		data->args.inode->i_ino);
7958 
7959 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7960 	task = rpc_run_task(&task_setup_data);
7961 	if (IS_ERR(task))
7962 		return PTR_ERR(task);
7963 	if (sync == false)
7964 		goto out;
7965 	status = nfs4_wait_for_completion_rpc_task(task);
7966 	if (status != 0)
7967 		goto out;
7968 	status = task->tk_status;
7969 	trace_nfs4_layoutcommit(data->args.inode, status);
7970 out:
7971 	dprintk("%s: status %d\n", __func__, status);
7972 	rpc_put_task(task);
7973 	return status;
7974 }
7975 
7976 /**
7977  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
7978  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
7979  */
7980 static int
7981 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7982 		    struct nfs_fsinfo *info,
7983 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7984 {
7985 	struct nfs41_secinfo_no_name_args args = {
7986 		.style = SECINFO_STYLE_CURRENT_FH,
7987 	};
7988 	struct nfs4_secinfo_res res = {
7989 		.flavors = flavors,
7990 	};
7991 	struct rpc_message msg = {
7992 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
7993 		.rpc_argp = &args,
7994 		.rpc_resp = &res,
7995 	};
7996 	struct rpc_clnt *clnt = server->client;
7997 	struct rpc_cred *cred = NULL;
7998 	int status;
7999 
8000 	if (use_integrity) {
8001 		clnt = server->nfs_client->cl_rpcclient;
8002 		cred = nfs4_get_clid_cred(server->nfs_client);
8003 		msg.rpc_cred = cred;
8004 	}
8005 
8006 	dprintk("--> %s\n", __func__);
8007 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8008 				&res.seq_res, 0);
8009 	dprintk("<-- %s status=%d\n", __func__, status);
8010 
8011 	if (cred)
8012 		put_rpccred(cred);
8013 
8014 	return status;
8015 }
8016 
8017 static int
8018 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8019 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8020 {
8021 	struct nfs4_exception exception = { };
8022 	int err;
8023 	do {
8024 		/* first try using integrity protection */
8025 		err = -NFS4ERR_WRONGSEC;
8026 
8027 		/* try to use integrity protection with machine cred */
8028 		if (_nfs4_is_integrity_protected(server->nfs_client))
8029 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8030 							  flavors, true);
8031 
8032 		/*
8033 		 * if unable to use integrity protection, or SECINFO with
8034 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8035 		 * disallowed by spec, but exists in deployed servers) use
8036 		 * the current filesystem's rpc_client and the user cred.
8037 		 */
8038 		if (err == -NFS4ERR_WRONGSEC)
8039 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8040 							  flavors, false);
8041 
8042 		switch (err) {
8043 		case 0:
8044 		case -NFS4ERR_WRONGSEC:
8045 		case -ENOTSUPP:
8046 			goto out;
8047 		default:
8048 			err = nfs4_handle_exception(server, err, &exception);
8049 		}
8050 	} while (exception.retry);
8051 out:
8052 	return err;
8053 }
8054 
8055 static int
8056 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8057 		    struct nfs_fsinfo *info)
8058 {
8059 	int err;
8060 	struct page *page;
8061 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8062 	struct nfs4_secinfo_flavors *flavors;
8063 	struct nfs4_secinfo4 *secinfo;
8064 	int i;
8065 
8066 	page = alloc_page(GFP_KERNEL);
8067 	if (!page) {
8068 		err = -ENOMEM;
8069 		goto out;
8070 	}
8071 
8072 	flavors = page_address(page);
8073 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8074 
8075 	/*
8076 	 * Fall back on "guess and check" method if
8077 	 * the server doesn't support SECINFO_NO_NAME
8078 	 */
8079 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8080 		err = nfs4_find_root_sec(server, fhandle, info);
8081 		goto out_freepage;
8082 	}
8083 	if (err)
8084 		goto out_freepage;
8085 
8086 	for (i = 0; i < flavors->num_flavors; i++) {
8087 		secinfo = &flavors->flavors[i];
8088 
8089 		switch (secinfo->flavor) {
8090 		case RPC_AUTH_NULL:
8091 		case RPC_AUTH_UNIX:
8092 		case RPC_AUTH_GSS:
8093 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8094 					&secinfo->flavor_info);
8095 			break;
8096 		default:
8097 			flavor = RPC_AUTH_MAXFLAVOR;
8098 			break;
8099 		}
8100 
8101 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8102 			flavor = RPC_AUTH_MAXFLAVOR;
8103 
8104 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8105 			err = nfs4_lookup_root_sec(server, fhandle,
8106 						   info, flavor);
8107 			if (!err)
8108 				break;
8109 		}
8110 	}
8111 
8112 	if (flavor == RPC_AUTH_MAXFLAVOR)
8113 		err = -EPERM;
8114 
8115 out_freepage:
8116 	put_page(page);
8117 	if (err == -EACCES)
