xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 63d63cbf)
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/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.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 "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69 
70 #include "nfs4trace.h"
71 
72 #define NFSDBG_FACILITY		NFSDBG_PROC
73 
74 #define NFS4_POLL_RETRY_MIN	(HZ/10)
75 #define NFS4_POLL_RETRY_MAX	(15*HZ)
76 
77 /* file attributes which can be mapped to nfs attributes */
78 #define NFS4_VALID_ATTRS (ATTR_MODE \
79 	| ATTR_UID \
80 	| ATTR_GID \
81 	| ATTR_SIZE \
82 	| ATTR_ATIME \
83 	| ATTR_MTIME \
84 	| ATTR_CTIME \
85 	| ATTR_ATIME_SET \
86 	| ATTR_MTIME_SET)
87 
88 struct nfs4_opendata;
89 static int _nfs4_proc_open(struct nfs4_opendata *data);
90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
96 			    struct nfs_fattr *fattr, struct iattr *sattr,
97 			    struct nfs4_state *state, struct nfs4_label *ilabel,
98 			    struct nfs4_label *olabel);
99 #ifdef CONFIG_NFS_V4_1
100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
101 		struct rpc_cred *);
102 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
103 		struct rpc_cred *);
104 #endif
105 
106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
107 static inline struct nfs4_label *
108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
109 	struct iattr *sattr, struct nfs4_label *label)
110 {
111 	int err;
112 
113 	if (label == NULL)
114 		return NULL;
115 
116 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
117 		return NULL;
118 
119 	err = security_dentry_init_security(dentry, sattr->ia_mode,
120 				&dentry->d_name, (void **)&label->label, &label->len);
121 	if (err == 0)
122 		return label;
123 
124 	return NULL;
125 }
126 static inline void
127 nfs4_label_release_security(struct nfs4_label *label)
128 {
129 	if (label)
130 		security_release_secctx(label->label, label->len);
131 }
132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
133 {
134 	if (label)
135 		return server->attr_bitmask;
136 
137 	return server->attr_bitmask_nl;
138 }
139 #else
140 static inline struct nfs4_label *
141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
142 	struct iattr *sattr, struct nfs4_label *l)
143 { return NULL; }
144 static inline void
145 nfs4_label_release_security(struct nfs4_label *label)
146 { return; }
147 static inline u32 *
148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
149 { return server->attr_bitmask; }
150 #endif
151 
152 /* Prevent leaks of NFSv4 errors into userland */
153 static int nfs4_map_errors(int err)
154 {
155 	if (err >= -1000)
156 		return err;
157 	switch (err) {
158 	case -NFS4ERR_RESOURCE:
159 	case -NFS4ERR_LAYOUTTRYLATER:
160 	case -NFS4ERR_RECALLCONFLICT:
161 		return -EREMOTEIO;
162 	case -NFS4ERR_WRONGSEC:
163 	case -NFS4ERR_WRONG_CRED:
164 		return -EPERM;
165 	case -NFS4ERR_BADOWNER:
166 	case -NFS4ERR_BADNAME:
167 		return -EINVAL;
168 	case -NFS4ERR_SHARE_DENIED:
169 		return -EACCES;
170 	case -NFS4ERR_MINOR_VERS_MISMATCH:
171 		return -EPROTONOSUPPORT;
172 	case -NFS4ERR_FILE_OPEN:
173 		return -EBUSY;
174 	default:
175 		dprintk("%s could not handle NFSv4 error %d\n",
176 				__func__, -err);
177 		break;
178 	}
179 	return -EIO;
180 }
181 
182 /*
183  * This is our standard bitmap for GETATTR requests.
184  */
185 const u32 nfs4_fattr_bitmap[3] = {
186 	FATTR4_WORD0_TYPE
187 	| FATTR4_WORD0_CHANGE
188 	| FATTR4_WORD0_SIZE
189 	| FATTR4_WORD0_FSID
190 	| FATTR4_WORD0_FILEID,
191 	FATTR4_WORD1_MODE
192 	| FATTR4_WORD1_NUMLINKS
193 	| FATTR4_WORD1_OWNER
194 	| FATTR4_WORD1_OWNER_GROUP
195 	| FATTR4_WORD1_RAWDEV
196 	| FATTR4_WORD1_SPACE_USED
197 	| FATTR4_WORD1_TIME_ACCESS
198 	| FATTR4_WORD1_TIME_METADATA
199 	| FATTR4_WORD1_TIME_MODIFY
200 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
202 	FATTR4_WORD2_SECURITY_LABEL
203 #endif
204 };
205 
206 static const u32 nfs4_pnfs_open_bitmap[3] = {
207 	FATTR4_WORD0_TYPE
208 	| FATTR4_WORD0_CHANGE
209 	| FATTR4_WORD0_SIZE
210 	| FATTR4_WORD0_FSID
211 	| FATTR4_WORD0_FILEID,
212 	FATTR4_WORD1_MODE
213 	| FATTR4_WORD1_NUMLINKS
214 	| FATTR4_WORD1_OWNER
215 	| FATTR4_WORD1_OWNER_GROUP
216 	| FATTR4_WORD1_RAWDEV
217 	| FATTR4_WORD1_SPACE_USED
218 	| FATTR4_WORD1_TIME_ACCESS
219 	| FATTR4_WORD1_TIME_METADATA
220 	| FATTR4_WORD1_TIME_MODIFY,
221 	FATTR4_WORD2_MDSTHRESHOLD
222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
223 	| FATTR4_WORD2_SECURITY_LABEL
224 #endif
225 };
226 
227 static const u32 nfs4_open_noattr_bitmap[3] = {
228 	FATTR4_WORD0_TYPE
229 	| FATTR4_WORD0_CHANGE
230 	| FATTR4_WORD0_FILEID,
231 };
232 
233 const u32 nfs4_statfs_bitmap[3] = {
234 	FATTR4_WORD0_FILES_AVAIL
235 	| FATTR4_WORD0_FILES_FREE
236 	| FATTR4_WORD0_FILES_TOTAL,
237 	FATTR4_WORD1_SPACE_AVAIL
238 	| FATTR4_WORD1_SPACE_FREE
239 	| FATTR4_WORD1_SPACE_TOTAL
240 };
241 
242 const u32 nfs4_pathconf_bitmap[3] = {
243 	FATTR4_WORD0_MAXLINK
244 	| FATTR4_WORD0_MAXNAME,
245 	0
246 };
247 
248 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
249 			| FATTR4_WORD0_MAXREAD
250 			| FATTR4_WORD0_MAXWRITE
251 			| FATTR4_WORD0_LEASE_TIME,
252 			FATTR4_WORD1_TIME_DELTA
253 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
254 			FATTR4_WORD2_LAYOUT_BLKSIZE
255 			| FATTR4_WORD2_CLONE_BLKSIZE
256 };
257 
258 const u32 nfs4_fs_locations_bitmap[3] = {
259 	FATTR4_WORD0_TYPE
260 	| FATTR4_WORD0_CHANGE
261 	| FATTR4_WORD0_SIZE
262 	| FATTR4_WORD0_FSID
263 	| FATTR4_WORD0_FILEID
264 	| FATTR4_WORD0_FS_LOCATIONS,
265 	FATTR4_WORD1_MODE
266 	| FATTR4_WORD1_NUMLINKS
267 	| FATTR4_WORD1_OWNER
268 	| FATTR4_WORD1_OWNER_GROUP
269 	| FATTR4_WORD1_RAWDEV
270 	| FATTR4_WORD1_SPACE_USED
271 	| FATTR4_WORD1_TIME_ACCESS
272 	| FATTR4_WORD1_TIME_METADATA
273 	| FATTR4_WORD1_TIME_MODIFY
274 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
275 };
276 
277 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
278 		struct nfs4_readdir_arg *readdir)
279 {
280 	__be32 *start, *p;
281 
282 	if (cookie > 2) {
283 		readdir->cookie = cookie;
284 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
285 		return;
286 	}
287 
288 	readdir->cookie = 0;
289 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
290 	if (cookie == 2)
291 		return;
292 
293 	/*
294 	 * NFSv4 servers do not return entries for '.' and '..'
295 	 * Therefore, we fake these entries here.  We let '.'
296 	 * have cookie 0 and '..' have cookie 1.  Note that
297 	 * when talking to the server, we always send cookie 0
298 	 * instead of 1 or 2.
299 	 */
300 	start = p = kmap_atomic(*readdir->pages);
301 
302 	if (cookie == 0) {
303 		*p++ = xdr_one;                                  /* next */
304 		*p++ = xdr_zero;                   /* cookie, first word */
305 		*p++ = xdr_one;                   /* cookie, second word */
306 		*p++ = xdr_one;                             /* entry len */
307 		memcpy(p, ".\0\0\0", 4);                        /* entry */
308 		p++;
309 		*p++ = xdr_one;                         /* bitmap length */
310 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
311 		*p++ = htonl(8);              /* attribute buffer length */
312 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
313 	}
314 
315 	*p++ = xdr_one;                                  /* next */
316 	*p++ = xdr_zero;                   /* cookie, first word */
317 	*p++ = xdr_two;                   /* cookie, second word */
318 	*p++ = xdr_two;                             /* entry len */
319 	memcpy(p, "..\0\0", 4);                         /* entry */
320 	p++;
321 	*p++ = xdr_one;                         /* bitmap length */
322 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
323 	*p++ = htonl(8);              /* attribute buffer length */
324 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
325 
326 	readdir->pgbase = (char *)p - (char *)start;
327 	readdir->count -= readdir->pgbase;
328 	kunmap_atomic(start);
329 }
330 
331 static long nfs4_update_delay(long *timeout)
332 {
333 	long ret;
334 	if (!timeout)
335 		return NFS4_POLL_RETRY_MAX;
336 	if (*timeout <= 0)
337 		*timeout = NFS4_POLL_RETRY_MIN;
338 	if (*timeout > NFS4_POLL_RETRY_MAX)
339 		*timeout = NFS4_POLL_RETRY_MAX;
340 	ret = *timeout;
341 	*timeout <<= 1;
342 	return ret;
343 }
344 
345 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
346 {
347 	int res = 0;
348 
349 	might_sleep();
350 
351 	freezable_schedule_timeout_killable_unsafe(
352 		nfs4_update_delay(timeout));
353 	if (fatal_signal_pending(current))
354 		res = -ERESTARTSYS;
355 	return res;
356 }
357 
358 /* This is the error handling routine for processes that are allowed
359  * to sleep.
360  */
361 static int nfs4_do_handle_exception(struct nfs_server *server,
362 		int errorcode, struct nfs4_exception *exception)
363 {
364 	struct nfs_client *clp = server->nfs_client;
365 	struct nfs4_state *state = exception->state;
366 	const nfs4_stateid *stateid = exception->stateid;
367 	struct inode *inode = exception->inode;
368 	int ret = errorcode;
369 
370 	exception->delay = 0;
371 	exception->recovering = 0;
372 	exception->retry = 0;
373 	switch(errorcode) {
374 		case 0:
375 			return 0;
376 		case -NFS4ERR_OPENMODE:
377 		case -NFS4ERR_DELEG_REVOKED:
378 		case -NFS4ERR_ADMIN_REVOKED:
379 		case -NFS4ERR_BAD_STATEID:
380 			if (inode) {
381 				int err;
382 
383 				err = nfs_async_inode_return_delegation(inode,
384 						stateid);
385 				if (err == 0)
386 					goto wait_on_recovery;
387 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
388 					exception->retry = 1;
389 					break;
390 				}
391 			}
392 			if (state == NULL)
393 				break;
394 			ret = nfs4_schedule_stateid_recovery(server, state);
395 			if (ret < 0)
396 				break;
397 			goto wait_on_recovery;
398 		case -NFS4ERR_EXPIRED:
399 			if (state != NULL) {
400 				ret = nfs4_schedule_stateid_recovery(server, state);
401 				if (ret < 0)
402 					break;
403 			}
404 		case -NFS4ERR_STALE_STATEID:
405 		case -NFS4ERR_STALE_CLIENTID:
406 			nfs4_schedule_lease_recovery(clp);
407 			goto wait_on_recovery;
408 		case -NFS4ERR_MOVED:
409 			ret = nfs4_schedule_migration_recovery(server);
410 			if (ret < 0)
411 				break;
412 			goto wait_on_recovery;
413 		case -NFS4ERR_LEASE_MOVED:
414 			nfs4_schedule_lease_moved_recovery(clp);
415 			goto wait_on_recovery;
416 #if defined(CONFIG_NFS_V4_1)
417 		case -NFS4ERR_BADSESSION:
418 		case -NFS4ERR_BADSLOT:
419 		case -NFS4ERR_BAD_HIGH_SLOT:
420 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
421 		case -NFS4ERR_DEADSESSION:
422 		case -NFS4ERR_SEQ_FALSE_RETRY:
423 		case -NFS4ERR_SEQ_MISORDERED:
424 			dprintk("%s ERROR: %d Reset session\n", __func__,
425 				errorcode);
426 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
427 			goto wait_on_recovery;
428 #endif /* defined(CONFIG_NFS_V4_1) */
429 		case -NFS4ERR_FILE_OPEN:
430 			if (exception->timeout > HZ) {
431 				/* We have retried a decent amount, time to
432 				 * fail
433 				 */
434 				ret = -EBUSY;
435 				break;
436 			}
437 		case -NFS4ERR_DELAY:
438 			nfs_inc_server_stats(server, NFSIOS_DELAY);
439 		case -NFS4ERR_GRACE:
440 		case -NFS4ERR_LAYOUTTRYLATER:
441 		case -NFS4ERR_RECALLCONFLICT:
442 			exception->delay = 1;
443 			return 0;
444 
445 		case -NFS4ERR_RETRY_UNCACHED_REP:
446 		case -NFS4ERR_OLD_STATEID:
447 			exception->retry = 1;
448 			break;
449 		case -NFS4ERR_BADOWNER:
450 			/* The following works around a Linux server bug! */
451 		case -NFS4ERR_BADNAME:
452 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
453 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
454 				exception->retry = 1;
455 				printk(KERN_WARNING "NFS: v4 server %s "
456 						"does not accept raw "
457 						"uid/gids. "
458 						"Reenabling the idmapper.\n",
459 						server->nfs_client->cl_hostname);
460 			}
461 	}
462 	/* We failed to handle the error */
463 	return nfs4_map_errors(ret);
464 wait_on_recovery:
465 	exception->recovering = 1;
466 	return 0;
467 }
468 
469 /* This is the error handling routine for processes that are allowed
470  * to sleep.
471  */
472 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
473 {
474 	struct nfs_client *clp = server->nfs_client;
475 	int ret;
476 
477 	ret = nfs4_do_handle_exception(server, errorcode, exception);
478 	if (exception->delay) {
479 		ret = nfs4_delay(server->client, &exception->timeout);
480 		goto out_retry;
481 	}
482 	if (exception->recovering) {
483 		ret = nfs4_wait_clnt_recover(clp);
484 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
485 			return -EIO;
486 		goto out_retry;
487 	}
488 	return ret;
489 out_retry:
490 	if (ret == 0)
491 		exception->retry = 1;
492 	return ret;
493 }
494 
495 static int
496 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
497 		int errorcode, struct nfs4_exception *exception)
498 {
499 	struct nfs_client *clp = server->nfs_client;
500 	int ret;
501 
502 	ret = nfs4_do_handle_exception(server, errorcode, exception);
503 	if (exception->delay) {
504 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
505 		goto out_retry;
506 	}
507 	if (exception->recovering) {
508 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
509 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
510 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
511 		goto out_retry;
512 	}
513 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
514 		ret = -EIO;
515 	return ret;
516 out_retry:
517 	if (ret == 0)
518 		exception->retry = 1;
519 	return ret;
520 }
521 
522 static int
523 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
524 			struct nfs4_state *state, long *timeout)
525 {
526 	struct nfs4_exception exception = {
527 		.state = state,
528 	};
529 
530 	if (task->tk_status >= 0)
531 		return 0;
532 	if (timeout)
533 		exception.timeout = *timeout;
534 	task->tk_status = nfs4_async_handle_exception(task, server,
535 			task->tk_status,
536 			&exception);
537 	if (exception.delay && timeout)
538 		*timeout = exception.timeout;
539 	if (exception.retry)
540 		return -EAGAIN;
541 	return 0;
542 }
543 
544 /*
545  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
546  * or 'false' otherwise.
547  */
548 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
549 {
550 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
551 
552 	if (flavor == RPC_AUTH_GSS_KRB5I ||
553 	    flavor == RPC_AUTH_GSS_KRB5P)
554 		return true;
555 
556 	return false;
557 }
558 
559 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
560 {
561 	spin_lock(&clp->cl_lock);
562 	if (time_before(clp->cl_last_renewal,timestamp))
563 		clp->cl_last_renewal = timestamp;
564 	spin_unlock(&clp->cl_lock);
565 }
566 
567 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
568 {
569 	struct nfs_client *clp = server->nfs_client;
570 
571 	if (!nfs4_has_session(clp))
572 		do_renew_lease(clp, timestamp);
573 }
574 
575 struct nfs4_call_sync_data {
576 	const struct nfs_server *seq_server;
577 	struct nfs4_sequence_args *seq_args;
578 	struct nfs4_sequence_res *seq_res;
579 };
580 
581 void nfs4_init_sequence(struct nfs4_sequence_args *args,
582 			struct nfs4_sequence_res *res, int cache_reply)
583 {
584 	args->sa_slot = NULL;
585 	args->sa_cache_this = cache_reply;
586 	args->sa_privileged = 0;
587 
588 	res->sr_slot = NULL;
589 }
590 
591 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
592 {
593 	args->sa_privileged = 1;
594 }
595 
596 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
597 			 struct nfs4_sequence_args *args,
598 			 struct nfs4_sequence_res *res,
599 			 struct rpc_task *task)
600 {
601 	struct nfs4_slot *slot;
602 
603 	/* slot already allocated? */
604 	if (res->sr_slot != NULL)
605 		goto out_start;
606 
607 	spin_lock(&tbl->slot_tbl_lock);
608 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
609 		goto out_sleep;
610 
611 	slot = nfs4_alloc_slot(tbl);
612 	if (IS_ERR(slot)) {
613 		if (slot == ERR_PTR(-ENOMEM))
614 			task->tk_timeout = HZ >> 2;
615 		goto out_sleep;
616 	}
617 	spin_unlock(&tbl->slot_tbl_lock);
618 
619 	slot->privileged = args->sa_privileged ? 1 : 0;
620 	args->sa_slot = slot;
621 	res->sr_slot = slot;
622 
623 out_start:
624 	rpc_call_start(task);
625 	return 0;
626 
627 out_sleep:
628 	if (args->sa_privileged)
629 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
630 				NULL, RPC_PRIORITY_PRIVILEGED);
631 	else
632 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
633 	spin_unlock(&tbl->slot_tbl_lock);
634 	return -EAGAIN;
635 }
636 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
637 
638 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
639 {
640 	struct nfs4_slot *slot = res->sr_slot;
641 	struct nfs4_slot_table *tbl;
642 
643 	tbl = slot->table;
644 	spin_lock(&tbl->slot_tbl_lock);
645 	if (!nfs41_wake_and_assign_slot(tbl, slot))
646 		nfs4_free_slot(tbl, slot);
647 	spin_unlock(&tbl->slot_tbl_lock);
648 
649 	res->sr_slot = NULL;
650 }
651 
652 static int nfs40_sequence_done(struct rpc_task *task,
653 			       struct nfs4_sequence_res *res)
654 {
655 	if (res->sr_slot != NULL)
656 		nfs40_sequence_free_slot(res);
657 	return 1;
658 }
659 
660 #if defined(CONFIG_NFS_V4_1)
661 
662 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
663 {
664 	struct nfs4_session *session;
665 	struct nfs4_slot_table *tbl;
666 	struct nfs4_slot *slot = res->sr_slot;
667 	bool send_new_highest_used_slotid = false;
668 
669 	tbl = slot->table;
670 	session = tbl->session;
671 
672 	/* Bump the slot sequence number */
673 	if (slot->seq_done)
674 		slot->seq_nr++;
675 	slot->seq_done = 0;
676 
677 	spin_lock(&tbl->slot_tbl_lock);
678 	/* Be nice to the server: try to ensure that the last transmitted
679 	 * value for highest_user_slotid <= target_highest_slotid
680 	 */
681 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
682 		send_new_highest_used_slotid = true;
683 
684 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
685 		send_new_highest_used_slotid = false;
686 		goto out_unlock;
687 	}
688 	nfs4_free_slot(tbl, slot);
689 
690 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
691 		send_new_highest_used_slotid = false;
692 out_unlock:
693 	spin_unlock(&tbl->slot_tbl_lock);
694 	res->sr_slot = NULL;
695 	if (send_new_highest_used_slotid)
696 		nfs41_notify_server(session->clp);
697 	if (waitqueue_active(&tbl->slot_waitq))
698 		wake_up_all(&tbl->slot_waitq);
699 }
700 
701 static int nfs41_sequence_process(struct rpc_task *task,
702 		struct nfs4_sequence_res *res)
703 {
704 	struct nfs4_session *session;
705 	struct nfs4_slot *slot = res->sr_slot;
706 	struct nfs_client *clp;
707 	bool interrupted = false;
708 	int ret = 1;
709 
710 	if (slot == NULL)
711 		goto out_noaction;
712 	/* don't increment the sequence number if the task wasn't sent */
713 	if (!RPC_WAS_SENT(task))
714 		goto out;
715 
716 	session = slot->table->session;
717 
718 	if (slot->interrupted) {
719 		slot->interrupted = 0;
720 		interrupted = true;
721 	}
722 
723 	trace_nfs4_sequence_done(session, res);
724 	/* Check the SEQUENCE operation status */
725 	switch (res->sr_status) {
726 	case 0:
727 		/* Update the slot's sequence and clientid lease timer */
728 		slot->seq_done = 1;
729 		clp = session->clp;
730 		do_renew_lease(clp, res->sr_timestamp);
731 		/* Check sequence flags */
732 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
733 				!!slot->privileged);
734 		nfs41_update_target_slotid(slot->table, slot, res);
735 		break;
736 	case 1:
737 		/*
738 		 * sr_status remains 1 if an RPC level error occurred.
739 		 * The server may or may not have processed the sequence
740 		 * operation..
741 		 * Mark the slot as having hosted an interrupted RPC call.
742 		 */
743 		slot->interrupted = 1;
744 		goto out;
745 	case -NFS4ERR_DELAY:
746 		/* The server detected a resend of the RPC call and
747 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
748 		 * of RFC5661.
749 		 */
750 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
751 			__func__,
752 			slot->slot_nr,
753 			slot->seq_nr);
754 		goto out_retry;
755 	case -NFS4ERR_BADSLOT:
756 		/*
757 		 * The slot id we used was probably retired. Try again
758 		 * using a different slot id.
759 		 */
760 		goto retry_nowait;
761 	case -NFS4ERR_SEQ_MISORDERED:
762 		/*
763 		 * Was the last operation on this sequence interrupted?
764 		 * If so, retry after bumping the sequence number.
765 		 */
766 		if (interrupted) {
767 			++slot->seq_nr;
768 			goto retry_nowait;
769 		}
770 		/*
771 		 * Could this slot have been previously retired?
772 		 * If so, then the server may be expecting seq_nr = 1!
773 		 */
774 		if (slot->seq_nr != 1) {
775 			slot->seq_nr = 1;
776 			goto retry_nowait;
777 		}
778 		break;
779 	case -NFS4ERR_SEQ_FALSE_RETRY:
780 		++slot->seq_nr;
781 		goto retry_nowait;
782 	default:
783 		/* Just update the slot sequence no. */
784 		slot->seq_done = 1;
785 	}
786 out:
787 	/* The session may be reset by one of the error handlers. */
788 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
789 out_noaction:
790 	return ret;
791 retry_nowait:
792 	if (rpc_restart_call_prepare(task)) {
793 		nfs41_sequence_free_slot(res);
794 		task->tk_status = 0;
795 		ret = 0;
796 	}
797 	goto out;
798 out_retry:
799 	if (!rpc_restart_call(task))
800 		goto out;
801 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
802 	return 0;
803 }
804 
805 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
806 {
807 	if (!nfs41_sequence_process(task, res))
808 		return 0;
809 	if (res->sr_slot != NULL)
810 		nfs41_sequence_free_slot(res);
811 	return 1;
812 
813 }
814 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
815 
816 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
817 {
818 	if (res->sr_slot == NULL)
819 		return 1;
820 	if (res->sr_slot->table->session != NULL)
821 		return nfs41_sequence_process(task, res);
822 	return nfs40_sequence_done(task, res);
823 }
824 
825 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
826 {
827 	if (res->sr_slot != NULL) {
828 		if (res->sr_slot->table->session != NULL)
829 			nfs41_sequence_free_slot(res);
830 		else
831 			nfs40_sequence_free_slot(res);
832 	}
833 }
834 
835 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
836 {
837 	if (res->sr_slot == NULL)
838 		return 1;
839 	if (!res->sr_slot->table->session)
840 		return nfs40_sequence_done(task, res);
841 	return nfs41_sequence_done(task, res);
842 }
843 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
844 
845 int nfs41_setup_sequence(struct nfs4_session *session,
846 				struct nfs4_sequence_args *args,
847 				struct nfs4_sequence_res *res,
848 				struct rpc_task *task)
849 {
850 	struct nfs4_slot *slot;
851 	struct nfs4_slot_table *tbl;
852 
853 	dprintk("--> %s\n", __func__);
854 	/* slot already allocated? */
855 	if (res->sr_slot != NULL)
856 		goto out_success;
857 
858 	tbl = &session->fc_slot_table;
859 
860 	task->tk_timeout = 0;
861 
862 	spin_lock(&tbl->slot_tbl_lock);
863 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
864 	    !args->sa_privileged) {
865 		/* The state manager will wait until the slot table is empty */
866 		dprintk("%s session is draining\n", __func__);
867 		goto out_sleep;
868 	}
869 
870 	slot = nfs4_alloc_slot(tbl);
871 	if (IS_ERR(slot)) {
872 		/* If out of memory, try again in 1/4 second */
873 		if (slot == ERR_PTR(-ENOMEM))
874 			task->tk_timeout = HZ >> 2;
875 		dprintk("<-- %s: no free slots\n", __func__);
876 		goto out_sleep;
877 	}
878 	spin_unlock(&tbl->slot_tbl_lock);
879 
880 	slot->privileged = args->sa_privileged ? 1 : 0;
881 	args->sa_slot = slot;
882 
883 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
884 			slot->slot_nr, slot->seq_nr);
885 
886 	res->sr_slot = slot;
887 	res->sr_timestamp = jiffies;
888 	res->sr_status_flags = 0;
889 	/*
890 	 * sr_status is only set in decode_sequence, and so will remain
891 	 * set to 1 if an rpc level failure occurs.
892 	 */
893 	res->sr_status = 1;
894 	trace_nfs4_setup_sequence(session, args);
895 out_success:
896 	rpc_call_start(task);
897 	return 0;
898 out_sleep:
899 	/* Privileged tasks are queued with top priority */
900 	if (args->sa_privileged)
901 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
902 				NULL, RPC_PRIORITY_PRIVILEGED);
903 	else
904 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
905 	spin_unlock(&tbl->slot_tbl_lock);
906 	return -EAGAIN;
907 }
908 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
909 
910 static int nfs4_setup_sequence(const struct nfs_server *server,
911 			       struct nfs4_sequence_args *args,
912 			       struct nfs4_sequence_res *res,
913 			       struct rpc_task *task)
914 {
915 	struct nfs4_session *session = nfs4_get_session(server);
916 	int ret = 0;
917 
918 	if (!session)
919 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
920 					    args, res, task);
921 
922 	dprintk("--> %s clp %p session %p sr_slot %u\n",
923 		__func__, session->clp, session, res->sr_slot ?
