xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision faa16bc4)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.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_BITMASK_SZ		3
75 
76 #define NFS4_POLL_RETRY_MIN	(HZ/10)
77 #define NFS4_POLL_RETRY_MAX	(15*HZ)
78 
79 /* file attributes which can be mapped to nfs attributes */
80 #define NFS4_VALID_ATTRS (ATTR_MODE \
81 	| ATTR_UID \
82 	| ATTR_GID \
83 	| ATTR_SIZE \
84 	| ATTR_ATIME \
85 	| ATTR_MTIME \
86 	| ATTR_CTIME \
87 	| ATTR_ATIME_SET \
88 	| ATTR_MTIME_SET)
89 
90 struct nfs4_opendata;
91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
96 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
97 			    struct nfs_fattr *fattr, struct iattr *sattr,
98 			    struct nfs_open_context *ctx, struct nfs4_label *ilabel,
99 			    struct nfs4_label *olabel);
100 #ifdef CONFIG_NFS_V4_1
101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
102 		struct rpc_cred *cred,
103 		struct nfs4_slot *slot,
104 		bool is_privileged);
105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
106 		struct rpc_cred *);
107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
108 		struct rpc_cred *, bool);
109 #endif
110 
111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
112 static inline struct nfs4_label *
113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
114 	struct iattr *sattr, struct nfs4_label *label)
115 {
116 	int err;
117 
118 	if (label == NULL)
119 		return NULL;
120 
121 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
122 		return NULL;
123 
124 	err = security_dentry_init_security(dentry, sattr->ia_mode,
125 				&dentry->d_name, (void **)&label->label, &label->len);
126 	if (err == 0)
127 		return label;
128 
129 	return NULL;
130 }
131 static inline void
132 nfs4_label_release_security(struct nfs4_label *label)
133 {
134 	if (label)
135 		security_release_secctx(label->label, label->len);
136 }
137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 {
139 	if (label)
140 		return server->attr_bitmask;
141 
142 	return server->attr_bitmask_nl;
143 }
144 #else
145 static inline struct nfs4_label *
146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
147 	struct iattr *sattr, struct nfs4_label *l)
148 { return NULL; }
149 static inline void
150 nfs4_label_release_security(struct nfs4_label *label)
151 { return; }
152 static inline u32 *
153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
154 { return server->attr_bitmask; }
155 #endif
156 
157 /* Prevent leaks of NFSv4 errors into userland */
158 static int nfs4_map_errors(int err)
159 {
160 	if (err >= -1000)
161 		return err;
162 	switch (err) {
163 	case -NFS4ERR_RESOURCE:
164 	case -NFS4ERR_LAYOUTTRYLATER:
165 	case -NFS4ERR_RECALLCONFLICT:
166 		return -EREMOTEIO;
167 	case -NFS4ERR_WRONGSEC:
168 	case -NFS4ERR_WRONG_CRED:
169 		return -EPERM;
170 	case -NFS4ERR_BADOWNER:
171 	case -NFS4ERR_BADNAME:
172 		return -EINVAL;
173 	case -NFS4ERR_SHARE_DENIED:
174 		return -EACCES;
175 	case -NFS4ERR_MINOR_VERS_MISMATCH:
176 		return -EPROTONOSUPPORT;
177 	case -NFS4ERR_FILE_OPEN:
178 		return -EBUSY;
179 	default:
180 		dprintk("%s could not handle NFSv4 error %d\n",
181 				__func__, -err);
182 		break;
183 	}
184 	return -EIO;
185 }
186 
187 /*
188  * This is our standard bitmap for GETATTR requests.
189  */
190 const u32 nfs4_fattr_bitmap[3] = {
191 	FATTR4_WORD0_TYPE
192 	| FATTR4_WORD0_CHANGE
193 	| FATTR4_WORD0_SIZE
194 	| FATTR4_WORD0_FSID
195 	| FATTR4_WORD0_FILEID,
196 	FATTR4_WORD1_MODE
197 	| FATTR4_WORD1_NUMLINKS
198 	| FATTR4_WORD1_OWNER
199 	| FATTR4_WORD1_OWNER_GROUP
200 	| FATTR4_WORD1_RAWDEV
201 	| FATTR4_WORD1_SPACE_USED
202 	| FATTR4_WORD1_TIME_ACCESS
203 	| FATTR4_WORD1_TIME_METADATA
204 	| FATTR4_WORD1_TIME_MODIFY
205 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
207 	FATTR4_WORD2_SECURITY_LABEL
208 #endif
209 };
210 
211 static const u32 nfs4_pnfs_open_bitmap[3] = {
212 	FATTR4_WORD0_TYPE
213 	| FATTR4_WORD0_CHANGE
214 	| FATTR4_WORD0_SIZE
215 	| FATTR4_WORD0_FSID
216 	| FATTR4_WORD0_FILEID,
217 	FATTR4_WORD1_MODE
218 	| FATTR4_WORD1_NUMLINKS
219 	| FATTR4_WORD1_OWNER
220 	| FATTR4_WORD1_OWNER_GROUP
221 	| FATTR4_WORD1_RAWDEV
222 	| FATTR4_WORD1_SPACE_USED
223 	| FATTR4_WORD1_TIME_ACCESS
224 	| FATTR4_WORD1_TIME_METADATA
225 	| FATTR4_WORD1_TIME_MODIFY,
226 	FATTR4_WORD2_MDSTHRESHOLD
227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
228 	| FATTR4_WORD2_SECURITY_LABEL
229 #endif
230 };
231 
232 static const u32 nfs4_open_noattr_bitmap[3] = {
233 	FATTR4_WORD0_TYPE
234 	| FATTR4_WORD0_FILEID,
235 };
236 
237 const u32 nfs4_statfs_bitmap[3] = {
238 	FATTR4_WORD0_FILES_AVAIL
239 	| FATTR4_WORD0_FILES_FREE
240 	| FATTR4_WORD0_FILES_TOTAL,
241 	FATTR4_WORD1_SPACE_AVAIL
242 	| FATTR4_WORD1_SPACE_FREE
243 	| FATTR4_WORD1_SPACE_TOTAL
244 };
245 
246 const u32 nfs4_pathconf_bitmap[3] = {
247 	FATTR4_WORD0_MAXLINK
248 	| FATTR4_WORD0_MAXNAME,
249 	0
250 };
251 
252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
253 			| FATTR4_WORD0_MAXREAD
254 			| FATTR4_WORD0_MAXWRITE
255 			| FATTR4_WORD0_LEASE_TIME,
256 			FATTR4_WORD1_TIME_DELTA
257 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
258 			FATTR4_WORD2_LAYOUT_BLKSIZE
259 			| FATTR4_WORD2_CLONE_BLKSIZE
260 };
261 
262 const u32 nfs4_fs_locations_bitmap[3] = {
263 	FATTR4_WORD0_CHANGE
264 	| FATTR4_WORD0_SIZE
265 	| FATTR4_WORD0_FSID
266 	| FATTR4_WORD0_FILEID
267 	| FATTR4_WORD0_FS_LOCATIONS,
268 	FATTR4_WORD1_OWNER
269 	| FATTR4_WORD1_OWNER_GROUP
270 	| FATTR4_WORD1_RAWDEV
271 	| FATTR4_WORD1_SPACE_USED
272 	| FATTR4_WORD1_TIME_ACCESS
273 	| FATTR4_WORD1_TIME_METADATA
274 	| FATTR4_WORD1_TIME_MODIFY
275 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
276 };
277 
278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
279 		struct inode *inode)
280 {
281 	unsigned long cache_validity;
282 
283 	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
284 	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
285 		return;
286 
287 	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
288 	if (!(cache_validity & NFS_INO_REVAL_FORCED))
289 		cache_validity &= ~(NFS_INO_INVALID_CHANGE
290 				| NFS_INO_INVALID_SIZE);
291 
292 	if (!(cache_validity & NFS_INO_INVALID_SIZE))
293 		dst[0] &= ~FATTR4_WORD0_SIZE;
294 
295 	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
296 		dst[0] &= ~FATTR4_WORD0_CHANGE;
297 }
298 
299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
300 		const __u32 *src, struct inode *inode)
301 {
302 	nfs4_bitmap_copy_adjust(dst, src, inode);
303 }
304 
305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
306 		struct nfs4_readdir_arg *readdir)
307 {
308 	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
309 	__be32 *start, *p;
310 
311 	if (cookie > 2) {
312 		readdir->cookie = cookie;
313 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
314 		return;
315 	}
316 
317 	readdir->cookie = 0;
318 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
319 	if (cookie == 2)
320 		return;
321 
322 	/*
323 	 * NFSv4 servers do not return entries for '.' and '..'
324 	 * Therefore, we fake these entries here.  We let '.'
325 	 * have cookie 0 and '..' have cookie 1.  Note that
326 	 * when talking to the server, we always send cookie 0
327 	 * instead of 1 or 2.
328 	 */
329 	start = p = kmap_atomic(*readdir->pages);
330 
331 	if (cookie == 0) {
332 		*p++ = xdr_one;                                  /* next */
333 		*p++ = xdr_zero;                   /* cookie, first word */
334 		*p++ = xdr_one;                   /* cookie, second word */
335 		*p++ = xdr_one;                             /* entry len */
336 		memcpy(p, ".\0\0\0", 4);                        /* entry */
337 		p++;
338 		*p++ = xdr_one;                         /* bitmap length */
339 		*p++ = htonl(attrs);                           /* bitmap */
340 		*p++ = htonl(12);             /* attribute buffer length */
341 		*p++ = htonl(NF4DIR);
342 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
343 	}
344 
345 	*p++ = xdr_one;                                  /* next */
346 	*p++ = xdr_zero;                   /* cookie, first word */
347 	*p++ = xdr_two;                   /* cookie, second word */
348 	*p++ = xdr_two;                             /* entry len */
349 	memcpy(p, "..\0\0", 4);                         /* entry */
350 	p++;
351 	*p++ = xdr_one;                         /* bitmap length */
352 	*p++ = htonl(attrs);                           /* bitmap */
353 	*p++ = htonl(12);             /* attribute buffer length */
354 	*p++ = htonl(NF4DIR);
355 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
356 
357 	readdir->pgbase = (char *)p - (char *)start;
358 	readdir->count -= readdir->pgbase;
359 	kunmap_atomic(start);
360 }
361 
362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
363 		nfs4_stateid *stateid,
364 		struct rpc_cred *cred)
365 {
366 	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
367 
368 	ops->test_and_free_expired(server, stateid, cred);
369 }
370 
371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
372 		nfs4_stateid *stateid,
373 		struct rpc_cred *cred)
374 {
375 	stateid->type = NFS4_REVOKED_STATEID_TYPE;
376 	nfs4_test_and_free_stateid(server, stateid, cred);
377 }
378 
379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
380 		const nfs4_stateid *stateid,
381 		struct rpc_cred *cred)
382 {
383 	nfs4_stateid tmp;
384 
385 	nfs4_stateid_copy(&tmp, stateid);
386 	__nfs4_free_revoked_stateid(server, &tmp, cred);
387 }
388 
389 static long nfs4_update_delay(long *timeout)
390 {
391 	long ret;
392 	if (!timeout)
393 		return NFS4_POLL_RETRY_MAX;
394 	if (*timeout <= 0)
395 		*timeout = NFS4_POLL_RETRY_MIN;
396 	if (*timeout > NFS4_POLL_RETRY_MAX)
397 		*timeout = NFS4_POLL_RETRY_MAX;
398 	ret = *timeout;
399 	*timeout <<= 1;
400 	return ret;
401 }
402 
403 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
404 {
405 	int res = 0;
406 
407 	might_sleep();
408 
409 	freezable_schedule_timeout_killable_unsafe(
410 		nfs4_update_delay(timeout));
411 	if (fatal_signal_pending(current))
412 		res = -ERESTARTSYS;
413 	return res;
414 }
415 
416 /* This is the error handling routine for processes that are allowed
417  * to sleep.
418  */
419 static int nfs4_do_handle_exception(struct nfs_server *server,
420 		int errorcode, struct nfs4_exception *exception)
421 {
422 	struct nfs_client *clp = server->nfs_client;
423 	struct nfs4_state *state = exception->state;
424 	const nfs4_stateid *stateid = exception->stateid;
425 	struct inode *inode = exception->inode;
426 	int ret = errorcode;
427 
428 	exception->delay = 0;
429 	exception->recovering = 0;
430 	exception->retry = 0;
431 
432 	if (stateid == NULL && state != NULL)
433 		stateid = &state->stateid;
434 
435 	switch(errorcode) {
436 		case 0:
437 			return 0;
438 		case -NFS4ERR_BADHANDLE:
439 		case -ESTALE:
440 			if (inode != NULL && S_ISREG(inode->i_mode))
441 				pnfs_destroy_layout(NFS_I(inode));
442 			break;
443 		case -NFS4ERR_DELEG_REVOKED:
444 		case -NFS4ERR_ADMIN_REVOKED:
445 		case -NFS4ERR_EXPIRED:
446 		case -NFS4ERR_BAD_STATEID:
447 			if (inode != NULL && stateid != NULL) {
448 				nfs_inode_find_state_and_recover(inode,
449 						stateid);
450 				goto wait_on_recovery;
451 			}
452 		case -NFS4ERR_OPENMODE:
453 			if (inode) {
454 				int err;
455 
456 				err = nfs_async_inode_return_delegation(inode,
457 						stateid);
458 				if (err == 0)
459 					goto wait_on_recovery;
460 				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
461 					exception->retry = 1;
462 					break;
463 				}
464 			}
465 			if (state == NULL)
466 				break;
467 			ret = nfs4_schedule_stateid_recovery(server, state);
468 			if (ret < 0)
469 				break;
470 			goto wait_on_recovery;
471 		case -NFS4ERR_STALE_STATEID:
472 		case -NFS4ERR_STALE_CLIENTID:
473 			nfs4_schedule_lease_recovery(clp);
474 			goto wait_on_recovery;
475 		case -NFS4ERR_MOVED:
476 			ret = nfs4_schedule_migration_recovery(server);
477 			if (ret < 0)
478 				break;
479 			goto wait_on_recovery;
480 		case -NFS4ERR_LEASE_MOVED:
481 			nfs4_schedule_lease_moved_recovery(clp);
482 			goto wait_on_recovery;
483 #if defined(CONFIG_NFS_V4_1)
484 		case -NFS4ERR_BADSESSION:
485 		case -NFS4ERR_BADSLOT:
486 		case -NFS4ERR_BAD_HIGH_SLOT:
487 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
488 		case -NFS4ERR_DEADSESSION:
489 		case -NFS4ERR_SEQ_FALSE_RETRY:
490 		case -NFS4ERR_SEQ_MISORDERED:
491 			dprintk("%s ERROR: %d Reset session\n", __func__,
492 				errorcode);
493 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
494 			goto wait_on_recovery;
495 #endif /* defined(CONFIG_NFS_V4_1) */
496 		case -NFS4ERR_FILE_OPEN:
497 			if (exception->timeout > HZ) {
498 				/* We have retried a decent amount, time to
499 				 * fail
500 				 */
501 				ret = -EBUSY;
502 				break;
503 			}
504 		case -NFS4ERR_DELAY:
505 			nfs_inc_server_stats(server, NFSIOS_DELAY);
506 		case -NFS4ERR_GRACE:
507 		case -NFS4ERR_LAYOUTTRYLATER:
508 		case -NFS4ERR_RECALLCONFLICT:
509 			exception->delay = 1;
510 			return 0;
511 
512 		case -NFS4ERR_RETRY_UNCACHED_REP:
513 		case -NFS4ERR_OLD_STATEID:
514 			exception->retry = 1;
515 			break;
516 		case -NFS4ERR_BADOWNER:
517 			/* The following works around a Linux server bug! */
518 		case -NFS4ERR_BADNAME:
519 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
520 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
521 				exception->retry = 1;
522 				printk(KERN_WARNING "NFS: v4 server %s "
523 						"does not accept raw "
524 						"uid/gids. "
525 						"Reenabling the idmapper.\n",
526 						server->nfs_client->cl_hostname);
527 			}
528 	}
529 	/* We failed to handle the error */
530 	return nfs4_map_errors(ret);
531 wait_on_recovery:
532 	exception->recovering = 1;
533 	return 0;
534 }
535 
536 /* This is the error handling routine for processes that are allowed
537  * to sleep.
538  */
539 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
540 {
541 	struct nfs_client *clp = server->nfs_client;
542 	int ret;
543 
544 	ret = nfs4_do_handle_exception(server, errorcode, exception);
545 	if (exception->delay) {
546 		ret = nfs4_delay(server->client, &exception->timeout);
547 		goto out_retry;
548 	}
549 	if (exception->recovering) {
550 		ret = nfs4_wait_clnt_recover(clp);
551 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
552 			return -EIO;
553 		goto out_retry;
554 	}
555 	return ret;
556 out_retry:
557 	if (ret == 0)
558 		exception->retry = 1;
559 	return ret;
560 }
561 
562 static int
563 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
564 		int errorcode, struct nfs4_exception *exception)
565 {
566 	struct nfs_client *clp = server->nfs_client;
567 	int ret;
568 
569 	ret = nfs4_do_handle_exception(server, errorcode, exception);
570 	if (exception->delay) {
571 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
572 		goto out_retry;
573 	}
574 	if (exception->recovering) {
575 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
576 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
577 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
578 		goto out_retry;
579 	}
580 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
581 		ret = -EIO;
582 	return ret;
583 out_retry:
584 	if (ret == 0)
585 		exception->retry = 1;
586 	return ret;
587 }
588 
589 static int
590 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
591 			struct nfs4_state *state, long *timeout)
592 {
593 	struct nfs4_exception exception = {
594 		.state = state,
595 	};
596 
597 	if (task->tk_status >= 0)
598 		return 0;
599 	if (timeout)
600 		exception.timeout = *timeout;
601 	task->tk_status = nfs4_async_handle_exception(task, server,
602 			task->tk_status,
603 			&exception);
604 	if (exception.delay && timeout)
605 		*timeout = exception.timeout;
606 	if (exception.retry)
607 		return -EAGAIN;
608 	return 0;
609 }
610 
611 /*
612  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
613  * or 'false' otherwise.
614  */
615 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
616 {
617 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
618 	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
619 }
620 
621 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
622 {
623 	spin_lock(&clp->cl_lock);
624 	if (time_before(clp->cl_last_renewal,timestamp))
625 		clp->cl_last_renewal = timestamp;
626 	spin_unlock(&clp->cl_lock);
627 }
628 
629 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
630 {
631 	struct nfs_client *clp = server->nfs_client;
632 
633 	if (!nfs4_has_session(clp))
634 		do_renew_lease(clp, timestamp);
635 }
636 
637 struct nfs4_call_sync_data {
638 	const struct nfs_server *seq_server;
639 	struct nfs4_sequence_args *seq_args;
640 	struct nfs4_sequence_res *seq_res;
641 };
642 
643 void nfs4_init_sequence(struct nfs4_sequence_args *args,
644 			struct nfs4_sequence_res *res, int cache_reply,
645 			int privileged)
646 {
647 	args->sa_slot = NULL;
648 	args->sa_cache_this = cache_reply;
649 	args->sa_privileged = privileged;
650 
651 	res->sr_slot = NULL;
652 }
653 
654 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
655 {
656 	struct nfs4_slot *slot = res->sr_slot;
657 	struct nfs4_slot_table *tbl;
658 
659 	tbl = slot->table;
660 	spin_lock(&tbl->slot_tbl_lock);
661 	if (!nfs41_wake_and_assign_slot(tbl, slot))
662 		nfs4_free_slot(tbl, slot);
663 	spin_unlock(&tbl->slot_tbl_lock);
664 
665 	res->sr_slot = NULL;
666 }
667 
668 static int nfs40_sequence_done(struct rpc_task *task,
669 			       struct nfs4_sequence_res *res)
670 {
671 	if (res->sr_slot != NULL)
672 		nfs40_sequence_free_slot(res);
673 	return 1;
674 }
675 
676 #if defined(CONFIG_NFS_V4_1)
677 
678 static void nfs41_release_slot(struct nfs4_slot *slot)
679 {
680 	struct nfs4_session *session;
681 	struct nfs4_slot_table *tbl;
682 	bool send_new_highest_used_slotid = false;
683 
684 	if (!slot)
685 		return;
686 	tbl = slot->table;
687 	session = tbl->session;
688 
689 	/* Bump the slot sequence number */
690 	if (slot->seq_done)
691 		slot->seq_nr++;
692 	slot->seq_done = 0;
693 
694 	spin_lock(&tbl->slot_tbl_lock);
695 	/* Be nice to the server: try to ensure that the last transmitted
696 	 * value for highest_user_slotid <= target_highest_slotid
697 	 */
698 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
699 		send_new_highest_used_slotid = true;
700 
701 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
702 		send_new_highest_used_slotid = false;
703 		goto out_unlock;
704 	}
705 	nfs4_free_slot(tbl, slot);
706 
707 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
708 		send_new_highest_used_slotid = false;
709 out_unlock:
710 	spin_unlock(&tbl->slot_tbl_lock);
711 	if (send_new_highest_used_slotid)
712 		nfs41_notify_server(session->clp);
713 	if (waitqueue_active(&tbl->slot_waitq))
714 		wake_up_all(&tbl->slot_waitq);
715 }
716 
717 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
718 {
719 	nfs41_release_slot(res->sr_slot);
720 	res->sr_slot = NULL;
721 }
722 
723 static int nfs41_sequence_process(struct rpc_task *task,
724 		struct nfs4_sequence_res *res)
725 {
726 	struct nfs4_session *session;
727 	struct nfs4_slot *slot = res->sr_slot;
728 	struct nfs_client *clp;
729 	bool interrupted = false;
730 	int ret = 1;
731 
732 	if (slot == NULL)
733 		goto out_noaction;
734 	/* don't increment the sequence number if the task wasn't sent */
735 	if (!RPC_WAS_SENT(task))
736 		goto out;
737 
738 	session = slot->table->session;
739 
740 	if (slot->interrupted) {
741 		if (res->sr_status != -NFS4ERR_DELAY)
742 			slot->interrupted = 0;
743 		interrupted = true;
744 	}
745 
746 	trace_nfs4_sequence_done(session, res);
747 	/* Check the SEQUENCE operation status */
748 	switch (res->sr_status) {
749 	case 0:
750 		/* Update the slot's sequence and clientid lease timer */
751 		slot->seq_done = 1;
752 		clp = session->clp;
753 		do_renew_lease(clp, res->sr_timestamp);
754 		/* Check sequence flags */
755 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
756 				!!slot->privileged);
757 		nfs41_update_target_slotid(slot->table, slot, res);
758 		break;
759 	case 1:
760 		/*
761 		 * sr_status remains 1 if an RPC level error occurred.
762 		 * The server may or may not have processed the sequence
763 		 * operation..
764 		 * Mark the slot as having hosted an interrupted RPC call.
765 		 */
766 		slot->interrupted = 1;
767 		goto out;
768 	case -NFS4ERR_DELAY:
769 		/* The server detected a resend of the RPC call and
770 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
771 		 * of RFC5661.
772 		 */
773 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
774 			__func__,
775 			slot->slot_nr,
776 			slot->seq_nr);
777 		goto out_retry;
778 	case -NFS4ERR_RETRY_UNCACHED_REP:
779 	case -NFS4ERR_SEQ_FALSE_RETRY:
780 		/*
781 		 * The server thinks we tried to replay a request.
782 		 * Retry the call after bumping the sequence ID.
783 		 */
784 		goto retry_new_seq;
785 	case -NFS4ERR_BADSLOT:
786 		/*
787 		 * The slot id we used was probably retired. Try again
788 		 * using a different slot id.
789 		 */
790 		if (slot->slot_nr < slot->table->target_highest_slotid)
791 			goto session_recover;
792 		goto retry_nowait;
793 	case -NFS4ERR_SEQ_MISORDERED:
794 		/*
795 		 * Was the last operation on this sequence interrupted?
796 		 * If so, retry after bumping the sequence number.
797 		 */
798 		if (interrupted)
799 			goto retry_new_seq;
800 		/*
801 		 * Could this slot have been previously retired?
802 		 * If so, then the server may be expecting seq_nr = 1!
803 		 */
804 		if (slot->seq_nr != 1) {
805 			slot->seq_nr = 1;
806 			goto retry_nowait;
807 		}
808 		goto session_recover;
809 	default:
810 		/* Just update the slot sequence no. */
811 		slot->seq_done = 1;
812 	}
813 out:
814 	/* The session may be reset by one of the error handlers. */
815 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
816 out_noaction:
817 	return ret;
818 session_recover:
819 	nfs4_schedule_session_recovery(session, res->sr_status);
820 	goto retry_nowait;
821 retry_new_seq:
822 	++slot->seq_nr;
823 retry_nowait:
824 	if (rpc_restart_call_prepare(task)) {
825 		nfs41_sequence_free_slot(res);
826 		task->tk_status = 0;
827 		ret = 0;
828 	}
829 	goto out;
830 out_retry:
831 	if (!rpc_restart_call(task))
832 		goto out;
833 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
834 	return 0;
835 }
836 
837 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
838 {
839 	if (!nfs41_sequence_process(task, res))
840 		return 0;
841 	if (res->sr_slot != NULL)
842 		nfs41_sequence_free_slot(res);
843 	return 1;
844 
845 }
846 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
847 
848 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
849 {
850 	if (res->sr_slot == NULL)
851 		return 1;
852 	if (res->sr_slot->table->session != NULL)
853 		return nfs41_sequence_process(task, res);
854 	return nfs40_sequence_done(task, res);
855 }
856 
857 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
858 {
859 	if (res->sr_slot != NULL) {
860 		if (res->sr_slot->table->session != NULL)
861 			nfs41_sequence_free_slot(res);
862 		else
863 			nfs40_sequence_free_slot(res);
864 	}
865 }
866 
867 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
868 {
869 	if (res->sr_slot == NULL)
870 		return 1;
871 	if (!res->sr_slot->table->session)
872 		return nfs40_sequence_done(task, res);
873 	return nfs41_sequence_done(task, res);
874 }
875 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
876 
877 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
878 {
879 	struct nfs4_call_sync_data *data = calldata;
880 
881 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
882 
883 	nfs4_setup_sequence(data->seq_server->nfs_client,
884 			    data->seq_args, data->seq_res, task);
885 }
886 
887 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
888 {
889 	struct nfs4_call_sync_data *data = calldata;
890 
891 	nfs41_sequence_done(task, data->seq_res);
892 }
893 
894 static const struct rpc_call_ops nfs41_call_sync_ops = {
895 	.rpc_call_prepare = nfs41_call_sync_prepare,
896 	.rpc_call_done = nfs41_call_sync_done,
897 };
898 
899 static void
900 nfs4_sequence_process_interrupted(struct nfs_client *client,
901 		struct nfs4_slot *slot, struct rpc_cred *cred)
902 {
903 	struct rpc_task *task;
904 
905 	task = _nfs41_proc_sequence(client, cred, slot, true);
906 	if (!IS_ERR(task))
907 		rpc_put_task_async(task);
908 }
909 
910 #else	/* !CONFIG_NFS_V4_1 */
911 
912 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
913 {
914 	return nfs40_sequence_done(task, res);
915 }
916 
917 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
918 {
919 	if (res->sr_slot != NULL)
920 		nfs40_sequence_free_slot(res);
921 }
922 
923 int nfs4_sequence_done(struct rpc_task *task,
924 		       struct nfs4_sequence_res *res)
925 {
926 	return nfs40_sequence_done(task, res);
927 }
928 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
929 
930 static void
931 nfs4_sequence_process_interrupted(struct nfs_client *client,
932 		struct nfs4_slot *slot, struct rpc_cred *cred)
933 {
934 	WARN_ON_ONCE(1);
935 	slot->interrupted = 0;
936 }
937 
938 #endif	/* !CONFIG_NFS_V4_1 */
939 
940 static
941 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
942 		struct nfs4_sequence_res *res,
943 		struct nfs4_slot *slot)
944 {
945 	if (!slot)
946 		return;
947 	slot->privileged = args->sa_privileged ? 1 : 0;
948 	args->sa_slot = slot;
949 
950 	res->sr_slot = slot;
951 	res->sr_timestamp = jiffies;
952 	res->sr_status_flags = 0;
953 	res->sr_status = 1;
954 
955 }
956 
957 int nfs4_setup_sequence(struct nfs_client *client,
958 			struct nfs4_sequence_args *args,
959 			struct nfs4_sequence_res *res,
960 			struct rpc_task *task)
961 {
962 	struct nfs4_session *session = nfs4_get_session(client);
963 	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
964 	struct nfs4_slot *slot;
965 
966 	/* slot already allocated? */
967 	if (res->sr_slot != NULL)
968 		goto out_start;
969 
970 	if (session) {
971 		tbl = &session->fc_slot_table;
972 		task->tk_timeout = 0;
973 	}
974 
975 	for (;;) {
976 		spin_lock(&tbl->slot_tbl_lock);
977 		/* The state manager will wait until the slot table is empty */
978 		if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
979 			goto out_sleep;
980 
981 		slot = nfs4_alloc_slot(tbl);
982 		if (IS_ERR(slot)) {
983 			/* Try again in 1/4 second */
984 			if (slot == ERR_PTR(-ENOMEM))
985 				task->tk_timeout = HZ >> 2;
986 			goto out_sleep;
987 		}
988 		spin_unlock(&tbl->slot_tbl_lock);
989 
990 		if (likely(!slot->interrupted))
991 			break;
992 		nfs4_sequence_process_interrupted(client,
993 				slot, task->tk_msg.rpc_cred);
994 	}
995 
996 	nfs4_sequence_attach_slot(args, res, slot);
997 
998 	trace_nfs4_setup_sequence(session, args);
999 out_start:
1000 	rpc_call_start(task);
1001 	return 0;
1002 
1003 out_sleep:
1004 	if (args->sa_privileged)
1005 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1006 				NULL, RPC_PRIORITY_PRIVILEGED);
1007 	else
1008 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1009 	spin_unlock(&tbl->slot_tbl_lock);
1010 	return -EAGAIN;
1011 }
1012 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1013 
1014 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1015 {
1016 	struct nfs4_call_sync_data *data = calldata;
1017 	nfs4_setup_sequence(data->seq_server->nfs_client,
1018 				data->seq_args, data->seq_res, task);
1019 }
1020 
1021 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1022 {
1023 	struct nfs4_call_sync_data *data = calldata;
1024 	nfs4_sequence_done(task, data->seq_res);
1025 }
1026 
1027 static const struct rpc_call_ops nfs40_call_sync_ops = {
1028 	.rpc_call_prepare = nfs40_call_sync_prepare,
1029 	.rpc_call_done = nfs40_call_sync_done,
1030 };
1031 
1032 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1033 				   struct nfs_server *server,
1034 				   struct rpc_message *msg,
1035 				   struct nfs4_sequence_args *args,
1036 				   struct nfs4_sequence_res *res)
1037 {
1038 	int ret;
1039 	struct rpc_task *task;
1040 	struct nfs_client *clp = server->nfs_client;
1041 	struct nfs4_call_sync_data data = {
1042 		.seq_server = server,
1043 		.seq_args = args,
1044 		.seq_res = res,
1045 	};
1046 	struct rpc_task_setup task_setup = {
1047 		.rpc_client = clnt,
1048 		.rpc_message = msg,
1049 		.callback_ops = clp->cl_mvops->call_sync_ops,
1050 		.callback_data = &data
1051 	};
1052 
1053 	task = rpc_run_task(&task_setup);
1054 	if (IS_ERR(task))
1055 		ret = PTR_ERR(task);
1056 	else {
1057 		ret = task->tk_status;
1058 		rpc_put_task(task);
1059 	}
1060 	return ret;
1061 }
1062 
1063 int nfs4_call_sync(struct rpc_clnt *clnt,
1064 		   struct nfs_server *server,
1065 		   struct rpc_message *msg,
1066 		   struct nfs4_sequence_args *args,
1067 		   struct nfs4_sequence_res *res,
1068 		   int cache_reply)
1069 {
1070 	nfs4_init_sequence(args, res, cache_reply, 0);
1071 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1072 }
1073 
1074 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1075 		unsigned long timestamp)
1076 {
1077 	struct nfs_inode *nfsi = NFS_I(dir);
1078 
1079 	spin_lock(&dir->i_lock);
1080 	nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1081 		| NFS_INO_INVALID_MTIME
1082 		| NFS_INO_INVALID_DATA;
1083 	if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1084 		nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1085 		nfsi->attrtimeo_timestamp = jiffies;
1086 	} else {
1087 		nfs_force_lookup_revalidate(dir);
1088 		if (cinfo->before != inode_peek_iversion_raw(dir))
1089 			nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1090 				NFS_INO_INVALID_ACL;
1091 	}
1092 	inode_set_iversion_raw(dir, cinfo->after);
1093 	nfsi->read_cache_jiffies = timestamp;
1094 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1095 	nfs_fscache_invalidate(dir);
1096 	spin_unlock(&dir->i_lock);
1097 }
1098 
1099 struct nfs4_open_createattrs {
1100 	struct nfs4_label *label;
1101 	struct iattr *sattr;
1102 	const __u32 verf[2];
1103 };
1104 
1105 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1106 		int err, struct nfs4_exception *exception)
1107 {
1108 	if (err != -EINVAL)
1109 		return false;
1110 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1111 		return false;
1112 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1113 	exception->retry = 1;
1114 	return true;
1115 }
1116 
1117 static u32
1118 nfs4_map_atomic_open_share(struct nfs_server *server,
1119 		fmode_t fmode, int openflags)
1120 {
1121 	u32 res = 0;
1122 
1123 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1124 	case FMODE_READ:
1125 		res = NFS4_SHARE_ACCESS_READ;
1126 		break;
1127 	case FMODE_WRITE:
1128 		res = NFS4_SHARE_ACCESS_WRITE;
1129 		break;
1130 	case FMODE_READ|FMODE_WRITE:
1131 		res = NFS4_SHARE_ACCESS_BOTH;
1132 	}
1133 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1134 		goto out;
1135 	/* Want no delegation if we're using O_DIRECT */
1136 	if (openflags & O_DIRECT)
1137 		res |= NFS4_SHARE_WANT_NO_DELEG;
1138 out:
1139 	return res;
1140 }
1141 
1142 static enum open_claim_type4
1143 nfs4_map_atomic_open_claim(struct nfs_server *server,
1144 		enum open_claim_type4 claim)
1145 {
1146 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1147 		return claim;
1148 	switch (claim) {
1149 	default:
1150 		return claim;
1151 	case NFS4_OPEN_CLAIM_FH:
1152 		return NFS4_OPEN_CLAIM_NULL;
1153 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1154 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1155 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1156 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1157 	}
1158 }
1159 
1160 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1161 {
1162 	p->o_res.f_attr = &p->f_attr;
1163 	p->o_res.f_label = p->f_label;
1164 	p->o_res.seqid = p->o_arg.seqid;
1165 	p->c_res.seqid = p->c_arg.seqid;
1166 	p->o_res.server = p->o_arg.server;
1167 	p->o_res.access_request = p->o_arg.access;
1168 	nfs_fattr_init(&p->f_attr);
1169 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1170 }
1171 
1172 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1173 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1174 		const struct nfs4_open_createattrs *c,
1175 		enum open_claim_type4 claim,
1176 		gfp_t gfp_mask)
1177 {
1178 	struct dentry *parent = dget_parent(dentry);
1179 	struct inode *dir = d_inode(parent);
1180 	struct nfs_server *server = NFS_SERVER(dir);
1181 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1182 	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1183 	struct nfs4_opendata *p;
1184 
1185 	p = kzalloc(sizeof(*p), gfp_mask);
1186 	if (p == NULL)
1187 		goto err;
1188 
1189 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1190 	if (IS_ERR(p->f_label))
1191 		goto err_free_p;
1192 
1193 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1194 	if (IS_ERR(p->a_label))
1195 		goto err_free_f;
1196 
1197 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1198 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1199 	if (IS_ERR(p->o_arg.seqid))
1200 		goto err_free_label;
1201 	nfs_sb_active(dentry->d_sb);
1202 	p->dentry = dget(dentry);
1203 	p->dir = parent;
1204 	p->owner = sp;
1205 	atomic_inc(&sp->so_count);
1206 	p->o_arg.open_flags = flags;
1207 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1208 	p->o_arg.umask = current_umask();
1209 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1210 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1211 			fmode, flags);
1212 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1213 	 * will return permission denied for all bits until close */
1214 	if (!(flags & O_EXCL)) {
1215 		/* ask server to check for all possible rights as results
1216 		 * are cached */
1217 		switch (p->o_arg.claim) {
1218 		default:
1219 			break;
1220 		case NFS4_OPEN_CLAIM_NULL:
1221 		case NFS4_OPEN_CLAIM_FH:
1222 			p->o_arg.access = NFS4_ACCESS_READ |
1223 				NFS4_ACCESS_MODIFY |
1224 				NFS4_ACCESS_EXTEND |
1225 				NFS4_ACCESS_EXECUTE;
1226 		}
1227 	}
1228 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1229 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1230 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1231 	p->o_arg.name = &dentry->d_name;
1232 	p->o_arg.server = server;
1233 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1234 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1235 	p->o_arg.label = nfs4_label_copy(p->a_label, label);
1236 	switch (p->o_arg.claim) {
1237 	case NFS4_OPEN_CLAIM_NULL:
1238 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1239 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1240 		p->o_arg.fh = NFS_FH(dir);
1241 		break;
1242 	case NFS4_OPEN_CLAIM_PREVIOUS:
1243 	case NFS4_OPEN_CLAIM_FH:
1244 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1245 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1246 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1247 	}
1248 	if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1249 		p->o_arg.u.attrs = &p->attrs;
1250 		memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1251 
1252 		memcpy(p->o_arg.u.verifier.data, c->verf,
1253 				sizeof(p->o_arg.u.verifier.data));
1254 	}
1255 	p->c_arg.fh = &p->o_res.fh;
1256 	p->c_arg.stateid = &p->o_res.stateid;
1257 	p->c_arg.seqid = p->o_arg.seqid;
1258 	nfs4_init_opendata_res(p);
1259 	kref_init(&p->kref);
1260 	return p;
1261 
1262 err_free_label:
1263 	nfs4_label_free(p->a_label);
1264 err_free_f:
1265 	nfs4_label_free(p->f_label);
1266 err_free_p:
1267 	kfree(p);
1268 err:
1269 	dput(parent);
1270 	return NULL;
1271 }
1272 
1273 static void nfs4_opendata_free(struct kref *kref)
1274 {
1275 	struct nfs4_opendata *p = container_of(kref,
1276 			struct nfs4_opendata, kref);
1277 	struct super_block *sb = p->dentry->d_sb;
1278 
1279 	nfs4_lgopen_release(p->lgp);
1280 	nfs_free_seqid(p->o_arg.seqid);
1281 	nfs4_sequence_free_slot(&p->o_res.seq_res);
1282 	if (p->state != NULL)
1283 		nfs4_put_open_state(p->state);
1284 	nfs4_put_state_owner(p->owner);
1285 
1286 	nfs4_label_free(p->a_label);
1287 	nfs4_label_free(p->f_label);
1288 
1289 	dput(p->dir);
1290 	dput(p->dentry);
1291 	nfs_sb_deactive(sb);
1292 	nfs_fattr_free_names(&p->f_attr);
1293 	kfree(p->f_attr.mdsthreshold);
1294 	kfree(p);
1295 }
1296 
1297 static void nfs4_opendata_put(struct nfs4_opendata *p)
1298 {
1299 	if (p != NULL)
1300 		kref_put(&p->kref, nfs4_opendata_free);
1301 }
1302 
1303 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1304 		fmode_t fmode)
1305 {
1306 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1307 	case FMODE_READ|FMODE_WRITE:
1308 		return state->n_rdwr != 0;
1309 	case FMODE_WRITE:
1310 		return state->n_wronly != 0;
1311 	case FMODE_READ:
1312 		return state->n_rdonly != 0;
1313 	}
1314 	WARN_ON_ONCE(1);
1315 	return false;
1316 }
1317 
1318 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1319 {
1320 	int ret = 0;
1321 
1322 	if (open_mode & (O_EXCL|O_TRUNC))
1323 		goto out;
1324 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1325 		case FMODE_READ:
1326 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1327 				&& state->n_rdonly != 0;
1328 			break;
1329 		case FMODE_WRITE:
1330 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1331 				&& state->n_wronly != 0;
1332 			break;
1333 		case FMODE_READ|FMODE_WRITE:
1334 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1335 				&& state->n_rdwr != 0;
1336 	}
1337 out:
1338 	return ret;
1339 }
1340 
1341 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1342 		enum open_claim_type4 claim)
1343 {
1344 	if (delegation == NULL)
1345 		return 0;
1346 	if ((delegation->type & fmode) != fmode)
1347 		return 0;
1348 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1349 		return 0;
1350 	switch (claim) {
1351 	case NFS4_OPEN_CLAIM_NULL:
1352 	case NFS4_OPEN_CLAIM_FH:
1353 		break;
1354 	case NFS4_OPEN_CLAIM_PREVIOUS:
1355 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1356 			break;
1357 	default:
1358 		return 0;
1359 	}
1360 	nfs_mark_delegation_referenced(delegation);
1361 	return 1;
1362 }
1363 
1364 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1365 {
1366 	switch (fmode) {
1367 		case FMODE_WRITE:
1368 			state->n_wronly++;
1369 			break;
1370 		case FMODE_READ:
1371 			state->n_rdonly++;
1372 			break;
1373 		case FMODE_READ|FMODE_WRITE:
1374 			state->n_rdwr++;
1375 	}
1376 	nfs4_state_set_mode_locked(state, state->state | fmode);
1377 }
1378 
1379 #ifdef CONFIG_NFS_V4_1
1380 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1381 {
1382 	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1383 		return true;
1384 	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1385 		return true;
1386 	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1387 		return true;
1388 	return false;
1389 }
1390 #endif /* CONFIG_NFS_V4_1 */
1391 
1392 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1393 {
1394 	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1395 		wake_up_all(&state->waitq);
1396 }
1397 
1398 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1399 		const nfs4_stateid *stateid)
1400 {
1401 	u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1402 	u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1403 
1404 	if (stateid_seqid == state_seqid + 1U ||
1405 	    (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1406 		nfs_state_log_update_open_stateid(state);
1407 	else
1408 		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1409 }
1410 
1411 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1412 {
1413 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1414 	bool need_recover = false;
1415 
1416 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1417 		need_recover = true;
1418 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1419 		need_recover = true;
1420 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1421 		need_recover = true;
1422 	if (need_recover)
1423 		nfs4_state_mark_reclaim_nograce(clp, state);
1424 }
1425 
1426 /*
1427  * Check for whether or not the caller may update the open stateid
1428  * to the value passed in by stateid.