8118 		return -EPERM;
8119 out:
8120 	return err;
8121 }
8122 
8123 static int _nfs41_test_stateid(struct nfs_server *server,
8124 		nfs4_stateid *stateid,
8125 		struct rpc_cred *cred)
8126 {
8127 	int status;
8128 	struct nfs41_test_stateid_args args = {
8129 		.stateid = stateid,
8130 	};
8131 	struct nfs41_test_stateid_res res;
8132 	struct rpc_message msg = {
8133 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8134 		.rpc_argp = &args,
8135 		.rpc_resp = &res,
8136 		.rpc_cred = cred,
8137 	};
8138 	struct rpc_clnt *rpc_client = server->client;
8139 
8140 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8141 		&rpc_client, &msg);
8142 
8143 	dprintk("NFS call  test_stateid %p\n", stateid);
8144 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8145 	nfs4_set_sequence_privileged(&args.seq_args);
8146 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8147 			&args.seq_args, &res.seq_res);
8148 	if (status != NFS_OK) {
8149 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8150 		return status;
8151 	}
8152 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8153 	return -res.status;
8154 }
8155 
8156 /**
8157  * nfs41_test_stateid - perform a TEST_STATEID operation
8158  *
8159  * @server: server / transport on which to perform the operation
8160  * @stateid: state ID to test
8161  * @cred: credential
8162  *
8163  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8164  * Otherwise a negative NFS4ERR value is returned if the operation
8165  * failed or the state ID is not currently valid.
8166  */
8167 static int nfs41_test_stateid(struct nfs_server *server,
8168 		nfs4_stateid *stateid,
8169 		struct rpc_cred *cred)
8170 {
8171 	struct nfs4_exception exception = { };
8172 	int err;
8173 	do {
8174 		err = _nfs41_test_stateid(server, stateid, cred);
8175 		if (err != -NFS4ERR_DELAY)
8176 			break;
8177 		nfs4_handle_exception(server, err, &exception);
8178 	} while (exception.retry);
8179 	return err;
8180 }
8181 
8182 struct nfs_free_stateid_data {
8183 	struct nfs_server *server;
8184 	struct nfs41_free_stateid_args args;
8185 	struct nfs41_free_stateid_res res;
8186 };
8187 
8188 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8189 {
8190 	struct nfs_free_stateid_data *data = calldata;
8191 	nfs41_setup_sequence(nfs4_get_session(data->server),
8192 			&data->args.seq_args,
8193 			&data->res.seq_res,
8194 			task);
8195 }
8196 
8197 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8198 {
8199 	struct nfs_free_stateid_data *data = calldata;
8200 
8201 	nfs41_sequence_done(task, &data->res.seq_res);
8202 
8203 	switch (task->tk_status) {
8204 	case -NFS4ERR_DELAY:
8205 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8206 			rpc_restart_call_prepare(task);
8207 	}
8208 }
8209 
8210 static void nfs41_free_stateid_release(void *calldata)
8211 {
8212 	kfree(calldata);
8213 }
8214 
8215 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8216 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8217 	.rpc_call_done = nfs41_free_stateid_done,
8218 	.rpc_release = nfs41_free_stateid_release,
8219 };
8220 
8221 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8222 		nfs4_stateid *stateid,
8223 		struct rpc_cred *cred,
8224 		bool privileged)
8225 {
8226 	struct rpc_message msg = {
8227 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8228 		.rpc_cred = cred,
8229 	};
8230 	struct rpc_task_setup task_setup = {
8231 		.rpc_client = server->client,
8232 		.rpc_message = &msg,
8233 		.callback_ops = &nfs41_free_stateid_ops,
8234 		.flags = RPC_TASK_ASYNC,
8235 	};
8236 	struct nfs_free_stateid_data *data;
8237 
8238 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8239 		&task_setup.rpc_client, &msg);
8240 
8241 	dprintk("NFS call  free_stateid %p\n", stateid);
8242 	data = kmalloc(sizeof(*data), GFP_NOFS);
8243 	if (!data)
8244 		return ERR_PTR(-ENOMEM);
8245 	data->server = server;
8246 	nfs4_stateid_copy(&data->args.stateid, stateid);
8247 
8248 	task_setup.callback_data = data;
8249 
8250 	msg.rpc_argp = &data->args;
8251 	msg.rpc_resp = &data->res;
8252 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8253 	if (privileged)
8254 		nfs4_set_sequence_privileged(&data->args.seq_args);
8255 
8256 	return rpc_run_task(&task_setup);
8257 }
8258 
8259 /**
8260  * nfs41_free_stateid - perform a FREE_STATEID operation
8261  *
8262  * @server: server / transport on which to perform the operation
8263  * @stateid: state ID to release
8264  * @cred: credential
8265  *
8266  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8267  * negative NFS4ERR value is returned.