924 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
925 
926 	ret = nfs41_setup_sequence(session, args, res, task);
927 
928 	dprintk("<-- %s status=%d\n", __func__, ret);
929 	return ret;
930 }
931 
932 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
933 {
934 	struct nfs4_call_sync_data *data = calldata;
935 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
936 
937 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
938 
939 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
940 }
941 
942 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
943 {
944 	struct nfs4_call_sync_data *data = calldata;
945 
946 	nfs41_sequence_done(task, data->seq_res);
947 }
948 
949 static const struct rpc_call_ops nfs41_call_sync_ops = {
950 	.rpc_call_prepare = nfs41_call_sync_prepare,
951 	.rpc_call_done = nfs41_call_sync_done,
952 };
953 
954 #else	/* !CONFIG_NFS_V4_1 */
955 
956 static int nfs4_setup_sequence(const struct nfs_server *server,
957 			       struct nfs4_sequence_args *args,
958 			       struct nfs4_sequence_res *res,
959 			       struct rpc_task *task)
960 {
961 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
962 				    args, res, task);
963 }
964 
965 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
966 {
967 	return nfs40_sequence_done(task, res);
968 }
969 
970 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
971 {
972 	if (res->sr_slot != NULL)
973 		nfs40_sequence_free_slot(res);
974 }
975 
976 int nfs4_sequence_done(struct rpc_task *task,
977 		       struct nfs4_sequence_res *res)
978 {
979 	return nfs40_sequence_done(task, res);
980 }
981 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
982 
983 #endif	/* !CONFIG_NFS_V4_1 */
984 
985 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
986 {
987 	struct nfs4_call_sync_data *data = calldata;
988 	nfs4_setup_sequence(data->seq_server,
989 				data->seq_args, data->seq_res, task);
990 }
991 
992 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
993 {
994 	struct nfs4_call_sync_data *data = calldata;
995 	nfs4_sequence_done(task, data->seq_res);
996 }
997 
998 static const struct rpc_call_ops nfs40_call_sync_ops = {
999 	.rpc_call_prepare = nfs40_call_sync_prepare,
1000 	.rpc_call_done = nfs40_call_sync_done,
1001 };
1002 
1003 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1004 				   struct nfs_server *server,
1005 				   struct rpc_message *msg,
1006 				   struct nfs4_sequence_args *args,
1007 				   struct nfs4_sequence_res *res)
1008 {
1009 	int ret;
1010 	struct rpc_task *task;
1011 	struct nfs_client *clp = server->nfs_client;
1012 	struct nfs4_call_sync_data data = {
1013 		.seq_server = server,
1014 		.seq_args = args,
1015 		.seq_res = res,
1016 	};
1017 	struct rpc_task_setup task_setup = {
1018 		.rpc_client = clnt,
1019 		.rpc_message = msg,
1020 		.callback_ops = clp->cl_mvops->call_sync_ops,
1021 		.callback_data = &data
1022 	};
1023 
1024 	task = rpc_run_task(&task_setup);
1025 	if (IS_ERR(task))
1026 		ret = PTR_ERR(task);
1027 	else {
1028 		ret = task->tk_status;
1029 		rpc_put_task(task);
1030 	}
1031 	return ret;
1032 }
1033 
1034 int nfs4_call_sync(struct rpc_clnt *clnt,
1035 		   struct nfs_server *server,
1036 		   struct rpc_message *msg,
1037 		   struct nfs4_sequence_args *args,
1038 		   struct nfs4_sequence_res *res,
1039 		   int cache_reply)
1040 {
1041 	nfs4_init_sequence(args, res, cache_reply);
1042 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1043 }
1044 
1045 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1046 {
1047 	struct nfs_inode *nfsi = NFS_I(dir);
1048 
1049 	spin_lock(&dir->i_lock);
1050 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1051 	if (!cinfo->atomic || cinfo->before != dir->i_version)
1052 		nfs_force_lookup_revalidate(dir);
1053 	dir->i_version = cinfo->after;
1054 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1055 	nfs_fscache_invalidate(dir);
1056 	spin_unlock(&dir->i_lock);
1057 }
1058 
1059 struct nfs4_opendata {
1060 	struct kref kref;
1061 	struct nfs_openargs o_arg;
1062 	struct nfs_openres o_res;
1063 	struct nfs_open_confirmargs c_arg;
1064 	struct nfs_open_confirmres c_res;
1065 	struct nfs4_string owner_name;
1066 	struct nfs4_string group_name;
1067 	struct nfs4_label *a_label;
1068 	struct nfs_fattr f_attr;
1069 	struct nfs4_label *f_label;
1070 	struct dentry *dir;
1071 	struct dentry *dentry;
1072 	struct nfs4_state_owner *owner;
1073 	struct nfs4_state *state;
1074 	struct iattr attrs;
1075 	unsigned long timestamp;
1076 	unsigned int rpc_done : 1;
1077 	unsigned int file_created : 1;
1078 	unsigned int is_recover : 1;
1079 	int rpc_status;
1080 	int cancelled;
1081 };
1082 
1083 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1084 		int err, struct nfs4_exception *exception)
1085 {
1086 	if (err != -EINVAL)
1087 		return false;
1088 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1089 		return false;
1090 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1091 	exception->retry = 1;
1092 	return true;
1093 }
1094 
1095 static u32
1096 nfs4_map_atomic_open_share(struct nfs_server *server,
1097 		fmode_t fmode, int openflags)
1098 {
1099 	u32 res = 0;
1100 
1101 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1102 	case FMODE_READ:
1103 		res = NFS4_SHARE_ACCESS_READ;
1104 		break;
1105 	case FMODE_WRITE:
1106 		res = NFS4_SHARE_ACCESS_WRITE;
1107 		break;
1108 	case FMODE_READ|FMODE_WRITE:
1109 		res = NFS4_SHARE_ACCESS_BOTH;
1110 	}
1111 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1112 		goto out;
1113 	/* Want no delegation if we're using O_DIRECT */
1114 	if (openflags & O_DIRECT)
1115 		res |= NFS4_SHARE_WANT_NO_DELEG;
1116 out:
1117 	return res;
1118 }
1119 
1120 static enum open_claim_type4
1121 nfs4_map_atomic_open_claim(struct nfs_server *server,
1122 		enum open_claim_type4 claim)
1123 {
1124 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1125 		return claim;
1126 	switch (claim) {
1127 	default:
1128 		return claim;
1129 	case NFS4_OPEN_CLAIM_FH:
1130 		return NFS4_OPEN_CLAIM_NULL;
1131 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1132 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1133 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1134 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1135 	}
1136 }
1137 
1138 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1139 {
1140 	p->o_res.f_attr = &p->f_attr;
1141 	p->o_res.f_label = p->f_label;
1142 	p->o_res.seqid = p->o_arg.seqid;
1143 	p->c_res.seqid = p->c_arg.seqid;
1144 	p->o_res.server = p->o_arg.server;
1145 	p->o_res.access_request = p->o_arg.access;
1146 	nfs_fattr_init(&p->f_attr);
1147 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1148 }
1149 
1150 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1151 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1152 		const struct iattr *attrs,
1153 		struct nfs4_label *label,
1154 		enum open_claim_type4 claim,
1155 		gfp_t gfp_mask)
1156 {
1157 	struct dentry *parent = dget_parent(dentry);
1158 	struct inode *dir = d_inode(parent);
1159 	struct nfs_server *server = NFS_SERVER(dir);
1160 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1161 	struct nfs4_opendata *p;
1162 
1163 	p = kzalloc(sizeof(*p), gfp_mask);
1164 	if (p == NULL)
1165 		goto err;
1166 
1167 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1168 	if (IS_ERR(p->f_label))
1169 		goto err_free_p;
1170 
1171 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1172 	if (IS_ERR(p->a_label))
1173 		goto err_free_f;
1174 
1175 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1176 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1177 	if (IS_ERR(p->o_arg.seqid))
1178 		goto err_free_label;
1179 	nfs_sb_active(dentry->d_sb);
1180 	p->dentry = dget(dentry);
1181 	p->dir = parent;
1182 	p->owner = sp;
1183 	atomic_inc(&sp->so_count);
1184 	p->o_arg.open_flags = flags;
1185 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1186 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1187 			fmode, flags);
1188 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1189 	 * will return permission denied for all bits until close */
1190 	if (!(flags & O_EXCL)) {
1191 		/* ask server to check for all possible rights as results
1192 		 * are cached */
1193 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1194 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1195 	}
1196 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1197 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1198 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1199 	p->o_arg.name = &dentry->d_name;
1200 	p->o_arg.server = server;
1201 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1202 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1203 	p->o_arg.label = nfs4_label_copy(p->a_label, label);
1204 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1205 	switch (p->o_arg.claim) {
1206 	case NFS4_OPEN_CLAIM_NULL:
1207 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1208 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1209 		p->o_arg.fh = NFS_FH(dir);
1210 		break;
1211 	case NFS4_OPEN_CLAIM_PREVIOUS:
1212 	case NFS4_OPEN_CLAIM_FH:
1213 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1214 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1215 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1216 	}
1217 	if (attrs != NULL && attrs->ia_valid != 0) {
1218 		__u32 verf[2];
1219 
1220 		p->o_arg.u.attrs = &p->attrs;
1221 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1222 
1223 		verf[0] = jiffies;
1224 		verf[1] = current->pid;
1225 		memcpy(p->o_arg.u.verifier.data, verf,
1226 				sizeof(p->o_arg.u.verifier.data));
1227 	}
1228 	p->c_arg.fh = &p->o_res.fh;
1229 	p->c_arg.stateid = &p->o_res.stateid;
1230 	p->c_arg.seqid = p->o_arg.seqid;
1231 	nfs4_init_opendata_res(p);
1232 	kref_init(&p->kref);
1233 	return p;
1234 
1235 err_free_label:
1236 	nfs4_label_free(p->a_label);
1237 err_free_f:
1238 	nfs4_label_free(p->f_label);
1239 err_free_p:
1240 	kfree(p);
1241 err:
1242 	dput(parent);
1243 	return NULL;
1244 }
1245 
1246 static void nfs4_opendata_free(struct kref *kref)
1247 {
1248 	struct nfs4_opendata *p = container_of(kref,
1249 			struct nfs4_opendata, kref);
1250 	struct super_block *sb = p->dentry->d_sb;
1251 
1252 	nfs_free_seqid(p->o_arg.seqid);
1253 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1254 	if (p->state != NULL)
1255 		nfs4_put_open_state(p->state);
1256 	nfs4_put_state_owner(p->owner);
1257 
1258 	nfs4_label_free(p->a_label);
1259 	nfs4_label_free(p->f_label);
1260 
1261 	dput(p->dir);
1262 	dput(p->dentry);
1263 	nfs_sb_deactive(sb);
1264 	nfs_fattr_free_names(&p->f_attr);
1265 	kfree(p->f_attr.mdsthreshold);
1266 	kfree(p);
1267 }
1268 
1269 static void nfs4_opendata_put(struct nfs4_opendata *p)
1270 {
1271 	if (p != NULL)
1272 		kref_put(&p->kref, nfs4_opendata_free);
1273 }
1274 
1275 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1276 {
1277 	int ret;
1278 
1279 	ret = rpc_wait_for_completion_task(task);
1280 	return ret;
1281 }
1282 
1283 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1284 		fmode_t fmode)
1285 {
1286 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1287 	case FMODE_READ|FMODE_WRITE:
1288 		return state->n_rdwr != 0;
1289 	case FMODE_WRITE:
1290 		return state->n_wronly != 0;
1291 	case FMODE_READ:
1292 		return state->n_rdonly != 0;
1293 	}
1294 	WARN_ON_ONCE(1);
1295 	return false;
1296 }
1297 
1298 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1299 {
1300 	int ret = 0;
1301 
1302 	if (open_mode & (O_EXCL|O_TRUNC))
1303 		goto out;
1304 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1305 		case FMODE_READ:
1306 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1307 				&& state->n_rdonly != 0;
1308 			break;
1309 		case FMODE_WRITE:
1310 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1311 				&& state->n_wronly != 0;
1312 			break;
1313 		case FMODE_READ|FMODE_WRITE:
1314 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1315 				&& state->n_rdwr != 0;
1316 	}
1317 out:
1318 	return ret;
1319 }
1320 
1321 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1322 		enum open_claim_type4 claim)
1323 {
1324 	if (delegation == NULL)
1325 		return 0;
1326 	if ((delegation->type & fmode) != fmode)
1327 		return 0;
1328 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1329 		return 0;
1330 	switch (claim) {
1331 	case NFS4_OPEN_CLAIM_NULL:
1332 	case NFS4_OPEN_CLAIM_FH:
1333 		break;
1334 	case NFS4_OPEN_CLAIM_PREVIOUS:
1335 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1336 			break;
1337 	default:
1338 		return 0;
1339 	}
1340 	nfs_mark_delegation_referenced(delegation);
1341 	return 1;
1342 }
1343 
1344 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1345 {
1346 	switch (fmode) {
1347 		case FMODE_WRITE:
1348 			state->n_wronly++;
1349 			break;
1350 		case FMODE_READ:
1351 			state->n_rdonly++;
1352 			break;
1353 		case FMODE_READ|FMODE_WRITE:
1354 			state->n_rdwr++;
1355 	}
1356 	nfs4_state_set_mode_locked(state, state->state | fmode);
1357 }
1358 
1359 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1360 {
1361 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1362 	bool need_recover = false;
1363 
1364 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1365 		need_recover = true;
1366 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1367 		need_recover = true;
1368 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1369 		need_recover = true;
1370 	if (need_recover)
1371 		nfs4_state_mark_reclaim_nograce(clp, state);
1372 }
1373 
1374 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1375 		nfs4_stateid *stateid)
1376 {
1377 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1378 		return true;
1379 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1380 		nfs_test_and_clear_all_open_stateid(state);
1381 		return true;
1382 	}
1383 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1384 		return true;
1385 	return false;
1386 }
1387 
1388 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1389 {
1390 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1391 		return;
1392 	if (state->n_wronly)
1393 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1394 	if (state->n_rdonly)
1395 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1396 	if (state->n_rdwr)
1397 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1398 	set_bit(NFS_OPEN_STATE, &state->flags);
1399 }
1400 
1401 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1402 		nfs4_stateid *arg_stateid,
1403 		nfs4_stateid *stateid, fmode_t fmode)
1404 {
1405 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1406 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1407 	case FMODE_WRITE:
1408 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1409 		break;
1410 	case FMODE_READ:
1411 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1412 		break;
1413 	case 0:
1414 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1415 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1416 		clear_bit(NFS_OPEN_STATE, &state->flags);
1417 	}
1418 	if (stateid == NULL)
1419 		return;
1420 	/* Handle races with OPEN */
1421 	if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1422 	    (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1423 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1424 		nfs_resync_open_stateid_locked(state);
1425 		return;
1426 	}
1427 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1428 		nfs4_stateid_copy(&state->stateid, stateid);
1429 	nfs4_stateid_copy(&state->open_stateid, stateid);
1430 }
1431 
1432 static void nfs_clear_open_stateid(struct nfs4_state *state,
1433 	nfs4_stateid *arg_stateid,
1434 	nfs4_stateid *stateid, fmode_t fmode)
1435 {
1436 	write_seqlock(&state->seqlock);
1437 	nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1438 	write_sequnlock(&state->seqlock);
1439 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1440 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1441 }
1442 
1443 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1444 {
1445 	switch (fmode) {
1446 		case FMODE_READ:
1447 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1448 			break;
1449 		case FMODE_WRITE:
1450 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1451 			break;
1452 		case FMODE_READ|FMODE_WRITE:
1453 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1454 	}
1455 	if (!nfs_need_update_open_stateid(state, stateid))
1456 		return;
1457 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1458 		nfs4_stateid_copy(&state->stateid, stateid);
1459 	nfs4_stateid_copy(&state->open_stateid, stateid);
1460 }
1461 
1462 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1463 {
1464 	/*
1465 	 * Protect the call to nfs4_state_set_mode_locked and
1466 	 * serialise the stateid update
1467 	 */
1468 	spin_lock(&state->owner->so_lock);
1469 	write_seqlock(&state->seqlock);
1470 	if (deleg_stateid != NULL) {
1471 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1472 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1473 	}
1474 	if (open_stateid != NULL)
1475 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1476 	write_sequnlock(&state->seqlock);
1477 	update_open_stateflags(state, fmode);
1478 	spin_unlock(&state->owner->so_lock);
1479 }
1480 
1481 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1482 {
1483 	struct nfs_inode *nfsi = NFS_I(state->inode);
1484 	struct nfs_delegation *deleg_cur;
1485 	int ret = 0;
1486 
1487 	fmode &= (FMODE_READ|FMODE_WRITE);
1488 
1489 	rcu_read_lock();
1490 	deleg_cur = rcu_dereference(nfsi->delegation);
1491 	if (deleg_cur == NULL)
1492 		goto no_delegation;
1493 
1494 	spin_lock(&deleg_cur->lock);
1495 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1496 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1497 	    (deleg_cur->type & fmode) != fmode)
1498 		goto no_delegation_unlock;
1499 
1500 	if (delegation == NULL)
1501 		delegation = &deleg_cur->stateid;
1502 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1503 		goto no_delegation_unlock;
1504 
1505 	nfs_mark_delegation_referenced(deleg_cur);
1506 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1507 	ret = 1;
1508 no_delegation_unlock:
1509 	spin_unlock(&deleg_cur->lock);
1510 no_delegation:
1511 	rcu_read_unlock();
1512 
1513 	if (!ret && open_stateid != NULL) {
1514 		__update_open_stateid(state, open_stateid, NULL, fmode);
1515 		ret = 1;
1516 	}
1517 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1518 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1519 
1520 	return ret;
1521 }
1522 
1523 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1524 		const nfs4_stateid *stateid)
1525 {
1526 	struct nfs4_state *state = lsp->ls_state;
1527 	bool ret = false;
1528 
1529 	spin_lock(&state->state_lock);
1530 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1531 		goto out_noupdate;
1532 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1533 		goto out_noupdate;
1534 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1535 	ret = true;
1536 out_noupdate:
1537 	spin_unlock(&state->state_lock);
1538 	return ret;
1539 }
1540 
1541 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1542 {
1543 	struct nfs_delegation *delegation;
1544 
1545 	rcu_read_lock();
1546 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1547 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1548 		rcu_read_unlock();
1549 		return;
1550 	}
1551 	rcu_read_unlock();
1552 	nfs4_inode_return_delegation(inode);
1553 }
1554 
1555 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1556 {
1557 	struct nfs4_state *state = opendata->state;
1558 	struct nfs_inode *nfsi = NFS_I(state->inode);
1559 	struct nfs_delegation *delegation;
1560 	int open_mode = opendata->o_arg.open_flags;
1561 	fmode_t fmode = opendata->o_arg.fmode;
1562 	enum open_claim_type4 claim = opendata->o_arg.claim;
1563 	nfs4_stateid stateid;
1564 	int ret = -EAGAIN;
1565 
1566 	for (;;) {
1567 		spin_lock(&state->owner->so_lock);
1568 		if (can_open_cached(state, fmode, open_mode)) {
1569 			update_open_stateflags(state, fmode);
1570 			spin_unlock(&state->owner->so_lock);
1571 			goto out_return_state;
1572 		}
1573 		spin_unlock(&state->owner->so_lock);
1574 		rcu_read_lock();
1575 		delegation = rcu_dereference(nfsi->delegation);
1576 		if (!can_open_delegated(delegation, fmode, claim)) {
1577 			rcu_read_unlock();
1578 			break;
1579 		}
1580 		/* Save the delegation */
1581 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1582 		rcu_read_unlock();
1583 		nfs_release_seqid(opendata->o_arg.seqid);
1584 		if (!opendata->is_recover) {
1585 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1586 			if (ret != 0)
1587 				goto out;
1588 		}
1589 		ret = -EAGAIN;
1590 
1591 		/* Try to update the stateid using the delegation */
1592 		if (update_open_stateid(state, NULL, &stateid, fmode))
1593 			goto out_return_state;
1594 	}
1595 out:
1596 	return ERR_PTR(ret);
1597 out_return_state:
1598 	atomic_inc(&state->count);
1599 	return state;
1600 }
1601 
1602 static void
1603 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1604 {
1605 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1606 	struct nfs_delegation *delegation;
1607 	int delegation_flags = 0;
1608 
1609 	rcu_read_lock();
1610 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1611 	if (delegation)
1612 		delegation_flags = delegation->flags;
1613 	rcu_read_unlock();
1614 	switch (data->o_arg.claim) {
1615 	default:
1616 		break;
1617 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1618 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1619 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1620 				   "returning a delegation for "
1621 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1622 				   clp->cl_hostname);
1623 		return;
1624 	}
1625 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1626 		nfs_inode_set_delegation(state->inode,
1627 					 data->owner->so_cred,
1628 					 &data->o_res);
1629 	else
1630 		nfs_inode_reclaim_delegation(state->inode,
1631 					     data->owner->so_cred,
1632 					     &data->o_res);
1633 }
1634 
1635 /*
1636  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1637  * and update the nfs4_state.
1638  */
1639 static struct nfs4_state *
1640 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1641 {
1642 	struct inode *inode = data->state->inode;
1643 	struct nfs4_state *state = data->state;
1644 	int ret;
1645 
1646 	if (!data->rpc_done) {
1647 		if (data->rpc_status) {
1648 			ret = data->rpc_status;
1649 			goto err;
1650 		}
1651 		/* cached opens have already been processed */
1652 		goto update;
1653 	}
1654 
1655 	ret = nfs_refresh_inode(inode, &data->f_attr);
1656 	if (ret)
1657 		goto err;
1658 
1659 	if (data->o_res.delegation_type != 0)
1660 		nfs4_opendata_check_deleg(data, state);
1661 update:
1662 	update_open_stateid(state, &data->o_res.stateid, NULL,
1663 			    data->o_arg.fmode);
1664 	atomic_inc(&state->count);
1665 
1666 	return state;
1667 err:
1668 	return ERR_PTR(ret);
1669 
1670 }
1671 
1672 static struct nfs4_state *
1673 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1674 {
1675 	struct inode *inode;
1676 	struct nfs4_state *state = NULL;
1677 	int ret;
1678 
1679 	if (!data->rpc_done) {
1680 		state = nfs4_try_open_cached(data);
1681 		trace_nfs4_cached_open(data->state);
1682 		goto out;
1683 	}
1684 
1685 	ret = -EAGAIN;
1686 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1687 		goto err;
1688 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1689 	ret = PTR_ERR(inode);
1690 	if (IS_ERR(inode))
1691 		goto err;
1692 	ret = -ENOMEM;
1693 	state = nfs4_get_open_state(inode, data->owner);
1694 	if (state == NULL)
1695 		goto err_put_inode;
1696 	if (data->o_res.delegation_type != 0)
1697 		nfs4_opendata_check_deleg(data, state);
1698 	update_open_stateid(state, &data->o_res.stateid, NULL,
1699 			data->o_arg.fmode);
1700 	iput(inode);
1701 out:
1702 	nfs_release_seqid(data->o_arg.seqid);
1703 	return state;
1704 err_put_inode:
1705 	iput(inode);
1706 err:
1707 	return ERR_PTR(ret);
1708 }
1709 
1710 static struct nfs4_state *
1711 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1712 {
1713 	struct nfs4_state *ret;
1714 
1715 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1716 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1717 	else
1718 		ret = _nfs4_opendata_to_nfs4_state(data);
1719 	nfs4_sequence_free_slot(&data->o_res.seq_res);
1720 	return ret;
1721 }
1722 
1723 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1724 {
1725 	struct nfs_inode *nfsi = NFS_I(state->inode);
1726 	struct nfs_open_context *ctx;
1727 
1728 	spin_lock(&state->inode->i_lock);
1729 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1730 		if (ctx->state != state)
1731 			continue;
1732 		get_nfs_open_context(ctx);
1733 		spin_unlock(&state->inode->i_lock);
1734 		return ctx;
1735 	}
1736 	spin_unlock(&state->inode->i_lock);
1737 	return ERR_PTR(-ENOENT);
1738 }
1739 
1740 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1741 		struct nfs4_state *state, enum open_claim_type4 claim)
1742 {
1743 	struct nfs4_opendata *opendata;
1744 
1745 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1746 			NULL, NULL, claim, GFP_NOFS);
1747 	if (opendata == NULL)
1748 		return ERR_PTR(-ENOMEM);
1749 	opendata->state = state;
1750 	atomic_inc(&state->count);
1751 	return opendata;
1752 }
1753 
1754 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1755 		fmode_t fmode)
1756 {
1757 	struct nfs4_state *newstate;
1758 	int ret;
1759 
1760 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1761 		return 0;
1762 	opendata->o_arg.open_flags = 0;
1763 	opendata->o_arg.fmode = fmode;
1764 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1765 			NFS_SB(opendata->dentry->d_sb),
1766 			fmode, 0);
1767 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1768 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1769 	nfs4_init_opendata_res(opendata);
1770 	ret = _nfs4_recover_proc_open(opendata);
1771 	if (ret != 0)
1772 		return ret;
1773 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1774 	if (IS_ERR(newstate))
1775 		return PTR_ERR(newstate);
1776 	if (newstate != opendata->state)
1777 		ret = -ESTALE;
1778 	nfs4_close_state(newstate, fmode);
1779 	return ret;
1780 }
1781 
1782 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1783 {
1784 	int ret;
1785 
1786 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1787 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1788 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1789 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1790 	/* memory barrier prior to reading state->n_* */
1791 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1792 	clear_bit(NFS_OPEN_STATE, &state->flags);
1793 	smp_rmb();
1794 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1795 	if (ret != 0)
1796 		return ret;
1797 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1798 	if (ret != 0)
1799 		return ret;
1800 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1801 	if (ret != 0)
1802 		return ret;
1803 	/*
1804 	 * We may have performed cached opens for all three recoveries.
1805 	 * Check if we need to update the current stateid.
1806 	 */
1807 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1808 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1809 		write_seqlock(&state->seqlock);
1810 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1811 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1812 		write_sequnlock(&state->seqlock);
1813 	}
1814 	return 0;
1815 }
1816 
1817 /*
1818  * OPEN_RECLAIM:
1819  * 	reclaim state on the server after a reboot.
1820  */
1821 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1822 {
1823 	struct nfs_delegation *delegation;
1824 	struct nfs4_opendata *opendata;
1825 	fmode_t delegation_type = 0;
1826 	int status;
1827 
1828 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1829 			NFS4_OPEN_CLAIM_PREVIOUS);
1830 	if (IS_ERR(opendata))
1831 		return PTR_ERR(opendata);
1832 	rcu_read_lock();
1833 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1834 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1835 		delegation_type = delegation->type;
1836 	rcu_read_unlock();
1837 	opendata->o_arg.u.delegation_type = delegation_type;
1838 	status = nfs4_open_recover(opendata, state);
1839 	nfs4_opendata_put(opendata);
1840 	return status;
1841 }
1842 
1843 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1844 {
1845 	struct nfs_server *server = NFS_SERVER(state->inode);
1846 	struct nfs4_exception exception = { };
1847 	int err;
1848 	do {
1849 		err = _nfs4_do_open_reclaim(ctx, state);
1850 		trace_nfs4_open_reclaim(ctx, 0, err);
1851 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1852 			continue;
1853 		if (err != -NFS4ERR_DELAY)
1854 			break;
1855 		nfs4_handle_exception(server, err, &exception);
1856 	} while (exception.retry);
1857 	return err;
1858 }
1859 
1860 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1861 {
1862 	struct nfs_open_context *ctx;
1863 	int ret;
1864 
1865 	ctx = nfs4_state_find_open_context(state);
1866 	if (IS_ERR(ctx))
1867 		return -EAGAIN;
1868 	ret = nfs4_do_open_reclaim(ctx, state);
1869 	put_nfs_open_context(ctx);
1870 	return ret;
1871 }
1872 
1873 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1874 {
1875 	switch (err) {
1876 		default:
1877 			printk(KERN_ERR "NFS: %s: unhandled error "
1878 					"%d.\n", __func__, err);
1879 		case 0:
1880 		case -ENOENT:
1881 		case -EAGAIN:
1882 		case -ESTALE:
1883 			break;
1884 		case -NFS4ERR_BADSESSION:
1885 		case -NFS4ERR_BADSLOT:
1886 		case -NFS4ERR_BAD_HIGH_SLOT:
1887 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1888 		case -NFS4ERR_DEADSESSION:
1889 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1890 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1891 			return -EAGAIN;
1892 		case -NFS4ERR_STALE_CLIENTID:
1893 		case -NFS4ERR_STALE_STATEID:
1894 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1895 		case -NFS4ERR_EXPIRED:
1896 			/* Don't recall a delegation if it was lost */
1897 			nfs4_schedule_lease_recovery(server->nfs_client);
1898 			return -EAGAIN;
1899 		case -NFS4ERR_MOVED:
1900 			nfs4_schedule_migration_recovery(server);
1901 			return -EAGAIN;
1902 		case -NFS4ERR_LEASE_MOVED:
1903 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1904 			return -EAGAIN;
1905 		case -NFS4ERR_DELEG_REVOKED:
1906 		case -NFS4ERR_ADMIN_REVOKED:
1907 		case -NFS4ERR_BAD_STATEID:
1908 		case -NFS4ERR_OPENMODE:
1909 			nfs_inode_find_state_and_recover(state->inode,
1910 					stateid);
1911 			nfs4_schedule_stateid_recovery(server, state);
1912 			return -EAGAIN;
1913 		case -NFS4ERR_DELAY:
1914 		case -NFS4ERR_GRACE:
1915 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1916 			ssleep(1);
1917 			return -EAGAIN;
1918 		case -ENOMEM:
1919 		case -NFS4ERR_DENIED:
1920 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1921 			return 0;
1922 	}
1923 	return err;
1924 }
1925 
1926 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1927 		struct nfs4_state *state, const nfs4_stateid *stateid,
1928 		fmode_t type)
1929 {
1930 	struct nfs_server *server = NFS_SERVER(state->inode);
1931 	struct nfs4_opendata *opendata;
1932 	int err = 0;
1933 
1934 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1935 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1936 	if (IS_ERR(opendata))
1937 		return PTR_ERR(opendata);
1938 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1939 	write_seqlock(&state->seqlock);
1940 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1941 	write_sequnlock(&state->seqlock);
1942 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1943 	switch (type & (FMODE_READ|FMODE_WRITE)) {
1944 	case FMODE_READ|FMODE_WRITE:
1945 	case FMODE_WRITE:
1946 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1947 		if (err)
1948 			break;
1949 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1950 		if (err)
1951 			break;
1952 	case FMODE_READ:
1953 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
1954 	}
1955 	nfs4_opendata_put(opendata);
1956 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1957 }
1958 
1959 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1960 {
1961 	struct nfs4_opendata *data = calldata;
1962 
1963 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1964 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1965 }
1966 
1967 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1968 {
1969 	struct nfs4_opendata *data = calldata;
1970 
1971 	nfs40_sequence_done(task, &data->c_res.seq_res);
1972 
1973 	data->rpc_status = task->tk_status;
1974 	if (data->rpc_status == 0) {
1975 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1976 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1977 		renew_lease(data->o_res.server, data->timestamp);
1978 		data->rpc_done = 1;
1979 	}
1980 }
1981 
1982 static void nfs4_open_confirm_release(void *calldata)
1983 {
1984 	struct nfs4_opendata *data = calldata;
1985 	struct nfs4_state *state = NULL;
1986 
1987 	/* If this request hasn't been cancelled, do nothing */
1988 	if (data->cancelled == 0)
1989 		goto out_free;
1990 	/* In case of error, no cleanup! */
1991 	if (!data->rpc_done)
1992 		goto out_free;
1993 	state = nfs4_opendata_to_nfs4_state(data);
1994 	if (!IS_ERR(state))
1995 		nfs4_close_state(state, data->o_arg.fmode);
1996 out_free:
1997 	nfs4_opendata_put(data);
1998 }
1999 
2000 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2001 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2002 	.rpc_call_done = nfs4_open_confirm_done,
2003 	.rpc_release = nfs4_open_confirm_release,
2004 };
2005 
2006 /*
2007  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2008  */
2009 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2010 {
2011 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2012 	struct rpc_task *task;
2013 	struct  rpc_message msg = {
2014 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2015 		.rpc_argp = &data->c_arg,
2016 		.rpc_resp = &data->c_res,
2017 		.rpc_cred = data->owner->so_cred,
2018 	};
2019 	struct rpc_task_setup task_setup_data = {
2020 		.rpc_client = server->client,
2021 		.rpc_message = &msg,
2022 		.callback_ops = &nfs4_open_confirm_ops,
2023 		.callback_data = data,
2024 		.workqueue = nfsiod_workqueue,
2025 		.flags = RPC_TASK_ASYNC,
2026 	};
2027 	int status;
2028 
2029 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2030 	kref_get(&data->kref);
2031 	data->rpc_done = 0;
2032 	data->rpc_status = 0;
2033 	data->timestamp = jiffies;
2034 	if (data->is_recover)
2035 		nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2036 	task = rpc_run_task(&task_setup_data);
2037 	if (IS_ERR(task))
2038 		return PTR_ERR(task);
2039 	status = nfs4_wait_for_completion_rpc_task(task);
2040 	if (status != 0) {
2041 		data->cancelled = 1;
2042 		smp_wmb();
2043 	} else
2044 		status = data->rpc_status;
2045 	rpc_put_task(task);
2046 	return status;
2047 }
2048 
2049 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2050 {
2051 	struct nfs4_opendata *data = calldata;
2052 	struct nfs4_state_owner *sp = data->owner;
2053 	struct nfs_client *clp = sp->so_server->nfs_client;
2054 	enum open_claim_type4 claim = data->o_arg.claim;
2055 
2056 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2057 		goto out_wait;
2058 	/*
2059 	 * Check if we still need to send an OPEN call, or if we can use
2060 	 * a delegation instead.
2061 	 */
2062 	if (data->state != NULL) {
2063 		struct nfs_delegation *delegation;
2064 
2065 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2066 			goto out_no_action;
2067 		rcu_read_lock();
2068 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2069 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2070 			goto unlock_no_action;
2071 		rcu_read_unlock();
2072 	}
2073 	/* Update client id. */
2074 	data->o_arg.clientid = clp->cl_clientid;
2075 	switch (claim) {
2076 	default:
2077 		break;
2078 	case NFS4_OPEN_CLAIM_PREVIOUS:
2079 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2080 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2081 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2082 	case NFS4_OPEN_CLAIM_FH:
2083 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2084 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2085 	}
2086 	data->timestamp = jiffies;
2087 	if (nfs4_setup_sequence(data->o_arg.server,
2088 				&data->o_arg.seq_args,
2089 				&data->o_res.seq_res,
2090 				task) != 0)
2091 		nfs_release_seqid(data->o_arg.seqid);
2092 
2093 	/* Set the create mode (note dependency on the session type) */
2094 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2095 	if (data->o_arg.open_flags & O_EXCL) {
2096 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2097 		if (nfs4_has_persistent_session(clp))
2098 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2099 		else if (clp->cl_mvops->minor_version > 0)
2100 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2101 	}
2102 	return;
2103 unlock_no_action:
2104 	trace_nfs4_cached_open(data->state);
2105 	rcu_read_unlock();
2106 out_no_action:
2107 	task->tk_action = NULL;
2108 out_wait:
2109 	nfs4_sequence_done(task, &data->o_res.seq_res);
2110 }
2111 
2112 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2113 {
2114 	struct nfs4_opendata *data = calldata;
2115 
2116 	data->rpc_status = task->tk_status;
2117 
2118 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2119 		return;
2120 
2121 	if (task->tk_status == 0) {
2122 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2123 			switch (data->o_res.f_attr->mode & S_IFMT) {
2124 			case S_IFREG:
2125 				break;
2126 			case S_IFLNK:
2127 				data->rpc_status = -ELOOP;
2128 				break;
2129 			case S_IFDIR:
2130 				data->rpc_status = -EISDIR;
2131 				break;
2132 			default:
2133 				data->rpc_status = -ENOTDIR;
2134 			}
2135 		}
2136 		renew_lease(data->o_res.server, data->timestamp);
2137 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2138 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2139 	}
2140 	data->rpc_done = 1;
2141 }
2142 
2143 static void nfs4_open_release(void *calldata)
2144 {
2145 	struct nfs4_opendata *data = calldata;
2146 	struct nfs4_state *state = NULL;
2147 
2148 	/* If this request hasn't been cancelled, do nothing */
2149 	if (data->cancelled == 0)
2150 		goto out_free;
2151 	/* In case of error, no cleanup! */
2152 	if (data->rpc_status != 0 || !data->rpc_done)
2153 		goto out_free;
2154 	/* In case we need an open_confirm, no cleanup! */
2155 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2156 		goto out_free;
2157 	state = nfs4_opendata_to_nfs4_state(data);
2158 	if (!IS_ERR(state))
2159 		nfs4_close_state(state, data->o_arg.fmode);
2160 out_free:
2161 	nfs4_opendata_put(data);
2162 }
2163 
2164 static const struct rpc_call_ops nfs4_open_ops = {
2165 	.rpc_call_prepare = nfs4_open_prepare,
2166 	.rpc_call_done = nfs4_open_done,
2167 	.rpc_release = nfs4_open_release,
2168 };
2169 
2170 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2171 {
2172 	struct inode *dir = d_inode(data->dir);
2173 	struct nfs_server *server = NFS_SERVER(dir);
2174 	struct nfs_openargs *o_arg = &data->o_arg;
2175 	struct nfs_openres *o_res = &data->o_res;
2176 	struct rpc_task *task;
2177 	struct rpc_message msg = {
2178 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2179 		.rpc_argp = o_arg,
2180 		.rpc_resp = o_res,
2181 		.rpc_cred = data->owner->so_cred,
2182 	};
2183 	struct rpc_task_setup task_setup_data = {
2184 		.rpc_client = server->client,
2185 		.rpc_message = &msg,
2186 		.callback_ops = &nfs4_open_ops,
2187 		.callback_data = data,
2188 		.workqueue = nfsiod_workqueue,
2189 		.flags = RPC_TASK_ASYNC,
2190 	};
2191 	int status;
2192 
2193 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2194 	kref_get(&data->kref);
2195 	data->rpc_done = 0;
2196 	data->rpc_status = 0;
2197 	data->cancelled = 0;
2198 	data->is_recover = 0;
2199 	if (isrecover) {
2200 		nfs4_set_sequence_privileged(&o_arg->seq_args);
2201 		data->is_recover = 1;
2202 	}
2203 	task = rpc_run_task(&task_setup_data);
2204         if (IS_ERR(task))
2205                 return PTR_ERR(task);
2206         status = nfs4_wait_for_completion_rpc_task(task);
2207         if (status != 0) {
2208                 data->cancelled = 1;
2209                 smp_wmb();
2210         } else
2211                 status = data->rpc_status;
2212         rpc_put_task(task);
2213 
2214 	return status;
2215 }
2216 
2217 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2218 {
2219 	struct inode *dir = d_inode(data->dir);
2220 	struct nfs_openres *o_res = &data->o_res;
2221         int status;
2222 
2223 	status = nfs4_run_open_task(data, 1);
2224 	if (status != 0 || !data->rpc_done)
2225 		return status;
2226 
2227 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2228 
2229 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2230 		status = _nfs4_proc_open_confirm(data);
2231 		if (status != 0)
2232 			return status;
2233 	}
2234 
2235 	return status;
2236 }
2237 
2238 /*
2239  * Additional permission checks in order to distinguish between an
2240  * open for read, and an open for execute. This works around the
2241  * fact that NFSv4 OPEN treats read and execute permissions as being
2242  * the same.
2243  * Note that in the non-execute case, we want to turn off permission
2244  * checking if we just created a new file (POSIX open() semantics).
2245  */
2246 static int nfs4_opendata_access(struct rpc_cred *cred,
2247 				struct nfs4_opendata *opendata,
2248 				struct nfs4_state *state, fmode_t fmode,
2249 				int openflags)
2250 {
2251 	struct nfs_access_entry cache;
2252 	u32 mask;
2253 
2254 	/* access call failed or for some reason the server doesn't
2255 	 * support any access modes -- defer access call until later */
2256 	if (opendata->o_res.access_supported == 0)
2257 		return 0;
2258 
2259 	mask = 0;
2260 	/*
2261 	 * Use openflags to check for exec, because fmode won't
2262 	 * always have FMODE_EXEC set when file open for exec.
2263 	 */
2264 	if (openflags & __FMODE_EXEC) {
2265 		/* ONLY check for exec rights */
2266 		mask = MAY_EXEC;
2267 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2268 		mask = MAY_READ;
2269 
2270 	cache.cred = cred;
2271 	cache.jiffies = jiffies;
2272 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2273 	nfs_access_add_cache(state->inode, &cache);
2274 
2275 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2276 		return 0;
2277 
2278 	/* even though OPEN succeeded, access is denied. Close the file */
2279 	nfs4_close_state(state, fmode);
2280 	return -EACCES;
2281 }
2282 
2283 /*
2284  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2285  */
2286 static int _nfs4_proc_open(struct nfs4_opendata *data)
2287 {
2288 	struct inode *dir = d_inode(data->dir);
2289 	struct nfs_server *server = NFS_SERVER(dir);
2290 	struct nfs_openargs *o_arg = &data->o_arg;
2291 	struct nfs_openres *o_res = &data->o_res;
2292 	int status;
2293 
2294 	status = nfs4_run_open_task(data, 0);
2295 	if (!data->rpc_done)
2296 		return status;
2297 	if (status != 0) {
2298 		if (status == -NFS4ERR_BADNAME &&
2299 				!(o_arg->open_flags & O_CREAT))
2300 			return -ENOENT;
2301 		return status;
2302 	}
2303 
2304 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2305 
2306 	if (o_arg->open_flags & O_CREAT) {
2307 		update_changeattr(dir, &o_res->cinfo);
2308 		if (o_arg->open_flags & O_EXCL)
2309 			data->file_created = 1;
2310 		else if (o_res->cinfo.before != o_res->cinfo.after)
2311 			data->file_created = 1;
2312 	}
2313 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2314 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2315 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2316 		status = _nfs4_proc_open_confirm(data);
2317 		if (status != 0)
2318 			return status;
2319 	}
2320 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2321 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2322 	return 0;
2323 }
2324 
2325 static int nfs4_recover_expired_lease(struct nfs_server *server)
2326 {
2327 	return nfs4_client_recover_expired_lease(server->nfs_client);
2328 }
2329 
2330 /*
2331  * OPEN_EXPIRED:
2332  * 	reclaim state on the server after a network partition.
