xref: /openbmc/linux/fs/nfs/nfs4proc.c (revision 7eec52db361a6ae6fbbd86c2299718586866b664)
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/nfs_idmap.h>
55 #include <linux/xattr.h>
56 #include <linux/utsname.h>
57 #include <linux/freezer.h>
58 
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68 
69 #include "nfs4trace.h"
70 
71 #define NFSDBG_FACILITY		NFSDBG_PROC
72 
73 #define NFS4_POLL_RETRY_MIN	(HZ/10)
74 #define NFS4_POLL_RETRY_MAX	(15*HZ)
75 
76 struct nfs4_opendata;
77 static int _nfs4_proc_open(struct nfs4_opendata *data);
78 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
79 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
80 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
85 			    struct nfs_fattr *fattr, struct iattr *sattr,
86 			    struct nfs4_state *state, struct nfs4_label *ilabel,
87 			    struct nfs4_label *olabel);
88 #ifdef CONFIG_NFS_V4_1
89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
90 		struct rpc_cred *);
91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
92 		struct rpc_cred *);
93 #endif
94 
95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
96 static inline struct nfs4_label *
97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
98 	struct iattr *sattr, struct nfs4_label *label)
99 {
100 	int err;
101 
102 	if (label == NULL)
103 		return NULL;
104 
105 	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106 		return NULL;
107 
108 	err = security_dentry_init_security(dentry, sattr->ia_mode,
109 				&dentry->d_name, (void **)&label->label, &label->len);
110 	if (err == 0)
111 		return label;
112 
113 	return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118 	if (label)
119 		security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123 	if (label)
124 		return server->attr_bitmask;
125 
126 	return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131 	struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144 	if (err >= -1000)
145 		return err;
146 	switch (err) {
147 	case -NFS4ERR_RESOURCE:
148 	case -NFS4ERR_LAYOUTTRYLATER:
149 	case -NFS4ERR_RECALLCONFLICT:
150 		return -EREMOTEIO;
151 	case -NFS4ERR_WRONGSEC:
152 	case -NFS4ERR_WRONG_CRED:
153 		return -EPERM;
154 	case -NFS4ERR_BADOWNER:
155 	case -NFS4ERR_BADNAME:
156 		return -EINVAL;
157 	case -NFS4ERR_SHARE_DENIED:
158 		return -EACCES;
159 	case -NFS4ERR_MINOR_VERS_MISMATCH:
160 		return -EPROTONOSUPPORT;
161 	case -NFS4ERR_FILE_OPEN:
162 		return -EBUSY;
163 	default:
164 		dprintk("%s could not handle NFSv4 error %d\n",
165 				__func__, -err);
166 		break;
167 	}
168 	return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175 	FATTR4_WORD0_TYPE
176 	| FATTR4_WORD0_CHANGE
177 	| FATTR4_WORD0_SIZE
178 	| FATTR4_WORD0_FSID
179 	| FATTR4_WORD0_FILEID,
180 	FATTR4_WORD1_MODE
181 	| FATTR4_WORD1_NUMLINKS
182 	| FATTR4_WORD1_OWNER
183 	| FATTR4_WORD1_OWNER_GROUP
184 	| FATTR4_WORD1_RAWDEV
185 	| FATTR4_WORD1_SPACE_USED
186 	| FATTR4_WORD1_TIME_ACCESS
187 	| FATTR4_WORD1_TIME_METADATA
188 	| FATTR4_WORD1_TIME_MODIFY,
189 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
190 	FATTR4_WORD2_SECURITY_LABEL
191 #endif
192 };
193 
194 static const u32 nfs4_pnfs_open_bitmap[3] = {
195 	FATTR4_WORD0_TYPE
196 	| FATTR4_WORD0_CHANGE
197 	| FATTR4_WORD0_SIZE
198 	| FATTR4_WORD0_FSID
199 	| FATTR4_WORD0_FILEID,
200 	FATTR4_WORD1_MODE
201 	| FATTR4_WORD1_NUMLINKS
202 	| FATTR4_WORD1_OWNER
203 	| FATTR4_WORD1_OWNER_GROUP
204 	| FATTR4_WORD1_RAWDEV
205 	| FATTR4_WORD1_SPACE_USED
206 	| FATTR4_WORD1_TIME_ACCESS
207 	| FATTR4_WORD1_TIME_METADATA
208 	| FATTR4_WORD1_TIME_MODIFY,
209 	FATTR4_WORD2_MDSTHRESHOLD
210 };
211 
212 static const u32 nfs4_open_noattr_bitmap[3] = {
213 	FATTR4_WORD0_TYPE
214 	| FATTR4_WORD0_CHANGE
215 	| FATTR4_WORD0_FILEID,
216 };
217 
218 const u32 nfs4_statfs_bitmap[3] = {
219 	FATTR4_WORD0_FILES_AVAIL
220 	| FATTR4_WORD0_FILES_FREE
221 	| FATTR4_WORD0_FILES_TOTAL,
222 	FATTR4_WORD1_SPACE_AVAIL
223 	| FATTR4_WORD1_SPACE_FREE
224 	| FATTR4_WORD1_SPACE_TOTAL
225 };
226 
227 const u32 nfs4_pathconf_bitmap[3] = {
228 	FATTR4_WORD0_MAXLINK
229 	| FATTR4_WORD0_MAXNAME,
230 	0
231 };
232 
233 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
234 			| FATTR4_WORD0_MAXREAD
235 			| FATTR4_WORD0_MAXWRITE
236 			| FATTR4_WORD0_LEASE_TIME,
237 			FATTR4_WORD1_TIME_DELTA
238 			| FATTR4_WORD1_FS_LAYOUT_TYPES,
239 			FATTR4_WORD2_LAYOUT_BLKSIZE
240 };
241 
242 const u32 nfs4_fs_locations_bitmap[3] = {
243 	FATTR4_WORD0_TYPE
244 	| FATTR4_WORD0_CHANGE
245 	| FATTR4_WORD0_SIZE
246 	| FATTR4_WORD0_FSID
247 	| FATTR4_WORD0_FILEID
248 	| FATTR4_WORD0_FS_LOCATIONS,
249 	FATTR4_WORD1_MODE
250 	| FATTR4_WORD1_NUMLINKS
251 	| FATTR4_WORD1_OWNER
252 	| FATTR4_WORD1_OWNER_GROUP
253 	| FATTR4_WORD1_RAWDEV
254 	| FATTR4_WORD1_SPACE_USED
255 	| FATTR4_WORD1_TIME_ACCESS
256 	| FATTR4_WORD1_TIME_METADATA
257 	| FATTR4_WORD1_TIME_MODIFY
258 	| FATTR4_WORD1_MOUNTED_ON_FILEID,
259 };
260 
261 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
262 		struct nfs4_readdir_arg *readdir)
263 {
264 	__be32 *start, *p;
265 
266 	if (cookie > 2) {
267 		readdir->cookie = cookie;
268 		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
269 		return;
270 	}
271 
272 	readdir->cookie = 0;
273 	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
274 	if (cookie == 2)
275 		return;
276 
277 	/*
278 	 * NFSv4 servers do not return entries for '.' and '..'
279 	 * Therefore, we fake these entries here.  We let '.'
280 	 * have cookie 0 and '..' have cookie 1.  Note that
281 	 * when talking to the server, we always send cookie 0
282 	 * instead of 1 or 2.
283 	 */
284 	start = p = kmap_atomic(*readdir->pages);
285 
286 	if (cookie == 0) {
287 		*p++ = xdr_one;                                  /* next */
288 		*p++ = xdr_zero;                   /* cookie, first word */
289 		*p++ = xdr_one;                   /* cookie, second word */
290 		*p++ = xdr_one;                             /* entry len */
291 		memcpy(p, ".\0\0\0", 4);                        /* entry */
292 		p++;
293 		*p++ = xdr_one;                         /* bitmap length */
294 		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
295 		*p++ = htonl(8);              /* attribute buffer length */
296 		p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
297 	}
298 
299 	*p++ = xdr_one;                                  /* next */
300 	*p++ = xdr_zero;                   /* cookie, first word */
301 	*p++ = xdr_two;                   /* cookie, second word */
302 	*p++ = xdr_two;                             /* entry len */
303 	memcpy(p, "..\0\0", 4);                         /* entry */
304 	p++;
305 	*p++ = xdr_one;                         /* bitmap length */
306 	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
307 	*p++ = htonl(8);              /* attribute buffer length */
308 	p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
309 
310 	readdir->pgbase = (char *)p - (char *)start;
311 	readdir->count -= readdir->pgbase;
312 	kunmap_atomic(start);
313 }
314 
315 static long nfs4_update_delay(long *timeout)
316 {
317 	long ret;
318 	if (!timeout)
319 		return NFS4_POLL_RETRY_MAX;
320 	if (*timeout <= 0)
321 		*timeout = NFS4_POLL_RETRY_MIN;
322 	if (*timeout > NFS4_POLL_RETRY_MAX)
323 		*timeout = NFS4_POLL_RETRY_MAX;
324 	ret = *timeout;
325 	*timeout <<= 1;
326 	return ret;
327 }
328 
329 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
330 {
331 	int res = 0;
332 
333 	might_sleep();
334 
335 	freezable_schedule_timeout_killable_unsafe(
336 		nfs4_update_delay(timeout));
337 	if (fatal_signal_pending(current))
338 		res = -ERESTARTSYS;
339 	return res;
340 }
341 
342 /* This is the error handling routine for processes that are allowed
343  * to sleep.
344  */
345 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
346 {
347 	struct nfs_client *clp = server->nfs_client;
348 	struct nfs4_state *state = exception->state;
349 	struct inode *inode = exception->inode;
350 	int ret = errorcode;
351 
352 	exception->retry = 0;
353 	switch(errorcode) {
354 		case 0:
355 			return 0;
356 		case -NFS4ERR_OPENMODE:
357 			if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
358 				nfs4_inode_return_delegation(inode);
359 				exception->retry = 1;
360 				return 0;
361 			}
362 			if (state == NULL)
363 				break;
364 			ret = nfs4_schedule_stateid_recovery(server, state);
365 			if (ret < 0)
366 				break;
367 			goto wait_on_recovery;
368 		case -NFS4ERR_DELEG_REVOKED:
369 		case -NFS4ERR_ADMIN_REVOKED:
370 		case -NFS4ERR_BAD_STATEID:
371 			if (state == NULL)
372 				break;
373 			ret = nfs4_schedule_stateid_recovery(server, state);
374 			if (ret < 0)
375 				break;
376 			goto wait_on_recovery;
377 		case -NFS4ERR_EXPIRED:
378 			if (state != NULL) {
379 				ret = nfs4_schedule_stateid_recovery(server, state);
380 				if (ret < 0)
381 					break;
382 			}
383 		case -NFS4ERR_STALE_STATEID:
384 		case -NFS4ERR_STALE_CLIENTID:
385 			nfs4_schedule_lease_recovery(clp);
386 			goto wait_on_recovery;
387 		case -NFS4ERR_MOVED:
388 			ret = nfs4_schedule_migration_recovery(server);
389 			if (ret < 0)
390 				break;
391 			goto wait_on_recovery;
392 		case -NFS4ERR_LEASE_MOVED:
393 			nfs4_schedule_lease_moved_recovery(clp);
394 			goto wait_on_recovery;
395 #if defined(CONFIG_NFS_V4_1)
396 		case -NFS4ERR_BADSESSION:
397 		case -NFS4ERR_BADSLOT:
398 		case -NFS4ERR_BAD_HIGH_SLOT:
399 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
400 		case -NFS4ERR_DEADSESSION:
401 		case -NFS4ERR_SEQ_FALSE_RETRY:
402 		case -NFS4ERR_SEQ_MISORDERED:
403 			dprintk("%s ERROR: %d Reset session\n", __func__,
404 				errorcode);
405 			nfs4_schedule_session_recovery(clp->cl_session, errorcode);
406 			goto wait_on_recovery;
407 #endif /* defined(CONFIG_NFS_V4_1) */
408 		case -NFS4ERR_FILE_OPEN:
409 			if (exception->timeout > HZ) {
410 				/* We have retried a decent amount, time to
411 				 * fail
412 				 */
413 				ret = -EBUSY;
414 				break;
415 			}
416 		case -NFS4ERR_GRACE:
417 		case -NFS4ERR_DELAY:
418 			ret = nfs4_delay(server->client, &exception->timeout);
419 			if (ret != 0)
420 				break;
421 		case -NFS4ERR_RETRY_UNCACHED_REP:
422 		case -NFS4ERR_OLD_STATEID:
423 			exception->retry = 1;
424 			break;
425 		case -NFS4ERR_BADOWNER:
426 			/* The following works around a Linux server bug! */
427 		case -NFS4ERR_BADNAME:
428 			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
429 				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
430 				exception->retry = 1;
431 				printk(KERN_WARNING "NFS: v4 server %s "
432 						"does not accept raw "
433 						"uid/gids. "
434 						"Reenabling the idmapper.\n",
435 						server->nfs_client->cl_hostname);
436 			}
437 	}
438 	/* We failed to handle the error */
439 	return nfs4_map_errors(ret);
440 wait_on_recovery:
441 	ret = nfs4_wait_clnt_recover(clp);
442 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
443 		return -EIO;
444 	if (ret == 0)
445 		exception->retry = 1;
446 	return ret;
447 }
448 
449 /*
450  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
451  * or 'false' otherwise.
452  */
453 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
454 {
455 	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
456 
457 	if (flavor == RPC_AUTH_GSS_KRB5I ||
458 	    flavor == RPC_AUTH_GSS_KRB5P)
459 		return true;
460 
461 	return false;
462 }
463 
464 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
465 {
466 	spin_lock(&clp->cl_lock);
467 	if (time_before(clp->cl_last_renewal,timestamp))
468 		clp->cl_last_renewal = timestamp;
469 	spin_unlock(&clp->cl_lock);
470 }
471 
472 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
473 {
474 	do_renew_lease(server->nfs_client, timestamp);
475 }
476 
477 struct nfs4_call_sync_data {
478 	const struct nfs_server *seq_server;
479 	struct nfs4_sequence_args *seq_args;
480 	struct nfs4_sequence_res *seq_res;
481 };
482 
483 static void nfs4_init_sequence(struct nfs4_sequence_args *args,
484 			       struct nfs4_sequence_res *res, int cache_reply)
485 {
486 	args->sa_slot = NULL;
487 	args->sa_cache_this = cache_reply;
488 	args->sa_privileged = 0;
489 
490 	res->sr_slot = NULL;
491 }
492 
493 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
494 {
495 	args->sa_privileged = 1;
496 }
497 
498 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
499 			 struct nfs4_sequence_args *args,
500 			 struct nfs4_sequence_res *res,
501 			 struct rpc_task *task)
502 {
503 	struct nfs4_slot *slot;
504 
505 	/* slot already allocated? */
506 	if (res->sr_slot != NULL)
507 		goto out_start;
508 
509 	spin_lock(&tbl->slot_tbl_lock);
510 	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
511 		goto out_sleep;
512 
513 	slot = nfs4_alloc_slot(tbl);
514 	if (IS_ERR(slot)) {
515 		if (slot == ERR_PTR(-ENOMEM))
516 			task->tk_timeout = HZ >> 2;
517 		goto out_sleep;
518 	}
519 	spin_unlock(&tbl->slot_tbl_lock);
520 
521 	args->sa_slot = slot;
522 	res->sr_slot = slot;
523 
524 out_start:
525 	rpc_call_start(task);
526 	return 0;
527 
528 out_sleep:
529 	if (args->sa_privileged)
530 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
531 				NULL, RPC_PRIORITY_PRIVILEGED);
532 	else
533 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
534 	spin_unlock(&tbl->slot_tbl_lock);
535 	return -EAGAIN;
536 }
537 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
538 
539 static int nfs40_sequence_done(struct rpc_task *task,
540 			       struct nfs4_sequence_res *res)
541 {
542 	struct nfs4_slot *slot = res->sr_slot;
543 	struct nfs4_slot_table *tbl;
544 
545 	if (slot == NULL)
546 		goto out;
547 
548 	tbl = slot->table;
549 	spin_lock(&tbl->slot_tbl_lock);
550 	if (!nfs41_wake_and_assign_slot(tbl, slot))
551 		nfs4_free_slot(tbl, slot);
552 	spin_unlock(&tbl->slot_tbl_lock);
553 
554 	res->sr_slot = NULL;
555 out:
556 	return 1;
557 }
558 
559 #if defined(CONFIG_NFS_V4_1)
560 
561 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
562 {
563 	struct nfs4_session *session;
564 	struct nfs4_slot_table *tbl;
565 	struct nfs4_slot *slot = res->sr_slot;
566 	bool send_new_highest_used_slotid = false;
567 
568 	tbl = slot->table;
569 	session = tbl->session;
570 
571 	spin_lock(&tbl->slot_tbl_lock);
572 	/* Be nice to the server: try to ensure that the last transmitted
573 	 * value for highest_user_slotid <= target_highest_slotid
574 	 */
575 	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
576 		send_new_highest_used_slotid = true;
577 
578 	if (nfs41_wake_and_assign_slot(tbl, slot)) {
579 		send_new_highest_used_slotid = false;
580 		goto out_unlock;
581 	}
582 	nfs4_free_slot(tbl, slot);
583 
584 	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
585 		send_new_highest_used_slotid = false;
586 out_unlock:
587 	spin_unlock(&tbl->slot_tbl_lock);
588 	res->sr_slot = NULL;
589 	if (send_new_highest_used_slotid)
590 		nfs41_server_notify_highest_slotid_update(session->clp);
591 }
592 
593 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
594 {
595 	struct nfs4_session *session;
596 	struct nfs4_slot *slot = res->sr_slot;
597 	struct nfs_client *clp;
598 	bool interrupted = false;
599 	int ret = 1;
600 
601 	if (slot == NULL)
602 		goto out_noaction;
603 	/* don't increment the sequence number if the task wasn't sent */
604 	if (!RPC_WAS_SENT(task))
605 		goto out;
606 
607 	session = slot->table->session;
608 
609 	if (slot->interrupted) {
610 		slot->interrupted = 0;
611 		interrupted = true;
612 	}
613 
614 	trace_nfs4_sequence_done(session, res);
615 	/* Check the SEQUENCE operation status */
616 	switch (res->sr_status) {
617 	case 0:
618 		/* Update the slot's sequence and clientid lease timer */
619 		++slot->seq_nr;
620 		clp = session->clp;
621 		do_renew_lease(clp, res->sr_timestamp);
622 		/* Check sequence flags */
623 		if (res->sr_status_flags != 0)
624 			nfs4_schedule_lease_recovery(clp);
625 		nfs41_update_target_slotid(slot->table, slot, res);
626 		break;
627 	case 1:
628 		/*
629 		 * sr_status remains 1 if an RPC level error occurred.
630 		 * The server may or may not have processed the sequence
631 		 * operation..
632 		 * Mark the slot as having hosted an interrupted RPC call.
633 		 */
634 		slot->interrupted = 1;
635 		goto out;
636 	case -NFS4ERR_DELAY:
637 		/* The server detected a resend of the RPC call and
638 		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
639 		 * of RFC5661.
640 		 */
641 		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
642 			__func__,
643 			slot->slot_nr,
644 			slot->seq_nr);
645 		goto out_retry;
646 	case -NFS4ERR_BADSLOT:
647 		/*
648 		 * The slot id we used was probably retired. Try again
649 		 * using a different slot id.
650 		 */
651 		goto retry_nowait;
652 	case -NFS4ERR_SEQ_MISORDERED:
653 		/*
654 		 * Was the last operation on this sequence interrupted?
655 		 * If so, retry after bumping the sequence number.
656 		 */
657 		if (interrupted) {
658 			++slot->seq_nr;
659 			goto retry_nowait;
660 		}
661 		/*
662 		 * Could this slot have been previously retired?
663 		 * If so, then the server may be expecting seq_nr = 1!
664 		 */
665 		if (slot->seq_nr != 1) {
666 			slot->seq_nr = 1;
667 			goto retry_nowait;
668 		}
669 		break;
670 	case -NFS4ERR_SEQ_FALSE_RETRY:
671 		++slot->seq_nr;
672 		goto retry_nowait;
673 	default:
674 		/* Just update the slot sequence no. */
675 		++slot->seq_nr;
676 	}
677 out:
678 	/* The session may be reset by one of the error handlers. */
679 	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
680 	nfs41_sequence_free_slot(res);
681 out_noaction:
682 	return ret;
683 retry_nowait:
684 	if (rpc_restart_call_prepare(task)) {
685 		task->tk_status = 0;
686 		ret = 0;
687 	}
688 	goto out;
689 out_retry:
690 	if (!rpc_restart_call(task))
691 		goto out;
692 	rpc_delay(task, NFS4_POLL_RETRY_MAX);
693 	return 0;
694 }
695 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
696 
697 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
698 {
699 	if (res->sr_slot == NULL)
700 		return 1;
701 	if (!res->sr_slot->table->session)
702 		return nfs40_sequence_done(task, res);
703 	return nfs41_sequence_done(task, res);
704 }
705 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
706 
707 int nfs41_setup_sequence(struct nfs4_session *session,
708 				struct nfs4_sequence_args *args,
709 				struct nfs4_sequence_res *res,
710 				struct rpc_task *task)
711 {
712 	struct nfs4_slot *slot;
713 	struct nfs4_slot_table *tbl;
714 
715 	dprintk("--> %s\n", __func__);
716 	/* slot already allocated? */
717 	if (res->sr_slot != NULL)
718 		goto out_success;
719 
720 	tbl = &session->fc_slot_table;
721 
722 	task->tk_timeout = 0;
723 
724 	spin_lock(&tbl->slot_tbl_lock);
725 	if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
726 	    !args->sa_privileged) {
727 		/* The state manager will wait until the slot table is empty */
728 		dprintk("%s session is draining\n", __func__);
729 		goto out_sleep;
730 	}
731 
732 	slot = nfs4_alloc_slot(tbl);
733 	if (IS_ERR(slot)) {
734 		/* If out of memory, try again in 1/4 second */
735 		if (slot == ERR_PTR(-ENOMEM))
736 			task->tk_timeout = HZ >> 2;
737 		dprintk("<-- %s: no free slots\n", __func__);
738 		goto out_sleep;
739 	}
740 	spin_unlock(&tbl->slot_tbl_lock);
741 
742 	args->sa_slot = slot;
743 
744 	dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
745 			slot->slot_nr, slot->seq_nr);
746 
747 	res->sr_slot = slot;
748 	res->sr_timestamp = jiffies;
749 	res->sr_status_flags = 0;
750 	/*
751 	 * sr_status is only set in decode_sequence, and so will remain
752 	 * set to 1 if an rpc level failure occurs.
753 	 */
754 	res->sr_status = 1;
755 	trace_nfs4_setup_sequence(session, args);
756 out_success:
757 	rpc_call_start(task);
758 	return 0;
759 out_sleep:
760 	/* Privileged tasks are queued with top priority */
761 	if (args->sa_privileged)
762 		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
763 				NULL, RPC_PRIORITY_PRIVILEGED);
764 	else
765 		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
766 	spin_unlock(&tbl->slot_tbl_lock);
767 	return -EAGAIN;
768 }
769 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
770 
771 static int nfs4_setup_sequence(const struct nfs_server *server,
772 			       struct nfs4_sequence_args *args,
773 			       struct nfs4_sequence_res *res,
774 			       struct rpc_task *task)
775 {
776 	struct nfs4_session *session = nfs4_get_session(server);
777 	int ret = 0;
778 
779 	if (!session)
780 		return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
781 					    args, res, task);
782 
783 	dprintk("--> %s clp %p session %p sr_slot %u\n",
784 		__func__, session->clp, session, res->sr_slot ?
785 			res->sr_slot->slot_nr : NFS4_NO_SLOT);
786 
787 	ret = nfs41_setup_sequence(session, args, res, task);
788 
789 	dprintk("<-- %s status=%d\n", __func__, ret);
790 	return ret;
791 }
792 
793 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
794 {
795 	struct nfs4_call_sync_data *data = calldata;
796 	struct nfs4_session *session = nfs4_get_session(data->seq_server);
797 
798 	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
799 
800 	nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
801 }
802 
803 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
804 {
805 	struct nfs4_call_sync_data *data = calldata;
806 
807 	nfs41_sequence_done(task, data->seq_res);
808 }
809 
810 static const struct rpc_call_ops nfs41_call_sync_ops = {
811 	.rpc_call_prepare = nfs41_call_sync_prepare,
812 	.rpc_call_done = nfs41_call_sync_done,
813 };
814 
815 #else	/* !CONFIG_NFS_V4_1 */
816 
817 static int nfs4_setup_sequence(const struct nfs_server *server,
818 			       struct nfs4_sequence_args *args,
819 			       struct nfs4_sequence_res *res,
820 			       struct rpc_task *task)
821 {
822 	return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
823 				    args, res, task);
824 }
825 
826 int nfs4_sequence_done(struct rpc_task *task,
827 		       struct nfs4_sequence_res *res)
828 {
829 	return nfs40_sequence_done(task, res);
830 }
831 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
832 
833 #endif	/* !CONFIG_NFS_V4_1 */
834 
835 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
836 {
837 	struct nfs4_call_sync_data *data = calldata;
838 	nfs4_setup_sequence(data->seq_server,
839 				data->seq_args, data->seq_res, task);
840 }
841 
842 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
843 {
844 	struct nfs4_call_sync_data *data = calldata;
845 	nfs4_sequence_done(task, data->seq_res);
846 }
847 
848 static const struct rpc_call_ops nfs40_call_sync_ops = {
849 	.rpc_call_prepare = nfs40_call_sync_prepare,
850 	.rpc_call_done = nfs40_call_sync_done,
851 };
852 
853 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
854 				   struct nfs_server *server,
855 				   struct rpc_message *msg,
856 				   struct nfs4_sequence_args *args,
857 				   struct nfs4_sequence_res *res)
858 {
859 	int ret;
860 	struct rpc_task *task;
861 	struct nfs_client *clp = server->nfs_client;
862 	struct nfs4_call_sync_data data = {
863 		.seq_server = server,
864 		.seq_args = args,
865 		.seq_res = res,
866 	};
867 	struct rpc_task_setup task_setup = {
868 		.rpc_client = clnt,
869 		.rpc_message = msg,
870 		.callback_ops = clp->cl_mvops->call_sync_ops,
871 		.callback_data = &data
872 	};
873 
874 	task = rpc_run_task(&task_setup);
875 	if (IS_ERR(task))
876 		ret = PTR_ERR(task);
877 	else {
878 		ret = task->tk_status;
879 		rpc_put_task(task);
880 	}
881 	return ret;
882 }
883 
884 int nfs4_call_sync(struct rpc_clnt *clnt,
885 		   struct nfs_server *server,
886 		   struct rpc_message *msg,
887 		   struct nfs4_sequence_args *args,
888 		   struct nfs4_sequence_res *res,
889 		   int cache_reply)
890 {
891 	nfs4_init_sequence(args, res, cache_reply);
892 	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
893 }
894 
895 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
896 {
897 	struct nfs_inode *nfsi = NFS_I(dir);
898 
899 	spin_lock(&dir->i_lock);
900 	nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
901 	if (!cinfo->atomic || cinfo->before != dir->i_version)
902 		nfs_force_lookup_revalidate(dir);
903 	dir->i_version = cinfo->after;
904 	nfs_fscache_invalidate(dir);
905 	spin_unlock(&dir->i_lock);
906 }
907 
908 struct nfs4_opendata {
909 	struct kref kref;
910 	struct nfs_openargs o_arg;
911 	struct nfs_openres o_res;
912 	struct nfs_open_confirmargs c_arg;
913 	struct nfs_open_confirmres c_res;
914 	struct nfs4_string owner_name;
915 	struct nfs4_string group_name;
916 	struct nfs_fattr f_attr;
917 	struct nfs4_label *f_label;
918 	struct dentry *dir;
919 	struct dentry *dentry;
920 	struct nfs4_state_owner *owner;
921 	struct nfs4_state *state;
922 	struct iattr attrs;
923 	unsigned long timestamp;
924 	unsigned int rpc_done : 1;
925 	unsigned int file_created : 1;
926 	unsigned int is_recover : 1;
927 	int rpc_status;
928 	int cancelled;
929 };
930 
931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
932 		int err, struct nfs4_exception *exception)
933 {
934 	if (err != -EINVAL)
935 		return false;
936 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
937 		return false;
938 	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
939 	exception->retry = 1;
940 	return true;
941 }
942 
943 static u32
944 nfs4_map_atomic_open_share(struct nfs_server *server,
945 		fmode_t fmode, int openflags)
946 {
947 	u32 res = 0;
948 
949 	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
950 	case FMODE_READ:
951 		res = NFS4_SHARE_ACCESS_READ;
952 		break;
953 	case FMODE_WRITE:
954 		res = NFS4_SHARE_ACCESS_WRITE;
955 		break;
956 	case FMODE_READ|FMODE_WRITE:
957 		res = NFS4_SHARE_ACCESS_BOTH;
958 	}
959 	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
960 		goto out;
961 	/* Want no delegation if we're using O_DIRECT */
962 	if (openflags & O_DIRECT)
963 		res |= NFS4_SHARE_WANT_NO_DELEG;
964 out:
965 	return res;
966 }
967 
968 static enum open_claim_type4
969 nfs4_map_atomic_open_claim(struct nfs_server *server,
970 		enum open_claim_type4 claim)
971 {
972 	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
973 		return claim;
974 	switch (claim) {
975 	default:
976 		return claim;
977 	case NFS4_OPEN_CLAIM_FH:
978 		return NFS4_OPEN_CLAIM_NULL;
979 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
980 		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
981 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
982 		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
983 	}
984 }
985 
986 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
987 {
988 	p->o_res.f_attr = &p->f_attr;
989 	p->o_res.f_label = p->f_label;
990 	p->o_res.seqid = p->o_arg.seqid;
991 	p->c_res.seqid = p->c_arg.seqid;
992 	p->o_res.server = p->o_arg.server;
993 	p->o_res.access_request = p->o_arg.access;
994 	nfs_fattr_init(&p->f_attr);
995 	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
996 }
997 
998 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
999 		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1000 		const struct iattr *attrs,
1001 		struct nfs4_label *label,
1002 		enum open_claim_type4 claim,
1003 		gfp_t gfp_mask)
1004 {
1005 	struct dentry *parent = dget_parent(dentry);
1006 	struct inode *dir = parent->d_inode;
1007 	struct nfs_server *server = NFS_SERVER(dir);
1008 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1009 	struct nfs4_opendata *p;
1010 
1011 	p = kzalloc(sizeof(*p), gfp_mask);
1012 	if (p == NULL)
1013 		goto err;
1014 
1015 	p->f_label = nfs4_label_alloc(server, gfp_mask);
1016 	if (IS_ERR(p->f_label))
1017 		goto err_free_p;
1018 
1019 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1020 	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1021 	if (IS_ERR(p->o_arg.seqid))
1022 		goto err_free_label;
1023 	nfs_sb_active(dentry->d_sb);
1024 	p->dentry = dget(dentry);
1025 	p->dir = parent;
1026 	p->owner = sp;
1027 	atomic_inc(&sp->so_count);
1028 	p->o_arg.open_flags = flags;
1029 	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1030 	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1031 			fmode, flags);
1032 	/* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1033 	 * will return permission denied for all bits until close */
1034 	if (!(flags & O_EXCL)) {
1035 		/* ask server to check for all possible rights as results
1036 		 * are cached */
1037 		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1038 				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1039 	}
1040 	p->o_arg.clientid = server->nfs_client->cl_clientid;
1041 	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1042 	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1043 	p->o_arg.name = &dentry->d_name;
1044 	p->o_arg.server = server;
1045 	p->o_arg.bitmask = nfs4_bitmask(server, label);
1046 	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1047 	p->o_arg.label = label;
1048 	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1049 	switch (p->o_arg.claim) {
1050 	case NFS4_OPEN_CLAIM_NULL:
1051 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1052 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1053 		p->o_arg.fh = NFS_FH(dir);
1054 		break;
1055 	case NFS4_OPEN_CLAIM_PREVIOUS:
1056 	case NFS4_OPEN_CLAIM_FH:
1057 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1058 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1059 		p->o_arg.fh = NFS_FH(dentry->d_inode);
1060 	}
1061 	if (attrs != NULL && attrs->ia_valid != 0) {
1062 		__u32 verf[2];
1063 
1064 		p->o_arg.u.attrs = &p->attrs;
1065 		memcpy(&p->attrs, attrs, sizeof(p->attrs));
1066 
1067 		verf[0] = jiffies;
1068 		verf[1] = current->pid;
1069 		memcpy(p->o_arg.u.verifier.data, verf,
1070 				sizeof(p->o_arg.u.verifier.data));
1071 	}
1072 	p->c_arg.fh = &p->o_res.fh;
1073 	p->c_arg.stateid = &p->o_res.stateid;
1074 	p->c_arg.seqid = p->o_arg.seqid;
1075 	nfs4_init_opendata_res(p);
1076 	kref_init(&p->kref);
1077 	return p;
1078 
1079 err_free_label:
1080 	nfs4_label_free(p->f_label);
1081 err_free_p:
1082 	kfree(p);
1083 err:
1084 	dput(parent);
1085 	return NULL;
1086 }
1087 
1088 static void nfs4_opendata_free(struct kref *kref)
1089 {
1090 	struct nfs4_opendata *p = container_of(kref,
1091 			struct nfs4_opendata, kref);
1092 	struct super_block *sb = p->dentry->d_sb;
1093 
1094 	nfs_free_seqid(p->o_arg.seqid);
1095 	if (p->state != NULL)
1096 		nfs4_put_open_state(p->state);
1097 	nfs4_put_state_owner(p->owner);
1098 
1099 	nfs4_label_free(p->f_label);
1100 
1101 	dput(p->dir);
1102 	dput(p->dentry);
1103 	nfs_sb_deactive(sb);
1104 	nfs_fattr_free_names(&p->f_attr);
1105 	kfree(p->f_attr.mdsthreshold);
1106 	kfree(p);
1107 }
1108 
1109 static void nfs4_opendata_put(struct nfs4_opendata *p)
1110 {
1111 	if (p != NULL)
1112 		kref_put(&p->kref, nfs4_opendata_free);
1113 }
1114 
1115 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1116 {
1117 	int ret;
1118 
1119 	ret = rpc_wait_for_completion_task(task);
1120 	return ret;
1121 }
1122 
1123 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1124 {
1125 	int ret = 0;
1126 
1127 	if (open_mode & (O_EXCL|O_TRUNC))
1128 		goto out;
1129 	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1130 		case FMODE_READ:
1131 			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1132 				&& state->n_rdonly != 0;
1133 			break;
1134 		case FMODE_WRITE:
1135 			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1136 				&& state->n_wronly != 0;
1137 			break;
1138 		case FMODE_READ|FMODE_WRITE:
1139 			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1140 				&& state->n_rdwr != 0;
1141 	}
1142 out:
1143 	return ret;
1144 }
1145 
1146 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1147 {
1148 	if (delegation == NULL)
1149 		return 0;
1150 	if ((delegation->type & fmode) != fmode)
1151 		return 0;
1152 	if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1153 		return 0;
1154 	nfs_mark_delegation_referenced(delegation);
1155 	return 1;
1156 }
1157 
1158 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1159 {
1160 	switch (fmode) {
1161 		case FMODE_WRITE:
1162 			state->n_wronly++;
1163 			break;
1164 		case FMODE_READ:
1165 			state->n_rdonly++;
1166 			break;
1167 		case FMODE_READ|FMODE_WRITE:
1168 			state->n_rdwr++;
1169 	}
1170 	nfs4_state_set_mode_locked(state, state->state | fmode);
1171 }
1172 
1173 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1174 {
1175 	struct nfs_client *clp = state->owner->so_server->nfs_client;
1176 	bool need_recover = false;
1177 
1178 	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1179 		need_recover = true;
1180 	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1181 		need_recover = true;
1182 	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1183 		need_recover = true;
1184 	if (need_recover)
1185 		nfs4_state_mark_reclaim_nograce(clp, state);
1186 }
1187 
1188 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1189 		nfs4_stateid *stateid)
1190 {
1191 	if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1192 		return true;
1193 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1194 		nfs_test_and_clear_all_open_stateid(state);
1195 		return true;
1196 	}
1197 	if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1198 		return true;
1199 	return false;
1200 }
1201 
1202 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1203 {
1204 	if (state->n_wronly)
1205 		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1206 	if (state->n_rdonly)
1207 		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1208 	if (state->n_rdwr)
1209 		set_bit(NFS_O_RDWR_STATE, &state->flags);
1210 }
1211 
1212 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1213 		nfs4_stateid *stateid, fmode_t fmode)
1214 {
1215 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1216 	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1217 	case FMODE_WRITE:
1218 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1219 		break;
1220 	case FMODE_READ:
1221 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1222 		break;
1223 	case 0:
1224 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1225 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1226 		clear_bit(NFS_OPEN_STATE, &state->flags);
1227 	}
1228 	if (stateid == NULL)
1229 		return;
1230 	/* Handle races with OPEN */
1231 	if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1232 	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1233 		nfs_resync_open_stateid_locked(state);
1234 		return;
1235 	}
1236 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1237 		nfs4_stateid_copy(&state->stateid, stateid);
1238 	nfs4_stateid_copy(&state->open_stateid, stateid);
1239 }
1240 
1241 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1242 {
1243 	write_seqlock(&state->seqlock);
1244 	nfs_clear_open_stateid_locked(state, stateid, fmode);
1245 	write_sequnlock(&state->seqlock);
1246 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1247 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1248 }
1249 
1250 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1251 {
1252 	switch (fmode) {
1253 		case FMODE_READ:
1254 			set_bit(NFS_O_RDONLY_STATE, &state->flags);
1255 			break;
1256 		case FMODE_WRITE:
1257 			set_bit(NFS_O_WRONLY_STATE, &state->flags);
1258 			break;
1259 		case FMODE_READ|FMODE_WRITE:
1260 			set_bit(NFS_O_RDWR_STATE, &state->flags);
1261 	}
1262 	if (!nfs_need_update_open_stateid(state, stateid))
1263 		return;
1264 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1265 		nfs4_stateid_copy(&state->stateid, stateid);
1266 	nfs4_stateid_copy(&state->open_stateid, stateid);
1267 }
1268 
1269 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1270 {
1271 	/*
1272 	 * Protect the call to nfs4_state_set_mode_locked and
1273 	 * serialise the stateid update
1274 	 */
1275 	write_seqlock(&state->seqlock);
1276 	if (deleg_stateid != NULL) {
1277 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
1278 		set_bit(NFS_DELEGATED_STATE, &state->flags);
1279 	}
1280 	if (open_stateid != NULL)
1281 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
1282 	write_sequnlock(&state->seqlock);
1283 	spin_lock(&state->owner->so_lock);
1284 	update_open_stateflags(state, fmode);
1285 	spin_unlock(&state->owner->so_lock);
1286 }
1287 
1288 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1289 {
1290 	struct nfs_inode *nfsi = NFS_I(state->inode);
1291 	struct nfs_delegation *deleg_cur;
1292 	int ret = 0;
1293 
1294 	fmode &= (FMODE_READ|FMODE_WRITE);
1295 
1296 	rcu_read_lock();
1297 	deleg_cur = rcu_dereference(nfsi->delegation);
1298 	if (deleg_cur == NULL)
1299 		goto no_delegation;
1300 
1301 	spin_lock(&deleg_cur->lock);
1302 	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1303 	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1304 	    (deleg_cur->type & fmode) != fmode)
1305 		goto no_delegation_unlock;
1306 
1307 	if (delegation == NULL)
1308 		delegation = &deleg_cur->stateid;
1309 	else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1310 		goto no_delegation_unlock;
1311 
1312 	nfs_mark_delegation_referenced(deleg_cur);
1313 	__update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1314 	ret = 1;
1315 no_delegation_unlock:
1316 	spin_unlock(&deleg_cur->lock);
1317 no_delegation:
1318 	rcu_read_unlock();
1319 
1320 	if (!ret && open_stateid != NULL) {
1321 		__update_open_stateid(state, open_stateid, NULL, fmode);
1322 		ret = 1;
1323 	}
1324 	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1325 		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1326 
1327 	return ret;
1328 }
1329 
1330 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1331 		const nfs4_stateid *stateid)
1332 {
1333 	struct nfs4_state *state = lsp->ls_state;
1334 	bool ret = false;
1335 
1336 	spin_lock(&state->state_lock);
1337 	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1338 		goto out_noupdate;
1339 	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1340 		goto out_noupdate;
1341 	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1342 	ret = true;
1343 out_noupdate:
1344 	spin_unlock(&state->state_lock);
1345 	return ret;
1346 }
1347 
1348 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1349 {
1350 	struct nfs_delegation *delegation;
1351 
1352 	rcu_read_lock();
1353 	delegation = rcu_dereference(NFS_I(inode)->delegation);
1354 	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1355 		rcu_read_unlock();
1356 		return;
1357 	}
1358 	rcu_read_unlock();
1359 	nfs4_inode_return_delegation(inode);
1360 }
1361 
1362 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1363 {
1364 	struct nfs4_state *state = opendata->state;
1365 	struct nfs_inode *nfsi = NFS_I(state->inode);
1366 	struct nfs_delegation *delegation;
1367 	int open_mode = opendata->o_arg.open_flags;
1368 	fmode_t fmode = opendata->o_arg.fmode;
1369 	nfs4_stateid stateid;
1370 	int ret = -EAGAIN;
1371 
1372 	for (;;) {
1373 		spin_lock(&state->owner->so_lock);
1374 		if (can_open_cached(state, fmode, open_mode)) {
1375 			update_open_stateflags(state, fmode);
1376 			spin_unlock(&state->owner->so_lock);
1377 			goto out_return_state;
1378 		}
1379 		spin_unlock(&state->owner->so_lock);
1380 		rcu_read_lock();
1381 		delegation = rcu_dereference(nfsi->delegation);
1382 		if (!can_open_delegated(delegation, fmode)) {
1383 			rcu_read_unlock();
1384 			break;
1385 		}
1386 		/* Save the delegation */
1387 		nfs4_stateid_copy(&stateid, &delegation->stateid);
1388 		rcu_read_unlock();
1389 		nfs_release_seqid(opendata->o_arg.seqid);
1390 		if (!opendata->is_recover) {
1391 			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1392 			if (ret != 0)
1393 				goto out;
1394 		}
1395 		ret = -EAGAIN;
1396 
1397 		/* Try to update the stateid using the delegation */
1398 		if (update_open_stateid(state, NULL, &stateid, fmode))
1399 			goto out_return_state;
1400 	}
1401 out:
1402 	return ERR_PTR(ret);
1403 out_return_state:
1404 	atomic_inc(&state->count);
1405 	return state;
1406 }
1407 
1408 static void
1409 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1410 {
1411 	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1412 	struct nfs_delegation *delegation;
1413 	int delegation_flags = 0;
1414 
1415 	rcu_read_lock();
1416 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1417 	if (delegation)
1418 		delegation_flags = delegation->flags;
1419 	rcu_read_unlock();
1420 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1421 		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1422 				   "returning a delegation for "
1423 				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1424 				   clp->cl_hostname);
1425 	} else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1426 		nfs_inode_set_delegation(state->inode,
1427 					 data->owner->so_cred,
1428 					 &data->o_res);
1429 	else
1430 		nfs_inode_reclaim_delegation(state->inode,
1431 					     data->owner->so_cred,
1432 					     &data->o_res);
1433 }
1434 
1435 /*
1436  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1437  * and update the nfs4_state.