1429  *
1430  * Note: This function relies heavily on the server implementing
1431  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1432  * correctly.
1433  * i.e. The stateid seqids have to be initialised to 1, and
1434  * are then incremented on every state transition.
1435  */
1436 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1437 		const nfs4_stateid *stateid)
1438 {
1439 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1440 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1441 		if (stateid->seqid == cpu_to_be32(1))
1442 			nfs_state_log_update_open_stateid(state);
1443 		else
1444 			set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1445 		return true;
1446 	}
1447 
1448 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1449 		nfs_state_log_out_of_order_open_stateid(state, stateid);
1450 		return true;
1451 	}
1452 	return false;
1453 }
1454 
1455 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1456 {
1457 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1458 		return;
1459 	if (state->n_wronly)
1460 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1461 	if (state->n_rdonly)
1462 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1463 	if (state->n_rdwr)
1464 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1465 	set_bit(NFS_OPEN_STATE, &state->flags);
1466 }
1467 
1468 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1469 		nfs4_stateid *stateid, fmode_t fmode)
1470 {
1471 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1472 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1473 	case FMODE_WRITE:
1474 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1475 		break;
1476 	case FMODE_READ:
1477 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1478 		break;
1479 	case 0:
1480 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1481 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1482 		clear_bit(NFS_OPEN_STATE, &state->flags);
1483 	}
1484 	if (stateid == NULL)
1485 		return;
1486 	/* Handle OPEN+OPEN_DOWNGRADE races */
1487 	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1488 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1489 		nfs_resync_open_stateid_locked(state);
1490 		goto out;
1491 	}
1492 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1493 		nfs4_stateid_copy(&state->stateid, stateid);
1494 	nfs4_stateid_copy(&state->open_stateid, stateid);
1495 	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1496 out:
1497 	nfs_state_log_update_open_stateid(state);
1498 }
1499 
1500 static void nfs_clear_open_stateid(struct nfs4_state *state,
1501 	nfs4_stateid *arg_stateid,
1502 	nfs4_stateid *stateid, fmode_t fmode)
1503 {
1504 	write_seqlock(&state->seqlock);
1505 	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1506 	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1507 		nfs_clear_open_stateid_locked(state, stateid, fmode);
1508 	write_sequnlock(&state->seqlock);
1509 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1510 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1511 }
1512 
1513 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1514 		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1515 {
1516 	DEFINE_WAIT(wait);
1517 	int status = 0;
1518 	for (;;) {
1519 
1520 		if (!nfs_need_update_open_stateid(state, stateid))
1521 			return;
1522 		if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1523 			break;
1524 		if (status)
1525 			break;
1526 		/* Rely on seqids for serialisation with NFSv4.0 */
1527 		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1528 			break;
1529 
1530 		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1531 		/*
1532 		 * Ensure we process the state changes in the same order
1533 		 * in which the server processed them by delaying the
1534 		 * update of the stateid until we are in sequence.
1535 		 */
1536 		write_sequnlock(&state->seqlock);
1537 		spin_unlock(&state->owner->so_lock);
1538 		rcu_read_unlock();
1539 		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1540 		if (!signal_pending(current)) {
1541 			if (schedule_timeout(5*HZ) == 0)
1542 				status = -EAGAIN;
1543 			else
1544 				status = 0;
1545 		} else
1546 			status = -EINTR;
1547 		finish_wait(&state->waitq, &wait);
1548 		rcu_read_lock();
1549 		spin_lock(&state->owner->so_lock);
1550 		write_seqlock(&state->seqlock);
1551 	}
1552 
1553 	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1554 	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1555 		nfs4_stateid_copy(freeme, &state->open_stateid);
1556 		nfs_test_and_clear_all_open_stateid(state);
1557 	}
1558 
1559 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1560 		nfs4_stateid_copy(&state->stateid, stateid);
1561 	nfs4_stateid_copy(&state->open_stateid, stateid);
1562 	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1563 	nfs_state_log_update_open_stateid(state);
1564 }
1565 
1566 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1567 		const nfs4_stateid *open_stateid,
1568 		fmode_t fmode,
1569 		nfs4_stateid *freeme)
1570 {
1571 	/*
1572 	 * Protect the call to nfs4_state_set_mode_locked and
1573 	 * serialise the stateid update
1574 	 */
1575 	write_seqlock(&state->seqlock);
1576 	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1577 	switch (fmode) {
1578 	case FMODE_READ:
1579 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1580 		break;
1581 	case FMODE_WRITE:
1582 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1583 		break;
1584 	case FMODE_READ|FMODE_WRITE:
1585 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1586 	}
1587 	set_bit(NFS_OPEN_STATE, &state->flags);
1588 	write_sequnlock(&state->seqlock);
1589 }
1590 
1591 static void nfs_state_set_delegation(struct nfs4_state *state,
1592 		const nfs4_stateid *deleg_stateid,
1593 		fmode_t fmode)
1594 {
1595 	/*
1596 	 * Protect the call to nfs4_state_set_mode_locked and
1597 	 * serialise the stateid update
1598 	 */
1599 	write_seqlock(&state->seqlock);
1600 	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1601 	set_bit(NFS_DELEGATED_STATE, &state->flags);
1602 	write_sequnlock(&state->seqlock);
1603 }
1604 
1605 static int update_open_stateid(struct nfs4_state *state,
1606 		const nfs4_stateid *open_stateid,
1607 		const nfs4_stateid *delegation,
1608 		fmode_t fmode)
1609 {
1610 	struct nfs_server *server = NFS_SERVER(state->inode);
1611 	struct nfs_client *clp = server->nfs_client;
1612 	struct nfs_inode *nfsi = NFS_I(state->inode);
1613 	struct nfs_delegation *deleg_cur;
1614 	nfs4_stateid freeme = { };
1615 	int ret = 0;
1616 
1617 	fmode &= (FMODE_READ|FMODE_WRITE);
1618 
1619 	rcu_read_lock();
1620 	spin_lock(&state->owner->so_lock);
1621 	if (open_stateid != NULL) {
1622 		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1623 		ret = 1;
1624 	}
1625 
1626 	deleg_cur = rcu_dereference(nfsi->delegation);
1627 	if (deleg_cur == NULL)
1628 		goto no_delegation;
1629 
1630 	spin_lock(&deleg_cur->lock);
1631 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1632 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1633 	    (deleg_cur->type & fmode) != fmode)
1634 		goto no_delegation_unlock;
1635 
1636 	if (delegation == NULL)
1637 		delegation = &deleg_cur->stateid;
1638 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1639 		goto no_delegation_unlock;
1640 
1641 	nfs_mark_delegation_referenced(deleg_cur);
1642 	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1643 	ret = 1;
1644 no_delegation_unlock:
1645 	spin_unlock(&deleg_cur->lock);
1646 no_delegation:
1647 	if (ret)
1648 		update_open_stateflags(state, fmode);
1649 	spin_unlock(&state->owner->so_lock);
1650 	rcu_read_unlock();
1651 
1652 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1653 		nfs4_schedule_state_manager(clp);
1654 	if (freeme.type != 0)
1655 		nfs4_test_and_free_stateid(server, &freeme,
1656 				state->owner->so_cred);
1657 
1658 	return ret;
1659 }
1660 
1661 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1662 		const nfs4_stateid *stateid)
1663 {
1664 	struct nfs4_state *state = lsp->ls_state;
1665 	bool ret = false;
1666 
1667 	spin_lock(&state->state_lock);
1668 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1669 		goto out_noupdate;
1670 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1671 		goto out_noupdate;
1672 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1673 	ret = true;
1674 out_noupdate:
1675 	spin_unlock(&state->state_lock);
1676 	return ret;
1677 }
1678 
1679 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1680 {
1681 	struct nfs_delegation *delegation;
1682 
1683 	fmode &= FMODE_READ|FMODE_WRITE;
1684 	rcu_read_lock();
1685 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1686 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1687 		rcu_read_unlock();
1688 		return;
1689 	}
1690 	rcu_read_unlock();
1691 	nfs4_inode_return_delegation(inode);
1692 }
1693 
1694 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1695 {
1696 	struct nfs4_state *state = opendata->state;
1697 	struct nfs_inode *nfsi = NFS_I(state->inode);
1698 	struct nfs_delegation *delegation;
1699 	int open_mode = opendata->o_arg.open_flags;
1700 	fmode_t fmode = opendata->o_arg.fmode;
1701 	enum open_claim_type4 claim = opendata->o_arg.claim;
1702 	nfs4_stateid stateid;
1703 	int ret = -EAGAIN;
1704 
1705 	for (;;) {
1706 		spin_lock(&state->owner->so_lock);
1707 		if (can_open_cached(state, fmode, open_mode)) {
1708 			update_open_stateflags(state, fmode);
1709 			spin_unlock(&state->owner->so_lock);
1710 			goto out_return_state;
1711 		}
1712 		spin_unlock(&state->owner->so_lock);
1713 		rcu_read_lock();
1714 		delegation = rcu_dereference(nfsi->delegation);
1715 		if (!can_open_delegated(delegation, fmode, claim)) {
1716 			rcu_read_unlock();
1717 			break;
1718 		}
1719 		/* Save the delegation */
1720 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1721 		rcu_read_unlock();
1722 		nfs_release_seqid(opendata->o_arg.seqid);
1723 		if (!opendata->is_recover) {
1724 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1725 			if (ret != 0)
1726 				goto out;
1727 		}
1728 		ret = -EAGAIN;
1729 
1730 		/* Try to update the stateid using the delegation */
1731 		if (update_open_stateid(state, NULL, &stateid, fmode))
1732 			goto out_return_state;
1733 	}
1734 out:
1735 	return ERR_PTR(ret);
1736 out_return_state:
1737 	atomic_inc(&state->count);
1738 	return state;
1739 }
1740 
1741 static void
1742 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1743 {
1744 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1745 	struct nfs_delegation *delegation;
1746 	int delegation_flags = 0;
1747 
1748 	rcu_read_lock();
1749 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1750 	if (delegation)
1751 		delegation_flags = delegation->flags;
1752 	rcu_read_unlock();
1753 	switch (data->o_arg.claim) {
1754 	default:
1755 		break;
1756 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1757 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1758 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1759 				   "returning a delegation for "
1760 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1761 				   clp->cl_hostname);
1762 		return;
1763 	}
1764 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1765 		nfs_inode_set_delegation(state->inode,
1766 				data->owner->so_cred,
1767 				data->o_res.delegation_type,
1768 				&data->o_res.delegation,
1769 				data->o_res.pagemod_limit);
1770 	else
1771 		nfs_inode_reclaim_delegation(state->inode,
1772 				data->owner->so_cred,
1773 				data->o_res.delegation_type,
1774 				&data->o_res.delegation,
1775 				data->o_res.pagemod_limit);
1776 }
1777 
1778 /*
1779  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1780  * and update the nfs4_state.
1781  */
1782 static struct nfs4_state *
1783 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1784 {
1785 	struct inode *inode = data->state->inode;
1786 	struct nfs4_state *state = data->state;
1787 	int ret;
1788 
1789 	if (!data->rpc_done) {
1790 		if (data->rpc_status)
1791 			return ERR_PTR(data->rpc_status);
1792 		/* cached opens have already been processed */
1793 		goto update;
1794 	}
1795 
1796 	ret = nfs_refresh_inode(inode, &data->f_attr);
1797 	if (ret)
1798 		return ERR_PTR(ret);
1799 
1800 	if (data->o_res.delegation_type != 0)
1801 		nfs4_opendata_check_deleg(data, state);
1802 update:
1803 	update_open_stateid(state, &data->o_res.stateid, NULL,
1804 			    data->o_arg.fmode);
1805 	atomic_inc(&state->count);
1806 
1807 	return state;
1808 }
1809 
1810 static struct inode *
1811 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1812 {
1813 	struct inode *inode;
1814 
1815 	switch (data->o_arg.claim) {
1816 	case NFS4_OPEN_CLAIM_NULL:
1817 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1818 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1819 		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1820 			return ERR_PTR(-EAGAIN);
1821 		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1822 				&data->f_attr, data->f_label);
1823 		break;
1824 	default:
1825 		inode = d_inode(data->dentry);
1826 		ihold(inode);
1827 		nfs_refresh_inode(inode, &data->f_attr);
1828 	}
1829 	return inode;
1830 }
1831 
1832 static struct nfs4_state *
1833 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1834 {
1835 	struct nfs4_state *state;
1836 	struct inode *inode;
1837 
1838 	inode = nfs4_opendata_get_inode(data);
1839 	if (IS_ERR(inode))
1840 		return ERR_CAST(inode);
1841 	if (data->state != NULL && data->state->inode == inode) {
1842 		state = data->state;
1843 		atomic_inc(&state->count);
1844 	} else
1845 		state = nfs4_get_open_state(inode, data->owner);
1846 	iput(inode);
1847 	if (state == NULL)
1848 		state = ERR_PTR(-ENOMEM);
1849 	return state;
1850 }
1851 
1852 static struct nfs4_state *
1853 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1854 {
1855 	struct nfs4_state *state;
1856 
1857 	if (!data->rpc_done) {
1858 		state = nfs4_try_open_cached(data);
1859 		trace_nfs4_cached_open(data->state);
1860 		goto out;
1861 	}
1862 
1863 	state = nfs4_opendata_find_nfs4_state(data);
1864 	if (IS_ERR(state))
1865 		goto out;
1866 
1867 	if (data->o_res.delegation_type != 0)
1868 		nfs4_opendata_check_deleg(data, state);
1869 	update_open_stateid(state, &data->o_res.stateid, NULL,
1870 			data->o_arg.fmode);
1871 out:
1872 	nfs_release_seqid(data->o_arg.seqid);
1873 	return state;
1874 }
1875 
1876 static struct nfs4_state *
1877 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1878 {
1879 	struct nfs4_state *ret;
1880 
1881 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1882 		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1883 	else
1884 		ret = _nfs4_opendata_to_nfs4_state(data);
1885 	nfs4_sequence_free_slot(&data->o_res.seq_res);
1886 	return ret;
1887 }
1888 
1889 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1890 {
1891 	struct nfs_inode *nfsi = NFS_I(state->inode);
1892 	struct nfs_open_context *ctx;
1893 
1894 	spin_lock(&state->inode->i_lock);
1895 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1896 		if (ctx->state != state)
1897 			continue;
1898 		get_nfs_open_context(ctx);
1899 		spin_unlock(&state->inode->i_lock);
1900 		return ctx;
1901 	}
1902 	spin_unlock(&state->inode->i_lock);
1903 	return ERR_PTR(-ENOENT);
1904 }
1905 
1906 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1907 		struct nfs4_state *state, enum open_claim_type4 claim)
1908 {
1909 	struct nfs4_opendata *opendata;
1910 
1911 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1912 			NULL, claim, GFP_NOFS);
1913 	if (opendata == NULL)
1914 		return ERR_PTR(-ENOMEM);
1915 	opendata->state = state;
1916 	atomic_inc(&state->count);
1917 	return opendata;
1918 }
1919 
1920 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1921 		fmode_t fmode)
1922 {
1923 	struct nfs4_state *newstate;
1924 	int ret;
1925 
1926 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1927 		return 0;
1928 	opendata->o_arg.open_flags = 0;
1929 	opendata->o_arg.fmode = fmode;
1930 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1931 			NFS_SB(opendata->dentry->d_sb),
1932 			fmode, 0);
1933 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1934 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1935 	nfs4_init_opendata_res(opendata);
1936 	ret = _nfs4_recover_proc_open(opendata);
1937 	if (ret != 0)
1938 		return ret;
1939 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1940 	if (IS_ERR(newstate))
1941 		return PTR_ERR(newstate);
1942 	if (newstate != opendata->state)
1943 		ret = -ESTALE;
1944 	nfs4_close_state(newstate, fmode);
1945 	return ret;
1946 }
1947 
1948 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1949 {
1950 	int ret;
1951 
1952 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1953 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1954 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1955 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1956 	/* memory barrier prior to reading state->n_* */
1957 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1958 	clear_bit(NFS_OPEN_STATE, &state->flags);
1959 	smp_rmb();
1960 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1961 	if (ret != 0)
1962 		return ret;
1963 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1964 	if (ret != 0)
1965 		return ret;
1966 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1967 	if (ret != 0)
1968 		return ret;
1969 	/*
1970 	 * We may have performed cached opens for all three recoveries.
1971 	 * Check if we need to update the current stateid.
1972 	 */
1973 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1974 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1975 		write_seqlock(&state->seqlock);
1976 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1977 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1978 		write_sequnlock(&state->seqlock);
1979 	}
1980 	return 0;
1981 }
1982 
1983 /*
1984  * OPEN_RECLAIM:
1985  * 	reclaim state on the server after a reboot.
1986  */
1987 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1988 {
1989 	struct nfs_delegation *delegation;
1990 	struct nfs4_opendata *opendata;
1991 	fmode_t delegation_type = 0;
1992 	int status;
1993 
1994 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1995 			NFS4_OPEN_CLAIM_PREVIOUS);
1996 	if (IS_ERR(opendata))
1997 		return PTR_ERR(opendata);
1998 	rcu_read_lock();
1999 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2000 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2001 		delegation_type = delegation->type;
2002 	rcu_read_unlock();
2003 	opendata->o_arg.u.delegation_type = delegation_type;
2004 	status = nfs4_open_recover(opendata, state);
2005 	nfs4_opendata_put(opendata);
2006 	return status;
2007 }
2008 
2009 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2010 {
2011 	struct nfs_server *server = NFS_SERVER(state->inode);
2012 	struct nfs4_exception exception = { };
2013 	int err;
2014 	do {
2015 		err = _nfs4_do_open_reclaim(ctx, state);
2016 		trace_nfs4_open_reclaim(ctx, 0, err);
2017 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2018 			continue;
2019 		if (err != -NFS4ERR_DELAY)
2020 			break;
2021 		nfs4_handle_exception(server, err, &exception);
2022 	} while (exception.retry);
2023 	return err;
2024 }
2025 
2026 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2027 {
2028 	struct nfs_open_context *ctx;
2029 	int ret;
2030 
2031 	ctx = nfs4_state_find_open_context(state);
2032 	if (IS_ERR(ctx))
2033 		return -EAGAIN;
2034 	ret = nfs4_do_open_reclaim(ctx, state);
2035 	put_nfs_open_context(ctx);
2036 	return ret;
2037 }
2038 
2039 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2040 {
2041 	switch (err) {
2042 		default:
2043 			printk(KERN_ERR "NFS: %s: unhandled error "
2044 					"%d.\n", __func__, err);
2045 		case 0:
2046 		case -ENOENT:
2047 		case -EAGAIN:
2048 		case -ESTALE:
2049 			break;
2050 		case -NFS4ERR_BADSESSION:
2051 		case -NFS4ERR_BADSLOT:
2052 		case -NFS4ERR_BAD_HIGH_SLOT:
2053 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2054 		case -NFS4ERR_DEADSESSION:
2055 			set_bit(NFS_DELEGATED_STATE, &state->flags);
2056 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2057 			return -EAGAIN;
2058 		case -NFS4ERR_STALE_CLIENTID:
2059 		case -NFS4ERR_STALE_STATEID:
2060 			set_bit(NFS_DELEGATED_STATE, &state->flags);
2061 			/* Don't recall a delegation if it was lost */
2062 			nfs4_schedule_lease_recovery(server->nfs_client);
2063 			return -EAGAIN;
2064 		case -NFS4ERR_MOVED:
2065 			nfs4_schedule_migration_recovery(server);
2066 			return -EAGAIN;
2067 		case -NFS4ERR_LEASE_MOVED:
2068 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2069 			return -EAGAIN;
2070 		case -NFS4ERR_DELEG_REVOKED:
2071 		case -NFS4ERR_ADMIN_REVOKED:
2072 		case -NFS4ERR_EXPIRED:
2073 		case -NFS4ERR_BAD_STATEID:
2074 		case -NFS4ERR_OPENMODE:
2075 			nfs_inode_find_state_and_recover(state->inode,
2076 					stateid);
2077 			nfs4_schedule_stateid_recovery(server, state);
2078 			return -EAGAIN;
2079 		case -NFS4ERR_DELAY:
2080 		case -NFS4ERR_GRACE:
2081 			set_bit(NFS_DELEGATED_STATE, &state->flags);
2082 			ssleep(1);
2083 			return -EAGAIN;
2084 		case -ENOMEM:
2085 		case -NFS4ERR_DENIED:
2086 			if (fl) {
2087 				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2088 				if (lsp)
2089 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2090 			}
2091 			return 0;
2092 	}
2093 	return err;
2094 }
2095 
2096 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2097 		struct nfs4_state *state, const nfs4_stateid *stateid,
2098 		fmode_t type)
2099 {
2100 	struct nfs_server *server = NFS_SERVER(state->inode);
2101 	struct nfs4_opendata *opendata;
2102 	int err = 0;
2103 
2104 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2105 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2106 	if (IS_ERR(opendata))
2107 		return PTR_ERR(opendata);
2108 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2109 	write_seqlock(&state->seqlock);
2110 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2111 	write_sequnlock(&state->seqlock);
2112 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2113 	switch (type & (FMODE_READ|FMODE_WRITE)) {
2114 	case FMODE_READ|FMODE_WRITE:
2115 	case FMODE_WRITE:
2116 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2117 		if (err)
2118 			break;
2119 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2120 		if (err)
2121 			break;
2122 	case FMODE_READ:
2123 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2124 	}
2125 	nfs4_opendata_put(opendata);
2126 	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2127 }
2128 
2129 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2130 {
2131 	struct nfs4_opendata *data = calldata;
2132 
2133 	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2134 			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2135 }
2136 
2137 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2138 {
2139 	struct nfs4_opendata *data = calldata;
2140 
2141 	nfs40_sequence_done(task, &data->c_res.seq_res);
2142 
2143 	data->rpc_status = task->tk_status;
2144 	if (data->rpc_status == 0) {
2145 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2146 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2147 		renew_lease(data->o_res.server, data->timestamp);
2148 		data->rpc_done = true;
2149 	}
2150 }
2151 
2152 static void nfs4_open_confirm_release(void *calldata)
2153 {
2154 	struct nfs4_opendata *data = calldata;
2155 	struct nfs4_state *state = NULL;
2156 
2157 	/* If this request hasn't been cancelled, do nothing */
2158 	if (!data->cancelled)
2159 		goto out_free;
2160 	/* In case of error, no cleanup! */
2161 	if (!data->rpc_done)
2162 		goto out_free;
2163 	state = nfs4_opendata_to_nfs4_state(data);
2164 	if (!IS_ERR(state))
2165 		nfs4_close_state(state, data->o_arg.fmode);
2166 out_free:
2167 	nfs4_opendata_put(data);
2168 }
2169 
2170 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2171 	.rpc_call_prepare = nfs4_open_confirm_prepare,
2172 	.rpc_call_done = nfs4_open_confirm_done,
2173 	.rpc_release = nfs4_open_confirm_release,
2174 };
2175 
2176 /*
2177  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2178  */
2179 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2180 {
2181 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2182 	struct rpc_task *task;
2183 	struct  rpc_message msg = {
2184 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2185 		.rpc_argp = &data->c_arg,
2186 		.rpc_resp = &data->c_res,
2187 		.rpc_cred = data->owner->so_cred,
2188 	};
2189 	struct rpc_task_setup task_setup_data = {
2190 		.rpc_client = server->client,
2191 		.rpc_message = &msg,
2192 		.callback_ops = &nfs4_open_confirm_ops,
2193 		.callback_data = data,
2194 		.workqueue = nfsiod_workqueue,
2195 		.flags = RPC_TASK_ASYNC,
2196 	};
2197 	int status;
2198 
2199 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2200 				data->is_recover);
2201 	kref_get(&data->kref);
2202 	data->rpc_done = false;
2203 	data->rpc_status = 0;
2204 	data->timestamp = jiffies;
2205 	task = rpc_run_task(&task_setup_data);
2206 	if (IS_ERR(task))
2207 		return PTR_ERR(task);
2208 	status = rpc_wait_for_completion_task(task);
2209 	if (status != 0) {
2210 		data->cancelled = true;
2211 		smp_wmb();
2212 	} else
2213 		status = data->rpc_status;
2214 	rpc_put_task(task);
2215 	return status;
2216 }
2217 
2218 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2219 {
2220 	struct nfs4_opendata *data = calldata;
2221 	struct nfs4_state_owner *sp = data->owner;
2222 	struct nfs_client *clp = sp->so_server->nfs_client;
2223 	enum open_claim_type4 claim = data->o_arg.claim;
2224 
2225 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2226 		goto out_wait;
2227 	/*
2228 	 * Check if we still need to send an OPEN call, or if we can use
2229 	 * a delegation instead.
2230 	 */
2231 	if (data->state != NULL) {
2232 		struct nfs_delegation *delegation;
2233 
2234 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2235 			goto out_no_action;
2236 		rcu_read_lock();
2237 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2238 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2239 			goto unlock_no_action;
2240 		rcu_read_unlock();
2241 	}
2242 	/* Update client id. */
2243 	data->o_arg.clientid = clp->cl_clientid;
2244 	switch (claim) {
2245 	default:
2246 		break;
2247 	case NFS4_OPEN_CLAIM_PREVIOUS:
2248 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2249 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2250 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2251 	case NFS4_OPEN_CLAIM_FH:
2252 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2253 	}
2254 	data->timestamp = jiffies;
2255 	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2256 				&data->o_arg.seq_args,
2257 				&data->o_res.seq_res,
2258 				task) != 0)
2259 		nfs_release_seqid(data->o_arg.seqid);
2260 
2261 	/* Set the create mode (note dependency on the session type) */
2262 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2263 	if (data->o_arg.open_flags & O_EXCL) {
2264 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2265 		if (nfs4_has_persistent_session(clp))
2266 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2267 		else if (clp->cl_mvops->minor_version > 0)
2268 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2269 	}
2270 	return;
2271 unlock_no_action:
2272 	trace_nfs4_cached_open(data->state);
2273 	rcu_read_unlock();
2274 out_no_action:
2275 	task->tk_action = NULL;
2276 out_wait:
2277 	nfs4_sequence_done(task, &data->o_res.seq_res);
2278 }
2279 
2280 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2281 {
2282 	struct nfs4_opendata *data = calldata;
2283 
2284 	data->rpc_status = task->tk_status;
2285 
2286 	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2287 		return;
2288 
2289 	if (task->tk_status == 0) {
2290 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2291 			switch (data->o_res.f_attr->mode & S_IFMT) {
2292 			case S_IFREG:
2293 				break;
2294 			case S_IFLNK:
2295 				data->rpc_status = -ELOOP;
2296 				break;
2297 			case S_IFDIR:
2298 				data->rpc_status = -EISDIR;
2299 				break;
2300 			default:
2301 				data->rpc_status = -ENOTDIR;
2302 			}
2303 		}
2304 		renew_lease(data->o_res.server, data->timestamp);
2305 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2306 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2307 	}
2308 	data->rpc_done = true;
2309 }
2310 
2311 static void nfs4_open_release(void *calldata)
2312 {
2313 	struct nfs4_opendata *data = calldata;
2314 	struct nfs4_state *state = NULL;
2315 
2316 	/* If this request hasn't been cancelled, do nothing */
2317 	if (!data->cancelled)
2318 		goto out_free;
2319 	/* In case of error, no cleanup! */
2320 	if (data->rpc_status != 0 || !data->rpc_done)
2321 		goto out_free;
2322 	/* In case we need an open_confirm, no cleanup! */
2323 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2324 		goto out_free;
2325 	state = nfs4_opendata_to_nfs4_state(data);
2326 	if (!IS_ERR(state))
2327 		nfs4_close_state(state, data->o_arg.fmode);
2328 out_free:
2329 	nfs4_opendata_put(data);
2330 }
2331 
2332 static const struct rpc_call_ops nfs4_open_ops = {
2333 	.rpc_call_prepare = nfs4_open_prepare,
2334 	.rpc_call_done = nfs4_open_done,
2335 	.rpc_release = nfs4_open_release,
2336 };
2337 
2338 static int nfs4_run_open_task(struct nfs4_opendata *data,
2339 			      struct nfs_open_context *ctx)
2340 {
2341 	struct inode *dir = d_inode(data->dir);
2342 	struct nfs_server *server = NFS_SERVER(dir);
2343 	struct nfs_openargs *o_arg = &data->o_arg;
2344 	struct nfs_openres *o_res = &data->o_res;
2345 	struct rpc_task *task;
2346 	struct rpc_message msg = {
2347 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2348 		.rpc_argp = o_arg,
2349 		.rpc_resp = o_res,
2350 		.rpc_cred = data->owner->so_cred,
2351 	};
2352 	struct rpc_task_setup task_setup_data = {
2353 		.rpc_client = server->client,
2354 		.rpc_message = &msg,
2355 		.callback_ops = &nfs4_open_ops,
2356 		.callback_data = data,
2357 		.workqueue = nfsiod_workqueue,
2358 		.flags = RPC_TASK_ASYNC,
2359 	};
2360 	int status;
2361 
2362 	kref_get(&data->kref);
2363 	data->rpc_done = false;
2364 	data->rpc_status = 0;
2365 	data->cancelled = false;
2366 	data->is_recover = false;
2367 	if (!ctx) {
2368 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2369 		data->is_recover = true;
2370 	} else {
2371 		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2372 		pnfs_lgopen_prepare(data, ctx);
2373 	}
2374 	task = rpc_run_task(&task_setup_data);
2375 	if (IS_ERR(task))
2376 		return PTR_ERR(task);
2377 	status = rpc_wait_for_completion_task(task);
2378 	if (status != 0) {
2379 		data->cancelled = true;
2380 		smp_wmb();
2381 	} else
2382 		status = data->rpc_status;
2383 	rpc_put_task(task);
2384 
2385 	return status;
2386 }
2387 
2388 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2389 {
2390 	struct inode *dir = d_inode(data->dir);
2391 	struct nfs_openres *o_res = &data->o_res;
2392 	int status;
2393 
2394 	status = nfs4_run_open_task(data, NULL);
2395 	if (status != 0 || !data->rpc_done)
2396 		return status;
2397 
2398 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2399 
2400 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2401 		status = _nfs4_proc_open_confirm(data);
2402 
2403 	return status;
2404 }
2405 
2406 /*
2407  * Additional permission checks in order to distinguish between an
2408  * open for read, and an open for execute. This works around the
2409  * fact that NFSv4 OPEN treats read and execute permissions as being
2410  * the same.
2411  * Note that in the non-execute case, we want to turn off permission
2412  * checking if we just created a new file (POSIX open() semantics).
2413  */
2414 static int nfs4_opendata_access(struct rpc_cred *cred,
2415 				struct nfs4_opendata *opendata,
2416 				struct nfs4_state *state, fmode_t fmode,
2417 				int openflags)
2418 {
2419 	struct nfs_access_entry cache;
2420 	u32 mask, flags;
2421 
2422 	/* access call failed or for some reason the server doesn't
2423 	 * support any access modes -- defer access call until later */
2424 	if (opendata->o_res.access_supported == 0)
2425 		return 0;
2426 
2427 	mask = 0;
2428 	/*
2429 	 * Use openflags to check for exec, because fmode won't
2430 	 * always have FMODE_EXEC set when file open for exec.
