xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 8b036556)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68 
69 #include "nfs4trace.h"
70 
71 #define NFSDBG_FACILITY		NFSDBG_PROC
72 
73 #define NFS4_POLL_RETRY_MIN	(HZ/10)
74 #define NFS4_POLL_RETRY_MAX	(15*HZ)
75 
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_FILE_OPEN:
162 		return -EBUSY;
163 	default:
164 		dprintk("%s could not handle NFSv4 error %d\n",
165 				__func__, -err);
166 		break;
167 	}
168 	return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175 	FATTR4_WORD0_TYPE
176 	| FATTR4_WORD0_CHANGE
177 	| FATTR4_WORD0_SIZE
178 	| FATTR4_WORD0_FSID
179 	| FATTR4_WORD0_FILEID,
180 	FATTR4_WORD1_MODE
181 	| FATTR4_WORD1_NUMLINKS
182 	| FATTR4_WORD1_OWNER
183 	| FATTR4_WORD1_OWNER_GROUP
184 	| FATTR4_WORD1_RAWDEV
185 	| FATTR4_WORD1_SPACE_USED
186 	| FATTR4_WORD1_TIME_ACCESS
187 	| FATTR4_WORD1_TIME_METADATA
188 	| FATTR4_WORD1_TIME_MODIFY,
189 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
190 	FATTR4_WORD2_SECURITY_LABEL
191 #endif
192 };
193 
194 static const u32 nfs4_pnfs_open_bitmap[3] = {
195 	FATTR4_WORD0_TYPE
196 	| FATTR4_WORD0_CHANGE
197 	| FATTR4_WORD0_SIZE
198 	| FATTR4_WORD0_FSID
199 	| FATTR4_WORD0_FILEID,
200 	FATTR4_WORD1_MODE
201 	| FATTR4_WORD1_NUMLINKS
202 	| FATTR4_WORD1_OWNER
203 	| FATTR4_WORD1_OWNER_GROUP
204 	| FATTR4_WORD1_RAWDEV
205 	| FATTR4_WORD1_SPACE_USED
206 	| FATTR4_WORD1_TIME_ACCESS
207 	| FATTR4_WORD1_TIME_METADATA
208 	| FATTR4_WORD1_TIME_MODIFY,
209 	FATTR4_WORD2_MDSTHRESHOLD
210 };
211 
212 static const u32 nfs4_open_noattr_bitmap[3] = {
213 	FATTR4_WORD0_TYPE
214 	| FATTR4_WORD0_CHANGE
215 	| FATTR4_WORD0_FILEID,
216 };
217 
218 const u32 nfs4_statfs_bitmap[3] = {
219 	FATTR4_WORD0_FILES_AVAIL
220 	| FATTR4_WORD0_FILES_FREE
221 	| FATTR4_WORD0_FILES_TOTAL,
222 	FATTR4_WORD1_SPACE_AVAIL
223 	| FATTR4_WORD1_SPACE_FREE
224 	| FATTR4_WORD1_SPACE_TOTAL
225 };
226 
227 const u32 nfs4_pathconf_bitmap[3] = {
228 	FATTR4_WORD0_MAXLINK
229 	| FATTR4_WORD0_MAXNAME,
230 	0
231 };
232 
233 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
234 			| FATTR4_WORD0_MAXREAD
235 			| FATTR4_WORD0_MAXWRITE
236 			| FATTR4_WORD0_LEASE_TIME,
237 			FATTR4_WORD1_TIME_DELTA
238 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
239 			FATTR4_WORD2_LAYOUT_BLKSIZE
240 };
241 
242 const u32 nfs4_fs_locations_bitmap[3] = {
243 	FATTR4_WORD0_TYPE
244 	| FATTR4_WORD0_CHANGE
245 	| FATTR4_WORD0_SIZE
246 	| FATTR4_WORD0_FSID
247 	| FATTR4_WORD0_FILEID
248 	| FATTR4_WORD0_FS_LOCATIONS,
249 	FATTR4_WORD1_MODE
250 	| FATTR4_WORD1_NUMLINKS
251 	| FATTR4_WORD1_OWNER
252 	| FATTR4_WORD1_OWNER_GROUP
253 	| FATTR4_WORD1_RAWDEV
254 	| FATTR4_WORD1_SPACE_USED
255 	| FATTR4_WORD1_TIME_ACCESS
256 	| FATTR4_WORD1_TIME_METADATA
257 	| FATTR4_WORD1_TIME_MODIFY
258 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
259 };
260 
261 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
262 		struct nfs4_readdir_arg *readdir)
263 {
264 	__be32 *start, *p;
265 
266 	if (cookie > 2) {
267 		readdir->cookie = cookie;
268 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
269 		return;
270 	}
271 
272 	readdir->cookie = 0;
273 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
274 	if (cookie == 2)
275 		return;
276 
277 	/*
278 	 * NFSv4 servers do not return entries for '.' and '..'
279 	 * Therefore, we fake these entries here.  We let '.'
280 	 * have cookie 0 and '..' have cookie 1.  Note that
281 	 * when talking to the server, we always send cookie 0
282 	 * instead of 1 or 2.
283 	 */
284 	start = p = kmap_atomic(*readdir->pages);
285 
286 	if (cookie == 0) {
287 		*p++ = xdr_one;                                  /* next */
288 		*p++ = xdr_zero;                   /* cookie, first word */
289 		*p++ = xdr_one;                   /* cookie, second word */
290 		*p++ = xdr_one;                             /* entry len */
291 		memcpy(p, ".\0\0\0", 4);                        /* entry */
292 		p++;
293 		*p++ = xdr_one;                         /* bitmap length */
294 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
295 		*p++ = htonl(8);              /* attribute buffer length */
296 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
297 	}
298 
299 	*p++ = xdr_one;                                  /* next */
300 	*p++ = xdr_zero;                   /* cookie, first word */
301 	*p++ = xdr_two;                   /* cookie, second word */
302 	*p++ = xdr_two;                             /* entry len */
303 	memcpy(p, "..\0\0", 4);                         /* entry */
304 	p++;
305 	*p++ = xdr_one;                         /* bitmap length */
306 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
307 	*p++ = htonl(8);              /* attribute buffer length */
308 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
309 
310 	readdir->pgbase = (char *)p - (char *)start;
311 	readdir->count -= readdir->pgbase;
312 	kunmap_atomic(start);
313 }
314 
315 static long nfs4_update_delay(long *timeout)
316 {
317 	long ret;
318 	if (!timeout)
319 		return NFS4_POLL_RETRY_MAX;
320 	if (*timeout <= 0)
321 		*timeout = NFS4_POLL_RETRY_MIN;
322 	if (*timeout > NFS4_POLL_RETRY_MAX)
323 		*timeout = NFS4_POLL_RETRY_MAX;
324 	ret = *timeout;
325 	*timeout <<= 1;
326 	return ret;
327 }
328 
329 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
330 {
331 	int res = 0;
332 
333 	might_sleep();
334 
335 	freezable_schedule_timeout_killable_unsafe(
336 		nfs4_update_delay(timeout));
337 	if (fatal_signal_pending(current))
338 		res = -ERESTARTSYS;
339 	return res;
340 }
341 
342 /* This is the error handling routine for processes that are allowed
343  * to sleep.
344  */
345 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
346 {
347 	struct nfs_client *clp = server->nfs_client;
348 	struct nfs4_state *state = exception->state;
349 	struct inode *inode = exception->inode;
350 	int ret = errorcode;
351 
352 	exception->retry = 0;
353 	switch(errorcode) {
354 		case 0:
355 			return 0;
356 		case -NFS4ERR_OPENMODE:
357 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
358 				nfs4_inode_return_delegation(inode);
359 				exception->retry = 1;
360 				return 0;
361 			}
362 			if (state == NULL)
363 				break;
364 			ret = nfs4_schedule_stateid_recovery(server, state);
365 			if (ret < 0)
366 				break;
367 			goto wait_on_recovery;
368 		case -NFS4ERR_DELEG_REVOKED:
369 		case -NFS4ERR_ADMIN_REVOKED:
370 		case -NFS4ERR_BAD_STATEID:
371 			if (state == NULL)
372 				break;
373 			ret = nfs4_schedule_stateid_recovery(server, state);
374 			if (ret < 0)
375 				break;
376 			goto wait_on_recovery;
377 		case -NFS4ERR_EXPIRED:
378 			if (state != NULL) {
379 				ret = nfs4_schedule_stateid_recovery(server, state);
380 				if (ret < 0)
381 					break;
382 			}
383 		case -NFS4ERR_STALE_STATEID:
384 		case -NFS4ERR_STALE_CLIENTID:
385 			nfs4_schedule_lease_recovery(clp);
386 			goto wait_on_recovery;
387 		case -NFS4ERR_MOVED:
388 			ret = nfs4_schedule_migration_recovery(server);
389 			if (ret < 0)
390 				break;
391 			goto wait_on_recovery;
392 		case -NFS4ERR_LEASE_MOVED:
393 			nfs4_schedule_lease_moved_recovery(clp);
394 			goto wait_on_recovery;
395 #if defined(CONFIG_NFS_V4_1)
396 		case -NFS4ERR_BADSESSION:
397 		case -NFS4ERR_BADSLOT:
398 		case -NFS4ERR_BAD_HIGH_SLOT:
399 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
400 		case -NFS4ERR_DEADSESSION:
401 		case -NFS4ERR_SEQ_FALSE_RETRY:
402 		case -NFS4ERR_SEQ_MISORDERED:
403 			dprintk("%s ERROR: %d Reset session\n", __func__,
404 				errorcode);
405 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
406 			goto wait_on_recovery;
407 #endif /* defined(CONFIG_NFS_V4_1) */
408 		case -NFS4ERR_FILE_OPEN:
409 			if (exception->timeout > HZ) {
410 				/* We have retried a decent amount, time to
411 				 * fail
412 				 */
413 				ret = -EBUSY;
414 				break;
415 			}
416 		case -NFS4ERR_GRACE:
417 		case -NFS4ERR_DELAY:
418 			ret = nfs4_delay(server->client, &exception->timeout);
419 			if (ret != 0)
420 				break;
421 		case -NFS4ERR_RETRY_UNCACHED_REP:
422 		case -NFS4ERR_OLD_STATEID:
423 			exception->retry = 1;
424 			break;
425 		case -NFS4ERR_BADOWNER:
426 			/* The following works around a Linux server bug! */
427 		case -NFS4ERR_BADNAME:
428 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
429 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
430 				exception->retry = 1;
431 				printk(KERN_WARNING "NFS: v4 server %s "
432 						"does not accept raw "
433 						"uid/gids. "
434 						"Reenabling the idmapper.\n",
435 						server->nfs_client->cl_hostname);
436 			}
437 	}
438 	/* We failed to handle the error */
439 	return nfs4_map_errors(ret);
440 wait_on_recovery:
441 	ret = nfs4_wait_clnt_recover(clp);
442 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
443 		return -EIO;
444 	if (ret == 0)
445 		exception->retry = 1;
446 	return ret;
447 }
448 
449 /*
450  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
451  * or 'false' otherwise.
452  */
453 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
454 {
455 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
456 
457 	if (flavor == RPC_AUTH_GSS_KRB5I ||
458 	    flavor == RPC_AUTH_GSS_KRB5P)
459 		return true;
460 
461 	return false;
462 }
463 
464 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
465 {
466 	spin_lock(&clp->cl_lock);
467 	if (time_before(clp->cl_last_renewal,timestamp))
468 		clp->cl_last_renewal = timestamp;
469 	spin_unlock(&clp->cl_lock);
470 }
471 
472 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
473 {
474 	do_renew_lease(server->nfs_client, timestamp);
475 }
476 
477 struct nfs4_call_sync_data {
478 	const struct nfs_server *seq_server;
479 	struct nfs4_sequence_args *seq_args;
480 	struct nfs4_sequence_res *seq_res;
481 };
482 
483 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
484 			       struct nfs4_sequence_res *res, int cache_reply)
485 {
486 	args->sa_slot = NULL;
487 	args->sa_cache_this = cache_reply;
488 	args->sa_privileged = 0;
489 
490 	res->sr_slot = NULL;
491 }
492 
493 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
494 {
495 	args->sa_privileged = 1;
496 }
497 
498 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
499 			 struct nfs4_sequence_args *args,
500 			 struct nfs4_sequence_res *res,
501 			 struct rpc_task *task)
502 {
503 	struct nfs4_slot *slot;
504 
505 	/* slot already allocated? */
506 	if (res->sr_slot != NULL)
507 		goto out_start;
508 
509 	spin_lock(&tbl->slot_tbl_lock);
510 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
511 		goto out_sleep;
512 
513 	slot = nfs4_alloc_slot(tbl);
514 	if (IS_ERR(slot)) {
515 		if (slot == ERR_PTR(-ENOMEM))
516 			task->tk_timeout = HZ >> 2;
517 		goto out_sleep;
518 	}
519 	spin_unlock(&tbl->slot_tbl_lock);
520 
521 	args->sa_slot = slot;
522 	res->sr_slot = slot;
523 
524 out_start:
525 	rpc_call_start(task);
526 	return 0;
527 
528 out_sleep:
529 	if (args->sa_privileged)
530 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
531 				NULL, RPC_PRIORITY_PRIVILEGED);
532 	else
533 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
534 	spin_unlock(&tbl->slot_tbl_lock);
535 	return -EAGAIN;
536 }
537 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
538 
539 static int nfs40_sequence_done(struct rpc_task *task,
540 			       struct nfs4_sequence_res *res)
541 {
542 	struct nfs4_slot *slot = res->sr_slot;
543 	struct nfs4_slot_table *tbl;
544 
545 	if (slot == NULL)
546 		goto out;
547 
548 	tbl = slot->table;
549 	spin_lock(&tbl->slot_tbl_lock);
550 	if (!nfs41_wake_and_assign_slot(tbl, slot))
551 		nfs4_free_slot(tbl, slot);
552 	spin_unlock(&tbl->slot_tbl_lock);
553 
554 	res->sr_slot = NULL;
555 out:
556 	return 1;
557 }
558 
559 #if defined(CONFIG_NFS_V4_1)
560 
561 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
562 {
563 	struct nfs4_session *session;
564 	struct nfs4_slot_table *tbl;
565 	struct nfs4_slot *slot = res->sr_slot;
566 	bool send_new_highest_used_slotid = false;
567 
568 	tbl = slot->table;
569 	session = tbl->session;
570 
571 	spin_lock(&tbl->slot_tbl_lock);
572 	/* Be nice to the server: try to ensure that the last transmitted
573 	 * value for highest_user_slotid <= target_highest_slotid
574 	 */
575 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
576 		send_new_highest_used_slotid = true;
577 
578 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
579 		send_new_highest_used_slotid = false;
580 		goto out_unlock;
581 	}
582 	nfs4_free_slot(tbl, slot);
583 
584 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
585 		send_new_highest_used_slotid = false;
586 out_unlock:
587 	spin_unlock(&tbl->slot_tbl_lock);
588 	res->sr_slot = NULL;
589 	if (send_new_highest_used_slotid)
590 		nfs41_server_notify_highest_slotid_update(session->clp);
591 }
592 
593 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
594 {
595 	struct nfs4_session *session;
596 	struct nfs4_slot *slot = res->sr_slot;
597 	struct nfs_client *clp;
598 	bool interrupted = false;
599 	int ret = 1;
600 
601 	if (slot == NULL)
602 		goto out_noaction;
603 	/* don't increment the sequence number if the task wasn't sent */
604 	if (!RPC_WAS_SENT(task))
605 		goto out;
606 
607 	session = slot->table->session;
608 
609 	if (slot->interrupted) {
610 		slot->interrupted = 0;
611 		interrupted = true;
612 	}
613 
614 	trace_nfs4_sequence_done(session, res);
615 	/* Check the SEQUENCE operation status */
616 	switch (res->sr_status) {
617 	case 0:
618 		/* Update the slot's sequence and clientid lease timer */
619 		++slot->seq_nr;
620 		clp = session->clp;
621 		do_renew_lease(clp, res->sr_timestamp);
622 		/* Check sequence flags */
623 		if (res->sr_status_flags != 0)
624 			nfs4_schedule_lease_recovery(clp);
625 		nfs41_update_target_slotid(slot->table, slot, res);
626 		break;
627 	case 1:
628 		/*
629 		 * sr_status remains 1 if an RPC level error occurred.
630 		 * The server may or may not have processed the sequence
631 		 * operation..
632 		 * Mark the slot as having hosted an interrupted RPC call.
633 		 */
634 		slot->interrupted = 1;
635 		goto out;
636 	case -NFS4ERR_DELAY:
637 		/* The server detected a resend of the RPC call and
638 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
639 		 * of RFC5661.
640 		 */
641 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
642 			__func__,
643 			slot->slot_nr,
644 			slot->seq_nr);
645 		goto out_retry;
646 	case -NFS4ERR_BADSLOT:
647 		/*
648 		 * The slot id we used was probably retired. Try again
649 		 * using a different slot id.
650 		 */
651 		goto retry_nowait;
652 	case -NFS4ERR_SEQ_MISORDERED:
653 		/*
654 		 * Was the last operation on this sequence interrupted?
655 		 * If so, retry after bumping the sequence number.
656 		 */
657 		if (interrupted) {
658 			++slot->seq_nr;
659 			goto retry_nowait;
660 		}
661 		/*
662 		 * Could this slot have been previously retired?
663 		 * If so, then the server may be expecting seq_nr = 1!
664 		 */
665 		if (slot->seq_nr != 1) {
666 			slot->seq_nr = 1;
667 			goto retry_nowait;
668 		}
669 		break;
670 	case -NFS4ERR_SEQ_FALSE_RETRY:
671 		++slot->seq_nr;
672 		goto retry_nowait;
673 	default:
674 		/* Just update the slot sequence no. */
675 		++slot->seq_nr;
676 	}
677 out:
678 	/* The session may be reset by one of the error handlers. */
679 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
680 	nfs41_sequence_free_slot(res);
681 out_noaction:
682 	return ret;
683 retry_nowait:
684 	if (rpc_restart_call_prepare(task)) {
685 		task->tk_status = 0;
686 		ret = 0;
687 	}
688 	goto out;
689 out_retry:
690 	if (!rpc_restart_call(task))
691 		goto out;
692 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
693 	return 0;
694 }
695 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
696 
697 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
698 {
699 	if (res->sr_slot == NULL)
700 		return 1;
701 	if (!res->sr_slot->table->session)
702 		return nfs40_sequence_done(task, res);
703 	return nfs41_sequence_done(task, res);
704 }
705 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
706 
707 int nfs41_setup_sequence(struct nfs4_session *session,
708 				struct nfs4_sequence_args *args,
709 				struct nfs4_sequence_res *res,
710 				struct rpc_task *task)
711 {
712 	struct nfs4_slot *slot;
713 	struct nfs4_slot_table *tbl;
714 
715 	dprintk("--> %s\n", __func__);
716 	/* slot already allocated? */
717 	if (res->sr_slot != NULL)
718 		goto out_success;
719 
720 	tbl = &session->fc_slot_table;
721 
722 	task->tk_timeout = 0;
723 
724 	spin_lock(&tbl->slot_tbl_lock);
725 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
726 	    !args->sa_privileged) {
727 		/* The state manager will wait until the slot table is empty */
728 		dprintk("%s session is draining\n", __func__);
729 		goto out_sleep;
730 	}
731 
732 	slot = nfs4_alloc_slot(tbl);
733 	if (IS_ERR(slot)) {
734 		/* If out of memory, try again in 1/4 second */
735 		if (slot == ERR_PTR(-ENOMEM))
736 			task->tk_timeout = HZ >> 2;
737 		dprintk("<-- %s: no free slots\n", __func__);
738 		goto out_sleep;
739 	}
740 	spin_unlock(&tbl->slot_tbl_lock);
741 
742 	args->sa_slot = slot;
743 
744 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
745 			slot->slot_nr, slot->seq_nr);
746 
747 	res->sr_slot = slot;
748 	res->sr_timestamp = jiffies;
749 	res->sr_status_flags = 0;
750 	/*
751 	 * sr_status is only set in decode_sequence, and so will remain
752 	 * set to 1 if an rpc level failure occurs.
753 	 */
754 	res->sr_status = 1;
755 	trace_nfs4_setup_sequence(session, args);
756 out_success:
757 	rpc_call_start(task);
758 	return 0;
759 out_sleep:
760 	/* Privileged tasks are queued with top priority */
761 	if (args->sa_privileged)
762 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
763 				NULL, RPC_PRIORITY_PRIVILEGED);
764 	else
765 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
766 	spin_unlock(&tbl->slot_tbl_lock);
767 	return -EAGAIN;
768 }
769 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
770 
771 static int nfs4_setup_sequence(const struct nfs_server *server,
772 			       struct nfs4_sequence_args *args,
773 			       struct nfs4_sequence_res *res,
774 			       struct rpc_task *task)
775 {
776 	struct nfs4_session *session = nfs4_get_session(server);
777 	int ret = 0;
778 
779 	if (!session)
780 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
781 					    args, res, task);
782 
783 	dprintk("--> %s clp %p session %p sr_slot %u\n",
784 		__func__, session->clp, session, res->sr_slot ?
785 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
786 
787 	ret = nfs41_setup_sequence(session, args, res, task);
788 
789 	dprintk("<-- %s status=%d\n", __func__, ret);
790 	return ret;
791 }
792 
793 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
794 {
795 	struct nfs4_call_sync_data *data = calldata;
796 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
797 
798 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
799 
800 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
801 }
802 
803 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
804 {
805 	struct nfs4_call_sync_data *data = calldata;
806 
807 	nfs41_sequence_done(task, data->seq_res);
808 }
809 
810 static const struct rpc_call_ops nfs41_call_sync_ops = {
811 	.rpc_call_prepare = nfs41_call_sync_prepare,
812 	.rpc_call_done = nfs41_call_sync_done,
813 };
814 
815 #else	/* !CONFIG_NFS_V4_1 */
816 
817 static int nfs4_setup_sequence(const struct nfs_server *server,
818 			       struct nfs4_sequence_args *args,
819 			       struct nfs4_sequence_res *res,
820 			       struct rpc_task *task)
821 {
822 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
823 				    args, res, task);
824 }
825 
826 int nfs4_sequence_done(struct rpc_task *task,
827 		       struct nfs4_sequence_res *res)
828 {
829 	return nfs40_sequence_done(task, res);
830 }
831 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
832 
833 #endif	/* !CONFIG_NFS_V4_1 */
834 
835 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
836 {
837 	struct nfs4_call_sync_data *data = calldata;
838 	nfs4_setup_sequence(data->seq_server,
839 				data->seq_args, data->seq_res, task);
840 }
841 
842 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
843 {
844 	struct nfs4_call_sync_data *data = calldata;
845 	nfs4_sequence_done(task, data->seq_res);
846 }
847 
848 static const struct rpc_call_ops nfs40_call_sync_ops = {
849 	.rpc_call_prepare = nfs40_call_sync_prepare,
850 	.rpc_call_done = nfs40_call_sync_done,
851 };
852 
853 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
854 				   struct nfs_server *server,
855 				   struct rpc_message *msg,
856 				   struct nfs4_sequence_args *args,
857 				   struct nfs4_sequence_res *res)
858 {
859 	int ret;
860 	struct rpc_task *task;
861 	struct nfs_client *clp = server->nfs_client;
862 	struct nfs4_call_sync_data data = {
863 		.seq_server = server,
864 		.seq_args = args,
865 		.seq_res = res,
866 	};
867 	struct rpc_task_setup task_setup = {
868 		.rpc_client = clnt,
869 		.rpc_message = msg,
870 		.callback_ops = clp->cl_mvops->call_sync_ops,
871 		.callback_data = &data
872 	};
873 
874 	task = rpc_run_task(&task_setup);
875 	if (IS_ERR(task))
876 		ret = PTR_ERR(task);
877 	else {
878 		ret = task->tk_status;
879 		rpc_put_task(task);
880 	}
881 	return ret;
882 }
883 
884 int nfs4_call_sync(struct rpc_clnt *clnt,
885 		   struct nfs_server *server,
886 		   struct rpc_message *msg,
887 		   struct nfs4_sequence_args *args,
888 		   struct nfs4_sequence_res *res,
889 		   int cache_reply)
890 {
891 	nfs4_init_sequence(args, res, cache_reply);
892 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
893 }
894 
895 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
896 {
897 	struct nfs_inode *nfsi = NFS_I(dir);
898 
899 	spin_lock(&dir->i_lock);
900 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
901 	if (!cinfo->atomic || cinfo->before != dir->i_version)
902 		nfs_force_lookup_revalidate(dir);
903 	dir->i_version = cinfo->after;
904 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
905 	nfs_fscache_invalidate(dir);
906 	spin_unlock(&dir->i_lock);
907 }
908 
909 struct nfs4_opendata {
910 	struct kref kref;
911 	struct nfs_openargs o_arg;
912 	struct nfs_openres o_res;
913 	struct nfs_open_confirmargs c_arg;
914 	struct nfs_open_confirmres c_res;
915 	struct nfs4_string owner_name;
916 	struct nfs4_string group_name;
917 	struct nfs_fattr f_attr;
918 	struct nfs4_label *f_label;
919 	struct dentry *dir;
920 	struct dentry *dentry;
921 	struct nfs4_state_owner *owner;
922 	struct nfs4_state *state;
923 	struct iattr attrs;
924 	unsigned long timestamp;
925 	unsigned int rpc_done : 1;
926 	unsigned int file_created : 1;
927 	unsigned int is_recover : 1;
928 	int rpc_status;
929 	int cancelled;
930 };
931 
932 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
933 		int err, struct nfs4_exception *exception)
934 {
935 	if (err != -EINVAL)
936 		return false;
937 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
938 		return false;
939 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
940 	exception->retry = 1;
941 	return true;
942 }
943 
944 static u32
945 nfs4_map_atomic_open_share(struct nfs_server *server,
946 		fmode_t fmode, int openflags)
947 {
948 	u32 res = 0;
949 
950 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
951 	case FMODE_READ:
952 		res = NFS4_SHARE_ACCESS_READ;
953 		break;
954 	case FMODE_WRITE:
955 		res = NFS4_SHARE_ACCESS_WRITE;
956 		break;
957 	case FMODE_READ|FMODE_WRITE:
958 		res = NFS4_SHARE_ACCESS_BOTH;
959 	}
960 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
961 		goto out;
962 	/* Want no delegation if we're using O_DIRECT */
963 	if (openflags & O_DIRECT)
964 		res |= NFS4_SHARE_WANT_NO_DELEG;
965 out:
966 	return res;
967 }
968 
969 static enum open_claim_type4
970 nfs4_map_atomic_open_claim(struct nfs_server *server,
971 		enum open_claim_type4 claim)
972 {
973 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
974 		return claim;
975 	switch (claim) {
976 	default:
977 		return claim;
978 	case NFS4_OPEN_CLAIM_FH:
979 		return NFS4_OPEN_CLAIM_NULL;
980 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
981 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
982 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
983 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
984 	}
985 }
986 
987 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
988 {
989 	p->o_res.f_attr = &p->f_attr;
990 	p->o_res.f_label = p->f_label;
991 	p->o_res.seqid = p->o_arg.seqid;
992 	p->c_res.seqid = p->c_arg.seqid;
993 	p->o_res.server = p->o_arg.server;
994 	p->o_res.access_request = p->o_arg.access;
995 	nfs_fattr_init(&p->f_attr);
996 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
997 }
998 
999 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1000 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1001 		const struct iattr *attrs,
1002 		struct nfs4_label *label,
1003 		enum open_claim_type4 claim,
1004 		gfp_t gfp_mask)
1005 {
1006 	struct dentry *parent = dget_parent(dentry);
1007 	struct inode *dir = parent->d_inode;
1008 	struct nfs_server *server = NFS_SERVER(dir);
1009 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1010 	struct nfs4_opendata *p;
1011 
1012 	p = kzalloc(sizeof(*p), gfp_mask);
1013 	if (p == NULL)
1014 		goto err;
1015 
1016 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1017 	if (IS_ERR(p->f_label))
1018 		goto err_free_p;
1019 
1020 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1021 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1022 	if (IS_ERR(p->o_arg.seqid))
1023 		goto err_free_label;
1024 	nfs_sb_active(dentry->d_sb);
1025 	p->dentry = dget(dentry);
1026 	p->dir = parent;
1027 	p->owner = sp;
1028 	atomic_inc(&sp->so_count);
1029 	p->o_arg.open_flags = flags;
1030 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1031 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1032 			fmode, flags);
1033 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1034 	 * will return permission denied for all bits until close */
1035 	if (!(flags & O_EXCL)) {
1036 		/* ask server to check for all possible rights as results
1037 		 * are cached */
1038 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1039 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1040 	}
1041 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1042 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1043 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1044 	p->o_arg.name = &dentry->d_name;
1045 	p->o_arg.server = server;
1046 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1047 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1048 	p->o_arg.label = label;
1049 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1050 	switch (p->o_arg.claim) {
1051 	case NFS4_OPEN_CLAIM_NULL:
1052 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1053 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1054 		p->o_arg.fh = NFS_FH(dir);
1055 		break;
1056 	case NFS4_OPEN_CLAIM_PREVIOUS:
1057 	case NFS4_OPEN_CLAIM_FH:
1058 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1059 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1060 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1061 	}
1062 	if (attrs != NULL && attrs->ia_valid != 0) {
1063 		__u32 verf[2];
1064 
1065 		p->o_arg.u.attrs = &p->attrs;
1066 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1067 
1068 		verf[0] = jiffies;
1069 		verf[1] = current->pid;
1070 		memcpy(p->o_arg.u.verifier.data, verf,
1071 				sizeof(p->o_arg.u.verifier.data));
1072 	}
1073 	p->c_arg.fh = &p->o_res.fh;
1074 	p->c_arg.stateid = &p->o_res.stateid;
1075 	p->c_arg.seqid = p->o_arg.seqid;
1076 	nfs4_init_opendata_res(p);
1077 	kref_init(&p->kref);
1078 	return p;
1079 
1080 err_free_label:
1081 	nfs4_label_free(p->f_label);
1082 err_free_p:
1083 	kfree(p);
1084 err:
1085 	dput(parent);
1086 	return NULL;
1087 }
1088 
1089 static void nfs4_opendata_free(struct kref *kref)
1090 {
1091 	struct nfs4_opendata *p = container_of(kref,
1092 			struct nfs4_opendata, kref);
1093 	struct super_block *sb = p->dentry->d_sb;
1094 
1095 	nfs_free_seqid(p->o_arg.seqid);
1096 	if (p->state != NULL)
1097 		nfs4_put_open_state(p->state);
1098 	nfs4_put_state_owner(p->owner);
1099 
1100 	nfs4_label_free(p->f_label);
1101 
1102 	dput(p->dir);
1103 	dput(p->dentry);
1104 	nfs_sb_deactive(sb);
1105 	nfs_fattr_free_names(&p->f_attr);
1106 	kfree(p->f_attr.mdsthreshold);
1107 	kfree(p);
1108 }
1109 
1110 static void nfs4_opendata_put(struct nfs4_opendata *p)
1111 {
1112 	if (p != NULL)
1113 		kref_put(&p->kref, nfs4_opendata_free);
1114 }
1115 
1116 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1117 {
1118 	int ret;
1119 
1120 	ret = rpc_wait_for_completion_task(task);
1121 	return ret;
1122 }
1123 
1124 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1125 {
1126 	int ret = 0;
1127 
1128 	if (open_mode & (O_EXCL|O_TRUNC))
1129 		goto out;
1130 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1131 		case FMODE_READ:
1132 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1133 				&& state->n_rdonly != 0;
1134 			break;
1135 		case FMODE_WRITE:
1136 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1137 				&& state->n_wronly != 0;
1138 			break;
1139 		case FMODE_READ|FMODE_WRITE:
1140 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1141 				&& state->n_rdwr != 0;
1142 	}
1143 out:
1144 	return ret;
1145 }
1146 
1147 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1148 {
1149 	if (delegation == NULL)
1150 		return 0;
1151 	if ((delegation->type & fmode) != fmode)
1152 		return 0;
1153 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1154 		return 0;
1155 	nfs_mark_delegation_referenced(delegation);
1156 	return 1;
1157 }
1158 
1159 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1160 {
1161 	switch (fmode) {
1162 		case FMODE_WRITE:
1163 			state->n_wronly++;
1164 			break;
1165 		case FMODE_READ:
1166 			state->n_rdonly++;
1167 			break;
1168 		case FMODE_READ|FMODE_WRITE:
1169 			state->n_rdwr++;
1170 	}
1171 	nfs4_state_set_mode_locked(state, state->state | fmode);
1172 }
1173 
1174 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1175 {
1176 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1177 	bool need_recover = false;
1178 
1179 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1180 		need_recover = true;
1181 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1182 		need_recover = true;
1183 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1184 		need_recover = true;
1185 	if (need_recover)
1186 		nfs4_state_mark_reclaim_nograce(clp, state);
1187 }
1188 
1189 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1190 		nfs4_stateid *stateid)
1191 {
1192 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1193 		return true;
1194 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1195 		nfs_test_and_clear_all_open_stateid(state);
1196 		return true;
1197 	}
1198 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1199 		return true;
1200 	return false;
1201 }
1202 
1203 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1204 {
1205 	if (state->n_wronly)
1206 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1207 	if (state->n_rdonly)
1208 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1209 	if (state->n_rdwr)
1210 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1211 }
1212 
1213 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1214 		nfs4_stateid *stateid, fmode_t fmode)
1215 {
1216 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1217 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1218 	case FMODE_WRITE:
1219 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1220 		break;
1221 	case FMODE_READ:
1222 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1223 		break;
1224 	case 0:
1225 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1226 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1227 		clear_bit(NFS_OPEN_STATE, &state->flags);
1228 	}
1229 	if (stateid == NULL)
1230 		return;
1231 	/* Handle races with OPEN */
1232 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1233 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1234 		nfs_resync_open_stateid_locked(state);
1235 		return;
1236 	}
1237 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1238 		nfs4_stateid_copy(&state->stateid, stateid);
1239 	nfs4_stateid_copy(&state->open_stateid, stateid);
1240 }
1241 
1242 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1243 {
1244 	write_seqlock(&state->seqlock);
1245 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1246 	write_sequnlock(&state->seqlock);
1247 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1248 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1249 }
1250 
1251 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1252 {
1253 	switch (fmode) {
1254 		case FMODE_READ:
1255 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1256 			break;
1257 		case FMODE_WRITE:
1258 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1259 			break;
1260 		case FMODE_READ|FMODE_WRITE:
1261 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1262 	}
1263 	if (!nfs_need_update_open_stateid(state, stateid))
1264 		return;
1265 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1266 		nfs4_stateid_copy(&state->stateid, stateid);
1267 	nfs4_stateid_copy(&state->open_stateid, stateid);
1268 }
1269 
1270 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1271 {
1272 	/*
1273 	 * Protect the call to nfs4_state_set_mode_locked and
1274 	 * serialise the stateid update
1275 	 */
1276 	write_seqlock(&state->seqlock);
1277 	if (deleg_stateid != NULL) {
1278 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1279 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1280 	}
1281 	if (open_stateid != NULL)
1282 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1283 	write_sequnlock(&state->seqlock);
1284 	spin_lock(&state->owner->so_lock);
1285 	update_open_stateflags(state, fmode);
1286 	spin_unlock(&state->owner->so_lock);
1287 }
1288 
1289 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1290 {
1291 	struct nfs_inode *nfsi = NFS_I(state->inode);
1292 	struct nfs_delegation *deleg_cur;
1293 	int ret = 0;
1294 
1295 	fmode &= (FMODE_READ|FMODE_WRITE);
1296 
1297 	rcu_read_lock();
1298 	deleg_cur = rcu_dereference(nfsi->delegation);
1299 	if (deleg_cur == NULL)
1300 		goto no_delegation;
1301 
1302 	spin_lock(&deleg_cur->lock);
1303 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1304 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1305 	    (deleg_cur->type & fmode) != fmode)
1306 		goto no_delegation_unlock;
1307 
1308 	if (delegation == NULL)
1309 		delegation = &deleg_cur->stateid;
1310 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1311 		goto no_delegation_unlock;
1312 
1313 	nfs_mark_delegation_referenced(deleg_cur);
1314 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1315 	ret = 1;
1316 no_delegation_unlock:
1317 	spin_unlock(&deleg_cur->lock);
1318 no_delegation:
1319 	rcu_read_unlock();
1320 
1321 	if (!ret && open_stateid != NULL) {
1322 		__update_open_stateid(state, open_stateid, NULL, fmode);
1323 		ret = 1;
1324 	}
1325 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1326 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1327 
1328 	return ret;
1329 }
1330 
1331 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1332 		const nfs4_stateid *stateid)
1333 {
1334 	struct nfs4_state *state = lsp->ls_state;
1335 	bool ret = false;
1336 
1337 	spin_lock(&state->state_lock);
1338 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1339 		goto out_noupdate;
1340 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1341 		goto out_noupdate;
1342 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1343 	ret = true;
1344 out_noupdate:
1345 	spin_unlock(&state->state_lock);
1346 	return ret;
1347 }
1348 
1349 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1350 {
1351 	struct nfs_delegation *delegation;
1352 
1353 	rcu_read_lock();
1354 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1355 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1356 		rcu_read_unlock();
1357 		return;
1358 	}
1359 	rcu_read_unlock();
1360 	nfs4_inode_return_delegation(inode);
1361 }
1362 
1363 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1364 {
1365 	struct nfs4_state *state = opendata->state;
1366 	struct nfs_inode *nfsi = NFS_I(state->inode);
1367 	struct nfs_delegation *delegation;
1368 	int open_mode = opendata->o_arg.open_flags;
1369 	fmode_t fmode = opendata->o_arg.fmode;
1370 	nfs4_stateid stateid;
1371 	int ret = -EAGAIN;
1372 
1373 	for (;;) {
1374 		spin_lock(&state->owner->so_lock);
1375 		if (can_open_cached(state, fmode, open_mode)) {
1376 			update_open_stateflags(state, fmode);
1377 			spin_unlock(&state->owner->so_lock);
1378 			goto out_return_state;
1379 		}
1380 		spin_unlock(&state->owner->so_lock);
1381 		rcu_read_lock();
1382 		delegation = rcu_dereference(nfsi->delegation);
1383 		if (!can_open_delegated(delegation, fmode)) {
1384 			rcu_read_unlock();
1385 			break;
1386 		}
1387 		/* Save the delegation */
1388 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1389 		rcu_read_unlock();
1390 		nfs_release_seqid(opendata->o_arg.seqid);
1391 		if (!opendata->is_recover) {
1392 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1393 			if (ret != 0)
1394 				goto out;
1395 		}
1396 		ret = -EAGAIN;
1397 
1398 		/* Try to update the stateid using the delegation */
1399 		if (update_open_stateid(state, NULL, &stateid, fmode))
1400 			goto out_return_state;
1401 	}
1402 out:
1403 	return ERR_PTR(ret);
1404 out_return_state:
1405 	atomic_inc(&state->count);
1406 	return state;
1407 }
1408 
1409 static void
1410 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1411 {
1412 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1413 	struct nfs_delegation *delegation;
1414 	int delegation_flags = 0;
1415 
1416 	rcu_read_lock();
1417 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1418 	if (delegation)
1419 		delegation_flags = delegation->flags;
1420 	rcu_read_unlock();
1421 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1422 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1423 				   "returning a delegation for "
1424 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1425 				   clp->cl_hostname);
1426 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1427 		nfs_inode_set_delegation(state->inode,
1428 					 data->owner->so_cred,
1429 					 &data->o_res);
1430 	else
1431 		nfs_inode_reclaim_delegation(state->inode,
1432 					     data->owner->so_cred,
1433 					     &data->o_res);
1434 }
1435 
1436 /*
1437  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1438  * and update the nfs4_state.