2333  * 	Assumes caller holds the appropriate lock
2334  */
2335 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2336 {
2337 	struct nfs4_opendata *opendata;
2338 	int ret;
2339 
2340 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2341 			NFS4_OPEN_CLAIM_FH);
2342 	if (IS_ERR(opendata))
2343 		return PTR_ERR(opendata);
2344 	ret = nfs4_open_recover(opendata, state);
2345 	if (ret == -ESTALE)
2346 		d_drop(ctx->dentry);
2347 	nfs4_opendata_put(opendata);
2348 	return ret;
2349 }
2350 
2351 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2352 {
2353 	struct nfs_server *server = NFS_SERVER(state->inode);
2354 	struct nfs4_exception exception = { };
2355 	int err;
2356 
2357 	do {
2358 		err = _nfs4_open_expired(ctx, state);
2359 		trace_nfs4_open_expired(ctx, 0, err);
2360 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2361 			continue;
2362 		switch (err) {
2363 		default:
2364 			goto out;
2365 		case -NFS4ERR_GRACE:
2366 		case -NFS4ERR_DELAY:
2367 			nfs4_handle_exception(server, err, &exception);
2368 			err = 0;
2369 		}
2370 	} while (exception.retry);
2371 out:
2372 	return err;
2373 }
2374 
2375 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2376 {
2377 	struct nfs_open_context *ctx;
2378 	int ret;
2379 
2380 	ctx = nfs4_state_find_open_context(state);
2381 	if (IS_ERR(ctx))
2382 		return -EAGAIN;
2383 	ret = nfs4_do_open_expired(ctx, state);
2384 	put_nfs_open_context(ctx);
2385 	return ret;
2386 }
2387 
2388 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2389 		const nfs4_stateid *stateid)
2390 {
2391 	nfs_remove_bad_delegation(state->inode, stateid);
2392 	write_seqlock(&state->seqlock);
2393 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2394 	write_sequnlock(&state->seqlock);
2395 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2396 }
2397 
2398 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2399 {
2400 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2401 		nfs_finish_clear_delegation_stateid(state, NULL);
2402 }
2403 
2404 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2405 {
2406 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2407 	nfs40_clear_delegation_stateid(state);
2408 	return nfs4_open_expired(sp, state);
2409 }
2410 
2411 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2412 		nfs4_stateid *stateid,
2413 		struct rpc_cred *cred)
2414 {
2415 	return -NFS4ERR_BAD_STATEID;
2416 }
2417 
2418 #if defined(CONFIG_NFS_V4_1)
2419 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2420 		nfs4_stateid *stateid,
2421 		struct rpc_cred *cred)
2422 {
2423 	int status;
2424 
2425 	status = nfs41_test_stateid(server, stateid, cred);
2426 
2427 	switch (status) {
2428 	case -NFS4ERR_EXPIRED:
2429 	case -NFS4ERR_ADMIN_REVOKED:
2430 	case -NFS4ERR_DELEG_REVOKED:
2431 		/* Ack the revoked state to the server */
2432 		nfs41_free_stateid(server, stateid, cred);
2433 	case -NFS4ERR_BAD_STATEID:
2434 		return status;
2435 	}
2436 	return NFS_OK;
2437 }
2438 
2439 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2440 {
2441 	struct nfs_server *server = NFS_SERVER(state->inode);
2442 	nfs4_stateid stateid;
2443 	struct nfs_delegation *delegation;
2444 	struct rpc_cred *cred;
2445 	int status;
2446 
2447 	/* Get the delegation credential for use by test/free_stateid */
2448 	rcu_read_lock();
2449 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2450 	if (delegation == NULL) {
2451 		rcu_read_unlock();
2452 		return;
2453 	}
2454 
2455 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2456 	if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2457 		rcu_read_unlock();
2458 		nfs_finish_clear_delegation_stateid(state, &stateid);
2459 		return;
2460 	}
2461 
2462 	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags)) {
2463 		rcu_read_unlock();
2464 		return;
2465 	}
2466 
2467 	cred = get_rpccred(delegation->cred);
2468 	rcu_read_unlock();
2469 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2470 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2471 	if (status != NFS_OK)
2472 		nfs_finish_clear_delegation_stateid(state, &stateid);
2473 
2474 	put_rpccred(cred);
2475 }
2476 
2477 /**
2478  * nfs41_check_open_stateid - possibly free an open stateid
2479  *
2480  * @state: NFSv4 state for an inode
2481  *
2482  * Returns NFS_OK if recovery for this stateid is now finished.
2483  * Otherwise a negative NFS4ERR value is returned.
2484  */
2485 static int nfs41_check_open_stateid(struct nfs4_state *state)
2486 {
2487 	struct nfs_server *server = NFS_SERVER(state->inode);
2488 	nfs4_stateid *stateid = &state->open_stateid;
2489 	struct rpc_cred *cred = state->owner->so_cred;
2490 	int status;
2491 
2492 	/* If a state reset has been done, test_stateid is unneeded */
2493 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2494 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2495 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2496 		return -NFS4ERR_BAD_STATEID;
2497 
2498 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2499 	trace_nfs4_test_open_stateid(state, NULL, status);
2500 	if (status != NFS_OK) {
2501 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2502 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2503 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2504 		clear_bit(NFS_OPEN_STATE, &state->flags);
2505 	}
2506 	return status;
2507 }
2508 
2509 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2510 {
2511 	int status;
2512 
2513 	nfs41_check_delegation_stateid(state);
2514 	status = nfs41_check_open_stateid(state);
2515 	if (status != NFS_OK)
2516 		status = nfs4_open_expired(sp, state);
2517 	return status;
2518 }
2519 #endif
2520 
2521 /*
2522  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2523  * fields corresponding to attributes that were used to store the verifier.
2524  * Make sure we clobber those fields in the later setattr call
2525  */
2526 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2527 				struct iattr *sattr, struct nfs4_label **label)
2528 {
2529 	const u32 *attrset = opendata->o_res.attrset;
2530 
2531 	if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2532 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2533 		sattr->ia_valid |= ATTR_ATIME;
2534 
2535 	if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2536 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2537 		sattr->ia_valid |= ATTR_MTIME;
2538 
2539 	/* Except MODE, it seems harmless of setting twice. */
2540 	if ((attrset[1] & FATTR4_WORD1_MODE))
2541 		sattr->ia_valid &= ~ATTR_MODE;
2542 
2543 	if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2544 		*label = NULL;
2545 }
2546 
2547 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2548 		fmode_t fmode,
2549 		int flags,
2550 		struct nfs_open_context *ctx)
2551 {
2552 	struct nfs4_state_owner *sp = opendata->owner;
2553 	struct nfs_server *server = sp->so_server;
2554 	struct dentry *dentry;
2555 	struct nfs4_state *state;
2556 	unsigned int seq;
2557 	int ret;
2558 
2559 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2560 
2561 	ret = _nfs4_proc_open(opendata);
2562 	if (ret != 0)
2563 		goto out;
2564 
2565 	state = nfs4_opendata_to_nfs4_state(opendata);
2566 	ret = PTR_ERR(state);
2567 	if (IS_ERR(state))
2568 		goto out;
2569 	if (server->caps & NFS_CAP_POSIX_LOCK)
2570 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2571 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2572 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2573 
2574 	dentry = opendata->dentry;
2575 	if (d_really_is_negative(dentry)) {
2576 		struct dentry *alias;
2577 		d_drop(dentry);
2578 		alias = d_exact_alias(dentry, state->inode);
2579 		if (!alias)
2580 			alias = d_splice_alias(igrab(state->inode), dentry);
2581 		/* d_splice_alias() can't fail here - it's a non-directory */
2582 		if (alias) {
2583 			dput(ctx->dentry);
2584 			ctx->dentry = dentry = alias;
2585 		}
2586 		nfs_set_verifier(dentry,
2587 				nfs_save_change_attribute(d_inode(opendata->dir)));
2588 	}
2589 
2590 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2591 	if (ret != 0)
2592 		goto out;
2593 
2594 	ctx->state = state;
2595 	if (d_inode(dentry) == state->inode) {
2596 		nfs_inode_attach_open_context(ctx);
2597 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2598 			nfs4_schedule_stateid_recovery(server, state);
2599 	}
2600 out:
2601 	return ret;
2602 }
2603 
2604 /*
2605  * Returns a referenced nfs4_state
2606  */
2607 static int _nfs4_do_open(struct inode *dir,
2608 			struct nfs_open_context *ctx,
2609 			int flags,
2610 			struct iattr *sattr,
2611 			struct nfs4_label *label,
2612 			int *opened)
2613 {
2614 	struct nfs4_state_owner  *sp;
2615 	struct nfs4_state     *state = NULL;
2616 	struct nfs_server       *server = NFS_SERVER(dir);
2617 	struct nfs4_opendata *opendata;
2618 	struct dentry *dentry = ctx->dentry;
2619 	struct rpc_cred *cred = ctx->cred;
2620 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2621 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2622 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2623 	struct nfs4_label *olabel = NULL;
2624 	int status;
2625 
2626 	/* Protect against reboot recovery conflicts */
2627 	status = -ENOMEM;
2628 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2629 	if (sp == NULL) {
2630 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2631 		goto out_err;
2632 	}
2633 	status = nfs4_recover_expired_lease(server);
2634 	if (status != 0)
2635 		goto err_put_state_owner;
2636 	if (d_really_is_positive(dentry))
2637 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2638 	status = -ENOMEM;
2639 	if (d_really_is_positive(dentry))
2640 		claim = NFS4_OPEN_CLAIM_FH;
2641 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2642 			label, claim, GFP_KERNEL);
2643 	if (opendata == NULL)
2644 		goto err_put_state_owner;
2645 
2646 	if (label) {
2647 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2648 		if (IS_ERR(olabel)) {
2649 			status = PTR_ERR(olabel);
2650 			goto err_opendata_put;
2651 		}
2652 	}
2653 
2654 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2655 		if (!opendata->f_attr.mdsthreshold) {
2656 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2657 			if (!opendata->f_attr.mdsthreshold)
2658 				goto err_free_label;
2659 		}
2660 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2661 	}
2662 	if (d_really_is_positive(dentry))
2663 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2664 
2665 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2666 	if (status != 0)
2667 		goto err_free_label;
2668 	state = ctx->state;
2669 
2670 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2671 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2672 		nfs4_exclusive_attrset(opendata, sattr, &label);
2673 		/*
2674 		 * send create attributes which was not set by open
2675 		 * with an extra setattr.
2676 		 */
2677 		if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2678 			nfs_fattr_init(opendata->o_res.f_attr);
2679 			status = nfs4_do_setattr(state->inode, cred,
2680 					opendata->o_res.f_attr, sattr,
2681 					state, label, olabel);
2682 			if (status == 0) {
2683 				nfs_setattr_update_inode(state->inode, sattr,
2684 						opendata->o_res.f_attr);
2685 				nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2686 			}
2687 		}
2688 	}
2689 	if (opened && opendata->file_created)
2690 		*opened |= FILE_CREATED;
2691 
2692 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2693 		*ctx_th = opendata->f_attr.mdsthreshold;
2694 		opendata->f_attr.mdsthreshold = NULL;
2695 	}
2696 
2697 	nfs4_label_free(olabel);
2698 
2699 	nfs4_opendata_put(opendata);
2700 	nfs4_put_state_owner(sp);
2701 	return 0;
2702 err_free_label:
2703 	nfs4_label_free(olabel);
2704 err_opendata_put:
2705 	nfs4_opendata_put(opendata);
2706 err_put_state_owner:
2707 	nfs4_put_state_owner(sp);
2708 out_err:
2709 	return status;
2710 }
2711 
2712 
2713 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2714 					struct nfs_open_context *ctx,
2715 					int flags,
2716 					struct iattr *sattr,
2717 					struct nfs4_label *label,
2718 					int *opened)
2719 {
2720 	struct nfs_server *server = NFS_SERVER(dir);
2721 	struct nfs4_exception exception = { };
2722 	struct nfs4_state *res;
2723 	int status;
2724 
2725 	do {
2726 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2727 		res = ctx->state;
2728 		trace_nfs4_open_file(ctx, flags, status);
2729 		if (status == 0)
2730 			break;
2731 		/* NOTE: BAD_SEQID means the server and client disagree about the
2732 		 * book-keeping w.r.t. state-changing operations
2733 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2734 		 * It is actually a sign of a bug on the client or on the server.
2735 		 *
2736 		 * If we receive a BAD_SEQID error in the particular case of
2737 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2738 		 * have unhashed the old state_owner for us, and that we can
2739 		 * therefore safely retry using a new one. We should still warn
2740 		 * the user though...
2741 		 */
2742 		if (status == -NFS4ERR_BAD_SEQID) {
2743 			pr_warn_ratelimited("NFS: v4 server %s "
2744 					" returned a bad sequence-id error!\n",
2745 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2746 			exception.retry = 1;
2747 			continue;
2748 		}
2749 		/*
2750 		 * BAD_STATEID on OPEN means that the server cancelled our
2751 		 * state before it received the OPEN_CONFIRM.
2752 		 * Recover by retrying the request as per the discussion
2753 		 * on Page 181 of RFC3530.
2754 		 */
2755 		if (status == -NFS4ERR_BAD_STATEID) {
2756 			exception.retry = 1;
2757 			continue;
2758 		}
2759 		if (status == -EAGAIN) {
2760 			/* We must have found a delegation */
2761 			exception.retry = 1;
2762 			continue;
2763 		}
2764 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2765 			continue;
2766 		res = ERR_PTR(nfs4_handle_exception(server,
2767 					status, &exception));
2768 	} while (exception.retry);
2769 	return res;
2770 }
2771 
2772 static int _nfs4_do_setattr(struct inode *inode,
2773 			    struct nfs_setattrargs *arg,
2774 			    struct nfs_setattrres *res,
2775 			    struct rpc_cred *cred,
2776 			    struct nfs4_state *state)
2777 {
2778 	struct nfs_server *server = NFS_SERVER(inode);
2779         struct rpc_message msg = {
2780 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2781 		.rpc_argp	= arg,
2782 		.rpc_resp	= res,
2783 		.rpc_cred	= cred,
2784         };
2785 	struct rpc_cred *delegation_cred = NULL;
2786 	unsigned long timestamp = jiffies;
2787 	fmode_t fmode;
2788 	bool truncate;
2789 	int status;
2790 
2791 	nfs_fattr_init(res->fattr);
2792 
2793 	/* Servers should only apply open mode checks for file size changes */
2794 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2795 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2796 
2797 	if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2798 		/* Use that stateid */
2799 	} else if (truncate && state != NULL) {
2800 		struct nfs_lockowner lockowner = {
2801 			.l_owner = current->files,
2802 			.l_pid = current->tgid,
2803 		};
2804 		if (!nfs4_valid_open_stateid(state))
2805 			return -EBADF;
2806 		if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2807 				&arg->stateid, &delegation_cred) == -EIO)
2808 			return -EBADF;
2809 	} else
2810 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2811 	if (delegation_cred)
2812 		msg.rpc_cred = delegation_cred;
2813 
2814 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2815 
2816 	put_rpccred(delegation_cred);
2817 	if (status == 0 && state != NULL)
2818 		renew_lease(server, timestamp);
2819 	trace_nfs4_setattr(inode, &arg->stateid, status);
2820 	return status;
2821 }
2822 
2823 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2824 			   struct nfs_fattr *fattr, struct iattr *sattr,
2825 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2826 			   struct nfs4_label *olabel)
2827 {
2828 	struct nfs_server *server = NFS_SERVER(inode);
2829         struct nfs_setattrargs  arg = {
2830                 .fh             = NFS_FH(inode),
2831                 .iap            = sattr,
2832 		.server		= server,
2833 		.bitmask = server->attr_bitmask,
2834 		.label		= ilabel,
2835         };
2836         struct nfs_setattrres  res = {
2837 		.fattr		= fattr,
2838 		.label		= olabel,
2839 		.server		= server,
2840         };
2841 	struct nfs4_exception exception = {
2842 		.state = state,
2843 		.inode = inode,
2844 		.stateid = &arg.stateid,
2845 	};
2846 	int err;
2847 
2848 	arg.bitmask = nfs4_bitmask(server, ilabel);
2849 	if (ilabel)
2850 		arg.bitmask = nfs4_bitmask(server, olabel);
2851 
2852 	do {
2853 		err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
2854 		switch (err) {
2855 		case -NFS4ERR_OPENMODE:
2856 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2857 				pr_warn_once("NFSv4: server %s is incorrectly "
2858 						"applying open mode checks to "
2859 						"a SETATTR that is not "
2860 						"changing file size.\n",
2861 						server->nfs_client->cl_hostname);
2862 			}
2863 			if (state && !(state->state & FMODE_WRITE)) {
2864 				err = -EBADF;
2865 				if (sattr->ia_valid & ATTR_OPEN)
2866 					err = -EACCES;
2867 				goto out;
2868 			}
2869 		}
2870 		err = nfs4_handle_exception(server, err, &exception);
2871 	} while (exception.retry);
2872 out:
2873 	return err;
2874 }
2875 
2876 static bool
2877 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2878 {
2879 	if (inode == NULL || !nfs_have_layout(inode))
2880 		return false;
2881 
2882 	return pnfs_wait_on_layoutreturn(inode, task);
2883 }
2884 
2885 struct nfs4_closedata {
2886 	struct inode *inode;
2887 	struct nfs4_state *state;
2888 	struct nfs_closeargs arg;
2889 	struct nfs_closeres res;
2890 	struct nfs_fattr fattr;
2891 	unsigned long timestamp;
2892 	bool roc;
2893 	u32 roc_barrier;
2894 };
2895 
2896 static void nfs4_free_closedata(void *data)
2897 {
2898 	struct nfs4_closedata *calldata = data;
2899 	struct nfs4_state_owner *sp = calldata->state->owner;
2900 	struct super_block *sb = calldata->state->inode->i_sb;
2901 
2902 	if (calldata->roc)
2903 		pnfs_roc_release(calldata->state->inode);
2904 	nfs4_put_open_state(calldata->state);
2905 	nfs_free_seqid(calldata->arg.seqid);
2906 	nfs4_put_state_owner(sp);
2907 	nfs_sb_deactive(sb);
2908 	kfree(calldata);
2909 }
2910 
2911 static void nfs4_close_done(struct rpc_task *task, void *data)
2912 {
2913 	struct nfs4_closedata *calldata = data;
2914 	struct nfs4_state *state = calldata->state;
2915 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2916 	nfs4_stateid *res_stateid = NULL;
2917 
2918 	dprintk("%s: begin!\n", __func__);
2919 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2920 		return;
2921 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2922         /* hmm. we are done with the inode, and in the process of freeing
2923 	 * the state_owner. we keep this around to process errors
2924 	 */
2925 	switch (task->tk_status) {
2926 		case 0:
2927 			res_stateid = &calldata->res.stateid;
2928 			if (calldata->roc)
2929 				pnfs_roc_set_barrier(state->inode,
2930 						     calldata->roc_barrier);
2931 			renew_lease(server, calldata->timestamp);
2932 			break;
2933 		case -NFS4ERR_ADMIN_REVOKED:
2934 		case -NFS4ERR_STALE_STATEID:
2935 		case -NFS4ERR_OLD_STATEID:
2936 		case -NFS4ERR_BAD_STATEID:
2937 		case -NFS4ERR_EXPIRED:
2938 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2939 						&state->open_stateid)) {
2940 				rpc_restart_call_prepare(task);
2941 				goto out_release;
2942 			}
2943 			if (calldata->arg.fmode == 0)
2944 				break;
2945 		default:
2946 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2947 				rpc_restart_call_prepare(task);
2948 				goto out_release;
2949 			}
2950 	}
2951 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
2952 			res_stateid, calldata->arg.fmode);
2953 out_release:
2954 	nfs_release_seqid(calldata->arg.seqid);
2955 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2956 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2957 }
2958 
2959 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2960 {
2961 	struct nfs4_closedata *calldata = data;
2962 	struct nfs4_state *state = calldata->state;
2963 	struct inode *inode = calldata->inode;
2964 	bool is_rdonly, is_wronly, is_rdwr;
2965 	int call_close = 0;
2966 
2967 	dprintk("%s: begin!\n", __func__);
2968 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2969 		goto out_wait;
2970 
2971 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2972 	spin_lock(&state->owner->so_lock);
2973 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2974 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2975 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2976 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2977 	/* Calculate the change in open mode */
2978 	calldata->arg.fmode = 0;
2979 	if (state->n_rdwr == 0) {
2980 		if (state->n_rdonly == 0)
2981 			call_close |= is_rdonly;
2982 		else if (is_rdonly)
2983 			calldata->arg.fmode |= FMODE_READ;
2984 		if (state->n_wronly == 0)
2985 			call_close |= is_wronly;
2986 		else if (is_wronly)
2987 			calldata->arg.fmode |= FMODE_WRITE;
2988 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2989 			call_close |= is_rdwr;
2990 	} else if (is_rdwr)
2991 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2992 
2993 	if (!nfs4_valid_open_stateid(state))
2994 		call_close = 0;
2995 	spin_unlock(&state->owner->so_lock);
2996 
2997 	if (!call_close) {
2998 		/* Note: exit _without_ calling nfs4_close_done */
2999 		goto out_no_action;
3000 	}
3001 
3002 	if (nfs4_wait_on_layoutreturn(inode, task)) {
3003 		nfs_release_seqid(calldata->arg.seqid);
3004 		goto out_wait;
3005 	}
3006 
3007 	if (calldata->arg.fmode == 0)
3008 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3009 	if (calldata->roc)
3010 		pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
3011 
3012 	calldata->arg.share_access =
3013 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3014 				calldata->arg.fmode, 0);
3015 
3016 	nfs_fattr_init(calldata->res.fattr);
3017 	calldata->timestamp = jiffies;
3018 	if (nfs4_setup_sequence(NFS_SERVER(inode),
3019 				&calldata->arg.seq_args,
3020 				&calldata->res.seq_res,
3021 				task) != 0)
3022 		nfs_release_seqid(calldata->arg.seqid);
3023 	dprintk("%s: done!\n", __func__);
3024 	return;
3025 out_no_action:
3026 	task->tk_action = NULL;
3027 out_wait:
3028 	nfs4_sequence_done(task, &calldata->res.seq_res);
3029 }
3030 
3031 static const struct rpc_call_ops nfs4_close_ops = {
3032 	.rpc_call_prepare = nfs4_close_prepare,
3033 	.rpc_call_done = nfs4_close_done,
3034 	.rpc_release = nfs4_free_closedata,
3035 };
3036 
3037 static bool nfs4_roc(struct inode *inode)
3038 {
3039 	if (!nfs_have_layout(inode))
3040 		return false;
3041 	return pnfs_roc(inode);
3042 }
3043 
3044 /*
3045  * It is possible for data to be read/written from a mem-mapped file
3046  * after the sys_close call (which hits the vfs layer as a flush).
3047  * This means that we can't safely call nfsv4 close on a file until
3048  * the inode is cleared. This in turn means that we are not good
3049  * NFSv4 citizens - we do not indicate to the server to update the file's
3050  * share state even when we are done with one of the three share
3051  * stateid's in the inode.
3052  *
3053  * NOTE: Caller must be holding the sp->so_owner semaphore!
3054  */
3055 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3056 {
3057 	struct nfs_server *server = NFS_SERVER(state->inode);
3058 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3059 	struct nfs4_closedata *calldata;
3060 	struct nfs4_state_owner *sp = state->owner;
3061 	struct rpc_task *task;
3062 	struct rpc_message msg = {
3063 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3064 		.rpc_cred = state->owner->so_cred,
3065 	};
3066 	struct rpc_task_setup task_setup_data = {
3067 		.rpc_client = server->client,
3068 		.rpc_message = &msg,
3069 		.callback_ops = &nfs4_close_ops,
3070 		.workqueue = nfsiod_workqueue,
3071 		.flags = RPC_TASK_ASYNC,
3072 	};
3073 	int status = -ENOMEM;
3074 
3075 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3076 		&task_setup_data.rpc_client, &msg);
3077 
3078 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3079 	if (calldata == NULL)
3080 		goto out;
3081 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3082 	calldata->inode = state->inode;
3083 	calldata->state = state;
3084 	calldata->arg.fh = NFS_FH(state->inode);
3085 	/* Serialization for the sequence id */
3086 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3087 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3088 	if (IS_ERR(calldata->arg.seqid))
3089 		goto out_free_calldata;
3090 	calldata->arg.fmode = 0;
3091 	calldata->arg.bitmask = server->cache_consistency_bitmask;
3092 	calldata->res.fattr = &calldata->fattr;
3093 	calldata->res.seqid = calldata->arg.seqid;
3094 	calldata->res.server = server;
3095 	calldata->roc = nfs4_roc(state->inode);
3096 	nfs_sb_active(calldata->inode->i_sb);
3097 
3098 	msg.rpc_argp = &calldata->arg;
3099 	msg.rpc_resp = &calldata->res;
3100 	task_setup_data.callback_data = calldata;
3101 	task = rpc_run_task(&task_setup_data);
3102 	if (IS_ERR(task))
3103 		return PTR_ERR(task);
3104 	status = 0;
3105 	if (wait)
3106 		status = rpc_wait_for_completion_task(task);
3107 	rpc_put_task(task);
3108 	return status;
3109 out_free_calldata:
3110 	kfree(calldata);
3111 out:
3112 	nfs4_put_open_state(state);
3113 	nfs4_put_state_owner(sp);
3114 	return status;
3115 }
3116 
3117 static struct inode *
3118 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3119 		int open_flags, struct iattr *attr, int *opened)
3120 {
3121 	struct nfs4_state *state;
3122 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3123 
3124 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3125 
3126 	/* Protect against concurrent sillydeletes */
3127 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3128 
3129 	nfs4_label_release_security(label);
3130 
3131 	if (IS_ERR(state))
3132 		return ERR_CAST(state);
3133 	return state->inode;
3134 }
3135 
3136 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3137 {
3138 	if (ctx->state == NULL)
3139 		return;
3140 	if (is_sync)
3141 		nfs4_close_sync(ctx->state, ctx->mode);
3142 	else
3143 		nfs4_close_state(ctx->state, ctx->mode);
3144 }
3145 
3146 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3147 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3148 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3149 
3150 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3151 {
3152 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3153 	struct nfs4_server_caps_arg args = {
3154 		.fhandle = fhandle,
3155 		.bitmask = bitmask,
3156 	};
3157 	struct nfs4_server_caps_res res = {};
3158 	struct rpc_message msg = {
3159 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3160 		.rpc_argp = &args,
3161 		.rpc_resp = &res,
3162 	};
3163 	int status;
3164 
3165 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3166 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3167 		     FATTR4_WORD0_LINK_SUPPORT |
3168 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3169 		     FATTR4_WORD0_ACLSUPPORT;
3170 	if (minorversion)
3171 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3172 
3173 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3174 	if (status == 0) {
3175 		/* Sanity check the server answers */
3176 		switch (minorversion) {
3177 		case 0:
3178 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3179 			res.attr_bitmask[2] = 0;
3180 			break;
3181 		case 1:
3182 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3183 			break;
3184 		case 2:
3185 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3186 		}
3187 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3188 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3189 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3190 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3191 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3192 				NFS_CAP_CTIME|NFS_CAP_MTIME|
3193 				NFS_CAP_SECURITY_LABEL);
3194 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3195 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3196 			server->caps |= NFS_CAP_ACLS;
3197 		if (res.has_links != 0)
3198 			server->caps |= NFS_CAP_HARDLINKS;
3199 		if (res.has_symlinks != 0)
3200 			server->caps |= NFS_CAP_SYMLINKS;
3201 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3202 			server->caps |= NFS_CAP_FILEID;
3203 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3204 			server->caps |= NFS_CAP_MODE;
3205 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3206 			server->caps |= NFS_CAP_NLINK;
3207 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3208 			server->caps |= NFS_CAP_OWNER;
3209 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3210 			server->caps |= NFS_CAP_OWNER_GROUP;
3211 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3212 			server->caps |= NFS_CAP_ATIME;
3213 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3214 			server->caps |= NFS_CAP_CTIME;
3215 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3216 			server->caps |= NFS_CAP_MTIME;
3217 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3218 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3219 			server->caps |= NFS_CAP_SECURITY_LABEL;
3220 #endif
3221 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3222 				sizeof(server->attr_bitmask));
3223 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3224 
3225 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3226 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3227 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3228 		server->cache_consistency_bitmask[2] = 0;
3229 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3230 			sizeof(server->exclcreat_bitmask));
3231 		server->acl_bitmask = res.acl_bitmask;
3232 		server->fh_expire_type = res.fh_expire_type;
3233 	}
3234 
3235 	return status;
3236 }
3237 
3238 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3239 {
3240 	struct nfs4_exception exception = { };
3241 	int err;
3242 	do {
3243 		err = nfs4_handle_exception(server,
3244 				_nfs4_server_capabilities(server, fhandle),
3245 				&exception);
3246 	} while (exception.retry);
3247 	return err;
3248 }
3249 
3250 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3251 		struct nfs_fsinfo *info)
3252 {
3253 	u32 bitmask[3];
3254 	struct nfs4_lookup_root_arg args = {
3255 		.bitmask = bitmask,
3256 	};
3257 	struct nfs4_lookup_res res = {
3258 		.server = server,
3259 		.fattr = info->fattr,
3260 		.fh = fhandle,
3261 	};
3262 	struct rpc_message msg = {
3263 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3264 		.rpc_argp = &args,
3265 		.rpc_resp = &res,
3266 	};
3267 
3268 	bitmask[0] = nfs4_fattr_bitmap[0];
3269 	bitmask[1] = nfs4_fattr_bitmap[1];
3270 	/*
3271 	 * Process the label in the upcoming getfattr
3272 	 */
3273 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3274 
3275 	nfs_fattr_init(info->fattr);
3276 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3277 }
3278 
3279 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3280 		struct nfs_fsinfo *info)
3281 {
3282 	struct nfs4_exception exception = { };
3283 	int err;
3284 	do {
3285 		err = _nfs4_lookup_root(server, fhandle, info);
3286 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3287 		switch (err) {
3288 		case 0:
3289 		case -NFS4ERR_WRONGSEC:
3290 			goto out;
3291 		default:
3292 			err = nfs4_handle_exception(server, err, &exception);
3293 		}
3294 	} while (exception.retry);
3295 out:
3296 	return err;
3297 }
3298 
3299 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3300 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3301 {
3302 	struct rpc_auth_create_args auth_args = {
3303 		.pseudoflavor = flavor,
3304 	};
3305 	struct rpc_auth *auth;
3306 	int ret;
3307 
3308 	auth = rpcauth_create(&auth_args, server->client);
3309 	if (IS_ERR(auth)) {
3310 		ret = -EACCES;
3311 		goto out;
3312 	}
3313 	ret = nfs4_lookup_root(server, fhandle, info);
3314 out:
3315 	return ret;
3316 }
3317 
3318 /*
3319  * Retry pseudoroot lookup with various security flavors.  We do this when:
3320  *
3321  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3322  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3323  *
3324  * Returns zero on success, or a negative NFS4ERR value, or a
3325  * negative errno value.
3326  */
3327 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3328 			      struct nfs_fsinfo *info)
3329 {
3330 	/* Per 3530bis 15.33.5 */
3331 	static const rpc_authflavor_t flav_array[] = {
3332 		RPC_AUTH_GSS_KRB5P,
3333 		RPC_AUTH_GSS_KRB5I,
3334 		RPC_AUTH_GSS_KRB5,
3335 		RPC_AUTH_UNIX,			/* courtesy */
3336 		RPC_AUTH_NULL,
3337 	};
3338 	int status = -EPERM;
3339 	size_t i;
3340 
3341 	if (server->auth_info.flavor_len > 0) {
3342 		/* try each flavor specified by user */
3343 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3344 			status = nfs4_lookup_root_sec(server, fhandle, info,
3345 						server->auth_info.flavors[i]);
3346 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3347 				continue;
3348 			break;
3349 		}
3350 	} else {
3351 		/* no flavors specified by user, try default list */
3352 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3353 			status = nfs4_lookup_root_sec(server, fhandle, info,
3354 						      flav_array[i]);
3355 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3356 				continue;
3357 			break;
3358 		}
3359 	}
3360 
3361 	/*
3362 	 * -EACCESS could mean that the user doesn't have correct permissions
3363 	 * to access the mount.  It could also mean that we tried to mount
3364 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3365 	 * existing mount programs don't handle -EACCES very well so it should
3366 	 * be mapped to -EPERM instead.
3367 	 */
3368 	if (status == -EACCES)
3369 		status = -EPERM;
3370 	return status;
3371 }
3372 
3373 /**
3374  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3375  * @server: initialized nfs_server handle
3376  * @fhandle: we fill in the pseudo-fs root file handle
3377  * @info: we fill in an FSINFO struct
3378  * @auth_probe: probe the auth flavours
3379  *
3380  * Returns zero on success, or a negative errno.
3381  */
3382 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3383 			 struct nfs_fsinfo *info,
3384 			 bool auth_probe)
3385 {
3386 	int status = 0;
3387 
3388 	if (!auth_probe)
3389 		status = nfs4_lookup_root(server, fhandle, info);
3390 
3391 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3392 		status = server->nfs_client->cl_mvops->find_root_sec(server,
3393 				fhandle, info);
3394 
3395 	if (status == 0)
3396 		status = nfs4_server_capabilities(server, fhandle);
3397 	if (status == 0)
3398 		status = nfs4_do_fsinfo(server, fhandle, info);
3399 
3400 	return nfs4_map_errors(status);
3401 }
3402 
3403 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3404 			      struct nfs_fsinfo *info)
3405 {
3406 	int error;
3407 	struct nfs_fattr *fattr = info->fattr;
3408 	struct nfs4_label *label = NULL;
3409 
3410 	error = nfs4_server_capabilities(server, mntfh);
3411 	if (error < 0) {
3412 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3413 		return error;
3414 	}
3415 
3416 	label = nfs4_label_alloc(server, GFP_KERNEL);
3417 	if (IS_ERR(label))
3418 		return PTR_ERR(label);
3419 
3420 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3421 	if (error < 0) {
3422 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3423 		goto err_free_label;
3424 	}
3425 
3426 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3427 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3428 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3429 
3430 err_free_label:
3431 	nfs4_label_free(label);
3432 
3433 	return error;
3434 }
3435 
3436 /*
3437  * Get locations and (maybe) other attributes of a referral.
3438  * Note that we'll actually follow the referral later when
3439  * we detect fsid mismatch in inode revalidation
3440  */
3441 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3442 			     const struct qstr *name, struct nfs_fattr *fattr,
3443 			     struct nfs_fh *fhandle)
3444 {
3445 	int status = -ENOMEM;
3446 	struct page *page = NULL;
3447 	struct nfs4_fs_locations *locations = NULL;
3448 
3449 	page = alloc_page(GFP_KERNEL);
3450 	if (page == NULL)
3451 		goto out;
3452 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3453 	if (locations == NULL)
3454 		goto out;
3455 
3456 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3457 	if (status != 0)
3458 		goto out;
3459 
3460 	/*
3461 	 * If the fsid didn't change, this is a migration event, not a
3462 	 * referral.  Cause us to drop into the exception handler, which
3463 	 * will kick off migration recovery.
3464 	 */
3465 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3466 		dprintk("%s: server did not return a different fsid for"
3467 			" a referral at %s\n", __func__, name->name);
3468 		status = -NFS4ERR_MOVED;
3469 		goto out;
3470 	}
3471 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3472 	nfs_fixup_referral_attributes(&locations->fattr);
3473 
3474 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3475 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3476 	memset(fhandle, 0, sizeof(struct nfs_fh));
3477 out:
3478 	if (page)
3479 		__free_page(page);
3480 	kfree(locations);
3481 	return status;
3482 }
3483 
3484 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3485 				struct nfs_fattr *fattr, struct nfs4_label *label)
3486 {
3487 	struct nfs4_getattr_arg args = {
3488 		.fh = fhandle,
3489 		.bitmask = server->attr_bitmask,
3490 	};
3491 	struct nfs4_getattr_res res = {
3492 		.fattr = fattr,
3493 		.label = label,
3494 		.server = server,
3495 	};
3496 	struct rpc_message msg = {
3497 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3498 		.rpc_argp = &args,
3499 		.rpc_resp = &res,
3500 	};
3501 
3502 	args.bitmask = nfs4_bitmask(server, label);
3503 
3504 	nfs_fattr_init(fattr);
3505 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3506 }
3507 
3508 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3509 				struct nfs_fattr *fattr, struct nfs4_label *label)
3510 {
3511 	struct nfs4_exception exception = { };
3512 	int err;
3513 	do {
3514 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3515 		trace_nfs4_getattr(server, fhandle, fattr, err);
3516 		err = nfs4_handle_exception(server, err,
3517 				&exception);
3518 	} while (exception.retry);
3519 	return err;
3520 }
3521 
3522 /*
3523  * The file is not closed if it is opened due to the a request to change
3524  * the size of the file. The open call will not be needed once the
3525  * VFS layer lookup-intents are implemented.
