xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision c1d45424)
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 #define NFSDBG_FACILITY		NFSDBG_PROC
70 
71 #define NFS4_POLL_RETRY_MIN	(HZ/10)
72 #define NFS4_POLL_RETRY_MAX	(15*HZ)
73 
74 struct nfs4_opendata;
75 static int _nfs4_proc_open(struct nfs4_opendata *data);
76 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
77 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
78 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
79 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
80 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
81 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
82 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
83 			    struct nfs_fattr *fattr, struct iattr *sattr,
84 			    struct nfs4_state *state, struct nfs4_label *ilabel,
85 			    struct nfs4_label *olabel);
86 #ifdef CONFIG_NFS_V4_1
87 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
88 		struct rpc_cred *);
89 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 #endif
92 
93 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
94 static inline struct nfs4_label *
95 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
96 	struct iattr *sattr, struct nfs4_label *label)
97 {
98 	int err;
99 
100 	if (label == NULL)
101 		return NULL;
102 
103 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
104 		return NULL;
105 
106 	if (NFS_SERVER(dir)->nfs_client->cl_minorversion < 2)
107 		return NULL;
108 
109 	err = security_dentry_init_security(dentry, sattr->ia_mode,
110 				&dentry->d_name, (void **)&label->label, &label->len);
111 	if (err == 0)
112 		return label;
113 
114 	return NULL;
115 }
116 static inline void
117 nfs4_label_release_security(struct nfs4_label *label)
118 {
119 	if (label)
120 		security_release_secctx(label->label, label->len);
121 }
122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
123 {
124 	if (label)
125 		return server->attr_bitmask;
126 
127 	return server->attr_bitmask_nl;
128 }
129 #else
130 static inline struct nfs4_label *
131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
132 	struct iattr *sattr, struct nfs4_label *l)
133 { return NULL; }
134 static inline void
135 nfs4_label_release_security(struct nfs4_label *label)
136 { return; }
137 static inline u32 *
138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
139 { return server->attr_bitmask; }
140 #endif
141 
142 /* Prevent leaks of NFSv4 errors into userland */
143 static int nfs4_map_errors(int err)
144 {
145 	if (err >= -1000)
146 		return err;
147 	switch (err) {
148 	case -NFS4ERR_RESOURCE:
149 	case -NFS4ERR_LAYOUTTRYLATER:
150 	case -NFS4ERR_RECALLCONFLICT:
151 		return -EREMOTEIO;
152 	case -NFS4ERR_WRONGSEC:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_ACCESS:
162 		return -EACCES;
163 	case -NFS4ERR_FILE_OPEN:
164 		return -EBUSY;
165 	default:
166 		dprintk("%s could not handle NFSv4 error %d\n",
167 				__func__, -err);
168 		break;
169 	}
170 	return -EIO;
171 }
172 
173 /*
174  * This is our standard bitmap for GETATTR requests.
175  */
176 const u32 nfs4_fattr_bitmap[3] = {
177 	FATTR4_WORD0_TYPE
178 	| FATTR4_WORD0_CHANGE
179 	| FATTR4_WORD0_SIZE
180 	| FATTR4_WORD0_FSID
181 	| FATTR4_WORD0_FILEID,
182 	FATTR4_WORD1_MODE
183 	| FATTR4_WORD1_NUMLINKS
184 	| FATTR4_WORD1_OWNER
185 	| FATTR4_WORD1_OWNER_GROUP
186 	| FATTR4_WORD1_RAWDEV
187 	| FATTR4_WORD1_SPACE_USED
188 	| FATTR4_WORD1_TIME_ACCESS
189 	| FATTR4_WORD1_TIME_METADATA
190 	| FATTR4_WORD1_TIME_MODIFY,
191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
192 	FATTR4_WORD2_SECURITY_LABEL
193 #endif
194 };
195 
196 static const u32 nfs4_pnfs_open_bitmap[3] = {
197 	FATTR4_WORD0_TYPE
198 	| FATTR4_WORD0_CHANGE
199 	| FATTR4_WORD0_SIZE
200 	| FATTR4_WORD0_FSID
201 	| FATTR4_WORD0_FILEID,
202 	FATTR4_WORD1_MODE
203 	| FATTR4_WORD1_NUMLINKS
204 	| FATTR4_WORD1_OWNER
205 	| FATTR4_WORD1_OWNER_GROUP
206 	| FATTR4_WORD1_RAWDEV
207 	| FATTR4_WORD1_SPACE_USED
208 	| FATTR4_WORD1_TIME_ACCESS
209 	| FATTR4_WORD1_TIME_METADATA
210 	| FATTR4_WORD1_TIME_MODIFY,
211 	FATTR4_WORD2_MDSTHRESHOLD
212 };
213 
214 static const u32 nfs4_open_noattr_bitmap[3] = {
215 	FATTR4_WORD0_TYPE
216 	| FATTR4_WORD0_CHANGE
217 	| FATTR4_WORD0_FILEID,
218 };
219 
220 const u32 nfs4_statfs_bitmap[3] = {
221 	FATTR4_WORD0_FILES_AVAIL
222 	| FATTR4_WORD0_FILES_FREE
223 	| FATTR4_WORD0_FILES_TOTAL,
224 	FATTR4_WORD1_SPACE_AVAIL
225 	| FATTR4_WORD1_SPACE_FREE
226 	| FATTR4_WORD1_SPACE_TOTAL
227 };
228 
229 const u32 nfs4_pathconf_bitmap[3] = {
230 	FATTR4_WORD0_MAXLINK
231 	| FATTR4_WORD0_MAXNAME,
232 	0
233 };
234 
235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
236 			| FATTR4_WORD0_MAXREAD
237 			| FATTR4_WORD0_MAXWRITE
238 			| FATTR4_WORD0_LEASE_TIME,
239 			FATTR4_WORD1_TIME_DELTA
240 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
241 			FATTR4_WORD2_LAYOUT_BLKSIZE
242 };
243 
244 const u32 nfs4_fs_locations_bitmap[3] = {
245 	FATTR4_WORD0_TYPE
246 	| FATTR4_WORD0_CHANGE
247 	| FATTR4_WORD0_SIZE
248 	| FATTR4_WORD0_FSID
249 	| FATTR4_WORD0_FILEID
250 	| FATTR4_WORD0_FS_LOCATIONS,
251 	FATTR4_WORD1_MODE
252 	| FATTR4_WORD1_NUMLINKS
253 	| FATTR4_WORD1_OWNER
254 	| FATTR4_WORD1_OWNER_GROUP
255 	| FATTR4_WORD1_RAWDEV
256 	| FATTR4_WORD1_SPACE_USED
257 	| FATTR4_WORD1_TIME_ACCESS
258 	| FATTR4_WORD1_TIME_METADATA
259 	| FATTR4_WORD1_TIME_MODIFY
260 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
261 };
262 
263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
264 		struct nfs4_readdir_arg *readdir)
265 {
266 	__be32 *start, *p;
267 
268 	if (cookie > 2) {
269 		readdir->cookie = cookie;
270 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
271 		return;
272 	}
273 
274 	readdir->cookie = 0;
275 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
276 	if (cookie == 2)
277 		return;
278 
279 	/*
280 	 * NFSv4 servers do not return entries for '.' and '..'
281 	 * Therefore, we fake these entries here.  We let '.'
282 	 * have cookie 0 and '..' have cookie 1.  Note that
283 	 * when talking to the server, we always send cookie 0
284 	 * instead of 1 or 2.
285 	 */
286 	start = p = kmap_atomic(*readdir->pages);
287 
288 	if (cookie == 0) {
289 		*p++ = xdr_one;                                  /* next */
290 		*p++ = xdr_zero;                   /* cookie, first word */
291 		*p++ = xdr_one;                   /* cookie, second word */
292 		*p++ = xdr_one;                             /* entry len */
293 		memcpy(p, ".\0\0\0", 4);                        /* entry */
294 		p++;
295 		*p++ = xdr_one;                         /* bitmap length */
296 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
297 		*p++ = htonl(8);              /* attribute buffer length */
298 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
299 	}
300 
301 	*p++ = xdr_one;                                  /* next */
302 	*p++ = xdr_zero;                   /* cookie, first word */
303 	*p++ = xdr_two;                   /* cookie, second word */
304 	*p++ = xdr_two;                             /* entry len */
305 	memcpy(p, "..\0\0", 4);                         /* entry */
306 	p++;
307 	*p++ = xdr_one;                         /* bitmap length */
308 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
309 	*p++ = htonl(8);              /* attribute buffer length */
310 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
311 
312 	readdir->pgbase = (char *)p - (char *)start;
313 	readdir->count -= readdir->pgbase;
314 	kunmap_atomic(start);
315 }
316 
317 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
318 {
319 	int res = 0;
320 
321 	might_sleep();
322 
323 	if (*timeout <= 0)
324 		*timeout = NFS4_POLL_RETRY_MIN;
325 	if (*timeout > NFS4_POLL_RETRY_MAX)
326 		*timeout = NFS4_POLL_RETRY_MAX;
327 	freezable_schedule_timeout_killable_unsafe(*timeout);
328 	if (fatal_signal_pending(current))
329 		res = -ERESTARTSYS;
330 	*timeout <<= 1;
331 	return res;
332 }
333 
334 /* This is the error handling routine for processes that are allowed
335  * to sleep.
336  */
337 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
338 {
339 	struct nfs_client *clp = server->nfs_client;
340 	struct nfs4_state *state = exception->state;
341 	struct inode *inode = exception->inode;
342 	int ret = errorcode;
343 
344 	exception->retry = 0;
345 	switch(errorcode) {
346 		case 0:
347 			return 0;
348 		case -NFS4ERR_OPENMODE:
349 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
350 				nfs4_inode_return_delegation(inode);
351 				exception->retry = 1;
352 				return 0;
353 			}
354 			if (state == NULL)
355 				break;
356 			ret = nfs4_schedule_stateid_recovery(server, state);
357 			if (ret < 0)
358 				break;
359 			goto wait_on_recovery;
360 		case -NFS4ERR_DELEG_REVOKED:
361 		case -NFS4ERR_ADMIN_REVOKED:
362 		case -NFS4ERR_BAD_STATEID:
363 			if (inode != NULL && nfs4_have_delegation(inode, FMODE_READ)) {
364 				nfs_remove_bad_delegation(inode);
365 				exception->retry = 1;
366 				break;
367 			}
368 			if (state == NULL)
369 				break;
370 			ret = nfs4_schedule_stateid_recovery(server, state);
371 			if (ret < 0)
372 				break;
373 			goto wait_on_recovery;
374 		case -NFS4ERR_EXPIRED:
375 			if (state != NULL) {
376 				ret = nfs4_schedule_stateid_recovery(server, state);
377 				if (ret < 0)
378 					break;
379 			}
380 		case -NFS4ERR_STALE_STATEID:
381 		case -NFS4ERR_STALE_CLIENTID:
382 			nfs4_schedule_lease_recovery(clp);
383 			goto wait_on_recovery;
384 #if defined(CONFIG_NFS_V4_1)
385 		case -NFS4ERR_BADSESSION:
386 		case -NFS4ERR_BADSLOT:
387 		case -NFS4ERR_BAD_HIGH_SLOT:
388 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
389 		case -NFS4ERR_DEADSESSION:
390 		case -NFS4ERR_SEQ_FALSE_RETRY:
391 		case -NFS4ERR_SEQ_MISORDERED:
392 			dprintk("%s ERROR: %d Reset session\n", __func__,
393 				errorcode);
394 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
395 			goto wait_on_recovery;
396 #endif /* defined(CONFIG_NFS_V4_1) */
397 		case -NFS4ERR_FILE_OPEN:
398 			if (exception->timeout > HZ) {
399 				/* We have retried a decent amount, time to
400 				 * fail
401 				 */
402 				ret = -EBUSY;
403 				break;
404 			}
405 		case -NFS4ERR_GRACE:
406 		case -NFS4ERR_DELAY:
407 			ret = nfs4_delay(server->client, &exception->timeout);
408 			if (ret != 0)
409 				break;
410 		case -NFS4ERR_RETRY_UNCACHED_REP:
411 		case -NFS4ERR_OLD_STATEID:
412 			exception->retry = 1;
413 			break;
414 		case -NFS4ERR_BADOWNER:
415 			/* The following works around a Linux server bug! */
416 		case -NFS4ERR_BADNAME:
417 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
418 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
419 				exception->retry = 1;
420 				printk(KERN_WARNING "NFS: v4 server %s "
421 						"does not accept raw "
422 						"uid/gids. "
423 						"Reenabling the idmapper.\n",
424 						server->nfs_client->cl_hostname);
425 			}
426 	}
427 	/* We failed to handle the error */
428 	return nfs4_map_errors(ret);
429 wait_on_recovery:
430 	ret = nfs4_wait_clnt_recover(clp);
431 	if (ret == 0)
432 		exception->retry = 1;
433 	return ret;
434 }
435 
436 
437 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
438 {
439 	spin_lock(&clp->cl_lock);
440 	if (time_before(clp->cl_last_renewal,timestamp))
441 		clp->cl_last_renewal = timestamp;
442 	spin_unlock(&clp->cl_lock);
443 }
444 
445 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
446 {
447 	do_renew_lease(server->nfs_client, timestamp);
448 }
449 
450 #if defined(CONFIG_NFS_V4_1)
451 
452 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
453 {
454 	struct nfs4_session *session;
455 	struct nfs4_slot_table *tbl;
456 	bool send_new_highest_used_slotid = false;
457 
458 	if (!res->sr_slot) {
459 		/* just wake up the next guy waiting since
460 		 * we may have not consumed a slot after all */
461 		dprintk("%s: No slot\n", __func__);
462 		return;
463 	}
464 	tbl = res->sr_slot->table;
465 	session = tbl->session;
466 
467 	spin_lock(&tbl->slot_tbl_lock);
468 	/* Be nice to the server: try to ensure that the last transmitted
469 	 * value for highest_user_slotid <= target_highest_slotid
470 	 */
471 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
472 		send_new_highest_used_slotid = true;
473 
474 	if (nfs41_wake_and_assign_slot(tbl, res->sr_slot)) {
475 		send_new_highest_used_slotid = false;
476 		goto out_unlock;
477 	}
478 	nfs4_free_slot(tbl, res->sr_slot);
479 
480 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
481 		send_new_highest_used_slotid = false;
482 out_unlock:
483 	spin_unlock(&tbl->slot_tbl_lock);
484 	res->sr_slot = NULL;
485 	if (send_new_highest_used_slotid)
486 		nfs41_server_notify_highest_slotid_update(session->clp);
487 }
488 
489 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
490 {
491 	struct nfs4_session *session;
492 	struct nfs4_slot *slot;
493 	struct nfs_client *clp;
494 	bool interrupted = false;
495 	int ret = 1;
496 
497 	/* don't increment the sequence number if the task wasn't sent */
498 	if (!RPC_WAS_SENT(task))
499 		goto out;
500 
501 	slot = res->sr_slot;
502 	session = slot->table->session;
503 
504 	if (slot->interrupted) {
505 		slot->interrupted = 0;
506 		interrupted = true;
507 	}
508 
509 	/* Check the SEQUENCE operation status */
510 	switch (res->sr_status) {
511 	case 0:
512 		/* Update the slot's sequence and clientid lease timer */
513 		++slot->seq_nr;
514 		clp = session->clp;
515 		do_renew_lease(clp, res->sr_timestamp);
516 		/* Check sequence flags */
517 		if (res->sr_status_flags != 0)
518 			nfs4_schedule_lease_recovery(clp);
519 		nfs41_update_target_slotid(slot->table, slot, res);
520 		break;
521 	case 1:
522 		/*
523 		 * sr_status remains 1 if an RPC level error occurred.
524 		 * The server may or may not have processed the sequence
525 		 * operation..
526 		 * Mark the slot as having hosted an interrupted RPC call.
527 		 */
528 		slot->interrupted = 1;
529 		goto out;
530 	case -NFS4ERR_DELAY:
531 		/* The server detected a resend of the RPC call and
532 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
533 		 * of RFC5661.
534 		 */
535 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
536 			__func__,
537 			slot->slot_nr,
538 			slot->seq_nr);
539 		goto out_retry;
540 	case -NFS4ERR_BADSLOT:
541 		/*
542 		 * The slot id we used was probably retired. Try again
543 		 * using a different slot id.
544 		 */
545 		goto retry_nowait;
546 	case -NFS4ERR_SEQ_MISORDERED:
547 		/*
548 		 * Was the last operation on this sequence interrupted?
549 		 * If so, retry after bumping the sequence number.
550 		 */
551 		if (interrupted) {
552 			++slot->seq_nr;
553 			goto retry_nowait;
554 		}
555 		/*
556 		 * Could this slot have been previously retired?
557 		 * If so, then the server may be expecting seq_nr = 1!
558 		 */
559 		if (slot->seq_nr != 1) {
560 			slot->seq_nr = 1;
561 			goto retry_nowait;
562 		}
563 		break;
564 	case -NFS4ERR_SEQ_FALSE_RETRY:
565 		++slot->seq_nr;
566 		goto retry_nowait;
567 	default:
568 		/* Just update the slot sequence no. */
569 		++slot->seq_nr;
570 	}
571 out:
572 	/* The session may be reset by one of the error handlers. */
573 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
574 	nfs41_sequence_free_slot(res);
575 	return ret;
576 retry_nowait:
577 	if (rpc_restart_call_prepare(task)) {
578 		task->tk_status = 0;
579 		ret = 0;
580 	}
581 	goto out;
582 out_retry:
583 	if (!rpc_restart_call(task))
584 		goto out;
585 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
586 	return 0;
587 }
588 
589 static int nfs4_sequence_done(struct rpc_task *task,
590 			       struct nfs4_sequence_res *res)
591 {
592 	if (res->sr_slot == NULL)
593 		return 1;
594 	return nfs41_sequence_done(task, res);
595 }
596 
597 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
598 		struct nfs4_sequence_res *res, int cache_reply)
599 {
600 	args->sa_slot = NULL;
601 	args->sa_cache_this = 0;
602 	args->sa_privileged = 0;
603 	if (cache_reply)
604 		args->sa_cache_this = 1;
605 	res->sr_slot = NULL;
606 }
607 
608 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
609 {
610 	args->sa_privileged = 1;
611 }
612 
613 int nfs41_setup_sequence(struct nfs4_session *session,
614 				struct nfs4_sequence_args *args,
615 				struct nfs4_sequence_res *res,
616 				struct rpc_task *task)
617 {
618 	struct nfs4_slot *slot;
619 	struct nfs4_slot_table *tbl;
620 
621 	dprintk("--> %s\n", __func__);
622 	/* slot already allocated? */
623 	if (res->sr_slot != NULL)
624 		goto out_success;
625 
626 	tbl = &session->fc_slot_table;
627 
628 	task->tk_timeout = 0;
629 
630 	spin_lock(&tbl->slot_tbl_lock);
631 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
632 	    !args->sa_privileged) {
633 		/* The state manager will wait until the slot table is empty */
634 		dprintk("%s session is draining\n", __func__);
635 		goto out_sleep;
636 	}
637 
638 	slot = nfs4_alloc_slot(tbl);
639 	if (IS_ERR(slot)) {
640 		/* If out of memory, try again in 1/4 second */
641 		if (slot == ERR_PTR(-ENOMEM))
642 			task->tk_timeout = HZ >> 2;
643 		dprintk("<-- %s: no free slots\n", __func__);
644 		goto out_sleep;
645 	}
646 	spin_unlock(&tbl->slot_tbl_lock);
647 
648 	args->sa_slot = slot;
649 
650 	dprintk("<-- %s slotid=%d seqid=%d\n", __func__,
651 			slot->slot_nr, slot->seq_nr);
652 
653 	res->sr_slot = slot;
654 	res->sr_timestamp = jiffies;
655 	res->sr_status_flags = 0;
656 	/*
657 	 * sr_status is only set in decode_sequence, and so will remain
658 	 * set to 1 if an rpc level failure occurs.
659 	 */
660 	res->sr_status = 1;
661 out_success:
662 	rpc_call_start(task);
663 	return 0;
664 out_sleep:
665 	/* Privileged tasks are queued with top priority */
666 	if (args->sa_privileged)
667 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
668 				NULL, RPC_PRIORITY_PRIVILEGED);
669 	else
670 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
671 	spin_unlock(&tbl->slot_tbl_lock);
672 	return -EAGAIN;
673 }
674 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
675 
676 int nfs4_setup_sequence(const struct nfs_server *server,
677 			struct nfs4_sequence_args *args,
678 			struct nfs4_sequence_res *res,
679 			struct rpc_task *task)
680 {
681 	struct nfs4_session *session = nfs4_get_session(server);
682 	int ret = 0;
683 
684 	if (session == NULL) {
685 		rpc_call_start(task);
686 		goto out;
687 	}
688 
689 	dprintk("--> %s clp %p session %p sr_slot %d\n",
690 		__func__, session->clp, session, res->sr_slot ?
691 			res->sr_slot->slot_nr : -1);
692 
693 	ret = nfs41_setup_sequence(session, args, res, task);
694 out:
695 	dprintk("<-- %s status=%d\n", __func__, ret);
696 	return ret;
697 }
698 
699 struct nfs41_call_sync_data {
700 	const struct nfs_server *seq_server;
701 	struct nfs4_sequence_args *seq_args;
702 	struct nfs4_sequence_res *seq_res;
703 };
704 
705 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
706 {
707 	struct nfs41_call_sync_data *data = calldata;
708 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
709 
710 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
711 
712 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
713 }
714 
715 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
716 {
717 	struct nfs41_call_sync_data *data = calldata;
718 
719 	nfs41_sequence_done(task, data->seq_res);
720 }
721 
722 static const struct rpc_call_ops nfs41_call_sync_ops = {
723 	.rpc_call_prepare = nfs41_call_sync_prepare,
724 	.rpc_call_done = nfs41_call_sync_done,
725 };
726 
727 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
728 				   struct nfs_server *server,
729 				   struct rpc_message *msg,
730 				   struct nfs4_sequence_args *args,
731 				   struct nfs4_sequence_res *res)
732 {
733 	int ret;
734 	struct rpc_task *task;
735 	struct nfs41_call_sync_data data = {
736 		.seq_server = server,
737 		.seq_args = args,
738 		.seq_res = res,
739 	};
740 	struct rpc_task_setup task_setup = {
741 		.rpc_client = clnt,
742 		.rpc_message = msg,
743 		.callback_ops = &nfs41_call_sync_ops,
744 		.callback_data = &data
745 	};
746 
747 	task = rpc_run_task(&task_setup);
748 	if (IS_ERR(task))
749 		ret = PTR_ERR(task);
750 	else {
751 		ret = task->tk_status;
752 		rpc_put_task(task);
753 	}
754 	return ret;
755 }
756 
757 #else
758 static
759 void nfs41_init_sequence(struct nfs4_sequence_args *args,
760 		struct nfs4_sequence_res *res, int cache_reply)
761 {
762 }
763 
764 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
765 {
766 }
767 
768 
769 static int nfs4_sequence_done(struct rpc_task *task,
770 			       struct nfs4_sequence_res *res)
771 {
772 	return 1;
773 }
774 #endif /* CONFIG_NFS_V4_1 */
775 
776 static
777 int _nfs4_call_sync(struct rpc_clnt *clnt,
778 		    struct nfs_server *server,
779 		    struct rpc_message *msg,
780 		    struct nfs4_sequence_args *args,
781 		    struct nfs4_sequence_res *res)
782 {
783 	return rpc_call_sync(clnt, msg, 0);
784 }
785 
786 static
787 int nfs4_call_sync(struct rpc_clnt *clnt,
788 		   struct nfs_server *server,
789 		   struct rpc_message *msg,
790 		   struct nfs4_sequence_args *args,
791 		   struct nfs4_sequence_res *res,
792 		   int cache_reply)
793 {
794 	nfs41_init_sequence(args, res, cache_reply);
795 	return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
796 						args, res);
797 }
798 
799 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
800 {
801 	struct nfs_inode *nfsi = NFS_I(dir);
802 
803 	spin_lock(&dir->i_lock);
804 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
805 	if (!cinfo->atomic || cinfo->before != dir->i_version)
806 		nfs_force_lookup_revalidate(dir);
807 	dir->i_version = cinfo->after;
808 	nfs_fscache_invalidate(dir);
809 	spin_unlock(&dir->i_lock);
810 }
811 
812 struct nfs4_opendata {
813 	struct kref kref;
814 	struct nfs_openargs o_arg;
815 	struct nfs_openres o_res;
816 	struct nfs_open_confirmargs c_arg;
817 	struct nfs_open_confirmres c_res;
818 	struct nfs4_string owner_name;
819 	struct nfs4_string group_name;
820 	struct nfs_fattr f_attr;
821 	struct nfs4_label *f_label;
822 	struct dentry *dir;
823 	struct dentry *dentry;
824 	struct nfs4_state_owner *owner;
825 	struct nfs4_state *state;
826 	struct iattr attrs;
827 	unsigned long timestamp;
828 	unsigned int rpc_done : 1;
829 	unsigned int is_recover : 1;
830 	int rpc_status;
831 	int cancelled;
832 };
833 
834 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
835 		int err, struct nfs4_exception *exception)
836 {
837 	if (err != -EINVAL)
838 		return false;
839 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
840 		return false;
841 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
842 	exception->retry = 1;
843 	return true;
844 }
845 
846 static enum open_claim_type4
847 nfs4_map_atomic_open_claim(struct nfs_server *server,
848 		enum open_claim_type4 claim)
849 {
850 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
851 		return claim;
852 	switch (claim) {
853 	default:
854 		return claim;
855 	case NFS4_OPEN_CLAIM_FH:
856 		return NFS4_OPEN_CLAIM_NULL;
857 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
858 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
859 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
860 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
861 	}
862 }
863 
864 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
865 {
866 	p->o_res.f_attr = &p->f_attr;
867 	p->o_res.f_label = p->f_label;
868 	p->o_res.seqid = p->o_arg.seqid;
869 	p->c_res.seqid = p->c_arg.seqid;
870 	p->o_res.server = p->o_arg.server;
871 	p->o_res.access_request = p->o_arg.access;
872 	nfs_fattr_init(&p->f_attr);
873 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
874 }
875 
876 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
877 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
878 		const struct iattr *attrs,
879 		struct nfs4_label *label,
880 		enum open_claim_type4 claim,
881 		gfp_t gfp_mask)
882 {
883 	struct dentry *parent = dget_parent(dentry);
884 	struct inode *dir = parent->d_inode;
885 	struct nfs_server *server = NFS_SERVER(dir);
886 	struct nfs4_opendata *p;
887 
888 	p = kzalloc(sizeof(*p), gfp_mask);
889 	if (p == NULL)
890 		goto err;
891 
892 	p->f_label = nfs4_label_alloc(server, gfp_mask);
893 	if (IS_ERR(p->f_label))
894 		goto err_free_p;
895 
896 	p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
897 	if (p->o_arg.seqid == NULL)
898 		goto err_free_label;
899 	nfs_sb_active(dentry->d_sb);
900 	p->dentry = dget(dentry);
901 	p->dir = parent;
902 	p->owner = sp;
903 	atomic_inc(&sp->so_count);
904 	p->o_arg.open_flags = flags;
905 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
906 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
907 	 * will return permission denied for all bits until close */
908 	if (!(flags & O_EXCL)) {
909 		/* ask server to check for all possible rights as results
910 		 * are cached */
911 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
912 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
913 	}
914 	p->o_arg.clientid = server->nfs_client->cl_clientid;
915 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
916 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
917 	p->o_arg.name = &dentry->d_name;
918 	p->o_arg.server = server;
919 	p->o_arg.bitmask = nfs4_bitmask(server, label);
920 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
921 	p->o_arg.label = label;
922 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
923 	switch (p->o_arg.claim) {
924 	case NFS4_OPEN_CLAIM_NULL:
925 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
926 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
927 		p->o_arg.fh = NFS_FH(dir);
928 		break;
929 	case NFS4_OPEN_CLAIM_PREVIOUS:
930 	case NFS4_OPEN_CLAIM_FH:
931 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
932 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
933 		p->o_arg.fh = NFS_FH(dentry->d_inode);
934 	}
935 	if (attrs != NULL && attrs->ia_valid != 0) {
936 		__be32 verf[2];
937 
938 		p->o_arg.u.attrs = &p->attrs;
939 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
940 
941 		verf[0] = jiffies;
942 		verf[1] = current->pid;
943 		memcpy(p->o_arg.u.verifier.data, verf,
944 				sizeof(p->o_arg.u.verifier.data));
945 	}
946 	p->c_arg.fh = &p->o_res.fh;
947 	p->c_arg.stateid = &p->o_res.stateid;
948 	p->c_arg.seqid = p->o_arg.seqid;
949 	nfs4_init_opendata_res(p);
950 	kref_init(&p->kref);
951 	return p;
952 
953 err_free_label:
954 	nfs4_label_free(p->f_label);
955 err_free_p:
956 	kfree(p);
957 err:
958 	dput(parent);
959 	return NULL;
960 }
961 
962 static void nfs4_opendata_free(struct kref *kref)
963 {
964 	struct nfs4_opendata *p = container_of(kref,
965 			struct nfs4_opendata, kref);
966 	struct super_block *sb = p->dentry->d_sb;
967 
968 	nfs_free_seqid(p->o_arg.seqid);
969 	if (p->state != NULL)
970 		nfs4_put_open_state(p->state);
971 	nfs4_put_state_owner(p->owner);
972 
973 	nfs4_label_free(p->f_label);
974 
975 	dput(p->dir);
976 	dput(p->dentry);
977 	nfs_sb_deactive(sb);
978 	nfs_fattr_free_names(&p->f_attr);
979 	kfree(p);
980 }
981 
982 static void nfs4_opendata_put(struct nfs4_opendata *p)
983 {
984 	if (p != NULL)
985 		kref_put(&p->kref, nfs4_opendata_free);
986 }
987 
988 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
989 {
990 	int ret;
991 
992 	ret = rpc_wait_for_completion_task(task);
993 	return ret;
994 }
995 
996 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
997 {
998 	int ret = 0;
999 
1000 	if (open_mode & (O_EXCL|O_TRUNC))
1001 		goto out;
1002 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1003 		case FMODE_READ:
1004 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1005 				&& state->n_rdonly != 0;
1006 			break;
1007 		case FMODE_WRITE:
1008 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1009 				&& state->n_wronly != 0;
1010 			break;
1011 		case FMODE_READ|FMODE_WRITE:
1012 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1013 				&& state->n_rdwr != 0;
1014 	}
1015 out:
1016 	return ret;
1017 }
1018 
1019 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1020 {
1021 	if (delegation == NULL)
1022 		return 0;
1023 	if ((delegation->type & fmode) != fmode)
1024 		return 0;
1025 	if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1026 		return 0;
1027 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1028 		return 0;
1029 	nfs_mark_delegation_referenced(delegation);
1030 	return 1;
1031 }
1032 
1033 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1034 {
1035 	switch (fmode) {
1036 		case FMODE_WRITE:
1037 			state->n_wronly++;
1038 			break;
1039 		case FMODE_READ:
1040 			state->n_rdonly++;
1041 			break;
1042 		case FMODE_READ|FMODE_WRITE:
1043 			state->n_rdwr++;
1044 	}
1045 	nfs4_state_set_mode_locked(state, state->state | fmode);
1046 }
1047 
1048 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1049 {
1050 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1051 		nfs4_stateid_copy(&state->stateid, stateid);
1052 	nfs4_stateid_copy(&state->open_stateid, stateid);
1053 	set_bit(NFS_OPEN_STATE, &state->flags);
1054 	switch (fmode) {
1055 		case FMODE_READ:
1056 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1057 			break;
1058 		case FMODE_WRITE:
1059 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1060 			break;
1061 		case FMODE_READ|FMODE_WRITE:
1062 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1063 	}
1064 }
1065 
1066 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1067 {
1068 	write_seqlock(&state->seqlock);
1069 	nfs_set_open_stateid_locked(state, stateid, fmode);
1070 	write_sequnlock(&state->seqlock);
1071 }
1072 
1073 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1074 {
1075 	/*
1076 	 * Protect the call to nfs4_state_set_mode_locked and
1077 	 * serialise the stateid update
1078 	 */
1079 	write_seqlock(&state->seqlock);
1080 	if (deleg_stateid != NULL) {
1081 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1082 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1083 	}
1084 	if (open_stateid != NULL)
1085 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1086 	write_sequnlock(&state->seqlock);
1087 	spin_lock(&state->owner->so_lock);
1088 	update_open_stateflags(state, fmode);
1089 	spin_unlock(&state->owner->so_lock);
1090 }
1091 
1092 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1093 {
1094 	struct nfs_inode *nfsi = NFS_I(state->inode);
1095 	struct nfs_delegation *deleg_cur;
1096 	int ret = 0;
1097 
1098 	fmode &= (FMODE_READ|FMODE_WRITE);
1099 
1100 	rcu_read_lock();
1101 	deleg_cur = rcu_dereference(nfsi->delegation);
1102 	if (deleg_cur == NULL)
1103 		goto no_delegation;
1104 
1105 	spin_lock(&deleg_cur->lock);
1106 	if (nfsi->delegation != deleg_cur ||
1107 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1108 	    (deleg_cur->type & fmode) != fmode)
1109 		goto no_delegation_unlock;
1110 
1111 	if (delegation == NULL)
1112 		delegation = &deleg_cur->stateid;
1113 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1114 		goto no_delegation_unlock;
1115 
1116 	nfs_mark_delegation_referenced(deleg_cur);
1117 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1118 	ret = 1;
1119 no_delegation_unlock:
1120 	spin_unlock(&deleg_cur->lock);
1121 no_delegation:
1122 	rcu_read_unlock();
1123 
1124 	if (!ret && open_stateid != NULL) {
1125 		__update_open_stateid(state, open_stateid, NULL, fmode);
1126 		ret = 1;
1127 	}
1128 
1129 	return ret;
1130 }
1131 
1132 
1133 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1134 {
1135 	struct nfs_delegation *delegation;
1136 
1137 	rcu_read_lock();
1138 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1139 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1140 		rcu_read_unlock();
1141 		return;
1142 	}
1143 	rcu_read_unlock();
1144 	nfs4_inode_return_delegation(inode);
1145 }
1146 
1147 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1148 {
1149 	struct nfs4_state *state = opendata->state;
1150 	struct nfs_inode *nfsi = NFS_I(state->inode);
1151 	struct nfs_delegation *delegation;
1152 	int open_mode = opendata->o_arg.open_flags;
1153 	fmode_t fmode = opendata->o_arg.fmode;
1154 	nfs4_stateid stateid;
1155 	int ret = -EAGAIN;
1156 
1157 	for (;;) {
1158 		if (can_open_cached(state, fmode, open_mode)) {
1159 			spin_lock(&state->owner->so_lock);
1160 			if (can_open_cached(state, fmode, open_mode)) {
1161 				update_open_stateflags(state, fmode);
1162 				spin_unlock(&state->owner->so_lock);
1163 				goto out_return_state;
1164 			}
1165 			spin_unlock(&state->owner->so_lock);
1166 		}
1167 		rcu_read_lock();
1168 		delegation = rcu_dereference(nfsi->delegation);
1169 		if (!can_open_delegated(delegation, fmode)) {
1170 			rcu_read_unlock();
1171 			break;
1172 		}
1173 		/* Save the delegation */
1174 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1175 		rcu_read_unlock();
1176 		nfs_release_seqid(opendata->o_arg.seqid);
1177 		if (!opendata->is_recover) {
1178 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1179 			if (ret != 0)
1180 				goto out;
1181 		}
1182 		ret = -EAGAIN;
1183 
1184 		/* Try to update the stateid using the delegation */
1185 		if (update_open_stateid(state, NULL, &stateid, fmode))
1186 			goto out_return_state;
1187 	}
1188 out:
1189 	return ERR_PTR(ret);
1190 out_return_state:
1191 	atomic_inc(&state->count);
1192 	return state;
1193 }
1194 
1195 static void
1196 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1197 {
1198 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1199 	struct nfs_delegation *delegation;
1200 	int delegation_flags = 0;
1201 
1202 	rcu_read_lock();
1203 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1204 	if (delegation)
1205 		delegation_flags = delegation->flags;
1206 	rcu_read_unlock();
1207 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1208 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1209 				   "returning a delegation for "
1210 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1211 				   clp->cl_hostname);
1212 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1213 		nfs_inode_set_delegation(state->inode,
1214 					 data->owner->so_cred,
1215 					 &data->o_res);
1216 	else
1217 		nfs_inode_reclaim_delegation(state->inode,
1218 					     data->owner->so_cred,
1219 					     &data->o_res);
1220 }
1221 
1222 /*
1223  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1224  * and update the nfs4_state.
