xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 1491eaf9)
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 	args->sa_slot = slot;
620 	res->sr_slot = slot;
621 
622 out_start:
623 	rpc_call_start(task);
624 	return 0;
625 
626 out_sleep:
627 	if (args->sa_privileged)
628 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
629 				NULL, RPC_PRIORITY_PRIVILEGED);
630 	else
631 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
632 	spin_unlock(&tbl->slot_tbl_lock);
633 	return -EAGAIN;
634 }
635 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
636 
637 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
638 {
639 	struct nfs4_slot *slot = res->sr_slot;
640 	struct nfs4_slot_table *tbl;
641 
642 	tbl = slot->table;
643 	spin_lock(&tbl->slot_tbl_lock);
644 	if (!nfs41_wake_and_assign_slot(tbl, slot))
645 		nfs4_free_slot(tbl, slot);
646 	spin_unlock(&tbl->slot_tbl_lock);
647 
648 	res->sr_slot = NULL;
649 }
650 
651 static int nfs40_sequence_done(struct rpc_task *task,
652 			       struct nfs4_sequence_res *res)
653 {
654 	if (res->sr_slot != NULL)
655 		nfs40_sequence_free_slot(res);
656 	return 1;
657 }
658 
659 #if defined(CONFIG_NFS_V4_1)
660 
661 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
662 {
663 	struct nfs4_session *session;
664 	struct nfs4_slot_table *tbl;
665 	struct nfs4_slot *slot = res->sr_slot;
666 	bool send_new_highest_used_slotid = false;
667 
668 	tbl = slot->table;
669 	session = tbl->session;
670 
671 	/* Bump the slot sequence number */
672 	if (slot->seq_done)
673 		slot->seq_nr++;
674 	slot->seq_done = 0;
675 
676 	spin_lock(&tbl->slot_tbl_lock);
677 	/* Be nice to the server: try to ensure that the last transmitted
678 	 * value for highest_user_slotid <= target_highest_slotid
679 	 */
680 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
681 		send_new_highest_used_slotid = true;
682 
683 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
684 		send_new_highest_used_slotid = false;
685 		goto out_unlock;
686 	}
687 	nfs4_free_slot(tbl, slot);
688 
689 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
690 		send_new_highest_used_slotid = false;
691 out_unlock:
692 	spin_unlock(&tbl->slot_tbl_lock);
693 	res->sr_slot = NULL;
694 	if (send_new_highest_used_slotid)
695 		nfs41_notify_server(session->clp);
696 	if (waitqueue_active(&tbl->slot_waitq))
697 		wake_up_all(&tbl->slot_waitq);
698 }
699 
700 static int nfs41_sequence_process(struct rpc_task *task,
701 		struct nfs4_sequence_res *res)
702 {
703 	struct nfs4_session *session;
704 	struct nfs4_slot *slot = res->sr_slot;
705 	struct nfs_client *clp;
706 	bool interrupted = false;
707 	int ret = 1;
708 
709 	if (slot == NULL)
710 		goto out_noaction;
711 	/* don't increment the sequence number if the task wasn't sent */
712 	if (!RPC_WAS_SENT(task))
713 		goto out;
714 
715 	session = slot->table->session;
716 
717 	if (slot->interrupted) {
718 		slot->interrupted = 0;
719 		interrupted = true;
720 	}
721 
722 	trace_nfs4_sequence_done(session, res);
723 	/* Check the SEQUENCE operation status */
724 	switch (res->sr_status) {
725 	case 0:
726 		/* Update the slot's sequence and clientid lease timer */
727 		slot->seq_done = 1;
728 		clp = session->clp;
729 		do_renew_lease(clp, res->sr_timestamp);
730 		/* Check sequence flags */
731 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
732 		nfs41_update_target_slotid(slot->table, slot, res);
733 		break;
734 	case 1:
735 		/*
736 		 * sr_status remains 1 if an RPC level error occurred.
737 		 * The server may or may not have processed the sequence
738 		 * operation..
739 		 * Mark the slot as having hosted an interrupted RPC call.
740 		 */
741 		slot->interrupted = 1;
742 		goto out;
743 	case -NFS4ERR_DELAY:
744 		/* The server detected a resend of the RPC call and
745 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
746 		 * of RFC5661.
747 		 */
748 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
749 			__func__,
750 			slot->slot_nr,
751 			slot->seq_nr);
752 		goto out_retry;
753 	case -NFS4ERR_BADSLOT:
754 		/*
755 		 * The slot id we used was probably retired. Try again
756 		 * using a different slot id.
757 		 */
758 		goto retry_nowait;
759 	case -NFS4ERR_SEQ_MISORDERED:
760 		/*
761 		 * Was the last operation on this sequence interrupted?
762 		 * If so, retry after bumping the sequence number.
763 		 */
764 		if (interrupted) {
765 			++slot->seq_nr;
766 			goto retry_nowait;
767 		}
768 		/*
769 		 * Could this slot have been previously retired?
770 		 * If so, then the server may be expecting seq_nr = 1!
771 		 */
772 		if (slot->seq_nr != 1) {
773 			slot->seq_nr = 1;
774 			goto retry_nowait;
775 		}
776 		break;
777 	case -NFS4ERR_SEQ_FALSE_RETRY:
778 		++slot->seq_nr;
779 		goto retry_nowait;
780 	default:
781 		/* Just update the slot sequence no. */
782 		slot->seq_done = 1;
783 	}
784 out:
785 	/* The session may be reset by one of the error handlers. */
786 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
787 out_noaction:
788 	return ret;
789 retry_nowait:
790 	if (rpc_restart_call_prepare(task)) {
791 		nfs41_sequence_free_slot(res);
792 		task->tk_status = 0;
793 		ret = 0;
794 	}
795 	goto out;
796 out_retry:
797 	if (!rpc_restart_call(task))
798 		goto out;
799 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
800 	return 0;
801 }
802 
803 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
804 {
805 	if (!nfs41_sequence_process(task, res))
806 		return 0;
807 	if (res->sr_slot != NULL)
808 		nfs41_sequence_free_slot(res);
809 	return 1;
810 
811 }
812 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
813 
814 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
815 {
816 	if (res->sr_slot == NULL)
817 		return 1;
818 	if (res->sr_slot->table->session != NULL)
819 		return nfs41_sequence_process(task, res);
820 	return nfs40_sequence_done(task, res);
821 }
822 
823 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
824 {
825 	if (res->sr_slot != NULL) {
826 		if (res->sr_slot->table->session != NULL)
827 			nfs41_sequence_free_slot(res);
828 		else
829 			nfs40_sequence_free_slot(res);
830 	}
831 }
832 
833 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
834 {
835 	if (res->sr_slot == NULL)
836 		return 1;
837 	if (!res->sr_slot->table->session)
838 		return nfs40_sequence_done(task, res);
839 	return nfs41_sequence_done(task, res);
840 }
841 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
842 
843 int nfs41_setup_sequence(struct nfs4_session *session,
844 				struct nfs4_sequence_args *args,
845 				struct nfs4_sequence_res *res,
846 				struct rpc_task *task)
847 {
848 	struct nfs4_slot *slot;
849 	struct nfs4_slot_table *tbl;
850 
851 	dprintk("--> %s\n", __func__);
852 	/* slot already allocated? */
853 	if (res->sr_slot != NULL)
854 		goto out_success;
855 
856 	tbl = &session->fc_slot_table;
857 
858 	task->tk_timeout = 0;
859 
860 	spin_lock(&tbl->slot_tbl_lock);
861 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
862 	    !args->sa_privileged) {
863 		/* The state manager will wait until the slot table is empty */
864 		dprintk("%s session is draining\n", __func__);
865 		goto out_sleep;
866 	}
867 
868 	slot = nfs4_alloc_slot(tbl);
869 	if (IS_ERR(slot)) {
870 		/* If out of memory, try again in 1/4 second */
871 		if (slot == ERR_PTR(-ENOMEM))
872 			task->tk_timeout = HZ >> 2;
873 		dprintk("<-- %s: no free slots\n", __func__);
874 		goto out_sleep;
875 	}
876 	spin_unlock(&tbl->slot_tbl_lock);
877 
878 	args->sa_slot = slot;
879 
880 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
881 			slot->slot_nr, slot->seq_nr);
882 
883 	res->sr_slot = slot;
884 	res->sr_timestamp = jiffies;
885 	res->sr_status_flags = 0;
886 	/*
887 	 * sr_status is only set in decode_sequence, and so will remain
888 	 * set to 1 if an rpc level failure occurs.
889 	 */
890 	res->sr_status = 1;
891 	trace_nfs4_setup_sequence(session, args);
892 out_success:
893 	rpc_call_start(task);
894 	return 0;
895 out_sleep:
896 	/* Privileged tasks are queued with top priority */
897 	if (args->sa_privileged)
898 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
899 				NULL, RPC_PRIORITY_PRIVILEGED);
900 	else
901 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
902 	spin_unlock(&tbl->slot_tbl_lock);
903 	return -EAGAIN;
904 }
905 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
906 
907 static int nfs4_setup_sequence(const struct nfs_server *server,
908 			       struct nfs4_sequence_args *args,
909 			       struct nfs4_sequence_res *res,
910 			       struct rpc_task *task)
911 {
912 	struct nfs4_session *session = nfs4_get_session(server);
913 	int ret = 0;
914 
915 	if (!session)
916 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
917 					    args, res, task);
918 
919 	dprintk("--> %s clp %p session %p sr_slot %u\n",
920 		__func__, session->clp, session, res->sr_slot ?
921 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
922 
923 	ret = nfs41_setup_sequence(session, args, res, task);
924 
925 	dprintk("<-- %s status=%d\n", __func__, ret);
926 	return ret;
927 }
928 
929 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
930 {
931 	struct nfs4_call_sync_data *data = calldata;
932 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
933 
934 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
935 
936 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
937 }
938 
939 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
940 {
941 	struct nfs4_call_sync_data *data = calldata;
942 
943 	nfs41_sequence_done(task, data->seq_res);
944 }
945 
946 static const struct rpc_call_ops nfs41_call_sync_ops = {
947 	.rpc_call_prepare = nfs41_call_sync_prepare,
948 	.rpc_call_done = nfs41_call_sync_done,
949 };
950 
951 #else	/* !CONFIG_NFS_V4_1 */
952 
953 static int nfs4_setup_sequence(const struct nfs_server *server,
954 			       struct nfs4_sequence_args *args,
955 			       struct nfs4_sequence_res *res,
956 			       struct rpc_task *task)
957 {
958 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
959 				    args, res, task);
960 }
961 
962 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
963 {
964 	return nfs40_sequence_done(task, res);
965 }
966 
967 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
968 {
969 	if (res->sr_slot != NULL)
970 		nfs40_sequence_free_slot(res);
971 }
972 
973 int nfs4_sequence_done(struct rpc_task *task,
974 		       struct nfs4_sequence_res *res)
975 {
976 	return nfs40_sequence_done(task, res);
977 }
978 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
979 
980 #endif	/* !CONFIG_NFS_V4_1 */
981 
982 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
983 {
984 	struct nfs4_call_sync_data *data = calldata;
985 	nfs4_setup_sequence(data->seq_server,
986 				data->seq_args, data->seq_res, task);
987 }
988 
989 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
990 {
991 	struct nfs4_call_sync_data *data = calldata;
992 	nfs4_sequence_done(task, data->seq_res);
993 }
994 
995 static const struct rpc_call_ops nfs40_call_sync_ops = {
996 	.rpc_call_prepare = nfs40_call_sync_prepare,
997 	.rpc_call_done = nfs40_call_sync_done,
998 };
999 
1000 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1001 				   struct nfs_server *server,
1002 				   struct rpc_message *msg,
1003 				   struct nfs4_sequence_args *args,
1004 				   struct nfs4_sequence_res *res)
1005 {
1006 	int ret;
1007 	struct rpc_task *task;
1008 	struct nfs_client *clp = server->nfs_client;
1009 	struct nfs4_call_sync_data data = {
1010 		.seq_server = server,
1011 		.seq_args = args,
1012 		.seq_res = res,
1013 	};
1014 	struct rpc_task_setup task_setup = {
1015 		.rpc_client = clnt,
1016 		.rpc_message = msg,
1017 		.callback_ops = clp->cl_mvops->call_sync_ops,
1018 		.callback_data = &data
1019 	};
1020 
1021 	task = rpc_run_task(&task_setup);
1022 	if (IS_ERR(task))
1023 		ret = PTR_ERR(task);
1024 	else {
1025 		ret = task->tk_status;
1026 		rpc_put_task(task);
1027 	}
1028 	return ret;
1029 }
1030 
1031 int nfs4_call_sync(struct rpc_clnt *clnt,
1032 		   struct nfs_server *server,
1033 		   struct rpc_message *msg,
1034 		   struct nfs4_sequence_args *args,
1035 		   struct nfs4_sequence_res *res,
1036 		   int cache_reply)
1037 {
1038 	nfs4_init_sequence(args, res, cache_reply);
1039 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1040 }
1041 
1042 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1043 {
1044 	struct nfs_inode *nfsi = NFS_I(dir);
1045 
1046 	spin_lock(&dir->i_lock);
1047 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1048 	if (!cinfo->atomic || cinfo->before != dir->i_version)
1049 		nfs_force_lookup_revalidate(dir);
1050 	dir->i_version = cinfo->after;
1051 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1052 	nfs_fscache_invalidate(dir);
1053 	spin_unlock(&dir->i_lock);
1054 }
1055 
1056 struct nfs4_opendata {
1057 	struct kref kref;
1058 	struct nfs_openargs o_arg;
1059 	struct nfs_openres o_res;
1060 	struct nfs_open_confirmargs c_arg;
1061 	struct nfs_open_confirmres c_res;
1062 	struct nfs4_string owner_name;
1063 	struct nfs4_string group_name;
1064 	struct nfs4_label *a_label;
1065 	struct nfs_fattr f_attr;
1066 	struct nfs4_label *f_label;
1067 	struct dentry *dir;
1068 	struct dentry *dentry;
1069 	struct nfs4_state_owner *owner;
1070 	struct nfs4_state *state;
1071 	struct iattr attrs;
1072 	unsigned long timestamp;
1073 	unsigned int rpc_done : 1;
1074 	unsigned int file_created : 1;
1075 	unsigned int is_recover : 1;
1076 	int rpc_status;
1077 	int cancelled;
1078 };
1079 
1080 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1081 		int err, struct nfs4_exception *exception)
1082 {
1083 	if (err != -EINVAL)
1084 		return false;
1085 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1086 		return false;
1087 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1088 	exception->retry = 1;
1089 	return true;
1090 }
1091 
1092 static u32
1093 nfs4_map_atomic_open_share(struct nfs_server *server,
1094 		fmode_t fmode, int openflags)
1095 {
1096 	u32 res = 0;
1097 
1098 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1099 	case FMODE_READ:
1100 		res = NFS4_SHARE_ACCESS_READ;
1101 		break;
1102 	case FMODE_WRITE:
1103 		res = NFS4_SHARE_ACCESS_WRITE;
1104 		break;
1105 	case FMODE_READ|FMODE_WRITE:
1106 		res = NFS4_SHARE_ACCESS_BOTH;
1107 	}
1108 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1109 		goto out;
1110 	/* Want no delegation if we're using O_DIRECT */
1111 	if (openflags & O_DIRECT)
1112 		res |= NFS4_SHARE_WANT_NO_DELEG;
1113 out:
1114 	return res;
1115 }
1116 
1117 static enum open_claim_type4
1118 nfs4_map_atomic_open_claim(struct nfs_server *server,
1119 		enum open_claim_type4 claim)
1120 {
1121 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1122 		return claim;
1123 	switch (claim) {
1124 	default:
1125 		return claim;
1126 	case NFS4_OPEN_CLAIM_FH:
1127 		return NFS4_OPEN_CLAIM_NULL;
1128 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1129 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1130 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1131 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1132 	}
1133 }
1134 
1135 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1136 {
1137 	p->o_res.f_attr = &p->f_attr;
1138 	p->o_res.f_label = p->f_label;
1139 	p->o_res.seqid = p->o_arg.seqid;
1140 	p->c_res.seqid = p->c_arg.seqid;
1141 	p->o_res.server = p->o_arg.server;
1142 	p->o_res.access_request = p->o_arg.access;
1143 	nfs_fattr_init(&p->f_attr);
1144 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1145 }
1146 
1147 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1148 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1149 		const struct iattr *attrs,
1150 		struct nfs4_label *label,
1151 		enum open_claim_type4 claim,
1152 		gfp_t gfp_mask)
1153 {
1154 	struct dentry *parent = dget_parent(dentry);
1155 	struct inode *dir = d_inode(parent);
1156 	struct nfs_server *server = NFS_SERVER(dir);
1157 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1158 	struct nfs4_opendata *p;
1159 
1160 	p = kzalloc(sizeof(*p), gfp_mask);
1161 	if (p == NULL)
1162 		goto err;
1163 
1164 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1165 	if (IS_ERR(p->f_label))
1166 		goto err_free_p;
1167 
1168 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1169 	if (IS_ERR(p->a_label))
1170 		goto err_free_f;
1171 
1172 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1173 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1174 	if (IS_ERR(p->o_arg.seqid))
1175 		goto err_free_label;
1176 	nfs_sb_active(dentry->d_sb);
1177 	p->dentry = dget(dentry);
1178 	p->dir = parent;
1179 	p->owner = sp;
1180 	atomic_inc(&sp->so_count);
1181 	p->o_arg.open_flags = flags;
1182 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1183 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1184 			fmode, flags);
1185 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1186 	 * will return permission denied for all bits until close */
1187 	if (!(flags & O_EXCL)) {
1188 		/* ask server to check for all possible rights as results
1189 		 * are cached */
1190 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1191 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1192 	}
1193 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1194 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1195 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1196 	p->o_arg.name = &dentry->d_name;
1197 	p->o_arg.server = server;
1198 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1199 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1200 	p->o_arg.label = nfs4_label_copy(p->a_label, label);
1201 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1202 	switch (p->o_arg.claim) {
1203 	case NFS4_OPEN_CLAIM_NULL:
1204 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1205 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1206 		p->o_arg.fh = NFS_FH(dir);
1207 		break;
1208 	case NFS4_OPEN_CLAIM_PREVIOUS:
1209 	case NFS4_OPEN_CLAIM_FH:
1210 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1211 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1212 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1213 	}
1214 	if (attrs != NULL && attrs->ia_valid != 0) {
1215 		__u32 verf[2];
1216 
1217 		p->o_arg.u.attrs = &p->attrs;
1218 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1219 
1220 		verf[0] = jiffies;
1221 		verf[1] = current->pid;
1222 		memcpy(p->o_arg.u.verifier.data, verf,
1223 				sizeof(p->o_arg.u.verifier.data));
1224 	}
1225 	p->c_arg.fh = &p->o_res.fh;
1226 	p->c_arg.stateid = &p->o_res.stateid;
1227 	p->c_arg.seqid = p->o_arg.seqid;
1228 	nfs4_init_opendata_res(p);
1229 	kref_init(&p->kref);
1230 	return p;
1231 
1232 err_free_label:
1233 	nfs4_label_free(p->a_label);
1234 err_free_f:
1235 	nfs4_label_free(p->f_label);
1236 err_free_p:
1237 	kfree(p);
1238 err:
1239 	dput(parent);
1240 	return NULL;
1241 }
1242 
1243 static void nfs4_opendata_free(struct kref *kref)
1244 {
1245 	struct nfs4_opendata *p = container_of(kref,
1246 			struct nfs4_opendata, kref);
1247 	struct super_block *sb = p->dentry->d_sb;
1248 
1249 	nfs_free_seqid(p->o_arg.seqid);
1250 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1251 	if (p->state != NULL)
1252 		nfs4_put_open_state(p->state);
1253 	nfs4_put_state_owner(p->owner);
1254 
1255 	nfs4_label_free(p->a_label);
1256 	nfs4_label_free(p->f_label);
1257 
1258 	dput(p->dir);
1259 	dput(p->dentry);
1260 	nfs_sb_deactive(sb);
1261 	nfs_fattr_free_names(&p->f_attr);
1262 	kfree(p->f_attr.mdsthreshold);
1263 	kfree(p);
1264 }
1265 
1266 static void nfs4_opendata_put(struct nfs4_opendata *p)
1267 {
1268 	if (p != NULL)
1269 		kref_put(&p->kref, nfs4_opendata_free);
1270 }
1271 
1272 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1273 {
1274 	int ret;
1275 
1276 	ret = rpc_wait_for_completion_task(task);
1277 	return ret;
1278 }
1279 
1280 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1281 		fmode_t fmode)
1282 {
1283 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1284 	case FMODE_READ|FMODE_WRITE:
1285 		return state->n_rdwr != 0;
1286 	case FMODE_WRITE:
1287 		return state->n_wronly != 0;
1288 	case FMODE_READ:
1289 		return state->n_rdonly != 0;
1290 	}
1291 	WARN_ON_ONCE(1);
1292 	return false;
1293 }
1294 
1295 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1296 {
1297 	int ret = 0;
1298 
1299 	if (open_mode & (O_EXCL|O_TRUNC))
1300 		goto out;
1301 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1302 		case FMODE_READ:
1303 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1304 				&& state->n_rdonly != 0;
1305 			break;
1306 		case FMODE_WRITE:
1307 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1308 				&& state->n_wronly != 0;
1309 			break;
1310 		case FMODE_READ|FMODE_WRITE:
1311 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1312 				&& state->n_rdwr != 0;
1313 	}
1314 out:
1315 	return ret;
1316 }
1317 
1318 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1319 		enum open_claim_type4 claim)
1320 {
1321 	if (delegation == NULL)
1322 		return 0;
1323 	if ((delegation->type & fmode) != fmode)
1324 		return 0;
1325 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1326 		return 0;
1327 	switch (claim) {
1328 	case NFS4_OPEN_CLAIM_NULL:
1329 	case NFS4_OPEN_CLAIM_FH:
1330 		break;
1331 	case NFS4_OPEN_CLAIM_PREVIOUS:
1332 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1333 			break;
1334 	default:
1335 		return 0;
1336 	}
1337 	nfs_mark_delegation_referenced(delegation);
1338 	return 1;
1339 }
1340 
1341 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1342 {
1343 	switch (fmode) {
1344 		case FMODE_WRITE:
1345 			state->n_wronly++;
1346 			break;
1347 		case FMODE_READ:
1348 			state->n_rdonly++;
1349 			break;
1350 		case FMODE_READ|FMODE_WRITE:
1351 			state->n_rdwr++;
1352 	}
1353 	nfs4_state_set_mode_locked(state, state->state | fmode);
1354 }
1355 
1356 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1357 {
1358 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1359 	bool need_recover = false;
1360 
1361 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1362 		need_recover = true;
1363 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1364 		need_recover = true;
1365 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1366 		need_recover = true;
1367 	if (need_recover)
1368 		nfs4_state_mark_reclaim_nograce(clp, state);
1369 }
1370 
1371 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1372 		nfs4_stateid *stateid)
1373 {
1374 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1375 		return true;
1376 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1377 		nfs_test_and_clear_all_open_stateid(state);
1378 		return true;
1379 	}
1380 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1381 		return true;
1382 	return false;
1383 }
1384 
1385 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1386 {
1387 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1388 		return;
1389 	if (state->n_wronly)
1390 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1391 	if (state->n_rdonly)
1392 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1393 	if (state->n_rdwr)
1394 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1395 	set_bit(NFS_OPEN_STATE, &state->flags);
1396 }
1397 
1398 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1399 		nfs4_stateid *arg_stateid,
1400 		nfs4_stateid *stateid, fmode_t fmode)
1401 {
1402 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1403 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1404 	case FMODE_WRITE:
1405 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1406 		break;
1407 	case FMODE_READ:
1408 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1409 		break;
1410 	case 0:
1411 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1412 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1413 		clear_bit(NFS_OPEN_STATE, &state->flags);
1414 	}
1415 	if (stateid == NULL)
1416 		return;
1417 	/* Handle races with OPEN */
1418 	if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1419 	    (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1420 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1421 		nfs_resync_open_stateid_locked(state);
1422 		return;
1423 	}
1424 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1425 		nfs4_stateid_copy(&state->stateid, stateid);
1426 	nfs4_stateid_copy(&state->open_stateid, stateid);
1427 }
1428 
1429 static void nfs_clear_open_stateid(struct nfs4_state *state,
1430 	nfs4_stateid *arg_stateid,
1431 	nfs4_stateid *stateid, fmode_t fmode)
1432 {
1433 	write_seqlock(&state->seqlock);
1434 	nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1435 	write_sequnlock(&state->seqlock);
1436 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1437 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1438 }
1439 
1440 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1441 {
1442 	switch (fmode) {
1443 		case FMODE_READ:
1444 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1445 			break;
1446 		case FMODE_WRITE:
1447 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1448 			break;
1449 		case FMODE_READ|FMODE_WRITE:
1450 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1451 	}
1452 	if (!nfs_need_update_open_stateid(state, stateid))
1453 		return;
1454 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1455 		nfs4_stateid_copy(&state->stateid, stateid);
1456 	nfs4_stateid_copy(&state->open_stateid, stateid);
1457 }
1458 
1459 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1460 {
1461 	/*
1462 	 * Protect the call to nfs4_state_set_mode_locked and
1463 	 * serialise the stateid update
1464 	 */
1465 	spin_lock(&state->owner->so_lock);
1466 	write_seqlock(&state->seqlock);
1467 	if (deleg_stateid != NULL) {
1468 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1469 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1470 	}
1471 	if (open_stateid != NULL)
1472 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1473 	write_sequnlock(&state->seqlock);
1474 	update_open_stateflags(state, fmode);
1475 	spin_unlock(&state->owner->so_lock);
1476 }
1477 
1478 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1479 {
1480 	struct nfs_inode *nfsi = NFS_I(state->inode);
1481 	struct nfs_delegation *deleg_cur;
1482 	int ret = 0;
1483 
1484 	fmode &= (FMODE_READ|FMODE_WRITE);
1485 
1486 	rcu_read_lock();
1487 	deleg_cur = rcu_dereference(nfsi->delegation);
1488 	if (deleg_cur == NULL)
1489 		goto no_delegation;
1490 
1491 	spin_lock(&deleg_cur->lock);
1492 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1493 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1494 	    (deleg_cur->type & fmode) != fmode)
1495 		goto no_delegation_unlock;
1496 
1497 	if (delegation == NULL)
1498 		delegation = &deleg_cur->stateid;
1499 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1500 		goto no_delegation_unlock;
1501 
1502 	nfs_mark_delegation_referenced(deleg_cur);
1503 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1504 	ret = 1;
1505 no_delegation_unlock:
1506 	spin_unlock(&deleg_cur->lock);
1507 no_delegation:
1508 	rcu_read_unlock();
1509 
1510 	if (!ret && open_stateid != NULL) {
1511 		__update_open_stateid(state, open_stateid, NULL, fmode);
1512 		ret = 1;
1513 	}
1514 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1515 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1516 
1517 	return ret;
1518 }
1519 
1520 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1521 		const nfs4_stateid *stateid)
1522 {
1523 	struct nfs4_state *state = lsp->ls_state;
1524 	bool ret = false;
1525 
1526 	spin_lock(&state->state_lock);
1527 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1528 		goto out_noupdate;
1529 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1530 		goto out_noupdate;
1531 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1532 	ret = true;
1533 out_noupdate:
1534 	spin_unlock(&state->state_lock);
1535 	return ret;
1536 }
1537 
1538 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1539 {
1540 	struct nfs_delegation *delegation;
1541 
1542 	rcu_read_lock();
1543 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1544 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1545 		rcu_read_unlock();
1546 		return;
1547 	}
1548 	rcu_read_unlock();
1549 	nfs4_inode_return_delegation(inode);
1550 }
1551 
1552 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1553 {
1554 	struct nfs4_state *state = opendata->state;
1555 	struct nfs_inode *nfsi = NFS_I(state->inode);
1556 	struct nfs_delegation *delegation;
1557 	int open_mode = opendata->o_arg.open_flags;
1558 	fmode_t fmode = opendata->o_arg.fmode;
1559 	enum open_claim_type4 claim = opendata->o_arg.claim;
1560 	nfs4_stateid stateid;
1561 	int ret = -EAGAIN;
1562 
1563 	for (;;) {
1564 		spin_lock(&state->owner->so_lock);
1565 		if (can_open_cached(state, fmode, open_mode)) {
1566 			update_open_stateflags(state, fmode);
1567 			spin_unlock(&state->owner->so_lock);
1568 			goto out_return_state;
1569 		}
1570 		spin_unlock(&state->owner->so_lock);
1571 		rcu_read_lock();
1572 		delegation = rcu_dereference(nfsi->delegation);
1573 		if (!can_open_delegated(delegation, fmode, claim)) {
1574 			rcu_read_unlock();
1575 			break;
1576 		}
1577 		/* Save the delegation */
1578 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1579 		rcu_read_unlock();
1580 		nfs_release_seqid(opendata->o_arg.seqid);
1581 		if (!opendata->is_recover) {
1582 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1583 			if (ret != 0)
1584 				goto out;
1585 		}
1586 		ret = -EAGAIN;
1587 
1588 		/* Try to update the stateid using the delegation */
1589 		if (update_open_stateid(state, NULL, &stateid, fmode))
1590 			goto out_return_state;
1591 	}
1592 out:
1593 	return ERR_PTR(ret);
1594 out_return_state:
1595 	atomic_inc(&state->count);
1596 	return state;
1597 }
1598 
1599 static void
1600 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1601 {
1602 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1603 	struct nfs_delegation *delegation;
1604 	int delegation_flags = 0;
1605 
1606 	rcu_read_lock();
1607 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1608 	if (delegation)
1609 		delegation_flags = delegation->flags;
1610 	rcu_read_unlock();
1611 	switch (data->o_arg.claim) {
1612 	default:
1613 		break;
1614 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1615 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1616 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1617 				   "returning a delegation for "
1618 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1619 				   clp->cl_hostname);
1620 		return;
1621 	}
1622 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1623 		nfs_inode_set_delegation(state->inode,
1624 					 data->owner->so_cred,
1625 					 &data->o_res);
1626 	else
1627 		nfs_inode_reclaim_delegation(state->inode,
1628 					     data->owner->so_cred,
1629 					     &data->o_res);
1630 }
1631 
1632 /*
1633  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1634  * and update the nfs4_state.