3526  *
3527  * Close is called when the inode is destroyed.
3528  * If we haven't opened the file for O_WRONLY, we
3529  * need to in the size_change case to obtain a stateid.
3530  *
3531  * Got race?
3532  * Because OPEN is always done by name in nfsv4, it is
3533  * possible that we opened a different file by the same
3534  * name.  We can recognize this race condition, but we
3535  * can't do anything about it besides returning an error.
3536  *
3537  * This will be fixed with VFS changes (lookup-intent).
3538  */
3539 static int
3540 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3541 		  struct iattr *sattr)
3542 {
3543 	struct inode *inode = d_inode(dentry);
3544 	struct rpc_cred *cred = NULL;
3545 	struct nfs4_state *state = NULL;
3546 	struct nfs4_label *label = NULL;
3547 	int status;
3548 
3549 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3550 	    sattr->ia_valid & ATTR_SIZE &&
3551 	    sattr->ia_size < i_size_read(inode))
3552 		pnfs_commit_and_return_layout(inode);
3553 
3554 	nfs_fattr_init(fattr);
3555 
3556 	/* Deal with open(O_TRUNC) */
3557 	if (sattr->ia_valid & ATTR_OPEN)
3558 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3559 
3560 	/* Optimization: if the end result is no change, don't RPC */
3561 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3562 		return 0;
3563 
3564 	/* Search for an existing open(O_WRITE) file */
3565 	if (sattr->ia_valid & ATTR_FILE) {
3566 		struct nfs_open_context *ctx;
3567 
3568 		ctx = nfs_file_open_context(sattr->ia_file);
3569 		if (ctx) {
3570 			cred = ctx->cred;
3571 			state = ctx->state;
3572 		}
3573 	}
3574 
3575 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3576 	if (IS_ERR(label))
3577 		return PTR_ERR(label);
3578 
3579 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3580 	if (status == 0) {
3581 		nfs_setattr_update_inode(inode, sattr, fattr);
3582 		nfs_setsecurity(inode, fattr, label);
3583 	}
3584 	nfs4_label_free(label);
3585 	return status;
3586 }
3587 
3588 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3589 		const struct qstr *name, struct nfs_fh *fhandle,
3590 		struct nfs_fattr *fattr, struct nfs4_label *label)
3591 {
3592 	struct nfs_server *server = NFS_SERVER(dir);
3593 	int		       status;
3594 	struct nfs4_lookup_arg args = {
3595 		.bitmask = server->attr_bitmask,
3596 		.dir_fh = NFS_FH(dir),
3597 		.name = name,
3598 	};
3599 	struct nfs4_lookup_res res = {
3600 		.server = server,
3601 		.fattr = fattr,
3602 		.label = label,
3603 		.fh = fhandle,
3604 	};
3605 	struct rpc_message msg = {
3606 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3607 		.rpc_argp = &args,
3608 		.rpc_resp = &res,
3609 	};
3610 
3611 	args.bitmask = nfs4_bitmask(server, label);
3612 
3613 	nfs_fattr_init(fattr);
3614 
3615 	dprintk("NFS call  lookup %s\n", name->name);
3616 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3617 	dprintk("NFS reply lookup: %d\n", status);
3618 	return status;
3619 }
3620 
3621 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3622 {
3623 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3624 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3625 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3626 	fattr->nlink = 2;
3627 }
3628 
3629 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3630 				   const struct qstr *name, struct nfs_fh *fhandle,
3631 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3632 {
3633 	struct nfs4_exception exception = { };
3634 	struct rpc_clnt *client = *clnt;
3635 	int err;
3636 	do {
3637 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3638 		trace_nfs4_lookup(dir, name, err);
3639 		switch (err) {
3640 		case -NFS4ERR_BADNAME:
3641 			err = -ENOENT;
3642 			goto out;
3643 		case -NFS4ERR_MOVED:
3644 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3645 			if (err == -NFS4ERR_MOVED)
3646 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3647 			goto out;
3648 		case -NFS4ERR_WRONGSEC:
3649 			err = -EPERM;
3650 			if (client != *clnt)
3651 				goto out;
3652 			client = nfs4_negotiate_security(client, dir, name);
3653 			if (IS_ERR(client))
3654 				return PTR_ERR(client);
3655 
3656 			exception.retry = 1;
3657 			break;
3658 		default:
3659 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3660 		}
3661 	} while (exception.retry);
3662 
3663 out:
3664 	if (err == 0)
3665 		*clnt = client;
3666 	else if (client != *clnt)
3667 		rpc_shutdown_client(client);
3668 
3669 	return err;
3670 }
3671 
3672 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3673 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3674 			    struct nfs4_label *label)
3675 {
3676 	int status;
3677 	struct rpc_clnt *client = NFS_CLIENT(dir);
3678 
3679 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3680 	if (client != NFS_CLIENT(dir)) {
3681 		rpc_shutdown_client(client);
3682 		nfs_fixup_secinfo_attributes(fattr);
3683 	}
3684 	return status;
3685 }
3686 
3687 struct rpc_clnt *
3688 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3689 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3690 {
3691 	struct rpc_clnt *client = NFS_CLIENT(dir);
3692 	int status;
3693 
3694 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3695 	if (status < 0)
3696 		return ERR_PTR(status);
3697 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3698 }
3699 
3700 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3701 {
3702 	struct nfs_server *server = NFS_SERVER(inode);
3703 	struct nfs4_accessargs args = {
3704 		.fh = NFS_FH(inode),
3705 		.bitmask = server->cache_consistency_bitmask,
3706 	};
3707 	struct nfs4_accessres res = {
3708 		.server = server,
3709 	};
3710 	struct rpc_message msg = {
3711 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3712 		.rpc_argp = &args,
3713 		.rpc_resp = &res,
3714 		.rpc_cred = entry->cred,
3715 	};
3716 	int mode = entry->mask;
3717 	int status = 0;
3718 
3719 	/*
3720 	 * Determine which access bits we want to ask for...
3721 	 */
3722 	if (mode & MAY_READ)
3723 		args.access |= NFS4_ACCESS_READ;
3724 	if (S_ISDIR(inode->i_mode)) {
3725 		if (mode & MAY_WRITE)
3726 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3727 		if (mode & MAY_EXEC)
3728 			args.access |= NFS4_ACCESS_LOOKUP;
3729 	} else {
3730 		if (mode & MAY_WRITE)
3731 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3732 		if (mode & MAY_EXEC)
3733 			args.access |= NFS4_ACCESS_EXECUTE;
3734 	}
3735 
3736 	res.fattr = nfs_alloc_fattr();
3737 	if (res.fattr == NULL)
3738 		return -ENOMEM;
3739 
3740 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3741 	if (!status) {
3742 		nfs_access_set_mask(entry, res.access);
3743 		nfs_refresh_inode(inode, res.fattr);
3744 	}
3745 	nfs_free_fattr(res.fattr);
3746 	return status;
3747 }
3748 
3749 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3750 {
3751 	struct nfs4_exception exception = { };
3752 	int err;
3753 	do {
3754 		err = _nfs4_proc_access(inode, entry);
3755 		trace_nfs4_access(inode, err);
3756 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3757 				&exception);
3758 	} while (exception.retry);
3759 	return err;
3760 }
3761 
3762 /*
3763  * TODO: For the time being, we don't try to get any attributes
3764  * along with any of the zero-copy operations READ, READDIR,
3765  * READLINK, WRITE.
3766  *
3767  * In the case of the first three, we want to put the GETATTR
3768  * after the read-type operation -- this is because it is hard
3769  * to predict the length of a GETATTR response in v4, and thus
3770  * align the READ data correctly.  This means that the GETATTR
3771  * may end up partially falling into the page cache, and we should
3772  * shift it into the 'tail' of the xdr_buf before processing.
3773  * To do this efficiently, we need to know the total length
3774  * of data received, which doesn't seem to be available outside
3775  * of the RPC layer.
3776  *
3777  * In the case of WRITE, we also want to put the GETATTR after
3778  * the operation -- in this case because we want to make sure
3779  * we get the post-operation mtime and size.
3780  *
3781  * Both of these changes to the XDR layer would in fact be quite
3782  * minor, but I decided to leave them for a subsequent patch.
3783  */
3784 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3785 		unsigned int pgbase, unsigned int pglen)
3786 {
3787 	struct nfs4_readlink args = {
3788 		.fh       = NFS_FH(inode),
3789 		.pgbase	  = pgbase,
3790 		.pglen    = pglen,
3791 		.pages    = &page,
3792 	};
3793 	struct nfs4_readlink_res res;
3794 	struct rpc_message msg = {
3795 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3796 		.rpc_argp = &args,
3797 		.rpc_resp = &res,
3798 	};
3799 
3800 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3801 }
3802 
3803 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3804 		unsigned int pgbase, unsigned int pglen)
3805 {
3806 	struct nfs4_exception exception = { };
3807 	int err;
3808 	do {
3809 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3810 		trace_nfs4_readlink(inode, err);
3811 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3812 				&exception);
3813 	} while (exception.retry);
3814 	return err;
3815 }
3816 
3817 /*
3818  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3819  */
3820 static int
3821 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3822 		 int flags)
3823 {
3824 	struct nfs4_label l, *ilabel = NULL;
3825 	struct nfs_open_context *ctx;
3826 	struct nfs4_state *state;
3827 	int status = 0;
3828 
3829 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3830 	if (IS_ERR(ctx))
3831 		return PTR_ERR(ctx);
3832 
3833 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3834 
3835 	sattr->ia_mode &= ~current_umask();
3836 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3837 	if (IS_ERR(state)) {
3838 		status = PTR_ERR(state);
3839 		goto out;
3840 	}
3841 out:
3842 	nfs4_label_release_security(ilabel);
3843 	put_nfs_open_context(ctx);
3844 	return status;
3845 }
3846 
3847 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3848 {
3849 	struct nfs_server *server = NFS_SERVER(dir);
3850 	struct nfs_removeargs args = {
3851 		.fh = NFS_FH(dir),
3852 		.name = *name,
3853 	};
3854 	struct nfs_removeres res = {
3855 		.server = server,
3856 	};
3857 	struct rpc_message msg = {
3858 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3859 		.rpc_argp = &args,
3860 		.rpc_resp = &res,
3861 	};
3862 	int status;
3863 
3864 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3865 	if (status == 0)
3866 		update_changeattr(dir, &res.cinfo);
3867 	return status;
3868 }
3869 
3870 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3871 {
3872 	struct nfs4_exception exception = { };
3873 	int err;
3874 	do {
3875 		err = _nfs4_proc_remove(dir, name);
3876 		trace_nfs4_remove(dir, name, err);
3877 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3878 				&exception);
3879 	} while (exception.retry);
3880 	return err;
3881 }
3882 
3883 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3884 {
3885 	struct nfs_server *server = NFS_SERVER(dir);
3886 	struct nfs_removeargs *args = msg->rpc_argp;
3887 	struct nfs_removeres *res = msg->rpc_resp;
3888 
3889 	res->server = server;
3890 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3891 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3892 
3893 	nfs_fattr_init(res->dir_attr);
3894 }
3895 
3896 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3897 {
3898 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3899 			&data->args.seq_args,
3900 			&data->res.seq_res,
3901 			task);
3902 }
3903 
3904 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3905 {
3906 	struct nfs_unlinkdata *data = task->tk_calldata;
3907 	struct nfs_removeres *res = &data->res;
3908 
3909 	if (!nfs4_sequence_done(task, &res->seq_res))
3910 		return 0;
3911 	if (nfs4_async_handle_error(task, res->server, NULL,
3912 				    &data->timeout) == -EAGAIN)
3913 		return 0;
3914 	update_changeattr(dir, &res->cinfo);
3915 	return 1;
3916 }
3917 
3918 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3919 {
3920 	struct nfs_server *server = NFS_SERVER(dir);
3921 	struct nfs_renameargs *arg = msg->rpc_argp;
3922 	struct nfs_renameres *res = msg->rpc_resp;
3923 
3924 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3925 	res->server = server;
3926 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3927 }
3928 
3929 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3930 {
3931 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3932 			&data->args.seq_args,
3933 			&data->res.seq_res,
3934 			task);
3935 }
3936 
3937 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3938 				 struct inode *new_dir)
3939 {
3940 	struct nfs_renamedata *data = task->tk_calldata;
3941 	struct nfs_renameres *res = &data->res;
3942 
3943 	if (!nfs4_sequence_done(task, &res->seq_res))
3944 		return 0;
3945 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3946 		return 0;
3947 
3948 	update_changeattr(old_dir, &res->old_cinfo);
3949 	update_changeattr(new_dir, &res->new_cinfo);
3950 	return 1;
3951 }
3952 
3953 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3954 {
3955 	struct nfs_server *server = NFS_SERVER(inode);
3956 	struct nfs4_link_arg arg = {
3957 		.fh     = NFS_FH(inode),
3958 		.dir_fh = NFS_FH(dir),
3959 		.name   = name,
3960 		.bitmask = server->attr_bitmask,
3961 	};
3962 	struct nfs4_link_res res = {
3963 		.server = server,
3964 		.label = NULL,
3965 	};
3966 	struct rpc_message msg = {
3967 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3968 		.rpc_argp = &arg,
3969 		.rpc_resp = &res,
3970 	};
3971 	int status = -ENOMEM;
3972 
3973 	res.fattr = nfs_alloc_fattr();
3974 	if (res.fattr == NULL)
3975 		goto out;
3976 
3977 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3978 	if (IS_ERR(res.label)) {
3979 		status = PTR_ERR(res.label);
3980 		goto out;
3981 	}
3982 	arg.bitmask = nfs4_bitmask(server, res.label);
3983 
3984 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3985 	if (!status) {
3986 		update_changeattr(dir, &res.cinfo);
3987 		status = nfs_post_op_update_inode(inode, res.fattr);
3988 		if (!status)
3989 			nfs_setsecurity(inode, res.fattr, res.label);
3990 	}
3991 
3992 
3993 	nfs4_label_free(res.label);
3994 
3995 out:
3996 	nfs_free_fattr(res.fattr);
3997 	return status;
3998 }
3999 
4000 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4001 {
4002 	struct nfs4_exception exception = { };
4003 	int err;
4004 	do {
4005 		err = nfs4_handle_exception(NFS_SERVER(inode),
4006 				_nfs4_proc_link(inode, dir, name),
4007 				&exception);
4008 	} while (exception.retry);
4009 	return err;
4010 }
4011 
4012 struct nfs4_createdata {
4013 	struct rpc_message msg;
4014 	struct nfs4_create_arg arg;
4015 	struct nfs4_create_res res;
4016 	struct nfs_fh fh;
4017 	struct nfs_fattr fattr;
4018 	struct nfs4_label *label;
4019 };
4020 
4021 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4022 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4023 {
4024 	struct nfs4_createdata *data;
4025 
4026 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4027 	if (data != NULL) {
4028 		struct nfs_server *server = NFS_SERVER(dir);
4029 
4030 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
4031 		if (IS_ERR(data->label))
4032 			goto out_free;
4033 
4034 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4035 		data->msg.rpc_argp = &data->arg;
4036 		data->msg.rpc_resp = &data->res;
4037 		data->arg.dir_fh = NFS_FH(dir);
4038 		data->arg.server = server;
4039 		data->arg.name = name;
4040 		data->arg.attrs = sattr;
4041 		data->arg.ftype = ftype;
4042 		data->arg.bitmask = nfs4_bitmask(server, data->label);
4043 		data->res.server = server;
4044 		data->res.fh = &data->fh;
4045 		data->res.fattr = &data->fattr;
4046 		data->res.label = data->label;
4047 		nfs_fattr_init(data->res.fattr);
4048 	}
4049 	return data;
4050 out_free:
4051 	kfree(data);
4052 	return NULL;
4053 }
4054 
4055 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4056 {
4057 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4058 				    &data->arg.seq_args, &data->res.seq_res, 1);
4059 	if (status == 0) {
4060 		update_changeattr(dir, &data->res.dir_cinfo);
4061 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4062 	}
4063 	return status;
4064 }
4065 
4066 static void nfs4_free_createdata(struct nfs4_createdata *data)
4067 {
4068 	nfs4_label_free(data->label);
4069 	kfree(data);
4070 }
4071 
4072 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4073 		struct page *page, unsigned int len, struct iattr *sattr,
4074 		struct nfs4_label *label)
4075 {
4076 	struct nfs4_createdata *data;
4077 	int status = -ENAMETOOLONG;
4078 
4079 	if (len > NFS4_MAXPATHLEN)
4080 		goto out;
4081 
4082 	status = -ENOMEM;
4083 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4084 	if (data == NULL)
4085 		goto out;
4086 
4087 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4088 	data->arg.u.symlink.pages = &page;
4089 	data->arg.u.symlink.len = len;
4090 	data->arg.label = label;
4091 
4092 	status = nfs4_do_create(dir, dentry, data);
4093 
4094 	nfs4_free_createdata(data);
4095 out:
4096 	return status;
4097 }
4098 
4099 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4100 		struct page *page, unsigned int len, struct iattr *sattr)
4101 {
4102 	struct nfs4_exception exception = { };
4103 	struct nfs4_label l, *label = NULL;
4104 	int err;
4105 
4106 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4107 
4108 	do {
4109 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4110 		trace_nfs4_symlink(dir, &dentry->d_name, err);
4111 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4112 				&exception);
4113 	} while (exception.retry);
4114 
4115 	nfs4_label_release_security(label);
4116 	return err;
4117 }
4118 
4119 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4120 		struct iattr *sattr, struct nfs4_label *label)
4121 {
4122 	struct nfs4_createdata *data;
4123 	int status = -ENOMEM;
4124 
4125 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4126 	if (data == NULL)
4127 		goto out;
4128 
4129 	data->arg.label = label;
4130 	status = nfs4_do_create(dir, dentry, data);
4131 
4132 	nfs4_free_createdata(data);
4133 out:
4134 	return status;
4135 }
4136 
4137 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4138 		struct iattr *sattr)
4139 {
4140 	struct nfs4_exception exception = { };
4141 	struct nfs4_label l, *label = NULL;
4142 	int err;
4143 
4144 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4145 
4146 	sattr->ia_mode &= ~current_umask();
4147 	do {
4148 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4149 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
4150 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4151 				&exception);
4152 	} while (exception.retry);
4153 	nfs4_label_release_security(label);
4154 
4155 	return err;
4156 }
4157 
4158 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4159 		u64 cookie, struct page **pages, unsigned int count, int plus)
4160 {
4161 	struct inode		*dir = d_inode(dentry);
4162 	struct nfs4_readdir_arg args = {
4163 		.fh = NFS_FH(dir),
4164 		.pages = pages,
4165 		.pgbase = 0,
4166 		.count = count,
4167 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4168 		.plus = plus,
4169 	};
4170 	struct nfs4_readdir_res res;
4171 	struct rpc_message msg = {
4172 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4173 		.rpc_argp = &args,
4174 		.rpc_resp = &res,
4175 		.rpc_cred = cred,
4176 	};
4177 	int			status;
4178 
4179 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4180 			dentry,
4181 			(unsigned long long)cookie);
4182 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4183 	res.pgbase = args.pgbase;
4184 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4185 	if (status >= 0) {
4186 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4187 		status += args.pgbase;
4188 	}
4189 
4190 	nfs_invalidate_atime(dir);
4191 
4192 	dprintk("%s: returns %d\n", __func__, status);
4193 	return status;
4194 }
4195 
4196 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4197 		u64 cookie, struct page **pages, unsigned int count, int plus)
4198 {
4199 	struct nfs4_exception exception = { };
4200 	int err;
4201 	do {
4202 		err = _nfs4_proc_readdir(dentry, cred, cookie,
4203 				pages, count, plus);
4204 		trace_nfs4_readdir(d_inode(dentry), err);
4205 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4206 				&exception);
4207 	} while (exception.retry);
4208 	return err;
4209 }
4210 
4211 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4212 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4213 {
4214 	struct nfs4_createdata *data;
4215 	int mode = sattr->ia_mode;
4216 	int status = -ENOMEM;
4217 
4218 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4219 	if (data == NULL)
4220 		goto out;
4221 
4222 	if (S_ISFIFO(mode))
4223 		data->arg.ftype = NF4FIFO;
4224 	else if (S_ISBLK(mode)) {
4225 		data->arg.ftype = NF4BLK;
4226 		data->arg.u.device.specdata1 = MAJOR(rdev);
4227 		data->arg.u.device.specdata2 = MINOR(rdev);
4228 	}
4229 	else if (S_ISCHR(mode)) {
4230 		data->arg.ftype = NF4CHR;
4231 		data->arg.u.device.specdata1 = MAJOR(rdev);
4232 		data->arg.u.device.specdata2 = MINOR(rdev);
4233 	} else if (!S_ISSOCK(mode)) {
4234 		status = -EINVAL;
4235 		goto out_free;
4236 	}
4237 
4238 	data->arg.label = label;
4239 	status = nfs4_do_create(dir, dentry, data);
4240 out_free:
4241 	nfs4_free_createdata(data);
4242 out:
4243 	return status;
4244 }
4245 
4246 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4247 		struct iattr *sattr, dev_t rdev)
4248 {
4249 	struct nfs4_exception exception = { };
4250 	struct nfs4_label l, *label = NULL;
4251 	int err;
4252 
4253 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4254 
4255 	sattr->ia_mode &= ~current_umask();
4256 	do {
4257 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4258 		trace_nfs4_mknod(dir, &dentry->d_name, err);
4259 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4260 				&exception);
4261 	} while (exception.retry);
4262 
4263 	nfs4_label_release_security(label);
4264 
4265 	return err;
4266 }
4267 
4268 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4269 		 struct nfs_fsstat *fsstat)
4270 {
4271 	struct nfs4_statfs_arg args = {
4272 		.fh = fhandle,
4273 		.bitmask = server->attr_bitmask,
4274 	};
4275 	struct nfs4_statfs_res res = {
4276 		.fsstat = fsstat,
4277 	};
4278 	struct rpc_message msg = {
4279 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4280 		.rpc_argp = &args,
4281 		.rpc_resp = &res,
4282 	};
4283 
4284 	nfs_fattr_init(fsstat->fattr);
4285 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4286 }
4287 
4288 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4289 {
4290 	struct nfs4_exception exception = { };
4291 	int err;
4292 	do {
4293 		err = nfs4_handle_exception(server,
4294 				_nfs4_proc_statfs(server, fhandle, fsstat),
4295 				&exception);
4296 	} while (exception.retry);
4297 	return err;
4298 }
4299 
4300 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4301 		struct nfs_fsinfo *fsinfo)
4302 {
4303 	struct nfs4_fsinfo_arg args = {
4304 		.fh = fhandle,
4305 		.bitmask = server->attr_bitmask,
4306 	};
4307 	struct nfs4_fsinfo_res res = {
4308 		.fsinfo = fsinfo,
4309 	};
4310 	struct rpc_message msg = {
4311 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4312 		.rpc_argp = &args,
4313 		.rpc_resp = &res,
4314 	};
4315 
4316 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4317 }
4318 
4319 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4320 {
4321 	struct nfs4_exception exception = { };
4322 	unsigned long now = jiffies;
4323 	int err;
4324 
4325 	do {
4326 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4327 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4328 		if (err == 0) {
4329 			nfs4_set_lease_period(server->nfs_client,
4330 					fsinfo->lease_time * HZ,
4331 					now);
4332 			break;
4333 		}
4334 		err = nfs4_handle_exception(server, err, &exception);
4335 	} while (exception.retry);
4336 	return err;
4337 }
4338 
4339 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4340 {
4341 	int error;
4342 
4343 	nfs_fattr_init(fsinfo->fattr);
4344 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4345 	if (error == 0) {
4346 		/* block layout checks this! */
4347 		server->pnfs_blksize = fsinfo->blksize;
4348 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
4349 	}
4350 
4351 	return error;
4352 }
4353 
4354 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4355 		struct nfs_pathconf *pathconf)
4356 {
4357 	struct nfs4_pathconf_arg args = {
4358 		.fh = fhandle,
4359 		.bitmask = server->attr_bitmask,
4360 	};
4361 	struct nfs4_pathconf_res res = {
4362 		.pathconf = pathconf,
4363 	};
4364 	struct rpc_message msg = {
4365 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4366 		.rpc_argp = &args,
4367 		.rpc_resp = &res,
4368 	};
4369 
4370 	/* None of the pathconf attributes are mandatory to implement */
4371 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4372 		memset(pathconf, 0, sizeof(*pathconf));
4373 		return 0;
4374 	}
4375 
4376 	nfs_fattr_init(pathconf->fattr);
4377 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4378 }
4379 
4380 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4381 		struct nfs_pathconf *pathconf)
4382 {
4383 	struct nfs4_exception exception = { };
4384 	int err;
4385 
4386 	do {
4387 		err = nfs4_handle_exception(server,
4388 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4389 				&exception);
4390 	} while (exception.retry);
4391 	return err;
4392 }
4393 
4394 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4395 		const struct nfs_open_context *ctx,
4396 		const struct nfs_lock_context *l_ctx,
4397 		fmode_t fmode)
4398 {
4399 	const struct nfs_lockowner *lockowner = NULL;
4400 
4401 	if (l_ctx != NULL)
4402 		lockowner = &l_ctx->lockowner;
4403 	return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4404 }
4405 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4406 
4407 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4408 		const struct nfs_open_context *ctx,
4409 		const struct nfs_lock_context *l_ctx,
4410 		fmode_t fmode)
4411 {
4412 	nfs4_stateid current_stateid;
4413 
4414 	/* If the current stateid represents a lost lock, then exit */
4415 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4416 		return true;
4417 	return nfs4_stateid_match(stateid, &current_stateid);
4418 }
4419 
4420 static bool nfs4_error_stateid_expired(int err)
4421 {
4422 	switch (err) {
4423 	case -NFS4ERR_DELEG_REVOKED:
4424 	case -NFS4ERR_ADMIN_REVOKED:
4425 	case -NFS4ERR_BAD_STATEID:
4426 	case -NFS4ERR_STALE_STATEID:
4427 	case -NFS4ERR_OLD_STATEID:
4428 	case -NFS4ERR_OPENMODE:
4429 	case -NFS4ERR_EXPIRED:
4430 		return true;
4431 	}
4432 	return false;
4433 }
4434 
4435 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4436 {
4437 	nfs_invalidate_atime(hdr->inode);
4438 }
4439 
4440 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4441 {
4442 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4443 
4444 	trace_nfs4_read(hdr, task->tk_status);
4445 	if (nfs4_async_handle_error(task, server,
4446 				    hdr->args.context->state,
4447 				    NULL) == -EAGAIN) {
4448 		rpc_restart_call_prepare(task);
4449 		return -EAGAIN;
4450 	}
4451 
4452 	__nfs4_read_done_cb(hdr);
4453 	if (task->tk_status > 0)
4454 		renew_lease(server, hdr->timestamp);
4455 	return 0;
4456 }
4457 
4458 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4459 		struct nfs_pgio_args *args)
4460 {
4461 
4462 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4463 		nfs4_stateid_is_current(&args->stateid,
4464 				args->context,
4465 				args->lock_context,
4466 				FMODE_READ))
4467 		return false;
4468 	rpc_restart_call_prepare(task);
4469 	return true;
4470 }
4471 
4472 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4473 {
4474 
4475 	dprintk("--> %s\n", __func__);
4476 
4477 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4478 		return -EAGAIN;
4479 	if (nfs4_read_stateid_changed(task, &hdr->args))
4480 		return -EAGAIN;
4481 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4482 				    nfs4_read_done_cb(task, hdr);
4483 }
4484 
4485 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4486 				 struct rpc_message *msg)
4487 {
4488 	hdr->timestamp   = jiffies;
4489 	if (!hdr->pgio_done_cb)
4490 		hdr->pgio_done_cb = nfs4_read_done_cb;
4491 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4492 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4493 }
4494 
4495 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4496 				      struct nfs_pgio_header *hdr)
4497 {
4498 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4499 			&hdr->args.seq_args,
4500 			&hdr->res.seq_res,
4501 			task))
4502 		return 0;
4503 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4504 				hdr->args.lock_context,
4505 				hdr->rw_ops->rw_mode) == -EIO)
4506 		return -EIO;
4507 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4508 		return -EIO;
4509 	return 0;
4510 }
4511 
4512 static int nfs4_write_done_cb(struct rpc_task *task,
4513 			      struct nfs_pgio_header *hdr)
4514 {
4515 	struct inode *inode = hdr->inode;
4516 
4517 	trace_nfs4_write(hdr, task->tk_status);
4518 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4519 				    hdr->args.context->state,
4520 				    NULL) == -EAGAIN) {
4521 		rpc_restart_call_prepare(task);
4522 		return -EAGAIN;
4523 	}
4524 	if (task->tk_status >= 0) {
4525 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4526 		nfs_writeback_update_inode(hdr);
4527 	}
4528 	return 0;
4529 }
4530 
4531 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4532 		struct nfs_pgio_args *args)
4533 {
4534 
4535 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4536 		nfs4_stateid_is_current(&args->stateid,
4537 				args->context,
4538 				args->lock_context,
4539 				FMODE_WRITE))
4540 		return false;
4541 	rpc_restart_call_prepare(task);
4542 	return true;
4543 }
4544 
4545 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4546 {
4547 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4548 		return -EAGAIN;
4549 	if (nfs4_write_stateid_changed(task, &hdr->args))
4550 		return -EAGAIN;
4551 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4552 		nfs4_write_done_cb(task, hdr);
4553 }
4554 
4555 static
4556 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4557 {
4558 	/* Don't request attributes for pNFS or O_DIRECT writes */
4559 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4560 		return false;
4561 	/* Otherwise, request attributes if and only if we don't hold
4562 	 * a delegation
4563 	 */
4564 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4565 }
4566 
4567 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4568 				  struct rpc_message *msg)
4569 {
4570 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4571 
4572 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4573 		hdr->args.bitmask = NULL;
4574 		hdr->res.fattr = NULL;
4575 	} else
4576 		hdr->args.bitmask = server->cache_consistency_bitmask;
4577 
4578 	if (!hdr->pgio_done_cb)
4579 		hdr->pgio_done_cb = nfs4_write_done_cb;
4580 	hdr->res.server = server;
4581 	hdr->timestamp   = jiffies;
4582 
4583 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4584 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4585 }
4586 
4587 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4588 {
4589 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4590 			&data->args.seq_args,
4591 			&data->res.seq_res,
4592 			task);
4593 }
4594 
4595 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4596 {
4597 	struct inode *inode = data->inode;
4598 
4599 	trace_nfs4_commit(data, task->tk_status);
4600 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4601 				    NULL, NULL) == -EAGAIN) {
4602 		rpc_restart_call_prepare(task);
4603 		return -EAGAIN;
4604 	}
4605 	return 0;
4606 }
4607 
4608 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4609 {
4610 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4611 		return -EAGAIN;
4612 	return data->commit_done_cb(task, data);
4613 }
4614 
4615 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4616 {
4617 	struct nfs_server *server = NFS_SERVER(data->inode);
4618 
4619 	if (data->commit_done_cb == NULL)
4620 		data->commit_done_cb = nfs4_commit_done_cb;
4621 	data->res.server = server;
4622 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4623 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4624 }
4625 
4626 struct nfs4_renewdata {
4627 	struct nfs_client	*client;
4628 	unsigned long		timestamp;
4629 };
4630 
4631 /*
4632  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4633  * standalone procedure for queueing an asynchronous RENEW.
4634  */
4635 static void nfs4_renew_release(void *calldata)
4636 {
4637 	struct nfs4_renewdata *data = calldata;
4638 	struct nfs_client *clp = data->client;
4639 
4640 	if (atomic_read(&clp->cl_count) > 1)
4641 		nfs4_schedule_state_renewal(clp);
4642 	nfs_put_client(clp);
4643 	kfree(data);
4644 }
4645 
4646 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4647 {
4648 	struct nfs4_renewdata *data = calldata;
4649 	struct nfs_client *clp = data->client;
4650 	unsigned long timestamp = data->timestamp;
4651 
4652 	trace_nfs4_renew_async(clp, task->tk_status);
4653 	switch (task->tk_status) {
4654 	case 0:
4655 		break;
4656 	case -NFS4ERR_LEASE_MOVED:
4657 		nfs4_schedule_lease_moved_recovery(clp);
4658 		break;
4659 	default:
4660 		/* Unless we're shutting down, schedule state recovery! */
4661 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4662 			return;
4663 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4664 			nfs4_schedule_lease_recovery(clp);
4665 			return;
4666 		}
4667 		nfs4_schedule_path_down_recovery(clp);
4668 	}
4669 	do_renew_lease(clp, timestamp);
4670 }
4671 
4672 static const struct rpc_call_ops nfs4_renew_ops = {
4673 	.rpc_call_done = nfs4_renew_done,
4674 	.rpc_release = nfs4_renew_release,
4675 };
4676 
4677 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4678 {
4679 	struct rpc_message msg = {
4680 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4681 		.rpc_argp	= clp,
4682 		.rpc_cred	= cred,
4683 	};
4684 	struct nfs4_renewdata *data;
4685 
4686 	if (renew_flags == 0)
4687 		return 0;
4688 	if (!atomic_inc_not_zero(&clp->cl_count))
4689 		return -EIO;
4690 	data = kmalloc(sizeof(*data), GFP_NOFS);
4691 	if (data == NULL)
4692 		return -ENOMEM;
4693 	data->client = clp;
4694 	data->timestamp = jiffies;
4695 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4696 			&nfs4_renew_ops, data);
4697 }
4698 
4699 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4700 {
4701 	struct rpc_message msg = {
4702 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4703 		.rpc_argp	= clp,
4704 		.rpc_cred	= cred,
4705 	};
4706 	unsigned long now = jiffies;
4707 	int status;
4708 
4709 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4710 	if (status < 0)
4711 		return status;
4712 	do_renew_lease(clp, now);
4713 	return 0;
4714 }
4715 
4716 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4717 {
4718 	return server->caps & NFS_CAP_ACLS;
4719 }
4720 
4721 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4722  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4723  * the stack.
