xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 110e6f26)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/file.h>
42 #include <linux/string.h>
43 #include <linux/ratelimit.h>
44 #include <linux/printk.h>
45 #include <linux/slab.h>
46 #include <linux/sunrpc/clnt.h>
47 #include <linux/nfs.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfs_fs.h>
50 #include <linux/nfs_page.h>
51 #include <linux/nfs_mount.h>
52 #include <linux/namei.h>
53 #include <linux/mount.h>
54 #include <linux/module.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69 
70 #include "nfs4trace.h"
71 
72 #define NFSDBG_FACILITY		NFSDBG_PROC
73 
74 #define NFS4_POLL_RETRY_MIN	(HZ/10)
75 #define NFS4_POLL_RETRY_MAX	(15*HZ)
76 
77 struct nfs4_opendata;
78 static int _nfs4_proc_open(struct nfs4_opendata *data);
79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_FILE_OPEN:
162 		return -EBUSY;
163 	default:
164 		dprintk("%s could not handle NFSv4 error %d\n",
165 				__func__, -err);
166 		break;
167 	}
168 	return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175 	FATTR4_WORD0_TYPE
176 	| FATTR4_WORD0_CHANGE
177 	| FATTR4_WORD0_SIZE
178 	| FATTR4_WORD0_FSID
179 	| FATTR4_WORD0_FILEID,
180 	FATTR4_WORD1_MODE
181 	| FATTR4_WORD1_NUMLINKS
182 	| FATTR4_WORD1_OWNER
183 	| FATTR4_WORD1_OWNER_GROUP
184 	| FATTR4_WORD1_RAWDEV
185 	| FATTR4_WORD1_SPACE_USED
186 	| FATTR4_WORD1_TIME_ACCESS
187 	| FATTR4_WORD1_TIME_METADATA
188 	| FATTR4_WORD1_TIME_MODIFY
189 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
190 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
191 	FATTR4_WORD2_SECURITY_LABEL
192 #endif
193 };
194 
195 static const u32 nfs4_pnfs_open_bitmap[3] = {
196 	FATTR4_WORD0_TYPE
197 	| FATTR4_WORD0_CHANGE
198 	| FATTR4_WORD0_SIZE
199 	| FATTR4_WORD0_FSID
200 	| FATTR4_WORD0_FILEID,
201 	FATTR4_WORD1_MODE
202 	| FATTR4_WORD1_NUMLINKS
203 	| FATTR4_WORD1_OWNER
204 	| FATTR4_WORD1_OWNER_GROUP
205 	| FATTR4_WORD1_RAWDEV
206 	| FATTR4_WORD1_SPACE_USED
207 	| FATTR4_WORD1_TIME_ACCESS
208 	| FATTR4_WORD1_TIME_METADATA
209 	| FATTR4_WORD1_TIME_MODIFY,
210 	FATTR4_WORD2_MDSTHRESHOLD
211 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
212 	| FATTR4_WORD2_SECURITY_LABEL
213 #endif
214 };
215 
216 static const u32 nfs4_open_noattr_bitmap[3] = {
217 	FATTR4_WORD0_TYPE
218 	| FATTR4_WORD0_CHANGE
219 	| FATTR4_WORD0_FILEID,
220 };
221 
222 const u32 nfs4_statfs_bitmap[3] = {
223 	FATTR4_WORD0_FILES_AVAIL
224 	| FATTR4_WORD0_FILES_FREE
225 	| FATTR4_WORD0_FILES_TOTAL,
226 	FATTR4_WORD1_SPACE_AVAIL
227 	| FATTR4_WORD1_SPACE_FREE
228 	| FATTR4_WORD1_SPACE_TOTAL
229 };
230 
231 const u32 nfs4_pathconf_bitmap[3] = {
232 	FATTR4_WORD0_MAXLINK
233 	| FATTR4_WORD0_MAXNAME,
234 	0
235 };
236 
237 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
238 			| FATTR4_WORD0_MAXREAD
239 			| FATTR4_WORD0_MAXWRITE
240 			| FATTR4_WORD0_LEASE_TIME,
241 			FATTR4_WORD1_TIME_DELTA
242 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
243 			FATTR4_WORD2_LAYOUT_BLKSIZE
244 			| FATTR4_WORD2_CLONE_BLKSIZE
245 };
246 
247 const u32 nfs4_fs_locations_bitmap[3] = {
248 	FATTR4_WORD0_TYPE
249 	| FATTR4_WORD0_CHANGE
250 	| FATTR4_WORD0_SIZE
251 	| FATTR4_WORD0_FSID
252 	| FATTR4_WORD0_FILEID
253 	| FATTR4_WORD0_FS_LOCATIONS,
254 	FATTR4_WORD1_MODE
255 	| FATTR4_WORD1_NUMLINKS
256 	| FATTR4_WORD1_OWNER
257 	| FATTR4_WORD1_OWNER_GROUP
258 	| FATTR4_WORD1_RAWDEV
259 	| FATTR4_WORD1_SPACE_USED
260 	| FATTR4_WORD1_TIME_ACCESS
261 	| FATTR4_WORD1_TIME_METADATA
262 	| FATTR4_WORD1_TIME_MODIFY
263 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
264 };
265 
266 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
267 		struct nfs4_readdir_arg *readdir)
268 {
269 	__be32 *start, *p;
270 
271 	if (cookie > 2) {
272 		readdir->cookie = cookie;
273 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
274 		return;
275 	}
276 
277 	readdir->cookie = 0;
278 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
279 	if (cookie == 2)
280 		return;
281 
282 	/*
283 	 * NFSv4 servers do not return entries for '.' and '..'
284 	 * Therefore, we fake these entries here.  We let '.'
285 	 * have cookie 0 and '..' have cookie 1.  Note that
286 	 * when talking to the server, we always send cookie 0
287 	 * instead of 1 or 2.
288 	 */
289 	start = p = kmap_atomic(*readdir->pages);
290 
291 	if (cookie == 0) {
292 		*p++ = xdr_one;                                  /* next */
293 		*p++ = xdr_zero;                   /* cookie, first word */
294 		*p++ = xdr_one;                   /* cookie, second word */
295 		*p++ = xdr_one;                             /* entry len */
296 		memcpy(p, ".\0\0\0", 4);                        /* entry */
297 		p++;
298 		*p++ = xdr_one;                         /* bitmap length */
299 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
300 		*p++ = htonl(8);              /* attribute buffer length */
301 		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
302 	}
303 
304 	*p++ = xdr_one;                                  /* next */
305 	*p++ = xdr_zero;                   /* cookie, first word */
306 	*p++ = xdr_two;                   /* cookie, second word */
307 	*p++ = xdr_two;                             /* entry len */
308 	memcpy(p, "..\0\0", 4);                         /* entry */
309 	p++;
310 	*p++ = xdr_one;                         /* bitmap length */
311 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
312 	*p++ = htonl(8);              /* attribute buffer length */
313 	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
314 
315 	readdir->pgbase = (char *)p - (char *)start;
316 	readdir->count -= readdir->pgbase;
317 	kunmap_atomic(start);
318 }
319 
320 static long nfs4_update_delay(long *timeout)
321 {
322 	long ret;
323 	if (!timeout)
324 		return NFS4_POLL_RETRY_MAX;
325 	if (*timeout <= 0)
326 		*timeout = NFS4_POLL_RETRY_MIN;
327 	if (*timeout > NFS4_POLL_RETRY_MAX)
328 		*timeout = NFS4_POLL_RETRY_MAX;
329 	ret = *timeout;
330 	*timeout <<= 1;
331 	return ret;
332 }
333 
334 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
335 {
336 	int res = 0;
337 
338 	might_sleep();
339 
340 	freezable_schedule_timeout_killable_unsafe(
341 		nfs4_update_delay(timeout));
342 	if (fatal_signal_pending(current))
343 		res = -ERESTARTSYS;
344 	return res;
345 }
346 
347 /* This is the error handling routine for processes that are allowed
348  * to sleep.
349  */
350 static int nfs4_do_handle_exception(struct nfs_server *server,
351 		int errorcode, struct nfs4_exception *exception)
352 {
353 	struct nfs_client *clp = server->nfs_client;
354 	struct nfs4_state *state = exception->state;
355 	struct inode *inode = exception->inode;
356 	int ret = errorcode;
357 
358 	exception->delay = 0;
359 	exception->recovering = 0;
360 	exception->retry = 0;
361 	switch(errorcode) {
362 		case 0:
363 			return 0;
364 		case -NFS4ERR_OPENMODE:
365 		case -NFS4ERR_DELEG_REVOKED:
366 		case -NFS4ERR_ADMIN_REVOKED:
367 		case -NFS4ERR_BAD_STATEID:
368 			if (inode && nfs_async_inode_return_delegation(inode,
369 						NULL) == 0)
370 				goto wait_on_recovery;
371 			if (state == NULL)
372 				break;
373 			ret = nfs4_schedule_stateid_recovery(server, state);
374 			if (ret < 0)
375 				break;
376 			goto wait_on_recovery;
377 		case -NFS4ERR_EXPIRED:
378 			if (state != NULL) {
379 				ret = nfs4_schedule_stateid_recovery(server, state);
380 				if (ret < 0)
381 					break;
382 			}
383 		case -NFS4ERR_STALE_STATEID:
384 		case -NFS4ERR_STALE_CLIENTID:
385 			nfs4_schedule_lease_recovery(clp);
386 			goto wait_on_recovery;
387 		case -NFS4ERR_MOVED:
388 			ret = nfs4_schedule_migration_recovery(server);
389 			if (ret < 0)
390 				break;
391 			goto wait_on_recovery;
392 		case -NFS4ERR_LEASE_MOVED:
393 			nfs4_schedule_lease_moved_recovery(clp);
394 			goto wait_on_recovery;
395 #if defined(CONFIG_NFS_V4_1)
396 		case -NFS4ERR_BADSESSION:
397 		case -NFS4ERR_BADSLOT:
398 		case -NFS4ERR_BAD_HIGH_SLOT:
399 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
400 		case -NFS4ERR_DEADSESSION:
401 		case -NFS4ERR_SEQ_FALSE_RETRY:
402 		case -NFS4ERR_SEQ_MISORDERED:
403 			dprintk("%s ERROR: %d Reset session\n", __func__,
404 				errorcode);
405 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
406 			goto wait_on_recovery;
407 #endif /* defined(CONFIG_NFS_V4_1) */
408 		case -NFS4ERR_FILE_OPEN:
409 			if (exception->timeout > HZ) {
410 				/* We have retried a decent amount, time to
411 				 * fail
412 				 */
413 				ret = -EBUSY;
414 				break;
415 			}
416 		case -NFS4ERR_DELAY:
417 			nfs_inc_server_stats(server, NFSIOS_DELAY);
418 		case -NFS4ERR_GRACE:
419 			exception->delay = 1;
420 			return 0;
421 
422 		case -NFS4ERR_RETRY_UNCACHED_REP:
423 		case -NFS4ERR_OLD_STATEID:
424 			exception->retry = 1;
425 			break;
426 		case -NFS4ERR_BADOWNER:
427 			/* The following works around a Linux server bug! */
428 		case -NFS4ERR_BADNAME:
429 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
430 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
431 				exception->retry = 1;
432 				printk(KERN_WARNING "NFS: v4 server %s "
433 						"does not accept raw "
434 						"uid/gids. "
435 						"Reenabling the idmapper.\n",
436 						server->nfs_client->cl_hostname);
437 			}
438 	}
439 	/* We failed to handle the error */
440 	return nfs4_map_errors(ret);
441 wait_on_recovery:
442 	exception->recovering = 1;
443 	return 0;
444 }
445 
446 /* This is the error handling routine for processes that are allowed
447  * to sleep.
448  */
449 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
450 {
451 	struct nfs_client *clp = server->nfs_client;
452 	int ret;
453 
454 	ret = nfs4_do_handle_exception(server, errorcode, exception);
455 	if (exception->delay) {
456 		ret = nfs4_delay(server->client, &exception->timeout);
457 		goto out_retry;
458 	}
459 	if (exception->recovering) {
460 		ret = nfs4_wait_clnt_recover(clp);
461 		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
462 			return -EIO;
463 		goto out_retry;
464 	}
465 	return ret;
466 out_retry:
467 	if (ret == 0)
468 		exception->retry = 1;
469 	return ret;
470 }
471 
472 static int
473 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
474 		int errorcode, struct nfs4_exception *exception)
475 {
476 	struct nfs_client *clp = server->nfs_client;
477 	int ret;
478 
479 	ret = nfs4_do_handle_exception(server, errorcode, exception);
480 	if (exception->delay) {
481 		rpc_delay(task, nfs4_update_delay(&exception->timeout));
482 		goto out_retry;
483 	}
484 	if (exception->recovering) {
485 		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
486 		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
487 			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
488 		goto out_retry;
489 	}
490 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
491 		ret = -EIO;
492 	return ret;
493 out_retry:
494 	if (ret == 0)
495 		exception->retry = 1;
496 	return ret;
497 }
498 
499 static int
500 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
501 			struct nfs4_state *state, long *timeout)
502 {
503 	struct nfs4_exception exception = {
504 		.state = state,
505 	};
506 
507 	if (task->tk_status >= 0)
508 		return 0;
509 	if (timeout)
510 		exception.timeout = *timeout;
511 	task->tk_status = nfs4_async_handle_exception(task, server,
512 			task->tk_status,
513 			&exception);
514 	if (exception.delay && timeout)
515 		*timeout = exception.timeout;
516 	if (exception.retry)
517 		return -EAGAIN;
518 	return 0;
519 }
520 
521 /*
522  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
523  * or 'false' otherwise.
524  */
525 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
526 {
527 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
528 
529 	if (flavor == RPC_AUTH_GSS_KRB5I ||
530 	    flavor == RPC_AUTH_GSS_KRB5P)
531 		return true;
532 
533 	return false;
534 }
535 
536 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
537 {
538 	spin_lock(&clp->cl_lock);
539 	if (time_before(clp->cl_last_renewal,timestamp))
540 		clp->cl_last_renewal = timestamp;
541 	spin_unlock(&clp->cl_lock);
542 }
543 
544 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
545 {
546 	struct nfs_client *clp = server->nfs_client;
547 
548 	if (!nfs4_has_session(clp))
549 		do_renew_lease(clp, timestamp);
550 }
551 
552 struct nfs4_call_sync_data {
553 	const struct nfs_server *seq_server;
554 	struct nfs4_sequence_args *seq_args;
555 	struct nfs4_sequence_res *seq_res;
556 };
557 
558 void nfs4_init_sequence(struct nfs4_sequence_args *args,
559 			struct nfs4_sequence_res *res, int cache_reply)
560 {
561 	args->sa_slot = NULL;
562 	args->sa_cache_this = cache_reply;
563 	args->sa_privileged = 0;
564 
565 	res->sr_slot = NULL;
566 }
567 
568 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
569 {
570 	args->sa_privileged = 1;
571 }
572 
573 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
574 			 struct nfs4_sequence_args *args,
575 			 struct nfs4_sequence_res *res,
576 			 struct rpc_task *task)
577 {
578 	struct nfs4_slot *slot;
579 
580 	/* slot already allocated? */
581 	if (res->sr_slot != NULL)
582 		goto out_start;
583 
584 	spin_lock(&tbl->slot_tbl_lock);
585 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
586 		goto out_sleep;
587 
588 	slot = nfs4_alloc_slot(tbl);
589 	if (IS_ERR(slot)) {
590 		if (slot == ERR_PTR(-ENOMEM))
591 			task->tk_timeout = HZ >> 2;
592 		goto out_sleep;
593 	}
594 	spin_unlock(&tbl->slot_tbl_lock);
595 
596 	args->sa_slot = slot;
597 	res->sr_slot = slot;
598 
599 out_start:
600 	rpc_call_start(task);
601 	return 0;
602 
603 out_sleep:
604 	if (args->sa_privileged)
605 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
606 				NULL, RPC_PRIORITY_PRIVILEGED);
607 	else
608 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
609 	spin_unlock(&tbl->slot_tbl_lock);
610 	return -EAGAIN;
611 }
612 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
613 
614 static int nfs40_sequence_done(struct rpc_task *task,
615 			       struct nfs4_sequence_res *res)
616 {
617 	struct nfs4_slot *slot = res->sr_slot;
618 	struct nfs4_slot_table *tbl;
619 
620 	if (slot == NULL)
621 		goto out;
622 
623 	tbl = slot->table;
624 	spin_lock(&tbl->slot_tbl_lock);
625 	if (!nfs41_wake_and_assign_slot(tbl, slot))
626 		nfs4_free_slot(tbl, slot);
627 	spin_unlock(&tbl->slot_tbl_lock);
628 
629 	res->sr_slot = NULL;
630 out:
631 	return 1;
632 }
633 
634 #if defined(CONFIG_NFS_V4_1)
635 
636 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
637 {
638 	struct nfs4_session *session;
639 	struct nfs4_slot_table *tbl;
640 	struct nfs4_slot *slot = res->sr_slot;
641 	bool send_new_highest_used_slotid = false;
642 
643 	tbl = slot->table;
644 	session = tbl->session;
645 
646 	spin_lock(&tbl->slot_tbl_lock);
647 	/* Be nice to the server: try to ensure that the last transmitted
648 	 * value for highest_user_slotid <= target_highest_slotid
649 	 */
650 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
651 		send_new_highest_used_slotid = true;
652 
653 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
654 		send_new_highest_used_slotid = false;
655 		goto out_unlock;
656 	}
657 	nfs4_free_slot(tbl, slot);
658 
659 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
660 		send_new_highest_used_slotid = false;
661 out_unlock:
662 	spin_unlock(&tbl->slot_tbl_lock);
663 	res->sr_slot = NULL;
664 	if (send_new_highest_used_slotid)
665 		nfs41_notify_server(session->clp);
666 }
667 
668 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
669 {
670 	struct nfs4_session *session;
671 	struct nfs4_slot *slot = res->sr_slot;
672 	struct nfs_client *clp;
673 	bool interrupted = false;
674 	int ret = 1;
675 
676 	if (slot == NULL)
677 		goto out_noaction;
678 	/* don't increment the sequence number if the task wasn't sent */
679 	if (!RPC_WAS_SENT(task))
680 		goto out;
681 
682 	session = slot->table->session;
683 
684 	if (slot->interrupted) {
685 		slot->interrupted = 0;
686 		interrupted = true;
687 	}
688 
689 	trace_nfs4_sequence_done(session, res);
690 	/* Check the SEQUENCE operation status */
691 	switch (res->sr_status) {
692 	case 0:
693 		/* Update the slot's sequence and clientid lease timer */
694 		++slot->seq_nr;
695 		clp = session->clp;
696 		do_renew_lease(clp, res->sr_timestamp);
697 		/* Check sequence flags */
698 		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
699 		nfs41_update_target_slotid(slot->table, slot, res);
700 		break;
701 	case 1:
702 		/*
703 		 * sr_status remains 1 if an RPC level error occurred.
704 		 * The server may or may not have processed the sequence
705 		 * operation..
706 		 * Mark the slot as having hosted an interrupted RPC call.
707 		 */
708 		slot->interrupted = 1;
709 		goto out;
710 	case -NFS4ERR_DELAY:
711 		/* The server detected a resend of the RPC call and
712 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
713 		 * of RFC5661.
714 		 */
715 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
716 			__func__,
717 			slot->slot_nr,
718 			slot->seq_nr);
719 		goto out_retry;
720 	case -NFS4ERR_BADSLOT:
721 		/*
722 		 * The slot id we used was probably retired. Try again
723 		 * using a different slot id.
724 		 */
725 		goto retry_nowait;
726 	case -NFS4ERR_SEQ_MISORDERED:
727 		/*
728 		 * Was the last operation on this sequence interrupted?
729 		 * If so, retry after bumping the sequence number.
730 		 */
731 		if (interrupted) {
732 			++slot->seq_nr;
733 			goto retry_nowait;
734 		}
735 		/*
736 		 * Could this slot have been previously retired?
737 		 * If so, then the server may be expecting seq_nr = 1!
738 		 */
739 		if (slot->seq_nr != 1) {
740 			slot->seq_nr = 1;
741 			goto retry_nowait;
742 		}
743 		break;
744 	case -NFS4ERR_SEQ_FALSE_RETRY:
745 		++slot->seq_nr;
746 		goto retry_nowait;
747 	default:
748 		/* Just update the slot sequence no. */
749 		++slot->seq_nr;
750 	}
751 out:
752 	/* The session may be reset by one of the error handlers. */
753 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
754 	nfs41_sequence_free_slot(res);
755 out_noaction:
756 	return ret;
757 retry_nowait:
758 	if (rpc_restart_call_prepare(task)) {
759 		task->tk_status = 0;
760 		ret = 0;
761 	}
762 	goto out;
763 out_retry:
764 	if (!rpc_restart_call(task))
765 		goto out;
766 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
767 	return 0;
768 }
769 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
770 
771 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
772 {
773 	if (res->sr_slot == NULL)
774 		return 1;
775 	if (!res->sr_slot->table->session)
776 		return nfs40_sequence_done(task, res);
777 	return nfs41_sequence_done(task, res);
778 }
779 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
780 
781 int nfs41_setup_sequence(struct nfs4_session *session,
782 				struct nfs4_sequence_args *args,
783 				struct nfs4_sequence_res *res,
784 				struct rpc_task *task)
785 {
786 	struct nfs4_slot *slot;
787 	struct nfs4_slot_table *tbl;
788 
789 	dprintk("--> %s\n", __func__);
790 	/* slot already allocated? */
791 	if (res->sr_slot != NULL)
792 		goto out_success;
793 
794 	tbl = &session->fc_slot_table;
795 
796 	task->tk_timeout = 0;
797 
798 	spin_lock(&tbl->slot_tbl_lock);
799 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
800 	    !args->sa_privileged) {
801 		/* The state manager will wait until the slot table is empty */
802 		dprintk("%s session is draining\n", __func__);
803 		goto out_sleep;
804 	}
805 
806 	slot = nfs4_alloc_slot(tbl);
807 	if (IS_ERR(slot)) {
808 		/* If out of memory, try again in 1/4 second */
809 		if (slot == ERR_PTR(-ENOMEM))
810 			task->tk_timeout = HZ >> 2;
811 		dprintk("<-- %s: no free slots\n", __func__);
812 		goto out_sleep;
813 	}
814 	spin_unlock(&tbl->slot_tbl_lock);
815 
816 	args->sa_slot = slot;
817 
818 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
819 			slot->slot_nr, slot->seq_nr);
820 
821 	res->sr_slot = slot;
822 	res->sr_timestamp = jiffies;
823 	res->sr_status_flags = 0;
824 	/*
825 	 * sr_status is only set in decode_sequence, and so will remain
826 	 * set to 1 if an rpc level failure occurs.
827 	 */
828 	res->sr_status = 1;
829 	trace_nfs4_setup_sequence(session, args);
830 out_success:
831 	rpc_call_start(task);
832 	return 0;
833 out_sleep:
834 	/* Privileged tasks are queued with top priority */
835 	if (args->sa_privileged)
836 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
837 				NULL, RPC_PRIORITY_PRIVILEGED);
838 	else
839 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
840 	spin_unlock(&tbl->slot_tbl_lock);
841 	return -EAGAIN;
842 }
843 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
844 
845 static int nfs4_setup_sequence(const struct nfs_server *server,
846 			       struct nfs4_sequence_args *args,
847 			       struct nfs4_sequence_res *res,
848 			       struct rpc_task *task)
849 {
850 	struct nfs4_session *session = nfs4_get_session(server);
851 	int ret = 0;
852 
853 	if (!session)
854 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
855 					    args, res, task);
856 
857 	dprintk("--> %s clp %p session %p sr_slot %u\n",
858 		__func__, session->clp, session, res->sr_slot ?
859 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
860 
861 	ret = nfs41_setup_sequence(session, args, res, task);
862 
863 	dprintk("<-- %s status=%d\n", __func__, ret);
864 	return ret;
865 }
866 
867 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
868 {
869 	struct nfs4_call_sync_data *data = calldata;
870 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
871 
872 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
873 
874 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
875 }
876 
877 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
878 {
879 	struct nfs4_call_sync_data *data = calldata;
880 
881 	nfs41_sequence_done(task, data->seq_res);
882 }
883 
884 static const struct rpc_call_ops nfs41_call_sync_ops = {
885 	.rpc_call_prepare = nfs41_call_sync_prepare,
886 	.rpc_call_done = nfs41_call_sync_done,
887 };
888 
889 #else	/* !CONFIG_NFS_V4_1 */
890 
891 static int nfs4_setup_sequence(const struct nfs_server *server,
892 			       struct nfs4_sequence_args *args,
893 			       struct nfs4_sequence_res *res,
894 			       struct rpc_task *task)
895 {
896 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
897 				    args, res, task);
898 }
899 
900 int nfs4_sequence_done(struct rpc_task *task,
901 		       struct nfs4_sequence_res *res)
902 {
903 	return nfs40_sequence_done(task, res);
904 }
905 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
906 
907 #endif	/* !CONFIG_NFS_V4_1 */
908 
909 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
910 {
911 	struct nfs4_call_sync_data *data = calldata;
912 	nfs4_setup_sequence(data->seq_server,
913 				data->seq_args, data->seq_res, task);
914 }
915 
916 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
917 {
918 	struct nfs4_call_sync_data *data = calldata;
919 	nfs4_sequence_done(task, data->seq_res);
920 }
921 
922 static const struct rpc_call_ops nfs40_call_sync_ops = {
923 	.rpc_call_prepare = nfs40_call_sync_prepare,
924 	.rpc_call_done = nfs40_call_sync_done,
925 };
926 
927 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
928 				   struct nfs_server *server,
929 				   struct rpc_message *msg,
930 				   struct nfs4_sequence_args *args,
931 				   struct nfs4_sequence_res *res)
932 {
933 	int ret;
934 	struct rpc_task *task;
935 	struct nfs_client *clp = server->nfs_client;
936 	struct nfs4_call_sync_data data = {
937 		.seq_server = server,
938 		.seq_args = args,
939 		.seq_res = res,
940 	};
941 	struct rpc_task_setup task_setup = {
942 		.rpc_client = clnt,
943 		.rpc_message = msg,
944 		.callback_ops = clp->cl_mvops->call_sync_ops,
945 		.callback_data = &data
946 	};
947 
948 	task = rpc_run_task(&task_setup);
949 	if (IS_ERR(task))
950 		ret = PTR_ERR(task);
951 	else {
952 		ret = task->tk_status;
953 		rpc_put_task(task);
954 	}
955 	return ret;
956 }
957 
958 int nfs4_call_sync(struct rpc_clnt *clnt,
959 		   struct nfs_server *server,
960 		   struct rpc_message *msg,
961 		   struct nfs4_sequence_args *args,
962 		   struct nfs4_sequence_res *res,
963 		   int cache_reply)
964 {
965 	nfs4_init_sequence(args, res, cache_reply);
966 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
967 }
968 
969 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
970 {
971 	struct nfs_inode *nfsi = NFS_I(dir);
972 
973 	spin_lock(&dir->i_lock);
974 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
975 	if (!cinfo->atomic || cinfo->before != dir->i_version)
976 		nfs_force_lookup_revalidate(dir);
977 	dir->i_version = cinfo->after;
978 	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
979 	nfs_fscache_invalidate(dir);
980 	spin_unlock(&dir->i_lock);
981 }
982 
983 struct nfs4_opendata {
984 	struct kref kref;
985 	struct nfs_openargs o_arg;
986 	struct nfs_openres o_res;
987 	struct nfs_open_confirmargs c_arg;
988 	struct nfs_open_confirmres c_res;
989 	struct nfs4_string owner_name;
990 	struct nfs4_string group_name;
991 	struct nfs4_label *a_label;
992 	struct nfs_fattr f_attr;
993 	struct nfs4_label *f_label;
994 	struct dentry *dir;
995 	struct dentry *dentry;
996 	struct nfs4_state_owner *owner;
997 	struct nfs4_state *state;
998 	struct iattr attrs;
999 	unsigned long timestamp;
1000 	unsigned int rpc_done : 1;
1001 	unsigned int file_created : 1;
1002 	unsigned int is_recover : 1;
1003 	int rpc_status;
1004 	int cancelled;
1005 };
1006 
1007 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1008 		int err, struct nfs4_exception *exception)
1009 {
1010 	if (err != -EINVAL)
1011 		return false;
1012 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1013 		return false;
1014 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1015 	exception->retry = 1;
1016 	return true;
1017 }
1018 
1019 static u32
1020 nfs4_map_atomic_open_share(struct nfs_server *server,
1021 		fmode_t fmode, int openflags)
1022 {
1023 	u32 res = 0;
1024 
1025 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1026 	case FMODE_READ:
1027 		res = NFS4_SHARE_ACCESS_READ;
1028 		break;
1029 	case FMODE_WRITE:
1030 		res = NFS4_SHARE_ACCESS_WRITE;
1031 		break;
1032 	case FMODE_READ|FMODE_WRITE:
1033 		res = NFS4_SHARE_ACCESS_BOTH;
1034 	}
1035 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1036 		goto out;
1037 	/* Want no delegation if we're using O_DIRECT */
1038 	if (openflags & O_DIRECT)
1039 		res |= NFS4_SHARE_WANT_NO_DELEG;
1040 out:
1041 	return res;
1042 }
1043 
1044 static enum open_claim_type4
1045 nfs4_map_atomic_open_claim(struct nfs_server *server,
1046 		enum open_claim_type4 claim)
1047 {
1048 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1049 		return claim;
1050 	switch (claim) {
1051 	default:
1052 		return claim;
1053 	case NFS4_OPEN_CLAIM_FH:
1054 		return NFS4_OPEN_CLAIM_NULL;
1055 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1056 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1057 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1058 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1059 	}
1060 }
1061 
1062 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1063 {
1064 	p->o_res.f_attr = &p->f_attr;
1065 	p->o_res.f_label = p->f_label;
1066 	p->o_res.seqid = p->o_arg.seqid;
1067 	p->c_res.seqid = p->c_arg.seqid;
1068 	p->o_res.server = p->o_arg.server;
1069 	p->o_res.access_request = p->o_arg.access;
1070 	nfs_fattr_init(&p->f_attr);
1071 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1072 }
1073 
1074 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1075 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1076 		const struct iattr *attrs,
1077 		struct nfs4_label *label,
1078 		enum open_claim_type4 claim,
1079 		gfp_t gfp_mask)
1080 {
1081 	struct dentry *parent = dget_parent(dentry);
1082 	struct inode *dir = d_inode(parent);
1083 	struct nfs_server *server = NFS_SERVER(dir);
1084 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1085 	struct nfs4_opendata *p;
1086 
1087 	p = kzalloc(sizeof(*p), gfp_mask);
1088 	if (p == NULL)
1089 		goto err;
1090 
1091 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1092 	if (IS_ERR(p->f_label))
1093 		goto err_free_p;
1094 
1095 	p->a_label = nfs4_label_alloc(server, gfp_mask);
1096 	if (IS_ERR(p->a_label))
1097 		goto err_free_f;
1098 
1099 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1100 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1101 	if (IS_ERR(p->o_arg.seqid))
1102 		goto err_free_label;
1103 	nfs_sb_active(dentry->d_sb);
1104 	p->dentry = dget(dentry);
1105 	p->dir = parent;
1106 	p->owner = sp;
1107 	atomic_inc(&sp->so_count);
1108 	p->o_arg.open_flags = flags;
1109 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1110 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1111 			fmode, flags);
1112 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1113 	 * will return permission denied for all bits until close */
1114 	if (!(flags & O_EXCL)) {
1115 		/* ask server to check for all possible rights as results
1116 		 * are cached */
1117 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1118 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1119 	}
1120 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1121 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1122 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1123 	p->o_arg.name = &dentry->d_name;
1124 	p->o_arg.server = server;
1125 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1126 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1127 	p->o_arg.label = nfs4_label_copy(p->a_label, label);
1128 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1129 	switch (p->o_arg.claim) {
1130 	case NFS4_OPEN_CLAIM_NULL:
1131 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1132 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1133 		p->o_arg.fh = NFS_FH(dir);
1134 		break;
1135 	case NFS4_OPEN_CLAIM_PREVIOUS:
1136 	case NFS4_OPEN_CLAIM_FH:
1137 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1138 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1139 		p->o_arg.fh = NFS_FH(d_inode(dentry));
1140 	}
1141 	if (attrs != NULL && attrs->ia_valid != 0) {
1142 		__u32 verf[2];
1143 
1144 		p->o_arg.u.attrs = &p->attrs;
1145 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1146 
1147 		verf[0] = jiffies;
1148 		verf[1] = current->pid;
1149 		memcpy(p->o_arg.u.verifier.data, verf,
1150 				sizeof(p->o_arg.u.verifier.data));
1151 	}
1152 	p->c_arg.fh = &p->o_res.fh;
1153 	p->c_arg.stateid = &p->o_res.stateid;
1154 	p->c_arg.seqid = p->o_arg.seqid;
1155 	nfs4_init_opendata_res(p);
1156 	kref_init(&p->kref);
1157 	return p;
1158 
1159 err_free_label:
1160 	nfs4_label_free(p->a_label);
1161 err_free_f:
1162 	nfs4_label_free(p->f_label);
1163 err_free_p:
1164 	kfree(p);
1165 err:
1166 	dput(parent);
1167 	return NULL;
1168 }
1169 
1170 static void nfs4_opendata_free(struct kref *kref)
1171 {
1172 	struct nfs4_opendata *p = container_of(kref,
1173 			struct nfs4_opendata, kref);
1174 	struct super_block *sb = p->dentry->d_sb;
1175 
1176 	nfs_free_seqid(p->o_arg.seqid);
1177 	if (p->state != NULL)
1178 		nfs4_put_open_state(p->state);
1179 	nfs4_put_state_owner(p->owner);
1180 
1181 	nfs4_label_free(p->a_label);
1182 	nfs4_label_free(p->f_label);
1183 
1184 	dput(p->dir);
1185 	dput(p->dentry);
1186 	nfs_sb_deactive(sb);
1187 	nfs_fattr_free_names(&p->f_attr);
1188 	kfree(p->f_attr.mdsthreshold);
1189 	kfree(p);
1190 }
1191 
1192 static void nfs4_opendata_put(struct nfs4_opendata *p)
1193 {
1194 	if (p != NULL)
1195 		kref_put(&p->kref, nfs4_opendata_free);
1196 }
1197 
1198 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1199 {
1200 	int ret;
1201 
1202 	ret = rpc_wait_for_completion_task(task);
1203 	return ret;
1204 }
1205 
1206 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1207 		fmode_t fmode)
1208 {
1209 	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1210 	case FMODE_READ|FMODE_WRITE:
1211 		return state->n_rdwr != 0;
1212 	case FMODE_WRITE:
1213 		return state->n_wronly != 0;
1214 	case FMODE_READ:
1215 		return state->n_rdonly != 0;
1216 	}
1217 	WARN_ON_ONCE(1);
1218 	return false;
1219 }
1220 
1221 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1222 {
1223 	int ret = 0;
1224 
1225 	if (open_mode & (O_EXCL|O_TRUNC))
1226 		goto out;
1227 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1228 		case FMODE_READ:
1229 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1230 				&& state->n_rdonly != 0;
1231 			break;
1232 		case FMODE_WRITE:
1233 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1234 				&& state->n_wronly != 0;
1235 			break;
1236 		case FMODE_READ|FMODE_WRITE:
1237 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1238 				&& state->n_rdwr != 0;
1239 	}
1240 out:
1241 	return ret;
1242 }
1243 
1244 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1245 		enum open_claim_type4 claim)
1246 {
1247 	if (delegation == NULL)
1248 		return 0;
1249 	if ((delegation->type & fmode) != fmode)
1250 		return 0;
1251 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1252 		return 0;
1253 	switch (claim) {
1254 	case NFS4_OPEN_CLAIM_NULL:
1255 	case NFS4_OPEN_CLAIM_FH:
1256 		break;
1257 	case NFS4_OPEN_CLAIM_PREVIOUS:
1258 		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1259 			break;
1260 	default:
1261 		return 0;
1262 	}
1263 	nfs_mark_delegation_referenced(delegation);
1264 	return 1;
1265 }
1266 
1267 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1268 {
1269 	switch (fmode) {
1270 		case FMODE_WRITE:
1271 			state->n_wronly++;
1272 			break;
1273 		case FMODE_READ:
1274 			state->n_rdonly++;
1275 			break;
1276 		case FMODE_READ|FMODE_WRITE:
1277 			state->n_rdwr++;
1278 	}
1279 	nfs4_state_set_mode_locked(state, state->state | fmode);
1280 }
1281 
1282 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1283 {
1284 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1285 	bool need_recover = false;
1286 
1287 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1288 		need_recover = true;
1289 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1290 		need_recover = true;
1291 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1292 		need_recover = true;