1439  */
1440 static struct nfs4_state *
1441 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1442 {
1443 	struct inode *inode = data->state->inode;
1444 	struct nfs4_state *state = data->state;
1445 	int ret;
1446 
1447 	if (!data->rpc_done) {
1448 		if (data->rpc_status) {
1449 			ret = data->rpc_status;
1450 			goto err;
1451 		}
1452 		/* cached opens have already been processed */
1453 		goto update;
1454 	}
1455 
1456 	ret = nfs_refresh_inode(inode, &data->f_attr);
1457 	if (ret)
1458 		goto err;
1459 
1460 	if (data->o_res.delegation_type != 0)
1461 		nfs4_opendata_check_deleg(data, state);
1462 update:
1463 	update_open_stateid(state, &data->o_res.stateid, NULL,
1464 			    data->o_arg.fmode);
1465 	atomic_inc(&state->count);
1466 
1467 	return state;
1468 err:
1469 	return ERR_PTR(ret);
1470 
1471 }
1472 
1473 static struct nfs4_state *
1474 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1475 {
1476 	struct inode *inode;
1477 	struct nfs4_state *state = NULL;
1478 	int ret;
1479 
1480 	if (!data->rpc_done) {
1481 		state = nfs4_try_open_cached(data);
1482 		goto out;
1483 	}
1484 
1485 	ret = -EAGAIN;
1486 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1487 		goto err;
1488 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1489 	ret = PTR_ERR(inode);
1490 	if (IS_ERR(inode))
1491 		goto err;
1492 	ret = -ENOMEM;
1493 	state = nfs4_get_open_state(inode, data->owner);
1494 	if (state == NULL)
1495 		goto err_put_inode;
1496 	if (data->o_res.delegation_type != 0)
1497 		nfs4_opendata_check_deleg(data, state);
1498 	update_open_stateid(state, &data->o_res.stateid, NULL,
1499 			data->o_arg.fmode);
1500 	iput(inode);
1501 out:
1502 	nfs_release_seqid(data->o_arg.seqid);
1503 	return state;
1504 err_put_inode:
1505 	iput(inode);
1506 err:
1507 	return ERR_PTR(ret);
1508 }
1509 
1510 static struct nfs4_state *
1511 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1512 {
1513 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1514 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1515 	return _nfs4_opendata_to_nfs4_state(data);
1516 }
1517 
1518 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1519 {
1520 	struct nfs_inode *nfsi = NFS_I(state->inode);
1521 	struct nfs_open_context *ctx;
1522 
1523 	spin_lock(&state->inode->i_lock);
1524 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1525 		if (ctx->state != state)
1526 			continue;
1527 		get_nfs_open_context(ctx);
1528 		spin_unlock(&state->inode->i_lock);
1529 		return ctx;
1530 	}
1531 	spin_unlock(&state->inode->i_lock);
1532 	return ERR_PTR(-ENOENT);
1533 }
1534 
1535 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1536 		struct nfs4_state *state, enum open_claim_type4 claim)
1537 {
1538 	struct nfs4_opendata *opendata;
1539 
1540 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1541 			NULL, NULL, claim, GFP_NOFS);
1542 	if (opendata == NULL)
1543 		return ERR_PTR(-ENOMEM);
1544 	opendata->state = state;
1545 	atomic_inc(&state->count);
1546 	return opendata;
1547 }
1548 
1549 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1550 {
1551 	struct nfs4_state *newstate;
1552 	int ret;
1553 
1554 	opendata->o_arg.open_flags = 0;
1555 	opendata->o_arg.fmode = fmode;
1556 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1557 			NFS_SB(opendata->dentry->d_sb),
1558 			fmode, 0);
1559 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1560 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1561 	nfs4_init_opendata_res(opendata);
1562 	ret = _nfs4_recover_proc_open(opendata);
1563 	if (ret != 0)
1564 		return ret;
1565 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1566 	if (IS_ERR(newstate))
1567 		return PTR_ERR(newstate);
1568 	nfs4_close_state(newstate, fmode);
1569 	*res = newstate;
1570 	return 0;
1571 }
1572 
1573 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1574 {
1575 	struct nfs4_state *newstate;
1576 	int ret;
1577 
1578 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1579 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1580 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1581 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1582 	/* memory barrier prior to reading state->n_* */
1583 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1584 	clear_bit(NFS_OPEN_STATE, &state->flags);
1585 	smp_rmb();
1586 	if (state->n_rdwr != 0) {
1587 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1588 		if (ret != 0)
1589 			return ret;
1590 		if (newstate != state)
1591 			return -ESTALE;
1592 	}
1593 	if (state->n_wronly != 0) {
1594 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1595 		if (ret != 0)
1596 			return ret;
1597 		if (newstate != state)
1598 			return -ESTALE;
1599 	}
1600 	if (state->n_rdonly != 0) {
1601 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1602 		if (ret != 0)
1603 			return ret;
1604 		if (newstate != state)
1605 			return -ESTALE;
1606 	}
1607 	/*
1608 	 * We may have performed cached opens for all three recoveries.
1609 	 * Check if we need to update the current stateid.
1610 	 */
1611 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1612 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1613 		write_seqlock(&state->seqlock);
1614 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1615 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1616 		write_sequnlock(&state->seqlock);
1617 	}
1618 	return 0;
1619 }
1620 
1621 /*
1622  * OPEN_RECLAIM:
1623  * 	reclaim state on the server after a reboot.
1624  */
1625 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1626 {
1627 	struct nfs_delegation *delegation;
1628 	struct nfs4_opendata *opendata;
1629 	fmode_t delegation_type = 0;
1630 	int status;
1631 
1632 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1633 			NFS4_OPEN_CLAIM_PREVIOUS);
1634 	if (IS_ERR(opendata))
1635 		return PTR_ERR(opendata);
1636 	rcu_read_lock();
1637 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1638 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1639 		delegation_type = delegation->type;
1640 	rcu_read_unlock();
1641 	opendata->o_arg.u.delegation_type = delegation_type;
1642 	status = nfs4_open_recover(opendata, state);
1643 	nfs4_opendata_put(opendata);
1644 	return status;
1645 }
1646 
1647 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1648 {
1649 	struct nfs_server *server = NFS_SERVER(state->inode);
1650 	struct nfs4_exception exception = { };
1651 	int err;
1652 	do {
1653 		err = _nfs4_do_open_reclaim(ctx, state);
1654 		trace_nfs4_open_reclaim(ctx, 0, err);
1655 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1656 			continue;
1657 		if (err != -NFS4ERR_DELAY)
1658 			break;
1659 		nfs4_handle_exception(server, err, &exception);
1660 	} while (exception.retry);
1661 	return err;
1662 }
1663 
1664 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1665 {
1666 	struct nfs_open_context *ctx;
1667 	int ret;
1668 
1669 	ctx = nfs4_state_find_open_context(state);
1670 	if (IS_ERR(ctx))
1671 		return -EAGAIN;
1672 	ret = nfs4_do_open_reclaim(ctx, state);
1673 	put_nfs_open_context(ctx);
1674 	return ret;
1675 }
1676 
1677 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1678 {
1679 	switch (err) {
1680 		default:
1681 			printk(KERN_ERR "NFS: %s: unhandled error "
1682 					"%d.\n", __func__, err);
1683 		case 0:
1684 		case -ENOENT:
1685 		case -ESTALE:
1686 			break;
1687 		case -NFS4ERR_BADSESSION:
1688 		case -NFS4ERR_BADSLOT:
1689 		case -NFS4ERR_BAD_HIGH_SLOT:
1690 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1691 		case -NFS4ERR_DEADSESSION:
1692 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1693 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1694 			return -EAGAIN;
1695 		case -NFS4ERR_STALE_CLIENTID:
1696 		case -NFS4ERR_STALE_STATEID:
1697 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1698 		case -NFS4ERR_EXPIRED:
1699 			/* Don't recall a delegation if it was lost */
1700 			nfs4_schedule_lease_recovery(server->nfs_client);
1701 			return -EAGAIN;
1702 		case -NFS4ERR_MOVED:
1703 			nfs4_schedule_migration_recovery(server);
1704 			return -EAGAIN;
1705 		case -NFS4ERR_LEASE_MOVED:
1706 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1707 			return -EAGAIN;
1708 		case -NFS4ERR_DELEG_REVOKED:
1709 		case -NFS4ERR_ADMIN_REVOKED:
1710 		case -NFS4ERR_BAD_STATEID:
1711 		case -NFS4ERR_OPENMODE:
1712 			nfs_inode_find_state_and_recover(state->inode,
1713 					stateid);
1714 			nfs4_schedule_stateid_recovery(server, state);
1715 			return -EAGAIN;
1716 		case -NFS4ERR_DELAY:
1717 		case -NFS4ERR_GRACE:
1718 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1719 			ssleep(1);
1720 			return -EAGAIN;
1721 		case -ENOMEM:
1722 		case -NFS4ERR_DENIED:
1723 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1724 			return 0;
1725 	}
1726 	return err;
1727 }
1728 
1729 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1730 {
1731 	struct nfs_server *server = NFS_SERVER(state->inode);
1732 	struct nfs4_opendata *opendata;
1733 	int err;
1734 
1735 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1736 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1737 	if (IS_ERR(opendata))
1738 		return PTR_ERR(opendata);
1739 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1740 	err = nfs4_open_recover(opendata, state);
1741 	nfs4_opendata_put(opendata);
1742 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1743 }
1744 
1745 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1746 {
1747 	struct nfs4_opendata *data = calldata;
1748 
1749 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1750 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1751 }
1752 
1753 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1754 {
1755 	struct nfs4_opendata *data = calldata;
1756 
1757 	nfs40_sequence_done(task, &data->c_res.seq_res);
1758 
1759 	data->rpc_status = task->tk_status;
1760 	if (data->rpc_status == 0) {
1761 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1762 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1763 		renew_lease(data->o_res.server, data->timestamp);
1764 		data->rpc_done = 1;
1765 	}
1766 }
1767 
1768 static void nfs4_open_confirm_release(void *calldata)
1769 {
1770 	struct nfs4_opendata *data = calldata;
1771 	struct nfs4_state *state = NULL;
1772 
1773 	/* If this request hasn't been cancelled, do nothing */
1774 	if (data->cancelled == 0)
1775 		goto out_free;
1776 	/* In case of error, no cleanup! */
1777 	if (!data->rpc_done)
1778 		goto out_free;
1779 	state = nfs4_opendata_to_nfs4_state(data);
1780 	if (!IS_ERR(state))
1781 		nfs4_close_state(state, data->o_arg.fmode);
1782 out_free:
1783 	nfs4_opendata_put(data);
1784 }
1785 
1786 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1787 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1788 	.rpc_call_done = nfs4_open_confirm_done,
1789 	.rpc_release = nfs4_open_confirm_release,
1790 };
1791 
1792 /*
1793  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1794  */
1795 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1796 {
1797 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1798 	struct rpc_task *task;
1799 	struct  rpc_message msg = {
1800 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1801 		.rpc_argp = &data->c_arg,
1802 		.rpc_resp = &data->c_res,
1803 		.rpc_cred = data->owner->so_cred,
1804 	};
1805 	struct rpc_task_setup task_setup_data = {
1806 		.rpc_client = server->client,
1807 		.rpc_message = &msg,
1808 		.callback_ops = &nfs4_open_confirm_ops,
1809 		.callback_data = data,
1810 		.workqueue = nfsiod_workqueue,
1811 		.flags = RPC_TASK_ASYNC,
1812 	};
1813 	int status;
1814 
1815 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1816 	kref_get(&data->kref);
1817 	data->rpc_done = 0;
1818 	data->rpc_status = 0;
1819 	data->timestamp = jiffies;
1820 	task = rpc_run_task(&task_setup_data);
1821 	if (IS_ERR(task))
1822 		return PTR_ERR(task);
1823 	status = nfs4_wait_for_completion_rpc_task(task);
1824 	if (status != 0) {
1825 		data->cancelled = 1;
1826 		smp_wmb();
1827 	} else
1828 		status = data->rpc_status;
1829 	rpc_put_task(task);
1830 	return status;
1831 }
1832 
1833 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1834 {
1835 	struct nfs4_opendata *data = calldata;
1836 	struct nfs4_state_owner *sp = data->owner;
1837 	struct nfs_client *clp = sp->so_server->nfs_client;
1838 
1839 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1840 		goto out_wait;
1841 	/*
1842 	 * Check if we still need to send an OPEN call, or if we can use
1843 	 * a delegation instead.
1844 	 */
1845 	if (data->state != NULL) {
1846 		struct nfs_delegation *delegation;
1847 
1848 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1849 			goto out_no_action;
1850 		rcu_read_lock();
1851 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1852 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1853 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1854 		    can_open_delegated(delegation, data->o_arg.fmode))
1855 			goto unlock_no_action;
1856 		rcu_read_unlock();
1857 	}
1858 	/* Update client id. */
1859 	data->o_arg.clientid = clp->cl_clientid;
1860 	switch (data->o_arg.claim) {
1861 	case NFS4_OPEN_CLAIM_PREVIOUS:
1862 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1863 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1864 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1865 	case NFS4_OPEN_CLAIM_FH:
1866 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1867 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1868 	}
1869 	data->timestamp = jiffies;
1870 	if (nfs4_setup_sequence(data->o_arg.server,
1871 				&data->o_arg.seq_args,
1872 				&data->o_res.seq_res,
1873 				task) != 0)
1874 		nfs_release_seqid(data->o_arg.seqid);
1875 
1876 	/* Set the create mode (note dependency on the session type) */
1877 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1878 	if (data->o_arg.open_flags & O_EXCL) {
1879 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1880 		if (nfs4_has_persistent_session(clp))
1881 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1882 		else if (clp->cl_mvops->minor_version > 0)
1883 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1884 	}
1885 	return;
1886 unlock_no_action:
1887 	rcu_read_unlock();
1888 out_no_action:
1889 	task->tk_action = NULL;
1890 out_wait:
1891 	nfs4_sequence_done(task, &data->o_res.seq_res);
1892 }
1893 
1894 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1895 {
1896 	struct nfs4_opendata *data = calldata;
1897 
1898 	data->rpc_status = task->tk_status;
1899 
1900 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1901 		return;
1902 
1903 	if (task->tk_status == 0) {
1904 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1905 			switch (data->o_res.f_attr->mode & S_IFMT) {
1906 			case S_IFREG:
1907 				break;
1908 			case S_IFLNK:
1909 				data->rpc_status = -ELOOP;
1910 				break;
1911 			case S_IFDIR:
1912 				data->rpc_status = -EISDIR;
1913 				break;
1914 			default:
1915 				data->rpc_status = -ENOTDIR;
1916 			}
1917 		}
1918 		renew_lease(data->o_res.server, data->timestamp);
1919 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1920 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1921 	}
1922 	data->rpc_done = 1;
1923 }
1924 
1925 static void nfs4_open_release(void *calldata)
1926 {
1927 	struct nfs4_opendata *data = calldata;
1928 	struct nfs4_state *state = NULL;
1929 
1930 	/* If this request hasn't been cancelled, do nothing */
1931 	if (data->cancelled == 0)
1932 		goto out_free;
1933 	/* In case of error, no cleanup! */
1934 	if (data->rpc_status != 0 || !data->rpc_done)
1935 		goto out_free;
1936 	/* In case we need an open_confirm, no cleanup! */
1937 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1938 		goto out_free;
1939 	state = nfs4_opendata_to_nfs4_state(data);
1940 	if (!IS_ERR(state))
1941 		nfs4_close_state(state, data->o_arg.fmode);
1942 out_free:
1943 	nfs4_opendata_put(data);
1944 }
1945 
1946 static const struct rpc_call_ops nfs4_open_ops = {
1947 	.rpc_call_prepare = nfs4_open_prepare,
1948 	.rpc_call_done = nfs4_open_done,
1949 	.rpc_release = nfs4_open_release,
1950 };
1951 
1952 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1953 {
1954 	struct inode *dir = data->dir->d_inode;
1955 	struct nfs_server *server = NFS_SERVER(dir);
1956 	struct nfs_openargs *o_arg = &data->o_arg;
1957 	struct nfs_openres *o_res = &data->o_res;
1958 	struct rpc_task *task;
1959 	struct rpc_message msg = {
1960 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1961 		.rpc_argp = o_arg,
1962 		.rpc_resp = o_res,
1963 		.rpc_cred = data->owner->so_cred,
1964 	};
1965 	struct rpc_task_setup task_setup_data = {
1966 		.rpc_client = server->client,
1967 		.rpc_message = &msg,
1968 		.callback_ops = &nfs4_open_ops,
1969 		.callback_data = data,
1970 		.workqueue = nfsiod_workqueue,
1971 		.flags = RPC_TASK_ASYNC,
1972 	};
1973 	int status;
1974 
1975 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1976 	kref_get(&data->kref);
1977 	data->rpc_done = 0;
1978 	data->rpc_status = 0;
1979 	data->cancelled = 0;
1980 	data->is_recover = 0;
1981 	if (isrecover) {
1982 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1983 		data->is_recover = 1;
1984 	}
1985 	task = rpc_run_task(&task_setup_data);
1986         if (IS_ERR(task))
1987                 return PTR_ERR(task);
1988         status = nfs4_wait_for_completion_rpc_task(task);
1989         if (status != 0) {
1990                 data->cancelled = 1;
1991                 smp_wmb();
1992         } else
1993                 status = data->rpc_status;
1994         rpc_put_task(task);
1995 
1996 	return status;
1997 }
1998 
1999 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2000 {
2001 	struct inode *dir = data->dir->d_inode;
2002 	struct nfs_openres *o_res = &data->o_res;
2003         int status;
2004 
2005 	status = nfs4_run_open_task(data, 1);
2006 	if (status != 0 || !data->rpc_done)
2007 		return status;
2008 
2009 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2010 
2011 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2012 		status = _nfs4_proc_open_confirm(data);
2013 		if (status != 0)
2014 			return status;
2015 	}
2016 
2017 	return status;
2018 }
2019 
2020 /*
2021  * Additional permission checks in order to distinguish between an
2022  * open for read, and an open for execute. This works around the
2023  * fact that NFSv4 OPEN treats read and execute permissions as being
2024  * the same.
2025  * Note that in the non-execute case, we want to turn off permission
2026  * checking if we just created a new file (POSIX open() semantics).
2027  */
2028 static int nfs4_opendata_access(struct rpc_cred *cred,
2029 				struct nfs4_opendata *opendata,
2030 				struct nfs4_state *state, fmode_t fmode,
2031 				int openflags)
2032 {
2033 	struct nfs_access_entry cache;
2034 	u32 mask;
2035 
2036 	/* access call failed or for some reason the server doesn't
2037 	 * support any access modes -- defer access call until later */
2038 	if (opendata->o_res.access_supported == 0)
2039 		return 0;
2040 
2041 	mask = 0;
2042 	/*
2043 	 * Use openflags to check for exec, because fmode won't
2044 	 * always have FMODE_EXEC set when file open for exec.
2045 	 */
2046 	if (openflags & __FMODE_EXEC) {
2047 		/* ONLY check for exec rights */
2048 		mask = MAY_EXEC;
2049 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2050 		mask = MAY_READ;
2051 
2052 	cache.cred = cred;
2053 	cache.jiffies = jiffies;
2054 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2055 	nfs_access_add_cache(state->inode, &cache);
2056 
2057 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2058 		return 0;
2059 
2060 	/* even though OPEN succeeded, access is denied. Close the file */
2061 	nfs4_close_state(state, fmode);
2062 	return -EACCES;
2063 }
2064 
2065 /*
2066  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2067  */
2068 static int _nfs4_proc_open(struct nfs4_opendata *data)
2069 {
2070 	struct inode *dir = data->dir->d_inode;
2071 	struct nfs_server *server = NFS_SERVER(dir);
2072 	struct nfs_openargs *o_arg = &data->o_arg;
2073 	struct nfs_openres *o_res = &data->o_res;
2074 	int status;
2075 
2076 	status = nfs4_run_open_task(data, 0);
2077 	if (!data->rpc_done)
2078 		return status;
2079 	if (status != 0) {
2080 		if (status == -NFS4ERR_BADNAME &&
2081 				!(o_arg->open_flags & O_CREAT))
2082 			return -ENOENT;
2083 		return status;
2084 	}
2085 
2086 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2087 
2088 	if (o_arg->open_flags & O_CREAT) {
2089 		update_changeattr(dir, &o_res->cinfo);
2090 		if (o_arg->open_flags & O_EXCL)
2091 			data->file_created = 1;
2092 		else if (o_res->cinfo.before != o_res->cinfo.after)
2093 			data->file_created = 1;
2094 	}
2095 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2096 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2097 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2098 		status = _nfs4_proc_open_confirm(data);
2099 		if (status != 0)
2100 			return status;
2101 	}
2102 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2103 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2104 	return 0;
2105 }
2106 
2107 static int nfs4_recover_expired_lease(struct nfs_server *server)
2108 {
2109 	return nfs4_client_recover_expired_lease(server->nfs_client);
2110 }
2111 
2112 /*
2113  * OPEN_EXPIRED:
2114  * 	reclaim state on the server after a network partition.
2115  * 	Assumes caller holds the appropriate lock
2116  */
2117 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2118 {
2119 	struct nfs4_opendata *opendata;
2120 	int ret;
2121 
2122 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2123 			NFS4_OPEN_CLAIM_FH);
2124 	if (IS_ERR(opendata))
2125 		return PTR_ERR(opendata);
2126 	ret = nfs4_open_recover(opendata, state);
2127 	if (ret == -ESTALE)
2128 		d_drop(ctx->dentry);
2129 	nfs4_opendata_put(opendata);
2130 	return ret;
2131 }
2132 
2133 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2134 {
2135 	struct nfs_server *server = NFS_SERVER(state->inode);
2136 	struct nfs4_exception exception = { };
2137 	int err;
2138 
2139 	do {
2140 		err = _nfs4_open_expired(ctx, state);
2141 		trace_nfs4_open_expired(ctx, 0, err);
2142 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2143 			continue;
2144 		switch (err) {
2145 		default:
2146 			goto out;
2147 		case -NFS4ERR_GRACE:
2148 		case -NFS4ERR_DELAY:
2149 			nfs4_handle_exception(server, err, &exception);
2150 			err = 0;
2151 		}
2152 	} while (exception.retry);
2153 out:
2154 	return err;
2155 }
2156 
2157 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2158 {
2159 	struct nfs_open_context *ctx;
2160 	int ret;
2161 
2162 	ctx = nfs4_state_find_open_context(state);
2163 	if (IS_ERR(ctx))
2164 		return -EAGAIN;
2165 	ret = nfs4_do_open_expired(ctx, state);
2166 	put_nfs_open_context(ctx);
2167 	return ret;
2168 }
2169 
2170 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2171 {
2172 	nfs_remove_bad_delegation(state->inode);
2173 	write_seqlock(&state->seqlock);
2174 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2175 	write_sequnlock(&state->seqlock);
2176 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2177 }
2178 
2179 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2180 {
2181 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2182 		nfs_finish_clear_delegation_stateid(state);
2183 }
2184 
2185 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2186 {
2187 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2188 	nfs40_clear_delegation_stateid(state);
2189 	return nfs4_open_expired(sp, state);
2190 }
2191 
2192 #if defined(CONFIG_NFS_V4_1)
2193 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2194 {
2195 	struct nfs_server *server = NFS_SERVER(state->inode);
2196 	nfs4_stateid stateid;
2197 	struct nfs_delegation *delegation;
2198 	struct rpc_cred *cred;
2199 	int status;
2200 
2201 	/* Get the delegation credential for use by test/free_stateid */
2202 	rcu_read_lock();
2203 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2204 	if (delegation == NULL) {
2205 		rcu_read_unlock();
2206 		return;
2207 	}
2208 
2209 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2210 	cred = get_rpccred(delegation->cred);
2211 	rcu_read_unlock();
2212 	status = nfs41_test_stateid(server, &stateid, cred);
2213 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2214 
2215 	if (status != NFS_OK) {
2216 		/* Free the stateid unless the server explicitly
2217 		 * informs us the stateid is unrecognized. */
2218 		if (status != -NFS4ERR_BAD_STATEID)
2219 			nfs41_free_stateid(server, &stateid, cred);
2220 		nfs_finish_clear_delegation_stateid(state);
2221 	}
2222 
2223 	put_rpccred(cred);
2224 }
2225 
2226 /**
2227  * nfs41_check_open_stateid - possibly free an open stateid
2228  *
2229  * @state: NFSv4 state for an inode
2230  *
2231  * Returns NFS_OK if recovery for this stateid is now finished.
2232  * Otherwise a negative NFS4ERR value is returned.
2233  */
2234 static int nfs41_check_open_stateid(struct nfs4_state *state)
2235 {
2236 	struct nfs_server *server = NFS_SERVER(state->inode);
2237 	nfs4_stateid *stateid = &state->open_stateid;
2238 	struct rpc_cred *cred = state->owner->so_cred;
2239 	int status;
2240 
2241 	/* If a state reset has been done, test_stateid is unneeded */
2242 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2243 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2244 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2245 		return -NFS4ERR_BAD_STATEID;
2246 
2247 	status = nfs41_test_stateid(server, stateid, cred);
2248 	trace_nfs4_test_open_stateid(state, NULL, status);
2249 	if (status != NFS_OK) {
2250 		/* Free the stateid unless the server explicitly
2251 		 * informs us the stateid is unrecognized. */
2252 		if (status != -NFS4ERR_BAD_STATEID)
2253 			nfs41_free_stateid(server, stateid, cred);
2254 
2255 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2256 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2257 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2258 		clear_bit(NFS_OPEN_STATE, &state->flags);
2259 	}
2260 	return status;
2261 }
2262 
2263 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2264 {
2265 	int status;
2266 
2267 	nfs41_check_delegation_stateid(state);
2268 	status = nfs41_check_open_stateid(state);
2269 	if (status != NFS_OK)
2270 		status = nfs4_open_expired(sp, state);
2271 	return status;
2272 }
2273 #endif
2274 
2275 /*
2276  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2277  * fields corresponding to attributes that were used to store the verifier.