2431 	 */
2432 	if (openflags & __FMODE_EXEC) {
2433 		/* ONLY check for exec rights */
2434 		if (S_ISDIR(state->inode->i_mode))
2435 			mask = NFS4_ACCESS_LOOKUP;
2436 		else
2437 			mask = NFS4_ACCESS_EXECUTE;
2438 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2439 		mask = NFS4_ACCESS_READ;
2440 
2441 	cache.cred = cred;
2442 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2443 	nfs_access_add_cache(state->inode, &cache);
2444 
2445 	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2446 	if ((mask & ~cache.mask & flags) == 0)
2447 		return 0;
2448 
2449 	return -EACCES;
2450 }
2451 
2452 /*
2453  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2454  */
2455 static int _nfs4_proc_open(struct nfs4_opendata *data,
2456 			   struct nfs_open_context *ctx)
2457 {
2458 	struct inode *dir = d_inode(data->dir);
2459 	struct nfs_server *server = NFS_SERVER(dir);
2460 	struct nfs_openargs *o_arg = &data->o_arg;
2461 	struct nfs_openres *o_res = &data->o_res;
2462 	int status;
2463 
2464 	status = nfs4_run_open_task(data, ctx);
2465 	if (!data->rpc_done)
2466 		return status;
2467 	if (status != 0) {
2468 		if (status == -NFS4ERR_BADNAME &&
2469 				!(o_arg->open_flags & O_CREAT))
2470 			return -ENOENT;
2471 		return status;
2472 	}
2473 
2474 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2475 
2476 	if (o_arg->open_flags & O_CREAT) {
2477 		if (o_arg->open_flags & O_EXCL)
2478 			data->file_created = true;
2479 		else if (o_res->cinfo.before != o_res->cinfo.after)
2480 			data->file_created = true;
2481 		if (data->file_created ||
2482 		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2483 			update_changeattr(dir, &o_res->cinfo,
2484 					o_res->f_attr->time_start);
2485 	}
2486 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2487 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2488 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2489 		status = _nfs4_proc_open_confirm(data);
2490 		if (status != 0)
2491 			return status;
2492 	}
2493 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2494 		nfs4_sequence_free_slot(&o_res->seq_res);
2495 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2496 				o_res->f_label, NULL);
2497 	}
2498 	return 0;
2499 }
2500 
2501 /*
2502  * OPEN_EXPIRED:
2503  * 	reclaim state on the server after a network partition.
2504  * 	Assumes caller holds the appropriate lock
2505  */
2506 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2507 {
2508 	struct nfs4_opendata *opendata;
2509 	int ret;
2510 
2511 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2512 			NFS4_OPEN_CLAIM_FH);
2513 	if (IS_ERR(opendata))
2514 		return PTR_ERR(opendata);
2515 	ret = nfs4_open_recover(opendata, state);
2516 	if (ret == -ESTALE)
2517 		d_drop(ctx->dentry);
2518 	nfs4_opendata_put(opendata);
2519 	return ret;
2520 }
2521 
2522 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2523 {
2524 	struct nfs_server *server = NFS_SERVER(state->inode);
2525 	struct nfs4_exception exception = { };
2526 	int err;
2527 
2528 	do {
2529 		err = _nfs4_open_expired(ctx, state);
2530 		trace_nfs4_open_expired(ctx, 0, err);
2531 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2532 			continue;
2533 		switch (err) {
2534 		default:
2535 			goto out;
2536 		case -NFS4ERR_GRACE:
2537 		case -NFS4ERR_DELAY:
2538 			nfs4_handle_exception(server, err, &exception);
2539 			err = 0;
2540 		}
2541 	} while (exception.retry);
2542 out:
2543 	return err;
2544 }
2545 
2546 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2547 {
2548 	struct nfs_open_context *ctx;
2549 	int ret;
2550 
2551 	ctx = nfs4_state_find_open_context(state);
2552 	if (IS_ERR(ctx))
2553 		return -EAGAIN;
2554 	ret = nfs4_do_open_expired(ctx, state);
2555 	put_nfs_open_context(ctx);
2556 	return ret;
2557 }
2558 
2559 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2560 		const nfs4_stateid *stateid)
2561 {
2562 	nfs_remove_bad_delegation(state->inode, stateid);
2563 	write_seqlock(&state->seqlock);
2564 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2565 	write_sequnlock(&state->seqlock);
2566 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2567 }
2568 
2569 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2570 {
2571 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2572 		nfs_finish_clear_delegation_stateid(state, NULL);
2573 }
2574 
2575 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2576 {
2577 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2578 	nfs40_clear_delegation_stateid(state);
2579 	return nfs4_open_expired(sp, state);
2580 }
2581 
2582 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2583 		nfs4_stateid *stateid,
2584 		struct rpc_cred *cred)
2585 {
2586 	return -NFS4ERR_BAD_STATEID;
2587 }
2588 
2589 #if defined(CONFIG_NFS_V4_1)
2590 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2591 		nfs4_stateid *stateid,
2592 		struct rpc_cred *cred)
2593 {
2594 	int status;
2595 
2596 	switch (stateid->type) {
2597 	default:
2598 		break;
2599 	case NFS4_INVALID_STATEID_TYPE:
2600 	case NFS4_SPECIAL_STATEID_TYPE:
2601 		return -NFS4ERR_BAD_STATEID;
2602 	case NFS4_REVOKED_STATEID_TYPE:
2603 		goto out_free;
2604 	}
2605 
2606 	status = nfs41_test_stateid(server, stateid, cred);
2607 	switch (status) {
2608 	case -NFS4ERR_EXPIRED:
2609 	case -NFS4ERR_ADMIN_REVOKED:
2610 	case -NFS4ERR_DELEG_REVOKED:
2611 		break;
2612 	default:
2613 		return status;
2614 	}
2615 out_free:
2616 	/* Ack the revoked state to the server */
2617 	nfs41_free_stateid(server, stateid, cred, true);
2618 	return -NFS4ERR_EXPIRED;
2619 }
2620 
2621 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2622 {
2623 	struct nfs_server *server = NFS_SERVER(state->inode);
2624 	nfs4_stateid stateid;
2625 	struct nfs_delegation *delegation;
2626 	struct rpc_cred *cred;
2627 	int status;
2628 
2629 	/* Get the delegation credential for use by test/free_stateid */
2630 	rcu_read_lock();
2631 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2632 	if (delegation == NULL) {
2633 		rcu_read_unlock();
2634 		return;
2635 	}
2636 
2637 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2638 	if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2639 		!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2640 			&delegation->flags)) {
2641 		rcu_read_unlock();
2642 		nfs_finish_clear_delegation_stateid(state, &stateid);
2643 		return;
2644 	}
2645 
2646 	cred = get_rpccred(delegation->cred);
2647 	rcu_read_unlock();
2648 	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2649 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2650 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2651 		nfs_finish_clear_delegation_stateid(state, &stateid);
2652 
2653 	put_rpccred(cred);
2654 }
2655 
2656 /**
2657  * nfs41_check_expired_locks - possibly free a lock stateid
2658  *
2659  * @state: NFSv4 state for an inode
2660  *
2661  * Returns NFS_OK if recovery for this stateid is now finished.
2662  * Otherwise a negative NFS4ERR value is returned.
2663  */
2664 static int nfs41_check_expired_locks(struct nfs4_state *state)
2665 {
2666 	int status, ret = NFS_OK;
2667 	struct nfs4_lock_state *lsp, *prev = NULL;
2668 	struct nfs_server *server = NFS_SERVER(state->inode);
2669 
2670 	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2671 		goto out;
2672 
2673 	spin_lock(&state->state_lock);
2674 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2675 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2676 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2677 
2678 			refcount_inc(&lsp->ls_count);
2679 			spin_unlock(&state->state_lock);
2680 
2681 			nfs4_put_lock_state(prev);
2682 			prev = lsp;
2683 
2684 			status = nfs41_test_and_free_expired_stateid(server,
2685 					&lsp->ls_stateid,
2686 					cred);
2687 			trace_nfs4_test_lock_stateid(state, lsp, status);
2688 			if (status == -NFS4ERR_EXPIRED ||
2689 			    status == -NFS4ERR_BAD_STATEID) {
2690 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2691 				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2692 				if (!recover_lost_locks)
2693 					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2694 			} else if (status != NFS_OK) {
2695 				ret = status;
2696 				nfs4_put_lock_state(prev);
2697 				goto out;
2698 			}
2699 			spin_lock(&state->state_lock);
2700 		}
2701 	}
2702 	spin_unlock(&state->state_lock);
2703 	nfs4_put_lock_state(prev);
2704 out:
2705 	return ret;
2706 }
2707 
2708 /**
2709  * nfs41_check_open_stateid - possibly free an open stateid
2710  *
2711  * @state: NFSv4 state for an inode
2712  *
2713  * Returns NFS_OK if recovery for this stateid is now finished.
2714  * Otherwise a negative NFS4ERR value is returned.
2715  */
2716 static int nfs41_check_open_stateid(struct nfs4_state *state)
2717 {
2718 	struct nfs_server *server = NFS_SERVER(state->inode);
2719 	nfs4_stateid *stateid = &state->open_stateid;
2720 	struct rpc_cred *cred = state->owner->so_cred;
2721 	int status;
2722 
2723 	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2724 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2725 			if (nfs4_have_delegation(state->inode, state->state))
2726 				return NFS_OK;
2727 			return -NFS4ERR_OPENMODE;
2728 		}
2729 		return -NFS4ERR_BAD_STATEID;
2730 	}
2731 	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2732 	trace_nfs4_test_open_stateid(state, NULL, status);
2733 	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2734 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2735 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2736 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2737 		clear_bit(NFS_OPEN_STATE, &state->flags);
2738 		stateid->type = NFS4_INVALID_STATEID_TYPE;
2739 		return status;
2740 	}
2741 	if (nfs_open_stateid_recover_openmode(state))
2742 		return -NFS4ERR_OPENMODE;
2743 	return NFS_OK;
2744 }
2745 
2746 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2747 {
2748 	int status;
2749 
2750 	nfs41_check_delegation_stateid(state);
2751 	status = nfs41_check_expired_locks(state);
2752 	if (status != NFS_OK)
2753 		return status;
2754 	status = nfs41_check_open_stateid(state);
2755 	if (status != NFS_OK)
2756 		status = nfs4_open_expired(sp, state);
2757 	return status;
2758 }
2759 #endif
2760 
2761 /*
2762  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2763  * fields corresponding to attributes that were used to store the verifier.
2764  * Make sure we clobber those fields in the later setattr call
2765  */
2766 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2767 				struct iattr *sattr, struct nfs4_label **label)
2768 {
2769 	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2770 	__u32 attrset[3];
2771 	unsigned ret;
2772 	unsigned i;
2773 
2774 	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2775 		attrset[i] = opendata->o_res.attrset[i];
2776 		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2777 			attrset[i] &= ~bitmask[i];
2778 	}
2779 
2780 	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2781 		sattr->ia_valid : 0;
2782 
2783 	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2784 		if (sattr->ia_valid & ATTR_ATIME_SET)
2785 			ret |= ATTR_ATIME_SET;
2786 		else
2787 			ret |= ATTR_ATIME;
2788 	}
2789 
2790 	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2791 		if (sattr->ia_valid & ATTR_MTIME_SET)
2792 			ret |= ATTR_MTIME_SET;
2793 		else
2794 			ret |= ATTR_MTIME;
2795 	}
2796 
2797 	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2798 		*label = NULL;
2799 	return ret;
2800 }
2801 
2802 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2803 		fmode_t fmode,
2804 		int flags,
2805 		struct nfs_open_context *ctx)
2806 {
2807 	struct nfs4_state_owner *sp = opendata->owner;
2808 	struct nfs_server *server = sp->so_server;
2809 	struct dentry *dentry;
2810 	struct nfs4_state *state;
2811 	unsigned int seq;
2812 	int ret;
2813 
2814 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2815 
2816 	ret = _nfs4_proc_open(opendata, ctx);
2817 	if (ret != 0)
2818 		goto out;
2819 
2820 	state = _nfs4_opendata_to_nfs4_state(opendata);
2821 	ret = PTR_ERR(state);
2822 	if (IS_ERR(state))
2823 		goto out;
2824 	ctx->state = state;
2825 	if (server->caps & NFS_CAP_POSIX_LOCK)
2826 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2827 	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2828 		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2829 
2830 	dentry = opendata->dentry;
2831 	if (d_really_is_negative(dentry)) {
2832 		struct dentry *alias;
2833 		d_drop(dentry);
2834 		alias = d_exact_alias(dentry, state->inode);
2835 		if (!alias)
2836 			alias = d_splice_alias(igrab(state->inode), dentry);
2837 		/* d_splice_alias() can't fail here - it's a non-directory */
2838 		if (alias) {
2839 			dput(ctx->dentry);
2840 			ctx->dentry = dentry = alias;
2841 		}
2842 		nfs_set_verifier(dentry,
2843 				nfs_save_change_attribute(d_inode(opendata->dir)));
2844 	}
2845 
2846 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2847 	if (ret != 0)
2848 		goto out;
2849 
2850 	if (d_inode(dentry) == state->inode) {
2851 		nfs_inode_attach_open_context(ctx);
2852 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2853 			nfs4_schedule_stateid_recovery(server, state);
2854 		else
2855 			pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2856 	}
2857 
2858 out:
2859 	nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2860 	return ret;
2861 }
2862 
2863 /*
2864  * Returns a referenced nfs4_state
2865  */
2866 static int _nfs4_do_open(struct inode *dir,
2867 			struct nfs_open_context *ctx,
2868 			int flags,
2869 			const struct nfs4_open_createattrs *c,
2870 			int *opened)
2871 {
2872 	struct nfs4_state_owner  *sp;
2873 	struct nfs4_state     *state = NULL;
2874 	struct nfs_server       *server = NFS_SERVER(dir);
2875 	struct nfs4_opendata *opendata;
2876 	struct dentry *dentry = ctx->dentry;
2877 	struct rpc_cred *cred = ctx->cred;
2878 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2879 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2880 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2881 	struct iattr *sattr = c->sattr;
2882 	struct nfs4_label *label = c->label;
2883 	struct nfs4_label *olabel = NULL;
2884 	int status;
2885 
2886 	/* Protect against reboot recovery conflicts */
2887 	status = -ENOMEM;
2888 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2889 	if (sp == NULL) {
2890 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2891 		goto out_err;
2892 	}
2893 	status = nfs4_client_recover_expired_lease(server->nfs_client);
2894 	if (status != 0)
2895 		goto err_put_state_owner;
2896 	if (d_really_is_positive(dentry))
2897 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2898 	status = -ENOMEM;
2899 	if (d_really_is_positive(dentry))
2900 		claim = NFS4_OPEN_CLAIM_FH;
2901 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2902 			c, claim, GFP_KERNEL);
2903 	if (opendata == NULL)
2904 		goto err_put_state_owner;
2905 
2906 	if (label) {
2907 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2908 		if (IS_ERR(olabel)) {
2909 			status = PTR_ERR(olabel);
2910 			goto err_opendata_put;
2911 		}
2912 	}
2913 
2914 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2915 		if (!opendata->f_attr.mdsthreshold) {
2916 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2917 			if (!opendata->f_attr.mdsthreshold)
2918 				goto err_free_label;
2919 		}
2920 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2921 	}
2922 	if (d_really_is_positive(dentry))
2923 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2924 
2925 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2926 	if (status != 0)
2927 		goto err_free_label;
2928 	state = ctx->state;
2929 
2930 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2931 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2932 		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
2933 		/*
2934 		 * send create attributes which was not set by open
2935 		 * with an extra setattr.
2936 		 */
2937 		if (attrs || label) {
2938 			unsigned ia_old = sattr->ia_valid;
2939 
2940 			sattr->ia_valid = attrs;
2941 			nfs_fattr_init(opendata->o_res.f_attr);
2942 			status = nfs4_do_setattr(state->inode, cred,
2943 					opendata->o_res.f_attr, sattr,
2944 					ctx, label, olabel);
2945 			if (status == 0) {
2946 				nfs_setattr_update_inode(state->inode, sattr,
2947 						opendata->o_res.f_attr);
2948 				nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2949 			}
2950 			sattr->ia_valid = ia_old;
2951 		}
2952 	}
2953 	if (opened && opendata->file_created)
2954 		*opened |= FILE_CREATED;
2955 
2956 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2957 		*ctx_th = opendata->f_attr.mdsthreshold;
2958 		opendata->f_attr.mdsthreshold = NULL;
2959 	}
2960 
2961 	nfs4_label_free(olabel);
2962 
2963 	nfs4_opendata_put(opendata);
2964 	nfs4_put_state_owner(sp);
2965 	return 0;
2966 err_free_label:
2967 	nfs4_label_free(olabel);
2968 err_opendata_put:
2969 	nfs4_opendata_put(opendata);
2970 err_put_state_owner:
2971 	nfs4_put_state_owner(sp);
2972 out_err:
2973 	return status;
2974 }
2975 
2976 
2977 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2978 					struct nfs_open_context *ctx,
2979 					int flags,
2980 					struct iattr *sattr,
2981 					struct nfs4_label *label,
2982 					int *opened)
2983 {
2984 	struct nfs_server *server = NFS_SERVER(dir);
2985 	struct nfs4_exception exception = { };
2986 	struct nfs4_state *res;
2987 	struct nfs4_open_createattrs c = {
2988 		.label = label,
2989 		.sattr = sattr,
2990 		.verf = {
2991 			[0] = (__u32)jiffies,
2992 			[1] = (__u32)current->pid,
2993 		},
2994 	};
2995 	int status;
2996 
2997 	do {
2998 		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
2999 		res = ctx->state;
3000 		trace_nfs4_open_file(ctx, flags, status);
3001 		if (status == 0)
3002 			break;
3003 		/* NOTE: BAD_SEQID means the server and client disagree about the
3004 		 * book-keeping w.r.t. state-changing operations
3005 		 * (OPEN/CLOSE/LOCK/LOCKU...)
3006 		 * It is actually a sign of a bug on the client or on the server.
3007 		 *
3008 		 * If we receive a BAD_SEQID error in the particular case of
3009 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3010 		 * have unhashed the old state_owner for us, and that we can
3011 		 * therefore safely retry using a new one. We should still warn
3012 		 * the user though...
3013 		 */
3014 		if (status == -NFS4ERR_BAD_SEQID) {
3015 			pr_warn_ratelimited("NFS: v4 server %s "
3016 					" returned a bad sequence-id error!\n",
3017 					NFS_SERVER(dir)->nfs_client->cl_hostname);
3018 			exception.retry = 1;
3019 			continue;
3020 		}
3021 		/*
3022 		 * BAD_STATEID on OPEN means that the server cancelled our
3023 		 * state before it received the OPEN_CONFIRM.
3024 		 * Recover by retrying the request as per the discussion
3025 		 * on Page 181 of RFC3530.
3026 		 */
3027 		if (status == -NFS4ERR_BAD_STATEID) {
3028 			exception.retry = 1;
3029 			continue;
3030 		}
3031 		if (status == -EAGAIN) {
3032 			/* We must have found a delegation */
3033 			exception.retry = 1;
3034 			continue;
3035 		}
3036 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3037 			continue;
3038 		res = ERR_PTR(nfs4_handle_exception(server,
3039 					status, &exception));
3040 	} while (exception.retry);
3041 	return res;
3042 }
3043 
3044 static int _nfs4_do_setattr(struct inode *inode,
3045 			    struct nfs_setattrargs *arg,
3046 			    struct nfs_setattrres *res,
3047 			    struct rpc_cred *cred,
3048 			    struct nfs_open_context *ctx)
3049 {
3050 	struct nfs_server *server = NFS_SERVER(inode);
3051 	struct rpc_message msg = {
3052 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3053 		.rpc_argp	= arg,
3054 		.rpc_resp	= res,
3055 		.rpc_cred	= cred,
3056 	};
3057 	struct rpc_cred *delegation_cred = NULL;
3058 	unsigned long timestamp = jiffies;
3059 	bool truncate;
3060 	int status;
3061 
3062 	nfs_fattr_init(res->fattr);
3063 
3064 	/* Servers should only apply open mode checks for file size changes */
3065 	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3066 	if (!truncate)
3067 		goto zero_stateid;
3068 
3069 	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3070 		/* Use that stateid */
3071 	} else if (ctx != NULL) {
3072 		struct nfs_lock_context *l_ctx;
3073 		if (!nfs4_valid_open_stateid(ctx->state))
3074 			return -EBADF;
3075 		l_ctx = nfs_get_lock_context(ctx);
3076 		if (IS_ERR(l_ctx))
3077 			return PTR_ERR(l_ctx);
3078 		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3079 						&arg->stateid, &delegation_cred);
3080 		nfs_put_lock_context(l_ctx);
3081 		if (status == -EIO)
3082 			return -EBADF;
3083 	} else {
3084 zero_stateid:
3085 		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3086 	}
3087 	if (delegation_cred)
3088 		msg.rpc_cred = delegation_cred;
3089 
3090 	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3091 
3092 	put_rpccred(delegation_cred);
3093 	if (status == 0 && ctx != NULL)
3094 		renew_lease(server, timestamp);
3095 	trace_nfs4_setattr(inode, &arg->stateid, status);
3096 	return status;
3097 }
3098 
3099 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
3100 			   struct nfs_fattr *fattr, struct iattr *sattr,
3101 			   struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3102 			   struct nfs4_label *olabel)
3103 {
3104 	struct nfs_server *server = NFS_SERVER(inode);
3105 	__u32 bitmask[NFS4_BITMASK_SZ];
3106 	struct nfs4_state *state = ctx ? ctx->state : NULL;
3107 	struct nfs_setattrargs	arg = {
3108 		.fh		= NFS_FH(inode),
3109 		.iap		= sattr,
3110 		.server		= server,
3111 		.bitmask = bitmask,
3112 		.label		= ilabel,
3113 	};
3114 	struct nfs_setattrres  res = {
3115 		.fattr		= fattr,
3116 		.label		= olabel,
3117 		.server		= server,
3118 	};
3119 	struct nfs4_exception exception = {
3120 		.state = state,
3121 		.inode = inode,
3122 		.stateid = &arg.stateid,
3123 	};
3124 	int err;
3125 
3126 	do {
3127 		nfs4_bitmap_copy_adjust_setattr(bitmask,
3128 				nfs4_bitmask(server, olabel),
3129 				inode);
3130 
3131 		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3132 		switch (err) {
3133 		case -NFS4ERR_OPENMODE:
3134 			if (!(sattr->ia_valid & ATTR_SIZE)) {
3135 				pr_warn_once("NFSv4: server %s is incorrectly "
3136 						"applying open mode checks to "
3137 						"a SETATTR that is not "
3138 						"changing file size.\n",
3139 						server->nfs_client->cl_hostname);
3140 			}
3141 			if (state && !(state->state & FMODE_WRITE)) {
3142 				err = -EBADF;
3143 				if (sattr->ia_valid & ATTR_OPEN)
3144 					err = -EACCES;
3145 				goto out;
3146 			}
3147 		}
3148 		err = nfs4_handle_exception(server, err, &exception);
3149 	} while (exception.retry);
3150 out:
3151 	return err;
3152 }
3153 
3154 static bool
3155 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3156 {
3157 	if (inode == NULL || !nfs_have_layout(inode))
3158 		return false;
3159 
3160 	return pnfs_wait_on_layoutreturn(inode, task);
3161 }
3162 
3163 struct nfs4_closedata {
3164 	struct inode *inode;
3165 	struct nfs4_state *state;
3166 	struct nfs_closeargs arg;
3167 	struct nfs_closeres res;
3168 	struct {
3169 		struct nfs4_layoutreturn_args arg;
3170 		struct nfs4_layoutreturn_res res;
3171 		struct nfs4_xdr_opaque_data ld_private;
3172 		u32 roc_barrier;
3173 		bool roc;
3174 	} lr;
3175 	struct nfs_fattr fattr;
3176 	unsigned long timestamp;
3177 };
3178 
3179 static void nfs4_free_closedata(void *data)
3180 {
3181 	struct nfs4_closedata *calldata = data;
3182 	struct nfs4_state_owner *sp = calldata->state->owner;
3183 	struct super_block *sb = calldata->state->inode->i_sb;
3184 
3185 	if (calldata->lr.roc)
3186 		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3187 				calldata->res.lr_ret);
3188 	nfs4_put_open_state(calldata->state);
3189 	nfs_free_seqid(calldata->arg.seqid);
3190 	nfs4_put_state_owner(sp);
3191 	nfs_sb_deactive(sb);
3192 	kfree(calldata);
3193 }
3194 
3195 static void nfs4_close_done(struct rpc_task *task, void *data)
3196 {
3197 	struct nfs4_closedata *calldata = data;
3198 	struct nfs4_state *state = calldata->state;
3199 	struct nfs_server *server = NFS_SERVER(calldata->inode);
3200 	nfs4_stateid *res_stateid = NULL;
3201 	struct nfs4_exception exception = {
3202 		.state = state,
3203 		.inode = calldata->inode,
3204 		.stateid = &calldata->arg.stateid,
3205 	};
3206 
3207 	dprintk("%s: begin!\n", __func__);
3208 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3209 		return;
3210 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3211 
3212 	/* Handle Layoutreturn errors */
3213 	if (calldata->arg.lr_args && task->tk_status != 0) {
3214 		switch (calldata->res.lr_ret) {
3215 		default:
3216 			calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3217 			break;
3218 		case 0:
3219 			calldata->arg.lr_args = NULL;
3220 			calldata->res.lr_res = NULL;
3221 			break;
3222 		case -NFS4ERR_OLD_STATEID:
3223 			if (nfs4_refresh_layout_stateid(&calldata->arg.lr_args->stateid,
3224 						calldata->inode))
3225 				goto lr_restart;
3226 			/* Fallthrough */
3227 		case -NFS4ERR_ADMIN_REVOKED:
3228 		case -NFS4ERR_DELEG_REVOKED:
3229 		case -NFS4ERR_EXPIRED:
3230 		case -NFS4ERR_BAD_STATEID:
3231 		case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3232 		case -NFS4ERR_WRONG_CRED:
3233 			calldata->arg.lr_args = NULL;
3234 			calldata->res.lr_res = NULL;
3235 			goto lr_restart;
3236 		}
3237 	}
3238 
3239 	/* hmm. we are done with the inode, and in the process of freeing
3240 	 * the state_owner. we keep this around to process errors
3241 	 */
3242 	switch (task->tk_status) {
3243 		case 0:
3244 			res_stateid = &calldata->res.stateid;
3245 			renew_lease(server, calldata->timestamp);
3246 			break;
3247 		case -NFS4ERR_ACCESS:
3248 			if (calldata->arg.bitmask != NULL) {
3249 				calldata->arg.bitmask = NULL;
3250 				calldata->res.fattr = NULL;
3251 				goto out_restart;
3252 
3253 			}
3254 			break;
3255 		case -NFS4ERR_OLD_STATEID:
3256 			/* Did we race with OPEN? */
3257 			if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3258 						state))
3259 				goto out_restart;
3260 			goto out_release;
3261 		case -NFS4ERR_ADMIN_REVOKED:
3262 		case -NFS4ERR_STALE_STATEID:
3263 		case -NFS4ERR_EXPIRED:
3264 			nfs4_free_revoked_stateid(server,
3265 					&calldata->arg.stateid,
3266 					task->tk_msg.rpc_cred);
3267 			/* Fallthrough */
3268 		case -NFS4ERR_BAD_STATEID:
3269 			break;
3270 		default:
3271 			task->tk_status = nfs4_async_handle_exception(task,
3272 					server, task->tk_status, &exception);
3273 			if (exception.retry)
3274 				goto out_restart;
3275 	}
3276 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3277 			res_stateid, calldata->arg.fmode);
3278 out_release:
3279 	task->tk_status = 0;
3280 	nfs_release_seqid(calldata->arg.seqid);
3281 	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3282 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3283 	return;
3284 lr_restart:
3285 	calldata->res.lr_ret = 0;
3286 out_restart:
3287 	task->tk_status = 0;
3288 	rpc_restart_call_prepare(task);
3289 	goto out_release;
3290 }
3291 
3292 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3293 {
3294 	struct nfs4_closedata *calldata = data;
3295 	struct nfs4_state *state = calldata->state;
3296 	struct inode *inode = calldata->inode;
3297 	bool is_rdonly, is_wronly, is_rdwr;
3298 	int call_close = 0;
3299 
3300 	dprintk("%s: begin!\n", __func__);
3301 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3302 		goto out_wait;
3303 
3304 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3305 	spin_lock(&state->owner->so_lock);
3306 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3307 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3308 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3309 	/* Calculate the change in open mode */
3310 	calldata->arg.fmode = 0;
3311 	if (state->n_rdwr == 0) {
3312 		if (state->n_rdonly == 0)
3313 			call_close |= is_rdonly;
3314 		else if (is_rdonly)
3315 			calldata->arg.fmode |= FMODE_READ;
3316 		if (state->n_wronly == 0)
3317 			call_close |= is_wronly;
3318 		else if (is_wronly)
3319 			calldata->arg.fmode |= FMODE_WRITE;
3320 		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3321 			call_close |= is_rdwr;
3322 	} else if (is_rdwr)
3323 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3324 
3325 	if (!nfs4_valid_open_stateid(state) ||
3326 	    !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3327 		call_close = 0;
3328 	spin_unlock(&state->owner->so_lock);
3329 
3330 	if (!call_close) {
3331 		/* Note: exit _without_ calling nfs4_close_done */
3332 		goto out_no_action;
3333 	}
3334 
3335 	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3336 		nfs_release_seqid(calldata->arg.seqid);
3337 		goto out_wait;
3338 	}
3339 
3340 	if (calldata->arg.fmode == 0)
3341 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3342 
3343 	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3344 		/* Close-to-open cache consistency revalidation */
3345 		if (!nfs4_have_delegation(inode, FMODE_READ))
3346 			calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3347 		else
3348 			calldata->arg.bitmask = NULL;
3349 	}
3350 
3351 	calldata->arg.share_access =
3352 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3353 				calldata->arg.fmode, 0);
3354 
3355 	if (calldata->res.fattr == NULL)
3356 		calldata->arg.bitmask = NULL;
3357 	else if (calldata->arg.bitmask == NULL)
3358 		calldata->res.fattr = NULL;
3359 	calldata->timestamp = jiffies;
3360 	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3361 				&calldata->arg.seq_args,
3362 				&calldata->res.seq_res,
3363 				task) != 0)
3364 		nfs_release_seqid(calldata->arg.seqid);
3365 	dprintk("%s: done!\n", __func__);
3366 	return;
3367 out_no_action:
3368 	task->tk_action = NULL;
3369 out_wait:
3370 	nfs4_sequence_done(task, &calldata->res.seq_res);
3371 }
3372 
3373 static const struct rpc_call_ops nfs4_close_ops = {
3374 	.rpc_call_prepare = nfs4_close_prepare,
3375 	.rpc_call_done = nfs4_close_done,
3376 	.rpc_release = nfs4_free_closedata,
3377 };
3378 
3379 /*
3380  * It is possible for data to be read/written from a mem-mapped file
3381  * after the sys_close call (which hits the vfs layer as a flush).
3382  * This means that we can't safely call nfsv4 close on a file until
3383  * the inode is cleared. This in turn means that we are not good
3384  * NFSv4 citizens - we do not indicate to the server to update the file's
3385  * share state even when we are done with one of the three share
3386  * stateid's in the inode.
3387  *
3388  * NOTE: Caller must be holding the sp->so_owner semaphore!
3389  */
3390 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3391 {
3392 	struct nfs_server *server = NFS_SERVER(state->inode);
3393 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3394 	struct nfs4_closedata *calldata;
3395 	struct nfs4_state_owner *sp = state->owner;
3396 	struct rpc_task *task;
3397 	struct rpc_message msg = {
3398 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3399 		.rpc_cred = state->owner->so_cred,
3400 	};
3401 	struct rpc_task_setup task_setup_data = {
3402 		.rpc_client = server->client,
3403 		.rpc_message = &msg,
3404 		.callback_ops = &nfs4_close_ops,
3405 		.workqueue = nfsiod_workqueue,
3406 		.flags = RPC_TASK_ASYNC,
3407 	};
3408 	int status = -ENOMEM;
3409 
3410 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3411 		&task_setup_data.rpc_client, &msg);
3412 
3413 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3414 	if (calldata == NULL)
3415 		goto out;
3416 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3417 	calldata->inode = state->inode;
3418 	calldata->state = state;
3419 	calldata->arg.fh = NFS_FH(state->inode);
3420 	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3421 		goto out_free_calldata;
3422 	/* Serialization for the sequence id */
3423 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3424 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3425 	if (IS_ERR(calldata->arg.seqid))
3426 		goto out_free_calldata;
3427 	nfs_fattr_init(&calldata->fattr);
3428 	calldata->arg.fmode = 0;
3429 	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3430 	calldata->res.fattr = &calldata->fattr;
3431 	calldata->res.seqid = calldata->arg.seqid;
3432 	calldata->res.server = server;
3433 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3434 	calldata->lr.roc = pnfs_roc(state->inode,
3435 			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3436 	if (calldata->lr.roc) {
3437 		calldata->arg.lr_args = &calldata->lr.arg;
3438 		calldata->res.lr_res = &calldata->lr.res;
3439 	}
3440 	nfs_sb_active(calldata->inode->i_sb);
3441 
3442 	msg.rpc_argp = &calldata->arg;
3443 	msg.rpc_resp = &calldata->res;
3444 	task_setup_data.callback_data = calldata;
3445 	task = rpc_run_task(&task_setup_data);
3446 	if (IS_ERR(task))
3447 		return PTR_ERR(task);
3448 	status = 0;
3449 	if (wait)
3450 		status = rpc_wait_for_completion_task(task);
3451 	rpc_put_task(task);
3452 	return status;
3453 out_free_calldata:
3454 	kfree(calldata);
3455 out:
3456 	nfs4_put_open_state(state);
3457 	nfs4_put_state_owner(sp);
3458 	return status;
3459 }
3460 
3461 static struct inode *
3462 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3463 		int open_flags, struct iattr *attr, int *opened)
3464 {
3465 	struct nfs4_state *state;
3466 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3467 
3468 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3469 
3470 	/* Protect against concurrent sillydeletes */
3471 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3472 
3473 	nfs4_label_release_security(label);
3474 
3475 	if (IS_ERR(state))
3476 		return ERR_CAST(state);
3477 	return state->inode;
3478 }
3479 
3480 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3481 {
3482 	if (ctx->state == NULL)
3483 		return;
3484 	if (is_sync)
3485 		nfs4_close_sync(ctx->state, ctx->mode);
3486 	else
3487 		nfs4_close_state(ctx->state, ctx->mode);
3488 }
3489 
3490 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3491 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3492 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3493 
3494 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3495 {
3496 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3497 	struct nfs4_server_caps_arg args = {
3498 		.fhandle = fhandle,
3499 		.bitmask = bitmask,
3500 	};
3501 	struct nfs4_server_caps_res res = {};
3502 	struct rpc_message msg = {
3503 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3504 		.rpc_argp = &args,
3505 		.rpc_resp = &res,
3506 	};
3507 	int status;
3508 	int i;
3509 
3510 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3511 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3512 		     FATTR4_WORD0_LINK_SUPPORT |
3513 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3514 		     FATTR4_WORD0_ACLSUPPORT;
3515 	if (minorversion)
3516 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3517 
3518 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3519 	if (status == 0) {
3520 		/* Sanity check the server answers */
3521 		switch (minorversion) {
3522 		case 0:
3523 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3524 			res.attr_bitmask[2] = 0;
3525 			break;
3526 		case 1:
3527 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3528 			break;
3529 		case 2:
3530 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3531 		}
3532 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3533 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3534 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3535 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3536 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3537 				NFS_CAP_CTIME|NFS_CAP_MTIME|
3538 				NFS_CAP_SECURITY_LABEL);
3539 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3540 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3541 			server->caps |= NFS_CAP_ACLS;
3542 		if (res.has_links != 0)
3543 			server->caps |= NFS_CAP_HARDLINKS;
3544 		if (res.has_symlinks != 0)
3545 			server->caps |= NFS_CAP_SYMLINKS;
3546 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3547 			server->caps |= NFS_CAP_FILEID;
3548 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3549 			server->caps |= NFS_CAP_MODE;
3550 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3551 			server->caps |= NFS_CAP_NLINK;
3552 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3553 			server->caps |= NFS_CAP_OWNER;
3554 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3555 			server->caps |= NFS_CAP_OWNER_GROUP;
3556 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3557 			server->caps |= NFS_CAP_ATIME;
3558 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3559 			server->caps |= NFS_CAP_CTIME;
3560 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3561 			server->caps |= NFS_CAP_MTIME;
3562 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3563 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3564 			server->caps |= NFS_CAP_SECURITY_LABEL;
3565 #endif
3566 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3567 				sizeof(server->attr_bitmask));
3568 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3569 
3570 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3571 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3572 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3573 		server->cache_consistency_bitmask[2] = 0;
3574 
3575 		/* Avoid a regression due to buggy server */
3576 		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3577 			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3578 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3579 			sizeof(server->exclcreat_bitmask));
3580 
3581 		server->acl_bitmask = res.acl_bitmask;
3582 		server->fh_expire_type = res.fh_expire_type;
3583 	}
3584 
3585 	return status;
3586 }
3587 
3588 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3589 {
3590 	struct nfs4_exception exception = { };
3591 	int err;
3592 	do {
3593 		err = nfs4_handle_exception(server,
3594 				_nfs4_server_capabilities(server, fhandle),
3595 				&exception);
3596 	} while (exception.retry);
3597 	return err;
3598 }
3599 
3600 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3601 		struct nfs_fsinfo *info)
3602 {
3603 	u32 bitmask[3];
3604 	struct nfs4_lookup_root_arg args = {
3605 		.bitmask = bitmask,
3606 	};
3607 	struct nfs4_lookup_res res = {
3608 		.server = server,
3609 		.fattr = info->fattr,
3610 		.fh = fhandle,
3611 	};
3612 	struct rpc_message msg = {
3613 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3614 		.rpc_argp = &args,
3615 		.rpc_resp = &res,
3616 	};
3617 
3618 	bitmask[0] = nfs4_fattr_bitmap[0];
3619 	bitmask[1] = nfs4_fattr_bitmap[1];
3620 	/*
3621 	 * Process the label in the upcoming getfattr
3622 	 */
3623 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3624 
3625 	nfs_fattr_init(info->fattr);
3626 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3627 }
3628 
3629 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3630 		struct nfs_fsinfo *info)
3631 {
3632 	struct nfs4_exception exception = { };
3633 	int err;
3634 	do {
3635 		err = _nfs4_lookup_root(server, fhandle, info);
3636 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3637 		switch (err) {
3638 		case 0:
3639 		case -NFS4ERR_WRONGSEC:
3640 			goto out;
3641 		default:
3642 			err = nfs4_handle_exception(server, err, &exception);
3643 		}
3644 	} while (exception.retry);
3645 out:
3646 	return err;
3647 }
3648 
3649 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3650 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3651 {
3652 	struct rpc_auth_create_args auth_args = {
3653 		.pseudoflavor = flavor,
3654 	};
3655 	struct rpc_auth *auth;
3656 
3657 	auth = rpcauth_create(&auth_args, server->client);
3658 	if (IS_ERR(auth))
3659 		return -EACCES;
3660 	return nfs4_lookup_root(server, fhandle, info);
3661 }
3662 
3663 /*
3664  * Retry pseudoroot lookup with various security flavors.  We do this when:
3665  *
3666  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3667  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3668  *
3669  * Returns zero on success, or a negative NFS4ERR value, or a
3670  * negative errno value.