1293 	if (need_recover)
1294 		nfs4_state_mark_reclaim_nograce(clp, state);
1295 }
1296 
1297 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1298 		nfs4_stateid *stateid)
1299 {
1300 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1301 		return true;
1302 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1303 		nfs_test_and_clear_all_open_stateid(state);
1304 		return true;
1305 	}
1306 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1307 		return true;
1308 	return false;
1309 }
1310 
1311 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1312 {
1313 	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1314 		return;
1315 	if (state->n_wronly)
1316 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1317 	if (state->n_rdonly)
1318 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1319 	if (state->n_rdwr)
1320 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1321 	set_bit(NFS_OPEN_STATE, &state->flags);
1322 }
1323 
1324 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1325 		nfs4_stateid *arg_stateid,
1326 		nfs4_stateid *stateid, fmode_t fmode)
1327 {
1328 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1329 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1330 	case FMODE_WRITE:
1331 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1332 		break;
1333 	case FMODE_READ:
1334 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1335 		break;
1336 	case 0:
1337 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1338 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1339 		clear_bit(NFS_OPEN_STATE, &state->flags);
1340 	}
1341 	if (stateid == NULL)
1342 		return;
1343 	/* Handle races with OPEN */
1344 	if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1345 	    (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1346 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1347 		nfs_resync_open_stateid_locked(state);
1348 		return;
1349 	}
1350 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1351 		nfs4_stateid_copy(&state->stateid, stateid);
1352 	nfs4_stateid_copy(&state->open_stateid, stateid);
1353 }
1354 
1355 static void nfs_clear_open_stateid(struct nfs4_state *state,
1356 	nfs4_stateid *arg_stateid,
1357 	nfs4_stateid *stateid, fmode_t fmode)
1358 {
1359 	write_seqlock(&state->seqlock);
1360 	nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1361 	write_sequnlock(&state->seqlock);
1362 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1363 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1364 }
1365 
1366 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1367 {
1368 	switch (fmode) {
1369 		case FMODE_READ:
1370 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1371 			break;
1372 		case FMODE_WRITE:
1373 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1374 			break;
1375 		case FMODE_READ|FMODE_WRITE:
1376 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1377 	}
1378 	if (!nfs_need_update_open_stateid(state, stateid))
1379 		return;
1380 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1381 		nfs4_stateid_copy(&state->stateid, stateid);
1382 	nfs4_stateid_copy(&state->open_stateid, stateid);
1383 }
1384 
1385 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1386 {
1387 	/*
1388 	 * Protect the call to nfs4_state_set_mode_locked and
1389 	 * serialise the stateid update
1390 	 */
1391 	spin_lock(&state->owner->so_lock);
1392 	write_seqlock(&state->seqlock);
1393 	if (deleg_stateid != NULL) {
1394 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1395 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1396 	}
1397 	if (open_stateid != NULL)
1398 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1399 	write_sequnlock(&state->seqlock);
1400 	update_open_stateflags(state, fmode);
1401 	spin_unlock(&state->owner->so_lock);
1402 }
1403 
1404 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1405 {
1406 	struct nfs_inode *nfsi = NFS_I(state->inode);
1407 	struct nfs_delegation *deleg_cur;
1408 	int ret = 0;
1409 
1410 	fmode &= (FMODE_READ|FMODE_WRITE);
1411 
1412 	rcu_read_lock();
1413 	deleg_cur = rcu_dereference(nfsi->delegation);
1414 	if (deleg_cur == NULL)
1415 		goto no_delegation;
1416 
1417 	spin_lock(&deleg_cur->lock);
1418 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1419 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1420 	    (deleg_cur->type & fmode) != fmode)
1421 		goto no_delegation_unlock;
1422 
1423 	if (delegation == NULL)
1424 		delegation = &deleg_cur->stateid;
1425 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1426 		goto no_delegation_unlock;
1427 
1428 	nfs_mark_delegation_referenced(deleg_cur);
1429 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1430 	ret = 1;
1431 no_delegation_unlock:
1432 	spin_unlock(&deleg_cur->lock);
1433 no_delegation:
1434 	rcu_read_unlock();
1435 
1436 	if (!ret && open_stateid != NULL) {
1437 		__update_open_stateid(state, open_stateid, NULL, fmode);
1438 		ret = 1;
1439 	}
1440 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1441 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1442 
1443 	return ret;
1444 }
1445 
1446 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1447 		const nfs4_stateid *stateid)
1448 {
1449 	struct nfs4_state *state = lsp->ls_state;
1450 	bool ret = false;
1451 
1452 	spin_lock(&state->state_lock);
1453 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1454 		goto out_noupdate;
1455 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1456 		goto out_noupdate;
1457 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1458 	ret = true;
1459 out_noupdate:
1460 	spin_unlock(&state->state_lock);
1461 	return ret;
1462 }
1463 
1464 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1465 {
1466 	struct nfs_delegation *delegation;
1467 
1468 	rcu_read_lock();
1469 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1470 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1471 		rcu_read_unlock();
1472 		return;
1473 	}
1474 	rcu_read_unlock();
1475 	nfs4_inode_return_delegation(inode);
1476 }
1477 
1478 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1479 {
1480 	struct nfs4_state *state = opendata->state;
1481 	struct nfs_inode *nfsi = NFS_I(state->inode);
1482 	struct nfs_delegation *delegation;
1483 	int open_mode = opendata->o_arg.open_flags;
1484 	fmode_t fmode = opendata->o_arg.fmode;
1485 	enum open_claim_type4 claim = opendata->o_arg.claim;
1486 	nfs4_stateid stateid;
1487 	int ret = -EAGAIN;
1488 
1489 	for (;;) {
1490 		spin_lock(&state->owner->so_lock);
1491 		if (can_open_cached(state, fmode, open_mode)) {
1492 			update_open_stateflags(state, fmode);
1493 			spin_unlock(&state->owner->so_lock);
1494 			goto out_return_state;
1495 		}
1496 		spin_unlock(&state->owner->so_lock);
1497 		rcu_read_lock();
1498 		delegation = rcu_dereference(nfsi->delegation);
1499 		if (!can_open_delegated(delegation, fmode, claim)) {
1500 			rcu_read_unlock();
1501 			break;
1502 		}
1503 		/* Save the delegation */
1504 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1505 		rcu_read_unlock();
1506 		nfs_release_seqid(opendata->o_arg.seqid);
1507 		if (!opendata->is_recover) {
1508 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1509 			if (ret != 0)
1510 				goto out;
1511 		}
1512 		ret = -EAGAIN;
1513 
1514 		/* Try to update the stateid using the delegation */
1515 		if (update_open_stateid(state, NULL, &stateid, fmode))
1516 			goto out_return_state;
1517 	}
1518 out:
1519 	return ERR_PTR(ret);
1520 out_return_state:
1521 	atomic_inc(&state->count);
1522 	return state;
1523 }
1524 
1525 static void
1526 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1527 {
1528 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1529 	struct nfs_delegation *delegation;
1530 	int delegation_flags = 0;
1531 
1532 	rcu_read_lock();
1533 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1534 	if (delegation)
1535 		delegation_flags = delegation->flags;
1536 	rcu_read_unlock();
1537 	switch (data->o_arg.claim) {
1538 	default:
1539 		break;
1540 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1541 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1542 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1543 				   "returning a delegation for "
1544 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1545 				   clp->cl_hostname);
1546 		return;
1547 	}
1548 	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1549 		nfs_inode_set_delegation(state->inode,
1550 					 data->owner->so_cred,
1551 					 &data->o_res);
1552 	else
1553 		nfs_inode_reclaim_delegation(state->inode,
1554 					     data->owner->so_cred,
1555 					     &data->o_res);
1556 }
1557 
1558 /*
1559  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1560  * and update the nfs4_state.
1561  */
1562 static struct nfs4_state *
1563 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1564 {
1565 	struct inode *inode = data->state->inode;
1566 	struct nfs4_state *state = data->state;
1567 	int ret;
1568 
1569 	if (!data->rpc_done) {
1570 		if (data->rpc_status) {
1571 			ret = data->rpc_status;
1572 			goto err;
1573 		}
1574 		/* cached opens have already been processed */
1575 		goto update;
1576 	}
1577 
1578 	ret = nfs_refresh_inode(inode, &data->f_attr);
1579 	if (ret)
1580 		goto err;
1581 
1582 	if (data->o_res.delegation_type != 0)
1583 		nfs4_opendata_check_deleg(data, state);
1584 update:
1585 	update_open_stateid(state, &data->o_res.stateid, NULL,
1586 			    data->o_arg.fmode);
1587 	atomic_inc(&state->count);
1588 
1589 	return state;
1590 err:
1591 	return ERR_PTR(ret);
1592 
1593 }
1594 
1595 static struct nfs4_state *
1596 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1597 {
1598 	struct inode *inode;
1599 	struct nfs4_state *state = NULL;
1600 	int ret;
1601 
1602 	if (!data->rpc_done) {
1603 		state = nfs4_try_open_cached(data);
1604 		trace_nfs4_cached_open(data->state);
1605 		goto out;
1606 	}
1607 
1608 	ret = -EAGAIN;
1609 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1610 		goto err;
1611 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1612 	ret = PTR_ERR(inode);
1613 	if (IS_ERR(inode))
1614 		goto err;
1615 	ret = -ENOMEM;
1616 	state = nfs4_get_open_state(inode, data->owner);
1617 	if (state == NULL)
1618 		goto err_put_inode;
1619 	if (data->o_res.delegation_type != 0)
1620 		nfs4_opendata_check_deleg(data, state);
1621 	update_open_stateid(state, &data->o_res.stateid, NULL,
1622 			data->o_arg.fmode);
1623 	iput(inode);
1624 out:
1625 	nfs_release_seqid(data->o_arg.seqid);
1626 	return state;
1627 err_put_inode:
1628 	iput(inode);
1629 err:
1630 	return ERR_PTR(ret);
1631 }
1632 
1633 static struct nfs4_state *
1634 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1635 {
1636 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1637 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1638 	return _nfs4_opendata_to_nfs4_state(data);
1639 }
1640 
1641 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1642 {
1643 	struct nfs_inode *nfsi = NFS_I(state->inode);
1644 	struct nfs_open_context *ctx;
1645 
1646 	spin_lock(&state->inode->i_lock);
1647 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1648 		if (ctx->state != state)
1649 			continue;
1650 		get_nfs_open_context(ctx);
1651 		spin_unlock(&state->inode->i_lock);
1652 		return ctx;
1653 	}
1654 	spin_unlock(&state->inode->i_lock);
1655 	return ERR_PTR(-ENOENT);
1656 }
1657 
1658 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1659 		struct nfs4_state *state, enum open_claim_type4 claim)
1660 {
1661 	struct nfs4_opendata *opendata;
1662 
1663 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1664 			NULL, NULL, claim, GFP_NOFS);
1665 	if (opendata == NULL)
1666 		return ERR_PTR(-ENOMEM);
1667 	opendata->state = state;
1668 	atomic_inc(&state->count);
1669 	return opendata;
1670 }
1671 
1672 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1673 		fmode_t fmode)
1674 {
1675 	struct nfs4_state *newstate;
1676 	int ret;
1677 
1678 	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1679 		return 0;
1680 	opendata->o_arg.open_flags = 0;
1681 	opendata->o_arg.fmode = fmode;
1682 	opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1683 			NFS_SB(opendata->dentry->d_sb),
1684 			fmode, 0);
1685 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1686 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1687 	nfs4_init_opendata_res(opendata);
1688 	ret = _nfs4_recover_proc_open(opendata);
1689 	if (ret != 0)
1690 		return ret;
1691 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1692 	if (IS_ERR(newstate))
1693 		return PTR_ERR(newstate);
1694 	if (newstate != opendata->state)
1695 		ret = -ESTALE;
1696 	nfs4_close_state(newstate, fmode);
1697 	return ret;
1698 }
1699 
1700 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1701 {
1702 	int ret;
1703 
1704 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1705 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1706 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1707 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1708 	/* memory barrier prior to reading state->n_* */
1709 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1710 	clear_bit(NFS_OPEN_STATE, &state->flags);
1711 	smp_rmb();
1712 	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1713 	if (ret != 0)
1714 		return ret;
1715 	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1716 	if (ret != 0)
1717 		return ret;
1718 	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1719 	if (ret != 0)
1720 		return ret;
1721 	/*
1722 	 * We may have performed cached opens for all three recoveries.
1723 	 * Check if we need to update the current stateid.
1724 	 */
1725 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1726 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1727 		write_seqlock(&state->seqlock);
1728 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1729 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1730 		write_sequnlock(&state->seqlock);
1731 	}
1732 	return 0;
1733 }
1734 
1735 /*
1736  * OPEN_RECLAIM:
1737  * 	reclaim state on the server after a reboot.
1738  */
1739 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1740 {
1741 	struct nfs_delegation *delegation;
1742 	struct nfs4_opendata *opendata;
1743 	fmode_t delegation_type = 0;
1744 	int status;
1745 
1746 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1747 			NFS4_OPEN_CLAIM_PREVIOUS);
1748 	if (IS_ERR(opendata))
1749 		return PTR_ERR(opendata);
1750 	rcu_read_lock();
1751 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1752 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1753 		delegation_type = delegation->type;
1754 	rcu_read_unlock();
1755 	opendata->o_arg.u.delegation_type = delegation_type;
1756 	status = nfs4_open_recover(opendata, state);
1757 	nfs4_opendata_put(opendata);
1758 	return status;
1759 }
1760 
1761 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1762 {
1763 	struct nfs_server *server = NFS_SERVER(state->inode);
1764 	struct nfs4_exception exception = { };
1765 	int err;
1766 	do {
1767 		err = _nfs4_do_open_reclaim(ctx, state);
1768 		trace_nfs4_open_reclaim(ctx, 0, err);
1769 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1770 			continue;
1771 		if (err != -NFS4ERR_DELAY)
1772 			break;
1773 		nfs4_handle_exception(server, err, &exception);
1774 	} while (exception.retry);
1775 	return err;
1776 }
1777 
1778 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1779 {
1780 	struct nfs_open_context *ctx;
1781 	int ret;
1782 
1783 	ctx = nfs4_state_find_open_context(state);
1784 	if (IS_ERR(ctx))
1785 		return -EAGAIN;
1786 	ret = nfs4_do_open_reclaim(ctx, state);
1787 	put_nfs_open_context(ctx);
1788 	return ret;
1789 }
1790 
1791 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1792 {
1793 	switch (err) {
1794 		default:
1795 			printk(KERN_ERR "NFS: %s: unhandled error "
1796 					"%d.\n", __func__, err);
1797 		case 0:
1798 		case -ENOENT:
1799 		case -EAGAIN:
1800 		case -ESTALE:
1801 			break;
1802 		case -NFS4ERR_BADSESSION:
1803 		case -NFS4ERR_BADSLOT:
1804 		case -NFS4ERR_BAD_HIGH_SLOT:
1805 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1806 		case -NFS4ERR_DEADSESSION:
1807 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1808 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1809 			return -EAGAIN;
1810 		case -NFS4ERR_STALE_CLIENTID:
1811 		case -NFS4ERR_STALE_STATEID:
1812 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1813 		case -NFS4ERR_EXPIRED:
1814 			/* Don't recall a delegation if it was lost */
1815 			nfs4_schedule_lease_recovery(server->nfs_client);
1816 			return -EAGAIN;
1817 		case -NFS4ERR_MOVED:
1818 			nfs4_schedule_migration_recovery(server);
1819 			return -EAGAIN;
1820 		case -NFS4ERR_LEASE_MOVED:
1821 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1822 			return -EAGAIN;
1823 		case -NFS4ERR_DELEG_REVOKED:
1824 		case -NFS4ERR_ADMIN_REVOKED:
1825 		case -NFS4ERR_BAD_STATEID:
1826 		case -NFS4ERR_OPENMODE:
1827 			nfs_inode_find_state_and_recover(state->inode,
1828 					stateid);
1829 			nfs4_schedule_stateid_recovery(server, state);
1830 			return -EAGAIN;
1831 		case -NFS4ERR_DELAY:
1832 		case -NFS4ERR_GRACE:
1833 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1834 			ssleep(1);
1835 			return -EAGAIN;
1836 		case -ENOMEM:
1837 		case -NFS4ERR_DENIED:
1838 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1839 			return 0;
1840 	}
1841 	return err;
1842 }
1843 
1844 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1845 		struct nfs4_state *state, const nfs4_stateid *stateid,
1846 		fmode_t type)
1847 {
1848 	struct nfs_server *server = NFS_SERVER(state->inode);
1849 	struct nfs4_opendata *opendata;
1850 	int err = 0;
1851 
1852 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1853 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1854 	if (IS_ERR(opendata))
1855 		return PTR_ERR(opendata);
1856 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1857 	write_seqlock(&state->seqlock);
1858 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1859 	write_sequnlock(&state->seqlock);
1860 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1861 	switch (type & (FMODE_READ|FMODE_WRITE)) {
1862 	case FMODE_READ|FMODE_WRITE:
1863 	case FMODE_WRITE:
1864 		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1865 		if (err)
1866 			break;
1867 		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1868 		if (err)
1869 			break;
1870 	case FMODE_READ:
1871 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
1872 	}
1873 	nfs4_opendata_put(opendata);
1874 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1875 }
1876 
1877 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1878 {
1879 	struct nfs4_opendata *data = calldata;
1880 
1881 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1882 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1883 }
1884 
1885 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1886 {
1887 	struct nfs4_opendata *data = calldata;
1888 
1889 	nfs40_sequence_done(task, &data->c_res.seq_res);
1890 
1891 	data->rpc_status = task->tk_status;
1892 	if (data->rpc_status == 0) {
1893 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1894 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1895 		renew_lease(data->o_res.server, data->timestamp);
1896 		data->rpc_done = 1;
1897 	}
1898 }
1899 
1900 static void nfs4_open_confirm_release(void *calldata)
1901 {
1902 	struct nfs4_opendata *data = calldata;
1903 	struct nfs4_state *state = NULL;
1904 
1905 	/* If this request hasn't been cancelled, do nothing */
1906 	if (data->cancelled == 0)
1907 		goto out_free;
1908 	/* In case of error, no cleanup! */
1909 	if (!data->rpc_done)
1910 		goto out_free;
1911 	state = nfs4_opendata_to_nfs4_state(data);
1912 	if (!IS_ERR(state))
1913 		nfs4_close_state(state, data->o_arg.fmode);
1914 out_free:
1915 	nfs4_opendata_put(data);
1916 }
1917 
1918 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1919 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1920 	.rpc_call_done = nfs4_open_confirm_done,
1921 	.rpc_release = nfs4_open_confirm_release,
1922 };
1923 
1924 /*
1925  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1926  */
1927 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1928 {
1929 	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1930 	struct rpc_task *task;
1931 	struct  rpc_message msg = {
1932 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1933 		.rpc_argp = &data->c_arg,
1934 		.rpc_resp = &data->c_res,
1935 		.rpc_cred = data->owner->so_cred,
1936 	};
1937 	struct rpc_task_setup task_setup_data = {
1938 		.rpc_client = server->client,
1939 		.rpc_message = &msg,
1940 		.callback_ops = &nfs4_open_confirm_ops,
1941 		.callback_data = data,
1942 		.workqueue = nfsiod_workqueue,
1943 		.flags = RPC_TASK_ASYNC,
1944 	};
1945 	int status;
1946 
1947 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1948 	kref_get(&data->kref);
1949 	data->rpc_done = 0;
1950 	data->rpc_status = 0;
1951 	data->timestamp = jiffies;
1952 	if (data->is_recover)
1953 		nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1954 	task = rpc_run_task(&task_setup_data);
1955 	if (IS_ERR(task))
1956 		return PTR_ERR(task);
1957 	status = nfs4_wait_for_completion_rpc_task(task);
1958 	if (status != 0) {
1959 		data->cancelled = 1;
1960 		smp_wmb();
1961 	} else
1962 		status = data->rpc_status;
1963 	rpc_put_task(task);
1964 	return status;
1965 }
1966 
1967 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1968 {
1969 	struct nfs4_opendata *data = calldata;
1970 	struct nfs4_state_owner *sp = data->owner;
1971 	struct nfs_client *clp = sp->so_server->nfs_client;
1972 	enum open_claim_type4 claim = data->o_arg.claim;
1973 
1974 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1975 		goto out_wait;
1976 	/*
1977 	 * Check if we still need to send an OPEN call, or if we can use
1978 	 * a delegation instead.
1979 	 */
1980 	if (data->state != NULL) {
1981 		struct nfs_delegation *delegation;
1982 
1983 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1984 			goto out_no_action;
1985 		rcu_read_lock();
1986 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1987 		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1988 			goto unlock_no_action;
1989 		rcu_read_unlock();
1990 	}
1991 	/* Update client id. */
1992 	data->o_arg.clientid = clp->cl_clientid;
1993 	switch (claim) {
1994 	default:
1995 		break;
1996 	case NFS4_OPEN_CLAIM_PREVIOUS:
1997 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1998 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1999 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2000 	case NFS4_OPEN_CLAIM_FH:
2001 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2002 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2003 	}
2004 	data->timestamp = jiffies;
2005 	if (nfs4_setup_sequence(data->o_arg.server,
2006 				&data->o_arg.seq_args,
2007 				&data->o_res.seq_res,
2008 				task) != 0)
2009 		nfs_release_seqid(data->o_arg.seqid);
2010 
2011 	/* Set the create mode (note dependency on the session type) */
2012 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2013 	if (data->o_arg.open_flags & O_EXCL) {
2014 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2015 		if (nfs4_has_persistent_session(clp))
2016 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2017 		else if (clp->cl_mvops->minor_version > 0)
2018 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2019 	}
2020 	return;
2021 unlock_no_action:
2022 	trace_nfs4_cached_open(data->state);
2023 	rcu_read_unlock();
2024 out_no_action:
2025 	task->tk_action = NULL;
2026 out_wait:
2027 	nfs4_sequence_done(task, &data->o_res.seq_res);
2028 }
2029 
2030 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2031 {
2032 	struct nfs4_opendata *data = calldata;
2033 
2034 	data->rpc_status = task->tk_status;
2035 
2036 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2037 		return;
2038 
2039 	if (task->tk_status == 0) {
2040 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2041 			switch (data->o_res.f_attr->mode & S_IFMT) {
2042 			case S_IFREG:
2043 				break;
2044 			case S_IFLNK:
2045 				data->rpc_status = -ELOOP;
2046 				break;
2047 			case S_IFDIR:
2048 				data->rpc_status = -EISDIR;
2049 				break;
2050 			default:
2051 				data->rpc_status = -ENOTDIR;
2052 			}
2053 		}
2054 		renew_lease(data->o_res.server, data->timestamp);
2055 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2056 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2057 	}
2058 	data->rpc_done = 1;
2059 }
2060 
2061 static void nfs4_open_release(void *calldata)
2062 {
2063 	struct nfs4_opendata *data = calldata;
2064 	struct nfs4_state *state = NULL;
2065 
2066 	/* If this request hasn't been cancelled, do nothing */
2067 	if (data->cancelled == 0)
2068 		goto out_free;
2069 	/* In case of error, no cleanup! */
2070 	if (data->rpc_status != 0 || !data->rpc_done)
2071 		goto out_free;
2072 	/* In case we need an open_confirm, no cleanup! */
2073 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2074 		goto out_free;
2075 	state = nfs4_opendata_to_nfs4_state(data);
2076 	if (!IS_ERR(state))
2077 		nfs4_close_state(state, data->o_arg.fmode);
2078 out_free:
2079 	nfs4_opendata_put(data);
2080 }
2081 
2082 static const struct rpc_call_ops nfs4_open_ops = {
2083 	.rpc_call_prepare = nfs4_open_prepare,
2084 	.rpc_call_done = nfs4_open_done,
2085 	.rpc_release = nfs4_open_release,
2086 };
2087 
2088 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2089 {
2090 	struct inode *dir = d_inode(data->dir);
2091 	struct nfs_server *server = NFS_SERVER(dir);
2092 	struct nfs_openargs *o_arg = &data->o_arg;
2093 	struct nfs_openres *o_res = &data->o_res;
2094 	struct rpc_task *task;
2095 	struct rpc_message msg = {
2096 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2097 		.rpc_argp = o_arg,
2098 		.rpc_resp = o_res,
2099 		.rpc_cred = data->owner->so_cred,
2100 	};
2101 	struct rpc_task_setup task_setup_data = {
2102 		.rpc_client = server->client,
2103 		.rpc_message = &msg,
2104 		.callback_ops = &nfs4_open_ops,
2105 		.callback_data = data,
2106 		.workqueue = nfsiod_workqueue,
2107 		.flags = RPC_TASK_ASYNC,
2108 	};
2109 	int status;
2110 
2111 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2112 	kref_get(&data->kref);
2113 	data->rpc_done = 0;
2114 	data->rpc_status = 0;
2115 	data->cancelled = 0;
2116 	data->is_recover = 0;
2117 	if (isrecover) {
2118 		nfs4_set_sequence_privileged(&o_arg->seq_args);
2119 		data->is_recover = 1;
2120 	}
2121 	task = rpc_run_task(&task_setup_data);
2122         if (IS_ERR(task))
2123                 return PTR_ERR(task);
2124         status = nfs4_wait_for_completion_rpc_task(task);
2125         if (status != 0) {
2126                 data->cancelled = 1;
2127                 smp_wmb();
2128         } else
2129                 status = data->rpc_status;
2130         rpc_put_task(task);
2131 
2132 	return status;
2133 }
2134 
2135 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2136 {
2137 	struct inode *dir = d_inode(data->dir);
2138 	struct nfs_openres *o_res = &data->o_res;
2139         int status;
2140 
2141 	status = nfs4_run_open_task(data, 1);
2142 	if (status != 0 || !data->rpc_done)
2143 		return status;
2144 
2145 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2146 
2147 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2148 		status = _nfs4_proc_open_confirm(data);
2149 		if (status != 0)
2150 			return status;
2151 	}
2152 
2153 	return status;
2154 }
2155 
2156 /*
2157  * Additional permission checks in order to distinguish between an
2158  * open for read, and an open for execute. This works around the
2159  * fact that NFSv4 OPEN treats read and execute permissions as being
2160  * the same.
2161  * Note that in the non-execute case, we want to turn off permission
2162  * checking if we just created a new file (POSIX open() semantics).
2163  */
2164 static int nfs4_opendata_access(struct rpc_cred *cred,
2165 				struct nfs4_opendata *opendata,
2166 				struct nfs4_state *state, fmode_t fmode,
2167 				int openflags)
2168 {
2169 	struct nfs_access_entry cache;
2170 	u32 mask;
2171 
2172 	/* access call failed or for some reason the server doesn't
2173 	 * support any access modes -- defer access call until later */
2174 	if (opendata->o_res.access_supported == 0)
2175 		return 0;
2176 
2177 	mask = 0;
2178 	/*
2179 	 * Use openflags to check for exec, because fmode won't
2180 	 * always have FMODE_EXEC set when file open for exec.
2181 	 */
2182 	if (openflags & __FMODE_EXEC) {
2183 		/* ONLY check for exec rights */
2184 		mask = MAY_EXEC;
2185 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2186 		mask = MAY_READ;
2187 
2188 	cache.cred = cred;
2189 	cache.jiffies = jiffies;
2190 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2191 	nfs_access_add_cache(state->inode, &cache);
2192 
2193 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2194 		return 0;
2195 
2196 	/* even though OPEN succeeded, access is denied. Close the file */
2197 	nfs4_close_state(state, fmode);
2198 	return -EACCES;
2199 }
2200 
2201 /*
2202  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2203  */
2204 static int _nfs4_proc_open(struct nfs4_opendata *data)
2205 {
2206 	struct inode *dir = d_inode(data->dir);
2207 	struct nfs_server *server = NFS_SERVER(dir);
2208 	struct nfs_openargs *o_arg = &data->o_arg;
2209 	struct nfs_openres *o_res = &data->o_res;
2210 	int status;
2211 
2212 	status = nfs4_run_open_task(data, 0);
2213 	if (!data->rpc_done)
2214 		return status;
2215 	if (status != 0) {
2216 		if (status == -NFS4ERR_BADNAME &&
2217 				!(o_arg->open_flags & O_CREAT))
2218 			return -ENOENT;
2219 		return status;
2220 	}
2221 
2222 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2223 
2224 	if (o_arg->open_flags & O_CREAT) {
2225 		update_changeattr(dir, &o_res->cinfo);
2226 		if (o_arg->open_flags & O_EXCL)
2227 			data->file_created = 1;
2228 		else if (o_res->cinfo.before != o_res->cinfo.after)
2229 			data->file_created = 1;
2230 	}
2231 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2232 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2233 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2234 		status = _nfs4_proc_open_confirm(data);
2235 		if (status != 0)
2236 			return status;
2237 	}
2238 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2239 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2240 	return 0;
2241 }
2242 
2243 static int nfs4_recover_expired_lease(struct nfs_server *server)
2244 {
2245 	return nfs4_client_recover_expired_lease(server->nfs_client);
2246 }
2247 
2248 /*
2249  * OPEN_EXPIRED:
2250  * 	reclaim state on the server after a network partition.
2251  * 	Assumes caller holds the appropriate lock
2252  */
2253 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2254 {
2255 	struct nfs4_opendata *opendata;
2256 	int ret;
2257 
2258 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2259 			NFS4_OPEN_CLAIM_FH);
2260 	if (IS_ERR(opendata))
2261 		return PTR_ERR(opendata);
2262 	ret = nfs4_open_recover(opendata, state);
2263 	if (ret == -ESTALE)
2264 		d_drop(ctx->dentry);
2265 	nfs4_opendata_put(opendata);
2266 	return ret;
2267 }
2268 
2269 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2270 {
2271 	struct nfs_server *server = NFS_SERVER(state->inode);
2272 	struct nfs4_exception exception = { };
2273 	int err;
2274 
2275 	do {
2276 		err = _nfs4_open_expired(ctx, state);
2277 		trace_nfs4_open_expired(ctx, 0, err);
2278 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2279 			continue;
2280 		switch (err) {
2281 		default:
2282 			goto out;
2283 		case -NFS4ERR_GRACE:
2284 		case -NFS4ERR_DELAY:
2285 			nfs4_handle_exception(server, err, &exception);
2286 			err = 0;
2287 		}
2288 	} while (exception.retry);
2289 out:
2290 	return err;
2291 }
2292 
2293 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2294 {
2295 	struct nfs_open_context *ctx;
2296 	int ret;
2297 
2298 	ctx = nfs4_state_find_open_context(state);
2299 	if (IS_ERR(ctx))
2300 		return -EAGAIN;
2301 	ret = nfs4_do_open_expired(ctx, state);
2302 	put_nfs_open_context(ctx);
2303 	return ret;
2304 }
2305 
2306 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2307 {
2308 	nfs_remove_bad_delegation(state->inode);
2309 	write_seqlock(&state->seqlock);
2310 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2311 	write_sequnlock(&state->seqlock);
2312 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2313 }
2314 
2315 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2316 {
2317 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2318 		nfs_finish_clear_delegation_stateid(state);
2319 }
2320 
2321 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2322 {
2323 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2324 	nfs40_clear_delegation_stateid(state);
2325 	return nfs4_open_expired(sp, state);
2326 }
2327 
2328 #if defined(CONFIG_NFS_V4_1)
2329 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2330 {
2331 	struct nfs_server *server = NFS_SERVER(state->inode);
2332 	nfs4_stateid stateid;
2333 	struct nfs_delegation *delegation;
2334 	struct rpc_cred *cred;
2335 	int status;
2336 
2337 	/* Get the delegation credential for use by test/free_stateid */
2338 	rcu_read_lock();
2339 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2340 	if (delegation == NULL) {
2341 		rcu_read_unlock();
2342 		return;
2343 	}
2344 
2345 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2346 	cred = get_rpccred(delegation->cred);
2347 	rcu_read_unlock();
2348 	status = nfs41_test_stateid(server, &stateid, cred);
2349 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2350 
2351 	if (status != NFS_OK) {
2352 		/* Free the stateid unless the server explicitly
2353 		 * informs us the stateid is unrecognized. */
2354 		if (status != -NFS4ERR_BAD_STATEID)
2355 			nfs41_free_stateid(server, &stateid, cred);
2356 		nfs_finish_clear_delegation_stateid(state);
2357 	}
2358 
2359 	put_rpccred(cred);
2360 }
2361 
2362 /**
2363  * nfs41_check_open_stateid - possibly free an open stateid
2364  *
2365  * @state: NFSv4 state for an inode
2366  *
2367  * Returns NFS_OK if recovery for this stateid is now finished.