1438  */
1439 static struct nfs4_state *
1440 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1441 {
1442 	struct inode *inode = data->state->inode;
1443 	struct nfs4_state *state = data->state;
1444 	int ret;
1445 
1446 	if (!data->rpc_done) {
1447 		if (data->rpc_status) {
1448 			ret = data->rpc_status;
1449 			goto err;
1450 		}
1451 		/* cached opens have already been processed */
1452 		goto update;
1453 	}
1454 
1455 	ret = nfs_refresh_inode(inode, &data->f_attr);
1456 	if (ret)
1457 		goto err;
1458 
1459 	if (data->o_res.delegation_type != 0)
1460 		nfs4_opendata_check_deleg(data, state);
1461 update:
1462 	update_open_stateid(state, &data->o_res.stateid, NULL,
1463 			    data->o_arg.fmode);
1464 	atomic_inc(&state->count);
1465 
1466 	return state;
1467 err:
1468 	return ERR_PTR(ret);
1469 
1470 }
1471 
1472 static struct nfs4_state *
1473 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1474 {
1475 	struct inode *inode;
1476 	struct nfs4_state *state = NULL;
1477 	int ret;
1478 
1479 	if (!data->rpc_done) {
1480 		state = nfs4_try_open_cached(data);
1481 		goto out;
1482 	}
1483 
1484 	ret = -EAGAIN;
1485 	if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1486 		goto err;
1487 	inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1488 	ret = PTR_ERR(inode);
1489 	if (IS_ERR(inode))
1490 		goto err;
1491 	ret = -ENOMEM;
1492 	state = nfs4_get_open_state(inode, data->owner);
1493 	if (state == NULL)
1494 		goto err_put_inode;
1495 	if (data->o_res.delegation_type != 0)
1496 		nfs4_opendata_check_deleg(data, state);
1497 	update_open_stateid(state, &data->o_res.stateid, NULL,
1498 			data->o_arg.fmode);
1499 	iput(inode);
1500 out:
1501 	nfs_release_seqid(data->o_arg.seqid);
1502 	return state;
1503 err_put_inode:
1504 	iput(inode);
1505 err:
1506 	return ERR_PTR(ret);
1507 }
1508 
1509 static struct nfs4_state *
1510 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1511 {
1512 	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1513 		return _nfs4_opendata_reclaim_to_nfs4_state(data);
1514 	return _nfs4_opendata_to_nfs4_state(data);
1515 }
1516 
1517 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1518 {
1519 	struct nfs_inode *nfsi = NFS_I(state->inode);
1520 	struct nfs_open_context *ctx;
1521 
1522 	spin_lock(&state->inode->i_lock);
1523 	list_for_each_entry(ctx, &nfsi->open_files, list) {
1524 		if (ctx->state != state)
1525 			continue;
1526 		get_nfs_open_context(ctx);
1527 		spin_unlock(&state->inode->i_lock);
1528 		return ctx;
1529 	}
1530 	spin_unlock(&state->inode->i_lock);
1531 	return ERR_PTR(-ENOENT);
1532 }
1533 
1534 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1535 		struct nfs4_state *state, enum open_claim_type4 claim)
1536 {
1537 	struct nfs4_opendata *opendata;
1538 
1539 	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1540 			NULL, NULL, claim, GFP_NOFS);
1541 	if (opendata == NULL)
1542 		return ERR_PTR(-ENOMEM);
1543 	opendata->state = state;
1544 	atomic_inc(&state->count);
1545 	return opendata;
1546 }
1547 
1548 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1549 {
1550 	struct nfs4_state *newstate;
1551 	int ret;
1552 
1553 	opendata->o_arg.open_flags = 0;
1554 	opendata->o_arg.fmode = fmode;
1555 	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1556 	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1557 	nfs4_init_opendata_res(opendata);
1558 	ret = _nfs4_recover_proc_open(opendata);
1559 	if (ret != 0)
1560 		return ret;
1561 	newstate = nfs4_opendata_to_nfs4_state(opendata);
1562 	if (IS_ERR(newstate))
1563 		return PTR_ERR(newstate);
1564 	nfs4_close_state(newstate, fmode);
1565 	*res = newstate;
1566 	return 0;
1567 }
1568 
1569 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1570 {
1571 	struct nfs4_state *newstate;
1572 	int ret;
1573 
1574 	/* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1575 	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1576 	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1577 	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1578 	/* memory barrier prior to reading state->n_* */
1579 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1580 	clear_bit(NFS_OPEN_STATE, &state->flags);
1581 	smp_rmb();
1582 	if (state->n_rdwr != 0) {
1583 		ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1584 		if (ret != 0)
1585 			return ret;
1586 		if (newstate != state)
1587 			return -ESTALE;
1588 	}
1589 	if (state->n_wronly != 0) {
1590 		ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1591 		if (ret != 0)
1592 			return ret;
1593 		if (newstate != state)
1594 			return -ESTALE;
1595 	}
1596 	if (state->n_rdonly != 0) {
1597 		ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1598 		if (ret != 0)
1599 			return ret;
1600 		if (newstate != state)
1601 			return -ESTALE;
1602 	}
1603 	/*
1604 	 * We may have performed cached opens for all three recoveries.
1605 	 * Check if we need to update the current stateid.
1606 	 */
1607 	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1608 	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1609 		write_seqlock(&state->seqlock);
1610 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1611 			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1612 		write_sequnlock(&state->seqlock);
1613 	}
1614 	return 0;
1615 }
1616 
1617 /*
1618  * OPEN_RECLAIM:
1619  * 	reclaim state on the server after a reboot.
1620  */
1621 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1622 {
1623 	struct nfs_delegation *delegation;
1624 	struct nfs4_opendata *opendata;
1625 	fmode_t delegation_type = 0;
1626 	int status;
1627 
1628 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1629 			NFS4_OPEN_CLAIM_PREVIOUS);
1630 	if (IS_ERR(opendata))
1631 		return PTR_ERR(opendata);
1632 	rcu_read_lock();
1633 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1634 	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1635 		delegation_type = delegation->type;
1636 	rcu_read_unlock();
1637 	opendata->o_arg.u.delegation_type = delegation_type;
1638 	status = nfs4_open_recover(opendata, state);
1639 	nfs4_opendata_put(opendata);
1640 	return status;
1641 }
1642 
1643 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1644 {
1645 	struct nfs_server *server = NFS_SERVER(state->inode);
1646 	struct nfs4_exception exception = { };
1647 	int err;
1648 	do {
1649 		err = _nfs4_do_open_reclaim(ctx, state);
1650 		trace_nfs4_open_reclaim(ctx, 0, err);
1651 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1652 			continue;
1653 		if (err != -NFS4ERR_DELAY)
1654 			break;
1655 		nfs4_handle_exception(server, err, &exception);
1656 	} while (exception.retry);
1657 	return err;
1658 }
1659 
1660 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1661 {
1662 	struct nfs_open_context *ctx;
1663 	int ret;
1664 
1665 	ctx = nfs4_state_find_open_context(state);
1666 	if (IS_ERR(ctx))
1667 		return -EAGAIN;
1668 	ret = nfs4_do_open_reclaim(ctx, state);
1669 	put_nfs_open_context(ctx);
1670 	return ret;
1671 }
1672 
1673 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1674 {
1675 	switch (err) {
1676 		default:
1677 			printk(KERN_ERR "NFS: %s: unhandled error "
1678 					"%d.\n", __func__, err);
1679 		case 0:
1680 		case -ENOENT:
1681 		case -ESTALE:
1682 			break;
1683 		case -NFS4ERR_BADSESSION:
1684 		case -NFS4ERR_BADSLOT:
1685 		case -NFS4ERR_BAD_HIGH_SLOT:
1686 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1687 		case -NFS4ERR_DEADSESSION:
1688 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1689 			nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1690 			return -EAGAIN;
1691 		case -NFS4ERR_STALE_CLIENTID:
1692 		case -NFS4ERR_STALE_STATEID:
1693 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1694 		case -NFS4ERR_EXPIRED:
1695 			/* Don't recall a delegation if it was lost */
1696 			nfs4_schedule_lease_recovery(server->nfs_client);
1697 			return -EAGAIN;
1698 		case -NFS4ERR_MOVED:
1699 			nfs4_schedule_migration_recovery(server);
1700 			return -EAGAIN;
1701 		case -NFS4ERR_LEASE_MOVED:
1702 			nfs4_schedule_lease_moved_recovery(server->nfs_client);
1703 			return -EAGAIN;
1704 		case -NFS4ERR_DELEG_REVOKED:
1705 		case -NFS4ERR_ADMIN_REVOKED:
1706 		case -NFS4ERR_BAD_STATEID:
1707 		case -NFS4ERR_OPENMODE:
1708 			nfs_inode_find_state_and_recover(state->inode,
1709 					stateid);
1710 			nfs4_schedule_stateid_recovery(server, state);
1711 			return -EAGAIN;
1712 		case -NFS4ERR_DELAY:
1713 		case -NFS4ERR_GRACE:
1714 			set_bit(NFS_DELEGATED_STATE, &state->flags);
1715 			ssleep(1);
1716 			return -EAGAIN;
1717 		case -ENOMEM:
1718 		case -NFS4ERR_DENIED:
1719 			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
1720 			return 0;
1721 	}
1722 	return err;
1723 }
1724 
1725 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1726 {
1727 	struct nfs_server *server = NFS_SERVER(state->inode);
1728 	struct nfs4_opendata *opendata;
1729 	int err;
1730 
1731 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
1732 			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1733 	if (IS_ERR(opendata))
1734 		return PTR_ERR(opendata);
1735 	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1736 	err = nfs4_open_recover(opendata, state);
1737 	nfs4_opendata_put(opendata);
1738 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1739 }
1740 
1741 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1742 {
1743 	struct nfs4_opendata *data = calldata;
1744 
1745 	nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1746 			     &data->c_arg.seq_args, &data->c_res.seq_res, task);
1747 }
1748 
1749 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1750 {
1751 	struct nfs4_opendata *data = calldata;
1752 
1753 	nfs40_sequence_done(task, &data->c_res.seq_res);
1754 
1755 	data->rpc_status = task->tk_status;
1756 	if (data->rpc_status == 0) {
1757 		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1758 		nfs_confirm_seqid(&data->owner->so_seqid, 0);
1759 		renew_lease(data->o_res.server, data->timestamp);
1760 		data->rpc_done = 1;
1761 	}
1762 }
1763 
1764 static void nfs4_open_confirm_release(void *calldata)
1765 {
1766 	struct nfs4_opendata *data = calldata;
1767 	struct nfs4_state *state = NULL;
1768 
1769 	/* If this request hasn't been cancelled, do nothing */
1770 	if (data->cancelled == 0)
1771 		goto out_free;
1772 	/* In case of error, no cleanup! */
1773 	if (!data->rpc_done)
1774 		goto out_free;
1775 	state = nfs4_opendata_to_nfs4_state(data);
1776 	if (!IS_ERR(state))
1777 		nfs4_close_state(state, data->o_arg.fmode);
1778 out_free:
1779 	nfs4_opendata_put(data);
1780 }
1781 
1782 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1783 	.rpc_call_prepare = nfs4_open_confirm_prepare,
1784 	.rpc_call_done = nfs4_open_confirm_done,
1785 	.rpc_release = nfs4_open_confirm_release,
1786 };
1787 
1788 /*
1789  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1790  */
1791 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1792 {
1793 	struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1794 	struct rpc_task *task;
1795 	struct  rpc_message msg = {
1796 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1797 		.rpc_argp = &data->c_arg,
1798 		.rpc_resp = &data->c_res,
1799 		.rpc_cred = data->owner->so_cred,
1800 	};
1801 	struct rpc_task_setup task_setup_data = {
1802 		.rpc_client = server->client,
1803 		.rpc_message = &msg,
1804 		.callback_ops = &nfs4_open_confirm_ops,
1805 		.callback_data = data,
1806 		.workqueue = nfsiod_workqueue,
1807 		.flags = RPC_TASK_ASYNC,
1808 	};
1809 	int status;
1810 
1811 	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1812 	kref_get(&data->kref);
1813 	data->rpc_done = 0;
1814 	data->rpc_status = 0;
1815 	data->timestamp = jiffies;
1816 	task = rpc_run_task(&task_setup_data);
1817 	if (IS_ERR(task))
1818 		return PTR_ERR(task);
1819 	status = nfs4_wait_for_completion_rpc_task(task);
1820 	if (status != 0) {
1821 		data->cancelled = 1;
1822 		smp_wmb();
1823 	} else
1824 		status = data->rpc_status;
1825 	rpc_put_task(task);
1826 	return status;
1827 }
1828 
1829 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1830 {
1831 	struct nfs4_opendata *data = calldata;
1832 	struct nfs4_state_owner *sp = data->owner;
1833 	struct nfs_client *clp = sp->so_server->nfs_client;
1834 
1835 	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1836 		goto out_wait;
1837 	/*
1838 	 * Check if we still need to send an OPEN call, or if we can use
1839 	 * a delegation instead.
1840 	 */
1841 	if (data->state != NULL) {
1842 		struct nfs_delegation *delegation;
1843 
1844 		if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1845 			goto out_no_action;
1846 		rcu_read_lock();
1847 		delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1848 		if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1849 		    data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1850 		    can_open_delegated(delegation, data->o_arg.fmode))
1851 			goto unlock_no_action;
1852 		rcu_read_unlock();
1853 	}
1854 	/* Update client id. */
1855 	data->o_arg.clientid = clp->cl_clientid;
1856 	switch (data->o_arg.claim) {
1857 	case NFS4_OPEN_CLAIM_PREVIOUS:
1858 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1859 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1860 		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1861 	case NFS4_OPEN_CLAIM_FH:
1862 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1863 		nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1864 	}
1865 	data->timestamp = jiffies;
1866 	if (nfs4_setup_sequence(data->o_arg.server,
1867 				&data->o_arg.seq_args,
1868 				&data->o_res.seq_res,
1869 				task) != 0)
1870 		nfs_release_seqid(data->o_arg.seqid);
1871 
1872 	/* Set the create mode (note dependency on the session type) */
1873 	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1874 	if (data->o_arg.open_flags & O_EXCL) {
1875 		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1876 		if (nfs4_has_persistent_session(clp))
1877 			data->o_arg.createmode = NFS4_CREATE_GUARDED;
1878 		else if (clp->cl_mvops->minor_version > 0)
1879 			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1880 	}
1881 	return;
1882 unlock_no_action:
1883 	rcu_read_unlock();
1884 out_no_action:
1885 	task->tk_action = NULL;
1886 out_wait:
1887 	nfs4_sequence_done(task, &data->o_res.seq_res);
1888 }
1889 
1890 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1891 {
1892 	struct nfs4_opendata *data = calldata;
1893 
1894 	data->rpc_status = task->tk_status;
1895 
1896 	if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1897 		return;
1898 
1899 	if (task->tk_status == 0) {
1900 		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1901 			switch (data->o_res.f_attr->mode & S_IFMT) {
1902 			case S_IFREG:
1903 				break;
1904 			case S_IFLNK:
1905 				data->rpc_status = -ELOOP;
1906 				break;
1907 			case S_IFDIR:
1908 				data->rpc_status = -EISDIR;
1909 				break;
1910 			default:
1911 				data->rpc_status = -ENOTDIR;
1912 			}
1913 		}
1914 		renew_lease(data->o_res.server, data->timestamp);
1915 		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1916 			nfs_confirm_seqid(&data->owner->so_seqid, 0);
1917 	}
1918 	data->rpc_done = 1;
1919 }
1920 
1921 static void nfs4_open_release(void *calldata)
1922 {
1923 	struct nfs4_opendata *data = calldata;
1924 	struct nfs4_state *state = NULL;
1925 
1926 	/* If this request hasn't been cancelled, do nothing */
1927 	if (data->cancelled == 0)
1928 		goto out_free;
1929 	/* In case of error, no cleanup! */
1930 	if (data->rpc_status != 0 || !data->rpc_done)
1931 		goto out_free;
1932 	/* In case we need an open_confirm, no cleanup! */
1933 	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1934 		goto out_free;
1935 	state = nfs4_opendata_to_nfs4_state(data);
1936 	if (!IS_ERR(state))
1937 		nfs4_close_state(state, data->o_arg.fmode);
1938 out_free:
1939 	nfs4_opendata_put(data);
1940 }
1941 
1942 static const struct rpc_call_ops nfs4_open_ops = {
1943 	.rpc_call_prepare = nfs4_open_prepare,
1944 	.rpc_call_done = nfs4_open_done,
1945 	.rpc_release = nfs4_open_release,
1946 };
1947 
1948 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1949 {
1950 	struct inode *dir = data->dir->d_inode;
1951 	struct nfs_server *server = NFS_SERVER(dir);
1952 	struct nfs_openargs *o_arg = &data->o_arg;
1953 	struct nfs_openres *o_res = &data->o_res;
1954 	struct rpc_task *task;
1955 	struct rpc_message msg = {
1956 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1957 		.rpc_argp = o_arg,
1958 		.rpc_resp = o_res,
1959 		.rpc_cred = data->owner->so_cred,
1960 	};
1961 	struct rpc_task_setup task_setup_data = {
1962 		.rpc_client = server->client,
1963 		.rpc_message = &msg,
1964 		.callback_ops = &nfs4_open_ops,
1965 		.callback_data = data,
1966 		.workqueue = nfsiod_workqueue,
1967 		.flags = RPC_TASK_ASYNC,
1968 	};
1969 	int status;
1970 
1971 	nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1972 	kref_get(&data->kref);
1973 	data->rpc_done = 0;
1974 	data->rpc_status = 0;
1975 	data->cancelled = 0;
1976 	data->is_recover = 0;
1977 	if (isrecover) {
1978 		nfs4_set_sequence_privileged(&o_arg->seq_args);
1979 		data->is_recover = 1;
1980 	}
1981 	task = rpc_run_task(&task_setup_data);
1982         if (IS_ERR(task))
1983                 return PTR_ERR(task);
1984         status = nfs4_wait_for_completion_rpc_task(task);
1985         if (status != 0) {
1986                 data->cancelled = 1;
1987                 smp_wmb();
1988         } else
1989                 status = data->rpc_status;
1990         rpc_put_task(task);
1991 
1992 	return status;
1993 }
1994 
1995 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1996 {
1997 	struct inode *dir = data->dir->d_inode;
1998 	struct nfs_openres *o_res = &data->o_res;
1999         int status;
2000 
2001 	status = nfs4_run_open_task(data, 1);
2002 	if (status != 0 || !data->rpc_done)
2003 		return status;
2004 
2005 	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2006 
2007 	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2008 		status = _nfs4_proc_open_confirm(data);
2009 		if (status != 0)
2010 			return status;
2011 	}
2012 
2013 	return status;
2014 }
2015 
2016 /*
2017  * Additional permission checks in order to distinguish between an
2018  * open for read, and an open for execute. This works around the
2019  * fact that NFSv4 OPEN treats read and execute permissions as being
2020  * the same.
2021  * Note that in the non-execute case, we want to turn off permission
2022  * checking if we just created a new file (POSIX open() semantics).
2023  */
2024 static int nfs4_opendata_access(struct rpc_cred *cred,
2025 				struct nfs4_opendata *opendata,
2026 				struct nfs4_state *state, fmode_t fmode,
2027 				int openflags)
2028 {
2029 	struct nfs_access_entry cache;
2030 	u32 mask;
2031 
2032 	/* access call failed or for some reason the server doesn't
2033 	 * support any access modes -- defer access call until later */
2034 	if (opendata->o_res.access_supported == 0)
2035 		return 0;
2036 
2037 	mask = 0;
2038 	/*
2039 	 * Use openflags to check for exec, because fmode won't
2040 	 * always have FMODE_EXEC set when file open for exec.
2041 	 */
2042 	if (openflags & __FMODE_EXEC) {
2043 		/* ONLY check for exec rights */
2044 		mask = MAY_EXEC;
2045 	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2046 		mask = MAY_READ;
2047 
2048 	cache.cred = cred;
2049 	cache.jiffies = jiffies;
2050 	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2051 	nfs_access_add_cache(state->inode, &cache);
2052 
2053 	if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2054 		return 0;
2055 
2056 	/* even though OPEN succeeded, access is denied. Close the file */
2057 	nfs4_close_state(state, fmode);
2058 	return -EACCES;
2059 }
2060 
2061 /*
2062  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2063  */
2064 static int _nfs4_proc_open(struct nfs4_opendata *data)
2065 {
2066 	struct inode *dir = data->dir->d_inode;
2067 	struct nfs_server *server = NFS_SERVER(dir);
2068 	struct nfs_openargs *o_arg = &data->o_arg;
2069 	struct nfs_openres *o_res = &data->o_res;
2070 	int status;
2071 
2072 	status = nfs4_run_open_task(data, 0);
2073 	if (!data->rpc_done)
2074 		return status;
2075 	if (status != 0) {
2076 		if (status == -NFS4ERR_BADNAME &&
2077 				!(o_arg->open_flags & O_CREAT))
2078 			return -ENOENT;
2079 		return status;
2080 	}
2081 
2082 	nfs_fattr_map_and_free_names(server, &data->f_attr);
2083 
2084 	if (o_arg->open_flags & O_CREAT) {
2085 		update_changeattr(dir, &o_res->cinfo);
2086 		if (o_arg->open_flags & O_EXCL)
2087 			data->file_created = 1;
2088 		else if (o_res->cinfo.before != o_res->cinfo.after)
2089 			data->file_created = 1;
2090 	}
2091 	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2092 		server->caps &= ~NFS_CAP_POSIX_LOCK;
2093 	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2094 		status = _nfs4_proc_open_confirm(data);
2095 		if (status != 0)
2096 			return status;
2097 	}
2098 	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2099 		nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2100 	return 0;
2101 }
2102 
2103 static int nfs4_recover_expired_lease(struct nfs_server *server)
2104 {
2105 	return nfs4_client_recover_expired_lease(server->nfs_client);
2106 }
2107 
2108 /*
2109  * OPEN_EXPIRED:
2110  * 	reclaim state on the server after a network partition.
2111  * 	Assumes caller holds the appropriate lock
2112  */
2113 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2114 {
2115 	struct nfs4_opendata *opendata;
2116 	int ret;
2117 
2118 	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2119 			NFS4_OPEN_CLAIM_FH);
2120 	if (IS_ERR(opendata))
2121 		return PTR_ERR(opendata);
2122 	ret = nfs4_open_recover(opendata, state);
2123 	if (ret == -ESTALE)
2124 		d_drop(ctx->dentry);
2125 	nfs4_opendata_put(opendata);
2126 	return ret;
2127 }
2128 
2129 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2130 {
2131 	struct nfs_server *server = NFS_SERVER(state->inode);
2132 	struct nfs4_exception exception = { };
2133 	int err;
2134 
2135 	do {
2136 		err = _nfs4_open_expired(ctx, state);
2137 		trace_nfs4_open_expired(ctx, 0, err);
2138 		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2139 			continue;
2140 		switch (err) {
2141 		default:
2142 			goto out;
2143 		case -NFS4ERR_GRACE:
2144 		case -NFS4ERR_DELAY:
2145 			nfs4_handle_exception(server, err, &exception);
2146 			err = 0;
2147 		}
2148 	} while (exception.retry);
2149 out:
2150 	return err;
2151 }
2152 
2153 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2154 {
2155 	struct nfs_open_context *ctx;
2156 	int ret;
2157 
2158 	ctx = nfs4_state_find_open_context(state);
2159 	if (IS_ERR(ctx))
2160 		return -EAGAIN;
2161 	ret = nfs4_do_open_expired(ctx, state);
2162 	put_nfs_open_context(ctx);
2163 	return ret;
2164 }
2165 
2166 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2167 {
2168 	nfs_remove_bad_delegation(state->inode);
2169 	write_seqlock(&state->seqlock);
2170 	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2171 	write_sequnlock(&state->seqlock);
2172 	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2173 }
2174 
2175 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2176 {
2177 	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2178 		nfs_finish_clear_delegation_stateid(state);
2179 }
2180 
2181 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2182 {
2183 	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2184 	nfs40_clear_delegation_stateid(state);
2185 	return nfs4_open_expired(sp, state);
2186 }
2187 
2188 #if defined(CONFIG_NFS_V4_1)
2189 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2190 {
2191 	struct nfs_server *server = NFS_SERVER(state->inode);
2192 	nfs4_stateid stateid;
2193 	struct nfs_delegation *delegation;
2194 	struct rpc_cred *cred;
2195 	int status;
2196 
2197 	/* Get the delegation credential for use by test/free_stateid */
2198 	rcu_read_lock();
2199 	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2200 	if (delegation == NULL) {
2201 		rcu_read_unlock();
2202 		return;
2203 	}
2204 
2205 	nfs4_stateid_copy(&stateid, &delegation->stateid);
2206 	cred = get_rpccred(delegation->cred);
2207 	rcu_read_unlock();
2208 	status = nfs41_test_stateid(server, &stateid, cred);
2209 	trace_nfs4_test_delegation_stateid(state, NULL, status);
2210 
2211 	if (status != NFS_OK) {
2212 		/* Free the stateid unless the server explicitly
2213 		 * informs us the stateid is unrecognized. */
2214 		if (status != -NFS4ERR_BAD_STATEID)
2215 			nfs41_free_stateid(server, &stateid, cred);
2216 		nfs_finish_clear_delegation_stateid(state);
2217 	}
2218 
2219 	put_rpccred(cred);
2220 }
2221 
2222 /**
2223  * nfs41_check_open_stateid - possibly free an open stateid
2224  *
2225  * @state: NFSv4 state for an inode
2226  *
2227  * Returns NFS_OK if recovery for this stateid is now finished.