4724  */
4725 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4726 
4727 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4728 		struct page **pages)
4729 {
4730 	struct page *newpage, **spages;
4731 	int rc = 0;
4732 	size_t len;
4733 	spages = pages;
4734 
4735 	do {
4736 		len = min_t(size_t, PAGE_SIZE, buflen);
4737 		newpage = alloc_page(GFP_KERNEL);
4738 
4739 		if (newpage == NULL)
4740 			goto unwind;
4741 		memcpy(page_address(newpage), buf, len);
4742                 buf += len;
4743                 buflen -= len;
4744 		*pages++ = newpage;
4745 		rc++;
4746 	} while (buflen != 0);
4747 
4748 	return rc;
4749 
4750 unwind:
4751 	for(; rc > 0; rc--)
4752 		__free_page(spages[rc-1]);
4753 	return -ENOMEM;
4754 }
4755 
4756 struct nfs4_cached_acl {
4757 	int cached;
4758 	size_t len;
4759 	char data[0];
4760 };
4761 
4762 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4763 {
4764 	struct nfs_inode *nfsi = NFS_I(inode);
4765 
4766 	spin_lock(&inode->i_lock);
4767 	kfree(nfsi->nfs4_acl);
4768 	nfsi->nfs4_acl = acl;
4769 	spin_unlock(&inode->i_lock);
4770 }
4771 
4772 static void nfs4_zap_acl_attr(struct inode *inode)
4773 {
4774 	nfs4_set_cached_acl(inode, NULL);
4775 }
4776 
4777 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4778 {
4779 	struct nfs_inode *nfsi = NFS_I(inode);
4780 	struct nfs4_cached_acl *acl;
4781 	int ret = -ENOENT;
4782 
4783 	spin_lock(&inode->i_lock);
4784 	acl = nfsi->nfs4_acl;
4785 	if (acl == NULL)
4786 		goto out;
4787 	if (buf == NULL) /* user is just asking for length */
4788 		goto out_len;
4789 	if (acl->cached == 0)
4790 		goto out;
4791 	ret = -ERANGE; /* see getxattr(2) man page */
4792 	if (acl->len > buflen)
4793 		goto out;
4794 	memcpy(buf, acl->data, acl->len);
4795 out_len:
4796 	ret = acl->len;
4797 out:
4798 	spin_unlock(&inode->i_lock);
4799 	return ret;
4800 }
4801 
4802 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4803 {
4804 	struct nfs4_cached_acl *acl;
4805 	size_t buflen = sizeof(*acl) + acl_len;
4806 
4807 	if (buflen <= PAGE_SIZE) {
4808 		acl = kmalloc(buflen, GFP_KERNEL);
4809 		if (acl == NULL)
4810 			goto out;
4811 		acl->cached = 1;
4812 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4813 	} else {
4814 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4815 		if (acl == NULL)
4816 			goto out;
4817 		acl->cached = 0;
4818 	}
4819 	acl->len = acl_len;
4820 out:
4821 	nfs4_set_cached_acl(inode, acl);
4822 }
4823 
4824 /*
4825  * The getxattr API returns the required buffer length when called with a
4826  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4827  * the required buf.  On a NULL buf, we send a page of data to the server
4828  * guessing that the ACL request can be serviced by a page. If so, we cache
4829  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4830  * the cache. If not so, we throw away the page, and cache the required
4831  * length. The next getxattr call will then produce another round trip to
4832  * the server, this time with the input buf of the required size.
4833  */
4834 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4835 {
4836 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4837 	struct nfs_getaclargs args = {
4838 		.fh = NFS_FH(inode),
4839 		.acl_pages = pages,
4840 		.acl_len = buflen,
4841 	};
4842 	struct nfs_getaclres res = {
4843 		.acl_len = buflen,
4844 	};
4845 	struct rpc_message msg = {
4846 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4847 		.rpc_argp = &args,
4848 		.rpc_resp = &res,
4849 	};
4850 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4851 	int ret = -ENOMEM, i;
4852 
4853 	/* As long as we're doing a round trip to the server anyway,
4854 	 * let's be prepared for a page of acl data. */
4855 	if (npages == 0)
4856 		npages = 1;
4857 	if (npages > ARRAY_SIZE(pages))
4858 		return -ERANGE;
4859 
4860 	for (i = 0; i < npages; i++) {
4861 		pages[i] = alloc_page(GFP_KERNEL);
4862 		if (!pages[i])
4863 			goto out_free;
4864 	}
4865 
4866 	/* for decoding across pages */
4867 	res.acl_scratch = alloc_page(GFP_KERNEL);
4868 	if (!res.acl_scratch)
4869 		goto out_free;
4870 
4871 	args.acl_len = npages * PAGE_SIZE;
4872 
4873 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4874 		__func__, buf, buflen, npages, args.acl_len);
4875 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4876 			     &msg, &args.seq_args, &res.seq_res, 0);
4877 	if (ret)
4878 		goto out_free;
4879 
4880 	/* Handle the case where the passed-in buffer is too short */
4881 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4882 		/* Did the user only issue a request for the acl length? */
4883 		if (buf == NULL)
4884 			goto out_ok;
4885 		ret = -ERANGE;
4886 		goto out_free;
4887 	}
4888 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4889 	if (buf) {
4890 		if (res.acl_len > buflen) {
4891 			ret = -ERANGE;
4892 			goto out_free;
4893 		}
4894 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4895 	}
4896 out_ok:
4897 	ret = res.acl_len;
4898 out_free:
4899 	for (i = 0; i < npages; i++)
4900 		if (pages[i])
4901 			__free_page(pages[i]);
4902 	if (res.acl_scratch)
4903 		__free_page(res.acl_scratch);
4904 	return ret;
4905 }
4906 
4907 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4908 {
4909 	struct nfs4_exception exception = { };
4910 	ssize_t ret;
4911 	do {
4912 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4913 		trace_nfs4_get_acl(inode, ret);
4914 		if (ret >= 0)
4915 			break;
4916 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4917 	} while (exception.retry);
4918 	return ret;
4919 }
4920 
4921 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4922 {
4923 	struct nfs_server *server = NFS_SERVER(inode);
4924 	int ret;
4925 
4926 	if (!nfs4_server_supports_acls(server))
4927 		return -EOPNOTSUPP;
4928 	ret = nfs_revalidate_inode(server, inode);
4929 	if (ret < 0)
4930 		return ret;
4931 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4932 		nfs_zap_acl_cache(inode);
4933 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4934 	if (ret != -ENOENT)
4935 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4936 		 * but no cached acl data, just the acl length */
4937 		return ret;
4938 	return nfs4_get_acl_uncached(inode, buf, buflen);
4939 }
4940 
4941 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4942 {
4943 	struct nfs_server *server = NFS_SERVER(inode);
4944 	struct page *pages[NFS4ACL_MAXPAGES];
4945 	struct nfs_setaclargs arg = {
4946 		.fh		= NFS_FH(inode),
4947 		.acl_pages	= pages,
4948 		.acl_len	= buflen,
4949 	};
4950 	struct nfs_setaclres res;
4951 	struct rpc_message msg = {
4952 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4953 		.rpc_argp	= &arg,
4954 		.rpc_resp	= &res,
4955 	};
4956 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4957 	int ret, i;
4958 
4959 	if (!nfs4_server_supports_acls(server))
4960 		return -EOPNOTSUPP;
4961 	if (npages > ARRAY_SIZE(pages))
4962 		return -ERANGE;
4963 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4964 	if (i < 0)
4965 		return i;
4966 	nfs4_inode_return_delegation(inode);
4967 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4968 
4969 	/*
4970 	 * Free each page after tx, so the only ref left is
4971 	 * held by the network stack
4972 	 */
4973 	for (; i > 0; i--)
4974 		put_page(pages[i-1]);
4975 
4976 	/*
4977 	 * Acl update can result in inode attribute update.
4978 	 * so mark the attribute cache invalid.
4979 	 */
4980 	spin_lock(&inode->i_lock);
4981 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4982 	spin_unlock(&inode->i_lock);
4983 	nfs_access_zap_cache(inode);
4984 	nfs_zap_acl_cache(inode);
4985 	return ret;
4986 }
4987 
4988 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4989 {
4990 	struct nfs4_exception exception = { };
4991 	int err;
4992 	do {
4993 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4994 		trace_nfs4_set_acl(inode, err);
4995 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4996 				&exception);
4997 	} while (exception.retry);
4998 	return err;
4999 }
5000 
5001 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5002 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5003 					size_t buflen)
5004 {
5005 	struct nfs_server *server = NFS_SERVER(inode);
5006 	struct nfs_fattr fattr;
5007 	struct nfs4_label label = {0, 0, buflen, buf};
5008 
5009 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5010 	struct nfs4_getattr_arg arg = {
5011 		.fh		= NFS_FH(inode),
5012 		.bitmask	= bitmask,
5013 	};
5014 	struct nfs4_getattr_res res = {
5015 		.fattr		= &fattr,
5016 		.label		= &label,
5017 		.server		= server,
5018 	};
5019 	struct rpc_message msg = {
5020 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5021 		.rpc_argp	= &arg,
5022 		.rpc_resp	= &res,
5023 	};
5024 	int ret;
5025 
5026 	nfs_fattr_init(&fattr);
5027 
5028 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5029 	if (ret)
5030 		return ret;
5031 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5032 		return -ENOENT;
5033 	if (buflen < label.len)
5034 		return -ERANGE;
5035 	return 0;
5036 }
5037 
5038 static int nfs4_get_security_label(struct inode *inode, void *buf,
5039 					size_t buflen)
5040 {
5041 	struct nfs4_exception exception = { };
5042 	int err;
5043 
5044 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5045 		return -EOPNOTSUPP;
5046 
5047 	do {
5048 		err = _nfs4_get_security_label(inode, buf, buflen);
5049 		trace_nfs4_get_security_label(inode, err);
5050 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5051 				&exception);
5052 	} while (exception.retry);
5053 	return err;
5054 }
5055 
5056 static int _nfs4_do_set_security_label(struct inode *inode,
5057 		struct nfs4_label *ilabel,
5058 		struct nfs_fattr *fattr,
5059 		struct nfs4_label *olabel)
5060 {
5061 
5062 	struct iattr sattr = {0};
5063 	struct nfs_server *server = NFS_SERVER(inode);
5064 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5065 	struct nfs_setattrargs arg = {
5066 		.fh             = NFS_FH(inode),
5067 		.iap            = &sattr,
5068 		.server		= server,
5069 		.bitmask	= bitmask,
5070 		.label		= ilabel,
5071 	};
5072 	struct nfs_setattrres res = {
5073 		.fattr		= fattr,
5074 		.label		= olabel,
5075 		.server		= server,
5076 	};
5077 	struct rpc_message msg = {
5078 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5079 		.rpc_argp       = &arg,
5080 		.rpc_resp       = &res,
5081 	};
5082 	int status;
5083 
5084 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5085 
5086 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5087 	if (status)
5088 		dprintk("%s failed: %d\n", __func__, status);
5089 
5090 	return status;
5091 }
5092 
5093 static int nfs4_do_set_security_label(struct inode *inode,
5094 		struct nfs4_label *ilabel,
5095 		struct nfs_fattr *fattr,
5096 		struct nfs4_label *olabel)
5097 {
5098 	struct nfs4_exception exception = { };
5099 	int err;
5100 
5101 	do {
5102 		err = _nfs4_do_set_security_label(inode, ilabel,
5103 				fattr, olabel);
5104 		trace_nfs4_set_security_label(inode, err);
5105 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5106 				&exception);
5107 	} while (exception.retry);
5108 	return err;
5109 }
5110 
5111 static int
5112 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5113 {
5114 	struct nfs4_label ilabel, *olabel = NULL;
5115 	struct nfs_fattr fattr;
5116 	struct rpc_cred *cred;
5117 	int status;
5118 
5119 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5120 		return -EOPNOTSUPP;
5121 
5122 	nfs_fattr_init(&fattr);
5123 
5124 	ilabel.pi = 0;
5125 	ilabel.lfs = 0;
5126 	ilabel.label = (char *)buf;
5127 	ilabel.len = buflen;
5128 
5129 	cred = rpc_lookup_cred();
5130 	if (IS_ERR(cred))
5131 		return PTR_ERR(cred);
5132 
5133 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5134 	if (IS_ERR(olabel)) {
5135 		status = -PTR_ERR(olabel);
5136 		goto out;
5137 	}
5138 
5139 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5140 	if (status == 0)
5141 		nfs_setsecurity(inode, &fattr, olabel);
5142 
5143 	nfs4_label_free(olabel);
5144 out:
5145 	put_rpccred(cred);
5146 	return status;
5147 }
5148 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
5149 
5150 
5151 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5152 				    nfs4_verifier *bootverf)
5153 {
5154 	__be32 verf[2];
5155 
5156 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5157 		/* An impossible timestamp guarantees this value
5158 		 * will never match a generated boot time. */
5159 		verf[0] = 0;
5160 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5161 	} else {
5162 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5163 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5164 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5165 	}
5166 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
5167 }
5168 
5169 static int
5170 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5171 {
5172 	size_t len;
5173 	char *str;
5174 
5175 	if (clp->cl_owner_id != NULL)
5176 		return 0;
5177 
5178 	rcu_read_lock();
5179 	len = 14 + strlen(clp->cl_ipaddr) + 1 +
5180 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5181 		1 +
5182 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5183 		1;
5184 	rcu_read_unlock();
5185 
5186 	if (len > NFS4_OPAQUE_LIMIT + 1)
5187 		return -EINVAL;
5188 
5189 	/*
5190 	 * Since this string is allocated at mount time, and held until the
5191 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5192 	 * about a memory-reclaim deadlock.
5193 	 */
5194 	str = kmalloc(len, GFP_KERNEL);
5195 	if (!str)
5196 		return -ENOMEM;
5197 
5198 	rcu_read_lock();
5199 	scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5200 			clp->cl_ipaddr,
5201 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5202 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5203 	rcu_read_unlock();
5204 
5205 	clp->cl_owner_id = str;
5206 	return 0;
5207 }
5208 
5209 static int
5210 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5211 {
5212 	size_t len;
5213 	char *str;
5214 
5215 	len = 10 + 10 + 1 + 10 + 1 +
5216 		strlen(nfs4_client_id_uniquifier) + 1 +
5217 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5218 
5219 	if (len > NFS4_OPAQUE_LIMIT + 1)
5220 		return -EINVAL;
5221 
5222 	/*
5223 	 * Since this string is allocated at mount time, and held until the
5224 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5225 	 * about a memory-reclaim deadlock.
5226 	 */
5227 	str = kmalloc(len, GFP_KERNEL);
5228 	if (!str)
5229 		return -ENOMEM;
5230 
5231 	scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5232 			clp->rpc_ops->version, clp->cl_minorversion,
5233 			nfs4_client_id_uniquifier,
5234 			clp->cl_rpcclient->cl_nodename);
5235 	clp->cl_owner_id = str;
5236 	return 0;
5237 }
5238 
5239 static int
5240 nfs4_init_uniform_client_string(struct nfs_client *clp)
5241 {
5242 	size_t len;
5243 	char *str;
5244 
5245 	if (clp->cl_owner_id != NULL)
5246 		return 0;
5247 
5248 	if (nfs4_client_id_uniquifier[0] != '\0')
5249 		return nfs4_init_uniquifier_client_string(clp);
5250 
5251 	len = 10 + 10 + 1 + 10 + 1 +
5252 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5253 
5254 	if (len > NFS4_OPAQUE_LIMIT + 1)
5255 		return -EINVAL;
5256 
5257 	/*
5258 	 * Since this string is allocated at mount time, and held until the
5259 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5260 	 * about a memory-reclaim deadlock.
5261 	 */
5262 	str = kmalloc(len, GFP_KERNEL);
5263 	if (!str)
5264 		return -ENOMEM;
5265 
5266 	scnprintf(str, len, "Linux NFSv%u.%u %s",
5267 			clp->rpc_ops->version, clp->cl_minorversion,
5268 			clp->cl_rpcclient->cl_nodename);
5269 	clp->cl_owner_id = str;
5270 	return 0;
5271 }
5272 
5273 /*
5274  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5275  * services.  Advertise one based on the address family of the
5276  * clientaddr.
5277  */
5278 static unsigned int
5279 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5280 {
5281 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5282 		return scnprintf(buf, len, "tcp6");
5283 	else
5284 		return scnprintf(buf, len, "tcp");
5285 }
5286 
5287 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5288 {
5289 	struct nfs4_setclientid *sc = calldata;
5290 
5291 	if (task->tk_status == 0)
5292 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5293 }
5294 
5295 static const struct rpc_call_ops nfs4_setclientid_ops = {
5296 	.rpc_call_done = nfs4_setclientid_done,
5297 };
5298 
5299 /**
5300  * nfs4_proc_setclientid - Negotiate client ID
5301  * @clp: state data structure
5302  * @program: RPC program for NFSv4 callback service
5303  * @port: IP port number for NFS4 callback service
5304  * @cred: RPC credential to use for this call
5305  * @res: where to place the result
5306  *
5307  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5308  */
5309 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5310 		unsigned short port, struct rpc_cred *cred,
5311 		struct nfs4_setclientid_res *res)
5312 {
5313 	nfs4_verifier sc_verifier;
5314 	struct nfs4_setclientid setclientid = {
5315 		.sc_verifier = &sc_verifier,
5316 		.sc_prog = program,
5317 		.sc_clnt = clp,
5318 	};
5319 	struct rpc_message msg = {
5320 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5321 		.rpc_argp = &setclientid,
5322 		.rpc_resp = res,
5323 		.rpc_cred = cred,
5324 	};
5325 	struct rpc_task *task;
5326 	struct rpc_task_setup task_setup_data = {
5327 		.rpc_client = clp->cl_rpcclient,
5328 		.rpc_message = &msg,
5329 		.callback_ops = &nfs4_setclientid_ops,
5330 		.callback_data = &setclientid,
5331 		.flags = RPC_TASK_TIMEOUT,
5332 	};
5333 	int status;
5334 
5335 	/* nfs_client_id4 */
5336 	nfs4_init_boot_verifier(clp, &sc_verifier);
5337 
5338 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5339 		status = nfs4_init_uniform_client_string(clp);
5340 	else
5341 		status = nfs4_init_nonuniform_client_string(clp);
5342 
5343 	if (status)
5344 		goto out;
5345 
5346 	/* cb_client4 */
5347 	setclientid.sc_netid_len =
5348 				nfs4_init_callback_netid(clp,
5349 						setclientid.sc_netid,
5350 						sizeof(setclientid.sc_netid));
5351 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5352 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5353 				clp->cl_ipaddr, port >> 8, port & 255);
5354 
5355 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
5356 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5357 		clp->cl_owner_id);
5358 	task = rpc_run_task(&task_setup_data);
5359 	if (IS_ERR(task)) {
5360 		status = PTR_ERR(task);
5361 		goto out;
5362 	}
5363 	status = task->tk_status;
5364 	if (setclientid.sc_cred) {
5365 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5366 		put_rpccred(setclientid.sc_cred);
5367 	}
5368 	rpc_put_task(task);
5369 out:
5370 	trace_nfs4_setclientid(clp, status);
5371 	dprintk("NFS reply setclientid: %d\n", status);
5372 	return status;
5373 }
5374 
5375 /**
5376  * nfs4_proc_setclientid_confirm - Confirm client ID
5377  * @clp: state data structure
5378  * @res: result of a previous SETCLIENTID
5379  * @cred: RPC credential to use for this call
5380  *
5381  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5382  */
5383 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5384 		struct nfs4_setclientid_res *arg,
5385 		struct rpc_cred *cred)
5386 {
5387 	struct rpc_message msg = {
5388 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5389 		.rpc_argp = arg,
5390 		.rpc_cred = cred,
5391 	};
5392 	int status;
5393 
5394 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5395 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5396 		clp->cl_clientid);
5397 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5398 	trace_nfs4_setclientid_confirm(clp, status);
5399 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5400 	return status;
5401 }
5402 
5403 struct nfs4_delegreturndata {
5404 	struct nfs4_delegreturnargs args;
5405 	struct nfs4_delegreturnres res;
5406 	struct nfs_fh fh;
5407 	nfs4_stateid stateid;
5408 	unsigned long timestamp;
5409 	struct nfs_fattr fattr;
5410 	int rpc_status;
5411 	struct inode *inode;
5412 	bool roc;
5413 	u32 roc_barrier;
5414 };
5415 
5416 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5417 {
5418 	struct nfs4_delegreturndata *data = calldata;
5419 
5420 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5421 		return;
5422 
5423 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5424 	switch (task->tk_status) {
5425 	case 0:
5426 		renew_lease(data->res.server, data->timestamp);
5427 	case -NFS4ERR_ADMIN_REVOKED:
5428 	case -NFS4ERR_DELEG_REVOKED:
5429 	case -NFS4ERR_BAD_STATEID:
5430 	case -NFS4ERR_OLD_STATEID:
5431 	case -NFS4ERR_STALE_STATEID:
5432 	case -NFS4ERR_EXPIRED:
5433 		task->tk_status = 0;
5434 		if (data->roc)
5435 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5436 		break;
5437 	default:
5438 		if (nfs4_async_handle_error(task, data->res.server,
5439 					    NULL, NULL) == -EAGAIN) {
5440 			rpc_restart_call_prepare(task);
5441 			return;
5442 		}
5443 	}
5444 	data->rpc_status = task->tk_status;
5445 }
5446 
5447 static void nfs4_delegreturn_release(void *calldata)
5448 {
5449 	struct nfs4_delegreturndata *data = calldata;
5450 	struct inode *inode = data->inode;
5451 
5452 	if (inode) {
5453 		if (data->roc)
5454 			pnfs_roc_release(inode);
5455 		nfs_iput_and_deactive(inode);
5456 	}
5457 	kfree(calldata);
5458 }
5459 
5460 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5461 {
5462 	struct nfs4_delegreturndata *d_data;
5463 
5464 	d_data = (struct nfs4_delegreturndata *)data;
5465 
5466 	if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5467 		return;
5468 
5469 	if (d_data->roc)
5470 		pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5471 
5472 	nfs4_setup_sequence(d_data->res.server,
5473 			&d_data->args.seq_args,
5474 			&d_data->res.seq_res,
5475 			task);
5476 }
5477 
5478 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5479 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5480 	.rpc_call_done = nfs4_delegreturn_done,
5481 	.rpc_release = nfs4_delegreturn_release,
5482 };
5483 
5484 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5485 {
5486 	struct nfs4_delegreturndata *data;
5487 	struct nfs_server *server = NFS_SERVER(inode);
5488 	struct rpc_task *task;
5489 	struct rpc_message msg = {
5490 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5491 		.rpc_cred = cred,
5492 	};
5493 	struct rpc_task_setup task_setup_data = {
5494 		.rpc_client = server->client,
5495 		.rpc_message = &msg,
5496 		.callback_ops = &nfs4_delegreturn_ops,
5497 		.flags = RPC_TASK_ASYNC,
5498 	};
5499 	int status = 0;
5500 
5501 	data = kzalloc(sizeof(*data), GFP_NOFS);
5502 	if (data == NULL)
5503 		return -ENOMEM;
5504 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5505 
5506 	nfs4_state_protect(server->nfs_client,
5507 			NFS_SP4_MACH_CRED_CLEANUP,
5508 			&task_setup_data.rpc_client, &msg);
5509 
5510 	data->args.fhandle = &data->fh;
5511 	data->args.stateid = &data->stateid;
5512 	data->args.bitmask = server->cache_consistency_bitmask;
5513 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5514 	nfs4_stateid_copy(&data->stateid, stateid);
5515 	data->res.fattr = &data->fattr;
5516 	data->res.server = server;
5517 	nfs_fattr_init(data->res.fattr);
5518 	data->timestamp = jiffies;
5519 	data->rpc_status = 0;
5520 	data->inode = nfs_igrab_and_active(inode);
5521 	if (data->inode)
5522 		data->roc = nfs4_roc(inode);
5523 
5524 	task_setup_data.callback_data = data;
5525 	msg.rpc_argp = &data->args;
5526 	msg.rpc_resp = &data->res;
5527 	task = rpc_run_task(&task_setup_data);
5528 	if (IS_ERR(task))
5529 		return PTR_ERR(task);
5530 	if (!issync)
5531 		goto out;
5532 	status = nfs4_wait_for_completion_rpc_task(task);
5533 	if (status != 0)
5534 		goto out;
5535 	status = data->rpc_status;
5536 	if (status == 0)
5537 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5538 	else
5539 		nfs_refresh_inode(inode, &data->fattr);
5540 out:
5541 	rpc_put_task(task);
5542 	return status;
5543 }
5544 
5545 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5546 {
5547 	struct nfs_server *server = NFS_SERVER(inode);
5548 	struct nfs4_exception exception = { };
5549 	int err;
5550 	do {
5551 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5552 		trace_nfs4_delegreturn(inode, stateid, err);
5553 		switch (err) {
5554 			case -NFS4ERR_STALE_STATEID:
5555 			case -NFS4ERR_EXPIRED:
5556 			case 0:
5557 				return 0;
5558 		}
5559 		err = nfs4_handle_exception(server, err, &exception);
5560 	} while (exception.retry);
5561 	return err;
5562 }
5563 
5564 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5565 {
5566 	struct inode *inode = state->inode;
5567 	struct nfs_server *server = NFS_SERVER(inode);
5568 	struct nfs_client *clp = server->nfs_client;
5569 	struct nfs_lockt_args arg = {
5570 		.fh = NFS_FH(inode),
5571 		.fl = request,
5572 	};
5573 	struct nfs_lockt_res res = {
5574 		.denied = request,
5575 	};
5576 	struct rpc_message msg = {
5577 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5578 		.rpc_argp       = &arg,
5579 		.rpc_resp       = &res,
5580 		.rpc_cred	= state->owner->so_cred,
5581 	};
5582 	struct nfs4_lock_state *lsp;
5583 	int status;
5584 
5585 	arg.lock_owner.clientid = clp->cl_clientid;
5586 	status = nfs4_set_lock_state(state, request);
5587 	if (status != 0)
5588 		goto out;
5589 	lsp = request->fl_u.nfs4_fl.owner;
5590 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5591 	arg.lock_owner.s_dev = server->s_dev;
5592 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5593 	switch (status) {
5594 		case 0:
5595 			request->fl_type = F_UNLCK;
5596 			break;
5597 		case -NFS4ERR_DENIED:
5598 			status = 0;
5599 	}
5600 	request->fl_ops->fl_release_private(request);
5601 	request->fl_ops = NULL;
5602 out:
5603 	return status;
5604 }
5605 
5606 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5607 {
5608 	struct nfs4_exception exception = { };
5609 	int err;
5610 
5611 	do {
5612 		err = _nfs4_proc_getlk(state, cmd, request);
5613 		trace_nfs4_get_lock(request, state, cmd, err);
5614 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5615 				&exception);
5616 	} while (exception.retry);
5617 	return err;
5618 }
5619 
5620 struct nfs4_unlockdata {
5621 	struct nfs_locku_args arg;
5622 	struct nfs_locku_res res;
5623 	struct nfs4_lock_state *lsp;
5624 	struct nfs_open_context *ctx;
5625 	struct file_lock fl;
5626 	struct nfs_server *server;
5627 	unsigned long timestamp;
5628 };
5629 
5630 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5631 		struct nfs_open_context *ctx,
5632 		struct nfs4_lock_state *lsp,
5633 		struct nfs_seqid *seqid)
5634 {
5635 	struct nfs4_unlockdata *p;
5636 	struct inode *inode = lsp->ls_state->inode;
5637 
5638 	p = kzalloc(sizeof(*p), GFP_NOFS);
5639 	if (p == NULL)
5640 		return NULL;
5641 	p->arg.fh = NFS_FH(inode);
5642 	p->arg.fl = &p->fl;
5643 	p->arg.seqid = seqid;
5644 	p->res.seqid = seqid;
5645 	p->lsp = lsp;
5646 	atomic_inc(&lsp->ls_count);
5647 	/* Ensure we don't close file until we're done freeing locks! */
5648 	p->ctx = get_nfs_open_context(ctx);
5649 	memcpy(&p->fl, fl, sizeof(p->fl));
5650 	p->server = NFS_SERVER(inode);
5651 	return p;
5652 }
5653 
5654 static void nfs4_locku_release_calldata(void *data)
5655 {
5656 	struct nfs4_unlockdata *calldata = data;
5657 	nfs_free_seqid(calldata->arg.seqid);
5658 	nfs4_put_lock_state(calldata->lsp);
5659 	put_nfs_open_context(calldata->ctx);
5660 	kfree(calldata);
5661 }
5662 
5663 static void nfs4_locku_done(struct rpc_task *task, void *data)
5664 {
5665 	struct nfs4_unlockdata *calldata = data;
5666 
5667 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5668 		return;
5669 	switch (task->tk_status) {
5670 		case 0:
5671 			renew_lease(calldata->server, calldata->timestamp);
5672 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5673 			if (nfs4_update_lock_stateid(calldata->lsp,
5674 					&calldata->res.stateid))
5675 				break;
5676 		case -NFS4ERR_BAD_STATEID:
5677 		case -NFS4ERR_OLD_STATEID:
5678 		case -NFS4ERR_STALE_STATEID:
5679 		case -NFS4ERR_EXPIRED:
5680 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5681 						&calldata->lsp->ls_stateid))
5682 				rpc_restart_call_prepare(task);
5683 			break;
5684 		default:
5685 			if (nfs4_async_handle_error(task, calldata->server,
5686 						    NULL, NULL) == -EAGAIN)
5687 				rpc_restart_call_prepare(task);
5688 	}
5689 	nfs_release_seqid(calldata->arg.seqid);
5690 }
5691 
5692 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5693 {
5694 	struct nfs4_unlockdata *calldata = data;
5695 
5696 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5697 		goto out_wait;
5698 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5699 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5700 		/* Note: exit _without_ running nfs4_locku_done */
5701 		goto out_no_action;
5702 	}
5703 	calldata->timestamp = jiffies;
5704 	if (nfs4_setup_sequence(calldata->server,
5705 				&calldata->arg.seq_args,
5706 				&calldata->res.seq_res,
5707 				task) != 0)
5708 		nfs_release_seqid(calldata->arg.seqid);
5709 	return;
5710 out_no_action:
5711 	task->tk_action = NULL;
5712 out_wait:
5713 	nfs4_sequence_done(task, &calldata->res.seq_res);
5714 }
5715 
5716 static const struct rpc_call_ops nfs4_locku_ops = {
5717 	.rpc_call_prepare = nfs4_locku_prepare,
5718 	.rpc_call_done = nfs4_locku_done,
5719 	.rpc_release = nfs4_locku_release_calldata,
5720 };
5721 
5722 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5723 		struct nfs_open_context *ctx,
5724 		struct nfs4_lock_state *lsp,
5725 		struct nfs_seqid *seqid)
5726 {
5727 	struct nfs4_unlockdata *data;
5728 	struct rpc_message msg = {
5729 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5730 		.rpc_cred = ctx->cred,
5731 	};
5732 	struct rpc_task_setup task_setup_data = {
5733 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5734 		.rpc_message = &msg,
5735 		.callback_ops = &nfs4_locku_ops,
5736 		.workqueue = nfsiod_workqueue,
5737 		.flags = RPC_TASK_ASYNC,
5738 	};
5739 
5740 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5741 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5742 
5743 	/* Ensure this is an unlock - when canceling a lock, the
5744 	 * canceled lock is passed in, and it won't be an unlock.