1225  */
1226 static struct nfs4_state *
1227 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1228 {
1229 	struct inode *inode = data->state->inode;
1230 	struct nfs4_state *state = data->state;
1231 	int ret;
1232 
1233 	if (!data->rpc_done) {
1234 		ret = data->rpc_status;
1235 		goto err;
1236 	}
1237 
1238 	ret = -ESTALE;
1239 	if (!(data->f_attr.valid & NFS_ATTR_FATTR_TYPE) ||
1240 	    !(data->f_attr.valid & NFS_ATTR_FATTR_FILEID) ||
1241 	    !(data->f_attr.valid & NFS_ATTR_FATTR_CHANGE))
1242 		goto err;
1243 
1244 	ret = -ENOMEM;
1245 	state = nfs4_get_open_state(inode, data->owner);
1246 	if (state == NULL)
1247 		goto err;
1248 
1249 	ret = nfs_refresh_inode(inode, &data->f_attr);
1250 	if (ret)
1251 		goto err;
1252 
1253 	nfs_setsecurity(inode, &data->f_attr, data->f_label);
1254 
1255 	if (data->o_res.delegation_type != 0)
1256 		nfs4_opendata_check_deleg(data, state);
1257 	update_open_stateid(state, &data->o_res.stateid, NULL,
1258 			    data->o_arg.fmode);
1259 
1260 	return state;
1261 err:
1262 	return ERR_PTR(ret);
1263 
1264 }
1265 
1266 static struct nfs4_state *
1267 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1268 {
1269 	struct inode *inode;
1270 	struct nfs4_state *state = NULL;
1271 	int ret;
1272 
1273 	if (!data->rpc_done) {
1274 		state = nfs4_try_open_cached(data);
1275 		goto out;
1276 	}
1277 
1278 	ret = -EAGAIN;
1279 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1280 		goto err;
1281 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1282 	ret = PTR_ERR(inode);
1283 	if (IS_ERR(inode))
1284 		goto err;
1285 	ret = -ENOMEM;
1286 	state = nfs4_get_open_state(inode, data->owner);
1287 	if (state == NULL)
1288 		goto err_put_inode;
1289 	if (data->o_res.delegation_type != 0)
1290 		nfs4_opendata_check_deleg(data, state);
1291 	update_open_stateid(state, &data->o_res.stateid, NULL,
1292 			data->o_arg.fmode);
1293 	iput(inode);
1294 out:
1295 	nfs_release_seqid(data->o_arg.seqid);
1296 	return state;
1297 err_put_inode:
1298 	iput(inode);
1299 err:
1300 	return ERR_PTR(ret);
1301 }
1302 
1303 static struct nfs4_state *
1304 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1305 {
1306 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1307 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1308 	return _nfs4_opendata_to_nfs4_state(data);
1309 }
1310 
1311 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1312 {
1313 	struct nfs_inode *nfsi = NFS_I(state->inode);
1314 	struct nfs_open_context *ctx;
1315 
1316 	spin_lock(&state->inode->i_lock);
1317 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1318 		if (ctx->state != state)
1319 			continue;
1320 		get_nfs_open_context(ctx);
1321 		spin_unlock(&state->inode->i_lock);
1322 		return ctx;
1323 	}
1324 	spin_unlock(&state->inode->i_lock);
1325 	return ERR_PTR(-ENOENT);
1326 }
1327 
1328 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1329 		struct nfs4_state *state, enum open_claim_type4 claim)
1330 {
1331 	struct nfs4_opendata *opendata;
1332 
1333 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1334 			NULL, NULL, claim, GFP_NOFS);
1335 	if (opendata == NULL)
1336 		return ERR_PTR(-ENOMEM);
1337 	opendata->state = state;
1338 	atomic_inc(&state->count);
1339 	return opendata;
1340 }
1341 
1342 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1343 {
1344 	struct nfs4_state *newstate;
1345 	int ret;
1346 
1347 	opendata->o_arg.open_flags = 0;
1348 	opendata->o_arg.fmode = fmode;
1349 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1350 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1351 	nfs4_init_opendata_res(opendata);
1352 	ret = _nfs4_recover_proc_open(opendata);
1353 	if (ret != 0)
1354 		return ret;
1355 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1356 	if (IS_ERR(newstate))
1357 		return PTR_ERR(newstate);
1358 	nfs4_close_state(newstate, fmode);
1359 	*res = newstate;
1360 	return 0;
1361 }
1362 
1363 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1364 {
1365 	struct nfs4_state *newstate;
1366 	int ret;
1367 
1368 	/* memory barrier prior to reading state->n_* */
1369 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1370 	clear_bit(NFS_OPEN_STATE, &state->flags);
1371 	smp_rmb();
1372 	if (state->n_rdwr != 0) {
1373 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
1374 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1375 		if (ret != 0)
1376 			return ret;
1377 		if (newstate != state)
1378 			return -ESTALE;
1379 	}
1380 	if (state->n_wronly != 0) {
1381 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1382 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1383 		if (ret != 0)
1384 			return ret;
1385 		if (newstate != state)
1386 			return -ESTALE;
1387 	}
1388 	if (state->n_rdonly != 0) {
1389 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1390 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1391 		if (ret != 0)
1392 			return ret;
1393 		if (newstate != state)
1394 			return -ESTALE;
1395 	}
1396 	/*
1397 	 * We may have performed cached opens for all three recoveries.
1398 	 * Check if we need to update the current stateid.
1399 	 */
1400 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1401 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1402 		write_seqlock(&state->seqlock);
1403 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1404 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1405 		write_sequnlock(&state->seqlock);
1406 	}
1407 	return 0;
1408 }
1409 
1410 /*
1411  * OPEN_RECLAIM:
1412  * 	reclaim state on the server after a reboot.
1413  */
1414 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1415 {
1416 	struct nfs_delegation *delegation;
1417 	struct nfs4_opendata *opendata;
1418 	fmode_t delegation_type = 0;
1419 	int status;
1420 
1421 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1422 			NFS4_OPEN_CLAIM_PREVIOUS);
1423 	if (IS_ERR(opendata))
1424 		return PTR_ERR(opendata);
1425 	rcu_read_lock();
1426 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1427 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1428 		delegation_type = delegation->type;
1429 	rcu_read_unlock();
1430 	opendata->o_arg.u.delegation_type = delegation_type;
1431 	status = nfs4_open_recover(opendata, state);
1432 	nfs4_opendata_put(opendata);
1433 	return status;
1434 }
1435 
1436 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1437 {
1438 	struct nfs_server *server = NFS_SERVER(state->inode);
1439 	struct nfs4_exception exception = { };
1440 	int err;
1441 	do {
1442 		err = _nfs4_do_open_reclaim(ctx, state);
1443 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1444 			continue;
1445 		if (err != -NFS4ERR_DELAY)
1446 			break;
1447 		nfs4_handle_exception(server, err, &exception);
1448 	} while (exception.retry);
1449 	return err;
1450 }
1451 
1452 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1453 {
1454 	struct nfs_open_context *ctx;
1455 	int ret;
1456 
1457 	ctx = nfs4_state_find_open_context(state);
1458 	if (IS_ERR(ctx))
1459 		return -EAGAIN;
1460 	ret = nfs4_do_open_reclaim(ctx, state);
1461 	put_nfs_open_context(ctx);
1462 	return ret;
1463 }
1464 
1465 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1466 {
1467 	switch (err) {
1468 		default:
1469 			printk(KERN_ERR "NFS: %s: unhandled error "
1470 					"%d.\n", __func__, err);
1471 		case 0:
1472 		case -ENOENT:
1473 		case -ESTALE:
1474 			break;
1475 		case -NFS4ERR_BADSESSION:
1476 		case -NFS4ERR_BADSLOT:
1477 		case -NFS4ERR_BAD_HIGH_SLOT:
1478 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1479 		case -NFS4ERR_DEADSESSION:
1480 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1481 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1482 			return -EAGAIN;
1483 		case -NFS4ERR_STALE_CLIENTID:
1484 		case -NFS4ERR_STALE_STATEID:
1485 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1486 		case -NFS4ERR_EXPIRED:
1487 			/* Don't recall a delegation if it was lost */
1488 			nfs4_schedule_lease_recovery(server->nfs_client);
1489 			return -EAGAIN;
1490 		case -NFS4ERR_DELEG_REVOKED:
1491 		case -NFS4ERR_ADMIN_REVOKED:
1492 		case -NFS4ERR_BAD_STATEID:
1493 		case -NFS4ERR_OPENMODE:
1494 			nfs_inode_find_state_and_recover(state->inode,
1495 					stateid);
1496 			nfs4_schedule_stateid_recovery(server, state);
1497 			return 0;
1498 		case -NFS4ERR_DELAY:
1499 		case -NFS4ERR_GRACE:
1500 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1501 			ssleep(1);
1502 			return -EAGAIN;
1503 		case -ENOMEM:
1504 		case -NFS4ERR_DENIED:
1505 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1506 			return 0;
1507 	}
1508 	return err;
1509 }
1510 
1511 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1512 {
1513 	struct nfs_server *server = NFS_SERVER(state->inode);
1514 	struct nfs4_opendata *opendata;
1515 	int err;
1516 
1517 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1518 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1519 	if (IS_ERR(opendata))
1520 		return PTR_ERR(opendata);
1521 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1522 	err = nfs4_open_recover(opendata, state);
1523 	nfs4_opendata_put(opendata);
1524 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1525 }
1526 
1527 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1528 {
1529 	struct nfs4_opendata *data = calldata;
1530 
1531 	data->rpc_status = task->tk_status;
1532 	if (data->rpc_status == 0) {
1533 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1534 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1535 		renew_lease(data->o_res.server, data->timestamp);
1536 		data->rpc_done = 1;
1537 	}
1538 }
1539 
1540 static void nfs4_open_confirm_release(void *calldata)
1541 {
1542 	struct nfs4_opendata *data = calldata;
1543 	struct nfs4_state *state = NULL;
1544 
1545 	/* If this request hasn't been cancelled, do nothing */
1546 	if (data->cancelled == 0)
1547 		goto out_free;
1548 	/* In case of error, no cleanup! */
1549 	if (!data->rpc_done)
1550 		goto out_free;
1551 	state = nfs4_opendata_to_nfs4_state(data);
1552 	if (!IS_ERR(state))
1553 		nfs4_close_state(state, data->o_arg.fmode);
1554 out_free:
1555 	nfs4_opendata_put(data);
1556 }
1557 
1558 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1559 	.rpc_call_done = nfs4_open_confirm_done,
1560 	.rpc_release = nfs4_open_confirm_release,
1561 };
1562 
1563 /*
1564  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1565  */
1566 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1567 {
1568 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1569 	struct rpc_task *task;
1570 	struct  rpc_message msg = {
1571 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1572 		.rpc_argp = &data->c_arg,
1573 		.rpc_resp = &data->c_res,
1574 		.rpc_cred = data->owner->so_cred,
1575 	};
1576 	struct rpc_task_setup task_setup_data = {
1577 		.rpc_client = server->client,
1578 		.rpc_message = &msg,
1579 		.callback_ops = &nfs4_open_confirm_ops,
1580 		.callback_data = data,
1581 		.workqueue = nfsiod_workqueue,
1582 		.flags = RPC_TASK_ASYNC,
1583 	};
1584 	int status;
1585 
1586 	kref_get(&data->kref);
1587 	data->rpc_done = 0;
1588 	data->rpc_status = 0;
1589 	data->timestamp = jiffies;
1590 	task = rpc_run_task(&task_setup_data);
1591 	if (IS_ERR(task))
1592 		return PTR_ERR(task);
1593 	status = nfs4_wait_for_completion_rpc_task(task);
1594 	if (status != 0) {
1595 		data->cancelled = 1;
1596 		smp_wmb();
1597 	} else
1598 		status = data->rpc_status;
1599 	rpc_put_task(task);
1600 	return status;
1601 }
1602 
1603 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1604 {
1605 	struct nfs4_opendata *data = calldata;
1606 	struct nfs4_state_owner *sp = data->owner;
1607 	struct nfs_client *clp = sp->so_server->nfs_client;
1608 
1609 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1610 		goto out_wait;
1611 	/*
1612 	 * Check if we still need to send an OPEN call, or if we can use
1613 	 * a delegation instead.
1614 	 */
1615 	if (data->state != NULL) {
1616 		struct nfs_delegation *delegation;
1617 
1618 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1619 			goto out_no_action;
1620 		rcu_read_lock();
1621 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1622 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1623 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1624 		    can_open_delegated(delegation, data->o_arg.fmode))
1625 			goto unlock_no_action;
1626 		rcu_read_unlock();
1627 	}
1628 	/* Update client id. */
1629 	data->o_arg.clientid = clp->cl_clientid;
1630 	switch (data->o_arg.claim) {
1631 	case NFS4_OPEN_CLAIM_PREVIOUS:
1632 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1633 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1634 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1635 	case NFS4_OPEN_CLAIM_FH:
1636 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1637 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1638 	}
1639 	data->timestamp = jiffies;
1640 	if (nfs4_setup_sequence(data->o_arg.server,
1641 				&data->o_arg.seq_args,
1642 				&data->o_res.seq_res,
1643 				task) != 0)
1644 		nfs_release_seqid(data->o_arg.seqid);
1645 
1646 	/* Set the create mode (note dependency on the session type) */
1647 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1648 	if (data->o_arg.open_flags & O_EXCL) {
1649 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1650 		if (nfs4_has_persistent_session(clp))
1651 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1652 		else if (clp->cl_mvops->minor_version > 0)
1653 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1654 	}
1655 	return;
1656 unlock_no_action:
1657 	rcu_read_unlock();
1658 out_no_action:
1659 	task->tk_action = NULL;
1660 out_wait:
1661 	nfs4_sequence_done(task, &data->o_res.seq_res);
1662 }
1663 
1664 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1665 {
1666 	struct nfs4_opendata *data = calldata;
1667 
1668 	data->rpc_status = task->tk_status;
1669 
1670 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1671 		return;
1672 
1673 	if (task->tk_status == 0) {
1674 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1675 			switch (data->o_res.f_attr->mode & S_IFMT) {
1676 			case S_IFREG:
1677 				break;
1678 			case S_IFLNK:
1679 				data->rpc_status = -ELOOP;
1680 				break;
1681 			case S_IFDIR:
1682 				data->rpc_status = -EISDIR;
1683 				break;
1684 			default:
1685 				data->rpc_status = -ENOTDIR;
1686 			}
1687 		}
1688 		renew_lease(data->o_res.server, data->timestamp);
1689 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1690 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1691 	}
1692 	data->rpc_done = 1;
1693 }
1694 
1695 static void nfs4_open_release(void *calldata)
1696 {
1697 	struct nfs4_opendata *data = calldata;
1698 	struct nfs4_state *state = NULL;
1699 
1700 	/* If this request hasn't been cancelled, do nothing */
1701 	if (data->cancelled == 0)
1702 		goto out_free;
1703 	/* In case of error, no cleanup! */
1704 	if (data->rpc_status != 0 || !data->rpc_done)
1705 		goto out_free;
1706 	/* In case we need an open_confirm, no cleanup! */
1707 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1708 		goto out_free;
1709 	state = nfs4_opendata_to_nfs4_state(data);
1710 	if (!IS_ERR(state))
1711 		nfs4_close_state(state, data->o_arg.fmode);
1712 out_free:
1713 	nfs4_opendata_put(data);
1714 }
1715 
1716 static const struct rpc_call_ops nfs4_open_ops = {
1717 	.rpc_call_prepare = nfs4_open_prepare,
1718 	.rpc_call_done = nfs4_open_done,
1719 	.rpc_release = nfs4_open_release,
1720 };
1721 
1722 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1723 {
1724 	struct inode *dir = data->dir->d_inode;
1725 	struct nfs_server *server = NFS_SERVER(dir);
1726 	struct nfs_openargs *o_arg = &data->o_arg;
1727 	struct nfs_openres *o_res = &data->o_res;
1728 	struct rpc_task *task;
1729 	struct rpc_message msg = {
1730 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1731 		.rpc_argp = o_arg,
1732 		.rpc_resp = o_res,
1733 		.rpc_cred = data->owner->so_cred,
1734 	};
1735 	struct rpc_task_setup task_setup_data = {
1736 		.rpc_client = server->client,
1737 		.rpc_message = &msg,
1738 		.callback_ops = &nfs4_open_ops,
1739 		.callback_data = data,
1740 		.workqueue = nfsiod_workqueue,
1741 		.flags = RPC_TASK_ASYNC,
1742 	};
1743 	int status;
1744 
1745 	nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1746 	kref_get(&data->kref);
1747 	data->rpc_done = 0;
1748 	data->rpc_status = 0;
1749 	data->cancelled = 0;
1750 	data->is_recover = 0;
1751 	if (isrecover) {
1752 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1753 		data->is_recover = 1;
1754 	}
1755 	task = rpc_run_task(&task_setup_data);
1756         if (IS_ERR(task))
1757                 return PTR_ERR(task);
1758         status = nfs4_wait_for_completion_rpc_task(task);
1759         if (status != 0) {
1760                 data->cancelled = 1;
1761                 smp_wmb();
1762         } else
1763                 status = data->rpc_status;
1764         rpc_put_task(task);
1765 
1766 	return status;
1767 }
1768 
1769 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1770 {
1771 	struct inode *dir = data->dir->d_inode;
1772 	struct nfs_openres *o_res = &data->o_res;
1773         int status;
1774 
1775 	status = nfs4_run_open_task(data, 1);
1776 	if (status != 0 || !data->rpc_done)
1777 		return status;
1778 
1779 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1780 
1781 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1782 		status = _nfs4_proc_open_confirm(data);
1783 		if (status != 0)
1784 			return status;
1785 	}
1786 
1787 	return status;
1788 }
1789 
1790 static int nfs4_opendata_access(struct rpc_cred *cred,
1791 				struct nfs4_opendata *opendata,
1792 				struct nfs4_state *state, fmode_t fmode,
1793 				int openflags)
1794 {
1795 	struct nfs_access_entry cache;
1796 	u32 mask;
1797 
1798 	/* access call failed or for some reason the server doesn't
1799 	 * support any access modes -- defer access call until later */
1800 	if (opendata->o_res.access_supported == 0)
1801 		return 0;
1802 
1803 	mask = 0;
1804 	/* don't check MAY_WRITE - a newly created file may not have
1805 	 * write mode bits, but POSIX allows the creating process to write.
1806 	 * use openflags to check for exec, because fmode won't
1807 	 * always have FMODE_EXEC set when file open for exec. */
1808 	if (openflags & __FMODE_EXEC) {
1809 		/* ONLY check for exec rights */
1810 		mask = MAY_EXEC;
1811 	} else if (fmode & FMODE_READ)
1812 		mask = MAY_READ;
1813 
1814 	cache.cred = cred;
1815 	cache.jiffies = jiffies;
1816 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
1817 	nfs_access_add_cache(state->inode, &cache);
1818 
1819 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
1820 		return 0;
1821 
1822 	/* even though OPEN succeeded, access is denied. Close the file */
1823 	nfs4_close_state(state, fmode);
1824 	return -EACCES;
1825 }
1826 
1827 /*
1828  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1829  */
1830 static int _nfs4_proc_open(struct nfs4_opendata *data)
1831 {
1832 	struct inode *dir = data->dir->d_inode;
1833 	struct nfs_server *server = NFS_SERVER(dir);
1834 	struct nfs_openargs *o_arg = &data->o_arg;
1835 	struct nfs_openres *o_res = &data->o_res;
1836 	int status;
1837 
1838 	status = nfs4_run_open_task(data, 0);
1839 	if (!data->rpc_done)
1840 		return status;
1841 	if (status != 0) {
1842 		if (status == -NFS4ERR_BADNAME &&
1843 				!(o_arg->open_flags & O_CREAT))
1844 			return -ENOENT;
1845 		return status;
1846 	}
1847 
1848 	nfs_fattr_map_and_free_names(server, &data->f_attr);
1849 
1850 	if (o_arg->open_flags & O_CREAT)
1851 		update_changeattr(dir, &o_res->cinfo);
1852 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1853 		server->caps &= ~NFS_CAP_POSIX_LOCK;
1854 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1855 		status = _nfs4_proc_open_confirm(data);
1856 		if (status != 0)
1857 			return status;
1858 	}
1859 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1860 		_nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
1861 	return 0;
1862 }
1863 
1864 static int nfs4_recover_expired_lease(struct nfs_server *server)
1865 {
1866 	return nfs4_client_recover_expired_lease(server->nfs_client);
1867 }
1868 
1869 /*
1870  * OPEN_EXPIRED:
1871  * 	reclaim state on the server after a network partition.