2228  * Otherwise a negative NFS4ERR value is returned.
2229  */
2230 static int nfs41_check_open_stateid(struct nfs4_state *state)
2231 {
2232 	struct nfs_server *server = NFS_SERVER(state->inode);
2233 	nfs4_stateid *stateid = &state->open_stateid;
2234 	struct rpc_cred *cred = state->owner->so_cred;
2235 	int status;
2236 
2237 	/* If a state reset has been done, test_stateid is unneeded */
2238 	if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2239 	    (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2240 	    (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2241 		return -NFS4ERR_BAD_STATEID;
2242 
2243 	status = nfs41_test_stateid(server, stateid, cred);
2244 	trace_nfs4_test_open_stateid(state, NULL, status);
2245 	if (status != NFS_OK) {
2246 		/* Free the stateid unless the server explicitly
2247 		 * informs us the stateid is unrecognized. */
2248 		if (status != -NFS4ERR_BAD_STATEID)
2249 			nfs41_free_stateid(server, stateid, cred);
2250 
2251 		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2252 		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2253 		clear_bit(NFS_O_RDWR_STATE, &state->flags);
2254 		clear_bit(NFS_OPEN_STATE, &state->flags);
2255 	}
2256 	return status;
2257 }
2258 
2259 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2260 {
2261 	int status;
2262 
2263 	nfs41_check_delegation_stateid(state);
2264 	status = nfs41_check_open_stateid(state);
2265 	if (status != NFS_OK)
2266 		status = nfs4_open_expired(sp, state);
2267 	return status;
2268 }
2269 #endif
2270 
2271 /*
2272  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2273  * fields corresponding to attributes that were used to store the verifier.
2274  * Make sure we clobber those fields in the later setattr call
2275  */
2276 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2277 {
2278 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2279 	    !(sattr->ia_valid & ATTR_ATIME_SET))
2280 		sattr->ia_valid |= ATTR_ATIME;
2281 
2282 	if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2283 	    !(sattr->ia_valid & ATTR_MTIME_SET))
2284 		sattr->ia_valid |= ATTR_MTIME;
2285 }
2286 
2287 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2288 		fmode_t fmode,
2289 		int flags,
2290 		struct nfs_open_context *ctx)
2291 {
2292 	struct nfs4_state_owner *sp = opendata->owner;
2293 	struct nfs_server *server = sp->so_server;
2294 	struct dentry *dentry;
2295 	struct nfs4_state *state;
2296 	unsigned int seq;
2297 	int ret;
2298 
2299 	seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2300 
2301 	ret = _nfs4_proc_open(opendata);
2302 	if (ret != 0)
2303 		goto out;
2304 
2305 	state = nfs4_opendata_to_nfs4_state(opendata);
2306 	ret = PTR_ERR(state);
2307 	if (IS_ERR(state))
2308 		goto out;
2309 	if (server->caps & NFS_CAP_POSIX_LOCK)
2310 		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2311 
2312 	dentry = opendata->dentry;
2313 	if (dentry->d_inode == NULL) {
2314 		/* FIXME: Is this d_drop() ever needed? */
2315 		d_drop(dentry);
2316 		dentry = d_add_unique(dentry, igrab(state->inode));
2317 		if (dentry == NULL) {
2318 			dentry = opendata->dentry;
2319 		} else if (dentry != ctx->dentry) {
2320 			dput(ctx->dentry);
2321 			ctx->dentry = dget(dentry);
2322 		}
2323 		nfs_set_verifier(dentry,
2324 				nfs_save_change_attribute(opendata->dir->d_inode));
2325 	}
2326 
2327 	ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2328 	if (ret != 0)
2329 		goto out;
2330 
2331 	ctx->state = state;
2332 	if (dentry->d_inode == state->inode) {
2333 		nfs_inode_attach_open_context(ctx);
2334 		if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2335 			nfs4_schedule_stateid_recovery(server, state);
2336 	}
2337 out:
2338 	return ret;
2339 }
2340 
2341 /*
2342  * Returns a referenced nfs4_state
2343  */
2344 static int _nfs4_do_open(struct inode *dir,
2345 			struct nfs_open_context *ctx,
2346 			int flags,
2347 			struct iattr *sattr,
2348 			struct nfs4_label *label,
2349 			int *opened)
2350 {
2351 	struct nfs4_state_owner  *sp;
2352 	struct nfs4_state     *state = NULL;
2353 	struct nfs_server       *server = NFS_SERVER(dir);
2354 	struct nfs4_opendata *opendata;
2355 	struct dentry *dentry = ctx->dentry;
2356 	struct rpc_cred *cred = ctx->cred;
2357 	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2358 	fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2359 	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2360 	struct nfs4_label *olabel = NULL;
2361 	int status;
2362 
2363 	/* Protect against reboot recovery conflicts */
2364 	status = -ENOMEM;
2365 	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2366 	if (sp == NULL) {
2367 		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2368 		goto out_err;
2369 	}
2370 	status = nfs4_recover_expired_lease(server);
2371 	if (status != 0)
2372 		goto err_put_state_owner;
2373 	if (dentry->d_inode != NULL)
2374 		nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
2375 	status = -ENOMEM;
2376 	if (dentry->d_inode)
2377 		claim = NFS4_OPEN_CLAIM_FH;
2378 	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2379 			label, claim, GFP_KERNEL);
2380 	if (opendata == NULL)
2381 		goto err_put_state_owner;
2382 
2383 	if (label) {
2384 		olabel = nfs4_label_alloc(server, GFP_KERNEL);
2385 		if (IS_ERR(olabel)) {
2386 			status = PTR_ERR(olabel);
2387 			goto err_opendata_put;
2388 		}
2389 	}
2390 
2391 	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2392 		if (!opendata->f_attr.mdsthreshold) {
2393 			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2394 			if (!opendata->f_attr.mdsthreshold)
2395 				goto err_free_label;
2396 		}
2397 		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2398 	}
2399 	if (dentry->d_inode != NULL)
2400 		opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
2401 
2402 	status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2403 	if (status != 0)
2404 		goto err_free_label;
2405 	state = ctx->state;
2406 
2407 	if ((opendata->o_arg.open_flags & O_EXCL) &&
2408 	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2409 		nfs4_exclusive_attrset(opendata, sattr);
2410 
2411 		nfs_fattr_init(opendata->o_res.f_attr);
2412 		status = nfs4_do_setattr(state->inode, cred,
2413 				opendata->o_res.f_attr, sattr,
2414 				state, label, olabel);
2415 		if (status == 0) {
2416 			nfs_setattr_update_inode(state->inode, sattr);
2417 			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
2418 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2419 		}
2420 	}
2421 	if (opendata->file_created)
2422 		*opened |= FILE_CREATED;
2423 
2424 	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2425 		*ctx_th = opendata->f_attr.mdsthreshold;
2426 		opendata->f_attr.mdsthreshold = NULL;
2427 	}
2428 
2429 	nfs4_label_free(olabel);
2430 
2431 	nfs4_opendata_put(opendata);
2432 	nfs4_put_state_owner(sp);
2433 	return 0;
2434 err_free_label:
2435 	nfs4_label_free(olabel);
2436 err_opendata_put:
2437 	nfs4_opendata_put(opendata);
2438 err_put_state_owner:
2439 	nfs4_put_state_owner(sp);
2440 out_err:
2441 	return status;
2442 }
2443 
2444 
2445 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2446 					struct nfs_open_context *ctx,
2447 					int flags,
2448 					struct iattr *sattr,
2449 					struct nfs4_label *label,
2450 					int *opened)
2451 {
2452 	struct nfs_server *server = NFS_SERVER(dir);
2453 	struct nfs4_exception exception = { };
2454 	struct nfs4_state *res;
2455 	int status;
2456 
2457 	do {
2458 		status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2459 		res = ctx->state;
2460 		trace_nfs4_open_file(ctx, flags, status);
2461 		if (status == 0)
2462 			break;
2463 		/* NOTE: BAD_SEQID means the server and client disagree about the
2464 		 * book-keeping w.r.t. state-changing operations
2465 		 * (OPEN/CLOSE/LOCK/LOCKU...)
2466 		 * It is actually a sign of a bug on the client or on the server.
2467 		 *
2468 		 * If we receive a BAD_SEQID error in the particular case of
2469 		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2470 		 * have unhashed the old state_owner for us, and that we can
2471 		 * therefore safely retry using a new one. We should still warn
2472 		 * the user though...
2473 		 */
2474 		if (status == -NFS4ERR_BAD_SEQID) {
2475 			pr_warn_ratelimited("NFS: v4 server %s "
2476 					" returned a bad sequence-id error!\n",
2477 					NFS_SERVER(dir)->nfs_client->cl_hostname);
2478 			exception.retry = 1;
2479 			continue;
2480 		}
2481 		/*
2482 		 * BAD_STATEID on OPEN means that the server cancelled our
2483 		 * state before it received the OPEN_CONFIRM.
2484 		 * Recover by retrying the request as per the discussion
2485 		 * on Page 181 of RFC3530.
2486 		 */
2487 		if (status == -NFS4ERR_BAD_STATEID) {
2488 			exception.retry = 1;
2489 			continue;
2490 		}
2491 		if (status == -EAGAIN) {
2492 			/* We must have found a delegation */
2493 			exception.retry = 1;
2494 			continue;
2495 		}
2496 		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2497 			continue;
2498 		res = ERR_PTR(nfs4_handle_exception(server,
2499 					status, &exception));
2500 	} while (exception.retry);
2501 	return res;
2502 }
2503 
2504 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2505 			    struct nfs_fattr *fattr, struct iattr *sattr,
2506 			    struct nfs4_state *state, struct nfs4_label *ilabel,
2507 			    struct nfs4_label *olabel)
2508 {
2509 	struct nfs_server *server = NFS_SERVER(inode);
2510         struct nfs_setattrargs  arg = {
2511                 .fh             = NFS_FH(inode),
2512                 .iap            = sattr,
2513 		.server		= server,
2514 		.bitmask = server->attr_bitmask,
2515 		.label		= ilabel,
2516         };
2517         struct nfs_setattrres  res = {
2518 		.fattr		= fattr,
2519 		.label		= olabel,
2520 		.server		= server,
2521         };
2522         struct rpc_message msg = {
2523 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2524 		.rpc_argp	= &arg,
2525 		.rpc_resp	= &res,
2526 		.rpc_cred	= cred,
2527         };
2528 	unsigned long timestamp = jiffies;
2529 	fmode_t fmode;
2530 	bool truncate;
2531 	int status;
2532 
2533 	arg.bitmask = nfs4_bitmask(server, ilabel);
2534 	if (ilabel)
2535 		arg.bitmask = nfs4_bitmask(server, olabel);
2536 
2537 	nfs_fattr_init(fattr);
2538 
2539 	/* Servers should only apply open mode checks for file size changes */
2540 	truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2541 	fmode = truncate ? FMODE_WRITE : FMODE_READ;
2542 
2543 	if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2544 		/* Use that stateid */
2545 	} else if (truncate && state != NULL) {
2546 		struct nfs_lockowner lockowner = {
2547 			.l_owner = current->files,
2548 			.l_pid = current->tgid,
2549 		};
2550 		if (!nfs4_valid_open_stateid(state))
2551 			return -EBADF;
2552 		if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2553 				&lockowner) == -EIO)
2554 			return -EBADF;
2555 	} else
2556 		nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2557 
2558 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2559 	if (status == 0 && state != NULL)
2560 		renew_lease(server, timestamp);
2561 	return status;
2562 }
2563 
2564 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2565 			   struct nfs_fattr *fattr, struct iattr *sattr,
2566 			   struct nfs4_state *state, struct nfs4_label *ilabel,
2567 			   struct nfs4_label *olabel)
2568 {
2569 	struct nfs_server *server = NFS_SERVER(inode);
2570 	struct nfs4_exception exception = {
2571 		.state = state,
2572 		.inode = inode,
2573 	};
2574 	int err;
2575 	do {
2576 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2577 		trace_nfs4_setattr(inode, err);
2578 		switch (err) {
2579 		case -NFS4ERR_OPENMODE:
2580 			if (!(sattr->ia_valid & ATTR_SIZE)) {
2581 				pr_warn_once("NFSv4: server %s is incorrectly "
2582 						"applying open mode checks to "
2583 						"a SETATTR that is not "
2584 						"changing file size.\n",
2585 						server->nfs_client->cl_hostname);
2586 			}
2587 			if (state && !(state->state & FMODE_WRITE)) {
2588 				err = -EBADF;
2589 				if (sattr->ia_valid & ATTR_OPEN)
2590 					err = -EACCES;
2591 				goto out;
2592 			}
2593 		}
2594 		err = nfs4_handle_exception(server, err, &exception);
2595 	} while (exception.retry);
2596 out:
2597 	return err;
2598 }
2599 
2600 struct nfs4_closedata {
2601 	struct inode *inode;
2602 	struct nfs4_state *state;
2603 	struct nfs_closeargs arg;
2604 	struct nfs_closeres res;
2605 	struct nfs_fattr fattr;
2606 	unsigned long timestamp;
2607 	bool roc;
2608 	u32 roc_barrier;
2609 };
2610 
2611 static void nfs4_free_closedata(void *data)
2612 {
2613 	struct nfs4_closedata *calldata = data;
2614 	struct nfs4_state_owner *sp = calldata->state->owner;
2615 	struct super_block *sb = calldata->state->inode->i_sb;
2616 
2617 	if (calldata->roc)
2618 		pnfs_roc_release(calldata->state->inode);
2619 	nfs4_put_open_state(calldata->state);
2620 	nfs_free_seqid(calldata->arg.seqid);
2621 	nfs4_put_state_owner(sp);
2622 	nfs_sb_deactive(sb);
2623 	kfree(calldata);
2624 }
2625 
2626 static void nfs4_close_done(struct rpc_task *task, void *data)
2627 {
2628 	struct nfs4_closedata *calldata = data;
2629 	struct nfs4_state *state = calldata->state;
2630 	struct nfs_server *server = NFS_SERVER(calldata->inode);
2631 	nfs4_stateid *res_stateid = NULL;
2632 
2633 	dprintk("%s: begin!\n", __func__);
2634 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2635 		return;
2636 	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2637         /* hmm. we are done with the inode, and in the process of freeing
2638 	 * the state_owner. we keep this around to process errors
2639 	 */
2640 	switch (task->tk_status) {
2641 		case 0:
2642 			res_stateid = &calldata->res.stateid;
2643 			if (calldata->arg.fmode == 0 && calldata->roc)
2644 				pnfs_roc_set_barrier(state->inode,
2645 						     calldata->roc_barrier);
2646 			renew_lease(server, calldata->timestamp);
2647 			break;
2648 		case -NFS4ERR_ADMIN_REVOKED:
2649 		case -NFS4ERR_STALE_STATEID:
2650 		case -NFS4ERR_OLD_STATEID:
2651 		case -NFS4ERR_BAD_STATEID:
2652 		case -NFS4ERR_EXPIRED:
2653 			if (!nfs4_stateid_match(&calldata->arg.stateid,
2654 						&state->stateid)) {
2655 				rpc_restart_call_prepare(task);
2656 				goto out_release;
2657 			}
2658 			if (calldata->arg.fmode == 0)
2659 				break;
2660 		default:
2661 			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2662 				rpc_restart_call_prepare(task);
2663 				goto out_release;
2664 			}
2665 	}
2666 	nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2667 out_release:
2668 	nfs_release_seqid(calldata->arg.seqid);
2669 	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2670 	dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2671 }
2672 
2673 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2674 {
2675 	struct nfs4_closedata *calldata = data;
2676 	struct nfs4_state *state = calldata->state;
2677 	struct inode *inode = calldata->inode;
2678 	bool is_rdonly, is_wronly, is_rdwr;
2679 	int call_close = 0;
2680 
2681 	dprintk("%s: begin!\n", __func__);
2682 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2683 		goto out_wait;
2684 
2685 	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2686 	spin_lock(&state->owner->so_lock);
2687 	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2688 	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2689 	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2690 	nfs4_stateid_copy(&calldata->arg.stateid, &state->stateid);
2691 	/* Calculate the change in open mode */
2692 	calldata->arg.fmode = 0;
2693 	if (state->n_rdwr == 0) {
2694 		if (state->n_rdonly == 0)
2695 			call_close |= is_rdonly;
2696 		else if (is_rdonly)
2697 			calldata->arg.fmode |= FMODE_READ;
2698 		if (state->n_wronly == 0)
2699 			call_close |= is_wronly;
2700 		else if (is_wronly)
2701 			calldata->arg.fmode |= FMODE_WRITE;
2702 	} else if (is_rdwr)
2703 		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2704 
2705 	if (calldata->arg.fmode == 0)
2706 		call_close |= is_rdwr;
2707 
2708 	if (!nfs4_valid_open_stateid(state))
2709 		call_close = 0;
2710 	spin_unlock(&state->owner->so_lock);
2711 
2712 	if (!call_close) {
2713 		/* Note: exit _without_ calling nfs4_close_done */
2714 		goto out_no_action;
2715 	}
2716 
2717 	if (calldata->arg.fmode == 0) {
2718 		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2719 		if (calldata->roc &&
2720 		    pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2721 			nfs_release_seqid(calldata->arg.seqid);
2722 			goto out_wait;
2723 		    }
2724 	}
2725 	calldata->arg.share_access =
2726 		nfs4_map_atomic_open_share(NFS_SERVER(inode),
2727 				calldata->arg.fmode, 0);
2728 
2729 	nfs_fattr_init(calldata->res.fattr);
2730 	calldata->timestamp = jiffies;
2731 	if (nfs4_setup_sequence(NFS_SERVER(inode),
2732 				&calldata->arg.seq_args,
2733 				&calldata->res.seq_res,
2734 				task) != 0)
2735 		nfs_release_seqid(calldata->arg.seqid);
2736 	dprintk("%s: done!\n", __func__);
2737 	return;
2738 out_no_action:
2739 	task->tk_action = NULL;
2740 out_wait:
2741 	nfs4_sequence_done(task, &calldata->res.seq_res);
2742 }
2743 
2744 static const struct rpc_call_ops nfs4_close_ops = {
2745 	.rpc_call_prepare = nfs4_close_prepare,
2746 	.rpc_call_done = nfs4_close_done,
2747 	.rpc_release = nfs4_free_closedata,
2748 };
2749 
2750 static bool nfs4_roc(struct inode *inode)
2751 {
2752 	if (!nfs_have_layout(inode))
2753 		return false;
2754 	return pnfs_roc(inode);
2755 }
2756 
2757 /*
2758  * It is possible for data to be read/written from a mem-mapped file
2759  * after the sys_close call (which hits the vfs layer as a flush).
2760  * This means that we can't safely call nfsv4 close on a file until
2761  * the inode is cleared. This in turn means that we are not good
2762  * NFSv4 citizens - we do not indicate to the server to update the file's
2763  * share state even when we are done with one of the three share
2764  * stateid's in the inode.
2765  *
2766  * NOTE: Caller must be holding the sp->so_owner semaphore!
2767  */
2768 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2769 {
2770 	struct nfs_server *server = NFS_SERVER(state->inode);
2771 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2772 	struct nfs4_closedata *calldata;
2773 	struct nfs4_state_owner *sp = state->owner;
2774 	struct rpc_task *task;
2775 	struct rpc_message msg = {
2776 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2777 		.rpc_cred = state->owner->so_cred,
2778 	};
2779 	struct rpc_task_setup task_setup_data = {
2780 		.rpc_client = server->client,
2781 		.rpc_message = &msg,
2782 		.callback_ops = &nfs4_close_ops,
2783 		.workqueue = nfsiod_workqueue,
2784 		.flags = RPC_TASK_ASYNC,
2785 	};
2786 	int status = -ENOMEM;
2787 
2788 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2789 		&task_setup_data.rpc_client, &msg);
2790 
2791 	calldata = kzalloc(sizeof(*calldata), gfp_mask);
2792 	if (calldata == NULL)
2793 		goto out;
2794 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2795 	calldata->inode = state->inode;
2796 	calldata->state = state;
2797 	calldata->arg.fh = NFS_FH(state->inode);
2798 	/* Serialization for the sequence id */
2799 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2800 	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2801 	if (IS_ERR(calldata->arg.seqid))
2802 		goto out_free_calldata;
2803 	calldata->arg.fmode = 0;
2804 	calldata->arg.bitmask = server->cache_consistency_bitmask;
2805 	calldata->res.fattr = &calldata->fattr;
2806 	calldata->res.seqid = calldata->arg.seqid;
2807 	calldata->res.server = server;
2808 	calldata->roc = nfs4_roc(state->inode);
2809 	nfs_sb_active(calldata->inode->i_sb);
2810 
2811 	msg.rpc_argp = &calldata->arg;
2812 	msg.rpc_resp = &calldata->res;
2813 	task_setup_data.callback_data = calldata;
2814 	task = rpc_run_task(&task_setup_data);
2815 	if (IS_ERR(task))
2816 		return PTR_ERR(task);
2817 	status = 0;
2818 	if (wait)
2819 		status = rpc_wait_for_completion_task(task);
2820 	rpc_put_task(task);
2821 	return status;
2822 out_free_calldata:
2823 	kfree(calldata);
2824 out:
2825 	nfs4_put_open_state(state);
2826 	nfs4_put_state_owner(sp);
2827 	return status;
2828 }
2829 
2830 static struct inode *
2831 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2832 		int open_flags, struct iattr *attr, int *opened)
2833 {
2834 	struct nfs4_state *state;
2835 	struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2836 
2837 	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2838 
2839 	/* Protect against concurrent sillydeletes */
2840 	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2841 
2842 	nfs4_label_release_security(label);
2843 
2844 	if (IS_ERR(state))
2845 		return ERR_CAST(state);
2846 	return state->inode;
2847 }
2848 
2849 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2850 {
2851 	if (ctx->state == NULL)
2852 		return;
2853 	if (is_sync)
2854 		nfs4_close_sync(ctx->state, ctx->mode);
2855 	else
2856 		nfs4_close_state(ctx->state, ctx->mode);
2857 }
2858 
2859 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2860 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2861 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2862 
2863 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2864 {
2865 	struct nfs4_server_caps_arg args = {
2866 		.fhandle = fhandle,
2867 	};
2868 	struct nfs4_server_caps_res res = {};
2869 	struct rpc_message msg = {
2870 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2871 		.rpc_argp = &args,
2872 		.rpc_resp = &res,
2873 	};
2874 	int status;
2875 
2876 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2877 	if (status == 0) {
2878 		/* Sanity check the server answers */
2879 		switch (server->nfs_client->cl_minorversion) {
2880 		case 0:
2881 			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2882 			res.attr_bitmask[2] = 0;
2883 			break;
2884 		case 1:
2885 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2886 			break;
2887 		case 2:
2888 			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2889 		}
2890 		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2891 		server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2892 				NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2893 				NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2894 				NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2895 				NFS_CAP_CTIME|NFS_CAP_MTIME|
2896 				NFS_CAP_SECURITY_LABEL);
2897 		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2898 				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2899 			server->caps |= NFS_CAP_ACLS;
2900 		if (res.has_links != 0)
2901 			server->caps |= NFS_CAP_HARDLINKS;
2902 		if (res.has_symlinks != 0)
2903 			server->caps |= NFS_CAP_SYMLINKS;
2904 		if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2905 			server->caps |= NFS_CAP_FILEID;
2906 		if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2907 			server->caps |= NFS_CAP_MODE;
2908 		if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2909 			server->caps |= NFS_CAP_NLINK;
2910 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2911 			server->caps |= NFS_CAP_OWNER;
2912 		if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2913 			server->caps |= NFS_CAP_OWNER_GROUP;
2914 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2915 			server->caps |= NFS_CAP_ATIME;
2916 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2917 			server->caps |= NFS_CAP_CTIME;
2918 		if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2919 			server->caps |= NFS_CAP_MTIME;
2920 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2921 		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2922 			server->caps |= NFS_CAP_SECURITY_LABEL;
2923 #endif
2924 		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2925 				sizeof(server->attr_bitmask));
2926 		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2927 
2928 		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2929 		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2930 		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2931 		server->cache_consistency_bitmask[2] = 0;
2932 		server->acl_bitmask = res.acl_bitmask;
2933 		server->fh_expire_type = res.fh_expire_type;
2934 	}
2935 
2936 	return status;
2937 }
2938 
2939 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2940 {
2941 	struct nfs4_exception exception = { };
2942 	int err;
2943 	do {
2944 		err = nfs4_handle_exception(server,
2945 				_nfs4_server_capabilities(server, fhandle),
2946 				&exception);
2947 	} while (exception.retry);
2948 	return err;
2949 }
2950 
2951 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2952 		struct nfs_fsinfo *info)
2953 {
2954 	u32 bitmask[3];
2955 	struct nfs4_lookup_root_arg args = {
2956 		.bitmask = bitmask,
2957 	};
2958 	struct nfs4_lookup_res res = {
2959 		.server = server,
2960 		.fattr = info->fattr,
2961 		.fh = fhandle,
2962 	};
2963 	struct rpc_message msg = {
2964 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2965 		.rpc_argp = &args,
2966 		.rpc_resp = &res,
2967 	};
2968 
2969 	bitmask[0] = nfs4_fattr_bitmap[0];
2970 	bitmask[1] = nfs4_fattr_bitmap[1];
2971 	/*
2972 	 * Process the label in the upcoming getfattr
2973 	 */
2974 	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2975 
2976 	nfs_fattr_init(info->fattr);
2977 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2978 }
2979 
2980 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2981 		struct nfs_fsinfo *info)
2982 {
2983 	struct nfs4_exception exception = { };
2984 	int err;
2985 	do {
2986 		err = _nfs4_lookup_root(server, fhandle, info);
2987 		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
2988 		switch (err) {
2989 		case 0:
2990 		case -NFS4ERR_WRONGSEC:
2991 			goto out;
2992 		default:
2993 			err = nfs4_handle_exception(server, err, &exception);
2994 		}
2995 	} while (exception.retry);
2996 out:
2997 	return err;
2998 }
2999 
3000 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3001 				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3002 {
3003 	struct rpc_auth_create_args auth_args = {
3004 		.pseudoflavor = flavor,
3005 	};
3006 	struct rpc_auth *auth;
3007 	int ret;
3008 
3009 	auth = rpcauth_create(&auth_args, server->client);
3010 	if (IS_ERR(auth)) {
3011 		ret = -EACCES;
3012 		goto out;
3013 	}
3014 	ret = nfs4_lookup_root(server, fhandle, info);
3015 out:
3016 	return ret;
3017 }
3018 
3019 /*
3020  * Retry pseudoroot lookup with various security flavors.  We do this when:
3021  *
3022  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3023  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3024  *
3025  * Returns zero on success, or a negative NFS4ERR value, or a
3026  * negative errno value.
3027  */
3028 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3029 			      struct nfs_fsinfo *info)
3030 {
3031 	/* Per 3530bis 15.33.5 */
3032 	static const rpc_authflavor_t flav_array[] = {
3033 		RPC_AUTH_GSS_KRB5P,
3034 		RPC_AUTH_GSS_KRB5I,
3035 		RPC_AUTH_GSS_KRB5,
3036 		RPC_AUTH_UNIX,			/* courtesy */
3037 		RPC_AUTH_NULL,
3038 	};
3039 	int status = -EPERM;
3040 	size_t i;
3041 
3042 	if (server->auth_info.flavor_len > 0) {
3043 		/* try each flavor specified by user */
3044 		for (i = 0; i < server->auth_info.flavor_len; i++) {
3045 			status = nfs4_lookup_root_sec(server, fhandle, info,
3046 						server->auth_info.flavors[i]);
3047 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3048 				continue;
3049 			break;
3050 		}
3051 	} else {
3052 		/* no flavors specified by user, try default list */
3053 		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3054 			status = nfs4_lookup_root_sec(server, fhandle, info,
3055 						      flav_array[i]);
3056 			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3057 				continue;
3058 			break;
3059 		}
3060 	}
3061 
3062 	/*
3063 	 * -EACCESS could mean that the user doesn't have correct permissions
3064 	 * to access the mount.  It could also mean that we tried to mount
3065 	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3066 	 * existing mount programs don't handle -EACCES very well so it should
3067 	 * be mapped to -EPERM instead.
3068 	 */
3069 	if (status == -EACCES)
3070 		status = -EPERM;
3071 	return status;
3072 }
3073 
3074 static int nfs4_do_find_root_sec(struct nfs_server *server,
3075 		struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3076 {
3077 	int mv = server->nfs_client->cl_minorversion;
3078 	return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3079 }
3080 
3081 /**
3082  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3083  * @server: initialized nfs_server handle
3084  * @fhandle: we fill in the pseudo-fs root file handle
3085  * @info: we fill in an FSINFO struct
3086  * @auth_probe: probe the auth flavours
3087  *
3088  * Returns zero on success, or a negative errno.
3089  */
3090 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3091 			 struct nfs_fsinfo *info,
3092 			 bool auth_probe)
3093 {
3094 	int status;
3095 
3096 	switch (auth_probe) {
3097 	case false:
3098 		status = nfs4_lookup_root(server, fhandle, info);
3099 		if (status != -NFS4ERR_WRONGSEC)
3100 			break;
3101 	default:
3102 		status = nfs4_do_find_root_sec(server, fhandle, info);
3103 	}
3104 
3105 	if (status == 0)
3106 		status = nfs4_server_capabilities(server, fhandle);
3107 	if (status == 0)
3108 		status = nfs4_do_fsinfo(server, fhandle, info);
3109 
3110 	return nfs4_map_errors(status);
3111 }
3112 
3113 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3114 			      struct nfs_fsinfo *info)
3115 {
3116 	int error;
3117 	struct nfs_fattr *fattr = info->fattr;
3118 	struct nfs4_label *label = NULL;
3119 
3120 	error = nfs4_server_capabilities(server, mntfh);
3121 	if (error < 0) {
3122 		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3123 		return error;
3124 	}
3125 
3126 	label = nfs4_label_alloc(server, GFP_KERNEL);
3127 	if (IS_ERR(label))
3128 		return PTR_ERR(label);
3129 
3130 	error = nfs4_proc_getattr(server, mntfh, fattr, label);
3131 	if (error < 0) {
3132 		dprintk("nfs4_get_root: getattr error = %d\n", -error);
3133 		goto err_free_label;
3134 	}
3135 
3136 	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3137 	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3138 		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3139 
3140 err_free_label:
3141 	nfs4_label_free(label);
3142 
3143 	return error;
3144 }
3145 
3146 /*
3147  * Get locations and (maybe) other attributes of a referral.
3148  * Note that we'll actually follow the referral later when
3149  * we detect fsid mismatch in inode revalidation
3150  */
3151 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3152 			     const struct qstr *name, struct nfs_fattr *fattr,
3153 			     struct nfs_fh *fhandle)
3154 {
3155 	int status = -ENOMEM;
3156 	struct page *page = NULL;
3157 	struct nfs4_fs_locations *locations = NULL;
3158 
3159 	page = alloc_page(GFP_KERNEL);
3160 	if (page == NULL)
3161 		goto out;
3162 	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3163 	if (locations == NULL)
3164 		goto out;
3165 
3166 	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3167 	if (status != 0)
3168 		goto out;
3169 
3170 	/*
3171 	 * If the fsid didn't change, this is a migration event, not a
3172 	 * referral.  Cause us to drop into the exception handler, which
3173 	 * will kick off migration recovery.
3174 	 */
3175 	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3176 		dprintk("%s: server did not return a different fsid for"
3177 			" a referral at %s\n", __func__, name->name);
3178 		status = -NFS4ERR_MOVED;
3179 		goto out;
3180 	}
3181 	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3182 	nfs_fixup_referral_attributes(&locations->fattr);
3183 
3184 	/* replace the lookup nfs_fattr with the locations nfs_fattr */
3185 	memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3186 	memset(fhandle, 0, sizeof(struct nfs_fh));
3187 out:
3188 	if (page)
3189 		__free_page(page);
3190 	kfree(locations);
3191 	return status;
3192 }
3193 
3194 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3195 				struct nfs_fattr *fattr, struct nfs4_label *label)
3196 {
3197 	struct nfs4_getattr_arg args = {
3198 		.fh = fhandle,
3199 		.bitmask = server->attr_bitmask,
3200 	};
3201 	struct nfs4_getattr_res res = {
3202 		.fattr = fattr,
3203 		.label = label,
3204 		.server = server,
3205 	};
3206 	struct rpc_message msg = {
3207 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3208 		.rpc_argp = &args,
3209 		.rpc_resp = &res,
3210 	};
3211 
3212 	args.bitmask = nfs4_bitmask(server, label);
3213 
3214 	nfs_fattr_init(fattr);
3215 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3216 }
3217 
3218 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3219 				struct nfs_fattr *fattr, struct nfs4_label *label)
3220 {
3221 	struct nfs4_exception exception = { };
3222 	int err;
3223 	do {
3224 		err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3225 		trace_nfs4_getattr(server, fhandle, fattr, err);
3226 		err = nfs4_handle_exception(server, err,
3227 				&exception);
3228 	} while (exception.retry);
3229 	return err;
3230 }
3231 
3232 /*
3233  * The file is not closed if it is opened due to the a request to change
3234  * the size of the file. The open call will not be needed once the
3235  * VFS layer lookup-intents are implemented.
3236  *
3237  * Close is called when the inode is destroyed.
3238  * If we haven't opened the file for O_WRONLY, we
3239  * need to in the size_change case to obtain a stateid.
3240  *
3241  * Got race?
3242  * Because OPEN is always done by name in nfsv4, it is
3243  * possible that we opened a different file by the same
3244  * name.  We can recognize this race condition, but we
3245  * can't do anything about it besides returning an error.
3246  *
3247  * This will be fixed with VFS changes (lookup-intent).