2368  * Otherwise a negative NFS4ERR value is returned.
2369  */
2370 static int nfs41_check_open_stateid(struct nfs4_state *state)
2371 {
2372 	struct nfs_server *server = NFS_SERVER(state->inode);
2373 	nfs4_stateid *stateid = &state->open_stateid;
2374 	struct rpc_cred *cred = state->owner->so_cred;
2375 	int status;
2376 
2377 	/* If a state reset has been done, test_stateid is unneeded */
2378 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2379 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2380 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2381 		return -NFS4ERR_BAD_STATEID;
2382 
2383 	status = nfs41_test_stateid(server, stateid, cred);
2384 	trace_nfs4_test_open_stateid(state, NULL, status);
2385 	if (status != NFS_OK) {
2386 		/* Free the stateid unless the server explicitly
2387 		 * informs us the stateid is unrecognized. */
2388 		if (status != -NFS4ERR_BAD_STATEID)
2389 			nfs41_free_stateid(server, stateid, cred);
2390 
2391 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2392 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2393 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2394 		clear_bit(NFS_OPEN_STATE, &state->flags);
2395 	}
2396 	return status;
2397 }
2398 
2399 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2400 {
2401 	int status;
2402 
2403 	nfs41_check_delegation_stateid(state);
2404 	status = nfs41_check_open_stateid(state);
2405 	if (status != NFS_OK)
2406 		status = nfs4_open_expired(sp, state);
2407 	return status;
2408 }
2409 #endif
2410 
2411 /*
2412  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2413  * fields corresponding to attributes that were used to store the verifier.
2414  * Make sure we clobber those fields in the later setattr call
2415  */
2416 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2417 				struct iattr *sattr, struct nfs4_label **label)
2418 {
2419 	const u32 *attrset = opendata->o_res.attrset;
2420 
2421 	if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2422 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2423 		sattr->ia_valid |= ATTR_ATIME;
2424 
2425 	if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2426 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2427 		sattr->ia_valid |= ATTR_MTIME;
2428 
2429 	/* Except MODE, it seems harmless of setting twice. */
2430 	if ((attrset[1] & FATTR4_WORD1_MODE))
2431 		sattr->ia_valid &= ~ATTR_MODE;
2432 
2433 	if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2434 		*label = NULL;
2435 }
2436 
2437 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2438 		fmode_t fmode,
2439 		int flags,
2440 		struct nfs_open_context *ctx)
2441 {
2442 	struct nfs4_state_owner *sp = opendata->owner;
2443 	struct nfs_server *server = sp->so_server;
2444 	struct dentry *dentry;
2445 	struct nfs4_state *state;
2446 	unsigned int seq;
2447 	int ret;
2448 
2449 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2450 
2451 	ret = _nfs4_proc_open(opendata);
2452 	if (ret != 0)
2453 		goto out;
2454 
2455 	state = nfs4_opendata_to_nfs4_state(opendata);
2456 	ret = PTR_ERR(state);
2457 	if (IS_ERR(state))
2458 		goto out;
2459 	if (server->caps & NFS_CAP_POSIX_LOCK)
2460 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2461 
2462 	dentry = opendata->dentry;
2463 	if (d_really_is_negative(dentry)) {
2464 		struct dentry *alias;
2465 		d_drop(dentry);
2466 		alias = d_exact_alias(dentry, state->inode);
2467 		if (!alias)
2468 			alias = d_splice_alias(igrab(state->inode), dentry);
2469 		/* d_splice_alias() can't fail here - it's a non-directory */
2470 		if (alias) {
2471 			dput(ctx->dentry);
2472 			ctx->dentry = dentry = alias;
2473 		}
2474 		nfs_set_verifier(dentry,
2475 				nfs_save_change_attribute(d_inode(opendata->dir)));
2476 	}
2477 
2478 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2479 	if (ret != 0)
2480 		goto out;
2481 
2482 	ctx->state = state;
2483 	if (d_inode(dentry) == state->inode) {
2484 		nfs_inode_attach_open_context(ctx);
2485 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2486 			nfs4_schedule_stateid_recovery(server, state);
2487 	}
2488 out:
2489 	return ret;
2490 }
2491 
2492 /*
2493  * Returns a referenced nfs4_state
2494  */
2495 static int _nfs4_do_open(struct inode *dir,
2496 			struct nfs_open_context *ctx,
2497 			int flags,
2498 			struct iattr *sattr,
2499 			struct nfs4_label *label,
2500 			int *opened)
2501 {
2502 	struct nfs4_state_owner  *sp;
2503 	struct nfs4_state     *state = NULL;
2504 	struct nfs_server       *server = NFS_SERVER(dir);
2505 	struct nfs4_opendata *opendata;
2506 	struct dentry *dentry = ctx->dentry;
2507 	struct rpc_cred *cred = ctx->cred;
2508 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2509 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2510 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2511 	struct nfs4_label *olabel = NULL;
2512 	int status;
2513 
2514 	/* Protect against reboot recovery conflicts */
2515 	status = -ENOMEM;
2516 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2517 	if (sp == NULL) {
2518 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2519 		goto out_err;
2520 	}
2521 	status = nfs4_recover_expired_lease(server);
2522 	if (status != 0)
2523 		goto err_put_state_owner;
2524 	if (d_really_is_positive(dentry))
2525 		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2526 	status = -ENOMEM;
2527 	if (d_really_is_positive(dentry))
2528 		claim = NFS4_OPEN_CLAIM_FH;
2529 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2530 			label, claim, GFP_KERNEL);
2531 	if (opendata == NULL)
2532 		goto err_put_state_owner;
2533 
2534 	if (label) {
2535 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2536 		if (IS_ERR(olabel)) {
2537 			status = PTR_ERR(olabel);
2538 			goto err_opendata_put;
2539 		}
2540 	}
2541 
2542 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2543 		if (!opendata->f_attr.mdsthreshold) {
2544 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2545 			if (!opendata->f_attr.mdsthreshold)
2546 				goto err_free_label;
2547 		}
2548 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2549 	}
2550 	if (d_really_is_positive(dentry))
2551 		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2552 
2553 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2554 	if (status != 0)
2555 		goto err_free_label;
2556 	state = ctx->state;
2557 
2558 	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2559 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2560 		nfs4_exclusive_attrset(opendata, sattr, &label);
2561 
2562 		nfs_fattr_init(opendata->o_res.f_attr);
2563 		status = nfs4_do_setattr(state->inode, cred,
2564 				opendata->o_res.f_attr, sattr,
2565 				state, label, olabel);
2566 		if (status == 0) {
2567 			nfs_setattr_update_inode(state->inode, sattr,
2568 					opendata->o_res.f_attr);
2569 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2570 		}
2571 	}
2572 	if (opened && opendata->file_created)
2573 		*opened |= FILE_CREATED;
2574 
2575 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2576 		*ctx_th = opendata->f_attr.mdsthreshold;
2577 		opendata->f_attr.mdsthreshold = NULL;
2578 	}
2579 
2580 	nfs4_label_free(olabel);
2581 
2582 	nfs4_opendata_put(opendata);
2583 	nfs4_put_state_owner(sp);
2584 	return 0;
2585 err_free_label:
2586 	nfs4_label_free(olabel);
2587 err_opendata_put:
2588 	nfs4_opendata_put(opendata);
2589 err_put_state_owner:
2590 	nfs4_put_state_owner(sp);
2591 out_err:
2592 	return status;
2593 }
2594 
2595 
2596 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2597 					struct nfs_open_context *ctx,
2598 					int flags,
2599 					struct iattr *sattr,
2600 					struct nfs4_label *label,
2601 					int *opened)
2602 {
2603 	struct nfs_server *server = NFS_SERVER(dir);
2604 	struct nfs4_exception exception = { };
2605 	struct nfs4_state *res;
2606 	int status;
2607 
2608 	do {
2609 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2610 		res = ctx->state;
2611 		trace_nfs4_open_file(ctx, flags, status);
2612 		if (status == 0)
2613 			break;
2614 		/* NOTE: BAD_SEQID means the server and client disagree about the
2615 		 * book-keeping w.r.t. state-changing operations
2616 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2617 		 * It is actually a sign of a bug on the client or on the server.
2618 		 *
2619 		 * If we receive a BAD_SEQID error in the particular case of
2620 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2621 		 * have unhashed the old state_owner for us, and that we can
2622 		 * therefore safely retry using a new one. We should still warn
2623 		 * the user though...
2624 		 */
2625 		if (status == -NFS4ERR_BAD_SEQID) {
2626 			pr_warn_ratelimited("NFS: v4 server %s "
2627 					" returned a bad sequence-id error!\n",
2628 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2629 			exception.retry = 1;
2630 			continue;
2631 		}
2632 		/*
2633 		 * BAD_STATEID on OPEN means that the server cancelled our
2634 		 * state before it received the OPEN_CONFIRM.
2635 		 * Recover by retrying the request as per the discussion
2636 		 * on Page 181 of RFC3530.
2637 		 */
2638 		if (status == -NFS4ERR_BAD_STATEID) {
2639 			exception.retry = 1;
2640 			continue;
2641 		}
2642 		if (status == -EAGAIN) {
2643 			/* We must have found a delegation */
2644 			exception.retry = 1;
2645 			continue;
2646 		}
2647 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2648 			continue;
2649 		res = ERR_PTR(nfs4_handle_exception(server,
2650 					status, &exception));
2651 	} while (exception.retry);
2652 	return res;
2653 }
2654 
2655 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2656 			    struct nfs_fattr *fattr, struct iattr *sattr,
2657 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2658 			    struct nfs4_label *olabel)
2659 {
2660 	struct nfs_server *server = NFS_SERVER(inode);
2661         struct nfs_setattrargs  arg = {
2662                 .fh             = NFS_FH(inode),
2663                 .iap            = sattr,
2664 		.server		= server,
2665 		.bitmask = server->attr_bitmask,
2666 		.label		= ilabel,
2667         };
2668         struct nfs_setattrres  res = {
2669 		.fattr		= fattr,
2670 		.label		= olabel,
2671 		.server		= server,
2672         };
2673         struct rpc_message msg = {
2674 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2675 		.rpc_argp	= &arg,
2676 		.rpc_resp	= &res,
2677 		.rpc_cred	= cred,
2678         };
2679 	unsigned long timestamp = jiffies;
2680 	fmode_t fmode;
2681 	bool truncate;
2682 	int status;
2683 
2684 	arg.bitmask = nfs4_bitmask(server, ilabel);
2685 	if (ilabel)
2686 		arg.bitmask = nfs4_bitmask(server, olabel);
2687 
2688 	nfs_fattr_init(fattr);
2689 
2690 	/* Servers should only apply open mode checks for file size changes */
2691 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2692 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2693 
2694 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2695 		/* Use that stateid */
2696 	} else if (truncate && state != NULL) {
2697 		struct nfs_lockowner lockowner = {
2698 			.l_owner = current->files,
2699 			.l_pid = current->tgid,
2700 		};
2701 		if (!nfs4_valid_open_stateid(state))
2702 			return -EBADF;
2703 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2704 				&lockowner) == -EIO)
2705 			return -EBADF;
2706 	} else
2707 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2708 
2709 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2710 	if (status == 0 && state != NULL)
2711 		renew_lease(server, timestamp);
2712 	trace_nfs4_setattr(inode, &arg.stateid, status);
2713 	return status;
2714 }
2715 
2716 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2717 			   struct nfs_fattr *fattr, struct iattr *sattr,
2718 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2719 			   struct nfs4_label *olabel)
2720 {
2721 	struct nfs_server *server = NFS_SERVER(inode);
2722 	struct nfs4_exception exception = {
2723 		.state = state,
2724 		.inode = inode,
2725 	};
2726 	int err;
2727 	do {
2728 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2729 		switch (err) {
2730 		case -NFS4ERR_OPENMODE:
2731 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2732 				pr_warn_once("NFSv4: server %s is incorrectly "
2733 						"applying open mode checks to "
2734 						"a SETATTR that is not "
2735 						"changing file size.\n",
2736 						server->nfs_client->cl_hostname);
2737 			}
2738 			if (state && !(state->state & FMODE_WRITE)) {
2739 				err = -EBADF;
2740 				if (sattr->ia_valid & ATTR_OPEN)
2741 					err = -EACCES;
2742 				goto out;
2743 			}
2744 		}
2745 		err = nfs4_handle_exception(server, err, &exception);
2746 	} while (exception.retry);
2747 out:
2748 	return err;
2749 }
2750 
2751 static bool
2752 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2753 {
2754 	if (inode == NULL || !nfs_have_layout(inode))
2755 		return false;
2756 
2757 	return pnfs_wait_on_layoutreturn(inode, task);
2758 }
2759 
2760 struct nfs4_closedata {
2761 	struct inode *inode;
2762 	struct nfs4_state *state;
2763 	struct nfs_closeargs arg;
2764 	struct nfs_closeres res;
2765 	struct nfs_fattr fattr;
2766 	unsigned long timestamp;
2767 	bool roc;
2768 	u32 roc_barrier;
2769 };
2770 
2771 static void nfs4_free_closedata(void *data)
2772 {
2773 	struct nfs4_closedata *calldata = data;
2774 	struct nfs4_state_owner *sp = calldata->state->owner;
2775 	struct super_block *sb = calldata->state->inode->i_sb;
2776 
2777 	if (calldata->roc)
2778 		pnfs_roc_release(calldata->state->inode);
2779 	nfs4_put_open_state(calldata->state);
2780 	nfs_free_seqid(calldata->arg.seqid);
2781 	nfs4_put_state_owner(sp);
2782 	nfs_sb_deactive(sb);
2783 	kfree(calldata);
2784 }
2785 
2786 static void nfs4_close_done(struct rpc_task *task, void *data)
2787 {
2788 	struct nfs4_closedata *calldata = data;
2789 	struct nfs4_state *state = calldata->state;
2790 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2791 	nfs4_stateid *res_stateid = NULL;
2792 
2793 	dprintk("%s: begin!\n", __func__);
2794 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2795 		return;
2796 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2797         /* hmm. we are done with the inode, and in the process of freeing
2798 	 * the state_owner. we keep this around to process errors
2799 	 */
2800 	switch (task->tk_status) {
2801 		case 0:
2802 			res_stateid = &calldata->res.stateid;
2803 			if (calldata->roc)
2804 				pnfs_roc_set_barrier(state->inode,
2805 						     calldata->roc_barrier);
2806 			renew_lease(server, calldata->timestamp);
2807 			break;
2808 		case -NFS4ERR_ADMIN_REVOKED:
2809 		case -NFS4ERR_STALE_STATEID:
2810 		case -NFS4ERR_OLD_STATEID:
2811 		case -NFS4ERR_BAD_STATEID:
2812 		case -NFS4ERR_EXPIRED:
2813 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2814 						&state->open_stateid)) {
2815 				rpc_restart_call_prepare(task);
2816 				goto out_release;
2817 			}
2818 			if (calldata->arg.fmode == 0)
2819 				break;
2820 		default:
2821 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2822 				rpc_restart_call_prepare(task);
2823 				goto out_release;
2824 			}
2825 	}
2826 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
2827 			res_stateid, calldata->arg.fmode);
2828 out_release:
2829 	nfs_release_seqid(calldata->arg.seqid);
2830 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2831 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2832 }
2833 
2834 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2835 {
2836 	struct nfs4_closedata *calldata = data;
2837 	struct nfs4_state *state = calldata->state;
2838 	struct inode *inode = calldata->inode;
2839 	bool is_rdonly, is_wronly, is_rdwr;
2840 	int call_close = 0;
2841 
2842 	dprintk("%s: begin!\n", __func__);
2843 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2844 		goto out_wait;
2845 
2846 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2847 	spin_lock(&state->owner->so_lock);
2848 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2849 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2850 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2851 	nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2852 	/* Calculate the change in open mode */
2853 	calldata->arg.fmode = 0;
2854 	if (state->n_rdwr == 0) {
2855 		if (state->n_rdonly == 0)
2856 			call_close |= is_rdonly;
2857 		else if (is_rdonly)
2858 			calldata->arg.fmode |= FMODE_READ;
2859 		if (state->n_wronly == 0)
2860 			call_close |= is_wronly;
2861 		else if (is_wronly)
2862 			calldata->arg.fmode |= FMODE_WRITE;
2863 	} else if (is_rdwr)
2864 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2865 
2866 	if (calldata->arg.fmode == 0)
2867 		call_close |= is_rdwr;
2868 
2869 	if (!nfs4_valid_open_stateid(state))
2870 		call_close = 0;
2871 	spin_unlock(&state->owner->so_lock);
2872 
2873 	if (!call_close) {
2874 		/* Note: exit _without_ calling nfs4_close_done */
2875 		goto out_no_action;
2876 	}
2877 
2878 	if (nfs4_wait_on_layoutreturn(inode, task)) {
2879 		nfs_release_seqid(calldata->arg.seqid);
2880 		goto out_wait;
2881 	}
2882 
2883 	if (calldata->arg.fmode == 0)
2884 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2885 	if (calldata->roc)
2886 		pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2887 
2888 	calldata->arg.share_access =
2889 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2890 				calldata->arg.fmode, 0);
2891 
2892 	nfs_fattr_init(calldata->res.fattr);
2893 	calldata->timestamp = jiffies;
2894 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2895 				&calldata->arg.seq_args,
2896 				&calldata->res.seq_res,
2897 				task) != 0)
2898 		nfs_release_seqid(calldata->arg.seqid);
2899 	dprintk("%s: done!\n", __func__);
2900 	return;
2901 out_no_action:
2902 	task->tk_action = NULL;
2903 out_wait:
2904 	nfs4_sequence_done(task, &calldata->res.seq_res);
2905 }
2906 
2907 static const struct rpc_call_ops nfs4_close_ops = {
2908 	.rpc_call_prepare = nfs4_close_prepare,
2909 	.rpc_call_done = nfs4_close_done,
2910 	.rpc_release = nfs4_free_closedata,
2911 };
2912 
2913 static bool nfs4_roc(struct inode *inode)
2914 {
2915 	if (!nfs_have_layout(inode))
2916 		return false;
2917 	return pnfs_roc(inode);
2918 }
2919 
2920 /*
2921  * It is possible for data to be read/written from a mem-mapped file
2922  * after the sys_close call (which hits the vfs layer as a flush).
2923  * This means that we can't safely call nfsv4 close on a file until
2924  * the inode is cleared. This in turn means that we are not good
2925  * NFSv4 citizens - we do not indicate to the server to update the file's
2926  * share state even when we are done with one of the three share
2927  * stateid's in the inode.
2928  *
2929  * NOTE: Caller must be holding the sp->so_owner semaphore!
2930  */
2931 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2932 {
2933 	struct nfs_server *server = NFS_SERVER(state->inode);
2934 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2935 	struct nfs4_closedata *calldata;
2936 	struct nfs4_state_owner *sp = state->owner;
2937 	struct rpc_task *task;
2938 	struct rpc_message msg = {
2939 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2940 		.rpc_cred = state->owner->so_cred,
2941 	};
2942 	struct rpc_task_setup task_setup_data = {
2943 		.rpc_client = server->client,
2944 		.rpc_message = &msg,
2945 		.callback_ops = &nfs4_close_ops,
2946 		.workqueue = nfsiod_workqueue,
2947 		.flags = RPC_TASK_ASYNC,
2948 	};
2949 	int status = -ENOMEM;
2950 
2951 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2952 		&task_setup_data.rpc_client, &msg);
2953 
2954 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2955 	if (calldata == NULL)
2956 		goto out;
2957 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2958 	calldata->inode = state->inode;
2959 	calldata->state = state;
2960 	calldata->arg.fh = NFS_FH(state->inode);
2961 	/* Serialization for the sequence id */
2962 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2963 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2964 	if (IS_ERR(calldata->arg.seqid))
2965 		goto out_free_calldata;
2966 	calldata->arg.fmode = 0;
2967 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2968 	calldata->res.fattr = &calldata->fattr;
2969 	calldata->res.seqid = calldata->arg.seqid;
2970 	calldata->res.server = server;
2971 	calldata->roc = nfs4_roc(state->inode);
2972 	nfs_sb_active(calldata->inode->i_sb);
2973 
2974 	msg.rpc_argp = &calldata->arg;
2975 	msg.rpc_resp = &calldata->res;
2976 	task_setup_data.callback_data = calldata;
2977 	task = rpc_run_task(&task_setup_data);
2978 	if (IS_ERR(task))
2979 		return PTR_ERR(task);
2980 	status = 0;
2981 	if (wait)
2982 		status = rpc_wait_for_completion_task(task);
2983 	rpc_put_task(task);
2984 	return status;
2985 out_free_calldata:
2986 	kfree(calldata);
2987 out:
2988 	nfs4_put_open_state(state);
2989 	nfs4_put_state_owner(sp);
2990 	return status;
2991 }
2992 
2993 static struct inode *
2994 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2995 		int open_flags, struct iattr *attr, int *opened)
2996 {
2997 	struct nfs4_state *state;
2998 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2999 
3000 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3001 
3002 	/* Protect against concurrent sillydeletes */
3003 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3004 
3005 	nfs4_label_release_security(label);
3006 
3007 	if (IS_ERR(state))
3008 		return ERR_CAST(state);
3009 	return state->inode;
3010 }
3011 
3012 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3013 {
3014 	if (ctx->state == NULL)
3015 		return;
3016 	if (is_sync)
3017 		nfs4_close_sync(ctx->state, ctx->mode);
3018 	else
3019 		nfs4_close_state(ctx->state, ctx->mode);
3020 }
3021 
3022 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3023 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3024 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3025 
3026 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3027 {
3028 	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3029 	struct nfs4_server_caps_arg args = {
3030 		.fhandle = fhandle,
3031 		.bitmask = bitmask,
3032 	};
3033 	struct nfs4_server_caps_res res = {};
3034 	struct rpc_message msg = {
3035 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3036 		.rpc_argp = &args,
3037 		.rpc_resp = &res,
3038 	};
3039 	int status;
3040 
3041 	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3042 		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3043 		     FATTR4_WORD0_LINK_SUPPORT |
3044 		     FATTR4_WORD0_SYMLINK_SUPPORT |
3045 		     FATTR4_WORD0_ACLSUPPORT;
3046 	if (minorversion)
3047 		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3048 
3049 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3050 	if (status == 0) {
3051 		/* Sanity check the server answers */
3052 		switch (minorversion) {
3053 		case 0:
3054 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3055 			res.attr_bitmask[2] = 0;
3056 			break;
3057 		case 1:
3058 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3059 			break;
3060 		case 2:
3061 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3062 		}
3063 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3064 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3065 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3066 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3067 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3068 				NFS_CAP_CTIME|NFS_CAP_MTIME|
3069 				NFS_CAP_SECURITY_LABEL);
3070 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3071 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3072 			server->caps |= NFS_CAP_ACLS;
3073 		if (res.has_links != 0)
3074 			server->caps |= NFS_CAP_HARDLINKS;
3075 		if (res.has_symlinks != 0)
3076 			server->caps |= NFS_CAP_SYMLINKS;
3077 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3078 			server->caps |= NFS_CAP_FILEID;
3079 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3080 			server->caps |= NFS_CAP_MODE;
3081 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3082 			server->caps |= NFS_CAP_NLINK;
3083 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3084 			server->caps |= NFS_CAP_OWNER;
3085 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3086 			server->caps |= NFS_CAP_OWNER_GROUP;
3087 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3088 			server->caps |= NFS_CAP_ATIME;
3089 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3090 			server->caps |= NFS_CAP_CTIME;
3091 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3092 			server->caps |= NFS_CAP_MTIME;
3093 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3094 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3095 			server->caps |= NFS_CAP_SECURITY_LABEL;
3096 #endif
3097 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3098 				sizeof(server->attr_bitmask));
3099 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3100 
3101 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3102 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3103 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3104 		server->cache_consistency_bitmask[2] = 0;
3105 		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3106 			sizeof(server->exclcreat_bitmask));
3107 		server->acl_bitmask = res.acl_bitmask;
3108 		server->fh_expire_type = res.fh_expire_type;
3109 	}
3110 
3111 	return status;
3112 }
3113 
3114 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3115 {
3116 	struct nfs4_exception exception = { };
3117 	int err;
3118 	do {
3119 		err = nfs4_handle_exception(server,
3120 				_nfs4_server_capabilities(server, fhandle),
3121 				&exception);
3122 	} while (exception.retry);
3123 	return err;
3124 }
3125 
3126 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3127 		struct nfs_fsinfo *info)
3128 {
3129 	u32 bitmask[3];
3130 	struct nfs4_lookup_root_arg args = {
3131 		.bitmask = bitmask,
3132 	};
3133 	struct nfs4_lookup_res res = {
3134 		.server = server,
3135 		.fattr = info->fattr,
3136 		.fh = fhandle,
3137 	};
3138 	struct rpc_message msg = {
3139 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3140 		.rpc_argp = &args,
3141 		.rpc_resp = &res,
3142 	};
3143 
3144 	bitmask[0] = nfs4_fattr_bitmap[0];
3145 	bitmask[1] = nfs4_fattr_bitmap[1];
3146 	/*
3147 	 * Process the label in the upcoming getfattr
3148 	 */
3149 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3150 
3151 	nfs_fattr_init(info->fattr);
3152 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3153 }
3154 
3155 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3156 		struct nfs_fsinfo *info)
3157 {
3158 	struct nfs4_exception exception = { };
3159 	int err;
3160 	do {
3161 		err = _nfs4_lookup_root(server, fhandle, info);
3162 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3163 		switch (err) {
3164 		case 0:
3165 		case -NFS4ERR_WRONGSEC:
3166 			goto out;
3167 		default:
3168 			err = nfs4_handle_exception(server, err, &exception);
3169 		}
3170 	} while (exception.retry);
3171 out:
3172 	return err;
3173 }
3174 
3175 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3176 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3177 {
3178 	struct rpc_auth_create_args auth_args = {
3179 		.pseudoflavor = flavor,
3180 	};
3181 	struct rpc_auth *auth;
3182 	int ret;
3183 
3184 	auth = rpcauth_create(&auth_args, server->client);
3185 	if (IS_ERR(auth)) {
3186 		ret = -EACCES;
3187 		goto out;
3188 	}
3189 	ret = nfs4_lookup_root(server, fhandle, info);
3190 out:
3191 	return ret;
3192 }
3193 
3194 /*
3195  * Retry pseudoroot lookup with various security flavors.  We do this when:
3196  *
3197  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3198  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3199  *
3200  * Returns zero on success, or a negative NFS4ERR value, or a
3201  * negative errno value.
3202  */
3203 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3204 			      struct nfs_fsinfo *info)
3205 {
3206 	/* Per 3530bis 15.33.5 */
3207 	static const rpc_authflavor_t flav_array[] = {
3208 		RPC_AUTH_GSS_KRB5P,
3209 		RPC_AUTH_GSS_KRB5I,
3210 		RPC_AUTH_GSS_KRB5,
3211 		RPC_AUTH_UNIX,			/* courtesy */
3212 		RPC_AUTH_NULL,
3213 	};
3214 	int status = -EPERM;
3215 	size_t i;
3216 
3217 	if (server->auth_info.flavor_len > 0) {
3218 		/* try each flavor specified by user */
3219 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3220 			status = nfs4_lookup_root_sec(server, fhandle, info,
3221 						server->auth_info.flavors[i]);
3222 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3223 				continue;
3224 			break;
3225 		}
3226 	} else {
3227 		/* no flavors specified by user, try default list */
3228 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3229 			status = nfs4_lookup_root_sec(server, fhandle, info,
3230 						      flav_array[i]);
3231 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3232 				continue;
3233 			break;
3234 		}
3235 	}
3236 
3237 	/*
3238 	 * -EACCESS could mean that the user doesn't have correct permissions
3239 	 * to access the mount.  It could also mean that we tried to mount
3240 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3241 	 * existing mount programs don't handle -EACCES very well so it should
3242 	 * be mapped to -EPERM instead.
3243 	 */
3244 	if (status == -EACCES)
3245 		status = -EPERM;
3246 	return status;
3247 }
3248 
3249 static int nfs4_do_find_root_sec(struct nfs_server *server,
3250 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3251 {
3252 	int mv = server->nfs_client->cl_minorversion;
3253 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3254 }
3255 
3256 /**
3257  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3258  * @server: initialized nfs_server handle
3259  * @fhandle: we fill in the pseudo-fs root file handle
3260  * @info: we fill in an FSINFO struct
3261  * @auth_probe: probe the auth flavours
3262  *
3263  * Returns zero on success, or a negative errno.
3264  */
3265 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3266 			 struct nfs_fsinfo *info,
3267 			 bool auth_probe)
3268 {
3269 	int status = 0;
3270 
3271 	if (!auth_probe)
3272 		status = nfs4_lookup_root(server, fhandle, info);
3273 
3274 	if (auth_probe || status == NFS4ERR_WRONGSEC)
3275 		status = nfs4_do_find_root_sec(server, fhandle, info);
3276 
3277 	if (status == 0)
3278 		status = nfs4_server_capabilities(server, fhandle);
3279 	if (status == 0)
3280 		status = nfs4_do_fsinfo(server, fhandle, info);
3281 
3282 	return nfs4_map_errors(status);
3283 }
3284 
3285 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3286 			      struct nfs_fsinfo *info)
3287 {
3288 	int error;
3289 	struct nfs_fattr *fattr = info->fattr;
3290 	struct nfs4_label *label = NULL;
3291 
3292 	error = nfs4_server_capabilities(server, mntfh);
3293 	if (error < 0) {
3294 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3295 		return error;
3296 	}
3297 
3298 	label = nfs4_label_alloc(server, GFP_KERNEL);
3299 	if (IS_ERR(label))
3300 		return PTR_ERR(label);
3301 
3302 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3303 	if (error < 0) {
3304 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3305 		goto err_free_label;
3306 	}
3307 
3308 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3309 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3310 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3311 
3312 err_free_label:
3313 	nfs4_label_free(label);
3314 
3315 	return error;
3316 }
3317 
3318 /*
3319  * Get locations and (maybe) other attributes of a referral.
3320  * Note that we'll actually follow the referral later when
3321  * we detect fsid mismatch in inode revalidation
3322  */
3323 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3324 			     const struct qstr *name, struct nfs_fattr *fattr,
3325 			     struct nfs_fh *fhandle)
3326 {
3327 	int status = -ENOMEM;
3328 	struct page *page = NULL;
3329 	struct nfs4_fs_locations *locations = NULL;
3330 
3331 	page = alloc_page(GFP_KERNEL);
3332 	if (page == NULL)
3333 		goto out;
3334 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3335 	if (locations == NULL)
3336 		goto out;
3337 
3338 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3339 	if (status != 0)
3340 		goto out;
3341 
3342 	/*
3343 	 * If the fsid didn't change, this is a migration event, not a
3344 	 * referral.  Cause us to drop into the exception handler, which
3345 	 * will kick off migration recovery.
3346 	 */
3347 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3348 		dprintk("%s: server did not return a different fsid for"
3349 			" a referral at %s\n", __func__, name->name);
3350 		status = -NFS4ERR_MOVED;
3351 		goto out;
3352 	}
3353 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3354 	nfs_fixup_referral_attributes(&locations->fattr);
3355 
3356 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3357 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3358 	memset(fhandle, 0, sizeof(struct nfs_fh));
3359 out:
3360 	if (page)
3361 		__free_page(page);
3362 	kfree(locations);
3363 	return status;
3364 }
3365 
3366 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3367 				struct nfs_fattr *fattr, struct nfs4_label *label)
3368 {
3369 	struct nfs4_getattr_arg args = {
3370 		.fh = fhandle,
3371 		.bitmask = server->attr_bitmask,
3372 	};
3373 	struct nfs4_getattr_res res = {
3374 		.fattr = fattr,
3375 		.label = label,
3376 		.server = server,
3377 	};
3378 	struct rpc_message msg = {
3379 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3380 		.rpc_argp = &args,
3381 		.rpc_resp = &res,
3382 	};
3383 
3384 	args.bitmask = nfs4_bitmask(server, label);
3385 
3386 	nfs_fattr_init(fattr);
3387 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3388 }
3389 
3390 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3391 				struct nfs_fattr *fattr, struct nfs4_label *label)
3392 {
3393 	struct nfs4_exception exception = { };
3394 	int err;
3395 	do {
3396 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3397 		trace_nfs4_getattr(server, fhandle, fattr, err);
3398 		err = nfs4_handle_exception(server, err,
3399 				&exception);
3400 	} while (exception.retry);
3401 	return err;
3402 }
3403 
3404 /*
3405  * The file is not closed if it is opened due to the a request to change
3406  * the size of the file. The open call will not be needed once the
3407  * VFS layer lookup-intents are implemented.