1872  * 	Assumes caller holds the appropriate lock
1873  */
1874 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1875 {
1876 	struct nfs4_opendata *opendata;
1877 	int ret;
1878 
1879 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1880 			NFS4_OPEN_CLAIM_FH);
1881 	if (IS_ERR(opendata))
1882 		return PTR_ERR(opendata);
1883 	ret = nfs4_open_recover(opendata, state);
1884 	if (ret == -ESTALE)
1885 		d_drop(ctx->dentry);
1886 	nfs4_opendata_put(opendata);
1887 	return ret;
1888 }
1889 
1890 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1891 {
1892 	struct nfs_server *server = NFS_SERVER(state->inode);
1893 	struct nfs4_exception exception = { };
1894 	int err;
1895 
1896 	do {
1897 		err = _nfs4_open_expired(ctx, state);
1898 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1899 			continue;
1900 		switch (err) {
1901 		default:
1902 			goto out;
1903 		case -NFS4ERR_GRACE:
1904 		case -NFS4ERR_DELAY:
1905 			nfs4_handle_exception(server, err, &exception);
1906 			err = 0;
1907 		}
1908 	} while (exception.retry);
1909 out:
1910 	return err;
1911 }
1912 
1913 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1914 {
1915 	struct nfs_open_context *ctx;
1916 	int ret;
1917 
1918 	ctx = nfs4_state_find_open_context(state);
1919 	if (IS_ERR(ctx))
1920 		return -EAGAIN;
1921 	ret = nfs4_do_open_expired(ctx, state);
1922 	put_nfs_open_context(ctx);
1923 	return ret;
1924 }
1925 
1926 #if defined(CONFIG_NFS_V4_1)
1927 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1928 {
1929 	struct nfs_server *server = NFS_SERVER(state->inode);
1930 	nfs4_stateid *stateid = &state->stateid;
1931 	struct nfs_delegation *delegation;
1932 	struct rpc_cred *cred = NULL;
1933 	int status = -NFS4ERR_BAD_STATEID;
1934 
1935 	/* If a state reset has been done, test_stateid is unneeded */
1936 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1937 		return;
1938 
1939 	/* Get the delegation credential for use by test/free_stateid */
1940 	rcu_read_lock();
1941 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1942 	if (delegation != NULL &&
1943 	    nfs4_stateid_match(&delegation->stateid, stateid)) {
1944 		cred = get_rpccred(delegation->cred);
1945 		rcu_read_unlock();
1946 		status = nfs41_test_stateid(server, stateid, cred);
1947 	} else
1948 		rcu_read_unlock();
1949 
1950 	if (status != NFS_OK) {
1951 		/* Free the stateid unless the server explicitly
1952 		 * informs us the stateid is unrecognized. */
1953 		if (status != -NFS4ERR_BAD_STATEID)
1954 			nfs41_free_stateid(server, stateid, cred);
1955 		nfs_remove_bad_delegation(state->inode);
1956 
1957 		write_seqlock(&state->seqlock);
1958 		nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1959 		write_sequnlock(&state->seqlock);
1960 		clear_bit(NFS_DELEGATED_STATE, &state->flags);
1961 	}
1962 
1963 	if (cred != NULL)
1964 		put_rpccred(cred);
1965 }
1966 
1967 /**
1968  * nfs41_check_open_stateid - possibly free an open stateid
1969  *
1970  * @state: NFSv4 state for an inode
1971  *
1972  * Returns NFS_OK if recovery for this stateid is now finished.
1973  * Otherwise a negative NFS4ERR value is returned.
1974  */
1975 static int nfs41_check_open_stateid(struct nfs4_state *state)
1976 {
1977 	struct nfs_server *server = NFS_SERVER(state->inode);
1978 	nfs4_stateid *stateid = &state->open_stateid;
1979 	struct rpc_cred *cred = state->owner->so_cred;
1980 	int status;
1981 
1982 	/* If a state reset has been done, test_stateid is unneeded */
1983 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1984 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1985 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1986 		return -NFS4ERR_BAD_STATEID;
1987 
1988 	status = nfs41_test_stateid(server, stateid, cred);
1989 	if (status != NFS_OK) {
1990 		/* Free the stateid unless the server explicitly
1991 		 * informs us the stateid is unrecognized. */
1992 		if (status != -NFS4ERR_BAD_STATEID)
1993 			nfs41_free_stateid(server, stateid, cred);
1994 
1995 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1996 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1997 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
1998 		clear_bit(NFS_OPEN_STATE, &state->flags);
1999 	}
2000 	return status;
2001 }
2002 
2003 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2004 {
2005 	int status;
2006 
2007 	nfs41_clear_delegation_stateid(state);
2008 	status = nfs41_check_open_stateid(state);
2009 	if (status != NFS_OK)
2010 		status = nfs4_open_expired(sp, state);
2011 	return status;
2012 }
2013 #endif
2014 
2015 /*
2016  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2017  * fields corresponding to attributes that were used to store the verifier.
2018  * Make sure we clobber those fields in the later setattr call
2019  */
2020 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2021 {
2022 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2023 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2024 		sattr->ia_valid |= ATTR_ATIME;
2025 
2026 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2027 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2028 		sattr->ia_valid |= ATTR_MTIME;
2029 }
2030 
2031 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2032 		fmode_t fmode,
2033 		int flags,
2034 		struct nfs_open_context *ctx)
2035 {
2036 	struct nfs4_state_owner *sp = opendata->owner;
2037 	struct nfs_server *server = sp->so_server;
2038 	struct dentry *dentry;
2039 	struct nfs4_state *state;
2040 	unsigned int seq;
2041 	int ret;
2042 
2043 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2044 
2045 	ret = _nfs4_proc_open(opendata);
2046 	if (ret != 0)
2047 		goto out;
2048 
2049 	state = nfs4_opendata_to_nfs4_state(opendata);
2050 	ret = PTR_ERR(state);
2051 	if (IS_ERR(state))
2052 		goto out;
2053 	if (server->caps & NFS_CAP_POSIX_LOCK)
2054 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2055 
2056 	dentry = opendata->dentry;
2057 	if (dentry->d_inode == NULL) {
2058 		/* FIXME: Is this d_drop() ever needed? */
2059 		d_drop(dentry);
2060 		dentry = d_add_unique(dentry, igrab(state->inode));
2061 		if (dentry == NULL) {
2062 			dentry = opendata->dentry;
2063 		} else if (dentry != ctx->dentry) {
2064 			dput(ctx->dentry);
2065 			ctx->dentry = dget(dentry);
2066 		}
2067 		nfs_set_verifier(dentry,
2068 				nfs_save_change_attribute(opendata->dir->d_inode));
2069 	}
2070 
2071 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2072 	if (ret != 0)
2073 		goto out;
2074 
2075 	ctx->state = state;
2076 	if (dentry->d_inode == state->inode) {
2077 		nfs_inode_attach_open_context(ctx);
2078 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2079 			nfs4_schedule_stateid_recovery(server, state);
2080 	}
2081 out:
2082 	return ret;
2083 }
2084 
2085 /*
2086  * Returns a referenced nfs4_state
2087  */
2088 static int _nfs4_do_open(struct inode *dir,
2089 			struct nfs_open_context *ctx,
2090 			int flags,
2091 			struct iattr *sattr,
2092 			struct nfs4_label *label)
2093 {
2094 	struct nfs4_state_owner  *sp;
2095 	struct nfs4_state     *state = NULL;
2096 	struct nfs_server       *server = NFS_SERVER(dir);
2097 	struct nfs4_opendata *opendata;
2098 	struct dentry *dentry = ctx->dentry;
2099 	struct rpc_cred *cred = ctx->cred;
2100 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2101 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2102 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2103 	struct nfs4_label *olabel = NULL;
2104 	int status;
2105 
2106 	/* Protect against reboot recovery conflicts */
2107 	status = -ENOMEM;
2108 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2109 	if (sp == NULL) {
2110 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2111 		goto out_err;
2112 	}
2113 	status = nfs4_recover_expired_lease(server);
2114 	if (status != 0)
2115 		goto err_put_state_owner;
2116 	if (dentry->d_inode != NULL)
2117 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2118 	status = -ENOMEM;
2119 	if (dentry->d_inode)
2120 		claim = NFS4_OPEN_CLAIM_FH;
2121 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2122 			label, claim, GFP_KERNEL);
2123 	if (opendata == NULL)
2124 		goto err_put_state_owner;
2125 
2126 	if (label) {
2127 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2128 		if (IS_ERR(olabel)) {
2129 			status = PTR_ERR(olabel);
2130 			goto err_opendata_put;
2131 		}
2132 	}
2133 
2134 	if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2135 		opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2136 		if (!opendata->f_attr.mdsthreshold)
2137 			goto err_free_label;
2138 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2139 	}
2140 	if (dentry->d_inode != NULL)
2141 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2142 
2143 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2144 	if (status != 0)
2145 		goto err_free_label;
2146 	state = ctx->state;
2147 
2148 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2149 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2150 		nfs4_exclusive_attrset(opendata, sattr);
2151 
2152 		nfs_fattr_init(opendata->o_res.f_attr);
2153 		status = nfs4_do_setattr(state->inode, cred,
2154 				opendata->o_res.f_attr, sattr,
2155 				state, label, olabel);
2156 		if (status == 0) {
2157 			nfs_setattr_update_inode(state->inode, sattr);
2158 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2159 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2160 		}
2161 	}
2162 
2163 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
2164 		*ctx_th = opendata->f_attr.mdsthreshold;
2165 	else
2166 		kfree(opendata->f_attr.mdsthreshold);
2167 	opendata->f_attr.mdsthreshold = NULL;
2168 
2169 	nfs4_label_free(olabel);
2170 
2171 	nfs4_opendata_put(opendata);
2172 	nfs4_put_state_owner(sp);
2173 	return 0;
2174 err_free_label:
2175 	nfs4_label_free(olabel);
2176 err_opendata_put:
2177 	kfree(opendata->f_attr.mdsthreshold);
2178 	nfs4_opendata_put(opendata);
2179 err_put_state_owner:
2180 	nfs4_put_state_owner(sp);
2181 out_err:
2182 	return status;
2183 }
2184 
2185 
2186 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2187 					struct nfs_open_context *ctx,
2188 					int flags,
2189 					struct iattr *sattr,
2190 					struct nfs4_label *label)
2191 {
2192 	struct nfs_server *server = NFS_SERVER(dir);
2193 	struct nfs4_exception exception = { };
2194 	struct nfs4_state *res;
2195 	int status;
2196 
2197 	do {
2198 		status = _nfs4_do_open(dir, ctx, flags, sattr, label);
2199 		res = ctx->state;
2200 		if (status == 0)
2201 			break;
2202 		/* NOTE: BAD_SEQID means the server and client disagree about the
2203 		 * book-keeping w.r.t. state-changing operations
2204 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2205 		 * It is actually a sign of a bug on the client or on the server.
2206 		 *
2207 		 * If we receive a BAD_SEQID error in the particular case of
2208 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2209 		 * have unhashed the old state_owner for us, and that we can
2210 		 * therefore safely retry using a new one. We should still warn
2211 		 * the user though...
2212 		 */
2213 		if (status == -NFS4ERR_BAD_SEQID) {
2214 			pr_warn_ratelimited("NFS: v4 server %s "
2215 					" returned a bad sequence-id error!\n",
2216 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2217 			exception.retry = 1;
2218 			continue;
2219 		}
2220 		/*
2221 		 * BAD_STATEID on OPEN means that the server cancelled our
2222 		 * state before it received the OPEN_CONFIRM.
2223 		 * Recover by retrying the request as per the discussion
2224 		 * on Page 181 of RFC3530.
2225 		 */
2226 		if (status == -NFS4ERR_BAD_STATEID) {
2227 			exception.retry = 1;
2228 			continue;
2229 		}
2230 		if (status == -EAGAIN) {
2231 			/* We must have found a delegation */
2232 			exception.retry = 1;
2233 			continue;
2234 		}
2235 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2236 			continue;
2237 		res = ERR_PTR(nfs4_handle_exception(server,
2238 					status, &exception));
2239 	} while (exception.retry);
2240 	return res;
2241 }
2242 
2243 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2244 			    struct nfs_fattr *fattr, struct iattr *sattr,
2245 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2246 			    struct nfs4_label *olabel)
2247 {
2248 	struct nfs_server *server = NFS_SERVER(inode);
2249         struct nfs_setattrargs  arg = {
2250                 .fh             = NFS_FH(inode),
2251                 .iap            = sattr,
2252 		.server		= server,
2253 		.bitmask = server->attr_bitmask,
2254 		.label		= ilabel,
2255         };
2256         struct nfs_setattrres  res = {
2257 		.fattr		= fattr,
2258 		.label		= olabel,
2259 		.server		= server,
2260         };
2261         struct rpc_message msg = {
2262 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2263 		.rpc_argp	= &arg,
2264 		.rpc_resp	= &res,
2265 		.rpc_cred	= cred,
2266         };
2267 	unsigned long timestamp = jiffies;
2268 	fmode_t fmode;
2269 	bool truncate;
2270 	int status;
2271 
2272 	arg.bitmask = nfs4_bitmask(server, ilabel);
2273 	if (ilabel)
2274 		arg.bitmask = nfs4_bitmask(server, olabel);
2275 
2276 	nfs_fattr_init(fattr);
2277 
2278 	/* Servers should only apply open mode checks for file size changes */
2279 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2280 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2281 
2282 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2283 		/* Use that stateid */
2284 	} else if (truncate && state != NULL && nfs4_valid_open_stateid(state)) {
2285 		struct nfs_lockowner lockowner = {
2286 			.l_owner = current->files,
2287 			.l_pid = current->tgid,
2288 		};
2289 		nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2290 				&lockowner);
2291 	} else
2292 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2293 
2294 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2295 	if (status == 0 && state != NULL)
2296 		renew_lease(server, timestamp);
2297 	return status;
2298 }
2299 
2300 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2301 			   struct nfs_fattr *fattr, struct iattr *sattr,
2302 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2303 			   struct nfs4_label *olabel)
2304 {
2305 	struct nfs_server *server = NFS_SERVER(inode);
2306 	struct nfs4_exception exception = {
2307 		.state = state,
2308 		.inode = inode,
2309 	};
2310 	int err;
2311 	do {
2312 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2313 		switch (err) {
2314 		case -NFS4ERR_OPENMODE:
2315 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2316 				pr_warn_once("NFSv4: server %s is incorrectly "
2317 						"applying open mode checks to "
2318 						"a SETATTR that is not "
2319 						"changing file size.\n",
2320 						server->nfs_client->cl_hostname);
2321 			}
2322 			if (state && !(state->state & FMODE_WRITE)) {
2323 				err = -EBADF;
2324 				if (sattr->ia_valid & ATTR_OPEN)
2325 					err = -EACCES;
2326 				goto out;
2327 			}
2328 		}
2329 		err = nfs4_handle_exception(server, err, &exception);
2330 	} while (exception.retry);
2331 out:
2332 	return err;
2333 }
2334 
2335 struct nfs4_closedata {
2336 	struct inode *inode;
2337 	struct nfs4_state *state;
2338 	struct nfs_closeargs arg;
2339 	struct nfs_closeres res;
2340 	struct nfs_fattr fattr;
2341 	unsigned long timestamp;
2342 	bool roc;
2343 	u32 roc_barrier;
2344 };
2345 
2346 static void nfs4_free_closedata(void *data)
2347 {
2348 	struct nfs4_closedata *calldata = data;
2349 	struct nfs4_state_owner *sp = calldata->state->owner;
2350 	struct super_block *sb = calldata->state->inode->i_sb;
2351 
2352 	if (calldata->roc)
2353 		pnfs_roc_release(calldata->state->inode);
2354 	nfs4_put_open_state(calldata->state);
2355 	nfs_free_seqid(calldata->arg.seqid);
2356 	nfs4_put_state_owner(sp);
2357 	nfs_sb_deactive(sb);
2358 	kfree(calldata);
2359 }
2360 
2361 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2362 		fmode_t fmode)
2363 {
2364 	spin_lock(&state->owner->so_lock);
2365 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
2366 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
2367 	case FMODE_WRITE:
2368 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2369 		break;
2370 	case FMODE_READ:
2371 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2372 		break;
2373 	case 0:
2374 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2375 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2376 		clear_bit(NFS_OPEN_STATE, &state->flags);
2377 	}
2378 	spin_unlock(&state->owner->so_lock);
2379 }
2380 
2381 static void nfs4_close_done(struct rpc_task *task, void *data)
2382 {
2383 	struct nfs4_closedata *calldata = data;
2384 	struct nfs4_state *state = calldata->state;
2385 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2386 
2387 	dprintk("%s: begin!\n", __func__);
2388 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2389 		return;
2390         /* hmm. we are done with the inode, and in the process of freeing
2391 	 * the state_owner. we keep this around to process errors
2392 	 */
2393 	switch (task->tk_status) {
2394 		case 0:
2395 			if (calldata->roc)
2396 				pnfs_roc_set_barrier(state->inode,
2397 						     calldata->roc_barrier);
2398 			nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2399 			renew_lease(server, calldata->timestamp);
2400 			nfs4_close_clear_stateid_flags(state,
2401 					calldata->arg.fmode);
2402 			break;
2403 		case -NFS4ERR_STALE_STATEID:
2404 		case -NFS4ERR_OLD_STATEID:
2405 		case -NFS4ERR_BAD_STATEID:
2406 		case -NFS4ERR_EXPIRED:
2407 			if (calldata->arg.fmode == 0)
2408 				break;
2409 		default:
2410 			if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2411 				rpc_restart_call_prepare(task);
2412 	}
2413 	nfs_release_seqid(calldata->arg.seqid);
2414 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2415 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2416 }
2417 
2418 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2419 {
2420 	struct nfs4_closedata *calldata = data;
2421 	struct nfs4_state *state = calldata->state;
2422 	struct inode *inode = calldata->inode;
2423 	int call_close = 0;
2424 
2425 	dprintk("%s: begin!\n", __func__);
2426 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2427 		goto out_wait;
2428 
2429 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2430 	calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2431 	spin_lock(&state->owner->so_lock);
2432 	/* Calculate the change in open mode */
2433 	if (state->n_rdwr == 0) {
2434 		if (state->n_rdonly == 0) {
2435 			call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2436 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2437 			calldata->arg.fmode &= ~FMODE_READ;
2438 		}
2439 		if (state->n_wronly == 0) {
2440 			call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2441 			call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2442 			calldata->arg.fmode &= ~FMODE_WRITE;
2443 		}
2444 	}
2445 	if (!nfs4_valid_open_stateid(state))
2446 		call_close = 0;
2447 	spin_unlock(&state->owner->so_lock);
2448 
2449 	if (!call_close) {
2450 		/* Note: exit _without_ calling nfs4_close_done */
2451 		goto out_no_action;
2452 	}
2453 
2454 	if (calldata->arg.fmode == 0) {
2455 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2456 		if (calldata->roc &&
2457 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2458 			nfs_release_seqid(calldata->arg.seqid);
2459 			goto out_wait;
2460 		    }
2461 	}
2462 
2463 	nfs_fattr_init(calldata->res.fattr);
2464 	calldata->timestamp = jiffies;
2465 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2466 				&calldata->arg.seq_args,
2467 				&calldata->res.seq_res,
2468 				task) != 0)
2469 		nfs_release_seqid(calldata->arg.seqid);
2470 	dprintk("%s: done!\n", __func__);
2471 	return;
2472 out_no_action:
2473 	task->tk_action = NULL;
2474 out_wait:
2475 	nfs4_sequence_done(task, &calldata->res.seq_res);
2476 }
2477 
2478 static const struct rpc_call_ops nfs4_close_ops = {
2479 	.rpc_call_prepare = nfs4_close_prepare,
2480 	.rpc_call_done = nfs4_close_done,
2481 	.rpc_release = nfs4_free_closedata,
2482 };
2483 
2484 /*
2485  * It is possible for data to be read/written from a mem-mapped file
2486  * after the sys_close call (which hits the vfs layer as a flush).
2487  * This means that we can't safely call nfsv4 close on a file until
2488  * the inode is cleared. This in turn means that we are not good
2489  * NFSv4 citizens - we do not indicate to the server to update the file's
2490  * share state even when we are done with one of the three share
2491  * stateid's in the inode.
2492  *
2493  * NOTE: Caller must be holding the sp->so_owner semaphore!
2494  */
2495 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2496 {
2497 	struct nfs_server *server = NFS_SERVER(state->inode);
2498 	struct nfs4_closedata *calldata;
2499 	struct nfs4_state_owner *sp = state->owner;
2500 	struct rpc_task *task;
2501 	struct rpc_message msg = {
2502 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2503 		.rpc_cred = state->owner->so_cred,
2504 	};
2505 	struct rpc_task_setup task_setup_data = {
2506 		.rpc_client = server->client,
2507 		.rpc_message = &msg,
2508 		.callback_ops = &nfs4_close_ops,
2509 		.workqueue = nfsiod_workqueue,
2510 		.flags = RPC_TASK_ASYNC,
2511 	};
2512 	int status = -ENOMEM;
2513 
2514 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2515 	if (calldata == NULL)
2516 		goto out;
2517 	nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2518 	calldata->inode = state->inode;
2519 	calldata->state = state;
2520 	calldata->arg.fh = NFS_FH(state->inode);
2521 	calldata->arg.stateid = &state->open_stateid;
2522 	/* Serialization for the sequence id */
2523 	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2524 	if (calldata->arg.seqid == NULL)
2525 		goto out_free_calldata;
2526 	calldata->arg.fmode = 0;
2527 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2528 	calldata->res.fattr = &calldata->fattr;
2529 	calldata->res.seqid = calldata->arg.seqid;
2530 	calldata->res.server = server;
2531 	calldata->roc = pnfs_roc(state->inode);
2532 	nfs_sb_active(calldata->inode->i_sb);
2533 
2534 	msg.rpc_argp = &calldata->arg;
2535 	msg.rpc_resp = &calldata->res;
2536 	task_setup_data.callback_data = calldata;
2537 	task = rpc_run_task(&task_setup_data);
2538 	if (IS_ERR(task))
2539 		return PTR_ERR(task);
2540 	status = 0;
2541 	if (wait)
2542 		status = rpc_wait_for_completion_task(task);
2543 	rpc_put_task(task);
2544 	return status;
2545 out_free_calldata:
2546 	kfree(calldata);
2547 out:
2548 	nfs4_put_open_state(state);
2549 	nfs4_put_state_owner(sp);
2550 	return status;
2551 }
2552 
2553 static struct inode *
2554 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2555 {
2556 	struct nfs4_state *state;
2557 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2558 
2559 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2560 
2561 	/* Protect against concurrent sillydeletes */
2562 	state = nfs4_do_open(dir, ctx, open_flags, attr, label);
2563 
2564 	nfs4_label_release_security(label);
2565 
2566 	if (IS_ERR(state))
2567 		return ERR_CAST(state);
2568 	return state->inode;
2569 }
2570 
2571 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2572 {
2573 	if (ctx->state == NULL)
2574 		return;
2575 	if (is_sync)
2576 		nfs4_close_sync(ctx->state, ctx->mode);
2577 	else
2578 		nfs4_close_state(ctx->state, ctx->mode);
2579 }
2580 
2581 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2582 {
2583 	struct nfs4_server_caps_arg args = {
2584 		.fhandle = fhandle,
2585 	};
2586 	struct nfs4_server_caps_res res = {};
2587 	struct rpc_message msg = {
2588 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2589 		.rpc_argp = &args,
2590 		.rpc_resp = &res,
2591 	};
2592 	int status;
2593 
2594 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2595 	if (status == 0) {
2596 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2597 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2598 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2599 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2600 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2601 				NFS_CAP_CTIME|NFS_CAP_MTIME);
2602 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2603 			server->caps |= NFS_CAP_ACLS;
2604 		if (res.has_links != 0)
2605 			server->caps |= NFS_CAP_HARDLINKS;
2606 		if (res.has_symlinks != 0)
2607 			server->caps |= NFS_CAP_SYMLINKS;
2608 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2609 			server->caps |= NFS_CAP_FILEID;
2610 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2611 			server->caps |= NFS_CAP_MODE;
2612 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2613 			server->caps |= NFS_CAP_NLINK;
2614 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2615 			server->caps |= NFS_CAP_OWNER;
2616 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2617 			server->caps |= NFS_CAP_OWNER_GROUP;
2618 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2619 			server->caps |= NFS_CAP_ATIME;
2620 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2621 			server->caps |= NFS_CAP_CTIME;
2622 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2623 			server->caps |= NFS_CAP_MTIME;
2624 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2625 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2626 			server->caps |= NFS_CAP_SECURITY_LABEL;
2627 #endif
2628 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2629 				sizeof(server->attr_bitmask));
2630 
2631 		if (server->caps & NFS_CAP_SECURITY_LABEL) {
2632 			server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2633 			res.attr_bitmask[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2634 		}
2635 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2636 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2637 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2638 		server->acl_bitmask = res.acl_bitmask;
2639 		server->fh_expire_type = res.fh_expire_type;
2640 	}
2641 
2642 	return status;
2643 }
2644 
2645 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2646 {
2647 	struct nfs4_exception exception = { };
2648 	int err;
2649 	do {
2650 		err = nfs4_handle_exception(server,
2651 				_nfs4_server_capabilities(server, fhandle),
2652 				&exception);
2653 	} while (exception.retry);
2654 	return err;
2655 }
2656 
2657 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2658 		struct nfs_fsinfo *info)
2659 {
2660 	u32 bitmask[3];
2661 	struct nfs4_lookup_root_arg args = {
2662 		.bitmask = bitmask,
2663 	};
2664 	struct nfs4_lookup_res res = {
2665 		.server = server,
2666 		.fattr = info->fattr,
2667 		.fh = fhandle,
2668 	};
2669 	struct rpc_message msg = {
2670 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2671 		.rpc_argp = &args,
2672 		.rpc_resp = &res,
2673 	};
2674 
2675 	bitmask[0] = nfs4_fattr_bitmap[0];
2676 	bitmask[1] = nfs4_fattr_bitmap[1];
2677 	/*
2678 	 * Process the label in the upcoming getfattr
2679 	 */
2680 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2681 
2682 	nfs_fattr_init(info->fattr);
2683 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2684 }
2685 
2686 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2687 		struct nfs_fsinfo *info)
2688 {
2689 	struct nfs4_exception exception = { };
2690 	int err;
2691 	do {
2692 		err = _nfs4_lookup_root(server, fhandle, info);
2693 		switch (err) {
2694 		case 0:
2695 		case -NFS4ERR_WRONGSEC:
2696 			goto out;
2697 		default:
2698 			err = nfs4_handle_exception(server, err, &exception);
2699 		}
2700 	} while (exception.retry);
2701 out:
2702 	return err;
2703 }
2704 
2705 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2706 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2707 {
2708 	struct rpc_auth *auth;
2709 	int ret;
2710 
2711 	auth = rpcauth_create(flavor, server->client);
2712 	if (IS_ERR(auth)) {
2713 		ret = -EACCES;
2714 		goto out;
2715 	}
2716 	ret = nfs4_lookup_root(server, fhandle, info);
2717 out:
2718 	return ret;
2719 }
2720 
2721 /*
2722  * Retry pseudoroot lookup with various security flavors.  We do this when:
2723  *
2724  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
2725  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
2726  *
2727  * Returns zero on success, or a negative NFS4ERR value, or a
2728  * negative errno value.
2729  */
2730 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2731 			      struct nfs_fsinfo *info)
2732 {
2733 	/* Per 3530bis 15.33.5 */
2734 	static const rpc_authflavor_t flav_array[] = {
2735 		RPC_AUTH_GSS_KRB5P,
2736 		RPC_AUTH_GSS_KRB5I,
2737 		RPC_AUTH_GSS_KRB5,
2738 		RPC_AUTH_UNIX,			/* courtesy */
2739 		RPC_AUTH_NULL,
2740 	};
2741 	int status = -EPERM;
2742 	size_t i;
2743 
2744 	for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
2745 		status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2746 		if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2747 			continue;
2748 		break;
2749 	}
2750 
2751 	/*
2752 	 * -EACCESS could mean that the user doesn't have correct permissions
2753 	 * to access the mount.  It could also mean that we tried to mount
2754 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2755 	 * existing mount programs don't handle -EACCES very well so it should
2756 	 * be mapped to -EPERM instead.
2757 	 */
2758 	if (status == -EACCES)
2759 		status = -EPERM;
2760 	return status;
2761 }
2762 
2763 static int nfs4_do_find_root_sec(struct nfs_server *server,
2764 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
2765 {
2766 	int mv = server->nfs_client->cl_minorversion;
2767 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
2768 }
2769 
2770 /**
2771  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
2772  * @server: initialized nfs_server handle
2773  * @fhandle: we fill in the pseudo-fs root file handle
2774  * @info: we fill in an FSINFO struct
2775  *
2776  * Returns zero on success, or a negative errno.
2777  */
2778 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2779 			 struct nfs_fsinfo *info)
2780 {
2781 	int status;
2782 
2783 	status = nfs4_lookup_root(server, fhandle, info);
2784 	if ((status == -NFS4ERR_WRONGSEC) &&
2785 	    !(server->flags & NFS_MOUNT_SECFLAVOUR))
2786 		status = nfs4_do_find_root_sec(server, fhandle, info);
2787 
2788 	if (status == 0)
2789 		status = nfs4_server_capabilities(server, fhandle);
2790 	if (status == 0)
2791 		status = nfs4_do_fsinfo(server, fhandle, info);
2792 
2793 	return nfs4_map_errors(status);
2794 }
2795 
2796 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2797 			      struct nfs_fsinfo *info)
2798 {
2799 	int error;
2800 	struct nfs_fattr *fattr = info->fattr;
2801 	struct nfs4_label *label = NULL;
2802 
2803 	error = nfs4_server_capabilities(server, mntfh);
2804 	if (error < 0) {
2805 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2806 		return error;
2807 	}
2808 
2809 	label = nfs4_label_alloc(server, GFP_KERNEL);
2810 	if (IS_ERR(label))
2811 		return PTR_ERR(label);
2812 
2813 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
2814 	if (error < 0) {
2815 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
2816 		goto err_free_label;
2817 	}
2818 
2819 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2820 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2821 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2822 
2823 err_free_label:
2824 	nfs4_label_free(label);
2825 
2826 	return error;
2827 }
2828 
2829 /*
2830  * Get locations and (maybe) other attributes of a referral.