3671  */
3672 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3673 			      struct nfs_fsinfo *info)
3674 {
3675 	/* Per 3530bis 15.33.5 */
3676 	static const rpc_authflavor_t flav_array[] = {
3677 		RPC_AUTH_GSS_KRB5P,
3678 		RPC_AUTH_GSS_KRB5I,
3679 		RPC_AUTH_GSS_KRB5,
3680 		RPC_AUTH_UNIX,			/* courtesy */
3681 		RPC_AUTH_NULL,
3682 	};
3683 	int status = -EPERM;
3684 	size_t i;
3685 
3686 	if (server->auth_info.flavor_len > 0) {
3687 		/* try each flavor specified by user */
3688 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3689 			status = nfs4_lookup_root_sec(server, fhandle, info,
3690 						server->auth_info.flavors[i]);
3691 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3692 				continue;
3693 			break;
3694 		}
3695 	} else {
3696 		/* no flavors specified by user, try default list */
3697 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3698 			status = nfs4_lookup_root_sec(server, fhandle, info,
3699 						      flav_array[i]);
3700 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3701 				continue;
3702 			break;
3703 		}
3704 	}
3705 
3706 	/*
3707 	 * -EACCESS could mean that the user doesn't have correct permissions
3708 	 * to access the mount.  It could also mean that we tried to mount
3709 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3710 	 * existing mount programs don't handle -EACCES very well so it should
3711 	 * be mapped to -EPERM instead.
3712 	 */
3713 	if (status == -EACCES)
3714 		status = -EPERM;
3715 	return status;
3716 }
3717 
3718 /**
3719  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3720  * @server: initialized nfs_server handle
3721  * @fhandle: we fill in the pseudo-fs root file handle
3722  * @info: we fill in an FSINFO struct
3723  * @auth_probe: probe the auth flavours
3724  *
3725  * Returns zero on success, or a negative errno.
3726  */
3727 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3728 			 struct nfs_fsinfo *info,
3729 			 bool auth_probe)
3730 {
3731 	int status = 0;
3732 
3733 	if (!auth_probe)
3734 		status = nfs4_lookup_root(server, fhandle, info);
3735 
3736 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3737 		status = server->nfs_client->cl_mvops->find_root_sec(server,
3738 				fhandle, info);
3739 
3740 	if (status == 0)
3741 		status = nfs4_server_capabilities(server, fhandle);
3742 	if (status == 0)
3743 		status = nfs4_do_fsinfo(server, fhandle, info);
3744 
3745 	return nfs4_map_errors(status);
3746 }
3747 
3748 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3749 			      struct nfs_fsinfo *info)
3750 {
3751 	int error;
3752 	struct nfs_fattr *fattr = info->fattr;
3753 	struct nfs4_label *label = NULL;
3754 
3755 	error = nfs4_server_capabilities(server, mntfh);
3756 	if (error < 0) {
3757 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3758 		return error;
3759 	}
3760 
3761 	label = nfs4_label_alloc(server, GFP_KERNEL);
3762 	if (IS_ERR(label))
3763 		return PTR_ERR(label);
3764 
3765 	error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3766 	if (error < 0) {
3767 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3768 		goto err_free_label;
3769 	}
3770 
3771 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3772 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3773 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3774 
3775 err_free_label:
3776 	nfs4_label_free(label);
3777 
3778 	return error;
3779 }
3780 
3781 /*
3782  * Get locations and (maybe) other attributes of a referral.
3783  * Note that we'll actually follow the referral later when
3784  * we detect fsid mismatch in inode revalidation
3785  */
3786 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3787 			     const struct qstr *name, struct nfs_fattr *fattr,
3788 			     struct nfs_fh *fhandle)
3789 {
3790 	int status = -ENOMEM;
3791 	struct page *page = NULL;
3792 	struct nfs4_fs_locations *locations = NULL;
3793 
3794 	page = alloc_page(GFP_KERNEL);
3795 	if (page == NULL)
3796 		goto out;
3797 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3798 	if (locations == NULL)
3799 		goto out;
3800 
3801 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3802 	if (status != 0)
3803 		goto out;
3804 
3805 	/*
3806 	 * If the fsid didn't change, this is a migration event, not a
3807 	 * referral.  Cause us to drop into the exception handler, which
3808 	 * will kick off migration recovery.
3809 	 */
3810 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3811 		dprintk("%s: server did not return a different fsid for"
3812 			" a referral at %s\n", __func__, name->name);
3813 		status = -NFS4ERR_MOVED;
3814 		goto out;
3815 	}
3816 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3817 	nfs_fixup_referral_attributes(&locations->fattr);
3818 
3819 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3820 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3821 	memset(fhandle, 0, sizeof(struct nfs_fh));
3822 out:
3823 	if (page)
3824 		__free_page(page);
3825 	kfree(locations);
3826 	return status;
3827 }
3828 
3829 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3830 				struct nfs_fattr *fattr, struct nfs4_label *label,
3831 				struct inode *inode)
3832 {
3833 	__u32 bitmask[NFS4_BITMASK_SZ];
3834 	struct nfs4_getattr_arg args = {
3835 		.fh = fhandle,
3836 		.bitmask = bitmask,
3837 	};
3838 	struct nfs4_getattr_res res = {
3839 		.fattr = fattr,
3840 		.label = label,
3841 		.server = server,
3842 	};
3843 	struct rpc_message msg = {
3844 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3845 		.rpc_argp = &args,
3846 		.rpc_resp = &res,
3847 	};
3848 
3849 	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3850 
3851 	nfs_fattr_init(fattr);
3852 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3853 }
3854 
3855 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3856 				struct nfs_fattr *fattr, struct nfs4_label *label,
3857 				struct inode *inode)
3858 {
3859 	struct nfs4_exception exception = { };
3860 	int err;
3861 	do {
3862 		err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3863 		trace_nfs4_getattr(server, fhandle, fattr, err);
3864 		err = nfs4_handle_exception(server, err,
3865 				&exception);
3866 	} while (exception.retry);
3867 	return err;
3868 }
3869 
3870 /*
3871  * The file is not closed if it is opened due to the a request to change
3872  * the size of the file. The open call will not be needed once the
3873  * VFS layer lookup-intents are implemented.
3874  *
3875  * Close is called when the inode is destroyed.
3876  * If we haven't opened the file for O_WRONLY, we
3877  * need to in the size_change case to obtain a stateid.
3878  *
3879  * Got race?
3880  * Because OPEN is always done by name in nfsv4, it is
3881  * possible that we opened a different file by the same
3882  * name.  We can recognize this race condition, but we
3883  * can't do anything about it besides returning an error.
3884  *
3885  * This will be fixed with VFS changes (lookup-intent).
3886  */
3887 static int
3888 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3889 		  struct iattr *sattr)
3890 {
3891 	struct inode *inode = d_inode(dentry);
3892 	struct rpc_cred *cred = NULL;
3893 	struct nfs_open_context *ctx = NULL;
3894 	struct nfs4_label *label = NULL;
3895 	int status;
3896 
3897 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3898 	    sattr->ia_valid & ATTR_SIZE &&
3899 	    sattr->ia_size < i_size_read(inode))
3900 		pnfs_commit_and_return_layout(inode);
3901 
3902 	nfs_fattr_init(fattr);
3903 
3904 	/* Deal with open(O_TRUNC) */
3905 	if (sattr->ia_valid & ATTR_OPEN)
3906 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3907 
3908 	/* Optimization: if the end result is no change, don't RPC */
3909 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3910 		return 0;
3911 
3912 	/* Search for an existing open(O_WRITE) file */
3913 	if (sattr->ia_valid & ATTR_FILE) {
3914 
3915 		ctx = nfs_file_open_context(sattr->ia_file);
3916 		if (ctx)
3917 			cred = ctx->cred;
3918 	}
3919 
3920 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3921 	if (IS_ERR(label))
3922 		return PTR_ERR(label);
3923 
3924 	/* Return any delegations if we're going to change ACLs */
3925 	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
3926 		nfs4_inode_make_writeable(inode);
3927 
3928 	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3929 	if (status == 0) {
3930 		nfs_setattr_update_inode(inode, sattr, fattr);
3931 		nfs_setsecurity(inode, fattr, label);
3932 	}
3933 	nfs4_label_free(label);
3934 	return status;
3935 }
3936 
3937 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3938 		const struct qstr *name, struct nfs_fh *fhandle,
3939 		struct nfs_fattr *fattr, struct nfs4_label *label)
3940 {
3941 	struct nfs_server *server = NFS_SERVER(dir);
3942 	int		       status;
3943 	struct nfs4_lookup_arg args = {
3944 		.bitmask = server->attr_bitmask,
3945 		.dir_fh = NFS_FH(dir),
3946 		.name = name,
3947 	};
3948 	struct nfs4_lookup_res res = {
3949 		.server = server,
3950 		.fattr = fattr,
3951 		.label = label,
3952 		.fh = fhandle,
3953 	};
3954 	struct rpc_message msg = {
3955 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3956 		.rpc_argp = &args,
3957 		.rpc_resp = &res,
3958 	};
3959 
3960 	args.bitmask = nfs4_bitmask(server, label);
3961 
3962 	nfs_fattr_init(fattr);
3963 
3964 	dprintk("NFS call  lookup %s\n", name->name);
3965 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3966 	dprintk("NFS reply lookup: %d\n", status);
3967 	return status;
3968 }
3969 
3970 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3971 {
3972 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3973 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3974 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3975 	fattr->nlink = 2;
3976 }
3977 
3978 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3979 				   const struct qstr *name, struct nfs_fh *fhandle,
3980 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3981 {
3982 	struct nfs4_exception exception = { };
3983 	struct rpc_clnt *client = *clnt;
3984 	int err;
3985 	do {
3986 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3987 		trace_nfs4_lookup(dir, name, err);
3988 		switch (err) {
3989 		case -NFS4ERR_BADNAME:
3990 			err = -ENOENT;
3991 			goto out;
3992 		case -NFS4ERR_MOVED:
3993 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3994 			if (err == -NFS4ERR_MOVED)
3995 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3996 			goto out;
3997 		case -NFS4ERR_WRONGSEC:
3998 			err = -EPERM;
3999 			if (client != *clnt)
4000 				goto out;
4001 			client = nfs4_negotiate_security(client, dir, name);
4002 			if (IS_ERR(client))
4003 				return PTR_ERR(client);
4004 
4005 			exception.retry = 1;
4006 			break;
4007 		default:
4008 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4009 		}
4010 	} while (exception.retry);
4011 
4012 out:
4013 	if (err == 0)
4014 		*clnt = client;
4015 	else if (client != *clnt)
4016 		rpc_shutdown_client(client);
4017 
4018 	return err;
4019 }
4020 
4021 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4022 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4023 			    struct nfs4_label *label)
4024 {
4025 	int status;
4026 	struct rpc_clnt *client = NFS_CLIENT(dir);
4027 
4028 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4029 	if (client != NFS_CLIENT(dir)) {
4030 		rpc_shutdown_client(client);
4031 		nfs_fixup_secinfo_attributes(fattr);
4032 	}
4033 	return status;
4034 }
4035 
4036 struct rpc_clnt *
4037 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4038 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4039 {
4040 	struct rpc_clnt *client = NFS_CLIENT(dir);
4041 	int status;
4042 
4043 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4044 	if (status < 0)
4045 		return ERR_PTR(status);
4046 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4047 }
4048 
4049 static int _nfs4_proc_lookupp(struct inode *inode,
4050 		struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4051 		struct nfs4_label *label)
4052 {
4053 	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4054 	struct nfs_server *server = NFS_SERVER(inode);
4055 	int		       status;
4056 	struct nfs4_lookupp_arg args = {
4057 		.bitmask = server->attr_bitmask,
4058 		.fh = NFS_FH(inode),
4059 	};
4060 	struct nfs4_lookupp_res res = {
4061 		.server = server,
4062 		.fattr = fattr,
4063 		.label = label,
4064 		.fh = fhandle,
4065 	};
4066 	struct rpc_message msg = {
4067 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4068 		.rpc_argp = &args,
4069 		.rpc_resp = &res,
4070 	};
4071 
4072 	args.bitmask = nfs4_bitmask(server, label);
4073 
4074 	nfs_fattr_init(fattr);
4075 
4076 	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4077 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4078 				&res.seq_res, 0);
4079 	dprintk("NFS reply lookupp: %d\n", status);
4080 	return status;
4081 }
4082 
4083 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4084 			     struct nfs_fattr *fattr, struct nfs4_label *label)
4085 {
4086 	struct nfs4_exception exception = { };
4087 	int err;
4088 	do {
4089 		err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4090 		trace_nfs4_lookupp(inode, err);
4091 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4092 				&exception);
4093 	} while (exception.retry);
4094 	return err;
4095 }
4096 
4097 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4098 {
4099 	struct nfs_server *server = NFS_SERVER(inode);
4100 	struct nfs4_accessargs args = {
4101 		.fh = NFS_FH(inode),
4102 		.access = entry->mask,
4103 	};
4104 	struct nfs4_accessres res = {
4105 		.server = server,
4106 	};
4107 	struct rpc_message msg = {
4108 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4109 		.rpc_argp = &args,
4110 		.rpc_resp = &res,
4111 		.rpc_cred = entry->cred,
4112 	};
4113 	int status = 0;
4114 
4115 	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4116 		res.fattr = nfs_alloc_fattr();
4117 		if (res.fattr == NULL)
4118 			return -ENOMEM;
4119 		args.bitmask = server->cache_consistency_bitmask;
4120 	}
4121 
4122 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4123 	if (!status) {
4124 		nfs_access_set_mask(entry, res.access);
4125 		if (res.fattr)
4126 			nfs_refresh_inode(inode, res.fattr);
4127 	}
4128 	nfs_free_fattr(res.fattr);
4129 	return status;
4130 }
4131 
4132 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4133 {
4134 	struct nfs4_exception exception = { };
4135 	int err;
4136 	do {
4137 		err = _nfs4_proc_access(inode, entry);
4138 		trace_nfs4_access(inode, err);
4139 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4140 				&exception);
4141 	} while (exception.retry);
4142 	return err;
4143 }
4144 
4145 /*
4146  * TODO: For the time being, we don't try to get any attributes
4147  * along with any of the zero-copy operations READ, READDIR,
4148  * READLINK, WRITE.
4149  *
4150  * In the case of the first three, we want to put the GETATTR
4151  * after the read-type operation -- this is because it is hard
4152  * to predict the length of a GETATTR response in v4, and thus
4153  * align the READ data correctly.  This means that the GETATTR
4154  * may end up partially falling into the page cache, and we should
4155  * shift it into the 'tail' of the xdr_buf before processing.
4156  * To do this efficiently, we need to know the total length
4157  * of data received, which doesn't seem to be available outside
4158  * of the RPC layer.
4159  *
4160  * In the case of WRITE, we also want to put the GETATTR after
4161  * the operation -- in this case because we want to make sure
4162  * we get the post-operation mtime and size.
4163  *
4164  * Both of these changes to the XDR layer would in fact be quite
4165  * minor, but I decided to leave them for a subsequent patch.
4166  */
4167 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4168 		unsigned int pgbase, unsigned int pglen)
4169 {
4170 	struct nfs4_readlink args = {
4171 		.fh       = NFS_FH(inode),
4172 		.pgbase	  = pgbase,
4173 		.pglen    = pglen,
4174 		.pages    = &page,
4175 	};
4176 	struct nfs4_readlink_res res;
4177 	struct rpc_message msg = {
4178 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4179 		.rpc_argp = &args,
4180 		.rpc_resp = &res,
4181 	};
4182 
4183 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4184 }
4185 
4186 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4187 		unsigned int pgbase, unsigned int pglen)
4188 {
4189 	struct nfs4_exception exception = { };
4190 	int err;
4191 	do {
4192 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4193 		trace_nfs4_readlink(inode, err);
4194 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4195 				&exception);
4196 	} while (exception.retry);
4197 	return err;
4198 }
4199 
4200 /*
4201  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4202  */
4203 static int
4204 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4205 		 int flags)
4206 {
4207 	struct nfs_server *server = NFS_SERVER(dir);
4208 	struct nfs4_label l, *ilabel = NULL;
4209 	struct nfs_open_context *ctx;
4210 	struct nfs4_state *state;
4211 	int status = 0;
4212 
4213 	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4214 	if (IS_ERR(ctx))
4215 		return PTR_ERR(ctx);
4216 
4217 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4218 
4219 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4220 		sattr->ia_mode &= ~current_umask();
4221 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4222 	if (IS_ERR(state)) {
4223 		status = PTR_ERR(state);
4224 		goto out;
4225 	}
4226 out:
4227 	nfs4_label_release_security(ilabel);
4228 	put_nfs_open_context(ctx);
4229 	return status;
4230 }
4231 
4232 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4233 {
4234 	struct nfs_server *server = NFS_SERVER(dir);
4235 	struct nfs_removeargs args = {
4236 		.fh = NFS_FH(dir),
4237 		.name = *name,
4238 	};
4239 	struct nfs_removeres res = {
4240 		.server = server,
4241 	};
4242 	struct rpc_message msg = {
4243 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4244 		.rpc_argp = &args,
4245 		.rpc_resp = &res,
4246 	};
4247 	unsigned long timestamp = jiffies;
4248 	int status;
4249 
4250 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4251 	if (status == 0)
4252 		update_changeattr(dir, &res.cinfo, timestamp);
4253 	return status;
4254 }
4255 
4256 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4257 {
4258 	struct nfs4_exception exception = { };
4259 	struct inode *inode = d_inode(dentry);
4260 	int err;
4261 
4262 	if (inode) {
4263 		if (inode->i_nlink == 1)
4264 			nfs4_inode_return_delegation(inode);
4265 		else
4266 			nfs4_inode_make_writeable(inode);
4267 	}
4268 	do {
4269 		err = _nfs4_proc_remove(dir, &dentry->d_name);
4270 		trace_nfs4_remove(dir, &dentry->d_name, err);
4271 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4272 				&exception);
4273 	} while (exception.retry);
4274 	return err;
4275 }
4276 
4277 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4278 {
4279 	struct nfs4_exception exception = { };
4280 	int err;
4281 
4282 	do {
4283 		err = _nfs4_proc_remove(dir, name);
4284 		trace_nfs4_remove(dir, name, err);
4285 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4286 				&exception);
4287 	} while (exception.retry);
4288 	return err;
4289 }
4290 
4291 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4292 		struct dentry *dentry,
4293 		struct inode *inode)
4294 {
4295 	struct nfs_removeargs *args = msg->rpc_argp;
4296 	struct nfs_removeres *res = msg->rpc_resp;
4297 
4298 	res->server = NFS_SB(dentry->d_sb);
4299 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4300 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4301 
4302 	nfs_fattr_init(res->dir_attr);
4303 
4304 	if (inode)
4305 		nfs4_inode_return_delegation(inode);
4306 }
4307 
4308 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4309 {
4310 	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4311 			&data->args.seq_args,
4312 			&data->res.seq_res,
4313 			task);
4314 }
4315 
4316 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4317 {
4318 	struct nfs_unlinkdata *data = task->tk_calldata;
4319 	struct nfs_removeres *res = &data->res;
4320 
4321 	if (!nfs4_sequence_done(task, &res->seq_res))
4322 		return 0;
4323 	if (nfs4_async_handle_error(task, res->server, NULL,
4324 				    &data->timeout) == -EAGAIN)
4325 		return 0;
4326 	if (task->tk_status == 0)
4327 		update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4328 	return 1;
4329 }
4330 
4331 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4332 		struct dentry *old_dentry,
4333 		struct dentry *new_dentry)
4334 {
4335 	struct nfs_renameargs *arg = msg->rpc_argp;
4336 	struct nfs_renameres *res = msg->rpc_resp;
4337 	struct inode *old_inode = d_inode(old_dentry);
4338 	struct inode *new_inode = d_inode(new_dentry);
4339 
4340 	if (old_inode)
4341 		nfs4_inode_make_writeable(old_inode);
4342 	if (new_inode)
4343 		nfs4_inode_return_delegation(new_inode);
4344 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4345 	res->server = NFS_SB(old_dentry->d_sb);
4346 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4347 }
4348 
4349 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4350 {
4351 	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4352 			&data->args.seq_args,
4353 			&data->res.seq_res,
4354 			task);
4355 }
4356 
4357 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4358 				 struct inode *new_dir)
4359 {
4360 	struct nfs_renamedata *data = task->tk_calldata;
4361 	struct nfs_renameres *res = &data->res;
4362 
4363 	if (!nfs4_sequence_done(task, &res->seq_res))
4364 		return 0;
4365 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4366 		return 0;
4367 
4368 	if (task->tk_status == 0) {
4369 		update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4370 		if (new_dir != old_dir)
4371 			update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4372 	}
4373 	return 1;
4374 }
4375 
4376 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4377 {
4378 	struct nfs_server *server = NFS_SERVER(inode);
4379 	__u32 bitmask[NFS4_BITMASK_SZ];
4380 	struct nfs4_link_arg arg = {
4381 		.fh     = NFS_FH(inode),
4382 		.dir_fh = NFS_FH(dir),
4383 		.name   = name,
4384 		.bitmask = bitmask,
4385 	};
4386 	struct nfs4_link_res res = {
4387 		.server = server,
4388 		.label = NULL,
4389 	};
4390 	struct rpc_message msg = {
4391 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4392 		.rpc_argp = &arg,
4393 		.rpc_resp = &res,
4394 	};
4395 	int status = -ENOMEM;
4396 
4397 	res.fattr = nfs_alloc_fattr();
4398 	if (res.fattr == NULL)
4399 		goto out;
4400 
4401 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
4402 	if (IS_ERR(res.label)) {
4403 		status = PTR_ERR(res.label);
4404 		goto out;
4405 	}
4406 
4407 	nfs4_inode_make_writeable(inode);
4408 	nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4409 
4410 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4411 	if (!status) {
4412 		update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4413 		status = nfs_post_op_update_inode(inode, res.fattr);
4414 		if (!status)
4415 			nfs_setsecurity(inode, res.fattr, res.label);
4416 	}
4417 
4418 
4419 	nfs4_label_free(res.label);
4420 
4421 out:
4422 	nfs_free_fattr(res.fattr);
4423 	return status;
4424 }
4425 
4426 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4427 {
4428 	struct nfs4_exception exception = { };
4429 	int err;
4430 	do {
4431 		err = nfs4_handle_exception(NFS_SERVER(inode),
4432 				_nfs4_proc_link(inode, dir, name),
4433 				&exception);
4434 	} while (exception.retry);
4435 	return err;
4436 }
4437 
4438 struct nfs4_createdata {
4439 	struct rpc_message msg;
4440 	struct nfs4_create_arg arg;
4441 	struct nfs4_create_res res;
4442 	struct nfs_fh fh;
4443 	struct nfs_fattr fattr;
4444 	struct nfs4_label *label;
4445 };
4446 
4447 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4448 		const struct qstr *name, struct iattr *sattr, u32 ftype)
4449 {
4450 	struct nfs4_createdata *data;
4451 
4452 	data = kzalloc(sizeof(*data), GFP_KERNEL);
4453 	if (data != NULL) {
4454 		struct nfs_server *server = NFS_SERVER(dir);
4455 
4456 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
4457 		if (IS_ERR(data->label))
4458 			goto out_free;
4459 
4460 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4461 		data->msg.rpc_argp = &data->arg;
4462 		data->msg.rpc_resp = &data->res;
4463 		data->arg.dir_fh = NFS_FH(dir);
4464 		data->arg.server = server;
4465 		data->arg.name = name;
4466 		data->arg.attrs = sattr;
4467 		data->arg.ftype = ftype;
4468 		data->arg.bitmask = nfs4_bitmask(server, data->label);
4469 		data->arg.umask = current_umask();
4470 		data->res.server = server;
4471 		data->res.fh = &data->fh;
4472 		data->res.fattr = &data->fattr;
4473 		data->res.label = data->label;
4474 		nfs_fattr_init(data->res.fattr);
4475 	}
4476 	return data;
4477 out_free:
4478 	kfree(data);
4479 	return NULL;
4480 }
4481 
4482 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4483 {
4484 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4485 				    &data->arg.seq_args, &data->res.seq_res, 1);
4486 	if (status == 0) {
4487 		update_changeattr(dir, &data->res.dir_cinfo,
4488 				data->res.fattr->time_start);
4489 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4490 	}
4491 	return status;
4492 }
4493 
4494 static void nfs4_free_createdata(struct nfs4_createdata *data)
4495 {
4496 	nfs4_label_free(data->label);
4497 	kfree(data);
4498 }
4499 
4500 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4501 		struct page *page, unsigned int len, struct iattr *sattr,
4502 		struct nfs4_label *label)
4503 {
4504 	struct nfs4_createdata *data;
4505 	int status = -ENAMETOOLONG;
4506 
4507 	if (len > NFS4_MAXPATHLEN)
4508 		goto out;
4509 
4510 	status = -ENOMEM;
4511 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4512 	if (data == NULL)
4513 		goto out;
4514 
4515 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4516 	data->arg.u.symlink.pages = &page;
4517 	data->arg.u.symlink.len = len;
4518 	data->arg.label = label;
4519 
4520 	status = nfs4_do_create(dir, dentry, data);
4521 
4522 	nfs4_free_createdata(data);
4523 out:
4524 	return status;
4525 }
4526 
4527 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4528 		struct page *page, unsigned int len, struct iattr *sattr)
4529 {
4530 	struct nfs4_exception exception = { };
4531 	struct nfs4_label l, *label = NULL;
4532 	int err;
4533 
4534 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4535 
4536 	do {
4537 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4538 		trace_nfs4_symlink(dir, &dentry->d_name, err);
4539 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4540 				&exception);
4541 	} while (exception.retry);
4542 
4543 	nfs4_label_release_security(label);
4544 	return err;
4545 }
4546 
4547 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4548 		struct iattr *sattr, struct nfs4_label *label)
4549 {
4550 	struct nfs4_createdata *data;
4551 	int status = -ENOMEM;
4552 
4553 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4554 	if (data == NULL)
4555 		goto out;
4556 
4557 	data->arg.label = label;
4558 	status = nfs4_do_create(dir, dentry, data);
4559 
4560 	nfs4_free_createdata(data);
4561 out:
4562 	return status;
4563 }
4564 
4565 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4566 		struct iattr *sattr)
4567 {
4568 	struct nfs_server *server = NFS_SERVER(dir);
4569 	struct nfs4_exception exception = { };
4570 	struct nfs4_label l, *label = NULL;
4571 	int err;
4572 
4573 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4574 
4575 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4576 		sattr->ia_mode &= ~current_umask();
4577 	do {
4578 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4579 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
4580 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4581 				&exception);
4582 	} while (exception.retry);
4583 	nfs4_label_release_security(label);
4584 
4585 	return err;
4586 }
4587 
4588 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4589 		u64 cookie, struct page **pages, unsigned int count, bool plus)
4590 {
4591 	struct inode		*dir = d_inode(dentry);
4592 	struct nfs4_readdir_arg args = {
4593 		.fh = NFS_FH(dir),
4594 		.pages = pages,
4595 		.pgbase = 0,
4596 		.count = count,
4597 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4598 		.plus = plus,
4599 	};
4600 	struct nfs4_readdir_res res;
4601 	struct rpc_message msg = {
4602 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4603 		.rpc_argp = &args,
4604 		.rpc_resp = &res,
4605 		.rpc_cred = cred,
4606 	};
4607 	int			status;
4608 
4609 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4610 			dentry,
4611 			(unsigned long long)cookie);
4612 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4613 	res.pgbase = args.pgbase;
4614 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4615 	if (status >= 0) {
4616 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4617 		status += args.pgbase;
4618 	}
4619 
4620 	nfs_invalidate_atime(dir);
4621 
4622 	dprintk("%s: returns %d\n", __func__, status);
4623 	return status;
4624 }
4625 
4626 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4627 		u64 cookie, struct page **pages, unsigned int count, bool plus)
4628 {
4629 	struct nfs4_exception exception = { };
4630 	int err;
4631 	do {
4632 		err = _nfs4_proc_readdir(dentry, cred, cookie,
4633 				pages, count, plus);
4634 		trace_nfs4_readdir(d_inode(dentry), err);
4635 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4636 				&exception);
4637 	} while (exception.retry);
4638 	return err;
4639 }
4640 
4641 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4642 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4643 {
4644 	struct nfs4_createdata *data;
4645 	int mode = sattr->ia_mode;
4646 	int status = -ENOMEM;
4647 
4648 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4649 	if (data == NULL)
4650 		goto out;
4651 
4652 	if (S_ISFIFO(mode))
4653 		data->arg.ftype = NF4FIFO;
4654 	else if (S_ISBLK(mode)) {
4655 		data->arg.ftype = NF4BLK;
4656 		data->arg.u.device.specdata1 = MAJOR(rdev);
4657 		data->arg.u.device.specdata2 = MINOR(rdev);
4658 	}
4659 	else if (S_ISCHR(mode)) {
4660 		data->arg.ftype = NF4CHR;
4661 		data->arg.u.device.specdata1 = MAJOR(rdev);
4662 		data->arg.u.device.specdata2 = MINOR(rdev);
4663 	} else if (!S_ISSOCK(mode)) {
4664 		status = -EINVAL;
4665 		goto out_free;
4666 	}
4667 
4668 	data->arg.label = label;
4669 	status = nfs4_do_create(dir, dentry, data);
4670 out_free:
4671 	nfs4_free_createdata(data);
4672 out:
4673 	return status;
4674 }
4675 
4676 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4677 		struct iattr *sattr, dev_t rdev)
4678 {
4679 	struct nfs_server *server = NFS_SERVER(dir);
4680 	struct nfs4_exception exception = { };
4681 	struct nfs4_label l, *label = NULL;
4682 	int err;
4683 
4684 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4685 
4686 	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4687 		sattr->ia_mode &= ~current_umask();
4688 	do {
4689 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4690 		trace_nfs4_mknod(dir, &dentry->d_name, err);
4691 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4692 				&exception);
4693 	} while (exception.retry);
4694 
4695 	nfs4_label_release_security(label);
4696 
4697 	return err;
4698 }
4699 
4700 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4701 		 struct nfs_fsstat *fsstat)
4702 {
4703 	struct nfs4_statfs_arg args = {
4704 		.fh = fhandle,
4705 		.bitmask = server->attr_bitmask,
4706 	};
4707 	struct nfs4_statfs_res res = {
4708 		.fsstat = fsstat,
4709 	};
4710 	struct rpc_message msg = {
4711 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4712 		.rpc_argp = &args,
4713 		.rpc_resp = &res,
4714 	};
4715 
4716 	nfs_fattr_init(fsstat->fattr);
4717 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4718 }
4719 
4720 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4721 {
4722 	struct nfs4_exception exception = { };
4723 	int err;
4724 	do {
4725 		err = nfs4_handle_exception(server,
4726 				_nfs4_proc_statfs(server, fhandle, fsstat),
4727 				&exception);
4728 	} while (exception.retry);
4729 	return err;
4730 }
4731 
4732 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4733 		struct nfs_fsinfo *fsinfo)
4734 {
4735 	struct nfs4_fsinfo_arg args = {
4736 		.fh = fhandle,
4737 		.bitmask = server->attr_bitmask,
4738 	};
4739 	struct nfs4_fsinfo_res res = {
4740 		.fsinfo = fsinfo,
4741 	};
4742 	struct rpc_message msg = {
4743 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4744 		.rpc_argp = &args,
4745 		.rpc_resp = &res,
4746 	};
4747 
4748 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4749 }
4750 
4751 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4752 {
4753 	struct nfs4_exception exception = { };
4754 	unsigned long now = jiffies;
4755 	int err;
4756 
4757 	do {
4758 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4759 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4760 		if (err == 0) {
4761 			nfs4_set_lease_period(server->nfs_client,
4762 					fsinfo->lease_time * HZ,
4763 					now);
4764 			break;
4765 		}
4766 		err = nfs4_handle_exception(server, err, &exception);
4767 	} while (exception.retry);
4768 	return err;
4769 }
4770 
4771 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4772 {
4773 	int error;
4774 
4775 	nfs_fattr_init(fsinfo->fattr);
4776 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4777 	if (error == 0) {
4778 		/* block layout checks this! */
4779 		server->pnfs_blksize = fsinfo->blksize;
4780 		set_pnfs_layoutdriver(server, fhandle, fsinfo);
4781 	}
4782 
4783 	return error;
4784 }
4785 
4786 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4787 		struct nfs_pathconf *pathconf)
4788 {
4789 	struct nfs4_pathconf_arg args = {
4790 		.fh = fhandle,
4791 		.bitmask = server->attr_bitmask,
4792 	};
4793 	struct nfs4_pathconf_res res = {
4794 		.pathconf = pathconf,
4795 	};
4796 	struct rpc_message msg = {
4797 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4798 		.rpc_argp = &args,
4799 		.rpc_resp = &res,
4800 	};
4801 
4802 	/* None of the pathconf attributes are mandatory to implement */
4803 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4804 		memset(pathconf, 0, sizeof(*pathconf));
4805 		return 0;
4806 	}
4807 
4808 	nfs_fattr_init(pathconf->fattr);
4809 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4810 }
4811 
4812 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4813 		struct nfs_pathconf *pathconf)
4814 {
4815 	struct nfs4_exception exception = { };
4816 	int err;
4817 
4818 	do {
4819 		err = nfs4_handle_exception(server,
4820 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4821 				&exception);
4822 	} while (exception.retry);
4823 	return err;
4824 }
4825 
4826 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4827 		const struct nfs_open_context *ctx,
4828 		const struct nfs_lock_context *l_ctx,
4829 		fmode_t fmode)
4830 {
4831 	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4832 }
4833 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4834 
4835 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4836 		const struct nfs_open_context *ctx,
4837 		const struct nfs_lock_context *l_ctx,
4838 		fmode_t fmode)
4839 {
4840 	nfs4_stateid current_stateid;
4841 
4842 	/* If the current stateid represents a lost lock, then exit */
4843 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4844 		return true;
4845 	return nfs4_stateid_match(stateid, &current_stateid);
4846 }
4847 
4848 static bool nfs4_error_stateid_expired(int err)
4849 {
4850 	switch (err) {
4851 	case -NFS4ERR_DELEG_REVOKED:
4852 	case -NFS4ERR_ADMIN_REVOKED:
4853 	case -NFS4ERR_BAD_STATEID:
4854 	case -NFS4ERR_STALE_STATEID:
4855 	case -NFS4ERR_OLD_STATEID:
4856 	case -NFS4ERR_OPENMODE:
4857 	case -NFS4ERR_EXPIRED:
4858 		return true;
4859 	}
4860 	return false;
4861 }
4862 
4863 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4864 {
4865 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4866 
4867 	trace_nfs4_read(hdr, task->tk_status);
4868 	if (task->tk_status < 0) {
4869 		struct nfs4_exception exception = {
4870 			.inode = hdr->inode,
4871 			.state = hdr->args.context->state,
4872 			.stateid = &hdr->args.stateid,
4873 		};
4874 		task->tk_status = nfs4_async_handle_exception(task,
4875 				server, task->tk_status, &exception);
4876 		if (exception.retry) {
4877 			rpc_restart_call_prepare(task);
4878 			return -EAGAIN;
4879 		}
4880 	}
4881 
4882 	if (task->tk_status > 0)
4883 		renew_lease(server, hdr->timestamp);
4884 	return 0;
4885 }
4886 
4887 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4888 		struct nfs_pgio_args *args)
4889 {
4890 
4891 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4892 		nfs4_stateid_is_current(&args->stateid,
4893 				args->context,
4894 				args->lock_context,
4895 				FMODE_READ))
4896 		return false;
4897 	rpc_restart_call_prepare(task);
4898 	return true;
4899 }
4900 
4901 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4902 {
4903 
4904 	dprintk("--> %s\n", __func__);
4905 
4906 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4907 		return -EAGAIN;
4908 	if (nfs4_read_stateid_changed(task, &hdr->args))
4909 		return -EAGAIN;
4910 	if (task->tk_status > 0)
4911 		nfs_invalidate_atime(hdr->inode);
4912 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4913 				    nfs4_read_done_cb(task, hdr);
4914 }
4915 
4916 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4917 				 struct rpc_message *msg)
4918 {
4919 	hdr->timestamp   = jiffies;
4920 	if (!hdr->pgio_done_cb)
4921 		hdr->pgio_done_cb = nfs4_read_done_cb;
4922 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4923 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
4924 }
4925 
4926 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4927 				      struct nfs_pgio_header *hdr)
4928 {
4929 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4930 			&hdr->args.seq_args,
4931 			&hdr->res.seq_res,
4932 			task))
4933 		return 0;
4934 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4935 				hdr->args.lock_context,
4936 				hdr->rw_mode) == -EIO)
4937 		return -EIO;
4938 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4939 		return -EIO;
4940 	return 0;
4941 }
4942 
4943 static int nfs4_write_done_cb(struct rpc_task *task,
4944 			      struct nfs_pgio_header *hdr)
4945 {
4946 	struct inode *inode = hdr->inode;
4947 
4948 	trace_nfs4_write(hdr, task->tk_status);
4949 	if (task->tk_status < 0) {
4950 		struct nfs4_exception exception = {
4951 			.inode = hdr->inode,
4952 			.state = hdr->args.context->state,
4953 			.stateid = &hdr->args.stateid,
4954 		};
4955 		task->tk_status = nfs4_async_handle_exception(task,
4956 				NFS_SERVER(inode), task->tk_status,
4957 				&exception);
4958 		if (exception.retry) {
4959 			rpc_restart_call_prepare(task);
4960 			return -EAGAIN;
4961 		}
4962 	}
4963 	if (task->tk_status >= 0) {
4964 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4965 		nfs_writeback_update_inode(hdr);
4966 	}
4967 	return 0;
4968 }
4969 
4970 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4971 		struct nfs_pgio_args *args)
4972 {
4973 
4974 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4975 		nfs4_stateid_is_current(&args->stateid,
4976 				args->context,
4977 				args->lock_context,
4978 				FMODE_WRITE))
4979 		return false;
4980 	rpc_restart_call_prepare(task);
4981 	return true;
4982 }
4983 
4984 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4985 {
4986 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4987 		return -EAGAIN;
4988 	if (nfs4_write_stateid_changed(task, &hdr->args))
4989 		return -EAGAIN;
4990 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4991 		nfs4_write_done_cb(task, hdr);
4992 }
4993 
4994 static
4995 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4996 {
4997 	/* Don't request attributes for pNFS or O_DIRECT writes */
4998 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4999 		return false;
5000 	/* Otherwise, request attributes if and only if we don't hold
5001 	 * a delegation
5002 	 */
5003 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5004 }
5005 
5006 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5007 				  struct rpc_message *msg,
5008 				  struct rpc_clnt **clnt)
5009 {
5010 	struct nfs_server *server = NFS_SERVER(hdr->inode);
5011 
5012 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5013 		hdr->args.bitmask = NULL;
5014 		hdr->res.fattr = NULL;
5015 	} else
5016 		hdr->args.bitmask = server->cache_consistency_bitmask;
5017 
5018 	if (!hdr->pgio_done_cb)
5019 		hdr->pgio_done_cb = nfs4_write_done_cb;
5020 	hdr->res.server = server;
5021 	hdr->timestamp   = jiffies;
5022 
5023 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5024 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5025 	nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5026 }
5027 
5028 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5029 {
5030 	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5031 			&data->args.seq_args,
5032 			&data->res.seq_res,
5033 			task);
5034 }
5035 
5036 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5037 {
5038 	struct inode *inode = data->inode;
5039 
5040 	trace_nfs4_commit(data, task->tk_status);
5041 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5042 				    NULL, NULL) == -EAGAIN) {
5043 		rpc_restart_call_prepare(task);
5044 		return -EAGAIN;
5045 	}
5046 	return 0;
5047 }
5048 
5049 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5050 {
5051 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5052 		return -EAGAIN;
5053 	return data->commit_done_cb(task, data);
5054 }
5055 
5056 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5057 				   struct rpc_clnt **clnt)
5058 {
5059 	struct nfs_server *server = NFS_SERVER(data->inode);
5060 
5061 	if (data->commit_done_cb == NULL)
5062 		data->commit_done_cb = nfs4_commit_done_cb;
5063 	data->res.server = server;
5064 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5065 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5066 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5067 }
5068 
5069 struct nfs4_renewdata {
5070 	struct nfs_client	*client;
5071 	unsigned long		timestamp;
5072 };
5073 
5074 /*
5075  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5076  * standalone procedure for queueing an asynchronous RENEW.