3408  *
3409  * Close is called when the inode is destroyed.
3410  * If we haven't opened the file for O_WRONLY, we
3411  * need to in the size_change case to obtain a stateid.
3412  *
3413  * Got race?
3414  * Because OPEN is always done by name in nfsv4, it is
3415  * possible that we opened a different file by the same
3416  * name.  We can recognize this race condition, but we
3417  * can't do anything about it besides returning an error.
3418  *
3419  * This will be fixed with VFS changes (lookup-intent).
3420  */
3421 static int
3422 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3423 		  struct iattr *sattr)
3424 {
3425 	struct inode *inode = d_inode(dentry);
3426 	struct rpc_cred *cred = NULL;
3427 	struct nfs4_state *state = NULL;
3428 	struct nfs4_label *label = NULL;
3429 	int status;
3430 
3431 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3432 	    sattr->ia_valid & ATTR_SIZE &&
3433 	    sattr->ia_size < i_size_read(inode))
3434 		pnfs_commit_and_return_layout(inode);
3435 
3436 	nfs_fattr_init(fattr);
3437 
3438 	/* Deal with open(O_TRUNC) */
3439 	if (sattr->ia_valid & ATTR_OPEN)
3440 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3441 
3442 	/* Optimization: if the end result is no change, don't RPC */
3443 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3444 		return 0;
3445 
3446 	/* Search for an existing open(O_WRITE) file */
3447 	if (sattr->ia_valid & ATTR_FILE) {
3448 		struct nfs_open_context *ctx;
3449 
3450 		ctx = nfs_file_open_context(sattr->ia_file);
3451 		if (ctx) {
3452 			cred = ctx->cred;
3453 			state = ctx->state;
3454 		}
3455 	}
3456 
3457 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3458 	if (IS_ERR(label))
3459 		return PTR_ERR(label);
3460 
3461 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3462 	if (status == 0) {
3463 		nfs_setattr_update_inode(inode, sattr, fattr);
3464 		nfs_setsecurity(inode, fattr, label);
3465 	}
3466 	nfs4_label_free(label);
3467 	return status;
3468 }
3469 
3470 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3471 		const struct qstr *name, struct nfs_fh *fhandle,
3472 		struct nfs_fattr *fattr, struct nfs4_label *label)
3473 {
3474 	struct nfs_server *server = NFS_SERVER(dir);
3475 	int		       status;
3476 	struct nfs4_lookup_arg args = {
3477 		.bitmask = server->attr_bitmask,
3478 		.dir_fh = NFS_FH(dir),
3479 		.name = name,
3480 	};
3481 	struct nfs4_lookup_res res = {
3482 		.server = server,
3483 		.fattr = fattr,
3484 		.label = label,
3485 		.fh = fhandle,
3486 	};
3487 	struct rpc_message msg = {
3488 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3489 		.rpc_argp = &args,
3490 		.rpc_resp = &res,
3491 	};
3492 
3493 	args.bitmask = nfs4_bitmask(server, label);
3494 
3495 	nfs_fattr_init(fattr);
3496 
3497 	dprintk("NFS call  lookup %s\n", name->name);
3498 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3499 	dprintk("NFS reply lookup: %d\n", status);
3500 	return status;
3501 }
3502 
3503 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3504 {
3505 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3506 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3507 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3508 	fattr->nlink = 2;
3509 }
3510 
3511 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3512 				   struct qstr *name, struct nfs_fh *fhandle,
3513 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3514 {
3515 	struct nfs4_exception exception = { };
3516 	struct rpc_clnt *client = *clnt;
3517 	int err;
3518 	do {
3519 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3520 		trace_nfs4_lookup(dir, name, err);
3521 		switch (err) {
3522 		case -NFS4ERR_BADNAME:
3523 			err = -ENOENT;
3524 			goto out;
3525 		case -NFS4ERR_MOVED:
3526 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3527 			if (err == -NFS4ERR_MOVED)
3528 				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3529 			goto out;
3530 		case -NFS4ERR_WRONGSEC:
3531 			err = -EPERM;
3532 			if (client != *clnt)
3533 				goto out;
3534 			client = nfs4_negotiate_security(client, dir, name);
3535 			if (IS_ERR(client))
3536 				return PTR_ERR(client);
3537 
3538 			exception.retry = 1;
3539 			break;
3540 		default:
3541 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3542 		}
3543 	} while (exception.retry);
3544 
3545 out:
3546 	if (err == 0)
3547 		*clnt = client;
3548 	else if (client != *clnt)
3549 		rpc_shutdown_client(client);
3550 
3551 	return err;
3552 }
3553 
3554 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3555 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3556 			    struct nfs4_label *label)
3557 {
3558 	int status;
3559 	struct rpc_clnt *client = NFS_CLIENT(dir);
3560 
3561 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3562 	if (client != NFS_CLIENT(dir)) {
3563 		rpc_shutdown_client(client);
3564 		nfs_fixup_secinfo_attributes(fattr);
3565 	}
3566 	return status;
3567 }
3568 
3569 struct rpc_clnt *
3570 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3571 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3572 {
3573 	struct rpc_clnt *client = NFS_CLIENT(dir);
3574 	int status;
3575 
3576 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3577 	if (status < 0)
3578 		return ERR_PTR(status);
3579 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3580 }
3581 
3582 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3583 {
3584 	struct nfs_server *server = NFS_SERVER(inode);
3585 	struct nfs4_accessargs args = {
3586 		.fh = NFS_FH(inode),
3587 		.bitmask = server->cache_consistency_bitmask,
3588 	};
3589 	struct nfs4_accessres res = {
3590 		.server = server,
3591 	};
3592 	struct rpc_message msg = {
3593 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3594 		.rpc_argp = &args,
3595 		.rpc_resp = &res,
3596 		.rpc_cred = entry->cred,
3597 	};
3598 	int mode = entry->mask;
3599 	int status = 0;
3600 
3601 	/*
3602 	 * Determine which access bits we want to ask for...
3603 	 */
3604 	if (mode & MAY_READ)
3605 		args.access |= NFS4_ACCESS_READ;
3606 	if (S_ISDIR(inode->i_mode)) {
3607 		if (mode & MAY_WRITE)
3608 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3609 		if (mode & MAY_EXEC)
3610 			args.access |= NFS4_ACCESS_LOOKUP;
3611 	} else {
3612 		if (mode & MAY_WRITE)
3613 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3614 		if (mode & MAY_EXEC)
3615 			args.access |= NFS4_ACCESS_EXECUTE;
3616 	}
3617 
3618 	res.fattr = nfs_alloc_fattr();
3619 	if (res.fattr == NULL)
3620 		return -ENOMEM;
3621 
3622 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3623 	if (!status) {
3624 		nfs_access_set_mask(entry, res.access);
3625 		nfs_refresh_inode(inode, res.fattr);
3626 	}
3627 	nfs_free_fattr(res.fattr);
3628 	return status;
3629 }
3630 
3631 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3632 {
3633 	struct nfs4_exception exception = { };
3634 	int err;
3635 	do {
3636 		err = _nfs4_proc_access(inode, entry);
3637 		trace_nfs4_access(inode, err);
3638 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3639 				&exception);
3640 	} while (exception.retry);
3641 	return err;
3642 }
3643 
3644 /*
3645  * TODO: For the time being, we don't try to get any attributes
3646  * along with any of the zero-copy operations READ, READDIR,
3647  * READLINK, WRITE.
3648  *
3649  * In the case of the first three, we want to put the GETATTR
3650  * after the read-type operation -- this is because it is hard
3651  * to predict the length of a GETATTR response in v4, and thus
3652  * align the READ data correctly.  This means that the GETATTR
3653  * may end up partially falling into the page cache, and we should
3654  * shift it into the 'tail' of the xdr_buf before processing.
3655  * To do this efficiently, we need to know the total length
3656  * of data received, which doesn't seem to be available outside
3657  * of the RPC layer.
3658  *
3659  * In the case of WRITE, we also want to put the GETATTR after
3660  * the operation -- in this case because we want to make sure
3661  * we get the post-operation mtime and size.
3662  *
3663  * Both of these changes to the XDR layer would in fact be quite
3664  * minor, but I decided to leave them for a subsequent patch.
3665  */
3666 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3667 		unsigned int pgbase, unsigned int pglen)
3668 {
3669 	struct nfs4_readlink args = {
3670 		.fh       = NFS_FH(inode),
3671 		.pgbase	  = pgbase,
3672 		.pglen    = pglen,
3673 		.pages    = &page,
3674 	};
3675 	struct nfs4_readlink_res res;
3676 	struct rpc_message msg = {
3677 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3678 		.rpc_argp = &args,
3679 		.rpc_resp = &res,
3680 	};
3681 
3682 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3683 }
3684 
3685 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3686 		unsigned int pgbase, unsigned int pglen)
3687 {
3688 	struct nfs4_exception exception = { };
3689 	int err;
3690 	do {
3691 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3692 		trace_nfs4_readlink(inode, err);
3693 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3694 				&exception);
3695 	} while (exception.retry);
3696 	return err;
3697 }
3698 
3699 /*
3700  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3701  */
3702 static int
3703 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3704 		 int flags)
3705 {
3706 	struct nfs4_label l, *ilabel = NULL;
3707 	struct nfs_open_context *ctx;
3708 	struct nfs4_state *state;
3709 	int status = 0;
3710 
3711 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3712 	if (IS_ERR(ctx))
3713 		return PTR_ERR(ctx);
3714 
3715 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3716 
3717 	sattr->ia_mode &= ~current_umask();
3718 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3719 	if (IS_ERR(state)) {
3720 		status = PTR_ERR(state);
3721 		goto out;
3722 	}
3723 out:
3724 	nfs4_label_release_security(ilabel);
3725 	put_nfs_open_context(ctx);
3726 	return status;
3727 }
3728 
3729 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3730 {
3731 	struct nfs_server *server = NFS_SERVER(dir);
3732 	struct nfs_removeargs args = {
3733 		.fh = NFS_FH(dir),
3734 		.name = *name,
3735 	};
3736 	struct nfs_removeres res = {
3737 		.server = server,
3738 	};
3739 	struct rpc_message msg = {
3740 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3741 		.rpc_argp = &args,
3742 		.rpc_resp = &res,
3743 	};
3744 	int status;
3745 
3746 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3747 	if (status == 0)
3748 		update_changeattr(dir, &res.cinfo);
3749 	return status;
3750 }
3751 
3752 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3753 {
3754 	struct nfs4_exception exception = { };
3755 	int err;
3756 	do {
3757 		err = _nfs4_proc_remove(dir, name);
3758 		trace_nfs4_remove(dir, name, err);
3759 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3760 				&exception);
3761 	} while (exception.retry);
3762 	return err;
3763 }
3764 
3765 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3766 {
3767 	struct nfs_server *server = NFS_SERVER(dir);
3768 	struct nfs_removeargs *args = msg->rpc_argp;
3769 	struct nfs_removeres *res = msg->rpc_resp;
3770 
3771 	res->server = server;
3772 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3773 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3774 
3775 	nfs_fattr_init(res->dir_attr);
3776 }
3777 
3778 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3779 {
3780 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3781 			&data->args.seq_args,
3782 			&data->res.seq_res,
3783 			task);
3784 }
3785 
3786 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3787 {
3788 	struct nfs_unlinkdata *data = task->tk_calldata;
3789 	struct nfs_removeres *res = &data->res;
3790 
3791 	if (!nfs4_sequence_done(task, &res->seq_res))
3792 		return 0;
3793 	if (nfs4_async_handle_error(task, res->server, NULL,
3794 				    &data->timeout) == -EAGAIN)
3795 		return 0;
3796 	update_changeattr(dir, &res->cinfo);
3797 	return 1;
3798 }
3799 
3800 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3801 {
3802 	struct nfs_server *server = NFS_SERVER(dir);
3803 	struct nfs_renameargs *arg = msg->rpc_argp;
3804 	struct nfs_renameres *res = msg->rpc_resp;
3805 
3806 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3807 	res->server = server;
3808 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3809 }
3810 
3811 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3812 {
3813 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3814 			&data->args.seq_args,
3815 			&data->res.seq_res,
3816 			task);
3817 }
3818 
3819 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3820 				 struct inode *new_dir)
3821 {
3822 	struct nfs_renamedata *data = task->tk_calldata;
3823 	struct nfs_renameres *res = &data->res;
3824 
3825 	if (!nfs4_sequence_done(task, &res->seq_res))
3826 		return 0;
3827 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3828 		return 0;
3829 
3830 	update_changeattr(old_dir, &res->old_cinfo);
3831 	update_changeattr(new_dir, &res->new_cinfo);
3832 	return 1;
3833 }
3834 
3835 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3836 {
3837 	struct nfs_server *server = NFS_SERVER(inode);
3838 	struct nfs4_link_arg arg = {
3839 		.fh     = NFS_FH(inode),
3840 		.dir_fh = NFS_FH(dir),
3841 		.name   = name,
3842 		.bitmask = server->attr_bitmask,
3843 	};
3844 	struct nfs4_link_res res = {
3845 		.server = server,
3846 		.label = NULL,
3847 	};
3848 	struct rpc_message msg = {
3849 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3850 		.rpc_argp = &arg,
3851 		.rpc_resp = &res,
3852 	};
3853 	int status = -ENOMEM;
3854 
3855 	res.fattr = nfs_alloc_fattr();
3856 	if (res.fattr == NULL)
3857 		goto out;
3858 
3859 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3860 	if (IS_ERR(res.label)) {
3861 		status = PTR_ERR(res.label);
3862 		goto out;
3863 	}
3864 	arg.bitmask = nfs4_bitmask(server, res.label);
3865 
3866 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3867 	if (!status) {
3868 		update_changeattr(dir, &res.cinfo);
3869 		status = nfs_post_op_update_inode(inode, res.fattr);
3870 		if (!status)
3871 			nfs_setsecurity(inode, res.fattr, res.label);
3872 	}
3873 
3874 
3875 	nfs4_label_free(res.label);
3876 
3877 out:
3878 	nfs_free_fattr(res.fattr);
3879 	return status;
3880 }
3881 
3882 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3883 {
3884 	struct nfs4_exception exception = { };
3885 	int err;
3886 	do {
3887 		err = nfs4_handle_exception(NFS_SERVER(inode),
3888 				_nfs4_proc_link(inode, dir, name),
3889 				&exception);
3890 	} while (exception.retry);
3891 	return err;
3892 }
3893 
3894 struct nfs4_createdata {
3895 	struct rpc_message msg;
3896 	struct nfs4_create_arg arg;
3897 	struct nfs4_create_res res;
3898 	struct nfs_fh fh;
3899 	struct nfs_fattr fattr;
3900 	struct nfs4_label *label;
3901 };
3902 
3903 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3904 		struct qstr *name, struct iattr *sattr, u32 ftype)
3905 {
3906 	struct nfs4_createdata *data;
3907 
3908 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3909 	if (data != NULL) {
3910 		struct nfs_server *server = NFS_SERVER(dir);
3911 
3912 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3913 		if (IS_ERR(data->label))
3914 			goto out_free;
3915 
3916 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3917 		data->msg.rpc_argp = &data->arg;
3918 		data->msg.rpc_resp = &data->res;
3919 		data->arg.dir_fh = NFS_FH(dir);
3920 		data->arg.server = server;
3921 		data->arg.name = name;
3922 		data->arg.attrs = sattr;
3923 		data->arg.ftype = ftype;
3924 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3925 		data->res.server = server;
3926 		data->res.fh = &data->fh;
3927 		data->res.fattr = &data->fattr;
3928 		data->res.label = data->label;
3929 		nfs_fattr_init(data->res.fattr);
3930 	}
3931 	return data;
3932 out_free:
3933 	kfree(data);
3934 	return NULL;
3935 }
3936 
3937 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3938 {
3939 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3940 				    &data->arg.seq_args, &data->res.seq_res, 1);
3941 	if (status == 0) {
3942 		update_changeattr(dir, &data->res.dir_cinfo);
3943 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3944 	}
3945 	return status;
3946 }
3947 
3948 static void nfs4_free_createdata(struct nfs4_createdata *data)
3949 {
3950 	nfs4_label_free(data->label);
3951 	kfree(data);
3952 }
3953 
3954 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3955 		struct page *page, unsigned int len, struct iattr *sattr,
3956 		struct nfs4_label *label)
3957 {
3958 	struct nfs4_createdata *data;
3959 	int status = -ENAMETOOLONG;
3960 
3961 	if (len > NFS4_MAXPATHLEN)
3962 		goto out;
3963 
3964 	status = -ENOMEM;
3965 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3966 	if (data == NULL)
3967 		goto out;
3968 
3969 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3970 	data->arg.u.symlink.pages = &page;
3971 	data->arg.u.symlink.len = len;
3972 	data->arg.label = label;
3973 
3974 	status = nfs4_do_create(dir, dentry, data);
3975 
3976 	nfs4_free_createdata(data);
3977 out:
3978 	return status;
3979 }
3980 
3981 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3982 		struct page *page, unsigned int len, struct iattr *sattr)
3983 {
3984 	struct nfs4_exception exception = { };
3985 	struct nfs4_label l, *label = NULL;
3986 	int err;
3987 
3988 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3989 
3990 	do {
3991 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3992 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3993 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3994 				&exception);
3995 	} while (exception.retry);
3996 
3997 	nfs4_label_release_security(label);
3998 	return err;
3999 }
4000 
4001 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4002 		struct iattr *sattr, struct nfs4_label *label)
4003 {
4004 	struct nfs4_createdata *data;
4005 	int status = -ENOMEM;
4006 
4007 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4008 	if (data == NULL)
4009 		goto out;
4010 
4011 	data->arg.label = label;
4012 	status = nfs4_do_create(dir, dentry, data);
4013 
4014 	nfs4_free_createdata(data);
4015 out:
4016 	return status;
4017 }
4018 
4019 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4020 		struct iattr *sattr)
4021 {
4022 	struct nfs4_exception exception = { };
4023 	struct nfs4_label l, *label = NULL;
4024 	int err;
4025 
4026 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4027 
4028 	sattr->ia_mode &= ~current_umask();
4029 	do {
4030 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4031 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
4032 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4033 				&exception);
4034 	} while (exception.retry);
4035 	nfs4_label_release_security(label);
4036 
4037 	return err;
4038 }
4039 
4040 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4041 		u64 cookie, struct page **pages, unsigned int count, int plus)
4042 {
4043 	struct inode		*dir = d_inode(dentry);
4044 	struct nfs4_readdir_arg args = {
4045 		.fh = NFS_FH(dir),
4046 		.pages = pages,
4047 		.pgbase = 0,
4048 		.count = count,
4049 		.bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4050 		.plus = plus,
4051 	};
4052 	struct nfs4_readdir_res res;
4053 	struct rpc_message msg = {
4054 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4055 		.rpc_argp = &args,
4056 		.rpc_resp = &res,
4057 		.rpc_cred = cred,
4058 	};
4059 	int			status;
4060 
4061 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4062 			dentry,
4063 			(unsigned long long)cookie);
4064 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4065 	res.pgbase = args.pgbase;
4066 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4067 	if (status >= 0) {
4068 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4069 		status += args.pgbase;
4070 	}
4071 
4072 	nfs_invalidate_atime(dir);
4073 
4074 	dprintk("%s: returns %d\n", __func__, status);
4075 	return status;
4076 }
4077 
4078 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4079 		u64 cookie, struct page **pages, unsigned int count, int plus)
4080 {
4081 	struct nfs4_exception exception = { };
4082 	int err;
4083 	do {
4084 		err = _nfs4_proc_readdir(dentry, cred, cookie,
4085 				pages, count, plus);
4086 		trace_nfs4_readdir(d_inode(dentry), err);
4087 		err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4088 				&exception);
4089 	} while (exception.retry);
4090 	return err;
4091 }
4092 
4093 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4094 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4095 {
4096 	struct nfs4_createdata *data;
4097 	int mode = sattr->ia_mode;
4098 	int status = -ENOMEM;
4099 
4100 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4101 	if (data == NULL)
4102 		goto out;
4103 
4104 	if (S_ISFIFO(mode))
4105 		data->arg.ftype = NF4FIFO;
4106 	else if (S_ISBLK(mode)) {
4107 		data->arg.ftype = NF4BLK;
4108 		data->arg.u.device.specdata1 = MAJOR(rdev);
4109 		data->arg.u.device.specdata2 = MINOR(rdev);
4110 	}
4111 	else if (S_ISCHR(mode)) {
4112 		data->arg.ftype = NF4CHR;
4113 		data->arg.u.device.specdata1 = MAJOR(rdev);
4114 		data->arg.u.device.specdata2 = MINOR(rdev);
4115 	} else if (!S_ISSOCK(mode)) {
4116 		status = -EINVAL;
4117 		goto out_free;
4118 	}
4119 
4120 	data->arg.label = label;
4121 	status = nfs4_do_create(dir, dentry, data);
4122 out_free:
4123 	nfs4_free_createdata(data);
4124 out:
4125 	return status;
4126 }
4127 
4128 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4129 		struct iattr *sattr, dev_t rdev)
4130 {
4131 	struct nfs4_exception exception = { };
4132 	struct nfs4_label l, *label = NULL;
4133 	int err;
4134 
4135 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
4136 
4137 	sattr->ia_mode &= ~current_umask();
4138 	do {
4139 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4140 		trace_nfs4_mknod(dir, &dentry->d_name, err);
4141 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4142 				&exception);
4143 	} while (exception.retry);
4144 
4145 	nfs4_label_release_security(label);
4146 
4147 	return err;
4148 }
4149 
4150 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4151 		 struct nfs_fsstat *fsstat)
4152 {
4153 	struct nfs4_statfs_arg args = {
4154 		.fh = fhandle,
4155 		.bitmask = server->attr_bitmask,
4156 	};
4157 	struct nfs4_statfs_res res = {
4158 		.fsstat = fsstat,
4159 	};
4160 	struct rpc_message msg = {
4161 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4162 		.rpc_argp = &args,
4163 		.rpc_resp = &res,
4164 	};
4165 
4166 	nfs_fattr_init(fsstat->fattr);
4167 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4168 }
4169 
4170 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4171 {
4172 	struct nfs4_exception exception = { };
4173 	int err;
4174 	do {
4175 		err = nfs4_handle_exception(server,
4176 				_nfs4_proc_statfs(server, fhandle, fsstat),
4177 				&exception);
4178 	} while (exception.retry);
4179 	return err;
4180 }
4181 
4182 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4183 		struct nfs_fsinfo *fsinfo)
4184 {
4185 	struct nfs4_fsinfo_arg args = {
4186 		.fh = fhandle,
4187 		.bitmask = server->attr_bitmask,
4188 	};
4189 	struct nfs4_fsinfo_res res = {
4190 		.fsinfo = fsinfo,
4191 	};
4192 	struct rpc_message msg = {
4193 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4194 		.rpc_argp = &args,
4195 		.rpc_resp = &res,
4196 	};
4197 
4198 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4199 }
4200 
4201 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4202 {
4203 	struct nfs4_exception exception = { };
4204 	unsigned long now = jiffies;
4205 	int err;
4206 
4207 	do {
4208 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4209 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4210 		if (err == 0) {
4211 			struct nfs_client *clp = server->nfs_client;
4212 
4213 			spin_lock(&clp->cl_lock);
4214 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4215 			clp->cl_last_renewal = now;
4216 			spin_unlock(&clp->cl_lock);
4217 			break;
4218 		}
4219 		err = nfs4_handle_exception(server, err, &exception);
4220 	} while (exception.retry);
4221 	return err;
4222 }
4223 
4224 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4225 {
4226 	int error;
4227 
4228 	nfs_fattr_init(fsinfo->fattr);
4229 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4230 	if (error == 0) {
4231 		/* block layout checks this! */
4232 		server->pnfs_blksize = fsinfo->blksize;
4233 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4234 	}
4235 
4236 	return error;
4237 }
4238 
4239 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4240 		struct nfs_pathconf *pathconf)
4241 {
4242 	struct nfs4_pathconf_arg args = {
4243 		.fh = fhandle,
4244 		.bitmask = server->attr_bitmask,
4245 	};
4246 	struct nfs4_pathconf_res res = {
4247 		.pathconf = pathconf,
4248 	};
4249 	struct rpc_message msg = {
4250 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4251 		.rpc_argp = &args,
4252 		.rpc_resp = &res,
4253 	};
4254 
4255 	/* None of the pathconf attributes are mandatory to implement */
4256 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4257 		memset(pathconf, 0, sizeof(*pathconf));
4258 		return 0;
4259 	}
4260 
4261 	nfs_fattr_init(pathconf->fattr);
4262 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4263 }
4264 
4265 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4266 		struct nfs_pathconf *pathconf)
4267 {
4268 	struct nfs4_exception exception = { };
4269 	int err;
4270 
4271 	do {
4272 		err = nfs4_handle_exception(server,
4273 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4274 				&exception);
4275 	} while (exception.retry);
4276 	return err;
4277 }
4278 
4279 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4280 		const struct nfs_open_context *ctx,
4281 		const struct nfs_lock_context *l_ctx,
4282 		fmode_t fmode)
4283 {
4284 	const struct nfs_lockowner *lockowner = NULL;
4285 
4286 	if (l_ctx != NULL)
4287 		lockowner = &l_ctx->lockowner;
4288 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4289 }
4290 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4291 
4292 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4293 		const struct nfs_open_context *ctx,
4294 		const struct nfs_lock_context *l_ctx,
4295 		fmode_t fmode)
4296 {
4297 	nfs4_stateid current_stateid;
4298 
4299 	/* If the current stateid represents a lost lock, then exit */
4300 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4301 		return true;
4302 	return nfs4_stateid_match(stateid, &current_stateid);
4303 }
4304 
4305 static bool nfs4_error_stateid_expired(int err)
4306 {
4307 	switch (err) {
4308 	case -NFS4ERR_DELEG_REVOKED:
4309 	case -NFS4ERR_ADMIN_REVOKED:
4310 	case -NFS4ERR_BAD_STATEID:
4311 	case -NFS4ERR_STALE_STATEID:
4312 	case -NFS4ERR_OLD_STATEID:
4313 	case -NFS4ERR_OPENMODE:
4314 	case -NFS4ERR_EXPIRED:
4315 		return true;
4316 	}
4317 	return false;
4318 }
4319 
4320 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4321 {
4322 	nfs_invalidate_atime(hdr->inode);
4323 }
4324 
4325 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4326 {
4327 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4328 
4329 	trace_nfs4_read(hdr, task->tk_status);
4330 	if (nfs4_async_handle_error(task, server,
4331 				    hdr->args.context->state,
4332 				    NULL) == -EAGAIN) {
4333 		rpc_restart_call_prepare(task);
4334 		return -EAGAIN;
4335 	}
4336 
4337 	__nfs4_read_done_cb(hdr);
4338 	if (task->tk_status > 0)
4339 		renew_lease(server, hdr->timestamp);
4340 	return 0;
4341 }
4342 
4343 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4344 		struct nfs_pgio_args *args)
4345 {
4346 
4347 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4348 		nfs4_stateid_is_current(&args->stateid,
4349 				args->context,
4350 				args->lock_context,
4351 				FMODE_READ))
4352 		return false;
4353 	rpc_restart_call_prepare(task);
4354 	return true;
4355 }
4356 
4357 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4358 {
4359 
4360 	dprintk("--> %s\n", __func__);
4361 
4362 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4363 		return -EAGAIN;
4364 	if (nfs4_read_stateid_changed(task, &hdr->args))
4365 		return -EAGAIN;
4366 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4367 				    nfs4_read_done_cb(task, hdr);
4368 }
4369 
4370 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4371 				 struct rpc_message *msg)
4372 {
4373 	hdr->timestamp   = jiffies;
4374 	hdr->pgio_done_cb = nfs4_read_done_cb;
4375 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4376 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4377 }
4378 
4379 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4380 				      struct nfs_pgio_header *hdr)
4381 {
4382 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4383 			&hdr->args.seq_args,
4384 			&hdr->res.seq_res,
4385 			task))
4386 		return 0;
4387 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4388 				hdr->args.lock_context,
4389 				hdr->rw_ops->rw_mode) == -EIO)
4390 		return -EIO;
4391 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4392 		return -EIO;
4393 	return 0;
4394 }
4395 
4396 static int nfs4_write_done_cb(struct rpc_task *task,
4397 			      struct nfs_pgio_header *hdr)
4398 {
4399 	struct inode *inode = hdr->inode;
4400 
4401 	trace_nfs4_write(hdr, task->tk_status);
4402 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4403 				    hdr->args.context->state,
4404 				    NULL) == -EAGAIN) {
4405 		rpc_restart_call_prepare(task);
4406 		return -EAGAIN;
4407 	}
4408 	if (task->tk_status >= 0) {
4409 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4410 		nfs_writeback_update_inode(hdr);
4411 	}
4412 	return 0;
4413 }
4414 
4415 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4416 		struct nfs_pgio_args *args)
4417 {
4418 
4419 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4420 		nfs4_stateid_is_current(&args->stateid,
4421 				args->context,
4422 				args->lock_context,
4423 				FMODE_WRITE))
4424 		return false;
4425 	rpc_restart_call_prepare(task);
4426 	return true;
4427 }
4428 
4429 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4430 {
4431 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4432 		return -EAGAIN;
4433 	if (nfs4_write_stateid_changed(task, &hdr->args))
4434 		return -EAGAIN;
4435 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4436 		nfs4_write_done_cb(task, hdr);
4437 }
4438 
4439 static
4440 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4441 {
4442 	/* Don't request attributes for pNFS or O_DIRECT writes */
4443 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4444 		return false;
4445 	/* Otherwise, request attributes if and only if we don't hold
4446 	 * a delegation
4447 	 */
4448 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4449 }
4450 
4451 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4452 				  struct rpc_message *msg)
4453 {
4454 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4455 
4456 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4457 		hdr->args.bitmask = NULL;
4458 		hdr->res.fattr = NULL;
4459 	} else
4460 		hdr->args.bitmask = server->cache_consistency_bitmask;
4461 
4462 	if (!hdr->pgio_done_cb)
4463 		hdr->pgio_done_cb = nfs4_write_done_cb;
4464 	hdr->res.server = server;
4465 	hdr->timestamp   = jiffies;
4466 
4467 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4468 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4469 }
4470 
4471 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4472 {
4473 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4474 			&data->args.seq_args,
4475 			&data->res.seq_res,
4476 			task);
4477 }
4478 
4479 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4480 {
4481 	struct inode *inode = data->inode;
4482 
4483 	trace_nfs4_commit(data, task->tk_status);
4484 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4485 				    NULL, NULL) == -EAGAIN) {
4486 		rpc_restart_call_prepare(task);
4487 		return -EAGAIN;
4488 	}
4489 	return 0;
4490 }
4491 
4492 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4493 {
4494 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4495 		return -EAGAIN;
4496 	return data->commit_done_cb(task, data);
4497 }
4498 
4499 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4500 {
4501 	struct nfs_server *server = NFS_SERVER(data->inode);
4502 
4503 	if (data->commit_done_cb == NULL)
4504 		data->commit_done_cb = nfs4_commit_done_cb;
4505 	data->res.server = server;
4506 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4507 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4508 }
4509 
4510 struct nfs4_renewdata {
4511 	struct nfs_client	*client;
4512 	unsigned long		timestamp;
4513 };
4514 
4515 /*
4516  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4517  * standalone procedure for queueing an asynchronous RENEW.
4518  */
4519 static void nfs4_renew_release(void *calldata)
4520 {
4521 	struct nfs4_renewdata *data = calldata;
4522 	struct nfs_client *clp = data->client;
4523 
4524 	if (atomic_read(&clp->cl_count) > 1)
4525 		nfs4_schedule_state_renewal(clp);
4526 	nfs_put_client(clp);
4527 	kfree(data);
4528 }
4529 
4530 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4531 {
4532 	struct nfs4_renewdata *data = calldata;
4533 	struct nfs_client *clp = data->client;
4534 	unsigned long timestamp = data->timestamp;
4535 
4536 	trace_nfs4_renew_async(clp, task->tk_status);
4537 	switch (task->tk_status) {
4538 	case 0:
4539 		break;
4540 	case -NFS4ERR_LEASE_MOVED:
4541 		nfs4_schedule_lease_moved_recovery(clp);
4542 		break;
4543 	default:
4544 		/* Unless we're shutting down, schedule state recovery! */
4545 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4546 			return;
4547 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4548 			nfs4_schedule_lease_recovery(clp);
4549 			return;
4550 		}
4551 		nfs4_schedule_path_down_recovery(clp);
4552 	}
4553 	do_renew_lease(clp, timestamp);
4554 }
4555 
4556 static const struct rpc_call_ops nfs4_renew_ops = {
4557 	.rpc_call_done = nfs4_renew_done,
4558 	.rpc_release = nfs4_renew_release,
4559 };
4560 
4561 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4562 {
4563 	struct rpc_message msg = {
4564 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4565 		.rpc_argp	= clp,
4566 		.rpc_cred	= cred,
4567 	};
4568 	struct nfs4_renewdata *data;
4569 
4570 	if (renew_flags == 0)
4571 		return 0;
4572 	if (!atomic_inc_not_zero(&clp->cl_count))
4573 		return -EIO;
4574 	data = kmalloc(sizeof(*data), GFP_NOFS);
4575 	if (data == NULL)
4576 		return -ENOMEM;
4577 	data->client = clp;
4578 	data->timestamp = jiffies;
4579 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4580 			&nfs4_renew_ops, data);
4581 }
4582 
4583 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4584 {
4585 	struct rpc_message msg = {
4586 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4587 		.rpc_argp	= clp,
4588 		.rpc_cred	= cred,
4589 	};
4590 	unsigned long now = jiffies;
4591 	int status;
4592 
4593 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4594 	if (status < 0)
4595 		return status;
4596 	do_renew_lease(clp, now);
4597 	return 0;
4598 }
4599 
4600 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4601 {
4602 	return server->caps & NFS_CAP_ACLS;
4603 }
4604 
4605 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4606  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4607  * the stack.