1635  */
1636 static struct nfs4_state *
1637 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1638 {
1639 	struct inode *inode = data->state->inode;
1640 	struct nfs4_state *state = data->state;
1641 	int ret;
1642 
1643 	if (!data->rpc_done) {
1644 		if (data->rpc_status) {
1645 			ret = data->rpc_status;
1646 			goto err;
1647 		}
1648 		/* cached opens have already been processed */
1649 		goto update;
1650 	}
1651 
1652 	ret = nfs_refresh_inode(inode, &data->f_attr);
1653 	if (ret)
1654 		goto err;
1655 
1656 	if (data->o_res.delegation_type != 0)
1657 		nfs4_opendata_check_deleg(data, state);
1658 update:
1659 	update_open_stateid(state, &data->o_res.stateid, NULL,
1660 			    data->o_arg.fmode);
1661 	atomic_inc(&state->count);
1662 
1663 	return state;
1664 err:
1665 	return ERR_PTR(ret);
1666 
1667 }
1668 
1669 static struct nfs4_state *
1670 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1671 {
1672 	struct inode *inode;
1673 	struct nfs4_state *state = NULL;
1674 	int ret;
1675 
1676 	if (!data->rpc_done) {
1677 		state = nfs4_try_open_cached(data);
1678 		trace_nfs4_cached_open(data->state);
1679 		goto out;
1680 	}
1681 
1682 	ret = -EAGAIN;
1683 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1684 		goto err;
1685 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1686 	ret = PTR_ERR(inode);
1687 	if (IS_ERR(inode))
1688 		goto err;
1689 	ret = -ENOMEM;
1690 	state = nfs4_get_open_state(inode, data->owner);
1691 	if (state == NULL)
1692 		goto err_put_inode;
1693 	if (data->o_res.delegation_type != 0)
1694 		nfs4_opendata_check_deleg(data, state);
1695 	update_open_stateid(state, &data->o_res.stateid, NULL,
1696 			data->o_arg.fmode);
1697 	iput(inode);
1698 out:
1699 	nfs_release_seqid(data->o_arg.seqid);
1700 	return state;
1701 err_put_inode:
1702 	iput(inode);
1703 err:
1704 	return ERR_PTR(ret);
1705 }
1706 
1707 static struct nfs4_state *
1708 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1709 {
1710 	struct nfs4_state *ret;
1711 
1712 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1713 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1714 	else
1715 		ret = _nfs4_opendata_to_nfs4_state(data);
1716 	nfs4_sequence_free_slot(&data->o_res.seq_res);
1717 	return ret;
1718 }
1719 
1720 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1721 {
1722 	struct nfs_inode *nfsi = NFS_I(state->inode);
1723 	struct nfs_open_context *ctx;
1724 
1725 	spin_lock(&state->inode->i_lock);
1726 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1727 		if (ctx->state != state)
1728 			continue;
1729 		get_nfs_open_context(ctx);
1730 		spin_unlock(&state->inode->i_lock);
1731 		return ctx;
1732 	}
1733 	spin_unlock(&state->inode->i_lock);
1734 	return ERR_PTR(-ENOENT);
1735 }
1736 
1737 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1738 		struct nfs4_state *state, enum open_claim_type4 claim)
1739 {
1740 	struct nfs4_opendata *opendata;
1741 
1742 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1743 			NULL, NULL, claim, GFP_NOFS);
1744 	if (opendata == NULL)
1745 		return ERR_PTR(-ENOMEM);
1746 	opendata->state = state;
1747 	atomic_inc(&state->count);
1748 	return opendata;
1749 }
1750 
1751 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1752 		fmode_t fmode)
1753 {
1754 	struct nfs4_state *newstate;
1755 	int ret;
1756 
1757 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1758 		return 0;
1759 	opendata->o_arg.open_flags = 0;
1760 	opendata->o_arg.fmode = fmode;
1761 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1762 			NFS_SB(opendata->dentry->d_sb),
1763 			fmode, 0);
1764 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1765 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1766 	nfs4_init_opendata_res(opendata);
1767 	ret = _nfs4_recover_proc_open(opendata);
1768 	if (ret != 0)
1769 		return ret;
1770 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1771 	if (IS_ERR(newstate))
1772 		return PTR_ERR(newstate);
1773 	if (newstate != opendata->state)
1774 		ret = -ESTALE;
1775 	nfs4_close_state(newstate, fmode);
1776 	return ret;
1777 }
1778 
1779 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1780 {
1781 	int ret;
1782 
1783 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1784 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1785 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1786 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1787 	/* memory barrier prior to reading state->n_* */
1788 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1789 	clear_bit(NFS_OPEN_STATE, &state->flags);
1790 	smp_rmb();
1791 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1792 	if (ret != 0)
1793 		return ret;
1794 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1795 	if (ret != 0)
1796 		return ret;
1797 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1798 	if (ret != 0)
1799 		return ret;
1800 	/*
1801 	 * We may have performed cached opens for all three recoveries.
1802 	 * Check if we need to update the current stateid.
1803 	 */
1804 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1805 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1806 		write_seqlock(&state->seqlock);
1807 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1808 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1809 		write_sequnlock(&state->seqlock);
1810 	}
1811 	return 0;
1812 }
1813 
1814 /*
1815  * OPEN_RECLAIM:
1816  * 	reclaim state on the server after a reboot.
1817  */
1818 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1819 {
1820 	struct nfs_delegation *delegation;
1821 	struct nfs4_opendata *opendata;
1822 	fmode_t delegation_type = 0;
1823 	int status;
1824 
1825 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1826 			NFS4_OPEN_CLAIM_PREVIOUS);
1827 	if (IS_ERR(opendata))
1828 		return PTR_ERR(opendata);
1829 	rcu_read_lock();
1830 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1831 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1832 		delegation_type = delegation->type;
1833 	rcu_read_unlock();
1834 	opendata->o_arg.u.delegation_type = delegation_type;
1835 	status = nfs4_open_recover(opendata, state);
1836 	nfs4_opendata_put(opendata);
1837 	return status;
1838 }
1839 
1840 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1841 {
1842 	struct nfs_server *server = NFS_SERVER(state->inode);
1843 	struct nfs4_exception exception = { };
1844 	int err;
1845 	do {
1846 		err = _nfs4_do_open_reclaim(ctx, state);
1847 		trace_nfs4_open_reclaim(ctx, 0, err);
1848 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1849 			continue;
1850 		if (err != -NFS4ERR_DELAY)
1851 			break;
1852 		nfs4_handle_exception(server, err, &exception);
1853 	} while (exception.retry);
1854 	return err;
1855 }
1856 
1857 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1858 {
1859 	struct nfs_open_context *ctx;
1860 	int ret;
1861 
1862 	ctx = nfs4_state_find_open_context(state);
1863 	if (IS_ERR(ctx))
1864 		return -EAGAIN;
1865 	ret = nfs4_do_open_reclaim(ctx, state);
1866 	put_nfs_open_context(ctx);
1867 	return ret;
1868 }
1869 
1870 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1871 {
1872 	switch (err) {
1873 		default:
1874 			printk(KERN_ERR "NFS: %s: unhandled error "
1875 					"%d.\n", __func__, err);
1876 		case 0:
1877 		case -ENOENT:
1878 		case -EAGAIN:
1879 		case -ESTALE:
1880 			break;
1881 		case -NFS4ERR_BADSESSION:
1882 		case -NFS4ERR_BADSLOT:
1883 		case -NFS4ERR_BAD_HIGH_SLOT:
1884 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1885 		case -NFS4ERR_DEADSESSION:
1886 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1887 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1888 			return -EAGAIN;
1889 		case -NFS4ERR_STALE_CLIENTID:
1890 		case -NFS4ERR_STALE_STATEID:
1891 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1892 		case -NFS4ERR_EXPIRED:
1893 			/* Don't recall a delegation if it was lost */
1894 			nfs4_schedule_lease_recovery(server->nfs_client);
1895 			return -EAGAIN;
1896 		case -NFS4ERR_MOVED:
1897 			nfs4_schedule_migration_recovery(server);
1898 			return -EAGAIN;
1899 		case -NFS4ERR_LEASE_MOVED:
1900 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1901 			return -EAGAIN;
1902 		case -NFS4ERR_DELEG_REVOKED:
1903 		case -NFS4ERR_ADMIN_REVOKED:
1904 		case -NFS4ERR_BAD_STATEID:
1905 		case -NFS4ERR_OPENMODE:
1906 			nfs_inode_find_state_and_recover(state->inode,
1907 					stateid);
1908 			nfs4_schedule_stateid_recovery(server, state);
1909 			return -EAGAIN;
1910 		case -NFS4ERR_DELAY:
1911 		case -NFS4ERR_GRACE:
1912 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1913 			ssleep(1);
1914 			return -EAGAIN;
1915 		case -ENOMEM:
1916 		case -NFS4ERR_DENIED:
1917 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1918 			return 0;
1919 	}
1920 	return err;
1921 }
1922 
1923 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1924 		struct nfs4_state *state, const nfs4_stateid *stateid,
1925 		fmode_t type)
1926 {
1927 	struct nfs_server *server = NFS_SERVER(state->inode);
1928 	struct nfs4_opendata *opendata;
1929 	int err = 0;
1930 
1931 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1932 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1933 	if (IS_ERR(opendata))
1934 		return PTR_ERR(opendata);
1935 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1936 	write_seqlock(&state->seqlock);
1937 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1938 	write_sequnlock(&state->seqlock);
1939 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1940 	switch (type & (FMODE_READ|FMODE_WRITE)) {
1941 	case FMODE_READ|FMODE_WRITE:
1942 	case FMODE_WRITE:
1943 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1944 		if (err)
1945 			break;
1946 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1947 		if (err)
1948 			break;
1949 	case FMODE_READ:
1950 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
1951 	}
1952 	nfs4_opendata_put(opendata);
1953 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1954 }
1955 
1956 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1957 {
1958 	struct nfs4_opendata *data = calldata;
1959 
1960 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1961 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1962 }
1963 
1964 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1965 {
1966 	struct nfs4_opendata *data = calldata;
1967 
1968 	nfs40_sequence_done(task, &data->c_res.seq_res);
1969 
1970 	data->rpc_status = task->tk_status;
1971 	if (data->rpc_status == 0) {
1972 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1973 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1974 		renew_lease(data->o_res.server, data->timestamp);
1975 		data->rpc_done = 1;
1976 	}
1977 }
1978 
1979 static void nfs4_open_confirm_release(void *calldata)
1980 {
1981 	struct nfs4_opendata *data = calldata;
1982 	struct nfs4_state *state = NULL;
1983 
1984 	/* If this request hasn't been cancelled, do nothing */
1985 	if (data->cancelled == 0)
1986 		goto out_free;
1987 	/* In case of error, no cleanup! */
1988 	if (!data->rpc_done)
1989 		goto out_free;
1990 	state = nfs4_opendata_to_nfs4_state(data);
1991 	if (!IS_ERR(state))
1992 		nfs4_close_state(state, data->o_arg.fmode);
1993 out_free:
1994 	nfs4_opendata_put(data);
1995 }
1996 
1997 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1998 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1999 	.rpc_call_done = nfs4_open_confirm_done,
2000 	.rpc_release = nfs4_open_confirm_release,
2001 };
2002 
2003 /*
2004  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2005  */
2006 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2007 {
2008 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2009 	struct rpc_task *task;
2010 	struct  rpc_message msg = {
2011 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2012 		.rpc_argp = &data->c_arg,
2013 		.rpc_resp = &data->c_res,
2014 		.rpc_cred = data->owner->so_cred,
2015 	};
2016 	struct rpc_task_setup task_setup_data = {
2017 		.rpc_client = server->client,
2018 		.rpc_message = &msg,
2019 		.callback_ops = &nfs4_open_confirm_ops,
2020 		.callback_data = data,
2021 		.workqueue = nfsiod_workqueue,
2022 		.flags = RPC_TASK_ASYNC,
2023 	};
2024 	int status;
2025 
2026 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2027 	kref_get(&data->kref);
2028 	data->rpc_done = 0;
2029 	data->rpc_status = 0;
2030 	data->timestamp = jiffies;
2031 	if (data->is_recover)
2032 		nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2033 	task = rpc_run_task(&task_setup_data);
2034 	if (IS_ERR(task))
2035 		return PTR_ERR(task);
2036 	status = nfs4_wait_for_completion_rpc_task(task);
2037 	if (status != 0) {
2038 		data->cancelled = 1;
2039 		smp_wmb();
2040 	} else
2041 		status = data->rpc_status;
2042 	rpc_put_task(task);
2043 	return status;
2044 }
2045 
2046 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2047 {
2048 	struct nfs4_opendata *data = calldata;
2049 	struct nfs4_state_owner *sp = data->owner;
2050 	struct nfs_client *clp = sp->so_server->nfs_client;
2051 	enum open_claim_type4 claim = data->o_arg.claim;
2052 
2053 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2054 		goto out_wait;
2055 	/*
2056 	 * Check if we still need to send an OPEN call, or if we can use
2057 	 * a delegation instead.
2058 	 */
2059 	if (data->state != NULL) {
2060 		struct nfs_delegation *delegation;
2061 
2062 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2063 			goto out_no_action;
2064 		rcu_read_lock();
2065 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2066 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2067 			goto unlock_no_action;
2068 		rcu_read_unlock();
2069 	}
2070 	/* Update client id. */
2071 	data->o_arg.clientid = clp->cl_clientid;
2072 	switch (claim) {
2073 	default:
2074 		break;
2075 	case NFS4_OPEN_CLAIM_PREVIOUS:
2076 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2077 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2078 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2079 	case NFS4_OPEN_CLAIM_FH:
2080 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2081 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2082 	}
2083 	data->timestamp = jiffies;
2084 	if (nfs4_setup_sequence(data->o_arg.server,
2085 				&data->o_arg.seq_args,
2086 				&data->o_res.seq_res,
2087 				task) != 0)
2088 		nfs_release_seqid(data->o_arg.seqid);
2089 
2090 	/* Set the create mode (note dependency on the session type) */
2091 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2092 	if (data->o_arg.open_flags & O_EXCL) {
2093 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2094 		if (nfs4_has_persistent_session(clp))
2095 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2096 		else if (clp->cl_mvops->minor_version > 0)
2097 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2098 	}
2099 	return;
2100 unlock_no_action:
2101 	trace_nfs4_cached_open(data->state);
2102 	rcu_read_unlock();
2103 out_no_action:
2104 	task->tk_action = NULL;
2105 out_wait:
2106 	nfs4_sequence_done(task, &data->o_res.seq_res);
2107 }
2108 
2109 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2110 {
2111 	struct nfs4_opendata *data = calldata;
2112 
2113 	data->rpc_status = task->tk_status;
2114 
2115 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2116 		return;
2117 
2118 	if (task->tk_status == 0) {
2119 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2120 			switch (data->o_res.f_attr->mode & S_IFMT) {
2121 			case S_IFREG:
2122 				break;
2123 			case S_IFLNK:
2124 				data->rpc_status = -ELOOP;
2125 				break;
2126 			case S_IFDIR:
2127 				data->rpc_status = -EISDIR;
2128 				break;
2129 			default:
2130 				data->rpc_status = -ENOTDIR;
2131 			}
2132 		}
2133 		renew_lease(data->o_res.server, data->timestamp);
2134 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2135 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2136 	}
2137 	data->rpc_done = 1;
2138 }
2139 
2140 static void nfs4_open_release(void *calldata)
2141 {
2142 	struct nfs4_opendata *data = calldata;
2143 	struct nfs4_state *state = NULL;
2144 
2145 	/* If this request hasn't been cancelled, do nothing */
2146 	if (data->cancelled == 0)
2147 		goto out_free;
2148 	/* In case of error, no cleanup! */
2149 	if (data->rpc_status != 0 || !data->rpc_done)
2150 		goto out_free;
2151 	/* In case we need an open_confirm, no cleanup! */
2152 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2153 		goto out_free;
2154 	state = nfs4_opendata_to_nfs4_state(data);
2155 	if (!IS_ERR(state))
2156 		nfs4_close_state(state, data->o_arg.fmode);
2157 out_free:
2158 	nfs4_opendata_put(data);
2159 }
2160 
2161 static const struct rpc_call_ops nfs4_open_ops = {
2162 	.rpc_call_prepare = nfs4_open_prepare,
2163 	.rpc_call_done = nfs4_open_done,
2164 	.rpc_release = nfs4_open_release,
2165 };
2166 
2167 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2168 {
2169 	struct inode *dir = d_inode(data->dir);
2170 	struct nfs_server *server = NFS_SERVER(dir);
2171 	struct nfs_openargs *o_arg = &data->o_arg;
2172 	struct nfs_openres *o_res = &data->o_res;
2173 	struct rpc_task *task;
2174 	struct rpc_message msg = {
2175 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2176 		.rpc_argp = o_arg,
2177 		.rpc_resp = o_res,
2178 		.rpc_cred = data->owner->so_cred,
2179 	};
2180 	struct rpc_task_setup task_setup_data = {
2181 		.rpc_client = server->client,
2182 		.rpc_message = &msg,
2183 		.callback_ops = &nfs4_open_ops,
2184 		.callback_data = data,
2185 		.workqueue = nfsiod_workqueue,
2186 		.flags = RPC_TASK_ASYNC,
2187 	};
2188 	int status;
2189 
2190 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2191 	kref_get(&data->kref);
2192 	data->rpc_done = 0;
2193 	data->rpc_status = 0;
2194 	data->cancelled = 0;
2195 	data->is_recover = 0;
2196 	if (isrecover) {
2197 		nfs4_set_sequence_privileged(&o_arg->seq_args);
2198 		data->is_recover = 1;
2199 	}
2200 	task = rpc_run_task(&task_setup_data);
2201         if (IS_ERR(task))
2202                 return PTR_ERR(task);
2203         status = nfs4_wait_for_completion_rpc_task(task);
2204         if (status != 0) {
2205                 data->cancelled = 1;
2206                 smp_wmb();
2207         } else
2208                 status = data->rpc_status;
2209         rpc_put_task(task);
2210 
2211 	return status;
2212 }
2213 
2214 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2215 {
2216 	struct inode *dir = d_inode(data->dir);
2217 	struct nfs_openres *o_res = &data->o_res;
2218         int status;
2219 
2220 	status = nfs4_run_open_task(data, 1);
2221 	if (status != 0 || !data->rpc_done)
2222 		return status;
2223 
2224 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2225 
2226 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2227 		status = _nfs4_proc_open_confirm(data);
2228 		if (status != 0)
2229 			return status;
2230 	}
2231 
2232 	return status;
2233 }
2234 
2235 /*
2236  * Additional permission checks in order to distinguish between an
2237  * open for read, and an open for execute. This works around the
2238  * fact that NFSv4 OPEN treats read and execute permissions as being
2239  * the same.
2240  * Note that in the non-execute case, we want to turn off permission
2241  * checking if we just created a new file (POSIX open() semantics).
2242  */
2243 static int nfs4_opendata_access(struct rpc_cred *cred,
2244 				struct nfs4_opendata *opendata,
2245 				struct nfs4_state *state, fmode_t fmode,
2246 				int openflags)
2247 {
2248 	struct nfs_access_entry cache;
2249 	u32 mask;
2250 
2251 	/* access call failed or for some reason the server doesn't
2252 	 * support any access modes -- defer access call until later */
2253 	if (opendata->o_res.access_supported == 0)
2254 		return 0;
2255 
2256 	mask = 0;
2257 	/*
2258 	 * Use openflags to check for exec, because fmode won't
2259 	 * always have FMODE_EXEC set when file open for exec.
2260 	 */
2261 	if (openflags & __FMODE_EXEC) {
2262 		/* ONLY check for exec rights */
2263 		mask = MAY_EXEC;
2264 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2265 		mask = MAY_READ;
2266 
2267 	cache.cred = cred;
2268 	cache.jiffies = jiffies;
2269 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2270 	nfs_access_add_cache(state->inode, &cache);
2271 
2272 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2273 		return 0;
2274 
2275 	/* even though OPEN succeeded, access is denied. Close the file */
2276 	nfs4_close_state(state, fmode);
2277 	return -EACCES;
2278 }
2279 
2280 /*
2281  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2282  */
2283 static int _nfs4_proc_open(struct nfs4_opendata *data)
2284 {
2285 	struct inode *dir = d_inode(data->dir);
2286 	struct nfs_server *server = NFS_SERVER(dir);
2287 	struct nfs_openargs *o_arg = &data->o_arg;
2288 	struct nfs_openres *o_res = &data->o_res;
2289 	int status;
2290 
2291 	status = nfs4_run_open_task(data, 0);
2292 	if (!data->rpc_done)
2293 		return status;
2294 	if (status != 0) {
2295 		if (status == -NFS4ERR_BADNAME &&
2296 				!(o_arg->open_flags & O_CREAT))
2297 			return -ENOENT;
2298 		return status;
2299 	}
2300 
2301 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2302 
2303 	if (o_arg->open_flags & O_CREAT) {
2304 		update_changeattr(dir, &o_res->cinfo);
2305 		if (o_arg->open_flags & O_EXCL)
2306 			data->file_created = 1;
2307 		else if (o_res->cinfo.before != o_res->cinfo.after)
2308 			data->file_created = 1;
2309 	}
2310 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2311 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2312 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2313 		status = _nfs4_proc_open_confirm(data);
2314 		if (status != 0)
2315 			return status;
2316 	}
2317 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2318 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2319 	return 0;
2320 }
2321 
2322 static int nfs4_recover_expired_lease(struct nfs_server *server)
2323 {
2324 	return nfs4_client_recover_expired_lease(server->nfs_client);
2325 }
2326 
2327 /*
2328  * OPEN_EXPIRED:
2329  * 	reclaim state on the server after a network partition.
2330  * 	Assumes caller holds the appropriate lock
2331  */
2332 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2333 {
2334 	struct nfs4_opendata *opendata;
2335 	int ret;
2336 
2337 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2338 			NFS4_OPEN_CLAIM_FH);
2339 	if (IS_ERR(opendata))
2340 		return PTR_ERR(opendata);
2341 	ret = nfs4_open_recover(opendata, state);
2342 	if (ret == -ESTALE)
2343 		d_drop(ctx->dentry);
2344 	nfs4_opendata_put(opendata);
2345 	return ret;
2346 }
2347 
2348 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2349 {
2350 	struct nfs_server *server = NFS_SERVER(state->inode);
2351 	struct nfs4_exception exception = { };
2352 	int err;
2353 
2354 	do {
2355 		err = _nfs4_open_expired(ctx, state);
2356 		trace_nfs4_open_expired(ctx, 0, err);
2357 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2358 			continue;
2359 		switch (err) {
2360 		default:
2361 			goto out;
2362 		case -NFS4ERR_GRACE:
2363 		case -NFS4ERR_DELAY:
2364 			nfs4_handle_exception(server, err, &exception);
2365 			err = 0;
2366 		}
2367 	} while (exception.retry);
2368 out:
2369 	return err;
2370 }
2371 
2372 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2373 {
2374 	struct nfs_open_context *ctx;
2375 	int ret;
2376 
2377 	ctx = nfs4_state_find_open_context(state);
2378 	if (IS_ERR(ctx))
2379 		return -EAGAIN;
2380 	ret = nfs4_do_open_expired(ctx, state);
2381 	put_nfs_open_context(ctx);
2382 	return ret;
2383 }
2384 
2385 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2386 {
2387 	nfs_remove_bad_delegation(state->inode);
2388 	write_seqlock(&state->seqlock);
2389 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2390 	write_sequnlock(&state->seqlock);
2391 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2392 }
2393 
2394 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2395 {
2396 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2397 		nfs_finish_clear_delegation_stateid(state);
2398 }
2399 
2400 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2401 {
2402 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2403 	nfs40_clear_delegation_stateid(state);
2404 	return nfs4_open_expired(sp, state);
2405 }
2406 
2407 #if defined(CONFIG_NFS_V4_1)
2408 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2409 {
2410 	struct nfs_server *server = NFS_SERVER(state->inode);
2411 	nfs4_stateid stateid;
2412 	struct nfs_delegation *delegation;
2413 	struct rpc_cred *cred;
2414 	int status;
2415 
2416 	/* Get the delegation credential for use by test/free_stateid */
2417 	rcu_read_lock();
2418 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2419 	if (delegation == NULL) {
2420 		rcu_read_unlock();
2421 		return;
2422 	}
2423 
2424 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2425 	cred = get_rpccred(delegation->cred);
2426 	rcu_read_unlock();
2427 	status = nfs41_test_stateid(server, &stateid, cred);
2428 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2429 
2430 	if (status != NFS_OK) {
2431 		/* Free the stateid unless the server explicitly
2432 		 * informs us the stateid is unrecognized. */
2433 		if (status != -NFS4ERR_BAD_STATEID)
2434 			nfs41_free_stateid(server, &stateid, cred);
2435 		nfs_finish_clear_delegation_stateid(state);
2436 	}
2437 
2438 	put_rpccred(cred);
2439 }
2440 
2441 /**
2442  * nfs41_check_open_stateid - possibly free an open stateid
2443  *
2444  * @state: NFSv4 state for an inode
2445  *
2446  * Returns NFS_OK if recovery for this stateid is now finished.
2447  * Otherwise a negative NFS4ERR value is returned.
2448  */
2449 static int nfs41_check_open_stateid(struct nfs4_state *state)
2450 {
2451 	struct nfs_server *server = NFS_SERVER(state->inode);
2452 	nfs4_stateid *stateid = &state->open_stateid;
2453 	struct rpc_cred *cred = state->owner->so_cred;
2454 	int status;
2455 
2456 	/* If a state reset has been done, test_stateid is unneeded */
2457 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2458 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2459 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2460 		return -NFS4ERR_BAD_STATEID;
2461 
2462 	status = nfs41_test_stateid(server, stateid, cred);
2463 	trace_nfs4_test_open_stateid(state, NULL, status);
2464 	if (status != NFS_OK) {
2465 		/* Free the stateid unless the server explicitly
2466 		 * informs us the stateid is unrecognized. */
2467 		if (status != -NFS4ERR_BAD_STATEID)
2468 			nfs41_free_stateid(server, stateid, cred);
2469 
2470 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2471 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2472 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2473 		clear_bit(NFS_OPEN_STATE, &state->flags);
2474 	}
2475 	return status;
2476 }
2477 
2478 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2479 {
2480 	int status;
2481 
2482 	nfs41_check_delegation_stateid(state);
2483 	status = nfs41_check_open_stateid(state);
2484 	if (status != NFS_OK)
2485 		status = nfs4_open_expired(sp, state);
2486 	return status;
2487 }
2488 #endif
2489 
2490 /*
2491  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2492  * fields corresponding to attributes that were used to store the verifier.
2493  * Make sure we clobber those fields in the later setattr call
2494  */
2495 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2496 				struct iattr *sattr, struct nfs4_label **label)
2497 {
2498 	const u32 *attrset = opendata->o_res.attrset;
2499 
2500 	if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2501 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2502 		sattr->ia_valid |= ATTR_ATIME;
2503 
2504 	if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2505 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2506 		sattr->ia_valid |= ATTR_MTIME;
2507 
2508 	/* Except MODE, it seems harmless of setting twice. */
2509 	if ((attrset[1] & FATTR4_WORD1_MODE))
2510 		sattr->ia_valid &= ~ATTR_MODE;
2511 
2512 	if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2513 		*label = NULL;
2514 }
2515 
2516 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2517 		fmode_t fmode,
2518 		int flags,
2519 		struct nfs_open_context *ctx)
2520 {
2521 	struct nfs4_state_owner *sp = opendata->owner;
2522 	struct nfs_server *server = sp->so_server;
2523 	struct dentry *dentry;
2524 	struct nfs4_state *state;
2525 	unsigned int seq;
2526 	int ret;
2527 
2528 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2529 
2530 	ret = _nfs4_proc_open(opendata);
2531 	if (ret != 0)
2532 		goto out;
2533 
2534 	state = nfs4_opendata_to_nfs4_state(opendata);
2535 	ret = PTR_ERR(state);
2536 	if (IS_ERR(state))
2537 		goto out;
2538 	if (server->caps & NFS_CAP_POSIX_LOCK)
2539 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2540 
2541 	dentry = opendata->dentry;
2542 	if (d_really_is_negative(dentry)) {
2543 		struct dentry *alias;
2544 		d_drop(dentry);
2545 		alias = d_exact_alias(dentry, state->inode);
2546 		if (!alias)
2547 			alias = d_splice_alias(igrab(state->inode), dentry);
2548 		/* d_splice_alias() can't fail here - it's a non-directory */
2549 		if (alias) {
2550 			dput(ctx->dentry);
2551 			ctx->dentry = dentry = alias;
2552 		}
2553 		nfs_set_verifier(dentry,
2554 				nfs_save_change_attribute(d_inode(opendata->dir)));
2555 	}
2556 
2557 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2558 	if (ret != 0)
2559 		goto out;
2560 
2561 	ctx->state = state;
2562 	if (d_inode(dentry) == state->inode) {
2563 		nfs_inode_attach_open_context(ctx);
2564 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2565 			nfs4_schedule_stateid_recovery(server, state);
2566 	}
2567 out:
2568 	return ret;
2569 }
2570 
2571 /*
2572  * Returns a referenced nfs4_state
2573  */
2574 static int _nfs4_do_open(struct inode *dir,
2575 			struct nfs_open_context *ctx,
2576 			int flags,
2577 			struct iattr *sattr,
2578 			struct nfs4_label *label,
2579 			int *opened)
2580 {
2581 	struct nfs4_state_owner  *sp;
2582 	struct nfs4_state     *state = NULL;
2583 	struct nfs_server       *server = NFS_SERVER(dir);
2584 	struct nfs4_opendata *opendata;
2585 	struct dentry *dentry = ctx->dentry;
2586 	struct rpc_cred *cred = ctx->cred;
2587 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2588 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2589 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2590 	struct nfs4_label *olabel = NULL;
2591 	int status;
2592 
2593 	/* Protect against reboot recovery conflicts */
2594 	status = -ENOMEM;
2595 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2596 	if (sp == NULL) {
2597 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2598 		goto out_err;
2599 	}
2600 	status = nfs4_recover_expired_lease(server);
2601 	if (status != 0)
2602 		goto err_put_state_owner;
2603 	if (d_really_is_positive(dentry))
2604 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2605 	status = -ENOMEM;
2606 	if (d_really_is_positive(dentry))
2607 		claim = NFS4_OPEN_CLAIM_FH;
2608 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2609 			label, claim, GFP_KERNEL);
2610 	if (opendata == NULL)
2611 		goto err_put_state_owner;
2612 
2613 	if (label) {
2614 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2615 		if (IS_ERR(olabel)) {
2616 			status = PTR_ERR(olabel);
2617 			goto err_opendata_put;
2618 		}
2619 	}
2620 
2621 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2622 		if (!opendata->f_attr.mdsthreshold) {
2623 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2624 			if (!opendata->f_attr.mdsthreshold)
2625 				goto err_free_label;
2626 		}
2627 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2628 	}
2629 	if (d_really_is_positive(dentry))
2630 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2631 
2632 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2633 	if (status != 0)
2634 		goto err_free_label;
2635 	state = ctx->state;
2636 
2637 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2638 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2639 		nfs4_exclusive_attrset(opendata, sattr, &label);
2640 		/*
2641 		 * send create attributes which was not set by open
2642 		 * with an extra setattr.
2643 		 */
2644 		if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2645 			nfs_fattr_init(opendata->o_res.f_attr);
2646 			status = nfs4_do_setattr(state->inode, cred,
2647 					opendata->o_res.f_attr, sattr,
2648 					state, label, olabel);
2649 			if (status == 0) {
2650 				nfs_setattr_update_inode(state->inode, sattr,
2651 						opendata->o_res.f_attr);
2652 				nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2653 			}
2654 		}
2655 	}
2656 	if (opened && opendata->file_created)
2657 		*opened |= FILE_CREATED;
2658 
2659 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2660 		*ctx_th = opendata->f_attr.mdsthreshold;
2661 		opendata->f_attr.mdsthreshold = NULL;
2662 	}
2663 
2664 	nfs4_label_free(olabel);
2665 
2666 	nfs4_opendata_put(opendata);
2667 	nfs4_put_state_owner(sp);
2668 	return 0;
2669 err_free_label:
2670 	nfs4_label_free(olabel);
2671 err_opendata_put:
2672 	nfs4_opendata_put(opendata);
2673 err_put_state_owner:
2674 	nfs4_put_state_owner(sp);
2675 out_err:
2676 	return status;
2677 }
2678 
2679 
2680 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2681 					struct nfs_open_context *ctx,
2682 					int flags,
2683 					struct iattr *sattr,
2684 					struct nfs4_label *label,
2685 					int *opened)
2686 {
2687 	struct nfs_server *server = NFS_SERVER(dir);
2688 	struct nfs4_exception exception = { };
2689 	struct nfs4_state *res;
2690 	int status;
2691 
2692 	do {
2693 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2694 		res = ctx->state;
2695 		trace_nfs4_open_file(ctx, flags, status);
2696 		if (status == 0)
2697 			break;
2698 		/* NOTE: BAD_SEQID means the server and client disagree about the
2699 		 * book-keeping w.r.t. state-changing operations
2700 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2701 		 * It is actually a sign of a bug on the client or on the server.
2702 		 *
2703 		 * If we receive a BAD_SEQID error in the particular case of
2704 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2705 		 * have unhashed the old state_owner for us, and that we can
2706 		 * therefore safely retry using a new one. We should still warn
2707 		 * the user though...
2708 		 */
2709 		if (status == -NFS4ERR_BAD_SEQID) {
2710 			pr_warn_ratelimited("NFS: v4 server %s "
2711 					" returned a bad sequence-id error!\n",
2712 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2713 			exception.retry = 1;
2714 			continue;
2715 		}
2716 		/*
2717 		 * BAD_STATEID on OPEN means that the server cancelled our
2718 		 * state before it received the OPEN_CONFIRM.
2719 		 * Recover by retrying the request as per the discussion
2720 		 * on Page 181 of RFC3530.