2831  * Note that we'll actually follow the referral later when
2832  * we detect fsid mismatch in inode revalidation
2833  */
2834 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2835 			     const struct qstr *name, struct nfs_fattr *fattr,
2836 			     struct nfs_fh *fhandle)
2837 {
2838 	int status = -ENOMEM;
2839 	struct page *page = NULL;
2840 	struct nfs4_fs_locations *locations = NULL;
2841 
2842 	page = alloc_page(GFP_KERNEL);
2843 	if (page == NULL)
2844 		goto out;
2845 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2846 	if (locations == NULL)
2847 		goto out;
2848 
2849 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2850 	if (status != 0)
2851 		goto out;
2852 	/* Make sure server returned a different fsid for the referral */
2853 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2854 		dprintk("%s: server did not return a different fsid for"
2855 			" a referral at %s\n", __func__, name->name);
2856 		status = -EIO;
2857 		goto out;
2858 	}
2859 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2860 	nfs_fixup_referral_attributes(&locations->fattr);
2861 
2862 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
2863 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2864 	memset(fhandle, 0, sizeof(struct nfs_fh));
2865 out:
2866 	if (page)
2867 		__free_page(page);
2868 	kfree(locations);
2869 	return status;
2870 }
2871 
2872 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
2873 				struct nfs_fattr *fattr, struct nfs4_label *label)
2874 {
2875 	struct nfs4_getattr_arg args = {
2876 		.fh = fhandle,
2877 		.bitmask = server->attr_bitmask,
2878 	};
2879 	struct nfs4_getattr_res res = {
2880 		.fattr = fattr,
2881 		.label = label,
2882 		.server = server,
2883 	};
2884 	struct rpc_message msg = {
2885 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2886 		.rpc_argp = &args,
2887 		.rpc_resp = &res,
2888 	};
2889 
2890 	args.bitmask = nfs4_bitmask(server, label);
2891 
2892 	nfs_fattr_init(fattr);
2893 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2894 }
2895 
2896 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
2897 				struct nfs_fattr *fattr, struct nfs4_label *label)
2898 {
2899 	struct nfs4_exception exception = { };
2900 	int err;
2901 	do {
2902 		err = nfs4_handle_exception(server,
2903 				_nfs4_proc_getattr(server, fhandle, fattr, label),
2904 				&exception);
2905 	} while (exception.retry);
2906 	return err;
2907 }
2908 
2909 /*
2910  * The file is not closed if it is opened due to the a request to change
2911  * the size of the file. The open call will not be needed once the
2912  * VFS layer lookup-intents are implemented.
2913  *
2914  * Close is called when the inode is destroyed.
2915  * If we haven't opened the file for O_WRONLY, we
2916  * need to in the size_change case to obtain a stateid.
2917  *
2918  * Got race?
2919  * Because OPEN is always done by name in nfsv4, it is
2920  * possible that we opened a different file by the same
2921  * name.  We can recognize this race condition, but we
2922  * can't do anything about it besides returning an error.
2923  *
2924  * This will be fixed with VFS changes (lookup-intent).
2925  */
2926 static int
2927 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2928 		  struct iattr *sattr)
2929 {
2930 	struct inode *inode = dentry->d_inode;
2931 	struct rpc_cred *cred = NULL;
2932 	struct nfs4_state *state = NULL;
2933 	struct nfs4_label *label = NULL;
2934 	int status;
2935 
2936 	if (pnfs_ld_layoutret_on_setattr(inode))
2937 		pnfs_commit_and_return_layout(inode);
2938 
2939 	nfs_fattr_init(fattr);
2940 
2941 	/* Deal with open(O_TRUNC) */
2942 	if (sattr->ia_valid & ATTR_OPEN)
2943 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2944 
2945 	/* Optimization: if the end result is no change, don't RPC */
2946 	if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2947 		return 0;
2948 
2949 	/* Search for an existing open(O_WRITE) file */
2950 	if (sattr->ia_valid & ATTR_FILE) {
2951 		struct nfs_open_context *ctx;
2952 
2953 		ctx = nfs_file_open_context(sattr->ia_file);
2954 		if (ctx) {
2955 			cred = ctx->cred;
2956 			state = ctx->state;
2957 		}
2958 	}
2959 
2960 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
2961 	if (IS_ERR(label))
2962 		return PTR_ERR(label);
2963 
2964 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
2965 	if (status == 0) {
2966 		nfs_setattr_update_inode(inode, sattr);
2967 		nfs_setsecurity(inode, fattr, label);
2968 	}
2969 	nfs4_label_free(label);
2970 	return status;
2971 }
2972 
2973 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2974 		const struct qstr *name, struct nfs_fh *fhandle,
2975 		struct nfs_fattr *fattr, struct nfs4_label *label)
2976 {
2977 	struct nfs_server *server = NFS_SERVER(dir);
2978 	int		       status;
2979 	struct nfs4_lookup_arg args = {
2980 		.bitmask = server->attr_bitmask,
2981 		.dir_fh = NFS_FH(dir),
2982 		.name = name,
2983 	};
2984 	struct nfs4_lookup_res res = {
2985 		.server = server,
2986 		.fattr = fattr,
2987 		.label = label,
2988 		.fh = fhandle,
2989 	};
2990 	struct rpc_message msg = {
2991 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2992 		.rpc_argp = &args,
2993 		.rpc_resp = &res,
2994 	};
2995 
2996 	args.bitmask = nfs4_bitmask(server, label);
2997 
2998 	nfs_fattr_init(fattr);
2999 
3000 	dprintk("NFS call  lookup %s\n", name->name);
3001 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3002 	dprintk("NFS reply lookup: %d\n", status);
3003 	return status;
3004 }
3005 
3006 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3007 {
3008 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3009 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3010 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3011 	fattr->nlink = 2;
3012 }
3013 
3014 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3015 				   struct qstr *name, struct nfs_fh *fhandle,
3016 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3017 {
3018 	struct nfs4_exception exception = { };
3019 	struct rpc_clnt *client = *clnt;
3020 	int err;
3021 	do {
3022 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3023 		switch (err) {
3024 		case -NFS4ERR_BADNAME:
3025 			err = -ENOENT;
3026 			goto out;
3027 		case -NFS4ERR_MOVED:
3028 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3029 			goto out;
3030 		case -NFS4ERR_WRONGSEC:
3031 			err = -EPERM;
3032 			if (client != *clnt)
3033 				goto out;
3034 
3035 			client = nfs4_create_sec_client(client, dir, name);
3036 			if (IS_ERR(client))
3037 				return PTR_ERR(client);
3038 
3039 			exception.retry = 1;
3040 			break;
3041 		default:
3042 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3043 		}
3044 	} while (exception.retry);
3045 
3046 out:
3047 	if (err == 0)
3048 		*clnt = client;
3049 	else if (client != *clnt)
3050 		rpc_shutdown_client(client);
3051 
3052 	return err;
3053 }
3054 
3055 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3056 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3057 			    struct nfs4_label *label)
3058 {
3059 	int status;
3060 	struct rpc_clnt *client = NFS_CLIENT(dir);
3061 
3062 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3063 	if (client != NFS_CLIENT(dir)) {
3064 		rpc_shutdown_client(client);
3065 		nfs_fixup_secinfo_attributes(fattr);
3066 	}
3067 	return status;
3068 }
3069 
3070 struct rpc_clnt *
3071 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3072 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3073 {
3074 	struct rpc_clnt *client = NFS_CLIENT(dir);
3075 	int status;
3076 
3077 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3078 	if (status < 0)
3079 		return ERR_PTR(status);
3080 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3081 }
3082 
3083 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3084 {
3085 	struct nfs_server *server = NFS_SERVER(inode);
3086 	struct nfs4_accessargs args = {
3087 		.fh = NFS_FH(inode),
3088 		.bitmask = server->cache_consistency_bitmask,
3089 	};
3090 	struct nfs4_accessres res = {
3091 		.server = server,
3092 	};
3093 	struct rpc_message msg = {
3094 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3095 		.rpc_argp = &args,
3096 		.rpc_resp = &res,
3097 		.rpc_cred = entry->cred,
3098 	};
3099 	int mode = entry->mask;
3100 	int status = 0;
3101 
3102 	/*
3103 	 * Determine which access bits we want to ask for...
3104 	 */
3105 	if (mode & MAY_READ)
3106 		args.access |= NFS4_ACCESS_READ;
3107 	if (S_ISDIR(inode->i_mode)) {
3108 		if (mode & MAY_WRITE)
3109 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3110 		if (mode & MAY_EXEC)
3111 			args.access |= NFS4_ACCESS_LOOKUP;
3112 	} else {
3113 		if (mode & MAY_WRITE)
3114 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3115 		if (mode & MAY_EXEC)
3116 			args.access |= NFS4_ACCESS_EXECUTE;
3117 	}
3118 
3119 	res.fattr = nfs_alloc_fattr();
3120 	if (res.fattr == NULL)
3121 		return -ENOMEM;
3122 
3123 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3124 	if (!status) {
3125 		nfs_access_set_mask(entry, res.access);
3126 		nfs_refresh_inode(inode, res.fattr);
3127 	}
3128 	nfs_free_fattr(res.fattr);
3129 	return status;
3130 }
3131 
3132 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3133 {
3134 	struct nfs4_exception exception = { };
3135 	int err;
3136 	do {
3137 		err = nfs4_handle_exception(NFS_SERVER(inode),
3138 				_nfs4_proc_access(inode, entry),
3139 				&exception);
3140 	} while (exception.retry);
3141 	return err;
3142 }
3143 
3144 /*
3145  * TODO: For the time being, we don't try to get any attributes
3146  * along with any of the zero-copy operations READ, READDIR,
3147  * READLINK, WRITE.
3148  *
3149  * In the case of the first three, we want to put the GETATTR
3150  * after the read-type operation -- this is because it is hard
3151  * to predict the length of a GETATTR response in v4, and thus
3152  * align the READ data correctly.  This means that the GETATTR
3153  * may end up partially falling into the page cache, and we should
3154  * shift it into the 'tail' of the xdr_buf before processing.
3155  * To do this efficiently, we need to know the total length
3156  * of data received, which doesn't seem to be available outside
3157  * of the RPC layer.
3158  *
3159  * In the case of WRITE, we also want to put the GETATTR after
3160  * the operation -- in this case because we want to make sure
3161  * we get the post-operation mtime and size.
3162  *
3163  * Both of these changes to the XDR layer would in fact be quite
3164  * minor, but I decided to leave them for a subsequent patch.
3165  */
3166 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3167 		unsigned int pgbase, unsigned int pglen)
3168 {
3169 	struct nfs4_readlink args = {
3170 		.fh       = NFS_FH(inode),
3171 		.pgbase	  = pgbase,
3172 		.pglen    = pglen,
3173 		.pages    = &page,
3174 	};
3175 	struct nfs4_readlink_res res;
3176 	struct rpc_message msg = {
3177 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3178 		.rpc_argp = &args,
3179 		.rpc_resp = &res,
3180 	};
3181 
3182 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3183 }
3184 
3185 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3186 		unsigned int pgbase, unsigned int pglen)
3187 {
3188 	struct nfs4_exception exception = { };
3189 	int err;
3190 	do {
3191 		err = nfs4_handle_exception(NFS_SERVER(inode),
3192 				_nfs4_proc_readlink(inode, page, pgbase, pglen),
3193 				&exception);
3194 	} while (exception.retry);
3195 	return err;
3196 }
3197 
3198 /*
3199  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3200  */
3201 static int
3202 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3203 		 int flags)
3204 {
3205 	struct nfs4_label l, *ilabel = NULL;
3206 	struct nfs_open_context *ctx;
3207 	struct nfs4_state *state;
3208 	int status = 0;
3209 
3210 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3211 	if (IS_ERR(ctx))
3212 		return PTR_ERR(ctx);
3213 
3214 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3215 
3216 	sattr->ia_mode &= ~current_umask();
3217 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel);
3218 	if (IS_ERR(state)) {
3219 		status = PTR_ERR(state);
3220 		goto out;
3221 	}
3222 out:
3223 	nfs4_label_release_security(ilabel);
3224 	put_nfs_open_context(ctx);
3225 	return status;
3226 }
3227 
3228 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3229 {
3230 	struct nfs_server *server = NFS_SERVER(dir);
3231 	struct nfs_removeargs args = {
3232 		.fh = NFS_FH(dir),
3233 		.name = *name,
3234 	};
3235 	struct nfs_removeres res = {
3236 		.server = server,
3237 	};
3238 	struct rpc_message msg = {
3239 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3240 		.rpc_argp = &args,
3241 		.rpc_resp = &res,
3242 	};
3243 	int status;
3244 
3245 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3246 	if (status == 0)
3247 		update_changeattr(dir, &res.cinfo);
3248 	return status;
3249 }
3250 
3251 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3252 {
3253 	struct nfs4_exception exception = { };
3254 	int err;
3255 	do {
3256 		err = nfs4_handle_exception(NFS_SERVER(dir),
3257 				_nfs4_proc_remove(dir, name),
3258 				&exception);
3259 	} while (exception.retry);
3260 	return err;
3261 }
3262 
3263 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3264 {
3265 	struct nfs_server *server = NFS_SERVER(dir);
3266 	struct nfs_removeargs *args = msg->rpc_argp;
3267 	struct nfs_removeres *res = msg->rpc_resp;
3268 
3269 	res->server = server;
3270 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3271 	nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
3272 
3273 	nfs_fattr_init(res->dir_attr);
3274 }
3275 
3276 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3277 {
3278 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3279 			&data->args.seq_args,
3280 			&data->res.seq_res,
3281 			task);
3282 }
3283 
3284 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3285 {
3286 	struct nfs_removeres *res = task->tk_msg.rpc_resp;
3287 
3288 	if (!nfs4_sequence_done(task, &res->seq_res))
3289 		return 0;
3290 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3291 		return 0;
3292 	update_changeattr(dir, &res->cinfo);
3293 	return 1;
3294 }
3295 
3296 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3297 {
3298 	struct nfs_server *server = NFS_SERVER(dir);
3299 	struct nfs_renameargs *arg = msg->rpc_argp;
3300 	struct nfs_renameres *res = msg->rpc_resp;
3301 
3302 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3303 	res->server = server;
3304 	nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
3305 }
3306 
3307 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3308 {
3309 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3310 			&data->args.seq_args,
3311 			&data->res.seq_res,
3312 			task);
3313 }
3314 
3315 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3316 				 struct inode *new_dir)
3317 {
3318 	struct nfs_renameres *res = task->tk_msg.rpc_resp;
3319 
3320 	if (!nfs4_sequence_done(task, &res->seq_res))
3321 		return 0;
3322 	if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3323 		return 0;
3324 
3325 	update_changeattr(old_dir, &res->old_cinfo);
3326 	update_changeattr(new_dir, &res->new_cinfo);
3327 	return 1;
3328 }
3329 
3330 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3331 		struct inode *new_dir, struct qstr *new_name)
3332 {
3333 	struct nfs_server *server = NFS_SERVER(old_dir);
3334 	struct nfs_renameargs arg = {
3335 		.old_dir = NFS_FH(old_dir),
3336 		.new_dir = NFS_FH(new_dir),
3337 		.old_name = old_name,
3338 		.new_name = new_name,
3339 	};
3340 	struct nfs_renameres res = {
3341 		.server = server,
3342 	};
3343 	struct rpc_message msg = {
3344 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3345 		.rpc_argp = &arg,
3346 		.rpc_resp = &res,
3347 	};
3348 	int status = -ENOMEM;
3349 
3350 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3351 	if (!status) {
3352 		update_changeattr(old_dir, &res.old_cinfo);
3353 		update_changeattr(new_dir, &res.new_cinfo);
3354 	}
3355 	return status;
3356 }
3357 
3358 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3359 		struct inode *new_dir, struct qstr *new_name)
3360 {
3361 	struct nfs4_exception exception = { };
3362 	int err;
3363 	do {
3364 		err = nfs4_handle_exception(NFS_SERVER(old_dir),
3365 				_nfs4_proc_rename(old_dir, old_name,
3366 					new_dir, new_name),
3367 				&exception);
3368 	} while (exception.retry);
3369 	return err;
3370 }
3371 
3372 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3373 {
3374 	struct nfs_server *server = NFS_SERVER(inode);
3375 	struct nfs4_link_arg arg = {
3376 		.fh     = NFS_FH(inode),
3377 		.dir_fh = NFS_FH(dir),
3378 		.name   = name,
3379 		.bitmask = server->attr_bitmask,
3380 	};
3381 	struct nfs4_link_res res = {
3382 		.server = server,
3383 		.label = NULL,
3384 	};
3385 	struct rpc_message msg = {
3386 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3387 		.rpc_argp = &arg,
3388 		.rpc_resp = &res,
3389 	};
3390 	int status = -ENOMEM;
3391 
3392 	res.fattr = nfs_alloc_fattr();
3393 	if (res.fattr == NULL)
3394 		goto out;
3395 
3396 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3397 	if (IS_ERR(res.label)) {
3398 		status = PTR_ERR(res.label);
3399 		goto out;
3400 	}
3401 	arg.bitmask = nfs4_bitmask(server, res.label);
3402 
3403 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3404 	if (!status) {
3405 		update_changeattr(dir, &res.cinfo);
3406 		status = nfs_post_op_update_inode(inode, res.fattr);
3407 		if (!status)
3408 			nfs_setsecurity(inode, res.fattr, res.label);
3409 	}
3410 
3411 
3412 	nfs4_label_free(res.label);
3413 
3414 out:
3415 	nfs_free_fattr(res.fattr);
3416 	return status;
3417 }
3418 
3419 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3420 {
3421 	struct nfs4_exception exception = { };
3422 	int err;
3423 	do {
3424 		err = nfs4_handle_exception(NFS_SERVER(inode),
3425 				_nfs4_proc_link(inode, dir, name),
3426 				&exception);
3427 	} while (exception.retry);
3428 	return err;
3429 }
3430 
3431 struct nfs4_createdata {
3432 	struct rpc_message msg;
3433 	struct nfs4_create_arg arg;
3434 	struct nfs4_create_res res;
3435 	struct nfs_fh fh;
3436 	struct nfs_fattr fattr;
3437 	struct nfs4_label *label;
3438 };
3439 
3440 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3441 		struct qstr *name, struct iattr *sattr, u32 ftype)
3442 {
3443 	struct nfs4_createdata *data;
3444 
3445 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3446 	if (data != NULL) {
3447 		struct nfs_server *server = NFS_SERVER(dir);
3448 
3449 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3450 		if (IS_ERR(data->label))
3451 			goto out_free;
3452 
3453 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3454 		data->msg.rpc_argp = &data->arg;
3455 		data->msg.rpc_resp = &data->res;
3456 		data->arg.dir_fh = NFS_FH(dir);
3457 		data->arg.server = server;
3458 		data->arg.name = name;
3459 		data->arg.attrs = sattr;
3460 		data->arg.ftype = ftype;
3461 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3462 		data->res.server = server;
3463 		data->res.fh = &data->fh;
3464 		data->res.fattr = &data->fattr;
3465 		data->res.label = data->label;
3466 		nfs_fattr_init(data->res.fattr);
3467 	}
3468 	return data;
3469 out_free:
3470 	kfree(data);
3471 	return NULL;
3472 }
3473 
3474 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3475 {
3476 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3477 				    &data->arg.seq_args, &data->res.seq_res, 1);
3478 	if (status == 0) {
3479 		update_changeattr(dir, &data->res.dir_cinfo);
3480 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3481 	}
3482 	return status;
3483 }
3484 
3485 static void nfs4_free_createdata(struct nfs4_createdata *data)
3486 {
3487 	nfs4_label_free(data->label);
3488 	kfree(data);
3489 }
3490 
3491 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3492 		struct page *page, unsigned int len, struct iattr *sattr,
3493 		struct nfs4_label *label)
3494 {
3495 	struct nfs4_createdata *data;
3496 	int status = -ENAMETOOLONG;
3497 
3498 	if (len > NFS4_MAXPATHLEN)
3499 		goto out;
3500 
3501 	status = -ENOMEM;
3502 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3503 	if (data == NULL)
3504 		goto out;
3505 
3506 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3507 	data->arg.u.symlink.pages = &page;
3508 	data->arg.u.symlink.len = len;
3509 	data->arg.label = label;
3510 
3511 	status = nfs4_do_create(dir, dentry, data);
3512 
3513 	nfs4_free_createdata(data);
3514 out:
3515 	return status;
3516 }
3517 
3518 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3519 		struct page *page, unsigned int len, struct iattr *sattr)
3520 {
3521 	struct nfs4_exception exception = { };
3522 	struct nfs4_label l, *label = NULL;
3523 	int err;
3524 
3525 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3526 
3527 	do {
3528 		err = nfs4_handle_exception(NFS_SERVER(dir),
3529 				_nfs4_proc_symlink(dir, dentry, page,
3530 							len, sattr, label),
3531 				&exception);
3532 	} while (exception.retry);
3533 
3534 	nfs4_label_release_security(label);
3535 	return err;
3536 }
3537 
3538 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3539 		struct iattr *sattr, struct nfs4_label *label)
3540 {
3541 	struct nfs4_createdata *data;
3542 	int status = -ENOMEM;
3543 
3544 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3545 	if (data == NULL)
3546 		goto out;
3547 
3548 	data->arg.label = label;
3549 	status = nfs4_do_create(dir, dentry, data);
3550 
3551 	nfs4_free_createdata(data);
3552 out:
3553 	return status;
3554 }
3555 
3556 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3557 		struct iattr *sattr)
3558 {
3559 	struct nfs4_exception exception = { };
3560 	struct nfs4_label l, *label = NULL;
3561 	int err;
3562 
3563 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3564 
3565 	sattr->ia_mode &= ~current_umask();
3566 	do {
3567 		err = nfs4_handle_exception(NFS_SERVER(dir),
3568 				_nfs4_proc_mkdir(dir, dentry, sattr, label),
3569 				&exception);
3570 	} while (exception.retry);
3571 	nfs4_label_release_security(label);
3572 
3573 	return err;
3574 }
3575 
3576 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3577 		u64 cookie, struct page **pages, unsigned int count, int plus)
3578 {
3579 	struct inode		*dir = dentry->d_inode;
3580 	struct nfs4_readdir_arg args = {
3581 		.fh = NFS_FH(dir),
3582 		.pages = pages,
3583 		.pgbase = 0,
3584 		.count = count,
3585 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3586 		.plus = plus,
3587 	};
3588 	struct nfs4_readdir_res res;
3589 	struct rpc_message msg = {
3590 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3591 		.rpc_argp = &args,
3592 		.rpc_resp = &res,
3593 		.rpc_cred = cred,
3594 	};
3595 	int			status;
3596 
3597 	dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3598 			dentry->d_parent->d_name.name,
3599 			dentry->d_name.name,
3600 			(unsigned long long)cookie);
3601 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3602 	res.pgbase = args.pgbase;
3603 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3604 	if (status >= 0) {
3605 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3606 		status += args.pgbase;
3607 	}
3608 
3609 	nfs_invalidate_atime(dir);
3610 
3611 	dprintk("%s: returns %d\n", __func__, status);
3612 	return status;
3613 }
3614 
3615 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3616 		u64 cookie, struct page **pages, unsigned int count, int plus)
3617 {
3618 	struct nfs4_exception exception = { };
3619 	int err;
3620 	do {
3621 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3622 				_nfs4_proc_readdir(dentry, cred, cookie,
3623 					pages, count, plus),
3624 				&exception);
3625 	} while (exception.retry);
3626 	return err;
3627 }
3628 
3629 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3630 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3631 {
3632 	struct nfs4_createdata *data;
3633 	int mode = sattr->ia_mode;
3634 	int status = -ENOMEM;
3635 
3636 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3637 	if (data == NULL)
3638 		goto out;
3639 
3640 	if (S_ISFIFO(mode))
3641 		data->arg.ftype = NF4FIFO;
3642 	else if (S_ISBLK(mode)) {
3643 		data->arg.ftype = NF4BLK;
3644 		data->arg.u.device.specdata1 = MAJOR(rdev);
3645 		data->arg.u.device.specdata2 = MINOR(rdev);
3646 	}
3647 	else if (S_ISCHR(mode)) {
3648 		data->arg.ftype = NF4CHR;
3649 		data->arg.u.device.specdata1 = MAJOR(rdev);
3650 		data->arg.u.device.specdata2 = MINOR(rdev);
3651 	} else if (!S_ISSOCK(mode)) {
3652 		status = -EINVAL;
3653 		goto out_free;
3654 	}
3655 
3656 	data->arg.label = label;
3657 	status = nfs4_do_create(dir, dentry, data);
3658 out_free:
3659 	nfs4_free_createdata(data);
3660 out:
3661 	return status;
3662 }
3663 
3664 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3665 		struct iattr *sattr, dev_t rdev)
3666 {
3667 	struct nfs4_exception exception = { };
3668 	struct nfs4_label l, *label = NULL;
3669 	int err;
3670 
3671 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3672 
3673 	sattr->ia_mode &= ~current_umask();
3674 	do {
3675 		err = nfs4_handle_exception(NFS_SERVER(dir),
3676 				_nfs4_proc_mknod(dir, dentry, sattr, label, rdev),
3677 				&exception);
3678 	} while (exception.retry);
3679 
3680 	nfs4_label_release_security(label);
3681 
3682 	return err;
3683 }
3684 
3685 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3686 		 struct nfs_fsstat *fsstat)
3687 {
3688 	struct nfs4_statfs_arg args = {
3689 		.fh = fhandle,
3690 		.bitmask = server->attr_bitmask,
3691 	};
3692 	struct nfs4_statfs_res res = {
3693 		.fsstat = fsstat,
3694 	};
3695 	struct rpc_message msg = {
3696 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3697 		.rpc_argp = &args,
3698 		.rpc_resp = &res,
3699 	};
3700 
3701 	nfs_fattr_init(fsstat->fattr);
3702 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3703 }
3704 
3705 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3706 {
3707 	struct nfs4_exception exception = { };
3708 	int err;
3709 	do {
3710 		err = nfs4_handle_exception(server,
3711 				_nfs4_proc_statfs(server, fhandle, fsstat),
3712 				&exception);
3713 	} while (exception.retry);
3714 	return err;
3715 }
3716 
3717 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3718 		struct nfs_fsinfo *fsinfo)
3719 {
3720 	struct nfs4_fsinfo_arg args = {
3721 		.fh = fhandle,
3722 		.bitmask = server->attr_bitmask,
3723 	};
3724 	struct nfs4_fsinfo_res res = {
3725 		.fsinfo = fsinfo,
3726 	};
3727 	struct rpc_message msg = {
3728 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3729 		.rpc_argp = &args,
3730 		.rpc_resp = &res,
3731 	};
3732 
3733 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3734 }
3735 
3736 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3737 {
3738 	struct nfs4_exception exception = { };
3739 	unsigned long now = jiffies;
3740 	int err;
3741 
3742 	do {
3743 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
3744 		if (err == 0) {
3745 			struct nfs_client *clp = server->nfs_client;
3746 
3747 			spin_lock(&clp->cl_lock);
3748 			clp->cl_lease_time = fsinfo->lease_time * HZ;
3749 			clp->cl_last_renewal = now;
3750 			spin_unlock(&clp->cl_lock);
3751 			break;
3752 		}
3753 		err = nfs4_handle_exception(server, err, &exception);
3754 	} while (exception.retry);
3755 	return err;
3756 }
3757 
3758 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3759 {
3760 	int error;
3761 
3762 	nfs_fattr_init(fsinfo->fattr);
3763 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3764 	if (error == 0) {
3765 		/* block layout checks this! */
3766 		server->pnfs_blksize = fsinfo->blksize;
3767 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3768 	}
3769 
3770 	return error;
3771 }
3772 
3773 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3774 		struct nfs_pathconf *pathconf)
3775 {
3776 	struct nfs4_pathconf_arg args = {
3777 		.fh = fhandle,
3778 		.bitmask = server->attr_bitmask,
3779 	};
3780 	struct nfs4_pathconf_res res = {
3781 		.pathconf = pathconf,
3782 	};
3783 	struct rpc_message msg = {
3784 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3785 		.rpc_argp = &args,
3786 		.rpc_resp = &res,
3787 	};
3788 
3789 	/* None of the pathconf attributes are mandatory to implement */
3790 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3791 		memset(pathconf, 0, sizeof(*pathconf));
3792 		return 0;
3793 	}
3794 
3795 	nfs_fattr_init(pathconf->fattr);
3796 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3797 }
3798 
3799 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3800 		struct nfs_pathconf *pathconf)
3801 {
3802 	struct nfs4_exception exception = { };
3803 	int err;
3804 
3805 	do {
3806 		err = nfs4_handle_exception(server,
3807 				_nfs4_proc_pathconf(server, fhandle, pathconf),
3808 				&exception);
3809 	} while (exception.retry);
3810 	return err;
3811 }
3812 
3813 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
3814 		const struct nfs_open_context *ctx,
3815 		const struct nfs_lock_context *l_ctx,
3816 		fmode_t fmode)
3817 {
3818 	const struct nfs_lockowner *lockowner = NULL;
3819 
3820 	if (l_ctx != NULL)
3821 		lockowner = &l_ctx->lockowner;
3822 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
3823 }
3824 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
3825 
3826 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
3827 		const struct nfs_open_context *ctx,
3828 		const struct nfs_lock_context *l_ctx,
3829 		fmode_t fmode)
3830 {
3831 	nfs4_stateid current_stateid;
3832 
3833 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode))
3834 		return false;
3835 	return nfs4_stateid_match(stateid, &current_stateid);
3836 }
3837 
3838 static bool nfs4_error_stateid_expired(int err)
3839 {
3840 	switch (err) {
3841 	case -NFS4ERR_DELEG_REVOKED:
3842 	case -NFS4ERR_ADMIN_REVOKED:
3843 	case -NFS4ERR_BAD_STATEID:
3844 	case -NFS4ERR_STALE_STATEID:
3845 	case -NFS4ERR_OLD_STATEID:
3846 	case -NFS4ERR_OPENMODE:
3847 	case -NFS4ERR_EXPIRED:
3848 		return true;
3849 	}
3850 	return false;
3851 }
3852 
3853 void __nfs4_read_done_cb(struct nfs_read_data *data)
3854 {
3855 	nfs_invalidate_atime(data->header->inode);
3856 }
3857 
3858 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3859 {
3860 	struct nfs_server *server = NFS_SERVER(data->header->inode);
3861 
3862 	if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3863 		rpc_restart_call_prepare(task);
3864 		return -EAGAIN;
3865 	}
3866 
3867 	__nfs4_read_done_cb(data);
3868 	if (task->tk_status > 0)
3869 		renew_lease(server, data->timestamp);
3870 	return 0;
3871 }
3872 
3873 static bool nfs4_read_stateid_changed(struct rpc_task *task,
3874 		struct nfs_readargs *args)
3875 {
3876 
3877 	if (!nfs4_error_stateid_expired(task->tk_status) ||
3878 		nfs4_stateid_is_current(&args->stateid,
3879 				args->context,
3880 				args->lock_context,
3881 				FMODE_READ))
3882 		return false;
3883 	rpc_restart_call_prepare(task);
3884 	return true;
3885 }
3886 
3887 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3888 {
3889 
3890 	dprintk("--> %s\n", __func__);
3891 
3892 	if (!nfs4_sequence_done(task, &data->res.seq_res))
3893 		return -EAGAIN;
3894 	if (nfs4_read_stateid_changed(task, &data->args))
3895 		return -EAGAIN;
3896 	return data->read_done_cb ? data->read_done_cb(task, data) :
3897 				    nfs4_read_done_cb(task, data);
3898 }
3899 
3900 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3901 {
3902 	data->timestamp   = jiffies;
3903 	data->read_done_cb = nfs4_read_done_cb;
3904 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3905 	nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3906 }
3907 
3908 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3909 {
3910 	if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3911 			&data->args.seq_args,
3912 			&data->res.seq_res,
3913 			task))
3914 		return;
3915 	nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3916 			data->args.lock_context, FMODE_READ);
3917 }
3918 
3919 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3920 {
3921 	struct inode *inode = data->header->inode;
3922 
3923 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3924 		rpc_restart_call_prepare(task);
3925 		return -EAGAIN;
3926 	}
3927 	if (task->tk_status >= 0) {
3928 		renew_lease(NFS_SERVER(inode), data->timestamp);
3929 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3930 	}
3931 	return 0;
3932 }
3933 
3934 static bool nfs4_write_stateid_changed(struct rpc_task *task,
3935 		struct nfs_writeargs *args)
3936 {
3937 
3938 	if (!nfs4_error_stateid_expired(task->tk_status) ||
3939 		nfs4_stateid_is_current(&args->stateid,
3940 				args->context,
3941 				args->lock_context,
3942 				FMODE_WRITE))
3943 		return false;
3944 	rpc_restart_call_prepare(task);
3945 	return true;
3946 }
3947 
3948 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3949 {
3950 	if (!nfs4_sequence_done(task, &data->res.seq_res))
3951 		return -EAGAIN;
3952 	if (nfs4_write_stateid_changed(task, &data->args))
3953 		return -EAGAIN;
3954 	return data->write_done_cb ? data->write_done_cb(task, data) :
3955 		nfs4_write_done_cb(task, data);
3956 }
3957 
3958 static
3959 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3960 {
3961 	const struct nfs_pgio_header *hdr = data->header;
3962 
3963 	/* Don't request attributes for pNFS or O_DIRECT writes */
3964 	if (data->ds_clp != NULL || hdr->dreq != NULL)
3965 		return false;
3966 	/* Otherwise, request attributes if and only if we don't hold
3967 	 * a delegation
3968 	 */
3969 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3970 }
3971 
3972 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3973 {
3974 	struct nfs_server *server = NFS_SERVER(data->header->inode);
3975 
3976 	if (!nfs4_write_need_cache_consistency_data(data)) {
3977 		data->args.bitmask = NULL;
3978 		data->res.fattr = NULL;
3979 	} else
3980 		data->args.bitmask = server->cache_consistency_bitmask;
3981 
3982 	if (!data->write_done_cb)
3983 		data->write_done_cb = nfs4_write_done_cb;
3984 	data->res.server = server;
3985 	data->timestamp   = jiffies;
3986 
3987 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3988 	nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3989 }
3990 
3991 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3992 {
3993 	if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3994 			&data->args.seq_args,
3995 			&data->res.seq_res,
3996 			task))
3997 		return;
3998 	nfs4_set_rw_stateid(&data->args.stateid, data->args.context,
3999 			data->args.lock_context, FMODE_WRITE);
4000 }
4001 
4002 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4003 {
4004 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4005 			&data->args.seq_args,
4006 			&data->res.seq_res,
4007 			task);
4008 }
4009 
4010 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4011 {
4012 	struct inode *inode = data->inode;
4013 
4014 	if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
4015 		rpc_restart_call_prepare(task);
4016 		return -EAGAIN;
4017 	}
4018 	return 0;
4019 }
4020 
4021 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4022 {
4023 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4024 		return -EAGAIN;
4025 	return data->commit_done_cb(task, data);
4026 }
4027 
4028 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4029 {
4030 	struct nfs_server *server = NFS_SERVER(data->inode);
4031 
4032 	if (data->commit_done_cb == NULL)
4033 		data->commit_done_cb = nfs4_commit_done_cb;
4034 	data->res.server = server;
4035 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4036 	nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4037 }
4038 
4039 struct nfs4_renewdata {
4040 	struct nfs_client	*client;
4041 	unsigned long		timestamp;
4042 };
4043 
4044 /*
4045  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4046  * standalone procedure for queueing an asynchronous RENEW.