4608  */
4609 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4610 
4611 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4612 		struct page **pages)
4613 {
4614 	struct page *newpage, **spages;
4615 	int rc = 0;
4616 	size_t len;
4617 	spages = pages;
4618 
4619 	do {
4620 		len = min_t(size_t, PAGE_SIZE, buflen);
4621 		newpage = alloc_page(GFP_KERNEL);
4622 
4623 		if (newpage == NULL)
4624 			goto unwind;
4625 		memcpy(page_address(newpage), buf, len);
4626                 buf += len;
4627                 buflen -= len;
4628 		*pages++ = newpage;
4629 		rc++;
4630 	} while (buflen != 0);
4631 
4632 	return rc;
4633 
4634 unwind:
4635 	for(; rc > 0; rc--)
4636 		__free_page(spages[rc-1]);
4637 	return -ENOMEM;
4638 }
4639 
4640 struct nfs4_cached_acl {
4641 	int cached;
4642 	size_t len;
4643 	char data[0];
4644 };
4645 
4646 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4647 {
4648 	struct nfs_inode *nfsi = NFS_I(inode);
4649 
4650 	spin_lock(&inode->i_lock);
4651 	kfree(nfsi->nfs4_acl);
4652 	nfsi->nfs4_acl = acl;
4653 	spin_unlock(&inode->i_lock);
4654 }
4655 
4656 static void nfs4_zap_acl_attr(struct inode *inode)
4657 {
4658 	nfs4_set_cached_acl(inode, NULL);
4659 }
4660 
4661 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4662 {
4663 	struct nfs_inode *nfsi = NFS_I(inode);
4664 	struct nfs4_cached_acl *acl;
4665 	int ret = -ENOENT;
4666 
4667 	spin_lock(&inode->i_lock);
4668 	acl = nfsi->nfs4_acl;
4669 	if (acl == NULL)
4670 		goto out;
4671 	if (buf == NULL) /* user is just asking for length */
4672 		goto out_len;
4673 	if (acl->cached == 0)
4674 		goto out;
4675 	ret = -ERANGE; /* see getxattr(2) man page */
4676 	if (acl->len > buflen)
4677 		goto out;
4678 	memcpy(buf, acl->data, acl->len);
4679 out_len:
4680 	ret = acl->len;
4681 out:
4682 	spin_unlock(&inode->i_lock);
4683 	return ret;
4684 }
4685 
4686 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4687 {
4688 	struct nfs4_cached_acl *acl;
4689 	size_t buflen = sizeof(*acl) + acl_len;
4690 
4691 	if (buflen <= PAGE_SIZE) {
4692 		acl = kmalloc(buflen, GFP_KERNEL);
4693 		if (acl == NULL)
4694 			goto out;
4695 		acl->cached = 1;
4696 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4697 	} else {
4698 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4699 		if (acl == NULL)
4700 			goto out;
4701 		acl->cached = 0;
4702 	}
4703 	acl->len = acl_len;
4704 out:
4705 	nfs4_set_cached_acl(inode, acl);
4706 }
4707 
4708 /*
4709  * The getxattr API returns the required buffer length when called with a
4710  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4711  * the required buf.  On a NULL buf, we send a page of data to the server
4712  * guessing that the ACL request can be serviced by a page. If so, we cache
4713  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4714  * the cache. If not so, we throw away the page, and cache the required
4715  * length. The next getxattr call will then produce another round trip to
4716  * the server, this time with the input buf of the required size.
4717  */
4718 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4719 {
4720 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4721 	struct nfs_getaclargs args = {
4722 		.fh = NFS_FH(inode),
4723 		.acl_pages = pages,
4724 		.acl_len = buflen,
4725 	};
4726 	struct nfs_getaclres res = {
4727 		.acl_len = buflen,
4728 	};
4729 	struct rpc_message msg = {
4730 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4731 		.rpc_argp = &args,
4732 		.rpc_resp = &res,
4733 	};
4734 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4735 	int ret = -ENOMEM, i;
4736 
4737 	/* As long as we're doing a round trip to the server anyway,
4738 	 * let's be prepared for a page of acl data. */
4739 	if (npages == 0)
4740 		npages = 1;
4741 	if (npages > ARRAY_SIZE(pages))
4742 		return -ERANGE;
4743 
4744 	for (i = 0; i < npages; i++) {
4745 		pages[i] = alloc_page(GFP_KERNEL);
4746 		if (!pages[i])
4747 			goto out_free;
4748 	}
4749 
4750 	/* for decoding across pages */
4751 	res.acl_scratch = alloc_page(GFP_KERNEL);
4752 	if (!res.acl_scratch)
4753 		goto out_free;
4754 
4755 	args.acl_len = npages * PAGE_SIZE;
4756 
4757 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4758 		__func__, buf, buflen, npages, args.acl_len);
4759 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4760 			     &msg, &args.seq_args, &res.seq_res, 0);
4761 	if (ret)
4762 		goto out_free;
4763 
4764 	/* Handle the case where the passed-in buffer is too short */
4765 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4766 		/* Did the user only issue a request for the acl length? */
4767 		if (buf == NULL)
4768 			goto out_ok;
4769 		ret = -ERANGE;
4770 		goto out_free;
4771 	}
4772 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4773 	if (buf) {
4774 		if (res.acl_len > buflen) {
4775 			ret = -ERANGE;
4776 			goto out_free;
4777 		}
4778 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4779 	}
4780 out_ok:
4781 	ret = res.acl_len;
4782 out_free:
4783 	for (i = 0; i < npages; i++)
4784 		if (pages[i])
4785 			__free_page(pages[i]);
4786 	if (res.acl_scratch)
4787 		__free_page(res.acl_scratch);
4788 	return ret;
4789 }
4790 
4791 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4792 {
4793 	struct nfs4_exception exception = { };
4794 	ssize_t ret;
4795 	do {
4796 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4797 		trace_nfs4_get_acl(inode, ret);
4798 		if (ret >= 0)
4799 			break;
4800 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4801 	} while (exception.retry);
4802 	return ret;
4803 }
4804 
4805 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4806 {
4807 	struct nfs_server *server = NFS_SERVER(inode);
4808 	int ret;
4809 
4810 	if (!nfs4_server_supports_acls(server))
4811 		return -EOPNOTSUPP;
4812 	ret = nfs_revalidate_inode(server, inode);
4813 	if (ret < 0)
4814 		return ret;
4815 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4816 		nfs_zap_acl_cache(inode);
4817 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4818 	if (ret != -ENOENT)
4819 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4820 		 * but no cached acl data, just the acl length */
4821 		return ret;
4822 	return nfs4_get_acl_uncached(inode, buf, buflen);
4823 }
4824 
4825 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4826 {
4827 	struct nfs_server *server = NFS_SERVER(inode);
4828 	struct page *pages[NFS4ACL_MAXPAGES];
4829 	struct nfs_setaclargs arg = {
4830 		.fh		= NFS_FH(inode),
4831 		.acl_pages	= pages,
4832 		.acl_len	= buflen,
4833 	};
4834 	struct nfs_setaclres res;
4835 	struct rpc_message msg = {
4836 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4837 		.rpc_argp	= &arg,
4838 		.rpc_resp	= &res,
4839 	};
4840 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4841 	int ret, i;
4842 
4843 	if (!nfs4_server_supports_acls(server))
4844 		return -EOPNOTSUPP;
4845 	if (npages > ARRAY_SIZE(pages))
4846 		return -ERANGE;
4847 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4848 	if (i < 0)
4849 		return i;
4850 	nfs4_inode_return_delegation(inode);
4851 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4852 
4853 	/*
4854 	 * Free each page after tx, so the only ref left is
4855 	 * held by the network stack
4856 	 */
4857 	for (; i > 0; i--)
4858 		put_page(pages[i-1]);
4859 
4860 	/*
4861 	 * Acl update can result in inode attribute update.
4862 	 * so mark the attribute cache invalid.
4863 	 */
4864 	spin_lock(&inode->i_lock);
4865 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4866 	spin_unlock(&inode->i_lock);
4867 	nfs_access_zap_cache(inode);
4868 	nfs_zap_acl_cache(inode);
4869 	return ret;
4870 }
4871 
4872 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4873 {
4874 	struct nfs4_exception exception = { };
4875 	int err;
4876 	do {
4877 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4878 		trace_nfs4_set_acl(inode, err);
4879 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4880 				&exception);
4881 	} while (exception.retry);
4882 	return err;
4883 }
4884 
4885 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4886 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4887 					size_t buflen)
4888 {
4889 	struct nfs_server *server = NFS_SERVER(inode);
4890 	struct nfs_fattr fattr;
4891 	struct nfs4_label label = {0, 0, buflen, buf};
4892 
4893 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4894 	struct nfs4_getattr_arg arg = {
4895 		.fh		= NFS_FH(inode),
4896 		.bitmask	= bitmask,
4897 	};
4898 	struct nfs4_getattr_res res = {
4899 		.fattr		= &fattr,
4900 		.label		= &label,
4901 		.server		= server,
4902 	};
4903 	struct rpc_message msg = {
4904 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4905 		.rpc_argp	= &arg,
4906 		.rpc_resp	= &res,
4907 	};
4908 	int ret;
4909 
4910 	nfs_fattr_init(&fattr);
4911 
4912 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4913 	if (ret)
4914 		return ret;
4915 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4916 		return -ENOENT;
4917 	if (buflen < label.len)
4918 		return -ERANGE;
4919 	return 0;
4920 }
4921 
4922 static int nfs4_get_security_label(struct inode *inode, void *buf,
4923 					size_t buflen)
4924 {
4925 	struct nfs4_exception exception = { };
4926 	int err;
4927 
4928 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4929 		return -EOPNOTSUPP;
4930 
4931 	do {
4932 		err = _nfs4_get_security_label(inode, buf, buflen);
4933 		trace_nfs4_get_security_label(inode, err);
4934 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4935 				&exception);
4936 	} while (exception.retry);
4937 	return err;
4938 }
4939 
4940 static int _nfs4_do_set_security_label(struct inode *inode,
4941 		struct nfs4_label *ilabel,
4942 		struct nfs_fattr *fattr,
4943 		struct nfs4_label *olabel)
4944 {
4945 
4946 	struct iattr sattr = {0};
4947 	struct nfs_server *server = NFS_SERVER(inode);
4948 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4949 	struct nfs_setattrargs arg = {
4950 		.fh             = NFS_FH(inode),
4951 		.iap            = &sattr,
4952 		.server		= server,
4953 		.bitmask	= bitmask,
4954 		.label		= ilabel,
4955 	};
4956 	struct nfs_setattrres res = {
4957 		.fattr		= fattr,
4958 		.label		= olabel,
4959 		.server		= server,
4960 	};
4961 	struct rpc_message msg = {
4962 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4963 		.rpc_argp       = &arg,
4964 		.rpc_resp       = &res,
4965 	};
4966 	int status;
4967 
4968 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4969 
4970 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4971 	if (status)
4972 		dprintk("%s failed: %d\n", __func__, status);
4973 
4974 	return status;
4975 }
4976 
4977 static int nfs4_do_set_security_label(struct inode *inode,
4978 		struct nfs4_label *ilabel,
4979 		struct nfs_fattr *fattr,
4980 		struct nfs4_label *olabel)
4981 {
4982 	struct nfs4_exception exception = { };
4983 	int err;
4984 
4985 	do {
4986 		err = _nfs4_do_set_security_label(inode, ilabel,
4987 				fattr, olabel);
4988 		trace_nfs4_set_security_label(inode, err);
4989 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4990 				&exception);
4991 	} while (exception.retry);
4992 	return err;
4993 }
4994 
4995 static int
4996 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4997 {
4998 	struct nfs4_label ilabel, *olabel = NULL;
4999 	struct nfs_fattr fattr;
5000 	struct rpc_cred *cred;
5001 	struct inode *inode = d_inode(dentry);
5002 	int status;
5003 
5004 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5005 		return -EOPNOTSUPP;
5006 
5007 	nfs_fattr_init(&fattr);
5008 
5009 	ilabel.pi = 0;
5010 	ilabel.lfs = 0;
5011 	ilabel.label = (char *)buf;
5012 	ilabel.len = buflen;
5013 
5014 	cred = rpc_lookup_cred();
5015 	if (IS_ERR(cred))
5016 		return PTR_ERR(cred);
5017 
5018 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5019 	if (IS_ERR(olabel)) {
5020 		status = -PTR_ERR(olabel);
5021 		goto out;
5022 	}
5023 
5024 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5025 	if (status == 0)
5026 		nfs_setsecurity(inode, &fattr, olabel);
5027 
5028 	nfs4_label_free(olabel);
5029 out:
5030 	put_rpccred(cred);
5031 	return status;
5032 }
5033 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
5034 
5035 
5036 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5037 				    nfs4_verifier *bootverf)
5038 {
5039 	__be32 verf[2];
5040 
5041 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5042 		/* An impossible timestamp guarantees this value
5043 		 * will never match a generated boot time. */
5044 		verf[0] = 0;
5045 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5046 	} else {
5047 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5048 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5049 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5050 	}
5051 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
5052 }
5053 
5054 static int
5055 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5056 {
5057 	size_t len;
5058 	char *str;
5059 
5060 	if (clp->cl_owner_id != NULL)
5061 		return 0;
5062 
5063 	rcu_read_lock();
5064 	len = 14 + strlen(clp->cl_ipaddr) + 1 +
5065 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5066 		1 +
5067 		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5068 		1;
5069 	rcu_read_unlock();
5070 
5071 	if (len > NFS4_OPAQUE_LIMIT + 1)
5072 		return -EINVAL;
5073 
5074 	/*
5075 	 * Since this string is allocated at mount time, and held until the
5076 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5077 	 * about a memory-reclaim deadlock.
5078 	 */
5079 	str = kmalloc(len, GFP_KERNEL);
5080 	if (!str)
5081 		return -ENOMEM;
5082 
5083 	rcu_read_lock();
5084 	scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5085 			clp->cl_ipaddr,
5086 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5087 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5088 	rcu_read_unlock();
5089 
5090 	clp->cl_owner_id = str;
5091 	return 0;
5092 }
5093 
5094 static int
5095 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5096 {
5097 	size_t len;
5098 	char *str;
5099 
5100 	len = 10 + 10 + 1 + 10 + 1 +
5101 		strlen(nfs4_client_id_uniquifier) + 1 +
5102 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5103 
5104 	if (len > NFS4_OPAQUE_LIMIT + 1)
5105 		return -EINVAL;
5106 
5107 	/*
5108 	 * Since this string is allocated at mount time, and held until the
5109 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5110 	 * about a memory-reclaim deadlock.
5111 	 */
5112 	str = kmalloc(len, GFP_KERNEL);
5113 	if (!str)
5114 		return -ENOMEM;
5115 
5116 	scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5117 			clp->rpc_ops->version, clp->cl_minorversion,
5118 			nfs4_client_id_uniquifier,
5119 			clp->cl_rpcclient->cl_nodename);
5120 	clp->cl_owner_id = str;
5121 	return 0;
5122 }
5123 
5124 static int
5125 nfs4_init_uniform_client_string(struct nfs_client *clp)
5126 {
5127 	size_t len;
5128 	char *str;
5129 
5130 	if (clp->cl_owner_id != NULL)
5131 		return 0;
5132 
5133 	if (nfs4_client_id_uniquifier[0] != '\0')
5134 		return nfs4_init_uniquifier_client_string(clp);
5135 
5136 	len = 10 + 10 + 1 + 10 + 1 +
5137 		strlen(clp->cl_rpcclient->cl_nodename) + 1;
5138 
5139 	if (len > NFS4_OPAQUE_LIMIT + 1)
5140 		return -EINVAL;
5141 
5142 	/*
5143 	 * Since this string is allocated at mount time, and held until the
5144 	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5145 	 * about a memory-reclaim deadlock.
5146 	 */
5147 	str = kmalloc(len, GFP_KERNEL);
5148 	if (!str)
5149 		return -ENOMEM;
5150 
5151 	scnprintf(str, len, "Linux NFSv%u.%u %s",
5152 			clp->rpc_ops->version, clp->cl_minorversion,
5153 			clp->cl_rpcclient->cl_nodename);
5154 	clp->cl_owner_id = str;
5155 	return 0;
5156 }
5157 
5158 /*
5159  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5160  * services.  Advertise one based on the address family of the
5161  * clientaddr.
5162  */
5163 static unsigned int
5164 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5165 {
5166 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5167 		return scnprintf(buf, len, "tcp6");
5168 	else
5169 		return scnprintf(buf, len, "tcp");
5170 }
5171 
5172 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5173 {
5174 	struct nfs4_setclientid *sc = calldata;
5175 
5176 	if (task->tk_status == 0)
5177 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5178 }
5179 
5180 static const struct rpc_call_ops nfs4_setclientid_ops = {
5181 	.rpc_call_done = nfs4_setclientid_done,
5182 };
5183 
5184 /**
5185  * nfs4_proc_setclientid - Negotiate client ID
5186  * @clp: state data structure
5187  * @program: RPC program for NFSv4 callback service
5188  * @port: IP port number for NFS4 callback service
5189  * @cred: RPC credential to use for this call
5190  * @res: where to place the result
5191  *
5192  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5193  */
5194 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5195 		unsigned short port, struct rpc_cred *cred,
5196 		struct nfs4_setclientid_res *res)
5197 {
5198 	nfs4_verifier sc_verifier;
5199 	struct nfs4_setclientid setclientid = {
5200 		.sc_verifier = &sc_verifier,
5201 		.sc_prog = program,
5202 		.sc_clnt = clp,
5203 	};
5204 	struct rpc_message msg = {
5205 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5206 		.rpc_argp = &setclientid,
5207 		.rpc_resp = res,
5208 		.rpc_cred = cred,
5209 	};
5210 	struct rpc_task *task;
5211 	struct rpc_task_setup task_setup_data = {
5212 		.rpc_client = clp->cl_rpcclient,
5213 		.rpc_message = &msg,
5214 		.callback_ops = &nfs4_setclientid_ops,
5215 		.callback_data = &setclientid,
5216 		.flags = RPC_TASK_TIMEOUT,
5217 	};
5218 	int status;
5219 
5220 	/* nfs_client_id4 */
5221 	nfs4_init_boot_verifier(clp, &sc_verifier);
5222 
5223 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5224 		status = nfs4_init_uniform_client_string(clp);
5225 	else
5226 		status = nfs4_init_nonuniform_client_string(clp);
5227 
5228 	if (status)
5229 		goto out;
5230 
5231 	/* cb_client4 */
5232 	setclientid.sc_netid_len =
5233 				nfs4_init_callback_netid(clp,
5234 						setclientid.sc_netid,
5235 						sizeof(setclientid.sc_netid));
5236 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5237 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5238 				clp->cl_ipaddr, port >> 8, port & 255);
5239 
5240 	dprintk("NFS call  setclientid auth=%s, '%s'\n",
5241 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5242 		clp->cl_owner_id);
5243 	task = rpc_run_task(&task_setup_data);
5244 	if (IS_ERR(task)) {
5245 		status = PTR_ERR(task);
5246 		goto out;
5247 	}
5248 	status = task->tk_status;
5249 	if (setclientid.sc_cred) {
5250 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5251 		put_rpccred(setclientid.sc_cred);
5252 	}
5253 	rpc_put_task(task);
5254 out:
5255 	trace_nfs4_setclientid(clp, status);
5256 	dprintk("NFS reply setclientid: %d\n", status);
5257 	return status;
5258 }
5259 
5260 /**
5261  * nfs4_proc_setclientid_confirm - Confirm client ID
5262  * @clp: state data structure
5263  * @res: result of a previous SETCLIENTID
5264  * @cred: RPC credential to use for this call
5265  *
5266  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5267  */
5268 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5269 		struct nfs4_setclientid_res *arg,
5270 		struct rpc_cred *cred)
5271 {
5272 	struct rpc_message msg = {
5273 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5274 		.rpc_argp = arg,
5275 		.rpc_cred = cred,
5276 	};
5277 	int status;
5278 
5279 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5280 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5281 		clp->cl_clientid);
5282 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5283 	trace_nfs4_setclientid_confirm(clp, status);
5284 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5285 	return status;
5286 }
5287 
5288 struct nfs4_delegreturndata {
5289 	struct nfs4_delegreturnargs args;
5290 	struct nfs4_delegreturnres res;
5291 	struct nfs_fh fh;
5292 	nfs4_stateid stateid;
5293 	unsigned long timestamp;
5294 	struct nfs_fattr fattr;
5295 	int rpc_status;
5296 	struct inode *inode;
5297 	bool roc;
5298 	u32 roc_barrier;
5299 };
5300 
5301 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5302 {
5303 	struct nfs4_delegreturndata *data = calldata;
5304 
5305 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5306 		return;
5307 
5308 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5309 	switch (task->tk_status) {
5310 	case 0:
5311 		renew_lease(data->res.server, data->timestamp);
5312 	case -NFS4ERR_ADMIN_REVOKED:
5313 	case -NFS4ERR_DELEG_REVOKED:
5314 	case -NFS4ERR_BAD_STATEID:
5315 	case -NFS4ERR_OLD_STATEID:
5316 	case -NFS4ERR_STALE_STATEID:
5317 	case -NFS4ERR_EXPIRED:
5318 		task->tk_status = 0;
5319 		if (data->roc)
5320 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5321 		break;
5322 	default:
5323 		if (nfs4_async_handle_error(task, data->res.server,
5324 					    NULL, NULL) == -EAGAIN) {
5325 			rpc_restart_call_prepare(task);
5326 			return;
5327 		}
5328 	}
5329 	data->rpc_status = task->tk_status;
5330 }
5331 
5332 static void nfs4_delegreturn_release(void *calldata)
5333 {
5334 	struct nfs4_delegreturndata *data = calldata;
5335 	struct inode *inode = data->inode;
5336 
5337 	if (inode) {
5338 		if (data->roc)
5339 			pnfs_roc_release(inode);
5340 		nfs_iput_and_deactive(inode);
5341 	}
5342 	kfree(calldata);
5343 }
5344 
5345 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5346 {
5347 	struct nfs4_delegreturndata *d_data;
5348 
5349 	d_data = (struct nfs4_delegreturndata *)data;
5350 
5351 	if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5352 		return;
5353 
5354 	if (d_data->roc)
5355 		pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5356 
5357 	nfs4_setup_sequence(d_data->res.server,
5358 			&d_data->args.seq_args,
5359 			&d_data->res.seq_res,
5360 			task);
5361 }
5362 
5363 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5364 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5365 	.rpc_call_done = nfs4_delegreturn_done,
5366 	.rpc_release = nfs4_delegreturn_release,
5367 };
5368 
5369 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5370 {
5371 	struct nfs4_delegreturndata *data;
5372 	struct nfs_server *server = NFS_SERVER(inode);
5373 	struct rpc_task *task;
5374 	struct rpc_message msg = {
5375 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5376 		.rpc_cred = cred,
5377 	};
5378 	struct rpc_task_setup task_setup_data = {
5379 		.rpc_client = server->client,
5380 		.rpc_message = &msg,
5381 		.callback_ops = &nfs4_delegreturn_ops,
5382 		.flags = RPC_TASK_ASYNC,
5383 	};
5384 	int status = 0;
5385 
5386 	data = kzalloc(sizeof(*data), GFP_NOFS);
5387 	if (data == NULL)
5388 		return -ENOMEM;
5389 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5390 
5391 	nfs4_state_protect(server->nfs_client,
5392 			NFS_SP4_MACH_CRED_CLEANUP,
5393 			&task_setup_data.rpc_client, &msg);
5394 
5395 	data->args.fhandle = &data->fh;
5396 	data->args.stateid = &data->stateid;
5397 	data->args.bitmask = server->cache_consistency_bitmask;
5398 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5399 	nfs4_stateid_copy(&data->stateid, stateid);
5400 	data->res.fattr = &data->fattr;
5401 	data->res.server = server;
5402 	nfs_fattr_init(data->res.fattr);
5403 	data->timestamp = jiffies;
5404 	data->rpc_status = 0;
5405 	data->inode = nfs_igrab_and_active(inode);
5406 	if (data->inode)
5407 		data->roc = nfs4_roc(inode);
5408 
5409 	task_setup_data.callback_data = data;
5410 	msg.rpc_argp = &data->args;
5411 	msg.rpc_resp = &data->res;
5412 	task = rpc_run_task(&task_setup_data);
5413 	if (IS_ERR(task))
5414 		return PTR_ERR(task);
5415 	if (!issync)
5416 		goto out;
5417 	status = nfs4_wait_for_completion_rpc_task(task);
5418 	if (status != 0)
5419 		goto out;
5420 	status = data->rpc_status;
5421 	if (status == 0)
5422 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5423 	else
5424 		nfs_refresh_inode(inode, &data->fattr);
5425 out:
5426 	rpc_put_task(task);
5427 	return status;
5428 }
5429 
5430 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5431 {
5432 	struct nfs_server *server = NFS_SERVER(inode);
5433 	struct nfs4_exception exception = { };
5434 	int err;
5435 	do {
5436 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5437 		trace_nfs4_delegreturn(inode, stateid, err);
5438 		switch (err) {
5439 			case -NFS4ERR_STALE_STATEID:
5440 			case -NFS4ERR_EXPIRED:
5441 			case 0:
5442 				return 0;
5443 		}
5444 		err = nfs4_handle_exception(server, err, &exception);
5445 	} while (exception.retry);
5446 	return err;
5447 }
5448 
5449 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5450 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5451 
5452 /*
5453  * sleep, with exponential backoff, and retry the LOCK operation.
5454  */
5455 static unsigned long
5456 nfs4_set_lock_task_retry(unsigned long timeout)
5457 {
5458 	freezable_schedule_timeout_killable_unsafe(timeout);
5459 	timeout <<= 1;
5460 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5461 		return NFS4_LOCK_MAXTIMEOUT;
5462 	return timeout;
5463 }
5464 
5465 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5466 {
5467 	struct inode *inode = state->inode;
5468 	struct nfs_server *server = NFS_SERVER(inode);
5469 	struct nfs_client *clp = server->nfs_client;
5470 	struct nfs_lockt_args arg = {
5471 		.fh = NFS_FH(inode),
5472 		.fl = request,
5473 	};
5474 	struct nfs_lockt_res res = {
5475 		.denied = request,
5476 	};
5477 	struct rpc_message msg = {
5478 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5479 		.rpc_argp       = &arg,
5480 		.rpc_resp       = &res,
5481 		.rpc_cred	= state->owner->so_cred,
5482 	};
5483 	struct nfs4_lock_state *lsp;
5484 	int status;
5485 
5486 	arg.lock_owner.clientid = clp->cl_clientid;
5487 	status = nfs4_set_lock_state(state, request);
5488 	if (status != 0)
5489 		goto out;
5490 	lsp = request->fl_u.nfs4_fl.owner;
5491 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5492 	arg.lock_owner.s_dev = server->s_dev;
5493 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5494 	switch (status) {
5495 		case 0:
5496 			request->fl_type = F_UNLCK;
5497 			break;
5498 		case -NFS4ERR_DENIED:
5499 			status = 0;
5500 	}
5501 	request->fl_ops->fl_release_private(request);
5502 	request->fl_ops = NULL;
5503 out:
5504 	return status;
5505 }
5506 
5507 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5508 {
5509 	struct nfs4_exception exception = { };
5510 	int err;
5511 
5512 	do {
5513 		err = _nfs4_proc_getlk(state, cmd, request);
5514 		trace_nfs4_get_lock(request, state, cmd, err);
5515 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5516 				&exception);
5517 	} while (exception.retry);
5518 	return err;
5519 }
5520 
5521 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5522 {
5523 	return locks_lock_inode_wait(inode, fl);
5524 }
5525 
5526 struct nfs4_unlockdata {
5527 	struct nfs_locku_args arg;
5528 	struct nfs_locku_res res;
5529 	struct nfs4_lock_state *lsp;
5530 	struct nfs_open_context *ctx;
5531 	struct file_lock fl;
5532 	struct nfs_server *server;
5533 	unsigned long timestamp;
5534 };
5535 
5536 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5537 		struct nfs_open_context *ctx,
5538 		struct nfs4_lock_state *lsp,
5539 		struct nfs_seqid *seqid)
5540 {
5541 	struct nfs4_unlockdata *p;
5542 	struct inode *inode = lsp->ls_state->inode;
5543 
5544 	p = kzalloc(sizeof(*p), GFP_NOFS);
5545 	if (p == NULL)
5546 		return NULL;
5547 	p->arg.fh = NFS_FH(inode);
5548 	p->arg.fl = &p->fl;
5549 	p->arg.seqid = seqid;
5550 	p->res.seqid = seqid;
5551 	p->lsp = lsp;
5552 	atomic_inc(&lsp->ls_count);
5553 	/* Ensure we don't close file until we're done freeing locks! */
5554 	p->ctx = get_nfs_open_context(ctx);
5555 	memcpy(&p->fl, fl, sizeof(p->fl));
5556 	p->server = NFS_SERVER(inode);
5557 	return p;
5558 }
5559 
5560 static void nfs4_locku_release_calldata(void *data)
5561 {
5562 	struct nfs4_unlockdata *calldata = data;
5563 	nfs_free_seqid(calldata->arg.seqid);
5564 	nfs4_put_lock_state(calldata->lsp);
5565 	put_nfs_open_context(calldata->ctx);
5566 	kfree(calldata);
5567 }
5568 
5569 static void nfs4_locku_done(struct rpc_task *task, void *data)
5570 {
5571 	struct nfs4_unlockdata *calldata = data;
5572 
5573 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5574 		return;
5575 	switch (task->tk_status) {
5576 		case 0:
5577 			renew_lease(calldata->server, calldata->timestamp);
5578 			do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5579 			if (nfs4_update_lock_stateid(calldata->lsp,
5580 					&calldata->res.stateid))
5581 				break;
5582 		case -NFS4ERR_BAD_STATEID:
5583 		case -NFS4ERR_OLD_STATEID:
5584 		case -NFS4ERR_STALE_STATEID:
5585 		case -NFS4ERR_EXPIRED:
5586 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5587 						&calldata->lsp->ls_stateid))
5588 				rpc_restart_call_prepare(task);
5589 			break;
5590 		default:
5591 			if (nfs4_async_handle_error(task, calldata->server,
5592 						    NULL, NULL) == -EAGAIN)
5593 				rpc_restart_call_prepare(task);
5594 	}
5595 	nfs_release_seqid(calldata->arg.seqid);
5596 }
5597 
5598 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5599 {
5600 	struct nfs4_unlockdata *calldata = data;
5601 
5602 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5603 		goto out_wait;
5604 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5605 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5606 		/* Note: exit _without_ running nfs4_locku_done */
5607 		goto out_no_action;
5608 	}
5609 	calldata->timestamp = jiffies;
5610 	if (nfs4_setup_sequence(calldata->server,
5611 				&calldata->arg.seq_args,
5612 				&calldata->res.seq_res,
5613 				task) != 0)
5614 		nfs_release_seqid(calldata->arg.seqid);
5615 	return;
5616 out_no_action:
5617 	task->tk_action = NULL;
5618 out_wait:
5619 	nfs4_sequence_done(task, &calldata->res.seq_res);
5620 }
5621 
5622 static const struct rpc_call_ops nfs4_locku_ops = {
5623 	.rpc_call_prepare = nfs4_locku_prepare,
5624 	.rpc_call_done = nfs4_locku_done,
5625 	.rpc_release = nfs4_locku_release_calldata,
5626 };
5627 
5628 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5629 		struct nfs_open_context *ctx,
5630 		struct nfs4_lock_state *lsp,
5631 		struct nfs_seqid *seqid)
5632 {
5633 	struct nfs4_unlockdata *data;
5634 	struct rpc_message msg = {
5635 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5636 		.rpc_cred = ctx->cred,
5637 	};
5638 	struct rpc_task_setup task_setup_data = {
5639 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5640 		.rpc_message = &msg,
5641 		.callback_ops = &nfs4_locku_ops,
5642 		.workqueue = nfsiod_workqueue,
5643 		.flags = RPC_TASK_ASYNC,
5644 	};
5645 
5646 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5647 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5648 
5649 	/* Ensure this is an unlock - when canceling a lock, the
5650 	 * canceled lock is passed in, and it won't be an unlock.