2721 		 */
2722 		if (status == -NFS4ERR_BAD_STATEID) {
2723 			exception.retry = 1;
2724 			continue;
2725 		}
2726 		if (status == -EAGAIN) {
2727 			/* We must have found a delegation */
2728 			exception.retry = 1;
2729 			continue;
2730 		}
2731 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2732 			continue;
2733 		res = ERR_PTR(nfs4_handle_exception(server,
2734 					status, &exception));
2735 	} while (exception.retry);
2736 	return res;
2737 }
2738 
2739 static int _nfs4_do_setattr(struct inode *inode,
2740 			    struct nfs_setattrargs *arg,
2741 			    struct nfs_setattrres *res,
2742 			    struct rpc_cred *cred,
2743 			    struct nfs4_state *state)
2744 {
2745 	struct nfs_server *server = NFS_SERVER(inode);
2746         struct rpc_message msg = {
2747 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2748 		.rpc_argp	= arg,
2749 		.rpc_resp	= res,
2750 		.rpc_cred	= cred,
2751         };
2752 	struct rpc_cred *delegation_cred = NULL;
2753 	unsigned long timestamp = jiffies;
2754 	fmode_t fmode;
2755 	bool truncate;
2756 	int status;
2757 
2758 	nfs_fattr_init(res->fattr);
2759 
2760 	/* Servers should only apply open mode checks for file size changes */
2761 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2762 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2763 
2764 	if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2765 		/* Use that stateid */
2766 	} else if (truncate && state != NULL) {
2767 		struct nfs_lockowner lockowner = {
2768 			.l_owner = current->files,
2769 			.l_pid = current->tgid,
2770 		};
2771 		if (!nfs4_valid_open_stateid(state))
2772 			return -EBADF;
2773 		if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2774 				&arg->stateid, &delegation_cred) == -EIO)
2775 			return -EBADF;
2776 	} else
2777 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2778 	if (delegation_cred)
2779 		msg.rpc_cred = delegation_cred;
2780 
2781 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2782 
2783 	put_rpccred(delegation_cred);
2784 	if (status == 0 && state != NULL)
2785 		renew_lease(server, timestamp);
2786 	trace_nfs4_setattr(inode, &arg->stateid, status);
2787 	return status;
2788 }
2789 
2790 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2791 			   struct nfs_fattr *fattr, struct iattr *sattr,
2792 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2793 			   struct nfs4_label *olabel)
2794 {
2795 	struct nfs_server *server = NFS_SERVER(inode);
2796         struct nfs_setattrargs  arg = {
2797                 .fh             = NFS_FH(inode),
2798                 .iap            = sattr,
2799 		.server		= server,
2800 		.bitmask = server->attr_bitmask,
2801 		.label		= ilabel,
2802         };
2803         struct nfs_setattrres  res = {
2804 		.fattr		= fattr,
2805 		.label		= olabel,
2806 		.server		= server,
2807         };
2808 	struct nfs4_exception exception = {
2809 		.state = state,
2810 		.inode = inode,
2811 		.stateid = &arg.stateid,
2812 	};
2813 	int err;
2814 
2815 	arg.bitmask = nfs4_bitmask(server, ilabel);
2816 	if (ilabel)
2817 		arg.bitmask = nfs4_bitmask(server, olabel);
2818 
2819 	do {
2820 		err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
2821 		switch (err) {
2822 		case -NFS4ERR_OPENMODE:
2823 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2824 				pr_warn_once("NFSv4: server %s is incorrectly "
2825 						"applying open mode checks to "
2826 						"a SETATTR that is not "
2827 						"changing file size.\n",
2828 						server->nfs_client->cl_hostname);
2829 			}
2830 			if (state && !(state->state & FMODE_WRITE)) {
2831 				err = -EBADF;
2832 				if (sattr->ia_valid & ATTR_OPEN)
2833 					err = -EACCES;
2834 				goto out;
2835 			}
2836 		}
2837 		err = nfs4_handle_exception(server, err, &exception);
2838 	} while (exception.retry);
2839 out:
2840 	return err;
2841 }
2842 
2843 static bool
2844 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2845 {
2846 	if (inode == NULL || !nfs_have_layout(inode))
2847 		return false;
2848 
2849 	return pnfs_wait_on_layoutreturn(inode, task);
2850 }
2851 
2852 struct nfs4_closedata {
2853 	struct inode *inode;
2854 	struct nfs4_state *state;
2855 	struct nfs_closeargs arg;
2856 	struct nfs_closeres res;
2857 	struct nfs_fattr fattr;
2858 	unsigned long timestamp;
2859 	bool roc;
2860 	u32 roc_barrier;
2861 };
2862 
2863 static void nfs4_free_closedata(void *data)
2864 {
2865 	struct nfs4_closedata *calldata = data;
2866 	struct nfs4_state_owner *sp = calldata->state->owner;
2867 	struct super_block *sb = calldata->state->inode->i_sb;
2868 
2869 	if (calldata->roc)
2870 		pnfs_roc_release(calldata->state->inode);
2871 	nfs4_put_open_state(calldata->state);
2872 	nfs_free_seqid(calldata->arg.seqid);
2873 	nfs4_put_state_owner(sp);
2874 	nfs_sb_deactive(sb);
2875 	kfree(calldata);
2876 }
2877 
2878 static void nfs4_close_done(struct rpc_task *task, void *data)
2879 {
2880 	struct nfs4_closedata *calldata = data;
2881 	struct nfs4_state *state = calldata->state;
2882 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2883 	nfs4_stateid *res_stateid = NULL;
2884 
2885 	dprintk("%s: begin!\n", __func__);
2886 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2887 		return;
2888 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2889         /* hmm. we are done with the inode, and in the process of freeing
2890 	 * the state_owner. we keep this around to process errors
2891 	 */
2892 	switch (task->tk_status) {
2893 		case 0:
2894 			res_stateid = &calldata->res.stateid;
2895 			if (calldata->roc)
2896 				pnfs_roc_set_barrier(state->inode,
2897 						     calldata->roc_barrier);
2898 			renew_lease(server, calldata->timestamp);
2899 			break;
2900 		case -NFS4ERR_ADMIN_REVOKED:
2901 		case -NFS4ERR_STALE_STATEID:
2902 		case -NFS4ERR_OLD_STATEID:
2903 		case -NFS4ERR_BAD_STATEID:
2904 		case -NFS4ERR_EXPIRED:
2905 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2906 						&state->open_stateid)) {
2907 				rpc_restart_call_prepare(task);
2908 				goto out_release;
2909 			}
2910 			if (calldata->arg.fmode == 0)
2911 				break;
2912 		default:
2913 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2914 				rpc_restart_call_prepare(task);
2915 				goto out_release;
2916 			}
2917 	}
2918 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
2919 			res_stateid, calldata->arg.fmode);
2920 out_release:
2921 	nfs_release_seqid(calldata->arg.seqid);
2922 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2923 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2924 }
2925 
2926 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2927 {
2928 	struct nfs4_closedata *calldata = data;
2929 	struct nfs4_state *state = calldata->state;
2930 	struct inode *inode = calldata->inode;
2931 	bool is_rdonly, is_wronly, is_rdwr;
2932 	int call_close = 0;
2933 
2934 	dprintk("%s: begin!\n", __func__);
2935 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2936 		goto out_wait;
2937 
2938 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2939 	spin_lock(&state->owner->so_lock);
2940 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2941 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2942 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2943 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2944 	/* Calculate the change in open mode */
2945 	calldata->arg.fmode = 0;
2946 	if (state->n_rdwr == 0) {
2947 		if (state->n_rdonly == 0)
2948 			call_close |= is_rdonly;
2949 		else if (is_rdonly)
2950 			calldata->arg.fmode |= FMODE_READ;
2951 		if (state->n_wronly == 0)
2952 			call_close |= is_wronly;
2953 		else if (is_wronly)
2954 			calldata->arg.fmode |= FMODE_WRITE;
2955 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
2956 			call_close |= is_rdwr;
2957 	} else if (is_rdwr)
2958 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2959 
2960 	if (!nfs4_valid_open_stateid(state))
2961 		call_close = 0;
2962 	spin_unlock(&state->owner->so_lock);
2963 
2964 	if (!call_close) {
2965 		/* Note: exit _without_ calling nfs4_close_done */
2966 		goto out_no_action;
2967 	}
2968 
2969 	if (nfs4_wait_on_layoutreturn(inode, task)) {
2970 		nfs_release_seqid(calldata->arg.seqid);
2971 		goto out_wait;
2972 	}
2973 
2974 	if (calldata->arg.fmode == 0)
2975 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2976 	if (calldata->roc)
2977 		pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2978 
2979 	calldata->arg.share_access =
2980 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2981 				calldata->arg.fmode, 0);
2982 
2983 	nfs_fattr_init(calldata->res.fattr);
2984 	calldata->timestamp = jiffies;
2985 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2986 				&calldata->arg.seq_args,
2987 				&calldata->res.seq_res,
2988 				task) != 0)
2989 		nfs_release_seqid(calldata->arg.seqid);
2990 	dprintk("%s: done!\n", __func__);
2991 	return;
2992 out_no_action:
2993 	task->tk_action = NULL;
2994 out_wait:
2995 	nfs4_sequence_done(task, &calldata->res.seq_res);
2996 }
2997 
2998 static const struct rpc_call_ops nfs4_close_ops = {
2999 	.rpc_call_prepare = nfs4_close_prepare,
3000 	.rpc_call_done = nfs4_close_done,
3001 	.rpc_release = nfs4_free_closedata,
3002 };
3003 
3004 static bool nfs4_roc(struct inode *inode)
3005 {
3006 	if (!nfs_have_layout(inode))
3007 		return false;
3008 	return pnfs_roc(inode);
3009 }
3010 
3011 /*
3012  * It is possible for data to be read/written from a mem-mapped file
3013  * after the sys_close call (which hits the vfs layer as a flush).
3014  * This means that we can't safely call nfsv4 close on a file until
3015  * the inode is cleared. This in turn means that we are not good
3016  * NFSv4 citizens - we do not indicate to the server to update the file's
3017  * share state even when we are done with one of the three share
3018  * stateid's in the inode.
3019  *
3020  * NOTE: Caller must be holding the sp->so_owner semaphore!
3021  */
3022 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3023 {
3024 	struct nfs_server *server = NFS_SERVER(state->inode);
3025 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3026 	struct nfs4_closedata *calldata;
3027 	struct nfs4_state_owner *sp = state->owner;
3028 	struct rpc_task *task;
3029 	struct rpc_message msg = {
3030 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3031 		.rpc_cred = state->owner->so_cred,
3032 	};
3033 	struct rpc_task_setup task_setup_data = {
3034 		.rpc_client = server->client,
3035 		.rpc_message = &msg,
3036 		.callback_ops = &nfs4_close_ops,
3037 		.workqueue = nfsiod_workqueue,
3038 		.flags = RPC_TASK_ASYNC,
3039 	};
3040 	int status = -ENOMEM;
3041 
3042 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3043 		&task_setup_data.rpc_client, &msg);
3044 
3045 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3046 	if (calldata == NULL)
3047 		goto out;
3048 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3049 	calldata->inode = state->inode;
3050 	calldata->state = state;
3051 	calldata->arg.fh = NFS_FH(state->inode);
3052 	/* Serialization for the sequence id */
3053 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3054 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3055 	if (IS_ERR(calldata->arg.seqid))
3056 		goto out_free_calldata;
3057 	calldata->arg.fmode = 0;
3058 	calldata->arg.bitmask = server->cache_consistency_bitmask;
3059 	calldata->res.fattr = &calldata->fattr;
3060 	calldata->res.seqid = calldata->arg.seqid;
3061 	calldata->res.server = server;
3062 	calldata->roc = nfs4_roc(state->inode);
3063 	nfs_sb_active(calldata->inode->i_sb);
3064 
3065 	msg.rpc_argp = &calldata->arg;
3066 	msg.rpc_resp = &calldata->res;
3067 	task_setup_data.callback_data = calldata;
3068 	task = rpc_run_task(&task_setup_data);
3069 	if (IS_ERR(task))
3070 		return PTR_ERR(task);
3071 	status = 0;
3072 	if (wait)
3073 		status = rpc_wait_for_completion_task(task);
3074 	rpc_put_task(task);
3075 	return status;
3076 out_free_calldata:
3077 	kfree(calldata);
3078 out:
3079 	nfs4_put_open_state(state);
3080 	nfs4_put_state_owner(sp);
3081 	return status;
3082 }
3083 
3084 static struct inode *
3085 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3086 		int open_flags, struct iattr *attr, int *opened)
3087 {
3088 	struct nfs4_state *state;
3089 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3090 
3091 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3092 
3093 	/* Protect against concurrent sillydeletes */
3094 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3095 
3096 	nfs4_label_release_security(label);
3097 
3098 	if (IS_ERR(state))
3099 		return ERR_CAST(state);
3100 	return state->inode;
3101 }
3102 
3103 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3104 {
3105 	if (ctx->state == NULL)
3106 		return;
3107 	if (is_sync)
3108 		nfs4_close_sync(ctx->state, ctx->mode);
3109 	else
3110 		nfs4_close_state(ctx->state, ctx->mode);
3111 }
3112 
3113 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3114 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3115 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3116 
3117 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3118 {
3119 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3120 	struct nfs4_server_caps_arg args = {
3121 		.fhandle = fhandle,
3122 		.bitmask = bitmask,
3123 	};
3124 	struct nfs4_server_caps_res res = {};
3125 	struct rpc_message msg = {
3126 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3127 		.rpc_argp = &args,
3128 		.rpc_resp = &res,
3129 	};
3130 	int status;
3131 
3132 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3133 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3134 		     FATTR4_WORD0_LINK_SUPPORT |
3135 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3136 		     FATTR4_WORD0_ACLSUPPORT;
3137 	if (minorversion)
3138 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3139 
3140 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3141 	if (status == 0) {
3142 		/* Sanity check the server answers */
3143 		switch (minorversion) {
3144 		case 0:
3145 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3146 			res.attr_bitmask[2] = 0;
3147 			break;
3148 		case 1:
3149 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3150 			break;
3151 		case 2:
3152 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3153 		}
3154 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3155 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3156 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3157 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3158 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3159 				NFS_CAP_CTIME|NFS_CAP_MTIME|
3160 				NFS_CAP_SECURITY_LABEL);
3161 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3162 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3163 			server->caps |= NFS_CAP_ACLS;
3164 		if (res.has_links != 0)
3165 			server->caps |= NFS_CAP_HARDLINKS;
3166 		if (res.has_symlinks != 0)
3167 			server->caps |= NFS_CAP_SYMLINKS;
3168 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3169 			server->caps |= NFS_CAP_FILEID;
3170 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3171 			server->caps |= NFS_CAP_MODE;
3172 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3173 			server->caps |= NFS_CAP_NLINK;
3174 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3175 			server->caps |= NFS_CAP_OWNER;
3176 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3177 			server->caps |= NFS_CAP_OWNER_GROUP;
3178 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3179 			server->caps |= NFS_CAP_ATIME;
3180 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3181 			server->caps |= NFS_CAP_CTIME;
3182 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3183 			server->caps |= NFS_CAP_MTIME;
3184 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3185 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3186 			server->caps |= NFS_CAP_SECURITY_LABEL;
3187 #endif
3188 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3189 				sizeof(server->attr_bitmask));
3190 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3191 
3192 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3193 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3194 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3195 		server->cache_consistency_bitmask[2] = 0;
3196 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3197 			sizeof(server->exclcreat_bitmask));
3198 		server->acl_bitmask = res.acl_bitmask;
3199 		server->fh_expire_type = res.fh_expire_type;
3200 	}
3201 
3202 	return status;
3203 }
3204 
3205 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3206 {
3207 	struct nfs4_exception exception = { };
3208 	int err;
3209 	do {
3210 		err = nfs4_handle_exception(server,
3211 				_nfs4_server_capabilities(server, fhandle),
3212 				&exception);
3213 	} while (exception.retry);
3214 	return err;
3215 }
3216 
3217 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3218 		struct nfs_fsinfo *info)
3219 {
3220 	u32 bitmask[3];
3221 	struct nfs4_lookup_root_arg args = {
3222 		.bitmask = bitmask,
3223 	};
3224 	struct nfs4_lookup_res res = {
3225 		.server = server,
3226 		.fattr = info->fattr,
3227 		.fh = fhandle,
3228 	};
3229 	struct rpc_message msg = {
3230 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3231 		.rpc_argp = &args,
3232 		.rpc_resp = &res,
3233 	};
3234 
3235 	bitmask[0] = nfs4_fattr_bitmap[0];
3236 	bitmask[1] = nfs4_fattr_bitmap[1];
3237 	/*
3238 	 * Process the label in the upcoming getfattr
3239 	 */
3240 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3241 
3242 	nfs_fattr_init(info->fattr);
3243 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3244 }
3245 
3246 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3247 		struct nfs_fsinfo *info)
3248 {
3249 	struct nfs4_exception exception = { };
3250 	int err;
3251 	do {
3252 		err = _nfs4_lookup_root(server, fhandle, info);
3253 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3254 		switch (err) {
3255 		case 0:
3256 		case -NFS4ERR_WRONGSEC:
3257 			goto out;
3258 		default:
3259 			err = nfs4_handle_exception(server, err, &exception);
3260 		}
3261 	} while (exception.retry);
3262 out:
3263 	return err;
3264 }
3265 
3266 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3267 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3268 {
3269 	struct rpc_auth_create_args auth_args = {
3270 		.pseudoflavor = flavor,
3271 	};
3272 	struct rpc_auth *auth;
3273 	int ret;
3274 
3275 	auth = rpcauth_create(&auth_args, server->client);
3276 	if (IS_ERR(auth)) {
3277 		ret = -EACCES;
3278 		goto out;
3279 	}
3280 	ret = nfs4_lookup_root(server, fhandle, info);
3281 out:
3282 	return ret;
3283 }
3284 
3285 /*
3286  * Retry pseudoroot lookup with various security flavors.  We do this when:
3287  *
3288  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3289  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3290  *
3291  * Returns zero on success, or a negative NFS4ERR value, or a
3292  * negative errno value.
3293  */
3294 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3295 			      struct nfs_fsinfo *info)
3296 {
3297 	/* Per 3530bis 15.33.5 */
3298 	static const rpc_authflavor_t flav_array[] = {
3299 		RPC_AUTH_GSS_KRB5P,
3300 		RPC_AUTH_GSS_KRB5I,
3301 		RPC_AUTH_GSS_KRB5,
3302 		RPC_AUTH_UNIX,			/* courtesy */
3303 		RPC_AUTH_NULL,
3304 	};
3305 	int status = -EPERM;
3306 	size_t i;
3307 
3308 	if (server->auth_info.flavor_len > 0) {
3309 		/* try each flavor specified by user */
3310 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3311 			status = nfs4_lookup_root_sec(server, fhandle, info,
3312 						server->auth_info.flavors[i]);
3313 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3314 				continue;
3315 			break;
3316 		}
3317 	} else {
3318 		/* no flavors specified by user, try default list */
3319 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3320 			status = nfs4_lookup_root_sec(server, fhandle, info,
3321 						      flav_array[i]);
3322 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3323 				continue;
3324 			break;
3325 		}
3326 	}
3327 
3328 	/*
3329 	 * -EACCESS could mean that the user doesn't have correct permissions
3330 	 * to access the mount.  It could also mean that we tried to mount
3331 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3332 	 * existing mount programs don't handle -EACCES very well so it should
3333 	 * be mapped to -EPERM instead.
3334 	 */
3335 	if (status == -EACCES)
3336 		status = -EPERM;
3337 	return status;
3338 }
3339 
3340 /**
3341  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3342  * @server: initialized nfs_server handle
3343  * @fhandle: we fill in the pseudo-fs root file handle
3344  * @info: we fill in an FSINFO struct
3345  * @auth_probe: probe the auth flavours
3346  *
3347  * Returns zero on success, or a negative errno.
3348  */
3349 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3350 			 struct nfs_fsinfo *info,
3351 			 bool auth_probe)
3352 {
3353 	int status = 0;
3354 
3355 	if (!auth_probe)
3356 		status = nfs4_lookup_root(server, fhandle, info);
3357 
3358 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3359 		status = server->nfs_client->cl_mvops->find_root_sec(server,
3360 				fhandle, info);
3361 
3362 	if (status == 0)
3363 		status = nfs4_server_capabilities(server, fhandle);
3364 	if (status == 0)
3365 		status = nfs4_do_fsinfo(server, fhandle, info);
3366 
3367 	return nfs4_map_errors(status);
3368 }
3369 
3370 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3371 			      struct nfs_fsinfo *info)
3372 {
3373 	int error;
3374 	struct nfs_fattr *fattr = info->fattr;
3375 	struct nfs4_label *label = NULL;
3376 
3377 	error = nfs4_server_capabilities(server, mntfh);
3378 	if (error < 0) {
3379 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3380 		return error;
3381 	}
3382 
3383 	label = nfs4_label_alloc(server, GFP_KERNEL);
3384 	if (IS_ERR(label))
3385 		return PTR_ERR(label);
3386 
3387 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3388 	if (error < 0) {
3389 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3390 		goto err_free_label;
3391 	}
3392 
3393 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3394 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3395 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3396 
3397 err_free_label:
3398 	nfs4_label_free(label);
3399 
3400 	return error;
3401 }
3402 
3403 /*
3404  * Get locations and (maybe) other attributes of a referral.
3405  * Note that we'll actually follow the referral later when
3406  * we detect fsid mismatch in inode revalidation
3407  */
3408 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3409 			     const struct qstr *name, struct nfs_fattr *fattr,
3410 			     struct nfs_fh *fhandle)
3411 {
3412 	int status = -ENOMEM;
3413 	struct page *page = NULL;
3414 	struct nfs4_fs_locations *locations = NULL;
3415 
3416 	page = alloc_page(GFP_KERNEL);
3417 	if (page == NULL)
3418 		goto out;
3419 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3420 	if (locations == NULL)
3421 		goto out;
3422 
3423 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3424 	if (status != 0)
3425 		goto out;
3426 
3427 	/*
3428 	 * If the fsid didn't change, this is a migration event, not a
3429 	 * referral.  Cause us to drop into the exception handler, which
3430 	 * will kick off migration recovery.
3431 	 */
3432 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3433 		dprintk("%s: server did not return a different fsid for"
3434 			" a referral at %s\n", __func__, name->name);
3435 		status = -NFS4ERR_MOVED;
3436 		goto out;
3437 	}
3438 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3439 	nfs_fixup_referral_attributes(&locations->fattr);
3440 
3441 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3442 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3443 	memset(fhandle, 0, sizeof(struct nfs_fh));
3444 out:
3445 	if (page)
3446 		__free_page(page);
3447 	kfree(locations);
3448 	return status;
3449 }
3450 
3451 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3452 				struct nfs_fattr *fattr, struct nfs4_label *label)
3453 {
3454 	struct nfs4_getattr_arg args = {
3455 		.fh = fhandle,
3456 		.bitmask = server->attr_bitmask,
3457 	};
3458 	struct nfs4_getattr_res res = {
3459 		.fattr = fattr,
3460 		.label = label,
3461 		.server = server,
3462 	};
3463 	struct rpc_message msg = {
3464 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3465 		.rpc_argp = &args,
3466 		.rpc_resp = &res,
3467 	};
3468 
3469 	args.bitmask = nfs4_bitmask(server, label);
3470 
3471 	nfs_fattr_init(fattr);
3472 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3473 }
3474 
3475 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3476 				struct nfs_fattr *fattr, struct nfs4_label *label)
3477 {
3478 	struct nfs4_exception exception = { };
3479 	int err;
3480 	do {
3481 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3482 		trace_nfs4_getattr(server, fhandle, fattr, err);
3483 		err = nfs4_handle_exception(server, err,
3484 				&exception);
3485 	} while (exception.retry);
3486 	return err;
3487 }
3488 
3489 /*
3490  * The file is not closed if it is opened due to the a request to change
3491  * the size of the file. The open call will not be needed once the
3492  * VFS layer lookup-intents are implemented.
3493  *
3494  * Close is called when the inode is destroyed.
3495  * If we haven't opened the file for O_WRONLY, we
3496  * need to in the size_change case to obtain a stateid.
3497  *
3498  * Got race?
3499  * Because OPEN is always done by name in nfsv4, it is
3500  * possible that we opened a different file by the same
3501  * name.  We can recognize this race condition, but we
3502  * can't do anything about it besides returning an error.
3503  *
3504  * This will be fixed with VFS changes (lookup-intent).
3505  */
3506 static int
3507 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3508 		  struct iattr *sattr)
3509 {
3510 	struct inode *inode = d_inode(dentry);
3511 	struct rpc_cred *cred = NULL;
3512 	struct nfs4_state *state = NULL;
3513 	struct nfs4_label *label = NULL;
3514 	int status;
3515 
3516 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3517 	    sattr->ia_valid & ATTR_SIZE &&
3518 	    sattr->ia_size < i_size_read(inode))
3519 		pnfs_commit_and_return_layout(inode);
3520 
3521 	nfs_fattr_init(fattr);
3522 
3523 	/* Deal with open(O_TRUNC) */
3524 	if (sattr->ia_valid & ATTR_OPEN)
3525 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3526 
3527 	/* Optimization: if the end result is no change, don't RPC */
3528 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3529 		return 0;
3530 
3531 	/* Search for an existing open(O_WRITE) file */
3532 	if (sattr->ia_valid & ATTR_FILE) {
3533 		struct nfs_open_context *ctx;
3534 
3535 		ctx = nfs_file_open_context(sattr->ia_file);
3536 		if (ctx) {
3537 			cred = ctx->cred;
3538 			state = ctx->state;
3539 		}
3540 	}
3541 
3542 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3543 	if (IS_ERR(label))
3544 		return PTR_ERR(label);
3545 
3546 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3547 	if (status == 0) {
3548 		nfs_setattr_update_inode(inode, sattr, fattr);
3549 		nfs_setsecurity(inode, fattr, label);
3550 	}
3551 	nfs4_label_free(label);
3552 	return status;
3553 }
3554 
3555 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3556 		const struct qstr *name, struct nfs_fh *fhandle,
3557 		struct nfs_fattr *fattr, struct nfs4_label *label)
3558 {
3559 	struct nfs_server *server = NFS_SERVER(dir);
3560 	int		       status;
3561 	struct nfs4_lookup_arg args = {
3562 		.bitmask = server->attr_bitmask,
3563 		.dir_fh = NFS_FH(dir),
3564 		.name = name,
3565 	};
3566 	struct nfs4_lookup_res res = {
3567 		.server = server,
3568 		.fattr = fattr,
3569 		.label = label,
3570 		.fh = fhandle,
3571 	};
3572 	struct rpc_message msg = {
3573 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3574 		.rpc_argp = &args,
3575 		.rpc_resp = &res,
3576 	};
3577 
3578 	args.bitmask = nfs4_bitmask(server, label);
3579 
3580 	nfs_fattr_init(fattr);
3581 
3582 	dprintk("NFS call  lookup %s\n", name->name);
3583 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3584 	dprintk("NFS reply lookup: %d\n", status);
3585 	return status;
3586 }
3587 
3588 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3589 {
3590 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3591 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3592 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3593 	fattr->nlink = 2;
3594 }
3595 
3596 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3597 				   const struct qstr *name, struct nfs_fh *fhandle,
3598 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3599 {
3600 	struct nfs4_exception exception = { };
3601 	struct rpc_clnt *client = *clnt;
3602 	int err;
3603 	do {
3604 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3605 		trace_nfs4_lookup(dir, name, err);
3606 		switch (err) {
3607 		case -NFS4ERR_BADNAME:
3608 			err = -ENOENT;
3609 			goto out;
3610 		case -NFS4ERR_MOVED:
3611 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3612 			if (err == -NFS4ERR_MOVED)
3613 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3614 			goto out;
3615 		case -NFS4ERR_WRONGSEC:
3616 			err = -EPERM;
3617 			if (client != *clnt)
3618 				goto out;
3619 			client = nfs4_negotiate_security(client, dir, name);
3620 			if (IS_ERR(client))
3621 				return PTR_ERR(client);
3622 
3623 			exception.retry = 1;
3624 			break;
3625 		default:
3626 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3627 		}
3628 	} while (exception.retry);
3629 
3630 out:
3631 	if (err == 0)
3632 		*clnt = client;
3633 	else if (client != *clnt)
3634 		rpc_shutdown_client(client);
3635 
3636 	return err;
3637 }
3638 
3639 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3640 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3641 			    struct nfs4_label *label)
3642 {
3643 	int status;
3644 	struct rpc_clnt *client = NFS_CLIENT(dir);
3645 
3646 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3647 	if (client != NFS_CLIENT(dir)) {
3648 		rpc_shutdown_client(client);
3649 		nfs_fixup_secinfo_attributes(fattr);
3650 	}
3651 	return status;
3652 }
3653 
3654 struct rpc_clnt *
3655 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3656 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3657 {
3658 	struct rpc_clnt *client = NFS_CLIENT(dir);
3659 	int status;
3660 
3661 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3662 	if (status < 0)
3663 		return ERR_PTR(status);
3664 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3665 }
3666 
3667 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3668 {
3669 	struct nfs_server *server = NFS_SERVER(inode);
3670 	struct nfs4_accessargs args = {
3671 		.fh = NFS_FH(inode),
3672 		.bitmask = server->cache_consistency_bitmask,
3673 	};
3674 	struct nfs4_accessres res = {
3675 		.server = server,
3676 	};
3677 	struct rpc_message msg = {
3678 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3679 		.rpc_argp = &args,
3680 		.rpc_resp = &res,
3681 		.rpc_cred = entry->cred,
3682 	};
3683 	int mode = entry->mask;
3684 	int status = 0;
3685 
3686 	/*
3687 	 * Determine which access bits we want to ask for...
3688 	 */
3689 	if (mode & MAY_READ)
3690 		args.access |= NFS4_ACCESS_READ;
3691 	if (S_ISDIR(inode->i_mode)) {
3692 		if (mode & MAY_WRITE)
3693 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3694 		if (mode & MAY_EXEC)
3695 			args.access |= NFS4_ACCESS_LOOKUP;
3696 	} else {
3697 		if (mode & MAY_WRITE)
3698 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3699 		if (mode & MAY_EXEC)
3700 			args.access |= NFS4_ACCESS_EXECUTE;
3701 	}
3702 
3703 	res.fattr = nfs_alloc_fattr();
3704 	if (res.fattr == NULL)
3705 		return -ENOMEM;
3706 
3707 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3708 	if (!status) {
3709 		nfs_access_set_mask(entry, res.access);
3710 		nfs_refresh_inode(inode, res.fattr);
3711 	}
3712 	nfs_free_fattr(res.fattr);
3713 	return status;
3714 }
3715 
3716 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3717 {
3718 	struct nfs4_exception exception = { };
3719 	int err;
3720 	do {
3721 		err = _nfs4_proc_access(inode, entry);
3722 		trace_nfs4_access(inode, err);
3723 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3724 				&exception);
3725 	} while (exception.retry);
3726 	return err;
3727 }
3728 
3729 /*
3730  * TODO: For the time being, we don't try to get any attributes
3731  * along with any of the zero-copy operations READ, READDIR,
3732  * READLINK, WRITE.
3733  *
3734  * In the case of the first three, we want to put the GETATTR
3735  * after the read-type operation -- this is because it is hard
3736  * to predict the length of a GETATTR response in v4, and thus
3737  * align the READ data correctly.  This means that the GETATTR
3738  * may end up partially falling into the page cache, and we should
3739  * shift it into the 'tail' of the xdr_buf before processing.
3740  * To do this efficiently, we need to know the total length
3741  * of data received, which doesn't seem to be available outside
3742  * of the RPC layer.
3743  *
3744  * In the case of WRITE, we also want to put the GETATTR after
3745  * the operation -- in this case because we want to make sure
3746  * we get the post-operation mtime and size.
3747  *
3748  * Both of these changes to the XDR layer would in fact be quite
3749  * minor, but I decided to leave them for a subsequent patch.