4047  */
4048 static void nfs4_renew_release(void *calldata)
4049 {
4050 	struct nfs4_renewdata *data = calldata;
4051 	struct nfs_client *clp = data->client;
4052 
4053 	if (atomic_read(&clp->cl_count) > 1)
4054 		nfs4_schedule_state_renewal(clp);
4055 	nfs_put_client(clp);
4056 	kfree(data);
4057 }
4058 
4059 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4060 {
4061 	struct nfs4_renewdata *data = calldata;
4062 	struct nfs_client *clp = data->client;
4063 	unsigned long timestamp = data->timestamp;
4064 
4065 	if (task->tk_status < 0) {
4066 		/* Unless we're shutting down, schedule state recovery! */
4067 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4068 			return;
4069 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4070 			nfs4_schedule_lease_recovery(clp);
4071 			return;
4072 		}
4073 		nfs4_schedule_path_down_recovery(clp);
4074 	}
4075 	do_renew_lease(clp, timestamp);
4076 }
4077 
4078 static const struct rpc_call_ops nfs4_renew_ops = {
4079 	.rpc_call_done = nfs4_renew_done,
4080 	.rpc_release = nfs4_renew_release,
4081 };
4082 
4083 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4084 {
4085 	struct rpc_message msg = {
4086 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4087 		.rpc_argp	= clp,
4088 		.rpc_cred	= cred,
4089 	};
4090 	struct nfs4_renewdata *data;
4091 
4092 	if (renew_flags == 0)
4093 		return 0;
4094 	if (!atomic_inc_not_zero(&clp->cl_count))
4095 		return -EIO;
4096 	data = kmalloc(sizeof(*data), GFP_NOFS);
4097 	if (data == NULL)
4098 		return -ENOMEM;
4099 	data->client = clp;
4100 	data->timestamp = jiffies;
4101 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4102 			&nfs4_renew_ops, data);
4103 }
4104 
4105 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4106 {
4107 	struct rpc_message msg = {
4108 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4109 		.rpc_argp	= clp,
4110 		.rpc_cred	= cred,
4111 	};
4112 	unsigned long now = jiffies;
4113 	int status;
4114 
4115 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4116 	if (status < 0)
4117 		return status;
4118 	do_renew_lease(clp, now);
4119 	return 0;
4120 }
4121 
4122 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4123 {
4124 	return (server->caps & NFS_CAP_ACLS)
4125 		&& (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
4126 		&& (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
4127 }
4128 
4129 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4130  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4131  * the stack.
4132  */
4133 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4134 
4135 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4136 		struct page **pages, unsigned int *pgbase)
4137 {
4138 	struct page *newpage, **spages;
4139 	int rc = 0;
4140 	size_t len;
4141 	spages = pages;
4142 
4143 	do {
4144 		len = min_t(size_t, PAGE_SIZE, buflen);
4145 		newpage = alloc_page(GFP_KERNEL);
4146 
4147 		if (newpage == NULL)
4148 			goto unwind;
4149 		memcpy(page_address(newpage), buf, len);
4150                 buf += len;
4151                 buflen -= len;
4152 		*pages++ = newpage;
4153 		rc++;
4154 	} while (buflen != 0);
4155 
4156 	return rc;
4157 
4158 unwind:
4159 	for(; rc > 0; rc--)
4160 		__free_page(spages[rc-1]);
4161 	return -ENOMEM;
4162 }
4163 
4164 struct nfs4_cached_acl {
4165 	int cached;
4166 	size_t len;
4167 	char data[0];
4168 };
4169 
4170 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4171 {
4172 	struct nfs_inode *nfsi = NFS_I(inode);
4173 
4174 	spin_lock(&inode->i_lock);
4175 	kfree(nfsi->nfs4_acl);
4176 	nfsi->nfs4_acl = acl;
4177 	spin_unlock(&inode->i_lock);
4178 }
4179 
4180 static void nfs4_zap_acl_attr(struct inode *inode)
4181 {
4182 	nfs4_set_cached_acl(inode, NULL);
4183 }
4184 
4185 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4186 {
4187 	struct nfs_inode *nfsi = NFS_I(inode);
4188 	struct nfs4_cached_acl *acl;
4189 	int ret = -ENOENT;
4190 
4191 	spin_lock(&inode->i_lock);
4192 	acl = nfsi->nfs4_acl;
4193 	if (acl == NULL)
4194 		goto out;
4195 	if (buf == NULL) /* user is just asking for length */
4196 		goto out_len;
4197 	if (acl->cached == 0)
4198 		goto out;
4199 	ret = -ERANGE; /* see getxattr(2) man page */
4200 	if (acl->len > buflen)
4201 		goto out;
4202 	memcpy(buf, acl->data, acl->len);
4203 out_len:
4204 	ret = acl->len;
4205 out:
4206 	spin_unlock(&inode->i_lock);
4207 	return ret;
4208 }
4209 
4210 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4211 {
4212 	struct nfs4_cached_acl *acl;
4213 	size_t buflen = sizeof(*acl) + acl_len;
4214 
4215 	if (buflen <= PAGE_SIZE) {
4216 		acl = kmalloc(buflen, GFP_KERNEL);
4217 		if (acl == NULL)
4218 			goto out;
4219 		acl->cached = 1;
4220 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4221 	} else {
4222 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4223 		if (acl == NULL)
4224 			goto out;
4225 		acl->cached = 0;
4226 	}
4227 	acl->len = acl_len;
4228 out:
4229 	nfs4_set_cached_acl(inode, acl);
4230 }
4231 
4232 /*
4233  * The getxattr API returns the required buffer length when called with a
4234  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4235  * the required buf.  On a NULL buf, we send a page of data to the server
4236  * guessing that the ACL request can be serviced by a page. If so, we cache
4237  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4238  * the cache. If not so, we throw away the page, and cache the required
4239  * length. The next getxattr call will then produce another round trip to
4240  * the server, this time with the input buf of the required size.
4241  */
4242 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4243 {
4244 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4245 	struct nfs_getaclargs args = {
4246 		.fh = NFS_FH(inode),
4247 		.acl_pages = pages,
4248 		.acl_len = buflen,
4249 	};
4250 	struct nfs_getaclres res = {
4251 		.acl_len = buflen,
4252 	};
4253 	struct rpc_message msg = {
4254 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4255 		.rpc_argp = &args,
4256 		.rpc_resp = &res,
4257 	};
4258 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4259 	int ret = -ENOMEM, i;
4260 
4261 	/* As long as we're doing a round trip to the server anyway,
4262 	 * let's be prepared for a page of acl data. */
4263 	if (npages == 0)
4264 		npages = 1;
4265 	if (npages > ARRAY_SIZE(pages))
4266 		return -ERANGE;
4267 
4268 	for (i = 0; i < npages; i++) {
4269 		pages[i] = alloc_page(GFP_KERNEL);
4270 		if (!pages[i])
4271 			goto out_free;
4272 	}
4273 
4274 	/* for decoding across pages */
4275 	res.acl_scratch = alloc_page(GFP_KERNEL);
4276 	if (!res.acl_scratch)
4277 		goto out_free;
4278 
4279 	args.acl_len = npages * PAGE_SIZE;
4280 	args.acl_pgbase = 0;
4281 
4282 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4283 		__func__, buf, buflen, npages, args.acl_len);
4284 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4285 			     &msg, &args.seq_args, &res.seq_res, 0);
4286 	if (ret)
4287 		goto out_free;
4288 
4289 	/* Handle the case where the passed-in buffer is too short */
4290 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4291 		/* Did the user only issue a request for the acl length? */
4292 		if (buf == NULL)
4293 			goto out_ok;
4294 		ret = -ERANGE;
4295 		goto out_free;
4296 	}
4297 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4298 	if (buf) {
4299 		if (res.acl_len > buflen) {
4300 			ret = -ERANGE;
4301 			goto out_free;
4302 		}
4303 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4304 	}
4305 out_ok:
4306 	ret = res.acl_len;
4307 out_free:
4308 	for (i = 0; i < npages; i++)
4309 		if (pages[i])
4310 			__free_page(pages[i]);
4311 	if (res.acl_scratch)
4312 		__free_page(res.acl_scratch);
4313 	return ret;
4314 }
4315 
4316 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4317 {
4318 	struct nfs4_exception exception = { };
4319 	ssize_t ret;
4320 	do {
4321 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4322 		if (ret >= 0)
4323 			break;
4324 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4325 	} while (exception.retry);
4326 	return ret;
4327 }
4328 
4329 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4330 {
4331 	struct nfs_server *server = NFS_SERVER(inode);
4332 	int ret;
4333 
4334 	if (!nfs4_server_supports_acls(server))
4335 		return -EOPNOTSUPP;
4336 	ret = nfs_revalidate_inode(server, inode);
4337 	if (ret < 0)
4338 		return ret;
4339 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4340 		nfs_zap_acl_cache(inode);
4341 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4342 	if (ret != -ENOENT)
4343 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4344 		 * but no cached acl data, just the acl length */
4345 		return ret;
4346 	return nfs4_get_acl_uncached(inode, buf, buflen);
4347 }
4348 
4349 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4350 {
4351 	struct nfs_server *server = NFS_SERVER(inode);
4352 	struct page *pages[NFS4ACL_MAXPAGES];
4353 	struct nfs_setaclargs arg = {
4354 		.fh		= NFS_FH(inode),
4355 		.acl_pages	= pages,
4356 		.acl_len	= buflen,
4357 	};
4358 	struct nfs_setaclres res;
4359 	struct rpc_message msg = {
4360 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4361 		.rpc_argp	= &arg,
4362 		.rpc_resp	= &res,
4363 	};
4364 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4365 	int ret, i;
4366 
4367 	if (!nfs4_server_supports_acls(server))
4368 		return -EOPNOTSUPP;
4369 	if (npages > ARRAY_SIZE(pages))
4370 		return -ERANGE;
4371 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4372 	if (i < 0)
4373 		return i;
4374 	nfs4_inode_return_delegation(inode);
4375 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4376 
4377 	/*
4378 	 * Free each page after tx, so the only ref left is
4379 	 * held by the network stack
4380 	 */
4381 	for (; i > 0; i--)
4382 		put_page(pages[i-1]);
4383 
4384 	/*
4385 	 * Acl update can result in inode attribute update.
4386 	 * so mark the attribute cache invalid.
4387 	 */
4388 	spin_lock(&inode->i_lock);
4389 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4390 	spin_unlock(&inode->i_lock);
4391 	nfs_access_zap_cache(inode);
4392 	nfs_zap_acl_cache(inode);
4393 	return ret;
4394 }
4395 
4396 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4397 {
4398 	struct nfs4_exception exception = { };
4399 	int err;
4400 	do {
4401 		err = nfs4_handle_exception(NFS_SERVER(inode),
4402 				__nfs4_proc_set_acl(inode, buf, buflen),
4403 				&exception);
4404 	} while (exception.retry);
4405 	return err;
4406 }
4407 
4408 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4409 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4410 					size_t buflen)
4411 {
4412 	struct nfs_server *server = NFS_SERVER(inode);
4413 	struct nfs_fattr fattr;
4414 	struct nfs4_label label = {0, 0, buflen, buf};
4415 
4416 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4417 	struct nfs4_getattr_arg args = {
4418 		.fh		= NFS_FH(inode),
4419 		.bitmask	= bitmask,
4420 	};
4421 	struct nfs4_getattr_res res = {
4422 		.fattr		= &fattr,
4423 		.label		= &label,
4424 		.server		= server,
4425 	};
4426 	struct rpc_message msg = {
4427 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4428 		.rpc_argp	= &args,
4429 		.rpc_resp	= &res,
4430 	};
4431 	int ret;
4432 
4433 	nfs_fattr_init(&fattr);
4434 
4435 	ret = rpc_call_sync(server->client, &msg, 0);
4436 	if (ret)
4437 		return ret;
4438 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4439 		return -ENOENT;
4440 	if (buflen < label.len)
4441 		return -ERANGE;
4442 	return 0;
4443 }
4444 
4445 static int nfs4_get_security_label(struct inode *inode, void *buf,
4446 					size_t buflen)
4447 {
4448 	struct nfs4_exception exception = { };
4449 	int err;
4450 
4451 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4452 		return -EOPNOTSUPP;
4453 
4454 	do {
4455 		err = nfs4_handle_exception(NFS_SERVER(inode),
4456 				_nfs4_get_security_label(inode, buf, buflen),
4457 				&exception);
4458 	} while (exception.retry);
4459 	return err;
4460 }
4461 
4462 static int _nfs4_do_set_security_label(struct inode *inode,
4463 		struct nfs4_label *ilabel,
4464 		struct nfs_fattr *fattr,
4465 		struct nfs4_label *olabel)
4466 {
4467 
4468 	struct iattr sattr = {0};
4469 	struct nfs_server *server = NFS_SERVER(inode);
4470 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4471 	struct nfs_setattrargs args = {
4472 		.fh             = NFS_FH(inode),
4473 		.iap            = &sattr,
4474 		.server		= server,
4475 		.bitmask	= bitmask,
4476 		.label		= ilabel,
4477 	};
4478 	struct nfs_setattrres res = {
4479 		.fattr		= fattr,
4480 		.label		= olabel,
4481 		.server		= server,
4482 	};
4483 	struct rpc_message msg = {
4484 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4485 		.rpc_argp       = &args,
4486 		.rpc_resp       = &res,
4487 	};
4488 	int status;
4489 
4490 	nfs4_stateid_copy(&args.stateid, &zero_stateid);
4491 
4492 	status = rpc_call_sync(server->client, &msg, 0);
4493 	if (status)
4494 		dprintk("%s failed: %d\n", __func__, status);
4495 
4496 	return status;
4497 }
4498 
4499 static int nfs4_do_set_security_label(struct inode *inode,
4500 		struct nfs4_label *ilabel,
4501 		struct nfs_fattr *fattr,
4502 		struct nfs4_label *olabel)
4503 {
4504 	struct nfs4_exception exception = { };
4505 	int err;
4506 
4507 	do {
4508 		err = nfs4_handle_exception(NFS_SERVER(inode),
4509 				_nfs4_do_set_security_label(inode, ilabel,
4510 				fattr, olabel),
4511 				&exception);
4512 	} while (exception.retry);
4513 	return err;
4514 }
4515 
4516 static int
4517 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4518 {
4519 	struct nfs4_label ilabel, *olabel = NULL;
4520 	struct nfs_fattr fattr;
4521 	struct rpc_cred *cred;
4522 	struct inode *inode = dentry->d_inode;
4523 	int status;
4524 
4525 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4526 		return -EOPNOTSUPP;
4527 
4528 	nfs_fattr_init(&fattr);
4529 
4530 	ilabel.pi = 0;
4531 	ilabel.lfs = 0;
4532 	ilabel.label = (char *)buf;
4533 	ilabel.len = buflen;
4534 
4535 	cred = rpc_lookup_cred();
4536 	if (IS_ERR(cred))
4537 		return PTR_ERR(cred);
4538 
4539 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4540 	if (IS_ERR(olabel)) {
4541 		status = -PTR_ERR(olabel);
4542 		goto out;
4543 	}
4544 
4545 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4546 	if (status == 0)
4547 		nfs_setsecurity(inode, &fattr, olabel);
4548 
4549 	nfs4_label_free(olabel);
4550 out:
4551 	put_rpccred(cred);
4552 	return status;
4553 }
4554 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4555 
4556 
4557 static int
4558 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
4559 {
4560 	struct nfs_client *clp = server->nfs_client;
4561 
4562 	if (task->tk_status >= 0)
4563 		return 0;
4564 	switch(task->tk_status) {
4565 		case -NFS4ERR_DELEG_REVOKED:
4566 		case -NFS4ERR_ADMIN_REVOKED:
4567 		case -NFS4ERR_BAD_STATEID:
4568 			if (state == NULL)
4569 				break;
4570 			nfs_remove_bad_delegation(state->inode);
4571 		case -NFS4ERR_OPENMODE:
4572 			if (state == NULL)
4573 				break;
4574 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4575 				goto stateid_invalid;
4576 			goto wait_on_recovery;
4577 		case -NFS4ERR_EXPIRED:
4578 			if (state != NULL) {
4579 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4580 					goto stateid_invalid;
4581 			}
4582 		case -NFS4ERR_STALE_STATEID:
4583 		case -NFS4ERR_STALE_CLIENTID:
4584 			nfs4_schedule_lease_recovery(clp);
4585 			goto wait_on_recovery;
4586 #if defined(CONFIG_NFS_V4_1)
4587 		case -NFS4ERR_BADSESSION:
4588 		case -NFS4ERR_BADSLOT:
4589 		case -NFS4ERR_BAD_HIGH_SLOT:
4590 		case -NFS4ERR_DEADSESSION:
4591 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4592 		case -NFS4ERR_SEQ_FALSE_RETRY:
4593 		case -NFS4ERR_SEQ_MISORDERED:
4594 			dprintk("%s ERROR %d, Reset session\n", __func__,
4595 				task->tk_status);
4596 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4597 			task->tk_status = 0;
4598 			return -EAGAIN;
4599 #endif /* CONFIG_NFS_V4_1 */
4600 		case -NFS4ERR_DELAY:
4601 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4602 		case -NFS4ERR_GRACE:
4603 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4604 			task->tk_status = 0;
4605 			return -EAGAIN;
4606 		case -NFS4ERR_RETRY_UNCACHED_REP:
4607 		case -NFS4ERR_OLD_STATEID:
4608 			task->tk_status = 0;
4609 			return -EAGAIN;
4610 	}
4611 	task->tk_status = nfs4_map_errors(task->tk_status);
4612 	return 0;
4613 stateid_invalid:
4614 	task->tk_status = -EIO;
4615 	return 0;
4616 wait_on_recovery:
4617 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4618 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4619 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4620 	task->tk_status = 0;
4621 	return -EAGAIN;
4622 }
4623 
4624 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4625 				    nfs4_verifier *bootverf)
4626 {
4627 	__be32 verf[2];
4628 
4629 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4630 		/* An impossible timestamp guarantees this value
4631 		 * will never match a generated boot time. */
4632 		verf[0] = 0;
4633 		verf[1] = (__be32)(NSEC_PER_SEC + 1);
4634 	} else {
4635 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4636 		verf[0] = (__be32)nn->boot_time.tv_sec;
4637 		verf[1] = (__be32)nn->boot_time.tv_nsec;
4638 	}
4639 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4640 }
4641 
4642 static unsigned int
4643 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4644 				   char *buf, size_t len)
4645 {
4646 	unsigned int result;
4647 
4648 	rcu_read_lock();
4649 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4650 				clp->cl_ipaddr,
4651 				rpc_peeraddr2str(clp->cl_rpcclient,
4652 							RPC_DISPLAY_ADDR),
4653 				rpc_peeraddr2str(clp->cl_rpcclient,
4654 							RPC_DISPLAY_PROTO));
4655 	rcu_read_unlock();
4656 	return result;
4657 }
4658 
4659 static unsigned int
4660 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4661 				char *buf, size_t len)
4662 {
4663 	char *nodename = clp->cl_rpcclient->cl_nodename;
4664 
4665 	if (nfs4_client_id_uniquifier[0] != '\0')
4666 		nodename = nfs4_client_id_uniquifier;
4667 	return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4668 				clp->rpc_ops->version, clp->cl_minorversion,
4669 				nodename);
4670 }
4671 
4672 /**
4673  * nfs4_proc_setclientid - Negotiate client ID
4674  * @clp: state data structure
4675  * @program: RPC program for NFSv4 callback service
4676  * @port: IP port number for NFS4 callback service
4677  * @cred: RPC credential to use for this call
4678  * @res: where to place the result
4679  *
4680  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4681  */
4682 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4683 		unsigned short port, struct rpc_cred *cred,
4684 		struct nfs4_setclientid_res *res)
4685 {
4686 	nfs4_verifier sc_verifier;
4687 	struct nfs4_setclientid setclientid = {
4688 		.sc_verifier = &sc_verifier,
4689 		.sc_prog = program,
4690 		.sc_cb_ident = clp->cl_cb_ident,
4691 	};
4692 	struct rpc_message msg = {
4693 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4694 		.rpc_argp = &setclientid,
4695 		.rpc_resp = res,
4696 		.rpc_cred = cred,
4697 	};
4698 	int status;
4699 
4700 	/* nfs_client_id4 */
4701 	nfs4_init_boot_verifier(clp, &sc_verifier);
4702 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4703 		setclientid.sc_name_len =
4704 				nfs4_init_uniform_client_string(clp,
4705 						setclientid.sc_name,
4706 						sizeof(setclientid.sc_name));
4707 	else
4708 		setclientid.sc_name_len =
4709 				nfs4_init_nonuniform_client_string(clp,
4710 						setclientid.sc_name,
4711 						sizeof(setclientid.sc_name));
4712 	/* cb_client4 */
4713 	rcu_read_lock();
4714 	setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4715 				sizeof(setclientid.sc_netid), "%s",
4716 				rpc_peeraddr2str(clp->cl_rpcclient,
4717 							RPC_DISPLAY_NETID));
4718 	rcu_read_unlock();
4719 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4720 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4721 				clp->cl_ipaddr, port >> 8, port & 255);
4722 
4723 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4724 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
4725 		setclientid.sc_name_len, setclientid.sc_name);
4726 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4727 	dprintk("NFS reply setclientid: %d\n", status);
4728 	return status;
4729 }
4730 
4731 /**
4732  * nfs4_proc_setclientid_confirm - Confirm client ID
4733  * @clp: state data structure
4734  * @res: result of a previous SETCLIENTID
4735  * @cred: RPC credential to use for this call
4736  *
4737  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4738  */
4739 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4740 		struct nfs4_setclientid_res *arg,
4741 		struct rpc_cred *cred)
4742 {
4743 	struct rpc_message msg = {
4744 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4745 		.rpc_argp = arg,
4746 		.rpc_cred = cred,
4747 	};
4748 	int status;
4749 
4750 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4751 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
4752 		clp->cl_clientid);
4753 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4754 	dprintk("NFS reply setclientid_confirm: %d\n", status);
4755 	return status;
4756 }
4757 
4758 struct nfs4_delegreturndata {
4759 	struct nfs4_delegreturnargs args;
4760 	struct nfs4_delegreturnres res;
4761 	struct nfs_fh fh;
4762 	nfs4_stateid stateid;
4763 	unsigned long timestamp;
4764 	struct nfs_fattr fattr;
4765 	int rpc_status;
4766 };
4767 
4768 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4769 {
4770 	struct nfs4_delegreturndata *data = calldata;
4771 
4772 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4773 		return;
4774 
4775 	switch (task->tk_status) {
4776 	case -NFS4ERR_STALE_STATEID:
4777 	case -NFS4ERR_EXPIRED:
4778 	case 0:
4779 		renew_lease(data->res.server, data->timestamp);
4780 		break;
4781 	default:
4782 		if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4783 				-EAGAIN) {
4784 			rpc_restart_call_prepare(task);
4785 			return;
4786 		}
4787 	}
4788 	data->rpc_status = task->tk_status;
4789 }
4790 
4791 static void nfs4_delegreturn_release(void *calldata)
4792 {
4793 	kfree(calldata);
4794 }
4795 
4796 #if defined(CONFIG_NFS_V4_1)
4797 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4798 {
4799 	struct nfs4_delegreturndata *d_data;
4800 
4801 	d_data = (struct nfs4_delegreturndata *)data;
4802 
4803 	nfs4_setup_sequence(d_data->res.server,
4804 			&d_data->args.seq_args,
4805 			&d_data->res.seq_res,
4806 			task);
4807 }
4808 #endif /* CONFIG_NFS_V4_1 */
4809 
4810 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4811 #if defined(CONFIG_NFS_V4_1)
4812 	.rpc_call_prepare = nfs4_delegreturn_prepare,
4813 #endif /* CONFIG_NFS_V4_1 */
4814 	.rpc_call_done = nfs4_delegreturn_done,
4815 	.rpc_release = nfs4_delegreturn_release,
4816 };
4817 
4818 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4819 {
4820 	struct nfs4_delegreturndata *data;
4821 	struct nfs_server *server = NFS_SERVER(inode);
4822 	struct rpc_task *task;
4823 	struct rpc_message msg = {
4824 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4825 		.rpc_cred = cred,
4826 	};
4827 	struct rpc_task_setup task_setup_data = {
4828 		.rpc_client = server->client,
4829 		.rpc_message = &msg,
4830 		.callback_ops = &nfs4_delegreturn_ops,
4831 		.flags = RPC_TASK_ASYNC,
4832 	};
4833 	int status = 0;
4834 
4835 	data = kzalloc(sizeof(*data), GFP_NOFS);
4836 	if (data == NULL)
4837 		return -ENOMEM;
4838 	nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4839 	data->args.fhandle = &data->fh;
4840 	data->args.stateid = &data->stateid;
4841 	data->args.bitmask = server->cache_consistency_bitmask;
4842 	nfs_copy_fh(&data->fh, NFS_FH(inode));
4843 	nfs4_stateid_copy(&data->stateid, stateid);
4844 	data->res.fattr = &data->fattr;
4845 	data->res.server = server;
4846 	nfs_fattr_init(data->res.fattr);
4847 	data->timestamp = jiffies;
4848 	data->rpc_status = 0;
4849 
4850 	task_setup_data.callback_data = data;
4851 	msg.rpc_argp = &data->args;
4852 	msg.rpc_resp = &data->res;
4853 	task = rpc_run_task(&task_setup_data);
4854 	if (IS_ERR(task))
4855 		return PTR_ERR(task);
4856 	if (!issync)
4857 		goto out;
4858 	status = nfs4_wait_for_completion_rpc_task(task);
4859 	if (status != 0)
4860 		goto out;
4861 	status = data->rpc_status;
4862 	if (status == 0)
4863 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4864 	else
4865 		nfs_refresh_inode(inode, &data->fattr);
4866 out:
4867 	rpc_put_task(task);
4868 	return status;
4869 }
4870 
4871 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4872 {
4873 	struct nfs_server *server = NFS_SERVER(inode);
4874 	struct nfs4_exception exception = { };
4875 	int err;
4876 	do {
4877 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4878 		switch (err) {
4879 			case -NFS4ERR_STALE_STATEID:
4880 			case -NFS4ERR_EXPIRED:
4881 			case 0:
4882 				return 0;
4883 		}
4884 		err = nfs4_handle_exception(server, err, &exception);
4885 	} while (exception.retry);
4886 	return err;
4887 }
4888 
4889 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4890 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4891 
4892 /*
4893  * sleep, with exponential backoff, and retry the LOCK operation.