8268  */
8269 static int nfs41_free_stateid(struct nfs_server *server,
8270 		nfs4_stateid *stateid,
8271 		struct rpc_cred *cred)
8272 {
8273 	struct rpc_task *task;
8274 	int ret;
8275 
8276 	task = _nfs41_free_stateid(server, stateid, cred, true);
8277 	if (IS_ERR(task))
8278 		return PTR_ERR(task);
8279 	ret = rpc_wait_for_completion_task(task);
8280 	if (!ret)
8281 		ret = task->tk_status;
8282 	rpc_put_task(task);
8283 	return ret;
8284 }
8285 
8286 static void
8287 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8288 {
8289 	struct rpc_task *task;
8290 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8291 
8292 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8293 	nfs4_free_lock_state(server, lsp);
8294 	if (IS_ERR(task))
8295 		return;
8296 	rpc_put_task(task);
8297 }
8298 
8299 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8300 		const nfs4_stateid *s2)
8301 {
8302 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8303 		return false;
8304 
8305 	if (s1->seqid == s2->seqid)
8306 		return true;
8307 	if (s1->seqid == 0 || s2->seqid == 0)
8308 		return true;
8309 
8310 	return false;
8311 }
8312 
8313 #endif /* CONFIG_NFS_V4_1 */
8314 
8315 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8316 		const nfs4_stateid *s2)
8317 {
8318 	return nfs4_stateid_match(s1, s2);
8319 }
8320 
8321 
8322 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8323 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8324 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8325 	.recover_open	= nfs4_open_reclaim,
8326 	.recover_lock	= nfs4_lock_reclaim,
8327 	.establish_clid = nfs4_init_clientid,
8328 	.detect_trunking = nfs40_discover_server_trunking,
8329 };
8330 
8331 #if defined(CONFIG_NFS_V4_1)
8332 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8333 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8334 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8335 	.recover_open	= nfs4_open_reclaim,
8336 	.recover_lock	= nfs4_lock_reclaim,
8337 	.establish_clid = nfs41_init_clientid,
8338 	.reclaim_complete = nfs41_proc_reclaim_complete,
8339 	.detect_trunking = nfs41_discover_server_trunking,
8340 };
8341 #endif /* CONFIG_NFS_V4_1 */
8342 
8343 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8344 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8345 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8346 	.recover_open	= nfs40_open_expired,
8347 	.recover_lock	= nfs4_lock_expired,
8348 	.establish_clid = nfs4_init_clientid,
8349 };
8350 
8351 #if defined(CONFIG_NFS_V4_1)
8352 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8353 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8354 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8355 	.recover_open	= nfs41_open_expired,
8356 	.recover_lock	= nfs41_lock_expired,
8357 	.establish_clid = nfs41_init_clientid,
8358 };
8359 #endif /* CONFIG_NFS_V4_1 */
8360 
8361 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8362 	.sched_state_renewal = nfs4_proc_async_renew,
8363 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8364 	.renew_lease = nfs4_proc_renew,
8365 };
8366 
8367 #if defined(CONFIG_NFS_V4_1)
8368 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8369 	.sched_state_renewal = nfs41_proc_async_sequence,
8370 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8371 	.renew_lease = nfs4_proc_sequence,
8372 };