3750  */
3751 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3752 		unsigned int pgbase, unsigned int pglen)
3753 {
3754 	struct nfs4_readlink args = {
3755 		.fh       = NFS_FH(inode),
3756 		.pgbase	  = pgbase,
3757 		.pglen    = pglen,
3758 		.pages    = &page,
3759 	};
3760 	struct nfs4_readlink_res res;
3761 	struct rpc_message msg = {
3762 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3763 		.rpc_argp = &args,
3764 		.rpc_resp = &res,
3765 	};
3766 
3767 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3768 }
3769 
3770 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3771 		unsigned int pgbase, unsigned int pglen)
3772 {
3773 	struct nfs4_exception exception = { };
3774 	int err;
3775 	do {
3776 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3777 		trace_nfs4_readlink(inode, err);
3778 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3779 				&exception);
3780 	} while (exception.retry);
3781 	return err;
3782 }
3783 
3784 /*
3785  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3786  */
3787 static int
3788 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3789 		 int flags)
3790 {
3791 	struct nfs4_label l, *ilabel = NULL;
3792 	struct nfs_open_context *ctx;
3793 	struct nfs4_state *state;
3794 	int status = 0;
3795 
3796 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3797 	if (IS_ERR(ctx))
3798 		return PTR_ERR(ctx);
3799 
3800 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3801 
3802 	sattr->ia_mode &= ~current_umask();
3803 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3804 	if (IS_ERR(state)) {
3805 		status = PTR_ERR(state);
3806 		goto out;
3807 	}
3808 out:
3809 	nfs4_label_release_security(ilabel);
3810 	put_nfs_open_context(ctx);
3811 	return status;
3812 }
3813 
3814 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3815 {
3816 	struct nfs_server *server = NFS_SERVER(dir);
3817 	struct nfs_removeargs args = {
3818 		.fh = NFS_FH(dir),
3819 		.name = *name,
3820 	};
3821 	struct nfs_removeres res = {
3822 		.server = server,
3823 	};
3824 	struct rpc_message msg = {
3825 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3826 		.rpc_argp = &args,
3827 		.rpc_resp = &res,
3828 	};
3829 	int status;
3830 
3831 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3832 	if (status == 0)
3833 		update_changeattr(dir, &res.cinfo);
3834 	return status;
3835 }
3836 
3837 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3838 {
3839 	struct nfs4_exception exception = { };
3840 	int err;
3841 	do {
3842 		err = _nfs4_proc_remove(dir, name);
3843 		trace_nfs4_remove(dir, name, err);
3844 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3845 				&exception);
3846 	} while (exception.retry);
3847 	return err;
3848 }
3849 
3850 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3851 {
3852 	struct nfs_server *server = NFS_SERVER(dir);
3853 	struct nfs_removeargs *args = msg->rpc_argp;
3854 	struct nfs_removeres *res = msg->rpc_resp;
3855 
3856 	res->server = server;
3857 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3858 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3859 
3860 	nfs_fattr_init(res->dir_attr);
3861 }
3862 
3863 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3864 {
3865 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
3866 			&data->args.seq_args,
3867 			&data->res.seq_res,
3868 			task);
3869 }
3870 
3871 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3872 {
3873 	struct nfs_unlinkdata *data = task->tk_calldata;
3874 	struct nfs_removeres *res = &data->res;
3875 
3876 	if (!nfs4_sequence_done(task, &res->seq_res))
3877 		return 0;
3878 	if (nfs4_async_handle_error(task, res->server, NULL,
3879 				    &data->timeout) == -EAGAIN)
3880 		return 0;
3881 	update_changeattr(dir, &res->cinfo);
3882 	return 1;
3883 }
3884 
3885 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3886 {
3887 	struct nfs_server *server = NFS_SERVER(dir);
3888 	struct nfs_renameargs *arg = msg->rpc_argp;
3889 	struct nfs_renameres *res = msg->rpc_resp;
3890 
3891 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3892 	res->server = server;
3893 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3894 }
3895 
3896 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3897 {
3898 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3899 			&data->args.seq_args,
3900 			&data->res.seq_res,
3901 			task);
3902 }
3903 
3904 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3905 				 struct inode *new_dir)
3906 {
3907 	struct nfs_renamedata *data = task->tk_calldata;
3908 	struct nfs_renameres *res = &data->res;
3909 
3910 	if (!nfs4_sequence_done(task, &res->seq_res))
3911 		return 0;
3912 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3913 		return 0;
3914 
3915 	update_changeattr(old_dir, &res->old_cinfo);
3916 	update_changeattr(new_dir, &res->new_cinfo);
3917 	return 1;
3918 }
3919 
3920 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3921 {
3922 	struct nfs_server *server = NFS_SERVER(inode);
3923 	struct nfs4_link_arg arg = {
3924 		.fh     = NFS_FH(inode),
3925 		.dir_fh = NFS_FH(dir),
3926 		.name   = name,
3927 		.bitmask = server->attr_bitmask,
3928 	};
3929 	struct nfs4_link_res res = {
3930 		.server = server,
3931 		.label = NULL,
3932 	};
3933 	struct rpc_message msg = {
3934 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3935 		.rpc_argp = &arg,
3936 		.rpc_resp = &res,
3937 	};
3938 	int status = -ENOMEM;
3939 
3940 	res.fattr = nfs_alloc_fattr();
3941 	if (res.fattr == NULL)
3942 		goto out;
3943 
3944 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3945 	if (IS_ERR(res.label)) {
3946 		status = PTR_ERR(res.label);
3947 		goto out;
3948 	}
3949 	arg.bitmask = nfs4_bitmask(server, res.label);
3950 
3951 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3952 	if (!status) {
3953 		update_changeattr(dir, &res.cinfo);
3954 		status = nfs_post_op_update_inode(inode, res.fattr);
3955 		if (!status)
3956 			nfs_setsecurity(inode, res.fattr, res.label);
3957 	}
3958 
3959 
3960 	nfs4_label_free(res.label);
3961 
3962 out:
3963 	nfs_free_fattr(res.fattr);
3964 	return status;
3965 }
3966 
3967 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
3968 {
3969 	struct nfs4_exception exception = { };
3970 	int err;
3971 	do {
3972 		err = nfs4_handle_exception(NFS_SERVER(inode),
3973 				_nfs4_proc_link(inode, dir, name),
3974 				&exception);
3975 	} while (exception.retry);
3976 	return err;
3977 }
3978 
3979 struct nfs4_createdata {
3980 	struct rpc_message msg;
3981 	struct nfs4_create_arg arg;
3982 	struct nfs4_create_res res;
3983 	struct nfs_fh fh;
3984 	struct nfs_fattr fattr;
3985 	struct nfs4_label *label;
3986 };
3987 
3988 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3989 		const struct qstr *name, struct iattr *sattr, u32 ftype)
3990 {
3991 	struct nfs4_createdata *data;
3992 
3993 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3994 	if (data != NULL) {
3995 		struct nfs_server *server = NFS_SERVER(dir);
3996 
3997 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3998 		if (IS_ERR(data->label))
3999 			goto out_free;
4000 
4001 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4002 		data->msg.rpc_argp = &data->arg;
4003 		data->msg.rpc_resp = &data->res;
4004 		data->arg.dir_fh = NFS_FH(dir);
4005 		data->arg.server = server;
4006 		data->arg.name = name;
4007 		data->arg.attrs = sattr;
4008 		data->arg.ftype = ftype;
4009 		data->arg.bitmask = nfs4_bitmask(server, data->label);
4010 		data->res.server = server;
4011 		data->res.fh = &data->fh;
4012 		data->res.fattr = &data->fattr;
4013 		data->res.label = data->label;
4014 		nfs_fattr_init(data->res.fattr);
4015 	}
4016 	return data;
4017 out_free:
4018 	kfree(data);
4019 	return NULL;
4020 }
4021 
4022 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4023 {
4024 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4025 				    &data->arg.seq_args, &data->res.seq_res, 1);
4026 	if (status == 0) {
4027 		update_changeattr(dir, &data->res.dir_cinfo);
4028 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4029 	}
4030 	return status;
4031 }
4032 
4033 static void nfs4_free_createdata(struct nfs4_createdata *data)
4034 {
4035 	nfs4_label_free(data->label);
4036 	kfree(data);
4037 }
4038 
4039 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4040 		struct page *page, unsigned int len, struct iattr *sattr,
4041 		struct nfs4_label *label)
4042 {
4043 	struct nfs4_createdata *data;
4044 	int status = -ENAMETOOLONG;
4045 
4046 	if (len > NFS4_MAXPATHLEN)
4047 		goto out;
4048 
4049 	status = -ENOMEM;
4050 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4051 	if (data == NULL)
4052 		goto out;
4053 
4054 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4055 	data->arg.u.symlink.pages = &page;
4056 	data->arg.u.symlink.len = len;
4057 	data->arg.label = label;
4058 
4059 	status = nfs4_do_create(dir, dentry, data);
4060 
4061 	nfs4_free_createdata(data);
4062 out:
4063 	return status;
4064 }
4065 
4066 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4067 		struct page *page, unsigned int len, struct iattr *sattr)
4068 {
4069 	struct nfs4_exception exception = { };
4070 	struct nfs4_label l, *label = NULL;
4071 	int err;
4072 
4073 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4074 
4075 	do {
4076 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4077 		trace_nfs4_symlink(dir, &dentry->d_name, err);
4078 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4079 				&exception);
4080 	} while (exception.retry);
4081 
4082 	nfs4_label_release_security(label);
4083 	return err;
4084 }
4085 
4086 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4087 		struct iattr *sattr, struct nfs4_label *label)
4088 {
4089 	struct nfs4_createdata *data;
4090 	int status = -ENOMEM;
4091 
4092 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4093 	if (data == NULL)
4094 		goto out;
4095 
4096 	data->arg.label = label;
4097 	status = nfs4_do_create(dir, dentry, data);
4098 
4099 	nfs4_free_createdata(data);
4100 out:
4101 	return status;
4102 }
4103 
4104 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4105 		struct iattr *sattr)
4106 {
4107 	struct nfs4_exception exception = { };
4108 	struct nfs4_label l, *label = NULL;
4109 	int err;
4110 
4111 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4112 
4113 	sattr->ia_mode &= ~current_umask();
4114 	do {
4115 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4116 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
4117 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4118 				&exception);
4119 	} while (exception.retry);
4120 	nfs4_label_release_security(label);
4121 
4122 	return err;
4123 }
4124 
4125 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4126 		u64 cookie, struct page **pages, unsigned int count, int plus)
4127 {
4128 	struct inode		*dir = d_inode(dentry);
4129 	struct nfs4_readdir_arg args = {
4130 		.fh = NFS_FH(dir),
4131 		.pages = pages,
4132 		.pgbase = 0,
4133 		.count = count,
4134 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4135 		.plus = plus,
4136 	};
4137 	struct nfs4_readdir_res res;
4138 	struct rpc_message msg = {
4139 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4140 		.rpc_argp = &args,
4141 		.rpc_resp = &res,
4142 		.rpc_cred = cred,
4143 	};
4144 	int			status;
4145 
4146 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4147 			dentry,
4148 			(unsigned long long)cookie);
4149 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4150 	res.pgbase = args.pgbase;
4151 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4152 	if (status >= 0) {
4153 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4154 		status += args.pgbase;
4155 	}
4156 
4157 	nfs_invalidate_atime(dir);
4158 
4159 	dprintk("%s: returns %d\n", __func__, status);
4160 	return status;
4161 }
4162 
4163 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4164 		u64 cookie, struct page **pages, unsigned int count, int plus)
4165 {
4166 	struct nfs4_exception exception = { };
4167 	int err;
4168 	do {
4169 		err = _nfs4_proc_readdir(dentry, cred, cookie,
4170 				pages, count, plus);
4171 		trace_nfs4_readdir(d_inode(dentry), err);
4172 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4173 				&exception);
4174 	} while (exception.retry);
4175 	return err;
4176 }
4177 
4178 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4179 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4180 {
4181 	struct nfs4_createdata *data;
4182 	int mode = sattr->ia_mode;
4183 	int status = -ENOMEM;
4184 
4185 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4186 	if (data == NULL)
4187 		goto out;
4188 
4189 	if (S_ISFIFO(mode))
4190 		data->arg.ftype = NF4FIFO;
4191 	else if (S_ISBLK(mode)) {
4192 		data->arg.ftype = NF4BLK;
4193 		data->arg.u.device.specdata1 = MAJOR(rdev);
4194 		data->arg.u.device.specdata2 = MINOR(rdev);
4195 	}
4196 	else if (S_ISCHR(mode)) {
4197 		data->arg.ftype = NF4CHR;
4198 		data->arg.u.device.specdata1 = MAJOR(rdev);
4199 		data->arg.u.device.specdata2 = MINOR(rdev);
4200 	} else if (!S_ISSOCK(mode)) {
4201 		status = -EINVAL;
4202 		goto out_free;
4203 	}
4204 
4205 	data->arg.label = label;
4206 	status = nfs4_do_create(dir, dentry, data);
4207 out_free:
4208 	nfs4_free_createdata(data);
4209 out:
4210 	return status;
4211 }
4212 
4213 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4214 		struct iattr *sattr, dev_t rdev)
4215 {
4216 	struct nfs4_exception exception = { };
4217 	struct nfs4_label l, *label = NULL;
4218 	int err;
4219 
4220 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4221 
4222 	sattr->ia_mode &= ~current_umask();
4223 	do {
4224 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4225 		trace_nfs4_mknod(dir, &dentry->d_name, err);
4226 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4227 				&exception);
4228 	} while (exception.retry);
4229 
4230 	nfs4_label_release_security(label);
4231 
4232 	return err;
4233 }
4234 
4235 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4236 		 struct nfs_fsstat *fsstat)
4237 {
4238 	struct nfs4_statfs_arg args = {
4239 		.fh = fhandle,
4240 		.bitmask = server->attr_bitmask,
4241 	};
4242 	struct nfs4_statfs_res res = {
4243 		.fsstat = fsstat,
4244 	};
4245 	struct rpc_message msg = {
4246 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4247 		.rpc_argp = &args,
4248 		.rpc_resp = &res,
4249 	};
4250 
4251 	nfs_fattr_init(fsstat->fattr);
4252 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4253 }
4254 
4255 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4256 {
4257 	struct nfs4_exception exception = { };
4258 	int err;
4259 	do {
4260 		err = nfs4_handle_exception(server,
4261 				_nfs4_proc_statfs(server, fhandle, fsstat),
4262 				&exception);
4263 	} while (exception.retry);
4264 	return err;
4265 }
4266 
4267 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4268 		struct nfs_fsinfo *fsinfo)
4269 {
4270 	struct nfs4_fsinfo_arg args = {
4271 		.fh = fhandle,
4272 		.bitmask = server->attr_bitmask,
4273 	};
4274 	struct nfs4_fsinfo_res res = {
4275 		.fsinfo = fsinfo,
4276 	};
4277 	struct rpc_message msg = {
4278 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4279 		.rpc_argp = &args,
4280 		.rpc_resp = &res,
4281 	};
4282 
4283 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4284 }
4285 
4286 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4287 {
4288 	struct nfs4_exception exception = { };
4289 	unsigned long now = jiffies;
4290 	int err;
4291 
4292 	do {
4293 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4294 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4295 		if (err == 0) {
4296 			nfs4_set_lease_period(server->nfs_client,
4297 					fsinfo->lease_time * HZ,
4298 					now);
4299 			break;
4300 		}
4301 		err = nfs4_handle_exception(server, err, &exception);
4302 	} while (exception.retry);
4303 	return err;
4304 }
4305 
4306 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4307 {
4308 	int error;
4309 
4310 	nfs_fattr_init(fsinfo->fattr);
4311 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4312 	if (error == 0) {
4313 		/* block layout checks this! */
4314 		server->pnfs_blksize = fsinfo->blksize;
4315 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4316 	}
4317 
4318 	return error;
4319 }
4320 
4321 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4322 		struct nfs_pathconf *pathconf)
4323 {
4324 	struct nfs4_pathconf_arg args = {
4325 		.fh = fhandle,
4326 		.bitmask = server->attr_bitmask,
4327 	};
4328 	struct nfs4_pathconf_res res = {
4329 		.pathconf = pathconf,
4330 	};
4331 	struct rpc_message msg = {
4332 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4333 		.rpc_argp = &args,
4334 		.rpc_resp = &res,
4335 	};
4336 
4337 	/* None of the pathconf attributes are mandatory to implement */
4338 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4339 		memset(pathconf, 0, sizeof(*pathconf));
4340 		return 0;
4341 	}
4342 
4343 	nfs_fattr_init(pathconf->fattr);
4344 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4345 }
4346 
4347 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4348 		struct nfs_pathconf *pathconf)
4349 {
4350 	struct nfs4_exception exception = { };
4351 	int err;
4352 
4353 	do {
4354 		err = nfs4_handle_exception(server,
4355 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4356 				&exception);
4357 	} while (exception.retry);
4358 	return err;
4359 }
4360 
4361 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4362 		const struct nfs_open_context *ctx,
4363 		const struct nfs_lock_context *l_ctx,
4364 		fmode_t fmode)
4365 {
4366 	const struct nfs_lockowner *lockowner = NULL;
4367 
4368 	if (l_ctx != NULL)
4369 		lockowner = &l_ctx->lockowner;
4370 	return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4371 }
4372 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4373 
4374 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4375 		const struct nfs_open_context *ctx,
4376 		const struct nfs_lock_context *l_ctx,
4377 		fmode_t fmode)
4378 {
4379 	nfs4_stateid current_stateid;
4380 
4381 	/* If the current stateid represents a lost lock, then exit */
4382 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4383 		return true;
4384 	return nfs4_stateid_match(stateid, &current_stateid);
4385 }
4386 
4387 static bool nfs4_error_stateid_expired(int err)
4388 {
4389 	switch (err) {
4390 	case -NFS4ERR_DELEG_REVOKED:
4391 	case -NFS4ERR_ADMIN_REVOKED:
4392 	case -NFS4ERR_BAD_STATEID:
4393 	case -NFS4ERR_STALE_STATEID:
4394 	case -NFS4ERR_OLD_STATEID:
4395 	case -NFS4ERR_OPENMODE:
4396 	case -NFS4ERR_EXPIRED:
4397 		return true;
4398 	}
4399 	return false;
4400 }
4401 
4402 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4403 {
4404 	nfs_invalidate_atime(hdr->inode);
4405 }
4406 
4407 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4408 {
4409 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4410 
4411 	trace_nfs4_read(hdr, task->tk_status);
4412 	if (nfs4_async_handle_error(task, server,
4413 				    hdr->args.context->state,
4414 				    NULL) == -EAGAIN) {
4415 		rpc_restart_call_prepare(task);
4416 		return -EAGAIN;
4417 	}
4418 
4419 	__nfs4_read_done_cb(hdr);
4420 	if (task->tk_status > 0)
4421 		renew_lease(server, hdr->timestamp);
4422 	return 0;
4423 }
4424 
4425 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4426 		struct nfs_pgio_args *args)
4427 {
4428 
4429 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4430 		nfs4_stateid_is_current(&args->stateid,
4431 				args->context,
4432 				args->lock_context,
4433 				FMODE_READ))
4434 		return false;
4435 	rpc_restart_call_prepare(task);
4436 	return true;
4437 }
4438 
4439 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4440 {
4441 
4442 	dprintk("--> %s\n", __func__);
4443 
4444 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4445 		return -EAGAIN;
4446 	if (nfs4_read_stateid_changed(task, &hdr->args))
4447 		return -EAGAIN;
4448 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4449 				    nfs4_read_done_cb(task, hdr);
4450 }
4451 
4452 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4453 				 struct rpc_message *msg)
4454 {
4455 	hdr->timestamp   = jiffies;
4456 	if (!hdr->pgio_done_cb)
4457 		hdr->pgio_done_cb = nfs4_read_done_cb;
4458 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4459 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4460 }
4461 
4462 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4463 				      struct nfs_pgio_header *hdr)
4464 {
4465 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4466 			&hdr->args.seq_args,
4467 			&hdr->res.seq_res,
4468 			task))
4469 		return 0;
4470 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4471 				hdr->args.lock_context,
4472 				hdr->rw_ops->rw_mode) == -EIO)
4473 		return -EIO;
4474 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4475 		return -EIO;
4476 	return 0;
4477 }
4478 
4479 static int nfs4_write_done_cb(struct rpc_task *task,
4480 			      struct nfs_pgio_header *hdr)
4481 {
4482 	struct inode *inode = hdr->inode;
4483 
4484 	trace_nfs4_write(hdr, task->tk_status);
4485 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4486 				    hdr->args.context->state,
4487 				    NULL) == -EAGAIN) {
4488 		rpc_restart_call_prepare(task);
4489 		return -EAGAIN;
4490 	}
4491 	if (task->tk_status >= 0) {
4492 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4493 		nfs_writeback_update_inode(hdr);
4494 	}
4495 	return 0;
4496 }
4497 
4498 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4499 		struct nfs_pgio_args *args)
4500 {
4501 
4502 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4503 		nfs4_stateid_is_current(&args->stateid,
4504 				args->context,
4505 				args->lock_context,
4506 				FMODE_WRITE))
4507 		return false;
4508 	rpc_restart_call_prepare(task);
4509 	return true;
4510 }
4511 
4512 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4513 {
4514 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4515 		return -EAGAIN;
4516 	if (nfs4_write_stateid_changed(task, &hdr->args))
4517 		return -EAGAIN;
4518 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4519 		nfs4_write_done_cb(task, hdr);
4520 }
4521 
4522 static
4523 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4524 {
4525 	/* Don't request attributes for pNFS or O_DIRECT writes */
4526 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4527 		return false;
4528 	/* Otherwise, request attributes if and only if we don't hold
4529 	 * a delegation
4530 	 */
4531 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4532 }
4533 
4534 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4535 				  struct rpc_message *msg)
4536 {
4537 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4538 
4539 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4540 		hdr->args.bitmask = NULL;
4541 		hdr->res.fattr = NULL;
4542 	} else
4543 		hdr->args.bitmask = server->cache_consistency_bitmask;
4544 
4545 	if (!hdr->pgio_done_cb)
4546 		hdr->pgio_done_cb = nfs4_write_done_cb;
4547 	hdr->res.server = server;
4548 	hdr->timestamp   = jiffies;
4549 
4550 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4551 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4552 }
4553 
4554 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4555 {
4556 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4557 			&data->args.seq_args,
4558 			&data->res.seq_res,
4559 			task);
4560 }
4561 
4562 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4563 {
4564 	struct inode *inode = data->inode;
4565 
4566 	trace_nfs4_commit(data, task->tk_status);
4567 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4568 				    NULL, NULL) == -EAGAIN) {
4569 		rpc_restart_call_prepare(task);
4570 		return -EAGAIN;
4571 	}
4572 	return 0;
4573 }
4574 
4575 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4576 {
4577 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4578 		return -EAGAIN;
4579 	return data->commit_done_cb(task, data);
4580 }
4581 
4582 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4583 {
4584 	struct nfs_server *server = NFS_SERVER(data->inode);
4585 
4586 	if (data->commit_done_cb == NULL)
4587 		data->commit_done_cb = nfs4_commit_done_cb;
4588 	data->res.server = server;
4589 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4590 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4591 }
4592 
4593 struct nfs4_renewdata {
4594 	struct nfs_client	*client;
4595 	unsigned long		timestamp;
4596 };
4597 
4598 /*
4599  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4600  * standalone procedure for queueing an asynchronous RENEW.
4601  */
4602 static void nfs4_renew_release(void *calldata)
4603 {
4604 	struct nfs4_renewdata *data = calldata;
4605 	struct nfs_client *clp = data->client;
4606 
4607 	if (atomic_read(&clp->cl_count) > 1)
4608 		nfs4_schedule_state_renewal(clp);
4609 	nfs_put_client(clp);
4610 	kfree(data);
4611 }
4612 
4613 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4614 {
4615 	struct nfs4_renewdata *data = calldata;
4616 	struct nfs_client *clp = data->client;
4617 	unsigned long timestamp = data->timestamp;
4618 
4619 	trace_nfs4_renew_async(clp, task->tk_status);
4620 	switch (task->tk_status) {
4621 	case 0:
4622 		break;
4623 	case -NFS4ERR_LEASE_MOVED:
4624 		nfs4_schedule_lease_moved_recovery(clp);
4625 		break;
4626 	default:
4627 		/* Unless we're shutting down, schedule state recovery! */
4628 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4629 			return;
4630 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4631 			nfs4_schedule_lease_recovery(clp);
4632 			return;
4633 		}
4634 		nfs4_schedule_path_down_recovery(clp);
4635 	}
4636 	do_renew_lease(clp, timestamp);
4637 }
4638 
4639 static const struct rpc_call_ops nfs4_renew_ops = {
4640 	.rpc_call_done = nfs4_renew_done,
4641 	.rpc_release = nfs4_renew_release,
4642 };
4643 
4644 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4645 {
4646 	struct rpc_message msg = {
4647 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4648 		.rpc_argp	= clp,
4649 		.rpc_cred	= cred,
4650 	};
4651 	struct nfs4_renewdata *data;
4652 
4653 	if (renew_flags == 0)
4654 		return 0;
4655 	if (!atomic_inc_not_zero(&clp->cl_count))
4656 		return -EIO;
4657 	data = kmalloc(sizeof(*data), GFP_NOFS);
4658 	if (data == NULL)
4659 		return -ENOMEM;
4660 	data->client = clp;
4661 	data->timestamp = jiffies;
4662 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4663 			&nfs4_renew_ops, data);
4664 }
4665 
4666 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4667 {
4668 	struct rpc_message msg = {
4669 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4670 		.rpc_argp	= clp,
4671 		.rpc_cred	= cred,
4672 	};
4673 	unsigned long now = jiffies;
4674 	int status;
4675 
4676 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4677 	if (status < 0)
4678 		return status;
4679 	do_renew_lease(clp, now);
4680 	return 0;
4681 }
4682 
4683 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4684 {
4685 	return server->caps & NFS_CAP_ACLS;
4686 }
4687 
4688 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4689  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4690  * the stack.
4691  */
4692 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4693 
4694 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4695 		struct page **pages)
4696 {
4697 	struct page *newpage, **spages;
4698 	int rc = 0;
4699 	size_t len;
4700 	spages = pages;
4701 
4702 	do {
4703 		len = min_t(size_t, PAGE_SIZE, buflen);
4704 		newpage = alloc_page(GFP_KERNEL);
4705 
4706 		if (newpage == NULL)
4707 			goto unwind;
4708 		memcpy(page_address(newpage), buf, len);
4709                 buf += len;
4710                 buflen -= len;
4711 		*pages++ = newpage;
4712 		rc++;
4713 	} while (buflen != 0);
4714 
4715 	return rc;
4716 
4717 unwind:
4718 	for(; rc > 0; rc--)
4719 		__free_page(spages[rc-1]);
4720 	return -ENOMEM;
4721 }
4722 
4723 struct nfs4_cached_acl {
4724 	int cached;
4725 	size_t len;
4726 	char data[0];
4727 };
4728 
4729 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4730 {
4731 	struct nfs_inode *nfsi = NFS_I(inode);
4732 
4733 	spin_lock(&inode->i_lock);
4734 	kfree(nfsi->nfs4_acl);
4735 	nfsi->nfs4_acl = acl;
4736 	spin_unlock(&inode->i_lock);
4737 }
4738 
4739 static void nfs4_zap_acl_attr(struct inode *inode)
4740 {
4741 	nfs4_set_cached_acl(inode, NULL);
4742 }
4743 
4744 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4745 {
4746 	struct nfs_inode *nfsi = NFS_I(inode);
4747 	struct nfs4_cached_acl *acl;
4748 	int ret = -ENOENT;
4749 
4750 	spin_lock(&inode->i_lock);
4751 	acl = nfsi->nfs4_acl;
4752 	if (acl == NULL)
4753 		goto out;
4754 	if (buf == NULL) /* user is just asking for length */
4755 		goto out_len;
4756 	if (acl->cached == 0)
4757 		goto out;
4758 	ret = -ERANGE; /* see getxattr(2) man page */
4759 	if (acl->len > buflen)
4760 		goto out;
4761 	memcpy(buf, acl->data, acl->len);
4762 out_len:
4763 	ret = acl->len;
4764 out:
4765 	spin_unlock(&inode->i_lock);
4766 	return ret;
4767 }
4768 
4769 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4770 {
4771 	struct nfs4_cached_acl *acl;
4772 	size_t buflen = sizeof(*acl) + acl_len;
4773 
4774 	if (buflen <= PAGE_SIZE) {
4775 		acl = kmalloc(buflen, GFP_KERNEL);
4776 		if (acl == NULL)
4777 			goto out;
4778 		acl->cached = 1;
4779 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4780 	} else {
4781 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4782 		if (acl == NULL)
4783 			goto out;
4784 		acl->cached = 0;
4785 	}
4786 	acl->len = acl_len;
4787 out:
4788 	nfs4_set_cached_acl(inode, acl);
4789 }
4790 
4791 /*
4792  * The getxattr API returns the required buffer length when called with a
4793  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4794  * the required buf.  On a NULL buf, we send a page of data to the server
4795  * guessing that the ACL request can be serviced by a page. If so, we cache
4796  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4797  * the cache. If not so, we throw away the page, and cache the required
4798  * length. The next getxattr call will then produce another round trip to
4799  * the server, this time with the input buf of the required size.
4800  */
4801 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4802 {
4803 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4804 	struct nfs_getaclargs args = {
4805 		.fh = NFS_FH(inode),
4806 		.acl_pages = pages,
4807 		.acl_len = buflen,
4808 	};
4809 	struct nfs_getaclres res = {
4810 		.acl_len = buflen,
4811 	};
4812 	struct rpc_message msg = {
4813 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4814 		.rpc_argp = &args,
4815 		.rpc_resp = &res,
4816 	};
4817 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4818 	int ret = -ENOMEM, i;
4819 
4820 	/* As long as we're doing a round trip to the server anyway,
4821 	 * let's be prepared for a page of acl data. */
4822 	if (npages == 0)
4823 		npages = 1;
4824 	if (npages > ARRAY_SIZE(pages))
4825 		return -ERANGE;
4826 
4827 	for (i = 0; i < npages; i++) {
4828 		pages[i] = alloc_page(GFP_KERNEL);
4829 		if (!pages[i])
4830 			goto out_free;
4831 	}
4832 
4833 	/* for decoding across pages */
4834 	res.acl_scratch = alloc_page(GFP_KERNEL);
4835 	if (!res.acl_scratch)
4836 		goto out_free;
4837 
4838 	args.acl_len = npages * PAGE_SIZE;
4839 
4840 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4841 		__func__, buf, buflen, npages, args.acl_len);
4842 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4843 			     &msg, &args.seq_args, &res.seq_res, 0);
4844 	if (ret)
4845 		goto out_free;
4846 
4847 	/* Handle the case where the passed-in buffer is too short */
4848 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4849 		/* Did the user only issue a request for the acl length? */
4850 		if (buf == NULL)
4851 			goto out_ok;
4852 		ret = -ERANGE;
4853 		goto out_free;
4854 	}
4855 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4856 	if (buf) {
4857 		if (res.acl_len > buflen) {
4858 			ret = -ERANGE;
4859 			goto out_free;
4860 		}
4861 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4862 	}
4863 out_ok:
4864 	ret = res.acl_len;
4865 out_free:
4866 	for (i = 0; i < npages; i++)
4867 		if (pages[i])
4868 			__free_page(pages[i]);
4869 	if (res.acl_scratch)
4870 		__free_page(res.acl_scratch);
4871 	return ret;
4872 }
4873 
4874 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4875 {
4876 	struct nfs4_exception exception = { };
4877 	ssize_t ret;
4878 	do {
4879 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4880 		trace_nfs4_get_acl(inode, ret);
4881 		if (ret >= 0)
4882 			break;
4883 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4884 	} while (exception.retry);
4885 	return ret;
4886 }
4887 
4888 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4889 {
4890 	struct nfs_server *server = NFS_SERVER(inode);
4891 	int ret;
4892 
4893 	if (!nfs4_server_supports_acls(server))
4894 		return -EOPNOTSUPP;
4895 	ret = nfs_revalidate_inode(server, inode);
4896 	if (ret < 0)
4897 		return ret;
4898 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4899 		nfs_zap_acl_cache(inode);
4900 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4901 	if (ret != -ENOENT)
4902 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4903 		 * but no cached acl data, just the acl length */
4904 		return ret;
4905 	return nfs4_get_acl_uncached(inode, buf, buflen);
4906 }
4907 
4908 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4909 {
4910 	struct nfs_server *server = NFS_SERVER(inode);
4911 	struct page *pages[NFS4ACL_MAXPAGES];
4912 	struct nfs_setaclargs arg = {
4913 		.fh		= NFS_FH(inode),
4914 		.acl_pages	= pages,
4915 		.acl_len	= buflen,
4916 	};
4917 	struct nfs_setaclres res;
4918 	struct rpc_message msg = {
4919 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4920 		.rpc_argp	= &arg,
4921 		.rpc_resp	= &res,
4922 	};
4923 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4924 	int ret, i;
4925 
4926 	if (!nfs4_server_supports_acls(server))
4927 		return -EOPNOTSUPP;
4928 	if (npages > ARRAY_SIZE(pages))
4929 		return -ERANGE;
4930 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4931 	if (i < 0)
4932 		return i;
4933 	nfs4_inode_return_delegation(inode);
4934 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4935 
4936 	/*
4937 	 * Free each page after tx, so the only ref left is
4938 	 * held by the network stack
4939 	 */
4940 	for (; i > 0; i--)
4941 		put_page(pages[i-1]);
4942 
4943 	/*
4944 	 * Acl update can result in inode attribute update.