5745 	 */
5746 	fl->fl_type = F_UNLCK;
5747 
5748 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5749 	if (data == NULL) {
5750 		nfs_free_seqid(seqid);
5751 		return ERR_PTR(-ENOMEM);
5752 	}
5753 
5754 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5755 	msg.rpc_argp = &data->arg;
5756 	msg.rpc_resp = &data->res;
5757 	task_setup_data.callback_data = data;
5758 	return rpc_run_task(&task_setup_data);
5759 }
5760 
5761 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5762 {
5763 	struct inode *inode = state->inode;
5764 	struct nfs4_state_owner *sp = state->owner;
5765 	struct nfs_inode *nfsi = NFS_I(inode);
5766 	struct nfs_seqid *seqid;
5767 	struct nfs4_lock_state *lsp;
5768 	struct rpc_task *task;
5769 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5770 	int status = 0;
5771 	unsigned char fl_flags = request->fl_flags;
5772 
5773 	status = nfs4_set_lock_state(state, request);
5774 	/* Unlock _before_ we do the RPC call */
5775 	request->fl_flags |= FL_EXISTS;
5776 	/* Exclude nfs_delegation_claim_locks() */
5777 	mutex_lock(&sp->so_delegreturn_mutex);
5778 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5779 	down_read(&nfsi->rwsem);
5780 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
5781 		up_read(&nfsi->rwsem);
5782 		mutex_unlock(&sp->so_delegreturn_mutex);
5783 		goto out;
5784 	}
5785 	up_read(&nfsi->rwsem);
5786 	mutex_unlock(&sp->so_delegreturn_mutex);
5787 	if (status != 0)
5788 		goto out;
5789 	/* Is this a delegated lock? */
5790 	lsp = request->fl_u.nfs4_fl.owner;
5791 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5792 		goto out;
5793 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5794 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5795 	status = -ENOMEM;
5796 	if (IS_ERR(seqid))
5797 		goto out;
5798 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5799 	status = PTR_ERR(task);
5800 	if (IS_ERR(task))
5801 		goto out;
5802 	status = nfs4_wait_for_completion_rpc_task(task);
5803 	rpc_put_task(task);
5804 out:
5805 	request->fl_flags = fl_flags;
5806 	trace_nfs4_unlock(request, state, F_SETLK, status);
5807 	return status;
5808 }
5809 
5810 struct nfs4_lockdata {
5811 	struct nfs_lock_args arg;
5812 	struct nfs_lock_res res;
5813 	struct nfs4_lock_state *lsp;
5814 	struct nfs_open_context *ctx;
5815 	struct file_lock fl;
5816 	unsigned long timestamp;
5817 	int rpc_status;
5818 	int cancelled;
5819 	struct nfs_server *server;
5820 };
5821 
5822 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5823 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5824 		gfp_t gfp_mask)
5825 {
5826 	struct nfs4_lockdata *p;
5827 	struct inode *inode = lsp->ls_state->inode;
5828 	struct nfs_server *server = NFS_SERVER(inode);
5829 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5830 
5831 	p = kzalloc(sizeof(*p), gfp_mask);
5832 	if (p == NULL)
5833 		return NULL;
5834 
5835 	p->arg.fh = NFS_FH(inode);
5836 	p->arg.fl = &p->fl;
5837 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5838 	if (IS_ERR(p->arg.open_seqid))
5839 		goto out_free;
5840 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5841 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5842 	if (IS_ERR(p->arg.lock_seqid))
5843 		goto out_free_seqid;
5844 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5845 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5846 	p->arg.lock_owner.s_dev = server->s_dev;
5847 	p->res.lock_seqid = p->arg.lock_seqid;
5848 	p->lsp = lsp;
5849 	p->server = server;
5850 	atomic_inc(&lsp->ls_count);
5851 	p->ctx = get_nfs_open_context(ctx);
5852 	get_file(fl->fl_file);
5853 	memcpy(&p->fl, fl, sizeof(p->fl));
5854 	return p;
5855 out_free_seqid:
5856 	nfs_free_seqid(p->arg.open_seqid);
5857 out_free:
5858 	kfree(p);
5859 	return NULL;
5860 }
5861 
5862 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5863 {
5864 	struct nfs4_lockdata *data = calldata;
5865 	struct nfs4_state *state = data->lsp->ls_state;
5866 
5867 	dprintk("%s: begin!\n", __func__);
5868 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5869 		goto out_wait;
5870 	/* Do we need to do an open_to_lock_owner? */
5871 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5872 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5873 			goto out_release_lock_seqid;
5874 		}
5875 		nfs4_stateid_copy(&data->arg.open_stateid,
5876 				&state->open_stateid);
5877 		data->arg.new_lock_owner = 1;
5878 		data->res.open_seqid = data->arg.open_seqid;
5879 	} else {
5880 		data->arg.new_lock_owner = 0;
5881 		nfs4_stateid_copy(&data->arg.lock_stateid,
5882 				&data->lsp->ls_stateid);
5883 	}
5884 	if (!nfs4_valid_open_stateid(state)) {
5885 		data->rpc_status = -EBADF;
5886 		task->tk_action = NULL;
5887 		goto out_release_open_seqid;
5888 	}
5889 	data->timestamp = jiffies;
5890 	if (nfs4_setup_sequence(data->server,
5891 				&data->arg.seq_args,
5892 				&data->res.seq_res,
5893 				task) == 0)
5894 		return;
5895 out_release_open_seqid:
5896 	nfs_release_seqid(data->arg.open_seqid);
5897 out_release_lock_seqid:
5898 	nfs_release_seqid(data->arg.lock_seqid);
5899 out_wait:
5900 	nfs4_sequence_done(task, &data->res.seq_res);
5901 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5902 }
5903 
5904 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5905 {
5906 	struct nfs4_lockdata *data = calldata;
5907 	struct nfs4_lock_state *lsp = data->lsp;
5908 
5909 	dprintk("%s: begin!\n", __func__);
5910 
5911 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5912 		return;
5913 
5914 	data->rpc_status = task->tk_status;
5915 	switch (task->tk_status) {
5916 	case 0:
5917 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5918 				data->timestamp);
5919 		if (data->arg.new_lock) {
5920 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5921 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
5922 				rpc_restart_call_prepare(task);
5923 				break;
5924 			}
5925 		}
5926 		if (data->arg.new_lock_owner != 0) {
5927 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5928 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5929 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5930 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5931 			rpc_restart_call_prepare(task);
5932 		break;
5933 	case -NFS4ERR_BAD_STATEID:
5934 	case -NFS4ERR_OLD_STATEID:
5935 	case -NFS4ERR_STALE_STATEID:
5936 	case -NFS4ERR_EXPIRED:
5937 		if (data->arg.new_lock_owner != 0) {
5938 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5939 						&lsp->ls_state->open_stateid))
5940 				rpc_restart_call_prepare(task);
5941 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5942 						&lsp->ls_stateid))
5943 				rpc_restart_call_prepare(task);
5944 	}
5945 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5946 }
5947 
5948 static void nfs4_lock_release(void *calldata)
5949 {
5950 	struct nfs4_lockdata *data = calldata;
5951 
5952 	dprintk("%s: begin!\n", __func__);
5953 	nfs_free_seqid(data->arg.open_seqid);
5954 	if (data->cancelled != 0) {
5955 		struct rpc_task *task;
5956 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5957 				data->arg.lock_seqid);
5958 		if (!IS_ERR(task))
5959 			rpc_put_task_async(task);
5960 		dprintk("%s: cancelling lock!\n", __func__);
5961 	} else
5962 		nfs_free_seqid(data->arg.lock_seqid);
5963 	nfs4_put_lock_state(data->lsp);
5964 	put_nfs_open_context(data->ctx);
5965 	fput(data->fl.fl_file);
5966 	kfree(data);
5967 	dprintk("%s: done!\n", __func__);
5968 }
5969 
5970 static const struct rpc_call_ops nfs4_lock_ops = {
5971 	.rpc_call_prepare = nfs4_lock_prepare,
5972 	.rpc_call_done = nfs4_lock_done,
5973 	.rpc_release = nfs4_lock_release,
5974 };
5975 
5976 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5977 {
5978 	switch (error) {
5979 	case -NFS4ERR_ADMIN_REVOKED:
5980 	case -NFS4ERR_BAD_STATEID:
5981 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5982 		if (new_lock_owner != 0 ||
5983 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5984 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5985 		break;
5986 	case -NFS4ERR_STALE_STATEID:
5987 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5988 	case -NFS4ERR_EXPIRED:
5989 		nfs4_schedule_lease_recovery(server->nfs_client);
5990 	};
5991 }
5992 
5993 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5994 {
5995 	struct nfs4_lockdata *data;
5996 	struct rpc_task *task;
5997 	struct rpc_message msg = {
5998 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5999 		.rpc_cred = state->owner->so_cred,
6000 	};
6001 	struct rpc_task_setup task_setup_data = {
6002 		.rpc_client = NFS_CLIENT(state->inode),
6003 		.rpc_message = &msg,
6004 		.callback_ops = &nfs4_lock_ops,
6005 		.workqueue = nfsiod_workqueue,
6006 		.flags = RPC_TASK_ASYNC,
6007 	};
6008 	int ret;
6009 
6010 	dprintk("%s: begin!\n", __func__);
6011 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6012 			fl->fl_u.nfs4_fl.owner,
6013 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6014 	if (data == NULL)
6015 		return -ENOMEM;
6016 	if (IS_SETLKW(cmd))
6017 		data->arg.block = 1;
6018 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6019 	msg.rpc_argp = &data->arg;
6020 	msg.rpc_resp = &data->res;
6021 	task_setup_data.callback_data = data;
6022 	if (recovery_type > NFS_LOCK_NEW) {
6023 		if (recovery_type == NFS_LOCK_RECLAIM)
6024 			data->arg.reclaim = NFS_LOCK_RECLAIM;
6025 		nfs4_set_sequence_privileged(&data->arg.seq_args);
6026 	} else
6027 		data->arg.new_lock = 1;
6028 	task = rpc_run_task(&task_setup_data);
6029 	if (IS_ERR(task))
6030 		return PTR_ERR(task);
6031 	ret = nfs4_wait_for_completion_rpc_task(task);
6032 	if (ret == 0) {
6033 		ret = data->rpc_status;
6034 		if (ret)
6035 			nfs4_handle_setlk_error(data->server, data->lsp,
6036 					data->arg.new_lock_owner, ret);
6037 	} else
6038 		data->cancelled = 1;
6039 	rpc_put_task(task);
6040 	dprintk("%s: done, ret = %d!\n", __func__, ret);
6041 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6042 	return ret;
6043 }
6044 
6045 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6046 {
6047 	struct nfs_server *server = NFS_SERVER(state->inode);
6048 	struct nfs4_exception exception = {
6049 		.inode = state->inode,
6050 	};
6051 	int err;
6052 
6053 	do {
6054 		/* Cache the lock if possible... */
6055 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6056 			return 0;
6057 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6058 		if (err != -NFS4ERR_DELAY)
6059 			break;
6060 		nfs4_handle_exception(server, err, &exception);
6061 	} while (exception.retry);
6062 	return err;
6063 }
6064 
6065 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6066 {
6067 	struct nfs_server *server = NFS_SERVER(state->inode);
6068 	struct nfs4_exception exception = {
6069 		.inode = state->inode,
6070 	};
6071 	int err;
6072 
6073 	err = nfs4_set_lock_state(state, request);
6074 	if (err != 0)
6075 		return err;
6076 	if (!recover_lost_locks) {
6077 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6078 		return 0;
6079 	}
6080 	do {
6081 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6082 			return 0;
6083 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6084 		switch (err) {
6085 		default:
6086 			goto out;
6087 		case -NFS4ERR_GRACE:
6088 		case -NFS4ERR_DELAY:
6089 			nfs4_handle_exception(server, err, &exception);
6090 			err = 0;
6091 		}
6092 	} while (exception.retry);
6093 out:
6094 	return err;
6095 }
6096 
6097 #if defined(CONFIG_NFS_V4_1)
6098 /**
6099  * nfs41_check_expired_locks - possibly free a lock stateid
6100  *
6101  * @state: NFSv4 state for an inode
6102  *
6103  * Returns NFS_OK if recovery for this stateid is now finished.
6104  * Otherwise a negative NFS4ERR value is returned.
6105  */
6106 static int nfs41_check_expired_locks(struct nfs4_state *state)
6107 {
6108 	int status, ret = -NFS4ERR_BAD_STATEID;
6109 	struct nfs4_lock_state *lsp;
6110 	struct nfs_server *server = NFS_SERVER(state->inode);
6111 
6112 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6113 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6114 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6115 
6116 			status = nfs41_test_and_free_expired_stateid(server,
6117 					&lsp->ls_stateid,
6118 					cred);
6119 			trace_nfs4_test_lock_stateid(state, lsp, status);
6120 			if (status != NFS_OK) {
6121 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6122 				ret = status;
6123 			}
6124 		}
6125 	};
6126 
6127 	return ret;
6128 }
6129 
6130 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6131 {
6132 	int status = NFS_OK;
6133 
6134 	if (test_bit(LK_STATE_IN_USE, &state->flags))
6135 		status = nfs41_check_expired_locks(state);
6136 	if (status != NFS_OK)
6137 		status = nfs4_lock_expired(state, request);
6138 	return status;
6139 }
6140 #endif
6141 
6142 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6143 {
6144 	struct nfs_inode *nfsi = NFS_I(state->inode);
6145 	struct nfs4_state_owner *sp = state->owner;
6146 	unsigned char fl_flags = request->fl_flags;
6147 	int status;
6148 
6149 	request->fl_flags |= FL_ACCESS;
6150 	status = locks_lock_inode_wait(state->inode, request);
6151 	if (status < 0)
6152 		goto out;
6153 	mutex_lock(&sp->so_delegreturn_mutex);
6154 	down_read(&nfsi->rwsem);
6155 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6156 		/* Yes: cache locks! */
6157 		/* ...but avoid races with delegation recall... */
6158 		request->fl_flags = fl_flags & ~FL_SLEEP;
6159 		status = locks_lock_inode_wait(state->inode, request);
6160 		up_read(&nfsi->rwsem);
6161 		mutex_unlock(&sp->so_delegreturn_mutex);
6162 		goto out;
6163 	}
6164 	up_read(&nfsi->rwsem);
6165 	mutex_unlock(&sp->so_delegreturn_mutex);
6166 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6167 out:
6168 	request->fl_flags = fl_flags;
6169 	return status;
6170 }
6171 
6172 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6173 {
6174 	struct nfs4_exception exception = {
6175 		.state = state,
6176 		.inode = state->inode,
6177 	};
6178 	int err;
6179 
6180 	do {
6181 		err = _nfs4_proc_setlk(state, cmd, request);
6182 		if (err == -NFS4ERR_DENIED)
6183 			err = -EAGAIN;
6184 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
6185 				err, &exception);
6186 	} while (exception.retry);
6187 	return err;
6188 }
6189 
6190 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6191 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6192 
6193 static int
6194 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6195 			struct file_lock *request)
6196 {
6197 	int		status = -ERESTARTSYS;
6198 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
6199 
6200 	while(!signalled()) {
6201 		status = nfs4_proc_setlk(state, cmd, request);
6202 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6203 			break;
6204 		freezable_schedule_timeout_interruptible(timeout);
6205 		timeout *= 2;
6206 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6207 		status = -ERESTARTSYS;
6208 	}
6209 	return status;
6210 }
6211 
6212 #ifdef CONFIG_NFS_V4_1
6213 struct nfs4_lock_waiter {
6214 	struct task_struct	*task;
6215 	struct inode		*inode;
6216 	struct nfs_lowner	*owner;
6217 	bool			notified;
6218 };
6219 
6220 static int
6221 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6222 {
6223 	int ret;
6224 	struct cb_notify_lock_args *cbnl = key;
6225 	struct nfs4_lock_waiter	*waiter	= wait->private;
6226 	struct nfs_lowner	*lowner = &cbnl->cbnl_owner,
6227 				*wowner = waiter->owner;
6228 
6229 	/* Only wake if the callback was for the same owner */
6230 	if (lowner->clientid != wowner->clientid ||
6231 	    lowner->id != wowner->id		 ||
6232 	    lowner->s_dev != wowner->s_dev)
6233 		return 0;
6234 
6235 	/* Make sure it's for the right inode */
6236 	if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6237 		return 0;
6238 
6239 	waiter->notified = true;
6240 
6241 	/* override "private" so we can use default_wake_function */
6242 	wait->private = waiter->task;
6243 	ret = autoremove_wake_function(wait, mode, flags, key);
6244 	wait->private = waiter;
6245 	return ret;
6246 }
6247 
6248 static int
6249 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6250 {
6251 	int status = -ERESTARTSYS;
6252 	unsigned long flags;
6253 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6254 	struct nfs_server *server = NFS_SERVER(state->inode);
6255 	struct nfs_client *clp = server->nfs_client;
6256 	wait_queue_head_t *q = &clp->cl_lock_waitq;
6257 	struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6258 				    .id = lsp->ls_seqid.owner_id,
6259 				    .s_dev = server->s_dev };
6260 	struct nfs4_lock_waiter waiter = { .task  = current,
6261 					   .inode = state->inode,
6262 					   .owner = &owner,
6263 					   .notified = false };
6264 	wait_queue_t wait;
6265 
6266 	/* Don't bother with waitqueue if we don't expect a callback */
6267 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6268 		return nfs4_retry_setlk_simple(state, cmd, request);
6269 
6270 	init_wait(&wait);
6271 	wait.private = &waiter;
6272 	wait.func = nfs4_wake_lock_waiter;
6273 	add_wait_queue(q, &wait);
6274 
6275 	while(!signalled()) {
6276 		status = nfs4_proc_setlk(state, cmd, request);
6277 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6278 			break;
6279 
6280 		status = -ERESTARTSYS;
6281 		spin_lock_irqsave(&q->lock, flags);
6282 		if (waiter.notified) {
6283 			spin_unlock_irqrestore(&q->lock, flags);
6284 			continue;
6285 		}
6286 		set_current_state(TASK_INTERRUPTIBLE);
6287 		spin_unlock_irqrestore(&q->lock, flags);
6288 
6289 		freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6290 	}
6291 
6292 	finish_wait(q, &wait);
6293 	return status;
6294 }
6295 #else /* !CONFIG_NFS_V4_1 */
6296 static inline int
6297 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6298 {
6299 	return nfs4_retry_setlk_simple(state, cmd, request);
6300 }
6301 #endif
6302 
6303 static int
6304 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6305 {
6306 	struct nfs_open_context *ctx;
6307 	struct nfs4_state *state;
6308 	int status;
6309 
6310 	/* verify open state */
6311 	ctx = nfs_file_open_context(filp);
6312 	state = ctx->state;
6313 
6314 	if (request->fl_start < 0 || request->fl_end < 0)
6315 		return -EINVAL;
6316 
6317 	if (IS_GETLK(cmd)) {
6318 		if (state != NULL)
6319 			return nfs4_proc_getlk(state, F_GETLK, request);
6320 		return 0;
6321 	}
6322 
6323 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6324 		return -EINVAL;
6325 
6326 	if (request->fl_type == F_UNLCK) {
6327 		if (state != NULL)
6328 			return nfs4_proc_unlck(state, cmd, request);
6329 		return 0;
6330 	}
6331 
6332 	if (state == NULL)
6333 		return -ENOLCK;
6334 
6335 	if ((request->fl_flags & FL_POSIX) &&
6336 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6337 		return -ENOLCK;
6338 
6339 	/*
6340 	 * Don't rely on the VFS having checked the file open mode,
6341 	 * since it won't do this for flock() locks.
6342 	 */
6343 	switch (request->fl_type) {
6344 	case F_RDLCK:
6345 		if (!(filp->f_mode & FMODE_READ))
6346 			return -EBADF;
6347 		break;
6348 	case F_WRLCK:
6349 		if (!(filp->f_mode & FMODE_WRITE))
6350 			return -EBADF;
6351 	}
6352 
6353 	status = nfs4_set_lock_state(state, request);
6354 	if (status != 0)
6355 		return status;
6356 
6357 	return nfs4_retry_setlk(state, cmd, request);
6358 }
6359 
6360 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6361 {
6362 	struct nfs_server *server = NFS_SERVER(state->inode);
6363 	int err;
6364 
6365 	err = nfs4_set_lock_state(state, fl);
6366 	if (err != 0)
6367 		return err;
6368 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6369 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6370 }
6371 
6372 struct nfs_release_lockowner_data {
6373 	struct nfs4_lock_state *lsp;
6374 	struct nfs_server *server;
6375 	struct nfs_release_lockowner_args args;
6376 	struct nfs_release_lockowner_res res;
6377 	unsigned long timestamp;
6378 };
6379 
6380 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6381 {
6382 	struct nfs_release_lockowner_data *data = calldata;
6383 	struct nfs_server *server = data->server;
6384 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6385 			     &data->args.seq_args, &data->res.seq_res, task);
6386 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6387 	data->timestamp = jiffies;
6388 }
6389 
6390 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6391 {
6392 	struct nfs_release_lockowner_data *data = calldata;
6393 	struct nfs_server *server = data->server;
6394 
6395 	nfs40_sequence_done(task, &data->res.seq_res);
6396 
6397 	switch (task->tk_status) {
6398 	case 0:
6399 		renew_lease(server, data->timestamp);
6400 		break;
6401 	case -NFS4ERR_STALE_CLIENTID:
6402 	case -NFS4ERR_EXPIRED:
6403 		nfs4_schedule_lease_recovery(server->nfs_client);
6404 		break;
6405 	case -NFS4ERR_LEASE_MOVED:
6406 	case -NFS4ERR_DELAY:
6407 		if (nfs4_async_handle_error(task, server,
6408 					    NULL, NULL) == -EAGAIN)
6409 			rpc_restart_call_prepare(task);
6410 	}
6411 }
6412 
6413 static void nfs4_release_lockowner_release(void *calldata)
6414 {
6415 	struct nfs_release_lockowner_data *data = calldata;
6416 	nfs4_free_lock_state(data->server, data->lsp);
6417 	kfree(calldata);
6418 }
6419 
6420 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6421 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6422 	.rpc_call_done = nfs4_release_lockowner_done,
6423 	.rpc_release = nfs4_release_lockowner_release,
6424 };
6425 
6426 static void
6427 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6428 {
6429 	struct nfs_release_lockowner_data *data;
6430 	struct rpc_message msg = {
6431 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6432 	};
6433 
6434 	if (server->nfs_client->cl_mvops->minor_version != 0)
6435 		return;
6436 
6437 	data = kmalloc(sizeof(*data), GFP_NOFS);
6438 	if (!data)
6439 		return;
6440 	data->lsp = lsp;
6441 	data->server = server;
6442 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6443 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6444 	data->args.lock_owner.s_dev = server->s_dev;
6445 
6446 	msg.rpc_argp = &data->args;
6447 	msg.rpc_resp = &data->res;
6448 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6449 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6450 }
6451 
6452 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6453 
6454 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6455 				   struct dentry *unused, struct inode *inode,
6456 				   const char *key, const void *buf,
6457 				   size_t buflen, int flags)
6458 {
6459 	return nfs4_proc_set_acl(inode, buf, buflen);
6460 }
6461 
6462 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6463 				   struct dentry *unused, struct inode *inode,
6464 				   const char *key, void *buf, size_t buflen)
6465 {
6466 	return nfs4_proc_get_acl(inode, buf, buflen);
6467 }
6468 
6469 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6470 {
6471 	return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6472 }
6473 
6474 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6475 
6476 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6477 				     struct dentry *unused, struct inode *inode,
6478 				     const char *key, const void *buf,
6479 				     size_t buflen, int flags)
6480 {
6481 	if (security_ismaclabel(key))
6482 		return nfs4_set_security_label(inode, buf, buflen);
6483 
6484 	return -EOPNOTSUPP;
6485 }
6486 
6487 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6488 				     struct dentry *unused, struct inode *inode,
6489 				     const char *key, void *buf, size_t buflen)
6490 {
6491 	if (security_ismaclabel(key))
6492 		return nfs4_get_security_label(inode, buf, buflen);
6493 	return -EOPNOTSUPP;
6494 }
6495 
6496 static ssize_t
6497 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6498 {
6499 	int len = 0;
6500 
6501 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6502 		len = security_inode_listsecurity(inode, list, list_len);
6503 		if (list_len && len > list_len)
6504 			return -ERANGE;
6505 	}
6506 	return len;
6507 }
6508 
6509 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6510 	.prefix = XATTR_SECURITY_PREFIX,
6511 	.get	= nfs4_xattr_get_nfs4_label,
6512 	.set	= nfs4_xattr_set_nfs4_label,
6513 };
6514 
6515 #else
6516 
6517 static ssize_t
6518 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6519 {
6520 	return 0;
6521 }
6522 
6523 #endif
6524 
6525 /*
6526  * nfs_fhget will use either the mounted_on_fileid or the fileid
6527  */
6528 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6529 {
6530 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6531 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6532 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6533 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6534 		return;
6535 
6536 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6537 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6538 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6539 	fattr->nlink = 2;
6540 }
6541 
6542 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6543 				   const struct qstr *name,
6544 				   struct nfs4_fs_locations *fs_locations,
6545 				   struct page *page)
6546 {
6547 	struct nfs_server *server = NFS_SERVER(dir);
6548 	u32 bitmask[3] = {
6549 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6550 	};
6551 	struct nfs4_fs_locations_arg args = {
6552 		.dir_fh = NFS_FH(dir),
6553 		.name = name,
6554 		.page = page,
6555 		.bitmask = bitmask,
6556 	};
6557 	struct nfs4_fs_locations_res res = {
6558 		.fs_locations = fs_locations,
6559 	};
6560 	struct rpc_message msg = {
6561 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6562 		.rpc_argp = &args,
6563 		.rpc_resp = &res,
6564 	};
6565 	int status;
6566 
6567 	dprintk("%s: start\n", __func__);
6568 
6569 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6570 	 * is not supported */
6571 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6572 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6573 	else
6574 		bitmask[0] |= FATTR4_WORD0_FILEID;
6575 
6576 	nfs_fattr_init(&fs_locations->fattr);
6577 	fs_locations->server = server;
6578 	fs_locations->nlocations = 0;
6579 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6580 	dprintk("%s: returned status = %d\n", __func__, status);
6581 	return status;
6582 }
6583 
6584 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6585 			   const struct qstr *name,
6586 			   struct nfs4_fs_locations *fs_locations,
6587 			   struct page *page)
6588 {
6589 	struct nfs4_exception exception = { };
6590 	int err;
6591 	do {
6592 		err = _nfs4_proc_fs_locations(client, dir, name,
6593 				fs_locations, page);
6594 		trace_nfs4_get_fs_locations(dir, name, err);
6595 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6596 				&exception);
6597 	} while (exception.retry);
6598 	return err;
6599 }
6600 
6601 /*
6602  * This operation also signals the server that this client is
6603  * performing migration recovery.  The server can stop returning
6604  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6605  * appended to this compound to identify the client ID which is
6606  * performing recovery.
6607  */
6608 static int _nfs40_proc_get_locations(struct inode *inode,
6609 				     struct nfs4_fs_locations *locations,
6610 				     struct page *page, struct rpc_cred *cred)
6611 {
6612 	struct nfs_server *server = NFS_SERVER(inode);
6613 	struct rpc_clnt *clnt = server->client;
6614 	u32 bitmask[2] = {
6615 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6616 	};
6617 	struct nfs4_fs_locations_arg args = {
6618 		.clientid	= server->nfs_client->cl_clientid,
6619 		.fh		= NFS_FH(inode),
6620 		.page		= page,
6621 		.bitmask	= bitmask,
6622 		.migration	= 1,		/* skip LOOKUP */
6623 		.renew		= 1,		/* append RENEW */
6624 	};
6625 	struct nfs4_fs_locations_res res = {
6626 		.fs_locations	= locations,
6627 		.migration	= 1,
6628 		.renew		= 1,
6629 	};
6630 	struct rpc_message msg = {
6631 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6632 		.rpc_argp	= &args,
6633 		.rpc_resp	= &res,
6634 		.rpc_cred	= cred,
6635 	};
6636 	unsigned long now = jiffies;
6637 	int status;
6638 
6639 	nfs_fattr_init(&locations->fattr);
6640 	locations->server = server;
6641 	locations->nlocations = 0;
6642 
6643 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6644 	nfs4_set_sequence_privileged(&args.seq_args);
6645 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6646 					&args.seq_args, &res.seq_res);
6647 	if (status)
6648 		return status;
6649 
6650 	renew_lease(server, now);
6651 	return 0;
6652 }
6653 
6654 #ifdef CONFIG_NFS_V4_1
6655 
6656 /*
6657  * This operation also signals the server that this client is
6658  * performing migration recovery.  The server can stop asserting
6659  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6660  * performing this operation is identified in the SEQUENCE
6661  * operation in this compound.
6662  *
6663  * When the client supports GETATTR(fs_locations_info), it can
6664  * be plumbed in here.
6665  */
6666 static int _nfs41_proc_get_locations(struct inode *inode,
6667 				     struct nfs4_fs_locations *locations,
6668 				     struct page *page, struct rpc_cred *cred)
6669 {
6670 	struct nfs_server *server = NFS_SERVER(inode);
6671 	struct rpc_clnt *clnt = server->client;
6672 	u32 bitmask[2] = {
6673 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6674 	};
6675 	struct nfs4_fs_locations_arg args = {
6676 		.fh		= NFS_FH(inode),
6677 		.page		= page,
6678 		.bitmask	= bitmask,
6679 		.migration	= 1,		/* skip LOOKUP */
6680 	};
6681 	struct nfs4_fs_locations_res res = {
6682 		.fs_locations	= locations,
6683 		.migration	= 1,
6684 	};
6685 	struct rpc_message msg = {
6686 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6687 		.rpc_argp	= &args,
6688 		.rpc_resp	= &res,
6689 		.rpc_cred	= cred,
6690 	};
6691 	int status;
6692 
6693 	nfs_fattr_init(&locations->fattr);
6694 	locations->server = server;
6695 	locations->nlocations = 0;
6696 
6697 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6698 	nfs4_set_sequence_privileged(&args.seq_args);
6699 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6700 					&args.seq_args, &res.seq_res);
6701 	if (status == NFS4_OK &&
6702 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6703 		status = -NFS4ERR_LEASE_MOVED;
6704 	return status;
6705 }
6706 
6707 #endif	/* CONFIG_NFS_V4_1 */
6708 
6709 /**
6710  * nfs4_proc_get_locations - discover locations for a migrated FSID
6711  * @inode: inode on FSID that is migrating
6712  * @locations: result of query
6713  * @page: buffer
6714  * @cred: credential to use for this operation
6715  *
6716  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6717  * operation failed, or a negative errno if a local error occurred.
6718  *
6719  * On success, "locations" is filled in, but if the server has
6720  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6721  * asserted.
6722  *
6723  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6724  * from this client that require migration recovery.
6725  */
6726 int nfs4_proc_get_locations(struct inode *inode,
6727 			    struct nfs4_fs_locations *locations,
6728 			    struct page *page, struct rpc_cred *cred)
6729 {
6730 	struct nfs_server *server = NFS_SERVER(inode);
6731 	struct nfs_client *clp = server->nfs_client;
6732 	const struct nfs4_mig_recovery_ops *ops =
6733 					clp->cl_mvops->mig_recovery_ops;
6734 	struct nfs4_exception exception = { };
6735 	int status;
6736 
6737 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6738 		(unsigned long long)server->fsid.major,
6739 		(unsigned long long)server->fsid.minor,
6740 		clp->cl_hostname);
6741 	nfs_display_fhandle(NFS_FH(inode), __func__);
6742 
6743 	do {
6744 		status = ops->get_locations(inode, locations, page, cred);
6745 		if (status != -NFS4ERR_DELAY)
6746 			break;
6747 		nfs4_handle_exception(server, status, &exception);
6748 	} while (exception.retry);
6749 	return status;
6750 }
6751 
6752 /*
6753  * This operation also signals the server that this client is
6754  * performing "lease moved" recovery.  The server can stop
6755  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6756  * is appended to this compound to identify the client ID which is
6757  * performing recovery.
6758  */
6759 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6760 {
6761 	struct nfs_server *server = NFS_SERVER(inode);
6762 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6763 	struct rpc_clnt *clnt = server->client;
6764 	struct nfs4_fsid_present_arg args = {
6765 		.fh		= NFS_FH(inode),
6766 		.clientid	= clp->cl_clientid,
6767 		.renew		= 1,		/* append RENEW */
6768 	};
6769 	struct nfs4_fsid_present_res res = {
6770 		.renew		= 1,
6771 	};
6772 	struct rpc_message msg = {
6773 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6774 		.rpc_argp	= &args,
6775 		.rpc_resp	= &res,
6776 		.rpc_cred	= cred,
6777 	};
6778 	unsigned long now = jiffies;
6779 	int status;
6780 
6781 	res.fh = nfs_alloc_fhandle();
6782 	if (res.fh == NULL)
6783 		return -ENOMEM;
6784 
6785 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6786 	nfs4_set_sequence_privileged(&args.seq_args);
6787 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6788 						&args.seq_args, &res.seq_res);
6789 	nfs_free_fhandle(res.fh);
6790 	if (status)
6791 		return status;
6792 
6793 	do_renew_lease(clp, now);
6794 	return 0;
6795 }
6796 
6797 #ifdef CONFIG_NFS_V4_1
6798 
6799 /*
6800  * This operation also signals the server that this client is
6801  * performing "lease moved" recovery.  The server can stop asserting
6802  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6803  * this operation is identified in the SEQUENCE operation in this
6804  * compound.
6805  */
6806 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6807 {
6808 	struct nfs_server *server = NFS_SERVER(inode);
6809 	struct rpc_clnt *clnt = server->client;
6810 	struct nfs4_fsid_present_arg args = {
6811 		.fh		= NFS_FH(inode),
6812 	};
6813 	struct nfs4_fsid_present_res res = {
6814 	};
6815 	struct rpc_message msg = {
6816 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6817 		.rpc_argp	= &args,
6818 		.rpc_resp	= &res,
6819 		.rpc_cred	= cred,
6820 	};
6821 	int status;
6822 
6823 	res.fh = nfs_alloc_fhandle();
6824 	if (res.fh == NULL)
6825 		return -ENOMEM;
6826 
6827 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6828 	nfs4_set_sequence_privileged(&args.seq_args);
6829 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6830 						&args.seq_args, &res.seq_res);
6831 	nfs_free_fhandle(res.fh);
6832 	if (status == NFS4_OK &&
6833 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6834 		status = -NFS4ERR_LEASE_MOVED;
6835 	return status;
6836 }
6837 
6838 #endif	/* CONFIG_NFS_V4_1 */
6839 
6840 /**
6841  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6842  * @inode: inode on FSID to check
6843  * @cred: credential to use for this operation
6844  *
6845  * Server indicates whether the FSID is present, moved, or not
6846  * recognized.  This operation is necessary to clear a LEASE_MOVED
6847  * condition for this client ID.
6848  *
6849  * Returns NFS4_OK if the FSID is present on this server,
6850  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6851  *  NFS4ERR code if some error occurred on the server, or a
6852  *  negative errno if a local failure occurred.
6853  */
6854 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6855 {
6856 	struct nfs_server *server = NFS_SERVER(inode);
6857 	struct nfs_client *clp = server->nfs_client;
6858 	const struct nfs4_mig_recovery_ops *ops =
6859 					clp->cl_mvops->mig_recovery_ops;
6860 	struct nfs4_exception exception = { };
6861 	int status;
6862 
6863 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6864 		(unsigned long long)server->fsid.major,
6865 		(unsigned long long)server->fsid.minor,
6866 		clp->cl_hostname);
6867 	nfs_display_fhandle(NFS_FH(inode), __func__);
6868 
6869 	do {
6870 		status = ops->fsid_present(inode, cred);
6871 		if (status != -NFS4ERR_DELAY)
6872 			break;
6873 		nfs4_handle_exception(server, status, &exception);
6874 	} while (exception.retry);
6875 	return status;
6876 }
6877 
6878 /**
6879  * If 'use_integrity' is true and the state managment nfs_client
6880  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6881  * and the machine credential as per RFC3530bis and RFC5661 Security
6882  * Considerations sections. Otherwise, just use the user cred with the
6883  * filesystem's rpc_client.
6884  */
6885 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6886 {
6887 	int status;
6888 	struct nfs4_secinfo_arg args = {
6889 		.dir_fh = NFS_FH(dir),
6890 		.name   = name,
6891 	};
6892 	struct nfs4_secinfo_res res = {
6893 		.flavors     = flavors,
6894 	};
6895 	struct rpc_message msg = {
6896 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6897 		.rpc_argp = &args,
6898 		.rpc_resp = &res,
6899 	};
6900 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6901 	struct rpc_cred *cred = NULL;
6902 
6903 	if (use_integrity) {
6904 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6905 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6906 		msg.rpc_cred = cred;
6907 	}
6908 
6909 	dprintk("NFS call  secinfo %s\n", name->name);
6910 
6911 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6912 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6913 
6914 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6915 				&res.seq_res, 0);
6916 	dprintk("NFS reply  secinfo: %d\n", status);
6917 
6918 	if (cred)
6919 		put_rpccred(cred);
6920 
6921 	return status;
6922 }
6923 
6924 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6925 		      struct nfs4_secinfo_flavors *flavors)
6926 {
6927 	struct nfs4_exception exception = { };
6928 	int err;
6929 	do {
6930 		err = -NFS4ERR_WRONGSEC;
6931 
6932 		/* try to use integrity protection with machine cred */
6933 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6934 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6935 
6936 		/*
6937 		 * if unable to use integrity protection, or SECINFO with
6938 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6939 		 * disallowed by spec, but exists in deployed servers) use
6940 		 * the current filesystem's rpc_client and the user cred.
6941 		 */
6942 		if (err == -NFS4ERR_WRONGSEC)
6943 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6944 
6945 		trace_nfs4_secinfo(dir, name, err);
6946 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6947 				&exception);
6948 	} while (exception.retry);
6949 	return err;
6950 }
6951 
6952 #ifdef CONFIG_NFS_V4_1
6953 /*
6954  * Check the exchange flags returned by the server for invalid flags, having
6955  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6956  * DS flags set.
6957  */
6958 static int nfs4_check_cl_exchange_flags(u32 flags)
6959 {
6960 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6961 		goto out_inval;
6962 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6963 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6964 		goto out_inval;
6965 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6966 		goto out_inval;
6967 	return NFS_OK;
6968 out_inval:
6969 	return -NFS4ERR_INVAL;
6970 }
6971 
6972 static bool
6973 nfs41_same_server_scope(struct nfs41_server_scope *a,
6974 			struct nfs41_server_scope *b)
6975 {
6976 	if (a->server_scope_sz == b->server_scope_sz &&
6977 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6978 		return true;
6979 
6980 	return false;
6981 }
6982 
6983 static void
6984 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6985 {
6986 }
6987 
6988 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6989 	.rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6990 };
6991 
6992 /*
6993  * nfs4_proc_bind_one_conn_to_session()
6994  *
6995  * The 4.1 client currently uses the same TCP connection for the
6996  * fore and backchannel.