8373 #endif
8374 
8375 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8376 	.get_locations = _nfs40_proc_get_locations,
8377 	.fsid_present = _nfs40_proc_fsid_present,
8378 };
8379 
8380 #if defined(CONFIG_NFS_V4_1)
8381 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8382 	.get_locations = _nfs41_proc_get_locations,
8383 	.fsid_present = _nfs41_proc_fsid_present,
8384 };
8385 #endif	/* CONFIG_NFS_V4_1 */
8386 
8387 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8388 	.minor_version = 0,
8389 	.init_caps = NFS_CAP_READDIRPLUS
8390 		| NFS_CAP_ATOMIC_OPEN
8391 		| NFS_CAP_CHANGE_ATTR
8392 		| NFS_CAP_POSIX_LOCK,
8393 	.init_client = nfs40_init_client,
8394 	.shutdown_client = nfs40_shutdown_client,
8395 	.match_stateid = nfs4_match_stateid,
8396 	.find_root_sec = nfs4_find_root_sec,
8397 	.free_lock_state = nfs4_release_lockowner,
8398 	.call_sync_ops = &nfs40_call_sync_ops,
8399 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8400 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8401 	.state_renewal_ops = &nfs40_state_renewal_ops,
8402 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8403 };
8404 
8405 #if defined(CONFIG_NFS_V4_1)
8406 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8407 	.minor_version = 1,
8408 	.init_caps = NFS_CAP_READDIRPLUS
8409 		| NFS_CAP_ATOMIC_OPEN
8410 		| NFS_CAP_CHANGE_ATTR
8411 		| NFS_CAP_POSIX_LOCK
8412 		| NFS_CAP_STATEID_NFSV41
8413 		| NFS_CAP_ATOMIC_OPEN_V1,
8414 	.init_client = nfs41_init_client,
8415 	.shutdown_client = nfs41_shutdown_client,
8416 	.match_stateid = nfs41_match_stateid,
8417 	.find_root_sec = nfs41_find_root_sec,
8418 	.free_lock_state = nfs41_free_lock_state,
8419 	.call_sync_ops = &nfs41_call_sync_ops,
8420 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8421 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8422 	.state_renewal_ops = &nfs41_state_renewal_ops,
8423 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8424 };
8425 #endif
8426 
8427 #if defined(CONFIG_NFS_V4_2)
8428 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8429 	.minor_version = 2,
8430 	.init_caps = NFS_CAP_READDIRPLUS
8431 		| NFS_CAP_ATOMIC_OPEN
8432 		| NFS_CAP_CHANGE_ATTR
8433 		| NFS_CAP_POSIX_LOCK
8434 		| NFS_CAP_STATEID_NFSV41
8435 		| NFS_CAP_ATOMIC_OPEN_V1
8436 		| NFS_CAP_ALLOCATE
8437 		| NFS_CAP_DEALLOCATE
8438 		| NFS_CAP_SEEK,
8439 	.init_client = nfs41_init_client,
8440 	.shutdown_client = nfs41_shutdown_client,
8441 	.match_stateid = nfs41_match_stateid,
8442 	.find_root_sec = nfs41_find_root_sec,
8443 	.free_lock_state = nfs41_free_lock_state,
8444 	.call_sync_ops = &nfs41_call_sync_ops,
8445 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8446 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8447 	.state_renewal_ops = &nfs41_state_renewal_ops,
8448 };
8449 #endif
8450 
8451 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8452 	[0] = &nfs_v4_0_minor_ops,
8453 #if defined(CONFIG_NFS_V4_1)
8454 	[1] = &nfs_v4_1_minor_ops,
8455 #endif
8456 #if defined(CONFIG_NFS_V4_2)
8457 	[2] = &nfs_v4_2_minor_ops,
8458 #endif
8459 };
8460 
8461 static const struct inode_operations nfs4_dir_inode_operations = {
8462 	.create		= nfs_create,
8463 	.lookup		= nfs_lookup,
8464 	.atomic_open	= nfs_atomic_open,
8465 	.link		= nfs_link,
8466 	.unlink		= nfs_unlink,
8467 	.symlink	= nfs_symlink,
8468 	.mkdir		= nfs_mkdir,
8469 	.rmdir		= nfs_rmdir,