2278  * Make sure we clobber those fields in the later setattr call
2279  */
2280 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2281 {
2282 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2283 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2284 		sattr->ia_valid |= ATTR_ATIME;
2285 
2286 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2287 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2288 		sattr->ia_valid |= ATTR_MTIME;
2289 }
2290 
2291 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2292 		fmode_t fmode,
2293 		int flags,
2294 		struct nfs_open_context *ctx)
2295 {
2296 	struct nfs4_state_owner *sp = opendata->owner;
2297 	struct nfs_server *server = sp->so_server;
2298 	struct dentry *dentry;
2299 	struct nfs4_state *state;
2300 	unsigned int seq;
2301 	int ret;
2302 
2303 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2304 
2305 	ret = _nfs4_proc_open(opendata);
2306 	if (ret != 0)
2307 		goto out;
2308 
2309 	state = nfs4_opendata_to_nfs4_state(opendata);
2310 	ret = PTR_ERR(state);
2311 	if (IS_ERR(state))
2312 		goto out;
2313 	if (server->caps & NFS_CAP_POSIX_LOCK)
2314 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2315 
2316 	dentry = opendata->dentry;
2317 	if (dentry->d_inode == NULL) {
2318 		/* FIXME: Is this d_drop() ever needed? */
2319 		d_drop(dentry);
2320 		dentry = d_add_unique(dentry, igrab(state->inode));
2321 		if (dentry == NULL) {
2322 			dentry = opendata->dentry;
2323 		} else if (dentry != ctx->dentry) {
2324 			dput(ctx->dentry);
2325 			ctx->dentry = dget(dentry);
2326 		}
2327 		nfs_set_verifier(dentry,
2328 				nfs_save_change_attribute(opendata->dir->d_inode));
2329 	}
2330 
2331 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2332 	if (ret != 0)
2333 		goto out;
2334 
2335 	ctx->state = state;
2336 	if (dentry->d_inode == state->inode) {
2337 		nfs_inode_attach_open_context(ctx);
2338 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2339 			nfs4_schedule_stateid_recovery(server, state);
2340 	}
2341 out:
2342 	return ret;
2343 }
2344 
2345 /*
2346  * Returns a referenced nfs4_state
2347  */
2348 static int _nfs4_do_open(struct inode *dir,
2349 			struct nfs_open_context *ctx,
2350 			int flags,
2351 			struct iattr *sattr,
2352 			struct nfs4_label *label,
2353 			int *opened)
2354 {
2355 	struct nfs4_state_owner  *sp;
2356 	struct nfs4_state     *state = NULL;
2357 	struct nfs_server       *server = NFS_SERVER(dir);
2358 	struct nfs4_opendata *opendata;
2359 	struct dentry *dentry = ctx->dentry;
2360 	struct rpc_cred *cred = ctx->cred;
2361 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2362 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2363 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2364 	struct nfs4_label *olabel = NULL;
2365 	int status;
2366 
2367 	/* Protect against reboot recovery conflicts */
2368 	status = -ENOMEM;
2369 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2370 	if (sp == NULL) {
2371 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2372 		goto out_err;
2373 	}
2374 	status = nfs4_recover_expired_lease(server);
2375 	if (status != 0)
2376 		goto err_put_state_owner;
2377 	if (dentry->d_inode != NULL)
2378 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2379 	status = -ENOMEM;
2380 	if (dentry->d_inode)
2381 		claim = NFS4_OPEN_CLAIM_FH;
2382 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2383 			label, claim, GFP_KERNEL);
2384 	if (opendata == NULL)
2385 		goto err_put_state_owner;
2386 
2387 	if (label) {
2388 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2389 		if (IS_ERR(olabel)) {
2390 			status = PTR_ERR(olabel);
2391 			goto err_opendata_put;
2392 		}
2393 	}
2394 
2395 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2396 		if (!opendata->f_attr.mdsthreshold) {
2397 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2398 			if (!opendata->f_attr.mdsthreshold)
2399 				goto err_free_label;
2400 		}
2401 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2402 	}
2403 	if (dentry->d_inode != NULL)
2404 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2405 
2406 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2407 	if (status != 0)
2408 		goto err_free_label;
2409 	state = ctx->state;
2410 
2411 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2412 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2413 		nfs4_exclusive_attrset(opendata, sattr);
2414 
2415 		nfs_fattr_init(opendata->o_res.f_attr);
2416 		status = nfs4_do_setattr(state->inode, cred,
2417 				opendata->o_res.f_attr, sattr,
2418 				state, label, olabel);
2419 		if (status == 0) {
2420 			nfs_setattr_update_inode(state->inode, sattr,
2421 					opendata->o_res.f_attr);
2422 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2423 		}
2424 	}
2425 	if (opendata->file_created)
2426 		*opened |= FILE_CREATED;
2427 
2428 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2429 		*ctx_th = opendata->f_attr.mdsthreshold;
2430 		opendata->f_attr.mdsthreshold = NULL;
2431 	}
2432 
2433 	nfs4_label_free(olabel);
2434 
2435 	nfs4_opendata_put(opendata);
2436 	nfs4_put_state_owner(sp);
2437 	return 0;
2438 err_free_label:
2439 	nfs4_label_free(olabel);
2440 err_opendata_put:
2441 	nfs4_opendata_put(opendata);
2442 err_put_state_owner:
2443 	nfs4_put_state_owner(sp);
2444 out_err:
2445 	return status;
2446 }
2447 
2448 
2449 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2450 					struct nfs_open_context *ctx,
2451 					int flags,
2452 					struct iattr *sattr,
2453 					struct nfs4_label *label,
2454 					int *opened)
2455 {
2456 	struct nfs_server *server = NFS_SERVER(dir);
2457 	struct nfs4_exception exception = { };
2458 	struct nfs4_state *res;
2459 	int status;
2460 
2461 	do {
2462 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2463 		res = ctx->state;
2464 		trace_nfs4_open_file(ctx, flags, status);
2465 		if (status == 0)
2466 			break;
2467 		/* NOTE: BAD_SEQID means the server and client disagree about the
2468 		 * book-keeping w.r.t. state-changing operations
2469 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2470 		 * It is actually a sign of a bug on the client or on the server.
2471 		 *
2472 		 * If we receive a BAD_SEQID error in the particular case of
2473 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2474 		 * have unhashed the old state_owner for us, and that we can
2475 		 * therefore safely retry using a new one. We should still warn
2476 		 * the user though...
2477 		 */
2478 		if (status == -NFS4ERR_BAD_SEQID) {
2479 			pr_warn_ratelimited("NFS: v4 server %s "
2480 					" returned a bad sequence-id error!\n",
2481 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2482 			exception.retry = 1;
2483 			continue;
2484 		}
2485 		/*
2486 		 * BAD_STATEID on OPEN means that the server cancelled our
2487 		 * state before it received the OPEN_CONFIRM.
2488 		 * Recover by retrying the request as per the discussion
2489 		 * on Page 181 of RFC3530.
2490 		 */
2491 		if (status == -NFS4ERR_BAD_STATEID) {
2492 			exception.retry = 1;
2493 			continue;
2494 		}
2495 		if (status == -EAGAIN) {
2496 			/* We must have found a delegation */
2497 			exception.retry = 1;
2498 			continue;
2499 		}
2500 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2501 			continue;
2502 		res = ERR_PTR(nfs4_handle_exception(server,
2503 					status, &exception));
2504 	} while (exception.retry);
2505 	return res;
2506 }
2507 
2508 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2509 			    struct nfs_fattr *fattr, struct iattr *sattr,
2510 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2511 			    struct nfs4_label *olabel)
2512 {
2513 	struct nfs_server *server = NFS_SERVER(inode);
2514         struct nfs_setattrargs  arg = {
2515                 .fh             = NFS_FH(inode),
2516                 .iap            = sattr,
2517 		.server		= server,
2518 		.bitmask = server->attr_bitmask,
2519 		.label		= ilabel,
2520         };
2521         struct nfs_setattrres  res = {
2522 		.fattr		= fattr,
2523 		.label		= olabel,
2524 		.server		= server,
2525         };
2526         struct rpc_message msg = {
2527 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2528 		.rpc_argp	= &arg,
2529 		.rpc_resp	= &res,
2530 		.rpc_cred	= cred,
2531         };
2532 	unsigned long timestamp = jiffies;
2533 	fmode_t fmode;
2534 	bool truncate;
2535 	int status;
2536 
2537 	arg.bitmask = nfs4_bitmask(server, ilabel);
2538 	if (ilabel)
2539 		arg.bitmask = nfs4_bitmask(server, olabel);
2540 
2541 	nfs_fattr_init(fattr);
2542 
2543 	/* Servers should only apply open mode checks for file size changes */
2544 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2545 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2546 
2547 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2548 		/* Use that stateid */
2549 	} else if (truncate && state != NULL) {
2550 		struct nfs_lockowner lockowner = {
2551 			.l_owner = current->files,
2552 			.l_pid = current->tgid,
2553 		};
2554 		if (!nfs4_valid_open_stateid(state))
2555 			return -EBADF;
2556 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2557 				&lockowner) == -EIO)
2558 			return -EBADF;
2559 	} else
2560 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2561 
2562 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2563 	if (status == 0 && state != NULL)
2564 		renew_lease(server, timestamp);
2565 	return status;
2566 }
2567 
2568 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2569 			   struct nfs_fattr *fattr, struct iattr *sattr,
2570 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2571 			   struct nfs4_label *olabel)
2572 {
2573 	struct nfs_server *server = NFS_SERVER(inode);
2574 	struct nfs4_exception exception = {
2575 		.state = state,
2576 		.inode = inode,
2577 	};
2578 	int err;
2579 	do {
2580 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2581 		trace_nfs4_setattr(inode, err);
2582 		switch (err) {
2583 		case -NFS4ERR_OPENMODE:
2584 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2585 				pr_warn_once("NFSv4: server %s is incorrectly "
2586 						"applying open mode checks to "
2587 						"a SETATTR that is not "
2588 						"changing file size.\n",
2589 						server->nfs_client->cl_hostname);
2590 			}
2591 			if (state && !(state->state & FMODE_WRITE)) {
2592 				err = -EBADF;
2593 				if (sattr->ia_valid & ATTR_OPEN)
2594 					err = -EACCES;
2595 				goto out;
2596 			}
2597 		}
2598 		err = nfs4_handle_exception(server, err, &exception);
2599 	} while (exception.retry);
2600 out:
2601 	return err;
2602 }
2603 
2604 struct nfs4_closedata {
2605 	struct inode *inode;
2606 	struct nfs4_state *state;
2607 	struct nfs_closeargs arg;
2608 	struct nfs_closeres res;
2609 	struct nfs_fattr fattr;
2610 	unsigned long timestamp;
2611 	bool roc;
2612 	u32 roc_barrier;
2613 };
2614 
2615 static void nfs4_free_closedata(void *data)
2616 {
2617 	struct nfs4_closedata *calldata = data;
2618 	struct nfs4_state_owner *sp = calldata->state->owner;
2619 	struct super_block *sb = calldata->state->inode->i_sb;
2620 
2621 	if (calldata->roc)
2622 		pnfs_roc_release(calldata->state->inode);
2623 	nfs4_put_open_state(calldata->state);
2624 	nfs_free_seqid(calldata->arg.seqid);
2625 	nfs4_put_state_owner(sp);
2626 	nfs_sb_deactive(sb);
2627 	kfree(calldata);
2628 }
2629 
2630 static void nfs4_close_done(struct rpc_task *task, void *data)
2631 {
2632 	struct nfs4_closedata *calldata = data;
2633 	struct nfs4_state *state = calldata->state;
2634 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2635 	nfs4_stateid *res_stateid = NULL;
2636 
2637 	dprintk("%s: begin!\n", __func__);
2638 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2639 		return;
2640 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2641         /* hmm. we are done with the inode, and in the process of freeing
2642 	 * the state_owner. we keep this around to process errors
2643 	 */
2644 	switch (task->tk_status) {
2645 		case 0:
2646 			res_stateid = &calldata->res.stateid;
2647 			if (calldata->arg.fmode == 0 && calldata->roc)
2648 				pnfs_roc_set_barrier(state->inode,
2649 						     calldata->roc_barrier);
2650 			renew_lease(server, calldata->timestamp);
2651 			break;
2652 		case -NFS4ERR_ADMIN_REVOKED:
2653 		case -NFS4ERR_STALE_STATEID:
2654 		case -NFS4ERR_OLD_STATEID:
2655 		case -NFS4ERR_BAD_STATEID:
2656 		case -NFS4ERR_EXPIRED:
2657 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2658 						&state->open_stateid)) {
2659 				rpc_restart_call_prepare(task);
2660 				goto out_release;
2661 			}
2662 			if (calldata->arg.fmode == 0)
2663 				break;
2664 		default:
2665 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2666 				rpc_restart_call_prepare(task);
2667 				goto out_release;
2668 			}
2669 	}
2670 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2671 out_release:
2672 	nfs_release_seqid(calldata->arg.seqid);
2673 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2674 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2675 }
2676 
2677 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2678 {
2679 	struct nfs4_closedata *calldata = data;
2680 	struct nfs4_state *state = calldata->state;
2681 	struct inode *inode = calldata->inode;
2682 	bool is_rdonly, is_wronly, is_rdwr;
2683 	int call_close = 0;
2684 
2685 	dprintk("%s: begin!\n", __func__);
2686 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2687 		goto out_wait;
2688 
2689 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2690 	spin_lock(&state->owner->so_lock);
2691 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2692 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2693 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2694 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2695 	/* Calculate the change in open mode */
2696 	calldata->arg.fmode = 0;
2697 	if (state->n_rdwr == 0) {
2698 		if (state->n_rdonly == 0)
2699 			call_close |= is_rdonly;
2700 		else if (is_rdonly)
2701 			calldata->arg.fmode |= FMODE_READ;
2702 		if (state->n_wronly == 0)
2703 			call_close |= is_wronly;
2704 		else if (is_wronly)
2705 			calldata->arg.fmode |= FMODE_WRITE;
2706 	} else if (is_rdwr)
2707 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2708 
2709 	if (calldata->arg.fmode == 0)
2710 		call_close |= is_rdwr;
2711 
2712 	if (!nfs4_valid_open_stateid(state))
2713 		call_close = 0;
2714 	spin_unlock(&state->owner->so_lock);
2715 
2716 	if (!call_close) {
2717 		/* Note: exit _without_ calling nfs4_close_done */
2718 		goto out_no_action;
2719 	}
2720 
2721 	if (calldata->arg.fmode == 0) {
2722 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2723 		if (calldata->roc &&
2724 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2725 			nfs_release_seqid(calldata->arg.seqid);
2726 			goto out_wait;
2727 		    }
2728 	}
2729 	calldata->arg.share_access =
2730 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2731 				calldata->arg.fmode, 0);
2732 
2733 	nfs_fattr_init(calldata->res.fattr);
2734 	calldata->timestamp = jiffies;
2735 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2736 				&calldata->arg.seq_args,
2737 				&calldata->res.seq_res,
2738 				task) != 0)
2739 		nfs_release_seqid(calldata->arg.seqid);
2740 	dprintk("%s: done!\n", __func__);
2741 	return;
2742 out_no_action:
2743 	task->tk_action = NULL;
2744 out_wait:
2745 	nfs4_sequence_done(task, &calldata->res.seq_res);
2746 }
2747 
2748 static const struct rpc_call_ops nfs4_close_ops = {
2749 	.rpc_call_prepare = nfs4_close_prepare,
2750 	.rpc_call_done = nfs4_close_done,
2751 	.rpc_release = nfs4_free_closedata,
2752 };
2753 
2754 static bool nfs4_roc(struct inode *inode)
2755 {
2756 	if (!nfs_have_layout(inode))
2757 		return false;
2758 	return pnfs_roc(inode);
2759 }
2760 
2761 /*
2762  * It is possible for data to be read/written from a mem-mapped file
2763  * after the sys_close call (which hits the vfs layer as a flush).
2764  * This means that we can't safely call nfsv4 close on a file until
2765  * the inode is cleared. This in turn means that we are not good
2766  * NFSv4 citizens - we do not indicate to the server to update the file's
2767  * share state even when we are done with one of the three share
2768  * stateid's in the inode.
2769  *
2770  * NOTE: Caller must be holding the sp->so_owner semaphore!
2771  */
2772 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2773 {
2774 	struct nfs_server *server = NFS_SERVER(state->inode);
2775 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2776 	struct nfs4_closedata *calldata;
2777 	struct nfs4_state_owner *sp = state->owner;
2778 	struct rpc_task *task;
2779 	struct rpc_message msg = {
2780 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2781 		.rpc_cred = state->owner->so_cred,
2782 	};
2783 	struct rpc_task_setup task_setup_data = {
2784 		.rpc_client = server->client,
2785 		.rpc_message = &msg,
2786 		.callback_ops = &nfs4_close_ops,
2787 		.workqueue = nfsiod_workqueue,
2788 		.flags = RPC_TASK_ASYNC,
2789 	};
2790 	int status = -ENOMEM;
2791 
2792 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2793 		&task_setup_data.rpc_client, &msg);
2794 
2795 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2796 	if (calldata == NULL)
2797 		goto out;
2798 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2799 	calldata->inode = state->inode;
2800 	calldata->state = state;
2801 	calldata->arg.fh = NFS_FH(state->inode);
2802 	/* Serialization for the sequence id */
2803 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2804 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2805 	if (IS_ERR(calldata->arg.seqid))
2806 		goto out_free_calldata;
2807 	calldata->arg.fmode = 0;
2808 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2809 	calldata->res.fattr = &calldata->fattr;
2810 	calldata->res.seqid = calldata->arg.seqid;
2811 	calldata->res.server = server;
2812 	calldata->roc = nfs4_roc(state->inode);
2813 	nfs_sb_active(calldata->inode->i_sb);
2814 
2815 	msg.rpc_argp = &calldata->arg;
2816 	msg.rpc_resp = &calldata->res;
2817 	task_setup_data.callback_data = calldata;
2818 	task = rpc_run_task(&task_setup_data);
2819 	if (IS_ERR(task))
2820 		return PTR_ERR(task);
2821 	status = 0;
2822 	if (wait)
2823 		status = rpc_wait_for_completion_task(task);
2824 	rpc_put_task(task);
2825 	return status;
2826 out_free_calldata:
2827 	kfree(calldata);
2828 out:
2829 	nfs4_put_open_state(state);
2830 	nfs4_put_state_owner(sp);
2831 	return status;
2832 }
2833 
2834 static struct inode *
2835 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2836 		int open_flags, struct iattr *attr, int *opened)
2837 {
2838 	struct nfs4_state *state;
2839 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2840 
2841 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2842 
2843 	/* Protect against concurrent sillydeletes */
2844 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2845 
2846 	nfs4_label_release_security(label);
2847 
2848 	if (IS_ERR(state))
2849 		return ERR_CAST(state);
2850 	return state->inode;
2851 }
2852 
2853 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2854 {
2855 	if (ctx->state == NULL)
2856 		return;
2857 	if (is_sync)
2858 		nfs4_close_sync(ctx->state, ctx->mode);
2859 	else
2860 		nfs4_close_state(ctx->state, ctx->mode);
2861 }
2862 
2863 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2864 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2865 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2866 
2867 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2868 {
2869 	struct nfs4_server_caps_arg args = {
2870 		.fhandle = fhandle,
2871 	};
2872 	struct nfs4_server_caps_res res = {};
2873 	struct rpc_message msg = {
2874 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2875 		.rpc_argp = &args,
2876 		.rpc_resp = &res,
2877 	};
2878 	int status;
2879 
2880 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2881 	if (status == 0) {
2882 		/* Sanity check the server answers */
2883 		switch (server->nfs_client->cl_minorversion) {
2884 		case 0:
2885 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2886 			res.attr_bitmask[2] = 0;
2887 			break;
2888 		case 1:
2889 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2890 			break;
2891 		case 2:
2892 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2893 		}
2894 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2895 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2896 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2897 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2898 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2899 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2900 				NFS_CAP_SECURITY_LABEL);
2901 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2902 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2903 			server->caps |= NFS_CAP_ACLS;
2904 		if (res.has_links != 0)
2905 			server->caps |= NFS_CAP_HARDLINKS;
2906 		if (res.has_symlinks != 0)
2907 			server->caps |= NFS_CAP_SYMLINKS;
2908 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2909 			server->caps |= NFS_CAP_FILEID;
2910 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2911 			server->caps |= NFS_CAP_MODE;
2912 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2913 			server->caps |= NFS_CAP_NLINK;
2914 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2915 			server->caps |= NFS_CAP_OWNER;
2916 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2917 			server->caps |= NFS_CAP_OWNER_GROUP;
2918 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2919 			server->caps |= NFS_CAP_ATIME;
2920 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2921 			server->caps |= NFS_CAP_CTIME;
2922 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2923 			server->caps |= NFS_CAP_MTIME;
2924 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2925 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2926 			server->caps |= NFS_CAP_SECURITY_LABEL;
2927 #endif
2928 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2929 				sizeof(server->attr_bitmask));
2930 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2931 
2932 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2933 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2934 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2935 		server->cache_consistency_bitmask[2] = 0;
2936 		server->acl_bitmask = res.acl_bitmask;
2937 		server->fh_expire_type = res.fh_expire_type;
2938 	}
2939 
2940 	return status;
2941 }
2942 
2943 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2944 {
2945 	struct nfs4_exception exception = { };
2946 	int err;
2947 	do {
2948 		err = nfs4_handle_exception(server,
2949 				_nfs4_server_capabilities(server, fhandle),
2950 				&exception);
2951 	} while (exception.retry);
2952 	return err;
2953 }
2954 
2955 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2956 		struct nfs_fsinfo *info)
2957 {
2958 	u32 bitmask[3];
2959 	struct nfs4_lookup_root_arg args = {
2960 		.bitmask = bitmask,
2961 	};
2962 	struct nfs4_lookup_res res = {
2963 		.server = server,
2964 		.fattr = info->fattr,
2965 		.fh = fhandle,
2966 	};
2967 	struct rpc_message msg = {
2968 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2969 		.rpc_argp = &args,
2970 		.rpc_resp = &res,
2971 	};
2972 
2973 	bitmask[0] = nfs4_fattr_bitmap[0];
2974 	bitmask[1] = nfs4_fattr_bitmap[1];
2975 	/*
2976 	 * Process the label in the upcoming getfattr
2977 	 */
2978 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2979 
2980 	nfs_fattr_init(info->fattr);
2981 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2982 }
2983 
2984 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2985 		struct nfs_fsinfo *info)
2986 {
2987 	struct nfs4_exception exception = { };
2988 	int err;
2989 	do {
2990 		err = _nfs4_lookup_root(server, fhandle, info);
2991 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2992 		switch (err) {
2993 		case 0:
2994 		case -NFS4ERR_WRONGSEC:
2995 			goto out;
2996 		default:
2997 			err = nfs4_handle_exception(server, err, &exception);
2998 		}
2999 	} while (exception.retry);
3000 out:
3001 	return err;
3002 }
3003 
3004 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3005 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3006 {
3007 	struct rpc_auth_create_args auth_args = {
3008 		.pseudoflavor = flavor,
3009 	};
3010 	struct rpc_auth *auth;
3011 	int ret;
3012 
3013 	auth = rpcauth_create(&auth_args, server->client);
3014 	if (IS_ERR(auth)) {
3015 		ret = -EACCES;
3016 		goto out;
3017 	}
3018 	ret = nfs4_lookup_root(server, fhandle, info);
3019 out:
3020 	return ret;
3021 }
3022 
3023 /*
3024  * Retry pseudoroot lookup with various security flavors.  We do this when:
3025  *
3026  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3027  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3028  *
3029  * Returns zero on success, or a negative NFS4ERR value, or a
3030  * negative errno value.
3031  */
3032 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3033 			      struct nfs_fsinfo *info)
3034 {
3035 	/* Per 3530bis 15.33.5 */
3036 	static const rpc_authflavor_t flav_array[] = {
3037 		RPC_AUTH_GSS_KRB5P,
3038 		RPC_AUTH_GSS_KRB5I,
3039 		RPC_AUTH_GSS_KRB5,
3040 		RPC_AUTH_UNIX,			/* courtesy */
3041 		RPC_AUTH_NULL,
3042 	};
3043 	int status = -EPERM;
3044 	size_t i;
3045 
3046 	if (server->auth_info.flavor_len > 0) {
3047 		/* try each flavor specified by user */
3048 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3049 			status = nfs4_lookup_root_sec(server, fhandle, info,
3050 						server->auth_info.flavors[i]);
3051 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3052 				continue;
3053 			break;
3054 		}
3055 	} else {
3056 		/* no flavors specified by user, try default list */
3057 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3058 			status = nfs4_lookup_root_sec(server, fhandle, info,
3059 						      flav_array[i]);
3060 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3061 				continue;
3062 			break;
3063 		}
3064 	}
3065 
3066 	/*
3067 	 * -EACCESS could mean that the user doesn't have correct permissions
3068 	 * to access the mount.  It could also mean that we tried to mount
3069 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3070 	 * existing mount programs don't handle -EACCES very well so it should
3071 	 * be mapped to -EPERM instead.
3072 	 */
3073 	if (status == -EACCES)
3074 		status = -EPERM;
3075 	return status;
3076 }
3077 
3078 static int nfs4_do_find_root_sec(struct nfs_server *server,
3079 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3080 {
3081 	int mv = server->nfs_client->cl_minorversion;
3082 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3083 }
3084 
3085 /**
3086  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3087  * @server: initialized nfs_server handle
3088  * @fhandle: we fill in the pseudo-fs root file handle
3089  * @info: we fill in an FSINFO struct
3090  * @auth_probe: probe the auth flavours
3091  *
3092  * Returns zero on success, or a negative errno.
3093  */
3094 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3095 			 struct nfs_fsinfo *info,
3096 			 bool auth_probe)
3097 {
3098 	int status;
3099 
3100 	switch (auth_probe) {
3101 	case false:
3102 		status = nfs4_lookup_root(server, fhandle, info);
3103 		if (status != -NFS4ERR_WRONGSEC)
3104 			break;
3105 	default:
3106 		status = nfs4_do_find_root_sec(server, fhandle, info);
3107 	}
3108 
3109 	if (status == 0)
3110 		status = nfs4_server_capabilities(server, fhandle);
3111 	if (status == 0)
3112 		status = nfs4_do_fsinfo(server, fhandle, info);
3113 
3114 	return nfs4_map_errors(status);
3115 }
3116 
3117 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3118 			      struct nfs_fsinfo *info)
3119 {
3120 	int error;
3121 	struct nfs_fattr *fattr = info->fattr;
3122 	struct nfs4_label *label = NULL;
3123 
3124 	error = nfs4_server_capabilities(server, mntfh);
3125 	if (error < 0) {
3126 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3127 		return error;
3128 	}
3129 
3130 	label = nfs4_label_alloc(server, GFP_KERNEL);
3131 	if (IS_ERR(label))
3132 		return PTR_ERR(label);
3133 
3134 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3135 	if (error < 0) {
3136 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3137 		goto err_free_label;
3138 	}
3139 
3140 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3141 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3142 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3143 
3144 err_free_label:
3145 	nfs4_label_free(label);
3146 
3147 	return error;
3148 }
3149 
3150 /*
3151  * Get locations and (maybe) other attributes of a referral.
3152  * Note that we'll actually follow the referral later when
3153  * we detect fsid mismatch in inode revalidation
3154  */
3155 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3156 			     const struct qstr *name, struct nfs_fattr *fattr,
3157 			     struct nfs_fh *fhandle)
3158 {
3159 	int status = -ENOMEM;
3160 	struct page *page = NULL;
3161 	struct nfs4_fs_locations *locations = NULL;
3162 
3163 	page = alloc_page(GFP_KERNEL);
3164 	if (page == NULL)
3165 		goto out;
3166 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3167 	if (locations == NULL)
3168 		goto out;
3169 
3170 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3171 	if (status != 0)
3172 		goto out;
3173 
3174 	/*
3175 	 * If the fsid didn't change, this is a migration event, not a
3176 	 * referral.  Cause us to drop into the exception handler, which
3177 	 * will kick off migration recovery.
3178 	 */
3179 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3180 		dprintk("%s: server did not return a different fsid for"
3181 			" a referral at %s\n", __func__, name->name);
3182 		status = -NFS4ERR_MOVED;
3183 		goto out;
3184 	}
3185 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3186 	nfs_fixup_referral_attributes(&locations->fattr);
3187 
3188 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3189 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3190 	memset(fhandle, 0, sizeof(struct nfs_fh));
3191 out:
3192 	if (page)
3193 		__free_page(page);
3194 	kfree(locations);
3195 	return status;
3196 }
3197 
3198 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3199 				struct nfs_fattr *fattr, struct nfs4_label *label)
3200 {
3201 	struct nfs4_getattr_arg args = {
3202 		.fh = fhandle,
3203 		.bitmask = server->attr_bitmask,
3204 	};
3205 	struct nfs4_getattr_res res = {
3206 		.fattr = fattr,
3207 		.label = label,
3208 		.server = server,
3209 	};
3210 	struct rpc_message msg = {
3211 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3212 		.rpc_argp = &args,
3213 		.rpc_resp = &res,
3214 	};
3215 
3216 	args.bitmask = nfs4_bitmask(server, label);
3217 
3218 	nfs_fattr_init(fattr);
3219 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3220 }
3221 
3222 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3223 				struct nfs_fattr *fattr, struct nfs4_label *label)
3224 {
3225 	struct nfs4_exception exception = { };
3226 	int err;
3227 	do {
3228 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3229 		trace_nfs4_getattr(server, fhandle, fattr, err);
3230 		err = nfs4_handle_exception(server, err,
3231 				&exception);
3232 	} while (exception.retry);
3233 	return err;
3234 }
3235 
3236 /*
3237  * The file is not closed if it is opened due to the a request to change
3238  * the size of the file. The open call will not be needed once the
3239  * VFS layer lookup-intents are implemented.
3240  *
3241  * Close is called when the inode is destroyed.
3242  * If we haven't opened the file for O_WRONLY, we
3243  * need to in the size_change case to obtain a stateid.
3244  *
3245  * Got race?
3246  * Because OPEN is always done by name in nfsv4, it is
3247  * possible that we opened a different file by the same
3248  * name.  We can recognize this race condition, but we
3249  * can't do anything about it besides returning an error.
3250  *
3251  * This will be fixed with VFS changes (lookup-intent).