5077  */
5078 static void nfs4_renew_release(void *calldata)
5079 {
5080 	struct nfs4_renewdata *data = calldata;
5081 	struct nfs_client *clp = data->client;
5082 
5083 	if (refcount_read(&clp->cl_count) > 1)
5084 		nfs4_schedule_state_renewal(clp);
5085 	nfs_put_client(clp);
5086 	kfree(data);
5087 }
5088 
5089 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5090 {
5091 	struct nfs4_renewdata *data = calldata;
5092 	struct nfs_client *clp = data->client;
5093 	unsigned long timestamp = data->timestamp;
5094 
5095 	trace_nfs4_renew_async(clp, task->tk_status);
5096 	switch (task->tk_status) {
5097 	case 0:
5098 		break;
5099 	case -NFS4ERR_LEASE_MOVED:
5100 		nfs4_schedule_lease_moved_recovery(clp);
5101 		break;
5102 	default:
5103 		/* Unless we're shutting down, schedule state recovery! */
5104 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5105 			return;
5106 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5107 			nfs4_schedule_lease_recovery(clp);
5108 			return;
5109 		}
5110 		nfs4_schedule_path_down_recovery(clp);
5111 	}
5112 	do_renew_lease(clp, timestamp);
5113 }
5114 
5115 static const struct rpc_call_ops nfs4_renew_ops = {
5116 	.rpc_call_done = nfs4_renew_done,
5117 	.rpc_release = nfs4_renew_release,
5118 };
5119 
5120 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5121 {
5122 	struct rpc_message msg = {
5123 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5124 		.rpc_argp	= clp,
5125 		.rpc_cred	= cred,
5126 	};
5127 	struct nfs4_renewdata *data;
5128 
5129 	if (renew_flags == 0)
5130 		return 0;
5131 	if (!refcount_inc_not_zero(&clp->cl_count))
5132 		return -EIO;
5133 	data = kmalloc(sizeof(*data), GFP_NOFS);
5134 	if (data == NULL) {
5135 		nfs_put_client(clp);
5136 		return -ENOMEM;
5137 	}
5138 	data->client = clp;
5139 	data->timestamp = jiffies;
5140 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5141 			&nfs4_renew_ops, data);
5142 }
5143 
5144 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
5145 {
5146 	struct rpc_message msg = {
5147 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5148 		.rpc_argp	= clp,
5149 		.rpc_cred	= cred,
5150 	};
5151 	unsigned long now = jiffies;
5152 	int status;
5153 
5154 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5155 	if (status < 0)
5156 		return status;
5157 	do_renew_lease(clp, now);
5158 	return 0;
5159 }
5160 
5161 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5162 {
5163 	return server->caps & NFS_CAP_ACLS;
5164 }
5165 
5166 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5167  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5168  * the stack.
5169  */
5170 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5171 
5172 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5173 		struct page **pages)
5174 {
5175 	struct page *newpage, **spages;
5176 	int rc = 0;
5177 	size_t len;
5178 	spages = pages;
5179 
5180 	do {
5181 		len = min_t(size_t, PAGE_SIZE, buflen);
5182 		newpage = alloc_page(GFP_KERNEL);
5183 
5184 		if (newpage == NULL)
5185 			goto unwind;
5186 		memcpy(page_address(newpage), buf, len);
5187 		buf += len;
5188 		buflen -= len;
5189 		*pages++ = newpage;
5190 		rc++;
5191 	} while (buflen != 0);
5192 
5193 	return rc;
5194 
5195 unwind:
5196 	for(; rc > 0; rc--)
5197 		__free_page(spages[rc-1]);
5198 	return -ENOMEM;
5199 }
5200 
5201 struct nfs4_cached_acl {
5202 	int cached;
5203 	size_t len;
5204 	char data[0];
5205 };
5206 
5207 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5208 {
5209 	struct nfs_inode *nfsi = NFS_I(inode);
5210 
5211 	spin_lock(&inode->i_lock);
5212 	kfree(nfsi->nfs4_acl);
5213 	nfsi->nfs4_acl = acl;
5214 	spin_unlock(&inode->i_lock);
5215 }
5216 
5217 static void nfs4_zap_acl_attr(struct inode *inode)
5218 {
5219 	nfs4_set_cached_acl(inode, NULL);
5220 }
5221 
5222 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5223 {
5224 	struct nfs_inode *nfsi = NFS_I(inode);
5225 	struct nfs4_cached_acl *acl;
5226 	int ret = -ENOENT;
5227 
5228 	spin_lock(&inode->i_lock);
5229 	acl = nfsi->nfs4_acl;
5230 	if (acl == NULL)
5231 		goto out;
5232 	if (buf == NULL) /* user is just asking for length */
5233 		goto out_len;
5234 	if (acl->cached == 0)
5235 		goto out;
5236 	ret = -ERANGE; /* see getxattr(2) man page */
5237 	if (acl->len > buflen)
5238 		goto out;
5239 	memcpy(buf, acl->data, acl->len);
5240 out_len:
5241 	ret = acl->len;
5242 out:
5243 	spin_unlock(&inode->i_lock);
5244 	return ret;
5245 }
5246 
5247 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5248 {
5249 	struct nfs4_cached_acl *acl;
5250 	size_t buflen = sizeof(*acl) + acl_len;
5251 
5252 	if (buflen <= PAGE_SIZE) {
5253 		acl = kmalloc(buflen, GFP_KERNEL);
5254 		if (acl == NULL)
5255 			goto out;
5256 		acl->cached = 1;
5257 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5258 	} else {
5259 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5260 		if (acl == NULL)
5261 			goto out;
5262 		acl->cached = 0;
5263 	}
5264 	acl->len = acl_len;
5265 out:
5266 	nfs4_set_cached_acl(inode, acl);
5267 }
5268 
5269 /*
5270  * The getxattr API returns the required buffer length when called with a
5271  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5272  * the required buf.  On a NULL buf, we send a page of data to the server
5273  * guessing that the ACL request can be serviced by a page. If so, we cache
5274  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5275  * the cache. If not so, we throw away the page, and cache the required
5276  * length. The next getxattr call will then produce another round trip to
5277  * the server, this time with the input buf of the required size.
5278  */
5279 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5280 {
5281 	struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5282 	struct nfs_getaclargs args = {
5283 		.fh = NFS_FH(inode),
5284 		.acl_pages = pages,
5285 		.acl_len = buflen,
5286 	};
5287 	struct nfs_getaclres res = {
5288 		.acl_len = buflen,
5289 	};
5290 	struct rpc_message msg = {
5291 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5292 		.rpc_argp = &args,
5293 		.rpc_resp = &res,
5294 	};
5295 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5296 	int ret = -ENOMEM, i;
5297 
5298 	if (npages > ARRAY_SIZE(pages))
5299 		return -ERANGE;
5300 
5301 	for (i = 0; i < npages; i++) {
5302 		pages[i] = alloc_page(GFP_KERNEL);
5303 		if (!pages[i])
5304 			goto out_free;
5305 	}
5306 
5307 	/* for decoding across pages */
5308 	res.acl_scratch = alloc_page(GFP_KERNEL);
5309 	if (!res.acl_scratch)
5310 		goto out_free;
5311 
5312 	args.acl_len = npages * PAGE_SIZE;
5313 
5314 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5315 		__func__, buf, buflen, npages, args.acl_len);
5316 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5317 			     &msg, &args.seq_args, &res.seq_res, 0);
5318 	if (ret)
5319 		goto out_free;
5320 
5321 	/* Handle the case where the passed-in buffer is too short */
5322 	if (res.acl_flags & NFS4_ACL_TRUNC) {
5323 		/* Did the user only issue a request for the acl length? */
5324 		if (buf == NULL)
5325 			goto out_ok;
5326 		ret = -ERANGE;
5327 		goto out_free;
5328 	}
5329 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5330 	if (buf) {
5331 		if (res.acl_len > buflen) {
5332 			ret = -ERANGE;
5333 			goto out_free;
5334 		}
5335 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5336 	}
5337 out_ok:
5338 	ret = res.acl_len;
5339 out_free:
5340 	for (i = 0; i < npages; i++)
5341 		if (pages[i])
5342 			__free_page(pages[i]);
5343 	if (res.acl_scratch)
5344 		__free_page(res.acl_scratch);
5345 	return ret;
5346 }
5347 
5348 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5349 {
5350 	struct nfs4_exception exception = { };
5351 	ssize_t ret;
5352 	do {
5353 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5354 		trace_nfs4_get_acl(inode, ret);
5355 		if (ret >= 0)
5356 			break;
5357 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5358 	} while (exception.retry);
5359 	return ret;
5360 }
5361 
5362 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5363 {
5364 	struct nfs_server *server = NFS_SERVER(inode);
5365 	int ret;
5366 
5367 	if (!nfs4_server_supports_acls(server))
5368 		return -EOPNOTSUPP;
5369 	ret = nfs_revalidate_inode(server, inode);
5370 	if (ret < 0)
5371 		return ret;
5372 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5373 		nfs_zap_acl_cache(inode);
5374 	ret = nfs4_read_cached_acl(inode, buf, buflen);
5375 	if (ret != -ENOENT)
5376 		/* -ENOENT is returned if there is no ACL or if there is an ACL
5377 		 * but no cached acl data, just the acl length */
5378 		return ret;
5379 	return nfs4_get_acl_uncached(inode, buf, buflen);
5380 }
5381 
5382 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5383 {
5384 	struct nfs_server *server = NFS_SERVER(inode);
5385 	struct page *pages[NFS4ACL_MAXPAGES];
5386 	struct nfs_setaclargs arg = {
5387 		.fh		= NFS_FH(inode),
5388 		.acl_pages	= pages,
5389 		.acl_len	= buflen,
5390 	};
5391 	struct nfs_setaclres res;
5392 	struct rpc_message msg = {
5393 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5394 		.rpc_argp	= &arg,
5395 		.rpc_resp	= &res,
5396 	};
5397 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5398 	int ret, i;
5399 
5400 	if (!nfs4_server_supports_acls(server))
5401 		return -EOPNOTSUPP;
5402 	if (npages > ARRAY_SIZE(pages))
5403 		return -ERANGE;
5404 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5405 	if (i < 0)
5406 		return i;
5407 	nfs4_inode_make_writeable(inode);
5408 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5409 
5410 	/*
5411 	 * Free each page after tx, so the only ref left is
5412 	 * held by the network stack
5413 	 */
5414 	for (; i > 0; i--)
5415 		put_page(pages[i-1]);
5416 
5417 	/*
5418 	 * Acl update can result in inode attribute update.
5419 	 * so mark the attribute cache invalid.
5420 	 */
5421 	spin_lock(&inode->i_lock);
5422 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5423 		| NFS_INO_INVALID_CTIME
5424 		| NFS_INO_REVAL_FORCED;
5425 	spin_unlock(&inode->i_lock);
5426 	nfs_access_zap_cache(inode);
5427 	nfs_zap_acl_cache(inode);
5428 	return ret;
5429 }
5430 
5431 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5432 {
5433 	struct nfs4_exception exception = { };
5434 	int err;
5435 	do {
5436 		err = __nfs4_proc_set_acl(inode, buf, buflen);
5437 		trace_nfs4_set_acl(inode, err);
5438 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5439 				&exception);
5440 	} while (exception.retry);
5441 	return err;
5442 }
5443 
5444 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5445 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5446 					size_t buflen)
5447 {
5448 	struct nfs_server *server = NFS_SERVER(inode);
5449 	struct nfs_fattr fattr;
5450 	struct nfs4_label label = {0, 0, buflen, buf};
5451 
5452 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5453 	struct nfs4_getattr_arg arg = {
5454 		.fh		= NFS_FH(inode),
5455 		.bitmask	= bitmask,
5456 	};
5457 	struct nfs4_getattr_res res = {
5458 		.fattr		= &fattr,
5459 		.label		= &label,
5460 		.server		= server,
5461 	};
5462 	struct rpc_message msg = {
5463 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5464 		.rpc_argp	= &arg,
5465 		.rpc_resp	= &res,
5466 	};
5467 	int ret;
5468 
5469 	nfs_fattr_init(&fattr);
5470 
5471 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5472 	if (ret)
5473 		return ret;
5474 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5475 		return -ENOENT;
5476 	if (buflen < label.len)
5477 		return -ERANGE;
5478 	return 0;
5479 }
5480 
5481 static int nfs4_get_security_label(struct inode *inode, void *buf,
5482 					size_t buflen)
5483 {
5484 	struct nfs4_exception exception = { };
5485 	int err;
5486 
5487 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5488 		return -EOPNOTSUPP;
5489 
5490 	do {
5491 		err = _nfs4_get_security_label(inode, buf, buflen);
5492 		trace_nfs4_get_security_label(inode, err);
5493 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5494 				&exception);
5495 	} while (exception.retry);
5496 	return err;
5497 }
5498 
5499 static int _nfs4_do_set_security_label(struct inode *inode,
5500 		struct nfs4_label *ilabel,
5501 		struct nfs_fattr *fattr,
5502 		struct nfs4_label *olabel)
5503 {
5504 
5505 	struct iattr sattr = {0};
5506 	struct nfs_server *server = NFS_SERVER(inode);
5507 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5508 	struct nfs_setattrargs arg = {
5509 		.fh		= NFS_FH(inode),
5510 		.iap		= &sattr,
5511 		.server		= server,
5512 		.bitmask	= bitmask,
5513 		.label		= ilabel,
5514 	};
5515 	struct nfs_setattrres res = {
5516 		.fattr		= fattr,
5517 		.label		= olabel,
5518 		.server		= server,
5519 	};
5520 	struct rpc_message msg = {
5521 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5522 		.rpc_argp	= &arg,
5523 		.rpc_resp	= &res,
5524 	};
5525 	int status;
5526 
5527 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5528 
5529 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5530 	if (status)
5531 		dprintk("%s failed: %d\n", __func__, status);
5532 
5533 	return status;
5534 }
5535 
5536 static int nfs4_do_set_security_label(struct inode *inode,
5537 		struct nfs4_label *ilabel,
5538 		struct nfs_fattr *fattr,
5539 		struct nfs4_label *olabel)
5540 {
5541 	struct nfs4_exception exception = { };
5542 	int err;
5543 
5544 	do {
5545 		err = _nfs4_do_set_security_label(inode, ilabel,
5546 				fattr, olabel);
5547 		trace_nfs4_set_security_label(inode, err);
5548 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
5549 				&exception);
5550 	} while (exception.retry);
5551 	return err;
5552 }
5553 
5554 static int
5555 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5556 {
5557 	struct nfs4_label ilabel, *olabel = NULL;
5558 	struct nfs_fattr fattr;
5559 	struct rpc_cred *cred;
5560 	int status;
5561 
5562 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5563 		return -EOPNOTSUPP;
5564 
5565 	nfs_fattr_init(&fattr);
5566 
5567 	ilabel.pi = 0;
5568 	ilabel.lfs = 0;
5569 	ilabel.label = (char *)buf;
5570 	ilabel.len = buflen;
5571 
5572 	cred = rpc_lookup_cred();
5573 	if (IS_ERR(cred))
5574 		return PTR_ERR(cred);
5575 
5576 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5577 	if (IS_ERR(olabel)) {
5578 		status = -PTR_ERR(olabel);
5579 		goto out;
5580 	}
5581 
5582 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5583 	if (status == 0)
5584 		nfs_setsecurity(inode, &fattr, olabel);
5585 
5586 	nfs4_label_free(olabel);
5587 out:
5588 	put_rpccred(cred);
5589 	return status;
5590 }
5591 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
5592 
5593 
5594 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5595 				    nfs4_verifier *bootverf)
5596 {
5597 	__be32 verf[2];
5598 
5599 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5600 		/* An impossible timestamp guarantees this value
5601 		 * will never match a generated boot time. */
5602 		verf[0] = cpu_to_be32(U32_MAX);
5603 		verf[1] = cpu_to_be32(U32_MAX);
5604 	} else {
5605 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5606 		u64 ns = ktime_to_ns(nn->boot_time);
5607 
5608 		verf[0] = cpu_to_be32(ns >> 32);
5609 		verf[1] = cpu_to_be32(ns);
5610 	}
5611 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
5612 }
5613 
5614 static int
5615 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5616 {
5617 	size_t len;
5618 	char *str;
5619 
5620 	if (clp->cl_owner_id != NULL)
5621 		return 0;
5622 
5623 	rcu_read_lock();
5624 	len = 14 +
5625 		strlen(clp->cl_rpcclient->cl_nodename) +
5626 		1 +
5627 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5628 		1;
5629 	rcu_read_unlock();
5630 	if (nfs4_client_id_uniquifier[0] != '\0')
5631 		len += strlen(nfs4_client_id_uniquifier) + 1;
5632 	if (len > NFS4_OPAQUE_LIMIT + 1)
5633 		return -EINVAL;
5634 
5635 	/*
5636 	 * Since this string is allocated at mount time, and held until the
5637 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5638 	 * about a memory-reclaim deadlock.
5639 	 */
5640 	str = kmalloc(len, GFP_KERNEL);
5641 	if (!str)
5642 		return -ENOMEM;
5643 
5644 	rcu_read_lock();
5645 	if (nfs4_client_id_uniquifier[0] != '\0')
5646 		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5647 			  clp->cl_rpcclient->cl_nodename,
5648 			  nfs4_client_id_uniquifier,
5649 			  rpc_peeraddr2str(clp->cl_rpcclient,
5650 					   RPC_DISPLAY_ADDR));
5651 	else
5652 		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5653 			  clp->cl_rpcclient->cl_nodename,
5654 			  rpc_peeraddr2str(clp->cl_rpcclient,
5655 					   RPC_DISPLAY_ADDR));
5656 	rcu_read_unlock();
5657 
5658 	clp->cl_owner_id = str;
5659 	return 0;
5660 }
5661 
5662 static int
5663 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5664 {
5665 	size_t len;
5666 	char *str;
5667 
5668 	len = 10 + 10 + 1 + 10 + 1 +
5669 		strlen(nfs4_client_id_uniquifier) + 1 +
5670 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5671 
5672 	if (len > NFS4_OPAQUE_LIMIT + 1)
5673 		return -EINVAL;
5674 
5675 	/*
5676 	 * Since this string is allocated at mount time, and held until the
5677 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5678 	 * about a memory-reclaim deadlock.
5679 	 */
5680 	str = kmalloc(len, GFP_KERNEL);
5681 	if (!str)
5682 		return -ENOMEM;
5683 
5684 	scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5685 			clp->rpc_ops->version, clp->cl_minorversion,
5686 			nfs4_client_id_uniquifier,
5687 			clp->cl_rpcclient->cl_nodename);
5688 	clp->cl_owner_id = str;
5689 	return 0;
5690 }
5691 
5692 static int
5693 nfs4_init_uniform_client_string(struct nfs_client *clp)
5694 {
5695 	size_t len;
5696 	char *str;
5697 
5698 	if (clp->cl_owner_id != NULL)
5699 		return 0;
5700 
5701 	if (nfs4_client_id_uniquifier[0] != '\0')
5702 		return nfs4_init_uniquifier_client_string(clp);
5703 
5704 	len = 10 + 10 + 1 + 10 + 1 +
5705 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5706 
5707 	if (len > NFS4_OPAQUE_LIMIT + 1)
5708 		return -EINVAL;
5709 
5710 	/*
5711 	 * Since this string is allocated at mount time, and held until the
5712 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5713 	 * about a memory-reclaim deadlock.
5714 	 */
5715 	str = kmalloc(len, GFP_KERNEL);
5716 	if (!str)
5717 		return -ENOMEM;
5718 
5719 	scnprintf(str, len, "Linux NFSv%u.%u %s",
5720 			clp->rpc_ops->version, clp->cl_minorversion,
5721 			clp->cl_rpcclient->cl_nodename);
5722 	clp->cl_owner_id = str;
5723 	return 0;
5724 }
5725 
5726 /*
5727  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5728  * services.  Advertise one based on the address family of the
5729  * clientaddr.
5730  */
5731 static unsigned int
5732 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5733 {
5734 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5735 		return scnprintf(buf, len, "tcp6");
5736 	else
5737 		return scnprintf(buf, len, "tcp");
5738 }
5739 
5740 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5741 {
5742 	struct nfs4_setclientid *sc = calldata;
5743 
5744 	if (task->tk_status == 0)
5745 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5746 }
5747 
5748 static const struct rpc_call_ops nfs4_setclientid_ops = {
5749 	.rpc_call_done = nfs4_setclientid_done,
5750 };
5751 
5752 /**
5753  * nfs4_proc_setclientid - Negotiate client ID
5754  * @clp: state data structure
5755  * @program: RPC program for NFSv4 callback service
5756  * @port: IP port number for NFS4 callback service
5757  * @cred: RPC credential to use for this call
5758  * @res: where to place the result
5759  *
5760  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5761  */
5762 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5763 		unsigned short port, struct rpc_cred *cred,
5764 		struct nfs4_setclientid_res *res)
5765 {
5766 	nfs4_verifier sc_verifier;
5767 	struct nfs4_setclientid setclientid = {
5768 		.sc_verifier = &sc_verifier,
5769 		.sc_prog = program,
5770 		.sc_clnt = clp,
5771 	};
5772 	struct rpc_message msg = {
5773 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5774 		.rpc_argp = &setclientid,
5775 		.rpc_resp = res,
5776 		.rpc_cred = cred,
5777 	};
5778 	struct rpc_task *task;
5779 	struct rpc_task_setup task_setup_data = {
5780 		.rpc_client = clp->cl_rpcclient,
5781 		.rpc_message = &msg,
5782 		.callback_ops = &nfs4_setclientid_ops,
5783 		.callback_data = &setclientid,
5784 		.flags = RPC_TASK_TIMEOUT,
5785 	};
5786 	int status;
5787 
5788 	/* nfs_client_id4 */
5789 	nfs4_init_boot_verifier(clp, &sc_verifier);
5790 
5791 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5792 		status = nfs4_init_uniform_client_string(clp);
5793 	else
5794 		status = nfs4_init_nonuniform_client_string(clp);
5795 
5796 	if (status)
5797 		goto out;
5798 
5799 	/* cb_client4 */
5800 	setclientid.sc_netid_len =
5801 				nfs4_init_callback_netid(clp,
5802 						setclientid.sc_netid,
5803 						sizeof(setclientid.sc_netid));
5804 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5805 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5806 				clp->cl_ipaddr, port >> 8, port & 255);
5807 
5808 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
5809 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5810 		clp->cl_owner_id);
5811 	task = rpc_run_task(&task_setup_data);
5812 	if (IS_ERR(task)) {
5813 		status = PTR_ERR(task);
5814 		goto out;
5815 	}
5816 	status = task->tk_status;
5817 	if (setclientid.sc_cred) {
5818 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5819 		put_rpccred(setclientid.sc_cred);
5820 	}
5821 	rpc_put_task(task);
5822 out:
5823 	trace_nfs4_setclientid(clp, status);
5824 	dprintk("NFS reply setclientid: %d\n", status);
5825 	return status;
5826 }
5827 
5828 /**
5829  * nfs4_proc_setclientid_confirm - Confirm client ID
5830  * @clp: state data structure
5831  * @res: result of a previous SETCLIENTID
5832  * @cred: RPC credential to use for this call
5833  *
5834  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5835  */
5836 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5837 		struct nfs4_setclientid_res *arg,
5838 		struct rpc_cred *cred)
5839 {
5840 	struct rpc_message msg = {
5841 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5842 		.rpc_argp = arg,
5843 		.rpc_cred = cred,
5844 	};
5845 	int status;
5846 
5847 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5848 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5849 		clp->cl_clientid);
5850 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5851 	trace_nfs4_setclientid_confirm(clp, status);
5852 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5853 	return status;
5854 }
5855 
5856 struct nfs4_delegreturndata {
5857 	struct nfs4_delegreturnargs args;
5858 	struct nfs4_delegreturnres res;
5859 	struct nfs_fh fh;
5860 	nfs4_stateid stateid;
5861 	unsigned long timestamp;
5862 	struct {
5863 		struct nfs4_layoutreturn_args arg;
5864 		struct nfs4_layoutreturn_res res;
5865 		struct nfs4_xdr_opaque_data ld_private;
5866 		u32 roc_barrier;
5867 		bool roc;
5868 	} lr;
5869 	struct nfs_fattr fattr;
5870 	int rpc_status;
5871 	struct inode *inode;
5872 };
5873 
5874 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5875 {
5876 	struct nfs4_delegreturndata *data = calldata;
5877 	struct nfs4_exception exception = {
5878 		.inode = data->inode,
5879 		.stateid = &data->stateid,
5880 	};
5881 
5882 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5883 		return;
5884 
5885 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5886 
5887 	/* Handle Layoutreturn errors */
5888 	if (data->args.lr_args && task->tk_status != 0) {
5889 		switch(data->res.lr_ret) {
5890 		default:
5891 			data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5892 			break;
5893 		case 0:
5894 			data->args.lr_args = NULL;
5895 			data->res.lr_res = NULL;
5896 			break;
5897 		case -NFS4ERR_OLD_STATEID:
5898 			if (nfs4_refresh_layout_stateid(&data->args.lr_args->stateid,
5899 						data->inode))
5900 				goto lr_restart;
5901 			/* Fallthrough */
5902 		case -NFS4ERR_ADMIN_REVOKED:
5903 		case -NFS4ERR_DELEG_REVOKED:
5904 		case -NFS4ERR_EXPIRED:
5905 		case -NFS4ERR_BAD_STATEID:
5906 		case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5907 		case -NFS4ERR_WRONG_CRED:
5908 			data->args.lr_args = NULL;
5909 			data->res.lr_res = NULL;
5910 			goto lr_restart;
5911 		}
5912 	}
5913 
5914 	switch (task->tk_status) {
5915 	case 0:
5916 		renew_lease(data->res.server, data->timestamp);
5917 		break;
5918 	case -NFS4ERR_ADMIN_REVOKED:
5919 	case -NFS4ERR_DELEG_REVOKED:
5920 	case -NFS4ERR_EXPIRED:
5921 		nfs4_free_revoked_stateid(data->res.server,
5922 				data->args.stateid,
5923 				task->tk_msg.rpc_cred);
5924 		/* Fallthrough */
5925 	case -NFS4ERR_BAD_STATEID:
5926 	case -NFS4ERR_STALE_STATEID:
5927 		task->tk_status = 0;
5928 		break;
5929 	case -NFS4ERR_OLD_STATEID:
5930 		if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
5931 			goto out_restart;
5932 		task->tk_status = 0;
5933 		break;
5934 	case -NFS4ERR_ACCESS:
5935 		if (data->args.bitmask) {
5936 			data->args.bitmask = NULL;
5937 			data->res.fattr = NULL;
5938 			goto out_restart;
5939 		}
5940 		/* Fallthrough */
5941 	default:
5942 		task->tk_status = nfs4_async_handle_exception(task,
5943 				data->res.server, task->tk_status,
5944 				&exception);
5945 		if (exception.retry)
5946 			goto out_restart;
5947 	}
5948 	data->rpc_status = task->tk_status;
5949 	return;
5950 lr_restart:
5951 	data->res.lr_ret = 0;
5952 out_restart:
5953 	task->tk_status = 0;
5954 	rpc_restart_call_prepare(task);
5955 }
5956 
5957 static void nfs4_delegreturn_release(void *calldata)
5958 {
5959 	struct nfs4_delegreturndata *data = calldata;
5960 	struct inode *inode = data->inode;
5961 
5962 	if (inode) {
5963 		if (data->lr.roc)
5964 			pnfs_roc_release(&data->lr.arg, &data->lr.res,
5965 					data->res.lr_ret);
5966 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5967 		nfs_iput_and_deactive(inode);
5968 	}
5969 	kfree(calldata);
5970 }
5971 
5972 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5973 {
5974 	struct nfs4_delegreturndata *d_data;
5975 
5976 	d_data = (struct nfs4_delegreturndata *)data;
5977 
5978 	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5979 		return;
5980 
5981 	nfs4_setup_sequence(d_data->res.server->nfs_client,
5982 			&d_data->args.seq_args,
5983 			&d_data->res.seq_res,
5984 			task);
5985 }
5986 
5987 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5988 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5989 	.rpc_call_done = nfs4_delegreturn_done,
5990 	.rpc_release = nfs4_delegreturn_release,
5991 };
5992 
5993 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5994 {
5995 	struct nfs4_delegreturndata *data;
5996 	struct nfs_server *server = NFS_SERVER(inode);
5997 	struct rpc_task *task;
5998 	struct rpc_message msg = {
5999 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6000 		.rpc_cred = cred,
6001 	};
6002 	struct rpc_task_setup task_setup_data = {
6003 		.rpc_client = server->client,
6004 		.rpc_message = &msg,
6005 		.callback_ops = &nfs4_delegreturn_ops,
6006 		.flags = RPC_TASK_ASYNC,
6007 	};
6008 	int status = 0;
6009 
6010 	data = kzalloc(sizeof(*data), GFP_NOFS);
6011 	if (data == NULL)
6012 		return -ENOMEM;
6013 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6014 
6015 	nfs4_state_protect(server->nfs_client,
6016 			NFS_SP4_MACH_CRED_CLEANUP,
6017 			&task_setup_data.rpc_client, &msg);
6018 
6019 	data->args.fhandle = &data->fh;
6020 	data->args.stateid = &data->stateid;
6021 	data->args.bitmask = server->cache_consistency_bitmask;
6022 	nfs_copy_fh(&data->fh, NFS_FH(inode));
6023 	nfs4_stateid_copy(&data->stateid, stateid);
6024 	data->res.fattr = &data->fattr;
6025 	data->res.server = server;
6026 	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6027 	data->lr.arg.ld_private = &data->lr.ld_private;
6028 	nfs_fattr_init(data->res.fattr);
6029 	data->timestamp = jiffies;
6030 	data->rpc_status = 0;
6031 	data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6032 	data->inode = nfs_igrab_and_active(inode);
6033 	if (data->inode) {
6034 		if (data->lr.roc) {
6035 			data->args.lr_args = &data->lr.arg;
6036 			data->res.lr_res = &data->lr.res;
6037 		}
6038 	} else if (data->lr.roc) {
6039 		pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6040 		data->lr.roc = false;
6041 	}
6042 
6043 	task_setup_data.callback_data = data;
6044 	msg.rpc_argp = &data->args;
6045 	msg.rpc_resp = &data->res;
6046 	task = rpc_run_task(&task_setup_data);
6047 	if (IS_ERR(task))
6048 		return PTR_ERR(task);
6049 	if (!issync)
6050 		goto out;
6051 	status = rpc_wait_for_completion_task(task);
6052 	if (status != 0)
6053 		goto out;
6054 	status = data->rpc_status;
6055 out:
6056 	rpc_put_task(task);
6057 	return status;
6058 }
6059 
6060 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
6061 {
6062 	struct nfs_server *server = NFS_SERVER(inode);
6063 	struct nfs4_exception exception = { };
6064 	int err;
6065 	do {
6066 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6067 		trace_nfs4_delegreturn(inode, stateid, err);
6068 		switch (err) {
6069 			case -NFS4ERR_STALE_STATEID:
6070 			case -NFS4ERR_EXPIRED:
6071 			case 0:
6072 				return 0;
6073 		}
6074 		err = nfs4_handle_exception(server, err, &exception);
6075 	} while (exception.retry);
6076 	return err;
6077 }
6078 
6079 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6080 {
6081 	struct inode *inode = state->inode;
6082 	struct nfs_server *server = NFS_SERVER(inode);
6083 	struct nfs_client *clp = server->nfs_client;
6084 	struct nfs_lockt_args arg = {
6085 		.fh = NFS_FH(inode),
6086 		.fl = request,
6087 	};
6088 	struct nfs_lockt_res res = {
6089 		.denied = request,
6090 	};
6091 	struct rpc_message msg = {
6092 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6093 		.rpc_argp	= &arg,
6094 		.rpc_resp	= &res,
6095 		.rpc_cred	= state->owner->so_cred,
6096 	};
6097 	struct nfs4_lock_state *lsp;
6098 	int status;
6099 
6100 	arg.lock_owner.clientid = clp->cl_clientid;
6101 	status = nfs4_set_lock_state(state, request);
6102 	if (status != 0)
6103 		goto out;
6104 	lsp = request->fl_u.nfs4_fl.owner;
6105 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6106 	arg.lock_owner.s_dev = server->s_dev;
6107 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6108 	switch (status) {
6109 		case 0:
6110 			request->fl_type = F_UNLCK;
6111 			break;
6112 		case -NFS4ERR_DENIED:
6113 			status = 0;
6114 	}
6115 	request->fl_ops->fl_release_private(request);
6116 	request->fl_ops = NULL;
6117 out:
6118 	return status;
6119 }
6120 
6121 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6122 {
6123 	struct nfs4_exception exception = { };
6124 	int err;
6125 
6126 	do {
6127 		err = _nfs4_proc_getlk(state, cmd, request);
6128 		trace_nfs4_get_lock(request, state, cmd, err);
6129 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6130 				&exception);
6131 	} while (exception.retry);
6132 	return err;
6133 }
6134 
6135 struct nfs4_unlockdata {
6136 	struct nfs_locku_args arg;
6137 	struct nfs_locku_res res;
6138 	struct nfs4_lock_state *lsp;
6139 	struct nfs_open_context *ctx;
6140 	struct nfs_lock_context *l_ctx;
6141 	struct file_lock fl;
6142 	struct nfs_server *server;
6143 	unsigned long timestamp;
6144 };
6145 
6146 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6147 		struct nfs_open_context *ctx,
6148 		struct nfs4_lock_state *lsp,
6149 		struct nfs_seqid *seqid)
6150 {
6151 	struct nfs4_unlockdata *p;
6152 	struct inode *inode = lsp->ls_state->inode;
6153 
6154 	p = kzalloc(sizeof(*p), GFP_NOFS);
6155 	if (p == NULL)
6156 		return NULL;
6157 	p->arg.fh = NFS_FH(inode);
6158 	p->arg.fl = &p->fl;
6159 	p->arg.seqid = seqid;
6160 	p->res.seqid = seqid;
6161 	p->lsp = lsp;
6162 	refcount_inc(&lsp->ls_count);
6163 	/* Ensure we don't close file until we're done freeing locks! */
6164 	p->ctx = get_nfs_open_context(ctx);
6165 	p->l_ctx = nfs_get_lock_context(ctx);
6166 	memcpy(&p->fl, fl, sizeof(p->fl));
6167 	p->server = NFS_SERVER(inode);
6168 	return p;
6169 }
6170 
6171 static void nfs4_locku_release_calldata(void *data)
6172 {
6173 	struct nfs4_unlockdata *calldata = data;
6174 	nfs_free_seqid(calldata->arg.seqid);
6175 	nfs4_put_lock_state(calldata->lsp);
6176 	nfs_put_lock_context(calldata->l_ctx);
6177 	put_nfs_open_context(calldata->ctx);
6178 	kfree(calldata);
6179 }
6180 
6181 static void nfs4_locku_done(struct rpc_task *task, void *data)
6182 {
6183 	struct nfs4_unlockdata *calldata = data;
6184 	struct nfs4_exception exception = {
6185 		.inode = calldata->lsp->ls_state->inode,
6186 		.stateid = &calldata->arg.stateid,
6187 	};
6188 
6189 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6190 		return;
6191 	switch (task->tk_status) {
6192 		case 0:
6193 			renew_lease(calldata->server, calldata->timestamp);
6194 			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6195 			if (nfs4_update_lock_stateid(calldata->lsp,
6196 					&calldata->res.stateid))
6197 				break;
6198 		case -NFS4ERR_ADMIN_REVOKED:
6199 		case -NFS4ERR_EXPIRED:
6200 			nfs4_free_revoked_stateid(calldata->server,
6201 					&calldata->arg.stateid,
6202 					task->tk_msg.rpc_cred);
6203 		case -NFS4ERR_BAD_STATEID:
6204 		case -NFS4ERR_OLD_STATEID:
6205 		case -NFS4ERR_STALE_STATEID:
6206 			if (!nfs4_stateid_match(&calldata->arg.stateid,
6207 						&calldata->lsp->ls_stateid))
6208 				rpc_restart_call_prepare(task);
6209 			break;
6210 		default:
6211 			task->tk_status = nfs4_async_handle_exception(task,
6212 					calldata->server, task->tk_status,
6213 					&exception);
6214 			if (exception.retry)
6215 				rpc_restart_call_prepare(task);
6216 	}
6217 	nfs_release_seqid(calldata->arg.seqid);
6218 }
6219 
6220 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6221 {
6222 	struct nfs4_unlockdata *calldata = data;
6223 
6224 	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6225 		nfs_async_iocounter_wait(task, calldata->l_ctx))
6226 		return;
6227 
6228 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6229 		goto out_wait;
6230 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6231 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6232 		/* Note: exit _without_ running nfs4_locku_done */
6233 		goto out_no_action;
6234 	}
6235 	calldata->timestamp = jiffies;
6236 	if (nfs4_setup_sequence(calldata->server->nfs_client,
6237 				&calldata->arg.seq_args,
6238 				&calldata->res.seq_res,
6239 				task) != 0)
6240 		nfs_release_seqid(calldata->arg.seqid);
6241 	return;
6242 out_no_action:
6243 	task->tk_action = NULL;
6244 out_wait:
6245 	nfs4_sequence_done(task, &calldata->res.seq_res);
6246 }
6247 
6248 static const struct rpc_call_ops nfs4_locku_ops = {
6249 	.rpc_call_prepare = nfs4_locku_prepare,
6250 	.rpc_call_done = nfs4_locku_done,
6251 	.rpc_release = nfs4_locku_release_calldata,
6252 };
6253 
6254 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6255 		struct nfs_open_context *ctx,
6256 		struct nfs4_lock_state *lsp,
6257 		struct nfs_seqid *seqid)
6258 {
6259 	struct nfs4_unlockdata *data;
6260 	struct rpc_message msg = {
6261 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6262 		.rpc_cred = ctx->cred,
6263 	};
6264 	struct rpc_task_setup task_setup_data = {
6265 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6266 		.rpc_message = &msg,
6267 		.callback_ops = &nfs4_locku_ops,
6268 		.workqueue = nfsiod_workqueue,
6269 		.flags = RPC_TASK_ASYNC,
6270 	};
6271 
6272 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6273 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6274 
6275 	/* Ensure this is an unlock - when canceling a lock, the
6276 	 * canceled lock is passed in, and it won't be an unlock.