3248  */
3249 static int
3250 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3251 		  struct iattr *sattr)
3252 {
3253 	struct inode *inode = dentry->d_inode;
3254 	struct rpc_cred *cred = NULL;
3255 	struct nfs4_state *state = NULL;
3256 	struct nfs4_label *label = NULL;
3257 	int status;
3258 
3259 	if (pnfs_ld_layoutret_on_setattr(inode) &&
3260 	    sattr->ia_valid & ATTR_SIZE &&
3261 	    sattr->ia_size < i_size_read(inode))
3262 		pnfs_commit_and_return_layout(inode);
3263 
3264 	nfs_fattr_init(fattr);
3265 
3266 	/* Deal with open(O_TRUNC) */
3267 	if (sattr->ia_valid & ATTR_OPEN)
3268 		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3269 
3270 	/* Optimization: if the end result is no change, don't RPC */
3271 	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3272 		return 0;
3273 
3274 	/* Search for an existing open(O_WRITE) file */
3275 	if (sattr->ia_valid & ATTR_FILE) {
3276 		struct nfs_open_context *ctx;
3277 
3278 		ctx = nfs_file_open_context(sattr->ia_file);
3279 		if (ctx) {
3280 			cred = ctx->cred;
3281 			state = ctx->state;
3282 		}
3283 	}
3284 
3285 	label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3286 	if (IS_ERR(label))
3287 		return PTR_ERR(label);
3288 
3289 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3290 	if (status == 0) {
3291 		nfs_setattr_update_inode(inode, sattr);
3292 		nfs_setsecurity(inode, fattr, label);
3293 	}
3294 	nfs4_label_free(label);
3295 	return status;
3296 }
3297 
3298 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3299 		const struct qstr *name, struct nfs_fh *fhandle,
3300 		struct nfs_fattr *fattr, struct nfs4_label *label)
3301 {
3302 	struct nfs_server *server = NFS_SERVER(dir);
3303 	int		       status;
3304 	struct nfs4_lookup_arg args = {
3305 		.bitmask = server->attr_bitmask,
3306 		.dir_fh = NFS_FH(dir),
3307 		.name = name,
3308 	};
3309 	struct nfs4_lookup_res res = {
3310 		.server = server,
3311 		.fattr = fattr,
3312 		.label = label,
3313 		.fh = fhandle,
3314 	};
3315 	struct rpc_message msg = {
3316 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3317 		.rpc_argp = &args,
3318 		.rpc_resp = &res,
3319 	};
3320 
3321 	args.bitmask = nfs4_bitmask(server, label);
3322 
3323 	nfs_fattr_init(fattr);
3324 
3325 	dprintk("NFS call  lookup %s\n", name->name);
3326 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3327 	dprintk("NFS reply lookup: %d\n", status);
3328 	return status;
3329 }
3330 
3331 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3332 {
3333 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3334 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3335 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3336 	fattr->nlink = 2;
3337 }
3338 
3339 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3340 				   struct qstr *name, struct nfs_fh *fhandle,
3341 				   struct nfs_fattr *fattr, struct nfs4_label *label)
3342 {
3343 	struct nfs4_exception exception = { };
3344 	struct rpc_clnt *client = *clnt;
3345 	int err;
3346 	do {
3347 		err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3348 		trace_nfs4_lookup(dir, name, err);
3349 		switch (err) {
3350 		case -NFS4ERR_BADNAME:
3351 			err = -ENOENT;
3352 			goto out;
3353 		case -NFS4ERR_MOVED:
3354 			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3355 			goto out;
3356 		case -NFS4ERR_WRONGSEC:
3357 			err = -EPERM;
3358 			if (client != *clnt)
3359 				goto out;
3360 			client = nfs4_negotiate_security(client, dir, name);
3361 			if (IS_ERR(client))
3362 				return PTR_ERR(client);
3363 
3364 			exception.retry = 1;
3365 			break;
3366 		default:
3367 			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3368 		}
3369 	} while (exception.retry);
3370 
3371 out:
3372 	if (err == 0)
3373 		*clnt = client;
3374 	else if (client != *clnt)
3375 		rpc_shutdown_client(client);
3376 
3377 	return err;
3378 }
3379 
3380 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3381 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3382 			    struct nfs4_label *label)
3383 {
3384 	int status;
3385 	struct rpc_clnt *client = NFS_CLIENT(dir);
3386 
3387 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3388 	if (client != NFS_CLIENT(dir)) {
3389 		rpc_shutdown_client(client);
3390 		nfs_fixup_secinfo_attributes(fattr);
3391 	}
3392 	return status;
3393 }
3394 
3395 struct rpc_clnt *
3396 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3397 			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3398 {
3399 	struct rpc_clnt *client = NFS_CLIENT(dir);
3400 	int status;
3401 
3402 	status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3403 	if (status < 0)
3404 		return ERR_PTR(status);
3405 	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3406 }
3407 
3408 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3409 {
3410 	struct nfs_server *server = NFS_SERVER(inode);
3411 	struct nfs4_accessargs args = {
3412 		.fh = NFS_FH(inode),
3413 		.bitmask = server->cache_consistency_bitmask,
3414 	};
3415 	struct nfs4_accessres res = {
3416 		.server = server,
3417 	};
3418 	struct rpc_message msg = {
3419 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3420 		.rpc_argp = &args,
3421 		.rpc_resp = &res,
3422 		.rpc_cred = entry->cred,
3423 	};
3424 	int mode = entry->mask;
3425 	int status = 0;
3426 
3427 	/*
3428 	 * Determine which access bits we want to ask for...
3429 	 */
3430 	if (mode & MAY_READ)
3431 		args.access |= NFS4_ACCESS_READ;
3432 	if (S_ISDIR(inode->i_mode)) {
3433 		if (mode & MAY_WRITE)
3434 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3435 		if (mode & MAY_EXEC)
3436 			args.access |= NFS4_ACCESS_LOOKUP;
3437 	} else {
3438 		if (mode & MAY_WRITE)
3439 			args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3440 		if (mode & MAY_EXEC)
3441 			args.access |= NFS4_ACCESS_EXECUTE;
3442 	}
3443 
3444 	res.fattr = nfs_alloc_fattr();
3445 	if (res.fattr == NULL)
3446 		return -ENOMEM;
3447 
3448 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3449 	if (!status) {
3450 		nfs_access_set_mask(entry, res.access);
3451 		nfs_refresh_inode(inode, res.fattr);
3452 	}
3453 	nfs_free_fattr(res.fattr);
3454 	return status;
3455 }
3456 
3457 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3458 {
3459 	struct nfs4_exception exception = { };
3460 	int err;
3461 	do {
3462 		err = _nfs4_proc_access(inode, entry);
3463 		trace_nfs4_access(inode, err);
3464 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3465 				&exception);
3466 	} while (exception.retry);
3467 	return err;
3468 }
3469 
3470 /*
3471  * TODO: For the time being, we don't try to get any attributes
3472  * along with any of the zero-copy operations READ, READDIR,
3473  * READLINK, WRITE.
3474  *
3475  * In the case of the first three, we want to put the GETATTR
3476  * after the read-type operation -- this is because it is hard
3477  * to predict the length of a GETATTR response in v4, and thus
3478  * align the READ data correctly.  This means that the GETATTR
3479  * may end up partially falling into the page cache, and we should
3480  * shift it into the 'tail' of the xdr_buf before processing.
3481  * To do this efficiently, we need to know the total length
3482  * of data received, which doesn't seem to be available outside
3483  * of the RPC layer.
3484  *
3485  * In the case of WRITE, we also want to put the GETATTR after
3486  * the operation -- in this case because we want to make sure
3487  * we get the post-operation mtime and size.
3488  *
3489  * Both of these changes to the XDR layer would in fact be quite
3490  * minor, but I decided to leave them for a subsequent patch.
3491  */
3492 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3493 		unsigned int pgbase, unsigned int pglen)
3494 {
3495 	struct nfs4_readlink args = {
3496 		.fh       = NFS_FH(inode),
3497 		.pgbase	  = pgbase,
3498 		.pglen    = pglen,
3499 		.pages    = &page,
3500 	};
3501 	struct nfs4_readlink_res res;
3502 	struct rpc_message msg = {
3503 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3504 		.rpc_argp = &args,
3505 		.rpc_resp = &res,
3506 	};
3507 
3508 	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3509 }
3510 
3511 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3512 		unsigned int pgbase, unsigned int pglen)
3513 {
3514 	struct nfs4_exception exception = { };
3515 	int err;
3516 	do {
3517 		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3518 		trace_nfs4_readlink(inode, err);
3519 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
3520 				&exception);
3521 	} while (exception.retry);
3522 	return err;
3523 }
3524 
3525 /*
3526  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3527  */
3528 static int
3529 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3530 		 int flags)
3531 {
3532 	struct nfs4_label l, *ilabel = NULL;
3533 	struct nfs_open_context *ctx;
3534 	struct nfs4_state *state;
3535 	int opened = 0;
3536 	int status = 0;
3537 
3538 	ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3539 	if (IS_ERR(ctx))
3540 		return PTR_ERR(ctx);
3541 
3542 	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3543 
3544 	sattr->ia_mode &= ~current_umask();
3545 	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3546 	if (IS_ERR(state)) {
3547 		status = PTR_ERR(state);
3548 		goto out;
3549 	}
3550 out:
3551 	nfs4_label_release_security(ilabel);
3552 	put_nfs_open_context(ctx);
3553 	return status;
3554 }
3555 
3556 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3557 {
3558 	struct nfs_server *server = NFS_SERVER(dir);
3559 	struct nfs_removeargs args = {
3560 		.fh = NFS_FH(dir),
3561 		.name = *name,
3562 	};
3563 	struct nfs_removeres res = {
3564 		.server = server,
3565 	};
3566 	struct rpc_message msg = {
3567 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3568 		.rpc_argp = &args,
3569 		.rpc_resp = &res,
3570 	};
3571 	int status;
3572 
3573 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3574 	if (status == 0)
3575 		update_changeattr(dir, &res.cinfo);
3576 	return status;
3577 }
3578 
3579 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3580 {
3581 	struct nfs4_exception exception = { };
3582 	int err;
3583 	do {
3584 		err = _nfs4_proc_remove(dir, name);
3585 		trace_nfs4_remove(dir, name, err);
3586 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3587 				&exception);
3588 	} while (exception.retry);
3589 	return err;
3590 }
3591 
3592 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3593 {
3594 	struct nfs_server *server = NFS_SERVER(dir);
3595 	struct nfs_removeargs *args = msg->rpc_argp;
3596 	struct nfs_removeres *res = msg->rpc_resp;
3597 
3598 	res->server = server;
3599 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3600 	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3601 
3602 	nfs_fattr_init(res->dir_attr);
3603 }
3604 
3605 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3606 {
3607 	nfs4_setup_sequence(NFS_SERVER(data->dir),
3608 			&data->args.seq_args,
3609 			&data->res.seq_res,
3610 			task);
3611 }
3612 
3613 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3614 {
3615 	struct nfs_unlinkdata *data = task->tk_calldata;
3616 	struct nfs_removeres *res = &data->res;
3617 
3618 	if (!nfs4_sequence_done(task, &res->seq_res))
3619 		return 0;
3620 	if (nfs4_async_handle_error(task, res->server, NULL,
3621 				    &data->timeout) == -EAGAIN)
3622 		return 0;
3623 	update_changeattr(dir, &res->cinfo);
3624 	return 1;
3625 }
3626 
3627 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3628 {
3629 	struct nfs_server *server = NFS_SERVER(dir);
3630 	struct nfs_renameargs *arg = msg->rpc_argp;
3631 	struct nfs_renameres *res = msg->rpc_resp;
3632 
3633 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3634 	res->server = server;
3635 	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3636 }
3637 
3638 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3639 {
3640 	nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3641 			&data->args.seq_args,
3642 			&data->res.seq_res,
3643 			task);
3644 }
3645 
3646 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3647 				 struct inode *new_dir)
3648 {
3649 	struct nfs_renamedata *data = task->tk_calldata;
3650 	struct nfs_renameres *res = &data->res;
3651 
3652 	if (!nfs4_sequence_done(task, &res->seq_res))
3653 		return 0;
3654 	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3655 		return 0;
3656 
3657 	update_changeattr(old_dir, &res->old_cinfo);
3658 	update_changeattr(new_dir, &res->new_cinfo);
3659 	return 1;
3660 }
3661 
3662 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3663 {
3664 	struct nfs_server *server = NFS_SERVER(inode);
3665 	struct nfs4_link_arg arg = {
3666 		.fh     = NFS_FH(inode),
3667 		.dir_fh = NFS_FH(dir),
3668 		.name   = name,
3669 		.bitmask = server->attr_bitmask,
3670 	};
3671 	struct nfs4_link_res res = {
3672 		.server = server,
3673 		.label = NULL,
3674 	};
3675 	struct rpc_message msg = {
3676 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3677 		.rpc_argp = &arg,
3678 		.rpc_resp = &res,
3679 	};
3680 	int status = -ENOMEM;
3681 
3682 	res.fattr = nfs_alloc_fattr();
3683 	if (res.fattr == NULL)
3684 		goto out;
3685 
3686 	res.label = nfs4_label_alloc(server, GFP_KERNEL);
3687 	if (IS_ERR(res.label)) {
3688 		status = PTR_ERR(res.label);
3689 		goto out;
3690 	}
3691 	arg.bitmask = nfs4_bitmask(server, res.label);
3692 
3693 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3694 	if (!status) {
3695 		update_changeattr(dir, &res.cinfo);
3696 		status = nfs_post_op_update_inode(inode, res.fattr);
3697 		if (!status)
3698 			nfs_setsecurity(inode, res.fattr, res.label);
3699 	}
3700 
3701 
3702 	nfs4_label_free(res.label);
3703 
3704 out:
3705 	nfs_free_fattr(res.fattr);
3706 	return status;
3707 }
3708 
3709 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3710 {
3711 	struct nfs4_exception exception = { };
3712 	int err;
3713 	do {
3714 		err = nfs4_handle_exception(NFS_SERVER(inode),
3715 				_nfs4_proc_link(inode, dir, name),
3716 				&exception);
3717 	} while (exception.retry);
3718 	return err;
3719 }
3720 
3721 struct nfs4_createdata {
3722 	struct rpc_message msg;
3723 	struct nfs4_create_arg arg;
3724 	struct nfs4_create_res res;
3725 	struct nfs_fh fh;
3726 	struct nfs_fattr fattr;
3727 	struct nfs4_label *label;
3728 };
3729 
3730 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3731 		struct qstr *name, struct iattr *sattr, u32 ftype)
3732 {
3733 	struct nfs4_createdata *data;
3734 
3735 	data = kzalloc(sizeof(*data), GFP_KERNEL);
3736 	if (data != NULL) {
3737 		struct nfs_server *server = NFS_SERVER(dir);
3738 
3739 		data->label = nfs4_label_alloc(server, GFP_KERNEL);
3740 		if (IS_ERR(data->label))
3741 			goto out_free;
3742 
3743 		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3744 		data->msg.rpc_argp = &data->arg;
3745 		data->msg.rpc_resp = &data->res;
3746 		data->arg.dir_fh = NFS_FH(dir);
3747 		data->arg.server = server;
3748 		data->arg.name = name;
3749 		data->arg.attrs = sattr;
3750 		data->arg.ftype = ftype;
3751 		data->arg.bitmask = nfs4_bitmask(server, data->label);
3752 		data->res.server = server;
3753 		data->res.fh = &data->fh;
3754 		data->res.fattr = &data->fattr;
3755 		data->res.label = data->label;
3756 		nfs_fattr_init(data->res.fattr);
3757 	}
3758 	return data;
3759 out_free:
3760 	kfree(data);
3761 	return NULL;
3762 }
3763 
3764 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3765 {
3766 	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3767 				    &data->arg.seq_args, &data->res.seq_res, 1);
3768 	if (status == 0) {
3769 		update_changeattr(dir, &data->res.dir_cinfo);
3770 		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3771 	}
3772 	return status;
3773 }
3774 
3775 static void nfs4_free_createdata(struct nfs4_createdata *data)
3776 {
3777 	nfs4_label_free(data->label);
3778 	kfree(data);
3779 }
3780 
3781 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3782 		struct page *page, unsigned int len, struct iattr *sattr,
3783 		struct nfs4_label *label)
3784 {
3785 	struct nfs4_createdata *data;
3786 	int status = -ENAMETOOLONG;
3787 
3788 	if (len > NFS4_MAXPATHLEN)
3789 		goto out;
3790 
3791 	status = -ENOMEM;
3792 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3793 	if (data == NULL)
3794 		goto out;
3795 
3796 	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3797 	data->arg.u.symlink.pages = &page;
3798 	data->arg.u.symlink.len = len;
3799 	data->arg.label = label;
3800 
3801 	status = nfs4_do_create(dir, dentry, data);
3802 
3803 	nfs4_free_createdata(data);
3804 out:
3805 	return status;
3806 }
3807 
3808 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3809 		struct page *page, unsigned int len, struct iattr *sattr)
3810 {
3811 	struct nfs4_exception exception = { };
3812 	struct nfs4_label l, *label = NULL;
3813 	int err;
3814 
3815 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3816 
3817 	do {
3818 		err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3819 		trace_nfs4_symlink(dir, &dentry->d_name, err);
3820 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3821 				&exception);
3822 	} while (exception.retry);
3823 
3824 	nfs4_label_release_security(label);
3825 	return err;
3826 }
3827 
3828 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3829 		struct iattr *sattr, struct nfs4_label *label)
3830 {
3831 	struct nfs4_createdata *data;
3832 	int status = -ENOMEM;
3833 
3834 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3835 	if (data == NULL)
3836 		goto out;
3837 
3838 	data->arg.label = label;
3839 	status = nfs4_do_create(dir, dentry, data);
3840 
3841 	nfs4_free_createdata(data);
3842 out:
3843 	return status;
3844 }
3845 
3846 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3847 		struct iattr *sattr)
3848 {
3849 	struct nfs4_exception exception = { };
3850 	struct nfs4_label l, *label = NULL;
3851 	int err;
3852 
3853 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3854 
3855 	sattr->ia_mode &= ~current_umask();
3856 	do {
3857 		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3858 		trace_nfs4_mkdir(dir, &dentry->d_name, err);
3859 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3860 				&exception);
3861 	} while (exception.retry);
3862 	nfs4_label_release_security(label);
3863 
3864 	return err;
3865 }
3866 
3867 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3868 		u64 cookie, struct page **pages, unsigned int count, int plus)
3869 {
3870 	struct inode		*dir = dentry->d_inode;
3871 	struct nfs4_readdir_arg args = {
3872 		.fh = NFS_FH(dir),
3873 		.pages = pages,
3874 		.pgbase = 0,
3875 		.count = count,
3876 		.bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3877 		.plus = plus,
3878 	};
3879 	struct nfs4_readdir_res res;
3880 	struct rpc_message msg = {
3881 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3882 		.rpc_argp = &args,
3883 		.rpc_resp = &res,
3884 		.rpc_cred = cred,
3885 	};
3886 	int			status;
3887 
3888 	dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3889 			dentry,
3890 			(unsigned long long)cookie);
3891 	nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3892 	res.pgbase = args.pgbase;
3893 	status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3894 	if (status >= 0) {
3895 		memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3896 		status += args.pgbase;
3897 	}
3898 
3899 	nfs_invalidate_atime(dir);
3900 
3901 	dprintk("%s: returns %d\n", __func__, status);
3902 	return status;
3903 }
3904 
3905 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3906 		u64 cookie, struct page **pages, unsigned int count, int plus)
3907 {
3908 	struct nfs4_exception exception = { };
3909 	int err;
3910 	do {
3911 		err = _nfs4_proc_readdir(dentry, cred, cookie,
3912 				pages, count, plus);
3913 		trace_nfs4_readdir(dentry->d_inode, err);
3914 		err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode), err,
3915 				&exception);
3916 	} while (exception.retry);
3917 	return err;
3918 }
3919 
3920 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3921 		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3922 {
3923 	struct nfs4_createdata *data;
3924 	int mode = sattr->ia_mode;
3925 	int status = -ENOMEM;
3926 
3927 	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3928 	if (data == NULL)
3929 		goto out;
3930 
3931 	if (S_ISFIFO(mode))
3932 		data->arg.ftype = NF4FIFO;
3933 	else if (S_ISBLK(mode)) {
3934 		data->arg.ftype = NF4BLK;
3935 		data->arg.u.device.specdata1 = MAJOR(rdev);
3936 		data->arg.u.device.specdata2 = MINOR(rdev);
3937 	}
3938 	else if (S_ISCHR(mode)) {
3939 		data->arg.ftype = NF4CHR;
3940 		data->arg.u.device.specdata1 = MAJOR(rdev);
3941 		data->arg.u.device.specdata2 = MINOR(rdev);
3942 	} else if (!S_ISSOCK(mode)) {
3943 		status = -EINVAL;
3944 		goto out_free;
3945 	}
3946 
3947 	data->arg.label = label;
3948 	status = nfs4_do_create(dir, dentry, data);
3949 out_free:
3950 	nfs4_free_createdata(data);
3951 out:
3952 	return status;
3953 }
3954 
3955 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3956 		struct iattr *sattr, dev_t rdev)
3957 {
3958 	struct nfs4_exception exception = { };
3959 	struct nfs4_label l, *label = NULL;
3960 	int err;
3961 
3962 	label = nfs4_label_init_security(dir, dentry, sattr, &l);
3963 
3964 	sattr->ia_mode &= ~current_umask();
3965 	do {
3966 		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3967 		trace_nfs4_mknod(dir, &dentry->d_name, err);
3968 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
3969 				&exception);
3970 	} while (exception.retry);
3971 
3972 	nfs4_label_release_security(label);
3973 
3974 	return err;
3975 }
3976 
3977 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3978 		 struct nfs_fsstat *fsstat)
3979 {
3980 	struct nfs4_statfs_arg args = {
3981 		.fh = fhandle,
3982 		.bitmask = server->attr_bitmask,
3983 	};
3984 	struct nfs4_statfs_res res = {
3985 		.fsstat = fsstat,
3986 	};
3987 	struct rpc_message msg = {
3988 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3989 		.rpc_argp = &args,
3990 		.rpc_resp = &res,
3991 	};
3992 
3993 	nfs_fattr_init(fsstat->fattr);
3994 	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3995 }
3996 
3997 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3998 {
3999 	struct nfs4_exception exception = { };
4000 	int err;
4001 	do {
4002 		err = nfs4_handle_exception(server,
4003 				_nfs4_proc_statfs(server, fhandle, fsstat),
4004 				&exception);
4005 	} while (exception.retry);
4006 	return err;
4007 }
4008 
4009 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4010 		struct nfs_fsinfo *fsinfo)
4011 {
4012 	struct nfs4_fsinfo_arg args = {
4013 		.fh = fhandle,
4014 		.bitmask = server->attr_bitmask,
4015 	};
4016 	struct nfs4_fsinfo_res res = {
4017 		.fsinfo = fsinfo,
4018 	};
4019 	struct rpc_message msg = {
4020 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4021 		.rpc_argp = &args,
4022 		.rpc_resp = &res,
4023 	};
4024 
4025 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4026 }
4027 
4028 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4029 {
4030 	struct nfs4_exception exception = { };
4031 	unsigned long now = jiffies;
4032 	int err;
4033 
4034 	do {
4035 		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4036 		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4037 		if (err == 0) {
4038 			struct nfs_client *clp = server->nfs_client;
4039 
4040 			spin_lock(&clp->cl_lock);
4041 			clp->cl_lease_time = fsinfo->lease_time * HZ;
4042 			clp->cl_last_renewal = now;
4043 			spin_unlock(&clp->cl_lock);
4044 			break;
4045 		}
4046 		err = nfs4_handle_exception(server, err, &exception);
4047 	} while (exception.retry);
4048 	return err;
4049 }
4050 
4051 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4052 {
4053 	int error;
4054 
4055 	nfs_fattr_init(fsinfo->fattr);
4056 	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4057 	if (error == 0) {
4058 		/* block layout checks this! */
4059 		server->pnfs_blksize = fsinfo->blksize;
4060 		set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4061 	}
4062 
4063 	return error;
4064 }
4065 
4066 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4067 		struct nfs_pathconf *pathconf)
4068 {
4069 	struct nfs4_pathconf_arg args = {
4070 		.fh = fhandle,
4071 		.bitmask = server->attr_bitmask,
4072 	};
4073 	struct nfs4_pathconf_res res = {
4074 		.pathconf = pathconf,
4075 	};
4076 	struct rpc_message msg = {
4077 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4078 		.rpc_argp = &args,
4079 		.rpc_resp = &res,
4080 	};
4081 
4082 	/* None of the pathconf attributes are mandatory to implement */
4083 	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4084 		memset(pathconf, 0, sizeof(*pathconf));
4085 		return 0;
4086 	}
4087 
4088 	nfs_fattr_init(pathconf->fattr);
4089 	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4090 }
4091 
4092 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4093 		struct nfs_pathconf *pathconf)
4094 {
4095 	struct nfs4_exception exception = { };
4096 	int err;
4097 
4098 	do {
4099 		err = nfs4_handle_exception(server,
4100 				_nfs4_proc_pathconf(server, fhandle, pathconf),
4101 				&exception);
4102 	} while (exception.retry);
4103 	return err;
4104 }
4105 
4106 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4107 		const struct nfs_open_context *ctx,
4108 		const struct nfs_lock_context *l_ctx,
4109 		fmode_t fmode)
4110 {
4111 	const struct nfs_lockowner *lockowner = NULL;
4112 
4113 	if (l_ctx != NULL)
4114 		lockowner = &l_ctx->lockowner;
4115 	return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4116 }
4117 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4118 
4119 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4120 		const struct nfs_open_context *ctx,
4121 		const struct nfs_lock_context *l_ctx,
4122 		fmode_t fmode)
4123 {
4124 	nfs4_stateid current_stateid;
4125 
4126 	/* If the current stateid represents a lost lock, then exit */
4127 	if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4128 		return true;
4129 	return nfs4_stateid_match(stateid, &current_stateid);
4130 }
4131 
4132 static bool nfs4_error_stateid_expired(int err)
4133 {
4134 	switch (err) {
4135 	case -NFS4ERR_DELEG_REVOKED:
4136 	case -NFS4ERR_ADMIN_REVOKED:
4137 	case -NFS4ERR_BAD_STATEID:
4138 	case -NFS4ERR_STALE_STATEID:
4139 	case -NFS4ERR_OLD_STATEID:
4140 	case -NFS4ERR_OPENMODE:
4141 	case -NFS4ERR_EXPIRED:
4142 		return true;
4143 	}
4144 	return false;
4145 }
4146 
4147 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4148 {
4149 	nfs_invalidate_atime(hdr->inode);
4150 }
4151 
4152 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4153 {
4154 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4155 
4156 	trace_nfs4_read(hdr, task->tk_status);
4157 	if (nfs4_async_handle_error(task, server,
4158 				    hdr->args.context->state,
4159 				    NULL) == -EAGAIN) {
4160 		rpc_restart_call_prepare(task);
4161 		return -EAGAIN;
4162 	}
4163 
4164 	__nfs4_read_done_cb(hdr);
4165 	if (task->tk_status > 0)
4166 		renew_lease(server, hdr->timestamp);
4167 	return 0;
4168 }
4169 
4170 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4171 		struct nfs_pgio_args *args)
4172 {
4173 
4174 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4175 		nfs4_stateid_is_current(&args->stateid,
4176 				args->context,
4177 				args->lock_context,
4178 				FMODE_READ))
4179 		return false;
4180 	rpc_restart_call_prepare(task);
4181 	return true;
4182 }
4183 
4184 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4185 {
4186 
4187 	dprintk("--> %s\n", __func__);
4188 
4189 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4190 		return -EAGAIN;
4191 	if (nfs4_read_stateid_changed(task, &hdr->args))
4192 		return -EAGAIN;
4193 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4194 				    nfs4_read_done_cb(task, hdr);
4195 }
4196 
4197 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4198 				 struct rpc_message *msg)
4199 {
4200 	hdr->timestamp   = jiffies;
4201 	hdr->pgio_done_cb = nfs4_read_done_cb;
4202 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4203 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4204 }
4205 
4206 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4207 				      struct nfs_pgio_header *hdr)
4208 {
4209 	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4210 			&hdr->args.seq_args,
4211 			&hdr->res.seq_res,
4212 			task))
4213 		return 0;
4214 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4215 				hdr->args.lock_context,
4216 				hdr->rw_ops->rw_mode) == -EIO)
4217 		return -EIO;
4218 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4219 		return -EIO;
4220 	return 0;
4221 }
4222 
4223 static int nfs4_write_done_cb(struct rpc_task *task,
4224 			      struct nfs_pgio_header *hdr)
4225 {
4226 	struct inode *inode = hdr->inode;
4227 
4228 	trace_nfs4_write(hdr, task->tk_status);
4229 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4230 				    hdr->args.context->state,
4231 				    NULL) == -EAGAIN) {
4232 		rpc_restart_call_prepare(task);
4233 		return -EAGAIN;
4234 	}
4235 	if (task->tk_status >= 0) {
4236 		renew_lease(NFS_SERVER(inode), hdr->timestamp);
4237 		nfs_post_op_update_inode_force_wcc(inode, &hdr->fattr);
4238 	}
4239 	return 0;
4240 }
4241 
4242 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4243 		struct nfs_pgio_args *args)
4244 {
4245 
4246 	if (!nfs4_error_stateid_expired(task->tk_status) ||
4247 		nfs4_stateid_is_current(&args->stateid,
4248 				args->context,
4249 				args->lock_context,
4250 				FMODE_WRITE))
4251 		return false;
4252 	rpc_restart_call_prepare(task);
4253 	return true;
4254 }
4255 
4256 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4257 {
4258 	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4259 		return -EAGAIN;
4260 	if (nfs4_write_stateid_changed(task, &hdr->args))
4261 		return -EAGAIN;
4262 	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4263 		nfs4_write_done_cb(task, hdr);
4264 }
4265 
4266 static
4267 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4268 {
4269 	/* Don't request attributes for pNFS or O_DIRECT writes */
4270 	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4271 		return false;
4272 	/* Otherwise, request attributes if and only if we don't hold
4273 	 * a delegation
4274 	 */
4275 	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4276 }
4277 
4278 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4279 				  struct rpc_message *msg)
4280 {
4281 	struct nfs_server *server = NFS_SERVER(hdr->inode);
4282 
4283 	if (!nfs4_write_need_cache_consistency_data(hdr)) {
4284 		hdr->args.bitmask = NULL;
4285 		hdr->res.fattr = NULL;
4286 	} else
4287 		hdr->args.bitmask = server->cache_consistency_bitmask;
4288 
4289 	if (!hdr->pgio_done_cb)
4290 		hdr->pgio_done_cb = nfs4_write_done_cb;
4291 	hdr->res.server = server;
4292 	hdr->timestamp   = jiffies;
4293 
4294 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4295 	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4296 }
4297 
4298 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4299 {
4300 	nfs4_setup_sequence(NFS_SERVER(data->inode),
4301 			&data->args.seq_args,
4302 			&data->res.seq_res,
4303 			task);
4304 }
4305 
4306 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4307 {
4308 	struct inode *inode = data->inode;
4309 
4310 	trace_nfs4_commit(data, task->tk_status);
4311 	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4312 				    NULL, NULL) == -EAGAIN) {
4313 		rpc_restart_call_prepare(task);
4314 		return -EAGAIN;
4315 	}
4316 	return 0;
4317 }
4318 
4319 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4320 {
4321 	if (!nfs4_sequence_done(task, &data->res.seq_res))
4322 		return -EAGAIN;
4323 	return data->commit_done_cb(task, data);
4324 }
4325 
4326 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4327 {
4328 	struct nfs_server *server = NFS_SERVER(data->inode);
4329 
4330 	if (data->commit_done_cb == NULL)
4331 		data->commit_done_cb = nfs4_commit_done_cb;
4332 	data->res.server = server;
4333 	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4334 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4335 }
4336 
4337 struct nfs4_renewdata {
4338 	struct nfs_client	*client;
4339 	unsigned long		timestamp;
4340 };
4341 
4342 /*
4343  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4344  * standalone procedure for queueing an asynchronous RENEW.
4345  */
4346 static void nfs4_renew_release(void *calldata)
4347 {
4348 	struct nfs4_renewdata *data = calldata;
4349 	struct nfs_client *clp = data->client;
4350 
4351 	if (atomic_read(&clp->cl_count) > 1)
4352 		nfs4_schedule_state_renewal(clp);
4353 	nfs_put_client(clp);
4354 	kfree(data);
4355 }
4356 
4357 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4358 {
4359 	struct nfs4_renewdata *data = calldata;
4360 	struct nfs_client *clp = data->client;
4361 	unsigned long timestamp = data->timestamp;
4362 
4363 	trace_nfs4_renew_async(clp, task->tk_status);
4364 	switch (task->tk_status) {
4365 	case 0:
4366 		break;
4367 	case -NFS4ERR_LEASE_MOVED:
4368 		nfs4_schedule_lease_moved_recovery(clp);
4369 		break;
4370 	default:
4371 		/* Unless we're shutting down, schedule state recovery! */
4372 		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4373 			return;
4374 		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4375 			nfs4_schedule_lease_recovery(clp);
4376 			return;
4377 		}
4378 		nfs4_schedule_path_down_recovery(clp);
4379 	}
4380 	do_renew_lease(clp, timestamp);
4381 }
4382 
4383 static const struct rpc_call_ops nfs4_renew_ops = {
4384 	.rpc_call_done = nfs4_renew_done,
4385 	.rpc_release = nfs4_renew_release,
4386 };
4387 
4388 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4389 {
4390 	struct rpc_message msg = {
4391 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4392 		.rpc_argp	= clp,
4393 		.rpc_cred	= cred,
4394 	};
4395 	struct nfs4_renewdata *data;
4396 
4397 	if (renew_flags == 0)
4398 		return 0;
4399 	if (!atomic_inc_not_zero(&clp->cl_count))
4400 		return -EIO;
4401 	data = kmalloc(sizeof(*data), GFP_NOFS);
4402 	if (data == NULL)
4403 		return -ENOMEM;
4404 	data->client = clp;
4405 	data->timestamp = jiffies;
4406 	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4407 			&nfs4_renew_ops, data);
4408 }
4409 
4410 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4411 {
4412 	struct rpc_message msg = {
4413 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4414 		.rpc_argp	= clp,
4415 		.rpc_cred	= cred,
4416 	};
4417 	unsigned long now = jiffies;
4418 	int status;
4419 
4420 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4421 	if (status < 0)
4422 		return status;
4423 	do_renew_lease(clp, now);
4424 	return 0;
4425 }
4426 
4427 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4428 {
4429 	return server->caps & NFS_CAP_ACLS;
4430 }
4431 
4432 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4433  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4434  * the stack.