5651 	 */
5652 	fl->fl_type = F_UNLCK;
5653 
5654 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5655 	if (data == NULL) {
5656 		nfs_free_seqid(seqid);
5657 		return ERR_PTR(-ENOMEM);
5658 	}
5659 
5660 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5661 	msg.rpc_argp = &data->arg;
5662 	msg.rpc_resp = &data->res;
5663 	task_setup_data.callback_data = data;
5664 	return rpc_run_task(&task_setup_data);
5665 }
5666 
5667 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5668 {
5669 	struct inode *inode = state->inode;
5670 	struct nfs4_state_owner *sp = state->owner;
5671 	struct nfs_inode *nfsi = NFS_I(inode);
5672 	struct nfs_seqid *seqid;
5673 	struct nfs4_lock_state *lsp;
5674 	struct rpc_task *task;
5675 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5676 	int status = 0;
5677 	unsigned char fl_flags = request->fl_flags;
5678 
5679 	status = nfs4_set_lock_state(state, request);
5680 	/* Unlock _before_ we do the RPC call */
5681 	request->fl_flags |= FL_EXISTS;
5682 	/* Exclude nfs_delegation_claim_locks() */
5683 	mutex_lock(&sp->so_delegreturn_mutex);
5684 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5685 	down_read(&nfsi->rwsem);
5686 	if (do_vfs_lock(inode, request) == -ENOENT) {
5687 		up_read(&nfsi->rwsem);
5688 		mutex_unlock(&sp->so_delegreturn_mutex);
5689 		goto out;
5690 	}
5691 	up_read(&nfsi->rwsem);
5692 	mutex_unlock(&sp->so_delegreturn_mutex);
5693 	if (status != 0)
5694 		goto out;
5695 	/* Is this a delegated lock? */
5696 	lsp = request->fl_u.nfs4_fl.owner;
5697 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5698 		goto out;
5699 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5700 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5701 	status = -ENOMEM;
5702 	if (IS_ERR(seqid))
5703 		goto out;
5704 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5705 	status = PTR_ERR(task);
5706 	if (IS_ERR(task))
5707 		goto out;
5708 	status = nfs4_wait_for_completion_rpc_task(task);
5709 	rpc_put_task(task);
5710 out:
5711 	request->fl_flags = fl_flags;
5712 	trace_nfs4_unlock(request, state, F_SETLK, status);
5713 	return status;
5714 }
5715 
5716 struct nfs4_lockdata {
5717 	struct nfs_lock_args arg;
5718 	struct nfs_lock_res res;
5719 	struct nfs4_lock_state *lsp;
5720 	struct nfs_open_context *ctx;
5721 	struct file_lock fl;
5722 	unsigned long timestamp;
5723 	int rpc_status;
5724 	int cancelled;
5725 	struct nfs_server *server;
5726 };
5727 
5728 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5729 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5730 		gfp_t gfp_mask)
5731 {
5732 	struct nfs4_lockdata *p;
5733 	struct inode *inode = lsp->ls_state->inode;
5734 	struct nfs_server *server = NFS_SERVER(inode);
5735 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5736 
5737 	p = kzalloc(sizeof(*p), gfp_mask);
5738 	if (p == NULL)
5739 		return NULL;
5740 
5741 	p->arg.fh = NFS_FH(inode);
5742 	p->arg.fl = &p->fl;
5743 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5744 	if (IS_ERR(p->arg.open_seqid))
5745 		goto out_free;
5746 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5747 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5748 	if (IS_ERR(p->arg.lock_seqid))
5749 		goto out_free_seqid;
5750 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5751 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5752 	p->arg.lock_owner.s_dev = server->s_dev;
5753 	p->res.lock_seqid = p->arg.lock_seqid;
5754 	p->lsp = lsp;
5755 	p->server = server;
5756 	atomic_inc(&lsp->ls_count);
5757 	p->ctx = get_nfs_open_context(ctx);
5758 	get_file(fl->fl_file);
5759 	memcpy(&p->fl, fl, sizeof(p->fl));
5760 	return p;
5761 out_free_seqid:
5762 	nfs_free_seqid(p->arg.open_seqid);
5763 out_free:
5764 	kfree(p);
5765 	return NULL;
5766 }
5767 
5768 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5769 {
5770 	struct nfs4_lockdata *data = calldata;
5771 	struct nfs4_state *state = data->lsp->ls_state;
5772 
5773 	dprintk("%s: begin!\n", __func__);
5774 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5775 		goto out_wait;
5776 	/* Do we need to do an open_to_lock_owner? */
5777 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5778 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5779 			goto out_release_lock_seqid;
5780 		}
5781 		nfs4_stateid_copy(&data->arg.open_stateid,
5782 				&state->open_stateid);
5783 		data->arg.new_lock_owner = 1;
5784 		data->res.open_seqid = data->arg.open_seqid;
5785 	} else {
5786 		data->arg.new_lock_owner = 0;
5787 		nfs4_stateid_copy(&data->arg.lock_stateid,
5788 				&data->lsp->ls_stateid);
5789 	}
5790 	if (!nfs4_valid_open_stateid(state)) {
5791 		data->rpc_status = -EBADF;
5792 		task->tk_action = NULL;
5793 		goto out_release_open_seqid;
5794 	}
5795 	data->timestamp = jiffies;
5796 	if (nfs4_setup_sequence(data->server,
5797 				&data->arg.seq_args,
5798 				&data->res.seq_res,
5799 				task) == 0)
5800 		return;
5801 out_release_open_seqid:
5802 	nfs_release_seqid(data->arg.open_seqid);
5803 out_release_lock_seqid:
5804 	nfs_release_seqid(data->arg.lock_seqid);
5805 out_wait:
5806 	nfs4_sequence_done(task, &data->res.seq_res);
5807 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5808 }
5809 
5810 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5811 {
5812 	struct nfs4_lockdata *data = calldata;
5813 	struct nfs4_lock_state *lsp = data->lsp;
5814 
5815 	dprintk("%s: begin!\n", __func__);
5816 
5817 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5818 		return;
5819 
5820 	data->rpc_status = task->tk_status;
5821 	switch (task->tk_status) {
5822 	case 0:
5823 		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5824 				data->timestamp);
5825 		if (data->arg.new_lock) {
5826 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5827 			if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5828 				rpc_restart_call_prepare(task);
5829 				break;
5830 			}
5831 		}
5832 		if (data->arg.new_lock_owner != 0) {
5833 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5834 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5835 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5836 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5837 			rpc_restart_call_prepare(task);
5838 		break;
5839 	case -NFS4ERR_BAD_STATEID:
5840 	case -NFS4ERR_OLD_STATEID:
5841 	case -NFS4ERR_STALE_STATEID:
5842 	case -NFS4ERR_EXPIRED:
5843 		if (data->arg.new_lock_owner != 0) {
5844 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5845 						&lsp->ls_state->open_stateid))
5846 				rpc_restart_call_prepare(task);
5847 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5848 						&lsp->ls_stateid))
5849 				rpc_restart_call_prepare(task);
5850 	}
5851 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5852 }
5853 
5854 static void nfs4_lock_release(void *calldata)
5855 {
5856 	struct nfs4_lockdata *data = calldata;
5857 
5858 	dprintk("%s: begin!\n", __func__);
5859 	nfs_free_seqid(data->arg.open_seqid);
5860 	if (data->cancelled != 0) {
5861 		struct rpc_task *task;
5862 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5863 				data->arg.lock_seqid);
5864 		if (!IS_ERR(task))
5865 			rpc_put_task_async(task);
5866 		dprintk("%s: cancelling lock!\n", __func__);
5867 	} else
5868 		nfs_free_seqid(data->arg.lock_seqid);
5869 	nfs4_put_lock_state(data->lsp);
5870 	put_nfs_open_context(data->ctx);
5871 	fput(data->fl.fl_file);
5872 	kfree(data);
5873 	dprintk("%s: done!\n", __func__);
5874 }
5875 
5876 static const struct rpc_call_ops nfs4_lock_ops = {
5877 	.rpc_call_prepare = nfs4_lock_prepare,
5878 	.rpc_call_done = nfs4_lock_done,
5879 	.rpc_release = nfs4_lock_release,
5880 };
5881 
5882 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5883 {
5884 	switch (error) {
5885 	case -NFS4ERR_ADMIN_REVOKED:
5886 	case -NFS4ERR_BAD_STATEID:
5887 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5888 		if (new_lock_owner != 0 ||
5889 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5890 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5891 		break;
5892 	case -NFS4ERR_STALE_STATEID:
5893 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5894 	case -NFS4ERR_EXPIRED:
5895 		nfs4_schedule_lease_recovery(server->nfs_client);
5896 	};
5897 }
5898 
5899 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5900 {
5901 	struct nfs4_lockdata *data;
5902 	struct rpc_task *task;
5903 	struct rpc_message msg = {
5904 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5905 		.rpc_cred = state->owner->so_cred,
5906 	};
5907 	struct rpc_task_setup task_setup_data = {
5908 		.rpc_client = NFS_CLIENT(state->inode),
5909 		.rpc_message = &msg,
5910 		.callback_ops = &nfs4_lock_ops,
5911 		.workqueue = nfsiod_workqueue,
5912 		.flags = RPC_TASK_ASYNC,
5913 	};
5914 	int ret;
5915 
5916 	dprintk("%s: begin!\n", __func__);
5917 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5918 			fl->fl_u.nfs4_fl.owner,
5919 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5920 	if (data == NULL)
5921 		return -ENOMEM;
5922 	if (IS_SETLKW(cmd))
5923 		data->arg.block = 1;
5924 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5925 	msg.rpc_argp = &data->arg;
5926 	msg.rpc_resp = &data->res;
5927 	task_setup_data.callback_data = data;
5928 	if (recovery_type > NFS_LOCK_NEW) {
5929 		if (recovery_type == NFS_LOCK_RECLAIM)
5930 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5931 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5932 	} else
5933 		data->arg.new_lock = 1;
5934 	task = rpc_run_task(&task_setup_data);
5935 	if (IS_ERR(task))
5936 		return PTR_ERR(task);
5937 	ret = nfs4_wait_for_completion_rpc_task(task);
5938 	if (ret == 0) {
5939 		ret = data->rpc_status;
5940 		if (ret)
5941 			nfs4_handle_setlk_error(data->server, data->lsp,
5942 					data->arg.new_lock_owner, ret);
5943 	} else
5944 		data->cancelled = 1;
5945 	rpc_put_task(task);
5946 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5947 	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
5948 	return ret;
5949 }
5950 
5951 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5952 {
5953 	struct nfs_server *server = NFS_SERVER(state->inode);
5954 	struct nfs4_exception exception = {
5955 		.inode = state->inode,
5956 	};
5957 	int err;
5958 
5959 	do {
5960 		/* Cache the lock if possible... */
5961 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5962 			return 0;
5963 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5964 		if (err != -NFS4ERR_DELAY)
5965 			break;
5966 		nfs4_handle_exception(server, err, &exception);
5967 	} while (exception.retry);
5968 	return err;
5969 }
5970 
5971 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5972 {
5973 	struct nfs_server *server = NFS_SERVER(state->inode);
5974 	struct nfs4_exception exception = {
5975 		.inode = state->inode,
5976 	};
5977 	int err;
5978 
5979 	err = nfs4_set_lock_state(state, request);
5980 	if (err != 0)
5981 		return err;
5982 	if (!recover_lost_locks) {
5983 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5984 		return 0;
5985 	}
5986 	do {
5987 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5988 			return 0;
5989 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5990 		switch (err) {
5991 		default:
5992 			goto out;
5993 		case -NFS4ERR_GRACE:
5994 		case -NFS4ERR_DELAY:
5995 			nfs4_handle_exception(server, err, &exception);
5996 			err = 0;
5997 		}
5998 	} while (exception.retry);
5999 out:
6000 	return err;
6001 }
6002 
6003 #if defined(CONFIG_NFS_V4_1)
6004 /**
6005  * nfs41_check_expired_locks - possibly free a lock stateid
6006  *
6007  * @state: NFSv4 state for an inode
6008  *
6009  * Returns NFS_OK if recovery for this stateid is now finished.
6010  * Otherwise a negative NFS4ERR value is returned.
6011  */
6012 static int nfs41_check_expired_locks(struct nfs4_state *state)
6013 {
6014 	int status, ret = -NFS4ERR_BAD_STATEID;
6015 	struct nfs4_lock_state *lsp;
6016 	struct nfs_server *server = NFS_SERVER(state->inode);
6017 
6018 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6019 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6020 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6021 
6022 			status = nfs41_test_stateid(server,
6023 					&lsp->ls_stateid,
6024 					cred);
6025 			trace_nfs4_test_lock_stateid(state, lsp, status);
6026 			if (status != NFS_OK) {
6027 				/* Free the stateid unless the server
6028 				 * informs us the stateid is unrecognized. */
6029 				if (status != -NFS4ERR_BAD_STATEID)
6030 					nfs41_free_stateid(server,
6031 							&lsp->ls_stateid,
6032 							cred);
6033 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6034 				ret = status;
6035 			}
6036 		}
6037 	};
6038 
6039 	return ret;
6040 }
6041 
6042 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6043 {
6044 	int status = NFS_OK;
6045 
6046 	if (test_bit(LK_STATE_IN_USE, &state->flags))
6047 		status = nfs41_check_expired_locks(state);
6048 	if (status != NFS_OK)
6049 		status = nfs4_lock_expired(state, request);
6050 	return status;
6051 }
6052 #endif
6053 
6054 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6055 {
6056 	struct nfs_inode *nfsi = NFS_I(state->inode);
6057 	unsigned char fl_flags = request->fl_flags;
6058 	int status = -ENOLCK;
6059 
6060 	if ((fl_flags & FL_POSIX) &&
6061 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6062 		goto out;
6063 	/* Is this a delegated open? */
6064 	status = nfs4_set_lock_state(state, request);
6065 	if (status != 0)
6066 		goto out;
6067 	request->fl_flags |= FL_ACCESS;
6068 	status = do_vfs_lock(state->inode, request);
6069 	if (status < 0)
6070 		goto out;
6071 	down_read(&nfsi->rwsem);
6072 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6073 		/* Yes: cache locks! */
6074 		/* ...but avoid races with delegation recall... */
6075 		request->fl_flags = fl_flags & ~FL_SLEEP;
6076 		status = do_vfs_lock(state->inode, request);
6077 		up_read(&nfsi->rwsem);
6078 		goto out;
6079 	}
6080 	up_read(&nfsi->rwsem);
6081 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6082 out:
6083 	request->fl_flags = fl_flags;
6084 	return status;
6085 }
6086 
6087 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6088 {
6089 	struct nfs4_exception exception = {
6090 		.state = state,
6091 		.inode = state->inode,
6092 	};
6093 	int err;
6094 
6095 	do {
6096 		err = _nfs4_proc_setlk(state, cmd, request);
6097 		if (err == -NFS4ERR_DENIED)
6098 			err = -EAGAIN;
6099 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
6100 				err, &exception);
6101 	} while (exception.retry);
6102 	return err;
6103 }
6104 
6105 static int
6106 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6107 {
6108 	struct nfs_open_context *ctx;
6109 	struct nfs4_state *state;
6110 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6111 	int status;
6112 
6113 	/* verify open state */
6114 	ctx = nfs_file_open_context(filp);
6115 	state = ctx->state;
6116 
6117 	if (request->fl_start < 0 || request->fl_end < 0)
6118 		return -EINVAL;
6119 
6120 	if (IS_GETLK(cmd)) {
6121 		if (state != NULL)
6122 			return nfs4_proc_getlk(state, F_GETLK, request);
6123 		return 0;
6124 	}
6125 
6126 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6127 		return -EINVAL;
6128 
6129 	if (request->fl_type == F_UNLCK) {
6130 		if (state != NULL)
6131 			return nfs4_proc_unlck(state, cmd, request);
6132 		return 0;
6133 	}
6134 
6135 	if (state == NULL)
6136 		return -ENOLCK;
6137 	/*
6138 	 * Don't rely on the VFS having checked the file open mode,
6139 	 * since it won't do this for flock() locks.
6140 	 */
6141 	switch (request->fl_type) {
6142 	case F_RDLCK:
6143 		if (!(filp->f_mode & FMODE_READ))
6144 			return -EBADF;
6145 		break;
6146 	case F_WRLCK:
6147 		if (!(filp->f_mode & FMODE_WRITE))
6148 			return -EBADF;
6149 	}
6150 
6151 	do {
6152 		status = nfs4_proc_setlk(state, cmd, request);
6153 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6154 			break;
6155 		timeout = nfs4_set_lock_task_retry(timeout);
6156 		status = -ERESTARTSYS;
6157 		if (signalled())
6158 			break;
6159 	} while(status < 0);
6160 	return status;
6161 }
6162 
6163 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6164 {
6165 	struct nfs_server *server = NFS_SERVER(state->inode);
6166 	int err;
6167 
6168 	err = nfs4_set_lock_state(state, fl);
6169 	if (err != 0)
6170 		return err;
6171 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6172 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6173 }
6174 
6175 struct nfs_release_lockowner_data {
6176 	struct nfs4_lock_state *lsp;
6177 	struct nfs_server *server;
6178 	struct nfs_release_lockowner_args args;
6179 	struct nfs_release_lockowner_res res;
6180 	unsigned long timestamp;
6181 };
6182 
6183 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6184 {
6185 	struct nfs_release_lockowner_data *data = calldata;
6186 	struct nfs_server *server = data->server;
6187 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6188 			     &data->args.seq_args, &data->res.seq_res, task);
6189 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6190 	data->timestamp = jiffies;
6191 }
6192 
6193 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6194 {
6195 	struct nfs_release_lockowner_data *data = calldata;
6196 	struct nfs_server *server = data->server;
6197 
6198 	nfs40_sequence_done(task, &data->res.seq_res);
6199 
6200 	switch (task->tk_status) {
6201 	case 0:
6202 		renew_lease(server, data->timestamp);
6203 		break;
6204 	case -NFS4ERR_STALE_CLIENTID:
6205 	case -NFS4ERR_EXPIRED:
6206 		nfs4_schedule_lease_recovery(server->nfs_client);
6207 		break;
6208 	case -NFS4ERR_LEASE_MOVED:
6209 	case -NFS4ERR_DELAY:
6210 		if (nfs4_async_handle_error(task, server,
6211 					    NULL, NULL) == -EAGAIN)
6212 			rpc_restart_call_prepare(task);
6213 	}
6214 }
6215 
6216 static void nfs4_release_lockowner_release(void *calldata)
6217 {
6218 	struct nfs_release_lockowner_data *data = calldata;
6219 	nfs4_free_lock_state(data->server, data->lsp);
6220 	kfree(calldata);
6221 }
6222 
6223 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6224 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6225 	.rpc_call_done = nfs4_release_lockowner_done,
6226 	.rpc_release = nfs4_release_lockowner_release,
6227 };
6228 
6229 static void
6230 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6231 {
6232 	struct nfs_release_lockowner_data *data;
6233 	struct rpc_message msg = {
6234 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6235 	};
6236 
6237 	if (server->nfs_client->cl_mvops->minor_version != 0)
6238 		return;
6239 
6240 	data = kmalloc(sizeof(*data), GFP_NOFS);
6241 	if (!data)
6242 		return;
6243 	data->lsp = lsp;
6244 	data->server = server;
6245 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6246 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6247 	data->args.lock_owner.s_dev = server->s_dev;
6248 
6249 	msg.rpc_argp = &data->args;
6250 	msg.rpc_resp = &data->res;
6251 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6252 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6253 }
6254 
6255 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6256 
6257 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6258 				   struct dentry *dentry, const char *key,
6259 				   const void *buf, size_t buflen,
6260 				   int flags)
6261 {
6262 	return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6263 }
6264 
6265 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6266 				   struct dentry *dentry, const char *key,
6267 				   void *buf, size_t buflen)
6268 {
6269 	return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6270 }
6271 
6272 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6273 {
6274 	return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6275 }
6276 
6277 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6278 
6279 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6280 				     struct dentry *dentry, const char *key,
6281 				     const void *buf, size_t buflen,
6282 				     int flags)
6283 {
6284 	if (security_ismaclabel(key))
6285 		return nfs4_set_security_label(dentry, buf, buflen);
6286 
6287 	return -EOPNOTSUPP;
6288 }
6289 
6290 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6291 				     struct dentry *dentry, const char *key,
6292 				     void *buf, size_t buflen)
6293 {
6294 	if (security_ismaclabel(key))
6295 		return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6296 	return -EOPNOTSUPP;
6297 }
6298 
6299 static ssize_t
6300 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6301 {
6302 	int len = 0;
6303 
6304 	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6305 		len = security_inode_listsecurity(inode, list, list_len);
6306 		if (list_len && len > list_len)
6307 			return -ERANGE;
6308 	}
6309 	return len;
6310 }
6311 
6312 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6313 	.prefix = XATTR_SECURITY_PREFIX,
6314 	.get	= nfs4_xattr_get_nfs4_label,
6315 	.set	= nfs4_xattr_set_nfs4_label,
6316 };
6317 
6318 #else
6319 
6320 static ssize_t
6321 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6322 {
6323 	return 0;
6324 }
6325 
6326 #endif
6327 
6328 /*
6329  * nfs_fhget will use either the mounted_on_fileid or the fileid
6330  */
6331 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6332 {
6333 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6334 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6335 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6336 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6337 		return;
6338 
6339 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6340 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6341 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6342 	fattr->nlink = 2;
6343 }
6344 
6345 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6346 				   const struct qstr *name,
6347 				   struct nfs4_fs_locations *fs_locations,
6348 				   struct page *page)
6349 {
6350 	struct nfs_server *server = NFS_SERVER(dir);
6351 	u32 bitmask[3] = {
6352 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6353 	};
6354 	struct nfs4_fs_locations_arg args = {
6355 		.dir_fh = NFS_FH(dir),
6356 		.name = name,
6357 		.page = page,
6358 		.bitmask = bitmask,
6359 	};
6360 	struct nfs4_fs_locations_res res = {
6361 		.fs_locations = fs_locations,
6362 	};
6363 	struct rpc_message msg = {
6364 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6365 		.rpc_argp = &args,
6366 		.rpc_resp = &res,
6367 	};
6368 	int status;
6369 
6370 	dprintk("%s: start\n", __func__);
6371 
6372 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6373 	 * is not supported */
6374 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6375 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6376 	else
6377 		bitmask[0] |= FATTR4_WORD0_FILEID;
6378 
6379 	nfs_fattr_init(&fs_locations->fattr);
6380 	fs_locations->server = server;
6381 	fs_locations->nlocations = 0;
6382 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6383 	dprintk("%s: returned status = %d\n", __func__, status);
6384 	return status;
6385 }
6386 
6387 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6388 			   const struct qstr *name,
6389 			   struct nfs4_fs_locations *fs_locations,
6390 			   struct page *page)
6391 {
6392 	struct nfs4_exception exception = { };
6393 	int err;
6394 	do {
6395 		err = _nfs4_proc_fs_locations(client, dir, name,
6396 				fs_locations, page);
6397 		trace_nfs4_get_fs_locations(dir, name, err);
6398 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6399 				&exception);
6400 	} while (exception.retry);
6401 	return err;
6402 }
6403 
6404 /*
6405  * This operation also signals the server that this client is
6406  * performing migration recovery.  The server can stop returning
6407  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6408  * appended to this compound to identify the client ID which is
6409  * performing recovery.
6410  */
6411 static int _nfs40_proc_get_locations(struct inode *inode,
6412 				     struct nfs4_fs_locations *locations,
6413 				     struct page *page, struct rpc_cred *cred)
6414 {
6415 	struct nfs_server *server = NFS_SERVER(inode);
6416 	struct rpc_clnt *clnt = server->client;
6417 	u32 bitmask[2] = {
6418 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6419 	};
6420 	struct nfs4_fs_locations_arg args = {
6421 		.clientid	= server->nfs_client->cl_clientid,
6422 		.fh		= NFS_FH(inode),
6423 		.page		= page,
6424 		.bitmask	= bitmask,
6425 		.migration	= 1,		/* skip LOOKUP */
6426 		.renew		= 1,		/* append RENEW */
6427 	};
6428 	struct nfs4_fs_locations_res res = {
6429 		.fs_locations	= locations,
6430 		.migration	= 1,
6431 		.renew		= 1,
6432 	};
6433 	struct rpc_message msg = {
6434 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6435 		.rpc_argp	= &args,
6436 		.rpc_resp	= &res,
6437 		.rpc_cred	= cred,
6438 	};
6439 	unsigned long now = jiffies;
6440 	int status;
6441 
6442 	nfs_fattr_init(&locations->fattr);
6443 	locations->server = server;
6444 	locations->nlocations = 0;
6445 
6446 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6447 	nfs4_set_sequence_privileged(&args.seq_args);
6448 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6449 					&args.seq_args, &res.seq_res);
6450 	if (status)
6451 		return status;
6452 
6453 	renew_lease(server, now);
6454 	return 0;
6455 }
6456 
6457 #ifdef CONFIG_NFS_V4_1
6458 
6459 /*
6460  * This operation also signals the server that this client is
6461  * performing migration recovery.  The server can stop asserting
6462  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6463  * performing this operation is identified in the SEQUENCE
6464  * operation in this compound.
6465  *
6466  * When the client supports GETATTR(fs_locations_info), it can
6467  * be plumbed in here.
6468  */
6469 static int _nfs41_proc_get_locations(struct inode *inode,
6470 				     struct nfs4_fs_locations *locations,
6471 				     struct page *page, struct rpc_cred *cred)
6472 {
6473 	struct nfs_server *server = NFS_SERVER(inode);
6474 	struct rpc_clnt *clnt = server->client;
6475 	u32 bitmask[2] = {
6476 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6477 	};
6478 	struct nfs4_fs_locations_arg args = {
6479 		.fh		= NFS_FH(inode),
6480 		.page		= page,
6481 		.bitmask	= bitmask,
6482 		.migration	= 1,		/* skip LOOKUP */
6483 	};
6484 	struct nfs4_fs_locations_res res = {
6485 		.fs_locations	= locations,
6486 		.migration	= 1,
6487 	};
6488 	struct rpc_message msg = {
6489 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6490 		.rpc_argp	= &args,
6491 		.rpc_resp	= &res,
6492 		.rpc_cred	= cred,
6493 	};
6494 	int status;
6495 
6496 	nfs_fattr_init(&locations->fattr);
6497 	locations->server = server;
6498 	locations->nlocations = 0;
6499 
6500 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6501 	nfs4_set_sequence_privileged(&args.seq_args);
6502 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6503 					&args.seq_args, &res.seq_res);
6504 	if (status == NFS4_OK &&
6505 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6506 		status = -NFS4ERR_LEASE_MOVED;
6507 	return status;
6508 }
6509 
6510 #endif	/* CONFIG_NFS_V4_1 */
6511 
6512 /**
6513  * nfs4_proc_get_locations - discover locations for a migrated FSID
6514  * @inode: inode on FSID that is migrating
6515  * @locations: result of query
6516  * @page: buffer
6517  * @cred: credential to use for this operation
6518  *
6519  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6520  * operation failed, or a negative errno if a local error occurred.
6521  *
6522  * On success, "locations" is filled in, but if the server has
6523  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6524  * asserted.
6525  *
6526  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6527  * from this client that require migration recovery.
6528  */
6529 int nfs4_proc_get_locations(struct inode *inode,
6530 			    struct nfs4_fs_locations *locations,
6531 			    struct page *page, struct rpc_cred *cred)
6532 {
6533 	struct nfs_server *server = NFS_SERVER(inode);
6534 	struct nfs_client *clp = server->nfs_client;
6535 	const struct nfs4_mig_recovery_ops *ops =
6536 					clp->cl_mvops->mig_recovery_ops;
6537 	struct nfs4_exception exception = { };
6538 	int status;
6539 
6540 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6541 		(unsigned long long)server->fsid.major,
6542 		(unsigned long long)server->fsid.minor,
6543 		clp->cl_hostname);
6544 	nfs_display_fhandle(NFS_FH(inode), __func__);
6545 
6546 	do {
6547 		status = ops->get_locations(inode, locations, page, cred);
6548 		if (status != -NFS4ERR_DELAY)
6549 			break;
6550 		nfs4_handle_exception(server, status, &exception);
6551 	} while (exception.retry);
6552 	return status;
6553 }
6554 
6555 /*
6556  * This operation also signals the server that this client is
6557  * performing "lease moved" recovery.  The server can stop
6558  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6559  * is appended to this compound to identify the client ID which is
6560  * performing recovery.
6561  */
6562 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6563 {
6564 	struct nfs_server *server = NFS_SERVER(inode);
6565 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6566 	struct rpc_clnt *clnt = server->client;
6567 	struct nfs4_fsid_present_arg args = {
6568 		.fh		= NFS_FH(inode),
6569 		.clientid	= clp->cl_clientid,
6570 		.renew		= 1,		/* append RENEW */
6571 	};
6572 	struct nfs4_fsid_present_res res = {
6573 		.renew		= 1,
6574 	};
6575 	struct rpc_message msg = {
6576 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6577 		.rpc_argp	= &args,
6578 		.rpc_resp	= &res,
6579 		.rpc_cred	= cred,
6580 	};
6581 	unsigned long now = jiffies;
6582 	int status;
6583 
6584 	res.fh = nfs_alloc_fhandle();
6585 	if (res.fh == NULL)
6586 		return -ENOMEM;
6587 
6588 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6589 	nfs4_set_sequence_privileged(&args.seq_args);
6590 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6591 						&args.seq_args, &res.seq_res);
6592 	nfs_free_fhandle(res.fh);
6593 	if (status)
6594 		return status;
6595 
6596 	do_renew_lease(clp, now);
6597 	return 0;
6598 }
6599 
6600 #ifdef CONFIG_NFS_V4_1
6601 
6602 /*
6603  * This operation also signals the server that this client is
6604  * performing "lease moved" recovery.  The server can stop asserting
6605  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6606  * this operation is identified in the SEQUENCE operation in this
6607  * compound.
6608  */
6609 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6610 {
6611 	struct nfs_server *server = NFS_SERVER(inode);
6612 	struct rpc_clnt *clnt = server->client;
6613 	struct nfs4_fsid_present_arg args = {
6614 		.fh		= NFS_FH(inode),
6615 	};
6616 	struct nfs4_fsid_present_res res = {
6617 	};
6618 	struct rpc_message msg = {
6619 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6620 		.rpc_argp	= &args,
6621 		.rpc_resp	= &res,
6622 		.rpc_cred	= cred,
6623 	};
6624 	int status;
6625 
6626 	res.fh = nfs_alloc_fhandle();
6627 	if (res.fh == NULL)
6628 		return -ENOMEM;
6629 
6630 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6631 	nfs4_set_sequence_privileged(&args.seq_args);
6632 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6633 						&args.seq_args, &res.seq_res);
6634 	nfs_free_fhandle(res.fh);
6635 	if (status == NFS4_OK &&
6636 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6637 		status = -NFS4ERR_LEASE_MOVED;
6638 	return status;
6639 }
6640 
6641 #endif	/* CONFIG_NFS_V4_1 */
6642 
6643 /**
6644  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6645  * @inode: inode on FSID to check
6646  * @cred: credential to use for this operation
6647  *
6648  * Server indicates whether the FSID is present, moved, or not
6649  * recognized.  This operation is necessary to clear a LEASE_MOVED
6650  * condition for this client ID.
6651  *
6652  * Returns NFS4_OK if the FSID is present on this server,
6653  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6654  *  NFS4ERR code if some error occurred on the server, or a
6655  *  negative errno if a local failure occurred.
6656  */
6657 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6658 {
6659 	struct nfs_server *server = NFS_SERVER(inode);
6660 	struct nfs_client *clp = server->nfs_client;
6661 	const struct nfs4_mig_recovery_ops *ops =
6662 					clp->cl_mvops->mig_recovery_ops;
6663 	struct nfs4_exception exception = { };
6664 	int status;
6665 
6666 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6667 		(unsigned long long)server->fsid.major,
6668 		(unsigned long long)server->fsid.minor,
6669 		clp->cl_hostname);
6670 	nfs_display_fhandle(NFS_FH(inode), __func__);
6671 
6672 	do {
6673 		status = ops->fsid_present(inode, cred);
6674 		if (status != -NFS4ERR_DELAY)
6675 			break;
6676 		nfs4_handle_exception(server, status, &exception);
6677 	} while (exception.retry);
6678 	return status;
6679 }
6680 
6681 /**
6682  * If 'use_integrity' is true and the state managment nfs_client
6683  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6684  * and the machine credential as per RFC3530bis and RFC5661 Security
6685  * Considerations sections. Otherwise, just use the user cred with the
6686  * filesystem's rpc_client.
6687  */
6688 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6689 {
6690 	int status;
6691 	struct nfs4_secinfo_arg args = {
6692 		.dir_fh = NFS_FH(dir),
6693 		.name   = name,
6694 	};
6695 	struct nfs4_secinfo_res res = {
6696 		.flavors     = flavors,
6697 	};
6698 	struct rpc_message msg = {
6699 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6700 		.rpc_argp = &args,
6701 		.rpc_resp = &res,
6702 	};
6703 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6704 	struct rpc_cred *cred = NULL;
6705 
6706 	if (use_integrity) {
6707 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6708 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6709 		msg.rpc_cred = cred;
6710 	}
6711 
6712 	dprintk("NFS call  secinfo %s\n", name->name);
6713 
6714 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6715 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6716 
6717 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6718 				&res.seq_res, 0);
6719 	dprintk("NFS reply  secinfo: %d\n", status);
6720 
6721 	if (cred)
6722 		put_rpccred(cred);
6723 
6724 	return status;
6725 }
6726 
6727 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6728 		      struct nfs4_secinfo_flavors *flavors)
6729 {
6730 	struct nfs4_exception exception = { };
6731 	int err;
6732 	do {
6733 		err = -NFS4ERR_WRONGSEC;
6734 
6735 		/* try to use integrity protection with machine cred */
6736 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6737 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6738 
6739 		/*
6740 		 * if unable to use integrity protection, or SECINFO with
6741 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6742 		 * disallowed by spec, but exists in deployed servers) use
6743 		 * the current filesystem's rpc_client and the user cred.
6744 		 */
6745 		if (err == -NFS4ERR_WRONGSEC)
6746 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6747 
6748 		trace_nfs4_secinfo(dir, name, err);
6749 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6750 				&exception);
6751 	} while (exception.retry);
6752 	return err;
6753 }
6754 
6755 #ifdef CONFIG_NFS_V4_1
6756 /*
6757  * Check the exchange flags returned by the server for invalid flags, having
6758  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6759  * DS flags set.
6760  */
6761 static int nfs4_check_cl_exchange_flags(u32 flags)
6762 {
6763 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6764 		goto out_inval;
6765 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6766 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6767 		goto out_inval;
6768 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6769 		goto out_inval;
6770 	return NFS_OK;
6771 out_inval:
6772 	return -NFS4ERR_INVAL;
6773 }
6774 
6775 static bool
6776 nfs41_same_server_scope(struct nfs41_server_scope *a,
6777 			struct nfs41_server_scope *b)
6778 {
6779 	if (a->server_scope_sz == b->server_scope_sz &&
6780 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6781 		return true;
6782 
6783 	return false;
6784 }
6785 
6786 static void
6787 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
6788 {
6789 }
6790 
6791 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
6792 	.rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
6793 };
6794 
6795 /*
6796  * nfs4_proc_bind_one_conn_to_session()
6797  *
6798  * The 4.1 client currently uses the same TCP connection for the
6799  * fore and backchannel.