6997  */
6998 static
6999 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7000 		struct rpc_xprt *xprt,
7001 		struct nfs_client *clp,
7002 		struct rpc_cred *cred)
7003 {
7004 	int status;
7005 	struct nfs41_bind_conn_to_session_args args = {
7006 		.client = clp,
7007 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
7008 	};
7009 	struct nfs41_bind_conn_to_session_res res;
7010 	struct rpc_message msg = {
7011 		.rpc_proc =
7012 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7013 		.rpc_argp = &args,
7014 		.rpc_resp = &res,
7015 		.rpc_cred = cred,
7016 	};
7017 	struct rpc_task_setup task_setup_data = {
7018 		.rpc_client = clnt,
7019 		.rpc_xprt = xprt,
7020 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
7021 		.rpc_message = &msg,
7022 		.flags = RPC_TASK_TIMEOUT,
7023 	};
7024 	struct rpc_task *task;
7025 
7026 	dprintk("--> %s\n", __func__);
7027 
7028 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7029 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7030 		args.dir = NFS4_CDFC4_FORE;
7031 
7032 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
7033 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
7034 		args.dir = NFS4_CDFC4_FORE;
7035 
7036 	task = rpc_run_task(&task_setup_data);
7037 	if (!IS_ERR(task)) {
7038 		status = task->tk_status;
7039 		rpc_put_task(task);
7040 	} else
7041 		status = PTR_ERR(task);
7042 	trace_nfs4_bind_conn_to_session(clp, status);
7043 	if (status == 0) {
7044 		if (memcmp(res.sessionid.data,
7045 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7046 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
7047 			status = -EIO;
7048 			goto out;
7049 		}
7050 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7051 			dprintk("NFS: %s: Unexpected direction from server\n",
7052 				__func__);
7053 			status = -EIO;
7054 			goto out;
7055 		}
7056 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7057 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
7058 				__func__);
7059 			status = -EIO;
7060 			goto out;
7061 		}
7062 	}
7063 out:
7064 	dprintk("<-- %s status= %d\n", __func__, status);
7065 	return status;
7066 }
7067 
7068 struct rpc_bind_conn_calldata {
7069 	struct nfs_client *clp;
7070 	struct rpc_cred *cred;
7071 };
7072 
7073 static int
7074 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7075 		struct rpc_xprt *xprt,
7076 		void *calldata)
7077 {
7078 	struct rpc_bind_conn_calldata *p = calldata;
7079 
7080 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7081 }
7082 
7083 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7084 {
7085 	struct rpc_bind_conn_calldata data = {
7086 		.clp = clp,
7087 		.cred = cred,
7088 	};
7089 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7090 			nfs4_proc_bind_conn_to_session_callback, &data);
7091 }
7092 
7093 /*
7094  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7095  * and operations we'd like to see to enable certain features in the allow map
7096  */
7097 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7098 	.how = SP4_MACH_CRED,
7099 	.enforce.u.words = {
7100 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7101 		      1 << (OP_EXCHANGE_ID - 32) |
7102 		      1 << (OP_CREATE_SESSION - 32) |
7103 		      1 << (OP_DESTROY_SESSION - 32) |
7104 		      1 << (OP_DESTROY_CLIENTID - 32)
7105 	},
7106 	.allow.u.words = {
7107 		[0] = 1 << (OP_CLOSE) |
7108 		      1 << (OP_OPEN_DOWNGRADE) |
7109 		      1 << (OP_LOCKU) |
7110 		      1 << (OP_DELEGRETURN) |
7111 		      1 << (OP_COMMIT),
7112 		[1] = 1 << (OP_SECINFO - 32) |
7113 		      1 << (OP_SECINFO_NO_NAME - 32) |
7114 		      1 << (OP_LAYOUTRETURN - 32) |
7115 		      1 << (OP_TEST_STATEID - 32) |
7116 		      1 << (OP_FREE_STATEID - 32) |
7117 		      1 << (OP_WRITE - 32)
7118 	}
7119 };
7120 
7121 /*
7122  * Select the state protection mode for client `clp' given the server results
7123  * from exchange_id in `sp'.
7124  *
7125  * Returns 0 on success, negative errno otherwise.
7126  */
7127 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7128 				 struct nfs41_state_protection *sp)
7129 {
7130 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7131 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7132 		      1 << (OP_EXCHANGE_ID - 32) |
7133 		      1 << (OP_CREATE_SESSION - 32) |
7134 		      1 << (OP_DESTROY_SESSION - 32) |
7135 		      1 << (OP_DESTROY_CLIENTID - 32)
7136 	};
7137 	unsigned int i;
7138 
7139 	if (sp->how == SP4_MACH_CRED) {
7140 		/* Print state protect result */
7141 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7142 		for (i = 0; i <= LAST_NFS4_OP; i++) {
7143 			if (test_bit(i, sp->enforce.u.longs))
7144 				dfprintk(MOUNT, "  enforce op %d\n", i);
7145 			if (test_bit(i, sp->allow.u.longs))
7146 				dfprintk(MOUNT, "  allow op %d\n", i);
7147 		}
7148 
7149 		/* make sure nothing is on enforce list that isn't supported */
7150 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7151 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7152 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7153 				return -EINVAL;
7154 			}
7155 		}
7156 
7157 		/*
7158 		 * Minimal mode - state operations are allowed to use machine
7159 		 * credential.  Note this already happens by default, so the
7160 		 * client doesn't have to do anything more than the negotiation.
7161 		 *
7162 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
7163 		 *       we're already using the machine cred for exchange_id
7164 		 *       and will never use a different cred.
7165 		 */
7166 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7167 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7168 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7169 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7170 			dfprintk(MOUNT, "sp4_mach_cred:\n");
7171 			dfprintk(MOUNT, "  minimal mode enabled\n");
7172 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7173 		} else {
7174 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7175 			return -EINVAL;
7176 		}
7177 
7178 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7179 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7180 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7181 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
7182 			dfprintk(MOUNT, "  cleanup mode enabled\n");
7183 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7184 		}
7185 
7186 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7187 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7188 			set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7189 				&clp->cl_sp4_flags);
7190 		}
7191 
7192 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7193 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7194 			dfprintk(MOUNT, "  secinfo mode enabled\n");
7195 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7196 		}
7197 
7198 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7199 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7200 			dfprintk(MOUNT, "  stateid mode enabled\n");
7201 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7202 		}
7203 
7204 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7205 			dfprintk(MOUNT, "  write mode enabled\n");
7206 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7207 		}
7208 
7209 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7210 			dfprintk(MOUNT, "  commit mode enabled\n");
7211 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7212 		}
7213 	}
7214 
7215 	return 0;
7216 }
7217 
7218 struct nfs41_exchange_id_data {
7219 	struct nfs41_exchange_id_res res;
7220 	struct nfs41_exchange_id_args args;
7221 	struct rpc_xprt *xprt;
7222 	int rpc_status;
7223 };
7224 
7225 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7226 {
7227 	struct nfs41_exchange_id_data *cdata =
7228 					(struct nfs41_exchange_id_data *)data;
7229 	struct nfs_client *clp = cdata->args.client;
7230 	int status = task->tk_status;
7231 
7232 	trace_nfs4_exchange_id(clp, status);
7233 
7234 	if (status == 0)
7235 		status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7236 
7237 	if (cdata->xprt && status == 0) {
7238 		status = nfs4_detect_session_trunking(clp, &cdata->res,
7239 						      cdata->xprt);
7240 		goto out;
7241 	}
7242 
7243 	if (status  == 0)
7244 		status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7245 
7246 	if (status == 0) {
7247 		clp->cl_clientid = cdata->res.clientid;
7248 		clp->cl_exchange_flags = cdata->res.flags;
7249 		/* Client ID is not confirmed */
7250 		if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7251 			clear_bit(NFS4_SESSION_ESTABLISHED,
7252 			&clp->cl_session->session_state);
7253 			clp->cl_seqid = cdata->res.seqid;
7254 		}
7255 
7256 		kfree(clp->cl_serverowner);
7257 		clp->cl_serverowner = cdata->res.server_owner;
7258 		cdata->res.server_owner = NULL;
7259 
7260 		/* use the most recent implementation id */
7261 		kfree(clp->cl_implid);
7262 		clp->cl_implid = cdata->res.impl_id;
7263 		cdata->res.impl_id = NULL;
7264 
7265 		if (clp->cl_serverscope != NULL &&
7266 		    !nfs41_same_server_scope(clp->cl_serverscope,
7267 					cdata->res.server_scope)) {
7268 			dprintk("%s: server_scope mismatch detected\n",
7269 				__func__);
7270 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7271 			kfree(clp->cl_serverscope);
7272 			clp->cl_serverscope = NULL;
7273 		}
7274 
7275 		if (clp->cl_serverscope == NULL) {
7276 			clp->cl_serverscope = cdata->res.server_scope;
7277 			cdata->res.server_scope = NULL;
7278 		}
7279 		/* Save the EXCHANGE_ID verifier session trunk tests */
7280 		memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7281 		       sizeof(clp->cl_confirm.data));
7282 	}
7283 out:
7284 	cdata->rpc_status = status;
7285 	return;
7286 }
7287 
7288 static void nfs4_exchange_id_release(void *data)
7289 {
7290 	struct nfs41_exchange_id_data *cdata =
7291 					(struct nfs41_exchange_id_data *)data;
7292 
7293 	nfs_put_client(cdata->args.client);
7294 	if (cdata->xprt) {
7295 		xprt_put(cdata->xprt);
7296 		rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7297 	}
7298 	kfree(cdata->res.impl_id);
7299 	kfree(cdata->res.server_scope);
7300 	kfree(cdata->res.server_owner);
7301 	kfree(cdata);
7302 }
7303 
7304 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7305 	.rpc_call_done = nfs4_exchange_id_done,
7306 	.rpc_release = nfs4_exchange_id_release,
7307 };
7308 
7309 /*
7310  * _nfs4_proc_exchange_id()
7311  *
7312  * Wrapper for EXCHANGE_ID operation.
7313  */
7314 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7315 			u32 sp4_how, struct rpc_xprt *xprt)
7316 {
7317 	nfs4_verifier verifier;
7318 	struct rpc_message msg = {
7319 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7320 		.rpc_cred = cred,
7321 	};
7322 	struct rpc_task_setup task_setup_data = {
7323 		.rpc_client = clp->cl_rpcclient,
7324 		.callback_ops = &nfs4_exchange_id_call_ops,
7325 		.rpc_message = &msg,
7326 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7327 	};
7328 	struct nfs41_exchange_id_data *calldata;
7329 	struct rpc_task *task;
7330 	int status = -EIO;
7331 
7332 	if (!atomic_inc_not_zero(&clp->cl_count))
7333 		goto out;
7334 
7335 	status = -ENOMEM;
7336 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7337 	if (!calldata)
7338 		goto out;
7339 
7340 	if (!xprt)
7341 		nfs4_init_boot_verifier(clp, &verifier);
7342 
7343 	status = nfs4_init_uniform_client_string(clp);
7344 	if (status)
7345 		goto out_calldata;
7346 
7347 	dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7348 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
7349 		clp->cl_owner_id);
7350 
7351 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7352 						GFP_NOFS);
7353 	status = -ENOMEM;
7354 	if (unlikely(calldata->res.server_owner == NULL))
7355 		goto out_calldata;
7356 
7357 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7358 					GFP_NOFS);
7359 	if (unlikely(calldata->res.server_scope == NULL))
7360 		goto out_server_owner;
7361 
7362 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7363 	if (unlikely(calldata->res.impl_id == NULL))
7364 		goto out_server_scope;
7365 
7366 	switch (sp4_how) {
7367 	case SP4_NONE:
7368 		calldata->args.state_protect.how = SP4_NONE;
7369 		break;
7370 
7371 	case SP4_MACH_CRED:
7372 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7373 		break;
7374 
7375 	default:
7376 		/* unsupported! */
7377 		WARN_ON_ONCE(1);
7378 		status = -EINVAL;
7379 		goto out_impl_id;
7380 	}
7381 	if (xprt) {
7382 		calldata->xprt = xprt;
7383 		task_setup_data.rpc_xprt = xprt;
7384 		task_setup_data.flags =
7385 				RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7386 		calldata->args.verifier = &clp->cl_confirm;
7387 	} else {
7388 		calldata->args.verifier = &verifier;
7389 	}
7390 	calldata->args.client = clp;
7391 #ifdef CONFIG_NFS_V4_1_MIGRATION
7392 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7393 	EXCHGID4_FLAG_BIND_PRINC_STATEID |
7394 	EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7395 #else
7396 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7397 	EXCHGID4_FLAG_BIND_PRINC_STATEID,
7398 #endif
7399 	msg.rpc_argp = &calldata->args;
7400 	msg.rpc_resp = &calldata->res;
7401 	task_setup_data.callback_data = calldata;
7402 
7403 	task = rpc_run_task(&task_setup_data);
7404 	if (IS_ERR(task)) {
7405 	status = PTR_ERR(task);
7406 		goto out_impl_id;
7407 	}
7408 
7409 	if (!xprt) {
7410 		status = rpc_wait_for_completion_task(task);
7411 		if (!status)
7412 			status = calldata->rpc_status;
7413 	} else	/* session trunking test */
7414 		status = calldata->rpc_status;
7415 
7416 	rpc_put_task(task);
7417 out:
7418 	if (clp->cl_implid != NULL)
7419 		dprintk("NFS reply exchange_id: Server Implementation ID: "
7420 			"domain: %s, name: %s, date: %llu,%u\n",
7421 			clp->cl_implid->domain, clp->cl_implid->name,
7422 			clp->cl_implid->date.seconds,
7423 			clp->cl_implid->date.nseconds);
7424 	dprintk("NFS reply exchange_id: %d\n", status);
7425 	return status;
7426 
7427 out_impl_id:
7428 	kfree(calldata->res.impl_id);
7429 out_server_scope:
7430 	kfree(calldata->res.server_scope);
7431 out_server_owner:
7432 	kfree(calldata->res.server_owner);
7433 out_calldata:
7434 	kfree(calldata);
7435 	goto out;
7436 }
7437 
7438 /*
7439  * nfs4_proc_exchange_id()
7440  *
7441  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7442  *
7443  * Since the clientid has expired, all compounds using sessions
7444  * associated with the stale clientid will be returning
7445  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7446  * be in some phase of session reset.
7447  *
7448  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7449  */
7450 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7451 {
7452 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7453 	int status;
7454 
7455 	/* try SP4_MACH_CRED if krb5i/p	*/
7456 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
7457 	    authflavor == RPC_AUTH_GSS_KRB5P) {
7458 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7459 		if (!status)
7460 			return 0;
7461 	}
7462 
7463 	/* try SP4_NONE */
7464 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7465 }
7466 
7467 /**
7468  * nfs4_test_session_trunk
7469  *
7470  * This is an add_xprt_test() test function called from
7471  * rpc_clnt_setup_test_and_add_xprt.
7472  *
7473  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7474  * and is dereferrenced in nfs4_exchange_id_release
7475  *
7476  * Upon success, add the new transport to the rpc_clnt
7477  *
7478  * @clnt: struct rpc_clnt to get new transport
7479  * @xprt: the rpc_xprt to test
7480  * @data: call data for _nfs4_proc_exchange_id.
7481  */
7482 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7483 			    void *data)
7484 {
7485 	struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7486 	u32 sp4_how;
7487 
7488 	dprintk("--> %s try %s\n", __func__,
7489 		xprt->address_strings[RPC_DISPLAY_ADDR]);
7490 
7491 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7492 
7493 	/* Test connection for session trunking. Async exchange_id call */
7494 	return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7495 }
7496 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7497 
7498 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7499 		struct rpc_cred *cred)
7500 {
7501 	struct rpc_message msg = {
7502 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7503 		.rpc_argp = clp,
7504 		.rpc_cred = cred,
7505 	};
7506 	int status;
7507 
7508 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7509 	trace_nfs4_destroy_clientid(clp, status);
7510 	if (status)
7511 		dprintk("NFS: Got error %d from the server %s on "
7512 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
7513 	return status;
7514 }
7515 
7516 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7517 		struct rpc_cred *cred)
7518 {
7519 	unsigned int loop;
7520 	int ret;
7521 
7522 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7523 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7524 		switch (ret) {
7525 		case -NFS4ERR_DELAY:
7526 		case -NFS4ERR_CLIENTID_BUSY:
7527 			ssleep(1);
7528 			break;
7529 		default:
7530 			return ret;
7531 		}
7532 	}
7533 	return 0;
7534 }
7535 
7536 int nfs4_destroy_clientid(struct nfs_client *clp)
7537 {
7538 	struct rpc_cred *cred;
7539 	int ret = 0;
7540 
7541 	if (clp->cl_mvops->minor_version < 1)
7542 		goto out;
7543 	if (clp->cl_exchange_flags == 0)
7544 		goto out;
7545 	if (clp->cl_preserve_clid)
7546 		goto out;
7547 	cred = nfs4_get_clid_cred(clp);
7548 	ret = nfs4_proc_destroy_clientid(clp, cred);
7549 	if (cred)
7550 		put_rpccred(cred);
7551 	switch (ret) {
7552 	case 0:
7553 	case -NFS4ERR_STALE_CLIENTID:
7554 		clp->cl_exchange_flags = 0;
7555 	}
7556 out:
7557 	return ret;
7558 }
7559 
7560 struct nfs4_get_lease_time_data {
7561 	struct nfs4_get_lease_time_args *args;
7562 	struct nfs4_get_lease_time_res *res;
7563 	struct nfs_client *clp;
7564 };
7565 
7566 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7567 					void *calldata)
7568 {
7569 	struct nfs4_get_lease_time_data *data =
7570 			(struct nfs4_get_lease_time_data *)calldata;
7571 
7572 	dprintk("--> %s\n", __func__);
7573 	/* just setup sequence, do not trigger session recovery
7574 	   since we're invoked within one */
7575 	nfs41_setup_sequence(data->clp->cl_session,
7576 			&data->args->la_seq_args,
7577 			&data->res->lr_seq_res,
7578 			task);
7579 	dprintk("<-- %s\n", __func__);
7580 }
7581 
7582 /*
7583  * Called from nfs4_state_manager thread for session setup, so don't recover
7584  * from sequence operation or clientid errors.
7585  */
7586 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7587 {
7588 	struct nfs4_get_lease_time_data *data =
7589 			(struct nfs4_get_lease_time_data *)calldata;
7590 
7591 	dprintk("--> %s\n", __func__);
7592 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7593 		return;
7594 	switch (task->tk_status) {
7595 	case -NFS4ERR_DELAY:
7596 	case -NFS4ERR_GRACE:
7597 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7598 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7599 		task->tk_status = 0;
7600 		/* fall through */
7601 	case -NFS4ERR_RETRY_UNCACHED_REP:
7602 		rpc_restart_call_prepare(task);
7603 		return;
7604 	}
7605 	dprintk("<-- %s\n", __func__);
7606 }
7607 
7608 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7609 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7610 	.rpc_call_done = nfs4_get_lease_time_done,
7611 };
7612 
7613 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7614 {
7615 	struct rpc_task *task;
7616 	struct nfs4_get_lease_time_args args;
7617 	struct nfs4_get_lease_time_res res = {
7618 		.lr_fsinfo = fsinfo,
7619 	};
7620 	struct nfs4_get_lease_time_data data = {
7621 		.args = &args,
7622 		.res = &res,
7623 		.clp = clp,
7624 	};
7625 	struct rpc_message msg = {
7626 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7627 		.rpc_argp = &args,
7628 		.rpc_resp = &res,
7629 	};
7630 	struct rpc_task_setup task_setup = {
7631 		.rpc_client = clp->cl_rpcclient,
7632 		.rpc_message = &msg,
7633 		.callback_ops = &nfs4_get_lease_time_ops,
7634 		.callback_data = &data,
7635 		.flags = RPC_TASK_TIMEOUT,
7636 	};
7637 	int status;
7638 
7639 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7640 	nfs4_set_sequence_privileged(&args.la_seq_args);
7641 	dprintk("--> %s\n", __func__);
7642 	task = rpc_run_task(&task_setup);
7643 
7644 	if (IS_ERR(task))
7645 		status = PTR_ERR(task);
7646 	else {
7647 		status = task->tk_status;
7648 		rpc_put_task(task);
7649 	}
7650 	dprintk("<-- %s return %d\n", __func__, status);
7651 
7652 	return status;
7653 }
7654 
7655 /*
7656  * Initialize the values to be used by the client in CREATE_SESSION
7657  * If nfs4_init_session set the fore channel request and response sizes,
7658  * use them.
7659  *
7660  * Set the back channel max_resp_sz_cached to zero to force the client to
7661  * always set csa_cachethis to FALSE because the current implementation
7662  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7663  */
7664 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7665 				    struct rpc_clnt *clnt)
7666 {
7667 	unsigned int max_rqst_sz, max_resp_sz;
7668 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7669 
7670 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7671 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7672 
7673 	/* Fore channel attributes */
7674 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7675 	args->fc_attrs.max_resp_sz = max_resp_sz;
7676 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7677 	args->fc_attrs.max_reqs = max_session_slots;
7678 
7679 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7680 		"max_ops=%u max_reqs=%u\n",
7681 		__func__,
7682 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7683 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7684 
7685 	/* Back channel attributes */
7686 	args->bc_attrs.max_rqst_sz = max_bc_payload;
7687 	args->bc_attrs.max_resp_sz = max_bc_payload;
7688 	args->bc_attrs.max_resp_sz_cached = 0;
7689 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7690 	args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7691 
7692 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7693 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7694 		__func__,
7695 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7696 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7697 		args->bc_attrs.max_reqs);
7698 }
7699 
7700 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7701 		struct nfs41_create_session_res *res)
7702 {
7703 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7704 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7705 
7706 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7707 		return -EINVAL;
7708 	/*
7709 	 * Our requested max_ops is the minimum we need; we're not
7710 	 * prepared to break up compounds into smaller pieces than that.
7711 	 * So, no point even trying to continue if the server won't
7712 	 * cooperate:
7713 	 */
7714 	if (rcvd->max_ops < sent->max_ops)
7715 		return -EINVAL;
7716 	if (rcvd->max_reqs == 0)
7717 		return -EINVAL;
7718 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7719 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7720 	return 0;
7721 }
7722 
7723 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7724 		struct nfs41_create_session_res *res)
7725 {
7726 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7727 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7728 
7729 	if (!(res->flags & SESSION4_BACK_CHAN))
7730 		goto out;
7731 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7732 		return -EINVAL;
7733 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7734 		return -EINVAL;
7735 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7736 		return -EINVAL;
7737 	if (rcvd->max_ops > sent->max_ops)
7738 		return -EINVAL;
7739 	if (rcvd->max_reqs > sent->max_reqs)
7740 		return -EINVAL;
7741 out:
7742 	return 0;
7743 }
7744 
7745 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7746 				     struct nfs41_create_session_res *res)
7747 {
7748 	int ret;
7749 
7750 	ret = nfs4_verify_fore_channel_attrs(args, res);
7751 	if (ret)
7752 		return ret;
7753 	return nfs4_verify_back_channel_attrs(args, res);
7754 }
7755 
7756 static void nfs4_update_session(struct nfs4_session *session,
7757 		struct nfs41_create_session_res *res)
7758 {
7759 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7760 	/* Mark client id and session as being confirmed */
7761 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7762 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7763 	session->flags = res->flags;
7764 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7765 	if (res->flags & SESSION4_BACK_CHAN)
7766 		memcpy(&session->bc_attrs, &res->bc_attrs,
7767 				sizeof(session->bc_attrs));
7768 }
7769 
7770 static int _nfs4_proc_create_session(struct nfs_client *clp,
7771 		struct rpc_cred *cred)
7772 {
7773 	struct nfs4_session *session = clp->cl_session;
7774 	struct nfs41_create_session_args args = {
7775 		.client = clp,
7776 		.clientid = clp->cl_clientid,
7777 		.seqid = clp->cl_seqid,
7778 		.cb_program = NFS4_CALLBACK,
7779 	};
7780 	struct nfs41_create_session_res res;
7781 
7782 	struct rpc_message msg = {
7783 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7784 		.rpc_argp = &args,
7785 		.rpc_resp = &res,
7786 		.rpc_cred = cred,
7787 	};
7788 	int status;
7789 
7790 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7791 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7792 
7793 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7794 	trace_nfs4_create_session(clp, status);
7795 
7796 	switch (status) {
7797 	case -NFS4ERR_STALE_CLIENTID:
7798 	case -NFS4ERR_DELAY:
7799 	case -ETIMEDOUT:
7800 	case -EACCES:
7801 	case -EAGAIN:
7802 		goto out;
7803 	};
7804 
7805 	clp->cl_seqid++;
7806 	if (!status) {
7807 		/* Verify the session's negotiated channel_attrs values */
7808 		status = nfs4_verify_channel_attrs(&args, &res);
7809 		/* Increment the clientid slot sequence id */
7810 		if (status)
7811 			goto out;
7812 		nfs4_update_session(session, &res);
7813 	}
7814 out:
7815 	return status;
7816 }
7817 
7818 /*
7819  * Issues a CREATE_SESSION operation to the server.
7820  * It is the responsibility of the caller to verify the session is
7821  * expired before calling this routine.
7822  */
7823 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7824 {
7825 	int status;
7826 	unsigned *ptr;
7827 	struct nfs4_session *session = clp->cl_session;
7828 
7829 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7830 
7831 	status = _nfs4_proc_create_session(clp, cred);
7832 	if (status)
7833 		goto out;
7834 
7835 	/* Init or reset the session slot tables */
7836 	status = nfs4_setup_session_slot_tables(session);
7837 	dprintk("slot table setup returned %d\n", status);
7838 	if (status)
7839 		goto out;
7840 
7841 	ptr = (unsigned *)&session->sess_id.data[0];
7842 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7843 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7844 out:
7845 	dprintk("<-- %s\n", __func__);
7846 	return status;
7847 }
7848 
7849 /*
7850  * Issue the over-the-wire RPC DESTROY_SESSION.
7851  * The caller must serialize access to this routine.
7852  */
7853 int nfs4_proc_destroy_session(struct nfs4_session *session,
7854 		struct rpc_cred *cred)
7855 {
7856 	struct rpc_message msg = {
7857 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7858 		.rpc_argp = session,
7859 		.rpc_cred = cred,
7860 	};
7861 	int status = 0;
7862 
7863 	dprintk("--> nfs4_proc_destroy_session\n");
7864 
7865 	/* session is still being setup */
7866 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7867 		return 0;
7868 
7869 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7870 	trace_nfs4_destroy_session(session->clp, status);
7871 
7872 	if (status)
7873 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7874 			"Session has been destroyed regardless...\n", status);
7875 
7876 	dprintk("<-- nfs4_proc_destroy_session\n");
7877 	return status;
7878 }
7879 
7880 /*
7881  * Renew the cl_session lease.
7882  */
7883 struct nfs4_sequence_data {
7884 	struct nfs_client *clp;
7885 	struct nfs4_sequence_args args;
7886 	struct nfs4_sequence_res res;
7887 };
7888 
7889 static void nfs41_sequence_release(void *data)
7890 {
7891 	struct nfs4_sequence_data *calldata = data;
7892 	struct nfs_client *clp = calldata->clp;
7893 
7894 	if (atomic_read(&clp->cl_count) > 1)
7895 		nfs4_schedule_state_renewal(clp);
7896 	nfs_put_client(clp);
7897 	kfree(calldata);
7898 }
7899 
7900 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7901 {
7902 	switch(task->tk_status) {
7903 	case -NFS4ERR_DELAY:
7904 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7905 		return -EAGAIN;
7906 	default:
7907 		nfs4_schedule_lease_recovery(clp);
7908 	}
7909 	return 0;
7910 }
7911 
7912 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7913 {
7914 	struct nfs4_sequence_data *calldata = data;
7915 	struct nfs_client *clp = calldata->clp;
7916 
7917 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7918 		return;
7919 
7920 	trace_nfs4_sequence(clp, task->tk_status);
7921 	if (task->tk_status < 0) {
7922 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7923 		if (atomic_read(&clp->cl_count) == 1)
7924 			goto out;
7925 
7926 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7927 			rpc_restart_call_prepare(task);
7928 			return;
7929 		}
7930 	}
7931 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7932 out:
7933 	dprintk("<-- %s\n", __func__);
7934 }
7935 
7936 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7937 {
7938 	struct nfs4_sequence_data *calldata = data;
7939 	struct nfs_client *clp = calldata->clp;
7940 	struct nfs4_sequence_args *args;
7941 	struct nfs4_sequence_res *res;
7942 
7943 	args = task->tk_msg.rpc_argp;
7944 	res = task->tk_msg.rpc_resp;
7945 
7946 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7947 }
7948 
7949 static const struct rpc_call_ops nfs41_sequence_ops = {
7950 	.rpc_call_done = nfs41_sequence_call_done,
7951 	.rpc_call_prepare = nfs41_sequence_prepare,
7952 	.rpc_release = nfs41_sequence_release,
7953 };
7954 
7955 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7956 		struct rpc_cred *cred,
7957 		bool is_privileged)
7958 {
7959 	struct nfs4_sequence_data *calldata;
7960 	struct rpc_message msg = {
7961 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7962 		.rpc_cred = cred,
7963 	};
7964 	struct rpc_task_setup task_setup_data = {
7965 		.rpc_client = clp->cl_rpcclient,
7966 		.rpc_message = &msg,
7967 		.callback_ops = &nfs41_sequence_ops,
7968 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7969 	};
7970 
7971 	if (!atomic_inc_not_zero(&clp->cl_count))
7972 		return ERR_PTR(-EIO);
7973 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7974 	if (calldata == NULL) {
7975 		nfs_put_client(clp);
7976 		return ERR_PTR(-ENOMEM);
7977 	}
7978 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7979 	if (is_privileged)
7980 		nfs4_set_sequence_privileged(&calldata->args);
7981 	msg.rpc_argp = &calldata->args;
7982 	msg.rpc_resp = &calldata->res;
7983 	calldata->clp = clp;
7984 	task_setup_data.callback_data = calldata;
7985 
7986 	return rpc_run_task(&task_setup_data);
7987 }
7988 
7989 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7990 {
7991 	struct rpc_task *task;
7992 	int ret = 0;
7993 
7994 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7995 		return -EAGAIN;
7996 	task = _nfs41_proc_sequence(clp, cred, false);
7997 	if (IS_ERR(task))
7998 		ret = PTR_ERR(task);
7999 	else
8000 		rpc_put_task_async(task);
8001 	dprintk("<-- %s status=%d\n", __func__, ret);
8002 	return ret;
8003 }
8004 
8005 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8006 {
8007 	struct rpc_task *task;
8008 	int ret;
8009 
8010 	task = _nfs41_proc_sequence(clp, cred, true);
8011 	if (IS_ERR(task)) {
8012 		ret = PTR_ERR(task);
8013 		goto out;
8014 	}
8015 	ret = rpc_wait_for_completion_task(task);
8016 	if (!ret)
8017 		ret = task->tk_status;
8018 	rpc_put_task(task);
8019 out:
8020 	dprintk("<-- %s status=%d\n", __func__, ret);
8021 	return ret;
8022 }
8023 
8024 struct nfs4_reclaim_complete_data {
8025 	struct nfs_client *clp;
8026 	struct nfs41_reclaim_complete_args arg;
8027 	struct nfs41_reclaim_complete_res res;
8028 };
8029 
8030 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8031 {
8032 	struct nfs4_reclaim_complete_data *calldata = data;
8033 
8034 	nfs41_setup_sequence(calldata->clp->cl_session,
8035 			&calldata->arg.seq_args,
8036 			&calldata->res.seq_res,
8037 			task);
8038 }
8039 
8040 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8041 {
8042 	switch(task->tk_status) {
8043 	case 0:
8044 	case -NFS4ERR_COMPLETE_ALREADY:
8045 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
8046 		break;
8047 	case -NFS4ERR_DELAY:
8048 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
8049 		/* fall through */
8050 	case -NFS4ERR_RETRY_UNCACHED_REP:
8051 		return -EAGAIN;
8052 	default:
8053 		nfs4_schedule_lease_recovery(clp);
8054 	}
8055 	return 0;
8056 }
8057 
8058 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8059 {
8060 	struct nfs4_reclaim_complete_data *calldata = data;
8061 	struct nfs_client *clp = calldata->clp;
8062 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
8063 
8064 	dprintk("--> %s\n", __func__);
8065 	if (!nfs41_sequence_done(task, res))
8066 		return;
8067 
8068 	trace_nfs4_reclaim_complete(clp, task->tk_status);
8069 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8070 		rpc_restart_call_prepare(task);
8071 		return;
8072 	}
8073 	dprintk("<-- %s\n", __func__);
8074 }
8075 
8076 static void nfs4_free_reclaim_complete_data(void *data)
8077 {
8078 	struct nfs4_reclaim_complete_data *calldata = data;
8079 
8080 	kfree(calldata);
8081 }
8082 
8083 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8084 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
8085 	.rpc_call_done = nfs4_reclaim_complete_done,
8086 	.rpc_release = nfs4_free_reclaim_complete_data,
8087 };
8088 
8089 /*
8090  * Issue a global reclaim complete.
8091  */
8092 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8093 		struct rpc_cred *cred)
8094 {
8095 	struct nfs4_reclaim_complete_data *calldata;
8096 	struct rpc_task *task;
8097 	struct rpc_message msg = {
8098 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8099 		.rpc_cred = cred,
8100 	};
8101 	struct rpc_task_setup task_setup_data = {
8102 		.rpc_client = clp->cl_rpcclient,
8103 		.rpc_message = &msg,
8104 		.callback_ops = &nfs4_reclaim_complete_call_ops,
8105 		.flags = RPC_TASK_ASYNC,
8106 	};
8107 	int status = -ENOMEM;
8108 
8109 	dprintk("--> %s\n", __func__);
8110 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8111 	if (calldata == NULL)
8112 		goto out;
8113 	calldata->clp = clp;
8114 	calldata->arg.one_fs = 0;
8115 
8116 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8117 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8118 	msg.rpc_argp = &calldata->arg;
8119 	msg.rpc_resp = &calldata->res;
8120 	task_setup_data.callback_data = calldata;
8121 	task = rpc_run_task(&task_setup_data);
8122 	if (IS_ERR(task)) {
8123 		status = PTR_ERR(task);
8124 		goto out;
8125 	}
8126 	status = nfs4_wait_for_completion_rpc_task(task);
8127 	if (status == 0)
8128 		status = task->tk_status;
8129 	rpc_put_task(task);
8130 	return 0;
8131 out:
8132 	dprintk("<-- %s status=%d\n", __func__, status);
8133 	return status;
8134 }
8135 
8136 static void
8137 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8138 {
8139 	struct nfs4_layoutget *lgp = calldata;
8140 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8141 	struct nfs4_session *session = nfs4_get_session(server);
8142 
8143 	dprintk("--> %s\n", __func__);
8144 	nfs41_setup_sequence(session, &lgp->args.seq_args,
8145 				&lgp->res.seq_res, task);
8146 	dprintk("<-- %s\n", __func__);
8147 }
8148 
8149 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8150 {
8151 	struct nfs4_layoutget *lgp = calldata;
8152 
8153 	dprintk("--> %s\n", __func__);
8154 	nfs41_sequence_process(task, &lgp->res.seq_res);
8155 	dprintk("<-- %s\n", __func__);
8156 }
8157 
8158 static int
8159 nfs4_layoutget_handle_exception(struct rpc_task *task,
8160 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8161 {
8162 	struct inode *inode = lgp->args.inode;
8163 	struct nfs_server *server = NFS_SERVER(inode);
8164 	struct pnfs_layout_hdr *lo;
8165 	int nfs4err = task->tk_status;
8166 	int err, status = 0;
8167 	LIST_HEAD(head);
8168 
8169 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8170 
8171 	switch (nfs4err) {
8172 	case 0:
8173 		goto out;
8174 
8175 	/*
8176 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8177 	 * on the file. set tk_status to -ENODATA to tell upper layer to
8178 	 * retry go inband.
8179 	 */
8180 	case -NFS4ERR_LAYOUTUNAVAILABLE:
8181 		status = -ENODATA;
8182 		goto out;
8183 	/*
8184 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8185 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8186 	 */
8187 	case -NFS4ERR_BADLAYOUT:
8188 		status = -EOVERFLOW;
8189 		goto out;
8190 	/*
8191 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8192 	 * (or clients) writing to the same RAID stripe except when
8193 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
8194 	 *
8195 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
8196 	 * while, and then eventually give up.