4945 	 * so mark the attribute cache invalid.
4946 	 */
4947 	spin_lock(&inode->i_lock);
4948 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4949 	spin_unlock(&inode->i_lock);
4950 	nfs_access_zap_cache(inode);
4951 	nfs_zap_acl_cache(inode);
4952 	return ret;
4953 }
4954 
4955 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4956 {
4957 	struct nfs4_exception exception = { };
4958 	int err;
4959 	do {
4960 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4961 		trace_nfs4_set_acl(inode, err);
4962 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4963 				&exception);
4964 	} while (exception.retry);
4965 	return err;
4966 }
4967 
4968 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4969 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4970 					size_t buflen)
4971 {
4972 	struct nfs_server *server = NFS_SERVER(inode);
4973 	struct nfs_fattr fattr;
4974 	struct nfs4_label label = {0, 0, buflen, buf};
4975 
4976 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4977 	struct nfs4_getattr_arg arg = {
4978 		.fh		= NFS_FH(inode),
4979 		.bitmask	= bitmask,
4980 	};
4981 	struct nfs4_getattr_res res = {
4982 		.fattr		= &fattr,
4983 		.label		= &label,
4984 		.server		= server,
4985 	};
4986 	struct rpc_message msg = {
4987 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4988 		.rpc_argp	= &arg,
4989 		.rpc_resp	= &res,
4990 	};
4991 	int ret;
4992 
4993 	nfs_fattr_init(&fattr);
4994 
4995 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4996 	if (ret)
4997 		return ret;
4998 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4999 		return -ENOENT;
5000 	if (buflen < label.len)
5001 		return -ERANGE;
5002 	return 0;
5003 }
5004 
5005 static int nfs4_get_security_label(struct inode *inode, void *buf,
5006 					size_t buflen)
5007 {
5008 	struct nfs4_exception exception = { };
5009 	int err;
5010 
5011 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5012 		return -EOPNOTSUPP;
5013 
5014 	do {
5015 		err = _nfs4_get_security_label(inode, buf, buflen);
5016 		trace_nfs4_get_security_label(inode, err);
5017 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5018 				&exception);
5019 	} while (exception.retry);
5020 	return err;
5021 }
5022 
5023 static int _nfs4_do_set_security_label(struct inode *inode,
5024 		struct nfs4_label *ilabel,
5025 		struct nfs_fattr *fattr,
5026 		struct nfs4_label *olabel)
5027 {
5028 
5029 	struct iattr sattr = {0};
5030 	struct nfs_server *server = NFS_SERVER(inode);
5031 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5032 	struct nfs_setattrargs arg = {
5033 		.fh             = NFS_FH(inode),
5034 		.iap            = &sattr,
5035 		.server		= server,
5036 		.bitmask	= bitmask,
5037 		.label		= ilabel,
5038 	};
5039 	struct nfs_setattrres res = {
5040 		.fattr		= fattr,
5041 		.label		= olabel,
5042 		.server		= server,
5043 	};
5044 	struct rpc_message msg = {
5045 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5046 		.rpc_argp       = &arg,
5047 		.rpc_resp       = &res,
5048 	};
5049 	int status;
5050 
5051 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5052 
5053 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5054 	if (status)
5055 		dprintk("%s failed: %d\n", __func__, status);
5056 
5057 	return status;
5058 }
5059 
5060 static int nfs4_do_set_security_label(struct inode *inode,
5061 		struct nfs4_label *ilabel,
5062 		struct nfs_fattr *fattr,
5063 		struct nfs4_label *olabel)
5064 {
5065 	struct nfs4_exception exception = { };
5066 	int err;
5067 
5068 	do {
5069 		err = _nfs4_do_set_security_label(inode, ilabel,
5070 				fattr, olabel);
5071 		trace_nfs4_set_security_label(inode, err);
5072 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5073 				&exception);
5074 	} while (exception.retry);
5075 	return err;
5076 }
5077 
5078 static int
5079 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5080 {
5081 	struct nfs4_label ilabel, *olabel = NULL;
5082 	struct nfs_fattr fattr;
5083 	struct rpc_cred *cred;
5084 	int status;
5085 
5086 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5087 		return -EOPNOTSUPP;
5088 
5089 	nfs_fattr_init(&fattr);
5090 
5091 	ilabel.pi = 0;
5092 	ilabel.lfs = 0;
5093 	ilabel.label = (char *)buf;
5094 	ilabel.len = buflen;
5095 
5096 	cred = rpc_lookup_cred();
5097 	if (IS_ERR(cred))
5098 		return PTR_ERR(cred);
5099 
5100 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5101 	if (IS_ERR(olabel)) {
5102 		status = -PTR_ERR(olabel);
5103 		goto out;
5104 	}
5105 
5106 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5107 	if (status == 0)
5108 		nfs_setsecurity(inode, &fattr, olabel);
5109 
5110 	nfs4_label_free(olabel);
5111 out:
5112 	put_rpccred(cred);
5113 	return status;
5114 }
5115 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
5116 
5117 
5118 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5119 				    nfs4_verifier *bootverf)
5120 {
5121 	__be32 verf[2];
5122 
5123 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5124 		/* An impossible timestamp guarantees this value
5125 		 * will never match a generated boot time. */
5126 		verf[0] = 0;
5127 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5128 	} else {
5129 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5130 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5131 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5132 	}
5133 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
5134 }
5135 
5136 static int
5137 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5138 {
5139 	size_t len;
5140 	char *str;
5141 
5142 	if (clp->cl_owner_id != NULL)
5143 		return 0;
5144 
5145 	rcu_read_lock();
5146 	len = 14 + strlen(clp->cl_ipaddr) + 1 +
5147 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5148 		1 +
5149 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5150 		1;
5151 	rcu_read_unlock();
5152 
5153 	if (len > NFS4_OPAQUE_LIMIT + 1)
5154 		return -EINVAL;
5155 
5156 	/*
5157 	 * Since this string is allocated at mount time, and held until the
5158 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5159 	 * about a memory-reclaim deadlock.
5160 	 */
5161 	str = kmalloc(len, GFP_KERNEL);
5162 	if (!str)
5163 		return -ENOMEM;
5164 
5165 	rcu_read_lock();
5166 	scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5167 			clp->cl_ipaddr,
5168 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5169 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5170 	rcu_read_unlock();
5171 
5172 	clp->cl_owner_id = str;
5173 	return 0;
5174 }
5175 
5176 static int
5177 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5178 {
5179 	size_t len;
5180 	char *str;
5181 
5182 	len = 10 + 10 + 1 + 10 + 1 +
5183 		strlen(nfs4_client_id_uniquifier) + 1 +
5184 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
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 	scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5199 			clp->rpc_ops->version, clp->cl_minorversion,
5200 			nfs4_client_id_uniquifier,
5201 			clp->cl_rpcclient->cl_nodename);
5202 	clp->cl_owner_id = str;
5203 	return 0;
5204 }
5205 
5206 static int
5207 nfs4_init_uniform_client_string(struct nfs_client *clp)
5208 {
5209 	size_t len;
5210 	char *str;
5211 
5212 	if (clp->cl_owner_id != NULL)
5213 		return 0;
5214 
5215 	if (nfs4_client_id_uniquifier[0] != '\0')
5216 		return nfs4_init_uniquifier_client_string(clp);
5217 
5218 	len = 10 + 10 + 1 + 10 + 1 +
5219 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5220 
5221 	if (len > NFS4_OPAQUE_LIMIT + 1)
5222 		return -EINVAL;
5223 
5224 	/*
5225 	 * Since this string is allocated at mount time, and held until the
5226 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5227 	 * about a memory-reclaim deadlock.
5228 	 */
5229 	str = kmalloc(len, GFP_KERNEL);
5230 	if (!str)
5231 		return -ENOMEM;
5232 
5233 	scnprintf(str, len, "Linux NFSv%u.%u %s",
5234 			clp->rpc_ops->version, clp->cl_minorversion,
5235 			clp->cl_rpcclient->cl_nodename);
5236 	clp->cl_owner_id = str;
5237 	return 0;
5238 }
5239 
5240 /*
5241  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5242  * services.  Advertise one based on the address family of the
5243  * clientaddr.
5244  */
5245 static unsigned int
5246 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5247 {
5248 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5249 		return scnprintf(buf, len, "tcp6");
5250 	else
5251 		return scnprintf(buf, len, "tcp");
5252 }
5253 
5254 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5255 {
5256 	struct nfs4_setclientid *sc = calldata;
5257 
5258 	if (task->tk_status == 0)
5259 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5260 }
5261 
5262 static const struct rpc_call_ops nfs4_setclientid_ops = {
5263 	.rpc_call_done = nfs4_setclientid_done,
5264 };
5265 
5266 /**
5267  * nfs4_proc_setclientid - Negotiate client ID
5268  * @clp: state data structure
5269  * @program: RPC program for NFSv4 callback service
5270  * @port: IP port number for NFS4 callback service
5271  * @cred: RPC credential to use for this call
5272  * @res: where to place the result
5273  *
5274  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5275  */
5276 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5277 		unsigned short port, struct rpc_cred *cred,
5278 		struct nfs4_setclientid_res *res)
5279 {
5280 	nfs4_verifier sc_verifier;
5281 	struct nfs4_setclientid setclientid = {
5282 		.sc_verifier = &sc_verifier,
5283 		.sc_prog = program,
5284 		.sc_clnt = clp,
5285 	};
5286 	struct rpc_message msg = {
5287 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5288 		.rpc_argp = &setclientid,
5289 		.rpc_resp = res,
5290 		.rpc_cred = cred,
5291 	};
5292 	struct rpc_task *task;
5293 	struct rpc_task_setup task_setup_data = {
5294 		.rpc_client = clp->cl_rpcclient,
5295 		.rpc_message = &msg,
5296 		.callback_ops = &nfs4_setclientid_ops,
5297 		.callback_data = &setclientid,
5298 		.flags = RPC_TASK_TIMEOUT,
5299 	};
5300 	int status;
5301 
5302 	/* nfs_client_id4 */
5303 	nfs4_init_boot_verifier(clp, &sc_verifier);
5304 
5305 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5306 		status = nfs4_init_uniform_client_string(clp);
5307 	else
5308 		status = nfs4_init_nonuniform_client_string(clp);
5309 
5310 	if (status)
5311 		goto out;
5312 
5313 	/* cb_client4 */
5314 	setclientid.sc_netid_len =
5315 				nfs4_init_callback_netid(clp,
5316 						setclientid.sc_netid,
5317 						sizeof(setclientid.sc_netid));
5318 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5319 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5320 				clp->cl_ipaddr, port >> 8, port & 255);
5321 
5322 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
5323 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5324 		clp->cl_owner_id);
5325 	task = rpc_run_task(&task_setup_data);
5326 	if (IS_ERR(task)) {
5327 		status = PTR_ERR(task);
5328 		goto out;
5329 	}
5330 	status = task->tk_status;
5331 	if (setclientid.sc_cred) {
5332 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5333 		put_rpccred(setclientid.sc_cred);
5334 	}
5335 	rpc_put_task(task);
5336 out:
5337 	trace_nfs4_setclientid(clp, status);
5338 	dprintk("NFS reply setclientid: %d\n", status);
5339 	return status;
5340 }
5341 
5342 /**
5343  * nfs4_proc_setclientid_confirm - Confirm client ID
5344  * @clp: state data structure
5345  * @res: result of a previous SETCLIENTID
5346  * @cred: RPC credential to use for this call
5347  *
5348  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5349  */
5350 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5351 		struct nfs4_setclientid_res *arg,
5352 		struct rpc_cred *cred)
5353 {
5354 	struct rpc_message msg = {
5355 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5356 		.rpc_argp = arg,
5357 		.rpc_cred = cred,
5358 	};
5359 	int status;
5360 
5361 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5362 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5363 		clp->cl_clientid);
5364 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5365 	trace_nfs4_setclientid_confirm(clp, status);
5366 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5367 	return status;
5368 }
5369 
5370 struct nfs4_delegreturndata {
5371 	struct nfs4_delegreturnargs args;
5372 	struct nfs4_delegreturnres res;
5373 	struct nfs_fh fh;
5374 	nfs4_stateid stateid;
5375 	unsigned long timestamp;
5376 	struct nfs_fattr fattr;
5377 	int rpc_status;
5378 	struct inode *inode;
5379 	bool roc;
5380 	u32 roc_barrier;
5381 };
5382 
5383 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5384 {
5385 	struct nfs4_delegreturndata *data = calldata;
5386 
5387 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5388 		return;
5389 
5390 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5391 	switch (task->tk_status) {
5392 	case 0:
5393 		renew_lease(data->res.server, data->timestamp);
5394 	case -NFS4ERR_ADMIN_REVOKED:
5395 	case -NFS4ERR_DELEG_REVOKED:
5396 	case -NFS4ERR_BAD_STATEID:
5397 	case -NFS4ERR_OLD_STATEID:
5398 	case -NFS4ERR_STALE_STATEID:
5399 	case -NFS4ERR_EXPIRED:
5400 		task->tk_status = 0;
5401 		if (data->roc)
5402 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5403 		break;
5404 	default:
5405 		if (nfs4_async_handle_error(task, data->res.server,
5406 					    NULL, NULL) == -EAGAIN) {
5407 			rpc_restart_call_prepare(task);
5408 			return;
5409 		}
5410 	}
5411 	data->rpc_status = task->tk_status;
5412 }
5413 
5414 static void nfs4_delegreturn_release(void *calldata)
5415 {
5416 	struct nfs4_delegreturndata *data = calldata;
5417 	struct inode *inode = data->inode;
5418 
5419 	if (inode) {
5420 		if (data->roc)
5421 			pnfs_roc_release(inode);
5422 		nfs_iput_and_deactive(inode);
5423 	}
5424 	kfree(calldata);
5425 }
5426 
5427 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5428 {
5429 	struct nfs4_delegreturndata *d_data;
5430 
5431 	d_data = (struct nfs4_delegreturndata *)data;
5432 
5433 	if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5434 		return;
5435 
5436 	if (d_data->roc)
5437 		pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5438 
5439 	nfs4_setup_sequence(d_data->res.server,
5440 			&d_data->args.seq_args,
5441 			&d_data->res.seq_res,
5442 			task);
5443 }
5444 
5445 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5446 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5447 	.rpc_call_done = nfs4_delegreturn_done,
5448 	.rpc_release = nfs4_delegreturn_release,
5449 };
5450 
5451 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5452 {
5453 	struct nfs4_delegreturndata *data;
5454 	struct nfs_server *server = NFS_SERVER(inode);
5455 	struct rpc_task *task;
5456 	struct rpc_message msg = {
5457 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5458 		.rpc_cred = cred,
5459 	};
5460 	struct rpc_task_setup task_setup_data = {
5461 		.rpc_client = server->client,
5462 		.rpc_message = &msg,
5463 		.callback_ops = &nfs4_delegreturn_ops,
5464 		.flags = RPC_TASK_ASYNC,
5465 	};
5466 	int status = 0;
5467 
5468 	data = kzalloc(sizeof(*data), GFP_NOFS);
5469 	if (data == NULL)
5470 		return -ENOMEM;
5471 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5472 
5473 	nfs4_state_protect(server->nfs_client,
5474 			NFS_SP4_MACH_CRED_CLEANUP,
5475 			&task_setup_data.rpc_client, &msg);
5476 
5477 	data->args.fhandle = &data->fh;
5478 	data->args.stateid = &data->stateid;
5479 	data->args.bitmask = server->cache_consistency_bitmask;
5480 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5481 	nfs4_stateid_copy(&data->stateid, stateid);
5482 	data->res.fattr = &data->fattr;
5483 	data->res.server = server;
5484 	nfs_fattr_init(data->res.fattr);
5485 	data->timestamp = jiffies;
5486 	data->rpc_status = 0;
5487 	data->inode = nfs_igrab_and_active(inode);
5488 	if (data->inode)
5489 		data->roc = nfs4_roc(inode);
5490 
5491 	task_setup_data.callback_data = data;
5492 	msg.rpc_argp = &data->args;
5493 	msg.rpc_resp = &data->res;
5494 	task = rpc_run_task(&task_setup_data);
5495 	if (IS_ERR(task))
5496 		return PTR_ERR(task);
5497 	if (!issync)
5498 		goto out;
5499 	status = nfs4_wait_for_completion_rpc_task(task);
5500 	if (status != 0)
5501 		goto out;
5502 	status = data->rpc_status;
5503 	if (status == 0)
5504 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5505 	else
5506 		nfs_refresh_inode(inode, &data->fattr);
5507 out:
5508 	rpc_put_task(task);
5509 	return status;
5510 }
5511 
5512 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5513 {
5514 	struct nfs_server *server = NFS_SERVER(inode);
5515 	struct nfs4_exception exception = { };
5516 	int err;
5517 	do {
5518 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5519 		trace_nfs4_delegreturn(inode, stateid, err);
5520 		switch (err) {
5521 			case -NFS4ERR_STALE_STATEID:
5522 			case -NFS4ERR_EXPIRED:
5523 			case 0:
5524 				return 0;
5525 		}
5526 		err = nfs4_handle_exception(server, err, &exception);
5527 	} while (exception.retry);
5528 	return err;
5529 }
5530 
5531 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5532 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5533 
5534 /*
5535  * sleep, with exponential backoff, and retry the LOCK operation.
5536  */
5537 static unsigned long
5538 nfs4_set_lock_task_retry(unsigned long timeout)
5539 {
5540 	freezable_schedule_timeout_killable_unsafe(timeout);
5541 	timeout <<= 1;
5542 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5543 		return NFS4_LOCK_MAXTIMEOUT;
5544 	return timeout;
5545 }
5546 
5547 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5548 {
5549 	struct inode *inode = state->inode;
5550 	struct nfs_server *server = NFS_SERVER(inode);
5551 	struct nfs_client *clp = server->nfs_client;
5552 	struct nfs_lockt_args arg = {
5553 		.fh = NFS_FH(inode),
5554 		.fl = request,
5555 	};
5556 	struct nfs_lockt_res res = {
5557 		.denied = request,
5558 	};
5559 	struct rpc_message msg = {
5560 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5561 		.rpc_argp       = &arg,
5562 		.rpc_resp       = &res,
5563 		.rpc_cred	= state->owner->so_cred,
5564 	};
5565 	struct nfs4_lock_state *lsp;
5566 	int status;
5567 
5568 	arg.lock_owner.clientid = clp->cl_clientid;
5569 	status = nfs4_set_lock_state(state, request);
5570 	if (status != 0)
5571 		goto out;
5572 	lsp = request->fl_u.nfs4_fl.owner;
5573 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5574 	arg.lock_owner.s_dev = server->s_dev;
5575 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5576 	switch (status) {
5577 		case 0:
5578 			request->fl_type = F_UNLCK;
5579 			break;
5580 		case -NFS4ERR_DENIED:
5581 			status = 0;
5582 	}
5583 	request->fl_ops->fl_release_private(request);
5584 	request->fl_ops = NULL;
5585 out:
5586 	return status;
5587 }
5588 
5589 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5590 {
5591 	struct nfs4_exception exception = { };
5592 	int err;
5593 
5594 	do {
5595 		err = _nfs4_proc_getlk(state, cmd, request);
5596 		trace_nfs4_get_lock(request, state, cmd, err);
5597 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5598 				&exception);
5599 	} while (exception.retry);
5600 	return err;
5601 }
5602 
5603 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5604 {
5605 	return locks_lock_inode_wait(inode, fl);
5606 }
5607 
5608 struct nfs4_unlockdata {
5609 	struct nfs_locku_args arg;
5610 	struct nfs_locku_res res;
5611 	struct nfs4_lock_state *lsp;
5612 	struct nfs_open_context *ctx;
5613 	struct file_lock fl;
5614 	struct nfs_server *server;
5615 	unsigned long timestamp;
5616 };
5617 
5618 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5619 		struct nfs_open_context *ctx,
5620 		struct nfs4_lock_state *lsp,
5621 		struct nfs_seqid *seqid)
5622 {
5623 	struct nfs4_unlockdata *p;
5624 	struct inode *inode = lsp->ls_state->inode;
5625 
5626 	p = kzalloc(sizeof(*p), GFP_NOFS);
5627 	if (p == NULL)
5628 		return NULL;
5629 	p->arg.fh = NFS_FH(inode);
5630 	p->arg.fl = &p->fl;
5631 	p->arg.seqid = seqid;
5632 	p->res.seqid = seqid;
5633 	p->lsp = lsp;
5634 	atomic_inc(&lsp->ls_count);
5635 	/* Ensure we don't close file until we're done freeing locks! */
5636 	p->ctx = get_nfs_open_context(ctx);
5637 	memcpy(&p->fl, fl, sizeof(p->fl));
5638 	p->server = NFS_SERVER(inode);
5639 	return p;
5640 }
5641 
5642 static void nfs4_locku_release_calldata(void *data)
5643 {
5644 	struct nfs4_unlockdata *calldata = data;
5645 	nfs_free_seqid(calldata->arg.seqid);
5646 	nfs4_put_lock_state(calldata->lsp);
5647 	put_nfs_open_context(calldata->ctx);
5648 	kfree(calldata);
5649 }
5650 
5651 static void nfs4_locku_done(struct rpc_task *task, void *data)
5652 {
5653 	struct nfs4_unlockdata *calldata = data;
5654 
5655 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5656 		return;
5657 	switch (task->tk_status) {
5658 		case 0:
5659 			renew_lease(calldata->server, calldata->timestamp);
5660 			do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5661 			if (nfs4_update_lock_stateid(calldata->lsp,
5662 					&calldata->res.stateid))
5663 				break;
5664 		case -NFS4ERR_BAD_STATEID:
5665 		case -NFS4ERR_OLD_STATEID:
5666 		case -NFS4ERR_STALE_STATEID:
5667 		case -NFS4ERR_EXPIRED:
5668 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5669 						&calldata->lsp->ls_stateid))
5670 				rpc_restart_call_prepare(task);
5671 			break;
5672 		default:
5673 			if (nfs4_async_handle_error(task, calldata->server,
5674 						    NULL, NULL) == -EAGAIN)
5675 				rpc_restart_call_prepare(task);
5676 	}
5677 	nfs_release_seqid(calldata->arg.seqid);
5678 }
5679 
5680 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5681 {
5682 	struct nfs4_unlockdata *calldata = data;
5683 
5684 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5685 		goto out_wait;
5686 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5687 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5688 		/* Note: exit _without_ running nfs4_locku_done */
5689 		goto out_no_action;
5690 	}
5691 	calldata->timestamp = jiffies;
5692 	if (nfs4_setup_sequence(calldata->server,
5693 				&calldata->arg.seq_args,
5694 				&calldata->res.seq_res,
5695 				task) != 0)
5696 		nfs_release_seqid(calldata->arg.seqid);
5697 	return;
5698 out_no_action:
5699 	task->tk_action = NULL;
5700 out_wait:
5701 	nfs4_sequence_done(task, &calldata->res.seq_res);
5702 }
5703 
5704 static const struct rpc_call_ops nfs4_locku_ops = {
5705 	.rpc_call_prepare = nfs4_locku_prepare,
5706 	.rpc_call_done = nfs4_locku_done,
5707 	.rpc_release = nfs4_locku_release_calldata,
5708 };
5709 
5710 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5711 		struct nfs_open_context *ctx,
5712 		struct nfs4_lock_state *lsp,
5713 		struct nfs_seqid *seqid)
5714 {
5715 	struct nfs4_unlockdata *data;
5716 	struct rpc_message msg = {
5717 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5718 		.rpc_cred = ctx->cred,
5719 	};
5720 	struct rpc_task_setup task_setup_data = {
5721 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5722 		.rpc_message = &msg,
5723 		.callback_ops = &nfs4_locku_ops,
5724 		.workqueue = nfsiod_workqueue,
5725 		.flags = RPC_TASK_ASYNC,
5726 	};
5727 
5728 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5729 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5730 
5731 	/* Ensure this is an unlock - when canceling a lock, the
5732 	 * canceled lock is passed in, and it won't be an unlock.
5733 	 */
5734 	fl->fl_type = F_UNLCK;
5735 
5736 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5737 	if (data == NULL) {
5738 		nfs_free_seqid(seqid);
5739 		return ERR_PTR(-ENOMEM);
5740 	}
5741 
5742 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5743 	msg.rpc_argp = &data->arg;
5744 	msg.rpc_resp = &data->res;
5745 	task_setup_data.callback_data = data;
5746 	return rpc_run_task(&task_setup_data);
5747 }
5748 
5749 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5750 {
5751 	struct inode *inode = state->inode;
5752 	struct nfs4_state_owner *sp = state->owner;
5753 	struct nfs_inode *nfsi = NFS_I(inode);
5754 	struct nfs_seqid *seqid;
5755 	struct nfs4_lock_state *lsp;
5756 	struct rpc_task *task;
5757 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5758 	int status = 0;
5759 	unsigned char fl_flags = request->fl_flags;
5760 
5761 	status = nfs4_set_lock_state(state, request);
5762 	/* Unlock _before_ we do the RPC call */
5763 	request->fl_flags |= FL_EXISTS;
5764 	/* Exclude nfs_delegation_claim_locks() */
5765 	mutex_lock(&sp->so_delegreturn_mutex);
5766 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5767 	down_read(&nfsi->rwsem);
5768 	if (do_vfs_lock(inode, request) == -ENOENT) {
5769 		up_read(&nfsi->rwsem);
5770 		mutex_unlock(&sp->so_delegreturn_mutex);
5771 		goto out;
5772 	}
5773 	up_read(&nfsi->rwsem);
5774 	mutex_unlock(&sp->so_delegreturn_mutex);
5775 	if (status != 0)
5776 		goto out;
5777 	/* Is this a delegated lock? */
5778 	lsp = request->fl_u.nfs4_fl.owner;
5779 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5780 		goto out;
5781 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5782 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5783 	status = -ENOMEM;
5784 	if (IS_ERR(seqid))
5785 		goto out;
5786 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5787 	status = PTR_ERR(task);
5788 	if (IS_ERR(task))
5789 		goto out;
5790 	status = nfs4_wait_for_completion_rpc_task(task);
5791 	rpc_put_task(task);
5792 out:
5793 	request->fl_flags = fl_flags;
5794 	trace_nfs4_unlock(request, state, F_SETLK, status);
5795 	return status;
5796 }
5797 
5798 struct nfs4_lockdata {
5799 	struct nfs_lock_args arg;
5800 	struct nfs_lock_res res;
5801 	struct nfs4_lock_state *lsp;
5802 	struct nfs_open_context *ctx;
5803 	struct file_lock fl;
5804 	unsigned long timestamp;
5805 	int rpc_status;
5806 	int cancelled;
5807 	struct nfs_server *server;
5808 };
5809 
5810 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5811 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5812 		gfp_t gfp_mask)
5813 {
5814 	struct nfs4_lockdata *p;
5815 	struct inode *inode = lsp->ls_state->inode;
5816 	struct nfs_server *server = NFS_SERVER(inode);
5817 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5818 
5819 	p = kzalloc(sizeof(*p), gfp_mask);
5820 	if (p == NULL)
5821 		return NULL;
5822 
5823 	p->arg.fh = NFS_FH(inode);
5824 	p->arg.fl = &p->fl;
5825 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5826 	if (IS_ERR(p->arg.open_seqid))
5827 		goto out_free;
5828 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5829 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5830 	if (IS_ERR(p->arg.lock_seqid))
5831 		goto out_free_seqid;
5832 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5833 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5834 	p->arg.lock_owner.s_dev = server->s_dev;
5835 	p->res.lock_seqid = p->arg.lock_seqid;
5836 	p->lsp = lsp;
5837 	p->server = server;
5838 	atomic_inc(&lsp->ls_count);
5839 	p->ctx = get_nfs_open_context(ctx);
5840 	get_file(fl->fl_file);
5841 	memcpy(&p->fl, fl, sizeof(p->fl));
5842 	return p;
5843 out_free_seqid:
5844 	nfs_free_seqid(p->arg.open_seqid);
5845 out_free:
5846 	kfree(p);
5847 	return NULL;
5848 }
5849 
5850 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5851 {
5852 	struct nfs4_lockdata *data = calldata;
5853 	struct nfs4_state *state = data->lsp->ls_state;
5854 
5855 	dprintk("%s: begin!\n", __func__);
5856 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5857 		goto out_wait;
5858 	/* Do we need to do an open_to_lock_owner? */
5859 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5860 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5861 			goto out_release_lock_seqid;
5862 		}
5863 		nfs4_stateid_copy(&data->arg.open_stateid,
5864 				&state->open_stateid);
5865 		data->arg.new_lock_owner = 1;
5866 		data->res.open_seqid = data->arg.open_seqid;
5867 	} else {
5868 		data->arg.new_lock_owner = 0;
5869 		nfs4_stateid_copy(&data->arg.lock_stateid,
5870 				&data->lsp->ls_stateid);
5871 	}
5872 	if (!nfs4_valid_open_stateid(state)) {
5873 		data->rpc_status = -EBADF;
5874 		task->tk_action = NULL;
5875 		goto out_release_open_seqid;
5876 	}
5877 	data->timestamp = jiffies;
5878 	if (nfs4_setup_sequence(data->server,
5879 				&data->arg.seq_args,
5880 				&data->res.seq_res,
5881 				task) == 0)
5882 		return;
5883 out_release_open_seqid:
5884 	nfs_release_seqid(data->arg.open_seqid);
5885 out_release_lock_seqid:
5886 	nfs_release_seqid(data->arg.lock_seqid);
5887 out_wait:
5888 	nfs4_sequence_done(task, &data->res.seq_res);
5889 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5890 }
5891 
5892 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5893 {
5894 	struct nfs4_lockdata *data = calldata;
5895 	struct nfs4_lock_state *lsp = data->lsp;
5896 
5897 	dprintk("%s: begin!\n", __func__);
5898 
5899 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5900 		return;
5901 
5902 	data->rpc_status = task->tk_status;
5903 	switch (task->tk_status) {
5904 	case 0:
5905 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5906 				data->timestamp);
5907 		if (data->arg.new_lock) {
5908 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5909 			if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5910 				rpc_restart_call_prepare(task);
5911 				break;
5912 			}
5913 		}
5914 		if (data->arg.new_lock_owner != 0) {
5915 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5916 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5917 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5918 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5919 			rpc_restart_call_prepare(task);
5920 		break;
5921 	case -NFS4ERR_BAD_STATEID:
5922 	case -NFS4ERR_OLD_STATEID:
5923 	case -NFS4ERR_STALE_STATEID:
5924 	case -NFS4ERR_EXPIRED:
5925 		if (data->arg.new_lock_owner != 0) {
5926 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5927 						&lsp->ls_state->open_stateid))
5928 				rpc_restart_call_prepare(task);
5929 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5930 						&lsp->ls_stateid))
5931 				rpc_restart_call_prepare(task);
5932 	}
5933 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5934 }
5935 
5936 static void nfs4_lock_release(void *calldata)
5937 {
5938 	struct nfs4_lockdata *data = calldata;
5939 
5940 	dprintk("%s: begin!\n", __func__);
5941 	nfs_free_seqid(data->arg.open_seqid);
5942 	if (data->cancelled != 0) {
5943 		struct rpc_task *task;
5944 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5945 				data->arg.lock_seqid);
5946 		if (!IS_ERR(task))
5947 			rpc_put_task_async(task);
5948 		dprintk("%s: cancelling lock!\n", __func__);
5949 	} else
5950 		nfs_free_seqid(data->arg.lock_seqid);
5951 	nfs4_put_lock_state(data->lsp);
5952 	put_nfs_open_context(data->ctx);
5953 	fput(data->fl.fl_file);
5954 	kfree(data);
5955 	dprintk("%s: done!\n", __func__);
5956 }
5957 
5958 static const struct rpc_call_ops nfs4_lock_ops = {
5959 	.rpc_call_prepare = nfs4_lock_prepare,
5960 	.rpc_call_done = nfs4_lock_done,
5961 	.rpc_release = nfs4_lock_release,
5962 };
5963 
5964 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5965 {
5966 	switch (error) {
5967 	case -NFS4ERR_ADMIN_REVOKED:
5968 	case -NFS4ERR_BAD_STATEID:
5969 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5970 		if (new_lock_owner != 0 ||
5971 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5972 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5973 		break;
5974 	case -NFS4ERR_STALE_STATEID:
5975 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5976 	case -NFS4ERR_EXPIRED:
5977 		nfs4_schedule_lease_recovery(server->nfs_client);
5978 	};
5979 }
5980 
5981 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5982 {
5983 	struct nfs4_lockdata *data;
5984 	struct rpc_task *task;
5985 	struct rpc_message msg = {
5986 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5987 		.rpc_cred = state->owner->so_cred,
5988 	};
5989 	struct rpc_task_setup task_setup_data = {
5990 		.rpc_client = NFS_CLIENT(state->inode),
5991 		.rpc_message = &msg,
5992 		.callback_ops = &nfs4_lock_ops,
5993 		.workqueue = nfsiod_workqueue,
5994 		.flags = RPC_TASK_ASYNC,
5995 	};
5996 	int ret;
5997 
5998 	dprintk("%s: begin!\n", __func__);
5999 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6000 			fl->fl_u.nfs4_fl.owner,
6001 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6002 	if (data == NULL)
6003 		return -ENOMEM;
6004 	if (IS_SETLKW(cmd))
6005 		data->arg.block = 1;
6006 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6007 	msg.rpc_argp = &data->arg;
6008 	msg.rpc_resp = &data->res;
6009 	task_setup_data.callback_data = data;
6010 	if (recovery_type > NFS_LOCK_NEW) {
6011 		if (recovery_type == NFS_LOCK_RECLAIM)
6012 			data->arg.reclaim = NFS_LOCK_RECLAIM;
6013 		nfs4_set_sequence_privileged(&data->arg.seq_args);
6014 	} else
6015 		data->arg.new_lock = 1;
6016 	task = rpc_run_task(&task_setup_data);
6017 	if (IS_ERR(task))
6018 		return PTR_ERR(task);
6019 	ret = nfs4_wait_for_completion_rpc_task(task);
6020 	if (ret == 0) {
6021 		ret = data->rpc_status;
6022 		if (ret)
6023 			nfs4_handle_setlk_error(data->server, data->lsp,
6024 					data->arg.new_lock_owner, ret);
6025 	} else
6026 		data->cancelled = 1;
6027 	rpc_put_task(task);
6028 	dprintk("%s: done, ret = %d!\n", __func__, ret);
6029 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6030 	return ret;
6031 }
6032 
6033 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6034 {
6035 	struct nfs_server *server = NFS_SERVER(state->inode);
6036 	struct nfs4_exception exception = {
6037 		.inode = state->inode,
6038 	};
6039 	int err;
6040 
6041 	do {
6042 		/* Cache the lock if possible... */
6043 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6044 			return 0;
6045 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6046 		if (err != -NFS4ERR_DELAY)
6047 			break;
6048 		nfs4_handle_exception(server, err, &exception);
6049 	} while (exception.retry);
6050 	return err;
6051 }
6052 
6053 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6054 {
6055 	struct nfs_server *server = NFS_SERVER(state->inode);
6056 	struct nfs4_exception exception = {
6057 		.inode = state->inode,
6058 	};
6059 	int err;
6060 
6061 	err = nfs4_set_lock_state(state, request);
6062 	if (err != 0)
6063 		return err;
6064 	if (!recover_lost_locks) {
6065 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6066 		return 0;
6067 	}
6068 	do {
6069 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6070 			return 0;
6071 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6072 		switch (err) {
6073 		default:
6074 			goto out;
6075 		case -NFS4ERR_GRACE:
6076 		case -NFS4ERR_DELAY:
6077 			nfs4_handle_exception(server, err, &exception);
6078 			err = 0;
6079 		}
6080 	} while (exception.retry);
6081 out:
6082 	return err;
6083 }
6084 
6085 #if defined(CONFIG_NFS_V4_1)
6086 /**
6087  * nfs41_check_expired_locks - possibly free a lock stateid
6088  *
6089  * @state: NFSv4 state for an inode
6090  *
6091  * Returns NFS_OK if recovery for this stateid is now finished.