3252  */
3253 static int
3254 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3255 		  struct iattr *sattr)
3256 {
3257 	struct inode *inode = dentry->d_inode;
3258 	struct rpc_cred *cred = NULL;
3259 	struct nfs4_state *state = NULL;
3260 	struct nfs4_label *label = NULL;
3261 	int status;
3262 
3263 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3264 	    sattr->ia_valid & ATTR_SIZE &&
3265 	    sattr->ia_size < i_size_read(inode))
3266 		pnfs_commit_and_return_layout(inode);
3267 
3268 	nfs_fattr_init(fattr);
3269 
3270 	/* Deal with open(O_TRUNC) */
3271 	if (sattr->ia_valid & ATTR_OPEN)
3272 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3273 
3274 	/* Optimization: if the end result is no change, don't RPC */
3275 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3276 		return 0;
3277 
3278 	/* Search for an existing open(O_WRITE) file */
3279 	if (sattr->ia_valid & ATTR_FILE) {
3280 		struct nfs_open_context *ctx;
3281 
3282 		ctx = nfs_file_open_context(sattr->ia_file);
3283 		if (ctx) {
3284 			cred = ctx->cred;
3285 			state = ctx->state;
3286 		}
3287 	}
3288 
3289 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3290 	if (IS_ERR(label))
3291 		return PTR_ERR(label);
3292 
3293 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3294 	if (status == 0) {
3295 		nfs_setattr_update_inode(inode, sattr, fattr);
3296 		nfs_setsecurity(inode, fattr, label);
3297 	}
3298 	nfs4_label_free(label);
3299 	return status;
3300 }
3301 
3302 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3303 		const struct qstr *name, struct nfs_fh *fhandle,
3304 		struct nfs_fattr *fattr, struct nfs4_label *label)
3305 {
3306 	struct nfs_server *server = NFS_SERVER(dir);
3307 	int		       status;
3308 	struct nfs4_lookup_arg args = {
3309 		.bitmask = server->attr_bitmask,
3310 		.dir_fh = NFS_FH(dir),
3311 		.name = name,
3312 	};
3313 	struct nfs4_lookup_res res = {
3314 		.server = server,
3315 		.fattr = fattr,
3316 		.label = label,
3317 		.fh = fhandle,
3318 	};
3319 	struct rpc_message msg = {
3320 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3321 		.rpc_argp = &args,
3322 		.rpc_resp = &res,
3323 	};
3324 
3325 	args.bitmask = nfs4_bitmask(server, label);
3326 
3327 	nfs_fattr_init(fattr);
3328 
3329 	dprintk("NFS call  lookup %s\n", name->name);
3330 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3331 	dprintk("NFS reply lookup: %d\n", status);
3332 	return status;
3333 }
3334 
3335 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3336 {
3337 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3338 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3339 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3340 	fattr->nlink = 2;
3341 }
3342 
3343 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3344 				   struct qstr *name, struct nfs_fh *fhandle,
3345 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3346 {
3347 	struct nfs4_exception exception = { };
3348 	struct rpc_clnt *client = *clnt;
3349 	int err;
3350 	do {
3351 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3352 		trace_nfs4_lookup(dir, name, err);
3353 		switch (err) {
3354 		case -NFS4ERR_BADNAME:
3355 			err = -ENOENT;
3356 			goto out;
3357 		case -NFS4ERR_MOVED:
3358 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3359 			goto out;
3360 		case -NFS4ERR_WRONGSEC:
3361 			err = -EPERM;
3362 			if (client != *clnt)
3363 				goto out;
3364 			client = nfs4_negotiate_security(client, dir, name);
3365 			if (IS_ERR(client))
3366 				return PTR_ERR(client);
3367 
3368 			exception.retry = 1;
3369 			break;
3370 		default:
3371 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3372 		}
3373 	} while (exception.retry);
3374 
3375 out:
3376 	if (err == 0)
3377 		*clnt = client;
3378 	else if (client != *clnt)
3379 		rpc_shutdown_client(client);
3380 
3381 	return err;
3382 }
3383 
3384 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3385 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3386 			    struct nfs4_label *label)
3387 {
3388 	int status;
3389 	struct rpc_clnt *client = NFS_CLIENT(dir);
3390 
3391 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3392 	if (client != NFS_CLIENT(dir)) {
3393 		rpc_shutdown_client(client);
3394 		nfs_fixup_secinfo_attributes(fattr);
3395 	}
3396 	return status;
3397 }
3398 
3399 struct rpc_clnt *
3400 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3401 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3402 {
3403 	struct rpc_clnt *client = NFS_CLIENT(dir);
3404 	int status;
3405 
3406 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3407 	if (status < 0)
3408 		return ERR_PTR(status);
3409 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3410 }
3411 
3412 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3413 {
3414 	struct nfs_server *server = NFS_SERVER(inode);
3415 	struct nfs4_accessargs args = {
3416 		.fh = NFS_FH(inode),
3417 		.bitmask = server->cache_consistency_bitmask,
3418 	};
3419 	struct nfs4_accessres res = {
3420 		.server = server,
3421 	};
3422 	struct rpc_message msg = {
3423 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3424 		.rpc_argp = &args,
3425 		.rpc_resp = &res,
3426 		.rpc_cred = entry->cred,
3427 	};
3428 	int mode = entry->mask;
3429 	int status = 0;
3430 
3431 	/*
3432 	 * Determine which access bits we want to ask for...
3433 	 */
3434 	if (mode & MAY_READ)
3435 		args.access |= NFS4_ACCESS_READ;
3436 	if (S_ISDIR(inode->i_mode)) {
3437 		if (mode & MAY_WRITE)
3438 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3439 		if (mode & MAY_EXEC)
3440 			args.access |= NFS4_ACCESS_LOOKUP;
3441 	} else {
3442 		if (mode & MAY_WRITE)
3443 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3444 		if (mode & MAY_EXEC)
3445 			args.access |= NFS4_ACCESS_EXECUTE;
3446 	}
3447 
3448 	res.fattr = nfs_alloc_fattr();
3449 	if (res.fattr == NULL)
3450 		return -ENOMEM;
3451 
3452 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3453 	if (!status) {
3454 		nfs_access_set_mask(entry, res.access);
3455 		nfs_refresh_inode(inode, res.fattr);
3456 	}
3457 	nfs_free_fattr(res.fattr);
3458 	return status;
3459 }
3460 
3461 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3462 {
3463 	struct nfs4_exception exception = { };
3464 	int err;
3465 	do {
3466 		err = _nfs4_proc_access(inode, entry);
3467 		trace_nfs4_access(inode, err);
3468 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3469 				&exception);
3470 	} while (exception.retry);
3471 	return err;
3472 }
3473 
3474 /*
3475  * TODO: For the time being, we don't try to get any attributes
3476  * along with any of the zero-copy operations READ, READDIR,
3477  * READLINK, WRITE.
3478  *
3479  * In the case of the first three, we want to put the GETATTR
3480  * after the read-type operation -- this is because it is hard
3481  * to predict the length of a GETATTR response in v4, and thus
3482  * align the READ data correctly.  This means that the GETATTR
3483  * may end up partially falling into the page cache, and we should
3484  * shift it into the 'tail' of the xdr_buf before processing.
3485  * To do this efficiently, we need to know the total length
3486  * of data received, which doesn't seem to be available outside
3487  * of the RPC layer.
3488  *
3489  * In the case of WRITE, we also want to put the GETATTR after
3490  * the operation -- in this case because we want to make sure
3491  * we get the post-operation mtime and size.
3492  *
3493  * Both of these changes to the XDR layer would in fact be quite
3494  * minor, but I decided to leave them for a subsequent patch.
3495  */
3496 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3497 		unsigned int pgbase, unsigned int pglen)
3498 {
3499 	struct nfs4_readlink args = {
3500 		.fh       = NFS_FH(inode),
3501 		.pgbase	  = pgbase,
3502 		.pglen    = pglen,
3503 		.pages    = &page,
3504 	};
3505 	struct nfs4_readlink_res res;
3506 	struct rpc_message msg = {
3507 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3508 		.rpc_argp = &args,
3509 		.rpc_resp = &res,
3510 	};
3511 
3512 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3513 }
3514 
3515 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3516 		unsigned int pgbase, unsigned int pglen)
3517 {
3518 	struct nfs4_exception exception = { };
3519 	int err;
3520 	do {
3521 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3522 		trace_nfs4_readlink(inode, err);
3523 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3524 				&exception);
3525 	} while (exception.retry);
3526 	return err;
3527 }
3528 
3529 /*
3530  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3531  */
3532 static int
3533 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3534 		 int flags)
3535 {
3536 	struct nfs4_label l, *ilabel = NULL;
3537 	struct nfs_open_context *ctx;
3538 	struct nfs4_state *state;
3539 	int opened = 0;
3540 	int status = 0;
3541 
3542 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3543 	if (IS_ERR(ctx))
3544 		return PTR_ERR(ctx);
3545 
3546 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3547 
3548 	sattr->ia_mode &= ~current_umask();
3549 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3550 	if (IS_ERR(state)) {
3551 		status = PTR_ERR(state);
3552 		goto out;
3553 	}
3554 out:
3555 	nfs4_label_release_security(ilabel);
3556 	put_nfs_open_context(ctx);
3557 	return status;
3558 }
3559 
3560 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3561 {
3562 	struct nfs_server *server = NFS_SERVER(dir);
3563 	struct nfs_removeargs args = {
3564 		.fh = NFS_FH(dir),
3565 		.name = *name,
3566 	};
3567 	struct nfs_removeres res = {
3568 		.server = server,
3569 	};
3570 	struct rpc_message msg = {
3571 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3572 		.rpc_argp = &args,
3573 		.rpc_resp = &res,
3574 	};
3575 	int status;
3576 
3577 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3578 	if (status == 0)
3579 		update_changeattr(dir, &res.cinfo);
3580 	return status;
3581 }
3582 
3583 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3584 {
3585 	struct nfs4_exception exception = { };
3586 	int err;
3587 	do {
3588 		err = _nfs4_proc_remove(dir, name);
3589 		trace_nfs4_remove(dir, name, err);
3590 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3591 				&exception);
3592 	} while (exception.retry);
3593 	return err;
3594 }
3595 
3596 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3597 {
3598 	struct nfs_server *server = NFS_SERVER(dir);
3599 	struct nfs_removeargs *args = msg->rpc_argp;
3600 	struct nfs_removeres *res = msg->rpc_resp;
3601 
3602 	res->server = server;
3603 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3604 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3605 
3606 	nfs_fattr_init(res->dir_attr);
3607 }
3608 
3609 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3610 {
3611 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3612 			&data->args.seq_args,
3613 			&data->res.seq_res,
3614 			task);
3615 }
3616 
3617 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3618 {
3619 	struct nfs_unlinkdata *data = task->tk_calldata;
3620 	struct nfs_removeres *res = &data->res;
3621 
3622 	if (!nfs4_sequence_done(task, &res->seq_res))
3623 		return 0;
3624 	if (nfs4_async_handle_error(task, res->server, NULL,
3625 				    &data->timeout) == -EAGAIN)
3626 		return 0;
3627 	update_changeattr(dir, &res->cinfo);
3628 	return 1;
3629 }
3630 
3631 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3632 {
3633 	struct nfs_server *server = NFS_SERVER(dir);
3634 	struct nfs_renameargs *arg = msg->rpc_argp;
3635 	struct nfs_renameres *res = msg->rpc_resp;
3636 
3637 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3638 	res->server = server;
3639 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3640 }
3641 
3642 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3643 {
3644 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3645 			&data->args.seq_args,
3646 			&data->res.seq_res,
3647 			task);
3648 }
3649 
3650 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3651 				 struct inode *new_dir)
3652 {
3653 	struct nfs_renamedata *data = task->tk_calldata;
3654 	struct nfs_renameres *res = &data->res;
3655 
3656 	if (!nfs4_sequence_done(task, &res->seq_res))
3657 		return 0;
3658 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3659 		return 0;
3660 
3661 	update_changeattr(old_dir, &res->old_cinfo);
3662 	update_changeattr(new_dir, &res->new_cinfo);
3663 	return 1;
3664 }
3665 
3666 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3667 {
3668 	struct nfs_server *server = NFS_SERVER(inode);
3669 	struct nfs4_link_arg arg = {
3670 		.fh     = NFS_FH(inode),
3671 		.dir_fh = NFS_FH(dir),
3672 		.name   = name,
3673 		.bitmask = server->attr_bitmask,
3674 	};
3675 	struct nfs4_link_res res = {
3676 		.server = server,
3677 		.label = NULL,
3678 	};
3679 	struct rpc_message msg = {
3680 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3681 		.rpc_argp = &arg,
3682 		.rpc_resp = &res,
3683 	};
3684 	int status = -ENOMEM;
3685 
3686 	res.fattr = nfs_alloc_fattr();
3687 	if (res.fattr == NULL)
3688 		goto out;
3689 
3690 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3691 	if (IS_ERR(res.label)) {
3692 		status = PTR_ERR(res.label);
3693 		goto out;
3694 	}
3695 	arg.bitmask = nfs4_bitmask(server, res.label);
3696 
3697 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3698 	if (!status) {
3699 		update_changeattr(dir, &res.cinfo);
3700 		status = nfs_post_op_update_inode(inode, res.fattr);
3701 		if (!status)
3702 			nfs_setsecurity(inode, res.fattr, res.label);
3703 	}
3704 
3705 
3706 	nfs4_label_free(res.label);
3707 
3708 out:
3709 	nfs_free_fattr(res.fattr);
3710 	return status;
3711 }
3712 
3713 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3714 {
3715 	struct nfs4_exception exception = { };
3716 	int err;
3717 	do {
3718 		err = nfs4_handle_exception(NFS_SERVER(inode),
3719 				_nfs4_proc_link(inode, dir, name),
3720 				&exception);
3721 	} while (exception.retry);
3722 	return err;
3723 }
3724 
3725 struct nfs4_createdata {
3726 	struct rpc_message msg;
3727 	struct nfs4_create_arg arg;
3728 	struct nfs4_create_res res;
3729 	struct nfs_fh fh;
3730 	struct nfs_fattr fattr;
3731 	struct nfs4_label *label;
3732 };
3733 
3734 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3735 		struct qstr *name, struct iattr *sattr, u32 ftype)
3736 {
3737 	struct nfs4_createdata *data;
3738 
3739 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3740 	if (data != NULL) {
3741 		struct nfs_server *server = NFS_SERVER(dir);
3742 
3743 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3744 		if (IS_ERR(data->label))
3745 			goto out_free;
3746 
3747 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3748 		data->msg.rpc_argp = &data->arg;
3749 		data->msg.rpc_resp = &data->res;
3750 		data->arg.dir_fh = NFS_FH(dir);
3751 		data->arg.server = server;
3752 		data->arg.name = name;
3753 		data->arg.attrs = sattr;
3754 		data->arg.ftype = ftype;
3755 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3756 		data->res.server = server;
3757 		data->res.fh = &data->fh;
3758 		data->res.fattr = &data->fattr;
3759 		data->res.label = data->label;
3760 		nfs_fattr_init(data->res.fattr);
3761 	}
3762 	return data;
3763 out_free:
3764 	kfree(data);
3765 	return NULL;
3766 }
3767 
3768 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3769 {
3770 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3771 				    &data->arg.seq_args, &data->res.seq_res, 1);
3772 	if (status == 0) {
3773 		update_changeattr(dir, &data->res.dir_cinfo);
3774 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3775 	}
3776 	return status;
3777 }
3778 
3779 static void nfs4_free_createdata(struct nfs4_createdata *data)
3780 {
3781 	nfs4_label_free(data->label);
3782 	kfree(data);
3783 }
3784 
3785 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3786 		struct page *page, unsigned int len, struct iattr *sattr,
3787 		struct nfs4_label *label)
3788 {
3789 	struct nfs4_createdata *data;
3790 	int status = -ENAMETOOLONG;
3791 
3792 	if (len > NFS4_MAXPATHLEN)
3793 		goto out;
3794 
3795 	status = -ENOMEM;
3796 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3797 	if (data == NULL)
3798 		goto out;
3799 
3800 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3801 	data->arg.u.symlink.pages = &page;
3802 	data->arg.u.symlink.len = len;
3803 	data->arg.label = label;
3804 
3805 	status = nfs4_do_create(dir, dentry, data);
3806 
3807 	nfs4_free_createdata(data);
3808 out:
3809 	return status;
3810 }
3811 
3812 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3813 		struct page *page, unsigned int len, struct iattr *sattr)
3814 {
3815 	struct nfs4_exception exception = { };
3816 	struct nfs4_label l, *label = NULL;
3817 	int err;
3818 
3819 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3820 
3821 	do {
3822 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3823 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3824 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3825 				&exception);
3826 	} while (exception.retry);
3827 
3828 	nfs4_label_release_security(label);
3829 	return err;
3830 }
3831 
3832 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3833 		struct iattr *sattr, struct nfs4_label *label)
3834 {
3835 	struct nfs4_createdata *data;
3836 	int status = -ENOMEM;
3837 
3838 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3839 	if (data == NULL)
3840 		goto out;
3841 
3842 	data->arg.label = label;
3843 	status = nfs4_do_create(dir, dentry, data);
3844 
3845 	nfs4_free_createdata(data);
3846 out:
3847 	return status;
3848 }
3849 
3850 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3851 		struct iattr *sattr)
3852 {
3853 	struct nfs4_exception exception = { };
3854 	struct nfs4_label l, *label = NULL;
3855 	int err;
3856 
3857 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3858 
3859 	sattr->ia_mode &= ~current_umask();
3860 	do {
3861 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3862 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3863 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3864 				&exception);
3865 	} while (exception.retry);
3866 	nfs4_label_release_security(label);
3867 
3868 	return err;
3869 }
3870 
3871 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3872 		u64 cookie, struct page **pages, unsigned int count, int plus)
3873 {
3874 	struct inode		*dir = dentry->d_inode;
3875 	struct nfs4_readdir_arg args = {
3876 		.fh = NFS_FH(dir),
3877 		.pages = pages,
3878 		.pgbase = 0,
3879 		.count = count,
3880 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3881 		.plus = plus,
3882 	};
3883 	struct nfs4_readdir_res res;
3884 	struct rpc_message msg = {
3885 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3886 		.rpc_argp = &args,
3887 		.rpc_resp = &res,
3888 		.rpc_cred = cred,
3889 	};
3890 	int			status;
3891 
3892 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3893 			dentry,
3894 			(unsigned long long)cookie);
3895 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3896 	res.pgbase = args.pgbase;
3897 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3898 	if (status >= 0) {
3899 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3900 		status += args.pgbase;
3901 	}
3902 
3903 	nfs_invalidate_atime(dir);
3904 
3905 	dprintk("%s: returns %d\n", __func__, status);
3906 	return status;
3907 }
3908 
3909 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3910 		u64 cookie, struct page **pages, unsigned int count, int plus)
3911 {
3912 	struct nfs4_exception exception = { };
3913 	int err;
3914 	do {
3915 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3916 				pages, count, plus);
3917 		trace_nfs4_readdir(dentry->d_inode, err);
3918 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3919 				&exception);
3920 	} while (exception.retry);
3921 	return err;
3922 }
3923 
3924 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3925 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3926 {
3927 	struct nfs4_createdata *data;
3928 	int mode = sattr->ia_mode;
3929 	int status = -ENOMEM;
3930 
3931 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3932 	if (data == NULL)
3933 		goto out;
3934 
3935 	if (S_ISFIFO(mode))
3936 		data->arg.ftype = NF4FIFO;
3937 	else if (S_ISBLK(mode)) {
3938 		data->arg.ftype = NF4BLK;
3939 		data->arg.u.device.specdata1 = MAJOR(rdev);
3940 		data->arg.u.device.specdata2 = MINOR(rdev);
3941 	}
3942 	else if (S_ISCHR(mode)) {
3943 		data->arg.ftype = NF4CHR;
3944 		data->arg.u.device.specdata1 = MAJOR(rdev);
3945 		data->arg.u.device.specdata2 = MINOR(rdev);
3946 	} else if (!S_ISSOCK(mode)) {
3947 		status = -EINVAL;
3948 		goto out_free;
3949 	}
3950 
3951 	data->arg.label = label;
3952 	status = nfs4_do_create(dir, dentry, data);
3953 out_free:
3954 	nfs4_free_createdata(data);
3955 out:
3956 	return status;
3957 }
3958 
3959 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3960 		struct iattr *sattr, dev_t rdev)
3961 {
3962 	struct nfs4_exception exception = { };
3963 	struct nfs4_label l, *label = NULL;
3964 	int err;
3965 
3966 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3967 
3968 	sattr->ia_mode &= ~current_umask();
3969 	do {
3970 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3971 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3972 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3973 				&exception);
3974 	} while (exception.retry);
3975 
3976 	nfs4_label_release_security(label);
3977 
3978 	return err;
3979 }
3980 
3981 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3982 		 struct nfs_fsstat *fsstat)
3983 {
3984 	struct nfs4_statfs_arg args = {
3985 		.fh = fhandle,
3986 		.bitmask = server->attr_bitmask,
3987 	};
3988 	struct nfs4_statfs_res res = {
3989 		.fsstat = fsstat,
3990 	};
3991 	struct rpc_message msg = {
3992 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3993 		.rpc_argp = &args,
3994 		.rpc_resp = &res,
3995 	};
3996 
3997 	nfs_fattr_init(fsstat->fattr);
3998 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3999 }
4000 
4001 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4002 {
4003 	struct nfs4_exception exception = { };
4004 	int err;
4005 	do {
4006 		err = nfs4_handle_exception(server,
4007 				_nfs4_proc_statfs(server, fhandle, fsstat),
4008 				&exception);
4009 	} while (exception.retry);
4010 	return err;
4011 }
4012 
4013 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4014 		struct nfs_fsinfo *fsinfo)
4015 {
4016 	struct nfs4_fsinfo_arg args = {
4017 		.fh = fhandle,
4018 		.bitmask = server->attr_bitmask,
4019 	};
4020 	struct nfs4_fsinfo_res res = {
4021 		.fsinfo = fsinfo,
4022 	};
4023 	struct rpc_message msg = {
4024 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4025 		.rpc_argp = &args,
4026 		.rpc_resp = &res,
4027 	};
4028 
4029 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4030 }
4031 
4032 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4033 {
4034 	struct nfs4_exception exception = { };
4035 	unsigned long now = jiffies;
4036 	int err;
4037 
4038 	do {
4039 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4040 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4041 		if (err == 0) {
4042 			struct nfs_client *clp = server->nfs_client;
4043 
4044 			spin_lock(&clp->cl_lock);
4045 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4046 			clp->cl_last_renewal = now;
4047 			spin_unlock(&clp->cl_lock);
4048 			break;
4049 		}
4050 		err = nfs4_handle_exception(server, err, &exception);
4051 	} while (exception.retry);
4052 	return err;
4053 }
4054 
4055 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4056 {
4057 	int error;
4058 
4059 	nfs_fattr_init(fsinfo->fattr);
4060 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4061 	if (error == 0) {
4062 		/* block layout checks this! */
4063 		server->pnfs_blksize = fsinfo->blksize;
4064 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4065 	}
4066 
4067 	return error;
4068 }
4069 
4070 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4071 		struct nfs_pathconf *pathconf)
4072 {
4073 	struct nfs4_pathconf_arg args = {
4074 		.fh = fhandle,
4075 		.bitmask = server->attr_bitmask,
4076 	};
4077 	struct nfs4_pathconf_res res = {
4078 		.pathconf = pathconf,
4079 	};
4080 	struct rpc_message msg = {
4081 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4082 		.rpc_argp = &args,
4083 		.rpc_resp = &res,
4084 	};
4085 
4086 	/* None of the pathconf attributes are mandatory to implement */
4087 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4088 		memset(pathconf, 0, sizeof(*pathconf));
4089 		return 0;
4090 	}
4091 
4092 	nfs_fattr_init(pathconf->fattr);
4093 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4094 }
4095 
4096 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4097 		struct nfs_pathconf *pathconf)
4098 {
4099 	struct nfs4_exception exception = { };
4100 	int err;
4101 
4102 	do {
4103 		err = nfs4_handle_exception(server,
4104 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4105 				&exception);
4106 	} while (exception.retry);
4107 	return err;
4108 }
4109 
4110 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4111 		const struct nfs_open_context *ctx,
4112 		const struct nfs_lock_context *l_ctx,
4113 		fmode_t fmode)
4114 {
4115 	const struct nfs_lockowner *lockowner = NULL;
4116 
4117 	if (l_ctx != NULL)
4118 		lockowner = &l_ctx->lockowner;
4119 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4120 }
4121 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4122 
4123 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4124 		const struct nfs_open_context *ctx,
4125 		const struct nfs_lock_context *l_ctx,
4126 		fmode_t fmode)
4127 {
4128 	nfs4_stateid current_stateid;
4129 
4130 	/* If the current stateid represents a lost lock, then exit */
4131 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4132 		return true;
4133 	return nfs4_stateid_match(stateid, &current_stateid);
4134 }
4135 
4136 static bool nfs4_error_stateid_expired(int err)
4137 {
4138 	switch (err) {
4139 	case -NFS4ERR_DELEG_REVOKED:
4140 	case -NFS4ERR_ADMIN_REVOKED:
4141 	case -NFS4ERR_BAD_STATEID:
4142 	case -NFS4ERR_STALE_STATEID:
4143 	case -NFS4ERR_OLD_STATEID:
4144 	case -NFS4ERR_OPENMODE:
4145 	case -NFS4ERR_EXPIRED:
4146 		return true;
4147 	}
4148 	return false;
4149 }
4150 
4151 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4152 {
4153 	nfs_invalidate_atime(hdr->inode);
4154 }
4155 
4156 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4157 {
4158 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4159 
4160 	trace_nfs4_read(hdr, task->tk_status);
4161 	if (nfs4_async_handle_error(task, server,
4162 				    hdr->args.context->state,
4163 				    NULL) == -EAGAIN) {
4164 		rpc_restart_call_prepare(task);
4165 		return -EAGAIN;
4166 	}
4167 
4168 	__nfs4_read_done_cb(hdr);
4169 	if (task->tk_status > 0)
4170 		renew_lease(server, hdr->timestamp);
4171 	return 0;
4172 }
4173 
4174 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4175 		struct nfs_pgio_args *args)
4176 {
4177 
4178 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4179 		nfs4_stateid_is_current(&args->stateid,
4180 				args->context,
4181 				args->lock_context,
4182 				FMODE_READ))
4183 		return false;
4184 	rpc_restart_call_prepare(task);
4185 	return true;
4186 }
4187 
4188 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4189 {
4190 
4191 	dprintk("--> %s\n", __func__);
4192 
4193 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4194 		return -EAGAIN;
4195 	if (nfs4_read_stateid_changed(task, &hdr->args))
4196 		return -EAGAIN;
4197 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4198 				    nfs4_read_done_cb(task, hdr);
4199 }
4200 
4201 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4202 				 struct rpc_message *msg)
4203 {
4204 	hdr->timestamp   = jiffies;
4205 	hdr->pgio_done_cb = nfs4_read_done_cb;
4206 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4207 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4208 }
4209 
4210 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4211 				      struct nfs_pgio_header *hdr)
4212 {
4213 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4214 			&hdr->args.seq_args,
4215 			&hdr->res.seq_res,
4216 			task))
4217 		return 0;
4218 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4219 				hdr->args.lock_context,
4220 				hdr->rw_ops->rw_mode) == -EIO)
4221 		return -EIO;
4222 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4223 		return -EIO;
4224 	return 0;
4225 }
4226 
4227 static int nfs4_write_done_cb(struct rpc_task *task,
4228 			      struct nfs_pgio_header *hdr)
4229 {
4230 	struct inode *inode = hdr->inode;
4231 
4232 	trace_nfs4_write(hdr, task->tk_status);
4233 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4234 				    hdr->args.context->state,
4235 				    NULL) == -EAGAIN) {
4236 		rpc_restart_call_prepare(task);
4237 		return -EAGAIN;
4238 	}
4239 	if (task->tk_status >= 0) {
4240 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4241 		nfs_writeback_update_inode(hdr);
4242 	}
4243 	return 0;
4244 }
4245 
4246 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4247 		struct nfs_pgio_args *args)
4248 {
4249 
4250 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4251 		nfs4_stateid_is_current(&args->stateid,
4252 				args->context,
4253 				args->lock_context,
4254 				FMODE_WRITE))
4255 		return false;
4256 	rpc_restart_call_prepare(task);
4257 	return true;
4258 }
4259 
4260 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4261 {
4262 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4263 		return -EAGAIN;
4264 	if (nfs4_write_stateid_changed(task, &hdr->args))
4265 		return -EAGAIN;
4266 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4267 		nfs4_write_done_cb(task, hdr);
4268 }
4269 
4270 static
4271 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4272 {
4273 	/* Don't request attributes for pNFS or O_DIRECT writes */
4274 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4275 		return false;
4276 	/* Otherwise, request attributes if and only if we don't hold
4277 	 * a delegation
4278 	 */
4279 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4280 }
4281 
4282 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4283 				  struct rpc_message *msg)
4284 {
4285 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4286 
4287 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4288 		hdr->args.bitmask = NULL;
4289 		hdr->res.fattr = NULL;
4290 	} else
4291 		hdr->args.bitmask = server->cache_consistency_bitmask;
4292 
4293 	if (!hdr->pgio_done_cb)
4294 		hdr->pgio_done_cb = nfs4_write_done_cb;
4295 	hdr->res.server = server;
4296 	hdr->timestamp   = jiffies;
4297 
4298 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4299 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4300 }
4301 
4302 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4303 {
4304 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4305 			&data->args.seq_args,
4306 			&data->res.seq_res,
4307 			task);
4308 }
4309 
4310 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4311 {
4312 	struct inode *inode = data->inode;
4313 
4314 	trace_nfs4_commit(data, task->tk_status);
4315 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4316 				    NULL, NULL) == -EAGAIN) {
4317 		rpc_restart_call_prepare(task);
4318 		return -EAGAIN;
4319 	}
4320 	return 0;
4321 }
4322 
4323 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4324 {
4325 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4326 		return -EAGAIN;
4327 	return data->commit_done_cb(task, data);
4328 }
4329 
4330 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4331 {
4332 	struct nfs_server *server = NFS_SERVER(data->inode);
4333 
4334 	if (data->commit_done_cb == NULL)
4335 		data->commit_done_cb = nfs4_commit_done_cb;
4336 	data->res.server = server;
4337 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4338 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4339 }
4340 
4341 struct nfs4_renewdata {
4342 	struct nfs_client	*client;
4343 	unsigned long		timestamp;
4344 };
4345 
4346 /*
4347  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4348  * standalone procedure for queueing an asynchronous RENEW.
4349  */
4350 static void nfs4_renew_release(void *calldata)
4351 {
4352 	struct nfs4_renewdata *data = calldata;
4353 	struct nfs_client *clp = data->client;
4354 
4355 	if (atomic_read(&clp->cl_count) > 1)
4356 		nfs4_schedule_state_renewal(clp);
4357 	nfs_put_client(clp);
4358 	kfree(data);
4359 }
4360 
4361 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4362 {
4363 	struct nfs4_renewdata *data = calldata;
4364 	struct nfs_client *clp = data->client;
4365 	unsigned long timestamp = data->timestamp;
4366 
4367 	trace_nfs4_renew_async(clp, task->tk_status);
4368 	switch (task->tk_status) {
4369 	case 0:
4370 		break;
4371 	case -NFS4ERR_LEASE_MOVED:
4372 		nfs4_schedule_lease_moved_recovery(clp);
4373 		break;
4374 	default:
4375 		/* Unless we're shutting down, schedule state recovery! */
4376 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4377 			return;
4378 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4379 			nfs4_schedule_lease_recovery(clp);
4380 			return;
4381 		}
4382 		nfs4_schedule_path_down_recovery(clp);
4383 	}
4384 	do_renew_lease(clp, timestamp);
4385 }
4386 
4387 static const struct rpc_call_ops nfs4_renew_ops = {
4388 	.rpc_call_done = nfs4_renew_done,
4389 	.rpc_release = nfs4_renew_release,
4390 };
4391 
4392 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4393 {
4394 	struct rpc_message msg = {
4395 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4396 		.rpc_argp	= clp,
4397 		.rpc_cred	= cred,
4398 	};
4399 	struct nfs4_renewdata *data;
4400 
4401 	if (renew_flags == 0)
4402 		return 0;
4403 	if (!atomic_inc_not_zero(&clp->cl_count))
4404 		return -EIO;
4405 	data = kmalloc(sizeof(*data), GFP_NOFS);
4406 	if (data == NULL)
4407 		return -ENOMEM;
4408 	data->client = clp;
4409 	data->timestamp = jiffies;
4410 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4411 			&nfs4_renew_ops, data);
4412 }
4413 
4414 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4415 {
4416 	struct rpc_message msg = {
4417 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4418 		.rpc_argp	= clp,
4419 		.rpc_cred	= cred,
4420 	};
4421 	unsigned long now = jiffies;
4422 	int status;
4423 
4424 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4425 	if (status < 0)
4426 		return status;
4427 	do_renew_lease(clp, now);
4428 	return 0;
4429 }
4430 
4431 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4432 {
4433 	return server->caps & NFS_CAP_ACLS;
4434 }
4435 
4436 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4437  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4438  * the stack.