4435  */
4436 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4437 
4438 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4439 		struct page **pages, unsigned int *pgbase)
4440 {
4441 	struct page *newpage, **spages;
4442 	int rc = 0;
4443 	size_t len;
4444 	spages = pages;
4445 
4446 	do {
4447 		len = min_t(size_t, PAGE_SIZE, buflen);
4448 		newpage = alloc_page(GFP_KERNEL);
4449 
4450 		if (newpage == NULL)
4451 			goto unwind;
4452 		memcpy(page_address(newpage), buf, len);
4453                 buf += len;
4454                 buflen -= len;
4455 		*pages++ = newpage;
4456 		rc++;
4457 	} while (buflen != 0);
4458 
4459 	return rc;
4460 
4461 unwind:
4462 	for(; rc > 0; rc--)
4463 		__free_page(spages[rc-1]);
4464 	return -ENOMEM;
4465 }
4466 
4467 struct nfs4_cached_acl {
4468 	int cached;
4469 	size_t len;
4470 	char data[0];
4471 };
4472 
4473 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4474 {
4475 	struct nfs_inode *nfsi = NFS_I(inode);
4476 
4477 	spin_lock(&inode->i_lock);
4478 	kfree(nfsi->nfs4_acl);
4479 	nfsi->nfs4_acl = acl;
4480 	spin_unlock(&inode->i_lock);
4481 }
4482 
4483 static void nfs4_zap_acl_attr(struct inode *inode)
4484 {
4485 	nfs4_set_cached_acl(inode, NULL);
4486 }
4487 
4488 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4489 {
4490 	struct nfs_inode *nfsi = NFS_I(inode);
4491 	struct nfs4_cached_acl *acl;
4492 	int ret = -ENOENT;
4493 
4494 	spin_lock(&inode->i_lock);
4495 	acl = nfsi->nfs4_acl;
4496 	if (acl == NULL)
4497 		goto out;
4498 	if (buf == NULL) /* user is just asking for length */
4499 		goto out_len;
4500 	if (acl->cached == 0)
4501 		goto out;
4502 	ret = -ERANGE; /* see getxattr(2) man page */
4503 	if (acl->len > buflen)
4504 		goto out;
4505 	memcpy(buf, acl->data, acl->len);
4506 out_len:
4507 	ret = acl->len;
4508 out:
4509 	spin_unlock(&inode->i_lock);
4510 	return ret;
4511 }
4512 
4513 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4514 {
4515 	struct nfs4_cached_acl *acl;
4516 	size_t buflen = sizeof(*acl) + acl_len;
4517 
4518 	if (buflen <= PAGE_SIZE) {
4519 		acl = kmalloc(buflen, GFP_KERNEL);
4520 		if (acl == NULL)
4521 			goto out;
4522 		acl->cached = 1;
4523 		_copy_from_pages(acl->data, pages, pgbase, acl_len);
4524 	} else {
4525 		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4526 		if (acl == NULL)
4527 			goto out;
4528 		acl->cached = 0;
4529 	}
4530 	acl->len = acl_len;
4531 out:
4532 	nfs4_set_cached_acl(inode, acl);
4533 }
4534 
4535 /*
4536  * The getxattr API returns the required buffer length when called with a
4537  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4538  * the required buf.  On a NULL buf, we send a page of data to the server
4539  * guessing that the ACL request can be serviced by a page. If so, we cache
4540  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4541  * the cache. If not so, we throw away the page, and cache the required
4542  * length. The next getxattr call will then produce another round trip to
4543  * the server, this time with the input buf of the required size.
4544  */
4545 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4546 {
4547 	struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4548 	struct nfs_getaclargs args = {
4549 		.fh = NFS_FH(inode),
4550 		.acl_pages = pages,
4551 		.acl_len = buflen,
4552 	};
4553 	struct nfs_getaclres res = {
4554 		.acl_len = buflen,
4555 	};
4556 	struct rpc_message msg = {
4557 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4558 		.rpc_argp = &args,
4559 		.rpc_resp = &res,
4560 	};
4561 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4562 	int ret = -ENOMEM, i;
4563 
4564 	/* As long as we're doing a round trip to the server anyway,
4565 	 * let's be prepared for a page of acl data. */
4566 	if (npages == 0)
4567 		npages = 1;
4568 	if (npages > ARRAY_SIZE(pages))
4569 		return -ERANGE;
4570 
4571 	for (i = 0; i < npages; i++) {
4572 		pages[i] = alloc_page(GFP_KERNEL);
4573 		if (!pages[i])
4574 			goto out_free;
4575 	}
4576 
4577 	/* for decoding across pages */
4578 	res.acl_scratch = alloc_page(GFP_KERNEL);
4579 	if (!res.acl_scratch)
4580 		goto out_free;
4581 
4582 	args.acl_len = npages * PAGE_SIZE;
4583 	args.acl_pgbase = 0;
4584 
4585 	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4586 		__func__, buf, buflen, npages, args.acl_len);
4587 	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4588 			     &msg, &args.seq_args, &res.seq_res, 0);
4589 	if (ret)
4590 		goto out_free;
4591 
4592 	/* Handle the case where the passed-in buffer is too short */
4593 	if (res.acl_flags & NFS4_ACL_TRUNC) {
4594 		/* Did the user only issue a request for the acl length? */
4595 		if (buf == NULL)
4596 			goto out_ok;
4597 		ret = -ERANGE;
4598 		goto out_free;
4599 	}
4600 	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4601 	if (buf) {
4602 		if (res.acl_len > buflen) {
4603 			ret = -ERANGE;
4604 			goto out_free;
4605 		}
4606 		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4607 	}
4608 out_ok:
4609 	ret = res.acl_len;
4610 out_free:
4611 	for (i = 0; i < npages; i++)
4612 		if (pages[i])
4613 			__free_page(pages[i]);
4614 	if (res.acl_scratch)
4615 		__free_page(res.acl_scratch);
4616 	return ret;
4617 }
4618 
4619 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4620 {
4621 	struct nfs4_exception exception = { };
4622 	ssize_t ret;
4623 	do {
4624 		ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4625 		trace_nfs4_get_acl(inode, ret);
4626 		if (ret >= 0)
4627 			break;
4628 		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4629 	} while (exception.retry);
4630 	return ret;
4631 }
4632 
4633 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4634 {
4635 	struct nfs_server *server = NFS_SERVER(inode);
4636 	int ret;
4637 
4638 	if (!nfs4_server_supports_acls(server))
4639 		return -EOPNOTSUPP;
4640 	ret = nfs_revalidate_inode(server, inode);
4641 	if (ret < 0)
4642 		return ret;
4643 	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4644 		nfs_zap_acl_cache(inode);
4645 	ret = nfs4_read_cached_acl(inode, buf, buflen);
4646 	if (ret != -ENOENT)
4647 		/* -ENOENT is returned if there is no ACL or if there is an ACL
4648 		 * but no cached acl data, just the acl length */
4649 		return ret;
4650 	return nfs4_get_acl_uncached(inode, buf, buflen);
4651 }
4652 
4653 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4654 {
4655 	struct nfs_server *server = NFS_SERVER(inode);
4656 	struct page *pages[NFS4ACL_MAXPAGES];
4657 	struct nfs_setaclargs arg = {
4658 		.fh		= NFS_FH(inode),
4659 		.acl_pages	= pages,
4660 		.acl_len	= buflen,
4661 	};
4662 	struct nfs_setaclres res;
4663 	struct rpc_message msg = {
4664 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4665 		.rpc_argp	= &arg,
4666 		.rpc_resp	= &res,
4667 	};
4668 	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4669 	int ret, i;
4670 
4671 	if (!nfs4_server_supports_acls(server))
4672 		return -EOPNOTSUPP;
4673 	if (npages > ARRAY_SIZE(pages))
4674 		return -ERANGE;
4675 	i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4676 	if (i < 0)
4677 		return i;
4678 	nfs4_inode_return_delegation(inode);
4679 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4680 
4681 	/*
4682 	 * Free each page after tx, so the only ref left is
4683 	 * held by the network stack
4684 	 */
4685 	for (; i > 0; i--)
4686 		put_page(pages[i-1]);
4687 
4688 	/*
4689 	 * Acl update can result in inode attribute update.
4690 	 * so mark the attribute cache invalid.
4691 	 */
4692 	spin_lock(&inode->i_lock);
4693 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4694 	spin_unlock(&inode->i_lock);
4695 	nfs_access_zap_cache(inode);
4696 	nfs_zap_acl_cache(inode);
4697 	return ret;
4698 }
4699 
4700 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4701 {
4702 	struct nfs4_exception exception = { };
4703 	int err;
4704 	do {
4705 		err = __nfs4_proc_set_acl(inode, buf, buflen);
4706 		trace_nfs4_set_acl(inode, err);
4707 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4708 				&exception);
4709 	} while (exception.retry);
4710 	return err;
4711 }
4712 
4713 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4714 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4715 					size_t buflen)
4716 {
4717 	struct nfs_server *server = NFS_SERVER(inode);
4718 	struct nfs_fattr fattr;
4719 	struct nfs4_label label = {0, 0, buflen, buf};
4720 
4721 	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4722 	struct nfs4_getattr_arg arg = {
4723 		.fh		= NFS_FH(inode),
4724 		.bitmask	= bitmask,
4725 	};
4726 	struct nfs4_getattr_res res = {
4727 		.fattr		= &fattr,
4728 		.label		= &label,
4729 		.server		= server,
4730 	};
4731 	struct rpc_message msg = {
4732 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4733 		.rpc_argp	= &arg,
4734 		.rpc_resp	= &res,
4735 	};
4736 	int ret;
4737 
4738 	nfs_fattr_init(&fattr);
4739 
4740 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4741 	if (ret)
4742 		return ret;
4743 	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4744 		return -ENOENT;
4745 	if (buflen < label.len)
4746 		return -ERANGE;
4747 	return 0;
4748 }
4749 
4750 static int nfs4_get_security_label(struct inode *inode, void *buf,
4751 					size_t buflen)
4752 {
4753 	struct nfs4_exception exception = { };
4754 	int err;
4755 
4756 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4757 		return -EOPNOTSUPP;
4758 
4759 	do {
4760 		err = _nfs4_get_security_label(inode, buf, buflen);
4761 		trace_nfs4_get_security_label(inode, err);
4762 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4763 				&exception);
4764 	} while (exception.retry);
4765 	return err;
4766 }
4767 
4768 static int _nfs4_do_set_security_label(struct inode *inode,
4769 		struct nfs4_label *ilabel,
4770 		struct nfs_fattr *fattr,
4771 		struct nfs4_label *olabel)
4772 {
4773 
4774 	struct iattr sattr = {0};
4775 	struct nfs_server *server = NFS_SERVER(inode);
4776 	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4777 	struct nfs_setattrargs arg = {
4778 		.fh             = NFS_FH(inode),
4779 		.iap            = &sattr,
4780 		.server		= server,
4781 		.bitmask	= bitmask,
4782 		.label		= ilabel,
4783 	};
4784 	struct nfs_setattrres res = {
4785 		.fattr		= fattr,
4786 		.label		= olabel,
4787 		.server		= server,
4788 	};
4789 	struct rpc_message msg = {
4790 		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4791 		.rpc_argp       = &arg,
4792 		.rpc_resp       = &res,
4793 	};
4794 	int status;
4795 
4796 	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4797 
4798 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4799 	if (status)
4800 		dprintk("%s failed: %d\n", __func__, status);
4801 
4802 	return status;
4803 }
4804 
4805 static int nfs4_do_set_security_label(struct inode *inode,
4806 		struct nfs4_label *ilabel,
4807 		struct nfs_fattr *fattr,
4808 		struct nfs4_label *olabel)
4809 {
4810 	struct nfs4_exception exception = { };
4811 	int err;
4812 
4813 	do {
4814 		err = _nfs4_do_set_security_label(inode, ilabel,
4815 				fattr, olabel);
4816 		trace_nfs4_set_security_label(inode, err);
4817 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4818 				&exception);
4819 	} while (exception.retry);
4820 	return err;
4821 }
4822 
4823 static int
4824 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4825 {
4826 	struct nfs4_label ilabel, *olabel = NULL;
4827 	struct nfs_fattr fattr;
4828 	struct rpc_cred *cred;
4829 	struct inode *inode = dentry->d_inode;
4830 	int status;
4831 
4832 	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4833 		return -EOPNOTSUPP;
4834 
4835 	nfs_fattr_init(&fattr);
4836 
4837 	ilabel.pi = 0;
4838 	ilabel.lfs = 0;
4839 	ilabel.label = (char *)buf;
4840 	ilabel.len = buflen;
4841 
4842 	cred = rpc_lookup_cred();
4843 	if (IS_ERR(cred))
4844 		return PTR_ERR(cred);
4845 
4846 	olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4847 	if (IS_ERR(olabel)) {
4848 		status = -PTR_ERR(olabel);
4849 		goto out;
4850 	}
4851 
4852 	status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4853 	if (status == 0)
4854 		nfs_setsecurity(inode, &fattr, olabel);
4855 
4856 	nfs4_label_free(olabel);
4857 out:
4858 	put_rpccred(cred);
4859 	return status;
4860 }
4861 #endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
4862 
4863 
4864 static int
4865 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4866 			struct nfs4_state *state, long *timeout)
4867 {
4868 	struct nfs_client *clp = server->nfs_client;
4869 
4870 	if (task->tk_status >= 0)
4871 		return 0;
4872 	switch(task->tk_status) {
4873 		case -NFS4ERR_DELEG_REVOKED:
4874 		case -NFS4ERR_ADMIN_REVOKED:
4875 		case -NFS4ERR_BAD_STATEID:
4876 		case -NFS4ERR_OPENMODE:
4877 			if (state == NULL)
4878 				break;
4879 			if (nfs4_schedule_stateid_recovery(server, state) < 0)
4880 				goto recovery_failed;
4881 			goto wait_on_recovery;
4882 		case -NFS4ERR_EXPIRED:
4883 			if (state != NULL) {
4884 				if (nfs4_schedule_stateid_recovery(server, state) < 0)
4885 					goto recovery_failed;
4886 			}
4887 		case -NFS4ERR_STALE_STATEID:
4888 		case -NFS4ERR_STALE_CLIENTID:
4889 			nfs4_schedule_lease_recovery(clp);
4890 			goto wait_on_recovery;
4891 		case -NFS4ERR_MOVED:
4892 			if (nfs4_schedule_migration_recovery(server) < 0)
4893 				goto recovery_failed;
4894 			goto wait_on_recovery;
4895 		case -NFS4ERR_LEASE_MOVED:
4896 			nfs4_schedule_lease_moved_recovery(clp);
4897 			goto wait_on_recovery;
4898 #if defined(CONFIG_NFS_V4_1)
4899 		case -NFS4ERR_BADSESSION:
4900 		case -NFS4ERR_BADSLOT:
4901 		case -NFS4ERR_BAD_HIGH_SLOT:
4902 		case -NFS4ERR_DEADSESSION:
4903 		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4904 		case -NFS4ERR_SEQ_FALSE_RETRY:
4905 		case -NFS4ERR_SEQ_MISORDERED:
4906 			dprintk("%s ERROR %d, Reset session\n", __func__,
4907 				task->tk_status);
4908 			nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4909 			goto wait_on_recovery;
4910 #endif /* CONFIG_NFS_V4_1 */
4911 		case -NFS4ERR_DELAY:
4912 			nfs_inc_server_stats(server, NFSIOS_DELAY);
4913 			rpc_delay(task, nfs4_update_delay(timeout));
4914 			goto restart_call;
4915 		case -NFS4ERR_GRACE:
4916 			rpc_delay(task, NFS4_POLL_RETRY_MAX);
4917 		case -NFS4ERR_RETRY_UNCACHED_REP:
4918 		case -NFS4ERR_OLD_STATEID:
4919 			goto restart_call;
4920 	}
4921 	task->tk_status = nfs4_map_errors(task->tk_status);
4922 	return 0;
4923 recovery_failed:
4924 	task->tk_status = -EIO;
4925 	return 0;
4926 wait_on_recovery:
4927 	rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4928 	if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4929 		rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4930 	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4931 		goto recovery_failed;
4932 restart_call:
4933 	task->tk_status = 0;
4934 	return -EAGAIN;
4935 }
4936 
4937 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4938 				    nfs4_verifier *bootverf)
4939 {
4940 	__be32 verf[2];
4941 
4942 	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4943 		/* An impossible timestamp guarantees this value
4944 		 * will never match a generated boot time. */
4945 		verf[0] = 0;
4946 		verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4947 	} else {
4948 		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4949 		verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4950 		verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4951 	}
4952 	memcpy(bootverf->data, verf, sizeof(bootverf->data));
4953 }
4954 
4955 static unsigned int
4956 nfs4_init_nonuniform_client_string(struct nfs_client *clp,
4957 				   char *buf, size_t len)
4958 {
4959 	unsigned int result;
4960 
4961 	if (clp->cl_owner_id != NULL)
4962 		return strlcpy(buf, clp->cl_owner_id, len);
4963 
4964 	rcu_read_lock();
4965 	result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4966 				clp->cl_ipaddr,
4967 				rpc_peeraddr2str(clp->cl_rpcclient,
4968 							RPC_DISPLAY_ADDR),
4969 				rpc_peeraddr2str(clp->cl_rpcclient,
4970 							RPC_DISPLAY_PROTO));
4971 	rcu_read_unlock();
4972 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4973 	return result;
4974 }
4975 
4976 static unsigned int
4977 nfs4_init_uniform_client_string(struct nfs_client *clp,
4978 				char *buf, size_t len)
4979 {
4980 	const char *nodename = clp->cl_rpcclient->cl_nodename;
4981 	unsigned int result;
4982 
4983 	if (clp->cl_owner_id != NULL)
4984 		return strlcpy(buf, clp->cl_owner_id, len);
4985 
4986 	if (nfs4_client_id_uniquifier[0] != '\0')
4987 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s/%s",
4988 				clp->rpc_ops->version,
4989 				clp->cl_minorversion,
4990 				nfs4_client_id_uniquifier,
4991 				nodename);
4992 	else
4993 		result = scnprintf(buf, len, "Linux NFSv%u.%u %s",
4994 				clp->rpc_ops->version, clp->cl_minorversion,
4995 				nodename);
4996 	clp->cl_owner_id = kstrdup(buf, GFP_KERNEL);
4997 	return result;
4998 }
4999 
5000 /*
5001  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5002  * services.  Advertise one based on the address family of the
5003  * clientaddr.
5004  */
5005 static unsigned int
5006 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5007 {
5008 	if (strchr(clp->cl_ipaddr, ':') != NULL)
5009 		return scnprintf(buf, len, "tcp6");
5010 	else
5011 		return scnprintf(buf, len, "tcp");
5012 }
5013 
5014 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5015 {
5016 	struct nfs4_setclientid *sc = calldata;
5017 
5018 	if (task->tk_status == 0)
5019 		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5020 }
5021 
5022 static const struct rpc_call_ops nfs4_setclientid_ops = {
5023 	.rpc_call_done = nfs4_setclientid_done,
5024 };
5025 
5026 /**
5027  * nfs4_proc_setclientid - Negotiate client ID
5028  * @clp: state data structure
5029  * @program: RPC program for NFSv4 callback service
5030  * @port: IP port number for NFS4 callback service
5031  * @cred: RPC credential to use for this call
5032  * @res: where to place the result
5033  *
5034  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5035  */
5036 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5037 		unsigned short port, struct rpc_cred *cred,
5038 		struct nfs4_setclientid_res *res)
5039 {
5040 	nfs4_verifier sc_verifier;
5041 	struct nfs4_setclientid setclientid = {
5042 		.sc_verifier = &sc_verifier,
5043 		.sc_prog = program,
5044 		.sc_cb_ident = clp->cl_cb_ident,
5045 	};
5046 	struct rpc_message msg = {
5047 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5048 		.rpc_argp = &setclientid,
5049 		.rpc_resp = res,
5050 		.rpc_cred = cred,
5051 	};
5052 	struct rpc_task *task;
5053 	struct rpc_task_setup task_setup_data = {
5054 		.rpc_client = clp->cl_rpcclient,
5055 		.rpc_message = &msg,
5056 		.callback_ops = &nfs4_setclientid_ops,
5057 		.callback_data = &setclientid,
5058 		.flags = RPC_TASK_TIMEOUT,
5059 	};
5060 	int status;
5061 
5062 	/* nfs_client_id4 */
5063 	nfs4_init_boot_verifier(clp, &sc_verifier);
5064 	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5065 		setclientid.sc_name_len =
5066 				nfs4_init_uniform_client_string(clp,
5067 						setclientid.sc_name,
5068 						sizeof(setclientid.sc_name));
5069 	else
5070 		setclientid.sc_name_len =
5071 				nfs4_init_nonuniform_client_string(clp,
5072 						setclientid.sc_name,
5073 						sizeof(setclientid.sc_name));
5074 	/* cb_client4 */
5075 	setclientid.sc_netid_len =
5076 				nfs4_init_callback_netid(clp,
5077 						setclientid.sc_netid,
5078 						sizeof(setclientid.sc_netid));
5079 	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5080 				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5081 				clp->cl_ipaddr, port >> 8, port & 255);
5082 
5083 	dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
5084 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5085 		setclientid.sc_name_len, setclientid.sc_name);
5086 	task = rpc_run_task(&task_setup_data);
5087 	if (IS_ERR(task)) {
5088 		status = PTR_ERR(task);
5089 		goto out;
5090 	}
5091 	status = task->tk_status;
5092 	if (setclientid.sc_cred) {
5093 		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5094 		put_rpccred(setclientid.sc_cred);
5095 	}
5096 	rpc_put_task(task);
5097 out:
5098 	trace_nfs4_setclientid(clp, status);
5099 	dprintk("NFS reply setclientid: %d\n", status);
5100 	return status;
5101 }
5102 
5103 /**
5104  * nfs4_proc_setclientid_confirm - Confirm client ID
5105  * @clp: state data structure
5106  * @res: result of a previous SETCLIENTID
5107  * @cred: RPC credential to use for this call
5108  *
5109  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5110  */
5111 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5112 		struct nfs4_setclientid_res *arg,
5113 		struct rpc_cred *cred)
5114 {
5115 	struct rpc_message msg = {
5116 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5117 		.rpc_argp = arg,
5118 		.rpc_cred = cred,
5119 	};
5120 	int status;
5121 
5122 	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5123 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
5124 		clp->cl_clientid);
5125 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5126 	trace_nfs4_setclientid_confirm(clp, status);
5127 	dprintk("NFS reply setclientid_confirm: %d\n", status);
5128 	return status;
5129 }
5130 
5131 struct nfs4_delegreturndata {
5132 	struct nfs4_delegreturnargs args;
5133 	struct nfs4_delegreturnres res;
5134 	struct nfs_fh fh;
5135 	nfs4_stateid stateid;
5136 	unsigned long timestamp;
5137 	struct nfs_fattr fattr;
5138 	int rpc_status;
5139 	struct inode *inode;
5140 	bool roc;
5141 	u32 roc_barrier;
5142 };
5143 
5144 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5145 {
5146 	struct nfs4_delegreturndata *data = calldata;
5147 
5148 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5149 		return;
5150 
5151 	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5152 	switch (task->tk_status) {
5153 	case 0:
5154 		renew_lease(data->res.server, data->timestamp);
5155 	case -NFS4ERR_ADMIN_REVOKED:
5156 	case -NFS4ERR_DELEG_REVOKED:
5157 	case -NFS4ERR_BAD_STATEID:
5158 	case -NFS4ERR_OLD_STATEID:
5159 	case -NFS4ERR_STALE_STATEID:
5160 	case -NFS4ERR_EXPIRED:
5161 		task->tk_status = 0;
5162 		if (data->roc)
5163 			pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5164 		break;
5165 	default:
5166 		if (nfs4_async_handle_error(task, data->res.server,
5167 					    NULL, NULL) == -EAGAIN) {
5168 			rpc_restart_call_prepare(task);
5169 			return;
5170 		}
5171 	}
5172 	data->rpc_status = task->tk_status;
5173 }
5174 
5175 static void nfs4_delegreturn_release(void *calldata)
5176 {
5177 	struct nfs4_delegreturndata *data = calldata;
5178 	struct inode *inode = data->inode;
5179 
5180 	if (inode) {
5181 		if (data->roc)
5182 			pnfs_roc_release(inode);
5183 		nfs_iput_and_deactive(inode);
5184 	}
5185 	kfree(calldata);
5186 }
5187 
5188 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5189 {
5190 	struct nfs4_delegreturndata *d_data;
5191 
5192 	d_data = (struct nfs4_delegreturndata *)data;
5193 
5194 	if (d_data->roc &&
5195 	    pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5196 		return;
5197 
5198 	nfs4_setup_sequence(d_data->res.server,
5199 			&d_data->args.seq_args,
5200 			&d_data->res.seq_res,
5201 			task);
5202 }
5203 
5204 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5205 	.rpc_call_prepare = nfs4_delegreturn_prepare,
5206 	.rpc_call_done = nfs4_delegreturn_done,
5207 	.rpc_release = nfs4_delegreturn_release,
5208 };
5209 
5210 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5211 {
5212 	struct nfs4_delegreturndata *data;
5213 	struct nfs_server *server = NFS_SERVER(inode);
5214 	struct rpc_task *task;
5215 	struct rpc_message msg = {
5216 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5217 		.rpc_cred = cred,
5218 	};
5219 	struct rpc_task_setup task_setup_data = {
5220 		.rpc_client = server->client,
5221 		.rpc_message = &msg,
5222 		.callback_ops = &nfs4_delegreturn_ops,
5223 		.flags = RPC_TASK_ASYNC,
5224 	};
5225 	int status = 0;
5226 
5227 	data = kzalloc(sizeof(*data), GFP_NOFS);
5228 	if (data == NULL)
5229 		return -ENOMEM;
5230 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5231 	data->args.fhandle = &data->fh;
5232 	data->args.stateid = &data->stateid;
5233 	data->args.bitmask = server->cache_consistency_bitmask;
5234 	nfs_copy_fh(&data->fh, NFS_FH(inode));
5235 	nfs4_stateid_copy(&data->stateid, stateid);
5236 	data->res.fattr = &data->fattr;
5237 	data->res.server = server;
5238 	nfs_fattr_init(data->res.fattr);
5239 	data->timestamp = jiffies;
5240 	data->rpc_status = 0;
5241 	data->inode = nfs_igrab_and_active(inode);
5242 	if (data->inode)
5243 		data->roc = nfs4_roc(inode);
5244 
5245 	task_setup_data.callback_data = data;
5246 	msg.rpc_argp = &data->args;
5247 	msg.rpc_resp = &data->res;
5248 	task = rpc_run_task(&task_setup_data);
5249 	if (IS_ERR(task))
5250 		return PTR_ERR(task);
5251 	if (!issync)
5252 		goto out;
5253 	status = nfs4_wait_for_completion_rpc_task(task);
5254 	if (status != 0)
5255 		goto out;
5256 	status = data->rpc_status;
5257 	if (status == 0)
5258 		nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5259 	else
5260 		nfs_refresh_inode(inode, &data->fattr);
5261 out:
5262 	rpc_put_task(task);
5263 	return status;
5264 }
5265 
5266 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5267 {
5268 	struct nfs_server *server = NFS_SERVER(inode);
5269 	struct nfs4_exception exception = { };
5270 	int err;
5271 	do {
5272 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5273 		trace_nfs4_delegreturn(inode, err);
5274 		switch (err) {
5275 			case -NFS4ERR_STALE_STATEID:
5276 			case -NFS4ERR_EXPIRED:
5277 			case 0:
5278 				return 0;
5279 		}
5280 		err = nfs4_handle_exception(server, err, &exception);
5281 	} while (exception.retry);
5282 	return err;
5283 }
5284 
5285 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5286 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5287 
5288 /*
5289  * sleep, with exponential backoff, and retry the LOCK operation.
5290  */
5291 static unsigned long
5292 nfs4_set_lock_task_retry(unsigned long timeout)
5293 {
5294 	freezable_schedule_timeout_killable_unsafe(timeout);
5295 	timeout <<= 1;
5296 	if (timeout > NFS4_LOCK_MAXTIMEOUT)
5297 		return NFS4_LOCK_MAXTIMEOUT;
5298 	return timeout;
5299 }
5300 
5301 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5302 {
5303 	struct inode *inode = state->inode;
5304 	struct nfs_server *server = NFS_SERVER(inode);
5305 	struct nfs_client *clp = server->nfs_client;
5306 	struct nfs_lockt_args arg = {
5307 		.fh = NFS_FH(inode),
5308 		.fl = request,
5309 	};
5310 	struct nfs_lockt_res res = {
5311 		.denied = request,
5312 	};
5313 	struct rpc_message msg = {
5314 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5315 		.rpc_argp       = &arg,
5316 		.rpc_resp       = &res,
5317 		.rpc_cred	= state->owner->so_cred,
5318 	};
5319 	struct nfs4_lock_state *lsp;
5320 	int status;
5321 
5322 	arg.lock_owner.clientid = clp->cl_clientid;
5323 	status = nfs4_set_lock_state(state, request);
5324 	if (status != 0)
5325 		goto out;
5326 	lsp = request->fl_u.nfs4_fl.owner;
5327 	arg.lock_owner.id = lsp->ls_seqid.owner_id;
5328 	arg.lock_owner.s_dev = server->s_dev;
5329 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5330 	switch (status) {
5331 		case 0:
5332 			request->fl_type = F_UNLCK;
5333 			break;
5334 		case -NFS4ERR_DENIED:
5335 			status = 0;
5336 	}
5337 	request->fl_ops->fl_release_private(request);
5338 	request->fl_ops = NULL;
5339 out:
5340 	return status;
5341 }
5342 
5343 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5344 {
5345 	struct nfs4_exception exception = { };
5346 	int err;
5347 
5348 	do {
5349 		err = _nfs4_proc_getlk(state, cmd, request);
5350 		trace_nfs4_get_lock(request, state, cmd, err);
5351 		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5352 				&exception);
5353 	} while (exception.retry);
5354 	return err;
5355 }
5356 
5357 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5358 {
5359 	int res = 0;
5360 	switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5361 		case FL_POSIX:
5362 			res = posix_lock_file_wait(file, fl);
5363 			break;
5364 		case FL_FLOCK:
5365 			res = flock_lock_file_wait(file, fl);
5366 			break;
5367 		default:
5368 			BUG();
5369 	}
5370 	return res;
5371 }
5372 
5373 struct nfs4_unlockdata {
5374 	struct nfs_locku_args arg;
5375 	struct nfs_locku_res res;
5376 	struct nfs4_lock_state *lsp;
5377 	struct nfs_open_context *ctx;
5378 	struct file_lock fl;
5379 	const struct nfs_server *server;
5380 	unsigned long timestamp;
5381 };
5382 
5383 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5384 		struct nfs_open_context *ctx,
5385 		struct nfs4_lock_state *lsp,
5386 		struct nfs_seqid *seqid)
5387 {
5388 	struct nfs4_unlockdata *p;
5389 	struct inode *inode = lsp->ls_state->inode;
5390 
5391 	p = kzalloc(sizeof(*p), GFP_NOFS);
5392 	if (p == NULL)
5393 		return NULL;
5394 	p->arg.fh = NFS_FH(inode);
5395 	p->arg.fl = &p->fl;
5396 	p->arg.seqid = seqid;
5397 	p->res.seqid = seqid;
5398 	p->lsp = lsp;
5399 	atomic_inc(&lsp->ls_count);
5400 	/* Ensure we don't close file until we're done freeing locks! */
5401 	p->ctx = get_nfs_open_context(ctx);
5402 	memcpy(&p->fl, fl, sizeof(p->fl));
5403 	p->server = NFS_SERVER(inode);
5404 	return p;
5405 }
5406 
5407 static void nfs4_locku_release_calldata(void *data)
5408 {
5409 	struct nfs4_unlockdata *calldata = data;
5410 	nfs_free_seqid(calldata->arg.seqid);
5411 	nfs4_put_lock_state(calldata->lsp);
5412 	put_nfs_open_context(calldata->ctx);
5413 	kfree(calldata);
5414 }
5415 
5416 static void nfs4_locku_done(struct rpc_task *task, void *data)
5417 {
5418 	struct nfs4_unlockdata *calldata = data;
5419 
5420 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5421 		return;
5422 	switch (task->tk_status) {
5423 		case 0:
5424 			renew_lease(calldata->server, calldata->timestamp);
5425 			do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5426 			if (nfs4_update_lock_stateid(calldata->lsp,
5427 					&calldata->res.stateid))
5428 				break;
5429 		case -NFS4ERR_BAD_STATEID:
5430 		case -NFS4ERR_OLD_STATEID:
5431 		case -NFS4ERR_STALE_STATEID:
5432 		case -NFS4ERR_EXPIRED:
5433 			if (!nfs4_stateid_match(&calldata->arg.stateid,
5434 						&calldata->lsp->ls_stateid))
5435 				rpc_restart_call_prepare(task);
5436 			break;
5437 		default:
5438 			if (nfs4_async_handle_error(task, calldata->server,
5439 						    NULL, NULL) == -EAGAIN)
5440 				rpc_restart_call_prepare(task);
5441 	}
5442 	nfs_release_seqid(calldata->arg.seqid);
5443 }
5444 
5445 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5446 {
5447 	struct nfs4_unlockdata *calldata = data;
5448 
5449 	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5450 		goto out_wait;
5451 	nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5452 	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5453 		/* Note: exit _without_ running nfs4_locku_done */
5454 		goto out_no_action;
5455 	}
5456 	calldata->timestamp = jiffies;
5457 	if (nfs4_setup_sequence(calldata->server,
5458 				&calldata->arg.seq_args,
5459 				&calldata->res.seq_res,
5460 				task) != 0)
5461 		nfs_release_seqid(calldata->arg.seqid);
5462 	return;
5463 out_no_action:
5464 	task->tk_action = NULL;
5465 out_wait:
5466 	nfs4_sequence_done(task, &calldata->res.seq_res);
5467 }
5468 
5469 static const struct rpc_call_ops nfs4_locku_ops = {
5470 	.rpc_call_prepare = nfs4_locku_prepare,
5471 	.rpc_call_done = nfs4_locku_done,
5472 	.rpc_release = nfs4_locku_release_calldata,
5473 };
5474 
5475 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5476 		struct nfs_open_context *ctx,
5477 		struct nfs4_lock_state *lsp,
5478 		struct nfs_seqid *seqid)
5479 {
5480 	struct nfs4_unlockdata *data;
5481 	struct rpc_message msg = {
5482 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5483 		.rpc_cred = ctx->cred,
5484 	};
5485 	struct rpc_task_setup task_setup_data = {
5486 		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5487 		.rpc_message = &msg,
5488 		.callback_ops = &nfs4_locku_ops,
5489 		.workqueue = nfsiod_workqueue,
5490 		.flags = RPC_TASK_ASYNC,
5491 	};
5492 
5493 	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5494 		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5495 
5496 	/* Ensure this is an unlock - when canceling a lock, the
5497 	 * canceled lock is passed in, and it won't be an unlock.