6092  * Otherwise a negative NFS4ERR value is returned.
6093  */
6094 static int nfs41_check_expired_locks(struct nfs4_state *state)
6095 {
6096 	int status, ret = -NFS4ERR_BAD_STATEID;
6097 	struct nfs4_lock_state *lsp;
6098 	struct nfs_server *server = NFS_SERVER(state->inode);
6099 
6100 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6101 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6102 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6103 
6104 			status = nfs41_test_stateid(server,
6105 					&lsp->ls_stateid,
6106 					cred);
6107 			trace_nfs4_test_lock_stateid(state, lsp, status);
6108 			if (status != NFS_OK) {
6109 				/* Free the stateid unless the server
6110 				 * informs us the stateid is unrecognized. */
6111 				if (status != -NFS4ERR_BAD_STATEID)
6112 					nfs41_free_stateid(server,
6113 							&lsp->ls_stateid,
6114 							cred);
6115 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6116 				ret = status;
6117 			}
6118 		}
6119 	};
6120 
6121 	return ret;
6122 }
6123 
6124 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6125 {
6126 	int status = NFS_OK;
6127 
6128 	if (test_bit(LK_STATE_IN_USE, &state->flags))
6129 		status = nfs41_check_expired_locks(state);
6130 	if (status != NFS_OK)
6131 		status = nfs4_lock_expired(state, request);
6132 	return status;
6133 }
6134 #endif
6135 
6136 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6137 {
6138 	struct nfs_inode *nfsi = NFS_I(state->inode);
6139 	struct nfs4_state_owner *sp = state->owner;
6140 	unsigned char fl_flags = request->fl_flags;
6141 	int status = -ENOLCK;
6142 
6143 	if ((fl_flags & FL_POSIX) &&
6144 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6145 		goto out;
6146 	/* Is this a delegated open? */
6147 	status = nfs4_set_lock_state(state, request);
6148 	if (status != 0)
6149 		goto out;
6150 	request->fl_flags |= FL_ACCESS;
6151 	status = do_vfs_lock(state->inode, request);
6152 	if (status < 0)
6153 		goto out;
6154 	mutex_lock(&sp->so_delegreturn_mutex);
6155 	down_read(&nfsi->rwsem);
6156 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6157 		/* Yes: cache locks! */
6158 		/* ...but avoid races with delegation recall... */
6159 		request->fl_flags = fl_flags & ~FL_SLEEP;
6160 		status = do_vfs_lock(state->inode, request);
6161 		up_read(&nfsi->rwsem);
6162 		mutex_unlock(&sp->so_delegreturn_mutex);
6163 		goto out;
6164 	}
6165 	up_read(&nfsi->rwsem);
6166 	mutex_unlock(&sp->so_delegreturn_mutex);
6167 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6168 out:
6169 	request->fl_flags = fl_flags;
6170 	return status;
6171 }
6172 
6173 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6174 {
6175 	struct nfs4_exception exception = {
6176 		.state = state,
6177 		.inode = state->inode,
6178 	};
6179 	int err;
6180 
6181 	do {
6182 		err = _nfs4_proc_setlk(state, cmd, request);
6183 		if (err == -NFS4ERR_DENIED)
6184 			err = -EAGAIN;
6185 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
6186 				err, &exception);
6187 	} while (exception.retry);
6188 	return err;
6189 }
6190 
6191 static int
6192 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6193 {
6194 	struct nfs_open_context *ctx;
6195 	struct nfs4_state *state;
6196 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6197 	int status;
6198 
6199 	/* verify open state */
6200 	ctx = nfs_file_open_context(filp);
6201 	state = ctx->state;
6202 
6203 	if (request->fl_start < 0 || request->fl_end < 0)
6204 		return -EINVAL;
6205 
6206 	if (IS_GETLK(cmd)) {
6207 		if (state != NULL)
6208 			return nfs4_proc_getlk(state, F_GETLK, request);
6209 		return 0;
6210 	}
6211 
6212 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6213 		return -EINVAL;
6214 
6215 	if (request->fl_type == F_UNLCK) {
6216 		if (state != NULL)
6217 			return nfs4_proc_unlck(state, cmd, request);
6218 		return 0;
6219 	}
6220 
6221 	if (state == NULL)
6222 		return -ENOLCK;
6223 	/*
6224 	 * Don't rely on the VFS having checked the file open mode,
6225 	 * since it won't do this for flock() locks.
6226 	 */
6227 	switch (request->fl_type) {
6228 	case F_RDLCK:
6229 		if (!(filp->f_mode & FMODE_READ))
6230 			return -EBADF;
6231 		break;
6232 	case F_WRLCK:
6233 		if (!(filp->f_mode & FMODE_WRITE))
6234 			return -EBADF;
6235 	}
6236 
6237 	do {
6238 		status = nfs4_proc_setlk(state, cmd, request);
6239 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6240 			break;
6241 		timeout = nfs4_set_lock_task_retry(timeout);
6242 		status = -ERESTARTSYS;
6243 		if (signalled())
6244 			break;
6245 	} while(status < 0);
6246 	return status;
6247 }
6248 
6249 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6250 {
6251 	struct nfs_server *server = NFS_SERVER(state->inode);
6252 	int err;
6253 
6254 	err = nfs4_set_lock_state(state, fl);
6255 	if (err != 0)
6256 		return err;
6257 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6258 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6259 }
6260 
6261 struct nfs_release_lockowner_data {
6262 	struct nfs4_lock_state *lsp;
6263 	struct nfs_server *server;
6264 	struct nfs_release_lockowner_args args;
6265 	struct nfs_release_lockowner_res res;
6266 	unsigned long timestamp;
6267 };
6268 
6269 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6270 {
6271 	struct nfs_release_lockowner_data *data = calldata;
6272 	struct nfs_server *server = data->server;
6273 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6274 			     &data->args.seq_args, &data->res.seq_res, task);
6275 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6276 	data->timestamp = jiffies;
6277 }
6278 
6279 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6280 {
6281 	struct nfs_release_lockowner_data *data = calldata;
6282 	struct nfs_server *server = data->server;
6283 
6284 	nfs40_sequence_done(task, &data->res.seq_res);
6285 
6286 	switch (task->tk_status) {
6287 	case 0:
6288 		renew_lease(server, data->timestamp);
6289 		break;
6290 	case -NFS4ERR_STALE_CLIENTID:
6291 	case -NFS4ERR_EXPIRED:
6292 		nfs4_schedule_lease_recovery(server->nfs_client);
6293 		break;
6294 	case -NFS4ERR_LEASE_MOVED:
6295 	case -NFS4ERR_DELAY:
6296 		if (nfs4_async_handle_error(task, server,
6297 					    NULL, NULL) == -EAGAIN)
6298 			rpc_restart_call_prepare(task);
6299 	}
6300 }
6301 
6302 static void nfs4_release_lockowner_release(void *calldata)
6303 {
6304 	struct nfs_release_lockowner_data *data = calldata;
6305 	nfs4_free_lock_state(data->server, data->lsp);
6306 	kfree(calldata);
6307 }
6308 
6309 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6310 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6311 	.rpc_call_done = nfs4_release_lockowner_done,
6312 	.rpc_release = nfs4_release_lockowner_release,
6313 };
6314 
6315 static void
6316 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6317 {
6318 	struct nfs_release_lockowner_data *data;
6319 	struct rpc_message msg = {
6320 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6321 	};
6322 
6323 	if (server->nfs_client->cl_mvops->minor_version != 0)
6324 		return;
6325 
6326 	data = kmalloc(sizeof(*data), GFP_NOFS);
6327 	if (!data)
6328 		return;
6329 	data->lsp = lsp;
6330 	data->server = server;
6331 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6332 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6333 	data->args.lock_owner.s_dev = server->s_dev;
6334 
6335 	msg.rpc_argp = &data->args;
6336 	msg.rpc_resp = &data->res;
6337 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6338 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6339 }
6340 
6341 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6342 
6343 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6344 				   struct dentry *unused, struct inode *inode,
6345 				   const char *key, const void *buf,
6346 				   size_t buflen, int flags)
6347 {
6348 	return nfs4_proc_set_acl(inode, buf, buflen);
6349 }
6350 
6351 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6352 				   struct dentry *unused, struct inode *inode,
6353 				   const char *key, void *buf, size_t buflen)
6354 {
6355 	return nfs4_proc_get_acl(inode, buf, buflen);
6356 }
6357 
6358 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6359 {
6360 	return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6361 }
6362 
6363 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6364 
6365 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6366 				     struct dentry *unused, struct inode *inode,
6367 				     const char *key, const void *buf,
6368 				     size_t buflen, int flags)
6369 {
6370 	if (security_ismaclabel(key))
6371 		return nfs4_set_security_label(inode, buf, buflen);
6372 
6373 	return -EOPNOTSUPP;
6374 }
6375 
6376 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6377 				     struct dentry *unused, struct inode *inode,
6378 				     const char *key, void *buf, size_t buflen)
6379 {
6380 	if (security_ismaclabel(key))
6381 		return nfs4_get_security_label(inode, buf, buflen);
6382 	return -EOPNOTSUPP;
6383 }
6384 
6385 static ssize_t
6386 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6387 {
6388 	int len = 0;
6389 
6390 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6391 		len = security_inode_listsecurity(inode, list, list_len);
6392 		if (list_len && len > list_len)
6393 			return -ERANGE;
6394 	}
6395 	return len;
6396 }
6397 
6398 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6399 	.prefix = XATTR_SECURITY_PREFIX,
6400 	.get	= nfs4_xattr_get_nfs4_label,
6401 	.set	= nfs4_xattr_set_nfs4_label,
6402 };
6403 
6404 #else
6405 
6406 static ssize_t
6407 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6408 {
6409 	return 0;
6410 }
6411 
6412 #endif
6413 
6414 /*
6415  * nfs_fhget will use either the mounted_on_fileid or the fileid
6416  */
6417 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6418 {
6419 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6420 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6421 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6422 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6423 		return;
6424 
6425 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6426 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6427 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6428 	fattr->nlink = 2;
6429 }
6430 
6431 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6432 				   const struct qstr *name,
6433 				   struct nfs4_fs_locations *fs_locations,
6434 				   struct page *page)
6435 {
6436 	struct nfs_server *server = NFS_SERVER(dir);
6437 	u32 bitmask[3] = {
6438 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6439 	};
6440 	struct nfs4_fs_locations_arg args = {
6441 		.dir_fh = NFS_FH(dir),
6442 		.name = name,
6443 		.page = page,
6444 		.bitmask = bitmask,
6445 	};
6446 	struct nfs4_fs_locations_res res = {
6447 		.fs_locations = fs_locations,
6448 	};
6449 	struct rpc_message msg = {
6450 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6451 		.rpc_argp = &args,
6452 		.rpc_resp = &res,
6453 	};
6454 	int status;
6455 
6456 	dprintk("%s: start\n", __func__);
6457 
6458 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6459 	 * is not supported */
6460 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6461 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6462 	else
6463 		bitmask[0] |= FATTR4_WORD0_FILEID;
6464 
6465 	nfs_fattr_init(&fs_locations->fattr);
6466 	fs_locations->server = server;
6467 	fs_locations->nlocations = 0;
6468 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6469 	dprintk("%s: returned status = %d\n", __func__, status);
6470 	return status;
6471 }
6472 
6473 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6474 			   const struct qstr *name,
6475 			   struct nfs4_fs_locations *fs_locations,
6476 			   struct page *page)
6477 {
6478 	struct nfs4_exception exception = { };
6479 	int err;
6480 	do {
6481 		err = _nfs4_proc_fs_locations(client, dir, name,
6482 				fs_locations, page);
6483 		trace_nfs4_get_fs_locations(dir, name, err);
6484 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6485 				&exception);
6486 	} while (exception.retry);
6487 	return err;
6488 }
6489 
6490 /*
6491  * This operation also signals the server that this client is
6492  * performing migration recovery.  The server can stop returning
6493  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6494  * appended to this compound to identify the client ID which is
6495  * performing recovery.
6496  */
6497 static int _nfs40_proc_get_locations(struct inode *inode,
6498 				     struct nfs4_fs_locations *locations,
6499 				     struct page *page, struct rpc_cred *cred)
6500 {
6501 	struct nfs_server *server = NFS_SERVER(inode);
6502 	struct rpc_clnt *clnt = server->client;
6503 	u32 bitmask[2] = {
6504 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6505 	};
6506 	struct nfs4_fs_locations_arg args = {
6507 		.clientid	= server->nfs_client->cl_clientid,
6508 		.fh		= NFS_FH(inode),
6509 		.page		= page,
6510 		.bitmask	= bitmask,
6511 		.migration	= 1,		/* skip LOOKUP */
6512 		.renew		= 1,		/* append RENEW */
6513 	};
6514 	struct nfs4_fs_locations_res res = {
6515 		.fs_locations	= locations,
6516 		.migration	= 1,
6517 		.renew		= 1,
6518 	};
6519 	struct rpc_message msg = {
6520 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6521 		.rpc_argp	= &args,
6522 		.rpc_resp	= &res,
6523 		.rpc_cred	= cred,
6524 	};
6525 	unsigned long now = jiffies;
6526 	int status;
6527 
6528 	nfs_fattr_init(&locations->fattr);
6529 	locations->server = server;
6530 	locations->nlocations = 0;
6531 
6532 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6533 	nfs4_set_sequence_privileged(&args.seq_args);
6534 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6535 					&args.seq_args, &res.seq_res);
6536 	if (status)
6537 		return status;
6538 
6539 	renew_lease(server, now);
6540 	return 0;
6541 }
6542 
6543 #ifdef CONFIG_NFS_V4_1
6544 
6545 /*
6546  * This operation also signals the server that this client is
6547  * performing migration recovery.  The server can stop asserting
6548  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6549  * performing this operation is identified in the SEQUENCE
6550  * operation in this compound.
6551  *
6552  * When the client supports GETATTR(fs_locations_info), it can
6553  * be plumbed in here.
6554  */
6555 static int _nfs41_proc_get_locations(struct inode *inode,
6556 				     struct nfs4_fs_locations *locations,
6557 				     struct page *page, struct rpc_cred *cred)
6558 {
6559 	struct nfs_server *server = NFS_SERVER(inode);
6560 	struct rpc_clnt *clnt = server->client;
6561 	u32 bitmask[2] = {
6562 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6563 	};
6564 	struct nfs4_fs_locations_arg args = {
6565 		.fh		= NFS_FH(inode),
6566 		.page		= page,
6567 		.bitmask	= bitmask,
6568 		.migration	= 1,		/* skip LOOKUP */
6569 	};
6570 	struct nfs4_fs_locations_res res = {
6571 		.fs_locations	= locations,
6572 		.migration	= 1,
6573 	};
6574 	struct rpc_message msg = {
6575 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6576 		.rpc_argp	= &args,
6577 		.rpc_resp	= &res,
6578 		.rpc_cred	= cred,
6579 	};
6580 	int status;
6581 
6582 	nfs_fattr_init(&locations->fattr);
6583 	locations->server = server;
6584 	locations->nlocations = 0;
6585 
6586 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6587 	nfs4_set_sequence_privileged(&args.seq_args);
6588 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6589 					&args.seq_args, &res.seq_res);
6590 	if (status == NFS4_OK &&
6591 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6592 		status = -NFS4ERR_LEASE_MOVED;
6593 	return status;
6594 }
6595 
6596 #endif	/* CONFIG_NFS_V4_1 */
6597 
6598 /**
6599  * nfs4_proc_get_locations - discover locations for a migrated FSID
6600  * @inode: inode on FSID that is migrating
6601  * @locations: result of query
6602  * @page: buffer
6603  * @cred: credential to use for this operation
6604  *
6605  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6606  * operation failed, or a negative errno if a local error occurred.
6607  *
6608  * On success, "locations" is filled in, but if the server has
6609  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6610  * asserted.
6611  *
6612  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6613  * from this client that require migration recovery.
6614  */
6615 int nfs4_proc_get_locations(struct inode *inode,
6616 			    struct nfs4_fs_locations *locations,
6617 			    struct page *page, struct rpc_cred *cred)
6618 {
6619 	struct nfs_server *server = NFS_SERVER(inode);
6620 	struct nfs_client *clp = server->nfs_client;
6621 	const struct nfs4_mig_recovery_ops *ops =
6622 					clp->cl_mvops->mig_recovery_ops;
6623 	struct nfs4_exception exception = { };
6624 	int status;
6625 
6626 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6627 		(unsigned long long)server->fsid.major,
6628 		(unsigned long long)server->fsid.minor,
6629 		clp->cl_hostname);
6630 	nfs_display_fhandle(NFS_FH(inode), __func__);
6631 
6632 	do {
6633 		status = ops->get_locations(inode, locations, page, cred);
6634 		if (status != -NFS4ERR_DELAY)
6635 			break;
6636 		nfs4_handle_exception(server, status, &exception);
6637 	} while (exception.retry);
6638 	return status;
6639 }
6640 
6641 /*
6642  * This operation also signals the server that this client is
6643  * performing "lease moved" recovery.  The server can stop
6644  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6645  * is appended to this compound to identify the client ID which is
6646  * performing recovery.
6647  */
6648 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6649 {
6650 	struct nfs_server *server = NFS_SERVER(inode);
6651 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6652 	struct rpc_clnt *clnt = server->client;
6653 	struct nfs4_fsid_present_arg args = {
6654 		.fh		= NFS_FH(inode),
6655 		.clientid	= clp->cl_clientid,
6656 		.renew		= 1,		/* append RENEW */
6657 	};
6658 	struct nfs4_fsid_present_res res = {
6659 		.renew		= 1,
6660 	};
6661 	struct rpc_message msg = {
6662 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6663 		.rpc_argp	= &args,
6664 		.rpc_resp	= &res,
6665 		.rpc_cred	= cred,
6666 	};
6667 	unsigned long now = jiffies;
6668 	int status;
6669 
6670 	res.fh = nfs_alloc_fhandle();
6671 	if (res.fh == NULL)
6672 		return -ENOMEM;
6673 
6674 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6675 	nfs4_set_sequence_privileged(&args.seq_args);
6676 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6677 						&args.seq_args, &res.seq_res);
6678 	nfs_free_fhandle(res.fh);
6679 	if (status)
6680 		return status;
6681 
6682 	do_renew_lease(clp, now);
6683 	return 0;
6684 }
6685 
6686 #ifdef CONFIG_NFS_V4_1
6687 
6688 /*
6689  * This operation also signals the server that this client is
6690  * performing "lease moved" recovery.  The server can stop asserting
6691  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6692  * this operation is identified in the SEQUENCE operation in this
6693  * compound.
6694  */
6695 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6696 {
6697 	struct nfs_server *server = NFS_SERVER(inode);
6698 	struct rpc_clnt *clnt = server->client;
6699 	struct nfs4_fsid_present_arg args = {
6700 		.fh		= NFS_FH(inode),
6701 	};
6702 	struct nfs4_fsid_present_res res = {
6703 	};
6704 	struct rpc_message msg = {
6705 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6706 		.rpc_argp	= &args,
6707 		.rpc_resp	= &res,
6708 		.rpc_cred	= cred,
6709 	};
6710 	int status;
6711 
6712 	res.fh = nfs_alloc_fhandle();
6713 	if (res.fh == NULL)
6714 		return -ENOMEM;
6715 
6716 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6717 	nfs4_set_sequence_privileged(&args.seq_args);
6718 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6719 						&args.seq_args, &res.seq_res);
6720 	nfs_free_fhandle(res.fh);
6721 	if (status == NFS4_OK &&
6722 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6723 		status = -NFS4ERR_LEASE_MOVED;
6724 	return status;
6725 }
6726 
6727 #endif	/* CONFIG_NFS_V4_1 */
6728 
6729 /**
6730  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6731  * @inode: inode on FSID to check
6732  * @cred: credential to use for this operation
6733  *
6734  * Server indicates whether the FSID is present, moved, or not
6735  * recognized.  This operation is necessary to clear a LEASE_MOVED
6736  * condition for this client ID.
6737  *
6738  * Returns NFS4_OK if the FSID is present on this server,
6739  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6740  *  NFS4ERR code if some error occurred on the server, or a
6741  *  negative errno if a local failure occurred.
6742  */
6743 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6744 {
6745 	struct nfs_server *server = NFS_SERVER(inode);
6746 	struct nfs_client *clp = server->nfs_client;
6747 	const struct nfs4_mig_recovery_ops *ops =
6748 					clp->cl_mvops->mig_recovery_ops;
6749 	struct nfs4_exception exception = { };
6750 	int status;
6751 
6752 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6753 		(unsigned long long)server->fsid.major,
6754 		(unsigned long long)server->fsid.minor,
6755 		clp->cl_hostname);
6756 	nfs_display_fhandle(NFS_FH(inode), __func__);
6757 
6758 	do {
6759 		status = ops->fsid_present(inode, cred);
6760 		if (status != -NFS4ERR_DELAY)
6761 			break;
6762 		nfs4_handle_exception(server, status, &exception);
6763 	} while (exception.retry);
6764 	return status;
6765 }
6766 
6767 /**
6768  * If 'use_integrity' is true and the state managment nfs_client
6769  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6770  * and the machine credential as per RFC3530bis and RFC5661 Security
6771  * Considerations sections. Otherwise, just use the user cred with the
6772  * filesystem's rpc_client.
6773  */
6774 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6775 {
6776 	int status;
6777 	struct nfs4_secinfo_arg args = {
6778 		.dir_fh = NFS_FH(dir),
6779 		.name   = name,
6780 	};
6781 	struct nfs4_secinfo_res res = {
6782 		.flavors     = flavors,
6783 	};
6784 	struct rpc_message msg = {
6785 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6786 		.rpc_argp = &args,
6787 		.rpc_resp = &res,
6788 	};
6789 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6790 	struct rpc_cred *cred = NULL;
6791 
6792 	if (use_integrity) {
6793 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6794 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6795 		msg.rpc_cred = cred;
6796 	}
6797 
6798 	dprintk("NFS call  secinfo %s\n", name->name);
6799 
6800 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6801 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6802 
6803 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6804 				&res.seq_res, 0);
6805 	dprintk("NFS reply  secinfo: %d\n", status);
6806 
6807 	if (cred)
6808 		put_rpccred(cred);
6809 
6810 	return status;
6811 }
6812 
6813 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6814 		      struct nfs4_secinfo_flavors *flavors)
6815 {
6816 	struct nfs4_exception exception = { };
6817 	int err;
6818 	do {
6819 		err = -NFS4ERR_WRONGSEC;
6820 
6821 		/* try to use integrity protection with machine cred */
6822 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6823 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6824 
6825 		/*
6826 		 * if unable to use integrity protection, or SECINFO with
6827 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6828 		 * disallowed by spec, but exists in deployed servers) use
6829 		 * the current filesystem's rpc_client and the user cred.
6830 		 */
6831 		if (err == -NFS4ERR_WRONGSEC)
6832 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6833 
6834 		trace_nfs4_secinfo(dir, name, err);
6835 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6836 				&exception);
6837 	} while (exception.retry);
6838 	return err;
6839 }
6840 
6841 #ifdef CONFIG_NFS_V4_1
6842 /*
6843  * Check the exchange flags returned by the server for invalid flags, having
6844  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6845  * DS flags set.
6846  */
6847 static int nfs4_check_cl_exchange_flags(u32 flags)
6848 {
6849 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6850 		goto out_inval;
6851 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6852 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6853 		goto out_inval;
6854 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6855 		goto out_inval;
6856 	return NFS_OK;
6857 out_inval:
6858 	return -NFS4ERR_INVAL;
6859 }
6860 
6861 static bool
6862 nfs41_same_server_scope(struct nfs41_server_scope *a,
6863 			struct nfs41_server_scope *b)
6864 {
6865 	if (a->server_scope_sz == b->server_scope_sz &&
6866 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6867 		return true;
6868 
6869 	return false;
6870 }
6871 
6872 static void
6873 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6874 {
6875 }
6876 
6877 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6878 	.rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6879 };
6880 
6881 /*
6882  * nfs4_proc_bind_one_conn_to_session()
6883  *
6884  * The 4.1 client currently uses the same TCP connection for the
6885  * fore and backchannel.
6886  */
6887 static
6888 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6889 		struct rpc_xprt *xprt,
6890 		struct nfs_client *clp,
6891 		struct rpc_cred *cred)
6892 {
6893 	int status;
6894 	struct nfs41_bind_conn_to_session_args args = {
6895 		.client = clp,
6896 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6897 	};
6898 	struct nfs41_bind_conn_to_session_res res;
6899 	struct rpc_message msg = {
6900 		.rpc_proc =
6901 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6902 		.rpc_argp = &args,
6903 		.rpc_resp = &res,
6904 		.rpc_cred = cred,
6905 	};
6906 	struct rpc_task_setup task_setup_data = {
6907 		.rpc_client = clnt,
6908 		.rpc_xprt = xprt,
6909 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
6910 		.rpc_message = &msg,
6911 		.flags = RPC_TASK_TIMEOUT,
6912 	};
6913 	struct rpc_task *task;
6914 
6915 	dprintk("--> %s\n", __func__);
6916 
6917 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6918 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6919 		args.dir = NFS4_CDFC4_FORE;
6920 
6921 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
6922 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
6923 		args.dir = NFS4_CDFC4_FORE;
6924 
6925 	task = rpc_run_task(&task_setup_data);
6926 	if (!IS_ERR(task)) {
6927 		status = task->tk_status;
6928 		rpc_put_task(task);
6929 	} else
6930 		status = PTR_ERR(task);
6931 	trace_nfs4_bind_conn_to_session(clp, status);
6932 	if (status == 0) {
6933 		if (memcmp(res.sessionid.data,
6934 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6935 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6936 			status = -EIO;
6937 			goto out;
6938 		}
6939 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6940 			dprintk("NFS: %s: Unexpected direction from server\n",
6941 				__func__);
6942 			status = -EIO;
6943 			goto out;
6944 		}
6945 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6946 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6947 				__func__);
6948 			status = -EIO;
6949 			goto out;
6950 		}
6951 	}
6952 out:
6953 	dprintk("<-- %s status= %d\n", __func__, status);
6954 	return status;
6955 }
6956 
6957 struct rpc_bind_conn_calldata {
6958 	struct nfs_client *clp;
6959 	struct rpc_cred *cred;
6960 };
6961 
6962 static int
6963 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6964 		struct rpc_xprt *xprt,
6965 		void *calldata)
6966 {
6967 	struct rpc_bind_conn_calldata *p = calldata;
6968 
6969 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6970 }
6971 
6972 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6973 {
6974 	struct rpc_bind_conn_calldata data = {
6975 		.clp = clp,
6976 		.cred = cred,
6977 	};
6978 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6979 			nfs4_proc_bind_conn_to_session_callback, &data);
6980 }
6981 
6982 /*
6983  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6984  * and operations we'd like to see to enable certain features in the allow map
6985  */
6986 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6987 	.how = SP4_MACH_CRED,
6988 	.enforce.u.words = {
6989 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6990 		      1 << (OP_EXCHANGE_ID - 32) |
6991 		      1 << (OP_CREATE_SESSION - 32) |
6992 		      1 << (OP_DESTROY_SESSION - 32) |
6993 		      1 << (OP_DESTROY_CLIENTID - 32)
6994 	},
6995 	.allow.u.words = {
6996 		[0] = 1 << (OP_CLOSE) |
6997 		      1 << (OP_OPEN_DOWNGRADE) |
6998 		      1 << (OP_LOCKU) |
6999 		      1 << (OP_DELEGRETURN) |
7000 		      1 << (OP_COMMIT),
7001 		[1] = 1 << (OP_SECINFO - 32) |
7002 		      1 << (OP_SECINFO_NO_NAME - 32) |
7003 		      1 << (OP_LAYOUTRETURN - 32) |
7004 		      1 << (OP_TEST_STATEID - 32) |
7005 		      1 << (OP_FREE_STATEID - 32) |
7006 		      1 << (OP_WRITE - 32)
7007 	}
7008 };
7009 
7010 /*
7011  * Select the state protection mode for client `clp' given the server results
7012  * from exchange_id in `sp'.