4439  */
4440 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4441 
4442 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4443 		struct page **pages, unsigned int *pgbase)
4444 {
4445 	struct page *newpage, **spages;
4446 	int rc = 0;
4447 	size_t len;
4448 	spages = pages;
4449 
4450 	do {
4451 		len = min_t(size_t, PAGE_SIZE, buflen);
4452 		newpage = alloc_page(GFP_KERNEL);
4453 
4454 		if (newpage == NULL)
4455 			goto unwind;
4456 		memcpy(page_address(newpage), buf, len);
4457                 buf += len;
4458                 buflen -= len;
4459 		*pages++ = newpage;
4460 		rc++;
4461 	} while (buflen != 0);
4462 
4463 	return rc;
4464 
4465 unwind:
4466 	for(; rc > 0; rc--)
4467 		__free_page(spages[rc-1]);
4468 	return -ENOMEM;
4469 }
4470 
4471 struct nfs4_cached_acl {
4472 	int cached;
4473 	size_t len;
4474 	char data[0];
4475 };
4476 
4477 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4478 {
4479 	struct nfs_inode *nfsi = NFS_I(inode);
4480 
4481 	spin_lock(&inode->i_lock);
4482 	kfree(nfsi->nfs4_acl);
4483 	nfsi->nfs4_acl = acl;
4484 	spin_unlock(&inode->i_lock);
4485 }
4486 
4487 static void nfs4_zap_acl_attr(struct inode *inode)
4488 {
4489 	nfs4_set_cached_acl(inode, NULL);
4490 }
4491 
4492 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4493 {
4494 	struct nfs_inode *nfsi = NFS_I(inode);
4495 	struct nfs4_cached_acl *acl;
4496 	int ret = -ENOENT;
4497 
4498 	spin_lock(&inode->i_lock);
4499 	acl = nfsi->nfs4_acl;
4500 	if (acl == NULL)
4501 		goto out;
4502 	if (buf == NULL) /* user is just asking for length */
4503 		goto out_len;
4504 	if (acl->cached == 0)
4505 		goto out;
4506 	ret = -ERANGE; /* see getxattr(2) man page */
4507 	if (acl->len > buflen)
4508 		goto out;
4509 	memcpy(buf, acl->data, acl->len);
4510 out_len:
4511 	ret = acl->len;
4512 out:
4513 	spin_unlock(&inode->i_lock);
4514 	return ret;
4515 }
4516 
4517 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4518 {
4519 	struct nfs4_cached_acl *acl;
4520 	size_t buflen = sizeof(*acl) + acl_len;
4521 
4522 	if (buflen <= PAGE_SIZE) {
4523 		acl = kmalloc(buflen, GFP_KERNEL);
4524 		if (acl == NULL)
4525 			goto out;
4526 		acl->cached = 1;
4527 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4528 	} else {
4529 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4530 		if (acl == NULL)
4531 			goto out;
4532 		acl->cached = 0;
4533 	}
4534 	acl->len = acl_len;
4535 out:
4536 	nfs4_set_cached_acl(inode, acl);
4537 }
4538 
4539 /*
4540  * The getxattr API returns the required buffer length when called with a
4541  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4542  * the required buf.  On a NULL buf, we send a page of data to the server
4543  * guessing that the ACL request can be serviced by a page. If so, we cache
4544  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4545  * the cache. If not so, we throw away the page, and cache the required
4546  * length. The next getxattr call will then produce another round trip to
4547  * the server, this time with the input buf of the required size.
4548  */
4549 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4550 {
4551 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4552 	struct nfs_getaclargs args = {
4553 		.fh = NFS_FH(inode),
4554 		.acl_pages = pages,
4555 		.acl_len = buflen,
4556 	};
4557 	struct nfs_getaclres res = {
4558 		.acl_len = buflen,
4559 	};
4560 	struct rpc_message msg = {
4561 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4562 		.rpc_argp = &args,
4563 		.rpc_resp = &res,
4564 	};
4565 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4566 	int ret = -ENOMEM, i;
4567 
4568 	/* As long as we're doing a round trip to the server anyway,
4569 	 * let's be prepared for a page of acl data. */
4570 	if (npages == 0)
4571 		npages = 1;
4572 	if (npages > ARRAY_SIZE(pages))
4573 		return -ERANGE;
4574 
4575 	for (i = 0; i < npages; i++) {
4576 		pages[i] = alloc_page(GFP_KERNEL);
4577 		if (!pages[i])
4578 			goto out_free;
4579 	}
4580 
4581 	/* for decoding across pages */
4582 	res.acl_scratch = alloc_page(GFP_KERNEL);
4583 	if (!res.acl_scratch)
4584 		goto out_free;
4585 
4586 	args.acl_len = npages * PAGE_SIZE;
4587 	args.acl_pgbase = 0;
4588 
4589 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4590 		__func__, buf, buflen, npages, args.acl_len);
4591 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4592 			     &msg, &args.seq_args, &res.seq_res, 0);
4593 	if (ret)
4594 		goto out_free;
4595 
4596 	/* Handle the case where the passed-in buffer is too short */
4597 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4598 		/* Did the user only issue a request for the acl length? */
4599 		if (buf == NULL)
4600 			goto out_ok;
4601 		ret = -ERANGE;
4602 		goto out_free;
4603 	}
4604 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4605 	if (buf) {
4606 		if (res.acl_len > buflen) {
4607 			ret = -ERANGE;
4608 			goto out_free;
4609 		}
4610 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4611 	}
4612 out_ok:
4613 	ret = res.acl_len;
4614 out_free:
4615 	for (i = 0; i < npages; i++)
4616 		if (pages[i])
4617 			__free_page(pages[i]);
4618 	if (res.acl_scratch)
4619 		__free_page(res.acl_scratch);
4620 	return ret;
4621 }
4622 
4623 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4624 {
4625 	struct nfs4_exception exception = { };
4626 	ssize_t ret;
4627 	do {
4628 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4629 		trace_nfs4_get_acl(inode, ret);
4630 		if (ret >= 0)
4631 			break;
4632 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4633 	} while (exception.retry);
4634 	return ret;
4635 }
4636 
4637 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4638 {
4639 	struct nfs_server *server = NFS_SERVER(inode);
4640 	int ret;
4641 
4642 	if (!nfs4_server_supports_acls(server))
4643 		return -EOPNOTSUPP;
4644 	ret = nfs_revalidate_inode(server, inode);
4645 	if (ret < 0)
4646 		return ret;
4647 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4648 		nfs_zap_acl_cache(inode);
4649 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4650 	if (ret != -ENOENT)
4651 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4652 		 * but no cached acl data, just the acl length */
4653 		return ret;
4654 	return nfs4_get_acl_uncached(inode, buf, buflen);
4655 }
4656 
4657 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4658 {
4659 	struct nfs_server *server = NFS_SERVER(inode);
4660 	struct page *pages[NFS4ACL_MAXPAGES];
4661 	struct nfs_setaclargs arg = {
4662 		.fh		= NFS_FH(inode),
4663 		.acl_pages	= pages,
4664 		.acl_len	= buflen,
4665 	};
4666 	struct nfs_setaclres res;
4667 	struct rpc_message msg = {
4668 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4669 		.rpc_argp	= &arg,
4670 		.rpc_resp	= &res,
4671 	};
4672 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4673 	int ret, i;
4674 
4675 	if (!nfs4_server_supports_acls(server))
4676 		return -EOPNOTSUPP;
4677 	if (npages > ARRAY_SIZE(pages))
4678 		return -ERANGE;
4679 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4680 	if (i < 0)
4681 		return i;
4682 	nfs4_inode_return_delegation(inode);
4683 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4684 
4685 	/*
4686 	 * Free each page after tx, so the only ref left is
4687 	 * held by the network stack
4688 	 */
4689 	for (; i > 0; i--)
4690 		put_page(pages[i-1]);
4691 
4692 	/*
4693 	 * Acl update can result in inode attribute update.
4694 	 * so mark the attribute cache invalid.
4695 	 */
4696 	spin_lock(&inode->i_lock);
4697 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4698 	spin_unlock(&inode->i_lock);
4699 	nfs_access_zap_cache(inode);
4700 	nfs_zap_acl_cache(inode);
4701 	return ret;
4702 }
4703 
4704 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4705 {
4706 	struct nfs4_exception exception = { };
4707 	int err;
4708 	do {
4709 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4710 		trace_nfs4_set_acl(inode, err);
4711 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4712 				&exception);
4713 	} while (exception.retry);
4714 	return err;
4715 }
4716 
4717 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4718 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4719 					size_t buflen)
4720 {
4721 	struct nfs_server *server = NFS_SERVER(inode);
4722 	struct nfs_fattr fattr;
4723 	struct nfs4_label label = {0, 0, buflen, buf};
4724 
4725 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4726 	struct nfs4_getattr_arg arg = {
4727 		.fh		= NFS_FH(inode),
4728 		.bitmask	= bitmask,
4729 	};
4730 	struct nfs4_getattr_res res = {
4731 		.fattr		= &fattr,
4732 		.label		= &label,
4733 		.server		= server,
4734 	};
4735 	struct rpc_message msg = {
4736 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4737 		.rpc_argp	= &arg,
4738 		.rpc_resp	= &res,
4739 	};
4740 	int ret;
4741 
4742 	nfs_fattr_init(&fattr);
4743 
4744 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4745 	if (ret)
4746 		return ret;
4747 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4748 		return -ENOENT;
4749 	if (buflen < label.len)
4750 		return -ERANGE;
4751 	return 0;
4752 }
4753 
4754 static int nfs4_get_security_label(struct inode *inode, void *buf,
4755 					size_t buflen)
4756 {
4757 	struct nfs4_exception exception = { };
4758 	int err;
4759 
4760 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4761 		return -EOPNOTSUPP;
4762 
4763 	do {
4764 		err = _nfs4_get_security_label(inode, buf, buflen);
4765 		trace_nfs4_get_security_label(inode, err);
4766 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4767 				&exception);
4768 	} while (exception.retry);
4769 	return err;
4770 }
4771 
4772 static int _nfs4_do_set_security_label(struct inode *inode,
4773 		struct nfs4_label *ilabel,
4774 		struct nfs_fattr *fattr,
4775 		struct nfs4_label *olabel)
4776 {
4777 
4778 	struct iattr sattr = {0};
4779 	struct nfs_server *server = NFS_SERVER(inode);
4780 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4781 	struct nfs_setattrargs arg = {
4782 		.fh             = NFS_FH(inode),
4783 		.iap            = &sattr,
4784 		.server		= server,
4785 		.bitmask	= bitmask,
4786 		.label		= ilabel,
4787 	};
4788 	struct nfs_setattrres res = {
4789 		.fattr		= fattr,
4790 		.label		= olabel,
4791 		.server		= server,
4792 	};
4793 	struct rpc_message msg = {
4794 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4795 		.rpc_argp       = &arg,
4796 		.rpc_resp       = &res,
4797 	};
4798 	int status;
4799 
4800 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4801 
4802 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4803 	if (status)
4804 		dprintk("%s failed: %d\n", __func__, status);
4805 
4806 	return status;
4807 }
4808 
4809 static int nfs4_do_set_security_label(struct inode *inode,
4810 		struct nfs4_label *ilabel,
4811 		struct nfs_fattr *fattr,
4812 		struct nfs4_label *olabel)
4813 {
4814 	struct nfs4_exception exception = { };
4815 	int err;
4816 
4817 	do {
4818 		err = _nfs4_do_set_security_label(inode, ilabel,
4819 				fattr, olabel);
4820 		trace_nfs4_set_security_label(inode, err);
4821 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4822 				&exception);
4823 	} while (exception.retry);
4824 	return err;
4825 }
4826 
4827 static int
4828 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4829 {
4830 	struct nfs4_label ilabel, *olabel = NULL;
4831 	struct nfs_fattr fattr;
4832 	struct rpc_cred *cred;
4833 	struct inode *inode = dentry->d_inode;
4834 	int status;
4835 
4836 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4837 		return -EOPNOTSUPP;
4838 
4839 	nfs_fattr_init(&fattr);
4840 
4841 	ilabel.pi = 0;
4842 	ilabel.lfs = 0;
4843 	ilabel.label = (char *)buf;
4844 	ilabel.len = buflen;
4845 
4846 	cred = rpc_lookup_cred();
4847 	if (IS_ERR(cred))
4848 		return PTR_ERR(cred);
4849 
4850 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4851 	if (IS_ERR(olabel)) {
4852 		status = -PTR_ERR(olabel);
4853 		goto out;
4854 	}
4855 
4856 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4857 	if (status == 0)
4858 		nfs_setsecurity(inode, &fattr, olabel);
4859 
4860 	nfs4_label_free(olabel);
4861 out:
4862 	put_rpccred(cred);
4863 	return status;
4864 }
4865 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4866 
4867 
4868 static int
4869 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4870 			struct nfs4_state *state, long *timeout)
4871 {
4872 	struct nfs_client *clp = server->nfs_client;
4873 
4874 	if (task->tk_status >= 0)
4875 		return 0;
4876 	switch(task->tk_status) {
4877 		case -NFS4ERR_DELEG_REVOKED:
4878 		case -NFS4ERR_ADMIN_REVOKED:
4879 		case -NFS4ERR_BAD_STATEID:
4880 		case -NFS4ERR_OPENMODE:
4881 			if (state == NULL)
4882 				break;
4883 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4884 				goto recovery_failed;
4885 			goto wait_on_recovery;
4886 		case -NFS4ERR_EXPIRED:
4887 			if (state != NULL) {
4888 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4889 					goto recovery_failed;
4890 			}
4891 		case -NFS4ERR_STALE_STATEID:
4892 		case -NFS4ERR_STALE_CLIENTID:
4893 			nfs4_schedule_lease_recovery(clp);
4894 			goto wait_on_recovery;
4895 		case -NFS4ERR_MOVED:
4896 			if (nfs4_schedule_migration_recovery(server) < 0)
4897 				goto recovery_failed;
4898 			goto wait_on_recovery;
4899 		case -NFS4ERR_LEASE_MOVED:
4900 			nfs4_schedule_lease_moved_recovery(clp);
4901 			goto wait_on_recovery;
4902 #if defined(CONFIG_NFS_V4_1)
4903 		case -NFS4ERR_BADSESSION:
4904 		case -NFS4ERR_BADSLOT:
4905 		case -NFS4ERR_BAD_HIGH_SLOT:
4906 		case -NFS4ERR_DEADSESSION:
4907 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4908 		case -NFS4ERR_SEQ_FALSE_RETRY:
4909 		case -NFS4ERR_SEQ_MISORDERED:
4910 			dprintk("%s ERROR %d, Reset session\n", __func__,
4911 				task->tk_status);
4912 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4913 			goto wait_on_recovery;
4914 #endif /* CONFIG_NFS_V4_1 */
4915 		case -NFS4ERR_DELAY:
4916 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4917 			rpc_delay(task, nfs4_update_delay(timeout));
4918 			goto restart_call;
4919 		case -NFS4ERR_GRACE:
4920 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4921 		case -NFS4ERR_RETRY_UNCACHED_REP:
4922 		case -NFS4ERR_OLD_STATEID:
4923 			goto restart_call;
4924 	}
4925 	task->tk_status = nfs4_map_errors(task->tk_status);
4926 	return 0;
4927 recovery_failed:
4928 	task->tk_status = -EIO;
4929 	return 0;
4930 wait_on_recovery:
4931 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4932 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4933 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4934 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4935 		goto recovery_failed;
4936 restart_call:
4937 	task->tk_status = 0;
4938 	return -EAGAIN;
4939 }
4940 
4941 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4942 				    nfs4_verifier *bootverf)
4943 {
4944 	__be32 verf[2];
4945 
4946 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4947 		/* An impossible timestamp guarantees this value
4948 		 * will never match a generated boot time. */
4949 		verf[0] = 0;
4950 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4951 	} else {
4952 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4953 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4954 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4955 	}
4956 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4957 }
4958 
4959 static unsigned int
4960 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4961 				   char *buf, size_t len)
4962 {
4963 	unsigned int result;
4964 
4965 	if (clp->cl_owner_id != NULL)
4966 		return strlcpy(buf, clp->cl_owner_id, len);
4967 
4968 	rcu_read_lock();
4969 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4970 				clp->cl_ipaddr,
4971 				rpc_peeraddr2str(clp->cl_rpcclient,
4972 							RPC_DISPLAY_ADDR),
4973 				rpc_peeraddr2str(clp->cl_rpcclient,
4974 							RPC_DISPLAY_PROTO));
4975 	rcu_read_unlock();
4976 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4977 	return result;
4978 }
4979 
4980 static unsigned int
4981 nfs4_init_uniform_client_string(struct nfs_client *clp,
4982 				char *buf, size_t len)
4983 {
4984 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4985 	unsigned int result;
4986 
4987 	if (clp->cl_owner_id != NULL)
4988 		return strlcpy(buf, clp->cl_owner_id, len);
4989 
4990 	if (nfs4_client_id_uniquifier[0] != '\0')
4991 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4992 				clp->rpc_ops->version,
4993 				clp->cl_minorversion,
4994 				nfs4_client_id_uniquifier,
4995 				nodename);
4996 	else
4997 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4998 				clp->rpc_ops->version, clp->cl_minorversion,
4999 				nodename);
5000 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
5001 	return result;
5002 }
5003 
5004 /*
5005  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5006  * services.  Advertise one based on the address family of the
5007  * clientaddr.
5008  */
5009 static unsigned int
5010 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5011 {
5012 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5013 		return scnprintf(buf, len, "tcp6");
5014 	else
5015 		return scnprintf(buf, len, "tcp");
5016 }
5017 
5018 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5019 {
5020 	struct nfs4_setclientid *sc = calldata;
5021 
5022 	if (task->tk_status == 0)
5023 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5024 }
5025 
5026 static const struct rpc_call_ops nfs4_setclientid_ops = {
5027 	.rpc_call_done = nfs4_setclientid_done,
5028 };
5029 
5030 /**
5031  * nfs4_proc_setclientid - Negotiate client ID
5032  * @clp: state data structure
5033  * @program: RPC program for NFSv4 callback service
5034  * @port: IP port number for NFS4 callback service
5035  * @cred: RPC credential to use for this call
5036  * @res: where to place the result
5037  *
5038  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5039  */
5040 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5041 		unsigned short port, struct rpc_cred *cred,
5042 		struct nfs4_setclientid_res *res)
5043 {
5044 	nfs4_verifier sc_verifier;
5045 	struct nfs4_setclientid setclientid = {
5046 		.sc_verifier = &sc_verifier,
5047 		.sc_prog = program,
5048 		.sc_cb_ident = clp->cl_cb_ident,
5049 	};
5050 	struct rpc_message msg = {
5051 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5052 		.rpc_argp = &setclientid,
5053 		.rpc_resp = res,
5054 		.rpc_cred = cred,
5055 	};
5056 	struct rpc_task *task;
5057 	struct rpc_task_setup task_setup_data = {
5058 		.rpc_client = clp->cl_rpcclient,
5059 		.rpc_message = &msg,
5060 		.callback_ops = &nfs4_setclientid_ops,
5061 		.callback_data = &setclientid,
5062 		.flags = RPC_TASK_TIMEOUT,
5063 	};
5064 	int status;
5065 
5066 	/* nfs_client_id4 */
5067 	nfs4_init_boot_verifier(clp, &sc_verifier);
5068 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5069 		setclientid.sc_name_len =
5070 				nfs4_init_uniform_client_string(clp,
5071 						setclientid.sc_name,
5072 						sizeof(setclientid.sc_name));
5073 	else
5074 		setclientid.sc_name_len =
5075 				nfs4_init_nonuniform_client_string(clp,
5076 						setclientid.sc_name,
5077 						sizeof(setclientid.sc_name));
5078 	/* cb_client4 */
5079 	setclientid.sc_netid_len =
5080 				nfs4_init_callback_netid(clp,
5081 						setclientid.sc_netid,
5082 						sizeof(setclientid.sc_netid));
5083 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5084 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5085 				clp->cl_ipaddr, port >> 8, port & 255);
5086 
5087 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5088 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5089 		setclientid.sc_name_len, setclientid.sc_name);
5090 	task = rpc_run_task(&task_setup_data);
5091 	if (IS_ERR(task)) {
5092 		status = PTR_ERR(task);
5093 		goto out;
5094 	}
5095 	status = task->tk_status;
5096 	if (setclientid.sc_cred) {
5097 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5098 		put_rpccred(setclientid.sc_cred);
5099 	}
5100 	rpc_put_task(task);
5101 out:
5102 	trace_nfs4_setclientid(clp, status);
5103 	dprintk("NFS reply setclientid: %d\n", status);
5104 	return status;
5105 }
5106 
5107 /**
5108  * nfs4_proc_setclientid_confirm - Confirm client ID
5109  * @clp: state data structure
5110  * @res: result of a previous SETCLIENTID
5111  * @cred: RPC credential to use for this call
5112  *
5113  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5114  */
5115 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5116 		struct nfs4_setclientid_res *arg,
5117 		struct rpc_cred *cred)
5118 {
5119 	struct rpc_message msg = {
5120 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5121 		.rpc_argp = arg,
5122 		.rpc_cred = cred,
5123 	};
5124 	int status;
5125 
5126 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5127 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5128 		clp->cl_clientid);
5129 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5130 	trace_nfs4_setclientid_confirm(clp, status);
5131 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5132 	return status;
5133 }
5134 
5135 struct nfs4_delegreturndata {
5136 	struct nfs4_delegreturnargs args;
5137 	struct nfs4_delegreturnres res;
5138 	struct nfs_fh fh;
5139 	nfs4_stateid stateid;
5140 	unsigned long timestamp;
5141 	struct nfs_fattr fattr;
5142 	int rpc_status;
5143 	struct inode *inode;
5144 	bool roc;
5145 	u32 roc_barrier;
5146 };
5147 
5148 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5149 {
5150 	struct nfs4_delegreturndata *data = calldata;
5151 
5152 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5153 		return;
5154 
5155 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5156 	switch (task->tk_status) {
5157 	case 0:
5158 		renew_lease(data->res.server, data->timestamp);
5159 	case -NFS4ERR_ADMIN_REVOKED:
5160 	case -NFS4ERR_DELEG_REVOKED:
5161 	case -NFS4ERR_BAD_STATEID:
5162 	case -NFS4ERR_OLD_STATEID:
5163 	case -NFS4ERR_STALE_STATEID:
5164 	case -NFS4ERR_EXPIRED:
5165 		task->tk_status = 0;
5166 		if (data->roc)
5167 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5168 		break;
5169 	default:
5170 		if (nfs4_async_handle_error(task, data->res.server,
5171 					    NULL, NULL) == -EAGAIN) {
5172 			rpc_restart_call_prepare(task);
5173 			return;
5174 		}
5175 	}
5176 	data->rpc_status = task->tk_status;
5177 }
5178 
5179 static void nfs4_delegreturn_release(void *calldata)
5180 {
5181 	struct nfs4_delegreturndata *data = calldata;
5182 	struct inode *inode = data->inode;
5183 
5184 	if (inode) {
5185 		if (data->roc)
5186 			pnfs_roc_release(inode);
5187 		nfs_iput_and_deactive(inode);
5188 	}
5189 	kfree(calldata);
5190 }
5191 
5192 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5193 {
5194 	struct nfs4_delegreturndata *d_data;
5195 
5196 	d_data = (struct nfs4_delegreturndata *)data;
5197 
5198 	if (d_data->roc &&
5199 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5200 		return;
5201 
5202 	nfs4_setup_sequence(d_data->res.server,
5203 			&d_data->args.seq_args,
5204 			&d_data->res.seq_res,
5205 			task);
5206 }
5207 
5208 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5209 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5210 	.rpc_call_done = nfs4_delegreturn_done,
5211 	.rpc_release = nfs4_delegreturn_release,
5212 };
5213 
5214 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5215 {
5216 	struct nfs4_delegreturndata *data;
5217 	struct nfs_server *server = NFS_SERVER(inode);
5218 	struct rpc_task *task;
5219 	struct rpc_message msg = {
5220 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5221 		.rpc_cred = cred,
5222 	};
5223 	struct rpc_task_setup task_setup_data = {
5224 		.rpc_client = server->client,
5225 		.rpc_message = &msg,
5226 		.callback_ops = &nfs4_delegreturn_ops,
5227 		.flags = RPC_TASK_ASYNC,
5228 	};
5229 	int status = 0;
5230 
5231 	data = kzalloc(sizeof(*data), GFP_NOFS);
5232 	if (data == NULL)
5233 		return -ENOMEM;
5234 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5235 	data->args.fhandle = &data->fh;
5236 	data->args.stateid = &data->stateid;
5237 	data->args.bitmask = server->cache_consistency_bitmask;
5238 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5239 	nfs4_stateid_copy(&data->stateid, stateid);
5240 	data->res.fattr = &data->fattr;
5241 	data->res.server = server;
5242 	nfs_fattr_init(data->res.fattr);
5243 	data->timestamp = jiffies;
5244 	data->rpc_status = 0;
5245 	data->inode = nfs_igrab_and_active(inode);
5246 	if (data->inode)
5247 		data->roc = nfs4_roc(inode);
5248 
5249 	task_setup_data.callback_data = data;
5250 	msg.rpc_argp = &data->args;
5251 	msg.rpc_resp = &data->res;
5252 	task = rpc_run_task(&task_setup_data);
5253 	if (IS_ERR(task))
5254 		return PTR_ERR(task);
5255 	if (!issync)
5256 		goto out;
5257 	status = nfs4_wait_for_completion_rpc_task(task);
5258 	if (status != 0)
5259 		goto out;
5260 	status = data->rpc_status;
5261 	if (status == 0)
5262 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5263 	else
5264 		nfs_refresh_inode(inode, &data->fattr);
5265 out:
5266 	rpc_put_task(task);
5267 	return status;
5268 }
5269 
5270 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5271 {
5272 	struct nfs_server *server = NFS_SERVER(inode);
5273 	struct nfs4_exception exception = { };
5274 	int err;
5275 	do {
5276 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5277 		trace_nfs4_delegreturn(inode, err);
5278 		switch (err) {
5279 			case -NFS4ERR_STALE_STATEID:
5280 			case -NFS4ERR_EXPIRED:
5281 			case 0:
5282 				return 0;
5283 		}
5284 		err = nfs4_handle_exception(server, err, &exception);
5285 	} while (exception.retry);
5286 	return err;
5287 }
5288 
5289 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5290 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5291 
5292 /*
5293  * sleep, with exponential backoff, and retry the LOCK operation.
5294  */
5295 static unsigned long
5296 nfs4_set_lock_task_retry(unsigned long timeout)
5297 {
5298 	freezable_schedule_timeout_killable_unsafe(timeout);
5299 	timeout <<= 1;
5300 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5301 		return NFS4_LOCK_MAXTIMEOUT;
5302 	return timeout;
5303 }
5304 
5305 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5306 {
5307 	struct inode *inode = state->inode;
5308 	struct nfs_server *server = NFS_SERVER(inode);
5309 	struct nfs_client *clp = server->nfs_client;
5310 	struct nfs_lockt_args arg = {
5311 		.fh = NFS_FH(inode),
5312 		.fl = request,
5313 	};
5314 	struct nfs_lockt_res res = {
5315 		.denied = request,
5316 	};
5317 	struct rpc_message msg = {
5318 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5319 		.rpc_argp       = &arg,
5320 		.rpc_resp       = &res,
5321 		.rpc_cred	= state->owner->so_cred,
5322 	};
5323 	struct nfs4_lock_state *lsp;
5324 	int status;
5325 
5326 	arg.lock_owner.clientid = clp->cl_clientid;
5327 	status = nfs4_set_lock_state(state, request);
5328 	if (status != 0)
5329 		goto out;
5330 	lsp = request->fl_u.nfs4_fl.owner;
5331 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5332 	arg.lock_owner.s_dev = server->s_dev;
5333 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5334 	switch (status) {
5335 		case 0:
5336 			request->fl_type = F_UNLCK;
5337 			break;
5338 		case -NFS4ERR_DENIED:
5339 			status = 0;
5340 	}
5341 	request->fl_ops->fl_release_private(request);
5342 	request->fl_ops = NULL;
5343 out:
5344 	return status;
5345 }
5346 
5347 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5348 {
5349 	struct nfs4_exception exception = { };
5350 	int err;
5351 
5352 	do {
5353 		err = _nfs4_proc_getlk(state, cmd, request);
5354 		trace_nfs4_get_lock(request, state, cmd, err);
5355 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5356 				&exception);
5357 	} while (exception.retry);
5358 	return err;
5359 }
5360 
5361 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5362 {
5363 	int res = 0;
5364 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5365 		case FL_POSIX:
5366 			res = posix_lock_file_wait(file, fl);
5367 			break;
5368 		case FL_FLOCK:
5369 			res = flock_lock_file_wait(file, fl);
5370 			break;
5371 		default:
5372 			BUG();
5373 	}
5374 	return res;
5375 }
5376 
5377 struct nfs4_unlockdata {
5378 	struct nfs_locku_args arg;
5379 	struct nfs_locku_res res;
5380 	struct nfs4_lock_state *lsp;
5381 	struct nfs_open_context *ctx;
5382 	struct file_lock fl;
5383 	const struct nfs_server *server;
5384 	unsigned long timestamp;
5385 };
5386 
5387 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5388 		struct nfs_open_context *ctx,
5389 		struct nfs4_lock_state *lsp,
5390 		struct nfs_seqid *seqid)
5391 {
5392 	struct nfs4_unlockdata *p;
5393 	struct inode *inode = lsp->ls_state->inode;
5394 
5395 	p = kzalloc(sizeof(*p), GFP_NOFS);
5396 	if (p == NULL)
5397 		return NULL;
5398 	p->arg.fh = NFS_FH(inode);
5399 	p->arg.fl = &p->fl;
5400 	p->arg.seqid = seqid;
5401 	p->res.seqid = seqid;
5402 	p->lsp = lsp;
5403 	atomic_inc(&lsp->ls_count);
5404 	/* Ensure we don't close file until we're done freeing locks! */
5405 	p->ctx = get_nfs_open_context(ctx);
5406 	memcpy(&p->fl, fl, sizeof(p->fl));
5407 	p->server = NFS_SERVER(inode);
5408 	return p;
5409 }
5410 
5411 static void nfs4_locku_release_calldata(void *data)
5412 {
5413 	struct nfs4_unlockdata *calldata = data;
5414 	nfs_free_seqid(calldata->arg.seqid);
5415 	nfs4_put_lock_state(calldata->lsp);
5416 	put_nfs_open_context(calldata->ctx);
5417 	kfree(calldata);
5418 }
5419 
5420 static void nfs4_locku_done(struct rpc_task *task, void *data)
5421 {
5422 	struct nfs4_unlockdata *calldata = data;
5423 
5424 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5425 		return;
5426 	switch (task->tk_status) {
5427 		case 0:
5428 			renew_lease(calldata->server, calldata->timestamp);
5429 			do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5430 			if (nfs4_update_lock_stateid(calldata->lsp,
5431 					&calldata->res.stateid))
5432 				break;
5433 		case -NFS4ERR_BAD_STATEID:
5434 		case -NFS4ERR_OLD_STATEID:
5435 		case -NFS4ERR_STALE_STATEID:
5436 		case -NFS4ERR_EXPIRED:
5437 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5438 						&calldata->lsp->ls_stateid))
5439 				rpc_restart_call_prepare(task);
5440 			break;
5441 		default:
5442 			if (nfs4_async_handle_error(task, calldata->server,
5443 						    NULL, NULL) == -EAGAIN)
5444 				rpc_restart_call_prepare(task);
5445 	}
5446 	nfs_release_seqid(calldata->arg.seqid);
5447 }
5448 
5449 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5450 {
5451 	struct nfs4_unlockdata *calldata = data;
5452 
5453 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5454 		goto out_wait;
5455 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5456 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5457 		/* Note: exit _without_ running nfs4_locku_done */
5458 		goto out_no_action;
5459 	}
5460 	calldata->timestamp = jiffies;
5461 	if (nfs4_setup_sequence(calldata->server,
5462 				&calldata->arg.seq_args,
5463 				&calldata->res.seq_res,
5464 				task) != 0)
5465 		nfs_release_seqid(calldata->arg.seqid);
5466 	return;
5467 out_no_action:
5468 	task->tk_action = NULL;
5469 out_wait:
5470 	nfs4_sequence_done(task, &calldata->res.seq_res);
5471 }
5472 
5473 static const struct rpc_call_ops nfs4_locku_ops = {
5474 	.rpc_call_prepare = nfs4_locku_prepare,
5475 	.rpc_call_done = nfs4_locku_done,
5476 	.rpc_release = nfs4_locku_release_calldata,
5477 };
5478 
5479 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5480 		struct nfs_open_context *ctx,
5481 		struct nfs4_lock_state *lsp,
5482 		struct nfs_seqid *seqid)
5483 {
5484 	struct nfs4_unlockdata *data;
5485 	struct rpc_message msg = {
5486 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5487 		.rpc_cred = ctx->cred,
5488 	};
5489 	struct rpc_task_setup task_setup_data = {
5490 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5491 		.rpc_message = &msg,
5492 		.callback_ops = &nfs4_locku_ops,
5493 		.workqueue = nfsiod_workqueue,
5494 		.flags = RPC_TASK_ASYNC,
5495 	};
5496 
5497 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5498 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5499 
5500 	/* Ensure this is an unlock - when canceling a lock, the
5501 	 * canceled lock is passed in, and it won't be an unlock.