4894  */
4895 static unsigned long
4896 nfs4_set_lock_task_retry(unsigned long timeout)
4897 {
4898 	freezable_schedule_timeout_killable_unsafe(timeout);
4899 	timeout <<= 1;
4900 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
4901 		return NFS4_LOCK_MAXTIMEOUT;
4902 	return timeout;
4903 }
4904 
4905 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4906 {
4907 	struct inode *inode = state->inode;
4908 	struct nfs_server *server = NFS_SERVER(inode);
4909 	struct nfs_client *clp = server->nfs_client;
4910 	struct nfs_lockt_args arg = {
4911 		.fh = NFS_FH(inode),
4912 		.fl = request,
4913 	};
4914 	struct nfs_lockt_res res = {
4915 		.denied = request,
4916 	};
4917 	struct rpc_message msg = {
4918 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4919 		.rpc_argp       = &arg,
4920 		.rpc_resp       = &res,
4921 		.rpc_cred	= state->owner->so_cred,
4922 	};
4923 	struct nfs4_lock_state *lsp;
4924 	int status;
4925 
4926 	arg.lock_owner.clientid = clp->cl_clientid;
4927 	status = nfs4_set_lock_state(state, request);
4928 	if (status != 0)
4929 		goto out;
4930 	lsp = request->fl_u.nfs4_fl.owner;
4931 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
4932 	arg.lock_owner.s_dev = server->s_dev;
4933 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4934 	switch (status) {
4935 		case 0:
4936 			request->fl_type = F_UNLCK;
4937 			break;
4938 		case -NFS4ERR_DENIED:
4939 			status = 0;
4940 	}
4941 	request->fl_ops->fl_release_private(request);
4942 out:
4943 	return status;
4944 }
4945 
4946 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4947 {
4948 	struct nfs4_exception exception = { };
4949 	int err;
4950 
4951 	do {
4952 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
4953 				_nfs4_proc_getlk(state, cmd, request),
4954 				&exception);
4955 	} while (exception.retry);
4956 	return err;
4957 }
4958 
4959 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4960 {
4961 	int res = 0;
4962 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4963 		case FL_POSIX:
4964 			res = posix_lock_file_wait(file, fl);
4965 			break;
4966 		case FL_FLOCK:
4967 			res = flock_lock_file_wait(file, fl);
4968 			break;
4969 		default:
4970 			BUG();
4971 	}
4972 	return res;
4973 }
4974 
4975 struct nfs4_unlockdata {
4976 	struct nfs_locku_args arg;
4977 	struct nfs_locku_res res;
4978 	struct nfs4_lock_state *lsp;
4979 	struct nfs_open_context *ctx;
4980 	struct file_lock fl;
4981 	const struct nfs_server *server;
4982 	unsigned long timestamp;
4983 };
4984 
4985 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4986 		struct nfs_open_context *ctx,
4987 		struct nfs4_lock_state *lsp,
4988 		struct nfs_seqid *seqid)
4989 {
4990 	struct nfs4_unlockdata *p;
4991 	struct inode *inode = lsp->ls_state->inode;
4992 
4993 	p = kzalloc(sizeof(*p), GFP_NOFS);
4994 	if (p == NULL)
4995 		return NULL;
4996 	p->arg.fh = NFS_FH(inode);
4997 	p->arg.fl = &p->fl;
4998 	p->arg.seqid = seqid;
4999 	p->res.seqid = seqid;
5000 	p->arg.stateid = &lsp->ls_stateid;
5001 	p->lsp = lsp;
5002 	atomic_inc(&lsp->ls_count);
5003 	/* Ensure we don't close file until we're done freeing locks! */
5004 	p->ctx = get_nfs_open_context(ctx);
5005 	memcpy(&p->fl, fl, sizeof(p->fl));
5006 	p->server = NFS_SERVER(inode);
5007 	return p;
5008 }
5009 
5010 static void nfs4_locku_release_calldata(void *data)
5011 {
5012 	struct nfs4_unlockdata *calldata = data;
5013 	nfs_free_seqid(calldata->arg.seqid);
5014 	nfs4_put_lock_state(calldata->lsp);
5015 	put_nfs_open_context(calldata->ctx);
5016 	kfree(calldata);
5017 }
5018 
5019 static void nfs4_locku_done(struct rpc_task *task, void *data)
5020 {
5021 	struct nfs4_unlockdata *calldata = data;
5022 
5023 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5024 		return;
5025 	switch (task->tk_status) {
5026 		case 0:
5027 			nfs4_stateid_copy(&calldata->lsp->ls_stateid,
5028 					&calldata->res.stateid);
5029 			renew_lease(calldata->server, calldata->timestamp);
5030 			break;
5031 		case -NFS4ERR_BAD_STATEID:
5032 		case -NFS4ERR_OLD_STATEID:
5033 		case -NFS4ERR_STALE_STATEID:
5034 		case -NFS4ERR_EXPIRED:
5035 			break;
5036 		default:
5037 			if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
5038 				rpc_restart_call_prepare(task);
5039 	}
5040 	nfs_release_seqid(calldata->arg.seqid);
5041 }
5042 
5043 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5044 {
5045 	struct nfs4_unlockdata *calldata = data;
5046 
5047 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5048 		goto out_wait;
5049 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5050 		/* Note: exit _without_ running nfs4_locku_done */
5051 		goto out_no_action;
5052 	}
5053 	calldata->timestamp = jiffies;
5054 	if (nfs4_setup_sequence(calldata->server,
5055 				&calldata->arg.seq_args,
5056 				&calldata->res.seq_res,
5057 				task) != 0)
5058 		nfs_release_seqid(calldata->arg.seqid);
5059 	return;
5060 out_no_action:
5061 	task->tk_action = NULL;
5062 out_wait:
5063 	nfs4_sequence_done(task, &calldata->res.seq_res);
5064 }
5065 
5066 static const struct rpc_call_ops nfs4_locku_ops = {
5067 	.rpc_call_prepare = nfs4_locku_prepare,
5068 	.rpc_call_done = nfs4_locku_done,
5069 	.rpc_release = nfs4_locku_release_calldata,
5070 };
5071 
5072 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5073 		struct nfs_open_context *ctx,
5074 		struct nfs4_lock_state *lsp,
5075 		struct nfs_seqid *seqid)
5076 {
5077 	struct nfs4_unlockdata *data;
5078 	struct rpc_message msg = {
5079 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5080 		.rpc_cred = ctx->cred,
5081 	};
5082 	struct rpc_task_setup task_setup_data = {
5083 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5084 		.rpc_message = &msg,
5085 		.callback_ops = &nfs4_locku_ops,
5086 		.workqueue = nfsiod_workqueue,
5087 		.flags = RPC_TASK_ASYNC,
5088 	};
5089 
5090 	/* Ensure this is an unlock - when canceling a lock, the
5091 	 * canceled lock is passed in, and it won't be an unlock.
5092 	 */
5093 	fl->fl_type = F_UNLCK;
5094 
5095 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5096 	if (data == NULL) {
5097 		nfs_free_seqid(seqid);
5098 		return ERR_PTR(-ENOMEM);
5099 	}
5100 
5101 	nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5102 	msg.rpc_argp = &data->arg;
5103 	msg.rpc_resp = &data->res;
5104 	task_setup_data.callback_data = data;
5105 	return rpc_run_task(&task_setup_data);
5106 }
5107 
5108 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5109 {
5110 	struct inode *inode = state->inode;
5111 	struct nfs4_state_owner *sp = state->owner;
5112 	struct nfs_inode *nfsi = NFS_I(inode);
5113 	struct nfs_seqid *seqid;
5114 	struct nfs4_lock_state *lsp;
5115 	struct rpc_task *task;
5116 	int status = 0;
5117 	unsigned char fl_flags = request->fl_flags;
5118 
5119 	status = nfs4_set_lock_state(state, request);
5120 	/* Unlock _before_ we do the RPC call */
5121 	request->fl_flags |= FL_EXISTS;
5122 	/* Exclude nfs_delegation_claim_locks() */
5123 	mutex_lock(&sp->so_delegreturn_mutex);
5124 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5125 	down_read(&nfsi->rwsem);
5126 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5127 		up_read(&nfsi->rwsem);
5128 		mutex_unlock(&sp->so_delegreturn_mutex);
5129 		goto out;
5130 	}
5131 	up_read(&nfsi->rwsem);
5132 	mutex_unlock(&sp->so_delegreturn_mutex);
5133 	if (status != 0)
5134 		goto out;
5135 	/* Is this a delegated lock? */
5136 	lsp = request->fl_u.nfs4_fl.owner;
5137 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5138 		goto out;
5139 	seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5140 	status = -ENOMEM;
5141 	if (seqid == NULL)
5142 		goto out;
5143 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5144 	status = PTR_ERR(task);
5145 	if (IS_ERR(task))
5146 		goto out;
5147 	status = nfs4_wait_for_completion_rpc_task(task);
5148 	rpc_put_task(task);
5149 out:
5150 	request->fl_flags = fl_flags;
5151 	return status;
5152 }
5153 
5154 struct nfs4_lockdata {
5155 	struct nfs_lock_args arg;
5156 	struct nfs_lock_res res;
5157 	struct nfs4_lock_state *lsp;
5158 	struct nfs_open_context *ctx;
5159 	struct file_lock fl;
5160 	unsigned long timestamp;
5161 	int rpc_status;
5162 	int cancelled;
5163 	struct nfs_server *server;
5164 };
5165 
5166 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5167 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5168 		gfp_t gfp_mask)
5169 {
5170 	struct nfs4_lockdata *p;
5171 	struct inode *inode = lsp->ls_state->inode;
5172 	struct nfs_server *server = NFS_SERVER(inode);
5173 
5174 	p = kzalloc(sizeof(*p), gfp_mask);
5175 	if (p == NULL)
5176 		return NULL;
5177 
5178 	p->arg.fh = NFS_FH(inode);
5179 	p->arg.fl = &p->fl;
5180 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5181 	if (p->arg.open_seqid == NULL)
5182 		goto out_free;
5183 	p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
5184 	if (p->arg.lock_seqid == NULL)
5185 		goto out_free_seqid;
5186 	p->arg.lock_stateid = &lsp->ls_stateid;
5187 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5188 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5189 	p->arg.lock_owner.s_dev = server->s_dev;
5190 	p->res.lock_seqid = p->arg.lock_seqid;
5191 	p->lsp = lsp;
5192 	p->server = server;
5193 	atomic_inc(&lsp->ls_count);
5194 	p->ctx = get_nfs_open_context(ctx);
5195 	memcpy(&p->fl, fl, sizeof(p->fl));
5196 	return p;
5197 out_free_seqid:
5198 	nfs_free_seqid(p->arg.open_seqid);
5199 out_free:
5200 	kfree(p);
5201 	return NULL;
5202 }
5203 
5204 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5205 {
5206 	struct nfs4_lockdata *data = calldata;
5207 	struct nfs4_state *state = data->lsp->ls_state;
5208 
5209 	dprintk("%s: begin!\n", __func__);
5210 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5211 		goto out_wait;
5212 	/* Do we need to do an open_to_lock_owner? */
5213 	if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
5214 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5215 			goto out_release_lock_seqid;
5216 		}
5217 		data->arg.open_stateid = &state->open_stateid;
5218 		data->arg.new_lock_owner = 1;
5219 		data->res.open_seqid = data->arg.open_seqid;
5220 	} else
5221 		data->arg.new_lock_owner = 0;
5222 	if (!nfs4_valid_open_stateid(state)) {
5223 		data->rpc_status = -EBADF;
5224 		task->tk_action = NULL;
5225 		goto out_release_open_seqid;
5226 	}
5227 	data->timestamp = jiffies;
5228 	if (nfs4_setup_sequence(data->server,
5229 				&data->arg.seq_args,
5230 				&data->res.seq_res,
5231 				task) == 0)
5232 		return;
5233 out_release_open_seqid:
5234 	nfs_release_seqid(data->arg.open_seqid);
5235 out_release_lock_seqid:
5236 	nfs_release_seqid(data->arg.lock_seqid);
5237 out_wait:
5238 	nfs4_sequence_done(task, &data->res.seq_res);
5239 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5240 }
5241 
5242 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5243 {
5244 	struct nfs4_lockdata *data = calldata;
5245 
5246 	dprintk("%s: begin!\n", __func__);
5247 
5248 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5249 		return;
5250 
5251 	data->rpc_status = task->tk_status;
5252 	if (data->arg.new_lock_owner != 0) {
5253 		if (data->rpc_status == 0)
5254 			nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
5255 		else
5256 			goto out;
5257 	}
5258 	if (data->rpc_status == 0) {
5259 		nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
5260 		set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
5261 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
5262 	}
5263 out:
5264 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5265 }
5266 
5267 static void nfs4_lock_release(void *calldata)
5268 {
5269 	struct nfs4_lockdata *data = calldata;
5270 
5271 	dprintk("%s: begin!\n", __func__);
5272 	nfs_free_seqid(data->arg.open_seqid);
5273 	if (data->cancelled != 0) {
5274 		struct rpc_task *task;
5275 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5276 				data->arg.lock_seqid);
5277 		if (!IS_ERR(task))
5278 			rpc_put_task_async(task);
5279 		dprintk("%s: cancelling lock!\n", __func__);
5280 	} else
5281 		nfs_free_seqid(data->arg.lock_seqid);
5282 	nfs4_put_lock_state(data->lsp);
5283 	put_nfs_open_context(data->ctx);
5284 	kfree(data);
5285 	dprintk("%s: done!\n", __func__);
5286 }
5287 
5288 static const struct rpc_call_ops nfs4_lock_ops = {
5289 	.rpc_call_prepare = nfs4_lock_prepare,
5290 	.rpc_call_done = nfs4_lock_done,
5291 	.rpc_release = nfs4_lock_release,
5292 };
5293 
5294 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5295 {
5296 	switch (error) {
5297 	case -NFS4ERR_ADMIN_REVOKED:
5298 	case -NFS4ERR_BAD_STATEID:
5299 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5300 		if (new_lock_owner != 0 ||
5301 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5302 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5303 		break;
5304 	case -NFS4ERR_STALE_STATEID:
5305 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5306 	case -NFS4ERR_EXPIRED:
5307 		nfs4_schedule_lease_recovery(server->nfs_client);
5308 	};
5309 }
5310 
5311 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5312 {
5313 	struct nfs4_lockdata *data;
5314 	struct rpc_task *task;
5315 	struct rpc_message msg = {
5316 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5317 		.rpc_cred = state->owner->so_cred,
5318 	};
5319 	struct rpc_task_setup task_setup_data = {
5320 		.rpc_client = NFS_CLIENT(state->inode),
5321 		.rpc_message = &msg,
5322 		.callback_ops = &nfs4_lock_ops,
5323 		.workqueue = nfsiod_workqueue,
5324 		.flags = RPC_TASK_ASYNC,
5325 	};
5326 	int ret;
5327 
5328 	dprintk("%s: begin!\n", __func__);
5329 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5330 			fl->fl_u.nfs4_fl.owner,
5331 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5332 	if (data == NULL)
5333 		return -ENOMEM;
5334 	if (IS_SETLKW(cmd))
5335 		data->arg.block = 1;
5336 	nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5337 	msg.rpc_argp = &data->arg;
5338 	msg.rpc_resp = &data->res;
5339 	task_setup_data.callback_data = data;
5340 	if (recovery_type > NFS_LOCK_NEW) {
5341 		if (recovery_type == NFS_LOCK_RECLAIM)
5342 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5343 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5344 	}
5345 	task = rpc_run_task(&task_setup_data);
5346 	if (IS_ERR(task))
5347 		return PTR_ERR(task);
5348 	ret = nfs4_wait_for_completion_rpc_task(task);
5349 	if (ret == 0) {
5350 		ret = data->rpc_status;
5351 		if (ret)
5352 			nfs4_handle_setlk_error(data->server, data->lsp,
5353 					data->arg.new_lock_owner, ret);
5354 	} else
5355 		data->cancelled = 1;
5356 	rpc_put_task(task);
5357 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5358 	return ret;
5359 }
5360 
5361 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5362 {
5363 	struct nfs_server *server = NFS_SERVER(state->inode);
5364 	struct nfs4_exception exception = {
5365 		.inode = state->inode,
5366 	};
5367 	int err;
5368 
5369 	do {
5370 		/* Cache the lock if possible... */
5371 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5372 			return 0;
5373 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5374 		if (err != -NFS4ERR_DELAY)
5375 			break;
5376 		nfs4_handle_exception(server, err, &exception);
5377 	} while (exception.retry);
5378 	return err;
5379 }
5380 
5381 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5382 {
5383 	struct nfs_server *server = NFS_SERVER(state->inode);
5384 	struct nfs4_exception exception = {
5385 		.inode = state->inode,
5386 	};
5387 	int err;
5388 
5389 	err = nfs4_set_lock_state(state, request);
5390 	if (err != 0)
5391 		return err;
5392 	do {
5393 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5394 			return 0;
5395 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5396 		switch (err) {
5397 		default:
5398 			goto out;
5399 		case -NFS4ERR_GRACE:
5400 		case -NFS4ERR_DELAY:
5401 			nfs4_handle_exception(server, err, &exception);
5402 			err = 0;
5403 		}
5404 	} while (exception.retry);
5405 out:
5406 	return err;
5407 }
5408 
5409 #if defined(CONFIG_NFS_V4_1)
5410 /**
5411  * nfs41_check_expired_locks - possibly free a lock stateid
5412  *
5413  * @state: NFSv4 state for an inode
5414  *
5415  * Returns NFS_OK if recovery for this stateid is now finished.
5416  * Otherwise a negative NFS4ERR value is returned.
5417  */
5418 static int nfs41_check_expired_locks(struct nfs4_state *state)
5419 {
5420 	int status, ret = -NFS4ERR_BAD_STATEID;
5421 	struct nfs4_lock_state *lsp;
5422 	struct nfs_server *server = NFS_SERVER(state->inode);
5423 
5424 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5425 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5426 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5427 
5428 			status = nfs41_test_stateid(server,
5429 					&lsp->ls_stateid,
5430 					cred);
5431 			if (status != NFS_OK) {
5432 				/* Free the stateid unless the server
5433 				 * informs us the stateid is unrecognized. */
5434 				if (status != -NFS4ERR_BAD_STATEID)
5435 					nfs41_free_stateid(server,
5436 							&lsp->ls_stateid,
5437 							cred);
5438 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5439 				ret = status;
5440 			}
5441 		}
5442 	};
5443 
5444 	return ret;
5445 }
5446 
5447 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5448 {
5449 	int status = NFS_OK;
5450 
5451 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5452 		status = nfs41_check_expired_locks(state);
5453 	if (status != NFS_OK)
5454 		status = nfs4_lock_expired(state, request);
5455 	return status;
5456 }
5457 #endif
5458 
5459 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5460 {
5461 	struct nfs4_state_owner *sp = state->owner;
5462 	struct nfs_inode *nfsi = NFS_I(state->inode);
5463 	unsigned char fl_flags = request->fl_flags;
5464 	unsigned int seq;
5465 	int status = -ENOLCK;
5466 
5467 	if ((fl_flags & FL_POSIX) &&
5468 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5469 		goto out;
5470 	/* Is this a delegated open? */
5471 	status = nfs4_set_lock_state(state, request);
5472 	if (status != 0)
5473 		goto out;
5474 	request->fl_flags |= FL_ACCESS;
5475 	status = do_vfs_lock(request->fl_file, request);
5476 	if (status < 0)
5477 		goto out;
5478 	down_read(&nfsi->rwsem);
5479 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5480 		/* Yes: cache locks! */
5481 		/* ...but avoid races with delegation recall... */
5482 		request->fl_flags = fl_flags & ~FL_SLEEP;
5483 		status = do_vfs_lock(request->fl_file, request);
5484 		goto out_unlock;
5485 	}
5486 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
5487 	up_read(&nfsi->rwsem);
5488 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5489 	if (status != 0)
5490 		goto out;
5491 	down_read(&nfsi->rwsem);
5492 	if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) {
5493 		status = -NFS4ERR_DELAY;
5494 		goto out_unlock;
5495 	}
5496 	/* Note: we always want to sleep here! */
5497 	request->fl_flags = fl_flags | FL_SLEEP;
5498 	if (do_vfs_lock(request->fl_file, request) < 0)
5499 		printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
5500 			"manager!\n", __func__);
5501 out_unlock:
5502 	up_read(&nfsi->rwsem);
5503 out:
5504 	request->fl_flags = fl_flags;
5505 	return status;
5506 }
5507 
5508 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5509 {
5510 	struct nfs4_exception exception = {
5511 		.state = state,
5512 		.inode = state->inode,
5513 	};
5514 	int err;
5515 
5516 	do {
5517 		err = _nfs4_proc_setlk(state, cmd, request);
5518 		if (err == -NFS4ERR_DENIED)
5519 			err = -EAGAIN;
5520 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5521 				err, &exception);
5522 	} while (exception.retry);
5523 	return err;
5524 }
5525 
5526 static int
5527 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5528 {
5529 	struct nfs_open_context *ctx;
5530 	struct nfs4_state *state;
5531 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5532 	int status;
5533 
5534 	/* verify open state */
5535 	ctx = nfs_file_open_context(filp);
5536 	state = ctx->state;
5537 
5538 	if (request->fl_start < 0 || request->fl_end < 0)
5539 		return -EINVAL;
5540 
5541 	if (IS_GETLK(cmd)) {
5542 		if (state != NULL)
5543 			return nfs4_proc_getlk(state, F_GETLK, request);
5544 		return 0;
5545 	}
5546 
5547 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5548 		return -EINVAL;
5549 
5550 	if (request->fl_type == F_UNLCK) {
5551 		if (state != NULL)
5552 			return nfs4_proc_unlck(state, cmd, request);
5553 		return 0;
5554 	}
5555 
5556 	if (state == NULL)
5557 		return -ENOLCK;
5558 	/*
5559 	 * Don't rely on the VFS having checked the file open mode,
5560 	 * since it won't do this for flock() locks.
5561 	 */
5562 	switch (request->fl_type) {
5563 	case F_RDLCK:
5564 		if (!(filp->f_mode & FMODE_READ))
5565 			return -EBADF;
5566 		break;
5567 	case F_WRLCK:
5568 		if (!(filp->f_mode & FMODE_WRITE))
5569 			return -EBADF;
5570 	}
5571 
5572 	do {
5573 		status = nfs4_proc_setlk(state, cmd, request);
5574 		if ((status != -EAGAIN) || IS_SETLK(cmd))
5575 			break;
5576 		timeout = nfs4_set_lock_task_retry(timeout);
5577 		status = -ERESTARTSYS;
5578 		if (signalled())
5579 			break;
5580 	} while(status < 0);
5581 	return status;
5582 }
5583 
5584 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
5585 {
5586 	struct nfs_server *server = NFS_SERVER(state->inode);
5587 	int err;
5588 
5589 	err = nfs4_set_lock_state(state, fl);
5590 	if (err != 0)
5591 		return err;
5592 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
5593 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
5594 }
5595 
5596 struct nfs_release_lockowner_data {
5597 	struct nfs4_lock_state *lsp;
5598 	struct nfs_server *server;
5599 	struct nfs_release_lockowner_args args;
5600 };
5601 
5602 static void nfs4_release_lockowner_release(void *calldata)
5603 {
5604 	struct nfs_release_lockowner_data *data = calldata;
5605 	nfs4_free_lock_state(data->server, data->lsp);
5606 	kfree(calldata);
5607 }
5608 
5609 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5610 	.rpc_release = nfs4_release_lockowner_release,
5611 };
5612 
5613 static int nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
5614 {
5615 	struct nfs_release_lockowner_data *data;
5616 	struct rpc_message msg = {
5617 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5618 	};
5619 
5620 	if (server->nfs_client->cl_mvops->minor_version != 0)
5621 		return -EINVAL;
5622 	data = kmalloc(sizeof(*data), GFP_NOFS);
5623 	if (!data)
5624 		return -ENOMEM;
5625 	data->lsp = lsp;
5626 	data->server = server;
5627 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5628 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5629 	data->args.lock_owner.s_dev = server->s_dev;
5630 	msg.rpc_argp = &data->args;
5631 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5632 	return 0;
5633 }
5634 
5635 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5636 
5637 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5638 				   const void *buf, size_t buflen,
5639 				   int flags, int type)
5640 {
5641 	if (strcmp(key, "") != 0)
5642 		return -EINVAL;
5643 
5644 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5645 }
5646 
5647 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5648 				   void *buf, size_t buflen, int type)
5649 {
5650 	if (strcmp(key, "") != 0)
5651 		return -EINVAL;
5652 
5653 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5654 }
5655 
5656 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5657 				       size_t list_len, const char *name,
5658 				       size_t name_len, int type)
5659 {
5660 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5661 
5662 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5663 		return 0;
5664 
5665 	if (list && len <= list_len)
5666 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5667 	return len;
5668 }
5669 
5670 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5671 static inline int nfs4_server_supports_labels(struct nfs_server *server)
5672 {
5673 	return server->caps & NFS_CAP_SECURITY_LABEL;
5674 }
5675 
5676 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
5677 				   const void *buf, size_t buflen,
5678 				   int flags, int type)
5679 {
5680 	if (security_ismaclabel(key))
5681 		return nfs4_set_security_label(dentry, buf, buflen);
5682 
5683 	return -EOPNOTSUPP;
5684 }
5685 
5686 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
5687 				   void *buf, size_t buflen, int type)
5688 {
5689 	if (security_ismaclabel(key))
5690 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
5691 	return -EOPNOTSUPP;
5692 }
5693 
5694 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
5695 				       size_t list_len, const char *name,
5696 				       size_t name_len, int type)
5697 {
5698 	size_t len = 0;
5699 
5700 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
5701 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
5702 		if (list && len <= list_len)
5703 			security_inode_listsecurity(dentry->d_inode, list, len);
5704 	}
5705 	return len;
5706 }
5707 
5708 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
5709 	.prefix = XATTR_SECURITY_PREFIX,
5710 	.list	= nfs4_xattr_list_nfs4_label,
5711 	.get	= nfs4_xattr_get_nfs4_label,
5712 	.set	= nfs4_xattr_set_nfs4_label,
5713 };
5714 #endif
5715 
5716 
5717 /*
5718  * nfs_fhget will use either the mounted_on_fileid or the fileid
5719  */
5720 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5721 {
5722 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5723 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5724 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5725 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5726 		return;
5727 
5728 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5729 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5730 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5731 	fattr->nlink = 2;
5732 }
5733 
5734 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5735 				   const struct qstr *name,
5736 				   struct nfs4_fs_locations *fs_locations,
5737 				   struct page *page)
5738 {
5739 	struct nfs_server *server = NFS_SERVER(dir);
5740 	u32 bitmask[3] = {
5741 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5742 	};
5743 	struct nfs4_fs_locations_arg args = {
5744 		.dir_fh = NFS_FH(dir),
5745 		.name = name,
5746 		.page = page,
5747 		.bitmask = bitmask,
5748 	};
5749 	struct nfs4_fs_locations_res res = {
5750 		.fs_locations = fs_locations,
5751 	};
5752 	struct rpc_message msg = {
5753 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5754 		.rpc_argp = &args,
5755 		.rpc_resp = &res,
5756 	};
5757 	int status;
5758 
5759 	dprintk("%s: start\n", __func__);
5760 
5761 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
5762 	 * is not supported */
5763 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5764 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5765 	else
5766 		bitmask[0] |= FATTR4_WORD0_FILEID;
5767 
5768 	nfs_fattr_init(&fs_locations->fattr);
5769 	fs_locations->server = server;
5770 	fs_locations->nlocations = 0;
5771 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5772 	dprintk("%s: returned status = %d\n", __func__, status);
5773 	return status;
5774 }
5775 
5776 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5777 			   const struct qstr *name,
5778 			   struct nfs4_fs_locations *fs_locations,
5779 			   struct page *page)
5780 {
5781 	struct nfs4_exception exception = { };
5782 	int err;
5783 	do {
5784 		err = nfs4_handle_exception(NFS_SERVER(dir),
5785 				_nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5786 				&exception);
5787 	} while (exception.retry);
5788 	return err;
5789 }
5790 
5791 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5792 {
5793 	int status;
5794 	struct nfs4_secinfo_arg args = {
5795 		.dir_fh = NFS_FH(dir),
5796 		.name   = name,
5797 	};
5798 	struct nfs4_secinfo_res res = {
5799 		.flavors     = flavors,
5800 	};
5801 	struct rpc_message msg = {
5802 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5803 		.rpc_argp = &args,
5804 		.rpc_resp = &res,
5805 	};
5806 
5807 	dprintk("NFS call  secinfo %s\n", name->name);
5808 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5809 	dprintk("NFS reply  secinfo: %d\n", status);
5810 	return status;
5811 }
5812 
5813 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5814 		      struct nfs4_secinfo_flavors *flavors)
5815 {
5816 	struct nfs4_exception exception = { };
5817 	int err;
5818 	do {
5819 		err = nfs4_handle_exception(NFS_SERVER(dir),
5820 				_nfs4_proc_secinfo(dir, name, flavors),
5821 				&exception);
5822 	} while (exception.retry);
5823 	return err;
5824 }
5825 
5826 #ifdef CONFIG_NFS_V4_1
5827 /*
5828  * Check the exchange flags returned by the server for invalid flags, having
5829  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5830  * DS flags set.