8470 	.mknod		= nfs_mknod,
8471 	.rename		= nfs_rename,
8472 	.permission	= nfs_permission,
8473 	.getattr	= nfs_getattr,
8474 	.setattr	= nfs_setattr,
8475 	.getxattr	= generic_getxattr,
8476 	.setxattr	= generic_setxattr,
8477 	.listxattr	= generic_listxattr,
8478 	.removexattr	= generic_removexattr,
8479 };
8480 
8481 static const struct inode_operations nfs4_file_inode_operations = {
8482 	.permission	= nfs_permission,
8483 	.getattr	= nfs_getattr,
8484 	.setattr	= nfs_setattr,
8485 	.getxattr	= generic_getxattr,
8486 	.setxattr	= generic_setxattr,
8487 	.listxattr	= generic_listxattr,
8488 	.removexattr	= generic_removexattr,
8489 };
8490 
8491 const struct nfs_rpc_ops nfs_v4_clientops = {
8492 	.version	= 4,			/* protocol version */
8493 	.dentry_ops	= &nfs4_dentry_operations,
8494 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8495 	.file_inode_ops	= &nfs4_file_inode_operations,
8496 	.file_ops	= &nfs4_file_operations,
8497 	.getroot	= nfs4_proc_get_root,
8498 	.submount	= nfs4_submount,
8499 	.try_mount	= nfs4_try_mount,
8500 	.getattr	= nfs4_proc_getattr,
8501 	.setattr	= nfs4_proc_setattr,
8502 	.lookup		= nfs4_proc_lookup,
8503 	.access		= nfs4_proc_access,
8504 	.readlink	= nfs4_proc_readlink,
8505 	.create		= nfs4_proc_create,
8506 	.remove		= nfs4_proc_remove,
8507 	.unlink_setup	= nfs4_proc_unlink_setup,
8508 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8509 	.unlink_done	= nfs4_proc_unlink_done,
8510 	.rename_setup	= nfs4_proc_rename_setup,
8511 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8512 	.rename_done	= nfs4_proc_rename_done,
8513 	.link		= nfs4_proc_link,
8514 	.symlink	= nfs4_proc_symlink,
8515 	.mkdir		= nfs4_proc_mkdir,
8516 	.rmdir		= nfs4_proc_remove,
8517 	.readdir	= nfs4_proc_readdir,
8518 	.mknod		= nfs4_proc_mknod,
8519 	.statfs		= nfs4_proc_statfs,
8520 	.fsinfo		= nfs4_proc_fsinfo,
8521 	.pathconf	= nfs4_proc_pathconf,
8522 	.set_capabilities = nfs4_server_capabilities,
8523 	.decode_dirent	= nfs4_decode_dirent,
8524 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8525 	.read_setup	= nfs4_proc_read_setup,
8526 	.read_done	= nfs4_read_done,
8527 	.write_setup	= nfs4_proc_write_setup,
8528 	.write_done	= nfs4_write_done,
8529 	.commit_setup	= nfs4_proc_commit_setup,
8530 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8531 	.commit_done	= nfs4_commit_done,
8532 	.lock		= nfs4_proc_lock,
8533 	.clear_acl_cache = nfs4_zap_acl_attr,
8534 	.close_context  = nfs4_close_context,
8535 	.open_context	= nfs4_atomic_open,
8536 	.have_delegation = nfs4_have_delegation,
8537 	.return_delegation = nfs4_inode_return_delegation,
8538 	.alloc_client	= nfs4_alloc_client,
8539 	.init_client	= nfs4_init_client,
8540 	.free_client	= nfs4_free_client,
8541 	.create_server	= nfs4_create_server,
8542 	.clone_server	= nfs_clone_server,
8543 };
8544 
8545 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8546 	.prefix	= XATTR_NAME_NFSV4_ACL,
8547 	.list	= nfs4_xattr_list_nfs4_acl,
8548 	.get	= nfs4_xattr_get_nfs4_acl,
8549 	.set	= nfs4_xattr_set_nfs4_acl,
8550 };
8551 
8552 const struct xattr_handler *nfs4_xattr_handlers[] = {
8553 	&nfs4_xattr_nfs4_acl_handler,
8554 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8555 	&nfs4_xattr_nfs4_label_handler,
8556 #endif
8557 	NULL
8558 };
8559 
8560 /*
8561  * Local variables:
8562  *  c-basic-offset: 8
8563  * End:
8564  */
8565