6277 	 */
6278 	fl->fl_type = F_UNLCK;
6279 	if (fl->fl_flags & FL_CLOSE)
6280 		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6281 
6282 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6283 	if (data == NULL) {
6284 		nfs_free_seqid(seqid);
6285 		return ERR_PTR(-ENOMEM);
6286 	}
6287 
6288 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6289 	msg.rpc_argp = &data->arg;
6290 	msg.rpc_resp = &data->res;
6291 	task_setup_data.callback_data = data;
6292 	return rpc_run_task(&task_setup_data);
6293 }
6294 
6295 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6296 {
6297 	struct inode *inode = state->inode;
6298 	struct nfs4_state_owner *sp = state->owner;
6299 	struct nfs_inode *nfsi = NFS_I(inode);
6300 	struct nfs_seqid *seqid;
6301 	struct nfs4_lock_state *lsp;
6302 	struct rpc_task *task;
6303 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6304 	int status = 0;
6305 	unsigned char fl_flags = request->fl_flags;
6306 
6307 	status = nfs4_set_lock_state(state, request);
6308 	/* Unlock _before_ we do the RPC call */
6309 	request->fl_flags |= FL_EXISTS;
6310 	/* Exclude nfs_delegation_claim_locks() */
6311 	mutex_lock(&sp->so_delegreturn_mutex);
6312 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6313 	down_read(&nfsi->rwsem);
6314 	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6315 		up_read(&nfsi->rwsem);
6316 		mutex_unlock(&sp->so_delegreturn_mutex);
6317 		goto out;
6318 	}
6319 	up_read(&nfsi->rwsem);
6320 	mutex_unlock(&sp->so_delegreturn_mutex);
6321 	if (status != 0)
6322 		goto out;
6323 	/* Is this a delegated lock? */
6324 	lsp = request->fl_u.nfs4_fl.owner;
6325 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6326 		goto out;
6327 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6328 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6329 	status = -ENOMEM;
6330 	if (IS_ERR(seqid))
6331 		goto out;
6332 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6333 	status = PTR_ERR(task);
6334 	if (IS_ERR(task))
6335 		goto out;
6336 	status = rpc_wait_for_completion_task(task);
6337 	rpc_put_task(task);
6338 out:
6339 	request->fl_flags = fl_flags;
6340 	trace_nfs4_unlock(request, state, F_SETLK, status);
6341 	return status;
6342 }
6343 
6344 struct nfs4_lockdata {
6345 	struct nfs_lock_args arg;
6346 	struct nfs_lock_res res;
6347 	struct nfs4_lock_state *lsp;
6348 	struct nfs_open_context *ctx;
6349 	struct file_lock fl;
6350 	unsigned long timestamp;
6351 	int rpc_status;
6352 	int cancelled;
6353 	struct nfs_server *server;
6354 };
6355 
6356 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6357 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6358 		gfp_t gfp_mask)
6359 {
6360 	struct nfs4_lockdata *p;
6361 	struct inode *inode = lsp->ls_state->inode;
6362 	struct nfs_server *server = NFS_SERVER(inode);
6363 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6364 
6365 	p = kzalloc(sizeof(*p), gfp_mask);
6366 	if (p == NULL)
6367 		return NULL;
6368 
6369 	p->arg.fh = NFS_FH(inode);
6370 	p->arg.fl = &p->fl;
6371 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6372 	if (IS_ERR(p->arg.open_seqid))
6373 		goto out_free;
6374 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6375 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6376 	if (IS_ERR(p->arg.lock_seqid))
6377 		goto out_free_seqid;
6378 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6379 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6380 	p->arg.lock_owner.s_dev = server->s_dev;
6381 	p->res.lock_seqid = p->arg.lock_seqid;
6382 	p->lsp = lsp;
6383 	p->server = server;
6384 	refcount_inc(&lsp->ls_count);
6385 	p->ctx = get_nfs_open_context(ctx);
6386 	memcpy(&p->fl, fl, sizeof(p->fl));
6387 	return p;
6388 out_free_seqid:
6389 	nfs_free_seqid(p->arg.open_seqid);
6390 out_free:
6391 	kfree(p);
6392 	return NULL;
6393 }
6394 
6395 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6396 {
6397 	struct nfs4_lockdata *data = calldata;
6398 	struct nfs4_state *state = data->lsp->ls_state;
6399 
6400 	dprintk("%s: begin!\n", __func__);
6401 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6402 		goto out_wait;
6403 	/* Do we need to do an open_to_lock_owner? */
6404 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6405 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6406 			goto out_release_lock_seqid;
6407 		}
6408 		nfs4_stateid_copy(&data->arg.open_stateid,
6409 				&state->open_stateid);
6410 		data->arg.new_lock_owner = 1;
6411 		data->res.open_seqid = data->arg.open_seqid;
6412 	} else {
6413 		data->arg.new_lock_owner = 0;
6414 		nfs4_stateid_copy(&data->arg.lock_stateid,
6415 				&data->lsp->ls_stateid);
6416 	}
6417 	if (!nfs4_valid_open_stateid(state)) {
6418 		data->rpc_status = -EBADF;
6419 		task->tk_action = NULL;
6420 		goto out_release_open_seqid;
6421 	}
6422 	data->timestamp = jiffies;
6423 	if (nfs4_setup_sequence(data->server->nfs_client,
6424 				&data->arg.seq_args,
6425 				&data->res.seq_res,
6426 				task) == 0)
6427 		return;
6428 out_release_open_seqid:
6429 	nfs_release_seqid(data->arg.open_seqid);
6430 out_release_lock_seqid:
6431 	nfs_release_seqid(data->arg.lock_seqid);
6432 out_wait:
6433 	nfs4_sequence_done(task, &data->res.seq_res);
6434 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6435 }
6436 
6437 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6438 {
6439 	struct nfs4_lockdata *data = calldata;
6440 	struct nfs4_lock_state *lsp = data->lsp;
6441 
6442 	dprintk("%s: begin!\n", __func__);
6443 
6444 	if (!nfs4_sequence_done(task, &data->res.seq_res))
6445 		return;
6446 
6447 	data->rpc_status = task->tk_status;
6448 	switch (task->tk_status) {
6449 	case 0:
6450 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6451 				data->timestamp);
6452 		if (data->arg.new_lock && !data->cancelled) {
6453 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6454 			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6455 				break;
6456 		}
6457 
6458 		if (data->arg.new_lock_owner != 0) {
6459 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
6460 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6461 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6462 			goto out_done;
6463 		} else if (nfs4_update_lock_stateid(lsp, &data->res.stateid))
6464 			goto out_done;
6465 
6466 		break;
6467 	case -NFS4ERR_BAD_STATEID:
6468 	case -NFS4ERR_OLD_STATEID:
6469 	case -NFS4ERR_STALE_STATEID:
6470 	case -NFS4ERR_EXPIRED:
6471 		if (data->arg.new_lock_owner != 0) {
6472 			if (nfs4_stateid_match(&data->arg.open_stateid,
6473 						&lsp->ls_state->open_stateid))
6474 				goto out_done;
6475 		} else if (nfs4_stateid_match(&data->arg.lock_stateid,
6476 						&lsp->ls_stateid))
6477 				goto out_done;
6478 	}
6479 	if (!data->cancelled)
6480 		rpc_restart_call_prepare(task);
6481 out_done:
6482 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6483 }
6484 
6485 static void nfs4_lock_release(void *calldata)
6486 {
6487 	struct nfs4_lockdata *data = calldata;
6488 
6489 	dprintk("%s: begin!\n", __func__);
6490 	nfs_free_seqid(data->arg.open_seqid);
6491 	if (data->cancelled) {
6492 		struct rpc_task *task;
6493 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6494 				data->arg.lock_seqid);
6495 		if (!IS_ERR(task))
6496 			rpc_put_task_async(task);
6497 		dprintk("%s: cancelling lock!\n", __func__);
6498 	} else
6499 		nfs_free_seqid(data->arg.lock_seqid);
6500 	nfs4_put_lock_state(data->lsp);
6501 	put_nfs_open_context(data->ctx);
6502 	kfree(data);
6503 	dprintk("%s: done!\n", __func__);
6504 }
6505 
6506 static const struct rpc_call_ops nfs4_lock_ops = {
6507 	.rpc_call_prepare = nfs4_lock_prepare,
6508 	.rpc_call_done = nfs4_lock_done,
6509 	.rpc_release = nfs4_lock_release,
6510 };
6511 
6512 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6513 {
6514 	switch (error) {
6515 	case -NFS4ERR_ADMIN_REVOKED:
6516 	case -NFS4ERR_EXPIRED:
6517 	case -NFS4ERR_BAD_STATEID:
6518 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6519 		if (new_lock_owner != 0 ||
6520 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6521 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6522 		break;
6523 	case -NFS4ERR_STALE_STATEID:
6524 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6525 		nfs4_schedule_lease_recovery(server->nfs_client);
6526 	};
6527 }
6528 
6529 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6530 {
6531 	struct nfs4_lockdata *data;
6532 	struct rpc_task *task;
6533 	struct rpc_message msg = {
6534 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6535 		.rpc_cred = state->owner->so_cred,
6536 	};
6537 	struct rpc_task_setup task_setup_data = {
6538 		.rpc_client = NFS_CLIENT(state->inode),
6539 		.rpc_message = &msg,
6540 		.callback_ops = &nfs4_lock_ops,
6541 		.workqueue = nfsiod_workqueue,
6542 		.flags = RPC_TASK_ASYNC,
6543 	};
6544 	int ret;
6545 
6546 	dprintk("%s: begin!\n", __func__);
6547 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6548 			fl->fl_u.nfs4_fl.owner,
6549 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6550 	if (data == NULL)
6551 		return -ENOMEM;
6552 	if (IS_SETLKW(cmd))
6553 		data->arg.block = 1;
6554 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6555 				recovery_type > NFS_LOCK_NEW);
6556 	msg.rpc_argp = &data->arg;
6557 	msg.rpc_resp = &data->res;
6558 	task_setup_data.callback_data = data;
6559 	if (recovery_type > NFS_LOCK_NEW) {
6560 		if (recovery_type == NFS_LOCK_RECLAIM)
6561 			data->arg.reclaim = NFS_LOCK_RECLAIM;
6562 	} else
6563 		data->arg.new_lock = 1;
6564 	task = rpc_run_task(&task_setup_data);
6565 	if (IS_ERR(task))
6566 		return PTR_ERR(task);
6567 	ret = rpc_wait_for_completion_task(task);
6568 	if (ret == 0) {
6569 		ret = data->rpc_status;
6570 		if (ret)
6571 			nfs4_handle_setlk_error(data->server, data->lsp,
6572 					data->arg.new_lock_owner, ret);
6573 	} else
6574 		data->cancelled = true;
6575 	rpc_put_task(task);
6576 	dprintk("%s: done, ret = %d!\n", __func__, ret);
6577 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6578 	return ret;
6579 }
6580 
6581 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6582 {
6583 	struct nfs_server *server = NFS_SERVER(state->inode);
6584 	struct nfs4_exception exception = {
6585 		.inode = state->inode,
6586 	};
6587 	int err;
6588 
6589 	do {
6590 		/* Cache the lock if possible... */
6591 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6592 			return 0;
6593 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6594 		if (err != -NFS4ERR_DELAY)
6595 			break;
6596 		nfs4_handle_exception(server, err, &exception);
6597 	} while (exception.retry);
6598 	return err;
6599 }
6600 
6601 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6602 {
6603 	struct nfs_server *server = NFS_SERVER(state->inode);
6604 	struct nfs4_exception exception = {
6605 		.inode = state->inode,
6606 	};
6607 	int err;
6608 
6609 	err = nfs4_set_lock_state(state, request);
6610 	if (err != 0)
6611 		return err;
6612 	if (!recover_lost_locks) {
6613 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6614 		return 0;
6615 	}
6616 	do {
6617 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6618 			return 0;
6619 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6620 		switch (err) {
6621 		default:
6622 			goto out;
6623 		case -NFS4ERR_GRACE:
6624 		case -NFS4ERR_DELAY:
6625 			nfs4_handle_exception(server, err, &exception);
6626 			err = 0;
6627 		}
6628 	} while (exception.retry);
6629 out:
6630 	return err;
6631 }
6632 
6633 #if defined(CONFIG_NFS_V4_1)
6634 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6635 {
6636 	struct nfs4_lock_state *lsp;
6637 	int status;
6638 
6639 	status = nfs4_set_lock_state(state, request);
6640 	if (status != 0)
6641 		return status;
6642 	lsp = request->fl_u.nfs4_fl.owner;
6643 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6644 	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6645 		return 0;
6646 	return nfs4_lock_expired(state, request);
6647 }
6648 #endif
6649 
6650 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6651 {
6652 	struct nfs_inode *nfsi = NFS_I(state->inode);
6653 	struct nfs4_state_owner *sp = state->owner;
6654 	unsigned char fl_flags = request->fl_flags;
6655 	int status;
6656 
6657 	request->fl_flags |= FL_ACCESS;
6658 	status = locks_lock_inode_wait(state->inode, request);
6659 	if (status < 0)
6660 		goto out;
6661 	mutex_lock(&sp->so_delegreturn_mutex);
6662 	down_read(&nfsi->rwsem);
6663 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6664 		/* Yes: cache locks! */
6665 		/* ...but avoid races with delegation recall... */
6666 		request->fl_flags = fl_flags & ~FL_SLEEP;
6667 		status = locks_lock_inode_wait(state->inode, request);
6668 		up_read(&nfsi->rwsem);
6669 		mutex_unlock(&sp->so_delegreturn_mutex);
6670 		goto out;
6671 	}
6672 	up_read(&nfsi->rwsem);
6673 	mutex_unlock(&sp->so_delegreturn_mutex);
6674 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6675 out:
6676 	request->fl_flags = fl_flags;
6677 	return status;
6678 }
6679 
6680 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6681 {
6682 	struct nfs4_exception exception = {
6683 		.state = state,
6684 		.inode = state->inode,
6685 	};
6686 	int err;
6687 
6688 	do {
6689 		err = _nfs4_proc_setlk(state, cmd, request);
6690 		if (err == -NFS4ERR_DENIED)
6691 			err = -EAGAIN;
6692 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
6693 				err, &exception);
6694 	} while (exception.retry);
6695 	return err;
6696 }
6697 
6698 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6699 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6700 
6701 static int
6702 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6703 			struct file_lock *request)
6704 {
6705 	int		status = -ERESTARTSYS;
6706 	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
6707 
6708 	while(!signalled()) {
6709 		status = nfs4_proc_setlk(state, cmd, request);
6710 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6711 			break;
6712 		freezable_schedule_timeout_interruptible(timeout);
6713 		timeout *= 2;
6714 		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6715 		status = -ERESTARTSYS;
6716 	}
6717 	return status;
6718 }
6719 
6720 #ifdef CONFIG_NFS_V4_1
6721 struct nfs4_lock_waiter {
6722 	struct task_struct	*task;
6723 	struct inode		*inode;
6724 	struct nfs_lowner	*owner;
6725 	bool			notified;
6726 };
6727 
6728 static int
6729 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6730 {
6731 	int ret;
6732 	struct nfs4_lock_waiter	*waiter	= wait->private;
6733 
6734 	/* NULL key means to wake up everyone */
6735 	if (key) {
6736 		struct cb_notify_lock_args	*cbnl = key;
6737 		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
6738 						*wowner = waiter->owner;
6739 
6740 		/* Only wake if the callback was for the same owner. */
6741 		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6742 			return 0;
6743 
6744 		/* Make sure it's for the right inode */
6745 		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6746 			return 0;
6747 
6748 		waiter->notified = true;
6749 	}
6750 
6751 	/* override "private" so we can use default_wake_function */
6752 	wait->private = waiter->task;
6753 	ret = autoremove_wake_function(wait, mode, flags, key);
6754 	wait->private = waiter;
6755 	return ret;
6756 }
6757 
6758 static int
6759 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6760 {
6761 	int status = -ERESTARTSYS;
6762 	unsigned long flags;
6763 	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6764 	struct nfs_server *server = NFS_SERVER(state->inode);
6765 	struct nfs_client *clp = server->nfs_client;
6766 	wait_queue_head_t *q = &clp->cl_lock_waitq;
6767 	struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6768 				    .id = lsp->ls_seqid.owner_id,
6769 				    .s_dev = server->s_dev };
6770 	struct nfs4_lock_waiter waiter = { .task  = current,
6771 					   .inode = state->inode,
6772 					   .owner = &owner,
6773 					   .notified = false };
6774 	wait_queue_entry_t wait;
6775 
6776 	/* Don't bother with waitqueue if we don't expect a callback */
6777 	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6778 		return nfs4_retry_setlk_simple(state, cmd, request);
6779 
6780 	init_wait(&wait);
6781 	wait.private = &waiter;
6782 	wait.func = nfs4_wake_lock_waiter;
6783 	add_wait_queue(q, &wait);
6784 
6785 	while(!signalled()) {
6786 		waiter.notified = false;
6787 		status = nfs4_proc_setlk(state, cmd, request);
6788 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6789 			break;
6790 
6791 		status = -ERESTARTSYS;
6792 		spin_lock_irqsave(&q->lock, flags);
6793 		if (waiter.notified) {
6794 			spin_unlock_irqrestore(&q->lock, flags);
6795 			continue;
6796 		}
6797 		set_current_state(TASK_INTERRUPTIBLE);
6798 		spin_unlock_irqrestore(&q->lock, flags);
6799 
6800 		freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6801 	}
6802 
6803 	finish_wait(q, &wait);
6804 	return status;
6805 }
6806 #else /* !CONFIG_NFS_V4_1 */
6807 static inline int
6808 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6809 {
6810 	return nfs4_retry_setlk_simple(state, cmd, request);
6811 }
6812 #endif
6813 
6814 static int
6815 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6816 {
6817 	struct nfs_open_context *ctx;
6818 	struct nfs4_state *state;
6819 	int status;
6820 
6821 	/* verify open state */
6822 	ctx = nfs_file_open_context(filp);
6823 	state = ctx->state;
6824 
6825 	if (IS_GETLK(cmd)) {
6826 		if (state != NULL)
6827 			return nfs4_proc_getlk(state, F_GETLK, request);
6828 		return 0;
6829 	}
6830 
6831 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6832 		return -EINVAL;
6833 
6834 	if (request->fl_type == F_UNLCK) {
6835 		if (state != NULL)
6836 			return nfs4_proc_unlck(state, cmd, request);
6837 		return 0;
6838 	}
6839 
6840 	if (state == NULL)
6841 		return -ENOLCK;
6842 
6843 	if ((request->fl_flags & FL_POSIX) &&
6844 	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6845 		return -ENOLCK;
6846 
6847 	/*
6848 	 * Don't rely on the VFS having checked the file open mode,
6849 	 * since it won't do this for flock() locks.
6850 	 */
6851 	switch (request->fl_type) {
6852 	case F_RDLCK:
6853 		if (!(filp->f_mode & FMODE_READ))
6854 			return -EBADF;
6855 		break;
6856 	case F_WRLCK:
6857 		if (!(filp->f_mode & FMODE_WRITE))
6858 			return -EBADF;
6859 	}
6860 
6861 	status = nfs4_set_lock_state(state, request);
6862 	if (status != 0)
6863 		return status;
6864 
6865 	return nfs4_retry_setlk(state, cmd, request);
6866 }
6867 
6868 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6869 {
6870 	struct nfs_server *server = NFS_SERVER(state->inode);
6871 	int err;
6872 
6873 	err = nfs4_set_lock_state(state, fl);
6874 	if (err != 0)
6875 		return err;
6876 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6877 	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
6878 }
6879 
6880 struct nfs_release_lockowner_data {
6881 	struct nfs4_lock_state *lsp;
6882 	struct nfs_server *server;
6883 	struct nfs_release_lockowner_args args;
6884 	struct nfs_release_lockowner_res res;
6885 	unsigned long timestamp;
6886 };
6887 
6888 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6889 {
6890 	struct nfs_release_lockowner_data *data = calldata;
6891 	struct nfs_server *server = data->server;
6892 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
6893 			   &data->res.seq_res, task);
6894 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6895 	data->timestamp = jiffies;
6896 }
6897 
6898 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6899 {
6900 	struct nfs_release_lockowner_data *data = calldata;
6901 	struct nfs_server *server = data->server;
6902 
6903 	nfs40_sequence_done(task, &data->res.seq_res);
6904 
6905 	switch (task->tk_status) {
6906 	case 0:
6907 		renew_lease(server, data->timestamp);
6908 		break;
6909 	case -NFS4ERR_STALE_CLIENTID:
6910 	case -NFS4ERR_EXPIRED:
6911 		nfs4_schedule_lease_recovery(server->nfs_client);
6912 		break;
6913 	case -NFS4ERR_LEASE_MOVED:
6914 	case -NFS4ERR_DELAY:
6915 		if (nfs4_async_handle_error(task, server,
6916 					    NULL, NULL) == -EAGAIN)
6917 			rpc_restart_call_prepare(task);
6918 	}
6919 }
6920 
6921 static void nfs4_release_lockowner_release(void *calldata)
6922 {
6923 	struct nfs_release_lockowner_data *data = calldata;
6924 	nfs4_free_lock_state(data->server, data->lsp);
6925 	kfree(calldata);
6926 }
6927 
6928 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6929 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6930 	.rpc_call_done = nfs4_release_lockowner_done,
6931 	.rpc_release = nfs4_release_lockowner_release,
6932 };
6933 
6934 static void
6935 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6936 {
6937 	struct nfs_release_lockowner_data *data;
6938 	struct rpc_message msg = {
6939 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6940 	};
6941 
6942 	if (server->nfs_client->cl_mvops->minor_version != 0)
6943 		return;
6944 
6945 	data = kmalloc(sizeof(*data), GFP_NOFS);
6946 	if (!data)
6947 		return;
6948 	data->lsp = lsp;
6949 	data->server = server;
6950 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6951 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6952 	data->args.lock_owner.s_dev = server->s_dev;
6953 
6954 	msg.rpc_argp = &data->args;
6955 	msg.rpc_resp = &data->res;
6956 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
6957 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6958 }
6959 
6960 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6961 
6962 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6963 				   struct dentry *unused, struct inode *inode,
6964 				   const char *key, const void *buf,
6965 				   size_t buflen, int flags)
6966 {
6967 	return nfs4_proc_set_acl(inode, buf, buflen);
6968 }
6969 
6970 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6971 				   struct dentry *unused, struct inode *inode,
6972 				   const char *key, void *buf, size_t buflen)
6973 {
6974 	return nfs4_proc_get_acl(inode, buf, buflen);
6975 }
6976 
6977 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6978 {
6979 	return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6980 }
6981 
6982 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6983 
6984 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6985 				     struct dentry *unused, struct inode *inode,
6986 				     const char *key, const void *buf,
6987 				     size_t buflen, int flags)
6988 {
6989 	if (security_ismaclabel(key))
6990 		return nfs4_set_security_label(inode, buf, buflen);
6991 
6992 	return -EOPNOTSUPP;
6993 }
6994 
6995 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6996 				     struct dentry *unused, struct inode *inode,
6997 				     const char *key, void *buf, size_t buflen)
6998 {
6999 	if (security_ismaclabel(key))
7000 		return nfs4_get_security_label(inode, buf, buflen);
7001 	return -EOPNOTSUPP;
7002 }
7003 
7004 static ssize_t
7005 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7006 {
7007 	int len = 0;
7008 
7009 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7010 		len = security_inode_listsecurity(inode, list, list_len);
7011 		if (list_len && len > list_len)
7012 			return -ERANGE;
7013 	}
7014 	return len;
7015 }
7016 
7017 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7018 	.prefix = XATTR_SECURITY_PREFIX,
7019 	.get	= nfs4_xattr_get_nfs4_label,
7020 	.set	= nfs4_xattr_set_nfs4_label,
7021 };
7022 
7023 #else
7024 
7025 static ssize_t
7026 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7027 {
7028 	return 0;
7029 }
7030 
7031 #endif
7032 
7033 /*
7034  * nfs_fhget will use either the mounted_on_fileid or the fileid
7035  */
7036 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7037 {
7038 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7039 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7040 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7041 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7042 		return;
7043 
7044 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7045 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7046 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7047 	fattr->nlink = 2;
7048 }
7049 
7050 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7051 				   const struct qstr *name,
7052 				   struct nfs4_fs_locations *fs_locations,
7053 				   struct page *page)
7054 {
7055 	struct nfs_server *server = NFS_SERVER(dir);
7056 	u32 bitmask[3];
7057 	struct nfs4_fs_locations_arg args = {
7058 		.dir_fh = NFS_FH(dir),
7059 		.name = name,
7060 		.page = page,
7061 		.bitmask = bitmask,
7062 	};
7063 	struct nfs4_fs_locations_res res = {
7064 		.fs_locations = fs_locations,
7065 	};
7066 	struct rpc_message msg = {
7067 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7068 		.rpc_argp = &args,
7069 		.rpc_resp = &res,
7070 	};
7071 	int status;
7072 
7073 	dprintk("%s: start\n", __func__);
7074 
7075 	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7076 	bitmask[1] = nfs4_fattr_bitmap[1];
7077 
7078 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
7079 	 * is not supported */
7080 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7081 		bitmask[0] &= ~FATTR4_WORD0_FILEID;
7082 	else
7083 		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7084 
7085 	nfs_fattr_init(&fs_locations->fattr);
7086 	fs_locations->server = server;
7087 	fs_locations->nlocations = 0;
7088 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7089 	dprintk("%s: returned status = %d\n", __func__, status);
7090 	return status;
7091 }
7092 
7093 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7094 			   const struct qstr *name,
7095 			   struct nfs4_fs_locations *fs_locations,
7096 			   struct page *page)
7097 {
7098 	struct nfs4_exception exception = { };
7099 	int err;
7100 	do {
7101 		err = _nfs4_proc_fs_locations(client, dir, name,
7102 				fs_locations, page);
7103 		trace_nfs4_get_fs_locations(dir, name, err);
7104 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
7105 				&exception);
7106 	} while (exception.retry);
7107 	return err;
7108 }
7109 
7110 /*
7111  * This operation also signals the server that this client is
7112  * performing migration recovery.  The server can stop returning
7113  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7114  * appended to this compound to identify the client ID which is
7115  * performing recovery.
7116  */
7117 static int _nfs40_proc_get_locations(struct inode *inode,
7118 				     struct nfs4_fs_locations *locations,
7119 				     struct page *page, struct rpc_cred *cred)
7120 {
7121 	struct nfs_server *server = NFS_SERVER(inode);
7122 	struct rpc_clnt *clnt = server->client;
7123 	u32 bitmask[2] = {
7124 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7125 	};
7126 	struct nfs4_fs_locations_arg args = {
7127 		.clientid	= server->nfs_client->cl_clientid,
7128 		.fh		= NFS_FH(inode),
7129 		.page		= page,
7130 		.bitmask	= bitmask,
7131 		.migration	= 1,		/* skip LOOKUP */
7132 		.renew		= 1,		/* append RENEW */
7133 	};
7134 	struct nfs4_fs_locations_res res = {
7135 		.fs_locations	= locations,
7136 		.migration	= 1,
7137 		.renew		= 1,
7138 	};
7139 	struct rpc_message msg = {
7140 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7141 		.rpc_argp	= &args,
7142 		.rpc_resp	= &res,
7143 		.rpc_cred	= cred,
7144 	};
7145 	unsigned long now = jiffies;
7146 	int status;
7147 
7148 	nfs_fattr_init(&locations->fattr);
7149 	locations->server = server;
7150 	locations->nlocations = 0;
7151 
7152 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7153 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7154 					&args.seq_args, &res.seq_res);
7155 	if (status)
7156 		return status;
7157 
7158 	renew_lease(server, now);
7159 	return 0;
7160 }
7161 
7162 #ifdef CONFIG_NFS_V4_1
7163 
7164 /*
7165  * This operation also signals the server that this client is
7166  * performing migration recovery.  The server can stop asserting
7167  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7168  * performing this operation is identified in the SEQUENCE
7169  * operation in this compound.
7170  *
7171  * When the client supports GETATTR(fs_locations_info), it can
7172  * be plumbed in here.
7173  */
7174 static int _nfs41_proc_get_locations(struct inode *inode,
7175 				     struct nfs4_fs_locations *locations,
7176 				     struct page *page, struct rpc_cred *cred)
7177 {
7178 	struct nfs_server *server = NFS_SERVER(inode);
7179 	struct rpc_clnt *clnt = server->client;
7180 	u32 bitmask[2] = {
7181 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7182 	};
7183 	struct nfs4_fs_locations_arg args = {
7184 		.fh		= NFS_FH(inode),
7185 		.page		= page,
7186 		.bitmask	= bitmask,
7187 		.migration	= 1,		/* skip LOOKUP */
7188 	};
7189 	struct nfs4_fs_locations_res res = {
7190 		.fs_locations	= locations,
7191 		.migration	= 1,
7192 	};
7193 	struct rpc_message msg = {
7194 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7195 		.rpc_argp	= &args,
7196 		.rpc_resp	= &res,
7197 		.rpc_cred	= cred,
7198 	};
7199 	int status;
7200 
7201 	nfs_fattr_init(&locations->fattr);
7202 	locations->server = server;
7203 	locations->nlocations = 0;
7204 
7205 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7206 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7207 					&args.seq_args, &res.seq_res);
7208 	if (status == NFS4_OK &&
7209 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7210 		status = -NFS4ERR_LEASE_MOVED;
7211 	return status;
7212 }
7213 
7214 #endif	/* CONFIG_NFS_V4_1 */
7215 
7216 /**
7217  * nfs4_proc_get_locations - discover locations for a migrated FSID
7218  * @inode: inode on FSID that is migrating
7219  * @locations: result of query
7220  * @page: buffer
7221  * @cred: credential to use for this operation
7222  *
7223  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7224  * operation failed, or a negative errno if a local error occurred.
7225  *
7226  * On success, "locations" is filled in, but if the server has
7227  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7228  * asserted.
7229  *
7230  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7231  * from this client that require migration recovery.
7232  */
7233 int nfs4_proc_get_locations(struct inode *inode,
7234 			    struct nfs4_fs_locations *locations,
7235 			    struct page *page, struct rpc_cred *cred)
7236 {
7237 	struct nfs_server *server = NFS_SERVER(inode);
7238 	struct nfs_client *clp = server->nfs_client;
7239 	const struct nfs4_mig_recovery_ops *ops =
7240 					clp->cl_mvops->mig_recovery_ops;
7241 	struct nfs4_exception exception = { };
7242 	int status;
7243 
7244 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7245 		(unsigned long long)server->fsid.major,
7246 		(unsigned long long)server->fsid.minor,
7247 		clp->cl_hostname);
7248 	nfs_display_fhandle(NFS_FH(inode), __func__);
7249 
7250 	do {
7251 		status = ops->get_locations(inode, locations, page, cred);
7252 		if (status != -NFS4ERR_DELAY)
7253 			break;
7254 		nfs4_handle_exception(server, status, &exception);
7255 	} while (exception.retry);
7256 	return status;
7257 }
7258 
7259 /*
7260  * This operation also signals the server that this client is
7261  * performing "lease moved" recovery.  The server can stop
7262  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7263  * is appended to this compound to identify the client ID which is
7264  * performing recovery.
7265  */
7266 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7267 {
7268 	struct nfs_server *server = NFS_SERVER(inode);
7269 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7270 	struct rpc_clnt *clnt = server->client;
7271 	struct nfs4_fsid_present_arg args = {
7272 		.fh		= NFS_FH(inode),
7273 		.clientid	= clp->cl_clientid,
7274 		.renew		= 1,		/* append RENEW */
7275 	};
7276 	struct nfs4_fsid_present_res res = {
7277 		.renew		= 1,
7278 	};
7279 	struct rpc_message msg = {
7280 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7281 		.rpc_argp	= &args,
7282 		.rpc_resp	= &res,
7283 		.rpc_cred	= cred,
7284 	};
7285 	unsigned long now = jiffies;
7286 	int status;
7287 
7288 	res.fh = nfs_alloc_fhandle();
7289 	if (res.fh == NULL)
7290 		return -ENOMEM;
7291 
7292 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7293 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7294 						&args.seq_args, &res.seq_res);
7295 	nfs_free_fhandle(res.fh);
7296 	if (status)
7297 		return status;
7298 
7299 	do_renew_lease(clp, now);
7300 	return 0;
7301 }
7302 
7303 #ifdef CONFIG_NFS_V4_1
7304 
7305 /*
7306  * This operation also signals the server that this client is
7307  * performing "lease moved" recovery.  The server can stop asserting
7308  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7309  * this operation is identified in the SEQUENCE operation in this
7310  * compound.