5498 	 */
5499 	fl->fl_type = F_UNLCK;
5500 
5501 	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5502 	if (data == NULL) {
5503 		nfs_free_seqid(seqid);
5504 		return ERR_PTR(-ENOMEM);
5505 	}
5506 
5507 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5508 	msg.rpc_argp = &data->arg;
5509 	msg.rpc_resp = &data->res;
5510 	task_setup_data.callback_data = data;
5511 	return rpc_run_task(&task_setup_data);
5512 }
5513 
5514 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5515 {
5516 	struct inode *inode = state->inode;
5517 	struct nfs4_state_owner *sp = state->owner;
5518 	struct nfs_inode *nfsi = NFS_I(inode);
5519 	struct nfs_seqid *seqid;
5520 	struct nfs4_lock_state *lsp;
5521 	struct rpc_task *task;
5522 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5523 	int status = 0;
5524 	unsigned char fl_flags = request->fl_flags;
5525 
5526 	status = nfs4_set_lock_state(state, request);
5527 	/* Unlock _before_ we do the RPC call */
5528 	request->fl_flags |= FL_EXISTS;
5529 	/* Exclude nfs_delegation_claim_locks() */
5530 	mutex_lock(&sp->so_delegreturn_mutex);
5531 	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5532 	down_read(&nfsi->rwsem);
5533 	if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5534 		up_read(&nfsi->rwsem);
5535 		mutex_unlock(&sp->so_delegreturn_mutex);
5536 		goto out;
5537 	}
5538 	up_read(&nfsi->rwsem);
5539 	mutex_unlock(&sp->so_delegreturn_mutex);
5540 	if (status != 0)
5541 		goto out;
5542 	/* Is this a delegated lock? */
5543 	lsp = request->fl_u.nfs4_fl.owner;
5544 	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5545 		goto out;
5546 	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5547 	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5548 	status = -ENOMEM;
5549 	if (IS_ERR(seqid))
5550 		goto out;
5551 	task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5552 	status = PTR_ERR(task);
5553 	if (IS_ERR(task))
5554 		goto out;
5555 	status = nfs4_wait_for_completion_rpc_task(task);
5556 	rpc_put_task(task);
5557 out:
5558 	request->fl_flags = fl_flags;
5559 	trace_nfs4_unlock(request, state, F_SETLK, status);
5560 	return status;
5561 }
5562 
5563 struct nfs4_lockdata {
5564 	struct nfs_lock_args arg;
5565 	struct nfs_lock_res res;
5566 	struct nfs4_lock_state *lsp;
5567 	struct nfs_open_context *ctx;
5568 	struct file_lock fl;
5569 	unsigned long timestamp;
5570 	int rpc_status;
5571 	int cancelled;
5572 	struct nfs_server *server;
5573 };
5574 
5575 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5576 		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5577 		gfp_t gfp_mask)
5578 {
5579 	struct nfs4_lockdata *p;
5580 	struct inode *inode = lsp->ls_state->inode;
5581 	struct nfs_server *server = NFS_SERVER(inode);
5582 	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5583 
5584 	p = kzalloc(sizeof(*p), gfp_mask);
5585 	if (p == NULL)
5586 		return NULL;
5587 
5588 	p->arg.fh = NFS_FH(inode);
5589 	p->arg.fl = &p->fl;
5590 	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5591 	if (IS_ERR(p->arg.open_seqid))
5592 		goto out_free;
5593 	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5594 	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5595 	if (IS_ERR(p->arg.lock_seqid))
5596 		goto out_free_seqid;
5597 	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5598 	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5599 	p->arg.lock_owner.s_dev = server->s_dev;
5600 	p->res.lock_seqid = p->arg.lock_seqid;
5601 	p->lsp = lsp;
5602 	p->server = server;
5603 	atomic_inc(&lsp->ls_count);
5604 	p->ctx = get_nfs_open_context(ctx);
5605 	memcpy(&p->fl, fl, sizeof(p->fl));
5606 	return p;
5607 out_free_seqid:
5608 	nfs_free_seqid(p->arg.open_seqid);
5609 out_free:
5610 	kfree(p);
5611 	return NULL;
5612 }
5613 
5614 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5615 {
5616 	struct nfs4_lockdata *data = calldata;
5617 	struct nfs4_state *state = data->lsp->ls_state;
5618 
5619 	dprintk("%s: begin!\n", __func__);
5620 	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5621 		goto out_wait;
5622 	/* Do we need to do an open_to_lock_owner? */
5623 	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5624 		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5625 			goto out_release_lock_seqid;
5626 		}
5627 		nfs4_stateid_copy(&data->arg.open_stateid,
5628 				&state->open_stateid);
5629 		data->arg.new_lock_owner = 1;
5630 		data->res.open_seqid = data->arg.open_seqid;
5631 	} else {
5632 		data->arg.new_lock_owner = 0;
5633 		nfs4_stateid_copy(&data->arg.lock_stateid,
5634 				&data->lsp->ls_stateid);
5635 	}
5636 	if (!nfs4_valid_open_stateid(state)) {
5637 		data->rpc_status = -EBADF;
5638 		task->tk_action = NULL;
5639 		goto out_release_open_seqid;
5640 	}
5641 	data->timestamp = jiffies;
5642 	if (nfs4_setup_sequence(data->server,
5643 				&data->arg.seq_args,
5644 				&data->res.seq_res,
5645 				task) == 0)
5646 		return;
5647 out_release_open_seqid:
5648 	nfs_release_seqid(data->arg.open_seqid);
5649 out_release_lock_seqid:
5650 	nfs_release_seqid(data->arg.lock_seqid);
5651 out_wait:
5652 	nfs4_sequence_done(task, &data->res.seq_res);
5653 	dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5654 }
5655 
5656 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5657 {
5658 	struct nfs4_lockdata *data = calldata;
5659 	struct nfs4_lock_state *lsp = data->lsp;
5660 
5661 	dprintk("%s: begin!\n", __func__);
5662 
5663 	if (!nfs4_sequence_done(task, &data->res.seq_res))
5664 		return;
5665 
5666 	data->rpc_status = task->tk_status;
5667 	switch (task->tk_status) {
5668 	case 0:
5669 		renew_lease(NFS_SERVER(data->ctx->dentry->d_inode),
5670 				data->timestamp);
5671 		if (data->arg.new_lock) {
5672 			data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5673 			if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5674 				rpc_restart_call_prepare(task);
5675 				break;
5676 			}
5677 		}
5678 		if (data->arg.new_lock_owner != 0) {
5679 			nfs_confirm_seqid(&lsp->ls_seqid, 0);
5680 			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5681 			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5682 		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5683 			rpc_restart_call_prepare(task);
5684 		break;
5685 	case -NFS4ERR_BAD_STATEID:
5686 	case -NFS4ERR_OLD_STATEID:
5687 	case -NFS4ERR_STALE_STATEID:
5688 	case -NFS4ERR_EXPIRED:
5689 		if (data->arg.new_lock_owner != 0) {
5690 			if (!nfs4_stateid_match(&data->arg.open_stateid,
5691 						&lsp->ls_state->open_stateid))
5692 				rpc_restart_call_prepare(task);
5693 		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5694 						&lsp->ls_stateid))
5695 				rpc_restart_call_prepare(task);
5696 	}
5697 	dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5698 }
5699 
5700 static void nfs4_lock_release(void *calldata)
5701 {
5702 	struct nfs4_lockdata *data = calldata;
5703 
5704 	dprintk("%s: begin!\n", __func__);
5705 	nfs_free_seqid(data->arg.open_seqid);
5706 	if (data->cancelled != 0) {
5707 		struct rpc_task *task;
5708 		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5709 				data->arg.lock_seqid);
5710 		if (!IS_ERR(task))
5711 			rpc_put_task_async(task);
5712 		dprintk("%s: cancelling lock!\n", __func__);
5713 	} else
5714 		nfs_free_seqid(data->arg.lock_seqid);
5715 	nfs4_put_lock_state(data->lsp);
5716 	put_nfs_open_context(data->ctx);
5717 	kfree(data);
5718 	dprintk("%s: done!\n", __func__);
5719 }
5720 
5721 static const struct rpc_call_ops nfs4_lock_ops = {
5722 	.rpc_call_prepare = nfs4_lock_prepare,
5723 	.rpc_call_done = nfs4_lock_done,
5724 	.rpc_release = nfs4_lock_release,
5725 };
5726 
5727 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5728 {
5729 	switch (error) {
5730 	case -NFS4ERR_ADMIN_REVOKED:
5731 	case -NFS4ERR_BAD_STATEID:
5732 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5733 		if (new_lock_owner != 0 ||
5734 		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5735 			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5736 		break;
5737 	case -NFS4ERR_STALE_STATEID:
5738 		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5739 	case -NFS4ERR_EXPIRED:
5740 		nfs4_schedule_lease_recovery(server->nfs_client);
5741 	};
5742 }
5743 
5744 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5745 {
5746 	struct nfs4_lockdata *data;
5747 	struct rpc_task *task;
5748 	struct rpc_message msg = {
5749 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5750 		.rpc_cred = state->owner->so_cred,
5751 	};
5752 	struct rpc_task_setup task_setup_data = {
5753 		.rpc_client = NFS_CLIENT(state->inode),
5754 		.rpc_message = &msg,
5755 		.callback_ops = &nfs4_lock_ops,
5756 		.workqueue = nfsiod_workqueue,
5757 		.flags = RPC_TASK_ASYNC,
5758 	};
5759 	int ret;
5760 
5761 	dprintk("%s: begin!\n", __func__);
5762 	data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5763 			fl->fl_u.nfs4_fl.owner,
5764 			recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5765 	if (data == NULL)
5766 		return -ENOMEM;
5767 	if (IS_SETLKW(cmd))
5768 		data->arg.block = 1;
5769 	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5770 	msg.rpc_argp = &data->arg;
5771 	msg.rpc_resp = &data->res;
5772 	task_setup_data.callback_data = data;
5773 	if (recovery_type > NFS_LOCK_NEW) {
5774 		if (recovery_type == NFS_LOCK_RECLAIM)
5775 			data->arg.reclaim = NFS_LOCK_RECLAIM;
5776 		nfs4_set_sequence_privileged(&data->arg.seq_args);
5777 	} else
5778 		data->arg.new_lock = 1;
5779 	task = rpc_run_task(&task_setup_data);
5780 	if (IS_ERR(task))
5781 		return PTR_ERR(task);
5782 	ret = nfs4_wait_for_completion_rpc_task(task);
5783 	if (ret == 0) {
5784 		ret = data->rpc_status;
5785 		if (ret)
5786 			nfs4_handle_setlk_error(data->server, data->lsp,
5787 					data->arg.new_lock_owner, ret);
5788 	} else
5789 		data->cancelled = 1;
5790 	rpc_put_task(task);
5791 	dprintk("%s: done, ret = %d!\n", __func__, ret);
5792 	return ret;
5793 }
5794 
5795 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5796 {
5797 	struct nfs_server *server = NFS_SERVER(state->inode);
5798 	struct nfs4_exception exception = {
5799 		.inode = state->inode,
5800 	};
5801 	int err;
5802 
5803 	do {
5804 		/* Cache the lock if possible... */
5805 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5806 			return 0;
5807 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5808 		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5809 		if (err != -NFS4ERR_DELAY)
5810 			break;
5811 		nfs4_handle_exception(server, err, &exception);
5812 	} while (exception.retry);
5813 	return err;
5814 }
5815 
5816 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5817 {
5818 	struct nfs_server *server = NFS_SERVER(state->inode);
5819 	struct nfs4_exception exception = {
5820 		.inode = state->inode,
5821 	};
5822 	int err;
5823 
5824 	err = nfs4_set_lock_state(state, request);
5825 	if (err != 0)
5826 		return err;
5827 	if (!recover_lost_locks) {
5828 		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5829 		return 0;
5830 	}
5831 	do {
5832 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5833 			return 0;
5834 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5835 		trace_nfs4_lock_expired(request, state, F_SETLK, err);
5836 		switch (err) {
5837 		default:
5838 			goto out;
5839 		case -NFS4ERR_GRACE:
5840 		case -NFS4ERR_DELAY:
5841 			nfs4_handle_exception(server, err, &exception);
5842 			err = 0;
5843 		}
5844 	} while (exception.retry);
5845 out:
5846 	return err;
5847 }
5848 
5849 #if defined(CONFIG_NFS_V4_1)
5850 /**
5851  * nfs41_check_expired_locks - possibly free a lock stateid
5852  *
5853  * @state: NFSv4 state for an inode
5854  *
5855  * Returns NFS_OK if recovery for this stateid is now finished.
5856  * Otherwise a negative NFS4ERR value is returned.
5857  */
5858 static int nfs41_check_expired_locks(struct nfs4_state *state)
5859 {
5860 	int status, ret = -NFS4ERR_BAD_STATEID;
5861 	struct nfs4_lock_state *lsp;
5862 	struct nfs_server *server = NFS_SERVER(state->inode);
5863 
5864 	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5865 		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5866 			struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5867 
5868 			status = nfs41_test_stateid(server,
5869 					&lsp->ls_stateid,
5870 					cred);
5871 			trace_nfs4_test_lock_stateid(state, lsp, status);
5872 			if (status != NFS_OK) {
5873 				/* Free the stateid unless the server
5874 				 * informs us the stateid is unrecognized. */
5875 				if (status != -NFS4ERR_BAD_STATEID)
5876 					nfs41_free_stateid(server,
5877 							&lsp->ls_stateid,
5878 							cred);
5879 				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5880 				ret = status;
5881 			}
5882 		}
5883 	};
5884 
5885 	return ret;
5886 }
5887 
5888 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5889 {
5890 	int status = NFS_OK;
5891 
5892 	if (test_bit(LK_STATE_IN_USE, &state->flags))
5893 		status = nfs41_check_expired_locks(state);
5894 	if (status != NFS_OK)
5895 		status = nfs4_lock_expired(state, request);
5896 	return status;
5897 }
5898 #endif
5899 
5900 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5901 {
5902 	struct nfs_inode *nfsi = NFS_I(state->inode);
5903 	unsigned char fl_flags = request->fl_flags;
5904 	int status = -ENOLCK;
5905 
5906 	if ((fl_flags & FL_POSIX) &&
5907 			!test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
5908 		goto out;
5909 	/* Is this a delegated open? */
5910 	status = nfs4_set_lock_state(state, request);
5911 	if (status != 0)
5912 		goto out;
5913 	request->fl_flags |= FL_ACCESS;
5914 	status = do_vfs_lock(request->fl_file, request);
5915 	if (status < 0)
5916 		goto out;
5917 	down_read(&nfsi->rwsem);
5918 	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
5919 		/* Yes: cache locks! */
5920 		/* ...but avoid races with delegation recall... */
5921 		request->fl_flags = fl_flags & ~FL_SLEEP;
5922 		status = do_vfs_lock(request->fl_file, request);
5923 		up_read(&nfsi->rwsem);
5924 		goto out;
5925 	}
5926 	up_read(&nfsi->rwsem);
5927 	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
5928 out:
5929 	request->fl_flags = fl_flags;
5930 	return status;
5931 }
5932 
5933 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5934 {
5935 	struct nfs4_exception exception = {
5936 		.state = state,
5937 		.inode = state->inode,
5938 	};
5939 	int err;
5940 
5941 	do {
5942 		err = _nfs4_proc_setlk(state, cmd, request);
5943 		trace_nfs4_set_lock(request, state, cmd, err);
5944 		if (err == -NFS4ERR_DENIED)
5945 			err = -EAGAIN;
5946 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
5947 				err, &exception);
5948 	} while (exception.retry);
5949 	return err;
5950 }
5951 
5952 static int
5953 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
5954 {
5955 	struct nfs_open_context *ctx;
5956 	struct nfs4_state *state;
5957 	unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
5958 	int status;
5959 
5960 	/* verify open state */
5961 	ctx = nfs_file_open_context(filp);
5962 	state = ctx->state;
5963 
5964 	if (request->fl_start < 0 || request->fl_end < 0)
5965 		return -EINVAL;
5966 
5967 	if (IS_GETLK(cmd)) {
5968 		if (state != NULL)
5969 			return nfs4_proc_getlk(state, F_GETLK, request);
5970 		return 0;
5971 	}
5972 
5973 	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
5974 		return -EINVAL;
5975 
5976 	if (request->fl_type == F_UNLCK) {
5977 		if (state != NULL)
5978 			return nfs4_proc_unlck(state, cmd, request);
5979 		return 0;
5980 	}
5981 
5982 	if (state == NULL)
5983 		return -ENOLCK;
5984 	/*
5985 	 * Don't rely on the VFS having checked the file open mode,
5986 	 * since it won't do this for flock() locks.
5987 	 */
5988 	switch (request->fl_type) {
5989 	case F_RDLCK:
5990 		if (!(filp->f_mode & FMODE_READ))
5991 			return -EBADF;
5992 		break;
5993 	case F_WRLCK:
5994 		if (!(filp->f_mode & FMODE_WRITE))
5995 			return -EBADF;
5996 	}
5997 
5998 	do {
5999 		status = nfs4_proc_setlk(state, cmd, request);
6000 		if ((status != -EAGAIN) || IS_SETLK(cmd))
6001 			break;
6002 		timeout = nfs4_set_lock_task_retry(timeout);
6003 		status = -ERESTARTSYS;
6004 		if (signalled())
6005 			break;
6006 	} while(status < 0);
6007 	return status;
6008 }
6009 
6010 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6011 {
6012 	struct nfs_server *server = NFS_SERVER(state->inode);
6013 	int err;
6014 
6015 	err = nfs4_set_lock_state(state, fl);
6016 	if (err != 0)
6017 		return err;
6018 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6019 	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6020 }
6021 
6022 struct nfs_release_lockowner_data {
6023 	struct nfs4_lock_state *lsp;
6024 	struct nfs_server *server;
6025 	struct nfs_release_lockowner_args args;
6026 	struct nfs_release_lockowner_res res;
6027 	unsigned long timestamp;
6028 };
6029 
6030 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6031 {
6032 	struct nfs_release_lockowner_data *data = calldata;
6033 	struct nfs_server *server = data->server;
6034 	nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6035 			     &data->args.seq_args, &data->res.seq_res, task);
6036 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6037 	data->timestamp = jiffies;
6038 }
6039 
6040 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6041 {
6042 	struct nfs_release_lockowner_data *data = calldata;
6043 	struct nfs_server *server = data->server;
6044 
6045 	nfs40_sequence_done(task, &data->res.seq_res);
6046 
6047 	switch (task->tk_status) {
6048 	case 0:
6049 		renew_lease(server, data->timestamp);
6050 		break;
6051 	case -NFS4ERR_STALE_CLIENTID:
6052 	case -NFS4ERR_EXPIRED:
6053 		nfs4_schedule_lease_recovery(server->nfs_client);
6054 		break;
6055 	case -NFS4ERR_LEASE_MOVED:
6056 	case -NFS4ERR_DELAY:
6057 		if (nfs4_async_handle_error(task, server,
6058 					    NULL, NULL) == -EAGAIN)
6059 			rpc_restart_call_prepare(task);
6060 	}
6061 }
6062 
6063 static void nfs4_release_lockowner_release(void *calldata)
6064 {
6065 	struct nfs_release_lockowner_data *data = calldata;
6066 	nfs4_free_lock_state(data->server, data->lsp);
6067 	kfree(calldata);
6068 }
6069 
6070 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6071 	.rpc_call_prepare = nfs4_release_lockowner_prepare,
6072 	.rpc_call_done = nfs4_release_lockowner_done,
6073 	.rpc_release = nfs4_release_lockowner_release,
6074 };
6075 
6076 static void
6077 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6078 {
6079 	struct nfs_release_lockowner_data *data;
6080 	struct rpc_message msg = {
6081 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6082 	};
6083 
6084 	if (server->nfs_client->cl_mvops->minor_version != 0)
6085 		return;
6086 
6087 	data = kmalloc(sizeof(*data), GFP_NOFS);
6088 	if (!data)
6089 		return;
6090 	data->lsp = lsp;
6091 	data->server = server;
6092 	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6093 	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6094 	data->args.lock_owner.s_dev = server->s_dev;
6095 
6096 	msg.rpc_argp = &data->args;
6097 	msg.rpc_resp = &data->res;
6098 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6099 	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6100 }
6101 
6102 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6103 
6104 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6105 				   const void *buf, size_t buflen,
6106 				   int flags, int type)
6107 {
6108 	if (strcmp(key, "") != 0)
6109 		return -EINVAL;
6110 
6111 	return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
6112 }
6113 
6114 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6115 				   void *buf, size_t buflen, int type)
6116 {
6117 	if (strcmp(key, "") != 0)
6118 		return -EINVAL;
6119 
6120 	return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
6121 }
6122 
6123 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6124 				       size_t list_len, const char *name,
6125 				       size_t name_len, int type)
6126 {
6127 	size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6128 
6129 	if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
6130 		return 0;
6131 
6132 	if (list && len <= list_len)
6133 		memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6134 	return len;
6135 }
6136 
6137 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6138 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6139 {
6140 	return server->caps & NFS_CAP_SECURITY_LABEL;
6141 }
6142 
6143 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6144 				   const void *buf, size_t buflen,
6145 				   int flags, int type)
6146 {
6147 	if (security_ismaclabel(key))
6148 		return nfs4_set_security_label(dentry, buf, buflen);
6149 
6150 	return -EOPNOTSUPP;
6151 }
6152 
6153 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6154 				   void *buf, size_t buflen, int type)
6155 {
6156 	if (security_ismaclabel(key))
6157 		return nfs4_get_security_label(dentry->d_inode, buf, buflen);
6158 	return -EOPNOTSUPP;
6159 }
6160 
6161 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6162 				       size_t list_len, const char *name,
6163 				       size_t name_len, int type)
6164 {
6165 	size_t len = 0;
6166 
6167 	if (nfs_server_capable(dentry->d_inode, NFS_CAP_SECURITY_LABEL)) {
6168 		len = security_inode_listsecurity(dentry->d_inode, NULL, 0);
6169 		if (list && len <= list_len)
6170 			security_inode_listsecurity(dentry->d_inode, list, len);
6171 	}
6172 	return len;
6173 }
6174 
6175 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6176 	.prefix = XATTR_SECURITY_PREFIX,
6177 	.list	= nfs4_xattr_list_nfs4_label,
6178 	.get	= nfs4_xattr_get_nfs4_label,
6179 	.set	= nfs4_xattr_set_nfs4_label,
6180 };
6181 #endif
6182 
6183 
6184 /*
6185  * nfs_fhget will use either the mounted_on_fileid or the fileid
6186  */
6187 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6188 {
6189 	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6190 	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6191 	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6192 	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6193 		return;
6194 
6195 	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6196 		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6197 	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6198 	fattr->nlink = 2;
6199 }
6200 
6201 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6202 				   const struct qstr *name,
6203 				   struct nfs4_fs_locations *fs_locations,
6204 				   struct page *page)
6205 {
6206 	struct nfs_server *server = NFS_SERVER(dir);
6207 	u32 bitmask[3] = {
6208 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6209 	};
6210 	struct nfs4_fs_locations_arg args = {
6211 		.dir_fh = NFS_FH(dir),
6212 		.name = name,
6213 		.page = page,
6214 		.bitmask = bitmask,
6215 	};
6216 	struct nfs4_fs_locations_res res = {
6217 		.fs_locations = fs_locations,
6218 	};
6219 	struct rpc_message msg = {
6220 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6221 		.rpc_argp = &args,
6222 		.rpc_resp = &res,
6223 	};
6224 	int status;
6225 
6226 	dprintk("%s: start\n", __func__);
6227 
6228 	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
6229 	 * is not supported */
6230 	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6231 		bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6232 	else
6233 		bitmask[0] |= FATTR4_WORD0_FILEID;
6234 
6235 	nfs_fattr_init(&fs_locations->fattr);
6236 	fs_locations->server = server;
6237 	fs_locations->nlocations = 0;
6238 	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6239 	dprintk("%s: returned status = %d\n", __func__, status);
6240 	return status;
6241 }
6242 
6243 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6244 			   const struct qstr *name,
6245 			   struct nfs4_fs_locations *fs_locations,
6246 			   struct page *page)
6247 {
6248 	struct nfs4_exception exception = { };
6249 	int err;
6250 	do {
6251 		err = _nfs4_proc_fs_locations(client, dir, name,
6252 				fs_locations, page);
6253 		trace_nfs4_get_fs_locations(dir, name, err);
6254 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6255 				&exception);
6256 	} while (exception.retry);
6257 	return err;
6258 }
6259 
6260 /*
6261  * This operation also signals the server that this client is
6262  * performing migration recovery.  The server can stop returning
6263  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6264  * appended to this compound to identify the client ID which is
6265  * performing recovery.
6266  */
6267 static int _nfs40_proc_get_locations(struct inode *inode,
6268 				     struct nfs4_fs_locations *locations,
6269 				     struct page *page, struct rpc_cred *cred)
6270 {
6271 	struct nfs_server *server = NFS_SERVER(inode);
6272 	struct rpc_clnt *clnt = server->client;
6273 	u32 bitmask[2] = {
6274 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6275 	};
6276 	struct nfs4_fs_locations_arg args = {
6277 		.clientid	= server->nfs_client->cl_clientid,
6278 		.fh		= NFS_FH(inode),
6279 		.page		= page,
6280 		.bitmask	= bitmask,
6281 		.migration	= 1,		/* skip LOOKUP */
6282 		.renew		= 1,		/* append RENEW */
6283 	};
6284 	struct nfs4_fs_locations_res res = {
6285 		.fs_locations	= locations,
6286 		.migration	= 1,
6287 		.renew		= 1,
6288 	};
6289 	struct rpc_message msg = {
6290 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6291 		.rpc_argp	= &args,
6292 		.rpc_resp	= &res,
6293 		.rpc_cred	= cred,
6294 	};
6295 	unsigned long now = jiffies;
6296 	int status;
6297 
6298 	nfs_fattr_init(&locations->fattr);
6299 	locations->server = server;
6300 	locations->nlocations = 0;
6301 
6302 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6303 	nfs4_set_sequence_privileged(&args.seq_args);
6304 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6305 					&args.seq_args, &res.seq_res);
6306 	if (status)
6307 		return status;
6308 
6309 	renew_lease(server, now);
6310 	return 0;
6311 }
6312 
6313 #ifdef CONFIG_NFS_V4_1
6314 
6315 /*
6316  * This operation also signals the server that this client is
6317  * performing migration recovery.  The server can stop asserting
6318  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6319  * performing this operation is identified in the SEQUENCE
6320  * operation in this compound.
6321  *
6322  * When the client supports GETATTR(fs_locations_info), it can
6323  * be plumbed in here.
6324  */
6325 static int _nfs41_proc_get_locations(struct inode *inode,
6326 				     struct nfs4_fs_locations *locations,
6327 				     struct page *page, struct rpc_cred *cred)
6328 {
6329 	struct nfs_server *server = NFS_SERVER(inode);
6330 	struct rpc_clnt *clnt = server->client;
6331 	u32 bitmask[2] = {
6332 		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6333 	};
6334 	struct nfs4_fs_locations_arg args = {
6335 		.fh		= NFS_FH(inode),
6336 		.page		= page,
6337 		.bitmask	= bitmask,
6338 		.migration	= 1,		/* skip LOOKUP */
6339 	};
6340 	struct nfs4_fs_locations_res res = {
6341 		.fs_locations	= locations,
6342 		.migration	= 1,
6343 	};
6344 	struct rpc_message msg = {
6345 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6346 		.rpc_argp	= &args,
6347 		.rpc_resp	= &res,
6348 		.rpc_cred	= cred,
6349 	};
6350 	int status;
6351 
6352 	nfs_fattr_init(&locations->fattr);
6353 	locations->server = server;
6354 	locations->nlocations = 0;
6355 
6356 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6357 	nfs4_set_sequence_privileged(&args.seq_args);
6358 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6359 					&args.seq_args, &res.seq_res);
6360 	if (status == NFS4_OK &&
6361 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6362 		status = -NFS4ERR_LEASE_MOVED;
6363 	return status;
6364 }
6365 
6366 #endif	/* CONFIG_NFS_V4_1 */
6367 
6368 /**
6369  * nfs4_proc_get_locations - discover locations for a migrated FSID
6370  * @inode: inode on FSID that is migrating
6371  * @locations: result of query
6372  * @page: buffer
6373  * @cred: credential to use for this operation
6374  *
6375  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6376  * operation failed, or a negative errno if a local error occurred.
6377  *
6378  * On success, "locations" is filled in, but if the server has
6379  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6380  * asserted.
6381  *
6382  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6383  * from this client that require migration recovery.
6384  */
6385 int nfs4_proc_get_locations(struct inode *inode,
6386 			    struct nfs4_fs_locations *locations,
6387 			    struct page *page, struct rpc_cred *cred)
6388 {
6389 	struct nfs_server *server = NFS_SERVER(inode);
6390 	struct nfs_client *clp = server->nfs_client;
6391 	const struct nfs4_mig_recovery_ops *ops =
6392 					clp->cl_mvops->mig_recovery_ops;
6393 	struct nfs4_exception exception = { };
6394 	int status;
6395 
6396 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6397 		(unsigned long long)server->fsid.major,
6398 		(unsigned long long)server->fsid.minor,
6399 		clp->cl_hostname);
6400 	nfs_display_fhandle(NFS_FH(inode), __func__);
6401 
6402 	do {
6403 		status = ops->get_locations(inode, locations, page, cred);
6404 		if (status != -NFS4ERR_DELAY)
6405 			break;
6406 		nfs4_handle_exception(server, status, &exception);
6407 	} while (exception.retry);
6408 	return status;
6409 }
6410 
6411 /*
6412  * This operation also signals the server that this client is
6413  * performing "lease moved" recovery.  The server can stop
6414  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6415  * is appended to this compound to identify the client ID which is
6416  * performing recovery.
6417  */
6418 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6419 {
6420 	struct nfs_server *server = NFS_SERVER(inode);
6421 	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6422 	struct rpc_clnt *clnt = server->client;
6423 	struct nfs4_fsid_present_arg args = {
6424 		.fh		= NFS_FH(inode),
6425 		.clientid	= clp->cl_clientid,
6426 		.renew		= 1,		/* append RENEW */
6427 	};
6428 	struct nfs4_fsid_present_res res = {
6429 		.renew		= 1,
6430 	};
6431 	struct rpc_message msg = {
6432 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6433 		.rpc_argp	= &args,
6434 		.rpc_resp	= &res,
6435 		.rpc_cred	= cred,
6436 	};
6437 	unsigned long now = jiffies;
6438 	int status;
6439 
6440 	res.fh = nfs_alloc_fhandle();
6441 	if (res.fh == NULL)
6442 		return -ENOMEM;
6443 
6444 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6445 	nfs4_set_sequence_privileged(&args.seq_args);
6446 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6447 						&args.seq_args, &res.seq_res);
6448 	nfs_free_fhandle(res.fh);
6449 	if (status)
6450 		return status;
6451 
6452 	do_renew_lease(clp, now);
6453 	return 0;
6454 }
6455 
6456 #ifdef CONFIG_NFS_V4_1
6457 
6458 /*
6459  * This operation also signals the server that this client is
6460  * performing "lease moved" recovery.  The server can stop asserting
6461  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6462  * this operation is identified in the SEQUENCE operation in this
6463  * compound.
6464  */
6465 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6466 {
6467 	struct nfs_server *server = NFS_SERVER(inode);
6468 	struct rpc_clnt *clnt = server->client;
6469 	struct nfs4_fsid_present_arg args = {
6470 		.fh		= NFS_FH(inode),
6471 	};
6472 	struct nfs4_fsid_present_res res = {
6473 	};
6474 	struct rpc_message msg = {
6475 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6476 		.rpc_argp	= &args,
6477 		.rpc_resp	= &res,
6478 		.rpc_cred	= cred,
6479 	};
6480 	int status;
6481 
6482 	res.fh = nfs_alloc_fhandle();
6483 	if (res.fh == NULL)
6484 		return -ENOMEM;
6485 
6486 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6487 	nfs4_set_sequence_privileged(&args.seq_args);
6488 	status = nfs4_call_sync_sequence(clnt, server, &msg,
6489 						&args.seq_args, &res.seq_res);
6490 	nfs_free_fhandle(res.fh);
6491 	if (status == NFS4_OK &&
6492 	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6493 		status = -NFS4ERR_LEASE_MOVED;
6494 	return status;
6495 }
6496 
6497 #endif	/* CONFIG_NFS_V4_1 */
6498 
6499 /**
6500  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6501  * @inode: inode on FSID to check
6502  * @cred: credential to use for this operation
6503  *
6504  * Server indicates whether the FSID is present, moved, or not
6505  * recognized.  This operation is necessary to clear a LEASE_MOVED
6506  * condition for this client ID.
6507  *
6508  * Returns NFS4_OK if the FSID is present on this server,
6509  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6510  *  NFS4ERR code if some error occurred on the server, or a
6511  *  negative errno if a local failure occurred.
6512  */
6513 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6514 {
6515 	struct nfs_server *server = NFS_SERVER(inode);
6516 	struct nfs_client *clp = server->nfs_client;
6517 	const struct nfs4_mig_recovery_ops *ops =
6518 					clp->cl_mvops->mig_recovery_ops;
6519 	struct nfs4_exception exception = { };
6520 	int status;
6521 
6522 	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6523 		(unsigned long long)server->fsid.major,
6524 		(unsigned long long)server->fsid.minor,
6525 		clp->cl_hostname);
6526 	nfs_display_fhandle(NFS_FH(inode), __func__);
6527 
6528 	do {
6529 		status = ops->fsid_present(inode, cred);
6530 		if (status != -NFS4ERR_DELAY)
6531 			break;
6532 		nfs4_handle_exception(server, status, &exception);
6533 	} while (exception.retry);
6534 	return status;
6535 }
6536 
6537 /**
6538  * If 'use_integrity' is true and the state managment nfs_client
6539  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6540  * and the machine credential as per RFC3530bis and RFC5661 Security
6541  * Considerations sections. Otherwise, just use the user cred with the
6542  * filesystem's rpc_client.