6800  */
6801 static
6802 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
6803 		struct rpc_xprt *xprt,
6804 		struct nfs_client *clp,
6805 		struct rpc_cred *cred)
6806 {
6807 	int status;
6808 	struct nfs41_bind_conn_to_session_args args = {
6809 		.client = clp,
6810 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6811 	};
6812 	struct nfs41_bind_conn_to_session_res res;
6813 	struct rpc_message msg = {
6814 		.rpc_proc =
6815 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6816 		.rpc_argp = &args,
6817 		.rpc_resp = &res,
6818 		.rpc_cred = cred,
6819 	};
6820 	struct rpc_task_setup task_setup_data = {
6821 		.rpc_client = clnt,
6822 		.rpc_xprt = xprt,
6823 		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
6824 		.rpc_message = &msg,
6825 		.flags = RPC_TASK_TIMEOUT,
6826 	};
6827 	struct rpc_task *task;
6828 
6829 	dprintk("--> %s\n", __func__);
6830 
6831 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6832 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6833 		args.dir = NFS4_CDFC4_FORE;
6834 
6835 	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
6836 	if (xprt != rcu_access_pointer(clnt->cl_xprt))
6837 		args.dir = NFS4_CDFC4_FORE;
6838 
6839 	task = rpc_run_task(&task_setup_data);
6840 	if (!IS_ERR(task)) {
6841 		status = task->tk_status;
6842 		rpc_put_task(task);
6843 	} else
6844 		status = PTR_ERR(task);
6845 	trace_nfs4_bind_conn_to_session(clp, status);
6846 	if (status == 0) {
6847 		if (memcmp(res.sessionid.data,
6848 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6849 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6850 			status = -EIO;
6851 			goto out;
6852 		}
6853 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6854 			dprintk("NFS: %s: Unexpected direction from server\n",
6855 				__func__);
6856 			status = -EIO;
6857 			goto out;
6858 		}
6859 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6860 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6861 				__func__);
6862 			status = -EIO;
6863 			goto out;
6864 		}
6865 	}
6866 out:
6867 	dprintk("<-- %s status= %d\n", __func__, status);
6868 	return status;
6869 }
6870 
6871 struct rpc_bind_conn_calldata {
6872 	struct nfs_client *clp;
6873 	struct rpc_cred *cred;
6874 };
6875 
6876 static int
6877 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
6878 		struct rpc_xprt *xprt,
6879 		void *calldata)
6880 {
6881 	struct rpc_bind_conn_calldata *p = calldata;
6882 
6883 	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
6884 }
6885 
6886 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6887 {
6888 	struct rpc_bind_conn_calldata data = {
6889 		.clp = clp,
6890 		.cred = cred,
6891 	};
6892 	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
6893 			nfs4_proc_bind_conn_to_session_callback, &data);
6894 }
6895 
6896 /*
6897  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6898  * and operations we'd like to see to enable certain features in the allow map
6899  */
6900 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6901 	.how = SP4_MACH_CRED,
6902 	.enforce.u.words = {
6903 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6904 		      1 << (OP_EXCHANGE_ID - 32) |
6905 		      1 << (OP_CREATE_SESSION - 32) |
6906 		      1 << (OP_DESTROY_SESSION - 32) |
6907 		      1 << (OP_DESTROY_CLIENTID - 32)
6908 	},
6909 	.allow.u.words = {
6910 		[0] = 1 << (OP_CLOSE) |
6911 		      1 << (OP_OPEN_DOWNGRADE) |
6912 		      1 << (OP_LOCKU) |
6913 		      1 << (OP_DELEGRETURN) |
6914 		      1 << (OP_COMMIT),
6915 		[1] = 1 << (OP_SECINFO - 32) |
6916 		      1 << (OP_SECINFO_NO_NAME - 32) |
6917 		      1 << (OP_LAYOUTRETURN - 32) |
6918 		      1 << (OP_TEST_STATEID - 32) |
6919 		      1 << (OP_FREE_STATEID - 32) |
6920 		      1 << (OP_WRITE - 32)
6921 	}
6922 };
6923 
6924 /*
6925  * Select the state protection mode for client `clp' given the server results
6926  * from exchange_id in `sp'.
6927  *
6928  * Returns 0 on success, negative errno otherwise.
6929  */
6930 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6931 				 struct nfs41_state_protection *sp)
6932 {
6933 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6934 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6935 		      1 << (OP_EXCHANGE_ID - 32) |
6936 		      1 << (OP_CREATE_SESSION - 32) |
6937 		      1 << (OP_DESTROY_SESSION - 32) |
6938 		      1 << (OP_DESTROY_CLIENTID - 32)
6939 	};
6940 	unsigned int i;
6941 
6942 	if (sp->how == SP4_MACH_CRED) {
6943 		/* Print state protect result */
6944 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6945 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6946 			if (test_bit(i, sp->enforce.u.longs))
6947 				dfprintk(MOUNT, "  enforce op %d\n", i);
6948 			if (test_bit(i, sp->allow.u.longs))
6949 				dfprintk(MOUNT, "  allow op %d\n", i);
6950 		}
6951 
6952 		/* make sure nothing is on enforce list that isn't supported */
6953 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6954 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6955 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6956 				return -EINVAL;
6957 			}
6958 		}
6959 
6960 		/*
6961 		 * Minimal mode - state operations are allowed to use machine
6962 		 * credential.  Note this already happens by default, so the
6963 		 * client doesn't have to do anything more than the negotiation.
6964 		 *
6965 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6966 		 *       we're already using the machine cred for exchange_id
6967 		 *       and will never use a different cred.
6968 		 */
6969 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6970 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6971 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6972 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6973 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6974 			dfprintk(MOUNT, "  minimal mode enabled\n");
6975 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6976 		} else {
6977 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6978 			return -EINVAL;
6979 		}
6980 
6981 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6982 		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
6983 		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
6984 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6985 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6986 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6987 		}
6988 
6989 		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
6990 			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
6991 			set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
6992 				&clp->cl_sp4_flags);
6993 		}
6994 
6995 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6996 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6997 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6998 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6999 		}
7000 
7001 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7002 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7003 			dfprintk(MOUNT, "  stateid mode enabled\n");
7004 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7005 		}
7006 
7007 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7008 			dfprintk(MOUNT, "  write mode enabled\n");
7009 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7010 		}
7011 
7012 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7013 			dfprintk(MOUNT, "  commit mode enabled\n");
7014 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7015 		}
7016 	}
7017 
7018 	return 0;
7019 }
7020 
7021 /*
7022  * _nfs4_proc_exchange_id()
7023  *
7024  * Wrapper for EXCHANGE_ID operation.
7025  */
7026 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7027 	u32 sp4_how)
7028 {
7029 	nfs4_verifier verifier;
7030 	struct nfs41_exchange_id_args args = {
7031 		.verifier = &verifier,
7032 		.client = clp,
7033 #ifdef CONFIG_NFS_V4_1_MIGRATION
7034 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7035 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
7036 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7037 #else
7038 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7039 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
7040 #endif
7041 	};
7042 	struct nfs41_exchange_id_res res = {
7043 		0
7044 	};
7045 	int status;
7046 	struct rpc_message msg = {
7047 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7048 		.rpc_argp = &args,
7049 		.rpc_resp = &res,
7050 		.rpc_cred = cred,
7051 	};
7052 
7053 	nfs4_init_boot_verifier(clp, &verifier);
7054 
7055 	status = nfs4_init_uniform_client_string(clp);
7056 	if (status)
7057 		goto out;
7058 
7059 	dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7060 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
7061 		clp->cl_owner_id);
7062 
7063 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7064 					GFP_NOFS);
7065 	if (unlikely(res.server_owner == NULL)) {
7066 		status = -ENOMEM;
7067 		goto out;
7068 	}
7069 
7070 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7071 					GFP_NOFS);
7072 	if (unlikely(res.server_scope == NULL)) {
7073 		status = -ENOMEM;
7074 		goto out_server_owner;
7075 	}
7076 
7077 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7078 	if (unlikely(res.impl_id == NULL)) {
7079 		status = -ENOMEM;
7080 		goto out_server_scope;
7081 	}
7082 
7083 	switch (sp4_how) {
7084 	case SP4_NONE:
7085 		args.state_protect.how = SP4_NONE;
7086 		break;
7087 
7088 	case SP4_MACH_CRED:
7089 		args.state_protect = nfs4_sp4_mach_cred_request;
7090 		break;
7091 
7092 	default:
7093 		/* unsupported! */
7094 		WARN_ON_ONCE(1);
7095 		status = -EINVAL;
7096 		goto out_impl_id;
7097 	}
7098 
7099 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7100 	trace_nfs4_exchange_id(clp, status);
7101 	if (status == 0)
7102 		status = nfs4_check_cl_exchange_flags(res.flags);
7103 
7104 	if (status == 0)
7105 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
7106 
7107 	if (status == 0) {
7108 		clp->cl_clientid = res.clientid;
7109 		clp->cl_exchange_flags = res.flags;
7110 		/* Client ID is not confirmed */
7111 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7112 			clear_bit(NFS4_SESSION_ESTABLISHED,
7113 					&clp->cl_session->session_state);
7114 			clp->cl_seqid = res.seqid;
7115 		}
7116 
7117 		kfree(clp->cl_serverowner);
7118 		clp->cl_serverowner = res.server_owner;
7119 		res.server_owner = NULL;
7120 
7121 		/* use the most recent implementation id */
7122 		kfree(clp->cl_implid);
7123 		clp->cl_implid = res.impl_id;
7124 		res.impl_id = NULL;
7125 
7126 		if (clp->cl_serverscope != NULL &&
7127 		    !nfs41_same_server_scope(clp->cl_serverscope,
7128 					     res.server_scope)) {
7129 			dprintk("%s: server_scope mismatch detected\n",
7130 				__func__);
7131 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7132 			kfree(clp->cl_serverscope);
7133 			clp->cl_serverscope = NULL;
7134 		}
7135 
7136 		if (clp->cl_serverscope == NULL) {
7137 			clp->cl_serverscope = res.server_scope;
7138 			res.server_scope = NULL;
7139 		}
7140 	}
7141 
7142 out_impl_id:
7143 	kfree(res.impl_id);
7144 out_server_scope:
7145 	kfree(res.server_scope);
7146 out_server_owner:
7147 	kfree(res.server_owner);
7148 out:
7149 	if (clp->cl_implid != NULL)
7150 		dprintk("NFS reply exchange_id: Server Implementation ID: "
7151 			"domain: %s, name: %s, date: %llu,%u\n",
7152 			clp->cl_implid->domain, clp->cl_implid->name,
7153 			clp->cl_implid->date.seconds,
7154 			clp->cl_implid->date.nseconds);
7155 	dprintk("NFS reply exchange_id: %d\n", status);
7156 	return status;
7157 }
7158 
7159 /*
7160  * nfs4_proc_exchange_id()
7161  *
7162  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7163  *
7164  * Since the clientid has expired, all compounds using sessions
7165  * associated with the stale clientid will be returning
7166  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7167  * be in some phase of session reset.
7168  *
7169  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7170  */
7171 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7172 {
7173 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7174 	int status;
7175 
7176 	/* try SP4_MACH_CRED if krb5i/p	*/
7177 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
7178 	    authflavor == RPC_AUTH_GSS_KRB5P) {
7179 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7180 		if (!status)
7181 			return 0;
7182 	}
7183 
7184 	/* try SP4_NONE */
7185 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7186 }
7187 
7188 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7189 		struct rpc_cred *cred)
7190 {
7191 	struct rpc_message msg = {
7192 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7193 		.rpc_argp = clp,
7194 		.rpc_cred = cred,
7195 	};
7196 	int status;
7197 
7198 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7199 	trace_nfs4_destroy_clientid(clp, status);
7200 	if (status)
7201 		dprintk("NFS: Got error %d from the server %s on "
7202 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
7203 	return status;
7204 }
7205 
7206 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7207 		struct rpc_cred *cred)
7208 {
7209 	unsigned int loop;
7210 	int ret;
7211 
7212 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7213 		ret = _nfs4_proc_destroy_clientid(clp, cred);
7214 		switch (ret) {
7215 		case -NFS4ERR_DELAY:
7216 		case -NFS4ERR_CLIENTID_BUSY:
7217 			ssleep(1);
7218 			break;
7219 		default:
7220 			return ret;
7221 		}
7222 	}
7223 	return 0;
7224 }
7225 
7226 int nfs4_destroy_clientid(struct nfs_client *clp)
7227 {
7228 	struct rpc_cred *cred;
7229 	int ret = 0;
7230 
7231 	if (clp->cl_mvops->minor_version < 1)
7232 		goto out;
7233 	if (clp->cl_exchange_flags == 0)
7234 		goto out;
7235 	if (clp->cl_preserve_clid)
7236 		goto out;
7237 	cred = nfs4_get_clid_cred(clp);
7238 	ret = nfs4_proc_destroy_clientid(clp, cred);
7239 	if (cred)
7240 		put_rpccred(cred);
7241 	switch (ret) {
7242 	case 0:
7243 	case -NFS4ERR_STALE_CLIENTID:
7244 		clp->cl_exchange_flags = 0;
7245 	}
7246 out:
7247 	return ret;
7248 }
7249 
7250 struct nfs4_get_lease_time_data {
7251 	struct nfs4_get_lease_time_args *args;
7252 	struct nfs4_get_lease_time_res *res;
7253 	struct nfs_client *clp;
7254 };
7255 
7256 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7257 					void *calldata)
7258 {
7259 	struct nfs4_get_lease_time_data *data =
7260 			(struct nfs4_get_lease_time_data *)calldata;
7261 
7262 	dprintk("--> %s\n", __func__);
7263 	/* just setup sequence, do not trigger session recovery
7264 	   since we're invoked within one */
7265 	nfs41_setup_sequence(data->clp->cl_session,
7266 			&data->args->la_seq_args,
7267 			&data->res->lr_seq_res,
7268 			task);
7269 	dprintk("<-- %s\n", __func__);
7270 }
7271 
7272 /*
7273  * Called from nfs4_state_manager thread for session setup, so don't recover
7274  * from sequence operation or clientid errors.
7275  */
7276 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7277 {
7278 	struct nfs4_get_lease_time_data *data =
7279 			(struct nfs4_get_lease_time_data *)calldata;
7280 
7281 	dprintk("--> %s\n", __func__);
7282 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7283 		return;
7284 	switch (task->tk_status) {
7285 	case -NFS4ERR_DELAY:
7286 	case -NFS4ERR_GRACE:
7287 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7288 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7289 		task->tk_status = 0;
7290 		/* fall through */
7291 	case -NFS4ERR_RETRY_UNCACHED_REP:
7292 		rpc_restart_call_prepare(task);
7293 		return;
7294 	}
7295 	dprintk("<-- %s\n", __func__);
7296 }
7297 
7298 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7299 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7300 	.rpc_call_done = nfs4_get_lease_time_done,
7301 };
7302 
7303 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7304 {
7305 	struct rpc_task *task;
7306 	struct nfs4_get_lease_time_args args;
7307 	struct nfs4_get_lease_time_res res = {
7308 		.lr_fsinfo = fsinfo,
7309 	};
7310 	struct nfs4_get_lease_time_data data = {
7311 		.args = &args,
7312 		.res = &res,
7313 		.clp = clp,
7314 	};
7315 	struct rpc_message msg = {
7316 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7317 		.rpc_argp = &args,
7318 		.rpc_resp = &res,
7319 	};
7320 	struct rpc_task_setup task_setup = {
7321 		.rpc_client = clp->cl_rpcclient,
7322 		.rpc_message = &msg,
7323 		.callback_ops = &nfs4_get_lease_time_ops,
7324 		.callback_data = &data,
7325 		.flags = RPC_TASK_TIMEOUT,
7326 	};
7327 	int status;
7328 
7329 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7330 	nfs4_set_sequence_privileged(&args.la_seq_args);
7331 	dprintk("--> %s\n", __func__);
7332 	task = rpc_run_task(&task_setup);
7333 
7334 	if (IS_ERR(task))
7335 		status = PTR_ERR(task);
7336 	else {
7337 		status = task->tk_status;
7338 		rpc_put_task(task);
7339 	}
7340 	dprintk("<-- %s return %d\n", __func__, status);
7341 
7342 	return status;
7343 }
7344 
7345 /*
7346  * Initialize the values to be used by the client in CREATE_SESSION
7347  * If nfs4_init_session set the fore channel request and response sizes,
7348  * use them.
7349  *
7350  * Set the back channel max_resp_sz_cached to zero to force the client to
7351  * always set csa_cachethis to FALSE because the current implementation
7352  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7353  */
7354 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7355 {
7356 	unsigned int max_rqst_sz, max_resp_sz;
7357 
7358 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7359 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7360 
7361 	/* Fore channel attributes */
7362 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7363 	args->fc_attrs.max_resp_sz = max_resp_sz;
7364 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7365 	args->fc_attrs.max_reqs = max_session_slots;
7366 
7367 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7368 		"max_ops=%u max_reqs=%u\n",
7369 		__func__,
7370 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7371 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7372 
7373 	/* Back channel attributes */
7374 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7375 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7376 	args->bc_attrs.max_resp_sz_cached = 0;
7377 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7378 	args->bc_attrs.max_reqs = NFS41_BC_MAX_CALLBACKS;
7379 
7380 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7381 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7382 		__func__,
7383 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7384 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7385 		args->bc_attrs.max_reqs);
7386 }
7387 
7388 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7389 		struct nfs41_create_session_res *res)
7390 {
7391 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7392 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7393 
7394 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7395 		return -EINVAL;
7396 	/*
7397 	 * Our requested max_ops is the minimum we need; we're not
7398 	 * prepared to break up compounds into smaller pieces than that.
7399 	 * So, no point even trying to continue if the server won't
7400 	 * cooperate:
7401 	 */
7402 	if (rcvd->max_ops < sent->max_ops)
7403 		return -EINVAL;
7404 	if (rcvd->max_reqs == 0)
7405 		return -EINVAL;
7406 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7407 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7408 	return 0;
7409 }
7410 
7411 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7412 		struct nfs41_create_session_res *res)
7413 {
7414 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7415 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7416 
7417 	if (!(res->flags & SESSION4_BACK_CHAN))
7418 		goto out;
7419 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7420 		return -EINVAL;
7421 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7422 		return -EINVAL;
7423 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7424 		return -EINVAL;
7425 	/* These would render the backchannel useless: */
7426 	if (rcvd->max_ops != sent->max_ops)
7427 		return -EINVAL;
7428 	if (rcvd->max_reqs != sent->max_reqs)
7429 		return -EINVAL;
7430 out:
7431 	return 0;
7432 }
7433 
7434 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7435 				     struct nfs41_create_session_res *res)
7436 {
7437 	int ret;
7438 
7439 	ret = nfs4_verify_fore_channel_attrs(args, res);
7440 	if (ret)
7441 		return ret;
7442 	return nfs4_verify_back_channel_attrs(args, res);
7443 }
7444 
7445 static void nfs4_update_session(struct nfs4_session *session,
7446 		struct nfs41_create_session_res *res)
7447 {
7448 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7449 	/* Mark client id and session as being confirmed */
7450 	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7451 	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7452 	session->flags = res->flags;
7453 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7454 	if (res->flags & SESSION4_BACK_CHAN)
7455 		memcpy(&session->bc_attrs, &res->bc_attrs,
7456 				sizeof(session->bc_attrs));
7457 }
7458 
7459 static int _nfs4_proc_create_session(struct nfs_client *clp,
7460 		struct rpc_cred *cred)
7461 {
7462 	struct nfs4_session *session = clp->cl_session;
7463 	struct nfs41_create_session_args args = {
7464 		.client = clp,
7465 		.clientid = clp->cl_clientid,
7466 		.seqid = clp->cl_seqid,
7467 		.cb_program = NFS4_CALLBACK,
7468 	};
7469 	struct nfs41_create_session_res res;
7470 
7471 	struct rpc_message msg = {
7472 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7473 		.rpc_argp = &args,
7474 		.rpc_resp = &res,
7475 		.rpc_cred = cred,
7476 	};
7477 	int status;
7478 
7479 	nfs4_init_channel_attrs(&args);
7480 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7481 
7482 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7483 	trace_nfs4_create_session(clp, status);
7484 
7485 	if (!status) {
7486 		/* Verify the session's negotiated channel_attrs values */
7487 		status = nfs4_verify_channel_attrs(&args, &res);
7488 		/* Increment the clientid slot sequence id */
7489 		if (clp->cl_seqid == res.seqid)
7490 			clp->cl_seqid++;
7491 		if (status)
7492 			goto out;
7493 		nfs4_update_session(session, &res);
7494 	}
7495 out:
7496 	return status;
7497 }
7498 
7499 /*
7500  * Issues a CREATE_SESSION operation to the server.
7501  * It is the responsibility of the caller to verify the session is
7502  * expired before calling this routine.
7503  */
7504 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7505 {
7506 	int status;
7507 	unsigned *ptr;
7508 	struct nfs4_session *session = clp->cl_session;
7509 
7510 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7511 
7512 	status = _nfs4_proc_create_session(clp, cred);
7513 	if (status)
7514 		goto out;
7515 
7516 	/* Init or reset the session slot tables */
7517 	status = nfs4_setup_session_slot_tables(session);
7518 	dprintk("slot table setup returned %d\n", status);
7519 	if (status)
7520 		goto out;
7521 
7522 	ptr = (unsigned *)&session->sess_id.data[0];
7523 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7524 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7525 out:
7526 	dprintk("<-- %s\n", __func__);
7527 	return status;
7528 }
7529 
7530 /*
7531  * Issue the over-the-wire RPC DESTROY_SESSION.
7532  * The caller must serialize access to this routine.
7533  */
7534 int nfs4_proc_destroy_session(struct nfs4_session *session,
7535 		struct rpc_cred *cred)
7536 {
7537 	struct rpc_message msg = {
7538 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7539 		.rpc_argp = session,
7540 		.rpc_cred = cred,
7541 	};
7542 	int status = 0;
7543 
7544 	dprintk("--> nfs4_proc_destroy_session\n");
7545 
7546 	/* session is still being setup */
7547 	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7548 		return 0;
7549 
7550 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7551 	trace_nfs4_destroy_session(session->clp, status);
7552 
7553 	if (status)
7554 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7555 			"Session has been destroyed regardless...\n", status);
7556 
7557 	dprintk("<-- nfs4_proc_destroy_session\n");
7558 	return status;
7559 }
7560 
7561 /*
7562  * Renew the cl_session lease.
7563  */
7564 struct nfs4_sequence_data {
7565 	struct nfs_client *clp;
7566 	struct nfs4_sequence_args args;
7567 	struct nfs4_sequence_res res;
7568 };
7569 
7570 static void nfs41_sequence_release(void *data)
7571 {
7572 	struct nfs4_sequence_data *calldata = data;
7573 	struct nfs_client *clp = calldata->clp;
7574 
7575 	if (atomic_read(&clp->cl_count) > 1)
7576 		nfs4_schedule_state_renewal(clp);
7577 	nfs_put_client(clp);
7578 	kfree(calldata);
7579 }
7580 
7581 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7582 {
7583 	switch(task->tk_status) {
7584 	case -NFS4ERR_DELAY:
7585 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7586 		return -EAGAIN;
7587 	default:
7588 		nfs4_schedule_lease_recovery(clp);
7589 	}
7590 	return 0;
7591 }
7592 
7593 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7594 {
7595 	struct nfs4_sequence_data *calldata = data;
7596 	struct nfs_client *clp = calldata->clp;
7597 
7598 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7599 		return;
7600 
7601 	trace_nfs4_sequence(clp, task->tk_status);
7602 	if (task->tk_status < 0) {
7603 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7604 		if (atomic_read(&clp->cl_count) == 1)
7605 			goto out;
7606 
7607 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7608 			rpc_restart_call_prepare(task);
7609 			return;
7610 		}
7611 	}
7612 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7613 out:
7614 	dprintk("<-- %s\n", __func__);
7615 }
7616 
7617 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7618 {
7619 	struct nfs4_sequence_data *calldata = data;
7620 	struct nfs_client *clp = calldata->clp;
7621 	struct nfs4_sequence_args *args;
7622 	struct nfs4_sequence_res *res;
7623 
7624 	args = task->tk_msg.rpc_argp;
7625 	res = task->tk_msg.rpc_resp;
7626 
7627 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7628 }
7629 
7630 static const struct rpc_call_ops nfs41_sequence_ops = {
7631 	.rpc_call_done = nfs41_sequence_call_done,
7632 	.rpc_call_prepare = nfs41_sequence_prepare,
7633 	.rpc_release = nfs41_sequence_release,
7634 };
7635 
7636 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7637 		struct rpc_cred *cred,
7638 		bool is_privileged)
7639 {
7640 	struct nfs4_sequence_data *calldata;
7641 	struct rpc_message msg = {
7642 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7643 		.rpc_cred = cred,
7644 	};
7645 	struct rpc_task_setup task_setup_data = {
7646 		.rpc_client = clp->cl_rpcclient,
7647 		.rpc_message = &msg,
7648 		.callback_ops = &nfs41_sequence_ops,
7649 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7650 	};
7651 
7652 	if (!atomic_inc_not_zero(&clp->cl_count))
7653 		return ERR_PTR(-EIO);
7654 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7655 	if (calldata == NULL) {
7656 		nfs_put_client(clp);
7657 		return ERR_PTR(-ENOMEM);
7658 	}
7659 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7660 	if (is_privileged)
7661 		nfs4_set_sequence_privileged(&calldata->args);
7662 	msg.rpc_argp = &calldata->args;
7663 	msg.rpc_resp = &calldata->res;
7664 	calldata->clp = clp;
7665 	task_setup_data.callback_data = calldata;
7666 
7667 	return rpc_run_task(&task_setup_data);
7668 }
7669 
7670 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7671 {
7672 	struct rpc_task *task;
7673 	int ret = 0;
7674 
7675 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7676 		return -EAGAIN;
7677 	task = _nfs41_proc_sequence(clp, cred, false);
7678 	if (IS_ERR(task))
7679 		ret = PTR_ERR(task);
7680 	else
7681 		rpc_put_task_async(task);
7682 	dprintk("<-- %s status=%d\n", __func__, ret);
7683 	return ret;
7684 }
7685 
7686 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7687 {
7688 	struct rpc_task *task;
7689 	int ret;
7690 
7691 	task = _nfs41_proc_sequence(clp, cred, true);
7692 	if (IS_ERR(task)) {
7693 		ret = PTR_ERR(task);
7694 		goto out;
7695 	}
7696 	ret = rpc_wait_for_completion_task(task);
7697 	if (!ret)
7698 		ret = task->tk_status;
7699 	rpc_put_task(task);
7700 out:
7701 	dprintk("<-- %s status=%d\n", __func__, ret);
7702 	return ret;
7703 }
7704 
7705 struct nfs4_reclaim_complete_data {
7706 	struct nfs_client *clp;
7707 	struct nfs41_reclaim_complete_args arg;
7708 	struct nfs41_reclaim_complete_res res;
7709 };
7710 
7711 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7712 {
7713 	struct nfs4_reclaim_complete_data *calldata = data;
7714 
7715 	nfs41_setup_sequence(calldata->clp->cl_session,
7716 			&calldata->arg.seq_args,
7717 			&calldata->res.seq_res,
7718 			task);
7719 }
7720 
7721 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7722 {
7723 	switch(task->tk_status) {
7724 	case 0:
7725 	case -NFS4ERR_COMPLETE_ALREADY:
7726 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7727 		break;
7728 	case -NFS4ERR_DELAY:
7729 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7730 		/* fall through */
7731 	case -NFS4ERR_RETRY_UNCACHED_REP:
7732 		return -EAGAIN;
7733 	default:
7734 		nfs4_schedule_lease_recovery(clp);
7735 	}
7736 	return 0;
7737 }
7738 
7739 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7740 {
7741 	struct nfs4_reclaim_complete_data *calldata = data;
7742 	struct nfs_client *clp = calldata->clp;
7743 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7744 
7745 	dprintk("--> %s\n", __func__);
7746 	if (!nfs41_sequence_done(task, res))
7747 		return;
7748 
7749 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7750 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7751 		rpc_restart_call_prepare(task);
7752 		return;
7753 	}
7754 	dprintk("<-- %s\n", __func__);
7755 }
7756 
7757 static void nfs4_free_reclaim_complete_data(void *data)
7758 {
7759 	struct nfs4_reclaim_complete_data *calldata = data;
7760 
7761 	kfree(calldata);
7762 }
7763 
7764 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7765 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7766 	.rpc_call_done = nfs4_reclaim_complete_done,
7767 	.rpc_release = nfs4_free_reclaim_complete_data,
7768 };
7769 
7770 /*
7771  * Issue a global reclaim complete.
7772  */
7773 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7774 		struct rpc_cred *cred)
7775 {
7776 	struct nfs4_reclaim_complete_data *calldata;
7777 	struct rpc_task *task;
7778 	struct rpc_message msg = {
7779 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7780 		.rpc_cred = cred,
7781 	};
7782 	struct rpc_task_setup task_setup_data = {
7783 		.rpc_client = clp->cl_rpcclient,
7784 		.rpc_message = &msg,
7785 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7786 		.flags = RPC_TASK_ASYNC,
7787 	};
7788 	int status = -ENOMEM;
7789 
7790 	dprintk("--> %s\n", __func__);
7791 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7792 	if (calldata == NULL)
7793 		goto out;
7794 	calldata->clp = clp;
7795 	calldata->arg.one_fs = 0;
7796 
7797 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7798 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7799 	msg.rpc_argp = &calldata->arg;
7800 	msg.rpc_resp = &calldata->res;
7801 	task_setup_data.callback_data = calldata;
7802 	task = rpc_run_task(&task_setup_data);
7803 	if (IS_ERR(task)) {
7804 		status = PTR_ERR(task);
7805 		goto out;
7806 	}
7807 	status = nfs4_wait_for_completion_rpc_task(task);
7808 	if (status == 0)
7809 		status = task->tk_status;
7810 	rpc_put_task(task);
7811 	return 0;
7812 out:
7813 	dprintk("<-- %s status=%d\n", __func__, status);
7814 	return status;
7815 }
7816 
7817 static void
7818 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7819 {
7820 	struct nfs4_layoutget *lgp = calldata;
7821 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7822 	struct nfs4_session *session = nfs4_get_session(server);
7823 	int ret;
7824 
7825 	dprintk("--> %s\n", __func__);
7826 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7827 	 * right now covering the LAYOUTGET we are about to send.
7828 	 * However, that is not so catastrophic, and there seems
7829 	 * to be no way to prevent it completely.
7830 	 */
7831 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7832 				&lgp->res.seq_res, task))
7833 		return;
7834 	ret = pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7835 					  NFS_I(lgp->args.inode)->layout,
7836 					  &lgp->args.range,
7837 					  lgp->args.ctx->state);
7838 	if (ret < 0)
7839 		rpc_exit(task, ret);
7840 }
7841 
7842 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7843 {
7844 	struct nfs4_layoutget *lgp = calldata;
7845 	struct inode *inode = lgp->args.inode;
7846 	struct nfs_server *server = NFS_SERVER(inode);
7847 	struct pnfs_layout_hdr *lo;
7848 	struct nfs4_state *state = NULL;
7849 	unsigned long timeo, now, giveup;
7850 
7851 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7852 
7853 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7854 		goto out;
7855 
7856 	switch (task->tk_status) {
7857 	case 0:
7858 		goto out;
7859 
7860 	/*
7861 	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
7862 	 * on the file. set tk_status to -ENODATA to tell upper layer to
7863 	 * retry go inband.
7864 	 */
7865 	case -NFS4ERR_LAYOUTUNAVAILABLE:
7866 		task->tk_status = -ENODATA;
7867 		goto out;
7868 	/*
7869 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7870 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7871 	 */
7872 	case -NFS4ERR_BADLAYOUT:
7873 		goto out_overflow;
7874 	/*
7875 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7876 	 * (or clients) writing to the same RAID stripe except when
7877 	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
7878 	 */
7879 	case -NFS4ERR_LAYOUTTRYLATER:
7880 		if (lgp->args.minlength == 0)
7881 			goto out_overflow;
7882 	/*
7883 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7884 	 * existing layout before getting a new one).
7885 	 */
7886 	case -NFS4ERR_RECALLCONFLICT:
7887 		timeo = rpc_get_timeout(task->tk_client);
7888 		giveup = lgp->args.timestamp + timeo;
7889 		now = jiffies;
7890 		if (time_after(giveup, now)) {
7891 			unsigned long delay;
7892 
7893 			/* Delay for:
7894 			 * - Not less then NFS4_POLL_RETRY_MIN.
7895 			 * - One last time a jiffie before we give up
7896 			 * - exponential backoff (time_now minus start_attempt)
7897 			 */
7898 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7899 				    min((giveup - now - 1),
7900 					now - lgp->args.timestamp));
7901 
7902 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7903 				__func__, delay);
7904 			rpc_delay(task, delay);
7905 			/* Do not call nfs4_async_handle_error() */
7906 			goto out_restart;
7907 		}
7908 		break;
7909 	case -NFS4ERR_EXPIRED:
7910 	case -NFS4ERR_BAD_STATEID:
7911 		spin_lock(&inode->i_lock);
7912 		if (nfs4_stateid_match(&lgp->args.stateid,
7913 					&lgp->args.ctx->state->stateid)) {
7914 			spin_unlock(&inode->i_lock);
7915 			/* If the open stateid was bad, then recover it. */
7916 			state = lgp->args.ctx->state;
7917 			break;
7918 		}
7919 		lo = NFS_I(inode)->layout;
7920 		if (lo && nfs4_stateid_match(&lgp->args.stateid,
7921 					&lo->plh_stateid)) {
7922 			LIST_HEAD(head);
7923 
7924 			/*
7925 			 * Mark the bad layout state as invalid, then retry
7926 			 * with the current stateid.