8197 	 */
8198 	case -NFS4ERR_LAYOUTTRYLATER:
8199 		if (lgp->args.minlength == 0) {
8200 			status = -EOVERFLOW;
8201 			goto out;
8202 		}
8203 		status = -EBUSY;
8204 		break;
8205 	case -NFS4ERR_RECALLCONFLICT:
8206 		status = -ERECALLCONFLICT;
8207 		break;
8208 	case -NFS4ERR_EXPIRED:
8209 	case -NFS4ERR_BAD_STATEID:
8210 		exception->timeout = 0;
8211 		spin_lock(&inode->i_lock);
8212 		lo = NFS_I(inode)->layout;
8213 		/* If the open stateid was bad, then recover it. */
8214 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8215 		    nfs4_stateid_match_other(&lgp->args.stateid,
8216 					&lgp->args.ctx->state->stateid)) {
8217 			spin_unlock(&inode->i_lock);
8218 			exception->state = lgp->args.ctx->state;
8219 			break;
8220 		}
8221 
8222 		/*
8223 		 * Mark the bad layout state as invalid, then retry
8224 		 */
8225 		pnfs_mark_layout_stateid_invalid(lo, &head);
8226 		spin_unlock(&inode->i_lock);
8227 		pnfs_free_lseg_list(&head);
8228 		status = -EAGAIN;
8229 		goto out;
8230 	}
8231 
8232 	err = nfs4_handle_exception(server, nfs4err, exception);
8233 	if (!status) {
8234 		if (exception->retry)
8235 			status = -EAGAIN;
8236 		else
8237 			status = err;
8238 	}
8239 out:
8240 	dprintk("<-- %s\n", __func__);
8241 	return status;
8242 }
8243 
8244 static size_t max_response_pages(struct nfs_server *server)
8245 {
8246 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8247 	return nfs_page_array_len(0, max_resp_sz);
8248 }
8249 
8250 static void nfs4_free_pages(struct page **pages, size_t size)
8251 {
8252 	int i;
8253 
8254 	if (!pages)
8255 		return;
8256 
8257 	for (i = 0; i < size; i++) {
8258 		if (!pages[i])
8259 			break;
8260 		__free_page(pages[i]);
8261 	}
8262 	kfree(pages);
8263 }
8264 
8265 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8266 {
8267 	struct page **pages;
8268 	int i;
8269 
8270 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8271 	if (!pages) {
8272 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8273 		return NULL;
8274 	}
8275 
8276 	for (i = 0; i < size; i++) {
8277 		pages[i] = alloc_page(gfp_flags);
8278 		if (!pages[i]) {
8279 			dprintk("%s: failed to allocate page\n", __func__);
8280 			nfs4_free_pages(pages, size);
8281 			return NULL;
8282 		}
8283 	}
8284 
8285 	return pages;
8286 }
8287 
8288 static void nfs4_layoutget_release(void *calldata)
8289 {
8290 	struct nfs4_layoutget *lgp = calldata;
8291 	struct inode *inode = lgp->args.inode;
8292 	struct nfs_server *server = NFS_SERVER(inode);
8293 	size_t max_pages = max_response_pages(server);
8294 
8295 	dprintk("--> %s\n", __func__);
8296 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
8297 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
8298 	put_nfs_open_context(lgp->args.ctx);
8299 	kfree(calldata);
8300 	dprintk("<-- %s\n", __func__);
8301 }
8302 
8303 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8304 	.rpc_call_prepare = nfs4_layoutget_prepare,
8305 	.rpc_call_done = nfs4_layoutget_done,
8306 	.rpc_release = nfs4_layoutget_release,
8307 };
8308 
8309 struct pnfs_layout_segment *
8310 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8311 {
8312 	struct inode *inode = lgp->args.inode;
8313 	struct nfs_server *server = NFS_SERVER(inode);
8314 	size_t max_pages = max_response_pages(server);
8315 	struct rpc_task *task;
8316 	struct rpc_message msg = {
8317 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8318 		.rpc_argp = &lgp->args,
8319 		.rpc_resp = &lgp->res,
8320 		.rpc_cred = lgp->cred,
8321 	};
8322 	struct rpc_task_setup task_setup_data = {
8323 		.rpc_client = server->client,
8324 		.rpc_message = &msg,
8325 		.callback_ops = &nfs4_layoutget_call_ops,
8326 		.callback_data = lgp,
8327 		.flags = RPC_TASK_ASYNC,
8328 	};
8329 	struct pnfs_layout_segment *lseg = NULL;
8330 	struct nfs4_exception exception = {
8331 		.inode = inode,
8332 		.timeout = *timeout,
8333 	};
8334 	int status = 0;
8335 
8336 	dprintk("--> %s\n", __func__);
8337 
8338 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8339 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
8340 
8341 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8342 	if (!lgp->args.layout.pages) {
8343 		nfs4_layoutget_release(lgp);
8344 		return ERR_PTR(-ENOMEM);
8345 	}
8346 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8347 
8348 	lgp->res.layoutp = &lgp->args.layout;
8349 	lgp->res.seq_res.sr_slot = NULL;
8350 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8351 
8352 	task = rpc_run_task(&task_setup_data);
8353 	if (IS_ERR(task))
8354 		return ERR_CAST(task);
8355 	status = nfs4_wait_for_completion_rpc_task(task);
8356 	if (status == 0) {
8357 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8358 		*timeout = exception.timeout;
8359 	}
8360 
8361 	trace_nfs4_layoutget(lgp->args.ctx,
8362 			&lgp->args.range,
8363 			&lgp->res.range,
8364 			&lgp->res.stateid,
8365 			status);
8366 
8367 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8368 	if (status == 0 && lgp->res.layoutp->len)
8369 		lseg = pnfs_layout_process(lgp);
8370 	nfs4_sequence_free_slot(&lgp->res.seq_res);
8371 	rpc_put_task(task);
8372 	dprintk("<-- %s status=%d\n", __func__, status);
8373 	if (status)
8374 		return ERR_PTR(status);
8375 	return lseg;
8376 }
8377 
8378 static void
8379 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8380 {
8381 	struct nfs4_layoutreturn *lrp = calldata;
8382 
8383 	dprintk("--> %s\n", __func__);
8384 	nfs41_setup_sequence(lrp->clp->cl_session,
8385 			&lrp->args.seq_args,
8386 			&lrp->res.seq_res,
8387 			task);
8388 }
8389 
8390 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8391 {
8392 	struct nfs4_layoutreturn *lrp = calldata;
8393 	struct nfs_server *server;
8394 
8395 	dprintk("--> %s\n", __func__);
8396 
8397 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8398 		return;
8399 
8400 	server = NFS_SERVER(lrp->args.inode);
8401 	switch (task->tk_status) {
8402 	default:
8403 		task->tk_status = 0;
8404 	case 0:
8405 		break;
8406 	case -NFS4ERR_DELAY:
8407 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8408 			break;
8409 		nfs4_sequence_free_slot(&lrp->res.seq_res);
8410 		rpc_restart_call_prepare(task);
8411 		return;
8412 	}
8413 	dprintk("<-- %s\n", __func__);
8414 }
8415 
8416 static void nfs4_layoutreturn_release(void *calldata)
8417 {
8418 	struct nfs4_layoutreturn *lrp = calldata;
8419 	struct pnfs_layout_hdr *lo = lrp->args.layout;
8420 	LIST_HEAD(freeme);
8421 
8422 	dprintk("--> %s\n", __func__);
8423 	spin_lock(&lo->plh_inode->i_lock);
8424 	if (lrp->res.lrs_present) {
8425 		pnfs_mark_matching_lsegs_invalid(lo, &freeme,
8426 				&lrp->args.range,
8427 				be32_to_cpu(lrp->args.stateid.seqid));
8428 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8429 	} else
8430 		pnfs_mark_layout_stateid_invalid(lo, &freeme);
8431 	pnfs_clear_layoutreturn_waitbit(lo);
8432 	spin_unlock(&lo->plh_inode->i_lock);
8433 	nfs4_sequence_free_slot(&lrp->res.seq_res);
8434 	pnfs_free_lseg_list(&freeme);
8435 	pnfs_put_layout_hdr(lrp->args.layout);
8436 	nfs_iput_and_deactive(lrp->inode);
8437 	kfree(calldata);
8438 	dprintk("<-- %s\n", __func__);
8439 }
8440 
8441 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8442 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
8443 	.rpc_call_done = nfs4_layoutreturn_done,
8444 	.rpc_release = nfs4_layoutreturn_release,
8445 };
8446 
8447 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8448 {
8449 	struct rpc_task *task;
8450 	struct rpc_message msg = {
8451 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8452 		.rpc_argp = &lrp->args,
8453 		.rpc_resp = &lrp->res,
8454 		.rpc_cred = lrp->cred,
8455 	};
8456 	struct rpc_task_setup task_setup_data = {
8457 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
8458 		.rpc_message = &msg,
8459 		.callback_ops = &nfs4_layoutreturn_call_ops,
8460 		.callback_data = lrp,
8461 	};
8462 	int status = 0;
8463 
8464 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8465 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8466 			&task_setup_data.rpc_client, &msg);
8467 
8468 	dprintk("--> %s\n", __func__);
8469 	if (!sync) {
8470 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8471 		if (!lrp->inode) {
8472 			nfs4_layoutreturn_release(lrp);
8473 			return -EAGAIN;
8474 		}
8475 		task_setup_data.flags |= RPC_TASK_ASYNC;
8476 	}
8477 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8478 	task = rpc_run_task(&task_setup_data);
8479 	if (IS_ERR(task))
8480 		return PTR_ERR(task);
8481 	if (sync)
8482 		status = task->tk_status;
8483 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8484 	dprintk("<-- %s status=%d\n", __func__, status);
8485 	rpc_put_task(task);
8486 	return status;
8487 }
8488 
8489 static int
8490 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8491 		struct pnfs_device *pdev,
8492 		struct rpc_cred *cred)
8493 {
8494 	struct nfs4_getdeviceinfo_args args = {
8495 		.pdev = pdev,
8496 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
8497 			NOTIFY_DEVICEID4_DELETE,
8498 	};
8499 	struct nfs4_getdeviceinfo_res res = {
8500 		.pdev = pdev,
8501 	};
8502 	struct rpc_message msg = {
8503 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8504 		.rpc_argp = &args,
8505 		.rpc_resp = &res,
8506 		.rpc_cred = cred,
8507 	};
8508 	int status;
8509 
8510 	dprintk("--> %s\n", __func__);
8511 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8512 	if (res.notification & ~args.notify_types)
8513 		dprintk("%s: unsupported notification\n", __func__);
8514 	if (res.notification != args.notify_types)
8515 		pdev->nocache = 1;
8516 
8517 	dprintk("<-- %s status=%d\n", __func__, status);
8518 
8519 	return status;
8520 }
8521 
8522 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8523 		struct pnfs_device *pdev,
8524 		struct rpc_cred *cred)
8525 {
8526 	struct nfs4_exception exception = { };
8527 	int err;
8528 
8529 	do {
8530 		err = nfs4_handle_exception(server,
8531 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
8532 					&exception);
8533 	} while (exception.retry);
8534 	return err;
8535 }
8536 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8537 
8538 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8539 {
8540 	struct nfs4_layoutcommit_data *data = calldata;
8541 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8542 	struct nfs4_session *session = nfs4_get_session(server);
8543 
8544 	nfs41_setup_sequence(session,
8545 			&data->args.seq_args,
8546 			&data->res.seq_res,
8547 			task);
8548 }
8549 
8550 static void
8551 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8552 {
8553 	struct nfs4_layoutcommit_data *data = calldata;
8554 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8555 
8556 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8557 		return;
8558 
8559 	switch (task->tk_status) { /* Just ignore these failures */
8560 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8561 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8562 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8563 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8564 		task->tk_status = 0;
8565 	case 0:
8566 		break;
8567 	default:
8568 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8569 			rpc_restart_call_prepare(task);
8570 			return;
8571 		}
8572 	}
8573 }
8574 
8575 static void nfs4_layoutcommit_release(void *calldata)
8576 {
8577 	struct nfs4_layoutcommit_data *data = calldata;
8578 
8579 	pnfs_cleanup_layoutcommit(data);
8580 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8581 					   data->res.fattr);
8582 	put_rpccred(data->cred);
8583 	nfs_iput_and_deactive(data->inode);
8584 	kfree(data);
8585 }
8586 
8587 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8588 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8589 	.rpc_call_done = nfs4_layoutcommit_done,
8590 	.rpc_release = nfs4_layoutcommit_release,
8591 };
8592 
8593 int
8594 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8595 {
8596 	struct rpc_message msg = {
8597 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8598 		.rpc_argp = &data->args,
8599 		.rpc_resp = &data->res,
8600 		.rpc_cred = data->cred,
8601 	};
8602 	struct rpc_task_setup task_setup_data = {
8603 		.task = &data->task,
8604 		.rpc_client = NFS_CLIENT(data->args.inode),
8605 		.rpc_message = &msg,
8606 		.callback_ops = &nfs4_layoutcommit_ops,
8607 		.callback_data = data,
8608 	};
8609 	struct rpc_task *task;
8610 	int status = 0;
8611 
8612 	dprintk("NFS: initiating layoutcommit call. sync %d "
8613 		"lbw: %llu inode %lu\n", sync,
8614 		data->args.lastbytewritten,
8615 		data->args.inode->i_ino);
8616 
8617 	if (!sync) {
8618 		data->inode = nfs_igrab_and_active(data->args.inode);
8619 		if (data->inode == NULL) {
8620 			nfs4_layoutcommit_release(data);
8621 			return -EAGAIN;
8622 		}
8623 		task_setup_data.flags = RPC_TASK_ASYNC;
8624 	}
8625 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8626 	task = rpc_run_task(&task_setup_data);
8627 	if (IS_ERR(task))
8628 		return PTR_ERR(task);
8629 	if (sync)
8630 		status = task->tk_status;
8631 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8632 	dprintk("%s: status %d\n", __func__, status);
8633 	rpc_put_task(task);
8634 	return status;
8635 }
8636 
8637 /**
8638  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8639  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8640  */
8641 static int
8642 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8643 		    struct nfs_fsinfo *info,
8644 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8645 {
8646 	struct nfs41_secinfo_no_name_args args = {
8647 		.style = SECINFO_STYLE_CURRENT_FH,
8648 	};
8649 	struct nfs4_secinfo_res res = {
8650 		.flavors = flavors,
8651 	};
8652 	struct rpc_message msg = {
8653 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8654 		.rpc_argp = &args,
8655 		.rpc_resp = &res,
8656 	};
8657 	struct rpc_clnt *clnt = server->client;
8658 	struct rpc_cred *cred = NULL;
8659 	int status;
8660 
8661 	if (use_integrity) {
8662 		clnt = server->nfs_client->cl_rpcclient;
8663 		cred = nfs4_get_clid_cred(server->nfs_client);
8664 		msg.rpc_cred = cred;
8665 	}
8666 
8667 	dprintk("--> %s\n", __func__);
8668 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8669 				&res.seq_res, 0);
8670 	dprintk("<-- %s status=%d\n", __func__, status);
8671 
8672 	if (cred)
8673 		put_rpccred(cred);
8674 
8675 	return status;
8676 }
8677 
8678 static int
8679 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8680 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8681 {
8682 	struct nfs4_exception exception = { };
8683 	int err;
8684 	do {
8685 		/* first try using integrity protection */
8686 		err = -NFS4ERR_WRONGSEC;
8687 
8688 		/* try to use integrity protection with machine cred */
8689 		if (_nfs4_is_integrity_protected(server->nfs_client))
8690 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8691 							  flavors, true);
8692 
8693 		/*
8694 		 * if unable to use integrity protection, or SECINFO with
8695 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8696 		 * disallowed by spec, but exists in deployed servers) use
8697 		 * the current filesystem's rpc_client and the user cred.
8698 		 */
8699 		if (err == -NFS4ERR_WRONGSEC)
8700 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8701 							  flavors, false);
8702 
8703 		switch (err) {
8704 		case 0:
8705 		case -NFS4ERR_WRONGSEC:
8706 		case -ENOTSUPP:
8707 			goto out;
8708 		default:
8709 			err = nfs4_handle_exception(server, err, &exception);
8710 		}
8711 	} while (exception.retry);
8712 out:
8713 	return err;
8714 }
8715 
8716 static int
8717 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8718 		    struct nfs_fsinfo *info)
8719 {
8720 	int err;
8721 	struct page *page;
8722 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8723 	struct nfs4_secinfo_flavors *flavors;
8724 	struct nfs4_secinfo4 *secinfo;
8725 	int i;
8726 
8727 	page = alloc_page(GFP_KERNEL);
8728 	if (!page) {
8729 		err = -ENOMEM;
8730 		goto out;
8731 	}
8732 
8733 	flavors = page_address(page);
8734 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8735 
8736 	/*
8737 	 * Fall back on "guess and check" method if
8738 	 * the server doesn't support SECINFO_NO_NAME
8739 	 */
8740 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8741 		err = nfs4_find_root_sec(server, fhandle, info);
8742 		goto out_freepage;
8743 	}
8744 	if (err)
8745 		goto out_freepage;
8746 
8747 	for (i = 0; i < flavors->num_flavors; i++) {
8748 		secinfo = &flavors->flavors[i];
8749 
8750 		switch (secinfo->flavor) {
8751 		case RPC_AUTH_NULL:
8752 		case RPC_AUTH_UNIX:
8753 		case RPC_AUTH_GSS:
8754 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8755 					&secinfo->flavor_info);
8756 			break;
8757 		default:
8758 			flavor = RPC_AUTH_MAXFLAVOR;
8759 			break;
8760 		}
8761 
8762 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8763 			flavor = RPC_AUTH_MAXFLAVOR;
8764 
8765 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8766 			err = nfs4_lookup_root_sec(server, fhandle,
8767 						   info, flavor);
8768 			if (!err)
8769 				break;
8770 		}
8771 	}
8772 
8773 	if (flavor == RPC_AUTH_MAXFLAVOR)
8774 		err = -EPERM;
8775 
8776 out_freepage:
8777 	put_page(page);
8778 	if (err == -EACCES)
8779 		return -EPERM;
8780 out:
8781 	return err;
8782 }
8783 
8784 static int _nfs41_test_stateid(struct nfs_server *server,
8785 		nfs4_stateid *stateid,
8786 		struct rpc_cred *cred)
8787 {
8788 	int status;
8789 	struct nfs41_test_stateid_args args = {
8790 		.stateid = stateid,
8791 	};
8792 	struct nfs41_test_stateid_res res;
8793 	struct rpc_message msg = {
8794 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8795 		.rpc_argp = &args,
8796 		.rpc_resp = &res,
8797 		.rpc_cred = cred,
8798 	};
8799 	struct rpc_clnt *rpc_client = server->client;
8800 
8801 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8802 		&rpc_client, &msg);
8803 
8804 	dprintk("NFS call  test_stateid %p\n", stateid);
8805 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8806 	nfs4_set_sequence_privileged(&args.seq_args);
8807 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8808 			&args.seq_args, &res.seq_res);
8809 	if (status != NFS_OK) {
8810 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8811 		return status;
8812 	}
8813 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8814 	return -res.status;
8815 }
8816 
8817 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8818 		int err, struct nfs4_exception *exception)
8819 {
8820 	exception->retry = 0;
8821 	switch(err) {
8822 	case -NFS4ERR_DELAY:
8823 		nfs4_handle_exception(server, err, exception);
8824 		break;
8825 	case -NFS4ERR_BADSESSION:
8826 	case -NFS4ERR_BADSLOT:
8827 	case -NFS4ERR_BAD_HIGH_SLOT:
8828 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8829 	case -NFS4ERR_DEADSESSION:
8830 		nfs4_do_handle_exception(server, err, exception);
8831 	}
8832 }
8833 
8834 /**
8835  * nfs41_test_stateid - perform a TEST_STATEID operation
8836  *
8837  * @server: server / transport on which to perform the operation
8838  * @stateid: state ID to test
8839  * @cred: credential
8840  *
8841  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8842  * Otherwise a negative NFS4ERR value is returned if the operation
8843  * failed or the state ID is not currently valid.
8844  */
8845 static int nfs41_test_stateid(struct nfs_server *server,
8846 		nfs4_stateid *stateid,
8847 		struct rpc_cred *cred)
8848 {
8849 	struct nfs4_exception exception = { };
8850 	int err;
8851 	do {
8852 		err = _nfs41_test_stateid(server, stateid, cred);
8853 		nfs4_handle_delay_or_session_error(server, err, &exception);
8854 	} while (exception.retry);
8855 	return err;
8856 }
8857 
8858 struct nfs_free_stateid_data {
8859 	struct nfs_server *server;
8860 	struct nfs41_free_stateid_args args;
8861 	struct nfs41_free_stateid_res res;
8862 };
8863 
8864 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8865 {
8866 	struct nfs_free_stateid_data *data = calldata;
8867 	nfs41_setup_sequence(nfs4_get_session(data->server),
8868 			&data->args.seq_args,
8869 			&data->res.seq_res,
8870 			task);
8871 }
8872 
8873 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8874 {
8875 	struct nfs_free_stateid_data *data = calldata;
8876 
8877 	nfs41_sequence_done(task, &data->res.seq_res);
8878 
8879 	switch (task->tk_status) {
8880 	case -NFS4ERR_DELAY:
8881 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8882 			rpc_restart_call_prepare(task);
8883 	}
8884 }
8885 
8886 static void nfs41_free_stateid_release(void *calldata)
8887 {
8888 	kfree(calldata);
8889 }
8890 
8891 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8892 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8893 	.rpc_call_done = nfs41_free_stateid_done,
8894 	.rpc_release = nfs41_free_stateid_release,
8895 };
8896 
8897 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8898 		nfs4_stateid *stateid,
8899 		struct rpc_cred *cred,
8900 		bool privileged)
8901 {
8902 	struct rpc_message msg = {
8903 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8904 		.rpc_cred = cred,
8905 	};
8906 	struct rpc_task_setup task_setup = {
8907 		.rpc_client = server->client,
8908 		.rpc_message = &msg,
8909 		.callback_ops = &nfs41_free_stateid_ops,
8910 		.flags = RPC_TASK_ASYNC,
8911 	};
8912 	struct nfs_free_stateid_data *data;
8913 
8914 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8915 		&task_setup.rpc_client, &msg);
8916 
8917 	dprintk("NFS call  free_stateid %p\n", stateid);
8918 	data = kmalloc(sizeof(*data), GFP_NOFS);
8919 	if (!data)
8920 		return ERR_PTR(-ENOMEM);
8921 	data->server = server;
8922 	nfs4_stateid_copy(&data->args.stateid, stateid);
8923 
8924 	task_setup.callback_data = data;
8925 
8926 	msg.rpc_argp = &data->args;
8927 	msg.rpc_resp = &data->res;
8928 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8929 	if (privileged)
8930 		nfs4_set_sequence_privileged(&data->args.seq_args);
8931 
8932 	return rpc_run_task(&task_setup);
8933 }
8934 
8935 /**
8936  * nfs41_free_stateid - perform a FREE_STATEID operation
8937  *
8938  * @server: server / transport on which to perform the operation
8939  * @stateid: state ID to release
8940  * @cred: credential
8941  *
8942  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8943  * negative NFS4ERR value is returned.
8944  */
8945 static int nfs41_free_stateid(struct nfs_server *server,
8946 		nfs4_stateid *stateid,
8947 		struct rpc_cred *cred)
8948 {
8949 	struct rpc_task *task;
8950 	int ret;
8951 
8952 	task = _nfs41_free_stateid(server, stateid, cred, true);
8953 	if (IS_ERR(task))
8954 		return PTR_ERR(task);
8955 	ret = rpc_wait_for_completion_task(task);
8956 	if (!ret)
8957 		ret = task->tk_status;
8958 	rpc_put_task(task);
8959 	return ret;
8960 }
8961 
8962 static void
8963 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8964 {
8965 	struct rpc_task *task;
8966 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8967 
8968 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8969 	nfs4_free_lock_state(server, lsp);
8970 	if (IS_ERR(task))
8971 		return;
8972 	rpc_put_task(task);
8973 }
8974 
8975 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8976 		const nfs4_stateid *s2)
8977 {
8978 	if (s1->type != s2->type)
8979 		return false;
8980 
8981 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8982 		return false;
8983 
8984 	if (s1->seqid == s2->seqid)
8985 		return true;
8986 	if (s1->seqid == 0 || s2->seqid == 0)
8987 		return true;
8988 
8989 	return false;
8990 }
8991 
8992 #endif /* CONFIG_NFS_V4_1 */
8993 
8994 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8995 		const nfs4_stateid *s2)
8996 {
8997 	return nfs4_stateid_match(s1, s2);
8998 }
8999 
9000 
9001 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9002 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9003 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
9004 	.recover_open	= nfs4_open_reclaim,
9005 	.recover_lock	= nfs4_lock_reclaim,
9006 	.establish_clid = nfs4_init_clientid,
9007 	.detect_trunking = nfs40_discover_server_trunking,
9008 };
9009 
9010 #if defined(CONFIG_NFS_V4_1)
9011 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9012 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9013 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
9014 	.recover_open	= nfs4_open_reclaim,
9015 	.recover_lock	= nfs4_lock_reclaim,
9016 	.establish_clid = nfs41_init_clientid,
9017 	.reclaim_complete = nfs41_proc_reclaim_complete,
9018 	.detect_trunking = nfs41_discover_server_trunking,
9019 };
9020 #endif /* CONFIG_NFS_V4_1 */
9021 
9022 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9023 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9024 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
9025 	.recover_open	= nfs40_open_expired,
9026 	.recover_lock	= nfs4_lock_expired,
9027 	.establish_clid = nfs4_init_clientid,
9028 };
9029 
9030 #if defined(CONFIG_NFS_V4_1)
9031 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9032 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9033 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
9034 	.recover_open	= nfs41_open_expired,
9035 	.recover_lock	= nfs41_lock_expired,
9036 	.establish_clid = nfs41_init_clientid,
9037 };
9038 #endif /* CONFIG_NFS_V4_1 */
9039 
9040 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9041 	.sched_state_renewal = nfs4_proc_async_renew,
9042 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9043 	.renew_lease = nfs4_proc_renew,
9044 };
9045 
9046 #if defined(CONFIG_NFS_V4_1)
9047 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9048 	.sched_state_renewal = nfs41_proc_async_sequence,
9049 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9050 	.renew_lease = nfs4_proc_sequence,
9051 };
9052 #endif
9053 
9054 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9055 	.get_locations = _nfs40_proc_get_locations,
9056 	.fsid_present = _nfs40_proc_fsid_present,
9057 };
9058 
9059 #if defined(CONFIG_NFS_V4_1)
9060 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9061 	.get_locations = _nfs41_proc_get_locations,
9062 	.fsid_present = _nfs41_proc_fsid_present,
9063 };
9064 #endif	/* CONFIG_NFS_V4_1 */
9065 
9066 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9067 	.minor_version = 0,
9068 	.init_caps = NFS_CAP_READDIRPLUS
9069 		| NFS_CAP_ATOMIC_OPEN
9070 		| NFS_CAP_POSIX_LOCK,
9071 	.init_client = nfs40_init_client,
9072 	.shutdown_client = nfs40_shutdown_client,
9073 	.match_stateid = nfs4_match_stateid,
9074 	.find_root_sec = nfs4_find_root_sec,
9075 	.free_lock_state = nfs4_release_lockowner,
9076 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
9077 	.alloc_seqid = nfs_alloc_seqid,
9078 	.call_sync_ops = &nfs40_call_sync_ops,
9079 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9080 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9081 	.state_renewal_ops = &nfs40_state_renewal_ops,
9082 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
9083 };
9084 
9085 #if defined(CONFIG_NFS_V4_1)
9086 static struct nfs_seqid *
9087 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9088 {
9089 	return NULL;
9090 }
9091 
9092 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9093 	.minor_version = 1,
9094 	.init_caps = NFS_CAP_READDIRPLUS
9095 		| NFS_CAP_ATOMIC_OPEN
9096 		| NFS_CAP_POSIX_LOCK
9097 		| NFS_CAP_STATEID_NFSV41
9098 		| NFS_CAP_ATOMIC_OPEN_V1,
9099 	.init_client = nfs41_init_client,
9100 	.shutdown_client = nfs41_shutdown_client,
9101 	.match_stateid = nfs41_match_stateid,
9102 	.find_root_sec = nfs41_find_root_sec,
9103 	.free_lock_state = nfs41_free_lock_state,
9104 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
9105 	.alloc_seqid = nfs_alloc_no_seqid,
9106 	.session_trunk = nfs4_test_session_trunk,
9107 	.call_sync_ops = &nfs41_call_sync_ops,
9108 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9109 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9110 	.state_renewal_ops = &nfs41_state_renewal_ops,
9111 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
9112 };
9113 #endif
9114 
9115 #if defined(CONFIG_NFS_V4_2)
9116 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9117 	.minor_version = 2,
9118 	.init_caps = NFS_CAP_READDIRPLUS
9119 		| NFS_CAP_ATOMIC_OPEN
9120 		| NFS_CAP_POSIX_LOCK
9121 		| NFS_CAP_STATEID_NFSV41
9122 		| NFS_CAP_ATOMIC_OPEN_V1
9123 		| NFS_CAP_ALLOCATE
9124 		| NFS_CAP_COPY
9125 		| NFS_CAP_DEALLOCATE
9126 		| NFS_CAP_SEEK
9127 		| NFS_CAP_LAYOUTSTATS
9128 		| NFS_CAP_CLONE,
9129 	.init_client = nfs41_init_client,
9130 	.shutdown_client = nfs41_shutdown_client,
9131 	.match_stateid = nfs41_match_stateid,
9132 	.find_root_sec = nfs41_find_root_sec,
9133 	.free_lock_state = nfs41_free_lock_state,
9134 	.call_sync_ops = &nfs41_call_sync_ops,
9135 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
9136 	.alloc_seqid = nfs_alloc_no_seqid,
9137 	.session_trunk = nfs4_test_session_trunk,
9138 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9139 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9140 	.state_renewal_ops = &nfs41_state_renewal_ops,
9141 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
9142 };
9143 #endif
9144 
9145 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9146 	[0] = &nfs_v4_0_minor_ops,
9147 #if defined(CONFIG_NFS_V4_1)
9148 	[1] = &nfs_v4_1_minor_ops,
9149 #endif
9150 #if defined(CONFIG_NFS_V4_2)
9151 	[2] = &nfs_v4_2_minor_ops,
9152 #endif
9153 };
9154 
9155 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9156 {
9157 	ssize_t error, error2;
9158 
9159 	error = generic_listxattr(dentry, list, size);
9160 	if (error < 0)
9161 		return error;
9162 	if (list) {
9163 		list += error;
9164 		size -= error;
9165 	}
9166 
9167 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9168 	if (error2 < 0)
9169 		return error2;
9170 	return error + error2;
9171 }
9172 
9173 static const struct inode_operations nfs4_dir_inode_operations = {
9174 	.create		= nfs_create,
9175 	.lookup		= nfs_lookup,
9176 	.atomic_open	= nfs_atomic_open,
9177 	.link		= nfs_link,
9178 	.unlink		= nfs_unlink,
9179 	.symlink	= nfs_symlink,
9180 	.mkdir		= nfs_mkdir,
9181 	.rmdir		= nfs_rmdir,
9182 	.mknod		= nfs_mknod,
9183 	.rename		= nfs_rename,
9184 	.permission	= nfs_permission,
9185 	.getattr	= nfs_getattr,
9186 	.setattr	= nfs_setattr,
9187 	.getxattr	= generic_getxattr,
9188 	.setxattr	= generic_setxattr,
9189 	.listxattr	= nfs4_listxattr,
9190 	.removexattr	= generic_removexattr,
9191 };
9192 
9193 static const struct inode_operations nfs4_file_inode_operations = {
9194 	.permission	= nfs_permission,
9195 	.getattr	= nfs_getattr,
9196 	.setattr	= nfs_setattr,
9197 	.getxattr	= generic_getxattr,
9198 	.setxattr	= generic_setxattr,
9199 	.listxattr	= nfs4_listxattr,
9200 	.removexattr	= generic_removexattr,
9201 };
9202 
9203 const struct nfs_rpc_ops nfs_v4_clientops = {
9204 	.version	= 4,			/* protocol version */
9205 	.dentry_ops	= &nfs4_dentry_operations,
9206 	.dir_inode_ops	= &nfs4_dir_inode_operations,
9207 	.file_inode_ops	= &nfs4_file_inode_operations,
9208 	.file_ops	= &nfs4_file_operations,
9209 	.getroot	= nfs4_proc_get_root,
9210 	.submount	= nfs4_submount,
9211 	.try_mount	= nfs4_try_mount,
9212 	.getattr	= nfs4_proc_getattr,
9213 	.setattr	= nfs4_proc_setattr,
9214 	.lookup		= nfs4_proc_lookup,
9215 	.access		= nfs4_proc_access,
9216 	.readlink	= nfs4_proc_readlink,
9217 	.create		= nfs4_proc_create,
9218 	.remove		= nfs4_proc_remove,
9219 	.unlink_setup	= nfs4_proc_unlink_setup,
9220 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9221 	.unlink_done	= nfs4_proc_unlink_done,
9222 	.rename_setup	= nfs4_proc_rename_setup,
9223 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9224 	.rename_done	= nfs4_proc_rename_done,
9225 	.link		= nfs4_proc_link,
9226 	.symlink	= nfs4_proc_symlink,
9227 	.mkdir		= nfs4_proc_mkdir,
9228 	.rmdir		= nfs4_proc_remove,
9229 	.readdir	= nfs4_proc_readdir,
9230 	.mknod		= nfs4_proc_mknod,
9231 	.statfs		= nfs4_proc_statfs,
9232 	.fsinfo		= nfs4_proc_fsinfo,
9233 	.pathconf	= nfs4_proc_pathconf,
9234 	.set_capabilities = nfs4_server_capabilities,
9235 	.decode_dirent	= nfs4_decode_dirent,
9236 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9237 	.read_setup	= nfs4_proc_read_setup,
9238 	.read_done	= nfs4_read_done,
9239 	.write_setup	= nfs4_proc_write_setup,
9240 	.write_done	= nfs4_write_done,
9241 	.commit_setup	= nfs4_proc_commit_setup,
9242 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9243 	.commit_done	= nfs4_commit_done,
9244 	.lock		= nfs4_proc_lock,
9245 	.clear_acl_cache = nfs4_zap_acl_attr,
9246 	.close_context  = nfs4_close_context,
9247 	.open_context	= nfs4_atomic_open,
9248 	.have_delegation = nfs4_have_delegation,
9249 	.return_delegation = nfs4_inode_return_delegation,
9250 	.alloc_client	= nfs4_alloc_client,
9251 	.init_client	= nfs4_init_client,
9252 	.free_client	= nfs4_free_client,
9253 	.create_server	= nfs4_create_server,
9254 	.clone_server	= nfs_clone_server,
9255 };
9256 
9257 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9258 	.name	= XATTR_NAME_NFSV4_ACL,
9259 	.list	= nfs4_xattr_list_nfs4_acl,
9260 	.get	= nfs4_xattr_get_nfs4_acl,
9261 	.set	= nfs4_xattr_set_nfs4_acl,
9262 };
9263 
9264 const struct xattr_handler *nfs4_xattr_handlers[] = {
9265 	&nfs4_xattr_nfs4_acl_handler,
9266 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9267 	&nfs4_xattr_nfs4_label_handler,
9268 #endif
9269 	NULL
9270 };
9271 
9272 /*
9273  * Local variables:
9274  *  c-basic-offset: 8
9275  * End:
9276  */
9277