5502 	 */
5503 	fl->fl_type = F_UNLCK;
5504 
5505 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5506 	if (data == NULL) {
5507 		nfs_free_seqid(seqid);
5508 		return ERR_PTR(-ENOMEM);
5509 	}
5510 
5511 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5512 	msg.rpc_argp = &data->arg;
5513 	msg.rpc_resp = &data->res;
5514 	task_setup_data.callback_data = data;
5515 	return rpc_run_task(&task_setup_data);
5516 }
5517 
5518 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5519 {
5520 	struct inode *inode = state->inode;
5521 	struct nfs4_state_owner *sp = state->owner;
5522 	struct nfs_inode *nfsi = NFS_I(inode);
5523 	struct nfs_seqid *seqid;
5524 	struct nfs4_lock_state *lsp;
5525 	struct rpc_task *task;
5526 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5527 	int status = 0;
5528 	unsigned char fl_flags = request->fl_flags;
5529 
5530 	status = nfs4_set_lock_state(state, request);
5531 	/* Unlock _before_ we do the RPC call */
5532 	request->fl_flags |= FL_EXISTS;
5533 	/* Exclude nfs_delegation_claim_locks() */
5534 	mutex_lock(&sp->so_delegreturn_mutex);
5535 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5536 	down_read(&nfsi->rwsem);
5537 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5538 		up_read(&nfsi->rwsem);
5539 		mutex_unlock(&sp->so_delegreturn_mutex);
5540 		goto out;
5541 	}
5542 	up_read(&nfsi->rwsem);
5543 	mutex_unlock(&sp->so_delegreturn_mutex);
5544 	if (status != 0)
5545 		goto out;
5546 	/* Is this a delegated lock? */
5547 	lsp = request->fl_u.nfs4_fl.owner;
5548 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5549 		goto out;
5550 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5551 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5552 	status = -ENOMEM;
5553 	if (IS_ERR(seqid))
5554 		goto out;
5555 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5556 	status = PTR_ERR(task);
5557 	if (IS_ERR(task))
5558 		goto out;
5559 	status = nfs4_wait_for_completion_rpc_task(task);
5560 	rpc_put_task(task);
5561 out:
5562 	request->fl_flags = fl_flags;
5563 	trace_nfs4_unlock(request, state, F_SETLK, status);
5564 	return status;
5565 }
5566 
5567 struct nfs4_lockdata {
5568 	struct nfs_lock_args arg;
5569 	struct nfs_lock_res res;
5570 	struct nfs4_lock_state *lsp;
5571 	struct nfs_open_context *ctx;
5572 	struct file_lock fl;
5573 	unsigned long timestamp;
5574 	int rpc_status;
5575 	int cancelled;
5576 	struct nfs_server *server;
5577 };
5578 
5579 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5580 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5581 		gfp_t gfp_mask)
5582 {
5583 	struct nfs4_lockdata *p;
5584 	struct inode *inode = lsp->ls_state->inode;
5585 	struct nfs_server *server = NFS_SERVER(inode);
5586 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5587 
5588 	p = kzalloc(sizeof(*p), gfp_mask);
5589 	if (p == NULL)
5590 		return NULL;
5591 
5592 	p->arg.fh = NFS_FH(inode);
5593 	p->arg.fl = &p->fl;
5594 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5595 	if (IS_ERR(p->arg.open_seqid))
5596 		goto out_free;
5597 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5598 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5599 	if (IS_ERR(p->arg.lock_seqid))
5600 		goto out_free_seqid;
5601 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5602 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5603 	p->arg.lock_owner.s_dev = server->s_dev;
5604 	p->res.lock_seqid = p->arg.lock_seqid;
5605 	p->lsp = lsp;
5606 	p->server = server;
5607 	atomic_inc(&lsp->ls_count);
5608 	p->ctx = get_nfs_open_context(ctx);
5609 	memcpy(&p->fl, fl, sizeof(p->fl));
5610 	return p;
5611 out_free_seqid:
5612 	nfs_free_seqid(p->arg.open_seqid);
5613 out_free:
5614 	kfree(p);
5615 	return NULL;
5616 }
5617 
5618 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5619 {
5620 	struct nfs4_lockdata *data = calldata;
5621 	struct nfs4_state *state = data->lsp->ls_state;
5622 
5623 	dprintk("%s: begin!\n", __func__);
5624 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5625 		goto out_wait;
5626 	/* Do we need to do an open_to_lock_owner? */
5627 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5628 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5629 			goto out_release_lock_seqid;
5630 		}
5631 		nfs4_stateid_copy(&data->arg.open_stateid,
5632 				&state->open_stateid);
5633 		data->arg.new_lock_owner = 1;
5634 		data->res.open_seqid = data->arg.open_seqid;
5635 	} else {
5636 		data->arg.new_lock_owner = 0;
5637 		nfs4_stateid_copy(&data->arg.lock_stateid,
5638 				&data->lsp->ls_stateid);
5639 	}
5640 	if (!nfs4_valid_open_stateid(state)) {
5641 		data->rpc_status = -EBADF;
5642 		task->tk_action = NULL;
5643 		goto out_release_open_seqid;
5644 	}
5645 	data->timestamp = jiffies;
5646 	if (nfs4_setup_sequence(data->server,
5647 				&data->arg.seq_args,
5648 				&data->res.seq_res,
5649 				task) == 0)
5650 		return;
5651 out_release_open_seqid:
5652 	nfs_release_seqid(data->arg.open_seqid);
5653 out_release_lock_seqid:
5654 	nfs_release_seqid(data->arg.lock_seqid);
5655 out_wait:
5656 	nfs4_sequence_done(task, &data->res.seq_res);
5657 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5658 }
5659 
5660 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5661 {
5662 	struct nfs4_lockdata *data = calldata;
5663 	struct nfs4_lock_state *lsp = data->lsp;
5664 
5665 	dprintk("%s: begin!\n", __func__);
5666 
5667 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5668 		return;
5669 
5670 	data->rpc_status = task->tk_status;
5671 	switch (task->tk_status) {
5672 	case 0:
5673 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode),
5674 				data->timestamp);
5675 		if (data->arg.new_lock) {
5676 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5677 			if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5678 				rpc_restart_call_prepare(task);
5679 				break;
5680 			}
5681 		}
5682 		if (data->arg.new_lock_owner != 0) {
5683 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5684 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5685 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5686 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5687 			rpc_restart_call_prepare(task);
5688 		break;
5689 	case -NFS4ERR_BAD_STATEID:
5690 	case -NFS4ERR_OLD_STATEID:
5691 	case -NFS4ERR_STALE_STATEID:
5692 	case -NFS4ERR_EXPIRED:
5693 		if (data->arg.new_lock_owner != 0) {
5694 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5695 						&lsp->ls_state->open_stateid))
5696 				rpc_restart_call_prepare(task);
5697 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5698 						&lsp->ls_stateid))
5699 				rpc_restart_call_prepare(task);
5700 	}
5701 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5702 }
5703 
5704 static void nfs4_lock_release(void *calldata)
5705 {
5706 	struct nfs4_lockdata *data = calldata;
5707 
5708 	dprintk("%s: begin!\n", __func__);
5709 	nfs_free_seqid(data->arg.open_seqid);
5710 	if (data->cancelled != 0) {
5711 		struct rpc_task *task;
5712 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5713 				data->arg.lock_seqid);
5714 		if (!IS_ERR(task))
5715 			rpc_put_task_async(task);
5716 		dprintk("%s: cancelling lock!\n", __func__);
5717 	} else
5718 		nfs_free_seqid(data->arg.lock_seqid);
5719 	nfs4_put_lock_state(data->lsp);
5720 	put_nfs_open_context(data->ctx);
5721 	kfree(data);
5722 	dprintk("%s: done!\n", __func__);
5723 }
5724 
5725 static const struct rpc_call_ops nfs4_lock_ops = {
5726 	.rpc_call_prepare = nfs4_lock_prepare,
5727 	.rpc_call_done = nfs4_lock_done,
5728 	.rpc_release = nfs4_lock_release,
5729 };
5730 
5731 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5732 {
5733 	switch (error) {
5734 	case -NFS4ERR_ADMIN_REVOKED:
5735 	case -NFS4ERR_BAD_STATEID:
5736 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5737 		if (new_lock_owner != 0 ||
5738 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5739 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5740 		break;
5741 	case -NFS4ERR_STALE_STATEID:
5742 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5743 	case -NFS4ERR_EXPIRED:
5744 		nfs4_schedule_lease_recovery(server->nfs_client);
5745 	};
5746 }
5747 
5748 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5749 {
5750 	struct nfs4_lockdata *data;
5751 	struct rpc_task *task;
5752 	struct rpc_message msg = {
5753 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5754 		.rpc_cred = state->owner->so_cred,
5755 	};
5756 	struct rpc_task_setup task_setup_data = {
5757 		.rpc_client = NFS_CLIENT(state->inode),
5758 		.rpc_message = &msg,
5759 		.callback_ops = &nfs4_lock_ops,
5760 		.workqueue = nfsiod_workqueue,
5761 		.flags = RPC_TASK_ASYNC,
5762 	};
5763 	int ret;
5764 
5765 	dprintk("%s: begin!\n", __func__);
5766 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5767 			fl->fl_u.nfs4_fl.owner,
5768 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5769 	if (data == NULL)
5770 		return -ENOMEM;
5771 	if (IS_SETLKW(cmd))
5772 		data->arg.block = 1;
5773 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5774 	msg.rpc_argp = &data->arg;
5775 	msg.rpc_resp = &data->res;
5776 	task_setup_data.callback_data = data;
5777 	if (recovery_type > NFS_LOCK_NEW) {
5778 		if (recovery_type == NFS_LOCK_RECLAIM)
5779 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5780 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5781 	} else
5782 		data->arg.new_lock = 1;
5783 	task = rpc_run_task(&task_setup_data);
5784 	if (IS_ERR(task))
5785 		return PTR_ERR(task);
5786 	ret = nfs4_wait_for_completion_rpc_task(task);
5787 	if (ret == 0) {
5788 		ret = data->rpc_status;
5789 		if (ret)
5790 			nfs4_handle_setlk_error(data->server, data->lsp,
5791 					data->arg.new_lock_owner, ret);
5792 	} else
5793 		data->cancelled = 1;
5794 	rpc_put_task(task);
5795 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5796 	return ret;
5797 }
5798 
5799 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5800 {
5801 	struct nfs_server *server = NFS_SERVER(state->inode);
5802 	struct nfs4_exception exception = {
5803 		.inode = state->inode,
5804 	};
5805 	int err;
5806 
5807 	do {
5808 		/* Cache the lock if possible... */
5809 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5810 			return 0;
5811 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5812 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5813 		if (err != -NFS4ERR_DELAY)
5814 			break;
5815 		nfs4_handle_exception(server, err, &exception);
5816 	} while (exception.retry);
5817 	return err;
5818 }
5819 
5820 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5821 {
5822 	struct nfs_server *server = NFS_SERVER(state->inode);
5823 	struct nfs4_exception exception = {
5824 		.inode = state->inode,
5825 	};
5826 	int err;
5827 
5828 	err = nfs4_set_lock_state(state, request);
5829 	if (err != 0)
5830 		return err;
5831 	if (!recover_lost_locks) {
5832 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5833 		return 0;
5834 	}
5835 	do {
5836 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5837 			return 0;
5838 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5839 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5840 		switch (err) {
5841 		default:
5842 			goto out;
5843 		case -NFS4ERR_GRACE:
5844 		case -NFS4ERR_DELAY:
5845 			nfs4_handle_exception(server, err, &exception);
5846 			err = 0;
5847 		}
5848 	} while (exception.retry);
5849 out:
5850 	return err;
5851 }
5852 
5853 #if defined(CONFIG_NFS_V4_1)
5854 /**
5855  * nfs41_check_expired_locks - possibly free a lock stateid
5856  *
5857  * @state: NFSv4 state for an inode
5858  *
5859  * Returns NFS_OK if recovery for this stateid is now finished.
5860  * Otherwise a negative NFS4ERR value is returned.
5861  */
5862 static int nfs41_check_expired_locks(struct nfs4_state *state)
5863 {
5864 	int status, ret = -NFS4ERR_BAD_STATEID;
5865 	struct nfs4_lock_state *lsp;
5866 	struct nfs_server *server = NFS_SERVER(state->inode);
5867 
5868 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5869 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5870 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5871 
5872 			status = nfs41_test_stateid(server,
5873 					&lsp->ls_stateid,
5874 					cred);
5875 			trace_nfs4_test_lock_stateid(state, lsp, status);
5876 			if (status != NFS_OK) {
5877 				/* Free the stateid unless the server
5878 				 * informs us the stateid is unrecognized. */
5879 				if (status != -NFS4ERR_BAD_STATEID)
5880 					nfs41_free_stateid(server,
5881 							&lsp->ls_stateid,
5882 							cred);
5883 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5884 				ret = status;
5885 			}
5886 		}
5887 	};
5888 
5889 	return ret;
5890 }
5891 
5892 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5893 {
5894 	int status = NFS_OK;
5895 
5896 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5897 		status = nfs41_check_expired_locks(state);
5898 	if (status != NFS_OK)
5899 		status = nfs4_lock_expired(state, request);
5900 	return status;
5901 }
5902 #endif
5903 
5904 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5905 {
5906 	struct nfs_inode *nfsi = NFS_I(state->inode);
5907 	unsigned char fl_flags = request->fl_flags;
5908 	int status = -ENOLCK;
5909 
5910 	if ((fl_flags & FL_POSIX) &&
5911 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5912 		goto out;
5913 	/* Is this a delegated open? */
5914 	status = nfs4_set_lock_state(state, request);
5915 	if (status != 0)
5916 		goto out;
5917 	request->fl_flags |= FL_ACCESS;
5918 	status = do_vfs_lock(request->fl_file, request);
5919 	if (status < 0)
5920 		goto out;
5921 	down_read(&nfsi->rwsem);
5922 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5923 		/* Yes: cache locks! */
5924 		/* ...but avoid races with delegation recall... */
5925 		request->fl_flags = fl_flags & ~FL_SLEEP;
5926 		status = do_vfs_lock(request->fl_file, request);
5927 		up_read(&nfsi->rwsem);
5928 		goto out;
5929 	}
5930 	up_read(&nfsi->rwsem);
5931 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5932 out:
5933 	request->fl_flags = fl_flags;
5934 	return status;
5935 }
5936 
5937 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5938 {
5939 	struct nfs4_exception exception = {
5940 		.state = state,
5941 		.inode = state->inode,
5942 	};
5943 	int err;
5944 
5945 	do {
5946 		err = _nfs4_proc_setlk(state, cmd, request);
5947 		trace_nfs4_set_lock(request, state, cmd, err);
5948 		if (err == -NFS4ERR_DENIED)
5949 			err = -EAGAIN;
5950 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5951 				err, &exception);
5952 	} while (exception.retry);
5953 	return err;
5954 }
5955 
5956 static int
5957 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5958 {
5959 	struct nfs_open_context *ctx;
5960 	struct nfs4_state *state;
5961 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5962 	int status;
5963 
5964 	/* verify open state */
5965 	ctx = nfs_file_open_context(filp);
5966 	state = ctx->state;
5967 
5968 	if (request->fl_start < 0 || request->fl_end < 0)
5969 		return -EINVAL;
5970 
5971 	if (IS_GETLK(cmd)) {
5972 		if (state != NULL)
5973 			return nfs4_proc_getlk(state, F_GETLK, request);
5974 		return 0;
5975 	}
5976 
5977 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5978 		return -EINVAL;
5979 
5980 	if (request->fl_type == F_UNLCK) {
5981 		if (state != NULL)
5982 			return nfs4_proc_unlck(state, cmd, request);
5983 		return 0;
5984 	}
5985 
5986 	if (state == NULL)
5987 		return -ENOLCK;
5988 	/*
5989 	 * Don't rely on the VFS having checked the file open mode,
5990 	 * since it won't do this for flock() locks.
5991 	 */
5992 	switch (request->fl_type) {
5993 	case F_RDLCK:
5994 		if (!(filp->f_mode & FMODE_READ))
5995 			return -EBADF;
5996 		break;
5997 	case F_WRLCK:
5998 		if (!(filp->f_mode & FMODE_WRITE))
5999 			return -EBADF;
6000 	}
6001 
6002 	do {
6003 		status = nfs4_proc_setlk(state, cmd, request);
6004 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6005 			break;
6006 		timeout = nfs4_set_lock_task_retry(timeout);
6007 		status = -ERESTARTSYS;
6008 		if (signalled())
6009 			break;
6010 	} while(status < 0);
6011 	return status;
6012 }
6013 
6014 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6015 {
6016 	struct nfs_server *server = NFS_SERVER(state->inode);
6017 	int err;
6018 
6019 	err = nfs4_set_lock_state(state, fl);
6020 	if (err != 0)
6021 		return err;
6022 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6023 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6024 }
6025 
6026 struct nfs_release_lockowner_data {
6027 	struct nfs4_lock_state *lsp;
6028 	struct nfs_server *server;
6029 	struct nfs_release_lockowner_args args;
6030 	struct nfs_release_lockowner_res res;
6031 	unsigned long timestamp;
6032 };
6033 
6034 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6035 {
6036 	struct nfs_release_lockowner_data *data = calldata;
6037 	struct nfs_server *server = data->server;
6038 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6039 			     &data->args.seq_args, &data->res.seq_res, task);
6040 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6041 	data->timestamp = jiffies;
6042 }
6043 
6044 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6045 {
6046 	struct nfs_release_lockowner_data *data = calldata;
6047 	struct nfs_server *server = data->server;
6048 
6049 	nfs40_sequence_done(task, &data->res.seq_res);
6050 
6051 	switch (task->tk_status) {
6052 	case 0:
6053 		renew_lease(server, data->timestamp);
6054 		break;
6055 	case -NFS4ERR_STALE_CLIENTID:
6056 	case -NFS4ERR_EXPIRED:
6057 		nfs4_schedule_lease_recovery(server->nfs_client);
6058 		break;
6059 	case -NFS4ERR_LEASE_MOVED:
6060 	case -NFS4ERR_DELAY:
6061 		if (nfs4_async_handle_error(task, server,
6062 					    NULL, NULL) == -EAGAIN)
6063 			rpc_restart_call_prepare(task);
6064 	}
6065 }
6066 
6067 static void nfs4_release_lockowner_release(void *calldata)
6068 {
6069 	struct nfs_release_lockowner_data *data = calldata;
6070 	nfs4_free_lock_state(data->server, data->lsp);
6071 	kfree(calldata);
6072 }
6073 
6074 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6075 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6076 	.rpc_call_done = nfs4_release_lockowner_done,
6077 	.rpc_release = nfs4_release_lockowner_release,
6078 };
6079 
6080 static void
6081 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6082 {
6083 	struct nfs_release_lockowner_data *data;
6084 	struct rpc_message msg = {
6085 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6086 	};
6087 
6088 	if (server->nfs_client->cl_mvops->minor_version != 0)
6089 		return;
6090 
6091 	data = kmalloc(sizeof(*data), GFP_NOFS);
6092 	if (!data)
6093 		return;
6094 	data->lsp = lsp;
6095 	data->server = server;
6096 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6097 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6098 	data->args.lock_owner.s_dev = server->s_dev;
6099 
6100 	msg.rpc_argp = &data->args;
6101 	msg.rpc_resp = &data->res;
6102 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6103 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6104 }
6105 
6106 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6107 
6108 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6109 				   const void *buf, size_t buflen,
6110 				   int flags, int type)
6111 {
6112 	if (strcmp(key, "") != 0)
6113 		return -EINVAL;
6114 
6115 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6116 }
6117 
6118 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6119 				   void *buf, size_t buflen, int type)
6120 {
6121 	if (strcmp(key, "") != 0)
6122 		return -EINVAL;
6123 
6124 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6125 }
6126 
6127 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6128 				       size_t list_len, const char *name,
6129 				       size_t name_len, int type)
6130 {
6131 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6132 
6133 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6134 		return 0;
6135 
6136 	if (list && len <= list_len)
6137 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6138 	return len;
6139 }
6140 
6141 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6142 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6143 {
6144 	return server->caps & NFS_CAP_SECURITY_LABEL;
6145 }
6146 
6147 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6148 				   const void *buf, size_t buflen,
6149 				   int flags, int type)
6150 {
6151 	if (security_ismaclabel(key))
6152 		return nfs4_set_security_label(dentry, buf, buflen);
6153 
6154 	return -EOPNOTSUPP;
6155 }
6156 
6157 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6158 				   void *buf, size_t buflen, int type)
6159 {
6160 	if (security_ismaclabel(key))
6161 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6162 	return -EOPNOTSUPP;
6163 }
6164 
6165 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6166 				       size_t list_len, const char *name,
6167 				       size_t name_len, int type)
6168 {
6169 	size_t len = 0;
6170 
6171 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6172 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6173 		if (list && len <= list_len)
6174 			security_inode_listsecurity(dentry->d_inode, list, len);
6175 	}
6176 	return len;
6177 }
6178 
6179 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6180 	.prefix = XATTR_SECURITY_PREFIX,
6181 	.list	= nfs4_xattr_list_nfs4_label,
6182 	.get	= nfs4_xattr_get_nfs4_label,
6183 	.set	= nfs4_xattr_set_nfs4_label,
6184 };
6185 #endif
6186 
6187 
6188 /*
6189  * nfs_fhget will use either the mounted_on_fileid or the fileid
6190  */
6191 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6192 {
6193 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6194 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6195 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6196 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6197 		return;
6198 
6199 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6200 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6201 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6202 	fattr->nlink = 2;
6203 }
6204 
6205 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6206 				   const struct qstr *name,
6207 				   struct nfs4_fs_locations *fs_locations,
6208 				   struct page *page)
6209 {
6210 	struct nfs_server *server = NFS_SERVER(dir);
6211 	u32 bitmask[3] = {
6212 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6213 	};
6214 	struct nfs4_fs_locations_arg args = {
6215 		.dir_fh = NFS_FH(dir),
6216 		.name = name,
6217 		.page = page,
6218 		.bitmask = bitmask,
6219 	};
6220 	struct nfs4_fs_locations_res res = {
6221 		.fs_locations = fs_locations,
6222 	};
6223 	struct rpc_message msg = {
6224 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6225 		.rpc_argp = &args,
6226 		.rpc_resp = &res,
6227 	};
6228 	int status;
6229 
6230 	dprintk("%s: start\n", __func__);
6231 
6232 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6233 	 * is not supported */
6234 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6235 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6236 	else
6237 		bitmask[0] |= FATTR4_WORD0_FILEID;
6238 
6239 	nfs_fattr_init(&fs_locations->fattr);
6240 	fs_locations->server = server;
6241 	fs_locations->nlocations = 0;
6242 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6243 	dprintk("%s: returned status = %d\n", __func__, status);
6244 	return status;
6245 }
6246 
6247 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6248 			   const struct qstr *name,
6249 			   struct nfs4_fs_locations *fs_locations,
6250 			   struct page *page)
6251 {
6252 	struct nfs4_exception exception = { };
6253 	int err;
6254 	do {
6255 		err = _nfs4_proc_fs_locations(client, dir, name,
6256 				fs_locations, page);
6257 		trace_nfs4_get_fs_locations(dir, name, err);
6258 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6259 				&exception);
6260 	} while (exception.retry);
6261 	return err;
6262 }
6263 
6264 /*
6265  * This operation also signals the server that this client is
6266  * performing migration recovery.  The server can stop returning
6267  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6268  * appended to this compound to identify the client ID which is
6269  * performing recovery.
6270  */
6271 static int _nfs40_proc_get_locations(struct inode *inode,
6272 				     struct nfs4_fs_locations *locations,
6273 				     struct page *page, struct rpc_cred *cred)
6274 {
6275 	struct nfs_server *server = NFS_SERVER(inode);
6276 	struct rpc_clnt *clnt = server->client;
6277 	u32 bitmask[2] = {
6278 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6279 	};
6280 	struct nfs4_fs_locations_arg args = {
6281 		.clientid	= server->nfs_client->cl_clientid,
6282 		.fh		= NFS_FH(inode),
6283 		.page		= page,
6284 		.bitmask	= bitmask,
6285 		.migration	= 1,		/* skip LOOKUP */
6286 		.renew		= 1,		/* append RENEW */
6287 	};
6288 	struct nfs4_fs_locations_res res = {
6289 		.fs_locations	= locations,
6290 		.migration	= 1,
6291 		.renew		= 1,
6292 	};
6293 	struct rpc_message msg = {
6294 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6295 		.rpc_argp	= &args,
6296 		.rpc_resp	= &res,
6297 		.rpc_cred	= cred,
6298 	};
6299 	unsigned long now = jiffies;
6300 	int status;
6301 
6302 	nfs_fattr_init(&locations->fattr);
6303 	locations->server = server;
6304 	locations->nlocations = 0;
6305 
6306 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6307 	nfs4_set_sequence_privileged(&args.seq_args);
6308 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6309 					&args.seq_args, &res.seq_res);
6310 	if (status)
6311 		return status;
6312 
6313 	renew_lease(server, now);
6314 	return 0;
6315 }
6316 
6317 #ifdef CONFIG_NFS_V4_1
6318 
6319 /*
6320  * This operation also signals the server that this client is
6321  * performing migration recovery.  The server can stop asserting
6322  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6323  * performing this operation is identified in the SEQUENCE
6324  * operation in this compound.
6325  *
6326  * When the client supports GETATTR(fs_locations_info), it can
6327  * be plumbed in here.
6328  */
6329 static int _nfs41_proc_get_locations(struct inode *inode,
6330 				     struct nfs4_fs_locations *locations,
6331 				     struct page *page, struct rpc_cred *cred)
6332 {
6333 	struct nfs_server *server = NFS_SERVER(inode);
6334 	struct rpc_clnt *clnt = server->client;
6335 	u32 bitmask[2] = {
6336 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6337 	};
6338 	struct nfs4_fs_locations_arg args = {
6339 		.fh		= NFS_FH(inode),
6340 		.page		= page,
6341 		.bitmask	= bitmask,
6342 		.migration	= 1,		/* skip LOOKUP */
6343 	};
6344 	struct nfs4_fs_locations_res res = {
6345 		.fs_locations	= locations,
6346 		.migration	= 1,
6347 	};
6348 	struct rpc_message msg = {
6349 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6350 		.rpc_argp	= &args,
6351 		.rpc_resp	= &res,
6352 		.rpc_cred	= cred,
6353 	};
6354 	int status;
6355 
6356 	nfs_fattr_init(&locations->fattr);
6357 	locations->server = server;
6358 	locations->nlocations = 0;
6359 
6360 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6361 	nfs4_set_sequence_privileged(&args.seq_args);
6362 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6363 					&args.seq_args, &res.seq_res);
6364 	if (status == NFS4_OK &&
6365 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6366 		status = -NFS4ERR_LEASE_MOVED;
6367 	return status;
6368 }
6369 
6370 #endif	/* CONFIG_NFS_V4_1 */
6371 
6372 /**
6373  * nfs4_proc_get_locations - discover locations for a migrated FSID
6374  * @inode: inode on FSID that is migrating
6375  * @locations: result of query
6376  * @page: buffer
6377  * @cred: credential to use for this operation
6378  *
6379  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6380  * operation failed, or a negative errno if a local error occurred.
6381  *
6382  * On success, "locations" is filled in, but if the server has
6383  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6384  * asserted.
6385  *
6386  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6387  * from this client that require migration recovery.
6388  */
6389 int nfs4_proc_get_locations(struct inode *inode,
6390 			    struct nfs4_fs_locations *locations,
6391 			    struct page *page, struct rpc_cred *cred)
6392 {
6393 	struct nfs_server *server = NFS_SERVER(inode);
6394 	struct nfs_client *clp = server->nfs_client;
6395 	const struct nfs4_mig_recovery_ops *ops =
6396 					clp->cl_mvops->mig_recovery_ops;
6397 	struct nfs4_exception exception = { };
6398 	int status;
6399 
6400 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6401 		(unsigned long long)server->fsid.major,
6402 		(unsigned long long)server->fsid.minor,
6403 		clp->cl_hostname);
6404 	nfs_display_fhandle(NFS_FH(inode), __func__);
6405 
6406 	do {
6407 		status = ops->get_locations(inode, locations, page, cred);
6408 		if (status != -NFS4ERR_DELAY)
6409 			break;
6410 		nfs4_handle_exception(server, status, &exception);
6411 	} while (exception.retry);
6412 	return status;
6413 }
6414 
6415 /*
6416  * This operation also signals the server that this client is
6417  * performing "lease moved" recovery.  The server can stop
6418  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6419  * is appended to this compound to identify the client ID which is
6420  * performing recovery.
6421  */
6422 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6423 {
6424 	struct nfs_server *server = NFS_SERVER(inode);
6425 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6426 	struct rpc_clnt *clnt = server->client;
6427 	struct nfs4_fsid_present_arg args = {
6428 		.fh		= NFS_FH(inode),
6429 		.clientid	= clp->cl_clientid,
6430 		.renew		= 1,		/* append RENEW */
6431 	};
6432 	struct nfs4_fsid_present_res res = {
6433 		.renew		= 1,
6434 	};
6435 	struct rpc_message msg = {
6436 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6437 		.rpc_argp	= &args,
6438 		.rpc_resp	= &res,
6439 		.rpc_cred	= cred,
6440 	};
6441 	unsigned long now = jiffies;
6442 	int status;
6443 
6444 	res.fh = nfs_alloc_fhandle();
6445 	if (res.fh == NULL)
6446 		return -ENOMEM;
6447 
6448 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6449 	nfs4_set_sequence_privileged(&args.seq_args);
6450 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6451 						&args.seq_args, &res.seq_res);
6452 	nfs_free_fhandle(res.fh);
6453 	if (status)
6454 		return status;
6455 
6456 	do_renew_lease(clp, now);
6457 	return 0;
6458 }
6459 
6460 #ifdef CONFIG_NFS_V4_1
6461 
6462 /*
6463  * This operation also signals the server that this client is
6464  * performing "lease moved" recovery.  The server can stop asserting
6465  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6466  * this operation is identified in the SEQUENCE operation in this
6467  * compound.
6468  */
6469 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6470 {
6471 	struct nfs_server *server = NFS_SERVER(inode);
6472 	struct rpc_clnt *clnt = server->client;
6473 	struct nfs4_fsid_present_arg args = {
6474 		.fh		= NFS_FH(inode),
6475 	};
6476 	struct nfs4_fsid_present_res res = {
6477 	};
6478 	struct rpc_message msg = {
6479 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6480 		.rpc_argp	= &args,
6481 		.rpc_resp	= &res,
6482 		.rpc_cred	= cred,
6483 	};
6484 	int status;
6485 
6486 	res.fh = nfs_alloc_fhandle();
6487 	if (res.fh == NULL)
6488 		return -ENOMEM;
6489 
6490 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6491 	nfs4_set_sequence_privileged(&args.seq_args);
6492 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6493 						&args.seq_args, &res.seq_res);
6494 	nfs_free_fhandle(res.fh);
6495 	if (status == NFS4_OK &&
6496 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6497 		status = -NFS4ERR_LEASE_MOVED;
6498 	return status;
6499 }
6500 
6501 #endif	/* CONFIG_NFS_V4_1 */
6502 
6503 /**
6504  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6505  * @inode: inode on FSID to check
6506  * @cred: credential to use for this operation
6507  *
6508  * Server indicates whether the FSID is present, moved, or not
6509  * recognized.  This operation is necessary to clear a LEASE_MOVED
6510  * condition for this client ID.
6511  *
6512  * Returns NFS4_OK if the FSID is present on this server,
6513  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6514  *  NFS4ERR code if some error occurred on the server, or a
6515  *  negative errno if a local failure occurred.
6516  */
6517 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6518 {
6519 	struct nfs_server *server = NFS_SERVER(inode);
6520 	struct nfs_client *clp = server->nfs_client;
6521 	const struct nfs4_mig_recovery_ops *ops =
6522 					clp->cl_mvops->mig_recovery_ops;
6523 	struct nfs4_exception exception = { };
6524 	int status;
6525 
6526 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6527 		(unsigned long long)server->fsid.major,
6528 		(unsigned long long)server->fsid.minor,
6529 		clp->cl_hostname);
6530 	nfs_display_fhandle(NFS_FH(inode), __func__);
6531 
6532 	do {
6533 		status = ops->fsid_present(inode, cred);
6534 		if (status != -NFS4ERR_DELAY)
6535 			break;
6536 		nfs4_handle_exception(server, status, &exception);
6537 	} while (exception.retry);
6538 	return status;
6539 }
6540 
6541 /**
6542  * If 'use_integrity' is true and the state managment nfs_client
6543  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6544  * and the machine credential as per RFC3530bis and RFC5661 Security
6545  * Considerations sections. Otherwise, just use the user cred with the
6546  * filesystem's rpc_client.