7311  */
7312 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7313 {
7314 	struct nfs_server *server = NFS_SERVER(inode);
7315 	struct rpc_clnt *clnt = server->client;
7316 	struct nfs4_fsid_present_arg args = {
7317 		.fh		= NFS_FH(inode),
7318 	};
7319 	struct nfs4_fsid_present_res res = {
7320 	};
7321 	struct rpc_message msg = {
7322 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7323 		.rpc_argp	= &args,
7324 		.rpc_resp	= &res,
7325 		.rpc_cred	= cred,
7326 	};
7327 	int status;
7328 
7329 	res.fh = nfs_alloc_fhandle();
7330 	if (res.fh == NULL)
7331 		return -ENOMEM;
7332 
7333 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7334 	status = nfs4_call_sync_sequence(clnt, server, &msg,
7335 						&args.seq_args, &res.seq_res);
7336 	nfs_free_fhandle(res.fh);
7337 	if (status == NFS4_OK &&
7338 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7339 		status = -NFS4ERR_LEASE_MOVED;
7340 	return status;
7341 }
7342 
7343 #endif	/* CONFIG_NFS_V4_1 */
7344 
7345 /**
7346  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7347  * @inode: inode on FSID to check
7348  * @cred: credential to use for this operation
7349  *
7350  * Server indicates whether the FSID is present, moved, or not
7351  * recognized.  This operation is necessary to clear a LEASE_MOVED
7352  * condition for this client ID.
7353  *
7354  * Returns NFS4_OK if the FSID is present on this server,
7355  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7356  *  NFS4ERR code if some error occurred on the server, or a
7357  *  negative errno if a local failure occurred.
7358  */
7359 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7360 {
7361 	struct nfs_server *server = NFS_SERVER(inode);
7362 	struct nfs_client *clp = server->nfs_client;
7363 	const struct nfs4_mig_recovery_ops *ops =
7364 					clp->cl_mvops->mig_recovery_ops;
7365 	struct nfs4_exception exception = { };
7366 	int status;
7367 
7368 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7369 		(unsigned long long)server->fsid.major,
7370 		(unsigned long long)server->fsid.minor,
7371 		clp->cl_hostname);
7372 	nfs_display_fhandle(NFS_FH(inode), __func__);
7373 
7374 	do {
7375 		status = ops->fsid_present(inode, cred);
7376 		if (status != -NFS4ERR_DELAY)
7377 			break;
7378 		nfs4_handle_exception(server, status, &exception);
7379 	} while (exception.retry);
7380 	return status;
7381 }
7382 
7383 /**
7384  * If 'use_integrity' is true and the state managment nfs_client
7385  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7386  * and the machine credential as per RFC3530bis and RFC5661 Security
7387  * Considerations sections. Otherwise, just use the user cred with the
7388  * filesystem's rpc_client.
7389  */
7390 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7391 {
7392 	int status;
7393 	struct nfs4_secinfo_arg args = {
7394 		.dir_fh = NFS_FH(dir),
7395 		.name   = name,
7396 	};
7397 	struct nfs4_secinfo_res res = {
7398 		.flavors     = flavors,
7399 	};
7400 	struct rpc_message msg = {
7401 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7402 		.rpc_argp = &args,
7403 		.rpc_resp = &res,
7404 	};
7405 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7406 	struct rpc_cred *cred = NULL;
7407 
7408 	if (use_integrity) {
7409 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7410 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7411 		msg.rpc_cred = cred;
7412 	}
7413 
7414 	dprintk("NFS call  secinfo %s\n", name->name);
7415 
7416 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7417 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7418 
7419 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7420 				&res.seq_res, 0);
7421 	dprintk("NFS reply  secinfo: %d\n", status);
7422 
7423 	if (cred)
7424 		put_rpccred(cred);
7425 
7426 	return status;
7427 }
7428 
7429 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7430 		      struct nfs4_secinfo_flavors *flavors)
7431 {
7432 	struct nfs4_exception exception = { };
7433 	int err;
7434 	do {
7435 		err = -NFS4ERR_WRONGSEC;
7436 
7437 		/* try to use integrity protection with machine cred */
7438 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7439 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
7440 
7441 		/*
7442 		 * if unable to use integrity protection, or SECINFO with
7443 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
7444 		 * disallowed by spec, but exists in deployed servers) use
7445 		 * the current filesystem's rpc_client and the user cred.
7446 		 */
7447 		if (err == -NFS4ERR_WRONGSEC)
7448 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
7449 
7450 		trace_nfs4_secinfo(dir, name, err);
7451 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
7452 				&exception);
7453 	} while (exception.retry);
7454 	return err;
7455 }
7456 
7457 #ifdef CONFIG_NFS_V4_1
7458 /*
7459  * Check the exchange flags returned by the server for invalid flags, having
7460  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7461  * DS flags set.
7462  */
7463 static int nfs4_check_cl_exchange_flags(u32 flags)
7464 {
7465 	if (flags & ~EXCHGID4_FLAG_MASK_R)
7466 		goto out_inval;
7467 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7468 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7469 		goto out_inval;
7470 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7471 		goto out_inval;
7472 	return NFS_OK;
7473 out_inval:
7474 	return -NFS4ERR_INVAL;
7475 }
7476 
7477 static bool
7478 nfs41_same_server_scope(struct nfs41_server_scope *a,
7479 			struct nfs41_server_scope *b)
7480 {
7481 	if (a->server_scope_sz != b->server_scope_sz)
7482 		return false;
7483 	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7484 }
7485 
7486 static void
7487 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7488 {
7489 }
7490 
7491 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7492 	.rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7493 };
7494 
7495 /*
7496  * nfs4_proc_bind_one_conn_to_session()
7497  *
7498  * The 4.1 client currently uses the same TCP connection for the
7499  * fore and backchannel.
7500  */
7501 static
7502 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7503 		struct rpc_xprt *xprt,
7504 		struct nfs_client *clp,
7505 		struct rpc_cred *cred)
7506 {
7507 	int status;
7508 	struct nfs41_bind_conn_to_session_args args = {
7509 		.client = clp,
7510 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
7511 	};
7512 	struct nfs41_bind_conn_to_session_res res;
7513 	struct rpc_message msg = {
7514 		.rpc_proc =
7515 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7516 		.rpc_argp = &args,
7517 		.rpc_resp = &res,
7518 		.rpc_cred = cred,
7519 	};
7520 	struct rpc_task_setup task_setup_data = {
7521 		.rpc_client = clnt,
7522 		.rpc_xprt = xprt,
7523 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
7524 		.rpc_message = &msg,
7525 		.flags = RPC_TASK_TIMEOUT,
7526 	};
7527 	struct rpc_task *task;
7528 
7529 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7530 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7531 		args.dir = NFS4_CDFC4_FORE;
7532 
7533 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
7534 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
7535 		args.dir = NFS4_CDFC4_FORE;
7536 
7537 	task = rpc_run_task(&task_setup_data);
7538 	if (!IS_ERR(task)) {
7539 		status = task->tk_status;
7540 		rpc_put_task(task);
7541 	} else
7542 		status = PTR_ERR(task);
7543 	trace_nfs4_bind_conn_to_session(clp, status);
7544 	if (status == 0) {
7545 		if (memcmp(res.sessionid.data,
7546 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7547 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
7548 			return -EIO;
7549 		}
7550 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7551 			dprintk("NFS: %s: Unexpected direction from server\n",
7552 				__func__);
7553 			return -EIO;
7554 		}
7555 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7556 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
7557 				__func__);
7558 			return -EIO;
7559 		}
7560 	}
7561 
7562 	return status;
7563 }
7564 
7565 struct rpc_bind_conn_calldata {
7566 	struct nfs_client *clp;
7567 	struct rpc_cred *cred;
7568 };
7569 
7570 static int
7571 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7572 		struct rpc_xprt *xprt,
7573 		void *calldata)
7574 {
7575 	struct rpc_bind_conn_calldata *p = calldata;
7576 
7577 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7578 }
7579 
7580 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7581 {
7582 	struct rpc_bind_conn_calldata data = {
7583 		.clp = clp,
7584 		.cred = cred,
7585 	};
7586 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7587 			nfs4_proc_bind_conn_to_session_callback, &data);
7588 }
7589 
7590 /*
7591  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7592  * and operations we'd like to see to enable certain features in the allow map
7593  */
7594 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7595 	.how = SP4_MACH_CRED,
7596 	.enforce.u.words = {
7597 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7598 		      1 << (OP_EXCHANGE_ID - 32) |
7599 		      1 << (OP_CREATE_SESSION - 32) |
7600 		      1 << (OP_DESTROY_SESSION - 32) |
7601 		      1 << (OP_DESTROY_CLIENTID - 32)
7602 	},
7603 	.allow.u.words = {
7604 		[0] = 1 << (OP_CLOSE) |
7605 		      1 << (OP_OPEN_DOWNGRADE) |
7606 		      1 << (OP_LOCKU) |
7607 		      1 << (OP_DELEGRETURN) |
7608 		      1 << (OP_COMMIT),
7609 		[1] = 1 << (OP_SECINFO - 32) |
7610 		      1 << (OP_SECINFO_NO_NAME - 32) |
7611 		      1 << (OP_LAYOUTRETURN - 32) |
7612 		      1 << (OP_TEST_STATEID - 32) |
7613 		      1 << (OP_FREE_STATEID - 32) |
7614 		      1 << (OP_WRITE - 32)
7615 	}
7616 };
7617 
7618 /*
7619  * Select the state protection mode for client `clp' given the server results
7620  * from exchange_id in `sp'.
7621  *
7622  * Returns 0 on success, negative errno otherwise.
7623  */
7624 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7625 				 struct nfs41_state_protection *sp)
7626 {
7627 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7628 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7629 		      1 << (OP_EXCHANGE_ID - 32) |
7630 		      1 << (OP_CREATE_SESSION - 32) |
7631 		      1 << (OP_DESTROY_SESSION - 32) |
7632 		      1 << (OP_DESTROY_CLIENTID - 32)
7633 	};
7634 	unsigned long flags = 0;
7635 	unsigned int i;
7636 	int ret = 0;
7637 
7638 	if (sp->how == SP4_MACH_CRED) {
7639 		/* Print state protect result */
7640 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7641 		for (i = 0; i <= LAST_NFS4_OP; i++) {
7642 			if (test_bit(i, sp->enforce.u.longs))
7643 				dfprintk(MOUNT, "  enforce op %d\n", i);
7644 			if (test_bit(i, sp->allow.u.longs))
7645 				dfprintk(MOUNT, "  allow op %d\n", i);
7646 		}
7647 
7648 		/* make sure nothing is on enforce list that isn't supported */
7649 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7650 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7651 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7652 				ret = -EINVAL;
7653 				goto out;
7654 			}
7655 		}
7656 
7657 		/*
7658 		 * Minimal mode - state operations are allowed to use machine
7659 		 * credential.  Note this already happens by default, so the
7660 		 * client doesn't have to do anything more than the negotiation.
7661 		 *
7662 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
7663 		 *       we're already using the machine cred for exchange_id
7664 		 *       and will never use a different cred.
7665 		 */
7666 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7667 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7668 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7669 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7670 			dfprintk(MOUNT, "sp4_mach_cred:\n");
7671 			dfprintk(MOUNT, "  minimal mode enabled\n");
7672 			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7673 		} else {
7674 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7675 			ret = -EINVAL;
7676 			goto out;
7677 		}
7678 
7679 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7680 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7681 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7682 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
7683 			dfprintk(MOUNT, "  cleanup mode enabled\n");
7684 			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7685 		}
7686 
7687 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7688 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7689 			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7690 		}
7691 
7692 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7693 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7694 			dfprintk(MOUNT, "  secinfo mode enabled\n");
7695 			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7696 		}
7697 
7698 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7699 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7700 			dfprintk(MOUNT, "  stateid mode enabled\n");
7701 			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7702 		}
7703 
7704 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7705 			dfprintk(MOUNT, "  write mode enabled\n");
7706 			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7707 		}
7708 
7709 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7710 			dfprintk(MOUNT, "  commit mode enabled\n");
7711 			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7712 		}
7713 	}
7714 out:
7715 	clp->cl_sp4_flags = flags;
7716 	return 0;
7717 }
7718 
7719 struct nfs41_exchange_id_data {
7720 	struct nfs41_exchange_id_res res;
7721 	struct nfs41_exchange_id_args args;
7722 };
7723 
7724 static void nfs4_exchange_id_release(void *data)
7725 {
7726 	struct nfs41_exchange_id_data *cdata =
7727 					(struct nfs41_exchange_id_data *)data;
7728 
7729 	nfs_put_client(cdata->args.client);
7730 	kfree(cdata->res.impl_id);
7731 	kfree(cdata->res.server_scope);
7732 	kfree(cdata->res.server_owner);
7733 	kfree(cdata);
7734 }
7735 
7736 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7737 	.rpc_release = nfs4_exchange_id_release,
7738 };
7739 
7740 /*
7741  * _nfs4_proc_exchange_id()
7742  *
7743  * Wrapper for EXCHANGE_ID operation.
7744  */
7745 static struct rpc_task *
7746 nfs4_run_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7747 			u32 sp4_how, struct rpc_xprt *xprt)
7748 {
7749 	struct rpc_message msg = {
7750 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7751 		.rpc_cred = cred,
7752 	};
7753 	struct rpc_task_setup task_setup_data = {
7754 		.rpc_client = clp->cl_rpcclient,
7755 		.callback_ops = &nfs4_exchange_id_call_ops,
7756 		.rpc_message = &msg,
7757 		.flags = RPC_TASK_TIMEOUT,
7758 	};
7759 	struct nfs41_exchange_id_data *calldata;
7760 	int status;
7761 
7762 	if (!refcount_inc_not_zero(&clp->cl_count))
7763 		return ERR_PTR(-EIO);
7764 
7765 	status = -ENOMEM;
7766 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7767 	if (!calldata)
7768 		goto out;
7769 
7770 	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7771 
7772 	status = nfs4_init_uniform_client_string(clp);
7773 	if (status)
7774 		goto out_calldata;
7775 
7776 	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7777 						GFP_NOFS);
7778 	status = -ENOMEM;
7779 	if (unlikely(calldata->res.server_owner == NULL))
7780 		goto out_calldata;
7781 
7782 	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7783 					GFP_NOFS);
7784 	if (unlikely(calldata->res.server_scope == NULL))
7785 		goto out_server_owner;
7786 
7787 	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7788 	if (unlikely(calldata->res.impl_id == NULL))
7789 		goto out_server_scope;
7790 
7791 	switch (sp4_how) {
7792 	case SP4_NONE:
7793 		calldata->args.state_protect.how = SP4_NONE;
7794 		break;
7795 
7796 	case SP4_MACH_CRED:
7797 		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7798 		break;
7799 
7800 	default:
7801 		/* unsupported! */
7802 		WARN_ON_ONCE(1);
7803 		status = -EINVAL;
7804 		goto out_impl_id;
7805 	}
7806 	if (xprt) {
7807 		task_setup_data.rpc_xprt = xprt;
7808 		task_setup_data.flags |= RPC_TASK_SOFTCONN;
7809 		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7810 				sizeof(calldata->args.verifier.data));
7811 	}
7812 	calldata->args.client = clp;
7813 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7814 	EXCHGID4_FLAG_BIND_PRINC_STATEID;
7815 #ifdef CONFIG_NFS_V4_1_MIGRATION
7816 	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
7817 #endif
7818 	msg.rpc_argp = &calldata->args;
7819 	msg.rpc_resp = &calldata->res;
7820 	task_setup_data.callback_data = calldata;
7821 
7822 	return rpc_run_task(&task_setup_data);
7823 
7824 out_impl_id:
7825 	kfree(calldata->res.impl_id);
7826 out_server_scope:
7827 	kfree(calldata->res.server_scope);
7828 out_server_owner:
7829 	kfree(calldata->res.server_owner);
7830 out_calldata:
7831 	kfree(calldata);
7832 out:
7833 	nfs_put_client(clp);
7834 	return ERR_PTR(status);
7835 }
7836 
7837 /*
7838  * _nfs4_proc_exchange_id()
7839  *
7840  * Wrapper for EXCHANGE_ID operation.
7841  */
7842 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7843 			u32 sp4_how)
7844 {
7845 	struct rpc_task *task;
7846 	struct nfs41_exchange_id_args *argp;
7847 	struct nfs41_exchange_id_res *resp;
7848 	int status;
7849 
7850 	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
7851 	if (IS_ERR(task))
7852 		return PTR_ERR(task);
7853 
7854 	argp = task->tk_msg.rpc_argp;
7855 	resp = task->tk_msg.rpc_resp;
7856 	status = task->tk_status;
7857 	if (status  != 0)
7858 		goto out;
7859 
7860 	status = nfs4_check_cl_exchange_flags(resp->flags);
7861 	if (status  != 0)
7862 		goto out;
7863 
7864 	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
7865 	if (status != 0)
7866 		goto out;
7867 
7868 	clp->cl_clientid = resp->clientid;
7869 	clp->cl_exchange_flags = resp->flags;
7870 	clp->cl_seqid = resp->seqid;
7871 	/* Client ID is not confirmed */
7872 	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
7873 		clear_bit(NFS4_SESSION_ESTABLISHED,
7874 			  &clp->cl_session->session_state);
7875 
7876 	if (clp->cl_serverscope != NULL &&
7877 	    !nfs41_same_server_scope(clp->cl_serverscope,
7878 				resp->server_scope)) {
7879 		dprintk("%s: server_scope mismatch detected\n",
7880 			__func__);
7881 		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7882 	}
7883 
7884 	swap(clp->cl_serverowner, resp->server_owner);
7885 	swap(clp->cl_serverscope, resp->server_scope);
7886 	swap(clp->cl_implid, resp->impl_id);
7887 
7888 	/* Save the EXCHANGE_ID verifier session trunk tests */
7889 	memcpy(clp->cl_confirm.data, argp->verifier.data,
7890 	       sizeof(clp->cl_confirm.data));
7891 out:
7892 	trace_nfs4_exchange_id(clp, status);
7893 	rpc_put_task(task);
7894 	return status;
7895 }
7896 
7897 /*
7898  * nfs4_proc_exchange_id()
7899  *
7900  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7901  *
7902  * Since the clientid has expired, all compounds using sessions
7903  * associated with the stale clientid will be returning
7904  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7905  * be in some phase of session reset.
7906  *
7907  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7908  */
7909 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7910 {
7911 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7912 	int status;
7913 
7914 	/* try SP4_MACH_CRED if krb5i/p	*/
7915 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
7916 	    authflavor == RPC_AUTH_GSS_KRB5P) {
7917 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7918 		if (!status)
7919 			return 0;
7920 	}
7921 
7922 	/* try SP4_NONE */
7923 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7924 }
7925 
7926 /**
7927  * nfs4_test_session_trunk
7928  *
7929  * This is an add_xprt_test() test function called from
7930  * rpc_clnt_setup_test_and_add_xprt.
7931  *
7932  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7933  * and is dereferrenced in nfs4_exchange_id_release
7934  *
7935  * Upon success, add the new transport to the rpc_clnt
7936  *
7937  * @clnt: struct rpc_clnt to get new transport
7938  * @xprt: the rpc_xprt to test
7939  * @data: call data for _nfs4_proc_exchange_id.
7940  */
7941 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7942 			    void *data)
7943 {
7944 	struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7945 	struct rpc_task *task;
7946 	int status;
7947 
7948 	u32 sp4_how;
7949 
7950 	dprintk("--> %s try %s\n", __func__,
7951 		xprt->address_strings[RPC_DISPLAY_ADDR]);
7952 
7953 	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7954 
7955 	/* Test connection for session trunking. Async exchange_id call */
7956 	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7957 	if (IS_ERR(task))
7958 		return PTR_ERR(task);
7959 
7960 	status = task->tk_status;
7961 	if (status == 0)
7962 		status = nfs4_detect_session_trunking(adata->clp,
7963 				task->tk_msg.rpc_resp, xprt);
7964 
7965 	rpc_put_task(task);
7966 	return status;
7967 }
7968 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7969 
7970 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7971 		struct rpc_cred *cred)
7972 {
7973 	struct rpc_message msg = {
7974 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7975 		.rpc_argp = clp,
7976 		.rpc_cred = cred,
7977 	};
7978 	int status;
7979 
7980 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7981 	trace_nfs4_destroy_clientid(clp, status);
7982 	if (status)
7983 		dprintk("NFS: Got error %d from the server %s on "
7984 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
7985 	return status;
7986 }
7987 
7988 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7989 		struct rpc_cred *cred)
7990 {
7991 	unsigned int loop;
7992 	int ret;
7993 
7994 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7995 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7996 		switch (ret) {
7997 		case -NFS4ERR_DELAY:
7998 		case -NFS4ERR_CLIENTID_BUSY:
7999 			ssleep(1);
8000 			break;
8001 		default:
8002 			return ret;
8003 		}
8004 	}
8005 	return 0;
8006 }
8007 
8008 int nfs4_destroy_clientid(struct nfs_client *clp)
8009 {
8010 	struct rpc_cred *cred;
8011 	int ret = 0;
8012 
8013 	if (clp->cl_mvops->minor_version < 1)
8014 		goto out;
8015 	if (clp->cl_exchange_flags == 0)
8016 		goto out;
8017 	if (clp->cl_preserve_clid)
8018 		goto out;
8019 	cred = nfs4_get_clid_cred(clp);
8020 	ret = nfs4_proc_destroy_clientid(clp, cred);
8021 	if (cred)
8022 		put_rpccred(cred);
8023 	switch (ret) {
8024 	case 0:
8025 	case -NFS4ERR_STALE_CLIENTID:
8026 		clp->cl_exchange_flags = 0;
8027 	}
8028 out:
8029 	return ret;
8030 }
8031 
8032 struct nfs4_get_lease_time_data {
8033 	struct nfs4_get_lease_time_args *args;
8034 	struct nfs4_get_lease_time_res *res;
8035 	struct nfs_client *clp;
8036 };
8037 
8038 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8039 					void *calldata)
8040 {
8041 	struct nfs4_get_lease_time_data *data =
8042 			(struct nfs4_get_lease_time_data *)calldata;
8043 
8044 	dprintk("--> %s\n", __func__);
8045 	/* just setup sequence, do not trigger session recovery
8046 	   since we're invoked within one */
8047 	nfs4_setup_sequence(data->clp,
8048 			&data->args->la_seq_args,
8049 			&data->res->lr_seq_res,
8050 			task);
8051 	dprintk("<-- %s\n", __func__);
8052 }
8053 
8054 /*
8055  * Called from nfs4_state_manager thread for session setup, so don't recover
8056  * from sequence operation or clientid errors.
8057  */
8058 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8059 {
8060 	struct nfs4_get_lease_time_data *data =
8061 			(struct nfs4_get_lease_time_data *)calldata;
8062 
8063 	dprintk("--> %s\n", __func__);
8064 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8065 		return;
8066 	switch (task->tk_status) {
8067 	case -NFS4ERR_DELAY:
8068 	case -NFS4ERR_GRACE:
8069 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8070 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
8071 		task->tk_status = 0;
8072 		/* fall through */
8073 	case -NFS4ERR_RETRY_UNCACHED_REP:
8074 		rpc_restart_call_prepare(task);
8075 		return;
8076 	}
8077 	dprintk("<-- %s\n", __func__);
8078 }
8079 
8080 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8081 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
8082 	.rpc_call_done = nfs4_get_lease_time_done,
8083 };
8084 
8085 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8086 {
8087 	struct rpc_task *task;
8088 	struct nfs4_get_lease_time_args args;
8089 	struct nfs4_get_lease_time_res res = {
8090 		.lr_fsinfo = fsinfo,
8091 	};
8092 	struct nfs4_get_lease_time_data data = {
8093 		.args = &args,
8094 		.res = &res,
8095 		.clp = clp,
8096 	};
8097 	struct rpc_message msg = {
8098 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8099 		.rpc_argp = &args,
8100 		.rpc_resp = &res,
8101 	};
8102 	struct rpc_task_setup task_setup = {
8103 		.rpc_client = clp->cl_rpcclient,
8104 		.rpc_message = &msg,
8105 		.callback_ops = &nfs4_get_lease_time_ops,
8106 		.callback_data = &data,
8107 		.flags = RPC_TASK_TIMEOUT,
8108 	};
8109 	int status;
8110 
8111 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8112 	task = rpc_run_task(&task_setup);
8113 
8114 	if (IS_ERR(task))
8115 		return PTR_ERR(task);
8116 
8117 	status = task->tk_status;
8118 	rpc_put_task(task);
8119 	return status;
8120 }
8121 
8122 /*
8123  * Initialize the values to be used by the client in CREATE_SESSION
8124  * If nfs4_init_session set the fore channel request and response sizes,
8125  * use them.
8126  *
8127  * Set the back channel max_resp_sz_cached to zero to force the client to
8128  * always set csa_cachethis to FALSE because the current implementation
8129  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8130  */
8131 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8132 				    struct rpc_clnt *clnt)
8133 {
8134 	unsigned int max_rqst_sz, max_resp_sz;
8135 	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8136 
8137 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8138 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8139 
8140 	/* Fore channel attributes */
8141 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
8142 	args->fc_attrs.max_resp_sz = max_resp_sz;
8143 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
8144 	args->fc_attrs.max_reqs = max_session_slots;
8145 
8146 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8147 		"max_ops=%u max_reqs=%u\n",
8148 		__func__,
8149 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8150 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8151 
8152 	/* Back channel attributes */
8153 	args->bc_attrs.max_rqst_sz = max_bc_payload;
8154 	args->bc_attrs.max_resp_sz = max_bc_payload;
8155 	args->bc_attrs.max_resp_sz_cached = 0;
8156 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8157 	args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
8158 
8159 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8160 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8161 		__func__,
8162 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8163 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8164 		args->bc_attrs.max_reqs);
8165 }
8166 
8167 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8168 		struct nfs41_create_session_res *res)
8169 {
8170 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
8171 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8172 
8173 	if (rcvd->max_resp_sz > sent->max_resp_sz)
8174 		return -EINVAL;
8175 	/*
8176 	 * Our requested max_ops is the minimum we need; we're not
8177 	 * prepared to break up compounds into smaller pieces than that.
8178 	 * So, no point even trying to continue if the server won't
8179 	 * cooperate:
8180 	 */
8181 	if (rcvd->max_ops < sent->max_ops)
8182 		return -EINVAL;
8183 	if (rcvd->max_reqs == 0)
8184 		return -EINVAL;
8185 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8186 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8187 	return 0;
8188 }
8189 
8190 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8191 		struct nfs41_create_session_res *res)
8192 {
8193 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
8194 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8195 
8196 	if (!(res->flags & SESSION4_BACK_CHAN))
8197 		goto out;
8198 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8199 		return -EINVAL;
8200 	if (rcvd->max_resp_sz < sent->max_resp_sz)
8201 		return -EINVAL;
8202 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8203 		return -EINVAL;
8204 	if (rcvd->max_ops > sent->max_ops)
8205 		return -EINVAL;
8206 	if (rcvd->max_reqs > sent->max_reqs)
8207 		return -EINVAL;
8208 out:
8209 	return 0;
8210 }
8211 
8212 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8213 				     struct nfs41_create_session_res *res)
8214 {
8215 	int ret;
8216 
8217 	ret = nfs4_verify_fore_channel_attrs(args, res);
8218 	if (ret)
8219 		return ret;
8220 	return nfs4_verify_back_channel_attrs(args, res);
8221 }
8222 
8223 static void nfs4_update_session(struct nfs4_session *session,
8224 		struct nfs41_create_session_res *res)
8225 {
8226 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8227 	/* Mark client id and session as being confirmed */
8228 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8229 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8230 	session->flags = res->flags;
8231 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8232 	if (res->flags & SESSION4_BACK_CHAN)
8233 		memcpy(&session->bc_attrs, &res->bc_attrs,
8234 				sizeof(session->bc_attrs));
8235 }
8236 
8237 static int _nfs4_proc_create_session(struct nfs_client *clp,
8238 		struct rpc_cred *cred)
8239 {
8240 	struct nfs4_session *session = clp->cl_session;
8241 	struct nfs41_create_session_args args = {
8242 		.client = clp,
8243 		.clientid = clp->cl_clientid,
8244 		.seqid = clp->cl_seqid,
8245 		.cb_program = NFS4_CALLBACK,
8246 	};
8247 	struct nfs41_create_session_res res;
8248 
8249 	struct rpc_message msg = {
8250 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8251 		.rpc_argp = &args,
8252 		.rpc_resp = &res,
8253 		.rpc_cred = cred,
8254 	};
8255 	int status;
8256 
8257 	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8258 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8259 
8260 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8261 	trace_nfs4_create_session(clp, status);
8262 
8263 	switch (status) {
8264 	case -NFS4ERR_STALE_CLIENTID:
8265 	case -NFS4ERR_DELAY:
8266 	case -ETIMEDOUT:
8267 	case -EACCES:
8268 	case -EAGAIN:
8269 		goto out;
8270 	};
8271 
8272 	clp->cl_seqid++;
8273 	if (!status) {
8274 		/* Verify the session's negotiated channel_attrs values */
8275 		status = nfs4_verify_channel_attrs(&args, &res);
8276 		/* Increment the clientid slot sequence id */
8277 		if (status)
8278 			goto out;
8279 		nfs4_update_session(session, &res);
8280 	}
8281 out:
8282 	return status;
8283 }
8284 
8285 /*
8286  * Issues a CREATE_SESSION operation to the server.
8287  * It is the responsibility of the caller to verify the session is
8288  * expired before calling this routine.
8289  */
8290 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8291 {
8292 	int status;
8293 	unsigned *ptr;
8294 	struct nfs4_session *session = clp->cl_session;
8295 
8296 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8297 
8298 	status = _nfs4_proc_create_session(clp, cred);
8299 	if (status)
8300 		goto out;
8301 
8302 	/* Init or reset the session slot tables */
8303 	status = nfs4_setup_session_slot_tables(session);
8304 	dprintk("slot table setup returned %d\n", status);
8305 	if (status)
8306 		goto out;
8307 
8308 	ptr = (unsigned *)&session->sess_id.data[0];
8309 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8310 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8311 out:
8312 	dprintk("<-- %s\n", __func__);
8313 	return status;
8314 }
8315 
8316 /*
8317  * Issue the over-the-wire RPC DESTROY_SESSION.
8318  * The caller must serialize access to this routine.
8319  */
8320 int nfs4_proc_destroy_session(struct nfs4_session *session,
8321 		struct rpc_cred *cred)
8322 {
8323 	struct rpc_message msg = {
8324 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8325 		.rpc_argp = session,
8326 		.rpc_cred = cred,
8327 	};
8328 	int status = 0;
8329 
8330 	dprintk("--> nfs4_proc_destroy_session\n");
8331 
8332 	/* session is still being setup */
8333 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8334 		return 0;
8335 
8336 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8337 	trace_nfs4_destroy_session(session->clp, status);
8338 
8339 	if (status)
8340 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8341 			"Session has been destroyed regardless...\n", status);
8342 
8343 	dprintk("<-- nfs4_proc_destroy_session\n");
8344 	return status;
8345 }
8346 
8347 /*
8348  * Renew the cl_session lease.
8349  */
8350 struct nfs4_sequence_data {
8351 	struct nfs_client *clp;
8352 	struct nfs4_sequence_args args;
8353 	struct nfs4_sequence_res res;
8354 };
8355 
8356 static void nfs41_sequence_release(void *data)
8357 {
8358 	struct nfs4_sequence_data *calldata = data;
8359 	struct nfs_client *clp = calldata->clp;
8360 
8361 	if (refcount_read(&clp->cl_count) > 1)
8362 		nfs4_schedule_state_renewal(clp);
8363 	nfs_put_client(clp);
8364 	kfree(calldata);
8365 }
8366 
8367 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8368 {
8369 	switch(task->tk_status) {
8370 	case -NFS4ERR_DELAY:
8371 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
8372 		return -EAGAIN;
8373 	default:
8374 		nfs4_schedule_lease_recovery(clp);
8375 	}
8376 	return 0;
8377 }
8378 
8379 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8380 {
8381 	struct nfs4_sequence_data *calldata = data;
8382 	struct nfs_client *clp = calldata->clp;
8383 
8384 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8385 		return;
8386 
8387 	trace_nfs4_sequence(clp, task->tk_status);
8388 	if (task->tk_status < 0) {
8389 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
8390 		if (refcount_read(&clp->cl_count) == 1)
8391 			goto out;
8392 
8393 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8394 			rpc_restart_call_prepare(task);
8395 			return;
8396 		}
8397 	}
8398 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8399 out:
8400 	dprintk("<-- %s\n", __func__);
8401 }
8402 
8403 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8404 {
8405 	struct nfs4_sequence_data *calldata = data;
8406 	struct nfs_client *clp = calldata->clp;
8407 	struct nfs4_sequence_args *args;
8408 	struct nfs4_sequence_res *res;
8409 
8410 	args = task->tk_msg.rpc_argp;
8411 	res = task->tk_msg.rpc_resp;
8412 
8413 	nfs4_setup_sequence(clp, args, res, task);
8414 }
8415 
8416 static const struct rpc_call_ops nfs41_sequence_ops = {
8417 	.rpc_call_done = nfs41_sequence_call_done,
8418 	.rpc_call_prepare = nfs41_sequence_prepare,
8419 	.rpc_release = nfs41_sequence_release,
8420 };
8421 
8422 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8423 		struct rpc_cred *cred,
8424 		struct nfs4_slot *slot,
8425 		bool is_privileged)
8426 {
8427 	struct nfs4_sequence_data *calldata;
8428 	struct rpc_message msg = {
8429 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8430 		.rpc_cred = cred,
8431 	};
8432 	struct rpc_task_setup task_setup_data = {
8433 		.rpc_client = clp->cl_rpcclient,
8434 		.rpc_message = &msg,
8435 		.callback_ops = &nfs41_sequence_ops,
8436 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8437 	};
8438 	struct rpc_task *ret;
8439 
8440 	ret = ERR_PTR(-EIO);
8441 	if (!refcount_inc_not_zero(&clp->cl_count))
8442 		goto out_err;
8443 
8444 	ret = ERR_PTR(-ENOMEM);
8445 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8446 	if (calldata == NULL)
8447 		goto out_put_clp;
8448 	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8449 	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8450 	msg.rpc_argp = &calldata->args;
8451 	msg.rpc_resp = &calldata->res;
8452 	calldata->clp = clp;
8453 	task_setup_data.callback_data = calldata;
8454 
8455 	ret = rpc_run_task(&task_setup_data);
8456 	if (IS_ERR(ret))
8457 		goto out_err;
8458 	return ret;
8459 out_put_clp:
8460 	nfs_put_client(clp);
8461 out_err:
8462 	nfs41_release_slot(slot);
8463 	return ret;
8464 }
8465 
8466 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8467 {
8468 	struct rpc_task *task;
8469 	int ret = 0;
8470 
8471 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8472 		return -EAGAIN;
8473 	task = _nfs41_proc_sequence(clp, cred, NULL, false);
8474 	if (IS_ERR(task))
8475 		ret = PTR_ERR(task);
8476 	else
8477 		rpc_put_task_async(task);
8478 	dprintk("<-- %s status=%d\n", __func__, ret);
8479 	return ret;
8480 }
8481 
8482 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8483 {
8484 	struct rpc_task *task;
8485 	int ret;
8486 
8487 	task = _nfs41_proc_sequence(clp, cred, NULL, true);
8488 	if (IS_ERR(task)) {
8489 		ret = PTR_ERR(task);
8490 		goto out;
8491 	}
8492 	ret = rpc_wait_for_completion_task(task);
8493 	if (!ret)
8494 		ret = task->tk_status;
8495 	rpc_put_task(task);
8496 out:
8497 	dprintk("<-- %s status=%d\n", __func__, ret);
8498 	return ret;
8499 }
8500 
8501 struct nfs4_reclaim_complete_data {
8502 	struct nfs_client *clp;
8503 	struct nfs41_reclaim_complete_args arg;
8504 	struct nfs41_reclaim_complete_res res;
8505 };
8506 
8507 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8508 {
8509 	struct nfs4_reclaim_complete_data *calldata = data;
8510 
8511 	nfs4_setup_sequence(calldata->clp,
8512 			&calldata->arg.seq_args,
8513 			&calldata->res.seq_res,
8514 			task);
8515 }
8516 
8517 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8518 {
8519 	switch(task->tk_status) {
8520 	case 0:
8521 		wake_up_all(&clp->cl_lock_waitq);
8522 		/* Fallthrough */
8523 	case -NFS4ERR_COMPLETE_ALREADY:
8524 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
8525 		break;
8526 	case -NFS4ERR_DELAY:
8527 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
8528 		/* fall through */
8529 	case -NFS4ERR_RETRY_UNCACHED_REP:
8530 		return -EAGAIN;
8531 	case -NFS4ERR_BADSESSION:
8532 	case -NFS4ERR_DEADSESSION:
8533 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8534 		nfs4_schedule_session_recovery(clp->cl_session,
8535 				task->tk_status);
8536 		break;
8537 	default:
8538 		nfs4_schedule_lease_recovery(clp);
8539 	}
8540 	return 0;
8541 }
8542 
8543 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8544 {
8545 	struct nfs4_reclaim_complete_data *calldata = data;
8546 	struct nfs_client *clp = calldata->clp;
8547 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
8548 
8549 	dprintk("--> %s\n", __func__);
8550 	if (!nfs41_sequence_done(task, res))
8551 		return;
8552 
8553 	trace_nfs4_reclaim_complete(clp, task->tk_status);
8554 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8555 		rpc_restart_call_prepare(task);
8556 		return;
8557 	}
8558 	dprintk("<-- %s\n", __func__);
8559 }
8560 
8561 static void nfs4_free_reclaim_complete_data(void *data)
8562 {
8563 	struct nfs4_reclaim_complete_data *calldata = data;
8564 
8565 	kfree(calldata);
8566 }
8567 
8568 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8569 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
8570 	.rpc_call_done = nfs4_reclaim_complete_done,
8571 	.rpc_release = nfs4_free_reclaim_complete_data,
8572 };
8573 
8574 /*
8575  * Issue a global reclaim complete.