5831  */
5832 static int nfs4_check_cl_exchange_flags(u32 flags)
5833 {
5834 	if (flags & ~EXCHGID4_FLAG_MASK_R)
5835 		goto out_inval;
5836 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5837 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5838 		goto out_inval;
5839 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5840 		goto out_inval;
5841 	return NFS_OK;
5842 out_inval:
5843 	return -NFS4ERR_INVAL;
5844 }
5845 
5846 static bool
5847 nfs41_same_server_scope(struct nfs41_server_scope *a,
5848 			struct nfs41_server_scope *b)
5849 {
5850 	if (a->server_scope_sz == b->server_scope_sz &&
5851 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5852 		return true;
5853 
5854 	return false;
5855 }
5856 
5857 /*
5858  * nfs4_proc_bind_conn_to_session()
5859  *
5860  * The 4.1 client currently uses the same TCP connection for the
5861  * fore and backchannel.
5862  */
5863 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5864 {
5865 	int status;
5866 	struct nfs41_bind_conn_to_session_res res;
5867 	struct rpc_message msg = {
5868 		.rpc_proc =
5869 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5870 		.rpc_argp = clp,
5871 		.rpc_resp = &res,
5872 		.rpc_cred = cred,
5873 	};
5874 
5875 	dprintk("--> %s\n", __func__);
5876 
5877 	res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5878 	if (unlikely(res.session == NULL)) {
5879 		status = -ENOMEM;
5880 		goto out;
5881 	}
5882 
5883 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5884 	if (status == 0) {
5885 		if (memcmp(res.session->sess_id.data,
5886 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5887 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
5888 			status = -EIO;
5889 			goto out_session;
5890 		}
5891 		if (res.dir != NFS4_CDFS4_BOTH) {
5892 			dprintk("NFS: %s: Unexpected direction from server\n",
5893 				__func__);
5894 			status = -EIO;
5895 			goto out_session;
5896 		}
5897 		if (res.use_conn_in_rdma_mode) {
5898 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
5899 				__func__);
5900 			status = -EIO;
5901 			goto out_session;
5902 		}
5903 	}
5904 out_session:
5905 	kfree(res.session);
5906 out:
5907 	dprintk("<-- %s status= %d\n", __func__, status);
5908 	return status;
5909 }
5910 
5911 /*
5912  * nfs4_proc_exchange_id()
5913  *
5914  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5915  *
5916  * Since the clientid has expired, all compounds using sessions
5917  * associated with the stale clientid will be returning
5918  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5919  * be in some phase of session reset.
5920  */
5921 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5922 {
5923 	nfs4_verifier verifier;
5924 	struct nfs41_exchange_id_args args = {
5925 		.verifier = &verifier,
5926 		.client = clp,
5927 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
5928 			EXCHGID4_FLAG_BIND_PRINC_STATEID,
5929 	};
5930 	struct nfs41_exchange_id_res res = {
5931 		0
5932 	};
5933 	int status;
5934 	struct rpc_message msg = {
5935 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5936 		.rpc_argp = &args,
5937 		.rpc_resp = &res,
5938 		.rpc_cred = cred,
5939 	};
5940 
5941 	nfs4_init_boot_verifier(clp, &verifier);
5942 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5943 							sizeof(args.id));
5944 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5945 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5946 		args.id_len, args.id);
5947 
5948 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5949 					GFP_NOFS);
5950 	if (unlikely(res.server_owner == NULL)) {
5951 		status = -ENOMEM;
5952 		goto out;
5953 	}
5954 
5955 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5956 					GFP_NOFS);
5957 	if (unlikely(res.server_scope == NULL)) {
5958 		status = -ENOMEM;
5959 		goto out_server_owner;
5960 	}
5961 
5962 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5963 	if (unlikely(res.impl_id == NULL)) {
5964 		status = -ENOMEM;
5965 		goto out_server_scope;
5966 	}
5967 
5968 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5969 	if (status == 0)
5970 		status = nfs4_check_cl_exchange_flags(res.flags);
5971 
5972 	if (status == 0) {
5973 		clp->cl_clientid = res.clientid;
5974 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5975 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5976 			clp->cl_seqid = res.seqid;
5977 
5978 		kfree(clp->cl_serverowner);
5979 		clp->cl_serverowner = res.server_owner;
5980 		res.server_owner = NULL;
5981 
5982 		/* use the most recent implementation id */
5983 		kfree(clp->cl_implid);
5984 		clp->cl_implid = res.impl_id;
5985 
5986 		if (clp->cl_serverscope != NULL &&
5987 		    !nfs41_same_server_scope(clp->cl_serverscope,
5988 					     res.server_scope)) {
5989 			dprintk("%s: server_scope mismatch detected\n",
5990 				__func__);
5991 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5992 			kfree(clp->cl_serverscope);
5993 			clp->cl_serverscope = NULL;
5994 		}
5995 
5996 		if (clp->cl_serverscope == NULL) {
5997 			clp->cl_serverscope = res.server_scope;
5998 			goto out;
5999 		}
6000 	} else
6001 		kfree(res.impl_id);
6002 
6003 out_server_owner:
6004 	kfree(res.server_owner);
6005 out_server_scope:
6006 	kfree(res.server_scope);
6007 out:
6008 	if (clp->cl_implid != NULL)
6009 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6010 			"domain: %s, name: %s, date: %llu,%u\n",
6011 			clp->cl_implid->domain, clp->cl_implid->name,
6012 			clp->cl_implid->date.seconds,
6013 			clp->cl_implid->date.nseconds);
6014 	dprintk("NFS reply exchange_id: %d\n", status);
6015 	return status;
6016 }
6017 
6018 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6019 		struct rpc_cred *cred)
6020 {
6021 	struct rpc_message msg = {
6022 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6023 		.rpc_argp = clp,
6024 		.rpc_cred = cred,
6025 	};
6026 	int status;
6027 
6028 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6029 	if (status)
6030 		dprintk("NFS: Got error %d from the server %s on "
6031 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6032 	return status;
6033 }
6034 
6035 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6036 		struct rpc_cred *cred)
6037 {
6038 	unsigned int loop;
6039 	int ret;
6040 
6041 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6042 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6043 		switch (ret) {
6044 		case -NFS4ERR_DELAY:
6045 		case -NFS4ERR_CLIENTID_BUSY:
6046 			ssleep(1);
6047 			break;
6048 		default:
6049 			return ret;
6050 		}
6051 	}
6052 	return 0;
6053 }
6054 
6055 int nfs4_destroy_clientid(struct nfs_client *clp)
6056 {
6057 	struct rpc_cred *cred;
6058 	int ret = 0;
6059 
6060 	if (clp->cl_mvops->minor_version < 1)
6061 		goto out;
6062 	if (clp->cl_exchange_flags == 0)
6063 		goto out;
6064 	if (clp->cl_preserve_clid)
6065 		goto out;
6066 	cred = nfs4_get_exchange_id_cred(clp);
6067 	ret = nfs4_proc_destroy_clientid(clp, cred);
6068 	if (cred)
6069 		put_rpccred(cred);
6070 	switch (ret) {
6071 	case 0:
6072 	case -NFS4ERR_STALE_CLIENTID:
6073 		clp->cl_exchange_flags = 0;
6074 	}
6075 out:
6076 	return ret;
6077 }
6078 
6079 struct nfs4_get_lease_time_data {
6080 	struct nfs4_get_lease_time_args *args;
6081 	struct nfs4_get_lease_time_res *res;
6082 	struct nfs_client *clp;
6083 };
6084 
6085 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
6086 					void *calldata)
6087 {
6088 	struct nfs4_get_lease_time_data *data =
6089 			(struct nfs4_get_lease_time_data *)calldata;
6090 
6091 	dprintk("--> %s\n", __func__);
6092 	/* just setup sequence, do not trigger session recovery
6093 	   since we're invoked within one */
6094 	nfs41_setup_sequence(data->clp->cl_session,
6095 			&data->args->la_seq_args,
6096 			&data->res->lr_seq_res,
6097 			task);
6098 	dprintk("<-- %s\n", __func__);
6099 }
6100 
6101 /*
6102  * Called from nfs4_state_manager thread for session setup, so don't recover
6103  * from sequence operation or clientid errors.
6104  */
6105 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
6106 {
6107 	struct nfs4_get_lease_time_data *data =
6108 			(struct nfs4_get_lease_time_data *)calldata;
6109 
6110 	dprintk("--> %s\n", __func__);
6111 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
6112 		return;
6113 	switch (task->tk_status) {
6114 	case -NFS4ERR_DELAY:
6115 	case -NFS4ERR_GRACE:
6116 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
6117 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
6118 		task->tk_status = 0;
6119 		/* fall through */
6120 	case -NFS4ERR_RETRY_UNCACHED_REP:
6121 		rpc_restart_call_prepare(task);
6122 		return;
6123 	}
6124 	dprintk("<-- %s\n", __func__);
6125 }
6126 
6127 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
6128 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
6129 	.rpc_call_done = nfs4_get_lease_time_done,
6130 };
6131 
6132 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
6133 {
6134 	struct rpc_task *task;
6135 	struct nfs4_get_lease_time_args args;
6136 	struct nfs4_get_lease_time_res res = {
6137 		.lr_fsinfo = fsinfo,
6138 	};
6139 	struct nfs4_get_lease_time_data data = {
6140 		.args = &args,
6141 		.res = &res,
6142 		.clp = clp,
6143 	};
6144 	struct rpc_message msg = {
6145 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
6146 		.rpc_argp = &args,
6147 		.rpc_resp = &res,
6148 	};
6149 	struct rpc_task_setup task_setup = {
6150 		.rpc_client = clp->cl_rpcclient,
6151 		.rpc_message = &msg,
6152 		.callback_ops = &nfs4_get_lease_time_ops,
6153 		.callback_data = &data,
6154 		.flags = RPC_TASK_TIMEOUT,
6155 	};
6156 	int status;
6157 
6158 	nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
6159 	nfs4_set_sequence_privileged(&args.la_seq_args);
6160 	dprintk("--> %s\n", __func__);
6161 	task = rpc_run_task(&task_setup);
6162 
6163 	if (IS_ERR(task))
6164 		status = PTR_ERR(task);
6165 	else {
6166 		status = task->tk_status;
6167 		rpc_put_task(task);
6168 	}
6169 	dprintk("<-- %s return %d\n", __func__, status);
6170 
6171 	return status;
6172 }
6173 
6174 /*
6175  * Initialize the values to be used by the client in CREATE_SESSION
6176  * If nfs4_init_session set the fore channel request and response sizes,
6177  * use them.
6178  *
6179  * Set the back channel max_resp_sz_cached to zero to force the client to
6180  * always set csa_cachethis to FALSE because the current implementation
6181  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
6182  */
6183 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
6184 {
6185 	unsigned int max_rqst_sz, max_resp_sz;
6186 
6187 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
6188 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
6189 
6190 	/* Fore channel attributes */
6191 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
6192 	args->fc_attrs.max_resp_sz = max_resp_sz;
6193 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
6194 	args->fc_attrs.max_reqs = max_session_slots;
6195 
6196 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
6197 		"max_ops=%u max_reqs=%u\n",
6198 		__func__,
6199 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
6200 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
6201 
6202 	/* Back channel attributes */
6203 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
6204 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
6205 	args->bc_attrs.max_resp_sz_cached = 0;
6206 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
6207 	args->bc_attrs.max_reqs = 1;
6208 
6209 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
6210 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
6211 		__func__,
6212 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
6213 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
6214 		args->bc_attrs.max_reqs);
6215 }
6216 
6217 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6218 {
6219 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
6220 	struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
6221 
6222 	if (rcvd->max_resp_sz > sent->max_resp_sz)
6223 		return -EINVAL;
6224 	/*
6225 	 * Our requested max_ops is the minimum we need; we're not
6226 	 * prepared to break up compounds into smaller pieces than that.
6227 	 * So, no point even trying to continue if the server won't
6228 	 * cooperate:
6229 	 */
6230 	if (rcvd->max_ops < sent->max_ops)
6231 		return -EINVAL;
6232 	if (rcvd->max_reqs == 0)
6233 		return -EINVAL;
6234 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
6235 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
6236 	return 0;
6237 }
6238 
6239 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
6240 {
6241 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
6242 	struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
6243 
6244 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
6245 		return -EINVAL;
6246 	if (rcvd->max_resp_sz < sent->max_resp_sz)
6247 		return -EINVAL;
6248 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
6249 		return -EINVAL;
6250 	/* These would render the backchannel useless: */
6251 	if (rcvd->max_ops != sent->max_ops)
6252 		return -EINVAL;
6253 	if (rcvd->max_reqs != sent->max_reqs)
6254 		return -EINVAL;
6255 	return 0;
6256 }
6257 
6258 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
6259 				     struct nfs4_session *session)
6260 {
6261 	int ret;
6262 
6263 	ret = nfs4_verify_fore_channel_attrs(args, session);
6264 	if (ret)
6265 		return ret;
6266 	return nfs4_verify_back_channel_attrs(args, session);
6267 }
6268 
6269 static int _nfs4_proc_create_session(struct nfs_client *clp,
6270 		struct rpc_cred *cred)
6271 {
6272 	struct nfs4_session *session = clp->cl_session;
6273 	struct nfs41_create_session_args args = {
6274 		.client = clp,
6275 		.cb_program = NFS4_CALLBACK,
6276 	};
6277 	struct nfs41_create_session_res res = {
6278 		.client = clp,
6279 	};
6280 	struct rpc_message msg = {
6281 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
6282 		.rpc_argp = &args,
6283 		.rpc_resp = &res,
6284 		.rpc_cred = cred,
6285 	};
6286 	int status;
6287 
6288 	nfs4_init_channel_attrs(&args);
6289 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
6290 
6291 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6292 
6293 	if (!status) {
6294 		/* Verify the session's negotiated channel_attrs values */
6295 		status = nfs4_verify_channel_attrs(&args, session);
6296 		/* Increment the clientid slot sequence id */
6297 		clp->cl_seqid++;
6298 	}
6299 
6300 	return status;
6301 }
6302 
6303 /*
6304  * Issues a CREATE_SESSION operation to the server.
6305  * It is the responsibility of the caller to verify the session is
6306  * expired before calling this routine.
6307  */
6308 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
6309 {
6310 	int status;
6311 	unsigned *ptr;
6312 	struct nfs4_session *session = clp->cl_session;
6313 
6314 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
6315 
6316 	status = _nfs4_proc_create_session(clp, cred);
6317 	if (status)
6318 		goto out;
6319 
6320 	/* Init or reset the session slot tables */
6321 	status = nfs4_setup_session_slot_tables(session);
6322 	dprintk("slot table setup returned %d\n", status);
6323 	if (status)
6324 		goto out;
6325 
6326 	ptr = (unsigned *)&session->sess_id.data[0];
6327 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
6328 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
6329 out:
6330 	dprintk("<-- %s\n", __func__);
6331 	return status;
6332 }
6333 
6334 /*
6335  * Issue the over-the-wire RPC DESTROY_SESSION.
6336  * The caller must serialize access to this routine.
6337  */
6338 int nfs4_proc_destroy_session(struct nfs4_session *session,
6339 		struct rpc_cred *cred)
6340 {
6341 	struct rpc_message msg = {
6342 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
6343 		.rpc_argp = session,
6344 		.rpc_cred = cred,
6345 	};
6346 	int status = 0;
6347 
6348 	dprintk("--> nfs4_proc_destroy_session\n");
6349 
6350 	/* session is still being setup */
6351 	if (session->clp->cl_cons_state != NFS_CS_READY)
6352 		return status;
6353 
6354 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6355 
6356 	if (status)
6357 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
6358 			"Session has been destroyed regardless...\n", status);
6359 
6360 	dprintk("<-- nfs4_proc_destroy_session\n");
6361 	return status;
6362 }
6363 
6364 /*
6365  * Renew the cl_session lease.
6366  */
6367 struct nfs4_sequence_data {
6368 	struct nfs_client *clp;
6369 	struct nfs4_sequence_args args;
6370 	struct nfs4_sequence_res res;
6371 };
6372 
6373 static void nfs41_sequence_release(void *data)
6374 {
6375 	struct nfs4_sequence_data *calldata = data;
6376 	struct nfs_client *clp = calldata->clp;
6377 
6378 	if (atomic_read(&clp->cl_count) > 1)
6379 		nfs4_schedule_state_renewal(clp);
6380 	nfs_put_client(clp);
6381 	kfree(calldata);
6382 }
6383 
6384 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6385 {
6386 	switch(task->tk_status) {
6387 	case -NFS4ERR_DELAY:
6388 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
6389 		return -EAGAIN;
6390 	default:
6391 		nfs4_schedule_lease_recovery(clp);
6392 	}
6393 	return 0;
6394 }
6395 
6396 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6397 {
6398 	struct nfs4_sequence_data *calldata = data;
6399 	struct nfs_client *clp = calldata->clp;
6400 
6401 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6402 		return;
6403 
6404 	if (task->tk_status < 0) {
6405 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
6406 		if (atomic_read(&clp->cl_count) == 1)
6407 			goto out;
6408 
6409 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6410 			rpc_restart_call_prepare(task);
6411 			return;
6412 		}
6413 	}
6414 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6415 out:
6416 	dprintk("<-- %s\n", __func__);
6417 }
6418 
6419 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6420 {
6421 	struct nfs4_sequence_data *calldata = data;
6422 	struct nfs_client *clp = calldata->clp;
6423 	struct nfs4_sequence_args *args;
6424 	struct nfs4_sequence_res *res;
6425 
6426 	args = task->tk_msg.rpc_argp;
6427 	res = task->tk_msg.rpc_resp;
6428 
6429 	nfs41_setup_sequence(clp->cl_session, args, res, task);
6430 }
6431 
6432 static const struct rpc_call_ops nfs41_sequence_ops = {
6433 	.rpc_call_done = nfs41_sequence_call_done,
6434 	.rpc_call_prepare = nfs41_sequence_prepare,
6435 	.rpc_release = nfs41_sequence_release,
6436 };
6437 
6438 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
6439 		struct rpc_cred *cred,
6440 		bool is_privileged)
6441 {
6442 	struct nfs4_sequence_data *calldata;
6443 	struct rpc_message msg = {
6444 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6445 		.rpc_cred = cred,
6446 	};
6447 	struct rpc_task_setup task_setup_data = {
6448 		.rpc_client = clp->cl_rpcclient,
6449 		.rpc_message = &msg,
6450 		.callback_ops = &nfs41_sequence_ops,
6451 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6452 	};
6453 
6454 	if (!atomic_inc_not_zero(&clp->cl_count))
6455 		return ERR_PTR(-EIO);
6456 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6457 	if (calldata == NULL) {
6458 		nfs_put_client(clp);
6459 		return ERR_PTR(-ENOMEM);
6460 	}
6461 	nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6462 	if (is_privileged)
6463 		nfs4_set_sequence_privileged(&calldata->args);
6464 	msg.rpc_argp = &calldata->args;
6465 	msg.rpc_resp = &calldata->res;
6466 	calldata->clp = clp;
6467 	task_setup_data.callback_data = calldata;
6468 
6469 	return rpc_run_task(&task_setup_data);
6470 }
6471 
6472 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6473 {
6474 	struct rpc_task *task;
6475 	int ret = 0;
6476 
6477 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6478 		return 0;
6479 	task = _nfs41_proc_sequence(clp, cred, false);
6480 	if (IS_ERR(task))
6481 		ret = PTR_ERR(task);
6482 	else
6483 		rpc_put_task_async(task);
6484 	dprintk("<-- %s status=%d\n", __func__, ret);
6485 	return ret;
6486 }
6487 
6488 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6489 {
6490 	struct rpc_task *task;
6491 	int ret;
6492 
6493 	task = _nfs41_proc_sequence(clp, cred, true);
6494 	if (IS_ERR(task)) {
6495 		ret = PTR_ERR(task);
6496 		goto out;
6497 	}
6498 	ret = rpc_wait_for_completion_task(task);
6499 	if (!ret) {
6500 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6501 
6502 		if (task->tk_status == 0)
6503 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6504 		ret = task->tk_status;
6505 	}
6506 	rpc_put_task(task);
6507 out:
6508 	dprintk("<-- %s status=%d\n", __func__, ret);
6509 	return ret;
6510 }
6511 
6512 struct nfs4_reclaim_complete_data {
6513 	struct nfs_client *clp;
6514 	struct nfs41_reclaim_complete_args arg;
6515 	struct nfs41_reclaim_complete_res res;
6516 };
6517 
6518 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6519 {
6520 	struct nfs4_reclaim_complete_data *calldata = data;
6521 
6522 	nfs41_setup_sequence(calldata->clp->cl_session,
6523 			&calldata->arg.seq_args,
6524 			&calldata->res.seq_res,
6525 			task);
6526 }
6527 
6528 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6529 {
6530 	switch(task->tk_status) {
6531 	case 0:
6532 	case -NFS4ERR_COMPLETE_ALREADY:
6533 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
6534 		break;
6535 	case -NFS4ERR_DELAY:
6536 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
6537 		/* fall through */
6538 	case -NFS4ERR_RETRY_UNCACHED_REP:
6539 		return -EAGAIN;
6540 	default:
6541 		nfs4_schedule_lease_recovery(clp);
6542 	}
6543 	return 0;
6544 }
6545 
6546 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6547 {
6548 	struct nfs4_reclaim_complete_data *calldata = data;
6549 	struct nfs_client *clp = calldata->clp;
6550 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
6551 
6552 	dprintk("--> %s\n", __func__);
6553 	if (!nfs41_sequence_done(task, res))
6554 		return;
6555 
6556 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6557 		rpc_restart_call_prepare(task);
6558 		return;
6559 	}
6560 	dprintk("<-- %s\n", __func__);
6561 }
6562 
6563 static void nfs4_free_reclaim_complete_data(void *data)
6564 {
6565 	struct nfs4_reclaim_complete_data *calldata = data;
6566 
6567 	kfree(calldata);
6568 }
6569 
6570 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6571 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
6572 	.rpc_call_done = nfs4_reclaim_complete_done,
6573 	.rpc_release = nfs4_free_reclaim_complete_data,
6574 };
6575 
6576 /*
6577  * Issue a global reclaim complete.
6578  */
6579 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
6580 		struct rpc_cred *cred)
6581 {
6582 	struct nfs4_reclaim_complete_data *calldata;
6583 	struct rpc_task *task;
6584 	struct rpc_message msg = {
6585 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6586 		.rpc_cred = cred,
6587 	};
6588 	struct rpc_task_setup task_setup_data = {
6589 		.rpc_client = clp->cl_rpcclient,
6590 		.rpc_message = &msg,
6591 		.callback_ops = &nfs4_reclaim_complete_call_ops,
6592 		.flags = RPC_TASK_ASYNC,
6593 	};
6594 	int status = -ENOMEM;
6595 
6596 	dprintk("--> %s\n", __func__);
6597 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6598 	if (calldata == NULL)
6599 		goto out;
6600 	calldata->clp = clp;
6601 	calldata->arg.one_fs = 0;
6602 
6603 	nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6604 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
6605 	msg.rpc_argp = &calldata->arg;
6606 	msg.rpc_resp = &calldata->res;
6607 	task_setup_data.callback_data = calldata;
6608 	task = rpc_run_task(&task_setup_data);
6609 	if (IS_ERR(task)) {
6610 		status = PTR_ERR(task);
6611 		goto out;
6612 	}
6613 	status = nfs4_wait_for_completion_rpc_task(task);
6614 	if (status == 0)
6615 		status = task->tk_status;
6616 	rpc_put_task(task);
6617 	return 0;
6618 out:
6619 	dprintk("<-- %s status=%d\n", __func__, status);
6620 	return status;
6621 }
6622 
6623 static void
6624 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6625 {
6626 	struct nfs4_layoutget *lgp = calldata;
6627 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6628 	struct nfs4_session *session = nfs4_get_session(server);
6629 
6630 	dprintk("--> %s\n", __func__);
6631 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
6632 	 * right now covering the LAYOUTGET we are about to send.
6633 	 * However, that is not so catastrophic, and there seems
6634 	 * to be no way to prevent it completely.