6543  */
6544 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6545 {
6546 	int status;
6547 	struct nfs4_secinfo_arg args = {
6548 		.dir_fh = NFS_FH(dir),
6549 		.name   = name,
6550 	};
6551 	struct nfs4_secinfo_res res = {
6552 		.flavors     = flavors,
6553 	};
6554 	struct rpc_message msg = {
6555 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6556 		.rpc_argp = &args,
6557 		.rpc_resp = &res,
6558 	};
6559 	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6560 	struct rpc_cred *cred = NULL;
6561 
6562 	if (use_integrity) {
6563 		clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6564 		cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6565 		msg.rpc_cred = cred;
6566 	}
6567 
6568 	dprintk("NFS call  secinfo %s\n", name->name);
6569 
6570 	nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6571 		NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6572 
6573 	status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6574 				&res.seq_res, 0);
6575 	dprintk("NFS reply  secinfo: %d\n", status);
6576 
6577 	if (cred)
6578 		put_rpccred(cred);
6579 
6580 	return status;
6581 }
6582 
6583 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6584 		      struct nfs4_secinfo_flavors *flavors)
6585 {
6586 	struct nfs4_exception exception = { };
6587 	int err;
6588 	do {
6589 		err = -NFS4ERR_WRONGSEC;
6590 
6591 		/* try to use integrity protection with machine cred */
6592 		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6593 			err = _nfs4_proc_secinfo(dir, name, flavors, true);
6594 
6595 		/*
6596 		 * if unable to use integrity protection, or SECINFO with
6597 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
6598 		 * disallowed by spec, but exists in deployed servers) use
6599 		 * the current filesystem's rpc_client and the user cred.
6600 		 */
6601 		if (err == -NFS4ERR_WRONGSEC)
6602 			err = _nfs4_proc_secinfo(dir, name, flavors, false);
6603 
6604 		trace_nfs4_secinfo(dir, name, err);
6605 		err = nfs4_handle_exception(NFS_SERVER(dir), err,
6606 				&exception);
6607 	} while (exception.retry);
6608 	return err;
6609 }
6610 
6611 #ifdef CONFIG_NFS_V4_1
6612 /*
6613  * Check the exchange flags returned by the server for invalid flags, having
6614  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6615  * DS flags set.
6616  */
6617 static int nfs4_check_cl_exchange_flags(u32 flags)
6618 {
6619 	if (flags & ~EXCHGID4_FLAG_MASK_R)
6620 		goto out_inval;
6621 	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6622 	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6623 		goto out_inval;
6624 	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6625 		goto out_inval;
6626 	return NFS_OK;
6627 out_inval:
6628 	return -NFS4ERR_INVAL;
6629 }
6630 
6631 static bool
6632 nfs41_same_server_scope(struct nfs41_server_scope *a,
6633 			struct nfs41_server_scope *b)
6634 {
6635 	if (a->server_scope_sz == b->server_scope_sz &&
6636 	    memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6637 		return true;
6638 
6639 	return false;
6640 }
6641 
6642 /*
6643  * nfs4_proc_bind_conn_to_session()
6644  *
6645  * The 4.1 client currently uses the same TCP connection for the
6646  * fore and backchannel.
6647  */
6648 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6649 {
6650 	int status;
6651 	struct nfs41_bind_conn_to_session_args args = {
6652 		.client = clp,
6653 		.dir = NFS4_CDFC4_FORE_OR_BOTH,
6654 	};
6655 	struct nfs41_bind_conn_to_session_res res;
6656 	struct rpc_message msg = {
6657 		.rpc_proc =
6658 			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6659 		.rpc_argp = &args,
6660 		.rpc_resp = &res,
6661 		.rpc_cred = cred,
6662 	};
6663 
6664 	dprintk("--> %s\n", __func__);
6665 
6666 	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6667 	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6668 		args.dir = NFS4_CDFC4_FORE;
6669 
6670 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6671 	trace_nfs4_bind_conn_to_session(clp, status);
6672 	if (status == 0) {
6673 		if (memcmp(res.sessionid.data,
6674 		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6675 			dprintk("NFS: %s: Session ID mismatch\n", __func__);
6676 			status = -EIO;
6677 			goto out;
6678 		}
6679 		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6680 			dprintk("NFS: %s: Unexpected direction from server\n",
6681 				__func__);
6682 			status = -EIO;
6683 			goto out;
6684 		}
6685 		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6686 			dprintk("NFS: %s: Server returned RDMA mode = true\n",
6687 				__func__);
6688 			status = -EIO;
6689 			goto out;
6690 		}
6691 	}
6692 out:
6693 	dprintk("<-- %s status= %d\n", __func__, status);
6694 	return status;
6695 }
6696 
6697 /*
6698  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6699  * and operations we'd like to see to enable certain features in the allow map
6700  */
6701 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6702 	.how = SP4_MACH_CRED,
6703 	.enforce.u.words = {
6704 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6705 		      1 << (OP_EXCHANGE_ID - 32) |
6706 		      1 << (OP_CREATE_SESSION - 32) |
6707 		      1 << (OP_DESTROY_SESSION - 32) |
6708 		      1 << (OP_DESTROY_CLIENTID - 32)
6709 	},
6710 	.allow.u.words = {
6711 		[0] = 1 << (OP_CLOSE) |
6712 		      1 << (OP_LOCKU) |
6713 		      1 << (OP_COMMIT),
6714 		[1] = 1 << (OP_SECINFO - 32) |
6715 		      1 << (OP_SECINFO_NO_NAME - 32) |
6716 		      1 << (OP_TEST_STATEID - 32) |
6717 		      1 << (OP_FREE_STATEID - 32) |
6718 		      1 << (OP_WRITE - 32)
6719 	}
6720 };
6721 
6722 /*
6723  * Select the state protection mode for client `clp' given the server results
6724  * from exchange_id in `sp'.
6725  *
6726  * Returns 0 on success, negative errno otherwise.
6727  */
6728 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6729 				 struct nfs41_state_protection *sp)
6730 {
6731 	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6732 		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6733 		      1 << (OP_EXCHANGE_ID - 32) |
6734 		      1 << (OP_CREATE_SESSION - 32) |
6735 		      1 << (OP_DESTROY_SESSION - 32) |
6736 		      1 << (OP_DESTROY_CLIENTID - 32)
6737 	};
6738 	unsigned int i;
6739 
6740 	if (sp->how == SP4_MACH_CRED) {
6741 		/* Print state protect result */
6742 		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6743 		for (i = 0; i <= LAST_NFS4_OP; i++) {
6744 			if (test_bit(i, sp->enforce.u.longs))
6745 				dfprintk(MOUNT, "  enforce op %d\n", i);
6746 			if (test_bit(i, sp->allow.u.longs))
6747 				dfprintk(MOUNT, "  allow op %d\n", i);
6748 		}
6749 
6750 		/* make sure nothing is on enforce list that isn't supported */
6751 		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6752 			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6753 				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6754 				return -EINVAL;
6755 			}
6756 		}
6757 
6758 		/*
6759 		 * Minimal mode - state operations are allowed to use machine
6760 		 * credential.  Note this already happens by default, so the
6761 		 * client doesn't have to do anything more than the negotiation.
6762 		 *
6763 		 * NOTE: we don't care if EXCHANGE_ID is in the list -
6764 		 *       we're already using the machine cred for exchange_id
6765 		 *       and will never use a different cred.
6766 		 */
6767 		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6768 		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6769 		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6770 		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6771 			dfprintk(MOUNT, "sp4_mach_cred:\n");
6772 			dfprintk(MOUNT, "  minimal mode enabled\n");
6773 			set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6774 		} else {
6775 			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6776 			return -EINVAL;
6777 		}
6778 
6779 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6780 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
6781 			dfprintk(MOUNT, "  cleanup mode enabled\n");
6782 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6783 		}
6784 
6785 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6786 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6787 			dfprintk(MOUNT, "  secinfo mode enabled\n");
6788 			set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6789 		}
6790 
6791 		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6792 		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6793 			dfprintk(MOUNT, "  stateid mode enabled\n");
6794 			set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6795 		}
6796 
6797 		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6798 			dfprintk(MOUNT, "  write mode enabled\n");
6799 			set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6800 		}
6801 
6802 		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6803 			dfprintk(MOUNT, "  commit mode enabled\n");
6804 			set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6805 		}
6806 	}
6807 
6808 	return 0;
6809 }
6810 
6811 /*
6812  * _nfs4_proc_exchange_id()
6813  *
6814  * Wrapper for EXCHANGE_ID operation.
6815  */
6816 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6817 	u32 sp4_how)
6818 {
6819 	nfs4_verifier verifier;
6820 	struct nfs41_exchange_id_args args = {
6821 		.verifier = &verifier,
6822 		.client = clp,
6823 #ifdef CONFIG_NFS_V4_1_MIGRATION
6824 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6825 			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
6826 			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6827 #else
6828 		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6829 			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
6830 #endif
6831 	};
6832 	struct nfs41_exchange_id_res res = {
6833 		0
6834 	};
6835 	int status;
6836 	struct rpc_message msg = {
6837 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6838 		.rpc_argp = &args,
6839 		.rpc_resp = &res,
6840 		.rpc_cred = cred,
6841 	};
6842 
6843 	nfs4_init_boot_verifier(clp, &verifier);
6844 	args.id_len = nfs4_init_uniform_client_string(clp, args.id,
6845 							sizeof(args.id));
6846 	dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
6847 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6848 		args.id_len, args.id);
6849 
6850 	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6851 					GFP_NOFS);
6852 	if (unlikely(res.server_owner == NULL)) {
6853 		status = -ENOMEM;
6854 		goto out;
6855 	}
6856 
6857 	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6858 					GFP_NOFS);
6859 	if (unlikely(res.server_scope == NULL)) {
6860 		status = -ENOMEM;
6861 		goto out_server_owner;
6862 	}
6863 
6864 	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6865 	if (unlikely(res.impl_id == NULL)) {
6866 		status = -ENOMEM;
6867 		goto out_server_scope;
6868 	}
6869 
6870 	switch (sp4_how) {
6871 	case SP4_NONE:
6872 		args.state_protect.how = SP4_NONE;
6873 		break;
6874 
6875 	case SP4_MACH_CRED:
6876 		args.state_protect = nfs4_sp4_mach_cred_request;
6877 		break;
6878 
6879 	default:
6880 		/* unsupported! */
6881 		WARN_ON_ONCE(1);
6882 		status = -EINVAL;
6883 		goto out_server_scope;
6884 	}
6885 
6886 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6887 	trace_nfs4_exchange_id(clp, status);
6888 	if (status == 0)
6889 		status = nfs4_check_cl_exchange_flags(res.flags);
6890 
6891 	if (status == 0)
6892 		status = nfs4_sp4_select_mode(clp, &res.state_protect);
6893 
6894 	if (status == 0) {
6895 		clp->cl_clientid = res.clientid;
6896 		clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
6897 		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
6898 			clp->cl_seqid = res.seqid;
6899 
6900 		kfree(clp->cl_serverowner);
6901 		clp->cl_serverowner = res.server_owner;
6902 		res.server_owner = NULL;
6903 
6904 		/* use the most recent implementation id */
6905 		kfree(clp->cl_implid);
6906 		clp->cl_implid = res.impl_id;
6907 
6908 		if (clp->cl_serverscope != NULL &&
6909 		    !nfs41_same_server_scope(clp->cl_serverscope,
6910 					     res.server_scope)) {
6911 			dprintk("%s: server_scope mismatch detected\n",
6912 				__func__);
6913 			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
6914 			kfree(clp->cl_serverscope);
6915 			clp->cl_serverscope = NULL;
6916 		}
6917 
6918 		if (clp->cl_serverscope == NULL) {
6919 			clp->cl_serverscope = res.server_scope;
6920 			goto out;
6921 		}
6922 	} else
6923 		kfree(res.impl_id);
6924 
6925 out_server_owner:
6926 	kfree(res.server_owner);
6927 out_server_scope:
6928 	kfree(res.server_scope);
6929 out:
6930 	if (clp->cl_implid != NULL)
6931 		dprintk("NFS reply exchange_id: Server Implementation ID: "
6932 			"domain: %s, name: %s, date: %llu,%u\n",
6933 			clp->cl_implid->domain, clp->cl_implid->name,
6934 			clp->cl_implid->date.seconds,
6935 			clp->cl_implid->date.nseconds);
6936 	dprintk("NFS reply exchange_id: %d\n", status);
6937 	return status;
6938 }
6939 
6940 /*
6941  * nfs4_proc_exchange_id()
6942  *
6943  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6944  *
6945  * Since the clientid has expired, all compounds using sessions
6946  * associated with the stale clientid will be returning
6947  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
6948  * be in some phase of session reset.
6949  *
6950  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
6951  */
6952 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
6953 {
6954 	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
6955 	int status;
6956 
6957 	/* try SP4_MACH_CRED if krb5i/p	*/
6958 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
6959 	    authflavor == RPC_AUTH_GSS_KRB5P) {
6960 		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
6961 		if (!status)
6962 			return 0;
6963 	}
6964 
6965 	/* try SP4_NONE */
6966 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
6967 }
6968 
6969 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
6970 		struct rpc_cred *cred)
6971 {
6972 	struct rpc_message msg = {
6973 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
6974 		.rpc_argp = clp,
6975 		.rpc_cred = cred,
6976 	};
6977 	int status;
6978 
6979 	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6980 	trace_nfs4_destroy_clientid(clp, status);
6981 	if (status)
6982 		dprintk("NFS: Got error %d from the server %s on "
6983 			"DESTROY_CLIENTID.", status, clp->cl_hostname);
6984 	return status;
6985 }
6986 
6987 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
6988 		struct rpc_cred *cred)
6989 {
6990 	unsigned int loop;
6991 	int ret;
6992 
6993 	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
6994 		ret = _nfs4_proc_destroy_clientid(clp, cred);
6995 		switch (ret) {
6996 		case -NFS4ERR_DELAY:
6997 		case -NFS4ERR_CLIENTID_BUSY:
6998 			ssleep(1);
6999 			break;
7000 		default:
7001 			return ret;
7002 		}
7003 	}
7004 	return 0;
7005 }
7006 
7007 int nfs4_destroy_clientid(struct nfs_client *clp)
7008 {
7009 	struct rpc_cred *cred;
7010 	int ret = 0;
7011 
7012 	if (clp->cl_mvops->minor_version < 1)
7013 		goto out;
7014 	if (clp->cl_exchange_flags == 0)
7015 		goto out;
7016 	if (clp->cl_preserve_clid)
7017 		goto out;
7018 	cred = nfs4_get_clid_cred(clp);
7019 	ret = nfs4_proc_destroy_clientid(clp, cred);
7020 	if (cred)
7021 		put_rpccred(cred);
7022 	switch (ret) {
7023 	case 0:
7024 	case -NFS4ERR_STALE_CLIENTID:
7025 		clp->cl_exchange_flags = 0;
7026 	}
7027 out:
7028 	return ret;
7029 }
7030 
7031 struct nfs4_get_lease_time_data {
7032 	struct nfs4_get_lease_time_args *args;
7033 	struct nfs4_get_lease_time_res *res;
7034 	struct nfs_client *clp;
7035 };
7036 
7037 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7038 					void *calldata)
7039 {
7040 	struct nfs4_get_lease_time_data *data =
7041 			(struct nfs4_get_lease_time_data *)calldata;
7042 
7043 	dprintk("--> %s\n", __func__);
7044 	/* just setup sequence, do not trigger session recovery
7045 	   since we're invoked within one */
7046 	nfs41_setup_sequence(data->clp->cl_session,
7047 			&data->args->la_seq_args,
7048 			&data->res->lr_seq_res,
7049 			task);
7050 	dprintk("<-- %s\n", __func__);
7051 }
7052 
7053 /*
7054  * Called from nfs4_state_manager thread for session setup, so don't recover
7055  * from sequence operation or clientid errors.
7056  */
7057 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7058 {
7059 	struct nfs4_get_lease_time_data *data =
7060 			(struct nfs4_get_lease_time_data *)calldata;
7061 
7062 	dprintk("--> %s\n", __func__);
7063 	if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7064 		return;
7065 	switch (task->tk_status) {
7066 	case -NFS4ERR_DELAY:
7067 	case -NFS4ERR_GRACE:
7068 		dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7069 		rpc_delay(task, NFS4_POLL_RETRY_MIN);
7070 		task->tk_status = 0;
7071 		/* fall through */
7072 	case -NFS4ERR_RETRY_UNCACHED_REP:
7073 		rpc_restart_call_prepare(task);
7074 		return;
7075 	}
7076 	dprintk("<-- %s\n", __func__);
7077 }
7078 
7079 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7080 	.rpc_call_prepare = nfs4_get_lease_time_prepare,
7081 	.rpc_call_done = nfs4_get_lease_time_done,
7082 };
7083 
7084 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7085 {
7086 	struct rpc_task *task;
7087 	struct nfs4_get_lease_time_args args;
7088 	struct nfs4_get_lease_time_res res = {
7089 		.lr_fsinfo = fsinfo,
7090 	};
7091 	struct nfs4_get_lease_time_data data = {
7092 		.args = &args,
7093 		.res = &res,
7094 		.clp = clp,
7095 	};
7096 	struct rpc_message msg = {
7097 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7098 		.rpc_argp = &args,
7099 		.rpc_resp = &res,
7100 	};
7101 	struct rpc_task_setup task_setup = {
7102 		.rpc_client = clp->cl_rpcclient,
7103 		.rpc_message = &msg,
7104 		.callback_ops = &nfs4_get_lease_time_ops,
7105 		.callback_data = &data,
7106 		.flags = RPC_TASK_TIMEOUT,
7107 	};
7108 	int status;
7109 
7110 	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7111 	nfs4_set_sequence_privileged(&args.la_seq_args);
7112 	dprintk("--> %s\n", __func__);
7113 	task = rpc_run_task(&task_setup);
7114 
7115 	if (IS_ERR(task))
7116 		status = PTR_ERR(task);
7117 	else {
7118 		status = task->tk_status;
7119 		rpc_put_task(task);
7120 	}
7121 	dprintk("<-- %s return %d\n", __func__, status);
7122 
7123 	return status;
7124 }
7125 
7126 /*
7127  * Initialize the values to be used by the client in CREATE_SESSION
7128  * If nfs4_init_session set the fore channel request and response sizes,
7129  * use them.
7130  *
7131  * Set the back channel max_resp_sz_cached to zero to force the client to
7132  * always set csa_cachethis to FALSE because the current implementation
7133  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7134  */
7135 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7136 {
7137 	unsigned int max_rqst_sz, max_resp_sz;
7138 
7139 	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7140 	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7141 
7142 	/* Fore channel attributes */
7143 	args->fc_attrs.max_rqst_sz = max_rqst_sz;
7144 	args->fc_attrs.max_resp_sz = max_resp_sz;
7145 	args->fc_attrs.max_ops = NFS4_MAX_OPS;
7146 	args->fc_attrs.max_reqs = max_session_slots;
7147 
7148 	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7149 		"max_ops=%u max_reqs=%u\n",
7150 		__func__,
7151 		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7152 		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7153 
7154 	/* Back channel attributes */
7155 	args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7156 	args->bc_attrs.max_resp_sz = PAGE_SIZE;
7157 	args->bc_attrs.max_resp_sz_cached = 0;
7158 	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7159 	args->bc_attrs.max_reqs = 1;
7160 
7161 	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7162 		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7163 		__func__,
7164 		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7165 		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7166 		args->bc_attrs.max_reqs);
7167 }
7168 
7169 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7170 		struct nfs41_create_session_res *res)
7171 {
7172 	struct nfs4_channel_attrs *sent = &args->fc_attrs;
7173 	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7174 
7175 	if (rcvd->max_resp_sz > sent->max_resp_sz)
7176 		return -EINVAL;
7177 	/*
7178 	 * Our requested max_ops is the minimum we need; we're not
7179 	 * prepared to break up compounds into smaller pieces than that.
7180 	 * So, no point even trying to continue if the server won't
7181 	 * cooperate:
7182 	 */
7183 	if (rcvd->max_ops < sent->max_ops)
7184 		return -EINVAL;
7185 	if (rcvd->max_reqs == 0)
7186 		return -EINVAL;
7187 	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7188 		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7189 	return 0;
7190 }
7191 
7192 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7193 		struct nfs41_create_session_res *res)
7194 {
7195 	struct nfs4_channel_attrs *sent = &args->bc_attrs;
7196 	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7197 
7198 	if (!(res->flags & SESSION4_BACK_CHAN))
7199 		goto out;
7200 	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7201 		return -EINVAL;
7202 	if (rcvd->max_resp_sz < sent->max_resp_sz)
7203 		return -EINVAL;
7204 	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7205 		return -EINVAL;
7206 	/* These would render the backchannel useless: */
7207 	if (rcvd->max_ops != sent->max_ops)
7208 		return -EINVAL;
7209 	if (rcvd->max_reqs != sent->max_reqs)
7210 		return -EINVAL;
7211 out:
7212 	return 0;
7213 }
7214 
7215 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7216 				     struct nfs41_create_session_res *res)
7217 {
7218 	int ret;
7219 
7220 	ret = nfs4_verify_fore_channel_attrs(args, res);
7221 	if (ret)
7222 		return ret;
7223 	return nfs4_verify_back_channel_attrs(args, res);
7224 }
7225 
7226 static void nfs4_update_session(struct nfs4_session *session,
7227 		struct nfs41_create_session_res *res)
7228 {
7229 	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7230 	session->flags = res->flags;
7231 	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7232 	if (res->flags & SESSION4_BACK_CHAN)
7233 		memcpy(&session->bc_attrs, &res->bc_attrs,
7234 				sizeof(session->bc_attrs));
7235 }
7236 
7237 static int _nfs4_proc_create_session(struct nfs_client *clp,
7238 		struct rpc_cred *cred)
7239 {
7240 	struct nfs4_session *session = clp->cl_session;
7241 	struct nfs41_create_session_args args = {
7242 		.client = clp,
7243 		.clientid = clp->cl_clientid,
7244 		.seqid = clp->cl_seqid,
7245 		.cb_program = NFS4_CALLBACK,
7246 	};
7247 	struct nfs41_create_session_res res;
7248 
7249 	struct rpc_message msg = {
7250 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7251 		.rpc_argp = &args,
7252 		.rpc_resp = &res,
7253 		.rpc_cred = cred,
7254 	};
7255 	int status;
7256 
7257 	nfs4_init_channel_attrs(&args);
7258 	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7259 
7260 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7261 	trace_nfs4_create_session(clp, status);
7262 
7263 	if (!status) {
7264 		/* Verify the session's negotiated channel_attrs values */
7265 		status = nfs4_verify_channel_attrs(&args, &res);
7266 		/* Increment the clientid slot sequence id */
7267 		if (clp->cl_seqid == res.seqid)
7268 			clp->cl_seqid++;
7269 		if (status)
7270 			goto out;
7271 		nfs4_update_session(session, &res);
7272 	}
7273 out:
7274 	return status;
7275 }
7276 
7277 /*
7278  * Issues a CREATE_SESSION operation to the server.
7279  * It is the responsibility of the caller to verify the session is
7280  * expired before calling this routine.
7281  */
7282 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7283 {
7284 	int status;
7285 	unsigned *ptr;
7286 	struct nfs4_session *session = clp->cl_session;
7287 
7288 	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7289 
7290 	status = _nfs4_proc_create_session(clp, cred);
7291 	if (status)
7292 		goto out;
7293 
7294 	/* Init or reset the session slot tables */
7295 	status = nfs4_setup_session_slot_tables(session);
7296 	dprintk("slot table setup returned %d\n", status);
7297 	if (status)
7298 		goto out;
7299 
7300 	ptr = (unsigned *)&session->sess_id.data[0];
7301 	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7302 		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7303 out:
7304 	dprintk("<-- %s\n", __func__);
7305 	return status;
7306 }
7307 
7308 /*
7309  * Issue the over-the-wire RPC DESTROY_SESSION.
7310  * The caller must serialize access to this routine.
7311  */
7312 int nfs4_proc_destroy_session(struct nfs4_session *session,
7313 		struct rpc_cred *cred)
7314 {
7315 	struct rpc_message msg = {
7316 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7317 		.rpc_argp = session,
7318 		.rpc_cred = cred,
7319 	};
7320 	int status = 0;
7321 
7322 	dprintk("--> nfs4_proc_destroy_session\n");
7323 
7324 	/* session is still being setup */
7325 	if (session->clp->cl_cons_state != NFS_CS_READY)
7326 		return status;
7327 
7328 	status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7329 	trace_nfs4_destroy_session(session->clp, status);
7330 
7331 	if (status)
7332 		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7333 			"Session has been destroyed regardless...\n", status);
7334 
7335 	dprintk("<-- nfs4_proc_destroy_session\n");
7336 	return status;
7337 }
7338 
7339 /*
7340  * Renew the cl_session lease.
7341  */
7342 struct nfs4_sequence_data {
7343 	struct nfs_client *clp;
7344 	struct nfs4_sequence_args args;
7345 	struct nfs4_sequence_res res;
7346 };
7347 
7348 static void nfs41_sequence_release(void *data)
7349 {
7350 	struct nfs4_sequence_data *calldata = data;
7351 	struct nfs_client *clp = calldata->clp;
7352 
7353 	if (atomic_read(&clp->cl_count) > 1)
7354 		nfs4_schedule_state_renewal(clp);
7355 	nfs_put_client(clp);
7356 	kfree(calldata);
7357 }
7358 
7359 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7360 {
7361 	switch(task->tk_status) {
7362 	case -NFS4ERR_DELAY:
7363 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7364 		return -EAGAIN;
7365 	default:
7366 		nfs4_schedule_lease_recovery(clp);
7367 	}
7368 	return 0;
7369 }
7370 
7371 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7372 {
7373 	struct nfs4_sequence_data *calldata = data;
7374 	struct nfs_client *clp = calldata->clp;
7375 
7376 	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7377 		return;
7378 
7379 	trace_nfs4_sequence(clp, task->tk_status);
7380 	if (task->tk_status < 0) {
7381 		dprintk("%s ERROR %d\n", __func__, task->tk_status);
7382 		if (atomic_read(&clp->cl_count) == 1)
7383 			goto out;
7384 
7385 		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7386 			rpc_restart_call_prepare(task);
7387 			return;
7388 		}
7389 	}
7390 	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7391 out:
7392 	dprintk("<-- %s\n", __func__);
7393 }
7394 
7395 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7396 {
7397 	struct nfs4_sequence_data *calldata = data;
7398 	struct nfs_client *clp = calldata->clp;
7399 	struct nfs4_sequence_args *args;
7400 	struct nfs4_sequence_res *res;
7401 
7402 	args = task->tk_msg.rpc_argp;
7403 	res = task->tk_msg.rpc_resp;
7404 
7405 	nfs41_setup_sequence(clp->cl_session, args, res, task);
7406 }
7407 
7408 static const struct rpc_call_ops nfs41_sequence_ops = {
7409 	.rpc_call_done = nfs41_sequence_call_done,
7410 	.rpc_call_prepare = nfs41_sequence_prepare,
7411 	.rpc_release = nfs41_sequence_release,
7412 };
7413 
7414 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7415 		struct rpc_cred *cred,
7416 		bool is_privileged)
7417 {
7418 	struct nfs4_sequence_data *calldata;
7419 	struct rpc_message msg = {
7420 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7421 		.rpc_cred = cred,
7422 	};
7423 	struct rpc_task_setup task_setup_data = {
7424 		.rpc_client = clp->cl_rpcclient,
7425 		.rpc_message = &msg,
7426 		.callback_ops = &nfs41_sequence_ops,
7427 		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7428 	};
7429 
7430 	if (!atomic_inc_not_zero(&clp->cl_count))
7431 		return ERR_PTR(-EIO);
7432 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7433 	if (calldata == NULL) {
7434 		nfs_put_client(clp);
7435 		return ERR_PTR(-ENOMEM);
7436 	}
7437 	nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7438 	if (is_privileged)
7439 		nfs4_set_sequence_privileged(&calldata->args);
7440 	msg.rpc_argp = &calldata->args;
7441 	msg.rpc_resp = &calldata->res;
7442 	calldata->clp = clp;
7443 	task_setup_data.callback_data = calldata;
7444 
7445 	return rpc_run_task(&task_setup_data);
7446 }
7447 
7448 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7449 {
7450 	struct rpc_task *task;
7451 	int ret = 0;
7452 
7453 	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7454 		return -EAGAIN;
7455 	task = _nfs41_proc_sequence(clp, cred, false);
7456 	if (IS_ERR(task))
7457 		ret = PTR_ERR(task);
7458 	else
7459 		rpc_put_task_async(task);
7460 	dprintk("<-- %s status=%d\n", __func__, ret);
7461 	return ret;
7462 }
7463 
7464 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7465 {
7466 	struct rpc_task *task;
7467 	int ret;
7468 
7469 	task = _nfs41_proc_sequence(clp, cred, true);
7470 	if (IS_ERR(task)) {
7471 		ret = PTR_ERR(task);
7472 		goto out;
7473 	}
7474 	ret = rpc_wait_for_completion_task(task);
7475 	if (!ret) {
7476 		struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
7477 
7478 		if (task->tk_status == 0)
7479 			nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
7480 		ret = task->tk_status;
7481 	}
7482 	rpc_put_task(task);
7483 out:
7484 	dprintk("<-- %s status=%d\n", __func__, ret);
7485 	return ret;
7486 }
7487 
7488 struct nfs4_reclaim_complete_data {
7489 	struct nfs_client *clp;
7490 	struct nfs41_reclaim_complete_args arg;
7491 	struct nfs41_reclaim_complete_res res;
7492 };
7493 
7494 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7495 {
7496 	struct nfs4_reclaim_complete_data *calldata = data;
7497 
7498 	nfs41_setup_sequence(calldata->clp->cl_session,
7499 			&calldata->arg.seq_args,
7500 			&calldata->res.seq_res,
7501 			task);
7502 }
7503 
7504 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7505 {
7506 	switch(task->tk_status) {
7507 	case 0:
7508 	case -NFS4ERR_COMPLETE_ALREADY:
7509 	case -NFS4ERR_WRONG_CRED: /* What to do here? */
7510 		break;
7511 	case -NFS4ERR_DELAY:
7512 		rpc_delay(task, NFS4_POLL_RETRY_MAX);
7513 		/* fall through */
7514 	case -NFS4ERR_RETRY_UNCACHED_REP:
7515 		return -EAGAIN;
7516 	default:
7517 		nfs4_schedule_lease_recovery(clp);
7518 	}
7519 	return 0;
7520 }
7521 
7522 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7523 {
7524 	struct nfs4_reclaim_complete_data *calldata = data;
7525 	struct nfs_client *clp = calldata->clp;
7526 	struct nfs4_sequence_res *res = &calldata->res.seq_res;
7527 
7528 	dprintk("--> %s\n", __func__);
7529 	if (!nfs41_sequence_done(task, res))
7530 		return;
7531 
7532 	trace_nfs4_reclaim_complete(clp, task->tk_status);
7533 	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7534 		rpc_restart_call_prepare(task);
7535 		return;
7536 	}
7537 	dprintk("<-- %s\n", __func__);
7538 }
7539 
7540 static void nfs4_free_reclaim_complete_data(void *data)
7541 {
7542 	struct nfs4_reclaim_complete_data *calldata = data;
7543 
7544 	kfree(calldata);
7545 }
7546 
7547 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7548 	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
7549 	.rpc_call_done = nfs4_reclaim_complete_done,
7550 	.rpc_release = nfs4_free_reclaim_complete_data,
7551 };
7552 
7553 /*
7554  * Issue a global reclaim complete.
7555  */
7556 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7557 		struct rpc_cred *cred)
7558 {
7559 	struct nfs4_reclaim_complete_data *calldata;
7560 	struct rpc_task *task;
7561 	struct rpc_message msg = {
7562 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7563 		.rpc_cred = cred,
7564 	};
7565 	struct rpc_task_setup task_setup_data = {
7566 		.rpc_client = clp->cl_rpcclient,
7567 		.rpc_message = &msg,
7568 		.callback_ops = &nfs4_reclaim_complete_call_ops,
7569 		.flags = RPC_TASK_ASYNC,
7570 	};
7571 	int status = -ENOMEM;
7572 
7573 	dprintk("--> %s\n", __func__);
7574 	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7575 	if (calldata == NULL)
7576 		goto out;
7577 	calldata->clp = clp;
7578 	calldata->arg.one_fs = 0;
7579 
7580 	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7581 	nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7582 	msg.rpc_argp = &calldata->arg;
7583 	msg.rpc_resp = &calldata->res;
7584 	task_setup_data.callback_data = calldata;
7585 	task = rpc_run_task(&task_setup_data);
7586 	if (IS_ERR(task)) {
7587 		status = PTR_ERR(task);
7588 		goto out;
7589 	}
7590 	status = nfs4_wait_for_completion_rpc_task(task);
7591 	if (status == 0)
7592 		status = task->tk_status;
7593 	rpc_put_task(task);
7594 	return 0;
7595 out:
7596 	dprintk("<-- %s status=%d\n", __func__, status);
7597 	return status;
7598 }
7599 
7600 static void
7601 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7602 {
7603 	struct nfs4_layoutget *lgp = calldata;
7604 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7605 	struct nfs4_session *session = nfs4_get_session(server);
7606 
7607 	dprintk("--> %s\n", __func__);
7608 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
7609 	 * right now covering the LAYOUTGET we are about to send.
7610 	 * However, that is not so catastrophic, and there seems
7611 	 * to be no way to prevent it completely.
7612 	 */
7613 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7614 				&lgp->res.seq_res, task))
7615 		return;
7616 	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7617 					  NFS_I(lgp->args.inode)->layout,
7618 					  &lgp->args.range,
7619 					  lgp->args.ctx->state)) {
7620 		rpc_exit(task, NFS4_OK);
7621 	}
7622 }
7623 
7624 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7625 {
7626 	struct nfs4_layoutget *lgp = calldata;
7627 	struct inode *inode = lgp->args.inode;
7628 	struct nfs_server *server = NFS_SERVER(inode);
7629 	struct pnfs_layout_hdr *lo;
7630 	struct nfs4_state *state = NULL;
7631 	unsigned long timeo, now, giveup;
7632 
7633 	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7634 
7635 	if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7636 		goto out;
7637 
7638 	switch (task->tk_status) {
7639 	case 0:
7640 		goto out;
7641 	/*
7642 	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7643 	 * (or clients) writing to the same RAID stripe
7644 	 */
7645 	case -NFS4ERR_LAYOUTTRYLATER:
7646 	/*
7647 	 * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7648 	 * existing layout before getting a new one).
7649 	 */
7650 	case -NFS4ERR_RECALLCONFLICT:
7651 		timeo = rpc_get_timeout(task->tk_client);
7652 		giveup = lgp->args.timestamp + timeo;
7653 		now = jiffies;
7654 		if (time_after(giveup, now)) {
7655 			unsigned long delay;
7656 
7657 			/* Delay for:
7658 			 * - Not less then NFS4_POLL_RETRY_MIN.