6547  */
6548 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6549 {
6550 	int status;
6551 	struct nfs4_secinfo_arg args = {
6552 		.dir_fh = NFS_FH(dir),
6553 		.name   = name,
6554 	};
6555 	struct nfs4_secinfo_res res = {
6556 		.flavors     = flavors,
6557 	};
6558 	struct rpc_message msg = {
6559 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6560 		.rpc_argp = &args,
6561 		.rpc_resp = &res,
6562 	};
6563 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6564 	struct rpc_cred *cred = NULL;
6565 
6566 	if (use_integrity) {
6567 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6568 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6569 		msg.rpc_cred = cred;
6570 	}
6571 
6572 	dprintk("NFS call  secinfo %s\n", name->name);
6573 
6574 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6575 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6576 
6577 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6578 				&res.seq_res, 0);
6579 	dprintk("NFS reply  secinfo: %d\n", status);
6580 
6581 	if (cred)
6582 		put_rpccred(cred);
6583 
6584 	return status;
6585 }
6586 
6587 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6588 		      struct nfs4_secinfo_flavors *flavors)
6589 {
6590 	struct nfs4_exception exception = { };
6591 	int err;
6592 	do {
6593 		err = -NFS4ERR_WRONGSEC;
6594 
6595 		/* try to use integrity protection with machine cred */
6596 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6597 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6598 
6599 		/*
6600 		 * if unable to use integrity protection, or SECINFO with
6601 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6602 		 * disallowed by spec, but exists in deployed servers) use
6603 		 * the current filesystem's rpc_client and the user cred.
6604 		 */
6605 		if (err == -NFS4ERR_WRONGSEC)
6606 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6607 
6608 		trace_nfs4_secinfo(dir, name, err);
6609 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6610 				&exception);
6611 	} while (exception.retry);
6612 	return err;
6613 }
6614 
6615 #ifdef CONFIG_NFS_V4_1
6616 /*
6617  * Check the exchange flags returned by the server for invalid flags, having
6618  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6619  * DS flags set.
6620  */
6621 static int nfs4_check_cl_exchange_flags(u32 flags)
6622 {
6623 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6624 		goto out_inval;
6625 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6626 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6627 		goto out_inval;
6628 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6629 		goto out_inval;
6630 	return NFS_OK;
6631 out_inval:
6632 	return -NFS4ERR_INVAL;
6633 }
6634 
6635 static bool
6636 nfs41_same_server_scope(struct nfs41_server_scope *a,
6637 			struct nfs41_server_scope *b)
6638 {
6639 	if (a->server_scope_sz == b->server_scope_sz &&
6640 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6641 		return true;
6642 
6643 	return false;
6644 }
6645 
6646 /*
6647  * nfs4_proc_bind_conn_to_session()
6648  *
6649  * The 4.1 client currently uses the same TCP connection for the
6650  * fore and backchannel.
6651  */
6652 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6653 {
6654 	int status;
6655 	struct nfs41_bind_conn_to_session_args args = {
6656 		.client = clp,
6657 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6658 	};
6659 	struct nfs41_bind_conn_to_session_res res;
6660 	struct rpc_message msg = {
6661 		.rpc_proc =
6662 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6663 		.rpc_argp = &args,
6664 		.rpc_resp = &res,
6665 		.rpc_cred = cred,
6666 	};
6667 
6668 	dprintk("--> %s\n", __func__);
6669 
6670 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6671 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6672 		args.dir = NFS4_CDFC4_FORE;
6673 
6674 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6675 	trace_nfs4_bind_conn_to_session(clp, status);
6676 	if (status == 0) {
6677 		if (memcmp(res.sessionid.data,
6678 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6679 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6680 			status = -EIO;
6681 			goto out;
6682 		}
6683 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6684 			dprintk("NFS: %s: Unexpected direction from server\n",
6685 				__func__);
6686 			status = -EIO;
6687 			goto out;
6688 		}
6689 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6690 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6691 				__func__);
6692 			status = -EIO;
6693 			goto out;
6694 		}
6695 	}
6696 out:
6697 	dprintk("<-- %s status= %d\n", __func__, status);
6698 	return status;
6699 }
6700 
6701 /*
6702  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6703  * and operations we'd like to see to enable certain features in the allow map
6704  */
6705 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6706 	.how = SP4_MACH_CRED,
6707 	.enforce.u.words = {
6708 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6709 		      1 << (OP_EXCHANGE_ID - 32) |
6710 		      1 << (OP_CREATE_SESSION - 32) |
6711 		      1 << (OP_DESTROY_SESSION - 32) |
6712 		      1 << (OP_DESTROY_CLIENTID - 32)
6713 	},
6714 	.allow.u.words = {
6715 		[0] = 1 << (OP_CLOSE) |
6716 		      1 << (OP_LOCKU) |
6717 		      1 << (OP_COMMIT),
6718 		[1] = 1 << (OP_SECINFO - 32) |
6719 		      1 << (OP_SECINFO_NO_NAME - 32) |
6720 		      1 << (OP_TEST_STATEID - 32) |
6721 		      1 << (OP_FREE_STATEID - 32) |
6722 		      1 << (OP_WRITE - 32)
6723 	}
6724 };
6725 
6726 /*
6727  * Select the state protection mode for client `clp' given the server results
6728  * from exchange_id in `sp'.
6729  *
6730  * Returns 0 on success, negative errno otherwise.
6731  */
6732 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6733 				 struct nfs41_state_protection *sp)
6734 {
6735 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6736 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6737 		      1 << (OP_EXCHANGE_ID - 32) |
6738 		      1 << (OP_CREATE_SESSION - 32) |
6739 		      1 << (OP_DESTROY_SESSION - 32) |
6740 		      1 << (OP_DESTROY_CLIENTID - 32)
6741 	};
6742 	unsigned int i;
6743 
6744 	if (sp->how == SP4_MACH_CRED) {
6745 		/* Print state protect result */
6746 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6747 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6748 			if (test_bit(i, sp->enforce.u.longs))
6749 				dfprintk(MOUNT, "  enforce op %d\n", i);
6750 			if (test_bit(i, sp->allow.u.longs))
6751 				dfprintk(MOUNT, "  allow op %d\n", i);
6752 		}
6753 
6754 		/* make sure nothing is on enforce list that isn't supported */
6755 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6756 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6757 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6758 				return -EINVAL;
6759 			}
6760 		}
6761 
6762 		/*
6763 		 * Minimal mode - state operations are allowed to use machine
6764 		 * credential.  Note this already happens by default, so the
6765 		 * client doesn't have to do anything more than the negotiation.
6766 		 *
6767 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6768 		 *       we're already using the machine cred for exchange_id
6769 		 *       and will never use a different cred.
6770 		 */
6771 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6772 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6773 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6774 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6775 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6776 			dfprintk(MOUNT, "  minimal mode enabled\n");
6777 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6778 		} else {
6779 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6780 			return -EINVAL;
6781 		}
6782 
6783 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6784 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6785 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6786 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6787 		}
6788 
6789 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6790 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6791 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6792 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6793 		}
6794 
6795 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6796 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6797 			dfprintk(MOUNT, "  stateid mode enabled\n");
6798 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6799 		}
6800 
6801 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6802 			dfprintk(MOUNT, "  write mode enabled\n");
6803 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6804 		}
6805 
6806 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6807 			dfprintk(MOUNT, "  commit mode enabled\n");
6808 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6809 		}
6810 	}
6811 
6812 	return 0;
6813 }
6814 
6815 /*
6816  * _nfs4_proc_exchange_id()
6817  *
6818  * Wrapper for EXCHANGE_ID operation.
6819  */
6820 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6821 	u32 sp4_how)
6822 {
6823 	nfs4_verifier verifier;
6824 	struct nfs41_exchange_id_args args = {
6825 		.verifier = &verifier,
6826 		.client = clp,
6827 #ifdef CONFIG_NFS_V4_1_MIGRATION
6828 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6829 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6830 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6831 #else
6832 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6833 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6834 #endif
6835 	};
6836 	struct nfs41_exchange_id_res res = {
6837 		0
6838 	};
6839 	int status;
6840 	struct rpc_message msg = {
6841 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6842 		.rpc_argp = &args,
6843 		.rpc_resp = &res,
6844 		.rpc_cred = cred,
6845 	};
6846 
6847 	nfs4_init_boot_verifier(clp, &verifier);
6848 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6849 							sizeof(args.id));
6850 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6851 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6852 		args.id_len, args.id);
6853 
6854 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6855 					GFP_NOFS);
6856 	if (unlikely(res.server_owner == NULL)) {
6857 		status = -ENOMEM;
6858 		goto out;
6859 	}
6860 
6861 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6862 					GFP_NOFS);
6863 	if (unlikely(res.server_scope == NULL)) {
6864 		status = -ENOMEM;
6865 		goto out_server_owner;
6866 	}
6867 
6868 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6869 	if (unlikely(res.impl_id == NULL)) {
6870 		status = -ENOMEM;
6871 		goto out_server_scope;
6872 	}
6873 
6874 	switch (sp4_how) {
6875 	case SP4_NONE:
6876 		args.state_protect.how = SP4_NONE;
6877 		break;
6878 
6879 	case SP4_MACH_CRED:
6880 		args.state_protect = nfs4_sp4_mach_cred_request;
6881 		break;
6882 
6883 	default:
6884 		/* unsupported! */
6885 		WARN_ON_ONCE(1);
6886 		status = -EINVAL;
6887 		goto out_server_scope;
6888 	}
6889 
6890 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6891 	trace_nfs4_exchange_id(clp, status);
6892 	if (status == 0)
6893 		status = nfs4_check_cl_exchange_flags(res.flags);
6894 
6895 	if (status == 0)
6896 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6897 
6898 	if (status == 0) {
6899 		clp->cl_clientid = res.clientid;
6900 		clp->cl_exchange_flags = res.flags;
6901 		/* Client ID is not confirmed */
6902 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6903 			clear_bit(NFS4_SESSION_ESTABLISHED,
6904 					&clp->cl_session->session_state);
6905 			clp->cl_seqid = res.seqid;
6906 		}
6907 
6908 		kfree(clp->cl_serverowner);
6909 		clp->cl_serverowner = res.server_owner;
6910 		res.server_owner = NULL;
6911 
6912 		/* use the most recent implementation id */
6913 		kfree(clp->cl_implid);
6914 		clp->cl_implid = res.impl_id;
6915 
6916 		if (clp->cl_serverscope != NULL &&
6917 		    !nfs41_same_server_scope(clp->cl_serverscope,
6918 					     res.server_scope)) {
6919 			dprintk("%s: server_scope mismatch detected\n",
6920 				__func__);
6921 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6922 			kfree(clp->cl_serverscope);
6923 			clp->cl_serverscope = NULL;
6924 		}
6925 
6926 		if (clp->cl_serverscope == NULL) {
6927 			clp->cl_serverscope = res.server_scope;
6928 			goto out;
6929 		}
6930 	} else
6931 		kfree(res.impl_id);
6932 
6933 out_server_owner:
6934 	kfree(res.server_owner);
6935 out_server_scope:
6936 	kfree(res.server_scope);
6937 out:
6938 	if (clp->cl_implid != NULL)
6939 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6940 			"domain: %s, name: %s, date: %llu,%u\n",
6941 			clp->cl_implid->domain, clp->cl_implid->name,
6942 			clp->cl_implid->date.seconds,
6943 			clp->cl_implid->date.nseconds);
6944 	dprintk("NFS reply exchange_id: %d\n", status);
6945 	return status;
6946 }
6947 
6948 /*
6949  * nfs4_proc_exchange_id()
6950  *
6951  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6952  *
6953  * Since the clientid has expired, all compounds using sessions
6954  * associated with the stale clientid will be returning
6955  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6956  * be in some phase of session reset.
6957  *
6958  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6959  */
6960 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6961 {
6962 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6963 	int status;
6964 
6965 	/* try SP4_MACH_CRED if krb5i/p	*/
6966 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6967 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6968 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6969 		if (!status)
6970 			return 0;
6971 	}
6972 
6973 	/* try SP4_NONE */
6974 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6975 }
6976 
6977 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6978 		struct rpc_cred *cred)
6979 {
6980 	struct rpc_message msg = {
6981 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6982 		.rpc_argp = clp,
6983 		.rpc_cred = cred,
6984 	};
6985 	int status;
6986 
6987 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6988 	trace_nfs4_destroy_clientid(clp, status);
6989 	if (status)
6990 		dprintk("NFS: Got error %d from the server %s on "
6991 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6992 	return status;
6993 }
6994 
6995 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6996 		struct rpc_cred *cred)
6997 {
6998 	unsigned int loop;
6999 	int ret;
7000 
7001 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7002 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7003 		switch (ret) {
7004 		case -NFS4ERR_DELAY:
7005 		case -NFS4ERR_CLIENTID_BUSY:
7006 			ssleep(1);
7007 			break;
7008 		default:
7009 			return ret;
7010 		}
7011 	}
7012 	return 0;
7013 }
7014 
7015 int nfs4_destroy_clientid(struct nfs_client *clp)
7016 {
7017 	struct rpc_cred *cred;
7018 	int ret = 0;
7019 
7020 	if (clp->cl_mvops->minor_version < 1)
7021 		goto out;
7022 	if (clp->cl_exchange_flags == 0)
7023 		goto out;
7024 	if (clp->cl_preserve_clid)
7025 		goto out;
7026 	cred = nfs4_get_clid_cred(clp);
7027 	ret = nfs4_proc_destroy_clientid(clp, cred);
7028 	if (cred)
7029 		put_rpccred(cred);
7030 	switch (ret) {
7031 	case 0:
7032 	case -NFS4ERR_STALE_CLIENTID:
7033 		clp->cl_exchange_flags = 0;
7034 	}
7035 out:
7036 	return ret;
7037 }
7038 
7039 struct nfs4_get_lease_time_data {
7040 	struct nfs4_get_lease_time_args *args;
7041 	struct nfs4_get_lease_time_res *res;
7042 	struct nfs_client *clp;
7043 };
7044 
7045 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7046 					void *calldata)
7047 {
7048 	struct nfs4_get_lease_time_data *data =
7049 			(struct nfs4_get_lease_time_data *)calldata;
7050 
7051 	dprintk("--> %s\n", __func__);
7052 	/* just setup sequence, do not trigger session recovery
7053 	   since we're invoked within one */
7054 	nfs41_setup_sequence(data->clp->cl_session,
7055 			&data->args->la_seq_args,
7056 			&data->res->lr_seq_res,
7057 			task);
7058 	dprintk("<-- %s\n", __func__);
7059 }
7060 
7061 /*
7062  * Called from nfs4_state_manager thread for session setup, so don't recover
7063  * from sequence operation or clientid errors.
7064  */
7065 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7066 {
7067 	struct nfs4_get_lease_time_data *data =
7068 			(struct nfs4_get_lease_time_data *)calldata;
7069 
7070 	dprintk("--> %s\n", __func__);
7071 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7072 		return;
7073 	switch (task->tk_status) {
7074 	case -NFS4ERR_DELAY:
7075 	case -NFS4ERR_GRACE:
7076 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7077 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7078 		task->tk_status = 0;
7079 		/* fall through */
7080 	case -NFS4ERR_RETRY_UNCACHED_REP:
7081 		rpc_restart_call_prepare(task);
7082 		return;
7083 	}
7084 	dprintk("<-- %s\n", __func__);
7085 }
7086 
7087 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7088 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7089 	.rpc_call_done = nfs4_get_lease_time_done,
7090 };
7091 
7092 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7093 {
7094 	struct rpc_task *task;
7095 	struct nfs4_get_lease_time_args args;
7096 	struct nfs4_get_lease_time_res res = {
7097 		.lr_fsinfo = fsinfo,
7098 	};
7099 	struct nfs4_get_lease_time_data data = {
7100 		.args = &args,
7101 		.res = &res,
7102 		.clp = clp,
7103 	};
7104 	struct rpc_message msg = {
7105 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7106 		.rpc_argp = &args,
7107 		.rpc_resp = &res,
7108 	};
7109 	struct rpc_task_setup task_setup = {
7110 		.rpc_client = clp->cl_rpcclient,
7111 		.rpc_message = &msg,
7112 		.callback_ops = &nfs4_get_lease_time_ops,
7113 		.callback_data = &data,
7114 		.flags = RPC_TASK_TIMEOUT,
7115 	};
7116 	int status;
7117 
7118 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7119 	nfs4_set_sequence_privileged(&args.la_seq_args);
7120 	dprintk("--> %s\n", __func__);
7121 	task = rpc_run_task(&task_setup);
7122 
7123 	if (IS_ERR(task))
7124 		status = PTR_ERR(task);
7125 	else {
7126 		status = task->tk_status;
7127 		rpc_put_task(task);
7128 	}
7129 	dprintk("<-- %s return %d\n", __func__, status);
7130 
7131 	return status;
7132 }
7133 
7134 /*
7135  * Initialize the values to be used by the client in CREATE_SESSION
7136  * If nfs4_init_session set the fore channel request and response sizes,
7137  * use them.
7138  *
7139  * Set the back channel max_resp_sz_cached to zero to force the client to
7140  * always set csa_cachethis to FALSE because the current implementation
7141  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7142  */
7143 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7144 {
7145 	unsigned int max_rqst_sz, max_resp_sz;
7146 
7147 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7148 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7149 
7150 	/* Fore channel attributes */
7151 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7152 	args->fc_attrs.max_resp_sz = max_resp_sz;
7153 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7154 	args->fc_attrs.max_reqs = max_session_slots;
7155 
7156 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7157 		"max_ops=%u max_reqs=%u\n",
7158 		__func__,
7159 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7160 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7161 
7162 	/* Back channel attributes */
7163 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7164 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7165 	args->bc_attrs.max_resp_sz_cached = 0;
7166 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7167 	args->bc_attrs.max_reqs = 1;
7168 
7169 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7170 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7171 		__func__,
7172 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7173 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7174 		args->bc_attrs.max_reqs);
7175 }
7176 
7177 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7178 		struct nfs41_create_session_res *res)
7179 {
7180 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7181 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7182 
7183 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7184 		return -EINVAL;
7185 	/*
7186 	 * Our requested max_ops is the minimum we need; we're not
7187 	 * prepared to break up compounds into smaller pieces than that.
7188 	 * So, no point even trying to continue if the server won't
7189 	 * cooperate:
7190 	 */
7191 	if (rcvd->max_ops < sent->max_ops)
7192 		return -EINVAL;
7193 	if (rcvd->max_reqs == 0)
7194 		return -EINVAL;
7195 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7196 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7197 	return 0;
7198 }
7199 
7200 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7201 		struct nfs41_create_session_res *res)
7202 {
7203 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7204 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7205 
7206 	if (!(res->flags & SESSION4_BACK_CHAN))
7207 		goto out;
7208 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7209 		return -EINVAL;
7210 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7211 		return -EINVAL;
7212 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7213 		return -EINVAL;
7214 	/* These would render the backchannel useless: */
7215 	if (rcvd->max_ops != sent->max_ops)
7216 		return -EINVAL;
7217 	if (rcvd->max_reqs != sent->max_reqs)
7218 		return -EINVAL;
7219 out:
7220 	return 0;
7221 }
7222 
7223 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7224 				     struct nfs41_create_session_res *res)
7225 {
7226 	int ret;
7227 
7228 	ret = nfs4_verify_fore_channel_attrs(args, res);
7229 	if (ret)
7230 		return ret;
7231 	return nfs4_verify_back_channel_attrs(args, res);
7232 }
7233 
7234 static void nfs4_update_session(struct nfs4_session *session,
7235 		struct nfs41_create_session_res *res)
7236 {
7237 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7238 	/* Mark client id and session as being confirmed */
7239 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7240 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7241 	session->flags = res->flags;
7242 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7243 	if (res->flags & SESSION4_BACK_CHAN)
7244 		memcpy(&session->bc_attrs, &res->bc_attrs,
7245 				sizeof(session->bc_attrs));
7246 }
7247 
7248 static int _nfs4_proc_create_session(struct nfs_client *clp,
7249 		struct rpc_cred *cred)
7250 {
7251 	struct nfs4_session *session = clp->cl_session;
7252 	struct nfs41_create_session_args args = {
7253 		.client = clp,
7254 		.clientid = clp->cl_clientid,
7255 		.seqid = clp->cl_seqid,
7256 		.cb_program = NFS4_CALLBACK,
7257 	};
7258 	struct nfs41_create_session_res res;
7259 
7260 	struct rpc_message msg = {
7261 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7262 		.rpc_argp = &args,
7263 		.rpc_resp = &res,
7264 		.rpc_cred = cred,
7265 	};
7266 	int status;
7267 
7268 	nfs4_init_channel_attrs(&args);
7269 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7270 
7271 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7272 	trace_nfs4_create_session(clp, status);
7273 
7274 	if (!status) {
7275 		/* Verify the session's negotiated channel_attrs values */
7276 		status = nfs4_verify_channel_attrs(&args, &res);
7277 		/* Increment the clientid slot sequence id */
7278 		if (clp->cl_seqid == res.seqid)
7279 			clp->cl_seqid++;
7280 		if (status)
7281 			goto out;
7282 		nfs4_update_session(session, &res);
7283 	}
7284 out:
7285 	return status;
7286 }
7287 
7288 /*
7289  * Issues a CREATE_SESSION operation to the server.
7290  * It is the responsibility of the caller to verify the session is
7291  * expired before calling this routine.
7292  */
7293 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7294 {
7295 	int status;
7296 	unsigned *ptr;
7297 	struct nfs4_session *session = clp->cl_session;
7298 
7299 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7300 
7301 	status = _nfs4_proc_create_session(clp, cred);
7302 	if (status)
7303 		goto out;
7304 
7305 	/* Init or reset the session slot tables */
7306 	status = nfs4_setup_session_slot_tables(session);
7307 	dprintk("slot table setup returned %d\n", status);
7308 	if (status)
7309 		goto out;
7310 
7311 	ptr = (unsigned *)&session->sess_id.data[0];
7312 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7313 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7314 out:
7315 	dprintk("<-- %s\n", __func__);
7316 	return status;
7317 }
7318 
7319 /*
7320  * Issue the over-the-wire RPC DESTROY_SESSION.
7321  * The caller must serialize access to this routine.
7322  */
7323 int nfs4_proc_destroy_session(struct nfs4_session *session,
7324 		struct rpc_cred *cred)
7325 {
7326 	struct rpc_message msg = {
7327 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7328 		.rpc_argp = session,
7329 		.rpc_cred = cred,
7330 	};
7331 	int status = 0;
7332 
7333 	dprintk("--> nfs4_proc_destroy_session\n");
7334 
7335 	/* session is still being setup */
7336 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7337 		return 0;
7338 
7339 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7340 	trace_nfs4_destroy_session(session->clp, status);
7341 
7342 	if (status)
7343 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7344 			"Session has been destroyed regardless...\n", status);
7345 
7346 	dprintk("<-- nfs4_proc_destroy_session\n");
7347 	return status;
7348 }
7349 
7350 /*
7351  * Renew the cl_session lease.
7352  */
7353 struct nfs4_sequence_data {
7354 	struct nfs_client *clp;
7355 	struct nfs4_sequence_args args;
7356 	struct nfs4_sequence_res res;
7357 };
7358 
7359 static void nfs41_sequence_release(void *data)
7360 {
7361 	struct nfs4_sequence_data *calldata = data;
7362 	struct nfs_client *clp = calldata->clp;
7363 
7364 	if (atomic_read(&clp->cl_count) > 1)
7365 		nfs4_schedule_state_renewal(clp);
7366 	nfs_put_client(clp);
7367 	kfree(calldata);
7368 }
7369 
7370 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7371 {
7372 	switch(task->tk_status) {
7373 	case -NFS4ERR_DELAY:
7374 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7375 		return -EAGAIN;
7376 	default:
7377 		nfs4_schedule_lease_recovery(clp);
7378 	}
7379 	return 0;
7380 }
7381 
7382 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7383 {
7384 	struct nfs4_sequence_data *calldata = data;
7385 	struct nfs_client *clp = calldata->clp;
7386 
7387 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7388 		return;
7389 
7390 	trace_nfs4_sequence(clp, task->tk_status);
7391 	if (task->tk_status < 0) {
7392 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7393 		if (atomic_read(&clp->cl_count) == 1)
7394 			goto out;
7395 
7396 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7397 			rpc_restart_call_prepare(task);
7398 			return;
7399 		}
7400 	}
7401 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7402 out:
7403 	dprintk("<-- %s\n", __func__);
7404 }
7405 
7406 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7407 {
7408 	struct nfs4_sequence_data *calldata = data;
7409 	struct nfs_client *clp = calldata->clp;
7410 	struct nfs4_sequence_args *args;
7411 	struct nfs4_sequence_res *res;
7412 
7413 	args = task->tk_msg.rpc_argp;
7414 	res = task->tk_msg.rpc_resp;
7415 
7416 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7417 }
7418 
7419 static const struct rpc_call_ops nfs41_sequence_ops = {
7420 	.rpc_call_done = nfs41_sequence_call_done,
7421 	.rpc_call_prepare = nfs41_sequence_prepare,
7422 	.rpc_release = nfs41_sequence_release,
7423 };
7424 
7425 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7426 		struct rpc_cred *cred,
7427 		bool is_privileged)
7428 {
7429 	struct nfs4_sequence_data *calldata;
7430 	struct rpc_message msg = {
7431 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7432 		.rpc_cred = cred,
7433 	};
7434 	struct rpc_task_setup task_setup_data = {
7435 		.rpc_client = clp->cl_rpcclient,
7436 		.rpc_message = &msg,
7437 		.callback_ops = &nfs41_sequence_ops,
7438 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7439 	};
7440 
7441 	if (!atomic_inc_not_zero(&clp->cl_count))
7442 		return ERR_PTR(-EIO);
7443 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7444 	if (calldata == NULL) {
7445 		nfs_put_client(clp);
7446 		return ERR_PTR(-ENOMEM);
7447 	}
7448 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7449 	if (is_privileged)
7450 		nfs4_set_sequence_privileged(&calldata->args);
7451 	msg.rpc_argp = &calldata->args;
7452 	msg.rpc_resp = &calldata->res;
7453 	calldata->clp = clp;
7454 	task_setup_data.callback_data = calldata;
7455 
7456 	return rpc_run_task(&task_setup_data);
7457 }
7458 
7459 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7460 {
7461 	struct rpc_task *task;
7462 	int ret = 0;
7463 
7464 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7465 		return -EAGAIN;
7466 	task = _nfs41_proc_sequence(clp, cred, false);
7467 	if (IS_ERR(task))
7468 		ret = PTR_ERR(task);
7469 	else
7470 		rpc_put_task_async(task);
7471 	dprintk("<-- %s status=%d\n", __func__, ret);
7472 	return ret;
7473 }
7474 
7475 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7476 {
7477 	struct rpc_task *task;
7478 	int ret;
7479 
7480 	task = _nfs41_proc_sequence(clp, cred, true);
7481 	if (IS_ERR(task)) {
7482 		ret = PTR_ERR(task);
7483 		goto out;
7484 	}
7485 	ret = rpc_wait_for_completion_task(task);
7486 	if (!ret) {
7487 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7488 
7489 		if (task->tk_status == 0)
7490 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7491 		ret = task->tk_status;
7492 	}
7493 	rpc_put_task(task);
7494 out:
7495 	dprintk("<-- %s status=%d\n", __func__, ret);
7496 	return ret;
7497 }
7498 
7499 struct nfs4_reclaim_complete_data {
7500 	struct nfs_client *clp;
7501 	struct nfs41_reclaim_complete_args arg;
7502 	struct nfs41_reclaim_complete_res res;
7503 };
7504 
7505 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7506 {
7507 	struct nfs4_reclaim_complete_data *calldata = data;
7508 
7509 	nfs41_setup_sequence(calldata->clp->cl_session,
7510 			&calldata->arg.seq_args,
7511 			&calldata->res.seq_res,
7512 			task);
7513 }
7514 
7515 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7516 {
7517 	switch(task->tk_status) {
7518 	case 0:
7519 	case -NFS4ERR_COMPLETE_ALREADY:
7520 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7521 		break;
7522 	case -NFS4ERR_DELAY:
7523 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7524 		/* fall through */
7525 	case -NFS4ERR_RETRY_UNCACHED_REP:
7526 		return -EAGAIN;
7527 	default:
7528 		nfs4_schedule_lease_recovery(clp);
7529 	}
7530 	return 0;
7531 }
7532 
7533 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7534 {
7535 	struct nfs4_reclaim_complete_data *calldata = data;
7536 	struct nfs_client *clp = calldata->clp;
7537 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7538 
7539 	dprintk("--> %s\n", __func__);
7540 	if (!nfs41_sequence_done(task, res))
7541 		return;
7542 
7543 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7544 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7545 		rpc_restart_call_prepare(task);
7546 		return;
7547 	}
7548 	dprintk("<-- %s\n", __func__);
7549 }
7550 
7551 static void nfs4_free_reclaim_complete_data(void *data)
7552 {
7553 	struct nfs4_reclaim_complete_data *calldata = data;
7554 
7555 	kfree(calldata);
7556 }
7557 
7558 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7559 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7560 	.rpc_call_done = nfs4_reclaim_complete_done,
7561 	.rpc_release = nfs4_free_reclaim_complete_data,
7562 };
7563 
7564 /*
7565  * Issue a global reclaim complete.
7566  */
7567 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7568 		struct rpc_cred *cred)
7569 {
7570 	struct nfs4_reclaim_complete_data *calldata;
7571 	struct rpc_task *task;
7572 	struct rpc_message msg = {
7573 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7574 		.rpc_cred = cred,
7575 	};
7576 	struct rpc_task_setup task_setup_data = {
7577 		.rpc_client = clp->cl_rpcclient,
7578 		.rpc_message = &msg,
7579 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7580 		.flags = RPC_TASK_ASYNC,
7581 	};
7582 	int status = -ENOMEM;
7583 
7584 	dprintk("--> %s\n", __func__);
7585 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7586 	if (calldata == NULL)
7587 		goto out;
7588 	calldata->clp = clp;
7589 	calldata->arg.one_fs = 0;
7590 
7591 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7592 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7593 	msg.rpc_argp = &calldata->arg;
7594 	msg.rpc_resp = &calldata->res;
7595 	task_setup_data.callback_data = calldata;
7596 	task = rpc_run_task(&task_setup_data);
7597 	if (IS_ERR(task)) {
7598 		status = PTR_ERR(task);
7599 		goto out;
7600 	}
7601 	status = nfs4_wait_for_completion_rpc_task(task);
7602 	if (status == 0)
7603 		status = task->tk_status;
7604 	rpc_put_task(task);
7605 	return 0;
7606 out:
7607 	dprintk("<-- %s status=%d\n", __func__, status);
7608 	return status;
7609 }
7610 
7611 static void
7612 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7613 {
7614 	struct nfs4_layoutget *lgp = calldata;
7615 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7616 	struct nfs4_session *session = nfs4_get_session(server);
7617 
7618 	dprintk("--> %s\n", __func__);
7619 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7620 	 * right now covering the LAYOUTGET we are about to send.
7621 	 * However, that is not so catastrophic, and there seems
7622 	 * to be no way to prevent it completely.
7623 	 */
7624 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7625 				&lgp->res.seq_res, task))
7626 		return;
7627 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7628 					  NFS_I(lgp->args.inode)->layout,
7629 					  &lgp->args.range,
7630 					  lgp->args.ctx->state)) {
7631 		rpc_exit(task, NFS4_OK);
7632 	}
7633 }
7634 
7635 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7636 {
7637 	struct nfs4_layoutget *lgp = calldata;
7638 	struct inode *inode = lgp->args.inode;
7639 	struct nfs_server *server = NFS_SERVER(inode);
7640 	struct pnfs_layout_hdr *lo;
7641 	struct nfs4_state *state = NULL;
7642 	unsigned long timeo, now, giveup;
7643 
7644 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7645 
7646 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7647 		goto out;
7648 
7649 	switch (task->tk_status) {
7650 	case 0:
7651 		goto out;
7652 	/*
7653 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7654 	 * (or clients) writing to the same RAID stripe
7655 	 */
7656 	case -NFS4ERR_LAYOUTTRYLATER:
7657 	/*
7658 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7659 	 * existing layout before getting a new one).
7660 	 */
7661 	case -NFS4ERR_RECALLCONFLICT:
7662 		timeo = rpc_get_timeout(task->tk_client);
7663 		giveup = lgp->args.timestamp + timeo;
7664 		now = jiffies;
7665 		if (time_after(giveup, now)) {
7666 			unsigned long delay;
7667 
7668 			/* Delay for:
7669 			 * - Not less then NFS4_POLL_RETRY_MIN.