7013  *
7014  * Returns 0 on success, negative errno otherwise.
7015  */
7016 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7017 				 struct nfs41_state_protection *sp)
7018 {
7019 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7020 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7021 		      1 << (OP_EXCHANGE_ID - 32) |
7022 		      1 << (OP_CREATE_SESSION - 32) |
7023 		      1 << (OP_DESTROY_SESSION - 32) |
7024 		      1 << (OP_DESTROY_CLIENTID - 32)
7025 	};
7026 	unsigned int i;
7027 
7028 	if (sp->how == SP4_MACH_CRED) {
7029 		/* Print state protect result */
7030 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7031 		for (i = 0; i <= LAST_NFS4_OP; i++) {
7032 			if (test_bit(i, sp->enforce.u.longs))
7033 				dfprintk(MOUNT, "  enforce op %d\n", i);
7034 			if (test_bit(i, sp->allow.u.longs))
7035 				dfprintk(MOUNT, "  allow op %d\n", i);
7036 		}
7037 
7038 		/* make sure nothing is on enforce list that isn't supported */
7039 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7040 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7041 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7042 				return -EINVAL;
7043 			}
7044 		}
7045 
7046 		/*
7047 		 * Minimal mode - state operations are allowed to use machine
7048 		 * credential.  Note this already happens by default, so the
7049 		 * client doesn't have to do anything more than the negotiation.
7050 		 *
7051 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
7052 		 *       we're already using the machine cred for exchange_id
7053 		 *       and will never use a different cred.
7054 		 */
7055 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7056 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7057 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7058 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7059 			dfprintk(MOUNT, "sp4_mach_cred:\n");
7060 			dfprintk(MOUNT, "  minimal mode enabled\n");
7061 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7062 		} else {
7063 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7064 			return -EINVAL;
7065 		}
7066 
7067 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7068 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7069 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7070 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
7071 			dfprintk(MOUNT, "  cleanup mode enabled\n");
7072 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7073 		}
7074 
7075 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7076 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7077 			set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7078 				&clp->cl_sp4_flags);
7079 		}
7080 
7081 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7082 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7083 			dfprintk(MOUNT, "  secinfo mode enabled\n");
7084 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7085 		}
7086 
7087 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7088 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7089 			dfprintk(MOUNT, "  stateid mode enabled\n");
7090 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7091 		}
7092 
7093 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7094 			dfprintk(MOUNT, "  write mode enabled\n");
7095 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7096 		}
7097 
7098 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7099 			dfprintk(MOUNT, "  commit mode enabled\n");
7100 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7101 		}
7102 	}
7103 
7104 	return 0;
7105 }
7106 
7107 /*
7108  * _nfs4_proc_exchange_id()
7109  *
7110  * Wrapper for EXCHANGE_ID operation.
7111  */
7112 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7113 	u32 sp4_how)
7114 {
7115 	nfs4_verifier verifier;
7116 	struct nfs41_exchange_id_args args = {
7117 		.verifier = &verifier,
7118 		.client = clp,
7119 #ifdef CONFIG_NFS_V4_1_MIGRATION
7120 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7121 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
7122 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7123 #else
7124 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7125 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
7126 #endif
7127 	};
7128 	struct nfs41_exchange_id_res res = {
7129 		0
7130 	};
7131 	int status;
7132 	struct rpc_message msg = {
7133 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7134 		.rpc_argp = &args,
7135 		.rpc_resp = &res,
7136 		.rpc_cred = cred,
7137 	};
7138 
7139 	nfs4_init_boot_verifier(clp, &verifier);
7140 
7141 	status = nfs4_init_uniform_client_string(clp);
7142 	if (status)
7143 		goto out;
7144 
7145 	dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7146 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
7147 		clp->cl_owner_id);
7148 
7149 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7150 					GFP_NOFS);
7151 	if (unlikely(res.server_owner == NULL)) {
7152 		status = -ENOMEM;
7153 		goto out;
7154 	}
7155 
7156 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7157 					GFP_NOFS);
7158 	if (unlikely(res.server_scope == NULL)) {
7159 		status = -ENOMEM;
7160 		goto out_server_owner;
7161 	}
7162 
7163 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7164 	if (unlikely(res.impl_id == NULL)) {
7165 		status = -ENOMEM;
7166 		goto out_server_scope;
7167 	}
7168 
7169 	switch (sp4_how) {
7170 	case SP4_NONE:
7171 		args.state_protect.how = SP4_NONE;
7172 		break;
7173 
7174 	case SP4_MACH_CRED:
7175 		args.state_protect = nfs4_sp4_mach_cred_request;
7176 		break;
7177 
7178 	default:
7179 		/* unsupported! */
7180 		WARN_ON_ONCE(1);
7181 		status = -EINVAL;
7182 		goto out_impl_id;
7183 	}
7184 
7185 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7186 	trace_nfs4_exchange_id(clp, status);
7187 	if (status == 0)
7188 		status = nfs4_check_cl_exchange_flags(res.flags);
7189 
7190 	if (status == 0)
7191 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
7192 
7193 	if (status == 0) {
7194 		clp->cl_clientid = res.clientid;
7195 		clp->cl_exchange_flags = res.flags;
7196 		/* Client ID is not confirmed */
7197 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7198 			clear_bit(NFS4_SESSION_ESTABLISHED,
7199 					&clp->cl_session->session_state);
7200 			clp->cl_seqid = res.seqid;
7201 		}
7202 
7203 		kfree(clp->cl_serverowner);
7204 		clp->cl_serverowner = res.server_owner;
7205 		res.server_owner = NULL;
7206 
7207 		/* use the most recent implementation id */
7208 		kfree(clp->cl_implid);
7209 		clp->cl_implid = res.impl_id;
7210 		res.impl_id = NULL;
7211 
7212 		if (clp->cl_serverscope != NULL &&
7213 		    !nfs41_same_server_scope(clp->cl_serverscope,
7214 					     res.server_scope)) {
7215 			dprintk("%s: server_scope mismatch detected\n",
7216 				__func__);
7217 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7218 			kfree(clp->cl_serverscope);
7219 			clp->cl_serverscope = NULL;
7220 		}
7221 
7222 		if (clp->cl_serverscope == NULL) {
7223 			clp->cl_serverscope = res.server_scope;
7224 			res.server_scope = NULL;
7225 		}
7226 	}
7227 
7228 out_impl_id:
7229 	kfree(res.impl_id);
7230 out_server_scope:
7231 	kfree(res.server_scope);
7232 out_server_owner:
7233 	kfree(res.server_owner);
7234 out:
7235 	if (clp->cl_implid != NULL)
7236 		dprintk("NFS reply exchange_id: Server Implementation ID: "
7237 			"domain: %s, name: %s, date: %llu,%u\n",
7238 			clp->cl_implid->domain, clp->cl_implid->name,
7239 			clp->cl_implid->date.seconds,
7240 			clp->cl_implid->date.nseconds);
7241 	dprintk("NFS reply exchange_id: %d\n", status);
7242 	return status;
7243 }
7244 
7245 /*
7246  * nfs4_proc_exchange_id()
7247  *
7248  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7249  *
7250  * Since the clientid has expired, all compounds using sessions
7251  * associated with the stale clientid will be returning
7252  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7253  * be in some phase of session reset.
7254  *
7255  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7256  */
7257 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7258 {
7259 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7260 	int status;
7261 
7262 	/* try SP4_MACH_CRED if krb5i/p	*/
7263 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
7264 	    authflavor == RPC_AUTH_GSS_KRB5P) {
7265 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7266 		if (!status)
7267 			return 0;
7268 	}
7269 
7270 	/* try SP4_NONE */
7271 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7272 }
7273 
7274 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7275 		struct rpc_cred *cred)
7276 {
7277 	struct rpc_message msg = {
7278 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7279 		.rpc_argp = clp,
7280 		.rpc_cred = cred,
7281 	};
7282 	int status;
7283 
7284 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7285 	trace_nfs4_destroy_clientid(clp, status);
7286 	if (status)
7287 		dprintk("NFS: Got error %d from the server %s on "
7288 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
7289 	return status;
7290 }
7291 
7292 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7293 		struct rpc_cred *cred)
7294 {
7295 	unsigned int loop;
7296 	int ret;
7297 
7298 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7299 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7300 		switch (ret) {
7301 		case -NFS4ERR_DELAY:
7302 		case -NFS4ERR_CLIENTID_BUSY:
7303 			ssleep(1);
7304 			break;
7305 		default:
7306 			return ret;
7307 		}
7308 	}
7309 	return 0;
7310 }
7311 
7312 int nfs4_destroy_clientid(struct nfs_client *clp)
7313 {
7314 	struct rpc_cred *cred;
7315 	int ret = 0;
7316 
7317 	if (clp->cl_mvops->minor_version < 1)
7318 		goto out;
7319 	if (clp->cl_exchange_flags == 0)
7320 		goto out;
7321 	if (clp->cl_preserve_clid)
7322 		goto out;
7323 	cred = nfs4_get_clid_cred(clp);
7324 	ret = nfs4_proc_destroy_clientid(clp, cred);
7325 	if (cred)
7326 		put_rpccred(cred);
7327 	switch (ret) {
7328 	case 0:
7329 	case -NFS4ERR_STALE_CLIENTID:
7330 		clp->cl_exchange_flags = 0;
7331 	}
7332 out:
7333 	return ret;
7334 }
7335 
7336 struct nfs4_get_lease_time_data {
7337 	struct nfs4_get_lease_time_args *args;
7338 	struct nfs4_get_lease_time_res *res;
7339 	struct nfs_client *clp;
7340 };
7341 
7342 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7343 					void *calldata)
7344 {
7345 	struct nfs4_get_lease_time_data *data =
7346 			(struct nfs4_get_lease_time_data *)calldata;
7347 
7348 	dprintk("--> %s\n", __func__);
7349 	/* just setup sequence, do not trigger session recovery
7350 	   since we're invoked within one */
7351 	nfs41_setup_sequence(data->clp->cl_session,
7352 			&data->args->la_seq_args,
7353 			&data->res->lr_seq_res,
7354 			task);
7355 	dprintk("<-- %s\n", __func__);
7356 }
7357 
7358 /*
7359  * Called from nfs4_state_manager thread for session setup, so don't recover
7360  * from sequence operation or clientid errors.
7361  */
7362 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7363 {
7364 	struct nfs4_get_lease_time_data *data =
7365 			(struct nfs4_get_lease_time_data *)calldata;
7366 
7367 	dprintk("--> %s\n", __func__);
7368 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7369 		return;
7370 	switch (task->tk_status) {
7371 	case -NFS4ERR_DELAY:
7372 	case -NFS4ERR_GRACE:
7373 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7374 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7375 		task->tk_status = 0;
7376 		/* fall through */
7377 	case -NFS4ERR_RETRY_UNCACHED_REP:
7378 		rpc_restart_call_prepare(task);
7379 		return;
7380 	}
7381 	dprintk("<-- %s\n", __func__);
7382 }
7383 
7384 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7385 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7386 	.rpc_call_done = nfs4_get_lease_time_done,
7387 };
7388 
7389 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7390 {
7391 	struct rpc_task *task;
7392 	struct nfs4_get_lease_time_args args;
7393 	struct nfs4_get_lease_time_res res = {
7394 		.lr_fsinfo = fsinfo,
7395 	};
7396 	struct nfs4_get_lease_time_data data = {
7397 		.args = &args,
7398 		.res = &res,
7399 		.clp = clp,
7400 	};
7401 	struct rpc_message msg = {
7402 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7403 		.rpc_argp = &args,
7404 		.rpc_resp = &res,
7405 	};
7406 	struct rpc_task_setup task_setup = {
7407 		.rpc_client = clp->cl_rpcclient,
7408 		.rpc_message = &msg,
7409 		.callback_ops = &nfs4_get_lease_time_ops,
7410 		.callback_data = &data,
7411 		.flags = RPC_TASK_TIMEOUT,
7412 	};
7413 	int status;
7414 
7415 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7416 	nfs4_set_sequence_privileged(&args.la_seq_args);
7417 	dprintk("--> %s\n", __func__);
7418 	task = rpc_run_task(&task_setup);
7419 
7420 	if (IS_ERR(task))
7421 		status = PTR_ERR(task);
7422 	else {
7423 		status = task->tk_status;
7424 		rpc_put_task(task);
7425 	}
7426 	dprintk("<-- %s return %d\n", __func__, status);
7427 
7428 	return status;
7429 }
7430 
7431 /*
7432  * Initialize the values to be used by the client in CREATE_SESSION
7433  * If nfs4_init_session set the fore channel request and response sizes,
7434  * use them.
7435  *
7436  * Set the back channel max_resp_sz_cached to zero to force the client to
7437  * always set csa_cachethis to FALSE because the current implementation
7438  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7439  */
7440 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7441 				    struct rpc_clnt *clnt)
7442 {
7443 	unsigned int max_rqst_sz, max_resp_sz;
7444 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7445 
7446 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7447 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7448 
7449 	/* Fore channel attributes */
7450 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7451 	args->fc_attrs.max_resp_sz = max_resp_sz;
7452 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7453 	args->fc_attrs.max_reqs = max_session_slots;
7454 
7455 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7456 		"max_ops=%u max_reqs=%u\n",
7457 		__func__,
7458 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7459 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7460 
7461 	/* Back channel attributes */
7462 	args->bc_attrs.max_rqst_sz = max_bc_payload;
7463 	args->bc_attrs.max_resp_sz = max_bc_payload;
7464 	args->bc_attrs.max_resp_sz_cached = 0;
7465 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7466 	args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7467 
7468 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7469 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7470 		__func__,
7471 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7472 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7473 		args->bc_attrs.max_reqs);
7474 }
7475 
7476 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7477 		struct nfs41_create_session_res *res)
7478 {
7479 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7480 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7481 
7482 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7483 		return -EINVAL;
7484 	/*
7485 	 * Our requested max_ops is the minimum we need; we're not
7486 	 * prepared to break up compounds into smaller pieces than that.
7487 	 * So, no point even trying to continue if the server won't
7488 	 * cooperate:
7489 	 */
7490 	if (rcvd->max_ops < sent->max_ops)
7491 		return -EINVAL;
7492 	if (rcvd->max_reqs == 0)
7493 		return -EINVAL;
7494 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7495 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7496 	return 0;
7497 }
7498 
7499 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7500 		struct nfs41_create_session_res *res)
7501 {
7502 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7503 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7504 
7505 	if (!(res->flags & SESSION4_BACK_CHAN))
7506 		goto out;
7507 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7508 		return -EINVAL;
7509 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7510 		return -EINVAL;
7511 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7512 		return -EINVAL;
7513 	/* These would render the backchannel useless: */
7514 	if (rcvd->max_ops != sent->max_ops)
7515 		return -EINVAL;
7516 	if (rcvd->max_reqs != sent->max_reqs)
7517 		return -EINVAL;
7518 out:
7519 	return 0;
7520 }
7521 
7522 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7523 				     struct nfs41_create_session_res *res)
7524 {
7525 	int ret;
7526 
7527 	ret = nfs4_verify_fore_channel_attrs(args, res);
7528 	if (ret)
7529 		return ret;
7530 	return nfs4_verify_back_channel_attrs(args, res);
7531 }
7532 
7533 static void nfs4_update_session(struct nfs4_session *session,
7534 		struct nfs41_create_session_res *res)
7535 {
7536 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7537 	/* Mark client id and session as being confirmed */
7538 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7539 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7540 	session->flags = res->flags;
7541 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7542 	if (res->flags & SESSION4_BACK_CHAN)
7543 		memcpy(&session->bc_attrs, &res->bc_attrs,
7544 				sizeof(session->bc_attrs));
7545 }
7546 
7547 static int _nfs4_proc_create_session(struct nfs_client *clp,
7548 		struct rpc_cred *cred)
7549 {
7550 	struct nfs4_session *session = clp->cl_session;
7551 	struct nfs41_create_session_args args = {
7552 		.client = clp,
7553 		.clientid = clp->cl_clientid,
7554 		.seqid = clp->cl_seqid,
7555 		.cb_program = NFS4_CALLBACK,
7556 	};
7557 	struct nfs41_create_session_res res;
7558 
7559 	struct rpc_message msg = {
7560 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7561 		.rpc_argp = &args,
7562 		.rpc_resp = &res,
7563 		.rpc_cred = cred,
7564 	};
7565 	int status;
7566 
7567 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7568 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7569 
7570 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7571 	trace_nfs4_create_session(clp, status);
7572 
7573 	switch (status) {
7574 	case -NFS4ERR_STALE_CLIENTID:
7575 	case -NFS4ERR_DELAY:
7576 	case -ETIMEDOUT:
7577 	case -EACCES:
7578 	case -EAGAIN:
7579 		goto out;
7580 	};
7581 
7582 	clp->cl_seqid++;
7583 	if (!status) {
7584 		/* Verify the session's negotiated channel_attrs values */
7585 		status = nfs4_verify_channel_attrs(&args, &res);
7586 		/* Increment the clientid slot sequence id */
7587 		if (status)
7588 			goto out;
7589 		nfs4_update_session(session, &res);
7590 	}
7591 out:
7592 	return status;
7593 }
7594 
7595 /*
7596  * Issues a CREATE_SESSION operation to the server.
7597  * It is the responsibility of the caller to verify the session is
7598  * expired before calling this routine.
7599  */
7600 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7601 {
7602 	int status;
7603 	unsigned *ptr;
7604 	struct nfs4_session *session = clp->cl_session;
7605 
7606 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7607 
7608 	status = _nfs4_proc_create_session(clp, cred);
7609 	if (status)
7610 		goto out;
7611 
7612 	/* Init or reset the session slot tables */
7613 	status = nfs4_setup_session_slot_tables(session);
7614 	dprintk("slot table setup returned %d\n", status);
7615 	if (status)
7616 		goto out;
7617 
7618 	ptr = (unsigned *)&session->sess_id.data[0];
7619 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7620 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7621 out:
7622 	dprintk("<-- %s\n", __func__);
7623 	return status;
7624 }
7625 
7626 /*
7627  * Issue the over-the-wire RPC DESTROY_SESSION.
7628  * The caller must serialize access to this routine.
7629  */
7630 int nfs4_proc_destroy_session(struct nfs4_session *session,
7631 		struct rpc_cred *cred)
7632 {
7633 	struct rpc_message msg = {
7634 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7635 		.rpc_argp = session,
7636 		.rpc_cred = cred,
7637 	};
7638 	int status = 0;
7639 
7640 	dprintk("--> nfs4_proc_destroy_session\n");
7641 
7642 	/* session is still being setup */
7643 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7644 		return 0;
7645 
7646 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7647 	trace_nfs4_destroy_session(session->clp, status);
7648 
7649 	if (status)
7650 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7651 			"Session has been destroyed regardless...\n", status);
7652 
7653 	dprintk("<-- nfs4_proc_destroy_session\n");
7654 	return status;
7655 }
7656 
7657 /*
7658  * Renew the cl_session lease.
7659  */
7660 struct nfs4_sequence_data {
7661 	struct nfs_client *clp;
7662 	struct nfs4_sequence_args args;
7663 	struct nfs4_sequence_res res;
7664 };
7665 
7666 static void nfs41_sequence_release(void *data)
7667 {
7668 	struct nfs4_sequence_data *calldata = data;
7669 	struct nfs_client *clp = calldata->clp;
7670 
7671 	if (atomic_read(&clp->cl_count) > 1)
7672 		nfs4_schedule_state_renewal(clp);
7673 	nfs_put_client(clp);
7674 	kfree(calldata);
7675 }
7676 
7677 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7678 {
7679 	switch(task->tk_status) {
7680 	case -NFS4ERR_DELAY:
7681 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7682 		return -EAGAIN;
7683 	default:
7684 		nfs4_schedule_lease_recovery(clp);
7685 	}
7686 	return 0;
7687 }
7688 
7689 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7690 {
7691 	struct nfs4_sequence_data *calldata = data;
7692 	struct nfs_client *clp = calldata->clp;
7693 
7694 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7695 		return;
7696 
7697 	trace_nfs4_sequence(clp, task->tk_status);
7698 	if (task->tk_status < 0) {
7699 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7700 		if (atomic_read(&clp->cl_count) == 1)
7701 			goto out;
7702 
7703 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7704 			rpc_restart_call_prepare(task);
7705 			return;
7706 		}
7707 	}
7708 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7709 out:
7710 	dprintk("<-- %s\n", __func__);
7711 }
7712 
7713 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7714 {
7715 	struct nfs4_sequence_data *calldata = data;
7716 	struct nfs_client *clp = calldata->clp;
7717 	struct nfs4_sequence_args *args;
7718 	struct nfs4_sequence_res *res;
7719 
7720 	args = task->tk_msg.rpc_argp;
7721 	res = task->tk_msg.rpc_resp;
7722 
7723 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7724 }
7725 
7726 static const struct rpc_call_ops nfs41_sequence_ops = {
7727 	.rpc_call_done = nfs41_sequence_call_done,
7728 	.rpc_call_prepare = nfs41_sequence_prepare,
7729 	.rpc_release = nfs41_sequence_release,
7730 };
7731 
7732 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7733 		struct rpc_cred *cred,
7734 		bool is_privileged)
7735 {
7736 	struct nfs4_sequence_data *calldata;
7737 	struct rpc_message msg = {
7738 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7739 		.rpc_cred = cred,
7740 	};
7741 	struct rpc_task_setup task_setup_data = {
7742 		.rpc_client = clp->cl_rpcclient,
7743 		.rpc_message = &msg,
7744 		.callback_ops = &nfs41_sequence_ops,
7745 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7746 	};
7747 
7748 	if (!atomic_inc_not_zero(&clp->cl_count))
7749 		return ERR_PTR(-EIO);
7750 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7751 	if (calldata == NULL) {
7752 		nfs_put_client(clp);
7753 		return ERR_PTR(-ENOMEM);
7754 	}
7755 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7756 	if (is_privileged)
7757 		nfs4_set_sequence_privileged(&calldata->args);
7758 	msg.rpc_argp = &calldata->args;
7759 	msg.rpc_resp = &calldata->res;
7760 	calldata->clp = clp;
7761 	task_setup_data.callback_data = calldata;
7762 
7763 	return rpc_run_task(&task_setup_data);
7764 }
7765 
7766 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7767 {
7768 	struct rpc_task *task;
7769 	int ret = 0;
7770 
7771 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7772 		return -EAGAIN;
7773 	task = _nfs41_proc_sequence(clp, cred, false);
7774 	if (IS_ERR(task))
7775 		ret = PTR_ERR(task);
7776 	else
7777 		rpc_put_task_async(task);
7778 	dprintk("<-- %s status=%d\n", __func__, ret);
7779 	return ret;
7780 }
7781 
7782 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7783 {
7784 	struct rpc_task *task;
7785 	int ret;
7786 
7787 	task = _nfs41_proc_sequence(clp, cred, true);
7788 	if (IS_ERR(task)) {
7789 		ret = PTR_ERR(task);
7790 		goto out;
7791 	}
7792 	ret = rpc_wait_for_completion_task(task);
7793 	if (!ret)
7794 		ret = task->tk_status;
7795 	rpc_put_task(task);
7796 out:
7797 	dprintk("<-- %s status=%d\n", __func__, ret);
7798 	return ret;
7799 }
7800 
7801 struct nfs4_reclaim_complete_data {
7802 	struct nfs_client *clp;
7803 	struct nfs41_reclaim_complete_args arg;
7804 	struct nfs41_reclaim_complete_res res;
7805 };
7806 
7807 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7808 {
7809 	struct nfs4_reclaim_complete_data *calldata = data;
7810 
7811 	nfs41_setup_sequence(calldata->clp->cl_session,
7812 			&calldata->arg.seq_args,
7813 			&calldata->res.seq_res,
7814 			task);
7815 }
7816 
7817 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7818 {
7819 	switch(task->tk_status) {
7820 	case 0:
7821 	case -NFS4ERR_COMPLETE_ALREADY:
7822 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7823 		break;
7824 	case -NFS4ERR_DELAY:
7825 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7826 		/* fall through */
7827 	case -NFS4ERR_RETRY_UNCACHED_REP:
7828 		return -EAGAIN;
7829 	default:
7830 		nfs4_schedule_lease_recovery(clp);
7831 	}
7832 	return 0;
7833 }
7834 
7835 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7836 {
7837 	struct nfs4_reclaim_complete_data *calldata = data;
7838 	struct nfs_client *clp = calldata->clp;
7839 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7840 
7841 	dprintk("--> %s\n", __func__);
7842 	if (!nfs41_sequence_done(task, res))
7843 		return;
7844 
7845 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7846 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7847 		rpc_restart_call_prepare(task);
7848 		return;
7849 	}
7850 	dprintk("<-- %s\n", __func__);
7851 }
7852 
7853 static void nfs4_free_reclaim_complete_data(void *data)
7854 {
7855 	struct nfs4_reclaim_complete_data *calldata = data;
7856 
7857 	kfree(calldata);
7858 }
7859 
7860 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7861 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7862 	.rpc_call_done = nfs4_reclaim_complete_done,
7863 	.rpc_release = nfs4_free_reclaim_complete_data,
7864 };
7865 
7866 /*
7867  * Issue a global reclaim complete.
7868  */
7869 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7870 		struct rpc_cred *cred)
7871 {
7872 	struct nfs4_reclaim_complete_data *calldata;
7873 	struct rpc_task *task;
7874 	struct rpc_message msg = {
7875 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7876 		.rpc_cred = cred,
7877 	};
7878 	struct rpc_task_setup task_setup_data = {
7879 		.rpc_client = clp->cl_rpcclient,
7880 		.rpc_message = &msg,
7881 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7882 		.flags = RPC_TASK_ASYNC,
7883 	};
7884 	int status = -ENOMEM;
7885 
7886 	dprintk("--> %s\n", __func__);
7887 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7888 	if (calldata == NULL)
7889 		goto out;
7890 	calldata->clp = clp;
7891 	calldata->arg.one_fs = 0;
7892 
7893 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7894 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7895 	msg.rpc_argp = &calldata->arg;
7896 	msg.rpc_resp = &calldata->res;
7897 	task_setup_data.callback_data = calldata;
7898 	task = rpc_run_task(&task_setup_data);
7899 	if (IS_ERR(task)) {
7900 		status = PTR_ERR(task);
7901 		goto out;
7902 	}
7903 	status = nfs4_wait_for_completion_rpc_task(task);
7904 	if (status == 0)
7905 		status = task->tk_status;
7906 	rpc_put_task(task);
7907 	return 0;
7908 out:
7909 	dprintk("<-- %s status=%d\n", __func__, status);
7910 	return status;
7911 }
7912 
7913 static void
7914 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7915 {
7916 	struct nfs4_layoutget *lgp = calldata;
7917 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7918 	struct nfs4_session *session = nfs4_get_session(server);
7919 
7920 	dprintk("--> %s\n", __func__);
7921 	nfs41_setup_sequence(session, &lgp->args.seq_args,
7922 				&lgp->res.seq_res, task);
7923 	dprintk("<-- %s\n", __func__);
7924 }
7925 
7926 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7927 {
7928 	struct nfs4_layoutget *lgp = calldata;
7929 
7930 	dprintk("--> %s\n", __func__);
7931 	nfs41_sequence_process(task, &lgp->res.seq_res);
7932 	dprintk("<-- %s\n", __func__);
7933 }
7934 
7935 static int
7936 nfs4_layoutget_handle_exception(struct rpc_task *task,
7937 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
7938 {
7939 	struct inode *inode = lgp->args.inode;
7940 	struct nfs_server *server = NFS_SERVER(inode);
7941 	struct pnfs_layout_hdr *lo;
7942 	int nfs4err = task->tk_status;
7943 	int err, status = 0;
7944 	LIST_HEAD(head);
7945 
7946 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7947 
7948 	switch (nfs4err) {
7949 	case 0:
7950 		goto out;
7951 
7952 	/*
7953 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7954 	 * on the file. set tk_status to -ENODATA to tell upper layer to
7955 	 * retry go inband.
7956 	 */
7957 	case -NFS4ERR_LAYOUTUNAVAILABLE:
7958 		status = -ENODATA;
7959 		goto out;
7960 	/*
7961 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7962 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7963 	 */
7964 	case -NFS4ERR_BADLAYOUT:
7965 		status = -EOVERFLOW;
7966 		goto out;
7967 	/*
7968 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7969 	 * (or clients) writing to the same RAID stripe except when
7970 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
7971 	 *
7972 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
7973 	 * while, and then eventually give up.