7659 			 * - One last time a jiffie before we give up
7660 			 * - exponential backoff (time_now minus start_attempt)
7661 			 */
7662 			delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7663 				    min((giveup - now - 1),
7664 					now - lgp->args.timestamp));
7665 
7666 			dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7667 				__func__, delay);
7668 			rpc_delay(task, delay);
7669 			task->tk_status = 0;
7670 			rpc_restart_call_prepare(task);
7671 			goto out; /* Do not call nfs4_async_handle_error() */
7672 		}
7673 		break;
7674 	case -NFS4ERR_EXPIRED:
7675 	case -NFS4ERR_BAD_STATEID:
7676 		spin_lock(&inode->i_lock);
7677 		lo = NFS_I(inode)->layout;
7678 		if (!lo || list_empty(&lo->plh_segs)) {
7679 			spin_unlock(&inode->i_lock);
7680 			/* If the open stateid was bad, then recover it. */
7681 			state = lgp->args.ctx->state;
7682 		} else {
7683 			LIST_HEAD(head);
7684 
7685 			/*
7686 			 * Mark the bad layout state as invalid, then retry
7687 			 * with the current stateid.
7688 			 */
7689 			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7690 			spin_unlock(&inode->i_lock);
7691 			pnfs_free_lseg_list(&head);
7692 
7693 			task->tk_status = 0;
7694 			rpc_restart_call_prepare(task);
7695 		}
7696 	}
7697 	if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7698 		rpc_restart_call_prepare(task);
7699 out:
7700 	dprintk("<-- %s\n", __func__);
7701 }
7702 
7703 static size_t max_response_pages(struct nfs_server *server)
7704 {
7705 	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7706 	return nfs_page_array_len(0, max_resp_sz);
7707 }
7708 
7709 static void nfs4_free_pages(struct page **pages, size_t size)
7710 {
7711 	int i;
7712 
7713 	if (!pages)
7714 		return;
7715 
7716 	for (i = 0; i < size; i++) {
7717 		if (!pages[i])
7718 			break;
7719 		__free_page(pages[i]);
7720 	}
7721 	kfree(pages);
7722 }
7723 
7724 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7725 {
7726 	struct page **pages;
7727 	int i;
7728 
7729 	pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7730 	if (!pages) {
7731 		dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7732 		return NULL;
7733 	}
7734 
7735 	for (i = 0; i < size; i++) {
7736 		pages[i] = alloc_page(gfp_flags);
7737 		if (!pages[i]) {
7738 			dprintk("%s: failed to allocate page\n", __func__);
7739 			nfs4_free_pages(pages, size);
7740 			return NULL;
7741 		}
7742 	}
7743 
7744 	return pages;
7745 }
7746 
7747 static void nfs4_layoutget_release(void *calldata)
7748 {
7749 	struct nfs4_layoutget *lgp = calldata;
7750 	struct inode *inode = lgp->args.inode;
7751 	struct nfs_server *server = NFS_SERVER(inode);
7752 	size_t max_pages = max_response_pages(server);
7753 
7754 	dprintk("--> %s\n", __func__);
7755 	nfs4_free_pages(lgp->args.layout.pages, max_pages);
7756 	pnfs_put_layout_hdr(NFS_I(inode)->layout);
7757 	put_nfs_open_context(lgp->args.ctx);
7758 	kfree(calldata);
7759 	dprintk("<-- %s\n", __func__);
7760 }
7761 
7762 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7763 	.rpc_call_prepare = nfs4_layoutget_prepare,
7764 	.rpc_call_done = nfs4_layoutget_done,
7765 	.rpc_release = nfs4_layoutget_release,
7766 };
7767 
7768 struct pnfs_layout_segment *
7769 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7770 {
7771 	struct inode *inode = lgp->args.inode;
7772 	struct nfs_server *server = NFS_SERVER(inode);
7773 	size_t max_pages = max_response_pages(server);
7774 	struct rpc_task *task;
7775 	struct rpc_message msg = {
7776 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7777 		.rpc_argp = &lgp->args,
7778 		.rpc_resp = &lgp->res,
7779 		.rpc_cred = lgp->cred,
7780 	};
7781 	struct rpc_task_setup task_setup_data = {
7782 		.rpc_client = server->client,
7783 		.rpc_message = &msg,
7784 		.callback_ops = &nfs4_layoutget_call_ops,
7785 		.callback_data = lgp,
7786 		.flags = RPC_TASK_ASYNC,
7787 	};
7788 	struct pnfs_layout_segment *lseg = NULL;
7789 	int status = 0;
7790 
7791 	dprintk("--> %s\n", __func__);
7792 
7793 	/* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7794 	pnfs_get_layout_hdr(NFS_I(inode)->layout);
7795 
7796 	lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7797 	if (!lgp->args.layout.pages) {
7798 		nfs4_layoutget_release(lgp);
7799 		return ERR_PTR(-ENOMEM);
7800 	}
7801 	lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7802 	lgp->args.timestamp = jiffies;
7803 
7804 	lgp->res.layoutp = &lgp->args.layout;
7805 	lgp->res.seq_res.sr_slot = NULL;
7806 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7807 
7808 	task = rpc_run_task(&task_setup_data);
7809 	if (IS_ERR(task))
7810 		return ERR_CAST(task);
7811 	status = nfs4_wait_for_completion_rpc_task(task);
7812 	if (status == 0)
7813 		status = task->tk_status;
7814 	trace_nfs4_layoutget(lgp->args.ctx,
7815 			&lgp->args.range,
7816 			&lgp->res.range,
7817 			status);
7818 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7819 	if (status == 0 && lgp->res.layoutp->len)
7820 		lseg = pnfs_layout_process(lgp);
7821 	rpc_put_task(task);
7822 	dprintk("<-- %s status=%d\n", __func__, status);
7823 	if (status)
7824 		return ERR_PTR(status);
7825 	return lseg;
7826 }
7827 
7828 static void
7829 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7830 {
7831 	struct nfs4_layoutreturn *lrp = calldata;
7832 
7833 	dprintk("--> %s\n", __func__);
7834 	nfs41_setup_sequence(lrp->clp->cl_session,
7835 			&lrp->args.seq_args,
7836 			&lrp->res.seq_res,
7837 			task);
7838 }
7839 
7840 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7841 {
7842 	struct nfs4_layoutreturn *lrp = calldata;
7843 	struct nfs_server *server;
7844 
7845 	dprintk("--> %s\n", __func__);
7846 
7847 	if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7848 		return;
7849 
7850 	server = NFS_SERVER(lrp->args.inode);
7851 	switch (task->tk_status) {
7852 	default:
7853 		task->tk_status = 0;
7854 	case 0:
7855 		break;
7856 	case -NFS4ERR_DELAY:
7857 		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7858 			break;
7859 		rpc_restart_call_prepare(task);
7860 		return;
7861 	}
7862 	dprintk("<-- %s\n", __func__);
7863 }
7864 
7865 static void nfs4_layoutreturn_release(void *calldata)
7866 {
7867 	struct nfs4_layoutreturn *lrp = calldata;
7868 	struct pnfs_layout_hdr *lo = lrp->args.layout;
7869 
7870 	dprintk("--> %s\n", __func__);
7871 	spin_lock(&lo->plh_inode->i_lock);
7872 	if (lrp->res.lrs_present)
7873 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7874 	pnfs_clear_layoutreturn_waitbit(lo);
7875 	clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags);
7876 	rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
7877 	lo->plh_block_lgets--;
7878 	spin_unlock(&lo->plh_inode->i_lock);
7879 	pnfs_put_layout_hdr(lrp->args.layout);
7880 	nfs_iput_and_deactive(lrp->inode);
7881 	kfree(calldata);
7882 	dprintk("<-- %s\n", __func__);
7883 }
7884 
7885 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7886 	.rpc_call_prepare = nfs4_layoutreturn_prepare,
7887 	.rpc_call_done = nfs4_layoutreturn_done,
7888 	.rpc_release = nfs4_layoutreturn_release,
7889 };
7890 
7891 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7892 {
7893 	struct rpc_task *task;
7894 	struct rpc_message msg = {
7895 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7896 		.rpc_argp = &lrp->args,
7897 		.rpc_resp = &lrp->res,
7898 		.rpc_cred = lrp->cred,
7899 	};
7900 	struct rpc_task_setup task_setup_data = {
7901 		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
7902 		.rpc_message = &msg,
7903 		.callback_ops = &nfs4_layoutreturn_call_ops,
7904 		.callback_data = lrp,
7905 	};
7906 	int status = 0;
7907 
7908 	dprintk("--> %s\n", __func__);
7909 	if (!sync) {
7910 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
7911 		if (!lrp->inode) {
7912 			nfs4_layoutreturn_release(lrp);
7913 			return -EAGAIN;
7914 		}
7915 		task_setup_data.flags |= RPC_TASK_ASYNC;
7916 	}
7917 	nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
7918 	task = rpc_run_task(&task_setup_data);
7919 	if (IS_ERR(task))
7920 		return PTR_ERR(task);
7921 	if (sync)
7922 		status = task->tk_status;
7923 	trace_nfs4_layoutreturn(lrp->args.inode, status);
7924 	dprintk("<-- %s status=%d\n", __func__, status);
7925 	rpc_put_task(task);
7926 	return status;
7927 }
7928 
7929 static int
7930 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
7931 		struct pnfs_device *pdev,
7932 		struct rpc_cred *cred)
7933 {
7934 	struct nfs4_getdeviceinfo_args args = {
7935 		.pdev = pdev,
7936 	};
7937 	struct nfs4_getdeviceinfo_res res = {
7938 		.pdev = pdev,
7939 	};
7940 	struct rpc_message msg = {
7941 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
7942 		.rpc_argp = &args,
7943 		.rpc_resp = &res,
7944 		.rpc_cred = cred,
7945 	};
7946 	int status;
7947 
7948 	dprintk("--> %s\n", __func__);
7949 	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
7950 	dprintk("<-- %s status=%d\n", __func__, status);
7951 
7952 	return status;
7953 }
7954 
7955 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
7956 		struct pnfs_device *pdev,
7957 		struct rpc_cred *cred)
7958 {
7959 	struct nfs4_exception exception = { };
7960 	int err;
7961 
7962 	do {
7963 		err = nfs4_handle_exception(server,
7964 					_nfs4_proc_getdeviceinfo(server, pdev, cred),
7965 					&exception);
7966 	} while (exception.retry);
7967 	return err;
7968 }
7969 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
7970 
7971 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
7972 {
7973 	struct nfs4_layoutcommit_data *data = calldata;
7974 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7975 	struct nfs4_session *session = nfs4_get_session(server);
7976 
7977 	nfs41_setup_sequence(session,
7978 			&data->args.seq_args,
7979 			&data->res.seq_res,
7980 			task);
7981 }
7982 
7983 static void
7984 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
7985 {
7986 	struct nfs4_layoutcommit_data *data = calldata;
7987 	struct nfs_server *server = NFS_SERVER(data->args.inode);
7988 
7989 	if (!nfs41_sequence_done(task, &data->res.seq_res))
7990 		return;
7991 
7992 	switch (task->tk_status) { /* Just ignore these failures */
7993 	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
7994 	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
7995 	case -NFS4ERR_BADLAYOUT:     /* no layout */
7996 	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
7997 		task->tk_status = 0;
7998 	case 0:
7999 		break;
8000 	default:
8001 		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8002 			rpc_restart_call_prepare(task);
8003 			return;
8004 		}
8005 	}
8006 }
8007 
8008 static void nfs4_layoutcommit_release(void *calldata)
8009 {
8010 	struct nfs4_layoutcommit_data *data = calldata;
8011 
8012 	pnfs_cleanup_layoutcommit(data);
8013 	nfs_post_op_update_inode_force_wcc(data->args.inode,
8014 					   data->res.fattr);
8015 	put_rpccred(data->cred);
8016 	nfs_iput_and_deactive(data->inode);
8017 	kfree(data);
8018 }
8019 
8020 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8021 	.rpc_call_prepare = nfs4_layoutcommit_prepare,
8022 	.rpc_call_done = nfs4_layoutcommit_done,
8023 	.rpc_release = nfs4_layoutcommit_release,
8024 };
8025 
8026 int
8027 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8028 {
8029 	struct rpc_message msg = {
8030 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8031 		.rpc_argp = &data->args,
8032 		.rpc_resp = &data->res,
8033 		.rpc_cred = data->cred,
8034 	};
8035 	struct rpc_task_setup task_setup_data = {
8036 		.task = &data->task,
8037 		.rpc_client = NFS_CLIENT(data->args.inode),
8038 		.rpc_message = &msg,
8039 		.callback_ops = &nfs4_layoutcommit_ops,
8040 		.callback_data = data,
8041 	};
8042 	struct rpc_task *task;
8043 	int status = 0;
8044 
8045 	dprintk("NFS: %4d initiating layoutcommit call. sync %d "
8046 		"lbw: %llu inode %lu\n",
8047 		data->task.tk_pid, sync,
8048 		data->args.lastbytewritten,
8049 		data->args.inode->i_ino);
8050 
8051 	if (!sync) {
8052 		data->inode = nfs_igrab_and_active(data->args.inode);
8053 		if (data->inode == NULL) {
8054 			nfs4_layoutcommit_release(data);
8055 			return -EAGAIN;
8056 		}
8057 		task_setup_data.flags = RPC_TASK_ASYNC;
8058 	}
8059 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8060 	task = rpc_run_task(&task_setup_data);
8061 	if (IS_ERR(task))
8062 		return PTR_ERR(task);
8063 	if (sync)
8064 		status = task->tk_status;
8065 	trace_nfs4_layoutcommit(data->args.inode, status);
8066 	dprintk("%s: status %d\n", __func__, status);
8067 	rpc_put_task(task);
8068 	return status;
8069 }
8070 
8071 /**
8072  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8073  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8074  */
8075 static int
8076 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8077 		    struct nfs_fsinfo *info,
8078 		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8079 {
8080 	struct nfs41_secinfo_no_name_args args = {
8081 		.style = SECINFO_STYLE_CURRENT_FH,
8082 	};
8083 	struct nfs4_secinfo_res res = {
8084 		.flavors = flavors,
8085 	};
8086 	struct rpc_message msg = {
8087 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8088 		.rpc_argp = &args,
8089 		.rpc_resp = &res,
8090 	};
8091 	struct rpc_clnt *clnt = server->client;
8092 	struct rpc_cred *cred = NULL;
8093 	int status;
8094 
8095 	if (use_integrity) {
8096 		clnt = server->nfs_client->cl_rpcclient;
8097 		cred = nfs4_get_clid_cred(server->nfs_client);
8098 		msg.rpc_cred = cred;
8099 	}
8100 
8101 	dprintk("--> %s\n", __func__);
8102 	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8103 				&res.seq_res, 0);
8104 	dprintk("<-- %s status=%d\n", __func__, status);
8105 
8106 	if (cred)
8107 		put_rpccred(cred);
8108 
8109 	return status;
8110 }
8111 
8112 static int
8113 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8114 			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8115 {
8116 	struct nfs4_exception exception = { };
8117 	int err;
8118 	do {
8119 		/* first try using integrity protection */
8120 		err = -NFS4ERR_WRONGSEC;
8121 
8122 		/* try to use integrity protection with machine cred */
8123 		if (_nfs4_is_integrity_protected(server->nfs_client))
8124 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8125 							  flavors, true);
8126 
8127 		/*
8128 		 * if unable to use integrity protection, or SECINFO with
8129 		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8130 		 * disallowed by spec, but exists in deployed servers) use
8131 		 * the current filesystem's rpc_client and the user cred.
8132 		 */
8133 		if (err == -NFS4ERR_WRONGSEC)
8134 			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8135 							  flavors, false);
8136 
8137 		switch (err) {
8138 		case 0:
8139 		case -NFS4ERR_WRONGSEC:
8140 		case -ENOTSUPP:
8141 			goto out;
8142 		default:
8143 			err = nfs4_handle_exception(server, err, &exception);
8144 		}
8145 	} while (exception.retry);
8146 out:
8147 	return err;
8148 }
8149 
8150 static int
8151 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8152 		    struct nfs_fsinfo *info)
8153 {
8154 	int err;
8155 	struct page *page;
8156 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8157 	struct nfs4_secinfo_flavors *flavors;
8158 	struct nfs4_secinfo4 *secinfo;
8159 	int i;
8160 
8161 	page = alloc_page(GFP_KERNEL);
8162 	if (!page) {
8163 		err = -ENOMEM;
8164 		goto out;
8165 	}
8166 
8167 	flavors = page_address(page);
8168 	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8169 
8170 	/*
8171 	 * Fall back on "guess and check" method if
8172 	 * the server doesn't support SECINFO_NO_NAME
8173 	 */
8174 	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8175 		err = nfs4_find_root_sec(server, fhandle, info);
8176 		goto out_freepage;
8177 	}
8178 	if (err)
8179 		goto out_freepage;
8180 
8181 	for (i = 0; i < flavors->num_flavors; i++) {
8182 		secinfo = &flavors->flavors[i];
8183 
8184 		switch (secinfo->flavor) {
8185 		case RPC_AUTH_NULL:
8186 		case RPC_AUTH_UNIX:
8187 		case RPC_AUTH_GSS:
8188 			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8189 					&secinfo->flavor_info);
8190 			break;
8191 		default:
8192 			flavor = RPC_AUTH_MAXFLAVOR;
8193 			break;
8194 		}
8195 
8196 		if (!nfs_auth_info_match(&server->auth_info, flavor))
8197 			flavor = RPC_AUTH_MAXFLAVOR;
8198 
8199 		if (flavor != RPC_AUTH_MAXFLAVOR) {
8200 			err = nfs4_lookup_root_sec(server, fhandle,
8201 						   info, flavor);
8202 			if (!err)
8203 				break;
8204 		}
8205 	}
8206 
8207 	if (flavor == RPC_AUTH_MAXFLAVOR)
8208 		err = -EPERM;
8209 
8210 out_freepage:
8211 	put_page(page);
8212 	if (err == -EACCES)
8213 		return -EPERM;
8214 out:
8215 	return err;
8216 }
8217 
8218 static int _nfs41_test_stateid(struct nfs_server *server,
8219 		nfs4_stateid *stateid,
8220 		struct rpc_cred *cred)
8221 {
8222 	int status;
8223 	struct nfs41_test_stateid_args args = {
8224 		.stateid = stateid,
8225 	};
8226 	struct nfs41_test_stateid_res res;
8227 	struct rpc_message msg = {
8228 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8229 		.rpc_argp = &args,
8230 		.rpc_resp = &res,
8231 		.rpc_cred = cred,
8232 	};
8233 	struct rpc_clnt *rpc_client = server->client;
8234 
8235 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8236 		&rpc_client, &msg);
8237 
8238 	dprintk("NFS call  test_stateid %p\n", stateid);
8239 	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8240 	nfs4_set_sequence_privileged(&args.seq_args);
8241 	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8242 			&args.seq_args, &res.seq_res);
8243 	if (status != NFS_OK) {
8244 		dprintk("NFS reply test_stateid: failed, %d\n", status);
8245 		return status;
8246 	}
8247 	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8248 	return -res.status;
8249 }
8250 
8251 /**
8252  * nfs41_test_stateid - perform a TEST_STATEID operation
8253  *
8254  * @server: server / transport on which to perform the operation
8255  * @stateid: state ID to test
8256  * @cred: credential
8257  *
8258  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8259  * Otherwise a negative NFS4ERR value is returned if the operation
8260  * failed or the state ID is not currently valid.
8261  */
8262 static int nfs41_test_stateid(struct nfs_server *server,
8263 		nfs4_stateid *stateid,
8264 		struct rpc_cred *cred)
8265 {
8266 	struct nfs4_exception exception = { };
8267 	int err;
8268 	do {
8269 		err = _nfs41_test_stateid(server, stateid, cred);
8270 		if (err != -NFS4ERR_DELAY)
8271 			break;
8272 		nfs4_handle_exception(server, err, &exception);
8273 	} while (exception.retry);
8274 	return err;
8275 }
8276 
8277 struct nfs_free_stateid_data {
8278 	struct nfs_server *server;
8279 	struct nfs41_free_stateid_args args;
8280 	struct nfs41_free_stateid_res res;
8281 };
8282 
8283 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8284 {
8285 	struct nfs_free_stateid_data *data = calldata;
8286 	nfs41_setup_sequence(nfs4_get_session(data->server),
8287 			&data->args.seq_args,
8288 			&data->res.seq_res,
8289 			task);
8290 }
8291 
8292 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8293 {
8294 	struct nfs_free_stateid_data *data = calldata;
8295 
8296 	nfs41_sequence_done(task, &data->res.seq_res);
8297 
8298 	switch (task->tk_status) {
8299 	case -NFS4ERR_DELAY:
8300 		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8301 			rpc_restart_call_prepare(task);
8302 	}
8303 }
8304 
8305 static void nfs41_free_stateid_release(void *calldata)
8306 {
8307 	kfree(calldata);
8308 }
8309 
8310 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8311 	.rpc_call_prepare = nfs41_free_stateid_prepare,
8312 	.rpc_call_done = nfs41_free_stateid_done,
8313 	.rpc_release = nfs41_free_stateid_release,
8314 };
8315 
8316 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8317 		nfs4_stateid *stateid,
8318 		struct rpc_cred *cred,
8319 		bool privileged)
8320 {
8321 	struct rpc_message msg = {
8322 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8323 		.rpc_cred = cred,
8324 	};
8325 	struct rpc_task_setup task_setup = {
8326 		.rpc_client = server->client,
8327 		.rpc_message = &msg,
8328 		.callback_ops = &nfs41_free_stateid_ops,
8329 		.flags = RPC_TASK_ASYNC,
8330 	};
8331 	struct nfs_free_stateid_data *data;
8332 
8333 	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8334 		&task_setup.rpc_client, &msg);
8335 
8336 	dprintk("NFS call  free_stateid %p\n", stateid);
8337 	data = kmalloc(sizeof(*data), GFP_NOFS);
8338 	if (!data)
8339 		return ERR_PTR(-ENOMEM);
8340 	data->server = server;
8341 	nfs4_stateid_copy(&data->args.stateid, stateid);
8342 
8343 	task_setup.callback_data = data;
8344 
8345 	msg.rpc_argp = &data->args;
8346 	msg.rpc_resp = &data->res;
8347 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8348 	if (privileged)
8349 		nfs4_set_sequence_privileged(&data->args.seq_args);
8350 
8351 	return rpc_run_task(&task_setup);
8352 }
8353 
8354 /**
8355  * nfs41_free_stateid - perform a FREE_STATEID operation
8356  *
8357  * @server: server / transport on which to perform the operation
8358  * @stateid: state ID to release
8359  * @cred: credential
8360  *
8361  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8362  * negative NFS4ERR value is returned.
8363  */
8364 static int nfs41_free_stateid(struct nfs_server *server,
8365 		nfs4_stateid *stateid,
8366 		struct rpc_cred *cred)
8367 {
8368 	struct rpc_task *task;
8369 	int ret;
8370 
8371 	task = _nfs41_free_stateid(server, stateid, cred, true);
8372 	if (IS_ERR(task))
8373 		return PTR_ERR(task);
8374 	ret = rpc_wait_for_completion_task(task);
8375 	if (!ret)
8376 		ret = task->tk_status;
8377 	rpc_put_task(task);
8378 	return ret;
8379 }
8380 
8381 static void
8382 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8383 {
8384 	struct rpc_task *task;
8385 	struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8386 
8387 	task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8388 	nfs4_free_lock_state(server, lsp);
8389 	if (IS_ERR(task))
8390 		return;
8391 	rpc_put_task(task);
8392 }
8393 
8394 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8395 		const nfs4_stateid *s2)
8396 {
8397 	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8398 		return false;
8399 
8400 	if (s1->seqid == s2->seqid)
8401 		return true;
8402 	if (s1->seqid == 0 || s2->seqid == 0)
8403 		return true;
8404 
8405 	return false;
8406 }
8407 
8408 #endif /* CONFIG_NFS_V4_1 */
8409 
8410 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8411 		const nfs4_stateid *s2)
8412 {
8413 	return nfs4_stateid_match(s1, s2);
8414 }
8415 
8416 
8417 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8418 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8419 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8420 	.recover_open	= nfs4_open_reclaim,
8421 	.recover_lock	= nfs4_lock_reclaim,
8422 	.establish_clid = nfs4_init_clientid,
8423 	.detect_trunking = nfs40_discover_server_trunking,
8424 };
8425 
8426 #if defined(CONFIG_NFS_V4_1)
8427 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8428 	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8429 	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
8430 	.recover_open	= nfs4_open_reclaim,
8431 	.recover_lock	= nfs4_lock_reclaim,
8432 	.establish_clid = nfs41_init_clientid,
8433 	.reclaim_complete = nfs41_proc_reclaim_complete,
8434 	.detect_trunking = nfs41_discover_server_trunking,
8435 };
8436 #endif /* CONFIG_NFS_V4_1 */
8437 
8438 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8439 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8440 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8441 	.recover_open	= nfs40_open_expired,
8442 	.recover_lock	= nfs4_lock_expired,
8443 	.establish_clid = nfs4_init_clientid,
8444 };
8445 
8446 #if defined(CONFIG_NFS_V4_1)
8447 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8448 	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8449 	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
8450 	.recover_open	= nfs41_open_expired,
8451 	.recover_lock	= nfs41_lock_expired,
8452 	.establish_clid = nfs41_init_clientid,
8453 };
8454 #endif /* CONFIG_NFS_V4_1 */
8455 
8456 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8457 	.sched_state_renewal = nfs4_proc_async_renew,
8458 	.get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8459 	.renew_lease = nfs4_proc_renew,
8460 };
8461 
8462 #if defined(CONFIG_NFS_V4_1)
8463 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8464 	.sched_state_renewal = nfs41_proc_async_sequence,
8465 	.get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8466 	.renew_lease = nfs4_proc_sequence,
8467 };
8468 #endif
8469 
8470 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8471 	.get_locations = _nfs40_proc_get_locations,
8472 	.fsid_present = _nfs40_proc_fsid_present,
8473 };
8474 
8475 #if defined(CONFIG_NFS_V4_1)
8476 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8477 	.get_locations = _nfs41_proc_get_locations,
8478 	.fsid_present = _nfs41_proc_fsid_present,
8479 };
8480 #endif	/* CONFIG_NFS_V4_1 */
8481 
8482 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8483 	.minor_version = 0,
8484 	.init_caps = NFS_CAP_READDIRPLUS
8485 		| NFS_CAP_ATOMIC_OPEN
8486 		| NFS_CAP_CHANGE_ATTR
8487 		| NFS_CAP_POSIX_LOCK,
8488 	.init_client = nfs40_init_client,
8489 	.shutdown_client = nfs40_shutdown_client,
8490 	.match_stateid = nfs4_match_stateid,
8491 	.find_root_sec = nfs4_find_root_sec,
8492 	.free_lock_state = nfs4_release_lockowner,
8493 	.alloc_seqid = nfs_alloc_seqid,
8494 	.call_sync_ops = &nfs40_call_sync_ops,
8495 	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8496 	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8497 	.state_renewal_ops = &nfs40_state_renewal_ops,
8498 	.mig_recovery_ops = &nfs40_mig_recovery_ops,
8499 };
8500 
8501 #if defined(CONFIG_NFS_V4_1)
8502 static struct nfs_seqid *
8503 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8504 {
8505 	return NULL;
8506 }
8507 
8508 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8509 	.minor_version = 1,
8510 	.init_caps = NFS_CAP_READDIRPLUS
8511 		| NFS_CAP_ATOMIC_OPEN
8512 		| NFS_CAP_CHANGE_ATTR
8513 		| NFS_CAP_POSIX_LOCK
8514 		| NFS_CAP_STATEID_NFSV41
8515 		| NFS_CAP_ATOMIC_OPEN_V1,
8516 	.init_client = nfs41_init_client,
8517 	.shutdown_client = nfs41_shutdown_client,
8518 	.match_stateid = nfs41_match_stateid,
8519 	.find_root_sec = nfs41_find_root_sec,
8520 	.free_lock_state = nfs41_free_lock_state,
8521 	.alloc_seqid = nfs_alloc_no_seqid,
8522 	.call_sync_ops = &nfs41_call_sync_ops,
8523 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8524 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8525 	.state_renewal_ops = &nfs41_state_renewal_ops,
8526 	.mig_recovery_ops = &nfs41_mig_recovery_ops,
8527 };
8528 #endif
8529 
8530 #if defined(CONFIG_NFS_V4_2)
8531 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8532 	.minor_version = 2,
8533 	.init_caps = NFS_CAP_READDIRPLUS
8534 		| NFS_CAP_ATOMIC_OPEN
8535 		| NFS_CAP_CHANGE_ATTR
8536 		| NFS_CAP_POSIX_LOCK
8537 		| NFS_CAP_STATEID_NFSV41
8538 		| NFS_CAP_ATOMIC_OPEN_V1
8539 		| NFS_CAP_ALLOCATE
8540 		| NFS_CAP_DEALLOCATE
8541 		| NFS_CAP_SEEK,
8542 	.init_client = nfs41_init_client,
8543 	.shutdown_client = nfs41_shutdown_client,
8544 	.match_stateid = nfs41_match_stateid,
8545 	.find_root_sec = nfs41_find_root_sec,
8546 	.free_lock_state = nfs41_free_lock_state,
8547 	.call_sync_ops = &nfs41_call_sync_ops,
8548 	.alloc_seqid = nfs_alloc_no_seqid,
8549 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8550 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8551 	.state_renewal_ops = &nfs41_state_renewal_ops,
8552 };
8553 #endif
8554 
8555 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8556 	[0] = &nfs_v4_0_minor_ops,
8557 #if defined(CONFIG_NFS_V4_1)
8558 	[1] = &nfs_v4_1_minor_ops,
8559 #endif
8560 #if defined(CONFIG_NFS_V4_2)
8561 	[2] = &nfs_v4_2_minor_ops,
8562 #endif
8563 };
8564 
8565 static const struct inode_operations nfs4_dir_inode_operations = {
8566 	.create		= nfs_create,
8567 	.lookup		= nfs_lookup,
8568 	.atomic_open	= nfs_atomic_open,
8569 	.link		= nfs_link,
8570 	.unlink		= nfs_unlink,
8571 	.symlink	= nfs_symlink,
8572 	.mkdir		= nfs_mkdir,
8573 	.rmdir		= nfs_rmdir,
8574 	.mknod		= nfs_mknod,
8575 	.rename		= nfs_rename,
8576 	.permission	= nfs_permission,
8577 	.getattr	= nfs_getattr,
8578 	.setattr	= nfs_setattr,
8579 	.getxattr	= generic_getxattr,
8580 	.setxattr	= generic_setxattr,
8581 	.listxattr	= generic_listxattr,
8582 	.removexattr	= generic_removexattr,
8583 };
8584 
8585 static const struct inode_operations nfs4_file_inode_operations = {
8586 	.permission	= nfs_permission,
8587 	.getattr	= nfs_getattr,
8588 	.setattr	= nfs_setattr,
8589 	.getxattr	= generic_getxattr,
8590 	.setxattr	= generic_setxattr,
8591 	.listxattr	= generic_listxattr,
8592 	.removexattr	= generic_removexattr,
8593 };
8594 
8595 const struct nfs_rpc_ops nfs_v4_clientops = {
8596 	.version	= 4,			/* protocol version */
8597 	.dentry_ops	= &nfs4_dentry_operations,
8598 	.dir_inode_ops	= &nfs4_dir_inode_operations,
8599 	.file_inode_ops	= &nfs4_file_inode_operations,
8600 	.file_ops	= &nfs4_file_operations,
8601 	.getroot	= nfs4_proc_get_root,
8602 	.submount	= nfs4_submount,
8603 	.try_mount	= nfs4_try_mount,
8604 	.getattr	= nfs4_proc_getattr,
8605 	.setattr	= nfs4_proc_setattr,
8606 	.lookup		= nfs4_proc_lookup,
8607 	.access		= nfs4_proc_access,
8608 	.readlink	= nfs4_proc_readlink,
8609 	.create		= nfs4_proc_create,
8610 	.remove		= nfs4_proc_remove,
8611 	.unlink_setup	= nfs4_proc_unlink_setup,
8612 	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8613 	.unlink_done	= nfs4_proc_unlink_done,
8614 	.rename_setup	= nfs4_proc_rename_setup,
8615 	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8616 	.rename_done	= nfs4_proc_rename_done,
8617 	.link		= nfs4_proc_link,
8618 	.symlink	= nfs4_proc_symlink,
8619 	.mkdir		= nfs4_proc_mkdir,
8620 	.rmdir		= nfs4_proc_remove,
8621 	.readdir	= nfs4_proc_readdir,
8622 	.mknod		= nfs4_proc_mknod,
8623 	.statfs		= nfs4_proc_statfs,
8624 	.fsinfo		= nfs4_proc_fsinfo,
8625 	.pathconf	= nfs4_proc_pathconf,
8626 	.set_capabilities = nfs4_server_capabilities,
8627 	.decode_dirent	= nfs4_decode_dirent,
8628 	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8629 	.read_setup	= nfs4_proc_read_setup,
8630 	.read_done	= nfs4_read_done,
8631 	.write_setup	= nfs4_proc_write_setup,
8632 	.write_done	= nfs4_write_done,
8633 	.commit_setup	= nfs4_proc_commit_setup,
8634 	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8635 	.commit_done	= nfs4_commit_done,
8636 	.lock		= nfs4_proc_lock,
8637 	.clear_acl_cache = nfs4_zap_acl_attr,
8638 	.close_context  = nfs4_close_context,
8639 	.open_context	= nfs4_atomic_open,
8640 	.have_delegation = nfs4_have_delegation,
8641 	.return_delegation = nfs4_inode_return_delegation,
8642 	.alloc_client	= nfs4_alloc_client,
8643 	.init_client	= nfs4_init_client,
8644 	.free_client	= nfs4_free_client,
8645 	.create_server	= nfs4_create_server,
8646 	.clone_server	= nfs_clone_server,
8647 };
8648 
8649 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8650 	.prefix	= XATTR_NAME_NFSV4_ACL,
8651 	.list	= nfs4_xattr_list_nfs4_acl,
8652 	.get	= nfs4_xattr_get_nfs4_acl,
8653 	.set	= nfs4_xattr_set_nfs4_acl,
8654 };
8655 
8656 const struct xattr_handler *nfs4_xattr_handlers[] = {
8657 	&nfs4_xattr_nfs4_acl_handler,
8658 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8659 	&nfs4_xattr_nfs4_label_handler,
8660 #endif
8661 	NULL
8662 };
8663 
8664 /*
8665  * Local variables:
8666  *  c-basic-offset: 8
8667  * End:
8668  */
8669