7670 			 * - One last time a jiffie before we give up
7671 			 * - exponential backoff (time_now minus start_attempt)
7672 			 */
7673 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7674 				    min((giveup - now - 1),
7675 					now - lgp->args.timestamp));
7676 
7677 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7678 				__func__, delay);
7679 			rpc_delay(task, delay);
7680 			task->tk_status = 0;
7681 			rpc_restart_call_prepare(task);
7682 			goto out; /* Do not call nfs4_async_handle_error() */
7683 		}
7684 		break;
7685 	case -NFS4ERR_EXPIRED:
7686 	case -NFS4ERR_BAD_STATEID:
7687 		spin_lock(&inode->i_lock);
7688 		lo = NFS_I(inode)->layout;
7689 		if (!lo || list_empty(&lo->plh_segs)) {
7690 			spin_unlock(&inode->i_lock);
7691 			/* If the open stateid was bad, then recover it. */
7692 			state = lgp->args.ctx->state;
7693 		} else {
7694 			LIST_HEAD(head);
7695 
7696 			/*
7697 			 * Mark the bad layout state as invalid, then retry
7698 			 * with the current stateid.
7699 			 */
7700 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7701 			spin_unlock(&inode->i_lock);
7702 			pnfs_free_lseg_list(&head);
7703 
7704 			task->tk_status = 0;
7705 			rpc_restart_call_prepare(task);
7706 		}
7707 	}
7708 	if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7709 		rpc_restart_call_prepare(task);
7710 out:
7711 	dprintk("<-- %s\n", __func__);
7712 }
7713 
7714 static size_t max_response_pages(struct nfs_server *server)
7715 {
7716 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7717 	return nfs_page_array_len(0, max_resp_sz);
7718 }
7719 
7720 static void nfs4_free_pages(struct page **pages, size_t size)
7721 {
7722 	int i;
7723 
7724 	if (!pages)
7725 		return;
7726 
7727 	for (i = 0; i < size; i++) {
7728 		if (!pages[i])
7729 			break;
7730 		__free_page(pages[i]);
7731 	}
7732 	kfree(pages);
7733 }
7734 
7735 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7736 {
7737 	struct page **pages;
7738 	int i;
7739 
7740 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7741 	if (!pages) {
7742 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7743 		return NULL;
7744 	}
7745 
7746 	for (i = 0; i < size; i++) {
7747 		pages[i] = alloc_page(gfp_flags);
7748 		if (!pages[i]) {
7749 			dprintk("%s: failed to allocate page\n", __func__);
7750 			nfs4_free_pages(pages, size);
7751 			return NULL;
7752 		}
7753 	}
7754 
7755 	return pages;
7756 }
7757 
7758 static void nfs4_layoutget_release(void *calldata)
7759 {
7760 	struct nfs4_layoutget *lgp = calldata;
7761 	struct inode *inode = lgp->args.inode;
7762 	struct nfs_server *server = NFS_SERVER(inode);
7763 	size_t max_pages = max_response_pages(server);
7764 
7765 	dprintk("--> %s\n", __func__);
7766 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7767 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7768 	put_nfs_open_context(lgp->args.ctx);
7769 	kfree(calldata);
7770 	dprintk("<-- %s\n", __func__);
7771 }
7772 
7773 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7774 	.rpc_call_prepare = nfs4_layoutget_prepare,
7775 	.rpc_call_done = nfs4_layoutget_done,
7776 	.rpc_release = nfs4_layoutget_release,
7777 };
7778 
7779 struct pnfs_layout_segment *
7780 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7781 {
7782 	struct inode *inode = lgp->args.inode;
7783 	struct nfs_server *server = NFS_SERVER(inode);
7784 	size_t max_pages = max_response_pages(server);
7785 	struct rpc_task *task;
7786 	struct rpc_message msg = {
7787 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7788 		.rpc_argp = &lgp->args,
7789 		.rpc_resp = &lgp->res,
7790 		.rpc_cred = lgp->cred,
7791 	};
7792 	struct rpc_task_setup task_setup_data = {
7793 		.rpc_client = server->client,
7794 		.rpc_message = &msg,
7795 		.callback_ops = &nfs4_layoutget_call_ops,
7796 		.callback_data = lgp,
7797 		.flags = RPC_TASK_ASYNC,
7798 	};
7799 	struct pnfs_layout_segment *lseg = NULL;
7800 	int status = 0;
7801 
7802 	dprintk("--> %s\n", __func__);
7803 
7804 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7805 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7806 
7807 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7808 	if (!lgp->args.layout.pages) {
7809 		nfs4_layoutget_release(lgp);
7810 		return ERR_PTR(-ENOMEM);
7811 	}
7812 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7813 	lgp->args.timestamp = jiffies;
7814 
7815 	lgp->res.layoutp = &lgp->args.layout;
7816 	lgp->res.seq_res.sr_slot = NULL;
7817 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7818 
7819 	task = rpc_run_task(&task_setup_data);
7820 	if (IS_ERR(task))
7821 		return ERR_CAST(task);
7822 	status = nfs4_wait_for_completion_rpc_task(task);
7823 	if (status == 0)
7824 		status = task->tk_status;
7825 	trace_nfs4_layoutget(lgp->args.ctx,
7826 			&lgp->args.range,
7827 			&lgp->res.range,
7828 			status);
7829 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7830 	if (status == 0 && lgp->res.layoutp->len)
7831 		lseg = pnfs_layout_process(lgp);
7832 	rpc_put_task(task);
7833 	dprintk("<-- %s status=%d\n", __func__, status);
7834 	if (status)
7835 		return ERR_PTR(status);
7836 	return lseg;
7837 }
7838 
7839 static void
7840 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7841 {
7842 	struct nfs4_layoutreturn *lrp = calldata;
7843 
7844 	dprintk("--> %s\n", __func__);
7845 	nfs41_setup_sequence(lrp->clp->cl_session,
7846 			&lrp->args.seq_args,
7847 			&lrp->res.seq_res,
7848 			task);
7849 }
7850 
7851 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7852 {
7853 	struct nfs4_layoutreturn *lrp = calldata;
7854 	struct nfs_server *server;
7855 
7856 	dprintk("--> %s\n", __func__);
7857 
7858 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7859 		return;
7860 
7861 	server = NFS_SERVER(lrp->args.inode);
7862 	switch (task->tk_status) {
7863 	default:
7864 		task->tk_status = 0;
7865 	case 0:
7866 		break;
7867 	case -NFS4ERR_DELAY:
7868 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7869 			break;
7870 		rpc_restart_call_prepare(task);
7871 		return;
7872 	}
7873 	dprintk("<-- %s\n", __func__);
7874 }
7875 
7876 static void nfs4_layoutreturn_release(void *calldata)
7877 {
7878 	struct nfs4_layoutreturn *lrp = calldata;
7879 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7880 
7881 	dprintk("--> %s\n", __func__);
7882 	spin_lock(&lo->plh_inode->i_lock);
7883 	if (lrp->res.lrs_present)
7884 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7885 	pnfs_clear_layoutreturn_waitbit(lo);
7886 	clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7887 	rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7888 	lo->plh_block_lgets--;
7889 	spin_unlock(&lo->plh_inode->i_lock);
7890 	pnfs_put_layout_hdr(lrp->args.layout);
7891 	nfs_iput_and_deactive(lrp->inode);
7892 	kfree(calldata);
7893 	dprintk("<-- %s\n", __func__);
7894 }
7895 
7896 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7897 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7898 	.rpc_call_done = nfs4_layoutreturn_done,
7899 	.rpc_release = nfs4_layoutreturn_release,
7900 };
7901 
7902 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7903 {
7904 	struct rpc_task *task;
7905 	struct rpc_message msg = {
7906 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7907 		.rpc_argp = &lrp->args,
7908 		.rpc_resp = &lrp->res,
7909 		.rpc_cred = lrp->cred,
7910 	};
7911 	struct rpc_task_setup task_setup_data = {
7912 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7913 		.rpc_message = &msg,
7914 		.callback_ops = &nfs4_layoutreturn_call_ops,
7915 		.callback_data = lrp,
7916 	};
7917 	int status = 0;
7918 
7919 	dprintk("--> %s\n", __func__);
7920 	if (!sync) {
7921 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
7922 		if (!lrp->inode) {
7923 			nfs4_layoutreturn_release(lrp);
7924 			return -EAGAIN;
7925 		}
7926 		task_setup_data.flags |= RPC_TASK_ASYNC;
7927 	}
7928 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7929 	task = rpc_run_task(&task_setup_data);
7930 	if (IS_ERR(task))
7931 		return PTR_ERR(task);
7932 	if (sync)
7933 		status = task->tk_status;
7934 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7935 	dprintk("<-- %s status=%d\n", __func__, status);
7936 	rpc_put_task(task);
7937 	return status;
7938 }
7939 
7940 static int
7941 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7942 		struct pnfs_device *pdev,
7943 		struct rpc_cred *cred)
7944 {
7945 	struct nfs4_getdeviceinfo_args args = {
7946 		.pdev = pdev,
7947 	};
7948 	struct nfs4_getdeviceinfo_res res = {
7949 		.pdev = pdev,
7950 	};
7951 	struct rpc_message msg = {
7952 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7953 		.rpc_argp = &args,
7954 		.rpc_resp = &res,
7955 		.rpc_cred = cred,
7956 	};
7957 	int status;
7958 
7959 	dprintk("--> %s\n", __func__);
7960 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7961 	dprintk("<-- %s status=%d\n", __func__, status);
7962 
7963 	return status;
7964 }
7965 
7966 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7967 		struct pnfs_device *pdev,
7968 		struct rpc_cred *cred)
7969 {
7970 	struct nfs4_exception exception = { };
7971 	int err;
7972 
7973 	do {
7974 		err = nfs4_handle_exception(server,
7975 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7976 					&exception);
7977 	} while (exception.retry);
7978 	return err;
7979 }
7980 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7981 
7982 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7983 {
7984 	struct nfs4_layoutcommit_data *data = calldata;
7985 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7986 	struct nfs4_session *session = nfs4_get_session(server);
7987 
7988 	nfs41_setup_sequence(session,
7989 			&data->args.seq_args,
7990 			&data->res.seq_res,
7991 			task);
7992 }
7993 
7994 static void
7995 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7996 {
7997 	struct nfs4_layoutcommit_data *data = calldata;
7998 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7999 
8000 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8001 		return;
8002 
8003 	switch (task->tk_status) { /* Just ignore these failures */
8004 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8005 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8006 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8007 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8008 		task->tk_status = 0;
8009 	case 0:
8010 		break;
8011 	default:
8012 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8013 			rpc_restart_call_prepare(task);
8014 			return;
8015 		}
8016 	}
8017 }
8018 
8019 static void nfs4_layoutcommit_release(void *calldata)
8020 {
8021 	struct nfs4_layoutcommit_data *data = calldata;
8022 
8023 	pnfs_cleanup_layoutcommit(data);
8024 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8025 					   data->res.fattr);
8026 	put_rpccred(data->cred);
8027 	nfs_iput_and_deactive(data->inode);
8028 	kfree(data);
8029 }
8030 
8031 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8032 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8033 	.rpc_call_done = nfs4_layoutcommit_done,
8034 	.rpc_release = nfs4_layoutcommit_release,
8035 };
8036 
8037 int
8038 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8039 {
8040 	struct rpc_message msg = {
8041 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8042 		.rpc_argp = &data->args,
8043 		.rpc_resp = &data->res,
8044 		.rpc_cred = data->cred,
8045 	};
8046 	struct rpc_task_setup task_setup_data = {
8047 		.task = &data->task,
8048 		.rpc_client = NFS_CLIENT(data->args.inode),
8049 		.rpc_message = &msg,
8050 		.callback_ops = &nfs4_layoutcommit_ops,
8051 		.callback_data = data,
8052 	};
8053 	struct rpc_task *task;
8054 	int status = 0;
8055 
8056 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
8057 		"lbw: %llu inode %lu\n",
8058 		data->task.tk_pid, sync,
8059 		data->args.lastbytewritten,
8060 		data->args.inode->i_ino);
8061 
8062 	if (!sync) {
8063 		data->inode = nfs_igrab_and_active(data->args.inode);
8064 		if (data->inode == NULL) {
8065 			nfs4_layoutcommit_release(data);
8066 			return -EAGAIN;
8067 		}
8068 		task_setup_data.flags = RPC_TASK_ASYNC;
8069 	}
8070 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8071 	task = rpc_run_task(&task_setup_data);
8072 	if (IS_ERR(task))
8073 		return PTR_ERR(task);
8074 	if (sync)
8075 		status = task->tk_status;
8076 	trace_nfs4_layoutcommit(data->args.inode, status);
8077 	dprintk("%s: status %d\n", __func__, status);
8078 	rpc_put_task(task);
8079 	return status;
8080 }
8081 
8082 /**
8083  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8084  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8085  */
8086 static int
8087 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8088 		    struct nfs_fsinfo *info,
8089 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8090 {
8091 	struct nfs41_secinfo_no_name_args args = {
8092 		.style = SECINFO_STYLE_CURRENT_FH,
8093 	};
8094 	struct nfs4_secinfo_res res = {
8095 		.flavors = flavors,
8096 	};
8097 	struct rpc_message msg = {
8098 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8099 		.rpc_argp = &args,
8100 		.rpc_resp = &res,
8101 	};
8102 	struct rpc_clnt *clnt = server->client;
8103 	struct rpc_cred *cred = NULL;
8104 	int status;
8105 
8106 	if (use_integrity) {
8107 		clnt = server->nfs_client->cl_rpcclient;
8108 		cred = nfs4_get_clid_cred(server->nfs_client);
8109 		msg.rpc_cred = cred;
8110 	}
8111 
8112 	dprintk("--> %s\n", __func__);
8113 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8114 				&res.seq_res, 0);
8115 	dprintk("<-- %s status=%d\n", __func__, status);
8116 
8117 	if (cred)
8118 		put_rpccred(cred);
8119 
8120 	return status;
8121 }
8122 
8123 static int
8124 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8125 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8126 {
8127 	struct nfs4_exception exception = { };
8128 	int err;
8129 	do {
8130 		/* first try using integrity protection */
8131 		err = -NFS4ERR_WRONGSEC;
8132 
8133 		/* try to use integrity protection with machine cred */
8134 		if (_nfs4_is_integrity_protected(server->nfs_client))
8135 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8136 							  flavors, true);
8137 
8138 		/*
8139 		 * if unable to use integrity protection, or SECINFO with
8140 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8141 		 * disallowed by spec, but exists in deployed servers) use
8142 		 * the current filesystem's rpc_client and the user cred.
8143 		 */
8144 		if (err == -NFS4ERR_WRONGSEC)
8145 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8146 							  flavors, false);
8147 
8148 		switch (err) {
8149 		case 0:
8150 		case -NFS4ERR_WRONGSEC:
8151 		case -ENOTSUPP:
8152 			goto out;
8153 		default:
8154 			err = nfs4_handle_exception(server, err, &exception);
8155 		}
8156 	} while (exception.retry);
8157 out:
8158 	return err;
8159 }
8160 
8161 static int
8162 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8163 		    struct nfs_fsinfo *info)
8164 {
8165 	int err;
8166 	struct page *page;
8167 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8168 	struct nfs4_secinfo_flavors *flavors;
8169 	struct nfs4_secinfo4 *secinfo;
8170 	int i;
8171 
8172 	page = alloc_page(GFP_KERNEL);
8173 	if (!page) {
8174 		err = -ENOMEM;
8175 		goto out;
8176 	}
8177 
8178 	flavors = page_address(page);
8179 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8180 
8181 	/*
8182 	 * Fall back on "guess and check" method if
8183 	 * the server doesn't support SECINFO_NO_NAME
8184 	 */
8185 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8186 		err = nfs4_find_root_sec(server, fhandle, info);
8187 		goto out_freepage;
8188 	}
8189 	if (err)
8190 		goto out_freepage;
8191 
8192 	for (i = 0; i < flavors->num_flavors; i++) {
8193 		secinfo = &flavors->flavors[i];
8194 
8195 		switch (secinfo->flavor) {
8196 		case RPC_AUTH_NULL:
8197 		case RPC_AUTH_UNIX:
8198 		case RPC_AUTH_GSS:
8199 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8200 					&secinfo->flavor_info);
8201 			break;
8202 		default:
8203 			flavor = RPC_AUTH_MAXFLAVOR;
8204 			break;
8205 		}
8206 
8207 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8208 			flavor = RPC_AUTH_MAXFLAVOR;
8209 
8210 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8211 			err = nfs4_lookup_root_sec(server, fhandle,
8212 						   info, flavor);
8213 			if (!err)
8214 				break;
8215 		}
8216 	}
8217 
8218 	if (flavor == RPC_AUTH_MAXFLAVOR)
8219 		err = -EPERM;
8220 
8221 out_freepage:
8222 	put_page(page);
8223 	if (err == -EACCES)
8224 		return -EPERM;
8225 out:
8226 	return err;
8227 }
8228 
8229 static int _nfs41_test_stateid(struct nfs_server *server,
8230 		nfs4_stateid *stateid,
8231 		struct rpc_cred *cred)
8232 {
8233 	int status;
8234 	struct nfs41_test_stateid_args args = {
8235 		.stateid = stateid,
8236 	};
8237 	struct nfs41_test_stateid_res res;
8238 	struct rpc_message msg = {
8239 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8240 		.rpc_argp = &args,
8241 		.rpc_resp = &res,
8242 		.rpc_cred = cred,
8243 	};
8244 	struct rpc_clnt *rpc_client = server->client;
8245 
8246 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8247 		&rpc_client, &msg);
8248 
8249 	dprintk("NFS call  test_stateid %p\n", stateid);
8250 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8251 	nfs4_set_sequence_privileged(&args.seq_args);
8252 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8253 			&args.seq_args, &res.seq_res);
8254 	if (status != NFS_OK) {
8255 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8256 		return status;
8257 	}
8258 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8259 	return -res.status;
8260 }
8261 
8262 /**
8263  * nfs41_test_stateid - perform a TEST_STATEID operation
8264  *
8265  * @server: server / transport on which to perform the operation
8266  * @stateid: state ID to test
8267  * @cred: credential
8268  *
8269  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8270  * Otherwise a negative NFS4ERR value is returned if the operation
8271  * failed or the state ID is not currently valid.
8272  */
8273 static int nfs41_test_stateid(struct nfs_server *server,
8274 		nfs4_stateid *stateid,
8275 		struct rpc_cred *cred)
8276 {
8277 	struct nfs4_exception exception = { };
8278 	int err;
8279 	do {
8280 		err = _nfs41_test_stateid(server, stateid, cred);
8281 		if (err != -NFS4ERR_DELAY)
8282 			break;
8283 		nfs4_handle_exception(server, err, &exception);
8284 	} while (exception.retry);
8285 	return err;
8286 }
8287 
8288 struct nfs_free_stateid_data {
8289 	struct nfs_server *server;
8290 	struct nfs41_free_stateid_args args;
8291 	struct nfs41_free_stateid_res res;
8292 };
8293 
8294 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8295 {
8296 	struct nfs_free_stateid_data *data = calldata;
8297 	nfs41_setup_sequence(nfs4_get_session(data->server),
8298 			&data->args.seq_args,
8299 			&data->res.seq_res,
8300 			task);
8301 }
8302 
8303 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8304 {
8305 	struct nfs_free_stateid_data *data = calldata;
8306 
8307 	nfs41_sequence_done(task, &data->res.seq_res);
8308 
8309 	switch (task->tk_status) {
8310 	case -NFS4ERR_DELAY:
8311 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8312 			rpc_restart_call_prepare(task);
8313 	}
8314 }
8315 
8316 static void nfs41_free_stateid_release(void *calldata)
8317 {
8318 	kfree(calldata);
8319 }
8320 
8321 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8322 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8323 	.rpc_call_done = nfs41_free_stateid_done,
8324 	.rpc_release = nfs41_free_stateid_release,
8325 };
8326 
8327 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8328 		nfs4_stateid *stateid,
8329 		struct rpc_cred *cred,
8330 		bool privileged)
8331 {
8332 	struct rpc_message msg = {
8333 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8334 		.rpc_cred = cred,
8335 	};
8336 	struct rpc_task_setup task_setup = {
8337 		.rpc_client = server->client,
8338 		.rpc_message = &msg,
8339 		.callback_ops = &nfs41_free_stateid_ops,
8340 		.flags = RPC_TASK_ASYNC,
8341 	};
8342 	struct nfs_free_stateid_data *data;
8343 
8344 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8345 		&task_setup.rpc_client, &msg);
8346 
8347 	dprintk("NFS call  free_stateid %p\n", stateid);
8348 	data = kmalloc(sizeof(*data), GFP_NOFS);
8349 	if (!data)
8350 		return ERR_PTR(-ENOMEM);
8351 	data->server = server;
8352 	nfs4_stateid_copy(&data->args.stateid, stateid);
8353 
8354 	task_setup.callback_data = data;
8355 
8356 	msg.rpc_argp = &data->args;
8357 	msg.rpc_resp = &data->res;
8358 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8359 	if (privileged)
8360 		nfs4_set_sequence_privileged(&data->args.seq_args);
8361 
8362 	return rpc_run_task(&task_setup);
8363 }
8364 
8365 /**
8366  * nfs41_free_stateid - perform a FREE_STATEID operation
8367  *
8368  * @server: server / transport on which to perform the operation
8369  * @stateid: state ID to release
8370  * @cred: credential
8371  *
8372  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8373  * negative NFS4ERR value is returned.
8374  */
8375 static int nfs41_free_stateid(struct nfs_server *server,
8376 		nfs4_stateid *stateid,
8377 		struct rpc_cred *cred)
8378 {
8379 	struct rpc_task *task;
8380 	int ret;
8381 
8382 	task = _nfs41_free_stateid(server, stateid, cred, true);
8383 	if (IS_ERR(task))
8384 		return PTR_ERR(task);
8385 	ret = rpc_wait_for_completion_task(task);
8386 	if (!ret)
8387 		ret = task->tk_status;
8388 	rpc_put_task(task);
8389 	return ret;
8390 }
8391 
8392 static void
8393 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8394 {
8395 	struct rpc_task *task;
8396 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8397 
8398 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8399 	nfs4_free_lock_state(server, lsp);
8400 	if (IS_ERR(task))
8401 		return;
8402 	rpc_put_task(task);
8403 }
8404 
8405 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8406 		const nfs4_stateid *s2)
8407 {
8408 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8409 		return false;
8410 
8411 	if (s1->seqid == s2->seqid)
8412 		return true;
8413 	if (s1->seqid == 0 || s2->seqid == 0)
8414 		return true;
8415 
8416 	return false;
8417 }
8418 
8419 #endif /* CONFIG_NFS_V4_1 */
8420 
8421 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8422 		const nfs4_stateid *s2)
8423 {
8424 	return nfs4_stateid_match(s1, s2);
8425 }
8426 
8427 
8428 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8429 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8430 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8431 	.recover_open	= nfs4_open_reclaim,
8432 	.recover_lock	= nfs4_lock_reclaim,
8433 	.establish_clid = nfs4_init_clientid,
8434 	.detect_trunking = nfs40_discover_server_trunking,
8435 };
8436 
8437 #if defined(CONFIG_NFS_V4_1)
8438 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8439 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8440 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8441 	.recover_open	= nfs4_open_reclaim,
8442 	.recover_lock	= nfs4_lock_reclaim,
8443 	.establish_clid = nfs41_init_clientid,
8444 	.reclaim_complete = nfs41_proc_reclaim_complete,
8445 	.detect_trunking = nfs41_discover_server_trunking,
8446 };
8447 #endif /* CONFIG_NFS_V4_1 */
8448 
8449 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8450 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8451 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8452 	.recover_open	= nfs40_open_expired,
8453 	.recover_lock	= nfs4_lock_expired,
8454 	.establish_clid = nfs4_init_clientid,
8455 };
8456 
8457 #if defined(CONFIG_NFS_V4_1)
8458 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8459 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8460 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8461 	.recover_open	= nfs41_open_expired,
8462 	.recover_lock	= nfs41_lock_expired,
8463 	.establish_clid = nfs41_init_clientid,
8464 };
8465 #endif /* CONFIG_NFS_V4_1 */
8466 
8467 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8468 	.sched_state_renewal = nfs4_proc_async_renew,
8469 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8470 	.renew_lease = nfs4_proc_renew,
8471 };
8472 
8473 #if defined(CONFIG_NFS_V4_1)
8474 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8475 	.sched_state_renewal = nfs41_proc_async_sequence,
8476 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8477 	.renew_lease = nfs4_proc_sequence,
8478 };
8479 #endif
8480 
8481 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8482 	.get_locations = _nfs40_proc_get_locations,
8483 	.fsid_present = _nfs40_proc_fsid_present,
8484 };
8485 
8486 #if defined(CONFIG_NFS_V4_1)
8487 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8488 	.get_locations = _nfs41_proc_get_locations,
8489 	.fsid_present = _nfs41_proc_fsid_present,
8490 };
8491 #endif	/* CONFIG_NFS_V4_1 */
8492 
8493 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8494 	.minor_version = 0,
8495 	.init_caps = NFS_CAP_READDIRPLUS
8496 		| NFS_CAP_ATOMIC_OPEN
8497 		| NFS_CAP_CHANGE_ATTR
8498 		| NFS_CAP_POSIX_LOCK,
8499 	.init_client = nfs40_init_client,
8500 	.shutdown_client = nfs40_shutdown_client,
8501 	.match_stateid = nfs4_match_stateid,
8502 	.find_root_sec = nfs4_find_root_sec,
8503 	.free_lock_state = nfs4_release_lockowner,
8504 	.alloc_seqid = nfs_alloc_seqid,
8505 	.call_sync_ops = &nfs40_call_sync_ops,
8506 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8507 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8508 	.state_renewal_ops = &nfs40_state_renewal_ops,
8509 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8510 };
8511 
8512 #if defined(CONFIG_NFS_V4_1)
8513 static struct nfs_seqid *
8514 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8515 {
8516 	return NULL;
8517 }
8518 
8519 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8520 	.minor_version = 1,
8521 	.init_caps = NFS_CAP_READDIRPLUS
8522 		| NFS_CAP_ATOMIC_OPEN
8523 		| NFS_CAP_CHANGE_ATTR
8524 		| NFS_CAP_POSIX_LOCK
8525 		| NFS_CAP_STATEID_NFSV41
8526 		| NFS_CAP_ATOMIC_OPEN_V1,
8527 	.init_client = nfs41_init_client,
8528 	.shutdown_client = nfs41_shutdown_client,
8529 	.match_stateid = nfs41_match_stateid,
8530 	.find_root_sec = nfs41_find_root_sec,
8531 	.free_lock_state = nfs41_free_lock_state,
8532 	.alloc_seqid = nfs_alloc_no_seqid,
8533 	.call_sync_ops = &nfs41_call_sync_ops,
8534 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8535 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8536 	.state_renewal_ops = &nfs41_state_renewal_ops,
8537 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8538 };
8539 #endif
8540 
8541 #if defined(CONFIG_NFS_V4_2)
8542 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8543 	.minor_version = 2,
8544 	.init_caps = NFS_CAP_READDIRPLUS
8545 		| NFS_CAP_ATOMIC_OPEN
8546 		| NFS_CAP_CHANGE_ATTR
8547 		| NFS_CAP_POSIX_LOCK
8548 		| NFS_CAP_STATEID_NFSV41
8549 		| NFS_CAP_ATOMIC_OPEN_V1
8550 		| NFS_CAP_ALLOCATE
8551 		| NFS_CAP_DEALLOCATE
8552 		| NFS_CAP_SEEK,
8553 	.init_client = nfs41_init_client,
8554 	.shutdown_client = nfs41_shutdown_client,
8555 	.match_stateid = nfs41_match_stateid,
8556 	.find_root_sec = nfs41_find_root_sec,
8557 	.free_lock_state = nfs41_free_lock_state,
8558 	.call_sync_ops = &nfs41_call_sync_ops,
8559 	.alloc_seqid = nfs_alloc_no_seqid,
8560 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8561 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8562 	.state_renewal_ops = &nfs41_state_renewal_ops,
8563 };
8564 #endif
8565 
8566 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8567 	[0] = &nfs_v4_0_minor_ops,
8568 #if defined(CONFIG_NFS_V4_1)
8569 	[1] = &nfs_v4_1_minor_ops,
8570 #endif
8571 #if defined(CONFIG_NFS_V4_2)
8572 	[2] = &nfs_v4_2_minor_ops,
8573 #endif
8574 };
8575 
8576 static const struct inode_operations nfs4_dir_inode_operations = {
8577 	.create		= nfs_create,
8578 	.lookup		= nfs_lookup,
8579 	.atomic_open	= nfs_atomic_open,
8580 	.link		= nfs_link,
8581 	.unlink		= nfs_unlink,
8582 	.symlink	= nfs_symlink,
8583 	.mkdir		= nfs_mkdir,
8584 	.rmdir		= nfs_rmdir,
8585 	.mknod		= nfs_mknod,
8586 	.rename		= nfs_rename,
8587 	.permission	= nfs_permission,
8588 	.getattr	= nfs_getattr,
8589 	.setattr	= nfs_setattr,
8590 	.getxattr	= generic_getxattr,
8591 	.setxattr	= generic_setxattr,
8592 	.listxattr	= generic_listxattr,
8593 	.removexattr	= generic_removexattr,
8594 };
8595 
8596 static const struct inode_operations nfs4_file_inode_operations = {
8597 	.permission	= nfs_permission,
8598 	.getattr	= nfs_getattr,
8599 	.setattr	= nfs_setattr,
8600 	.getxattr	= generic_getxattr,
8601 	.setxattr	= generic_setxattr,
8602 	.listxattr	= generic_listxattr,
8603 	.removexattr	= generic_removexattr,
8604 };
8605 
8606 const struct nfs_rpc_ops nfs_v4_clientops = {
8607 	.version	= 4,			/* protocol version */
8608 	.dentry_ops	= &nfs4_dentry_operations,
8609 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8610 	.file_inode_ops	= &nfs4_file_inode_operations,
8611 	.file_ops	= &nfs4_file_operations,
8612 	.getroot	= nfs4_proc_get_root,
8613 	.submount	= nfs4_submount,
8614 	.try_mount	= nfs4_try_mount,
8615 	.getattr	= nfs4_proc_getattr,
8616 	.setattr	= nfs4_proc_setattr,
8617 	.lookup		= nfs4_proc_lookup,
8618 	.access		= nfs4_proc_access,
8619 	.readlink	= nfs4_proc_readlink,
8620 	.create		= nfs4_proc_create,
8621 	.remove		= nfs4_proc_remove,
8622 	.unlink_setup	= nfs4_proc_unlink_setup,
8623 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8624 	.unlink_done	= nfs4_proc_unlink_done,
8625 	.rename_setup	= nfs4_proc_rename_setup,
8626 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8627 	.rename_done	= nfs4_proc_rename_done,
8628 	.link		= nfs4_proc_link,
8629 	.symlink	= nfs4_proc_symlink,
8630 	.mkdir		= nfs4_proc_mkdir,
8631 	.rmdir		= nfs4_proc_remove,
8632 	.readdir	= nfs4_proc_readdir,
8633 	.mknod		= nfs4_proc_mknod,
8634 	.statfs		= nfs4_proc_statfs,
8635 	.fsinfo		= nfs4_proc_fsinfo,
8636 	.pathconf	= nfs4_proc_pathconf,
8637 	.set_capabilities = nfs4_server_capabilities,
8638 	.decode_dirent	= nfs4_decode_dirent,
8639 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8640 	.read_setup	= nfs4_proc_read_setup,
8641 	.read_done	= nfs4_read_done,
8642 	.write_setup	= nfs4_proc_write_setup,
8643 	.write_done	= nfs4_write_done,
8644 	.commit_setup	= nfs4_proc_commit_setup,
8645 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8646 	.commit_done	= nfs4_commit_done,
8647 	.lock		= nfs4_proc_lock,
8648 	.clear_acl_cache = nfs4_zap_acl_attr,
8649 	.close_context  = nfs4_close_context,
8650 	.open_context	= nfs4_atomic_open,
8651 	.have_delegation = nfs4_have_delegation,
8652 	.return_delegation = nfs4_inode_return_delegation,
8653 	.alloc_client	= nfs4_alloc_client,
8654 	.init_client	= nfs4_init_client,
8655 	.free_client	= nfs4_free_client,
8656 	.create_server	= nfs4_create_server,
8657 	.clone_server	= nfs_clone_server,
8658 };
8659 
8660 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8661 	.prefix	= XATTR_NAME_NFSV4_ACL,
8662 	.list	= nfs4_xattr_list_nfs4_acl,
8663 	.get	= nfs4_xattr_get_nfs4_acl,
8664 	.set	= nfs4_xattr_set_nfs4_acl,
8665 };
8666 
8667 const struct xattr_handler *nfs4_xattr_handlers[] = {
8668 	&nfs4_xattr_nfs4_acl_handler,
8669 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8670 	&nfs4_xattr_nfs4_label_handler,
8671 #endif
8672 	NULL
8673 };
8674 
8675 /*
8676  * Local variables:
8677  *  c-basic-offset: 8
8678  * End:
8679  */
8680