6635 	 */
6636 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
6637 				&lgp->res.seq_res, task))
6638 		return;
6639 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6640 					  NFS_I(lgp->args.inode)->layout,
6641 					  lgp->args.ctx->state)) {
6642 		rpc_exit(task, NFS4_OK);
6643 	}
6644 }
6645 
6646 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6647 {
6648 	struct nfs4_layoutget *lgp = calldata;
6649 	struct inode *inode = lgp->args.inode;
6650 	struct nfs_server *server = NFS_SERVER(inode);
6651 	struct pnfs_layout_hdr *lo;
6652 	struct nfs4_state *state = NULL;
6653 	unsigned long timeo, giveup;
6654 
6655 	dprintk("--> %s\n", __func__);
6656 
6657 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
6658 		goto out;
6659 
6660 	switch (task->tk_status) {
6661 	case 0:
6662 		goto out;
6663 	case -NFS4ERR_LAYOUTTRYLATER:
6664 	case -NFS4ERR_RECALLCONFLICT:
6665 		timeo = rpc_get_timeout(task->tk_client);
6666 		giveup = lgp->args.timestamp + timeo;
6667 		if (time_after(giveup, jiffies))
6668 			task->tk_status = -NFS4ERR_DELAY;
6669 		break;
6670 	case -NFS4ERR_EXPIRED:
6671 	case -NFS4ERR_BAD_STATEID:
6672 		spin_lock(&inode->i_lock);
6673 		lo = NFS_I(inode)->layout;
6674 		if (!lo || list_empty(&lo->plh_segs)) {
6675 			spin_unlock(&inode->i_lock);
6676 			/* If the open stateid was bad, then recover it. */
6677 			state = lgp->args.ctx->state;
6678 		} else {
6679 			LIST_HEAD(head);
6680 
6681 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
6682 			spin_unlock(&inode->i_lock);
6683 			/* Mark the bad layout state as invalid, then
6684 			 * retry using the open stateid. */
6685 			pnfs_free_lseg_list(&head);
6686 		}
6687 	}
6688 	if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
6689 		rpc_restart_call_prepare(task);
6690 out:
6691 	dprintk("<-- %s\n", __func__);
6692 }
6693 
6694 static size_t max_response_pages(struct nfs_server *server)
6695 {
6696 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6697 	return nfs_page_array_len(0, max_resp_sz);
6698 }
6699 
6700 static void nfs4_free_pages(struct page **pages, size_t size)
6701 {
6702 	int i;
6703 
6704 	if (!pages)
6705 		return;
6706 
6707 	for (i = 0; i < size; i++) {
6708 		if (!pages[i])
6709 			break;
6710 		__free_page(pages[i]);
6711 	}
6712 	kfree(pages);
6713 }
6714 
6715 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6716 {
6717 	struct page **pages;
6718 	int i;
6719 
6720 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6721 	if (!pages) {
6722 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6723 		return NULL;
6724 	}
6725 
6726 	for (i = 0; i < size; i++) {
6727 		pages[i] = alloc_page(gfp_flags);
6728 		if (!pages[i]) {
6729 			dprintk("%s: failed to allocate page\n", __func__);
6730 			nfs4_free_pages(pages, size);
6731 			return NULL;
6732 		}
6733 	}
6734 
6735 	return pages;
6736 }
6737 
6738 static void nfs4_layoutget_release(void *calldata)
6739 {
6740 	struct nfs4_layoutget *lgp = calldata;
6741 	struct inode *inode = lgp->args.inode;
6742 	struct nfs_server *server = NFS_SERVER(inode);
6743 	size_t max_pages = max_response_pages(server);
6744 
6745 	dprintk("--> %s\n", __func__);
6746 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
6747 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
6748 	put_nfs_open_context(lgp->args.ctx);
6749 	kfree(calldata);
6750 	dprintk("<-- %s\n", __func__);
6751 }
6752 
6753 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6754 	.rpc_call_prepare = nfs4_layoutget_prepare,
6755 	.rpc_call_done = nfs4_layoutget_done,
6756 	.rpc_release = nfs4_layoutget_release,
6757 };
6758 
6759 struct pnfs_layout_segment *
6760 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6761 {
6762 	struct inode *inode = lgp->args.inode;
6763 	struct nfs_server *server = NFS_SERVER(inode);
6764 	size_t max_pages = max_response_pages(server);
6765 	struct rpc_task *task;
6766 	struct rpc_message msg = {
6767 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6768 		.rpc_argp = &lgp->args,
6769 		.rpc_resp = &lgp->res,
6770 		.rpc_cred = lgp->cred,
6771 	};
6772 	struct rpc_task_setup task_setup_data = {
6773 		.rpc_client = server->client,
6774 		.rpc_message = &msg,
6775 		.callback_ops = &nfs4_layoutget_call_ops,
6776 		.callback_data = lgp,
6777 		.flags = RPC_TASK_ASYNC,
6778 	};
6779 	struct pnfs_layout_segment *lseg = NULL;
6780 	int status = 0;
6781 
6782 	dprintk("--> %s\n", __func__);
6783 
6784 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6785 	if (!lgp->args.layout.pages) {
6786 		nfs4_layoutget_release(lgp);
6787 		return ERR_PTR(-ENOMEM);
6788 	}
6789 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6790 	lgp->args.timestamp = jiffies;
6791 
6792 	lgp->res.layoutp = &lgp->args.layout;
6793 	lgp->res.seq_res.sr_slot = NULL;
6794 	nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6795 
6796 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
6797 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
6798 
6799 	task = rpc_run_task(&task_setup_data);
6800 	if (IS_ERR(task))
6801 		return ERR_CAST(task);
6802 	status = nfs4_wait_for_completion_rpc_task(task);
6803 	if (status == 0)
6804 		status = task->tk_status;
6805 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
6806 	if (status == 0 && lgp->res.layoutp->len)
6807 		lseg = pnfs_layout_process(lgp);
6808 	rpc_put_task(task);
6809 	dprintk("<-- %s status=%d\n", __func__, status);
6810 	if (status)
6811 		return ERR_PTR(status);
6812 	return lseg;
6813 }
6814 
6815 static void
6816 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6817 {
6818 	struct nfs4_layoutreturn *lrp = calldata;
6819 
6820 	dprintk("--> %s\n", __func__);
6821 	nfs41_setup_sequence(lrp->clp->cl_session,
6822 			&lrp->args.seq_args,
6823 			&lrp->res.seq_res,
6824 			task);
6825 }
6826 
6827 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6828 {
6829 	struct nfs4_layoutreturn *lrp = calldata;
6830 	struct nfs_server *server;
6831 
6832 	dprintk("--> %s\n", __func__);
6833 
6834 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
6835 		return;
6836 
6837 	server = NFS_SERVER(lrp->args.inode);
6838 	if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6839 		rpc_restart_call_prepare(task);
6840 		return;
6841 	}
6842 	dprintk("<-- %s\n", __func__);
6843 }
6844 
6845 static void nfs4_layoutreturn_release(void *calldata)
6846 {
6847 	struct nfs4_layoutreturn *lrp = calldata;
6848 	struct pnfs_layout_hdr *lo = lrp->args.layout;
6849 
6850 	dprintk("--> %s\n", __func__);
6851 	spin_lock(&lo->plh_inode->i_lock);
6852 	if (lrp->res.lrs_present)
6853 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6854 	lo->plh_block_lgets--;
6855 	spin_unlock(&lo->plh_inode->i_lock);
6856 	pnfs_put_layout_hdr(lrp->args.layout);
6857 	kfree(calldata);
6858 	dprintk("<-- %s\n", __func__);
6859 }
6860 
6861 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6862 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
6863 	.rpc_call_done = nfs4_layoutreturn_done,
6864 	.rpc_release = nfs4_layoutreturn_release,
6865 };
6866 
6867 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6868 {
6869 	struct rpc_task *task;
6870 	struct rpc_message msg = {
6871 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6872 		.rpc_argp = &lrp->args,
6873 		.rpc_resp = &lrp->res,
6874 		.rpc_cred = lrp->cred,
6875 	};
6876 	struct rpc_task_setup task_setup_data = {
6877 		.rpc_client = lrp->clp->cl_rpcclient,
6878 		.rpc_message = &msg,
6879 		.callback_ops = &nfs4_layoutreturn_call_ops,
6880 		.callback_data = lrp,
6881 	};
6882 	int status;
6883 
6884 	dprintk("--> %s\n", __func__);
6885 	nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6886 	task = rpc_run_task(&task_setup_data);
6887 	if (IS_ERR(task))
6888 		return PTR_ERR(task);
6889 	status = task->tk_status;
6890 	dprintk("<-- %s status=%d\n", __func__, status);
6891 	rpc_put_task(task);
6892 	return status;
6893 }
6894 
6895 /*
6896  * Retrieve the list of Data Server devices from the MDS.
6897  */
6898 static int _nfs4_getdevicelist(struct nfs_server *server,
6899 				    const struct nfs_fh *fh,
6900 				    struct pnfs_devicelist *devlist)
6901 {
6902 	struct nfs4_getdevicelist_args args = {
6903 		.fh = fh,
6904 		.layoutclass = server->pnfs_curr_ld->id,
6905 	};
6906 	struct nfs4_getdevicelist_res res = {
6907 		.devlist = devlist,
6908 	};
6909 	struct rpc_message msg = {
6910 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6911 		.rpc_argp = &args,
6912 		.rpc_resp = &res,
6913 	};
6914 	int status;
6915 
6916 	dprintk("--> %s\n", __func__);
6917 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6918 				&res.seq_res, 0);
6919 	dprintk("<-- %s status=%d\n", __func__, status);
6920 	return status;
6921 }
6922 
6923 int nfs4_proc_getdevicelist(struct nfs_server *server,
6924 			    const struct nfs_fh *fh,
6925 			    struct pnfs_devicelist *devlist)
6926 {
6927 	struct nfs4_exception exception = { };
6928 	int err;
6929 
6930 	do {
6931 		err = nfs4_handle_exception(server,
6932 				_nfs4_getdevicelist(server, fh, devlist),
6933 				&exception);
6934 	} while (exception.retry);
6935 
6936 	dprintk("%s: err=%d, num_devs=%u\n", __func__,
6937 		err, devlist->num_devs);
6938 
6939 	return err;
6940 }
6941 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6942 
6943 static int
6944 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
6945 		struct pnfs_device *pdev,
6946 		struct rpc_cred *cred)
6947 {
6948 	struct nfs4_getdeviceinfo_args args = {
6949 		.pdev = pdev,
6950 	};
6951 	struct nfs4_getdeviceinfo_res res = {
6952 		.pdev = pdev,
6953 	};
6954 	struct rpc_message msg = {
6955 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6956 		.rpc_argp = &args,
6957 		.rpc_resp = &res,
6958 		.rpc_cred = cred,
6959 	};
6960 	int status;
6961 
6962 	dprintk("--> %s\n", __func__);
6963 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6964 	dprintk("<-- %s status=%d\n", __func__, status);
6965 
6966 	return status;
6967 }
6968 
6969 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
6970 		struct pnfs_device *pdev,
6971 		struct rpc_cred *cred)
6972 {
6973 	struct nfs4_exception exception = { };
6974 	int err;
6975 
6976 	do {
6977 		err = nfs4_handle_exception(server,
6978 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
6979 					&exception);
6980 	} while (exception.retry);
6981 	return err;
6982 }
6983 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6984 
6985 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6986 {
6987 	struct nfs4_layoutcommit_data *data = calldata;
6988 	struct nfs_server *server = NFS_SERVER(data->args.inode);
6989 	struct nfs4_session *session = nfs4_get_session(server);
6990 
6991 	nfs41_setup_sequence(session,
6992 			&data->args.seq_args,
6993 			&data->res.seq_res,
6994 			task);
6995 }
6996 
6997 static void
6998 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6999 {
7000 	struct nfs4_layoutcommit_data *data = calldata;
7001 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7002 
7003 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7004 		return;
7005 
7006 	switch (task->tk_status) { /* Just ignore these failures */
7007 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7008 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7009 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7010 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7011 		task->tk_status = 0;
7012 		break;
7013 	case 0:
7014 		nfs_post_op_update_inode_force_wcc(data->args.inode,
7015 						   data->res.fattr);
7016 		break;
7017 	default:
7018 		if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
7019 			rpc_restart_call_prepare(task);
7020 			return;
7021 		}
7022 	}
7023 }
7024 
7025 static void nfs4_layoutcommit_release(void *calldata)
7026 {
7027 	struct nfs4_layoutcommit_data *data = calldata;
7028 
7029 	pnfs_cleanup_layoutcommit(data);
7030 	put_rpccred(data->cred);
7031 	kfree(data);
7032 }
7033 
7034 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
7035 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
7036 	.rpc_call_done = nfs4_layoutcommit_done,
7037 	.rpc_release = nfs4_layoutcommit_release,
7038 };
7039 
7040 int
7041 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
7042 {
7043 	struct rpc_message msg = {
7044 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
7045 		.rpc_argp = &data->args,
7046 		.rpc_resp = &data->res,
7047 		.rpc_cred = data->cred,
7048 	};
7049 	struct rpc_task_setup task_setup_data = {
7050 		.task = &data->task,
7051 		.rpc_client = NFS_CLIENT(data->args.inode),
7052 		.rpc_message = &msg,
7053 		.callback_ops = &nfs4_layoutcommit_ops,
7054 		.callback_data = data,
7055 		.flags = RPC_TASK_ASYNC,
7056 	};
7057 	struct rpc_task *task;
7058 	int status = 0;
7059 
7060 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
7061 		"lbw: %llu inode %lu\n",
7062 		data->task.tk_pid, sync,
7063 		data->args.lastbytewritten,
7064 		data->args.inode->i_ino);
7065 
7066 	nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
7067 	task = rpc_run_task(&task_setup_data);
7068 	if (IS_ERR(task))
7069 		return PTR_ERR(task);
7070 	if (sync == false)
7071 		goto out;
7072 	status = nfs4_wait_for_completion_rpc_task(task);
7073 	if (status != 0)
7074 		goto out;
7075 	status = task->tk_status;
7076 out:
7077 	dprintk("%s: status %d\n", __func__, status);
7078 	rpc_put_task(task);
7079 	return status;
7080 }
7081 
7082 static int
7083 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7084 		    struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
7085 {
7086 	struct nfs41_secinfo_no_name_args args = {
7087 		.style = SECINFO_STYLE_CURRENT_FH,
7088 	};
7089 	struct nfs4_secinfo_res res = {
7090 		.flavors = flavors,
7091 	};
7092 	struct rpc_message msg = {
7093 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
7094 		.rpc_argp = &args,
7095 		.rpc_resp = &res,
7096 	};
7097 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7098 }
7099 
7100 static int
7101 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
7102 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
7103 {
7104 	struct nfs4_exception exception = { };
7105 	int err;
7106 	do {
7107 		err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
7108 		switch (err) {
7109 		case 0:
7110 		case -NFS4ERR_WRONGSEC:
7111 		case -NFS4ERR_NOTSUPP:
7112 			goto out;
7113 		default:
7114 			err = nfs4_handle_exception(server, err, &exception);
7115 		}
7116 	} while (exception.retry);
7117 out:
7118 	return err;
7119 }
7120 
7121 static int
7122 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
7123 		    struct nfs_fsinfo *info)
7124 {
7125 	int err;
7126 	struct page *page;
7127 	rpc_authflavor_t flavor;
7128 	struct nfs4_secinfo_flavors *flavors;
7129 
7130 	page = alloc_page(GFP_KERNEL);
7131 	if (!page) {
7132 		err = -ENOMEM;
7133 		goto out;
7134 	}
7135 
7136 	flavors = page_address(page);
7137 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
7138 
7139 	/*
7140 	 * Fall back on "guess and check" method if
7141 	 * the server doesn't support SECINFO_NO_NAME
7142 	 */
7143 	if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
7144 		err = nfs4_find_root_sec(server, fhandle, info);
7145 		goto out_freepage;
7146 	}
7147 	if (err)
7148 		goto out_freepage;
7149 
7150 	flavor = nfs_find_best_sec(flavors);
7151 	if (err == 0)
7152 		err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
7153 
7154 out_freepage:
7155 	put_page(page);
7156 	if (err == -EACCES)
7157 		return -EPERM;
7158 out:
7159 	return err;
7160 }
7161 
7162 static int _nfs41_test_stateid(struct nfs_server *server,
7163 		nfs4_stateid *stateid,
7164 		struct rpc_cred *cred)
7165 {
7166 	int status;
7167 	struct nfs41_test_stateid_args args = {
7168 		.stateid = stateid,
7169 	};
7170 	struct nfs41_test_stateid_res res;
7171 	struct rpc_message msg = {
7172 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
7173 		.rpc_argp = &args,
7174 		.rpc_resp = &res,
7175 		.rpc_cred = cred,
7176 	};
7177 
7178 	dprintk("NFS call  test_stateid %p\n", stateid);
7179 	nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
7180 	nfs4_set_sequence_privileged(&args.seq_args);
7181 	status = nfs4_call_sync_sequence(server->client, server, &msg,
7182 			&args.seq_args, &res.seq_res);
7183 	if (status != NFS_OK) {
7184 		dprintk("NFS reply test_stateid: failed, %d\n", status);
7185 		return status;
7186 	}
7187 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
7188 	return -res.status;
7189 }
7190 
7191 /**
7192  * nfs41_test_stateid - perform a TEST_STATEID operation
7193  *
7194  * @server: server / transport on which to perform the operation
7195  * @stateid: state ID to test
7196  * @cred: credential
7197  *
7198  * Returns NFS_OK if the server recognizes that "stateid" is valid.
7199  * Otherwise a negative NFS4ERR value is returned if the operation
7200  * failed or the state ID is not currently valid.
7201  */
7202 static int nfs41_test_stateid(struct nfs_server *server,
7203 		nfs4_stateid *stateid,
7204 		struct rpc_cred *cred)
7205 {
7206 	struct nfs4_exception exception = { };
7207 	int err;
7208 	do {
7209 		err = _nfs41_test_stateid(server, stateid, cred);
7210 		if (err != -NFS4ERR_DELAY)
7211 			break;
7212 		nfs4_handle_exception(server, err, &exception);
7213 	} while (exception.retry);
7214 	return err;
7215 }
7216 
7217 struct nfs_free_stateid_data {
7218 	struct nfs_server *server;
7219 	struct nfs41_free_stateid_args args;
7220 	struct nfs41_free_stateid_res res;
7221 };
7222 
7223 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
7224 {
7225 	struct nfs_free_stateid_data *data = calldata;
7226 	nfs41_setup_sequence(nfs4_get_session(data->server),
7227 			&data->args.seq_args,
7228 			&data->res.seq_res,
7229 			task);
7230 }
7231 
7232 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
7233 {
7234 	struct nfs_free_stateid_data *data = calldata;
7235 
7236 	nfs41_sequence_done(task, &data->res.seq_res);
7237 
7238 	switch (task->tk_status) {
7239 	case -NFS4ERR_DELAY:
7240 		if (nfs4_async_handle_error(task, data->server, NULL) == -EAGAIN)
7241 			rpc_restart_call_prepare(task);
7242 	}
7243 }
7244 
7245 static void nfs41_free_stateid_release(void *calldata)
7246 {
7247 	kfree(calldata);
7248 }
7249 
7250 const struct rpc_call_ops nfs41_free_stateid_ops = {
7251 	.rpc_call_prepare = nfs41_free_stateid_prepare,
7252 	.rpc_call_done = nfs41_free_stateid_done,
7253 	.rpc_release = nfs41_free_stateid_release,
7254 };
7255 
7256 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
7257 		nfs4_stateid *stateid,
7258 		struct rpc_cred *cred,
7259 		bool privileged)
7260 {
7261 	struct rpc_message msg = {
7262 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
7263 		.rpc_cred = cred,
7264 	};
7265 	struct rpc_task_setup task_setup = {
7266 		.rpc_client = server->client,
7267 		.rpc_message = &msg,
7268 		.callback_ops = &nfs41_free_stateid_ops,
7269 		.flags = RPC_TASK_ASYNC,
7270 	};
7271 	struct nfs_free_stateid_data *data;
7272 
7273 	dprintk("NFS call  free_stateid %p\n", stateid);
7274 	data = kmalloc(sizeof(*data), GFP_NOFS);
7275 	if (!data)
7276 		return ERR_PTR(-ENOMEM);
7277 	data->server = server;
7278 	nfs4_stateid_copy(&data->args.stateid, stateid);
7279 
7280 	task_setup.callback_data = data;
7281 
7282 	msg.rpc_argp = &data->args;
7283 	msg.rpc_resp = &data->res;
7284 	nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
7285 	if (privileged)
7286 		nfs4_set_sequence_privileged(&data->args.seq_args);
7287 
7288 	return rpc_run_task(&task_setup);
7289 }
7290 
7291 /**
7292  * nfs41_free_stateid - perform a FREE_STATEID operation
7293  *
7294  * @server: server / transport on which to perform the operation
7295  * @stateid: state ID to release
7296  * @cred: credential
7297  *
7298  * Returns NFS_OK if the server freed "stateid".  Otherwise a
7299  * negative NFS4ERR value is returned.
7300  */
7301 static int nfs41_free_stateid(struct nfs_server *server,
7302 		nfs4_stateid *stateid,
7303 		struct rpc_cred *cred)
7304 {
7305 	struct rpc_task *task;
7306 	int ret;
7307 
7308 	task = _nfs41_free_stateid(server, stateid, cred, true);
7309 	if (IS_ERR(task))
7310 		return PTR_ERR(task);
7311 	ret = rpc_wait_for_completion_task(task);
7312 	if (!ret)
7313 		ret = task->tk_status;
7314 	rpc_put_task(task);
7315 	return ret;
7316 }
7317 
7318 static int nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
7319 {
7320 	struct rpc_task *task;
7321 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
7322 
7323 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
7324 	nfs4_free_lock_state(server, lsp);
7325 	if (IS_ERR(task))
7326 		return PTR_ERR(task);
7327 	rpc_put_task(task);
7328 	return 0;
7329 }
7330 
7331 static bool nfs41_match_stateid(const nfs4_stateid *s1,
7332 		const nfs4_stateid *s2)
7333 {
7334 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
7335 		return false;
7336 
7337 	if (s1->seqid == s2->seqid)
7338 		return true;
7339 	if (s1->seqid == 0 || s2->seqid == 0)
7340 		return true;
7341 
7342 	return false;
7343 }
7344 
7345 #endif /* CONFIG_NFS_V4_1 */
7346 
7347 static bool nfs4_match_stateid(const nfs4_stateid *s1,
7348 		const nfs4_stateid *s2)
7349 {
7350 	return nfs4_stateid_match(s1, s2);
7351 }
7352 
7353 
7354 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
7355 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7356 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
7357 	.recover_open	= nfs4_open_reclaim,
7358 	.recover_lock	= nfs4_lock_reclaim,
7359 	.establish_clid = nfs4_init_clientid,
7360 	.get_clid_cred	= nfs4_get_setclientid_cred,
7361 	.detect_trunking = nfs40_discover_server_trunking,
7362 };
7363 
7364 #if defined(CONFIG_NFS_V4_1)
7365 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
7366 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
7367 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
7368 	.recover_open	= nfs4_open_reclaim,
7369 	.recover_lock	= nfs4_lock_reclaim,
7370 	.establish_clid = nfs41_init_clientid,
7371 	.get_clid_cred	= nfs4_get_exchange_id_cred,
7372 	.reclaim_complete = nfs41_proc_reclaim_complete,
7373 	.detect_trunking = nfs41_discover_server_trunking,
7374 };
7375 #endif /* CONFIG_NFS_V4_1 */
7376 
7377 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
7378 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7379 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
7380 	.recover_open	= nfs4_open_expired,
7381 	.recover_lock	= nfs4_lock_expired,
7382 	.establish_clid = nfs4_init_clientid,
7383 	.get_clid_cred	= nfs4_get_setclientid_cred,
7384 };
7385 
7386 #if defined(CONFIG_NFS_V4_1)
7387 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
7388 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
7389 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
7390 	.recover_open	= nfs41_open_expired,
7391 	.recover_lock	= nfs41_lock_expired,
7392 	.establish_clid = nfs41_init_clientid,
7393 	.get_clid_cred	= nfs4_get_exchange_id_cred,
7394 };
7395 #endif /* CONFIG_NFS_V4_1 */
7396 
7397 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
7398 	.sched_state_renewal = nfs4_proc_async_renew,
7399 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
7400 	.renew_lease = nfs4_proc_renew,
7401 };
7402 
7403 #if defined(CONFIG_NFS_V4_1)
7404 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
7405 	.sched_state_renewal = nfs41_proc_async_sequence,
7406 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
7407 	.renew_lease = nfs4_proc_sequence,
7408 };
7409 #endif
7410 
7411 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
7412 	.minor_version = 0,
7413 	.init_caps = NFS_CAP_READDIRPLUS
7414 		| NFS_CAP_ATOMIC_OPEN
7415 		| NFS_CAP_CHANGE_ATTR
7416 		| NFS_CAP_POSIX_LOCK,
7417 	.call_sync = _nfs4_call_sync,
7418 	.match_stateid = nfs4_match_stateid,
7419 	.find_root_sec = nfs4_find_root_sec,
7420 	.free_lock_state = nfs4_release_lockowner,
7421 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
7422 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
7423 	.state_renewal_ops = &nfs40_state_renewal_ops,
7424 };
7425 
7426 #if defined(CONFIG_NFS_V4_1)
7427 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
7428 	.minor_version = 1,
7429 	.init_caps = NFS_CAP_READDIRPLUS
7430 		| NFS_CAP_ATOMIC_OPEN
7431 		| NFS_CAP_CHANGE_ATTR
7432 		| NFS_CAP_POSIX_LOCK
7433 		| NFS_CAP_STATEID_NFSV41
7434 		| NFS_CAP_ATOMIC_OPEN_V1,
7435 	.call_sync = nfs4_call_sync_sequence,
7436 	.match_stateid = nfs41_match_stateid,
7437 	.find_root_sec = nfs41_find_root_sec,
7438 	.free_lock_state = nfs41_free_lock_state,
7439 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7440 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7441 	.state_renewal_ops = &nfs41_state_renewal_ops,
7442 };
7443 #endif
7444 
7445 #if defined(CONFIG_NFS_V4_2)
7446 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
7447 	.minor_version = 2,
7448 	.init_caps = NFS_CAP_READDIRPLUS
7449 		| NFS_CAP_ATOMIC_OPEN
7450 		| NFS_CAP_CHANGE_ATTR
7451 		| NFS_CAP_POSIX_LOCK
7452 		| NFS_CAP_STATEID_NFSV41
7453 		| NFS_CAP_ATOMIC_OPEN_V1,
7454 	.call_sync = nfs4_call_sync_sequence,
7455 	.match_stateid = nfs41_match_stateid,
7456 	.find_root_sec = nfs41_find_root_sec,
7457 	.free_lock_state = nfs41_free_lock_state,
7458 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
7459 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
7460 	.state_renewal_ops = &nfs41_state_renewal_ops,
7461 };
7462 #endif
7463 
7464 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
7465 	[0] = &nfs_v4_0_minor_ops,
7466 #if defined(CONFIG_NFS_V4_1)
7467 	[1] = &nfs_v4_1_minor_ops,
7468 #endif
7469 #if defined(CONFIG_NFS_V4_2)
7470 	[2] = &nfs_v4_2_minor_ops,
7471 #endif
7472 };
7473 
7474 const struct inode_operations nfs4_dir_inode_operations = {
7475 	.create		= nfs_create,
7476 	.lookup		= nfs_lookup,
7477 	.atomic_open	= nfs_atomic_open,
7478 	.link		= nfs_link,
7479 	.unlink		= nfs_unlink,
7480 	.symlink	= nfs_symlink,
7481 	.mkdir		= nfs_mkdir,
7482 	.rmdir		= nfs_rmdir,
7483 	.mknod		= nfs_mknod,
7484 	.rename		= nfs_rename,
7485 	.permission	= nfs_permission,
7486 	.getattr	= nfs_getattr,
7487 	.setattr	= nfs_setattr,
7488 	.getxattr	= generic_getxattr,
7489 	.setxattr	= generic_setxattr,
7490 	.listxattr	= generic_listxattr,
7491 	.removexattr	= generic_removexattr,
7492 };
7493 
7494 static const struct inode_operations nfs4_file_inode_operations = {
7495 	.permission	= nfs_permission,
7496 	.getattr	= nfs_getattr,
7497 	.setattr	= nfs_setattr,
7498 	.getxattr	= generic_getxattr,
7499 	.setxattr	= generic_setxattr,
7500 	.listxattr	= generic_listxattr,
7501 	.removexattr	= generic_removexattr,
7502 };
7503 
7504 const struct nfs_rpc_ops nfs_v4_clientops = {
7505 	.version	= 4,			/* protocol version */
7506 	.dentry_ops	= &nfs4_dentry_operations,
7507 	.dir_inode_ops	= &nfs4_dir_inode_operations,
7508 	.file_inode_ops	= &nfs4_file_inode_operations,
7509 	.file_ops	= &nfs4_file_operations,
7510 	.getroot	= nfs4_proc_get_root,
7511 	.submount	= nfs4_submount,
7512 	.try_mount	= nfs4_try_mount,
7513 	.getattr	= nfs4_proc_getattr,
7514 	.setattr	= nfs4_proc_setattr,
7515 	.lookup		= nfs4_proc_lookup,
7516 	.access		= nfs4_proc_access,
7517 	.readlink	= nfs4_proc_readlink,
7518 	.create		= nfs4_proc_create,
7519 	.remove		= nfs4_proc_remove,
7520 	.unlink_setup	= nfs4_proc_unlink_setup,
7521 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7522 	.unlink_done	= nfs4_proc_unlink_done,
7523 	.rename		= nfs4_proc_rename,
7524 	.rename_setup	= nfs4_proc_rename_setup,
7525 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7526 	.rename_done	= nfs4_proc_rename_done,
7527 	.link		= nfs4_proc_link,
7528 	.symlink	= nfs4_proc_symlink,
7529 	.mkdir		= nfs4_proc_mkdir,
7530 	.rmdir		= nfs4_proc_remove,
7531 	.readdir	= nfs4_proc_readdir,
7532 	.mknod		= nfs4_proc_mknod,
7533 	.statfs		= nfs4_proc_statfs,
7534 	.fsinfo		= nfs4_proc_fsinfo,
7535 	.pathconf	= nfs4_proc_pathconf,
7536 	.set_capabilities = nfs4_server_capabilities,
7537 	.decode_dirent	= nfs4_decode_dirent,
7538 	.read_setup	= nfs4_proc_read_setup,
7539 	.read_pageio_init = pnfs_pageio_init_read,
7540 	.read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7541 	.read_done	= nfs4_read_done,
7542 	.write_setup	= nfs4_proc_write_setup,
7543 	.write_pageio_init = pnfs_pageio_init_write,
7544 	.write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7545 	.write_done	= nfs4_write_done,
7546 	.commit_setup	= nfs4_proc_commit_setup,
7547 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7548 	.commit_done	= nfs4_commit_done,
7549 	.lock		= nfs4_proc_lock,
7550 	.clear_acl_cache = nfs4_zap_acl_attr,
7551 	.close_context  = nfs4_close_context,
7552 	.open_context	= nfs4_atomic_open,
7553 	.have_delegation = nfs4_have_delegation,
7554 	.return_delegation = nfs4_inode_return_delegation,
7555 	.alloc_client	= nfs4_alloc_client,
7556 	.init_client	= nfs4_init_client,
7557 	.free_client	= nfs4_free_client,
7558 	.create_server	= nfs4_create_server,
7559 	.clone_server	= nfs_clone_server,
7560 };
7561 
7562 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7563 	.prefix	= XATTR_NAME_NFSV4_ACL,
7564 	.list	= nfs4_xattr_list_nfs4_acl,
7565 	.get	= nfs4_xattr_get_nfs4_acl,
7566 	.set	= nfs4_xattr_set_nfs4_acl,
7567 };
7568 
7569 const struct xattr_handler *nfs4_xattr_handlers[] = {
7570 	&nfs4_xattr_nfs4_acl_handler,
7571 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7572 	&nfs4_xattr_nfs4_label_handler,
7573 #endif
7574 	NULL
7575 };
7576 
7577 /*
7578  * Local variables:
7579  *  c-basic-offset: 8
7580  * End:
7581  */
7582