7974 	 */
7975 	case -NFS4ERR_LAYOUTTRYLATER:
7976 		if (lgp->args.minlength == 0) {
7977 			status = -EOVERFLOW;
7978 			goto out;
7979 		}
7980 		status = -EBUSY;
7981 		break;
7982 	case -NFS4ERR_RECALLCONFLICT:
7983 		status = -ERECALLCONFLICT;
7984 		break;
7985 	case -NFS4ERR_EXPIRED:
7986 	case -NFS4ERR_BAD_STATEID:
7987 		exception->timeout = 0;
7988 		spin_lock(&inode->i_lock);
7989 		lo = NFS_I(inode)->layout;
7990 		/* If the open stateid was bad, then recover it. */
7991 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
7992 		    nfs4_stateid_match_other(&lgp->args.stateid,
7993 					&lgp->args.ctx->state->stateid)) {
7994 			spin_unlock(&inode->i_lock);
7995 			exception->state = lgp->args.ctx->state;
7996 			break;
7997 		}
7998 
7999 		/*
8000 		 * Mark the bad layout state as invalid, then retry
8001 		 */
8002 		pnfs_mark_layout_stateid_invalid(lo, &head);
8003 		spin_unlock(&inode->i_lock);
8004 		pnfs_free_lseg_list(&head);
8005 		status = -EAGAIN;
8006 		goto out;
8007 	}
8008 
8009 	err = nfs4_handle_exception(server, nfs4err, exception);
8010 	if (!status) {
8011 		if (exception->retry)
8012 			status = -EAGAIN;
8013 		else
8014 			status = err;
8015 	}
8016 out:
8017 	dprintk("<-- %s\n", __func__);
8018 	return status;
8019 }
8020 
8021 static size_t max_response_pages(struct nfs_server *server)
8022 {
8023 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8024 	return nfs_page_array_len(0, max_resp_sz);
8025 }
8026 
8027 static void nfs4_free_pages(struct page **pages, size_t size)
8028 {
8029 	int i;
8030 
8031 	if (!pages)
8032 		return;
8033 
8034 	for (i = 0; i < size; i++) {
8035 		if (!pages[i])
8036 			break;
8037 		__free_page(pages[i]);
8038 	}
8039 	kfree(pages);
8040 }
8041 
8042 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8043 {
8044 	struct page **pages;
8045 	int i;
8046 
8047 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8048 	if (!pages) {
8049 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8050 		return NULL;
8051 	}
8052 
8053 	for (i = 0; i < size; i++) {
8054 		pages[i] = alloc_page(gfp_flags);
8055 		if (!pages[i]) {
8056 			dprintk("%s: failed to allocate page\n", __func__);
8057 			nfs4_free_pages(pages, size);
8058 			return NULL;
8059 		}
8060 	}
8061 
8062 	return pages;
8063 }
8064 
8065 static void nfs4_layoutget_release(void *calldata)
8066 {
8067 	struct nfs4_layoutget *lgp = calldata;
8068 	struct inode *inode = lgp->args.inode;
8069 	struct nfs_server *server = NFS_SERVER(inode);
8070 	size_t max_pages = max_response_pages(server);
8071 
8072 	dprintk("--> %s\n", __func__);
8073 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
8074 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
8075 	put_nfs_open_context(lgp->args.ctx);
8076 	kfree(calldata);
8077 	dprintk("<-- %s\n", __func__);
8078 }
8079 
8080 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8081 	.rpc_call_prepare = nfs4_layoutget_prepare,
8082 	.rpc_call_done = nfs4_layoutget_done,
8083 	.rpc_release = nfs4_layoutget_release,
8084 };
8085 
8086 struct pnfs_layout_segment *
8087 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8088 {
8089 	struct inode *inode = lgp->args.inode;
8090 	struct nfs_server *server = NFS_SERVER(inode);
8091 	size_t max_pages = max_response_pages(server);
8092 	struct rpc_task *task;
8093 	struct rpc_message msg = {
8094 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8095 		.rpc_argp = &lgp->args,
8096 		.rpc_resp = &lgp->res,
8097 		.rpc_cred = lgp->cred,
8098 	};
8099 	struct rpc_task_setup task_setup_data = {
8100 		.rpc_client = server->client,
8101 		.rpc_message = &msg,
8102 		.callback_ops = &nfs4_layoutget_call_ops,
8103 		.callback_data = lgp,
8104 		.flags = RPC_TASK_ASYNC,
8105 	};
8106 	struct pnfs_layout_segment *lseg = NULL;
8107 	struct nfs4_exception exception = {
8108 		.inode = inode,
8109 		.timeout = *timeout,
8110 	};
8111 	int status = 0;
8112 
8113 	dprintk("--> %s\n", __func__);
8114 
8115 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8116 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
8117 
8118 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8119 	if (!lgp->args.layout.pages) {
8120 		nfs4_layoutget_release(lgp);
8121 		return ERR_PTR(-ENOMEM);
8122 	}
8123 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8124 
8125 	lgp->res.layoutp = &lgp->args.layout;
8126 	lgp->res.seq_res.sr_slot = NULL;
8127 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8128 
8129 	task = rpc_run_task(&task_setup_data);
8130 	if (IS_ERR(task))
8131 		return ERR_CAST(task);
8132 	status = nfs4_wait_for_completion_rpc_task(task);
8133 	if (status == 0) {
8134 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8135 		*timeout = exception.timeout;
8136 	}
8137 
8138 	trace_nfs4_layoutget(lgp->args.ctx,
8139 			&lgp->args.range,
8140 			&lgp->res.range,
8141 			&lgp->res.stateid,
8142 			status);
8143 
8144 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8145 	if (status == 0 && lgp->res.layoutp->len)
8146 		lseg = pnfs_layout_process(lgp);
8147 	nfs4_sequence_free_slot(&lgp->res.seq_res);
8148 	rpc_put_task(task);
8149 	dprintk("<-- %s status=%d\n", __func__, status);
8150 	if (status)
8151 		return ERR_PTR(status);
8152 	return lseg;
8153 }
8154 
8155 static void
8156 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8157 {
8158 	struct nfs4_layoutreturn *lrp = calldata;
8159 
8160 	dprintk("--> %s\n", __func__);
8161 	nfs41_setup_sequence(lrp->clp->cl_session,
8162 			&lrp->args.seq_args,
8163 			&lrp->res.seq_res,
8164 			task);
8165 }
8166 
8167 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8168 {
8169 	struct nfs4_layoutreturn *lrp = calldata;
8170 	struct nfs_server *server;
8171 
8172 	dprintk("--> %s\n", __func__);
8173 
8174 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8175 		return;
8176 
8177 	server = NFS_SERVER(lrp->args.inode);
8178 	switch (task->tk_status) {
8179 	default:
8180 		task->tk_status = 0;
8181 	case 0:
8182 		break;
8183 	case -NFS4ERR_DELAY:
8184 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8185 			break;
8186 		nfs4_sequence_free_slot(&lrp->res.seq_res);
8187 		rpc_restart_call_prepare(task);
8188 		return;
8189 	}
8190 	dprintk("<-- %s\n", __func__);
8191 }
8192 
8193 static void nfs4_layoutreturn_release(void *calldata)
8194 {
8195 	struct nfs4_layoutreturn *lrp = calldata;
8196 	struct pnfs_layout_hdr *lo = lrp->args.layout;
8197 	LIST_HEAD(freeme);
8198 
8199 	dprintk("--> %s\n", __func__);
8200 	spin_lock(&lo->plh_inode->i_lock);
8201 	if (lrp->res.lrs_present) {
8202 		pnfs_mark_matching_lsegs_invalid(lo, &freeme,
8203 				&lrp->args.range,
8204 				be32_to_cpu(lrp->args.stateid.seqid));
8205 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8206 	} else
8207 		pnfs_mark_layout_stateid_invalid(lo, &freeme);
8208 	pnfs_clear_layoutreturn_waitbit(lo);
8209 	spin_unlock(&lo->plh_inode->i_lock);
8210 	nfs4_sequence_free_slot(&lrp->res.seq_res);
8211 	pnfs_free_lseg_list(&freeme);
8212 	pnfs_put_layout_hdr(lrp->args.layout);
8213 	nfs_iput_and_deactive(lrp->inode);
8214 	kfree(calldata);
8215 	dprintk("<-- %s\n", __func__);
8216 }
8217 
8218 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8219 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
8220 	.rpc_call_done = nfs4_layoutreturn_done,
8221 	.rpc_release = nfs4_layoutreturn_release,
8222 };
8223 
8224 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8225 {
8226 	struct rpc_task *task;
8227 	struct rpc_message msg = {
8228 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8229 		.rpc_argp = &lrp->args,
8230 		.rpc_resp = &lrp->res,
8231 		.rpc_cred = lrp->cred,
8232 	};
8233 	struct rpc_task_setup task_setup_data = {
8234 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
8235 		.rpc_message = &msg,
8236 		.callback_ops = &nfs4_layoutreturn_call_ops,
8237 		.callback_data = lrp,
8238 	};
8239 	int status = 0;
8240 
8241 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8242 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8243 			&task_setup_data.rpc_client, &msg);
8244 
8245 	dprintk("--> %s\n", __func__);
8246 	if (!sync) {
8247 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8248 		if (!lrp->inode) {
8249 			nfs4_layoutreturn_release(lrp);
8250 			return -EAGAIN;
8251 		}
8252 		task_setup_data.flags |= RPC_TASK_ASYNC;
8253 	}
8254 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8255 	task = rpc_run_task(&task_setup_data);
8256 	if (IS_ERR(task))
8257 		return PTR_ERR(task);
8258 	if (sync)
8259 		status = task->tk_status;
8260 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8261 	dprintk("<-- %s status=%d\n", __func__, status);
8262 	rpc_put_task(task);
8263 	return status;
8264 }
8265 
8266 static int
8267 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8268 		struct pnfs_device *pdev,
8269 		struct rpc_cred *cred)
8270 {
8271 	struct nfs4_getdeviceinfo_args args = {
8272 		.pdev = pdev,
8273 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
8274 			NOTIFY_DEVICEID4_DELETE,
8275 	};
8276 	struct nfs4_getdeviceinfo_res res = {
8277 		.pdev = pdev,
8278 	};
8279 	struct rpc_message msg = {
8280 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8281 		.rpc_argp = &args,
8282 		.rpc_resp = &res,
8283 		.rpc_cred = cred,
8284 	};
8285 	int status;
8286 
8287 	dprintk("--> %s\n", __func__);
8288 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8289 	if (res.notification & ~args.notify_types)
8290 		dprintk("%s: unsupported notification\n", __func__);
8291 	if (res.notification != args.notify_types)
8292 		pdev->nocache = 1;
8293 
8294 	dprintk("<-- %s status=%d\n", __func__, status);
8295 
8296 	return status;
8297 }
8298 
8299 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8300 		struct pnfs_device *pdev,
8301 		struct rpc_cred *cred)
8302 {
8303 	struct nfs4_exception exception = { };
8304 	int err;
8305 
8306 	do {
8307 		err = nfs4_handle_exception(server,
8308 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
8309 					&exception);
8310 	} while (exception.retry);
8311 	return err;
8312 }
8313 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8314 
8315 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8316 {
8317 	struct nfs4_layoutcommit_data *data = calldata;
8318 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8319 	struct nfs4_session *session = nfs4_get_session(server);
8320 
8321 	nfs41_setup_sequence(session,
8322 			&data->args.seq_args,
8323 			&data->res.seq_res,
8324 			task);
8325 }
8326 
8327 static void
8328 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8329 {
8330 	struct nfs4_layoutcommit_data *data = calldata;
8331 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8332 
8333 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8334 		return;
8335 
8336 	switch (task->tk_status) { /* Just ignore these failures */
8337 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8338 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8339 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8340 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8341 		task->tk_status = 0;
8342 	case 0:
8343 		break;
8344 	default:
8345 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8346 			rpc_restart_call_prepare(task);
8347 			return;
8348 		}
8349 	}
8350 }
8351 
8352 static void nfs4_layoutcommit_release(void *calldata)
8353 {
8354 	struct nfs4_layoutcommit_data *data = calldata;
8355 
8356 	pnfs_cleanup_layoutcommit(data);
8357 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8358 					   data->res.fattr);
8359 	put_rpccred(data->cred);
8360 	nfs_iput_and_deactive(data->inode);
8361 	kfree(data);
8362 }
8363 
8364 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8365 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8366 	.rpc_call_done = nfs4_layoutcommit_done,
8367 	.rpc_release = nfs4_layoutcommit_release,
8368 };
8369 
8370 int
8371 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8372 {
8373 	struct rpc_message msg = {
8374 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8375 		.rpc_argp = &data->args,
8376 		.rpc_resp = &data->res,
8377 		.rpc_cred = data->cred,
8378 	};
8379 	struct rpc_task_setup task_setup_data = {
8380 		.task = &data->task,
8381 		.rpc_client = NFS_CLIENT(data->args.inode),
8382 		.rpc_message = &msg,
8383 		.callback_ops = &nfs4_layoutcommit_ops,
8384 		.callback_data = data,
8385 	};
8386 	struct rpc_task *task;
8387 	int status = 0;
8388 
8389 	dprintk("NFS: initiating layoutcommit call. sync %d "
8390 		"lbw: %llu inode %lu\n", sync,
8391 		data->args.lastbytewritten,
8392 		data->args.inode->i_ino);
8393 
8394 	if (!sync) {
8395 		data->inode = nfs_igrab_and_active(data->args.inode);
8396 		if (data->inode == NULL) {
8397 			nfs4_layoutcommit_release(data);
8398 			return -EAGAIN;
8399 		}
8400 		task_setup_data.flags = RPC_TASK_ASYNC;
8401 	}
8402 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8403 	task = rpc_run_task(&task_setup_data);
8404 	if (IS_ERR(task))
8405 		return PTR_ERR(task);
8406 	if (sync)
8407 		status = task->tk_status;
8408 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8409 	dprintk("%s: status %d\n", __func__, status);
8410 	rpc_put_task(task);
8411 	return status;
8412 }
8413 
8414 /**
8415  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8416  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8417  */
8418 static int
8419 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8420 		    struct nfs_fsinfo *info,
8421 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8422 {
8423 	struct nfs41_secinfo_no_name_args args = {
8424 		.style = SECINFO_STYLE_CURRENT_FH,
8425 	};
8426 	struct nfs4_secinfo_res res = {
8427 		.flavors = flavors,
8428 	};
8429 	struct rpc_message msg = {
8430 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8431 		.rpc_argp = &args,
8432 		.rpc_resp = &res,
8433 	};
8434 	struct rpc_clnt *clnt = server->client;
8435 	struct rpc_cred *cred = NULL;
8436 	int status;
8437 
8438 	if (use_integrity) {
8439 		clnt = server->nfs_client->cl_rpcclient;
8440 		cred = nfs4_get_clid_cred(server->nfs_client);
8441 		msg.rpc_cred = cred;
8442 	}
8443 
8444 	dprintk("--> %s\n", __func__);
8445 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8446 				&res.seq_res, 0);
8447 	dprintk("<-- %s status=%d\n", __func__, status);
8448 
8449 	if (cred)
8450 		put_rpccred(cred);
8451 
8452 	return status;
8453 }
8454 
8455 static int
8456 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8457 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8458 {
8459 	struct nfs4_exception exception = { };
8460 	int err;
8461 	do {
8462 		/* first try using integrity protection */
8463 		err = -NFS4ERR_WRONGSEC;
8464 
8465 		/* try to use integrity protection with machine cred */
8466 		if (_nfs4_is_integrity_protected(server->nfs_client))
8467 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8468 							  flavors, true);
8469 
8470 		/*
8471 		 * if unable to use integrity protection, or SECINFO with
8472 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8473 		 * disallowed by spec, but exists in deployed servers) use
8474 		 * the current filesystem's rpc_client and the user cred.
8475 		 */
8476 		if (err == -NFS4ERR_WRONGSEC)
8477 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8478 							  flavors, false);
8479 
8480 		switch (err) {
8481 		case 0:
8482 		case -NFS4ERR_WRONGSEC:
8483 		case -ENOTSUPP:
8484 			goto out;
8485 		default:
8486 			err = nfs4_handle_exception(server, err, &exception);
8487 		}
8488 	} while (exception.retry);
8489 out:
8490 	return err;
8491 }
8492 
8493 static int
8494 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8495 		    struct nfs_fsinfo *info)
8496 {
8497 	int err;
8498 	struct page *page;
8499 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8500 	struct nfs4_secinfo_flavors *flavors;
8501 	struct nfs4_secinfo4 *secinfo;
8502 	int i;
8503 
8504 	page = alloc_page(GFP_KERNEL);
8505 	if (!page) {
8506 		err = -ENOMEM;
8507 		goto out;
8508 	}
8509 
8510 	flavors = page_address(page);
8511 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8512 
8513 	/*
8514 	 * Fall back on "guess and check" method if
8515 	 * the server doesn't support SECINFO_NO_NAME
8516 	 */
8517 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8518 		err = nfs4_find_root_sec(server, fhandle, info);
8519 		goto out_freepage;
8520 	}
8521 	if (err)
8522 		goto out_freepage;
8523 
8524 	for (i = 0; i < flavors->num_flavors; i++) {
8525 		secinfo = &flavors->flavors[i];
8526 
8527 		switch (secinfo->flavor) {
8528 		case RPC_AUTH_NULL:
8529 		case RPC_AUTH_UNIX:
8530 		case RPC_AUTH_GSS:
8531 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8532 					&secinfo->flavor_info);
8533 			break;
8534 		default:
8535 			flavor = RPC_AUTH_MAXFLAVOR;
8536 			break;
8537 		}
8538 
8539 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8540 			flavor = RPC_AUTH_MAXFLAVOR;
8541 
8542 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8543 			err = nfs4_lookup_root_sec(server, fhandle,
8544 						   info, flavor);
8545 			if (!err)
8546 				break;
8547 		}
8548 	}
8549 
8550 	if (flavor == RPC_AUTH_MAXFLAVOR)
8551 		err = -EPERM;
8552 
8553 out_freepage:
8554 	put_page(page);
8555 	if (err == -EACCES)
8556 		return -EPERM;
8557 out:
8558 	return err;
8559 }
8560 
8561 static int _nfs41_test_stateid(struct nfs_server *server,
8562 		nfs4_stateid *stateid,
8563 		struct rpc_cred *cred)
8564 {
8565 	int status;
8566 	struct nfs41_test_stateid_args args = {
8567 		.stateid = stateid,
8568 	};
8569 	struct nfs41_test_stateid_res res;
8570 	struct rpc_message msg = {
8571 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8572 		.rpc_argp = &args,
8573 		.rpc_resp = &res,
8574 		.rpc_cred = cred,
8575 	};
8576 	struct rpc_clnt *rpc_client = server->client;
8577 
8578 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8579 		&rpc_client, &msg);
8580 
8581 	dprintk("NFS call  test_stateid %p\n", stateid);
8582 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8583 	nfs4_set_sequence_privileged(&args.seq_args);
8584 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8585 			&args.seq_args, &res.seq_res);
8586 	if (status != NFS_OK) {
8587 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8588 		return status;
8589 	}
8590 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8591 	return -res.status;
8592 }
8593 
8594 /**
8595  * nfs41_test_stateid - perform a TEST_STATEID operation
8596  *
8597  * @server: server / transport on which to perform the operation
8598  * @stateid: state ID to test
8599  * @cred: credential
8600  *
8601  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8602  * Otherwise a negative NFS4ERR value is returned if the operation
8603  * failed or the state ID is not currently valid.
8604  */
8605 static int nfs41_test_stateid(struct nfs_server *server,
8606 		nfs4_stateid *stateid,
8607 		struct rpc_cred *cred)
8608 {
8609 	struct nfs4_exception exception = { };
8610 	int err;
8611 	do {
8612 		err = _nfs41_test_stateid(server, stateid, cred);
8613 		if (err != -NFS4ERR_DELAY)
8614 			break;
8615 		nfs4_handle_exception(server, err, &exception);
8616 	} while (exception.retry);
8617 	return err;
8618 }
8619 
8620 struct nfs_free_stateid_data {
8621 	struct nfs_server *server;
8622 	struct nfs41_free_stateid_args args;
8623 	struct nfs41_free_stateid_res res;
8624 };
8625 
8626 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8627 {
8628 	struct nfs_free_stateid_data *data = calldata;
8629 	nfs41_setup_sequence(nfs4_get_session(data->server),
8630 			&data->args.seq_args,
8631 			&data->res.seq_res,
8632 			task);
8633 }
8634 
8635 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8636 {
8637 	struct nfs_free_stateid_data *data = calldata;
8638 
8639 	nfs41_sequence_done(task, &data->res.seq_res);
8640 
8641 	switch (task->tk_status) {
8642 	case -NFS4ERR_DELAY:
8643 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8644 			rpc_restart_call_prepare(task);
8645 	}
8646 }
8647 
8648 static void nfs41_free_stateid_release(void *calldata)
8649 {
8650 	kfree(calldata);
8651 }
8652 
8653 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8654 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8655 	.rpc_call_done = nfs41_free_stateid_done,
8656 	.rpc_release = nfs41_free_stateid_release,
8657 };
8658 
8659 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8660 		nfs4_stateid *stateid,
8661 		struct rpc_cred *cred,
8662 		bool privileged)
8663 {
8664 	struct rpc_message msg = {
8665 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8666 		.rpc_cred = cred,
8667 	};
8668 	struct rpc_task_setup task_setup = {
8669 		.rpc_client = server->client,
8670 		.rpc_message = &msg,
8671 		.callback_ops = &nfs41_free_stateid_ops,
8672 		.flags = RPC_TASK_ASYNC,
8673 	};
8674 	struct nfs_free_stateid_data *data;
8675 
8676 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8677 		&task_setup.rpc_client, &msg);
8678 
8679 	dprintk("NFS call  free_stateid %p\n", stateid);
8680 	data = kmalloc(sizeof(*data), GFP_NOFS);
8681 	if (!data)
8682 		return ERR_PTR(-ENOMEM);
8683 	data->server = server;
8684 	nfs4_stateid_copy(&data->args.stateid, stateid);
8685 
8686 	task_setup.callback_data = data;
8687 
8688 	msg.rpc_argp = &data->args;
8689 	msg.rpc_resp = &data->res;
8690 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8691 	if (privileged)
8692 		nfs4_set_sequence_privileged(&data->args.seq_args);
8693 
8694 	return rpc_run_task(&task_setup);
8695 }
8696 
8697 /**
8698  * nfs41_free_stateid - perform a FREE_STATEID operation
8699  *
8700  * @server: server / transport on which to perform the operation
8701  * @stateid: state ID to release
8702  * @cred: credential
8703  *
8704  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8705  * negative NFS4ERR value is returned.
8706  */
8707 static int nfs41_free_stateid(struct nfs_server *server,
8708 		nfs4_stateid *stateid,
8709 		struct rpc_cred *cred)
8710 {
8711 	struct rpc_task *task;
8712 	int ret;
8713 
8714 	task = _nfs41_free_stateid(server, stateid, cred, true);
8715 	if (IS_ERR(task))
8716 		return PTR_ERR(task);
8717 	ret = rpc_wait_for_completion_task(task);
8718 	if (!ret)
8719 		ret = task->tk_status;
8720 	rpc_put_task(task);
8721 	return ret;
8722 }
8723 
8724 static void
8725 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8726 {
8727 	struct rpc_task *task;
8728 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8729 
8730 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8731 	nfs4_free_lock_state(server, lsp);
8732 	if (IS_ERR(task))
8733 		return;
8734 	rpc_put_task(task);
8735 }
8736 
8737 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8738 		const nfs4_stateid *s2)
8739 {
8740 	if (s1->type != s2->type)
8741 		return false;
8742 
8743 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8744 		return false;
8745 
8746 	if (s1->seqid == s2->seqid)
8747 		return true;
8748 	if (s1->seqid == 0 || s2->seqid == 0)
8749 		return true;
8750 
8751 	return false;
8752 }
8753 
8754 #endif /* CONFIG_NFS_V4_1 */
8755 
8756 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8757 		const nfs4_stateid *s2)
8758 {
8759 	return nfs4_stateid_match(s1, s2);
8760 }
8761 
8762 
8763 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8764 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8765 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8766 	.recover_open	= nfs4_open_reclaim,
8767 	.recover_lock	= nfs4_lock_reclaim,
8768 	.establish_clid = nfs4_init_clientid,
8769 	.detect_trunking = nfs40_discover_server_trunking,
8770 };
8771 
8772 #if defined(CONFIG_NFS_V4_1)
8773 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8774 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8775 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8776 	.recover_open	= nfs4_open_reclaim,
8777 	.recover_lock	= nfs4_lock_reclaim,
8778 	.establish_clid = nfs41_init_clientid,
8779 	.reclaim_complete = nfs41_proc_reclaim_complete,
8780 	.detect_trunking = nfs41_discover_server_trunking,
8781 };
8782 #endif /* CONFIG_NFS_V4_1 */
8783 
8784 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8785 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8786 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8787 	.recover_open	= nfs40_open_expired,
8788 	.recover_lock	= nfs4_lock_expired,
8789 	.establish_clid = nfs4_init_clientid,
8790 };
8791 
8792 #if defined(CONFIG_NFS_V4_1)
8793 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8794 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8795 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8796 	.recover_open	= nfs41_open_expired,
8797 	.recover_lock	= nfs41_lock_expired,
8798 	.establish_clid = nfs41_init_clientid,
8799 };
8800 #endif /* CONFIG_NFS_V4_1 */
8801 
8802 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8803 	.sched_state_renewal = nfs4_proc_async_renew,
8804 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8805 	.renew_lease = nfs4_proc_renew,
8806 };
8807 
8808 #if defined(CONFIG_NFS_V4_1)
8809 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8810 	.sched_state_renewal = nfs41_proc_async_sequence,
8811 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8812 	.renew_lease = nfs4_proc_sequence,
8813 };
8814 #endif
8815 
8816 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8817 	.get_locations = _nfs40_proc_get_locations,
8818 	.fsid_present = _nfs40_proc_fsid_present,
8819 };
8820 
8821 #if defined(CONFIG_NFS_V4_1)
8822 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8823 	.get_locations = _nfs41_proc_get_locations,
8824 	.fsid_present = _nfs41_proc_fsid_present,
8825 };
8826 #endif	/* CONFIG_NFS_V4_1 */
8827 
8828 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8829 	.minor_version = 0,
8830 	.init_caps = NFS_CAP_READDIRPLUS
8831 		| NFS_CAP_ATOMIC_OPEN
8832 		| NFS_CAP_POSIX_LOCK,
8833 	.init_client = nfs40_init_client,
8834 	.shutdown_client = nfs40_shutdown_client,
8835 	.match_stateid = nfs4_match_stateid,
8836 	.find_root_sec = nfs4_find_root_sec,
8837 	.free_lock_state = nfs4_release_lockowner,
8838 	.alloc_seqid = nfs_alloc_seqid,
8839 	.call_sync_ops = &nfs40_call_sync_ops,
8840 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8841 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8842 	.state_renewal_ops = &nfs40_state_renewal_ops,
8843 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8844 };
8845 
8846 #if defined(CONFIG_NFS_V4_1)
8847 static struct nfs_seqid *
8848 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8849 {
8850 	return NULL;
8851 }
8852 
8853 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8854 	.minor_version = 1,
8855 	.init_caps = NFS_CAP_READDIRPLUS
8856 		| NFS_CAP_ATOMIC_OPEN
8857 		| NFS_CAP_POSIX_LOCK
8858 		| NFS_CAP_STATEID_NFSV41
8859 		| NFS_CAP_ATOMIC_OPEN_V1,
8860 	.init_client = nfs41_init_client,
8861 	.shutdown_client = nfs41_shutdown_client,
8862 	.match_stateid = nfs41_match_stateid,
8863 	.find_root_sec = nfs41_find_root_sec,
8864 	.free_lock_state = nfs41_free_lock_state,
8865 	.alloc_seqid = nfs_alloc_no_seqid,
8866 	.call_sync_ops = &nfs41_call_sync_ops,
8867 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8868 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8869 	.state_renewal_ops = &nfs41_state_renewal_ops,
8870 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8871 };
8872 #endif
8873 
8874 #if defined(CONFIG_NFS_V4_2)
8875 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8876 	.minor_version = 2,
8877 	.init_caps = NFS_CAP_READDIRPLUS
8878 		| NFS_CAP_ATOMIC_OPEN
8879 		| NFS_CAP_POSIX_LOCK
8880 		| NFS_CAP_STATEID_NFSV41
8881 		| NFS_CAP_ATOMIC_OPEN_V1
8882 		| NFS_CAP_ALLOCATE
8883 		| NFS_CAP_COPY
8884 		| NFS_CAP_DEALLOCATE
8885 		| NFS_CAP_SEEK
8886 		| NFS_CAP_LAYOUTSTATS
8887 		| NFS_CAP_CLONE,
8888 	.init_client = nfs41_init_client,
8889 	.shutdown_client = nfs41_shutdown_client,
8890 	.match_stateid = nfs41_match_stateid,
8891 	.find_root_sec = nfs41_find_root_sec,
8892 	.free_lock_state = nfs41_free_lock_state,
8893 	.call_sync_ops = &nfs41_call_sync_ops,
8894 	.alloc_seqid = nfs_alloc_no_seqid,
8895 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8896 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8897 	.state_renewal_ops = &nfs41_state_renewal_ops,
8898 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8899 };
8900 #endif
8901 
8902 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8903 	[0] = &nfs_v4_0_minor_ops,
8904 #if defined(CONFIG_NFS_V4_1)
8905 	[1] = &nfs_v4_1_minor_ops,
8906 #endif
8907 #if defined(CONFIG_NFS_V4_2)
8908 	[2] = &nfs_v4_2_minor_ops,
8909 #endif
8910 };
8911 
8912 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8913 {
8914 	ssize_t error, error2;
8915 
8916 	error = generic_listxattr(dentry, list, size);
8917 	if (error < 0)
8918 		return error;
8919 	if (list) {
8920 		list += error;
8921 		size -= error;
8922 	}
8923 
8924 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8925 	if (error2 < 0)
8926 		return error2;
8927 	return error + error2;
8928 }
8929 
8930 static const struct inode_operations nfs4_dir_inode_operations = {
8931 	.create		= nfs_create,
8932 	.lookup		= nfs_lookup,
8933 	.atomic_open	= nfs_atomic_open,
8934 	.link		= nfs_link,
8935 	.unlink		= nfs_unlink,
8936 	.symlink	= nfs_symlink,
8937 	.mkdir		= nfs_mkdir,
8938 	.rmdir		= nfs_rmdir,
8939 	.mknod		= nfs_mknod,
8940 	.rename		= nfs_rename,
8941 	.permission	= nfs_permission,
8942 	.getattr	= nfs_getattr,
8943 	.setattr	= nfs_setattr,
8944 	.getxattr	= generic_getxattr,
8945 	.setxattr	= generic_setxattr,
8946 	.listxattr	= nfs4_listxattr,
8947 	.removexattr	= generic_removexattr,
8948 };
8949 
8950 static const struct inode_operations nfs4_file_inode_operations = {
8951 	.permission	= nfs_permission,
8952 	.getattr	= nfs_getattr,
8953 	.setattr	= nfs_setattr,
8954 	.getxattr	= generic_getxattr,
8955 	.setxattr	= generic_setxattr,
8956 	.listxattr	= nfs4_listxattr,
8957 	.removexattr	= generic_removexattr,
8958 };
8959 
8960 const struct nfs_rpc_ops nfs_v4_clientops = {
8961 	.version	= 4,			/* protocol version */
8962 	.dentry_ops	= &nfs4_dentry_operations,
8963 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8964 	.file_inode_ops	= &nfs4_file_inode_operations,
8965 	.file_ops	= &nfs4_file_operations,
8966 	.getroot	= nfs4_proc_get_root,
8967 	.submount	= nfs4_submount,
8968 	.try_mount	= nfs4_try_mount,
8969 	.getattr	= nfs4_proc_getattr,
8970 	.setattr	= nfs4_proc_setattr,
8971 	.lookup		= nfs4_proc_lookup,
8972 	.access		= nfs4_proc_access,
8973 	.readlink	= nfs4_proc_readlink,
8974 	.create		= nfs4_proc_create,
8975 	.remove		= nfs4_proc_remove,
8976 	.unlink_setup	= nfs4_proc_unlink_setup,
8977 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8978 	.unlink_done	= nfs4_proc_unlink_done,
8979 	.rename_setup	= nfs4_proc_rename_setup,
8980 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8981 	.rename_done	= nfs4_proc_rename_done,
8982 	.link		= nfs4_proc_link,
8983 	.symlink	= nfs4_proc_symlink,
8984 	.mkdir		= nfs4_proc_mkdir,
8985 	.rmdir		= nfs4_proc_remove,
8986 	.readdir	= nfs4_proc_readdir,
8987 	.mknod		= nfs4_proc_mknod,
8988 	.statfs		= nfs4_proc_statfs,
8989 	.fsinfo		= nfs4_proc_fsinfo,
8990 	.pathconf	= nfs4_proc_pathconf,
8991 	.set_capabilities = nfs4_server_capabilities,
8992 	.decode_dirent	= nfs4_decode_dirent,
8993 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8994 	.read_setup	= nfs4_proc_read_setup,
8995 	.read_done	= nfs4_read_done,
8996 	.write_setup	= nfs4_proc_write_setup,
8997 	.write_done	= nfs4_write_done,
8998 	.commit_setup	= nfs4_proc_commit_setup,
8999 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9000 	.commit_done	= nfs4_commit_done,
9001 	.lock		= nfs4_proc_lock,
9002 	.clear_acl_cache = nfs4_zap_acl_attr,
9003 	.close_context  = nfs4_close_context,
9004 	.open_context	= nfs4_atomic_open,
9005 	.have_delegation = nfs4_have_delegation,
9006 	.return_delegation = nfs4_inode_return_delegation,
9007 	.alloc_client	= nfs4_alloc_client,
9008 	.init_client	= nfs4_init_client,
9009 	.free_client	= nfs4_free_client,
9010 	.create_server	= nfs4_create_server,
9011 	.clone_server	= nfs_clone_server,
9012 };
9013 
9014 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9015 	.name	= XATTR_NAME_NFSV4_ACL,
9016 	.list	= nfs4_xattr_list_nfs4_acl,
9017 	.get	= nfs4_xattr_get_nfs4_acl,
9018 	.set	= nfs4_xattr_set_nfs4_acl,
9019 };
9020 
9021 const struct xattr_handler *nfs4_xattr_handlers[] = {
9022 	&nfs4_xattr_nfs4_acl_handler,
9023 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9024 	&nfs4_xattr_nfs4_label_handler,
9025 #endif
9026 	NULL
9027 };
9028 
9029 /*
9030  * Local variables:
9031  *  c-basic-offset: 8
9032  * End:
9033  */
9034