7927 			 */
7928 			set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7929 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7930 			spin_unlock(&inode->i_lock);
7931 			pnfs_free_lseg_list(&head);
7932 		} else
7933 			spin_unlock(&inode->i_lock);
7934 		goto out_restart;
7935 	}
7936 	if (nfs4_async_handle_error(task, server, state, &lgp->timeout) == -EAGAIN)
7937 		goto out_restart;
7938 out:
7939 	dprintk("<-- %s\n", __func__);
7940 	return;
7941 out_restart:
7942 	task->tk_status = 0;
7943 	rpc_restart_call_prepare(task);
7944 	return;
7945 out_overflow:
7946 	task->tk_status = -EOVERFLOW;
7947 	goto out;
7948 }
7949 
7950 static size_t max_response_pages(struct nfs_server *server)
7951 {
7952 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7953 	return nfs_page_array_len(0, max_resp_sz);
7954 }
7955 
7956 static void nfs4_free_pages(struct page **pages, size_t size)
7957 {
7958 	int i;
7959 
7960 	if (!pages)
7961 		return;
7962 
7963 	for (i = 0; i < size; i++) {
7964 		if (!pages[i])
7965 			break;
7966 		__free_page(pages[i]);
7967 	}
7968 	kfree(pages);
7969 }
7970 
7971 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7972 {
7973 	struct page **pages;
7974 	int i;
7975 
7976 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7977 	if (!pages) {
7978 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7979 		return NULL;
7980 	}
7981 
7982 	for (i = 0; i < size; i++) {
7983 		pages[i] = alloc_page(gfp_flags);
7984 		if (!pages[i]) {
7985 			dprintk("%s: failed to allocate page\n", __func__);
7986 			nfs4_free_pages(pages, size);
7987 			return NULL;
7988 		}
7989 	}
7990 
7991 	return pages;
7992 }
7993 
7994 static void nfs4_layoutget_release(void *calldata)
7995 {
7996 	struct nfs4_layoutget *lgp = calldata;
7997 	struct inode *inode = lgp->args.inode;
7998 	struct nfs_server *server = NFS_SERVER(inode);
7999 	size_t max_pages = max_response_pages(server);
8000 
8001 	dprintk("--> %s\n", __func__);
8002 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
8003 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
8004 	put_nfs_open_context(lgp->args.ctx);
8005 	kfree(calldata);
8006 	dprintk("<-- %s\n", __func__);
8007 }
8008 
8009 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8010 	.rpc_call_prepare = nfs4_layoutget_prepare,
8011 	.rpc_call_done = nfs4_layoutget_done,
8012 	.rpc_release = nfs4_layoutget_release,
8013 };
8014 
8015 struct pnfs_layout_segment *
8016 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
8017 {
8018 	struct inode *inode = lgp->args.inode;
8019 	struct nfs_server *server = NFS_SERVER(inode);
8020 	size_t max_pages = max_response_pages(server);
8021 	struct rpc_task *task;
8022 	struct rpc_message msg = {
8023 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8024 		.rpc_argp = &lgp->args,
8025 		.rpc_resp = &lgp->res,
8026 		.rpc_cred = lgp->cred,
8027 	};
8028 	struct rpc_task_setup task_setup_data = {
8029 		.rpc_client = server->client,
8030 		.rpc_message = &msg,
8031 		.callback_ops = &nfs4_layoutget_call_ops,
8032 		.callback_data = lgp,
8033 		.flags = RPC_TASK_ASYNC,
8034 	};
8035 	struct pnfs_layout_segment *lseg = NULL;
8036 	int status = 0;
8037 
8038 	dprintk("--> %s\n", __func__);
8039 
8040 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8041 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
8042 
8043 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8044 	if (!lgp->args.layout.pages) {
8045 		nfs4_layoutget_release(lgp);
8046 		return ERR_PTR(-ENOMEM);
8047 	}
8048 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8049 	lgp->args.timestamp = jiffies;
8050 
8051 	lgp->res.layoutp = &lgp->args.layout;
8052 	lgp->res.seq_res.sr_slot = NULL;
8053 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8054 
8055 	task = rpc_run_task(&task_setup_data);
8056 	if (IS_ERR(task))
8057 		return ERR_CAST(task);
8058 	status = nfs4_wait_for_completion_rpc_task(task);
8059 	if (status == 0)
8060 		status = task->tk_status;
8061 	trace_nfs4_layoutget(lgp->args.ctx,
8062 			&lgp->args.range,
8063 			&lgp->res.range,
8064 			&lgp->res.stateid,
8065 			status);
8066 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8067 	if (status == 0 && lgp->res.layoutp->len)
8068 		lseg = pnfs_layout_process(lgp);
8069 	rpc_put_task(task);
8070 	dprintk("<-- %s status=%d\n", __func__, status);
8071 	if (status)
8072 		return ERR_PTR(status);
8073 	return lseg;
8074 }
8075 
8076 static void
8077 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8078 {
8079 	struct nfs4_layoutreturn *lrp = calldata;
8080 
8081 	dprintk("--> %s\n", __func__);
8082 	nfs41_setup_sequence(lrp->clp->cl_session,
8083 			&lrp->args.seq_args,
8084 			&lrp->res.seq_res,
8085 			task);
8086 }
8087 
8088 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8089 {
8090 	struct nfs4_layoutreturn *lrp = calldata;
8091 	struct nfs_server *server;
8092 
8093 	dprintk("--> %s\n", __func__);
8094 
8095 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8096 		return;
8097 
8098 	server = NFS_SERVER(lrp->args.inode);
8099 	switch (task->tk_status) {
8100 	default:
8101 		task->tk_status = 0;
8102 	case 0:
8103 		break;
8104 	case -NFS4ERR_DELAY:
8105 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8106 			break;
8107 		rpc_restart_call_prepare(task);
8108 		return;
8109 	}
8110 	dprintk("<-- %s\n", __func__);
8111 }
8112 
8113 static void nfs4_layoutreturn_release(void *calldata)
8114 {
8115 	struct nfs4_layoutreturn *lrp = calldata;
8116 	struct pnfs_layout_hdr *lo = lrp->args.layout;
8117 	LIST_HEAD(freeme);
8118 
8119 	dprintk("--> %s\n", __func__);
8120 	spin_lock(&lo->plh_inode->i_lock);
8121 	pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8122 	pnfs_mark_layout_returned_if_empty(lo);
8123 	if (lrp->res.lrs_present)
8124 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8125 	pnfs_clear_layoutreturn_waitbit(lo);
8126 	spin_unlock(&lo->plh_inode->i_lock);
8127 	pnfs_free_lseg_list(&freeme);
8128 	pnfs_put_layout_hdr(lrp->args.layout);
8129 	nfs_iput_and_deactive(lrp->inode);
8130 	kfree(calldata);
8131 	dprintk("<-- %s\n", __func__);
8132 }
8133 
8134 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8135 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
8136 	.rpc_call_done = nfs4_layoutreturn_done,
8137 	.rpc_release = nfs4_layoutreturn_release,
8138 };
8139 
8140 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8141 {
8142 	struct rpc_task *task;
8143 	struct rpc_message msg = {
8144 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8145 		.rpc_argp = &lrp->args,
8146 		.rpc_resp = &lrp->res,
8147 		.rpc_cred = lrp->cred,
8148 	};
8149 	struct rpc_task_setup task_setup_data = {
8150 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
8151 		.rpc_message = &msg,
8152 		.callback_ops = &nfs4_layoutreturn_call_ops,
8153 		.callback_data = lrp,
8154 	};
8155 	int status = 0;
8156 
8157 	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8158 			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8159 			&task_setup_data.rpc_client, &msg);
8160 
8161 	dprintk("--> %s\n", __func__);
8162 	if (!sync) {
8163 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8164 		if (!lrp->inode) {
8165 			nfs4_layoutreturn_release(lrp);
8166 			return -EAGAIN;
8167 		}
8168 		task_setup_data.flags |= RPC_TASK_ASYNC;
8169 	}
8170 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8171 	task = rpc_run_task(&task_setup_data);
8172 	if (IS_ERR(task))
8173 		return PTR_ERR(task);
8174 	if (sync)
8175 		status = task->tk_status;
8176 	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8177 	dprintk("<-- %s status=%d\n", __func__, status);
8178 	rpc_put_task(task);
8179 	return status;
8180 }
8181 
8182 static int
8183 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8184 		struct pnfs_device *pdev,
8185 		struct rpc_cred *cred)
8186 {
8187 	struct nfs4_getdeviceinfo_args args = {
8188 		.pdev = pdev,
8189 		.notify_types = NOTIFY_DEVICEID4_CHANGE |
8190 			NOTIFY_DEVICEID4_DELETE,
8191 	};
8192 	struct nfs4_getdeviceinfo_res res = {
8193 		.pdev = pdev,
8194 	};
8195 	struct rpc_message msg = {
8196 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8197 		.rpc_argp = &args,
8198 		.rpc_resp = &res,
8199 		.rpc_cred = cred,
8200 	};
8201 	int status;
8202 
8203 	dprintk("--> %s\n", __func__);
8204 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8205 	if (res.notification & ~args.notify_types)
8206 		dprintk("%s: unsupported notification\n", __func__);
8207 	if (res.notification != args.notify_types)
8208 		pdev->nocache = 1;
8209 
8210 	dprintk("<-- %s status=%d\n", __func__, status);
8211 
8212 	return status;
8213 }
8214 
8215 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8216 		struct pnfs_device *pdev,
8217 		struct rpc_cred *cred)
8218 {
8219 	struct nfs4_exception exception = { };
8220 	int err;
8221 
8222 	do {
8223 		err = nfs4_handle_exception(server,
8224 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
8225 					&exception);
8226 	} while (exception.retry);
8227 	return err;
8228 }
8229 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8230 
8231 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8232 {
8233 	struct nfs4_layoutcommit_data *data = calldata;
8234 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8235 	struct nfs4_session *session = nfs4_get_session(server);
8236 
8237 	nfs41_setup_sequence(session,
8238 			&data->args.seq_args,
8239 			&data->res.seq_res,
8240 			task);
8241 }
8242 
8243 static void
8244 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8245 {
8246 	struct nfs4_layoutcommit_data *data = calldata;
8247 	struct nfs_server *server = NFS_SERVER(data->args.inode);
8248 
8249 	if (!nfs41_sequence_done(task, &data->res.seq_res))
8250 		return;
8251 
8252 	switch (task->tk_status) { /* Just ignore these failures */
8253 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8254 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8255 	case -NFS4ERR_BADLAYOUT:     /* no layout */
8256 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
8257 		task->tk_status = 0;
8258 	case 0:
8259 		break;
8260 	default:
8261 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8262 			rpc_restart_call_prepare(task);
8263 			return;
8264 		}
8265 	}
8266 }
8267 
8268 static void nfs4_layoutcommit_release(void *calldata)
8269 {
8270 	struct nfs4_layoutcommit_data *data = calldata;
8271 
8272 	pnfs_cleanup_layoutcommit(data);
8273 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8274 					   data->res.fattr);
8275 	put_rpccred(data->cred);
8276 	nfs_iput_and_deactive(data->inode);
8277 	kfree(data);
8278 }
8279 
8280 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8281 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8282 	.rpc_call_done = nfs4_layoutcommit_done,
8283 	.rpc_release = nfs4_layoutcommit_release,
8284 };
8285 
8286 int
8287 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8288 {
8289 	struct rpc_message msg = {
8290 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8291 		.rpc_argp = &data->args,
8292 		.rpc_resp = &data->res,
8293 		.rpc_cred = data->cred,
8294 	};
8295 	struct rpc_task_setup task_setup_data = {
8296 		.task = &data->task,
8297 		.rpc_client = NFS_CLIENT(data->args.inode),
8298 		.rpc_message = &msg,
8299 		.callback_ops = &nfs4_layoutcommit_ops,
8300 		.callback_data = data,
8301 	};
8302 	struct rpc_task *task;
8303 	int status = 0;
8304 
8305 	dprintk("NFS: initiating layoutcommit call. sync %d "
8306 		"lbw: %llu inode %lu\n", sync,
8307 		data->args.lastbytewritten,
8308 		data->args.inode->i_ino);
8309 
8310 	if (!sync) {
8311 		data->inode = nfs_igrab_and_active(data->args.inode);
8312 		if (data->inode == NULL) {
8313 			nfs4_layoutcommit_release(data);
8314 			return -EAGAIN;
8315 		}
8316 		task_setup_data.flags = RPC_TASK_ASYNC;
8317 	}
8318 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8319 	task = rpc_run_task(&task_setup_data);
8320 	if (IS_ERR(task))
8321 		return PTR_ERR(task);
8322 	if (sync)
8323 		status = task->tk_status;
8324 	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8325 	dprintk("%s: status %d\n", __func__, status);
8326 	rpc_put_task(task);
8327 	return status;
8328 }
8329 
8330 /**
8331  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8332  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8333  */
8334 static int
8335 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8336 		    struct nfs_fsinfo *info,
8337 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8338 {
8339 	struct nfs41_secinfo_no_name_args args = {
8340 		.style = SECINFO_STYLE_CURRENT_FH,
8341 	};
8342 	struct nfs4_secinfo_res res = {
8343 		.flavors = flavors,
8344 	};
8345 	struct rpc_message msg = {
8346 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8347 		.rpc_argp = &args,
8348 		.rpc_resp = &res,
8349 	};
8350 	struct rpc_clnt *clnt = server->client;
8351 	struct rpc_cred *cred = NULL;
8352 	int status;
8353 
8354 	if (use_integrity) {
8355 		clnt = server->nfs_client->cl_rpcclient;
8356 		cred = nfs4_get_clid_cred(server->nfs_client);
8357 		msg.rpc_cred = cred;
8358 	}
8359 
8360 	dprintk("--> %s\n", __func__);
8361 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8362 				&res.seq_res, 0);
8363 	dprintk("<-- %s status=%d\n", __func__, status);
8364 
8365 	if (cred)
8366 		put_rpccred(cred);
8367 
8368 	return status;
8369 }
8370 
8371 static int
8372 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8373 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8374 {
8375 	struct nfs4_exception exception = { };
8376 	int err;
8377 	do {
8378 		/* first try using integrity protection */
8379 		err = -NFS4ERR_WRONGSEC;
8380 
8381 		/* try to use integrity protection with machine cred */
8382 		if (_nfs4_is_integrity_protected(server->nfs_client))
8383 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8384 							  flavors, true);
8385 
8386 		/*
8387 		 * if unable to use integrity protection, or SECINFO with
8388 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8389 		 * disallowed by spec, but exists in deployed servers) use
8390 		 * the current filesystem's rpc_client and the user cred.
8391 		 */
8392 		if (err == -NFS4ERR_WRONGSEC)
8393 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8394 							  flavors, false);
8395 
8396 		switch (err) {
8397 		case 0:
8398 		case -NFS4ERR_WRONGSEC:
8399 		case -ENOTSUPP:
8400 			goto out;
8401 		default:
8402 			err = nfs4_handle_exception(server, err, &exception);
8403 		}
8404 	} while (exception.retry);
8405 out:
8406 	return err;
8407 }
8408 
8409 static int
8410 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8411 		    struct nfs_fsinfo *info)
8412 {
8413 	int err;
8414 	struct page *page;
8415 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8416 	struct nfs4_secinfo_flavors *flavors;
8417 	struct nfs4_secinfo4 *secinfo;
8418 	int i;
8419 
8420 	page = alloc_page(GFP_KERNEL);
8421 	if (!page) {
8422 		err = -ENOMEM;
8423 		goto out;
8424 	}
8425 
8426 	flavors = page_address(page);
8427 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8428 
8429 	/*
8430 	 * Fall back on "guess and check" method if
8431 	 * the server doesn't support SECINFO_NO_NAME
8432 	 */
8433 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8434 		err = nfs4_find_root_sec(server, fhandle, info);
8435 		goto out_freepage;
8436 	}
8437 	if (err)
8438 		goto out_freepage;
8439 
8440 	for (i = 0; i < flavors->num_flavors; i++) {
8441 		secinfo = &flavors->flavors[i];
8442 
8443 		switch (secinfo->flavor) {
8444 		case RPC_AUTH_NULL:
8445 		case RPC_AUTH_UNIX:
8446 		case RPC_AUTH_GSS:
8447 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8448 					&secinfo->flavor_info);
8449 			break;
8450 		default:
8451 			flavor = RPC_AUTH_MAXFLAVOR;
8452 			break;
8453 		}
8454 
8455 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8456 			flavor = RPC_AUTH_MAXFLAVOR;
8457 
8458 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8459 			err = nfs4_lookup_root_sec(server, fhandle,
8460 						   info, flavor);
8461 			if (!err)
8462 				break;
8463 		}
8464 	}
8465 
8466 	if (flavor == RPC_AUTH_MAXFLAVOR)
8467 		err = -EPERM;
8468 
8469 out_freepage:
8470 	put_page(page);
8471 	if (err == -EACCES)
8472 		return -EPERM;
8473 out:
8474 	return err;
8475 }
8476 
8477 static int _nfs41_test_stateid(struct nfs_server *server,
8478 		nfs4_stateid *stateid,
8479 		struct rpc_cred *cred)
8480 {
8481 	int status;
8482 	struct nfs41_test_stateid_args args = {
8483 		.stateid = stateid,
8484 	};
8485 	struct nfs41_test_stateid_res res;
8486 	struct rpc_message msg = {
8487 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8488 		.rpc_argp = &args,
8489 		.rpc_resp = &res,
8490 		.rpc_cred = cred,
8491 	};
8492 	struct rpc_clnt *rpc_client = server->client;
8493 
8494 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8495 		&rpc_client, &msg);
8496 
8497 	dprintk("NFS call  test_stateid %p\n", stateid);
8498 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8499 	nfs4_set_sequence_privileged(&args.seq_args);
8500 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8501 			&args.seq_args, &res.seq_res);
8502 	if (status != NFS_OK) {
8503 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8504 		return status;
8505 	}
8506 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8507 	return -res.status;
8508 }
8509 
8510 /**
8511  * nfs41_test_stateid - perform a TEST_STATEID operation
8512  *
8513  * @server: server / transport on which to perform the operation
8514  * @stateid: state ID to test
8515  * @cred: credential
8516  *
8517  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8518  * Otherwise a negative NFS4ERR value is returned if the operation
8519  * failed or the state ID is not currently valid.
8520  */
8521 static int nfs41_test_stateid(struct nfs_server *server,
8522 		nfs4_stateid *stateid,
8523 		struct rpc_cred *cred)
8524 {
8525 	struct nfs4_exception exception = { };
8526 	int err;
8527 	do {
8528 		err = _nfs41_test_stateid(server, stateid, cred);
8529 		if (err != -NFS4ERR_DELAY)
8530 			break;
8531 		nfs4_handle_exception(server, err, &exception);
8532 	} while (exception.retry);
8533 	return err;
8534 }
8535 
8536 struct nfs_free_stateid_data {
8537 	struct nfs_server *server;
8538 	struct nfs41_free_stateid_args args;
8539 	struct nfs41_free_stateid_res res;
8540 };
8541 
8542 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8543 {
8544 	struct nfs_free_stateid_data *data = calldata;
8545 	nfs41_setup_sequence(nfs4_get_session(data->server),
8546 			&data->args.seq_args,
8547 			&data->res.seq_res,
8548 			task);
8549 }
8550 
8551 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8552 {
8553 	struct nfs_free_stateid_data *data = calldata;
8554 
8555 	nfs41_sequence_done(task, &data->res.seq_res);
8556 
8557 	switch (task->tk_status) {
8558 	case -NFS4ERR_DELAY:
8559 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8560 			rpc_restart_call_prepare(task);
8561 	}
8562 }
8563 
8564 static void nfs41_free_stateid_release(void *calldata)
8565 {
8566 	kfree(calldata);
8567 }
8568 
8569 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8570 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8571 	.rpc_call_done = nfs41_free_stateid_done,
8572 	.rpc_release = nfs41_free_stateid_release,
8573 };
8574 
8575 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8576 		nfs4_stateid *stateid,
8577 		struct rpc_cred *cred,
8578 		bool privileged)
8579 {
8580 	struct rpc_message msg = {
8581 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8582 		.rpc_cred = cred,
8583 	};
8584 	struct rpc_task_setup task_setup = {
8585 		.rpc_client = server->client,
8586 		.rpc_message = &msg,
8587 		.callback_ops = &nfs41_free_stateid_ops,
8588 		.flags = RPC_TASK_ASYNC,
8589 	};
8590 	struct nfs_free_stateid_data *data;
8591 
8592 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8593 		&task_setup.rpc_client, &msg);
8594 
8595 	dprintk("NFS call  free_stateid %p\n", stateid);
8596 	data = kmalloc(sizeof(*data), GFP_NOFS);
8597 	if (!data)
8598 		return ERR_PTR(-ENOMEM);
8599 	data->server = server;
8600 	nfs4_stateid_copy(&data->args.stateid, stateid);
8601 
8602 	task_setup.callback_data = data;
8603 
8604 	msg.rpc_argp = &data->args;
8605 	msg.rpc_resp = &data->res;
8606 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8607 	if (privileged)
8608 		nfs4_set_sequence_privileged(&data->args.seq_args);
8609 
8610 	return rpc_run_task(&task_setup);
8611 }
8612 
8613 /**
8614  * nfs41_free_stateid - perform a FREE_STATEID operation
8615  *
8616  * @server: server / transport on which to perform the operation
8617  * @stateid: state ID to release
8618  * @cred: credential
8619  *
8620  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8621  * negative NFS4ERR value is returned.
8622  */
8623 static int nfs41_free_stateid(struct nfs_server *server,
8624 		nfs4_stateid *stateid,
8625 		struct rpc_cred *cred)
8626 {
8627 	struct rpc_task *task;
8628 	int ret;
8629 
8630 	task = _nfs41_free_stateid(server, stateid, cred, true);
8631 	if (IS_ERR(task))
8632 		return PTR_ERR(task);
8633 	ret = rpc_wait_for_completion_task(task);
8634 	if (!ret)
8635 		ret = task->tk_status;
8636 	rpc_put_task(task);
8637 	return ret;
8638 }
8639 
8640 static void
8641 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8642 {
8643 	struct rpc_task *task;
8644 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8645 
8646 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8647 	nfs4_free_lock_state(server, lsp);
8648 	if (IS_ERR(task))
8649 		return;
8650 	rpc_put_task(task);
8651 }
8652 
8653 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8654 		const nfs4_stateid *s2)
8655 {
8656 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8657 		return false;
8658 
8659 	if (s1->seqid == s2->seqid)
8660 		return true;
8661 	if (s1->seqid == 0 || s2->seqid == 0)
8662 		return true;
8663 
8664 	return false;
8665 }
8666 
8667 #endif /* CONFIG_NFS_V4_1 */
8668 
8669 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8670 		const nfs4_stateid *s2)
8671 {
8672 	return nfs4_stateid_match(s1, s2);
8673 }
8674 
8675 
8676 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8677 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8678 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8679 	.recover_open	= nfs4_open_reclaim,
8680 	.recover_lock	= nfs4_lock_reclaim,
8681 	.establish_clid = nfs4_init_clientid,
8682 	.detect_trunking = nfs40_discover_server_trunking,
8683 };
8684 
8685 #if defined(CONFIG_NFS_V4_1)
8686 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8687 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8688 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8689 	.recover_open	= nfs4_open_reclaim,
8690 	.recover_lock	= nfs4_lock_reclaim,
8691 	.establish_clid = nfs41_init_clientid,
8692 	.reclaim_complete = nfs41_proc_reclaim_complete,
8693 	.detect_trunking = nfs41_discover_server_trunking,
8694 };
8695 #endif /* CONFIG_NFS_V4_1 */
8696 
8697 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8698 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8699 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8700 	.recover_open	= nfs40_open_expired,
8701 	.recover_lock	= nfs4_lock_expired,
8702 	.establish_clid = nfs4_init_clientid,
8703 };
8704 
8705 #if defined(CONFIG_NFS_V4_1)
8706 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8707 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8708 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8709 	.recover_open	= nfs41_open_expired,
8710 	.recover_lock	= nfs41_lock_expired,
8711 	.establish_clid = nfs41_init_clientid,
8712 };
8713 #endif /* CONFIG_NFS_V4_1 */
8714 
8715 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8716 	.sched_state_renewal = nfs4_proc_async_renew,
8717 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8718 	.renew_lease = nfs4_proc_renew,
8719 };
8720 
8721 #if defined(CONFIG_NFS_V4_1)
8722 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8723 	.sched_state_renewal = nfs41_proc_async_sequence,
8724 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8725 	.renew_lease = nfs4_proc_sequence,
8726 };
8727 #endif
8728 
8729 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8730 	.get_locations = _nfs40_proc_get_locations,
8731 	.fsid_present = _nfs40_proc_fsid_present,
8732 };
8733 
8734 #if defined(CONFIG_NFS_V4_1)
8735 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8736 	.get_locations = _nfs41_proc_get_locations,
8737 	.fsid_present = _nfs41_proc_fsid_present,
8738 };
8739 #endif	/* CONFIG_NFS_V4_1 */
8740 
8741 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8742 	.minor_version = 0,
8743 	.init_caps = NFS_CAP_READDIRPLUS
8744 		| NFS_CAP_ATOMIC_OPEN
8745 		| NFS_CAP_POSIX_LOCK,
8746 	.init_client = nfs40_init_client,
8747 	.shutdown_client = nfs40_shutdown_client,
8748 	.match_stateid = nfs4_match_stateid,
8749 	.find_root_sec = nfs4_find_root_sec,
8750 	.free_lock_state = nfs4_release_lockowner,
8751 	.alloc_seqid = nfs_alloc_seqid,
8752 	.call_sync_ops = &nfs40_call_sync_ops,
8753 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8754 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8755 	.state_renewal_ops = &nfs40_state_renewal_ops,
8756 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8757 };
8758 
8759 #if defined(CONFIG_NFS_V4_1)
8760 static struct nfs_seqid *
8761 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8762 {
8763 	return NULL;
8764 }
8765 
8766 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8767 	.minor_version = 1,
8768 	.init_caps = NFS_CAP_READDIRPLUS
8769 		| NFS_CAP_ATOMIC_OPEN
8770 		| NFS_CAP_POSIX_LOCK
8771 		| NFS_CAP_STATEID_NFSV41
8772 		| NFS_CAP_ATOMIC_OPEN_V1,
8773 	.init_client = nfs41_init_client,
8774 	.shutdown_client = nfs41_shutdown_client,
8775 	.match_stateid = nfs41_match_stateid,
8776 	.find_root_sec = nfs41_find_root_sec,
8777 	.free_lock_state = nfs41_free_lock_state,
8778 	.alloc_seqid = nfs_alloc_no_seqid,
8779 	.call_sync_ops = &nfs41_call_sync_ops,
8780 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8781 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8782 	.state_renewal_ops = &nfs41_state_renewal_ops,
8783 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8784 };
8785 #endif
8786 
8787 #if defined(CONFIG_NFS_V4_2)
8788 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8789 	.minor_version = 2,
8790 	.init_caps = NFS_CAP_READDIRPLUS
8791 		| NFS_CAP_ATOMIC_OPEN
8792 		| NFS_CAP_POSIX_LOCK
8793 		| NFS_CAP_STATEID_NFSV41
8794 		| NFS_CAP_ATOMIC_OPEN_V1
8795 		| NFS_CAP_ALLOCATE
8796 		| NFS_CAP_DEALLOCATE
8797 		| NFS_CAP_SEEK
8798 		| NFS_CAP_LAYOUTSTATS
8799 		| NFS_CAP_CLONE,
8800 	.init_client = nfs41_init_client,
8801 	.shutdown_client = nfs41_shutdown_client,
8802 	.match_stateid = nfs41_match_stateid,
8803 	.find_root_sec = nfs41_find_root_sec,
8804 	.free_lock_state = nfs41_free_lock_state,
8805 	.call_sync_ops = &nfs41_call_sync_ops,
8806 	.alloc_seqid = nfs_alloc_no_seqid,
8807 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8808 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8809 	.state_renewal_ops = &nfs41_state_renewal_ops,
8810 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8811 };
8812 #endif
8813 
8814 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8815 	[0] = &nfs_v4_0_minor_ops,
8816 #if defined(CONFIG_NFS_V4_1)
8817 	[1] = &nfs_v4_1_minor_ops,
8818 #endif
8819 #if defined(CONFIG_NFS_V4_2)
8820 	[2] = &nfs_v4_2_minor_ops,
8821 #endif
8822 };
8823 
8824 ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
8825 {
8826 	ssize_t error, error2;
8827 
8828 	error = generic_listxattr(dentry, list, size);
8829 	if (error < 0)
8830 		return error;
8831 	if (list) {
8832 		list += error;
8833 		size -= error;
8834 	}
8835 
8836 	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
8837 	if (error2 < 0)
8838 		return error2;
8839 	return error + error2;
8840 }
8841 
8842 static const struct inode_operations nfs4_dir_inode_operations = {
8843 	.create		= nfs_create,
8844 	.lookup		= nfs_lookup,
8845 	.atomic_open	= nfs_atomic_open,
8846 	.link		= nfs_link,
8847 	.unlink		= nfs_unlink,
8848 	.symlink	= nfs_symlink,
8849 	.mkdir		= nfs_mkdir,
8850 	.rmdir		= nfs_rmdir,
8851 	.mknod		= nfs_mknod,
8852 	.rename		= nfs_rename,
8853 	.permission	= nfs_permission,
8854 	.getattr	= nfs_getattr,
8855 	.setattr	= nfs_setattr,
8856 	.getxattr	= generic_getxattr,
8857 	.setxattr	= generic_setxattr,
8858 	.listxattr	= nfs4_listxattr,
8859 	.removexattr	= generic_removexattr,
8860 };
8861 
8862 static const struct inode_operations nfs4_file_inode_operations = {
8863 	.permission	= nfs_permission,
8864 	.getattr	= nfs_getattr,
8865 	.setattr	= nfs_setattr,
8866 	.getxattr	= generic_getxattr,
8867 	.setxattr	= generic_setxattr,
8868 	.listxattr	= nfs4_listxattr,
8869 	.removexattr	= generic_removexattr,
8870 };
8871 
8872 const struct nfs_rpc_ops nfs_v4_clientops = {
8873 	.version	= 4,			/* protocol version */
8874 	.dentry_ops	= &nfs4_dentry_operations,
8875 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8876 	.file_inode_ops	= &nfs4_file_inode_operations,
8877 	.file_ops	= &nfs4_file_operations,
8878 	.getroot	= nfs4_proc_get_root,
8879 	.submount	= nfs4_submount,
8880 	.try_mount	= nfs4_try_mount,
8881 	.getattr	= nfs4_proc_getattr,
8882 	.setattr	= nfs4_proc_setattr,
8883 	.lookup		= nfs4_proc_lookup,
8884 	.access		= nfs4_proc_access,
8885 	.readlink	= nfs4_proc_readlink,
8886 	.create		= nfs4_proc_create,
8887 	.remove		= nfs4_proc_remove,
8888 	.unlink_setup	= nfs4_proc_unlink_setup,
8889 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8890 	.unlink_done	= nfs4_proc_unlink_done,
8891 	.rename_setup	= nfs4_proc_rename_setup,
8892 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8893 	.rename_done	= nfs4_proc_rename_done,
8894 	.link		= nfs4_proc_link,
8895 	.symlink	= nfs4_proc_symlink,
8896 	.mkdir		= nfs4_proc_mkdir,
8897 	.rmdir		= nfs4_proc_remove,
8898 	.readdir	= nfs4_proc_readdir,
8899 	.mknod		= nfs4_proc_mknod,
8900 	.statfs		= nfs4_proc_statfs,
8901 	.fsinfo		= nfs4_proc_fsinfo,
8902 	.pathconf	= nfs4_proc_pathconf,
8903 	.set_capabilities = nfs4_server_capabilities,
8904 	.decode_dirent	= nfs4_decode_dirent,
8905 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8906 	.read_setup	= nfs4_proc_read_setup,
8907 	.read_done	= nfs4_read_done,
8908 	.write_setup	= nfs4_proc_write_setup,
8909 	.write_done	= nfs4_write_done,
8910 	.commit_setup	= nfs4_proc_commit_setup,
8911 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8912 	.commit_done	= nfs4_commit_done,
8913 	.lock		= nfs4_proc_lock,
8914 	.clear_acl_cache = nfs4_zap_acl_attr,
8915 	.close_context  = nfs4_close_context,
8916 	.open_context	= nfs4_atomic_open,
8917 	.have_delegation = nfs4_have_delegation,
8918 	.return_delegation = nfs4_inode_return_delegation,
8919 	.alloc_client	= nfs4_alloc_client,
8920 	.init_client	= nfs4_init_client,
8921 	.free_client	= nfs4_free_client,
8922 	.create_server	= nfs4_create_server,
8923 	.clone_server	= nfs_clone_server,
8924 };
8925 
8926 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8927 	.name	= XATTR_NAME_NFSV4_ACL,
8928 	.list	= nfs4_xattr_list_nfs4_acl,
8929 	.get	= nfs4_xattr_get_nfs4_acl,
8930 	.set	= nfs4_xattr_set_nfs4_acl,
8931 };
8932 
8933 const struct xattr_handler *nfs4_xattr_handlers[] = {
8934 	&nfs4_xattr_nfs4_acl_handler,
8935 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8936 	&nfs4_xattr_nfs4_label_handler,
8937 #endif
8938 	NULL
8939 };
8940 
8941 /*
8942  * Local variables:
8943  *  c-basic-offset: 8
8944  * End:
8945  */
8946