8576  */
8577 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8578 		struct rpc_cred *cred)
8579 {
8580 	struct nfs4_reclaim_complete_data *calldata;
8581 	struct rpc_task *task;
8582 	struct rpc_message msg = {
8583 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8584 		.rpc_cred = cred,
8585 	};
8586 	struct rpc_task_setup task_setup_data = {
8587 		.rpc_client = clp->cl_rpcclient,
8588 		.rpc_message = &msg,
8589 		.callback_ops = &nfs4_reclaim_complete_call_ops,
8590 		.flags = RPC_TASK_ASYNC,
8591 	};
8592 	int status = -ENOMEM;
8593 
8594 	dprintk("--> %s\n", __func__);
8595 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8596 	if (calldata == NULL)
8597 		goto out;
8598 	calldata->clp = clp;
8599 	calldata->arg.one_fs = 0;
8600 
8601 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8602 	msg.rpc_argp = &calldata->arg;
8603 	msg.rpc_resp = &calldata->res;
8604 	task_setup_data.callback_data = calldata;
8605 	task = rpc_run_task(&task_setup_data);
8606 	if (IS_ERR(task)) {
8607 		status = PTR_ERR(task);
8608 		goto out;
8609 	}
8610 	status = rpc_wait_for_completion_task(task);
8611 	if (status == 0)
8612 		status = task->tk_status;
8613 	rpc_put_task(task);
8614 out:
8615 	dprintk("<-- %s status=%d\n", __func__, status);
8616 	return status;
8617 }
8618 
8619 static void
8620 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8621 {
8622 	struct nfs4_layoutget *lgp = calldata;
8623 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8624 
8625 	dprintk("--> %s\n", __func__);
8626 	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8627 				&lgp->res.seq_res, task);
8628 	dprintk("<-- %s\n", __func__);
8629 }
8630 
8631 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8632 {
8633 	struct nfs4_layoutget *lgp = calldata;
8634 
8635 	dprintk("--> %s\n", __func__);
8636 	nfs41_sequence_process(task, &lgp->res.seq_res);
8637 	dprintk("<-- %s\n", __func__);
8638 }
8639 
8640 static int
8641 nfs4_layoutget_handle_exception(struct rpc_task *task,
8642 		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8643 {
8644 	struct inode *inode = lgp->args.inode;
8645 	struct nfs_server *server = NFS_SERVER(inode);
8646 	struct pnfs_layout_hdr *lo;
8647 	int nfs4err = task->tk_status;
8648 	int err, status = 0;
8649 	LIST_HEAD(head);
8650 
8651 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8652 
8653 	switch (nfs4err) {
8654 	case 0:
8655 		goto out;
8656 
8657 	/*
8658 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8659 	 * on the file. set tk_status to -ENODATA to tell upper layer to
8660 	 * retry go inband.
8661 	 */
8662 	case -NFS4ERR_LAYOUTUNAVAILABLE:
8663 		status = -ENODATA;
8664 		goto out;
8665 	/*
8666 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8667 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8668 	 */
8669 	case -NFS4ERR_BADLAYOUT:
8670 		status = -EOVERFLOW;
8671 		goto out;
8672 	/*
8673 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8674 	 * (or clients) writing to the same RAID stripe except when
8675 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
8676 	 *
8677 	 * Treat it like we would RECALLCONFLICT -- we retry for a little
8678 	 * while, and then eventually give up.
8679 	 */
8680 	case -NFS4ERR_LAYOUTTRYLATER:
8681 		if (lgp->args.minlength == 0) {
8682 			status = -EOVERFLOW;
8683 			goto out;
8684 		}
8685 		status = -EBUSY;
8686 		break;
8687 	case -NFS4ERR_RECALLCONFLICT:
8688 		status = -ERECALLCONFLICT;
8689 		break;
8690 	case -NFS4ERR_DELEG_REVOKED:
8691 	case -NFS4ERR_ADMIN_REVOKED:
8692 	case -NFS4ERR_EXPIRED:
8693 	case -NFS4ERR_BAD_STATEID:
8694 		exception->timeout = 0;
8695 		spin_lock(&inode->i_lock);
8696 		lo = NFS_I(inode)->layout;
8697 		/* If the open stateid was bad, then recover it. */
8698 		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8699 		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8700 			spin_unlock(&inode->i_lock);
8701 			exception->state = lgp->args.ctx->state;
8702 			exception->stateid = &lgp->args.stateid;
8703 			break;
8704 		}
8705 
8706 		/*
8707 		 * Mark the bad layout state as invalid, then retry
8708 		 */
8709 		pnfs_mark_layout_stateid_invalid(lo, &head);
8710 		spin_unlock(&inode->i_lock);
8711 		nfs_commit_inode(inode, 0);
8712 		pnfs_free_lseg_list(&head);
8713 		status = -EAGAIN;
8714 		goto out;
8715 	}
8716 
8717 	nfs4_sequence_free_slot(&lgp->res.seq_res);
8718 	err = nfs4_handle_exception(server, nfs4err, exception);
8719 	if (!status) {
8720 		if (exception->retry)
8721 			status = -EAGAIN;
8722 		else
8723 			status = err;
8724 	}
8725 out:
8726 	dprintk("<-- %s\n", __func__);
8727 	return status;
8728 }
8729 
8730 size_t max_response_pages(struct nfs_server *server)
8731 {
8732 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8733 	return nfs_page_array_len(0, max_resp_sz);
8734 }
8735 
8736 static void nfs4_layoutget_release(void *calldata)
8737 {
8738 	struct nfs4_layoutget *lgp = calldata;
8739 
8740 	dprintk("--> %s\n", __func__);
8741 	nfs4_sequence_free_slot(&lgp->res.seq_res);
8742 	pnfs_layoutget_free(lgp);
8743 	dprintk("<-- %s\n", __func__);
8744 }
8745 
8746 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8747 	.rpc_call_prepare = nfs4_layoutget_prepare,
8748 	.rpc_call_done = nfs4_layoutget_done,
8749 	.rpc_release = nfs4_layoutget_release,
8750 };
8751 
8752 struct pnfs_layout_segment *
8753 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8754 {
8755 	struct inode *inode = lgp->args.inode;
8756 	struct nfs_server *server = NFS_SERVER(inode);
8757 	struct rpc_task *task;
8758 	struct rpc_message msg = {
8759 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8760 		.rpc_argp = &lgp->args,
8761 		.rpc_resp = &lgp->res,
8762 		.rpc_cred = lgp->cred,
8763 	};
8764 	struct rpc_task_setup task_setup_data = {
8765 		.rpc_client = server->client,
8766 		.rpc_message = &msg,
8767 		.callback_ops = &nfs4_layoutget_call_ops,
8768 		.callback_data = lgp,
8769 		.flags = RPC_TASK_ASYNC,
8770 	};
8771 	struct pnfs_layout_segment *lseg = NULL;
8772 	struct nfs4_exception exception = {
8773 		.inode = inode,
8774 		.timeout = *timeout,
8775 	};
8776 	int status = 0;
8777 
8778 	dprintk("--> %s\n", __func__);
8779 
8780 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8781 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
8782 
8783 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8784 
8785 	task = rpc_run_task(&task_setup_data);
8786 	if (IS_ERR(task))
8787 		return ERR_CAST(task);
8788 	status = rpc_wait_for_completion_task(task);
8789 	if (status == 0) {
8790 		status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8791 		*timeout = exception.timeout;
8792 	}
8793 
8794 	trace_nfs4_layoutget(lgp->args.ctx,
8795 			&lgp->args.range,
8796 			&lgp->res.range,
8797 			&lgp->res.stateid,
8798 			status);
8799 
8800 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8801 	if (status == 0 && lgp->res.layoutp->len)
8802 		lseg = pnfs_layout_process(lgp);
8803 	rpc_put_task(task);
8804 	dprintk("<-- %s status=%d\n", __func__, status);
8805 	if (status)
8806 		return ERR_PTR(status);
8807 	return lseg;
8808 }
8809 
8810 static void
8811 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8812 {
8813 	struct nfs4_layoutreturn *lrp = calldata;
8814 
8815 	dprintk("--> %s\n", __func__);
8816 	nfs4_setup_sequence(lrp->clp,
8817 			&lrp->args.seq_args,
8818 			&lrp->res.seq_res,
8819 			task);
8820 }
8821 
8822 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8823 {
8824 	struct nfs4_layoutreturn *lrp = calldata;
8825 	struct nfs_server *server;
8826 
8827 	dprintk("--> %s\n", __func__);
8828 
8829 	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8830 		return;
8831 
8832 	server = NFS_SERVER(lrp->args.inode);
8833 	switch (task->tk_status) {
8834 	case -NFS4ERR_OLD_STATEID:
8835 		if (nfs4_refresh_layout_stateid(&lrp->args.stateid,
8836 					lrp->args.inode))
8837 			goto out_restart;
8838 		/* Fallthrough */
8839 	default:
8840 		task->tk_status = 0;
8841 		/* Fallthrough */
8842 	case 0:
8843 		break;
8844 	case -NFS4ERR_DELAY:
8845 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8846 			break;
8847 		goto out_restart;
8848 	}
8849 	dprintk("<-- %s\n", __func__);
8850 	return;
8851 out_restart:
8852 	task->tk_status = 0;
8853 	nfs4_sequence_free_slot(&lrp->res.seq_res);
8854 	rpc_restart_call_prepare(task);
8855 }
8856 
8857 static void nfs4_layoutreturn_release(void *calldata)
8858 {
8859 	struct nfs4_layoutreturn *lrp = calldata;
8860 	struct pnfs_layout_hdr *lo = lrp->args.layout;
8861 
8862 	dprintk("--> %s\n", __func__);
8863 	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8864 			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8865 	nfs4_sequence_free_slot(&lrp->res.seq_res);
8866 	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8867 		lrp->ld_private.ops->free(&lrp->ld_private);
8868 	pnfs_put_layout_hdr(lrp->args.layout);
8869 	nfs_iput_and_deactive(lrp->inode);
8870 	kfree(calldata);
8871 	dprintk("<-- %s\n", __func__);
8872 }
8873 
8874 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8875 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
8876 	.rpc_call_done = nfs4_layoutreturn_done,
8877 	.rpc_release = nfs4_layoutreturn_release,
8878 };
8879 
8880 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8881 {
8882 	struct rpc_task *task;
8883 	struct rpc_message msg = {
8884 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8885 		.rpc_argp = &lrp->args,
8886 		.rpc_resp = &lrp->res,
8887 		.rpc_cred = lrp->cred,
8888 	};
8889 	struct rpc_task_setup task_setup_data = {
8890 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
8891 		.rpc_message = &msg,
8892 		.callback_ops = &nfs4_layoutreturn_call_ops,
8893 		.callback_data = lrp,
8894 	};
8895 	int status = 0;
8896 
8897 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8898 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8899 			&task_setup_data.rpc_client, &msg);
8900 
8901 	dprintk("--> %s\n", __func__);
8902 	if (!sync) {
8903 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8904 		if (!lrp->inode) {
8905 			nfs4_layoutreturn_release(lrp);
8906 			return -EAGAIN;
8907 		}
8908 		task_setup_data.flags |= RPC_TASK_ASYNC;
8909 	}
8910 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
8911 	task = rpc_run_task(&task_setup_data);
8912 	if (IS_ERR(task))
8913 		return PTR_ERR(task);
8914 	if (sync)
8915 		status = task->tk_status;
8916 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8917 	dprintk("<-- %s status=%d\n", __func__, status);
8918 	rpc_put_task(task);
8919 	return status;
8920 }
8921 
8922 static int
8923 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8924 		struct pnfs_device *pdev,
8925 		struct rpc_cred *cred)
8926 {
8927 	struct nfs4_getdeviceinfo_args args = {
8928 		.pdev = pdev,
8929 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
8930 			NOTIFY_DEVICEID4_DELETE,
8931 	};
8932 	struct nfs4_getdeviceinfo_res res = {
8933 		.pdev = pdev,
8934 	};
8935 	struct rpc_message msg = {
8936 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8937 		.rpc_argp = &args,
8938 		.rpc_resp = &res,
8939 		.rpc_cred = cred,
8940 	};
8941 	int status;
8942 
8943 	dprintk("--> %s\n", __func__);
8944 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8945 	if (res.notification & ~args.notify_types)
8946 		dprintk("%s: unsupported notification\n", __func__);
8947 	if (res.notification != args.notify_types)
8948 		pdev->nocache = 1;
8949 
8950 	dprintk("<-- %s status=%d\n", __func__, status);
8951 
8952 	return status;
8953 }
8954 
8955 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8956 		struct pnfs_device *pdev,
8957 		struct rpc_cred *cred)
8958 {
8959 	struct nfs4_exception exception = { };
8960 	int err;
8961 
8962 	do {
8963 		err = nfs4_handle_exception(server,
8964 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
8965 					&exception);
8966 	} while (exception.retry);
8967 	return err;
8968 }
8969 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8970 
8971 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8972 {
8973 	struct nfs4_layoutcommit_data *data = calldata;
8974 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8975 
8976 	nfs4_setup_sequence(server->nfs_client,
8977 			&data->args.seq_args,
8978 			&data->res.seq_res,
8979 			task);
8980 }
8981 
8982 static void
8983 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8984 {
8985 	struct nfs4_layoutcommit_data *data = calldata;
8986 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8987 
8988 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8989 		return;
8990 
8991 	switch (task->tk_status) { /* Just ignore these failures */
8992 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8993 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8994 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8995 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8996 		task->tk_status = 0;
8997 	case 0:
8998 		break;
8999 	default:
9000 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9001 			rpc_restart_call_prepare(task);
9002 			return;
9003 		}
9004 	}
9005 }
9006 
9007 static void nfs4_layoutcommit_release(void *calldata)
9008 {
9009 	struct nfs4_layoutcommit_data *data = calldata;
9010 
9011 	pnfs_cleanup_layoutcommit(data);
9012 	nfs_post_op_update_inode_force_wcc(data->args.inode,
9013 					   data->res.fattr);
9014 	put_rpccred(data->cred);
9015 	nfs_iput_and_deactive(data->inode);
9016 	kfree(data);
9017 }
9018 
9019 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9020 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
9021 	.rpc_call_done = nfs4_layoutcommit_done,
9022 	.rpc_release = nfs4_layoutcommit_release,
9023 };
9024 
9025 int
9026 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9027 {
9028 	struct rpc_message msg = {
9029 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9030 		.rpc_argp = &data->args,
9031 		.rpc_resp = &data->res,
9032 		.rpc_cred = data->cred,
9033 	};
9034 	struct rpc_task_setup task_setup_data = {
9035 		.task = &data->task,
9036 		.rpc_client = NFS_CLIENT(data->args.inode),
9037 		.rpc_message = &msg,
9038 		.callback_ops = &nfs4_layoutcommit_ops,
9039 		.callback_data = data,
9040 	};
9041 	struct rpc_task *task;
9042 	int status = 0;
9043 
9044 	dprintk("NFS: initiating layoutcommit call. sync %d "
9045 		"lbw: %llu inode %lu\n", sync,
9046 		data->args.lastbytewritten,
9047 		data->args.inode->i_ino);
9048 
9049 	if (!sync) {
9050 		data->inode = nfs_igrab_and_active(data->args.inode);
9051 		if (data->inode == NULL) {
9052 			nfs4_layoutcommit_release(data);
9053 			return -EAGAIN;
9054 		}
9055 		task_setup_data.flags = RPC_TASK_ASYNC;
9056 	}
9057 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9058 	task = rpc_run_task(&task_setup_data);
9059 	if (IS_ERR(task))
9060 		return PTR_ERR(task);
9061 	if (sync)
9062 		status = task->tk_status;
9063 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9064 	dprintk("%s: status %d\n", __func__, status);
9065 	rpc_put_task(task);
9066 	return status;
9067 }
9068 
9069 /**
9070  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9071  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9072  */
9073 static int
9074 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9075 		    struct nfs_fsinfo *info,
9076 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9077 {
9078 	struct nfs41_secinfo_no_name_args args = {
9079 		.style = SECINFO_STYLE_CURRENT_FH,
9080 	};
9081 	struct nfs4_secinfo_res res = {
9082 		.flavors = flavors,
9083 	};
9084 	struct rpc_message msg = {
9085 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9086 		.rpc_argp = &args,
9087 		.rpc_resp = &res,
9088 	};
9089 	struct rpc_clnt *clnt = server->client;
9090 	struct rpc_cred *cred = NULL;
9091 	int status;
9092 
9093 	if (use_integrity) {
9094 		clnt = server->nfs_client->cl_rpcclient;
9095 		cred = nfs4_get_clid_cred(server->nfs_client);
9096 		msg.rpc_cred = cred;
9097 	}
9098 
9099 	dprintk("--> %s\n", __func__);
9100 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9101 				&res.seq_res, 0);
9102 	dprintk("<-- %s status=%d\n", __func__, status);
9103 
9104 	if (cred)
9105 		put_rpccred(cred);
9106 
9107 	return status;
9108 }
9109 
9110 static int
9111 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9112 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9113 {
9114 	struct nfs4_exception exception = { };
9115 	int err;
9116 	do {
9117 		/* first try using integrity protection */
9118 		err = -NFS4ERR_WRONGSEC;
9119 
9120 		/* try to use integrity protection with machine cred */
9121 		if (_nfs4_is_integrity_protected(server->nfs_client))
9122 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9123 							  flavors, true);
9124 
9125 		/*
9126 		 * if unable to use integrity protection, or SECINFO with
9127 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
9128 		 * disallowed by spec, but exists in deployed servers) use
9129 		 * the current filesystem's rpc_client and the user cred.
9130 		 */
9131 		if (err == -NFS4ERR_WRONGSEC)
9132 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9133 							  flavors, false);
9134 
9135 		switch (err) {
9136 		case 0:
9137 		case -NFS4ERR_WRONGSEC:
9138 		case -ENOTSUPP:
9139 			goto out;
9140 		default:
9141 			err = nfs4_handle_exception(server, err, &exception);
9142 		}
9143 	} while (exception.retry);
9144 out:
9145 	return err;
9146 }
9147 
9148 static int
9149 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9150 		    struct nfs_fsinfo *info)
9151 {
9152 	int err;
9153 	struct page *page;
9154 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9155 	struct nfs4_secinfo_flavors *flavors;
9156 	struct nfs4_secinfo4 *secinfo;
9157 	int i;
9158 
9159 	page = alloc_page(GFP_KERNEL);
9160 	if (!page) {
9161 		err = -ENOMEM;
9162 		goto out;
9163 	}
9164 
9165 	flavors = page_address(page);
9166 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9167 
9168 	/*
9169 	 * Fall back on "guess and check" method if
9170 	 * the server doesn't support SECINFO_NO_NAME
9171 	 */
9172 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9173 		err = nfs4_find_root_sec(server, fhandle, info);
9174 		goto out_freepage;
9175 	}
9176 	if (err)
9177 		goto out_freepage;
9178 
9179 	for (i = 0; i < flavors->num_flavors; i++) {
9180 		secinfo = &flavors->flavors[i];
9181 
9182 		switch (secinfo->flavor) {
9183 		case RPC_AUTH_NULL:
9184 		case RPC_AUTH_UNIX:
9185 		case RPC_AUTH_GSS:
9186 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9187 					&secinfo->flavor_info);
9188 			break;
9189 		default:
9190 			flavor = RPC_AUTH_MAXFLAVOR;
9191 			break;
9192 		}
9193 
9194 		if (!nfs_auth_info_match(&server->auth_info, flavor))
9195 			flavor = RPC_AUTH_MAXFLAVOR;
9196 
9197 		if (flavor != RPC_AUTH_MAXFLAVOR) {
9198 			err = nfs4_lookup_root_sec(server, fhandle,
9199 						   info, flavor);
9200 			if (!err)
9201 				break;
9202 		}
9203 	}
9204 
9205 	if (flavor == RPC_AUTH_MAXFLAVOR)
9206 		err = -EPERM;
9207 
9208 out_freepage:
9209 	put_page(page);
9210 	if (err == -EACCES)
9211 		return -EPERM;
9212 out:
9213 	return err;
9214 }
9215 
9216 static int _nfs41_test_stateid(struct nfs_server *server,
9217 		nfs4_stateid *stateid,
9218 		struct rpc_cred *cred)
9219 {
9220 	int status;
9221 	struct nfs41_test_stateid_args args = {
9222 		.stateid = stateid,
9223 	};
9224 	struct nfs41_test_stateid_res res;
9225 	struct rpc_message msg = {
9226 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9227 		.rpc_argp = &args,
9228 		.rpc_resp = &res,
9229 		.rpc_cred = cred,
9230 	};
9231 	struct rpc_clnt *rpc_client = server->client;
9232 
9233 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9234 		&rpc_client, &msg);
9235 
9236 	dprintk("NFS call  test_stateid %p\n", stateid);
9237 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9238 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9239 			&args.seq_args, &res.seq_res);
9240 	if (status != NFS_OK) {
9241 		dprintk("NFS reply test_stateid: failed, %d\n", status);
9242 		return status;
9243 	}
9244 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9245 	return -res.status;
9246 }
9247 
9248 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9249 		int err, struct nfs4_exception *exception)
9250 {
9251 	exception->retry = 0;
9252 	switch(err) {
9253 	case -NFS4ERR_DELAY:
9254 	case -NFS4ERR_RETRY_UNCACHED_REP:
9255 		nfs4_handle_exception(server, err, exception);
9256 		break;
9257 	case -NFS4ERR_BADSESSION:
9258 	case -NFS4ERR_BADSLOT:
9259 	case -NFS4ERR_BAD_HIGH_SLOT:
9260 	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9261 	case -NFS4ERR_DEADSESSION:
9262 		nfs4_do_handle_exception(server, err, exception);
9263 	}
9264 }
9265 
9266 /**
9267  * nfs41_test_stateid - perform a TEST_STATEID operation
9268  *
9269  * @server: server / transport on which to perform the operation
9270  * @stateid: state ID to test
9271  * @cred: credential
9272  *
9273  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9274  * Otherwise a negative NFS4ERR value is returned if the operation
9275  * failed or the state ID is not currently valid.
9276  */
9277 static int nfs41_test_stateid(struct nfs_server *server,
9278 		nfs4_stateid *stateid,
9279 		struct rpc_cred *cred)
9280 {
9281 	struct nfs4_exception exception = { };
9282 	int err;
9283 	do {
9284 		err = _nfs41_test_stateid(server, stateid, cred);
9285 		nfs4_handle_delay_or_session_error(server, err, &exception);
9286 	} while (exception.retry);
9287 	return err;
9288 }
9289 
9290 struct nfs_free_stateid_data {
9291 	struct nfs_server *server;
9292 	struct nfs41_free_stateid_args args;
9293 	struct nfs41_free_stateid_res res;
9294 };
9295 
9296 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9297 {
9298 	struct nfs_free_stateid_data *data = calldata;
9299 	nfs4_setup_sequence(data->server->nfs_client,
9300 			&data->args.seq_args,
9301 			&data->res.seq_res,
9302 			task);
9303 }
9304 
9305 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9306 {
9307 	struct nfs_free_stateid_data *data = calldata;
9308 
9309 	nfs41_sequence_done(task, &data->res.seq_res);
9310 
9311 	switch (task->tk_status) {
9312 	case -NFS4ERR_DELAY:
9313 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9314 			rpc_restart_call_prepare(task);
9315 	}
9316 }
9317 
9318 static void nfs41_free_stateid_release(void *calldata)
9319 {
9320 	kfree(calldata);
9321 }
9322 
9323 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9324 	.rpc_call_prepare = nfs41_free_stateid_prepare,
9325 	.rpc_call_done = nfs41_free_stateid_done,
9326 	.rpc_release = nfs41_free_stateid_release,
9327 };
9328 
9329 /**
9330  * nfs41_free_stateid - perform a FREE_STATEID operation
9331  *
9332  * @server: server / transport on which to perform the operation
9333  * @stateid: state ID to release
9334  * @cred: credential
9335  * @is_recovery: set to true if this call needs to be privileged
9336  *
9337  * Note: this function is always asynchronous.
9338  */
9339 static int nfs41_free_stateid(struct nfs_server *server,
9340 		const nfs4_stateid *stateid,
9341 		struct rpc_cred *cred,
9342 		bool privileged)
9343 {
9344 	struct rpc_message msg = {
9345 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9346 		.rpc_cred = cred,
9347 	};
9348 	struct rpc_task_setup task_setup = {
9349 		.rpc_client = server->client,
9350 		.rpc_message = &msg,
9351 		.callback_ops = &nfs41_free_stateid_ops,
9352 		.flags = RPC_TASK_ASYNC,
9353 	};
9354 	struct nfs_free_stateid_data *data;
9355 	struct rpc_task *task;
9356 
9357 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9358 		&task_setup.rpc_client, &msg);
9359 
9360 	dprintk("NFS call  free_stateid %p\n", stateid);
9361 	data = kmalloc(sizeof(*data), GFP_NOFS);
9362 	if (!data)
9363 		return -ENOMEM;
9364 	data->server = server;
9365 	nfs4_stateid_copy(&data->args.stateid, stateid);
9366 
9367 	task_setup.callback_data = data;
9368 
9369 	msg.rpc_argp = &data->args;
9370 	msg.rpc_resp = &data->res;
9371 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9372 	task = rpc_run_task(&task_setup);
9373 	if (IS_ERR(task))
9374 		return PTR_ERR(task);
9375 	rpc_put_task(task);
9376 	return 0;
9377 }
9378 
9379 static void
9380 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9381 {
9382 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9383 
9384 	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9385 	nfs4_free_lock_state(server, lsp);
9386 }
9387 
9388 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9389 		const nfs4_stateid *s2)
9390 {
9391 	if (s1->type != s2->type)
9392 		return false;
9393 
9394 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9395 		return false;
9396 
9397 	if (s1->seqid == s2->seqid)
9398 		return true;
9399 
9400 	return s1->seqid == 0 || s2->seqid == 0;
9401 }
9402 
9403 #endif /* CONFIG_NFS_V4_1 */
9404 
9405 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9406 		const nfs4_stateid *s2)
9407 {
9408 	return nfs4_stateid_match(s1, s2);
9409 }
9410 
9411 
9412 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9413 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9414 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
9415 	.recover_open	= nfs4_open_reclaim,
9416 	.recover_lock	= nfs4_lock_reclaim,
9417 	.establish_clid = nfs4_init_clientid,
9418 	.detect_trunking = nfs40_discover_server_trunking,
9419 };
9420 
9421 #if defined(CONFIG_NFS_V4_1)
9422 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9423 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9424 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
9425 	.recover_open	= nfs4_open_reclaim,
9426 	.recover_lock	= nfs4_lock_reclaim,
9427 	.establish_clid = nfs41_init_clientid,
9428 	.reclaim_complete = nfs41_proc_reclaim_complete,
9429 	.detect_trunking = nfs41_discover_server_trunking,
9430 };
9431 #endif /* CONFIG_NFS_V4_1 */
9432 
9433 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9434 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9435 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
9436 	.recover_open	= nfs40_open_expired,
9437 	.recover_lock	= nfs4_lock_expired,
9438 	.establish_clid = nfs4_init_clientid,
9439 };
9440 
9441 #if defined(CONFIG_NFS_V4_1)
9442 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9443 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9444 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
9445 	.recover_open	= nfs41_open_expired,
9446 	.recover_lock	= nfs41_lock_expired,
9447 	.establish_clid = nfs41_init_clientid,
9448 };
9449 #endif /* CONFIG_NFS_V4_1 */
9450 
9451 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9452 	.sched_state_renewal = nfs4_proc_async_renew,
9453 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9454 	.renew_lease = nfs4_proc_renew,
9455 };
9456 
9457 #if defined(CONFIG_NFS_V4_1)
9458 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9459 	.sched_state_renewal = nfs41_proc_async_sequence,
9460 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9461 	.renew_lease = nfs4_proc_sequence,
9462 };
9463 #endif
9464 
9465 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9466 	.get_locations = _nfs40_proc_get_locations,
9467 	.fsid_present = _nfs40_proc_fsid_present,
9468 };
9469 
9470 #if defined(CONFIG_NFS_V4_1)
9471 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9472 	.get_locations = _nfs41_proc_get_locations,
9473 	.fsid_present = _nfs41_proc_fsid_present,
9474 };
9475 #endif	/* CONFIG_NFS_V4_1 */
9476 
9477 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9478 	.minor_version = 0,
9479 	.init_caps = NFS_CAP_READDIRPLUS
9480 		| NFS_CAP_ATOMIC_OPEN
9481 		| NFS_CAP_POSIX_LOCK,
9482 	.init_client = nfs40_init_client,
9483 	.shutdown_client = nfs40_shutdown_client,
9484 	.match_stateid = nfs4_match_stateid,
9485 	.find_root_sec = nfs4_find_root_sec,
9486 	.free_lock_state = nfs4_release_lockowner,
9487 	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
9488 	.alloc_seqid = nfs_alloc_seqid,
9489 	.call_sync_ops = &nfs40_call_sync_ops,
9490 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9491 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9492 	.state_renewal_ops = &nfs40_state_renewal_ops,
9493 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
9494 };
9495 
9496 #if defined(CONFIG_NFS_V4_1)
9497 static struct nfs_seqid *
9498 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9499 {
9500 	return NULL;
9501 }
9502 
9503 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9504 	.minor_version = 1,
9505 	.init_caps = NFS_CAP_READDIRPLUS
9506 		| NFS_CAP_ATOMIC_OPEN
9507 		| NFS_CAP_POSIX_LOCK
9508 		| NFS_CAP_STATEID_NFSV41
9509 		| NFS_CAP_ATOMIC_OPEN_V1
9510 		| NFS_CAP_LGOPEN,
9511 	.init_client = nfs41_init_client,
9512 	.shutdown_client = nfs41_shutdown_client,
9513 	.match_stateid = nfs41_match_stateid,
9514 	.find_root_sec = nfs41_find_root_sec,
9515 	.free_lock_state = nfs41_free_lock_state,
9516 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
9517 	.alloc_seqid = nfs_alloc_no_seqid,
9518 	.session_trunk = nfs4_test_session_trunk,
9519 	.call_sync_ops = &nfs41_call_sync_ops,
9520 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9521 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9522 	.state_renewal_ops = &nfs41_state_renewal_ops,
9523 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
9524 };
9525 #endif
9526 
9527 #if defined(CONFIG_NFS_V4_2)
9528 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9529 	.minor_version = 2,
9530 	.init_caps = NFS_CAP_READDIRPLUS
9531 		| NFS_CAP_ATOMIC_OPEN
9532 		| NFS_CAP_POSIX_LOCK
9533 		| NFS_CAP_STATEID_NFSV41
9534 		| NFS_CAP_ATOMIC_OPEN_V1
9535 		| NFS_CAP_LGOPEN
9536 		| NFS_CAP_ALLOCATE
9537 		| NFS_CAP_COPY
9538 		| NFS_CAP_DEALLOCATE
9539 		| NFS_CAP_SEEK
9540 		| NFS_CAP_LAYOUTSTATS
9541 		| NFS_CAP_CLONE,
9542 	.init_client = nfs41_init_client,
9543 	.shutdown_client = nfs41_shutdown_client,
9544 	.match_stateid = nfs41_match_stateid,
9545 	.find_root_sec = nfs41_find_root_sec,
9546 	.free_lock_state = nfs41_free_lock_state,
9547 	.call_sync_ops = &nfs41_call_sync_ops,
9548 	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
9549 	.alloc_seqid = nfs_alloc_no_seqid,
9550 	.session_trunk = nfs4_test_session_trunk,
9551 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9552 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9553 	.state_renewal_ops = &nfs41_state_renewal_ops,
9554 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
9555 };
9556 #endif
9557 
9558 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9559 	[0] = &nfs_v4_0_minor_ops,
9560 #if defined(CONFIG_NFS_V4_1)
9561 	[1] = &nfs_v4_1_minor_ops,
9562 #endif
9563 #if defined(CONFIG_NFS_V4_2)
9564 	[2] = &nfs_v4_2_minor_ops,
9565 #endif
9566 };
9567 
9568 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9569 {
9570 	ssize_t error, error2;
9571 
9572 	error = generic_listxattr(dentry, list, size);
9573 	if (error < 0)
9574 		return error;
9575 	if (list) {
9576 		list += error;
9577 		size -= error;
9578 	}
9579 
9580 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9581 	if (error2 < 0)
9582 		return error2;
9583 	return error + error2;
9584 }
9585 
9586 static const struct inode_operations nfs4_dir_inode_operations = {
9587 	.create		= nfs_create,
9588 	.lookup		= nfs_lookup,
9589 	.atomic_open	= nfs_atomic_open,
9590 	.link		= nfs_link,
9591 	.unlink		= nfs_unlink,
9592 	.symlink	= nfs_symlink,
9593 	.mkdir		= nfs_mkdir,
9594 	.rmdir		= nfs_rmdir,
9595 	.mknod		= nfs_mknod,
9596 	.rename		= nfs_rename,
9597 	.permission	= nfs_permission,
9598 	.getattr	= nfs_getattr,
9599 	.setattr	= nfs_setattr,
9600 	.listxattr	= nfs4_listxattr,
9601 };
9602 
9603 static const struct inode_operations nfs4_file_inode_operations = {
9604 	.permission	= nfs_permission,
9605 	.getattr	= nfs_getattr,
9606 	.setattr	= nfs_setattr,
9607 	.listxattr	= nfs4_listxattr,
9608 };
9609 
9610 const struct nfs_rpc_ops nfs_v4_clientops = {
9611 	.version	= 4,			/* protocol version */
9612 	.dentry_ops	= &nfs4_dentry_operations,
9613 	.dir_inode_ops	= &nfs4_dir_inode_operations,
9614 	.file_inode_ops	= &nfs4_file_inode_operations,
9615 	.file_ops	= &nfs4_file_operations,
9616 	.getroot	= nfs4_proc_get_root,
9617 	.submount	= nfs4_submount,
9618 	.try_mount	= nfs4_try_mount,
9619 	.getattr	= nfs4_proc_getattr,
9620 	.setattr	= nfs4_proc_setattr,
9621 	.lookup		= nfs4_proc_lookup,
9622 	.lookupp	= nfs4_proc_lookupp,
9623 	.access		= nfs4_proc_access,
9624 	.readlink	= nfs4_proc_readlink,
9625 	.create		= nfs4_proc_create,
9626 	.remove		= nfs4_proc_remove,
9627 	.unlink_setup	= nfs4_proc_unlink_setup,
9628 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9629 	.unlink_done	= nfs4_proc_unlink_done,
9630 	.rename_setup	= nfs4_proc_rename_setup,
9631 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9632 	.rename_done	= nfs4_proc_rename_done,
9633 	.link		= nfs4_proc_link,
9634 	.symlink	= nfs4_proc_symlink,
9635 	.mkdir		= nfs4_proc_mkdir,
9636 	.rmdir		= nfs4_proc_rmdir,
9637 	.readdir	= nfs4_proc_readdir,
9638 	.mknod		= nfs4_proc_mknod,
9639 	.statfs		= nfs4_proc_statfs,
9640 	.fsinfo		= nfs4_proc_fsinfo,
9641 	.pathconf	= nfs4_proc_pathconf,
9642 	.set_capabilities = nfs4_server_capabilities,
9643 	.decode_dirent	= nfs4_decode_dirent,
9644 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9645 	.read_setup	= nfs4_proc_read_setup,
9646 	.read_done	= nfs4_read_done,
9647 	.write_setup	= nfs4_proc_write_setup,
9648 	.write_done	= nfs4_write_done,
9649 	.commit_setup	= nfs4_proc_commit_setup,
9650 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9651 	.commit_done	= nfs4_commit_done,
9652 	.lock		= nfs4_proc_lock,
9653 	.clear_acl_cache = nfs4_zap_acl_attr,
9654 	.close_context  = nfs4_close_context,
9655 	.open_context	= nfs4_atomic_open,
9656 	.have_delegation = nfs4_have_delegation,
9657 	.alloc_client	= nfs4_alloc_client,
9658 	.init_client	= nfs4_init_client,
9659 	.free_client	= nfs4_free_client,
9660 	.create_server	= nfs4_create_server,
9661 	.clone_server	= nfs_clone_server,
9662 };
9663 
9664 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9665 	.name	= XATTR_NAME_NFSV4_ACL,
9666 	.list	= nfs4_xattr_list_nfs4_acl,
9667 	.get	= nfs4_xattr_get_nfs4_acl,
9668 	.set	= nfs4_xattr_set_nfs4_acl,
9669 };
9670 
9671 const struct xattr_handler *nfs4_xattr_handlers[] = {
9672 	&nfs4_xattr_nfs4_acl_handler,
9673 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9674 	&nfs4_xattr_nfs4_label_handler,
9675 #endif
9676 	NULL
9677 };
9678 
9679 /*
9680  * Local variables:
9681  *  c-basic-offset: 8
9682  * End:
9683  */
9684